QuarkGluon Plasma 4
This page intentionally left blank
QuarkGluon Plasma 4 edited by
Rudolph C Hwa University of Oregon, USA
XinNian Wang Lawrence Berkeley National Laboratory, USA
World Scientific NEW JERSEY
•
LONDON
•
SINGAPORE
•
BEIJING
•
SHANGHAI
•
HONG KONG
•
TA I P E I
•
CHENNAI
Published by World Scientific Publishing Co. Pte. Ltd. 5 Toh Tuck Link, Singapore 596224 USA office 27 Warren Street, Suite 401402, Hackensack, NJ 07601 UK office 57 Shelton Street, Covent Garden, London WC2H 9HE
British Library CataloguinginPublication Data A catalogue record for this book is available from the British Library.
QUARKGLUON PLASMA 4 Copyright © 2010 by World Scientific Publishing Co. Pte. Ltd. All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means, electronic or mechanical, including photocopying, recording or any information storage and retrieval system now known or to be invented, without written permission from the Publisher.
For photocopying of material in this volume, please pay a copying fee through the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to photocopy is not required from the publisher.
ISBN13 9789814293280 ISBN10 9814293288
Printed in Singapore.
Preface
The beginning of this decade was also the beginning of a new era when RHIC opened up a new vista in the physics of dense matter. The major theoretical insights gained soon after were reviewed in the previous volume of this series. Now as the decade draws to its end and as RHIC moves on to its mature phase, it is remarkable that new discoveries are still being made and the theoretical understanding shows no sign of reaching a saturation point. The new ideas that continue to spring up are to be tested by future experiments in the next energy domain offered at LHC. A notable claim that has been made is that RHIC has produced the most perfect fluid achievable, which the cover picture of this volume is intended to symbolize. As a springboard to LHC, it is an opportune moment to make an objective assessment of what has been learnt and what to expect next. We are happy to have assembled a group of experts who have all made significant contributions to the various aspects of the subject on many fronts, although regretfully we were not able to cover some fields that are also important. To all the contributors we are very grateful for their time and energy spent in giving the community comprehensive descriptions of the present status of QGP at what will be an auspicious beginning of the next era.
Rudolph C. Hwa Eugene, Oregon XinNian Wang Berkeley, California November 2009
v
This page intentionally left blank
Contents Preface
v
Energy Loss in a Strongly Coupled Thermal Medium and the GaugeString Duality
1
S. S. Gubser, S. S. Pufu, F. D. Rocha and A. Yarom 1. 2. 3. 4. 5.
6.
7. 8. 9.
Introduction . . . . . . . . . . . . . . . . . . . . . . The Thermal Medium as a Black Hole . . . . . . . . The Trailing String . . . . . . . . . . . . . . . . . . The Magnitude of the Drag Force . . . . . . . . . . The Perturbed Einstein Equations . . . . . . . . . . 5.1. Metric perturbations in axial gauge . . . . . . 5.2. Boundary conditions . . . . . . . . . . . . . . . 5.3. The boundary stressenergy tensor . . . . . . . Asymptotics . . . . . . . . . . . . . . . . . . . . . . 6.1. Long distance asymptotics . . . . . . . . . . . 6.2. Short distance asymptotics . . . . . . . . . . . Numerical Results for the Holographic Stress Tensor Hadronization, JetBroadening, and JetSplitting . . Conclusions . . . . . . . . . . . . . . . . . . . . . . . Appendix A. Notation . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
Quarkonium at Finite Temperature
1 2 4 8 13 16 22 24 28 29 33 38 42 51 52 61
A. Bazavov, P. Petreczky and A. Velytsky 1. 2. 3. 4.
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . pNRQCD at Finite Temperature . . . . . . . . . . . . . . . . . . . . . . Basics of Lattice Gauge Theory . . . . . . . . . . . . . . . . . . . . . . . Correlation Functions of Static Quarks in Lattice Gauge Theory . . . . 4.1. Static meson correlators . . . . . . . . . . . . . . . . . . . . . . . . 4.2. Lattice results on static meson correlators . . . . . . . . . . . . . . 4.3. Color singlet correlator in SU(2) gauge theory at low temperatures 4.4. Color singlet free energy in the deconfined phase . . . . . . . . . . 4.5. Color adjoint free energy . . . . . . . . . . . . . . . . . . . . . . . 5. Quarkonium Spectral Functions . . . . . . . . . . . . . . . . . . . . . . 5.1. Meson correlators and spectral functions . . . . . . . . . . . . . . vii
. . . . . . . . . . .
61 62 64 66 66 69 71 74 76 78 78
QuarkGluon Plasma 4
viii
5.2. Lattice formulations for charmonium physics . . . . . . . . . . . . 5.3. Bayesian analysis of meson correlators . . . . . . . . . . . . . . . . 5.4. Charmonium spectral functions at zero temperature . . . . . . . . 5.5. Charmonium correlators at finite temperature . . . . . . . . . . . 5.6. Charmonium spectral functions at finite temperature . . . . . . . 5.7. Charmonium correlators and spectral functions at finite momenta 5.8. Bottomonium spectral functions at zero temperature . . . . . . . . 5.9. Bottomonium at finite temperature . . . . . . . . . . . . . . . . . 5.10. Zero modes contribution . . . . . . . . . . . . . . . . . . . . . . . 6. Potential Models at Finite Temperature . . . . . . . . . . . . . . . . . . 7. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. 80 . 82 . 83 . 86 . 91 . 94 . 95 . 97 . 99 . 103 . 105
Heavy Quarks in the QuarkGluon Plasma
111
R. Rapp and H. van Hees 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. HeavyQuark Interactions in QCD Matter . . . . . . . . . . . . 2.1. Heavyquark diffusion in the quarkgluon plasma . . . . . 2.2. Perturbative QCD approaches . . . . . . . . . . . . . . . 2.3. Nonperturbative interactions . . . . . . . . . . . . . . . . 2.4. String theoretical evaluations of heavyquark diffusion . . 2.5. Comparison of elastic diffusion approaches . . . . . . . . 2.6. Collisional versus radiative energy loss . . . . . . . . . . . 2.7. D mesons in the hadronic phase . . . . . . . . . . . . . . 3. HeavyQuark Observables in Relativistic HeavyIon Collisions 3.1. Relativistic Langevin simulations . . . . . . . . . . . . . . 3.2. Background medium in heavyion collisions . . . . . . . . 3.3. Initial conditions and hadronization . . . . . . . . . . . . 3.4 Model comparisons of heavyquark spectra at RHIC . . . 3.5. Heavymeson and electron observables . . . . . . . . . . . 3.6. Viscosity? . . . . . . . . . . . . . . . . . . . . . . . . . . . 4. Heavy Quarkonia in Medium . . . . . . . . . . . . . . . . . . . 4.1. Spectral properties of quarkonia in the QGP . . . . . . . 4.2. Quarkonium production in heavyion collisions . . . . . . 5. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Viscous Hydrodynamics and the Quark Gluon Plasma
111 117 119 122 128 139 141 144 147 149 150 153 156 159 166 172 174 176 185 197 207
D. A. Teaney 1. Introduction . . . . . . . . . . . . . . . . . . . 1.1. Experimental overview . . . . . . . . . . . 1.2. An interpretation of elliptic flow . . . . . 2. Elliptic Flow — Measurements and Definitions
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
207 207 209 210
Contents
3. 4.
5. 6.
7.
ix
2.1. Measurements and definitions . . . . . . . . . . . . . . . . . 2.2. Interpretation . . . . . . . . . . . . . . . . . . . . . . . . . 2.3. The eccentricity and fluctuations . . . . . . . . . . . . . . . 2.4. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Shear Viscosity in QCD . . . . . . . . . . . . . . . . . . . . Hydrodynamic Description of Heavy Ion Collisions . . . . . . . . 4.1. Ideal hydrodynamics . . . . . . . . . . . . . . . . . . . . . . 4.2. Ideal Bjorken evolutions and three dimensional estimates . 4.3. Viscous Bjorken evolution and three dimensional estimates 4.4. The applicability of hydrodynamics and η/s . . . . . . . . . 4.5. Time evolution . . . . . . . . . . . . . . . . . . . . . . . . . 4.6. Second order hydrodynamics . . . . . . . . . . . . . . . . . 4.7. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . Kinetic Theory Description . . . . . . . . . . . . . . . . . . . . . Viscous Hydrodynamic Models of Heavy Ion Collisions . . . . . 6.1. Initial conditions . . . . . . . . . . . . . . . . . . . . . . . . 6.2. Corrections to the hydrodynamic flow . . . . . . . . . . . . 6.3. Convergence of the gradient expansion . . . . . . . . . . . . 6.4. Kinetic theory and hydrodynamic simulations . . . . . . . . 6.5. Particle spectra . . . . . . . . . . . . . . . . . . . . . . . . . Summary and Outlook . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . .
Hadron Correlations in Jets and Ridges Through Parton Recombination R. C. Hwa 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2. Hadronization by Recombination . . . . . . . . . . . . . 2.1. A historical perspective . . . . . . . . . . . . . . . 2.2. Shower partons . . . . . . . . . . . . . . . . . . . . 2.3. Parton distributions before recombination . . . . . 3. Large Baryon/Meson Ratios . . . . . . . . . . . . . . . 3.1. Intermediate pT in heavyion collisions . . . . . . . 3.2. Cronin effect . . . . . . . . . . . . . . . . . . . . . 3.3. Forward production in dAu collisions . . . . . . . 3.4. Forward production in AuAu collisions . . . . . . 3.5. Recombination of adjacent jets at LHC . . . . . . 4. Ridgeology Phenomenology of Ridges . . . . . . . . . . 4.1. Experimental features of ridges . . . . . . . . . . . 4.2. Recombination of enchanced thermal partons . . . 4.3. Trigger from the ridge . . . . . . . . . . . . . . . . 4.4. Dependence of ridge formation on trigger azimuth 5. Azimuthal Anisotropy . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
210 214 216 218 218 223 224 224 230 231 232 234 239 240 244 246 247 249 250 255 260
267 . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
267 269 269 272 274 275 275 278 280 282 283 284 285 288 290 292 296
QuarkGluon Plasma 4
x
5.1. Effects of ridge formation at low pT . . . . . 5.2. Effects of shower partons at intermediate pT 5.3. Breaking of quark number scaling . . . . . . 6. Hadron Correlation in Dijet Production . . . . . . 6.1. Distribution of dynamical path length . . . . 6.2. Nearside and awayside yields per trigger . . 6.3. Medium effects on dijets . . . . . . . . . . . . 6.4. Symmetric dijets and tangential jets . . . . . 6.5. Unsymmetric dijets and tomography . . . . . 7. Conclusion . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
Elliptic Flow: A Study of SpaceMomentum Correlations in Relativistic Nuclear Collisions
296 299 301 302 302 305 307 311 312 314
323
P. Sorenson 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1. Two decades in time and five decades in beam energy 1.2. Initial geometry: the reaction plane and eccentricity . 2. Review of Recent Data . . . . . . . . . . . . . . . . . . . . 2.1. Differential elliptic flow . . . . . . . . . . . . . . . . . 2.2. High pT . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3. Multiply strange hadrons and heavy flavor . . . . . . 2.4. Fluctuations and correlations . . . . . . . . . . . . . . 2.5. Scaling observations . . . . . . . . . . . . . . . . . . . 3. Confronting the Hydrodynamic Paradigm with RHIC Data 3.1. Transport model fits . . . . . . . . . . . . . . . . . . . 3.2. Viscous hydrodynamics . . . . . . . . . . . . . . . . . 3.3. Fluctuating initial conditions . . . . . . . . . . . . . . 3.4. Addressing uncertainties . . . . . . . . . . . . . . . . . 4. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
Predictions for the HeavyIon Programme at the Large Hadron Collider
323 326 327 332 335 339 343 346 351 359 362 364 366 368 368
375
N. Armesto 1. Introduction . . . . . . . . . . . . . . . . . . . . . 2. Qualitative Expectations . . . . . . . . . . . . . . 3. Bulk Observables . . . . . . . . . . . . . . . . . . . 3.1. Multiplicities . . . . . . . . . . . . . . . . . . 3.2. Collective flow . . . . . . . . . . . . . . . . . 3.3. Hadrochemistry at low transverse momentum 3.4. Correlations . . . . . . . . . . . . . . . . . . 3.5. Fluctuations . . . . . . . . . . . . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
375 379 389 389 395 399 403 405
Contents
4. Hard and Electromagnetic Probes . . . . . 4.1. Particle production at large transverse 4.2. Heavy quarks and quarkonia . . . . . 4.3. Photons and dileptons . . . . . . . . . 5. pA Collisions . . . . . . . . . . . . . . . . . 6. Summary and Discussion . . . . . . . . . .
xi
. . . . . . . . . . . . momentum and jets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
406 407 414 421 425 428
ENERGY LOSS IN A STRONGLY COUPLED THERMAL MEDIUM AND THE GAUGESTRING DUALITY
STEVEN S. GUBSER∗ , SILVIU S. PUFU† , FABIO D. ROCHA‡ and AMOS YAROM§ Joseph Henry Laboratories, Princeton University Princeton, NJ 08540, USA ∗
[email protected] †
[email protected] ‡
[email protected] §
[email protected] We review methods developed in the gaugestring duality to treat energy loss by energetic probes of a strongly coupled thermal medium. After introducing the black hole description of the thermal medium, we discuss the trailing string behind a heavy quark and the drag force that it implies. We then explain how to solve the linearized Einstein equations in the presence of the trailing string and extract from the solutions the energy density and the Poynting vector of the dual gauge theory. We summarize some efforts to compare these calculations to heavy ion phenomenology.
1. Introduction A longstanding hope, as yet unrealized, is that quantum chromodynamics (QCD) will be reformulated as a string theory. The gaugestring duality1–3 provides the closest approach to that goal so far attained. It provides useful computational methods for studying strongly coupled gauge theories. The theory that is most accessible via these methods is N = 4 superYangMills theory (SYM) in the limit of a large number of colors and large ’t Hooft coupling.a Aspects of the progress in using the gaugestring duality to understand QCD have recently been reviewed at a pedestrian level.4 The aim of the current article is a more focused review of efforts to understand energy loss by energetic probes of a thermal medium in a strongly coupled gauge theory, such as SYM, which has a string theory dual; and to review how energy loss in SYM can be compared to energy loss in QCD. At least in simple cases like SYM, the response of an infinite, static, strongly coupled thermal medium to an energetic probe can be presumed to be hydrodynamical far from the probe, because hydrodynamic perturbations are the only longwavelength modes available. In an infinite, interacting thermal medium, there is no aN
= 4 superYangMills theory is a gauge theory whose matter content consists of gluons, four Majorana fermions in the adjoint representation of the gauge group, and six real scalars, also in the adjoint representation. All the fields are related to one another by the N = 4 supersymmetry, which completely fixes the Lagrangian once the gauge group and gauge coupling are chosen. Our interest is in the gauge group SU (N ). 1
2
S. S. Gubser et al.
radiation, because there are no asymptotic states. In a strongly coupled medium, it is not clear that there is a gaugeinvariant distinction between collisional energy loss and radiative energy loss. So the main questions are: (1) What is the rate of energy loss from an energetic probe? (2) What is the hydrodynamical response far from the energetic probe? (3) What gaugeinvariant information can be extracted using the gaugestring duality about the nonhydrodynamic region near the probe? (4) Do the rate and pattern of energy loss have some meaningful connection to heavy ion phenomenology? In Sec. 2 we briefly review the dual description of the thermal state of SYM as an AdS5 Schwarzschild black hole. The reader interested in a more extensive discussion of the AdS/CFT duality is referred to various reviews in the literature.5–7 In Sec. 3 we explain how to describe heavy quarks in N = 4 SYM using strings in AdS5 , and in Sec. 4 we extract the drag force acting on the quark via an “obvious” and “alternative” identification of parameters between SYM and QCD. Also in Sec. 4, we consider how the string theory estimates of drag force relate to the measured nuclear modification factor for heavy quarks. The response of the stress tensor to the motion of the quark is dual to the metric perturbations around an AdS black hole. These are studied in Sec. 5, where we also discuss how the metric perturbations map into the stress tensor of the plasma. In Sec. 6 we provide analytic approximations to the stress tensor both near to the moving quark and far from it. The full numerical solution is described in Sec. 7, and its application to heavyion phenomenology can be found in Sec. 8. In Appendix A we provide a glossary of mathematical notations used in the main text. 2. The Thermal Medium as a Black Hole N = 4 SU (N ) superYangMills theory at finite temperature can be described in terms of N D3branes near extremality.8 D3branes are 3 + 1 dimensional objects on which strings may end.9, 10 Each string can end on one of the N D3branes, which eventually gives rise to the SU (N ) gauge symmetry of SYM.11 Having a nonvanishing tension, Dbranes themselves warp the spacetime around them. The near horizon geometry produced by the D3branes is AdS5 Schwarzschild times a fivesphere, threaded by N units of flux of the selfdual RamondRamond fiveform of type IIB supergravity.12 For most calculations of interest to us in this review, one can ignore the fivesphere and work just with the five noncompact dimensions. Our starting point is the Einstein equations of type IIB supergravity in the noncompact directions, which can be recovered from the action Z √ 1 12 5 Sbulk = 2 d x −G R + 2 . (1) 2κ L
Here κ is the fivedimensional gravitational coupling, G = det Gµν , and L is the radius of the S 5 . Standard relations based ultimately on the quantized charge of
Energy Loss and the GaugeString Duality
3
the D3brane lead to L3 = κ2
N 2π
2
.
(2)
AdS5 Schwarzschild is a solution of the Einstein equations following from (1). Its line element is L2 dz 2 2 µ ν 2 2 ds = Gµν dx dx = 2 −h(z)dt + d~x + , (3) z h(z) where the “blackening function” h(z) is given by h(z) = 1 −
z4 4 . zH
(4)
In this coordinate system, the conformal boundary of AdS5 is located at z = 0. The AdS5 Schwarzschild solution has a horizon at z = zH , whose temperature is 1 T = . (5) πzH According to the gaugestring duality, the temperature of the horizon (5) is also the temperature of the thermal medium in the dual gauge theory.3 This medium is infinite and static. Its energy density equals the mass per unit coordinate volume, d3 x, of the black hole, and its pressure p can also be straightforwardly computed as minus the free energy of the black hole. To leading order in the number of colors N and the ’t Hooft coupling λ, one finds π2 2 4 =p= N T . 3 8
(6)
These relations reflect the conformal invariance of N = 4 superYangMills theory, which is exact even at finite N and λ. If we send zH → ∞, or equivalently T → 0, we end up with a pure AdS5 geometry and no black hole. Perturbations of the AdS5 Schwarzschild black hole with wavelengths much longer than 1/T are described by relativistic fluid dynamics,13, 14 and the viscosity is known to be remarkably small15 : η 1 = , s 4π
(7)
again to leading order in the limit of large N and λ. In summary: A thermal medium of SU (N ) N = 4 superYangMills theory can be represented as a black hole in AdS5 in the limit of large N and large ’t Hooft coupling. Relations between the radius of AdS5 and the fivedimensional gravitational coupling, between the temperature and the position of the horizon, between the energy density and the temperature, and between the viscosity and the entropy density, can all be derived starting from tendimensional type IIB string theory.
4
S. S. Gubser et al.
3. The Trailing String N = 4 superYangMills theory has no fundamentally charged quarks. Instead, its field content is the gluon and its superpartners under N = 4 supersymmetry: namely, four Majorana fermions and six real scalars, all in the adjoint representation of SU (N ). The AdS5 description of SYM makes no direct reference to these colored dynamical fields. The magic of AdS/CFT is to replace the strong coupling dynamics of open strings, whose lowenergy quanta participate in the gauge theory, with the gravitational dynamics of closed strings (gravitons, for example) in a weakly curved background. For example, as we saw in Sec. 2, a finite temperature bath in the gauge theory is replaced by a black hole in AdS5 . The absence of colored degrees of freedom in the gravitational theory makes it less than obvious how to discuss energetic colored probes of the medium. A clue comes from the treatment of Wilson loops in the gaugestring duality.1, 16 A static, infinitely massive, fundamentally charged quark can be represented as a string hanging straight down from the boundary of AdS5 , at z = 0, into the horizon. There is no contradiction with the previous statement that N = 4 superYangMills has no fundamentally charged quarks, because the quark is an external probe of the theory, not part of the theory. An antifundamentally charged quark is represented in the same way as a quark, except that the string runs the other way. (In type IIB string theory, strings are oriented, so this statement makes sense.) If a quark and an antiquark are both present, it is possible for the strings running down into AdS5 to lower their total energy by connecting into a shape like a catenary. From the point of view of the boundary theory, the string configuration gives rise to an attractive potential with a 1/r dependence, as conformal invariance says it must. Let’s go back to a single isolated quark — which makes sense in N = 4 even at zero temperature because there is no confinement. The string dual to the quark can be described by its embedding in AdS5 . The twodimensional spacetime manifold swept out by the string is called a worldsheet and can be parameterized by two coordinates. If we use t and z to parameterize the string worldsheet, then the string in AdS5 , or in AdS5 Schwarzschild, dual to a static quark at ~x = 0, is described by the equation static quark:
~x(t, z) = 0 .
(8)
This is a solution to the string equations of motion with boundary conditions such that the string endpoint on the boundary is stationary. By symmetry, (8) is the only possible solution with these boundary conditions. In the absence of a medium, it must be that the description of a moving quark can be obtained by performing a Lorentz boost on the description of a static quark. If the boost is in the x1 direction, and we continue to use t and z to parameterize the worldsheet, then the equations describing the string dual to a quark moving in the absence of a medium are quark moving in vacuum:
x1 (t, z) = vt ,
~x⊥ (t, z) = 0 ,
(9)
Energy Loss and the GaugeString Duality
5
where v is the velocity of the quark and ~x⊥ = (x2 , x3 ).b Given that (8) solves the equations of motion of classical string theory, it’s guaranteed that (9) does too, because the equations are invariant under the boost in the x1 direction, and the line element (3) with h = 1 is invariant under any boost acting on the (t, ~x) coordinates. When h 6= 1, signaling the presence of a medium, a moving quark is not the same as a static one. Naively, one might nevertheless try to represent the moving quark in terms of the string shape described in (9). Let’s see why this is problematic. String dynamics is defined in terms of the metric on the worldsheet. We chose to parameterize the worldsheet by t and z, but any coordinates σ α = (σ 1 , σ 2 ) could have been chosen. For a general embedding xµ = xµ (σ 1 , σ 2 ) of a string worldsheet into spacetime, the worldsheet metric is gαβ =
∂xµ ∂xν Gµν . ∂σ α ∂σ β
(10)
What (10) says is that times and distances along the string are measured the same way as in the ambient spacetime. Using σ α = (t, z) and the ansatz (9), one immediately finds L2 −h + v 2 0 gαβ = 2 . (11) 1 0 z h This shows that the z direction on the worldsheet is always spacelike and that the t direction is timelike only when h > v 2 . Using (4), this condition is equivalent to p 4 z < z∗ ≡ zH 1 − v 2 . (12)
The part of the string worldsheet with z > z∗ is purely spacelike, which means that the string is locally moving faster than the speed of light. This does not make sense if we are aiming to describe the classical dynamics of a string moving in real Minkowski time. There is a simpler way to arrive at (12): any trajectory of a point particle in the bulk of AdS5 Schwarzschild with x1 = vt must be spacelike — that is, the speed of the particle will exceed the local speed of light — if the inequality z < z∗ is violated. This conclusion holds even when there are other components of the velocity, but the inequality is sharp only in the case where there aren’t. Physically, what we learn from (12) is that it is harder and harder to move forward as one approaches the horizon. This evokes the idea that there must be some drag force from the medium. But how does one get at that drag force? An answer was provided by two groups,17, 18 and closely related work on fluctuations appeared at the same time.19 The string does not hang straight down from the quark: rather, it trails out behind it. If the shape is assumed not to change as b It
would be more proper to use (x1 , x2 , x3 ) in place of (x1 , x2 , x3 ), and reserve the notation xµ for Gµν xν . However, it simplifies notation to set xi = xi for i = 1, 2, 3, and we will do this consistently.
6
S. S. Gubser et al.
the quark moves forward, and if it respects the SO(2) symmetry rotating the ~x⊥ coordinates, then it must be specified by a small variant of (9): x1 (t, z) = vt + ξ(z)
(13)
for some function ξ(z). If we insist that the quark’s location on the boundary is x1 = vt, then we must have ξ → 0 as z → 0. To determine ξ(z), one must resort to the classical equations of motion for the string. These follow from the action Z √ 1 d2 σ −g , (14) Sstring = − 0 2πα
where g = det gαβ . The parameter α0 is related to L and the ’t Hooft coupling by L4 = λα02 .
(15)
2 Here we define λ = gYM N , so that a quantity analogous to the coupling αs in QCD is αYM = λ/4πN . The classical equations of motion following from the action (13) take the form
∇α pα µ = 0 ,
(16)
where 1 αβ g Gµν ∂β xν (17) 2πα0 is the momentum current on the worldsheet conjugate to the position xµ . Plugging (13) into (17), one straightforwardly finds that s πξ dξ h − v2 = , (18) 4 L 2 dz h z 4 h − πξ pα µ ≡ −
where πξ is a constant of integration. In order for ξ(z) to be real, the right hand side of (18) must be real. There are three ways this can happen in a manner consistent with the assumption of steadystate behaviorc : • One can choose πξ = 0. This leads to ξ = 0, which shows that (9) is formally a solution of the equations of motion. But it is not a physical solution — at least, not in the context of classical motions of a string — because of the problem with the signature of the worldsheet metric. The action for this solution is complex. • One can arrange for the worldsheet never to go below z = z∗ , and choose πξ small enough that the denominator inside the square root in (18) is always positive. There is indeed a oneparameter family of such solutions, and they describe a heavy quark and antiquark in a color singlet state propagating c Technically,
there is a fourth way, but its significance is obscure to us. A string can lead down to z = z∗ in the shape of the trailing string, (20), and then turn around and retrace its path back up again. The energy localized at the kink must grow linearly with time, which means that this configuration is not quite a steadystate solution.
Energy Loss and the GaugeString Duality
7
without drag (at the level of the current treatment), one behind the other. Similar states were studied by other groups,20–22 but they do not capture the dynamics of a single quark propagating through the plasma, so we do not consider them further here. • One can choose v L2 p L2 (19) πξ = ± 2 h(z∗ ) = ± √ 2 , z∗ 1 − v 2 zH so that the denominator inside the square root in (18) changes sign at the same value of z as the numerator, namely z = z∗ , rendering the ratio inside the square root everywhere positive and finite. Choosing the sign that makes πξ positive means that the string trails out in front of the quark instead of behind it. Although this is technically a solution, it does not describe energy loss and should be discarded. Choosing πξ negative leads to the trailing string solution that we are interested in. Equation (18) then straightforwardly leads to 1 − iy 1+y zH v ξ=− log + i log , (20) 4i 1 + iy 1−y where we have introduced a rescaled depth variable, z y= . zH
(21)
There is another way to justify the choice of the minus sign, not the plus sign, in (19)23 : The solution (20) is nonsingular at the future event horizon, whereas the solution one would get with the opposite sign choice is singular. To understand this point, it is convenient to pass to Kruskal coordinates, defined implicitly by the equations UV = −
1 − y −2 tan−1 y e 1+y
V = −e4t/zH . U
(22)
In the region outside the horizon, U < 0 and V > 0, while in the region inside the future horizon, U > 0 and V > 0. The trailing string solution (13), with ξ given by (20), can be extended to a nonsingular solution over the union of these two regions: v x1 = log V + v tan−1 y . (23) 2 Thus, the logarithmic singularity in ξ at y = 1 (meaning z = zH ) is a singularity not at the future horizon, which is at U = 0, but at the past horizon, which is at V = 0. Reversing the sign choice in (19) would lead to a solution that is singular at the future horizon but not the past horizon. Causal dynamics in the presence of a black hole horizon can generally be described in terms of functions which are smooth at the future horizon. At the level of our presentation, it has been assumed rather than demonstrated that the trailing string is a stable, steadystate configuration representing the late
8
S. S. Gubser et al.
time behavior of a string attached to a moving quark on the boundary of AdS5 Schwarzschild. In fact, this has been fairly well checked.17, 24 The description we have given of the trailing string is not limited to Schwarzschild black holes in AdS5 . One may extend this analysis to various other black hole geometries which asymptote to AdS5 near their boundary.25–30 These geometries describe theories which are deformations of SYM. The literature also includes a discussion of the distribution of energy along the string31 and an interpretation of the shape of the string in terms of a rapid cascade of strongly coupled partons.32 4. The Magnitude of the Drag Force As we explained in the previous section, an infinitely massive, fundamentally charged quark moving at speed v in the x1 direction through an infinite, static, thermal medium of N = 4 superYangMills theory can be described at strong coupling in terms of the trailing string solution (20). The quark cannot slow down because it has infinite mass. However, it does lose energy and momentum at a finite, calculable rate. In fivedimensional terms, this energy can be thought of as flowing down the string toward the black hole horizon. In fourdimensional terms, energy and momentum emanates from the quark and eventually thermalizes. To calculate the fourmomentum ∆pm delivered from the quark to the bath over a time ∆t, one can integrate the conserved worldsheet current pα m of spacetime energymomentum over an appropriate linesegment I on the worldsheet. I should cover a time interval ∆t, and it can be chosen to lie at a definite depth z0 in AdS5 . The fourmomentum ∆pm isd Z √ ∆pm = − dt −g pz m . (24) I
Because the trailing string is a steadystate configuration, fourmomentum is lost at a constant rate: √ dpm = − −g pz m . (25) dt In particular, the drag force can be defined as Fdrag =
√ L2 v dp1 √ = − −gpz 1 = − . 2 0 dt 2πzH α 1 − v 2
(26)
Using (5) and (15), one obtains Fdrag d There
√ π λ 2 v =− T √ . 2 1 − v2
(27)
is an explicit minus sign in (24) which doesn’t appear in the analogous equation of one of the original works.18 This is due to use of the z variable, which increases as one goes deeper into AdS5 , instead of the r = L/z variable, which increases as one goes out toward the boundary.
Energy Loss and the GaugeString Duality
9
If, instead of an infinitely massive quark, we consider a quark with finite but large mass m, then using the standard relativistic expression mv p= √ (28) 1 − v2
leads to
Fdrag = −
√ π λ 2p T . 2 m
(29)
It has been explained17 that (28) receives corrections when a finite mass quark is described as a string ending at a definite depth z = z∗ on a D7brane. While these corrections are interesting, it would take us too far from the main purpose of this review to give a proper explanation of how the D7branes modify the physics. Our discussion is formal because we derived the result (27) in the strict m = ∞ limit and then applied it to finite mass quarks. From (29) it is clear that the drag force causes the momentum of a quark to fall off exponentially: p(t) = p(0)e−t/tquark
where
2 m tquark = √ . π λ T2
(30)
In order to make a physical prediction for QCD, we must plug in sensible values for m, λ, and T . The effective quark mass in the thermal medium is already nontrivial to specify precisely, but mc = 1.5 GeV for charm and mb = 4.8 GeV for bottom are reasonably representative values which were used in a recent phenomenological study.33 We will review here two approaches34 to specifying λ and T . The first approach, used earlier in a calculation of the jetquenching parameter qˆ from a lightlike Wilson loop in N = 4 superYangMills theory,35 is to identify the temperature TSYM with the temperature TQCD , and then identifying the gauge coupling gYM of superYangMills with the gauge coupling gs of QCD evaluated at temperatures typical of RHIC. Because of the proximity of the confinement transition, gs has substantial uncertainty. A standard choice is αs = 0.5, corresponding to gs ≈ 2.5. With the number of colors N set equal to 3, one finds λ ≈ 6π. We will refer to this as the “obvious scheme.” The second approach, called the “alternative scheme,” is based on two ideas. The first idea is that it may make more sense to compare SYM to QCD at fixed energy density than fixed temperature. SYM has ∝ T 4 , and so does QCD, approximately: this approximation is surprisingly good for T ≥ 1.2Tc , according to lattice data.36 But the constant of proportionality is about 2.7 times bigger for SYM than for QCD.e That is, SYM has about 2.7 times as many degrees of freedom as QCD above the confinement transition. So SYM = QCD implies TSYM ≈√TQCD /(2.7)1/4 . This identification leads to a suppression of Fdrag by a factor of 2.7 relative to e This mismatch has previously been stated34 as a factor of 3 rather than 2.7. Some uncertainty exists on both the SYM and the QCD sides, because of finite coupling effects and time discretization, respectively; but 2.7 is probably closer to the true figure.
10
S. S. Gubser et al. Table 1. The obvious and alternative schemes for comparing SYM and QCD.
Scheme
TQCD MeV
TSYM MeV
QCD GeV/fm3
SYM GeV/fm3
λ GeV
mc GeV
mb fm
tc fm
tb
obvious alternative
250a 250a
250 195
5.6b 5.6b
15 5.6
6π 5.5
1.5 1.5
4.8 4.8
0.69 2.1
2.2 6.8
a
We set TQCD = 250 MeV because this is a typical temperature scale for heavy ion collisions at √ sNN = 200 GeV.
b
We use /T 4 ≈ 11 for QCD.
the obvious scheme. The second idea behind the alternative scheme is that the ’t Hooft parameter λ in string theory can be determined by comparing the static force between a quark and an antiquark, as calculated in string theory, to the same force calculated in lattice gauge theory. The string theory calculation is based on a Ushaped string connecting the quark and the antiquark. This string pulls on the static quarks in a fashion that is similar to how the trailing string pulls on a moving quark. The lattice calculation is based on computing the excess free energy due to the presence of an external quark and antiquark in a thermal bath. There is a significant difficulty: in the simplest string theory calculation, based only on the Ushaped string, the force between the quark and antiquark vanishes for separations larger than some limiting distance r∗ , and this distance is quite small: r∗ ≈ 0.24 fm when TSYM ≈ 195 MeV (corresponding to TQCD = 250 MeV). It has been pointed out37 that exchange of closed strings between two long strings describing the quark and antiquark at separations r > r∗ contribute to the quarkantiquark force at the same order in N as the Ushaped string. Unfortunately, it is hard to compute the contribution of closed string exchange. The approach34 is therefore to match the Ushaped string computation to lattice data38 near the limiting distance r∗ . The result of this matching is λ ≈ 5.5. As we show in Table 1, heavy quark relaxation times tc and tb are remarkably short when one uses the obvious scheme, and somewhat larger in the alternative scheme. The uncertainties in tc and tb are substantial: even if one accepts the ideas behind the alternative scheme, one should probably regard the resulting relaxation times as uncertain by a factor of 1.5. An experimental study39 favors a model40 in which tc is roughly 4.5 fm at TQCD = 250 MeV, as estimated from plots from a detailed exposition of that model.41 This seems to indicate that the string theory estimates of tc and tb , even in the alternative scheme, are too short. However, the results of a recent phenomenological study33 favor a range of parameters that is consistent with the string theory predictions translated to QCD using the alternative scheme. Let us briefly review the recent study.33 The starting point is the Langevin equation, which in the Itˆ o discretization scheme takes the form ∆~x(t) =
p ~ ∆t E
~ . ∆~ p(t) = −Γ~ p∆t + ξ(t)
(31)
Energy Loss and the GaugeString Duality
11
~ Here ξ(t) is a stochastic force, assumed to be Gaussian and uncorrelated from one timestep to the next. The strength of the stochastic force is related to the drag coefficient Γ by demanding that the relativistic MaxwellBoltzmann distribution p is preserved by the time evolution (31). Ordinary relativistic kinematics, E = p~ 2 + m2 , are assumed. It is also assumed that Γ=γ
T2 , m
(32)
where γ is a dimensionless quantity with no p dependence. Evidently, tquark = 1/Γ. The temperature in (32) is TQCD , whereas the temperature in (30) is TSYM . Comparing these two equations, one finds that the string theory prediction is √ 2 ( 6.8 obvious scheme π λ TSYM γ= = (33) 2 TQCD 2.2 alternative scheme. The alternative scheme value in (33) is fractionally larger than the one quoted in the study under discussion,33 due to the use here of the factor 2.7 for the ratio of degrees of freedom between SYM and QCD, as compared to 3 in previous work.34 (It is a numerical coincidence that the dimensionless factor γ is the same, in the alternative scheme, as tc in femtometers when TQCD = 250 MeV.) The next step of the study33 is to compare Langevin dynamics of heavy quarks in a hydrodynamically expanding plasma to PHENIX39 and STAR42 data on the nuclear modification factor RAA for nonphotonic electrons — meaning electrons and positrons coming from decays of heavyquark mesons. Because of the treatment of hadronization, the theoretical results are deemed trustworthy only when the transverse momentum pT of the nonphotonic electron is at least 3 GeV.f For fairly central collisions (impact parameter b = 3.1 fm), agreement between theory (using the alternative scheme) and experiment is best for γ between 1 and 3: see Fig. 1. Thus the prediction (33) of string theory in the alternative scheme can reasonably be said to agree with data to within the uncertainties of the calculations. These uncertainties stem in large part from the difficulty of comparing SYM to QCD; however, it is also clear that the treatment of hadronization is a significant hurdle.33 Fluctuations of the trailing string19, 23, 24, 43 provide direct access to the stochastic forces in (31). In the nonrelativistic limit, the size of these forces, relative to the drag force, is exactly what is needed to equilibrate to a thermal distribution. Indeed, the original calculations of drag force17, 18 and stochastic forces19 were done independently. For relativistic quarks, the stochastic forces are enhanced by powers of 1/(1 − v 2 ), including enhancement of longitudinal stochastic forces by the startlingly large factor 1/(1 − v 2 )5/4 . This is larger by 1/(1 − v 2 )3/4 than what is needed to equilibrate to a thermal distribution. Another issue is that the correlation f The
electron carries only a fraction of the pT of the charm quark that led to its production. This fraction varies, but a reasonable ruleofthumb value is 1/2.
12
S. S. Gubser et al.
1.8 1.6
(a) b=3.1fm, c+b>e±
RAA
1.4 1.2 1 0.8 0.6 0.4 0.2 0 0
1
2
3
4 5 6 p [GeV]
γ = 0.3 γ = 1.0 γ = 3.0 PHENIX(010%) STAR(05%)
7
8
9 10
T
Fig. 1. (Color online) Comparison33 between experimental data for goldgold collisions with impact parameter b = 3.1 fm and theoretical predictions based on Langevin dynamics of heavy quarks in a hydrodynamically expanding plasma. The open circles correspond to central collisions in a PHENIX experiment,39 while the solid circles come from a STAR experiment.42 Theoretical predictions are plotted for γ = 0.3 (red), 1 (blue), and 3 (green), with γ is as defined in (32). Reprinted with permission from the authors from Y. Akamatsu, T. Hatsuda, and T. Hirano, “Heavy Quark Diffusion with Relativistic Langevin Dynamics in the QuarkGluon Fluid,” 0809.1499.
time tcor for these stochastic forces grows with velocity: based on results for the relevant Green’s function24 one may estimate tcor ≈
1 . πT (1 − v 2 )1/4
(34)
If we use tcor < tquark as a criterion of validity for the Langevin dynamics, then we get a limit on the Lorentz boost factor: 1 4 m2 √ < . λ T2 1 − v2
(35)
Approximately the same inequality arises from demanding that when the string ends on a D7brane, its endpoint should not move superluminally. The same inequality, but with m replaced by some fixed scale µ, also arises from demanding that the worldsheet horizon23, 24 should be at a depth in AdS5 corresponding to a scale µ where the dynamics of QCD is far from weakly coupled. Plugging in numbers in the √ alternative scheme, with µ = 1.2 GeV and TQCD = 250 GeV, one finds that 1/ 1 − v 2 < ∼ 30. This inequality should be understood as quite a rough estimate, because of the quadratic dependence on the quantity µ which is only qualitatively defined. For a charm quark, the corresponding limit on the transverse
Energy Loss and the GaugeString Duality
13
momentum of the nonphotonic electron is pT < ∼ 20 GeV. This is well in excess of the highest momentum for which there are statistically significant data; moreover, for√pT more than a few GeV, bottom quarks dominate, and for them the bound < 1/ 1 − v 2 < ∼ 30 translates to pT ∼ 70 GeV for nonphotonic electrons. The upshot is that heavy quarks at RHIC do not obviously fall outside the regime of validity of a selfconsistent Langevin treatment based on the trailing string — except for the troublesome scaling of longitudinal stochastic forces as 1/(1 − v 2 )5/4 , whose consequences, we feel, are illunderstood. In another study,44 it is argued that energy loss in the perturbative and strongly coupled regimes have experimentally distinguishable signatures for large enough transverse momenta, pT mb , which will be attained at the LHC. A convenient observable that distinguishes between predictions of string theory and perturbative QCD is the ratio of the nuclear modification factors for b and c quarks, Rcb =
c RAA (pT ) . b RAA (pT )
(36)
In particular, the drag force formula (29) implies that at large enough pT , one has mc cb RAdS ≈ ≈ 0.3 , (37) mb where we used the bottom and charm masses quoted right after equation (30). In contrast, perturbative QCD predicts that pcb cb RpQCD ≈1− (38) pT at large pT , where pcb is a relevant momentum scale. So according to perturbative QCD, Rcb should approach unity at large pT . At sufficiently large pT , the trailing string treatment presumably fails, and perturbative QCD presumably is correct. But as discussed following (35), it is difficult to give a good estimate of the characteristic value of pT where the trailing string fails. Absent a reliable estimate of the characteristic pT , the upshot is that if the measured Rcb is significantly below the perturbative prediction, the trailing string should be considered as a candidate explanation. 5. The Perturbed Einstein Equations Given that an external quark dual to the trailing string described in Sec. 3 experiences drag, one might ask what happens to the energy that the quark deposits in the medium. At scales much larger than the inverse temperature, one expects the excitations present in the medium due to interactions with the moving quark to be welldescribed by linearized hydrodynamics. Earlier investigations of linearized hydrodynamics revealed the presence, for generic sources, of both a sonic boom and a diffusion wake.45 The sonic boom is a directional structure, which, in an ideal fluid, is concentrated on the Mach cone, but in a real fluid there is broadening because of viscous effects. It appears only when the probe is moving faster than the
14
S. S. Gubser et al.
T
mn
q
v
R3,1 AdS 5 −Schwarzschild
h mn
fundamental string
black hole horizon Fig. 2. (Color online) A visual summary46 of the calculation of the response hTmn i of a strongly coupled thermal medium to a heavy quark moving with a speed v via AdS/CFT. hmn is a perturbation of the metric caused by the trailing string.
speed of sound in the medium, and it comes from constructive interference among spherical waves sourced by the quark along its trajectory. The diffusion wake is a flow of the medium behind the quark in the direction of the quark’s motion. It too is broadened by viscous effects. We will discuss the hydrodynamic limit in more detail in Sec. 6.1.2. Apart from a qualitative understanding of energy loss at large distances, little is known a priori about what happens, for example, at small distances close to the quark. An allscales description can be achieved using the gaugestring duality, where one computes the disturbances in the stressenergy tensor due to the presence of the quark. This was done in a series of papers.46–53 The linearized response of the lagrangian density in the dual field theory was also computed in Fourier space.54, 55 The purpose of this section and the next two is to present a reasonably selfcontained summary of how the gauge theory stressenergy tensor is computed starting from the trailing string. A visual summary of the main elements of the computation is shown in Fig. 2. In the context of the AdS/CFT duality, the stressenergy tensor hTmn i of the boundary theory is dual to fluctuations of the metric in the bulk. So in order to compute the expectation value of hTmn i, one first needs to compute to linear order the backreaction of the string describing the quark on√the metric. Nonlinear corrections to the Einstein equations will be suppressed by λ/N . The total action describing both the string and the metric is S = Sbulk + Sstring , which one can write as √ Z Z −G(R + 12/L2) 1 2 √ 5 µ µ S = d5 x − d σ −g δ (x − x (σ)) , (39) ∗ 2κ2 2πα0 where xµ∗ (σ) is the embedding function of the string in AdS5 Schwarzschild. In a gauge where we parameterize the string worldsheet by σ α = (t, z), xµ∗ (σ) is
Energy Loss and the GaugeString Duality
15
given by xµ∗ = t vt + ξ(z) 0 0 z ,
(40)
with ξ(z) as given in (20). The equations of motion following from (39) are just Einstein’s equations: 6 1 Rµν − Gµν R − 2 Gµν = τ µν , 2 L
(41)
where τ µν = −
3 p κ2 z H y 3 1 − v 2 δ(x1 − vt − ξ(z)) δ(x2 )δ(x3 )∂α xµ∗ ∂ α xν∗ 0 3 2πα L
(42)
is the bulk stressenergy tensor of the trailing string. To compute the backreaction of the string on the metric, we write Gµν = G(0) µν + hµν
(43)
(0)
where Gµν is the unperturbed AdS5 Schwarzschild metric given in (3), and plug this into (41) to obtain the linearized equations of motion for the metric perturbations hµν . The resulting equations take the form Lhµν = τ µν ,
(44)
νσ 56 where hµν = Gµρ (0) G(0) hρσ and L is the differential operator given by ρ Lhµν = −hµν − 2Rµρνσ hρσ + 2R(µ hν)ρ − ∇µ ∇ν h + 2∇(µ ∇ρ hν)ρ 6 ρσ ρσ + Gµν (−∇ ∇ h + h + R h ) − − R hµν . ρ σ ρσ (0) L2
(45)
The covariant derivatives, the Riemann and Ricci tensors, and the Ricci scalar (0) appearing in (45) are computed using the background metric Gµν , and we have (0) µν denoted h = Gµν h . For a steadystate solution of (44), hµν depends on x1 and t only through the combination x1 − vt. We therefore pass to comoving Fourier space variables by writing τ
µν
1
2
3
(t, x , x , x , z) =
Z
1 2 3 d3 K µν τ (z)ei[K1 (x −vt)+K2 x +K3 x ]/zH , (2π)3 K
(46)
~ ≡ ~kzH = ~k/πT . We make where we have defined the dimensionless wavevector K a similar expansion for hµν . In Fourier space, (44) can then be written as µν LK hµν K = τK ,
(47)
16
S. S. Gubser et al.
µν with τK being given by
v2 y2 h + v2 y4 v 0 0 h2 h h v 2 2 v 0 0 vy 2 −iK1 ξ(z)/zH 2 5 z y κ e h µν H , √ (48) τK = 0 0 0 0 0 2πα0 L5 1 − v2 0 0 0 0 0 2 2 v y vy 2 0 0 v 2 − h h where y is the rescaled depth coordinate defined in (21), but the tensor components are given in the (t, x1 , x2 , x3 , z) coordinate system. The explicit form of LK is too complicated to be reproduced here. We will decouple equations (47) by passing to a gauge where hµz = 0, which we will refer to as “axial gauge.” Note that hµz = 0 leaves some residual gauge freedom. We will discuss this shortly. The rest of this section is organized as follows. In Sec. 5.1 we explain how to decouple equations (47). In Sec. 5.2 we explain the boundary conditions needed to solve these equations. Lastly, in Sec. 5.3 we explain how the onepoint function of the SYM stressenergy tensor is related to the asymptotic behavior of the metric perturbations near y = 0.
5.1. Metric perturbations in axial gauge As mentioned above, we choose a gauge where hµz H00 H01 H10 H11 1 κ2 L √ hK = H20 H21 µν 2πα0 1 − v 2 z 2 H30 H31 0 0
= 0, and let H02 H12 H22 H32 0
H03 H13 H23 H33 0
0 0 0 . 0
(49)
0
Rotational symmetry around the direction of motion of the quark allows us to set ~ = K1 K⊥ 0 with K⊥ > 0. Defining K q K⊥ 2 K = K12 + K⊥ ϑ = tan−1 , (50) K1 one can form the following linear combinations of metric perturbations: −H11 + 2 cot ϑH12 − cot2 ϑH22 + csc2 ϑH33 2v 2 H03 H13 + tan ϑH23 B1 = B2 = − v v2 A=
C = − sin ϑH13 + cos ϑH23
(51a) (51b) (51c)
Energy Loss and the GaugeString Duality
H01 − cot ϑH02 −H11 + 2 cot 2ϑH12 + H22 D2 = 2 2v 2v 2 1 3 H01 + tan ϑH02 E1 = − H00 + H11 + H22 + H33 E2 = 2 h 2v
D1 =
H11 + H22 + H33 2 −H11 − H22 + 3 cos 2ϑ(−H11 + H22 ) + 2H33 − 6 sin 2ϑH12 . E4 = 4 E3 =
17
(51d) (51e) (51f) (51g)
Using these new variables, the Einstein equations (47) decouple into five sets46 : K2 3 h0 y ∂y + 2 v 2 cos2 ϑ − h A = e−iK1 ξ/zH (52a) ∂y2 + − + y h h h "
∂y2
+
− y3 0
0 − y3 +
h0 h
!
K2 + 2 h
# −h v 2 h cos2 ϑ B1 0 = −1 v 2 cos2 ϑ B2 0
B10 − hB20 = 0 "
∂y2
+
− y3
∂y2
K2 2 3 h0 2 + − + ∂y + 2 v cos ϑ − h C = 0 y h h 0
− y3 +
0
h0 h
!
K2 + 2 h
2 ∂y +
− y3 +
3h0 2h
0
0 0
0
0
0
− y3
0
0
0
0
(52c) (52d)
# y −iK1 ξ/zH 1 −h v 2 h cos2 ϑ D1 = e −1 v 2 cos2 ϑ D2 1 h (52e) D10 − hD20 = 0
(52b)
− y3 + 0
h0 2h
0 −3 y
+
(52f)
h0 h
∂y
−2h 12v 2 cos2 ϑ 6v 2 cos2 ϑ + 2h 0 E1 2 K 0 0 2h h E2 + 2 0 0 −2h −h E3 3h 2 2 2 2 2h −12v cos ϑ 0 3v cos ϑ + h E4
=
y −iK1 ξ/zH e h
2
1 + vh 1 2 −1 + v −
v2 h 1+3 cos 2ϑ 2 v 2
(52g)
18
S. S. Gubser et al.
0 1 1 0 −h 0 −3v 2 cos2 ϑ − h −h ∂y h 0 2 0 +
=
0
0
0 −6h −3h 0 1 3(3v 2 cos2 ϑ + h)h0 0 −3hh0 18v 2 cos2 ϑh0 6h 2K 2 yh −2K 2 v 2 y cos2 ϑ −2K 2 y(3v 2 cos2 ϑ − h) 2K 2 yh
−ivy sec ϑ h0 −iK1 ξ/zH e 3ivy cos ϑ(v 2 + h) . 4Kyh K(v 2 − h)
E1 E2 E3
E4
(52h)
In (52), A, Bi , C, Di , and Ei are all functions of the rescaled AdS depth y ≡ z/zH , and primes denote derivatives with respect to y. A few comments are in order. Let’s start by counting the equations. Einstein’s equations (47) consist of fifteen linearly independent equations that split between the A, B, C, D, and E sets as follows: the A and C sets each consist of one second order equation for one unknown function; the B and D sets each consist of two second order equations and one first order constraint for two unknown functions; lastly, the E set consists of four second order equations and three first order constraints for four unknown functions. At first glance, the B, D, and E systems of equations might seem overdetermined. A more careful analysis shows that the constraints are consistent with the second order equations in the sense that if they hold at a particular value of y, they continue to hold at all y. We can therefore think of the constraint equations as reducing by five the number of integration constants in the second order equations. There are fifteen remaining integration constants that are fixed by the boundary conditions which we discuss in Sec. 5.2. Because the Bi and C equations, (52b)–(52d), have no source terms, one can consistently set B1 = B2 = C = 0. This identification is in fact enforced if we insist that the response of the medium should respect the same axial symmetry around the direction of motion of the quark that the trailing string does. In order to keep the discussion of perturbations, boundary conditions, and integration constants general, let us not discard the Bi and C fields just yet. A more general source would force them to be nonzero. Equations (52) can be reduced to just five equations using the residual gauge symmetry.53, 57 The action (39) is reparameterization invariant, and infinitesimal gauge transformations act by sending hµν → hµν + ∇µ ζν + ∇ν ζµ
(53)
for any oneform ζµ dxµ . This allowed us to pass to axial gauge in the first place: if we had started with some arbitrary metric perturbations hµν we could solve, at least locally, ∇z ζµ + ∇µ ζz = −hµz ,
(54)
Energy Loss and the GaugeString Duality
19
so applying the infinitesimal gauge transformation (53) gives hµz = 0. Residual gauge symmetry arises from the fact that (54) specifies ζµ only up to five integration constants, which in general are functions of t and ~x. Put differently, there are five linearly independent gauge transformations that preserve the axial gauge condition hµz = 0, and the corresponding ζµ are given by the linearly independent solutions to (54) with hµz = 0. For the steadystate solution of (47) the allowed gauge transformations are the m ones where the t and ~x dependence of ζµ is of the form eiKm x /zH , where we have defined Km = −K1 v K1 K2 K3 . (55) As in previous sections, our convention is that lower case Roman indices m, n, . . . take values in (t, x1 , x2 , x3 ). The first four of these gauge transformations are parameterized by m
µ ζ(a) =
κ2 zH eiKm x /zH µ √ δ , 2πα0 L 1 − v 2 a
where a = 0, 1, 2, or 3, while the fifth is given by m 2 √ µ κ2 zH eiKm x /zH µ µ 2 µ 0y j √ ζ(5) = 2y hδ5 − iK δ − iK arcsin y δj . 2πα0 L 1 − v 2 h 0
(56)
(57)
The corresponding pure gauge solutions are given by 2 (0) z Hmn(a) = −2iK(mGn)a 2 a = 0, 1, 2, 3 L √ √ 0 0 Hmn(5) = 4 h ηmn − 2Km Kn arcsin y 2 − 4y 4 h δm δn h √ i 2 2 0 + 2K K(m ηn)0 arcsin y − y h ,
(58)
(59)
where ηmn is the Minkowski metric. One can straightforwardly check that (58)–(59) satisfy equations (47). Using the definitions (51), we can work out the pure gauge solutions in the ABCDE variables. We find that A and C are invariant, which was to be expected since their equations of motion are already fully decoupled. The other variables, however, do transform nontrivially under (53) with (56) and (57): for example the Bi and Di variables vary by iK1 (λ1 − λ2 cot ϑ) 2v
δB1 = iλ3 K1
δD1 =
iλ3 K1 δB2 = 2 v cos2 ϑ
iK1 (λ1 − λ2 cot ϑ) , δD2 = 2v 2 cos2 ϑ sin2 ϑ
(60)
where λa are arbitrary constants multiplying the pure gauge solutions parameterµ ized by ζ(a) .
20
S. S. Gubser et al.
From these transformation laws it is easy to see that there is a linear combination of B1 and B2 that is gaugeinvariant (and the same is true for D1 and D2 ). A choice of gauge invariants is given by B = B1 − cos2 ϑ v 2 B2
(61)
D = D1 − cos2 ϑ v 2 D2 .
The transformation law for the Ei variables is simple to derive but its exact form is not very enlightening. The corresponding gaugeinvariant can be taken to be E = 4E3 + E4 + (12E2 − 3E4 )v 2 cos2 ϑ − (2E1 + E4 )h(z) .
(62)
From (52) we obtain the equations of motion for the B, D, and E invariants: 2 K 2 (v 2 cos2 ϑ − h) 3h + (h − 4)v 2 cos2 ϑ ∂ + B=0 (63) ∂y2 + y yh(v 2 cos2 ϑ − h) h2
∂y2
3h2 + (h − 4)v 2 cos2 ϑ K 2 (h − v 2 cos2 ϑ) ∂ + D y yh(v 2 cos2 ϑ − h) h2 4ivy 5 2 2 −iK1 ξ/zH y 1 − v cos ϑ + =e h K(v 2 cos2 ϑ − h)
∂y2 +
+
1 4 16y 4 − + y yh 4 − 6v 2 sin2 ϑ + 2h
−
∂y
K12 v 2 32y 8 K1 + − E h2 y 2 h(4 − 6v 2 sin2 ϑ + 2h) h cos2 ϑ
y 3v 2 cos2 ϑ − 2 − v 2 h 2 + h − 3v 2 cos2 ϑ 8iy 5 4 4 2 2 8 × 9v cos ϑ − 18v cos ϑ 1 + −y +9 . 3vK1
=e
(64)
−iK1 ξ zH
(65)
The gauge invariants we just described are unique up to an overall zdependent factor, provided we only consider linear combinations of the Bi , Di , and Ei variables. If we also allow derivatives of the latter there are many other choices of gauge invariants that might prove useful. For instance, 1 2 K E 3 2 → 2v sin ϑzH D10 Z1 = vzH B10 Z0 =
← → Z2 =
−A sin2 ϑv 2 −C
(66) −C
A sin2 ϑv 2
!
.
Energy Loss and the GaugeString Duality
21
is another set of gauge invariants which has been used in the literature.53 Yet another set, consisting of “master fields,”47 is ψTodd = − ψTeven = − ψVodd = ψVeven
3 zH αv C L6 3 2 zH v αv A sin2 ϑ L6
2 zH vαv hB20 2L4
(67)
z 2 vαv = H 4 hD20 sin ϑ L
2 00 K 2 zH αv 3h yE4 + 6h2 E20 − 3h(3 + y 4 )E40 6L2 hy(K 2 + 6y 2 ) + 2hyK 2 (E1 − E3 ) + 3y 2hy 2 − K12 v 2 (4E2 − E4 ) , √ where αv = 1/2πα0 1 − v 2 . The gauge invariants (67) will prove useful for the asymptotic analysis of Sec. 6.2. Below, we list their equations of motion: 3 h0 K 2 (v 2 cos2 ϑ − h) 00 ψT + − + ψT0 + ψT = −JT (68a) y h h2 3 h0 K 2 v 2 y 2 cos2 ϑ + h(3 − K 2 y 2 + 9y 4 ) 00 ψV + − + ψV0 + ψV = −JV (68b) y h y 2 h2 0 2 2 h 3 K 2 (4 − K 2 y 2 ) + 12y 2 (6 − y 4 ) K1 v ψS00 + − ψS0 + + ψS = −JS (68c) h y h2 y 2 h(K 2 + 6y 2 )2 ψS = −
where JTodd = JVodd = 0 JTeven =
3 zH αv v 2 y sin2 ϑ −iK1 ξ/zH e L6 h
2 zH αv e−iK1 ξ/zH v sin ϑ(5y 4 − 1 − iKvy 3 cos ϑ) + iKy 3 tan ϑ L4 h zH αv K 2 ye−iK1 ξ/zH h JS = (2 + v 2 ) 2y 2 (K 4 − 45) − 3K 2 4 2 2 2 6L (K + 6y )
JVeven =
(69) (70) (71)
+ 3K 2 (2 − 5v 2 )y 4 + 18(2 + 3v 2 )y 6 − 3iKv 3 y 3 (K 2 − 12y 2 ) cos ϑ + 3v 2 90y 2 − K 4 (4 + v 2 )y 2 − 54y 6 + 3K 2 (1 + y 4 − 2v 2 y 4 ) cos2 ϑ + 9iKv 3 y 3 (K 2 − 18y 2 ) cos3 ϑ + 9K 2 v 4 y 2 (K 2 + 6y 2 ) cos4 ϑ i − 6ivy 3 K(K 2 − 2y 2 ) sec ϑ .
(72)
22
S. S. Gubser et al.
By omitting the “even” and “odd” superscripts in (68a) and (68b) we mean that these equations take the same form (up to the different source terms as indicated in (69)–(71)). 5.2. Boundary conditions In axial gauge, the system of equations (47) consists of ten second order differential equations in y and five first order constraints, so we need to specify fifteen integration constants. The purpose of this section is to show that five of these are fixed by imposing boundary conditions at the horizon (y = 1) and ten of them are fixed from the boundary conditions at the conformal boundary (y = 0). Consider first the tensor field A, defined in (51a). The near horizon solution to (52a) takes the form ivK1
e− 8 (π−ln 4) (1 − y)1−ivK1 /4 +UA (1−y)−ivK1 /4 +VA (1−y)ivK1 /4 +. . . , (73) A= 1 4 1 − ivK 2
where . . . indicates terms which are subleading to one of the ones shown. The first term in (73) is a particular solution to (52a) characterizing the response of A to the trailing string source. The second and third terms are solutions to the homogeneous equation, and UA and VA are constants of integration. The correct boundary condition at the horizon is VA = 0. This corresponds to requiring that there are no outgoing modes at the horizon, and it can be justified by the fact that classical horizons don’t radiate. The result of choosing purely infalling conditions is that in the dual gauge theory, we describe a causal response of the medium to the probe. To see that the UA term in (73) is infalling, let’s define a new coordinate y∗ = log(1 − y) that ranges from −∞ when y = 1 (the horizon) to zero when y = 0 (the conformal boundary of AdS5 ). Recalling that the time dependence of metric perturbations was assumed to be e−iK1 vt/zH , one sees immediately that the (1 − y)−ivK1 /4 term in (73) corresponds to certain metric components behaving as hmn ∼ e−ivK1 (y∗ +4t/zH )/4
(74)
at large negative y∗ . This behavior describes a wave traveling towards negative y∗ , i.e., falling into the black hole horizon. Similarly, the VA term in (73) corresponds to an outgoing mode. The same story goes through for the C combination of metric components defined in (51c), except that since there is no source term, there will be no analog to the first term in (74). A subtlety arises in the horizon boundary conditions for the B, D, and E sets: for each set, in addition to a single infalling solution and a single outgoing solution, there are the pure gauge solutions discussed around (58) and (59). The pure gauge solutions are neither infalling nor outgoing at the horizon. The correct boundary conditions are to exclude the outgoing solution and to permit both the infalling solution and the pure gauge solutions. If one passes to a description only in terms
Energy Loss and the GaugeString Duality
23
of gaugeinvariants, then this subtlety is avoided: each gauge invariant field has only an infalling and outgoing solution, and the latter is excluded by the horizon boundary conditions. Having fixed five integration constants (one for each of the ABCDE sets) using infalling boundary conditions at the horizon, we now discuss the boundary conditions at the boundary of AdS. Close to y = 0, Einstein’s equations can be solved in a series expansion in y. The two homogeneous solutions are (1) Hmn (y) = Rmn 1 + O(y 2 )
(2) Hmn (y) = Qmn y 4 + O(y 6 )
(75)
where Rmn and Qmn are arbitrary constants. The full solution then takes the form (1) (2) Hmn = Hmn (y) + Hmn (y) +
Pmn 3 y + O(y 5 ) . 3
(76)
The components of Pmn are given by 2
+ v2 ) −2v 0 0
3 (2
Pmn =
−2v 2 (1 + 2v 2 ) 3 0 0
0 0 2 2 3 (1 − v ) 0
0 0 . 0 2 2 3 (1 − v )
(77)
The boundary conditions we impose are that Rmn = 0,
(78)
thus fixing the remaining ten integration constants. Allowing nonzero Rmn would correspond to deformations of the gauge theory lagrangian. With this condition imposed, the Qmn are related to the expectation value of the stress tensor in the gauge theory, as we will see in Sec. 5.3. Equation (78) can be translated into boundary conditions for the ABCDE variables. Once the horizon boundary conditions are also taken into account, it follows that each set of equations in (52) is supplemented by just enough boundary conditions to uniquely fix a solution. For example, (78) implies boundary conditions on A which fix one constant of integration. Another integration constant is fixed by the horizon boundary conditions. Since the underlying equation (51a) is second order and linear, the solution is unique. A more complicated example is the E set, where the boundary conditions from (78) fix four integration constants and the horizon boundary conditions fix one more. Since the four second order equations of motion (52g) are subject to three constraints (52h), the number of integration constants available is five. Thus one again finds a unique solution. If we pass to a description in terms of gauge invariants, then (78) fixes a single constant of integration for each gaugeinvariant field. Because the horizon fixes another constant and the underlying equation for the gaugeinvariant is always second order, we again recover a unique solution.
24
S. S. Gubser et al.
5.3. The boundary stressenergy tensor The AdS/CFT duality offers a prescription2, 3 for computing the stress tensor of the boundary theory from the bulk action58, 59 : 2 δStotal hTmn i = lim √ . →0 −g δg mn
(79)
We now explain what Stotal , gmn and are. gmn is the metric on the conformal boundary of AdS5 . After taking the variational derivative, gmn is set equal to the metric of the boundary theory, i.e., the Minkowski metric ηmn in our case. Stotal is given by Stotal = Sbulk + SGH + Scounter .
(80)
Here Sbulk is the bulk action (1); SGH is the GibbonsHawking boundary term Z √ 1 SGH = 2 d4 x GΣ K Σ , (81) κ where Σ is a codimension one surface close to the AdS boundary with outward normal nµ , induced metric GΣ µν ≡ Gµν − nµ nν ,
(82)
Σ ρ Kµν ≡ −GΣ µρ ∇ nν ;
(83)
and extrinsic curvature tensor
and Scounter is an additional boundary term which renders the onshell action finite for geometries which do not induce a trace anomaly in the boundary theory. An explicit expression for this term is Z p 2 L2 Σ 1 4 Σ Scounter = 2 d x −G − R , (84) κ L 4
with RΣ the Ricci scalar constructed from GΣ mn . Usually, there are additional terms in (84) coming from the matter action. Since we are working in the probe limit, we do not need to worry about these extra terms.60 The coordinate in (79) specifies a hypersurface Σ() which coincides with the boundary of AdS5 as we take the → 0 limit. Since we are working with a flat boundary metric, the variation of Scounter will not contribute to hTmn i, and we can choose Σ to be a surface of constant z. The outward normal form is L nµ dxµ = − √ dz , z h
(85)
√ z h ∂Gmn = . 2L ∂z
(86)
and then Σ Kmn
Energy Loss and the GaugeString Duality
25
Equation (79) now reads hTmn i = lim
z→0
L2 Σ Kmn − K Σ GΣ mn . 2 2 z κ
(87)
Using (87), the expectation value of the stressenergy tensor in the absence of the quark is that of a thermal bath hTmn ibath =
π2 2 4 N T diag{3, 1, 1, 1} . 8
(88)
The presence of the quark generates two additional contributions: writing Z 3 1 2 3 d K K K hTmn i = hTmn ibath + hTmn idiv + hTmn i ei[K1 (x −vt)+K2 x +K3 x ]/zH , (2π)3 (89) and using the definitions (49) and (76), as well as the boundary condition Rmn = 0 and the AdS/CFT identities (2) and (15), one obtains √ 1 π2 T 3 λ K √ (90) hTmn idiv = Pmn − ηmn Pl l 4 1 − v 2 √ π3 T 4 λ K (91) hTmn i = √ Qmn − ηmn Qll . 2 1−v
Let’s first understand the divergent contribution. Plugging (77) into (90), it is not hard to see that in position space hTmn idiv takes the form of a contact term √ p λ (92) hTmn idiv = um un 1 − v 2 δ(x1 − vt)δ(x2 )δ(x3 ) 2π
1 where um = √1−v (1, ~v ) is the fourvelocity of the quark. A divergence of this form 2 was to be expected, and it is associated to having an infinitely massive quark. In the dual gravity language, the mass of the quark can be identified with the energy of the trailing string. But this energy is both IR and UV divergent. The UV divergence corresponds to the “bare mass” of the quark and is exactly given by √ λ M= , (93) 2π
where is the UV cutoff. Equation (92) then takes the form of the stressenergy tensor of a particle with mass M moving with velocity v along the x1 direction. It can be shown that any string configuration whose endpoint lies on the boundary of AdS, whether it is stationary, moving with constant velocity relative to the plasma, or accelerating, will generate a divergent contribution of the form (92).60 The finite contribution to the stressenergy tensor given in (91) can be further simplified by using the 55 Einstein equation. Using the series expansion (76) with Rmn = 0, the 55 Einstein equation imposes a tracelessness relation on the Qmn : −Q00 + Q11 + Q22 + Q33 = 0 .
(94)
26
S. S. Gubser et al.
This equation implies further that K hTmn i
√ π3 T 4 λ = √ Qmn . 1 − v2
(95)
K Note that even from (91) one can see that hTmn i is traceless, so (94) does not imply tracelessness of the boundary stressenergy tensor. What (94) does is that it allows us to write the onepoint function of the stress tensor in the simplified form (95). The 5m Einstein equations imply four more relations among the Qmn : iv K m Qmn = (96) v −1 0 0 . 2 Using (95), (96) shows that the boundary stressenergy tensor fails to be conserved: √ iv π 3 T 4 λ √ K m hTmn i = (97) v −1 0 0 . 2 1 − v2
This nonconservation comes from the fact that the quark is prescribed to move at constant velocity. The drag force can be interpreted as minus the force exerted by the quark on the medium. With this interpretation in mind, we can check explicitly that the drag force (27) can be recovered from (97), as follows. Given the stressenergy tensor, the external force acting on a region V can be computed from Z I Z d in n 3 0n 2 F = d x hT i + d a ni hT i = d3 x ∂m hT mn i . (98) dt V ∂V V
The region V is assumed not to depend on time in the asymptotic rest frame of the plasma. The first term in (98) gives the rate of change of energymomentum in this region, while the second term corresponds to the energymomentum flux through the boundary of V . To obtain the last equality we used the divergence theorem. Using (89) and taking the limit where the volume of V goes to infinity, as appropriate for computing the total force on the system, one can see that √ v πT 2 λ mn n 2 F = izH lim Km hTK i = √ . (99) v 1 0 0 ~ 2 1 − v2 K→0
F 1 is then indeed minus the drag force, as can be easily checked by comparing (99) to (27). The prescription (95) gives the stress tensor in terms of the coefficients Qmn that appear in the near boundary asymptotics of the components of Hmn . We can give similar prescriptions in terms of the near boundary asymptotics of the other linear combinations of metric perturbations that we discussed in Sec. 5.1 — all that we need to do is to relate the Qmn to a near boundary expansion of these variables. For the ABCDE variables defined in (51), the near boundary behavior is
PX 3 y + QX y 4 + O(y 5 ) where X = A, Bi , C, Di , Ei , (100) 3 Here, as before, the QX and RX are arbitrary constants while the PX are set by (52). The boundary condition Rmn = 0 is equivalent to RX = 0. X = RX +
Energy Loss and the GaugeString Duality
27
It is easy to invert the definition (51) and use the near boundary expansions (100) and (76) to write the Qmn components in terms of the QX . However, it is useful to notice that we need not specify all of the QX . The reason for this is that the constraint equations (52c), (52f), and (52h) imply relations among the QX , namely QD1 − QD2 = −
QB1 − QB2 = 0 QE1 − 2QE2 = −
iv 2K1
i 4vK1
QE1 + 2QE3 = 0
(1 − 3v 2 cos2 ϑ)QE1 + 2QE4 =
3iv(1 + v 2 ) cos ϑ . 2K
(101)
Using (101), we can write Qmn in terms of only five of the QX , one from each set. This will be nothing more than a parameterization of the most general Qmn obeying constraints (94) and (96). Recalling that the response to the trailing string has B1 = B2 = C = 0 everywhere because of symmetry, we conclude that we can write Qmn in terms of QA , QD ≡ QD1 and QE ≡ QE1 g : Qmn = amn QA + dmn QD + emn QE + pmn ,
(102)
where
amn
0 0 0 2 v 2 sin2 ϑ 0 −2 sin ϑ sin 2ϑ = 0 sin 2ϑ −2 cos2 ϑ 2 0 0 0
0 4 sin2 ϑ −2 sin 2ϑ 2 2 v 4 sin ϑ −2 sin 2ϑ sin 4ϑ = sin 4ϑ 2 sin2 2ϑ 2 −2 sin 2ϑ 0 0 0 2
dmn
emn
(103)
2
0 0 0 0
−4 4v cos2 ϑ 2v sin 2ϑ 0 1 4v cos2 ϑ 4e11 4e12 0 = 4e12 4e22 0 4 2v sin 2ϑ 2 2 0 0 0 2v cos ϑ − 2
(104)
1 −1 + (1 + v 2 ) cos2 ϑ − 3v 2 cos4 ϑ 2 1 = sin 2ϑ 1 − 3v 2 cos2 ϑ 4 1 = cos2 ϑ −1 − 2v 2 + 3v 2 cos2 ϑ 2
e11 = e12 e22 g In
0 0 0
(105)
later sections, we will continue to use QD to mean QD1 , and likewise QE = QE1 . If the gaugeinvariants D and E defined in (61) and (62) were expanded in powers of y, like in (100), the coefficients of y 4 would be related to QD1 and QE1 , but not identically equal.
28
S. S. Gubser et al.
pmn
0 4v 4v tan ϑ 0 iv cos ϑ ℘11 ℘12 0 4v = 4v tan ϑ ℘12 ℘22 0 8K 0 0 0 2 + 2v 2
(106)
℘11 = (1 − 3v 2 ) cos 2ϑ − 5 − v 2
℘12 = (1 − 3v 2 ) sin 2ϑ − 4 tan ϑ
℘22 = (3v 2 − 1) cos 2ϑ + 3 − v 2 .
For the master fields (67), the near boundary behavior is PT QT ψTeven = 3 y 3 + 8 y 4 + O(y 5 ) L L ψVeven =
PV 2 QV 3 y + 6 y + O(y 4 ) L4 L
(107a) (107b)
QS PS y + 4 y 2 + O(y 3 ) . (107c) 2 L L Here we write asymptotics after having imposed the condition Rmn = 0. Similarly to before, the Q coefficients are arbitrary while the P are set by (68). The odd master fields ψTodd and ψVodd have similar expansions, but since their equations of motion are homogeneous, we can set them to be identically zero and worry about them no longer. Using the definitions (67) and (101) it is easy to relate the Q coefficients for the master fields to the QX . The result is QT QA = − (108a) 3 sin2 ϑ 2 2 αv v L zH ψS =
i QV + (108b) 2 sin ϑ 4vK1 4αv vL2 zH iv QS [5pt]QE = − − . (108c) 2K1 2αv L2 zH Similar formulas relating the QX to the asymptotics of the other gaugeinvariants discussed can be easily derived, but we will not need them. QD = −
6. Asymptotics Equations (52) or (68) are difficult to solve exactly and we eventually resort to numerics to obtain a full solution. However, there are various approximations which can be used in order to get a handle on the large and small momentum asymptotics of the metric fluctuations. These may be Fourier transformed to real space, giving us approximations to the near and farfield behavior of the boundary theory stressenergy tensor. In Sec. 6.1 we focus on the small momentum asymptotics of the solution: in Sec. 6.1.1 we construct a small momentum series expansion of the metric perturbations, while in Sec. 6.1.2 we explain how this maps into hydrodynamic behavior on the boundary theory. In Sec. 6.2 we focus on the nearfield of the stressenergy tensor.
Energy Loss and the GaugeString Duality
29
6.1. Long distance asymptotics 6.1.1. Momentum space analysis The small momentum asymptotics of the solution can be obtained by formally expanding the appropriate fields in power series in the momentum K. In what follows we will go over such an expansion in the A, B, C, D, E variables of (51).46 A similar construction can be carried out for the other parameterizations of the Einstein equations47, 51 given in (66) and (67). We start by formally expanding the A variable in (51a) such that A=
∞ X
αn K n ,
(109)
n=0
~ which are invariant under rescalings K ~ → λK. ~ where the αn are functions of K ~ ~ That is, the αn only depend on the direction of K, not its magnitude K = K. Plugging the expansion (109) into (52a) and collecting terms with identical powers of K, we obtain a set of equations for αn of the form y3 h ∂y ∂y αn = Sn , h y3 where Sn can depend on αm with m < n. For example, 2 y y cos ϑξ y cos ϑ v 2 cos2 ϑ − h S0 = S1 = −i S2 = − − α0 . h h hzH h zH h2 The most general solution to (110) is Z y Z y˜ h(y˜˜) y˜3 Sn (y˜˜) . dy˜˜ αn = d˜ y h(˜ y ) y1 y˜˜3 y0
(110)
(111)
(112)
In order to satisfy the boundary conditions (78) at the asymptotically AdS boundary, we require that y0 = 0. The other integration constant in (112) is obtained by matching the near horizon behavior of (112) to the series approximation (73) with VA = 0, expanded at small K. For example, at order n = 0, we find that 1 1−y 1 + y2 −1 α0 = 2 tan y + log + y1 log . (113) 4 1+y 1 − y2 Expanding (113) near the horizon and matching it to a small K expansion of (73), A = (1 − y) + UA + O(K) ,
(114)
we find that we need to set y1 = 1. According to (95) and (102), to obtain the stress tensor we will eventually need QA . Expanding (113) for small y and recalling (100), we can read off that QA = 1/4 + O(K). The next order corrections to QA can be obtained in the same manner. We find that to order O(K), QA =
1 i ln 2 − vK1 + O(K 2 ) . 4 8
(115)
30
S. S. Gubser et al.
A similar treatment gives us the asymptotic values of D1 and E1 . The result is QD = − QE =
i sec ϑ sec2 ϑ − 4v 2 − + O(K 2 ) 4vK 16v 2
3iv(1 + v 2 ) cos ϑ 3v 2 cos2 ϑ(2 + v 2 (1 − 3 cos2 ϑ)) − . 2 2 2K(1 − 3v cos ϑ) 2(1 − 3v 2 cos2 ϑ)2
(116) (117)
Note that QD exhibits a pole structure at K = 0, while QE exhibits a pole structure at K 2 = 3K12 v 2 . We will see shortly that these correspond to the diffusion pole and sound pole expected of the hydrodynamic behavior of the plasma far from the moving quark. 6.1.2. Relating the large momentum asymptotics to hydrodynamics At scales much larger than the mean free path, we expect to be able to describe a thermal gauge theory by effective, hydrodynamic, slowly varying degrees of freedom. Let’s first see what such a description entails, and then check how well our asymptotics match it. Consider a conformal theory for which the hydrodynamic energymomentum tensor (T hydro )mn is traceless: (T hydro )m m = 0.
(118)
A static configuration of the fluid will be given by 0 (T hydro )mn = diag {3, 1, 1, 1} . (119) 3 If we now perturb this fluid slightly, and choose our hydrodynamic variables to be = (T hydro )00 and Si = (T hydro )0i with i = 1, 2, or 3, then we should be able to write the remaining spacespace components of the energymomentum tensor, (T hydro )ij , in terms of gradients of and Si . The only possible combination consistent with the tracelessness condition (118) is 1 3 δij − Γik(i Sj) + O(k 2 ) , (120) 3 2 where the free parameter Γ is the sound attenuation length, related to the shear viscosity, temperature, and entropy density through Γ = 4η/3sT . Using (7) we find that Γ = 1/3πT in the theories we are considering. Brackets denote the symmetric traceless combination, i.e., (T hydro )ij =
1 1 (Mmn + Mnm ) − ηmn η pq Mpq . (121) 2 4 The energy density and the energy flux Si can now be computed from the conservation equations, M(mn) =
hydro ikn (T hydro )nm = fm ,
(122)
hydro where fm is the source which perturbs the hydrodynamic stressenergy tensor. Separating the spatial part of the source, f~ hydro , into longitudinal and transverse
Energy Loss and the GaugeString Duality
31
ˆ kˆ and f~ hydro = f~ hydro − f~ hydro , with (k) ˆ i = ki /k, components, f~Lhydro = (f~ hydro · k) T L we find from (120) and (122) that −3i f~ hydro · ~k − f0hydro ω − 3f0 k 2 Γ + O(f k) (123a) = k 2 − 3ω 2 − 3iΓk 2 ω hydro 2 k + 3f~ hydro · ~kω ~ · ~k = −f0 S + O(f k) k 2 − 3ω 2 − 3iΓk 2 ω
(123b)
~ hydro ˆ kˆ = −ifT ~ − (S ~ · k) S + O(f k) . ω + 34 Γik 2
(123c)
In real space, the pole at approximately ω ∼ 0 corresponds to a diffusive wake behind the source, and the pole at roughly k 2 ∼ 3ω 2 corresponds to the shock wave which appears along the Mach cone behind the moving probe. The displacement of the two poles from the real axis due to the shear viscosity of the fluid is responsible for viscous broadening of the wake and Mach cone. We can now compare the small momentum results of Sec. 6.1.1 to the hydrodynamic behavior (123). To do so, we use (115)–(117) in (102) and (95) to obtain the small momentum stressenergy tensor. We find −S1 −S⊥ 0 √ (πT )4 λ K + Tmn + O(K) −S1 (124) i= √ hTmn 2 −S⊥ Tij 1−v 0 with
3K12 v 2 −3K12 v 2 + K 2 (2 + v 2 ) 3iK1 v(1 + v 2 ) =− + 2 2π (K 2 − 3K12 v 2 ) 2π (K 2 − 3K12 v 2 ) −6K14 v 5 − K 4 v + K12 K 2 6v 2 + 1 v iK1 v 2 + 1 + S1 = − i 2 2π (K 2 − 3K12 v 2 ) 2π (K 2 − 3K12 v 2 ) +
2 K⊥ i + 2K1 π 32K12 πv 2
(125b)
K1 K2 v K 2 + 3K12 v 4 iK2 (1 + v 2 ) S⊥ = − + 2 2π (K 2 − 3K1 v 2 ) 2π(K 2 − 3K12 v 2 )2 −
(125a)
K2 8πK1 v
(125c)
and Tij = Tmn =
1 1 δij − iK(i Sj) 3 6
(126)
v2 diag {0, −2, 1, 1} . 6π
(127)
32
S. S. Gubser et al.
Note that we may always carry out a resummation K2
O(K 4 ) αi Ki + O(K 2 ) αi Ki + = 2 2 2 2 2 2 2 − 3K1 v (K − 3K1 v ) K − 3K12 v 2 − iK1 K 2 v
(128)
(where αi are constants) or
1 1 + O(K) O(K 2 ) . + 2 2 = vK1 v K1 vK1 + 41 iK 2
(129)
~ in (125) take the form (123) After such a resummation, the expressions for and S with 1 (130) fnhydro = −v 2 v 0 0 n − iK m Tmn 2π and Γ = 1/3πT . Thus, if we identify the first term in the parenthesis on the right hand side of (124) with the hydrodynamic contribution to the stress tensor of the SYM theory, (T hydro,SYM)mn , then we find that (T hydro,SYM)mn satisfies iK m (T hydro,SYM)mn = fnhydro,SYM .
(131)
This should be compared to the full conservation law (99), K iK m hTmn i = fn .
(132)
The extra term Tmn in (124) holds information on the deviation of the stressenergy tensor from its hydrodynamic form. Alternately, Jn − ik m Tmn gives us an effective hydrodynamic fourforce which sources the hydrodynamic stressenergy tensor. It is no coincidence that the large distance asymptotics of the stressenergy tensor agree with a hydrodynamic expansion. In fact, it can be shown that generic probesources excite the metric in such a way that the resulting large distance asymptotics of the boundary theory stress tensor will have hydrodynamic behavior.60 There is also mounting evidence that such a connection between hydrodynamics and gravity goes beyond the linearized approximation.14 To see that the pole structures in (128) and in (129) really correspond to a laminar wake and a shock wave, we Fourier transform them to real space. Consider the last two terms in (125b), resummed as in (129). Using n−d/2 Z d ~ ~ d K eiK·X 2 X = Kn−d/2 (µX) , (133) (2π)d (K 2 + µ2 )n (4π)d/2 Γ(n) 2µ
we can Fourier transform the resummed expression to position space: Z 3 1 2 3 d K 2v ei(K1 (x −vt)+K2 x +K3 x )/zH 3 2 (2π) π(K − 4iK1v) √ − 2v (x1 −vt)+ (x1 −vt)2 +x2⊥ e zH vzH p = . 2 2 1 2 2π (x − vt) + x⊥
(134)
As expected of a diffusion wake, we find that the configuration (134) exhibits a directional energy flow, with a parabolic shape far behind the moving quark.
Energy Loss and the GaugeString Duality
33
Fourier transforming the resummed sound pole (128) is difficult due to the cubic terms in the denominator, and we eventually resort to numerics to convert such expressions to real space. To see that the pole at K 2 ∼ 3K12 v 2 really corresponds to a shock wave, it is sufficient to Fourier transform only the leading order contributions to these poles. Since we are neglecting the viscous contribution to the pole structure, this corresponds to the inviscid limit. Using contour integration and the identities61 Z ∞ 0 0 Z0 ,
(145)
and then tn (Z) =
Z
dZ 0 G(Z, Z 0 )(JT )n (Z 0 ) .
(146)
Energy Loss and the GaugeString Duality
For n = 0 and n = 1 these integrals may be carried out exactly.61 We find 3 2 zH v αv 2α 3 2 2 2 sin ϑ Z L2 (αZ) − Z I2 (αZ) + Z t0 (Z) = − L6 3π t1 (Z) = 0
z 3 v 2 αv − H 6 L
(147)
4
iK1 vZ , 3Kα2 with L2 a modified Struve function. Recalling (107a) we can read off q ivK1 1 3 2 QT = −zH v αv sin2 ϑL2 π K 2 − K12 v 2 − 16 3(K 2 − K12 v 2 ) t2 (Z) =
35
sin2 ϑ −
(148)
from (147). The equation of motion for t0 coincides with the one that would have been obtained starting from a string hanging straight down from the boundary of AdS space, boosted to a velocity v in the x1 direction. The solution t0 then corresponds to the tensor mode metric perturbation in response to this string and, as we will see shortly, it captures the nearfield physics of the stressenergy tensor in response to a massive quark. At scales much smaller than the mean free path, one may effectively ignore the interaction of the quark with the plasma. The function t2 corresponds to the first thermal corrections to the nearfield of the quark. The computation of the large K asymptotics for the vector and scalar modes follows in a similar manner. Let ψVeven = K −4
0 X
vn K −n ,
(149)
n=−∞
similar to (138). Expanding (68b) at large K, we find that the vn ’s satisfy 3 ∂Z vn + (3 − Z 2 α2 )vn = −(JV )n . Z The first few terms in −(JV )n are given by ∂Z2 vn −
αv vK⊥ Z 4 K5 −(JV )1 = 0 −(JV )0 =
−(JV )2 = −i
(150)
(151) (152) 2
4K12 v 2 )
αv K⊥ (3K − 3K 2 K1
.
(153)
The homogeneous solutions to (150) are v (1) = Z 2 I1 (αZ)
v (2) = Z 2 K1 (αZ) ,
(154)
and using the Green’s function method we find ψVeven =
αv K⊥ πv 2 Z (L1 (αZ) − I1 (αZ)) 2Kα2 +
iαv K⊥ (3K 2 − 4K12 v 2 )Z 3 (8 + Z 2 α2 ) 3 Z , 3K1 K 2 α4
(155)
S. S. Gubser et al.
36
which, recalling (107b), implies " # p π k 2 − v 2 k12 L2 αv vk⊥ 3k 2 − 4k12 v 2 L4 −4 QV = − − + 2 + O(zH ) . (156) k 4 3ivkk1 (k 2 − k12 v 2 ) zH The details of the computation of the scalar modes can be found elsewhere.47, 49, 50 The final result is " π 2 + v 2 (k 2 − k12 v 2 ) − v 2 (1 − v 2 )k12 p QS = αv − 12 k 2 − k12 v 2 # iv 9 k1 (5 − 11v 2 ) 2v 2 (1 − v 2 )k13 −4 − − + O(zH ) (157) − 2 9zH k1 k 2 − v 2 k12 (k 2 − k12 v 2 )2 where ψS = PS L−2 y + QS L−4 y 2 + O(y 5 ) .
(158)
With QT , QV , and QS at hand we can use (108), (102), and (95) to obtain the leading large momentum asymptotics of the stressenergy tensor. The momentum space expressions for the energy density and Poynting vector are p √ " π3 T 4 λ (2 + v 2 ) K 2 − K12 v 2 K12 v 2 (1 − v 2 ) K − + p hT00 i = − √ 24 1 − v2 24 K 2 − K12 v 2
iK1 v(11v 2 − 5) 14 + 7v 2 iK13 v 3 (1 − v 2 ) + + 18π(K 2 − K12 v 2 ) 24 (K 2 − K 2 v 2 )3/2 9π(K 2 − K12 v 2 )2 1 # K12 v 2 10v 2 − 1 K14 v 4 (1 − v 2 ) + − (159) 24(K 2 − K12 v 2 )5/2 8(K 2 − K12 v 2 )7/2 p √ " v K 2 − K12 v 2 π3 T 4 λ K 2 v(1 − v 2 ) K hT01 i = − √ + p1 − 8 1 − v2 24 K 2 − K12 v 2 −
iK1 v 2 9v iK13 v 2 (1 − v 2 ) + + 2 2 3π(K 2 − K1 v 2 ) 16(K 2 − K1 v 2 )3/2 9(K 2 − K12 v 2 )2 π # K12 v(1 + 17v 2 ) K14 v 3 (1 − v 2 ) + − 48(K 2 − K12 v 2 )5/2 8(K 2 − K12 v 2 )7/2 −
(160)
and
√ " 3 4 π T λ K1 K2 v(1 − v 2 ) iK2 v 2 K p hT02 i = −√ − 2π(K 2 − K12 v 2 ) 1 − v 2 24 K 2 − K12 v 2
# iK12 K2 v 2 (1 − v 2 ) K1 K2 v(1 + 14v 2 ) K13 K2 v 3 (1 − v 2 ) + + − . 9π(K 2 − K12 v 2 )2 48(K 2 − K12 v 2 )5/2 8(K 2 − K12 v 2 )7/2
(161)
Energy Loss and the GaugeString Duality
37
K The expression for hT03 i can be obtained from (161) by exchanging K2 with K3 . In real space, using the notation in (91), we find that to leading order in xT , K hTmn i = Λ−1 T quark Λ
(162)
mn
where Λmn represents a Lorentz transformation with boost parameter v in the x1 quark direction and Tmn is given by
quark Tmn
1 x4 √ 0 λ = 12π 2 0 0
0 x2⊥
− x6
0 x21
2x1 x2 − 6 x −
2x1 x3 x6
−
2x1 x2 x6
x21 + x23 − x22 x6 −
2x2 x3 x6
0
2x1 x3 − 6 x . 2x2 x3 − 6 x 2 2 2 x1 + x2 − x3 x6
(163)
quark Tmn is the stressenergy tensor of a stationary heavy quark. Up to the overall multiplicative factor, it can be determined by the requirement that it is conserved and satisfies conformal symmetry. Thus, the leading short distance behavior of the near field of our quark is a boosted version of the stressenergy tensor of a stationary quark. At distances much shorter than the typical length scale of the fluid, the quark does not see the plasma it is moving through, and behaves as if it were in vacuum. Of more interest are the subleading corrections to the stressenergy tensor. These are given by
√ 2 λT v(x − vt) x2⊥ (−5 + 13v 2 − 8v 4 ) + (−5 + 11v 2 )(x − vt)2 hTtt i = √ 5/2 1 − v2 72 [x2⊥ (1 − v 2 ) + (x − vt)2 ] (164a) √ 2 2 2 2 2 λT v (x − vt) (1 − v )x⊥ + 2(x − vt) hTtx1 i = − √ (164b) 24[x2⊥ (1 − v 2 ) + (x − vt)2 ]5/2 1 − v2 √ 2 λT x⊥ (1 − v 2 )v 2 8x2⊥ (1 − v 2 ) + 11(x − vt)2 hTtx⊥ i = − √ (164c) 5/2 1 − v2 72 [x2⊥ (1 − v 2 ) + (x − vt)2 ] √ 2 λT v(x − vt) x2⊥ (8 − 13v 2 + 5v 4 ) + (11 − 5v 2 )(x − vt)2 hTx1 x1 i = √ (164d) 5/2 1 − v2 72 [x2⊥ (1 − v 2 ) + (x − vt)2 ] √ 2 λT v(1 − v 2 ) 8x2⊥ (1 − v 2 ) + 11(x − vt)2 hTx1 x⊥ i = √ (164e) 5/2 1 − v2 72 [x2⊥ (1 − v 2 ) + (x − vt)2 ]
38
S. S. Gubser et al.
√ 2 λT v(1 − v 2 )(x − vt) 5x2⊥ (1 − v 2 ) + 8x2⊥ hTx⊥ x⊥ i = − √ 5/2 1 − v2 72 [x2⊥ (1 − v 2 ) + (x − vt)2 ] √ 2 λT v(1 − v 2 )(x − vt)x2⊥ hTϕϕ i = − √ 2 1 − v 9 [r2 (1 − v 2 ) + (x − vt)2 ]5/2
(164f)
(164g)
where (x⊥ , ϑ) are polar coordinates for the x2 x3 plane. A strange feature of (164a) is that it exhibits a transition from a region of energy depletion behind the quark, to a region of energy depletion in front of it as the quarks velocity decreases. When v 2 > 5/8 there is a buildup of energy density ahead of it, forming a “bulldozer effect.” See Fig. 3. As it slows down extra lobelike features appear until v 2 < 5/13 where the energy buildup is behind the quark, creating an “inversebulldozer” effect. See Fig. 4. Recall that the speed of sound in a conformal fluid is v 2 = 1/3, so that this transition occurs at velocities which are higher than the speed of sound. This indicates that the features we are seeing are not hydrodynamic in nature. A more detailed analysis of the deviation of the energy density from linearized hydrodynamics can be found in the literature.63 We will see in Sec. 8 that it is probably the near field of the stressenergy tensor which dominates highangle emission of hadrons. It would certainly be interesting to understand the physical mechanism behind this nearfield behavior. 7. Numerical Results for the Holographic Stress Tensor Expression (164) and the Fourier transform of (124) capture the nearfield and farfield asymptotics of the stressenergy tensor. In 6.1.2, we have seen an indication that far from the moving source the energymomentum tensor exhibits hydrodynamic behavior. In 6.2, we have seen that the nearfield stress tensor exhibits nonhydrodynamic behavior with interesting features, like the multilobe structure in Figure 4. In the intermediate regime, there is a transition region between hydrodynamics and whatever shortdistance physics governs the near field. To probe this region, one needs solutions to (44) for values of K where no analytic asymptotic treatment is available. We have obtained such solutions numerically. First, (52a), (52b), (52d), (52e), and (52g) were solved, and QA , QD , and QE were obtained.46 Then, the resulting momentum space stressenergy tensor was passed through an FFT to position space using a 1283 grid. Such a computation has been carried out for the energy density50, 51 and for the Poynting vector.52, 53 Consider the normalized energy density √ 1 − v2 K √ hT00 E= i, (165) (πT )4 λ K where hTmn i is defined in (89) as the stressenergy tensor of the system, minus the stressenergy of the thermal bath, minus the divergent deltafunction contribution (92) at the position of the moving quark. It is convenient to decompose this rescaled
Energy Loss and the GaugeString Duality
39
energy density into a Coulombic term, a nearfield (large momentum) term, a farfield (small momentum) term, and a residual term: E = ECoulomb + EUV + EIR + Eres .
(166)
The Coulombic term represents the contribution coming from the near field of the quark: p K 2 − K12 v 2 v 2 K12 2 + (1 − v 2 ) , (167) ECoulomb = − 2 + v 24π 24π which is what we found in (162) converted to momentum space. It can be read off of the O(K) terms in (159). The farfield term, EIR , scales like O(K −5 ) at large momentum and asymptotes to (125a) at small momenta.h There are many possible expressions which satisfy the above criteria. Taking note of the resummation (168), we used EIR = −
1 3ivK1 (1 + v 2 ) − 3v 2 K12 1 3ivK1 (1 + v 2 ) − 3v 2 K12 + 2 2π K 2 − 3v 2 K1 − ivK 2 K1 2π K 2 − 3v 2 K12 − ivK 2 K1 + µ2IR
(168)
where µIR is a typical scale where (125a) stops being valid. We used µIR = 1. Similarly, EUV scales like O(K 1 ) at small momenta, and asymptotes to (159) at large momenta. To regulate the large momentum expressions in the IR we made the replacement −n/2 1 1 µ2UV = 1− 2 (169) K − K12 v 2 + µ2UV (K 2 − v 2 K12 )n/2 (K 2 − K12 v 2 + µ2UV )n/2
where the last term in (169) is expanded to order O(K −5 ). Similar to (168), µUV is a cutoff scale which we set to 1. After the replacement (169), the first two terms in a series expansion of the energy density take the form p (2 + v 2 ) K 2 − K12 v 2 + µ2UV 2K12 v 2 (1 − v 2 ) − (2 + v 2 )µ2UV p + ... . EUV = − + 24 48 K 2 − K12 v 2 µ2UV (170) Once EUV , EIR , and ECoulomb are known we can numerically compute Eres , which can be fed through a threedimensional FFT with controllable errors because it is absolutely integrable. We then add back to the real space numerical expression for Eres the real space version of EIR , EUV , and ECoulomb to obtain the energy density in position space. The Fourier transform of the large momentum asymptotic expressions can be carried out using (133). As explained in Sec. 6.1.2, Fourier transforming the sound pole is difficult due to the cubic term in the denominator — the term associated with the shear viscosity. To convert the the sound pole structure of (168) to real space we first rewrote it as EIR = h Actually,
2 K⊥
A(K1 ) , + m(K1 )2
(171)
we require that it asymptote to (125a) up to a momentumindependent constant.
40
S. S. Gubser et al.
E for v= 0.58 0.1
0.08
Xp
0.06
0.04
0.02
0
0.1
0.05
0
X1
0.05
0.1
Fig. 3. (Color online) A contour plot of the energy density near the moving quark, with the bath and the Coulombic contributions subtracted.50 Red signifies energies above the background value of the plasma while blue signifies energies below the background value of the plasma. The black dot at X1 = Xp = 0 marks the location of the quark which is moving at a constant velocity of v = 0.58, just above the speed of sound. We work in dimensionless units where X1 = πT x1 and Xp = πT x⊥ . The dashed green line shows the presumed location of the Mach cone.
then Fourier transformed in the K⊥ direction using (133), and resorted to numerics to FFT the remaining K1 coordinate. This was carried out on a line with 1944 points with K1 ranging from −20 to 20. In Figures 3–5 we show the energy density (with the Coulombic field ECoulomb subtracted) at various spatial scales. The components of the energy flux can be treated in a similar manner: we define √ 1 − v2 K √ hT0i i (172) Si = − (πT )4 λ and decompose S~ into
S~ = S~Coulomb + S~UV + S~IR + S~res .
(173)
The Coulombic expression for the Poynting vector is given by the O(K) terms in (160) and (161). The small momentum expressions are given by S1 IR = − + S⊥ IR = −
1 i(1 + v 2 )K1 + vK 2 − 2v 3 K12 1 i(1 + v 2 )K1 + vK 2 − 2v 3 K12 + 2π K 2 − 3v 2 K12 − ivK 2 K1 2π K 2 − 3v 2 K12 − ivK 2 K1 + µ2IR 2v 1 + iK1 /4v 2v 1 + iK1 /4v − 2 2 π K − 4ivK1 π K − 4ivK1 + µ2IR
(174)
1 i(1 + v 2 )K⊥ + b2 K1 K⊥ ) 1 iK⊥ + + (regulators) 2π K 2 − 3v 2 K12 − ivK 2 K1 2π K 2 − 4ivK1
(175)
where we have set µIR = 1 and by “(regulators)” we mean terms containing the regulator µIR , analogous to those in (168) and (174). The large momentum expressions are given by applying (169) to (160) and (161). As was the case for the
Energy Loss and the GaugeString Duality
41
E for v= 0.75 0.1
0.08
Xp
0.06
0.04
0.02
0
0.05
0.1
0
0.05
X1
0.1
Fig. 4. (Color 0 4online) A contour plot of the energy density near the moving quark, with the bath and the Coulombic contributions subtracted.50 Red signifies energies above the background value of the plasma while blue signifies energies below the background value of the plasma. The black 2 dot at X1 = Xp = 0 marks the location of the quark which is moving at a constant velocity of v = 0.75, well above the speed of sound. We work in dimensionless units where X1 = πT x1 and Xp = πT x⊥ . The dashed green line shows the presumed location of the Mach cone. 0
E for v = 0.75 14
12
Xp
10 8 6 4 2
0 15
10
5
0
X1
5
10
15
Fig. 5. (Color online) A contour plot of the energy density far from the moving quark, with the bath and the Coulombic contributions subtracted.50 Red signifies energies above the background value of the plasma while blue signifies energies below the background value of the plasma. The black dot at X1 = Xp = 0 marks the location of the quark which is moving at a constant velocity of v = 0.75, well above the speed of sound. We work in dimensionless units where X1 = πT x1 and Xp = πT x⊥ . The dashed green line shows the presumed location of the Mach cone.
42
S. S. Gubser 0. et al.
Xp
S for v =0.75 14 12 10 8 6 4 2 0 15
10
5
0
X1
5
10
15
Fig. 6. (Color online) Contour plot of the magnitude of the Poynting vector, with the Coulombic contribution subtracted.52 The magnitude of the Poynting vector goes from red (large) to white (zero) while the arrows show its direction. The dashed green line shows the presumed location of the Mach angle, and the blue line shows the location of the laminar wake — as dictated by its large distance asymptotics.
energy density, we used µUV = 1. The real space results for the Poynting vector for v 2 = 3/4 are shown in Fig. 6. 8. Hadronization, JetBroadening, and JetSplitting There is a significant gap between the results of Secs. 5–7 and experimental data. Before reviewing recent attempts to bridge this gap, let’s briefly summarize some of the relevant data. There are of course more authoritative summaries in the experimental literature.64–69 The data seem to reveal a phenomenon of “jetsplitting,” whereby an energetic parton traversing the medium deposits so much of its energy through highangle emission that — with appropriate momentum cuts and subtractions — the extra particle production due to the parton is at a minimum in the direction of its motion, and reaches a maximum at an angle roughly 1.2 radians away. Jetsplitting is most simply illustrated through twopoint histograms of the azimuthal angle ∆φ separating a pair of energetic hadrons close to midrapidity. To understand the phenomena better, it is useful to examine two landmark studies of these histograms: one from STAR64 and one from PHENIX.65 In the STAR analysis, fairly inclusive momentum cuts were considered: under one set of cuts, the less energetic of the two hadrons was required to have transverse momentum greater than 150 MeV/c. The resulting data show a peak for nearly collinear hadrons that is approximately the same shape for central goldgold collisions as for protonproton collisions: see Fig. 7. This “nearside jet” feature can reasonably be supposed to arise from vacuum fragmentation effects. The twopoint
A
pa rs per tr gger: 1/N dNAB(d jet)/d(∆φ)
Energy Loss and the GaugeString Duality
0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0
0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0
0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0
0
0.5
(a) 05%
(b) 510%
(c) 1020%
(d) 2040%
(e) 4060%
(f) 6090%
1
1.5 2 ∆φ (rad)
2.5
3
0.5
1
1.5 2 ∆φ (rad)
2.5
43
3
Fig. 7. (Color online) Top: The STAR analysis64 shows substantial broadening of the awayside jet. Reprinted Fig. 1 with permission from J. Adams et al., Phys. Rev. Lett. 95, 152301 (2005), http://link.aps.org/abstract/PRL/v95/p152301. Copyright 2005 by the American Physical Society. Bottom: The PHENIX analysis65 shows jetsplitting for sufficiently central events. Reprinted Fig. 2 with permission from S. S. Adler et al., Phys. Rev. Lett. 97, 052301 (2006), http://link.aps.org/abstract/PRL/v97/p052301. Copyright 2006 by the American Physical Society.
44
S. S. Gubser et al.
histogram also shows an “awayside jet” feature around ∆φ = π which is substantially broader for central goldgold collisions than for protonproton. Neither the “nearside jet” nor the “awayside jet” are reconstructed jets in the usual sense; instead, they are ideas that help explain the main features of histograms assembled from millions of events. It is usually assumed that the typical event contributing to the histograms involves a hard scattering event where one parton escaped the medium without much interaction, producing the highly energetic “trigger hadron,” and the other parton interacted substantially with the medium before generating an “associated hadron” in the vicinity of ∆φ = π. The upshot is that with inclusive momentum cuts, there is substantial broadening of the awayside jet, but not jetsplitting: associated hadron production is still maximized, or statistically indistinguishable from its maximum, at ∆φ = π. With tighter momentum cuts on the associated hadron, the data used in the particular STAR analysis under discussion show striking jetbroadening, but the scatter in the data is sufficient to prevent firm conclusions from being drawn — from this particular study — about whether there is jetsplitting. (Subsequent STAR analyses of both two and threepoint hadron correlators provide strong evidence in favor of jetsplitting.67–69 ) The PHENIX analysis65 is similar to the STAR analysis,64 but with more restrictive cuts: in particular, the less energetic hadron was required to have transverse momentum greater than 2.5 GeV/c. The resulting histograms, with a zeroyieldatminimum (ZYAM) subtraction, show a distinct minimum in associated hadron production at ∆φ = π, with a broad maximum in the ballpark of ∆φ = π − 1.2. This jetsplitting persists down to roughly 50% centrality, meaning that it occurs for events where the impact parameter is less than about 10 fm. The ZYAM subtraction is an important part of the analysis, especially for the less central events. The reason a subtraction is needed is that for a noncentral collision, there is an angular modulation of singleparticle yields, approximately proportional to 1 + v2 cos 2φ, where the zero of φ coincides with the azimuthal direction of the impact parameter, and the elliptic flow coefficient v2 depends on the transverse momentum and species. Twopoint hadron correlators receive a contribution from singlehadron yields. The ZYAM scheme is to subtract a multiple of the appropriate product of singleparticle yields. The multiple is chosen so that the resulting histogram has one bin with zero net events, while all other bins have a positive net number of events. A natural hypothesis is that the highangle emission leading to either jetsplitting or jetbroadening can be described in terms of a sonic boom in the medium.70, 71 Two related difficulties afflict this idea. First, it’s hard to get a sonic boom with a big enough amplitude to account for the data65 with reasonable rates of energy loss;45, 72 it should be noted however that not all investigators agree on this point,73 and that there are some phenomenological models based on sonic booms that fit the data.74, 75 Second, one usually finds a diffusion wake with comparable strength to the sonic boom.30, 52, 71, 76 At least in a static medium, it is hard to get jetsplitting in the presence of a significant diffusion wake.
Energy Loss and the GaugeString Duality
45
To compute the relative strength of the diffusion wake and sonic boom for the heavy quark in the SYM theory, we go back to the conservation equation (99), 2 mn F n = izH lim Km hTK i.
(176)
~ K→0
In Sec. 6.1.2 we saw that at small K, the components of the stressenergy tensor may be decomposed into terms containing sound poles at K 2 ∼ 3K12 v 2 and terms mn associated with a wake which have a pole at K1 ∼ 0. If hTIR i is the leading, small K contribution to the stressenergy tensor, then we may decompose mn mn mn hTIR i = hTsound i + hTwake i,
(177)
where 3iK1 v iK1 iK2 iK3 √ iK1 iK1 v (πT )4 λ 1 + v2 0 0 mn i=−√ hTsound 2 2 2 2 iK2 0 iK1 v 0 1 − v 2π(K − 3K1 v ) iK3 0 0 iK1 v
(178)
and
0 √ 4 (πT ) λ i mn 1 i= √ hTwake 2 1 − v 2πK1 0 0
1 0 0 0
0 0 0 0
0 0 . 0
(179)
0
Since only the terms in (177) contribute to the total drag force in (176), this gives a natural division of the total drag force: n 2 mn Fsound = izH lim Km hTsound i ~ K→0
n 2 mn Fwake = izH lim Km hTwake i, ~ K→0
From (124), (125), (126), and (127), we find that 1 1 0 0 Fsound = − 2F0 Fwake = 1 + 2 F0 . v v
(180)
(181)
In our conventions, the zero component of F n gives us the total rate of change in the energy density plus any energy flux going out of the system. Thus, the ratio of energy going into sound waves to energy going into the wake is 1 + v 2 : −1 .
(182)
While sound modes carry energy away from the moving quark, the wake feeds energy in toward the quark. While this may seem counterintuitive, in some sense it’s obvious: the diffusion wake consists of a flow of the medium forward toward the quark. The forwardmoving momentum in the diffusion wake is the momentum deposited by the quark at earlier times. Qualitatively, (182) says that the diffusion wake and the sonic boom have comparable strength. When comparing (182) to the scenarios of energy loss in the literature,71, 72 one finds that (182) quantitatively matches a scenario where the relative strength of the wake is so large that it washes out features of jetsplitting associated with the sonic boom. It may be significant,
46
S. S. Gubser et al.
associated hadron
φ
x2
x1
ϕ
x3
θ beamline
trigger hadron Fig. 8. (Color online) A sketch of the coordinate systems used to describe awayside hadron production. If the trigger hadron is at φ = 0, then a helpful relation at midrapidity (θ = π/2) is ϑ = π − φ, where ∆φ is the azimuthal separation between the trigger and associated hadrons.
however, that the medium is infinite and static, both in our work and in the linear hydrodynamic scenario71 that our results match onto at large length scales. In light of the difficulties in explaining the data with a “boom and wake” model, focused on the hydrodynamic regime, it is natural to investigate the effect on hadron production of the region of the medium close to the moving quark where hydrodynamics is inapplicable. This has been pursued in a series of works,63, 77–83 which we briefly summarize in the next few paragraphs. The first idea is to subtract away the leading order Coulombic contribution to hTmn i. Up to an overall multiplicative rescaling, these are the quantities we denoted ECoulomb and S~Coulomb in Sec. 7. The justification for this is that these fields describe the energy of the energetic parton itself, not the energy lost from it. The remaining energy density, which we will denote as sub , can be split up as sub = bath + ∆. (Note that in contrast to our definitions of E and its variants, sub explicitly includes the contribution from the ~sub , with the Coulombic contribution subtracted away, bath.) The Poynting vector S is nonzero only because of the presence of the quark. The basic plan is to use the ~sub ) into a CooperFrye algorithm84 to convert string theory predictions for (sub , S spectrum of hadrons. The CooperFrye algorithm is based on converting a fluid element at temperature T and with local fourvelocity U m into hadrons according a Maxwell
Energy Loss and the GaugeString Duality
Boltzmann distribution in the local rest frame: Z µ dN =− dΣµ P µ eU Pµ /T f (pT , φ) = pT dpT dφ y=0 R3 Z ∞ Z 2π Z ∞ µ = x⊥ dx⊥ dϕ dx1 EeU Pµ /T , 0
0
47
(183)
−∞
where N is the number of hadrons, and we have set P m = pT
pT cos(π − φ) pT sin(π − φ) 0
and
(184)
U m = U 0 U 1 U⊥ cos ϕ U⊥ sin ϕ .
(185)
tanh y = cos θ .
(186)
Note that because of our choice of mostly plus signature, the energy of the hadron in the local rest frame of the fluid is −U m Pm . Also because of this choice of signature, we are obliged to include an explicit minus sign in the first integral expression of (183). To understand (183)–(185), it helps to refer to Fig. 8. The momentum of the associated hadron is P m , and (183) is written in the approximation that the associated hadron is massless — an excellent approximation since a typical hadron of interest is a pion with pT ∼ 3 GeV/c. The rapidity y is related to the angle from the beamline θ by
(Note that rapidity y has nothing to do with the depth coordinate y = z/zH used in previous sections.) The freezeout surface is chosen to be a slice of constant x0 in (183). This is the best motivated choice for an infinite, asymptotically static medium. In an expanding medium, a more usual choice is a fixedtemperature surface with the temperature set close to the QCD scale. For isochronous freezeout, the measure dΣm is simply dx1 dx2 dx3 1 0 0 0 , and in passing to the second line of (183) we have simply expressed the metric on R3 in radial coordinates. It is important to realize that the azimuthal angle ϕ around the direction of motion of the parton (assumed to be in the +x1 direction, as usual) is different from the azimuthal angle φ around the beamline. We have omitted in (183) a subtraction of the contribution of the bath to hadron production, which depends on pT but not on φ. We have also not attempted to normalize f (pT , φ): doing so would involve partitioning over the spectrum of hadrons. The information from string theory enters into (183) in two ways. First, the local fourvelocity U m is the local rest frame of the medium, in which the Poynting vector vanishes. Second, the temperature T is the temperature in this local rest sub m n frame, deduced by plugging the energy density Tmn U U into the equation of mn state. Evidently, one needs all components of Tsub in order to precisely determine U m and T . This is a problem since only the m = 0 row of the stress tensor has been computed in full.52, 53 Another problem is that close to the quark, the medium is
48
S. S. Gubser et al.
presumably far from equilibrium, so using CooperFrye seems somewhat perilous. We will return to a discussion of these two issues below. To understand how hydrodynamical and nonhydrodynamical effects contribute to the spectrum of produced hadrons, one must have some notion of where the boundary is between hydrodynamical and nonhydrodynamical regimes. This boundary is presumably not sharp. Three considerations have gone into identifying an appropriate boundary: (1) The nonhydrodynamical region can be chosen as the region where ∆/bath is less than some constant of order unity. For v = 0.9 ,
λ = 5.5 ,
N = 3,
TSYM = 200 MeV ,
(187)
a preferred choice is ∆/bath ≤ 0.3 .
(188)
Here and below, we will describe the region defined by (188) with the parameter choices (187) as the “Neck.” It extends roughly over −1 ≤ X1 ≤ 0.5 and 0 ≤ X⊥ ≤ 1.7. (For TSYM = 200 MeV, X = 1 corresponds to x = 1/πTSYM = 0.31 fm.) (2) The Neck region can be compared with the region where the Knudsen number exceeds some constant of order unity. An appropriate version of the Knudsen number in the current context is ~sub  ∇ · S . (189) Kn ≡ Γ ~sub  S Here the sound attenuation length Γ is the same as the one discussed following (120): Γ = 4η/3sT = 1/3πT . For the choice of parameters (187), examination of the nearfield expressions (164) shows that the region where Kn > ∼ 1/3 is somewhat bigger in the X1 direction than the Neck. However, corrections to ~sub may not be negligible for X1 and/or X⊥ the nearfield approximation to S of order unity. (3) The Neck region can be compared with the region where the constitutive relamn tions of hydrodynamics break down. Given all components Tsub of the stress tensor, with the Coulomb field subtracted away, there is a straightforward procedure for testing the constitutive relations. First determine the local velocity field U m by passing to the local rest frame of the fluid. Let (T mn )L be the subtracted stressenergy tensor in the local rest frame. The energy density is read off immediately as (T 00 )L ; the pressure is deduced from the equation of state; and the shear viscosity contribution to the spacespace parts of the stress tensor can be determined from U m and its gradient. Deviations of the spacespace components of (T mn )L from the combined contribution of pressure and shear viscosity are measures of the failure of hydrodynamics. A study78 of the nearfield expressions (164) for a somewhat different choice of parameters from (187) (namely v = 0.99, λ = 3π, N = 3) concludes that
Energy Loss and the GaugeString Duality
49
deviations from hydrodynamics are appreciable out as far as X ∼ 8. However, the nearfield expressions definitely cannot be trusted at such large distances.i This analysis could therefore be considerably improved if all components of mn Tsub were computed directly from string theory. The main conclusion to draw from points 2 and 3 is that in the Neck region, the subtracted stress tensor is essentially unrelated to hydrodynamics. Instead, the physics may be presumed to be dominated by strong coherent color fields combined with responses of the medium to strong field gradients: hence the term “chromoviscous neck.” To return to hadronization: The CooperFrye integral over R3 can be split into the Neck region and the “Mach” region — which is everything else. In the Mach region, where the energy density comes mainly from the bath, a good approximation to the local rest frame can be found by setting ~ ~ = 3 Ssub . U 4 bath
(190)
In the neck region, this approximation is less reliable, but because spacespace components of T mn are not available from a string theory calculation, it is hard to give a better motivated prescription for determining the local rest frame. With the choice (190), the result is that the Neck contribution to the CooperFrye integral leads to a distinctive doublepeaked structure in f (pT , φ) for pT . This is remarkable when compared to the singlepeaked structure emerging from a computation in a perturbative QCD framework based on Joule heating,81 which is similarly passed through the CooperFrye hadronization algorithm. See Fig. 9. The doublepeaked structure from the Neck region of the trailing string stress tensor has nothing to do with the Mach cone. It doesn’t occur at the same angle: for example, at v = 0.58, the Mach angle ϑ = cos−1 cs /v is very nearly zero, but the Neck region still produces a double peak structure (not shown in Fig. 9) about a radian away from φ = π. When v is very close to 1, the double peaks get closer to φ = π. This is reminiscent of the structure observed at large K in Fourier space,46, 54 but the peaks observed in the predicted hadron spectra are more widely separated than the ones in Fourier space. While the hadronization studies63, 77–83 give valuable insight into the relation of the trailing string to highangle hadron emission from an energetic parton, it is not claimed that the results are fully realistic, or that a direct comparison to untagged dihadron histograms, like the ones in Fig. 7, is justified. Let us review the potential difficulties. First, the trailing string describes an infinitely massive quark that propagates at a constant velocity through an infinite, static, thermal medium. Fluctuations leading to stochastic motions of finite mass quarks may significantly affect i A computation of the energy radiated from a moving quark53 shows that agreement with linearized hydrodynamics is already fairly good at X ∼ 5. Based on results of an earlier study,50 the onset of reasonable agreement with linearized hydro occurs near X = 3.
50
S. S. Gubser et al.
Fig. 9. (Color online) Hadron production based on a CooperFrye hadronization of a perturbative QCD calculation and of trailing string results (AdS/CFT).81 Bath contributions have been subtracted away, and the curves have all been normalized to have the same maximum when the CooperFrye integral is carried out over the entire volume accessible to each computation. The Neck region for the perturbative QCD calculation is again defined as a region close to the quark where deviations from hydrodynamics are significant.
the results. Also, there may be an effective “form factor” for massive quarks that partially smears out the field close to the quark. In addition, it is not obvious that CooperFrye is justified, because the crucial effect comes from the nonequilibrium part of the medium (the Neck). Subtracting away the Coulomb field is certainly wellmotivated physically, but it is possible to maintain some skepticism about whether it is the correct prescription in combination with CooperFrye. Finally, the approximation (190) to the local rest frame is imprecise in the Neck region. Despite these potentially serious issues, the punchline of the phenomenological studies63, 77–83 seems to us likely to be robust: the nearquark region has a substantially greater tendency toward high angle emission in the trailing string treatment
Energy Loss and the GaugeString Duality
51
than in the perturbative QCD treatment based on Joule heating. Modulo concerns already expressed, the nearfield contribution to highangle emission is stronger than the contribution of the hydrodynamical regime, and it results in a significant doublehump structure, reminiscent of jetsplitting.
9. Conclusions Let us conclude by addressing the four main questions we raised in the introduction: (1) What is the rate of energy loss from an energetic probe? For√ a heavy quark moving at a velocity v, the drag force is Fdrag = v − π 2 λ T 2 √1−v . This is explained in more detail in Sec. 4. 2 (2) What is the hydrodynamical response far from the energetic probe? √ There is a sonic boom with Mach angle ϑ = cos−1 cs /v, where cs = 1/ 3 is the speed of sound dictated by conformal invariance. There is also a diffusion wake of comparable strength. These points are explained in Secs. 6.1.2 and 7. (3) What gaugeinvariant information can be extracted using the gaugestring duality about the nonhydrodynamic region near the probe? The expectation values hT m0 i of the energy density and the Poynting vector of the gauge theory stress tensor have been computed with uniformly good accuracy across all length scales for several values of the velocity of the heavy quark, as we review in Sec. 7. Analytic approximations to the spacespace components of hT mn i are also available at small length scales: see Sec. 6.2. (4) Do the rate and pattern of energy loss have some meaningful connection to heavy ion phenomenology? A suitable translation of parameters from SYM to QCD results in estimates of energy loss for c and b quarks which are not far from realistic, or which may be fully realistic. We summarize these estimates in Sec. 4. Studies of hadronization starting from the string theory predictions for the energy density and Poynting vector indicate that the trailing string leads to significant highangle emission from the Neck region, close to the quark, suggestive of jetsplitting. We describe these studies in Sec. 8. Although it is premature to make detailed comparisons to data, it is clearly worthwhile to extend and refine both the string theory analysis and the phenomenological studies.
Acknowledgments We thank M. Gyulassy and J. Noronha for useful correspondence. This work was supported in part by the Department of Energy under Grant No. DEFG0291ER40671 and by the NSF under award number PHY0652782. Fabio De Rocha was also supported in part by the FCT grant SFRH/BD/30374/2006.
52
S. S. Gubser et al.
Appendix A. Notation In this appendix we present short explanations of some of the nomenclature and mathematical notations used in the main text. xµ : The five spacetime coordinates of AdS5 Schwarzschild, usually (t, x1 , x2 , x3 , z). m x , pm : The fourvectors for position and momentum in R3,1 . We use mostly plus signature, so (for example) η mn pm pn = −E 2 + p~2 = −m2 . ~x, p~: The threevectors for position and momentum: spatial components of xm and pm . 2 x1 : This could mean either G1µ xµ = Lz2 x1 or η1ν xν = x1 . Our convention is to prefer the latter; likewise x2 = x2 and x3 = x3 . x⊥ : p The radial distance from the quark in the x2 , x3 plane: x⊥ = x22 + x23 . Occasionally we consider the twovector ~x⊥ = (x2 , x3 ). z: This is the depth coordinate in AdS5 or AdS5 Schwarzschild which is 0 at the boundary and has dimensions of length. The AdS5 metric 2 is ds2 = Lz2 (−dt2 + d~x2 + dz 2 ). zH : The depth of the horizon in AdS5 Schwarzschild, related to the temperature by T = 1/πzH . y: Usually, a rescaled depth coordinate in AdS5 Schwarzschild, defined by y = z/zH . But in Sec. 8 we use y to denote rapidity, i.e. tanh y = pz /E where pz is the momentum along the beampipe and E is the energy. r: We use r to indicate a radial separation in R3 . Some authors use r to denote the depth coordinate r = L2 /z in AdS5 . AdS5 : Fivedimensional antide Sitter space, the maximally symmetric negatively curved spacetime in 4 + 1 dimensions. Its metric is given by (3) with h = 1. SYM: An abbreviation for “N = 4 superYangMills theory in four dimensions,” which is the theory controlling the lowenergy excitations of D3branes. N : Usually, the number of colors: N = 3 in QCD. An exception is that in Sec. 8, we use N to indicate the number of hadrons predicted by the CooperFrye algorithm. 2 gYM : The gauge coupling of SYM, normalized so that gYM N = L4 /α02 , where N is the number of colors. gs : The gauge coupling of QCD. We also use αs = gs2 /4π. 2 λ: The ’t Hooft coupling, λ = gYM N. L: The radius of curvature of AdS5 . α0 : The Regge slope parameter of fundamental strings, see (14). κ: The fivedimensional gravitational coupling. Gµν : The spacetime metric of AdS5 or AdS5 Schwarzschild. R: The Ricci scalar in AdS5 or AdS5 Schwarzschild. We also use the
Energy Loss and the GaugeString Duality
h:
T:
gαβ : σα : v: ξ: xµ∗ (σ): RAA :
τµν :
K τµν : hµν : hK µν : Axial gauge: Hmn :
A:
ψS :
53
Ricci tensor Rµν and the Riemann tensor Rαµβν . Our conventions are R = Gµν Rµν and Rµν = Gαβ Rαµβν , with signs arranged so that Rµν = − L42 Gµν in AdS5 of radius L. The “blackening function” for AdS5 Schwarzschild, whose metric is 4 given in (3). It is given by h(z) = 1 − z 4 /zH . We sometimes think of h as a function of the depth z, and sometimes as a function of y = z/zH . h0 always means h0 (y) = −4y 3 . The temperature in the dual field theory, which is the same as the Hawking temperature of the dual black hole background. The temperature of the AdS5 Schwarzschild background (3) is T = 1/πzH . The worldsheet metric of a string. Coordinates on the string worldsheet. The speed of a moving quark. Usually we take this motion to be in the +x1 direction. Gives the shape of the string that describes the quark in the fivedimensional geometry. See (13) and (20). The embedding function for a classical string in AdS5 Schwarzschild. When no ambiguity is possible, we denote this embedding function more simply as xµ (σ). The nuclear modification factor, defined as the number of particles produced (usually at a particular value of pT and in a specified range of rapidity) in a collision of two nuclei with atomic number A, divided by the number produced in a protonproton collision scaled up by the effective number of binary nucleonnucleon collisions in the heavyion collision. The fivedimensional stressenergy tensor, not to be confused with the expectation value of the boundary stressenergy tensor hTmn i. For the trailing string, τµν is given in (42). The Fourier components of τµν . See (46) and (48). Small metric perturbations around AdS5 Schwarzschild. The Fourier components of hµν defined by analogy with (46). A gauge choice for the metric perturbations where hµz = 0. The Fourier components of hK µν in axial gauge, up to a normalization factor. See (49). We think of the Hmn as functions of y = z/zH , not of z. The even tensor mode combination of metric perturbations in axial gauge: see (51a). Similar definitions for Bi , C, Di , and Ei follow, and gaugeinvariant combinations B, D, and E can be found in (61)(62). These are all functions of y = z/zH . The scalar master field. A master field is a gaugeinvariant combination of metric fluctuations in AdS5 Schwarzschild with a simple equation of motion. We encountered four other master fields: ψVeven , ψVodd , ψTeven , and ψTodd : see (67).
54
S. S. Gubser et al.
√ αv : A recurring normalization factor given by αv = 1/ 2πα0 1 − v 2 . ~ The momentum conjugate to ~x/zH ; used as a dimensionless waveK: number to parameterize the threedimensional Fourier space used to describe the medium’s response to the quark. ~ perpendicular to the motion K⊥ : The magnitude of the component of K of the quark. Rmn : The asymptotic values of Hmn at y = 0. The boundary condition Rmn = 0 says that the fourdimensional metric which the boundary gauge theory experiences is flat Minkowski space. Pmn : The coefficients of y 3 in a small y expansion of Hmn . See (76) and (77). They are related to the divergent contribution to the boundary stressenergy tensor given in (90). Qmn : The coefficients of y 4 in a small y expansion of Hmn when Rmn = 0. They are related to the expectation value of the boundary stressenergy tensor by (91). gmn : The metric of the boundary conformal field theory, usually set equal to the Minkowski metric ηmn with mostly plus signature. um : The fourvelocity of a heavy quark moving through the thermal medium. hTmn i: The onepoint function of the stressenergy tensor in SYM in the presence of the moving quark. We find it convenient to decompose it into three pieces given in (89). hydro (T )mn : A stress tensor which satisfies the hydrodynamic constitutive relations, (120). (T hydro,SYM)mn is the hydrodynamic contribution to the stress tensor of the SYM theory. K hTmn i: The Fourier modes of the contribution of the moving quark to the stressenergy tensor in the dual field theory minus the divergent piece corresponding to the infinite mass of the quark. See (89). It can be computed from Qmn through (95). fn : The source term for the energymomentum tensor, iKm T mn = f n . Various superscripts specify which contribution of the energymomentum tensor is being sourced. For example, fnhydro sources (T hydro )mn . RX : The coefficient of the leading homogeneous solution for various linear combinations of Hmn and their derivatives, such as A, B, Bi , ψT , etc. Our boundary condition is RX = 0. PX : The analog of Pmn for various linear combinations of Hmn and their derivatives. QX : The analog of Qmn for various linear combinations of Hmn and their derivatives. See (102) and (108). Also, we use the notations QD = QD1 and QE = QE1 . pT : The component of momentum perpendicular to the beamline.
Energy Loss and the GaugeString Duality
55
ϑ: An angular coordinate in momentum space, sin ϑ = K1 /K, or in real space, sin ϑ = x1 /x. ϕ: The azimuthal angle around the direction of motion of an energetic quark. Our usual convention is that the energetic quark moves in the +x1 direction, and then tan ϕ = K3 /K2 or x3 /x2 . θ: The angle of a trajectory relative to the beam. θ = π/2 is midrapidity. φ: The azimuthal angle around the beam. The angular variable ∆φ in dihadron histograms is the separation in φ between two hadrons. I, K: Modified Bessel functions of the first and second kind. J is a Bessel function of the first kind, and L is a modified Struve function. Neck: The neck is the region near a moving quark where the response of the medium is nonhydrodynamical. In practice, for the choice of parameters (187), the Neck can be defined, as in (188), as the region where the energy density, excluding the Coulombic contribution, exceeds 1.3 times the asymptotic energy density of an infinite static bath. U m : The fourvelocity of a fluid element, usually defined so that it vanishes in precisely the same Lorentz frame in which the Poynting vector vanishes. ~ The energy density and Poynting vector, rescaled to make them E, S: dimensionless, with contributions from the thermal bath excluded: see (165) and (172). E can be decomposed into a sum of contributions from the Coulomb field of the quark, subleading UV effects, IR effects, and a residual quantity Eres , as in (166). An analogous ~ decomposition can be performed on S. References 1. J. M. Maldacena, The large N limit of superconformal field theories and supergravity, Adv. Theor. Math. Phys. 2 (1998) 231–252, hepth/9711200. 2. S. S. Gubser, I. R. Klebanov, and A. M. Polyakov, Gauge theory correlators from noncritical string theory, Phys. Lett. B428 (1998) 105–114, hepth/9802109. 3. E. Witten, Antide Sitter space and holography, Adv. Theor. Math. Phys. 2 (1998) 253–291, hepth/9802150. 4. S. S. Gubser and A. Karch, From gaugestring duality to strong interactions: a Pedestrian’s Guide, 0901.0935. 5. O. Aharony, S. S. Gubser, J. M. Maldacena, H. Ooguri, and Y. Oz, Large N field theories, string theory and gravity, Phys. Rept. 323 (2000) 183–386, hepth/9905111. 6. I. R. Klebanov, TASI lectures: Introduction to the AdS/CFT correspondence, hepth/0009139. 7. E. D’Hoker and D. Z. Freedman, Supersymmetric gauge theories and the AdS/CFT correspondence, hepth/0201253. 8. S. S. Gubser, I. R. Klebanov, and A. W. Peet, Entropy and temperature of black 3branes, Phys. Rev. D54 (1996) 3915–3919, hepth/9602135.
56
S. S. Gubser et al.
9. J. Dai, R. G. Leigh, and J. Polchinski, New connections between string theories, Mod. Phys. Lett. A4 (1989) 2073–2083. 10. J. Polchinski, Dirichletbranes and Ramond–Ramond charges, Phys. Rev. Lett. 75 (1995) 4724–4727, hepth/9510017. 11. E. Witten, Bound states of strings and pbranes, Nucl. Phys. B460 (1996) 335–350, hepth/9510135. 12. J. H. Schwarz, Covariant field equations of chiral N = 2 D = 10 supergravity, Nucl. Phys. B226 (1983) 269. 13. G. Policastro, D. T. Son, and A. O. Starinets, From AdS/CFT correspondence to hydrodynamics, JHEP 09 (2002) 043, hepth/0205052. 14. S. Bhattacharyya, V. E. Hubeny, S. Minwalla, and M. Rangamani, Nonlinear fluid dynamics from gravity, JHEP 02 (2008) 045, 0712.2456. 15. G. Policastro, D. T. Son, and A. O. Starinets, The shear viscosity of strongly coupled N = 4 supersymmetric Yang–Mills plasma, Phys. Rev. Lett. 87 (2001) 081601, hepth/0104066. 16. S.J. Rey and J.T. Yee, Macroscopic strings as heavy quarks in large N gauge theory and antide Sitter supergravity, Eur. Phys. J. C22 (2001) 379–394, hepth/9803001. 17. C. P. Herzog, A. Karch, P. Kovtun, C. Kozcaz, and L. G. Yaffe, Energy loss of a heavy quark moving through N = 4 supersymmetric Yang–Mills plasma, JHEP 07 (2006) 013, hepth/0605158. 18. S. S. Gubser, Drag force in AdS/CFT, Phys. Rev. D74 (2006) 126005, hepth/ 0605182. 19. J. CasalderreySolana and D. Teaney, Heavy quark diffusion in strongly coupled N = 4 Yang–Mills, hepph/0605199. 20. K. Peeters, J. Sonnenschein, and M. Zamaklar, Holographic melting and related properties of mesons in a quark gluon plasma, Phys. Rev. D74 (2006) 106008, hepth/0606195. 21. H. Liu, K. Rajagopal, and U. A. Wiedemann, An AdS/CFT calculation of screening in a hot wind, hepph/0607062. 22. M. Chernicoff, J. A. Garcia, and A. Guijosa, The energy of a moving quarkantiquark pair in an N = 4 SYM plasma, hepth/0607089. 23. J. CasalderreySolana and D. Teaney, Transverse momentum broadening of a fast quark in a N = 4 Yang–Mills plasma, JHEP 04 (2007) 039, hepth/0701123. 24. S. S. Gubser, Momentum fluctuations of heavy quarks in the gaugestring duality, Nucl. Phys. B790 (2008) 175–199, hepth/0612143. 25. C. P. Herzog, Energy loss of heavy quarks from asymptotically AdS geometries, JHEP 09 (2006) 032, hepth/0605191. 26. P. Talavera, Drag force in a string model dual to largeN QCD, JHEP 01 (2007) 086, hepth/0610179. 27. E. Caceres and A. Guijosa, Drag force in charged N = 4 SYM plasma, JHEP 11 (2006) 077, hepth/0605235. 28. G. Michalogiorgakis, Heavy ion collisions, quasinormal modes and nonlinear sigma models, PhD thesis, Princeton University, 2007. 29. C. P. Herzog and A. Vuorinen, Spinning dragging strings, JHEP 10 (2007) 087, 0708.0609. 30. S. S. Gubser and A. Yarom, Universality of the diffusion wake in the gaugestring duality, Phys. Rev. D77 (2008) 066007, 0709.1089. 31. M. Chernicoff and A. Guijosa, Acceleration, energy loss and screening in stronglycoupled gauge theories, JHEP 06 (2008) 005, 0803.3070.
Energy Loss and the GaugeString Duality
57
32. Y. Hatta, E. Iancu, and A. H. Mueller, Jet evolution in the N = 4 SYM plasma at strong coupling, JHEP 05 (2008) 037, 0803.2481. 33. Y. Akamatsu, T. Hatsuda, and T. Hirano, Heavy quark diffusion with relativistic Langevin dynamics in the quarkgluon fluid, 0809.1499. 34. S. S. Gubser, Comparing the drag force on heavy quarks in N = 4 superYang–Mills theory and QCD, hepth/0611272. 35. H. Liu, K. Rajagopal, and U. A. Wiedemann, Calculating the jet quenching parameter from AdS/CFT, hepph/0605178. 36. F. Karsch, Lattice QCD at high temperature and density, Lect. Notes Phys. 583 (2002) 209–249, heplat/0106019. 37. D. Bak, A. Karch, and L. G. Yaffe, Debye screening in strongly coupled N = 4 supersymmetric Yang–Mills plasma, JHEP 08 (2007) 049, 0705.0994. 38. O. Kaczmarek and F. Zantow, Static quark antiquark interactions in zero and finite temperature QCD. I: Heavy quark free energies, running coupling and quarkonium binding, Phys. Rev. D71 (2005) 114510, heplat/0503017. 39. PHENIX Collaboration, A. Adare et al., Energy loss and flow of heavy quarks √ in Au+Au collisions at sN N = 200 GeV, Phys. Rev. Lett. 98 (2007) 172301, nuclex/0611018. 40. H. van Hees, V. Greco, and R. Rapp, Heavyquark probes of the quarkgluon plasma at RHIC, Phys. Rev. C73 (2006) 034913, nuclth/0508055. 41. H. van Hees and R. Rapp, Thermalization of heavy quarks in the quarkgluon plasma, Phys. Rev. C71 (2005) 034907, nuclth/0412015. 42. STAR Collaboration, B. I. Abelev et al., Transverse momentum and centrality dependence of highpt nonphotonic electron suppression in Au+Au collisions at √ sN N = 200 GeV, Phys. Rev. Lett. 98 (2007) 192301, nuclex/0607012. 43. J. de Boer, V. E. Hubeny, M. Rangamani, and M. Shigemori, Brownian motion in AdS/CFT, 0812.5112. 44. W. A. Horowitz and M. Gyulassy, Testing AdS/CFT Deviations from pQCD Heavy Quark Energy Loss with Pb+Pb at LHC, 0706.2336. 45. J. CasalderreySolana, E. V. Shuryak, and D. Teaney, Hydrodynamic flow from fast particles, hepph/0602183. 46. J. J. Friess, S. S. Gubser, G. Michalogiorgakis, and S. S. Pufu, The stress tensor of a quark moving through N = 4 thermal plasma, Phys. Rev. D75 (2007) 106003, hepth/0607022. 47. S. S. Gubser and S. S. Pufu, Master field treatment of metric perturbations sourced by the trailing string, Nucl. Phys. B790 (2008) 42–71, hepth/0703090. 48. S. S. Gubser, S. S. Pufu, and A. Yarom, Shock waves from heavyquark mesons in AdS/CFT, JHEP 07 (2008) 108, 0711.1415. 49. A. Yarom, On the energy deposited by a quark moving in an N = 4 SYM plasma, Phys. Rev. D75 (2007) 105023, hepth/0703095. 50. S. S. Gubser, S. S. Pufu, and A. Yarom, Energy disturbances due to a moving quark from gaugestring duality, JHEP 09 (2007) 108, 0706.0213. 51. P. M. Chesler and L. G. Yaffe, The wake of a quark moving through a stronglycoupled N = 4 supersymmetric Yang–Mills plasma, Phys. Rev. Lett. 99 (2007) 152001, 0706.0368. 52. S. S. Gubser, S. S. Pufu, and A. Yarom, Sonic booms and diffusion wakes generated by a heavy quark in thermal AdS/CFT, Phys. Rev. Lett. 100 (2008) 012301, 0706.4307. 53. P. M. Chesler and L. G. Yaffe, The stressenergy tensor of a quark moving through a stronglycoupled N = 4 supersymmetric Yang–Mills plasma: comparing hydrodynamics and AdS/CFT, Phys. Rev. D78 (2008) 045013, 0712.0050.
58
S. S. Gubser et al.
54. J. J. Friess, S. S. Gubser, and G. Michalogiorgakis, Dissipation from a heavy quark moving through N = 4 superYang–Mills plasma, JHEP 09 (2006) 072, hepth/0605292. 55. Y.H. Gao, W.S. Xu, and D.F. Zeng, Wake of color fields in charged N = 4 SYM plasmas, hepth/0606266. 56. J. MartinezMorales, Eigentensors of the Lichnerowicz operator in Euclidean Schwarzschild metrics, Annalen der Physik 15 (2006), no. 9 653–662. 57. P. K. Kovtun and A. O. Starinets, Quasinormal modes and holography, Phys. Rev. D72 (2005) 086009, hepth/0506184. 58. V. Balasubramanian and P. Kraus, A stress tensor for antide Sitter gravity, Commun. Math. Phys. 208 (1999) 413–428, hepth/9902121. 59. S. de Haro, S. N. Solodukhin, and K. Skenderis, Holographic reconstruction of spacetime and renormalization in the AdS/CFT correspondence, Commun. Math. Phys. 217 (2001) 595–622, hepth/0002230. 60. S. S. Gubser and A. Yarom, Linearized hydrodynamics from probesources in the gaugestring duality, 0803.0081. 61. I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and Products, Academic Press, 7th edn., 2007. Edited by A. Jeffrey and D. Zwillinger. 62. A. Yarom, The high momentum behavior of a quark wake, Phys. Rev. D75 (2007) 125010, hepth/0702164. 63. J. Noronha, G. Torrieri, and M. Gyulassy, Near zone Navier–Stokes analysis of heavy quark jet quenching in an N = 4 SYM plasma, Phys. Rev. C78 (2008) 024903, 0712.1053. 64. STAR Collaboration, J. Adams et al., Distributions of charged hadrons associated with high transverse momentum particles in p p and Au + Au collisions at s(NN)**(1/2) = 200GeV, Phys. Rev. Lett. 95 (2005) 152301, nuclex/0501016. 65. PHENIX Collaboration, S. S. Adler et al., Modifications to dijet hadron pair correlations in Au + Au collisions at s(NN)**(1/2) = 200GeV, Phys. Rev. Lett. 97 (2006) 052301, nuclex/0507004. 66. PHENIX Collaboration, A. Adare et al., Dihadron azimuthal correlations in Au+Au √ collisions at sN N = 200 GeV, Phys. Rev. C78 (2008) 014901, 0801.4545. 67. J. G. Ulery, Two and ThreeParticle JetLike Correlations, 0801.4904. 68. STAR Collaboration, B. I. Abelev et al., Indications of Conical Emission of Charged Hadrons at RHIC, 0805.0622. 69. J. G. Ulery, Conical emission in heavy ion collisions, J. Phys. G35 (2008) 104032, 0807.1613. 70. H. Stoecker, Collective flow signals the quark gluon plasma, Nucl. Phys. A750 (2005) 121–147, nuclth/0406018. 71. J. CasalderreySolana, E. V. Shuryak, and D. Teaney, Conical flow induced by quenched QCD jets, J. Phys. Conf. Ser. 27 (2005) 22–31, hepph/0411315. 72. A. K. Chaudhuri and U. Heinz, Effect of jet quenching on the hydrodynamical evolution of QGP, Phys. Rev. Lett. 97 (2006) 062301, nuclth/0503028. 73. J. CasalderreySolana and E. V. Shuryak, Conical flow in a medium with variable speed of sound, hepph/0511263. 74. T. Renk and J. Ruppert, Mach cones in an evolving medium, Phys. Rev. C73 (2006) 011901, hepph/0509036. 75. T. Renk and J. Ruppert, The rapidity structure of Mach cones and other large angle correlations in heavyion collisions, Phys. Lett. B646 (2007) 19–23, hepph/0605330. 76. B. Betz et al., Universality of the Diffusion Wake from Stopped and PunchThrough Jets in HeavyIon Collisions, 0812.4401.
Energy Loss and the GaugeString Duality
59
77. J. Noronha and M. Gyulassy, Machlike Angular Correlations Arise Only from the Head Zone of AdS/CFT String Jets, 0806.4374. 78. J. Noronha, M. Gyulassy, and G. Torrieri, Near zone hydrodynamics of AdS/CFT jet wakes, J. Phys. G35 (2008) 104061, 0806.4665. 79. J. Noronha, M. Gyulassy, and G. Torrieri, NonMach Conical DiJet Correlations Associated with Heavy Quark Jets in AdS/CFT, 0807.1038. 80. M. Gyulassy, J. Noronha, and G. Torrieri, Conical DiJet Correlations from a ChromoViscous Neck in AdS/CFT, 0807.2235. 81. B. Betz, M. Gyulassy, J. Noronha, and G. Torrieri, Anomalous Conical Dijet Correlations in pQCD versus AdS/CFT, 0807.4526. 82. B. Betz, J. Noronha, M. Gyulassy, and G. Torrieri, Jet Energy Loss and Mach Cones in pQCD versus AdS/CFT, 0812.1905. 83. G. Torrieri, B. Betz, J. Noronha, and M. Gyulassy, Mach cones in heavy ion collisions, Acta Phys. Polon. B39 (2008) 3281–3308, 0901.0230. 84. F. Cooper and G. Frye, Comment on the single particle distribution in the hydrodynamic and statistical thermodynamic models of multiparticle production, Phys. Rev. D10 (1974) 186.
QUARKONIUM AT FINITE TEMPERATURE
ALEXEI BAZAVOV Department of Physics, University of Arizona, Tucson, AZ 85721, USA ´ PETER PETRECZKY RIKENBNL Research Center and Physics Department, Brookhaven National Laboratory, Upton, NY 11973, USA ALEXANDER VELYTSKY∗ Enrico Fermi Institute, University of Chicago, 5640 S. Ellis Ave., Chicago, IL 60637, USA and HEP Division and Physics Division, Argonne National Laboratory, 9700 Cass Ave., Argonne, IL 60439, USA
We discuss properties of heavy quarkonium states at high temperatures based on lattice QCD and potential models. We review recent progress made in lattice calculations of spatial static quark antiquark correlators as well as quarkonium correlators in Euclidean time. Recent developments in effective field theory approach and potential models are also discussed.
1. Introduction There was considerable interest in the properties and the fate of heavy quarkonium states at finite temperature since the famous conjecture by Matsui and Satz.1 It has been argued that color screening in medium will lead to quarkonium dissociation above deconfinement, which in turn can signal quark gluon plasma formation in heavy ion collisions. The basic assumption behind the conjecture by Matsui and Satz was the fact that medium effects can be understood in terms of a temperature dependent heavy quark potential. Color screening makes the potential exponentially suppressed at distances larger than the Debye radius and it therefore cannot bind the heavy quark and antiquark once the temperature is sufficiently high. Based on this idea potential models at finite temperature with different temperature dependent potentials have been used over the last two decades to study quarkonium properties at finite temperature (see Ref. 2 for a recent review). It was not until recently that effective field theory approach, the socalled thermal pNRQCD, has been developed to justify the use of potential models at finite temperature.3 This approach, however, is based on the weak coupling techniques and will be discussed ∗ Current
address: Physics Department, Brookhaven National Laboratory, Upton, NY 11973, USA. 61
62
A. Bazavov, P. Petreczky & A. Velytsky
in the next section. To understand the nonperturbative aspects of color screening lattice calculations of the spatial correlation functions of static quarks are needed. Recently a lot of progress has been made in this direction which will be the topic of Sec. 4. To prepare the reader for this in Sec. 3 we review the basics of lattice gauge theory. In principle it is possible to study the problem of quarkonium dissolution without any use of potential models. Inmedium properties of different quarkonium states and/or their dissolution are encoded in spectral functions. Spectral functions are related to Euclidean meson correlation functions which can be calculated on the lattice. Reconstruction of the spectral functions from the lattice meson correlators turns out to be very difficult, and despite several attempts its outcome still remains inconclusive. One remarkable feature of the studies of the lattice meson correlators is their small temperature dependence despite the expected color screening. This seems to be puzzling. We will discuss the possible resolution of this puzzle in Sec. 6, while the current status of the lattice calculations of the Euclidean correlators and the corresponding meson spectral functions will be presented in Sec. 5. The summary and outlook will be given in Sec. 7. 2. pNRQCD at Finite Temperature There are different scales in the heavy quark bound state problem related to the heavy quark mass m, the inverse size ∼ mv and the binding energy mv 2 . Here v is the typical heavy quark velocity in the bound state and is considered to be a small parameter. Therefore it is possible to derive a sequence of effective field theories using this separation of scales (see Refs. 4 and 5 for recent reviews). Integrating out modes at the highest energy scale ∼ m leads to an effective field theory called nonrelativistic QCD or NRQCD, where the pair creation of heavy quarks is suppressed by powers of the inverse mass and the heavy quarks are described by nonrelativistic Pauli spinors.6 At the next step, when the large scale related to the inverse size is integrated out, the potential NRQCD or pNRQCD appears. In this effective theory the dynamical fields include the singlet S(r, R) and octet O(r, R) fields corresponding to the heavy quark antiquark pair in singlet and octet states respectively, as well as light quarks and gluon fields at the lowest scale ∼ mv 2 . The Lagrangian of this effective field theory has the form ( nf Z ∇2r 1 a a µν X 3 † − Vs (r) S + q¯i iD / qi + d r Tr S i∂0 + L = − Fµν F 4 m i=1 +O
+
†
) n o ∇2r ~ S + S†~r · g E ~O − Vo (r) O + VA Tr O†~r · g E iD0 + m
o VB n † ~ O + O† O~r · g E ~ + ... . Tr O ~r · g E 2
(1)
Quarkonium at Finite Temperature
63
Here the dots correspond to terms which are higher order in the multipole expansion.5 The relative distance r between the heavy quark and antiquark plays a role of a label, the light quark and gluon fields depend only on the centerofmass coordinate R. The singlet Vs (r) and octet Vo (r) heavy quark potentials appear as matching coefficients in the Lagrangian of the effective field theory and therefore can be rigorously defined in QCD at any order of the perturbative expansion. At leading order Vs (r) = −
1 αs N 2 − 1 αs , Vo (r) = 2N r 2N r
(2)
and VA = VB = 1. The free field equation for the singlet field is ∇2 i∂0 + r − Vs (r) S(r, R) = 0, m
(3)
i.e. has the form of a Schr¨ odinger equation with the potential Vs (r). In this sense, potential models emerge from the pNRQCD. Note, however, that pNRQCD also accounts for interaction of the soft gluons which cannot be included in potential models, i.e. it can describe retardation effects. One can generalize this approach to finite temperature. However, the presence of additional scales makes the analysis more complicated.3 The effective Lagrangian will have the same form as above, but the matching coefficients may be temperature dependent. In the weak coupling regime there are three different thermal scales : T , gT and g 2 T . The calculations of the matching coefficients depend on the relation of these thermal scales to the heavy quark bound state scales.3 To simplify the analysis the static approximation has been used, in which case the scale mv is replaced by the inverse distance 1/r between the static quark and antiquark. The binding energy in the static limit becomes Vo − Vs ≃ N αs /(2r). When the binding energy is larger than the temperature the derivation of pNRQCD proceeds in the same way as at zero temperature and there is no medium modifications of the heavy quark potential.3 But bound state properties will be affected by the medium through interactions with ultrasoft gluons, in particular, the binding energy will be reduced and a finite thermal width will appear due to medium induced singletoctet transitions arising from the dipole interactions in the pNRQCD Lagrangian3 (c.f. Eq. (1)). When the binding energy is smaller than one of the thermal scales the singlet and octet potential will be temperature dependent and will acquire an imaginary part.3 The imaginary part of the potential arises because of the singletoctet transitions induced by the dipole vertex as well as due to the Landau damping in the plasma, i.e. scattering of the gluons with spacelike momentum off the thermal excitations in the plasma. In general, the thermal corrections to the potential go like (rT )n and (mD r)n ,3 where mD denotes the Debye mass. Only for distances r > 1/mD there is an exponential screening. In this region the singlet potential has
64
A. Bazavov, P. Petreczky & A. Velytsky
a simple form αs −mD r 2 e + iCF αs T Vs (r) = −CF r rmD 2
Z
∞
dx
0
sin(mD r x) − CF αs (mD + iT ) , (x2 + 1)2
CF = (N − 1)/(2N )
(4)
The real part of the singlet potential coincides with the leading order result of the socalled singlet free energy.7 The imaginary part of the singlet potential in this limit has been first calculated in Ref. 8. For small distances the imaginary part vanishes, while at large distances it is twice the damping rate of a heavy quark.9 This fact was first noted in Ref. 10 for thermal QED. The effective field theory at finite temperature has been derived in the weak coupling regime assuming the separation of different thermal scales as well as ΛQCD . In practice the separation of these scales is not evident and one needs lattice techniques to test the approach. Therefore Sec. 4 will be dedicated to the study of static quarks at finite temperature on the lattice. To prepare the reader for this discussion some basics of the lattice gauge theory will be given in the next section. 3. Basics of Lattice Gauge Theory To study nonperturbative aspects of QCD we use lattice gauge theory.11 In this formalism a field theory is defined in a gaugeinvariant way on a discrete spacetime domain. This serves at least two purposes: a) to provide an ultraviolet cutoff for the theory, restricting highest momentum to π/a (a being the lattice spacing), and b) to evaluate the path integrals in the Euclidean formulation stochastically using importance sampling. On the lattice the fundamental degrees of freedom of a theory with local SU (N ) gauge symmetry are fermion fields ψx that reside on the sites of the lattice and carry flavor, color and Dirac indeces, which we suppress through the most of this paper, and gauge, bosonic degrees of freedom that in the form of SU (N ) matrices Ux,µ reside on links. Sites on a fourdimensional lattice are labeled with x ≡ (~x, t). The theory is defined by the partition function Z ¯ Z = DU DψDψ exp(−S) (5) where the action
S = Sg + Sf
(6)
contains gauge, Sg and fermionic, Sf parts. The latter part is bilinear in fields and has the form ¯ ψ Sf = ψM
(7)
where M is the fermion matrix. In the simplest formulation the lattice gauge action can be written as X 1 (8) Sg = β 1 − TrUP , N P
Quarkonium at Finite Temperature
65
where UP is the socalled plaquette, a product of link variables along the elementary square and β = 2N/g 2 with g 2 being the bare gauge coupling. This is the Wilson gauge action.11 The explicit form of the fermion action that is often used in lattice QCD calculations will be discussed in Sec. 5.2. ˆ is given then by The expectation value of an operator O ˆ = hOi
1 Z
Z
¯ ˆ exp(−S). DU DψDψ O
(9)
Integration over the fermion fields (which are Grasmann variables) can be carried out explicitly: Z=
Z
DU det M [U ] exp(−Sg ) ≡
Z
DU exp(−Sef f ),
(10)
where Sef f = Sg − ln det M [U ] is the effective action. The fermion determinant det M [U ] describes the vacuum polarization effects due to the dynamical quarks and makes the effective action nonlocal in gauge variables. For this reason simulations with dynamical quarks are very resource demanding and the quenched approximation is often employed, where det M [U ] is set to 1. To evaluate the path integral (9) stochastically, an ensemble of NU gauge configurations, weighted with exp(−Sef f ), is generated using Monte Carlo or Molecular Dynamics techniques. The expectation value of the operator is then approximated by the ensemble average: ˆ ≃ hOi
NU 1 X Oi (U ), NU i=1
(11)
ˆ calculated on ith configuration. (When where Oi (U ) is the value of the operator O, ˆ the operator O depends explicitly on the quark fields extra factors of M −1 appear as shown for a meson correlator below.) Consider a meson (quark antiquark pair) operator of a general form ¯ x, t)ΓU(~x, ~y; t)ψ(~y , t), J(~x, ~y; t) = ψ(~
(12)
where Γ determines the spin structure and U is a gauge connection that corresponds to the excitations of the gluonic field. Dirac and color indeces in (12) are suppressed. The propagation of such meson from time t = 0 to t is described by the correlation function hJ(~x1 , ~y1 ; 0)J(~x2 , ~y2 ; t)i =
1 Z
Z
¯ DU DψDψ exp(−S)
¯ x1 , 0)ΓU(~x1 , ~y1 ; 0)ψ(~y1 , 0)ψ(~ ¯ y2 , t)Γ† U † (~x2 , ~y2 ; t)ψ(~x2 , t). (13) × ψ(~
66
A. Bazavov, P. Petreczky & A. Velytsky
Again, integration over the quark fields can be carried out resulting in hJ(~x1 , ~y1 ; 0)J(~x2 , ~y2 ; t)i = hTr M −1 (~x2 , t; ~x1 , 0)ΓU(~x1 , ~y1 ; 0)M −1 (~y1 , 0; ~y2 , t)Γ† U † (~x2 , ~y2 ; t) i −1 − hTr M (~y1 , 0; ~x1 , 0)ΓU(~x1 , ~y1 ; 0) i × hTr M −1 (~x2 , t; ~y2 , t)Γ† U † (~x2 , ~y2 ; t) i.
(14)
The inverse of the fermion matrix, M −1 has meaning of a fermion propagator. Gauge transporters U(~x, ~y; t) can be taken as weighted sums of different paths connecting points ~x and ~ y . By choosing paths of certain shape or combinations of different paths it is possible to achieve a better overlap of the meson operator with a given state. One of the possibilities is to construct the gauge transporters by using APE smearing12 on spatial links: a link variable Ux,µ is replaced by a weighted average of itself and a sum of the 3link paths connecting the same sites as Ux,µ : X † ′ Ux,µ → Ux,µ = (1 − 6c)Ux,µ + c Ux,µ Ux+ˆν ,µ Ux+ˆ (15) µ,ν . ν6=µ
This procedure can be applied iteratively. Then a gauge transporter U(~x, ~y ; t), taken as a product of smeared links, is equivalent to a weighted sum of differently shaped paths connecting the sites ~x and ~y. In the following we will consider meson correlators at finite temperature. The finite temperature is introduced by compactifying the the Eucliden time direction, i.e. T = 1/(Nτ a) with Nτ being the number of temporal time slices. Gauge fields and fermion fields obey periodic and antiperiodic boundary conditions in the temporal direction. In the next section we consider a spinless static quark antiquark pair, Γ = I that can propagate only in time. In this case second term in (14) vanishes. In Sec. 5 we consider local meson operators with ~xi = ~yi , i = 1, 2 which means U(~x, ~y ; t) = I. 4. Correlation Functions of Static Quarks in Lattice Gauge Theory 4.1. Static meson correlators Consider static (infinitely heavy) quarks. The position of heavy quark antiquark pair is fixed in space and propagation happens only along the time direction. In this limit the second term on the right hand side of (14) vanishes. We are interested in a spinless state and set Γ = I in this section. With respect to the color the meson can be in a singlet or adjoint state. These states are described by the following gauge connections U(~x, ~y ; t) = U (~x, ~y ; t), a
(16) a
U (~x, ~y ; t) = U (~x, ~x0 ; t)T U (~x0 , ~y; t),
(17)
where U (~x, ~y ) is a spatial gauge transporter, – the product of the gauge variables along the path connecting ~x and ~y , ~x0 is the coordinate of the center of mass of the meson and T a are the SU (N ) group generators.
Quarkonium at Finite Temperature
67
The meson operators are given then by Eq. (12) with ~x1 = ~x2 , ~y1 = ~y2 for static quarks ¯ x, t)U (~x, ~y; t)ψ(~y , t), J(~x, ~y ; t) = ψ(~ ¯ x, t)U (~x, ~x0 ; t)T a U (~x0 , ~y ; t)ψ(~y , t). J a (~x, ~y ; t) = ψ(~
(18) (19)
Substituting expressions (18) into Eq. (14) and noting that for a static quark the propagator M −1 (~x, 0; ~x, t) ∼ L(~x), where the temporal Wilson line L(~x) = QNτ −1 x,t),0 with U(~ x,t),0 being the temporal links, we get for the meson correlat=0 U(~ tors at t = 1/T : 1 ¯ x, ~y ; 1/T )i hJ(~x, ~y ; 0)J(~ N 1 = hTr L† (~x)U (~x, ~y ; 0)L(~y)U † (~x, ~y , 1/T ) i, N 2 NX −1 1 hJ a (~x, ~y; 0)J¯a (~x, ~y; 1/T )i Ga (r, T ) ≡ 2 N − 1 a=1 G1 (r, T ) ≡
=
1 hTrL† (x)TrL(y)i −1 1 − hTr L† (x)U (x, y; 0)L(y)U † (x, y, 1/T ) i, 2 N (N − 1) r = ~x − ~y .
(20)
N2
(21)
The correlators depend on the choice of the spatial transporters U (~x, ~y ; t). Typically, a straight line connecting points ~x and ~y is used as a path in the gauge transporters, i.e. one deals with timelike rectangular cyclic Wilson loops. This object has been calculated at finite temperature in hard thermal loop (HTL) petrurbation theory in context of perturbative calculations of the singlet potential introduced in the previous section and quarkonium spectral functions.8,13,14 In the special gauge, where U (~x, ~y ; t) = 1 the above correlators give the standard definition of the singlet ¯ pair and adjoint free energies of a static QQ 1 hTr[L† (x)L(y)]i, N 1 hTrL† (x)TrL(y)i exp(−Fa (r, T )/T ) = 2 N −1 1 − hTr L† (x)L(y) i. N (N 2 − 1) exp(−F1 (r, T )/T ) =
(22)
(23)
The singlet and adjoint free energies can be calculated at high temperature in leading order HTL approximation7 resulting in (N 2 − 1)αs mD N 2 − 1 αs exp(−mD r) − , 2N r 2N 1 αs (N 2 − 1)αs mD exp(−mD r) − , Fa (r, T ) = 2N r 2N F1 (r, T ) = −
(24) (25)
68
A. Bazavov, P. Petreczky & A. Velytsky
p with mD = gT (N/3 + Nf /6) being the leading order Debye mass and Nf is the number of quark flavors. At this order F1 and Fa are gauge independent or, in other words, do not depend on the choice of the parallel transporters U (~x, ~y; t). Note that at small distances (rmD ≪ 1) the singlet free energy F1 (r, T ) ≃ −
N 2 − 1 αs 2N r
(26)
is temperature independent and coincides with the zero temperature potential, while the adjoint free energy Fa (r, T ) ≃
N 1 αs − αs mD 2N r 2
(27)
depends on the temperature. The physical free energy of a static quark antiquark pair, i.e. the one related to the work that has to be done to separate the static charges by certain distance is given by the thermal average of the singlet and adjoint free energies15 exp(−F (r, T )/T ) = =
1 N2 − 1 exp(−F1 (r, T )/T ) + exp(−Fa (r, T )/T ) 2 N N2 1 1 hTr [L(x)TrL(y)]i ≡ 2 G(r, T ). N2 N
(28)
This quantity is explicitly gauge independent. In leading order HTL approximation the free energy is F (r, T ) = −
(N 2 − 1) α2s exp(−2mD r). 8N 2 r2 T
(29)
The 1/r2 behavior is due to partial cancellation between the singlet and adjoint contribution15,16 and has been confirmed by lattice calculations in the intermediate distance regime above the deconfinement transition.17,18 Using the transfer matrix one can show that in the confined phase G1 (r, T ) =
∞ X
cn (r)e−En (r,T )/T ,
(30)
e−En (r,T )/T ,
(31)
n=1
G(r, T ) =
∞ X
n=1
where En are the energy levels of static quark and antiquark pair.19 The coefficients cn (r) depend on the choice of U (x, y; t) entering the static meson operator in Eqs. (1819). Since the color averaged correlator G(r, T ) corresponds to a gauge invariant measurable quantity it does not contain cn . The lowest energy level is the usual static quark antiquark potential, while the higher energy levels correspond to hybrid potentials.20–23 Using multipole expansion in pNRQCD one can show that at short distances the hybrid potential corresponds to the adjoint potential
Quarkonium at Finite Temperature
69
up to nonperturbative constants.24 Indeed, lattice calculations of the hybrid potentials indicate a repulsive short distance part.20–23 Furthermore, the gap between the static potential and the first hybrid potential can be estimated fairly well at short distances in perturbation theory.25 If c1 = 1 the dominant contribution to Ga would be the first excited state E2 , i.e. the lowest hybrid potential which at short distances is related to the adjoint potential. In this sense Ga is related to static mesons with quark antiquark in adjoint state. Numerical calculations show, however, that c1 is rdependent and in general c1 (r) 6= 1. Thus Ga also receives contribution from E1 .19 The lattice data suggests that c1 approaches unity at short distances19 in accord with expectations based on perturbation theory, where c1 = 1 up to O(α3s ) corrections.24 Therefore at short distances, r ≪ 1/T the color singlet and color averaged free energy are related F (r, T ) = F1 (r, T ) + T ln(N 2 − 1). In the following we consider SU (2) and SU (3) gauge theories and refer to the adjoint state as triplet and octet, correspondingly. 4.2. Lattice results on static meson correlators Correlation function of static quarks have been extensively studied on the lattice since the pioneering work by Mclerran and Svetistky.15 Most of these studies, however, considered only the color averaged correlator, i.e. the correlation function of two Polyakov loops (for the most complete analysis see Ref. 26 and references therein). Since both color singlet and color octet degrees of freedom contribute to the Polyakov loop correlator it has large temperature dependence even at short distances and it is not a very useful quantity if we want to learn something about quarkonium properties at high temperatures. As this has been pointed out in Refs. 27 and 28 we need to know the quark antiquark interactions in the singlet channel in order to learn about inmedium quarkonium properties. Therefore in recent years the singlet correlator has been computed on the lattice in SU (2) and SU (3) gauge theories18,29–32 as well as in full QCD with 3 and 2 flavors of dynamical quarks.33,34 Preliminary results also exist for 2+1 flavor QCD with physical value of the strange quark mass and light u, dquark masses corresponding to pion mass of about 220M eV .35,36 All these calculations use Coulomb gauge definition of the singlet correlator, i.e. the definition (22) with Coulomb gauge fixing. The numerical results for SU (3) gauge theory are presented in Fig. 1 and compared with the zero temperature potential. Here we used the string Ansatz for the zero temperature potential π + σr, (32) V (r) = − 12r since this form gives a very good description of the lattice data in SU (3) gauge the√ ory.37 To convert lattice units to physical units the value σ = 420MeV has been used for the string tension. To remove the additive renormalization in the singlet free energy the lattice data have been normalized to V (r) given by Eq. (32) at the
70
A. Bazavov, P. Petreczky & A. Velytsky
F1(r,T) [GeV]
2 1.5 1 0.5
r [fm]
0.87Tc 0.91Tc 0.94Tc 0.98Tc 1.05Tc 1.50Tc 3.00Tc
1.5
2
0 0.5 1 0
0.5
1
2.5
3
Fig. 1. The color singlet free energy singlet free energy in quenched QCD (SU(3) gauge theory) calculated in Coulomb gauge.29,31,32 The solid line shows the zero temperature potential.
shortest distance available. As one can see from Fig. 1 the singlet free energy is temperature independent at short distances and coincides with the zero temperature potential. Below the deconfinement transition, T < Tc , it rises linearly, indicating confinement. The string tension at finite temperature is smaller than the zero temperature string tension σ. At large distances the singlet free energy is the same as the free energy determined from the Polyakov loop correlators and therefore the finite temperature string tension agrees with findings of Ref. 26. Medium effects set in at distance rmed ≃ 0.4 fm/(T /Tc ) and we see exponential screening at distance r > 1/T .32 In the intermediate distance regime 0.5 fm < r < 1.5 fm we see that the singlet free energy is enhanced relative to the zero temperature potential. This is not a real physical effect but an artifact of the calculations. Similar effect has also been seen in SU (2) gauge theory.18,19,30 At these distances the color singlet correlator is sensitive to the value of the coefficients cn (r) in Eq. (30). Below we will show that the enhancement of the singlet free energy is due to the fact that c1 (r) < 1 in this region. Numerical results for the singlet free energy in 2+1 flavor QCD are shown in Fig. 2. In this case a different normalization procedure has been used. The additive renormalization has been determined at zero temperature for each value of the lattice spacing used in the finite temperature calculations by matching the zero temperature potential to the string Ansatz (32) at distance r = r0 , with r0 being the Sommer scale.38 This additive renormalization then has been used for the singlet free energy. For detailed discussion of the calculation of the static
Quarkonium at Finite Temperature
1
71
F1(r,T) [GeV]
0.8 0.6 0.4 0.2
T/Tc = 0.82
0.89 0.97 1.02 1.09 1.58 1.95 2.54 3.29
0 0.2 0.4 0.6 0.8 0
0.5
1
1.5
2
2.5
r [fm] Fig. 2. The color singlet free energy in 2+1 flavor QCD.35,36 The solid line is the parametrization of the lattice data on the zero temperature potential from Ref. 39.
potential and its renormalization see Ref. 39. As in quenched QCD the singlet free energy is temperature independent at short distances and coincides with the zero temperature potential. At distances larger than the inverse temperature it is exponentially screened. The novel feature of the singlet free energy in full QCD is the string breaking, i.e. the fact that it approaches a constant at large separation. This happens because if the energy in the string exceeds the binding energy of a heavylight meson the static quark and antiquark are screened due to pair creation from the vacuum. This happens at distances of about 0.8 fm according to the figure. As temperature increases the distance where the singlet free energy flattens out becomes smaller and for sufficiently high temperatures turns out to be inversely proportional to the temperature. Thus string breaking smoothly turns into color screening as temperature increases. 4.3. Color singlet correlator in SU(2) gauge theory at low temperatures For a better understanding of the temperature dependence of the singlet correlator and its physical interpretation the values of the overlap factors cn (r) should be estimated. To extract the overlap factors we need to calculate the singlet correlators in the low temperature region, where only few energy levels contribute to the correlator. The calculations of the singlet correlator at low temperatures is difficult because of the rapidly deceasing signal to noise ratio. To overcome this difficulty
72
A. Bazavov, P. Petreczky & A. Velytsky
5.5
0.95Tc 0.76Tc 0.63Tc 0.54Tc 0.48Tc 0.42Tc T=0
5
F1(r,T)/σ1/2
4.5 4 3.5
F1(r,T)/σ1/2
4
3 2.5
3
2
2
r σ1/2
0
0.5
1
1.5
2
1.5 0
0.5
1
1.5
2
1/2
rσ
Fig. 3. The color singlet free energy in SU (2) gauge theory below the deconfinement temperature at β = 2.5 calculated on 323 × Nτ lattices. Also shown is the T = 0 potential. The inset shows the color singlet free energy from which the contribution from the matrix element T ln c1 has been subtracted.
it has been suggested to calculate Wilson loops with multilevel LuescherWeisz algorithm40 instead of the Coulomb gauge fixing.41 In this case the color singlet correlator defined by Eq. (20) is an expectation value of the gaugeinvariant Wilson loop with gauge transporter U (~x, ~y; t) being a product of the iteratively smeared spatial links. Numerical calculations have been performed in SU (2) gauge theory using standard Wilson gauge action.41 With the use of the multilevel algorithm it was possible to go down to temperatures as low as 0.32Tc not accessible in the previous studies. It is well known that smearing increases the overlap with the ground state by removing the short distance fluctuations in the spatial links.42 For this reason smearing also reduces the breaking of the rotational invariance to the level expected in the free theory. When no smearing is used the color singlet free energy, −T ln G1 (r, T ) shows a small but visible temperature dependence. The temperature dependence of the singlet free energy is significantly reduced when APE smearing is applied. The color singlet free energy for β = 2.5 and 10 APE smearings is shown in Fig. 3. As one can see from the figure the color singlet free energy shows very mild temperature dependence in the confined phase with noticeable temperature effects appearing only at T = 0.95Tc.
Quarkonium at Finite Temperature
73
1.2 c1(r) 1 0.8 0.6 0.4 0.2
Ns=24 Ns=32 Ns=24,APE10 Ns=32,APE10
0 0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
r σ1/2 1.2 c1(r) 1 0.8 0.6 0.4 Ns=32 Ns=32, APE10 Ns=32, APE20
0.2 0.2
0.4
0.6
0.8 rσ
1
1.2
1/2
Fig. 4. The preexponential factor of the color singlet correlators as function of distance r for β = 2.5 (top) and β = 2.7 (bottom). Shown are results for unsmeared spatial links and 10 and 20 steps of APE smearing.
Consider the expansion (30). The dominant contribution comes from the ground state, so it is reasonable to fit the singlet correlator to the form G1 (r, T ) = c1 (r) exp(−E1 (r)/T ).
(33)
This allows to extract the matrix element c1 (r) using a simple exponential fit, which is shown in Fig. 4. When no APE smearing is used the value of c1 (r) strongly depends on the separation r. At small distances it shows a tendency of approaching unity as one would expect in perturbation theory. However, c1 (r) decreases with increasing distance r. At large distance its value is around 0.3 − 0.5. Similar results for c1 (r) have been obtained in the study of SU (2) gauge theory in 3 dimensions.19
74
A. Bazavov, P. Petreczky & A. Velytsky
When APE smearing is applied the rdependence of the amplitude c1 (r) is largely reduced and its value is close to unity both for β = 2.5 and β = 2.7. For β = 2.7 we also see that increasing the number of smearing steps from 10 to 20 reduces the deviation of c1 (r) from unity. As discussed in Sec. 4.1 perturbation theory predicts that the deviation of c1 (r) from unity is of order α3s . Therefore it can be made arbitrarily small by going to sufficiently small distances. It is known, however, that lattice perturbation theory converges very poorly. The main reason for this has been identified with the short distance fluctuations of the link variables, which makes their mean value very different from unity.43 Smearing removes these short distance fluctuations and this is the reason why c1 (r) is much closer to unity when APE smearing is applied. Thus, almost the entire temperature dependence of the singlet free energy at dis√ tances 0.5 < r σ < 2 is due to the deviation of c1 from unity and can be largely reduced by applying APE smearing to the links in the spatial gauge connections. To further demonstrate this point in the inset of Fig. 3 we show the results for F1 (r, T ) + T ln c1 (r). Clearly no temperature dependence can be seen in this quan√ tity up to 0.95Tc, where we see temperature dependence at distances r σ ≥ 1.5 corresponding to the expected drop of the effective string tension. 4.4. Color singlet free energy in the deconfined phase The behavior of the color singlet free energy in the deconfined phase has been studied in Coulomb gauge18,29,32–34 and from cyclic Wilson loops.41 As discussed above at short distances it is temperature independent and coincides with the zero temperature potential. At large distances it approaches a constant F∞ (T ), which monotonically decreases with the temperature. The constant F∞ (T ) is the free energy of two isolated static quarks, or equivalently of a quark antiquark pair at infinite separation. Its value is therefore independent of the definition of the singlet correlator G1 (r, T ) and is related to the renormalized Polyakov loop Lren (T ) = exp(−F∞ (T )/(2T )).29 At leading order F1 (r, T ) − F∞ (T ) is of Yukawa form (c.f. Eq. (24)). Therefore it is useful to define a quantity called the screening function S(r, T ) = r · (F1 (r, T ) − F∞ (T )).
(34)
This quantity shows exponential decay at distances r > 1/T in the deconfined phase both in pure gauge theories18,32 and full QCD.34 From its exponential decay the Debye screening mass has been determined and turns out to be about 40% larger than the leading order perturbative value. It is interesting to study the screening function using the free energy determined from cyclic Wilson loops and make comparison with Coulomb gauge results. This analysis has been recently done in Ref. 41 for SU (2) gauge theory. The behavior of the screening function at different temperatures is shown in Fig. 5. At short distances (rT < 0.5) the singlet free energy does not depend on the smearing level.
Quarkonium at Finite Temperature
75
1 S(r,T)
0.1
3.4Tc 2.3Tc 1.7Tc 1.4Tc 1.1Tc
0.01
0.001 0
0.2 0.4 0.6 0.8
1
1.2 1.4 1.6 1.8
2
1/2
rσ 1
1.2Tc, β=2.3533 1.3Tc, β=2.5000 1.5Tc, β=2.4215 1.4Tc, β=2.7 1.7Tc, β=2.7
S(r,T)
3
0.1
m1/T
2.5 2 1.5 T/Tc
1 1
2
3
4
5
6
0.01 0
0.2
0.4
0.6
0.8
1
1.2
1.4
rT Fig. 5. The screening function S(r, T ) = r(F1 (r, T ) − F∞ (T )) in SU (2) gauge theory at different √ temperatures calculated for β = 2.7 as function of r σ (top) and as function of rT (bottom).
Furthermore, it is very close to the free energy calculated in Coulomb gauge. At large distances the screening function S(r, T ) shows an exponential decay determined by a temperature dependent screening mass m1 (T ), which is equal to the leading order Debye mass up to the nonperturbative g 2 corrections: m1 = mD + O(g 2 ).44,45 There is some dependence on the smearing level at larger distances which, however, disappears at high temperatures and with increasing the smearing level. In particular, for β 6 2.5 it turns out that there is no dependence on smearing level for 5 or more smearing steps. For β = 2.7 10–20 steps are needed, depending on the temperature. Fitting the large distance behavior of the screening function by an exponential form exp(−m1 (T )r) allows to determine the screening mass m1 (T ).
76
A. Bazavov, P. Petreczky & A. Velytsky
In the inset of Fig. 5 the color singlet screening masses extracted from the fits are shown in comparison with the results obtained in Coulomb gauge in Ref. 18. The solid line is the leading order Debye mass calculated using 2loop gauge coupling g(µ = 2πT ) in M Sscheme. As we see from the figure the screening masses are smaller than those calculated in Coulomb gauge and agree well with the leading order perturbative prediction. 4.5. Color adjoint free energy The structure of Eqs. (20) and (21) shows that color adjoint correlator is given by a difference of the color averaged and singlet correlators. The color adjoint correlator in Coulomb gauge has been studied in pure gauge theory,7,18 3flavor QCD33 and 2flavor QCD.34 At low temperatures the color adjoint free energy turned out to be significantly smaller than the first hybrid potential contrary to the expectations. In fact at sufficiently large distance it was found to be identical to the singlet free energy. As has been pointed out in Ref. 19 this is due to the nontrivial rdependence of the overlap factor c1 (r) and its deviation from unity. We have shown in Sec. 4.3 that deviations of the overlap factor c1 (r) from unity can be greatly reduced when one uses Wilson loops with the smeared spatial gauge connection. Therefore it is interesting to see how the adjoint free energy behaves in this approach. The numerical analysis has been done in SU (2) gauge theory41 therefore below we will refer to the adjoint free energy as the triplet free energy. If we assume that only two states contribute to the Eqs. (30) and (31), then from Eq. (21) it follows that 1 ∆E(r)/T , (35) F3 (r, T ) = E2 (r) − T ln 1 − c2 (r) + (1 − c1 (r))e 3 with ∆E(r) = E2 (r) − E1 (r). We have seen in Sec. 4.3 that the temperature dependence of the singlet free energy is quite small. In any case it is considerably smaller than the temperature dependence of the averaged free energy. Therefore the contribution of the excited states to G1 (r, T ) is quite small and it is reasonable to assume that c2 (r) ≪ 1. From the analysis of the multipole expansion we also expect that at small distances, c2 (r) ∼ (rΛQCD )4 .46 Thus, the temperature dependence of F3 (r, T ) and its deviation from the hybrid state E2 (r) is due to small deviation of c1 (r) from unity. At low temperatures, when ∆E ≫ T these small deviations are amplified by the exponential factor. This can be easily verified by subtracting the correction T ln(1 + 13 (1 − c1 )e∆E/T ) from the triplet free energy and assuming that E1 (r) is given by the ground state potential and E2 (r) is given by the first hybrid potential as calculated in Ref. 23. The numerical results are summarized in Fig. 6 which shows that after this correction is accounted for in the confined phase the triplet free energy at low temperatures agrees reasonably well with the first hybrid potential. As temperature increases more excited states contribute. In particular, at 0.76Tc the value of the triplet free energy can be accounted for by including the
Quarkonium at Finite Temperature
77
6.5 6 1.90Tc 1.27Tc 0.95Tc 0.76Tc 0.63Tc 0.48Tc 0.42Tc T=0
F3(r,T)/σ1/2
5.5 5 4.5 4 3.5 3 0
0.5
1
1.5
2
2.5
3
1/2
rσ
Fig. 6. The triplet free energy at different temperatures calculated at β = 2.5. The filled symbols correspond to calculations in Coulomb gauge. Also shown is the first hybrid potential calculated in Ref. 23.
next hybrid state.23 However, at 0.95Tc there are large temperature effects, which cannot be explained by including the contribution from only few excited states. In Fig. 6 we also show the triplet free energy above the deconfinement temperature compared to the calculations in Coulomb gauge.18 It turns out to be much smaller than in the confined phase and agrees well with Coulomb gauge results. This means that the small deviation of the overlap factor c1 (r) from unity is unimportant in this case. The triplet free energy monotonically decreases with increasing temperature as expected in HTL perturbation theory (c.f. Eq. (25)). In the limit of high temperatures and short distances, r ≪ 1/T we have E2 (r) = αs /(4r), ∆E(r) = αs /r, c2 (r) ≃ 0 and c1 (r) = 1 + O(α3s ). Therefore we can expand the logarithm in Eq. (35) to get F3 (r, T ) = +
1 αs + O(α3s T ) + O(αs mD ). 4 r
(36)
Thus the correction due to c1 (r) 6= 0 is much smaller than the expected leading order thermal effects in the triplet free energy. The strong temperature dependence of the adjoint free energy has been also observed in SU (3) gauge theory7 and in full QCD.33,34 Although the adjoint free energy is strongly temperature dependent even at short distances its rdependence is almost the same as rdependence of the singlet free energy. The derivative of the adjoint free energy with respect to r is identical within errors to that of the singlet free energy at short distances up to the Casimir factor, i.e. Fa′ (r, T )/F1′ (r, T ) = −1/(N 2 − 1).7
78
A. Bazavov, P. Petreczky & A. Velytsky
5. Quarkonium Spectral Functions 5.1. Meson correlators and spectral functions Now we focus the discussion on the relation between the Euclidean meson correlators and spectral functions at finite temperature. The zero temperature limit is straightforward. Most dynamic properties of the finite temperature system are incorporated in the spectral function. The spectral function σH (p0 , ~p) for a given mesonic channel H in a system at temperature T can be defined through the Fourier transform of the real time two point functions D> and D< or equivalently as the imaginary part of the Fourier transformed retarded correlation function,47 1 < (D> (p0 , ~p) − DH (p0 , p~)) 2π H 1 R = ImDH (p0 , ~p) π Z d4 p ip·x >(( (x0 , ~x) = hJH (x0 , ~x), JH (0, ~0)i DH
< DH (x0 , ~x) = hJH (0, ~0), JH (x0 , ~x)i, x0 > 0
(38)
In essence σH is the Fourier transformation of the thermal average of the commutator [J(x), J(0)]. In the present paper we study local meson operators of the form (c.f. (12) with U = I) JH (t, x) = q¯(t, x)ΓH q(t, x)
(39)
with q a continuous realtime fermion position operator and ΓH = 1, γ5 , γµ , γ5 γµ , γµ γν
(40)
for scalar, pseudoscalar, vector, axialvector and tensor channels. The relation of these quantum number channels to different meson states is given in Table 1. Table 1. Meson states in different channels.
Γ
2S+1
γ5
1
γs
3
γs γs′
1
1 γ5 γs
LJ
JPC
uu
cc(n = 1)
cc(n = 2)
bb(n = 1)
bb(n = 2)
ηc
ηb
ηb′
ψ′
Υ(1S)
Υ(2S)
′
S0
−+
0
π
ηc
S1
1−−
ρ
J/ψ
P1
+−
1
b1
hc
hb
3
P0
0++
a0
χc0
χb0 (1P )
χb0 (2P )
3
P1
1++
a1
χc1
χb1 (1P )
χb1 (2P )
χc2
χb2 (1P )
χb2 (2P )
++
2
Quarkonium at Finite Temperature >( < DH (x0 , ~x) = DH (x0 + i/T, ~x).
Inserting a complete set of states and using Eq. (41), one gets the expansion (2π)2 X −En /T (e ± e−Em /T ) σH (p0 , p~) = Z m,n × hnJH (0)mi2 δ 4 (pµ − kµn + kµm )
(41)
(42)
where Z is the partition function, and k n(m) refers to the fourmomenta of the state n(m)i. A stable mesonic state contributes a δ functionlike peak to the spectral function: σH (p0 , ~ p) = h0JH Hi2 ǫ(p0 )δ(p2 − m2H ),
(43)
where mH is the mass of the state and ǫ(p0 ) is the sign function. For a quasiparticle in the medium one gets a smeared peak, with the width being the thermal width. As one increases the temperature the width increases and at sufficiently high temperatures, the contribution from the meson state in the spectral function may be sufficiently broad so that it is not very meaningful to speak of it as a well defined state any more. The spectral function as defined in Eq. (42) can be directly accessible by high energy heavy ion experiments. For example, the spectral function for the vector current is directly related to the differential thermal cross section for the production of dilepton pairs:48 1 5α2em dW σV (p0 , ~p). (44) = 3 dp0 d p p~=0 27π 2 p20 (ep0 /T − 1)
Then presence or absence of a bound state in the spectral function will manifest itself in the peak structure of the differential dilepton rate. In finite temperature lattice calculations, one calculates Euclidean time propagators, usually projected to a given spatial momentum: Z GH (τ, p~) = d3 xei~p.~x hTτ JH (τ, ~x)JH (0, ~0)i (45) > This quantity is an analytical continuation of DH (x0 , ~p) > GH (τ, p~) = DH (−iτ, ~p).
(46)
Using this equation and the KMS condition one can easily show that GH (τ, p~) is related to the spectral function, Eq. (37), by an integral equation (see e.g. appendix B of Ref. 49): Z ∞ GH (τ, p~) = dωσ(ω, p~)H K(ω, τ ) 0
K(ω, τ ) =
cosh(ω(τ − 1/2T )) . sinh(ω/2T )
(47)
80
A. Bazavov, P. Petreczky & A. Velytsky
This equation is the basic equation for extracting the spectral function from meson correlators. Equation (47) is valid in the continuum. Formally the same spectral representation can be written for the Euclidean correlator calculated on the lattice Glat ~). The corresponding spectral function, however, will be distorted by the H (τ, p effect of the finite lattice spacing. These distortions have been calculated in the free theory.50,51 When discussing the numerical results in following sections the subscript H denoting different channels for meson correlators and spectral functions will be omitted. 5.2. Lattice formulations for charmonium physics The quarkonium system at zero and finite temperature was studied in Refs. 52–55 using Wilsontype fermions # " X X c SW Wilson W ilson ¯ σµν Fµν ψ(x), (48) Sq = ψ(x) m0 + D 6 − 2 µ,ν x where the Dirac operator is defined as 1 DµWilson = ∇µ − γµ ∆µ 2
(49)
with 1 Uµ (x)ψ(x + µ) − Uµ† (x − µ)ψ(x − µ) 2 ∆µ ψ(x) = Uµ (x)ψ(x + µ) + Uµ† (x − µ)ψ(x − µ) − 2ψ(x) .
∇µ ψ(x) =
(50)
Furthermore, σµ,ν = {γµ , γν } and the field strength tensor is defined as i Fµν (x) = − [Qµν − Q†µν ] 2
(51)
4Qµν (x) = Uµ (x)Uν (x + µ ˆ)Uµ† (x + νˆ)Uν† (x) + Uν (x)Uµ† (x − µ ˆ + νˆ)Uν† (x − µ ˆ)Uµ (x − µ ˆ) + Uµ† (x − µ ˆ)Uν† (x − µ ˆ − νˆ)Uµ (x − µ ˆ − νˆ)Uν (x − νˆ) + Uν† (x − νˆ)Uµ (x − νˆ)Uν (x + µ ˆ − νˆ)Uµ† (x) .
(52)
The last term in the brackets in Eq. (48) helps to suppress O(a) lattice artifacts and is called the clover term. Formulations with cSW 6= 0 usually referred to as clover action. The standard Wilson action for fermions corresponds to cSW = 0. In the lattice literature usually the form with the hopping parameter κ = 1/(2am0 + 8) is √ used. In this form the fields need to be accordingly normalized ψ(x) → ψ(x)a−3/2 / 2κ. In Ref. 55 the anisotropic Fermilab formulation is used for the heavy quarks. The anisotropy allows to use fine temporal lattice spacings without significant increase in the computational cost. The study uses the quenched approximation and the
Quarkonium at Finite Temperature
81
standard Wilson action in the gauge sector for which the relation between the bare ξ0 and the renormalized anisotropy ξ = as /at is known in a wide range of the gauge coupling β = 6/g 2 .56 The anisotropic clover action57 is " X νs X Wilson ξ ¯ D 6 s Sq = ψ(x) m0 + νt D 6 Wilson + t ξ0 s x 1 − 2
t Csw
X s
Cs X σts Fts + sw σss′ Fss′ ξ0 s<s′
!#
ψ(x) .
(53)
Because the lattice spacings in space and time directions are different the spatial and temporal Dirac operators as well as the clover terms have different coefficients. By tuning the clover coefficients according to the Fermilab prescription57,58 it is possible to reduce O(at m0 ) discretization errors. We will refer to this formulation as anisotropic Fermilab formulation. In the following we will show results from Ref. 55, where anisotropic Fermilab formulation was used as well as results from the study that used isotropic clover action with nonperturbatively determined values of cSW .53 In the study with anisotropic Fermilab action two values for the renormalized anisotropy ξ = 2, 4 as well as β = 5.7, 5.9, 6.1, 6.5 were used. These parameters correspond to temporal lattice spacings a−1 t = 1.905−14.12 GeV. To set the scale for the lattice spacing the traditional phenomenological value r0 = 0.5 fm for the Sommer scale38 was used. The Sommer scale r0 has also been calculated for anisotropic Wilson action for β = 5.5 − 6.1.59 Alternatively one can estimate the lattice spacing from the difference between the mass of 1 P1 state and the spin averaged 1S mass: ∆M (1 P1 − 1S). To a very good approximation this mass difference is not affected by fine and hyperfine splitting and thus is not very sensitive to quenching errors. It was found that close to the continuum limit the lattice spacing determined from ∆M (1 P1 − 1S) is different from that determined from r0 by 10%57,60 if the phenomenological value r = 0.5 fm is used. Using the value of r0 = 0.469(7) determined in full QCD61 would give a value for ∆M (1 P1 − 1S) splitting which is closer to the experimental one, however, the ∆M (2S − 1S) splitting would be even further away from the experimental value.60 This problem is due to the quenched approximation. In the studies with isotropic clover action the following values of the lattice gauge couplings have been used β = 6.499, 6.640 and 7.192.53,62–64 These correspond to √ lattice spacings a−1 = 4.04, 4.86 and 9.72 GeV respectively if the value σ = 420MeV for the string tension is used. The continuum meson current in Eq. (39), JH is related to the lattice current as ¯ H ψ, JH = ZH a3s ψΓ
(54)
where ψ is the lattice quark field in Eq. (48). The renormalization constant ZH
82
A. Bazavov, P. Petreczky & A. Velytsky
can be calculated in perturbation theory or nonperturbatively. For isotropic clover action this has been done with both methods (see Ref. 53 and references therein). 5.3. Bayesian analysis of meson correlators The obvious difficulty in the reconstruction of the spectral function from Eq. (47) is the fact that the Euclidean correlator is calculated only at O(10) data points on the lattice, while for a reasonable discretization of the integral in Eq. (47) we need O(100) degrees of freedom. The problem can be solved using Bayesian analysis of the correlator, where one looks for a spectral function which maximizes the conditional probability P [σDH] of having the spectral function σ given the data D and some prior knowledge H (for reviews see Refs. 65 and 66). Different Bayesian methods differ in the choice of the prior knowledge. One version of this analysis which is extensively used in the literature is the Maximum Entropy Method (MEM).67,68 It has been used to study different correlation functions in Quantum Field Theory at zero and finite temperature.52–54,62,65,68–79 In this method the basic prior knowledge is the positivity of the spectral function and the prior knowledge is given by the ShannonJanes entropy Z σ(ω) . S = dω σ(ω) − m(ω) − σ(ω) ln m(ω) The real function m(ω) is called the default model and parametrizes all additional prior knowledge about the spectral functions, e.g. such as the asymptotic behavior at high energy.65,68 For this case the conditional probability becomes 1 2 (55) P [σDH] = exp − χ + αS , 2 with χ2 being the standard likelihood function and α a real parameter. Previously in the MEM analysis of the meson spectral functions the Bryan’s algorithm was used.67 A new algorithm was introduced in Ref. 55. It is worth to make connection between this method and the Bryan algorithm. In both cases the true problem is numberofdata dimensional — in more dimensions the problem would be underdetermined. To find the relevant subspace, the Bryan algorithm uses singular value decomposition, while the new algorithm finds the same relevant subspace by exact mathematical transformations. Although the method of identifying the subspace is different, the result is the same, and in both cases one proceeds with solving the original problem in this restricted subspace. The advantage of the new algorithm is that it is more stable numerically when one reconstructs quarkonium spectral functions at zero temperature. The comparison of the two algorithms was done for the pseudoscalar spectral function for β = 6.1, ξ = 4.55 The problem with the Bryan algorithm is that it does not work well for charmonium correlators if the time extent is sufficiently large, which is the case at low temperatures; the iterative procedure does not always
Quarkonium at Finite Temperature
83
converge. For instance at β = 6.1, ξ = 4 and 163 × 96 lattice it was possible to get the spectral functions using the Bryan algorithm only when using τmax = 24 data points in the time direction. With the new algorithm there is no restriction on τmax which can be as large as Nτ /2. 5.4. Charmonium spectral functions at zero temperature 5.4.1. Pseudoscalar and vector spectral functions at zero temperature In this subsection we discuss the results of Ref. 55 on charmonium spectral functions obtained using MEM. The zero temperature spectral functions for three different lattice spacings obtained by the new method from Ref. 55 are summarized in Fig. 7. Here the simple default model m(ω) = 1 (in units of the spatial lattice spacing) is used. To get a feeling for the statistical errors in the spectral functions its mean value in some interval I is calculated: R dωσ(ω) . (56) σ ¯ = IR dω I
Then the error on σ ¯ is calculated using standard jackknife method. These errors are shown in Fig. 7, where the length of the intervals are shown as horizontal error bars. As one can see from the figure, the ηc (1S) can be identified very well. The second peak is likely to correspond to excited states. Because of the heavy quark mass the splitting between different radial excitations is small and MEM cannot resolve different excitations individually but rather produces a second broad peak to which all radial excitation contribute. This can be seen from the fact that the amplitude 0.4 0.35
ηc(1S)
at1= 4.11GeV,ξ=2
0.3
at1= 8.18GeV,ξ=4
0.25
at =14.12GeV,ξ=4
σ(ω)/ω2
1
0.2 0.15 0.1
2S+3S+4S
0.05 0 5
10
15
20
ω[GeV] Fig. 7. The pseudoscalar spectral function at zero temperature for three finest lattice spacings. The horizontal line corresponds to the spectral function in the free massless limit at zero lattice spacing.
84
A. Bazavov, P. Petreczky & A. Velytsky
0.4 0.1 0.35 m(ω)/ω2
0.08
0.3 σ(ω)/ω2
0.25
0.06 0.04 0.02
0.2
0 0
0.15
5
10 15 20 25 30 ω [GeV]
0.1 0.05 0 5
10
15
20
ω [GeV] Fig. 8. The default model dependence of the pseudoscalar spectral function at the finest lattice spacing (β = 6.5). In the inset the default models corresponding to different spectral functions are shown.
of the second peak (i.e. the area under the peak) is more than two times larger than the first one. Physical considerations tell us that it should be smaller than the first amplitude if it was a 2S state. When comparing amplitudes and peak positions from MEM analysis and from double exponential fits a very good agreement for the first peak and a fair agreement for the second peak are found. This gives confidence that at zero temperature charmonium properties can be reproduced well with MEM. For energies larger than 5 GeV one probably sees a continuum in the spectral functions which is distorted by finite lattice spacing. In particular the spectral function is zero above some energy which scales roughly as a−1 s . Note that for ω < 5 GeV the spectral function does not depend on the lattice spacing. One should control how the result depends on the default model. In Fig. 8 we show the spectral function for three different default models. One can see that the default model dependence is significant only for ω > 5 GeV. This is not surprising as there are very few time slices which are sensitive to the spectral functions at ω > 5 GeV, while the first peak is well determined by the large distance behavior of the correlator. The spectral function in the vector channel defined as 1X σii (ω), (57) σV (ω) = 3 i was also calculated and is shown in Fig. 9 for the three finest lattice spacings. The conclusions which can be derived from this figure are similar to the ones discussed above for the pseudoscalar channel. The first peak corresponds to the J/ψ(1S) state, the second peak most likely is a combination of 2S and higher excited states,
Quarkonium at Finite Temperature
85
0.25 J/ψ(1S)
at1= 4.11GeV,ξ=2
0.2
at1= 8.18GeV,ξ=4
σ(ω)/ω2
at1=14.12GeV,ξ=4 0.15
0.1 2S+3S+4S 0.05
0 5
10
15
20
ω[GeV] Fig. 9. The vector spectral function at zero temperature for three finest lattice spacings. The horizontal line is the spectral function in the free massless limit.
finally there is a continuum above 5 GeV which is, however, distorted by lattice artifacts. The similarity between the pseudoscalar and vector channel is, of course, expected. The lower lying states in theses channels differ only by small hyperfine splitting.
5.4.2. Spectral functions for P states The spectral functions in the scalar, axialvector and tensor channels which have the 1P charmonia as the ground state were also calculated in Ref. 55. The scalar spectral functions reconstructed using MEM are shown in Fig. 10. The first peak corresponds to χc0 state, but it is not resolved as well as the ground state in the pseudoscalar channel. This is due to the fact that the scalar correlator is considerably more noisy than the pseudoscalar or vector correlator. This can be understood as follows. For the heavy quark mass the contribution of the ground state in the scalar channel is suppressed as 1/m2 relative to the ground state contribution in the pseudoscalar and vector channels, and therefore it is considerably smaller than the continuum contribution to the scalar correlator. For the two finest lattice spacings there is a second peak which may correspond to a combination of excited P states. Above ω > 5 GeV we see a continuum which is strongly distorted by lattice artifacts and probably also by MEM. The spectral functions in the axialvector and tensor channels are shown in Fig. 11. They look similar to the scalar spectral functions. As in the scalar channel the first peak is less pronounced than in the case of Swave charmonium spectral functions, and it corresponds to χc1 and hc state, respectively. The continuum part of the spectral function is again strongly distorted.
December 16, 2009 8:43 WSPC/INSTRUCTION FILE
86
qgp4˙bpv
A. Bazavov, P. Petreczky & A. Velytsky
0.09
σ(ω)/ω2
0.08
χc0(1P)
at1= 4.11GeV,ξ=2
0.07
at1= 8.18GeV,ξ=4
0.06
at1=14.12GeV,ξ=4
0.05 0.04 0.03 0.02 0.01 0 5
10
15
20
ω[GeV] Fig. 10. The scalar spectral function at zero temperature for three finest lattice spacings.
Fig. 11. The axialvector (left) and tensor (right) spectral functions at zero temperature for three lattice spacings.
5.5. Charmonium correlators at finite temperature The reconstruction of the spectral functions from lattice correlators is difficult already at zero temperatures. At finite temperature it is even more difficult to control the systematic errors in the spectral functions reconstructed from MEM. This is because with increasing temperature the maximal time extent τmax is decreasing as 1/T . Also the number of data points available for the analysis becomes smaller. Therefore other methods which can give some information about the change of the spectral functions as the temperature is increasing are often used. The temperature dependence of the spectral function will manifest itself in the temperature dependence of the lattice correlator G(τ, T ). Looking at Eq. (47) it is easy to see that the temperature dependence of G(τ, T ) comes from the temperature dependence of the
Quarkonium at Finite Temperature
87
G/Grecon
1 0.95 β=6.1
0.9
G/Grecon
1 0.95 β=6.5
0.9 0.1
0.2
0.3
0.87Tc 1.07Tc 1.16Tc 1.39Tc 1.73Tc 2.31Tc 1.09Tc 1.20Tc 1.50Tc 1.99Tc 2.39Tc 2.99Tc 0.4
0.5
τ [fm] Fig. 12. The ratio G/Grecon for the pseudoscalar channel for the two finer ξ = 4 lattices.
spectral function σ(ω, T ) and the temperature dependence of the kernel K(τ, ω, T ). To separate out the trivial temperature dependence due to K(τ, ω, T ) one calculates the reconstructed correlator53 Z ∞ Grecon (τ, T ) = dωσ(ω, T = 0)K(τ, ω, T ). (58) 0
If the spectral function does not change with increasing temperature we expect G(τ, T )/Grecon (τ, T ) = 1. In Ref. 55 an extensive study of the temperature dependence of this ratio for different channels at different lattice spacings was carried out. 5.5.1. The pseudoscalar correlators First we present results from Ref. 55 for the temperature dependence of the pseudoscalar correlators. In Fig. 12 we show numerical results for G/Grecon on lattices with ξ = 4. We see almost no change in the pseudoscalar correlator till temperatures as high as 1.2Tc. The temperature dependence of the pseudoscalar correlator remains small for temperatures below 1.5Tc. Medium modifications of the correlator slowly turn on as we increase the temperature above this value. From the figures it is clear that the temperature dependence of the correlators is not affected significantly by the finite lattice spacing. The very small temperature dependence of the pseudoscalar correlator suggests that the corresponding ground state ηc (1S) may survive till temperatures as high a 1.5Tc. The temperature dependence of the correlator found in this study is similar to that of Ref. 53, where isotropic lattices
88
A. Bazavov, P. Petreczky & A. Velytsky
with very small lattice spacings, a−1 = 4.86, 9.72 GeV have been used. However, at temperatures higher than 1.5Tc the deviations of G/Grecon from unity become slightly larger than those found in Ref. 53. This is possibly due to the fact that cutoff effects are more important at higher temperatures. Thus despite similarities of the temperature dependence of the pseudoscalar correlator to findings of Ref. 53 there are quantitative differences. One should note, however, statistical errors and systematic uncertainties are larger in the analysis presented in Ref. 53 than in Ref. 55. In the study on isotropic lattices the ratio G/Grecon starts to depend more strongly on the temperature only around 3Tc.53 5.5.2. The Pwave correlators Next we present the temperature dependence of the scalar, axialvector and tensor correlators corresponding to P states. The τ dependence of the scalar correlator was studied on ξ = 2, 4 lattices.55 The numerical results on fine lattices with ξ = 4 are shown in Fig. 13. We see some differences in G/Grecon calculated at β = 6.1 and β = 6.5. Thus the cutoff dependence of G/Grecon is larger in the scalar channel than in the pseudoscalar one. For β = 6.1 and ξ = 4 the calculations were done on 243 × 24 lattice to check finite volume effects. The corresponding results are shown in Fig. 13 indicating that the finite volume effects are small. On the finest lattice the enhancement of the scalar correlator is very similar to that found in calculations done on isotropic lattices,53 but small quantitative differences can be identified. In Figs. 14 and 15 we show the temperature dependence of the axialvector and tensor correlators respectively for ξ = 4. Qualitatively their behavior is very similar to the scalar correlator but the enhancement over the zero temperature result is larger. The results for the axialvector correlators again are very similar to those published in Ref. 53. The difference in G/Grecon calculated at β = 6.1 and β = 6.5 are smaller than in the scalar channel. The large increase in the scalar, axialvector and tensor correlators may be interpreted as indicator of strong modification of the corresponding spectral function and, possibly the dissolution of 1P charmonia states. However, as we will see in the next sections the situation is more complicated. It has been noticed in Ref. 80 that the increase in G/Grecon may be due to the zero mode contribution. 5.5.3. The vector correlator The numerical results for the vector correlator of Ref. 55 are shown in Fig. 16 for ξ = 4. As one can see from the figures the temperature dependence of G/Grec is different from the pseudoscalar case and this ratio is larger than unity for all lattice spacings. Similar results have been obtained on isotropic lattices.63 The enhancement of the vector correlator is due to the presence of the transport contribution in the spectral function49,81 and will be discussed in Sec. 5.10. Since the vector current is conserved the vector correlator also caries information about
Quarkonium at Finite Temperature
0.87Tc 1.07Tc 1.16Tc Ns=24 1.16Tc 1.39Tc 1.73Tc β=6.1 2.31Tc
G/Grecon
2.2 1.9 1.6 1.3
G/Grecon
1 2.2 1.9 1.6 1.3
β=6.5
1 0.1
0.2
0.3 τ [fm]
1.09Tc 1.20Tc 1.50Tc 1.99Tc 2.39Tc 2.99Tc 0.4
0.5
Fig. 13. The ratio G/Grecon for the scalar channel for the two finer ξ = 4 lattices.
G/Grecon
3 2.5 β=6.1
2 1.5
G/Grecon
1 3 2.5 2 1.5
β=6.5
1 0.1
0.2
0.3 τ [fm]
0.87Tc 1.07Tc 1.16Tc 1.39Tc 1.73Tc 2.31Tc 1.09Tc 1.20Tc 1.50Tc 1.99Tc 2.39Tc 2.99Tc 0.4
0.5
Fig. 14. The ratio G/Grecon for the axialvector channel for ξ = 4 lattices.
89
A. Bazavov, P. Petreczky & A. Velytsky
β=6.1
G/Grecon
3 2.5 2 1.5
0.87Tc 1.07Tc 1.16Tc 1.39Tc 1.73Tc 2.31Tc
1 G/Grecon
3 2.5 2 1.5
β=6.5
1 0.1
0.2
0.3 τ [fm]
1.09Tc 1.20Tc 1.50Tc 1.99Tc 2.39Tc 2.99Tc 0.4
0.5
Fig. 15. The ratio G/Grecon for the tensor channel for ξ = 4 lattices.
G/Grecon
β=6.1
G/Grecon
90
1.06 1.03 1
1.06 1.03 1
β=6.5 0.1
0.2
0.3 τ [fm]
0.87Tc 1.07Tc 1.16Tc 1.39Tc 1.73Tc 2.31Tc 1.09Tc 1.20Tc 1.50Tc 1.99Tc 2.39Tc 2.99Tc 0.4
0.5
Fig. 16. The ratio G/Grecon for the vector channel for the two finer ξ = 4 lattices.
Quarkonium at Finite Temperature
91
the transport of heavy quarks in the plasma. This shows up as a peak in the spectral function at zero energy, leading to the observed enhancement in G/Grecon . 5.6. Charmonium spectral functions at finite temperature In Sec. 5.4 we have seen that using MEM one can reconstruct well the main features of the spectral function, in particular the ground state properties. At finite temperature the situation becomes worse because the temporal extent is decreasing. The maximal time separation is τmax = 1/(2T ). As a consequence it is no longer possible to isolate the ground state well. Also the number of available data points becomes smaller. While the later limitation can be overcome by using smaller and smaller lattice spacings in time direction the former limitation is always present. Therefore we should investigate systematic effects due to limited extent of the temporal direction. It appears that the pseudoscalar channel is the most suitable case for this investigation as at zero temperature it is well under control and there is no contribution from heavy quark transport. To estimate the effect of limited temporal extent in Ref. 55 the spectral function at zero temperature is calculated considering only the first Ndata timeslices in the analysis for β = 6.5, ξ = 4. The result of this calculation is shown in Fig. 17 where Ndata = 80, 40, 20 and 16. The last two values correspond to the finite temperature lattices in the deconfined phase. In this case we see the first peak quite clearly. As one can see from the figure already for Ndata = 40 and τmax = 0.56 fm the second peak corresponding to radial excitation is no longer visible and the first peak becomes significantly broader. The position of the first peak, however, is unchanged. As the number of data points is further decreased (Ndata = 20, 16) we see further broadening of the first peak and a small shift of the peak position to higher energies. These systematic effects should be
Ndata= 80, τmax=1.12fm
0.12
Ndata= 40, τmax=0.56fm
σ(ω)/ω2
0.1
Ndata= 20, τmax=0.28fm Ndata= 16, τmax=0.22fm
0.08 0.06 0.04 0.02 0 5
10
15
20
ω[GeV] Fig. 17. The dependence of the reconstructed pseudoscalar spectral function on the maximal temporal extent for β = 6.5. In the analysis the default model m(ω) = 1 has been used.
92
A. Bazavov, P. Petreczky & A. Velytsky
taken into account when analyzing the spectral functions at finite temperature. Therefore when studying the spectral functions at finite temperature we always compare with the zero temperature spectral functions reconstructed with the same number of data points and τmax as available at that temperature. In Fig. 18 the spectral functions at different temperatures are shown together with the zero temperature spectral functions.55 As a default model for T = 1.2Tc and T = 1.5Tc m(ω) = 0.01 is used. For T = 2.0Tc m(ω) = 1 is used since the use of m(ω) = 0.01 resulted in numerical problems in the MEM analysis. The figure shows that the pseudoscalar spectral function is not modified till 1.5Tc within the errors of the calculations. This is consistent with the conclusions of Ref. 54, 53. One should note, however, that it is difficult to make any conclusive statement based on the shape of the spectral functions as this was done in the above mentioned works. The dependence of the reconstructed spectral functions on the default model m(ω) is much stronger at finite temperature. The spectral functions were reconstructed using different types of default models. For all temperatures T ≤ 1.5Tc the difference between the finite temperature spectral function and the zero temperature one is very small compared to the statistical errors for all default models considered here. In particular a use is made of the default models constructed from the high energy part of the lattice spectral functions calculated at zero temperature as this was done in Ref. 53. The idea is that at sufficiently high energy the spectral function is dominated by the continuum and is temperature independent. Therefore it is suitable to provide the prior knowledge, i.e. the default model. With this default model the spectral functions were calculated at T = (1.07 − 1.5)Tc .55 Very little temperature dependence of the spectral functions was found, see Fig. 19. Note, however, that for this choice of the default model no clear peak can be identified in the spectral functions if T ≥ 1.2Tc. The spectral function in the vector channel is also calculated.55 The results are shown in Fig. 20 for the default model m(ω) = 0.01. As this was already discussed in the previous section the basic difference between the pseudoscalar and vector spectral functions at finite temperature is the presence of the transport peak at ω ≃ 0. The difference of the temperature dependence of the vector and pseudoscalar correlators is consistent with this assumption. The vector spectral function reconstructed with MEM shows no evidence of the transport peak at ω ≃ 0. On the other hand the spectral function at 1.2Tc differs from the zero temperature spectral function, in particular the first peak is shifted to smaller ω values. We believe that this is a problem of the MEM analysis which cannot resolve the peak at ω ≃ 0 but instead mimics its effect by shifting the J/ψ peak to smaller ω. Also at 2.4Tc the spectral function extends to smaller ω values than in the pseudoscalar correlator which again indicates some structure at ω ≃ 0. The analysis of the vector spectral functions using other choices for the default model always indicates that the spectral functions at finite temperature differs from the zero temperature spectral functions and extend to significantly smaller ω values.
Quarkonium at Finite Temperature
93
0.14 T=0, Ndata=20
0.12
T=1.2Tc
σ(ω)/ω2
0.1 0.08 0.06 0.04 0.02 0 5
10
15
20
ω[GeV] 0.14 T=0, Ndata=16
0.12
T=1.5Tc
σ(ω)/ω2
0.1 0.08 0.06 0.04 0.02 0 5
10
15
20
ω[GeV] 0.07 T=0, Ndata=12
0.06
T=2.0Tc
σ(ω)/ω2
0.05 0.04 0.03 0.02 0.01 0 0
5
10
15
20
ω[GeV]
Fig. 18. The pseudoscalar spectral function for β = 6.5 and Nt = 40, 32, 20 corresponding to temperatures 1.2Tc , 1.5Tc and 2.0Tc . In the analysis the default model m(ω) = 0.01 has been used for 1.2Tc and 1.5Tc , while for 2.0T c we used m(ω) = 1.
94
A. Bazavov, P. Petreczky & A. Velytsky
0.07
0.07 T=0, Ndata=13
0.06
σ(ω)/ω2
σ(ω)/ω2
T=1.16Tc
0.05
0.04 0.03
0.04 0.03
0.02
0.02
0.01
0.01
0
0 0
2
4
6
8 10 ω[GeV]
12
14
16
18
0
0.07
2
4
6
8 10 ω[GeV]
12
14
16
18
0.07 T=0, Ndata=20
0.06
T=0, Ndata=16
0.06
T=1.2Tc
T=1.5Tc
0.05 σ(ω)/ω2
0.05 σ(ω)/ω2
T=0, Ndata=12
0.06
T=1.07Tc
0.05
0.04 0.03
0.04 0.03
0.02
0.02
0.01
0.01
0
0 0
5
10 15 ω[GeV]
20
0
5
10 15 ω[GeV]
20
Fig. 19. The pseudoscalar spectral function at different temperatures together with the zero temperature spectral functions reconstructed using default model coming from the high energy part of the zero temperature spectral function.
5.7. Charmonium correlators and spectral functions at finite momenta So far we reviewed charmonia at zero spatial momentum, i.e. charmonia at rest in the heatbath’s rest frame. It is certainly of interest to study the temperature dependence of correlators and spectral functions at nonzero spatial momentum. Such a study has been done using isotropic lattices with lattice spacing a−1 = 4.86 GeV and 9.72 GeV.63 It has been found that the pseudoscalar correlators are enhanced compared to the zero temperature correlators for nonvanishing spatial momenta, see Fig. 21. Furthermore, the enhancement of vector correlator at finite spatial momentum is larger than at zero spatial momentum. In Ref. 55 the finite momentum pseudoscalar correlators are calculated on anisotropic lattices at β = 6.1, ξ = 4 for different temperatures. The differences in G/Grecon calculated in this work and in Refs. 53 and 63 are present already at zero momentum and are presumably due to finite lattice spacing errors. Apart from this the momentum dependence of the pseudoscalar correlators is similar to the findings of Refs. 53 and 63. It would be interesting to see if the difference in the temperature dependence of the correlators at zero and finite spatial momenta is due to a contribution to the spectral functions below the light cone at finite temperature.50,51
Quarkonium at Finite Temperature
95
0.14 0.12 T=0, Ndata=20
σ(ω)/ω2
0.1
T=1.2Tc
0.08 0.06 0.04 0.02 0 0
5
10
15
20
ω[GeV] 0.14 0.12 T=0, Ndata=10
σ(ω)/ω2
0.1
T=2.4Tc
0.08 0.06 0.04 0.02 0 0
5
10
15
20
ω[GeV] Fig. 20. The vector spectral function for β = 6.5 and Nt = 40, 20 corresponding to temperatures 1.2Tc and 2.4Tc . In the analysis the default model m(ω) = 0.01 has been used.
5.8. Bottomonium spectral functions at zero temperature The use of Fermilab formulation described in the previous sections allows for a study of bottomonium for the same range of lattice spacings. Usually bottomonium is studied using lattice NRQCD (see e.g. Ref. 82). First study of bottomonium within the relativistic framework was presented in Ref. 83. More recently it was studied in Ref. 55, where as before the lattice spacing is fixed by the Sommer scale r0 , assuming r0 = 0.5 fm. Note that the lattice spacings determined from r0 are about 20% smaller than in Ref. 83 where it was determined from bottomonium 1 P1 − 1S mass splitting. As the result the new estimates of the Υ mass are smaller than those in Ref. 83. However, one finds good agreement if the values of the lattice spacing quoted in Ref. 83 are used to calculate the physical masses.
96
A. Bazavov, P. Petreczky & A. Velytsky
1.4 G(τ)/Grecon(τ)
0 0.64 1.28 1.92
ηc
1.5Tc
G(τ)/Grecon(τ)
1.6
p[GeV]
p[GeV]
(+0.2)
0 0.64 1.28 1.92
J/ψT
1.5Tc (+0 3)
1.2 1.3
1.12Tc
1.1Tc 1
1
0.1
0.2
0.3
0.05
0.1
τ[fm]
0.15
0.2 τ[fm]
0.25
0.3
0.35
Fig. 21. Comparison of correlators at 1.1 Tc and 1.5 Tc for pseudoscalar and (transverse) vector charmonia with those reconstructed from spectral function at 0.75 Tc .
0.16
ηb(1S)
0.14
at1= 5.88GeV,ξ=2 at1= 8.18GeV,ξ=4
σ(ω)/ω2
0.12
at1=10.89GeV,ξ=4
0.1 0.08 0.06 0.04 0.02 0 8
10
12
14
16
18
20
22
ω[GeV] Fig. 22. The pseudoscalar bottomonium spectral function at zero temperature for different lattice spacings.
Using MEM the spectral functions in different channels for three lattice spacings were analyzed. In Fig. 22 we show the spectral functions in the pseudoscalar channel. Since the physical quark mass is different at different lattice spacings the horizontal scale was shifted by the difference of the calculated Υmass and the corresponding experimental value. We can see that the first peak in the spectral function corresponds to the ηb (1S) state and its position is independent of the lattice spacing. The remaining details of the spectral functions are cutoff dependent and we cannot distinguish the excited states from the continuum. The position and the amplitude of the first peak in the spectral functions is in good agreement with the results of simple exponential fit. As in the charmonium case the maximal
Quarkonium at Finite Temperature
97
0.1 at1= 5.88GeV, ξ=2
σ(ω)/ω2
0.08
at1= 8.18GeV, ξ=4
χb(1P)
at1=10.89GeV, ξ=4
0.06
0.04
0.02
0 8
10
12
14 16 ω[GeV]
18
20
22
Fig. 23. The scalar bottomonium spectral function at zero temperature for different lattice spacings.
energy ωmax for which the spectral function is nonzero scales approximately as a−1 s . Similar results have been obtained in the vector channel. The spectral function in the scalar channel is shown in Fig. 23. As it was the case for charmonium the correlators in this channel are more noisy than in the pseudoscalar channel and as a result it is more difficult to reconstruct the spectral function. Nevertheless we are able to reconstruct the χb0 state which is the first peak in the spectral function. The peak position and the amplitude are in reasonable agreement with the result of the simple exponential fit. 5.9. Bottomonium at finite temperature Besides calculating the bottomonium spectral function at zero temperature Ref. 55 presents a study of the temperature dependence of bottomonium correlators to see medium modification of bottomonia properties. In Fig. 24 we show G/Grecon for vector and pseudoscalar channel at different lattice spacings. This ratio appears to be temperature independent and very close to unity up to quite high temperatures. This is consistent with the expectation that 1S bottomonia are smaller than 1S charmonia and thus are less affected by the medium. They could survive till significantly higher temperatures. Compared to charmonium case the difference between the pseudoscalar and vector channels is smaller. This is also expected as the transport contribution which is responsible for this difference is proportional to ∼ exp(−mc,b /T ), and thus is much smaller for bottom quarks (see the discussion in the next section). Similar temperature dependence of the pseudoscalar bottomonium correlator has been found in calculations with isotropic clover action.84 The temperature dependence of the scalar correlator is shown in Fig. 25. Contrary to the pseudoscalar and vector correlators it shows strong temperature
A. Bazavov, P. Petreczky & A. Velytsky
G/Grecon
98
1.01 1 β=5.9
G/Grecon
0.99
1.16Tc 1.39Tc 1.73Tc
1.01 1 β=6.1
0.99 G/Grecon
1.25Tc 1.66Tc 2.49Tc 3.32Tc
1.01 1
β=6.3
0.99 0.1
0.2
1.15Tc 1.54Tc 1.85Tc 2.31Tc
0.3
τ [fm] G/Grecon
1.02
G/Grecon
1 1.02
G/Grecon
1 1.02
β=5.9
1.25Tc 1.66Tc 2.49Tc 3.32Tc
β=6.1
1.16Tc 1.39Tc 1.73Tc
β=6.3
1.15Tc 1.54Tc 1.85Tc 2.31Tc
1 0.1
0.2
0.3
τ [fm] Fig. 24. The ratio G/Grecon in the pseudoscalar (top) and vector channels at different lattice spacings.
dependence and G/Grecon is significantly larger than unity already at 1.1Tc . Again, similar enhancement in G/Grecon has been observed in isotropic lattice calculations.84 The spectral functions were reconstructed at finite temperature for β = 6.3. For the pseudoscalar channel the results are shown in Fig. 26. As in the charmonium case we compare the finite temperature spectral function with the zero temperature spectral function obtained with the same number of data points and time interval. As expected the spectral function shows no temperature dependence within errors. On the other hand it was not possible to reliably reconstruct the scalar spectral function at finite temperature due to numerical problems. Presumably much more statistics is needed to get some information about the scalar spectral function.
G/Grecon
Quarkonium at Finite Temperature
3
β=5.9
1.25Tc 1.66Tc 2.49Tc 3.32Tc
β=6.1
1.16Tc 1.39Tc 1.73Tc
2
99
G/Grecon
1 3 2
G/Grecon
1 1.15Tc 1.54Tc 1.85Tc 2.31Tc
β=6.3
3 2 1 0.1
0.2
0.3
τ [fm] Fig. 25. The ratio G/Grecon in the scalar channel for different lattice spacings.
0.1 T=0, Ndata=16 0.08 σ(ω)/ω
2
T=1.15Tc
0.06 0.04 0.02 0 8
10
12
14
16
18
20
22
ω[GeV] Fig. 26. The pseudoscalar bottomonium spectral function at finite temperature.
5.10. Zero modes contribution At finite temperature quarkonium spectral functions contain information about states containing a quark antiquark pair as well as scattering of the external probe off a heavy quark from the medium. The later gives a contribution to the spectral function below the light cone (ω < k). In the limit of zero momentum it becomes χi (T )ωδ(ω) in the free theory. The generalized susceptibilities χi (T ) were calculated in Ref. 51 in the free theory. Interaction with the medium leads to the broadening of the delta function, which becomes a Lorentzian with a small width.81 Because
100
A. Bazavov, P. Petreczky & A. Velytsky
the quark antiquark pair contributes to the spectral function at energies ω > 2m it is reasonable to separate quarkonium spectral function into two terms64 i i σ i (ω, T ) = σhigh (ω, T ) + σlow (ω, T ).
(59)
i Here σhigh (ω, T ) is the high energy part of the spectral functions which is nonzero only for ω > 2m and describes the propagation of bound or unbound quark i antiquark pairs. On the other hand σlow (ω, T ) receives the dominant contribution at ω ≃ 0. Because the width of the peak at ω ≃ 0 is small the later gives an almost constant contribution to the Euclidean correlator, which is called the zero mode contribution. We can write an analogous decomposition for the Euclidean correlator
Gi (τ, T ) = Gihigh (τ, T ) + Gilow (τ, T ).
(60)
To a very good approximation Gilow (τ, T ) = χi (T )T , i.e. constant. In the previous sections we used the ratio G(τ, T )/Grecon (τ, T ) to study the temperature dependence of the correlators. This dependence comes separately from the high energy part and low energy part, which gives the zero mode contribution. The zero mode contribution is absent in the derivative of the correlator with respect to τ . Therefore one can study the temperature dependence of the correlators induced by change of bound state properties and/or its dissolution by considering the ratio of the derivatives of the correlators G′ (τ, T )/G′recon (τ, T ). The temperature dependence of scalar (13) and axialvector correlators (14) has been presented in previous sections in terms of G/Grecon . It is temperature independent in the confined phase and shows large enhancement in the deconfined phase. This large enhancement is present both in charmonium and bottomonium correlators. To eliminate the zero mode contribution the derivative of the correlators and the corresponding ratio G′ (τ, T )/G′recon (τ, T ) have been calculated in Ref. 64 using isotropic lattices. The numerical results for this ratio at β = 7.192 are shown in Fig. 27. The results from anisotropic lattices55 are also shown. There is good agreement between the results obtained from isotropic and anisotropic lattices. As one can see from Fig. 27 G′ (τ, T )/G′recon (τ, T ) shows very little temperature dependence and is close to unity. This means that almost the entire temperature dependence of the scalar and axialvector correlators is due to zero mode contribution and Ghigh (τ, T ) is temperature independent. The temperature dependence of the Sstates seen in previous sections is also greatly reduced and the agreement between isotropic and anisotropic calculations is better for the ratio of derivatives. For the pseudoscalar channel this behavior can be explained by the presence of a small negative zero mode.64 Since the high energy part of the quarkonium correlators turns out to be temperature independent to very good approximation one can assume that Gihigh (τ, T ) ≃ Girecon (τ, T ). Then the low energy part of the correlators, i.e. the zero mode contribution can be evaluated as Gilow (τ, T ) = Gi (τ, T ) − Girecon (τ, T ).64 Current lattice
December 16, 2009 8:43 WSPC/INSTRUCTION FILE
qgp4˙bpv
Quarkonium at Finite Temperature
101
Fig. 27. The ratio of the derivatives G′ (τ, T )/G′recon (τ, T ) in the scalar channel (left) and axialvector channel (right) calculated at β = 7.192. The results from anisotropic lattice calculations at β = 6.555 are also shown (open symbols).
data are not precise enough to see a clear τ dependence in Gilow (τ, T ) and it is compatible with being a constant. Therefore, one could take the value of low energy part of the correlator at the midpoint as an estimate for the zero mode contribution, Gilow (τ = 1/(2T ), T ) = T χi (T ). The zero mode contribution is related to the propagation of single (unbound) heavy quark in the medium. Therefore, it is natural to describe its temperature dependence in terms of a quasiparticle model with effective temperature dependent heavy quark masses. Since the temporal component of the vector correlator has no high energy part, i.e. Gvc0 (τ, T ) = −T χvc0 (T ) = −T χ(T ) it is most suitable for fixing the effective quark mass. Matching the lattice data to the free theory expression for χ(T ) one can determine the effective quasiparticle masses meff .64 The results of this analysis are shown in Fig. 28 for different values of the constituent heavy quark mass, including the bottom quarks, in terms of thermal mass correction δmeff (T ) = meff (T )−m. As one can see from the figure the thermal mass correction decreases monotonically with increasing the constituent quark mass and increasing temperature for bottom quarks it is not incompatible with the leadingorder perturbative prediction δmeff = −
4 g 2 (T ) mD , 3 4π
(61)
with mD = g(T )T being the perturbative Debye mass (Nf = 0 because we work in the quenched approximation). Having determined the effective heavy quark mass we can study the zero mode contribution in other channels. If the quasiparticle model is correct the zero mode contribution should be a function of meff /T only. Therefore in Fig. 29 the temperature dependence of zero mode contribution for the vector and axialvector channel is shown as function of T /meff . Indeed, all the lattice data seem to fall on one curve within errors, which agrees with the quasiparticle model prediction shown
December 16, 2009 8:43 WSPC/INSTRUCTION FILE
102
qgp4˙bpv
A. Bazavov, P. Petreczky & A. Velytsky
Fig. 28. The thermal mass correction as function of the temperature. The dashed line and the band correspond to the perturbative prediction of the thermal mass correction.
Fig. 29. The thermal velocity of heavy quarks (left) and the zero mode contribution to the axial3 ax 2 vector correlator Gax low /T = χ /T (right). The lines show the prediction of the quasiparticle model with meff (T ). The open symbols show the thermal velocity squared estimated on anisotropic lattices.55
as the black lines. For the vector channel we show the data in terms of the ravc0 tio Gvc , where Gvc low (T )/G low (T ) is the sum over all spatial components. This is because this quantity does not depend on the renormalization64 and has simple 2 physical interpretation in terms of averaged thermal velocity squared vth . The later follows from the fact that due to the large quark mass the Boltzmann approximation can be used and we have Z Z 2 Gvc 3 p −Ep /T 3 −Ep /T 2 low (T ) ≃ d p e d pe = vth . (62) Gvc0 (T ) Ep2 The zero mode contribution in the vector channel has also been studied on anisotropic lattices55 and in Fig. 29 we also show the corresponding results for thermal velocity squared.
Quarkonium at Finite Temperature
103
6. Potential Models at Finite Temperature Quarkonium properties at finite temperature have been studied in potential models since the famous paper of Matsui and Satz1 (for recent review see Ref. 2). The basic idea behind this is that color screening will modify the potential, which becomes short range and cannot support bound states of heavy quarks at sufficiently high temperatures. As discussed in Sec. 2 this approach can be justified if there is a separation of the scales related to binding energy and other scales in the problem, like the inverse size of the bound states and the temperature. Close to the QCD transition this separation of scales is not obvious, however, the lattice calculation of static quark correlators, discussed in Sec. 4 show that screening effects are very strong already in the transition region. In fact, correlation functions of static quark antiquark pair show significant temperature modifications already at distances similar to quarkonium size. Therefore, we may expect the most of quarkonium bound states dissolve in the deconfined phase at temperatures close to the transition temperature. This seemingly contradicts to the small temperature dependence of quarkonium correlators and spectral functions discussed in the previous section. Due to the heavy quark mass quarkonium spectral functions can be calculated in the potential approach by relating the spectral functions in the threshold region to the nonrelativistic Green function14,85,86 6 (63) σ(ω) = K ImGnr (~r, r~′ , E)~r=r~′ =0 , π σ(ω) = K
6 1 ~ ·∇ ~ ′ Gnr (~r, r~′ , E) ~′ , Im∇ ~ r =r =0 π m2c
(64)
for Swave, and P wave quarkonia, respectively. Here E = ω − 2m. The prefactor K accounts for relativistic and radiative corrections. The nonrelativistic Green function is calculated from the Schr¨odinger equation with a deltafunction on the right hand side. Away from the threshold the spectral function is matched to the perturbative result. It has been shown that this approach can provide a fair description of quarkonium correlators at zero temperature.87 Other approaches to calculate quarkonium spectral functions in potential models were proposed in Refs. 49, 88–91. Close to the transition temperature the perturbative calculations of the potential are not applicable. Therefore in Ref. 85 the singlet free energy calculated in quenched lattice QCD has been used to construct the potential. The drawback of this approach is that there is no one to one correspondence between the singlet free energy and the potential in the nonperturbative domain. Therefore in Ref. 85 several possible forms of the potential compatible with the lattice data have been considered. Furthermore, lattice calculations do not have much information about the imaginary part of the potential. Therefore the imaginary part of the potential has been approximated by a small constant term. The results of the calculations of the charmonium and bottomonium spectral functions for Swave are shown in Fig. 30. In the case of charmonium all bound states are melted at temperatures higher than 1.2Tc. We see, however a significant threshold enhancement, i.e. near
104
A. Bazavov, P. Petreczky & A. Velytsky
0.16
σ(ω)/ω2
G/Grec
1.2Tc 1.5Tc 2.0Tc free lattice, 1.5Tc
0.14 0.12
1 0.99
0.1
τ [fm]
0.98 0.08
0.05
0.15
0.25
0.06 0.04 0.02 ω [GeV] 0 2
0.8
3
4
σ(ω)/ω2
5
6
7
8
9
1.02 G/Grec
0.7
1.01
0.6
1
0.5
0.99
τ [fm] 0.05
0.4
0.1
0.15
0.2
T=0 1.2Tc 1.5Tc 2.0Tc free
0.3 0.2 0.1 0 9
10
11
12
13
14
ω [GeV]
Fig. 30. The charmonium (top) and bottomonium (bottom) spectral functions at different temperatures. For charmonium we also show the spectral functions from lattice QCD obtained from the MEM at 1.5Tc . The errorbars on the lattice spectral function correspond to the statistical error of the spectral function integrated in the ωinterval corresponding to the horizontal errorbars. The insets show the corresponding ratio G/Grecon together with the results from anisotropic lattice calculations.55 For charmonium, lattice calculations of G/Grecon are shown for T = 1.2Tc (squares), 1.5Tc (circles), and 2.0Tc (triangles). For bottomonium lattice data are shown for T = 1.5Tc (circles) and 1.8Tc (triangles).
the threshold the spectral function is much larger than in the free case. In the bottomonium case only the ground state survives in the deconfined phase. At temperatures above 2Tc we see the melting of the ground state as well. Quarkonium spectral functions calculated using the perturbative potential, where the imaginary part is fully taken into account, show similar qualitative features.14,86 Here too a significant threshold enhancement is seen. The imaginary part of the potential plays an important role in weakening the bound state peak
Quarkonium at Finite Temperature
105
or transforming it to mere threshold enhancement. Let us note that the nexttoleading order perturbative correction to the quarkonium spectral functions also give rise to significant enhancement in the threshold region.92 From the spectral functions we can calculate the quarkonium correlation functions in Euclidean time G(τ, T ) and compare them to the available lattice data. This comparison is shown in Fig. 30 for the ratio G(τ, T )/Grecon (τ, T ). As one can see from the figure the melting of bound states does not lead to large change in the Euclidean correlation functions. The ratio G(τ, T )/Grecon (τ, T ) calculated in the potential model is flat and temperature independent in agreement with lattice calculations. This means that threshold enhancement can compensate for melting of bound states in terms of the Euclidean correlators. Spectral functions of P wave quarkonium have been also calculated14,85 and show significant threshold enhancement as well. As the consequence the ratio of the derivatives G′ (τ, T )/G′recon (τ, T ) is temperature independent and close to unity85 in agreement with lattice calculations shown in the previous section. The analysis discussed above has been done in quenched QCD. This is because only in quenched QCD we have sufficiently precise lattice calculations of quarkonium correlators. Potential model calculations of the spectral functions have been extended to 2+1 flavor QCD using the lattice data discussed in Sec. 4. These calculations show that all quarkonium states except the ground state bottomonium dissolve in the quark gluon plasma.93 The upper limits on the dissociation temperatures for different quarkonium states obtained in this analysis are given in Table 2. Table 2. Upper bound on the dissociation temperatures for different quarkonium states in 2+1 flavor QCD.93
State
χc
ψ′
J/ψ
Υ′
χb
Υ
Tdis
≤ Tc
≤ Tc
1.2Tc
1.2Tc
1.3Tc
2Tc
7. Conclusion In this paper we discussed quarkonium properties in quark gluon plasma. We discussed how color screening can be studied nonperturbatively on the lattice using spatial correlation functions of static quark and antiquark. We discussed the importance of singlet and adjoint (triplet for SU (2) or octet for SU (3)) degrees of freedom for understanding the temperature dependence of static meson (quarkantiquark) correlators. It has been shown how color singlet and adjoint degrees of freedom can be defined in the effective field theory framework, the socalled thermal pNRQCD. We have seen that the singlet correlators of static quark and antiquark show strong screening effects at distances comparable to quarkonium size. Thus it is natural to expect that most quarkonium states melt in quark gluon plasma. Quite surprisingly lattice calculations of quarkonium correlators in Euclidean time show
106
A. Bazavov, P. Petreczky & A. Velytsky
very little temperature dependence. We reviewed the current status of these calculations in quenched QCD. It turns out that the spatial lattice spacing should be smaller than (4 GeV)−1 to have full control over the discretization errors. Therefore the extension of these calculations to full QCD pioneered in Ref. 94 will be quite difficult. The only source of significant temperature dependence of quarkonium correlators is the zero mode contribution which is not related to bound states but to transport properties of heavy quarks in the medium.81 We have shown that this contribution can be well described by a quasiparticle model with temperature dependent heavy quark mass. We have presented the calculations of quarkonium spectral functions in potential model and have shown that all quarkonium states, except the ground state bottomonium melt in the deconfined phase. We have also shown how the seemingly existing contradiction of strong color screening leading to quarkonium melting and very weak temperature dependence of quarkonium correlators can be resolved within potential models. It turns out that strong threshold enhancement of quarkonium spectral functions can compensate the absence of bound state and result in Euclidean correlarors, which are almost temperature independent. The fact that no charmonium bound states can exist in quark gluon plasma has important consequences for describing charmonium production in heavy ion collisions. In absence of bound states heavy quarks in the plasma can be treated quasiclassically using Langevin dynamics. In this scenario the residual correlation between the heavy quark and antiquark, visible in the threshold enhancement and the finite life time of the plasma play an important role.95 It turns out that using these ideas and the interactions of heavy quark determined in lattice QCD it is possible to uderstand the J/ψ suppression pattern at RHIC.95 Acknowledgements This work was supported by U.S. Department of Energy under Contract No. DEAC0298CH10886. The work of A.B. was supported by grants DOE DEFC0206ER41439 and NSF 0555397. A.V. work was supported by the Joint Theory Institute funded together by Argonne National Laboratory and the University of Chicago, and in part by the U.S. Department of Energy, Division of High Energy Physics and Office of Nuclear Physics, under Contract DEAC0206CH11357. References 1. T. Matsui and H. Satz, J/psi suppression by quarkgluon plasma formation, Phys. Lett. B178, 416 (1986). 2. A. Mocsy, Potential Models for Quarkonia (2008), eprint 0811.0337 [hepph]. 3. N. Brambilla, J. Ghiglieri, A. Vairo, and P. Petreczky, Static quarkantiquark pairs at finite temperature, Phys. Rev. D78, 014017 (2008). 4. N. Brambilla et al., Heavy quarkonium physics (2004), eprint hepph/0412158. 5. N. Brambilla, A. Pineda, J. Soto, and A. Vairo, Effective field theories for heavy quarkonium, Rev. Mod. Phys. 77, 1423 (2005).
Quarkonium at Finite Temperature
107
6. W. E. Caswell and G. P. Lepage, Effective Lagrangians for bound state problems in QED, QCD, and other field theories, Phys. Lett. B167, 437 (1986). 7. P. Petreczky, Heavy quark potentials and quarkonia binding, Eur. Phys. J. C43, 51–57 (2005). 8. M. Laine, O. Philipsen, P. Romatschke, and M. Tassler, Realtime static potential in hot QCD, JHEP 03, 054 (2007). 9. R. D. Pisarski, Damping rates for moving particles in hot QCD, Phys. Rev. D47, 5589–5600 (1993). ¯ correlators in 10. A. Beraudo, J. P. Blaizot, and C. Ratti, Real and imaginarytime QQ a thermal medium, Nucl. Phys. A806, 312–338 (2008). 11. K. G. Wilson, Confinement of quarks, Phys. Rev. D10, 2445–2459 (1974). 12. M. Albanese et al., Glueball masses and string tension in lattice QCD, Phys. Lett. B192, 163 (1987). 13. M. Laine, O. Philipsen, and M. Tassler, Thermal imaginary part of a realtime static potential from classical lattice gauge theory simulations, JHEP 09, 066 (2007). 14. Y. Burnier, M. Laine, and M. Vepsalainen, Heavy quarkonium in any channel in resummed hot QCD, JHEP 01, 043 (2008). 15. L. D. McLerran and B. Svetitsky, Quark liberation at high temperature: A Monte Carlo study of SU(2) gauge theory, Phys. Rev. D24, 450 (1981). 16. S. Nadkarni, Nonabelian Debye Screening. 1. The Color Averaged Potential, Phys. Rev. D33, 3738 (1986). 17. P. Petreczky et al., Lattice calculation of medium effects at short and long distances, Nucl. Phys. A698, 400–403 (2002). 18. S. Digal, S. Fortunato, and P. Petreczky, Heavy quark free energies and screening in SU(2) gauge theory, Phys. Rev. D68, 034008 (2003). 19. O. Jahn and O. Philipsen, The Polyakov loop and its relation to static quark potentials and free energies, Phys. Rev. D70, 074504 (2004). 20. G. S. Bali, QCD forces and heavy quark bound states, Phys. Rept. 343, 1–136 (2001). 21. C. J. Morningstar, K. J. Juge, and J. Kuti, Where is the string limit in QCD?, Nucl. Phys. Proc. Suppl. 73, 590–595 (1999). 22. K. J. Juge, J. Kuti, and C. Morningstar, Fine structure of the QCD string spectrum, Phys. Rev. Lett. 90, 161601 (2003). 23. C. Michael and S. J. Perantonis, Potentials and glueballs at large beta in SU(2) pure gauge theory, J. Phys. G18, 1725–1736 (1992). 24. N. Brambilla, A. Pineda, J. Soto, and A. Vairo, Potential NRQCD: An effective theory for heavy quarkonium, Nucl. Phys. B566, 275 (2000). 25. G. S. Bali and A. Pineda, QCD phenomenology of static sources and gluonic excitations at short distances, Phys. Rev. D69, 094001 (2004). 26. O. Kaczmarek, F. Karsch, E. Laermann, and M. Lutgemeier, Heavy quark potentials in quenched QCD at high temperature, Phys. Rev. D62, 034021 (2000). 27. S. Digal, P. Petreczky, and H. Satz, String breaking and quarkonium dissociation at finite temperatures, Phys. Lett. B514, 57–62 (2001). 28. S. Digal, P. Petreczky, and H. Satz, Quarkonium feeddown and sequential suppression, Phys. Rev. D64, 094015 (2001). 29. O. Kaczmarek, F. Karsch, P. Petreczky, and F. Zantow, Heavy quark antiquark free energy and the renormalized polyakov loop, Phys. Lett. B543, 41–47 (2002). 30. O. Philipsen, Nonperturbative formulation of the static color octet potential, Phys. Lett. B535, 138–144 (2002). 31. O. Kaczmarek, F. Karsch, P. Petreczky, and F. Zantow, Heavy quark free energies, potentials and the renormalized Polyakov loop, Nucl. Phys. Proc. Suppl. 129, 560–562 (2004).
108
A. Bazavov, P. Petreczky & A. Velytsky
32. O. Kaczmarek, F. Karsch, F. Zantow, and P. Petreczky, Static quark antiquark free energy and the running coupling at finite temperature, Phys. Rev. D70, 074505 (2004). 33. P. Petreczky and K. Petrov, Free energy of a static quark antiquark pair and the renormalized Polyakov loop in three flavor QCD, Phys. Rev. D70, 054503 (2004). 34. O. Kaczmarek and F. Zantow, Static quark antiquark interactions in zero and finite temperature QCD. I: Heavy quark free energies, running coupling and quarkonium binding, Phys. Rev. D71, 114510 (2005). 35. O. Kaczmarek, Screening at finite temperature and density, PoS CPOD07, 043 (2007). 36. RBC Collaboration, work in progress. 37. S. Necco and R. Sommer, The N(f) = 0 heavy quark potential from short to intermediate distances, Nucl. Phys. B622, 328–346 (2002). 38. R. Sommer, A New way to set the energy scale in lattice gauge theories and its applications to the static force and alphas in SU(2) Yang–Mills theory, Nucl. Phys. B411, 839–854 (1994). 39. M. Cheng et al., The QCD equation of state with almost physical quark masses, Phys. Rev. D77, 014511 (2008). 40. M. Luscher and P. Weisz, Locality and exponential error reduction in numerical lattice gauge theory, JHEP 09, 010 (2001). 41. A. Bazavov, P. Petreczky, and A. Velytsky, Static quark antiquark pair in SU(2) gauge theory, Phys. Rev. D78, 114026 (2008). 42. S. P. Booth et al., Towards the continuum limit of SU(2) lattice gauge theory, Phys. Lett. B275, 424–428 (1992). 43. G. P. Lepage and P. B. Mackenzie, On the viability of lattice perturbation theory, Phys. Rev. D48, 2250–2264 (1993). 44. A. K. Rebhan, The NonAbelian Debye mass at nexttoleading order, Phys. Rev. D48, 3967–3970 (1993). 45. A. K. Rebhan, NonAbelian Debye screening in one loop resummed perturbation theory, Nucl. Phys. B430, 319–344 (1994). 46. A. Vairo, private communication (2008). 47. M. Le Bellac, Thermal Field Theory, Cambridge monographs on mathematical physics (Cambridge Univ. Press, Cambridge, 1996). 48. E. Braaten, R. D. Pisarski, and T.C. Yuan, Production of soft dileptons in the quarkgluon plasma, Phys. Rev. Lett. 64, 2242 (1990). 49. A. Mocsy and P. Petreczky, Quarkonia correlators above deconfinement, Phys. Rev. D73, 074007 (2006). 50. F. Karsch, E. Laermann, P. Petreczky, and S. Stickan, Infinite temperature limit of meson spectral functions calculated on the lattice, Phys. Rev. D68, 014504 (2003). 51. G. Aarts and J. M. Martinez Resco, Continuum and lattice meson spectral functions at nonzero momentum and high temperature, Nucl. Phys. B726, 93–108 (2005). 52. T. Umeda, K. Nomura, and H. Matsufuru, Charmonium at finite temperature in quenched lattice QCD, Eur. Phys. J. C39S1, 9–26 (2005). 53. S. Datta, F. Karsch, P. Petreczky, and I. Wetzorke, Behavior of charmonium systems after deconfinement, Phys. Rev. D69, 094507 (2004). 54. M. Asakawa and T. Hatsuda, Jpsi and etac in the deconfined plasma from lattice QCD, Phys. Rev. Lett. 92, 012001 (2004). 55. A. Jakovac, P. Petreczky, K. Petrov, and A. Velytsky, Quarkonium correlators and spectral functions at zero and finite temperature, Phys. Rev. D75, 014506 (2007).
Quarkonium at Finite Temperature
109
56. T. R. Klassen, The anisotropic Wilson gauge action, Nucl. Phys. B533, 557–575 (1998). 57. P. Chen, Heavy quarks on anisotropic lattices: The charmonium spectrum, Phys. Rev. D64, 034509 (2001). 58. A. X. ElKhadra, A. S. Kronfeld, and P. B. Mackenzie, Massive fermions in lattice gauge theory, Phys. Rev. D55, 3933–3957 (1997). 59. R. G. Edwards, U. M. Heller, and T. R. Klassen, unpublished (1999). 60. M. Okamoto et al., Charmonium spectrum from quenched anisotropic lattice QCD, Phys. Rev. D65, 094508 (2002). 61. A. Gray et al., The Upsilon spectrum and mb from full lattice QCD, Phys. Rev. D72, 094507 (2005). 62. S. Datta, F. Karsch, P. Petreczky, and I. Wetzorke, A study of charmonium systems across the deconfinement transition, Nucl. Phys. Proc. Suppl. 119, 487–489 (2003). 63. S. Datta, F. Karsch, S. Wissel, P. Petreczky, and I. Wetzorke, Charmonia at finite momenta in a deconfined plasma (2004), eprint heplat/0409147. 64. P. Petreczky, On temperature dependence of quarkonium correlators (2008), eprint 0810.0258 [heplat]. 65. M. Asakawa, T. Hatsuda, and Y. Nakahara, Maximum entropy analysis of the spectral functions in lattice QCD, Prog. Part. Nucl. Phys. 46, 459–508 (2001). 66. G. P. Lepage et al., Constrained curve fitting, Nucl. Phys. Proc. Suppl. 106, 12–20 (2002). 67. R. K. Bryan, Maximum entropy analysis of oversampled data problems, Eur. Biophys. J. 18, 165–174 (1990). 68. Y. Nakahara, M. Asakawa, and T. Hatsuda, Hadronic spectral functions in lattice QCD, Phys. Rev. D60, 091503 (1999). 69. T. Yamazaki et al., Spectral function and excited states in lattice QCD with maximum entropy method, Phys. Rev. D65, 014501 (2002). 70. F. Karsch, E. Laermann, P. Petreczky, S. Stickan, and I. Wetzorke, A lattice calculation of thermal dilepton rates, Phys. Lett. B530, 147–152 (2002). 71. F. Karsch et al., Hadron correlators, spectral functions and thermal dilepton rates from lattice QCD, Nucl. Phys. A715, 701–704 (2003). 72. M. Asakawa, T. Hatsuda, and Y. Nakahara, Hadronic spectral functions above the QCD phase transition, Nucl. Phys. A715, 863–866 (2003). 73. T. Blum and P. Petreczky, Cutoff effects in meson spectral functions, Nucl. Phys. Proc. Suppl. 140, 553–555 (2005). 74. P. Petreczky, Lattice calculations of meson correlators and spectral functions at finite temperature, J. Phys. G30, S431–S440 (2004). 75. P. Petreczky, F. Karsch, E. Laermann, S. Stickan, and I. Wetzorke, Temporal quark and gluon propagators: Measuring the quasiparticle masses, Nucl. Phys. Proc. Suppl. 106, 513–515 (2002). 76. C. R. Allton, J. E. Clowser, S. J. Hands, J. B. Kogut, and C. G. Strouthos, Application of the maximum entropy method to the (2+1)d fourfermion model, Phys. Rev. D66, 094511 (2002). 77. K. Langfeld, H. Reinhardt, and J. Gattnar, Gluon propagators and quark confinement, Nucl. Phys. B621, 131–156 (2002). 78. H. R. Fiebig, Spectral density analysis of time correlation functions in lattice QCD using the maximum entropy method, Phys. Rev. D65, 094512 (2002). 79. K. Sasaki, S. Sasaki, and T. Hatsuda, Spectral analysis of excited nucleons in lattice QCD with maximum entropy method, Phys. Lett. B623, 208–217 (2005).
110
A. Bazavov, P. Petreczky & A. Velytsky
80. T. Umeda, A constant contribution in meson correlators at finite temperature, Phys. Rev. D75, 094502 (2007). 81. P. Petreczky and D. Teaney, Heavy quark diffusion from the lattice, Phys. Rev. D73, 014508 (2006). 82. C. T. H. Davies et al., Precision Upsilon spectroscopy from nonrelativistic lattice QCD, Phys. Rev. D50, 6963–6977 (1994). 83. X. Liao and T. Manke, Relativistic bottomonium spectrum from anisotropic lattices, Phys. Rev. D65, 074508 (2002). 84. S. Datta, A. Jakovac, F. Karsch, and P. Petreczky, Quarkonia in a deconfined gluonic plasma, AIP Conf. Proc. 842, 35–37 (2006). 85. A. Mocsy and P. Petreczky, Can quarkonia survive deconfinement?, Phys. Rev. D77, 014501 (2008). 86. M. Laine, A resummed perturbative estimate for the quarkonium spectral function in hot QCD, JHEP 05, 028 (2007). 87. A. Mocsy and P. Petreczky, Describing charmonium correlation functions in Euclidean time, Eur. Phys. J. ST. 155, 101–106 (2008). 88. A. Mocsy and P. Petreczky, Heavy quarkonia survival in potential model, Eur. Phys. J. C43, 77–80 (2005). 89. W. M. Alberico, A. Beraudo, A. De Pace, and A. Molinari, Potential models and lattice correlators for quarkonia at finite temperature, Phys. Rev. D77, 017502 (2008). 90. D. Cabrera and R. Rapp, Tmatrix approach to quarkonium correlation functions in the QGP, Phys. Rev. D76, 114506 (2007). 91. H. T. Ding, O. Kaczmarek, F. Karsch, and H. Satz, Charmonium correlators and spectral functions at finite temperature (2009), eprint 0901.3023 [heplat]. 92. Y. Burnier, M. Laine, and M. Vepsalainen, Heavy quark medium polarization at nexttoleading order, JHEP 02, 008 (2009). 93. A. Mocsy and P. Petreczky, Color screening melts quarkonium, Phys. Rev. Lett. 99, 211602 (2007). 94. G. Aarts, C. Allton, M. B. Oktay, M. Peardon, and J.I. Skullerud, Charmonium at high temperature in twoflavor QCD, Phys. Rev. D76, 094513 (2007). 95. C. Young and E. Shuryak, Charmonium in strongly coupled quarkgluon plasma (2008), eprint 0803.2866 [nuclth].
HEAVY QUARKS IN THE QUARKGLUON PLASMA
RALF RAPP Cyclotron Institute and Physics Department, Texas A&M University, College Station, Texas 778433366, USA
[email protected] HENDRIK VAN HEES Institut f¨ ur Theoretische Physik, JustusLiebigUniversit¨ at Giessen, D35392 Giessen, Germany
[email protected] Heavyflavor particles are believed to provide valuable probes of the medium produced in ultrarelativistic collisions of heavy nuclei. In this article we review recent progress in our understanding of the interactions of charm and bottom quarks in the QuarkGluon Plasma (QGP). For individual heavy quarks, we focus on elastic interactions for which the large quark mass enables a Brownian motion treatment. This opens a unique access to thermalization mechanisms for heavy quarks at low momentum, and thus to their transport coefficients in the quarkgluon fluid. Different approaches to evaluate heavyquark diffusion are discussed and compared, including perturbative QCD, effective potential models utilizing input from lattice QCD and stringtheoretic estimates in conformal field theories. Applications to heavyquark observables in heavyion collisions are realized via relativistic Langevin simulations, where we illustrate the important role of a realistic medium evolution to quantitatively extract the heavyquark diffusion constant. In the heavy quarkonium sector, we briefly review the current status in potentialmodel based interpretations of correlation functions computed in lattice QCD, followed by an evaluation of quarkonium dissociation reactions in the QGP. The discussion of the phenomenology in heavyion reactions focuses on thermal model frameworks paralleling the open heavyflavor sector. We also emphasize connections to the heavyquark diffusion problem in both potential models and quarkonium regeneration processes.
1. Introduction The investigation of strongly interacting matter constitutes a major challenge in modern nuclear and particle physics. Of particular interest are phase changes between hadronic and quarkgluon matter, similar to the one which is believed to have occurred in the early Universe at a few microseconds after its birth. While the theory of the strong force is by now well established in terms of Quantum Chromodynamics (QCD),1–3 two of its major manifestations in the world around us — the confinement of quarks and gluons and the generation of hadronic masses — are subject of vigorous contemporary research. Both phenomena occur at energymomentum scales of Q2 . 1 GeV2 where the QCD coupling constant is rather large, αs & 0.3, and therefore perturbation theory is not reliable and/or applicable. In a hot and dense medium at sufficiently large temperature (T ) and/or 111
112
R. Rapp & H. van Hees
quark chemical potential (µq ), one expects the finitesize hadrons to be dissolved into a deconfined QuarkGluon Plasma (QGP) where the condensates underlying hadronicmass generation have melted. Numerical simulations of latticediscretized QCD (lQCD) at finite temperature predict the phase change from hadronic to quarkgluon matter to occur at a “pseudocritical” temperature of Tc ' 200 MeV.4 This appears to be a rather small scale for a “perturbative QGP” (pQGP) of weakly interacting quarks and gluons to be realized, even though the computed energy density matches that of an ideal (noninteracting and massless) QGP within 20% or so for T & 1.2 Tc. In the laboratory, one hopes to create a QGP by colliding heavy atomic nuclei at ultrarelativistic energies, with a centerofmass energy per colliding nucleon pair √ well above the nucleon rest mass, s/A MN . If the energy deposition in the reaction zone is large enough, and if the interactions of the produced particles are strong enough, the notion of an interacting medium may be justified, despite its transient nature. This notion has been convincingly verified in nuclear collision experiments over the last ∼ 25 years at the SuperProtonSynchrotron (SPS) at the European Organization for Nuclear Research (CERN)5 and at the Relativistic HeavyIon Collider (RHIC) at Brookhaven National Laboratory (BNL).6 Transversemomentum (pT ) spectra of different hadron species in the lowpT regime (pT . 2–3 GeV) reveal that the produced medium explodes collectively reaching expansion velocities in excess of half the speed of light. In the highpT regime (pT & 5 GeV), which in the heavyion environment became available at RHIC for the first time, hadron spectra are suppressed by up to a factor of ∼ 5 relative to p–p collisions, indicative for a strong absorption of highenergy partons traversing the medium. 7 The inclusive production of charmquark bound states (J/ψ mesons) is suppressed by a factor of 35 at both SPS and RHIC, indicative for their dissolution in the medium (possibly related to deconfinement).8–10 A large excess of electromagnetic radiation (photons and dileptons) is observed, indicative for medium temperatures around 200 MeV and a “melting” of the ρmeson resonance (possibly related to hadronic mass degeneration).11,12 A more differential analysis of hadron spectra in noncentral Au–Au collisions at RHIC reveals a large elliptic asymmetry of the collective flow (“elliptic flow”): the spatial asymmetry of the initial nuclear overlap zone is converted into an opposite asymmetry in the final hadron pT spectra. Within a hydrodynamic modeling of the exploding fireball this observation requires a rapid thermalization and a very small viscosity of the interacting medium. 13–16 Only then can spatial pressure gradients build up fast enough to facilitate an effective conversion into azimuthal asymmetries in the energymomentum tensor of the system. The agreement of hydrodynamic predictions with ellipticflow data at RHIC led to the notion of an “almost perfect liquid”, with a ratio of viscosity to entropy density close to a conjectured lower bound of any quantum mechanical system.17 The microscopic mechanisms underlying these rather remarkable transport properties are yet to be determined. In this context, heavy quarks (charm and
Heavy Quarks in the QuarkGluon Plasma
113
bottom, Q = c and b) and their bound states (charmonia and bottomonia) are recognized as particularly suitable probes of the medium produced in ultrarelativistic heavyion collisions (URHICs).18,a In the present article we will attempt to review the current status of the theory and phenomenology of this promise. Let us first focus on the sector of individual heavy quarks (open heavy flavor). The fact that their masses are well above the typical temperature of the system, mQ T , has at least three important implications: ¯ production process is essentially restricted to primordial (1) The (hard) QQ N –N collisions,20 i.e., reinteractions in the subsequently evolving medium are not expected to change the number of heavy quarks (reminiscent of the “factorization theorem” of perturbative QCD21 ); this is borne out experimentally by a scaling of c¯ c production, Nc¯c , with the number of binary N –N collisions, Ncoll , at different collision centralities.22 (2) The thermal relaxation time of heavy quarks ought to be larger than for light quarks, parameterically by a factor ∼ mQ /T ≈ 5–20. With a lightquark and gluon thermalization time of τq,g ' 0.3–1 fm/c (as indirectly inferred from hydrodynamic modeling at RHIC) and an estimated QGP lifetime of τQGP ' 5 fm/c in central Au–Au collisions, one expects τc (τb ) to be on the same order as (significantly larger than) τQGP . Thus, charm (and especially bottom) quarks are not expected to reach thermal equilibrium, but their reinteractions should impart noticeable modifications on the initial momentum spectrum (less pronounced for bottom). The final heavyquark (HQ) spectra may therefore encode a “memory” of the interaction history throughout the evolving fireball, by operating in between the limits of thermalization and free streaming. (3) The theoretical task of describing HQ interactions is amenable to a diffusion treatment, i.e., Brownian motion of a heavy test particle in a bath of a lightparticle fluid. Nonrelativistically, the typical thermal momentum of a heavy quark is p2th ' 3mQ T T 2 , and therefore much larger than the typical momentum transfer from the medium, Q2 ∼ T 2 . This allows to expand the Boltzmann equation in momentum transfer to arrive at a Fokker–Planck description of HQ diffusion in the QGP, which directly yields the pertinent transport coefficients as well. The above three points provide a welldefined framework to construct inmedium HQ interactions in QCD matter and test them against observables in URHICs (quantitative comparisons additionally require to account for effects of hadronization of the quarks, as well as reinteractions in the hadronic medium). The Fokker– a The
(weakdecay) lifetime of the top quark of ∼ 0.1 fm/c is too short to render it a viable probe in URHICs; thus, heavy quarks will exclusively refer to charm and bottom in this article. Strange quarks are in between the heavy and lightquark limit, forming their own complex of valuable observables.19
114
R. Rapp & H. van Hees
Planck approach is readily implemented for the case of elastic p + Q → p + Q scattering off partons in the medium (p = q, q¯, g).23–28 In the lighthadron sector, however, the large suppression of highpT spectra is believed to be largely caused by radiative energy loss of highenergy partons traversing the QGP, i.e., mediuminduced gluon radiation of type q + g → q + g + g.7,29,30 Even in the lowpT regime, perturbative 2, 3 ↔ 3 scattering processes have been suggested to facilitate the rapid thermalization required by phenomenology (albeit in connection with rather large coupling constants of αs ' 0.5).31 The situation could be quite different in the HQ sector. In the lowmomentum limit, gluonBremsstrahlung of a heavy quark is suppressed32 and the dominant momentumtransfer reaction is elastic scattering.26 As is well known from classical electrodynamics, the radiative energy loss of a muon is suppressed relative to an electron by a mass ratio (me /mµ )4 . In perturbative QCD (pQCD) it is currently an open question at what momentum scale radiative energy loss of a heavy quark takes over from the collisional one (which, most likely, will depend on additional parameters such as temperature, path length, etc.). In fact, this may not even be a welldefined question since nonperturbative processes at moderate momentum transfers may supersede perturbative ones before the elastic part of the latter dominates over the radiative one. The relation between perturbative and nonperturbative interactions is one of the key issues to be addressed in this review. Experimental signatures for the modifications of HQ spectra in URHICs are currently encoded in singleelectron (e± ) spectra associated with the semileptonic decays of charm and bottom hadrons, D, B, Λc , . . . → eνX. These measurements require a careful subtraction of all possible “photonic” sources of electrons, such as photon conversions in the detector material, Dalitz decays of π and η, vectormeson decays, and others. The modifications of the “nonphotonic” electron spectra (associated with heavyflavor decays) in Au–Au collisions are then quantified by the stane dard nuclear modification factor, RAA , and elliptic flow coefficient, v2e . The available √ RHIC data in semicentral and central Au–Au collisions at sN N = 200 GeV exe hibit a substantial elliptic flow of up to v2 ' 10% and a large highpT suppression e down to RAA ' 0.25, respectively.22,33–35 Both values are quite comparable to those for light hadrons (the pion v2 reaches somewhat higher, to about 15%). Radiative energyloss models36 based on perturbative QCD cannot explain the e± data. These data were, in fact, instrumental37 in reconsidering elastic scattering as a significant source of parton energy loss in the QGP.25–27,38 The combination of pQCD elastic and radiative scattering does not suffice either to reproduce the observed suppression once a realistic bottom component is accounted for in the electron spectra. 38 Elastic scattering based on nonperturbative interactions, as proposed in Refs. 25 and 28, simultaneously accounts for the e± elliptic flow and suppression reasonably well.22 This has reinforced the hope that HQ observables provide the promised precision tool to characterize transport properties of the “strongly coupled QGP” m (sQGP). E.g., if a clear mass hierarchy in thermal relaxation times, τQ ∝ TQ τq (as well as τb /τc = mb /mc ) emerges from a quantitative analysis of URHIC data, it
Heavy Quarks in the QuarkGluon Plasma
115
would be suggestive for a universal behavior of light and heavyquark transport in the QGP. However, there is still a substantial way to go before such a program can be realized, as discussed below.b It should not be surprising if inmedium properties of open heavy flavor, especially at low momentum, are closely related to medium modifications of heavy quarkonia. The latter have a long history as “probes” of the QGP in heavyion collisions, especially as potential indicators of the deconfinement transition, cf. Refs. 8– 10 for a broad uptodate coverage of this topic. In particular, progress in finitetemperature lattice QCD41–43 has triggered vigorous reconsideration of the question whether quarkonia, especially their ground states, can survive in the QGP significantly above the critical temperature. These developments include the application of potential models at finite temperature, coupled with the hope that heavyquark free energies as computed in thermal lQCD can serve as a modelindependent input for the lowenergy heavyquark interaction. If charmonium binding indeed remains sufficiently strong in the QGP to support bound states up to rather high temperatures, it is conceivable that the underlying interaction is of a more general relevance and therefore also operative in heavylight45 and maybe even lightlight44 systems. Especially in the former case, from the point of view of elastic (onshell) scattering of a heavy quark in the medium, the conditions for momentum transfer are comparable to the heavyheavy interaction governing quarkonium properties. Since lowmomentum HQ interactions determine their transport properties, one immediately recognizes an intimate relation between HQ transport and inmedium quarkonia. These connections are also being exploited in the analysis of thermal lQCD computations of quarkonium correlation functions.46 In addition to the binding properties, the inelastic reaction rates of quarkonia with surrounding partons or hadrons are a key ingredient for a quantitative description of their spectral function in QCD matter (also here “quasielastic” scattering of thermal partons with a heavy quark inside the quarkonium bound states may play an important role, especially if the binding energy becomes small47 ). A good control over all of these aspects is mandatory to utilize quarkonium properties as diagnostic tool in heavyion collisions and eventually deduce more general properties of the medium produced in these reactions. As in the open heavyflavor sector, this has to be built on a solid knowledge of the spacetime history of nuclear collisions, as well as of the initial conditions on quarkonium spectra. The latter aspect could be more involved than for single heavy quarks, since (a) measurements in p–A collisions show that coldnuclearmatter (CNM) effects from the incoming nuclei (e.g., the socalled nuclear absorption) affect the primordial charmonium number significantly (e.g., with up to 60% suppression for J/ψ at SPS energies when extrapolated to central Pb– Pb collisions); (b) the boundstate formation time introduces another rather long time scale (soft energy scale) which is easily of the order of (or longer than) the bA
recent review article39 addresses similar topics but from a more elementary perspective; see also Ref. 40.
116
R. Rapp & H. van Hees
thermalization time of the medium (at least for charmonia and excited bottomonia at RHIC energies and higher). Charmonium suppression beyond the level of CNM effects has been established in semi/central Pb–Pb and Au–Au collisions at SPS48 and RHIC,49 respectively. An intriguing finding is that the observed suppression pattern and magnitude is very comparable at SPS and RHIC, despite the different collision energies which lead to substantial variations in, e.g., lighthadron observables (most notably a factor ∼ 2 larger charged particle rapidity density and stronger collective phenomena at RHIC). This “degeneracy” was predicted47 as a consequence of charmonium regeneration mechanisms:50–52 a stronger suppression in the hotter and denser medium at RHIC is compensated by the coalescence of c and c¯ quarks in the QGP and/or at hadronization (the c¯ c production cross section at RHIC is about a factor of ∼ 100 larger than at SPS energies). While an “extra” source of charmonia increases the complexity of pertinent observables in heavyion reactions, it also provides another, rather direct, connection between the open and hidden heavyflavor sectors. The secondary charmonium yield from c–¯ c coalescence necessarily carries imprints of the charmquark distributions, both in its magnitude (softer cquark spectra are expected to result in larger coalescence probabilities) and in its momentum spectra (including elliptic flow). A comprehensive theoretical and phenomenological analysis of open and hidden heavy flavor is thus becoming an increasingly pressing and challenging issue. As a final remark on quarkonia in this introduction, we point out that bottomonium production in heavyion reactions is less likely to receive regeneration contributions (at least at RHIC and possibly neither at LHC). In addition, the increase in bottomonium binding energies (compared to charmonia) render them rather sensitive probes of color screening which strongly influences its dissociation rates.53 Bottomonia thus remain a promising observable to realize the originally envisaged “spectral analysis of strongly interacting matter”.54 Our review is organized as follows: In Sec. 2 we outline the theoretical framework of evaluating HQ diffusion in equilibrium QCD matter. We first recall basic steps in setting up the HQ diffusion equation (Sec. 2.1) which determines the time evolution of the HQ distribution function in terms of pertinent transport coefficients based on elastic scattering amplitudes. This is followed by a discussion of several microscopic approaches to calculate the HQ friction and diffusion coefficients in the QGP: perturbative QCD (Sec. 2.2) at leading (2.2.1, 2.2.2, 2.2.3) and nexttoleading order (2.2.4) as well as for threebody scattering (2.2.5); nonperturbative calculations (Sec. 2.3) implementing resonancelike correlations in the QGP using HQ effective theory (Sec. 2.3.1), inmedium T matrices with HQ potentials estimated from thermal lattice QCD (Sec. 2.3.2), or collisionaldissociation mechanisms of heavy mesons (Sec. 2.3.3); and stringtheoretic evaluations based on the conjectured correspondence to conformal field theories (Sec. 2.4). The variety of the proposed approaches calls for an attempt to reconcile the underlying assumptions
Heavy Quarks in the QuarkGluon Plasma
117
and basic interactions (Sec. 2.5). This is followed by a discussion of inelastic (radiative) energyloss calculations and their relation to elastic ones (Sec. 2.6). We briefly consider interactions of open heavyflavor hadrons in hadronic matter (Sec. 2.7). In Sec. 3 we discuss applications of HQ diffusion to URHICs using relativistic Langevin simulations of the Fokker–Planck equation within an expanding finitesize thermal medium (Sec. 3.1). A realistic description of the latter (utilizing hydrodynamics, transport models or suitable parameterizations thereof) is an essential prerequisite to enable a quantitative extraction of transport properties of the QCD medium (Sec. 3.2). Further ingredients are reliable initial conditions (possibly modified by nuclear effects) and the conversion of quarks to hadrons (Sec. 3.3). Implementations of different HQ diffusion coefficients in various spacetime models are quantitatively analyzed in terms of the resulting HQ spectra at RHIC, in particular their nuclear modification factor and elliptic flow (Sec. 3.4). Including effects of hadronization (as well as semileptonic electron decays), a quantitative comparison of these calculations to singleelectron spectra at RHIC is conducted (Sec. 3.5). We emphasize the importance of a consistent (simultaneous) description of pt spectra and elliptic flow. Only then can these observables be converted into a meaningful (albeit preliminary) estimate of charm and bottomquark diffusion coefficients in the QGP. We finish the discussion on open heavy flavor with an attempt to utilize these coefficients for a schematic estimate of the ratio of shear viscosity to entropy density in the QGP (Sec. 3.6). In Sec. 4 we elaborate on theoretical and phenomenological analyses of quarkonia in medium and their production in heavyion collisions. We first address spectral properties of quarkonia in equilibrium matter (Sec. 4.1); Euclidean correlation functions computed in lattice QCD with good precision have been analyzed in terms of potential models based on screened HQ potentials (Sec. 4.1.1). The interplay of color screening and partoninduced dissociation reactions has important consequences for the evaluation of quarkonium dissociation widths (Sec. 4.1.2). In light of the charmonium equilibrium properties the current status of the phenomenology in heavyion collisions is discussed (Sec. 4.2). First, quarkonium transport equations are introduced along with their main ingredients, i.e., dissociation widths and equilibrium numbers using relative chemical equilibrium at fixed HQ number (Sec. 4.2.1); this is followed by model comparisons to J/ψ data at SPS and RHIC, scrutinizing suppression vs. regeneration mechanisms and their transversemomentum dependencies (Sec. 4.2.2), and a brief illustration of predictions for Υ production at RHIC. In Sec. 5 we recollect the main points of this article and conclude. 2. HeavyQuark Interactions in QCD Matter At an energy scale of the (pseudo) critical QCD transition temperature, the large charm and bottomquark masses imply that the HQ diffusion problem is a nonrelativistic one (unless initial conditions bring in an additional large scale). In the weakcoupling regime this further implies that the dominant interactions of the
118
R. Rapp & H. van Hees
heavy quark are elastic scattering (gluon radiation is suppressed by an extra power in αs and cannot be compensated by a large momentum transfer as could be the case for a fast quark; see, e.g., the discussion in Ref. 26). It turns out, however, that the perturbative expansion of the charmquark diffusion coefficient, evaluated using thermal field theory, is not well convergent even for a strong coupling constant as low as αs = 0.1.55 Thus, nonperturbative methods, e.g., resummations of large contributions or interactions beyond perturbation theory, are necessary to improve the estimates of HQ diffusion. This is not surprising since transport coefficients usually involve the zeromomentum limit of correlation functions rendering them susceptible to threshold effects which may increase with the mass of the particles. A simple example of such kind are Coulomblike bound states (e.g., heavy quarkonia), where the binding energy increases with increasing HQ mass, B ∝ α2s mQ , to be compared to thermal effects, e.g., at a scale ∼ gT for Debye screening (to leading order in g) or at ∼ T for inelastic dissociation reactions with thermal partons. An interesting question in this context is whether potential models are a viable means to evaluate HQ interactions in the QGP. If a suitable formulation of a potential at finite temperature can be established, a promising opportunity arises by extracting these from first principle lattice computations of the HQ free energy. In the heavyquarkonium sector such a program has been initiated a few years ago56–60 with fair success, although several open questions remain.58,61–63 If applicable, potential models have the great benefit of allowing for nonperturbative solutions utilizing Schr¨ odinger or Lippmann–Schwinger equations; the calculated scattering amplitudes can then be straightforwardly related to transport coefficients. A key issue in this discussion is the transition to the (ultra)relativistic regime, which becomes inevitable in applications to experiment toward high momentum. While relativistic kinematics can be readily accounted for, the opening of inelastic (radiative) channels poses major problems. However, here the contact to perturbative calculations may be possible and provide a valuable interface to match the different regimes, at least parametrically (e.g., in the limit of a small coupling constant and/or high temperature). This reiterates the importance of identifying the common grounds of seemingly different calculations for HQ properties in medium. We start the discussion in this Section by setting up the Brownian Motion framework for heavy quarks in the QGP (Sec. 2.1). The main part of this Section is devoted to the evaluation of the Fokker–Planck transport coefficients. We focus on elastic interactions, classified into (various levels of) perturbative (Sec. 2.2) and nonperturbative approaches (Secs. 2.3 and 2.4). As we will see, there is considerable conceptual overlap in the calculations available in the literature, the main difference being that they are carried out in different approximation schemes (Sec. 2.5). Our presentation also encompasses inelastic reactions with an additional gluon in the final and/or initial state, i.e., radiative energyloss calculations within perturbative QCD (Sec. 2.6). This raises the issue of their relative magnitude compared to elastic interactions which has recently received considerable reconsideration even for light
Heavy Quarks in the QuarkGluon Plasma
119
quarks and gluons. Finally, we address interactions of hadrons carrying charm or bottom in hadronic matter (Sec. 2.7). Bottomup extrapolations in temperature (or density) in the hadronic world are useful complements to topdown ones in the QGP, to reveal qualitative trends of, e.g., the HQ diffusion coefficient toward T c . 2.1. Heavyquark diffusion in the quarkgluon plasma As emphasized in the Introduction, an attractive feature in analyzing HQ motion in a QGP is the ensuing simplification to a Brownian motion framework.23 The latter is characterized by a Fokker–Planck equation where HQ interactions are conveniently encoded in transport coefficients. These, in turn, are readily related to underlying (elastic) scattering matrix elements on light partons in the QGP which allow for direct comparisons of microscopic models of the HQ interaction (as elaborated in subsequent sections). Starting point for the derivation of the Fokker–Planck equation23 is the Boltzmann equation for the HQ phasespace distribution, fQ , p ∂ ∂ ∂ + +F fQ (t, x, p) = C[fQ ], (1) ∂t ωp ∂x ∂p where ωp =
q
m2Q + p2 denotes the energy of a heavy quark with threemomentum
p, F is the meanfield force, and C[fQ ] summarizes the collision integral which will be analyzed in more detail below. In the following, meanfield effects will be neglected, and by integration over the fireball volume, Eq. (1) simplifies to an equation for the momentum distribution, ∂ fQ (t, p) = C[fQ ], ∂t
(2)
where fQ (t, p) =
Z
d3 xfQ (t, x, p).
(3)
The collision integral on the righthand side of Eq. (2) encodes the transition rate of heavy quarks due to collisions into and out of a small momentum cell d3 p around the HQ momentum p, Z C[f ] = d3 k[w(p + k, k)fQ (p + k) − w(p, k)fQ (p)] . (4) Here w(p, k) is the transition rate for collisions of a heavy quark with heatbath particles with momentum transfer k, changing the HQ momentum from p to p − k. Accordingly the first (gain) term in the integral describes the transition rate for HQ scattering from a state with momentum p + k, into a state with momentum p, while the second (loss) term the scattering out of the momentum state p.
120
R. Rapp & H. van Hees
The transition rate, w, can be expressed through the cross section of the collision processes in the heat bath. For elastic scattering of a heavy quark with momentum p on a light quark in the heat bath with momentum q, one finds Z dσ d3 q fq,g (q)vrel (p, q → p − k, q + k), (5) w(p, k) = γq,g (2π)3 dΩ where fq,g are the Fermi or Bose distributions for thermal light quarks or gluons, and γq = 6 or γg = 16 the respective spincolor degeneracy factors. The relative velocity is defined as p (p · q)2 − (mQ mq )2 , (6) vrel = ωQ ωq where p = (ωp , p) and q = (ωq , q) are the four momenta of the incoming heavy and light quark, respectively. Upon expressing the invariant differential cross section, P dσ/dΩ, in Eq. (5) in terms of the spincolor summed matrix element, M2 , the collision term, Eq. (4), takes the form Z Z Z 1 d3 q 0 d3 q d3 p 0 1 X C[fQ ] = M2 3 3 3 2ωp (2π) 2ωq (2π) 2ωp0 (2π) 2ωq0 γQ × (2π)4 δ (4) (p + q − p0 − q 0 )[fQ (p0 )fq,g (q 0 ) − fQ (p)fq,g (q)]
(7)
with k = p − p0 = q 0 − q. The key approximation is now that the relevant momentum transfers to the heavy quark obey k p. This enables to expand the HQ momentum distribution function, fQ , and the first argument of the transition rate, w, in the collision integral, Eq. (4), with respect to k up to second order,c w(p + k, k)fQ (p + k, k) ' w(p, k)fQ (p) + k +
∂ [w(p, k)fQ (p)] ∂p
1 ∂2 ki kj [w(p, k)fQ (p)] 2 ∂pi ∂pj
(8)
(i, j = 1, 2, 3 denote the spatial components of the 3vectors, with standard summation convention for repeated indices). The collision integral then simplifies to Z ∂ 1 ∂2 3 C[fQ ] ' d k ki + ki kj [w(p, k)fQ (p)] , (9) ∂pi 2 pi pj i.e., the Boltzmann equation (2) is approximated by the Fokker–Planck equation, ∂ ∂ ∂ fQ (t, p) = Ai (p)fQ (t, p) + [Bij (p)fQ (t, p)] . (10) ∂t ∂pi ∂pj c According
to the Pawula theorem64 any truncation of the collision integral at finite order is only consistent with fundamental properties of Markov processes if the truncation is made at the 2nd order term.
Heavy Quarks in the QuarkGluon Plasma
The drag and diffusion coefficients are given according to Eq. (9) by Z Ai (p) = d3 kw(p, k)ki , Bij (p) =
1 2
Z
121
(11)
d3 kw(p, k)ki kj .
For an isotropic background medium, especially in the case of (local) equilibrium (implying that the coefficients are defined in the local rest frame of the heat bath), rotational symmetry enables to simplify the coefficients to Ai (p) = A(p)pi , k
Bij (p) = B0 (p)Pij (p) + B1 (p)Pij⊥ (p) ,
(12)
where the projection operators on the longitudinal and transverse momentum components read k
Pij (p) =
pi pj , p2
Pij⊥ (p) = δij −
pi pj . p2
(13)
Implementing these simplifications into the collision integral, Eq. (7), the scalar drag and diffusion coefficients in Eq. (12) are given by integrals of the form Z Z Z d3 q 0 1 d3 q d3 p 0 1 X 0 hX(p )i = M2 2ωp (2π)3 2ωq (2π)3 2ωp0 (2π)3 2ωq0 γQ g,q × (2π)4 δ (4) (p + q − p0 − q 0 )fq,g (q)X(p0 ) . In this notation, the coefficients can be written as pp0 A(p) = 1 − 2 , p 1 (p0 p)2 02 B0 (p) = p − , 4 p2 1 (p0 p)2 0 2 B1 (p) = − 2p p + p . 2 p2
(14)
(15)
Note that Eq. (14) includes the sum over gluons and light quarks (u, d, s). The physical meaning of the coefficients becomes clear in the nonrelativistic approximation of constant coefficients, γ ≡ A(p) = const and D ≡ B0 (p) = B1 (p) = const, in which case the Fokker–Planck equation further simplifies to ∂ ∂ fQ (t, p) = γ [pi fQ (t, p)] + D∆p fQ (t, p) . ∂t ∂pi
(16)
E.g., for an initial condition fQ (t = 0, p) = δ (3) (p − p0 ) ,
(17)
122
R. Rapp & H. van Hees
the solution takes the form of a Gaussian distribution, o−3/2 n γ γ [p − p0 exp(−γt)]2 [1 − exp(−2γt)] exp − . fQ (t, p) = 2πD 2D 1 − exp(−2γt)
(18)
From the equation for the mean momentum,
hpi = p0 exp(−γt) ,
(19)
one sees that γ determines the relaxation rate of the average momentum to its equilibrium value, i.e., it is a drag or friction coefficient. The standard deviation of the momentum evolves according to
3D 2 p2 − hpi = [1 − exp(−2γt)] , γ
(20)
i.e., D is the momentumdiffusion constant, describing the momentum fluctuations. In the limit t → ∞, Eq. (18) approaches the (nonrelativistic) Boltzmann distribution, 3/2 2πD γp2 fQ (t, p) = exp − . (21) γ 2D Since in thermal equilibrium the heavy quarks have to obey an equilibrium distribution with the temperature, T , of the heat bath, the drag and diffusion coefficients should satisfy the Einstein dissipationfluctuation relation, D = mQ γT .
(22)
The relativistic Fokker–Planck equation will be discussed in Sec. 3.1 in connection with its formulation in terms of stochastic Langevin equations. We note that the spatial diffusion coefficient, Ds , which describes the broadening of the spatial distribution with time,
2 2 x (t) − hx(t)i ' 6Ds t , (23) is related to the drag and momentumdiffusion coefficient through Ds =
T T2 = . mQ γ D
(24)
2.2. Perturbative QCD approaches In a first step to evaluate HQ diffusion in a QGP perturbation theory has been applied, thereby approximating the medium as a weakly interacting system of quark and gluon quasiparticles. Such a treatment is expected to be reliable if the temperature is large enough for the typical momentum transfers, Q2 ∼ T 2 , to be in the perturbative regime, Q2 ≥ 2 GeV2 or so. This is most likely not satisfied for matter conditions realized at SPS and RHIC. For more realistic applications to experiment several amendments of the perturbative approach have been suggested which are discussed subsequently (focusing again on elastic HQ scattering on light partons).
123
Heavy Quarks in the QuarkGluon Plasma
g
g
g
g
Q
Q
Q
Q
g
g
q
q
Q
Q
Q
Q
Fig. 1. Feynman diagrams for leadingorder perturbative HQ scattering off light partons.
2.2.1. Schematic leading order The initial estimates of equilibration times and energy loss of heavy quarks in the QGP23 have started from the leadingorder (LO) perturbative diagrams involving the minimum of two stronginteraction vertices, as displayed in Fig. 1. Pertinent matrix elements65 figuring into Eq. (14) in the vacuum have been computed in Ref. 65. The dominant contribution arises from gluon tchannel exchange, i.e., the 3rd and 4th diagram in Fig. 1. For forward scattering, the gluon propagator develops the wellknown infrared singularity which has been regularized by introducing a Debyescreening mass, G(t) =
1 1 → , t t − µD
µD = gT ,
(25)
√ where g = 4παs denotes the strong coupling constant. Even for a value as large as αs = 0.4, and at a temperature of T = 300 MeV (typical for the early stages in heavyion collisions at RHIC), the thermal relaxation time, τeq = 1/γ, for charm (bottom) quarks turns out around ∼ 15(40) fm/c (and therefore much larger than a typical QGP lifetime of ∼ 5 fm/c at RHIC), see, e.g., right panel of Fig. 7 (in Ref. 24 the corrections due to quantumequilibrium distributions (Bose/Fermi) have been investigated and found to be small). Note that with the above gluon propagator, the pertinent total HQparton cross section is parametrically given by σQp ∝ α2s /µ2D , i.e., it essentially increases only linearly in αs (p = q, q¯, g). 2.2.2. Leading order with hard thermal loop resummation In Ref. 26, the schematic introduction of the Debye mass into the tchannel gluonexchange propagator has been extended by a LO hardthermal loop (HTL) calculation of the charmquark drag and diffusion coefficients in the QGP. In this approach, the screening of the gluon propagator in the tchannel diagrams (Fig. 1) is realized by inserting the HTL gluon propagator for the region of small momentum exchange. In Coulomb gauge, with q = q, this propagator is given by Gµν (ω, q) = −
δµ0 δν0 δij − qi qj /q 2 + 2 , + Π00 q − ω 2 + ΠT
q2
(26)
R. Rapp & H. van Hees 0.8 Nf = 0 0.7 0.6 Nf = 2
0.5
ηD(p) / ηD(p= 0)
D (αs2 T)
124
2.2
mD/T = 0.5 mD/T = 1 .0 mD/T = 1 .5
2 1.8 1.6 1.4 1.2
0.4 Nf = 3 0.3
1
d p∝p dt
0.8
Nf = 6
0.2
0.6
d p∝v dt
0.4
0.1 0 0
0.2
0.5
1
1 .5
2
2.5 mD/T
0 0
0.5
1
1.5
2
2.5
3 p/M
Fig. 2. (Color online) HQ transport coefficients in HTL improved perturbation theory. 26 Left panel: spatial diffusion coefficient at p = 0 as a function of an independently varied Debye mass, mD ≡ µD , figuring into the tchannel gluon exchange propagator, for different quarkflavor content of the (Q)GP. Right panel: momentum dependence of the drag coefficient, ηD (p) ≡ A(p), for three values of µD in tchannel gluon exchange; the lower curve, with dp/dt ∝ v, resembles a calculation in the nonrelativistic limit (M ≡ mc = 1.4 GeV in the xaxis label denotes the charmquark mass).
where the i, j ∈ {1, 2, 3} denote the spatial components of µ, ν ∈ {0, 1, 2, 3}. The HTL selfenergies read 2 ω ω(q 2 − ω 2 ) q+ω 2 ΠT (ω, q) = µD + ln − iπ , 2q 2 4q 3 q−ω (27) q+ω ω 2 ln − iπ . Π00 (ω, q) = µD 1 − 2q q−ω For small energy transfers, ω, and a slowly moving heavy quark, v 1, only the time component of the propagator contributes to the squared matrix elements which in this limit reduces to the Debyescreened Coulomblike propagator, Eq. (25). Figure 2 shows the spatial diffusion coefficient and the momentum dependence of the drag coefficient resulting from this calculation. Compared to the screening description with a constant Debye mass, the drag coefficient shows a slight increase for an intermediate range of momenta (cf., e.g., the pQCD curves in Fig. 12). 2.2.3. Leading order with running coupling As indicated in the Introduction, the current data situation at RHIC does not allow for an understanding of the electron data in terms of LO pQCD with reasonably small coupling constant (say, αs ≤ 0.4). This was a motivation for more recent studies,66,67 augmenting the LO pQCD framework in search for stronger effects. Two basic amendments have been introduced. First, the idea of Ref. 26 of introducing a reduced screening mass in the gluon propagator was made more quantitative.
Heavy Quarks in the QuarkGluon Plasma
125
Starting from an ansatz for the screened gluon propagator, Gr (t) ∝
1 . t − rµ2D
(28)
the objective is to obtain an estimate for the constant r (it was denoted κ in Refs. 66 and 67; we changed the notation to avoid conflicts in what follows below). This has been done in analogy to a corresponding QED calculation,68,69 by requiring that the energy loss of a highenergy quark obtained in a LOpQCD calculation with the screened propagator, Eq. (28), matches a calculation where for low momentum transfers, t < t∗ , the HTL propagator, Eq. (26), and for t > t∗  the perturbative gluon propagator, Eq. (25), is used; t∗  is a momentumtransfer scale between g 2 T 2 and T 2 . The QED calculation68,69 yields an energy loss which is independent of the matching scale t∗ , while this is not the case in QCD. This problem is treated by introducing an infraredregulator mass into the hard part of the energyloss integrals involving the tchannel exchangematrix elements, chosen such that the dependence on t∗  is weak for t∗  < T 2 (the validity range of the HTL approximation). This translates into effective values for the r coefficient in Eq. (28) of r ' 0.15–0.2. Second, a running strong coupling constant is introduced well into the nonperturbative regime but with an infraredfinite limit. The justification for such a procedure70 is that it can account for (lowenergy) physical observables (e.g., in e+ e− annihilation71 ) in an effective way. The parameterization adopted in Refs. 66 and 67 is based on an extrapolation of Ref. 70 into the spacelike regime, ( for Q2 ≤ 0 4π L−1 − 2 αeff (Q ) = (29) β0 1/2 − π −1 arctan(L+ /π) for Q2 > 0, where β0 = 11−2Nf /3, Nf = 3, and L± = ln(±Q2 /Λ2 ). The pertinent substitution in the tchannel gluonexchange matrix elements amounts to αeff (t) α → , t t−µ ˜2
(30)
where the regulator mass is chosen as µ ˜ 2 ∈ [1/2, 2]˜ µ2D , while the Debyescreening mass is determined selfconsistently from the equation Nc Nf 2 µ ˜D = + 4πα(−˜ µ2D )T 2 . (31) 3 6 To find the optimal value for the regulator mass a similar strategy of matching the energy loss with a Born approximation has been employed, using the substitution, Eq. (30), in the tchannel diagrams, with a HTL calculation along the same lines as summarized above for the calculation with nonrunning αs . The results for the drag coefficients for charm quarks under the various model assumptions described above are depicted in Fig. 3. Changing the screening mass from the standard Debye mass, µD , to that reproducing the HTL energy loss, with r = 0.15 in Eq. (28),
126
R. Rapp & H. van Hees
αS µ2 A 0.3 m2D B αS (2πT ) m2D C αS (2πT ) 0.15 × m2D D running (Eq. (29)) m ˜ 2D E running (Eq. (29)) 0.2 × m ˜ 2D F running (Eq. (29)) 0.11 × 6π αeff (t) T 2
line form dotted thin dashed thin full thin dashed bold full bold dashed dotted bold
figure color black black black red red purple
Fig. 3. (Color online) The drag coefficient as a function of HQ threemomentum in the amended pQCD scheme with reduced infrared regulator and running coupling constant (left panel). 66,67 The corresponding legend (right panel) details the different parameter choices in the calculation.
increases the drag coefficient by a factor of 2. In view of the large reduction in r this appears to be a rather moderate effect. This is simply due to the fact that the change mostly enhances forward scattering which is little effective in thermalizing (isotropizing) a given momentum distribution. Implementing the runningcoupling scheme with a small screening mass yields a substantial enhancement by a factor of ∼ 5. 2.2.4. Nexttoleading order The rather large values of the coupling constant employed in the calculations discussed in the previous sections imminently raise questions on the convergence of the perturbative series. This problem has been addressed in a rigorous nexttoleadingorder (NLO) calculation for the HQ momentumdiffusion coefficient, κ = 2D, in Refs. 55 and 72. This work starts from the definition of κ as the mean squared momentum transfer per unit time, which in gauge theories is given by the timeintegrated correlator of colorelectricfield operators connected by fundamental Wilson lines: Z
g2 a b κ= dtTrH W † (t, 0)Eia (t)TH W (t, 0)Eib (0)TH ; (32) 3dH W (t; 0) denotes a fundamental Wilson line running from t0 = 0 to t along the a static trajectory of the heavy quark, TH are the generators of the gauge group in the representation of the heavy quark and dH its dimension. In leading order this reduces to a Wightmantwopoint function of A0 fields at zero frequency, i.e., in the usual realtime propagator notation, κ'
CH g 2 3
Z
d3 p 2 >00 p G (ω = 0, p) , (2π)3
(33)
with CH = 4/3 the Casimir operator of the HQ representation. The integral is IR regulated by HTL corrections, i.e., a Debye mass, µ2D = g 2 T 2 (Nc + Nf /2)/3. In
Heavy Quarks in the QuarkGluon Plasma
127
αs 0.01
Q
0.05
0.1
0.2
0.3
0.4
0.6
P
R
Q
P
0.5
R (B)
Q P
P Q
0.4
κ/g 4 T 3
(A)
Nexttoleading order Leading order Truncated leading order
0.3 0.2 0.1 0
(C)
(D)
0.5
1
1.5
2
2.5
gs
Fig. 4. NLO calculations for HQ diffusion in the QGP.55 Left panel: NLO diagrams for the momentumdiffusion coefficient, κ; the double line represents the heavy quark, all propagators are soft and HTL resummed, and all vertices include HTL vertices. Right panel: comparison of LO to NLO result for κ as a function of the strong coupling, αs .
the left panel of Fig. 4 the NLO corrections to the LO result, Eq. (33), are depicted in terms of Feynman diagrams. The double line represents the heavy quark, and all propagators and vertices include HTL corrections, leading to a gauge invariant expression as it should be the case for an observable quantity like κ. The diagrams are evaluated in Coulomb gauge within the closedtime path (realtime) Keldysh formalism of thermal quantumfield theory (TQFT). The real part of diagram (A) provides a correction to the Debye mass. Diagrams (C) and (D) take into account real and virtual corrections by additional soft scattering or plasmon emission/absorption of the light or heavy scatterer, respectively. Diagram (B) represents interference between scattering events occurring on the light scatterer’s and on the heavy quark’s side. Contrary to naive power counting, the NLO calculation provides O(g) corrections due to scattering with soft gluons with momentum, q ' µD , and due to overlapping scattering events, dominated by tchannel Coulombic scatterings involving soft momentum transfers, ' µD ∝ gT . The right panel of Fig. 4 shows that the NLO correction to κ is positive, i.e., the momentumdiffusion coefficient becomes larger compared to the LO calculation. The convergence is poor even for rather small coupling constants. A rigorous resummation scheme to cure this behavior is not known to date, especially to establish convergence in the typical range of coupling constants under conditions in relativistic heavyion collisions, αs ' 0.3–0.4. In Ref. 72 the investigation of NLO corrections is extended to the weakcoupling limit of N = 4 supersymmetric YangMills (SYM) theory. Also in this case the perturbative series turns out to be poorly convergent, even for low couplings. 2.2.5. Threebody elastic scattering Another step in the (wouldbe) perturbative hierarchy are threebody collisions, which are expected to become increasingly important at high parton density. An
128
R. Rapp & H. van Hees
(a)
(b)
(c)
Fig. 5. Different topological classes of diagrams for threebody elastic scattering of a heavy quark (thick lines) off light quarks and antiquarks (thin lines).73
attempt to assess the effects of threebody elastic scattering for HQ diffusion has been conducted in Ref. 73, with pertinent Feynman diagrams as depicted in Fig. 5. Special care has to be taken in regularizing contributions from diagrams with intermediate particles going onshell; these can lead to divergent real parts in the scattering amplitude and represent successive twobody scatterings (rather than genuine threebody scattering). Therefore, in Ref. 73 the intermediate quark lines in diagram (a) and (b) are supplemented with an inmedium collisional width, and only the real part of their propagator is kept in the evaluation of the diagrams. For threebody elastic processes involving one or two gluons, it has been assumed that the dominant contributions arise from diagrams with similar topology as diagram (b) in Fig. 5 for Qqq scattering; all other contributions are neglected. To compare with twobody gluoradiative inelastic scattering the LO diagrams have been used to evaluate matrix elements for Qq → Qqg, Q¯ q → Q¯ qg and Qg → Qgg processes. Within this scheme, at temperatures T = 200–300 MeV, threebody elastic scattering processes are estimated to contribute to the c and bquark friction coefficients with a magnitude comparable to twobody elastic scattering. Again, this raises the question of how to control the perturbative series for HQ diffusion. As a byproduct, the friction coefficient for radiative scattering, Qp → Qpg, was estimated to exceed the one from elastic twobody scattering for HQ momenta p & 12 GeV (for both charm and bottom). 2.3. Nonperturbative interactions The evidence for the formation of a strongly coupled QGP (sQGP) at RHIC has motivated vigorous theoretical studies of the possible origin of the interaction strength (see, e.g., Ref. 74 for a recent review). In particular, several lattice QCD computations of hadronic correlation functions at finite temperature have found indications that hadronic resonances (or bound states) survive up to temperatures of twice the critical one or more (for both a gluon plasma (GP) and a QGP),41,42,75,76 cf. also Sec. 4.1.1 of this article. Pertinent spectral functions (extracted from Euclidean correlators using probabilistic methods, i.e., the maximum entropy method) exhibit ¯ and q q¯ channels. The consequences of hadronic resoresonance peaks in both QQ nances in the QGP for HQ transport have been elaborated in Refs. 25, 28, 77 and 45. The starting point in Refs. 25, 28 and 77 is the postulate that heavylight quark
Heavy Quarks in the QuarkGluon Plasma
129
(Q¯ q ) resonances, i.e., “D” and “B” mesons, persist in the QGP. In Refs. 25 and 28 this has been realized within an effective resonance model for Q–¯ q scattering (Sec. 2.3.1) while in Ref. 77 HQ fragmentation into mesons and their subsequent momentum broadening was considered (Sec. 2.3.3). The phenomenological success of these models (cf., e.g., the right panel of Fig. 26) called for a more microscopic evaluation of the heavylight quark correlations. This was realized in Ref. 45 where inmedium heavylight quark T matrices were computed with interaction potentials estimated from HQ free energies in lattice QCD (cf. Sec. 2.3.2). This approach is the direct analog to the potential models used in the heavy quarkonium context (cf. Sec. 4.1.1). We finish this section with a brief discussion of a recent suggestion to extract information on HQ diffusion more directly from thermal lattice QCD (Sec. 2.3.4), which would constitute a valuable benchmark for both perturbative and nonperturbative calculations. 2.3.1. Effective Q¯ qresonance model The heavylight quark resonance model25 has been set up by combining HQ effective theory (HQET) with chiral symmetry in the lightquark sector, q = (u, d), based on the Lagrangian, / / 5 1+v 0 0 ∗1+v c − q¯γ Φ c + h.c. LDcq =LD + Lc,q − iGS q¯Φ0 2 2 1 + v/ 1 + v/ − GV q¯γ µ Φ∗µ c − q¯γ 5 γ µ Φ1µ c + h.c. , (34) 2 2 written in the charm sector (an equivalent one in the bottom sector follows via the replacements c → b and D → B for the HQ and resonance fields, respectively; v: HQ fourvelocity). The pertinent free Lagrangians read 0 Lc,q = c¯(i/ ∂ − mc )c + q¯ i/ ∂q , ∗ LD0 = (∂µ Φ† )(∂ µ Φ) + (∂µ Φ0 ∗† )(∂ µ Φ∗0 ) − m2S (Φ† Φ + Φ∗† 0 Φ0 )
(35)
1 2 ∗† ∗µ Φ∗µν + Φ†1µν Φµν − (Φ∗† + Φ†1µ Φµ1 ) . 1 ) + mV (Φµ Φ 2 µν Φ and Φ∗0 denote the pseudoscalar and scalar meson fields (corresponding to D and D0∗ mesons) which are assumed to be degenerate chiral partners (mass mS ) as a consequence of chiral restoration in the QGP. The same reasoning applies to the vector and axialvector states (mass mV ), Φ∗µ and Φ1µ (corresponding to D ∗ and D1∗ ). HQ spin symmetry furthermore asserts the degeneracy of spin0 and 1 states with identical angular momentum, implying mS = mV and the equality of the coupling constants, GS = GV . In the strangequark sector only the pseudoscalar (Ds ) and vector (Ds∗ ) resonance states are considered (i.e., chiral symmetry is not imposed).
130
R. Rapp & H. van Hees
q
D k
l−k l
D
D
k
s
u
c
c
c
c
q¯
q
q¯
c
D
q
Fig. 6. Left panel: oneloop diagram representing the Dmeson selfenergy in the QGP within the effective resonance model.25 Right and middle panels: elastic Q¯ q and Qq scattering diagrams for s and uchannel resonance exchange, respectively.
The bosonresonance propagators are dressed with heavylight quark selfenergies at the oneloop level (cf. left panel of Fig. 6). To leading order in HQET, in accordance with spin symmetry, the selfenergies for the vector/axialvector resonances are given by ΠD∗ ,µν = (vµ vν − gµν )ΠD (s),
(36)
where s = p2 denotes the meson’s four momentum, and ΠD is the selfenergy of the pseudoscalar/scalar resonances. Its imaginary part reads Im ΠD (s) = −
3G2 (s − m2c )2 Θ(s − m2c ) , 8π s
(37)
while the real part is calculated from a twicesubtracted dispersion relation with the wavefunction and mass counter terms adjusted such that the following renormalization conditions hold, (ren)
∂s ΠD
(s)s=0 = 0 ,
(ren)
Re ΠD
(s)s=m2D = 0 .
As an alternative regularization scheme, dipole form factors, 2 2Λ2 , F (q) = 2Λ2 + q 2
(38)
(39)
have been supplemented to simulate finitesize vertices of the resonance model, (ff)
Im ΠD (s) = Im ΠD (s)F 2 (q) , (40) √ with q = (s − m2c )/(2 s). In this scheme, the real part is calculated from an unsubtracted dispersion relation, while the bare resonance mass is adjusted to obey the second renormalization condition in Eq. (38). With charm and bottomquark masses of mc = 1.5 GeV and mb = 4.5 GeV, the physical resonance masses are adjusted to mD = 2 GeV and mB = 5 GeV, respectively. This is in approximate accordance with earlier T matrix models of heavylight quark interactions.78,79 Likewise, the coupling constant, G, is adjusted
Heavy Quarks in the QuarkGluon Plasma
131
12
60
τ[fm]
σ [mb]
8
charm, pQCD charm, reso (Γ=0 4 GeV) + pQCD bottom, pQCD bottom, reso (Γ=0 4 GeV) + pQCD
80
reso schannel reso uchannel pQCD qc scatt. pQCD gc scatt.
10
6
40
4 20
2 0 1.5
2
3 2.5 1/2 s [GeV]
3.5
4
02
0 25
03 T [GeV]
0 35
04
Fig. 7. (Color online) Left panel: total HQ scattering cross sections off light partons in LO pQCD (blue lines) and within the effective resonance model (red lines). Right panel: thermalization times, τ = 1/A(p = 0), for charm and bottom quarks in LO pQCD with αs = 0.4 and Debyescreening mass µD = gT , compared to the results from the resonance+pQCD model, as a function of QGP temperature.
such that the resonance widths vary as ΓD,B = 0.4 . . . 0.75 GeV. The resulting heavylight quark scattering matrix elements (cf. middle and right panels of Fig. 6) have been injected into Eq. (15) to calculate HQ drag and diffusion coefficients. In the left and right panel of Fig. 7 we compare the total HQ elastic scattering cross sections and resulting thermal relaxation times, τeq = 1/A(p = 0), of the resonance model with LO pQCD (cf. the diagrams in Fig. 1). Although the total cross sections are not very different in magnitude, the thermalization times decrease by around a factor of ∼ 3–4 when adding resonant scattering, for all temperatures T = 1–2 T c . The main reason for this behavior is that schannel Q¯ q scattering is isotropic in the rest frame of the resonance, while the pQCD cross section is largely forwardpeaked (tchannel gluon exchange), and thus produces a much less efficient transport cross section (which encodes an extra angular weight). The charmquark equilibration c times in the resonance+pQCD model, τeq = 2–10 fm/c, are comparable to the expected QGP lifetime at RHIC of around τQGP ' 5 fm/c. Thus, at least for charm quarks, substantial modifications of their pt spectra towards local equilibrium in the flowing medium can be expected. The consistency of the Fokker–Planck approach can be checked with the dissipationfluctuation relation, Eq. (77), at p = 0, cf. left panel of Fig. 8. For the forwardpeaked pQCDmatrix elements, the relation is fulfilled within 3%, while with the isotropic resonance scattering deviations reach up to 11% in the renormalization scheme and up to 26% in the formfactorcutoff scheme at the highest temperatures considered (T = 400 MeV). Note however, that for a typical thermal evolution at RHIC, average fireball temperatures above T = 250 MeV are only present within the first fm/c39 ; below this temperature, the deviations are less than 5% for all cases. The right panel of Fig. 8 illustrates that (for identical resonance
132
R. Rapp & H. van Hees
05
form factor Λ=1 GeV renormalized 02
04 A[1/fm]
D/(γmQ)[GeV]
0 25
charm: total (renormalized) charm: pQCD bottom: total charm: total (form factor)
03
T=250 MeV
0 15 01
T=200 MeV
02 0 05
01 02
03 T[GeV]
0 25
0 35
0
04
1
05
15
p[GeV]
Fig. 8. Left panel: check of the dissipationfluctuation relation at p = 0 for c and b quarks with and without resonance interactions and in the renormalization and formfactor cutoff schemes. Right panel: momentum dependence of the drag coefficient in the renormalization and the formfactorcutoff scheme, where the coupling constants have been chosen in both schemes as to obtain a resonance width Γ = 0.4 GeV.
02
0 06
resonances Γ=0 3 GeV resonances Γ=0 4 GeV resonances Γ=0 5 GeV pQCD: αs=0 3 pQCD: αs=0 4 pQCD: αs=0 5
T=200 MeV
0 05 2
B0[GeV /fm]
0 15
A[1/fm]
0 07
resonances Γ=0 3 GeV resonances Γ=0 4 GeV resonances Γ=0 5 GeV pQCD: αs=0 3 pQCD: αs=0 4 pQCD: αs=0 5
T=200 MeV
01
0 04 0 03 0 02
0 05 0 01
0
1
05 p[GeV]
15
0
1
05
15
p[GeV]
Fig. 9. Drag (left panel) and transversediffusion coefficient (right panel) for pQCD and resonance model with varying interactions strengths as a function of the HQ momentum at a temperature of T = 200 MeV.
widths) the formfactor regularization scheme leads to somewhat larger (smaller) friction coefficients at low (high) momentum than the renormalization scheme. In Fig. 9 the momentum dependence of the drag and transverse diffusion coefficients is depicted using either resonancescattering or pQCDmatrix elements. Resonance scattering becomes relatively less efficient for higher HQ momenta since the centerofmass energy in collisions with thermal light antiquarks increasingly exceeds the resonance pole. The variations of the coefficients with the strong coupling constant in the pQCD scatteringmatrix elements or the resonancecoupling constant in the effective resonancescattering model are rather moderate. This is due to compensating effects of an increase of the matrix elements with α2s or G4 , on the one hand, and the accordingly increased Debyescreening mass for pQCD scattering or the broadening of the resonances widths, on the other hand.
Heavy Quarks in the QuarkGluon Plasma
q, q¯ T
=
V
+
V
=
Σglu
+
133
T
Q
Σ
T
Fig. 10. (Color online) Diagrammatic representation of the Brueckner manybody calculation for the coupled system of the T matrix based on the lQCD static internal potential energy as the interaction kernel and the HQ selfenergy.
2.3.2. Inmedium T matrix with lQCDbased potentials The idea of utilizing HQ free energies computed in lattice QCD to extract a driving kernel for heavylight quark interactions in the QGP has been carried out in Ref. 45, with the specific goal of evaluating HQ diffusion. Since the latter is, in principle, determined by lowenergy HQ interactions, the potentialmodel framework appears to be suitable for this task. Moreover, with a potential extracted from lQCD, the calculation could be essentially parameterfree. Currently, however, such an approach bears significant uncertainty, both from principle and practical points of view, e.g., whether a welldefined potential description can be constructed in medium80–82 and, if so, how to extract this information from, say, the HQ free energy. In the vacuum, both questions have been answered positively,83,84 thus validating the 30 yearold phenomenological approaches using Cornell potentials for heavy quarkonia, which provide a very successful spectroscopy.85 The potential approach has been extended to heavylight mesons in Refs. 86 and 87. A Bruecknerlike inmedium T matrix approach for heavylight quark scattering in the QGP has been applied in Ref. 45, diagrammatically represented in Fig. 10. The underlying (static) twobody potential has been identified with the internal energy U1 (r, T ) = F1 (r, T ) + T S1 (r, T ) = F1 (r, T ) − T
∂F1 (r, T ) , ∂T
(41)
extracted from two lQCD computations of the colorsinglet HQ free energy above Tc , for quenched88 and twoflavor89 QCD (pertinent parameterizations are given in Refs. 90 and 44, hereafter referred to as [Wo] and [SZ], respectively). This choice (rather than, e.g., the free energy) provides an upper limit for the interaction strength.44,58,90,91 To use Eq. (41) as a potential in a T matrix calculation, the internal energy has to be subtracted such that it vanishes for r → ∞, V1 (r, T ) ≡ U1 (r, T ) − U1 (r → ∞, T ) ,
(42)
134
R. Rapp & H. van Hees
which is dictated by the convergence of the T matrix integral in momentum space. It is suggestive to interpret the asymptotic value U1∞ ≡ U1 (r → ∞, T ) as an inmedium HQmass, 1 mQ (T ) = m0Q + U1∞ , 2
(43)
where m0Q denotes the bare HQ mass (e.g., m0c ' 1.25 GeV92 for the bare cquark mass). However, close to Tc , the values for U1∞ (T ) extracted from lQCD calculations develop a rather pronounced peak structure,93,94 which renders a mass interpretation problematic. Progress in understanding these properties is closely connected with the proper identification of the potential. First lQCD estimates of the inmedium HQ mass (extracted by relating zeromode contributions to quarkonium correlators to the HQ susceptibility) indicate a moderate increase when approaching Tc from above.46 In Ref. 45 constant (average) inmedium charm and bottomquark masses of mc = 1.5 GeV and mb = 4.5 GeV, respectively, have been employed. The T matrix approach is readily generalized to colorconfigurations other than the singlet channel of the qQ pair. The complete set of color states for Q¯ q (singlet and octet) and Qq (antitriplet and sextet) pairs has been taken into account assuming Casimir scaling as in pQCD, 1 V8 = − V1 , 8
V¯3 =
1 V1 , 2
1 V 6 = − V1 , 4
(44)
which is, in fact, supported by finiteT lQCD.95,96 To augment the static (colorelectric) potentials with a minimal relativistic (magnetic) correction for moving quarks,97 the socalled Breit correction as known from electrodynamics98 has been implemented via the substitution Va → Va (1 − α ˆ1 · α ˆ2 ),
(45)
where α ˆ 1,2 are quasiparticle velocity operators. The above constructed heavylight potentials can now be resummed in a twobody scattering equation. In accordance with the static nature of the potentials, it is appropriate to use a threedimensional reduction of the full fourdimensional BetheSalpeter equation. This leads to the wellknown ladder series which is resummed by the Lippmann–Schwinger (LS) integral equation for the T matrix In the q–Q centerofmass (CM) frame it takes the form Z d3 k 0 0 Ta (E; q , q) = Va (q , q) − Va (q 0 , k) GqQ (E; k) (2π)3 × Ta (E; k, q) [1 − fq (ωkq ) − fQ (ωkQ )] .
(46)
The driving kernel (potential) can now be identified with the Fourier transform of the coordinatespace potential extracted from lQCD, Z Va (q 0 , q) = d3 r Va (r) exp[i(q − q 0 ) · r] (47)
135
Heavy Quarks in the QuarkGluon Plasma
in a given color channel, a ∈ {1, ¯ 3, 6, 8}. The concrete form of the intermediate q–Q (or q¯–Q) propagator, GqQ (E, k), depends on the reduction scheme of the underlying BetheSalpeter equation. It has been verified45,91 that, e.g., the Thompson99 and BlancenbeclerSugar100 scheme lead to very similar results in the present context (as was found for nucleonnucleon scattering). In the former, the twoparticle propagator is given by GqQ (E, k) =
1 E−
(ωkq
+
iΣqI )
− (ωkQ + iΣQ I )
,
(48)
where E and k denote the CM energy and relative momentum of the qQ pair, respectively. The quasiparticle widths are chosen as Γq,Q = −2Σq,Q = 200 MeV, I I and the light quark masses as constant at mq = 0.25 GeV, with onshell energies q (49) ωkq,Q = m2q,Q + k 2 .
The latter figure into the Pauli blocking factor with FermiDirac distributions, fq,Q (ω q,Q ) =
1 exp(ω q,Q /T ) + 1
(50)
(at the considered temperatures their impact is negligible). The solution of the T matrix Eq. (46) is facilitated by a an expansion into partial waves, l, X Va (q 0 , q) = 4π (2l + 1) Va,l (q 0 , q) Pl [cos ∠(q, q 0 )] , l
0
Ta (E; q , q) = 4π
X
(2l + 1) Ta,l (E; q 0 , q) Pl [cos ∠(q, q 0 )] ,
(51)
l
which yields a onedimensional LS equation, Z 2 Ta,l (E; q 0 , q) = Va,l (q 0 , q) + dkk 2 Va,l (q 0 , k) GQq (E; k) π × Ta,l (E; k, q) [1 − fF (ωkQ ) − fF (ωkq )] ,
(52)
for the partialwave components, Ta,l , of the T matrix. Equation (52) can be solved numerically by discretization and subsequent matrixinversion with the algorithm of Haftel and Tabakin.101 The resulting Swave (l = 0) T matrices indeed show resonance structures in a QGP in the channels where the potential is attractive, i.e., in the meson (colorsinglet) and diquark (colorantitriplet) channels. The pertinent peaks in the imaginary part of the T matrix develop close to the Q–q threshold, and melt with increasing temperature at around 1.7 Tc and 1.4 Tc , respectively (cf. left panel of Fig. 11). In the repulsive channels, as well as for P waves, the T matrices carry much reduced (nonresonant) strength. The increasing strength in the meson and diquark channels (the latter relevant for baryon binding) when approaching Tc from above is suggestive for “prehadronic” correlations toward the hadronization transition.
136
R. Rapp & H. van Hees
6 5
0
4 ΣQ (MeV)
2
Im T (GeV )
50
sing, T=1 2 Tc sing, T=1 5 Tc trip, T=1 2 Tc trip, T=1 5 Tc
s wave
3 2
50 100 T=1 1 Tc, real T=1 1 Tc, imag T=1 4 Tc, real T=1 4 Tc, imag T=1 8 Tc, real T=1 8 Tc, imag
150
1 200
0 0
1
2 Ecm (GeV)
3
4
0
1
2 p (GeV)
3
4
Fig. 11. (Color online) Results of a Bruecknertype approach for cquarks in a QGP 45 based on a potential corresponding to the internal energy extracted from quenched lattice QCD. 88,90 Left panel: imaginary part of the inmedium T matrices for Swave c–q scattering in colorsinglet and triplet channels at two different temperatures; right panel: real and imaginary parts of cquark selfenergies based on the T matrices in the left panel.
The next step is to use the T matrices to compute the lightquark contribution to HQ selfenergies, i.e., the last diagram in the second line of the Brueckner scheme illustrated in Fig. 10. In a given colorchannel, a, the T matrix induced selfenergy is given by Z 2 k dkdx dSI da ΣQ [fF (ωk ) + fB (ω + ωk )] Ta (E; p, k) , (53) (ω, p) = a 6 4π 2 where dSI = 4(2l + 1)Nf denotes the spinisospin and angular momentum degeneracy of all Qq (or Q¯ q) configurations (assuming spin and lightflavor symmetry) and da the color degeneracy of channel a; the factor 1/6 averages over the incoming HQ colorspin degrees of freedom. The resulting charmquark selfenergies (summed over all light quarks and antiquarks) are displayed in the right panel of Fig. 11. One finds rather small corrections to the HQ mass (presented by the real part of Σ), but the imaginary parts are substantial, Γc = −2 Im Σc ' 100–300 MeV for temperatures T = 1.1–1.8 Tc (with the largest values attained close to Tc ). These values were the motivation for the choice of input widths in the propagator, Eq. (48), of the T matrix equation, thus providing a rough selfconsistency. The inmedium mass corrections, on the other hand, are associated with the gluonic contribution to the HQ selfenergy (corresponding to the first term in the lower line Fig. 10), which have not been calculated explicitly in Ref. 45, but roughly represent the (average) (1) asymptotic values of the HQ potential, U∞ (as discussed above). The final step is to implement the T matrix elements into a calculation of HQ drag and diffusion coefficients via Eq. (15); one finds X
M2 =
64π (s − m2q + m2Q )2 (s − m2Q + m2q )2 s2 X × Nf da (Ta,l=0 (s)2 + 3Ta,l=1 (s) cos(θcm )2 ) . a
(54)
Heavy Quarks in the QuarkGluon Plasma 0 15
0 15
Tmatrix [Wo]: 1 1 Tc Tmatrix [Wo]: 1 5 Tc Tmatrix [Wo]: 1 8 Tc pQCD: 1 1 Tc pQCD: 1 4 Tc pQCD: 1 8 Tc
Tmatrix [SZ]: 1 1 Tc Tmatrix [SZ]: 1 5 Tc Tmatrix [SZ]: 1 8 Tc pQCD: 1 1 Tc pQCD: 1 4 Tc pQCD: 1 8 Tc
01 Α (1/fm)
Α (1/fm)
01
0 05
137
0 05
0
0 0
1
2
3
4
5
0
1
p (GeV)
2
3
4
5
p (GeV)
Fig. 12. (Color online) The drag coefficients at different temperatures, using the parameterization of the HQ potential from [Wo] (left panel) and [SZ] (right panel) compared to LO pQCD with αs = 0.4 and µD = gT .
The resulting friction coefficients are summarized in Fig. 12 as a function of momentum for three temperatures and for two potential extractions from lQCD. 44,90 Generally, the Qq T matrix based coefficients are largest at low HQ momentum, as to be expected from the resonance formation close to threshold. The values exceed the LOpQCD coefficients at small temperatures and for both potentials by a factor of ∼ 3–5. At higher temperatures the enhancement reduces considerably, to a factor of less than 2 in the [SZ] potential and to essentially equal strength for the [Wo] potential.d In fact, the coefficients computed with the [SZ] potential have a slightly (30%) increasing trend with T , while the [Wo] potential leads to a decreasing trend. This difference needs to be scrutinized by future systematic comparisons of lQCD input potentials. Compared to the resonance model (cf., e.g., left panel of Fig. 9), the T matrix calculations yield quantitatively similar results at temperatures not too far above Tc but become smaller at higher T due to resonance melting (which is presumably a more realistic feature of a nonperturbative interaction strength). 2.3.3. Collisional dissociation of heavy mesons in the QGP In Ref. 77, a socalled reactionoperator (GLV) approach has been applied to resum multiple elastic scatterings of a fast Q¯ q pair. The quenching of heavy quarks in a QGP is calculated by solving coupled rate equations for the fragmentation of c and b quarks into D and B mesons and their dissociation in the QGP. The main mechanisms for HQ energy loss are collisional broadening of the meson’s transverse d Recall
that the LOpQCD calculations employ a rather large coupling (α s = 0.4), and are dominated by scattering off gluons in the heat bath; thus a minimal merging of the gluon sector with the T matrix calculations would consist of adding the gluonic part of LOpQCD; this procedure is adopted below whenever combined results are shown or utilized. In principle, a nonperturbative treatment should also be applied to HQgluon scattering.
138
R. Rapp & H. van Hees
momentum and the distortion of its intrinsic lightcone wave function. The latter is modeled in a Cornelltype potential ansatz87 (note that this bears some similarity to the T matrix approach discussed in Sec. 2.3.2). This inmedium HQ/heavymeson fragmentation/dissociation mechanism leads to comparable highpt suppression for B and D mesons, which is quite contrary to perturbative calculations for both collisional and radiative energy loss (where the suppression of b quarks is significantly less pronounced than for c quarks). This feature largely results from the much smaller formation times of Bmesons compared to Dmesons, leading to a faster fragmentationdissociation cycle for b quarks/B mesons. 2.3.4. Estimates of HQ diffusion in lattice QCD It has recently been suggested that, unlike in the case of other transport coefficients (e.g., the shear viscosity), the HQ diffusion coefficient might be amenable to a determination within lQCD, based on an analytic continuation of the colorelectricfield correlator along a Polyakov loop102 (see also Ref. 103 for earlier related work). The starting point of these considerations is the spectral function of the HQ current correlator, h Z Z i 1 µ ν 3 J (t, x), J (0, 0) , (55) ρµν (ω) = dt exp(iωt) d x Q V 2 Q
µ where JQ denotes the HQ current operator in the Heisenberg picture. The spatial diffusion coefficient, Ds , can be extracted from this spectral function by the pole position at ω = −iDs k2 , where k is the HQ momentum. The condition for a pole leads to the Kubo relation
Ds =
3 X 1 ρii V (ω) lim , 00 3χ ω→0 i=1 ω
where χ00 is the conservedcharge susceptibility Z
0 1 00 0 χ = d3 x JQ (t, x)JQ (0, 0) . T
(56)
(57)
For a heavy quark, the spectral function, Eq. (55), is expected to develop a sharp peak around ω = 0 which can be described by a Lorentzian function close to this point. The width of this function is given by the drag coefficient, which obeys the fluctuationdissipation relations, discussed in Sec. 2.1. Using HQ effective theory techniques it is shown that in the static limit the momentumdiffusion coefficient, 0 κ = 2D, is given by a correlator involving colorelectric fields and JQ operators whose Euclidean analogue can be mapped to an expectation value involving Wilson lines and colorelectric fields, similar to Eq. (32). This purely gluonic correlation function can in principle be evaluated in lQCD. Another latticebased approach to assess HQ diffusion has been suggested in Ref. 104 in terms of (discretized) classical gauge theory. The limitation of this
Heavy Quarks in the QuarkGluon Plasma
139
approach is set by the thermal (hard) scale ∼ πT where quantum theory suppresses excitations. Since the HQ thermalization rate is expected to be governed by the electric screening scale ∼ gT , this limitation may not be severe for small and moderate coupling (gT T ), and thus allow for valuable insights. First, it has been verified that, in the weak coupling limit, the discretized (nonperturbative) classical computation indeed agrees well with pQCD. Upon increasing the coupling strength, the classical lattice results for the HQ momentum diffusion coefficient increasingly exceed the LO perturbative result, by about an order of magnitude for a moderate coupling strength corresponding to αs ' 0.2. Next, the NLO term (with slightly increased strength to account for HTL effects), as calculated in Ref. 55, has been added to the LO calculation which extends the agreement of pQCD with the classical lattice results to larger (but still weak) coupling. For αs ' 0.2 the increase over LO amounts to a factor ∼ 2, which means that the classical lattice result remains substantially larger (by a factor of ∼ 5) than the NLO value. Besides reconfirming the poor convergence of pQCD, this also suggests that the perturbative series is not alternating but that higher order terms keep increasing the value of the HQ momentum diffusion coefficient. Semiquantitatively, such an enhancement is in the ball park of the factor ∼ 3–4 found in the effective resonance model (Sec. 2.3.1) or T matrix approach (Sec. 2.3.2). 2.4. String theoretical evaluations of heavyquark diffusion The conjectured correspondence between certain classes of string theories, formulated in fivedimensional AntideSitter space (AdS5 ), and gauge theories with conformal invariance (conformal field theory, CFT) has opened interesting possibilities to address nonperturbative aspects of QCD. The socalled AdS/CFT correspondence implies a “duality” of a weakly coupled gravity to a strongly coupled supersymmetric (and conformal) gauge theory, specifically N = 4 SU(Nc ) superYangMills (SYM) theory. This connection has been exploited to formulate the problem of HQ diffusion at finite temperature and extract an “exact” nonperturbative result for HQ transport coefficients in the SYM plasma.105–107 The translation to QCD matter is beset with several caveats,108 e.g., the particle content of the SYM medium is quite different compared to the QGP. While this may be corrected for by a suitable rescaling of the temperature by matching, e.g., the energy densities, e a more problematic difference is the absence of a scale (other than temperature) in conformal SYM. Thus, the latter does not possess a breaking of scale invariance, a running coupling constant, confinement nor spontaneous chiral symmetry breaking, and consequently no notion of a critical temperature, either. Thus SYM is quite different from QCD in the zero and lowtemperature regimes. However, at sufficiently high T , where the QCD medium deconfines its fundamental charges, e This
procedure works quite well when comparing quantities in quenched and unquenched lattice QCD computations, e.g. for the critical temperature.
140
R. Rapp & H. van Hees
the resemblance to SYM might be much closer. E.g., the pressure in SYM in the strong coupling limit amounts to about 75% of the StefanBoltzmann limit, close to what is found in thermal lattice QCD for a wide range above Tc . In addition, the finding of an extremely small shear viscosity in strongly coupled SYM, η/s = 1/4π (conjectured to be a universal lower bound),17 and the apparently lowviscosity QCD medium deduced from the success of hydrodynamic models at RHIC, is another good reason to further pursue exact nonperturbative calculations in SYM for quantities that are relevant for RHIC phenomenology. If nonperturbative effects in the strongly coupled QGP at moderate temperatures, T = 1–2 Tc , are ultimately connected to the presence of the phase change(s) (and thus inherently to the critical temperature as a relevant scale), the CFTQCD connection would not be a rigorous one. But even in this case, the nonperturbative computation of transport coefficients of a strongly coupled system at a given reference temperature “not too close” to Tc should provide useful insights. The first step in computing HQ diffusion for CFT is the introduction of a heavy quark into the conformal field theory. This can be achieved by either introducing a heavy charge via breaking the gauge group from Nc + 1 to Nc (which, strictly speaking, generates (2Nc + 1) “Higgsed” “W ” bosons),109 or by adding a finitemass N = 2 hypermultiplet with charges in the fundamental representation as a “probe” of the CFT medium. In either case, the pertinent object on the 4dimensional boundary of the 5dimensional AdS space represents a fundamental charge. In Refs. 105 and 106, the HQ drag has been evaluated by computing its momentum degradation, dp/dt = −γp, through the force on the trailing string, resulting in a friction (or drag) coefficient, γAdS/CFT
√ 2 π λTSYM , = 2mQ
(58)
2 where λ = gSYM Nc denotes the ’t Hooft coupling constant. Alternatively, in Ref. 107 the problem was formulated focusing on the diffusion term. For time scales longer than the thermal relaxation time of the medium, but short compared to the HQ relaxation time, the fluctuation term in the Langevin equation (63) dominates over the drag term. The evaluation of the noise (or force) correlator is then carried out via the fluctuations of the string, resulting in a noise coefficient which is directly related to the diffusion coefficient (cf. Eqs. (65) and (74) below). Furthermore, the latter can be related to the friction coefficient using the Einstein relation, Eq. (22); it turns out that the result is identical to Eq. (58), which also verifies that a Langevin process consistent with the fluctuationdissipation theorem applies in the SYM theory (see, however, Refs. 110 and 111, where the applicability of the Langevin framework in AdS/CFT for highmomentum quarks is discussed). The squareroot dependence of γAdS/CFT on the coupling constant λ clearly characterizes its nonperturbative nature; in this sense it is parametrically large for comparatively small coupling constants. The temperature dependence is rather “conventional”, as to be expected
Heavy Quarks in the QuarkGluon Plasma
141
since there are no additional scales in the problem (the HQ mass in the denominator implies the standard suppression of the HQ relaxation rate by ∼ T /mQ ). The next question is how to convert the result into a (semi) quantitative estimate for the QCD plasma. Naively, one may just insert the values of the strong coupling constant, gs , and QGP temperature, T , for gSYM and TSYM , respectively. A more suitable identification probably consists of matching physical quantities which leads to somewhat different parameter values. E.g., in Ref. 108, comparable temperatures were identified by matching the energy densities (ε) of the QGP and SYMplasma. Since the latter has a factor ∼ 3 larger particle content (degeneracy factor), one has a smaller temperature at the same ε, TSYM ' T /31/4 . For the coupling constant, one can exploit the fact that in AdS/CFT the potential between a heavy charge and anticharge is essentially of Coulombtype, both at zero 109 and finite temperature.112,113 In the latter case, the potential goes to zero at some finite range, characteristic for Debyescreening behavior. This range can be used to identify the length scale in comparison to typical screening radii of heavyquark free energies as computed in thermal lattice QCD (although some ambiguity remains).108 Matching the magnitude of the potentials at the screening radius then allows for a matching of the coupling constants. This leads to significantly smaller values for λ (by a factor of 36) than the naive identification with αs = 0.5. In connection with the redefined temperature, the improved AdS/CFTbased estimate for the HQ friction coefficient in QCD amounts to γ ' 0.3–0.9 c/fm at T = 250 MeV, which is significantly smaller than the “naive” estimate of ∼ 2 c/fm. 2.5. Comparison of elastic diffusion approaches In view of the recent proliferation of seemingly different approaches to evaluate HQ transport coefficients in the QGP it becomes mandatory to ask to what extent they are related and encode similar microscopic mechanisms.114 It turns out that all of the approaches discussed above incorporate a color Coulombtype interaction. This is rather obvious for the T matrix approach, where the input potentials from lattice QCD clearly exhibit the Coulomb part at sufficiently small distance (including effects of color screening). The onegluon exchange in pQCD (which is the dominant contribution to HQ rescattering, recall the two right diagrams in Fig. 1), also recovers the Coulomb potential in the static limit (color screening enters via the Debye mass in the spacelike gluonexchange propagator). The collisional dissociation mechanism involves the Cornell potential for the D and Bmeson wave functions and thus incorporates a Coulomb interaction as well; the emphasis in this approach is on formationtime effects essentially caused by the different (free) binding energies of D and B mesons. In addition, the confining part of the Cornell potential may play a role (as in the T matrix approach). Finally, in conformal field theory (AdS/CFT), the absence of any scale promotes the Coulomb potential to the unique form of a potential, V (r) ∝ 1/r (this is the only way of generating a quantity with units of energy). On the other hand, scalebreaking effects are present in the
142
R. Rapp & H. van Hees
QCDbased approaches in terms of a running coupling constant (pQCD), while the Cornell and lQCDbased potentials additionally feature linear terms ∝ σr where the string tension introduces a further (nonperturbative) scale. In fact, in Ref. 94 it has been argued, based on an analysis of lQCD results for the heavyquark free (and internal) energy, that “remnants” of the confining force play a prominent role for temperatures not too far above Tc (e.g., for heavy quarkonium binding). If one assumes the prevalence of the Coulomb interaction, the obvious first question is with what strength (coupling constant) it figures into the different approaches, which should be fairly straightforward to determine. A more involved issue is to scrutinize the underlying approximation schemes and their applicability. E.g., perturbative approaches with large (running) coupling constants have poor (if any) control over higherorder corrections. As usual in such situations, diagrams with large contributions should be identified and resummed (which is, of course, a nontrivial task, e.g., maintaining gauge invariance); it would be illuminating to extract a static gluonexchange (Coulomb) potential for a given set of parameters. The T matrix approach performs a resummation of the ladder series of a static (colorelectric) potential; magnetic interactions are implemented in a simplified manner using the Breit currentcurrent interaction from electrodynamics. It has been verified that for large centerofmass energies, the qQ T matrix recovers the result for perturbative scattering. However, a number of effects are neglected and need to be scrutinized, including the interactions with gluons beyond pQCD, retardation, extra gluon or particle/antiparticle emission (e.g., in a coupled channel treatment) and the validity (and/or accuracy) of a potential approach at finite temperature (this issue will reappear in the context of heavy quarkonia in Sec. 4). In the collisional dissociation approach, it would be interesting to explore medium effects in the employed potential (i.e., on the mesonic wave function). Ideally, by improving on specific assumptions in a given approach, an agreement would emerge establishing a common result. Explicit connections with the AdS/CFT results are more difficult to identify. Maybe it is possible to push the T matrix approach into a regime of “large” coupling, or study the existence and properties of (D and B) bound states in the string theory setting. In Fig. 13 we summarize the drag coefficients as function of momentum (for 3 temperatures, left panel) and temperature (for p = 0, right panel) resulting from the approaches discussed above, i.e., (i) leadingorder pQCD calculations with fixed αs = 0.4 and Debyescreening mass, µD = gT , in the tchannel gluonexchange contributions to the matrix elements for elastic gQ and qQ scattering, (ii) inmedium T matrix calculations using lQCDbased qQ potentials, augmented by the leadingorder pQCD matrix elements for elastic gQ scattering,45 (iii) pQCD calculations with running αs and reduced screening mass,66,67 and (iv) the AdS/CFT correspondence matched to QCD108 with γQCD = CT 2 /mQ for C = 1.5–2.6.115
Heavy Quarks in the QuarkGluon Plasma 1
1 1 Tc 1 4 Tc 1 8 Tc
0.5
0.8
γ (1/fm)
γ (1/fm)
0.4
0.3
143
pQCD, αs=0.30.4 reso+pQCD Tmatrix+pQCD AdS/CFT pQCD, αs run
0.6
0.4
0.2
0.2
0.1
0 0
1
2
3 p (GeV)
4
5
0 0.2
0.25 T (GeV)
0.3
Fig. 13. (Color online) Charmquark friction coefficients, γ, in the QGP. Left panel: threemomentum dependence at three temperatures (color code) for: LOpQCD with fixed α s = 0.4 and µD = gT (dashdotted lines), heavylight quark T matrix plus LOpQCD for gluons (solid lines),45 and pQCD with running αs and reduced infrared regulator (dashed lines).66,67 Right panel: temperature dependence of γ for LOpQCD, T matrix plus LOpQCD (gluons only), pQCD with running αs , and from AdS/CFT correspondence matched to QCD108 with C = 1.5–2.6.115
At all temperatures, the T matrix approach, (ii), produces significantly more HQ interaction strength than LO pQCD, (i), while for T > 0.2 GeV the thermalization rate for the T matrix is a factor of ∼ 2–4 less than for AdS/CFT, (iv), or for LOpQCD with running coupling and reduced infrared regulator, (iii). Close to Tc ' 180 MeV, however, the 3 approaches (ii), (iii) and (iv) are not much different and share overlap around γ' 0.2 c/fm. The spread in the numerical results reiterates the necessity for systematic checks as indicated above. Finally, one can convert the drag coefficients into estimates of other HQ transport coefficients of the QGP. Within the Fokker–Planck approach the spatial diffusion coefficient, Ds , is directly related to the drag coefficient, γ, as given by Eq. (24). Figure 14 shows the dimensionless quantity 2πT Ds for charm (left panel) and bottom quarks (right panel) as a function of temperature for LO pQCD, LO pQCD with running coupling and reduced infrared regulator, effective resonance model and T matrix approach. The former three are fairly constant as a function of temperature while the T matrix approach exhibits a significant increase with temperature, indicating maximal interaction strength close to Tc . This originates from the increasing potential strength (decrease in colorscreening) with decreasing temperature, enhancing resonance correlations at lower temperature. It is tempting to interpret this feature as a precursor phenomenon of hadronization. However, its robustness needs to be checked with a broader range of lattice potentials. We recall that the internalenergy based potentials probably provide an upper estimate for the strength of the interaction. It is interesting to note that for all approaches the results for b quarks coincide with the ones for c quarks within ∼ 20–30%. The largest deviation is seen in the T matrix approach, where the (spatial) diffusion coefficient is smaller for b quarks than for c quarks (Bmeson resonances survive until higher temperatures than D resonances). This is qualitatively similar to what has been
144
R. Rapp & H. van Hees
30 charm quarks
20
2πT Ds
2πT Ds
20
30 Tmat+pQCD reso + pQCD pQCD pQCD run αs
10
Tmat + pQCD reso + pQCD pQCD pQCD run αs
bottom quarks
10
0
0 02
0 25
03 T (GeV)
0 35
04
02
0 25
03 T (GeV)
0 35
04
Fig. 14. (Color online) Spatial diffusion coefficient, Ds = T /(γmQ ), for c (left) and b quarks (right) in a QGP for: LOpQCD with fixed αs = 0.4 (dashed lines), effective resonance model + LOpQCD (bands for ΓD,B = 0.4–0.75 GeV)25 , T matrix approach + LOpQCD (gluons only)45 and pQCD with running αs (dashdotted line).66,67 The AdS/CFT result corresponds to 2πT Ds = 2π/C ' 1.5 − 4 (not shown in the plots).
found for the collisional dissociation mechanism, where the relative enhancement of the bquark energy loss (compared to charm) is due to smaller Bmeson formation times. Since the latter are related to larger Bmeson binding energies, the dynamical origin of the smaller Ds for b quarks appears to be of similar origin as in the T matrix approach. The relative magnitudes of the various approaches reflect what we discussed before for the drag coefficient. 2.6. Collisional versus radiative energy loss For slowly moving heavy quarks in the QGP, the parametrically dominant interaction is elastic scattering. However, at high pT , radiative scattering is believed to eventually become the prevailing energyloss mechanism. It is currently not known at which pT this transition occurs. Therefore, it is important to assess the relative importance between elastic and inelastic scattering processes in the medium, even at the level of perturbative scattering only. Toward this purpose, we first recollect basic results on the gluonBremsstrahlung mechanism for lightparton, and then HQ, energy loss in the QGP, followed by a direct comparison to collisional energy loss for heavy quarks. A seminal perturbative treatment of gluoradiative energy loss (Eloss) of highenergy partons in the QGP has been given in Refs. 116 and 117 (BDMPS). The medium is modeled as static scattering centers which implies that the Eloss is purely radiative. The key finding is that the Eloss due to multiple inmedium scattering of a highenergy parton grows as L2 , where L is the path length of the parton traversing the medium. The static scattering centers, at positions xi , are described by screened Coulomb potentials, Vi (q) =
q2
g exp(−iqxi ) . + µ2D
(59)
Heavy Quarks in the QuarkGluon Plasma
145
The range of the potentials is assumed to be small compared to the mean free path, λ, of the scattered parton, i.e., 1/µD λ. In this case successive scatterings can be considered as independent, thus enabling an eikonal approximation for the elastic scattering on static centers, i.e., a classical propagation of the particle with energy E µD , undergoing independent kicks, thereby radiating Bremsstrahlung gluons. In analogy to the QED case an important ingredient is the coherent resummation of the multiplescattering Bremsstrahlung amplitudes (“LandauPomeranchukMigdal effect”) which can be formulated as a diffusion equation for the effective scattering amplitudes (or pertinent currents). The total radiative Eloss of a highenergy parton traversing a medium of path length L is then given by ∆E =
αs 2 qˆL , 2
(60)
where qˆ is the diffusion coefficient for transversemomentum broadening in scat tering off the static scattering centers, qT2 = qˆL. Perturbative calculations of the transport coefficient result in a value of about qˆ ' 1 GeV/fm2 at typical energy densities of ' 10 GeV/fm3 (translating into T ' 250 MeV) relevant for the QGP at RHIC.118 It turns out, however, that the description of highmomentum pion suppression at RHIC in the BDMPS formalism requires an approximate tenfold increase of the perturbative value for qˆ.29 Recent calculations of perturbative Eloss including both elastic and radiative contributions within a thermalfieldtheory framework indicate that collisional Eloss may be significant even for highpT light partons.30 This would imply a reduction of the value required for qˆ from RHIC phenomenology. An early calculation119 of radiative charmquark Eloss, −dE/dx, in the QGP has found that it dominates over the elastic one down to rather small momenta, p ≤ 2 GeV.24 In Ref. 32 it has been pointed out that the application of radiative Eloss to heavy quarks leads to the appearance of the socalled “dead cone”, i.e., the suppression of forward gluon radiation for Θ < mQ /E, where Θ denotes the direction of motion of the gluon with energy E, relative to the direction of the HQ momentum.32 It has been predicted that the reduced HQ Eloss leads to a heavytolight hadron ratio above one in the highpT regime accessible at RHIC. Within the BDMPS model, extended to heavy quarks, it has been argued,120 however, that mediuminduced gluon radiation tends to fill the dead cone. As will be discussed in Sec. 3.5, a similar value for qˆ as in the lighthadron sector is necessary to come near the observed suppression of highpT electrons from HQ decays in terms of radiative Eloss alone.36 The BDMPS formalism for light partons has been generalized to resum an expansion of gluoradiative parton Eloss in the GP with opacity, n ¯ = L/λ, employˆ n is ing a socalled reaction operator approach121 (GLV). A reaction operator R th constructed that relates the n power in a opacityinclusive radiation probability distribution to classes of diagrams of order n−1. This results in a recursion relation for the radiation probability distribution, corresponding to a certain resummation
146
R. Rapp & H. van Hees
1 dNg/dy = 1000, L = 5 fm
dNg/dy = 1000
st la E
Rad D
GLV
0.6
c Ela
stic
b
b
RQ(pT)
c  Rad DGLV
ic
Quark ∆E / E
only
tic las
d u,
0.2
Rad
+E
0.8 u,d 
DGLV rad + TG elastic
d Ra
L=5fm
0.3
u c 0.4
BT TG
0.1 b
GLV  Rad D
TG
u c
0.2
g
b Elastic BT
g 0
5
15
10 E (GeV)
20
0
0
2
4
6
8
10 12 pT (GeV)
14
16
18
20
Fig. 15. (Color online) Left panel: average relative Eloss, ∆E/E, for u, c and , b quarks as a function of jet energy, E, in a longitudinally (Bjorken) expanding QGP, with fixed path length L = 5 fm, initial gluon √ rapiditiy density dNg /dy = 1000 and fixed αs = 0.3; the gluon (lightquark) mass is set to µD / 2 (µD /2), the c(b)quark mass to mc = 1.2(4.75) GeV (solid lines: radiatve Eloss, dashed bands: elastic Eloos in 2 schemes as discussed in the text). Right panel: parton nuclear modification factor, RAA ≡RQ , for gluons, u, c, and bquarks as a function of pT for a fixed path length and dNg /dy = 1000 (dashed lines: radiative Eloss, solid lines: radiative+elastic Eloss).
to all orders in opacity, which can be implemented in MonteCarlo simulations for jet quenching. The GLV reactionoperator method for lightparton radiative Eloss in the QGP has been extended to heavy quarks in Ref. 122 (DGLV), implementing the kinematical suppression of gluon radiation by the HQ mass in the “dead cone”. A direct study of the relative magnitude of collisional (elastic) and radiative pQCD HQ Eloss in the GP has been undertaken in Ref. 38. For the elastic Eloss of a parton with color Casimir constant, CR , the leading logarithm expression in an ideal QGP with Nf effective quark flavors at temperature T , dE el Nf = CR πα2s T 2 1 + f (v) ln(Bc ), (61) dx 6 has been used. In an ultrarelativistic gas of massless partons the jetvelocity function is given by 1 1 2 1+v f (v) = 2 v + (v − 1) ln , (62) v 2 1−v while estimates for Bc are taken from Refs. 123, 124, 68 and 69. The different values for Bc obtained in these models are considered as reflecting theoretical uncertainties. The radiative Eloss within the DGLV reactionoperator approach is calculated in Ref. 122 based on Refs. 121 and 125. The left panel of Fig. 15 compares pQCD radiative and collisional Eloss for various quark flavors (masses) at high p T > 5 GeV
Heavy Quarks in the QuarkGluon Plasma
147
in a gluon plasma (GP) with T ' 240 MeV. For light and charm (bottom) quarks the elastic Eloss is comparable to the radiative one up to pT ' 10(20) GeV, and still significant above. The right panel of Fig. 15 reiterates that, within pQCD, collisional Eloss is an essential component in calculating the suppression of lightparton and especially HQ spectra at RHIC. Note that the relative importance of collisional Eloss is expected to increase if nonperturbative effects become relevant (which predominantly figure toward lower pT ), or if the GP is replaced by a QGP. 2.7. D mesons in the hadronic phase To complete the discussion of open charm in QCD matter we briefly address medium modifications of charm hadrons in hadronic matter. Pertinent studies may be divided into calculations for cold nuclear matter as well as for hot meson matter. Early studies of Dmesons in cold nuclear matter focused on possible mass shifts due to scalar and vector mean fields acting on the lightquark content of the meson.126 At normal nuclear matter density %N ≡ % = 0.16 fm−3 , attractive mass shifts of up to − 100 MeV have been reported for D + and D0 mesons (where both mean fields contribute with the same sign) while the mass change of the D − and ¯ 0 turned out to be small due to a cancellation of the mean fields. Similar findings D have been reported in QCD sum rule calculations127 where the (isospinaveraged) Dmeson mass is reduced by about − 50 MeV, mostly as a consequence of the reduction in the lightquark condensate. Rather different results are obtained in microscopic calculations of Dmeson selfenergies (or spectral functions) based on coupled channel T matrices for DN scattering in nuclear matter.128,129 These calculations incorporate hadronic manybody effects, most notably DN excitations into charmbaryon resonances not too far from the DN threshold, e.g., Λc (2593) and Σc (2625), as well as charm exchange into πΛc and πΣc channels. In Ref. 128 separable mesonbaryon interactions have been employed with parameters constrained to dynamically generate the Λc (2593) state. Since the inmedium Dmeson spectral function figures back into the T matrix, one is facing a selfconsistency problem (much like for the heavylight quark T matrix discussed in Sec. 2.3.2). Selfconsistent calculations including nucleon Pauli blocking and dressing of intermediate pion and nucleon propagators result in Dmeson spectral functions with a significant broadening of up to ΓD ' 100 MeV but a rather small shift of the peak position of about − 10 MeV (for %N ≡ % = 0.16 fm−3 ). In Ref. 129, a somewhat stronger coupling of DN to the Λc (2593) results in a stronger collective DN −1 Λc (2593) mode (about 250 MeV below the free Dmeson mass) and a pertinent level repulsion which pushes up the “elementary” Dpeak by ∼ 30 MeV. Also in this calculation the broadening is significant, by about ∼ 50 MeV. The D − was found to be rather little affected, neither in mass nor in width. Investigations in the selfconsistent coupledchannel framework have been extended to a nucleon gas at finite temperature130 with a more complete treatment of DN scattering, cf. left and middle panels of Fig. 16. The thermal motion of nucleons implies that a larger kinematic regime in the center
148
R. Rapp & H. van Hees
80
40
ρ0 2ρ0 ρ0 (Σ=0) 2 ρ0 (Σ=0)
100
40 (MeV)
50
30 20
150
ImU(q=0)[MeV]
D meson
50
50
10 0
80 80
10
40
20 30
100
0
100 50 T[MeV]
150
50
ReΣ/2M
Γ
D*
0
ReΣ/2M
40
40 150
Γ
D
0 40
(MeV)
Re U(q=0)[MeV]

D meson
0
0
100 50 T[MeV]
150
80 0
50
100 150 T (MeV)
200
Fig. 16. (Color online) Real and imaginary parts of Dmeson selfenergies in dense and/or hot hadronic matter. Left and middle panels: selfconsistent coupledchannel calculations for inmedium D + (left) and D − (right) potentials based on DN scattering in a hot nucleon gas as a function ¯ of temperature130 for two different nuclear densities (and for two different inputs for the DN scattering length); the real (upper panels) and imaginary (lower panels) part of the potentials on (q), q)/2ω on with ω on the are defined in terms of the onshell selfenergy via U (q) = ΣD (ωD D D quasiparticle energy; note that the width is given by Γ = −2 Im U . Right panel: width (upper ∗ lines) and mass shift (lower lines) of D (top) and D (2010) (bottom) mesons in a hot pion gas based on resonant scattering via chiral partners131 ; note that the isospin symmetric pion gas ¯ 0 ). implies equal effects on Dmesons (D + , D 0 ) and antiD mesons (D − , D
ofmass (cm) energies in the scattering amplitude is probed (compared to T = 0). For the real parts this leads to a further averaging of the positive and negative parts of the amplitude, while the imaginary parts are negative definite (some loss of interaction strength may occur in channels with resonances close to threshold). More quantitatively, at normal nuclear matter density, the resulting mass shifts are ¯ decreasing to about 10–20 MeV at T = 0 (attractive for D and repulsive for D), half (or less) at T = 150 MeV. On the other hand, the Dmeson width is around ¯ width is small at T = 0 100 MeV at both zero and finite temperature, while the D but increases to about 30 MeV at T = 150 MeV. Medium modifications of Dmesons in a hot pion gas have been studied in Ref. 131. The main idea in this work is to implement the recently discovered scalar D0∗ (2310) and axialvector D10 (2430) states as chiral partners of the pseudoscalar D and vector D∗ (2010) mesons, respectively. Their large widths of 200–400 MeV are primarily attributed to Swave pion decays into D and D ∗ . In a thermal pion gas, D0∗ (2310) and D10 (2430) therefore act as strong resonances in Dπ and D ∗ π scattering, which have been treated in BreitWigner approximation. In addition, Dwave resonances, D1 (2420) and D2∗ (2460), have been accounted for. The resulting collisional widths of D and D ∗ reach up to about ∼ 40–60 MeV for temperatures around T = 175 MeV, while the mass shifts are attractive up to − 20 MeV. It can be expected that (e.g., in a selfconsistent calculation) the inclusion of medium
Heavy Quarks in the QuarkGluon Plasma
149
effects on the resonances (e.g., chiral partners are expected to approach degeneracy towards chiral restoration) will lead to a reduction in the mass shift (not so much for the widths). Above the critical temperature, it is then natural to switch to a quarkbased description, i.e., c + q scattering, much like in the effective resonance model discussed in Sec. 2.3.1. When combining the effects of pion and nucleon scattering on Dmesons in hot hadronic matter, their total width at temperatures around Tc adds up to Γtot D (T = 180 MeV) ' 150 MeV. This is only by about a factor of ∼ 2 smaller than what was found for cquarks at T = 1.1 Tc in the T matrix approach for c–q scattering, cf. right panel of Fig. 11. Since it can be expected that other excited hadrons contribute to Dmeson rescattering (albeit with less strength), Dmeson transport properties may not be much different from those of cquarks in the QGP, at least at temperatures close to Tc . It is therefore of considerable interest to employ inmedium Dmeson T matrices to evaluate heavyflavor transport coefficients in hadronic matter. 3. HeavyQuark Observables in Relativistic HeavyIon Collisions One of the main motivations for the vigorous theoretical studies of HQ diffusion in the QGP is the possibility of utilizing HQ observables in ultrarelativistic heavyion collisions as a quantitative probe of the matter produced in these reactions. If the latter reaches approximate local thermal equilibrium, such applications can be performed by solving the Fokker–Planck equation for a heavy quark diffusing within a collectively expanding background medium with spacetime dependent temperature and flow field (applicable for “sufficiently slow” charm and bottom quarks). This is typically realized with a MonteCarlo simulation using a testparticle ansatz for an equivalent stochastic Langevin equation. In such a formulation, a direct relation between the input in terms of a (temperaturedependent) HQ diffusion coefficient and the modifications of HQ spectra in the evolution can be established. In addition, the Langevin formulation admits an efficient implementation of the dissipationfluctuation relation for relativistic kinematics. Alternatively, the modifications of HQ spectra in URHICs have been evaluated by implementing test particles into numerical transport simulations of the background medium.132–134 This, in principle, accounts for nonequilibrium effects in the medium evolution (which could be particularly relevant for highpt particles in the bulk), but the connection to the diffusion concept becomes less direct (HQ cross sections need to be evaluated in an equilibrium medium to extract “equivalent” diffusion coefficients). However, when analyzing theoretical predictions of HQ spectra we also compare to results of transport models for the bulk evolution. We start our presentation in this Section by briefly outlining how the Fokker– Planck equation is implemented into numerical simulations based on a relativistic Langevin process (Sec. 3.1). This is followed by a discussion of different models
150
R. Rapp & H. van Hees
for the background medium in relativistic heavyion collisions, where we focus on thermal models including hydrodynamics and expanding fireballs (Sec. 3.2). The main task here is to provide realistic benchmarks for the conversion of transport coefficients into modifications of HQ spectra. This furthermore requires the definition of controlled initial conditions for the HQ spectra (Sec. 3.3), usually taken from p–p collisions, possibly augmented by nuclear effects (in particular a “Cronin” p t broadening). Available Langevin simulations combining different inputs are quantitatively compared at the level of the finalstate HQ spectra resulting from the QGP and “mixed” phases in central and semicentral Au–Au collisions at RHIC (Sec. 3.4). Even though HQ spectra are not observable, they provide the cleanest theoretical level of comparison, before further processing through hadronization, electron decay and charm/bottom composition occurs. The latter three steps are necessary to enable comparisons to currently available electron data (Sec. 3.5), and thus arrive at an empirical estimate of the HQ diffusion coefficient characterizing the QCD medium produced at RHIC. In a more speculative step, the extracted HQ transport coefficient may be used to schematically estimate the ratio of shear viscosity to entropy density (Sec. 3.6), which has recently received considerable attention in connection with viscous hydrodynamic simulations at RHIC. 3.1. Relativistic Langevin simulations The Fokker–Planck equation, introduced in Sec. 2.1, is equivalent to an ordinary stochastic differential equation. Neglecting meanfield effects of the medium, the force acting on the heavy particle is divided into a “deterministic” part, describing its average interactions with the light particles in the medium (friction or drag), and a “stochastic” part, taking into account fluctuations around the average on the level of the standard deviation. Thus the relativistic equations of motion for a heavy quark become a coupled set of stochastic differential equations, which for an isotropic medium can be written in the form pj dxj = dt , E (63) √ dpj = −Γpj dt + dtCjk ρk , where E = (m2Q + p2 )1/2 , and Γ and Cjk are functions of (t, x, p) with j, k = 1, 2, 3; they are related to the transport coefficients A and B (discussed in the previous section) below. Γ and Cjk describe the deterministic friction (drag) force and the stochastic force in terms of independent Gaussiannormal distributed random variables ρ = (ρ1 , ρ2 , ρ3 ), 3 ρ2 1 exp − , (64) P (ρ) = 2π 2 In the limit dt → 0, the covariance of the fluctuating force is thus given by D E (fl) (fl) Fj (t)Fk (t0 ) = Cjl Ckl δ(t − t0 ) .
(65)
Heavy Quarks in the QuarkGluon Plasma
151
However, with these specifications the stochastic process is not yet uniquely defined, but depends on the specific choice of the momentum argument of the covariance matrix, Cjk , in Eq. (63),135 i.e., the definition of the stochastic integral. Usual schemes are given by the prepoint Ito, the midpoint StratonovicFisk, and the postpoint Ito (or H¨ anggiKlimontovich136) interpretation of the stochastic integral. We can summarize all these realizations of the stochastic process by specifying the actual momentum argument in the covariance matrix by Cjk → Cjk (t, x, p + ξdp) ,
(66)
where ξ = 0, 1/2, 1 corresponds to the prepoint Ito, the midpoint Stratonovic, and the postpoint Ito realizations, respectively. The equation for the corresponding phasespace distribution function can be found by calculating the average change of an arbitrary phasespace function, g(x, p), with time. According to Eq. (63), with the specification Eq. (66) of the stochastic process, we find ∂g ∂Cjk ∂g pj + −Γpj + ξ Clk hg(x + dx, p + dp) − g(x, p)i = ∂xj E ∂pj ∂pl 2 1 ∂ g + Cjl Ckl dt + O(dt3/2 ) . (67) 2 ∂pj ∂pk Here, the arguments of both, Γ and Cjk , have to be taken at (t, x, p) since the corrections are of the neglected order, O(dt3/2 ). In the derivation of this equation the statistical properties of the random variables ρi , implied by Eq. (64), hρj i = 0 ,
hρj ρk i = δjk ,
(68)
have been used. It follows that the average of an arbitrary phasespace function is by definition given by the phasespace distribution function for the heavy particle (in our context a heavy quark), fQ (t, x, p), e.g., Z Z ∂ d hg(x, p)i = d3 x d3 p g(x, p) fQ (t, x, p) . (69) dt ∂t After integrations by parts, and since Eq. (67) holds for any function g, one finally arrives at the Fokker–Planck equation, ∂fQ pj ∂fQ ∂ ∂Cjk 1 ∂2 + = Γpj − ξClk fQ + (Cjl Ckl fQ ) . (70) ∂t E ∂xj ∂pj ∂pl 2 ∂pj ∂pk The drag term, i.e., the first term on the righthand side of this equation, depends on the definition of the stochastic integral in terms of the parameter ξ. Comparison with Eq. (10) shows that, independent of the choice of ξ, the covariance matrix is related to the diffusion matrix by p p k ⊥ + 2B1 Pjk , Cjk = 2B0 Pjk (71)
152
R. Rapp & H. van Hees
while the friction force is given by Γpj = Apj − ξClk
∂Cjk . ∂pl
(72)
Numerical investigations have shown that the drag and diffusion coefficients inferred from microscopic models according to Eqs. (15) in general do not warrant a good agreement of the longtime limit of the solution to the Fokker–Planck evolution with the relativistic equilibrium J¨ uttnerBoltzmann distribution (where the temperature is given by the background medium). The problem is with the longitudinal diffusion coefficient, B1 , which induces an overestimate of the corresponding fluctuating forces. Thus, one typically adjusts the drag coefficient by choosing B 1 in Eq. (70) to satisfy the asymptotic equilibration condition.26,137,138 To find the dissipationfluctuation relation, imposed by the equilibration condition, we first study the heavy quark’s motion in a heat bath in thermal equilibrium in its rest frame. Then the momentum distribution of the heavy quarks should become a J¨ uttnerBoltzmann distribution, eq fQ (p) ∝ exp (−E/T )
(73)
with the temperature, T , imposed by the heat bath. For a Langevin process with B0 = B1 = D, i.e., p (74) Cjk = 2D(E)δjk ,
where the diffusion coefficient has been written as a function of the heavy quark’s energy, E, the equilibration condition for a given parameter ξ in Eq. (66) is obtained by using Eqs. (73) and Eq. (74) in Eq. (70): A(E)ET − D(E) + T (1 − ξ)D 0 (E) = 0 .
(75)
Since the drag and diffusion coefficients are usually given numerically, the most convenient update rule for the Langevin process is achieved by setting ξ = 1, i.e., using the postpoint Ito (H¨ anggiKlimontovich) rule for the stochastic integral in Eq. (63) and imposing the simple relativistic dissipationfluctuation relation, D = AET .
(76)
This guarantees the proper approach of the heavy quark’s phasespace distribution to the appropriate equilibrium distribution with the temperature imposed by the heat bath. For the more general form of the covariance matrix, Eq. (71), the postpoint Ito value, ξ = 1, has been chosen in Ref. 28, and the longitudinal diffusion coefficient is set to B1 = AET ,
(77)
Heavy Quarks in the QuarkGluon Plasma
153
while the drag coefficient A as well as the transverse diffusion coefficient, B0 , are used as given by Eq. (15) for the various microscopic models for HQ scattering in the QGP. Comparing to Eq. (75), one finds that this is equivalent to the strategy followed in Ref. 26 of using the prepointIto rule, ξ = 0, but to adjust the drag coefficient according to the dissipationfluctuation relation Eq. (75). 3.2. Background medium in heavyion collisions For HQ transport coefficients computed in an equilibrium QGP, the natural and consistent framework to describe the evolving medium in heavyion collisions are hydrodynamic simulations, formulated in the same (thermodynamic) variables. This choice is further rendered attractive by the success of ideal hydrodynamics in describing bulk observables at RHIC, in particular pT spectra and elliptic flow of the most abundant species of hadrons.13–16 The agreement with meson and baryon spectra typically extends to pT ' 2–3 GeV, respectively. At the parton level, this converts into a momentum of pt ' 1 GeV, which approximately coincides with the “levelingoff” of the experimentally observed v2 (pt ) (at higher momenta hydrodynamics overestimates the elliptic flow). On the one hand, this appears as a rather small momentum in view of the ambition of describing HQ spectra out to, say, pt ' 5 GeV. However, one should realize that (a) more than 90% of the bulk matter is comprised of light partons with momenta below pt ' 1 GeV, and (b) the velocity of a pt = 5 GeV charm quark (with mc = 1.5 GeV) is very similar to a pt = 1 GeV light quark (with mq = 0.3 GeV). This suggests that most of the interactions of a pt = 5 GeV charm quark actually occur with soft light partons (which are well described by a hydrodynamic bulk). This has been verified by explicit calculations139 and is, after all, a prerequisite for the applicability of the Fokker–Planck approach (i.e., small momentum transfer per collision). On the other hand, one may be concerned that the overestimate of the experimentally observed elliptic flow at intermediate and high pt within hydrodynamics may exaggerate the HQ elliptic flow in Langevin simulations. This is, however, not necessarily the case, since the transfer of v2 from the bulk to the heavy quark critically depends on the lightparton phase space density (the drag coefficient is proportional to it); since the hydrodynamic spectra fall significantly below the experimental ones at higher pT , the phase space density of the hydrodynamic component is relatively small. It is therefore not clear whether the (small) fraction of thermalized particles at high pt implies an overestimate of the total v2 ; this may be judged more quantitatively by comparing to transport calculations. In parallel to hydrodynamic descriptions of the bulk medium, expanding fireball models have been employed. These are simplified (and thus convenient) parameterizations of a full hydrodynamic calculation in terms of an expanding volume and spatial flowvelocity field, but otherwise based on similar principles and variables. E.g., entropy conservation is used to convert a given volume into a temperature via an underlying equation of state (EoS). The reliability of a fireball model
154
R. Rapp & H. van Hees
Table 1. Survey of parameters figuring into hydrodynamic and fireball evolutions employed in the Langevin simulations of HQ spectra for semicentral Au–Au collisions at RHIC, corresponding to MT,26,140 HGR,28,45,142 AHH115,141 and GA.13,67
MT
HGR
AHH
GA
τ0 [fm/c ]
1.0
0.33
0.6
0.6
T0 [MeV]
265 (max)
340 (avg)
250 (avg)
330 (max) 260 (avg)
Tc [MeV]
165
180
170
165
initial spatial
wounded nucleon
isotropic
lin. comb. of Ncoll and Npart
lin. comb. of Ncoll and Npart
b [fm]
6.5
7
5.5
7
bulkv2
5
5.5%
3%
4.75%
τFB [fm/c ]
∼9
∼5
∼9
∼7
QGPEoS
massless (Nf = 3)
massless (Nf = 2.5)
massless (Nf = 3)
massless (Nf = 3)
HQ Int.
pQCD HTL
pQCD+reso pQCD+Tmat
AdS/CFT
pQCD run. αs
mc,b [GeV]
1.4
1.5, 4.5
1.5, 4.8
1.5, 5.1
crucially hinges on a realistic choice of the parameters, in connection with a proper description of the final state in terms of particle production and collective flow. In principle, a fireball model offers some additional flexibility in varying the evolution, which may provide useful checks of the sensitivity to specific aspects of the expansion. Several key parameters of thermal medium evolution models employed in HQ Langevin simulations are compiled in Table 1. The starting point of both hydro and fireball models are the initial conditions of the thermal medium, characterized by a formation time when the medium is first assumed to be (locally) equilibrated. At RHIC, typical formation times are estimated to be in the range of τ0 ' 0.3–1 fm/c. With a total entropy fixed to reproduce the measured rapidity density of hadrons at a given centrality, e.g., at impact parameter b ' 7 fm/c, these formation times translate into average initial temperatures of T0 ' 250–350 MeV. If the entropy density scales as s ∝ T 3 , one can roughly compare the initial conditions in different approaches using S = s0 V0 and V0 ∝ τ0 . E.g., an initial T0max = 265 MeV based on τ0 = 1 fm/c increases by a factor of 31/3 upon decreasing τ0 = 0.33 fm/c, resulting in T0max ' 382 MeV; if the number of light flavors in the EoS is reduced, T0 increases as well; e.g., T0avg = 260 MeV based on τ0 = 0.6 fm/c and Nf = 3 (as in Ref. 67) increases to T0avg = 260 MeV (0.6/0.33)1/3 (47.5/42.25)1/3 ' 330 MeV for τ0 = 0.33 fm/c and Nf = 2.5 (reasonably consistent with Ref. 28, cf. Table 1). The QGPdominated evolution lasts for about 2–4 fm/c, followed by a mixed phase
Heavy Quarks in the QuarkGluon Plasma
155
of similar duration at a critical temperature Tc ' 165–180 MeV. The effects of a continuous (crossover) transition, as well as of the hadronic phase, have received little attention thus far, but are not expected to leave a large imprint on HQ observables. After all, the crossover transition found in lQCD exhibits a marked change in energy density over a rather narrow temperature interval. A more important aspect is the consistency between the EoS used to extract the temperature of the bulk evolution and the corresponding degrees of freedom figuring into the calculation of the HQ transport coefficients. In hydrodynamical backgrounds used thus far26,67,115 the evolution is described with a 2+1dimensional boostinvariant simulation with an ideal masslessgas EoS. The initial state is typically determined by distributing the entropy in the transverse plane according to the wounded nucleon density. Unfortunately, the impact parameters in current Langevin simulations vary somewhat, which is particularly critical for the magnitude of the subsequently developed elliptic flow. The value of the critical temperature has some influence on the QGP lifetime (lower temperatures leading to larger QGP duration), as does the hadrongas EoS (more hadronic states imply a larger entropy density at T c and thus a reduced duration of the mixed phase). The termination point of the evolution (beginning, middle or end of mixed phase) is rather significant, especially for the HQ v2 which needs about 5 fm/c to build up most of its magnitude. In Refs. 28 and 45 the medium is parameterized, guided by the detailed hydrodynamic calculations of Ref. 142, as a homogeneous thermal elliptic “fire cylinder” of volume V (t). The QGP temperature is determined via the QGP entropy density, s, under the assumption of isentropic expansion (total S = const), s=
4π 2 3 S = T (16 + 10.5Nf ) . V (t) 90
(78)
The thus inferred temperature is used in Eq. (15) to compute the friction coefficients, A, and transverse diffusion coefficient, B0 , with the longitudinal diffusion coefficient fixed by the dissipationfluctuation relation, Eq. (77). In the mixed phase at Tc = 180 MeV the QGP drag and diffusion coefficients are scaled by a factor ∝ %2/3 to account for the reduction in parton densities (rather than using hadronic calculations). Special care has to be taken in the parameterization of the elliptic flow in noncentral Au–Au collisions: the contours of constant flow velocity are taken as confocal ellipses in the transverse plane with the pertinent transverse flow set consistently in perpendicular direction. The time evolution of the surface velocity of the semiaxes of the elliptic fire cylinder parameterizes the corresponding results of the hydrodynamic calculations in Ref. 142, in particular the timedependence of the ellipticflow parameter, v2 , for the light quarks. The parameters are adjusted such (s) that the average surface velocity reaches v⊥ = 0.5c and the anisotropy parameter v2 = 5.5% at the end of the mixed phase. Finally, the velocity field is specified by scaling the boundary velocity linearly with distance from the center of the fireball, again in accordance with the hydrodynamic calculation.142
156
R. Rapp & H. van Hees
2
3
10
1/(2πpT) dN/dpT [a u ]
STAR D * STAR prelim D (×2 5) cquark (mod PYTHIA) cquark (CompHEP)
4
10
2
1/(2 πpT) d N/dpT dy [a u ]
10
0
10
5
10
6
10
d+Au √sNN=200 GeV
10
10 10
STAR (prel, pp) STAR (prel, d+Au/7 5) STAR (pp) D→e B→e sum
±
e
5
6
10 10 10 10
7 8 9
10
σbb/σcc= 4.9 x10
3
11
7
10 0
3
4
1
2
3
4 5 pT[GeV]
6
7
8
10 0
1
2
3
4 pT [GeV]
5
6
7
8
Fig. 17. (Color online) Fits of D and D ∗ meson pT spectra in 200AGeV dAu collisions at RHIC with a modified PYTHIA simulation (left panel) and the corresponding nonphotonic singleelectron pt spectra in p–p and dAu collisions.144 The missing yield of highpT electrons is fitted with the analogous Bmeson decay spectra, thus fixing the bottomcharm ratio at σ b¯b /σc¯c ' 4.9 · 10−3 .
3.3. Initial conditions and hadronization The Langevin simulations of HQ diffusion in the QGP require initial conditions for charm and bottomquark phasespace distributions. For the spatial part of the initial distribution in the transverse plane all calculations adopt binarycollision scaling following from a Glauber model, reflecting a hard process for the primordial production mechanism. Furthermore, all calculations thus far focus on a limited rapidity window around midrapidity, where the longitudinal distribution is assumed to be uniform in spacetime rapidity. As for the initial HQ pt spectra, Ref. 26 employs a fit to a leadingorder partonmodel calculation from the CompHEP package, 143 dN 1 ∝ δ(η − y) 2 , 2 dydηd pt (pt + Λ2 )α
(79)
with α = 3.5 and Λ = 1.849 GeV. In Refs. 28 and 45, PYTHIA results for cquark spectra have been tuned to reproduce available Dmeson spectra in dAu collisions at RHIC (assuming δfunction fragmentation, cf. left panel of Fig. 17). The pertinent semileptonic singleelectron decay spectra approximately account for p–p and dAu spectra up to pT = 4 GeV; the missing part at higher pT is then supplemented by Bmeson contributions. This procedure results in a crossing of the D and Bmeson decay electrons at pT ' 5 GeV and a crosssection ratio of σb¯b /σc¯c ' 4.9 · 10−3 (see right panel of Fig. 17), which is within the range of pQCD predictions.145 With initial conditions and bulk medium evolution in place, one can evolve HQ phasespace distributions through the QGP (and mixed phase) of a heavyion collision. The final HQ spectra, however, require further processing before comparisons to observables can be made. First, one has to address the hadronization of the HQ spectra into charm and bottom hadrons (D, D ∗ , Λc etc.). Two basic mechanisms
Heavy Quarks in the QuarkGluon Plasma
157
have been widely considered in hadronic collisions, i.e., fragmentation of an individual quark and recombination with an extra quark from the environment. In general, the former is mostly applicable for highenergy partons while the latter requires a sufficient overlap of the mesonic wave function with the phasespace density of surrounding quarks and is therefore more relevant toward lower momentum. The fragmentation of a quark is implemented by applying the factorization theorem of QCD.21 At large transverse momenta, the production process of a parton occurs on a short time scale, τprod ' 1/pt , while hadronization occurs at the considerably larger time scale τhad ' 1/ΛQCD . Thus the production cross section for a hadron can be factorized into an elementary partonproduction cross section (hard process) and a phenomenological universal transition probability distribution, Dh/i (z), for a parton i of momentum pi to convert into a hadron with momentum fraction z = ph /pi ≤ 1. For light quarks and gluons the fragmentation functions, Dh/i , are rather broad distributions around z ' 0.5, but for heavy quarks they become rather sharply peaked toward z = 1 and are sometimes even approximated by a δfunction, D(z) = δ(1 − z). The mechansim of recombination of a produced quark with other quarks or antiquarks in its environment (e.g., the valence quarks of the colliding hadrons) has first been introduced in the late 1970’s to explain flavor asymmetries in π and K meson production in hadronic collisions at forward rapidities.146 In particular, the recombination idea has been rather successful in describing flavor asymmetries in the charm sector,147,148 even close to midrapidity. In the context of heavyion collisions, quark coalescence models, applied at the hadronization transition, provide a simple and intuitive explanation for the observed constituentquark number scaling (CQNS) of the elliptic flow of light hadrons149,150 and the (unex0 ¯ pectedly) large baryontomeson ratios (e.g., p/π ' 1 or (Λ + Λ)/(4K S ) ' 1.3 in central 200 AGeV Au–Au collisions at RHIC) at intermediate transverse momenta (2 GeV . pT . 5 GeV).151–154 CQNS refers to a scaling property of the hadronic elliptic flow, v2,h (pT ), in terms of a universal function v2,q (pT /n) = v2,h (pT )/n, where n denotes the number of constituent quarks in a given hadron, h. CQNS naturally emerges from the recombination of approximately comoving quarks and antiquarks in a collectively flowing medium. Thus, within this picture, v2,q (pt ) is interpreted as a universal elliptic flow of the quarks with transverse momentum pt at the moment of hadronization (typically assumed to be the quarkhadron transition at Tc ). It can be expected that the phenomenologically very successful coalescence concept also applies in the HQ sector of heavyion collisions.155,156 Note that, unlike quark fragmentation, quark recombination adds momentum and elliptic flow to the produced hadron (through the quark picked up from the environment). At this point it might be instructive to reiterate a conceptual connection between the quark coalescence model and the idea of resonance correlations in the QGP. The latter were found to be an efficient mechanism for arriving at a small HQ diffusion constant, both within the effective resonance model (Sec. 2.3.1) and within the T matrix approach (Sec. 2.3.2). Especially in the T matrix approach, the
158
R. Rapp & H. van Hees
resonance correlations were found to strengthen toward the expected hadronization transition, and thus provide a natural emergence of heavylight quark coalescence at Tc . These ideas have recently been implemented in a resonancebased description of the coalescence process in kinetic theory based on a Boltzmann equation. 157 This approach improves instantaneous coalescence formulations in that it respects energy conservation and establishes a welldefined equilibrium limit in the coalescence process (i.e., the thermal distribution for the formed meson). Subsequently, resonancerecombination has been combined with “realistic” quark phasespace distributions, as generated in relativistic Langevin simulations.158 In particular, it was found that CQNS could be recovered under the inclusion of spacemomentum correlations in the quark phasespace distributions. In Refs. 28 and 45 HQ spectra at RHIC have been hadronized in a combined coalescence plus fragmentation scheme. For the hadronization of, e.g., charm quarks into D mesons one obtains the Dmeson spectra as coal tot dND dNcfrag dND = + . dy d2 pT dy d2 pT dy d2 pT
(80)
For the first term on the righthand side, the quarkcoalescence model of Ref. 155 has been employed, where the pT spectrum of a D meson follows from a convolution of light antiquark and charmquark phasespace distributions, fq¯,c , as coal dND = gD dyd2 pT
Z
p · dσ (2π)3
Z
d3 qfD (q, x)fq¯(pq¯, rq¯)fc (pc , rc ) .
(81)
Here, p = pq¯ + pc is the Dmeson momentum and gD a combinatorial factor accounting for colorneutrality and spin averaging. The Dmeson Wigner function, fD (q, x), is assumed as a double Gaussian in relative momentum pc − pq¯ and size, r c − rq¯, and dσ is the hypersurface element 4vector of the hadronization volume. The charmquark distribution corresponds to the Langevin output at the end of the mixed phase of the fireball model, while the lightquark distributions are taken from previous applications of the coalescence model to lighthadron observables at RHIC.152 This represents a parameterfree conversion of HQ distributions into heavymeson spectra (note that the final state of the expanding fireball model28 has been matched to the parameterization of collective velocity and elliptic flow for the lightquark distributions in the coalescence model152 ). The coalescence mechanism does not exhaust all heavy quarks in the hadronization process, especially toward higher pt (where the lightquark phasespace density becomes small). Therefore, the remaining heavy quarks are hadronized using fragmentation, which for simplicity is treated in δfunction approximation (as has been done in connection with the initial conditions). The formation of baryons containing heavy quarks (e.g., Λc ) has been neglected since it has been found to give only small contributions, i.e., Λc /D 1. Quantitative refinements should, however, include these processes, see, e.g., Refs. 159 and 160.
Heavy Quarks in the QuarkGluon Plasma
159
Finally, the comparison to electron spectra requires to compute semileptonic decays of heavyflavor hadrons. Thus far, these have been approximated in threebody kinematics, e.g., D → eνK. An important finding in this context is that the resulting electron v2 (pT ) traces the one of the parent meson rather accurately,155,161 implying that electron spectra essentially carry the full information on the heavymeson v2 . In the pT spectra, the decay electrons appear at roughly half of the momentum of the parent meson. It has also been pointed out159,160 that Λc baryons have a significantly smaller branching fraction into electrons (about 4–5%) than D mesons (7% and 17% for neutral and charged D’s, respectively). Thus, in case of a large Λc /D enhancement, a net electron “loss” could mimic a stronger suppression than actually present at the HQ level. In fact, even variations in the neutral to charged chemistry from p–p to A–A collisions148 could be quantitatively relevant. 3.4. Model comparisons of heavyquark spectra at RHIC We are now in position to conduct quantitative comparisons of diffusion calculations using transport simulations for HQ spectra in 200 AGeV Au–Au collisions at RHIC. We focus on Langevin simulations but also allude to Boltzmann transport models. The modifications of the initial spectra are routinely quantified in terms of the nuclear modification factor, RAA , and ellipticflow parameter, v2 , defined by RAA (pt ; b) =
v2 (pt ; b) =
AA dNQ (b)/dpt , pp Ncoll (b) dNQ /dpt
R
dN AA (b)
dφ dptQdydφ cos(2φ) , R dN AA (b) dφ dptQdydφ
(82)
AA respectively; dNQ (b)/dpt denotes the HQ pt spectrum in an A–A collision at impp pact parameter, b, which is scaled by the spectrum dNQ /dpt from p–p collisions times the number of binary nucleonnucleon collisions, Ncoll (b) (to account for the same number of heavy quarks). Thus, any deviation of RAA from one indicates nuclear effects (from the QGP but possibly also in the nuclear initial conditions or from the preequilibrium stages). The ellipticflow parameter, v2 (pt ), is the second Fourier coefficient in the expansion of the (final) momentum distributions in the azimuthal angle, φ, relative to the reaction plane (x–z plane) of the nuclear collision. At midrapidity, where the “directed” flow (v1 ) is expected to vanish, the v2 coefficient is the leading source of azimuthal asymmetries. A nonzero v2 is only expected to occur in noncentral A–A collisions due to an “almond”shaped nuclear overlap zone (with a long (short) axis in y (x) direction). Typical sources for a nonzero elliptic flow are a pathlength difference for absorption of particles traversing the reaction zone or an asymmetry in the collective (hydrodynamic) flow due to stronger pressure gradients across the short axis. Both effects convert the initial spatial anisotropy, v2 , in a positive momentum anisotropy in the particle pt spectra.
R. Rapp & H. van Hees 1.4 1.2 1
D (2πT) = 1.5 D (2πT) = 3 D (2πT) = 6 D (2πT) = 12 D (2πT) = 24
(a)
v2 (pT)
RAA
160
0.22 0.2 0.18
LO QCD
(b)
0.16 0.14
0.8
0.12 0.1
0.6
0.08
0.4
0.06 0.04
0.2
0.02
0 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 pt (GeV)
0 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 pt (GeV)
Fig. 18. (Color online) Nuclear modification factor (left panel) and elliptic flow (right panel) of charm quarks as a function of transverse momentum in semicentral (b = 6.5 fm) Au–Au collisions using a hydrodynamic evolution of the bulk medium at RHIC.26 The calculations are performed for HTLimproved LOpQCD scattering with variable strong coupling and fixed Debyescreening mass in tchannel gluonexchange scatteringi (µD = 1.5 T ). The relation of the spatial diffusion coefficient, Ds (denoted D in the figure legend), to the strong coupling constant, αs , is given by the approximate relation 2πT Ds ≈ 6(0.5/αs )2 .
While the former mechanism is usually associated with highpt particles (typical leading to a v2 < 5%), the latter is driven by collective expansion due to thermal pressure mostly applicable to lowpt particles (with significantly larger v2 values, in excess of 5%). Since in the Langevin simulations heavy quarks are assumed to be exclusively produced in primordial N –N collisions (i.e., their number is conserved subsequently), the HQ RAA can be simply calculated as the ratio of the HQ pt distribution function at the moment of hadronization to the initial distribution (taken from p–p collisions), fQ (thad , pt ) , fQ (t0 , pt ) R dφfQ (thad , pt , φ) cos(2φ) , v2 (pt ) = fQ (thad , pt ) RAA =
(83)
while the v2 is computed using its definition given above. The next five figures (1822) encompass calculations of RAA and v2 in semicentral Au–Au collisions at RHIC for the following approaches: (i) Figure 18 [MT]26 displays Langevin simulations for c quarks (with the prepoint Ito realization of the stochastic integral) using a hydrodynamic evolution for b = 6.5 fm; the HQ drag and diffusion coefficients are based on LO hardthermal loop scattering matrix elements with variable αs but fixed Debye screening mass.
Heavy Quarks in the QuarkGluon Plasma
15 v2 [%]
RAA
20
c, reso (Γ=0 40 75 GeV) c, pQCD, αs=0 4 b, reso (Γ=0 40 75 GeV)
1.5
161
1
10
0.5
c, reso (Γ=0.40.75 GeV) c, pQCD, αs=0.4 b, reso (Γ=0.40.75 GeV) AuAu √s=200 GeV (b=7 fm)
5 AuAu √s=200 GeV (b=7 fm)
0 0
1
2 3 pt [GeV]
4
5
0 0
1
2 3 pt [GeV]
4
5
Fig. 19. (Color online) The HQ RAA (left panel) and v2 (right panel) for semicentral (b = 7 fm) Au–Au collisions at RHIC within the effective resonance + pQCD model compared to results from LO √ pQCD elastic scattering only with αs = 0.4 and corresponding Debyescreening mass µD = 4παs T .
(ii) Figure 19 [HGR]28 displays Langevin simulations for c and b quarks (with the postpoint Ito (H¨ anggiKlimontovich) realization) using a thermal fireball expansion for b = 7 fm; the HQ drag and diffusion coefficients are based on the effective resonance+pQCD model25 for variable resonance width (coupling strength) and αs = 0.4 in the pQCD part. (iii) Figure 20 [AHH]115 displays Langevin simulations for c quarks (with the prepoint Ito realization) using a hydrodynamic expansion for b = 5.5 fm; the HQ drag and diffusion coefficients are based on the strongcoupling limit with AdS/CFT correspondence with a variable coupling strength estimated from matching to QCD.108 (iv) Figure 21 [HMGR]45 displays Langevin simulations as under (ii) but with HQ transport coefficients based on the T matrix+pQCD approach for two lQCDbased input potentials. (v) Figure 22 [Mol]134 displays Boltzmann transport simulations using a covariant transport model for b = 8 fm; the HQ interactions are modeled by schematic LO pQCD cross sections, including upscaling by “K factors”. Before going into details, let us try to extract generic features of the calculations. In all cases there is a definite correlation between a reduction in RAA (pt > 3 GeV) and an increase in v2 (pt ), i.e., both features are coupled to an increase in interaction strength (decrease in the spatial HQ diffusion coefficient). Furthermore, the v2 (pt ) shows a typical, essentially linear, increase reminiscent of a quasithermal regime followed by a saturation characteristic for the transition to a kinetic regime. In all Langevin calculations the saturation for charm quarks occurs at about p t = 2–3 GeV. For the largest interaction strength considered (Ds ' 1/(2πT )), the left panels of Figs. 18 and 20 even suggest a turnover of v2 (at this point one should recall that all calculations displayed in this section utilize elastic scattering only which is expected to receive appreciable corrections at high pt due to radiative
162
R. Rapp & H. van Hees
1.4
b=5.5fm, f =0.5 0
γ =0.3 (charm) γ =1.0 γ =3.0 γ =30.0
1.2
0.8
b=5.5fm, f =0.5 0
γ =0.3 (charm) γ =1.0 γ =3.0 γ =30.0
vc2
RcAA
1
0.14 0.12 0.1 0.08 0.06 0.04 0.02 0 0.02 0.04 0
0.6 0.4 0.2 0 0
1
2
3
4 5 6 7 pout [GeV]
8
9 10
1
2
3
t
4 5 6 7 pout [GeV]
8
9 10
t
Fig. 20. (Color online) RAA (left) and v2 (right) of charm quarks resulting from hydrodynamic √ simulations of b = 5.5 fm Au–Au ( sN N = 200 GeV) collisions using AdS/CFTmotivated charmquark diffusion constants with variable strength parameter, γ,115 which corresponds to the constant C in Eq. (84).
Table 2. Overview of model approaches (1st column) and input parameters (2nd column: spatial charmquark diffusion coefficient, 3rd column: nuclear impact parameter) for Langevin simulations of charmquark spectra in Au–Au collisions at RHIC; selected values for the resulting elliptic flow (v2max ' v2 (pt = 5 GeV)) and nuclear modification factor are quoted in columns 4 and 5. The last two rows represent charmquark transport calculations in a transport model for the bulk.
Model [Ref.] hydro + LOpQCD26 hydro + LOpQCD
26
fireball + LOpQCD28
Ds (2πT )
b [fm]
v2max
RAA (pt = 5 GeV)
24
6.5
1.5%
0.7
6
6.5
5%
0.25
∼ 30
7
2%
0.65
28
∼6
7
6%
0.3
115
21
7.1
1.5–2%
∼ 0.7
hydro + “AdS/CFT” (84)115
2π
7.1
4%
∼ 0.3
∼ 30
8
∼ 2%
∼ 0.65
∼7
8
10%
∼ 0.4
fireball + reso+LOpQCD hydro + “AdS/CFT” (84)
transport + LO pQCD
134
transport + LO pQCD134
processes). On the other hand, for pt = 2–3 GeV the nuclear modification factor is still significantly falling, leveling off only at larger pt ' 5–6 GeV. As expected, bottom quarks exhibit much reduced effects for comparable diffusion constants due to their factor ∼ 3 larger mass (see Fig. 19 and lower panels in Fig. 21). Next, we attempt more quantitative comparisons. Some representative numbers for the resulting RAA and v2 values are compiled in Table 2.114 First we compare the LOpQCD calculations for HQ diffusion in the hydrodynamic and fireball backgrounds corresponding to Figs. 18 and 19, respectively; for a comparable
Heavy Quarks in the QuarkGluon Plasma 16
15
Tmatrix [Wo] Tmatrix [SZ] Reso, Γ=0 40 75 GeV pQCD, αs=0 4
14 12
v2 (%)
RAA
Tmatrix [Wo] Tmatrix [SZ] Reso, Γ=0 40 75 GeV pQCD, αs=0 4 AuAu √s=200 GeV (b=7 fm) charm quarks
10
1 08 06
163
5
04 AuAu √s=200 GeV (b=7 fm) charm quarks
02 0 0
1
18
4
5
0
14
5
2
3
4
5
Tmatrix [Wo] Tmatrix [SZ] Reso, Γ=0 40 75 GeV pQCD, αs=0 4 AuAu √s=200 GeV (b=7 fm) bottom quarks
4 3 v2 (%)
12
1
pt (GeV)
Tmatrix [Wo] Tmatrix [SZ] Reso, Γ=0 40 75 GeV pQCD, αs=0 4
16
RAA
0
2 3 pt (GeV)
1 08 06
2 1
04
0
AuAu √s=200 GeV (b=7 fm) bottom quarks
02 0
1 0
1
2
3
4
5
0
pt (GeV)
1
2
3
4
5
pt (GeV)
Fig. 21. (Color online) Charm (top row) and bottomquark (bottom row) spectra in Au– √ Au( sN N = 200 GeV) collisions at RHIC using Langevin simulations for HQ diffusion in an expanding fireball model. The results using HQ diffusion based on LOpQCD and resonance+pQCD approaches (see also Fig. 19) are compared to the T matrix+pQCD calculations 45 (the pertinent uncertainty band reflects different input potentials based on parameterizations of lQCD HQ free energies as given in Refs. 90 and 44. The left (right) panels show the R AA (v2 ) for central (semicentral) collisions.
spatial diffusion coefficient, Ds ' 24–30/(2πT ), both calculations show a maximal v2 of about 2% and a RAA (pt = 5 GeV) ' 0.7 (recall the smaller b = 6.5 fm in [MT] vs. 7 fm in [HGR] which may lead to somewhat smaller v2 , and the smaller T0 = 265 MeV [MT] vs. 340 MeV in [HGR] which entails somewhat less suppression). For the [AHH] hydro calculation with an AdS/CFTmotivated ansatz for the HQ friction constant, γ=C
T2 , mQ
(84)
a diffusion constant of Ds = 21/(2πT ) leads to similar results (note that Table 2 contains results for b = 7.1 fm162 while Fig. 20 is calculated for b = 5.5 fm). Let us now turn to stronger coupling, still focusing on the three Langevin simulations in Figs. 18, 19 and 20 which all utilize friction coefficients with a similar temperature dependence, essentially γ ∝ T 2 (recall right panel of Fig. 13), corresponding to an approximately constant spatial diffusion constant times temperature, Ds (2πT ) ' const (recall left panel of Fig. 14). For Ds = 6/(2πT ), all calculations are again in semi/quantitative agreement, with a maximum v2 of 4–6% and
164
R. Rapp & H. van Hees
RAA (pt = 5 GeV) ' 0.25–0.3. The 4% value for Ref. 115 [AHH] will increase somewhat if the hydro evolution is run to the end of the mixed phase rather than terminated in the middle of the mixed phase (this is supported by the discussion in Sec. 3.5.2). We also note that in the fireball model of Ref. 28 [HGR] the inclusive (pt integrated) v2 at the end of the mixed phase was adjusted to the experimentally observed lightparticle v2 ' 5.5–6% at an impact parameter of b = 7 fm, i.e., it presumably includes an extra 20% of bulkv2 compared to the hydrodynamic calculations.f Such an amount is typically built up in the subsequent hadronic phase of hydrodynamic evolutions and thus not present in pertinent HQ simulations within a QGP (+ mixed) phase. We now make some comments specific to individual calculations. The Langevin calculations using HQ T matrix interactions (supplemented with pQCD scattering off gluons) shown in Fig. 21 are rather close to the effective resonance model, even though they do not involve tunable parameters. However, they are still beset with substantial uncertainty, as indicated by the use of two different input potentials (in addition, the use of the free energy, F1 , instead of the internal energy, U1 , as potential significantly reduces the effects). One also notices that the v2 at low pt is very similar to the resonance model while the suppression at high pt is somewhat less pronounced. This is so since the T matrix transport coefficients (a) fall off stronger with 3momentum (the resonant correlations are close to the Q–q threshold), and (b) decrease with increasing temperature (resonance melting). The latter combines with the facts that the suppression is primarily built up in the very early stages (where the T matrix is less strong) while the bulk v2 takes a few fm/c to build up (at which point the T matrix has become stronger). Furthermore, the T matrix calculations lead to stronger medium effects on b quarks than the effective resonance model; this reflects the stronger binding due to the mass effect in the T matrix calculation. A principal limitation of the Langevin approach is the treatment of fluctuations which are by definition implemented in Gaussian approximation. The latter arises due to enforcing the dissipationfluctuation relation (mandatory to ensure the HQ distributions to approach equilibrium) which tends to underestimate the momentum fluctuations especially at high momentum, compared to a full transport calculation. This leads to an overestimate of the quenching effect at high pt even for the same average energy loss. One may assess these limitations more quantitatively by comparing to Boltzmann simulations including partonic phases,133,134 an example of which is displayed in Fig. 22 for charm quarks in b = 8 fm Au–Au collisions at RHIC. The baseline LOpQCD calculations indicated by the crosses in Fig. 22, labeled by “1.33 mb”, may be compared to the fireballLangevin simulations represented by the blue lines in Fig. 19. In both calculations the underlying elastic partonHQ cross sections correspond to a strong coupling constant of αs ' 0.4. f This
adjustment ensures compatibility of the fireball freezeout with the coalescence model. 152
Heavy Quarks in the QuarkGluon Plasma
1.4
0.4
1.2
1.33 mb 5.33 mb 20 mb
charm
0.3
0.8 v2
RAA
1
1.33mb 5.33 mb 20 mb
165
charm
0.2
0.6 0.1
0.4 0.2
0
0 0
2
4
6
8
10
12
14
0
1
pT [GeV]
2
3
4
5
pT [GeV]
Fig. 22. (Color online) RAA (left) and v2 (right) of charm quarks resulting from covariant transport √ simulations of b = 8 fm Au–Au ( sN N = 200 GeV) collisions.134
2 1.5
20 15
v2 [%]
RAA
25
c, pQCD+rad+reso (Γ=0 40 75 GeV) c, pQCD+rad b, pQCD+rad+reso (Γ=0 40 75 GeV)
1
c, pQCD+rad+reso (Γ=0.40.75 GeV) c, pQCD+rad b pQCD+reso+rad (Γ=0.4−0.75 GeV) AuAu √s=200 GeV (b=7 fm)
10
0.5
5 AuAu √s=200 GeV (b=7 fm)
0 0
1
2 3 pt [GeV]
4
5
0 0
1
2 3 pt [GeV]
4
5
Fig. 23. (Color online) The nuclear modification factor, RAA (left panel), and elliptic flow, v2 (right panel), for heavy quarks from collisional28 and radiative Eloss,121 cf. Ref. 163.
The quenching and elliptic flow come out quite similar in both calculations at least up to pt ' 5 GeV, especially when accounting for the slightly different centrality. E.g., in the Boltzmann treatment, the RAA for pt = 5 GeV charm quarks is about 0.6–0.7 with a v2 of a few percent. For a fourfold increase of the cross section (which would roughly correspond to a reduction of Ds (2πT ) from ∼ 30 to ∼ 7), one finds RAA (pt = 5 GeV) ' 0.4 and a maximum v2 of close to 10%. While the latter value is somewhat larger than the Langevin predictions, the agreement is not too bad. Finally, Fig. 23 shows results from an exploratory calculation in the Langevin approach where HQ drag and diffusion coefficients from elastic scattering in the effective resonance model (cf. Sec. 2.3.1) are combined with induced gluon radiation in the DGLV Eloss formalism (cf. Sec. 2.6).163 One of the uncertainties in
166
R. Rapp & H. van Hees
this calculation is the extrapolation of the radiative Eloss into the lowmomentum regime, where it still contributes rather substantially; e.g., the elliptic flow of charm quarks is increased over elastic pQCD+resonance model by ca. 40%, and even more (ca. 100%) relative to pQCD elastic scattering only. Another limitation is the above mentioned caveat in Langevin theory of underestimating the (Eloss) fluctuations implying an overestimate of the quenching at high pt . This can also be seen when comparing to the pQCD radiative Eloss calculations,7,38,164 where the gluon radiation is treated microscopically within an opacity expansion. A consistent merging of radiative and elastic processes in HQ transport thus remains a challenging task. 3.5. Heavymeson and electron observables To compare to observables, the HQ spectra discussed in the preceding section need to be converted into spectra of colorneutral finalstate particles. At the minimal level, this requires hadronization into charm and bottom mesons and baryons. Thus, a measurement of identified HQ hadrons constitutes the most direct way to make contact with theoretical predictions. Currently, the richest source of information on HQ spectra in Au–Au collisions at RHIC are singleelectron (e± ) spectra, which, after the subtraction of sources coupling to a photon (“photonic sources”), are associated with semileptonic decays of HQ hadrons. As discussed in Sec. 3.3, the decay electrons largely preserve the modifications of the parent hadron spectra, albeit shifted in pt (by roughly a factor of ∼ 2). The more severe complication is the composition of the e± spectra, most notably the partition into charm and bottom parents.165 Since the heavier bottom quarks are, in general, less affected by the medium, their contribution significantly influences the resulting e± spectra. Unless otherwise stated, the calculations discussed below include a “realistic” input for the charm/bottom partition, i.e., either in terms of pQCD predictions for p–p spectra or via empirical estimates from Dmeson and electron spectra in p–p and dAu. Within the current theoretical and experimental uncertainties, both procedures agree, with an expected crossing of charm ad bottom electrons at pt ' 3–6 GeV in p–p collisions at RHIC energy. Almost all of the approaches for computing HQ diffusion and/or energy loss introduced in Sec. 2 have been applied to e± data at RHIC. We organize the following discussion into (mainly perturbative) Eloss calculations (usually applied within a static geometry of the nuclear reaction zone) as well as perturbative and nonpertubative diffusion calculations using Langevin simulations for an expanding medium. 3.5.1. Energyloss calculations Radiative energy loss (Eloss) of highenergy partons in the QGP is believed to be the prevalent mechanism in the suppression of light hadrons with high pT ≥ 6 GeV. It turns out that the application of this picture to the HQ sector (Sec. 2.6) cannot account for the observed suppression in the nonphotonic e± spectra.
167
1.2
Mass and scale uncertainties
1
PHENIX STAR (Prelim)
1.0 STAR PHENIX
dNg/dy 1000
v2
RAA
Heavy Quarks in the QuarkGluon Plasma
0.25
0.8
0.2
DGLV Ra Electron RAA(p )
d only
0.6
Rad
PHENIX preliminary
other effects dominant
STAR preliminary c, q = 1 GeV2/fm
2
b, q = 1 GeV2/fm
q=14 GeV/fm
0.8
0.15 + Ela
stic
e/(b+c) +G
0.6
eom
etry
0.4
0
2
4
6 p (GeV)
8
GeV2/fm
AuAu, 080%
0
sNN
0.050
10
0 0
b, q =
0.05
0.4
BT
0.2 0.0
b, q = 1 GeV2/fm
c
0.1
BT TG e/c TG
0.2
c
1
2
3
4
5
200 GeV
6 7 8 9 10 electron p [GeV] T
2
4
6
8
10
pT [GeV]
Fig. 24. (Color online) PQCD Eloss calculations for heavy quarks compared to singleelectron observables in Au–Au collisions at RHIC33,34,166 Left panel: e± nuclear modification factor, ReAA , in central collisions in the DGLV approach38 (upper band: radiative Eloss only; lower band: radiative plus elastic Eloss) for a gluon plasma with rapidity density dN g /dy = 1000; the bands represent the uncertainty in the leadinglogarithm approximation of the elastic part (as described in Sec. 2.6). Middle and right panel: ReAA in central and v2e in 080% central collisions within the BDMPS approach using a transport coefficient of qˆ = 14 GeV2 /fm36 ; dashed and dashdotted curves represent the results for c and bquark contributions separately, while the solid curve is the combined result with a band indicating the pQCD uncertainty in the charm/bottom partition. 145
In the DGLV formalism, the highpT e± suppression due to radiative Eloss of c and b quarks falls short of the data by about a factor of 3, cf. left panel of Fig. 24. 38 This led the authors to consider elastic Eloss (see also Refs. 27 and 167) which was found to be comparable to the radiative one out to the highest electron pT measured thus far (∼ 10 GeV). Their combined effect still underestimates the measured suppression by about a factor of ∼ 2 for pt > 4 GeV. Similar findings were reported within the BDMPS approach: for a transport coefficient of qˆ = 14 GeV2 /fm,g the e± spectra cannot be reproduced either, unless an unrealistic assumption of neglecting the bottom contribution is made, cf. middle panel of Fig. 24.36 Both Eloss calculations36,38 are performed for a static (timeaveraged) medium of gluons, with fragmentation as the sole mechanism for hadronization. This is expected to be a good approximation at high pT . Processes leading to an energy gain in the spectra, e.g., due to drag effects or coalescence with a light quark, are not included. Such processes lead to an increase in the pT of the finalstate hadron and thus to an increase in the electron RAA , which would augment the discrepancy with data. The neglect of the diffusive term becomes particularly apparent in the elliptic flow. In the Eloss treatment the only source of an azimuthal asymmetry in the pT spectra in noncentral Au–Au collisions is the spatial geometry of the overlap zone: particles traveling along the short axis are less likely to be absorbed g This
value is a factor 5–10 larger than predicted by pQCD, and at the upper limit of being compatible with light hadron suppression.29
168
R. Rapp & H. van Hees 1 1/2 NN
s
RAA(pT)
0 75
1/2 NN
s
200 GeV
200 GeV
B+D mesons+baryons B mesons+baryons D mesons+baryons
ξ
−
2−3
05
−
PHENIX 0.5(e +e ), 0−10% Au+Au − STAR 0.5(e +e ), 0−5% Au+Au − STAR 0.5(e +e ), 0−12% Au+Au
08 g
Central Au+Au at RHIC (dN /dy
g
Central Cu+Cu at RHIC (dN /dy
1175)
RAA(pT)
0 25 350)
0 1/2 NN
s
1/2 NN
5500 GeV
s
5500 GeV
0 75
RAA(pT)
0.5(e +e ), QGP dissociation ξ 2−3
No nuclear effect
1
06 04
B
05
02
0 25 g
Central Pb+Pb at LHC (dN /dy
0
0
Central Au+Au
D
10 20 pT [GeV]
g
2000)
Central Pb+Pb at LHC (dN /dy
0
10 20 pT [GeV]
0 0
3500)
30
2
4
6 pT [GeV]
8
10
12
Fig. 25. (Color online) Energyloss calculations employing the collisional dissociation mechanism for HQ fragmentation into heavy mesons in the QGP.77 Left panels: nuclear modification factor for D and B mesons in central 200 AGeV Au–Au and Cu–Cu at RHIC (upper panels) for gluonrapidity densities of dN g /dy = 1175 and 350, respectively, and in central 5.5 ATeV Pb–Pb collisions at LHC for dN g /dy = 2000 and 3500 (lower panels). Right panel: RAA for e± in central 200 AGeV Au–Au collisions at RHIC compared to PHENIX and STAR data 22,34 ; the yellow bands reflect theoretical uncertainties due to an impact parameter expansion of the heavylight quark interaction.
than those moving along the long axis of the approximately elliptic reaction zone. The positive v2 generated by this mechanism amounts to up to a few percent and significantly falls short of the observed electron v2 , see right panel in Fig. 24. As an alternative mechanism, the collisional dissociation of D and B mesons from HQ fragmentation in the QGP (Sec. 2.3.3) has been implemented into an Eloss calculation.77,h A rather striking prediction of this calculation is that the shorter formation time of B mesons leads to stronger suppression than for D mesons above hadron momenta of pt ' 15 GeV at RHIC, cf. left panels in Fig. 25. This turns out to be an important ingredient in the successful reproduction of the e± suppression data as shown in the right panel of Fig. 25. 3.5.2. Langevin simulations The importance of elastic scattering for HQ diffusion and Eloss has been emphasized, prior to quantitative measurements of e± spectra, in Refs. 25 and 26, albeit within rather different realizations of the underlying HQ interaction (recall Secs. 2.3.1 and 2.2.2, respectively).
h The
compatibility with the radiative picture of highpt lighthadron suppression is presumably little affected since the Lorentzdilated formation times of light quarks largely result in hadronization beyond the QGP lifetime.
Heavy Quarks in the QuarkGluon Plasma
169
20 15
PHENIX prel (min bias)
c+b reso c+b pQCD c reso
central AuAu √s=200 GeV
15 AuAu √s=200 GeV (b=7 fm)
c+b reso c+b pQCD
v2 [%]
RAA
PHENIX prel (010%) ◆ STAR prel (020%)
1
10 5
05
0 0 0
1
2
3
4
5
6
7
5 0
8
1
2
3
4
5
pT [GeV]
pT [GeV] 20 15
PHENIX prel (min bias)
c+b reso c+b pQCD c reso
central AuAu √s=200 GeV
15 AuAu √s=200 GeV (b=7 fm)
c+b reso c+b pQCD
1
v2 [%]
RAA
PHENIX prel (010%) ◆ STAR prel (020%)
10 5
05
0 0 0
1
2
3
4
pT [GeV]
5
6
7
8
5 0
1
2
3
4
5
pT [GeV]
Fig. 26. (Color online) Nuclear modification factor (left panels) and elliptic flow (right panels) of e ± spectra from heavyflavor decays in Au–Au collisions at RHIC, as computed within HQ Langevin simulations in a thermal fireball employing the resonance+pQCD model for HQ transport in the QGP.28,168 The upper panels include effects from heavylight quark coalescence while the latter are switched off in the lower panels. The bands represent the full results for D and Bmeson resonance widths of Γ = 0.4–0.75 GeV; the blue lines are obtained for LOpQCD interactions only, while the purple lines neglect bottom contributions. The data are from Refs. 33, 166 and 169.
The HQ spectra of the effective resonance + LOpQCD model25 (using Langevin simulations within an expanding fireball, recall Fig. 19) have been converted into e± spectra utilizing a combined coalescence/fragmentation scheme at Tc followed by heavymeson threebody decays.28,168 The predicted e± spectra and elliptic flow show approximate agreement with 2005 PHENIX33,166 and STAR169 data, see upper left and right panel of Fig. 26, respectively. Compared to the results for LO pQCD interactions only (blue lines), the resonance interactions (red bands) turn out to be instrumental in generating the required suppression and elliptic flow (see upper panels of Fig. 26). LOpQCD scattering alone, even with a strong coupling of αs = 0.4, does not produce sufficient coupling to the bulk medium to suppress the primordial quark spectra, nor to drag the heavy quarks along with the collective flow of the expanding fireball. The effect of heavylight quark coalescence is illustrated by a calculation where only fragmentation is used as a hadronization mechanism (lower panels in Fig. 26). In this case, the shape of the e± RAA and the magnitude
RAA
170
R. Rapp & H. van Hees 1.8 1.6
(a)
0 10% central
theory [Wo] frag only [Wo] theory [SZ] PHENIX STAR
Armesto et al. (I)
15
1.4
van Hees et al. (II)
1.2
12/(2πT) Teaney (III) 3/(2πT)
Moore &
RAA
1 0.8
1
0.6
05
0.4
vHF 2
0.2
Au+Au @ sNN
Au+Au √s=200 AGeV (010% central)
200 GeV
0 0 15
0.2
0.15
(b)
π 0 RAA , p > 4 GeV/c
minimum bias
π 0 v2, p > 2 GeV/c
T
T
0.1
RAA ,
e±
PHENIX PHENIX QM08
0 05
0.05
0
0
0
minimum bias
01
vHF 2
v2
e±
1
2
3
4
5
6
7
8
9
p [GeV/c]
0
1
2 3 pT [GeV]
4
5
Fig. 27. (Color online) Left panel: Comparison of PHENIX e± data22 to theoretical predictions based on (i) Langevin simulations with the resonance+LOpQCD interactions plus quark coalescence (bands),28 (ii) Langevin simulations with (upscaled) HTLimproved LOpQCD interactions (dashdotted and dotted lines),26 and (iii) radiative Eloss calculations (dashed lines).36 For further illustration, PHENIX data for π 0 suppression and elliptic flow are shown.170 Right panel: Comparison of PHENIX22,35 and STAR34 e± data to Langevinfireball simulations for HQ transport using T matrices with lQCDbased heavyquark potentials45 ; an uncertainty due to different lQCD internal energies is indicated by the solid and dashdotted curve; the dashed lines are obtained without the effects of heavylight quark coalescence at Tc .
of the v2 are not well reproduced. Coalescence processes add both momentum and v2 to the meson (and thus to the e± ) spectra, i.e., the suppression becomes smaller. It is furthermore instructive to compare the LOpQCD results with fragmentation only (lower left panel in Fig. 26) to the pQCD Eloss calculations, especially to the elastic DGLV results where αs = 0.3 has been used (left panel of Fig. 24). The suppression level in the pertinent electron RAA is quite comparable for a rather large range in pT . The increasing trend in the Langevin calculations for pT & 5 GeV is presumably due to the dominant bquark contribution (which is barely suppressed even in the resonance model up to pT & 5 GeV, see left of Fig. 19). Let us also estimate the impact of radiative contributions on the resonance model. Within DGLV the electron suppression due to radiative Eloss alone amounts to about 0.6–0.8 for pT ' 4–10 GeV. Upon multiplying the RAA for the resonance+pQCD model in the upper left panel of Fig. 26 with this factor, the result would be compatible with current RHIC data. The PHENIX collaboration has conducted a comprehensive comparison of their 2006 e± data22 to theoretical calculations predicting both RAA and v2 ,28,36,171 cf. left panel of Fig. 27. The interpretation reiterates some of the main points made above: (i) the missing drag in (radiative) Eloss calculations entails a substantial underprediction of the v2 ; (ii) Langevin calculations using elastic scattering require rather small HQ diffusion coefficients, Ds (2πT ) ' 46, to be compatible with the observed
Heavy Quarks in the QuarkGluon Plasma
171
level of suppression and elliptic flow, and, (iii) quark coalescence improves the simultaneous description of these two observables. The heavylight quark T matrix approach, based on input potentials estimated from thermal latticeQCD, has been applied within the same Langevinfireball + coalescence/fragmentation scheme as the effective resonance model.45 The pertinent e± spectra (cf. right panel of Fig. 27) exhibit a comparable level of agreement with current RHIC data22,34,35 as in the resonance model, with a similar uncertainty due to different extractions of the HQ internal energy. Although the T matrix calculations involve, in principle, no tunable parameters, the inherent theoretical uncertainties are appreciable (e.g., in the definition of the inmedium potential in terms of internal or free energy). Let us, however, recall a rather general feature of the T matrix approach which was visible already at the level of the HQ spectra in Fig. 21: the weak temperature dependence of, e.g., the friction coefficient implies that the HQ coupling to the medium remains rather strong in the later QGP and mixed phase stages of the evolution. Since the bulk v2 is largest in these later stages, while the suppression largely occurs in the first 12 fm/c,39 the T matrix interactions generate relatively more v2 than suppression compared to, e.g., the resonance model (or, alternatively: for the same v2 , the suppression in the T matrix approach is smaller). This traces back to the increasing colorDebye screening with increasing temperature, which leads to a gradual melting of the resonance correlations and a marked increase of the spatial diffusion constant, Ds /(2πT ), with temperature (recall Fig. 14). Such a temperature dependence appears to improve the consistency in the simultaneous description of the e± RAA and v2 , but more precise data are needed to scrutinize this feature (including dA collisions to quantify the Cronin effect, which could increase the RAA without noticably affecting v2 ). Finally, we reproduce in Fig. 28 selected results of the Boltzmanntransport approach of HQ diffusion67 with the background medium described by the hydrodynamical model of Ref. 13; hadronization is treated in a combined coalescence+fragmentation approach similar to the one in Refs. 28 and 45. The HQ interactions in the QGP are implemented via the elastic pQCD scattering amplitudes described in Sec. 2.2.3. The left panels in Fig. 28 refer to a model with fixed coupling constant, αs (2πT ), at given temperature and standard screening mass (r = 1). A large K factor of K = 12 is needed to simultaneously reproduce the electron RAA and v2 data. The final elliptic flow is found to be rather sensitive to the late QGP stages of the evolution, favoring hadronization at the end of the mixed phase (i.e., at a small transition energydensity, trans ); this is consistent with the findings of Ref. 39 and the above discussion of the T matrix interaction. It thus corroborates that anisotropic matterflow can only be transferred to the heavy quarks if the latter is sufficiently large, while Eloss (reflected in highpT suppression) is mostly effective when the fireball density is high. The simulations find little impact of the initialstate Cronin effect on the final heavyquark v2 , but the suppression is somewhat reduced primarily for e± momenta of pT ' 1–3 GeV. The right panel
December 18, 2009 10:18 WSPC/INSTRUCTION FILE
172
hqqgp4
R. Rapp & H. van Hees
Fig. 28. (Color online) Boltzmanntransport model results for electron R AA (upper panels) and v2 (lower panels)67 in a hydrodynamic evolution using different versions of LOpQCD HQ interactions, compared to PHENIX data22,35 in 200 AGeV Au–Au collisions at RHIC. The curves in the left panels are computed with fixed αs (2πT ) at given temperature, conventional infrared regulator (˜ µ2D = rµ2D with r≡κ = 1) and large Kfactors; the curves in the right panels are computed with a running αs , reduced IR regulator (r = 0.2) and reduced K factors of 1.5–2 (or 2–3), as represented by the bands. trmax or trmin indicate freezeout at the beginning or end of the mixed phase, respectively.
of Fig. 28 illustrates that similar results can be achieved with smaller Kfactors, K = 1.5–2, if the pQCD cross sections are augmented by a running coupling, αs (t) (t: 4momentum transfer in the elastic scattering process), and a small infrared regulator, µ ˜ 2D = rµ2D with r = 0.2, in tchannel gluonexchange scattering. 3.6. Viscosity? In this section we utilize the quantitative estimates for the HQ diffusion coefficient as extracted from current RHIC data to obtain a rough estimate of the ratio of shear viscosity to entropy density, η/s, in the QGP. This quantity has received considerable attention recently since (a) it allows to quantify deviations from the predictions of ideal fluid dynamics for observables like the elliptic flow, and (b) conformal field theories in the strong coupling limit are conjectured to set a universal lower bound for any liquid, given by η/s = 1/(4π),17 referred to as KSS bound. In
Heavy Quarks in the QuarkGluon Plasma
173
the following we will bracket the estimates derived from HQ observables by using relations of Ds and η/s in the weak and strongcoupling limit (the latter assumed to be given by the AdS/CFT correspondence). Following the discussion in Sec. 2.4, the strong coupling limit in the AdS/CFT framework results in a (spatial) HQ diffusion constant of Ds ' 1/(2πT ). Combining this with the lower bound of the viscositytoentropydensity quoted above, one obtains 1 η = T Ds . s 2
(85)
For a weakly coupled dilute gas, an estimate for a relation between Ds and η can be inferred starting from the calculation of the shear viscosity from kinetic theory for an ultrarelativistic gas,172,173 η≈
4 n hpi λtr , 15
(86)
where n denotes the particle density, hpi the average momentum of the gas particles and λtr its transportmean free path. With the estimate, n hpi ' , for the energy density and the corresponding equation of state, T s = + P = 4/3, one arrives at η 1 ' T λtr . s 5
(87)
Finally, equating the transport meanfree path λtr to the meanfree time τtr and taking into account the delay due to the mass effect of the heavy quark on the thermalization time, τQ ≈ τtr T /mQ , one finds η 1 ≈ T Ds . s 5
(88)
In comparison to the “strongcoupling” estimate within AdS/CFT, Eq. (86), the shear viscosity appears to be underestimated when the kinetic theory for a dilute gas is applied to liquids. These estimates are now applied to several of the HQ diffusion calculations discussed above, see the left panel of Fig. 29. Since η/s ∝ Ds (2πT ), the main features of Fig. 14 are transmitted to η/s, in particular the weak temperature dependence of the LOpQCD calculations and the effective resonance model. Of course, the absolute values of these calculations differ considerably. A different behavior is only found for the T matrix+pQCD model, which suggests a transition from a strongly coupled regime close to Tc to relatively weak coupling above ∼ 2 Tc . In fact, the uncertainty band has been constructed as follows: for the lower limit, the weakcoupling estimate Eq. (88) is used; for the upper limit, the strongcoupling limit estimate, Eq. (86), at T = 0.2 GeV is linearly interpolated with the LOpQCD weakcoupling limit at T = 0.4 GeV (the strongcoupling estimate for T matrix+pQCD overshoots the LOpQCD result at this temperature). As for the spatial diffusion constant, the increase of η/s with temperature is related to colorDebye screening of the lQCDbased potentials which entails a gradual melting of the
174 2 15
R. Rapp & H. van Hees 2.0
Tmat + pQCD reso + pQCD pQCD pQCD run αs KSS bound
η s
1.5 1.0
05
0.5
24
3 3
8 8
Perturbative Theory
η/s
1
16
KSS bound
0.0
0 charm quarks
0.5
0 5 02
0 25
03 T (GeV)
0 35
04
1
1.5
2
T Tc
2.5
3
Fig. 29. (Color online) The ratio of shear viscosity to entropy density, η/s. Left panel: schematic estimates using charmquark diffusion constants based on (a) schematic LO pQCD elastic scattering (αs = 0.4) in the weakly interacting limit, Eq. (88) (dashed line), (b) pQCD elastic scattering with running coupling constant and small IR regulator (band enclosed by dashdotted lines using the weak and strongcoupling limits), (c) the effective resonance + pQCD model in the strongcoupling limit, Eq. (86) (band enclosed by longdashed lines for Γ = 0.4–0.75 GeV), and (d) the latticeQCD potential based T matrix approach augmented by pQCD scattering off gluons (band enclosed by solid lines constructed from the weak and strongcoupling limits). Right panel: lattice QCD computations in a gluon plasma174 compared to results inferred from perturbation theory.175,176
dynamically generated resonances in the heavylight quark T matrix. It is tempting to interpret the decrease of η/s when approaching Tc from above as a precursorphenomenon of hadronization and thus connected to the phase transition itself. It remains to be seen whether a similar mechanism is operative in the lightquark and/or gluon sector (threebody interactions are unlikely to produce this due to the decrease in particle density when approaching Tc from above). Such a behavior is rather general in that it has been observed around phasetransition points for a large variety of substances, see, e.g., the discussion in Refs. 177 and 178. Finally we show in the right panel of Fig. 29 a quenched lQCD computation of η/s.174 The error bars are appreciable but the results tend to favor η/s values which are below LOpQCD calculations. The specific pQCD result included in this plot employs a nexttoleading logarithm calculation for the shear viscosity 176 and a selfconsistent hardthermalloop calculation for the entropy density.175 It is rather close to the schematic LO calculation (using αs = 0.4) in the left panel of Fig. 29.
4. Heavy Quarkonia in Medium In recent years it has become increasingly evident that observables in the heavyquarkonium and open heavyflavor sectors are intimately connected. In the original picture of charmonium suppression as a probe of colorscreening in hot and dense QCD matter179 there are no obvious such connections. Several recent developments have changed this situation. Thermal lattice QCD calculations find
Heavy Quarks in the QuarkGluon Plasma
175
that charmonium correlation functions are remarkably stable up to temperatures of ∼ 2 Tc or higher, suggestive for the survival of the ground state (ηc , J/ψ) well into the QGP. This interpretation is supported by probabilistic extractions of the pertinent quarkonium spectral functions. It implies that quarkonia can not only dissociate but also regenerate in the QGP.i It immediately follows that the yield and spectra of regenerated quarkonia are, in principle, sensitive to the abundance and momentum spectra of opencharm states in the system. E.g., for a fixed total charm number in the system, a softening of the heavyquark spectra is expected to increase c–¯ c overlap in phase space and thus enhance the probability for charmonium formation. At the same time, elliptic flow of charm quarks will imprint itself on regenerated charmonia. Furthermore, HQ interactions with light quarks (and possibly gluons) may be closely related to the interaction (or potential) between two heavy quarks. E.g., the T matrix approach discussed in the previous section is directly based on potentials which are extracted from the HQ free energy computed in lattice QCD. As we argued there, this approach to evaluate HQ diffusion has several attractive features, both theoretically (it may provide maximal interaction strength in the vicinity of Tc ) and phenomenologically (it describes current HQ observables at RHIC fairly well). In the remainder of this section we address several aspects of quarkonia in medium and in heavyion collisions with a focus on connections to the open heavyflavor sector. More extensive reviews on quarkonia in medium have recently been given in Refs. 8–10, which we do not attempt to reproduce here. In Sec. 4.1 we give a brief review of theoretical issues in the understanding of inmedium quarkonium spectral properties, in terms of thermal lattice QCD results for correlation and spectral functions and their interpretation using effective potential models (Sec. 4.1.1). The latter are employing input potentials extracted from heavyquark free energies computed in lattice QCD, thus enabling, in principle, an internal consistency check, provided a suitable potential can be defined. While color screening is a key medium effect in the potentials (governing the binding energy of the bound states), a quantitative assessment of spectral functions requires the inclusion of finitewidth effects induced by dissociation reactions and possibly elastic scattering (Sec. 4.1.2). In Sec. 4.2 we elaborate on recent developments in describing heavyquarkonium production in heavyion collisions. The main focus is on transport models which track the dissociation and regeneration of charmonia (and bottomonia) through the QGP, mixed and hadronic phases (Sec. 4.2.1), complemented by a brief discussion of initial conditions as affected by coldnuclearmatter (CNM) effects (shadowing, Cronin effect and nuclear absorption). This is followed by an assessment of the current status of charmonium phenomenology at SPS and RHIC.
i Note
that higher dissociation rates also imply higher formation rates.
176
R. Rapp & H. van Hees
4.1. Spectral properties of quarkonia in the QGP 4.1.1. Lattice QCD and potential models The phenomenological Cornell potential85 for the interaction between two heavyquark (color) charges in the colorsinglet channel, VQQ ¯ (r; T = 0) = −
4 αs + σr , 3 r
(89)
has been very successful in reproducing the vacuum spectroscopy of charmonium and bottomonium bound states. It consists of a (color) Coulomb plus a (linear) confining part with a very limited number of parameters, i.e., a strong coupling constant, αs , and string tension, σ (in addition, an effective HQ mass, mQ , needs to be specified). Subsequent developments have put this framework on a more rigorous footing by showing that (a) the potential description can be recovered as a lowenergy effective theory of QCD with heavy quarks,83,84 and (b) lattice QCD computations of the colorsinglet heavyquark free energy, F1 , have found excellent agreement with the functional form (and parameters) of the Cornell potential.180 ¯ boundstate properties in the QGP have suppleEarly calculations181 of Q–Q mented the Cornell potential by a phenomenological ansatz for color screening of both the Coulomb and confining parts, VQQ ¯ (r; T ) =
4α σ s −µD (T )r 1 − e−µD (T )r − e . µD (T ) 3r
(90)
The key quantity carrying the temperature dependence is the Debye screening mass, µD (∝ gT in thermal pQCD). Already at that time the possibility was established that groundstate charmonia (and even more so bottomonia) can survive until temperatures (well) above Tc . More recently, quantitative lQCD computations of the finitetemperature colorsinglet free energy of a HQ pair, F1 (r; T ), have become available, see, e.g., Fig. 30. The results nicely illustrate the colorscreening effect and its gradual penetration to smaller distances. When inserting the inmedium free energy as an improved estimate of the finitetemperature HQ potential into a Schr¨ odinger equation, the “melting” temperature of the J/ψ (ψ 0 , χc ) was found to be just above (below) Tc .183 Further progress in thermal lQCD came with the computation of heavy quarkonium correlation functions, Gα (τ, r) = hhjα (τ, r)jα† (0, 0)ii
(91)
(also referred to as temporal correlators), as a function of Euclidean time, τ . j α represent the creation/annihilation operators of a hadronic current of given quantum numbers, α. In the pseudoscalar and vector charmonium channels (corresponding to c–¯ c Swaves with ηc and J/ψ states, respectively), the Euclidean correlators were found to exhibit a surprisingly weak temperature dependence up to ∼ 2 Tc , even
Heavy Quarks in the QuarkGluon Plasma
1.4 12 1
F1,GeV
08 06 0.4 02
3
138 150 163 176 190 205 221 254 291 330 410
fQ− Q,0 /σ 1
1/2
T=0
2 1 T/Tc = 1 002 T/Tc = 1 02 T/Tc = 1 07 T/Tc = 1 16 T/Tc = 1 22 T/Tc = 1 50 T/Tc = 1 98 T/Tc = 3 00 T/Tc = 4 01
0
0
T/Tc = 0 81 T/Tc = 0 87 T/Tc = 0 90 T/Tc = 0 96
1
0 2
2
0.4 02
177
0.4
06
08
1
12
1.4
rσ 0
1
1/2
2
3
4
5
6
r [fm]
Fig. 30. (Color online) Free energy of a static colorsinglet HQ pair as computed in lattice QCD for Nf = 393 (left) and Nf = 2182 flavors (right). The critical temperature is Tc = 193(170) MeV for the Nf = 3(2) calculation, and the string tension typically amounts to σ 1/2 ' 420 MeV = 1/(0.45 fm).
at large τ , suggestive for rather stable bound states. The temporal correlators are related to the physical spectral function, σα (E, p; T ), via Gα (τ, p; T ) =
Z∞
dE σα (E, p; T ) K(E, τ ; T )
(92)
0
with a thermal integral kernel K(E, τ ; T ) =
cosh[E(τ − 1/2T )] . sinh[E/2T ]
(93)
Equation (92) implies that the extraction of the spectral function from the Euclidean correlator requires a nontrivial integral inversion. Especially at finite T , where periodic boundary conditions limit the information on Gα (τ, p; T ) to a finite interval, 0 ≤ τ ≤ 1/T , and for a finite number of τ points, the unambiguous inversion to obtain σα (E, p; T ) becomes an illdefined problem. However, using probabilistic methods (in particular the socalled maximum entropy method (MEM)), a statistical reconstruction of σα (E, p; T ) is possible and has been applied.41,42 The approximate constancy of the temporal correlators leads to spectral functions with rather stable groundstate peaks corroborating the notion of surviving ground states well above Tc . To resolve the apparent discrepancy with the low dissociation temperature found in the potential model discussed above, it has been suggested to employ as potential the internal rather than the free energy, which are related via F (r; T ) = U (r; T ) − T S(r; T ) .
(94)
Especially in the colorsinglet channel, the (positive) entropy contribution rises sig¯ separation, r, thus producing “deeper” potentials (cf. Fig. 31) nificantly with Q–Q
178
R. Rapp & H. van Hees
U1 [MeV]
F1 [MeV] 1500 1000
(a)
(b) 0.76Tc 0.81Tc 0.90Tc 0.96Tc 1.00Tc 1.02Tc 1.07Tc 1.23Tc 1.50Tc 1.98Tc 4.01Tc
500
0
r [fm]
500 0
0.5
1
1.5
2
2.5
1000
500
0 r [fm] 3
0
0.5
1
1.09Tc 1.13Tc 1.19Tc 1.29Tc 1.43Tc 1.57Tc 1.89Tc 1.5
2
Fig. 31. HQ free energy in the colorsinglet channel computed in thermal N f = 2 lattice QCD (left)180 and corresponding numerically extracted internal energy (right). 94
entailing stronger binding. Consequently, pertinent evaluations of quarkonium spectra lead to larger dissociation temperatures, which seemingly agree better with the lQCD spectral functions. These assertions have been made more quantitative 56–63 by employing potential models to calculate inmedium spectral functions, perform the straightforward integral in Eq. (92) and compare to the rather precise temporal correlators from lQCD. It is important to realize that the Euclidean correlators involve the pertinent spectral function at all energies. In Ref. 56 the inmedium boundstate spectrum obtained from a Schr¨ odinger equation (using either a screened Cornell potential or lQCD internal energies) has been combined with a ¯ continuum above threshold. No agreement with perturbative ansatz for the Q–Q lQCD correlators could be established. In Refs. 57 and 58 the importance of rescat¯ continuum was emphasized and implemented tering effects for the interacting Q–Q ¯ spectral functions. In Ref. 57 continuum correlainto the calculations of the Q–Q tions were implemented via Gamov resonance states in Breit–Wigner approximation, while in Ref. 58 a thermodynamic T matrix approach was employed, Tα (E) = Vα +
Z
d3 k ¯ Vα GQQ¯ (E; k) Tα (E) [1 − fQ (ωkQ ) − fQ (ωkQ )] , 3 (2π)
(95)
exactly as introduced in the context of HQ diffusion in Sec. 2.3.2, recall Eq. (46). The T matrix approach enables a consistent treatment of bound and continuum states on equal footing, as well as the implementation of medium effects (selfenergies) into the intermediate twoparticle propagator, GQQ¯ , recall Eq. (48). Pertinent results for S and P wave charmonium spectral functions, using the internal energy extracted from the Nf = 3 free energy93 (left panel of Fig. 30), are shown in Fig. 32 for a constant (T independent) charmquark mass of mc = 1.7 GeV. One clearly recognizes the reduction in binding energy as a result of color screening by the upward moving boundstate peak position with increasing temperature (the c¯ c threshold is fixed at 2mc = 3.4 GeV). Also note that nonperturbative rescattering effects close to and above threshold induce a substantial enhancement in
Heavy Quarks in the QuarkGluon Plasma
179
3.0 14
1.1 Tc 1.5 Tc 2.0 Tc 2.5 Tc 3.0 Tc 3.3 Tc uncorrelated
8
2
10
2.5
Im G [GeV ]
2
Im G [GeV ]
12
6
1.1 Tc 1.5 Tc 2.0 Tc uncorrelated
2.0 1.5 1.0
4 0.5
2 0
2.8
3.0
3.2
3.4
ω
[GeV]
3.6
3.8
4.0
0.0 2.6
2.8
3.0
3.2
3.4
ω
[GeV]
3.6
3.8
4.0
Fig. 32. (Color online) Charmonium spectral functions computed in a T matrix approach 58 using internal HQ free energies extracted from Nf = 3 thermal lattice QCD.93 For numerical purposes, ¯ a small charmquark width of Γc = 20 MeV has been implemented into the intermediate QQ propagator in the scattering equation (46).
¯ spectral function over the noninteracting continuum (indicated by the the Q–Q red longdashed lines), an effect which is of prime importance in the calculation of HQ diffusion in q–Q scattering as well. When applied to the calculation of Euclidean correlators58 in Eq. (92), the upward shift of lowenergy strength due to the moving bound states in the Swave spectral function shown in the left panel of Fig. 32 entails a suppression of G(τ ) with temperature which disagrees with the weak temperature dependence found in lQCD. Another important ingredient to understand the behavior of the correlators is the temperature dependence of the HQ mass. Schematically, the, say, J/ψ boundstate mass may be written as mJ/ψ = 2m∗c − εB .
(96)
This illustrates that a small (large) binding energy, εB , can be compensated by a small (large) effective quark mass in a way that the groundstate mass stays approximately constant. Indeed, when interpreting the asymptotic value of the inmedium potential as an effective HQ mass correction, m∗c = m0c + ∆mc ,
∆mc ≡ X(r = ∞; T )/2 ,
(97)
with X = U or F (or an appropriate combination thereof), the use of U implies strong binding with large effective quark masses while the use of F leads to weak binding with small m∗c (recall from Fig. 31 that the “U potential” is deeper than the “F potential” but features a larger asymptotic value for r → ∞). Consequently, it has been found that reasonable agreement with lQCD correlators can be achieved with different spectral functions, covering a rather large range of dissociation temperatures, e.g., slightly above Tc using screened Cornell potentials similar to the free
180
R. Rapp & H. van Hees
0.12
2
24
G/Grec
1.2Tc 1.5Tc 2.0Tc free lattice, 1.5Tc
1 0.99
0.1
τ [fm]
0.98 0.08
0.05
0.15
0.25
0.06 0.04
1.1 Tc 1.5 Tc 2.0 Tc 2.5 Tc 3.0 Tc T=0
20 16
2
σ(ω)/ω
0.14
Im G [GeV ] (Swave)
0.16
12 8 4
0.02 ω [GeV] 0 2
3
4
5
6
7
8
9
0
26
28
30
32
34
36
38
40
42
44
Ecm [GeV]
Fig. 33. (Color online) Swave charmonium spectral functions (η c or J/ψ) computed employing: (i) a screened Cornell potential within a nonrelativistic Green’s function approach (left), 61,62 and (ii) a lQCDbased Nf = 3 internal energy93 within a T matrix approach.58 Both calculations account for inmedium charmquark masses, and approximately reproduce the weak temperature dependence of the temporal correlators computed in lQCD up to temperatures of at least 2T c (see, e.g., inset in the left panel). In calculation (i) the boundstate has disappeared at temperatures below 1.5Tc , while in calculation (ii) it is visible up to ∼ 2.5 Tc .
energy (F1 ),61,62 ∼ 1.5 Tc using a linear combination of free and internal energy,57,63 or up to ∼ 2.5 Tc when using internal energies.58j , cf. Fig. 33. To resolve this redundancy, it will be necessary to develop independent means of determining the inmedium quark mass and the appropriate quantity to be identified with the HQ potential. First estimates of the HQ mass from thermal lQCD have been obtained by approximating the HQ number susceptibility within a quasiparticle model with effective quark mass.46 The results suggest a rather moderate temperature variation of the latter, which deviates significantly from the perturbative predictions up to T '3 Tc . In Refs. 80–82 hardthermalloop and HQ effective theory techniques have been applied to derive the leading terms in a perturbative and HQ mass expansion of a finite temperature potential. An interesting finding of these investigations is that the potential develops an imaginary part in the medium which arises from the Landau damping of the exchanged gluons, representing a decay channel of the HQ bound state. A more general discussion of the inmedium decay width of heavy quarkonia, which plays a central role for phenomenology in heavyion collisions, is the subject of the following section. The impact of finitewidth effects on charmonium correlators has been studied within the T matrix approach in Ref. 58, by implementing an imaginary part into the charmquark propagators. A broadening of charmonium spectral functions j We
do not address here the issue of socalled zeromode contributions to quarkonium correlators, which arise on the lattice due to the periodic boundary conditions in temporal direction (and physically correspond to HQ scattering).184 These contributions are essential to obtain quantitative agreement with the lQCD correlators in all mesonic quantumnumber channels except the pseudoscalar one (ηc ); they are rather straightforward to implement in quasiparticle approximation.
Heavy Quarks in the QuarkGluon Plasma
181
leads to an enhancement of the temporal correlators (due to additional strength at lower energies), which, however, is only a few percent for a charmonium width on the order of ΓΨ ' 100 MeV. On the one hand, such a value for the width is phenomenologically significant, as it implies that about 60% of the charmonia decay within a time of 2 fm/c. On the other hand, for larger widths, their impact on the correlators should be accounted for in quantitative comparisons to lQCD “data”. 4.1.2. Dissociation widths The spectral width of a quarkonium state propagating through matter can, in principle, receive contributions from elastic and inelastic reactions with the medium particles. Elastic scattering affects the momentum distribution of the quarkonium while inelastic interactions change its abundance (via dissociation or formation). More formally, the quarkonium acquires a complex selfenergy which can be expressed via the inmedium scattering amplitude, MΨi , folded over the (thermal) distribution, fi , of the medium particles, ΣΨ (p) =
XZ i
d3 k fi (ωi (k); T ) MΨi (p, k) . (2π)3 2ωi (k)
(98)
The real part of ΣΨ characterizes inmedium changes of the quarkonium mass while the imaginary part determines its width, ΓΨ (E) = −2 Im ΣΨ (E).k Most of the attention thus far has been directed to the inelastic reactions (rather than elastic scattering). Using the optical theorem to relate the imaginary part of the forward scattering amplitude to the cross section, one arrives at the wellknown expression ΓΨ =
XZ i
d3 k diss fi (ωk , T ) vrel σΨi (s) , (2π)3
(99)
where vrel denotes the relative velocity of the incoming particles and s = (p + k)2 the squared centerofmass energy of the Ψ–i collision. The first evaluation of the inelastic quarkonium reaction cross section with gluons was conducted in Refs. 185 and 186. Employing Coulomb wave functions for the quarkonium bound state, ¯ is the analog of photodissociation of the leadingorder process, Ψ + g → Q + Q, hydrogen, see left panel of Fig. 34. For an Swave Ψ bound state with binding energy εB , the cross section is given as a function of incoming gluon energy, k0 , by 2π σgΨ (k0 ) = 3
32 3
2
mQ εB
1/2
1 (k0 /εB − 1)3/2 . m2Q (k0 /εB )5
(100)
¯ level via inmedium effects on the addition, mass and width changes are induced at the Q–Q ¯ potential and direct Ψ → Q + Q ¯ decays, respectively. These effects can be accounted for, Q–Q ¯ T matrix, Eq. (95). e.g., in the underlying Q–Q
k In
182
R. Rapp & H. van Hees
Fig. 34. Diagrams for quarkoniumdissociation reactions via parton impact; left panel: gluodissociation185,186 ; right panel: quasifree dissociation.47
This expression has a rather pronounced maximum structure with the peak cross section reached for a gluon energy, k0max = 10 7 εB . The applicability of the gluodissociation formula should be reasonable for the free bottomonium ground state (εΥ ' 1 GeV), but borderline for J/ψ (εJ/ψ ' 0.6 GeV). Taken at face value for QGP temperatures of T = 300–400 MeV, where the typical thermal energy of (massless) gluons is around k0 = 1 GeV, the convolution of the gluodissociation cross section with a thermal gluon distribution function in Eq. (99) results in an inelastic J/ψ width (lifetime) of ΓJ/ψ ' 150–400 MeV (τJ/ψ ' 0.5–1.3 fm/c), see the dashed line in the right panel of Fig. 35. The situation changes if the quarkonium binding energy decreases due to colorscreening as discussed in the previous section (or for excited charmonia which are weakly bound even in vacuum). In this case, the peak of the gluodissociation cross section moves to smaller energies and becomes rather narrow; the loss of phase space can be basically understood by the fact that for a loosely bound Ψ state, the absorption of an onshell gluon on an (almost) onshell quark is kinematically impossible (suppressed). Consequently, with decreasing binding energy, the cross section has less overlap with the thermal gluon spectrum,47,187 leading to a decreasing width with temperature (cf. dotted line in the left panel of Fig. 35). This unphysical behavior signals the presence of other inelastic processes taking over. In Ref. 47 the socalled “quasifree” dissociation mechanism has been suggested, J/ψ+p → c+¯ c+p, where a thermal parton (p = g, q, q¯) scatters “quasielastically” off an individual heavy quark in the bound state (see right panel of Fig. 34). The p–Q scattering amplitude has been evaluated in leadingorder (LO) perturbation theory,65 including thermal parton and Debye masses and slightly modified kinematics due to the small but finite binding energy. While naively of nexttoleading order (NLO) in αs compared to gluodissociation, the additional outgoing parton opens a large phase space rendering the quasifree process significantly more efficient for weakly bound states. Therefore, it readily applies to excited states as well, enabling a treatment of all charmonia on an equal footing (which is essential even for J/ψ observables, since the latter receive significant feeddown contributions from χc and ψ 0 states, see Sec. 4.2.1 below). For a coupling constant of αs ' 0.25, the quasifree dissociation rate reaches ΓJ/ψ = 100–200 MeV for temperatures, T = 300–400 MeV (cf. solid line in the left panel of Fig. 35). These values could be substantially
Heavy Quarks in the QuarkGluon Plasma
183
QG
ΓΨ (MeV)
100
10
med
quasifree (εB ) vac gluodiss (εB ) med
gluodiss (εB ) 1
0 15
02
0 25
03
T [GeV]
0 35
04
0 45
Fig. 35. (Color online) Partoninduced dissociation widths for J/ψ at rest in a QGP as a function of temperature. Left panel: gluodissociation width based on the cross section, Eq. (100), with free (dashed line) and inmedium decreasing binding energy (dotted line), as well as “quasifree dissociation” width (J/ψ + p → c + c¯ + p) with inmedium decreasing εB (solid line).47,187 Right panel: full NLO calculation for (a) quark and (b) gluoninduced dissociation for different thermal parton masses.188
enhanced if nonperturbative p–Q scattering mechanisms are operative, much like the ones discussed in Sec. 2.3. A complete NLO calculation for partoninduced charmonium destruction has recently been carried out in Ref. 188, including the effects of inmedium reduced binding energies. The right panels of Fig. 35 show the results for quark and gluoninduced breakup of the J/ψ for αs = 0.5 and different (fixed) thermal parton masses. For temperatures around 250 MeV, the sum of both contributions (ΓJ/ψ ' 350 MeV) is about a factor ∼ 4 larger than the quasifree results (ΓJ/ψ ' 80 MeV) in the left panel, calculated for αs = 0.25.47 Thus, there is good agreement between these two calculations, since the rate is basically ∝ α 2s (for T = 250 MeV the T dependent Debyemass in Ref. 47, µD = gT , amounts to µD ' 440 MeV). The threemomentum dependence of the dissociation rate of a moving quarkonium, ΓΨ (p), has been calculated for full NLO and the quasifree rates in Refs. 189 and 190, respectively. In both calculations a weak increase of the rate with increasing threemomentum is found. Since the quasifree cross section is essentially constant, this increase is caused by the increasing flux of thermal partons encountered by the moving bound state. A similar result has been obtained in a calculation employing the AdS/CFT correspondence191 (recall Sec. 2.4 for more details on this framework and its caveats). On the other hand, gluodissociation leads to a rather pronounced decrease of the dissociation rate with increasing threemomentum, since the pertinent cross section is peaked at relatively low energies and falls off rapidly at large centerofmass energies, s = (p + k)2 . Of course, if gluodissociation becomes ineffective, its threemomentum dependence becomes immaterial. A decreasing pdependence should also be expected in quasifree dissociation if nonperturbative
184
R. Rapp & H. van Hees
(resonancelike) q–Q interactions are operative (recall Sec. 2.3), since the pertinent √ cross sections are concentrated at small s as well. To utilize quarkonium observables as a probe of QGP formation in URHICs, it is mandatory to have good control over the modifications of quarkonia in hadronic matter, in particular their inelastic reaction rates. From current lattice QCD calculations it is very difficult to extract information on excited states (say, ψ 0 ). In addition, the results in the P wave (scalar and axialvector) channels (corresponding to χc,0 and χc,1 , respectively) are sensitive to the socalled zeromode contributions which are not directly related to the boundstate properties (as briefly mentioned in a previous footnote). Potential models find that ψ 0 , χc,0 and χc,1 “melt” close to or even below Tc , suggesting substantial modifications in the hadronic phase. Even for the J/ψ, hadronic dissociation may lead to significant suppression (in addition to suppressed feeddown from ψ 0 and χc states). One main obstacle in a reliable assess¯ ment of these reactions is that lowenergy reactions of the type h+J/ψ → D+ D+X constitute a nonperturbative problem with little experimental information available to constrain effective models. An initial estimate of quarkonium dissociation by light hadrons has been obtained by using the gluodissociation cross section, Eq. (100), convoluted over the gluon distribution inside hadrons.192 Since the latter is rather soft (k0 ' 0.1 GeV), the gluon energy is in general not sufficient to break up the J/ψ, leading to a (lowinel energy) cross section of order σhJ/ψ ' 0.1 mb. Quarkexchange reactions,193 e.g., ¯ (including excited D mesons in the in mesoninduced breakup, h + J/ψ → D + D finalstate), are presumably more relevant. Effective quark models predict dissociation cross sections of order 12 mb, see, e.g., Refs. 194 and 195. An alternative approach is to construct effective hadronic models, pioneered in Ref. 196. Guiding principles are basic symmetries including gauge invariance for vector mesons (J/ψ, ρ) as well as flavor symmetries, most notably SU(4) (albeit explicitly broken by the charmquark mass) and chiral symmetry which is operative in interactions with (pseudo) Goldstone bosons (π and K).197–202 The main uncertainty in these models remains a reliable determination of the cutoff scales figuring into the hadronic vertex form factors. With cutoff values of around 1 GeV, the agreement with quark models is quite reasonable; dissociation reactions induced by ρ mesons appear to be the most important channel. Their thermal density in hadronic matter (i.e., for temperatures of ∼ 180 MeV) is not very large, so that the total J/ψ width does not exceed a few MeV, and therefore is substantially smaller than in the QGP. E.g., for a total hadron density of %h = 3%0 and a thermally averaged cross section diss of hσhJ/ψ i = 1 mb (corresponding to significantly larger peak cross sections), a diss rough estimate for the dissociation rate gives Γdiss hJ/ψ = hσhJ/ψ vrel i %h ' 5 MeV. An interesting possibility to constrain effective hadronic vertices in a rather modelindependent way is to use QCD sum rules.200,203 Pertinent estimates yield, e.g., diss a thermally averaged πJ/ψ dissociation cross section of hσπJ/ψ vrel i = 0.3 mb at T = 150 MeV, in the same range as the above estimate. We finally remark that
Heavy Quarks in the QuarkGluon Plasma
185
inmedium effects, e.g., modified spectral distributions of Dmesons, can increase the finalstate phase space and lead to an appreciable increase of the dissociation rate.204 This is especially pertinent to excited charmonia like the ψ 0 , whose mass ¯ threshold, so that a slight reduction (or broadening) in the is close to the free D D ¯ 205 Dmeson mass can open the direct decay channel, ψ 0 → DD. 4.2. Quarkonium production in heavyion collisions Similar to the open heavyflavor sector, a key objective (and challenge) in the quarkonium sector is to connect their equilibrium properties to observables in heavyion collisions, and eventually deduce more general insights about basic properties of QCD matter, e.g., color de/confinement and Debye screening. Since quarkonium states in heavyion collisions are even more rare than individual heavy quarks, it is suitable to adopt a transport treatment for their distribution functions in a realistic “background medium” whose evolution is not affected by the heavy quarks or quarkonia. The connection between observables extracted from the distribution function (after its “transport” through the medium) and the equilibrium properties discussed in the previous section is given by the coefficients and equilibrium limit of the transport equation, as elaborated in the following section, 4.2.1. The current status in comparing various model implementations to charmonium data at SPS and RHIC will be assessed in Sec. 4.2.2. 4.2.1. Quarkonium transport in heavyion collisions The (classical) Boltzmann equation describing the time evolution of the phasespace distribution function, fΨ (r, τ ; p), of an (onshell) quarkonium state, Ψ (with energy p0 = ωp = (p2 + m2Ψ )1/2 ), may be written as pµ ∂µ fΨ (r, τ ; p) = −ωp ΓΨ (r, τ ; p) fΨ (r, τ ; p) + ωp βΨ (r, τ ; p)
(101)
(a meanfield term has been neglected assuming that the real part of the Ψ selfenergy is small). When focusing on inelastic reactions, ΓΨ (r, τ ; p) represents the dissociation rate discussed in Sec. 4.1.2 above, which governs the loss term, i.e., the first term on the righthandside (rhs) of Eq. (101). The (r, τ ) dependence of ΓΨ typically converts into a temperature dependence via the fireball evolution of √ a heavyion reaction for given projectile/target (A/B), collision energy ( s) and impact parameter (b). The second term on the rhs of Eq. (101) is the gain term accounting for the formation of quarkonia. For a 2 → 2 process (as, e.g., realized ¯ → g + Ψ), it takes the form206 via the inverse of gluodissociation, Q + Q βΨ (p; r, τ ) =
1 2p0
Z
d3 pQ¯ d3 k d3 p Q ¯ f Q (pQ ; r, τ ) f Q (pQ¯ ; r, τ ) (2π)3 2ωk (2π)3 2ωpQ (2π)3 2ωpQ
4 (4) × WQgΨ (p + k − pQ − pQ¯ ) . ¯ (s) Θ[Tdiss − T (r, τ )] (2π) δ Q
(102)
186
R. Rapp & H. van Hees
To ensure detailed balance, the cross section figuring into the formation probabilform ity, WQgΨ vrel 4 ωpQ ωpQ , has to be the same (up to a kinematic and ¯ ¯ (s) = σQQ→gΨ Q statistical factor) as the one used in the dissociation rate, Eq. (99). For reactions ¯ → p + Ψ), the microscopic beyond 2 ↔ 2 (such as the quasifree process, p + Q + Q evaluation of the gain term becomes more involved. Note the explicit dependence ¯ on the HQ phasespace distribution functions, f Q,Q , in Eq. (102), whose modifications in heavyion reactions are the central theme in Sec. 3 of this review. The temperaturedependent step function in Eq. (102) signifies the limit set by the dissociation temperature, Tdiss , above which a welldefined Ψ state no longer exists and thus formation reactions are not meaningful. It is both instructive and useful for practical applications to simplify the gain term by integrating out its spatial and threemomentum dependence. This is possible under the assumption of a homogeneous medium and thermally equilibrated HQ distribution functions; one obtains207 dNΨ eq = −ΓΨ (NΨ − NΨ ), dτ
(103)
which now clearly exhibits detailed balance in terms of the approach to the equieq librium limit, NΨ , of the state Ψ. The latter quantity is given by Z d3 p eq eq 2 NΨ = VFB nΨ (mΨ ; T, γQ ) = dΨ γQ fΨ (ωp ; T ) , (104) (2π)3 carrying the explicit dependence on the (inmedium) quarkonium mass, mΨ (which, ¯ binding energy and inmedium HQ in turn, depends on a combination of Q–Q mass, cf. Eq. (96)); dΨ denotes the spin degeneracy of the Ψ state and VFB the (timedependent) fireball volume. The appearance of a HQ fugacity, γQ = γQ¯ , owes ¯ its origin to the (theoretically and experimentally supported) postulate that Q Q production is restricted to hard N –N collisions upon initial nuclear impact. The number NQQ¯ = NQ = NQ¯ of heavy anti/quarks is then conserved in the subsequent fireball evolution (separately for charm and bottom), which is achieved by introducing γQ at given fireball volume and temperature into the thermal densities of open and hidden HQ states, i.e., NQQ¯ =
X eq 1 I1 (Nop ) 2 Nop + VFB γQ nΨ (T ) , 2 I0 (Nop )
(105)
Ψ
for either charm (Q = c) or bottom (Q = b). The thermal opencharm (bottom) number, Nop , depends on whether one is evaluating it in terms of individual quark ∗ states, Nop = VFB γQ 2neq for a (weakly interacting) QGP), Q (mQ , T ) (appropriate P eq or in terms of hadronic states, Nop = VFB γQ α nα (T, µB ). This is, in principle, a nontrivial issue, since both hadronic and partonic evaluations of Nop can be subject to corrections, see, e.g., Refs. 207, 53 and 208. In the hadronic phase one expects the spectral functions of D mesons, Λc baryons, etc. to undergo significant medium effects, e.g., reduced masses and/or increased widths. In the partonic phase,
Heavy Quarks in the QuarkGluon Plasma
187
especially close to Tc , it is not inconceivable that hadronic bound states are still present and thus an approximation with weakly interacting quasiquarks may not be an accurate one. Even within a quasiquark description, significant uncertainty is associated with the value of the HQ mass adopted in the calculation of Nop and thus in the quantitative determination of γQ . The general trend is that for a given temperature, volume and NQQ¯ , a smaller value for m∗Q results in a larger eq value for neq Q and thus in a smaller value for γQ , which, in turn, reduces NΨ quadratically. The underlying physics is that of relative chemical equilibrium: for a given number of heavy anti/quarks, the latter preferentially occupy the states of the lowest energy. In the simplest case, where a quasiquark description applies and the Ψ mass is given by the expression (96), the Ψ number is essentially determined eq by its binding energy (larger εB implying larger NΨ ). The gain term as written in Eq. (102) is, strictly speaking, only applicable in the quasiquark approximation. If additional resonances are present in the medium (e.g., Dmeson resonances or cq diquark states), additional reaction channels would have to be included in a coupled rateequation framework to account for the competition of these resonances to harbor c quarks. In the simplified treatment given by Eq. (103), this competition eq is included via the cquark fugacity figuring into NΨ . A slightly different view on regeneration and suppression processes in the QGP is advocated in Ref. 209, based on the strongly coupled nature of the QGP (sQGP) as produced at SPS and RHIC (i.e., at not too high temperatures). It is argued that a small charmquark diffusion constant (cf. Secs. 2 and 3) inhibits the separation of the produced c and c¯ pair in the sQGP. In connection with the survival of J/ψ bound states well above Tc (as, e.g., in the right panel of Fig. 33), this enhances the probability for a produced c¯ c pair to bind into a charmonium state (relative to p–p collisions). In particular, this approach accounts for the possibility that the pairwise produced c and c¯ quarks do not explore the entire fireball volume as usually assumed in equilibrium models. Such an effect has also been implemented in a more simplified manner in the thermalrate equation approach of Refs. 207 and 190 in terms of a timedependent correlation volume. Let us briefly discuss the initial conditions for the quarkonium distribution functions. Starting point are measured quarkonium spectra in p–p collisions. In a heavyion collision, these are subject to modifications before the medium can be approximated with a thermal evolution. “Preequilibrium” effects may be distin¯ guished according to whether they occur before or after the hard QQproduction process takes place. The former include nuclear modifications of the parton distribution functions generically denoted as “shadowing”, as well as pt broadening (Cronin effect) attributed to a scattering of the projectile/target partons on their ¯ way through the target/projectile nucleus prior to the fusion reaction into Q– Q. In a randomwalk picture, the accumulated transverse momentum is approximated by ∆p2t = agN hli, where hli is an average nuclear path length of both gluons before the hard scattering, and agN parameterizes the transversemomentum kick per path length in gluonnucleon scattering. Both “prefusion” effects are in principle
188
R. Rapp & H. van Hees
¯ production process. In p–p collisions, universal, i.e., not directly linked to the QQ ¯ a fraction of 1–2% of c¯ c pairs (∼ 0.1% of bb pairs) develop a correlation that leads to the formation of a charmonium (bottomonium) state.210 In nuclear collisions ¯ pair with passingby nucleons can destroy inelastic collisions of the produced QQ this correlation. This socalled nuclear absorption may be parameterized by an efΨN l fective absorptioncross section, σabs . As is wellknown, the finite (and different) ¯ pair interacting with a nucleon formation times of quarkonia imply that the QQ does, in general, not represent a fully formed quarkonium, but rather a preresonance state. A microscopic description of nuclear absorption is therefore a rather challenging task.211,212 At a minimal level, finite formation times imply that the values for effective nuclear absorption cross sections should be expected to depend √ on collision energy ( s), rapidity (y) and boundstate quantum numbers (since different binding energies imply different formation times). A careful measurement and systematic interpretation of quarkonium suppression in p–A collisions, where the formation of a thermal medium is not expected (at least at SPS and RHIC), is therefore an inevitable prerequisite for quantitative interpretations of heavyion data, see, e.g., Refs. 213–215 for recent work. Preequilibrium effects not only modify the momentum dependence of the quarkonium distribution functions but also their spatial dependence.
4.2.2. Quarkonium phenomenology in heavyion collisions As discussed in the Introduction, there is ample evidence for both chemical and thermal equilibration in the lowpt regime of (bulk) particle production in ultrarelativistic heavyion collisions. A well defined set of thermodynamic variables characterizing the temperature evolution and flow fields of the fireball greatly facilitates the comparison of independent calculations for quarkonium production and maintains direct contact to their inmedium properties in equilibrated QCD matter. In this section we therefore focus on rateequation approaches implemented into thermal background media. We recall that the experimental quarkonium yields usually include feeddown contributions due to decays of higher resonances. E.g., for J/ψ production in p–p collisions about 30% (10%) of the observed number arises from decays of χc (ψ 0 ) states.216,217 The standard assumption in heavyion collisions is that primordial production fractions of excited states scale as in p–p collisions, but subsequent suppression (and/or regeneration) will change these ratios (due to different inelastic cross sections at all stages). This needs to be taken into account for realistic comparisons to heavyion data (unless otherwise stated, it is included in the theoretical l Nuclear
absorption typically occurs at a rather large Ψ–N centerofmass energy (comparable to √ the s of primordial N –N collisions) and is therefore in a very different energy regime than the lowenergy hadronic absorption cross section relevant for the later hadrongas stage of the fireball evolution.
Bµµσ(J/Ψ)/σ(DY)2.94.5
NA50 pT indep
PbPb (17.3 A GeV)
35
eq
coal τc ~7fm/c
30
eq τc ~3fm/c
coal total Nuc Abs
25 20
106 x (dNJ/ψ /dy) / Ncoll
Heavy Quarks in the QuarkGluon Plasma
1.6
189
√sNN =17.3 GeV NA50 data
1.4
2004 (norm.) 1998
1.2 1 0.8
15 0.6
10 0.4
5 0.2
0
0
20
40
60
80
100
120
0
ET (GeV) (GeV/c)
PbPb (17.3 A GeV)
150
200
250
300
350
400
1.25
1.225
2
√sNN =17.3 GeV
1.2
1.175
1.8
1.15
NA50 1995 NA50 1996 NA50 2000 Nuc. Abs. direct coalescence total
1.6
2
2
100
Npart
2.2
(GeV )
SHM, dσcc/dy=5.7 µb (2x pQCD) 50
1.4 1.2 1
0
1.125 1.1
40
60
80
ET (GeV)
100
120
T=160 MeV, β=0.22
1.05 1.025 1
20
NA50 data
1.075
50
100
150
200
250
300
350
400
Npart
Fig. 36. (Color online) J/ψ production at SPS in the thermal rateequation approach 190,207 (left panels) and the statistical hadronization model221 (right panels). Upper panels display the centrality dependence of the inclusive J/ψ yield (normalized to DrellYan production or the number of binary N –N collisions), lower panels the average J/ψ transverse momentum (squared in the left panel).
models discussed below). After thermal freezeout, the decay branchings are assumed to be as in vacuum (since the J/ψ lifetime after freezeout (ca. 2000 fm/c) is about a factor of ∼ 200 larger than the fireball lifetime, inmedium dilepton decays contribute a small fraction to the spectrum observed in the detectors). √ Let us start by analyzing J/ψ production in Pb–Pb( s = 17.3 AGeV) collisions at SPS in the context of NA50 data,48,218–220 cf. Fig. 36. The left panels display the outcome of thermalrate equation calculations,190,207 where quasifree dissociation rates in the QGP (cf. left panel of Fig. 35) and hadronic SU(4) cross sections for mesoninduced dissociation in the HG are evolved over an expanding fireball model (adjusted to empirical information on hadron production and flow velocities). The prevalent effect is identified as suppression in the QGP, controlled by an effective strong coupling constant, αs ' 0.25, in the quasifree rate. This value is adjusted to reproduce the suppression level in central collisions (where the average initial temperature amounts to about T0 ' 210 MeV). Regeneration is a rather small effect, based on a p–p opencharm cross section of σc¯c = 5.5 µb distributed over
190
R. Rapp & H. van Hees
two fireballs (the covered rapidity window amounts to ∆y = 3.6, resulting in a rapidity density of dσc¯c /dy ' 1.53 µb. A recent compilation222 of charm production at fixedtarget energies finds a total cross section of σc¯c ' 3.6–5.2 µb; with an experimental rapidity width of around ∆y = 2,223 the resulting rapidity density is approximately dσc¯c /dy = (2.2 ± 0.5) µb). In addition, a correction for incomplete charmquark thermalization has been implemented207 via a kinetic relaxation time (τceq ) reducing the equilibrium J/ψ number. The lower left panel of Fig. 36 suggests that the centrality dependence of the average J/ψ’s transversemomentum squared, hp2t i, is largely governed by the Cronin effect as extracted from experimental p–A data.219 The quasifree charmonium dissociation rates, which increase with threemomentum,190 lead to a slight suppression of hp2t i for central collisions. The right panels in Fig. 36 are calculated within the statistical hadronization model,221 assuming that all primordial charmonia are suppressed and production entirely occurs at the critical temperature for hadronization based on relative chemical equilibrium of open and hiddencharm hadrons (with Nc¯c fixed as in Eq. (105)). This also implies that the charmquark momentum distributions are kinetically equilibrated. With a rapidity density for the p–p charm cross section of dσc¯c /dy = 5.7 µb the NA50 data can be reproduced reasonably well. This input c¯ c number is larger by a factor of ∼ 4 compared to the input in the left panels, which accounts for most of the difference to the regeneration yield in the rateequation calculation for central collisions (remaining discrepancies are largely due to the cquark relaxation correction which becomes more pronounced toward more peripheral collisions).m The interpretation of the J/ψ’s average pt is also rather different, in that it entirely stems from a thermal source (with moderate collective flow) in the vicinity of Tc (the resulting hp2t i is quite consistent with the regeneration component in the lower left panel of Fig. 36). NA50 has also measured ψ 0 production.224,225,227 Using p–A collisions, the exψ0 tracted nuclear absorption cross section has been updated227 to σnuc = (7.7 ± J/ψ 0.9) mb, which is significantly larger than for J/ψ, σnuc = (4.2 ± 0.5) mb. In Pb–Pb √ collisions ( s = 17.3 AGeV), the ratio ψ 0 /(J/ψ) is suppressed substantially already in rather peripheral collisions, cf. Fig. 37.224,225 Within the thermal rateequation approach,207 this behavior cannot be explained by inelastic reactions in the QGP alone, since very little (if any) QGP is formed in peripheral Pb–Pb collisions at SPS. However, hadronic dissociation of the ψ 0 can account for the suppression pattern, but only if inmedium effects are included (cf. left panel of Fig. 37),207,226 ¯ threshold which accelerates ψ 0 suppression due specifically a reduction of the D D ¯ (a similar effect can result from a to the opening of the direct decay mode, ψ 0 → DD broadening of the Dmeson spectral functions as discussed in Sec. 2.7). The updated (larger) ψ 0 nuclear absorption cross section227 also plays a significant role in the quantitative description of the lowcentrality data. The statistical hadronization m The
charm ensemble at SPS is in the canonical limit, Nop 1, for which I1 (Nop )/I0 (Nop ) ' 0.5Nop in Eq. (105), and thus Nψ ∝ Nc¯c .
December 18, 2009 10:18 WSPC/INSTRUCTION FILE
hqqgp4
Bµµ(Ψ’)σΨ’/Bµµ(J/Ψ)σJ/Ψ
Heavy Quarks in the QuarkGluon Plasma
0.02
Pb(158AGeV)Pb
191
Pb Pb 95 PbPb 96 PbPb QM04 mD m*
average pp and pA
D Ψ’
0.015
σabs=7.9±0.6mb
0.01
0.005
0
thermal ratio
0
50
100
150
200
250
300
350
400
450
Npart √ Fig. 37. (Color online) NA50 data for the ψ 0 /(J/ψ) ratio in Pb–Pb( s = 17.3 AGeV) collisions224,225 compared to calculations within the thermal rateequation approach 207,226 (left panel) and the statistical hadronization model221 (right panel). In the left panel, the dashed (solid) curve is obtained without (with) the inclusion of hadronic medium effects (modeled by reduced Dmeson 0
J/ψ
ψ masses); both calculations utilize a ψ 0 nuclear absorptioncross section of σnuc = σnuc = 4.4 mb, while in the third calculation (dashdotted line) the value has been updated to 7.9 mb (with the associated band indicating a ±0.6 mb uncertainty). In the right panel, the statistical model yield involves either the full hadronization volume (dashdotted line) or excludes a dilute nuclear surface (“corona”), where neither suppression nor regeneration is operative (solid line).
model predicts a flat ψ 0 /(J/ψ) ratio, given by the thermal densities at the hadronization temperature. The shape and magnitude of the calculated ratio is rather consistent with the NA50 data for central and semicentral collisions where hadronization from a QGP can be expected to be applicable; deviations occur for more peripheral centralities. Thus, the ψ 0 /(J/ψ) ratio does not provide a clear discrimination of regeneration and suppressiondominated scenarios at SPS. One of the controversies in the interpretation of the NA50 data has been whether they feature any “sharp” drop in their centrality dependence, e.g., around ET ' 35 GeV (or Npart ' 120) in the upper left (right) panel of Fig. 36. Such a drop has been associated with a threshold behavior for QGP formation resulting in an abrupt “melting” of the χc states due to color screening229 (recall that χc feeddown presumably makes up ∼ 30% of the inclusive J/ψ yield). The investigation of this question was one of the main objectives of the successor experiment of NA50, NA60, where J/ψ production in a medium size nuclear system (InIn) has been measured.228 Figure 38 compares the NA60 J/ψ data as a function of centrality to three theoretical predictions,187,229,230 all of which reproduce the NA50 data reasonably well. The predictions of the thermal rateequation approach187,207 roughly reproduce the onset and magnitude of the suppression (except for the most central data points); the thresholdmelting scenario229 misplaces the onset of the suppression (which in the data is below Npart = 100, contrary to the Pb–Pb system) and the comover calculation230 overpredicts the suppression. The levelingoff (or even
R. Rapp & H. van Hees
Measured / Expected J/ψ yield
192
1.3 1.2 1.1 1 0.9 0.8 0.7 0.6 0.5 0
50
100
150
200
250
Npart
√ Fig. 38. (Color online) NA60 data for the centrality dependence of J/ψ production in InIn ( s = 228 17.3 AGeV) collisions, compared to theoretical predictions based on (from top to bottom): (a) the thresholdmelting scenario229 (dashed line), (b) the thermal rateequation approach187 (dashdotted line) and (c) the comover suppression approach230 (dotted line). Nuclear absorption effects have been divided out of the data and calculations based on measured suppression in p–A collisions.
increasing trend) of the data for Npart ≥ 150 is somewhat unexpected and deserves further study. The thermal rateequation framework has been used to predict J/ψ production at RHIC.207,n With updates190 for the experimental input (a smaller nuclear absorption cross section232 and a larger J/ψ number in p–p collisions49 which figures into the denominator of the nuclear modification factor and leads to a relative reduction of the regeneration yield), an approximate agreement with current PHENIX data on the centrality dependence of inclusive J/ψ production and pt spectra emerges, see left panels of Fig. 39 (the underlying charm cross section in p–p, σc¯c = 570 µb, translates into dσc¯c /dy ' 100 µb, consistent with PHENIX measurements233,234 ). The main features of this interpretation are an about equal share of (suppressed) primordial and regenerated J/ψ’s in central Au–Au collisions (where the average initial temperature is about T0 = 370 MeV), as well as a significant reduction of the average p2t due to secondary production, as compared to primordial production with an estimated Cronin effect (the latter is not yet accurately determined from p–A data). Consequently, the regeneration component is concentrated in the lowpt regime of the spectra. The dependence on the kinetic relaxation time for c quarks is rather moderate while the inclusion of the momentum dependence in the quasifree dissociation rate190 has little effect.
n The
underlying fireball model is the same as used in the open heavyflavor sector in connection with Figs. 19, 21, etc., with average initial temperatures of T0 ' 340–370 MeV for semi/central Au–Au collisions.
Heavy Quarks in the QuarkGluon Plasma
1.4
AuAu (200 A GeV) y ∼ 0.4 .
(148)
This bound is quite safe and has been found by all groups which have tried to reproduce the observed flow. • The preferred value of η/s is (see Fig. 15 and Fig. 21) η/s ' (1 ↔ 3) ×
1 . 4π
To reduce these constraints further several items need to studied and quantified. • At the end of the collision kinetic assumptions need to be made. The uncertainties involved with the particle content and the quadratic ansatz to the distribution need to be quantified to a much greater extent than has been done so far. The theoretical motivations and limitations of the quadratic ansatz have been discussed in Sec. 5. In Sec. 6.5 we have taken the nascent steps to quantify these uncertainties with an independent analysis of Fig. 23 due to Luzum and Romatschke. • The nucleusnucleus collision terminates as the system enters the transition region and starts expanding three dimensionally. Some estimates for this transition process have been given in Sec. 4.5. Clearly the transition region sets a very definite scale e ' 0.6 ↔ 1.2 GeV/fm3 which can not be ignored. It is difficult to separate the rapid η(e) dependence in this region from the shear
262
D. A. Teaney
viscosity well into the QGP phase (See Fig. 21). This scale influences the dependence of v2 on centrality and hampers an extraction of η/s from this size dependence. It will be important to categorize in a model independent way (e.g. η(e)) how this scale influences the final flow and the final estimates of η/s. Finally, while this review has focused squarely on elliptic flow, the short transport time scales estimated here have implications for a large number of other observables — energy loss,1, 2 the ridge and Mach cones,116–118 heavy quarks,119, 120 and many more. Ultimately these observables will provide a more complete picture of strongly coupled dynamics near the QCD phase transition. Acknowledgments I gratefully acknowledge help from Raimond Snellings, Kevin Dusling, Paul Romatschke, Peter Petreczky, Denes Molnar, Guy Moore, Pasi Huovinen, Tetsufumi Hirano, Peter Steinberg, Ulrich Heinz, Thomas Schaefer, and Carsten Greiner. Any misstatements and errors reflect the shortcomings of the author. I also am grateful to the organizers and participants of the Nearly Perfect Fluids Workshop which clarified the appropriate content of this review. D.T. is supported by the U.S. Department of Energy under an OJI grant DEFG0208ER41540 and as a Sloan Fellow. References 1. J. Adams et al. [STAR Collaboration], Nucl. Phys. A 757, 102 (2005) [arXiv:nuclex/0501009]. 2. K. Adcox et al. [PHENIX Collaboration], Nucl. Phys. A 757, 184 (2005) [arXiv:nuclex/0410003]. 3. D. Molnar and M. Gyulassy, Nucl. Phys. A 697, 495 (2002) [Erratumibid. A 703, 893 (2002)] [arXiv:nuclth/0104073]. 4. D. Teaney, Phys. Rev. C 68, 034913 (2003) [arXiv:nuclth/0301099]. 5. P. Romatschke and U. Romatschke, Phys. Rev. Lett. 99, 172301 (2007) [arXiv:0706.1522 [nuclth]]. 6. Z. Xu, C. Greiner and H. Stocker, Phys. Rev. Lett. 101, 082302 (2008) [arXiv:0711.0961 [nuclth]]. 7. Z. Xu and C. Greiner, Phys. Rev. C 79, 014904 (2009) [arXiv:0811.2940 [hepph]]. 8. H. J. Drescher, A. Dumitru, C. Gombeaud and J. Y. Ollitrault, Phys. Rev. C 76, 024905 (2007) [arXiv:0704.3553 [nuclth]]. 9. H. Song and U. W. Heinz, Phys. Rev. C 77, 064901 (2008) [arXiv:0712.3715 [nuclth]]. 10. K. Dusling and D. Teaney, Phys. Rev. C 77, 034905 (2008) [arXiv:0710.5932 [nuclth]]. 11. D. Molnar and P. Huovinen, J. Phys. G 35, 104125 (2008) [arXiv:0806.1367 [nuclth]]. 12. G. Ferini, M. Colonna, M. Di Toro and V. Greco, Phys. Lett. B 670, 325 (2009) [arXiv:0805.4814 [nuclth]]. 13. P. Danielewicz and M. Gyulassy, Phys. Rev. D 31, 53 (1985).
Viscous Hydrodynamics and the Quark Gluon Plasma
263
14. G. Policastro, D. T. Son and A. O. Starinets, Phys. Rev. Lett. 87, 081601 (2001) [arXiv:hepth/0104066]. 15. P. Kovtun, D. T. Son and A. O. Starinets Phys. Rev. Lett. 94, 111601 (2005) [arXiv:hepth/0405231]. 16. JeanYves Ollitrault, Phys. Rev. D 46, 229245 (1992). 17. JeanYves Ollitrault, private communication. 18. T. Hirano and Y. Nara, Nucl. Phys. A 743, 305 (2004) [arXiv:nuclth/0404039]. 19. D. Teaney, J. Lauret and E. V. Shuryak, arXiv:nuclth/0110037. ibid., Phys. Rev. Lett. 86, 4783 (2001) 20. P. F. Kolb, P. Huovinen, U. W. Heinz and H. Heiselberg, Phys. Lett. B 500, 232 (2001) [arXiv:hepph/0012137]. 21. P. Huovinen, P. F. Kolb, U. W. Heinz, P. V. Ruuskanen and S. A. Voloshin, Phys. Lett. B 503, 58 (2001) [arXiv:hepph/0101136]. 22. C. Nonaka and S. A. Bass, Phys. Rev. C 75, 014902 (2007). 23. P. Huovinen and P. V. Ruuskanen, Ann. Rev. Nucl. Part. Sci. 56, 163 (2006) [arXiv:nuclth/0605008]. 24. P. F. Kolb and U. W. Heinz, arXiv:nuclth/0305084. 25. R. Baier, P. Romatschke, D. T. Son, A. O. Starinets and M. A. Stephanov, JHEP 0804, 100 (2008) [arXiv:0712.2451 [hepth]]. 26. M. A. York and G. D. Moore, arXiv:0811.0729 [hepph]. 27. B. Betz, D. Henkel and D. H. Rischke, arXiv:0812.1440 [nuclth]. 28. S. Bhattacharyya, V. E. Hubeny, S. Minwalla and M. Rangamani, JHEP 0802, 045 (2008) [arXiv:0712.2456 [hepth]]. 29. P. Huovinen and D. Molnar, Phys. Rev. C 79, 014906 (2009) [arXiv:0808.0953 [nuclth]]. 30. H. Song and U. W. Heinz, Phys. Rev. C 78, 024902 (2008) [arXiv:0805.1756 [nuclth]]. 31. M. Luzum and P. Romatschke, Phys. Rev. C 78, 034915 (2008) [arXiv:0804.4015 [nuclth]]. 32. V. Greco, M. Colonna, M. Di Toro and G. Ferini, arXiv:0811.3170 [hepph]. 33. Z. Xu and C. Greiner, Phys. Rev. C 71, 064901 (2005) [arXiv:hepph/0406278]. 34. C. Gombeaud and J. Y. Ollitrault, Phys. Rev. C 77, 054904 (2008) [arXiv:nuclth/0702075]. 35. M. L. Miller, K. Reygers, S. J. Sanders and P. Steinberg, Ann. Rev. Nucl. Part. Sci. 57, 205 (2007) [arXiv:nuclex/0701025]. 36. P. Steinberg, private communication. 37. see for example, S. A. Voloshin, A. M. Poskanzer and R. Snellings, arXiv:0809.2949 [nuclex]. 38. R. Snellings, private communication. 39. The data presented are from the thesis Dr. Yuting Bai. 40. see for example, B. I. Abelev et al. [STAR Collaboration], Phys. Rev. C 77, 054901 (2008) [arXiv:0801.3466 [nuclex]]. 41. H. J. Drescher, A. Dumitru and J. Y. Ollitrault, arXiv:0706.1707 [nuclth]; in N. Armesto et al., “Heavy Ion Collisions at the LHC  Last Call for Predictions,” J. Phys. G 35, 054001 (2008). 42. For discussion see, A. Dumitru et al., Phys. Lett. B 460, 411 (1999). 43. S. S. Shi, for the STAR Collaboration. Presented at Quark Matter 2009. 44. For a recent account see, D. Kharzeev, E. Levin and M. Nardi, arXiv:0707.0811 [hepph]. 45. D. Kharzeev, E. Levin and M. Nardi, Nucl. Phys. A 730, 448 (2004) [Erratumibid.
264
D. A. Teaney
A 743, 329 (2004)] [arXiv:hepph/0212316]. 46. T. Hirano, U. W. Heinz, D. Kharzeev, R. Lacey and Y. Nara, Phys. Lett. B 636, 299 (2006) [arXiv:nuclth/0511046]. 47. H. J. Drescher, A. Dumitru, A. Hayashigaki and Y. Nara, The eccentricity in heavyion collisions from color glass condensate, Phys. Rev. C 74, 044905 (2006) [arXiv:nuclth/0605012]. 48. T. Lappi and R. Venugopalan, Phys. Rev. C 74, 054905 (2006) [arXiv:nuclth/0609021]. 49. J. Y. Ollitrault, A. M. Poskanzer and S. A. Voloshin, arXiv:0904.2315 [nuclex]. 50. R. S. Bhalerao and J. Y. Ollitrault, Phys. Lett. B 641, 260 (2006) [arXiv:nuclth/0607009]. 51. B. Alver et al. [PHOBOS Collaboration], Phys. Rev. Lett. 98, 242302 (2007) [arXiv:nuclex/0610037]. 52. M. Miller and R. Snellings, “Eccentricity fluctuations and its possible effect on elliptic flow arXiv:nuclex/0312008. 53. S. A. Voloshin, A. M. Poskanzer, A. Tang and G. Wang, Phys. Lett. B 659, 537 (2008) [arXiv:0708.0800 [nuclth]]. 54. B. Alver et al., Phys. Rev. C 77, 014906 (2008) [arXiv:0711.3724 [nuclex]]. 55. A. Bazavov et al., arXiv:0903.4379 [heplat]. 56. K. Itakura, O. Morimatsu and H. Otomo, Phys. Rev. D 77, 014014 (2008) [arXiv:0711.1034 [hepph]]. 57. D. Teaney, Phys. Rev. D 74, 045025 (2006). 58. A. J. Kox, S. R. de Groot, and W. A. van Leeuwen, Physica 84A, 155 (1976). 59. M. Prakash, M. Prakash, R. Venugopalan and G. Welke, Phys. Rept. 227, 321 (1993); see also R. Venugopalan, Ph.D. thesis, Stony Brook (1992). 60. E. Beth and G. E. Uhlenbeck, Physica 4, 915 (1937). 61. R. Venugopalan and M. Prakahsh, Nucl. Phys. A 546, 718 (1992). 62. N. Demir and S. A. Bass, arXiv:0812.2422 [nuclth]. 63. P. Arnold, G. D. Moore and L. G. Yaffe, JHEP 0305, 051 (2003) [arXiv:hepph/0302165]. 64. G. Baym, H. Monien, C. J. Pethick and D. G. Ravenhall, Phys. Rev. Lett. 64 (1990) 1867. 65. R. Baier, A. H. Mueller, D. Schiff and D. T. Son, Phys. Lett. B 502, 51 (2001) [arXiv:hepph/0009237]. 66. P. Aurenche, F. Gelis, R. Kobes and H. Zaraket, Phys. Rev. D 58, 085003 (1998) [arXiv:hepph/9804224]. 67. See for example, M. Le Bellac, Thermal Field Theory (Cambridge University Press, Cambridge, England, 1996). 68. H. B. Meyer, arXiv:0809.5202 [heplat]. 69. G. Aarts, C. Allton, J. Foley, S. Hands and S. Kim, Phys. Rev. Lett. 99, 022002 (2007) [arXiv:heplat/0703008]. 70. G. Aarts and J. M. Martinez Resco, JHEP 0204, 053 (2002) [arXiv:hepph/0203177]. 71. P. Petreczky and D. Teaney, Phys. Rev. D 73, 014508 (2006). 72. see for example, T. Schaefer and D. Teaney, arXiv:0904.3107 [hepph]. 73. P. Arnold, C. Dogan and G. D. Moore, Phys. Rev. D 74, 085021 (2006) [arXiv:hepph/0608012]. 74. D. Kharzeev and K. Tuchin, JHEP 0809, 093 (2008) [arXiv:0705.4280 [hepph]]. 75. G. D. Moore and O. Saremi, JHEP 0809, 015 (2008) [arXiv:0805.4201 [hepph]]. 76. K. Huebner, F. Karsch and C. Pica, Phys. Rev. D 78, 094501 (2008) [arXiv:0808.1127 [heplat]].
Viscous Hydrodynamics and the Quark Gluon Plasma
265
77. J. D. Bjorken, Phys. Rev. D 27, 140 (1983). 78. G. Baym, Phys. Lett. B 138 (1984) 18. 79. M. Gyulassy, Y. Pang and B. Zhang, Nucl. Phys. A 626, 999 (1997) [arXiv:nuclth/9709025]. 80. see for example, A. Milov [PHENIX Collaboration], J. Phys. Conf. Ser. 5, 17 (2005) [arXiv:nuclex/0409023]. 81. P. F. Kolb, J. Sollfrank, and U. Heinz, Phys. Rev. C 65, 054909 (2000). 82. T. Hirano, private communication. 83. F. Karsch, E. Laermann, P. Petreczky, S. Stickan and I. Wetzorke, Phys. Lett. B 530, 147 (2002) [arXiv:heplat/0110208]. 84. M. Asakawa, T. Hatsuda and Y. Nakahara, Prog. Part. Nucl. Phys. 46, 459 (2001) [arXiv:heplat/0011040]. 85. W. Hiscock and L. Lindblom, Phys. Rev. D31, 725 (1985). 86. W. A. Hiscock and L. Lindblom, Annals Phys. 151, 466 (1983). 87. L. P. Kadanoff and P. C. Martin, Ann. Phys. 24, 419 (1963). 88. D. Forster, Hydrodynamics, Fluctuations, Broken Symmetry, and Correlation Functions (Perseus Books, 1990). 89. Lee Lindblom, arXiv:grqc:9508058. 90. W. Israel and J. M. Stewart, Annals Phys. 118, 341 (1979). 91. R. Geroch and L. Lindblom, Phys. Rev. D 41, 1855 (1990). 92. D. Pav´ on, D. Jou and J. CasasV´ asquez, Ann. Inst. Henri Poincar´e, Sect. A 36, 79 (1982). ¨ 93. H. C. Ottinger, Physica A 254 (1998) 433450. 94. R. Geroch and L. Lindblom, Annals. Phys. 207, 394 (19901. 95. W. Israel, Ann. Phys. 100 (1976) 310. 96. I. M¨ uller, Z. Phys. 198, 329 (1967). 97. M. Natsuume and T. Okamura, Phys. Rev. D 77, 066014 (2008) [Erratumibid. D 78, 089902 (2008)] [arXiv:0712.2916 [hepth]]. 98. A. Muronga and D. H. Rischke, arXiv:nuclth/0407114. 99. Peter Arnold, Guy D. Moore, and Laurence G. Yaffe, J. High Energy Phys. 11, 001 (2000). 100. S. R. De Groot, W. A. Van Leeuwen and C. G. Van Weert, Relativistic Kinetic Theory. Principles And Applications, Amsterdam, Netherlands: Northholland (1980). 101. F. Cooper and G. Frye, Phys. Rev. D 10, 186 (1974). 102. E. Braaten and M. H. Thoma, Phys. Rev. D 44, 2625 (1991); Phys. Rev. D 44, 1298 (1991). 103. Guy D. Moore, private communication. See also Ref. 26. 104. D. Molnar, private communication. 105. J. Vredevoogd and S. Pratt, arXiv:0810.4325 [nuclth]. 106. P. F. Kolb, U. W. Heinz, P. Huovinen, K. J. Eskola and K. Tuominen, Nucl. Phys. A 696, 197 (2001) [arXiv:hepph/0103234]. 107. Derek Teaney, J. Phys. G 30, S1247–S1250 (2004). 108. M. Martinez and M. Strickland, arXiv:0902.3834 [hepph]. 109. U. W. Heinz, K. S. Lee and M. J. RhoadesBrown, Phys. Rev. Lett. 58, 2292 (1987). 110. C. M. Hung and E. V. Shuryak, Phys. Rev. C 57, 1891 (1998) [arXiv:hepph/9709264]. 111. F. Reif, Fundamentals of Statistical Physics (McGrawHill, 1965). 112. A. El, A. Muronga, Z. Xu and C. Greiner, arXiv:0812.2762 [hepph].
266
D. A. Teaney
113. Z. Xu and C. Greiner, Phys. Rev. Lett. 100, 172301 (2008) [arXiv:0710.5719 [nuclth]]. 114. K. Dusling, private communication. 115. D. Teaney, and K. Dusling, in progress. 116. S. S. Adler et al. [PHENIX Collaboration], Phys. Rev. Lett. 97, 052301 (2006) [arXiv:nuclex/0507004]. 117. B. Alver et al. [PHOBOS Collaboration], arXiv:0812.1172 [nuclex]. 118. B. I. Abelev et al. [STAR Collaboration], Phys. Rev. Lett. 102, 052302 (2009) [arXiv:0805.0622 [nuclex]]. 119. B. I. Abelev et al. [STAR Collaboration], Phys. Rev. Lett. 98, 192301 (2007) [arXiv:nuclex/0607012]. 120. A. Adare et al. [PHENIX Collaboration], Phys. Rev. Lett. 98, 172301 (2007) [arXiv:nuclex/0611018].
HADRON CORRELATIONS IN JETS AND RIDGES THROUGH PARTON RECOMBINATION
RUDOLPH C. HWA Institute of Theoretical Science and Department of Physics University of Oregon, Eugene, OR 974035203, USA
[email protected] Hadron correlations in jets, ridges and opposite dijets at all pT above 2 GeV/c are discussed. Since abundant data are available from RHIC at intermediate p T , a reliable hadronization scheme at that pT range is necessary in order to relate the semihard partonic processes to the observables. The recombination model is therefore first reviewed for that purpose. Finalstate interaction is shown to be important for the Cronin effect, large B/M ratio and forward production. The effect of semihard partons on the medium is then discussed with particular emphasis on the formation of ridge with or without trigger. Azimuthal anisotropy can result from ridges without early thermalization. Dynamical path length distribution is derived for any centrality. Dihadron correlations in jets on the same or opposite side are shown to reveal detail properties of trigger and antitrigger biases with the inference that tangential jets dominate the dijets accessible to observation.
1. Introduction Among the many properties of the dense medium that have been studied at RHIC, the nature of jetmedium interaction has become the subject of particular current interest.1,2 Jet quenching, proposed as a means to reveal the effect of the hot medium produced in heavyion collisions on the hard parton traversing that medium,3,4 has been confirmed by experiments5,6 and has thereby been referred to as a piece of strong evidence for the medium being a deconfined plasma of quarks and gluons.7–10 By the time of Quark Matter 2006 the frontier topic has moved beyond the suppression of singleparticle distribution at high pT and into the correlation of hadrons on both the near side and the away side of jets.11,12 The data on dihadron and trihadron correlations are currently analyzed for low and intermediate pT , so the characteristic of hydrodynamical flow is involved in its interplay with semihard partons propagating through the medium. The physics issues are therefore broadened from the medium effects on jets to include also the effect of jets on medium. Theoretical studies of those problems can no longer be restricted to perturbative QCD that is reliable only at high pT or to hydrodynamics that is relevant only at low pT . In the absence of any theory based on first principles that is suitable for intermediate pT , phenomenological modeling is thus inevitable. A sample of some of the papers published before 2008 are given in Refs. 13–31. 267
268
R. C. Hwa
Hadron correlation at intermediate pT involves essentially every complication that can be listed in heavyion collisions. First, there is the characteristic of the medium created. Then there is the hard or semihard scattering that generates partons propagating through it. The interaction of those partons with the medium not only results in the degradation of the parton momentum, but also gives rise to ridges in association with trigger jets and to broad structure on the away side. Those features observed are in the correlations among hadrons, so hadronization is an unavoidable subprocess that stands between the partonic subprocess and the detected hadrons. Any realistic model must deal with all aspects of the various subprocesses involved. Without an accurate description of hadronization, observed data on hadron correlation cannot be reliably related to the partonic origin of ridges and jet structure. It is generally accepted that fragmentation is the hadronization subprocess at high pT , as in leptoninitiated processes. At intermediate and lower pT recombination or coalescence subprocess (ReCo) in heavyion collisions has been found to be more relevant.32–35 Despite differences in detail, the three formulations of ReCo are physically very similar. In Refs. 33 and 34 the descriptions are 3dimensional and treat recombination and fragmentation as independent additive components of hadronization. In Ref. 35 the formulation is 1D on the basis that acollinear partons have low probability of coalescence, and is simple enough to incorporate fragmentation as a component of recombination (of shower partons) so that there is a smooth transition from low to high pT . Since the discussions on jetmedium interaction in the main part of this review are based largely on the formalism developed in Ref. 35 that emphasizes the role of shower partons at intermediate pT , the background of the subject of recombination along that line is first summarized along with an outline of how nontrivial recombination functions are determined. Some questions raised by critics, concerning such topics as entropy and how partons are turned into constituent quarks, are addressed. More importantly, how shower partons are determined is discussed. Large baryontomeson ratio observed in heavyion collisions is a signature of ReCo, since the physical reason for it to be higher than in fragmentation is the same in all three formulations.32–34 The discussion here that follows the formulation of the recombination model (RM) by HwaYang should not be taken to imply less significance of the other two, but only the limits of the scope of this review. Considerable space is given to the topics of the Cronin effect (to correct a prevailing misconception) and to forward production at low and intermediate pT in Sec. 3. The large B/M ratio observed at forward production cements the validity of recombination so that one can move on to the main topic of jetmedium interaction. The two aspects of the jetmedium interaction, namely, the effect of jets on the medium and that of the medium on jets, are discussed in Secs. 4 and 6, respectively. In between those two sections we insert a section on azimuthal anisotropy because semihard jets can affect what is conventionally referred to as elliptic flow at low p T and also because ridge formation can depend on the trigger azimuth at intermediate
Hadron Correlations in Jets and Ridges Through Parton Recombination
269
pT . Much theoretical attention has been given in the past year to the phenomena of ridges on the near side and double hump on the away side of triggers at intermediate pT .36 – 50 Our aim here is not to review the various approaches of those studies, but to give an overview of what has been accomplished on these topics in the RM. The focus is necessary in order to cover a range of problems that depend on a reliable description of the hadronization subprocess. This review is complementary to the one given recently by Majumder,51 which emphasizes the region of pT much higher than what is considered here, so that factorized fragmentation can be applied. Due to space limitations this review cannot go into the mathematical details of either the basic formalism or the specific problems. Adequate referencing is provided to guide the interested reader to the original papers where details can be found. The discussions will mainly be qualitative, thus rendering an opportunity to describe the motivations, assumptions and physical ideas that underlie the model calculations. For example, the shower parton is an important ingredient in this approach that interpolates between what are soft (thermalthermal) and hard (showershower), but we have neither space nor inclination to revisit the precise scheme in which the showerparton distributions are derived from the fragmentation functions. The concept of thermalshower recombination and its application to intermediatepT physics are more important than the numerical details. Similarly, we emphasize the role that the ridges play (without triggers) in the inclusive distributions of single particles because of the pervasiveness of semihard scattering, the discussion of which can only be phenomenological. Attempts are made to distinguish our approach from conceptions and interpretations that are generally regarded as conventional wisdom. Some examples of what is conventional are: (1) Cronin effect is due to initialstate transverse broadening; (2) large B/M ratio is anomalous; (3) azimuthal anisotropy is due to asymmetric high pressure gradient at early time; (4) recombination implies quark number scaling (QNS) of v2 ; (5) dijets probe the medium interior. In each case evidences are given to support an alternative interpretation. In (4), it is the other way around: QNS confirms recombination but the breaking of QNS does not imply the failure of recombination. Other topics are more current, so no standard views have been developed yet. Indeed, there exist a wide variety of approaches to jetmedium interaction, and what is described here is only one among many possibilities. 2. Hadronization by Recombination 2.1. A historical perspective In the 70s when inclusive cross sections were beginning to be measured in hadronic processes the only theoretical scheme to treat hadronization was fragmentation for leptoninitiated processes for which the interaction of quark were known to be the basic subprocess responsible for multiparticle production. The same fragmentation process was applied also to the production of highpT particles in hadronic collisions.52 Local partonhadron duality was also invoked as a way to avoid focusing on
270
R. C. Hwa
the issue of hadronization.53 In dual parton model where color strings are stretched between quarks and diquarks, the fragmentation functions (FFs) are attached to the ends of the strings to materialize the partons to hadrons, even if one of the ends is a diquark.54 However, at low pT in pp collisions quarks are not isolated objects in the parton model since there are gluons and weepartons at small x,55 so the justification for the confinement of color flux to a narrow string is less cogent than at high pT . A more physically realistic description of hadronization seemed wanting. The first serious alternative to fragmentation against the prevalent scheme for hadronization was the suggestion that pion production at low pT in pp collisions can be treated by recombination.56 The simple equation that describes it is Z dx1 dx2 dN π = Fqq¯(x1 , x2 )Rπ (x1 , x2 , x) , (1) x dx x1 x2 where Fqq¯(x1 , x2 ) is the q q¯ distribution, taken to be the product Fq (x1 )Fq¯(x2 ) of the q and q¯ distributions already known at the time among the parton distributions of a proton. The recombination function (RF) Rπ (x1 , x2 , x) contains the momentum conserving δ(x1 + x2 − x) with a multiplicative factor that is constrained by the counting rule developed for quarks in hadrons. That simple treatment of hadronization turned out to produce results that agreed with the existing data very well. The next important step in solidifying the treatment of recombination is the detailed study of the RF. If RF is circumscribed by the characteristics of the wave function of the hadron formed, then it should be related to the timereversed process of describing the structure of that hadron. In dealing with that relationship it also becomes clear that the distinction between partons and constituent quarks must be recognized and then bridged — a problem that has puzzled some users of the RM even in recent years. Since hadron structure is the basis for RF, it became essential to have a description of the constituents of a hadron in a way that interpolates between the hadronic scale and the partonic scale. It is in the context of filling that need that the concept of valons was proposed.57 The origin of the notion of constituent quarks (CQs) is rooted in solving the boundstate problem of hadrons. However, in describing the structure of a nucleon in deep inelastic scattering the role of CQs seems to be totally absent in the structure functions F, such as νW2 (x, Q2 ). The two descriptions are not merely due to the difference in reference frames, CQs being in the rest frame, the partons in a highmomentum frame. Also important is that the bound state is a problem at the hadronic scale, i.e. low Q2 , while deep inelastic scattering is at high Q2 . The two aspects of the problem can be connected by the introduction of valons as the dressed valence quarks, i.e., each being a valence quark with its cloud of gluons and sea quarks which can be resolved only by highQ2 probes. At low Q2 the internal structure of a valon cannot be resolved, so a valon becomes what a CQ would be in the momentumfraction variable in an infinitemomentum frame. Thus the valon distribution in a hadron is the wavefunction squared of the CQs, whose structure
Hadron Correlations in Jets and Ridges Through Parton Recombination
271
functions are described by pQCD at high Q2 . Note that the usual description of Q2 evolution by DGLAP has no prescription within the theory for the boundary condition at low Q2 . That distribution at low Q2 is precisely what the valon distribution specifies. In summary, the structure function F h (x, Q2 ) of a hadron is a convolution of the valon distribution Gv/h (y) and the structure function F v (z, Q2 ) of a valon58 XZ 1 dyGv/h (y)F v (x/y, Q2 ) , (2) F h (x, Q2 ) = ν
x
where y is the momentum fraction (not rapidity) of a valon in the hadron h. The first description of the properties of Gv/h (y) is given in Refs. 57 and 58, derived from the early data F h (x, Q2 ). More recent determination of Gv/h (y) is described in Ref. 59 where more modern parton distribution functions have been used.60 Gv/h (y) is the singlevalon inclusive distribution in hadron h, and is the appropriate integral of the exclusive distribution, Gv/π (y1 , y2 ) for pion and Gv/p (y1 , y2 , y3 ) for proton. More specifically, Gv/π (y1 , y2 ) is the absolute square of the pion wave function hv1 (y1 )v2 (y2 )πi in the infinitemomentum frame. Once we have that, it is trivial to get the RF for pion (i.e., by complex conjugation), since it is the timereversed process. Thus for pion and proton, we have x x x1 x2 2 1 , , (3) Rπ (x1 , x2 , x) = 2 Gv/π x x x x x x x1 x2 x3 2 3 1 Rp (x1 , x2 , x3 , x) = G , , , (4) v/p x3 x x x
where the factors on the RHS are due to the fact that the RFs are invariant distributions defined in the phase space element Πi dxi /xi , whereas Gv/h are noninvariant defined in Πi dyi , as seen in (2). The exclusive distribution Gv/h contains the moP mentum conserving δ( i yi −1). For pion there is nothing else, but for other hadrons the prefactors are given in Refs. 59 and 61. Having determined the RF, the natural question next is how partons turn into valons before recombination in a scattering process. Let us suppose that we can calculate the multiparton distribution F (x1 , x2 ) for a q and q¯ moving in the same direction, whether at low or high pT . If their momentum vectors are not parallel, with relative transverse momentum larger than the inverse hadronic size, then the probability of recombination is negligible. Relative longitudinal momentum need not be small, since the RF allows for the variation in the momentum fractions, just as the partons in a hadron can have various momentum fractions. Now, as the q and q¯ move out of the interaction region, they may undergo color mutation by soft gluon radiation as well as dress themselves with gluon emission and reabsorption with the possibility of creating virtual q q¯ pairs, none of which can be made precise without a high Q2 probe. The net effect is that given enough time before hadronization the quarks convert themselves to valons with essentially the same original momenta, assuming that the energy loss in vacuum due to soft gluon radiation is negligible
272
R. C. Hwa
(even though color mutation is not negligible). For that reason we may simply write F (x1 , x2 )R(x1 , x2 , x) as multiplicative factors, as done in (1), while treating F (x1 , x2 ) as the distribution of partons and R(x1 , x2 , x) as the RF of valons. The detail of this is explained in Ref. 58. The question of entropy conservation has been raised at times, especially by those with experience in nuclear physics. In elementary processes, such as q+ q¯ → π, unlike a nuclear process p + n → d, the color degree of freedom is important. Since a pion is colorless, the q and q¯ that recombine must have opposite color. If they do not, they cannot travel in vacuum without dragging a color flux tube behind them. The most energyefficient way for them to evolve is to emit soft gluons thereby mutating their color charges until the q q¯ pair becomes colorless and recombine. Such soft processes leave behind color degrees of freedom from the q q¯ system whose entropy is consequently not conserved. It is therefore pointless to pursue the question of entropy conservation in recombination, since the problem is uncalculable and puts no constraint on the kinematics of the formation of hadrons. Besides, the entropy principle should not be applied locally. A global consideration must recognize that the bulk volume is increasing during the hadronization process, and thus this compensates any decrease of local entropy density. After the extensive discussion given above on the RF, we have come to the point of being able to assert that the main issue about recombination is the determination of the multiparton distribution, such as Fqq¯(x1 , x2 ) in (1), of the quarks that recombine. Related to that is the question about the role of gluons which have to hadronize also. By moving the focus to the distributions of partons that hadronize, the investigation can then concentrate on the more relevant issues in heavyion collisions concerning the effect of the nuclear medium. 2.2. Shower partons At low pT in the forward direction the partons that recombine are closely related to the lowQ2 partons in the projectile. It is a subject to be discussed in a following section. At intermediate and high pT the partons are divided into two types: thermal (T) and shower (S). The former contains the medium effect; the latter is due to semihard and hard scattered partons. The consideration of shower partons is a unique feature of our approach to recombination, which is empowered by the possibility to include fragmentation process as SS or SSS recombination. The jetmedium interaction is taken into account at the hadronization stage by TS recombination, although at an earlier stage the energy loss of the partons before emerging from the medium is another effect of the interaction that is, of course, also important. A quantitative theoretical study of that energy loss in realistic heavyion collisions at fixed centrality cannot be carried out and compared with data without a reliable description of hadronization. At intermediate pT there is no evidence that fragmentation is applicable because the baryon/meson ratio would be too small, as we shall describe in the next section.
Hadron Correlations in Jets and Ridges Through Parton Recombination
273
The fragmentation function (FF), D(x), is a phenomenological quantity whose Q2 evolution is calculable in pQCD; however, at some low Q2 before evolution the distribution in x is parametrized by fitting the data. With that reality in mind it is reasonable to consider an alternative way of treating the FF, one that builds in more dynamical content by regarding fragmentation as a recombination process. That is, if we replace the LHS of Eq. (1) by the invariant function xDiπ (x), then the corresponding twoparton distribution in the integrand on the RHS is the product distributions of two shower partons in a jet initiated by a parton of type i. To be specific, consider, for example, the fragmentation of gluon to pion Z x2 dx1 dx2 q q¯ π Rqπq¯(x1 , x2 , x) , (5) S (x1 )Sg xDg (x) = x1 x2 g 1 − x1 where the q and q¯ distributions in a shower initiated by the gluon are the same, but their momentumfraction dependencies are such that if one (x1 ) is a leading quark, the other (x2 ) has to be from the remainder (1 − x1 ) of the parton pool. With xDgπ (x) being a phenomenological input, it is possible to solve (5) numerically to obtain Sgq (z). It has been shown in Ref. 62 that there are enough FFs known from analyzing leptonic processes to render feasible the determination of various shower parton distributions (SPDs), which are denoted collectively by Sij with i = q, q¯, g and j = q, s, q¯, s¯, where q can be either u or d. If in i the initiating hard parton is an s quark, it is treated as q. That is not the case if s is in the produced shower. The parameterization of Sij has the form Sij (z) = Az a (1 − z)b (1 + cz d ) ,
(6)
where the dependence of A, a, etc., on i and j are given in a Table in Ref. 62. Those parameters were determined from fitting the FFs at Q = 10 GeV, and have been used for all hadronization processes without further consideration of their dependence on pT . It should be recognized that those shower partons are not to be identified with the ones due to gluon radiation at very high virtuality calculable in pQCD, which is not applicable for the description of hadronization at low virtuality. To sum up, in the conventional approach the FF is treated as a black box with a parton going in and a hadron going out, whereas in the RM we open up the black box and treat the outgoing hadron as the product of recombination of shower partons, whose distributions are to be determined from the FFs. Once the SPDs are known, one can then consider the possibility that a shower parton may recombine with a thermal parton in the vicinity of a jet and thus give a more complete description of hadronization at intermediatepT region, especially in the case of nuclear collisions. The SPDs parametrized by (6), being derived from the meson FFs, open up the question of what happens if, instead of j ¯j 0 → M , three quarks in the shower recombine, e.g., uud → p, or uds → Λ. A selfconsistent scheme of hadronization would have to demand that the formation of baryons is a possibility in a fragmentation process and that the SPDs already determined should give an unambiguous prediction of what baryon FFs are. The calculation has been carried out in Ref. 63
274
R. C. Hwa
where the results for g → p and g → Λ in gluon jets are in good agreement with data64 without the use of any adjustable parameters. To be able to relate meson and baryon FFs is an attribute of our formalism for hadronization that has not been achieved in other theoretical approaches, and provides further evidence that the SPDs are reliable for use at the hadronization scale. 2.3. Parton distributions before recombination In the study of shower partons in a jet we have assumed the validity of the approximation that the fragmentation process is essentially one dimensional. One may question whether the recombination process in a nuclear collision for a hadron produced at high pT may necessitate a 3D consideration, since two different length scales seem to be involved, one being that of the hadron produced, the other being the size of nuclear medium at hadronization time. Indeed, recombination schemes formulated in 3D have been proposed, and various groups have independently found satisfactory results that are similar to one another.32–34 The essence of recombination is, however, not in the 2D transverse plane normal to the direction of hadron momentum because if the coalescing parton momenta are not roughly parallel, then the relative momentum would have a large component in that transverse plane. If that component is larger than the inverse of the hadron size, then the two (or three) partons cannot recombine. Thus partons from regions of the nuclear medium that are far apart cannot form a hadron, rendering the concern over different length scales in the problem inessential. Only collinear partons emanating from the same region of the dense medium can recombine. For that reason the 1D formulation of recombination is adequate, as simple as expressed in (1). If one asks why the relative momentum can be large in the hadron direction, but not transverse to it, the answer lies in the foundation of the parton model where the momentum fraction can vary from 0 to 1, while the transverse parton momentum kT is limited to ∼ (hadron radius)−1 . The RFs in (3) and (4) are related to the 1D wave function in that framework. Having justified the 1D formulation of recombination, let us now focus on the distributions of the recombining partons at low pT , and later at high pT . Since pQCD cannot be applied to multiparticle production at low pT , our consideration of the problem is based on Feynman’s parton model, which was originally proposed for hadron production at low pT .55 In a pp collision there are valence and sea quarks and gluons whose xdistributions at low Q2 are known.60 Without hard scattering their momenta carry them forward, and they must hadronize in the fragmentation region of the initial proton. RM has provided a quantitative treatment of the singlehadron inclusive distribution in xF not only for pp, but also for all realistic hadronic collisions.56,58,61 What is to be remarked here is how gluons hadronize. In an incident proton as in other hadrons, the gluons carry about half the momentum of the host. Since gluons cannot hadronize by themselves, but can virtually turn to q q¯ pairs in the sea, we require that all gluons be converted to the sea quarks (thus
Hadron Correlations in Jets and Ridges Through Parton Recombination
275
saturating the sea) before recombination. This idea was originally suggested by Duke and Taylor,65 and was implemented quantitatively in the valon model in Refs. 58 and 61. The simplest way to achieve that is to increase the normalization of the q q¯ sea quarks without changing their x distributions so that the total momentum of the valence (unchanged) and sea quarks (enhanced) exhausts the initial momentum of the hadron without any left over for gluons. With the Fqq¯(x1 , x2 ) thus obtained, the use of (1) results in an inclusive π distribution that agrees with data in both normalization and x spectrum.58,61 Using the appropriate valon distributions of pion and kaon, the success extends beyond p → π ± to π + → π − , K + → π ± , and π + → K ± in hadronic collisions at low pT . In nuclear collisions there is the additional complication arising from momentum degradation when partons traverse nuclear medium. It is a subject that will be brought up in Sec. 3.4. When pT is not small, then there has to be a semihard or hard scattering at the partonic level so that a parton with kT > 3 GeV/c has to be created. In that case shower partons are developed in addition to the thermal partons, so the partons before recombination can be separated into the following types: TT+TS+SS for mesons and TTT+TTS+TSS+SSS for baryons.35 The thermal partons have kT mainly < 2 GeV/c. If one has a reliable scheme to calculate the thermal partons, then their kT distributions can, of course, be used in the recombination equation. It does not mean that hydrodynamics is a necessary input in the RM. In pA collisions, for instance, hydrodynamics is not reliable, yet the Cronin effect can be understood in the RM for both proton and pion production without associating the effect with initialstate scattering — a departure from the conventional thinking that will be discussed in the next section. In most applications reviewed here, the distributions of thermal partons are determined from fitting the data at low pT , and are then used in the RM to describe the behavior of hadrons at pT & 3 GeV/c. When we consider correlation at a later section, careful attention will be given to the enhancement of thermal partons due to the energy loss of a semihard or hard parton passing through the nuclear medium. It is only in the framework of a reliable hadronization scheme can one learn from the detected hadrons the nature of jetmedium interaction, as aspired in jet tomography. 3. Large Baryon/Meson Ratios 3.1. Intermediate pT in heavyion collisions A wellknown signature of the RM is that the baryon/meson (B/M) ratio is large — larger than what is customarily expected in fragmentation. The p/π ratio of the FFs, i.e., Dp/q (x)/Dπ/q (x), is at most 0.2 at x ' 0.3, and is much lower at other values of x.64 However, for inclusive distributions in heavyion collisions at RHIC the ratio Rp/π is as large as ∼ 1 at pT ' 3 GeV/c,66,67 as shown in Fig. 1. Thus hadronization at intermediate pT cannot be due to parton fragmentation. Three groups (TAM, Duke and Oregon) have studied the problem in the Recombination/Coalescence (ReCo) model32–35 and found large Rp/π in agreement with
276
R. C. Hwa
Fig. 1. Comparison of baryon/meson ratio from STAR data to theoretical curves from ReCo. 33–35 Left panel is from Ref. 66; right panel is from Ref. 67.
the data. The underlying reason that is common in all versions of ReCo is that for p and π at the same pT the three quarks that form the p has average momentum pT /3, while the q and q¯ that form the π has pT /2. Since parton distributions are suppressed severely at increasing kT , there are more quarks at pT /3 than at pT /2, so the formation of proton is not at a disadvantage compared to that of a pion despite the difference in the RFs. For either hadron the recombination process is at an advantage over fragmentation because of the addivity of momenta. Fragmentation suffers from two penalties: first, the initiating parton must have a momentum higher than pT , and second, the FFs are suppressed at any momentum fraction, more for proton than for pion. Thus the yield from parton fragmentation is lower compared to that from parton recombination at intermediate pT , even apart from the issue of B/M ratio. When faced with the question why baryon production is so efficient, the proponents of pion fragmentation regard it as an anomaly. Despite efforts to explain the enhancement in terms of baryon junction,68,69 the program has not been successful in establishing it as a viable mechanism for the formation of baryons.70 From the point of view of ReCo there is nothing anomalous. A simple way to understand the pT dependence of Rp/π (pT ) is to consider the 1D formulation of ReCo given in Ref. 35, where the invariant distributions of meson and baryon production are expressed as ! Z Y 2 M dqi 0 dN p = Fqq¯(q1 , q2 )RM (q1 , q2 , p) , (7) dp q i i=1 p
0 dN
B
dp
=
Z
3 Y dqi q i=1 i
!
F3q (q1 , q2 , q3 )RB (q1 , q2 , q3 , p) ,
(8)
in which all quarks are collinear with the hadron momentum p. We assume that the rapidity y is ≈ 0, so the transverse momenta are the only essential variables, for which the subscripts T of all momenta are therefore omitted, for brevity. Mass
Hadron Correlations in Jets and Ridges Through Parton Recombination
277
effect at low pT renders the approximation poor and the 1D description inadequate. However, in order to gain a transparent picture analytically, let us ignore those complications and assume provisionally that all hadrons are massless. Then the experimental observation of exponential behavior of the pT distribution of pions at low pT , i.e., dN π /pdp ∝ exp(−p/T ), implies that the thermal partons behave as T (q) = q
dN th = Cqe−q/T , dq
(9)
where C has dimension (GeV)−1 , and Rπ given in (3) is dimensionless. When thermal partons dominate Fqq¯ and F3q , the multiparton distributions can be written as products: T (q1 )T (q2 ) and T (q1 )T (q2 )T (q3 ), respectively. It is then clear from the dimensionlessness of the quantities in (7) and (8) that with the proton distribution having C 3 dependence, as opposed to the pion distribution being ∝ C 2 , the p/π ratio has the property Rp/π (p) =
dN p /pdp ∝ Cp , dN π /pdp
(10)
so long as thermal recombination dominates. This linear rise with p is the behavior seen in Fig. 1, although the mass effect of proton makes it less trivial in pT . Nevertheless, this simple feature is embodied in the more detailed computation until shower partons become important for pT > 3 GeV/c.35 From the above analysis which should apply to any baryon and meson, it follows that the ratios Λ/K and Ω/φ should also increase with pT in a way similar to p/π. Such behaviors have indeed been observed by STAR,71,72 as have been obtained in theoretical calculation.73 Taken altogether, it means that without TS and TTS recombination the B/M ratios would continue to rise with pT . But the data all show that the ratios peak at around pT ≈ 3 GeV/c. In the RM the bendover is due to the increase of the TS component of the meson earlier than the TTS component of the baryon, since two thermal partons in the latter have more weight than the single thermal parton in the former. The shower parton distribution S(q) in heavyion collisions is a convolution of the hard parton distribution fi (k) and the S distribution derived from FF, discussed in Sec. 2.2, i.e., XZ S(q) = ζ dkkfi (k)Si (q/k). (11) i
fi (k) is the transversemomentum distribution of hard parton i at midrapidity and contains the shadowing effect of the parton distribution in nuclear collisions. A simple parametrization of it is given in Ref. 74 as follows fi (k) = K
A , (1 + k/B)n
(12)
where K = 2.5 and A, B, n are tabulated for each parton type i for nuclear collisions at RHIC and LHC. The parameter ζ = 0.07 is the average suppression factor that
278
R. C. Hwa
can be related to the nuclear modification factor RAA , and was denoted by ξ in Ref. 35 and other references thereafter. Since fi (k) has a powerlaw dependence on k, so does S(q2 ) on q2 in contrast to the exponential behavior of the thermal partons, T (q1 ). This upward bending of S(q2 ) relative to T (q1 ) is the beginning of the dominance of TS and TTS components over TT and TTT components, resulting in a peak in the B/M ratio at around pT ∼ 3 GeV/c. Detailed descriptions of these calculations are given in Refs. 35 and 73. We add here that the effort made to consider the shower partons before recombination is motivated by our concern that a hard parton with high virtuality cannot hadronize by coalescing with a soft parton with low virtuality. The introduction of shower partons is our way to bring the effects of hard scattering to the hadronization scale. At the same time the formalism does not exclude fragmentation by a hard parton, since SS and SSS recombination at high pT are equivalent to fragmentation but in a language that has dynamical content at the hadronization scale. One could ask how the RM can be applied reliably in the intermediatepT region before the shower partons were introduced. The approach adopted in Ref. 32 does not involve the determination of the hard parton distribution by perturbative calculation, but uses the pion data as input to extract the parton distribution at the hadronization scale at all pT in the framework of the RM. It is on the basis of the extracted parton distribution (which must in hindsight contain the shower partons) that the proton inclusive distribution is calculated. Thus the procedure is selfconsistent. The result is that the p/π ratio is large at pT ∼ 3 GeV/c in agreement with data; furthermore, it was a prediction that the ratio would decrease as pT increases beyond 3 GeV/c, as confirmed later by data shown in Fig. 1. 3.2. Cronin effect The conventional explanation of the Cronin effect,75 i.e., the enhancement of hadron spectra at intermediate pT in pA collisions with increasing nuclear size, is that it is due to multiple scattering of projectile partons as they propagate through the target nucleus, thus acquiring transverse momenta, and that a moderately largek T parton hadronizes by fragmentation.76 The emphasis has been on the transverse broadening of the parton in the initialstate interaction (ISI) and not on the finalstate interaction (FSI). In fact, the Cronin effect has become synonymous to ISI effect in certain circles. However, that line of interpretation ignores another part of the original discovery75 where the A dependence of hadrons produced in pA collisions, when parameterized as dN (pA → hX) ∝ Aαh (pT ) , dpT
(13)
has the property that αp (pT ) > απ (pT ) for all pT measured. That experimental result alone is sufficient to invalidate the application of fragmentation to the hadronization process, since if the A dependence in (13) arises mainly from the ISI,
279
π
Hadron Correlations in Jets and Ridges Through Parton Recombination
Fig. 2. Centraltoperipheral ratios for the production of pion (left panel) and proton (right panel) in dAu collisions. Data are from Ref. 77 and lines are from Refs. 78 and 79.
where the multiplyscattered parton picks up it kT , then the transverse broadening of that parton should have no knowledge of whether the parton would hadronize into a proton or a pion, so αh should be independent of the hadron type h. A modern version of the Cronin effect is given in terms of the centraltoperipheral nuclear modification factor for dAu collisions at midrapidity C 1/Ncoll dN h /pT dpT (C) h RCP , (14) (pT ) = P 1/Ncoll dN h /pT dpT (P )
where C and P denote central and peripheral, respectively, and Ncoll is the average number of inelastic N N collisions. If hadronization is by fragmentation, which is a factorizable subprocess, the FFs for any given h should cancel in the ratio of p h (14), so RCP should be independent of h. However, the data show that RCP (pT ) > π RCP (pT ) for all pT > 1 GeV/c when C = 0–20% and P = 60–90% centralities.77 See Fig. 2. Clearly ISI is not able to explain this phenomenon, which strongly suggests the mediumdependence of hadronization. The data further indicate that the pT h dependence of RCP (pT ) peaks at pT ∼ 3 GeV/c for both p and π, reminiscent of the p/π ratio at fixed centrality in AuAu collisions although the C/P ratio for dAu collisions is distinctly different. Hadron production at intermediate pT and η ∼ 0 in dAu collisions can be treated in the RM in a similar way as for AuAu collisions. Although no hot and dense medium is produced in a dAu collision, so thermal partons are not generated in the same sense as in AuAu collisions, nevertheless soft partons are present to give rise to the lowpT hadrons. For notational uniformity we continue to refer to them as thermal partons. We apply the same formalism developed in Ref. 35 to the dAu problem and consider the TT+TS+SS contributions to π production (TTT+TTS+TSS+SSS for p). The thermal T distribution is determined by fitting the pT spectra at pT < 1 GeV/c for each centrality; the showerparton S distribution is calculated as before but without nuclear suppression. Unlike the dense
280
R. C. Hwa
thermal system created in AuAu collisions, the T distribution in this case is weaker; its parameters C and T (inverse slope) that correspond to the ones in Eq. (9) are smaller. Furthermore, C decreases with increasing peripherality, while T remains unchanged at 0.21 GeV/c.78 Thus thermalshower recombination becomes important at pT & 1 GeV/c, which is earlier than in AuAu collisions. As a consequence, π RCP (pT ) becomes > 1 at pT > 1 GeV/c. That is the Cronin effect, but not due to ISI. The same situation occurs for proton production, only stronger.79 The calculated results for the inclusive distributions of both π and p agree well with data p π h at all centralities, hence also RCP (pT ) and RCP (pT ). Figure 2 shows RCP (pT ) for 77 C = 0–20% and P = 60–90% in dAu collisions for h = π and p; the lines are the p π results obtained in the RM.78,79 The reason for RCP > RCP can again be traced to 3quark recombination for p and only 2 quarks for π. When pT is large, fragmenh tation dominates (i.e. SS and SSS), and both RCP approach 1, since FFs cancel and the yields are normalized by Ncoll . No exotic mechanism need be invoked to explain the p production process. FSI alone is sufficient to provide the underlying physics for the Cronin effect. 3.3. Forward production in dAu collisions Hadron production at forward rapidities in dAu collisions was regarded as a fertile ground for exposing the physics of ISI, especially saturation physics,80,81 since the nuclear effect in the deuteron fragmentation region was thought to cause minimal FSI. It was further thought that the difference in nuclear media for the Au side (η < 0) and the d side (η > 0) would lead to backwardforward asymmetry in particle yield in such a way as to reveal a transition in basic physics from multiple scattering in ISI for η . 0 to gluon saturation for η > 0. The observation by BRAHMS82 that RCP decreases with increasing η was regarded as an indication supporting that view.83 That line of thinking, however, assumes that FSI is invariant under changes in η so that any dependence on η observed is a direct signal from ISI. Such an assumption is inconsistent with the result of a study of forward production in dAu collisions in the RM, where both RCP (pT , η) and RB/F (pT ) are shown to be well reproduced by considering FSI only.84 Any inference on ISI from the data must first perform a subtraction of the effect of FSI, and just as in the case of the Cronin effect there is essentially nothing left after the subtraction. In Sec. 3.2 the Cronin effect at midrapidity (η ≈ 0) is considered. The extension to η > 0 along the same line involves no new physics. However, it is necessary to determine the η dependencies of the soft and hard parton spectra at various centralities. For the soft partons, use is made of the data on dN/dη to modify the normalization of T (q, η) already determined at η = 0. For the hard partons, modified parametrizations of their distributions fi (kT , η) are obtained from leading order minijet calculations using the CTEQ5 pdf 85 and the EKS98 shadowing.86 A notable feature of the result is that fi (kT , η) falls rapidly with kT as η increases, especially near the kinematical boundary kT = 8.13 GeV/c and η = 3.2. Thus TS
Hadron Correlations in Jets and Ridges Through Parton Recombination
(a)
281
(b)
Fig. 3. (a) RCP (pT , η) for forward production. Data are from Ref. 82 and lines from Ref. 84. (b) Back/Forward ratio RB/F (pT ) for 0.5 < η < 1.0 that shows agreement of theoretical result84 with preliminary data87 for pT < 2 GeV/c but not for higher pT , but later data88 (inset) show agreement for pT > 2 GeV/c also.
and SS components are negligible compared to TT at large η for any pT and any centrality, even though the TT component is exponentially suppressed. In central collisions there is the additional suppression due to momentum degraduation of the forward partons going through the nuclear medium of the target Au. Putting the various features together leads to the ratio RCP (pT , η) shown in Fig. 3(a), where the data are from Ref. 82 and the curves from the calculation in Ref. 84. It is evident that the decrease of RCP (pT , η) at pT > 2 GeV/c as η increased from η = 0 to η = 3.2 is well reproduced in the RM. Only one new parameter is introduced to describe the centrality and η dependence of the inverse slope T of the soft partons, but no new physics has been added. The suppression of RCP (pT , η) at η > 1 is due mainly to the reduction of the density of soft partons in the forward direction, where hard partons are suppressed. Extending the consideration to the backward region and using the same T (η) extrapolated to η < 0, the backward/forward ratio of the yield can be calculated. 84 For η = ±0.75 corresponding to the data of STAR at 0.5 < η < 1.0 and 0–20% centrality,87 the calculated result on RB/F for π + + π − + p + p¯ is shown by the solid line in Fig. 3(b). While it agrees with the data very well for pT < 2 GeV/c, it is noticably lower than the data for all charged particles for pT > 2 GeV/c. However, more recent data on RB/F (pT ) for π + +π − +p+ p¯,88 shown in the inset of Fig. 3(b), exhibit excellent agreement with the same theoretical curve that should be regarded as a prediction. The fact that RB/F is > 1 for all pT measured may be regarded as a proof against initial transverse broadening of partons, since forward partons of d have more nuclear matter of Au to go through than the backward partons of Au. Thus if ISI is responsible for the acquisition of pT of the finalstate hadrons, then RB/F should be < 1. The data clearly indicate otherwise.
282
R. C. Hwa
3.4. Forward production in AuAu collisions Theoretical study of hadron production in the forward direction in heavyion collision is a difficult problem for several reasons. The parton momentum distribution at low Q2 and large momentum fraction x in nuclear collisions is hard to determine, especially when momentum degradation that accounts for what is called “baryon stopping” cannot be ignored. Furthermore, degradation of highmomentum partons in the nuclear medium implies the regeneration of soft partons at lower x; that is hard to treat also. The use of data as input to constrain unknown parameters is unavoidable; however, existent data have their own limitations. Measurement at fixed η cannot be used to provide information on xF dependence unless pT is known.89 Measurement of both η and pT has been limited to charged hadrons90 that cannot easily be separated into baryons and mesons. For these various reasons forward production in AA collisions has not been an active area of theoretical investigation. However, there are gross features at large η that suggest important physics at play and deserve explanation. PHOBOS data show that particles are detected at η 0 > 0 where η 0 is the shifted pseudorapidity defined by η 0 = η − ybeam .89 It is significant because it suggests that 0 if hpT i is not too small, it corresponds to xF > 1, where xF = (pT /mp )eη . Instead of violation of momentum conservation, the interpretations in the RM is that a proton can be produced in the xF > 1 region, if three quarks from three different nucleons in the projectile nucleus, each with xi < 1, recombine to form a nucleon with xF = P 91 That kinematical region is referred to as transfragmentation region i xi > 1. (TFR), which is not accessible, if hadronization is by fragmentation. The theoretical calculation in the RM involves an unknown parameter, κ, which quantifies the degree of momentum degradation of lowkT partons, in the forward direction. For κ in a reasonable range, not only can nucleons be produced continuously across the xF boundary, but also can p/π ratio attain an amazingly large value.91 BRAHMS has determined the pT distribution of all charged particles at η = 3.2.90 For hpT i = 1 GeV/c, the corresponding values of xF for pion and proton are, respectively, 0.4 and 0.54. Taking the preliminary value of the p¯/p ratio at 0.05 into account, it is possible to estimate the value of κ and then calculate the pT distribution of p + p¯ + π + + π − .92 The p/π ratio was predicted to be ∼ 1 at pT ∼ 1 GeV/c. However, at QM2008 the more recent data on Rp/p was reported to have a ¯ 93 lower value at 0.02 and on Rp/π a higher value at ∼ 4 at pT ∼ 1 GeV/c.94 Those new data prompted a reexamination of the problem in the RM; with appropriate changes in the treatment of degradation, regeneration and transverse momentum, the very large p/π ratio can be understood.95 Since p, p¯ and π production at large η depends sensitively on q and q¯ distributions, which in turn depend strongly on the dynamical process of momentum degradation and softparton regeneration (the parameterization of which requires phenomenological inputs), the procedure in Ref. 95 is to use Rp/π and Rp/p as input ¯ in order to determine κ and then calculate the x distributions of the hadrons. At
Hadron Correlations in Jets and Ridges Through Parton Recombination
(a)
283
(b)
Fig. 4. (a) Proton/pion ratio in forward production at η = 3.2 showing agreement between data 94 and solid line from the RM;95 the dashed line is the contribution from the longitudinal components at fixed η. (b) Comparison of the pT distribution of charged particles at η = 3.2 from BRAHMS90 with calculated result from the RM.95
fixed η the x and pT distributions are related. It turns out that the result on the x distribution leads to a large contribution to the pT distribution of Rp/π (pT ) shown by the dashed line in Fig. 4(a). The additional enhancement shown by the solid line arises from the mass dependence of the inverse slopes Th due to flow. While the ratio Rp/π is insensitive to the absolute normalizations of the yields, the inclusive distribution of all charged particles is not. In Fig. 4(b) is shown the good agreement between the calculated result and the data in both normalization and shape with no extra parameters beyond κ already fixed. It should be noted that the p/π ratio, shown in Fig. 4(a), is extremely large at η = 3.2 and modest pT < 2 GeV/c. The underlying physics is clearly the suppression of q¯ at medium x and the enhancement of p due to 3q recombination, where the (valence) quarks are from three different nucleons in the projectile. No other hadronization mechanisms are known to be able to reproduce the data on the large Rp/π at large η.
3.5. Recombination of adjacent jets at LHC So far we have considered only the physics at RHIC energies and the recombination of thermal and shower partons, either between them or among themselves. At RHIC highpT jets are rare, so the shower partons are from one jet at most in an event. At LHC, however, highpT jets are copiously produced for pT < 20 GeV/c. When the jet density is high, the recombination of shower partons in neighboring jets becomes more probable and can make a significant contribution to the spectra of hadrons in the 10 < pT < 20 GeV/c range, high by RHIC standard, but intermediate at LHC. If that turns out to be true, then a remarkable signature is predicted and is easily measurable: the p/π ratio will be huge, perhaps as high as 20.96
284
R. C. Hwa
If a hard parton of momentum kT is produced, shower partons in its jet with P momenta qi are limited by the constraint i qi < kT , so that the recombination of those shower partons can produce a hadron with momentum pT not exceeding kT . However, if there are two adjacent jets with hardparton momenta k1T and k2T , then to form a hadron at pT from shower partons in those two jets, neither k1T nor k2T need to be larger than pT , so the rate of such a process would be higher. Furthermore, to form a proton at pT the shower parton qi can be lower than those for pion formation at the same pT , so kiT can be even lower. Thus Rp/π in 2jet recombination can be much higher than the ratio in 1jet fragmentation. The probability for 2jet recombination, however, also depends on the overlap of jet cones, since the coalescing shower partons must be nearly collinear. That overlap decreases with increasing kiT , so there is a suppression factor in the SS or SSS recombination integral that depends on the widths of the jet cones. Using some reasonable estimates on all the factors involved, it is found that Rp/π can be between 5 and 20 in the range 10 < pT < 20 GeV/c, decreasing with increasing pT .96 Although exact numbers are unreliable, the approximate value of Rp/π is about 2 orders of magnitude higher than what is expected in the usual scenario of fragmentation from single hard partons. The origin of the large Rp/π at LHC discussed above is basically the same as that for forward production in AuAu collisions at RHIC. In both cases it is the multisource supply of the recombining partons that enhances the proton production. At large pT at LHC there are more than one jet going in the same direction; at large pL at RHIC there are more than one nucleon going in the forward direction. In the latter case we already have data supporting our view that Rp/π should be large as shown in Fig. 3(a). It would be surprising that our prediction of large Rp/π at LHC turns out to be untrue. 4. Ridgeology
Phenomenology of Ridges
In the previous section the topics of discussion have been exclusively on the singleparticle distributions in various regions of phase space. Everywhere it is found that the B/M ratio is large when pT is in the intermediate range. We now consider twoparticle correlations, on which there is a wealth of data as a result of the general consensus in both the experimental and theoretical communities that more can be learned about the dense medium when one studies the system’s effect on (and response to) penetrating probes. The strong interaction between energetic partons and the medium they traverse, resulting in jet quenching, is the underlying physics that can be revealed in the jet tomography program.4,97 To calibrate the medium effect theoretically, it is necessary to have a reliable framework in which to do calculation from first principles, and that is perturbative QCD. Although many studies in pQCD have been carried out to learn about the modification of jets in dense medium in various approximation schemes,51,98 they are mainly concerned with the effect of the medium on jets at high pT , and the results can only be
Hadron Correlations in Jets and Ridges Through Parton Recombination
285
compared with data on singleparticle distributions, such as RAA (pT ). The response of the medium to the passage of hard partons is not what can be calculated in pQCD, since it involves soft physics. That is, however, the physical origin of most of the characteristics in the correlation data. An understanding of that response is one of the objectives of studying correlations. Without the reliable theory to describe correlation, especially at low to intermediate pT where abundant data exist, it becomes necessary to use phenomenological models to relate various features of correlation. When all the features can consistently be explained in the framework of a model, then one may feel that a few parameters are a small price to pay for the elucidation of the jetmedium interaction. On twoparticle correlation the most active area in recent years has been the use of triggers at intermediate or high pT to select a restricted class of events and the observation of associated particles at various values of η and φ relative to the trigger.2,99 Among the new features found, the discovery of ridges on the near side has stimulated extensive interest and activities.100 We review in this section only those aspects in which recombination plays an important role, which in turn makes inferences on the origin of the ridges. We start with a summary of the experimental facts. 4.1. Experimental features of ridges The distribution of particles associated with a trigger at intermediate pT exhibits a peak at small ∆η and ∆φ sitting on top of a ridge that has a wide range in ∆η, where ∆η and ∆φ are, respectively, the differences of η and φ of the associated particle from those of the trigger.99,101 A 2D correlation function in (∆η, ∆φ) first shown by Putschke101 at QM06 is reproduced here in Fig. 5(a). STAR has been able to separate the ridge (R) from the peak (J), where J refers to Jet, although both are features associated with jets. The structure shown in Fig. 5(a) is for 3 < ptrig < 4 GeV/c and passoc > 2 GeV/c in central AuAu collisions. The ridge T T yield integrated over ∆η and ∆φ decreases with decreasing Npart , until it vanishes at the lowest Npart corresponding to pp collisions, so R depends strongly on the nuclear medium. That is not the case with J. On the other hand, R is also strongly correlated to jet production, since the ridge yield is insensitive to ptrig T . Thus the ridge is a manifestation of jetmedium interaction. Putschke further showed101 that the ridge yield is exponential in its dependence on passoc and that the slope in the semilog plot is essentially independent of ptrig T T . That is shown by the solid lines in Fig. 5(b). The inverse slope parameterized by T is slightly higher than T0 of the inclusive distribution, also shown in that figure. Since the pT range in that figure is between 2 and 4 GeV/c, we know from singleparticle distribution that the shape of the inclusive spectrum is at the transition from pure exponential on the low side to powerlaw behavior on the high side. The last data point at passoc = 4 GeV/c being above the straight line is an indication T of that. Thus the value T0 of the pure exponential part for the bulk is lower than
286
R. C. Hwa
(b) (a)
Fig. 5. (a) Jet structure from Ref. 101 for charged particles associated with a trigger. (b) Dependence of ridge yield101 on passoc for various ptrig T T .
what that straight line suggests. The exponential behavior of R should be taken to mean that the particles in the ridge are emitted from a thermal source. Usually thermal partons are regarded as begin uncorrelated. In the case of R they are all correlated to the semihard parton that initiates the jet. We thus interpret the observed characteristics as indicating that the ridge is from a thermal source at T , enhanced by the energy lost by the semihard parton transversing the medium at T0 . The B/M ratio of particles in the ridge is found to be even higher than the same ratio of the inclusive distributions in AuAu collisions at 200 GeV. More specifically, (p + p¯)/(π + + π − ) in R for ptrig > 4 GeV/c and 2 < passoc < ptrig is about 1 at T T T 102 pT = 4 GeV/c. In contrast, that ratio in J is more than 5 times lower. There is indication that the Λ/K ratio in the ridge is just as large.71 As discussed in Sec. 3, it is hard to find any way to explain the large B/M ratio outside the framework of recombination. Since the exponential behavior in pT implies the hadronization of thermal partons, the application of recombination very naturally gives rise to large B/M ratio, as we have seen in Sec. 3.1. Putting together all the experimental features discussed above, we can construct a coherent picture of the dynamical origin of the R and J components of the jet structure, although no part of it can be rigorously proved for lack of a calculationally effective theory of soft physics. There are several stages of the dynamical process. (i) A hard or semihard scattering takes place in the medium resulting in a parton directed outward in the transverse plane at midrapidity. Because of energy loss to the medium, those originating in the interior are not able to transverse the medium as effectively as those created near the surface. That leads to trigger bias. (ii) Whatever the nature of the jetmedium interaction is, the energy lost from the semihard parton goes to the enhancement of the thermal energy of the partons in the near vicinity of the passing trajectory. Those enhanced thermal partons
Hadron Correlations in Jets and Ridges Through Parton Recombination
287
are swept by the local collective movement outward whether or not the flow can be described by equilibrated hydrodynamics initially. (iii) Since the initial scattering takes place at η < 0.7, which is the pseudorapidity range of the trigger acceptance, the shower (S) partons associated with the jet are restricted to the same range of η. However, the enhanced thermal partons that interact strongly with the medium can be carried by the highη initial partons that they encounter on the way out and be boosted to higher η. Thus the distribution of the enhanced thermal partons is elongated in ∆η, but not in ∆φ because the expansion of the bulk system is in longitudinal and radial directions, not in the azimuthal direction. Consequently, the hadronization of the enhanced thermal partons has the shape of a ridge. (iv) In terms of recombination the ridge is formed by TT and TTT recombination, while the peak J is formed largely by TS and TTS (or TSS) recombination, and possibly also by fragmentation (SS and SSS), depending on pT and centrality. Since the J component involves S, it is restricted to a narrow cone in ∆η and ∆φ. An initial attempt to incorporate all these properties in the RM was made in Ref. 15 before the ridge data were reported in QM06.101 By the time of QM08 ridgeology has become an intensely studied subject, as evidence by the talks in Ref. 100. 4.2. Recombination of enchanced thermal partons Although the properties of ridges described in the above subsection are derived from events with triggers, it should be recognized that ridges are present with or without triggers. That is because the ridges are induced by semihard scattering which can take place whether or not a hadron in a chosen pT range is used to select a subset of events. Experimentally, it is known that the peak and ridge structure is seen in autocorrelation where no triggers are used.103 The implication is that the ridge hadrons are pervasive and are always present in the singleparticle spectra. Hard scattering of partons can occur at all virtuality Q2 , with increasing probability at lower and lower Q2 . When the parton kT is < 3 GeV/c, the rate of such semihard scattering can be high, while the time scale involved is low enough (∼ 0.1 fm/c) to be sensitive to the initial spatial configuration of the collision system. Thus for noncentral collisions there can be nontrivial φ dependence, which we shall discuss in Sec. 5. Hadron formation does not take place until much later, so it is important to bear in mind the two time scales involved in ridgeology. Ridges are the hadronization products of enhanced thermal partons at late time, which are stimulated by semihard parton created at early time. In the absence of a theoretical framework to calculate the degree of enhancement due to energy loss, we extract the characteristics of the thermal distributions from the data. Although hydrodynamics may be a valid description of the collective flow after local thermal equilibrium is established, it does not take semihard scattering into consideration and assumes
288
R. C. Hwa
fast thermaliztion without firmly grounded justification. If the semihard scattering occurs in the interior of the dense medium, the energy of the scattered parton is dissipated in the medium and contributes to the thermalization of the bulk (B). That process may take some time to complete. If the semihard scattering occurs near the surface of the medium, its effect can be detected as J + R in these events selected by a trigger with the trigger direction not far from the local flow direction, a point to be discussed in more detail later in Sec. 4.4. Inclusive distribution averages over all events without triggers, including all manifestation of hard and semihard scatterings; hence, it is the sum of B + R + J. Since J is associated with the shower partons (S), we identify J with the recombination of TS+SS for the mesons and TTS+TSS+SSS for the baryons, leaving TT+TTT for B +R. Thus the exponential behavior of the thermal partons is revealed in the exponential behavior of B + R in pT , for which we emphasize the inclusion of the ridge contribution to the inclusive distribution. In noncentral collisions the ridges are not produced uniformly throughout all azimuth,104 so dN/dpT that averages over all φ has varying proportions of B and R contributions depending on centrality. To be certain that we can get a measure of the R contribution independent of φ, we focus on only the most central collisions in this and the next subsections. Continuing to use the notation kT for the transverse momentum of the semihard parton at the point of creation in the medium, qT for that at the point of exit from the medium, and pT for the hadron outside, we have for thermal partons the distribution given in (9) just before recombination. Our first point to stress here is that the inverse slope T in (9) includes the effects of both B and R. Putting that expression into (7) where one takes Fqq¯(q1T , q2T ) = T (q1T )T (q2T ) and being more explicit with the RF for pion in (3), i.e., q1T q2 q1 q2 + T −1 , Rπ (q1T , q2T , pT ) = T 2 T δ pT pT pT
(15)
(16)
one obtains35 dNπB+R C 2 −pT /T = e , pT dpT 6
(17)
although in 2004 no one was aware of the existence of ridges. From the data105 for identified hadrons, one can fit the π + distribution for 0–5% centrality in the range 1 < pT < 3 GeV/c and get T = 0.3 GeV/c. This value is slightly lower than the one given in Ref. 101 which takes the slope of the inclusive distribution in the range 2 < pT < 3.5 GeV/c. Reference 105 provides data for K and p also, which have the same value of T as above for 1 < pT < 3 GeV/c, thus confirming that the exponential behaviors of the hadronic spectra can be traced to the common thermal distribution in (9) through recombination. At lower pT the spectra for K
Hadron Correlations in Jets and Ridges Through Parton Recombination
289
and p deviate from exponential behavior because of mass effect, which can largely be taken into account by using ET instead of pT , where ET (pT ) = mT − m0 ,
mT = p2T + m20
1/2
,
(18)
m0 being the hadron rest mass. Thus we write for all hadrons 0 dNhB+R = Ah (pT )e−ET (pT )/Th , pT dpT
(19)
where Aπ (pT ) = C 2 /6 is a constant for pion, but for proton Ap (pT ) = C 3 A0 p2T /p0 where A0 is a numerical factor that arises from the wave functions (valon distribution) of the proton.35 Note that the inverse slope is now denoted by Th0 , since the data105 show dependence on hadron type when the distributions are plotted as functions of ET . Furthermore, Tp0 is found to depend on centrality, which is a feature that can be understood in the RM as being due to the nonfactorizability of the thermal parton distributions of uud at very peripheral collisions where the density of thermal partons is low.106 For central collisions, Tp0 = 0.35 GeV. We summarize the empirical results for π and p as follows: Tπ0 = 0.3 GeV,
(20)
Tp0 = 0.35(1 − 0.5c) GeV,
(21)
where c denotes % centrality, e. g., c = 0.1 for 10%. We shall hereafter use Th0 to denote the inverse slope in ET for B + R, and Th for that in ET for B only, i.e., dNhB = Bh (pT ) = Ah (pT )e−ET (pT )/Th . pT dpT
(22)
It is hard to find data that describes the bulk contribution only, since the effect of semihard scattering cannot easily be filtered out. Indeed, as pT → 0, there is no operational way without using trigger to distinguish B(pT ) from all inclusive. For that reason the prefactor Ah (pT ) in (22) is the same as that in (19). In events with trigger above a threshold momentum, semihard partons with lower momenta than that threshold can contribute to R; it becomes a part of the background, which is experimentally treated as B. Thus the only meaningful way to isolate R quantitatively is by use of correlation, while accepting the difficulty of separating B and R outside the momentum ranges where the correlated particles are measured. Another way of stating that attitude is to accept the experimental paradigm of regarding the mixed events as a measure of the background (hence, by definition, the bulk), and treating R as only that associated with a trigger. Our cautionary point to make is that such a background can contain untriggered ridges. In practice, one can take the difference between (19) and (22) and identify it as the ridge yield
290
R. C. Hwa
i h 00 0 dNhR = Rh (pT ) = Ah (pT )e−ET (pT )/Th 1 − e−ET (pT )/Th , pT dpT
(23)
where 1 1 ∆Th 1 = , − 0 = 00 Th Th Th Th Th0
∆Th = Th0 − Th .
(24)
If ∆Th Th , then the quantity in the square bracket makes a small correction to the exp [−ET (pT )/Th0 ] behavior, and one can determine Th0 from the data. The only data available that address the ridge distribution are in Ref. 101 where the associated particles are in the range 2 < passoc < 4 GeV/c, exhibiting an approxiT mately exponential behavior. It is shown in Ref. 106 by using the data for trigger momentum in the range 4 < ptrig < 5 GeV/c that with ∆Th = 45 MeV in (24) T the ridge distributions can be well fitted. The expression for Rh (pT ) in (23) has no explicit dependence on ptrig T , as is roughly the case with the data. It does have strong dependence on passoc , which is pT in (23). Experimental exploration of the T lower passoc region would provide further validation that (23) needs. The physics T basis for that distribution is the recombination of thermal partons given in (9). 4.3. Trigger from the ridge We have discussed above the observation of ridge in triggered events, but to have a trigger from the ridge seems to put the horse behind the cart. There must be a phenomenological motivation for that role reversal. Let us start with the singlepion inclusive distribution that shows an exponential decrease in pT followed by a powerlaw behavior. The boundary between the two regions is at ∼ 2 GeV/c. We have associated the exponential region to TT recombination and the powerlaw region to TS+SS. We have also discussed the contribution to T from the enhanced thermal partons arising from the medium response to semihard partons. In order to be able to investigate the TT component better without the interference from the shower contribution so that one can examine the B + R components cleanly, it would be desirable to be able to push the TS+SS components out of the way. That is not possible with the light u and d quarks, but not impossible with the s quark, since the heavier quark is suppressed in hard scattering. If one observes the hadrons formed from only the s quarks, either φ or Ω, one finds exponential behavior at all pT measured, which in the case Ω extends to as high as 5.5 GeV/c.107 – 113 The absence of any indication of upbending of the distributions clearly suggests that the source of the s quarks is thermal in nature and that no shower partons participate in the formation of φ and Ω. That problem is studied in Ref. 114 along with K and Λ production. Indeed, the data can be well reproduced by TT for φ, TTT for Ω, TT+TS for K, and TTT+TTS+TSS for Λ.
Hadron Correlations in Jets and Ridges Through Parton Recombination
291
Since s quarks in S make insignificant contribution to Ω production for pT < 6 GeV/c, and since thermal partons are uncorrelated, it is reasonable to expect that the Ω observed has no correlated particles. It was therefore predicted 115 that if Ω in 3 < pT < 6 GeV/c is treated as a trigger particle, there should be no associated particles appearing as a peak in ∆φ on the near side. Within a year that prediction was falsified by a report at QM06 showing that there is a nearside peak after background subtraction in ∆φ for Ω in the range 2.5 < ptrig < 4.5 GeV/c in T 116 central AuAu collisions with charged particles in the range 1.5 < passoc < ptrig T T . The data created a dilemma: is Ω created by a jet or not? If it is, why is the pT distribution strictly exponential with no hint of jet characteristics? If it is not, why is there a ∆φ peak in azimuthal correlation? The dilemma became known as the Ω puzzle.117 The resolution of that puzzle is in the recognition that both the trigger Ω and the associated particles are in the ridge, first conjectured in Ref. 117 and later quantified in Ref. 118. Jets are involved, since without jets there can be no ridge. But not all jet structures exhibit a prominent peak above a ridge. It depends on the trigger particle and the ranges of ptrig and passoc . Consider the jet yield compared to T T trig the ridge yield at 3 < pT < 6 GeV/c in AuAu collisions at 0–10% centrality.71,119 The J/R ratio at passoc ∼ 1.2 GeV/c decreases as the trigger particle goes from h T ¯ For Λ/Λ ¯ trigger and unidentified charged h associated, the to KS0 and then to Λ/Λ. −1 ¯ trigger must J/R ratio is ≤ 10 for ∆ηJ < 0.7 and ∆ηR < 1.7. Since the Λ/Λ contain an s quark which is absent in the shower, the participating s quark must be a thermal parton. For ptrig near 3 GeV/c, thermal s quark around 1 GeV/c or less T can be quite abundant, and the initiating semihard parton need not be very hard. With passoc as low as ∼ 1 GeV/c, the light hadrons in R dominate over those in J, so T J/R is small. As the strangeness in the trigger increases, more thermal s quarks are involved with less dependence on shower. J/R is likely to be even smaller, although present data with Ω trigger lack statistics to show the ∆η distribution. If the jet structure shows mainly a ridge with negligible peak (J) in ∆η, we have referred to it as a phantom jet,117 i.e., a Jetless jet. The corresponding ∆φ distribution should then be dominantly R. Since the initiating semihard partons are either gluon or light quarks, the usual jet structure may still be seen if the trigger particle is ordinary. But for Ω trigger the structure is very different. Now, we can address the Ω puzzle. The enhanced thermal partons generated by the semihard parton contain s as well as u and d quarks. Three s quarks anywhere in the ridge can recombine to form a trigger Ω. Other quarks, in particular the light quarks, in the ridge can form associated particles. The pool of enhanced thermal partons are all correlated to the semihard parton in every event selected by the Ω trigger, so the associated particles are all restricted to ∆φ < 1. Since the s quarks that form the Ω are thermal, the Ω spectrum in pT is exponential. Hence the Ω puzzle is solved. The trigger can be from the ridge. The above description outlines a detailed calculation of the ∆φ distribution of charged hadrons produced in association with Ω.118 The background is calculated
292
R. C. Hwa
(a)
(b)
Fig. 6. (a) (Left panel: Data119 on associated particle distribution in ∆φ for three hyperon triggers at 2.5 < ptrig < 4.5 GeV/c and 1.5 < passoc < ptrig T T T . (b) Right panel: Nearside ridge yield associated with Ω trigger. The lines are from calculations in the RM118 for particles associated with Ω trigger, with the solid and dashed lines for slightly different normalization constants of the Gaussian peak.
in the RM using previous parametrizations, and the height agrees with the data. That is important, since the ridge signal is less than 4% of the background height. Two adjustable parameters are used to fit the ∆φ distribution of the ridge, but then the yield/trigger is calculated as a function of ptrig without further unknowns T in the model, and the result is in agreement with the data,71,119 as shown in Fig. 6. The solid and dashed curves in the two panels of that figure are the results of the calculation using two values of the strength of enhanced thermal partons that differ by only 1%, yet the height of the ridge varies by about 20%. That is because the ridge is the difference between large numbers of B + R and B. Such accuracy is beyond the scope of any dynamical theory to achieve. The phenomenological approach adopted in Ref. 118 has been the only one that offers a quantitative understanding of the Ω problem. Recently, PHENIX has shown data120 that can be interpreted as support for the notion of trigger from the ridge. At ptrig < 4 GeV/c (for unidentified trigger) T the pertrigger yield of the associated particles is found to be less than expected from fragmentation, and the “dilution” effect is attributed to the increase of the number of triggers due to soft processes. In our language fragmentation is the SS component, and the TS component due to medium effect gives an increase already over fragmentation for pT < 6 GeV/c. The additional dilution effect at pT < 4 GeV/c is due to TT recombination, which is enhanced by the ridge contribution. 4.4. Dependence of ridge formation on trigger azimuth So far our consideration in ridgeology has been concerned mainly with the dependence on pT . Now, let us turn to azimuthal correlation, although our discussion of the main topic of azimuthal anisotropy of singleparticle distributions is deferred
Hadron Correlations in Jets and Ridges Through Parton Recombination
293
to the following section. Our present focus is on the correlation between the directions of the trigger and ridge particles in the transverse plane, a topic that can be discussed prior to φ anistropy because it is mainly a problem in ridge formation. The subject was stimulated by the report104 that the ridge yield depends on the azimuthal angle φs between the trigger angle φT and the reaction plane ΨRP , even for nearly central collisions, but especially for noncentral collisions. Since geometry is an important factor that influences the φs dependence, it is necessary to treat carefully the initial configuration of the problem: (a) the point of origin (x0 , y0 ) of the semihard parton in the almondshaped overlap region, (b) the angle φs of the parton’s trajectory, (c) the density of the medium, D(x, y), along that trajectory, and (d) the point of exit from the medium. In the approximation that the medium does not expand very much during the time that the semihard parton near the surface traverses the medium, it is not difficult to calculate the path length, but it is much more difficult to calculate the energy loss that depends on D(x, y). Even if there is a reliable way to account for the effect of the medium on the semihard parton, there is no known way to translate that to the effect of the parton on the medium. The enhancement of the thermal partons that lead to the ridge particles takes time to develop, during which the medium expands. Local flow direction depends on where the enhanced thermal partons are in the overlap, which evolves into elliptical geometry. Each of the various parts of the process can be represented by a factor that can be expressed in terms of variables that have reasonable physical relevance. Without entering into the details that are described in Ref. 121 we can outline the essence here. Let P (x0 , y0 , t) denote the probability of detecting a parton emerging from the medium, where t is the path length measured from the initial point (x0 , y0 ) to the surface along a straightline trajectory at angle φs . Let C(ψ(x, y), φs ) be a function that describes the correlation between the enhanced thermal partons along the flow direction ψ(x, y) and the semihard parton direction φs . Finally, let Γ(x, y, φ) describe the fluctuation of the angle φ of a ridge particle from the average flow direction. For fixed (x0 , y0 ) and φs , the ridgeparticle distribution in φ is then R(φ, φs , x0 , y0 ) = N P (x0 , y0 , t)t
Z
1
dξD(xξ , yξ )C(xξ , yξ , φs )Γ(xξ , yξ , φ) , (25) 0
where N is a normalization constant related to the rate of production of the ridge particle and ξ is the fraction of the path length t along the trajectory starting at 0 at (x0 , y0 ) and ending at 1. For the observed distribution it is necessary to integrate over all (x0 , y0 ). Not every semihard parton included in that integration gets out of the medium to generate a particle that triggers the event. The ridge distribution per trigger is therefore normalized by the probability of the ridgegenerating parton emerging from the medium R dx0 dy0 R(φ, φs , x0 , y0 ) , (26) R(φ, φs ) = R dx0 dy0 P (x0 , y0 , t)
294
R. C. Hwa
where the integration is over the entire region of initial overlap. There is no explicit dependence on ptrig and passoc in (25), but the parameters specifying the different T T factors in the equation do. We leave the momenta fixed in the ranges 3 < ptrig 0 most of the points along the trajectory have ψ(x, y) < φs , since ψ(x, y) is generally normal to the surface. Hence, the ridge particles are mostly at ∆φ < 0. The reverse is true for φs < 0. The left shift of the peaks in ∆φ in Fig. 8(a) has been confirmed by STAR recently.122
(a)
(b)
Fig. 8. (a) Left shift of the ridge yield in ∆φ for φs > 0 in CEM.121 (b) Insideoutside asymmetry function defined in Eq. (29).
296
R. C. Hwa
Another way to test the asymmetry is to measure a quantity A(φs ) called insideoutside asymmetry function.121 To that end, define for 0 < φs < π/2 Z φs Z φs +1 Y+ (φs ) = dφR(φ, φs ), Y− (φs ) = dφR(φ, φs ) (28) φs −1
φs
and for −π/2 < φs < 0 reverse the definition. Then for any φs define A(φs ) =
Y+ (φs ) − Y− (φs ) . Y+ (φs ) + Y− (φs )
(29)
This asymmetry function should always be positive whether there is a left shift for φs < 0 or a right shift for φs > 0. By reflection symmetry it vanishes at φs = 0 and ±π/2. In CEM the properties of A(φs ) at two centralities are shown in Fig. 8(b). STAR reported very recently at QM09 that such an asymmetry has indeed been found to exist in the data on the ridges for various φs .122,123 5. Azimuthal Anisotropy The azimuthal dependence of singleparticle distribution has been studied ever since the beginning of heavyion physics.124 – 127 Hydrodynamical model at low pT 128 – 132 and ReCo at intermediate pT 33,34,133,134 have been able to describe the data on elliptic flow very well. The only points worthy of comments here are those outside the realm of what has been covered in the references given above. There are then only three points: (a) Is early thermalization necessary? (b) What is the role of the shower partons? (c) At what pT does the quark number scaling begin to break down? 5.1. Effects of ridge formation at low pT The usual hydrodynamical treatment of elliptic flow assumes rapid thermalization with initial time of expansion set at τ0 = 0.6 fm/c. Such an early time of equilibration has never been shown to be the consequence of any dynamical process that is firmly grounded and commonly accepted. The question then is whether the azimuthal anisotropy can be driven initially not by pressure gradient at τ < 1 fm/c, but by some other mechanism that is sensitive to the early spatial configuration. That mechanism is suggested in Ref. 135 to be semihard scattering, the rate of which can be high for parton kT around 2–3 GeV/c, while the time scale involved is low (∼ 0.1 fm/c). Semihard partons created near the surface of the nuclear overlap can lead to a continuous range of ridges that is shaped by the initial geometry. The effects of such ridges are not considered in the usual studies in hydrodynamics, but should not be ignored. It is shown in Refs. 135 and 106 that the azimuthal anisotropy of the ridges produces the observed v2 at low pT without the use of hydrodynamics. Elliptic flow at some point of the expansion of the system may well be describable by hydrodynamics, but early thermalization is not required.
Hadron Correlations in Jets and Ridges Through Parton Recombination
297
For dN/dφ of singleparticle distribution, no triggers are used and ridges are generated by many semihard partons produced in each event. Semihard scatterings take place throughout the overlap, but only those occurring near the surface and directed normal to the surface can lead to the development of ridges, as discussed in Sec. 4.4. If a drastic simplification is made to require ridge particles to be also directed only in the the directions normal to the almondshaped initial boundary in the transverse plane, then the φ distribution of all ridges, R(pT , φ), is restricted to the region φ ∈ R, which is a set of angles defined by φ < Φ(b)
and
π − φ < Φ(b) ,
(30)
where Φ(b) = cos−1 (b/2RA ).135 Fluctuation from the restricted range is, of course, possible, not only because the semihard partons can have any scattered angle, but also because the ridge particles can fluctuate from the directions of the parton trajectories. It is shown in the Appendix of Ref. 106 that the region R is nevertheless a good approximation even when all those effects plus elliptic geometry are taken into account, provided that the inaccuracy in the regions around φ ∼ π/2 for noncentral collisions is unimportant — which is indeed the case, since the density at the upper and lower tips of the ellipse is low so ridge production there is suppressed. Thus in the box approximation the ridge distribution is R(pT , φ) = R(pT )Θ(φ) ,
(31)
Θ(φ) = θ(Φ − φ) + θ(Φ − π − φ) .
(32)
where
In the assumption that the above anisotropy from the ridge is the only φ dependence in dN/dφ, replacing the usual assumption of rapid thermalization and the consequent pressure gradient, the bulk medium is then isotropic and the singleparticle distribution at low pT can be written in the form dN = B(pT ) + R(pT )Θ(φ) . pT dpT dφ
(33)
The normalized second harmonic in φ can then be calculated analytically, yielding v2 (pT , b) = hcos 2φib =
sin 2Φ(b) . πB(pT )/R(pT ) + 2Φ(b)
(34)
At low pT the first term in the denominator is much larger than the second, so (34) is reduced to the even simpler formula v2 (pT , b) '
R(pT ) sin 2Φ(b) , πB(pT )
(35)
This is such a compact formula that its validity should be checked regardless of its derivation for the sake of having available a simple description of what is usually called elliptic flow after extensive computation.
298
R. C. Hwa
The pT distributions of B(pT ) and R(pT ) are given, respectively, in (22) and (23). The latter can be written in a form that factors out B(pT ) so that R/B has the simple form 00 R(pT ) = eET (pT )/Th − 1 , B(pT )
(36)
which becomes ET (pT )/Th00 at low pT , since Th00 is large. It then follows from (35) that for small ET v2h (ET , b) =
ET sin 2Φ(b) . πTh00
(37)
Thus the initial slope in ET depends only on Th00 (which sets the scale) and the geometric factor sin 2Φ(b). That factorizability is in agreement with the data for pion, but not so well for proton for midcentral to peripheral collisions. That is because Tp00 develops a b dependence when the centrality c is above 0.3, as can be seen in (21). While Tπ00 is essentially constant Tp00 (c) can be approximated by a polynomial in c106 Tπ00 = 1.7 GeV,
Tp00 = 2.37 (1 − 1.05c + 0.26c2 ) GeV .
(38)
For ET ∼ 1 GeV the full expression in (36) should be used in (35), instead of the small ET approximation in (37). The results for π and p are shown in Fig. 9, and reproduce the data124 very well. The reason why Tp0 (c) decreases with increasing peripherality, resulting in similar trend in Tp00 (c), is that 3quark recombination is more difficult when the thermal partons become less abundant at lower medium density as the collisions get more peripheral. At fixed pT ∼ 0.5 GeV/c the b dependences of v2h for h = π and p are studied in Ref. 106 in which it is shown that the characteristics of the data are well captured by the simple formula in (34).
(a)
(b)
Fig. 9. (a) Pion v2 and (b) proton v2 , calculated in the RM106 by taking ridge effect into account. The data are from Ref. 124.
Hadron Correlations in Jets and Ridges Through Parton Recombination
299
5.2. Effects of shower partons at intermediate pT As pT is increased to above 2 GeV/c, it is necessary to consider the role played by the shower partons, which introduce φ dependence due to jet quenching of hard parton that depends on path length in the medium. The dominance of TS + SS recombination over TT leads to a change in the pT dependence of v2 (pT ). The shower parton distribution S(q), given in (11), due to a hard parton produced at momentum k, is for central collisions and is averaged over all φ. Now for noncentral collisions with φ anisotropy, that formula needs to √ be generalized. Assuming that the energy loss of a hard parton is proportional to k 0 where k 0 is the initial parton momentum,34,136 one can write ∆k = k 0 − k in the form √ ˆ φ) k 0 , (39) ∆k = (b)`(b, ˆ φ) is the normalized path length and (b) is the energyloss coefficient where `(b, that depends, apart from geometrical factors, a parameter 0 that is to be determined. After solving (39) for k 0 , and replacing ζfi (k) in (11) by fi (k 0 ) at the shifted momentum, one can keep the first two nonvanishing terms in the harmonic expansion of fi (k 0 (k, b, φ)) and get fi (k 0 (k, b, φ)) = fi (k) [g0 (k, b) + 2g2 (k, b) cos 2φ] ,
(40)
where g0 and g2 can be determined explicitly in terms of 0 , k and b.106 The shower contribution to the singlepion distribution, when averaged over all φ, may be written in the symbolic form dNπT S+SS (b) 1 = g0 (k, b)fi (k) ⊗ (T S + SS) . pT dpT 2π
(41)
The contribution of the shower component to v2π (pT , b) is v2π,sh (pT , b) =
g2 (k, b)fi (k) ⊗ (T S + SS) . g0 (k, b)fi (k) ⊗ (T S + SS)
(42)
The thermal component v2π,th (pT , b) is as given in (34) and (36). The overall v2π is obtained from the above with the help of an interpolating function W (pT , b) v2π (pT , b) = v2π,th (pT , b)W (pT , b) + v2π,sh (pT , b) [1 − W (pT , b)] ,
(43)
where W (pT , b) =
TT , T T + T S + SS
(44)
with T T being the thermal, and T S +SS the shower, components of the φaveraged dNπ /pT dpT . By fitting the singlepion distribution at 0–10% centrality over the range 1 < pT < 6 GeV/c using all three T T + T S + SS terms, 0 is found to be 0.55 GeV1/2 . It is then possible to calculate v2π (pT , b) without any further adjustment; the result is shown in Fig. 10(a). The data points for ET > 1 GeV are from
300
R. C. Hwa
(a)
(b)
Fig. 10. (a) Left panel: pion v2 and (b) Right panel: proton v2 , both at higher pT where the effects of shower partons are taken into account. The data are Refs. 126 and 137.
Ref. 137. The saturation of v2π in that range is thus interpreted in the RM as being due to the shower partons, where W (pT , b) is suppressed and g2 is much smaller than g0 at high k. For proton v2 the general procedure in the calculation is similar to that for pion, except for an additional thermal parton to incorporate. There is also the complication of b dependence in the inverse slope Tp0 in (21). Taking them all into account the result for v2p is shown in Fig. 10(b). Comparison with data126,137 is acceptable, although more accurate data are needed to check the calculated results at high ET and b. What is learned from this study is that the main source of φ anisotropy is the pathlength dependence of jet quenching, which is parametrized by one unknown 0 that is determined by fitting the inclusive distribution at one value of pT . The characteristics of v2h (pT , b) in Fig. 10 are obtained without any more adjustable parameters. It should be noted that we have refrained from using the term “elliptic flow”, except in reference to past work based on hydrodynamics. In Sec. 5.1 the emphasis is on the effect of ridges due to semihard scattering that are not taken into account by earlytime hydrodynamics, although the exponential behavior of the thermal partons may well be the result of latetime hydrodynamical flow. Then in this section hard scattering is incorporated in the treatment of jetmedium interaction through TS recombination before fragmentation dominates. Such interaction is outside hydrodynamics, so the overall characteristics of v2h (pT , b) are not the properties of flow. In Ref. 34 the effects of jet quenching on the hard partons are also considered, but since only the fragmentation of those partons is included, the transitional contribution from TS interaction is not explicitly taken into account. That turns out to be important in the intermediate pT region.
Hadron Correlations in Jets and Ridges Through Parton Recombination
301
5.3. Breaking of quark number scaling Quark number scaling (QNS) has long been regarded as a signature of recombination33,34,133 and has been verified in a number of experiments.124,126,127 It has been regarded as a statement of the universality of v2h (pT /nq )/nq , where nq is the number of constituent quarks in the hadron h. It is based on the assumption of factorizability of the distribution of the quarks that recombine, i.e., the multiquark distributions in (7) and (8) (but with φi dependencies included) can be written as Fnq (q1 , φ1 ; q2 , φ2 ; · · · ) =
nq Y
i=1
Fi (qi , φi ) =
nq Y
Fi (qi )[1 + 2v2i (qi ) cos 2φi ] .
(45)
i=1
Coupled with the assumption that the RF has the simple form δ(qi − pT /nq ), it then follows trivially from (7) and (8) that v2h (pT ) = nq v2q (pT /nq ) ,
(46)
if the v2i (qi ) of all quarks in (45) are the same, denoted by v2q (q). From our discussions throughout this paper it is clear that none of the above assumptions are valid under close examination. TT+TS+SS for pion and TTT+TTS+TSS+SSS for proton are obviously not factorizable. Even at low p T where only the recombination of thermal partons is important, the inverse slopes Th0 , given in (20) and (21), are not the same for π and p. Consequently, the R/B ratios for π and p are different, as seen in (36) and (38), resulting in different v2h (ET , b). Furthermore, the wave functions of π and p are very different, since the pion is a tightly bounds state of the constituent quarks, while the proton is loosely bound. That means the momentum fractions of the quarks (valons) are not 1/2 for pion and 1/3 for proton. It is then a very rough approximation to write the momentum conservation δfunction, δ(Σqi − pT ), as δ(nq qT − pT ) with a common qT . At intermediate pT where shower partons become important, we have seen that they acquire the φ dependence of the hard parton, given in (40), so v2S (qi ) for the shower is different from v2T (qj ) for the thermal parton. Even if TS and TTS contributions dominate, one can at best, by ignoring all other complications, have v2M (pT ) = v2T (q1 ) + v2S (q2 ),
v2B (pT ) = v2T (q1 ) + v2T (q2 ) + v2S (q3 ) .
(47)
They would not lead to QNS, as expressed in (46). Most data in support of QNS are for minimum bias and at low ET . It is shown in Ref. 106 that the calculated result at ET /nq < 0.5 GeV does exhibit QNS in agreement with the data, but the scaling is broken above that. The breaking of QNS is due primarily to the nonequivalence of the φ dependencies of the thermal and shower partons. The important point to stress here is that the breaking of QNS at intermediate ET does not imply the failure of recombination (in fact, it is expected), but the validity of QNS at lower ET does confirm recombination as the proper hadronization process.
302
R. C. Hwa
6. Hadron Correlation in Dijet Production In Sec. 4 the subject of discussion is ridgeology, which is the study of the effects of jets on the medium. In this section we consider the reverse, i.e., the effects of the medium on the jets for pT not so extremely high as to exclude TS recombination. In triggered events the ridge is the broad pedestal on top of which sits the peak. The structure of that peak, when compared to the jet structure in pp collisions, reveals the medium effect on hard partons. Since the peak is restricted to a narrow cone around the trigger direction, it is natural in the RM to associate it with shower partons, while the ridge being broad is associated with the thermal partons. Hadrons closely correlated to a trigger particle as jets exhibit peaks on both the same and away sides.138 – 142 How the structures of the jets on the two opposite sides differ from each other is a strong indication of the difference in energy losses in the two jets, since their path lengths in the medium are generally different. In realistic collisions even when centrality is chosen to be within a narrow range, the path lengths can vary significantly depending on the location and angle of a scattered hard parton. Thus a careful study of the properties of the near and awayside jets must start with finding a good description of the variation of energy loss within each class of centrality. 6.1. Distribution of dynamical path length By dynamical path length we mean not only the geometrical length of a trajectory, but also the medium effect along that trajectory. To quantify that in an analytic expression, we need to revisit the singleparticle distribution discussed earlier for central collision, but now formulated in a way appropriate for any centrality. For pion production we have for y ∼ 0 Z p π p C 2 −p/T 1 X dq dNπ c = e + 2 Fi (q) TS(q, p) + Di , pdp 6 p i q q q
(48)
where all momenta are in the transverse plane with the subscript T omitted. The first term on the right side is from (17) for TT recombination; the centrality dependence of C is given in Ref. 143. The first term in the square bracket is for TS recombination which will be detailed below, and the second term is the fragmentation function that is equivalent to SS recombination. Fi (q) is the distribution of parton i at the surface of the medium with q denoting the momentum of the hard parton there. It differs from the distribution fi (k) given in (12), that describes the hard parton with momentum k at the point of creation. In Secs. 2 and 3 where only central collisions are considered, an average suppression factor ζ is used, as shown in (11). We now replace ζfi (k) by a pathdependent term related to Fi (q) by Fi (q) =
Z
L 0
dt L
Z
dkkfi (k)G(q, k, t) ,
(49)
Hadron Correlations in Jets and Ridges Through Parton Recombination
303
where G(q, k, t) is a degradation factor that describes the decrease of parton momentum from k to q as the hard parton traverses a distance t through the medium.144 L is the average maximum length of that trajectory. In the limit L → 0, Fi (q) should become the parton distribution function Fi (k) for pp collisions. For energy loss we seek a form that is consistent, on one hand, with ∆E ∝ L as suggested in Ref. 97 for 1D expansion, and on the other hand, with hdE/dLi ∝ E in Ref. 145 for 6 < E < 12 GeV. A reasonable approximation of the differential energy loss is then ∆E = β∆L , E
(50)
which translates to our variables as k − q = kβt .
(51)
at small t with β being an adjustable parameter. For larger t we exponentiate the above to q = ke−βt ,
(52)
and let G(q, k, t) take the simple form G(q, k, t) = qδ(q − ke−βt ) .
(53)
The δ function can be broadened to account for fluctuations, but we shall take (53) as an adequate approximation of the complicated processes involved in the partonmedium interaction, the justification of which rests ultimately on how well the calculated result can agree with the data for 2 < pT < 12 GeV/c. Using (53) in (49) yields 1 Fi (q) = βL
Z
qeβL
dkkfi (k) ,
(54)
q
which shows explicitly how fi (k) is transformed to Fi (q) by the nuclear effect parametrized by βL, while fi (k) itself contains the hidden modifications due to such effects as Ncoll dependence and shadowing.74 Since fi (k) is dNihard /kdkdyy=0 , (54) becomes in the limit of L → 0 the invariant distribution for hard parton i production in pp collision. In heavyion collisions Eq. (11), written for centrality ∼ 0%, can now be improved to the form Z dq S(q1 ) = Fi (q)Sij (q1 /q) (55) q for any centrality with L being the geometrical path length. We can now write the TS recombination term in (48) as Z Z 0 dq1 j q1 c p) = Si dq2 C¯j e−q2 /T Rj¯j (q1 , q2 , p) , (56) TS(q, q1 q where Rj¯j is the RF given in (16).
304
R. C. Hwa
Equation (48) is now totally specified, relating the observable pion spectrum to the nuclear parameter βL through Fi (q) given in (54). We emphasize that because all three components of TT+TS+SS recombination are included in (48) it can describe the pion distribution at all pT using only one parameter βL for each centrality. Figure 11(a) shows the fits of the π 0 distributions for nine bins of centrality,146 with a different value of βL used for each centrality.144 Those values of βL are shown by the nine points in Fig. 11(b) (ignoring the line for the moment). Evidently, the agreement with data is excellent over such a wide range of pT and centrality c. Apart from concluding that the RM works well, the jetmedium interaction has been effectively summarized by one phenomenological function βL(c), which could not have been extracted from the data without a reliable way to relate energy loss to the suppression of pion production at all pT through appropriate hadronization. Having obtained βL(c) we now go a step further to inquire what kind of variation of the dynamical path length can the nuclear overlap generate for any given centrality. That is, for a fixed c, βL(c) is the average of a variable ξ over a probability distribution P (ξ, c) that describes the likelihood that a particular trajectory occurs at centrality c with an effective energy loss such that Z βL(c) = dξξP (ξ, c). (57) Thus ξ plays the role of βL except that it can vary from 0 to a maximum for every fixed initial elliptic geometry depending on the initial point and orientation of the
(b) (a)
(c)
Fig. 11. (a) Pion spectra146 for various centralities lowered by a factor of 0.2 for each step of increase of 10%. The solid lines are fits in the RM144 with one parameter for each centrality c. ¯ that is the average Those parameters are shown by dots in (b), which are fitted by the curve ξ(c) of the dynamical path length ξ over P (ξ, c) defined in Eq. (58). (c) ξ distributions for various c.
Hadron Correlations in Jets and Ridges Through Parton Recombination
305
¯ hard parton. So ξ is the dynamical path length of a trajectory whose average ξ(c) is βL(c). In Ref. 144 P (ξ, c) is chosen to have the form P (ξ, c) = N ξ(ξ0 − ξ)αc ,
(58)
where N normalizes the total probability to 1. The two parameters ξ0 and α are adjusted to fit the nine points in Fig. 11(b), with the result shown by the curve that renders an excellent fit for ξ0 = 5.42,
α = 15.2 .
(59)
Thus (58) is a very efficient way to describe the energy loss effect for all centralities. The shapes of P (ξ, c) are shown in Fig. 11(c) for 6 values of c, exhibiting the expected peak that decreases with increasing c or shrinking ellipse. In view of the difficulty of deriving βL(c) from first principles, let alone P (ξ, c), it is very convenient to have the pathdependent quenching effect be represented by the simple description in (58) and (59). The singlepion distribution at midrapidity for centrality c can now be expressed as Z dNπ (c, ξ) dNπ (c) = dξP (ξ, c) , (60) pdp pdp where (48) is to be used for dNπ (c, ξ)/pdp, provided that Fi (q) in it is not as given in (54), but with βL replaced by ξ, and, of course, fi (q) scaled by Ncoll (c). It should be clear that (48) has basically two parts: Fi (q) that describes the hard parton part and the rest that describes the hadronization part. We have in this subsection incorporated the centralitydependent energyloss effect on Fi (q) by the use of just two dimensionless parameters given in (59). Having successfully formulated the treatment of the singlehadron distribution, we are now ready to proceed to the study of dihadron correlation in near and awayside jets. 6.2. Nearside and awayside yields per trigger For dihadron correlation many momentum vectors of partons and hadrons are involved. To depict clearly their relationships with one another, we show in Fig. 12 a pictorial representation of all of them. The near side is on the right and the away side on the left. The vectors k, q, and pt are, respectively, the momenta of the initiating hard parton, of the same parton as it leaves the medium, and of the trigger hadron. The associated hadron on the same side is labeled by pa . On the away side the corresponding momenta are k 0 , q 0 , and pb , there being no trigger on that side. If there are no transverse momenta in the beam partons, k and k 0 should be equal and opposite in every event; however, their averages over triggered events may not be the same, so it is better to label them distinctly from the start. The momenta of interest in the following are such that the hadronic pT are in the range 2 < pT < 10 GeV/c but with special emphasis on pT < 6 GeV/c.
306
R. C. Hwa
Fig. 12. (Color online) A sketch of momentum vectors of partons (in red) and hadrons (in blue) with near side being on the right and away side on the left.
For two pions in the same jet, neither of which are in the ridge, we can leave out the TT contribution, and write as a generalization of (48) X Z dq pt π pt dNππ (ξ) 1 c c − p t , pa ) TS(q = Fi (q, ξ) TS(q, pt ) + Di pt pa dpt dpa (pt pa )2 i q q q ) p p p p p p a t a a t a π π c − p a , pt ) D , +TS(q + 2 D2 , (61) q i q qj q q
where D2 (z1 , z2 ) is the dihadron fragmentation function.144 For notational brevity (61) is for a fixed ξ, the averaging over P (ξ, c) being a process that can be applied when c is fixed. For two pions on opposite sides the recoil parton must be considered explicitly, so (49) should be generalized to ξ
Fi0 (q, q 0 , ξ) =
Z
=
1 ξ
0
Z
βdt ξ qeξ q
Z
dkkfi (k)G(q, k, t)G(q 0 , k 0 ,
dkkfi (k)qq 0 δ(qq 0 − kk 0 e−ξ ) .
ξ − t) β (62)
with k 0 = k. Thus the dipion distribution for pb on the side away from pt is X Z dq dq 0 1 dNππ (ξ) 0 0 c p t ) + pt D π pt = F TS(q, (q, q , ξ) pt pb dpt dpb (pt pb )2 i q q0 i q i q c 0 , pb ) + pb Dπ0 pb . (63) × TS(q q0 i q0
The δ function in (62) restricts the integration of q 0 in (63) to the range from qe−ξ to qeξ that correspond to the hard scattering point being on the nearside boundary to the farside boundary. The yield per trigger at fixed centrality can now be obtained for the near side as R dξP (ξ, c)dNππ (ξ)/pt pa dpt dpa near R Yππ (pt , pa , c) = , (64) dξP (ξ, c)dNπ (ξ)/pt dpt
Hadron Correlations in Jets and Ridges Through Parton Recombination
(a)
307
(b)
Fig. 13. (a) Nearside jet yield per trigger vs trigger momentum pt and associated particle momentum pa . (b) Awayside jet yield per trigger vs pt and pb . away and similarly for Yππ (pt , pb , c) with dNππ (ξ)/pt pb dpt dpb being used in the numerator. Since every factor is already specified, it remains only for the computation to be carried out. The results are shown in Fig. 13(a) for near side and (b) for away side. The three sheets for c = 0.05, 0.35, and 0.86 are separated by a factor of 10 between sheets for clarity’s sake. Superficially, the two figures may look similar in a vertical scale that spans over 4 orders of magnitude, but there are significant differences that only by closer examination can one learn from them the nature of the medium effects on the jets.
6.3. Medium effects on dijets Let us first cut the two figures in Fig. 13 by three fixedpt planes at pt = 4, 6, and 8 GeV/c, and show the results in Fig. 14(a) and (b) for c = 0.05 and 0.35. Note (i) the near dependence of Yππ (pa ) on pa is more sensitive to pt than that of the dependence away of Yππ (pb ) on pb ; (ii) the increase of the yield with pt is more pronounced for the near near side than for the away side; and (iii) Yππ has negligible dependence on c, but away Yππ increases by roughly a factor of 2 when c changes from 0.05 to 0.35. Let us discuss these three features separately. On item (i) one can determine the nearside average inverse slope Ta of the approximate exponential behavior in the range 2 < pa < 4 GeV/c for c = 0.05, for which there are data. The result is shown by the line in Fig. 15(a), and agrees well with the data.71,101,147 Evidently, the spectrum of the associated particles in a jet becomes harder as the trigger momentum increases, as one would expect. It should be remarked that the line in Fig. 15(a) is for pions, while the data are for all charged particles. However, since pions dominate in jet peaks (unlike the ridges), the comparison is not unreasonable. The inverse slope Tb for the awayside jet at c = 0.05 for the same ranges of pt and pb is shown in Fig. 15(b), exhibiting a lower value compared to Ta and a mild decrease with pt . There exist no suitable data
308
R. C. Hwa
Fig. 14. (a) Left panel: nearside jet yield per trigger vs pa for fixed pt and c. Data points are from Ref. 6.11. (b) Right panel: the same for away side vs pb .
(a)
(b)
Fig. 15. Inverse slopes of associated particles on the (a) near side and (b) away side. Data in (a) are from Refs. 101 and 147.
for comparison. The strong difference between Ta and Tb has good reasons, as will be discussed below. Two data points from Ref. 148 are included in Fig. 14(a), the details of which are discussed in Ref. 144. They lend support to the theoretical curve in both magnitude and pa dependences. Note (ii) about the pt dependence of the yield is related to note (i) about the near away pt dependencies of the shapes in pa and pb . First of all, Yππ is larger than Yππ in magnitude, meaning that there is more suppression on the away side than on the near side. To quantify that interpretation let ξ be fixed at 2.9 corresponding to the maximum probability for c = 0.05 so that the suppression effect is not partially hidden by averaging over ξ, which can vary from 0 to ξ0 . Let the suppression factor for the near side be defined by Γnear (pT ) = he−βt ipT = hq/kipT ,
(65)
Hadron Correlations in Jets and Ridges Through Parton Recombination
309
where the average is performed over dNπ /pT dpT given in (48), and the last expression follows from (53). Γnear (pT ) gives a measure of the fraction of momentum retained after energy loss reduces k to q. For the away side the suppression factor is defined by Γaway (pt , pb ) = he−ξ+βt ipt ,pb ,
(66)
where −ξ + βt is equivalent to −β(L − t) if ξ is denoted by βL for a fixed medium length L, in which t is the portion from the hardscattering point to the nearside surface, and L − t is the distance to the awayside surface. The average in (66) is done using the dihadron distribution given in (63).
(a)
(b)
Fig. 16. Suppression factors for the (a) near side and (b) away side.
The calculated results for Γnear (pT ) and Γaway (pt , pb ) are shown in Figs. 16(a) and 16(b). They clearly indicate that Γnear (pT ) is much larger than Γaway (pt , pb ) with the implication that there is less suppression on the near side than on the away side. The physics of the phenomenon is clear: at large pT the point of creation of hard parton with large k is predominantly close to the nearside surface in order to minimize energy loss with the consequence that the distance to the awayside surface is longer, thus more suppression for pb on that side. Since Γnear (pT ) saturates at around 0.85, only 15% of the parton energy is lost to the medium on the near side. The corresponding hβti is less than 0.2, so hβti/ξ = hti/L ≈ 0.065, meaning that the hard partons are created within a layer of thickness ∼ 13% of L from the surface. That is a quantitative description of trigger bias. On the other hand, the behavior of Γaway (pt , pb ) reveals the opposite: at fixed pb it decreases with increasing pt , implying more suppression as the hardscattering point is pulled closer to the nearside surface. That is antitrigger bias. At fixed pt , Γaway (pt , pb ) increases with pb , since higher pb demands higher q 0 , which can be satisfied only if hL − ti is reduced or hk 0 i increased, actually both. Lowering hL − ti is, of course, a way to
310
R. C. Hwa
reduce energy loss by pulling the scattering point closer to the awayside surface. But the increase of hk 0 i with pb is another aspect of antitrigger bias, whose details are described in Ref. 144. Eventbyevent momentum conservation requires k 0 = k; however, when averaged over all events, hk 0 i depends on pb , while hki does not. In general, hk 0 i is far greater than hki because any trigger favors shorter hti, and any finite pb pushes up hk 0 i/hki though not hk 0 /ki. Because of the difficulty of producing an associated particle on the away side relative to one on the near side, Tb is lower and decreases with increasing pt in Fig. 15(b), while Ta is higher and increases with pt in Fig. 15(a). Now on note (iii) about centrality dependence Fig. 14(b) shows that the depenaway dencies of Yππ (pt , pb , c) on pt and pb are essentially the same whether c = 0.05 or 0.35, except that the magnitude of the yield is increased due to the reduction of path length at higher c. For the near side there is essentially no dependence of near Yππ (pt , pa , c) on c. That is shown more explicitly in Fig. 17, where the lower three lines are for pt = 4 GeV/c and the upper line is for pt = 6 GeV/c. The solid lines are for integrated yields with 2 < pa < 4 GeV/c. The near independence on c is a manifestation of the trigger bias, since the hardparton production point, being restricted to a layer roughly 13% of L just inside the nearside surface, is insensitive to how large the main body of the medium is. Actually, the decrease of the TS component with c balances the increase of the SS component with c so that the net yield being their sum is approximately constant in c. The data points in Fig. 17 support the calculated result in both the magnitude and the c dependence near of Yππ (pt , c) when pa is integrated from 2 to 4 GeV/c.101 In summary the discussion above gives quantitative demonstration that the trigger bias is the preference for the hardscattering point to be close to the nearside surface and that the antitrigger bias is the consequence: hk 0 i is much larger
Fig. 17. Nearside yield per trigger vs centrality c. The solid lines are for integrated yields with 2 < pa < 4 GeV/c. The data points are from Ref. 101.
Hadron Correlations in Jets and Ridges Through Parton Recombination
(a)
311
(b)
Fig. 18. (a) The ratio hq 0 i/hk 0 i at the symmetry point p = pt = pb for four values of centrality. (b) Distribution of associated pion (pb ) in the awayside jet for six values of pion trigger momentum (pt ) in GeV/c for ξ fixed at 2.9.
than hki, and even larger than pt or pb . Those are the properties of the events selected by a trigger at pt with an associated particle on either the near or away side.
6.4. Symmetric dijets and tangential jets We now follow the summary comment at the end of Sec. 6.3 with the question on what if we select events with symmetric dijets where pt = pb . Instead of studying the properties of a third particle in association with the two trigger particles (for which we need trihadron correlation function), we can nevertheless learn a great deal from examining closely various calculable quantities in the dihadron correlation problem. Let p be the momentum of the symmetric dijets, p = pt = pb . One can calculate hk 0 i(p, c) and hq 0 i(p, c) for various centralities, even though they are not directly measurable. The result is that they both increase almost linearly with p and that there is essentially no dependence on c. Their ratio hq 0 i/hk 0 i is therefore approximately constant in p with a value of about 0.8 as shown in Fig. 18(a). That behavior is similar to the property of Γnear (p), shown in Fig. 16(a), which, according to (65), is also hq/ki. It is important to bear in mind that hq 0 i and hk 0 i are averages over the twoparticle distribution dNππ (c)/pt pb dpt dpb with pt = pb , given in (63) for fixed ξ, followed by averaging over P (ξ, c), whereas hqi, hki and hq/ki are averages over the singleparticle distribution dNπ /pdp, as noted after (65). The nearside averages know nothing about the awayside analysis, so Γnear (p) describes only the suppression associated with trigger bias. The fact that the suppression on the away side is about the same as on the near side, when the average hq 0 /k 0 i is over symmetric dijets and hq/ki is averaged over nearside jet, and that it is true for any centrality, has only one important implication: the dijets are created very near the surface on both sides. It means that the symmetric dijets are dominated
312
R. C. Hwa
by tangential jets, which behave similarly at all centralities. That is a striking conclusion. Hard scattering can, of course, occur anywhere in the overlap region. But those partons from the interior lose most of their momenta on the way out. Those that are created on the near side would not give rise to symmetric dijets. Thus only those created along the rim and directed tangentially to the surface can lead to events with pt = pb . There is some experimental evidence to support this finding. In 2jet + 1 correlation studied149 it is found that the third particle does not show any ridge structure 2/3 and that the centrality dependence goes as Npart . The latter means that the dijets are created near the surface, while the former means that the dijets are tangential because we have already seen in Sec. 4.4 that ridge production depends on the local flow direction to match the trigger direction due to the correlation function in (27). The flow direction near the surface is normal to the surface and is therefore normal to tangential jets. In the experimental analysis149 the two jets have pT cuts at pT1 > 5 GeV/c and pT2 > 4 GeV/c, so they are not exactly symmetrical. If ridge formation from those dijets is to be discovered in the future, it would pose a serious challenge to the treatment of ridgeology in Sec. 4. But so far the data are in accord with the results both here and on ridges. 6.5. Unsymmetric dijets and tomography The conclusion in the last section on the dominance in symmetric dijets by tangential jets leads one to consider the only option left for probing the dense medium by parton jets apart from using direct γ, and that is the study of awayside jets in unsymmetric dijets. If hard or semihard partons created near the surface are most responsible for the nearside jet, then the awayside jet should experience fully the effect of the medium. That is indeed the case when one calculates hq 0 i and hk 0 i for various values of pt and pb at fixed ξ. It is found in Ref. 144 that for ξ = 2.9, hq 0 i/hk 0 i is in the range of 0.15 to 0.3 for 4 < pt < 10 GeV/c and 2 < pb < 6 GeV/c, and gives a quantitative measure of the degree of energy loss. Unfortunately, that cannot be checked experimentally, since only centrality can be selected in realistic collisions, not the dynamical path length ξ. At fixed centrality the value of ξ can fluctuate over a wide range, as can be ¯ is 2.9, but the characteristic of the seen in Fig. 11(c). For c = 0.05 the average ξ(c) awayside jet is dominated by the lower ξ portion of the range. The consequence away is that Yππ (pt , pb , c) does not depend sensitively on pt , as Fig. 14(b) has already indicated for any fixed c. That is in sharp contrast from the case where ξ is fixed away at 2.9, and Yππ (pt , pb , ξ) decreases by an order of magnitude as pt is decreased from 9 to 4 GeV/c at any fixed pb , as shown in Fig. 18(b). That diminishing yield is because hki is lower at lower pt and the energy loss by the recoil parton traversing the thick medium results in reduced probability of producing a pion at fixed pb . That is not the case in Fig. 14(b). At fixed c, the decrease of pt does not lead to away significantly lower Yππ (pt , pb , c) because the hardscattering point is already in a
Hadron Correlations in Jets and Ridges Through Parton Recombination
313
Fig. 19. (a) Left panel: average dynamical path length of recoil parton directed toward the away side. (b) Right panel: ratio of the average parton momenta of recoil parton at the surface to that at the creation point.
region of lower ξ to minimize energy loss. Lowering pt increases the yield at fixed pb due to softer parton, but the number of triggers is also higher, so the yield per trigger remain nearly unchanged. In other words, allowing ξ to be small even at small c removes energy loss as a decisive factor in the problem. That is what makes dijet tomography ineffective as a probe to learn about a medium that has no fixed thickness. Further insight can be gained by studying hβt0 i at fixed c, where t0 denotes the distance from the hardscattering point to the awayside surface. Figure 19(a) shows hβt0 i vs pb for various pt and c.144 A general impression from that figure is that hβt0 i is low, less than approximately 0.4. That is much lower than βL = 2.9, which is the average dynamical path length determined from fitting the singlepion distribution from c = 0.05 [see Fig. 11(b)]. The height of that distribution at large pT [see Fig. 11(a)] is what renders RAA ≈ 0.2, a number that pQCD calculations aim to obtain. The fit in Ref. 144 is achieved by setting βL = 2.9 in order to obtain the correct normalization for dNπ /pt dpt at large pT , for which the contributions from all partons, near or far, hard or semihard, are counted. The result that hβt0 i βL shown in Fig. 19(a) is therefore indicative of the fact that conditional probability with pt and pb fixed is highly restricted compared to the inclusive probability. With ht0 i being much less than L at fixed c, one is led to conclude that the unsymmetric dijets are also produced near the surface and are essentially tangential, as with symmetric dijets. Thus the medium interior is not probed. This conclusion is distinctively different from the case where ξ is fixed. Figure 16(b) has shown that the suppression on the away side can be large (hence Γaway small) for unsymmetric dijets at ξ = 2.9. In that case hβt0 i would not be small, as we now have for any fixed c.
314
R. C. Hwa
Experimental data149 on slightly asymmetric dijets at high pT show no essential difference in the structures of the two jets and are in support of the findings in Ref. 144. Another way to come to the same conclusion is to study the energy loss of the awayside jet. In Fig. 19(b) is shown hq 0 i/hk 0 i for c = 0.05 and 0.35.144 It is the fraction of average momentum retained by the hard or semihard parton after traversing the medium on the away side. Being around 0.8 implies that only about 20% of the parton momentum is lost to the medium, not much more than the fractional energy loss on the near side. At fixed pt Fig. 19(b) shows an increase of hq 0 i/hk 0 i with increasing pb because the hard scattering point is pulled more to the away side, but it shows a decrease of the ratio at increasing pt since the point is then pulled to the near side. This pushandpull effect of pt and pb is clearly what one expects in the oppositelydirected jets when the path lengths on the two sides are comparable. At the symmetry point pt = pb = 4 GeV/c, hq 0 i/hk 0 i for c = 0.05 is only slightly lower than its value for c = 0.35, implying strongly that the fractional energy loss on both sides remains about the same regardless of centrality. That can only mean that the hard partons are created near the surface and directed tangentially. Making pt 6= pb in unsymmetric dijets does not change hq 0 i/hk 0 i drastically. Thus so long as the medium thickness cannot be controlled, there seems to be no useful tomography that can be done with partoninitiated dijets. If the awayside jets are dominated by those created near the awayside surface (such as the tangential jets), then the events triggered by direct γ on the near side are also likely to be dominated by those where the hard scattering takes place near the awayside surface (not just tangential) so long as an associated particle with significant pb is required on the away side. The πtriggered and γtriggered distributions, IAA , of the associated particles should be roughly the same. There is some experimental evidence for that similarity.150 It is important to note that the above comment is for tomography only, i.e., medium effect on jets. The jet effect on medium, such as Mach cone, is a difference matter that depends on different physics and may well reveal properties of the medium interior. 7. Conclusion In this review many problems in heavyion collisions have been examined in many parts of the phase space. The good agreement between theoretical calculation and experimental data in almost all cases cannot but affirm that the successes of the theoretical approach adopted cannot be all fortuitous and that the interpretations given to the physical origins of the measured phenomena are not without some degree of realism. In some cases there are no apparent alternative schemes to explain the data. Of course, a phenomenological model is not a theory based on first principles, but its usefulness should not be overlooked when its scope is out of reach by any theory commonly accepted.
Hadron Correlations in Jets and Ridges Through Parton Recombination
315
In the first part of this review, mainly in Secs. 2 and 3, we have summarized the essential basis of the recombination model, answering some questions raised by critics, and pinning down some of the parameters, while pushing the frontier to kinematical regions no other models have attempted, and establishing parton recombination as a universal hadronization mechanism. Then in the second and main part recombination is used in an essential way to relate what are observed about jets and ridges to two complementary aspects about the dense medium: the effects of hard or semihard partons on the medium and the converse. Those aspects of the medium effects cannot be made empirically relevant unless there is a reliable description of hadronization for all pT where correlation data exist. Some of the new insights gained that are of particular current interest are: (a) (b) (c) (d) (e)
correlation between trigger and ridge (Sec. 4.4), independence of φ anisotropy on fast thermalization (Sec. 5.1), recombination does not guarantee quark number scaling (Sec. 5.3), different properties of the near and awayside jets (Sec. 6.3), dominance of tangential jets in symmetric and unsymmetric dijets (Secs. 6.4 and 6.5).
If there is to be one unifying conclusion to be made as a result of these findings, it is that most observables on jets and ridges are due to hard or semihard partons created near the surface. Partons created in the interior of the medium that lead to dijets are not able to compete effectively with those that have shorter distance to traverse. So long as the observables allow the parton creation points to include regions that offer the partons a choice of paths of least resistance, they will take it and dominate. That is the reality faced by experiments that can only fix centrality, not medium thickness. With that recognition the efficacy of jet tomography is called into question. Unlike Xray scanning of organic or inorganic substances, there is no control of the sources of the hard partons, so the necessary averaging process underweighs the contribution from the region of the medium that one wants to learn most about. The above comment refers to partoninitiated dijets. Of course, for single hadron at large pT the nuclear modification factor RAA has long been used as a measure of energy loss in dense medium in experiments and in theory. As soon as a condition is imposed on the detector of another particle on the away side, the region of the system probed in changed. If that hadron’s pT is low, it may be in the doublehumped shoulder region, which does provide some information on what the effect of the awayside jet is on the medium. However, to learn about the effect of the main body of the medium on the jet from dihadron correlation is more difficult. Ridgeology addresses a different set of problems, quite distinct from dijets. Ridges are stimulated by jets, but are not a part of the jets that are characterized by the participation of shower partons. Considerable attention has recently been drawn to the study of ridges. At this point there is no consensus in their theoretical interpretation. The connection between the ridges found in triggered
316
R. C. Hwa
events101,151 and those found in autocorrelation without triggers103,152 is in our opinion tenuous, until the pT dependence of the ridge is clarified. Minijets153 peaking at yt ∼ 2.8 correspond to hpT i ∼ 1.2 GeV/c, which is much lower than the pT range of the triggers used in ridge analysis. The ridge found in autocorrelation has no pT cut, while those found in triggered events have passoc > 2 GeV/c.101 T Nevertheless, there may exist a connection between autocorrelation and the ridges without trigger discussed in Sec. 4.2. Another phenomenon of some current interest is the observation of extended ridge at large ∆η.151 Before one concludes that such a longrange correlation can only arise from the mechanism of strings (or color flux tubes) being stretched between forward and backwardgoing quarks, it seems prudent to allow firstly the possibility of other types of earlytime dynamics, and secondly a broader view of the problem of particle production at large η. In particular, one should consider the role played by the hard parton that leads to both the trigger and the ridge. Moreover, one should consider the issue of large p/π ratio at η = 3.2 (see Sec. 3.4) as a part of the solution of the bigger problem. If the observed large p/π ratio is a feature of the finalstate interaction (FSI), one should not regard the ridge phenomenon as a manifestation of the initialstate interaction (ISI) only without taking into consideration also the effects of FSI. As we have noted at a number of places throughout this review, hadronization is an important link between the observables and partonic dynamics. The characteristic of ridge formation in azimuthal correlation discussed in Sec. 4.4 is a good example of the interplay between ISI and FSI. It would be surprising if the same does not hold true for the correlation in rapidity. The considerations given in this review to jets and ridges may not be directly relevant at LHC where jets are copiously produced — unless pT is extremely high. At RHIC the background to a rare jet is thermal, but at LHC the background to a highpT jet, say at pT ∼ 100 GeV/c, includes many other lowerpT jets. The admixture of thermal and shower partons in the background introduces new complications to the notion of enhanced thermal partons, and renders what is simply conceived for RHIC inadequate for LHC. That new energy frontier will indeed open up a wide new horizon. Note Added in Proof A recent study of the azimuthal and centrality dependencies of jet production has shed more light on the subject discussed in Sec. 6.1 and therefore on the nuclear modification factor RAA .154 The dynamical path length ξ considered in (57) and (58) is averaged over φ and its properties are determined phenomenologically as shown in Fig. 11. In Ref. 154 a theoretical study of the dynamical path length for every given pair of φ and c is carried out, taking into account the medium density along the hard parton’s trajectory as well as the probability of producing a hard parton at the creation point. Although the same symbol ξ is used, it is a very different quantity describing the degradation of the parton momentum k by
317
Hadron Correlations in Jets and Ridges Through Parton Recombination
10
R
−1
10
10−20% 20−30% 30−40% 40−50% 50−60%
PHENIX
0
R
AA
10
10−20% 20−30% 30−40% 40−50% 50−60%
AA
PHENIX
0
0
pT=4−5 GeV/c 0.2
−1
0.4 ξ 0.6
0.8
1
10
0
pT=7−8 =4−5 = − GeV/c 0.2
(a)
0.4 ξ 0.6
0.8
1
(b)
Fig. 20. Experimental data on RAA from Ref. 155 for (a) pT = 4–5 GeV/c and (b) pT = 7– ¯ Points having the same symbol are for φ = nπ/24 8 GeV/c, plotted against the scaling variable ξ. with n = 1, 3, . . . , 11 from left to right. The solid line is the result of theoretical calculation. 154
1
2
LHC p =10 GeV/c
0.8
AA
R
RAA
0.4
1
0.5
0.2 0 0
T
1.5
Γ=0.001
0.6
p =10 GeV/c
LHC
T
Γ=0.1 0.2
0.4 ξ 0.6 (a)
0.8
1
0 0
0.2
0.4
ξ
0.6
0.8
1
(b)
Fig. 21. Theoretical on RAA at LHC, plotted against the scaling variable ξ¯ for two possible values of the 2jet overlap factor Γ: (a) Γ = 10−3 and (b) Γ = 10−1 . Symbol are the same as in Fig. 20 for various values of c and φ. results154
q = ke−ξ . What is noteworthy is that there is a scaling behavior of RAA (pT , φ, c) ¯ c), i.e., the dependence on φ and c exhibits when expressed in terms of the mean ξ(φ, ¯ a universal behavior in ξ. The PHENIX data155 on π 0 RAA are shown in Fig. 20 for two pT ranges and five bins of centrality; each set of symbols has six evenlyspaced φ values from 0 to π/2, appearing from left to right in the order of increasing φ. The solid lines are the results of the theoretical calculation in the recombination π model. An approximate but simple description of the scaling behavior is RAA ≈ 154 ¯ exp(−2.6ξ). A similar behavior is found also in the awayside yield per trigger for backtoback dijets. On the subject of recombination of adjacent jets at LHC discussed in Sec. 3.5, further study has shown that pion production at pT ∼ 10 GeV/c due to 2jet recombination can lead to spectacular deviation from the prediction of 1jet fragmentation. Figure 21 shows RAA at LHC for two different values of the 2jet overlap
318
R. C. Hwa
factor Γ.154 When Γ = 10−3 , hadronization from single jets dominates and a scaling behavior in ξ¯ is found. But when Γ = 10−1 , 2jet recombination dominates and ¯ ξscaling is broken. Moreover, the value of RAA can exceed 1, which would be a striking signal for the hadronization process. Acknowledgment I am grateful to many who have helped me in the studies described here, but especially to Charles Chiu and Chunbin Yang. This work was supported, in part, by the U.S. Department of Energy under Grant No. DEFG0292ER40972. References 1. See many papers on jetmedium interaction presented at Quark Matter 2008, Jaipur, India (2208), in particular the plenary talk by J. Jia (for Phenix Collaboration), J. Phys. G: Nucl. Part. Phys. 35 (2008) 104033. See also J. Jia, Int. J. Mod. Phys. E 16 (2007) 3058. 2. M. van Leeuwen, J. Phys. G: Nucl. Part. Phys. 34 (2007) S559; Int. J. Mod. Phys. E 16 (2007) 3049. 3. X. N. Wang and M. Gyulassy, Phys. Rev. D 44 (1991) 3501. 4. X. N. Wang and M. Gyulassy, Phys. Rev. Lett. 68 (1992) 1480. 5. C. Adler et al., (STAR Collaboration), Phys. Rev. Lett. 89 (2002) 202301. 6. S. S. Adler et al., (PHENIX Collaboration), Phys. Rev. Lett. 91 (2003) 072301; Phys. Rev. C 69 (2004) 034910. 7. I. Arsene et al., (BRAHMS Collaboration), Nucl. Phys. A 757 (2005) 1. 8. B. B. Back et al., (PHOBOS Collaboration), Nucl. Phys. A 757 (2005) 28. 9. J. Adams et al., (STAR Collaboration), Nucl. Phys. A 757 (2005) 102. 10. K. Adcox et al., (PHENIX Collaboration), Nucl. Phys. A 757 (2005) 184. 11. L. Ruan (for STAR Collaboration), J. Phys. G: Nucl. Part. Phys. 34 (2007) S199. 12. F. Wang (for STAR Collaboration), J. Phys. G: Nucl. Part. Phys. 34 (2007) S225. 13. N. Armesto, C. A. Salgado and U. A. Wiedemann, Phys. Rev. Lett. 93 (2004) 242301; Phys. Rev. C 72 (2005) 064910. 14. S. A. Voloshin, Nucl. Phys. A 749 (2005) 287; Phys. Lett. B 632 (2006) 490. 15. C. B. Chiu and R. C. Hwa, Phys. Rev. C 72 (2005) 034903. 16. P. Romatsche, Phys. Rev. C 75 (2007) 014901. 17. A. Majumder, B. M¨ uller and S. A. Bass, Phys. Rev. Lett. 99 (2007) 042301. 18. E. V. Shuryak, Phys. Rev. C 76 (2007) 047901. 19. C. Y. Wong, Phys. Rev. C 76 (2007) 054908. 20. I. Vitev, Phys. Lett. B 630 (2005) 78. 21. C. B. Chiu and R. C. Hwa, Phys. Rev. C 74 (2006) 064909. 22. R. C. Hwa, Int. J. Mod. Phys. E 16 (2007) 3176. 23. A. D. Polosa and C. A. Salgado, Phys. Rev. C 75 (2007) 041901 (R). 24. I. Dremin, JETP Lett. 30 (1979) 140; Nucl. Phys. A 767 (2006) 233. 25. V. Koch, A. Majumder and X. N. Wang, Phys. Rev. Lett. 96 (2006) 172302. 26. J. CasalderreySolana, E. V. Shuryak and D. Teaney, J. Phys. Conf. Ser. 27 (2005) 22; hepph/0602183. 27. H. Stoecker, Nucl. Phys. A 750 (2005) 121. 28. T. Renk and J. Ruppert, Phys. Rev. C 73 (2006) 011901; Phys. Lett. B 646 (2007) 19.
Hadron Correlations in Jets and Ridges Through Parton Recombination
29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64.
65. 66. 67.
319
V. S. Pantuev, JETP Lett. 85 (2007) 104. A. Majumder, B. M¨ uller and X. N. Wang, Phys. Rev. Lett. 99 (2007) 192301. J. CasalderreySolana, J. Phys. G: Nucl. Part. Phys. 34 (2007) S345. R. C. Hwa and C. B. Yang, Phys. Rev. C 67 (2003) 034902. V. Greco, C. M. Ko and P. L´evai, Phys. Rev. Lett. 90 (2003) 202302; Phys. Rev. C 68 (2003) 034904. R. J. Fries, B. M¨ uller, C. Nonaka and S. A. Bass, Phys. Rev. Lett. 90 (2003) 202303; Phys. Rev. C 68 (2003) 044902. R. C. Hwa and C. B. Yang, Phys. Rev. C 70 (2004) 024905. C. Y. Wong, Chin. Phys. Lett. C 25 (2008) 3936; Phys. Rev. C 78 (2008) 064905. A. Dumitru, F. Gelis, L. McLerran and R. Venugopalan, Nucl. Phys. A 810 (2008) 91. S. Gavin and G. Moschelli, J. Phys. G: Nucl. Part. Phys. 35 (2008) 104084. S. Gavin, L. McLerran and G. Moschelli, arXiv:0806.4718. R. Mizukawa, T. Hirano, M. Isse, Y. Nara and A. Ohnishi, J. Phys. G: Nucl. Part. Phys. 35 (2008) 104083. R. C. Hwa, J. Phys. G: Nucl. Part. Phys. 35 (2008) 104017. B. Betz, M. Gyulassy, D. H. Rischke, H. St¨ ocker and G. Torrieri, J. Phys. G: Nucl. Part. Phys. 35 (2008) 104106. G. Torrieri, B. Betz, J. Noronha and M. Gyulassy, Acta Phys. Polo. B 29 (2008) 3281. T. Renk, Phys. Rev. C 78 (2008) 014903. J. Liao and E. Shuryak, arXiv:0810.4116. B. M¨ uller and R. B. Neufeld, J. Phys. G: Nucl. Part. Phys. 35 (2008) 104108. R. B. Neufeld, B. M¨ uller and J. Ruppert, Phys. Rev. C 78 (2008) 041901. R. B. Neufeld and B. M¨ uller, arXiv:0902.2950. K. Dusling, D. FernandezFraile and R. Venugopalan, arXiv:0902.4435. J. Jia and R. Lacey, Phys. Rev. C 79 (2009) 011901. A. Majumder, plenary talk given at Quark Matter 2009, Knoxville, TN. R. D. Field, Applications of Perturbative QCD (AddisonWesley, 1989). Ya. I. Azimov, Yu. L. Dokshitzer, V. A. Khoze and S. I. Troyan, Z. Phys. C 27 (1985) 65. A. Capella, U. Sukhatme, C.I. Tan and J. Tran Thanh Van, Phys. Rep. 236 (1994) 225. R. P. Feynman, Phys. Rev. Lett. bf 23 (1969) 1415. K. P. Das and R. C. Hwa, Phys. Lett. 68B (1977) 459. R. C. Hwa, Phys. Rev. D 22 (1980) 759. R. C. Hwa, Phys. Rev. D 22 (1980) 1593. R. C. Hwa and C. B. Yang, Phys. Rev. C 66 (2002) 025204. CTEQ4LQ, http://zebu.uoregon.edu/∼parton/partongraph.html R. C. Hwa and C. B. Yang, Phys. Rev. C 66 (2002) 025205. R. C. Hwa and C. B. Yang, Phys. Rev. C 70 (2004) 024904. R. C. Hwa and C. B. Yang, Phys. Rev. C 73 (2006) 064904. B. A. Kniehl, G. Kramer and B. P¨ otter, Nucl. Phys. B 582 (2000) 514; D. de Florian, M. Stratmaan and W. Vogelsang, Phys. Rev. D 57 (1998) 5811; S. Albino, B. A. Kniehl and G. Kramer, Nucl. Phys. B 734 (2006) 50. D. W. Duke and F. E. Taylor, Phys. Rev. D 17 (1978) 1788. B. I. Abelev et al., (STAR Collaboration), Phys. Rev. Lett. 97 (2006) 152301. J. Bielcikova (for STAR Collaboration), talk given at Hard Probes 2006, Asilomar, Pacific Grove, CA, Nucl. Phys. A 783 (2007) 565c.
320
R. C. Hwa
68. D. Kharzeev, Phys. Lett. B 378 (1996) 238. 69. S. E. Vance and M. Gyulassy, Phys. Rev. Lett. 83 (1999) 1735. 70. V. Topor Pop, M. Gyulassy, J. Barrette, C. Gale, X. N. Wang and N. Xu, Phys. Rev. C 70 (2004) 064906. 71. J. Bielcikova (for STAR Collaboration), Proc. 23rd Winter Workshop on Nuclear Dynamics (2007), arXiv:0707.3100. 72. B. I. Abelev et al., (STAR Collaboration), Phys. Rev. Lett. 99 (2007) 112301. 73. R. C. Hwa and C. B. Yang, Phys. Rev. C 75 (2007) 054904. 74. D. K. Srivastava, C. Gale and R. J. Fries, Phys. Rev. C 67 (2003) 034903. 75. J. W. Cronin et al., Phys. Rev. D 11 (1975) 3105. 76. A. Accardi and M. Gyulassy, Phys. Lett. B 586 (2004) 244. 77. F. Matathias, J. Phys. G: Nucl. Part. Phys. 30 (2004) S1113. 78. R. C. Hwa and C. B. Yang, Phys. Rev. Lett. 93 (2004) 082302. 79. R. C. Hwa and C. B. Yang, Phys. Rev. C 70 (2004) 037901. 80. L. McLerran and R. Venugopalan, Phys. Rev. D 49 (1994) 2232, 3352; 50 (1994) 2225. 81. E. Iancu and R. Venugopalan, in QuarkGluon Plasma 3, eds. R. C. Hwa and X. N. Wang (World Scientific, Singapore, 2004). 82. I. Arsene et al., (BRAHMS Collaboration), Phys. Rev. Lett. 93 (2004) 242303. 83. J. JalilianMarian, J. Phys. G: Nucl. Part. Phys. 30 (2004) S751. 84. R. C. Hwa, C. B. Yang and R. J. Fries, Phys. Rev. C 71 (2005) 024902. 85. H. L. Lai et al. (CTEQ Collaboration), Eur. Phys. J. C 12 (2000) 375. 86. K. J. Eskola, V. J. Kolhinen and C. A. Salgado, Eur. Phys. J. C 9 (1999) 61. 87. J. Adams et al., (STAR Collaboration), Phys. Rev. C 70 (2004) 064907. 88. B. I. Abelev et al., (STAR Collaboration), Phys. Rev. C 76 (2007) 054903 . 89. B. B. Back et al., (PHOBOS Collaboration), Phys. Rev. Lett. 91 (2003) 052303; Phys. Rev. C 74 (2006) 021901. 90. I. C. Arsene et al. (BRAHMS Collaboration), nuclex/0602018. 91. R. C. Hwa and C. B. Yang, Phys. Rev. C 73 (2006) 044913. 92. R. C. Hwa and C. B. Yang, Phys. Rev. C 76 (2007) 014901. 93. I. C. Arsene (for BRAHMS Collaboration), Quark Matter 2008, talk presented at Quark Matter 2008, Jaipur, India, J. Phys. G: Nucl. Part. Phys. 35 (2008) 104056. 94. N. Katry´ nska and P. Staszel (for BRAHMS Collaboration), poster presentation at Quark Matter 2008, Jaipur, India, arXiv:0806.1162. 95. R. C. Hwa and L. L. Zhu, Phys. Rev. C 78 (2008) 024907. 96. R. C. Hwa and C. B. Yang, Phys. Rev. Lett. 97 (2006) 042301. 97. M. Gyulassy, I. Vitev, X. N. Wang and B. W. Zhang, in Quark Gluon Plasma 3, eds. R. C. Hwa and X. N. Wang (World Scientific, Singapore, 2004). 98. A. Majumder, J. Phys. G: Nucl. Part. Phys. 34 (2007) S377, and see other references cited therein. 99. J. Adams et al. (STAR Collaboration), Phys. Rev. Lett. 95 (2005) 152301. 100. See talks given separately by B. Mohanty, P. K. Netrakanti, R. C. Hwa, J. Jia, E. Wenger, M. P. McCumber, A. Feng, R. Mizukawa, S. Gavin, C. Y. Wong and C. E. Nattrass at Quark Matter 2008, J. Phys. G: Nucl. Part. Phys. 35 (2008) 104001–104167. 101. J. Putschke (for STAR Collaboration), J. Phys. G: Nucl. Part. Phys. 34 (2007) S679, talk given at QM06. 102. C. Suarez, (for STAR Collaboration), poster at Quark Matter 2008, Jaipur, India (2008). 103. J. Adams et al., (STAR Collaboration), Phys. Rev. C 73 (2006) 064907.
Hadron Correlations in Jets and Ridges Through Parton Recombination
321
104. A. Feng, (for STAR Collaboration), talk given at Quark Matter 2008, Jaipur, India (2008), J. Phys. G: Nucl. Part. Phys. 35 (2008) 104082. 105. C. Adler et al., (STAR Collaboration), Phys. Rev. Lett. 90 (2003) 082302. 106. C. B. Chiu, R. C. Hwa and C. B. Yang, Phys. Rev. C 78 (2008) 044903. 107. J. Adams et al., (STAR Collaboration), Phys. Lett. B 612 (2005) 181. 108. S. S. Adler et al., (PHENIX Collaboration), Phys. Rev. C 72 (2005) 014903. 109. J. Adams et al., (STAR Collaboration), Phys. Rev. Lett. 92 (2004) 182301. 110. S. L. Blyth (for STAR Collaboration), J. Phys. G 32 (2006) S461; 34 (2007) S933. 111. B. I. Abelev et al., (STAR Collaboration), Phys. Rev. Lett. 99 (2007) 112301. 112. J. Adams et al. (STAR Collaboration), Nucl. Phys. A 757 (2005) 102. 113. J. Adams et al. (STAR Collaboration), Phys. Rev. Lett. 98 (2007) 062301 . 114. R. C. Hwa and C. B. Yang, Phys. Rev. C 75 (2007) 054904. 115. R. C. Hwa, BNL/RIKEN Workshop on Strangeness in Collisions, February 2006 (unpublished); R. C. Hwa and C. B. Yang, Nuclth/0602024 (unpublished first version of Ref. 114). 116. J. Bielcikova (for STAR Collaboration), J. Phys. G: Nucl. Part. Phys. 34 (2007) S929. 117. R. C. Hwa, J. Phys. G: Nucl. Part. Phys. 34 (2007) S789. 118. C. B. Chiu and R. C. Hwa, Phys. Rev. C 76 (2007) 024904. 119. J. Bielcikova (for STAR Collaboration), Eur. Phys. J. C 49 (2007) 321. 120. A. Adare et al., (PHENIX Collaboration), Phys. Rev. C 78 (2008) 014901. 121. C. B. Chiu and R. C. Hwa, Phys. Rev. C 79 (2009) 034901. 122. P. Netrakanti, parallel talk given at Quark Matter 2009, Knoxville, TN. 123. J. Putschke, plenary talk given at Quark Matter 2009, Knoxville, TN. 124. J. Adams et al., (STAR Collaboration), Nucl. Phys. A 757 (2005) 102; Phys. Rev. C 72 (2005) 014904. 125. K. Adcox et al., (PHENIX Collaboration), Nucl. Phys. A 757 (2005) 184; Phys. Rev. C 69 (2004) 024904. 126. B. I. Abelev et al., (STAR Collaboration), Phys. Rev. C 75 (2007) 054906. 127. A. Adare et al., (PHENIX Collaboration), Phys. Rev. Lett. 98 (2007) 162301. 128. D. Teaney, J. Lauret and E. V. Shuryak, Phys. Rev. Lett. 86 (2001) 4783. 129. P. Houvinen, P. F. Kolb, U. Heinz, P. V. Ruuskanen and S. A. Voloshin, Phys. Lett. B 503 (2001) 58. 130. P. F. Kolb, U. Heinz, P. Houvinen, K. J. Eskola and K. Tuominen, Nucl. Rev. A 696 (2001) 197. 131. P. F. Kolb and U. Heinz, in QuarkGluon Plasma 3, eds. R. C. Hwa and X.N. Wang (World Scientific, Singapore, 2004), p. 634. 132. U. Heinz, arXiv:0901.4355. 133. D. Moln´ ar and S. A. Voloshin, Phys. Rev. Lett. 91 (2003) 092301. 134. V. Greco and C. M. Ko, Phys. Rev. C 70 (2004) 024901. 135. R. C. Hwa, Phys. Lett. B 666 (2008) 228. 136. R. Baier, Y. L. Dokshitzer, A. H. Mueller and D. Schiff, J. High Energy Phys. 0109 (2001) 033. 137. PHENIX data plot, id p0636 (www.phenix.bnl.gov). 138. J. Adams et al., (STAR Collaboration), Phys. Rev. Lett. 95 (2005) 152301. 139. S. S. Adler et al. (PHENIX Collaboration), Phys. Rev. Lett. 97 (2006) 052301. 140. J. Adams et al., (STAR Collaboration), Phys. Rev. Lett. 97 (2006) 162301. 141. A. Adare et al. (PHENIX Collaboration), Phys. Rev. Lett. 98 (2007) 232302. 142. A. Adare et al., (PHENIX Collaboration), Phys. Rev. C 77 (2008) 011901 (R). 143. R. C. Hwa and Z. Tan, Phys. Rev. C 72 (2005) 024908; 72 (2005) 057902.
322
R. C. Hwa
144. 145. 146. 147. 148. 149.
R. C. Hwa and C. B. Yang, Phys. Rev. C 79 (2009) 044908. X.N. Wang, Phys. Lett. B 595 (2004) 165. S. S. Adler et al. (PHENIX Collaboration), Phys. Rev. C 76 (2007) 034904. J. Bielcikova (for STAR Collaboration), Nucl. Phys. A 783 (2007) 565c. C. Nattrass (for STAR Collaboration), Eur. Phys. J. C 62 (2009) 265. O. Barannikova, (for STAR Collaboration), talk given at Quark Matter 2008, Jaipur, India, J. Phys. G: Nucl. Part. Phys. 35 (2008) 104086. A. Sickles, plenary talk given at Quark Matter 2009, Knoxville, TN. E. Wenger (for PHOBOS Collaboration), J. Phys. G: Nucl. Part. Phys. 35 (2008) 104080; B. Alver et al. (PHOBOS Collaboration), arXiv:0903.2811. M. Daugherty (for STAR Collaboration) J. Phys. G: Nucl. Part. Phys. 35 (2008) 104090. J. Adams et al. (STAR Collaboration), Phys. Rev. D 74 (2006) 032006. R. C. Hwa and C. B. Yang, arXiv: 0912.1377. S. Afanasiev et al., (PHENIX Collaboration), arXiv: 0903.4886.
150. 151. 152. 153. 154. 155.
ELLIPTIC FLOW: A STUDY OF SPACEMOMENTUM CORRELATIONS IN RELATIVISTIC NUCLEAR COLLISIONS
PAUL SORENSEN Physics Department, Brookhaven National Laboratory, Building 510, Upton, NY 11973, USA
[email protected] Here I review measurements of v2 , the second component in a Fourier decomposition of the azimuthal dependence of particle production relative to the reaction plane in heavyion collisions. v2 is an observable central to the interpretation of the subsequent expansion of heavyion collisions. Its large value indicates significant spacemomentum correlations, consistent with the rapid expansion of a strongly interacting Quark Gluon √ Plasma. Data is reviewed for collision energies from sN N = 2 to 200 GeV. Scaling observations and comparisons to hydrodynamic models are discussed.
1. Introduction Collisions of heavy nuclei have been exploited for decades to search for and study the transition of hadronic matter to quark gluon plasma.1,2 In these collisions, the extended overlap area, where the nuclei intersect and initial interactions occur, does not possess sphrerical symmetry in the transverse plane. Rather, for noncentral collisions, the overlap area is roughly elliptic in shape. If individual nucleonnucleon collisions within the interaction region are independent of each other (e.g. pointlike) and no subsequent interactions occur, this spatial anisotropy will not be reflected in the momentum distribution of particles emitted from the interaction region. On the other hand, if the initial interactions are not independent, or if there are subsequent interactions after the initial collisions, then the spatial anisotropy can be converted into an anisotropy in momentumspace. The extent to which this conversion takes place allows one to study how the system created in the collision of heavy nuclei deviates from a pointlike, noninteracting system. The existence and nature of spacemomentum correlations is therefore an interesting subject in the study of heavy ion collisions and the nature of the matter created in those collisions.3,4 Figure 1 shows an illustration of the possible stages of a heavyion collisions starting with some initial energy density deposited at midrapidity, followed by a QGP expansion, a hadronization phase boundary, a kinetic freezeout boundary and finally the observation of particle trajectories in a detector. One can consider a number of ways to study spacemomentum correlations: e.g. twoparticle correlations5 and HBT.6,7 In this review we discuss elliptic flow v2 ; an observable that has been central in the interpretation of heavyion data 323
324
P. Sorensen
Fig. 1. A schematic diagram of the expansion after an ultrarelativistic heavyion collision.
and QGP formation.8 Given the predominantly elliptic shape of the initial overlap region, it is natural to ask whether this shape also shows up in the distribution of particles in momentumspace. Figure 2 shows a schematic illustration of the conversion of coordinatespace anisotropy to anisotropy in momentumspace. The left panel shows the position of nucleons in two colliding nuclei at the moment of impact. The overlap region is outlined and shaded. A Fourier decomposition can be used to describe the azimuthal dependence of the final triple momentumspace distributions:9 d3 N 1 dN = × [1 + 2v1 cos(φ − Ψ) + 2v2 cos(2(φ − Ψ)) + . . . ] , pT dpT dyd(φ − Ψ) 2π pT dpT dy (1) where φ is the azimuth angle of the particle, y the longitudinal rapidity variable, pT the transverse momentum, and Ψ is the reaction plane angle defined by the vector connecting the centers of the two colliding nuclei. Positive v2 implies that more particles are emitted along the short axis of the overlap region. To study the extent to which spacemomentum correlations develop in heavyion collisions, one can measure the second component v2 and compare it to the initial spatial eccentricity.3,10 The right panel of Fig. 2 shows the final azimuthal distribution
Review of Elliptic Flow v2
y
325
py pT (GeV/c) 1.5
x
1.0 0.75 0.5 0.25
px
√ Fig. 2. Schematic illustrations of a sN N = 200 GeV Au+Au collision with a 6 fm impact parameter. The left panel shows the nucleons of the two colliding nuclei with an ellipse outlining the approximate interaction region. The right panel shows a momentumspace representation of v 2 . The average radius of each successive ring represents the pT of the particles while the anisotropy of the ring represents the magnitude of v2 . The highest pT particles (outerring) exhibit the strongest v2 while the lowest pT particles (innerring) exhibit a vanishingly small v2 .
of particles in momentumspace. The curves represent the anisotropy at different pT values measured in 200 GeV Au+Au collisions11 : i.e. f (pT , φ) = pT ∗ (1 + 2v2 (pT ) cos(atan2(py , px ))). The goal of v2 measurements is to study how the initial spatial anisotropy in the left panel is converted to the momentumspace anisotropy in the right panel. In this review, a summary of v2 data for different colliding systems, different centerofmass energies, and different centralities is given. This review will focus on results from the first four years of operation of the Relativistic Heavy Ion Collider (RHIC). We start with a brief discussion of the beam energy dependence of v2 and some ideas about what physics might be relevant. Even before considering physics scenarios to explain how a spacemomentum correlation develops, one can see that to interpret v2 it is important to understand the initial geometry and how it varies with the collision centrality and systemsize. Since, the concept of the reactionplane is so central to the definition of v2 and eccentricity is so central to it’s interpretation, I discuss the two in a subsection below. Then a review of RHIC data is provided. This will include the dependence of v2 on centerofmass energy, centrality, colliding system, pseudorapidity, pT , particle mass, constituent quark number and various scaling laws. In the following section, I will discuss comparisons to models and the emergence of the hydrodynamic paradigm at RHIC. Particular emphasis is given to uncertainties in the model comparisons. In that section I will also discuss current attempts to extract viscosity and future directions of investigation. Voloshin, Poskanzer, and Snellings recently wrote a review article12 on collective phenomena in noncentral nuclear collisions that deals with a similar subject matter. That article provides valuable detail on technical aspects of measuring v2 . In this review I will attempt to avoid duplicating that work by discussing interpretations of v2 more extensively and refer the reader to that review where appropriate.
326
P. Sorensen
0.1
v2 0.05
inplane 0 outofplane
FOPI EOS E895 E877 CERES NA49 STAR Phenix Phobos
0.05
0.1
101
1
10
102
103
104
Ebeam/A (GeV) Fig. 3. The beam energy dependence of elliptic flow measurements. The RHIC and E877 data are for charged hadrons, independent of species. The NA49 data is for charged pions. The E895 data are for protons and the FOPI data are for atomic number Z=1. At each energy, the sample of particles is close to the total charge. Positive v2 values indicate that particles tend to be more aligned with the reactionplane (inplane). RHIC and SPS data suggest a smooth trend of inplane √ √ v2 growing with log( snn ) above Ebeam /A ≈ 20 GeV or sN N ≈ 6 GeV.
1.1. Two decades in time and five decades in beam energy Positive values of v2 imply that particles tend to be produced more abundantly in the xdirection than in the ydirection. This is referred to as inplane flow. Figure 3 shows v2 measured in an interval of beam energies covering five orders of magnitude.13–22 For Ebeam /A ranging from approximately 0.12 − 5 GeV √ (1.92 < snn < 3.3 GeV), v2 is negative. For this energy range, spectator protons and neutrons are still passing the interaction region while particles are being produced. Their presence inhibits particle emission in the inplane direction leading to the phenomenon termed squeezeout. At still lower energies, v2 is positive as the rotation of the matter leads to fragments being emmitted inplane. At this energy beam rapidity and midrapidity are essentially indistinguishable with ybeam < 0.41 units. Inplane flow: As the beam energy is increased, the nuclei become more Lorentz contracted and the time it takes the spectators to pass each other decreases.
Review of Elliptic Flow v2
327
It was predicted by Ollitrault3 that at high enough beam energy, the squeezeout phenomena would cease and v2 would take on positive values. Positive v2 values √ were measured at the AGS for energies above Ebeam /A = 5 GeV ( snn = 3.3 GeV). √ For energies above Ebeam /A ≈ 20 GeV ( snn = 6.3 GeV), v2 exhibits a steady log√ linear increase: v2 ≈ 0.01 + 0.0042 log(Ebeam /A) or v2 ≈ 0.008 + 0.0084 log( snn ) where the data represented are from intermediate impact parameter A+A collsions. It appears therefore that RHIC v2 data may be part of a smooth trend that began at SPS energies. This trend was noted previously at least once.23 Understanding the physics that underlies that trend is one of the challenges of heavyion physics. One class of models that has provided an illustrative reference for heavyion collisions are hydrodynamic models which are used to model the expansion the matter remaining in the fireball after the initial collisions.24–33 This model can be used to determine how matter with a vanishingly small mean free path would convert the initial eccentricity into v2 . These models typically treat all elliptic flow as arising from the final state expansion rather than from some initial state effects. 34–37 In the hydrodynamic models, large pressure gradients in the inplane direction lead to a preferential flow of matter in the inplane direction. In this review, we will use hydrodynamic models as a convenient reference. Other models providing a valuable reference for measurements include hadronic and partonic cascades and transport models.38–46 Additionally, the blastwave model provides a successful parametrization of low pT heavyion data, including v2 , HBT, and spectra in terms of several freezeout parameters.30,47
1.2. Initial geometry: the reaction plane and eccentricity In the collision of two symmetric nuclei, a unique vector (the yaxis) can be defined by applying the righthandrule to the momentum vector of one nucleus and the vector pointing to the center of the other nucleus. The yaxis is a pseudovector. The reactionplane is then the plane perpendicular to the yaxis containing the points at the center of the two nuclei. The reactionplane and the righthanded coordinate system are illustrated in Fig. 4. The figure contains perspective illustrations of two nuclei approaching with an impact parameter of 6 fm. The impact parameter is the distance between the centers of the two nuclei at the moment of their closest approach. The two nuclei in this illustration are Lorentz contracted by a Lorentz gamma factor of 10 which roughly corresponds to the appropriate gamma for top SPS energies. The reactionplane is not directly observed in experiments, however, and this introduces a systematic uncertainty into the measurement of v2 . One often relies instead on indirect observations to estimate v2 .48–52 For example, when forming two particle azimuthal correlations such as d(φdN , a nonzero v2 value will lead 1 −φ2 ) to a modulation in ∆φ = φ1 − φ2 of the form 1 + 2hv22 i cos(2∆φ). Figure 5 shows the correlation function for hadrons produced at midrapidity at RHIC.53,54 The
328
P. Sorensen
y y
y
y
x z
x
z
z
z x
x
y
y
y z x
z
y
x
x z
zx
√ Fig. 4. Schematic illustrations of a sN N = 200 GeV Au+Au collision with a 6 fm impact parameter. The righthanded coordinate systems defined by the momentum of the nucleus (here the zaxis) crossed with the vector pointing to the center of the approaching nucleus (the xaxis) is shown for each nucleus. The reactionplane is the plane normal to the yaxis, containing the centers of the two colliding nuclei. The nucleons are distributed inside the nucleus according to a WoodsSaxon distribution. The nuclei are Lorentz contracted in the zdirection.
panels show different centralities. The area normalized correlation function is R AB Y (∆φ) Y AB (∆φ) dN AB R Mixed C(∆φ) ≡ Same × ∝ (2) AB AB d(∆φ) YMixed (∆φ) YSame (∆φ)
AB AB where YSame (∆φ) and YMixed (∆φ) are, respectively, the uncorrected yields of pairs in the same and in mixed events within each data sample. C(∆φ) shows a clear cos(2∆φ) dependence. We note here that what is measured in these correlation functions is hv22 i = hv2 i2 + σv22 in anticipation of a discussion of v2 fluctuations. v2 will not be the only contribution to the azimuthal dependence of the twoparticle azimuthal correlations. Other processes that are not related to the reactionplane can give rise to structures in the shape of the twoparticle ∆φ distribution as well. These nonreaction plane contributions are commonly called “nonflow”. The subject of nonflow is an important one and will be discussed throughout this review. The contribution of nonflow can be seen more clearly by looking at very peripheral collisions or by selecting high momentum particles to increase the chance that a particular pair of hadrons are correlated to a hard scattered parton (jet). Figure 6 shows the correlation function for higher momentum particles.55 The solid line shows what a pure v2 correlation would look like.56 The difference between those curves and the data are often taken as a measurement of jet correlations. 55,57,58
Review of Elliptic Flow v2
Fig. 5. Charged hadron correlation functions in Au+Au collisions at centrality intervals and two pT ranges.
329
√ snn = 200 GeV for two
Even if the reactionplane were known with precision, there is no first principles calculation of the initial matter distribution in the overlap region, so the eccentricity is uncertain. Various models can be used to calculate the initial spatial eccentricity which can then be compared to v2 . Defining the yaxis according to the righthandrule, the eccentricity εs is traditionally calculated as: εs =
hy 2 − x2 i , hx2 + y 2 i
(3)
where the average represents a weighted mean. Other eccentricity definitions have also been considered.59 The weights can be some physical quantity in a model such as energy or entropy density, or simply the position of nucleons participating in the collision. One popular method for calculating the eccentricity is to use a Monte Carlo Glauber model. Details can be found in a recent review.60 In that model, a finite number of nucleons are distributed in a nucleus according to a WoodsSaxon distribution. Then two nuclei are overlaid with a fixed impact parameter and the x and y positions of the participating nucleons is determined based on whether the nucleons overlap in the transverse plane; each nucleon is considered to be a √ disk with an area determined by the s dependent nucleonnucleon crosssection. The x and y coordinates of the participating nucleons are then used to calculate the eccentricity. Those nucleons that do not participate in this initial interaction are called spectators. One can anticipate that due to the finite number of nucleons
330
P. Sorensen
1.06 1.04
(a) 05%
(b) 510%
(c) 1020%
(d) 2040%
(e) 4060%
(f) 6090%
1.02 1 0.98 0.96 0.94
C(∆ φ)
1.1 1.05 1 0.95 0.9
1.3 1.2 1.1 1 0.9 0.8
0
0.5
1
1.5 2 ∆ φ (rad)
2.5
3
0.5
1
1.5 2 ∆ φ (rad)
2.5
3
Fig. 6. Charged hadron correlations for a variety of centrality intervals. The correlation function is formed between two samples of hadrons based on their pT : the “trigger” particle sample is selected with 2.5 GeV/c < pT < 4.0 GeV/c while the “associated” particle sample is selected with 1 GeV/c < pT < 2.5 GeV/c.
in this model, the initial geometry will fluctuate. Other models used to determine the initial matter distribution including HIJING,61 NEXUS,62 and Color Glass Condensate models63–66 also reach the same conclusions; the initial overlap region is expected to be lumpy rather than smooth. Figure 7 shows the gluon density in the transverse plane which is probed by a 0.2 fm quarkantiquark dipole at two √ different x values in the IPsat CGC model63 (x = 2pT / sN N is 10−5 in the left panel and 10−3 in the right panel). The lumpiness is immediately apparent. Until recently however,67,68 v2 data was compared almost exclusively to calculations assuming an infinitely smooth initial matter distribution (for example in initializing a hydrodynamic expansion). Improving on that approximation may be d2 N important for understanding the shape expected for the d∆ηd∆φ distribution.5,69 This distribution is also investigated in heavy ion collisions in order to search for jets. In any scenario where spacemomentum correlations develop, the correlations dN distribution and fluctuations in the initial geometry can be manifested in the d∆φ and understanding these correlations is important for interpreting heavyion collisions. Fluctuations in the initial geometry have also led to the idea of measuring particle distributions relative to the participantplane rather than the reactionplane.59,70 The participantplane is defined by the major axis of the eccentricity
Review of Elliptic Flow v2
331
Fig. 7. Gluon density in the transverse plane when the nucleus is probed at different x values by a 0.2 fm quarkantiquark dipole in the IPsat CGC model.
εstd
1
Monte Carlo Glauber Model
εpart
0.8
1
Monte Carlo Glauber Model
0.8 0.6 0.4
0.6
0.2
0.4
0 0.2
0.2
0.4 0
2
4
6
8
10
Impact Parameter (fm)
12
14
0 0
2
4
6
8
10
12
Impact Parameter (fm)
Fig. 8. The distribution of eccentricity on an eventbyevent basis when calculated relative to the reactionplane (left) and the participant plane (right).
which, due to fluctuations, can deviate from the reactionplane. The eccentricity relative to the participantplane is a positive definite quantity and is always larger than the eccentricity relative to the reactionplane; the participantplane is defined by rotating to the axis that maximizes the eccentricity. Figure 8 shows the eventbyevent distribution of the standard eccentricity (left panel) and the participant eccentricity (right panel) as a function of impact parameter determined from a MonteCarlo Glauber calculation. The fluctuations in this model are large as illustrated by the widths of the distributions. The relationship between the different definitions of eccentricity and their fluctuations are explained clearly in two recent papers.59,71 Different models for the initial matter distribution yield different estimates of hεi p and hε2 i. The deviations in the hεi for different models can be of the order of 30% and strongly centrality dependent.72 That uncertainty in hεi leads to an inherent uncertainty when comparing models to v2 /ε. This level of uncertainty becomes
332
P. Sorensen
important when attempting to estimate transport properties of the matter based on comparisons of the observed v2 to the initial ε. 2. Review of Recent Data √ The first paper published on RHIC data was on elliptic flow in sN N = 130 GeV Au+Au collisions.18 Figure 9 shows that data on the centrality dependence of v2 . The values of v2 reach a maximum of approximately 6% for peripheral collisions where the initial eccentricity of the system is largest. That value is 50% larger than the values reached at SPS energies21 and the v2 values are a factor of two larger than the those predicted by the RQMD transportcascade model.42 For central collisions, the measurements approach the zero meanfreepath limit estimated from the eccentricity shown in the figure as open boxes. The boxes represent the eccentricity scaled by 0.19 (bottom edge of the boxes) and 0.25 (top edge of the boxes). Those values are chosen to represent the typical conversion of eccentricity √ to v2 in hydrodynamic models. At lower sN N energies, the RQMD model provided a better description of the data, while hydrodynamic models significantly overpredicted the data. The conclusion based on this early comparison, therefore,
0.1
yd
H ro
v2
dy ic
m
na
0.08
od M el
0.06
13
0G
0.04
eV
+A u
0.02 0 0
Au
0.2
0.4
0.6
n ch /n max
Da
ta
0.8
1
Fig. 9. First measurements of v2 versus centrality at RHIC. Positive values are observed which are largest for events with the largest eccentricity and decrease for more central, symmetric collisions. The trend with centrality clearly indicates a spacemomentum correlation driven by the eccentricity of the initial overlap zone. Centrality is expressed in terms of the observed multiplicity of a give event relative to the highest multiplicity observed nch /nmax . Data are compared to eccentricity scaled by 0.19 (bottom edge of the boxes) and 0.25 (top edge of the boxes). The values are chosen to represent the typical conversion of eccentricity to v2 in a hydrodynamic model.
Review of Elliptic Flow v2
333
v2 /ε 0.25
HYDRO limits
0.2 0.15 0.1
Elab /A=11.8 GeV, E877 Elab /A=40 GeV, NA49 Elab /A=158 GeV, NA49
0.05
sNN =130 GeV, STAR sNN =200 GeV, STAR Preliminary
0 0
5
10
15
20
25
30
35
(1/S)dNch /dy Fig. 10. v2 scaled by a Monte Carlo Glauber model calculation of the initial overlapp eccentricity. The ratio is plotted versus transverse particle density (1/S)dNch /dy, where S = π hx2 ihy 2 i is a weighted average area calculated with the same model as the eccentricity. Data are taken from √ different sN N values and different centralities. Plotted in this format, the data suggest v 2 /ε for different energies and overlap geometries is determined by the transverse particle density, and approaches a zero meanfreepath hydrodynamic limit for most central top energy collisions at RHIC. This conclusion is not universally accepted and is still being investigated.
was that heavyion collisions approximately satisfy the assumptions made in the hydrodynamic models: 1) zero meanfreepath between interactions, and 2) early local thermal equilibrium.33 These conclusions remain at the center of scientific debate in the heavyion community. In Fig. 10, v2 is scaled by model calculations of the initial eccentricity and 21 plotted versus transverse particle density S1 dN This facilitates comparisons of v2 dy . √ across different sN N energies, collision centralities and systemsizes.10,41 For the case of ballistic expansion of the system — that is an expansion for which the produced particles escape the initial overlap zone without interactions — v2 should only reflect the spacemomentum correlations that arise from the initial conditions. Those can exist in the case that the initial interactions are not pointlike34 but rather involve crosstalk between different N + N interactions within the overlap zone. The opposite extreme from the ballistic expansion limit is the zero meanfreepath limit represented by ideal hydrodynamic models. Lacking a length scale, the zero meanfreepath models should not depend on systemsize and instead should be a function of density. The measurements of v2 are expected to rise from values near the ballistic expansion limit and asymptotically approach the zero meanfreepath limit as the
334
P. Sorensen
density of the system is increased. Data in Fig. 10 exhibit such a behavior with the most central collisions at full RHIC energy apparently becoming consistent with the hydrodynamic model. This conclusion however depends on the model calculations for the initial eccentricity and on the assumption that the observed v2 dominantly arises from an expansion phase where anisotropic pressure gradients are the origin of the spacemomentum correlations. Different models for the eccentricity yield ε results that deviate both in their centrality dependence and in their overall magnitude. Reasonable models for the eccentricity can easily give magnitudes 30% larger than those used in Fig. 10 with a stronger centrality dependence. The ratio v2 /ε can therefore be smaller than what is shown and have a different shape.72 √ Given this level of uncertainty, the conclusion that heavyion collisions at sN N = 200 GeV approximately satisfy the assumptions made in the hydrodynamic models i.e. early local thermal equilibrium and interactions near the zero meanfree path limit, would be more convincing if an asymptotic approach to a limiting value were observed. Rather, for the eccentricity calculation used in Fig. 10, the data suggest a nearly linear rise with no indication of asymptotic behavior. In the hydrodynamic picture, one might also expect that v2 /ε versus S1 dN dy will be sensitive to the equationofstate of the matter formed during the expansion phase. Since v2 is expected to reflect spacemomentum correlation developed due to pressure gradients and S1 dN dy is a measure of the transverse particle density, 1 dN v2 /ε versus S dy could be considered as a proxy for the pressure versus energy density or the equationofstate. It’s difficult to identify in the data the features that are expected in the equationofstate. Figure 11 shows the equationofstate calculated in a recent lattice QCD calculation.73 The onset of the QGP phase is RHIC
LHC
0.35 0.30
p/ε
0.25
243 6 323 8 fit: p/ε HRG: p/ε c2s
0.20 0.15 0.10 0.05
ε [GeV/fm3]
0.00 1
10
100
1000
Fig. 11. The QCD equationofstate (pressure over energy density versus the fourth power of the energy density) as determined in Lattice calculations.
Review of Elliptic Flow v2
335
seen to lead to an increase in the pressure as the energy density is increased above a critical value. The energy density in heavyion collisions is often estimated from the Bjorken formula74 : Bj =
1 dET Aτ dy
(4)
T which depends only on the experimentally accessible quantity dE dy , the overlap area of the nuclei and τ , the unknown formation time which is often assumed to be 1 fm. The Bjorken estimate for the energy density is closely related to the transverse particle density (1/S)(dN/dy).
2.1. Differential elliptic flow In addition to studying how v2 integrated over all particles depends on the centrality √ or sN N of the collision, one can study how v2 depends on the kinematics of the produced particles (differential elliptic flow). Figure 12 shows the centrality and
2.5
Peripheral/Central
v 2 0.1 2
0.08
1.5 2
1
0
η
1
2
0.06
0.04
0.02
Peripheral (2550%) Midcentral (1525%)
Central (315%)
0
4
2
0
η
2
4
Fig. 12. v2 (η) for 200 GeV Au+Au collisions for three centrality intervals. The inset shows the ratio of v2 (η) for central collisions over peripheral collisions. Open symbols are data reflected to negative η.
336
P. Sorensen
pseudorapidity dependence of v2 for 200 GeV Au+Au collisions.75 v2 is largest at midrapidity where the transverse particle density is largest and then falls off at larger η values. This behavior is therefore consistent with the trends seen in integrated v2 where v2 /ε appears to increase with increasing transverse particle density. The fall off of v2 (η) with increasing η is common to the three centrality intervals studied. The inset of the figure shows the ratio of v2 in peripheral over central collisions. Within errors the ratio is flat indicating a similar shape for all centralities with v2 (η) only changing by a scale factor. Scaling of v2 (η) for different energies and system sizes will be discussed in a later section. Figure 13 shows v2 for a variety of particle species as a function of their transverse momentum pT .76–81 In the region below pT ∼ 2 GeV/c, v2 follows mass ordering with heavier particles having smaller v2 at a given pT . Above this range, the mass ordering is broken and the heavier baryons take on larger v2 values. A hydrodynamic model for v2 (pT ) is also shown which describes the v2 in the lower pT region well. This mass ordering is a feature expected for particle emission from a boosted source. In the case that particles move with a collective velocity, more massive particles will receive a larger pT kick. As the particles are shifted to higher pT , the lower momentum regions become depopulated with a larger reduction
v2
0.3
PHENIX Data π++π+ K +K p+p + h +h
Hydro model π K p Λ
STAR Data π++π0 KS p+p Λ+Λ
0.2
0.1
0 0
2
4
6
8
10
Transverse Momentum pT (GeV/c) Fig. 13. v2 (pT ) for a variety of identified particle species and inclusive charged hadrons. Data are for Au+Au collisions at 200 GeV averaged over all centralities. At low momentum v 2 (pT ) exhibits mass ordering while at larger pT the identified particle v2 appears to be grouped according to constituent quark number. The mass ordering at low pT is approximately reproduced by the hydrodynamic calculation.
Review of Elliptic Flow v2
0.1
337
−−
Au + Au at √sNN = 200 GeV
0.08
Tfo (MeV) 130
0.06 163
0.04
0.02
Hydrodynamic calculation TC=165 MeV π
0
0
K/KS0
0.5
Λ
p
1
Ξ
Ω
1.5
2 2
Particle Mass (GeV/c ) Fig. 14. pT integrated v2 versus mass for a variety of identified particle species. Yellow bands show hydrodynamic calculations with different freezeout temperatures.
in the direction with the largest boost (inplane). This reduction reduces v2 at a given pT , with the reduction largest for more massive particles. Note that this does not imply that the more massive particles have a smaller integrated v2 value, and in fact the opposite is true. Figure 14 shows v2 for identified particles integrated over all pT .82,83 The integration shows that v2 increases with particle mass. This is because the more massive particles have a larger hpT i and v2 is generally increasing with pT in the pT region where the bulk of the particles are produced. The hydrodynamic model also exhibits this trend.
2.1.1. Identified particle v2 (pT ): RHIC versus SPS √ Figure 15 shows pion and proton v2 from sN N = 62.4 Au+Au81 and 17.3 GeV Pb+Pb collisions.21 The centrality intervals have been chosen similarly for the 17.3 GeV and 62.4 GeV data. The STAR data at 62.4 GeV are measured within the pseudorapidity interval η < 1.0 and the 17.3 GeV data are from the rapidity interval 0 < y < 0.7. These intervals represent similar y/ybeam intervals. It has been shown that v2 data for pions and kaons at 62.4 GeV are similar to 200 GeV data; the 62.4 GeV data only tending to be about 5% smaller than the 200 GeV data. Appreciable differences are seen between the 17.3 GeV and 62.4 GeV data. At pT > 0.5 GeV/c, for both pions and protons, the v2 values measured at 62.4 GeV
338
P Sorensen
STAR (62.4 GeV; 040%; midrapidity)
v2
π++πp+p
0.15
0.1
0.05
NA49 (17.3 GeV; 044%; 0 7 due to statistical and systematic uncertainties. At these higher pT values one expects that the dominant process giving rise to v2 is jetquenching85 where hadron suppression is larger along the long axis of the overlap region than along the short axis.86–88 For very large energy loss, the value of v2 should be dominated by the geometry of the collision region. Figure 16 shows a comparison of v2 data89 for 3 < pT < 6 GeV/c compared to several geometric models.90 This comparison seems to indicate that v2 in this intermediate pT range is still too large to be related exclusively to quenching. In the higher pT regions, significant azimuthal structure will arise from jets, so nonflow correlations are thought to be significant in this region.58 These effects have been studied in several ways. The fourparticle cumulant v2 has been studied as a function of pT and the ratio of the four and twoparticle cumulants v2 {4}/v2 {2} is found to decrease with increasing pT .91,92 This decrease is identified with a gradual
v2{4} (3 GeV/c < pt < 6 GeV/c)
0.25
v2{4} 0.2 0.15 Ha
0.1
el
l
Hard Sphere
0.05 0 0
rd
Sh
Wo o d s  Sax
2
4
6
8
10
on
12
b (fm) Fig. 16. The fourparticle cumulant v2 at 3 < pT < 6 GeV/c versus impact parameter, b, compared to models of v2 based on geometry alone. Data in this region does not agree well with these surface emission pictures.
340
P. Sorensen
sNN = 130 GeV
sNN = 200 GeV
(a)
v 2[4]/v2[2]; 570%
0.08
0.06
〈v2〉
Ratio
1
2
3
pT (GeV/c)
4
5
0.6 0.4 0.5
1
0.04
0.02
1
0.8
0.2 0
0.8
0
1
(b) Ratio
1.2
0 0
Analysis Method 4th order cumulant reaction plane 0.5
1
Nch/Nmax
Fig. 17. The left panel shows the ratio v2 {4}/v2 {2} for the 0%–70% centrality interval in 200 GeV Au+Au collisions. The ratio falling below unity indicates the importance of nonflow and v 2 fluctuations. The larger reduction at high pT seems indicative of an increase in nonflow due to jets. The right panel shows the centrality dependence of the v2 {4} and v2 measured with respect to the event plane in 130 GeV Au+Au collisions. The inset shows the ratio of the two.
increase in the contribution of jets to v2 {2} (see Fig. 17 left panel). The fourparticle cumulant suppresses contributions due to intrajet correlations but the statistical errors of the measurement are larger. One can also suppress jet structure in the v 2 measurement by implementing a ∆η cut in the pairs of particles being used in the analysis.11 In this case, a high pT particle is correlated with other particles in the event that are separated by a minimum ∆η. This method relies on the assumption that jet correlations do not extend beyond a given ∆η range. Interactions of jets with the medium in nuclear collisions however can change the structure of jets and extend the correlations in ∆η beyond the widths observed in p + p collisions.93 This method therefore is not guaranteed to eliminate nonflow from jets. The problem of measuring v2 without nonflow and of measuring modifications of jet structure by the medium are entirely coupled. If one is known, the other is trivial. Other methods for suppressing nonflow include measuring correlations between particles at midrapidity and and an eventplane determined from particles observed at forward rapidity.94 In the extreme and the most effective case, the eventplane was reconstructed from spectator neutrons in a ZeroDegree Calorimeter to measure v2 of produced particles near η = 0. An extension of analyses based on the change in correlations across various rapidity intervals is the analysis of the two dimensional correlation landscape for twoparticle correlations e.g. d2 N/∆φ∆η.95 After unfolding the two particle correlations one can attempt to identify various structures with known physics such as jets, resonance decay, or HBT based on their width in η and φ. The remaining cos(2∆φ) structure can then be used to estimate hv2 i2 + σv22 . This method will be discussed below.
341
T
i
Σcos (2( φp  φ )) 〉
Review of Elliptic Flow v2
〈
i
1
Au+Au 100%80% d+Au Minbias p+p Minbias
0.1
0
2
4
6
8 10 12 0
Au+Au 60%20% d+Au Minbias p+p Minbias
2
4
6
8 10 12 0
Au+Au top 5% d+Au Minbias p+p Minbias
2
4
6
8 10 12
p T (GeV/c) Fig. 18. Second harmonic azimuthal correlations in p + p, d+Au, and Au+Au collisions. The quantity hΣi cos(n(φpT − Φi ))i = M hv2 v2 (pT )i + M δ2 facilitates comparisons between different systems particularly where a reactionplane may not be well defined.
As we study progressively smaller systems the connection between the nucleusnucleus reactionplane and the azimuthal structure breaks down. In the limit that one proton from each nucleus participates in the interaction, the reactionplane defined by the colliding protons will not necessarily be related to the reactionplane defined by the vector connecting the centers of the colliding nuclei. In order to facilitate a comparison between the pT dependence of azimuthal correlations in large systems and small systems, the scalar product huQ∗ i is used where u = ei2φ and Q∗ = e−i2Ψ .89 The mean of uQ∗ therefore yields a quantity that depends on hv2 v2 (pT )i and nonflow as follows: huQ∗ i = hΣi cos(n(φpT − Φi ))i = M hv2 v2 (pT )i + M δ2 ,
(5)
where M is the multiplicity used in the sum. Figure 18 shows this quantity for p+p, d+Au, and three Au+Au centrality intervals. The p+p and d+Au data are repeated in each of the panels. The most peripheral Au+Au collisions are shown in the left panel. The centrality bin shown is not usually presented since trigger inefficiencies for low multiplicity events makes it difficult to define the actual centrality range sampled. In this case, the data has been published in order to compare uQ∗ between the most peripheral sample of events and p + p collisions. The data in Au+Au has a similar shape and magnitude as the data in p + p. This suggests that peripheral collisions are dominated by the same azimuthal structure as p + p collisions; an
342
P. Sorensen
observation consistent with twoparticle ∆η, ∆φ correlations.96 The data from midcentral Au+Au collisions shown in the middle panel however, exhibit a magnitude and shape clearly different than p+p collisions. While uQ∗ for p+p, d+Au and very peripheral Au+Au collisions rises monotonically with pT , for mid central Au+Au collisions, the data rises to a maximum at pT = 3 GeV/c and then falls. For central collisions shown in the right panel, a similar feature is seen with data rising to a maximum at pT = 3 GeV/c, then falling until pT = 6 GeV/c, where it begins rising again. This second rise is presumably a manifestation of nonflow at high pT in central collisions. These data suggest that azimuthal structure in Au+Au collisions above pT = 6 GeV/c is dominated by jets. This is also consistent with the conclusions reached by examining the particle type dependence of v2 and RCP .78 In Fig. 18 the p + p data is replotted in each panel to facilitate a comparison between the shape and magnitude in p + p to that in Au+Au. In the absence of jetquenching however, nonflow at high pT is expected to scale with the number binary nucleonnucleon collisions Nbinary . The plotting format in Fig. 18 on the other hand, assumes that δ2 ∝ 1/M rather than Nbin /M 2 as would be expected for hard scattering. The multiplicity has been shown to scale as (1 − xhard )Npart + xhard Nbinary with xhard ≈ 0.11.97,98 This is referred to as the twocomponent model. In order to compare azimuthal structure in Au+Au collisions to Nbinary scaling of p + p collisions we can form a ratio in analogy with RAA for single hadrons: ∗
uQ RAA =
huQ∗ iAA /MAA Nbinary huQ∗ ipp /Mpp
(6)
where MAA and Mpp are the multiplicities in A + A and p + p collisions with MAA taken according to the two component model. In the case that jet production in Au+Au collisions scales with the number of binary collisions, as hard processes uQ∗ uQ∗ are expected to, RAA should be unity. The right panel of Fig. 19 shows RAA for charged hadrons in 0%−−5% central Au+Au collisions. For comparison, RAA from uQ∗ single particle charged hadron spectra is also shown in the figure. RAA first rises abruptly with pT to a maximum of 2 at pT ≈ 0.5 GeV/c and then falls to a value of uQ∗ 0.25 at pT ≈ 5 GeV/c. At pT > 5 GeV/c RAA is similar to RAA . This shows that jetquenching suppresses the charged hadron spectra, and the azimuthal structure by a similar amount; confirming that the single hadron suppression is indeed related uQ∗ to jetquenching. RAA is complimentary to studies of IAA , the ratio of dihadron correlations in Au+Au and p + p collisions.57 We note the presence of what appears to be a local minimum and local maximum pT ≈ 1.5 and 2.0 GeV/c respectively. It is not clear if this is a real feature or simply an artifact largely caused by the shape of the p + p data. In the case that it is a real feature, it is possibly related to the changing particle composition in Au+Au collisions where baryons with larger v2 values become more prominent. At pT = 3 GeV/c, baryons and mesons in p + p collisions are created in the proportion 1:3 while at the same pT in central Au+Au collisions the proportion is approximately uQ∗ 1:1. RAA will be an interesting quantity to investigate for identified particles. One
Review of Elliptic Flow v2 〈uQ*〉
〈uQ*〉 ≈ M 〈v(p )v〉 + Mδ
RAA =
T
〈uQ*〉
0%5%
〈uQ*〉p+p
343
NbinMpp MAA
1
1
Au+Au (0%5%) 101
p+p minbias 101 0
2
4
6
8
10
12
0
2
4
6
8
10
12
pT (GeV/c) huQ∗i
Fig. 19. The left panel reproduces the data in Fig. 18. The right panel shows R AA
. Binary
huQ∗i RAA
scaling of δ2 from p + p collisions corresponds to = 1. In comparison to binary scaling, nonflow above pT ∼ 5 GeV/c in Au+Au collisions is suppressed by a factor of 5. Also shown for comparison is RAA for single particle spectra of charged hadrons.
can anticipate a quark number dependence at intermediate pT as seen in RCP and v2 . 2.3. Multiply strange hadrons and heavy flavor The buildup of spacemomentum correlations throughout the collision evolution is cumulative. Information about spacemomentum correlations developed during a QuarkGluonPlasma phase can be masked by interactions during a later hadronic phase. For studying a QGP phase, it is useful to use a probe that is less sensitive to the hadronic phase. Multistrange hadrons have hadronic crosssections smaller than the equivalent nonstrange hadrons, and the v2 values measured for hadrons such as φmesons (ss) and Ωbaryons (sss) are therefore thought to be more sensitive to a quarkgluonplasma phase than to a hadronic phase.32 Figure 20 shows v2 (pT ) for the φmeson.99,100 The v2 rises with pT and reaches a maximum of approximately 15% at pT near 2 GeV/c. At intermediate pT , the φmeson v2 appears to follow a trend similar to the other meson KS0 . This observation suggests that either the φmeson cross section is larger than anticipated or rescattering during the hadronic phase does not contribute significantly to v2 . The latter possibility requires that v2 is established prior to a hadronic phase, suggestive of development of v2 during a QGP phase.
344
P. Sorensen
v2 0.25 0.2
Au+Au 200 GeV Λ φ 0 Ks
0.15 0.1 0.05 0 0
1
2
3
4
5
pT (GeV/c) Fig. 20. v2 (pT ) for the φ meson. The φ is composed of an ss pair and is expected to have a smaller hadronic crosssection than nonstrange, or singlystrange hadrons. The φ v 2 is compared to KS0 and Λ v2 .
Measurements of multiply strange hadrons are interesting because they should be less coupled to the matter in a hadronic phase and therefore a better reflection of the QGP phase. Heavy quarks on the other hand (e.g. charm and bottom quarks) may be less coupled to even the QGP matter.101,102 It’s not a priori obvious that heavy quarks will couple significantly to the medium and be influenced by its apparent expansion. The extent to which they do couple to the medium should be reflected in how large v2 for heavy flavor hadrons becomes and how much the nuclear modification (RAA ) deviates from unity. Precision measurements of Heavy Flavor mesons or baryons are not yet available from the RHIC experiments. As a proxy for identifying Dmesons, the STAR and PHENIX experiments have measured nonphotonic electrons.103,104 Nonphotonic electrons are generated from the weakdecays of heavy flavor hadrons and after various backgrounds have been accounted for can, with some caveats,105 be used to infer the RAA and v2 of Dmesons. The top panel (a) of Fig. 21 shows RAA for nonphotonic electrons.106,107 Prior to the measurement of nonphotonic electron RAA , it was expected that heavyflavor hadrons would be significantly less suppressed than light flavor hadrons. These expectations based on a decrease in the coupling of charm quarks to the medium because of the deadcone effect,101 are contradicted by the data; At pT ≈ 5 GeV/c, nonphotonic electrons are as suppressed as pions. This suppression suggests a
Review of Elliptic Flow v2
345
1.8
RAA
1.6
(a)
Moore & Teaney 3/(2πT)
1.4
12/(2 πT)
1.2 1 0.8 0.6 0.4 0.2
v2HF 0.2
(b)
π 0 RAA π 0 v 2, p > 2 GeV/c
0.15
T
e± RAA , e± vHF 2
0.1
0.05
0
0
1
2
3
4
5
p T [GeV/c]
6
7
8
9
Fig. 21. RAA and v2 for nonphotonic electrons. The two measurements can be used in conjunction to provide constraints on models. Data are compared to a model with the conjectured minimum viscositytoentropy ratio η/s and one with η/s four times that value.
stronger than expected coupling of charm quarks to the medium. This coupling apparently also leads to significant v2 as seen in Fig. 21 (b). Also shown in the figure is a calculation of v2 and RAA based on a Langevin model.108 In that model, the strength of the energy loss and momentum diffusion of charm quarks is characterized in terms of a diffusion coefficient (D). RAA and v2 for charm quarks is then computed for several values of D. Two of these values are shown in Fig. 21. Although neither curve provides an entirely satisfactory simultaneous description of v2 and RAA , the comparison suggests that the diffusion coefficient is large. This comparison only achieves rough agreement, but the calculation illustrates the sensitivity of heavy flavor hadrons to transport coefficients of the QGP and v2 is an important quantity to measure for these hadrons. This is also in agreement with H. van Hees, et al.109 where coalescence at the hadronization phase boundary is also considered and found to help improve the agreement with data.
346
P. Sorensen
2.4. Fluctuations and correlations Comparisons between data and models are complicated by uncertainties in the initial eccentricity and by uncertainties in the data. Estimating transport quantities from the data may require a precision comparison between eccentricity and v 2 so it is important to reduce the uncertainties in both. As discussed in the previous sections, a CGC model of the initial conditions yields eccentricity values typically 30% larger than a Gluaber model while the fluctuations (σε ) are still of the same width. The ratio of σε /ε in a CGC model is therefore smaller than in a Glauber model.65,66 One can expect the statistical fluctuations in eccentricity to show up as dynamical fluctuations in v2 measurements. Measuring the dynamic v2 fluctuations in conjunction with hv2 i can therefore provide an additional constraint on the initial conditions.110–113 Several methods have been employed for measuring v2 and the various methods have different dependencies on nonflow correlations and v2 fluctuations.59,68,71,114–118 The differences between these measurements give information on nonflow correlations and v2 fluctuations. If one uses a two particle correlation to estimate v2 , then one finds v2 {2}2 = hcos(2(φi − φj ))i = hv2 i2 + σv22 + δ2 where the average is over all unique pairs of particles. v2 is the single particle anisotropy with respect to the reaction plane v2 = hcos(2(φ − Ψ))i and δ2 is the nonflow parameter which summarizes the contributions to hcos(2(φi − φj ))i from correlations not related to the reaction plane. If one uses a 4particle cumulant v2 {4} calculation, then for most cases the nonflow term will be suppressed by large combinatorial factors and v2 fluctuations will contribute with the opposite sign. For Gaussian fluctuations, v2 {4}2 ≈ hv2 i2 − σv22 .71 Without knowing δ or σv2 , one cannot determine the exact value of hv2 i. Rather, hv2 i2 could lie anywhere between v2 {4}2 and (v2 {2}2 + v2 {4}2)/2. It is advantageous to confront various models with the data that is experimentally accessible. The difference between the two and fourparticle cumulants in the case of Gaussian v2 fluctuations is: v2 {2}2 − v2 {4}2 ≈ δ2 + 2σv22 .
(7)
The term δ + 2σv22 is also approximately equivalent to the nonstatistical width of the distribution of the length of the flow vector distribution (dN/dq2 ) and is 2 called σtot . The flow vector for the nth harmonic is defined as qn,x = Σi cos(nφi ) 2 and qn,y = Σi sin(nφi ). Figure 22 shows σtot extracted from the difference be11 tween the two and fourparticle cumulants. In the case of Gaussian fluctuations, 2 higher cumulants such as v2 {6} are equal to v2 {4}. In this case, the quantity σtot and v2 {2}2 summarizes the information available experimentally from the second harmonic flow vector distribution. No more information can be accessed without applying more differential techniques or by making assumptions about the shape or centrality dependence of flow, nonflow, or flow fluctuations. An example of a more differential analysis is also shown in Fig. 22, where two particle correlations have
Review of Elliptic Flow v2 2 σtot
2
0.003
347
2
v2{2} v 2{4} 2 2 v2{1D} v 2{2D}
0.002
0.001
0 0
2
4
6
8
10
12
14
impact parameter 〈b〉 (fm) 2 = δ +2σ 2 estimated from the from the difference v {2}2 −v {4}2 . Also Fig. 22. The quantity σtot 2 2 2 v2 shown is a parametrization of v2 {1D}2 − v2 {2D}2 ; the difference between hcos(2∆φ)i extracted for all correlations and that extracted by first fitting other structures in the correlations identified with various nonflow sources.
been fit in ∆φ∆η space.119 Terms identified with various nonflow sources have been included with the fit and the remaining cos(2∆φ) modulation is then identified as v2 {2D}2 . In the case that the sources of nonflow are correctly parametrized, δ = v2 {1D}2 −v2 {2D}2 , where v2 {1D}2 is hcos(2∆φ)i integrated over all azimuthal structure. Then v2 {2D}2 = hv2 i2 + σv22 . This procedure is discussed below. Even without attempting to disentangle flow fluctuations from nonflow correlations, the assumption that nonflow is a positive quantity (consistent with 2 v2 {1D}2 − v2 {2D}2 ) can be used with σtot to provide an upper limit on v2 fluctuations σv22
5 where RCP for protons is slightly larger than RCP for pions143 and the v2 measurements are not yet precise enough to conclude whether the baryon v2 is also smaller than the meson v2 . This is a topic that needs to be studied further. In a coalescence picture, the final momentum of the observed hadron would depend on the momentum of the coalescing constituent quarks. The exact dependence is not known but a relatively good scaling of v2 for KS0 and Λ was found when v2 /n was plotted as a function of pT /n. Such a scaling implies that the momentum of the hadron is simply the sum of the momenta of the coalescing quarks. Figure 27 shows v2 /n versus pT /n for KS0 mesons and Λbaryons. The scaling appears to be good throughout the whole pT range but part of this perception is due to the decrease of v2 for both particles at small pT . When a ratio is taken between the v2 /n(pT /n) values, a clear deviation from scaling is seen in the lower pT region. A combination of the mT − m0 scaling in Fig. 26 and the v2 /n scaling in Fig. 27 will lead to a good scaling over the whole measured momentum range; since v2 (mT − m0 ) for all particles fall on a single line at low mT − m0 , dividing the x and yaxis by n will
v2 /n
Λ + Λ (n=3)
K0S (n=2) 0.1
0 0
1
2
3
Transverse Momentum pT /n (GeV/c) Fig. 27. v2 for KS0 and Λ scaled by the number of their constituent quarks (n) and plotted versus pT /n. The data appear to fall on a universal curve which has been taken as an indication of hadron formation via coalescence of quarks from a flowing medium.
Review of Elliptic Flow v2
(a) 40%80%
0.15
355
(b)
1.5
0.1 1 0.05 0
Λ+Λ
KS
0
+
(c) 10%40%
0.15
v2/nq

Ξ +Ξ
0.5
Ratio to Fitting Function (d)
1.5
0.1 1 0.05
Fitting Function
0
(e) 0%10%
0.15
0.05
(f)
1.5
Hydro model π K Λ Ξ
0.1
0.5
1
0.5
0 0
0.5
1
1.5
2
2.5
3
0
0.5
(m  m)/nq
1
1.5
2
(m  m)/nq
T
T
Fig. 28. A more detailed study of quark number scaling in Au+Au collisions. In the left panels (a, c, and e) v2 /n is shown versus (mT − m0 )/n for three centrality classes. Hydrodynamic models are also shown for comparison. Data are fit to a single curve. In the right panels (b, d and f) the ratio of the data and hydro model to the fit function are shown.
not destroy that scaling seen in Fig. 26. Plotting v2 /n versus (mT − m0 )/n should therefore provide a good scaling across a large kinematic range. Figure 28 shows v2 /n versus (mT − m0 )/n for 200 GeV Au+Au collisions in three different centrality intervals.80 Data for KS0 mesons, Λbaryons and Ξbaryons are shown. The left panels show the data with a hydrodynamic calculation and a fitting function. The phenomenologically motivated function v2 /n =
a + bx + cx2 a −(x−d) − 2 1 + exp e
(13)
with x = (mT − m0 )/n, describes the data well for the three centralities. The function captures the rise then saturation and steady decline seen in the data. We
356
P. Sorensen
Table 1. The fit parameters describing the curves in Fig. 28.
Centrality
a
40%–80% 10%–40% 0%–10%
20.0e02 16.4e02 8.96e02
Fit parameters for Eq. (13) b c d 0.0 −4.53e03 −4.08e03
0.0 0.0 0.0
−1.19e02 2.61e02 6.52e02
e 2.37e01 2.40e01 2.70e01
ascribe no physical meaning to the function or the five fit parameters but simply use it as a convenient reference. The right panels show the ratio of the data and the hydro model to the fit function. For reference, the fit parameters are shown in Table 1. The data is in good agreement with the fit function for all centralities while this hydro model calculation does not agree well with the data in any centrality. There is a systematic deviation from the ideal n scaling at (mT − m0 )/n > 0.8 GeV/c2 with KS0 mesons having slightly larger v2 /n values than Λ baryons. This deviation from ideal scaling was predicted based on the inclusion of higher fock states in the hadrons or the inclusion of a finite width in the hadron wave function.144,145 Deviations can also arise in a hadronic phase when the hadronic cross sections are relevant. In the case that hadronic crosssections are an important factor, higher statistics data for Ω baryons and φ mesons should deviate from their respective groups. We also note that hadronic cascade models also obtain approximate v2 /n scaling due to the use of the additive quark model for hadronic crosssections.146 On the other hand, these models underpredict the integrated v2 by a factor of two. We also note that nonflow contributions can affect the scaling observed in this range and the particletype dependence of nonflow sources is still being investigated. In Fig. 29 we investigate the breaking of ideal scaling in more detail with data integrated over a larger centrality interval. While this reduces the statistical uncertainty, it also introduces uncertainties due to the large centrality bin width. In particular, when particle yields have different centrality dependencies, the average eccentricity of events producing a particle can deviate from particle to particle. For example, the enhancement of baryons in central collisions will mean that the average baryon comes from a more central event than the average meson. Given the decrease of v2 with centrality, this can lead to a decrease of baryon v2 simply due to the wide centrality bin. Although there are caveats and systematic errors still to be quantified, we note that the baryons in Fig. 29 appear to lie systematically and significantly below the mesons. Selfsimilar curves are fit to mesons and baryons. The curves appear to describe the data. We note that the two selfsimilar curves shown in Fig. 29 can be nearly unified if we replace n with n+1. This demonstrates that the naive constituent quark scaling is violated to the extent that baryon v 2 is actually closer to 4/3 the meson v2 rather than 3/2. The connection of the baryon versus meson dependence and the number of constituent quark scaling appears to
Review of Elliptic Flow v2
v2 n
357
0.08
0.06
0.04
π++πK 0S
0.02
p+p Λ+Λ Ξ+Ξ Ω+Ω
0 0
0.5
1
1.5
2
25
2
(mT m0)/n [GeV/c ] Fig. 29. v2 /n versus (mT − m0 )/n for minimum bias Au+Au collisions. The quark number scaling appears to be violated when integrating over a wide centrality bin. Mesons and Baryons are fit with the same functional form but with different parameters. The scaling is violated to the extent that n + 1 would give a better agreement i.e. baryon v2 is closer to fourthirds the meson v2 rather than threehalves.
not be as directly connected to the number of constituent quarks as originally conceived. Whether this is indicative of higher fock states, the wavefunction of the hadrons, an as yet unaccounted for experimental systematic error, or something else is yet to be determined. The systematic uncertainties based on the particletype dependence of nonflow are still being investigated. 2.5.3. Systemsize scaling The systemsize dependence of v2 can be studied by looking at the centrality dependence of v2 or by colliding smaller nuclei. Ideal hydro predictions, having a zero meanfreepath assumption, should be independent of the systemsize. In this case, given the same eccentricity, the v2 should be independent of system size. One can try to account for the change in eccentricity by dividing v2 by eccentricity from a model but this introduces a large amount of uncertainty. Another approach is to study the shape of v2 (pT ) to see if that varies.147 The left panel of Fig. 30 shows v2 measured in Au+Au and Cu+Cu collisions for several centrality intervals.129 In the right panel, v2 (pT ) is scaled by 3.1 times the mean v2 for that data set. 3.1hv2 i was taken as a proxy for the eccentricity of the collision system, and this proxy is not inconsistent with models of eccentricity which are quite uncertain. What is best demonstrated by this scaling, is that although the magnitude of v2 changes significantly for the different centralities and systems, the shape of v2 (pT ) is very similar.
358
P. Sorensen
0.25
1
v2(p )
3.1〈v2〉
0.2
0.8
0.15
0.6
0.1
0.4
010% 1020 2030 3040
0.05
0 0
v2(p ) T
T
1
2
3
0.2 open symbols: Cu+Cu closed symbols: Au+Au 0 0
1
2
3
Transverse Momentum pT (GeV/c) Fig. 30. Left panel: v2 versus pT in Au+Au and Cu+Cu collisions for four centrality intervals. Right panel: the same v2 scaled by 3.1 times the pT integrated v2 . The scaling demonstrates that the shape of v2 (pT ) is approximately independent of centrality and system size. A function is fit to the Au+Au data and shown as a solid orange line.
The invariance of v2 (pT ) with systemsize can be taken as an indication that the viscosity of the expanding medium created in heavyion collisions can not be large when v2 is established; Large viscous effects should introduce a systemsize dependence to v2 (pT ) with viscosity causing v2 to saturate at lower pT values in the smaller system.147 Hydrodynamic calculations including viscosity confirm this idea.148–153 To look more carefully for a systemsize dependence in the shape of v2 we plot the ratio of the scaled data to a curve fit to the Au+Au data. The results are shown in Fig. 31. The Cu+Cu data systematically deviate from the Au+Au data. The pT dependence of the ratio indicates that the Cu+Cu data begins to saturate before the Au+Au data. This leads to a ratio that first rises then falls. This would happen the other way around if the Au+Au data saturated first. The uncertainties in the figure are large but the shapes are still significantly different. The systemsize dependence of v2 (pT ) may be a valuable tool for estimating the viscosity of the matter created in heavyion collisions. Although the data on v2 includes many particle types, a wide kinematic range in pT and η, a variety of systemsizes and a wide range in centerofmass energy, we’ve been able to identify several regular features of the data. These include a √ nearly linear rise of v2 at midrapidity with log( sN N ): √ v2 = 0.008 + 0.0084 log( sN N )
(14)
for 0%–20% central Au+Au or Pb+Pb collisions and a pT , mass and particletype
Review of Elliptic Flow v2
359
v2(p )/3.1〈v 〉 2
T
Fit Function 1.1
1
0.9
0
open symbols: Cu+Cu closed symbols: Au+Au 1
2
3
Transverse Momentum pT (GeV/c) Fig. 31. The v2 (pT )/3.1hv2 i data from the right panel of Fig. 30 scaled by a function fit to the Au+Au data. This figure illustrates that their seems to be a significant difference between the shape of v2 (pT ) in Au+Au and Cu+Cu collisions with the Cu+Cu collisions exhibiting a more abrupt turnover – i.e. Cu+Cu data first rises faster with pT then falls faster at pT > 1 GeV/c.
dependence that can be parametrized by v2 /n =
a + bx + cx2 a −(x−d) − , 2 1 + exp e
(15)
where x = (mT −m0 )/n, while v2 (η) to good approximation decreases linearly from it’s maximum at mid rapidity to beam rapidity. This linear rise may be a trivial √ consequence of the log( snn ) dependence of midrapidity v2 or viceversa. 3. Confronting the Hydrodynamic Paradigm with RHIC Data We have discussed the hydrodynamic model extensively in this review as a convenient reference for how well the matter produced in heavyion collision converts spatial deformation into momentum space anisotropy. Hydrodynamic models of heavyion collisions have many uncertainties. These include, uncertain initial conditions, uncertain thermalization times, and uncertain freezeout conditions. A successful description of data using a hydrodynamic model offers the promise of not
360
P Sorensen
(1/2 π)d2N/dpTdy 10
π , PHENIX sqrt(s) 200, 05% π QGP EOS+RQMD, Teaney et al.
v2 /ε
π QGP EOS +PCE, Hirano et al. π QGP EOS +PCE, Ko b et al. π QGP EOS, Huovinen et al.
2
π RG+mixed EOS, Teaney et al. π RG EOS, Huovinen et al.
10 1 10 10 10
1
2
pions
3
0
0.5
1
1.5
2
p PHENIX
(1/2 π)d2N/dpTdy
2.5
3
3.5
s NN=200 GeV, 05%
p QGP EOS RQMD, Teaney et al. (pbar/0.75) QGP EOS PCE, Hirano et al. (pbar/0.75) QGP EOS PCE, Kolb et al.
10
p QGP EOS, Huovinen et al. p RG mixed EOS, Teaney et al. p RG EOS, Huovinen et al.
1
10
10
10
1
2
protons
3
0
05
1
1.5
2
p T (GeV/c)
2.5
3
3.5
v2 /ε
1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0
1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0
π PHENIX sNN 200 GeV, minBias π STAR sNN 200 GeV, minBias π QGP EOS+RQMD, Teaney et al. π QGP EOS +PCE, Hirano et al. π QGP EOS +PCE, Kolb et al. π QGP EOS, Huovinen et al. π RG+mixed EOS, Teaney et al. π RG EOS, Huovinen et al.
pions 05
1
1.5
2
2.5
p PHENIX sNN 200 GeV, minBias p STAR sNN 200 GeV, minBias p QGP EOS+RQMD, Teaney et al. p QGP EOS +PCE, Hirano et al. p QGP EOS +PCE, Ko b et al. p QGP EOS, Huovinen et al. p RG+mixed EOS, Teaney et al. p RG EOS, Huovinen et al.
protons 0.5
1 1.5 p T (GeV/c)
2
2.5
Fig. 32. The four panels show pT spectra and v2 (top and bottom) for pions and protons (left and right). Data from 200 GeV Au+Au collisions is compared to a variety of hydrodynamic models. Most models do not agree with pT spectra and v2 simultaneously.
only establishing the attainment of local equilibrium but also the promise of providing information on the EquationofState of the matter and its transport properties. The uncertainty in the models, however, are large and it has not yet been possible to extract this desired information with satisfactory certainty. In addition, the possibility that significant v2 arises from initialstate effects34,37 could call into question the applicability of hydrodynamics and the need for prolific finalstate rescattering. Measurements of two particle correlations, which have often been interpreted as arising from minijets,96,119 need to be reconciled with the idea of a locally thermalized matter with extensive finalstate rescattering. If the hydrodynamic models and data are irreconcilable, the paradigm will, of course, have to be abandoned. To check for consistency with hydrodynamic models,29,30,154–156 the PHENIX collaboration created a comprehensive comparison between heavyion data on p T spectra and v2 /ε.8 The inclusion of a comparison to HBT data was hampered by the lack of predictions from some of the models. The comparison to pT spectra and v2 is shown in Fig. 32. The left panels show pT spectra with pions in the top panel and protons in the bottom. The right panels show v2 for the same particles. The combination of data on v2 and spectra provide a stringent test for the models as
Review of Elliptic Flow v2
361
some models can reproduce one quantity but only by adjusting parameters in such a way that the agreement with other observables is spoiled. The models shown in the figure differ in several ways. Models that include a phase transition and a QGP phase are shown with solid lines while models without a pure QGP phase are shown as dotted lines. Including this phase transition acts to reduce the value of v2 since the equation of state is soft during the transition. This means that the speed of sound drops (in these models to zero), so that conversion of coordinate space eccentricity to momentum space anisotropy is halted during the phase transition. In the case that the models, do approximately match the pion spectra and v2 , the most directly observable consequence of the lack of a phase transition is on the proton spectra and proton v2 . The proton spectra end up being too soft, and the splitting between proton and pion v2 is reduced with the proton v2 becoming larger. This is somewhat counterintuitive but is a consequence of fixing the parameters to match central data. The models also differ in their treatments of the final hadronic stage. The calculations from Teaney et al. include a hybrid model that uses a hadronic cascade (RQMD) for the final hadronic evolution. Hirano and Kolb do not use such an afterburner but allow the particle abundances to stop changing at a temperature above the temperature at which they stop interacting; chemical freezeout happens before kinetic freezeout. Huovinen on the other hand, maintains chemical and kinetic equilibrium throughout the expansion. These different treatments have very important consequences for the particletype dependence of the pT spectra and v2 . Huovinen’s treatment can reproduce the v2 for pions and protons, but only at the expense of underpredicting the number of protons; a direct consequence of maintaining chemical equilibrium until the final freezeout at a relatively low temperature. The only model which compares well to all the data is Teaney’s model including a QGP phase, a phase transition, and a hadronic phase modeled with RQMD. Such a hybrid model adds significantly to the number of tunable parameters as compared for example to Huovinen’s model. On the other hand, the Teaney model shows that some particle types are less affected by the hadronic phase and therefore less sensitive to some of the uncertainty in freezeout prescription. Figure 33 shows the Teaney calculation with Hydro only versus Hydro+RQMD. The particle species least affected by the inclusion of a hadronic afterburner, are the φmeson and the Ωbaryon. This arises presumably from the small hadronic crosssection for these hadrons. This suggests highstatistics measurements for these particles are a viable way to avoid uncertainties in the effects of hadronic rescattering. Besides the uncertainty in the freezeout prescription, there is uncertainty on the eccentricity of the expanding fireball at the start of the conjectured hydrodynamic evolution. Figure 34 shows a hybrid hydro+cascade model compared to v 2 data.72 Two model curves are shown: one with a ColorGlassCondensate (CGC) initial eccentricity,66 the other with a MonteCarloGlauber (MCG) eccentricity. As discussed previously the CGC eccentricity is larger than the MCG eccentricity; this
362
P. Sorensen
Tslope (GeV) 0.7
Hydro Only
0.6
Hydro+RQMD
Low Mt
Low Mt
High Mt
High Mt
0.5
0.4
0.3
0.2 0.4
K
p φ Λ
+
0.6
0.8
1
Ξ 1.2
Ω 1.4
1.6
1.8
Mass (GeV) Fig. 33. The inverse slope parameter Tslope for a variety of particle species when a heavyion collision is modeled with only a hydrodynamic model compared to a hybrid model which includes a stage modeled with a hadronic cascade model RQMD.
leads to an overprediction for v2 . On the other hand, this hybrid model does not include viscous effects in the QGP phase so the difference between the hybrid+CGC prediction could be related to viscosity. In fact, since viscosity acts to reduce v 2 , the hybrid+MCG curve shows that there is no room for viscosity in this model. This violates the lower bound on viscosity derived based on quantum mechanical arguments158 and also later from string theory.159 Clearly, to estimate the viscosity allowed, or required by the data, the uncertainty on the initial conditions must be 2 reduced. As discussed previously, the measured quantity σtot = δ2 + 2σv22 provides a sensitive test of the models of the initial conditions and needs to be carefully compared to the hydrodynamic model predictions with various initial conditions. 3.1. Transport model fits An approach to circumvent the uncertainties in the hydrodynamic models has been outlined in Refs. 160 and 161 where v2 /ε is fit as a function of S1 dN dy . The fit function is used to infer how close the data come to a saturated value in the collisions with the highest density achieved. The fit function is constrained by how v2 /ε should
Review of Elliptic Flow v2
v2
363
0.2 hydro+cascade, CGC hydro+cascade, Glauber PHOBOS (hit) PHOBOS (track)
0.18 0.16 0.14 0.12 0.1 0.08 0.06 0.04 0.02 0 0
50
100
150
200
250
300
350
400
Npart Fig. 34. A hybrid hydrodynamic model showing the uncertainty in the model calculations arising from two different models for the initial conditions.
approach the high density and the low density limits. One can construct different equations but in a transport code, the following was found to represent the approach to the zero meanfreepath limit well:162 v2 v sat 1 = 2 ε ε 1 + K/K0
(16)
where K is the Knudsen number and K0 is a constant of order one. Figure 35 shows the data and fit in the left panel and the inferred Knudsen number in the right panel. Based on this procedure it is found that RHIC v2 data are still some 20% below the saturation value anticipated within the fit function. This conclusion however, not only depends on the assumptions built into the transport model approach but also the centrality dependence of the eccentricity. The Color Glass Condensate model for example predicts a stronger centrality dependence for the eccentricity than the Monte Carlo Glauber model. As a consequence, this fit implies that if the initial conditions at RHIC are described by the CGC model, then the v2 data is closer to its saturation limit than if the MCG gives the correct description. This is counter intuitive and opposite to the conclusions reached based on real hydrodynamic calculations, which indicate that the larger CGC initial eccentricity allows more room for viscous effects in the QGP phase.148,150 The transport based fit circumvents the actual solving of hydrodynamics but the conclusions are dependent on the centrality dependence of the initial eccentricity which is strongly model dependent. The fit also includes the speed of sound as a free parameter. This
364
P. Sorensen
0.3
v2 εCGC
10
Knudsen Number
8
Hydro Limit 0.2 6
4 0.1
2
0 0
0.2
0.4
0.6
0.8
1
1.2
1.4
0 0
0.2
0.4
0.6
08
1
1.2
1.4
1
(1/S)(dN/dy) (mb ) Fig. 35. Left panel: v2 in Au+Au and Cu+Cu collisions scaled by eccentricity calculated in a CGC framework and plotted versus (1/S)(dN/dy). The fit function and hydrodynamic limit are explained in the text. Right panel: The effective Knudsen extracted from the data and fit in the left panel.
effectively leads to an equation of state which has no phase transition but which is allowed to vary in the fit. A complimentary and perhaps better method for accessing the Knudsen number and the viscosity is to study the shape change of v2 (pT ) for different systemsizes which avoids the uncertainty in the eccentricity.147 This is a work currently in progress. 3.2. Viscous hydrodynamics The apparent success of ideal hydrodynamic models to describe the gross features of RHIC data has led to the inference of small viscosity and the claim of the discovery of the perfect liquid at RHIC. The perfect liquid announcement was listed as the top physics story of 2005 by the American Institute of Physics and was widely covered in the popular press. Much recent work has gone towards including viscous effects in hydrodynamic calculations so that the viscosity can be more accurately estimated.148–153 Figure 36 shows one such calculation. The top panel shows results when the hydrodynamic evolution starts from Glauber initial conditions while the bottom panel shows the case of CGC initial conditions. The results that come closest to the STAR data81 are given by η/s = 0.08 for the Glauber initial conditions and η/s = 0.16 for the CGC initial conditions. The STAR nonflow corrected data
Review of Elliptic Flow v2
365
Fig. 36. Hydrodynamic calculations including viscous effects. The top panel shows v 2 (pT ) for the case that the initial conditions are described with a Monte Carlo Glauber model. The bottom panel is based on Color Glass Condensate initial conditions. The curves show results for different values of η/s, the ratio of shear viscosity to entropy.
are from Fig. 4 of Ref. 80. The 10%–40% central data from that reference was scaled to account for the difference between the 10%–40% centrality interval and the 0%–80% (minbias) centrality interval. The larger η/s inferred based on the CGC model arises from the larger initial eccentricity which leaves more room for viscous effects that tend to reduce the v2 . This contradicts the conclusions drawn from the transport model inspired fit, which allows the equationofstate to change for the two different initial conditions. The pT dependence of the data is also better captured in the larger viscosity CGC scenario. The larger viscosity inferred from the
366
P. Sorensen
CGC initial conditions gives a more pronounced turn over of v2 (pT ) which better describes the pT dependence of v2 . The comparison shown in Fig. 36 shows that hydrodynamic models including viscosity have a good chance of reproducing RHIC 1 2 where 4π is data as long as the shear viscosity to entropy ratio η/s is less than 4π the conjectured lower limit. 3.3. Fluctuating initial conditions The comparison of v2 and other RHIC data to hydrodynamic models seems to indicate that when viscous corrections are included, a successful description of the data may be possible. There is uncertainty in this comparison, however, related to uncertainties in the initial conditions and in the freezeout prescription. The uncertainty in the initial conditions can be addressed experimentally with measurements of v2 fluctuations which in turn require an understanding of nonflow correlations; The experimentally accessible information appears to reduce to v 2 {2}2 and v2 {2}2 −v2 {4}4 = δ2 +2σv22 . An alternative approach may be for hydrodynamic models to predict v2 {2} and v2 {4} by including correlations and fluctuations in the models. Progress has been made in this direction. Early work relating to the effect of fluctuations in the initial conditions on hydrodynamic calculations was carried out using the NeXSPheRIO hydrodynamic model.67,163 The initial eccentricity fluctuations were indeed found to lead to v2 fluctuations as shown in Fig. 37. Later it was suggested that correlations in the initial conditions could lead to vn fluctuations of even and odd orders of n that would manifest themselves as nonsinusoidal, apparently nontrivial, twoparticle correlations as seen in the RHIC data. 69,164 Subsequent work following through on this idea shows that hydrodynamic models with fluctuating initial conditions do lead to twoparticle correlations with structure beyond a simple cos(2∆φ) shape.165 The correlation structure arising from the fluctuations in the initial conditions is shown in the right panel of Fig. 37. The model exhibits many of the features seen in the data including a jetlike peak, a nearside ridge, and an awayside ridge shifted away from ∆φ = π. All this structure arises without the explicit inclusion of jets in the model. The apparently exotic correlations do not appear in the model when a smooth initial condition is used. This calculation illustrates the importance of accounting for fluctuations in the initial conditions when interpreting the correlation landscape. It also demonstrates that complex interactions between jets and the medium, including machcones, are not needed to explain the correlations data nor is the concept of minijets necessarily required. In light of the NeXSPheRIO calculations, the highly structured correlation landscape at RHIC should not necessarily be taken as an invalidation of the hydrodynamic models. The correlations may simply reflect the need to abandon certain approximations, including the approximation of infinitely smooth initial conditions. Besides comparing to twoparticle correlation data, these models can be used to calculate v2 {2} and v2 {4} to directly compare to data. It will be interesting to see how
Review of Elliptic Flow v2
367
50
Au+Au, s1/2=130A GeV, b=10fm, EoS= QGP+RG =0.0515 δv2= 0.0199
number of events
40
30
20
10
0 −0.1
−0 05
0
0 05
0.1
0.15
0.2
v2
1 d 2N Ntrig d∆ηd∆φ
0.6 0.4
6 4
0.2 2 0
0 3
2
2 1
0
∆φ
1
∆η
4 2
3
6
Fig. 37. Hydrodynamic calculations including fluctuations in the initial conditions. The left panel shows the eventbyevent v2 distribution. The right panel shows the twoparticle correlations that arise due to the correlations in the initial energy density.
the correlation landscape in this model depends on the parameters of the model, in particular, the thermalization time and the freezeout time. The connection of vn fluctuations (related to twoparticle correlations) to the lifetime of the system was first pointed out by Mishra et al.166 In that reference the authors also introduce p the anaology between hvn2 i fluctuations and the power spectrum of the Cosmic Microwave Background Radiation.
368
P. Sorensen
3.4. Addressing uncertainties Uncertainties in the freezeout prescription and the effects of the hadronic phase can be experimentally addressed through precise measurements of φmesons and Ωbaryons. Models indicate that due to their small hadronic crosssections, these hadrons are minimally influenced by the hadronic phase and reflect well the QGP phase. In addition, heavy flavor hadrons may help determine or provide a crosscheck for the transport properties of the QGP. Another approach to extracting the viscosity is by studying the shape of v2 (pT ) versus system size. This approach does not rely on a model for the initial eccentricity. Uncertainties in the eccentricity and the initial conditions can be reduced through measurements of v2 fluctuations and twoparticle correlations. These studies are ongoing. One can also measure v n fluctuations for arbitrary n value. These are of course related to the twoparticle correlation landscape which has already been extensively studied at RHIC. It will be of great interest to see how the correlation landscape predicted in hydrodynamic models with fluctuating initial conditions changes depending on the model parameters. The correlations data may help constrain quanties like the lifetime of the system. The studies listed above, along with a beamenergy scan at RHIC and the first data from LHC, will allow for more progress in understanding the matter created in heavyion collisions and its subsequent evolution.
4. Summary In this review, v2 measurements were presented as a method for studying spacemomentum correlations in heavyion collisions. The measurements of v2 indicate the eccentricity in the initial overlap region is transferred efficiently to momentumspace. At top RHIC energy, the conversion is near that expectated from zero meanfreepath hydrodynamic predictions. The comparisons of data to hydrodynamics, however, depends on model calculations of the initial eccentricity. Several models for the initial eccentricity have been discussed. The mass, and pT dependence of v2 at pT < 1 GeV/c is found to be consistent with emission from a boosted source. Above that, the particle type dependence of v2 exhibits a dependence on the number of constituent quarks in the hadron, with baryons obtaining v2 values larger than mesons. The relationship between twoparticle correlations, v2 , and v2 fluctuations has also been discussed. Calculations showing that some of the structures in twoparticle correlations can be ascribed to fluctuations in the initial conditions, have been reviewed. Measurements of correlations and v2 fluctuations can therefore be used to constrain models for the initial conditions. These constraints, along with improved measurements of the shape of v2 (pT ) as a function of systemsize, improved measurements of φ and Ω v2 , measurements of v2 for heavyflavor hadrons, measurements at LHC energies, and a beamenergy scan at RHIC will further improve our understanding of the properties of the matter created in heavyion collisions.
Review of Elliptic Flow v2
369
Acknowledgements I’m grateful to Yan Lu and Alexander Wetzler for providing the code to produce several figures. My thanks go to Tuomas Lappi for providing Figure 7 and to Paul Romatschke, Peter Filip, Arthur Poskanzer, and Anton Andronic for comments provided on the text. I also want to thank Arthur Poskanzer for the advice he gives me. I always appreciate it even though sometimes I fail to follow it (usually to my own disadvantage). I also appreciate the many discussions held with the STAR Flow Group and the STAR BulkCorrelations Physics Working Group. I want to thank all the people who have contributed to the study of elliptic flow and its interpretation, some of whom, I have failed to recognize in this review. I refer the reader to a presentation on the history of flow measurements167 for more context. References 1. W. Reisdorf and H. G. Ritter, Ann. Rev. Nucl. Part. Sci. 47, 663 (1997). 2. N. Herrmann, J. P. Wessels and T. Wienold, Ann. Rev. Nucl. Part. Sci. 49, 581 (1999). 3. J. Y. Ollitrault, Phys. Rev. D 46, 229 (1992). 4. J. Y. Ollitrault, Nucl. Phys. A 590, 561C (1995). 5. S. A. Voloshin, Phys. Lett. B 632, 490 (2006) [arXiv:nuclth/0312065]. 6. M. A. Lisa, S. Pratt, R. Soltz and U. Wiedemann, Ann. Rev. Nucl. Part. Sci. 55, 357 (2005) [arXiv:nuclex/0505014]. 7. S. A. Voloshin and W. E. Cleland, Phys. Rev. C 53, 896 (1996) [arXiv:nuclth/9509025];Phys. Rev. C 54, 3212 (1996) [arXiv:nuclth/9606033]. 8. J. Adams et al. [STAR Collaboration], Nucl. Phys. A 757, 102 (2005) [arXiv:nuclex/0501009]; B. B. Back et al., Nucl. Phys. A 757, 28 (2005) [arXiv:nuclex/0410022]; I. Arsene et al. [BRAHMS Collaboration], Nucl. Phys. A 757, 1 (2005) [arXiv:nuclex/0410020]; K. Adcox et al. [PHENIX Collaboration], Nucl. Phys. A 757, 184 (2005) [arXiv:nuclex/0410003]. 9. S. Voloshin and Y. Zhang, Z. Phys. C 70, 665 (1996) [arXiv:hepph/9407282]. 10. S. A. Voloshin and A. M. Poskanzer, Phys. Lett. B 474, 27 (2000) [arXiv:nuclth/9906075]. 11. J. Adams et al. [STAR Collaboration], Phys. Rev. C 72, 014904 (2005) [arXiv:nuclex/0409033]. 12. S. A. Voloshin, A. M. Poskanzer and R. Snellings, arXiv:0809.2949 [nuclex]. 13. A. Wetzler, private communication (2005). 14. J. Barrette et al. [E877 Collaboration], Phys. Rev. Lett. 73, 2532 (1994) [arXiv:hepex/9405003]. 15. J. Barrette et al. [E877 Collaboration], Phys. Rev. C 55, 1420 (1997) [Erratumibid. C 56, 2336 (1997)] [arXiv:nuclex/9610006]. 16. C. Pinkenburg et al. [E895 Collaboration], Phys. Rev. Lett. 83, 1295 (1999) [arXiv:nuclex/9903010]. 17. P. Chung et al. [E895 Collaboration], Phys. Rev. C 66, 021901 (2002) [arXiv:nuclex/0112002]. 18. K. H. Ackermann et al. [STAR Collaboration], Phys. Rev. Lett. 86, 402 (2001) [arXiv:nuclex/0009011]. 19. S. Manly et al. [PHOBOS Collaboration], Quark Matter ’02, Nucl. Phys. A715, 611c (2003) [arXiv:nuclex/0210036].
370
P. Sorensen
20. J. Slivova [CERES Collaboration], Quark Matter ’02, Nucl. Phys. A 715, 615c (2003) [arXiv:nuclex/0212013]. 21. C. Alt et al. [NA49 Collaboration], Phys. Rev. C 68, 034903 (2003) [arXiv:nuclex/0303001]. 22. A. Andronic et al. [FOPI Collaboration], Phys. Lett. B 612, 173 (2005) [arXiv:nuclex/0411024]. 23. A. Andronic and P. BraunMunzinger, Lect. Notes Phys. 652, 35 (2004) [arXiv:hepph/0402291]. 24. L. D. Landau, Izv. Akad. Nauk Ser. Fiz. 17, 51 (1953). 25. P. F. Kolb, J. Sollfrank and U. W. Heinz, Phys. Lett. B 459, 667 (1999) [arXiv:nuclth/9906003]. 26. D. Teaney and E. V. Shuryak, Phys. Rev. Lett. 83, 4951 (1999) [arXiv:nuclth/9904006]. 27. P. F. Kolb, J. Sollfrank and U. W. Heinz, Phys. Rev. C 62, 054909 (2000) [arXiv:hepph/0006129]. 28. P. F. Kolb, P. Huovinen, U. W. Heinz and H. Heiselberg, Phys. Lett. B 500, 232 (2001) [arXiv:hepph/0012137]. 29. D. Teaney, J. Lauret and E. V. Shuryak, Phys. Rev. Lett. 86, 4783 (2001) [arXiv:nuclth/0011058]. 30. P. Huovinen, P. F. Kolb, U. W. Heinz, P. V. Ruuskanen and S. A. Voloshin, Phys. Lett. B 503, 58 (2001) [arXiv:hepph/0101136]. 31. D. Teaney, J. Lauret and E. V. Shuryak, arXiv:nuclth/0110037. 32. D. Teaney, J. Lauret and E. V. Shuryak, Nucl. Phys. A 698, 479 (2002) [arXiv:nuclth/0104041]. 33. U. W. Heinz and P. F. Kolb, Nucl. Phys. A 702, 269 (2002) [arXiv:hepph/0111075]. 34. K. G. Boreskov, A. B. Kaidalov and O. V. Kancheli, arXiv:0809.0625 [hepph]. 35. B. Z. Kopeliovich, A. H. Rezaeian and I. Schmidt, arXiv:0809.4327 [hepph]. 36. D. Teaney and R. Venugopalan, Phys. Lett. B 539, 53 (2002) [arXiv:hepph/0203208]. 37. A. Krasnitz, Y. Nara and R. Venugopalan, Phys. Lett. B 554, 21 (2003) [arXiv:hepph/0204361]. 38. Y. Pang, T. J. Schlagel and S. H. Kahana, Phys. Rev. Lett. 68, 2743 (1992). 39. K. Geiger and B. Muller, Nucl. Phys. B 369, 600 (1992). 40. K. Geiger, Comput. Phys. Commun. 104, 70 (1997) [arXiv:hepph/9701226]. 41. H. Sorge, Phys. Rev. Lett. 78, 2309 (1997) [arXiv:nuclth/9610026]. 42. M. Bleicher et al., J. Phys. G 25, 1859 (1999) [arXiv:hepph/9909407]. 43. S. A. Bass et al., Prog. Part. Nucl. Phys. 41, 255 (1998) [Prog. Part. Nucl. Phys. 41, 225 (1998)] [arXiv:nuclth/9803035]. 44. G. Burau, J. Bleibel, C. Fuchs, A. Faessler, L. V. Bravina and E. E. Zabrodin, Phys. Rev. C 71, 054905 (2005) [arXiv:nuclth/0411117]. 45. B. Zhang, Comput. Phys. Commun. 109, 193 (1998) [arXiv:nuclth/9709009]. 46. D. Molnar and M. Gyulassy, Nucl. Phys. A 697, 495 (2002) [Erratumibid. A 703, 893 (2002)] [arXiv:nuclth/0104073]. 47. F. Retiere and M. A. Lisa, Phys. Rev. C 70, 044907 (2004) [arXiv:nuclth/0312024]. 48. J. Y. Ollitrault, arXiv:nuclex/9711003. 49. J. Y. Ollitrault, Nucl. Phys. A 638, 195 (1998) [arXiv:nuclex/9802005]. 50. A. M. Poskanzer and S. A. Voloshin, Phys. Rev. C 58, 1671 (1998) [arXiv:nuclex/9805001]. 51. P. Danielewicz and G. Odyniec, Phys. Lett. B 157, 146 (1985). 52. J. Y. Ollitrault, Phys. Rev. D 48, 1132 (1993) [arXiv:hepph/9303247].
Review of Elliptic Flow v2
371
53. S. Wang et al., Phys. Rev. C 44, 1091 (1991). 54. K. Adcox et al. [PHENIX Collaboration], Phys. Rev. Lett. 89, 212301 (2002) [arXiv:nuclex/0204005]. 55. S. S. Adler et al. [PHENIX Collaboration], Phys. Rev. Lett. 97, 052301 (2006) [arXiv:nuclex/0507004]. 56. J. Bielcikova, S. Esumi, K. Filimonov, S. Voloshin and J. P. Wurm, Phys. Rev. C 69, 021901 (2004) [arXiv:nuclex/0311007]. 57. C. Adler et al. [STAR Collaboration], Phys. Rev. Lett. 90, 082302 (2003) [arXiv:nuclex/0210033]. 58. C. Adler et al. [STAR Collaboration], Phys. Rev. Lett. 90, 032301 (2003) [arXiv:nuclex/0206006]. 59. R. S. Bhalerao and J. Y. Ollitrault, Phys. Lett. B 641, 260 (2006) [arXiv:nuclth/0607009]. 60. M. L. Miller, K. Reygers, S. J. Sanders and P. Steinberg, Ann. Rev. Nucl. Part. Sci. 57, 205 (2007) [arXiv:nuclex/0701025]. 61. X. N. Wang and M. Gyulassy, Phys. Rev. D 44, 3501 (1991). 62. H. J. Drescher, S. Ostapchenko, T. Pierog and K. Werner, Phys. Rev. C 65, 054902 (2002) [arXiv:hepph/0011219]. 63. H. Kowalski, T. Lappi and R. Venugopalan, Phys. Rev. Lett. 100, 022303 (2008) [arXiv:0705.3047 [hepph]]. 64. H. Kowalski and D. Teaney, Phys. Rev. D 68, 114005 (2003) [arXiv:hepph/0304189]. 65. T. Lappi and R. Venugopalan, Phys. Rev. C 74, 054905 (2006) [arXiv:nuclth/0609021]. 66. H. J. Drescher, A. Dumitru, A. Hayashigaki and Y. Nara, Phys. Rev. C 74, 044905 (2006) [arXiv:nuclth/0605012]. 67. C. E. Aguiar, Y. Hama, T. Kodama and T. Osada, Nucl. Phys. A 698, 639 (2002) [arXiv:hepph/0106266]. 68. M. Miller and R. Snellings, arXiv:nuclex/0312008. 69. P. Sorensen, arXiv:0808.0503 [nuclex]. 70. S. Manly et al. [PHOBOS Collaboration], Nucl. Phys. A 774, 523 (2006) [arXiv:nuclex/0510031]. 71. S. A. Voloshin, A. M. Poskanzer, A. Tang and G. Wang, Phys. Lett. B 659, 537 (2008) [arXiv:0708.0800 [nuclth]]. 72. T. Hirano, U. W. Heinz, D. Kharzeev, R. Lacey and Y. Nara, Phys. Lett. B 636, 299 (2006) [arXiv:nuclth/0511046]. 73. C. Schmidt, PoS C POD2006 (2006) 002 [arXiv:heplat/0701019]; F. Karsch, PoS C POD07 (2007) 026 [arXiv:0711.0656 [heplat]]; PoS LAT2007 (2007) 015 [arXiv:0711.0661 [heplat]]. 74. J.D. Bjorken, FERMILABPUB82059THY (1982). 75. B. B. Back et al. [PHOBOS Collaboration], Phys. Rev. C 72, 051901 (2005) [arXiv:nuclex/0407012]. 76. C. Adler et al. [STAR Collaboration], Phys. Rev. Lett. 87, 182301 (2001) [arXiv:nuclex/0107003]. 77. C. Adler et al. [STAR Collaboration], Phys. Rev. Lett. 89, 132301 (2002) [arXiv:hepex/0205072]. 78. J. Adams et al. [STAR Collaboration], Phys. Rev. Lett. 92, 052302 (2004) [arXiv:nuclex/0306007]. 79. S. S. Adler et al. [PHENIX Collaboration], Phys. Rev. Lett. 91, 182301 (2003) [arXiv:nuclex/0305013].
372
P. Sorensen
80. B. I. Abelev et al. [STAR Collaboration], Phys. Rev. C 77, 054901 (2008) [arXiv:0801.3466 [nuclex]]. 81. B. I. Abelev et al. [the STAR Collaboration], Phys. Rev. C 75, 054906 (2007) [arXiv:nuclex/0701010]. 82. P. R. Sorensen, arXiv:nuclex/0309003. 83. J. Adams et al. [STAR Collaboration], Phys. Rev. Lett. 95, 122301 (2005) [arXiv:nuclex/0504022]. 84. S. S. Adler et al. [PHENIX Collaboration], Phys. Rev. Lett. 94, 232302 (2005). 85. X. N. Wang and M. Gyulassy, Phys. Rev. Lett. 68, 1480 (1992). 86. X. N. Wang, Phys. Rev. C 63, 054902 (2001) [arXiv:nuclth/0009019]. 87. M. Gyulassy, I. Vitev, X. N. Wang and P. Huovinen, Phys. Lett. B 526, 301 (2002) [arXiv:nuclth/0109063]. 88. M. Gyulassy, I. Vitev and X. N. Wang, Phys. Rev. Lett. 86, 2537 (2001) [arXiv:nuclth/0012092]. 89. J. Adams et al. [STAR Collaboration], Phys. Rev. Lett. 93, 252301 (2004) [arXiv:nuclex/0407007]. 90. A. Drees, H. Feng and J. Jia, Phys. Rev. C 71, 034909 (2005) [arXiv:nuclth/0310044]. 91. C. Adler et al. [STAR Collaboration], Phys. Rev. C 66, 034904 (2002) [arXiv:nuclex/0206001]. 92. N. Borghini, P. M. Dinh and J. Y. Ollitrault, arXiv:nuclex/0110016. 93. J. Putschke, J. Phys. G 34, S679 (2007) [arXiv:nuclex/0701074]. 94. S. A. Voloshin [STAR Collaboration], J. Phys. G 34, S883 (2007) [arXiv:nuclex/0701038]. 95. T. A. Trainor and D. T. Kettler, Int. J. Mod. Phys. E 17 (2008) 1219 [arXiv:0704.1674 [hepph]]. 96. J. Adams et al. [STAR Collaboration], Phys. Rev. C 73, 064907 (2006) [arXiv:nuclex/0411003]. 97. D. Kharzeev and M. Nardi, Phys. Lett. B 507, 121 (2001) [arXiv:nuclth/0012025]. 98. B. B. Back et al. [PHOBOS Collaboration], Phys. Rev. C 65, 031901 (2002) [arXiv:nuclex/0105011]. 99. B. I. Abelev et al. [STAR Collaboration], Phys. Rev. Lett. 99, 112301 (2007) [arXiv:nuclex/0703033]. 100. S. Afanasiev et al. [PHENIX Collaboration], Phys. Rev. Lett. 99, 052301 (2007) [arXiv:nuclex/0703024]. 101. Y. L. Dokshitzer and D. E. Kharzeev, Phys. Lett. B 519, 199 (2001) [arXiv:hepph/0106202]. 102. M. Djordjevic and M. Gyulassy, Nucl. Phys. A 733, 265 (2004) [arXiv:nuclth/0310076]. 103. S. S. Adler et al. [PHENIX Collaboration], Phys. Rev. Lett. 96, 032301 (2006) [arXiv:nuclex/0510047]. 104. B. I. Abelev et al. [STAR Collaboration], Phys. Rev. Lett. 98, 192301 (2007) [arXiv:nuclex/0607012]. 105. P. R. Sorensen and X. Dong, Phys. Rev. C 74, 024902 (2006) [arXiv:nuclth/0512042]. 106. S. S. Adler et al. [PHENIX Collaboration], Phys. Rev. C 72, 024901 (2005) [arXiv:nuclex/0502009]. 107. A. Adare et al. [PHENIX Collaboration], Phys. Rev. Lett. 98, 172301 (2007) [arXiv:nuclex/0611018]. 108. G. D. Moore and D. Teaney, Phys. Rev. C 71, 064904 (2005) [arXiv:hepph/0412346].
Review of Elliptic Flow v2
373
109. H. van Hees, V. Greco and R. Rapp, Phys. Rev. C 73, 034913 (2006) [arXiv:nuclth/0508055]; H. van Hees, M. Mannarelli, V. Greco and R. Rapp, Phys. Rev. Lett. 100, 192301 (2008) [arXiv:0709.2884 [hepph]]. 110. P. Sorensen [STAR Collaboration], J. Phys. G 34, S897 (2007) [arXiv:nuclex/0612021]. 111. B. Alver et al. [PHOBOS Collaboration], arXiv:nuclex/0702036. 112. P. Sorensen [STAR Collaboration], arXiv:0808.0356 [nuclex]. 113. B. Alver et al., Phys. Rev. C 77, 014906 (2008) [arXiv:0711.3724 [nuclex]]. 114. N. Borghini, P. M. Dinh and J. Y. Ollitrault, Phys. Rev. C 64, 054901 (2001) [arXiv:nuclth/0105040]. 115. R. S. Bhalerao, N. Borghini and J. Y. Ollitrault, Nucl. Phys. A 727, 373 (2003) [arXiv:nuclth/0310016]. 116. N. Borghini, R. S. Bhalerao and J. Y. Ollitrault, J. Phys. G 30, S1213 (2004) [arXiv:nuclth/0402053]. 117. X. l. Zhu, M. Bleicher and H. Stoecker, Phys. Rev. C 72, 064911 (2005) [arXiv:nuclth/0509081]. 118. A. Bilandzic, N. van der Kolk, J. Y. Ollitrault and R. Snellings, arXiv:0801.3915 [nuclex]. 119. T. A. Trainor, Mod. Phys. Lett. A 23, 569 (2008) [arXiv:0708.0792 [hepph]]. 120. S. Mrowczynski and E. V. Shuryak, Acta Phys. Polon. B 34, 4241 (2003) [arXiv:nuclth/0208052]. 121. W. Broniowski, P. Bozek and M. Rybczynski, Phys. Rev. C 76, 054905 (2007) [arXiv:0706.4266 [nuclth]]. 122. S. Vogel, G. Torrieri and M. Bleicher, arXiv:nuclth/0703031. 123. M. Daugherity [STAR Collaboration], J. Phys. G 35, 104090 (2008) [arXiv:0806.2121 [nuclex]]. 124. B. Alver et al. [PHOBOS Collaboration], Phys. Rev. Lett. 98, 242302 (2007) [arXiv:nuclex/0610037]. 125. P. Sorensen [STAR Collaboration], PoS CPOD2006 (2006) 019 [arXiv:nuclex/0701028]; T. Satogata, et.al., PoS CPOD2007 (2007) 051. 126. H. Sorge, Phys. Rev. Lett. 82, 2048 (1999) [arXiv:nuclth/9812057]. 127. H. Stocker, arXiv:0710.5089 [hepph]. 128. P. Sorensen [STAR Collaboration], J. Phys. G 30, S217 (2004). 129. A. Adare et al. [PHENIX Collaboration], Phys. Rev. Lett. 98, 162301 (2007) [arXiv:nuclex/0608033]. 130. S. A. Voloshin, Nucl. Phys. A 715, 379 (2003) [arXiv:nuclex/0210014]. 131. Z. w. Lin and C. M. Ko, Phys. Rev. C 65, 034904 (2002) [arXiv:nuclth/0108039]. 132. D. Molnar and S. A. Voloshin, Phys. Rev. Lett. 91, 092301 (2003) [arXiv:nuclth/0302014]. 133. R. C. Hwa and C. B. Yang, Phys. Rev. C 67, 064902 (2003). 134. R. J. Fries, B. Muller, C. Nonaka and S. A. Bass, Phys. Rev. C 68, 044902 (2003). 135. V. Greco, C. M. Ko and P. Levai, Phys. Rev. C 68, 034904 (2003). 136. V. Greco, C. M. Ko and P. Levai, Phys. Rev. Lett. 90, 202302 (2003). 137. R. J. Fries, B. Muller, C. Nonaka and S. A. Bass, Phys. Rev. Lett. 90, 202303 (2003). 138. S. Pratt and S. Pal, Nucl. Phys. A 749, 268 (2005) [Phys. Rev. C 71, 014905 (2005)] [arXiv:nuclth/0409038]. 139. L. Ravagli, H. van Hees and R. Rapp, arXiv:0806.2055 [hepph]. 140. R. J. Fries, V. Greco and P. Sorensen, arXiv:0807.4939 [nuclth]. 141. S. S. Adler et al. [PHENIX Collaboration], Phys. Rev. Lett. 91, 172301 (2003) [arXiv:nuclex/0305036].
374
P. Sorensen
142. S. J. Brodsky and A. Sickles, Phys. Lett. B 668, 111 (2008) [arXiv:0804.4608 [hepph]]. 143. B. I. Abelev et al. [STAR Collaboration], Phys. Lett. B 655, 104 (2007) [arXiv:nuclex/0703040]. 144. V. Greco and C. M. Ko, Phys. Rev. C 70, 024901 (2004). 145. B. Muller, R. J. Fries and S. A. Bass, Phys. Lett. B 618, 77 (2005). 146. Y. Lu et al., J. Phys. G 32, 1121 (2006) [arXiv:nuclth/0602009]. 147. D. Teaney, Phys. Rev. C 68, 034913 (2003) [arXiv:nuclth/0301099]. 148. P. Romatschke and U. Romatschke, Phys. Rev. Lett. 99, 172301 (2007) [arXiv:0706.1522 [nuclth]]. 149. M. Luzum and P. Romatschke, Phys. Rev. C 78, 034915 (2008) [arXiv:0804.4015 [nuclth]]. 150. H. Song and U. W. Heinz, Phys. Lett. B 658, 279 (2008) [arXiv:0709.0742 [nuclth]]. 151. H. Song and U. W. Heinz, Phys. Rev. C 77, 064901 (2008) [arXiv:0712.3715 [nuclth]]. 152. K. Dusling and D. Teaney, Phys. Rev. C 77, 034905 (2008) [arXiv:0710.5932 [nuclth]]. 153. H. Song and U. W. Heinz, Phys. Rev. C 78, 024902 (2008) [arXiv:0805.1756 [nuclth]]. 154. P. F. Kolb and R. Rapp, Phys. Rev. C 67, 044903 (2003) [arXiv:hepph/0210222]. 155. T. Hirano and K. Tsuda, Phys. Rev. C 66, 054905 (2002) [arXiv:nuclth/0205043]. 156. T. Hirano and Y. Nara, Nucl. Phys. A 743, 305 (2004) [arXiv:nuclth/0404039]. 157. H. J. Drescher and Y. Nara, Phys. Rev. C 76, 041903 (2007) [arXiv:0707.0249 [nuclth]]. 158. P. Danielewicz and M. Gyulassy, Phys. Rev. D 31, 53 (1985). 159. P. Kovtun, D. T. Son and A. O. Starinets, Phys. Rev. Lett. 94, 111601 (2005) [arXiv:hepth/0405231]. 160. R. S. Bhalerao, J. P. Blaizot, N. Borghini and J. Y. Ollitrault, Phys. Lett. B 627, 49 (2005) [arXiv:nuclth/0508009]. 161. H. J. Drescher, A. Dumitru, C. Gombeaud and J. Y. Ollitrault, Phys. Rev. C 76, 024905 (2007) [arXiv:0704.3553 [nuclth]]. 162. C. Gombeaud and J. Y. Ollitrault, Phys. Rev. C 77, 054904 (2008) [arXiv:nuclth/0702075]. 163. R. Andrade, F. Grassi, Y. Hama, T. Kodama and O. . J. Socolowski, Phys. Rev. Lett. 97, 202302 (2006) [arXiv:nuclth/0608067]. 164. P. Sorensen, arXiv:0811.2959 [nuclex]. 165. J. Takahashi, B. M. Tavares, W. L. Qian, F. Grassi, Y. Hama, T. Kodama and N. Xu, arXiv:0902.4870 [nuclth]. 166. A. P. Mishra, R. K. Mohapatra, P. S. Saumia and A. M. Srivastava, Phys. Rev. C 77, 064902 (2008) [arXiv:0711.1323 [hepph]]. 167. A. M. Poskanzer, http://rnc.lbl.gov/Poskanzer/FlowHistory.ppt (2004).
PREDICTIONS FOR THE HEAVYION PROGRAMME AT THE LARGE HADRON COLLIDER
´ NESTOR ARMESTO Departamento de F´ısica de Part´ıculas and Instituto Galego de F´ısica de Altas Enerx´ıas, Universidade de Santiago de Compostela, 15706 Santiago de Compostela, Spain nestor.armesto@usc.es
I review the main predictions for the heavyion programme at the Large Hadron Collider (LHC) at CERN, as available in early April 2009. I begin by remembering the standard claims made in view of the experimental data measured at the Super Proton Synchrotron (SPS) at CERN and at the Relativistic Heavy Ion Collider (RHIC) at the BNL. These claims will be used for later discussion of the new opportunities at the LHC. Next I review the generic, qualitative expectations for the LHC. Then I turn to quantitative predictions: First I analyze observables which characterize directly the medium produced in the collisions — bulk observables or soft probes: multiplicities, collective flow, hadrochemistry at low transverse momentum, correlations and fluctuations. Second, I move to calibrated probes of the medium i.e. typically those whose expectation in the absence of any medium can be described in Quantum Chromodynamics (QCD) using perturbative techniques (pQCD), usually called hard probes. I discuss particle production at large transverse momentum and jets, heavyquark and quarkonium production, and photons and dileptons. Finally, after a brief review of pA collisions, I end with a summary and a discussion about the potentiality of the measurements at the LHC — particularly those made during the first run — to further substantiate or, on the contrary, disproof the picture of the medium that has arisen from the confrontation between the SPS and RHIC data, and theoretical models.
1. Introduction The experimental programme for the study of ultrarelativistic heavyion collisions started in 1986 at the Super Proton Synchrotron (SPS) at CERN. It accelerated protons and ions (up to Pb), at plab ≤ 158 GeV per nucleon in the case of Pb.a The next step was the Relativistic Heavy Ion Collider (RHIC) at the BNL, which √ began in 2000, accelerating protons and ions up to AuAu collisions at sN N = 200 GeV. Both experimental programmes have allowed for the extraction of important conclusions about the properties of the strongly interacting matter produced in such collisions.1–5 The next step in the near future, apart from RHIC upgrades6 and the energy and collision species scan at the SPS,7 is the heavyion programme at the Large Hadron Collider (LHC) at CERN.8 It will accelerate ions as heavy as Pb (A = 208, a Natural
units ~ = c = 1, and kB = 1 will be used throughout this manuscript. 375
376
N. Armesto
Z = 82), with energies √
sN N = 2
Z × 7 TeV ≃ 5.5 TeV for PbPb, A
(1)
with a total centerofmass energy of 1.15 PeV. The nominal peak luminosity will be L0 = 1027 cm−2 s−1 , with hLi/L0 = 0.5 and a estimated running time 106 s/year.b Collisions of other ions and asymmetric collisions like pPb9 are possible, the latter with a shift in the centerofmass rapidity with respect to the rapidity in the laboratory given by δy =
1 Z1 A2 ln 2 Z2 A1
(2)
A2 1 for A Z1 A Z2 B collisions. While the first proton beams circulated along the LHC ring in September 2008 and the first pp collisions are expected for autumn 2009, the first PbPb collisions are only expected for the second half of 2010. Three out of the four large experiments at the LHC: ALICE, ATLAS and CMS, will measure PbPb collisions.10–13 While ALICE is a dedicated experiment to nucleusnucleus collisions, both ATLAS and CMS will offer detector capabilities complementary to each other and to ALICE. They will provide a wide range of measurements covering all the main relevant observables in heavyion collisions. Measurements in pp and nucleusnucleus collisions at roughly the same unexplored top energy will be, for the first time, performed using the same accelerator and detectors. The increase in centerofmass energy of almost a factor 30 with respect to RHIC, together with the complementary detector capabilities, will offer new measurements with respect to those presently available. As evident examples:
• The yield of particles with large mass or transverse momentum — hard probes9,14–16 — will be much more abundant, and some of them will be measured for the first time (with high statistics) in heavyion collisions, like Υ or Z 0 + jet production (see Fig. 1, taken from Refs. 17 and 18). • Calorimeter capabilities of ATLAS and CMS12,13 will allow for measuring jets both at central and noncentral rapidities.c ALICE will also be equipped with an electromagnetic calorimeter.18 • The kinematical coverage of the parton densities inside proton and nuclei will greatly exceed that available at the SPS and RHIC (see Fig. 2, taken from Refs. 9 and 21). The aim of this review is to present a comprehensive compilation of the existing predictions for the heavyion programme at the LHC, not to discuss the current b These
numbers are to be compared with those for pp collisions: L0 = 1034 cm−2 s−1 , with hLi/L0 closer to 1 and an estimated running time 8 · 106 s/year. c Jets in heavyion collisions have been measured for the first time by the STAR Collaboration at RHIC.19,20
σ (mb)
Predictions for the HeavyIon Programme at the Large Hadron Collider
106 PbPb hard cross sections
cc
σPbPb = σpp(NLO) × A2
105
bb
σgeom
104
377
J/ψ
103
LHC
102
Jet E >100 GeV
ϒ
10
W
RHIC
Z
1 101
SPS tt
102 103 104 10
100
1000
10000 sNN(GeV)
10
104
pmax = s/2 exp( η) T
pp @ 14 TeV
pA DY
LHC, pA HQ
10 8 6 4 2 0
2
2
/GeV
10
2
tion
8 10
10
~0
SPS DY
1
satura
10
LHC, Ap DY
E772 DY
2 mb
10
6
RHIC pA DY
yQ
2
0
satura
tion,
2
mc
Pb
p
DIS NMC E665
140m 240m 420m
ZDC, LHCf TOTEM RPs ALFA RPs FP420
HF/FCal, T1
4
14 17 m
CASTOR, T2
ALICE / LHCb
ALICE
CASTOR, T2
FP420 ALFA RPs TOTEM RPs ZDC, LHCf
HF/FCal, T1
102 10
~3
10
11 m
101
10
LHC from HQ
3
yQ
10
1
4
ATLAS,CMS
3
T
dp /dη (GeV/c)
hard = A2 σ hard ) in PbPb minimum Fig. 1. Left: Cross sections for various hard processes (σPbPb pp √ bias collisions in the range sNN = 0.01 ÷ 14 TeV. Figure taken from Ref. 17. Right: Expected annual yields in the ALICE EMCal acceptance for various hard processes for minimum bias PbPb collisions at 5.5 TeV. Figure taken from Ref. 18.
1
2
10
5
10
4
10
3
10
2
10
1
10
0
η
Fig. 2. Left: Approximate pT − η coverage of current and proposed spectrometers and calorimeters at the LHC. Figure taken from 21 . Right: Resolution power (Q2 ) × momentum fraction (x) coverage of the SPS, RHIC and LHC experiments for parton densities (grey bands and solid lines), compared with the regions covered by previous leptonnucleus and protonnucleus experiments (colored markers). Figure taken from Ref. 9.
interpretation of available experimental data. Nevertheless, I will briefly indicate the main, ‘standard’ claims extracted from the experimental programmes at the SPS and RHIC. Let me stress that none of these claims are devoid of alternative explanations, and that their presentation will doubtlessly contain some personal
378
N. Armesto
bias. I will use them only to motivate the discussion of the new opportunities at the LHC and the discriminating power of the forthcoming measurements there. The standard claims at RHIC are the following (the reader may find extensive discussions and references to the relevant experimental data in Refs. 1–5, 22–25): • Multiplicities at RHIC are much lower than preRHIC expectations.26,27 The standard interpretation is that particle production in the collisions shows a large degree of coherence due to initial state effects. • The elliptic flow measured in the collisions can be well reproduced by calculations within ideal hydrodynamics with a very early thermalization (or isotropization) time and small room for shear viscosity. This is currently interpreted in terms of the creation of some form of matter which (nearly) equilibrates very early and behaves like a quasiideal fluid. • The yield of high transverse momentum particles of different species measured at RHIC is strongly depleted in comparison with the expectations of an incoherent superposition of nucleonnucleon collisions (as suggested by the collinear factorization theorems and confirmed by experimental data on weakly interacting perturbative probes). This fact, named jet quenching, together with the absence of such depletion in dAu collisions, is understood as the creation of a partonic medium, very opaque to energetic partons traversing it. On the basis of these observations, it has been claimed that partonic matter, with an energy density larger than required by lattice QCD (see e.g. Refs. 28–30) for the phase transition from hadronic matter to the QuarkGluon Plasma (QGP) to occur, has been formed. Such matter is extremely opaque to fast color charges traversing it, and its collective expansion closely resembles that of an ideal fluid. These two latter facts suggest that the produced matter is strongly coupled, which is in opposition to the naive picture of the QGP as an ideal parton gas and is not contradicted by lattice data which show some deviation from the StefanBoltzmann law and a finite value of the conformal anomaly up to temperatures larger than several times the deconfinement temperature. Many questions remain open both in the experiment (e.g. suppression of heavyflavor production or unbiased jet measurements, in nucleusnucleus collisions) and on the theory side (can the observed phenomena be explained within pQCD or do they require strong coupling; what is the correct implementation and actual role of bulk and shear viscosity in hydrodynamical calculations; how can such an early isotropization be achieved; can the initial state — the nuclear wave function — be described by perturbative methods; how can we compute particle production in such a dense environment; etc.). In this review of predictions for the heavyion programme at the LHC.d I will classify them into different groups according to the following scheme: Those d Similar
efforts done for RHIC can be found in Refs. 26 and 27.
Predictions for the HeavyIon Programme at the Large Hadron Collider
379
observables which characterize the produced medium itself, which I will call bulk observables (or soft probes, as they refer to particles with momentum scales of the order of the typical momentum scale of the medium — the temperature if thermalization were achieved); and those whose expectation in the absence of any medium can be calculated by perturbative methods in QCD (pQCD) (thus characterized by a momentum scale much larger than both ΛQCD and the ‘temperature’ of the medium), commonly referred to as hard probes.9,14–16 Not being the subject of this review, I will provide few references to introduce the different subjects — I refer the reader to Refs. 31–33. I will start this review by some qualitative expectations for the LHC, based on simple arguments (in this respect see also Ref. 34). Through such discussion I aim to show how a single observable — charged multiplicity at midrapidity — strongly influences most other predictions. Then I will turn to detailed predictions on bulk observables. I will review those on multiplicities, collective flow, hadrochemistry at low transverse momentum, correlations and fluctuations. Next I will discuss hard and electromagnetic probes: particle production at large transverse momentum and jets, heavy quarks and quarkonia, and photons and dileptons.e Then I will review briefly pA collisions.9 I will conclude with a summary and a discussion about the potentiality of the measurements at the LHC — particularly those made during the first run — to further substantiate or, on the contrary, disproof the picture of the medium that has arisen from the SPS and RHIC. Most of the material that I will review is based on what was presented at the CERN Theory Institute on Heavy Ion Collisions at the LHC — Last Call for Predictions, held at CERN from May 14th to June 8th 2007, coorganized by Nicolas Borghini, Sangyong Jeon, Urs Achim Wiedemann and myself.35,36 I apologize in advance to those whose contributions I may unwillingly skip. I also apologize for not including any prediction for ultraperipheral collisions (UPC) — see recent excellent reviews in Refs. 37 and 38. 2. Qualitative Expectations In principle, the reliability of the predictions for a given observable made within the framework of a given model is as good as the understanding of the existing experimental situation on that observable and related ones — provided the model contains the physical ingredients relevant for the extrapolation. It turns out that predictions for most observables, both for soft and hard probes, demand some parameter fixing which, in the most favorable case, can be related to a single measurable quantity. Such a quantity is usually the charged multiplicity at midrapidity or pseudorapidity which, in a more or less modeldependent way, can be related with energy densities, temperatures, etc. of the medium at some given time. e Concerning
photons, their production at low momentum cannot not be described within pQCD but they have customarily become part of the general item of hard and electromagnetic probes.
380
N. Armesto
In this Section, I will review some qualitative or semiquantitative expectations for central PbPb collisions at the LHC. The aim here is not to provide realistic or definite numbers (actually I will be most conservative in the estimates, so very probably the quantities for the LHC are underestimated in comparison to those at RHIC), but more or less stringent bounds, and to show explicitly how different predictions become affected or determined by a single observable, namely charged multiplicity at midrapidity. Let me note that a collection of datadriven predictions can be found in Ref. 34. While this latter collection, in its aim to being as modelindependent as possible, is complementary to the one to be presented in the next Sections, it overlaps in spirit what will be presented here. In Table 1 I show the results within the Monte Carlo code39 for the number of participants, of collisions and the charged multiplicity at midrapidity and pseu√ dorapidity in central PbPb collisions at sN N = 5.5 TeV. While these quantities are obtained in the framework of a given simulator, they will serve for the purpose of illustration in this Section. They will also be employed to better allow a comparison among different predictions for multiplicities at mid(pseudo)rapidity in Subsection 3.1.
Table 1. Results in the Monte Carlo code in Ref. 39 for the mean impact parameter, number of participants and binary nucleonnucleon collisions, and charged multiplicity at mid(pseudo)rapidity, for different centrality classes defined by the number of √ participants, in central PbPb collisions at sNN = 5.5 TeV.
%
hbi (fm)
hNpart i
hNcoll i
dNch /dyy=0
dNch /dηη=0
0÷3
1.9
390
1584
3149
2633
0÷5
2.4
375
1490
2956
2472
0÷6
2.7
367
1447
2872
2402
0 ÷ 7.5
3.0
357
1390
2759
2306
0 ÷ 8.5
3.1
350
1354
2686
2245
0÷9
3.2
347
1336
2649
2214
0 ÷ 10
3.4
340
1303
2583
2159
For the purpose of fixing one reference centrality class, I will define it by a number of participants Npart = 350. In the following and unless otherwise stated, when referring to RHIC and the LHC I will be making reference to AuAu collisions at top RHIC energy, and PbPb collisions at the LHC, for a central centrality class defined by Npart = 350. Let me start with multiplicities, as they are a key observable which will determine many other predictions. As stated previously, expectations for other observables from collective flow to jet quenching, depend on the scaling of certain quantities e.g. initial energy density, which are related in some more or less direct way with the final multiplicity measured in the event. Thus, many predictions
Predictions for the HeavyIon Programme at the Large Hadron Collider
381
are provided for some specific values of parameters which may be linked with a multiplicity. Predictions for multiplicities can be discussed in the following way: A lower bound comes from the wounded nucleon model40 in which the multiplicity in nuclear collisions is expected to be proportional to the number of participant nucleons. This proportionality is also the limiting value expected by models which consider extremely strong shadowing effects. On the other hand, an upper limit can be set by the proportionality to the number of binary nucleonnucleon collisions Ncoll , as expected both in models of particle production which suppose a dominance of hard, perturbative processes (using the collinear factorization theorem,41,42 inclusive particle production is proportional to the product of the fluxes of partons in projectile and target which in the totally incoherent limit is proportional to the number of nucleonnucleon collisions) and in soft models of particle production in absence of shadowing corrections (see e.g. Ref. 43) through the cutting rules.44 On the basis of these considerations, the multiplicity can then be written in the following way (see also the discussions in Ref. 27): NN AA dNch 1−x dNch N (3) = + xN part coll , 0 < x < 1, dη η=0 dη η=0 2
with the superscript N N referring to nucleonnucleon collisions — an average of pp, pn and nn.f Shadowing effects and energymomentum constraints43 tend to decrease √ x. As an example, values extracted from RHIC data at sN N = 19.6 and 200 GeV45 are x ≃ 0.13. For nucleonnucleon collisions, I will use the proton(anti)proton data shown in Fig. 3. The three lines correspond to the parametrization of Sp¯ pS and Tevatron data by CDF46 NN dNch (CDF) = 2.5 − 0.25 ln sN N + 0.023 ln2 sN N , (4) dη η=0 to the parametrization in Ref. 47 NN dNch (ASW) = 0.47 (sN N )0.144 (Npart )0.089 = 0.50 (sN N )0.144 dη η=0
(5)
and to the PHOBOS parametrization in the contribution by Busza in Ref. 35, NN dNch (PHOBOS) = −0.5 + 0.39 ln sN N (6) dη η=0
(note that this parametrization was obtained from fits to nucleusnucleus data and thus it was not intended to describe nucleonnucleon data). I will also assume, as f At
large energies and at central rapidities, particle production should be determined by partons √ with small momentum fraction (which can be estimated using 2 → 1 kinematics as x ∼ mT / sNN , q with mT = p2T + m2 the transverse mass of the produced particle). At such small momentum fractions, isospin symmetry is expected to hold.
382
N. Armesto
suggested by RHIC data,45 that the energy and centrality dependences of charged particle yields at midrapidity decouple. Considering all this, I show in Table 2 some naive predictions for the LHC.g The predictions from the wounded nucleon model (x = 0) lie in the range 900 ÷ 1100, while those from a scaling with the number of collisions lie in the range 6800 ÷ 8400. The latter agree with the expectations in 1995 as shown in the ALICE Technical Proposal.49 The former roughly coincide with the expectations (1100,34 ) from limiting fragmentation (extended longitudinal scaling) and a selfsimilar trapezoidal shape of the ηdistribution between RHIC and LHC energies. Let us note that, as discussed in Ref. 34, charged multiplicities larger than ∼ 1650 will be difficult to reconcile with limiting fragmentation.
dNNN ch /dη 10
η=0
9
triangle: PYTHIA6.4 (pp)
8
square: PSM (pp)
7
solid: CDF (pp)
6
dashed: ASW (NN)
5
dotted: PHOBOS (NN)
4 3 2 1
102
3
10
4
Ecm (GeV) 10
Fig. 3. Charged multiplicity at midpseudorapidity in nucleonnucleon collisions versus centerofmass energy, from different parametrizations (CDF for p¯ p collisions 46 , ASW 47 and PHOBOS 35 for nucleonnucleon collisions) and Monte Carlo simulators (PSM1.0 39 , and PYTHIA6.4 48 as shown in 12 , for pp collisions; these two points are included just for the purpose of illustration as they depend on the set of parameters used for the simulation).
Now one can try to estimate a lower bound for the energy density in this reference centrality class defined by Npart = 350. For this and for the forthcoming discussions in this Section, I will consider three possibilities for multiplicities: • Case I, the smallest one in Table 2, 900; • Case II, the maximum multiplicity allowed by limiting fragmentation,34 1650; • And Case III, a value of 2600 which is representative of the highest recent predictions for the LHC (see Subsection 3.1). g For the same centrality class defined by N part = 350, the corresponding charged multiplicity at η = 0 at top RHIC energy from the PHOBOS parametrization35 is 635.
Predictions for the HeavyIon Programme at the Large Hadron Collider
383
Table 2. Charged multiplicity at central pseudorapidity in PbPb collisions at LHC energy for Npart = 350 (Ncoll = 1354) from Eq. (3), for three different predictions of the corresponding multiplicity in pp collisions, see Fig. 3.
pp extrapolation
pp dNch /dηη=0
ASW ASW
x
P bP b dNch /dηη=0
5.97
0
1050
5.97
0.13
1950
ASW CDF
5.97 5.02
1 0
8100 900
CDF
5.02
0.13
1650
CDF
5.02
1
6800
PHOBOS
6.22
0
1100
PHOBOS
6.22
0.13
2050
PHOBOS
6.22
1
8400
I use the Bjorken estimate50 and the arguments about the average formation (proper) time for particle production, hτf orm i, in Ref. 2: measured 1 measured dN hǫi(hτf orm i) ≥ hmT i hτf orm iA dη η=0 measured hmmeasured i2 3 dNch T ≈ . (7) 2 πRA 2 dη η=0
2 In this equation hτf orm i ≈ hmmeasured i−1 , A = πRA is an upper bound for the T overlapping area for central collisions with RA = 1.12 A1/3 fm the nuclear radius, mT is the transverse mass and the superindex measured indicate that these are the final quantities measured in the detectors. For top RHIC energy, using hmmeasured i = 0.57 GeV as given by PHENIX2 (this quantity is weakly depenT measured dent on centrality), and taking dNch /dηη=0 = 635 as given by the PHOBOS parametrization in Ref. 35, I get hǫi(hτf orm i = 0.35 fm) ≥ 12 GeV/fm3 . For the LHC, one has to estimate the increase in hmmeasured i with collision energy. For T √ that, I use the parametrization for hpT i ( s) by UA151 and adjust the hadron mass to get the value given by PHENIX i.e. √ √ √ s/GeV [GeV], hpT i s = 0.4 − 0.03 ln s/GeV + 0.0053 ln2 (8)
and m = 0.42 GeV. Then I get, for Case I, hǫi(hτf orm i = 0.29 fm) ≥ 22 GeV/fm3 . Therefore, the most conservative estimates for the LHC indicate a multiplicity increase of a factor 900/635 ≃ 1.4 and an increase of a factor ∼ 2 in energy density at formation time, with respect to top RHIC energy (or hǫi(hτf orm i = 0.29 fm) ≥ 42 and 66 GeV/fm3 for Cases II and III respectively). Now, and for the purpose of illustrating some qualitative behaviors, I turn to the eventual equilibration and dynamical evolution of the created system. For that
384
N. Armesto
I will use generic arguments based on the Bjorken ideal hydrodynamical scenario in one spatial dimension,50 see Refs. 33 and 52 for reviews of the hydrodynamical description of heavyion collisions. First, one needs the energy density at the time when hydrodynamical evolution is initialized, i.e. a thermalization or isotropization time. The estimates at RHIC lie in the range 0.17÷1 fm2–5,33,53 in ideal hydro (0.17 fm is the crossing time of two Au nuclei at RHIC), and similar values for studies RHIC including viscosity.54–56h I will take an intermediate time τtherm ≃ 0.6 fm as a reference value. To extrapolate to the LHC, it looks plausible that a system with larger density thermalizes faster. Using the ideas58,59 in the Color Glass Condensate (CGC, see the review in Ref. 33),i I will assume that the thermalization time scales like the inverse square root of the multiplicity at η = 0. Therefore one expects LHC hτtherm i ≃ 0.85, 0.62, 0.49 RHIC hτtherm i
(9)
for Cases I, II, III respectively. So the thermalization time at the LHC is τtherm . 0.5 fm. Assuming free streaming (ǫ ∝ 1/τ in the onedimensional case) from formation time to thermalization, the corresponding lower bound for the energy density is 3
LHC hǫi(hτtherm i ≃ 0.5 fm) ≥ 12 GeV/fm ,
(10)
again factor ∼ 2 larger than the one obtained for RHIC. If one assumes that the thermalization time decreases with increasing particle density, then larger multiplicities at the LHC will favor smaller thermalization times and thus larger energy densities at thermalization. For example, in the model used for illustration, measured onedimensional free streaming plus CGC, hǫi(hτtherm i ∝ (dNch /dηη=0 )3/2 (modulo logarithmic corrections). Now I will consider the evolution of the system, in order to illustrate the typical scales for the different phases of the system. To do so, I assume an ultrarelativistic ideal gas of 3 light quarks and gluons in the deconfined phase, and of 8 pseudoscalar mesons in the confined phase. Using Bjorken estimate (7) and the StefanBoltzmann law, the values of the relevant quantities at hτtherm i are given in Table 3. Then I consider the evolution of the system using the Bjorken ideal hydrodynamical scenario in one spatial dimension50 (see Ref. 60 for recent developments in 1+1 ideal hydrodynamics) for both the confined and deconfined phases, but with a free parameter α to mimic a larger dilution rate due to transverse expansion: h On
T T0
3
=
τ α 0
τ
,
τ 4α/3 ǫ 0 = , ..., ǫ0 τ
(11)
viscous hydrodynamics, see the recent review in Ref. 57. the CGC, the multiplicity is proportional to the saturation scale squared Q2s and its energy and centrality dependences roughly factorize, while the thermalization time is expected to be inversely proportional to Qs . i In
Predictions for the HeavyIon Programme at the Large Hadron Collider
385
Table 3. Values of the energy density, the temperature and the entropy density at hτtherm i for RHIC and the LHC. AA dNch /dηη=0
hτtherm i (fm)
hǫi (GeV/fm3 )
Ti (GeV)
s0 (fm−3 ) 38
RHIC
635
0.6
6.8
0.241
LHC
900 (I)
0.51
12.6
0.281
60
LHC
1650 (II)
0.37
32.7
0.356
123
LHC
2600 (III)
0.30
64.4
0.422
204
with α = 1 corresponding to a pure longitudinal expansion. For a firstorder phase transition, and assuming a deconfinement temperature of 170 MeV and a freezeout temperature of 140 MeV, the evolution of the temperature is shown in Fig. 4. While the numbers shown in both the Figure and in Table 3 are most rough estimations, the plot illustrates some features common to more involved calculations: at the LHC the deconfined phase will last longer than at RHIC. The hadronic phase is not comparatively shorter than at RHIC (in this very schematic calculation using powerlaw evolutions of thermodynamical quantities), but its impact on some final observables (e.g. on photon or dilepton emission) could be expected to be smaller than at RHIC, due to the fact that the hadronic phase is restricted to the same range of temperatures but the partonic phase reaches higher T at the LHC than at RHIC. On the other hand, it clearly shows that the larger the multiplicities, the longerlived the deconfined phase will be. Now I will focus on the dependence on multiplicity of the elliptic flow v2 integrated over transverse momenta.j According to general arguments, see e.g. Ref. 61, in the lowdensity limit the distortion of the azimuthal spectra with respect to the reaction plane XZ (and thus the elliptic flow v2 ) is proportional to the space anisotropy ǫx =
hy 2 − x2 i hy 2 + x2 i
(12)
and to the density of scattering centers (or particle density) in the transverse plane XY , 1 dNch v2 ∝ , (13) ǫx Sover dy y=0
with Sover the overlap area for a given centrality class and the average in (12) is done over the transverse energy density profile and, eventually, over the number of events. This relation is fulfilled by experimental data from lowest SPS to highest RHIC energies, see Ref. 62, and is illustrated in Fig. 5 left. It allows for a semiquantitative relation between the multiplicity and the elliptic flow: I will assume that for AuAu or j The
discussion of the behavior of v2 (pT ) requires a parallel discussion of the hadronization process which is far more involved.
386
N. Armesto
Fig. 4. Temperature versus proper time in the Bjorken model for the four scenarios in Table 3 and for two values of α in Eq. (11): α = 1 and 1.2.
PbPb collisions at a given centrality class the spatial anisotropy, mainly determined by the geometry of the collision, and the overlap area are approximately the same and do not vary substantially with energy. Taking the slope of the experimental trend ∼ 0.005 and for a point lying at (22,0.16),k increases in multiplicity by factors 1.5, 2.5, 4l translates into increases in v2 /ǫx of ≃ 35, 100, 205% respectively. On the other hand, ideal hydrodynamics calculations64,65 indicate a saturation or limiting value of v2 /ǫ versus (1/Sover )dNch /dyy=0 . The detailed value depends on the equation of state, on the details of initialization (see e.g. Ref. 66 for a study of the influence of different initial conditions on the spatial anisotropy) and hadronization prescription, and on the treatment of the confined phase. The inclusion of viscous effects further reduces such limiting value.54 Besides, as illustrated in Fig. 5 right, for a fixed initial spatial anisotropy, higher initial energy densities or temperatures imply larger density gradients which increase the final momentum
k These
values are roughly those of the experimental data (v2 = 0.051, ǫx = 0.319)63 for a 20÷30% centrality class, which corresponds53 to an impact parameter ∼ 7.5 fm and Npart ∼ 160 both for RHIC and the LHC. l These numbers are illustrative of the predictions for charged multiplicities at midpseudorapidity for Npart = 350 at the LHC, 900, 1650 and 2600 — Cases I, II and III respectively — compared with 635 at RHIC, and are applicable to other centralities provided the factorization between the centrality and energy dependences holds, see previous discussions.
Predictions for the HeavyIon Programme at the Large Hadron Collider
387
Fig. 5. Left: schematic plot showing the experimental trend (black) and the hydrodynamical limit (red line) of v2 /ǫx versus (1/Sover )dNch /dyy=0 . Right: schematic behavior of the spatial ǫx and momentum ǫp anisotropies versus proper time τ for lower (dashed) and higher (solid lines) energy densities at a fixed initial space anisotropy.
anisotropy33,52 defined as ǫp =
hTxx − Tyy i , hTxx + Tyy i
(14)
and thus increase v2 /ǫx , as this momentum anisotropy is known65 to be related with the observed v2 ≃ ǫp /2. Numerical results53,67 within ideal hydrodynamics indicate increases in the transverse momentum integrated v2 at b ∼ 7.5 fm from RHIC to the LHC, ranging from ∼ 15%67 for charged multiplicities at midrapidity around 1200, to ∼ 40÷60% in Ref. 53 for twice this multiplicity for central PbPb collisions. Results in viscous hydro54 yield increases ∼ 10% for a charged multiplicity of 1800. Let us finally discuss very briefly the influence of multiplicities on the standard observable for jet quenching, namely single inclusive particle suppression usually studied through the nuclear modification factor, defined for a given particle k = h± (ch), π 0 , . . . as RAA (y, pT ) =
dNkAA dydpT dNkN N hNcoll i dydp T
,
(15)
with hNcoll i the average number of binary nucleonnucleon collisions in the considered centrality class. A simple modelm to discuss this is the following: Let us assume m This
simplistic model, whose sole aim is allowing for a discussion of the competing effects of density increase and different biases, is by no means quantitative. For example, it does not consider in any detail geometrical biases like the surface bias, it does not take into account fragmentation and it assumes a pure power law behavior of the hadronic spectra which is true neither in data nor in pQCD. It is based on ideas developed in models of radiative energy loss in e.g. Refs. 68 and 69 but not restricted to these models — e.g. models with collisional energy loss also result in some probability of no energy loss.
388
N. Armesto
for a fixed geometry (i.e. fixed length or eventual dynamical expansion) that partons can escape the medium without losing any energy with probability p0 , while they may lose some energy ∆E with probability 1 − p0 . Considering a spectrum ∝ 1/pnT (n = 8 roughly describes the spectrum in pp collisions at midrapidity at RHIC2 ) I get RAA (y, pT ) = p0 +
∆E 1 − p0 , ǫ= . (1 + ǫ)n pT
(16)
In Fig. 6 I show the ratio of nuclear modification factors at the LHC and at RHIC. For RHIC, I have chosen p0 = 0.1 and ǫ = 0.3 which produce a flat RAA (pT ) ≃ 0.21 which qualitatively corresponds with that observed for π 0 ’s in central AuAu collisions at RHIC for 10 GeV < pT < 20 GeV. To extrapolate to the LHC situation, I have chosen two values of n = 6, 5 and either the same values of p0 and ǫ, or these values modified by the expected ratio of multiplicities in Cases I and II (Case III is not illustrated for clarity of the plot), or a modification in which ∆E, and not ǫ, scales with multiplicity at midrapidity. Different options produce evidently different results (e.g. the flatter the spectrum, n = 5 compared to n = 6, the larger the ratio; the larger the multiplicity, the smaller the ratio), a fact which stresses the need of a control of the reference spectrum and of the geometry or
Fig. 6. Ratio of nuclear modification factors at the LHC and at RHIC from Eq. (16). Different line styles refer to different parameters p0 and ǫ or ∆E, see the legends in the plot, while lower and upper lines of each style correspond to spectral powerlaw exponents n = 6 and 5 respectively.
Predictions for the HeavyIon Programme at the Large Hadron Collider
389
dynamical behavior of the medium in order to extract quantitative conclusions about the medium properties from measurements of the nuclear modification factor. 3. Bulk Observables Now I turn to the predictions for observables which directly characterize the medium produced in the collisions. These bulk observables correspond to particles with momentum scales of the order of the typical scales of the medium — the temperature if thermalization is achieved — thus the name of soft probes that has been used to designate them. In the following, the use of names of authors will correspond usually to those predictions contained in the compilation,35,36 while those predictions not contained there will be referenced in the standard way. I refer the reader to the compilation35 for further information and model description of the former — a given contribution in Ref. 35 can be found by looking for the name of the authors in the Section devoted to the corresponding observable. In this Section I will review consecutively: multiplicities, collective flow, hadrochemistry at low transverse momentum, correlations and fluctuations. 3.1. Multiplicities Charged particle multiplicity at mid(pseudo)rapidity is a firstday observable at the LHC. Many groups have produced such predictions, see a compilation in Fig. 7 where 25 predictions are shown.n Different groups provide predictions for different centrality classes. For a more accurate comparison, I rescale them to a common observable (dNch /dηη=0 ) and centrality class (hNpart i = 350) using the model.39 The rescaling factors can be read off Table 1 and the corrected results found in Fig. 8. The rescaling being made using a given model, its accuracy cannot be taken as very high, but it should reduce the uncertainties in the comparison to a 10% level. Let me start by describing briefly the different predictions presented in the plots. A rough classification, for mere organizational purposes, can be made into the following items: (1) Monte Carlo simulators of nuclear collisions. These models include many different physical ingredients to be combined in a consistent manner. They all take into account energymomentum and quantum number conservation in a detailed way. While sometimes the physical ingredients are similar between different models, the details of the implementation lead to different results. • The PSM model39 contains a soft component with contributions from both the number of collisions and of participants which lead to the creation of nA
compilation containing a smaller number of predictions can be found in Ref. 36.
390
N. Armesto
Charged multiplicity for η=0 in central PbPb at 5.5 TeV 25
20
15
10
5
0
Wolschin et al. Sarkisyan et al. Sa et al. Porteboeuf et al. Mitrovski et al. Lokhtin et al. Kharzeev et al. Jeon et al. Humanic Fujii et al. Eskola et al. El et al. Dias de Deus et al. Chen et al. Capella et al. Chaudhuri Bzdak Busza Bopp et al. Topor Pop et al. Armesto et al. Armesto et al. Arleo et al. Albacete Abreu et al.
0
dN/dη, 6% dN/dη, N =350 part dN/dη, 10% dN/dη, N =350 part dN/dη, 5% dN/dη, 5% dN/dη, N =350 part dN/dη, N =350 part dN/dη, 5% dN/dη, N =350 part dN/dη, 5% dN/dη, 3% dN/dη, N =350 part dN/dη, b 8 GeV at the LHC, are shown210 (see also Ref. 211). The disentanglement of a thermal component on the background looks defying and demands a very detailed understanding of the background sources. Now I turn to dilepton production. Dileptons offer interesting information both in the low mass region M < 1 GeV and in the intermediate mass region 1 GeV < M < MJ/ψ , see e.g. Refs. 212, 213 and 214 and references therein. In the former they are expected to reflect the changes of resonances (masses, widths) in the medium. In the latter a window of sizable thermal emission has been speculated. In Fig. 35,35,215 the different contributions to the dilepton spectra at large (top) and small (bottom) transverse momentum is shown. The contribution from heavyquark decays seems to dominate all masses but M < 0.5 GeV, where hadronic contributions are very large. Therefore, the identification of thermal sources looks defying. A larger multiplicity should be linked with a larger temperature but also with a larger hadronic background. It seems that only a larger light multiplicity — if originating from a larger temperature — linked with a smaller heavyquark cross section — leading to a smaller background — would improve the situation for detection of thermal dileptons in the intermediate mass region. All calculations shown until now for photons or dileptons in the lowpT region assume a very small thermalization time, < 1 fm, at which the system is isotropized/thermalized. The physical mechanisms which could make such fast
N. Armesto
424
4
10
2
2
/dydM (GeV )

e e
010% Central Ti= 845 MeV
LHC
5
10
yd =0, ye < 0.5
6
10
7
10
8
10
heavy quarks decay DrellYan (direct) jettherm, HTL jetfrag (Eloss) thermtherm, HTL
dN
2 G1 in d H e m 1 1 D l Y r e a n 3 4 2 5 3 6 9
10
10
10
11
10 0
1
2
3
4
PT > 8 GeV
6
5
7
8
9
10
M (GeV)
Central Pb+Pb √s=5.5 ATeV ye0.2GeV
10
10
Central Pb+Pb √s=5.5 ATeV
Hadron Gas QGP
dNch/dy=1400
10
sum corr. charm
ye 8 GeV, with different sources indicated, for a scenario corresponding to dNch /dyy=0 = T 5625, from Fries et al. Bottom: Id. for ppair > 0.2 GeV, for the low (left) and intermediate (right) T mass regions, for a scenario corresponding to dNch /dyy=0 = 1400, from van Hees et al. Figures taken from Ref. 35.
isotropization feasible, are not understood yet. Therefore some authors have studied the possibility of a later isotropization time and a previous evolution in an anisotropic stage. For example, the authors in Refs. 216 and 217 consider a model in which the system evolves from an early formation time ∼ 0.1 fm in an anisotropic stage (a collisionallybroadened expansion) to an isotropization reached at 2 fm. They find a signal of such anisotropic behavior (depending on the kinematical cuts applied) in the enhancement of dileptons with large transverse momentum at y = 0, and a suppression of the pT integrated yield (larger for forward rapidities), compared to the early isotropization scenario, see Fig. 36. Similar considerations for photons can be found in Ref. 218. To conclude, much information can be obtained from real and virtual photons at the LHC but an accurate understanding of backgrounds is required.y y The
ratio of real to virtual photons has been argued, Alam et al. in Refs. 35 and 219, to develop a plateau at transverse momentum greater than ∼ 2 GeV, quite insensitive to details of the model and reflecting the initial temperature of the system.
Dilepton modification , Φ(τiso )
Predictions for the HeavyIon Programme at the Large Hadron Collider
425
1 09 08 07 06 05
τiso = 0 088 fm/c
04
τiso = 1 fm/c
03
τiso = 2 fm/c
02 0
1
2
3
4
5
y
6
7
8
9
10
Fig. 36. Ratio of dilepton yield in the anisotropic situation over the isotropic one, versus rapidity, for M ≥ 2 GeV and pT ≥ 0.1 GeV. The initial temperature is fixed to 845 MeV for τiso = 0.088 fm in the early isotropization scenario, and the parameters for the other situations are determined in order to keep the same final multiplicity. Figure taken from Ref. 217.
5. pA Collisions While pA collisions will not take place until several successful datataking heavyion runs have occurred, they offer a vast amount of information (see Ref. 9 and references therein) which finally may turn out to be essential for the interpretation of the PbPb data, as it was the case with dAu collisions at RHIC. They should establish the benchmark for the cold nuclear matter effects on top of which the eventual signals of a dense partonic stage are to be searched. I do not intend to give a full overview of all the possibilities of the pA programme, but rather focus on some selected aspects. First, pA collisions offer the possibility of constraining the nuclear parton densities in kinematical regions, see Fig. 2 right, which will not be explored in leptonnucleus collisions unless future colliders220,221 become eventually available. This is a key ingredient for hard probes, and the present situation of the parton densities in the x region of interest for the LHC (10−4 . x . 10−2 ) derived from DGLAP analysis (see e.g. the review222 or the recent work223 and references therein) is far from being satisfactory, see Fig. 37.223 As evident from this figure, the nuclear gluon densities at a low virtuality for x < 0.05 are very badly constrained.z The inclusion of pA data from the LHC in the fits can only improve this situation. On the other hand, isolated photons offer the possibility of a direct access to the gluon distribution (Fig. 38), see Arleo in Refs. 35 and 224, where it is shown that the nuclear modification factor for isolated photons closely follows the nuclear modifications factors for the gluon distribution and structure function F2 . z The
situation is similar for NLO analysis, see Ref. 223. On the other hand, DGLAP evolution reduces the uncertainties at larger scales.
426
N. Armesto Pb
Pb
1.4
1.4
1.2
1.2
1.0
1.0
Pb
( ,
2
2
=1.69 GeV )
Pb
0.8
0.8 This work, EPS09LO EKS98 HKN07 (LO) EPS08 nDS (LO)
0.6 0.4 0.2 0.0 4 10
0.6 0.4 0.2 0.0
3
10
2
4
1
10
10
10
3
10
2
1
10
10
4
10
3
10
2
10
1
10
1
Fig. 37. Comparison of different parametrizations at lowest order, for the ratio of parton densities (valence, sea and gluons from left to right) per nucleon in Pb over that in p, at Q2 = 1.69 GeV2 . The band represents the uncertainty extracted from the error analysis in the EPS09 parametrizations. Figure taken from Ref. 223.
1.2 1
RpPb (x⊥)
0.8 0.6
√s = 8.8 TeV y = 0
0.4 0.2
isolated γ NLO w/ nDSg 0.5 ( RF2 + RG )
0 0.2 3 10
x⊥
10
2
√ Fig. 38. RpA (y = 0) (solid) for isolated photons in pPb collisions at sNN = 8.8 TeV, and (RF2 + √ Rg )/2 for Pb (dashed), versus x⊥ = 2pT / sNN . The lower (dasheddotted) curve corresponds to the difference between these two quantities divided by RpA . Figure taken from Ref. 35.
Another aspect which has raised large interest is the possibility to check the ideas of gluon saturation as proposed in the framework of the CGC, see the review in Ref. 33. Generically, the saturation scale which characterizes the momentum below which the gluon densities are expected to be maximal, is expected to increase with increasing rapidity or energy, reaching values in the range from 1 to several GeV in heavyion collisions at the LHC. Therefore saturation effects should become visible in a region usually considered within the range of applicability of pQCD. Specifically, the CGC predicts105,106 that the Cronin effect (the fact that
December 18, 2009 10:47 WSPC/INSTRUCTION FILE
lhcpredictionsv2
Predictions for the HeavyIon Programme at the Large Hadron Collider
427
Fig. 39. Left: RpA (y = 0) versus pT for pions in dAu collisions at RHIC (upper panel) and in pPb collisions at the LHC, by Kopeliovich et al. Solid and dashed lines correspond to the calculation with and without gluon shadowing respectively. Right: RpA for pions versus pT in dAu collisions at RHIC and in pPb collisions at the LHC, for different η, from Tuchin. Figures taken from Ref. 35.
the nuclear modification factor is larger than 1) observed at midrapidity in dAu collisions at RHIC disappears with increasing rapidity — as observed at RHIC, see Ref. 4 — and increasing energy. While there is no consensus on this suppression at forward rapidities at RHIC being a clear signal of saturation in the CGC, see e.g. Ref. 225 or Bravina et al. in Ref. 35 for an alternative approach to shadowing, pA collisions at the LHC offer the possibility of further tests. In Fig. 39 left, I show the predictions by Kopeliovich et al.35,226 in which the Cronin effect at midrapidity is still present in pPb collisions at the LHC. On the other hand, in Fig. 39 right predictions are shown within the CGC framework by Tuchin.35,227 While these predictions are for light flavors (see also De Boer et al.,35 or Refs. 228, 229 and 230 for predictions for DrellYan and photons), Tuchin also provides predictions for heavyflavor production with similar features, namely a marked suppression of ratios both at mid and forward rapidities in pPb collisions at the LHC. Clearly different scenarios should be discriminated by LHC data. Finally, in the framework of the CGC, the nuclear modification factor in pPb collisions at the LHC offers the possibility of establishing the relevance of different effects. For example, considering a running coupling instead of a fixed coupling in the CGC evolution equations at high energies or small momentum fractions x (the BK equation, see Ref. 33) leads231 to a nuclear modification factor at very large −(1−γ)/3 rapidities which goes from A ln A1/3 (γ ≃ 0.63, fixed coupling) to A−1/3 (running coupling, called total shadowing), with A the mass number of the nucleus. The same total shadowing is achieved when fluctuations (or pomeron loops, see Refs. 232 and 233) are included.234 Both effects are illustrated in Fig. 40.
N. Armesto
428
RpA
RpA 0.5
1 @A lnHA13 LDH1ΓL3
0.4
Y1
0.3
Y2
Η=3
Y3
0.2 13
A
0.1
Y4
1 1
A3
9
10
11
12
13
14
15
p2¦
Qs (Y1 )
Qs (Y4 )
k⊥
Fig. 40. Left: RpA versus p2T for pPb at the LHC at η = 3 from smallx evolution for fixed coupling (dashed) and running coupling (solid black), from Iancu et al. The lower, blue line corresponds to total gluon shadowing. Right: Schematic plot of RpA versus transverse momentum k⊥ showing the increasing effect of fluctuations with increasing rapidities and the approach to total shadowing, by Kozlov et al. Figures taken from Ref. 35.
6. Summary and Discussion In this work I have reviewed the predictions for the heavyion programme at the LHC, as available in early April 2009. After an introduction I have discussed some qualitative expectations with the aim of illustrating how a single observable, namely charged multiplicity at midrapidity, influences predictions for the energy density and other thermodynamical quantities, the evolution of the system, predictions for elliptic flow (v2 ) or the nuclear modification factor (RAA ) in models of energy loss. Then I have turned to a compilation of results (additional information can be found in Refs. 9, 14–16, 27, 34–36). Referring to PbPb collisions at the LHC and, otherwise stated, to observables at midrapidity, a summary of what was presented is: (1) In Subsection 3.1 I have discussed the predictions for charged multiplicity at midpseudorapidity. Most predictions (for Npart ∼ 350, ∼ 10% more central collisions) now lie below 2000 — a value sizably smaller than preRHIC predictions,27 and they include a large degree of coherence in particle production through saturation, strong gluon shadowing, strong color fields, etc. On the other hand, the expectations for net protons at η = 0 are systematically below 4. (2) In Subsection 3.2 I have analyzed the results for elliptic flow in several models. pT integrated v2 increases in all models when going from RHIC to the LHC, but this increase is usually smaller in hydrodynamical models than in naive expectations,34 and Section 2, and in some nonequilibrium, transport models. For v2 (pT ), hydrodynamical models indicate a value for pT . 2 GeV which is very close for pions, while a decrease is expected for protons. A strong decrease would be interpreted — once the initial conditions are settled — as an increase
Predictions for the HeavyIon Programme at the Large Hadron Collider
(3)
(4)
(5)
(6)
(7)
(8)
429
in viscous effects. On the other hand, nonequilibrium models generically result in an increase of v2 (pT ). In Subsection 3.3 predictions for hadrochemistry are reviewed. Different versions of the statistical models result in slightly different predictions, and nonequilibrium scenarios show distinctive features for resonance production. Hydrodynamical and recombination models predict large baryontomeson ratios at moderate pT . Approaches with strong color fields or percolation show Cronin effect for protons in central PbPb collisions at the LHC. In Subsection 3.4 I have reviewed the predictions for correlations. HBT radii are expected to increase from RHIC to the LHC. The predictive power of ideal hydrodynamics is reduced by the limitations that appear in its description of RHIC data. The role of viscosity in the hydrodynamical descriptions of HBT radii is still to be clarified. On the other hand, correlations in rapidity are expected to extend along large intervals and offer additional possibilities of constraining the multiparticle production mechanism. In Subsection 3.5 I have shown the existing predictions for multiplicity fluctuations — few predictions are available as the evidence of a nonstatistical or nontrivial origin of fluctuations at SPS and RHIC is still under debate. Fluctuations also hold discriminative power between different mechanisms of particle production e.g. different statistical ensembles. In Subsection 4.1 I have enumerated the predictions for the nuclear modification factor for highpT charged particles or pions in central collisions. They generically lie, for radiative or collisional energy loss models, in the range 0.15 ÷ 0.25 at pT = 20 GeV and increasing with increasing pT . Then I have commented on the possibilities of discriminating between the energy loss mechanism offered by jets, by hadrochemistry at large pT where several mechanisms like energy loss and parton conversions may be simultaneously at work, and by the study of correlations. In Subsection 4.2 results from different models with radiative or collisional energy loss for the nuclear modification factor of heavy flavors have been shown. They offer the possibility to further test the energy loss mechanism, as the energy diminution of a heavy quark traveling through the produced medium is different from that of a massless parton. On the other hand, predictions for quarkonium production are uncertain due to the lack of knowledge of both cold nuclear matter (nuclear parton densities and nuclear absorption) and hot nuclear matter (pattern of dissociation, recombination mechanism at work, etc.) effects. The identification of different quarkonium states and the large pT reach at the LHC, may help to settle the dissociation pattern and the role of recombination. But predictions for the nuclear modification factor of J/ψ are plagued with uncertainties due to e.g. nuclear shadowing or the c¯ c cross section for recombination. In Subsection 4.3 I have reviewed the available predictions for photon and dilepton production. While the large initial temperature or energy density implies a
430
N. Armesto
large yield of thermal real and virtual photons, the huge backgrounds make the disentanglement of a thermal component in the final spectrum challenging — a very precise knowledge of the pp baseline will be required. Effects beyond the usual equilibrium scenarios like anisotropies in the preequilibrium stage may modify the yields with respect to the early thermalization expectations. (9) In Section 5 I have analyzed the usefulness of the pA programme at the LHC for the purpose of reducing the uncertainties in the nuclear parton distributions which weaken the capabilities of hard probes to characterize the medium produced in the collisions. I have also discussed the possibilities of studies of high gluon density QCD through measurements of the nuclear modification factor in pPb collisions in a large rapidity interval. To put in context the predictions with respect to our current interpretation of existing data, let me draw a set of rough predictions for the LHC with respect to every ‘standard’ claim based on the experimental findings at RHIC (see Section 1): Finding at RHIC
‘Standard’ interpretation
Prediction for the LHC
multiplicities smaller than expectations
highly coherent particle production (expected e.g. in CGC)
P bP b dNch /dηη=0 < 2000
v2 in agreement with ideal hydro
quasiideal fluid (strongly coupled QGP)
v2 (pT ) for pT < 2 GeV similar or smaller than at RHIC
strong jet quenching
very opaque medium
light RAA ∼ 0.2 at pT ∼ 20 GeV and increasing with pT
Obviously, neither the standard interpretations nor the predictions presented in this Table are free from problems and uncertainties, even more when the predictions tend to disagree with naive, datadriven expectations which would suggest multiplicities of order 1000, and sizably larger v2 (pT < 2 GeV) and smaller RAA than at RHIC. Finally, I find it tempting to speculate on possible scenarios based on the firstday measurement of charged particle production at midpseudorapidity in central PbPb collisions. Without any intention beyond showing how our understanding may become affected by the very first data and having in mind the present experimental situation and its ‘standard’ interpretation, three rough possibilities can be discussed: P bP b • A low multiplicity scenario, dNch /dηη=0 < 1000, which would be close to the wounded nucleon model expectations and even smaller than most datadriven expectations. It would imply a extremely coherent particle production, difficult to describe even in saturation models. The conditions for collective flow would be relatively close to those at RHIC, and differentiating between naive extrapolations and hydrodynamical behaviors for v2 more involved, as their
Predictions for the HeavyIon Programme at the Large Hadron Collider
431
predictions would not be so different. On high transverse momentum particle production, the fact that the densities are close to RHIC ones, would imply that the difference e.g. in RAA from RHIC would be driven by the different transverse momentum spectra — the trigger bias, so the expectation would be an RAA larger than at RHIC for the same large transverse momentum (e.g. of the order or greater than 20 GeV). The low multiplicity implies a small background for jet and correlation studies. A small light multiplicity could also be a good scenario for recombination models for quarkonia (for a fixed heavy quark cross section). P bP b • An intermediate multiplicity scenario, 1000 < dNch /dηη=0 < 2000 as predicted by most models with a large degree of coherence and by datadriven extrapolations. The differences between naive predictions and results of hydrodynamical models for v2 would be more noticeable. RAA should be more similar than at RHIC, for the same large transverse momentum, than in the previous scenario. P bP b • A large multiplicity scenario, 2000 < dNch /dηη=0 . This scenario would defy naive extrapolations based on logarithmic increases and limiting fragmentation, and would be very problematic for saturation physics. Discriminating between naive predictions and results of hydrodynamical models for v2 should be relatively easy. In this case, a strong decrease of v2 at fixed small pT with respect to RHIC, would strongly suggest viscous effects. RAA at large pT , of the order or greater than 20 GeV, could be smaller than at RHIC for the same transverse momentum. Jet and correlation studies might be more defying due to the larger background. On the other hand, this scenario would imply larger temperatures and energy densities which may be welcome, even in spite of the larger background, for electromagnetic probes. To conclude, the heavyion programme at the LHC will offer most valuable information for improving our understanding of highdensity QCD matter — and, in a wider context, on the behavior of the strong interaction at high energies — from the very first day of data taking. But it should be kept in mind that the usefulness of some observables will be restricted by our lack of knowledge of the pp and pA benchmarks, in particular to constrain the parton densities in nuclei. It seems plausible that a pA run will be needed — as it was the case at RHIC — in order to understand the effects of cold nuclear matter at LHC energies before strong conclusions about the heavyion programme can be drawn. A large amount of work has already been done to extrapolate existing models to the LHC situation. Still much work is needed in order to deal with some observables e.g. viscous hydrodynamical calculations or transport models for collective flow, or Monte Carlo tools for jet analysis, just to mention two obvious ongoing developments. The first LHC data will reduce much of the available freedom in model parameters. The more restricted model predictions done after those very first data will indicate, when confronted with subsequent data on other observables, whether
432
N. Armesto
the physics at the LHC is qualitatively similar to that at the SPS and RHIC or, on the contrary, new aspects appear which will require new ideas. Acknowledgements I thank J. Albacete, J. AlvarezMuniz, F. Bopp, W. Busza, R. Concei¸cao, L. Cunqueiro, A. Dainese, J. Dias de Deus, A. El, D. d’Enterria, K. Eskola, V. Gon¸calves, U. Heinz, C.M. Ko, I. Lokhtin, M. Martinez, J. G. Milhano, C. Pajares, V. Pantuev, T. Renk, E. Sarkisyan, V. Topor Pop, K. Tywoniuk, R. Venugopalan, I. Vitev, X. N. Wang, K. Werner and G. Wolschin for information on their predictions, P. Brogueira, J. Dias de Deus and J. G. Milhano for an updated version of a figure in their paper,113 and U. Heinz, A. Mischke, E. Saridakis, H. Song and D. Srivastava for useful comments. Special thanks are due to David d’Enterria, Guilherme Milhano, Carlos Pajares, Carlos Salgado and Konrad Tywoniuk for a critical reading of the manuscript. This work has been supported by Ministerio de Ciencia e Innovaci´ on of Spain under projects FPA200501963, FPA200801177 and a contract Ram´ on y Cajal, by Xunta de Galicia (Conseller´ıa de Educaci´on) and through grant PGIDIT07PXIB206126PR, and by the Spanish ConsoliderIngenio 2010 Programme CPAN (CSD200700042). References 1. 2. 3. 4. 5. 6. 7.
8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22.
U. W. Heinz and M. Jacob, arXiv:nuclth/0002042. K. Adcox et al. [PHENIX Collaboration], Nucl. Phys. A 757 (2005) 184. B. B. Back et al. [PHOBOS Collaboration], Nucl. Phys. A 757 (2005) 28. I. Arsene et al. [BRAHMS Collaboration], Nucl. Phys. A 757 (2005) 1. J. Adams et al. [STAR Collaboration], Nucl. Phys. A 757 (2005) 102. Midterm plan for RHIC: http://www.bnl.gov/HENP/docs/RHICplanning/RHIC Midtermplan print.pdf. J. Bartke et al. [NA49future Collaboration], A New Experimental Programme with Nuclei and Proton Beams at the CERN SPS, preprint CERNSPSC2003038 (SPSCEOI01). J. M. Jowett, J. Phys. G 35 (2008) 104028. A. Accardi et al., arXiv:hepph/0308248. F. Carminati et al. [ALICE Collaboration], J. Phys. G 30 (2004) 1517. B. Alessandro et al. [ALICE Collaboration], J. Phys. G 32 (2006) 1295. D. G. d’Enterria et al. [CMS Collaboration], J. Phys. G 34 (2007) 2307. P. Steinberg [ATLAS Collaboration], J. Phys. G 34 (2007) S527. A. Accardi et al., arXiv:hepph/0310274. M. Bedjidian et al., arXiv:hepph/0311048. F. Arleo et al., arXiv:hepph/0311131. D. d’Enterria [CMS Collaboration], J. Phys. G 35 (2008) 104039. P. Cortese et al. [ALICE Collaboration], “ALICE electromagnetic calorimeter technical design report”, preprint CERNLHCC2008014/CERNALICETDR014. S. Salur [for the STAR Collaboration], arXiv:0809.1609 [nuclex]. J. Putschke [for the STAR Collaboration], arXiv:0809.1419 [nuclex]. D. d’Enterria, AIP Conf. Proc. 1038 (2008) 95. M. Gyulassy and L. McLerran, Nucl. Phys. A 750 (2005) 30.
Predictions for the HeavyIon Programme at the Large Hadron Collider
23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33.
34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65.
433
P. Jacobs and X. N. Wang, Prog. Part. Nucl. Phys. 54 (2005) 443. B. Muller, Acta Phys. Polon. B 38 (2007) 3705. D. G. d’Enterria, J. Phys. G 34 (2007) S53. S. A. Bass et al., Nucl. Phys. A 661 (1999) 205. N. Armesto and C. Pajares, Int. J. Mod. Phys. A 15 (2000) 2019. F. Karsch, Lect. Notes Phys. 583 (2002) 209. M. Cheng et al., Phys. Rev. D 77 (2008) 014511. Z. Fodor, PoS C POD07 (2007) 027. R. Hwa (ed.), QuarkGluon Plasma, Vol. 1 (World Scientific, Singapore, 1990). R. Hwa (ed.), QuarkGluon Plasma, Vol. 2 (World Scientific, Singapore, 1995). R. Hwa and X. N. Wang (eds.), QuarkGluon Plasma, Vol. 3 (World Scientific, Singapore, 2004); P. F. Kolb and U. W. Heinz, arXiv:nuclth/0305084; A. Kovner and U. A. Wiedemann, arXiv:hepph/0304151; M. Gyulassy, I. Vitev, X. N. Wang and B. W. Zhang, arXiv:nuclth/0302077; E. Iancu and R. Venugopalan, arXiv:hepph/0303204. N. Borghini and U. A. Wiedemann, J. Phys. G 35 (2008) 023001. N. Armesto et al., J. Phys. G 35 (2008) 054001. N. Armesto, J. Phys. G 35 (2008) 104042. K. Hencken et al., Phys. Rept. 458 (2008) 1. D. d’Enterria, M. Klasen and K. Piotrzkowski, Nucl. Phys. Proc. Suppl. B 179 (2008) 1. N. S. Amelin, N. Armesto, C. Pajares and D. Sousa, Eur. Phys. J. C 22 (2001) 149. A. Bialas, M. Bleszynski and W. Czyz, Nucl. Phys. B 111 (1976) 461. J. C. Collins, D. E. Soper and G. Sterman, Nucl. Phys. B 263 (1986) 37. J. C. Collins, D. E. Soper and G. Sterman, Nucl. Phys. B 261 (1985) 104. A. Capella, A. Kaidalov and J. Tran Thanh Van, Heavy Ion Phys. 9 (1999) 169. V. A. Abramovsky, V. N. Gribov and O. V. Kancheli, Yad. Fiz. 18 (1973) 595 [Sov. J. Nucl. Phys. 18 (1974)] 308. B. B. Back et al. [PHOBOS Collaboration], Phys. Rev. C 70 (2004) 021902. F. Abe et al. [CDF Collaboration], Phys. Rev. D 41 (1990) 2330. N. Armesto, C. A. Salgado and U. A. Wiedemann, Phys. Rev. Lett. 94 (2005) 022002. T. Sjostrand, S. Mrenna and P. Skands, JHEP 0605 (2006) 026. N. Ahmad et al. [ALICE Collaboration], preprint CERN/LHCC 1995071. J. D. Bjorken, Phys. Rev. D 27 (1983) 140. C. Albajar et al. [UA1 Collaboration], Nucl. Phys. B 335 (1990) 261. T. Hirano, N. van der Kolk and A. Bilandzic, arXiv:0808.2684 [nuclth]. H. Niemi, K. J. Eskola and P. V. Ruuskanen, arXiv:0806.1116 [hepph]. M. Luzum and P. Romatschke, arXiv:0901.4588 [nuclth]. A. K. Chaudhuri, Phys. Lett. B 672 (2009) 126. H. Song and U. W. Heinz, arXiv:0812.4274 [nuclth]. P. Romatschke, arXiv:0902.3663 [hepph]. D. Kharzeev and K. Tuchin, Nucl. Phys. A 753 (2005) 316. D. Kharzeev and M. Nardi, Phys. Lett. B 507 (2001) 121. G. Beuf, R. Peschanski and E. N. Saridakis, Phys. Rev. C 78 (2008) 064909. S. A. Voloshin and A. M. Poskanzer, Phys. Lett. B 474 (2000) 27. C. Alt et al. [NA49 Collaboration], Phys. Rev. C 68 (2003) 034903. C. Adler et al. [STAR Collaboration], Phys. Rev. C 66 (2002) 034904. J. Y. Ollitrault, Phys. Rev. D 46 (1992) 229. P. F. Kolb, J. Sollfrank and U. W. Heinz, Phys. Lett. B 459 (1999) 667.
434
66. 67. 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85. 86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96. 97. 98. 99. 100. 101. 102. 103. 104.
N. Armesto
T. Lappi and R. Venugopalan, Phys. Rev. C 74 (2006) 054905. G. Kestin and U. W. Heinz, arXiv:0806.4539 [nuclth]. R. Baier, Y. L. Dokshitzer, A. H. Mueller and D. Schiff, JHEP 0109 (2001) 033. C. A. Salgado and U. A. Wiedemann, Phys. Rev. D 68 (2003) 014008. V. Topor Pop, M. Gyulassy, J. Barrette, C. Gale, S. Jeon and R. Bellwied, Phys. Rev. C 75 (2007) 014904. F. W. Bopp, J. Ranft, R. Engel and S. Roesler, arXiv:hepph/0403084. Z. W. Lin, C. M. Ko, B. A. Li, B. Zhang and S. Pal, Phys. Rev. C 72 (2005) 064901. X. N. Wang and M. Gyulassy, Phys. Rev. D 44 (1991) 3501. I. P. Lokhtin, L. V. Malinina, S. V. Petrushanko, A. M. Snigirev, I. Arsene and K. Tywoniuk, arXiv:0903.0525 [hepph]. M. Mitrovski, T. Schuster, G. Graf, H. Petersen and M. Bleicher, arXiv:0812.2041 [hepph]. H. J. Drescher, M. Hladik, S. Ostapchenko, T. Pierog and K. Werner, Phys. Rept. 350 (2001) 93. B. H. Sa, D. M. Zhou, B. G. Dong, Y. L. Yan, H. L. Ma and X. M. Li, J. Phys. G 36 (2009) 025007. J. Dias de Deus and J. G. Milhano, Nucl. Phys. A 795 (2007) 98. J. L. Albacete, Phys. Rev. Lett. 99 (2007) 262301. D. Kharzeev, E. Levin and M. Nardi, Nucl. Phys. A 747 (2005) 609. A. Bzdak, Acta Phys. Polon. B 39 (2008) 1977. A. El, Z. Xu and C. Greiner, Nucl. Phys. A 806 (2008) 287. T. J. Humanic, arXiv:0810.0621 [nuclth]. E. K. G. Sarkisyan and A. S. Sakharov, AIP Conf. Proc. 828 (2006) 35. C. Y. Wong, arXiv:0809.0517 [nuclth]. R. Kuiper and G. Wolschin, Annalen Phys. 16 (2007) 67. I. G. Bearden et al. [BRAHMS Collaboration], Phys. Rev. Lett. 93 (2004) 102301. Y. MehtarTani and G. Wolschin, arXiv:0811.1721 [hepph]. I. C. Arsene [BRAHMS Collaboration], arXiv:0901.0872 [nuclex]. J. AlvarezMuniz, R. Conceicao, J. D. de Deus, M. C. E. Santo, J. G. Milhano and M. Pimenta, arXiv:0903.0957 [hepph]. S. S. Gubser, S. S. Pufu and A. Yarom, arXiv:0902.4062 [hepth]. G. Torrieri, I. Mishustin and B. Tomasik, arXiv:0901.0226 [nuclth]. G. S. Denicol, T. Kodama, T. Koide and Ph. Mota, arXiv:0903.3595 [hepph]. A. Monnai and T. Hirano, arXiv:0903.4436 [nuclth]. H. J. Drescher, A. Dumitru, C. Gombeaud and J. Y. Ollitrault, Phys. Rev. C 76 (2007) 024905. B. B. Back et al. [PHOBOS Collaboration], Phys. Rev. C 72 (2005) 051901. M. Chojnacki, W. Florkowski, W. Broniowski and A. Kisiel, Phys. Rev. C 78 (2008) 014905. H. Petersen and M. Bleicher, arXiv:0901.3821 [nuclth]. D. Molnar and M. Gyulassy, Nucl. Phys. A 697 (2002) 495 [Erratumibid. A 703 (2002) 893]. G. Policastro, D. T. Son and A. O. Starinets, Phys. Rev. Lett. 87 (2001) 081601. P. Arnold, G. D. Moore and L. G. Yaffe, JHEP 0011 (2000) 001. I. Kraus, J. Cleymans, H. Oeschler, K. Redlich and S. Wheaton, arXiv:0902.0873 [hepph]. I. Kraus, J. Cleymans, H. Oeschler, K. Redlich and S. Wheaton, Phys. Rev. C 76 (2007) 064903. J. Rafelski and J. Letessier, Eur. Phys. J. C 45 (2006) 61.
Predictions for the HeavyIon Programme at the Large Hadron Collider
435
105. J. L. Albacete, N. Armesto, A. Kovner, C. A. Salgado and U. A. Wiedemann, Phys. Rev. Lett. 92 (2004) 082001. 106. J. JalilianMarian and Y. V. Kovchegov, Prog. Part. Nucl. Phys. 56 (2006) 104. 107. M. A. Lisa and S. Pratt, arXiv:0811.1352 [nuclex]. 108. A. Kisiel, W. Broniowski, M. Chojnacki and W. Florkowski, Phys. Rev. C 79 (2009) 014902. 109. A. Muronga and D. H. Rischke, arXiv:nuclth/0407114. 110. S. Pratt, arXiv:0811.3363 [nuclth]. 111. C. Gombeaud, T. Lappi and J. Y. Ollitrault, arXiv:0901.4908 [nuclth]. 112. A. Capella and A. Krzywicki, Phys. Rev. D 18 (1978) 4120. 113. P. Brogueira, J. Dias de Deus and J. G. Milhano, Phys. Rev. C 76 (2007) 064901. 114. B. K. Srivastava [STAR Collaboration], Int. J. Mod. Phys. E 16 (2008) 3371. 115. P. Brogueira, J. D. de Deus and C. Pajares, arXiv:0901.0997 [hepph]. 116. N. Armesto, L. McLerran and C. Pajares, Nucl. Phys. A 781 (2007) 201. 117. M. van Leeuwen [STAR collaboration], arXiv:0808.4096 [nuclex]. 118. A. Dumitru, F. Gelis, L. McLerran and R. Venugopalan, Nucl. Phys. A 810 (2008) 91. 119. R. Vogt, Eur. Phys. J. ST 155 (2008) 213. 120. R. Baier, D. Schiff and B. G. Zakharov, Ann. Rev. Nucl. Part. Sci. 50 (2000) 37. 121. D. d’Enterria, arXiv:0902.2011 [nuclex]. 122. J. CasalderreySolana and C. A. Salgado, Acta Phys. Polon. B 38 (2007) 3731. 123. A. Majumder, J. Phys. G 34 (2007) S377. 124. S. S. Gubser, S. S. Pufu, F. D. Rocha and A. Yarom, arXiv:0902.4041 [hepth]. 125. R. Baier, Nucl. Phys. A 715 (2003) 209. 126. H. Liu, K. Rajagopal and U. A. Wiedemann, Phys. Rev. Lett. 97 (2006) 182301. 127. J. CasalderreySolana and X. N. Wang, Phys. Rev. C 77 (2008) 024902. 128. D. Antonov and H. J. Pirner, Eur. Phys. J. C 55 (2008) 439. 129. F. Arleo, JHEP 0707 (2007) 032. 130. A. Dainese, C. Loizides and G. Paic, Eur. Phys. J. C 38 (2005) 461. 131. T. Renk and K. Eskola, Phys. Rev. C 75 (2007) 054910. 132. S. Jeon, J. JalilianMarian and I. Sarcevic, Phys. Lett. B 562 (2003) 45. 133. I. Vitev, Phys. Lett. B 639 (2006) 38. 134. H. Zhang, J. F. Owens, E. Wang and X. N. Wang, Phys. Rev. Lett. 98 (2007) 212301. 135. G. Y. Qin, J. Ruppert, S. Turbide, C. Gale, C. Nonaka and S. A. Bass, Phys. Rev. C 76 (2007) 064907. 136. S. Wicks, W. Horowitz, M. Djordjevic and M. Gyulassy, Nucl. Phys. A 783 (2007) 493. 137. I. P. Lokhtin and A. M. Snigirev, Eur. Phys. J. C 45 (2006) 211. 138. B. G. Zakharov, arXiv:0811.0445 [hepph]. 139. W. Liu, C. M. Ko and B. W. Zhang, Phys. Rev. C 75 (2007) 051901. 140. A. Capella and E. G. Ferreiro, Phys. Rev. C 75 (2007) 024905. 141. L. Cunqueiro, J. Dias de Deus, E. G. Ferreiro and C. Pajares, Eur. Phys. J. C 53 (2008) 585. 142. B. Z. Kopeliovich, H. J. Pirner, I. K. Potashnikova and I. Schmidt, Phys. Lett. B 662 (2008) 117. 143. V. S. Pantuev, Phys. Atom. Nucl. 71 (2008) 1625. 144. M. Cacciari and G. P. Salam, Phys. Lett. B 659 (2008) 119. 145. N. Borghini and U. A. Wiedemann, arXiv:hepph/0506218. 146. N. Borghini, arXiv:0902.2951 [hepph]. 147. I. M. Dremin and O. S. Shadrin, J. Phys. G 32 (2006) 963.
436
N. Armesto
148. 149. 150. 151. 152.
N. Armesto, C. Pajares and P. QuirogaArias, arXiv:0809.4428 [hepph]. R. P. Ramos, arXiv:0811.2418 [hepph]. N. Armesto, L. Cunqueiro, C. A. Salgado and W. C. Xiang, JHEP 0802 (2008) 048. A. Majumder, arXiv:0901.4516 [nuclth]. S. Domdey, G. Ingelman, H. J. Pirner, J. Rathsman, J. Stachel and K. Zapp, Nucl. Phys. A 808 (2008) 178. C. A. Salgado and U. A. Wiedemann, Phys. Rev. Lett. 93 (2004) 042301. I. Vitev, S. Wicks and B. W. Zhang, JHEP 0811 (2008) 093. N. Armesto, L. Cunqueiro and C. A. Salgado, arXiv:0809.4433 [hepph]. K. Zapp, G. Ingelman, J. Rathsman, J. Stachel and U. A. Wiedemann, arXiv:0804.3568 [hepph]. T. Renk, Phys. Rev. C 78 (2008) 034908. S. Sapeta and U. A. Wiedemann, Eur. Phys. J. C 55 (2008) 293. S. Albino et al., arXiv:0804.2021 [hepph]. R. C. Hwa and C. B. Yang, Phys. Rev. Lett. 97 (2006) 042301. T. Renk and K. J. Eskola, Phys. Rev. C 77 (2008) 044905. B. I. Abelev et al. [STAR Collaboration], arXiv:0805.0622 [nuclex]. R. Rapp and H. van Hees, arXiv:0903.1096 [hepph]. Y. L. Dokshitzer and D. E. Kharzeev, Phys. Lett. B 519 (2001) 199. N. Armesto, A. Dainese, C. A. Salgado and U. A. Wiedemann, Phys. Rev. D 71 (2005) 054027. M. G. Mustafa, Phys. Rev. C 72 (2005) 014905. M. Djordjevic and U. W. Heinz, Phys. Rev. Lett. 101 (2008) 022302. A. Adare et al. [PHENIX Collaboration], Phys. Rev. Lett. 98 (2007) 172301. B. I. Abelev et al. [STAR Collaboration], Phys. Rev. Lett. 98 (2007) 192301. A. Mischke [STAR Collaboration], J. Phys. G 35 (2008) 104117. A. Mischke, Phys. Lett. B 671 (2009) 361. A. Adare et al. [PHENIX Collaboration], arXiv:0903.4851 [hepex]. N. Armesto, M. Cacciari, A. Dainese, C. A. Salgado and U. A. Wiedemann, Phys. Lett. B 637 (2006) 362. A. Adil and I. Vitev, Phys. Lett. B 649 (2007) 139. R. Sharma, I. Vitev and B. W. Zhang, arXiv:0904.0032 [hepph]. R. Rapp, V. Greco and H. van Hees, Nucl. Phys. A 774 (2006) 685. P. B. Gossiaux, R. Bierkandt and J. Aichelin, arXiv:0901.0946 [hepph]. Z. Conesa del Valle, A. Dainese, H. T. Ding, G. Martinez Garcia and D. C. Zhou, Phys. Lett. B 663 (2008) 202. W. A. Horowitz and M. Gyulassy, Phys. Lett. B 666 (2008) 320. W. Liu and R. J. Fries, Phys. Rev. C 78 (2008) 037902. X. Zhu, N. Xu and P. Zhuang, Phys. Rev. Lett. 100 (2008) 152301. G. Tsiledakis and K. Schweda, arXiv:0901.4296 [nuclex]. V. T. Pop, J. Barrette and M. Gyulassy, arXiv:0902.4028 [hepph]. T. Matsui and H. Satz, Phys. Lett. B 178 (1986) 416. F. Arleo and V. N. Tram, Eur. Phys. J. C 55 (2008) 449. C. Lourenco, R. Vogt and H. K. Woehri, JHEP 0902 (2009) 014. M. A. Braun, C. Pajares, C. A. Salgado, N. Armesto and A. Capella, Nucl. Phys. B 509 (1998) 357. B. Kopeliovich, A. Tarasov and J. Hufner, Nucl. Phys. A 696 (2001) 669. A. Capella and E. G. Ferreiro, Phys. Rev. C 76 (2007) 064906. O. Kaczmarek, F. Karsch, F. Zantow and P. Petreczky, Phys. Rev. D 70 (2004) 074505 [Erratumibid. D 72 (2005) 059903].
153. 154. 155. 156. 157. 158. 159. 160. 161. 162. 163. 164. 165. 166. 167. 168. 169. 170. 171. 172. 173. 174. 175. 176. 177. 178. 179. 180. 181. 182. 183. 184. 185. 186. 187. 188. 189. 190.
Predictions for the HeavyIon Programme at the Large Hadron Collider
191. 192. 193. 194. 195. 196. 197. 198. 199. 200. 201. 202. 203. 204. 205. 206. 207. 208. 209. 210. 211. 212. 213. 214. 215. 216. 217. 218. 219. 220. 221.
222. 223. 224. 225. 226. 227. 228. 229. 230. 231. 232. 233. 234.
437
A. Mocsy, arXiv:0811.0337 [hepph]. H. Liu, K. Rajagopal and U. A. Wiedemann, Phys. Rev. Lett. 98 (2007) 182301. A. Adare et al. [PHENIX Collaboration], Phys. Rev. Lett. 98 (2007) 232301. A. Andronic, P. BraunMunzinger, K. Redlich and J. Stachel, Nucl. Phys. A 789 (2007) 334. R. L. Thews and M. L. Mangano, Phys. Rev. C 73 (2006) 014904. A. Capella, L. Bravina, E. G. Ferreiro, A. B. Kaidalov, K. Tywoniuk and E. Zabrodin, Eur. Phys. J. C 58 (2008) 437. I. Kuznetsova and J. Rafelski, Eur. Phys. J. C 51 (2007) 113. R. Chatterjee, L. Bhattacharya and D. K. Srivastava, arXiv:0901.3610 [nuclth]. I. Tserruya, arXiv:0903.0415 [nuclex]. S. S. Adler et al. [PHENIX Collaboration], Phys. Rev. Lett. 94 (2005) 232301. A. Adare et al. [PHENIX Collaboration], arXiv:0804.4168 [nuclex]. D. G. d’Enterria and D. Peressounko, Eur. Phys. J. C 46 (2006) 451. I. Vitev and B. W. Zhang, Phys. Lett. B 669 (2008) 337. F. Arleo, JHEP 0609 (2006) 015. R. J. Fries, B. Muller and D. K. Srivastava, Phys. Rev. Lett. 90 (2003) 132301. W. Liu and R. J. Fries, Phys. Rev. C 77 (2008) 054902. R. Chatterjee, D. K. Srivastava and S. Jeon, arXiv:0902.1036 [nuclth]. R. Chatterjee, D. K. Srivastava and U. Heinz, arXiv:0901.3270 [nuclth]. R. Chatterjee and D. K. Srivastava, Phys. Rev. C 79 (2009) 021901. L. Bhattacharya and P. Roy, arXiv:0806.4033 [hepph]. S. Turbide, C. Gale, S. Jeon and G. D. Moore, Phys. Rev. C 72 (2005) 014906. H. J. Specht, Nucl. Phys. A 805 (2008) 338. J. Ruppert, C. Gale, T. Renk, P. Lichard and J. I. Kapusta, Phys. Rev. Lett. 100 (2008) 162301. H. van Hees and R. Rapp, Nucl. Phys. A 806 (2008) 339. S. Turbide, C. Gale, D. K. Srivastava and R. J. Fries, Phys. Rev. C 74 (2006) 014903. M. Martinez and M. Strickland, Phys. Rev. C 78 (2008) 034917. M. Martinez and M. Strickland, arXiv:0808.3969 [hepph]. L. Bhattacharya and P. Roy, arXiv:0809.4596 [hepph]. J. K. Nayak, J. Alam, S. Sarkar and B. Sinha, Phys. Rev. C 78 (2008) 034903. The Electron Ion Collider Working Group Collaboration, C. Aidala et al., A High Luminosity, High Energy Electron Ion Collider, http://web.mit.edu/eicc/. M. Klein et. al., Prospects for a Large Hadron Electron Collider (LHeC) at the LHC, EPAC’08, 11th European Particle Accelerator Conference, 23–27 June 2008, Genoa, Italy. N. Armesto, J. Phys. G 32 (2006) R367. K. J. Eskola, H. Paukkunen and C. A. Salgado, arXiv:0902.4154 [hepph]. F. Arleo and T. Gousset, Phys. Lett. B 660 (2008) 181. V. Guzey, M. Strikman and W. Vogelsang, Phys. Lett. B 603 (2004) 173. B. Z. Kopeliovich, J. Nemchik, A. Schafer and A. V. Tarasov, Phys. Rev. Lett. 88 (2002) 232303. D. Kharzeev, Y. V. Kovchegov and K. Tuchin, Phys. Lett. B 599 (2004) 23. J. JalilianMarian, Nucl. Phys. A 739 (2004) 319. M. A. Betemps and M. B. Gay Ducati, Phys. Rev. D 70 (2004) 116005. M. A. Betemps and V. P. Goncalves, JHEP 0809 (2008) 019. E. Iancu, K. Itakura and D. N. Triantafyllopoulos, Nucl. Phys. A 742 (2004) 182. A. Kovner, Acta Phys. Polon. B 36 (2005) 3551. D. N. Triantafyllopoulos, Acta Phys. Polon. B 36 (2005) 3593. M. Kozlov, A. I. Shoshi and B. W. Xiao, Nucl. Phys. A 792 (2007) 170.