Trace Chemical Sensing of Explosives

TRACE CHEMICAL SENSING OF EXPLOSIVES Edited by RONALD L. WOODFIN WILEY-INTERSCIENCE A JOHN WILEY & SONS, INC., PUBLICAT...

Author: Ronald L. Woodfin

281 downloads 2319 Views 4MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

TRACE CHEMICAL SENSING OF EXPLOSIVES Edited by RONALD L. WOODFIN

WILEY-INTERSCIENCE A JOHN WILEY & SONS, INC., PUBLICATION

TRACE CHEMICAL SENSING OF EXPLOSIVES

TRACE CHEMICAL SENSING OF EXPLOSIVES Edited by RONALD L. WOODFIN

WILEY-INTERSCIENCE A JOHN WILEY & SONS, INC., PUBLICATION

c 2007 by John Wiley & Sons, Inc. All rights reserved. Copyright Published by John Wiley & Sons, Inc., Hoboken, New Jersey. Published simultaneously in Canada. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning, or otherwise, except as permitted under Section 107 or 108 of the 1976 United States Copyright Act, without either the prior written permission of the Publisher, or authorization through payment of the appropriate per-copy fee to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, (978) 750-8400, fax (978) 750-4470, or on the web at www.copyright.com. Requests to the Publisher for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, (201) 748-6011, fax (201) 748-6008, or online at http://www.wiley.com/go/permission. Limit of Liability/Disclaimer of Warranty: While the publisher and author have used their best efforts in preparing this book, they make no representations or warranties with respect to the accuracy or completeness of the contents of this book and specifically disclaim any implied warranties of merchantability or fitness for a particular purpose. No warranty may be created or extended by sales representatives or written sales materials. The advice and strategies contained herein may not be suitable for your situation. You should consult with a professional where appropriate. Neither the publisher nor author shall be liable for any loss of profit or any other commercial damages, including but not limited to special, incidental, consequential, or other damages. For general information on our other products and services or for technical support, please contact our Customer Care Department within the United States at (800) 762-2974, outside the United States at (317) 572-3993 or fax (317) 572-4002. Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic formats. For more information about Wiley products, visit our web site at www.wiley.com. Library of Congress Cataloging-in-Publication Data: Trace chemical sensing of explosives / edited by Ronald L. Woodfin. p. cm. Includes bibliographical references. ISBN-13: 978-0-471-73839-8 (cloth) ISBN-10: 0-471-73839-5 (cloth) 1. Chemical detectors. 2. Explosives—Detection. 3. Terrorism—Prevention. I. Woodfin, Ronald L. TP159.C46T73 2006 662 .20287—dc22 2006027027

Printed in the United States of America. 10 9 8 7 6 5 4 3 2 1

To all those courageous men and women, worldwide, who put themselves in harm’s way to search for hidden explosives in order that others may not suffer harm. May the concepts discussed herein assist them.

CONTENTS

FOREWORD

xv

PREFACE

xvii

LIST OF CONTRIBUTORS

xxiii

PART I 1

FUNDAMENTAL CONSIDERATIONS

CHEMICAL SENSING

1 3

Ronald L. Woodfin

1.1 1.2 1.3

1.4

1.5

What Is Chemical Sensing? / 3 Types of Sensing Systems / 3 Sensing Possibilities / 4 1.3.1 Bulk Sensors / 4 1.3.2 Trace Sensors / 5 Aromas / 6 1.4.1 Biosensors / 6 1.4.2 Electronic Sensors / 10 1.4.3 Other Indirect Methods (Switch of Molecules) / 12 1.4.4 Target Possibilities / 12 1.4.5 Sensitivity and the Problem of False Positives / 13 Configuring an Electronic Trace Sensor / 15 vii

viii

CONTENTS

1.6

2

1.5.1 Required Elements / 16 1.5.2 Integration and Packaging / 18 Issue of Concentration / 18 1.6.1 Nomenclature / 18 1.6.2 Source to Sample / 23 1.6.3 Catch, Count, and Release Cycle / 23 1.6.4 Sensor Sensitivity Versus Sampling Time / 23 1.6.5 The Concentration Gap / 26 1.6.6 Sensitivity Comparison / 27 References / 32

WHAT TO DETECT?

35

Jimmie C. Oxley

References / 41 3

DANGEROUS INNOVATIONS

43

Kirk Yeager

3.1 3.2 3.3 3.4

3.5 3.6

4

Introduction / 43 Theory of Improvised Explosives / 43 History and the Anarchist Literature / 45 Fertilizer-Based IEs / 51 3.4.1 Ammonium Nitrate IEs / 51 3.4.2 Urea Nitrate / 54 Peroxide Explosives / 55 The Next Wave / 63 3.6.1 Improvised Detonators / 63 3.6.2 Peroxide Main Charges / 64 3.6.3 Fringe Mixtures / 65 3.6.4 On the Horizon / 66 References / 67

WHERE SHOULD WE LOOK FOR EXPLOSIVE MOLECULES? Ronald L. Woodfin

4.1

4.2

Introduction / 69 4.1.1 Where Did the Molecules Come from and How Did They Get Here? / 69 4.1.2 Objects Other Than Buried Landmines / 70 4.1.3 Questions That Beg for Answers / 70 Source of the Molecules / 71

69

ix

CONTENTS

4.3

4.4

4.5

4.6

4.7

5

4.2.1 How the Molecules Diffuse or Leak from a Munition / 71 4.2.2 Example of Landmines / 73 4.2.3 Other Munitions / 76 Transport of the Molecules / 76 4.3.1 Buried Sources / 77 4.3.2 Concentration Estimates from Buried Sources / 94 4.3.3 Other Environments / 96 4.3.4 Odor Plumes / 97 EF&T Implications for Search and Sampling Strategies / 99 4.4.1 Sources Buried on Land / 99 4.4.2 Sources Producing Plumes / 99 Open Questions and Fruitful Areas for Future Research / 101 4.5.1 Objects Buried in the Sea Bottom / 102 4.5.2 Sampling Plant Material / 102 Role of Computer Modeling / 102 4.6.1 Soil Transport Models / 103 4.6.2 Plume Transport Models / 104 4.6.3 Plume Search Models / 104 Conclusions / 104 References / 105

STRUCTURE OF TURBULENT CHEMICAL PLUMES

109

Donald R. Webster

5.1 5.2 5.3 5.4 5.5

Turbulent Mixing / 109 Instantaneous Structure / 111 Time-Averaged Characteristics / 115 Information for Tracking Chemical Odor Plumes / 118 Variation of the Plume Structure / 125 Acknowledgments / 127 References / 127

PART II FIELD EXPERIENCE 6

DETECTION OF TRACE EXPLOSIVE SIGNATURES IN THE MARINE ENVIRONMENT Mark Fisher and Matthew Dock

6.1 6.2 6.3

Introduction / 133 Overview of Fate and Transport of Explosives Released from UUXO / 134 Sampling and Sensing Methodology / 135

131

133

x

CONTENTS

6.4

6.5

6.6

7

SeaDog Sensor Configurations / 137 6.4.1 Prototype Integrated with a Robotic Crawler Platform / 137 6.4.2 Diver-Deployed SeaDog and Initial Integration with the REMUS / 139 6.4.3 SeaDog Miniaturization: The SeaPup / 142 Results of Sensor Tests Conducted in the Marine Environment / 143 6.5.1 Tests of the Sensor Prototype on a Crawler Vehicle / 143 6.5.2 Tests of the Diver-Deployed SeaDog Sensor and Initial Integration to the REMUS / 145 6.5.3 Tests of the SeaPup Sensor Integrated on the REMUS / 146 Conclusions / 148 Acknowledgments / 149 References / 149

EXPLOSIVES DETECTION USING ULTRASENSITIVE ELECTRONIC VAPOR SENSORS: FIELD EXPERIENCE

151

Mark Fisher

7.1 7.2 7.3 7.4

7.5 7.6

8

Introduction / 151 Relevance of Field Testing To Sensor Development / 153 Overview of the Vapor Signatures of Explosives / 154 Landmine Detection / 158 7.4.1 Introduction to the Mine Problem / 158 7.4.2 Discussion of Landmine Chemical Vapor Signatures / 159 7.4.3 Landmine Detection Field Test Results / 164 Comparison of Fido with Canines Using High-Volume Sampling Methods (REST) / 170 Conclusions / 172 Acknowledgments / 172 References / 172

REFLECTIONS ON HUNTING MINES BY AROMA SENSING Vernon Joynt

8.1 8.2

Editor’s Note / 177 Interview / 177 References / 192

177

CONTENTS

PART III EXAMPLE SENSING TECHNOLOGIES 9

EXPLOSIVES DETECTION BASED ON AMPLIFYING FLUORESCENCE POLYMERS

xi

193

195

Colin Cumming

9.1

9.2 9.3 9.4 9.5 9.6 9.7 9.8 9.9

Introduction / 195 9.1.1 AFP Principle of Operation / 196 9.1.2 AFP Technology / 197 History of AFP and Fido® / 201 Fido Sensitivity / 201 Fido Performance / 202 Performance Limitations / 202 Physical Parameters / 203 Latest Implementations / 203 Maturity of the Fido Technology / 208 Funding / 208 References / 209

10 ION MOBILITY SPECTROMETRY

211

Ronald L. Woodfin

10.1 10.2 10.3 10.4

Introduction / 211 Brief Description of Principle of Operation / 211 Some Recent Developments / 214 Some IMS Manufacturers / 216 Acknowledgment / 217 References / 218

11 MASS SPECTROMETRY FOR SECURITY SCREENING OF EXPLOSIVES Jack A. Syage and Karl A. Hanold

11.1 Introduction / 219 11.2 Detection Methods / 220 11.2.1 Explosives Trace Detection / 220 11.2.2 Sampling Methods / 222 11.2.3 Quantitative vs. Screening Analysis / 222 11.3 Mass Spectrometry / 223 11.3.1 Primer / 223 11.3.2 QitTOF Mass Spectrometry / 226 11.3.3 Mass Spectrometry Versus Ion Mobility Spectrometry / 228

219

xii

CONTENTS

11.4

11.5

11.6

11.7

11.3.4 Other MS Analyzers Used for Explosives Detection / 230 Results / 231 11.4.1 Mass Spectral Signatures / 231 11.4.2 MS/MS Analysis / 232 11.4.3 Limits of Detection / 233 Detection Accuracy—A Model / 234 11.5.1 False-Positive Analysis / 234 11.5.2 Receiver Operator Characteristics (ROC) / 236 11.5.3 MS vs. IMS Accuracy / 236 Applications / 240 11.6.1 Personnel Screening / 240 11.6.2 Other Applications / 242 Summary and Conclusion / 242 Acknowledgments / 242 References / 243

12 EXPLOSIVE VAPOR DETECTION USING MICROCANTILEVER SENSORS

245

Thomas Thundat

12.1 Introduction / 245 12.2 Modes of Operation and Theory / 247 12.2.1 Resonance Frequency / 248 12.2.2 Thermal Motions of a Cantilever / 249 12.2.3 Thermal Effects—Deflagration / 250 12.3 Apparatus / 250 12.3.1 Cantilevers / 250 12.3.2 Excitation Techniques / 250 12.3.3 Readout Techniques / 250 12.3.4 Selectivity / 252 12.4 Results and Discussion / 254 12.5 Deflagration / 256 12.6 Conclusions / 259 Acknowledgments / 259 References / 259 13 LAB-ON-A-CHIP DETECTION OF EXPLOSIVES Greg E. Collins, Joseph Wang, and Christopher A. Tipple

13.1 Introduction / 261

261

CONTENTS

xiii

13.2 Lab-on-a-Chip Explosives Detection by Electrochemical Detection / 265 13.2.1 Amperometry for Explosives Detection / 266 13.2.2 Contactless Conductivity Detection / 268 13.2.3 Dual Amperometric/Conductivity Detection for Simultaneous Monitoring of Ionic and Organic Explosives / 270 13.3 Lab-on-a-Chip Explosives Detection Utilizing Optical Methods / 271 13.4 Lab-on-a-chip Sampling of Explosives / 277 13.5 Conclusions / 281 Acknowledgments / 281 References / 281 14 NANOSCALE SENSING ASSEMBLIES USING QUANTUM DOT–PROTEIN BIOCONJUGATES

285

Hedi Mattoussi, Aaron R. Clapp, and Igor L. Medintz

14.1 14.2 14.3 14.4 14.5 14.6 14.7

Introduction / 285 Quantum Dot–Protein Bioconjugates / 286 Förster Formalism and Quantum Dots as Energy Donors / 287 Quantum Dots as FRET Donors / 290 Quantum-Dot-Based FRET Nanosensors / 294 Surface-Attached QD–FRET Nanoassemblies / 296 Conclusions / 300 Acknowledgments / 300 References / 300

15 REMOTE SENSING OF EXPLOSIVE MATERIALS USING DIFFERENTIAL REFLECTION SPECTROSCOPY ¨ Rolf E. Hummel, Anna M. Fuller, Claus Schollhorn, and Paul H. Holloway

15.1 15.2 15.3 15.4

Introduction / 303 Differential Reflectometry / 304 Results / 305 Conclusions / 309 Acknowledgments / 310 References / 310

303

xiv

CONTENTS

PART IV SUPPLEMENTARY MATERIAL APPENDIX: ORGANIZATIONS INVOLVED IN SEARCHING FOR HIDDEN EXPLOSIVES

311

313

Charles O. Schmidt

International and Nongovernmental Organizations / 313 Commercial Demining / 315 Governments / 317 Military Systems / 318 Equipment / 319 University Research / 321 Information/Data Bases and Links / 322 Emeritus / 323 DEFINITIONS, SYMBOLS AND ABBREVIATIONS

325

Acronyms / 325 Symbols and Abbreviations / 330 EXPLOSIVES DEFINITIONS

331

BIBLIOGRAPHY

333

INDEX

351

FOREWORD

Dr. Woodfin and his associates who authored Trace Chemical Sensing of Explosives have performed a useful service for both the scientific and nonscientific communities. This volume takes a systems approach to the problem of chemical sensing of explosives. Considering Dr. Woodfin’s long background at Sandia National Laboratories in system engineering research, his approach to this volume is not surprising. The volume does not pretend to undertake a detailed, fundamental examination of the elements of sampling, concentration, detection, and understanding and interpretation of the results, which will probably disappoint many individual technical experts. But each technical expert in these individual scientific fields of the system should look at the entire system to understand how their individual specialty interacts with other technical specialties to make the total system more effective and all readers will benefit. This volume can best be understood by considering the olfactory senses of animals. Everyone has witnessed the ability of a dog to use its nose with its superb olfactory sense to detect trace odors. A “holy grail” of detection science has been to duplicate a dog’s nose electronically. Dr. Woodfin’s volume tries to assess where the scientific community stands in its search for this holy grail. The authors of this volume are using this chemical sensing approach to detect explosives. Explosives, as used in this context, are defined as materials that concentrate releasable energy; controllable or uncontrollable. A major detriment to successful detection of any explosive in trace concentrations has always been the problem of too many false-positive alarms. This book attempts to compare various chemical sensing approaches to alleviate this longterm bane of detection. The volume compares various approaches to sampling, concentrating, and finally sensing small amounts of material. The volume also xv

xvi

FOREWORD

attempts to understand how these various components of a sensing system can be optimized into a major detection unit. Dr. Ronald L. Woodfin’s unique background makes him highly qualified to edit a volume of this nature. His career at Sandia National Laboratories, universities, industry, naval laboratories, serving as consultant, in addition to being an ordained minister, makes him highly qualified to organize such a volume as this. I hope this volume will alleviate or reduce the dangers of hidden mines in future times. Dr. Fred E. Saalfeld Former Chief Scientist, Director, Technical Director, Deputy Chief of Naval Research, and Executive Director of the Office of Naval Research (ONR)

PREFACE

WHY THIS BOOK?

For all of recorded history mankind has located and identified certain items by their aroma. Whether it is a dead mouse in the closet or a freshly baking loaf of bread in the oven, we often make the identification correctly without seeing or touching the item. We have considered this sense so useful that when we find our own sense of smell to have inadequate sensitivity for a certain task we often “borrow” the more acute sense of smell from some animal. For centuries we have used dogs for hunting and pigs for truffle harvesting. Recent research has been directed toward finding ways to replace some of these noses, especially the borrowed ones, with electronic devices. In this book we will examine some of the technologies that have been developed and some that are currently being developed. The field is progressing rapidly enough that it is likely that other technologies that could be included will be announced while this book is in press. Nevertheless, we will provide here a basis for comparison of technologies, so that one who plans to develop or use a device will be able to make more informed choices. While many of these other applications have great value, this book will focus specifically on finding and identifying explosive-bearing objects. As is often the case in the history of technology, work of this nature began in several places at about the same time. Those of us working at Sandia National Labs were asked by a Navy/Marine Corps committee to adapt technology developed for other purposes to their need for better ways to search for mines, both land and sea mines. As we considered technologies currently applied to that effort, it became clear most of them found too many targets. The rate of false positives was so high as to greatly reduce the actual effectiveness of the devices being used. For xvii

xviii

PREFACE

example, a metal detector will find most mines, but it also finds every piece of metallic battlefield debris, bullets, fragments of exploded munitions, and general metallic trash. We recognized the need for a sensor that would focus on the real issue. The objects that one needs to find are those that contain explosives! This rather obvious observation led us to begin work on chemical sensors that look for the actual explosive molecules. Undoubtedly, other researchers came to similar conclusions because several programs were begun at about the same time to develop this basic idea. Chapter 4 provides a brief glimpse at the history of some of these programs. Explosives, like many technological developments in history, provide us with a way to concentrate energy to accomplish tasks more quickly. The benefits of explosives in industry, particularly in mining, in construction, and in agriculture, are substantial. However, many people associate explosives most often with military, paramilitary, or terrorist operations. As with any technology that uses high levels of power, explosives can be turned to dangerous, even nefarious purposes. When this happens it is often desirable to locate the explosives before they are actuated. This book explores some of the technologies that have been developed, or currently under development, for the purpose of locating explosives. This exploration concentrates primarily on the search for explosives by portable means. Chapter 9 forms an exception to this rule, describing some technologies that are directed to fixed position screening operations. The group of technologies included herein is by no means exhaustive. They are somewhat representative of the directions that research and development has taken recently. Some of the research has been done or sponsored by government agencies. Some has been commercially led. Some of the work will result, or has resulted, in systems committed to production. Our emphasis is not on products but on technologies. Our anticipation is that work will build on this and products will become available to enhance the search for hidden explosives.

PLAN OF BOOK

This book is composed of four basic divisions. Part I, Fundamental Considerations, includes Chapters 1 through 5 and discusses the problems and issues of finding and identifying small quantities of explosive materials within a given environment, with the hope of then being able to locate their hidden source. Part II, Field Experience, Chapters 6, 7 and 8, describes actual conditions encountered and procedures used. Part III, Example Sensing Technologies, Chapters 9 through 15, provides descriptions of some of the technologies under development at the time of this writing, and seeks to display some of the many ways that innovative researchers have applied the general principles described in Part I. Chapters 12 through 14 of this part are discussions of sensors at microscale and smaller. Part IV, Supplementary Material, Appendix 5 and following, includes contact information for organizations, definitions of acronyms and explosives, and a collective bibliography and index.

PREFACE

xix

In all the technology chapters, 9 through 15, the authors were free to organize and discuss their technologies as they considered best. These authors are actively working to develop the technology they are describing. Consequently, it is natural that each may show somewhat more enthusiasm for his own technology than a dispassionate observer reviewing several technologies might show. Competing technologies that are not represented are not omitted in order to favor those that are included. In some cases researchers with other technologies declined to contribute, usually from press of other obligations. The editor takes full responsibility for omissions that are due to his own ignorance and offers apology to those whose work he failed to include for that reason. It was done in ignorance, not in malice. The technologies presented provide a sampling of the possibilities, and an opportunity to examine the systems issues that the developer of any trace chemical sensing system must confront. The editor would welcome contact from researchers who are developing other technologies. The broadest exposure of these new ideas is beneficial to all except the terrorists; perhaps a sequel to this book is in order. WHAT IS IN THIS BOOK?

Part I, Fundamental Considerations Chapter 1, “Chemical Sensing,” introduces the subject with a general discussion of chemical sensing, exploring some of the possible avenues available. This chapter examines types and classes of sensing possibilities. It considers the required elements of a generic system and discusses the relationship of chemical concentration and sensor sensitivity. It includes a final section on the nomenclature of concentration and related issues. Chapter 2 “What To Detect?” Serving as a refresher in the chemistry of explosives, this chapter provides a basis for considering what molecules we may need to seek. Chapter 3, “Dangerous Innovations” This chapter continues the discussion of the chemistry of explosives by discussing some nontraditional explosives that are now being encountered in connection with various terrorist and paramilitary groups. It may become necessary to include these chemicals within our universe of compounds being sought, especially in improvised explosive devices (IEDs). Chapter 4, “Where Should We Look for Explosive Molecules?” In this chapter we examine the processes that affect the release of explosive molecules from a munition and their movement to some location where a sensing system may encounter them. The effects of environment and transport through soil, air, and water are explored. Chapter 5, “The Structure of Chemical Plumes,” continues the thrust of the preceding chapter by describing in detail the way small concentrations of chemicals are transported through water or air. Part II, Field Experience. Chapter 6, “Detection of Trace Explosive Signatures in the Marine Environment.” Since the main subject is underwater experience with the SeaDog sensor,

xx

PREFACE

it might sound like a collection of nautical yarns. In reality, it describes the results of applying a trace sensing technology (AFP) in SeaDog for plume tracking in an autonomous underwater vehicle (AUV), named REMUS, in a search for explosives hidden under water. Chapter 7, “Explosive Detection Using Ultrasensitive Electronic Vapor Sensors: Field Experience,” applies some of the concepts of the preceding chapters by describing system development work in field conditions. It describes some surprising head-to-head comparisons between dogs and electronic sensors. Chapter 8, “Reflections on Hunting Mines by Aroma Sensing” records some of the unique practical experience that Dr. Vernon Joynt has accumulated over many years of laboratory and field work. Chapter 9, “Explosives Detection Based on Amplifying Fluorescence Polymers,” describes a relatively new technology with very high sensitivity that has been applied successfully to several real-world search efforts. This is the sensing technology used in Chapters 6 and 7. Chapter 10, “Ion Mobility Spectrometry,” describes one of the most commonly used technologies in chemical sensing systems, with specific application to explosives, the one perhaps currently most used in portable systems. Chapter 11, “Mass Spectrometry for Security Screening of Explosives” This chapter is a little different from the other technology descriptions since it describes systems that are not portable. These are systems used to locate explosives in containers, or on personnel, that pass by a fixed point. The chapter also compares the features of mass spectrometry (MS) and ion mobility spectrometry (IMS) that often cause systems developers to choose one or the other for a specific application. Chapter 12, “Explosive Vapor Detection Using Microcantilever Sensors,” is representative of a general class of sensors “Surface Effect Microsensors” that use the change in some mechanical or electrical property of an ultra-small structure to sense and identify a wide variety of molecules. Chapter 13, “Lab-on-a-Chip Detection of Explosives,” describes one particular technology that has been successfully developed in a very small size. It too thus becomes representative of the possibilities that are offered by sensor systems of this genre. Chapter 14, “Nanoscale Sensing Assemblies Using Quantum Dot–Protein Bioconjugates,” also describes a kind of sensor technology in that ultra-small size, but one using a very different approach. Chapter 15, “Differential Reflection Spectroscopy for Detection of Explosive Materials,” describes a quite different type of technology that became available just as this book was in press. It is different from all the others described herein because it seeks to remotely locate and identify the explosive molecules in situ, whereas all the other trace sensing technologies require that some molecules be taken into the apparatus, ingested, in order to be sensed. This approach presents exciting possibilities, but is just emerging, with no field experience yet.

PREFACE

xxi

Part IV, Supplementary Material, begins with an Appendix, “Organizations Representing Those Who Search for Hidden Explosives,” cataloging some of those organizations, with contact information. Also included are a few pages for quick reference to acronyms and abbreviations, plus one to define all the explosives mentioned in the text. This section is called Definitions, Symbols, and Abbreviations. The Bibliography collects all the references cited in the text and adds a few. A Subject Index completes the work.

ACKNOWLEDGMENTS

A number of persons assisted or encouraged me in this effort. My wife, Martha, in addition to patiently enduring months of clutter and my reduced attention to normal household repairs, has proofread the manuscript several times in its various revisions. Her advice was always on track and I appreciate all her help. RADM John Pearson, USN (retired) and Prof. Al Bottoms of the Mine Warfare Association (MINWARA) were instrumental in getting this project started when they provided me an opportunity to chair sessions at the MINWARA conferences in 2002 and 2004. I initiated contact with several contributors as a result. RADM Pearson also got me the opportunity to be invited to lecture at the Naval Postgraduate School in 2003. I am grateful for both those opportunities, doubly so because that gave us the opportunities to visit our grandchildren. Gunner Schmidt, in addition to compiling the Appendix, read and commented on early drafts and helped me find references and people. Tom Obrien, my manager at Los Alamos Technical Associates (LATA), where I do occasional consulting, was kind enough to allow me the use of some computer software that I do not own. Paul Cooper of Sandia National Laboratories provided some valuable guidance in finding a publisher, directing me to Amy Byers, of Wiley, who has been most helpful. Jim Phelan of Sandia National Laboratories was kind enough to review the chapter I wrote about his work, Chapter 5. Since he was unavailable to write it himself, his assistance in “making sure I got it right” is greatly appreciated. The congregation of Cedar Crest Baptist Church has been encouraging, even though most probably regarded this as a distraction. Their patience is appreciated. Finally, I want to thank each contributor, particularly those whom I have never met face to face, for the work they have presented here. It is most obvious that their contributions are the true value of this book.


Aaron R. Clapp, Ph.D., Naval Research Laboratory Optical Sciences Division, Code 5611 Washington, DC 20376 Greg E. Collins, Ph.D., Naval Research Laboratory Chemistry Division, Code 6112 4555 Overlook Ave., S.W. Washington D.C. 20375-5342 Colin Cumming, MS, Nomadics Inc., an ICx Technologies Company 1024 S. Innovation Way Stillwater, OK 74074 Matthew Dock, Nomadics Inc., an ICx Technologies Company 1024 S. Innovation Way Stillwater, OK 74074 Mark Fisher, Ph.D., Nomadics Inc., an ICx Technologies Company 1024 S. Innovation Way Stillwater, OK 74074 Anna Fuller, Materials Science and Engineering College of Engineering University of Florida P.O. Box 116400 Gainesville, FL 32608 xxiii

xxiv


Karl A. Hanold, Ph.D., Syagen Technology, Inc. 1411 Warner Ave. Tustin, CA 92780 Paul H. Holloway, Ph.D., Department of Materials Science and Engineering Gale Lemerand Drive P.O. Box 116400 University of Florida Gainesville FL 32611-6400 Rolf E. Hummel, Ph.D., Department of Materials Science and Engineering P.O. Box 116400 University of Florida Gainesville, FL 32611-6400 Vernon Joynt, Ph.D., Force Protection, Inc. 9801 Highway 78, #3 Ladson, SC 29456 Hedi Mattoussi, Ph.D., Naval Research Laboratory Optical Sciences Division, Code 5611 Washington, DC 20376 Igor L. Medintz, Ph.D., Naval Research Laboratory Center for Biomolecular Science and Eng., Code 6900 Washington, DC 20376 Jimmie C. Oxley, Ph.D., University of Rhode Island Department of Chemistry 51 Lower College Road Kingston, RI 02881 Jack A. Syage, Ph.D., Syagen Technology, Inc. 1411 Warner Ave. Tustin, CA 92780 Charles O. Schmidt, CWO-4, USN (Ret.) Los Alamos Technical Associates Inc. 2400 Louisiana Blvd, Suite 400, Bldg 1 Albuquerque, NM 87110 ¨ Claus Schollhorn, Department of Materials Science and Engineering P.O. Box 116400 University of Florida Gainesville, FL 32611-6400 Christopher Tipple, Ph.D., U.S. Department of Defense 200 MacDill Blvd. ATTN: CLAR Washington, DC 20340


Thomas Thundat, Ph.D., Oak Ridge National Laboratory Nanoscale Science and Devices Group Mail stop-6123, Oak Ridge, TN 37831-6123 Kirk Yeager, Ph.D., FBI/EU 2501 Investigation Parkway Quantico, VA 22135 Joseph Wang, Ph.D., Arizona State University Departments of Chemical & Materials Engineering and Chemistry Ira A. Fulton School of Engineering Director Center for Bioelectronics and Biosensors, Biodesign Institute P.O. Box 876006 Tempe, AZ 85287-6006, USA Donald R. Webster, Ph.D., Georgia Institute of Technology School of Civil and Environmental Engineering 790 Atlantic Drive Atlanta, Georgia 30332-0355 Ronald L. Woodfin, Ph.D., Sandia National Laboratories, retired PO Box 55 Sandia Park, New Mexico 87047

xxv

PART I

FUNDAMENTAL CONSIDERATIONS

CHAPTER 1

CHEMICAL SENSING RONALD L. WOODFIN Sandia National Laboratories (retired)

We do a lot of chemical sensing. Mostly we do it without special equipment. But, whether it is the pleasant notification of tortillas toasting or bread baking or the jarring realization of a skunk or a cigar on the golf course, sensing, analyzing, and cataloging chemical signals are a part of our daily lives. We also use other means for chemical sensing. When we travel, we judge the local water hardness by the amount of soap we need to wash or shampoo. We use test kits to measure the chemistry of a swimming pool or a flower bed. In this chapter we will examine some basic ideas of chemical sensing as they apply to finding explosives. 1.1

WHAT IS CHEMICAL SENSING?

When we speak here of chemical sensing we mean the direct sensing of chemicals, rather than sensing of something else by means of chemicals. By that choice we will not consider those techniques often called “wet chemistry.” We then have a fairly clear idea that we mean to search for a specific chemical or suite of chemicals. Sometimes the search is for single chemicals like chemical warfare agents such as mustard or sarin. At other times combinations of chemicals, such as found in marijuana, are the object of the search. A short digression to consider the broader range of techniques and targets is appropriate in order to limit more precisely the scope of this book to the specific type of search indicated by its title. We broadly classify the possible types of sensing system according to activity and according to focus. 1.2

TYPES OF SENSING SYSTEMS

First, we may class sensors according to the activity required of the sensing system. Hence a system may be passive, active, or semiactive. Passive sensors Trace Chemical Sensing of Explosives, Edited by Ronald L. Woodfin c 2007 John Wiley & Sons, Inc. Copyright

3

4

CHEMICAL SENSING

do not produce any emanations directed toward the target; they are based on sensing something emanated or released by the target without stimulation by the sensing system. Conversely, active sensors need no emanation or release of anything from the target; they provide some sort of emission that interacts with the target. The active sensor then senses the alteration in its emission by the target. Semiactive sensors operate in yet another way. They provide a stimulation that produces some sensible emission or emanation from the target, which they then sense in the manner of a passive sensor. 1.3 1.3.1

SENSING POSSIBILITIES Bulk Sensors

There are two basic ways to look for explosive material. They differ in their point of focus. Some sensors seek the mass of explosive material within a device. These are particularly useful when the device is well sealed and its surface is well cleaned of stray explosive molecules, or when the explosive being used is nonaromatic, that is, it does not readily release molecules from its bulk. We will refer to these as bulk sensors. They include X-ray techniques, both transmission and backscatter; neutron activation in several techniques; γ -ray excitation, in either transmission or backscatter modes; and nuclear resonance techniques, either nuclear magnetic resonance (NMR) or nuclear quadrupole resonance (NQR). Bruschini [1] has described these thoroughly. They are also described by the staff of the Jet Propulsion Laboratory [2]. The following forms a very brief synopsis. 1.3.1.1 X-ray Techniques X-ray techniques are familiar because of their use in medical diagnosis. The basic concept is that material of different densities or chemical compositions absorb and scatter X-rays differently. When the X-rays pass through the materials and strike the film or detector, they form a gray-scale image. After proper calibration a bulk charge of explosive may be inferred from this image. In many applications, such as seeking buried munitions, it is not possible to place the X-ray source and the detector on opposite sides of the objects being investigated. In this case techniques have been developed to form images from the X-rays that are scattered back toward the source, or backscattered. 1.3.1.2 Neutron or γ Sensors Neutron- or γ -based sensors are similar in concept to the X-ray sensors. They use different forms of excitation and different detectors, but the basic forms of transmission or backscatter follow the pattern described above. Both normally rely on extensive computation for signal processing called computed tomography, where the detector signals are combined to synthesize an image of the irradiated object. 1.3.1.3 Electromagnetic Techniques Sensors based on the related principles of NMR and NQR have been successfully developed. These are active techniques that excite the electromagnetic interactions between the atomic nuclei

SENSING POSSIBILITIES

5

by means of an external magnetic field. In NMR the magnetic field is a dc, or fixed, field, whereas in NQR it is an oscillating, or ac, field. The detector is tuned to respond to resonant frequency of a molecule of interest, such as RDX [3]. Both portable and fixed systems have been demonstrated. It may be possible that ground-penetrating radar (GPR) could be tuned to preferentially indicate masses of particular chemicals, but such work is not known to have been reported. 1.3.1.4 Bulk Sensor Targets All the bulk sensor technologies have a common thread. They seek to find a mass of material with certain physical properties and to distinguish its shape when other materials obscure it. The object may be buried in the ground or contained within a vehicle, a structure, or container such as a crate or luggage. Potential targets are mines, unexploded ordnance (UXO), improvised explosive devices (IEDs), or drug caches. We are perhaps most familiar with the medical search for tumors or foreign objects within a body using computer-aided tomography, or CAT scan, and magnetic resonance imaging (MRI). 1.3.2

Trace Sensors

Bulk sensors certainly have a role in chemical sensing of explosives, but the subject of this book is the other basic type sensor, one that seeks molecules released from the bulk of the explosive material in an object. We will refer to these as trace chemical sensors. They are sometimes called vapor sensors, but that seems a less accurate description when they are applied to explosive molecules, which may not always be found in a vapor state. As we shall see in Chapter 5, that requires us to understand where and how to look for these molecules. It will become apparent upon a little reflection that the two types of sensors are complementary and are best used in different situations. Furthermore, even when trace sensors are used, in some situations sampling of particles of soil or vegetation or sampling from surfaces may prove to be more productive that vapor sampling. For underwater sources the term vapor sensing is also inappropriate. In fact, with the very recent addition of differential reflection spectroscopy (DRS) to the suite of applicable technologies, as described in Chapter 15, we now have the possibility of sensing trace quantities of explosives where they are most often found in the environment, adsorbed to solid surfaces. Technologies that can, like DRS, locate these traces in situ offer a very different way to approach the problem. There have been several recent attempts to do this in situ detection from some distance away. To date the DRS seems the most successful. It has demonstrated detection at a range of a few meters. The marked advantage of an in situ detection system is that it does not ingest the molecules; hence it does not disturb the area, does not require as close approach to an explosive device, and it should produce a faster response time than ingesting systems. The disadvantage of in situ systems is also that they do not ingest the molecules; hence they have no means of concentrating the sample

6

CHEMICAL SENSING

for sensing. Thus they must have greater direct sensitivities in order to produce comparable results. This book was in press before the first publication of the DRS results. Therefore, the following chapters were written with the expectation that most sensor systems would require ingesting samples. This condition is expected to continue, since in situ sensors will most likely supplement, not replace, ingesting sensors. Nevertheless, the discussion applies to the in situ systems by recognizing the high proportion of molecules that are normally adsorbed to surfaces, such as dust, plants, manmade objects, and the like. These form the target for the in situ systems. Trying to apply in situ systems like DRS to vapors and plumes is unlikely to lead to success; but, when vapors and plumes strike solid objects, there is a marked tendency for the entrained molecules to adsorb to the surface, producing opportunity for the in situ sensor. 1.4

AROMAS

The ability to recognize, locate, or distinguish among specific materials by their characteristic aromas can be useful in a variety of applications. Our first warning of a fire is frequently the smell of smoke. Wine and tea tasters regularly use their sense of smell to assist in quality control of these food items. Similarly, when we buy fruits such as peaches, we often sniff them to judge quality. We become aware of poor water quality before tasting the water because it “smells bad.” It is well known by medical personnel that certain diseases and conditions can be diagnosed quickly from the patient’s aroma. Farmers sometimes judge when to harvest from the aroma of the produce. We often see television news reports of trained dogs searching for contraband illegal drugs. Trained soldiers can recognize some chemical agent attacks from the first whiff of the agent, hopefully in time to protect themselves. Dogs and other animals are being effectively used in searching for hidden explosives in a variety of situations, including humanitarian demining and antiterrorist patrols. In each of these situations, sensing the aroma, which, for the purposes of this book, we will consider to consist of a specific molecule or suite of molecules that is uniquely produced by its source, provides the means of identification and/or location of the source. In each of these examples the user would gain significant advantage if a very sensitive and specifically “tuned” electronic sensor, which could accurately and reliably identify the characteristic aroma for that application, were available. 1.4.1

Biosensors

The historical use of the biologically based sensors mentioned above might lead us to the conclusion that finding hidden sources of explosives simply means training some kind of animal. Indeed, that conclusion has merit, and a great deal of success has been recorded in that way. This is discussed somewhat more fully in Chapter 8. There are, of course, disadvantages to this technique. The

AROMAS

7

possibilities of biosensors have certainly not been exhausted in the search for explosives or other trace chemicals. As an aside, designed to pique a hope of even broader application for technologies developed primarily to search for explosives, consider a recent finding by researchers at Amersham Hospital in the United Kingdom. [4] They report success in using dogs to detect cancers. For centuries people, particularly nurses, have been aware that certain conditions, for example, gangrene, could be detected and diagnosed by smell. This latest finding offers hope that research will make us able to apply electronic vapor sensing to medical diagnosis as well as our immediate objective of finding hidden explosives. There have been efforts directed toward using bacteria in the search for explosives1 [5]. Whether this approach offers a potential for developing practical sensing systems based on bacteria is an open question. 1.4.1.1 Mammals Perhaps because we have a long history of domesticating mammals for various tasks, the preponderance of explosive search to date has been by mammals, mostly dogs [6, pp. 165–174], though some work has been done using rats [6, pp. 175–193] and pigs [7]. There have been two principal approaches used, which may be characterized as direct and indirect search. These techniques are discussed in detail in McLean [6]. Direct Mammalian Search In the direct search technique the animal, usually a dog, is brought to the suspected location of the hidden explosive. The animal is trained to go to the strongest scent of the aromas presented during training. The animal is trained to adopt a particular behavior when nearest the source. With dogs, this is usually to sit near the source. This technique has the advantage of immediate results; hence it is the method most often adopted by military personnel or police forces when the threat is immediate. This direct approach has several distinct disadvantages as well. When animals such as dogs are used in a field environment, they become subject to the hazards of that environment. Approaching an explosive device puts both animal and handler in the danger zone. Of course, in such a zone there are often related dangers. Also, because of the danger, the animal is restrained, usually by training, from actual contact with the source. Hence there is always an area of uncertainty surrounding the actual location of the source, particularly a buried one. Animals find a multitude of interesting distractions in a field environment. Since they operate in a world of aromas undetectable to us, it is hard to realize how many distractions there may be. Certainly food aromas and sexual stimulation aromas are present. Animals are subject to physical ailments, just as we are. A dog with the equivalent of a “head cold” is not as effective in search. Providing proper care for the animal in such locations often becomes difficult and expensive. Training is also time consuming and expensive. In spite of all 1 http://www.ornl.gov/info/ornlreview/meas tech/threat.htm and http://eerc.ra.utk.edu/insites/ins54.htm#critters. Both sites visited 9/28/05.

8

CHEMICAL SENSING

these difficulties, the use of dogs in direct search for hidden explosives remains one of the most effective, arguably the most effective, method for finding hidden explosives. The limit of detection2 (LOD) for explosive searching dogs has been a subject for active debate for years. Chapter 8 includes some discussion on this matter. One report [8] attempts, with partial success, to quantify this LOD. Indirect Mammalian Search In the indirect method, called in general terms the REST concept, for remote explosive scent tracking [6, pp. 53–107, 9], samples of chemicals in air are trapped in special filter cartridges by passing large volumes of air from the search area through them. In principle, the concept could be extended to water also. The filter cartridges are then taken to a laboratory, or other protected area, where they are offered to trained mammals, sometimes dogs and sometimes rats, for sniffing. When the animal detects the aroma for which it has been trained, it gives the proper response. The sample is correlated with its collection location so that more detailed sampling or direct search can be initiated. This technique was first developed in the late 1980s in South Africa, largely through the work of Dr. Vernon Joynt and colleagues at Mechem Consultants, a division of Denel (Pty) Ltd. They called it MEDDS (Mechem Explosives and Drug Detection System). The technique has been called EVD (Explosives Vapour Detection) by Norwegian People’s Aid (NPA) and Checkmate elsewhere. The term REST is now applied generically. REST was conceived for solving the most pressing question in Humanitarian Demining, or in any postconflict ordnance cleanup operation: “Where are there no explosives?” The most immediate need is always to identify those areas that are free of explosives, so that they can quickly be returned to normal activity, such as agriculture and transportation. Then the detailed search and removal efforts can be concentrated on the areas that actually contain explosives. REST has been used with success in several postconflict areas, especially in Africa. A typical REST application mounts an apparatus on a vehicle that can move safely through the suspect area. The apparatus has a vacuum pump that pulls air through the removable filter cartridge. A material for which explosive materials have an affinity is chosen for the filter. That is a material on whose surface the explosive molecules tend to tightly adhere. Most often, that tendency is a function of temperature, so the molecules adhere more tenaciously at lower temperatures and release as the temperature is increased. This characteristic is one often exploited in electronic sensing systems as well. When a certain region, say a predetermined length of a particular road, has been sampled, the filter cartridge is removed and stored. A new filter cartridge is inserted and the next area sampled. Filter cartridges are then offered to the trained mammal in a controlled environment. This technique is essentially similar to the one used with the 2 We

can conveniently refer to the smallest quantity of any given compound that an animal or a technology can detect as its LOD.

AROMAS

9

cancer seeking dogs [4]. Large areas can be surveyed quickly, safely, and inexpensively. REST has also been effective in searching for contraband at border crossings. REST need not necessarily involve any animals at all. If electronic sensors of adequate sensitivity are developed, they can replace the animals. Certainly, the history of electronic instrument development shows that the earlier generations of any device are more suited for laboratory than field use, and that laboratory units can normally be expected to show better performance than portable ones. Calibration of an electronic instrument, which corresponds to training of a mammal, should become more precise and dependable than that training. 1.4.1.2 Insects It has long been recognized that many insects may have a more acute sense of smell than any mammal. A consensus found in articles by many researchers seems to indicate that only a few molecules are required for odor detection by some insects. We have little experience in domesticating insects. Usually, we simply gather the ones we deem useful, mostly bees and silkworms (the larvae of an insect, the Bombycid moth, Bombyx mori ), to a convenient location and let them do their normal activity. We then take their produce when it is available. Even within this limited process we can find ways to use them for chemical search. Researchers have found that bees can be trained to seek out sources of explosive molecules by the simple expedient of feeding them sugarwater laced with the target molecules [10]. Training bees is thus much simpler and quicker than training mammals. Bees have an obvious advantage when entering a minefield or similar danger area. They are too small to actuate any sort of mine, so they are completely safe. Thus, they have one of the most desirable characteristics for sensing systems. They do not put the searchers at risk. With an appropriate method for tracking the bees and geographically fixing the points where a concentration of foraging bees indicates a source of explosive molecules, a minefield can be mapped for efficient search and removal. Such techniques have been offered [10]. After discovering that bees tend to forage along a particular vector from the hive, the researchers found it possible to “aim” the bees in a general direction and track them with a LIDAR3 system developed at Sandia National Laboratories [11]. Recently, wasps, Microplitis croceipes, have been conditioned to detect explosives [12]. Training is similar to that of the bees, but instead of allowing the wasps to range freely, a handheld, air-ingesting sensor has been developed with the wasps as the sensing element, with a video system for monitoring behavioral indications of detection [13]. The device, called Wasp Hound, was developed originally for other purposes, primarily agricultural, but has been successfully demonstrated in detecting 2,4 DNT. No detailed information on filed LOD for explosive compounds is yet available, but comparison with one commercial electronic sensor, Cyranose 320, [14] seems to indicate somewhat better LODs for some chemicals. Thus, Wasp Hound can be classified as a hybrid sensor, part biological and part electronic. 3

LIDAR: light detection and ranging.

10

1.4.2

CHEMICAL SENSING

Electronic Sensors

The term electronic is used here in the most generic sense. Since the objective is to find and identify small quantities of explosive molecules by extracting them from their environment, the problem is analogous to many solved with electronic devices. Common examples include radio and television receivers and electronic instruments for measuring physical quantities such as weight, acceleration, or temperature. The principal difference here lies in the sensing modules that form part of the systems we will consider. Of course, it is conceivable that a system can be developed that uses no electronics, perhaps using color changes, or something else, but the systems discussed herein all use some electronic modules to enable the observer to know when explosive molecules have been located. Whatever system may be developed, the basic objective is the same. Different technologies may be employed, but the reason for the system is always to locate and identify explosive molecules in air, water or soil, or on a surface, in order to pinpoint the source of these molecules. Ideally, the source is located before it can be actuated in a harmful way. Consequently, every system developed for this purpose will have some characteristics in common with any other developed for the same purpose. Therefore, it is valuable to consider those common elements in a general way. We are faced with locating molecules that are sparse within the environment, hence the term trace. There are several basic steps that are necessary, independent of the design of the specific system or the technology being incorporated. Table 1.1 lists those steps and correlates them with more common electronic instrumentation nomenclature. Some of the actions will involve TABLE 1.1 Common Processes in Locating and Identifying Explosive Molecules within the Environment Step Sample

Action

Introduce a quantity of explosive-bearing medium into the system Separate Pass the sample through some process that preferentially selects the explosive molecules Concentrate Collect the separated molecules while additional sampling and separation continues Identify and quantify Apply some signal processing procedure or algorithm to infer the in situ concentration of explosive molecules in the samples Present Organize the results in a form that can be interpreted by an operator or that can be used to guide further activities in autonomous devices

Electrical Analog Receiver or transducer Tuner

Filter/amplifier

Signal processor

Display

AROMAS

11

physical processes, such as pumping air or water past a collector. Some may use chemical processes, such as tagging the explosive molecules with other molecules that can be sensed more easily or sensitively. Some will use electronic processing; some may use radiation, such as ultraviolet light. The variety of possibilities has not yet been fully explored. Several types of systems have been, or are being, developed. The following descriptions are presented as a brief summary. Detailed descriptions will be found in the chapters devoted to each technology. 1.4.2.1 Spectrometry “Spectroscopy4 is basically an experimental subject and is concerned with the adsorption, emission or scattering of electromagnetic radiation by atoms or molecules” [15, p. 1]. A wide variety of applications of this concept have been applied in analyzing many substances. In the particular case of explosive molecules the most prominent are several forms of mass spectrometry and ion mobility spectrometry. Each has certain advantages and disadvantages. Each is discussed in detail in a later chapter. The former is most often used in fixed applications; the latter, in both fixed and portable applications. 1.4.2.2 Surface Effect Sensors Surface effect sensors have been offered in several forms. In each design they rely on using a substrate, often a microstructure, coated with a material having an affinity for adsorbing specific molecules. When the molecules adhere to it, they alter some property of the substrate. Depending on the particular concept the property exploited may be physical or electrical. Some designs measure altered resistance; some, altered capacitance. Some designs measure alterations in physical effects, such as the frequency of vibration of a microcantilever or the frequency or amplitude of an induced surface acoustic wave (SAW) [16]. Sensors of this type tend to be very good at identifying a broad range of chemicals, particularly complex aromas that contain suites of compounds. They have been used to sort fruits for ripeness and can distinguish among fruits. These sensors use an array of sensing elements, each coated with a different material, to adsorb different molecules. They then form signature patterns for different aromas in a calibration process not completely unlike a dog’s training. Signal processing algorithms then match patterns of sampled aromas to known patterns for identification. Acoustic wave devices have been used, primarily as electronic bandpass filters, for more than 60 years, but the first reported use as a chemical sensor appeared in 1979 [16, p7]. A form of surface effect sensor that exploits altered surface resistance, or chemiresistors, forms the surface from a mixture of tailored polymers and a finely divided conductive material, such as carbon black, as a thin film on a substrate. They use a number of polymers, 32 in one implementation, with different properties to form an array of chemiresistors. When a vapor is passed over the array, 4 Spectroscopy

uses photographic means to record the spectrum; spectrometry uses photoelectric recording [15, p. 67].

12

CHEMICAL SENSING

the film swells. Each film swells differently. The pattern of altered resistances forms a signature for that vapor. Sophisticated signal processing matches these signatures to known aromas, much as fingerprints are matched. Chemiresistors have been successfully employed to detect undesirable odors in food packaging materials and to distinguish among five varieties of diesel fuel [17]. When the design focus becomes this kind of specificity among aromas, the sensitivity for simple compounds such as explosives may be reduced. A design focused directly on explosive molecules will likely prove more adaptable to the search for hidden explosives. Chapter 12 describes one of these sensors in detail. To date none of these sensors has produced an LOD adequate for field search for explosives. 1.4.2.3 Amplifying Fluorescence Polymers (AFPs) The development of materials that fluoresce intensely in the presence of ultraviolet light when no nitroaromatic explosive compounds are present, but are prevented from fluorescing when those compounds are introduced, has led to the development of a new class of sensors. Upon encountering a nitroaromatic molecule, the tailored fluorescing polymer binds with it. The bound pair no longer fluoresces when struck by a photon. This behavior is exhibited by a class of polymer compounds called chemophores, each having an affinity for a particular class of analytes. The characteristic that distinguishes AFPs from other chemophores is their forming of long chains with as many binding sites as molecules forming the chain. A single explosive molecule, binding at any one of these sites, is sufficient to induce quenching of the fluorescence along the entire chain; thus, the appellation of “amplifying.” The intensity of the quenching produced by the binding of one explosive molecule is thus multiplied many times. Very low concentrations of explosive molecules are thus detectable. Chapter 9 describes the way AFPs have been incorporated into a family of fielded sensing systems. Chapter 6 examines the experience gained in using one of this family in an unmanned underwater vehicle (UUV) called SeaDog. Chapter 7 describes field experience with AFPs in air, as implemented in the Fido® sensor. 1.4.3

Other Indirect Methods (Switch of Molecules)

Methods similar to AFPs in that they introduce other molecules have been suggested. The concept is that molecules that are easier to detect than explosive molecules, but only when the target molecules are present, can be introduced to increase the sensitivity of the detector system. With the introduction of AFPs that concept is clearly proven. Other adaptations are to be expected in the future. 1.4.4

Target Possibilities

When we consider possible targets in the context of system design we may group them into three broad classes, all of which may be considered as unexploded ordnance, UXO. Classes 1 and 2, and sometimes all three, are termed explosive remnants of war, ERW by the United Nations, UN.

AROMAS

TABLE 1.2

Possibility Space for System Indication of Detected Explosive

System Indication

Explosive No Explosive

13

Actual Condition Explosive Present True Positive False Negative

No Explosive Present False Positive True Negative

Class 1 includes any munition, whether standard or IED, that is deployed with the intent of causing damage to an opponent during a belligerency of any nature. These form two subclasses, those that are “fresh” and those that are abandoned. Clearly, those fresh ones constitute the most urgent targets. Those that are abandoned may be functional, as with mines, or malfunctioning units. These “duds” may still constitute an immediate danger, depending on the reason for their malfunction. Class 2 includes those munitions that have become lost or discarded during military operations or transport. Major quantities of ordnance are lost when ships are sunk or blow up. When munition storage areas are attacked, some munitions are ejected to substantial distances. These munitions may be fully assembled and functional, as when an aircraft ejects them. Others may be incomplete, likely unfuzed. Class 3 includes those ordnance items that are intentionally buried, dumped at sea, or otherwise disposed of either in an intact or in a partially assembled condition. It also includes military or civilian storage and military practice ranges. Each class has distinct characteristics that should be considered in optimizing a system design. The level of urgency varies greatly. The consequences of false negatives also vary greatly. The mention of false negatives begs a bit of explanation. During a search for hidden explosives the sensing system continually operates within a two-by-two space of possibilities, as diagrammed in Table 1.2. When the system provides the correct indication, whether positive or negative, it is functioning properly. However, there are two different error conditions that can be realized. These are indicated by the shaded blocks. A false negative means a missed target. Obviously, that can lead to immediate danger for the operator or others. A false positive induces a different kind of danger, deferred danger, due to loss of operator confidence in the system. 1.4.5

Sensitivity and the Problem of False Positives

One of the primary characteristics of any sensor is its sensitivity. In principle, increased sensitivity should lead to increased performance, that is, a greater likelihood of finding the object being sought. However, increasing the sensitivity in a cluttered environment often leads instead to an increase in “false positive.” A false positive occurs when the sensor system correctly detects an item of the

14

CHEMICAL SENSING

class it is designed to detect, but an item that is not the item being sought. For example, when a metal detector is used to search for buried landmines, it also detects odd bits of metallic debris in addition to the mines. This can greatly impede the search for the mines. It produces deferred danger when the operator, having encountered several false positives, becomes skeptical of a true positive. One solution is to reduce the sensitivity of the sensor. Such a reduction in sensitivity may also reduce the probability of detecting the mines. Therefore, in any search the probability of detection, PD , must be balanced against the probability of false positive, PFA . This is discussed in a little more depth in Chapter 11. For some sensors seeking explosive molecules, an increase in sensitivity can lead to the detection of molecules other than explosives. These can be interpreted as false positives. Clearly, if there is unburned explosive material present in an area of interest, then the trace chemical sensor will register a false positive. Similarly, some bulk chemical sensors may provide false positives when they encounter a mass of nitrogen-rich material, such as fertilizer or feces. In either case the rate of false positives is still likely to be less that the false-positive rate of a metal detector. Adjusting the sensor sensitivity is usually only partially successful. There is another way to overcome the problem of false positives. A suite of sensors that are “orthogonal in principle” can be used together. 1.4.5.1 Orthogonality In geometry lines or planes are called orthogonal when they meet at a right angle. Values or conditions on one are therefore independent of those on the other. We consider two sensors to be orthogonal when they depend on different principles of chemistry or physics. In that sense, a chemical sensor is clearly orthogonal to a metal detector. When searching for hidden, explosive-bearing objects, false positives can be dangerous, as finding too many can cause an operator to lose concentration. Therefore, when one needs to identify an object detected by any sensor, that means that indications from two or more sensors, each of which exploits a different physical or chemical property of the target, need to be compared before identification can be considered definitive. For example, a set of three sensors that are orthogonal in this sense is illustrated in Figure 1.1.

Metal Detector (Senses changes in an induced electric field caused by the presence of metal)

Visual Sensor (Senses the shape, size, texture, and appearance of an object)

Trace Chemical Sensor (Senses presence of specific molecules)

Figure 1.1

Orthogonal sensor suite.

CONFIGURING AN ELECTRONIC TRACE SENSOR

15

It is normally advisable to adhere to the principal of orthogonality in some way. In the above example the three sensors could be independently applied, not necessarily part of a single package or system, especially if the object in question is where it can be seen directly. 1.5


Before anyone begins development of a sensor system, the first task is to determine what is being sought. That is not necessarily the procedure when developing a sensing technology. Thus, it is common for a technology to find its most effective application in systems that were not originally envisioned. Perhaps the most apparent need that led to the development of some of the trace chemical sensing technologies was the need for a method of warning troops, or populations, when they were being attacked with chemical weapons. While this still forms a major application of these technologies we will consider one that was not as apparent originally. The original need that led to development of trace chemical sensors for explosives was the need to restore land that had been abandoned to public or private use. This land was abandoned because of the presence of, or perception of the presence of, mines or unexploded ordnance, often called UXO. These potentially lethal items could be the result of some earlier armed conflict. In that case it is now common to refer to them as explosive remnants of war, or ERW. In some cases the war that left its remnants was concluded many years ago. Dangerous ERW are still found on World War I battlefields, and occasionally on U.S. Civil War battlefields, though the latter munitions are usually deteriorated beyond holding any dangerous explosive residue. UXO are also found on former military training facilities, such as Ft. Ord, California, or Kahoolawe, Hawaii. Both these sites have undergone restoration in recent years. Since September 11, 2001, the realization that explosive devices might be deposited almost anywhere by terrorists or other malefactors has added a new group of devices to the list of targets. These are improvised explosive devices, or IEDs. An IED may take any form, subject only to the imagination of its fabricator. Many contain traditional munitions but are constructed to permit delivery and/or actuation in nonstandard, or improvised, ways. Examples are “dud” artillery shells or aviation bombs that have been salvaged and equipped with a fuze for use as a mine or booby trap. Car bombs are also common examples of IEDs. When IEDs are constructed using salvaged munitions, they will naturally include conventional explosives; hence they may be appropriate targets for trace chemical sensors. However, IEDs are not always made in this way. Chapter 3 discusses a variety of nontraditional explosives that are being discovered in IEDs. In this book we will focus our attention on applications that require portable sensors, sensors that must be transported to the vicinity of the target. These applications include mine and UXO removal and the search for and identification of IEDs. Chapter 11 forms a notable exception to this focus, where fixed screening of airport portals is the principal application to date. Fixed sensing stations

16

CHEMICAL SENSING

are considered more fully by Oxley [18]. Our focus is due to the very different conditions in the two applications. We will devote a chapter, Chapter 4, to the peculiar subject of the environment fate and transport, or EF&T, of that elusive quarry, the free explosive molecule. These free molecules are the trace chemicals that we are trying to sense. We will explore how they become free, the environmental factors that affect their mobility, and where we may expect to find them in sufficient concentration to be detectable.

1.5.1

Required Elements

There are some common elements that any trace chemical sensing system must have. Each system will have a defined purpose and be subject to specific size, weight, power, mobility, performance, and cost constraints. These, along with the imagination of the system designer, will steer the designs of different systems in individual ways, but the basic elements will remain, though their forms may be very different among a group of systems. The terminology used here is that commonly used in electronic instrumentation, but in some cases the element may have no electronic components. The elements are described in sequential order. In each case, except as noted, the function of each follows that of the preceding one. Conceptually, one samples, then concentrates, then senses, and so forth. No inference about the movement of material from one stage to the next is implied. That movement is a unique property of the system design. 1.5.1.1 Sampling Unit Since we are seeking the free molecules that form a trace chemical “signal,” we must have a unit that collects them from their environment. For some systems that environment is air, for others water or soil. For some, it may be a surface, such as a parcel, a vehicle, a leaf, or a wall. It will be necessary for the system designer to devise a way to collect the molecules from that environment of interest. While most of the system elements are more or less independent of the search environment, the sampling unit is entirely dependent on it. Sampling units may be as simple as a piece of chemically treated paper used to “swipe” a surface or one with multiple parts such as pumps, valves, scoops, and heating units. It may be physically integrated into the rest of the system or operated separately, as in the case of the swipe paper. The sampling unit will always be required to accept some of the environment along with the quarry molecules. Because these bits of the environment obscure the trace signal, it is necessary to sort through the sample in some way to collect as many as possible of the quarry molecules together while excluding as much as possible of the environment. This process is the function of the concentrating unit. In many practical systems the sampling unit and the concentrating unit may be so integrated as to be physically inseparable. Nevertheless, their functions remain conceptually distinct.


17

1.5.1.2 Concentrating Unit As described in the preceding paragraph, the concentrating unit sorts through the sample taken by the sampling unit to extract the molecules of interest. Sections 1.6.2 to 1.6.4 discuss some of the considerations that determine the design of a sampling unit and a concentrating unit. Concentrator materials of choice are often polymers. Polydimethylsiloxane (PDMS), or a polydimethylsiloxane/divinylbenzene copolymer (PDMS/DVB) are favored choices for explosive molecules. PDMS are often used in the form of solid-phase microextraction (SPME) fibers. PDMS/DVB is often used in the form of microspheres with diameters in the 50- to 75-μm range. Detailed considerations for use of SPME fibers is given on a website maintained by the University of Western England [19]. It references a more complete treatise [20]. Other geometries, such as stacked spheres, have also been used successfully [21]. Some concentrating units use chemical solvents; some use mechanical methods, sometimes coating a surface with a material for which the quarry molecules have a distinct affinity. One of the most convenient characteristics of explosive molecules is the way temperature affects their adhesion to surfaces. They adhere readily to cool surfaces but are easily released by a modest rise in temperature. Concentrating units often exploit this characteristic by alternately chilling and warming a collection surface. The surface is chilled while sampling and warmed for sensing. 1.5.1.3 Sensing Unit The heart of the sensor system is the sensing unit. It is here that the wide variety of sensing technologies discussed in Chapters 8 through 14 are employed in different systems. As systems are designed and integrated, the various units may be merged physically, but the sensing unit is where the quarry molecules are identified as present or not in a concentrated sample. The success of the entire system is determined by the performance of the sensing unit, but does not function without the other units. We often associate the overall performance of the system with the sensitivity of the sensing unit. Design of the other units can either enhance or degrade that overall performance by large factors. Nevertheless, the search for more sensitive sensing technologies is ongoing and needed. Along with sensitivity, the other dominant characteristic of a sensing technology is specificity. A technology that is very sensitive but lacks specificity leads to many false positives, or a high probability of false positive (sometimes called false alarms), PFA ; one that lacks sensitivity provides a lower probability of detection, PD . False positives cause the search to become diverted in an attempt to positively identify each one. They slow the search and can become dangerous if they obscure the real target or lull an operator into complacency. A low PD becomes dangerous when a potential hazard, such as a mine, is left undetected. A design that maximizes the PD while minimizing PFA is the goal for sensing technologies. 1.5.1.4 Amplifier Some systems may use an amplifier. This unit may use a process of chemical replacement or reaction to produce secondary molecules from the few quarry molecules collected. These molecules may be more numerous,

18

CHEMICAL SENSING

easier to detect, or exhibit a characteristic that permits a different principle of detection to be employed. This kind of amplifier may precede the sensing unit. Another kind of amplifier may be used, one that simply increases the intensity of the sensing unit output when a quarry molecule is detected. This type amplifier will follow the sensing unit. 1.5.1.5 Signal Processor Electronic instruments normally employ signal processing of some form. Sometimes the calculations are quite extensive and complex. Often the object is to extract a signal from its background. In our case, that is also the function of the concentrating unit. However, it may be possible to further improve the system performance by some form of signal processing at this stage. This process is sometimes termed chemometrics. 1.5.1.6 Output and/or Display Unit Finally, the results of the process to this point must be formatted in a manner appropriate to the purpose of the system, so that some action can be taken. If the system has a human operator the display may be presented in any number of visual and/or audible forms. The operator, upon recognizing the message of the display, decides on his next course of action. If the system is a part of an unmanned or robotic vehicle, the output may be in the form of a digital value or a flag, depending on whether the decision algorithms are contained within the vehicle or are part of the signal processing of the sensor system. 1.5.2

Integration and Packaging

There are always practical considerations when designing a system. Some of these relate to optimizing its performance, and some derive from its anticipated field service environment. In many instruments increased sensitivity to the target often corresponds to increased sensitivity to environmental effects such as temperature, humidity, electromagnetic interference, shock, and vibration. It sometimes requires heroic measures to mitigate these effects. Systems can become large or clumsy. The systems we are principally considering, portable, field-operable trace chemical sensing systems, have some common constraints. Portable systems must operate on a portable power source. Often this simply means batteries. When heating becomes a part of the operating sequence, batteries can be consumed quickly. A well-defined energy budget is necessary early in the design process. Portable systems are often introduced into small spaces in search of the quarry. They also tend to be handled rather roughly by their operators. Current trends in microminiaturization offer relief to the system designer in both these areas. 1.6 1.6.1

ISSUE OF CONCENTRATION Nomenclature

In this book we will often consider very small quantities of materials. In an effort to ensure understanding, Table 1.3 presents the notation used for mass.

ISSUE OF CONCENTRATION

TABLE 1.3 Abbreviation g mg μg ng pg fg ag

19

Mass Nomenclature Name gram milligram microgram nanogram picogram femtogram attogram

Scientific Notation 100 g 10−3 g 10−6 g 10−9 g 10−12 g 10−15 g 10−18 g

Decimal Notation 1g 0.001 g 0.000001 g 0.000000001 g 0.000000000001 g 0.000000000000001 g 0.000000000000000001 g

1.6.1.1 Concentration Most often these mass terms will be used to describe concentration of the chemical of interest in some medium, usually air, water, or soil. When we talk of concentration, we have in mind some sort of normalization, so instead of describing an absolute quantity we describe the quantity of the chemical of interest contained in a fixed quantity of the containing medium. When that medium is air or water, we most frequently use a specified volume; when it is soil, we use a specified mass. Thus we will use terms like ng/L (nanograms of chemical per liter of air or water) in air or water and ng/g (nanograms of chemical per gram of soil) in soil, since soil varies in density, depending on weather and compaction. It is quite common practice to express concentrations as a ratio, using the “parts per” method of describing the ratio. Thus, in soil the ratio is simple, 1 ng/g, or (10−9 g/g), is a ppb, or one part per billion5 . In water or air, the issue is a bit more complex since it is necessary to reconcile the units of mass per unit volume to units of mass per unit mass. We make this reconciliation by dividing by the density of the fluid. For water this is, for purposes of this book, taken as 10−3 grams per liter. Table 1.4 illustrates this reconciliation. It may be worthwhile to illustrate the actual meaning of these concentrations: 1 ppb concentration in soil means that there is one gram of chemical contained within one billion grams (1000 tonnes) of soil. The implication is that the chemical is distributed more or less homogeneously within the medium. In that case, how much soil is one billion grams? Using a nominal value of 1.4 for the specific gravity of soil, we find that this comprises about 650 m3 of soil, which would fill about a 100 large dump trucks. Similarly, 1 ppb in water implies that a gram of chemical is dissolved in the quantity of water contained in an Olympic-size pool, 50 m long by 10 m wide by 2 m deep. One ppt means the gram of chemical is dissolved in 1000 of these pools. Since the parts per nomenclature for air would be expected to be a volume/volume ratio, in order to reconcile these we must account for the molecular 5

Note the use of common U.S. practice, defining billion as 109 ; other terms follow the same pattern.

20

CHEMICAL SENSING

TABLE 1.4

Reconciling Concentration Nomenclature in Soil and Water

Parts per. . . Nomenclature

In Soil

Parts per . . .

1 part in:

Abbreviation

million billion trillion quadrillion

106 109 1012 1015

1 1 1 1

ppm = ppb = ppt = ppq =

Mass/Mass 1 1 1 1

μg/g ng/g pg/g fg/g

In Water Mass/Volume

Mass/Mass

1 1 1 1

1 1 1 1

mg/L μg/L ng/L pg/L

μg/g ng/g pg/g fg/g

TABLE 1.5 Molecular Weights and Vapor Molar Densities for Some Common Explosives and Associated Compounds Compound DNB DNT TNB TNT RDX PETN Tetryl

168 182 213 227 222 316 287

Molecular Weight

Mass of Single Molecule (g), m0

Molar Density (Vapor) (g/L)

[22, [23, [22, [24, [24, [24, [24,

2.8 × 10−22 3.0 × 10−22 3.6 × 10−22 3.8 × 10−22 3.7 × 10−22 5.3 × 10−22 4.8 × 10−22

6.87 7.44 8.71 9.28 9.08 12.92 11.74

pp. 3–43] p. 1180] p. 3–66] p. 125] p. 125] p. 125] p. 125]

weight of the explosive molecule. Thus we divide the mass per mole6 for the molecule of interest by the molar volume. The result may logically be termed the molar density. First, however, since, as we shall see in Chapter 4, the search for explosive molecules at 0◦ C is normally futile, we will adjust the molar volume to a temperature more often encountered, 25◦ C. The molar volume at 25◦ C and 76 cm Hg is 24.45 liters: mass ratio volume molecular weight = molar volume molecular weight (g) = 24.45 L

Molar density =

(1.1)

Table 1.5 lists the molecular weights and the calculated molar densities for some of the most common molecules of interest. 6 A mole contains 6 × 1023 molecules. It has mass in grams numerically equal to the molecular weight. The molar volume is the volume occupied by a mole of any gas, measured at standard temperature and pressure (0◦ C, 101.325 kPa = 76 cm Hg). It has a value of 22.414 L [22, pp. 1–19, 1–5, 1–14, 1–16].

21


TABLE 1.6

Correlation of Concentration Nomenclatures

Mass/Volume

TNT in Air

TNB in Air

DNT in Air

DNB in Air

1 1 1 1

108 108 108 108

115 115 115 115

135 135 135 135

146 146 146 146

mg/L μg/L ng/L pg/L

TABLE 1.7 Parts per . . . 1 1 1 1

ppm = ppb = ppt = ppq =

ppm ppb ppt ppq

ppm ppb ppt ppq

ppm ppb ppt ppq

ppm ppb ppt ppq

Reconciling Concentration Nomenclaturesa TNT in Air 9.3 9.3 9.3 9.3

μg/L ng/L pg/L fg/L

TNB in Air 8.7 8.7 8.7 8.7


DNT in Air 7.4 7.4 7.4 7.4


DNB in Air 6.9 6.9 6.9 6.9


a Since parts per must remain dimensionless, this table actually contains “volume/volume ratios” even though the concentrations are stated in mass/volume units. The correction is made using Eq. (1.2).

With the molar densities we can calculate the volume/volume ratios in air. Equation (1.2) [25, p. 10] yields the data tabulated in Table 1.6: ppt =

1000cA molar density

(1.2)

where cA is the concentration in ng/L. It is often convenient to make the correlations “in the other direction.” This is shown in Table 1.7, where the data are essentially the reciprocals of the correlations shown in Table 1.6, but can be calculated directly from Table 1.5 values. 1.6.1.2 Minimum Possible Concentrations Obviously, the absolute minimum possible concentration of any chemical within any medium occurs when there is only one molecule of the subject chemical in the sample. Knowledge of this concentration will provide a recognizable lower limit to possible sensing. We can find that concentration [25, p. 11] as follows: If one gram of soil contains a single molecule of the subject chemical of molecular mass m0 (Table 1.5), then the mass/mass concentration is numerically equal to m0 . For example, using TNT (trinitrotoluene) at molecular weight (MW) of 227,

m0 =

227 grams per mole = 3.78 × 10−22 g 6 × 1023 molecules per mole

(1.3)

22

CHEMICAL SENSING

yields the mass/mass concentration for one TNT molecule in one gram of soil as c0 = 3.8 × 10−22 g/g. We can express this in terms from Table 1.3 in several ways: c0 = 3.8 × 10−22 g/g = 3.8 × 10−10 × 10−12 g/g = 3.8 × 10−10 pg/g = 3.8 × 10−13 ng/g = 3.8 × 10−7 fg/g

(1.4)

By reference to Table 1.4 we can also use the parts per nomenclature, recognizing that 1 ng/g is 1 ppb, and so forth: c0 = 3.8 × 10−22 g/g = 3.8 × 10−16 ppm = 3.8 × 10−13 ppb = 3.8 × 10−10 ppt = 3.8 × 10−7 ppq

(1.5)

We find similar results in water if we consider one milliliter of water to contain one molecule. Since the 1 mL of water has a mass of 1 g, the mass/mass concentration is again c0 = 3.8 × 10−22 g/g, numerically the same as the mass/volume concentration, 3.8 × 10−22 g/mL. However, expressing this concentration in the units previously used, g/L, yields cL = 3.8 × 10−22 g/mL = 3.8 × 10−19 g/L = 3.8 × 10−7 pg/L

(1.6)

This concentration describes a concentration of one thousand molecules per liter. Alternatively, considering the case of one molecule in a liter of water, the concentration is then cL = 3.8 × 10−22 g/L. As before, this can also be expressed in several ways: cL = 3.8 × 10−22 g/L = 3.8 × 10−9 × 10−13 g/L = 3.8 × 10−13 ng/L = 3.8 × 10−10 pg/L = 3.8 × 10−7 fg/L

(1.7)

Or, from Table 1.4, using the mass/mass concentration, for the 1000 g of water in a liter: cL = 3.8 × 10−22 g/L = 3.8 × 10−25 g/g = 3.8 × 10−13 × 10−12 g/g = 3.8 × 10−13 pg/g = 3.8 × 10−13 ppt

(1.8)


23

Therefore, on the basis of one molecule of the target compound present in a standard sample, a gram for soil, or a liter for water or air, we can calculate the minimum possible concentration for that compound in that size sample of soil water or air. Table 1.8 presents those results. This provides an objective lower bound for the LODs. 1.6.1.3 Flux Rates We use the term flux to describe the movement of material. Flux rate is then the quantity of material moved por unit time. In the literature we find rates quoted in different time units, sometimes seconds, sometimes minutes, or hours, or days. In particular, flux rates are sometimes presented in μg/cm2 -day and sometimes in fg/cm2 -sec. We can simplify this a bit by normalizing these to cm2 . Thus, Table 1.9 provides a correlation of the rates. 1.6.2

Source to Sample

There are a wide variety of possible sources of explosive molecules. No matter what that source is, or where it is located, a similar process may be recognized. First, the molecules must become free from the bulk of explosive at the source. Then they must travel through the local environment to a location where they are accessible to the sampling unit. This journey is full of pitfalls that stop many of the released molecules, so the initial small flux rates may be reduced even more. Chapter 4 discusses the process in considerable detail. 1.6.3

Catch, Count, and Release Cycle

A useful strategy adapted in many sensing systems is to use what we might term “minibatch” processing. By this we mean that the sampling unit takes in material for a short period of time, delivering this “batch” of material to the concentrating unit. The batch may be quite small in order to enable more or less continual motion of the system in search. The batch is delivered to the concentrator where the quarry molecules are collected. Then, through either thermal or chemical cycling, the concentrated molecules are stripped off the concentrator and sent to the sensing unit. While this process is proceeding, another batch may be being collected by the sampling unit. Properly timed, such a strategy can increase overall response rate substantially. 1.6.4

Sensor Sensitivity Versus Sampling Time

Concentration is a most important issue. Each of the sensing technologies discussed in this book, and any other conceivable sensing technology, has some limit of detection, below which it will not provide reliable results. But for most sensors this sensitivity threshold, the lowest concentration of a given molecule that can be reliability detected, is not a static value. The threshold is also a function of the sampling time, in the most general sense. Heuristically, we consider that a certain number of molecules need to be ingested by the sampling unit to

24

Volume/volume, from Eq. (1.2).

a

−13

3.8 × 10 3.8 × 10−10 3.8 × 10−13 3.8 × 10−13 3.8 × 10−13 4.1 × 10−11

ng/g ppt ng/L ppt ng/L ppt

g soil g soil L water L water L air L aira

1 1 1 1 1 1

TNT

Units −13

3.6 × 10 3.6 × 10−10 3.6 × 10−13 3.6 × 10−13 3.6 × 10−13 4.1 × 10−11

TNB 3.0 × 10 3.0 × 10−10 3.0 × 10−13 3.0 × 10−13 3.0 × 10−13 4.0 × 10−11

−13

2.8 × 10 2.9 × 10−10 2.8 × 10−13 2.8 × 10−13 2.8 × 10−13 4.1 × 10−11

−13

3.7 × 10 3.7 × 10−10 3.7 × 10−13 3.7 × 10−13 3.7 × 10−13 4.1 × 10−11

−13

Concentration per Quantity of Medium DNT DNB RDX

Minimum Possible Concentration, Represented by One Molecule of Explosive

Medium

TABLE 1.8

5.3 × 10 5.3 × 10−10 5.3 × 10−13 5.3 × 10−13 5.3 × 10−13 4.1 × 10−11

−13

PETN

4.8 × 10−13 4.8 × 10−10 4.8 × 10−13 4.8 × 10−13 4.8 × 10−13 4.1 × 10−11

Tetryl

25

1 103 106 109 1.16 × 10−5 1.16 × 10−2 1.16 × 101 1.16 × 104

fg/sec

TABLE 1.9

10 10−3 1 103 1.16 × 10−11 1.16 × 10−8 1.16 × 10−5 1.16 × 10−2

−6

−3

10 1 103 106 1.16 × 10−8 1.16 × 10−5 1.16 × 10−2 1.16 × 10−1

ng/sec

pg/sec 10 10−6 10−3 1 1.16 × 10−14 1.16 × 10−11 1.16 × 10−8 1.16 × 10−5

−9

μg/sec

Correlation of Flux Rate Nomenclature (per cm2 )

4

8.64 × 10 8.64 × 107 8.64 × 1010 8.64 × 1013 1 103 106 109

fg/day 1

8.64 × 10 8.64 × 104 8.64 × 107 8.64 × 1010 10−3 1 103 106

pg/day

−2

8.64 × 10 8.64 × 101 8.64 × 104 8.64 × 107 10−6 10−3 1 103

ng/day

8.64 × 10−5 8.64 × 10−2 8.64 × 101 8.64 × 104 10−9 10−6 10−3 1

μg/day

CHEMICAL SENSING

Sensitivity

26

Sampling Time

Figure 1.2 Conceptual relationship between sensitivity and sampling time for a fixed probability of detection and fixed sampling rate.

provide the sensing unit with enough product to form an acceptable probability of detection, PD . A balance thus needs to be struck between the quantity sampled and the sensitivity of the sensing unit. For a given sampling rate and a given PD , the relationship is visualized as in Figure 1.2. In any real sensing operation the sensing system is either being moved to search different locations or different objects are being brought to the system for examination. In either case the system will operate in a sequential batch mode. That is, the system will cycle through the series of actions described in Table 1.1. It is necessary to process through the entire sequence of actions in order to extract a determination of each location or object. Sampling time is only one part of the sensing cycle. Detectors or sensing units must be purged of the molecules collected in the previous cycle before a new cycle can continue. Some systems may be configured with multiple, alternating, sensing units in order to reduce or eliminate the “dead time” in the sensing cycle. Operational application and economics often determine trade-off choices between sensitivity and sampling time. Since sensing units have historically tended to increase in cost and fragility with sensitivity, increasing sampling time may offer a desirable option for some system designs. 1.6.5

The Concentration Gap

As we shall see in Chapter 4 the anticipated concentration of explosive molecules in many search situation, such as for buried landmines, may be very low, perhaps 1 pg/L (or 100 ppq, or 1 in 1013 molecules). Most sensing systems are not capable of detecting such low concentrations directly. Hence there usually exists a gap between the available sensitivity of existing systems and our perceived needed sensitivity.


Projected IMS & AFP

Buried Munitions Vapor AFP

IMS

QitTof MS

SAW

27

Chemiresistors

Explosive Vapor Concentration 10 ag/L

1 part in: 1018

10 fg/L

10 pg/L

10 ng/L

10 mg/L

1015

1012

109

106

ppq

ppt

ppb

ppm

Figure 1.3 Illustration of concentration gap.

There are three potential ways to reduce this gap: (1) The basic sensitivity of the sensing unit could be increased or (2) the degree of concentration accomplished in the concentrating unit could be increased. While it is easy to make a statement like that, actually increasing the performance of either unit will require a great deal of innovative research and development. Alternatively, since explosive molecules tend to stick to solid surfaces such as structures, vegetation, or soil particles, (3) a sampling process that extracts them from those locations, rather than directly from the air, could exploit that natural concentration. See the discussion of this in Chapter 8. Figure 1.3 illustrates the current situation. Figure 1.3 diagrams the problem faced by system designers. If the sensitivity of the sensing unit is insufficient to detect the available concentration of explosive vapor there is a concentration gap. This requires the system to include a concentrator unit to bridge the gap. In the example shown, assuming a part per quadrillion target vapor source, those sensing systems that have lower sensitivities fall to the right of the source concentration and must take action to concentrate the sample in order to successfully detect the target. 1.6.6

Sensitivity Comparison

Many different technologies have been adapted to construct detection systems for finding explosives. The result is a variety of instruments being developed or marketed. The potential user can become bewildered by the variety. Several reviews have been published to categorize the technologies and the available instruments. The following information is extracted from four [26–29] of those reviews. This provides a representative, though by no means exhaustive, comparison of the limit of detection7 (LOD) of these technologies. Three of the four references describe the technologies and present a reported [26] or advertised [27, 28] 7

As earlier noted, we refer to the smallest quantity of any given compound that a technology can detect as its LOD. The concept of LOD for mass spectrometery is more rigorously defined in Chapter 11, Section 11.4.3. The concept is used here in a parallel way for the other technologies, without specifying measurement procedures.

28

2004 2004 2004 2003

Ion trap ion mobility spectrometer

Nonlinear dependence of ion mobility

Thermo-redox

Electrochemicalb

2004 2004

2004

Gas chromatography/mass spectrometer

Ion mobility spectrometer

2004

Gas chromatography/differential ion mobility spectrometer

Gas chromatography/surface acoustic wave

2004 2004

Gas chromatography/chemiluminesence

2004

Chemiluminesence detector a.k.a. thermal energy analyzer

Chemical reagent-based (color change)

2005 2005

2005

Ion mobility spectrometer

Quadrupole ion trap—time-of-flight mass spectrometer

2005

Amplifying fluorescent polymer

Lab-on-a-chip/high-performance liquid chromatography

Published

Electrochemical

TR

NLDIMS

ITIMS

IMS

GC/SAW

GC/MS

GC/DMS

GC/CL

Color

CL or TEA

QitTof MS

LOC/HPLC

IMS

AFP

Abbreviation

Published or Advertised Limits of Detection for Sensing Technologies

Technology

TABLE 1.10

1000

1000

1000

1000

215

LOD (ppt)

1000

1

1

0.001

1

1

1000

0.9

0.05

0.001

LOD (pg)

[29, p. 37]

[28, p. 42]

[28, p. 42]

[28, p. 42]

[28, p. 42]

[28, p. 42]

[28, p. 42]

[28, p. 42]

[28, p. 42]

[28, p. 42]

[28, p. 42; 23, p. 34]

Syage, 11.3.4

Collins, 13.4

Woodfin 10.2

Cumming, 9.3

Sourcea

29

0.25

1000 10000

100

10 10

LOD (ppt)

20 200

0.5

10 0.01

50

500

1 10000 1000000

LOD (pg) p. p. p. p. p. p. p. p. p. p. p. p.

37] 37] 37] 26] 26] 26] 26] 26] 26] 40] 66] 54]

[26, p. 60] [26, p. 73] [26, p. 73]

[26, p. 53]

[29, [29, [29, [27, [27, [27, [27, [27, [27, [27, [26, [26,

Sourcea

When a name and section number is quoted, the quoted datum is found in that section of this book. Otherwise, the datum is to be found on the given page in the noted reference work. b Presented as a general class, rather than for any specific system. c Text mentions only AFP.

a

ASGDI MS/MS ECNIMS IMS BL

1996 1995 1993 1989

Fluorescent Piezoelectric Spectroscopic FIS GC/ECD GC/IMS GC/MS IMS TR GC/SAW IRS API/TOF MS

2003 2003 2003 1999 1999 1999 1999 1999 1999 1998 1997 1996

Fluorescentb,c Piezoelectricb Spectroscopicb Field ion spectrometer Gas chromatography/electron capture detector Gas chromatography/ion mobility spectrometer Gas chromatography/mass spectrometer Ion mobility spectrometer Thermo-redox Gas chromatography/surface acoustic wave Infrared spectroscopy Atmospheric pressure ionization time-of-flight mass spectrometer Atmospheric sampling-glow discharge-ion trap-tandem mass spectrometer Electron-capture negative-ion mass spectrometry Ion mobility spectrometer Bioluminescence detector

Abbreviation

Published

Technology

TABLE 1.10 (continued )

30

CHEMICAL SENSING

ppt 0.00001 0.0001 AFP (2005)

0.01

0.001

0.1

1

10

100

1000

ppq

10000 100000 1000000 1E + 07

ppb

ppm

GC/MS (2004) API/TOF MS (1995) IMS (2005) ASGDI MS/MS (1996) QitT of MS (2005) Fluorescent (2003) GC/CL (2004) GC/DMS (2004) IMS (2004) ITIMS (2004) BL (1989) IRS (1997) ECNIMS (1995) IMS (1999) FIS (1999) GC/ECD (1999) IMS (1993) GC/IMS (1999) Color (2004) Electrochemical (2003) GC/MS (1999) LOC-HPLC (2005) CL (2004) GC/SAW (2004) NLDIMS (2004) Piezoelectric (2003) TR (1999) TR (2004) GC/SAW (1998) Spectroscopic (2003)

Figure 1.4

Normalized limits of detection reported or advertised.

LOD for each. MacDonald et al. [29], however, aggregate all trace detection technologies into four broad categories. The rated technologies are presented in Table 1.10. In many reports and advertisements the LOD is presented as a range of values. In each of those cases the most favorable value offered is shown in Table 1.10. However, some are presented in “parts per” nomenclature and some in mass nomenclature. In order to present all in a similar form Figure 1.4 presents them normalized. Normalizing required an assumption that may not be valid in every case. It was assumed that


31

Average Volume of Air Required in Sample to Detect Buried Land Mine (L) [Explosive Concentration ~ 0.9 pg/L] ppt 0.0001

0.001

0.01

0.1

1

10

100

1000

AFP (2005) GC/MS(2004) API/TOF MS (1995) IMS (2005) ASGDI MS/MS (1996) QitTof MS (2005) Fluorescent (2003) GC/CL (2004) GC/DMS (2004) IMS (2004) ITIMS (2004) BL (1989) IRS (1997) ECNIMS (1995) IMS (1999) FIS (1999) GC/ECD (1999) IMS (1993) GC/IMS (1999)

Figure 1.5 Expected volume of air sample required to detect landmine signature.

when a LOD was quoted as a mass, that that mass of explosive was contained in one liter of air. No attempt was made to normalize all to the same explosive molecule, although it is recognized that there may be as much as an order of magnitude between the LODs of two different molecules using the same detector. Therefore, Figure 1.4 should be used with caution as a relative indicator of LODs. Furthermore, the author apologizes for the omission of those technologies not included; there are constantly emerging technologies that have not been included. A further caution is offered. Since the LODs are based on published or advertised values, there is no way to know whether the one who published or advertised the value was being conservative or optimistic in the disclosure. Likely, there are some of each in the group. 1.6.6.1 Required Sampling Volume Based on this normalization, we can get some idea of the quantity of air that needs to be sampled in order to detect a given concentration of explosive. Of course, this can only be an average value estimate. For an example we can use that concentration predicted for the TNT vapor signature of a buried landmine as given in Chapter 4, Section 4.3.2, Table 4.5, cm ≈ 10−3 ng/L. To estimate the necessary volume of air containing TNT at concentration cm that must be sampled by a system with a given LOD to detect the TNT, we merely divide the LOD of that system by the concentration, cm , as in Eq. (1.9):

32

CHEMICAL SENSING

Vol (L) ≈

LOD LOD (ng) = ≈ 1000 LOD (L) cm (ng/L) 9.4 × 10−4

(1.9)

Sample volume estimates for the technologies reporting the better LODs listed in Table 1.10 are shown in Figure 1.5. Examination of Figures 1.4 and 1.5 indicates the significant progress made in lowering the LODs in the last 5 years or so. Only two technologies reported before 2000 are in the most sensitive group. If progress continues at this rate, within another 5 years trace chemical sensing technologies should be able to detect buried or hidden explosives much more reliably and quickly.

REFERENCES 1. Bruschini, C. Commercial Systems for the Direct Detection of Explosives (for Ordnance Disposal Tasks), ExploStudy, Final Report, Ecole Polytechnique Federale de Lausanne, Lausanne, Switzerland, Feb 17, 2001. 2. Sensor Technology Assessment for Ordnance and Explosive Waste Detection and Location, JPL D-11367, rev B, Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA, March 1995. 3. Hibbs, A. D., G. A. Barrall, P. V. Czipott, A. J. Drew, D. Gregory, D. K. Lathrop, Y. K. Lee, E. E. Magnuson, R. Matthews, D. C. Skvoretz, S. A. Vierkötter, and D. O. Walsh. Detection of TNT and RDX Landmines by Stand-off Nuclear Quadrupole Resonance, preprint given author by Andy Hibbs, Quantum Magnetics Inc., San Diego, CA, 1999. 4. Willis, C. M. Olfactory detection of human bladder cancer by dogs: Proof of principle study. Br. Med. J. 329 (7468), 712–714 (2004). 5. Fliermans, C. B. and G. Lopez-de-Victoria. Microbial Mine Detection System (MMMDS), A. C. Dubey, J. F. Harvey, and J. T. Broach, Eds. in Proceedings of the SPIE 12th Annual International Symposium on Aerospace/Defense Sensing, Simulation and Controls, Detection and Remediation Technologies for Mines and Minelike Targets III, April 13–17, 1998. 6. McLean, I. G., Ed. Mine Detection Dogs: Training, Operations and Odour Detection, Geneva International Centre for Humanitarian Demining (GICHD), Geneva, 2003. 7. Townsend, J. Pigs, a demining tool of the future? J. Mine Action 7.3, 43–46 (2003). 8. Phelan, J. M. and J. L. Barnett. Chemical Sensing Thresholds for Mine Detection Dogs, SAND2002-0692C, Sandia Laboratories Report, Albuquerque, NM, 2002. ˚ 9. McLean, I. G., H. Bach, R. Fjellanger, and C. Akerblom. Bringing the minefield to the detector: Updating the REST concept, Proceedings of EUDEM2-SCOT–2003, Vol. 1, 2003, pp. 156–161. 10. Bromenshenk, J. J., C. B. Henderson, R. A. Seccomb, S. D. Rice, R. T. Etter, S. F. A. Bender, P. J. Rodacy, J. A. Shaw, N. L. Seldomridge, L. H. Spangler, and J. J. Wilson. Can Honeybees Assist in Area Reduction and Landmine Detection?, J. Mine Action 7.3, 2003. http://maic.jmu.edu/journal7.3/bromoersheuk/bromenshenk.htm Visited 8/9/06.

REFERENCES

33

11. Bender, S. F. A., P. J. Rodacy, R. L. Schmitt, P. J. Hargis, Jr., M. S. Johnson, J. R. Klarkowski, G. I. Magee, and G. L. Bender. Tracking Honey Bees Using LIDAR (Light Detection and Ranging) Technology, SAND2003-0184, Sandia National Laboratories Report, Albuquerque, NM, 2003. 12. Tomberlin, J. K., M. Tertuliano, G. Rains and W. J. Lewis, Conditioned Microplitis croceipes Cresson (Hymenopteria: Braconidae) Detect and Respond to 2,4 DNT: Development of a Biological Sensor, J. Forensic Sci., Sept. 2005, Vol. 50, No. 5, 5 pages. Paper ID JPS2005014, Available online at: www.astm.org 13. Rains, G. C., S. L. Utley and W. J. Lewis, Behavioral Monitoring of Trained Insects for Chemical Detection, Biotechnol. Prog. 2006, 22, 2–8. 14. Rains, G. C., Tomberlin, J. K., M. D’Alessandro and W. J. Lewis, Limits of Volatile Chemical Detection of a Parasitoid Wasp, Microplitis croceipes, and an Electronic Nose: A Comparative Study, Transactions of the American Society of Agricultural Engineers, Vol. 47(6): 2145–2152, 2004. 15. Hollas, J. M., Ed. Modern Spectroscopy, 4th ed., Wiley, Chichester, West Sussex, 2004. 16. Drafts, B. Acoustic Wave Technology Sensors, Sensors Magazine Online, October 2000; visited 8/9/04; http://www.sensorsmag.com/sensors/articles/article Detail.jsp? id=327349. 17. Li, J. The Cyranose Chemical Vapor Analayzer, Sensors Magazine Online, August 2000; visited 8/9/06; http://www.sensorsmag.com/articles/0800/56/main.shtml. 18. Oxley, J. Detection of Illicit Chemicals and Explosives, Elseveir, in Press. 19. Garner, K. and S. Smith. Volatile Organic Compounds, the good, the bad and the analysis, SPME, University of the West of England, Bristol; updated 9/23/04; visited 9/19/05; http://www.chemsoc.org/exemplarchem/entries/2004/westengland smith/ExempWeb/methdev.htm. 20. Janusz, P. Solid Phase Microextraction: Theory and Practice. Wiley-VCH, New York, April, 1997. 21. Chambers, W. B., P. J. Rodacy, E. E. Jones, B. J. Gomez, and R. L. Woodfin. Chemical Sensing System for Classification of Minelike Objects by Explosive Detection, in A. C. Dubey, J. F. Harvey, J. T. Broach, Eds. Proceedings of the SPIE 12th Annual International Symposium on Aerospace/Defense Sensing, Simulation and Controls, Detection and Remediation Technologies for Mines and Minelike Targets III, April 13–17, 1998. 22. Handbook of Chemistry and Physics, 80th ed., Chemical Rubber Publishing Co., Boca Raton, FL, 2000. 23. Handbook of Chemistry and Physics, 37th ed., Chemical Rubber Publishing Co., Cleveland, 1955–1956. 24. Cooper, P. W. Explosives Engineering, Wiley-VCH, New York, 1996. 25. Phelan, J. M. and S. W. Webb. Chemical Sensing for Buried Landmines—Fundamental Processes Influencing Trace Chemical Detection. SAND2002-0909, Sandia National Laboratories, Albuquerque, NM, 2002. 26. Yinon, J. Forensic and Environmental Detection of Explosives. Wiley, New York 1999.

34

CHEMICAL SENSING

27. Rhykerd, C. L., D. W. Hannum, D. W. Murray, and J. E. Parmeter. Guide for the Selection of Commercial Explosives Detection Systems for Law Enforcement Applications. NIJ Guide 100-99, NCJ 178913, National Institute of Justice, Office of Science and Technology, Washington, DC, 1999. 28. Theisen, L., D. W. Hannum, D. W. Murray, and J. E. Parmeter. Survey of Commercially Available Explosives Detection Technologies and Equipment 2004, Document No. 208861, National Law Enforcement and Correction Technology Center, a Program of the National Institute of Justice, U.S. Department of Justice, Washington, DC, 2005. 29. MacDonald, J., J. R. Lockwood, J. McFee, T. Altshuler, T. Broach, L. Carin, R. Harmon, C. Rappaport, W. Scott, and R. Weaver. Alternatives for Landmine Detection, RAND Science and Technology Policy Institute, Santa Monica, 2003.

CHAPTER 2

WHAT TO DETECT? JIMMIE C. OXLEY University of Rhode Island

The process of detecting explosives is generally broken down into bulk and trace detection. Bulk detection looks for a mass with certain properties considered indicative of an explosive—high nitrogen and/or oxygen content, high bulk density. Naturally, there will be explosive compounds that do not match these target characteristics [e.g., triacetone triperoxide (TATP)]; and there will be innocuous compounds that do (e.g., sausage). In detecting the presence of an explosive compound at trace levels, the general approach has been to look for a specific chemical from a library of target compounds rather than for a general property. This means that the probability of a false alarm is significantly lower than for bulk detection, but that a positive detection is limited to the provided library and makes no immediate allowance for terrorist innovations. A positive detection may also be misleading. The example that comes to mind is the reported positive detection on the wreckage of TWA flight 800. In that case, it was explained that explosives had been present in the aircraft many days earlier for purposes of a training exercise, but that none were present upon takeoff. In theory, any chemical analysis scheme should be applicable to trace detection of explosives, but the realities of explosive detection require a degree of rapidity and robustness that limit the type of usable instrumentation. Perhaps the place to begin in a book about trace chemical sensing of explosives is to define trace and explosive. If trace is defined as a submilligram quantity, then it should be noted that there are no chemicals that are explosive at trace levels. To clarify this point we need to elaborate on what makes a chemical an explosive. An energetic material is defined as one that releases energy upon decomposition. This material could be an explosive, a propellant, a pyrotechnic, or a fruit cocktail. For an energetic material to be an explosive chemical or composition its must be capable of undergoing decomposition with extremely Trace Chemical Sensing of Explosives, Edited by Ronald L. Woodfin c 2007 John Wiley & Sons, Inc. Copyright

35

36

WHAT TO DETECT?

TABLE 2.1 Energy vs. Power

Burning petroleum Detonating dynamite

J/g

W/cm2

50 5

103 1010

rapid release of energy and gas. An energetic material that burns may overall release significantly more energy than one that detonates (see Table 2.1). It is not the amount of energy (joules) released but the rate of the energy release that is essential for explosive behavior. Many materials and devices “explode” (e.g., a hot water heater, pressure cooker). Only explosives can “detonate,” that is, form a special type of shock wave that travels through the material at supersonic speed. The problem, in terms of our definition of an explosive, is that the same chemical can be formulated or configured so that it detonates or it doesn’t. Because the rapid energy release is necessary to “support” the detonation front, much like a piston, configuration of the chemical can be such that this is not possible. Typically, the problem arises at edges (regions of density change). A high-density region will bounce the shock wave back much as water hitting the wall of a swimming pool. The bounced back waves (rarefaction waves) degrade the shock wave, so that at such edges the shock wave is slowed and an overall curvature to the wave develops. If the width of the explosive is narrow, the rarefaction waves may be sufficient to kill the shock wave. This leads to the concept of critical diameter, where the explosive sample is too small to support a detonation wave. Thus, 200 g of explosive in a cylindrical configuration is probably detonable; the same amount sprinkled across the table top is probably not. This is, of course, the principle explosive disruptors employ—breaking a large charge into small ones without initiating detonation. With the size requirement in mind, let us consider the types of materials that may be explosive. The U.S. Department of Transportation (DoT) provides some guidance, specifying characteristics that, if met, require a material to be screened for explosive performance (see Table 2.2). Materials that typically fall in the explosive category are those where oxygen is incorporated within the molecule in nitro (NO2 ) groups. The presence of oxygen

TABLE 2.2

DoT Requirements for Class 1 Screening1

If a compound contains certain chemical groups: C=C unsaturation; M–C or M–N; contiguous N or O atoms; N–O; N–X or O–X (where X = halogen, M = metal) AND if it has an oxygen balance (O B) greater than −200, where O B = [−1600(2C + (H/2)–O)]/molecular weight; OR if its exothermic energy >500 J/g and onset below 500◦ C, then it must undergo DoT series one testing.

37

WHAT TO DETECT?

ONO2

ONO2 ONO2

O2NO

O O2NO

O O2NO ONO2

Nitroglycerin

ONO2

X

ONO2

O2NO

Nitrocellulose

Pentyerthritol tetrnitrate (PETN)

Figure 2.1 Nitrate esters.

within the molecule allows for rapid self-oxidation: C–H + O → H2 O + CO or

CO2

(2.1)

The nitrogen primarily ends up as gaseous N2 . The amount of oxygen available for these self-oxidation reactions is calculated as the oxygen balance. As can be seen from the DoT requirements above, a material can be quite oxygen deficient [trinitrotoluene (TNT) has an oxygen balance (OB) of −80] and still be detonable. Nevertheless, we are comfortable using 2-ethylhexyl nitrate ester as a cetane additive in diesel fuels, whereas we would not consider using glycerin trinitrate ester (nitroglycerin) for that application. CH3 CH2 | CH3 CH2 CH2 CH2 CHCH2 ONO2

2-Ethylhexyl Nitrate Ester

CH2 ONO2 | CH–ONO2 | CH2 ONO2

(2.2)

Nitroglycerin

For reasons of safety, as well as performance, U.S. military explosives require a high level of stability and a low level of sensitivity. These requirements result in years of formulation and testing. Thus, most of the chemicals used by the military have been known for the last 100 years. Most military explosives can be classed by the attachment of the nitro group: nitrate esters (O–NO2 ), nitrarenes (C–NO2 ), or nitramines (N–NO2 ) (Figs. 2.1 to 2.3).1 Figure 2.4 shows a comparison of powders that are similar in appearance: sugar, salt, AN, HMX, RDX, TNT, PETN, TATP, and HMTD.2 1 In

structural formula the angle represents carbon with appropriate number of hydrogen. These are common abbreviations for the following: TNT 2,4,6-trinitrotoluene; AN ammonium nitrate; PETN pentaerythritol tetranitrate; HMX octahydro-1,3,5,7,-tetranitro-1,3,4,5-tetrazocine; RDX hexahydro-1,3,5-trinitro-s-triazine; HMTD hexamethylene triperoxide diamine; TATP triacetone triperoxide. 2

38

WHAT TO DETECT?

OH

CH3

O2N

NO2

O2N

H3C

NO2

NO2 N

O2N

NO2

NO2

Picric acid

NO2

2,4,6-Trinitrotoluene (TNT) Figure 2.2

NO2

Tetryl

Nitroarenes.

NO2 O2N

N

N

NO2

N N

N

O2 N

NO2

NO2

N N NO2

Figure 2.3

Nitramines.

Figure 2.4 Comparison of powders that are similar in appearance. Left to right: sugar, salt, AN, HMX, RDX, TNT, PETN, TATP, and HMTD.

Picric acid and tetryl, both yellow powders, are no longer used by the military, though “do-it-yourself” books outline the synthesis of picric acid for the wouldbe criminal/terrorist and tetryl is still found in old munitions. Most of the military explosives are white-colored powders (TNT is cream colored). Since all, but TNT, decompose upon or instead of melting, they require some sort of compounding in order to be shapeable. They can be blended into TNT in a variety of ratios to make the formulations listed in Table 2.3. They can also be formulated in wax or plasticizer. The use of plasticizer is preferred because less dilution of the explosive occurs. (In the world of performance, TNT, with detonation velocity of 6900 m/s is considered a dilutant of HMX, detonation velocity of 9100 m/s.) In the detonation process, temperatures reach 3000 to 4000 K so that residue, if any, consists of unreacted explosive and, possibly, carbon if the explosive is

39

WHAT TO DETECT?

TABLE 2.3

TNT-Based Explosive Formulations

Ingredient Added to TNT

Name

% TNT

Aluminum (Al) Ammonium nitrate (AN) RDX RDX + 1% wax Comp B(+18%Al) HMX PETN Barium nitrate

Tritonal Amatol Comp B2 Comp B Torpex-2 Octol Pentolite Baratol

20–80 20–60 40 39 12 25–30 50 24

greatly underoxidized. Presumably any unreacted explosive recovered is material that has been thrown out from the perimeter of the charge and not experienced the detonation front. The amount of unreacted explosive residue depends on the efficiency of the initiation. Studies focusing on the decomposition products of explosives are not studying detonation products; they are looking for products of thermal decomposition. These may form during long-term or high-temperature storage. These products can be produced in laboratory studies by heating the explosive to 200 to 400◦ C. For detection purposes, thermal decomposition products may be important, if they are sufficiently volatile. One thermal decomposition pathway accessible to most explosives is homolytic cleavage of the nitro group: O–NO2 → O• + • NO2 C–NO2 → C• + • NO2

(2.3)

N–NO2 → N• + • NO2 The energy required for this process depends on the exact linkage. In general, the C–NO2 linkage found in nitroarenes is quite strong. Other decomposition paths, which require lower energies, dominate. For nitrate esters, the homolytic cleavage of the nitro group is a fairly low energy pathway. Furthermore, the evolved nitro group can accelerate the decomposition of the nitrate ester; thus, nitrate esters such as nitrocellulose, are stored with stabilizers, which function to bind evolved NO2 . Regardless of whether evolution of NO2 is a dominant or minor thermal decomposition pathway, nitro-containing explosives can be detected by the use of chemiluminescence. In fact, chemiluminescence is the detection scheme used in the thermal energy analyzer (TEA), a detector used in conjunction with gas chromatographic analysis by forensic explosive laboratories. It is highly specific and sensitive to nitro-containing explosives. The difficulty is immediately obvious—what if the explosive does not contain a nitro group? There are four types of explosives that do not contain nitro groups: inorganic oxidizers (in combination with any fuel), peroxide explosives, primary explosives, and newly developed high-nitrogen explosives. Chapter 3 discusses both the inorganic oxidizers—principally nitrate or chlorate-based and peroxide explosives—TATP and HMTD. We have previously reported on the thermal stabilities

40

WHAT TO DETECT?

TABLE 2.4 Stable Sensitive Volatile

Definitions Does not readily decompose at elevated temperatures or over time Easily initiated to decomposition or detonation by flame, impact, friction, or spark Transitions from solid to gas at a relatively low temperature; it can be returned to its original state upon cooling

and decomposition pathways of these materials [1,2]. Detection of the inorganic oxidizers is challenging since they have extremely low volatility, though if the cation is ammonium or urea, there may be trace amounts of ammonia or urea present. TATP is readily detectable because of its high volatility—0.05 torr at room temperature [3], about 10,000 times the vapor pressure of TNT, which can be detected by most “sniffing” instrumentation, including canines. Unfortunately, there is a tendency to equate “volatility,” “stability,” and “sensitivity” (see Table 2.4). Pure TATP is thermally stable; it does not decompose any faster than PETN. However, TATP is highly volatile, readily subliming to the atmosphere at room temperature if its container is open. For the terrorist, TATP and HMTD offer easy sources of primary explosives. Consulting the “do-it-yourself” literature, it can be seen that there are two other commonly recommended primary explosives—lead azide Pb(N3 )2 and mercury fulminate Hg(ONC)2 , but these are difficult to prepare cleanly. The synthesis of diazodinitrophenol (DDNP) (Fig. 2.5), common in commercial detonators, is reported in such publications, but apparently is rarely attempted by clandestine chemists. Typically, the brisance of a primary is less than TNT, but the efficacy is the fact that a shock wave can result from a relatively mild insult. Synthesis of new chemical explosives continues. The trend has been toward compounds with high positive heats of formation, high nitrogen content, and/or large ring strain (Figs. 2.6 and 2.7). However, synthesis of these requires sophisticated chemical expertise and, generally, is not attractive to the clandestine chemist. For the next decade we can expect that the challenge will remain to detect nitrated (X–NO2 ) materials and composite materials. As detection limits

O

N N

O2N

NO2

Figure 2.5

Primary explosive DDNP.

REFERENCES

41

N

N

N N

N

N

N

Tetrazolidine

[1,2,4]Triazolidine

Figure 2.6 Nitrogen rings are basis of new explosive compounds. O 2N

N H2N

NO2

O

X N N

H2N

N

N

N

N

N

O

O

X = O, SOn (n=0,1,2), N=N, N(O)-N, N(NO2)CH2N(NO2)

N

N

O2N

O

O2N

NO2

N

N

N

N

N

N

HN

N

+ N

NO2

N

NH

N

NO2

O2N

NO2

NH2

N

H2N

N

NO2 N

NO2

N

N

N N

N

Figure 2.7 Examples of fused ring systems.

become lower and lower, the significance of minute levels compared to background or other environmental exposition will require careful consideration. REFERENCES 1. Oxley, J. C., J. L. Smith, and H. Chen. Decomposition of multi-peroxidic compound: Triacetone triperoxides (TATP). Propellants, Explosives Pyrotechnics 27, 209–216 (2002). 2. Oxley, J. C., J. L. Smith, H. Chen, and E. Cioffi. Decomposition of multi-peroxidic compounds: Part II: Hexamethylene triperoxide diamine (HMTD). Thermochem. Acta 388(1–2), 215–225 (2002). 3. Oxley, J. C., J. L. Smith, J. Moran, and K. Shinde. Determination of the vapor density of triacetone triperoxide (TATP) using a gas chromatography headspace technique. Propellants, Explosives, Protechnics 30(2), 127–130 (2005).

CHAPTER 3

DANGEROUS INNOVATIONS KIRK YEAGER

3.1

INTRODUCTION

The use of explosives by terrorists is not a new innovation. Examples of groups attempting to violently address political grievances range from the Gunpowder Plot to blow up the British Houses of Parliament in 1605 to the attack on the USS Cole in Aden Harbor in October 2000. Explosives have been throughout the last half-century, and remain to this day, the primary tool of terrorists. What is fairly new in this arena is the sophistication of the disparate terrorist groups throughout the world in terms of producing their own explosive formulations. These homemade explosives are often referred to as improvised explosives (IEs). Broader experimentation with IEs has been coupled with more documented cross communication between terrorist groups and the systematic training of a new generation of bomb makers. Put plainly, the bad guys are getting better at what they do. To address the threat they pose, it is imperative that those who fight this ever-evolving enemy also advance in sophistication. This chapter is designed to give a perspective on the improvised explosives being utilized by terrorist groups throughout the globe. It will cover the scientific theory behind manufacture of improvised explosives, the history of their use, and an overview of the most common materials encountered by law enforcement worldwide. Some details will be excluded out of necessity so as to not make this a “how-to” chapter to be co-opted by the wrong individuals. 3.2

THEORY OF IMPROVISED EXPLOSIVES

Typically, just the mention of “improvised explosives” leaves the reader with a picture of a bushy-bearded mad bomber. Stereotypes surrounding IEs have Trace Chemical Sensing of Explosives, Edited by Ronald L. Woodfin c 2007 John Wiley & Sons, Inc. Copyright

43

44


been put forth by a sensationalist press, unrealistic Hollywood exaggerations, and even cartoons for years. This has not given the field a great deal of credibility. If any area of explosives research requires meticulous scientific understanding, it is IEs. Both commercial and military explosives are designed to be safely stored, handled, and used. Dealing with these materials in many cases requires no significant insight into their properties or chemistry. IEs are usually much less stable and forgiving than their military and commercial counterparts. To work safely with these materials, a much greater insight into their properties is required. Often the terrorist/criminal’s lack of understanding of these characteristics creates a self-solving problem. Those forced to deal with these materials (especially in law enforcement) should never make the same error. The goal of this section is to provide greater insight into the science of IEs, and, hopefully, help those forced to deal with them. Before anything else can be said about IEs, some rudimentary chemistry is needed. From a “cookbook” perspective, all explosives (be they military, commercial, or improvised) require the same chemical building blocks, which consist of a fuel and an oxidizer. Some explosives have the fuel and oxidizer as part of the same molecule, such as trinitrotoluene (TNT), and some explosives are comprised of mixtures of separate fuels and oxidizers, such as ammonium nitrate–fuel oil (ANFO). The oxidizer employed by the vast majority of explosives tends to be the NO2 (nitro) group. It is so predominant as an explosive ingredient that the primary focus of detection methods traditionally has been to look for nitroderived properties. IEs tend to utilize a more diverse range of oxidizers. Table 3.1 gives a list of the numerous oxidizer possibilities. Many of the oxidizers should be familiar to those experienced in investigating bombings. Prime examples are NH4 NO3 (ammonium nitrate) and KClO3 (potassium chlorate), both of which are utilized to make bombs on a regular basis. Other materials may appear more exotic but have equal destructive potential. It should be noted that some oxidizer families such as chlorates, nitrites, and peroxides are very reactive. When used in IE formulations, these oxidizers can create very sensitive mixtures, which need to be approached with extreme caution.

TABLE 3.1

Potential Oxidizers for Improvised Explosives

Oxidizer Name Chlorates Perchlorates Hypochlorites Nitrates Chromates Dichromates Iodates Permanganates Peroxides

Formula −

ClO3 ClO4 − OCl− NO3 − CrO4 −2 Cr2 O7 −2 IO3 − MnO4 − O2 −2

Common Counter Ions +

+

+

+2

K , Li , Na , Ba K+ , NH4 + , Na+ , Ba+2 , Ca+2 Ca+2 K+ , NH4 + , Na+ , Ba+2 , Ca+2 , Sr+2 Ba+2 , Ca+2 , K+ , Pb+2 K+ K+ , NH4 + K+ Na+ , Ba+2 , H+

Examples KClO3 , Ba(ClO3 )2 NH4 ClO4 , KClO4 Ca(OCl)2 NH4 NO3 , Ba(NO3 )2 K2 CrO4 , PbCrO4 K2 Cr2 O7 NH4 IO3 KMnO4 Na2 O2 , H2 O2 , BaO2

HISTORY AND THE ANARCHIST LITERATURE

45

To make an IE the oxidizer must be combined with a suitable fuel. The majority of fuels available for use in IEs are hydrocarbons (materials containing mostly carbon and hydrogen). These materials are often referred to as organic compounds due to their prevalence in living organisms. Potential fuels can be broken down into three categories: hydrocarbons, energetic hydrocarbons, and “elementals.” The category of pure hydrocarbons has too many materials to list. Anything that is typically burned to produce heat or energy can be applied as a fuel for an IE mixture. Some high-profile examples would include petroleum products such as diesel fuel and kerosene, plant- and animal-based oils, sugars, glycerin, wax/paraffin, sawdust/wood pulp, Vaseline, shellac, and rosin. Energetic hydrocarbons are materials with carbon, hydrogen, and nitro groups all in the same molecule. Because they contain their own oxidizer premixed at the molecular level in the form of the nitro group, energetic hydrocarbons tend to make some of the more powerful IEs when mixed with a suitable oxidizer. Prime examples of energetic hydrocarbons are nitromethane (NM) and nitrobenzene (NB). Due to the incorporation of the nitro group, many of these energetic hydrocarbons can be made into explosives in their own right, without addition of an oxidizer. The liquid NM is such a material. The elemental fuels applicable to IE production are not as obvious as the hydrocarbons described above and simply have to be learned and memorized. Those most prevalent are powdered metals (aluminum, magnesium, titanium), carbon disulfide, phosphorus, sulfur, and antimony sulfide. One important side note is that, for the most part, these elemental fuels produce IE formulations that are very sensitive and often unstable. Mixtures incorporating any of them should be treated with extra caution.

3.3


One of the best ways of understanding the improvised explosives being exploited by current-day terrorist groups is to become a student of history. It is a fair generalization to state “If it is simple, it has already been done.” This holds true for the formulation of explosives. Since the early 1800s, and the development of better techniques for chemical purification, research into explosives has mushroomed. Countless concoctions have been prepared and examined, especially over the last 150 years. The best candidates prepared over the years have gone on to commercial and military applications. All those formulations that were unstable, untrustworthy, or easily substituted with better alternatives fell into obscurity. Their ghosts haunt us to this day as IEs. To fully appreciate this transition, an exploration of some of the explosives developed in the last century is useful. Many of the formulations presented in historical terms will be discussed in further detail in the subsequent sections dealing with current terrorist IE threats. In 1871 Herman Sprengel patented a series of mining explosives that consisted of an oxidizing substance mixed with a fuel. The novelty of this approach was

46


that the components could be stored separately as nonexplosives, mixed on site to produce an explosive, and then shot with a cap (detonator). Taking into account the educational levels of miners back in the late 1800s, the isolation of explosives was a very prudent safety measure. Sprengel utilized potassium chlorate (KClO3 ), strong nitric acid (HNO3 ), and liquid nitrogen dioxide (N2 O4 ) as his oxidizing agents. He chose nitrobenzene (NB), nitronapthalene, carbon disulfide, petroleum, and picric acid (PA) as fuels. A study of the oxidizers reveals the most historical significance. Over the years, numerous formulations were prepared based on HNO3 . Usually, fuming nitric acid (HNO3 > 90% in water) was utilized in their preparation. Some historically significant formulations were: • •

Oxanite: PA (58 parts)/HNO3 (42 parts) Hellhoffite: NB (28 parts)/HNO3 (72 parts)

These explosives could be quite effective. Hellhoffite had ∼70% the power of TNT and was more shock sensitive (easier to initiate). Although the theory was sound, the practice left a good deal to be desired. The main drawback of these mixtures was the nitric acid, which was very corrosive, and exceedingly unpleasant to work with due to its choking fumes and tendency to produce bad chemical burns. It also ate through the copper blasting caps of the time, sometimes creating a premature detonation of the fulminate in them. Then, as now, nothing is as unnerving as an explosive picking its own time to detonate. Numerous attempts were made to get around the nitric acid problems. At the time, dynamite was well understood, with one variety produced consisting of nitroglycerin absorbed on Kieselguhr (diatomaceous earth). Attempts were made to absorb Hellhoffite on Kieselguhr. Fumes were reduced, but the resulting fill still attacked wrapping paper, wood pulp, metal, and skin. Eventually, the idea of nitric-acid-based explosives was abandoned due to the existence of kinder and gentler alternatives. The most successful of Sprengel’s formulations, from the standpoint of “historical longevity,” would be those based on KClO3 . Chlorates have a rich history in the energetic materials field. Perhaps a commercial formulation called Rack-aRock is the most notable example. It consisted of 21 parts NB and 79 parts KClO3 . In 1885 a combination of 42,331 lb of dynamite and 240,399 lb of Rack-a-Rock was used to blast out the Hell Gate Channel in New York Harbor in one single blast. An original advertisement for this formulation is seen in Figure 3.1 [1]. The majority of Sprengel’s fuels were energetic hydrocarbons. Chlorate-based composite explosives tend to be very sensitive to friction and shock. When mixed with energetic fuels, the chlorate formulations are even more sensitive. To address this problem, historically two paths were taken. The chlorate was either mixed with a nonenergetic hydrocarbon or the mixture was phlegmatized (lubricated) if an energetic fuel was desired.


Figure 3.1

47

Rack a rock advertisement.

Friction is one means of initiating energetic materials. When mixing solids, or handling and transferring powders, friction can be difficult to avoid entirely. In general, friction can be minimized with the application of a lubricant. The lubricant helps particles, such as crystals of potassium chlorate, glide over each other with minimized heat production. This reduces the chances of accidental initiation. Initially, castor oil was utilized to phlegmatize chlorate mixtures. The advantage of using an organic oil, above its ready availability and low cost, was that it could also serve as a fuel component and a water repellant. A group of explosives known as Cheddites was based on this principle. •

Example: KClO3 (80%)/nitronapthalene (12%)/castor oil (8%)

Some experimentalists chose to maximize precautions by eliminating energetic fuels and phlegmatizing their mixes. A classic explosive group that took this approach was the Minélites. • •

Example 1: KClO3 (90%)/heavy petroleum oil (3%)/paraffin (7%) Example 2: KClO3 (90%)/Vaseline (3%)/paraffin (7%)

48


The initial example utilized petroleum oil to lubricate, while the later employed Vaseline. Both lubricants served a dual function as a fuel. In the end, the sensitivity of the chlorate mixtures could only be reduced to a certain extent. Eventually, safer and more reliable explosives pushed the chlorate variants out of vogue. It should be mentioned that whereas chlorates faded away from the explosives scene, they found a broad range of applications in the pyrotechnic industry. Numerous commercial flash powders consist of KClO3 and aluminum to this day. They are still exceedingly sensitive and are perhaps one of the biggest sources of accidents in the industry. Over time, numerous formulations started using potassium perchlorate (KClO4 ) as a substitute. Perchlorates tend to be much less sensitive to insult than chlorates. Before returning to the present day, one more group of explosives deserves some historical exploration. This group consists of organic peroxides. Although a tremendous number of organic peroxides exist, only a small number are detonable. A subsection of this group has the properties of primary explosives. Unlike the common nitrated explosives, which derive their explosive potential from the NO2 group, peroxides utilize the–O–O–group. This is a much more reactive group than nitrate and as such makes peroxides much less stable than the more common explosives. Due to their emerging popularity within the IE community, the materials that will be discussed include hexamethylene triperoxidediamine (HMTD) and triacetone triperoxide (TATP). Their significance will also be addressed later in the chapter. HMTD was first prepared by Baeyer and Villiger in 1900 [2]. It is the only organic peroxide that ever drew serious consideration as a viable explosive. In the early 1920s it was examined by U.S. Army ordnance as a potential military explosive [3]. HMTD’s ease of initiation, coupled with its capability to crush approximately 2.5 times the amount of sand as mercury fulminate led to its consideration as a primary explosive for detonators. Both HMTD and mercury fulminate [Hg(ONC)2 ] were placed in caps and pressed to 1000 psi (pounds per square inch) to compare their ability to initiate base charges. These caps were used to initiate 0.4-g portions of various high explosives. The amount of primary was varied to determine the minimum amount necessary to facilitate initiation. The results are summarized in Davis [4]. Important highlights are given in Table 3.2. It was obvious that HMTD was much more efficient as an initiator. Although there is some dispute about the sensitivity of HMTD, the general consensus is

TABLE 3.2

Comparisons of Initiating Powers of Primary Explosives Minimum Charge (g) of Primary with Reenforcing Cap to Initiate

High Explosive TNT PA Tetryl

Mercury Fulminate (g)

HMTD (g)

0.26 0.21 0.24

0.08 0.05 0.05


49

that it is more sensitive than mercury fulminate. Literature drophammer results for a 2-kg weight are as follows: HMTD (3 cm) [3] and Hg(ONC)2 (5 cm) [5]. Meyer also shows HMTD as being more sensitive [5], with an impact initiation of 0.6 Nm versus 1 to 2 Nm for Hg(ONC)2 . The power of HMTD, rated by lead block expansion, varies from 60% [3] to 116% [5] of TNT. Despite the good performance characteristics of HMTD, it had several significant faults. As mentioned earlier, the peroxide bond is very reactive. This made HMTD incompatible with most metals. It actively attacked aluminum, tin, zinc, brass, copper, iron, and lead. HMTD was also very unstable when stored, exhibiting tremendous weight loss over short periods of time. In the end, it was judged both too reactive and too thermally unstable for any practical usage. It fell into obscurity in the explosives community in the early 1950s. TATP is unique in that it is a material that never received any serious consideration for military or commercial applications. It was studied by numerous groups, but primarily for academic reasons. Minimal literature references exist on it. TATP was first reported by Wolfenstein in 1895 [6]. Since that time, numerous recipes have been developed for its preparation. One of the most useful studies of its properties as an explosive was conducted by Rohrlich and Sauermilch [7]. They determined that TATP had a TNT equivalency of approximately 88% based on lead block expansion. They prepared a firing train consisting of 0.05 g TATP (pressed at 250 kg/cm2 ) in contact with pentaerythrital tetranitrate (PETN) to produce reliable blasting caps. Other experiments determined that a 0.16-g portion of the peroxide (density = 1.35 g/cm3 ) could initiate TNT. TATP displayed less reactivity with metals than HMTD, but it was less thermally stable. Sublimation (the direct transition of a material from the solid to the gaseous state) was the main pathway for weight loss in TATP. Even at room temperature, notable weight loss could be witnessed within a day. The high sensitivity and thermal instability of TATP precluded its use in any practical applications. Although of limited academic interest, it never took hold in the explosives community. Having shown that both Sprengel explosives and organic peroxides have been seriously studied, and in numerous cases applied as commercial energetic materials, their current status needs to be addressed. As stressed in each case, these materials were too unstable, unfriendly, and unpredictable for continued use in a field populated with much better alternatives. Terrorists, however, often do not have ready access to the commercial and military alternatives in use today. They are not deterred by the factors that forced the Sprengel explosives and the peroxides into mainstream obscurity. An examination of the current “anarchist” literature enforces this point. The term anarchist literature is used to refer to the dozens of books that have historically been available through mail order companies such as Desert Publications and Paladin Press dealing with the production of explosives and bombs. Many of these sources are still copied today as primer material to train current Al Qaeda bomb makers. The same information has been disseminated over the Internet in numerous bomb-making websites.

50


To illustrate how improvised explosives are often related to historically rejected formulations; a sampling of the anarchist literature will be utilized. The small cross section examined will consist of Guerrilla’s Arsenal (GA) [8], Improvised Munitions Black Book Vol. 1 (BB1) [9], Ragnar’s Homemade Detonators (RHD) [10], and FMX The Revised Black Book (FMX) [11]. It should be stressed that this small cross section examined represents only a fraction of the readily available information accessible through the Internet. These examples are utilized solely to show the overlap of historical explosive formulations and current terrorist recipes. More examples of current-day usage of these improvised materials will be included in the next sections of this chapter. Sprengel explosives are very popular in this cross section of anarchist literature. The mixture of NB/HNO3 (Hellhoffite) is actually given by name in BB1 and FMX. Chlorate-based formulations also abound. Examples are provided in GA, BB1, and FMX. GA gives a recipe for a PIRA (Provisional Irish Republican Army)-based mixture consisting of sodium chlorate and NB called CO-OP. It then goes on to relate this mixture back to the potassium-chlorate-based variant Rack-a-Rock discussed earlier. The author goes to the extent of being helpful enough to notify the reader that if potassium chlorate is problematic to find, sodium chlorate can be acquired in the form of “Solidox” oxygen pellets for home welders. Luckily Solidox is no longer on the market. BB1 gives instructions on how to prepare a chlorate-based mixture that is quite prevalent in the anarchist literature. This formulation consists of KClO3 phlegmatized with Vaseline. Typically this particular formulation is given the nickname “Poor Man’s C-4.” It does not have the power of C-4 but does posses

Figure 3.2

Poor man’s C-4.

FERTILIZER-BASED IEs

51

a degree of moldability. A sample prepared by the author is shown in Figure 3.2. The relationship of this mixture with the second of the Minélite formulations provided earlier is obvious. Numerous terrorist groups still apply chlorate-based explosives to this day. Sodium chlorate is a principle ingredient found in many weed killers in Europe. As seen, the applicability of chlorates to IEs has been well documented. This has not been overlooked by bomb makers throughout Europe. One of the more popular mixtures involves combining the chlorate with icing sugar. This fuel will be encountered again in the next section. The recipes for organic peroxides also crop up throughout the anarchist press. All four references taken for the cross-section analysis have recipes for HMTD. Keeping with the established research, all four references indicate that HMTD serves as a good initiating explosive. GA goes as far as drawing a homemade detonator with a firing train composed of flash powder, HMTD, and PA. A preparation for picric acid based on aspirin as the crucial starting material is provided as well for the curious reader. Not to be outdone, Ragnar Benson includes pictures of the three readily available components (hydrogen peroxide, hexamine, and citric acid) needed to fabricate HMTD. With a fair bit of accuracy, he refers to HMTD as “the primer mixture they don’t want you to know about.” As a final note on the anarchist literature, it should be mentioned that there is a tremendous amount of information available to the would-be bomber. A good portion of it is, at best, dangerous to the bomb maker. Some of it is excellent. In either case, the anarchist literature should never be written off as information for “crackpots” by “crackpots.” The best example of this is FMX. A quick look through the contents by an individual experienced in the field of explosives shows a great deal of insight provided by the author. Upon looking at the bibliography provided, the source of this insight becomes apparent. Some of the books utilized to prepare FMX mirror texts used to prepare this chapter [4,5]. Numerous other references given would be found on the shelves of any researcher engaged in the serious study of explosives. The terrorist does have sources of well thought-out and documented work, and this should never be forgotten.

3.4 3.4.1

FERTILIZER-BASED IEs Ammonium Nitrate IEs

Terrorists have become exceedingly adept at utilizing any precursor native to their environment to produce IEs. For simplicity, bombings can be broken down into large scale [such as vehicle-borne IE devices (IEDs)] or small scale (such as package bombs). The mixtures utilized by terrorist groups to manufacture small-scale charges are far too numerous to cover in any detail. A sampling was provided above in the review of historical precedence for the anarchist literature. However, when terrorists wish to fashion a charge in the hundreds or thousands of pounds, their available precursors become fairly limited.

52


Historically, and to this day, terrorists tend to utilize fertilizer precursors in their explosive production for large-scale charges. The ready availability of these materials in large quantities, coupled with the low cost per pound, makes them attractive to many bomb makers. It needs to be stressed that bomb makers will utilize what they have at their ready disposal. Most of the ingredients they adapt to their nefarious purposes will have legitimate uses. Fertilizers are common materials found throughout the entire world. With minimal processing they have shown themselves to be effective weapons in the terrorist arsenal. The most common fertilizer that can be used to produce explosive formulations is ammonium nitrate (AN). In special circumstances, pure AN itself can be made into an explosive. It contains both a fuel (hydrogen atoms) and an oxidizer (the nitrate group) in the same molecule. Although AN is a very common fertilizer, it is important to remember that it is classified as an oxidizer. Like any other oxidizer, mixing a fuel with it will produce an energetic material. As mentioned earlier, AN brings a small amount of fuel with it in the form of hydrogen. Nevertheless, even if the available oxygen reacted with all the hydrogen present, the oxygen still would not be totally consumed. Adding a fuel to AN gives the excess oxygen something to react with to produce additional energy. The types of fuels that can be added to AN to produce explosive mixtures are limited only by the imagination of the bomber. ANFO, the most common explosive produced from ammonium nitrate, is made by mixing it with common diesel fuel. This mixture employs 94% (by weight) AN and 6% No. 2 diesel. The AN used to produce ANFO comes in a very specific 1 form. It is produced in small “beads” (called prills) which are approximately 16 inch in diameter. There are two types of AN prills manufactured. The prill type intended for production of explosive mixtures is referred to as explosive-grade (or “commercial”) prills. These prills are designed to act like miniature sponges. As such, they are relatively porous (6 to 10%) and will readily absorb liquids such as diesel fuel. AN prills produced to be utilized as fertilizer are referred to as FGAN (fertilizer-grade AN). These prills are much denser than explosivegrade prills and usually have less than 3% porosity. Due to their low porosity they will not absorb liquids efficiently. If 6% fuel oil is added to FGAN prills, a good portion of it will pool at the bottom of the mixing container. The same qualities that make ANFO an attractive explosive for commercial applications also draw terrorists to it. It is easy to prepare and can be made from cheap readily available ingredients. Because of this, ANFO has been utilized in terrorist attacks worldwide for many years. It has been one of the most common IEs utilized in large-scale bomb attacks. When producing an improvised explosive from AN, there are four basic options dictated by the nature of the starting materials. The AN used can either be prilled or powdered. The fuels used can be either solid or liquid. In theory, the use of molten AN or gaseous fuels is possible, but the odds are really not in favor of it. Thus the four obvious combinations for improvised formulations are: (1) prilled AN plus a liquid fuel, (2) prilled AN plus a solid fuel, (3) powdered

FERTILIZER-BASED IEs

53

AN plus a liquid fuel, and (4) powdered AN plus a solid fuel. Each combination has unique characteristics, which will be covered separately. Prilled AN will be the first option examined. Prills have certain physical characteristics that produce similar behaviors in any explosives based on them. Most importantly they are little beads separated by air gaps that make it difficult for a shock wave to travel from prill to prill. Anyone who has committed to a step without realizing that they were walking off a curb can understand how energy can be disrupted by empty space. To get any explosive based on prills initiated, a good deal of energy is needed to overcome this physical air gap barrier. This is why a booster is utilized in ANFO. The air gaps dissipate the “punch” that the cap is trying to deliver. Any improvised explosives based on AN prills and an absorbed liquid fuel will run into the same need for a strong initiation source. Detonation velocities of these mixes will also tend to be lower due to the energy extracted by the air pockets throughout the explosive. Again, the choice of liquid fuel is limited only by imagination. Energetic fuels such as nitrobenzene (NB) can provide additional energy to the mixture. Common hydrocarbons such as kerosene can also be used. In general, bombers will most likely stick with diesel as the fuel, as ANFO is the formulation that gets the most “press.” Life is not simple. As stated earlier, prills come in two forms. When using liquid fuels, it is best to utilize the porous (explosive) grade AN. Its ability to absorb liquids ensures good mixing. It is possible to utilize FGAN prills to make ANFO-like mixtures. FGAN will not absorb liquids effectively; so mixes made from it will not have the power of analogous formulations produced with explosive-grade prills. In either case, it should be realized that bombs can be made utilizing fertilizer-grade AN and liquid fuels. The author has made ton quantities of ANFO with the densest AN prills available (E2 prills). Even though a good amount of diesel pools at the bottom of the charge container in FGANbased ANFOs, these charges still give a respectable yield. Fertilizer-grade AN is far from benign. Probably the most rare combination is the mixture of AN prills with a solid fuel. To get a good reaction between a fuel and an oxidizer an intimate mixture is needed. This is why monomolecular explosives (with fuel and oxidizer on the same molecule) tend to have higher detonation velocities than composite explosives (where fuel and oxidizer are two separate components). This is also why grain dust explosions occur; yet, a pile of wheat is not easy to light on fire. 1 inch in diameter) will Mixing any solid fuel with prills of AN (which are 16 obviously not produce a very good mixture of reactive components. The next set of explosives that need to be examined are improvised mixtures based on powdered AN. Once the AN is pulverized (powdered), the air gaps that dictated many of the properties of the prilled mixtures vanish. Thus any mixture based on powdered AN will take on different characteristics than a prilled mix. This has been noted by groups ranging from PIRA to Al Qaeda.

54


There are far too many potential fuels that could be added to ground AN to make improvised explosives. It would be impossible to discuss them all. Instead, a few important general principles will be covered. First, ground AN can be mixed with liquid fuels. A common example prepared commercially is ground AN mixed with nitromethane. This produces the binary explosive known as KinePak. Diesel fuel can also be used, as it was with prills to produce ANFO. This brings up the last potential combination: powdered AN and a powdered fuel. Again, any fuel capable of reacting with the excess oxygen released by AN will produce an effective improvised explosive. For this reason, one of the common solid fuels historically encountered is icing sugar. PIRA has used icing sugar as its fuel of choice to mix with AN since 1991. The resulting IE mixture is referred to as ANIS (ammonium nitrate icing sugar). 3.4.2

Urea Nitrate

The second fertilizer that is often used as the basis for an explosive is urea. Pure urea is neither an explosive nor oxidizer. In fact, it is primarily a fuel. It can be mixed with AN to form an explosive mixture. In portions of the United States urea is routinely sold to de-ice sidewalks during the winter months. It comes in prill form, just like AN. When urea is mixed with nitric acid it chemically reacts, and one molecule of urea attaches to one molecule of nitric acid. This forms the explosive urea nitrate (UNi), with the chemical formula NH2 CONH2 –HNO3 . In this explosive, the urea serves as the fuel and the nitric acid the oxidizer. The literature lists UNi as being 90% as powerful as TNT [5]. Small-scale air-blast work displayed a far-field equivalency closer to 65% [12]. Literature detonation velocities listed range from 3.4 to 4.7 km/s (∼11,200 to 15,400 ft/s) [13]. The author has seen several different velocities in this region depending on the charge size and confinement. UNi possesses a low loading density (typically around 0.7 g/cm3 ). It is highly corrosive and readily attacks most metals. In terms of sensitivity, UNi is typically a tertiary explosives such as ANFO and requires a booster to be initiated. Urea nitrate secured its place on the U.S. landscape by being the explosive used in the World Trade Center (WTC) bombing in February of 1993. Although it was not commonly seen in the United States at the time, terrorist groups in other parts of the world had long since known about it. Both the Shining Path in Peru and numerous mideastern terrorists utilized UNi in their attacks well before the WTC incident. The use by the Shining Path was so predominant that sales of urea were outlawed in Peru in 1992 to curtail explosive production. The Palestinians have utilized UNi as a main charge explosive for many years. Along with TATP (to be discussed in the next section), it remains one of the main explosives in the Palestinian arsenal. A photograph taken by the author in the Gaza Strip shows a sample of Palestinian UNi (see Fig. 3.3). Urea in fertilizer form is seldom pure. It carries a good number of impurities that create a beige color in the final UNi.

PEROXIDE EXPLOSIVES

55

Figure 3.3 Palestinian urea nitrate.

Finally, urea nitrate can be taken one step further chemically. Through the process of dehydration (chemical removal of water), urea nitrate can be transformed into nitrourea (NH2 CONH–NO2 ). Typically UNi is added to concentrated sulfuric acid to facilitate this reaction. In theory, the loss of water should make nitrourea more powerful than UNi. There is not a great deal of experimental data available about this material. It was theorized that the group that bombed the World Trade Center may have planned to conduct this reaction. No conclusive evidence exists to prove this speculation. To date, nitrourea has not been seen in any terrorist bombs. 3.5

PEROXIDE EXPLOSIVES

The newest group of explosives to take hold in the arsenals of terrorist groups across the entire world has been from the peroxide family. Each year new groups add these materials to their bomb-making repertoire. The author has personally encountered them in the past couple of years in Morocco, Turkey, Israel, Jordan, and Uzbekistan. Much of the emergence of these materials can be traced back to the cross pollination of ideas from terrorists coming out of Al Qaeda training camps with smaller already established groups. As seen in the history section, the two main peroxides TATP and HMTD have been around for over a century. The information on their manufacture is far from new. Neither of these materials was encountered with any degree of regularity until about the mid-1990s. Of the two, TATP has been the predominant one adopted for reasons that will become clear during the HMTD discussion. Two forms of acetone peroxides are known to exist: a dimer called diacetone diperoxide (DADP) and a trimer called triacetone triperoxide (TATP). Figure 3.4 displays the chemical structures of each.

56


Figure 3.4 Chemical structures of acetone peroxide molecules.

Although both dimer and trimer are known to exist, the trimer is the more easily produced. Typically, the dimer is only seen as a by-product of the trimer synthesis. Unfortunately, over the past decade law enforcement agencies have encountered TATP more frequently both in the United States and abroad. Numerous questions regarding its synthesis and properties have surfaced in Internet news groups as well, and there is no sign of abatement in interest about the compound. TATP is prepared by the proper combination of three chemicals: hydrogen peroxide, acetone, and sulfuric acid. The exact preparation will not be detailed for obvious reasons. All the precursors needed to prepare TATP are readily available in most countries. Acetone is a very common solvent and industrial chemical. Sulfuric acid of approximately 35% concentration is utilized as battery acid. Aqueous hydrogen peroxide (H2 O2 ) can be procured from a number of sources. Most drug stores sell dilute solutions of H2 O2 ranging from 3 to 6%. Hairdressers, who bleach hair, can buy concentrated solutions of peroxide (up to 60 volume) to make their own dilutions. Recently, one of the trends that have cropped up in health circles is to “oxygenate” the body. To do this, it is recommended that baths be taken in dilute peroxide. To facilitate this, multiple Internet companies that will sell and ship 35% peroxide (commonly referred to as “food grade”) have been formed. All percentages of peroxide can be used to produce TATP. Higher concentrations obviously produce better yields. As mentioned earlier, TATP is currently becoming more predominant. The Internet has been accused of being the root of most modern evils. Luckily, it can also serve as a valuable tool for those following trends in improvised explosives. For example, it is possible to track how many times any given explosive is mentioned in forums such as newsgroups. Newsgroups are computer bulletin boards dedicated to a variety of subjects of interest where members can log in and exchange information. The field of explosives has two main newsgroups dedicated to it known as alt.engr.explosives and rec.pyrotechnics. The first group was designed for explosives engineers (such as blasters) and the second for hobby pyrotechnicians.

PEROXIDE EXPLOSIVES

57

700 600 RDX TATP

Number of Citations

500 400 300 200 100 0 1996

1997 YEAR

1998

Figure 3.5 Analysis of number of citations occurring in explosives newsgroups.

Using an Internet search engine called DejaNews it was possible to search any newsgroup for keywords. The search would show how many times a given keyword was mentioned. Figure 3.5 shows an analysis of the two explosiverelated newsgroup citations for TATP and RDX from 1996 to 1998. As mentioned earlier, TATP has been known for over 100 years. It was not until the mid-1990s that it began to emerge as a recurring terrorist and criminal threat. Figure 3.5 focuses on the time frame when this increase in incidents involving TATP was occurring. Before any conclusions are drawn, the reason for analyzing RDX must be explained. Each year more and more people get “hooked-up” to the Internet. Looking at the number of references for any one material, like TATP, in isolation could be misleading. It would be hard to tell if an increasing number of citations was simply caused by more people having computer access to newsgroups. RDX would be a very common explosive to be mentioned in such a venue. Using RDX as a control, the increased Internet access during any given time frame can be normalized. Between 1996 and 1998 there was a 323% increase in the number of references to RDX in the explosive-related newsgroups. This increase can be attributed to an increase in the number of computer users. During the same period of time, the references to TATP showed a 1372% increase. The increase in the interest in TATP can therefore be attributed to a real problem and not just the fact that more people have access to newsgroups. TATP was definitely an emerging threat at that time. It remains a strong threat to this day.

58


TATP is very sensitive to impurities. The primary destabilizing impurity present in the TATP synthesis is the residual sulfuric acid catalyst. TATP purity can be related to the extent of which the acid is removed or washed out of the final product. The TATP produced by the author is made in an exceedingly meticulous fashion and is as free of impurities as it can physically be. The common criminal or terrorist cannot be counted on to be as careful in his or her TATP preparations. To better understand the dangers of TATP, the author has over the years made and studied the sensitivity of TATP of various purities. A “crude” TATP would be one in which minimal attempt to remove the residual acid catalyst was made. Crude TATP might have a simple water wash applied to the product crystals. A “washed” TATP would be one in which some chemical step has been conducted with the intent of trying to neutralize the residual acid. Utilizing a bicarbonate solution to wash the product crystals would produce a washed product. “Pure” TATP would refer to the product of a process where multiple steps were utilized to clean the material. Pure TATP would have undergone both a neutralizing wash and a recrystallization of the washed material. Heat, impact, friction, and electrostatic discharge (ESD) can all initiate explosives. Impact sensitivity can be tested with a device called a drophammer. This device employs a 2.5-kg weight raised to varying heights and dropped on 35-mg samples of the material being tested. How high the weight has to be raised to initiate the sample 50% of the time is used to indicate sensitivity to impact. This value is referred to as H50 (in centimeters). To judge sensitivity to friction, a line of sample (for lack of better analogy, like a line of cocaine on a mirror) is first placed on a movable metal plate. A locked metal wheel is lowered onto the front of the line of powder. The wheel can have an adjustable pressure exerted on it with a “piston.” The plate is then impacted with a falling pendulum, forcing it to travel underneath the locked wheel. This creates a sliding frictional force on the material. Sensitivity to friction is measured by how much pressure is placed on the wheel to create initiations 50% of the time. This value is called P50 (in pounds per square inch). Finally, ESD sensitivity is measured by discharging a spark of known energy through a sample. The spark energy needed to initiate the material 50% of the time is calculated and referred to as E50 (in Joules). Table 3.3 displays the results from these sensitivity tests on the various forms of TATP. Data for PETN provides a baseline for all comparisons. Overall, the most pure form of TATP is just slightly more sensitive than dry PETN. Although the

TABLE 3.3

Sensitivity of TATP at Different Purities

Material Tested Crude TATP Washed TATP Pure TATP PETN (Reference)

Melting (◦ C)

Impact H50 (cm)

Friction P50 (psi)

ESD E50 (J)

73–79 Decomposes 85–90 94–95 141 Decomposes

6.5 ± 3.2 5.3 ± 1.0 10.0 ± 1.4 12.5 ± 0.2

∼80 − 100 21 ± 12 37 ± 3 645 ± 70

— — 0.16 ± 0.05 0.325

PEROXIDE EXPLOSIVES

59

sensitivity of pure TATP is very high (a quality that classifies it as a primary explosive), the freshly prepared TATP does not exhibit “hair-raising” sensitivity. Unfortunately, this one pristine state is the only one that is not unnerving, since the less pure materials were even more sensitive to insult. As mentioned before, it is doubtful that the “home-brew” versions of TATP will be as clean as the pure samples. A brief clarification needs to be made about ESD sensitivity. As ESD is the most difficult form of insult to guard against, great attention needs to be paid to materials that exhibit high sensitivity to it. ESD initiation sensitivity was measured for only the purest form of TATP. As in all other forms of insult, the pure TATP was only slightly more sensitive than PETN. A standard deviation was done for the measurements. Using simple statistical analysis, it can be shown that TATP will initiate approximately 1 to 2% of the time with energies as low as 0.06 J. Note: A good sized human body under the right conditions can generate up to 0.08 J. This is more than ample energy to initiate TATP a small percentage of the time. Although odds are better than at a roulette table, it still does not seem a wise gamble to take. TATP needs to be treated as if it is static sensitive. As the TATP decreased in purity, an increase in initiation sensitivity was noticed. For a point of reference, an increase in initiation sensitivity means that a material is easier to initiate. Simply skipping the recrystallization stage reduced the drophammer of washed TATP to half that of the pure sample. This simple difference also increased friction sensitivity substantially (by ∼43%). Generally speaking, TATP goes from being very sensitive when pure to being dangerously sensitive in a less purified form. The TATP variation most likely encountered in the field would be analogous to the crude TATP. The amount of water used to wash the explosive would change from maker to maker, but high levels of impurities should be anticipated in the final product. It has been mentioned to the author (who, having never seen crack cannot verify this) that certain forms of TATP resemble crack cocaine. Spot tests for drugs such as crack contain strong mineral acid solutions. These solutions, when brought into contact with TATP, produce impressive decompositions. Extreme care should be taken to use small amounts of any unknown material when testing for the presence of drugs. If any of the starting materials for TATP are present, the use of these spot tests is not recommended. Pictures of TATP prepared by the author are provided in Figure 3.6. TATP can have a wide variety of appearances that are effected by the way the product is washed and allowed to crystallize. In addition, its appearance will be altered by any additives placed in the material. TATP has been mixed with ammonium nitrate, glue, TNT, mineral oil, and numerous other materials by terrorist groups in an attempt to increase either its power or its handling safety. Figure 3.7 shows photos taken by the author of TATP in two different countries. Detection of TATP in the field has been problematic. Technologies have been developed that focus on the peroxide group. Of these tests, the most commonly available are field detection kits that rely on color changes brought on by chemical reactions. Mistral’s PDK (Peroxide Detection Kit) offers chemical screening

60


Figure 3.6

(Left) Washed TATP and (right) pure TATP.

Figure 3.7 (Left) TATP (Amman, Jordan) and (right) TATP (Ankara, Turkey).

for peroxides [14]. As in all colorimetric testing, the chance for false positives always exists. False positives also plague more advanced technologies such as ion mobility spectrometry (IMS) trace detection. This remains an area open for active study. While little serious research has been conducted on the properties of TATP, this is not the case for HMTD. As previously mentioned, tests conducted by U.S. Army ordnance illustrated that HMTD was a tremendously powerful initiating explosive, exhibiting between three and four times the strength of mercury fulminate. Unfortunately, HMTD was too thermally unstable and too chemically

PEROXIDE EXPLOSIVES

61

reactive to be utilized in any practical fashion. Due to this, no military or commercial application of HMTD ever occurred. The hazards associated with HMTD have not, however, dissuaded criminals, terrorists, or juvenile experimenters from manufacturing and using it. It has been recommended in numerous anarchist publications as the improvised explosive of choice for the preparation of homemade initiators. Over the past several years it has been encountered with increasing frequency by law enforcement officers worldwide. For this reason, as noted for TATP earlier, a basic knowledge of its properties is essential for those in the explosive community. Of all the improvised explosives the author has made, HMTD probably has the simplest preparation. It requires three basic ingredients: hexamethylene tetramine (hexamine), citric acid, and hydrogen peroxide. Hexamine is a common fuel used in the pyrotechnic industry and can be purchased from numerous chemical supply houses. It is also the principle ingredient used by many camping stoves and can be purchased in tablet form for this purpose. Citric acid is a common flavorant additive and can be purchased from many drugstores. Aqueous hydrogen peroxide availability was discussed earlier. HMTD has many of the same molecular constituents of other high explosives (such as nitrogen, oxygen, and carbon). It should be noted that, like TATP, there are no nitro groups associated with HMTD. The structure of HMTD is provided in Figure 3.8. Like TATP, HMTD is characterized as a primary explosive. It is highly sensitive to external inputs of energy. A summary of the sensitivities to small-scale testing exhibited by HMTD is provided in Table 3.4. Values for PETN, RDX, and TATP have been included to provide reference points. PETN was obtained from Reynolds Corp. It was Risi “Hi-Surface” Lot F-2961. RDX was Type B Class G 0206-00 from Holston.

Figure 3.8

TABLE 3.4

Chemical structure of HMTD.

Sensitivity of HMTD Compared to TATP and Reference Standards

Impact H 50 (cm) Friction P 50 (psi) ESD E 50 (J)

HMTD

TATP

PETN

RDX

18.1 ± 3.2 0 !!! 0.217 ± 0.027

10.0 ± 1.4 37.0 ± 3.0 0.16 ± 0.05

12.5 ± 0.2 645 ± 70 0.325 ± 0.032

22.5 ± 1.9 >1100 0.265 ± 0.064

62


HMTD displays some interesting characteristics. It was surprisingly fairly insensitive to impact compared to PETN and TATP. The impact sensitivity measured lay between the values for PETN and RDX. Even more surprising was the ESD sensitivity. HMTD is a very fine powder when isolated. It has an appearance almost like that of talc. Fine powders are well known for being prone to static charge buildup. When transferring from one container to another, the static buildup in the HMTD was obvious by the way the powder stuck to the walls of the vessels involved. HMTD was more static sensitive than either of the standards tested, but it was less static sensitive than TATP. This was unexpected, and at the moment cannot be explained. The values obtained seem to indicate that HMTD has only moderate static sensitivity. CAUTION: Even though the experiments showed that HMTD was not overly sensitive to static, when dealing with improvised primary explosives, it is always wise to take static precautions. Terrorists do not always make the cleanest products. The truly unnerving characteristic of HMTD was its sensitivity to friction. Earlier, the test apparatus used to determine friction sensitivity was briefly described. Usually, some form of pressure needs to be exerted on the wheel in order to create enough frictional force to initiate an explosive. In the case of HMTD, a sample could be placed on the plate and the wheel placed over it with no pressure exerted at all. The minute the plate was slid under the wheel, the HMTD detonated. CAUTION: HMTD is by far one of the most friction-sensitive improvised explosives currently being produced by criminals and terrorists. Only contact explosives such as Armstrong’s mixture (chlorate and red phosphorous) rival it in sensitivity. As mentioned earlier, HMTD is a very fine white powder. Fresh, it resembles talc or flour. As it ages HMTD begins to decompose and clumps together. Examples of both fresh and aged HMTD can be seen in Figure 3.9.

Figure 3.9 (Left) Fresh HMTD and (right) aged HMTD recovered from Ahmed Ressam (see 3.6.1).

THE NEXT WAVE

63

Many sources provide instruction on producing homemade detonators from HMTD. They recommend placing it in holders such a cartridge casings. Recommending that someone press an explosive as friction sensitive as HMTD into anything is doing quite a disservice to the reader. In addition, HMTD is very reactive with most metals. It will not sit well in something made with copper or iron. As cartridge casings are made of brass, there is an immediate incompatibility of HMTD with this type of housing. When found, homemade initiators of this type should be treated as potentially self-initiating. Like TATP, detection of HMTD is problematic. HMTD, being an amine, does give off a unique odor. The smell emitted by HMTD has been likened to that of fish. As it gets older, the odor gets more predominant and unpleasant. 3.6

THE NEXT WAVE

Terrorist groups have over the years shown the ability to adapt and learn through experimentation. PIRA was well known for many years of experimenting with and perfecting improvised explosives based on ammonium nitrate. The current group of Al-Qaeda-trained terrorists is carrying on that tradition. In addition, a broader interaction between what were previously isolated terrorist groups has led to an unprecedented cross pollination of ideas. As a result, today’s law enforcement agencies are presented with an increasingly complicated explosive arsenal that presents totally new safety considerations to address. 3.6.1

Improvised Detonators

One of the newer additions to the terrorist’s lexicon is the ability to produce reliable improvised initiators. Only a decade ago the blasting cap was one of the most difficult to procure items necessary to produce an IED. This hurdle has been almost totally eliminated. Not surprisingly this change was precipitated by the “discovery” of TATP and HMTD. Both the millennium bomber Ahmed Ressam and the shoe bomber Richard Reid provide examples of terrorists utilizing this new “technology.” Both TATP and HMTD have been applied by terrorist groups worldwide in the production of homemade detonators. TATP is, however, seen in more instances. Examples of improvised caps can be seen in Figure 3.10. The photograph gives an illustration of the standard paper detonators (in this case produced from yellow lined paper and Christmas tree bulbs) and also shows a variety produced from syringes. The detonators in the photograph were recovered in Amman, Jordan. Jordanian officials discovered in their investigation that the bomb-making factory also served as a training facility for those involved in the terrorist cell. The detonators recovered were filled with either TATP or HMTD and were manufactured by members of the cell learning how to produce such improvised initiators. The paper detonators matched those produced by known Al-Qaeda-trained terrorists. Similar detonators were seen in the Gaza Strip, Uzbekistan, and Turkey by the author.

64


Figure 3.10

3.6.2

Improvised paper and syringe detonators.

Peroxide Main Charges

Although highly sensitive, TATP has been utilized as the main charge in numerous attacks over the last 10 years. The Palestinians have experimented with TATP and created numerous formulations in an attempt to increase its handling safety. Many explosive production labs met with a violent end during the course of this experimentation. This has motivated more Palestinian bomb makers to return again to urea nitrate as a main charge, although TATP is still routinely encountered in Israel. In an example of how quickly technology travels from one terrorist group to another, the attacks in Casablanca, Morocco, in May of 2003 utilized a TATP formulation first seen a couple years earlier in Palestinian attacks in Israel. TATP has also become one of the main charges utilized by Communist terrorist groups in Turkey over the last couple of years. TATP charges ranging from 1 to 20 lb are not uncommon. The proliferation of TATP as a main charge explosive remains relatively new. It provides a sobering illustration of how far bomb builders are willing to go to produce an IED with limited available starting materials. The risks of utilizing a highly reactive primary explosive for a main charge have not deterred these groups. Unfortunately, law enforcement is increasingly encountering labs producing these materials and having to engineer ways to safely deal with them. Due to the current trends in TATP usage, the need for reliable field-use trace and

THE NEXT WAVE

65

bulk detection methods geared toward peroxides has become more of an issue throughout the globe. 3.6.3

Fringe Mixtures

The trend seen with TATP has been recently mirrored with other highly dangerous mixtures. As commercial explosives become more difficult to procure, by both legitimate and illicit means, terrorist groups determined to conduct bombing operations are forced to work with more dangerous and unpredictable formulations. A prime example of this can be seen in the major bombings carried out in Indonesia over the last two and a half years. On October 12, 2002, a series of bombs shattered the tranquility of the Indonesian island of Bali. Three devices in all were detonated over the period of a few minutes. The devices ranged in size from a couple kilograms to approximately a ton. Of particular note was the vehicle borne IED (VBIED), which was initiated outside of Sari’s nightclub. The explosive filler in this bomb would mark the first time terrorists applied a chlorate-based improvised explosive in a large vehicle bomb. Figure 3.11 shows the aftermath of the bombing. The ruins in the lower right-hand corner mark the location of Sari’s nightclub. As discussed in the history section, chlorates have been long utilized by terrorist groups for small bombing attacks. Sodium chlorate is a very common weed killer throughout many European countries. The mixture of sodium chlorate and nitrobenzene (CO-OP) by PIRA was discussed earlier. Most commonly, chlorate is mixed with sugar to make an explosive filler for pipe bombs. As mentioned earlier, chlorates are also a major ingredient in various pyrotechnic formulations. Typically potassium chlorate is utilized in this capacity due to the more hygroscopic nature of the sodium analog. Mixed with powdered aluminum and/or sulfur, potassium chlorate produces a mixture commonly referred

Figure 3.11

See color plates. Aerial view of Bali bombing scene.

66


to as flash powder. The low ionization energy of aluminum creates mixtures when combined with chlorate that are very static sensitive. Addition of the low melting sulfur creates mixtures with high friction sensitivity. Combining both sulfur and aluminum with chlorate creates horrendously sensitive formulations. Interviews of suspects associated with the Bali bombings indicated that the bomb makers produced approximately 1000 kg of a chlorate flash powder incorporating both aluminum powder and sulfur. This device was delivered via vehicle to the targeted nightclub. Precursor chemicals were easily purchased in Indonesia due to the geographical proximity to China, which remains one of the largest producers of pyrotechnics in the world. Before the Bali attack it would have been almost unthinkable for a group to produce a VBIED filled with a ton of flash powder. The hazards inherent with such a device would have dissuaded most bomb builders. However, current trends show that terrorists have become more willing to take risks to produce explosive charges with whatever chemicals are readily available regardless of the inherent dangers. Bomb making has historically never been an overly safe occupation. As more “exotic” formulations begin to wend their way into the terrorist’s arsenal, the job of dealing with explosives and recovered IEDs will become more complicated. It should be noted that the use of chlorate has not been limited to Bali. Two more attacks using chlorate-based improvised explosives occurred in Indonesia followed the Bali bombing. On August 5, 2003, the J.W. Marriott hotel was bombed in Jakarta. In an even bolder attack the Australian Embassy in Jakarta was bombed on September 9, 2004. This continued reliance on a highly sensitive main charge in Indonesia illustrates that terrorists will utilize whatever tools are at their disposal regardless of the risks. 3.6.4

On the Horizon

The production of large charges based on chlorates (or primary explosives such as TATP) has raised the threshold of danger exponentially. With this trend, law enforcement can no longer discount any mixture as being too dangerous for terrorists to attempt to utilize. Past detection techniques, both bulk and trace, have tended to focus on nitrate groups. At the time this approach was sound and logical, as most bombing attacks utilized traditional explosive charges such as dynamite or TNT. This scenario has been forever altered. Materials that in the past were never considered to pose a viable threat have been successfully utilized in numerous attacks. If explosive screening technology hopes to match the growing threats, it will have to expand into the realm of reactive chemicals, which were previously discounted as nonviable for illicit use. The list of precursors that can be utilized to produce improvised explosives is large. Luckily oxidizers efficacious for this purpose tend to fall into smaller groups or families. Thus the family of peroxides would constitute a logical next step for screening technologies to focus on. Chlorate, perchlorates, and permanganates would also need to be further addressed.

REFERENCES

67

Terrorists have shown an ability to learn, improvise, and adapt. What only 10 years ago would have been considered too improbable a threat to consider addressing has in many cases become almost routine. Technology and training must take their cue from this trend and attempt to counter the next wave of threats in a proactive fashion. The alternative is to allow the enemy to exploit a weakness that thoughtful anticipation might have been able to eliminate. REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14.

Photograph courtesy of Robert Hopler, Powderman Consulting Inc. Baeyer, A. and V. Villiger. Ber. Dtsch. Chem. Ges. 33, 2479 (1900). Taylor and Rickenbach, Army Ordnance 5, 463 (1924). Davis, T. L. The Chemistry of Powder and Explosives. Wiley, New York, 1956. Meyer, R. Explosives, 3rd ed., VCH, New York, 1987. Wolfenstein, R. Ber. Dtsch. Chem. Ges. 28, 2265 (1895). Rohrlich, M. and W. Sauermilch. Z. Gesamte Schiess Sprengstoffwesen 38, 97 (1943). Harber, D. Guerrilla’s Arsenal: Advanced Techniques for Making Explosives and Time Delay Bombs. Paladin Press, Boulder, CO, 1994. Improvised Munitions Black Book, Vol. 1. Desert Publications, El Dorado, AR, 1978. Benson, R. Ragnar’s Homemade Detonators: How to Make ‘Em, How to Salvage ‘Em, How to Detonate ‘Em. Paladin Press, Boulder, CO, 1993. Wallace, W. FMX the Revised Black Book, Paladin Press, Boulder, CO, 1995. Yeager, K. Data-Base of Range Evaluated Improvised Explosives (D-BREIE) Phase I: Ammonium Nitrate Based Explosives. EMRTC Report to TSWG, July 1999. Kage, S., Ed. Encyclopedia of Explosives and Related Items Vol. 10. U.S. Army ARRADCOM, Dover, NJ, 1983, p. U103. MISTRAL Detection Israel, www.mistralgroup.com.

CHAPTER 4

WHERE SHOULD WE LOOK FOR EXPLOSIVE MOLECULES? RONALD L. WOODFIN Sandia National Laboratories (retired)

J. M. Phelan did a major part of this research. Much of this chapter, through Section 4.1, is drawn primarily from his work [1]. This work is also reproduced in [1a]. Since he was not available to actually write this chapter, the author, who merely coordinated this work as part of a broader project, gratefully acknowledges his debt for these important contributions to him and his co-workers. S. W. Webb, J. L. Barnett, and W. B. Chambers of Sandia National Laboratories and T. F. Jenkins, D. C. Leggett, and their colleagues of the U.S. Army Engineer Research and Development Center—Cold Regions Research and Engineering Laboratory (CRRDC), also made major contributions in this area.

4.1

INTRODUCTION

4.1.1 Where Did the Molecules Come from and How Did They Get Here?

We most often employ a portable trace chemical sensor in an effort to find an object containing an explosive charge. Normally, when we find explosive molecules, the immediate question becomes: “Where did these originate?” In choosing this kind of device we make a tacit assumption that some molecules will be released from the explosive charge and find their way into the surrounding environment, where they may be located and identified, using the sensor. It becomes crucial then, when we employ a trace chemical sensor, to recognize the mechanisms that affect these molecules after release. Therefore, we need to understand the environment through which these molecules migrate, the transport mechanisms that produce this migration, and the final condition of the molecules as they come to the positions where a searcher may encounter them. As we began Trace Chemical Sensing of Explosives, Edited by Ronald L. Woodfin c 2007 John Wiley & Sons, Inc. Copyright

69

70

WHERE SHOULD WE LOOK FOR EXPLOSIVE MOLECULES?

to study this part of the problem at Sandia National Laboratories (SNL), we came to refer to this study as “environmental fate and transport,” or EF&T. Of course, the object that contains and releases the explosive can be almost anything. However, since we began this study from the perspective of searching for mines, initially, landmines buried in soil, much of this chapter will use that specific work to illustrate the approach. We have gained a level of understanding for this particular application. From there we can adapt the methods to understand the EF&T of molecules in other situations. We can think of numerous examples of other applications where we might need to locate the object that is releasing explosive molecules. This is a problem that is most easily worked in reverse. That is, we find it most convenient to establish a source of molecules in a particular environment and follow the migration of the molecules as they are released under changes in that environment. With this understanding we have a reasonable expectation that we can predict the location of the source in a similar environment when we locate some of the released molecules. 4.1.2

Objects Other Than Buried Landmines

These obviously include UXO,1 IEDs,2 and other ERW,3 but also may include similar objects underwater, buried in the seabed. There are also abandoned mining and construction sites that may need to be searched for old explosives. No matter which type object we may hypothesize as the source of molecules found, we need to follow similar reasoning to locate that source. In particular, the study of “plumes” of molecules released in water from a submerged source has been the focus of substantial research. The migration of the molecules from within or from the surface of a submerged, buried munition is similar to the case of munitions buried in the ground. Following that release there is little similarity. Almost no work has been done to date to examine the processes that affect the molecules released from an object buried in the seabed. This may form a fruitful area for research. 4.1.3

Questions That Beg for Answers

We find a series of questions that arise in each situation: • • •

How and why do the molecules we find with the sensor leave their source? What forces cause the molecules to move away from the source? How quickly does this process occur, and how long does it last?

1 Unexploded

ordnance, usually referring to items such as bombs or artillery shells that failed to explode when employed, but also referring to items containing explosives that were abandoned or misplaced. 2 Improvised explosive devices, usually one-of-a-kind items or adapted UXO. 3 The United Nations has adopted the term “explosive remnants of war” to describe a variety of items that includes mines and unexploded ordnance.

SOURCE OF THE MOLECULES

• • • •

71

Is it a transient process or does it reach a “steady state”? Do the molecules tend to move in predictable directions? Do the molecules tend to concentrate in predictable locations? Are there some times or conditions more favorable for detecting molecules?

This chapter will explore these questions in a general way. Because of the nature of the processes involved, much of the basic research used to develop experiments and models for the migration of molecules was originally done for other purposes. In particular, much comes from agricultural research and much from research into environmental pollution. More details may be found in the references listed at the end of the chapter. 4.2 4.2.1

SOURCE OF THE MOLECULES How the Molecules Diffuse or Leak from a Munition

Henceforth we will consider any object containing an explosive charge as a munition. These may take any of the forms noted above. For every munition we assume that the explosive charge is contained within some sort of case. The forms of these cases vary widely. Some are as simple as a rubber or plastic bag inside a wooden box. Some are specially designed plastic or metal cases. Some are metal shells sealed with rubber o-rings or gaskets. All have two things in common: They have some explosive molecules adhering to their external surface and they have paths by which molecules can escape from the interior, either through leakage or diffusion. (It may be that some sea mines are so well sealed that no paths exist. This has not been well established.) 4.2.1.1 Surface Contamination During the process of filling a munition with explosive, some molecules will adhere to the external surface. Since these molecules are inherently “sticky,” they tend to remain there. In addition, when munitions are stored in magazines, or handled in work areas where explosives are regularly exposed to the air, the effect is for all surfaces to have explosive molecules adhering to their surfaces. (See Fig. 4.1.) (This is also true of the people who work in these areas.) Munitions may also receive surface contamination from being fired from an artillery piece or a rocket launcher, or from becoming damaged in the act of being “placed.” The term placed is used to designate any method involved in delivering the munition to where it is intended to function. This includes direct burial or hiding, firing of guns or rockets, dropping of bombs, and the like. A munition loses this surface contamination within days or weeks after being placed. This time depends on many factors, but it is clear that there may be an initially higher level of molecules entering the local environment from a recently placed, or “fresh” munition. This is a transient effect that normally gives way to a more steady-state supply of molecules. In munitions that have painted surfaces,

72


Figure 4.1 mine.

TABLE 4.1 Mine Type TM62-P TM62-P TM62-P TM62-P TMM-1 TMM-1 TMM-1 TMM-1

Sampling for surface contamination on a Soviet-type TM-62P2 anti-tank

Surface Contamination on Mines Chemical

Case

Concentration (ng/cm2 )

TNT TNT DNT DNT TNT DNT TNT DNT

Bakelte Polyethylene Bakelte Polyethylene Metal Metal Metal Metal Average

6 3 28 5 10 10 62 20 18

Source [3] [3] [3] [3] [4] [4] [4] [4]

this transient effect is to be expected. Paint samples scraped from unused U.S. mortar and artillery rounds contained from 1 to 45 μg/g of trinitrotoluene (TNT) and dinitrotoluene (DNT) [2]. Researchers at Sandia [3] and at the Cold Regions Research and Engineering Laboratory (CRRDC) [1] measured surface contamination on mine cases. Some of the results are highlighted in Table 4.1, based on data summarized by Phelan and Webb [1, p. 20].


73

4.2.1.2 Diffusion through Cases A principal mechanism for this steadystate release is direct diffusion of the explosive molecules through the munition case. Clearly, the rate of this diffusion is dependent on the case material, through a property called the diffusivity. Metal cases offer essentially zero diffusion, but cases made of some polymers or natural rubber have high enough diffusivities to enable substantial diffusion rates. The diffusivity of cases varies widely. Phelan and Webb [1] quote Mark and Kroschwitz [5] to illustrate that a natural rubber case may have a diffusion rate of explosive molecules about 9 orders of magnitude greater than a polyvinylchloride (PVC) case under the same conditions. However, there seem to be many more munitions with PVC than natural rubber cases. Most nonmetallic munition cases fall in between these extremes. There are some general statements possible about diffusion rates. Diffusivity or permeability of the case material increases with temperature. A 5◦ C rise in temperature can result in a 30 to 50% rise in permeability [1 p. 18]. This immediately implies that we would expect more molecules available for detection in warmer environments, if all other factors were equal. 4.2.1.3 Leakage around Seals When munitions, such as those illustrated in Figure 4.2, are fabricated in metal cases, there is essentially no diffusion through the case. However, since a munition needs a fuze to actuate it, and since it must somehow be filled with its explosive charge, there must be penetrations in the metal case. These may be closed with threaded plugs that often contain o-rings or gaskets. While the leakage is usually small around these plugs, direct samplings indicate a higher level of explosive material around them. This is observed for both fresh and old munitions. While the leak rate may be too small to produce a supply of molecules sufficient for detection, once the munition is found, it is sometimes possible to separate explosive bearing munitions from nonexplosive bearing munitions by sampling around these penetrations [6]. However, considering the consequences of misidentifying a loaded munition as inert, this technique is considered as insufficiently robust for that purpose [7]. Two groups of UXO are pictured in Figure 4.2. The upper group has been recovered from a burial site; the lower group has mixed duds and inert practice rounds from an artillery training impact area. 4.2.2

Example of Landmines

When we began this line of research at Sandia National Laboratories, we were primarily concerned with landmines. One of our first efforts was to characterize the surface contamination on some typical mines. For this work we obtained some Soviet mines from the U.S. Marine Corps. Leggett and colleagues at the U.S. Army CRRDC also measured surface contamination. In addition, they measured flux of molecules from mines in both air and water. Mines form one of the larger groups of explosive bearing targets for trace chemical sensors. Therefore, these measurements provide some insight into the concentrations at the source

74


Figure 4.2

Typical UXO items.

where the migration of the explosive molecules through their environment begins. Hopefully, they eventually arrive at a location accessible to sampling, whether the actual source or target is a mine or something else. Surface contamination was found to vary widely from mine to mine, but effective average contamination from a number of mines of several national origins seems to be about 15 to 18 ng/cm2 . Contamination in the painted surface of mortar and artillery shells was similar [8]. Flux measurements by Leggett and colleagues are briefly summarized in Tables 4.2 and 4.3 [9]. Two features are immediately apparent: (1) The flux of TNT is smaller that that of dinitrobenzene (DNB), while the greatest flux is always for DNT, and (2) the flux from any is greater when the mine is submerged in water, compared to the same mine in air. They also determined that,


TABLE 4.2 Mine TM5 PPM2 TMM1 PMA1A PMA2 VS-50

75

Landmine Flux into Air at 20◦ C (ng/mine/day) Construction

TNT

DNT

DNB

Polystyrene Unknown Metal Polyvinylchloride Polystyrene Unknown

1,380 128 740 207 24

15,100 12,800 1,720 1,550 282

4,500 3,480 282 358 332

RDX

14

From [9] Leggett, et al., 2001.

TABLE 4.3 Mine Type PMA-2 PPM2 VS-50

Comparison of Mine Flux into Air and Water at 22◦ C (ng/mine/day) Medium

TNT

DNT

DNB

Air Water Air Water Air Water

21 1,270 2,040 4,640

240 720 2,110 6,690

240 1,300 460 1,000

RDX

8 1,460

From [9] Leggett, et al., 2001.

at least for these mines, the flux rate follows an exponential temperature dependence, where the exponent, surprisingly, is not dependent on the case material or mine type [1, p. 23]: φ ∝ e0.11T where φ is the flux and T the temperature difference (in ◦ C). Summary of Landmine Flux Results Since no one has devised a method of directly measuring the flux of explosive molecules from a mine, whether in situ or in the laboratory, several laboratory measurements have been reported in which the mine was placed in a sealed container, surrounded by soil, water, or air. The concentrations of explosive molecules in the surrounding media were then measured at intervals of several days and the flux inferred from the total concentration divided by the elapsed time. This likely provides the best estimate that can be expected. The various measurements have substantial variation, depending on the techniques and media used. Phelan and Webb describe several experiments [1, pp. 23, 24]. It appears that a reasonable expectation of flux of explosive compounds from a buried landmine that move into the surrounding soil will be in the range of 1 to 200 μg/day. There are some complications, of course, since the surrounding soil produces a level of resistance, or “back pressure,” to the flux of the molecules. While the mechanisms are complex, the net effect is that wet soil permits a lower diffusive flux than dry

76


soil [1, pp. 24, 25], primarily because biological activity is greater in wet soil. These, and other environmental effects, begin to act on the molecules as they leave the source. Transient and Steady-State Conditions From the landmine studies we readily conclude that the “source term” for these molecules has an initial “spike,” or increased rate, in the days or weeks after the mine is placed. This rate then decreases to some more or less constant level and may remain at that level for years. The initial spike comes from surface contamination, while the long-term rate is primarily from diffusion through the case and seals or leakage through imperfections or damage. The rates are clearly subject to environmental factors, principally temperature and soil wetness. Nevertheless, it seems clear that, at least in the case of landmines, there is a continuing flux of molecules that provide a potential for detection. 4.2.3

Other Munitions

Most of the research accomplished to date on source flux rates has been directed toward landmines. This is generally because global publicity has focused attention, and hence funding, on them. Other ERWs are acknowledged, but similar measurements are more limited. We can recognize that many of the same processes will be operative. However, since landmines are intended to be placed, most often, by hand, they can be manufactured effectively from various polymer materials, rather than steel. Munitions that are designed to be dynamically placed, such as bombs, mortar rounds, or artillery shells, must naturally have stronger, usually steel, cases. These munitions, when they appear either as duds or as abandoned UXO, will have no significant diffusion, except through seals or o-rings. This diffusion should be expected to be considerably less than that available from mines. Duds often have case damage that can provide leak paths for the molecules. Naturally, this is not quantifiable in any predictive way. The other munition type of major interest, IEDs, are usually assembled a short while before they are placed, hence any molecules available for detection must be expected to come primarily from surface contamination and leaks, rather than from diffusive processes. Fortunately, for the operator of the sensing system, IEDs, by their nature, are seldom assembled with the care or precision of a manufactured device. Hence, we may perhaps harbor a reasonable hope that the source produces a flux of molecules adequate to permit detection. 4.3

TRANSPORT OF THE MOLECULES

Once a source of molecules exists, that source becomes the beginning of a pathway that leads, hopefully, to the sampling portion of a sensor system. That pathway may be tortuous; it may be broad. There are a number of effects that alter the rate at which molecules become available to a detector. That rate, expressed as a concentration level per unit time within the medium where the detection


77

is being attempted, largely determines the success or failure of a given system in locating the source. The medium of search may be air, water, soil particles, or perhaps even plant or animal4 material, depending upon system design and the suspected location of the source. In this section we will examine some of the main processes that affect the concentration of molecules available for detection. 4.3.1

Buried Sources

4.3.1.1 Competing Effects When we began studying these effects at Sandia in the early 1990s we used a diagram similar to Figure 4.3. This was an aid to isolating the various effects so that they could be studied more effectively. Because we were concentrating our study on landmines, we diagrammed the process for a buried mine. Of course, the basic processes would be the same for any buried munition; but the source flux rates would vary with the different munition construction. Figure 4.3 illustrates some of the conditions and processes that determine the concentration of molecules available to the sensor intake. This diagram was developed by Phelan and uses terms from environmental and agricultural science. Understanding of each of these processes is needed to make accurate estimates of the expected concentration at any given time and location. Phelan and Webb discuss these processes in some detail [1, p. 54ff]. We will consider them briefly in the following sections. Figure 4.4 is a similar diagram that would apply to a mine or other munition buried in the bed of the sea, a lake, or a stream. In Figure 4.4 the munition is shown partially buried, both to illustrate the different possibilities and to remind us that, while partially buried objects on dry land are usually found visually, submerged objects demand different search techniques. The exploitation of plumes will be discussed in Section 4.3.4 and in Chapter 5. 4.3.1.2 Partitioning of Molecules in Ground Partitioning is used to describe the sorting of the molecules according to their relative absorption or solubility. The three double-headed horizontal arrows in Figure 4.3 indicate the partitioning opportunities available to the molecules. These processes are reversible; the molecules may reside in one state for a while, then move to the other, later to return to the former state. The vertical, single-headed arrow associated with the mine represents the irreversible efflux of molecules from the source. Some molecules will be adsorbed on the surface of soil particles, some will be in solution in the water present in the spaces between the soil particles, and some will exist in the vapor phase, as free molecules, neither adsorbed nor in solution. Partitioning is a dynamic process, changing as soil conditions change. The principle dynamic parameters at a given location are moisture content and temperature. 4 Researchers

have intercepted bees returning to a hive and pureed them for sampling in an attempt to find molecules collected by the insects during their foraging.

78


Figure 4.3

Processes affecting molecules released from a landmine.

Of course, in a different location, with a different soil type the partitioning will form differently. The single-headed horizontal arrow at the lower right indicates the irreversible processes of plant uptake and biological decomposition of the molecules. These processes will be discussed in later sections. Partitioning takes on a more limited role in Figure 4.4, where the vapor state is assumed negligible. There is still some partitioning between the liquid and solid conditions, but even this is neglected when the leak is directly into the water above the seabed, as from the left end of the seamine. Since this condition has not been studied in detail, when someone does study it we may find that Figure 4.4 needs substantial revision. Such a study is certainly a fertile field for research. The processes diagrammed are constantly working, always seeking an equilibrium condition, but as local environmental factors change, so does that equilibrium. Hence the molecules released from the source are subject to continual change of their situation. In the following paragraphs we briefly discuss the processes at work and how they determine the availability of molecules for sampling. It will become apparent that, much like the search for the “mother lode”


79

Figure 4.4 See color plates. Processes affecting molecules released from a sea mine, partially embedded in the seafloor.

of a gold mine, the easiest place to sample is not always the richest source of molecules. Sorbing, Solution, and Vapor In the case of buried sources, such as landmines, the dominant processes are the sorbing of molecules onto the surface of solid particles and the dissolving of molecules into solution in the moisture available in the soil. Normal soil consists of particles packed together with spaces, often called pores, between them. The relative size of these pores varies with the compaction of the soil. In most soils these pores between particles are filled with a mixture of air and moisture, called vapor and liquid in Figure 4.3. The pores form potential paths for migration of molecules. The processes controlling migration through these paths are discussed in the following sections of this chapter. Fortunately, for anyone searching for these molecules as an indicator of the presence of the source, these processes drive most of them upward, toward the surface. Explosive molecules tend to be rather “sticky,” that is, they readily sorb, or adhere, to available surfaces. They are also soluble in water. When water is present in a pore adjacent to a particle, some molecules entering that pore will sorb to the particle and some will enter solution in the water. Since the pore is not entirely filled with water, but also contains some air, some molecules will also be found in the vapor state, simply as free molecules, neither sorbed to a

80


Mass Fraction, Solid 95%

Mass Fraction, Vapor 0.00001%

Figure 4.5

Mass Fraction, Liquid 5%

Explosive molecule distribution in soil at 50% pore saturation, 23◦ C, Kd ∼ 4.

soil particle nor in solution in the soil moisture. Figure 4.5 illustrates5 a typical equilibrium condition in a soil with 50% pore saturation, that is, for a soil that has half its maximum possible moisture content. Clearly, the great majority of the molecules is sorbed to soil particles, and very few are in the vapor state. Here is a point of major importance for anyone designing a system for finding these molecules. When any chemical is distributed between vapor and liquid in the soil pores, the proportion residing in each phase is an important parameter, perhaps more important in agriculture than in any other area of study. Researchers in agriculture have adopted a measure of this proportion called the Henry law constant, with symbol KH . It is defined as KH =

CG chemical concentration in gas phase(g/cm3 air) = CL chemical concentration in liquid phase(g/cm3 water)

(4.1)

That characteristic of a chemical called vapor pressure determines the quantity of that chemical that can be contained in a specific volume of air at a given temperature and pressure. Vapor pressures of the explosive chemicals of interest are strong functions of temperature. See Chapter 2. In fact, it has been shown [10] that, for the entire temperature range of interest, the vapor pressure of DNT is about 20 times that of TNT, but that the vapor pressures of both show similar temperature dependence. The vapor pressure of each increases by about a factor of 4 for each decade increase in temperature, from about 5◦ C to about 50◦ C. Also, the solubility of these chemicals is weakly dependent on temperature below 20◦ C, but above 20◦ C the temperature dependence increases substantially [11]. In both ranges DNT is about one and a half times as soluble as TNT. 5 The

unpublished diagram in Figure 4.5 was developed during the research at Sandia in the 1990s, based principally on work by Phelan.


81

Phelan and Webb combined these effects [1, p. 27] to determine the dependence of KH on temperature. Their chart is reproduced as Figure 4.6. Examination of this chart reveals several facts of interest. First, we see that the molecule concentration in the vapor phase is always several orders of magnitude smaller than that in the liquid phase. Next, we see that there tends always to be a higher proportion of DNT in the vapor phase relative to the proportions for TNT, by about a factor of 10. At the higher temperatures we can expect to find nearly 100 times the proportion of molecules in the vapor phase as at the lowest temperatures. Of course, these are relative, rather than absolute values. Also, as the proportion of molecules in the vapor phase increases, there must be a corresponding decrease in the proportion in the liquid phase. Just as the Henry law constant is used to quantify the partitioning between air and water, it is useful to define a parameter to quantify the partitioning between soil and water. The linear adsorption coefficient, Kd , has often been defined for this purpose: Kd =

CS sorbed concentration (μg/g) = CL aqueous phase concentration (μg/mL)

(4.2)

Kd has units of milliliters per gram (mL/g). They then summarize the work of several research teams and individuals working to experimentally determine values for Kd [1, pp. 28–30]. These efforts were made in a number of different soils, with different moisture levels, different organic content, and different soil chemistry. Published results for these conditions naturally differ widely; however, it appears that values of 0.5 ≤ Kd ≤ 4.0

Henry's Law Constant

1.E−04

1.E−05

1.E−06 KH DNT KH TNT

1.E−07

1.E−08

0

5

10

15

20

25

30

35

40

Temperature (°C)

Figure 4.6

Henry’s law constant for DNT and TNT.

45

50

82


describe most soil conditions. The differing experimental techniques account for some of the variation. No dominant parameter causing the variations appears, but TNT seems to always have a higher value of Kd than DNT for the same conditions. In general, though [1, p. 26]: Soils with greater amounts of organic matter (agricultural or forest soils) or minerals (compared to desert sand) will sorb greater landmine signature chemicals,6 leaving less available for transfer to the air for vapor sensing.

As discussed earlier, the proportion of the molecules in the vapor state within the soil near a source is around 6 orders of magnitude less than the sorbed or dissolved portions. This condition was stated for 50% moisture saturation. For extremely dry soils, this ratio becomes even more pronounced. Phelan and Barnett defined a soil–air partitioning coefficient, Kd , in a manner parallel to the soil–water partitioning coefficient, Kd [12]: Kd (w) =

CS concentration sorbed on soil particles (g/g) = CG concentration in vapor state in soil pores (g/mL)

(4.3)

Note here that Kd is indicated to be a function of w, the gravimetric moisture content in grams/grams. That dependence on soil moisture is most pronounced in very dry soils, up to about 11% saturation, above which Kd seems to asymptotically approach moisture content independence. In the most dry soils Kd may be 8 or more orders of magnitude greater than at 50% saturation. A plausible explanation for this is that, as the soil becomes wetter, water displaces some of the explosive molecules sorbed on the soil particles, causing them to be released into the vapor state. In the driest soils almost all the molecules sorb to the soil particles. Dry soils, therefore, tend to act as reservoirs, or storage, for explosive molecules. These molecules can be released into the vapor state when the soil becomes moistened. Clearly, though, in totally saturated soil, where there is no air in the pores, there can be no molecules in the vapor state. Therefore, both at near-zero saturation and at full saturation the proportion of the molecules in the vapor state becomes vanishingly small. The concentration is substantially larger at some intermediate values of saturation. Therefore, we should expect to find an optimum condition of soil moisture, at which there is the highest concentration of molecules in the vapor state. Phelan and Webb show results from algorithms they developed [13–15] predicting those optimum conditions. They offer partitioning charts like Figures 4.7 and 4.8 [1, pp. 30–35]. The charts shown represent only one soil type, that is, only one value of Kd . Still, they give us a sense of the dynamics of partitioning as moisture content changes. The charts are based on the concept that “the whole is equal to the sum of the parts.” They illustrate the relative concentrations in each phase as functions of soil moisture. Their results indicate that a saturation of about 17.5% provides the highest concentration of molecules in the vapor state. 6

Principally TNT, DNT, and DNB.


83

1.0E-05 DNT 1.0E-06

Mass Fraction

1.0E-07 TNT 1.0E-08 1.0E-09 1.0E-10 1.0E-11 1.0E-12

0

10

20

30

Figure 4.7

40 50 60 70 Soil Saturation (%)

80

90

100

Soil vapor mass fraction.

1.00 0.90 TNT Soil Solid Phase

0.80 Mass Fraction

0.70 DNT Soil Solid Phase

0.60 0.50 DNT Soil Liquid Phase

0.40 0.30 0.20

TNT Soil Liquid Phase

0.10 0.00

0

10

20

30


80

90

100

Figure 4.8 Soil solid and liquid phase mass fractions, Kd = 0.9.

Phelan and Webb further show that while increasing temperature in the range from 5 to 45◦ C substantially increases the actual concentration in the vapor state, the maximum value still tends to be at about 17.5% saturation. In fact, the vapor state concentration at 23◦ C is about 10 times that at 5◦ C; at 45◦ C the mass fraction in the vapor state is about 5 times more than at 23◦ C [1, pp. 30–35].

84


Thus, it would seem that for anyone planning to sample for molecules in the vapor state, a very warm day with 17.5% soil moisture would be ideal. Figure 4.9 illustrates the results of another of Phelan and Webb’s [1, p 31, Table 13] calculations mentioned above. When we compare Figure 4.9 with Figure 4.5, the effect of soil type, as represented by the differing Kd values, becomes very apparent. As mentioned earlier, the foregoing discussion has significant implications for anyone designing a system for finding these molecules. We have seen that the dominant factors include soil type, including organic material, and soil moisture and temperature. But this is not a static condition. The molecules are migrating, forced by several processes. 4.3.1.3 Diffusion One of these processes is diffusion, the dominant transport mechanism for that portion of the molecules that are in the vapor state. As discussed above, this is a minor proportion of the total number of molecules in the soil. However, it is the mechanism that brings molecules directly to the open air above the source, where they are available to the most straightforward collection techniques, whether by mechanical or animal sensor systems. This process is indicated in the left-hand oval of Figure 4.3. Diffusion also occurs in the liquid state, as indicated in the center oval of Figure 4.3. In this phase it forms a minor, rather than the major, mechanism. In fact, it is difficult to separate vapor diffusion and liquid diffusion, and for this application separation is not really required. Therefore, Phelan and Webb [13], combining earlier work of Hamaker, Millington and Quirk, and Jury, defined an effective diffusivity that combines the two processes. They produced Figure 4.10 to illustrate.

Mass Fraction, Solid 86%

Mass Fraction, Vapor 0.00003%

Figure 4.9 Kd = 0.9.

Mass Fraction, Liquid 14%

Partitioning of explosive molecules in soil at 15% pore Saturation, 23◦ C,


85

1 Vapor Diffusivity Solute Diffusivity Effective Diffusivity

Diffusivity (cm 2/day)

0.1 0.01 0.001 0.0001 0.00001 0.000001 0.0000001

0

Figure 4.10

10

20

30


80

90

100

TNT vapor, solute, and effective diffusivity.

However, as we discussed above, when the soil becomes very dry, below about 11% saturation, the molecules tend primarily to sorb onto the soil particles. This reduces the effect of vapor diffusion. Therefore the validity of Figure 4.10 is reduced in those conditions. Realizing this, Webb and his colleagues explored it in more detail [14]. This work showed “a dramatic decline in effective diffusivity below 15% saturation” [1, p. 43]. This conclusion is of paramount importance. It means that there is no effective mechanism for transporting molecules from a buried source to the surface when the soil is very dry. The transport process can begin anew when rainfall, or perhaps artificial introduction of water to the soil, raises the saturation to levels where diffusion can again occur. The reservoir of sorbed molecules simply awaits a mechanism. 4.3.1.4 Convection In addition to the double-headed arrow labeled “Diffusion,” the center oval in Figure 4.3 contains a downward arrow labeled “Precipitation” and an upward arrow labeled “Evapotranspiration.” Both these latter terms refer to movement of the water within the soil. They form the major drivers of convection within the soil. In fact Phelan and Webb list three key conclusions concerning transport of the subject molecules within the soil [1, p. 42]: • Movement of landmine signature chemicals is controlled by chemical and soil

properties, and driven mostly by the movement of water in soils. • Water transports more TNT, DNT and DNB by convection than occurs by either vapor or solute diffusion. • Conditions that cause upward evaporation of soil water in proximity to the landmine will be most beneficial to chemical sensing.

In general, water applied to the soil surface tends to move deeper under the influence of gravity, taking the molecules into solution and carrying them lower.

86


This would seem, at first glance, to make them even more inaccessible. That would be true if the influx of water never stopped. However, when it does stop, the process of evaporation begins, driven mostly by solar heating. This causes the water to move to the surface, carrying with it the molecules in solution. Of course, plants take in some of the water, which then moves through their systems. Molecules dissolved in that portion of the water seem to remain within the plants rather than being released into the air. Effects of Moisture, Rain, Dew, and Evaporation These weather effects become major factors in the transport of the molecules from the source to the surface of the soil, and thence into the air. Phelan and Webb again list some key conclusions [1, p. 26]: • Soil moisture has a tremendous effect on soil-vapor sorption. Dry soils will

sorb about 10,000 times more landmine signature chemicals than damp soils. This depresses the vapor levels the same amount. This process is reversible, so daily morning dew is valuable for vapor sensing, and afternoon drying is detrimental for vapor sensing. • The soil acts as a temporary storage reservoir for the landmine signature chemicals, releasing them when dew or rain falls, and collecting more as soil water evaporates.

When we began this study at Sandia, we did a number of experiments to gain an understanding of the migration of explosive molecules under field conditions. We also did some laboratory experiments to extend this understanding. One of the more enlightening experiments is illustrated in Figure 4.11. Phelan and his colleagues described in detail the techniques of the experiment and the complete results [15]. They also developed a predictive mathematical model along with the experiments. The experiment vividly illustrated results that have been regularly observed in field trials and are reflected in the last quotation above. These results clearly show the major effect of convection in the transport process. A direct implication of this experiment becomes apparent: It is quite possible to determine favorable and unfavorable times for vapor flux sampling, based on moisture events. Furthermore, it may sometimes be possible to artificially enhance sampling by appropriate local introduction of moisture.7 Otherwise, it may prove more effective, in some environments, to sample surface soil particles rather than vapor. The simulated weather changes that produced the variations in surface vapor flux rate shown in Figure 4.11 began with a soil sample with about 50% pore moisture saturation in a chamber under an artificial atmosphere. A quantity of DNT was introduced beneath the surface, at a depth typical of a landmine, about 3.5 cm. The atmosphere in the chamber was controlled to 50% relative 7 An

issue of curiosity, engendering some little debate, is whether dogs do this with their exhaled breath before each inhaled breath collects samples for the aroma being sought.


87

2, 4-DNT Surface Flux Artificial Drying Begins

1000000

2, 4-DNT (pg/min)

100000 10000 Relative Humidity Reduced to 0%

1000 100 Relative Humidity 50%

10

Artificial Wetting Begins

1 0

10

20

30

40

50

60

70

80

90

100

Time (days)

Figure 4.11 Effect of wetting and drying on surface vapor flux of 2,4-DNT.

humidity. In a bit less than a week flux was measurable above the surface. That flux increased rapidly until it began to approach an equilibrium state. From prior experiments it was understood that equilibrium was expected after about 35 days. Therefore, on day 35 the relative humidity was abruptly reduced to 0%. Two effects are visible. There is the immediate reduction in flux as the surface dries and molecules are sorbed to the surface particles. This is followed almost immediately by a marked increase in flux as evaporation draws moisture into the atmosphere above the surface, bringing dissolved molecules with it, tending toward a new equilibrium. On day 44 the soil was subjected to a drying process. As the moisture leaves the soil, the molecules have more opportunity to remain in both the vapor state in the pores, as air replaces water, and in the solid phase, sorbed to the surface of soil particles. Drying was continued until day 69, simulating the normal process in soil between rain or dew events. At that point water was introduced to simulate precipitation. The dramatic spike in flux rate reflects the ability of the newly introduced water to transport molecules to the surface. This is consistent with the previous discussion where we saw that very dry soil holds on to its molecules. The drying–wetting cycle was repeated with similar results on day 79. The latter two cycles exhibit behavior similar to a field condition where there is a substantial diurnal cycle of heating and cooling, with morning dew. Once molecules reach the surface, whether from a buried source as in the foregoing discussion or by some other transport mechanism from a nonburied source, they tend to sorb onto the surface particles. Some will be in the vapor state and some in solution, but as any water evaporates, they must revert to vapor state or sorb to surface particles.

88


Figure 4.12

4.3.1.5

Unfuzed Soviet-type TM-62P2 being placed for experiments at Sandia.

Residue on Surface

SNL Results from Experiments Among the experiments we conducted at Sandia during the early stages of our investigations was the establishing of a sample minefield, using Soviet mines supplied by the U.S. Marine Corps. We buried them according to published Polish military doctrine for the particular mine type. They were buried, unfused, in the normal prairie loam found in Albuquerque, New Mexico. Figure 4.12 shows a mine being placed. The snap-in plastic cover over the fuze well is not a seal. It was left in place to approximate the presence of a fuze for overburden thickness control. The fuze well is a continuous part of the case and forms a watertight seal over the enclosed explosive. This mine was placed on February 7, 1997. After several months, on June 26, 1997, the concentrations were measured as shown on Figure 4.13. Soil samples for the vertical series of measurements were taken directly above and below the mine, at 1-inch intervals. The horizontal samples were taken at the upper surface burial depth, 1, 2, and 3 inches from the mine case. As noted on the diagram, TNT concentrated near the surface. In this case we find the DNT has not migrated as far from the source as the TNT. As this was early in the process, no real explanations were offered for the difference, and insufficient detail on weather effects were noted. Nevertheless, this diagram illustrates a condition found by several deminers as they used dogs to search for mines.8 As they studied effects of success and failure and realized that very often the molecules they were searching for were in the surface layer, rather than being available in the air. This led to some interesting realizations of the relationship between physical locations of surface concentrations and the buried source. 8

Personal communication with Dr. Vernon Joynt in April 2004.


89

Figure 4.13 TNT and DNT concentrations measured near the buried TM-62P2 antitank mine.

Dried Puddles Concentrate Molecules In our common experience, we have all observed the formation of puddles after a rain. We realize, without much analysis, that the puddle is not formed from rain that fell in only that location. It contains water that fell nearby and flowed to that area, which is at the locally lowest elevation. If the rain falls on an area that has buried sources of explosive molecules, then some of those that were sorbed to the surface particles above the source will be dissolved and carried into the puddle. When puddles dry they leave a concentration of molecules on the surface soil particles. Thus, an irregularly shaped area of relatively high concentration of molecules may appear some distance from any buried source. See the discussion on Figure 8.2 p 182. This concentration in puddles does not require a buried source. If a source on, or above, the surface receives rain, concentrating puddles can form. Since the presumed reason for searching for trace explosive molecules is to locate the source, some ingenuity may need to be exercised to complete that task. In fact, Phelan and Webb [1, pp. 70, 71] report on work by Hewitt et al. [16] where they buried mines on a gentle slope. The signatures from these mines were found to form in patterns where concentration decreased with distance (a few feet) from the mine as the surface water flowed down the slope away from the mine. 4.3.1.6

Above the Surface of the Ground

Boundary Layers Molecules may reach the surface in each of the three states. Those dissolved in water will, at some point, either be released into the vapor state or become sorbed to surfaces as the water evaporates, unless the water

90


becomes part of a larger volume that does not evaporate, but flows into a stream or pond. Transport of molecules that remain dissolved in water will be discussed in a later section. Molecules that originate from a buried source, upon reaching the surface, may be carried away by wind currents. This is not always the case since they emerge from their upward journey through the pores in the soil into a region of relatively stagnant air called a boundary layer. Ludwig Prandtl introduced the concept of boundary layers in 1904. Since they play a large role in determining the parameters of fluid flow, they have been extensively studied. A boundary layer is that region near a surface where the fluid flow is dominated by the presence of the surface. The fluid cannot flow through the surface; but there is always some attraction between the molecules of the fluid and those of the surface, the surface tension effect. In addition, at low velocities, the viscous forces in the fluid dominate the kinetic forces. Therefore, the fluid immediately adjacent to the surface is restrained in its normal tendency to move with the rest of the fluid. The result of this restraint is a velocity gradient. The velocity increases from effectively zero at the surface to the nominal fluid velocity at some distance away. The ability of the moving fluid to produce motion of objects on or very near the surface, from molecules to grains of sand, depends moving that object out of the boundary layer into the main flow field. This normally requires turbulence on the appropriate scale. The thickness of the boundary layer and the turbulence within it are determined by a number of factors, including flow velocity and surface roughness. Beginning with Prandtl’s original work, there have been many empirical equations proposed for estimating the boundary layer thickness. All recognize that the thickness is inversely related to some fractional power of the Reynold’s number. δ∝

1 (NRe )1/n

where d is the boundary layer thickness, NRe is the Reynold’s number, and n varies from 2 to about 7 among the theories. Since the Reynold’s number is defined as NRe =

flow velocity × characteristic length × fluid density V lρ = μ fluid viscosity

(4.4)

the thickness is observed to depend inversely on the velocity. That means that boundary layers are thicker for slower flows. This explains our common observation that higher wind velocities raise dust from locations that lower winds do not. The higher winds producing a thinner boundary layer are able to move particles that remained within the thicker boundary layer at lower wind velocity. Since searching for trace vapor signals in other than nearly calm conditions is almost certainly a futile exercise, the boundary layers of interest are relatively thick, perhaps a few centimeters.


91

However, the molecules percolating up into the boundary layer from beneath the soil surface tend to become trapped in the stagnant laminar sublayer of the boundary layer. This sublayer is usually much thinner than the overall turbulent boundary layer, since it is dominated by viscous and surface tension forces, rather than by velocity. Phelan and Webb call this the “chemical boundary layer” and state categorically that “there will generally be no chemical signature” above this chemical boundary layer [1, p. 52]. Other factors do intervene. Significant solar heating of the soil surface, so that the soil becomes warmer than the air, causes vertical thermal convection currents to develop within the boundary layers. This introduces turbulence or instability that acts to move the chemical signature up into the free air. When the molecules are moved into the free flow of the air, the effect is to reduce the concentration by dilution. Conversely, when the soil surface is cooler than the air, thermal convection is inhibited, with the result that the molecules are effectively trapped in the boundary layer. This effect is strengthened by the cooling of the air adjacent to the surface, which increases its viscosity. Higher viscosity lowers the Reynold’s number, thus decreasing boundary layer thickness. Thus, we can expect that the flux of molecules that form the chemical signature at any given location will vary with time and weather conditions. This can even include intermittent transients. In addition, mechanically disturbing the surface can result in moving particles or molecules out of the stagnant part of the boundary layer where they can be moved by the wind. This effect can often be observed in desert regions. As a wind increases, it is often possible to see dust being lifted from footprints left since the last wind. If there is vegetation, such as short grass, the boundary layer may be largely submerged in it. Molecules do not readily find their way through this boundary layer; hence, it may be necessary to force them into a collection device by suction. That appears an integral part of a mammal’s technique. In most cases the molecules that form the desired chemical signature will be found very close to the surface. Realizing this will enable more effective sampler designs to be fielded. Some of the molecules that do make their way into the free air above the boundary layer are likely to sorb onto the surface of any object that is in their flow path. Once this happens, that molecule is effectively lost for collection by vapor sampling techniques, reducing the available concentration in a sample. In many search areas plants form the most available surfaces for molecules to fall upon. Hence, it is possible that plant surfaces near a source might form a reservoir for molecules that could be profitably exploited by innovative sampling techniques. Certainly, it is well recognized that when plants take in water through their root system that they may be also taking in the molecules released from a nearby source [17]. 4.3.1.7 Plant Uptake When the molecules are taken into a plant’s system, they may become concentrated in parts of the plant or they may begin to break down. These processes have scarcely been examined. When they are, it is possible

92


that new opportunities will be revealed. This area needs thorough investigation. This process is represented in Figure 4.3 by the lower right oval. That oval also includes microbiological processes that cause breakdown of the molecules. 4.3.1.8 Soil Type Effects Those microbiological processes represented in that lower right-hand oval of Figure 4.3 depend heavily on soil types. They are even more strongly dependent on moisture level and temperature. In addition to microbiological degradation, there is abiotic degradation from chemical reactions with soil components. These processes are most pronounced when the soil contains iron-bearing minerals [18]. Phelan and Webb report work from several studies [1, p. 36] and summarize the results: • Landmine signature chemicals change form, chemical properties, and eventu-

ally become eliminated from soil systems by microbiological and soil mineral degradation reactions. • Both biologic and abiotic reactions require water for degradation reactions to proceed.

Degradation times are normally stated in terms of half-life.9 Measured results vary substantially, depending on experimental technique, but in wet soil at warm (>20◦ C) temperatures both TNT and 2,4-DNT can show a half-life of one day or less. At temperatures just above freezing, this increases to a couple of weeks; while at temperatures below freezing, the half-life extends to years. Different soils show varying rates of degradation. Phelan and Webb’s summary included both laboratory and field measurements from multiple sources. Results from laboratory and field results correlate reasonably well. Continuing their summary [1, p. 36]: • Laboratory measurements of TNT, DNT, and DNB found degradation rates

to be dependent on soil type, soil moisture content and temperature. • Higher clay content and organic matter content soils have higher degrada-

tion rates. • Soil moisture contents greater than 1% cause very fast degradation rates (half

the amount degrades over the period of one day).10 • Soil moisture contents less than 1% preserve landmine signature chemicals (half the amount degrades over the period of 3 years) • Only subzero (◦ C) conditions limit degradation. The greater the temperature, the greater the degradation rate.

Table 4.4 [1, p. 37] lists the principal degradation products of the signature chemicals. See the additional discussion in Chapter 2. It may be useful to consider searching for these compounds in addition to the parent compounds when conditions for rapid degradation are present, or when the source being sought may have been in place for some extended time. 9 Half-life is the time, usually quoted in days for this application, during which the concentration is reduced to one-half its previous value. 10 At temperatures above 20◦ C (author’s clarification).


TABLE 4.4

93

Parent and Degradation By-products of TNT, DNT, and DNB

Parent Compound

Degradation By-product (abbreviation)

2,4,6-TNT

4-Amino-2,6-dinitrotoluene (4A-DNT) 2-Amino-4,6-dinitrotoluene (2A-DNT) 2-Amino-4-nitrotoluene (2A-NT) 4-Amino-2-nitrotoluene (4A-NT) 3-Nitroaniline (3-NA)

2,4-DNT 1,3-DNB

4.3.1.9 Climatic Effects Weather conditions cause several effects on the concentration level of the chemical signature at the soil surface and immediately above it. Earlier discussion focused on the cycles of rain or dew and the associated evaporation. This is the largest effect, but it is closely associated with the diurnal cycle of heating and cooling of the soil. Direct solar radiation and long-wave radiation from the atmosphere heat the surface during the day, while long-wave radiative transfer to space cools it at night. Soil surface temperature may also be affected by air temperature as warm or cold fronts move through the area. The amount of temperature change in each diurnal cycle is greatly influenced by cloud cover and relative humidity. Soil color greatly affects the magnitude of temperature changes due to radiation. Plant coverage may provide enough insulation to greatly reduce that effect. In addition to the diurnal cycle, the seasonal cycle produces much variation in the surface temperature, hence in the processes that control concentration. Concentrations tend to be much higher in the warm summer temperatures when evaporation is driving percolation of the molecules to the surface, than in the winter, when activity may become halted by freezing. Even if no freezing occurs, the processes are slower at colder temperatures, producing lower signature concentrations. These weather changes affect the signature concentration in several ways. Evaporation increases with increasing surface temperature and decreases with increasing humidity. Boundary layers are less stable when the surface is undergoing direct heating, causing vertical convection currents that carry molecules through the boundary layers. Microbiological processes producing molecular degradation are much more pronounced in warmer soils. These processes are greatly restricted when temperatures fall below freezing (0◦ C). Continued heating reduces soil moisture. When this results in the saturation dropping below 8 to 10%, the available signature drops dramatically, for reasons discussed previously. Drought conditions may cause this to persist for long periods. Conversely, continual rainfall will drive the soil saturation to a level that precludes molecules in the vapor state. In either condition, without a change in weather to permit evaporation no signature will be available to vapor state collectors. Prolonged or heavy rainfall may produce erosion, moving the surface soil to other locations. It will almost certainly produce runoff that carries dissolved

94


TABLE 4.5 Estimated Vapor Concentrations in the Air Boundary Layer from Surface Soils Residue Data and Ks/a Values TMA5 Chemical 2,4,6-TNT 9.4 × 10−4 2,4-DNT 2A-DNT 4A-DNT

PMA1A

ng/L

ppt

ng/L

ppt

9.4 × 10−4 1.5 × 10−1 9.0 × 10−3 1.1 × 10−2

0.1 20 1 1

8.5 × 10−4 3.0 × 10−1 2.3 × 10−2 2.0 × 10−2

0.1 40 3 2

molecules elsewhere, as discussed above. In either case, this condition can produce signature concentrations that are difficult to relate to a source. These can, in effect, produce a form of false-positive search results. The concentration produces a true chemical signature, but not in a location readily associated with a source. One vivid photo, no longer available, exhibited this extreme condition, showing a mine partially uncovered by erosion in the bank of a wadi in Israel. The chemical signature might well have been discernable in the bottom of the wadi some meters distant from the actual source, misleading at best. 4.3.2

Concentration Estimates from Buried Sources

At the time of the studies described, and up to the present, a vapor sensing system with sensitivity adequate to routinely discover the vapor signature from buried landmines has not been fielded in quantity. Therefore, it is necessary to estimate the concentrations that may be expected, so that system developers may form realistic design goals. Jenkins and his colleagues estimated the air concentrations, for one kind of soil and two types of mines, as shown in Table 4.5 [8]. The quantity Ks/a is equivalent to the Kd previously defined. It was calculated as the ratio of soil residue to vapor concentration in their experimental samples. They were working at the limits of sensitivity of available instrumentation to produce this work. It is well established11 within the demining community that mammals have a vapor sensing capability several orders of magnitude more sensitive than current electronic systems, even laboratory systems, though recent developments (Chapter 7–9) are beginning to change that. Most mammalian fieldwork has been done using dogs, but some work indicates that rats may be even more capable [19]. Other biological sensors may be even more capable. The well-known12 ability of some insects to detect one or two molecules of some chemicals provides a goal for sensor developers. 11 Several

conversations, 1996 through 2004, with Dr. Vernon Joynt, who has researched the use of dogs and rats as well as electronic systems in South Africa for locating explosives for many years. See Chapter 8. 12 Mentioned by Paul Patton in lecture notes on website: http://soma.npa.uiuc.edu/courses/bio303/ Ch13.html.


95

Phelan and Webb [1, pp. 73–78] report experiments that tried to determine the available concentrations by using established mine hunting methods employing a dog as the sensor system. They used several indirect methods, where the vapor in a suspected or known minefield is collected on filters and presented to dogs at a convenient site. The experiments compensated for the concentrating effect of the filters inherent in these methods. Among others, they used the MEDDS [20] system and the REST [21, 22] system. When they did those experiments in 2001, they found the best laboratory sampling and analysis capable of about 0.5 ppt (0.005 ng/L). They further note that some dogs could sometimes sense concentrations as low as 10−10 ppt [1, p. 75]. See Chapter 8. Unfortunately, even with best efforts no one is yet able to provide a definitive value that a sensing system developer can use as the available concentration near a buried mine. It will continue to be necessary to develop more sensitive sensors. However, it also becomes increasingly valuable to use them more astutely, based on the behavior of the molecules as discussed here. Whether using artificial or biological sensors to search for buried explosives, a few things become apparent as important. Among them are: • • • • • • • •

Damp soil releases more vapor than very dry soil or saturated soil. The concentration of molecules may be expected to be higher in the soil surface than in the air above it. Vapor signatures become stronger or weaker at some times, depending on the moisture and temperature cycles. The air and chemical boundary layers tend to keep chemical vapor close to the surface; farther from the surface dilution becomes the dominant factor. There may be advantages to artificially inducing local moisture cycles, perhaps even in the immediate vicinity of the sample collection. There will be no vapor signature when the surface temperature falls below freezing. Molecules migrate on the surface after rains, following drainage patterns. Molecules persist in dry soils for a long time but degrade more rapidly in wet soils.

With these considerations, and others gleaned from the research, system developers can use innovative sample collection techniques to extend the capability of their systems to detect lower concentrations and locate or identify buried explosive bearing objects. The issue of buried objects has attracted a great deal of attention, especially the worldwide proliferation of landmines; hence, there have been funds for research. This research will have application beyond landmines. Much UXO is buried, some because it was buried for disposal, some because it became buried in the course of the conflict. However, understanding the way the molecules are released and how they migrate after release will also assist in applications where the munitions are not buried but are hidden in various ways. There are also other environments worthy of consideration.

96

4.3.3


Other Environments

We may reasonably expect IEDs to be disguised. Penetrating the disguise may frustrate the efforts of their user. Some objects may be hidden by debris. Expected vapor signatures from these items required search in air. Many ordnance items are used in water; others are disposed of there. Search for traces of explosives from them may be expected to be somewhat different. 4.3.3.1 Underwater Objects We made a brief allusion to another environment in commenting on Figure 4.4. That figure shows one example of an environment that is very different from a simple buried munition. Several things become apparent immediately. Clearly, the aqueous state is likely to dominate the entire process. There will be some sorbing onto particles of the bottom, but even for objects totally buried in the bottom that will be reduced, since the bottom will always be at a saturated state. Therefore, it is reasonable to focus development attention on finding molecules in the aqueous state, rather than the vapor state. The Office of Naval Research (ONR) has sponsored substantial work of this nature. Some of it is reported in detail in Chapter 6. A look at Figure 4.4 illustrates some considerations when the object is submerged. It is worth noting that, while seamines are intentionally placed underwater for military purposes, they are not the only munition that may be found submerged. Of course, UXO is commonly found underwater, either having been dumped there, having been cargo on ships that sunk, or having fallen there as duds during conflict. The sea is not the only recipient of these ERWs. During conflicts some ordnance will be intentionally or incidentally placed in rivers, canals, streams, ponds, lakes, or irrigated fields. For locating these ERWs, the most important characteristic of a body of water is its current. Currents may be nearly nonexistent in some stagnant ponds or irrigated fields. Currents in streams and rivers are usually consistent in direction but variable in velocity. Currents in tidal areas oscillate in direction and vary in velocity. Rapid currents have higher Reynolds numbers, hence more turbulence. Even rather modest velocities in water produce turbulence because the Reynolds number is higher with the higher density than in air. Turbulence is not the asset in water that it is in air. In air, we saw that some turbulence is required to bring the molecules out of the chemical boundary layer. That may also be needed in water to move molecules away from the bottom surface. Turbulence away from the surface tends to break up the plumes of molecules that are diagrammed in Figure 4.4. Those plumes are the key to successful detection of an underwater object that is releasing the molecules of interest. One result of the ONR experiment at San Clemente Island, off San Diego, California, was a better understanding of the formation, persistence, and dissipation of these plumes. When a well-formed plume is available, it often becomes possible to follow it to its source; see Chapters 5 and 6. But how does the plume form? What processes control its persistence and its integrity? Are there other processes that control the concentration of molecules that provide the signature within the plume? Can we reasonably expect to find


97

detectable concentrations outside a plume? There has been some research directed toward answering these questions specifically for explosive signatures. However, there has been much research into similar plumes by workers in biological studies. They refer to them as “odor plumes.” The following section discusses plumes in more detail. There has been almost no research for objects buried in the seafloor13 that parallels the work of Phelan and Jenkins and their colleagues on land. Therefore, we can only surmise that processes similar to those they describe are at work. Clearly, as diagrammed in Figure 4.4, there will be diffusion and convection in the liquid phase. For buried objects, there must still be partitioning. There will be microbial degradation, in fact, almost certainly at a greater rate than for objects buried on land. That microbial degradation, much more than in air, can actively continue as the molecules enter the free water above the seafloor, whether they are in a plume or not. In addition, some molecules in the water will be ingested by larger aquatic organisms. In stagnant water no plume may form. This could cause a condition where the signature is ample for detection but so distributed that no source location can be inferred. In tidal areas the plumes will reverse directions with the tidal currents. Clearly, there are enough unanswered questions to offer ample research opportunities. 4.3.3.2 Surface Source or Hidden Source Sometimes source munitions are hidden, either intentionally or unintentionally, with neither soil nor water above them. In this case, we may expect that chemical signatures will be found in the vapor state. It is possible that sampling of surfaces could lead to detection, but normally a search will be in air. Many animals routinely face this same task. Whether looking for a meal or a mate, a common approach is to follow an odor plume to its source. 4.3.4

Odor Plumes

A plume is a collection of fluid molecules within a larger body of that fluid that, with their suspended particles and dissolved chemicals, are moving as a more or less distinct ribbon through the containing body of fluid. While some might call it tubular, it is often styled as a ribbon because of its high aspect ratio. The term “plume” is normally applied to a portion of the fluid that is being rendered distinct by the injection, over a period of time, of the characteristic dissolved chemical or suspended matter. We think of plumes of smoke from a chimney or a stream of muddy water joining a flowing stream of clear water. A classic form of plume, familiar to everyone, is the smoke from a just extinguished candle. Close observation of this plume in still air reveals that it is tightly constrained and very smooth for some distance above the wick. A few inches above the wick the character of the plume changes. It begins 13 The

term seafloor is used generically to refer to the bottom of any body of water, even a field flooded for irrigation.

98


to twist, widen, and become less distinct. It rolls about itself and forms gaps, finally disappearing altogether. This behavior is characteristic of chemical plumes or odor plumes as well. If we think in very simple terms, we may regard the tightly constrained portion to demonstrate laminar flow, with the introduction of turbulence causing the observed breakup of the plume. While that may be generally correct, things are, of course, not that simple. There is extensive literature on plumes as related to stack emissions and other environmental pollution issues in air, in water, and in groundwater. There are several government agencies [23–26] that maintain websites to direct inquirers to data, analyses, and models describing plume formation in all levels of detail. A web search [27] will quickly locate more references. Unfortunately, all these are at the wrong scale for the sensing of trace chemicals. They are focused on large-scale events where quantities are measured in grams to tonnes and distances in kilometers. In contrast, when seeking plumes of trace chemicals, the quantities are in nanograms or less and distances are in centimeters to meters. Fortunately, there is another body of literature, one that is focused on scales compatible with our interest. Researchers who are exploring the mysteries of behavior in a variety of animals have found that following odor plumes forms an essential part of the activity patterns of many, from crabs to moths to foxes to elephants. Odor plumes provide a path to food and to potential mates. They also provide warning of danger from predators. There is an extensive literature reporting this research. Much of it is specific to habits and behavior of the animal being studied, but much of it has direct application to locating chemical signatures. Attention is directed to only a few parts of this research that have immediate application to our subject. Perusal of the reference lists in the publications cited here will lead one into that larger volume of literature. 4.3.4.1 Plumes in Water Several fine researchers have approached the study of underwater plumes with different objectives. While all this work is undoubtedly instructive, a series of articles [28–32] produced by Webster and Weissberg and their colleagues may apply most directly. They have examined the structure [28] of plumes in controlled experiments and produced photographs of dye plumes to study their development. They also took the point of view of a hungry crab [32]. In its attempt to find the food source indicated by the plume, the crab manipulates its sensors within the plume. The structure of the plume makes it necessary. Chapter 5 is devoted to a description of these plumes. 4.3.4.2 Plumes in Air It is clear that, if the supply of molecules is adequate, plumes can form in air as well. While there may not be enough concentration in the air above buried sources to form plumes, in the case of unburied explosives, such as in IEDs or UXO covered loosely with rubble, there may be exploitable plumes. When such plumes develop, there will often be some level of urgency associated with locating the source. Sometimes, when logs seem to be following

EF&T IMPLICATIONS FOR SEARCH AND SAMPLING STRATEGIES

99

a plume they may be sampling from residue on plants on surfaces that act as concentrators of molecules deposited by very weak plumes over an extended period.

4.4 EF&T IMPLICATIONS FOR SEARCH AND SAMPLING STRATEGIES 4.4.1

Sources Buried on Land

The primary purpose for this discussion of EF&T of the molecules is to provide the one who employs a trace chemical sensor with means to increase the probability of locating and identifying the source. A thorough understanding of the transport processes presents options to the system designer and the operator to make a system more successful. Since “there is no effective mechanism for transporting molecules from a buried source to the surface when the soil is very dry” (see Section 4.3.1.3) when faced with such conditions, the system operator may need to become innovative. In addition, unless there is some mechanism for horizontal transport of molecules in moist soil, the operator should expect the strongest surface concentration to appear directly above the buried source. Search strategies that exploit the natural tendency of the molecules to remain adsorbed on surface particles and to become trapped in the boundary layers near the surface are likely to be most productive. If odor plumes from buried sources do develop, they are likely to remain close to the surface until they dissipate from turbulence. When seeking buried sources, the material presented here should make it clear that some times and some weather conditions are more favorable than others. Alternately, when circumstances do not permit waiting for a more propitious time or weather condition, it may be that artificially increasing either moisture or temperature will prove useful. 4.4.2

Sources Producing Plumes

4.4.2.1 Weather Effects When following plumes to a source, weather can be a major factor. Obviously, this applies in water as well as in air. When turbulence is heavy, the plume will become more dispersed. This turbulence can result from a temporary, weather-induced condition, but may force delay in plume tracing. Judging the degree of turbulence in water is somewhat easier than in air, simply because we can see the water. Substantial research on environmental pollution plume persistence and dispersion in air has produced a classification system called Pasquill’s stability classes for plume stability in air. Unfortunately, the system does not apply as directly as we might wish since it applies at the much larger scale of stack exhausts and the like. However, some insight from the system is available.

100


TABLE 4.6 Stability of Air in Various Meteorological Conditions1 Daytime Solar Heating

Nighttime Cloud Cover

Wind

Much

Some

Slight

Sparse

Heavy

Calm

Most Unstable Very Unstable Somewhat Unstable

Very Unstable Unstable

Unstable

Moderately Stable Somewhat Unstable Somewhat Unstable

Most Stable Moderately Stable Somewhat Unstable

Gentle Breezy to Strong 1

Somewhat Unstable

Somewhat Unstable Somewhat Unstable

Lighter shading denotes more stable air; heavier shading, more unstable.

This system classifies the local turbulence conditions into classes A through F, where class A denotes the most anticipated turbulence, or the most unstable conditions, and class F denotes the most stable conditions. The class is determined for a particular location at a particular time, based on wind speed, cloud cover, and sun angle. At least two websites currently present tables for determining the local stability class. A government site14 presents a step-by-step procedure to precisely determine the class. A commercial site15 presents a more generalized table. A much-simplified classification, based on these tables, is shown in Table 4.6. The scale of interest is very different for trace chemical or odor plumes from that for which the original classification system was developed. Therefore, it is proposed that the simplified classification scheme of Table 4.6 will be sufficiently applicable to provide an indication of the more and the less favorable conditions for following plumes. In this presentation the lighter shading denotes those conditions that are generally more favorable; darker shading indicates more instability and hence less well defined plumes. There may be some surprises when considering those conditions that are the most favorable for stable odor plumes. Best conditions occur on cloudless nights with the lightest breezes. Worst conditions for plume stability are on bright, sunny days with the lightest breezes. It appears that surface heating is the source of the daytime turbulence. Conversely, clear nights, with radiant cooling to space produce the least turbulence from the surface. Whether the plume is being sought in air or water, weather conditions will definitely affect the formation of the plumes. Appropriate scheduling holds promise for more success. 4.4.2.2 Effects of Plume Characteristics Chapter 5 describes the behavior of plumes in more detail, but a brief discussion of their effect on search is appropriate here. Two characteristics of plumes appear significantly important 14 http://www.eglin.af.mil/weather/Pasquills.html. 15 http://www.air-dispersion.com/formulas.html#stability

heavier shading, more unstable.

Lighter shading denotes more stable air;

OPEN QUESTIONS AND FRUITFUL AREAS FOR FUTURE RESEARCH

101

in the search for chemical signatures. Since, in experiments, a dye plume often seems to consist of long, twisting filaments of color moving through clear water, we might call the first “filamentation.” We use this term to qualitatively describe the degree to which the plume resembles a ribbon suspended in the water. This filamentation often dominates the plume’s appearance near its source, but as the plume encounters turbulence, it becomes less like a ribbon and more like a cloud. Since the plume appears to be rather uniform within its bounds, we might call the other characteristic homogeneity. Turbulence also effects plume homogeneity. Turbulence interacts with the plume on different scales. It may be induced by the plume’s flowing past a surface, by thermal convection cells, or by cross currents in the water. When the scale is large relative to the diameter of the plume, the effect is somewhat like twisting or rolling a ribbon. When the scale is comparable to the diameter of the plume, it directly affects the cross-sectional homogeneity of the plume. If we were able to place a sufficiently capable, sufficiently small sensor at a spatially fixed point within a plume and measure the concentration as the plume moves past us, we would observe a variation with time. If the degree of filamentation were high enough we would see the signal vary abruptly from strong to essentially zero, then back again to strong, continuing in a random manner. On the other hand, if there were little filamentation, in a more homogeneous plume, the signal would vary more gradually, but still might disappear completely at times. Therefore, any finite sampling time produces a time average of the concentration. The longer sampling times needed to compensate for lower sensor sensitivity may actually be collecting an integrated average of instantaneous concentrations that vary widely. For a broader look at the plume, suppose we form an array of those perfect sensors in the preceding paragraph, and orient the array in a plane normal to the axis of the plume. This is the approach adopted by Ishida et al. [33] In a plume with pronounced filamentation they observed the concentration to vary randomly among the quadrants of an eight-element array. They also photographed that plume. A single, stop-action photo vividly showed the filamentation, but a multiple exposure (6000 frames at 10 frames per second) from the same spot produced an image that seemed completely homogeneous. The crab [32] maneuvers its sensors in an apparent effort to capture the strongest signals. Time averaging does not seem to be a technique used by crabs, but it may be an effective strategy for artificial sensing systems

4.5 OPEN QUESTIONS AND FRUITFUL AREAS FOR FUTURE RESEARCH

As the technology for sensing smaller concentrations of explosive molecules develops further, and new systems are fielded and used, the need for better understanding of where to find these elusive molecules will become ever more

102


evident. This will be particularly evident when intelligent machines form part of the system. These machines must be programmed or instructed to follow certain rules and procedures. Better understanding of the EF&T of the quarry molecules enables more effective and efficient programming. Some obvious open questions must certainly be addressed. 4.5.1

Objects Buried in the Sea Bottom

As mentioned in Section 4.1.1, the extensive work of Phelan and Webb has no parallel for objects buried in the seafloor.16 Whether searching rice paddies or oceans, if a munition has been dropped from the surface, or lain in the water for very long, it is likely to have become at least partially buried in the seafloor. The understanding of the processes at work in these conditions is rudimentary at best. Many sensors commonly used to find objects on the seafloor prove ineffective when the objects become buried. With a more thorough understanding of the EF&T of the explosive molecules in these situations, trace chemical sensors may be able to provide more success in locating these objects. 4.5.2

Sampling Plant Material

Except for extreme desert areas, most locations where munitions are buried will have plants. When the plants are growing directly over the munition, some of the explosive molecules released will certainly be drawn into the plant with its water intake. Whether these molecules are metabolized or concentrated has been studied for a few plants [17]. In either case, there may be opportunities for exploitation. If the explosive molecules are metabolized, the plant may produce some characteristic chemical signature that could aid in locating the source. If the molecules are concentrated, then it may be plausible to use the plant as a preconcentrator and sample the plant tissue to search for explosive compounds.

4.6

ROLE OF COMPUTER MODELING

Several types of computer models have been developed for estimating the expected concentrations of the chemicals of interest as they move away from the source. Soil transport models attempt to estimate the expected concentration at the surface above buried sources. Plume transport models attempt to estimate the concentrations within a plume, along with its shape and position. A different form of model is designed to guide a search pattern for employing a sensing system to trace a plume. 16 Recall

that the term seafloor in this work designates the soil of the bottom of any body of water, whether sea, lake, pond, river, stream, or irrigated field.

ROLE OF COMPUTER MODELING

4.6.1

103

Soil Transport Models

Webb and his colleagues [14] have had considerable success in modeling the relative concentrations of molecules in their migration through the soil. They developed a code called T2TNT to include the effects of weather. This type of analysis can provide field workers searching for signatures with indications of expected success as moisture changes in a particular area, or as the search moves to a different type of soil. For example, they analyzed the conditions illustrated in Figure 4.11. Figure 4.14 shows the degree of success of the predictions. Therefore, we may infer that, if the soil and moisture conditions are well known, the surface flux can be estimated with considerable accuracy. Phelan and Webb [1, pp. 65, 66] present results of diurnal concentration predictions for periods up to a year. The results indicate that diurnal liquid phase concentrations tend to oscillate by about 8 orders of magnitude; gas-phase levels, by about 6. This oscillation is superimposed on a weather cycle and seasonal variation. These longer cycle effects move the diurnal concentrations up or down by about 2 orders of magnitude. They often reduce the diurnal variations to only 1 or 2 orders of magnitude, sometimes on the high side, sometimes on the low. Hence, it would appear that field workers would find it to their advantage to employ these modeling techniques, realizing that they must input good data to expect good results. In any given field of buried munitions there will be times that produce much higher concentrations than at other times. Searchers may be able to increase efficiency, and perhaps safety, by using these predictive tools.

1000000

Surface Flux (pg /min)

100000 10000 1000 100 10 1

0

10

Figure 4.14

20

30

40 50 Time (days)

60

70

T2TNT predictions compared to data.

80

90

104


4.6.2

Plume Transport Models

Much research has been applied toward understanding the ways that various animals use odor plumes. It is not within the scope of this book to exhaustively survey the resulting literature. One aspect of this research has direct bearing on the use of plumes to locate sources of chemical molecules such as explosives. As they are studying their animal subjects, these researchers construct two kinds of mathematical models to assist them. One of these models the plume itself, attempting to provide a predictive capability for describing the characteristics of the plume in a measured environment. Knowing the motion characteristics of the medium—air or water—the models seek to predict the concentration and dimensions of the plume at distances downstream. Of course, the nature of the problem quickly leads to use of statistical descriptions. This provides a model quite adequate for constructing tracking algorithms. An example of this work is that of Farrell and co-workers [34]. They present a rather complex model to attempt to account for the effects of fluid motion and turbulence in three different levels of scale, relative to the plume. They begin with classical equations of motion, but by breaking their particle velocity vector into three components related to the three scales of interest, they are able to introduce appropriate statistical descriptions for the components. The result is a model that retains both the diffusive and the filamentary nature of the plume. 4.6.3

Plume Search Models

The other type model attempts to track the plume using techniques inspired by observation of animal subjects. Some of the models actually emulate the search behaviors observed. A variation on this type model, perhaps more appropriately called an adaptation, is that of Ishida and co-workers [33]. In follow-on papers to [34], Farrell and co-workers [35, 36] first use hidden Markov methods (HMM) to develop a model for mapping a plume, presupposing a searcher moving into the plume’s vicinity in search of it. Their model guides the searcher from first discovery on a path that enhances the searcher’s probability of maintaining continuous contact with the plume, while directing the search toward its source. They compare this mapping with the plume described in the model of [35] with satisfactory results. In the third paper [36] they present algorithms for use in plume tracing by an autonomous vehicle. They show their level of success by presenting experimental results. Their procedure differs markedly from those of Chapter 6 and of Ishida and co-workers [33].

4.7

CONCLUSIONS

When the value a sensor system is based on its ability to find very sparse concentrations of a few very specific molecules, every available advantage needs to be taken to ensure even modest success. We have all heard the story of the drunk

REFERENCES

105

searching for his keys under the street light, just because the light was better, even though he knew he lost them down the block. Without proper application of the principles discussed in this chapter, someone employing a chemical sensor could be following the drunk’s procedure. It is important to search where the molecules are most likely to be found. It may be helpful to schedule the search to take maximum advantage of favorable conditions, but this also can be guided by knowledge.

REFERENCES 1. Phelan, J. M. and S. W. Webb. Chemical Sensing for Buried Landmines—Fundamental Processes Influencing Trace Chemical Detection, SAND2002-0909, Sandia National Laboratories, Albuquerque, NM, 2002 1a. Phelza, J. M. and S. W. Webb. Chemical Sensing for Buried Landmines: Fundamental Processes Influencing Trace Chemical Detection, in Mine Detection Dogs, Training, Operations and Odour Detection, Geneva International Centre for Humanitarian Demining, Geneva, 2003, p. 209–285 2. Phelan, J. M, P. J. Rodacy, and J. L. Barnett. Explosive Chemical Signatures from Military Ordnance. Sandia National Laboratories Report SAND2001-0755, Albuquerque, NM, April 2001. 3. Chambers, W. B., P. J. Rodacy, E. E. Jones, B. J. Gomez, and R. L. Woodfin. Chemical sensing system for classification of minelike objects by explosive detection, in A. C. Dubey, J. F. Harvey, and J. T. Broach, Eds. Proceedings of the SPIE 12th Annual International Symposium on Aerospace/Defense Sensing, Simulation and Controls, Detection and Remediation Technologies for Mines and Minelike Targets III, April 13–17, 1998. 4. Leggett, D. C., T. F. Jenkins, A. Hogan, T. A. Ranney, and P. H. Miyares, External Contamination on Landmines by Organic Nitro-Compounds. U.S. Army Engineer Research and Development Center—Cold Regions Research and Engineering Laboratory, ERDC-CRREL Technical Report TR-00-2, Hanover, NH, March 2000. 5. Mark, H. F. and Kroschwitz J. Encyclopedia of Polymer Science and Technology. Wiley, New York, 1985. 6. Rodacy, P. J., P. K. Walker, S. D. Reber, J. Phelan, and J. V Andre. Explosive Detection in the Marine Environment and on Land Using Ion Mobility Spectrometry, A Summary of Field Tests, SAND2000-0921, Sandia National Laboratories Report, Albuquerque, NM, 2000 7. Phelan, J., P. Rodacy, and J. Barnett. Explosive Chemical Signatures from Military Ordnance, SAND2001-0755. Sandia National Laboratories, Albuquerque, NM, 2001. 8. Jenkins, T. F., M. E. Walsh, P. H. Miyares, J. Kopzynski, T. Ranney, V. George, J. Pennington, and T. Berry. Analysis of Explosive-Related Chemical Signatures in Soil Samples Collected Near Buried Landmines. U.S. Army Engineer Research and Development Center—Cold Regions Research and Engineering Laboratory, ERDCCRREL, Report ERDC TR-00-5, Hanover, NM, March 2000. 9. Leggett, D. C., J. H. Cragin, T. F. Jenkins, and T. A. Ranney. Release of ExplosiveRelated Vapors from Landmines. U.S. Army Engineer Research and Development

106

10. 11. 12. 13.

14.

15.

16.

17. 18.

19.

20.

21.


Center—Cold Regions Research and Engineering Laboratory, ERDC-CRREL Technical Report TR-00-2, Hanover, NH, February 2001. Pella, P. A. Measurement of the vapor pressures of pressures of TNT, 2,4-DNT, 2,6-DNT and EGDN. J. Chem. Thermodyna. 9, 301–305 (1977). Phelan, J. M. and J. L. Barnett. Solubility of 2,4-dinitrotoluene and 2,4,6trinitrotoluene in water. J. Chem Eng. Data Mar/Apr (2001). Phelan, J. M. and J. L. Barnett. Phase Partitioning of TNT and DNT in Soils. Sandia National Laboratories Report SAND2001-0310, Albuquerque, NM, February 2001 Phelan, J. M. and S. W. Webb. Environmental Fate and Transport of Chemical Signatures from Buried Landmines—Screening Model Formulation and Initial Simulations. Sandia National Laboratories Report, SAND97-1426, Albuquerque, NM, June 1997. Webb, S. W., K. Preuss, J. M. Phelan, and S. Finsterle, Development of a mechanistic model for the movement of chemical signatures from buried landmines/UXO, in A. C. Dubey, J. F. Harvey, J. T. Broach, and R. E. Dugan, Eds. Proceedings of the SPIE 14th Annual International Symposium on Aerospace/Defense Sensing, Simulation and Controls, Detection and Remediation Technologies for Mines and Minelike Targets IV, April 5–9, 1999. Phelan, J. M., M. Gozdor, S. W. Webb, and M. Cal. Laboratory data and model comparisons of the transport of chemical signatures from buried landmines/UX, in A. C. Dubey, J. F. Harvey, J. T. Broach, and R. E. Dugan, Eds. Proceedings of the SPIE 14th Annual International Symposium on Aerospace/Defense Sensing, Simulation and Controls, Detection and Remediation Technologies for Mines and Minelike Targets IV, April 5–9, 1999. Hewitt, A. D., T. F. Jenkins, and T. A. Raney. Field Gas Chromatography/Thermionic Detector System for On-site Determination of Explosives in Soils. U.S Army Corps of Engineers, Engineer Research and Development Center, Cold Regions Research and Engineering Laboratory Technical Report ERDC/CRREL TR01-9, Hanover, NH, 2001. Thompson, P. L., L. A. Raner, and J. L. Schnoor. Uptake and transformation of TNT by hybrid poplar Trees. Environ. Sci. Tech. 32 (7), 975–980 (1998). Hundal, L. S., J. Singh, E. L. Bier, P. J. Shea, S. D. Comfort, and W. L. Powers. Removal of TNT and RDX from water and soil using iron metal. Environ. Pollution 97 (1–2), 55–64 (1997). Verhagen, R., C. Cox, R. Machangu, B. Weetjens, and M. Billet. Preliminary results on the use of Cricetomys rats as indicators of buried explosives in field conditions, I. G. McLean, Ed. in Mine Detection Dogs: Training, Operations and Odour Detection. Geneva International Centre for Humanitarian Demining (GICHD), Geneva, 2003, pp. 175–193. Joynt, V. The Mechem explosive and drug detection system (MEDDS), in I. G. McLean, Ed. Mine Detection Dogs: Training, Operations and Odour Detection. Geneva International Centre for Humanitarian Demining (GICHD), Geneva, 2003, pp. 165–174. Fjellaner, R. The REST Concept, in I. G. McLean, Ed. Mine Detection Dogs: Training, Operations and Odour Detection. Geneva International Centre for Humanitarian Demining (GICHD), Geneva, 2003, pp. 53–107.

REFERENCES

107

˚ 22. McLean, I. G., H. Bach, R. Fjellanger, and C. Akerblom. Bringing the minefield to the detector: Updating the REST concept. Proceedings of EUDEM2-SCOT—2003, vol. 1, 2003, pp. 156–161. 23. National Atmospheric Release Advisory Center. http://www.nrac.llnl.gov. 24. Subcommittee on Consequence Assessment and Protective Action (SCAPA) of the Emergency Management Advisory Committee of the U.S. Department of Energy. http://www.orau.gov/emi/scapa. 25. Atmospheric Dispersion Modeling Liaison Committee (ADMLC). http://www.ad mlc.org.uk. 26. Federal Coordinator for Meteorological Services and Supporting Research, Directory of Atmospheric Transport and Diffusion Consequence Assessment Models, FCMI3-1999, Washington, DC, March 1999. http://www.ofcm.gov. 27. Dispersion Modeling Feature Articles. http://www.environmental-expert.com/articleindex.htp. 28. Webster, D. R. and M. J. Weissburg. Chemosensory guidance cues in a turbulent chemical odor plume. Limnol. Oceanogr. 46(5), 1034–1047 (2001). 29. Zimmer-Faust, R. K., C. M. Finelli, N. D. Pentcheff, and D. S. Wethey. Odor plumes and animal navigation in turbulent water flow: A field study. Biol. Bull. 188, 111–116 (1995). 30. Horner, A. J., M. J. Weissburg, and C. D. Derby. Dual antennular chemosensory pathways can mediate orientation by Caribbean spiny lobsters in naturalistic flow conditions. J. Exp. Biol. 207(21), 3785–3796 (2004). 31. Dasi, L. P. Statistical Characteristics of Turbulent Chemical Plumes. Master’s Thesis, Georgia Institute of Technology, Atlanta, Georgia, July 2000. 32. Weissberg, M. J., C. P. James, D. L. Smee, and D. R. Webster. Fluid mechanics produces conflicting constraints during olfactory navigation of blue crabs, Callinectes sapidus. J. Exp. Biol. 206, 171–180 (2003). 33. Ishida, H., T. Nakamoto, T. Moriizumi, T. Kikas, and J. Janata, Plume-tracking robots: A new application of chemical sensors. Biol. Bull. 222–226 (April 2001). 34. Farrell, J. A., J. Murlis, X. Long, W. Li, and R. T. Cardé. Filament-based atmospheric dispersion model to achieve short time scale structure of odor plumes. Environ. Fluid Mech. 2, 143–169 (2002). 35. Farrell, J. A., S. Pang, and W. Li. Plume mapping via hidden Markov methods. Ieee Trans. Syst., Man, and Cybernetics—Part B: Cybernetics 33(6), 850–863 (2003). 36. Farrell, J. A., S. Pang, and W. Li. Chemical lume tracing via an autonomous underwater vehicle. IEEE J. Oceanic Eng. 30(1) 1–15 (2005).

CHAPTER 5

STRUCTURE OF TURBULENT CHEMICAL PLUMES DONALD R. WEBSTER School of Civil & Environmental Engineering, Georgia Institute of Technology

Sensing systems that track chemical plumes have great practical importance because they, for instance, facilitate locating sources of explosive compounds or improve detection of other hazards, such as pollutants leaking into the environment. An important aspect of understanding the process of sensing and tracking chemical plumes is characterizing the environment into which the chemical compounds are released. For instance, chemicals leaking from a point source are transported by the local fluid flow. Unless the velocity is extremely slow, the flow is likely to be turbulent. Thus, the transport of chemicals is a combination of advection by the bulk flow and turbulent mixing. The primary objective of this chapter is to introduce the basic characteristics and structure of turbulent chemical plumes. The second objective is to discuss the sensory signals that provide information for locating the source of the plume. 5.1

TURBULENT MIXING

Turbulent mixing is the process by which a fluctuating turbulent flow dilutes and homogenizes quantities such as chemical concentration, heat, and momentum. This chapter focuses on the turbulent mixing of chemical concentration, but there are strong analogies with other scalar quantities. Turbulence is difficult to define precisely, but turbulent flows have several common features. For instance, all turbulent flows are characterized by random, unpredictable fluctuations of velocity that are spatially organized into swirling packets called eddies. These velocity fluctuations advect chemicals in a spatially and temporally random fashion. Note that advection means transport by fluid movement (i.e., the fluid motion carries the quantity) and is often used synonymously with the term convection for transport of heat. Trace Chemical Sensing of Explosives, Edited by Ronald L. Woodfin c 2007 John Wiley & Sons, Inc. Copyright

109

110


Turbulent mixing can be thought of as a two-step process, but in reality both steps are occurring simultaneously and continuously. The first step consists of advection of a filament (or patch) of concentration. Because the flow field is spatially varying, neighboring regions of the filament are advected away from each other. Thus, the filament undergoes a process called turbulent stirring, which consists of stretching and folding by the random velocity fluctuations. Turbulent stirring continuously creates very steep gradients of concentration as the filament shape becomes more irregular. The second step of turbulent mixing is molecular diffusion. Fick’s law for molecular diffusion states that the flux of concentration is proportional to the strength of the concentration gradient. Diffusion acts across the large concentration gradients created by turbulent stirring to smooth the gradients, which effectively dilutes the filament and expands its volume. Thus, turbulent mixing can be conceptualized as continual random stirring that creates sharp concentration gradients, which rapidly diffuse. Turbulent mixing is significantly more effective than molecular diffusion alone. Consider the mixing process in a coffee cup. The time for molecular diffusion to mix a small patch of milk throughout the cup is several days, at which time the coffee would be also cold and unpleasant. However, by stirring with a spoon, and hence creating a turbulent-like flow, the mixing process is complete within seconds. Turbulent fields are often described in terms of spatial length scales that relate to the size of the eddies and filaments in the flow. The range of scales extends from the largest flow structures to the scales at which molecular diffusion acts to smooth gradients of momentum or chemical concentration. The integral scale of the velocity field describes the largest turbulent eddies in the flow. In an open-channel flow a good estimate of the integral length scale is half of the water depth, simply because the presence of larger eddies is constrained by the physical boundaries. In an atmospheric boundary layer, the layer thickness similarly constrains the largest eddy size. Diffusion of momentum of the velocity fluctuations (or dissipation of turbulent kinetic energy) occurs at the Kolmogorov scale, which is estimated as 3 1/4 ν η= (5.1) ε where ν is the fluid kinematic viscosity and ε is the dissipation rate of turbulent kinetic energy [1]. It is important to note that this is an order of magnitude estimate rather than equality. Thus, the Kolmogorov scale provides a rough estimate of the size of the smallest eddies present in the turbulent flow. The length scales for the turbulent concentration field range from the plume width to the scale at which molecular diffusion acts to homogenize the distribution (or dissipate the variance of the scalar fluctuations). The smallest length scale is referred to as the Batchelor scale and is estimated as 2 1/4 νD LB ∼ = η(Sc)−1/2 (5.2) ε

INSTANTANEOUS STRUCTURE

111

where D is the molecular diffusivity of the chemical in the fluid [2]. Equation (5.2) also shows that the Batchelor and Kolmogorov scales are related by the square root of the Schmidt number (Sc = ν/D). The majority of laboratory research on turbulent chemical plumes has occurred in water flows, while field research has occurred in both air and water environments. Plumes in these fluid environments are similar in many regards, but they differ in some important ways due to differences in the fluid properties. For instance, the kinematic viscosity of water is ν = 1 × 10−6 m2 /s at 20◦ C, whereas the kinematic viscosity of air is ν = 1.5 × 10−5 m2 /s at 20◦ C. Thus, even in the scenario of identical dissipation rates of turbulent kinetic energy, the Kolmogorov scale would differ between air and water flows. Similarly, the diffusivity of many chemical compounds in water is around 1 × 10−9 m2 /s, whereas the diffusivity of compounds in air is often around 1 × 10−5 m2 /s. This difference has significant effects on the turbulent mixing process because of differences in the Batchelor length scale. In water flows, the size of the fine-scale chemical concentration structure is roughly 35 times smaller than the Kolmogorov scale, whereas in air the fine-scale structure of the velocity and concentration fields is roughly the same size. 5.2


Figure 5.1 shows a photograph of a chemical plume in a turbulent water flow. The dye in this example has been released from a small orifice upstream of

Figure 5.1 Closeup, overhead view of a plume released isokinetically into a turbulent boundary layer in an open-channel flow. Flow direction is from left to right.

112


the photograph location. The photograph reveals the basic processes described in the previous section. The filament forms an irregular pattern that wanders randomly across the image, which is evidence of the turbulent stirring process. The concentration within the filament is clearly more dilute in some regions compared to others, which is consistent with molecular diffusion occurring at the boundary of the filament. The distribution of concentration consists of complex shapes that are continually evolving. A moment after this image was collected, the plume had a completely different appearance. The flow visualization image (Fig. 5.1) is an integrated view of the threedimensional plume structure projected onto the image plane. The image provides a good qualitative sense of the plume structure but lacks quantitative information. Webster et al. [3] employed the laser-induced fluorescence (LIF) technique to quantify the instantaneous concentration field. For the demonstration data shown herein, the flow was water in an open channel with a fully developed turbulent boundary layer on the bed. Fluorescent dye was released from a point source elevated 25 mm above the bed in the inertial region of the turbulent boundary layer. The velocity of the fluid released from the source nozzle matched the flow in the channel and is called an isokinetic release. The test section was illuminated by a laser sheet of monochromatic light with a wavelength in the absorption range of the dye. The dye was excited by the laser light and emitted light at a longer wavelength. A digital camera collected the emitted light, the intensity of which was proportional to the dye concentration and laser light intensity. Thus, after a calibration process, the intensity of the light recorded by each pixel yielded a concentration measurement in the field. For the data shown in this chapter, a onemegapixel digital camera effectively collected a million samples of concentration in each image. Figure 5.2 shows an example concentration field for the plane illuminated by the laser sheet. In this case, the horizontal plane is aligned with the source nozzle

Figure 5.2 See color plates. Sample instantaneous concentration field near the plume source. H is the channel depth and equals 20 cm. The contour values are normalized by the source concentration. (Figure adapted from data in Webster et al. [3].)


113

height. It is important to remember that the data correspond to a planar slice through the three-dimensional plume structure. The coordinate system employed defines x for the streamwise direction and y for the transverse direction with the source nozzle located at the origin. Note that the contour levels have exponential spacing to fully reveal the broad range of concentration levels present in the field. The concentration values are normalized by the source concentration and vary between zero and one. Near the source, the plume width was narrow, like the flow visualization image in Figure 5.1. The peak concentration values in this sample field are in the range of 30 to 40% of the source concentration. However, the highest contour shown is 10% of the source concentration because the regions of higher concentration are extremely small. The field appears to have randomly distributed islands of relatively high concentration. In reality the three-dimensional plume filament is continuously connected outside of the illuminated plane, and the measurement technique reveals only the intersections of the structure with the measurement plane. Keep in mind that this field is a snapshot at an arbitrary instant and that the concentration distribution changes continually. Also, note that the peak concentration values decrease significantly over the length of the field due to the turbulent mixing process. Figure 5.3 shows a sample concentration field at a location farther downstream (the x/H = 10 location corresponds to 2 m downstream of the source nozzle). The upper end of the concentration range decreased by more than an order of magnitude compared to the field in Figure 5.2 (note that the contour levels have changed). Also, the plume structure covers a much wider transverse distance. Thus, over a fairly short distance from the source, the concentration decreases and the plume size increases significantly. These attributes are consistent with the turbulent mixing process described in the previous section. A flow visualization image of this region of the plume appears to be a homogeneous cloud of dye [4]. However, the concentration measurements reveal that the

Figure 5.3 See color plates. Sample instantaneous concentration field. H is the channel depth and equals 20 cm. The contour values are normalized by the source concentration.

114


Figure 5.4 See color plates. Fine resolution perspective of a sample instantaneous concentration field. H is the channel depth and equals 20 cm. The contour values are normalized by the source concentration.

instantaneous field continues to have patchy structure with significant gradients of concentration. Molecular diffusion acts across strong gradients in the field to dilute the concentration of a filament. Figure 5.4 shows a high-resolution view of the plume to demonstrate the fine-scale concentration structure. The resolution of this measurement is close to the Batchelor scale, and hence the field resolves the finest structure in the concentration field. Conversely, Figures 5.2 and 5.3 have resolutions slightly coarser than the Kolmorogov scale and, thus, only reveal the large-scale characteristics of the plume. In the high-resolution image, the field appears as disconnected islands of high concentration. Again, the individual islands are continuously connected outside of the measurement plane. At the edge of these filaments, the concentration gradient is very steep. In this particular example, the concentration changes from a peak of roughly 10% of the source concentration to zero over a distance of half of a millimeter or less. Molecular diffusion transports very effectively across such steep gradients, which explains the rapid decrease in peak concentration shown in the large-scale fields.

TIME-AVERAGED CHARACTERISTICS

5.3

115


While the concentration fields help to describe the spatial structure of the plume, the instantaneous snapshots do not reveal the temporal variation. Figure 5.5 shows the first 50 s of a sample time record that was collected along the plume centerline. The time record was highly intermittent with low or zero concentration for long periods separated by extremely high spikes of concentration, which correspond to a concentration filament meandering over the sensor location. The time-averaged concentration for this record is much smaller than the concentration peaks due to the extended periods of zero concentration. Thus, the signal consists of a small time-averaged value with huge fluctuations. Since negative concentration values are impossible, the statistical record of concentration is a highly skewed distribution [4]. The time-averaged concentration field is calculated by ensemble averaging a long series of individual concentration fields, such as Figure 5.2. In this example, 6000 consecutive concentration fields were averaged to produce the time-averaged field shown in Figure 5.6. As surmised from the discussion of the time record, the magnitude of the time-averaged concentration is much smaller than the instantaneous values (a factor of 10 or more between Fig. 5.2 and 5.6). While the instantaneous concentration fields consist of random islands, the time-averaged concentration is a much smoother distribution. The distribution appears nearly symmetric about the plume centerline (y = 0). Because the plume was unbounded in the transverse direction, the transverse profiles possess this symmetry and, in fact, follow a Gaussian profile shape. Figure 5.7 shows the transverse timeaveraged concentration profiles at four distances from the source. The transverse coordinate is normalized by the square root of the second central moment of the profile (i.e., the standard deviation, σ ), which is a measure of the half-width of the time-averaged plume. The concentration is normalized by the value on the centerline for each profile. With this normalization, the profiles are self-similar, which means that they are coincident and have the same shape. The profiles also agree very well with the Gaussian profile shape shown with the solid line: y2 (5.3) c = ccenterline exp − 2 2σ where an overbar indicates a time-averaged quantity.

c/Co

0.2

0.1

0

0

10

20

30

40

50

Time (s)

Figure 5.5 Time record of concentration at a point along the plume centerline. (Adapted from data in Webster et al. [5].)

116


Figure 5.6 Time-averaged concentration field. The contour values are normalized by the source concentration.

centerline

Gaussian profile

Figure 5.7 Profiles of the time-averaged concentration at four downstream locations. The profiles are self-similar and the solid line corresponds to a Gaussian profile shape.

As described above, the plume becomes wider and more dilute as it evolves in the streamwise direction, thus ccenterline and σ are changing with x. The decrease of the time-averaged concentration along the centerline of the plume follows a x −1 profile for x/H > 2 (Fig. 5.8). This power law decrease agrees well with the time-averaged concentration field predicted by modeling efforts that assume


117

Figure 5.8 Time-averaged concentration along the plume centerline. Also shown is a power law curve.

constant eddy diffusivity [6]. Closer to the source, the data do not follow the x −1 line, rather the profile shows a greater rate of decrease. The change in behavior at x/H ≈ 2 resulted from a transition of the size of the plume compared to the size of the turbulent eddies in the flow [3]. While the time-averaged concentration is decreasing rapidly in the x direction, the magnitude of the fluctuations is decreasing even more rapidly. Figure 5.9 shows the standard deviation of the concentration fluctuations along the plume centerline. The profile shown only tells part of the story about the magnitude of the concentration fluctuations. At the source, the concentration fluctuations are zero due to the homogeneous nature of the plume at the release. As the filament begins to distort and mix, the concentration starts to fluctuate and the standard deviation increases rapidly. This increase occurs upstream of the measurement region shown. As the mixing process continues, the plume becomes more homogeneous and the peak concentration in the filaments decreases. The combined effect decreases the standard deviation. As shown, the profile follows a power law decrease (x −1.75 ). At x/H = 2 the standard deviation is roughly 2.5 times the value of the time-averaged concentration. Thus, at this location the time record consists of large fluctuations above a small time-averaged value as discussed above. The standard deviation decreases at a faster rate than the time-averaged concentration as indicated by the exponent of the power law. Thus, the plume

118


Figure 5.9 Standard deviation of the concentration fluctuations along the plume centerline. Also shown is a power law curve.

becomes homogeneous faster than the time-averaged concentration dilutes. The sample field shown in Figure 5.3 provides support for this conclusion because the field is more homogeneous compared to the field closer to the source shown in Figure 5.2.

5.4

INFORMATION FOR TRACKING CHEMICAL ODOR PLUMES

The task of locating the point source of a turbulent chemical plume is a difficult challenge. Even in the scenario that perfect sensors are available to detect the appropriate chemical compounds, using the acquired sensory information to rapidly locate the source with a high percentage of success is far from a trivial endeavor. Nevertheless, animals have developed an incredible ability to track chemical plumes to facilitate feeding and mating processes. In recent years, research has been performed to better understand the turbulent chemical plume tracking ability and strategies of moths [7], blue crabs [8, 9], lobsters [10, 11], and other animals (e.g. [12, 13]). The robust tracking ability of these organisms provides inspiration for engineers attempting to design systems to perform similar tasks [14, 15].


119

The tracking problem can be distilled into three distinct tasks: (1) making contact with the chemical plume, (2) using information in the chemical plume structure to move toward the source, and (3) determining when the source has been reached. Making contact with the plume (1) is essentially a search for a randomly located goal, and the effectiveness of various strategies is discussed in numerous publications (e.g., [16]). Realizing that the source has been reached (3) initially seems like a simple task, but it is actually quite difficult in practice and has received little research attention to date. The discussion in this section is focused on the second phase of the tracking problem: extracting information from the plume structure in order to move successfully toward the source. Animal performance indicates that they are effectively extracting useful information from the structure of chemical plumes. For instance, after contacting a turbulent plume similar to the example in the previous sections, blue crabs find the source via chemosensory tracking with a success rate of 80 to 90% from a distance of 1.5 m within a period of 10 to 30 s [17]. Thus, the basic question is: What information in the intermittent and chaotic plume structure is providing cues to guide tracking? The smooth and predictable gradient in the time-averaged concentration field seems useful for tracking at first glance (Fig. 5.6). A tracker would simply move up the concentration gradient by sequentially comparing the response of a single sensor to eventually locate the position of the highest concentration, which is coincident with the source. This strategy is called chemical klinotaxis, and small organisms, such as bacteria and nematodes, are believed to employ this strategy in laminar flow environments [16]. Unfortunately, this strategy is ineffective for rapid tracking of turbulent chemical plumes for the simple reason that the information shown in Figure 5.6 is unavailable. The time period necessary to determine the time-averaged concentration with sufficient accuracy to perceive the gradient direction is many minutes [4]. To follow the gradient in Figure 5.6, a tracker would have to collect chemosensory information for several minutes, compare to the time-averaged concentration previously collected to assess the gradient direction, and then move and start the process again. Figure 5.10 demonstrates that the convergence of the time-averaged concentration calculation takes roughly 10 min for the example plume. To generate this figure, the total time record (example segment shown in Fig. 5.5) was divided into numerous shorter periods, and the time-averaged value was calculated for each of the subperiods. For sample periods of 4.7 s, for instance, the time-averaged values vary by a huge range, roughly 200% larger to 100% smaller than the converged value. Thus, to accurately assess the concentration value compared to the concentration at the previous location, a much longer sampling period is needed. The data scatter is not solely a result of the number of samples that are collected, as one might expect from statistical considerations. Rather, it results directly from the intermittent nature of the concentration field. In other words, a sensor must sample over a time period sufficient for numerous filaments to pass by in order to determine a converged time-averaged value. Thus, the intermittent nature of

120


Figure 5.10 Time-averaged concentration as a function of sampling period at an arbitrary location in the flow. The total time record of 600 s was divided into numerous shorter periods to demonstrate the slow convergence of the time-averaged concentration calculation. (Adapted from data in Webster and Weissburg [4].)

the concentration field necessitates very long sampling periods in order to evaluate any time-averaged quantity, including the time-averaged concentration shown here. Obviously, a strategy that employs the gradient in a time-averaged quantity is extremely tedious. It is clear that blue crabs and other organisms are employing strategies that are much more efficient than waiting for time-averaged quantities to statistically converge. Thus, they appear to be acquiring information from the fluctuating plume structure. Several hypotheses have been suggested about the nature of the sensory information available in turbulent chemical plumes. For instance, Moore and Atema [18] reported that the rise slope of a concentration burst (see time record in Fig. 5.5) varies with distance from the source and hence provides ranging information during tracking. Webster and Weissburg [4] confirmed that the timeaveraged value of the rise slope changes systematically as a function of distance from the source. However, like the concentration value, the rise slope at any location varies as a random variable and a long period of sampling is required to achieve a converged time-averaged value. This is, in fact, the underlying problem with any cue that requires a time-averaged value: The convergence time in an intermittent plume is very long and the required monitoring prevents rapid tracking progress. The instantaneous concentration or rise slope at a particular location in the plume is any value within a broad range and hence does not provide predictable information. If sequential sampling seems incongruent with the performance of rapid trackers, then what other information is available? Several researchers have suggested


121

that animals are employing a bilateral comparison to adjust their transverse position in the plume [19–21]. Blue crabs, for instance, have chemosensors located on their leg appendages. Thus, by comparing the signal received by the left and right legs, the location of the plume edge may be deduced. Weissburg [22] refined this idea to include the fact that the sensor spacing (which he called the spatial integration factor, SIF) needs to be relatively large compared to the plume size. The plume structure data introduced in the previous sections provide the opportunity to assess whether multiple sensors that are spatially separated can rapidly acquire useful information [5]. The correlation function for the instantaneous concentration acquired by two sensors at locations p and q, separated in the transverse direction, is defined as cp cq =

1 T

T

c(y = p, t) × c(y = q, t) dt

(5.4)

0

where c is the instantaneous concentration, and T is the total period of the time record. Figure 5.11 shows the correlation function for four distances downstream of the source for the case of one sensor fixed to the plume centerline and the other transversely displaced. The correlation function is normalized by the value at y = 0 (i.e., when the sensors are coincident along the centerline), thus the maximum value of the normalized correlation function is one. With increased

Figure 5.11 Correlation function in the transverse direction at four downstream locations.

122


distance from the source, the profile becomes wider and the value of the correlation function increases. This trend results from turbulent mixing increasing the plume width and creating a more homogeneous field (see Figs 5.2 and 5.3). The integral length scale provides a simple measure of the plume width and homogeneity and is calculated based on the area under the correlation function profiles: ∞ c0 cy L= dy (5.5) c0 c0 0 Figure 5.12 shows the integral length scale as a function of distance from the source. As expected, the integral length scale increases in the downstream direction. Further, and more importantly, the integral length scale provides a measure of the plume width to which the chemical sensor separation can be compared. Figure 5.13 shows the correlation function for sensors that are separated by a fixed distance equal to L. The inner sensor moves throughout the field and its position is denoted by y, as shown in the sketch. The sensor location is normalized by the standard deviation of the time-averaged profiles (see Fig. 5.7). With this scaling, the profiles at the four downstream locations are coincident, which suggests that the integral length scale is the correct scaling length for the sensor separation and that the sensor location is properly scaled by the width of the time-averaged plume.

Figure 5.12

Integral length scale as a function of distance from the source.


123

Figure 5.13 Correlation function with sensor spacing equal to L in the transverse direction at four locations.

For the process of identifying the edge of the plume during tracking, the ideal situation is to have high contrast between the sensors, which corresponds to a low correlation function value. The correlation function shown in Figure 5.13 for sensor spacing equal to L is fairly high, approaching 0.3 at the plume centerline. By increasing the sensor spacing to four times the integral length scale, the value of the correlation function is greatly reduced (Fig. 5.14). Again, the profiles for the four locations are coincident, which reinforces the validity of the scaling. Thus, high contrast is perceived by the sensors, if the sensors are separated by a “large enough” distance. The measure of “large enough” for the turbulent plume is provided by comparing to the integral length scale. The definition of the correlation function includes the same time-averaging procedure that requires long sampling periods and hence is not directly useful for rapid tracking. The correlation function is useful, however, to define the plume width and appreciate that large spacing leads to contrast between the instantaneous signal received by the sensors. Essentially, if the sensor spacing is large, as measured relative to the integral length scale, then the correlation function is low, and instantaneous comparisons of the sensor signal values provides insight to the location of the plume centerline. In support of this conclusion, Grasso et al. [14] reported that sensor spacing affected path tortuousity of their autonomous lobsterlike robot tracker. Note that the integral length scale varies with distance from

124


Figure 5.14 Correlation function with sensor spacing equal to 4L in the transverse direction at four locations.

the source (Fig. 5.12), thus the ability to adjust the sensor spacing or compare among multiple sensors in an array is potentially advantageous. Spatial comparison appears to be useful to guide a tracker to the centerline of the plume and the information appears to be quickly available provided the sensor spacing is large. But, bilateral comparison alone is not enough to perform a successful, rapid search of a turbulent plume source. For instance, it is believed that blue crabs employ an odor-gated rheotaxis with bilateral comparison strategy. Rheotaxis means that they are sensing and using the flow direction to head upstream, and odor-gated means that they move upstream only when receiving chemosensory information of interest. Odor-gated rheotaxis alone does not explain the high performance of blue crabs because, without the ability to effectively correct their position relative to the plume centerline, the tracker frequently misses the source as it continues to move upstream. However, the added ability to steer and adjust their position relative to the plume centerline via a bilateral comparison improves tracking performance to be in accord with observed animal behavior [23]. Keller et al. [17] showed that upstream movement of blue crabs largely resulted from stimulation at the antennule chemosensors, while the leg chemosensors are largely responsible for the steering ability. Further, the simulations of Weissburg and Dusenbery [23] showed that tracking performance is critically affected by the balance between employing flow and chemical signals, which confirms the importance of both cues.

VARIATION OF THE PLUME STRUCTURE

125

The odor-gated rheotaxis with bilateral comparison strategy appears to be effective for rapid plume tracking in straight plumes and is congruent with blue crab performance. Of course, other strategies may be as effective, or perhaps more effective, for turbulent plume tracking. Indeed, other organisms employ alternate strategies due to constraints such as body size and sensor spacing. For example, moths cast in the cross-wind direction because they are much smaller than a typical plume width and hence cannot instantaneously compare sensory input across the plume [24]. In addition, the odor-gated rheotaxis with bilateral comparison strategy has not been tested as the flow conditions become more complex. 5.5

VARIATION OF THE PLUME STRUCTURE

The sample data presented in this chapter were collected for fairly simple flow conditions. The flow was a unidirectional open-channel flow without large-scale flow meander, and the release condition was isokinetic in the direction of the bulk flow. Thus, chemical filaments were advected by the bulk flow in the streamwise direction, while turbulent mixing acted to expand the plume size and dilute the chemical concentration. Changes in the flow and release conditions lead to significant variation in the plume characteristics and structure. Several laboratory studies have contributed to our understanding of turbulent chemical plumes and the effects of various flow configurations. Fackrell and Robins [25] released an isokinetic neutrally buoyant plume in a wind tunnel at elevated and bed-level locations. Bara et al. [26], Yee et al. [27], Crimaldi and Koseff [28], and Crimaldi et al. [29] studied plumes released in water channels from bed-level and elevated positions. Airborne plumes in atmospheric boundary layers also have been studied in the field by Murlis and Jones [30], Jones [31], Murlis [32], Hanna and Insley [33], Mylne [34, 35], and Yee et al. [36, 37]. In addition, aqueous plumes in coastal environments have been studied by Stacey et al. [38] and Fong and Stacey [39]. The combined information of these and other studies reveals that the plume structure is influenced by several factors including the bulk velocity, fluid environment, release conditions, bed conditions, flow meander, and surface waves. The release location influences the vertical distribution of the time-averaged concentration and fluctuations. For a bed-level release, vertical profiles of the time-averaged concentration are self-similar and agreed well with gradient diffusion theory [26]. In contrast, the vertical profiles for an elevated release have a peak value above the bed and are not self-similar because the distance from the source to the bed introduces a finite length scale [3, 25, 37]. Additionally, it is clear that the size and relative velocity of the chemical release affects both the mean and fluctuating concentration [4]. The orientation of the release also appears to influence the plume structure. The shape of the profiles of the standard deviation of the concentration fluctuations is different in the study of Crimaldi et al. [29] compared with those of Fackrell and Robins [25] and Bara et al. [26]. Crimaldi et al. [29] attributed the difference to the release orientation, which was vertically upward from a flush-mounted orifice at the bed in their study.

126


The Reynolds number is the ratio of inertial to viscous forces and depends on the fluid properties, bulk velocity, and boundary layer thickness. Turbulence characteristics vary with Reynolds number in boundary layers [40]. Thus, variation in the contributing factors for the Reynolds number ultimately influences the turbulent mixing and plume structure. Further, the fluid environment, air or water, affects both the Reynolds number and the molecular diffusivity of the chemical compounds. Bed conditions also influence the plume by altering the turbulence intensity and distribution in the boundary layer. Rahman and Webster [41] systematically varied the bed roughness to evaluate the effect of boundary layer characteristics on a plume from an elevated isokinetic chemical release. The average concentration was lower along the centerline for increasing bed roughness due to increased mixing and dilution. Similarly, the concentration fluctuation intensity was smaller and decreased faster for the rougher beds. Shifts in the flow direction lead to large-scale meander of the plume. The mixing and dispersion of chemicals in the plume are then dictated by meander of the plume centerline and mixing about the plume centerline [39]. Figure 5.15 shows two plumes in the wake of a circular obstacle and a plume for the same flow

Figure 5.15 Overhead view of plumes released isokinetically into a turbulent boundary layer with and without (top image) an obstacle upstream. Meander in the middle and bottom images was created by the wake of a small and large circular obstacle, respectively. Flow direction is from left to right.

REFERENCES

127

conditions without an upstream obstacle. The addition of an obstacle induced a Kármán vortex street that periodically advected the plume in different directions and resulted in the meandering plume shown. The flow pattern also modified the turbulence characteristics and hence affected the mixing process. The overall effect is a change in the shape of the centerline of the plume and a change in the mixing and dilution of the chemical field. Similarly, the imposition of surface waves on an open-channel flow, as expected in coastal flows, affects both the bulk velocity and the turbulence characteristics [42]. The full effect of periodic meander and wave perturbations on the plume structure is not currently known. Based on the discussed example parameters that influence the turbulent mixing process, it is clear that plume structure can be described in general terms, but the specific characteristics are likely to be case dependent. Nevertheless, certain characteristics, such as those employed by the odor-gated rheotaxis with bilateral comparison strategy, may be similar enough to allow animals and engineered systems to track chemical plumes for a wide range of flow conditions. ACKNOWLEDGMENTS

The research conducted in support of this chapter was financially supported by ONR/DARPA (N00014-98-1-0776) and NSF (IBN-0321444). Special thanks to S. Rahman and L.P. Dasi for data collection and analysis and to M.J. Weissburg and C.B. Woodson for helpful discussion. REFERENCES 1. Kolmogorov, A. N. The local structure of turbulence in incompressible viscous fluid for very large Reynolds numbers. Dolk. Akad. Nauk SSSR 30, 301 (1941). 2. Batchelor, G. K. Small-scale variation of convected quantities like temperature in turbulent fluid. Part 1. General discussion and the case of small conductivity. J. Fluid Mech. 5, 113–133 (1959). 3. Webster, D. R., S. Rahman, and L. P. Dasi. Laser-induced fluorescence measurements of a turbulent plume. J. Engng. Mech. 129, 1130–1137 (2003). 4. Webster, D. R. and M. J. Weissburg. Chemosensory guidance cues in a turbulent chemical odor plume. Limnol. Oceanogr. 46, 1034–1047 (2001). 5. Webster, D. R., S. Rahman, and L. P. Dasi. On the usefulness of bilateral comparison to tracking turbulent chemical odor plumes. Limnol. Oceanogr. 46, 1048–1053 (2001). 6. Rahman, S. Effect of Bed Roughness on Scalar Mixing in Turbulent Boundary Layers. Ph.D. Thesis, Georgia Institute of Technology, 2002. 7. Mafra-Neto, A. and R. T. Cardé. Fine-scale structure of pheromone plumes modulates upwind orientation of flying moths. Nature 369, 142–144 (1994). 8. Weissburg, M. J. and R. K. Zimmer-Faust. Life and death in moving fluids: Hydrodynamic effects on chemosensory-mediated predation. Ecology 74, 1428–1443 (1993). 9. Weissburg, M. J. and R. K. Zimmer-Faust. Odor plumes and how blue crabs use them in finding prey. J. Exp. Biol. 197, 349–375 (1994).

128


10. Moore, P. A., N. Scholz, and J. Atema. Chemical orientation of lobsters, Homarus americanus, in turbulent odor plumes. J. Chem. Ecol. 17, 1293–1307 (1991). 11. Koehl, M. A. R., J. R. Koseff, J. P. Crimaldi, M. G. McCay, T. Cooper, M. B. Wiley, and P. A. Moore, Lobster sniffing: Antennule design and hydrodynamic filtering of information in an odor plume. Science 294, 1948–1951 (2001). 12. Moore, P. A. and D. M. E. Lepper. Role of chemical signals in the orientation behavior of the sea star Asterias forbesi. Biol. Bull. 192, 410–417 (1997). 13. Ferner, M. C. and M. J. Weissburg. Slow-moving predatory gastropods track prey odors in fast and turbulent flow. J. Exp. Biol. 208, 809–819 (2005). 14. Grasso, F. W., T. R. Consi, D. C. Mountain, and J. Atema. Biomimetic robot lobster performs chemo-orientation in turbulence using a pair of spatially separated sensors: Progress and challenges. Robot. Auton. Syst. 30, 115–131 (2000). 15. Hayes, A. T., A. Martinoli, and R. M. Goodman. Swarm robotic odor localization: Off-line optimization and validation with real robots, Robotica 21, 427–441 (2003). 16. Dusenbery, D. B. Sensory Ecology, W.H. Freeman, New York, 1992. 17. Keller, T. A., I. Powell, and M. J. Weissburg. Role of olfactory appendages in chemically mediated orientation of blue crabs. Mar. Ecol. Prog. Ser. 261, 217–231 (2003). 18. Moore, P. A. and J. Atema. Spatial information contained in three-dimensional fine structure of an aquatic odor plume. Biol. Bull. 181, 408–418 (1991). 19. Reeder, P. B. and B. W. Ache. Chemotaxis in the Florida spiny lobster, Panulirus argus. Anim. Behav. 28, 831–839 (1980). 20. Zimmer-Faust, R. K., C. M. Finelli, N. D. Pentcheff, and D. Wethey. Odor plumes and animal navigation in turbulent water flows: A field study. Biol. Bull. 188, 111–116 (1995). 21. Atema, J. Eddy chemotaxis and odor landscapes: Exploration of nature with animal sensors. Biol. Bull. 191, 129–138 (1996). 22. Weissburg, M. J. The fluid dynamical context of chemosensory mediated behavior. Biol. Bull. 198, 188–202 (2000). 23. Weissburg, M. J. and D. B. Dusenbery. Behavioral observations and computer simulations of blue crab movement to a chemical source in a controlled turbulent flow. J. Exp. Biol. 205, 3387–3398 (2002). 24. Vickers, N. J. and T. C. Baker. Reiterative responses to single strands of odor promote sustained upwind flight and odor source location by moths. Proc. Natl. Acad. Sci. USA 91, 5756–5760 (1994). 25. Fackrell, J. E. and A. G. Robins. Concentration fluctuations and fluxes in plumes from point sources in a turbulent boundary layer. J. Fluid Mech. 117, 1–26 (1982). 26. Bara, B. M., D. J. Wilson, and B. W. Zelt. Concentration fluctuation profiles from a water channel simulation of a ground level release. Atmos. Environ. 26A, 1053–1062 (1992). 27. Yee, E., D. J. Wilson, and B. W. Zelt. Probability-distributions of concentration fluctuations of a weakly diffusive passive plume in turbulent boundary-layer. Bound.-Lay. Meteorol. 64, 321–354 (1993). 28. Crimaldi, J. P. and J. R. Koseff. High resolution measurements of the spatial and temporal scalar structure of a turbulent plume. Exp. Fluids 31, 90–102 (2001).

REFERENCES

129

29. Crimaldi, J. P., M. B. Wiley, and J. R. Koseff. The relationship between mean and instantaneous structure in turbulent passive scalar plumes. J. Turbul. 3(14), 1–24 (2002). 30. Murlis, J. and C. D. Jones. Fine-scale structure of odour plumes in relation to insect orientation to distant pheromone and other attractant sources. Physiol. Entomol. 6, 71–86 (1981). 31. Jones, C. D. On the structure of instantaneous plumes in the atmosphere. J. Hazard. Mater. 7, 87–112 (1983). 32. Murlis, J. The structure of odour plumes, in T. L. Payne, M. C. Birch, and C. E. J. Kennedy, Eds. Mechanisms in Insect Olfaction. Clarendon Press, Oxford, 1986, pp. 27–38. 33. Hanna, S. R. and E. M. Insley. Time series analysis of concentration and wind fluctuations. Bound.-Lay. Meteorol. 47, 131–147 (1989). 34. Mylne, K. R. Concentration fluctuation measurements in a plume dispersing in a stable surface layer. Bound.-Lay. Meteorol. 60, 15–48 (1992). 35. Mylne, K. R. The vertical profile of concentration fluctuations in near-surface plumes. Bound.-Lay. Meteorol. 65, 111–136 (1993). 36. Yee, E., R. Chan, P. R. Kosteniuk, G. M. Chandler, C. A. Biltoft, and J. F. Bowers. Experimental measurements of concentration fluctuations and scales in a dispersing plume in the atmospheric surface layer obtained using a very fast response concentration detector. J. Appl. Meteorol. 33, 996–1016 (1994). 37. Yee, E., R. Chan, P. R. Kosteniuk, G. M. Chandler, C. A. Biltoft, and J. F. Bowers. The vertical structure of concentration fluctuation statistics in plumes dispersion in the atmospheric surface layer. Bound.-Lay. Meteorol. 76, 41–67 (1995). 38. Stacey, M. T., E. A. Cowen, T. M. Powell, E. Dobbins, S. G. Monismith, and J. R. Koseff. Plume dispersion in a stratified, near-coastal flow: Measurements and modeling. Cont. Shelf Res. 20, 637–663 (2000). 39. Fong, D. A. and M. T. Stacey. Horizontal dispersion of a near-bed coastal plume. J. Fluid Mech. 489, 239–267 (2003). 40. DeGraaff, D. B. and J. K. Eaton. Reynolds-number scaling of the flat-plate turbulent boundary layer. J. Fluid Mech. 422, 319–346 (2000). 41. Rahman, S. and D. R. Webster. The effect of bed roughness on scalar fluctuations in turbulent boundary layers. Exp. Fluids 38, 372–384 (2005). 42. Mead, K. S., M. B. Wiley, M. A. R. Koehl, and J. R. Koseff. Fine-scale patterns of odor encounter by the antennules of mantis shrimp tracking turbulent plumes in wave-affected and unidirectional flow. J. Exp. Biol. 206, 181–193 (2003).

PART II

FIELD EXPERIENCE

CHAPTER 6

DETECTION OF TRACE EXPLOSIVE SIGNATURES IN THE MARINE ENVIRONMENT MARK FISHER AND MATTHEW DOCK Nomadics, Inc., an ICx Technologies Company

6.1

INTRODUCTION

Underwater explosive devices typically possess a combination of detectable expressions, including sonar, electromagnetic, and chemical. While significant progress has been made in the areas of advanced high-resolution sonar, synthetic aperture subbottom sonar, magnetic anomaly detection, and electrical anomaly detection, the area of chemical detection of explosives is still in an early stage of development. The characteristics of explosives submerged in water have been studied, the dynamics of the underwater chemical plumes have been analyzed, and now the sensors capable of detecting underwater explosives have been developed. Many of these advances in underwater chemical sensing have been a result of the Chemical Sensing in the Marine Environment (CSME) program sponsored by the Office of Naval Research (ONR). A goal of this program was to develop new tools for detecting underwater unexploded ordnance (UUXO). During the program, a number of trinitrotoluene (TNT) sensors were developed and evaluated by investigators. The explosive contained in UUXO is a potential source for generation of a chemical signature that could be exploited as a means of determining the presence and location of these devices. Because of the low solubility of explosives such as TNT in water, the concentrations of explosive in the water near UUXO are likely to be very low (possibly orders of magnitude lower than the solubility limit of the explosive in water). Once the explosive chemical signature is released into the water, it rapidly mixes into the water column. If there is a prevailing underwater current (which is frequently the case), the plume of explosive is dispersed away from the source, primarily through convective (turbulent) transport processes. This dispersion results in additional dilution of the signature as it is Trace Chemical Sensing of Explosives, Edited by Ronald L. Woodfin c 2007 John Wiley & Sons, Inc. Copyright

133

134

DETECTION OF TRACE EXPLOSIVE SIGNATURES IN THE MARINE ENVIRONMENT

transported away from the source. If there is a prevailing current, dispersion of the plume will be highly directional, which makes it necessary to position the sensor downcurrent from the source in order to facilitate detection. The challenges in detection are further compounded by the large variety of naturally occurring chemical, biological, and particulate matter encountered in the marine environment. These materials represent potential interferents that further complicate the detection process. Hence, sensors with exquisite sensitivity and selectivity are required for this sensing application. This chapter discusses the development and testing of the Nomadics SeaDog underwater explosives detection system, developed through funding from ONR during the CSME program. This system is an adaptation of Nomadics Fido vapor sensing technology originally developed for the detection of landmines (described in Chapters 7 and 9). The sensor utilizes a proprietary amplifying fluorescent polymer (AFP) technology developed by collaborators at the Massachusetts Institute of Technology (MIT). The fluorescence of AFP is greatly reduced when exposed to target explosive compounds. In addition, the response of the polymer is selective, responding to a very limited subset of chemical constituents found in the environment. The selectivity and sensitivity of the sensor allow the AFP to detect the chemical signature emanating from underwater UXO in real time and without the use of preconcentrators. To further increase the utility of the sensor for underwater sensing applications, the sensor has been deployed on a number of autonomous platforms, enabling real-time detection of underwater explosive plumes. An overview of the sensor platform is presented, along with results of tests conducted at sites in the Gulf of Mexico and the Pacific and Atlantic oceans. 6.2 OVERVIEW OF FATE AND TRANSPORT OF EXPLOSIVES RELEASED FROM UUXO

During World War II, copious quantities of ordnance were lost into the harbor at Halifax, Nova Scotia. Decades later, these UUXO now present a significant environmental contamination problem. Studies conducted on this ordnance by Sandia National Laboratories [1] suggest that there may be sufficient concentrations of explosive chemical signature compounds emanating from UUXO to enable detection with chemical sensors. Some UUXO in Halifax Harbor have been shown to produce parts-per-billion levels of explosives in the water near the ordnance. In addition to the parent explosive compound (TNT), other explosiverelated compounds such as 2,4-dinitrotoluene (2,4-DNT) were detected, as were degradation products of TNT such as 4-amino-2,6-dinitrotoluene (4-ADNT), and 2-amino-4,6-dinitrotoluene (2-ADNT). In addition to the contamination detected in the water column, contamination was more frequently detected in sediment from the seafloor near UUXO. It was also noted that the casings of some of the munitions had corroded, exposing the explosive fill material directly to seawater. In some rounds that had corroded, the explosive fill had completely dissolved.

SAMPLING AND SENSING METHODOLOGY

135

In another study, Darrach et al. [2] reported that samples collected near intact UUXO targets contained traces of explosives at up to parts per billion (ppb) concentration levels. The samples were analyzed in the laboratory, using solid-phase microextraction (SPME) to extract target analytes from the samples. The samples were then processed using a reversal electron attachment detection (READ) technique. If the levels of contamination found in these studies are representative of that emanating from most UUXO, the implication is that sensitive chemical sensors such as the SeaDog may be useful for detecting UUXO. Another important consideration in chemical detection of explosives in the marine environment is the nature of the chemical plume emanating from UUXO. Work completed during the CSME program has shown that chemical signatures can persist for long distances away from a source because of the relatively low rates of molecular diffusion of explosives in the aqueous environment. A number of advanced algorithms for predicting how chemical signatures are mixed and transported through the water were also developed during this program. Results from these predictive models were consistent with quantitative experimental data collected from plumes of fluorescent dyes released into the ocean. Plumes of fluorescent dye were shown to persist at detectable levels (using sensitive fluorometers) for over 1000 m from the source. These studies also pointed out that there is great variability in the plume structure. Most simple models of plume dynamics assume a Gaussian distribution of explosive concentrations across and downcurrent in the plume. However, this study reveals that most underwater plumes will be very heterogeneous in structure, consisting of dispersed “filaments” of explosive at high concentrations relative to the remainder of the water (which contains little to no explosive) within the plume. These filaments may contain concentrations of explosive significantly higher than the concentrations expected assuming a Gaussian distribution of the explosive. 6.3

SAMPLING AND SENSING METHODOLOGY

As described in Section 6.2, it has been shown that convective transport is the primary mechanism for dispersal of explosive signatures in water, resulting in a plume of small filaments of dissolved explosive emanating away from the source. Hence, the challenge is to search a large volume of water for small, widely dispersed plume filaments containing dissolved explosive material. This places extreme demands on sensor technology used to detect underwater chemical plumes. The technology must be highly sensitive and selective. In addition, it must be amenable to deployment in seawater. Finally, the sensor must also respond rapidly to the analyte and quickly reset after a response. The SeaDog sensor utilized in this work is capable of near real-time detection of low concentrations of explosives in water. The sensor utilizes novel sensing materials originally developed by collaborators at MIT. These materials are fluorescent polymers that are highly emissive when deployed as solid-state thin films. When the polymers interact with nitroaromatic explosives such as TNT, the fluorescence is quenched [3–5]. The response of these materials to target analytes

136


is amplified because a single molecule of explosive binding anywhere along a single chain of polymer quenches the emission of multiple polymer repeat units, each of which is a fluorophore [6]. The quenching effect is not confined to the polymer chain to which a molecule of explosive is bound. Rather, the quenching effect is delocalized over multiple polymer chains, resulting in quenching of a three-dimensional volume of polymer comprising multiple fluorophores. The sensor responds to other nitroaromatic species often found in the environment near UUXO, such as 2,4-DNT. In addition, the sensor also responds to TNT breakdown products such as trinitrobenzene (a photochemical degradation product of TNT), and amino-dinitrotoluenes (microbial degradation products). The response of the sensor to these compounds is rapid and reversible, enabling the sensor to quickly respond to multiple exposures to target analytes. The underwater sensor platform is derived from the Fido explosives vapor sensor, originally developed under the Defense Advanced Research Projects Agency (DARPA) Dog’s Nose Program. The vapor sensor, whose operation is discussed in Chapters 7 and 9 and in other publications [7–9], was developed for the task of landmine detection. The underwater adaptation of the sensor is very similar to the vapor sensor. In the underwater implementation of the sensor, thin films of polymers are deposited onto glass or sapphire substrates. The emission intensity of these films is monitored as water (rather than air) flows past the substrate. If the concentration of TNT in the water beings to rise, the polymer will exhibit a measurable reduction in fluorescence intensity. The reduction in emission intensity is proportional to the concentration of target analyte in the water. Because the sensor is small, lightweight, and consumes little power, it proved to be ideal for deployment on autonomous platforms. The initial challenge during development of the sensor was to determine the pros and cons of aqueous versus vapor-based detection architectures. Because the partitioning of explosive into the polymer films from the vapor phase is much greater than from the aqueous phase, the sensor response to a given concentration of TNT vapor is orders of magnitude larger than to the same concentration of TNT in the aqueous phase. Hence, to maximize sensor response, the original approach was to develop methods of extracting explosive from the water, and to then vaporize the extracted sample into a stream of carrier gas (air) that would transport the explosive vapor into the sensor for analysis. Extraction of explosive from the water proved to be slow (on the order of 10 s to minutes depending on the method) and inefficient, with only a small portion of the explosive being extracted from the sample in a reasonable length of time. Hence, the sensitivity advantage obtained by analyzing in the vapor phase was largely (but not completely) offset by poor extraction efficiencies from the aqueous phase. This unfortunate result had further ramifications. One of the goals of the CSME program was to develop sensors that could be deployed on autonomous platforms for the purpose of tracing explosives plumes to their source. The autonomous vehicle utilized in the CSME program was not designed to hover at a fixed position. The vehicle was designed to move forward continuously through the water, which is advantageous for interrogating a large underwater area in a short time.

SEADOG SENSOR CONFIGURATIONS

137

However, because the vehicle could not pause or even move slowly through the water, sensors deployed on the vehicle were forced to respond to the presence of explosive very rapidly in order to provide accurate, high-resolution concentration maps of the plume. Assuming that sample collection, analyte extraction, and analysis is a serial process, data points can at best be generated at a rate equal to the total time required to perform each of these three steps. Because the vehicle is moving during all steps of the sensor duty cycle, the length of time required to achieve all three steps must be short or the resolution of the concentration map will be compromised. Since the extraction procedure described in the previous paragraph was a relatively slow process, it would not have been possible to achieve high-resolution concentration mapping using this method. Hence, the method of vapor-phase detection after extraction from samples was abandoned in favor of direct detection from water, which can be accomplished in near real time. Direct sensing from seawater was also attractive because the hardware required to sense explosives directly in water was much less complex than the hardware required to extract explosive and analyze it in the vapor phase. 6.4 6.4.1

SEADOG SENSOR CONFIGURATIONS Prototype Integrated with a Robotic Crawler Platform

A crude schematic of the first prototype sensor design is shown in Figure 6.1. The prototype consisted of two watertight boxes housing the sensor hardware. The first box contained the sensing head and associated hardware to facilitate sample (seawater) passage through the head. Inside the sensing head is a cylindrical glass waveguide coated on one end with a thin film of AFP. The AFP film is illuminated with light from a laser diode source that causes the polymer to fluoresce. The light emitted from the film is internally reflected down the length of the cylindrical waveguide. The emitted light exits out the opposite (uncoated) end of the waveguide, passes through an 80/20 mirror, and then impinges onto an optical filter that blocks passage of stray light from the laser excitation source, but allows light emitted from the AFP to pass. The light passing through the filter impinges on a photomultiplier tube, enabling the intensity of the light emitted from the polymer to be quantified. Water is pumped through the sensor head and across the AFP film. When explosives and related compounds such as 2,4-DNT or amino-DNTs are present in the water, the intensity of the emission from the AFP film is dramatically reduced in proportion to the concentration of target analyte in the sample. The change in emission intensity is measured with a photomultiplier tube (PMT). The output from the PMT is processed via an onboard digital signal processor (DSP). The processed signal is then transferred to a handheld computer housed in one of the watertight boxes. The handheld computer controls sensor functions and allows logging of the data stream generated by the sensor. The data can be downloaded for postprocessing after a mission. In addition, the handheld computer generates a graphical display of sensor response that can be viewed in real time through the enclosure by a diver if so desired.

138


Figure 6.1 Schematic of the first SeaDog sensor design.

The second box contains a peristaltic pump and a servoactuator. Both the pump and servo are controlled electrically from the sensing head and are powered from the same power supply. The separation of the sensing head from the pump and servo provide electrical and mechanical isolation and address space constraints associated with mounting the system on the autonomous underwater vehicle. The peristaltic pump enables operation at a variable flow rate and has bidirectional flow capability. The servo actuates a movable sample inlet tube that can be raised or lowered by remote control to enable precise positioning of the inlet relative to the source or in the source plume. A photograph of the entire system, mounted on a metal plate that attaches to the robot, is shown in Figure 6.2. Initial testing of the sensor was successfully performed in our laboratory prior to integration with the vehicle. Integration of the sensor with a Foster-Miller Talon was then performed in Panama City, Florida, at Coastal Systems Station (CSS). Figure 6.3 illustrates the sensor assembly, consisting of the two boxes and the robot cover plate, attached to the robotic crawler. In this configuration, data from the SeaDog is transferred to the robot control computer via a dedicated RS-232 link. The sensor and crawler were connected to a towed float that contained system batteries and communications hardware to transmit sensor and crawler data to and from remote computers. A video camera was also integrated with the system to enable visual monitoring of the underwater tests. After initial testing of system integration in the robotics lab at CSS, the system was taken to a sandy beach area for mobility and integration trials. Other than


139

Figure 6.2 SeaDog sensor and pump housings mounted for robotic integration.

Figure 6.3

SeaDog mounted on the Foster-Miller crawler during testing.

a minor communications problem that was rapidly diagnosed and corrected, the integration trials progressed smoothly. The system was then tested for basic operation underwater. After moving the crawler and sensor into the water, the pump system was activated and the operation of the servo-controlled inlet was confirmed. These tests were successful, enabling sea trials of the system. The results of these trials are presented in Section 6.5.1. 6.4.2

Diver-Deployed SeaDog and Initial Integration with the REMUS

After successful testing of the sensor prototype described in Section 6.4.1, the prototype sensor was reconfigured for mounting onto an autonomous underwater

140


Figure 6.4

SeaDog deployed on the REMUS during a sea trial.

vehicle (AUV). The platform selected for integration was the Remote Environmental Monitoring UnitS (REMUS) platform developed by Woods Hole Oceanographic Institute (WHOI) (Fig. 6.4). This adaptable platform is capable of carrying a variety of payloads including sonar and sensors of various types. The initial sensor prototype described in Section 6.4.1 was modified to fit into a sealed enclosure that was designed to interface with the REMUS vehicle. The enclosure contained a pump and an electronics package very similar to the prototype sensor. Minor changes were made to the optical setup. The cylindrical waveguide was replaced with a circular sapphire plate coated with AFP on its front face. The plate was positioned inside a sample chamber through which seawater was pumped at a rate of 180 mL per minute. The sensor module also contained a fluorometer designed to detect marker dyes that were introduced into the water during some missions. The dye plume gave a visual indication as to the location of the plume of explosives emanating from the source. The integrated fluorometer was able to detect the marker dye, signaling when the AUV passed through the plume. If the plume also contained explosives, the SeaDog and fluorometer would respond simultaneously, confirming the sensor response to explosives. An optical transmittance sensor was included in the package to measure turbidity of the water. A high-resolution temperature sensor was also onboard to monitor the temperature of the water during testing. Finally, a generic REMUS-compatible communications protocol was developed through collaboration with WHOI and the Space and Naval Warfare Systems Center (SPAWAR). This generic protocol enabled communication between sensors in the payload with the onboard computer in the REMUS. All data collected during a mission was logged by the system and was downloaded to an external computer for analysis after completion of a mission. The SeaDog sensor payload as deployed on the REMUS is illustrated in Figure 6.5. The REMUS sensor suite was also configured for diver deployment (Fig. 6.6). The sensor module was easily detachable from the REMUS and could be mounted


Figure 6.5

141

SeaDog sensor payload configured for deployment on the REMUS.

Figure 6.6

SeaDog sensor package configured for diver deployment.

onto a plate equipped with a handle for the diver to grasp. A transparent box was also attached to the plate to house the control system, a handheld computer, and a battery to power the system. The data collected during a diver-deployed mission could be stored to the handheld computer, or viewed in real time on the graphical interface of the handheld computer. Audio output was also added to the system for use during diver deployment. The audio output consisted of a

142


series of “clicks” whose frequency was proportional to the response of the sensor. The audio output was audible underwater, enabling feedback of sensor response to the diver during turbid or low-light operating conditions when the graphical interface was not visible. Test results for the diver-deployed and initial REMUS-deployed sensor tests are presented in Section 6.5.2. 6.4.3

SeaDog Miniaturization: The SeaPup

Lessons learned during sea trials with the SeaDog led to development of the SeaPup sensor, which is a miniaturized version of the SeaDog sensor. In addition to being significantly smaller than its predecessor (see Fig. 6.7), the SeaPup was more sensitive (by a factor of approximately 6) than the SeaDog. More importantly, the new design exhibited greatly enhanced response kinetics, making it possible to better resolve the structure of plumes. The accelerated rate of response was due largely to a redesign of the sensor sample cell. The new proprietary design achieved these gains in performance by optimizing the mass transport of target analyte from the sample volume to the AFP-coated substrate. The fluorometer and optical absorbance sensors were eliminated from this design, further reducing the size of the sensor. The new sensor design made it possible for the first time to map a plume of explosive emanating from a source in the marine environment. The results for sea trials using the SeaPup sensor design are presented in Section 6.5.3.

Figure 6.7

Comparison of the size of the SeaDog (left) to the SeaPup (right).

RESULTS OF SENSOR TESTS CONDUCTED IN THE MARINE ENVIRONMENT

143

6.5 RESULTS OF SENSOR TESTS CONDUCTED IN THE MARINE ENVIRONMENT 6.5.1

Tests of the Sensor Prototype on a Crawler Vehicle

The first tests of the prototype sensor described in Section 6.4.1 were conducted during February of 2002 in conjunction with CSS. During these tests, the prototype sensor mounted on a Foster-Miller crawler robot (provided by CSS) was used to detect traces of explosive emanating from underwater targets. Figure 6.8 is a photo of the test site, illustrating the sensor on the robot, and the towed float (being carried by the two test personnel located in the center of the image) housing batteries and communications hardware. This photo was taken just prior to deployment of the system for testing. All targets in these tests were secured to a post located in 4 ft of water. Targets were suspended from the post by a rope so that the targets were approximately 1 ft from the seafloor. For initial tests, a target was constructed using a commercially available TNT simulant. The simulant consists of TNT coated onto particles of sand, but the quantity of TNT in the simulant is insufficient to sustain a detonation. This target was placed inside a water-permeable fabric bag and was affixed to the pole as previously described. For the initial test, the crawler approached the target from a direction upcurrent of the prevailing current so that any chemical signature emanating from the target would be transported away from the sensor. No sensor response was noted as the sensor approached the target from upcurrent. The robot was then repositioned by executing a number of turns, placing the robot and sensor approximately 3 ft downcurrent from the target. Shortly after arriving in this position, two sensor responses (seen in Fig. 6.9 near the 230 time position) were observed. These responses were consistent with that of TNT in the water. Note that the response to TNT was rapid and reversible.

Figure 6.8

Sensor/crawler system prior to deployment at the test site.

144


Figure 6.9 Response to TNT 3 ft downcurrent of target.

Immediately after these responses were noted, the robot was positioned closer to the target, approaching to a distance of approximately one foot downcurrent from the source. The response at this position is indicated in Figure 6.10. A comparison of Figures 6.9 and 6.10 reveal that the frequency of encounters between TNT-containing plume filaments increases closer to the target. This is consistent with the assumption that the distance between plume filaments containing detectable concentrations of TNT increases farther from the target as the plume filaments become more widely dispersed. Also note that while the number of responses per unit time is greater near the target, the magnitude of the sensor response to each filament is similar at both distances investigated. This result suggests that while the plume filaments become progressively farther apart, the concentration of explosive within the plume filaments does not rapidly decrease with distance from the source.

Figure 6.10 Sensor response 1 ft downcurrent of target.


145

To confirm that the response was to TNT and not a chemical interferent in the water, the sensor was moved to numerous locations around the target that were upcurrent from the prevailing current near the target. No responses consistent with the presence of TNT were observed unless the sensor was downcurrent from the target. Water samples were also collected at the site and were later analyzed in our lab for the presence of TNT. The samples collected were extracted into toluene and analyzed by gas chromatography (GC) using an electron capture detector (ECD) and by a second chromatographic system using a Fido vapor sensor as the detector. The results from both analytical systems confirmed the presence of TNT in the samples collected near the target. The concentration of TNT in samples collected from 1 to 3 ft from the target ranged from 1 to 20 ppb by mass (as measured by GC-ECD). On the final day of testing a TNT demolition charge was substituted for the test target constructed from the TNT simulant. The block of explosive was prepared in a manner consistent with how the block would normally be deployed for use. The block was then deployed from the post as previously described. The sensor was positioned at the downcurrent side of the post at a distance of approximately 3 ft. The sensor immediately indicated a strong response to the explosive. Initial tests of the Nomadics SeaDog sensor integrated with a Foster-Miller crawler were extremely promising. The sensor was readily integrated with the crawler, enabling tests to be conducted seamlessly. The sensor was able to detect plumes of TNT emanating from test targets. This was accomplished without sample preconcentration and in real time. A more detailed account of this test is presented by Dock and co-workers [10, 11]. 6.5.2 Tests of the Diver-Deployed SeaDog Sensor and Initial Integration to the REMUS

The SeaDog sensor was tested in September 2002 at a test site in the Pacific Ocean. During this test, the SeaDog sensor module was tested in the diverdeployed configuration described in Section 6.4.2. The diver-deployed sensor module successfully demonstrated detection of TNT near an underwater target during these tests. After completion of the diver-deployed portion of the testing, the sensor module was then reconfigured for operation on the REMUS. Due to time constraints (the REMUS was being utilized for multiple tasks during this test), it was not possible to test the SeaDog integrated with the REMUS in the marine environment. However, the sensor was integrated to the REMUS and was tested on the benchtop. The system integration was successful except for a minor software problem that made it impossible to download data from the system computer after a mission. This software issue was resolved prior to a second series of tests conducted at the same site 2 months later (November 2002). During the second round of tests conducted in November of 2002, detection of a plume of TNT at a significant standoff distance was demonstrated for the first time. During these tests, a TNT source was constructed and deployed in the ocean at a depth of approximately 30 m. Using the diver-deployed SeaDog module (shown in Fig. 6.11), divers were able to detect a plume of TNT at distances of up to 30 m

146


Figure 6.11

Diver-deployed SeaDog sensor during sea trials.

from the source. For the first 6 min of data collection (the upper data trace in Fig. 6.12), the diver was located upcurrent from the TNT source. As can be seen from the upper data trace, no sensor responses were noted, as would be expected when the sensor is located upcurrent from the source. No responses suggestive of chemical interferents in the water were observed. However, numerous responses were observed when the sensor was positioned downcurrent of the source in the TNT plume. The lower data trace in Figure 6.12 was collected at distances of up to 30 m downcurrent from the source as the diver transected the plume numerous times with the sensor. Large responses to the TNT plume were obtained. During the second round of testing, it was possible to deploy the sensor on the REMUS. Figure 6.13 is a photograph of the system while operational in the marine environment during the test. While no positive indications of TNT were observed during the test, the sensor was successfully operated on the REMUS in the marine environment. Lessons learned during this deployment led to development of the SeaPup sensor, which was constructed and tested in early 2003. 6.5.3

Tests of the SeaPup Sensor Integrated on the REMUS

In June 2003, the SeaPup sensor was tested at a site off the Atlantic coast. As was discussed in Section 6.4.3, the SeaPup sensor showed almost an order of magnitude improvement in sensitivity over the SeaDog due to design enhancements incorporated into the system. In addition, the SeaPup responds much more rapidly to TNT than the SeaDog, which is an important advantage for mapping chemical plumes in the marine environment.


147

Figure 6.12 Data collected with diver-deployed sensor during sea trials.

Figure 6.13 SeaDog deployed on the REMUS AUV during sea trials.

In these tests the SeaPup sensor was mounted on the REMUS AUV and commanded to interrogate an area greater than 22,000 m2 . The REMUS was programmed to execute a back-and-forth “mow-the-lawn” style of mapping mission perpendicular to the direction of the prevailing current near the target. As the REMUS followed the programmed mission, the time-domain TNT concentration data is integrated with the longitude–latitude positioning data to generate a concentration map of the area. Figure 6.14 is a concentration map generated near one of the test targets. The TNT plume emanating from the target is clearly visible in the map, with the concentration of TNT decreasing with distance from

148


Figure 6.14 See color plates. Detection of TNT plume with the SeaPup sensor mounted on the REMUS AUV off the Atlantic Coast of the U.S. in June 2003.

the source. The plume was detected at over 100 m from the source. To our knowledge, the SeaPup is the only sensor to demonstrate real-time detection of an explosive plume from an AUV in the marine environment from any distance. While not demonstrated during this work, another interesting capability of the REMUS AUV system is that a sensitive, real-time explosive sensor could supply explosive concentration information to an adaptive mission planner that is capable of modifying the course of the AUV, so that the AUV tracks the explosive plume to its source. This capability, if further developed, could enable detection of UUXO by tracking explosive plumes to source. During these tests, the SeaPup sensor had a third-party verified, in-field TNT sensitivity of 4 ppb. While it is generally believed that a sensitivity of 4 ppb is not adequate for detecting all UUXO, the results of the studies of UUXO in Halifax Harbor [1, 2] suggest that this level of sensitivity will enable detection of some UUXO items. Further improvements in sensitivity of detectors may soon make detection of UUXO possible with chemical sensors, providing an orthogonal detection capability for detection methods such as sonar. 6.6

CONCLUSIONS

To our knowledge, this is the first demonstration of a sensor capable of realtime detection of a TNT plume in the marine environment at standoff distances (up to 100 m from the source) while deployed on an autonomous underwater vehicle. The sensor has shown virtually no sensitivity to chemical interferent during testing in the marine environment. While the sensitivity of the detector is excellent, its sensitivity is not adequate at its present state of development to

REFERENCES

149

detect all types of underwater unexploded ordnance. However, based on published accounts of the concentration of explosive signature compounds in the environment near some UUXO, the system should have adequate sensitivity to detect some UUXO, especially if the UXO has been underwater for decades. While not discussed here, new AFPs are being developed for enhanced detection of RDX and other explosives not detectable with the sensors at the time they were tested. With further development, chemical sensor systems such as the SeaDog could be a valuable tool for detection of UUXO, especially when deployed on autonomous underwater platforms. ACKNOWLEDGMENTS

We would like thank Dr. Keith Ward and Dr. Angela Ervin, formerly at the Office of Naval Research (ONR) for sponsoring this work, and Don Robeson, Rudy Arrieta, and their colleagues at USN Coastal Systems Station (Panama City) for their efforts in the planning and conduct of the trials using their test assets. We also wish to thank the Woods Hole Oceanographic Institute (WHOI) and the Space and Naval Warfare Systems Center (SPAWAR) for technical support during integration of the SeaDog sensor with the REMUS, and for support during sea trials. REFERENCES 1. Rodacy, P., P. K. Walker, S. D. Reber, J. Phelan and J. V. Andre. Chemical Sensing of Explosive Targets in the Bedford Basin, Halifax, Nova Scotia, Sandia National Laboratories, SAND2001-3569, November 20, 2001. 2. Darrach, M. R., A. Chutjian, and G. A. Plett. Trace explosives signatures from World War II unexploded undersea ordnance. Jet Propulsion Laboratory, Environmental Science Technology 32(9), 1354 (1998). 3. Yang, J. S. and T. M. Swager. Porous shape persistent fluorescent polymer films: An approach to TNT sensory materials. J. Am. Chem. Soc. 120, 5321–5322 (1998). 4. Yang, J. S. and T. M. Swager. Fluorescent porous polymer films as TNT chemosensors: Electronic and structural effects. J. Am. Chem. Soc. 120, 11864–11873 (1998). 5. Williams, V. and T. M. Swager. Iptycene-containing poly(aryleneethynylene)s. Macromolecules 33, 4069–4073 (2000). 6. Zhou, Q. and T. Swager. Methodology for enhancing the sensitivity of fluorescent chemosensors: Energy migration in conjugated polymers. J. Am. Chem Soc. 117, 7017–7018 (1995). 7. Cumming, C., C. Aker, M. Fisher, M. Fox, M. la Grone, D. Reust, M. Rockley, T. Swager, E. Towers, and V. Williams. Using novel fluorescent polymers as sensory materials for above-ground sensing of chemical signature compounds emanating from buried landmines. IEEE Trans. Geosci. Remote Sensing 39(6), 1119–1128 (2001). 8. Fisher, M. and C. Cumming. Detection of trace concentrations of vapor phase nitroaromatic explosives by fluorescence quenching of novel polymer materials, in Proceedings of 7th International Symposium on the Analysis and Detection of Explosives, Defense Evaluation and Research Agency, Edinburgh, Scotland, UK, June, 2001.

150


9. Fisher, M., M. la Grone, and J. Sikes. Implementation of serial amplifying fluorescent polymer arrays for enhanced chemical vapor sensing of landmines, in Proceedings of UXO/Countermine Forum 2003, Orlando, Florida, September, 2003. 10. Dock, M., M. Fisher, and C. Cumming. Novel detection apparatus for locating underwater unexploded ordnance, in Proceedings of the 5th International Symposium on Technology and the Mine Problem, Mine Warfare Association, Monterrey, California, April 2002. 11. Dock, M., M. Fisher, and C. Cumming. Sensor for real-time detection of underwater unexploded ordnance, in Proceedings of UXO/Countermine Forum 2002, September 2002, Orlando, Florida.

CHAPTER 7

EXPLOSIVES DETECTION USING ULTRASENSITIVE ELECTRONIC VAPOR SENSORS: FIELD EXPERIENCE MARK FISHER Nomadics, Inc., an ICx Technologies Company

7.1

INTRODUCTION

This chapter summarizes almost a decade of field experience using trace-level ® electronic vapor sensors (the Nomadics Fido ) to detect explosives and explosive devices. This sensor was originally developed as part of the Defense Advanced Research Projects Agency (DARPA) Dog’s Nose Program. The objective of this program was to develop advanced sensor technology for the purpose of detection of low-metal-content (plastic) antipersonnel landmines that are difficult to detect with conventional landmine detection systems. Since the completion of this program, the sensor has undergone further development for landmine detection applications with funding from the U.S. Army Night Vision and Electronic Sensors Directorate (NVESD). Experience gained as a result of field testing conducted as part of these programs will be presented. In addition to landmine detection, the sensor has been evaluated and found to have utility for other explosive sensing applications, but due to the sensitive nature of much of this work it cannot be presented here. Significant effort has been devoted to development of extremely sensitive trace detection systems capable of detecting a variety of explosive threats. While sensitive detectors are required to detect these threats, our experience and that of others working in this area have shown that a sensitive detector is not alone sufficient to ensure detection of targets containing explosives. The methods by which trace samples are collected and presented to the sensor are critically important, and often determine whether or not a target will be detected. Sample acquisition is often mistakenly regarded as trivial by comparison to sensor development, and is hence largely ignored during sensor design. To neglect sampling issues during development of explosives detection systems virtually guarantees that the system Trace Chemical Sensing of Explosives, Edited by Ronald L. Woodfin c 2007 John Wiley & Sons, Inc. Copyright

151

152

ULTRASENSITIVE ELECTRONIC VAPOR SENSORS, FIELD EXPERIENCE

will be of little value operationally, regardless of the sensitivity of the detector. Our experience has been that a majority of the difficulties encountered during field trials have not been related to sensor hardware, but rather to sampling issues. Hence, issues related to sampling will be presented throughout this chapter where relevant. Having a basic understanding of the nature of the explosive chemical signature emanating from a given target is also critical to achieving high probabilities of detection of explosive devices. The parameter most often considered is the concentration (or mass flux) of explosive emanating from the target. Generally, the sensitivity of a detector can be measured in a laboratory setting and will have been established prior to use in the field. Similarly, the rate of release of explosive chemical signature constituents emanating from an explosive-containing target can be measured in the controlled environment of a laboratory. However, when this target is translated from a laboratory to a field environment, the influence of environmental factors can significantly alter the rate of release of explosive from the target. Moreover, the difficulty in measuring the explosive “source term” for targets in the field is compounded by the fact that environmental factors affecting the rate of release of explosive can change rapidly. Further, there is no “standard” explosive device. For example, the vapor signature produced for two apparently identical landmines can vary significantly, even though they were both manufactured to the same specifications. Subtle differences in how the devices are deployed can also have an impact on detection. An understanding of how the vapor signature evolves after it is released into the environment is also important. For example, the concentration of explosive vapor released into the boundary layer of air in contact with an explosive device may be quite large, approaching the equilibrium concentration value of the explosive contained within the device. However, the concentration of the explosive signature can diminish by orders of magnitude over a relatively short distance from the target. Hence, the signature becomes progressively more difficult to detect as the distance from the target increases. If the detection application requires tracing a “vapor plume” to its source from a distance, this factor can be a significant limitation. In addition, environmental influences can degrade the chemical constituents into other substances, and if the detector being utilized cannot detect these degradation products, detection probabilities will be compromised. The “fate and transport” of explosives in operational environments is a complex subject that has not been extensively studied for many detection applications and should be a focal point of ongoing explosives detection research. As will be discussed, explosives fate and transport issues and the environmental factors that influence them play a key role in designing sampling strategies, and also must be considered during development of training materials used to teach sensor operators how to effectively use a sensor system. The emphasis of this chapter will be on vapor-phase detection of explosives. While the data and results presented here were obtained using the Nomadics Fido sensor, many of the lessons learned are applicable to other trace sensor systems and also to detection using canines.

RELEVANCE OF FIELD TESTING TO SENSOR DEVELOPMENT

7.2

153

RELEVANCE OF FIELD TESTING TO SENSOR DEVELOPMENT

In 1997, DARPA launched the Dog’s Nose Program. The program brought together researchers from a variety of science and engineering disciplines to address the problem of landmine detection. In this program canines, regarded by many as the “gold standard” in mine detection, were used as a model detection system from which to base sensor design and performance. A critical component of this program was field testing. DARPA maintained a test minefield for the purpose of testing sensor performance. All systems were tested at the facility at least twice a year. Some investigators were not involved directly in sensor development, but fieldwork was still a component of most of their statements of work. For example, experts in canine olfaction studied the canine olfactory system and the performance of mine detection dogs. Some of this work was done in a laboratory, but in the latter stages of the program, canines were taken to the field for testing alongside of sensors in order to evaluate and compare their performance under field conditions. There was a widely held belief prior to this program that chemical vapor sensors could never achieve the level of sensitivity required to detect landmines in the same manner as canines. This assertion was shown to be false, as an electronic vapor sensor developed during this program (the Nomadics Fido) demonstrated for the first time detection of landmines at performance levels comparable to experienced mine detection canines during blind field trials. The motivation provided by mandatory biannual field testing, coupled with lessons learned in the field, greatly accelerated development of the sensor, driving the achievement of canine-comparable detection capability with less than 2 years of sensor development [1]. It should also be noted that the work done by other researchers was instrumental in reaching the objective of detection of landmines with electronic sensors. In particular, the results of researchers investigating the mechanisms of explosive release from mines into the soil, and how these explosive “chemical signatures” are modified as they are transported through the soil were very useful [2–4]. These factors strongly influence probabilities of detection because explosive traces must be released from the mine and then transported to the surface of the ground where they can be sampled and detected. This is true whether the sensor is electronic or a canine. An important outcome of this work was the identification of key chemical constituents of the mine chemical signature and their concentrations, as well as how the signature is influenced by environmental factors [5–8]. While some of these experiments were performed in laboratories, much of this data was collected from test minefields that differed from real minefields in only one respect—the landmines were fitted with inert detonators (or a storage plug that sealed the detonator well) to enable testing to be conducted without accidental detonation of mines. From this data, sensor developers gained important insights into which explosive signature constituents were most useful as target analytes and how to effectively sample these signatures. It is understood that new technology must be validated through laboratory and field testing before it can be fielded operationally. However, sensor developers

154


are often reluctant to subject new technology to field testing until the technology is regarded as reasonably mature. It is true that the rigors of testing a sensor in the field will likely accentuate known weaknesses in the design, but field testing will also reveal issues that would likely never manifest in a laboratory environment. It is often much easier to correct weaknesses in a system in the early stages of development while the design is still evolving rather than to address them late in the design process. An additional benefit of field testing that is often overlooked is that testing is often conducted with potential end users of the equipment present, many of which have considerable operational experience. These individuals are uniquely qualified to provide expert feedback on the sensor design and potential concepts of operation. Sensor developers rarely have operational experience using the equipment they are developing. Hence, a common problem that arises is that the equipment, while functional, is not always practical to use in the field. This problem could be largely avoided if developers and end users are both involved in the design cycle. Involvement of end users in the design process increases the probability that the equipment will be suitable for operational deployment. Field exercises provide an excellent opportunity for this to occur. In fact, the feedback from experienced operators and the lessons learned from the field tests have driven many of the system enhancements that have been incorporated into the sensor. 7.3

OVERVIEW OF THE VAPOR SIGNATURES OF EXPLOSIVES

The Fido sensor [9–13], described in detail in Chapter 9, is an extremely sensitive detector for trinitrotoluene (TNT) and explosives formulations containing TNT such as Composition-B (a mixture of 60% RDX and 40% TNT). The polymeric sensing material used in the sensor was originally developed for detection of nitroaromatic explosives (TNT) [14–17], but through additional development collaborators at the Massachusetts Institute of Technology (MIT) have developed new sensing materials that possess excellent sensitivity to RDX and pentaerythritol tetranitrate (PETN). These explosives are utilized in many types of plastic explosives including Semtex, C-4, and PE-4. TNT and plastic explosives are widely utilized in military and civilian applications. These explosives are also commonly used by terrorists. When a source of explosive is not readily available, terrorists are more frequently turning to the use of improvised explosives, including triacetone triperoxide (TATP) or hexamethylene triperoxide diamine (HMTD). These explosives can be synthesized from readily available materials. Hence, they are being used frequently in terrorist attacks. The explosives mentioned here are only a few selected from a long list of materials that can be used as explosives. This presents an unusual detection challenge. The chemical and physical properties of explosives vary widely, so it is a challenge to design a sensor that can detect all explosives equally well. One such property is the equilibrium vapor pressure of explosives. From Figure 7.1, which is a plot of the equilibrium vapor pressures of selected explosives at 25◦ C,


155

1.00E+00 EGDN

Equilibrium Vapor Pressure

1.00E−01

NG

TNT

RDX

PETN

1.00E−02 1.00E−03 1.00E−04 1.00E−05 1.00E−06 1.00E−07 1.00E−08 1.00E−09 1.00E−10

Figure 7.1 Equilibrium vapor pressures of common explosives (in Torr) at 25◦ C.

it can be seen that the vapor pressures of these common explosives span 9 orders of magnitude [18]. The equilibrium vapor concentration of an explosive at a given temperature represents an upper limit to the concentration of explosive vapor available for detection. Many factors serve to lower the vapor-phase concentration of explosive near an explosive device to less than the equilibrium value. First, the flux rate of explosive emanating from the device may not be adequate to maintain equilibrium concentrations in the boundary layer of air at the surface of the device. Regardless of the concentration of explosive in the boundary layer, the vapor is rapidly dispersed into the ambient air near the device, further lowering the concentration of analyte. In practice, the average concentration of explosive vapor emanating from an explosive device is orders of magnitude lower than the equilibrium value, and the average concentration at a given point in space decreases rapidly as the distance from the device increases. Hence, detection of explosives in the vapor phase requires extremely sensitive detectors and/or the use of sample preconcentration methods. It is often assumed that transport of explosive vapor away from a source results in a Gaussian-like distribution of explosive downwind of the source. This is approximately correct if one is considering the average concentration of explosive at a given point in the plume as a function of time. However, the instantaneous concentration of explosive in the plume at a given position in space away from the source can range from zero to values approaching the source concentration. Convective dispersion (due to turbulent flow of ambient air) can quickly transport vapors a significant distance from their source. When this occurs, the vapor plume, which may be fairly homogeneous very near the source, is fragmented into smaller and smaller filaments by the turbulent flow. These filaments eventually

156


1.005

Normalized Response

1 0.995 0.99 0.985 0.98 0.975 0.97 0.965 0.96 0.955 0

50

100

150

200

250

Elapsed Time (s)

Figure 7.2

Response of Fido to a vapor plume 2 m downwind of a TNT source.

disperse due to the action of molecular diffusion. However, molecular diffusion is a slow transport process compared to turbulent convection, so before the plume filaments disperse via molecular diffusion, they can be transported a significant distance from the source while retaining a significant fraction of the source concentration. Molecular diffusion eventually becomes the dominant mode of dispersion, but this does not occur until the plume filaments become quite small. The heterogeneous nature of a TNT vapor plume is presented in Figure 7.2. This figure illustrates the response of Fido to a vapor plume encountered during an outdoor test of the sensor at an arid test facility. The source for the plume was three 1-lb demolition blocks of TNT. The blocks were stacked on the ground 2 m upwind of the sensor. The TNT source and the detector remained stationary throughout the sampling event. Wind during sampling was light (less than 2 m/s) and variable. Fido responded strongly when plume filaments were drawn into the sensor (at between 120 and 180 s into the sample), but a vast majority of the time the sensor did not respond because it was sampling relatively “clean” air. A good visual example of this type of plume behavior can be observed by watching smoke from a cigarette. Near the cigarette, the smoke is very concentrated and reasonably homogeneous in composition, but the smoke plume quickly fragments into filaments that can persist for a significant distance from the cigarette. The concentration of smoke in the filaments is large relative to the concentration in the air surrounding the filament. An important point should be made here. If the average concentration of TNT in the air over time were computed at the point in space where the data in Figure 7.2 was collected, the average concentration would be at or below the minimum detection limits of the detector. However, the sensor was able to detect explosive because it was not measuring the average concentration but the instantaneous (real-time) concentration of explosive at that point in space, which at times is significantly greater than the average concentration.


157

The heterogeneous nature of vapor plumes has several important implications for sampling and sensing. First, plume filaments containing concentrations of explosive slightly less than the concentration at the source can persist a significant distance from the source. This enables the possibility of standoff detection, as demonstrated by the data shown in Figure 7.2. However, the greater the distance from the source the filaments are sampled, the smaller the concentration of explosive in the filaments, and the number of plume filaments per unit volume of air also decreases as the distance from the source increases. This means that the greater the distance between the point of sampling and the source, the larger the volume of air that must be processed in order to ensure that the plume filaments have been sampled. Except for laboratory reference materials that have been rigorously purified, explosives are not pure materials. For example, TNT contains synthesis byproducts that are the result of incomplete nitration of toluene [19]. Depending on the synthesis process and how much effort was made to purify the explosive, mono- and di-substituted nitrotoluenes may be present in TNT at levels ranging from a trace up to a few percent by mass. This has important implications for detection because both mono-and dinitrotoluenes (DNT) have higher vapor pressures than TNT. Hence, the vapor-phase concentrations of these contaminants may be higher than that of TNT near an explosive device utilizing TNT as the explosive. Plastic explosives are also mixtures containing one or more explosives mixed with a variety of other compounds, including materials that may be more volatile than the explosive in the mixture. If these higher volatility compounds are unique to explosives, they could be used as a means of identifying the explosive. However, if the compound in question is widely used in a variety of nonexplosive products, the sensor could be prone to false alarms and would therefore be of little utility. The composition of the vapor signature of a given type of explosive can be quite variable. For example, the relative abundance of explosive signature constituents for a given type of explosive can vary significantly depending on the source of the explosive [4]. For example, TNT from different sources can contain a range of concentrations of 2,4-DNT, which could affect probabilities of detection if 2,4-DNT were the preferred analyte for detection. In addition, the composition of explosives can change with age. Degradation of the explosive into other compounds does occur, and escape of more volatile constituents from an explosive can occur with age, changing the chemical vapor signature of the explosive over time. Environmental factors can also change the chemical signature of an explosive. For example, in soils near landmines TNT can be quickly converted to amino-dinitrotoluenes by the action of soil microbes, altering the chemical signature of the mine [3]. Ideally, any compound found in an explosive that is unique to the explosive, and is always found in the vapor signature of the explosive regardless of its source or age could be used as a means of detecting the explosive. Unfortunately, these requirements severely limit the potential analyte set for consistent, reliable detection. One could argue that the only compound that can be guaranteed to be present in an explosive headspace are vapors of the

158


actual explosive, and as such should be the target analyte for a vapor detector. However, if certain relatively uncommon signature compounds were consistently present in the headspace of a given type of explosive, and these compounds could be reliably detected when present, target analytes other than the parent explosive could be used to detect certain explosive devices, especially when the parent explosive has a very low vapor pressure. 7.4 7.4.1

LANDMINE DETECTION Introduction to the Mine Problem

The worldwide landmine problem is multifaceted, encompassing a myriad of humanitarian, political, and technical challenges. As a result of decades of armed conflicts around the world, it is estimated that 60 to 70 million mines are deployed in approximately 70 countries. It has been estimated by the International Committee of the Red Cross that a person is maimed or killed by a mine every 20 min [20]. In addition to the physical and mental trauma inflicted by landmines, landmines restrict access to land and critical natural resources, which can have a disastrous effect on the economy of mined areas. Clearing minefields is also expensive, making it difficult for the governments of most countries affected by mines to subsidize a significant fraction of needed demining operations. Demining operations are largely subsidized through humanitarian organizations and governments located outside mine-affected areas. In general, the level of funding has not been adequate nor consistent enough to clear minefields in a timely manner. Hence, minefields often remain long after the conflict that resulted in the mines being emplaced has ended. The landmine problem is also not without technical challenges. In particular, modern plastic-cased antipersonnel mines with very low metal content are much more difficult to detect than metal-cased mines, which can routinely be detected with a metal detector or ground-penetrating radar. Hence, new technologies with enhanced capability for detecting plastic-cased mines is needed. Acoustic, infrared, and spectroscopic methods including laser-induced breakdown spectroscopy (LIBS) are examples of technology that have demonstrated detection of plastic-cased mines with limited success. In addition to new technologies, significant improvements have recently been made in decades-old mine detection technology (metal detection and ground-penetrating radar in particular). Nevertheless, widespread implementation of the enhanced technology in demining operations has been limited due to the high cost and complexity of fielding the new equipment. In spite of recent advances in detector technology, metal detectors, mine probes, and canines remain as the most widely utilized mine detection tools. All of these methods have been utilized with some success, but all have limitations. For example, metal detectors must be operated at extremely high gain settings to obtain adequate sensitivity to detect low-metal-content mines. This is an issue because minefields are often littered with metallic objects (e.g., fragments from exploded munitions). Hence, false alarms are common, limiting the utility

LANDMINE DETECTION

159

of metal detectors in these environments. “Probing,” a technique that involves “feeling” for mines by methodically inserting rigid rods into the soil, is also utilized. When a solid object is encountered in the soil, it is then excavated to determine if the object is a mine. This is an extremely dangerous and time-consuming method of detecting landmines. Canines have also been used for landmine detection for decades [21]. Even though canines can be very effective for this purpose, their use is not without problems [22]. Logistical problems can be significant. Dogs and handlers are expensive to train and maintain. Dogs (and handlers) do not perform well under all environmental conditions. In addition, the performance of the dog–handler team can vary from day to day, and performance at any given instant in time can be difficult to verify. A single solution to the problem does not exist. Hence, new mine detection technologies are desperately needed. In an effort to address the problem of detection of low-metal-content mines, DARPA funded development of new sensor technologies that focused on detection of the single material unique to all landmines—the explosive. The Fido sensor was one of several explosives vapor detectors developed under this program. Unless specified otherwise, the data presented in this section was collected during field tests of this sensor. 7.4.2

Discussion of Landmine Chemical Vapor Signatures

This section is a brief literature review of laboratory results obtained by a number of investigators studying landmine chemical signatures. These results are relevant to a discussion of field experiences because they have proven to be extremely beneficial in understanding the performance of the sensor and sampling system under field conditions. In addition, many of the characteristics of landmine signatures documented in the laboratory investigations have been observed under field conditions, adding a measure of validity to both the field and lab results. 7.4.2.1 Chemical Composition of Landmine Signatures Almost 80% of the types of mines manufactured worldwide contain TNT, or mixtures of explosives containing TNT such as Composition-B [23]. Approximately 85% of the number of mines now deployed contain at least some TNT as a constituent of its explosive charge. Of these mines, approximately 86% contain at least 50 g of TNT [24]. Because of the relatively low volatility of TNT, it would take many decades for 50 g of TNT to completely sublime. Hence, the quantity of TNT found in most landmines is sufficient to generate a chemical vapor signature that can be released into the soil for decades. Like canines, the Fido sensor is an extremely sensitive detector for TNT, making it ideally suited for detection of landmines. As discussed previously, military-grade TNT is not a pure material, containing additional nitroaromatic constituents at concentrations of up to several percent by mass [4]. Some of these explosive-related compounds (ERCs) have been shown to be significant contributors to the chemical fingerprint of a landmine [5, 7]. Of the ERCs found in TNT, those that are most prevalent in the headspace vapor of TNT include 2,4-DNT, 2,6-DNT, 1,3-dinitrobenzene (1,3-DNB), and 1,3,5-trinitrobenzene (1,3,5-TNB) [5].

160


Many of the trace contaminants found in TNT have higher equilibrium vapor concentrations than TNT, which has an equilibrium vapor concentration of 70 pg/mL of air at 298 K [18]. For example, analysis of TNT from a PMA1A1 . (Soviet Block antipersonnel landmine) found the headspace concentration of 2,4-DNT to be 20 times that of TNT even though the 2,4-DNT accounted for less than 1% of the explosive by mass [5]. 2,4-DNT is also more stable in the environment than TNT [3]. It has also been shown that 2,4-DNT is the compound most likely detected by canines trained to detect TNT [25, 26]. For these reasons, some investigators have singled out 2,4-DNT as the analyte of choice for vapor phase sensing of landmines. However, our laboratory analysis (gas chromatography using electron-capture detection) of acetonitrile extracts of soil samples collected from minefields around the world has shown that soils found to contain detectable levels of TNT do not always contain detectable levels of 2,4-DNT. As will be discussed in the next section, explosive and ERC-contaminated soils near mines provide the source material for production of the landmine chemical signature that can be detected by canines and the Fido sensor. Hence, we conclude that the best approach is not to focus on detection of a single analyte but on detection of both TNT and ERCs derived from it. Our field results with Fido support this conclusion. The sensor can readily differentiate 2,4-DNT responses from responses to TNT. Depending on field conditions, responses consistent with that of TNT are sometimes more prevalent, while at other times 2,4-DNT appears to be the more prevalent analyte. In addition to synthesis by-products, the environment in the vicinity of a landmine may contain additional nitroaromatic constituents that are derived from other nitroaromatic species as a consequence of chemical degradation. In the environment, nitroaromatic compounds are subject to various forms of degradation. For example, in addition to being a contaminant in TNT, 1,3,5-TNB trinitrobenzene is a photochemical degradation product of TNT. 2-Amino-4,6dinitrotoluene, 2-ADNT, and 4-amino-2,6-dinitrotolune, 4ADNT, are microbial degradation products of TNT [3]. The amino-DNTs are frequently found in the soil over landmines, particularly during warm weather when microbial activity is high. None of these products are common in the environment. Fido is capable of detecting all of these compounds. Hence, detection of any of these compounds could be used to signal the presence of a landmine. 7.4.2.2 Factors Affecting the Release of Explosive into the Soil There are multiple sources for release of explosive signatures into the soil from a mine. The first source is surface contamination that is on the outside of the mine casing at the time of emplacement. Mine casings are likely to be heavily contaminated on their outer surfaces at the time of burial. This contamination could have been placed on the exterior of the mine at the time of manufacture, or could be due to contamination from other sources during storage or handling of other explosives or explosive devices. This source of contamination is likely to be 1

Details of the mines mentioned are included in [23].

LANDMINE DETECTION

161

quickly dispersed into the soil matrix surrounding the mine once it is buried. This type of surface contamination on a plastic-cased Yugoslavian PMA-1A AP mine was measured and reported in the literature [5]. The surface of the mine was contaminated with 1,3-DNB, 2,4-DNT, and TNT. The total mass of these analytes distributed over the entire surface of the mine was 6.5 μg. This mass of analyte is insufficient to sustain long-term release of signature compounds into the soil. This level of contamination is, however, possibly sufficient to facilitate detection of recently buried mines for a short time after their burial until this source is depleted. The remaining mechanism for explosive release into the environment is via vapor phase diffusion of analyte through mine structural materials and by leakage through cracks, seams, and holes in the mine. Because of the low vapor pressure of the explosives, and because the quantity of explosive in most mines is large, these are both long-term release mechanisms. The total vapor-phase flux of signature compounds from several types of landmine cases was also measured as part of the mine surface contamination study described in the previous paragraph [5]. The flux in this case was the total flux due to diffusion through structural materials and leakage through cracks and holes. For the PMA-1A landmine encased in polyvinyl chloride (PVC), the vapor-phase flux of 2,4-DNT into air was 3.4 μg per mine per day at 296 K, while the flux for the less volatile TNT was only 0.3 μg per mine per day. Flux rates were also measured into water. The flux rate of 2,4-DNT from a polystyrene-cased PMA2 AP mine into water was a factor of 30 larger than into air, and for TNT was a factor of 400 larger. It was estimated that the flux rate into wet soil would be intermediate between the air and water values. This has important implications for sensing mines since soil moisture content will likely influence the rate of release of ERC from mines. The vapor-phase flux through a mine case also varies depending on the type of material the mine is constructed from and the structural design of the mine [5, 7]. Depending on the type of plastic used, flux rates from plastic-cased mines are typically higher than from metal-cased mines. For a metal-cased Yugoslavian TMM-1 Soviet Block antitank AT mine, the flux rate for 2,4-DNT was 2.3 μg per mine per day at 296 K. This compares with 3.4 μg per mine per day for the plastic-cased PMA-1A whose total surface area is approximately 6.5 times smaller than that of the TMM-1. Release of ERC through intact metal-cased mines is presumably through seams, seals, and other nonmetallic structural materials. 7.4.2.3 Fate and Transport of Signature Compounds in Soil In order for vapor detectors to detect buried landmines, there must be vapors of explosive or ERCs present in the air over mines. The source for this vapor is from soils at the surface of the ground that have become contaminated with explosives. This implies that once explosive is released from a mine casing into the soil, there are mechanisms that facilitate transport of the explosive from the surface of the mine casing through the soil, and onto the soil at the surface of the ground. The fate and transport of ERC released into the soil from buried landmines has been the focus of a number of recent experimental and theoretical studies [2, 5–7].

162


These studies have resulted in a better understanding of the conditions that are necessary to produce vapor concentrations of ERC at the surface of the ground sufficient for mine detection using our sensor. Figure 7.3 illustrates some of the processes that occur in the soil near a buried landmine. Signature compounds escape from the mine casing and are quickly adsorbed onto soil solids or dissolve into soil water. Because explosives and ERCs have a high affinity for sorption onto surfaces [27], a large fraction of these compounds quickly sorb onto soil solid phases. The remainder of the explosive partitions into either soil liquid or gas phases. The fraction of the signature partitioning into these three soil phases is primarily a function of soil moisture content, followed by temperature and soil type [5, 7]. For a typical soil, approximately 95% of the total mass of ERC is adsorbed onto soil solids, followed by approximately 5% into soil water, with a trace (approximately 1 × 10−6 %) partitioning into the vapor phase [2]. Hence, the concentration of ERCs in the soil is orders of magnitude greater than the concentration of ERCs in the vapor phase. Some of the ERC sorbed by soil solid phases is irreversibly bound, preventing it from reaching the surface of the ground. Some of the material is degraded by the action of microbes in the soil [3]. The half-life of TNT in certain soils when the conditions for microbial degradation are favorable is short, on the order of a few days [28], while 2,4-DNT is much more stable in the environment. Some of these degradation products (namely amino-dinitrotoluenes, ADNTs) are detectable by Fido, but some interact more strongly with the soil than the parent materials, further reducing the concentration of signature available for detection. Explosive-related compounds are transported to the surface of the soil primarily through the movement of water in the soil. Molecular diffusion plays only a

Chemical Detection System Precipitation

nsp potr a

Microbial / Chemical / Physical Transformation or Loss

Eva

ion

fus

Dif

iratio

n

Photodegradation

n

sio

fu Dif

Liquid Phase

Vapor Phase

Soil Solid Phases

Land Mine

Figure 7.3 Processes affecting the fate and transport of landmine signature compounds in soil.

LANDMINE DETECTION

163

minor role in movement of ERCs through the soil column [2, 6]. During periods of precipitation, water tends to carry the signature downward through the soil, or transports the signature at the surface of the ground away from the mine in surface runoff. When precipitation ends, water evaporating from the surface of the ground tends to transport contaminated subsurface water upward toward the surface of the ground (evapotranspiration). As the water evaporates, it deposits signature compounds onto soil at the surface of the ground. This mechanism also describes transport of low-volatility pesticides through soil [29]. The chemical properties of pesticides and herbicides are similar to explosives, so it is not surprising that the same transport mechanisms are involved. Once the explosive signature molecules are deposited on surface soils, they can escape into the boundary layer of air near the ground over the mine. The concentration of ERCs released into the air depends on the solid-to-vapor partition coefficient, which is again primarily a function of soil moisture content, followed by temperature and soil type. Explosives sorbed on surface soils are also subject to photochemical degradation when exposed to sunlight. Some of the photodegradation products (such as 1,3,5-TNB) are Fido detectable. Due to the complexity of the vapor signature of military-grade TNT, and because of the microbial and photochemical degradation products that result from degradation of signature compounds, it is not surprising that 20 different ERCs have been detected in surface soil samples collected over landmines [6, 7]. Most of these 20 compounds are infrequently detected in land-mine-contaminated soils. Four of these compounds (2,4-DNT, TNT, 4-amino-2,6-DNT, and 2-amino-4,6-DNT) are found much more frequently than the other compounds. Soil moisture content is an important parameter affecting the concentration of ERCs in the headspace over explosive-contaminated soils. Lab experiments have shown that for a given soil contaminated with ERC, the concentration of ERC in the headspace over the soil can span five orders of magnitude as a function of soil moisture content [7]. Hence, the water content of surface soils is a very important consideration when sampling vapor-phase ERC to locate landmines. The concentration of TNT in the boundary layer of air over a landmine has been estimated at 3 to 6 orders of magnitude below its equilibrium vapor concentration [2, 6]. This places the concentration of TNT in the air over a landmine in the parts-per-trillion (ppt) to parts-per-quadrillion (ppq) range. In terms of mass of analyte per milliliter of air, the concentrations range from femtograms (10−15 g) to low attograms (10−18 g) per milliliter of air. More volatile constituents such as 2,4-DNT would be present at slightly higher concentrations. Nevertheless, vapor detectors must be extremely sensitive in order to detect these low concentrations of signature compounds. The distribution of signature compounds in the environment can be highly heterogeneous, often with small areas of relatively high contamination dispersed among a larger area of little measurable contamination. A surprising finding is that the mine chemical signature is not necessarily strongest directly over the mine. Further, the mine signature often extends past the perimeter of the mine, up to several meters from the mine center. The direction of dispersal of signature

164


has also been observed to be nonsymmetrical and can be influenced by factors such as the topography of the soil near the mine [30]. Consequently, it is extremely difficult to pinpoint the exact location of the mine using trace chemical detection methods alone. This is analogous to the current use of dogs for mine detection. Virtually all mine dogs are trained to alert when they have the landmine’s scent, not when they are directly on top of the mine [31]. Nevertheless, evidence currently available indicates that it may be possible to isolate a mine location to within a small, well-defined area. 7.4.3

Landmine Detection Field Test Results

7.4.3.1 Results of Sensor Blind Testing Blind field testing of Fido against buried landmines has been conducted on several occasions in a test minefield at Ft. Leonard Wood, Missouri. Blind test lanes (one containing TMA-5 Soviet Block antitank mines and the other containing PMA-1A antipersonnel mines) were established by marking potential target positions in the test field. At each test position, two surveying flags were placed approximately 50 cm apart. At some of the test positions, mines were buried, centered between the flags. Other test positions contained no mines. The mines used for the tests were authentic but had the detonators removed and the fuze wells capped with shipping plugs. Fido was used to sample between the flags at each test location. The sensor responses at each location were recorded and submitted to DARPA for scoring. The best performance recorded at this facility was in the antitank mine lane, with a probability of detection (PD ) of 0.90, with a probability of false alarm (PFA ) of 0.10. Tests of this type have been conducted on numerous occasions, with sensor performance ranging from poor to excellent. There is a high degree of correlation between sensor performance and environmental conditions, which have been shown to strongly influence the strength of landmine chemical vapor signatures. Warm to hot temperatures with damp soil conditions and light winds appear to be most favorable, while cold temperatures, with dry soil conditions and high winds appear to be least favorable. Conversations with dog handlers (at the site and elsewhere) confirmed that canines tend to perform well (or poorly) under similar operational conditions. 7.4.3.2 Landmine Detection Performance Comparison to Canines On one occasion, the sensor was tested in conjunction with two experienced teams of canines. One dog was trained to detect explosives, and the other was an experienced mine detection dog. The team trained to detect explosives (i.e., bombs) withdrew from the test because of extreme difficulty in locating the mines. The team of canines with actual landmine detection experience in Bosnia and Mozambique performed much better. The field conditions at the time of the test were very hot with dry soil, conditions under which the dog handlers stated that canine performance was typically poor. Comparisons of Fido performance were made to that of the dogs. Probabilities of detection and false alarm rates were in general slightly better for the sensor than that of the experienced canine landmine

LANDMINE DETECTION

165

detection team. While the testing was limited in scope, the performance of the sensor was very promising. A more detailed account of this testing has been published elsewhere [7]. To the knowledge of the author, this is the first time a chemical vapor “sniffer” had demonstrated the ability to detect landmines under field conditions with performance comparable to that of canines. 7.4.3.3 Influence of Environmental Factors on Detection In general, the performance of the sensor correlates well with what are now understood to be the most relevant environmental factors that affect landmine vapor signature concentrations. In general, the sensor performs better when conditions were warm to hot, with the soil in a drying phase after rain had fallen. This is consistent with the results of the fate and transport simulations conducted by Phelan and ebb [2] and the experimental work of Jenkins et al. [5–7] During one field test, a subset of mines at the Ft. Leonard Wood minefield were sampled at 10:00, 14:00, and 16:00 on three consecutive days in order to evaluate the variability in mine signatures as a function of changing environmental conditions throughout the day. For these tests, a high-volume sampling method known as Remote Explosive Scent Tracing (REST) was utilized. The REST method is derived from the MECHEM Explosive and Drug Detection System (MEDDS), used successfully for almost two decades by MECHEM, a South African demining company. MEDDS has been used to detect roadside improvised explosive devices (IEDs), weapons caches, landmines, and various types of contraband. Using this methodology, an area suspected of being mined is sampled by drawing large volumes of air and entrained soil particulates from a suspect area through a specially designed filter designed to trap vapors of explosives and soil particulates. Portable, highvolume air pumps are used to draw air through the filters. After collecting a sample on the filter, which is inexpensive and disposable, the filter is presented to highly trained dogs for analysis. These dogs are trained to detect traces of explosive that may have been collected on the filter during sampling of a suspect area. Standard REST filters are not compatible with electronic sensors, so a Fido-compatible version of the filter, shown in Figure 7.4, was developed. The filter is also compatible for use with canines. Because canines were not available, samples were analyzed only with the Fido sensor. On day 1, samples were collected at 10:00 and analyzed. Soil conditions were relatively dry, resulting in nonideal detection conditions. Only 60% of the mines were detected. Prior to the scheduled 14:00 sampling on day 1, heavy rainfall fell, resulting in no samples being collected the remainder of day 1. At 10:00 on day 2, sample collection was resumed. The percent of mines detected had fallen to 20%, consistent with the signature being carried into the soil by the heavy rainfall. Days 2 and 3 were sunny with highs in the upper 80s, resulting in rapid drying of the soil. As can be seen from the data in Figure 7.5, the percentage of mines detected rose steadily from a low of 20% detected at 10:00 on day 2 to 90% detected at 18:00 on day 3. This finding is consistent with an increase in the concentration of ERCs in surface soils due to evapotranspiration, resulting

166


Flow (Air)

Packed Bed of Beads

Trapped Analyte

Flow (Air + Analyte)

Figure 7.4 Nomadics version of the REST high-volume sampling filter.

100 Day 1

Day 2

Day 3

20 10

18:00

15:00

30

10:00

40

18:00

50

14:00

60

10:00

Percent Detected

70

10:00

80

Heavy Rain. No Samples Collected

90

0

Figure 7.5 Effect of changing environmental conditions on mine detection.

in enhanced transport of the mine signature to the surface of the ground. The surface soil remained moist during days 2 and 3 of the test, resulting in favorable conditions for release of ERCs from surface soils into the vapor headspace over the mines. The results of this test are consistent with the explosive fate and transport results presented in Sections 7.4.2.3, 4.3.1, and 4.3.2 of Chapter 4. Our experience suggests that during periods of low precipitation when soil conditions are dry, sensor performance was typically better early in the morning, presumably due to the fact that surface soils contain a higher level of moisture early in the morning. As the temperature rises during the day, the soil presumably dries out, resulting in a drop in headspace ERC concentrations, thereby reducing

LANDMINE DETECTION

167

performance of the sensor. Canines seemed to perform better during the early morning as well. 7.4.3.4 Distribution of Mine Signatures in the Vicinity of Mines Another interesting result that was observed consistently during field tests was that mine signatures are not necessarily the strongest directly over the top of a mine. This counterintuitive result was especially true for large antitank mines. The signature was consistently stronger around the perimeter of the antitank mines, forming a “halo” around the mine. It was also observed that when a mine was buried on a slope, the signature was consistently stronger on the downhill side of the slope. This was presumably due to the runoff of surface water carrying the signature in a downhill direction. Mine signatures were heterogeneously distributed in the soil near mines. Areas of contamination resulting in large Fido responses were often very localized, and not necessarily over the top of the mine. Moving the sensor a few centimeters in any direction could change the sensor response dramatically. All of these findings are consistent with the results of laboratory analysis of soil samples taken from these sites during testing [30]. During one field exercise, when mine detection dogs indicated the presence of a mine, the location at which the dog indicated was marked. The sensor was then used to search for the mine. The sensor consistently detected the strongest mine signatures in roughly the same area as the dog. Interestingly, the location of strongest response was not always directly over the mine. The strongest response was, however, usually within a meter of the mine position. 7.4.3.5 Investigation of Area Reduction Methods As pointed out in the previous section, over time mine signatures seem to spread over a fairly large area extending outward from the mine. As previously mentioned, this makes it difficult to use trace chemical sensors to pinpoint the exact location of a mine. A study funded by the U.S. Army Humanitarian Demining Program at NVESD (Ft. Belvoir, Virginia) enabled an investigation of the extent of spread of contamination from mines. The study involved Nomadics and MECHEM, and utilized the high-volume sampling (REST) method previously described in Section 7.4.3.3 [32–35]. REST samples were anlayzed by MECHEM’s canines and were also analyzed by Fido. Three high-volume samples were collected at 3-, 7-, and 11-m radii from the center of four different mine types (metal AT, metal AP, two plastic AP) at three different depths (10, 15, and 20 cm). These were real mines with detonators installed but rendered inert. In tests conducted during the spring of 2003, 82% of samples from the test area were found to contain explosive traces. Explosive signatures were detected out to 11 m from the mine center. Tests were not conducted at larger distances due to the layout of the test field. Data supporting the conclusion that mine chemical signatures do not remain localized in the immediate vicinity of mines were collected at an arid test facility in the United States. In this test, multiple soil samples were collected along a line that ran perpendicular to the length of a mine lane. This line was selected so that

168


Sample Position (Feet From Mine Center)

Sensor Response

−12 −10 −8 −6 −4 −2 2 4 6 8 10 12

0 0 0 0 1 3 1 0 0 0 0 0

Figure 7.6

Sensor responses for samples collected perpendicular to mine lane.

it crossed over the top of a mine in the lane. Twelve samples were taken at 2, 4, 6, 8, 10, and 12 ft north and 2, 4, 6, 8, 10, and 12 ft south from the mine. Solvent extracts of the soil samples were analyzed by gas chromatography using the Fido sensor as a detector. Explosives were detected in only four of the samples. TNT and 2,4-DNT were found in samples at 2 ft south (refer to Fig. 7.6). TNT was recorded in samples taken at 4 ft south and 2 ft north. Hence, samples collected at points along the line on both sides of the lane (but outside of the lane) were blank, but explosive signatures were detected in the samples taken from inside the mine lane. The data indicate that the detection radius of Fido around this mine was ∼1.75 m ± 1.0 m (with 95% confidence). These test results suggest that while it may be difficult to pinpoint the exact location of mines using vapor detection, it may be feasible to detect areas that are mined. This detection paradigm differs from most in use today that rely on various mine-detecting technologies to locate the exact position of targets. Approaches that seek to locate exact target positions require that every square meter of an area be screened for individual mines. In contrast, an area reduction technique could be used to determine whether or not dispersed explosive signatures are present in an area, indicating that a minefield may be present. Once an area is deemed suspect, technologies that are good at pinpointing the exact location of the mines can be used to facilitate removal of the mines. If successful, this will enable demining efforts to focus on areas that actually contain mines, rather than consuming significant resources in areas that are not mined. This need for rapid area reduction has been identified by humanitarian demining organizations as a top priority. 7.4.3.6 Landmine Soil Probe Test Results The concentration of TNT and other ERCs are often many orders of magnitude higher in the soil near a landmine than in the air immediately above the mine. In the soil, the concentration of landmine chemical signature compounds is often highest in the soil that is directly adjacent to the sides, top, and bottom of the mine casing [6, 30]. See Figure 4.13,

LANDMINE DETECTION

169

p88. However, this soil is not easily accessible for vapor sampling analyses because of the overlying covering of soil in which the mine is buried. Hence, a prototype soil-sampling probe has been designed to sample TNT and ERCs in the subsurface soil near a mine where the concentrations are likely to be the highest. The first test of the probe was conducted at Fort Leonard Wood, Missouri, in May 2003. The results of this test were very encouraging, which led to further testing of the probe design at Fort AP VA Hill. The probe at this stage of development consists of a rigid, tapered stainless steel tube that can be inserted into the soil near a mine. The tip of the probe is perforated to allow vapor to diffuse from the soil into the interior of the probe. The probe is designed to accommodate a standard solid-phase microextraction (SPME) sampler that is deployed inside the probe during sampling [36]. The functional portion of the SPME sampler is a fused silica fiber coated with a material that preferentially binds semivolatile organic compounds such as TNT. The sorbent coating is exposed to sample vapors that diffuse or are pumped into the tip of the probe through perforations in the probe tip. As vapors of explosives diffuse through the perforations in the tip of the probe, the coating on the SPME fiber sorbs (through adsorptive and absorptive mechanisms) constituents of the vapor sample. Because of the high affinity of the SPME coating for target analytes, the SPME fiber serves as a vapor preconcentrator. The mass of TNT sorbed is proportional to the amount of time the fiber is exposed to sample, enabling accumulation of sufficient analyte for detection even when the concentration of vapor-phase analyte is very low. To collect a sample, the probe with a SPME fiber installed is inserted into the soil. Air is pumped through the probe, drawing subsurface soil vapors into the probe tip and across the SPME fiber. Pumping air across the fiber increases uptake of target analytes by the SPME fiber relative to what is collected by molecular diffusion alone. Once a sample is collected, the SPME fiber is removed from the probe for analysis. To analyze the sample, the SPME fiber is inserted into a modified inlet system attached to the Fido sensor. The modified inlet serves to heat the SPME fiber, causing rapid and quantitative desorption of trapped molecules of analyte. The vapor-phase analyte is then swept into the sensor for analysis by a flow of carrier gas. Figure 7.7 is a comparison of the response of Fido to a high-volume vapor (REST) sample and a sample collected with the soil vapor probe. The samples were both collected near a TM-62P3 landmine. The REST sample generated a weak sensor response, but the sample collected with the soil probe resulted in a strong Fido response. This data illustrates the improvement in sensor response that can be achieved with the soil probe. More importantly, all mines sampled with the probe (a total of four) were detected. Obviously, there are operational issues with using such an approach for detecting mines, specifically the danger associated with inserting probes into the ground near mines. However, a soil probe could be mounted on a robotic platform equipped with ground penetrating radar (GPR). The GPR could be used to locate a suspicious target. Once located, the GPR could be used to guide insertion of the probe into the soil near the target for sampling. This method could allow

170


High-Volume Vapor (HVV) Sample (0.55% Quench)

1.02 Normalized Response

1 0.98 0.96 0.94 0.92 Soil Probe Sample (13.3% Quench)

0.9 0.88 0.86 0

5

10

15 Time (s)

20

25

30

Figure 7.7 Comparison of Fido responses to samples from the soil probe and MEDDS samples of the same landmine.

for enhanced detection capability without unnecessarily endangering demining personnel. 7.5 COMPARISON OF FIDO WITH CANINES USING HIGH-VOLUME SAMPLING METHODS (REST)

A comparison of the performance of the Fido sensor to MECHEM MEDDS canines was performed at the MECHEM MEDDS facility in Pretoria, South Africa, in February, 2003. These tests were conducted using the Nomadics version of the REST filter previously described in Section 7.4.3.3. At the time of testing, the MECHEM canines had been trained on the Nomadics filter for a period of approximately 4 months. Two dogs, with previous REST experience, were retrained to detect explosives on the Nomadics filters. Positive samples were prepared by placing TNT vapor strips (filter paper coated with a few milligrams of TNT) inside cardboard boxes. Once the vapor strips were placed into a box, the box was sealed using shipping tape. After allowing the vapor to accumulate inside the box for a measured amount of time (typically 30 to 120 min), the top of the box was punctured to allow insertion of a sampling probe into the box. A sample was drawn onto a Nomadics filter inserted in the inlet of the sampling probe. Air was drawn through the filter at a rate of 1 L per second. Samples were collected for varying time intervals ranging from 30 to 120 s. After sampling was completed, the filter was inserted into its plastic storage container until analysis. The concentration of TNT vapor in the box was not determined. However, by increasing or reducing the number of vapor strips in the box, by increasing or reducing the “soak” time prior to sampling, or by varying the sampling time, it was possible to vary the quantity of TNT trapped in the filters.

COMPARISON OF FIDO WITH CANINES USING HIGH-VOLUME SAMPLING METHODS (REST)

171

Samples of potential chemical interferences were also prepared by placing samples of selected materials inside clean cardboard boxes. The potential “interferent” items were materials commonly encountered in demining operations such as petrol. These materials were sampled in an identical fashion as the positive samples. Blank samples were also prepared by sampling empty cardboard boxes. All samples were marked by sampling personnel in a manner that made it impossible for analysts to determine the composition of the sample during analysis. Nomadics personnel and dog handlers were not given any information on sample identity until analysis of samples was completed and results were submitted for scoring (i.e., the tests were conducted in a “blind” fashion). After samples were prepared, they were split into 2 separate batches, with each batch containing positive, blank, and interferent samples. The number of positive, blank, and interferent samples per batch was not revealed to the dog handlers or sensor operators. All samples were first analyzed by the canines. After the canine analysis was completed, Fido was then used to analyze the same batch of samples. Batch 1 of filters contained a total of 25 samples, 4 of which were positive. Both Fido and the canines detected 3 of the 4 positive samples. Interestingly, the sensor and the dogs missed the same positive sample. All samples from batch 1 were analyzed at room temperature. In the second batch of samples, 3 samples out of 24 were positive. Fido and the canines detected all 3 positive samples. Prior to presentation of samples in the second batch to Fido, the samples were heated slightly to enhance the vapor-phase concentration of target analytes in the samples. As would be expected, responses to the positive samples that were heated were stronger than the samples analyzed at room temperature. The performance of Fido and the canines against interferents was also identical. The interferent samples were selected from materials that would be routinely encountered during a demining operation. Of the 20 potential interferents included in the test, Fido and the canines responded to the same interferents, giving responses to 2 of the 20 interferents. Although the sensor responded to 2 test interferents, the sensor response to these 2 interferents was not consistent with that of TNT or ERCs. For the 2 materials that false indications were registered (none of which were pure materials), it is not known which of the chemical compounds in the samples was causing the sensor and the dogs to alarm, nor was it known if they were responding to the same compound. Nevertheless, the response to interferences is interesting and should be studied further. The results of this comparison were promising. The performance of the sensor during this series of tests was comparable to that of the canines. One outcome of these tests was the notion that the Fido sensor could possibly be used as a canine training tool. For example, when positive samples are prepared there is currently no easy way to determine if the samples are actually positive. The sample that was missed by the canines and by Fido was prepared in exactly the same manner as the three samples that were detected, yet this sample was not detected. If the sample in question were used as a positive sample during training, but were actually blank, confusion of the dog could occur, reducing the effectiveness of the training session. In addition, a properly designed electronic sensor should

172


exhibit reproducible and quantifiable levels of performance from day to day. The performance of canines can vary for a variety of reasons, and it can be difficult to determine when a dog is not performing at its best. The sensor could possibly be used to help verify the performance of canines. This is not to say that the performance of Fido is presently adequate to replace dogs in certain roles, but it may have a role in enhancing and complimenting the performance of dogs. Dogs can be trained to detect a wide range of substances in a short period of time. In this respect, dogs could be used to cover critical gaps in sensing applications while sensory materials for detection of other analytes of interest are developed. For sensing tasks in which Fido and dogs could both be used, the sensor could be used in a confirmatory sensing role to improve the performance of both methods. 7.6

CONCLUSIONS

The field experiences presented here represent a cross section of lessons learned during fieldwork performed over a period of approximately 8 years. While laboratory testing of explosives detection equipment is necessary, laboratory testing cannot substitute for testing under realistic field conditions. In addition, the phenomenology associated with chemical signatures of explosive devices is complex and cannot always be studied in a laboratory setting without significant risk of inadvertently altering the characteristics of the signature. ACKNOWLEDGMENTS

We wish to thank the Defense Advanced Research Projects Agency and the U.S. Army Night Vision and Electronic Sensors Directorate for funding the work presented here and for providing technical and programmatic guidance during the execution of this work. We also wish to thank Professor Tim Swager at MIT for his contribution in the form of development of sensing materials for the Fido sensor.

REFERENCES 1. Cumming, C., C. Aker, M. Fisher, M. Fox, M. la Grone, D. Reust, M. Rockley, T. Swager, E. Towers, and V. Williams. Using novel fluorescent polymers as sensory materials for above-ground sensing of chemical signature compounds emanating from buried landmines. IEEE Trans. Geosci. Remote Sensing 39(6), 1119–1128 (2001). 2. Phelan, J. M. and S. W. Webb. Environmental Fate and Transport of Chemical Signatures from Buried Landmines—Screening Model Formulation and Initial Simulations. Sandia Report SAND97-1426, Sandia National Laboratories, Albuquerque, NM June 1997. 3. Walsh, M. E. and T. F. Jenkins. Identification of TNT Transformation Products in Soil. SR92-16, U.S. Army Corps of Engineers, Cold Regions Research and Engineering Laboratory, Hanover, NH 1992.

REFERENCES

173

4. Leggett, D. C., T. F. Jenkins, and R. P Murmann. Composition of Vapors Evolved from Military TNT as Influenced by Temperature, Solid Composition, Age, and Source. SR 77-16/AD A040632, U.S. Army Corps of Engineers, Cold Regions Research and Engineering Laboratory, Hanover, NH, 1977. 5. George, V., T. F. Jenkins, D. C. Leggett, J. H. Cragin, J. Phelan, J. Oxley, and J. Pennington. “Progress on Determining the Vapor Signature of a Buried Landmine”, Proc. SPIE, Detection and Remediation Technologies for Mines and Minelike Targets IV, vol. 3710, part 2, p258, 1999. 6. Jenkins, T. F., M. E. Walsh, P. H. Miyares, J. A. Kopczynski, T. A. Ranney, V. George, J. C. Pennington, and T. E. Berry. Analysis of Explosives-Related Signature Chemicals in Soil Samples Collected Near Buried Landmines. ERDC Technical Report, Cold Regions Research and Engineering Laboratory, Hanover, NH, 2000. 7. George, V., T. F. Jenkins, J. M. Phelan, D. C. Leggett, J. Oxley, S. W. Webb, P. H. Miyares, J. H. Cragin, J. Smith, and T. E. Berry. Progress on determining the vapor signature of a buried landmine. Proc. SPIE, Detection and Remediation Technologies for Mines and Minelike Targets V, 2000, Volume 4038, part 1, p. 290 (2000), Orlando, FL. 8. Cumming, C., C. Aker, M. Fisher, M. Fox, M. la Grone, D. Reust, M. Rockley, T. Swager, E. Towers, and V. Williams. Using novel fluorescent polymers as sensory materials for above-ground sensing of chemical signature compounds emanating from buried landmines, in Proceedings of the UXO/Countermine Forum, New Orleans, Louisiana, April, 2001. 9. la Grone, M., C. Cumming, C. Aker, M. Fisher, M. Fox, D. Reust, M. Rockley, T. Swager, E. Towers, and V. Williams. Developments in the use of an amplified fluorescent polymer-based sensor for the detection of landmines, in Proceedings of Aerosense 2001, Volume 4394-111, International Society for Optical Engineering, Orlando, Florida, April, 2001. 10. Fisher, M. and C. Cumming. Detection of trace concentrations of vapor phase nitroaromatic explosives by fluorescence quenching of novel polymer materials, in Proceedings of 7th International Symposium on the Analysis and Detection of Explosives, Defense Evaluation and Research Agency, Edinburgh, Scotland, UK, June, 2001. 11. Fisher, M. and C. Cumming. Trace detection of nitroaromatic explosives by fluorescence quenching of novel polymer materials, in Proceedings of the U.S. Federal Aviation Administration’s Third International Aviation Security Technology Symposium, Atlantic City, NJ, November 27–30, 2001. 12. Fisher, M. and C. Cumming. Utilization of novel fluorescent polymer materials for trace level vapor-phase detection of nitroaromatic explosives, in Proceedings of the U.S. Federal Aviation Administration’s Third International Aviation Security Technology Symposium, Atlantic City, NJ, November 27–30, 2001. 13. Fisher, M., M. la Grone, and J. Sikes. Implementation of serial amplifying fluorescent polymer arrays for enhanced chemical vapor sensing of landmines, in Proceedings of UXO/Countermine Forum 2003, Orlando, Florida, September 2003. 14. Yang, J. S. and T. M. Swager. Porous shape persistent fluorescent polymer films: An approach to TNT sensory materials. J. Am. Chem. Soc. 120, 5321–5322, 1998. 15. Yang, J. S. and T. M. Swager. Fluorescent porous polymer films as TNT chemosensors: Electronic and structural effects. J. Am. Chem. Soc. 120, 11864–11873, 1998.

174


16. Williams, V. and T. M. Swager. Iptycene-containing poly(aryleneethynylene)s. Macromolecules 33, 4069–4073 (2000). 17. Zhou, Q. and T. Swager. Methodology for enhancing the sensitivity of fluorescent chemosensors: Energy migration in conjugated polymers. J. Am. Chem Soc. 117, 7017–7018 (1995). 18. Yinon, J. and S. Zitrin. Modern Methods and Applications in Analysis of Explosives. Wiley, New York, 1993. 19. Kirk-Othmer Encyclopedia of Chemical Technology, Vol. 9, 3rd ed. Wiley, New York, 1980. 20. U.S. Department of State. Hidden Killers: The Global Demining Crisis. U.S. Department of State Publication 190575, Washington, DC, 1998. 21. Field Engineering and Mine Warfare, Pamphlet No. 6, Detection and Clearance of Mines, British War Office, London, UK 1947. 22. Nolan, R. V. and D. L. Gravitte. Mine Detecting Canines. Report 2217, U.S. Army Mobility Research and Development Command, Ft Belvoir, VA 1977. 23. King, C., Ed. Jane’s Mines and Mine Clearance, 4th ed., Jane’s Information Group, Surrey, UK. 1999–2000. 24. Mine Facts. CD-ROM, v. 1.2, U.S. Department of Defense, Washington, DC, 1997. 25. Harper, R. J., J. R. Almirall, and K. G. Furton. Identification of dominant odor chemicals emanating from explosives for use in developing optimal training aid combinations and mimics for canine detection. Talanta 67, 313–327 (2005). 26. Johnston, J. M., M. Williams, L. P. Waggoner, C. C. Edge, R. E. Dugan, and S. F. Hallowell. Canine detection odor signatures for mine-related explosives. Proc. SPIE, Detection and Remediation Technologies for Mines and Minelike Targets III, 3392(1), 490–501 (1998). 27. Kenna, B. T. and F. J. Conrad. Studies of the Adsorption/Desorption Behavior of Explosive-Like Molecules. Sandia Report SAND86-0141, Sandia National Laboratories, Albuquerque, NM 1986. 28. Grant, C. L., T. F. Jenkins, and S. M. Golden. Experimental Assessment of Analytical Holding Times for Nitroaromatic and Nitramine Explosives in Soil. SR 93–11, U.S. Army Corps of Engineers, Cold Regions Research and Engineering Laboratory, Hanover, NH June 1993. 29. Spencer, W. F., M. M. Cliath, and S. R. Yates. Soil-pesticide interactions and their impact on the volatilization process, in Environmental Impact of Soil Component Interactions—Natural and Anthropogenic Organics, Vol. 1, CRC Press, Boca Raton, FL, 1995, pp. 371–381. 30. Hewitt, A. D., T. F. Jenkins, and T. A. Ranney. ERDC/CRREL TR-01-09, U.S. Army Corps of Engineers, Cold Regions Research and Engineering Laboratory, Hanover, NH May 2001. 31. Horwood, C. The Use of Dogs for Operations Related to Humanitarian Mine Clearance. Handicap International Mines Coordination Unit, Paris, France, 1998. 32. Fisher, M., M. Prather, and J. Sikes. Serial amplifying fluorescent polymer arrays for enhanced chemical vapor sensing of landmines. EUDEM-2/SCOT Conference, Brussels, Belgium, September, 2003. 33. Fisher, M. and J. Sikes. Minefield edge detection using a novel chemical vapor sensing technique, in R. S. Harmon, J.H. Holloway, Jr., J.T. Broach, Eds. Detection

REFERENCES

175

and Remediation Technologies for Mines and Minelike Targets VIII, Proceedings of SPIE Vol. 5089, pp. 1078–1087, April 21–25, 2003, Orlando, FL. 34. la Grone, M. C. Cumming, M. Fisher, and E. Towers. Investigation of an area reduction method for suspected minefields using an ultra-sensitive chemical vapor detector, in Proceedings of Aerosense 2002, Volume 4742, International Society for Optical Engineering, Orlando, Florida, April, 2002. 35. Fisher, M. M. la Grone, C. Cumming, and E. Towers. Utilization of chemical vapor detection of explosives as a means of rapid minefield area reduction, in Proceedings of the 5th International Symposium on Technology and the Mine Problem, Mine Warfare Association, Monterrey, California, April 2002. 36. Pawliszyn, J. Solid Phase Microextraction—Theory and Practice. Wiley, New York, 1997.

CHAPTER 8

REFLECTIONS ON HUNTING MINES BY AROMA SENSING VERNON JOYNT Mechem Division of Denel, South Africa (retired)

8.1

EDITOR’S NOTE

This chapter is organized in the form of an interview with Dr. Vernon Joynt. Dr. Joynt has many years’ experience clearing minefields in southern Africa. In recent years he has brought his experience to bear in other areas, notably in the Balkans. He was a, perhaps better stated the, pioneer in using trace chemical sensing to locate mines in the field. His techniques were developed with animals, first dogs, later adding rats. The design of this interview is the highlighting of some of the practical considerations that only actual field experience can teach. The following pages contain a very informal set of questions and answers, conducted over several months, mostly by e-mail.

8.2

INTERVIEW

Editor: You mentioned a workshop where you recently gave a lecture on a subject that sounds very much like the subject we want to include in the book. What was the most pertinent part of that workshop? VJ: The workshop that addressed the problem would have been quite interesting to you, as it was a shoot-out between chemical and physical test procedures and animal noses. I was asked to represent the noses. Editor: I would have found that discussion extremely interesting. I know you have compared the ability of animals to detect trace chemicals with the sensitivity of various chemical techniques. You have discussed this with Trace Chemical Sensing of Explosives, Edited by Ronald L. Woodfin c 2007 John Wiley & Sons, Inc. Copyright

177

178

REFLECTIONS ON HUNTING MINES BY AROMA SENSING

many chemists. What is your personal assessment of the current comparison, and what trend do you see? VJ: To answer the question as a gut reaction just this: More and more biological detectors like bees, elephants, and others that I know less about are shown to be viable detectors. In all cases the best practical results are achieved by the addition of a scientific enhancement like an electronic eye to observe the bee’s proboscis flick in and out, or the satellite thermal imager tracing elephants on the move. And then there are, of course, filters to help dogs and rats. So combination systems are the most useful. There are many more stories, but a key is to use combinations not shoot-outs. Animal sensors add a component of speed, sensitivity, and cost reduction to a system. The scientific instrument in turn enhances accuracy and credibility. Editor: You have, for quite some time, worked in the laboratory and the field developing the MEDDS (Mechem1 Explosive and Drug Detection System) that has led to the REST (remote explosive sensing techniques) concept. What is the current state of development of these? VJ: In Africa I attended the REST workshop at Morogoro in Tanzania. This workshop was attended by top chemists from Nomadics2 (Dr. Mark Fischer), Sandia National Laboratories3 (Jim Phelan), and FOI4 (Lena Sarholm). The world REST fraternity was also well represented: The Geneva International Centre for Humanitarian Demining (GICHD), who was running the workshop under H˚avard Bach, APOPO5 under Christophe Cox, Mechem with Kip Schulz, NOKSH,6 and NPA7 with Rune Fjellanger. There was also an excellent summary of the workshop presentations issued by Dr. Ian McLean of the GICHD. At a large workshop of about 150 dog people in Ljubljana, Slovenia, some years ago, the case for mine detection dogs (MDDs) was debated. No consensus was reached. There were just too many opinions, stresses, and uncertainties. The work group at the GICHD started basically with the same problems but through the persistence of H˚avard Bach and others the problems became less. The old Mechem was part of that GICHD MDD effort and I pushed our MEDDS with a passion. It was only accepted once the South African government 1

MECHEM, a division of DENEL (Pty) Ltd (South Africa), was established in the late 1960s, originally as a research unit of the Council for Scientific and Industrial Research (CSIR) in South Africa, culminating into a division of DENEL (South Africa). 2 See Chapters 6 and 7. 3 See Chapter 4. 4 FOI, Totalförsvarets forskningsinstitut, Sweden. 5 The APOPO project: African Giant Pouched Rats, SUA APOPO (prime contractor). 6 Norsk Kompetansesenter for Spesialsøkshund (NOKSH), Norway. 7 Norwegian Peoples’ Aid.

INTERVIEW

179

offered the use of the system to all humanitarian demining (HD) operations with no strings attached. This was done in Tokyo, during a United Nations meeting there, by the SA8 Foreign Affairs Department. The technology in this context was first transferred to NPA who named it explosive vapor detection (EVD), and then later H˚avard Bach renamed and broadened the concept to REST. MEDDS was more scientifically controllable because the training and use of the dogs were separated from the field and sampling problems. With MDD, setting up controlled tests was extremely difficult when one is trying to do this in real operational demining conditions. With MEDDS this was easier and the results more acceptable. MEDDS resulted in our partaking in the DARPA9 Electronic nose project. This project delivered two significant products: • •

The Nomadics/MIT FIDO®10 electronic nose The Quantum Magnetics NQR11 explosive detector

A NVESD12 contract over the past two years with the new Mechem and Nomadics attempted to repeat more scientifically what the original development and demining contracts had taught us about the system. They had some success. MEDDS taught us many things about MDD and their problems, but most important of all it taught us the problems of what happens to the explosives once they are out of the landmine and left to the environmental influences. A nightmare! MEDDS was often used to provide a dog-nose benchmark to a vapor sensor development or evaluation. Many unresolved arguments resulted as to the sensitivity levels that a dog’s nose can achieve. According to Jim Phelan and us, dogs in Mechem and Global13 (San Antonio) could do ppq (parts per quadrillion). This is detecting 10−15 grams of TNT [trinitrotoluene] or femtograms. A picogram is 10−12 grams, so dogs are at least 1000 times below classic chemistry, which can normally do picograms. Nomadics have, however, now cast some doubt on the way this figure was measured. And so it goes on. So it is fair to say that the whole field of REST is improving and moving forward. Editor: Clearly, establishing what you call sensitivity, or others may call limit of detection, or LOD, is a very challenging task, one that is never ending as technology improves. In your opinion, what is the best way to quantitatively establish this LOD for any system? 8

South Africa. (U.S.) Defense Advanced Projects Agency. 10 See Chapter 9. 11 Nuclear quadrupole resonance, a bulk sensing technology. Quantuum Magnetics, Inc., San Diego, California. 12 U.S. Army Night Vision and Electronic Sensors Directorate, Ft. Belvoir, Virginia. 13 Global Training Academy, Somerset, Texas. 9

180


VJ: My determination went back some years when the UK sent a team of forensic scientists under Dr. Dick Lacey and Tim Sheldon to come to South Africa in about 1988. My group was still in the CSIR14 of South Africa, and we had just started developing MEDDS, then called SOS, which in Afrikaans is Springstof Opsporing Stelsel, for tracing the smuggling of explosives into the country. For obvious reasons, the British police were interested. We hired them a chemical laboratory on the CSIR site, and they proceeded to prepare vapor samples using a technique of adsorptions onto polyethylene beads and a split-dilute-equilibrate-split again, etc., till they got down to vapor concentration levels of 10−14 grams RDX (cyclotrimethylenetrinitrzmine). This was the technique they had used to evaluate other chemical systems, like the Condor system. Our dogs almost broke their necks indicating the samples we had sucked onto filters from their measured source! (Exaggeration! All four dogs indicated, repeatedly, no errors.) I asked they drop the concentration. Equilibrations after further splits take exponentially longer. But, they went away, and after some days came back with a 10−15 grams RDX sample, which the dogs also found without error. However, from judgment of previous tests we had concluded with those dogs they would go tenfold down before some would start missing. Thus 10−16 grams is possible or even better with very carefully trained animals. Like with athletes, we found practice improves performance. Their final conclusion was positive, and further visits took place in which operational facilities at country border stations of the SA Police and SA Defense Force were visited. Today some systems in the UK and France based on this original cooperation is still ongoing. Editor: In this set of experiments you used RDX. It is a more common practice to use TNT in such experiments. Do the MDDs distinguish between the two? VJ: The MEDDS dogs could never be taught to discriminate between RDX and TNT. This suited us fine for mine clearance. In fact right at the onset in 1987 we had contracted a scientifically run dog school to train dogs for us that would discriminate between the various common explosives. After two years of trying they could only show the dogs were able to separate the nitro esters like PETN (pentzerythritoltetranitrate) and nitrocellulose on the one hand and organic nitro compounds like RDX and TNT on the other hand, but not between compounds inside a group. This made me conclude that the smell was a combination of the compound and its decomposition products. The nitro esters are decomposed when exposed to acids while the nitro explosives are decomposed by exposure to alkaline conditions. Maybe the alkaline nature of mucus in the mouth and nose produce the 14

Council of Scientific and Industrial Research, South Africa.

INTERVIEW

181

difference they can recognize. We only pursued that research for two years and it was inconclusive. Editor: Has there been an occasion in your experience where some electronic sensor has achieved an LOD comparable to those of the animals with which you have worked? VJ: Yes, some of my work included comparisons between dogs and chemical methods. I participated in an important one sponsored by NVESD. What makes this NVESD test important is that under some of those test conditions the Fido® actually gave more sensitive results than the dogs. My explanation, which was accepted by Prof. Vladimir Knapp of CROMAC15 who was also one of the scientific advisors on the project, was logical (says I!) but the conclusions were never or, as yet, have not been made public.∗ Editor: What is your explanation? VJ: The training of the Mechem MEDDS dogs in the test were rigidly controlled to find only TNT and RDX, and none of the degradation products like the amino-dinitrobenzene products, while the Fido could find that type of product. The life of these degradation products in wet soil was orders of magnitude longer than that of the TNT being exuded out of the test mines. The result was that the Fido could indicate the presence of the test mines at longer distances than the MEDDS dogs. In all probability, if the dogs were also trained on the other chemicals, they would have been at least as good, but the test records stand. It must also be said that the tests were typical of real-world conditions with storms and other factors spreading the chemicals around in the test site. Therefore, the comparisons were difficult to scientifically duplicate, and the conclusion we made was only supported by percentage calculations. None of these were 90% or above. More like 65% for Fido and 35% for the dogs on specific mines in the test. Editor: You have found that the animals sense the chemicals in some ways that we don’t always anticipate. You mentioned vultures once. How do they fit in? VJ: There are also some likely stories that vultures rely more on smell to alert them about a dead animal than their legendary eye sight that has encouraged some of us to fit REST-based systems on RP16 helicopters to search for arms caches and mine fields. Editor: That brings two questions to mind. First, what success have you had with the RP helicopters? 15

Croatian Mine Action Center, Sisak, Croatia. Remotely piloted. ∗ Editor’s note: That comparison is discussed in Section 7.4.3, page 162.

16

182


VJ: As yet we have only done wind tunnel tests of sampling and they look promising, no more. Editor: The other question brought to mind by the vulture reference concerns odor plumes. An entire, Chapter 5, of the book is devoted to the structure of plumes in air and water. Most animals, not just vultures, have an instinctive sense of how to follow a plume to its source. As you have used animals in the field, have you observed any behaviors that seem to indicate plume following? How far have you seen animals effectively follow an explosive plume? VJ: Plumes work better in water. But in air, particulate sensing with dust and wind is more likely than vapor and diffusion. From the word go, Mechem and myself urged for the testing of vegetation instead of soils when searching for explosive vapors that originate out of the soil. In our mine clearance contracts in Mozambique and Angola, where we cleared more than 5000 miles of mined roads in UN contracts, we never failed a QA17 done behind us by manual deminers of another demining company, who had the QA contract from UN. Mechem’s own internal QA, however, found the MEDDS system missing mines. Then we would go back and look for the errors. Inevitably the misses would be on clean soil, mostly dry or when the sun was baking down. In all cases a retest on vegetation in the vicinity of the missed mine would indicate the explosive as positive, while a retest above the soil above the mine would still be negative. We changed our sampling to always include sucking filters in the vegetation on the roadsides. This worked well. What upset me, is that most of the scientific testing going on with dogs and instruments were on clean soil. Vegetation was not being included, because it was said that the problem then became too complex, and one step at a time was the way to go, before the vegetation steps would be brought to bear. What was good about the NVESD tests in Croatia was that vegetation was part of the test. When a local farmer accidentally cut a part of the test site’s long grass, we suddenly found no explosive near the mines; but when Mark Fisher, of Nomadics, took the Fido to the heap of cut grass, he had immediate positive tests in the cut grass. It is likely that dust particles get stuck to leaves and the TNT is protected there from both soil bacteria and intense sunlight. When one looks at studies on how the dog’s nose moves the dust particles [1] to the smell sensors with his nose hairs and mucus, then one realizes that the water in the mucus dissolves the TNT off the dust and offers it to the olfactory sensors. Fjellanger also quoted literature of how a dog’s nose uses side muscles to pulse air at a frequency of 3 to 5 puffs per second at the slits on the side of his nostrils when sniffing. The dog then sucks the stirred up dust with a long 17

Quality Assurance.

INTERVIEW

183

Figure 8.1 See color plates. Diagram of a dog’s sniffing passages. Substitute dust with TNT adsorbed for sausages! (Courtesy of Rune Fjellanger, NPA. Used by permission.)

suck through his main nostril into his nose. The olfactory sensors sit in the roof of the nose and the mucus moving past is flushed down the back into his throat. Figure 8.1 is from Rune’s lecture in Tanzania. Editor: That seems to correlate with our observation that the molecules tend to adsorb on soil particles and dust, rather than remain in the air as explosive vapor. I take it that you mean that some wind is needed to move these particles in order to make them available above ground level to the animal. So, a pure vapor plume in air is not as rich a source as wind carried dust and soil particles. Is that correct? VJ: The story about the importance of vegetation seems to bear that out. Editor: In Chapter 4 of this book we examine the issue of leakage of explosive from mines. We attribute a comment to you about the resulting pattern of dispersal of the explosive molecules on the ground surface. Of course, the details of the local topography have a dominant effect, but you indicated that in nearly flat areas you routinely notice a peculiar pattern. Please address this issue. VJ: The explosives leak and leach out of buried mines and then spread out into and onto the surrounding soil. How does it spread? Where does it find itself after a time? What happens to it in different conditions? A million questions but the mere fact of MDD successes made it important to study this. This has been a bugbear all along, and in the early 1990s we tried to establish how the contours of detection around a buried mine, or mines, looked. The shape we established, I described later, could be as irregular as a camouflage pattern. Compare this with the lines determined during one NVESD test in Croatia recently (Fig. 8.2): The grid is in inches and feet. There were some deep ruts made by wheel marks along the service road through the site. This seems to have produced the 40-cm depth.

184


Figure 8.2 Pattern of concentrated explosive residue, Croatia, Area Reduction Test Site, May/September 2002. Shaded areas represent standing and flowing water to 40 cm deep (Courtesy of Mark Fisher, Nomadics, Inc. Used by permission.)

Although this represented standing water, tests done showed the TNT to be strongest in the water. This is to be expected because TNT is about a million times more soluble in water than in air. What is important is to note that when the water dried up the spread of TNT remained in that shape. This would explain the so-called camouflage patterns we had determined ten years before. The scientific tests and results obtained in the early 1990s by the old Mechem were never published and only recorded in broad detail. What was shared was the results and developed techniques. Editor: When you say “the old Mechem,” do you mean while it was part of CSIR? VJ: The term old and new Mechem refers to the split when the larger old Mechem in Denel had its R&D group split off and moved into the CSIR some four years ago. I then retired and was hired by CSIR to work with the 70 strong R&D group while the smaller new Mechem component that remained at Denel did mostly only mine clearance contracts for commercial gain. The NVESD contract started before the split but went on mostly after the split and I then had only an advisor role. CSIR did have an interest in MEDDS but not in mine clearance, which remained at new Mechem. CSIR was involved in using dogs and filters to do cancer detection in humans. Editor: This discussion brings to mind the question of how the animals sniff in the molecules. You have addressed this question from the point of view of field practice more than physiology, which may help us understand how better to use electronic systems. You indicated that recent data from Norway18 indicates that the dogs more often make correct identification of explosives when the humidity is low, but that the temperature seems to have no effect. How does this compare with the greater explosive vapor flux rates measured after rain, as discussed in Chapter 4 and Chapter 7? VJ: We used to believe MEDDS worked better in humid conditions and avoided hot dry conditions. This came from our MDD experiences. Dogs smell better when the humidity is above 30%; but REST is sampled with a pump and filter that then is presented to a dog in a humidified room. 18

Norsk Kompetansesenter for Spesialsøkshund (NOKSH).

INTERVIEW

185

Editor: You used the MEDDS system to explore this in terms of the new ScentPrint REST filter system. Who developed this system and why is it an improvement? VJ: The CSIR R&D group developed the ScentPrint™ filter for the cancer work, but we then explored the possibility of using it more widely. New Mechem never adapted to its use. Mechem retained the dogs, as the CSIR ruled that they would not use dogs on their site; that is a registered game sanctuary. The ScentPrint filter is a two-stage system. It consists of a tubewithin-a-tube, about 3 cm in diameter by about 20 cm long. It comes with lids for each end that are removed for use. Air is drawn through the tubes to collect the sample. The inner tube is perforated to release the trapped vapor sample. The tube is also designed to house any vapor capturing material suited for the appropriate circumstance. The inner tube is a holding reservoir with strong absorption while the cotton wool-like wads have the correct surface adsorption properties. This is to hold enough transferred reservoir material ready on the surface of the wool for the hot moist animal breath to desorb the smell and get it into his nose. Editor: Your reference to “the appropriate circumstance” recalls that the system is used for vapors other than explosives. What are some of these? VJ: The CSIR, in collaboration with a U.S. company, is detecting cancer in humans by using dogs, and the ScentPrint two-stage filter system the CSIR has developed for multiple uses of REST. The U.S. company has already some trained dogs doing the cancer work and the addition of a filter step has shown an improvement. The CSIR is also involved with tuberculosis diagnosis but lacks an animal training company as a partner. One needs also to consider the APOPO rats and mention they are also using the rats to detect tuberculosis. So biological detection, or perhaps, the chemical signature produced by biological processes, is in the cards for REST. APOPO is also running a rat detection program to do TB diagnosis. For this they found a second-stage absorber that works better than the Mechem filter, to be polyester wool. The CSIR found that similar polyester wool coated with a silicone polymer was best for cancer detection’s second stage in the ScentPrint filter. Editor: So, do there seem to be materials that work better for specific vapors? VJ: Scientifically, the filter problem is that a good absorber is a good reservoir for large quantities of the active chemical or bouquet vapors. This seldom is a good adsorbent for releasing the ingredient to the animal when it smells. Good adsorbents often hold too little sample for a series of animals to make passes, and one then finds the second stage “fading.” So the second stage can have more than one “wool filter” to use in sequence of smelling. (Now you see why Jim Phelan has coined the word “sorption” so one need not specify!)

186


Figure 8.3

Fixture for comparing filter materials using parallel adsorption filters.

APOPO uses a polyester fiber wad for their rats to smell so we ran a parallel adsorption filter configuration (Fig. 8.3) and let the rats smell both the MEDDS and the APOPO filter. Figure 8.3 is a picture of the parallel filter system. Chemical detectors can and are using filters in mine detection to do a similar function, to concentrate samples, and also to separate the rough and tumble of field sampling from the analysis step. The various companies use different solutions, but inevitably they have to have a holding stage followed by a quick release into the detector. Like the cold finger and flash heater of commercial chemical analyzers. Editor: You discussed the camouflage patterns of leaked explosives in soil and vegetation. How long after the puddles dry are MDDs able to find the scent? It would seem that, since animals must be trained with particular suites of molecules, or aromas, that these should match the expected field aromas. Do decomposition products become more important and change over time? VJ: Both Sandia and FOI analyzed the decomposition products of TNT in various soils and conditions. An example of a FOI result is shown in Figure 8.4. These are products extracted out of soil contaminated with TNT, RDX, and TETRYL (Trinitrophenyl-n-methylnitramine). For TNT a decomposition product that had a longer half-life in the soil was 4-amino-dinitrotoluene, so in doing area reduction one should include this product. Editor: This begs for a much more extensive explanation and interpretation. What does it teach us? VJ: The only comment I would like to make is that of the hundreds of soil possibilities put into a bucket with the hundreds of landmine types put into

INTERVIEW

187

TNT

Real Sample from Afghanistan Kharga Area June-03 029AFG30

8e + 4

2, 4-DNA

Intensity

1e + 5

4-ADNT RDX 2-ADNT

6e + 4

4e + 4 5

10

15

20

Time (min)

TNT

3e + 5

Real Sample from Kharga Area Afghanistan 029AFG29

2e + 5

2, 4-DNT

1e + 5 5e + 4 5

10

15

4-ADNT RDX 2-ADNT

2e + 5 2, 4-DNA

Intensity

3e + 5

20

Time (min) (a)

Figure 8.4(a) Analysis of field samples from the vicinity of abandoned landmines in Afghanistan. The figure shows the chromatograms of two samples where TNT, 2,4-DNT, 4-amino-2,6-DNT and 2-amino-4,6-DNT have been detected and quantified using GC and thermoionic detection. In sample B, the TNT concentration caused the GC detector out of range. RDX was also detected in both samples. (2,4-DNA = 2,4-dinitroanisole used as internal standard.) The detected sample background was in both cases very low. As an extra validation of results, the presence of TNT in sample B was verified with LC/MS analysis. In cases where samples only contained trace amounts, several samples with similar detected TNT and related compounds profile were pooled to obtain concentrations high enough for LC/MS analysis. (Courtesy of Ann Kjellström, FOI.)

188


0.065 LC

0.060

O2N N O2N N

UV 2B5 nm

4-NT NO2 N NO2 N

O

O−

0.055 N

0.050 0.045

O

0.040

N−

O

O

N−

0.025

N−

N+ O

2, 4-dinitro6-NT Tetryl 3-NT

O−

N

0.020 0.015 0.010 0.005

−

O

N−

N

O−

O

+

N+

N O−

O−

O

NB O−

0.030 O

O−

O−

N O

N+

3, 4-DNT

N−

N

O

N+ O

NO2

O

N− N

N+ O− O

−

N

−

0.035 A.U

N

O2N

N

N

N−

1, 2-DNB O 2-NT

N+ O 2, 5-DNT 1, 4-DNB 2, 4-DNT

3, 4DNT

1, 3-DNB

O−

O

2, 3-DNT

RDX HMX

CL 20

O−

2-amino-4, 6-DNT

3, 5-DNT TNT

4-amino-2, 6-DNT 2, 6-dinitro4-NT

TNB

0.000 5.00

10.00

15.00

20.00 25.00 30.00 Time (min) (b)

35.00

40.00

46.00

Figure 8.4(b) HPLC-UV chromatogram of organic explosives (standards) using a porous graphitic carbon (PGC) column in reversed phase mode (UV detection 290 nm). Gradient elution using a water–acetonitrile–methanol–toluene system. A quartenary HPLC system was used to deliver the solvent mixture. The complete separation of all the analytes, including the DNT isomers, was carried out on a porous graphitic carbon column (150 mm, 4.6 mm I.D., 5 μm particles). UV detection was conducted by using a Waters UV Detector Model 996 photodiode array detector. The best signal-to-noise ratio was found at 290 nm. (Courtesy of Erik Holmgren, FOI.)

that bucket of soil, you would never get results that would help you with a field system. The only significant result for me was the recognition that decomposition products are abundant and some have much better survival and mobility away from the source of explosive. Results we observed in the Croatian tests support this observation. Editor: Can you illustrate just how the REST system is used in the field? VJ: While working with the company CGTVA19 in Mozambique we used a REST to sample and then using the Morogoro rats to analyze the filters. We surveyed 70 kilometers of road verge using the pump on a hand-pushed cart (Fig. 8.5).

19

Carlos Gassmann Tecnologias De Vanguarda Aplicadas Lda, Portugal.

INTERVIEW

189

Figure 8.5 See color plates. Searching for mines and UXO in Mozambique using REST procedures.

In the picture Osario Serveniano, the ex-director of the Mozambique mine action center and then the manager of the CGTVA operation, can be seen doing the sampling. He is standing in a footpath made by locals so is quite safe. The area to be cleared extended some 3 meters each side of the road and some UXO (unexploded ordnance) and mines were located. The road had to be widened with a new tar strip and the road building machines followed the demining operation. No mines or UXO were missed. Editor: You have mentioned these rats several times. Can you describe them and how you use them? How is it different from using MDDs? VJ: Figure 8.6 is a picture of me petting one. I suggest you get a copy of one of APOPO’s publications. APOPO had won the World Bank competition in

Figure 8.6 APOPO African pouched rat (Cricetomys gambianus), a friendly creature.

190


the face of more than a thousand entries, I believe, for this system to help in the mine clearance problem. The story should be out there. APOPO has a website [2]. Editor: Your success with dogs, both in the minefield and using the MEDDS system, is well documented. You have also used the rats and most electronic systems. What does your experience indicate, overall, of the relative sensitivities of the animals and the electronic systems? VJ: Animals are more sensitive and a lot quicker but their communication and training is flawed. Editor: Do you have a feeling, based on your extensive field experience, of the relative values of sensing a single chemical compound or a group of related chemicals that form a complex aroma? VJ: What is important in the project we discussed is that, to my mind, it indicated what chemicals had staying power in soil and spread out certain distances using plants, rain, wind, and diffusion to be picked up by an animal or chemical detector under real world conditions. In the test, the dogs were trained to find only TNT and nothing else. Because Fido could find more chemicals, it could afford to lose on straight TNT detection in this real-world test. Editor: Based on your experience with animals and with some of the best current chemical sensing technology, what would you project as the necessary practical sensitivity needed in a chemical sensor: VJ: Around 10−15 grams of TNT for a detection. Less sensitive than that will require a concentration step that would take time. For Mechem to do clearance contracts looking for “wild mines” we had to move at about 20 kilometers a day while sampling a 10-meter-wide strip. Manual demining worked orders slower. Rough estimates showed that in Cambodia, Angola, and Croatia, manual demining will take 50 years or more to remove the mines, using the most optimistic calculations. Mechem’s methods could clear roads at a practical rate. However, wild mines are spread over open countryside. To do area clearance, Mechem’s techniques would have to speed up at least fivefold to be really useful. Manual demining, using hand-held detectors, at present rates has no hope in humanitarian demining. Military objectives are much more restricted. Editor: You mentioned earlier that animals are faster. What kind of cycle time do electronic instruments need to have to become competitive? This may be best answered in terms of area coverage rate, or something like that. VJ: Quite a question. Let’s look at MEDDS as we used it: First, we do a survey checking 400 meters by 10 meters per filter. A filter suction sampler can reliably find explosive vapors (on dust) on the vegetation at 5 meters distances away from the source of the explosive, so one pass with a vehicle gives a 10-meter-wide check. The use of an absorption filter concentrates the explosive vapor. We suck at a liter per second continuously through the filter.

INTERVIEW

191

A dog sucks maybe 100 milliliters per sniff, so he must move slower if he needs to pick up enough dust from say 5 meters away from the source. Surveying a 10-meter-wide path at 2 kilometers per hour gave 5 pairs of simultaneous samples in filters; we could check 400 filters a day with a team of MEDDS dogs. All samples were taken double. The MRV20 sampling vehicle worked 10 hours a day, thus doing 20 kilometers, and 200 samples in the 10 hours. Doing the test directly in the vehicle while on the move would require an analysis every 3 minutes. Each filter would then represent 400 meters times 10 meters. The positive sections are then rechecked using a normal MDD. A dog takes a sniff and in a second and gives the answer. The problem with MDDs is that they are not allowed to follow their noses but the handlers have to work them in a rigid pattern to satisfy contractual requirements. At that rate it takes them about a minute to cover a 10-meter by 10-meter section. So if an instrument that can work a 10-meter by 10-meter section and declare it positive or mine free in a minute, it can do the same. A manual deminer does the final positioning and opening of the mine. Not all companies using MDD on squares work the same as Mechem, to let the dog only identify the positive section and then let a manual deminer use a handheld detector and search the whole 10-meter by 10-meter positive section to find and position the mine. Some other companies try and rely on the dog indicating the mine within a 1-meter distance and then let the man just open the mine The dog process is then slower, but the manual deminer’s job much shorter. In my opinion this method is less reliable because of the movement of the explosive smells away from their source and allowing conclusions that dogs had given a false signal when there is no mine within a meter. They then move to another block often leaving a mine behind in the positive block. Alternatively for an instrument to fit into our process, a third different approach can be used: Redo the long positive section with filters but now use much smaller areas to narrow down the position of mines, for example, sections of 10 meters long by 10 meters wide. If the answer would come directly through an analyses of the previous section while one is now sampling the next section, then the process would be continuous. The problem with this third approach was that the changing of the filters between sections becomes a real time-limiting step. Doing 10 meters at 2 kilometers per hour would require a test answer every 18 seconds. Editor: You estimated 3 minutes per analysis on board the MRV. You also state an analysis cycle time of 18 seconds. How do these two times relate to each other? VJ: They are supporting different steps in the process. In summary, the process is as follows: 20

Mine resistant vehicle.

192


1. Survey the road or strip at 2 kilometers per hour in 400-meter lengths, 10 meters wide (3 minutes per strip). 2. Go back to the positive strips and ignore the negative strips. 3. Narrow the positive in the 400-meter strip to 10-meter by 10-meter blocks that are testing positive (18 seconds per block). 4. Use manual deminers in the 10-meter by 10-meter positive blocks to locate the precise position. Editor: You mentioned the MRV. You have spent a great deal of effort developing these vehicles. Having seen your videos of the Casspir moving through a minefield exploding mines as it goes, but without damage, is very impressive. Did you use them extensively to do the route surveys you mentioned? VJ: The key to the Mechem mine clearance operations have been the use of these vehicles because they can carry detectors through mined areas safely without exposing the operator. Why I liked rats is because the REST system using them could be built into a MRV like a Buffalo or Casspir using a team of rats on board. The new Fido can now also do this. MRV are also sometimes called MPV or Mine Protected Vehicles but I prefer MRV. Incidentally in contrast to LAV light armored vehicles like Bradley and Stryker the MRV used in Africa and built in South Africa, Namibia and Rhodesia numbered around 19 000 and are now the only vehicles that suit conditions in Afghanistan and Iraq. Editor: Thank you for the benefit of your experience. Hopefully, these insights will help make the development and employment of new technologies more efficient and more effective. REFERENCES 1. Fjellaner, R. The REST concept, in I. G. McLean, Ed. Mine Detection Dogs: Training, Operations and Odour Detection. Geneva International Centre for Humanitarian Demining (GICHD), Geneva, 2003, pp. 53–107. 2. http://www.apopo.org/newsite/content/, visited 12/11/05.

PART III

EXAMPLE SENSING TECHNOLOGIES

CHAPTER 9

EXPLOSIVES DETECTION BASED ON AMPLIFYING FLUORESCENCE POLYMERS COLIN CUMMING Nomadics, Inc., an ICx Technologies Company.

9.1

INTRODUCTION

One of the most promising technologies for explosives detection is built upon fluorescent polymers developed at the Massachusetts Institute of Technology (MIT) by Professor Timothy Swager. When excited by a light source, the polymers fluoresce. However, should a molecule of explosive bind with the polymer, the fluorescence is quenched. Further, because the polymer can be structured so that the polymeric chains are electronically linked, a binding event anywhere along this “molecular wire” of conjugated polymer quenches the fluorescence all along the polymer network. Consequently, even a single binding event can result in a reduction of overall fluorescence that is readily detectable with relatively simple and inexpensive optics. For this reason, the technology is referred to as amplifying fluorescent polymer (AFP) [1]. The fluorescence-quenching phenomenon is clearly shown in Figure 9.1. Both vials contain AFP in water and are being exposed to an excitation source. However, the vial on the right also contains particles of TNT; as a result, the fluorescent emissions have been shut down. The use of MIT’s AFP for explosives detection has been licensed to Nomadics, Inc. The company has developed a family of explosives detection systems based on the AFP sensor platform and is expanding this collection of devices for additional applications. Because these systems emulate the sensing of explosives demonstrated by dogs, the Nomadics detection technology is called Fido®. Thin films of AFP materials coated onto a suitable substrate form the sensory element of the Fido explosives detection systems. Through extensive testing, Fido has demonstrated the detection of explosives at levels of concentration that are three orders of magnitude less than those detectable by the most sensitive commercially Trace Chemical Sensing of Explosives, Edited by Ronald L. Woodfin c 2007 John Wiley & Sons, Inc. Copyright

195

196


Figure 9.1 Two vials containing the amplifying fluorescent polymer (AFP) that has been developed to detect nitroaromatic explosives. They fluoresce when excited by light of a given wavelength as shown by the vial on the right. The vial on the left contains trace amounts of TNT, which has quenched the fluorescence.

available laboratory instruments [2]. The Fido sensor is the only trace chemical vapor sensor that has demonstrated the ability to detect explosives with performance levels similar to that of trained canines. High explosives (HE) characteristically have very low vapor pressures. Thus, in order to detect explosive by means of its chemical signature, trace sensors typically sample large volumes of the low-concentration analyte in order to collect sufficient material to ensure detection. Fido sensors use a different approach. Rather than collecting large samples of potential analyte or preconcentrating samples in order to obtain enough explosive to generate a detection signal, Fido relies on the amplification provided by the AFP so that an extremely small amount of analyte—a few femtograms of explosive—creates a detectable signal. While many conventional detection methods use other approaches to amplification, such as electronic amplification, these are generally ineffective for detecting vapors from explosives. Because such amplification occurs subsequent to the transduction step, both the signal and any noise present are amplified, providing no advantage in terms of sensitivity or minimum detection level. Fido’s AFPbased technology is highly significant in that it exhibits intrinsic amplification so that the transduction process is amplified. 9.1.1

AFP Principle of Operation

Structurally, the Fido polymers consist of a conjugated backbone with rigid, threedimensional pentiptycene groups (Fig. 9.2). The rigidity and three-dimensional

INTRODUCTION

197

C14H29 O

n

O C14H29

Figure 9.2 Chemical structure of the AFP (left) and a representation of the physical structure that has been engineered to bind to vapors of nitroaromatic explosives (right).

structure of the polymer forms cavities in the sensor films that accommodate small molecules and helps to enhance diffusion of small molecules into the films. Because of the overlap among polymer backbones, there is a risk that the polymer will quench itself due to π stacking. However, the AFPs have been engineered so that the same structural features that help create cavities for analyte capture also minimize self-quenching [3]. The polymer depicted in Figure 9.2 is only one of several dozen AFPs that have been synthesized for explosives detection, each with slightly different responses to target analytes. Because the performance needs vary for different explosives detection applications, these different AFP formulations seek to optimize the material’s performance characteristics for specific needs, such as the attachment of the material to the substrate used, the adsorption of analyte, duration of polymer operational life, temperature stability, and mechanical robustness. 9.1.2

AFP Technology

The sensory mechanism of AFP is not unlike that employed by conventional fluorescent chromophores. Typically, individual chromophores fluoresce upon excitation by a light source. However, if the chromophore binds with a target analyte, then excitation is not possible and the chromophore does not emit photons. This is illustrated in the top panel of Figure 9.3. The signal amplification exhibited by the Fido polymers is largely the result of the ease with which excited electrons can migrate along the AFP backbone. When thin films of AFP absorb photons of light, the excited state electrons or excitons form and begin propagating along the polymer backbones due to their higher energy. This energy transfer is highly efficient in AFP because the polymer’s electrons are extremely delocalized. Consequently, a single exciton propagates through many polymer repeat units during its excited state lifetime. These polymer units contain electron receptor sites, and eventually the exciton is likely to become trapped in a receptor so that it transitions back to a ground state, resulting in fluorescence. [3]

198


Traditional chemosensor (isolated chromophores): Binding of an analyte molecule quenches only the chromophore to which it binds. hn

Legend

hn

hn′

hn′

hn′ +

= Emissive polymer = Quenched polymer

−

= Analyte molecule

hn′

hn = Light at excitation wavelength

hn′

hn′ = Light at emission wavelength

hn′

hn′

Receptors wired in series (molecular wire arrangement): A single molecular binding event quenches the fluorescence of many chromophores, amplifying the chemosensory response by orders of magnitude. hn

hn′ + n

−

hn n

Figure 9.3 Comparison of conventional fluorescent chromophore sensors versus the molecular wire response of the AFP.

When an electron-deficient (i.e., electron-accepting) molecule binds to one of the polymer’s receptor sites, a low-energy trap is formed. If an exciton migrates to the site of the bound electron-deficient molecule, the exciton will become trapped and unable to return to a ground state. Thus, no fluorescence will result from the excitation event. One key to the amplification provided by the AFP technology is that once exciton migration has been halted anywhere along the polymer chain, fluorescence is prevented along the entire chain. This is the mechanism illustrated in the lower panel of Figure 9.3. In both examples, the same number of chromophores are present when a single binding event occurs, but the reduction in fluorescence is substantially greater in the case of the AFP chain. [3] Another factor in the enhanced sensitivity of AFP is the extent of exciton propagation. Individual excitons are not confined to the polymer chain upon which the original photon absorption event occurred. Excitons propagate three dimensionally and follow a “random walk” path, which means that they can pass through the same polymer repeat unit more than once. Because an exciton migrates through many receptor sites, the probability that it will encounter an occupied receptor site is greatly increased. Hence, the likelihood that a photon absorption event will result in fluorescence is greatly reduced even when small amounts of analyte have bound to the polymer receptor sites. While in theory an AFP chain of N polymer repeat units would be N times more sensitive than the same number of monomer chromophores, in practice the Fido polymers exhibit an effective amplification in response of between 100 and 1000 times as compared to monomeric quenching mechanisms. [4] The Fido polymers have also proven to be quite selective in responding to specific analytes. This characteristic has been engineered into the materials via

INTRODUCTION

199

three principal paths. The first path is through the reduction of steric constraints. Because fluorescence quenching depends on the transfer of electrons from the electronically excited polymer to the target analyte, measures must be taken to enhance the transfer of electrons. The rate of Förster energy transfer is greatly influenced by the distance between the electron donor and the acceptor, with the rate varying with the distance between the donor and acceptor to the inverse sixth power. Small molecules, such as the target explosives, fit into the cavities within the AFP films and, as electron acceptors, come into close proximity to the electrons migrating along the donor polymer backbone. Larger molecules are excluded, greatly reducing the possibility that these nontargeted analytes will quench the polymer. A second mechanism to enhance selectivity is the electrostatic complementarity between the polymer and target analytes. The polymers, which are electron rich, appear to bind reversibly to electron-deficient explosives through an electrostatic-type interaction. Compounds that have a strong affinity for the polymer (i.e., those that have a large binding constant, Kb ) are more likely to bind strongly with the polymer. To enhance binding of target compounds, receptor sites that are electrostatic mirror images of the target molecule have been specifically synthesized into the polymer backbone. A third avenue to enhanced selectivity is the fact that, for electron transfer to the bound molecule to occur spontaneously, the overall free energy change (G0 ) for this process must be negative. Thus, the bound molecule must have a standard reduction potential large enough to cause G0 to be negative. In equation form, the magnitude of fluorescence quenching (FQ) can be approximated as FQ ∝ (C)[exp(−G0 )2 ](Kb ) where (C) is the concentration of the quencher in the sample. Hence, for significant quenching to occur, an analyte must bind strongly to the film and have an appropriately high reduction potential. Consequently, compounds that bind weakly to the AFP film and those without favorable reduction potentials are not likely to result in significant quenches, except in very high concentrations. 9.1.2.1 Fido Detector Operation The Fido technology has been in development since 1998, leading to several products with more on the way. Implementing AFP as a sensor platform and integrating it into an electronic and optical system capable of highly sensitive and selective detection of explosives under a number of scenarios has been an iterative process. While numerous optical geometries and sampling configurations have been developed, Figure 9.4 shows the general configuration of components prominent in most designs. The configuration illustrated is built around a glass capillary whose inner surface is coated with AFP material. The device is configured in such a way that the system pump pulls air from the environment through the capillary. Because molecules of many explosives tend to stick to surfaces, the inlet is heated to ensure that analyte does not become trapped in the inlet. The heat also helps to improve system response [5].

200


Detector

Filter Fluorescence

Contaminated Air LED Activated Chromophore Coating Capillary Waveguide

Contaminated Air

Figure 9.4

Block diagram of the Fido explosives detection system.

As air moves through the capillary, the molecules of explosives diffuse to the AFP where they are adsorbed. External to the capillary is a light source. Various Fido models have used blue light-emitting diodes (LEDs), ultraviolet LEDs, and laser diodes. The light source is matched to the particular AFP used for a specific analyte and application. The excited AFP emits fluorescence, which the optics system waveguides to a detector aligned with the axis of the capillary. Both photomultipliers (PMTs) and semiconductor photodetectors have been used in various Fido models. The optics subsystem also includes a filter ahead of the detector to reject the excitation light. Reductions in fluorescence are processed by the Fido controller and software, and data indicating whether explosives are present in the sample are presented to the operator. The entire process from sample collection to notification of a sample containing explosives takes a matter of a few seconds [6]. Additionally, the AFP is recoverable—that is, once the analyte is removed (by airflow), the sensor is ready to analyze another sample. Recovery time is generally less than one minute. In field experience, AFP films have been exposed to analyte, recovered, and exposed again through dozens of cycles before the intensity of response diminishes enough to require replacement of the AFP cartridge [7, 8].

FIDO SENSITIVITY

9.2

201

HISTORY OF AFP AND FIDO®

The initial use of AFP for explosives detection came as part of the Defense Advanced Research Program Agency (DARPA) Dog’s Nose Program. The goal of this project was to identify technologies that could locate buried landmines based on the minute amounts of explosives that escape to the surface, emulating the sensory methods theorized to be used by dogs. Nomadics proved the feasibility of explosives detection with AFP in the laboratory using a rudimentary sensor. As a result of this genesis, the AFP platform became known as Fido. The first field tests of early prototypes took place in 1999 at Fort Leonard Wood, Missouri. Since that time, the technology has been further refined and tailored for use in various other explosives detection applications [8]. In 2001, the Fido system was modified to operate underwater and became known as the SeaDog. The U.S. Navy Office of Naval Research (ONR), under its Chemical Sensing in the Marine Environment (CSME) Program, funded the integration of the SeaDog with an autonomous underwater vehicle (AUV). The integrated system was able to map a plume of trinitrotoluene (TNT) in open water in real time. This was the first demonstration of the mapping of an explosive plume underwater in real time [9, 10]. The SeaDog technology proved to have applications for environmental monitoring as well. In 2002, the Strategic Environmental Research and Development Program (SERDP) funded Nomadics to develop a Fido configuration that could be used to monitor explosives contamination of groundwater. SERDP is a joint effort by the Department of Defense (DoD), the Environmental Protection Agency, and the Department of Energy focused on remediating former military sites for public use. SERDP has also funded development of Fido devices that can assist in locating unexploded ordnance in former range sites. Current Fido development efforts and new explosives detection applications are directed toward cargo, vehicle, and personnel screening for covert explosives and improvised explosive devices (IEDs). In short, Fido is being considered for use in nearly all explosives detection application in which dogs have been used [11]. 9.3

FIDO SENSITIVITY

The Fido system has demonstrated the ability to detect low (100,000) Nearly unlimited mass range Slow scan rate Cost very high

Fourier transform MS is an indispensable analyzer for large molecules such as proteins but is not well suited for fast and sensitive screening of target compounds for the purposes of threat analysis. 11.3.1.5 Hybrid MS It is often desirable to combine mass analyzers into hybrid instruments that can perform additional levels of mass analysis. There are two major reasons for doing this: (1) to obtain structure analysis by isolating specific ion masses and then causing them to break into identifiable fragment ions and (2) additional sensitivity can be gained by removing all ions except the target ions, thus creating a clean background on which a known fragment ion can be sensitively observed. The major tandem or hybrid MS instruments include: •

•

•

Triple quadrupole: In the standard configuration, the first quadrupole is used to select a specific ion mass, the second quadrupole is a collision cell that induces the selected ion mass to fragment, and the third quadrupole is used to analyze the fragments. The triple quadrupole is a very popular choice for very sensitive and quantitative analysis of complex mixtures (e.g., drug analysis). Quadrupole/TOF: A version of the triple quadrupole is to replace the third quadrupole with a TOFMS. This provides potentially faster analysis and higher mass resolution. Ion trap/TOF: This combination is advantageous because it enables the MSn capabilities of an ion trap with the high mass accuracy and speed of a TOF analyzer. The ion trap can either be a conventional hyperbolic Penning-type device or a linear trap device. Below we describe the deployment of the former type ion trap/TOF for explosives detection.

11.3.2

QitTof Mass Spectrometry

The quadrupole ion trap, time-of-flight (QitTofMS) analyzer was first developed by Lubman [12,13] and co-workers. QitTofMS components were first sold commercially by R. M Jordan Company and a full commercial instrument first

MASS SPECTROMETRY

227

Ion Optics

Discharge Ionization Source

Ion Trap

DI source • Selectivity toward high electron affinity (−NO2) • Minimum charge competition (1000:1 for known interferents)

Detector

Quadrupole Ion Trap (QIT) • Storage of ions from continuous source • MS/MS capable

Reflectron

TOFMS • Detection of all ions at once • Exact mass if desired (10 mDa)

Figure 11.2 Schematic of the QitTofMS detection system showing the individual components and their respective advantages. DI = discharge ionization.

introduced by Syagen [14,15] and then by Shimadzu [16]. A schematic of the QitTof configuration is shown in Figure 11.2. The principal role of the ion trap is to accumulate ions that are continuously generated by the ionizer. These accumulated ions are then pulsed into the TOFMS analyzer at relatively high repetition rates (e.g., 10 to 100 Hz). Because the TOFMS analysis is performed independently of the ion accumulation into the trap other than the short RF shutoff and pulse out time (about 50 to 100 μs), the duty cycle for collecting ions in the total analysis cycle is >99%. This is to be compared to an ITMS analyzer where duty cycle is typically

Trace Chemical Sensing of Explosives

Explosives and Chemical Weapons Identification

Explosives and Chemical Weapons Identification

Explosives

Explosives

Trace

Trace

Chemical Reviews: Modern Topics in Chemical Sensing issue 2

Trace

Trace

Trace

Trace

Trace

Trace

Trace

Ecotoxicology of Explosives

Trace

Trace

Trace

Trace

Trace

Aspects of Explosives Detection

Trace

Trace

Trace

Trace

Trace

Trace

Trace

Trace

Trace

Trace Chemical Sensing of Explosives

Explosives and Chemical Weapons Identification

Explosives and Chemical Weapons Identification

Explosives

Explosives

Trace

Trace

Chemical Reviews: Modern Topics in Chemical Sensing issue 2

Trace

Trace

Trace

Trace

Trace

Trace

Trace

Ecotoxicology of Explosives

Trace

Trace

Trace

Trace

Trace

Aspects of Explosives Detection

Trace

Trace

Trace

Trace

Trace

Trace

Trace

Trace

Trace

Recommend Documents