INTEGRATION BETWEEN THE LEBESGUE INTEGRAL AND THE HENSTOCK-KURZWEIL INTEGRAL Its Relation to Local Convex Vector Spaces ...
15 downloads
910 Views
3MB 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
INTEGRATION BETWEEN THE LEBESGUE INTEGRAL AND THE HENSTOCK-KURZWEIL INTEGRAL Its Relation to Local Convex Vector Spaces
SERIES IN REAL ANALYSIS
Vol. 1:
Lectures on the Theory of Integration R Henstock
Vol. 2:
Lanzhou Lectures on Henstock Integration Lee Peng Yee
Vol. 3:
The Theory of the Denjoy Integral & Some Applications V G Celidze &AG Dzvarseisvili translated by P S Bullen
Vol. 4:
Linear Functional Analysis WOrlicz
Vol. 5:
Generalized ODE S Schwabik
Vol. 6:
Uniqueness & Nonuniqueness Criteria in ODE R P Agarwal & V Lakshmikantham
Vol. 7:
Henstock-Kurzweil Integration: Its Relation to Topological Vector Spaces Jaroslav Kurzweil
Vol. 8:
Integration between the Lebesgue Integral and the Henstock-Kurzweil Integral: Its Relation to Local Convex Vector Spaces Jaroslav Kurzweil
Series in Real Analysis - Volume 8 INTEGRATION BETWEEN THE LEBESGUE INTEGRAL AND THE HENSTOCK-KURZWEIL INTEGRAL Its Relation to Local Convex Vector Spaces
Jaroslav Kurzweil Mathematical Institute of the Academy of Sciences of the Czech Republic
V f e World Scientific wb
Singapore • Hong Kong Sinqapore • New Jersey • London L
Published by World Scientific Publishing Co. Pte. Ltd. P O Box 128, Farrer Road, Singapore 912805 USA office: Suite IB, 1060 Main Street, River Edge, NJ 07661 UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE
British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library.
INTEGRATION BETWEEN THE LEBESGUE INTEGRAL AND THE HENSTOCK-KURZWEIL INTEGRAL Its Relation to Local Convex Vector Spaces Copyright © 2002 by World Scientific Publishing Co. Pte. Ltd. All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means, electronic or mechanical, including photocopying, recording or any information storage and retrieval system now known or to be invented, without written permission from the Publisher.
For photocopying of material in this volume, please pay a copying fee through the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to photocopy is not required from the publisher.
ISBN 981-238-046-9
Printed by Fulsland Offset Printing (S) Pte Ltd, Singapore
PREFACE
The approach to integration by Riemannian sums was rehabilitated in the fifties of the 20th century by a new interpretation of the concept of a "fine" or "6-fme" partition of the integration interval. It is well known that both the Lebesgue integration and the Henstock-Kurzweil integration can be obtained by the same method, only the integration bases are different. The concept of an integration basis y is very flexible and results in a rich class of ^-integrations. To every integration basis 3^ there corresponds the vector space Py of primitives of 3^-integrable functions (on a fixed interval / = [a, b] C K), a concept of Ey-convergent sequence of functions from Py, and ULci^y) which is the finest locally convex topology on Py such that every Ky-convergent sequence is convergent in (Py:ULc(^y))Lebesgue integration is obtained by a suitable choice of y, y = C. Then Pc is the space of absolutely continuous functions and ULC(^-C) is induced by the norm ||.F||var = vari*1. Hence (Pci^Lci^c)) is a complete space. If y — TiK, then HenstockKurzweil integration is obtained: The topology Uici^wc) is induced by the norm H-FHsup and (PHKMLC^HK)) is not complete. The problem whether (Py ,UL,c(J&y)) is complete is the central problem of this book. A theory is developed which gives an answer for a broad class of J^'s and to an extended problem which includes integrations introduced by Bongiorno and Pfeffer in 1992 and by Bongiorno in 1996.
V
PREFACE
VI
Topics connected with the Riemann approach to integration were reported and discussed in the Seminar on Differential Equations and Real Functions of the Mathematical Institute of the Academy of Sciences of the Czech Republic since the beginning of this approach. I wish to thank the participants of the seminar for their contributions and comments. I express sincere thanks to J. Jarnik and S. Schwabik, who read the manuscript and suggested several improvements. I am grateful to S. Schwabik who encouraged me and transformed the manuscript into the camera ready form. The research which resulted in publishing this book was supported by the grant No. 210/01/1199 of the Grant Agency of the Czech Republic.
Prague, March 2002
Jaroslav Kurzweil
CONTENTS
Preface
v
0. Introduction
1
1. Basic concepts and properties of ^-integration
9
2. Convergence
21
3. Convergence and locally convex spaces
32
4. An auxiliary locally convex space
42
5. £-integration
52
6. .M-integration
69
7. Noncompleteness
76
8. 5-integration
86
9. ^-integration
104
10. An extension of the concept of ^-integration
109
11. Differentiation and integration
116
References
135
List of symbols
137
Index
139
vn
0
INTRODUCTION
The approach to integration which is based on approximation of the integral by Riemannian sums is rather flexible. If the set of partitions which are used in the formation of Riemannian sums is rich then Lebesgue integration is obtained. On the other end of the spectrum a poor set of partitions leads to an integration which is called Henstock-Kurzweil and which is equivalent to Denjoy integration in the restricted sense and to Perron integration. In this book integrations are studied for various sets of partitions. If y is a set of partitions we denote by Py the set of primitives of 3^-integrable functions. For every 3? some sequences Fi G Py are called Ey- convergent IF"
to a limit FQ G Py, Fi —> FQ. Therefore there exists a unique locally convex topology ULc(^y) on Py which is the finest one among locally convex topologies T on Py with the property IF
that Fi —> Fo implies that Fi —> FQ in (Py,T). The problem whether (Py,Uic(^y)) is complete is crucial for this book; the answer depends on y. Let I = [a,b] C R. A set A = {(ti,Ai);i = 1 , 2 , . . . , A;} is called a partition in I if k G N and if (0.1) U G / f o r i = 1,2,...,*;, (0.2) Ai C I is a figure, i.e. a finite union of closed intervals, % =
i , 2 , . . . , K,
(0.3) figures Ai,Aj are nonoverlapping for i ^ j (i.e. \Ai D Aj\ = 0 for i ^ j where \E\ is the Lebesgue measure of 1
2
INTEGRATION BETWEEN L AND H - K
E C R). A is called a partition of / if (in addition) it
|J At = I.
(0.4)
Denote by C the set of partitions in / and by TiK, the subset of C which consists of A such that (0.5) Ai is a closed interval, i = 1, 2 , . . . , k, (0.6) U eAi,i
=
l,2,...,k.
+
Let ( : I -* R - A G £ is called (-fine if (0.7) At C (U - {(U),U + C(tx)) for i =
l,2,...,k.
f : I —» R is called Tf/C-integrable (£-integrable, respectively) if there exists j G R and for every £ G R + there exists ( : / -> R + such that
| 7 -^/(^)l^|| yk), k (=. N. The concept of ^-integration is an extension of the concept of 3^-integration and is studied in Chapter 10. Noncompleteness results are obtained if X = (<S(Ai), «S(A2),(Afc),... ), \ k £ A for k € N and if X = (K(u>i),K(uj2),K(ujk),...), uk € ft for k keN. If \k(a) = 2 a iora>0,X = (<S(A1),<S(A2),«S(Afc), • • •) then ^-integration is the integration which was introduced in [B
8
INTEGRATION BETWEEN L AND H - K
1996], if Lok(a) = 2~k for a > 0, X = ( f t ^ ) , ^ ^ ) , - ^ ^ , . . . ) then ^-integration is the integration which was studied in [B-Pf 1992]. Noncompleteness results in Chapters 9 and 10 are obtained by comparison with <S(A)-integration for a suitable A € A. 3^-differentiation is introduced in Chapter 11 and a general result concerning the relation of ^-differentiation and ^ - i n t e g ration is obtained. Chapter 11 is concluded by a specialization to £-integration. This Chapter will be closed by a list of four parts of the book which are not necessary for understanding the text which follows them: 1. Sections 4.7, 4.8, pp. 49-51 (a necessary and sufficient condition for completeness of (Py,ULc(Qy)) for a subclass of 3>'s); 2. Sections 5.8 - 5.11, pp. 62-68 (relations between convergences in (PC,ULC(QC)), in Qc and in Ec); 3. Chapter 6, pp. 69-75 (A^-integration); 4. Sections 8.9 - 8.12, pp. 91-103 (dependence of 5(A)integration on A).
1 BASIC CONCEPTS PROPERTIES OF
AND
^-INTEGRATION
1.1 N o t a t i o n . By R, R + , N, Z we denote the set of reals, the set of positive reals, the set of positive integers, the set of integers. For E C R, t G R let i n t E , c\E, \E\, d i a m E , d i s t ( t , £ ) be the interior of E, the closure of E, the outer Lebesgue measure of E, the diameter of E, the distance of t from E. Let N be the set of N C R such that |7V| = 0. By intervals we mean compact nondegenerate intervals in R, e.g. L = [c,d]. An open interval is denoted by (c, d) while [c, d), (c, d] are semiopen intervals. A figure is a finite union of intervals. T h e symbol #M stands for the number of elements of a finite set M. Let I — [a, b] be an interval. If K C I , then Iv(A') is the set of intervals L G K and Fig(A') is the set of figures A C K. We shall write Iv, Fig instead of Iv(J), Fig(J). Let K C I. A finite set A = {(i;,Aj); z = l , 2 , . . . , f c } , shortly A = {(i, A)} is called a partition in K if (1.1) R be ^-i-ntegrable and iet F : Fig —> R be its primitive. Then F is additive. Proof. Let KUK2 £ Fig, 1 ^ n K2\ = 0, j £ N. Let 6 £ D* correspond to / by Definition 1.3. There exist A^ = {(£, A)} £ y(Ki,Ki,6(j, •)), Ai being a partition of A^, z = 1,2. Then
(1.11)
|F(/v 2 )-5]/WI^I E be ^-integrable in I, F : Fig —> R being its primitive. Let 6 G -D* correspond to f by Definition 1.3 and let j G N, A = {(£, A)} G y(I,I,6(j,-)). Then (1.13)
|£(F(A)-/( ( A ) + ] T F ( c l L m ) . A
m€M
BASIC CONCEPTS
15
Therefore (cf. (1.16), (1.17)) | £ ( F ( A ) - f(t)\A\)\
< 2-> + # M • 2 - ' " - '
A
and (1.13) holds since p may be arbitrarily large. Let A1 = {(t,A)eA-F(A)-f(t)\A\>0}, A 2 = A \ Ai. By (1.13) we obtain £
\F(A) - / ( 0 | A | | = | £ 0 F ( A ) - f(t)\A\)\
Ai
< 2~\
Ai
J ] |F(A) - /(0|A|| = | £(F(A) ~ f(*)\A\)\ < 2"J" A2
A2
so that (1.14) holds. The proof is complete 1.8 Definition. Let H : Fig -* R, s € J. if is called J continuous at s if for every e £ R + there exists < e , a < % + 2 , 5 ) , 2 a | / ( s ) | < 2"-''- 1 . By (1.14) we have J2 \F(A)\
< •2-^-1 + \f{s)\J2
A
\A\ < 2-i-1
+ \f(s)\2 R. Py is a vector space, Py C AdC (cf. Theorem 1.6, Theorem 1.9). 1.11 D e f i n i t i o n . Let H : Fig -> R or H : Iv - • R, s <E / , (3 £ R. H is called differentiable at S and /? is the derivative of H at s if for every e <E R + there exists R be y-integrable be its primitive. Then (1.18) F(t) exists almost everywhere, erywhere, f is measurable.
F(t) = f(t)
and let F almost
ev-
(1.18) is a consequence of [K 2000], Theorem 1.17 since 7i)C C y and / is 7Y/C-integrable. 1.13 L e m m a . Let h : I - • R, N G M, e G R+. Then exists ( : I —•> R+ such that
there
£>(0PI<e A
forA = {(t,A)}€£(I,JV,C). # m i (cf. [K 2000], Lemma 1.15). For i G N there exist open sets d CR such that N C Gt, \G{\ < e2~2i. Put E0 = 0, Et = {t G iV; |fc(*)| < 21}. L e t ( : I ^ R + fulfil (*-((*),*+((*)) C Gt for t e Ei\ Ei-i, i G N. Then £ does the job. 1.14 T h e o r e m , (i) Let / , # : I —> K be ^-integrable, (1.19)
(3>) / / ( s ) d 5 = (y) Ja
I g{s)ds for < G / . •/ a
Then f = g almost everywhere. (ii) Let / , g : / —> R. Assume £ba£ / is [V-mtegrabie and / = g almost everywhere. Then g is y-integrable and (1.19) holds. Proof, (i) is a consequence of Theorem 1.12 and (ii) follows by Lemma 1.13. 1.15 L e m m a . Let K G Fig, / : K -»• R, F : Fig(A') -> R. Assume that F is additive. Then the following conditions are equivalent: (1.20) / is y-integrable
and F is its
primitive,
18
INTEGRATION B E T W E E N L AND H - K
(1.21) there exists 9 £ D* such that
Y,\F(A)-f(t)\A\\ A
for j £ N, A = {(t, A)} £ y(K, K, 0(j, •)), (1.22) there exist M £ M and r\ £ D* such that (i)
£|F(A)-/(*)|A|| R be additive and fulfil (2.14) and (2.15). Then F is the y-primitive of f.
Proof. Let M be defined by (2.15). By Lemma 2.7 there exists £ € D* such that
2^/jr,^ *(M)
z
e
for j € N, 0 = {(t, A)} € y(I,M,Z(J,-))Hence (1.22) holds if r](j,t) < min{0(j + l,t),£(j + 1,*)} and Lemma 2.8 is a consequence of Lemma 1.15. T h e proof is complete. 2.9 T h e o r e m .
U QyW) = py9 R7 Go '• Fig —> EL Assume that (2.17)
G 8 (A) -»• G0(A)
for A e Fig,z -> oo.
T i e n there exists go : i" —• R. such that (2.18)
^i —• ^o ha measure for i —> oo.
Note. The theorem has several consequences: (2.19) there exists a subsequence i(k), k £ N and Mo (E TV such that 9i(k)(t) -* #o(0 for k -+ oo a.e., (2.20) G 0 is additive (by (2.17), since G; are additive), (2.21) Go e Qy(0) (since we may write G ,-(£), oo by (2.17), (2.19) after having made a suitable choice of M ) , (2.22) Qy(8) is closed in AdC (cf. (2.21), (2.17), (2.19)), (2.23) Gi(fc) —>• Go (since G m e Qy{6) for m = 0 , 1 , 2 , . . . and (2.19) holds, there exist M G Af, rj G D* such that (2.9) (i) and (iii) hold. Moreover, (2.9) (ii) follows from (2.14) by Lemma 1.13 so that (2.23) holds by Lemma 2.3). Proof. In order to prove (2.18) it is sufficient to show that the sequence gi is Cauchy in measure. Let us suppose the opposite, i.e. that there exists a 6 R + such that for every r € N there are p = p(r), q = q(r) such that p,q > r and (2.24)
\{tel;
\gp(t)-gg(t)\>a}\>2a.
CONVERGENCE
27
Let h £ N be chosen such that 2~h+2b-^}.
Then J 5 ^ C Xi forfc< / and (Jj.^Xfc = / so that lim \Xk\ = b — a. k—>oo
Find k eN such that \Xk\ > b — a — a. Denote EVA = {te I ; |flf„(*)-*}By Theorem 1.12 2 00. Denote by Qy the set of couples ((F 8 , i G N ) , F 0 ) fulfilling (2.31) and (2.32). Qy is called a convergence and the sequence Pi, z G N is said to be convergent
to FQ in Qy, Ft —> Fo.
2.15 L e m m a . Let F{ ^U F0, Gt ^ Then (2.33)
(2.34)
(2.35)
G0, a G M.
Pi(jfe) — • Po for any subsequence
i^ + G ^ F o
+ Go,
a f ^ a F o .
i(k), k = 1, 2, 3 . . . ,
CONVERGENCE
31
Proof. By Definition 2.14 there exist 771,772 £ D* such that FleQy(r]l),GleQy(r]2)
for i = 0 , 1 , 2 , . . .
and 11-ft - -Po| I sup —> 0, ||Gi - Go||sup —>• 0 for z —• 00. Obviously, (2.33) is valid. (2.34) holds since there exists 773 £ D* (cf. Lemma 2.13) such that Ft + Gi GQyM
for i = 0 , 1 , 2 , . . .
and H^ + G i - F o - G o l l s u p - O . Let a £ R, k £ N, | a | < 2*. Making use of Lemma 2.13 repeatedly we conclude that there exists 0 £ D* such that 2*ft GQy(O)
for i = 0 , 1 , 2 , . . . .
Hence a f t £ # y ( 0 ) for z = 0 , 1 , 2 , . . . and (2.35) holds since \\aFi — a-Fo||sup —• 0. T h e proof is complete. 2.16 T h e o r e m , (i) If F{ ^
F0, then ft ^
(ii) If Fi — • Fo then there exists a subsequence such that Fi(k) ^
F0. i(k), k £ N
FQ.
Proof. Let Fi —^» -Fo- By Definition 2.2 there exist / m , m = 0 , 1 , 2 , . . . and 6 £ D* such that (2.6) and (2.7) hold. Moreover, there exists 9 £ D* such that
0(j,t) fc. Let T\,T2 be topologies on F . 7i is called ./mer than T2 if 7i D T 2 . From now on let Y be a vector space over R. Let T be a topology on F . If y G F , t/ G T implies that y + U C T, then T is called invariant with respect to shifts. 32
CONVERGENCE AND LCS
33
Let 53 C 2 y fulfil (3.4)
(3.5)
(3.6)
03 ^ 0, 0 i 03,
every V from 03 is radial, circled and convex,
for Vi, V2 G 03 there exists V3 G 03
such that v3 c Vi n v2, (3.7)
(3.8)
for V G 03 there exists Vi G 03 such that 2Vi C V,
for x G F , x 7^ 0 there exists V G 03 such that x ^ V.
Let V(03) be the set of U C Y such that x G U if and only if x + V C U for some V G 03. V(03) is a topology on F which is invariant with respect to shifts and Hausdorff. V(03) is called the topology induced by 03. (Y,T) is called a locally convex space if there exists 03 such that T = V(03). Let r : Y -+ [0, oo). If (3.9)
(3.10)
r(x + y) < r(x) + r(y)
for
i,t/6F,
r ( ^ x ) = |)u|r(x) for x G F, /u G R
then r is called a seminorm. Let i J b e a set of seminomas. Assume that for every x G F there exists r G -R such that r(x) > 0. The set (3.11)
{x G F ; r,(x) < £ for i = 1 , 2 , . . . , k}
INTEGRATION BETWEEN L AND H - K
34
where k € N, r; G R, e G R + is convex, circled and radial. Let 2G be the set of sets (3.11). 2U fulfils (3.4) - (3.8) and V(2U) is called the topology induced by R. Let P* C AdC be a vector space. Let A be a set the elements of which are couples ((Fi,i G N),Po) where Fm G P* for m = 0 12 V,
-L, - : , . . . •
We shall write (3.12) Ft - ^ F0 instead of ((Fi,i G N ) , P 0 ) G A and assume (3.13) i f P , - ^ P 0 , G 4 - ^ G 0 , a , / ? G E t h e n ( a ^ + / ? G O ^ (aF0+/3G0),
aFi -^
aF0,
(3.14) if Ft - ^ P 0 then ||P, - P 0 | | s u p - • 0. A is called a convergence on P* and the sequence Pj-,i G N is said to be convergent
to F0 in A, Fi
A
> P0.
3.2 L e m m a . Let P* C P2* C AdC, Pi*,.P2* being spaces. Let A* be a convergence on P ? , i = 1,2. T i e n (3.15)
Ax C A 2 iff P, - ^ P 0 implies Ft-^
vector
F0.
This is just a reformulation of (3.12). 3 . 3 D e f i n i t i o n . Let P * C AdC be a vector space, let A be a convergence on P * and let T be a topology on P * . T is called tolerant to A if Ft - ^ F0 implies Ft -> P 0 in
{P*,T).
3.4 N o t a t i o n . Denote by 7^ up the topology on AdC which is induced by the norm || • ||Sup- Let P * C AdC be a vector space and let A be a convergence on P * . By U L C ( A ) let us denote the set of topologies T on P * such that (3.16) (3.17)
(P*,T)
is a locally convex space, y
is tolerant to A.
Note. U L C ( A ) / 0 since { 0 , P * } G U L C ( A ) .
CONVERGENCE AND LCS
35
3.5 L e m m a . Let P* C AdC be a vector space and let A be a convergence on P*. Then there exists ULC(A) G U L C ( A ) such that (3.18)
W L C (A) is finer than any T G U L c ( A ) .
3.6 N o t e . Uic{A) is unique. Moreover, ULC(A) dorff since it is finer than T SU p|p*.
is Haus-
Proof. If T G U L C ( A ) then (P*,T) is a locally convex space and there exists 03 G 2 P * fulfilling (3.4) - (3.8) such that T = V(03). For a finite set 7i = V(03i), T2 = V(03 2 ), . . . , Tk = V(03fc) G U L C ( A ) and Vi G 3 such that F G ( 1 - £ ) F SO that F + — F C ( 1 - — )V. m m Let / € N, 2' > m. Making use of (3.7) repeatedly we conclude that there exists V £ 53 such that 2lV C V. Hence
i>c2 _ 'v c — v m and F + VC(1-—)V
m
CV0
so that Vo is open. The proof is complete. 3.9 T h e o r e m . Let T be a topology on Py. tolerant to Qy if and only if it is tolerant to Ey.
Then T is
CONVERGENCE AND LCS
Corollary. ULC{Ey)
= VLC(Qy),
ULC(Ey)
37
=
ULC(Qy).
Proof. If T is tolerant to Qy then it is tolerant to Ey by Theorem 2.16 (i). Assume that T is tolerant to Ey and that there exist Fm G Py, m = 0 , 1 , 2 , . . . such that Fi —> FQ but the sequence Ft, i G N is not convergent to F0 in ( P y , T ) . Hence there exist U G T and a subsequence i(k), k G N such that FQ £ U and (3.20) However, F^)
F i ( f e ) £ U for fc G N. ~~^ -Fo by (2.33) and Theorem 2.16 (ii) implies
that there exists a subsequence &(/), / G N such that Fj(fc(;)) —> FQ. Therefore there exists IQ G N such that F;(j.(/)) £ U for I > IQ, which contradicts (3.20). This contradiction proves that T is tolerant to Q y . T h e proof is complete. 3 . 1 0 L e m m a . For a G R + p u t 5 ( a ) = {F G AdC; | | F | | s u p < a}. Let 0 G 17 G ^ L c ( Q y ) , 0 £ D*. Then there exists a G R + such that Qy(0) D £( < 2~\ Hence Fi > 0 which is impossible since ULci'Q'y ) is tolerant to Q y . This contradiction makes t h e proof complete. 3.11 L e m m a . For E C AdC denote by c o n v F the convex hull of E. LetOeUe ULC(Qy)Then there exists ( : D* -> + R such that conv | J Q y ( 6 ) n B(C(6)) C U. seD*
INTEGRATION BETWEEN L AND H - K
38
Proof. Since 0 G U € WLC(Q;>;) and WLC(Q3>) = V(2J) for some R + such that | J Qy(6) n B(((6))
C Vo
and conv \J Qy{8) n # ( ( ( £ ) ) C Vo C V C f/. seD* The proof is complete. 3 . 1 2 L e m m a . For ( e D* define V>(C) : N x J -+ R+ by
(3.21)
^(C)0',0 = |C0" + 1,0-
Then (3.22)
il>:D*^
(3.23) V>(C) = * iff <M,t) = 26(j-l,t) and ((l,t) (3.24)
2Qy(C)
D*,
for j =2,3,4,...,t
> ((2,t)
C Qy(H0)
el
for t e I, for
C € #*•
Proof. (3.22) and (3.23) are immediate consequences of (3.21). Let F e Qy(0- Then
CONVERGENCE AND LCS
39
for j G N, A = {(t, A)} G y{I, I, CO', •)) and there exist M G AT and / : I —> R such that ^|F(A)-/(t)|A||" 1 + £
E. 4.2 T h e o r e m . Assume that (4.3)
P y C Ty,
(4.4)
Ty\py
(4.5)
TKF
(4.6)
for A' e Iv
is tolerant to
Qy,
£ Py for K € Iv ,F £ P y , there exists
I I F / c P b , , < K\\F\\y,„
K£ R
such that
FePy.se
Then (4.7)
ULC{Qy)
= Ty\Py.
T h e proof will be performed in several steps.
I.
INTEGRATION BETWEEN L AND H - K
44
4 . 3 L e m m a . Let K(i) = [ci,dt] G Iv, K(i + 1) C K(i) zeN, s e l , f]ienK(i) = {s}. Let Ft G Py fulfil
for
(4.8) Fi(A) = 0 if A G Fig , A C [a,ct] U [d,-, 6], « G N, (4.9) for e G R+ there exists a G R + such that
E l*K^)l ^ £ A
fori€N,A = {(s,A)}6y(I,{s}, R and Mt e JV for z G N such that
(4.10)
JX^I^+^J^ A
A
u
'
y
for i , ; G N, A = {(*,4)} G ? ( / , J , * ( j , •)),
(4.11)
^lF^)-/«(t)lAH^2"i A
for i,JGN, A = { ( t , A ) } G ^ J \ M , , ^ 0 V ) ) . Without loss of generality we may assume that (4.12) fi(t) = 0 for a < t < a, di < t < b and t = s, i G N, (4.13) Sk(j,t) < 6i(j,t) for i,j,ke N,k>i,tel.
AN AUXILIARY LCS
Put (4.14) ui(t) = dist(«, K(i)) for i e N, t e I and define 9 as follows: (4.15) 0U,t)=u1{t)
for j G N , t G / \ / i ( l ) ,
9(j,t) =min{u)t+1(t),6t(j foriJtN,
+ i + l,t)}
f € A ^ ) \ A ' ( « + l),
9(j,s) is so small that ^|Fi(A)|