Probability and Its Applications

Andrei N. Borodin
Paavo Salminen

Handbook of
Brownian
Motion – Facts
and Formulae
Second Edition
Probability and Its Applications

Series Editors
Davar Khoshnevisan
Andreas Kyprianou
Sydney I. Resnick

More information about this series at http://www.springer.com/series/4893

Andrei N. Borodin • Paavo Salminen

Handbook of Brownian
Motion - Facts and Formulae
Second Edition
Andrei N. Borodin Paavo Salminen
Steklov Mathematical Institute Åbo Akademi University
St. Petersburg, Russia Faculty of Science and Engineering
and Mathematics and Statistics
Department of Mathematics and Mechanics Turku, Finland
St. Petersburg State University
St. Petersburg, Russia

2015 corrected reprint of the second edition

ISSN 2297-0371 ISSN 2297-0398 (electronic)

Probability and Its Applications
ISBN 978-3-7643-6705-3 ISBN 978-3-0348-8163-0 (eBook)
DOI 10.1007/978-3-0348-8163-0

Library of Congress Control Number: 2015936694

Mathematics Subject Classification (2010): 60J65, 60J60, 60H05, 60H10, 60J55

Springer Basel Heidelberg New York Dordrecht London

© Springer Basel 2002, corrected printing 2015
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the
material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now
known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does
not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective
laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book are
believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors
give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions
that may have been made.

Printed on acid-free paper

Springer Basel AG is part of Springer Science+Business Media (www.birkhauser-science.com)

 CONTENTS
Preface to the first edition ix
Preface to the second edition xi
Preface to the corrected second printing of the second
edition xiii
Notation xiv

Part I: THEORY

Chapter I. Stochastic processes in general 1

1. Basic definitions 1
2. Markov processes, transition functions, resolvents,
and generators 3
3. Feller processes, Feller–Dynkin processes,
and the strong Markov property 5
4. Martingales 7

Chapter II. Linear diffusions 12

1. Basic facts 12
2. Local time 21
3. Passage times 24
4. Additive functionals and killing 27
5. Excessive functions 32
6. Ergodic results 36

Chapter III. Stochastic calculus 39

1. Stochastic integration with respect to Brownian motion 39
2. The Itô and Tanaka formulae 43
3. Stochastic differential equations – strong solutions 45
4. Stochastic differential equations – weak solutions 47
5. One-dimensional stochastic differential equations 48
6. The Cameron–Martin–Girsanov transformation of measure 49

Chapter IV. Brownian motion 53

1. Definition and basic properties 53
2. Brownian local time 56
3. Excursions 59
4. Brownian bridge 66
5. Brownian motion with drift 70
6. Bessel processes 74
7. Geometric Brownian motion 80

v
vi CONTENTS

Chapter V. Local time as a Markov process 84

1. Diffusion local time 84

2. Local time of Brownian motion 87
3. Local time of Brownian motion with drift 93
4. Local time of Bessel process 97
5. Summarizing tables 102

Chapter VI. Differential systems associated to

Brownian motion 106

1. The Feynman–Kac formula 106

2. Exponential stopping 108
3. Stopping at first exit time 112
4. Stopping at inverse additive functional 116
5. Stopping at first range time 120

Appendix 1. Briefly on some diffusions 123

Part II: TABLES OF DISTRIBUTIONS OF FUNCTIONALS

OF BROWNIAN MOTION AND RELATED PROCESSES

Introduction 151

1. List of functionals 152

2. Comments and references 154

1. Brownian motion 159

1. Exponential stopping 159

2. Stopping at first hitting time 204
3. Stopping at first exit time 218
4. Stopping at inverse local time 235
5. Stopping at first range time 248

2. Brownian motion with drift 256

1. Exponential stopping 256

2. Stopping at first hitting time 301
3. Stopping at first exit time 315
4. Stopping at inverse local time 329
 CONTENTS vii

3. Reflecting Brownian motion 339

1. Exponential stopping 339
2. Stopping at first hitting time 361
4. Stopping at inverse local time 370

4. Bessel process of order ν 379

1. Exponential stopping, ν ≥ 0 379
2. Stopping at first hitting time, ν > 0 404
3. Stopping at first exit time, ν > 0 415
4. Stopping at inverse local time, ν > 0 426

5. Bessel process of order 1/2 435

1. Exponential stopping 435
2. Stopping at first hitting time 469
3. Stopping at first exit time 486
4. Stopping at inverse local time 499

6. Bessel process of order zero 513

2. Stopping at first hitting time 513
3. Stopping at first exit time 518
4. Stopping at inverse local time 525

7. Ornstein–Uhlenbeck process 528

1. Exponential stopping 528
2. Stopping at first hitting time 548
3. Stopping at first exit time 554
4. Stopping at inverse local time 563

8. Radial Ornstein–Uhlenbeck process 571

1. Exponential stopping 571
2. Stopping at first hitting time 587
3. Stopping at first exit time 593
4. Stopping at inverse local time 603

9. Geometric Brownian motion 612

1. Exponential stopping 612
2. Stopping at first hitting time 628
3. Stopping at first exit time 633
4. Stopping at inverse local time 639

Appendix 2. Special functions 643

Appendix 3. Inverse Laplace transforms 659

viii CONTENTS

Appendix 4. Differential equations and their solutions 662

Appendix 5. Formulae for n-fold differentiation 669

Bibliography 671

Subject index 682

 PREFACE TO THE FIRST EDITION

There are two parts in this book. The first part is devoted mainly to the proper-
ties of linear diffusions in general and Brownian motion in particular. The second
part consists of tables of distributions of functionals of Brownian motion and re-
lated processes.
The primary aim of this book is to give an easy reference to a large number of
facts and formulae associated to Brownian motion. We have tried to do this in
a “handbook–style”. By this we mean that results are given without proofs but
are equipped with a reference, where a proof or a derivation can be found. It is
our belief and experience that such a material would be very much welcome by
students and people working with applications of diffusions and Brownian motion.
In discussions with many of our colleagues we have found that they share this point
of view.
Our original plan included more things than we were able to realize. It turned out
very soon when trying to put the plan into practice that the material would be too
wide to be published under one cover. Excursion theory, which most of the recent
results concerning linear Brownian motion and diffusions can be classified as, is only
touched upon slightly here, not to mention Brownian motion in several dimensions
which enters only through the discussion of Bessel processes. On the other hand,
much attention is given to the theory of local time. The Markov property of local
time, the so called Ray–Knight theorems, are discussed in a chapter of their own,
and one can find a large number of distributions associated with local times in
the tables. The concept of local time has proved to be of importance in many
applications of diffusions; for instance, in economics, mathematical finance, and
optimal control. We hope that the material presented in this book will be useful
in these and other applications.
In the tables there are more than 1550 numbered formulae associated directly
or via some transformations to Brownian motion. The systematic structure of the
tables is explained in Introduction on p. 1201 . Some of the formulae are certainly
obvious, tautological, and perhaps too complicated to be of any real use. In spite
of this a decision was made to keep the structure and not to expand the tables with
more “exotic” formulae. The main reason for this decision was that nothing like
these tables has been published before, and so we think that one should start from
the “bottom”. We also believe that the structure of the tables makes them easy to
1 On p. 151 in this edition.

ix
x PREFACE TO THE FIRST EDITION

use, especially for people who do not deal with Brownian motion or diffusions on
a daily basis. It is, further, our hope that the presentation has some pedagogical
merits when used in teaching and by students.
As mentioned above results in this book are usually given without proofs. There
are many superb research monographs and text books, where diffusions and
Brownian motion play the central role, and where proofs of the theorems in our
book are given. In particular, we have been using books by Dynkin (1965), Blu-
menthal and Getoor (1968), McKean (1969), Ito and McKean (1974, 2nd ed.),
Gikhman and Skorokhod (1974, 1975, 1979), Williams (1979), Ikeda and Watan-
abe (1981), Knight (1981), Chung and Williams (1983), Doob (1984), Rogers and
Williams (1987), Revuz and Yor (1991), and Karatzas and Shreve (1991, 2nd ed.).
Many of these, e.g., Doob (1984), Rogers and Williams (1987), Revuz and Yor
(1991), Karatzas and Shreve (1991, 2nd ed.), give, besides the proofs, also histor-
ical comments and notes concerning the origin and development of the theorems.
Therefore, whenever possible, a reference is given to a monograph. We want to
stress heavily (the obvious thing) that the presentation in Part I is not everywhere
meant to be mathematically complete. To achieve this one must in several places
study the different sources given in the text and via them, perhaps, also the original
research papers.
Paul McGill, Jim Pitman and Marc Yor read parts of the material of the first and
latter versions. We thank them for their comments, criticism and encouragement.
We are also grateful to Göran Högnas and Juha Vuolle–Apiala for their comments
to a more final version.
During the course of this work we have received financial support from the
Academy of Finland, the Research Institute of the Åbo Akademi Foundation, the
Foundation of the University of Turku, and the Russian Foundation of Fundamental
Research, grant No. 93-01-01454. This support is gratefully acknowledged. It is
also our pleasure to thank the mathematical departments at the Steklov Institute,
St. Petersburg’s branch, the University of Turku, and the Åbo Akademi University
for their hospitality.
Our special thanks goes to P.E. Baskakova for helping us with the complicated
typing work.
Finally, we want to thank the staff of Birkhauser for their patience when this
project was dragging in time, and for their excellent work when preparing the real
thing. It is also our pleasure to thank the copy–editor whose careful work made
the text more readable.

St. Petersburg, Russia, and Åbo (Turku), Finland, March 1996

A.N. Borodin P. Salminen

 PREFACE TO THE SECOND EDITION

Compared with the first edition published 1996 this second edition is revised
and considerably expanded. The number of new formulae in the tables is more
than 1000 and some of the old formulae in the tables and especially in Appendix 1
have found a new equivalent form.
It is our feeling that the first edition has been used, in some extent, for problems
arising in financial mathematics. To meet better requirements from this field we
have included in the theory part a new subsection on some recent results on geo-
metric Brownian motion and in the formula part there is also a new section with
ca. 200 formulae devoted to geometric Brownian motion.
Section V on Ray–Knight theorems has undergone much revision. Aim has been
to make it more readable and also to present more detailed explanations. We have
also included summarizing tables for ease of reference.
There is a new subsection in Section VI on Feynman–Kac formula concerning
functionals stopped at the first range time. This is the random time when the
maximum minus minimum of a process reaches a given level. We think that the
first range time is an important stopping time with many applications on different
fields. A number of formulae associated to the first range time of Brownian motion
are displayed in the formula part of the book.
New processes are added to Appendix 1 “Briefly on some diffusions”. In particu-
lar, we have studied radial Ornstein–Uhlenbeck processes and their squares. These
processes are stationary with interesting equilibrium probability distributions; like
gamma-, Rayleigh-, Maxwell- and χ2 -distributions. Radial Ornstein–Uhlenbeck
processes appear also in the formula part with ca. 160 new formulae.
We have also expanded the introduction to the tables with new references and
some explicit examples. We hope that these examples make it easier to use and
understand the results in Section VI. There are also three new appendices and the
old appendices have been expanded. The first new appendix is on formulae for
inverse Laplace transforms, the next one on the second order ordinary differential
equations and their fundamental solutions, and the third one on formulae for n-fold
differentiation of the product and composition of two functions. The appendix on
differential equations is needed when applying the Feynman–Kac formula, and the
appendix on differentiation is useful when computing, e.g., moments of a functional
of a random process by differentiating a Laplace transform.
This second edition has about 200 pages more than the first edition. However,
the added material covers more than that. To keep the book in a reasonable size
we have not included some formulae from the first edition to the second one. From

xi
xii PREFACE TO THE SECOND EDITION

our point of view the excluded formulae are simple consequences of others and can
be constructed easily, if needed.
We apologize sincerely for the misprints in the first edition and hope that the
positive experiences with the book have weighted more than the negative ones. We
have worked hard to correct all the misprints and other errors. However, due to
the nature of the book and the fast modern word processing technology it would
be too optimistic to claim that this second edition does not have any defects. We
therefore appreciate to have feedback from our readers also on this edition. Our
e-mail addresses are printed on the information page of this book.
We express our warm thanks to Marc Yor for encouragement, many concrete
proposals for improvements, pointing out misprints, and giving new references.
We are also grateful to Jim Pitman for comments and interest in this work. Andrei
Sudakov helped us with the secrets of the AMS-TEXpackage and we thank him for
his help.
Financial support from the Research Institute of the Åbo Akademi Foundation,
the Magnus Ehrnrooth’s Foundation (Finland), Vaisala Foundation (Finland), and
the Russian Foundation of Fundamental Research, grant No. 99-01-0112 are grate-
fully acknowledged.

St. Petersburg, Russia, and Åbo (Turku), Finland, October 2001

A.N. Borodin P. Salminen

 PREFACE TO THE CORRECTED SECOND
PRINTING OF THE SECOND EDITION

Our main aim with this printing has been to correct some typos and flaws still
remaining in the second edition. For a list of the corrections, see our home pages.
In this process it was very tempting to digress from the aim, include new material
and steer toward a third edition. Although the immense mathematical richness of
Brownian motion and many important applications would have made this possible
and, perhaps, desirable we feel that it is time to close the project. However, there
are small bonuses with the new printing, e.g., a discussion on perpetual integral
functionals of Brownian motion with drift in Chapter 4 No.’s 34 and 35 and a
presentation of CEV processes in Appendix 1 No. 27.
Financial support from the Magnus Ehrnrooth’s Foundation (Finland) and from
Svenska kulturfonden via Stiftelsernas professorspool (Finland) are gratefully ac-
knowledged.

St. Petersburg, Russia, and Åbo (Turku), Finland, June 2014

A.N. Borodin P. Salminen

xiii
 NOTATION

Chapters are referred to with Roman numerals. Inside the chapters we use
running Arabic numbering for the items. For example, V.6 is item 6 in the fifth
chapter, A1.1 is item 1 in the first appendix which is at the end of Part 1, and
A2.2 is item 2 in the second appendix at the end of Part 2. When referring to an
item in the same chapter the Roman numeral is omitted and we write, e.g., No. 6.

The symbol
:= means “is defined to be equal to”,
∼ means “is identical in law to”.

(Ω, F, P) – probability space.

ω – element in Ω.

t 7→ ω(t) – continuous function in the canonical framework, see I.12.

(Ω, F, {Ft }, P) – filtered probability space, see I.3.

{Ft : t ≥ 0}, {Ft } – filtration, see I.3.

{Ftc : t ≥ 0}, {Ftc } – canonical filtration, see I.12.

{Fto : t ≥ 0}, {Fto } – natural filtration, see I.4.

B(E) – Borel σ-algebra of subsets of E.

C(E) – continuous functions from E to R.

Cb (E) – continuous and bounded functions from E to R.

C0 (E) – continuous and bounded functions from E to R vanishing at infinity.

C(R+ , Rd ) – continuous functions from R+ to Rd .

Exp(α) – exponential distribution with parameter α.

τ ∼ Exp(α) – random variable τ is exponentially distributed with parameter α.

I – interval on R.

l, r – left and right, respectively, endpoint of I.

int I – interior of I.

C 1 (I) – continuously differentiable functions from I to R.

xiv
 NOTATION xv

C 2 (I) – two times continuously differentiable functions from I to R.

Cb2 (I) – bounded two times continuously differentiable functions from I to R.

BB (I) – bounded Borel-measurable functions from I to R.

Leb – Lebesgue measure.

ε{x} – Dirac measure at x.

δ0 (·) – Dirac δ-function.

Px – probability measure associated with a process started at x.

Ex – expectation operator associated with a process started at x.

θt – shift operator, see I.13.

1IA (ω) – indicator function.

Ex ξ(ω); A := Ex ξ(ω)1IA (ω) .

@

Ex ξ(ω); η(ω) ∈ dy := Ex ξ(ω); η(ω) < y dy.
@y

G – infinitesimal generator, see I.8, II.7.

m – speed measure, see II.4.

k – killing measure, see II.4.

s – scale function, see II.4.

p(t; x, y) – transition density with respect to the speed measure, see II.4.

Gα (x, y), G(x, y) := G0 (x, y) – Green function, see II.11.

ψα – fundamental increasing solution, see II.10.

ϕα – fundamental decreasing solution, see II.10.

f + – right derivative of f with respect to the scale function.

f − – left derivative of f with respect to the scale function.

`(t, x) – local time with respect to the Lebesgue measure.

%(t, z) = inf{s : `(s, z) > t } – inverse local time.

xvi NOTATION

L(t, x) – local time with respect to the speed measure, see II.13.

a ∨ b := max{a, b}.

a ∧ b := min{a, b}.

M̌t := sup{Xs : 0 ≤ s ≤ t } – maximum of the process X.

M̂t := inf{Xs : 0 ≤ s ≤ t } – minimum of the process X.

Ȟ(t) := inf{s < t : Xs = M̌t } – first location of the maximum before time t.

Ĥ(t) := inf{s < t : Xs = M̂t } – first location of the minimum before time t.

Hz := inf{t : Xt = z} – first hitting time of z.

Ha,b := inf{t : Xs ∈
/ (a, b)} – first exit time from (a, b).

L−1
γ – inverse Laplace transform with respect to γ.

Z t
f (t) ∗ g(t) := f (s)g(t − s) ds – convolution of the functions f and g.
0

sh, ch, th, cth – hyperbolic functions, see A2.1.

Γ – Gamma function, see A2.2.

Jν , Yν – Bessel functions, see A2.3.

Iν , Kν – modified Bessel functions, see A2.4.

Ai – Airy function, see A2.5.

Hen – Hermite polynomials, see A2.6.

Erf, Erfc, Erfi, Erfid – error functions, see A2.8.

Dν – parabolic cylinder functions, see A2.9.

M (a, b, ·), U (a, b, ·) – Kummer functions, see A2.10.

Mn,m , Wn,m – Whittaker functions, see A2.10.

F (α, β, γ, ·), G(α, β, γ, ·) – hypergeometric functions, see A2.11.

Ppq , Q
e qp – Legendre functions, see A2.12.

For further notations associated to special functions, see A2.

 Part I

THEORY

CHAPTER I

STOCHASTIC PROCESSES IN GENERAL

1. Basic definitions

1. Let (Ω, F, P) be a probability space, (E, E) a measurable space, and Σ an arbi-

trary set. A stochastic process is a family X = {Xt : t ∈ Σ} of random variables
Xt : Ω 7→ E. The space (E, E) is referred to as the state space of X. Given a
point ω in the sample space Ω, the mapping t 7→ Xt (ω) is called a trajectory or
realization or sample path of the process X. For the general theory of stochastic
processes see Dellacherie and Meyer (1976, 1980, 1983, 1987) and Gihman and
Skorohod (1974, 1975, 1979).
Two stochastic processes X and Y defined on the same probability space are
said to be modifications of each other if P(Xt = Yt ) = 1 for all t ∈ Σ.
Two stochastic processes X and Y taking values in the same state space (E, E)
but not necessarily defined on the same probability space are said to be iden-
tical in law or versions of each other if they have the same finite dimensional
distributions. In other words, for all ti ∈ Σ, and Ai ∈ E, i = 1, 2, . . . , n,

P(Xt1 ∈ A1 , . . . , Xtn ∈ An ) = P(Yt1 ∈ A1 , . . . , Ytn ∈ An ).

Two stochastic processes X and Y defined on the same probability space are
said to be indistinguishable or equivalent if there exists a set N ∈ F such that
P(N ) = 0 and X· (ω) = Y· (ω) for all ω ∈ Ω \ N.
A stochastic process with Σ = R or R+ is called stationary if the finite dimen-
sional distributions are invariant under translations of the time parameter, that
is,
P(Xt1 ∈ A1 , . . . , Xtn ∈ An ) = P(Xt1 +h ∈ A1 , . . . , Xtn +h ∈ An )
for all h ∈ Σ, ti ∈ Σ, and Ai ∈ E, i = 1, 2, . . . , n.
In the general theory of stochastic processes Σ is usually a subset of R and E is
a complete separable metrizable space (Polish) or a locally compact space with
countable base (LCCB). In this framework, the σ-algebra E consists of the Borel
subsets of E, denoted by B(E).
We assume – if nothing else is said – that Σ = [0, ∞).
2. Let X be a stochastic process taking values in Rd . Assume that Rd is equipped
with the usual topology. Then X is said to be a.s. continuous or to have a.s.

© Springer Basel 2015 1

A.N. Borodin, P. Salminen, Handbook of Brownian Motion - Facts and Formulae,
Probability and Its Applications, DOI 10.1007/978-3-0348-8163-0_1
2 I STOCHASTIC PROCESSES IN GENERAL

continuous sample paths if there exists a set N ∈ F such that P(N ) = 0 and the
mapping t 7→ Xt (ω) is continuous for all ω ∈ Ω \ N. An analogous statement is
used to define, e.g., a process which is a.s. right continuous and has left limits
(for short, r.c.l.l.).
The process X is said to be stochastically continuous or continuous in probability
if for every t > 0 and ε > 0,

lim P(|Xs − Xt | > ε) = 0,

s→t

and continuous in mean square if for every t ≥ 0,

lim E|Xt − Xs |2 = 0.
s→t

Kolmogorov’s criterion for continuity. Let the time parameter of X be

n-dimensional, that is t = (t1 , . . . , tn ), ti ≥ 0. Assume that there exist positive
constants α, β, and Mk , k = 1, 2, . . . , such that

(a) E|Xt − Xs |α ≤ Mk |t − s|n+β

for every s and t with si , ti ≤ k, i = 1, . . . , n. Then X has a continuous modifi-

cation. For a proof in the one-dimensional case, see Ikeda and Watanabe (1981)
p. 20. The n-dimensional case is proved similarly. See also Meyer (1981).
3. A family of σ-algebras {Ft : t ≥ 0} (for short, {Ft }) on (Ω, F) is called a
filtration if
Fs ⊆ Ft ⊆ F for every s ≤ t.
A filtration is said to be right continuous if for every s,
\
Fs = Fs+ := Ft .
t>s

The collection (Ω, F, {Ft }, P) is called a filtered probability space. It is said to

satisfy the usual conditions if
(a) F is P-complete,
(b) F0 contains all P-null sets of F,
(c) {Ft } is right continuous.
The meaning of (a) is that a set A ⊆ Ω belongs to F if there exist A1 and A2
in F such that A1 ⊆ A ⊆ A2 and P(A1 ) = P(A2 ).
The usual augmentation of a filtered probability space (Ω, F, {Ft }, P) is a filtered
probability space (Ω, G, {Gt }, P), where
(?) G is the P-completion of F,
(?) Gt is for every t the σ-algebra generated by Ft+ and the P-null sets of G.
 2. MARKOV PROCESSES 3

By definition, A ⊆ Ω belongs to G if there exist A1 and A2 in F such that

A1 ⊆ A ⊆ A2 and P(A1 ) = P(A2 ).
It can be proved (Rogers and Williams (1994) p. 172) that (Ω, G, {Gt }, P) sat-
isfies the usual conditions.
4. A stochastic process X defined on a filtered probability space (Ω, F, {Ft }, P) is
said to be progressively measurable if for every t the mapping (s, ω) 7→ Xs (ω)
from [0, t]×Ω to E is B([0, t])×Ft -measurable. We say that X is adapted (to the
filtration {Ft }) if for every t the random variable Xt is Ft -measurable. Notice
that X is always adapted to its natural filtration {Fto := σ{Xs : s ≤ t}}. An
adapted process with right or left continuous paths is progressively measurable
(see Revuz and Yor (2001) p. 44).
In general, a stochastic process X is called measurable if the mapping (t, ω) 7→
Xt (ω) is B([0, ∞)) × F-measurable.
5. A stopping time with respect to a filtration {Ft } is a mapping T : Ω 7→ [0, ∞) ∪
{∞} such that {T ≤ t} ∈ Ft for every t.
In the case {Ft } is right continuous, T is a stopping time if and only if {T <
t} ∈ Ft for every t.
The σ-algebra FT which, informally, consists of all events occurring before or at
a given stopping time T is defined by

A ∈ FT ⇔ A ∈ F and A ∩ {T ≤ t} ∈ Ft for all t.

2. Markov processes, transition functions,

resolvents, and generators

6. An adapted stochastic process X = {Xt : t ≥ 0} defined on a filtered probability

space (Ω, F, {Ft }, P) and taking values in a general state space (E, E) is called
a time-homogeneous Markov process (Revuz and Yor (2001) p. 81) if for every t
there exists a kernel Pt : E × E 7→ [0, 1], called a transition function, such that
(a) for all A ∈ E and t, s ≥ 0,

P(Xt+s ∈ A|Ft ) = Ps (Xt , A) a.s.,

(b) for all A ∈ E and t the mapping x 7→ Pt (x, A) is measurable,

(c) for all x ∈ E and t the mapping A 7→ Pt (x, A) is a measure,
(d) for all x ∈ E and A ∈ E,
x ∈ A,

1,
P0 (x, A) = ε{x} (A) :=
0, x 6∈ A,

(e) the Chapman–Kolmogorov equation holds:

Z
Pt+s (x, A) = Pt (x, dy)Ps (y, A), x ∈ E, A ∈ E, t, s ≥ 0.
E
4 I STOCHASTIC PROCESSES IN GENERAL

The time-homogeneity and the Markov property are expressed in (a). Intuitively,
the Markov property means that the behavior of X in the future depends only
on the present state of X and not on the behavior in the past. In other words,
the future and the past are independent given the present.
We remark that there are also more general definitions of a Markov process, and
definitions in which the approach is not directly via transition functions. For
these see, e.g., Dynkin (1965), Blumenthal and Getoor (1968), and Ethier and
Kurtz (1986) p. 156.
Properties (b) and (c) are the defining properties of a kernel. In general, a
mapping K : E × E 7→ [0, ∞) is called a kernel if
(?) for every A ∈ E the mapping x 7→ K(x, A) is measurable,
(?) for every x ∈ E the mapping A 7→ K(x, A) is a measure.
A Markov process is called normal if limt→0 Pt (x, E) = 1 for all x ∈ E, and
conservative (or honest) if Pt (x, E) = 1 for all t and x ∈ E (see Dynkin (1965)
Vol I p. 47 and 85). A non-conservative process is often made conservative
by adjoining an extra state ∂ (called cemetery or coffin state) to E. Let E ∂ :=
E ∪ {∂} and E ∂ := σ{E, ∂}. Then the conservative transition function {Pt∂ } is
introduced in the state space (E ∂ , E ∂ ) as follows:
(?) Pt∂ (x, A) := Pt (x, A) x ∈ E, A ∈ E,
(?) Pt∂ (x, ∂) := 1 − Pt (x, E), x ∈ E,
(?) Pt∂ (∂, A) := ε{∂} (A), A ∈ E∂.
7. A transition function P is called jointly measurable if for all A ∈ E the mapping
(t, x) 7→ Pt (x, A) is B([0, ∞)) × E-measurable. In this case we may construct a
family of operators {Rλ : λ > 0} from BE (E) to BE (E) by setting
Z ∞
Rλ f (x) := e−λt Pt f (x)dt,
0

where Z
Pt f (x) := Pt (x, dy)f (y),
E

and BE (E) is the set of bounded E-measurable functions from E to R.

The family {Rλ } is called the resolvent of {Pt }. It has the following properties
(see Gihman and Skorohod (1975) vol. II p. 92, Rogers and Williams (1994)
p. 233):
(?) the resolvent equation

Rα − Rβ = (β − α)Rβ Rα , α, β > 0,

and, consequently, Rα Rβ = Rβ Rα .
dn
(?) R f = (−1)n n!Rλn+1 f,
dn λ
 3. FELLER PROCESSES 5
∞
X
(?) Rλ f = (λo − λ)n Rλn+1
o
f for all λ such that |λ − λo | < λo with λo > 0,
n=0

(?) kRλ k ≤ 1/λ, where kRλ k := sup{kRλ f k : kf k = 1} and kf k := sup |f (x)|,

(?) if fn ↓ 0 (pointwise on E) then Rλ fn ↓ 0 (pointwise on E) for all λ > 0.
8. The (strong) infinitesimal generator G of a Markov process X (or its transition
function P ) is defined as
Pt f − f
Gf := lim
t↓0 t
for f ∈ BE (E) such that the limit exists in the sup norm || · ||. The set of all
these f is called the domain of G.

3. Feller processes, Feller–Dynkin processes, and

the strong Markov property

9. The definition of a Markov process in No. 6 does not give enough structure to
build a rich theory. In particular, more assumptions are needed to be able to
prove regularity properties of sample paths such as right continuity and exis-
tence of limits from the left (r.c.l.l.). Regularity is of importance, e.g., when
extending the Markov property to be valid at stopping times. Also in stochas-
tic calculus regularity of sample paths is paramount. In the general theory of
Markov processes one usually works with processes having a r.c.l.l. modification
(see, e.g., the definition of a standard process in Blumenthal and Getoor (1968)
p. 45 or Sharpe (1988) p. 220).
In this section, E is a LCCB space and E its Borel σ-algebra. The notation
Cb (E) is used for the set of bounded continuous E-measurable functions from
E to R, and C0 (E) is the subset of Cb (E) consisting of functions vanishing at
infinity.
10. A Feller semigroup is a family {Pt } of positive linear operators mapping Cb (E)
into Cb (E) such that
(a) P0 = I (the identity operator),
(b) 0 ≤ Pt f ≤ 1 for all f ∈ Cb (E) such that 0 ≤ f ≤ 1,
(c) Ps+t = Ps Pt ,
(d) limt↓0 kPt f − f k = 0 for all f ∈ Cb (E).
A Markov process whose transition function is a Feller semigroup is called a
Feller process. A Feller process does not, in general, have a r.c.l.l. modification.
(Dynkin (1965) Theorem 3.6, p. 92 gives sufficient conditions for this to hold).
11. A Feller–Dynkin semigroup is a family {Pt } of positive linear operators mapping
C0 (E) into C0 (E) such that (a)–(d) in No. 10 hold when Cb (E) is replaced by
C0 (E). Property (d) in No. 10, the strong continuity condition, follows (Rogers
and Williams (1994) p. 241) in this case from the pointwise continuity condition:
 6 I STOCHASTIC PROCESSES IN GENERAL

limPt f (x) = f (x) for all f ∈ C0 (E) and x ∈ E.

t↓0

A Markov process whose transition function is a Feller–Dynkin semigroup is

called a Feller–Dynkin process. A Feller–Dynkin process (in a LCCB space)
always has a r.c.l.l. modification (Rogers and Williams (1994) p. 243, Revuz
and Yor (2001) p. 91, Blumenthal and Getoor (1968) p. 46, and Dynkin (1965)
p. 92).
12. Let X be a Feller process or a Feller–Dynkin process, and consider it in the
state space (E ∂ , E ∂ ) and assume, therefore, that X is conservative (see No. 6).
Let P ∂ = {Pt∂ } be its transition function and ν(·) = P(X0 ∈ ·) its initial
distribution. A canonical realization of X (or of the pair (ν, P ∂ ) ) is a process
Y in the probability space (E ∂,∞ , F∞ c
, Pν ), where E ∂,∞ is the set of functions
∂ c
ω : [0, ∞) → E such that Yt (ω) := ω(t), F∞ := σ{ω(t) : t ≥ 0}, and Pν is the
unique probability measure (Rogers and Williams (1994) p. 244) on (E ∂,∞ , F∞ c
)
∂
such that for all 0 ≤ t1 ≤ t2 ≤ · · · ≤ tn , and Ai ∈ E ,

Pν (Y0 ∈ A0 , . . . , Ytn ∈ An )
Z Z
= ··· ν(dx0 )Pt∂1 (x0 , dx1 ) · . . . · Pt∂n −tn−1 (xn−1 , dxn ).
A0 An

It can be proved (Revuz and Yor (2001) p. 81) that Y is a Markov process in
the filtered probability space (E ∂,∞ , F∞
c
, {Ftc := σ{ω(s) : s ≤ t}}, Pν ) for any
initial distribution ν. We let Px denote Pν when ν = ε{x} .
In the case X is a Feller–Dynkin process the space E ∂,∞ can be replaced by
the space D, the so-called Skorohod space, which consists of all r.c.l.l. functions
ω : [0, ∞) 7→ E ∂ such that ω(t) = ∂ for all t > s if ω(s) = ∂ or limu↑s ω(u) = ∂.
Let ω∂ denote the function which is identically equal to ∂. Functions in D are
often also extended to [0, ∞] by setting ω(∞) = ∂. Moreover, a usual convention
c
is that η(ω∂ ) = 0 for any F∞ -measurable random variable η.
The lifetime ζ of a path ω ∈ D is defined by ζ(ω) := inf{t : ωt = ∂}. Observe
c
that ζ is an Ft+ -stopping time.
13. The Markov property of a r.c.l.l. Feller process or a Feller–Dynkin process X in
the canonical framework can be expressed as follows: Let ξ and η be, respec-
tively, bounded Ftc - and F∞
c
-measurable random variables. Then for any initial
distribution ν,

(a), Eν (ξ η ◦ θt ) = Eν (ξ EXt η),

where θt is the so-called shift operator, i.e., a mapping defined for every t via

θt (ω) = ω 0 , ω 0 (s) := ω(t + s) for every s ≥ 0,

and η ◦ θt (ω) := η(θt (ω)). Note that in (a) it is necessary that x 7→ Ex (η) is
E-measurable. This follows from (b) in No. 6 by the monotone class theorem.
 4. MARTINGALES 7

The process X has also the strong Markov property. To state this let T be an
c
Ft+ -stopping time, and introduce the σ-algebra FTc + (see. No. 5) by setting

A ∈ FTc + ⇔ c
A ∈ F∞ and A ∩ {T < t} ∈ Ftc for all t.

Then the value of X at time T, denoted XT (= ∂ if T = ∞), is FTc + -measurable

(Rogers and Williams (1994) p. 183). Let ξ and η be, respectively, FTc + - and
c
F∞ -measurable bounded random variables. The strong Markov property says
that for any initial distribution ν,

Eν (ξ η ◦ θT ) = Eν (ξ EXT η),

where θT stands for the shift operator evaluated at T with the convention
θT (ω) = ω∂ if T = ∞.
14. It is an important fact – especially when working with discontinuous processes
(Rogers and Williams (1994) p. 248) – that the above presented Markov prop-
erties can be extended to be valid with respect to appropriately completed fil-
trations.
ν c
To present this, let F∞ be the Pν -completion of F∞ and for every t let Ftν be
the σ-algebra containing Ftc and the Pν -null sets from F∞ν
. We also introduce
the universal completions
\ \
Ft := Ftν , F∞ := F∞ν
,
ν ν

where ν runs through the set of all probability measures on (E ∂ , E ∂ ).

Assume that X is a r.c.l.l. Feller process or a Feller–Dynkin process. Then it
can be proved (Revuz and Yor (2001) p. 93 for Feller–Dynkin processes) that
{Ftν } and {Ft } are right continuous. Therefore, (E ∂,∞ , F∞ ν
, {Ftν }, Pν ) is the
∂,∞ c c
usual augmentation of (E , F∞ , {Ft }, Pν ). Let T be a stopping time with
respect to the filtration {Ft }. Then for any initial distribution ν (Dynkin (1965)
p. 99, p. 104 or Rogers and Williams (1994) p. 250) X has the strong Markov
property
Eν (ξ η ◦ θT ) = Eν (ξ EXT (η)),
where ξ and η are, respectively, FT - and F∞ -measurable bounded random vari-
ables, and FT is defined as in No. 5 with respect to {Ft }. An equivalent formu-
lation is
Eν (η ◦ θT |FT ) = EXT (η) Pν -a.s. on {T < ∞}.

4. Martingales

15. In this section it is assumed that (Ω, F, {Ft }, P) is a filtered probability space
satisfying the usual conditions (see No. 3). A real valued stochastic process
M = {Mt : t ≥ 0} defined on this space is called a martingale (or (P, Ft )-
martingale or martingale with respect to {Ft }) if for every t ≥ 0,
 8 I STOCHASTIC PROCESSES IN GENERAL

(a) M is adapted, that is, Mt is Ft -measurable,

(b) E|Mt | < ∞,
(c) E(Mt |Fs ) = Ms a.s. for every s ≤ t.
The process M is called a supermartingale if it has properties (a), (b), and
(d) E(Mt |Fs ) ≤ Ms a.s. for every s ≤ t.
The process M is called a submartingale if it has properties (a), (b), and
(e) E(Mt |Fs ) ≥ Ms a.s. for every s ≤ t.
By the definition of conditional expectation the martingale property (c) (or (d)
or (e)) is equivalent to

E(Mt 1IA ) = E(Ms 1IA ) (or E(Mt 1IA ) ≤ E(Ms 1IA ) or E(Mt 1IA ) ≥ E(Ms 1IA )),

respectively, which should hold for all s and t such that s ≤ t and for all A ∈ Fs .
Clearly, if M is a submartingale, then −M is a supermartingale. A martingale
is both a super- and a submartingale.
Under the assumption that (Ω, F, {Ft }, P) satisfies the usual conditions it can
be proved (Rogers and Williams (1994) p. 173) that a sub- or supermartingale
M has a r.c.l.l. modification if and only if the mapping

t 7→ E(Mt )

is right continuous. In particular, a martingale has in this case always a r.c.l.l.

modification.
For the rest of this section all of the sub- and supermartigales we consider are
assumed to be r.c.l.l.

16. Convergence. Let M be a sub- or supermartingale. Assume that M is bounded

in L1 , that is
sup E|Mt | < ∞.
t

Then

(a) lim Mt =: M∞ exists a.s. and E|M∞ | < ∞.

t→∞

This is due to J.L. Doob, and is based on the so-called upcrossing lemma. See
Doob (1953) p. 319 and 316, where the result is formulated for martingales in
discrete time. For more recent reproductions see Ikeda and Watanabe (1981)
p. 30, and Rogers and Williams (1994) p. 176. We refer also to Chung (1998)
for an article on Doob’s fundamental contributions in probability theory.
Next recall that a stochastic process X is called uniformly integrable if for every
ε > 0 there exists a number c = c(ε) > 0 such that for every t ≥ 0,

E(|Xt | ; |Xt | > c) < ε.

 4. MARTINGALES 9

Notice that a uniformly integrable stochastic process is bounded in L1 .

Now assume additionally that M is uniformly integrable. Then Mt → M∞ also
in L1 , that is

(b) lim E|Mt − M∞ | = 0.

t→∞

Moreover, if M is a martingale,

(c) Mt = E(M∞ |Ft ) a.s.

Observe that a non-negative supermartingale is always convergent (but not nec-

essarily uniformly integrable).
A martingale which is bounded in L2 is uniformly integrable, and also bounded
in L1 . Therefore, (a), (b), and (c) hold for martingales bounded in L2 .

17. Consider an arbitrary real valued stochastic process M = {Mt : t ≥ 0} on

the probability space (Ω, F, {Ft }, P). Assume that M0 is F0 -measurable. Then
M is called a local martingale if there exists an increasing sequence of stopping
times {Tn } such that Tn ↑ ∞ and each stopped process {MTn ∧t − M0 : t ≥ 0} is
a martingale, where 
MTn , Tn < t,
MTn ∧t =
Mt , Tn ≥ t.
A martingale is always a local martingale. In fact, we have the following general
result. Let M be a sub- or supermartingale and T a stopping time. Then
{MT ∧t : t ≥ 0} is also a sub- or supermartingale, respectively, with respect to
{Ft }. For the proof see, e.g., Revuz and Yor (2001) p. 71, or Doob (1984) p.
471. In particular, we have, respectively,

EMT ∧t = EM0 , EMT ∧t ≥ EM0 , EMT ∧t ≤ EM0 .

18. Optional stopping. Let M be a uniformly integrable or non-negative super-

martingale and M∞ := limt→∞ Mt . Then for all stopping times S and T such
that S ≤ T we have
(a) E(MT |FS ) ≤ MS a.s.
In particular,
(b) EMT ≤ EMS ≤ EM0 .
In (a) and (b) it is understood that MT = M∞ on the set {T = ∞}. For a proof
in discrete time, see Doob (1984) p. 440, and for continuous time Gihman and
Skorohod (1979) p. 11, and Rogers and Williams (1994) p. 189.
If there exists a constant C such that S ≤ T ≤ C a.s. then (a) holds because M
restricted to a finite time interval is uniformly integrable. Furthermore, if M is
bounded it is uniformly integrable and hence (a) holds.
 10 I STOCHASTIC PROCESSES IN GENERAL

For a uniformly integrable martingale the inequalities above can be replaced by

equalities.
19. Maximal inequalities. Let t > 0 and α > 0 be given. The following inequali-
ties hold for a submartingale M :
   
αP sup Ms ≥ α ≤ E Mt ; sup Ms ≥ α
s≤t s≤t

≤ E(Mt ∨ 0) ≤ E|Mt |,

   
αP inf Ms ≤ −α ≤ E Mt ; inf Ms > −α − E(M0 )
s≤t s≤t

≤ E(Mt ∨ 0) − E(M0 ) ≤ E|Mt | + E|M0 |.

If M is a supermartingale, then −M is a submartingale and, therefore,
 
αP inf Ms ≤ −α ≤ −E(Mt ∧ 0) ≤ E|Mt |,
s≤t

 
αP sup Ms ≥ α ≤ −E(Mt ∧ 0) + E(M0 ) ≤ E|Mt | + E|M0 |.
s≤t

Especially, if M is a positive supermartingale we have the inequality

 
αP sup Ms ≥ α ≤ E(M0 ).
s≤t

For proofs, see Doob (1984) p. 442, and Ikeda and Watanabe (1981) p. 27.
20. Lp -inequalities. Let M be a positive submartingale. Assume that EMtp < ∞
for a given t > 0 and for some p ≥ 1. Then for α > 0,
   
αp P sup Ms ≥ α ≤ E Mtp ; sup Ms ≥ α ≤ E(Mtp ),
s≤t s≤t

and if p > 1, then

p p
   
(a) E sup Msp ≤ EMtp .
s≤t p−1

The inequality (a) for p = 2 is called the Doob–Kolmogorov inequality. Notice

that if N is a martingale, then |N | is a positive submartingale. From (a) it
follows (for p > 1) that sups≤t Ms is in Lp if and only if Mt is in Lp .
For proofs, see Chung and Williams (1990) p. 15 or Doob (1984) p. 444.
21. Doob–Meyer decomposition. A stochastic process X defined on the proba-
bility space (Ω, F, {Ft }, P) is said to be of class (D) if the family

{XT : T a finite stopping time }

 4. MARTINGALES 11

is uniformly integrable. The Doob–Meyer decomposition theorem says that a

r.c.l.l. adapted process Z is a submartingale belonging to the class (D) if and only
if there exist a uniformly integrable martingale M, M0 = 0, and a predictable
(see III.2) integrable increasing process A, A0 = 0, such that a.s. for all t,

Zt = Z0 + Mt + At .

Moreover, for a given submartingale Z in the class (D) this decomposition is

unique, and A is called the compensator of Z. We remark that uniformly inte-
grable martingales belong to the class (D). In particular, martingales bounded
in Lp , p > 1, are uniformly integrable, and, hence, in (D).
We refer to Rogers and Williams (2000) p. 367 for a proof and further devel-
opments. For a definition of a smaller class (DL), and the Doob–Meyer de-
composition for submartingales in this class, see Ikeda and Watanabe (1981)
p. 35. Moreover, the Doob–Meyer decomposition can be formulated for local
submartingales; see Rogers and Williams (2000) p. 375.
22. The Burkholder–Davis–Gundy inequality. Let M, M0 = 0, be a con-
tinuous local martingale. Then M 2 is a local submartingale. Let hM i be
the predictable increasing process associated to M 2 by the local form of the
Doob–Meyer decomposition. The process hM i is continuous in this case. Let
Mt∗ := sups≤t |Ms |. The Burkholder–Davis–Gundy inequality says that for every
p > 0 there exist constants cp and Cp such that for all stopping times T,

cp E(hM ipT ) ≤ E((MT∗ )2p ) ≤ Cp E(hM ipT ).

For a direct proof of the existence of hM i see Revuz and Yor (2001) p. 124. For a
proof and more general formulations of the Burkholder–Davis–Gundy inequality
see Lenglart, Lépingle and Pratelli (1980) and Revuz and Yor (2001) p. 160.
23. An adapted r.c.l.l. process X defined on the probability space (Ω, F, {Ft }, P) is
said to be a semimartingale (see Rogers and Williams (2000) p. 23, Revuz and
Yor (2001) p. 127) if there exist a local martingale M, M0 = 0, and an adapted
process A, A0 = 0, of finite variation such that a.s. for all t,

Xt = X0 + Mt + At .

Because A could also be a (discontinuous) martingale the semimartingale repre-

sentation is not in general unique. However, if the semimartingale representation
holds with A predictable, then A is the unique predictable process for which the
representation is valid. Also if X is a continuous semimartingale and M in its
representation is continuous, then A is unique.
 CHAPTER II

LINEAR DIFFUSIONS

1. Basic facts

1. We adopt here the general approach to linear diffusions as presented in the book
by K. Itô and H.P. McKean which appeared in 1965. References are made to
the second printing from 1974.
Let I be an interval with left endpoint l ≥ −∞ and right endpoint r ≤ ∞.
Let X be a time-homogeneous Markov process taking values in I (see I.6) and
F c its canonical filtration (see I.12). Further, let Px denote the probability
measure associated to X when started at x ∈ I. Then X is called a linear (or
one-dimensional) diffusion if for all x ∈ I,
(a) t 7→ Xt (ω) is continuous on [0, ζ) Px -a.s.,
(b) Ex (η ◦ θT |FTc + ) = EXT (η) Px -a.s.,
c
where ζ is the lifetime of X, η is a bounded F∞ -measurable random variable, T
c
is an Ft+ -stopping time, and θ. is the shift operator (see I.13). The σ-algebra
FTc + is constructed as in I.13. Recall also the conventions in I.12 concerning the
cemetery state ∂. In particular, if T = ∞ both the left- and the right-hand side
in (b) are taken to be equal to zero.
For the definition above we refer to Itô and McKean (1974) p. 84, Freedman
(1971) p. 103, and Rogers and Williams (2000) p. 271. For an approach based on
extended generators, martingale problems and stochastic differential equations,
see Stroock and Varadhan (1979) p. 136, and Revuz and Yor (2001) p. 294,
Øksendal (1998) p. 108.

2. Roughly speaking, a linear diffusion is a strong Markov process with continuous

paths taking values on an interval. Notice, or see Itô and McKean (1974) p. 84,
that if τ ∼ Exp(1), then our strong Markov property excludes processes of the
type
t ≤ τ,

0,
Xt =
t − τ, t > τ.
The strong Markov property also implies that for all x ∈ I,


(a) Px ∃ ε(ω) > 0 such that Xt (ω) ≥ x ∀ t ∈ (0, ε(ω)) = 0 or 1,


(b) Px ∃ ε(ω) > 0 such that Xt (ω) ≤ x ∀ t ∈ (0, ε(ω)) = 0 or 1.

© Springer Basel 2015 12

A.N. Borodin, P. Salminen, Handbook of Brownian Motion - Facts and Formulae,
Probability and Its Applications, DOI 10.1007/978-3-0348-8163-0_2
 1. BASIC FACTS 13

In fact, the Blumenthal 0–1 law (Itô and McKean(1974) p. 85) holds:

c
Px (A) = 0 or 1, for all x ∈ I and A ∈ F0+ .

3. Unless otherwise stated, we consider only regular diffusions (Dynkin (1965) Vol
II p. 121); letting Hy := inf{t : Xt = y}, then X is a regular diffusion if for
every x, y ∈ I,

(a) Px (Hy < ∞) > 0.

It can be proved (Rogers and Williams (2000) p. 273) that if X is regular, then
the probabilities in (a) and (b) in No. 2 equal 0 for every x ∈ intI. It holds also
for every x ∈ I that


Px ∃ ε(ω) > 0 such that Xt (ω) ∈ I ∀ t ∈ (0, ε(ω)) = 1,

that is, the lifetime ζ is positive a.s. Moreover, if l ∈ I, then

Pl ∃ε(ω) > 0 such that Xt (ω) > l ∀t ∈ (0, ε(ω)) = 0,

and an analogous equality holds for r.

Notice also that (a) means that the state space of a regular diffusion consists of
only one communicating class (Itô and McKean (1974) p. 92). Using terminology
from the theory of Markov chains, regularity here means irreducibility.
We refer to Blumenthal and Getoor (1968) p. 61 for a general definition of a
regular point.
We stress that our definition of regularity does not allow absorbing boundaries
and is hence slightly more restrictive than the definition in Freedman (1971) p.
106 and Rogers and Williams (2000) p. 271, or in Karlin and Taylor (1981) p.
158.

4. Every diffusion has three basic characteristics: speed measure, scale function,
and killing measure.
The speed measure m is a measure on B(I) such that 0 < m((a, b)) < ∞, l <
a < b < r. For every t > 0 and x ∈ I the measure A 7→ Pt (x, A), A ∈ B(I), (cf.
I.6) is absolutely continuous with respect to m :
Z
Pt (x, A) = p(t; x, y)m(dy).
A

The density p may be taken to be positive, jointly continuous in all variables,

and symmetric, that is p(t; x, y) = p(t; y, x). For a proof see Itô and McKean
(1974) p. 149. Due to this symmetry X is in duality with itself relative to m
(for a definition of duality see Blumenthal and Getoor (1968) p. 253).
14 II LINEAR DIFFUSIONS

The killing measure k is a measure on B(I) such that k((a, b)) < ∞, l < a <
b < r. It is associated to the distribution of the location of the process at its
lifetime ζ := inf{t : Xt ∈
/ I} :
Z t Z
Px (Xζ− ∈ A; ζ < t) = ds k(dy) p(s; x, y), A ∈ B(I).
0 A

The scale function s is an increasing continuous function from I to R. It is

connected to the drift of the diffusion. In particular, if k((a, b)) = 0 for an
interval (a, b) ⊂ I, then for a ≤ x ≤ b,

s(b) − s(x)
Px (Ha < Hb ) = 1 − Px (Hb < Ha ) = .
s(b) − s(a)

We say that X is in natural scale if s(x) = x. In this case and if, e.g., k ≡ 0 and
I = R, the diffusion X is a local martingale and can be constructed as a random
time change of a standard Brownian motion (see No. 16).
In No. 9 we relate m, k, and s – in the absolutely continuous case – with the
infinitesimal parameters of X.

5. The transition semigroup {Pt } of a diffusion maps Cb (I) into Cb (I) (Breiman
(1968) p. 358, Freedman (1971) p. 116, Rogers and Williams (2000) p. 291)
and is strongly continuous (Dynkin (1965) Vol. II p. 137 (I closed), p. 162 (I
open)). It follows that a linear diffusion is a Feller process (cf. I.10).
The semigroup {Pt } does not in general map C0 (I) into C0 (I), and hence a
diffusion is not in general a Feller–Dynkin process (cf. I.11) (see Rogers and
Williams (2000) p. 292 for an example).
A linear diffusion is not in general a semimartingale
p (cf. I.23). An example of
such a diffusion (Yor (1978)) is X := { |Wt | : t ≥ 0}, where W is standard
Brownian
√ motion. Notice that X is a diffusion because |W | is a diffusion and
x 7→ x, x ≥ 0, is a one-to-one mapping.
A linear diffusion can be viewed as a standard process in the sense of the defi-
nition (9.2) in Blumenthal and Getoor (1968) p. 45.

6. Boundary classification. The behavior of X in the vicinity of the endpoints

of I is determined by the properties of the basic characteristics m, s, and k. To
explain this let z be such that l < z < r. Then the following terminology is used
(Itô and Mckean (1974) p. 108, Ethier and Kurtz (1986) p. 366):
The left-hand endpoint l is called exit if
Z
(m((a, z)) + k((a, z)))s(da) < ∞,
(l,z)

and entrance if Z
(s(z) − s(a))(m(da) + k(da)) < ∞.
(l,z)
 1. BASIC FACTS 15

The right-hand endpoint r is called exit if

Z
(m((z, a)) + k((z, a)))s(da) < ∞,
(z,r)

and entrance if Z
(s(a) − s(z))(m(da) + k(da)) < ∞.
(z,r)

A boundary point which is both entrance and exit is called non-singular. A

diffusion reaches its non-singular boundaries with positive probability, and it is,
a priori, possible to start a diffusion from a non-singular boundary. In this case
the basic characteristics alone do not determine the process uniquely and the
description of the process must be completed by giving a boundary condition at
each non-singular boundary point. This is discussed in the next section.
A boundary point which is neither entrance nor exit is called natural. A natural
boundary point cannot be reached in finite time, and hence does not belong to
the state space of X.
Consider the endpoint l and suppose that it is natural. Then l is said to be attrac-
tive if lima↓l s(a) > −∞. In this case and if, e.g., k ≡ 0, we have limt↑∞ Xt = l
with positive probability.
A boundary point which is entrance but not exit, for short entrance-not-exit,
cannot be reached from an interior point of I, and hence does not belong to the
state space of a regular diffusion. The difference between an entrance-not-exit
and a natural boundary is that it is possible in the former case, but not in the
latter, to start the process from the boundary. Analytically this means that for
every t > 0 and entrance boundary l, the measure Pt (x, ·) converges (weakly) to
a (sub)probability measure Pt (l, ·) 6≡ 0 as x tends to l.
A boundary point which is exit but not entrance, for short exit-not-entrance, is
reached from an interior point of I with positive probability. It is not, however,
possible to start the process from an exit-not-entrance boundary point.

7. The (weak) infinitesimal generator of X is the operator G • defined by

Pt f − f
G • f := lim
t↓0 t
applied to f ∈ Cb (I) for which the limit exists pointwise, is in Cb (I), and
Pt f − f
sup < ∞.
t>0 t

Let D(G • ) denote this set of functions.

Assume that both l and r are non-singular and belong to I. Let f + and f −
denote the right and left derivative of a function f with respect to the scale
function s:
f (x + h) − f (x) f (x) − f (x − h)
f + (x) := lim , f − (x) := lim .
h↓0 s(x + h) − s(x) h↓0 s(x) − s(x − h)
16 II LINEAR DIFFUSIONS

We define a set of functions D(G) by saying that a function f ∈ Cb (I) belongs

to D(G) if there exists a function g ∈ Cb (I) such that for every l < a < b < r,
Z Z
(a) g(x)m(dx) = f − (b) − f − (a) − f (x)k(dx),
[a,b) [a,b)
Z Z
(b) g(x)m(dx) = f + (b) − f + (a) − f (x)k(dx),
(a,b] (a,b]

(c) g(l)m({l}) = f + (l+) − f (l)k({l}), if m({l}) < ∞, k({l}) < ∞,

(d) g(r)m({r}) = −f − (r−) − f (r)k({r}), if m({r}) < ∞, k({r}) < ∞.
Let Gf := g for f ∈ D(G). The existence of f + and f − is part of the definition
of D(G). From (a) and (b) it is seen that for f ∈ D(G) the derivatives f + and f −
are right and left continuous, respectively. The result (Itô and McKean (1974)
p. 100, 117, and 135, Mandl (1968), Freedman (1971) p. 131, Ethier and Kurtz
(1986) ) is
G = G • , D(G) = D(G • ).
The following terminology concerning non-singular boundaries is often used.
Consider the left endpoint l. It is called
(?) reflecting, if m({l}) = k({l}) = 0,
(?) sticky, if m({l}) > 0, k({l}) = 0,
(?) elastic, if m({l}) = 0, k({l}) > 0.
A diffusion “spends no time” and “does not die” at a reflecting boundary point.
Supposing that l is reflecting, then the first statement means that for all t ≥ 0
and x ∈ I,
Px (Leb({s ≤ t : Xs = l}) = 0) = 1,
and the meaning of the second one is

Px (Xζ− = l) = 0.

If l is sticky, then X does not die at l but spends a positive amount of time
there, i.e.,
Px (Leb({s ≤ t : Xs = l}) > 0) > 0.
If l is elastic, then X does not spend any time at l but

Px (Xζ− = l) > 0.

The condition in (c) can be given meaning also when m({l}) = ∞ and/or
k({l}) = ∞. Indeed, dividing in (c) by m({l}) and letting m({l}) tend to infinity
gives the condition
(e) g(l) = Gf (l) = 0.
In this case l is called absorbing, and we may define Pl (Xt = l for all t) = 1.
Now X is not regular.
 1. BASIC FACTS 17

Dividing in (c) by k({l}) and letting k({l}) tend to infinity gives the condition
(f) f (l) = 0.
In this case l is called a killing boundary, meaning that the diffusing particle is
– if it hits l – immediately sent to the cemetery state ∂. Here l ∈
/ I.
The conditions (e) and (f) can also be combined
g(l) = −γ(l)f (l), γ(l) > 0.
The picture here is that the particle – if it hits l – stays there a random amount
of time which is exponentially distributed with parameter γ(l) and independent
of the motion prior to hitting of l. After that time the particle is sent to the
cemetery state ∂. We call this kind of boundary point a trap, and remark that
in Itô and McKean (1974) p. 91 a more general meaning is given to this term.
We refer also to Itô and McKean (1974) p. 186 and Rogers and Williams (2000)
p. 439, where general boundary conditions are discussed in the context of Brown-
ian motion (so-called Feller’s Brownian motions).

8. If the boundary l is not non-singular, the infinitesimal generator is still given as

in (a) and (b) in No. 7 but the condition (c) must be replaced by:
(?) f + (l+) = 0 if l is entrance-not-exit,
(?) f (l+) = 0 if l is exit-not-entrance,
(?) no condition is needed if l is natural.

9. We consider here the special case in which the basic characteristics are absolutely
continuous with respect to the Lebesgue measure and have smooth derivatives.
In other words,
Z x
m(dx) = m(x)dx, k(dx) = k(x)dx, s(x) = s0 (y)dy,

where the functions m and s0 are continuous and positive and k is continuous
and non-negative. Moreover, if s00 is continuous, then the infinitesimal generator
G : D(G) 7→ Cb (I) is a second order differential operator

1 d2 d
Gf (x) = a2 (x) f (x) + b(x) f (x) − c(x)f (x).
2 dx2 dx
The functions a, b and c – the infinitesimal parameters of X – are related to
m, k and s via the formulae

m(x) = 2 a−2 (x)eB(x) , s0 (x) = e−B(x) , k(x) = 2 a−2 (x) c(x)eB(x)

Rx
with B(x) := 2 a−2 (y) b(y)dy.
The domain of G consists of all functions in Cb (I) such that Gf ∈ Cb (I) together
with the appropriate boundary conditions as explained in No. 7 and 8. Because
 18 II LINEAR DIFFUSIONS

of absolute continuity, the boundary condition at a non-singular boundary can

– strictly speaking – only be reflection or killing.
The function a2 is called the infinitesimal variance, b the infinitesimal mean,
and c the infinitesimal killing rate because, in many cases,

1
lim Ex (Xt − x)2 = a2 (x),
t→0 t
1
lim Ex (Xt − x) = b(x),
t→0 t
and
1 
lim Px (ζ > t) − 1 = −c(x),
t→0 t

see, e.g., Lamperti (1977) p. 130, and Karlin and Taylor (1981) p. 159. Notice
that the scale function s has the property

1 2 d2 d
a (x) 2 s(x) + b(x) s(x) = 0,
2 dx dx
and the function m – the density of the speed measure – has the adjoint property

1 d2 2 d
(a (x)m(x)) − (b(x)m(x)) = 0.
2 dx2 dx

10. Let Hz := inf{t : Xt = z} be the first hitting time of z ∈ I. Then (Itô and
McKean (1974) p. 128) for α > 0,

(x)

, x ≤ z,
(z )


Ex (e−αHz ) =
 ' (x)
, x ≥ z,

' (z )
where ψα and ϕα are continuous solutions of the generalized differential equation

(a) Gu = αu.

This equation should be read as follows: u is a function which satisfies

Z Z
α u(x)m(dx) = u− (b) − u− (a) − u(x)k(dx)
[a,b) [a,b)

for all a and b such that l < a < b < r. In the absolute continuous case u is a
solution of the ODE
1 2 d2 d
a (x) 2 u(x) + b(x) u(x) − c(x)u(x) = αu(x).
2 dx dx
The functions ψα and ϕα can be characterized as the unique (up to a multiplica-
tive constant) positive solutions of (a) by firstly demanding that ψα is increasing
 1. BASIC FACTS 19

and ϕα is decreasing, and secondly posing conditions at the non-singular bound-

ary points. For ψα a boundary condition is only needed at l:
if l ∈ I,
αψα (l)m({l}) = ψα+ (l) − ψα (l)k({l}),
if l ∈
/ I,
ψα (l+) = 0.
In the latter case l is a killing boundary. For ϕα we have, similarly,
if r ∈ I,
αϕα (r)m({r}) = −ϕ−
α (r) − ϕα (r)k({r}),

if r ∈
/ I,
ϕα (r−) = 0.
The functions ψα and ϕα also have the following properties at l (and analogous
properties at r):
if l is entrance-not-exit,

ψα (l+) > 0, ψα+ (l+) = 0, ϕα (l+) = +∞, ϕ+

α (l+) > −∞,

if l is exit-not-entrance,

ψα (l+) = 0, ψα+ (l+) > 0, ϕα (l+) < +∞, ϕ+

α (l+) = −∞,

if l is natural,

ψα (l+) = 0, ψα+ (l+) = 0, ϕα (l+) = +∞, ϕ+

α (l+) = −∞.

11. The functions ψα and ϕα are called the fundamental solutions of (a) in No. 10.
They are linearly independent and all solutions can be expressed as their linear
combinations. Moreover, the so-called Wronskian
− −
wα := ψα+ (x)ϕα (x) − ψα (x)ϕ+
α (x) = ψα (x)ϕα (x) − ψα (x)ϕα (x)

is independent of x.
Let p be the transition density of X with respect to the speed measure (see No.
4) and introduce the Green function
Z ∞
Gα (x, y) := e−αt p(t; x, y)dt.
0

Then
wα−1 ψα (x)ϕα (y), x ≤ y,

Gα (x, y) =
wα−1 ψα (y)ϕα (x), x ≥ y.
 20 II LINEAR DIFFUSIONS

If the killing measure k ≡ 0 and non-singular boundaries are killing boundaries,

then Z ∞
(s(x) − s(a))(s(b) − s(y))
G0 (x, y) := p(t; x, y)dt = lim , x ≤ y.
0 a↓l,b↑r s(b) − s(a)
Under the same assumptions but with l reflecting, we have ψ0 ≡ 1 and hence
G0 (x, y) = lim(s(b) − s(y)), x ≤ y.
b↑r

It can be proved (Itô and McKean (1974) p. 149, Lamperti (1977) p. 127)
that (t, x) 7→ p(t; x, ·) solves for t > 0 and x ∈ intI the generalized parabolic
differential equation
@
(a) u(t, x) = Gu(t, x).
@t
The boundary conditions for p(t; x, ·) at l and r can be deduced from the corre-
sponding conditions for ψα and ϕα given in No. 10.
The equation (a) is the so-called backward Kolmogorov equation for X introduced
in Kolmogorov (1931) and further studied in Feller (1936). See Lamperti (1977)
p. 127 for a short historical view of Kolmogorov’s equation, and Shiryayev (1989)
for an article on Kolmogorov’s life and activities. We refer also to Jacobsen
(1996) who discusses Laplace’s early work on an urn model leading to a partial
differential equation which is the forward Kolmogorov equation for an Ornstein–
Uhlenbeck process.
12. A diffusion X is said to be recurrent if Px (Hy < ∞) = 1 for all x, y ∈ I. Notice
that
Px (Hy < ∞) = lim Ex (e−αHy ).
α→0
A diffusion which is not recurrent is called transient. It can be proved (Itô and
McKean (1974) p. 159 for the case x = y) that X is transient if and only if
lim Gα (x, y) < ∞
α↓0
for some x, y ∈ I, and hence for all x, y ∈ I. From the resolvent equation it
follows that this limit always exists or equals +∞. Therefore X is recurrent if
and only if
lim Gα (x, y) = ∞.
α↓0

A recurrent diffusion is called null recurrent if Ex (Hy ) = ∞ for all x, y ∈ I

and positively recurrent if Ex (Hy ) < ∞ for all x, y ∈ I. It can be proved that
a recurrent diffusion is positively recurrent if and only if m(I) < ∞. In the
recurrent case we have (see Salminen (1993))
1
(a) lim αGα (x, x) = .
α↓0 m(I )
The speed measure m is also easily seen to be a stationary (or equilibrium or
invariant) measure of a recurrent diffusion, that is, it satisfies for all t and
A ∈ B(I), Z
mPt (A) := m(dx) Pt (x, A) = m(A).
I
 2. LOCAL TIME 21

2. Local time

13. For a regular diffusion X there exists (see Itô and McKean (1974) p. 174 and p.
183, Freedman (1971) p. 159 and Rogers and Williams (2000) p. 289) a family
of random variables
{L(t, x) : x ∈ I, t ≥ 0},
called the local time of X, such that
Rt R
(a) 0 1IA (Xs ) ds = A L(t, x) m(dx) a.s., A ∈ B(I),
Leb({s < t : x −  < Xs < x + })
(b) L(t, x) = lim↓0 a.s.,
m((x − ; x + ))
(c) L(t, x, ω) = L(s, x, ω) + L(t − s, x, θs (ω)) a.s.,
where in (c) s < t, and θ. is the shift operator (see I.13). If m({x}) > 0, then
L(t, x) is proportional to the amount of time X spends at x up to t.
We have the following connection between the local time and the transition
density:
Z t
(d) Ex (L(t, y)) = p(s; x, y) ds.
0

In the transient case, Ex (L(ζ, y)) = Ex (L(∞, y)) = G0 (x, y). In fact, (see No.
27)  
Px (L(∞, x) > α) = exp − .
G0 (x; x)

For a fixed x the process Lx = {L(t, x) : t ≥ 0} is called the local time process
of X at the point x. Using the definition in No. 21 it is seen that Lx is a
continuous additive functional of X. Because Lx is non-decreasing it induces a
(random) measure on the time axis. The support of this measure, that is, the
complement of the largest open set of measure zero, is a.s. equal to the set
{t : Xt = x}. This fact is often expressed by saying that Lx increases only when
X is at x, and can be proved using the random time change technique (see No.
16) from the corresponding property of the Brownian local time (see IV.7).
Regularity of a point and existence of local time at that point are equivalent
properties for a large class of stochastic processes; see Blumenthal and Getoor
(1968) p. 216. For a survey article on local times in general see Geman and
Horowitz (1980).

14. Assume that X is started at x, and introduce the inverse local time process
ρx = {ρ(t, x) : t ≥ 0} at x by setting

ρ(t, x) := inf{s : L(s, x) > t}.

Then ρx has the properties:

(a) ρ(0, x) = 0 a.s.,
 22 II LINEAR DIFFUSIONS

(b) t 7→ ρ(t, x) is right-continuous and non-decreasing,

(c) ρ(t + s, x) − ρ(t, x) ∼ ρ(s, x) for all t, s ≥ 0,
(d) for 0 ≤ t1 < · · · < tn the random variables

ρ(t1 , x), ρ(t2 , x) − ρ(t1 , x), . . . , ρ(tn , x) − ρ(tn−1 , x)

are independent.
In general, a stochastic process having the properties (b)–(d) is called a sub-
ordinator, or a non-decreasing Lévy process. Due to (c) and (d) we say that
ρx has stationary and independent increments. Subordinators play prominent
roles, e.g., in the excursion theory of Markov processes.
To give the Lévy–Khintchine representation of ρx let nx± be the measures defined
by
Py (Hx < ∞) − Ey (e− Hx )
Z ∞
(1 − e−αl )nx± (dl) = lim ,
0 y→x± ±(s(y ) − s(x))

and
1 − Py (Hx < ∞)
nx± (∞) := lim .
y→x± ±(s(y ) − s(x))

Then the Lévy–Khintchine representation is

  Z ∞ 
−αρ(t,x) −αl x x
Ex (e ) = exp −t m({x})α + (1 − e )n (dl) + n (∞) ,
0

where
nx := nx− + nx+ , nx (∞) := nx+ (∞) + nx− (∞).

Note that
Px (ρ(t, x) < ∞) = exp(−t nx (∞)).
It follows that X is transient if and only if nx (∞) > 0 for some x. Computing
the limits above (see Itô and McKean (1974) p. 214, where the recurrent case
is studied, see also Pitman and Yor (1999a)) shows that

t
 
Ex (e−αρ(t,x) ) = exp − .
G (x; x)

Notice that, if X is recurrent, it follows from (a) in No. 12 that for all x ∈ I,

Ex (ρ(t, x)) = m(I) t.

15. Let X be recurrent and 0 ∈ I. Let ρ := ρ0 be the inverse local time process at
0 and consider for a given x ∈ I the process

L(ρ) = {L(ρ(t), x) : t ≥ 0}.

 2. LOCAL TIME 23

Using the strong Markov property it is seen that L(ρ) is a subordinator. Its
Lévy–Khintchine representation is

t
 
E0 (e−αL(ρ(t),x) ) = exp −
1 + |s(x) − s(0)|
 Z ∞ 
= exp −t (1 − e−αl )n(dl) ,
0

where
n(dl) = β 2 e−βl dl, β := |s(x) − s(0)|−1 .
For the proof when X is a Brownian motion see Itô and McKean (1974) p. 74.
General recurrent diffusions are considered in Csáki and Salminen (1996). It is
also possible to have analogous representations for transient diffusions. In this
case, however, the process ρ is exploding.
Notice that the representation above does not depend on the speed measure
and because n([0, ∞)) < ∞ the process L(ρ) is a compound Poisson process.
Straightforward computations show that

E0 (L(ρ(t), x)) = t, E0 (L2 (ρ(t), x)) = t2 + 2t|s(x) − s(0)|.

16. Consider a diffusion X in natural scale, i.e., when s(x) = x, and assume that
k ≡ 0. This imposes some restrictions on the boundary behavior. For instance,
a finite boundary point cannot be entrance-not-exit and an infinite boundary
point cannot be exit. As usual, let m denote the speed measure of X. Further,
let W be a standard Brownian motion on R and assume that X0 = W0 . We
explain here how to obtain a version of X from W via a random time change
(or a random time substitution.) This construction makes the fundamental role
of local time in the theory of diffusions apparent.
Let {L(t, x) : x ∈ R, t ≥ 0} be the local time of W and extend m to B(R) by
setting m(A) = 0 if A ∩ I = ∅. Consider the functional
Z
(m)
L (t) := L(t, y) m(dy).
R

In the case m(dx) = 2m(x) dx we have

Z t
(m)
L (t) = m(Wu ) du.
0

It can be proved (Itô and McKean (1974) p. 168) that t 7→ L(m) (t) is non-
decreasing and continuous (when finite). This also follows from No. 21 because
L(m) constitutes an additive functional of W. It may happen that L(m) (t) = ∞
with positive probability for some t. In fact, we force this to happen in an
important special case. Indeed, assume that l is non-singular for X. Then it is
impossible on the basis of m alone to say whether l is reflecting or killing. In
 24 II LINEAR DIFFUSIONS

order to distinguish between the two classes, it is assumed that in the latter
case m({l}) = ∞, i.e., l is regarded as an absorbing state. This implies that
L(m) (t) = ∞ when t > Hl := inf{t : Wt = l}.
Now let
ρ(m) (t) := inf{u : L(m) (u) > t}

be the right continuous inverse of L(m) . If l is absorbing, then ρ(m) (t) = Hl for
all t ≥ L(m) (Hl −). The random time change of W based on L(m) is the process

W (m) := {Wρ(m) (t) : t ≥ 0}.

Then, when killing boundaries are changed to absorbing ones,

X ∼ W (m) .

Further, letting {LX (t, x) : x ∈ I, t ≥ 0} be the local time of X with respect

to the speed measure m, then
 

LX (t, x) : t ≥ 0, x ∈ I ∼ L ρ(m) (t), x : t ≥ 0, x ∈ I .

If X is not in natural scale the process defined by X et := s(Xt ), where s is the

scale function of X, is a diffusion in natural scale for which the results above
apply. In particular, if m
e is the speed measure of X,
e then

s(−1) Wρ(m)
 

Xt : t ≥ 0 ∼ e (t) : t ≥ 0 ,

where s(−1) (x) := inf{y : s(y) > x} is the inverse of s.

These results are due to Itô and McKean (1974) p. 167–174 and Volkonskij
(1958). For proofs see also Rogers and Williams (2000) p. 277–283 and Freedman
(1971) p. 151–159.
From the definition of L(m) it is evident that if a Brownian particle is moving
at time t on the region, where m is “big”, then L(m) is increasing rapidly. Con-
sequently, ρ(m) (t + h) ' ρ(m) (t) for “large” h > 0, and hence the increment
Xt+h − Xt is “small”. Due to this property m has been given the name speed
measure. An extreme case is obtained at the points with positive m-mass (so-
called sticky points, cf. the definition of a sticky boundary in No. 7) in the
vicinity of which X will “spend much time”.

3. Passage times

17. Assume that X0 = x < r and let Sa+ := inf{t : Xt > a}, a > x. The first
passage process
S + := {Sa+ : a ≥ x}
 3. PASSAGE TIMES 25

is non-decreasing and, due to the strong Markov property, has independent

increments. Using No. 10 and 14 it is seen (Itô and McKean (1974) p. 144)
that S + has the Lévy–Khintchine representation
 Z a Z ∞ 
+
Ex (e−αSa ) = exp − s(dy) (1 − e−αl )ny− (dl) + ny− (∞) .
x 0

Analogously, for Sa− := inf{t : X(t) < a}, a < x, we have

 Z x Z ∞ 
−αSa− −αl y y
Ex (e ) = exp − s(dy) (1 − e )n+ (dl) + n+ (∞) .
a 0

For the definitions of ny± , see No. 14.

18. Assume that X is recurrent and 0 ∈ I. Using the strong Markov property it is
seen that the process
L(H) := {L(Hb , 0) : b ≥ 0}
has independent increments. To find its Lévy–Khintchine representation observe
that for 0 ≤ a < b,

E0 (e−α(L(Hb ,0)−L(Ha ,0)) ) = Pa (Hb < H0 ) + Pa (H0 < Hb )E0 (e−αL(Hb ,0) ).

But (see No. 27)

1
E0 (e−αL(Hb ,0) ) = ,
(s(b) − s(0)) + 1
giving
(s(a) − s(0)) + 1
E0 (e−α(L(Hb ,0)−L(Ha ,0)) ) =
(s(b) − s(0)) + 1
 Z ∞ 
−αl
= exp − (1 − e )n(dl) ,
0

where

n(dl) := l−1 (e−β(b) l − e−β(a) l ) dl, β(x) := (s(x) − s(0))−1 .

For Brownian motion with a = 0 this result is proved in Itô and McKean (1974)
p. 72. The proof therein is based on the scaling property.

19. Here we consider the joint distribution of the maximum up to time t and its
location. Assume that the diffusion X is conservative and for t > 0 let

M̌t := sup{Xs : s ≤ t}.

Then for x < z,

Z t
Px (M̌t ≥ z) = Px (Hz < t) = nx (z, s)ds,
0
 26 II LINEAR DIFFUSIONS

where nx (z, ·) is the Px -density of Hz (Itô and McKean (1974) p. 154). Taking
the Laplace transform with respect to t gives
∞
1 (x)
Z
e−αt Px (M̌t ≥ z)dt = .
0
(z )

Notice that if τ ∼ Exp(α) and is independent of X, then

Px (M̌τ ≥ z) = Px (Hz < τ ) = Ex (exp(−αHz )).

In particular, if X does not hit r it follows that

r
(x )
Z
Ex (M̌τ ) = x + dz.
x
(z )

For z < r we have

Z ∞
(x)
Z
e−αt Px (M̌t ∈ dz)dt = 2 (z ) s(dz) ψα (y)m(dy),
0 [l,z)∩I

and ∞
(x) (y)
Z
e−αt Px (M̌t ∈ dz, Xt ∈ dy)dt = 2 (z ) s(dz) m(dy).
0

Let
Ȟ(t) := inf{s < t : Xs = M̌t }
be the first time point that X attains its maximum before t. (In fact, the time
point of the maximum is a.s. unique in the case M̌t < r.) Then for x ∨ y < z
and v < t,

Px (M̌t ∈ dz, Xt ∈ dy, Ȟ(t) ∈ dv) = nx (z, v)ny (z, t − v) s(dz) m(dy) dv.

This formula is valid for all (regular) diffusions, not just for conservative ones.
For a proof, see Csáki, Földes and Salminen (1987). Notice, however, that the
distribution might have an atom at r :

Px (M̌t = r, Xt ∈ dy, Ȟ(t) ∈ dv) = Px (Hr ∈ dv, Xt ∈ dy)

= nx (r, v)p(t − v; r, y) m(dy) dv.

Observe that the distribution of the global maximum M̌ := sup Xt is (cf. No.
10)
(x)
Px (M̌ > z) = Px (Hz < ∞) = lim Ex (e−αHz ) = lim .
α→0 α→0 (z )

20. Let y ∈ I be given and define the last passage or last exit time at y by

λy := sup{t : Xt = y}.
 4. ADDITIVE FUNCTIONALS AND KILLING 27

If X does not visit y at all we define λy = 0. Therefore, {λy = 0} = {Hy = ∞}.

If X is recurrent, then λy = ∞ a.s. for all y ∈ I. In the transient case λy < ∞
a.s. for all y ∈ I, we have (see Pitman and Yor (1981), Salminen (1984))

p(u; x; y)
Z t
(a) Px (0 < λy ≤ t) = du.
0
G0 (y; y)

Further
G0 (x; y)
Px (λy = 0) = Px (Hy = ∞) = 1 − Px (0 < λy < ∞) = 1 − .
G0 (y; y)

Now let t be fixed and define λty := sup{u < t : Xu = y}. To find the distribution
of λty let τ be an exponentially distributed random variable with parameter α,
independent of X, and let λτy := sup{u < τ : Xu = y}. From (a),

e− u p(u; x; x)
Px (λτx ∈ du) = du.
G (x; x)

Let f (·, x) be the inverse of the Laplace transform 1/αGα (x, x). Then

Px (λtx ∈ du) = f (t − u, x) p(u; x, x)du, t > u.

Notice also that

G + (x; x)
Ex (exp(−βλτx )) = .
G (x; x)
We refer to Chung and Getoor (1977), Chung (1982b), and (1995) for interpre-
tations and applications of last exit times in potential theory.

4. Additive functionals and killing

21. Let Pν denote the probability measure associated to a diffusion X when X is

started with the initial distribution ν. Let Ft and F∞ be the universal com-
pletions of the canonical filtration Ftc and the σ-algebra F∞ c
, respectively, (cf.
I.14). Following, e.g., Blumenthal and Getoor (1968) p. 148 or Revuz and Yor
(2001) p. 401, a stochastic process A = {At : t ≥ 0}, A0 = 0, taking values in
[0, ∞] is called an additive functional of X if it has the following properties for
every x ∈ I:
(a) At is Ft -measurable for every t ≥ 0,
(b) t 7→ At is Px -a.s. non-decreasing, right-continuous, and

At = Aζ− , t ≥ ζ,

(c) for every s and t,

As+t = At + 1I[0,ζ) (t) As ◦ θt , Px -a.s.

 28 II LINEAR DIFFUSIONS

It can be proved that every additive functional of a regular diffusion is a.s.

continuous. Indeed, we notice first that all additive functionals of a diffusion
are natural (Blumenthal and Getoor (1968) p. 153). Combining this with the
fact that all α-excessive functions are continuous (see No. 29) the claim follows
using Exercise (4.17) in Blumenthal and Getoor (1968) p. 293. The continuity
at t = ζ is clear from (b).
An additive functional A of a diffusion is said to have the strong Markov property
if
(c’) for every Ft -stopping time T and non-negative random variable S

AT (ω)+S(ω) (ω) = AT (ω) (ω) + 1I[0,ζ(ω)) (T (ω)) AS(ω) (θT (ω)) Px -a.s.

From Proposition (1.13) in Blumenthal and Getoor (1968) p. 152 it follows that
every additive functional of a diffusion has the strong Markov property.

22. Let X̂ be a diffusion having a killing measure k 6≡ 0. We describe here how a

version of X̂ can be constructed from a diffusion X having the same speed and
scale as X̂ but for which the killing measure equals zero. It is assumed that
both X̂ and X attain values on the same interval I.
Let {L(x, t) : x ∈ I, t ≥ 0} be the local time of X, and define
Z
L(k) (t) := L(t, y) k(dy).
I

From the properties of local time in No. 13 it follows that L(k) is an additive
functional. Let τ ∼ Exp(1) independent of X, and introduce

Xt , L(k) (t) < τ,


Xt• :=
∂, L(k) (t) ≥ τ.

Then X̂ and X • are identical in law. We say that X is killed according to the
additive functional L(k) . For a proof, see Itô and McKean (1974) p. 179–183.

23. Let A be an arbitrary additive functional of a diffusion X. We assume, for

simplicity, that the killing measure of X equals zero. Let τ be as in No. 22 and
define 
Xt , At < τ,
X̂t :=
∂, At ≥ τ.

Then X̂ is a Markov process with continuous paths. Moreover, because A has

the strong Markov property it is seen that X̂ is a diffusion. Letting k denote
the killing measure of X̂ it follows from No. 22 that for every t a.s.
Z
At = L(t, x) k(dx).
I
 4. ADDITIVE FUNCTIONALS AND KILLING 29

Due to this property k is also called the representing measure of the additive
functional A.

24. Let X and L(k) be as in No. 22. The distribution of L(k) can be characterized
by choosing f ≡ 1 on I in the following double Laplace transform:
Z ∞   Z
(k)
e−γt Ex e−η L (t) f (Xt ) dt = G̃γ (x, y) f (y) m(dy),
0 I

where G̃ is the Green function of a diffusion with the characteristics m, s and

k̃ := η k. The equality above is a compact form of the famous Feynman–Kac
formula for diffusions. In Chapter VI this formula is studied for Brownian motion
in a number of special cases. Recall also that if k(dx) = k(x)m(dx), then the
occupation time formula (a) in No. 13 gives
Z Z t
L(k) (t) = L(t, y)k(y)m(dy) = k(Xs ) ds.
I 0

In particular, to find the distribution of L(t, z) for given t and z (under the
assumption m({z}) = 0), let the killing measure in No. 22 be the Dirac measure
at z. Then functions ψ and ϕ in terms of which the corresponding Green function
G̃ is expressed are solutions of the generalized differential equation (see No’s. 7
and 10)
Gu = γu
such that
u+ (z+) − u+ (z−) = ηu(z),
or, equivalently,
u− (z+) − u− (z−) = ηu(z).
In the absolutely continuous case (cf. No. 9) we may consider the local time
with respect to the Lebesgue measure. Letting ` denote this local time we have

`(t, x) = m(x)L(t, x).

Now ϕ and ψ, associated to the distribution of ` via the Feynman–Kac formula,

are solutions of the ODE
1 2
a (x) u00 (x) + b(x) u0 (x) = γu(x)
2
such that
2
u0 (z+) − u0 (z−) = 2 u(z).
a (z )

The distribution of L(t, x) is obtained more easily, perhaps, by choosing ζ in

(a) in No. 27 to be exponentially distributed and independent of X. Then the
result (a) in No. 27 gives the inversion with respect to η of the double Laplace
transform computed above.
 30 II LINEAR DIFFUSIONS

25. Let A be an additive functional of X and ko the measure associated to it as

explained in No. 23. Then by (d) in No. 13,
Z Z t
(a) Ex At = ko (dy) ds p(s; x, y).
I 0

To find the second moment notice that

Z t Z t  Z t Z t 
Ex A2t = Ex dAt1 dAt2 = 2 Ex dAt1 dAt2
0 0 0 t1
Z Z t
=2 ko (dy) dt1 p(t1 ; x, y)Ey (At−t1 ).
I 0

We may now proceed recursively and obtain

Z t Z t n−1  Z t Z t n−1 
Ex Ant = Ex dAt1 dAt2 = n Ex dAt1 dAt2
0 0 0 t1
Z Z t
=n ko (dy) dt1 p(t1 ; x, y)Ey (An−1
t−t1 ).
I 0

This formula is called the Kac moment formula, see Kac (1949, 1951), where a
special case is treated. Here and in the next two sections our main reference is
Fitzsimmons and Pitman (1999) in which more general processes are considered.
Now let A(1) and A(2) be two additive functionals with representing measures k1
and k2 , respectively. Then
Z t Z t 
Ex (A(1) (2)
t At ) = Ex dA(1)
t1 dA(2)
t2
0 0
Z t Z t  Z t Z t 
(1) (2) (2)
= Ex dAt1 dAt2 + Ex dAt2 dA(1)
t1
0 t1 0 t2
Z Z t
= k1 (dy) dt1 p(t1 ; x, y)Ey (A(2)t−t1 )
I 0
Z Z t
+ k2 (dy) dt2 p(t2 ; x, y)Ey (A(1)
t−t2 ).
I 0

In particular, for local times we have

Z t
Ex (L(t, y)n ) = n dt1 p(t1 ; x, y)Ey (L(t − t1 , y)n−1 )
0

and
Z t
Ex (L(t, y)L(t, z)) = dt1 p(t1 ; x, y)Ey (L(t − t1 , z))
0
Z t
+ dt1 p(t1 ; x, z)Ez (L(t − t1 , y)).
0
 4. ADDITIVE FUNCTIONALS AND KILLING 31

Further, for f ∈ BB (I),

Z Z t
Ex (At f (Xt )) = ko (dy) ds p(s; x, y)Ey (f (Xt−s )).
I 0

Choosing here f ≡ 1 on I gives Ex (Ant ; t < ζ). In general,

Z Z t
Ex (Ant f (Xt )) = n ko (dy) ds p(s; x, y)Ey (An−1
t−s f (Xt−s )).
I 0

26. We present here some formulae associated to the location of X at its lifetime ζ.
First, recall the formula displayed for indicators in No. 4:
Z t Z
(a) Ex (f (Xζ− ); ζ ≤ t) = ds k(dy) p(s; x, y)f (y), f ∈ BB (I).
0 I

Because f (∂) = 0 this does not, in general, give the distribution of ζ. On the
other hand, Z
Px (ζ > t) = p(t; x, y)m(dy).
I

From (a) we obtain

Z
Ex (e−γζ f (Xζ− )) = Gγ (x, y)f (y)k(dy).
I

Using the properties of ψ0 and ϕ0 given in No. 10 it follows that for l < a <
b ≤ x ≤ c < d < r,

Px (a ≤ Xζ− < b) = ϕ0 (x)(ψ0− (b) − ψ0− (a)),

Px (c < Xζ− ≤ d) = ψ0 (x)(ϕ+ +
0 (d) − ϕ0 (c)),

where ψ0 and ϕ0 are normalized so that

G0 (x, y) = ψ0 (y)ϕ0 (x), x ≥ y.

Furthermore, in the case ζ < ∞ and Xζ− ∈ I a.s., we have

Z
Gγ (x, y)(k(dy) + γ m(dy)) = 1.
I

27. Let A be an additive functional of X and ko the measure associated to it as

explained in No. 23. Then from (a) in No. 25,
Z
Ex Aζ = G0 (x, y)ko (dy)
I
 32 II LINEAR DIFFUSIONS

and Z
Ex (Anζ ) =n G0 (x, y)Ey (An−1
ζ )ko (dy).
I

In particular,
Ex (L(ζ, x)n ) = n! G0 (x, x)n ,
which leads to


(a) Px (L(ζ, x) > α) = exp − .
G0 (x; x)

Further,
Z
(b) Ex (Anζ f (Xζ− )) = n G0 (x, y)Ey (An−1
ζ f (Xζ− ))ko (dy).
I

Choosing here f ≡ 1 on I gives Ex (Anζ ; Xζ− ∈ I).

5. Excessive functions

28. A non-negative measurable function h : I 7→ R ∪ {∞} is called α-excessive,

α ≥ 0, for X if it has the following two properties:
(?) e−αt Ex (h(Xt )) ≤ h(x) for all x ∈ I and t ≥ 0,
(?) e−αt Ex (h(Xt )) → h(x) for all x ∈ I as t ↓ 0.
An α-excessive function h is called α-invariant if for all x ∈ I and t ≥ 0,

e−αt Ex (h(Xt )) = h(x).

A 0-excessive function is simply called excessive and, similarly, a 0-invariant

function is called invariant. Recall the convention that a function f defined on
I is extended to I ∂ := I ∪ {∂} by setting f (∂) = 0

29. A regular diffusion is standard in the sense of Dynkin (1965) Vol I p. 104
and Blumenthal and Getoor (1968). Consequently, using Dynkin (1965) Vol II
Theorems 15.5 p. 134 and 12.4 p. 7 (see also Blumenthal and Getoor (1968)
p. 93) it is seen that a non-negative function h is α-excessive if and only if h is
continuous and satisfies for all x ∈ I,

(a) Ex (e−αT h(XT )) ≤ h(x),

where T := inf{t : Xt ∈ Γ} and Γ is an arbitrary compact subset of I (recall

the convention h(XT ) = 0 if T = ∞). Because XHz = z a.s. when Hz < ∞, we
obtain from No. 10 and (a) that an α-excessive function h satisfies the inequality
(x )

, x ≤ z,
h(x) (z )


≥
h(z )   ' (x) , x ≥ z.
' (z )
 5. EXCESSIVE FUNCTIONS 33

From (a) and because X is assumed to be regular it follows that, if h is α-

excessive and h(x) = 0 for some x ∈ I, then h ≡ 0. Further, if h is α-excessive
and h(x) = ∞ for some x ∈ I, then h ≡ ∞ on I. These two functions are called
trivial α-excessive functions.
If X is recurrent (see No. 12), then a non-negative measurable function h is
excessive if and only if it is constant on I. The following converse is also true: if
X is a diffusion such that every excessive function for X is constant on I, then
X is recurrent. See Getoor (1980) for a discussion of recurrence and transience
for general Markov processes.
We remark also that if h is excessive (invariant), then, by the Markov property,
{h(Xt ) : t ≥ 0} is a positive supermartingale (martingale) with respect to the
natural filtration of X. In particular, it follows from the martingale convergence
theorem that limt→∞ h(Xt ) exists a.s., and from the optional stopping theorem

Ex (h(XT ) ≤ Ex (h(XS ))

for all stopping times S and T such that S ≤ T.

30. Using (a) in No. 29 it is seen that for every y ∈ (l, r) the function x 7→ Gα (x, y)
is α-excessive, as are x 7→ ψα (x) and x 7→ ϕα (x).
These functions are minimal in the sense that an arbitrary non-trivial α-exces-
sive function h can be represented as a linear combination of them. To give
the exact statement let xo ∈ I be given and consider a non-negative measurable
function h on I such that h(xo ) = 1. Then h is α-excessive if and only if there
exists a probability measure µ on [l, r] such that for all x ∈ I,

G (x; y) ' (x) (x)

Z
(a) h(x) = µ(dy) + µ({l}) + µ({r}).
(l,r)
G (x o ; y ) ' ( x o ) (xo )

The measure µ is called the representing measure of h.

From (a) it can be deduced (see Salminen (1985)) that α-excessive functions of
regular diffusions are continuous. Moreover, if ϕα and ψα are differentiable at a
given point z ∈ I then for any α-excessive function h it holds that h+ and h−
exist and satisfy
h− (z) ≥ h+ (z).
Furthermore, h is differentiable with respect to the scale function if and only if
the representing measure of h does not charge z.
In general, the idea to represent α-excessive functions in terms of the minimal
ones comes from classical potential theory. The corresponding problem therein
is to represent positive harmonic functions on Greenian subsets of Rn in terms
of the minimal harmonic functions. See Doob (1984) Chapter XII: The Martin
Boundary, p. 195–225, Bass (1995), and Chung and Walsh (2005) Chapter
14. One-dimensional diffusions are especially treated in, e.g., Lai (1973) and
Salminen (1984, 1985)).
 34 II LINEAR DIFFUSIONS

31. Let h, xo , and µ be as in No. 30. Recall that the sample space of X is the
canonical space (I ∂,∞ , F∞
c
). The elements of this space are denoted by ω and
they are functions mapping [0, ∞) to I ∂ (see I.12). Using the probability measure
P associated to X we construct a new probability measure Ph by setting, for
A ∈ Ftc ∩ {ζ > t},
h(!(t))
 
Phx (A) := e−αt Ex ;A .
h(x)
c
Let T be an Ft+ -stopping time and recall the definition of FTc + in No. 1. Then
c
for A ∈ FT + ∩ {ζ > T },

h(!(T ))
 
Phx (A) = Ex e−αT ;A .
h(x)

The co-ordinate process under the measure Ph is called Doob’s h-transform or

an α-excessive transform of X. It is a regular diffusion and we let X h denote
this process.
Let m be the speed measure of X and p the transition density with respect to
m. Then the semigroup associated to X h is

h(!(t))
 
Pth f (x) := Ehx (f (ω(t))) = e−αt Ex f (ω(t))
h(x)
h(y )
Z
= e−αt p(t; x, y)f (y) m(dy), f ∈ Cb (I).
I
h(x)

In many cases, we can readily check that the basic characteristics of X h are
given as follows:
(a) speed measure mh (dy) = h2 (y)m(dy),
(b) scale measure sh (dy) = h−2 (y)s(dy),
G (x; y)
(c) killing measure k h (dy) = (Gh (xo , y))−1 µ(dy), y ∈ I, Gh (x, y) := .
h(x)h(y)
The transition density ph of X h with respect to mh is given by

e− t p(t; x; y)
ph (t; x, y) := ,
h(x)h(y)

and, hence, Gh as defined above is the Green function of X h . We refer to Ja-

cobsen (1974), Williams (1974), and Salminen (1984), where h-transforms based
on the scale function and the Green kernel are appearing and analysed. For the
connection between the representing measure of an excessive function and the
terminal distribution of the corresponding h-transform in a general Markovian
setting, see Kunita and Watanabe (1965). For diffusions the terminal distribu-
tion can be expressed, on the other hand, in terms of the killing measure as
explained in No. 26. Combining these two approaches results to the formula
for the klling measure of the h-transform presented in (c). See also Langer and
Schenk (1990).
 5. EXCESSIVE FUNCTIONS 35

We stress furthermore that the boundary behaviors of X and X h can be very

different. For example, an entrance boundary of X might turn into a non-
singular killing boundary for X h .

32. Constructing an excessive transform of a diffusion, or more generally of a Markov

process, can be viewed as conditioning the process to behave in some desired
fashion.
For example, assume that the diffusion X is transient but k ≡ 0. Then
 
Px lim Xt = l or r = 1.
t→ζ
+
Now let (X , P+
· )
be the excessive transform of X obtained using the function
ψ0 . It can be proved that this ψ0 -transform has the property
 
+
P+x lim Xt = r = 1.
t→ζ

If Px (limt→ζ Xt = r) > 0, the function ψ0 (x) is a constant multiple of this prob-

ability, and consequently, X + is identical in law to X given that limt→ζ X(t) = r.
An illustrative case is when X is a Brownian motion with drift µ < 0. The
ψ0 -transform is a Brownian motion with drift −µ > 0. For examples of excessive
transforms of other types see IV.23. See also Doob (1984) Chapter X.

33. Excessive transforms also arise naturally when we are considering time reversed
processes.
Let X be as in No. 32. Fix two points a, b ∈ I and let ha := G0 (·, a) and
hb := G0 (·, b). Furthermore, let (X a , Pa· ) and (X b , Pb· ) denote the ha - and hb -
transforms of X, respectively. Then
{Xta , 0 < t < ζ; Pab } ∼ {Xζ−t
b
, 0 < t < ζ; Pba }.
We remark also that
{Xta , 0 ≤ t < ζ; Paa } ∼ {Xt , 0 ≤ t < λa ; Pa },
where λa := sup{s : Xs = a} is the last exit time at a (cf. No. 20).
An important special case is obtained when we take X to be a Brownian motion
living in R+ and killed at the first hitting time to zero. Then, letting a = 0 and
b > 0, we have h0 ≡ 1 and hb (x) = x ∧ b. Hence, the time reversal of X, we
assume that X0 = b, is identical in law to the hb -transform of X (started at 0).
Simple computations show further that this transform is identical in law to a
3-dimensional Bessel process killed at the last exit time λb . More generally, for
ν > 0 let R(ν) be a Bessel process of order ν (see IV.39) starting at 0 and R(−ν)
a Bessel process of order −ν starting at a point a > 0; if −1 < −ν we let 0 be a
killing boundary for R(−ν) . Then
{Rλ(ν)a −t : 0 ≤ t ≤ λa } ∼ {Rt(−ν) : 0 ≤ t ≤ H0 }.
We refer to Williams (1974) for additional information, see also Meyer, Smythe
and Walsh (1971), Jacobsen (1974), Pitman and Yor (1981), Sharpe (1980),
Salminen (1984). These results can also be obtained in the general time reversal
framework of Nagasawa (1964).
 36 II LINEAR DIFFUSIONS

6. Ergodic results

34. Let X be a recurrent diffusion and {L(t, x) : x ∈ I, t ≥ 0} its local time (with
respect to the speed measure). Let µ1 and µ2 be two measures on B(I) and
introduce, for i = 1, 2,
Z
L(i) (t) := L(t, x) µi (dx).
I

Then in the case 0 < µ2 (I) < ∞ we have (Itô and McKean (1974) p. 228–229,
Revuz and Yor (2001) p. 427, Rogers and Williams (2000) p. 300) a.s.

L(1) (t)  (I )
lim = 1 .
t→∞ L(2) (t) 2 (I )

35. We present here a number of special cases of the general result in No. 34. All
limits below exist a.s.
Consider first the case in which µi (dx) = fi (x)m(dx), where fi ∈ BB (I), fi ≥ 0,
and m is the speed measure of X. Due to the occupation time formula (a) in
No. 13 we have Z t
L(i) (t) = fi (Xs )ds.
0
R
Consequently, if I
f2 (x) m(dx) < ∞,
t f (X )ds f1 (x)m(dx)
1 s
R R
(a) lim R0t = RI .
t→∞ 0 f2 (Xs )ds I 2 (x)m(dx)
f

Letting Ai ∈ B(I) and taking fi = 1IAi in (a) gives in the case 0 < m(A2 ) < ∞,

Leb{s < t : Xs ∈ A1 } m(A1 )

lim = .
t→∞ Leb{s < t : Xs ∈ A2 } m(A2 )
Let x ∈ I be given and choose µ2 = ε{x} , i.e., µ2 is the Dirac measure at x.
Then L(2) (t) = L(t, x) and we have

L(1) (t)
(b) lim = µ1 (I).
t→∞ L(t; x)

Let µ2 be the Dirac measure at x as in (b) and µ1 the Dirac measure at another
point y ∈ I. Then
L(t; y)
lim = 1.
t→∞ L(t; x)

Letting µ2 = m, the speed measure of X, and using the occupation time formula
give L(2) (t) = t. Consequently, in the case µ1 (I) < ∞, we have

L(1) (t)  (I )
lim = 1 ,
t→∞ t m(I )
 6. ERGODIC RESULTS 37

where the right-hand side is defined to be zero when m(I) = ∞. Taking here µ1
to be the Dirac measure at y gives
L(t; x) 1
(c) lim = .
t→∞ t m(I )

36. Let X be as in No. 34, and let ν1 and ν2 be two initial probability distributions.
Introduce for i = 1, 2 and A ∈ B(I),
Z
νi (t, A) := Px (Xt ∈ A)νi (dx), ν(t, A) := ν1 (t, A) − ν2 (t, A).
I

Consider the total variation of the signed measure ν(t, ·) :

kν(t)k := sup{|ν(t, A)| + |ν(t, I \ A)| : A ∈ B(I)}

= 2 sup{|ν(t, A)| : A ∈ B(I)}.

Then

(a) lim kν(t)k = 0.

t→∞

For a proof, see Lindvall (1983), where rates of convergence are also derived.
See also Rogers and Williams (2000) p. 301. We remark that for (a) to hold for
a general regular diffusion it is necessary and sufficient that the tail σ-field of
the diffusion is trivial. This is the case, e.g., for recurrent diffusions, see Rösler
(1979).
Assume now that X is positively recurrent. Then, m(I) < ∞ and π(·) :=
m(·)/m(I) is the stationary probability measure of X. Let x ∈ I be given and
choose ν1 = ε{x} and ν2 = π. Then the result above gives

lim kPt (x, ·) − πk = 0.

t→∞

In particular, we have for every f ∈ BB (I),

Z
(b) lim Ex (f (Xt )) = f (y) π(dy).
t→∞ I

37. Let X be positively recurrent. Then using (b) in No. 36 and l’Hopital’s rule we
have

1
Z t  Z
(a) lim Ex f (Xs ) ds = f (y) π(dy), f ∈ BB (I).
t→∞ t 0 I

Combining (a) with (a) in No. 35, where we take f1 = f and f2 ≡ 1, shows that
the convergence
1 t
Z Z
lim f (Xs ) ds = f (y) π(dy)
t→∞ t 0 I
38 II LINEAR DIFFUSIONS

takes place also in L1 . We remark that (a) can also be deduced using the
Chacon–Ornstein ergodic theorem, see Revuz and Yor (2001) (3.18) Exercise p.
429. From this approach it also follows that

1 1 t
1
Z
lim E(L(t, x)) = lim p(s; x, y) ds = ;
t→∞ t t→∞ t 0
m(I )

hence, (c) in No. 35 holds also in L1 . Moreover, it can be proved using spectral
representations (see Salminen (1996)) that

1
lim p(t; x, y) = ∀x, y ∈ I.
t→∞ m(I )
 CHAPTER III

STOCHASTIC CALCULUS

1. Stochastic integration with respect to Brownian motion

1. We present here very briefly the basic facts of the theory of stochastic integration
in the case, where the integrator is a Brownian motion. Let (Ω, F, {Ft }, P) be a
probability space satisfying the usual conditions (see I.3) and W an Ft -Brownian
motion (see IV.1) in this space. The goal is to give some meaning to stochastic
integrals of the type
Z t
(a), f (Ws ) dWs ,
0

where f is a “smooth” function. Because t 7→ Wt is of infinite variation on any

interval, it is not possible to define such integrals by using classical approaches
from the theory of integration. This difficulty was overcome by Kiyosi Itô. His
far-reaching basic idea presented in Itô (1944) is that stochastic integrals of the
type (a) can be defined via an isometry. The notion to which this approach
leads us is called the Itô integral and the theory is called the Itô calculus. In
particular, using the differentiation rule in the Itô calculus (see No. 8), it is seen
that for all t ≥ 0, Z t
2 Ws dWs = Wt2 − t a.s.
0

In this chapter the words “a.s.” in the statements of the above kind are usually
omitted.
A classical textbook on Itô’s theory is McKean (1969). Our discussion of sto-
chastic integrals follows mainly Chung and Williams (1990).

2. We begin by introducing the so-called predictable (or previsible) processes, a

subset of which appears as the set of integrands in the general theory of stochastic
integration. To consider only predictable integrands instead of progressively
measurable (see I.4) is not a loss of generality because stochastic integrals with
progressively measurable integrands are indistinguishable from integrals with
predictable integrands (see Ikeda and Watanabe (1981) p. 45, Revuz and Yor
(2001) p. 172).
The predictable σ-field P is the smallest σ-field which contains all sets of the
form {0} × A and (a, b] × B, where 0 < a < b, A ∈ F0 , and B ∈ Fa . A function
F : [0, ∞) × Ω 7→ R is called predictable if it is P-measurable. Analogously, a

© Springer Basel 2015 39

A.N. Borodin, P. Salminen, Handbook of Brownian Motion - Facts and Formulae,
Probability and Its Applications, DOI 10.1007/978-3-0348-8163-0_3
40 III STOCHASTIC CALCULUS

stochastic process X = {Xt : t ≥ 0} is predictable if the mapping (t, ω) 7→ Xt (ω)

is predictable. By the monotone class theorem, predictable processes are adapted
(and progressively measurable).
We remark that P is also the smallest σ-field on [0, ∞) × Ω making all adapted
and (left) continuous processes measurable (Revuz and Yor (2001) p. 171). The
smallest σ-field on [0, ∞) × Ω making all adapted and right continuous processes
measurable is called the optional σ-field. Using for this the notation O, we have

P ⊂ O ⊂ P rog,

where P rog stands for the σ-field generated by the progressively measurable
processes.
Let (ai , bi ] × Ai , i = 1, 2, . . . , n, be predictable such that 0 < a1 < b1 ≤ a2 <
b2 ≤ · · · ≤ an < bn , and define the predictable process C = {Ct : t ≥ 0} by
setting
n
X
Ct (ω) = λi 1I(ai ,bi ] (t)1IAi (ω) + λ0 1I0 (t)1IA0 (ω), λi ∈ R.
i=1

Processes of this type are called simple predictable processes.

3. The stochastic integral of C with respect to W is defined to be

Z n
X
Cs dWs := λi 1IAi (ω)(Wbi (ω) − Wai (ω)).
i=1

Because of the simple predictable structure of C we have

Z 2 Z ∞
E Cs dWs =E Cs2 ds.
0

This formula is of key importance for further development.

4. Let L2 be the space of square integrable random variables. Then L2 is a Hilbert

space when equipped with the norm

k X k2 := (EX 2 )1/2 , X ∈ L2 .

Further, let L2 be the space of predictable processes such that

Z ∞
(a) E Xs2 ds < ∞.
0

Then L2 is a Hilbert space when equipped with the norm

 Z ∞ 1/2
kX k◦2 := E Xs2 ds , X ∈ L2 .
0
 1. STOCHASTIC INTEGRATION 41

In particular, if C is a simple predictable process, then C ∈ L2 and the mapping

Z
(b) C 7→ Cs dWs

is an isometry from a subset of L2 into L2 . The set of simple predictable pro-

cesses is dense in (L2 , k · k◦2 ) (Chung and Williams (1990) Lemma 2.4 p. 37).
Consequently, the linear isometry in (b) can be extended uniquely to a linear
isometry from the whole of L2 into L2 , thus defining the stochastic integral of
X ∈ L2 with respect to the Brownian motion W.
We remark that if {C (n) } is a sequence of simple predictable processes converging
in norm k · k◦2 to a predictable process X ∈ L2 , then the corresponding sequence
of stochastic integrals converges in L2 to the stochastic integral of X.

5. Let L2` be the set of predictable processes X such that Y (t) := X 1I[0,t] ∈ L2
for each t ≥ 0. The stochastic integral of Y (t) is well defined for each t and we
denote it by
Z t Z
Xs dWs := Ys(t) dWs .
0

By the isometry property we have

Z 0 Z
Xs dWs := Ys(0) dWs = 0.
0

The set L2` of integrands can be further extended by a localization argument.

Indeed, let L2loc be the set of predictable processes X for which there exists a
sequence of stopping times {Tk } such that Tk ↑ ∞ and
Z ∞
E Xs2 1I[0,Tk ] ds < ∞,
0

where
[0, Tk ] := {(t, ω) : 0 ≤ t ≤ Tk (ω), ω ∈ Ω}.

Then viewing W as a local martingale and recalling that [0, Tk ] is predictable

(see Chung and Williams (1990) p. 28) the stochastic integral of X ∈ L2loc with
respect to W can be introduced on [0, Tk ] for every k, similarly as is done above
for deterministic t. These different constructions are consistent, and hence lead
to the stochastic integral of X on the whole time axis. See, e.g., Chung and
Williams (1990) p. 44 or Revuz and Yor (2001) p. 140. We remark (see Rogers
and Williams (2000) p. 55) that L2loc consists precisely of predictable processes
such that for all t a.s. Z t
Xs2 ds < ∞.
0
42 III STOCHASTIC CALCULUS

6. We give here some properties of the stochastic integral.

Additivity I. For X ∈ L2loc
Z t+s Z t Z t+s
Xu dWu = Xu dWu + Xu dWu .
0 0 t

Additivity II. Let A and B be a.s. finite F0 −measurable random variables and
X, Y ∈ L2loc . Then
Z t Z t Z t
(A Xs + B Ys ) dWs = A Xs dWs + B Ys dWs .
0 0 0

Covariance. For X, Y ∈ L2`

Z t Z t  Z t 
E Xs dWs Ys dWs = E Xs Ys ds .
0 0 0

Martingale property and continuity. For X ∈ L2loc

nZ t o
X · W := Xs dWs : t ≥ 0
0
defines a continuous local martingale. Furthermore, if X ∈ L2` , then X · W is a
continuous L2 −martingale (Chung and Williams (1990) p. 38).
Riemann sums. Let X ∈ L2loc be left continuous and {t(in) : i = 1, 2, . . . , n} a
sequence of subdivisions of the interval [0, t] such that
sup(t(in) − t(i−1
n)
)→0 as n → ∞.
i
Then (Revuz and Yor (2001) p. 142)
Z t Xn
Xs dWs = lim Xt(n) (Wt(n) − Wt(n) ),
0 n→∞ i i+1 i
i=1
where the limit is in probability.

7. Let {Gto }be the natural completed filtration of Brownian motion W, and let
L2 (W ) be the class of predictable processes X with respect to {Gto } such that
(a) in No. 4 holds. Then the Itô representation theorem states (see, e.g., Revuz
and Yor (2001) p. 199 or Øksendal (1998) p. 51) that for any square integrable
o
G∞ -measurable random variable F there exists a unique H ∈ L2 (W ) such that
Z ∞
F = E(F ) + Hs dWs .
0
o
Next, let M be a square integrable Gt -martingale and L2` (W ) the class of pre-
dictable processes X such that X 1I[0,t] ∈ L2 (W ) for all t ≥ 0. Then there exists
a unique H ∈ L2` (W ) such that
Z t
Mt = E(M0 ) + Hs dWs .
0
This result, the martingale representation theorem, can also be formulated for
local martingales, see Revuz and Yor (2001) p. 199. An important consequence
is that all local Gto -martingales have continuous versions.
 ^ AND TANAKA FORMULAE
2. THE ITO 43

2. The Itô and Tanaka formulae

8. It is often of interest to study the properties of the process f (W ) = {f (Wt ) :

t ≥ 0}, where f is a given smooth function. In the case, where f is twice
continuously differentiable, the celebrated Itô formula says that
t
1 t
Z Z
f (Wt ) − f (W0 ) = f 0 (Ws ) dWs + f 00 (Ws ) ds.
0
2 0

Notice that this is also the semimartingale decomposition of f (W ). The stochas-

tic integral is well defined because f 0 (W ) ∈ L2loc due to continuity.

9. For a more general form of the Itô formula consider a real valued function
(x, y) 7→ f (x, y) such that fx0 , fxx
00
, and fy0 are continuous. Let V be a con-
tinuous Ft -adapted process of finite variation on finite intervals. Then

f (Wt , Vt ) − f (W0 , V0 )
1 t 00
Z t Z t Z
0 0
= fx (Ws , Vs ) dWs + fy (Ws , Vs ) dVs + fxx (Ws , Vs ) ds,
0 0
2 0

where the second integral on the right-hand side is a Lebesgue–Stieltjes integral.

10. Consider a family of continuous semimartingales {Z (i) ; i = 1, 2, . . . , n} given by

Z t Z t
(i) (i) (i)
Zt := Z0 + Xs dWs + Ys(i) ds,
0 0

where Z0(i) are F0 -measurable random variables, X (i) ∈ L2loc , and Y (i) are Ft -
adapted and B[0, ∞) × F-measurable processes such that for all t > 0,
Z t
|Ys(i) | ds < ∞.
0

Let f : R+ × Rn 7→ R be a continuous function of (t, x1 , . . . , xn ) such that the

partial derivatives

@f @f @2f
f00 := fi0 := 00
fij :=
@t @xi @xi @xj

~ t := (Zt(1) , Zt(2) , . . . , Zt(n) ) we have

exist and are continuous. Then setting Z
Z t
~ t ) − f (0, Z
f (t,Z ~0) = ~ s ) ds
f00 (s, Z
0
n Z n Z
X t
~ s )(Xs(i) 1 X t 00 ~
+ fi0 (s, Z dWs + Ys (i)
ds) + fij (s, Zs )Xs(i) Xs(j ) ds.
0
2 0
i=1 i,j=1
 44 III STOCHASTIC CALCULUS
→
11. Let W := (W (1) , W (2) , . . . , W (d) ) be a d-dimensional Brownian motion, that
is, W (i) are independent linear Brownian motions. For a continuous function
f : Rd 7→ R such that fi0 and fii00 exist and are continuous for every i, we extend
the Itô formula to Rd :
d Z d
t
1X t
Z
→ → X → →
f (W t ) − f (W 0 ) = fi0 (W s ) dWs(i) + fii00 (W s ) ds
0
2 0
i=1 i=1
t
1 t
Z Z
→ → →
= ∇f (W s ) dW s + ∆f (W s ) ds.
0
2 0

12. Let `(·, x) denote the Brownian local time at the point x ∈ R such that

1 t
Z
`(t, x) = lim 1I(x−,x+) (Ws ) ds,
↓0 2 0

where the limit exists a.s. and in L2 (cf. IV.8). We may and do take ` to be
jointly continuous and Ft -adapted. Setting sgn(y) := 1I(0,+∞) (y) − 1I(−∞,0) (y),
the Tanaka formula says that
Z t
(a) |Wt − x| − |W0 − x| = sgn(Ws − x) dWs + `(t, x).
0

There is also a one-sided version

t
1
Z
(Wt − x) ∨ 0 − (W0 − x) ∨ 0 = 1I(0,∞) (Ws − x) dWs + `(t, x).
0
2

13. To generalize (a) in No. 12, let F be a linear combination of convex functions.
Recall (Roberts and Varberg (1973) p. 23) that this is the case if and only if
there exists a function f of finite variation such that
Z x
F (x) = F (0) + f (y)dy.
0

Then letting f (dy) be the signed measure associated to f via its representation
as a difference of two increasing functions, we have
∞
t
1
Z Z
F (Wt ) − F (W0 ) = f (Ws ) dWs + `(t, y) f (dy).
0
2 −∞

Next, consider a function F given by

Z x
(a) F (x) := f (y)dy,
0
 3. STRONG SOLUTIONS 45

where f is locally bounded and measurable. Then (see Bouleau and Yor (1981))
∞
t
1
Z Z
F (Wt ) = f (Ws ) dWs − f (y) `(t, dy).
0
2 −∞

Finally, letting F be as in (a) with f locally square integrable, it is proved in

Föllmer, Protter and Shiryayev (1995) that
t
1
Z
F (Wt ) = f (Ws ) dWs + [f (W ), W ]t ,
0
2

where
n
X
[f (W ), W ]t := lim (f (Wt(n) ) − f (Wt(n) ))(Wt(n) − Wt(n) )
n→∞ i+1 i i+1 i
i=1

and the limit exists in probability along a sequence of subdivisions {t(in) } as

introduced in No. 6. In Föllmer and Protter (2000) this result is generalised for
multidimensional Brownian motion.

3. Stochastic differential equations – strong solutions

14. Let (Ω, F, P) be a probability space on which a d−dimensional Brownian motion

W = {Wt : t ≥ 0}, and a finite-valued random variable ξ ∈ Rr with law µ
are given. Assume that W is started from the origin and that ξ and W are
independent. Introduce for t ≥ 0,

Ft := σ{ξ, {Ws : s ≤ t}},

and let (Ω, G, {Gt }, P) be the usual augmentation of (Ω, F, {Ft }, P). Then, fol-
lowing Rogers and Williams (2000) p. 125, the object (Ω, G, {Gt }, P, W, ξ) is
called a minimal set-up.
Let a : R+ ×Rr 7→ Rr ×Rd and b : R+ ×Rr 7→ Rr be Borel-measurable functions.
A process X, X0 = ξ, on the set-up (Ω, G, {Gt }, P) taking values in Rr is said to
be a strong non-exploding solution with the initial distribution µ of the stochastic
differential equation (SDE)

(†) dXt = a(t, Xt )dWt + b(t, Xt )dt

if
(a) X is Gt −adapted,
Rt
(b) 0 (|a(s, Xs )|2 + |b(s, Xs )|) ds < ∞ for all t a.s.,
Rt Rt
(c) Xt = ξ + 0 a(s, Xs ) dWs + 0 b(s, Xs ) ds for all t a.s.
Note that due to (b) the integrals in (c) are well defined. It is also clear from
(c) that X is a continuous semimartingale – in fact, a (non-time-homogeneous)
 46 III STOCHASTIC CALCULUS

diffusion. Non-exploding means that X is well defined for all t. The notation | · |
stands for the usual Euclidian norm, i.e., if A = (Ai,j ) is an r × d matrix, then
X
|A|2 := |Ai,j |2 .
i,j

Strong uniqueness is said to hold for the equation (†) if any two strong solutions
in any given minimal set-up are indistinguishable or equivalently, because of the
continuity, modifications of each other. Observe that to verify strong uniqueness
we have to consider all minimal set-ups.
Remark. Introducing the notion of strong uniqueness we follow the approach
in Karatzas and Shreve (1991) Chapter 5. It is also common to work with the
more general notion of pathwise uniqueness (see No. 18).

15. Assume that the functions a and b satisfy the “global” Lipschitz condition
(a) there exists a constant C such that for all t and x, y ∈ Rr
|a(t, x) − a(t, y)| + |b(t, x) − b(t, y)| ≤ C|x − y|,
and the boundedness condition
(b) for all T > 0 there exists a constant CT such that for all t ≤ T
|a(t, 0)| + |b(t, 0)| ≤ CT .
Under these conditions the SDE (†) has a unique strong non-exploding solution,
see Itô (1951) and, e.g., Rogers and Williams (2000) p. 128, where the SDE (†)
is in this case said to be pathwise exact.
We remark also that in this case the strong solution X can be expressed a.s. as
X = F (ξ, W ),
where F is a measurable mapping from Rr × C(R+ , Rd ) to C(R+ , Rr ) (for a proof
see Rogers and Williams (2000) p. 136; for the definition of measurability in
this context see ibid. p. 127).

16. Assume that the functions a and b satisfy the “local” Lipschitz condition
(a) for every N > 0 there exists a constant CN such that for all t and |x|, |y| ≤
N,
|a(t, x) − a(t, y)| + |b(t, x) − b(t, y)| ≤ CN |x − y|,
and the growth condition
(b) for all T > 0 there exists a constant CT such that for all t ≤ T and x ∈ Rr ,
|a(t, x)| + |b(t, x)| ≤ CT (1 + |x|).
Under these conditions the SDE (†) has a unique strong non-exploding solution
(Rogers and Williams (2000) p. 132). The condition (b) is needed to guarantee
the solution to be non-exploding. Without (b) the equation still has a unique
strong solution up to the eventual explosion time.
 4. WEAK SOLUTIONS 47

4. Stochastic differential equations – weak solutions

17. We say that the SDE (†) in No. 14 has a weak solution (X, W ) with initial
distribution µ if there exists a filtered probability space (Ω, F, {Ft }, P) satis-
fying the usual conditions and carrying a Brownian motion W, a continuous
semimartingale X, and a random variable ξ having the law µ such that

(?) X and W are Ft −adapted,

Rt
(?) 0 (|a(s, Xs )|2 + |b(s, Xs )|) ds < ∞ for all t a.s.,
Rt Rt
(?) Xt = ξ + 0
a(s, Xs ) dWs + 0
b(s, Xs ) ds for all t a.s.

It is clear that a strong solution is always also a weak solution. See Ikeda and
Watanabe (1981) p. 152, Barlow (1982), (1988) for examples showing that the
converse is not true.

18. Let (X, W ) and (X 0 , W 0 ) be any two weak solutions of the SDE (†) defined on
the same filtered probability space such that W = W 0 and X0 = X00 a.s. Then
we say that pathwise uniqueness holds for (†) if X and X 0 are indistinguishable.

Let (X, W ) and (X 0 , W 0 ) be any two weak solutions of (†) with initial distribu-
tion µ. Then we say that uniqueness in law holds for (†) if X and X 0 are versions
of each other, i.e., have the same finite dimensional distributions.

It can be proved that pathwise uniqueness implies uniqueness in law. Moreover,

given that a weak solution exists for every initial distribution µ and that pathwise
uniqueness holds, then there also exists a unique strong solution. These results
are due to Yamada and Watanabe (1971); for proofs see Ikeda and Watanabe
(1981) p. 149 and 152, and Rogers and Williams (2000) p. 151.

19. Assume that the functions a and b in the SDE (†) are jointly continuous and
bounded in all variables (in the Euclidian norm | · |), and that µ has a compact
support. Then the equation (†) has a non-exploding weak solution (X, W ) with
initial distribution µ (Ikeda and Watanabe (1981) p. 155).

20. Assume that a and b in the SDE (†) do not depend explicitly on t. Let a? be
the transpose of the (matrix) function a, and suppose that a · a? is bounded,
continuous and uniformly positive definite, i.e., there exists α > 0 such that for
all x, λ ∈ Rr ,
λ? · a(x) · a? (x) · λ ≥ α λ? · λ .

Assume further that the function b is bounded. Then uniqueness in law holds
for the equation (†). This is due to Stroock and Varadhan (1969); see Ikeda and
Watanabe (1981) p. 171.
 48 III STOCHASTIC CALCULUS

5. One-dimensional stochastic differential equations

21. Let us take d = r = 1 in the SDE (†), and assume that a and b therein are
bounded. Then pathwise uniqueness holds if the following conditions are ful-
filled:
(a) there exists an increasing function ρ : [0, ∞) 7→ [0, ∞) such that ρ(0) = 0
and for all  > 0, Z 
ρ−2 (u) du = ∞,
0

and for every t, and x, y ∈ R,

|a(t, x) − a(t, y)| ≤ ρ(|x − y|),

(b) there exists a non-decreasing concave function κ : [0, ∞) 7→ [0, ∞) such

that κ(0) = 0 and for all  > 0,
Z 
κ−1 (u)du = ∞,
0

and for every t, and x, y ∈ R,

|b(t, x) − b(t, y)| ≤ κ(|x − y|).

If, moreover, a and b are continuous, the equation (†) has a unique strong solution
(cf. No. 18 and 19). For a proof see Yamada and Watanabe (1971). In Ikeda
and Watanabe (1981) p. 168 a proof in the time-homogeneous case is given, and
in Karatzas and Shreve (1991) p. 291 the result is discussed under the more
restrictive assumption that b is globally Lipschitz in the sense of No. 15 (a).
We remark that pathwise uniqueness still holds without (b) if (a) holds and if
there exists ε > 0 such that |a(t, x)| > ε for all t and x (LeGall (1983)). See also
Perkins (1982).

22. Assume, as in No. 21, that a and b are bounded. Then pathwise uniqueness
holds if a satisfies the following conditions:
(?) there exists ε > 0 such that a(t, x) > ε for all t and x,
(?) there exists a bounded increasing function f such that for all t, x, and y

|a(t, x) − a(t, y)|2 ≤ |f (x) − f (y)|.

See LeGall (1983) or Revuz and Yor (2001) p. 390, where the time-homogeneous
case is discussed.

23. Here we assume that a and b are independent of t. In other words, (†) has the
form

(a) dXt = a(Xt )dWt + b(Xt )dt.

 6. TRANSFORMATION OF MEASURE 49

If b is bounded and measurable and a is Lipschitz continuous such that a(x) > ε
for some ε > 0 and all x ∈ R then (a) has a unique strong solution. This due to
Zvonkin (1974), see also Rogers and Wiiliams (2000) p. 179.

24. If a and b are continuous and bounded in No. 23 (a) and if a2 (x) > ε for all x
for some ε > 0, then (a) has a weak solution which is unique in law (see No. 19
and 20). This solution is non-exploding and, in fact, a linear diffusion (in the
sense of the definition in II.1). The infinitesimal generator is the second order
differential operator
1 d2 d
Gf (x) := a2 (x) f (x) + b(x) f (x)
2 dx2 dx
acting on the domain D := {f : f, Gf ∈ Cb (R)}. Due to the assumptions on a
and b the boundary points ±∞ are natural (cf. II.6). This is an alternative way
to express the fact that the solution is non-exploding.

25. Let a be a measurable function and consider the equation

(a) dXt = a(Xt )dWt .

Every solution of (a) is non-exploding (in the case a ∈ C 1 (R), see McKean
(1969) p. 57, and, in general, Karatzas and Shreve (1991) p. 332). Recall that
a solution X such that P(Xt = ξ for all t) = 1 is called trivial and otherwise
non-trivial. By Engelbert and Schmidt (1985) the SDE (a) has for every initial
distribution µ a non-trivial weak solution if and only if x 7→ a−2 (x) is locally
integrable. Moreover, uniqueness in law holds if and only if a2 (x) > 0 for all x.

26. Assume now that a and b in No. 23 (a) are measurable and satisfy the following
conditions:
(?) a2 (x) > 0 for all x ∈ R,
(?) x 7→ a−2 (x) and x 7→ |b(x)|/a2 (x) are locally integrable.
Then using the method of transformation of drift (cf. No. 27) on the SDE (a)
in No. 25 it can be shown that for every initial distribution µ the equation (a)
in No. 24 has a weak solution defined up to an explosion time and the solution
is unique in law (Engelbert and Schmidt (1984, 1991) or Karatzas and Shreve
(1991) p. 341).

6. The Cameron–Martin–Girsanov transformation of measure

27. We consider here W up to a given time T in its canonical framework. For this let
CT be the space of continuous functions ω : [0, T ] 7→ R and FTc := σ{ω(s); 0 ≤
s ≤ T } the smallest σ-field making all co-ordinate mappings measurable. In-
troduce also the filtration {Ftc }, Ftc := σ{ω(s); s ≤ t}. Let P be a probability
measure on (CT , FTc , {Ftc }) such that the process W defined for 0 ≤ t ≤ T by

Wt (ω) := ω(t), W0 (ω) := ω(0) = 0,

 50 III STOCHASTIC CALCULUS

is a standard Brownian motion. Further, let (CT , FT , {Ft }, P) be the usual

augmentation of (CT , FTc , {Ftc }, P). Then W is a (stopped) Brownian motion
also in this larger space.
A probability measure Q on (CT , FT ) is said to be absolutely continuous with
respect to P if for all A ∈ FT such that P(A) = 0 it is also the case that
Q(A) = 0. We are interested in the structure of a measure Q, when restricted
to Ft , which has this property.
Let µ = {µ(t) : 0 ≤ t ≤ T } be in L2loc and introduce
t
1 t
Z Z 
2
Zt := exp µ(s) dω(s) − µ (s) ds .
0
2 0

Assume that µ is such that {Zt : 0 ≤ t ≤ T } is a martingale. This is the case

if the Novikov condition

1
 Z T 
E exp µ2 (s)ds < +∞
2 0

holds (see Ikeda and Watanabe (1981) p.142, and Liptser and Shiryaev (1977)
Section 6.2 which contains also many illuminating examples).
The Cameron–Martin–Girsanov theorem states that there exists a unique mea-
sure Q on (CT , FT ) such that it is absolutely continuous with respect to P and Zt
is a version of its Radon–Nikodym derivative when the measures are restricted
to Ft , i.e.,
dQ
= Zt P − a.s.
dP Ft
Rt
Further, under Q, the process {ω(t) − 0 µ(s) ds : 0 ≤ t ≤ T } is a (stopped)
standard Brownian motion.
For the proof see Rogers and Williams (2000) p. 82, where uncompleted filtra-
tions are used and the result is formulated in the case T = ∞. In Revuz and
Yor (2001) Chapter VIII one can find a generalization to the semi-martingale
framework.

28. Consider in R the SDEs

dXt = a(t, Xt )dWt + b1 (t, Xt )dt,

dYt = a(t, Yt )dWt + b2 (t, Yt )dt,

where 0 ≤ t ≤ T as in No. 27 and X0 = Y0 = 0. We assume (for simplicity)

that the functions a, b1 , and b2 satisfy the “global” conditions in No. 15. Then
both equations have non-exploding strong unique solutions. Let PX and PY
be the probability measures in (CT , FT ) associated to the solutions X and Y,
respectively.
Assume further that
 6. TRANSFORMATION OF MEASURE 51

(?) a(t, x) 6= 0 for all 0 ≤ t ≤ T and x ∈ R,

the function α(t, ω(t)) := a−1 (t, ω(t)) b1 (t, ω(t)) − b2 (t, ω(t)) is bounded


(?)
PX -a.s. for all t,
(?) {Zt : 0 ≤ t ≤ T } is a PX -martingale, where
 Z t
a−2 (s, ω(s)) b1 (s, ω(s)) − b2 (s, ω(s)) dω(s)


Zt := exp −
0
1 t
Z 
a−2 (s, ω(s))) b21 (s, ω(s)) − b22 (s, ω(s)) ds ,


+
2 0

(?) for i = 1, 2,
Z T 
PX a−2 (s, ω(s)) b2i (s, ω(s)) ds < ∞ = 1.
0

Then for every 0 ≤ t ≤ T the measure PY is absolutely continuous with respect

to PX when restricted to Ft with the Radon–Nikodym derivative Zt :

dPY
= Zt PX − a.s.
dPX Ft

For a proof, see Liptser and Shiryaev (1977) p. 275, where also the derivative of
PX with respect to PY is given. We refer also to Gihman and Skorohod (1979)
p. 261.
Notice that the process Z can also be expressed as

1 t
 Z t Z 
2
Zt = exp − α(s, ω(s)) dβs − α(s, ω(s)) ds ,
0
2 0

where β is a Brownian motion in (CT , FT , PX ).

29. To formulate an important special case of the result in No. 28 let X be a diffusion
on R having the generator
1 d2 d
+ b(x) ,
2 dx2 dx
where b is continuous (but not necessarily bounded). Assume also that both
−∞ and +∞ are not-exit boundaries, i.e., letting m be the speed measure and
s the scale function (see II.9) we have (see II.6) for any z ∈ R,
Z Z
(a) m((z, a))s(da) = m((a, z))s(da) = ∞.
(z,+∞) (−∞,z)

This is the one-dimensional version of Khasminskii’s test for explosion (see

Rogers and Williams (2000) p. 297). Assuming (a) we have for all t ≥ 0,
Z t  Z t 
X 2
P b (ω(s))ds < ∞ = P b2 (ω(s))ds < ∞ = 1,
0 0
52 III STOCHASTIC CALCULUS

where PX and P are the measures associated to X and the Wiener process,
respectively. Using Theorem 7.6 in Liptser and Shiryayev (1977) p. 246 it is
seen that PX is absolutely continuous with respect to P when restricted to Ft
with the Radon–Nikodym derivative

dPX t
1 t
Z Z 
2
= exp b(ω(s)) dω(s) − b (ω(s)) ds P − a.s.
dP Ft 0
2 0
 CHAPTER IV

BROWNIAN MOTION

1. Definition and basic properties

1. Having defined linear diffusions (cf. II.1), standard one-dimensional Brownian

motion W = {Wt : t ≥ 0} can be defined as a linear diffusion on R with the
speed measure m(dx) = 2 dx and the scale function s(x) = x. This is equivalent
to saying that the infinitesimal generator of W is the second order differential
operator
1 d2
Gf (x) := f (x)
2 dx2
acting on the domain D(G) = {f ∈ Cb2 (R) : f 00 ∈ Cb (R)}. For the basic data of
W see A1.1.
From the early days of Brownian motion we mention the papers by Bache-
lier (1900) and Einstein (1905). See Nelson (1967) for a discussion on Robert
Brown’s and Albert Einstein’s works, and Taqqu (2001) for an article on Louis
Bachelier and his times. There is an English translation of Bachelier’s disser-
tation in Cootner (1964). It was, however, Norbert Wiener who in his paper
from 1923 “put the Brownian motion on a firm mathematical foundation” (Itô
and McKean (1974) p. viii). We refer to Doob (1966) and Kac (1966) for arti-
cles concerning Wiener’s contributions to Brownian motion and related topics.
Much of the modern analysis of Brownian motion is based on the ideas of Paul
Lévy; see Loeve (1973) and Schwartz (1988) for memorial articles on Lévy.
There are now several different constructions of Brownian motion, in other
words, proofs for its existence; see, e.g., Freedman (1971), Knight (1981), and
Karatzas and Shreve (1991). In these the starting point is usually the following
definition:
Standard one-dimensional Brownian motion initiated at x on a probability space
(Ω, F, P) is a stochastic process W such that
(a) W0 = x a.s.,
(b) s 7→ Ws is continuous a.s.,
(c) for all 0 = t0 < t1 < · · · < tn the increments

Wtn − Wtn−1 , Wtn−1 − Wtn−2 , . . . , Wt1 − Wt0

are normally distributed with

E(Wti − Wti−1 ) = 0, E(Wti − Wti−1 )2 = ti − ti−1 .

© Springer Basel 2015 53

A.N. Borodin, P. Salminen, Handbook of Brownian Motion - Facts and Formulae,
Probability and Its Applications, DOI 10.1007/978-3-0348-8163-0_4
54 IV BROWNIAN MOTION

Because uncorrelated normally distributed random variables are independent it

follows from (c) that the increments of Brownian motion are independent.
When W is to be defined on a filtered probability space (Ω, F, {Ft }, P) it is
stated that W is Ft -adapted, has the properties (a), (b), and (c), where (c) can
be reformulated in the following compressed form:

2
 
(c’) E(exp(i µ (Wt − Ws ))|Fs ) = exp − (t − s) for all µ ∈ R.
2
In this setting we often say that W is an Ft -Brownian motion. Of course, W is
always an Fto -Brownian motion, where {Fto } is the natural filtration generated
by W. An example (see Yor (1992a) p. 3) of a Brownian motion which is not an
Fto -Brownian motion is given by
t
Ws
Z
W̃t := Wt − ds, t > 0, W̃0 := 0.
0
s

Notice that t 7→ W̃t is continuous (use, e.g., the law of iterated logarithm, see
No. 5). The fact that W̃ is not an Fto -Brownian motion follows from No. 3.
In the sequel we consider only standard Brownian motions, and the word
“standard” will be omitted. It is also assumed, if nothing else is said, that
W is given in its canonical framework (cf. II.1).

2. Here is a list of basic distributional properties of W. Assume that W0 = 0.

Spatial homogeneity. For every x ∈ R the process x + W is a Brownian
motion initiated at x.
Symmetry. −W is a Brownian motion.
c
Reflection principle. Let T be a finite stopping time with respect to {Ft+ }.
Then the process given by

t ≤ T,

Wt ,
Yt :=
2WT − Wt , t ≥ T,

is a Brownian motion.
√
Scaling. For every c > 0 the process { c Wt/c : t ≥ 0} is a Brownian motion.
Time inversion. The process given by

0, t = 0,
Zt :=
t W1/t , t > 0,

is a Brownian motion.
Time reversibility. For a given t > 0,

{Ws : 0 ≤ s ≤ t} ∼ {Wt−s − Wt : 0 ≤ s ≤ t}.

 1. DEFINITION AND BASIC PROPERTIES 55

3. Lévy’s martingale characterization. A continuous real-valued process X in

a filtered probability space (Ω, F, {Ft }, P) is an Ft -Brownian motion if and only
if both X itself and {Xt2 − t : t ≥ 0} are Ft -martingales (Revuz and Yor (2001)
p. 150).

4. Strong law of large numbers.

Wt
lim =0 a.s.
t→∞ t

5. Laws of the iterated logarithm (Revuz and Yor (2001) p. 53).

Wt
lim sup p = 1 a.s.
t↓0 2t ln ln(1=t)

Wt
lim inf = −1 a.s.
2t ln ln(1=t)
p
t↓0

Wt
lim sup √ =1 a.s.
t→∞ 2t ln ln t
Wt
lim inf √ = −1 a.s.
t→∞ 2t ln ln t

6. Local properties of sample paths. For proofs see, e.g., Freedman (1971),
Karatzas and Shreve (1991), Knight (1981), Revuz and Yor (2001).
Continuity. The first property, already stated above as a defining property, is
that Brownian paths t 7→ Wt are continuous a.s.
Hölder continuity. In fact, t 7→ Wt is locally Hölder continuous of order α for
every α < 1/2. In other words, for all T > 0, 0 < α < 1/2 and almost all ω
there exists a constant CT,α (ω) such that for all t, s < T,

|Wt (ω) − Ws (ω)| ≤ CT,α (ω)|t − s|α .

Modulus of continuity. Recall that for a continuous function f : [0, ∞) 7→ R

its modulus of continuity on the interval [0, T ] is a function δ 7→ ε(δ) such that

|f (t2 ) − f (t1 )| < ε(δ) for all t1 , t2 ∈ ∆δ := {(s, t) : 0 < |t − s| < δ, s, t < T }.

For Brownian paths

|Wt2 − Wt1 |
lim sup sup =1 a.s.
2 ln(1=)
p
δ→0 ∆δ

Nowhere differentiability. Brownian paths are a.s. nowhere locally Hölder

continuous of order α > 1/2. In particular, Brownian paths are nowhere differ-
entiable.
56 IV BROWNIAN MOTION

Variation. Brownian paths are of infinite variation on intervals, that is, for
every s ≤ t a.s. X
sup |Wti − Wti−1 | = ∞,

where the supremum is over all subdivisions s ≤ t1 ≤ · · · ≤ tn ≤ t of the interval

(s, t). On the other hand, let ∆n := {t(in) } be a sequence of subdivisions of [s, t]
such that ∆n ⊆ ∆n+1 and

lim max |t(in) − t(i−1

n)
| = 0.
n→∞ i

Then a.s. and in L2 ,

X
lim |Wt(n) − Wt(n) |2 = t − s.
n→∞ i i−1

This fact is expressed by saying that the quadratic variation of W over [s, t] is
t − s.
Local maxima and minima. Recall that for a continuous function f :
[0, ∞) → R a point t is called a point of local (strict) maximum if there exists
ε > 0 such that for all s ∈ (t − ε, t + ε) we have f (s) ≤ f (t) (f (s) < f (t), s 6= t).
A point of local minimum is defined analogously. Then for almost every ω ∈ Ω
the set of points of local maxima for the Brownian path W. (ω) is countable and
dense in [0, ∞), each local maximum is strict, and there is no interval on which
the path is monotone.
Points of increase and decrease. Recall that for a continuous function
f : [0, ∞) → R a point t is called a point of increase if there exists ε > 0 such
that for all s ∈ (0, ε) we have f (t − s) ≤ f (t) ≤ f (t + s). A point of decrease
is defined analogously. Then for almost every ω ∈ Ω the Brownian path W. (ω)
has no points of increase or decrease.
Level sets. For a given ω and x ∈ R let Zx (ω) := {t : Wt (ω) = x}. Then a.s.
the random set Zx (ω) is unbounded and of the Lebesgue measure 0. It is closed
and has no isolated points, i.e., is dense in itself. A set with these properties is
called perfect. The Hausdorff dimension of Zx is 1/2 (Itô and McKean (1974)
p. 50). Its complement Zxc is a countable union of open intervals; so-called
excursion intervals.

2. Brownian local time

7. For t > 0 let Z0 (t) := Z0 ∩ [0, t], where Z0 is the level set of 0. As explained
in No. 6, the Lebesgue measure is not an appropriate tool to measure Z0 (t).
Paul Lévy’s ingenious observation is that there exists, however, a nondecreasing
(random) function determined by Z0 . This function, called the Brownian local
time at 0, is an additive functional of Brownian motion and as such unique
up to a normalizing constant. Every point in Z0 is a (left and/or right) strict
increase point for the local time and it is constant on every open interval in Z0c .
 2. BROWNIAN LOCAL TIME 57

Therefore, the local time at 0 measures the sets Z0 (t). In fact, it is the exact
Hausdorff measure of the level sets (Taylor and Wendel (1966), Perkins (1981)).
To motivate briefly the existence of the local time consider the process

W + = {|Wt | : t ≥ 0}.

This is called reflecting Brownian motion. It is a linear diffusion; see A1.2 for
its basic characteristics. Let

W ◦ = {M̌t − Wt : t ≥ 0},

where M̌t := sup{Ws : s ≤ t}. Then by Lévy (for a proof see Itô and McKean
(1974) p. 41)

(a) W + ∼ W ◦.

Consider now the set Z0◦ := {t : M̌t − Wt = 0}. If there exists a function
determined by Z0◦ with the desired properties, then there exists by (a) also
a function determined by Z0 = {t : |Wt | = 0} having the same properties.
The claim is that M̌ = {M̌t : t ≥ 0} is what we want. Because the asserted
properties of M̌ are obvious the real problem is to construct for every t the value
of M̌t using only the set Z0◦ ∩ [0, t]. For this see Itô and McKean (1974) p. 43,
Balkema and Chung (1991), and Karatzas and Shreve (1991) Section 6.2.

8. Let t 7→ `(t, 0) denote the local time at zero constructed above. It is clear that
we can find in a similar way the local time at an arbitrary point x. Lévy also
proved that a.s.

1 t
Z
(a) `(t, x) = lim 1I(x−ε,x+ε) (Ws )ds.
ε↓0 2" 0

This gives the occupation time formula

Z t Z
1IA (Ws ) ds = `(t, x) dx, A ∈ B(R).
0 A

Further, (t, x) 7→ `(t, x) is continuous a.s. This is due to Trotter (1958) (Itô and
McKean (1974) p. 63, Ikeda and Watanabe (1981) p. 113). Convergence in (a)
holds also in L2 (Chung and Williams (1990) p. 145).
We restate using our new notation the basic distributional property explained
in No. 7,

{(|Wt |, `(t, 0)) : t ≥ 0} ∼ {(M̌t − Wt , M̌t ) : t ≥ 0}.

Notice also that under the usual conditions {`(t, x) : t ≥ 0} can by (a) be taken
to be Ft -adapted. For survey articles on Brownian local time, see McKean
(1975) and Borodin (1989a).
 58 IV BROWNIAN MOTION

9. We describe two additional constructions of `(t, 0) due to Lévy. Assume that

W0 = 0.
(A) For ε > 0 let T (ε, t) be the real number obtained by adding the lengths
smaller than ε of all open intervals in the complement of Z0 (t). Then (Itô and
McKean (1974) p. 43 or Karatzas and Shreve (1991) p. 414)

q
lim T (ε, t) = `(t, 0) a.s.
ε↓0 2"
(B) Introduce for ε > 0 and n = 2, 3, . . . ,
Hε(1) := inf{s : Ws = ε}, H0(1) := inf{s > Hε(1) : Ws = 0},
Hε(n) := inf{s > H0(n−1) : Ws = ε}, H0(n) := inf{s > Hε(n) : Ws = 0}.
Then D(ε, t) := sup{n : H0(n) ≤ t} is called the number of downcrossings of
(0, ε) up to time t and we have (Itô and McKean (1974) p. 48 and p. 222,
Karatzas and Shreve (1991) p. 416; see also Williams (1977) and Walsh (1978))
1
lim ε D(ε, t) = `(t, 0) a.s.
ε↓0 2

10. We conclude this subsection with some miscellaneous properties of `.

Semimartingale property. For every t > 0 the process x 7→ `(t, x) is a
semimartingale, see Perkins (1982).
Scaling. For c > 0,
√ √
{ c `(t/c, x/ c) : 0 ≤ t < ∞, x ∈ R} ∼ {`(t, x) : 0 ≤ t < ∞, x ∈ R}.

Laws of the iterated logarithm.

`(t; 0)
lim sup √ = 1 a.s.
t→∞ 2t ln ln t
sup `(t; y)
lim sup √y∈R = 1 a.s.
t→∞ 2t ln ln t
√
ln ln t
lim inf √ sup `(t, y) = γ a.s.
t→∞ t y∈R

For proofs see Kesten (1965).

√ In Csáki and Földes (1986) the value of the
constant γ is found to be j1 2, where j1 is the smallest positive zero of the
Bessel function J0 .
Modulus of continuity. The exact modulus of continuity uniform in x of
Brownian local time `(t, x) with respect to t is (Perkins (1981)):
|`(t + h; x) − `(t; x)|
lim sup sup = 1 a.s.
2h ln(1=h)
p
h↓0 (t,x)∈[0,T ]×R

The exact modulus of continuity uniform in t on intervals [0, T ] of Brownian

local time `(t, x) with respect to x is (Borodin (1985)):
sup0<t<T |`(t; x + ) − `(t; x)|
lim sup sup = 1 a.s.
δ↓0 x∈R 2((` (T; x) + `(T; x + )) ln(1=))1=2
 3. EXCURSIONS 59

3. Excursions

11. In this section we study excursions of Brownian motion W which start and end
at zero. Our basic reference is Revuz and Yor (2001) Chapter XII. We refer also
to Blumenthal (1992) and Yor (1995).
Let E be the set which consists of continuous functions f : [0, ∞) 7→ R, f (0) = 0,
with the property

∀ f ∈ E ∃ ζ = ζ(f ) < ∞ : f (s) = 0 for all s ≥ ζ(f )

and f (s) 6= 0 for all 0 < s < ζ(f ).

Let δ denote the function which is identically equal to 0. It is assumed that

δ ∈ E. The elements in E are called excursions and the function ζ : E 7→ [0, ∞)
is called the excursion length. Let E be the smallest σ-algebra on E making
all co-ordinate mappings measurable. The measurable space (E, E) is called the
excursion space.
Let ` be the Brownian local time at 0 with the properties given in No. 8 and

%t := inf{s : `(s, 0) > t}

its right continuous inverse. The excursion process Ξ = {Ξs : s ≥ 0} for the
excursions of W from 0 to 0 is a stochastic process taking values in (E, E) and
defined on the sample space of W as follows
(a) if ls := %s − %s− > 0, then Ξs is the mapping r 7→ 1I[0,ls ] (r)W (%s− + r),
(b) if ls = 0 or if the mapping in (a) is not in E, then Ξs (ω) = δ.
Notice that %t is for every t an {Fs }-stopping time, and let F%t be the σ-algebra
as defined in I.5 using %t as T. Then the famous result due to Itô (1970) is that
Ξ is a stationary F%t -Poisson point process. The proof is based on the strong
Markov property, see, e.g., Revuz and Yor (2001) p. 481.

12. A time-homogeneous Poisson point process is completely characterized by its

mean measure. The mean measure of Ξ is called the Itô measure (or character-
istic measure) and we denote it by n. For A ∈ E and s < t let

N (A; s, t) := |{(u, Ξu ) : u ∈ (s, t], Ξu ∈ A}|,

that is, N is the number of points in the set in the braces. Then

E(N (A; s, t)) = (t − s) n(A)

(see Ikeda and Watanabe (1981) p. 43). The Itô measure n is an (infinite) σ-
finite measure. To characterize it let p̂ be the transition density of the diffusion
obtained from W by killing at the first hitting time H0 :
( 1  
(x − y)2  
(x + y)2 
√ exp − − exp − , xy > 0,
p̂(t; x, y) = 2t 2t 2t
0, xy ≤ 0.
 60 IV BROWNIAN MOTION

Recall also that

 2
|x| x
nx (0, t) := Px (H0 ∈ dt)/dt = √ exp − .
2t3 2t
Then for 0 < t1 < · · · < tk , Ai ∈ B(R), i = 1, 2, . . . , k, and

Γ := {f ∈ E : f (t1 ) ∈ A1 , . . . , f (tk ) ∈ Ak } ∈ E

we have
Z Z
n(Γ) = dx1 · · · dxk nx1 (0, t1 ) p̂(t2 − t1 ; x1 , x2 )
(a) A1 Ak
· . . . · p̂(tk − tk−1 ; xk−1 , xk ),

see Revuz and Yor (2001) p. 494.

13. We remark also that from a realization of Ξ we can construct the corresponding
realization of W. This is achieved by first setting
X X
%t := ζ(Ξs ), %t− := ζ(Ξs ).
s≤t s<t

Next let Ξs (r) be the value of Ξs at time 0 ≤ r < ∞. Then for t ≥ 0

Wt := Ξs (t − %s− ), %s− ≤ t ≤ %s ,

defines a Brownian motion and the continuous inverse of % gives its local time
at 0 (see Revuz and Yor (2001) p. 482).

14. In No. 15 and 16 below two explicit characterizations of n are discussed in more
detail. These are called Itô’s and Williams’ descriptions, respectively. There
is also a description due to Bismut (1985); this is presented in Revuz and Yor
(2001) p. 502. We also refer to Biane and Yor (1987) for further developments
and as a source for many formulae associated to the excursion theory of Brownian
motion.
It is convenient to study separately positive and negative excursions. Therefore
introduce
E + := {f ∈ E : f (s) ≥ 0 for all s} ∪ {δ},
E − := {f ∈ E : f (s) ≤ 0 for all s} ∪ {δ},
and let n+ := n(· ∩ E + ) and n− := n(· ∩ E − ) be the restrictions of n on E +
and E − , respectively.

15. Here we take a closer look at the Itô measure n from the point of view of the
lengths of the excursions. We call this Itô’s description. It is also in the spirit
of Lévy (1948), pp. 225–237, see also Itô and McKean (1974) pp. 75–79 and
Revuz and Yor (2001).
 3. EXCURSIONS 61

In (a) in No. 12 let n = 1, A1 = R and integrate over x1 . This gives

2
Z
n(ζ > t) = nx (0, t) dx = √ .
R 2t

Because excursions take place during the constant periods of ` this formula is also
obtained directly from the Lévy measure in the Lévy–Khintchine representation
of %t (cf. II.14; notice, however, that the local time is here taken with respect
to the Lebesgue measure):
∞
1
 Z 
E0 (exp(−λ%t )) = exp − t (1 − e−λl ) √ 3 dl
0 2l
√
= exp(−t 2λ).

Due to the spatial symmetry of W we have

1
(a) n+ (ζ > t) = n− (ζ > t) = √ .
2t

To proceed we need the notion of excursion bridge. Let Ŵ be a Brownian

motion on (0, ∞) killed when it hits 0. Let Py,l,0 be the probability measure
associated to Ŵ when started at y and conditioned (in the sense of Doob’s h-
transforms) to hit 0 at time l. In other words with canonical notations, letting
for 0 < t1 < · · · < tk < l and Ai ∈ B((0, ∞)),

Γl := {ω : ωt1 ∈ A1 , . . . , ωtk ∈ Ak },

we have
Z Z
Py,l,0 (Γl ) = dx1 · · · dxk p̂(t1 ; y, x1 )
A1 Ak
nxk (0; l − tk )
· . . . · p̂(tk − tk−1 ; xk−1 , xk ) .
ny (0; l)

The process governed by Py,l,0 is a non-time-homogeneous strong Markov pro-

cess. The family of measures {Py,l,0 : y > 0} can be extended by defining

P+
0,l,0 := lim Py,l,0 .
y↓0

The limit is in the sense of the convergence of finite dimensional distributions.

The process governed by P+ 0,l,0 is called the positive excursion bridge of length l
for the excursions from 0. It has the finite dimensional distributions
Z Z
P+0,l,0 (Γl ) = dx1 · · · dxk nx1 (0, t1 ) p̂(t2 − t1 ; x1 , x2 )
A1 Ak
nxk (0; l − tk )
· . . . · p̂(tk − tk−1 ; xk−1 , xk )
n0+ (l)
 62 IV BROWNIAN MOTION

with
1
n0+ (l) := n+ (ζ ∈ dl)/dl = √ .
2 2l3
Then for Γ+ := {f ∈ E + : f (t1 ) ∈ A1 , . . . , f (tk ) ∈ Ak } Itô’s description of n+
is
Z ∞
n+ (Γ+ ) = n+ (Γ+ ; ζ ∈ dl)
tk
Z ∞
= P+ + +
0,l,0 (Γ ) n (ζ ∈ dl).
tk

An analogous description, with the obvious modifications, is valid for n− .

Now let R(3) , R0(3) = 0, be a 3-dimensional Bessel process, Q0 the probability
measure associated to it, and r its transition density with respect to the speed
measure (see A1.22). Introduce the function
( r(l − t; z; 0)
, t < l,
h(z, t; l, 0) := r(l; 0; 0)
0, t ≥ l.
Let Q0,l,0 denote the law of the h-transform of R(3) . This is constructed similarly
to the Brownian bridge in No. 23 and called the 3-dimensional Bessel bridge of
length l from 0 to 0. It is a straightforward computation to check that the laws
Q0,l,0 and P+0,l,0 are the same.

16. We turn now to Williams’ description of n (see Williams (1979) p. 98). For
f ∈ E + let
m(f ) := sup{f (t) : t ≥ 0}.
The starting point in Williams’ approach is the n+ - distribution of m = m(f )
1
(a) n+ (m > z) = .
2z
For proofs see Rogers and Williams (1987) p. 425 or Revuz and Yor (2001)
p. 492. Still another way to prove this is to compute the distribution of the
maximum of a 3-dimensional Bessel bridge and to use Itô’s description of n+ .
Indeed, using absolute continuity and the strong Markov property it is seen that
√
2l3 l
Z
(b), Q0,l,0 (M̌ > z) = r0 (z, t) nz (0, l − t) dt,
z 0

where M̌ is the maximum and

r0 (z, t) := Q0 (Hz ∈ dt)/dt.
Putting this into Itô’s description and integrating over l gives the desired for-
mula. We remark that in Chung (1976) (see also Kennedy (1976), and Durrett
and Iglehart (1977)) it is proved that
∞ 
X 4k2 z 2  
2k2 z 2 
Q0,l,0 (M̌ > z) = 2 − 1 exp − .
l l
k=1
 3. EXCURSIONS 63

For further results on the maximum of a Bessel bridge see Pitman and Yor
(1999b).
To proceed with Williams’ description we notice that Ŵ on (0, ∞) conditioned
in the sense of Doob’s h-transforms to drift to ∞ is identical in law to a 3-
dimensional Bessel process. The excessive function used is simply the scale
function s(x) = x of W. This observation makes the appearance of 3-dimensional
Bessel processes in Williams’ description plausible. Now let R(3,1) and R(3,2) be
two independent 3-dimensional Bessel processes starting from 0, and for a given
z > 0 let
Hz(i) := inf{t : Rt(3,i) = z}, i = 1, 2.
Introduce the process
 (3,1)
R , 0 ≤ t ≤ Hz(1) ,
 t

(3,2)
Zt(z) := z − Rt−H (1) , Hz ≤ t ≤ Hz + Hz ,
(1) (1) (2)

 z

Hz(1) + Hz(2) ≤ t,

0,

and let K(z, Γ+ ), Γ+ ∈ E + , be the kernel (see I.6) associated to Z (z) . Then
Williams’ description for positive excursions is
Z ∞
n+ (Γ+ ) = K(z, Γ+ ) n+ (m ∈ dz).
0

An analogous description is valid for negative excursions.

17. Here we give a few additional formulae derived from No. 15 (a) and 16 (a),(b).
First,
Z l
dl
n+ (m > z, ζ ∈ dl) = r (z, t) nz (0, l − t) dt.
2z 0 0
Next recall that the Laplace transform of the Q0 -distribution of the first hitting
time for a 3-dimensional Bessel process is given by (see formula 5.2.0.1)
√
2 x
Q0 (exp(−λHx )) = √ .
sh( 2 x)
From Williams’ description we obtain
√

2 z  2 1
n+ (e−λζ ; m ∈ dz) = √ dz,
sh( 2 z ) 2z 2
and, consequently,
l
dl dz
Z
+
n (m ∈ dz, ζ ∈ dl) = 2 r0 (z, t) r0 (z, l − t) dt.
2z 0

18. Excursions straddling a fixed time. Let t > 0 be given, and introduce the
random variables

Gt := sup{s < t : Ws = 0}, Dt := inf{s > t : Ws = 0}.

64 IV BROWNIAN MOTION

Then
Ws+Gt , 0 ≤ s < Dt − Gt ,

Us :=
∂, s ≥ Dt − Gt ,
defines a process U called the Brownian excursion process straddling t.
We characterize the law of U under the measure P0 . For proofs see Chung
(1976), and, in a more general framework, Getoor (1979b) and Getoor and
Sharpe (1982).
To start with, Gt has the arcsine distribution (cf. II.20), that is
du 2 u
q
P0 (Gt ∈ du) = ( or P0 (Gt ≤ u) = arcsin ).
 u(t − u)  t
p

Using the Markov property it is seen that for v > t,

Z
P0 (Dt ∈ dv) = P0 (Wt ∈ dy)Py (H0 ∈ dv − t)
R
1 t
r
= dv.
v v − t

The joint distribution is

du dv
P0 (Gt ∈ du, Dt ∈ dv) = , 0 < u < t < v,
2 u (v − u)3
p

giving
du dl
P0 (Gt ∈ du, Dt − Gt ∈ dl) = √ , (t − l) ∨ 0 < u < t,
2 u l3
and √
t− (t − l ) ∨ 0
p
P0 (Dt − Gt ∈ dl) = √ dl.
 l3
Recall the definitions of p̂(t; x, y) and nx (0, t) in No. 12. Then for 0 < u1 <
· · · < uk < l, , u < t, u+l > t, and y1 , y2 , . . . , yk either all positive or all negative
we have

P0 (Gt ∈ du, Uu1 ∈ dy1 , Uu2 ∈ dy2 , . . . , Uuk ∈ dyk , Dt − Gt ∈ dl)

1
= √ du ny1 (0, u1 ) dy1 p̂(u2 − u1 ; y1 , y2 ) dy2
2u
· . . . · p̂(uk − uk−1 ; yk−1 , yk ) dyk nyk (0, l − un ) dl.

This leads to

P0 (Uu1 ∈ dy1 , Uu2 ∈ dy2 , . . . , Uuk ∈ dyk | Gt = u, Dt − Gt = l)

√
= 2πl3 ny1 (0, u1 ) dy1 p̂(u2 − u1 ; y1 , y2 ) dy2
· . . . · p̂(uk − uk−1 ; yk−1 , yk ) dyk nyk (0, l − uk )
1
= Q0,l,0 (ω(u1 ) ∈ dy1 , ω(u2 ) ∈ dy2 , . . . , ω(uk ) ∈ dyk ),
2
 3. EXCURSIONS 65

where Q0,l,0 is the law (in the canonical setting) of the 3-dimensional Bessel
bridge as introduced in No. 15.

19. Here we take a closer look at W prior to time t given that Gt = u.

First, given Gt = u the process {Ws : 0 ≤ s ≤ Gt } is independent of U and has
the law of Brownian bridge from 0 to 0 with length u (Revuz and Yor (2001) p.
492).
Second, let
Ws+Gt , 0 ≤ s < t − Gt ,

Vs :=
∂, s ≥ t − Gt .
This defines a process V called the Brownian meandering process ending at t.
To analyze V we have for 0 < u1 < · · · < uk < t − u the joint distributions

P0 (Gt ∈ du, Vu1 ∈ dy1 , Vu2 ∈ dy2 , . . . , Vuk ∈ dyk , Wt ∈ dy)

1
= √ du ny1 (0, u1 ) dy1 p̂(u2 − u1 ; y1 , y2 ) dy2
2u
· . . . · p̂(uk − uk−1 ; yk−1 , yk ) dyk p̂(t − u − uk ; yk , y) dy.

In particular,

|y | y2
 
P0 (Gt ∈ du, Wt ∈ dy) = exp − du dy, 0 < u < t,
2 u(t − u)3 2(t − u)
p

leading for y > 0 to the conditional law

P0 (Vu1 ∈ dy1 , Vu2 ∈ dy2 , . . . , Vuk ∈ dyk | Gt = u, Wt = y)

= Q0,t−u,y ω(u1 ) ∈ dy1 , ω(u2 ) ∈ dy2 , . . . , ω(uk ) ∈ dyk ,

where Q0,l,y is the measure (in the canonical setting) associated to the 3-dimen-
sional Bessel bridge from 0 to y having length l. By symmetry, this also charac-
terizes the case y < 0. Further, we have the following absolute continuity relation
(see Imhof (1984))

P0 (Vu1 ∈ dy1 , Vu2 ∈ dy2 , . . . , Vuk ∈ dyk , Wt ∈ dy | Gt = u)

(a) c

= Q0 ω(u1 ) ∈ dy1 , ω(u2 ) ∈ dy2 , . . . , ω(uk ) ∈ dyk , ω(t − u) ∈ dy ,
y

where Q0 is the measure (in the canonical setting) associated to the 3-dimen-
sional Bessel process starting at 0 and c = (π(t − u)/2)1/2 .
A Brownian meandering process ending at t is often taken to be non-negative
and normalized to have length 1 and, hence, defined as
n
1 o
√ |WGt +u(t−Gt ) | : 0 ≤ u ≤ 1 ,
t − Gt
 66 IV BROWNIAN MOTION

and called simply Brownian meander (see Yor (1992a) p. 41). The absolute
continuity property displayed in (a) takes in this setting the form
dQm  1
q
= Q0 -a.s.,
dQ0 Fv+c 2 !(1)

where v ≤ 1, Qm is the measure associated to the Brownian meander, and {Fvc }

is the canonical filtration in the space of continuous functions.
For results concerning Brownian meander see, e.g., Chung (1976, 1982a), Imhof
(1984), Biane and Yor (1987), (1988), Yor (1992a), Bertoin and Pitman (1994),
and Pitman and Yor (1996).

4. Brownian bridge

20. Let x, y ∈ R and l > 0 be given. A continuous Gaussian process X = {Xt : 0 ≤

t ≤ l}, X0 = x, such that
t st
E Xt = x + (y − x) , Cov(Xt , Xs ) = (s ∧ t) −
l l
is called a Brownian bridge from x to y of length l.
To show that there indeed exists a continuous modification of a Gaussian process
having the stated covariance Kolmogorov’s criterion in I.2 can be used (see also
Gihman and Skorohod (1974) p. 193).
We indicate the parameters of X by writing X x,l,y . The law is denoted by Px,l,y .
Observe that EXl = y and Cov(Xt , Xs ) = 0 if s = l or t = l. Therefore, a.s.
Xl = y. Because of this property X is called a Brownian bridge. Other names
are pinned Brownian motion and tied-down Brownian motion.

21. Let W be a standard Brownian motion started at 0. Then we have the following
two descriptions of a Brownian bridge:
t
X x,l,x ∼ x + Wt − Wl : 0 ≤ t ≤ l ,

(a)
l

l−t lt
n   o
(b) X x,l,x ∼ x+ W : 0≤t≤l .
l l−t
In (b) the process in braces is defined to be equal to x when t = l.
Moreover, X x,l,y can be viewed as X x,l,x with a drift. The exact statement is
t
X x,l,y ∼ Xtx,l,x + (y − x) : 0 ≤ t ≤ l .

l

22. A Brownian bridge of length l from x to y can also be characterized as the

pathwise unique solution of the SDE (Ikeda and Watanabe (1981) p. 229)
( y − Xt
dXt = dt + dWt , 0 ≤ t < l,
l−t
X0 = x.
 4. BROWNIAN BRIDGE 67

The solution is
t
t dWs
Z
Xt = x + (y − x) + (l − t) .
l 0
l−s

We refer to Yor (1997a) p. 34 for an approach to the Brownian bridge via

enlargement of filtrations, see also Jeulin and Yor (1979), and Jeulin (1980) p.
119.

23. One more approach to the Brownian bridge is via Doob’s h-transform. To ex-
plain this let p be the transition density of W and notice that the function

p(l − t; x, y), 0 ≤ t < l,


(x, t) 7→ h(x, t; y, l) =
0, t ≥ l,

is for every y ∈ R and l > 0 space-time excessive for W. Indeed, letting x̄ = (s, x)
be the starting state,

h(x, s; y, l), 0 ≤ s ≤ t < l,


Ex̄ (h(Wt , t; y, l)) =
0, otherwise.

Using the function h we build the h-transform of W. This is a non-time-homo-

geneous Markov process W h = {Wth : 0 ≤ t < l} having the finite dimensional
distributions

Phx (Wth1 ∈ dx1 , Wth2 ∈ dx2 , . . . , Wthn ∈ dxn )

n
Y h(xn ; tn ; y; l)
= p(ti − ti−1 ; xi−1 , xi ) ,
h(x; 0; y; l)
i=1

where 0 = t0 < t1 < · · · < tn < l, x0 := x and xi ∈ R. It is a straightforward

computation to check that the measures Phx and Px,l,y are the same.
Intuitively, a Brownian bridge X x,l,y is a Brownian motion W started at x and
conditioned to be at y at time l. The h-transform techniques makes this kind of
conditioning with respect to a set of probability 0 plausible.
From the h-transform approach we also get the absolute continuity result

dPx,l,y h(!(t); t; y; l)
= Px -a.s.,
dPx c
Ft+ h(x; 0; y; l)

where t < l and {Ftc } is the canonical filtration in the space of continuous
functions.

24. Vervaat’s path transformation. Let X denote X 0,1,0 , and introduce

M̂ := inf{Xt : 0 ≤ t ≤ 1} and Ĥ := inf{t : Xt = M̂ }.

 68 IV BROWNIAN MOTION

Then Ĥ ∼ U (0, 1), that is, Ĥ is uniformly distributed on (0, 1). Define next
the process Y by
(
XĤ+t − M̂ , 0 ≤ t ≤ 1 − Ĥ,
Yt :=
Xt−(1−Ĥ) − M̂ , 1 − Ĥ ≤ t ≤ 1.

Let R denote the 3-dimensional Bessel bridge from 0 to 0 of length 1 (see No.
15). A result due to Vervaat (1979) is that Ĥ and Y are independent and

Y ∼ R.

Conversely, let ξ ∼ U (0, 1) and independent of R, and set

Rξ+t − Rξ , 0 ≤ t ≤ 1 − ξ,

Zt :=
Rt−(1−ξ) − Rξ , 1 − ξ ≤ t ≤ 1.

Then (Biane (1986))

Z ∼ X.
An immediate implication of Vervaat’s result is that the range of X and the
maximum of R have the same laws. For the law of the maximum see No. 16,
and for the distribution of the range the formula 1.1.15.8 (2), see also Bass and
Koshnevisan (1995).
Verwaat’s path transformation can be used (see Chaumont, Hobson, and Yor
(2001)) to explain the following puzzling identity presented in Donati-Martin,
Matsumoto and Yor (2002)
Z 1 −1 
E exp(α Xu ) du = 1 for all α ∈ R,
0

see also Donati-Martin, Matsumoto and Yor (2000b).

25. Let X and R be as in No. 24, and let `R (x) be the total local time at x of R
such that Z 1 Z
1IB (Rs ) ds = `R (x) 2dx, B ∈ B(R).
0 B

Let, further, Z 1
AR
x := 1I(−∞,x) (Rs ) ds
0

be the time R spends below the level x and introduce for 0 ≤ t ≤ 1 its inverse

αtR := inf{x : AR
x ≥ t}.

Then (Jeulin (1985) p. 264)

{`R (αtR ) : 0 ≤ t ≤ 1} ∼ R.
 4. BROWNIAN BRIDGE 69

In particular,

sup{`R (x) : x ≥ 0} ∼ sup{Rt : 0 ≤ t ≤ 1}.

Now let `X (x) be the total local time at x of X taken with respect to 2dx. As
above, let AX X
x be the time X spends below the level x and αt its inverse. Then
the result is also as above (Biane (1986)), that is,

{`X (αtX ) : 0 ≤ t ≤ 1} ∼ R.

In particular,

sup{`X (x) : x ∈ R} ∼ sup{Rt : 0 ≤ t ≤ 1}.

For further results concerning the Brownian bridge and related notions see Biane
and Yor (1987), Pitman and Yor (1996), and Pitman (1999).

26. We have not explicitly displayed formulae for Brownian bridges in Part II of the
book. Instead, we give the joint distributions of functionals up to time t and
the location of W at time t. The distributions of the corresponding functionals
of a Brownian bridge can easily be obtained from these. For example, by the
formula 1.1.1.8 from Part II for x, z ≤ b,
 
Px sup Wt < b, Wl ∈ dz
0≤t≤l
1   (z − x)2  
(z + x − 2b)2 
= √ exp − − exp − dz.
2 l 2l 2l

Therefore, dividing above by Px (Wt ∈ dz) gives

  
(z + x − 2b)2 (z − x)2 
Px,l,z sup Xtx,l,z < b = 1 − exp − + .
0≤t≤l 2l 2l

Another approach is to use absolute continuity, namely,

p(l − t; b; z )
  Z l  
Px,l,z sup Xtx,l,z >b = Ex ; Hb ∈ dt
0≤t≤l 0
p(l; x; z )
1
Z l
= p(l − t; b, z) Px (Hb ∈ dt).
p(l; x; z ) 0

The integral can be computed, e.g., by taking the Laplace transform.

 70 IV BROWNIAN MOTION

5. Brownian motion with drift

27. Let µ ∈ R be given. A linear diffusion on R having the operator

1 d2 d
G= +µ
2 dx2 dx
as the infinitesimal generator acting on the domain

D(G) = {f : f, Gf ∈ Cb (R)}

is called a Brownian motion with drift µ. We let W (µ) = {Wt(µ) : t ≥ 0} denote

this process. For the basic data of W (µ) see A1.14.

28. Brownian motion with drift can be constructed from W by adding a drift term

W (µ) ∼ {Wt + µt : t ≥ 0}.

It can also be obtained as an h-transform of W (in space-time). Indeed, the

function
2 t
 
(x, t) 7→ h(x, t; µ) := exp µx −
2
is space-time invariant for W, i.e.,

Es,x (h(Wt , t; µ)) = h(x, s; µ).

This is seen immediately from (c’) in No. 1. Using the function h we build the
h-transform of W. It turns out that this is a time-homogeneous Markov process
identical in law to W (µ) .
Notice also that if W is an Ft -Brownian motion, then the process

{h(Wt , t; µ) : t ≥ 0}

is a (P, Ft )-martingale.

29. The following properties of W (µ) can be deduced from the corresponding prop-
erties of W :
Independent increments. For all 0 = t0 < t1 < · · · < tn the increments

Wt(nµ) − Wt(n−1
µ) µ)
, Wt(n−1 µ)
− Wt(n−2 , . . . , Wt(1µ) − Wt(0µ)

are independent, and normally distributed with

E(Wt(iµ) − Wt(i−1
µ)
) = µ(ti − ti−1 ), Var(Wt(iµ) − Wt(i−1
µ)
) = ti − ti−1 .

Spatial homogeneity. Let x ∈ R be given and assume that W0(µ) = 0. Then

x + W (µ) is a Brownian motion with drift µ started at x.
 5. BROWNIAN MOTION WITH DRIFT 71

Symmetry. −W (µ) is a Brownian motion with drift −µ.

√ √
( cµ)
Scaling. For every c > 0 the process { c Wt/c : t ≥ 0} started at 0 is
(µ)
identical in law to W started at 0.
Time inversion. The process defined by
(
µ, t = 0,
Zt := (µ)
tW1/t , t > 0,

is a Brownian motion started at µ.

Time reversibility. For a given t > 0,
(−µ)
{Ws(µ) − W0(µ) : 0 ≤ s ≤ t} ∼ {Wt−s − Wt(−µ) : 0 ≤ s ≤ t}.

Strong law of large numbers.

Wt()
lim =µ a.s.
t→∞ t

Laws of the iterated logarithm.

Wt()
lim sup p =1 a.s.
t↓0 2t ln ln(1=t)

Wt()
lim inf = −1 a.s.
2t ln ln(1=t)
p
t↓0

All local properties of the sample paths as presented in No. 6 hold except that
the level sets Zx (ω) := {t : Wt(µ) (ω) = x} are bounded a.s. when µ 6= 0.

30. The h-transform approach leads to the following absolute continuity result (cf.
the Cameron–Martin–Girsanov transformation of measure in III.26–28):

dP(xµ) h(!(t); t; ) 2 t
 
= = exp µ(ω(t) − x) − Px -a.s.,
dPx c
Ft+ h(x; 0; ) 2

where P(xµ) and Px are the measures associated to W (µ) and W, respectively.

31. Consider the last exit time λy := sup{t : Wt(µ) = y}. By the general results in
II.20 (for the Green function G0 and the transition density p, see A1.14),

p(t; x; y)
Px (λy ∈ dt) = dt
G0 (y; y)
| | 2 t (x − y)2
 
=√ exp µ(y − x) − − dt,
2 t 2 2t
 72 IV BROWNIAN MOTION

and

1 − e−2µ (x−y) , µ(x − y) ≥ 0,


Px (λy = 0) = Px (Hy = ∞) =
0, µ(x − y) ≤ 0.

32. Suppose that µ > 0 and W (µ) is started at 0. Let

M̂ := inf{Wt(µ) : t ≥ 0} and Ĥ := inf{t : Wt(µ) = M̂ }.

We find here the joint distribution of M̂ and Ĥ. Let

M̂t := inf{Ws(µ) : 0 ≤ s ≤ t} and Ĥ(t) := inf{s ≤ t : Ws(µ) = M̂t }.

Then M̂t → M̂ and Ĥt → Ĥ a.s. as t → ∞. Using II.19 it is seen that for
α, β, γ > 0,
Z ∞
e−γ t E0 (exp(αM̂t − β Ĥt )) dt
0
∞
' (y)
Z 0 Z
αz
= s(dz) e m(dy)
−∞ z
' + (z ) ' (z )
1 2 + 2 + 
p
= =: F (γ).
2( + ) + 2 + + 
p

Consequently,

E0 (exp(αM̂ − β Ĥ)) = lim E0 (exp(αM̂t − β Ĥt ))

t→∞
2
= lim γ F (γ) = .
2 + 2 + + 
p
γ→0

Because M̂ ∼ Exp(2 µ) on (−∞, 0) it is seen that Ĥ given M̂ = z is distributed

as Hz for the process W (−µ) .

33. The result in No. 32 can be expanded considerably. This leads to the path
decomposition of W (µ) due to David Williams (see Rogers and Williams (2000)
p. 436).
First, we construct from W (µ) a new process living on (0, ∞) by conditioning
W (µ) to not hit 0. For this let

h(x) := Px (H0 = ∞) = 1 − e−2µ x , x > 0.

Then the desired conditioning is achieved by building the excessive transform

of W (µ) using the function h. The transform is a time-homogeneous diffusion on
(0, ∞) with the generator

1 d2 d
G+ = + µ cth(µ x) .
2 dx2 dx
 5. BROWNIAN MOTION WITH DRIFT 73

Straightforward computations show that for this process 0 is an entrance-not-

exit boundary.
Let W + denote the diffusion having the generator G + and which is started at
0. Further, let S ∼ Exp(2 µ) on (−∞, 0). Assume that W + , S and W (−µ)
are defined on the same probability space and that they are independent. Let
HS := inf{t; Wt(−µ) = S} and define

Wt(−µ) , 0 ≤ t < HS ,

Xt(µ) := +
S+ Wt−HS
, t ≥ HS .

Then, if X0(µ) = W0(µ) = 0, the Williams path decomposition theorem says that

X (µ) ∼ W (µ) .

This result can be viewed as a canonical example of the various path decomposi-
tion theorems. It is possible to formulate similar explicit results for all transient
diffusions. We refer to Rogers and Williams (2000) for references, and, espe-
cially, for the path decomposition of killed Brownian motion. See also Williams
(1974), Biane (1993), and Pitman and Yor (1996).

34. Consider the perpetual integral functional

Z ∞
I∞ (f ) := f (Wt(µ) ) dt,
0

where f is a non-negative, locally bounded, and Borel measurable function and

µ > 0. It holds that
Z ∞
(a) I∞ (f ) < ∞ a.s. ⇔ f (x)dx < ∞,

R∞
(b) f (x) dx < ∞ and ∃m > 0 f (x) = O(|x|m )x→−∞

⇒ ∀n ∈ N, ∀x Ex I∞ (f )n < ∞,

R∞
(c) f (x) dx < ∞ and f (x) = O(1)x→−∞

⇒ ∃γ > 0, ∀x Ex exp(γI∞ (f ))) < ∞,

Z ∞
(d) sup Ex I∞ (f )) < ∞ ⇔ f (x) dx < ∞.
x −∞

For proofs based on Ray-Knight theorems and Jeulin’s lemma, see Salminen and
Yor (2005a); for (a), see also Engelbert and Senf (1991).
 74 IV BROWNIAN MOTION

35. It is interesting that perpetual integral functionals of W (µ) with µ > 0 can be
viewed as first hitting times for a properly scaled and time changed diffusions.
To explain this, let f be a monotone, twice continuously differentiable function
such that r := limx→∞ f (x) exists and introduce the additive functional
Z s
Js := (f 0 (Wu(µ) ))2 du.
0

Assume further that f 0 (x) 6= 0 and

Z ∞
(f 0 (x))2 dx < ∞.

Let Z be a diffusion given by

Zt := f (Wα(µt ) ),

for t ≥ 0 such that

αt := inf{s : Js > t} < ∞.
Then a.s. Z ∞
(f 0 (Ws(µ) ))2 ds = inf{t : Zt = r}.
0

Moreover, Z is a solution of the SDE

dZt = dβt + G ◦ f −1 (Zt ) dt, Z0 = f (0),

where β is a Brownian motion and

1 1 
G(x) := f 00 (x) + µ f 0 (x) .
(f 0 (x))2 2

For a proof, see Salminen and Yor (2005b). In Khosnevisan et al. (2006) the
result is generalized for fairly general diffusions. In Salminen and Wallin (2007)
numerical metods for calculating the distributions of perpetual functionals are
studied. For other related works, see Salminen and Yor (2003), Decamps et al.
(2005), and Borodin and Salminen (2006). These papers contain also many ex-
amples, in particular Salminen and Yor (2005b). A famous example – Dufresne’s
identity – is presented in No. 50.

6. Bessel processes

36. Let {W (i) : i = 1, 2, . . . , n} be a family of independent Brownian motions. The

process R(n) defined by
q
Rt(n) := (Wt(1) )2 + (Wt(2) )2 + · · · + (Wt(n) )2 , t ≥ 0,
 6. BESSEL PROCESSES 75

is called an n-dimensional Bessel process or a Bessel process of order ν := n/2−1.

Clearly, R(n) is the radial part of an n-dimensional Brownian motion. Special
properties of Brownian motion imply (Itô and McKean (1974) p. 60) that R(n)
is a linear diffusion. Its generator is

1 d2 n−1 d
G (n) = +
2 dx2 2x dx
with the domain D(n) = {f : f, G (n) f ∈ Cb [0, ∞)} for n ≥ 2 and D(1) = {f :
f ∈ Cb2 [0, ∞), f 0 (0+) = 0}.

37. Notice that R(1) is identical in law to a reflected Brownian motion and 0 is in this
case a non-singular boundary point. For n ≥ 2 the boundary point 0 is entrance-
not-exit. This means that the n-dimensional Brownian motion (for n ≥ 2)
started from the origin does not hit the origin after time 0. However, because
R(2) is recurrent the 2-dimensional Brownian motion when started outside the
origin hits every ball with center at the origin. For n > 2 the process R(n) is
transient and drifts to ∞. It follows that n-dimensional Brownian motion (for
n > 2) is transient.

38. To justify the name “Bessel process” we remark that the fundamental solutions
ϕα and ψα (see II.10 and 11) are given in terms of the modified Bessel functions
I and K (for these and other basic characteristics of R(n) see A1.21). In fact, if
the ODE
G (n) u = αu
is multiplied by x2 we obtain a form of the (modified) Bessel differential equation.
See Abramowitz and Stegun (1972) Section 9.6 for details concerning Bessel’s
ODE and its solutions. See also A2.4 and A4.10, A4.11.

39. The property of n-dimensional Brownian motion that its radial part is a linear
diffusion is a special one. We take a closer look at this in general.
First, notice that if X is a linear diffusion on R such that X ∼ −X, then X 2
is also a diffusion. In particular, W 2 is a diffusion with the generator

d2 d
2x + .
dx2 dx

Secondly, we remark that the requirement that the sum of two independent non-
negative diffusions X1 and X2 is also a diffusion is very exclusive. It has been
proved by Shiga and Watanabe (1973) that this is the case if and only if the
generators of X1 and X2 are of the corresponding forms

d2 d
ax + (bx + cαi ) , i = 1, 2,
dx2 dx
where a > 0, b ∈ R, c ≥ 0 and αi ≥ 0, i = 1, 2. The boundary point 0 is
(?) exit-not-entrance if cαi = 0,
 76 IV BROWNIAN MOTION

(?) non-singular if 0 < cαi < a,

(?) entrance-not-exit if cαi ≥ a.
In the non-singular case we let the boundary condition be reflection. The gen-
erator of X1 + X2 is, then
d2 d
(a) ax + (bx + c(α1 + α2 )) .
dx2 dx
√
√ if X is a non-negative diffusion, then X is also a diffusion because
Thirdly,
x 7→ x is increasing.
√ Moreover, if the generator of X is of the form in (a), then
the generator of X is
a d2 bx 2c( 1 + 2 ) − a d
 
(b) + + .
4 dx2 2 4x dx
Diffusions associated to the generator in (b) can be viewed as radial Ornstein–
Uhlenbeck processes; see A1.25 for their basic characteristics. Diffusions associ-
ated to the generator in (a) – squared radial Ornstein–Uhlenbeck processes, see
A1.26 – play a prominent role in the Ray–Knight theorems; see Chapter 5. In
mathematical finance they are often called CIR-processes, due to the work of
Cox, Ingersoll and Ross (1985).

40. Let M̌t := sup{Ws : s ≤ t}, where W is initiated at 0. Then, see No. 7, a result
due to Lévy is that

R(1) ∼ |W | ∼ {M̌t − Wt : t ≥ 0}.

In Pitman (1975) (see also Williams (1979) p. 96) a similar construction is given
for R(3) :
R(3) ∼ {2M̌t − Wt : t ≥ 0}.
The filtration generated by {M̌t −Wt : t ≥ 0} is equal to the Brownian filtration,
but the filtration generated by {2M̌t − Wt : t ≥ 0} is strictly smaller than the
Brownian filtration (see Revuz and Yor (2001) p. 253). We remark also that if
c 6∈ {0, 1, 2}, then the process {cM̌t − Wt : t ≥ 0} is not Markovian.
There are extensions of these results when W is replaced with W (µ) . We have
that
{M̌t(µ) − Wt(µ) : t ≥ 0}
is identical in law with a reflecting Brownian motion on R+ with drift −µ (see,
e.g., Harrison (1985)). For the basic characteristics of this process see A1.16.
Further (see Rogers and Pitman (1981)), the process

{2M̌t(µ) − Wt(µ) : t ≥ 0}

is a linear diffusion with the generator (cf. No. 33)

1 d2 d
+ |µ| cth(|µ| x) .
2 dx2 dx
 6. BESSEL PROCESSES 77

For further extensions of Pitman’s theorem we refer to Rogers (1981), Salmi-

nen (1983), Saisho and Tanemura (1990), Bertoin (1992), Rauscher (1997) and
Matsumoto and Yor (2000),(2001) (for the last one see also No. 53).

41. Let n = 2ν + 2 for ν ∈ R. Then a linear diffusion having the generator

1 d2 2 + 1 d
G (n) = + ,
2 dx2 2x dx
is called a Bessel process of order ν. Clearly, the Bessel processes in No. 35
are obtained from these when ν = −1/2, 0, 1/2, 1, . . . . With a slight abuse of
notation we let R(ν ) below denote a diffusion with the generator G (n) .
For the basic characteristics of Bessel processes see A1.21. Recall that the bound-
ary point 0 is:
(?) exit-not-entrance if ν ≤ −1,
(?) non-singular if −1 < ν < 0,
(?) entrance-not-exit if ν ≥ 0.

42. Assume that R0(ν ) = 0 and ν > −1. For −1 < ν < 0 let 0 be reflecting. Then
R(ν ) has the following properties.
Scaling. For every c > 0,
√ (ν )
R(ν ) ∼ { c Rt/c : t ≥ 0}.

Time inversion. Letting

(
0, t = 0,
Zt := (ν )
t R1/t , t > 0,

it holds (see Shiga and Watanabe (1973)) that

R(ν ) ∼ {Zt : t ≥ 0}.

From this we obtain

Rt( )
lim =0 a.s.
t→∞ t

43. Laws of the iterated logarithm. Let ν ≥ 0; then

Rt( )
(a) lim sup p =1 a.s.,
t↓0 2t ln ln(1=t)

R( )
(b) lim sup √ t =1 a.s.
t→∞ 2t ln ln t
 78 IV BROWNIAN MOTION

For −1/2 ≤ ν < 0, (a) and (b) hold if the boundary condition is reflection. For
a proof of (a) for ν = (n − 2)/2, n = 1, 2, . . . , see Itô and McKean (1974) p. 61.
For arbitrary ν ≥ −1/2, a proof of (a) is indicated in Revuz and Yor (2001) p.
450. Observe that (b) follows from (a) by time inversion.

44. For ν 6= −1/2 the drift term in the generator of a Bessel process is (only) locally
Lipschitz. Therefore, we can a priori expect that R(ν ) is characterizable via the
stochastic differential equation

 dRt(ν ) = dWt + 2 + 1

( )
dt, t > 0,
(a) 2 Rt
 (ν )
R0 = x > 0,

only up to the “explosion” time H0 := inf{t : Rt(ν ) = 0}.

In the case ν ≥ 0 the boundary point 0 is entrance-not-exit, and so H0 = ∞ a.s.
It can be proved, in this case, that R(ν ) is the unique strong solution of (a) (see
Chung and Williams (1990) p. 252).
We remark that the description via the SDE in (a) is also sufficient to charac-
terize R(ν ) in the case ν ≤ −1 because then 0 is exit-not-entrance.
When −1 < ν < 0 a boundary condition at 0 is involved. For instance, if
ν = −1/2 and the condition at 0 is reflection, then

R(ν ) ∼ |W |,

and there does not exist a SDE in the classical sense for R(ν ) . See Skorokhod’s
equation, e.g., in Ikeda and Watanabe (1981) p. 120. We refer also to Bertoin
(1990) and Yor (1997a) p. 6.

45. The scale function of R(ν ) , ν > 0, is s(x) = −x−2ν /(2ν). Hence, the function
h(x) := s(∞) − s(x) = x−2ν /(2ν) is excessive for R(ν ) . Standard computations
show that the corresponding h-transform (cf. II.31) of R(ν ) is R(−ν ) . If 0 < ν < 1,
in this setting, the boundary condition at 0 for R(−ν ) is killing.
Let P(ν ) and P(−ν ) be the measures associated to R(ν ) and R(−ν ) , respectively.
From the h-transform property we get the absolute continuity property
h(!(t))
 
P(x−ν ) (A; H0 > t) = E(xν ) c
; A , A ∈ Ft+ .
h(x)

46. The speed measure of R(ν ) , ν > 0, is m(dx) = 2 x2ν+1 dx, and the transition
density with respect to m is (see A1.21)
 2
1 x + y2 xy

p(ν ) (t; x, y) = (xy)−ν exp − Iν ( ).
2t 2t t
Because for small values of z,
1  z ν
Iν (z) '
( + 1) 2
 6. BESSEL PROCESSES 79

we have
t− −1
 2
y
p(ν ) (t; 0, y) = lim p(ν ) (t; x, y) =  +1 exp − .
x→0 2 ( + 1) 2t
From II.20 we obtain the distribution of the last exit time
p( ) (t; x; y)
P(xν ) (λy ∈ dt) = dt.
−s(y )

The time reversal property in II.33 gives further (see also Getoor and Sharpe
(1979) and Getoor (1979a))

P(y−ν ) (H0 ∈ dt) = P(0ν ) (λy ∈ dt).

47. We conclude this chapter by presenting a special case of a formula due to Pitman
and Yor (1982) (see also Yor (1992a)). For ν ≥ −1 let S (ν ) := (R(ν ) )2 be the
squared Bessel process of order ν, see A1.23. When −1 < ν < 0 the boundary
condition at 0 is reflection. Then
2 t
 Z


E(xν ) exp − αSt(ν ) − Ss(ν ) ds
2 0
x (1 + 2 −1 cth( t))

−ν−1  
= ch(βt) + 2αβ −1 sh(βt) exp − 1
2 (cth( t) + 2 − )
and
2 t
 Z 
E(xν ) exp − Ss(ν ) ds St(ν ) = y


2 0
 √xy 
 I
t x+y sh( t)

= exp (1 − βt cth(βt))  √xy  .
sh( t) 2t
I
t
Choosing here ν = −1/2 gives the classical Cameron–Martin formula for Brown-
ian motion (see also 1.1.9.3):
2 t
 Z


E0 exp − Ws2 ds = (ch(βt))−1/2 .
2 0

When ν = 0 we obtain Lévy’s area formula for 2-dimensional Brownian motion,

i.e., letting W (1) and W (2) be two independent Brownian motions started at 0,
then
 Z 1 
(Ws(1) dWs(2) − Ws(2) dWs(1) ) W1(1) = x, W1(2) = y


E exp iβ
0
2 1
 Z 
= E(00) exp − Ss(0) ds S1(0) = x2 + y 2


2 0
 2
x + y2

= exp (1 − β cth β) .
sh 2
 80 IV BROWNIAN MOTION

7. Geometric Brownian motion

48. For σ 6= 0, µ ∈ R, and x > 0 consider the linear SDE

dXt = σ Xt dWt + µ Xt dt, X0 = x,

where W is a standard Brownian started at 0. By III.15 this equation has

a strong unique non-exploding solution, and it is a simple application of Itô’s
formula to show that the solution is given by
2
  
Vt := x exp σ Wt + µ − t , t ≥ 0.
2
The process V = {Vt : t ≥ 0} is called a geometric or exponential Brownian
motion. It is a linear diffusion with the generator
1 2 2 d2 d
σ x + µx .
2 dx2 dx
Geometric Brownian motion is of primary interest in mathematical finance and
insurance, where it is used as an (unrealistic but computional) model for vari-
ous economical processes. Many recent results on geometric Brownian motion
originate from this setting, see, e.g., Geman and Yor (1993), Dufresne (1990),
Paulsen (1993), Shepp and Shiryaev (1993), and Yor (1992a).
For the basic characteristics of geometric Brownian motion, see A1.20. Part II
Section 9 contains formulae associated to geometric Brownian motion; therein
2
the drift coefficient µ is, for typographical reasons, taken to be σ 2 ν + .
2
49. We consider here transformations of geometric Brownian motion inspired by
Lévy’s construction of reflected Brownian motion and Pitman’s construction of
3-dimensional Bessel process (see No. 40).
For µ ∈ R let Wt(µ) := Wt + µ t and define the process Y (1,µ) by setting

Yt(1,µ) := exp −Wt(µ) sup exp Ws(µ) .



s≤t

Then, using the fact that the process

n o
sup Ws(µ) − Wt(µ) : t ≥ 0
s≤t

is identical in law to a reflecting Brownian motion with drift −µ (cf. No. 40), it
is seen that Y (1,µ) is a reflecting geometric Brownian motion with the generator
1 2 d2 1
d
G := x + −µ x
2 dx2 2 dx
acting on the domain

D = {f : f, Gf ∈ Cb ([1, ∞), f 0 (1+) = 0}.

 7. GEOMETRIC BROWNIAN MOTION 81

The filtrations generated by W and Y (1,µ) are the same. The process Y (1,µ) is
used in Shepp and Shiryaev (1994) when analyzing “Russian” options, and is
associated to the notion of “dual martingale measure”.
To formulate a similar extension of Pitman’s theorem introduce Y (2,µ) by setting

Yt(2,µ) := exp −Wt(µ) sup exp 2 Ws(µ) .



s≤t

Then, using Itô’s formula, it is seen that Y (2,µ) is a linear diffusion on [1, ∞)
starting at 1 (which is an entrance-not-exit boundary) with the generator

1 2 d2 1
d
x 2 + + |µ| cth(|µ| ln x) x .
2 dx 2 dx

The filtration generated by Y (2,µ) is strictly smaller than the filtration generated
by W.

50. Dufresne’s identity. For a 6= 0, b > 0 we have (see Dufresne (1990))

∞
2
Z
ea Ws −b s ds ∼ Z −1 ,
0
a2

where W0 = 0 and Z is a Γ(α, 1)-distributed random variable with α = 2b/a2 ,

i.e.,
1 α−1 −z
P(Z ∈ dz) = z e dz
( )
(cf. 9.1.8.4.(1)). The proof of Dufresne’s identity in Yor (1992c) is based on
the representation of geometric Brownian motion as a random time change of a
Bessel process (the so-called Lamperti relation). To explain this (cf. No.35), let
R be a Bessel process of order −ν < 0, started at 1 and independent of W. Then

eWt −ν t : t ≥ 0
 
(a) ∼ RA(−ν) : t ≥ 0 ,
t

where Z t
A(t−ν ) := e2(Ws −νs) ds.
0

As t → ∞
eWt −ν t → 0 a.s.
and it follows from (a) that

−ν )
A(∞ ∼ H0(−ν ) := inf{t : Rt(−ν ) = 0}.

The distribution of H0(−ν ) can be computed using standard methods (cf. II.10),
and this leads to Dufresne’s identity. Another, perhaps shorter, approach is by
time reversal using last exit times as explained in No. 45 and 46.
 82 IV BROWNIAN MOTION

51. The law of the additive functional

Z t
(ν )
At := e2 (Ws +ν s) ds
0

is of interest in mathematical finance (especially in studies concerning Asian

options), in mathematical physics (in studies on one-dimensional disorder sys-
tems, see Comtet, Monthus, and Yor (1998) and references therein), and in the
theory of hyperbolic Brownian motion (see Yor (1997b)). We notice that the
distribution of A(tν ) is given in 9.1.8.4. and the joint distribution of A(tν ) and
Wt(ν ) in 9.1.8.8. In particular, for ν = 0 we have (At := A(t0) )

1 + e2x  du dx
(a) P0 (At ∈ du, Wt ∈ dx) = exp − θex /u (t) .
2u u
The function t 7→ θr (t) in (a) is the (unnormalized) Hartman–Watson density
characterized via its Laplace transform:
Z ∞
1

exp − β 2 t θr (t) dt = Iβ (r),
0
2
where Iβ is the usual modified Bessel function. We have also the integral repre-
sentation due to Yor (1980),
 2
 − y2
Z ∞
r y


θr (t) = √ 3 exp − r ch(y) sh(y) sin dy.
2 t 0 2t t

In A2.14 and in the tables, iy (z) denotes this function and the connection be-
tween the different notations is

iy (z) = 2 θz (2y).

For further results (especially for positive and negative moments of At ) see
Yor (1980), (1992b), Dufresne (1999), and Donati-Martin, Matsumoto and Yor
(2000a), (2000b).

52. Bougerol’s identity. Consider the functional t 7→ At as introduced in No. 51.

Then, letting W ◦ be another standard Brownian motion started at 0 independent
of W , we have (see Bougerol (1983)) for every fixed t ≥ 0,

(a) WA◦ t ∼ sh (Wt ).

An equivalent formulation of (a) is

Z t
eWs dWs◦ ∼ sh (Wt ).
0

In Alili, Dufresne and Yor (1997) it is proved that (a) can be extended to an
equivalence in law for processes:
 Wt t −Ws
Z
dWs◦ : t ≥ 0

e e ∼ sh(Wt ) : t ≥ 0 .
0
 7. GEOMETRIC BROWNIAN MOTION 83

These results can be generalized for Brownian motion with drift. To give a flavor
of this, we quote from Alili, Dufresne and Yor (1997) the result
Z t
eWs +s dWs◦ ∼ sh (Wt + ε t),
0
where ε is a random variable independent of W such that
P(ε = +1) = P(ε = −1) = 1/2.

We refer to Yor (1992b), Alili, Dufresne and Yor (1997), Alili and Gruet (1997),
Comtet, Monthus and Yor (1998), and Matsumoto and Yor (1998) for proofs,
further results and generalizations (especially for Brownian motion with drift)
concerning Bougerol’s identity.

53. We present here two additional extensions of Lévy’s and Pitman’s theorems (cf.
No. 40 and No. 49). For the first one define the process Z (1,µ) by setting

t
Z
(1,µ) (µ)
exp Ws(µ) ds, µ ∈ R.


Zt := exp −Wt
0

Then (see Kramkov and Mordecki (1994) and Matsumoto and Yor (2000)) Z (1,µ)
is a linear diffusion with the generator
1 2 d2 
1

d
x + − µ x + 1 .
2 dx2 2 dx
Moreover, the filtration generated by Z (1,µ) is the same as the filtration generated
by W.
In Kramkov and Mordecki (1994) Z (1,µ) is used as a tool to solve the optimal
stopping problem associated to the pricing of an American perpetual integral
option.
For the second one introduce
Z t
Zt(2,µ) := exp −Wt(µ) exp 2 Ws(µ) ds,



µ ∈ R.
0

Then (see Matsumoto and Yor (1999), (2000), (2001)) the process Z (2,µ) is a
linear diffusion with the generator
1 2 d2 
1
K1+ (1=x) d

(a) x + − µ x + ,
2 dx2 2 K (1=x) dx
where Kµ is the modified Bessel function of order µ. Using the relationship
K1+µ (x) = µ x−1 Kµ (x) − Kµ0 (x)
the generator in (a) takes the form
1 2 d2 
1 K0 (1=x)  d
x + x− .
2 dx2 2 K (1=x) dx
It is also proved in Matsumoto and Yor (2000) that the filtration generated
by Z (2,µ) is strictly smaller than the filtration generated by W ; in analog with
Pitman’s theorem, see No. 40.
 CHAPTER V

LOCAL TIME AS A MARKOV PROCESS

1. Diffusion local time

1. Let X be a (regular) diffusion on an interval I. It is assumed that the scale

function s is in C 2 (I) and that the infinitesimal generator (see II.7) is given by
the ordinary differential operator

1 2 d2 d
a (x) 2 + b(x)
2 dx dx
with continuous a and b. For simplicity we also assume that a2 (x) > 0 for all
x ∈ I. Then s0 > 0 in the interior of I, and the measure m? given by
Z
?
m (A) = a−2 (x) dx, A ∈ B(I)
A

is well defined and finite on compact subsets of I. From the existence of the local
time L (see II.13) it follows that X has a local time {`(t, x) : t ≥ 0, x ∈ I} with
respect to the measure m? such that a.s. for all t ≥ 0 and A ∈ B(I),
Z t Z
1IA (Xs ) ds = `(t, x) m? (dx).
0 A

In fact, letting B be as in II.9

`(t, x) = 2 eB(x) L(t, x).

Below we give results concerning the Markovian character of the process

{`(ζ, x) : x ∈ I},

where ζ refers to the lifetime of X killed in an appropriate way. These kinds

of results are usually called Ray–Knight theorems, due to the fundamental con-
tributions of Ray (1963) and Knight (1963). Our basic reference, however, is
Section 4 in Walsh (1978). For a different approach based on the representation
theorems in the excursion filtrations see Jeulin and Yor (1979), McGill (1982)
and Jeulin (1985).
The results in No. 2 and 3 are obtained from No. 4 – which is Walsh’s theorem
4.1 – by an appropriate limiting procedure. The local time used by Walsh is

© Springer Basel 2015 84

A.N. Borodin, P. Salminen, Handbook of Brownian Motion - Facts and Formulae,
Probability and Its Applications, DOI 10.1007/978-3-0348-8163-0_5
 1. DIFFUSION LOCAL TIME 85

constructed from downcrossings, as explained in the case of Brownian motion in

IV.9 (see also Itô and McKean (1974) p. 222). Notice that our local time ` is
twice Walsh’s local time. The infinitesimal generators in this chapter are taken
in the extended sense, see, e.g., Revuz and Yor (2001) p. 285.

2. We assume first that X is transient and Xζ− = r a.s. Recall that r denotes the
right endpoint of I. It is also assumed that r is not elastic (the case with elastic
r is covered in No. 4). Let X0 = x and M̂ := inf{Xt : t ≥ 0}. Then the process

{`(ζ, y) : y ≥ M̂ }

is a non-time-homogeneous diffusion absorbed at 0 and having the space-time

infinitesimal generator:
(a) for M̂ < y ≤ x,
@2 s (y) 2 s0 (y)   @ @
  00
2z + 4 − + z + ,
@z 2 s0 (y) s(r) − s(y) @z @y

(b) for x < y < r,

@2 s (y) 2 s0 (y)   @ @
  00
2z + 2 − + z + .
@z 2 s0 (y) s(r) − s(y) @z @y

The process in (b) is started from the position of the process in (a) at time x
but is otherwise independent of it. The distribution of M̂ is
s(r) − s(x)
Px (M̂ ≤ y) = .
s(r) − s(y)
If l is non-singular, then due to the assumption Xζ− = r a.s., the boundary
condition at l is reflection. To complete the description in this case the Px -
distribution of `(ζ, l) is needed. This can be computed as explained in II.27 (a).
It can also be obtained by considering −X and using the result in No. 3.

3. Let X be as in No. 2 but now let Xζ− = l a.s., where l is the left endpoint of I.
Then the process
{`(ζ, y) : y > l}
is a non-time-homogeneous diffusion absorbed at 0 having the space-time infin-
itesimal generator:
(a) for l < y ≤ x,
@2 s00 (y) @ @
 
2z + 2 − z + ,
@z 2 s 0 (y ) @z @y
(b) for x ≤ y < r,
@2 s00 (y) @ @
2z 2 − 0 z + .
@z s (y) @z @y
The process in (b) is started from the position of the process in (a) at time x but
is otherwise independent of it. In case l is elastic the Px -distribution of `(ζ, l)
is needed to complete the description (cf. II.27 (a)).
86 V LOCAL TIME AS A MARKOV PROCESS

4. Next assume that X, X0 = x, is recurrent. Let L be the local time of X with

respect to the speed measure and τ ∼ Exp(c), c > 0, independent of X. For a
given zo ∈ I, a new diffusion X̂ is constructed by killing X at time

ζ = inf{t : L(t, zo ) ≥ τ }.

The killing measure of X̂ is c ε{zo } . We normalize the scale function s so that

s(zo ) = 0. Consider for x ≤ zo ,

' (zo )
Êzo (e−αHx ) = ,
' (x)

where ϕα is the fundamental decreasing solution associated to X̂. Let α → 0 to

obtain
1 1
P̂zo (Hx < ∞) = = , x ≤ zo ,
'0 ( x ) 1 − c s(x)
(see II.24; we may take ϕ0 ≡ 1 for x ≥ zo ). Following Walsh (1978), we introduce

1, x ≥ zo ,
(
p(x) := 1
, x ≤ zo .
1 − c s(x)
Then
{`(ζ, y) : y > M̂ }
is a non-time-homogeneous diffusion absorbed at 0 having the space-time infin-
itesimal generator:
(a) for M̂ < y ≤ x ∧ zo ,

@2 s00 (y) 2c s0 (y)
 @ @


2z + 4 − + z + ,
@z 2 s 0 (y ) 1 − c s(y) @z @y

(b) for x ∧ zo < y ≤ x ∨ zo ,

@2 (p2 (y) s0 (y))0  @ @


2z + 2 − z + ,
@z 2 p2 (y) s0 (y) @z @y

(c) for x ∨ zo < y < r,

@2 s00 (y) @ @
2z 2 − 0 z + .
@z s (y) @z @y

The processes in (a), (b), and (c) are knotted together in the same manner as
in No’s. 2 and 3. Observe also that if we study −X̂ instead of X̂ the result
above characterizes the local time process (or any part of it) in the reversed
y-direction. This is practical in some computations. Notice that the coefficient
of z in the drift terms can in every case be written as in (b).

5. As can be seen from above, the squared Bessel processes of dimensions 0, 2, and
4 (or of orders –1, 0, and 1, respectively) have important roles in the Ray–Knight
 2. BROWNIAN MOTION 87

theorems. We let (Z (δ) , Q(δ) ), δ = 0, 2, 4, respectively, denote these processes.

The corresponding generators are

d2 d
2z +δ , δ = 0, 2, 4.
dz 2 dz
These processes appear in No’s. 6, 10 and 17 below. For basic characteristics of
squared Bessel processes see A1.23.
For θ ∈ R we let (Z (δ,θ) , Q(δ,θ) ), δ = 0, 2, 4, denote diffusions with generators

d2 d
2z + (δ − θz) , δ = 0, 2, 4,
dz 2 dz
respectively. These are particular cases of squared radial Ornstein–Uhlenbeck
processes (see A1.26 for their basic characteristics, and also IV.37), and are used
below in No’s. 9, 11, 15 and 16 when treating Brownian motion with drift.
To construct non-time-homogeneous diffusions with generators of the form dis-
played, e.g., in (a) and (b) in No. 3, a deterministic space-time transformation
can be applied. Indeed, let g be an increasing function in C 2 (I) such that g 0 > 0.
Then for a ∈ I the process
n
1 (δ) o
(a) Zg(y)−g(a) : y ≥ a
g (y)
0

has the extended generator

@2 g00 (y) @ @
 
2z + δ − z +
@z 2 g0 (y) @z @y

(cf. McGill (1982)). To give an example, notice that

n o
Z (δ,θ) ∼ e−θt Z(e
(δ)
θt −1)/θ : t ≥ 0 .

Deterministic space-time transformations as in (a) are used to compute distri-

butions of the transformed “complicated” process in terms of the corresponding
known distributions of the original “simple” process (see the remarks in No’s.
6, 8, 17, 18 and 19).

2. Local time of Brownian motion

6. Local time stopped at the first hitting time. Let b ∈ R be given and
suppose that W0 = x ≤ b. We are interested in characterizing the law of the
process
`b := {`(Hb , y) : y ≤ b}.
Recall that the scale function of W is s(x) = x and the measure m? is the
Lebesgue measure. In the case (A) below the y-direction is from right to left,
i.e, from b to −∞. In the case (B) it is the opposite.
88 V LOCAL TIME AS A MARKOV PROCESS

(A) From No. 3 by considering −W we obtain

{`(Hb , b − y) : 0 ≤ y ≤ b − x} ∼ {Zy(2) : 0 ≤ y ≤ b − x},

{`(Hb , x − y) : y ≥ 0} ∼ {Zy(0) : y ≥ 0}.

The squared Bessel process Z (2) is started at its entrance boundary 0 and Z (0)
is started from the position of Z (2) at time b − x; otherwise Z (0) and Z (2) are
independent. The process Z (0) is run until it hits its exit boundary 0.
(B) Given that M̂ := inf{Wt : t ≤ Hb } = m we have from No. 2,

{`(Hb , m + y) : 0 ≤ y ≤ x − m} ∼ {Yy(4,b−m) : 0 ≤ y ≤ x − m},

{`(Hb , x + y) : 0 ≤ y < b − x} ∼ {Yy(2,b−x) : 0 ≤ y < b − x},

where Y0(4,b−m) := 0 and the generators of Y (4,b−m) and Y (2,b−x) are

@2 2z  @ @

2z 2 + 4− + ,
@z b − m − y @z @y
@2 2z  @ @

2z 2 + 2 − + ,
@z b − x − y @z @y
respectively. The process Y (2,b−x) is started from the position of Y (4,b−m) at time
x − m but otherwise Y (2,b−x) and Y (4,b−m) are independent.
(2)
Remark 1. The distribution of Zb−x can be found in A1.23. Hence (see also
II.27 (a)),
v
 
µ((v, ∞)) := Px (`(Hb , x) > v) = Q(2) (Z (2)
> v) = exp − .
0 b−x 2(b − x)
Using, e.g., time reversal as explained in the case of Bessel processes in IV.46
we have
y
(a) Q(0)
y (H0 ∈ dt) = e−y/2t dt.
2t2
Let M̂ := inf{Wt : t ≤ Hb }. Then `(Hb , M̂ ) = 0, and it can be checked using
(a) that
Z ∞
b−x
Px (M̂ < m) = = µ(dy)Q(0)y (H0 ≥ x − m), m ≤ x.
b−m 0

Remark 2. Let for a > 0 and 0 ≤ y < a,

1 1
y 7→ g(y, a) := − .
a−y a
From No. 5,
n
1 (4)
o
Y (4,b−m) ∼ Zg(y,b−m) : 0≤y ≤x−m ,
g0 (y; b − m)
 2. BROWNIAN MOTION 89

with Z0(4) = 0, and

n
1 (2)
o
Y (2,b−x) ∼ Zg(y,b−x) : 0≤y <b−x ,
g (y; b − x)
0

1
where Z (2) has the law µm given in (b) below. It follows from the
g0 (0; b − x) 0
SLLN (cf. IV.4) that limy→b−x Yy(2,b−x) = 0 a.s.
Remark 3. Using the explicit expressions for the transition density of Z (4) it
is seen that Y (4,b−m) , when considered on the time axis [0, b − m), is identical in
law to Z (4) started at 0 and conditioned to hit 0 at time b − m. The construction
is similar to the construction of the 3-dimensional Bessel bridge (see IV.15) and
the Brownian bridge (see IV.23). From this description of Y (4,b−m) it follows
that

(b) µm (dv) := Px (`(Hb , x) ∈ dv|M̂ = m) = λ2 v e−λv dv

with λ = (b − m)/2(x − m)(b − x). Moreover, it is seen that Y (2,b−x) is identical

in law to Z (2) started with the initial distribution µm and conditioned to hit 0
at time b − x.

7. Local time stopped at the first range time. Assume that W is started at
x. For r > 0 define the first range time (of the level r) via
n o
θr := inf t : sup Ws − inf Ws = r .
0≤s≤t 0≤s≤t

Given Wθr = z > x the process

`θ := {`(θr , y) : z − r ≤ y ≤ z}.

is identical in law to the process

{`(Hz , y) : z − r ≤ y ≤ z}

given M̂ := inf{Wt : t ≤ Hz } = z − r. Clearly, we have a similar description

in the case Wθr = z < x. Therefore, the law of `θ can be deduced from (B) in
No. 6.
Remark 1. From Remark 3 in No. 6, using symmetry,

Px (`(θr , x) ∈ dv|Wθr = z) = η 2 v e−ηv dv

with η = r/2(r − |x − z|)|z − x|. Further, from VI.22,

|z − x|
Px (Wθr ∈ dz) = dz, |z − x| ≤ r.
r2
90 V LOCAL TIME AS A MARKOV PROCESS

Remark 2. Because Y (4,b−m) , as introduced in No. 6, is a diffusion bridge

starting from 0 and ending at 0 at time b − m it has the time reversal property
(see Salminen (1997))
(4,b−m)
(a) {Yy(4,b−m) : 0 ≤ y ≤ b − m} ∼ {Yb−m−y : 0 ≤ y ≤ b − m}.

Applying (a) with b = z and m = z − r we obtain, given Wθr = z > x,

(4,r)
{`(θr , x − y) : 0 ≤ y ≤ x − z + r} ∼ {Yz−x+y : 0 ≤ y ≤ x − z + r}.

This identity in law is used in the computations in VI.23.

8. Local time stopped at the first exit time from a finite interval. Let
a < b be given and suppose that W0 = x ∈ (a, b). Consider

`a,b := {`(Ha,b , y) : a ≤ y ≤ b},

given WHa,b = b. Under this condition {Wt : 0 ≤ t < Ha,b } is a linear diffusion
R which is killed when it hits b. The generator of R is

1 d2 1 d
+ , x > a,
2 dx2 x − a dx
the scale function is s(x) = −(x − a)−1 , and the measure m? is the Lebesgue
measure. Therefore, the local time of R up to Hb (with respect to m? ) is identical
in law to `a,b given WHa,b = b and the results in No’s. 2 and 3 are applicable for
`a,b . Notice also that the boundary point a is entrance-not-exit for the diffusion
R.
(A) From No. 3, conditioning on WHa,b = b,

{`(Ha,b , b − y) : 0 ≤ y ≤ b − x} ∼ {Yy(2,b−a) : 0 ≤ y ≤ b − x},

{`(Ha,b , x − y) : 0 ≤ y < x − a} ∼ {Yy(0,x−a) : 0 ≤ y < x − a},

where Y0(2,b−a) := 0 and the generators of Y (2,b−a) and Y (0,x−a) are

@2 2z  @ @

2z + 2 − + ,
@z 2 b − a − y @z @y

@2 2z @ @
2z − + ,
@z 2 x − a − y @z @y
respectively. The process Y (0,x−a) is started from the position of Y (2,b−a) at time
b − x but otherwise Y (0,x−a) and Y (2,b−a) are independent.
(B) Here we consider the same process as in (A) but with the y-direction
reversed. Using No. 2 and (B) in No. 6 it is seen that given WHa,b = b and
M̂a,b := inf{Wt : t ≤ Ha,b } = m,

{`(Ha,b , m + y) : 0 ≤ y ≤ b − m}
 2. BROWNIAN MOTION 91

is identical in law to

{`(Hb , m + y) : 0 ≤ y ≤ b − m}

given M̂b := inf{Wt : t ≤ Hb } = m. Of course, the (conditional) distributions

of M̂b and M̂a,b are different.
Remark 1. Comparing the generator of Y (2,b−a) with the generator of Y (2,b−x) in
(B) in No. 6 reveals that Y (2,b−a) can be described similarly as Y (2,b−x) . Indeed,
Y (2,b−a) is identical in law to Z (2) started from 0 and conditioned to hit 0 at time
b − a. This gives

(b − a)v

µ((v, ∞)) := Px (`(Ha,b , x) > v | WHa,b = b) = exp − .
2(b − x)(x − a)

An alternative way to find this distribution is to use II.27 (a) (see also 1.3.3.6).
Remark 2. We show that Y (0,x−a) is identical in law to Z (0) started with the
initial distribution µ and conditioned to hit 0 before time x − a. To realize the
conditioning we use Doob’s h-transform techniques. For this, assume first that
Z (0) is started at xo and let nxo (0, ·) be the Q(0)
xo -density of H0 , see (a) in No. 6.
Define for ξ > 0,
( nz (0;  − t)
, 0 ≤ t < ξ,
?
h (z, t; 0, ξ) := nxo (0;  )
0, t ≥ ξ,

and introduce for ξ < x − a the measure

xo xo xo
 
κ(dξ) := Q(0)
xo (H0 ∈ dξ|H0 < x − a) = exp − + dξ.
2 2 2 2(x − a)

Keeping z and t fixed we integrate the function ξ 7→ h? (z, t; 0, ξ) with respect to

κ to obtain
x−a
z xo
Z  
h(z, t) := h? (z, t; 0, ξ)κ(dξ) = exp − + .
0
2(x − a − t) 2(x − a)

The function h is proportional to the probability that starting Z (0) from the
space-time point (z, t), t < x − a, the hitting time H0 will be less than x − a.
The desired conditioning can now be realized using this function h. Because

−h0t (z, t) = 2z h00zz (z, t)

it is seen that the h-transform of Z (0) has the same extended generator as
Y (0,x−a) .

9. Local time stopped at an exponential time. Let τ be an exponentially

distributed random variable (with parameter λ > 0) independent of W. By No. 4
 92 V LOCAL TIME AS A MARKOV PROCESS

we obtain (see the remark below) that under the conditions W0 = x, Wτ = zo ≥

x, and M̂τ := inf{Wt : t ≤ τ } = m,
{`(τ, m + y) : 0 ≤ y ≤ x − m} ∼ {Zy(4,θ) : 0 ≤ y ≤ x − m},
{`(τ, x + y) : 0 ≤ y ≤ zo − x} ∼ {Zy(2,θ) : 0 ≤ y ≤ zo − x},
{`(τ, zo + y) : y ≥ 0} ∼ {Zy(0,θ) : y ≥ 0},
√
where the processes Z (δ,θ) , δ = 0, 2, 4, are as in No. 5 with θ = 2 2λ.
Remark. To show that No. 4 can be applied let Ŵ be W killed at time τ.
Then Ŵ conditioned on Wτ = zo is a linear √ diffusion W̄ obtained from Ŵ by
constructing the h-transform with h = 2 2λ Gλ (·, zo ), where G is the Green
function of W : √
1
Gλ (x, y) = √ exp(− 2λ|x − y|).
2 2
From II.31 it is seen that
√ √
mh (dy) = exp(−2 2λ|y − zo |) 2dy, sh (dy) = exp(2 2λ|y − zo |) dy,
√
k h (dy) = 2 2λ εzo (dy)
can be taken to be the speed, scale, and killing measure, respectively, of the
h-transform W̄ . Hence, W̄ satisfies the assumptions in No. 4. Consequently,
letting ζ be the lifetime of W̄ and `o the local time of W̄ with respect to the
Lebesgue measure, we have conditionally on Wτ = zo ,
{`(τ, y) : y ∈ R} ∼ {`o (ζ, y) : y ∈ R}.
Next, we remark that for all x ≤ y ≤ zo ,
√
Px (`(τ, y) ≥ v | Wτ = zo ) = exp(− 2λ v).
Finally, observe by considering −W that Z (0,θ) and Z (4,θ) are, in a sense (cf.
II.33), time reversals of each other.

10. Local time stopped at the inverse local time. For zo ∈ R let
% := %(t, zo ) := inf{s : `(s, zo ) > t}
be the inverse Brownian local time evaluated at t ≥ 0. Assume that W0 = x <
zo . Then
{`(%, zo − y) : 0 ≤ y ≤ zo − x} ∼ {Zy(2) : 0 ≤ y ≤ zo − x},
{`(%, x − y) : y ≥ 0} ∼ {Zy(0 | 1) : y ≥ 0},
{`(%, zo + y) : y ≥ 0} ∼ {Zy(0 | 2) : y ≥ 0},
where Z (0 | 1) and Z (0 | 2) are 0-dimensional squared Bessel processes. The pro-
cesses Z (2) and Z (0 | 2) are started at t and Z (0 | 1) from the position of Z (2) at
time zo − x but otherwise Z (0 | 1) , Z (0 | 2) and Z (2) are independent.
Remark. The statement above cannot be proved directly using the general
results in No’s. 2, 3, and 4, and we refer to Knight (1981) p. 137 for a proof.
However, the Laplace transforms of the finite dimensional distributions can be
obtained by “inverting” in No. 9 with respect to λ, see also VI.16.
 3. BROWNIAN MOTION WITH DRIFT 93

3. Local time of Brownian motion with drift

11. Total local time. Let W (µ) be a Brownian motion with drift µ > 0 started at
x and `(µ) its local time with respect to the Lebesgue measure. Because W (µ) is
transient its total local time `(µ) (∞, y) is a.s. finite for all y. Here we characterize
the law of
{`(µ) (∞, y) : y ∈ R}.
The processes Z (δ,θ) , δ = 0, 2, 4, appearing below are as in No. 5 with θ = 2µ.
(A) By No. 3,

{`(µ) (∞, −y) : −∞ < y ≤ −x} ∼ {Zy(2,θ) : −∞ < y ≤ −x},

{`(µ) (∞, x − y) : y ≥ 0} ∼ {Zy(0,θ) : y ≥ 0}.

(B) Next, by No. 2, given M̂ := inf{Wt(µ) : t ≥ 0} = m

{`(µ) (∞, m + y) : 0 ≤ y ≤ x − m} ∼ {Zy(4,θ) : 0 ≤ y ≤ x − m},

{`(µ) (∞, x + y) : y ≥ 0} ∼ {Zy(2,θ) : y ≥ 0}.

Remark. Notice that Z (2,θ) is positively recurrent and its stationary probability
distribution is an exponential distribution with parameter µ. Because Z (2,θ) in
(A) is considered on the time axis (−∞, −x), it is, in fact, considered in its
stationary state. This means that the finite dimensional distributions of Z (2,θ)
are given for t1 < t2 < · · · < tn < −x by

Q(2,θ) (2,θ)
m̂ (Zt1 ∈ dz1 , Zt(2,θ)
2
∈ dz2 , . . . , Zt(2,θ)
n
∈ dzn )
= m̂(dz1 ) r̂(t2 − t1 ; z1 , z2 ) m(dz2 ) · . . . · r̂(tn − tn−1 ; zn−1 , zn ) m(dzn ),

where m̂, m, and r̂ are the stationary probability measure, speed measure, and
transition density of Z (2,θ) , respectively. In particular,

Px (`(µ) (∞, y) ≥ v) = Px (`(µ) (∞, y) ≥ v, Hy < ∞)

= Px (Hy < ∞)Py (`(µ) (∞, y) ≥ v)
 −µ v
e , y ≥ x,
=
e−2µ (x−y)−µ v , y ≤ x,

(cf. II.27 (a) for y = x).

12. Local time stopped at the first hitting time, part 1. The next Ray–Knight
theorem for W (µ) , µ > 0, gives the law of

{`(µ) (Hb , y) : y ≤ b}

when W (µ) is initiated at x < b. The processes Z (δ,θ) , δ = 0, 2, 4, are as in No. 11.
 94 V LOCAL TIME AS A MARKOV PROCESS

(A) Considering −W (µ) it is seen from No. 3 that

{`(µ) (Hb , b − y) : 0 ≤ y ≤ b − x} ∼ {Zy(2,θ) : 0 ≤ y ≤ b − x},

{`(µ) (Hb , x − y) : y ≥ 0} ∼ {Zy(0,θ) : y ≥ 0}.

(B) Given that M̂ := inf{Wt(µ) : t ≤ Hb } = m,

{`(µ) (Hb , m + y) : 0 ≤ y ≤ x − m} ∼ {Yy(4,µ,b−m) : 0 ≤ y ≤ x − m},

{`(µ) (Hb , x + y) : 0 ≤ y < b − x} ∼ {Yy(2,µ,b−x) : 0 ≤ y < b − x},

where Y0(4,µ,b−m) := 0 and the generators of Y (4,µ,b−m) and Y (2,µ,b−x) are

@2 @ @
2z + (4 − 2µ z cth(µ(b − m − y))) + ,
@z 2 @z @y

@2 @ @
2z + (2 − 2µ z cth(µ (b − x − y))) + ,
@z 2 @z @y
respectively. The process Y (2,µ,b−x) is started from the position of Y (4,µ,b−m) at
time x − m but otherwise Y (2,µ,b−x) and Y (4,µ,b−m) are independent.
Remark 1. In particular,


 
(a) Px (`(µ) (Hb , x) ≥ v) = exp − v
1 − exp(−2 (b − x))

(see also 2.2.3.4).

Remark 2. It can be proved that Y (2,µ,b−x) is identical in law to Z (2,2µ) started
with the exponential distribution given in (a) and conditioned to hit 0 at time
b − x.

13. Local time stopped at the first hitting time, part 2. Let W (µ) , µ > 0,
be started at x > a and killed if it hits a. For this process we study the total
local time
µ)
`(∞ := {`(µ) (∞, y) : y ≥ a}.

Conditioned on Ha < ∞, the process W (µ) is identical in law to W (−µ) killed

µ)
when it hits a. It follows that the law of `(∞ is as described in No. 12. On the
other hand, given Ha = ∞ and M̂ := inf{Wt : t ≥ 0} = m, we obtain from
No. 2,

{`(µ) (∞, m + y) : 0 ≤ y ≤ x − m} ∼ {Zy(4,θ) : 0 ≤ y ≤ x − m},

{`(µ) (∞, x + y) : y ≥ 0} ∼ {Zy(2,θ) : y ≥ 0},

where Z (δ,θ) , δ = 0, 2, 4, are as in No’s. 11 and 12.
 3. BROWNIAN MOTION WITH DRIFT 95

Remark 1. The process W (µ) given Ha = ∞ is identical in law (cf. IV.33) to

the diffusion X + associated to the generator

1 d2 d
+ µ cth(µ (x − a)) .
2 dx2 dx

The scale function of X + is s(x) = 1 − cth(µ (x − a)). Notice that s(a) = −∞

and s(+∞) = 0. Let `+ denote the total local time of X + . Then `(∞µ)
under the
+
condition Ha = ∞ is identical in law to ` , for which No. 2 can be applied.
Using the scale function of X + it is seen that

Px (M̂ ∈ dm | Ha = ∞) = µ e2µ (m−a) (cth(µ (x − a)) − 1) dm,

and so, evoking the explicit form of the transition density of Z (4,θ) (see A1.26)
and integrating, we obtain

 exp( (x − a))
 
(a) Px (`(µ) (∞, x) ≥ v | Ha = ∞) = exp − v .
2 sh( (x − a))

An alternative and more direct way to compute this distribution is to use II.27
(a) (see also 2.2.3.2.(1)).
Remark 2. We leave the complete description in the other y-direction to the
reader. From the point of view of explicit computation it is, however, good to
know that

{`(µ) (∞, x − y) : 0 ≤ y < x − a} ∼ {Yy(0,µ,x−a) : 0 ≤ y < x − a},

where the generator of Y (0,µ,x−a) is

@2 @ @
2z − 2µ z cth(µ(x − a − y)) + .
@z 2 @z @y

The process Y (0,µ,x−a) is started at time 0 with the exponential distribution given
in (a). We remark that Y (0,µ,x−a) is identical in law to Z (0,2µ) conditioned to hit
0 before time x − a.

14. Local time stopped at the first exit time from a finite interval. Consider
W (µ) started at x ∈ (a, b) and killed when it exits (a, b) (cf. No. 7). Conditioned
(µ)
on WH a,b
= a, this process is a diffusion with the generator (cf. No. 13)

1 d2 d
− |µ| cth(|µ| (b − x)) ,
2 dx2 dx

and having a as a killing boundary. Therefore, the Ray–Knight theorem here

is very similar to the one in No. 12. For the y-direction from left to right we
(µ)
obtain from No. 3 that, given WH a,b
= a,

{`(µ) (Ha,b , a + y) : 0 ≤ y ≤ x − a} ∼ {Yy(2,|µ|,b−a) : 0 ≤ y ≤ x − a},

 96 V LOCAL TIME AS A MARKOV PROCESS

{`(µ) (Ha,b , x + y) : 0 ≤ y < b − x} ∼ {Yy(0,|µ|,b−x) : 0 ≤ y < b − x},

where Y0(2,|µ|,b−a) := 0 and the generators of Y (2,|µ|,b−a) and Y (0,|µ|,b−x) are

@2
@ @
2z 2 + 2 − 2|µ| z cth(|µ|(b − a − y)) + ,
@z @z @y

@2 @ @
2z − 2|µ| z cth(|µ|(b − x − y)) + ,
@z 2 @z @y
respectively.
Remark 1. The process Y (2,|µ|,b−a) is identical in law to Z (2,2|µ|) started at 0
and conditioned to hit 0 at time b − a. This gives, e.g.,

|| sh(||(b − a))

 
(µ)
(a) Px (`(µ) (Ha,b , x) ≥ v | WH = a) = exp − v .
a,b 2 sh(||(x − a)) sh(||(b − x))

The process Y (0,|µ|,b−x) is identical in law to Z (0,2|µ|) started with the exponential
distribution given in (a) and conditioned to hit 0 before time b − x.
Remark 2. We leave the complete description in the other y-direction also
here to the interested reader. Notice, however, that from the characterization
of Y (2,|µ|,b−a) (or from No. 2) it follows that

{`(µ) (Ha,b , x − y) : 0 ≤ y < x − a} ∼ {Yy(2,|µ|,x−a) : 0 ≤ y < x − a},

where Y (2,|µ|,x−a) is started at time 0 with the distribution given in (a) and the
generator of Y (2,|µ|,x−a) is

@2
@ @
2z 2 + 2 − 2|µ| z cth(|µ|(x − a − y)) + .
@z @z @y

Hence Y (2,|µ|,x−a) is identical in law to Z (2,2|µ|) conditioned to hit 0 at time x − a.

15. Local time stopped at an exponential time. Let τ ∼ Exp(λ), independent

of W (µ) , and let Ŵ (µ) be W (µ) killed at time τ. Recall (see A1.14) that the Green
function of W (µ) is
√ √
1 −( 2α+µ2 +µ) x ( 2α+µ2 −µ) y
Gα (x, y) = p e e , x ≥ y.
2 2 + 2

Then, Ŵ (µ) under the condition Wτ(µ) = zo is a diffusion which can be con-
structed as an h-transform of Ŵ (µ) using

G (y; zo )
h(y) = .
G (zo ; zo )

Computing the basic characteristics of this h-process and comparing these with
the corresponding characteristics in No. 9 it is seen that the conditioned pro-
cesses are very similar in both cases. Essentially, all we have to do is to change
 4. BESSEL PROCESS 97

the parameter λ in No. 9 to λ + µ2 /2. Once this change is made we obtain the
desired description for
{`(µ) (τ, y) : y ∈ R}.

16. Local time stopped at the inverse local time. For zo ∈ R let

% := %(µ) (t, zo ) := inf{s : `(µ) (s, zo ) > t}

be the inverse local time evaluated at t ≥ 0. Assume that W0(µ) = x < zo , and
let Z (0,θ) and Z (2,θ) be as in No. 5 with θ = 2|µ|. Then given % < ∞,

{`(µ) (%, zo − y) : 0 ≤ y ≤ zo − x} ∼ {Zy(2,θ) : 0 ≤ y ≤ zo − x},

{`(µ) (%, x − y) : y ≥ 0} ∼ {Zy(0,θ | 1) : y ≥ 0},

{`(µ) (%, zo + y) : y ≥ 0} ∼ {Zy(0,θ | 2) : y ≥ 0},
where Z (0,θ | 1) and Z (0,θ | 2) are independent copies of Z (0,θ) . The processes Z (2,θ)
and Z (0,θ | 2) are started at t, and Z (0,θ | 1) is initiated from the position of Z (2,θ)
at time zo − x.

4. Local time of Bessel process

17. Total local time when ν > 0. Let R(ν ) denote a Bessel process of order ν
and `(ν ) its local time with respect to the Lebesgue measure. For ν > 0, R(ν ) is
transient and we may consider its total local time `(ν ) (∞, y), y > 0. We assume
here (and below) R0(ν ) = x > 0, and leave the descriptions in the case x = 0 to
the reader.
(A) Applying No. 3 for −R(ν ) gives

{`(ν ) (∞, −y) : −∞ < y ≤ −x} ∼ {X̂y(2,0) : −∞ < y ≤ −x},

{`(ν ) (∞, x − y) : 0 ≤ y < x} ∼ {X̂y(0,x) : 0 ≤ y < x},

where the extended generators of X̂ (2,0) and X̂ (0,x) are

@2 (2 + 1) z  @ @

2z 2 + 2− + ,
@z −y @z @y

@2 (2 + 1) z @ @
2z − + ,
@z 2 x − y @z @y
respectively.
(B) Given M̂ := inf{Rt(ν ) : t ≥ 0} = m we obtain from No. 2,

{`(ν ) (∞, m + y) : 0 ≤ y ≤ x − m} ∼ {X̃y(4,m) : 0 ≤ y ≤ x − m},

{`(ν ) (∞, x + y) : y ≥ 0} ∼ {X̃y(2,x) : y ≥ 0},

 98 V LOCAL TIME AS A MARKOV PROCESS

where X̃0(4,m) := 0 and the generators of X̃ (4,m) and X̃ (2,x) are

@2 (2 − 1) z  @ @

2z + 4 − + ,
@z 2 m+y @z @y

@2 (2 − 1) z  @ @

2z + 2 − + ,
@z 2 x+y @z @y
respectively. In the 3-dimensional case, that is, ν = 1/2,

X̃ (4,m) ∼ Z (4) , X̃ (2,x) ∼ Z (2) .

Remark 1. For the description in (A) it is essential to have the P(xν ) -distribution
of `(ν ) (∞, y), y ≥ x, which can be found using II.27 (a):


(a) P(xν ) (`(ν ) (∞, y) ≥ v) = P(yν ) (`(ν ) (∞, y) ≥ v) = exp − v .
y

Using this we can write down the finite-dimensional distributions of X̂ (2,0) sim-
ilarly as explained for Brownian motion with drift in the Remark in No. 11.
Remark 2. Let for a ≥ 0 and 0 < y < a

y 7→ h(y, a) := (a − y)−2ν − a−2ν .

Then for all b > x, assuming that the initial distributions coincide, we have from
No. 5,

(2,0)
n
1 (2)
o
{X̂−b+y : 0 ≤ y ≤ b − x} ∼ Z : 0 ≤ y ≤ b − x ,
h0 (y; b) h(y,b)
n
1 (0)
o
{X̂y(0,x) : 0 ≤ y < x} ∼ Zh(y,x) : 0≤y<x .
h (y; x)
0

In the 3-dimensional case X̂ (0,x) is identical in law to Z (0) conditioned to hit 0

before time x (cf. No. 8).
Remark 3. Similarly as in Remark 2, introduce for a > 0 and y ≥ 0 the
function
y 7→ g(y, a) := (a + y)2ν − a2ν .
Then from No. 5 assuming that the initial distributions coincide,
n
1 (4)
o
{X̃y(4,m) : 0 ≤ y ≤ x − m} ∼ Z : 0 ≤ y ≤ x − m ,
g0 (y; m) g(y,m)
n
1 (2)
o
{X̃y(2,x) : y ≥ 0} ∼ Zg(y,x) : y≥0 .
g (y; x)
0

18. Total local time when ν < 0. Let `(ν ) be as in No. 17 with ν < 0. If ν > −1
then 0 is non-singular and we pose on it the boundary condition of killing. The
reflecting case is considered in No’s. 19 and 20.
 4. BESSEL PROCESS 99

Using the result in No. 3 (we leave the other y-direction to the reader) it is seen
that
{`(ν ) (∞, y) : 0 ≤ y ≤ x} ∼ {X̄y(2,0) : 0 ≤ y ≤ x},

{`(ν ) (∞, x + y) : y ≥ 0} ∼ {X̄y(0,x) : y ≥ 0},

where the generators of X̄ (2,0) and X̄ (0,x) are

@2 (2 + 1) z  @ @

2z + 2 + + ,
@z 2 y @z @y

@2 (2 + 1) z @ @
2z + + ,
@z 2 x + y @z @y
respectively. For ν = −1/2 the Bessel process R(ν ) is a Brownian motion on
(0, ∞) killed when it hits 0. The above result, in this case, has already been
presented in No. 6.
Remark 1. To express X̄ (2,0) and X̄ (0,x) in terms of Z (2) and Z (0) , introduce for
a ≥ 0 and y ≥ 0,
y 7→ g(y, a) := (a + y)−2ν − a−2ν .
Then n o n
1 (2)
o
X̄y(2,0) : 0 < y ≤ x ∼ Zg(y,0) : 0<y≤x ,
g (y; 0)
0

n o n
1 (0)
o
X̄y(0,x) : y ≥ 0 ∼ Zg(y,x) : y≥0 ,
g0 (y; x)
where Z (2) is started at 0.
Remark 2. Notice that
| |
P(xν ) (`(ν ) (∞, x) ≥ v) = exp −


v
x

which coincides with the corresponding distribution in the case ν > 0, see (a) in
No. 17. This can be explained by the h-transform property of Bessel processes
(cf. IV.43) and the last exit decompositions.

19. Local time stopped at the first hitting time. Next we study the process

{`(ν ) (Hb , y) : 0 ≤ y ≤ b}

under the condition R0(ν ) = x < b. We consider only the case when the y-direction
is from right to left.
Assume first that ν > 0 or −1 < ν < 0 with 0 reflecting. Then from No. 3, by
considering −R(ν ) ,

{`(ν ) (Hb , b − y) : 0 ≤ y ≤ b − x} ∼ {X̂y(2,b) : 0 ≤ y ≤ b − x},

{`(ν ) (Hb , x − y) : 0 ≤ y ≤ x} ∼ {X̂y(0,x) : 0 ≤ y ≤ x},

 100 V LOCAL TIME AS A MARKOV PROCESS

where the generators of X̂ (2,b) and X̂ (0,x) (cf. (A) in No. 17) are

@2 (2 + 1) z  @ @

2z 2 + 2− + ,
@z b−y @z @y

@2 (2 + 1) z @ @
2z − + ,
@z 2 x − y @z @y
respectively. Notice that for ν = −1/2 we have the Ray–Knight theorem for
reflecting Brownian motion.
Assume next ν ≤ −1 or −1 < ν < 0 with 0 a killing boundary. It is now
possible that Hb = ∞; this case is discussed in No. 20. However, conditioning
on Hb < ∞,
{Rt(ν ) : t < Hb } ∼ {Rt(−ν ) : t < Hb }
(see IV.43), and hence, under the condition Hb < ∞,

{`(ν ) (Hb , y) : 0 ≤ y ≤ b} ∼ {`(−ν ) (Hb , y) : 0 ≤ y ≤ b}.

For ν = −1/2, i.e., R(ν ) is a killed Brownian motion, these results are already
given in No. 6.
Remark. Using the characterization of X̂ (2,b) in terms of Z (2) (see Remark 2 in
No. 17) we obtain for ν > 0 and x < b

h0 (b − x; b)
 
P(xν ) (`(ν ) (Hb , x) ≥ v) = exp − v ,
2 h(b − x; b)

where
y 7→ h(y, b) := (b − y)−2ν − b−2ν .
By II.27 (a) the parameter of this exponential distribution can also be expressed
as 1/m(x)G(x, x), where m is the speed measure and G := G0 is the Green
function of R(ν ) killed at Hb .

20. Local time stopped at the first exit time from a finite interval. Let
0 < a < b and x ∈ (a, b) be given and consider the process

`(ν ) (Ha,b ) := {`(ν ) (Ha,b , y) : a ≤ y ≤ b}.

We are interested in characterizing the law of `(ν ) (Ha,b ) under the condition
Ha,b = Ha . The complete result is not given but we indicate the facts needed to
be able to apply No. 3. For ν > 0 let

s(b) − s(x)
g(x) := = P(xν ) (Ha,b = Ha ),
s(b) − s(a)

where s(x) = −x−2ν /2ν is the scale function of R(ν ) (see A1.21). Use g to
build the excessive transform of R(ν ) . This transform is identical in law to R(ν )
 4. BESSEL PROCESS 101

conditioned by Ha,b = Ha . Standard computations show that the generator of

the transform is
1 d2 
2 + 1 g0 (x)  d
2 + + .
2 dx 2x g(x) dx
Observe that this makes sense also when a = 0 if we take g(x) = s(b) − s(x).
The scale function of the h-transform can be obtained, e.g., from the general
formula in II.31. The diffusion characterization of `(ν ) (Ha,b ) given Ha,b = Ha
follows by applying No. 3 to the excessive transform.
Similar computations when ν < 0 show that the results above do not depend on
the sign of ν, that is, conditioned on the event Ha,b = Ha ,

{`(ν ) (Ha,b , y) : a ≤ y ≤ b} ∼ {`(−ν ) (Ha,b , y) : a ≤ y ≤ b}.

When ν = −1/2 the Bessel process R(ν ) is a Brownian motion on R+ killed

when it hits 0. Therefore, the above result in the case ν = ±1/2 has already
been discussed in No. 8.

21. Local time stopped at an exponential time. The discussion here follows
that of No. 9 (see also No. 15). Assume that ν > 0, and let τ ∼ Exp(λ)
independent of R(ν ) . First, R(ν ) started at x and conditioned on Rτ(ν ) = zo > x,
is the excessive transform of R(ν ) obtained by using
G (x; zo )
h(x) := .
G (zo ; zo )

Let R̄(ν ) denote this h-transform and let ζ and `¯(ν ) be its lifetime and local time,
respectively. Then under the condition Rτ(ν ) = zo ,

{`(ν ) (τ, y) : y ≥ 0} ∼ {`¯(ν ) (ζ, y) : y ≥ 0}.

The latter process can be characterized as explained in No. 4. A properly

normalized scale function is
G (y; zo ) −2
Z x 
h
s (x) = s(dy)
zo
G (zo ; zo )

 (x)  (zo )
  
 ϕ2λ (zo ) ' (x) − ' (z ) , x ≥ zo ,

  o
=
 ψλ2 (zo ) ' (zo ) − ' (x) , x ≤ zo ,
 

 (zo )  (x)

where s(dy) = y −2ν−1 dy is the scale measure (cf. A1.21), and

√ √
ϕλ (x) = x−ν Kν (x 2λ), ψλ (x) = x−ν Iν (x 2λ).

The function p appearing in No. 4 takes the form

1, x ≥ zo ,
(
p(x) = ' (zo )  (x)
, x ≤ zo ,
' (x)  (zo )
102 V LOCAL TIME AS A MARKOV PROCESS

and hence, for all x,

√
(p2 (x) (sh (x))0 )0 s00 (x) '0 (x) 1 √ K 0 (x 2)
= − 2 = − − 2λ  √ .
p2 (x) (sh (x))0 s 0 (x ) ' (x) x K (x 2)
Putting this into the expressions in No. 4 gives us the Ray–Knight theorem for
R̄(ν ) , leading to the desired characterization. In particular, for y ∈ [x, zo ],
v
 
P(xν ) (`(ν ) (τ, y) ≥ v|R(ν ) (τ ) = zo ) = exp − ,
m(y) G (y; y)
where m is the density of the speed measure.
Now let ν < 0. Straightforward computations show that R̄(ν ) is identical in law
to R̄(−ν ) . Therefore, from the results above we also obtain results for R(ν ) , ν < 0,
by doing some obvious changes. The case ν = 0 is left to the reader.
We conclude by considering the case ν = 1/2. Now (see A1.22)

(p2 (x) (sh (x))0 )0 √

= 2 2λ.
p2 (x) (sh (x))0
Consequently, the description is very similar to the one in No. 9. Notice,
however, that the process

{`(1/2) (τ, x − y) : 0 ≤ y ≤ x}

must be different. In fact, it is identical in law to a non-time-homogeneous

diffusion having the generator
@2 √ √ @ @
2v 2 − 2 2λ z cth( 2λ(x − y)) + .
@z @z @y
√
It can be proved that this is also the generator of Z (0,θ) with θ = 2 2λ condi-
tioned to hit 0 before time x (cf. Remark 1 in No. 14).

5. Summarizing tables

Below we review in a table form some of the Ray–Knight results presented

above. For Bessel processes of positive order we have included only three cases,
and Bessel processes of negative order do not appear at all. For these (as well
as for reflecting Brownian motion) we refer to No’s. 17–21. In the following
tables references are given to the sections, where the result is discussed in more
detail and where one can also find the unexplained notations explained. For
typographical reasons the partial differential operator ∂/∂y associated to the
time-parameter y (in non-time-homogeneous cases) is not displayed. Recall that
Ha is the first hitting time of a, Ha,b is the first exit time from the interval (a, b),
τ ∼ Exp(λ) independent of the considered process, % is the inverse local time
at zo , the starting state of the process is denoted by x, and m is the value of the
infimum of the process up to the considered stopping time.
 5. SUMMARIZING TABLES 103

Brownian motion

Local time process Generator

No. 6 (A)
d2 d
{`(Hb , b − y) : 0 ≤ y ≤ b − x} 2z +2
dz 2 dz
d2
{`(Hb , x − y) : y ≥ 0} 2z
dz 2
No. 6 (B)
@2 2z  @

{`(Hb , m + y) : 0 ≤ y ≤ x − m} 2z 2 + 4−
@z b − m − y @z
@2 2z  @

{`(Hb , x + y) : 0 ≤ y < b − x} 2z 2 + 2 −
@z b − x − y @z

No. 8 Given Ha,b = Hb ,

@2 2z  @

{`(Ha,b , b − y) : 0 ≤ y ≤ b − x} 2z 2 + 2−
@z b − a − y @z
@2 2z @
{`(Ha,b , x − y) : 0 ≤ y < x − a} 2z 2 −
@z x − a − y @z

No. 9 Given Wτ = zo ,
d2 √ d
{`(τ, m + y) : 0 ≤ y ≤ x − m} 2z 2 + (4 − 2z 2λ)
dz dz
d2 √ d
{`(τ, x + y) : 0 ≤ y ≤ zo − x} 2z 2 + (2 − 2z 2λ)
dz dz
d2 √ d
{`(τ, zo + y) : y ≥ 0} 2z 2 − 2z 2λ
dz dz
No. 10
d2 d
{`(%, zo − y) : 0 ≤ y ≤ zo − x} 2z +2
dz 2 dz
d2
{`(%, x − y) : y ≥ 0} 2z 2
dz
d2
{`(%, zo + y) : y ≥ 0} 2z 2
dz
104 V LOCAL TIME AS A MARKOV PROCESS

Brownian motion with drift µ > 0

Local time process Generator

No. 11 (A)
d2 d
{`(µ) (∞, −y) : −∞ < y ≤ −x} 2z + (2 − 2µz)
dz 2 dz
d2 d
{`(µ) (∞, x − y) : y ≥ 0} 2z 2 − 2µz
dz dz
No. 11 (B)
d2 d
{`(µ) (∞, m + y) : 0 ≤ y ≤ x − m} 2z + (4 − 2µz)
dz 2 dz
d2 d
{`(µ) (∞, x + y) : y ≥ 0} 2z 2 + (2 − 2µz)
dz dz
No. 12 (A)
d2 d
{`(µ) (Hb , b − y) : 0 ≤ y ≤ b − x} 2z + (2 − 2µz)
dz 2 dz
d2 d
{`(µ) (Hb , x − y) : y ≥ 0} 2z 2 − 2µz
dz dz
No. 12 (B)
@2 @
{`(µ) (Hb , m + y) : 0 ≤ y ≤ x − m} 2z + (4 − 2µ z cth(µ y1 ))
@z 2 @z
@2 @
{`(µ) (Hb , x + y) : 0 ≤ y < b − x} 2z 2 + (2 − 2µ z cth(µ y2 ))
@z @z
y1 := b − m − y, y2 := b − x − y

No. 13 Given Ha = ∞,
d2 d
{`(µ) (∞, m + y) : 0 ≤ y ≤ x − m} 2z + (4 − 2µz)
dz 2 dz
d2 d
{`(µ) (∞, x + y) : y ≥ 0} 2z 2 + (2 − 2µz)
dz dz
No. 14 Given Ha,b = Ha ,
@2
@
{`(µ) (Ha,b , a + y) : 0 ≤ y ≤ x − a} 2z 2 + 2 − 2|µ| z cth(|µ|y1 )
@z @z
@2 @
{`(µ) (Ha,b , x + y) : 0 ≤ y < b − x} 2z 2 − 2|µ| z cth(|µ|y2 )
@z @z
y1 := b − a − y, y2 := b − x − y
 5. SUMMARIZING TABLES 105

Brownian motion with drift µ > 0 (cont’d)

Local time process Generator

No. 15 Given Wτ(µ) = zo ,

d2 p d
{`(µ) (τ, m + y) : 0 ≤ y ≤ x − m} 2z + (4 − 2z 2λ + µ2 )
dz 2 dz
d2 p d
{`(µ) (τ, x + y) : 0 ≤ y ≤ zo − x} 2z 2 + (2 − 2z 2λ + µ2 )
dz dz
d2 p d
{`(µ) (τ, zo + y) : y ≥ 0} 2z 2 − 2z 2λ + µ2
dz dz
No. 16 Given % < ∞,
d2 d
{`(µ) (%, zo − y) : 0 ≤ y ≤ zo − x} 2z + (2 − 2|µ|z)
dz 2 dz
d2 d
{`(µ) (%, x − y) : y ≥ 0} 2z 2 − 2|µ|z
dz dz
d2 d
{`(µ) (%, zo + y) : y ≥ 0} 2z 2 − 2|µ|z
dz dz

Bessel process of order ν > 0

Local time process Generator

No. 17 (A)
@2 (2 + 1) z  @

{`(ν ) (∞, −y) : −∞ < y ≤ −x} 2z + 2 −
@z 2 −y @z
@ 2 (2 + 1) z @
{`(µ) (∞, x − y) : 0 ≤ y < x} 2z 2 −
@z x − y @z

No. 17 (B)
@2 (2 − 1) z  @

{`(ν ) (∞, m + y) : 0 ≤ y ≤ x − m} 2z + 4 −
@z 2 m+y @z
@2 (2 − 1) z  @

{`(ν ) (∞, x + y) : y ≥ 0} 2z 2 + 2 −
@z x+y @z

No. 19
@2 (2 + 1) z  @

{`(ν ) (Hb , b − y) : 0 ≤ y ≤ b − x} 2z + 2 −
@z 2 b−y @z
@2 (2 + 1) z @
{`(ν ) (Hb , x − y) : 0 ≤ y ≤ x} 2z 2 −
@z x − y @z
 CHAPTER VI

DIFFERENTIAL SYSTEMS
ASSOCIATED TO BROWNIAN MOTION

1. The Feynman–Kac formula

1. For T > 0 let F : R 7→ R, f : [0, T ] × R 7→ R+ and g : [0, T ] × R 7→ R be

given functions. Assume that g is bounded and that F is bounded and Hölder
continuous. If, moreover, g and f are Hölder continuous locally in t, that is, e.g.
for g, there exist for any N < ∞ constants C > 0 and α > 0 such that

|g(t, x) − g(t, y)| < C |x − y|α

for all t < N, then the Feynman–Kac formula (cf. II.24) says that
  Z T 
v(t, x) := E F (WT ) exp − f (s, Ws ) ds
t
Z T  Z s  
+ g(s, Ws ) exp − f (v, Wv ) dv ds Wt = x
t t

is the unique solution of the problem

1
−u0t (t, x) = u00xx (t, x) − f (t, x)u(t, x) + g(t, x), (t, x) ∈ (0, T ) × R,
2
u(T, x) = F (x), x ∈ R.

For a proof (for more general diffusions than Brownian motion) see, e.g., Ka-
ratzas and Shreve (1991) p. 366–368. The Hölder conditions above are needed
to guarantee enough smoothness on v; see Durrett (1996) p. 130–141. We also
remark that the boundedness assumption can be relaxed to a polynomial growth
condition.
Another much used form of the Feynman–Kac formula says that
  Z t 
(t, x) 7→ Ex F (Wt ) exp − f (t − s, Ws ) ds
0
Z t  Z s  
+ g(t − s, Ws ) exp − f (t − v, Wv ) dv ds
0 0

is the unique solution of the problem

1
u0t (t, x) = u00xx (t, x) − f (t, x)u(t, x) + g(t, x), t > 0, x ∈ R,
2
u(0, x) = F (x), x ∈ R.

© Springer Basel 2015 106

A.N. Borodin, P. Salminen, Handbook of Brownian Motion - Facts and Formulae,
Probability and Its Applications, DOI 10.1007/978-3-0348-8163-0_6
 1. FEYNMAN{KAC FORMULA 107

For proofs we refer to Durrett (1996) p. 130–141, and Karatzas and Shreve
(1991) p. 268–270.
The issue 14(7)(1986) of Annals of Probability is dedicated to the memory of
Mark Kac. In particular, Kesten (1986) discusses the influence of Mark Kac on
probability theory.

2. In this chapter we present a number of differential systems for finding distri-

butions of various functionals associated to W. Our basic tool hereby is the
Feynman–Kac formula. We are mainly working in the time-homogeneous case,
and exploit the Feynman–Kac formula as presented for general diffusions in II.24.
This approach leads to ordinary differential problems which are more suitable
for analytical solutions than parabolic systems.
For a vector ~x = (x1 , x2 , . . . , xk ) and a non-negative piecewise continuous func-
tion f consider the random vector
Z t 
~ := R(f,
R ~ ~x) := f (Ws )ds, `(t, x1 ), `(t, x2 ), . . . , `(t, xk ) ,
0

where `(·, xi ) is the Brownian local time (with respect to Lebesgue measure) at
~ at certain stopping times.
the point xi . Our aim is to characterize the law of R
For this we introduce the additive functional
Z t k
X
t 7→ A(f, ~x, t) := f (Ws )ds + γi `(t, xi ), γi ≥ 0.
0 i=1

The functional A is considered under the side condition

C(a, b, t) :={a < inf Ws < sup Ws < b}

s≤t s≤t
={Ha,b > t}.

It is possible to generalize the treatment of A presented in the sections below

and to recursively use the Markov property to characterize the distribution of
non-time-homogeneous functionals of the type

n Z
X tj k
X 
A(f~n , ~x, ~tn ) := fj (Ws )ds + γi,j (`(tj , xi ) − `(tj−1 , xi )) ,
j=1 tj−1 i=1

where fj : R 7→ R+ are non-negative piecewise continuous functions, γi,j ≥ 0 and

0 = t0 < t1 ≤ · · · ≤ tn . However, this is fairly straightforward and is omitted.

3. We also comment below how the corresponding differential systems for a Brown-
ian motion with drift are obtained. There are two standard general approaches:
the first one is based on absolute continuity or the Cameron–Martin–Girsanov
formula (see III.26–28 and IV.30), and in the second W (µ) is considered as a
108 VI DIFFERENTIAL SYSTEMS

linear diffusion and the Laplace operator appearing in the differential equations
for Brownian motion is replaced by
1 d2 d
G (µ) = +µ .
2 dx2 dx

To demonstrate these two approaches in the Feynman–Kac formula consider the

time-homogeneous case and assume also that g ≡ 0. Then by absolute continuity
  Z t 
vµ (t, x) : = Ex F (Wt(µ) ) exp − f (Ws(µ) ) ds
0
  Z t 
= Ex F (Wt ) exp µ(Wt − x) − µ2 t/2 − f (Ws ) ds
0
  Z t 
= e−µx Ex eµ Wt F (Wt ) exp − (f (Ws ) + µ2 /2) ds
0
−µx
=: e v(t, x).
Therefore, v is the solution of the problem
 2
1 

u0t (t, x) = u00xx (t, x) − + f (x) u(t, x), t > 0, x ∈ R,
2 2
u(0, x) = eµx F (x), x ∈ R.

On the other hand, using the differential operator G (µ) it is seen that vµ is the
solution of the problem
1
u0t (t, x) = u00xx (t, x) + µ u0x (t, x) − f (x)u(t, x), t > 0, x ∈ R,
2
u(0, x) = F (x), x ∈ R.

We emphasize that the local time of W (µ) is taken with respect to Lebesgue
measure. As will be seen, this leads to seemingly simpler conditions than some
other normalizations.

2. Exponential stopping

In this section τ ∼ Exp(λ), independent of W, and Ŵ denotes W killed when

it exits the interval (a, b).

4. Let F be a piecewise continuous bounded function and let A := A(f, ~x, ·) be as

in No. 2. Then


(a) v(x) := Ex F (Wτ ) exp(−A(f, ~x, τ )); C(a, b, τ ) ,
is the solution of the problem
1 00
(b) u (x) − (λ + f (x))u(x) = −λF (x), x ∈ (a, b) \ {x1 , . . . , xk },
2
0 0
(c) u (xi +) − u (xi −) = 2γi u(xi ), i = 1, . . . , k,
(d) u(a+) = u(b−) = 0.
 2. EXPONENTIAL STOPPING 109

Remark. When we say (here and below in this chapter) that “v is the solution
of the problem” we mean that v is the unique continuous bounded function
which satisfies the ODE (b) and the conditions (c) and (d). If f and/or F have
discontinuity points, then (b) holds on every interval of continuity of f and F.
At the discontinuity points v 00 does not in general exist but v 0 exists and is
continuous.
To ensure that the function v given in (a) is the correct one, let p̃ and G̃λ be
the transition density and the Green function, respectively, of the process W̃
obtained from Ŵ by killing it (additionally) according to the additive functional
A (see II.22 and 23). The Green function G̃λ is given (see II.11) in terms of the
functions ψ̃λ and ϕ̃λ . These satisfy
1 00
(b’) u (x) − (λ + f (x))u(x) = 0 x ∈ (a, b) \ {x1 , . . . , xk },
2
and (c) and are such that ψ̃λ (a) = 0 and ϕ̃λ (b) = 0 (see II.24). Notice that
Z ∞ Z b Z b
v(x) = dt λe−λt 2dy p̃(t; x, y) F (y) = λ G̃λ (x, y)F (y) 2dy,
0 a a

and it can be proved by straightforward differentiation that v has the claimed

properties.
Next, replace W with W (µ) in (a), and let vµ be the corresponding function.
Then
Z b
(e) vµ (x) = λ G̃(λµ) (x, y)F (y) m(y) dy,
a

2µy
where m(y) = 2 e is the density of the speed measure. The Green function is
computed in the usual way. Notice that in this case the fundamental solutions
also satisfy (c) due to the normalization of `(µ) .
For a = −∞, say, the condition “u(a+) = 0” must be replaced by the condition
“u is bounded in a neighborhood of −∞”.

5. Especially, choose in No. 4

F (x) = 1I(−∞,z) (x), z ∈ (a, b).

Then
@
vz0 (x, z) := vz0 (x) = E (exp(−A(f, ~x, τ )); C(a, b, τ ), Wτ ≤ z)
@z x
= 2λG̃λ (x, z).

For a Brownian motion with drift the factor 2 above must be replaced by m(z) =
2e2µz (see (e) in No. 4). Recalling the definition of the Wronskian it is seen that
for both W and W (µ) and for z 6∈ {x1 , . . . , xk }
@ 0 @
v (z+, z) − vz0 (z−, z) = −2λ.
@x z @x
110 VI DIFFERENTIAL SYSTEMS

6. Here we characterize the first two moments of the functional A as solutions of a

differential system. For this let (cf. 1.3.0.1)


g(x) : = Px (τ ≤ Ha,b ) = 1 − Ex exp(−λHa,b )
ch((b + a − 2x) =2)
p
=1− .
ch((b − a) =2)
p

Then

v1 (x) := Ex A(f, ~x, τ ); C(a, b, τ )
is the solution of the problem (see Remark in No. 4)
1 00
u (x) − λu(x) = −f (x)g(x), x ∈ (a, b) \ {x1 , . . . , xk },
2
0 0
u (xi +) − u (xi −) = −2γi g(xi ), i = 1, 2, . . . , k,
u(a+) = u(b−) = 0.

To briefly motivate this, let Ĝ be the Green function of Ŵ . Then (see also (b)
in II.27) by the Markov property
Z τ  Z b
Ex f (Ws ) ds; τ < Ha,b = Ĝλ (x, y)f (y)g(y) 2dy,
0 a

and

Ex `(τ, xi ); τ < Ha,b = 2Ĝλ (x, xi )g(xi ),
explaining the above condition involving first derivatives.
The second moment is obtained by iteration. The result is that

v2 (x) := Ex A2 (f, ~x, τ ); C(a, b, τ )

is the solution of the problem

1 00
u (x) − λu(x) = −2f (x)v1 (x), x ∈ (a, b) \ {x1 , . . . , xk },
2
u0 (xi +) − u0 (xi −) = −4γi v1 (xi ), i = 1, 2, . . . , k,
u(a+) = u(b−) = 0.

In the case a = −∞ and/or b = ∞, some condition must be posed on f, e.g.,

that it does not grow faster than exponentially, to guarantee existence of the
expectations. Notice also that if a = −∞, say, then g above must be replaced
by its limit as a → −∞ :

lim Px (τ ≤ Ha,b ) = Px (τ ≤ Hb ).
a→−∞

7. Combining No’s. 5 and 6 shows that

@

v1 (x) := E A(f, ~x, τ ); C(a, b, τ ), Wτ ≤ z
@z x
 2. EXPONENTIAL STOPPING 111

is the solution of the problem (see Remark in No. 4)

1 00
u (x) − λu(x) = −2f (x)Ĝλ (x, z), x ∈ (a, b) \ {x1 , . . . , xk },
2
u0 (xi +) − u0 (xi −) = −4γi Ĝλ (xi , z), i = 1, 2, . . . , k,
u(a+) = u(b−) = 0.

Furthermore,

@
E A2 (f, ~x, τ ); C(a, b, τ ), Wτ ≤ z


v2 (x) :=
@z x

is the solution of the problem

1 00
u (x) − λu(x) = −2f (x)v1 (x), x ∈ (a, b) \ {x1 , . . . , xk },
2
0 0
u (xi +) − u (xi −) = −4γi v1 (xi ), i = 1, 2, . . . , k,
u(a+) = u(b−) = 0.

We remark that
Z b k
X
v1 (x) = Ĝλ (x, y) f (y) 2Ĝλ (y, z) 2dy + 4γi Ĝλ (x, xi )Ĝλ (xi , z).
a i=1

8. Let f : R+ 7→ R+ , f (0) = 0, be piecewise continuous and introduce

Z +∞
V (f, τ ) := f (`(τ, y)) dy.
−∞

Then using the Ray–Knight theorem in V.9 it can be proved (Borodin (1989b))
that for h > 0,
  Z h √ √
Ex exp(−V (f, τ )); sup `(τ, y) < h = 2λe− 2λl
v0 (l) v2 (l) dl,
y 0

where v0 is the solution of the problem

√
2x u00 (x) − 2 2λ x u0 (x) − f (x)u(x) = 0, x ∈ (0, h),
u(0) = 1, u(h) = 0,

and v2 is the solution of the problem

√
2xu00 (x) + (2 − 2 2λ x)u0 (x)
√ √
− ( 2λ + f (x))u(x) = − 2λv0 (x), x ∈ (0, h),
u(h) = 0.
112 VI DIFFERENTIAL SYSTEMS

Choosing f ≡ 0 here gives the distribution sup `(τ, y) (see 1.1.11.2 and Borodin
y
(1989b)).
p
For Brownian motion with drift use V.15 to obtain (γ := 2λ + µ2 )
 
Ex exp(−V (f, τ )); sup `(µ) (τ, y) < h
y
h h
 
Z Z
= e−γl
v0 (l)v2− (l) dl + e−γl v0 (l)v2+ (l) dl,
− 0
+ 0

where v0 is the solution of the problem

2x u00 (x) − 2γx u0 (x) − f (x)u(x) = 0, x ∈ (0, h),

u(0) = 1, u(h) = 0.

The functions v2+ and v2− are the solutions of the problems

2xu00 (x) + (2 − 2γx)u0 (x)

− (γ ± µ + f (x))u(x) = −(γ ± µ)v0 (x), x ∈ (0, h),
u(h) = 0,

respectively.

3. Stopping at first exit time

9. The function


v(x) := Ex exp(−A(f, ~x, Ha,b ))

is the solution of the problem

1 00
u (x) − f (x)u(x) = 0, x ∈ (a, b) \ {x1 , . . . , xk },
2
0 0
u (xi +) − u (xi −) = 2γi u(xi ), i = 1, . . . , k,
u(a+) = u(b−) = 1.

To explain this let W̃ , ϕ̃0 , and ψ̃0 be as in No. 4. Then

~ (x) '~0 (x)

v(x) = Px (W̃ζ− = a or b) = ~0 + ,
0 (b) '~0 (a)

and the result follows. From this it is also obvious that

~ (x )
x 7→ Ex exp(−A(f, ~x, Ha,b )); WHa,b = b = ~0


0 (b)
 3. STOPPING AT FIRST EXIT TIME 113

is the solution of the problem

1 00
u (x) − f (x)u(x) = 0, x ∈ (a, b) \ {x1 , . . . , xk },
2
0 0
u (xi +) − u (xi −) = 2γi u(xi ), i = 1, . . . , k,
u(a+) = 0, u(b−) = 1.

A similar statement holds for the function

'~ (x)
x 7→ Ex exp(−A(f, ~x, Ha,b )); WHa,b = a = 0 .
'~0 (a)

10. It follows from No. 9 that for c ∈ (a, b),

x 7→ Ex exp(−A(f, ~x, Hc )); C(a, b, Hc )

is the solution of the problem

1 00
u (x) − f (x)u(x) = 0, x ∈ (a, b) \ {c, x1 , . . . , xk },
2
0 0
u (xi +) − u (xi −) = 2γi u(xi ), i = 1, . . . , k,
u(a+) = u(b−) = 0, u(c) = 1.

11. We characterize the first two moments of A(f, ~x, Ha,b ) as solutions to differential
problems. The first moment


v(x) := Ex A(f, ~x, Ha,b )

is the solution of the problem

1 00
u (x) = −f (x), x ∈ (a, b) \ {x1 , . . . , xk },
2
0 0
u (xi +) − u (xi −) = −2γi , i = 1, . . . , k,
u(a+) = u(b−) = 0.

The second moment is obtained by iteration. The result is that

x 7→ Ex A2 (f, ~x, Ha,b )

is the solution of the problem

1 00
u (x) = −2f (x) v(x), x ∈ (a, b) \ {x1 , . . . , xk },
2
0 0
u (xi +) − u (xi −) = −4γi v(xi ), i = 1, 2, . . . , k,
u(a+) = u(b−) = 0.
 114 VI DIFFERENTIAL SYSTEMS

For an explanation of these, see II.27 and recall (cf. II.11) that the Green kernel
is given by
(b − x)(y − a)
G0 (x, y) := , x ≥ y.
b−a

12. Next we consider the first two moments of the functional A(f, ~x, Ha,b ) under the
side condition WHa,b = b. Recall that

x−a
g(x) := Px (WHa,b = b) = Px Hb < Ha = .
b−a

Then

v(x) := Ex A(f, ~x, Ha,b ); WHa,b = b
is the solution of
1 00
u (x) = −f (x)g(x), x ∈ (a, b) \ {x1 , . . . , xk },
2
0 0
u (xi +) − u (xi −) = −2γi g(xi ), i = 1, . . . , k,
u(a+) = u(b−) = 0.

Further,
x 7→ Ex A2 (f, ~x, Ha,b ); WHa,b = b

is the solution of the problem

1 00
u (x) = −2f (x) v(x), x ∈ (a, b) \ {x1 , . . . , xk },
2
0 0
u (xi +) − u (xi −) = −4γi v(xi ), i = 1, . . . , k,
u(a+) = u(b−) = 0.

13. Let f be as in No. 8 and introduce

Z +∞
V (f, Hz ) := f (`(Hz , y)) dy.
−∞

For h > 0 and x < z consider the function

 
v(x, z) := Ex exp(−V (f, Hz )); sup `(Hz , y) < h .
y

By the spatial homogeneity of Brownian motion v depends on x and z only

through the difference z − x. Let v(u) := v(x, z), where u = z − x. Then using
the Ray–Knight theorem in V.6 it can be proved (Borodin (1989b), see also
Salminen and Vallois (2005) and cf. No. 23) that
∞ h
1
Z Z
−αu
(a) e v(u) du = G̃α (0, y)v0 (y) dy,
0 0
2
 3. STOPPING AT FIRST EXIT TIME 115

where   Z ∞  
v0 (y) := Ey exp − f (Zs(0) )ds ; sup Zs(0) < h
0 s≥0
(0)
and Z is a squared 0-dimensional Bessel process. Therefore, v0 is the solution
of the problem
2x u00 (x) − f (x)u(x) = 0, x ∈ (0, h),
u(0) = 1, u(h) = 0.

The Green function G̃ appearing in (a) is associated to the process Z̃ (2) obtained
dx
from the squared 2-dimensional Bessel process Z (2) (recall that is the speed
2
measure of Z (2) ) killed when it hits h and which is also killed according to the
additive functional Z t
t 7→ A(f, t) := f (Zs(2) ) ds.
0

In other words, Z̃ (2) is defined via

 (2)
(2)
Zt , t < ζ ∧ Hh ,
Z̃t =
∂, t ≥ ζ ∧ Hh ,
where
Hh := inf{t : Zt(2) = h} and ζ := inf{t : A(f, t) > τ }
with τ ∼ Exp(1), independent of Z (2) .

14. Consider for h > 0 and a < 0 < b the function

 
vb (a, b) := E0 exp(−V (f, Ha,b )); sup `(Ha,b , y) < h, WHa,b = b ,
y

where f is as in No. 8 and

Z +∞
V (f, Ha,b ) := f (`(Ha,b , y)) dy.
−∞

Using the Ray–Knight theorem in V.8 it can be proved that (Borodin 1989b)
Z ∞
1 h
Z 0 Z
db e−αb da eβa vb (a, b) = G̃α (0, y)v0 (y) dy,
0 −∞
2 0

where G̃ is as in No. 13 and v0 is the solution of the problem

2x u00 (x) − (β + f (x))u(x) = 0, x ∈ (0, h),
u(0) = 1, u(h) = 0.
The function
 
va (a, b) := E0 exp(−V (f, Ha,b )); sup `(Ha,b , y) < h, WHa,b = a
y

has a similar representation with the roles of α and β changed.

 116 VI DIFFERENTIAL SYSTEMS

4. Stopping at inverse additive functional

15. Let g be a non-negative piecewise continuous function and introduce

Z t
A(g, t) := g(Ws )ds.
0

For t > 0 we consider the stopping time ν(t) := inf s : A(g, s) > t . Let
τ ∼ Exp(λ) independent of W. Then the function


v(x) := Ex exp(−A(f, ~x, ν(τ ))); C(a, b, ν(τ ))

is the solution of the problem

1 00
u (x) − (λg(x) + f (x))u(x) = −λg(x), x ∈ (a, b) \ {x1 , . . . , xk },
2
0 0
u (xi +) − u (xi −) = 2γi u(xi ), i = 1, . . . , k,
u(a+) = u(b−) = 0.

To briefly motivate this, let Ŵ be as in No. 4 and construct W̄ from Ŵ by

killing it (see II.22) according to the additive functional

t 7→ A(f, ~x, t) + λA(g, t).

Hence
k
X
k(dy) = 2(f (y) + λg(y))dy + 2γi ε{xi } (dy)
i=1

is the killing measure of W̄ . Let Ḡ and ζ̄ be the Green function (with respect to
2dx) and the lifetime, respectively, of W̄ . Then (see II.26)
Z b
v(x) = Px (ζ̄ = ν(τ )) = λ Ḡ0 (x, z)g(z)2dz,
a

and the result follows from No. 4.

16. For t ≥ 0 and z ∈ (a, b) let %(t, z) := inf{s : `(s, z) > t} be the inverse local
time and let τ be as in No. 15. Then


v(x) := Ex exp(−A(f, ~x, %(τ, z))); C(a, b, %(τ, z))

is the solution of the problem

1 00
u (x) − f (x)u(x) = 0, x ∈ (a, b) \ {z, x1 , . . . , xk },
2
0 0
u (z+) − u (z−) = 2λu(z) − 2λ,
u0 (xi +) − u0 (xi −) = 2γi u(xi ), i = 1, . . . , k,
u(a+) = u(b−) = 0,
 4. STOPPING AT INVERSE ADDITIVE FUNCTIONAL 117

where it is assumed that z 6= xi , i = 1, 2, . . . , k. If z = xi for some i, then

u0 (z+) − u0 (z−) = 2(λ + γi )u(z) − 2λ.

This can be ensured similarly as the result in No. 15.

17. Consider the functional in No. 16 for Brownian motion with drift W (µ) . Due to
transience of W (µ) it may happen that %(µ) (τ, z) = ∞, and if a = −∞ and µ < 0,
for instance, this appears in the formula. Indeed, set C(b, ·) := C(−∞, b, ·) and
%(µ) := %(µ) (τ, z). Then

v(x) :=Ex exp(−A(µ) (f, ~x, %(µ) )); C(b, %(µ) )

=Ex exp(−A(µ) (f, ~x, %(µ) )); C(b, %(µ) ), %(µ) < ∞

+ Ex exp(−A(µ) (f, ~x, ∞)); C(b, ∞), %(µ) = ∞

= : v1 (x) + v2 (x).

Let W̄ (µ) be the process obtained from W (µ) by killing at Hb and also according
to the additive functional

t 7→ A(µ) (f, ~x, t) + λ`(µ) (t, z).

Then if z 6= xi , i = 1, . . . , k,
(µ)
v1 (x) = Px (W̄ζ− = z) = λm(z)Ḡ(0µ) (x, z),

and
'0 (x)
v2 (x) = Px ( lim W̄t(µ) = −∞) = lim ,
t→∞ y→−∞ '0 (y)
where m(z) = 2e2µz is the density of the speed measure and ϕ̄0 is the non-
increasing fundamental solution of the system
1 00
u (x) + µu0 (x) − f (x)u(x) = 0, x ∈ (−∞, b) \ {z, x1 , . . . , xk },
2
0 0
u (z+) − u (z−) = 2λu(z),
u0 (xi +) − u0 (xi −) = 2γi u(xi ), i = 1, . . . , k,
u(b) = 0.

18. The Laplace transform v in No. 16 can be inverted explicitly. Analytically

this is due to the fact that λ appears in a simple form only in one of the side
conditions. However, there is also a more probabilistic explanation. For this let
%̃(·, z) be the inverse local time at z of the process W̃ obtained from Ŵ by killing
it according to the additive functional A(f, ~x, ·) (cf. No. 4). Then, letting ζ̃ be
the lifetime of W̃ , we have

Pz (%̃(t, z) < ∞) = Pz (%̃(t, z) < ζ̃)

= Ez exp(−A(f, ~x, %(t, z))); C(a, b, %(t, z)) .
 118 VI DIFFERENTIAL SYSTEMS

From II.14 we have

t
 
Pz (%̃(t, z) < ∞) = exp − ~ ,
2G0 (z; z )

where G̃ is the Green function of W̃ . Let H̃ be the first hitting time functional
for W̃ . Then, by the strong Markov property,

Px (%̃(t, z) < ∞) = Px (H̃z < ∞, %̃(t, z) < ∞)

= Px (H̃z < ∞)Pz (%̃(t, z) < ∞).

Because  ~
0 (x)
 ~0 (z ) ,

 x ≤ z,
Px (H̃z < ∞) =
 '~0 (x) ,

x ≥ z,

'~0 (z )
we have done the promised inversion. Here, ψ̃0 and ϕ̃0 are the fundamental
solutions of the differential system associated to W̃ .
The description is very similar for W (µ) . In this case we have
t
 
Pz (%̃(µ) (t, z) < ∞) = exp − ,
m(z )G~ (0) (z; z )

where G̃(µ) is the Green function of the diffusion obtained from W (µ) similarly
as W̃ is constructed from W.
Observe that {%(µ) (τ, z) = ∞} = {`(µ) (∞, z) < τ } and so the function v2 in No.
17 can be written as

v2 (x) = Ex exp(−A(µ) (f, ~x, ∞) − λ`(µ) (∞, z)); C(b, ∞) .

In some cases this can be inverted to find an expression for

Ex exp(−A(µ) (f, ~x, ∞)); C(b, ∞), %(µ) (t, z) = ∞ .

19. The result in No. 16 can be generalized to stopping times of the type

%̂(t1 , t2 , . . . , tn ) := %(t1 , z1 ) ∧ %(t2 , z2 ) ∧ · · · ∧ %(tn , zn ).

Consider the case n = 2, and let τ1 and τ2 be two independent exponentially

distributed random variables (with parameters λ1 and λ2 , respectively,) inde-
pendent of W. Then in the case z1 , z2 6∈ {x1 , . . . , xk } and z1 6= z2 ,


v(x) :=Ex exp(−A(f, ~x, %̂(τ1 , τ2 ))); C(a, b, %̂(τ1 , τ2 ))
=Px (W̄ζ− = z1 or z2 )
=2λ1 Ḡ0 (x, z1 ) + 2λ2 Ḡ0 (x, z2 ),
 4. STOPPING AT INVERSE ADDITIVE FUNCTIONAL 119

where Ḡ is the Green function of the process W̄ obtained from Ŵ by killing

according to the additive functional
t 7→ A(f, ~x, t) + λ1 `(t, z1 ) + λ2 `(t, z2 ).

20. Analogously to the generalization in No. 19, we can generalize the result in No.
15. Consider the case n = 2 and let ν̂(t1 , t2 ) := ν1 (t1 ) ∧ ν2 (t2 ), where
 Z s 
νj (t) := inf s : gj (Wv )dv > t , j = 1, 2,
0

and gj , j = 1, 2, are non-negative piecewise continuous functions. Then

v(x) := Ex exp(−A(f, ~x, ν̂(τ1 , τ2 ))); C(a, b, ν̂(τ1 , τ2 ))
is the solution of the problem
1 00  
u (x) − λ1 g1 (x) + λ2 g2 (x) + f (x) u(x)
2
= −λ1 g1 (x) − λ2 g2 (x), x ∈ (a, b) \ {x1 , . . . , xk },
0 0
u (xi +) − u (xi −) = 2γi u(xi ), i = 1, . . . , k,
u(a+) = u(b−) = 0.

21. Consider for h > 0 the function

 
v(x) := Ex exp(−V (f, %(t, z))); sup `(%(t, z), y) < h ,
y

where f is as in No. 8 and

Z +∞
V (f, %(t, z)) := f (`(%(t, z), y)) dy.
−∞

This generalizes in a sense the result in No. 13 and we use the notation therein.
By the Ray–Knight theorem in V.10 we have
v(x) = v0 (t) v2 (t, |z − x|), t < h,
where v0 is as in No. 13 and
  Z y  
(2)
v2 (t, y) := Et v0 (Zy ) exp − f (Zu(2) )du ; sup Zu(2) < h .
0 u≤y

Then for G̃ as in No. 13,

Z ∞ h
1
Z
e−αy v2 (t, y) dy = G̃α (t, u) v0 (u) du.
0 0
2

In the case z = x we simply have

 
Ex exp(−V (f, %(t, x))); sup `(%(t, x), y) < h = v02 (t), t < h.
y
 120 VI DIFFERENTIAL SYSTEMS

5. Stopping at first range time

22. Let r > 0 and consider the first range time

n o
θr := inf t : sup Ws − inf Ws = r .
0≤s≤t 0≤s≤t

For results on first range time see Feller (1951), Imhof (1985), Vallois (1993),
(1995), Borodin (1999), and Chong, Cowan and Holst (2000), and Salminen
and Vallois (2005). We have the following fundamental identity which holds for
z ≤ x ≤ z + r,

@ @
 
(a) Px (θr ∈ dt, Wθr ≤ z) = Px Hz ∈ dt, sup Ws − inf Ws ≤ r .
@z @r 0≤s≤t 0≤s≤t

Clearly,
 
Px Hz ∈ dt, sup Ws − inf Ws ≤ r = Px (Hz ∈ dt, Hz+r > t).
0≤s≤t 0≤s≤t

Using this in (a) and integrating over t gives for x − r ≤ z ≤ x,

@ @ @ z+r−x x−z
P (W ≤ z) = Px (Hz < Hz+r ) = = 2 ,
@z x θr @r @r r r

and, further, by symmetry

|z − x|
Px (Wθr ∈ dz) = dz, |z − x| ≤ r.
r2

The identity (a) can be generalized to (see Borodin (1999))

@


E exp −A(f, ~x, θr ) ; θr ∈ dt, Wθr < z
@z x
@


= Ex exp −A(f, ~x, Hz ) ; Hz ∈ dt, Hz+r > t ,
@r

where z < x < z + r. Integrating over t and using the results and the notation
from No. 9 (with a = z and b = z + r) lead to

@

@ '~0 (x)
(b) Ex exp −A(f, ~x, θr ) ; Wθr < z = .
@z @r '~0 (z )

Analogously, for z − r < x < z (now a = z − r and b = z in No. 9)

@

@ ~0 (x)
(c) Ex exp −A(f, ~x, θr ) ; Wθr < z = .
@z @r ~0 (z )

Notice that the functions ϕ̃0 and ψ̃0 depend on r via the boundary conditions.
It is possible to write the right-hand sides of (b) and (c) in a form, where the
 5. STOPPING AT FIRST RANGE TIME 121

dependence on r is more explicit. To explain this assume that f 6≡ 0 and/or

γi 6= 0 for some i. Then the functions ϕ̃0 and ψ̃0 can be expressed as

ϕ̃0 (x) = ψ(z + r)ϕ(x) − ψ(x)ϕ(z + r),

ψ̃0 (x) = ψ(x)ϕ(z − r) − ψ(z − r)ϕ(x),

where ϕ := ϕ0 and ψ := ψ0 are fundamental solutions of the system

1 00
u (x) − f (x)u(x) = 0, x ∈ R \ {x1 , . . . , xk },
2
0 0
u (xi +) − u (xi −) = 2γi u(xi ), i = 1, . . . , k.

Using these expressions we can, in fact, compute the derivatives in (b) and (c).
To do this, define
ρ(x, y) := ψ(x)ϕ(y) − ψ(y)ϕ(x).
Then for any a, b and c,

@ @
ρ(b, c) ρ(a, b) − ρ(a, b) ρ(b, c) = −ωρ(a, c),
@b @b

where
ω := ψ 0 (x)ϕ(x) − ψ(x)ϕ0 (x)
is a constant (Wronskian). Consequently, for z < x < z+r (the case z−r < x < z
is left to the reader),

@ '~0 (x) @ (z + r; x) @ (x; z + r) !(x; z )

= =− = 2 ,
@r '~0 (z ) @r (z + r; z ) @r (z + r; z )  (z + r; z )

and we have the formulae

!(x; z )

 2 (z + r; z ) , x − r < z ≤ x,
@




E exp −A(f, ~x, θr ) ; Wθr < z =
@z x  !(z; x) ,

x ≤ z < x + r,
2 (z; z − r)

and

F (z )(z; x) + F (z − r)(x; z − r)



Z x+r
Ex F (Wθr ) exp −A(f, ~x, θr ) = ω dz,
x
2 (z; z − r)

where F is measurable and bounded. The corresponding formulae for Brownian

motion with drift can be derived using absolute continuity.
23. Let f : R+ 7→ R+ , f (0) = 0, be piecewise continuous and consider
Z +∞
V (f, θr ) := f (`(θr , y)) dy.
−∞
122 VI DIFFERENTIAL SYSTEMS

Using the Ray–Knight theorem for Brownian motion stopped at θr (see V.7) we
obtain (see Borodin (1999) – also for Brownian motion with drift – and Salminen
and Vallois (2005))
Z ∞   Z h
e−αr Ex exp(−V (f, θr )); sup `(θr , y) < h dr = y G̃α (0, y) Ĝα (0, y) dy,
0 y 0

where G̃α is as in No. 13 and the Green function Ĝα is associated to the process
Z̃ (4) obtained from the squared Bessel process Z (4) killed when it hits h and
which is also killed according to the additive functional
Z t
t 7→ f (Zs(4) ) ds.
0
 APPENDIX 1

BRIEFLY ON SOME DIFFUSIONS

In this appendix we give basic characteristics and other facts and formulae as-
sociated with some diffusions appearing frequently in the literature – and in our
own text. This collection may also serve as a sample of the examples highlighting
different aspects in the general theory of linear diffusions.
Diffusions below are considered in the canonical framework. In particular, sam-
ple paths are continuous functions t 7→ ω(t) taking values in R up to the lifetime.
See No. I.12.
Transition densities are given with respect to speed measures. This normaliza-
tion gives us symmetric densities (cf. No. II.4). To get the density with respect
to Lebesgue measure our densities must be multiplied by the speed density (with
respect to Lebesgue measure). See also No. II.9.
The Green functions are obtained from the transition densities, and vice versa.
We present below the Green functions in the form

Gα (x, y) = wα−1 ϕα (x)ψα (y), x ≥ y.

By symmetry, this description is sufficient. Notice also that with this notation
the decreasing fundamental solution is always a function of x, and the increasing
one is always a function of y. Recall from No. II.10 the connection with the
hitting times and the fundamental solutions.
For properties of special functions, Laplace transforms, and inverse Laplace
transforms we refer to Abramowitz and Stegun (1970), and Erdelyi and et al.
(1953), (1954). At the end of the formula part of this book there is a list of
properties of some special functions (Appendix 2) appearing in the tables and
in this appendix and also there is a collection of the most used formulae for the
inverse Laplace transforms (Appendix 3).

1. Brownian motion

Speed measure: m(dx) = 2 dx.

Scale function: s(x) = x.
1 d2 f
Generator: Gf = .
2 dx2
Domain: D = {f : f, Gf ∈ Cb (R)}.

123
124 APPENDIX 1
√ √
Green function: Gα (x, y) = wα−1 e−x 2α y 2α
e , x ≥ y.
√
Wronskian: wα = 2 2α.
1 
(x − y )2 
Transition density w.r.t. m: p(t; x, y) = √ exp − .
2 2t 2t
Spectral representation of the transition density:
∞
1
Z
2
e−λ t/2


p(t; x, y) = cos(λx) cos(λy) + sin(λx) sin(λy) dλ
4 −∞
∞
1
Z
2
= e−λ t/2
cos(λ(x − y)) dλ.
2 0

2. Brownian motion on [a, ∞) reflected at a

Speed measure: m(dx) = 2 dx.

Scale function: s(x) = x.
1 d2 f 1
Generator: Gf = , x > a, Gf (a) = f 00 (a+).
2 dx2 2
Domain: D = {f : f, Gf ∈ Cb ([a, ∞)), f 0 (a+) = 0}.
√ √
Green function: Gα (x, y) = wα−1 e−(x−a) 2α
ch((y − a) 2α) , x ≥ y ≥ a.
√
Wronskian: wα = 2α.
Transition density w.r.t. m:

1   (x − y)2  
(x + y − 2a)2 
p(t; x, y) = √ exp − + exp − .
2 2t 2t 2t

Spectral representation of the transition density:

∞
1
Z
2
p(t; x, y) = e−λ t/2
cos(λ(x − a)) cos(λ(y − a)) dλ.
 0

Let W be a Brownian motion process. Then {|Wt − a| + a : t ≥ 0} is the

Brownian motion on [a, ∞) reflected at a.

3. Brownian motion on (a, ∞) killed at a

Speed measure: m(dx) = 2 dx.

Scale function: s(x) = x.
1 d2 f
Generator: Gf = , x > a.
2 dx2
Domain: D = {f : f, Gf ∈ Cb ((a, ∞)), f (a+) = 0}.
 BRIEFLY ON SOME DIFFUSIONS 125
√ √
Green function: Gα (x, y) = wα−1 e−(x−a) 2α
sh((y − a) 2α), x ≥ y > a.
√
Wronskian: wα = 2α.
Transition density w.r.t. m:

1   (x − y)2  
(x + y − 2a)2 
p(t; x, y) = √ exp − − exp − .
2 2t 2t 2t

Spectral representation of the transition density:

1 ∞ −λ2 t/2
Z
p(t; x, y) = e sin(λ(x − a)) sin(λ(y − a)) dλ.
 0

4. Brownian motion on [a, ∞) absorbed at a

Speed measure: m(dx) = 2 dx.

Scale function: s(x) = x.
1 d2 f 1
Generator: Gf = , x > a, Gf (a) = f 00 (a+).
2 dx2 2
Domain: D = {f : f, Gf ∈ Cb ([a, ∞)), f 00 (a+) = 0}.
√ √
Green function: Gα (x, y) = wα−1 e−(x−a) 2α
sh((y − a) 2α), x ≥ y > a.
√
Wronskian: wα = 2α.
1
Resolvent at a: Rα f (a) = f (a).
Transition density w.r.t. m and absorption probability:

1   (x − y)2  
(x + y − 2a)2 
p(t; x, y) = √ exp − − exp − , x > a, y > a,
2 2t 2t 2t

x−a
(  
Erfc √ , x > a,
Px (ω(t) = a) = 2t
1, x = a.

5. Brownian motion on [a, b] reflected at a and b

Speed measure: m(dx) = 2 dx.

Scale function: s(x) = x.
1 d2
Generator: Gf = , a < x < b,
2 dx2
1 1
Gf (a) = f 00 (a+), Gf (b) = f 00 (b−).
2 2
Domain: D = {f : f, Gf ∈ Cb ([a, b]), f 0 (a+) = f 0 (b−) = 0}.
126 APPENDIX 1

Green function:
√ √
Gα (x, y) = wα−1 ch((b − x) 2α) ch((y − a) 2α), b ≥ x ≥ y ≥ a.

√ √
Wronskian: wα = 2α sh((b − a) 2α).
Transition density w.r.t. m:
∞ 
1 X 
(x − y + 2n(b − a))2 
p(t; x, y) = √ exp −
2 2t 2t
n=−∞

(x + y − 2a + 2n(b − a))2 
+ exp − .
2t
Spectral representation of the transition density:
∞
1 1 X n2 2 n n
     
p(t; x, y) = + exp − 2 t cos (x − a) cos (y − a) .
b−a 2 2(b − a) b−a b−a
n=1

6. Brownian motion on (a, b) killed at a or b

Speed measure: m(dx) = 2 dx.

Scale function: s(x) = x.
1 d2 f
Generator: Gf = , a < x < b.
2 dx2
Domain: D = {f : f, Gf ∈ Cb ((a, b)), f (a+) = f (b−) = 0}.
Green function:
√ √
Gα (x, y) = wα−1 sh((b − x) 2α) sh((y − a) 2α), b>x≥y>a.

√ √
Wronskian: wα = 2α sh((b − a) 2α).
Transition density w.r.t. m:
∞ 
1 X 
(x − y + 2n(b − a))2 
p(t; x, y) = √ exp −
2 2t 2t
n=−∞

(x + y − 2a + 2n(b − a))2 
− exp − .
2t

Spectral representation of the transition density:

∞
1 X n2 2 n n
     
p(t; x, y) = exp − 2 t sin (x − a) sin (y − a) .
b−a 2(b − a) b−a b−a
n=1
 BRIEFLY ON SOME DIFFUSIONS 127

7. Brownian motion killed elastically at 0

Speed measure: m(dx) = 2 dx.

Scale function: s(x) = x.
Killing measure: κ(dx) = γ ε{0} (dx), γ > 0.
1 d2 f 1 1
Generator: Gf = , x 6= 0, Gf (0) = f 00 (0+) = f 00 (0−).
2 dx2 2 2
Domain: D = {f : f, Gf ∈ Cb (R), f 0 (0+) − f 0 (0−) = γf (0)}.
Green function:
  √ √  √
wα−1 e−x 2α − √ sh(x 2α) ey 2α , 0 ≥ x ≥ y,
2



 √ √
Gα (x, y) = wα−1 e−x 2α ey 2α , x ≥ 0 ≥ y,

 √  √ √ 
wα−1 e−x 2α √ sh(y 2α) + ey 2α
x ≥ y ≥ 0,

 ,
2

1  −|x−y|√2α √ 
= √ e − e−(|x|+|y|) 2α .
2 2 w
√
Wronskian: wα = 2 2α + γ.
Transition density w.r.t. m:
1 
(x − y)2 
p(t; x, y) = √ exp −
2 2t 2t

|x| + |y | 2t  
2|x| + 2|y| + t 
− exp + Erfc √ .
8 2 8 2 2t

8. Brownian motion sticky at 0

Speed measure: m(dx) = 2 dx + 2c ε{0} (dx), c > 0.

Scale function: s(x) = x.
1 d2 f 1 1
Generator: Gf = , x 6= 0, Gf (0) = f 00 (0+) = f 00 (0−).
2 dx2 2 2
Domain: D = {f : f, Gf ∈ Cb (R), c f 00 (0+) = f 0 (0+) − f 0 (0−)}.
Green function:

−1
 √
−x 2α
√ √  y√2α


 wα e − c 2α sh(x 2α e , 0 ≥ x ≥ y,
 √ √
Gα (x, y) = wα−1 e−x 2α ey 2α , x ≥ 0 ≥ y,

 √  √ √ √ 
wα−1 e−x 2α c 2α sh(y 2α) y 2α
x ≥ y ≥ 0,

 +e ,

1  −|x−y|√2α 2 c −(|x|+|y|)√2α 
= √ e − e .
2 2 w
128 APPENDIX 1
√
Wronskian: wα = 2 2α + 2α c.
Transition density w.r.t. m:
1   (x − y)2  
(|x| + |y|)2 
p(t; x, y) = √ exp − − exp −
2 2t 2t 2t
√ 
1 
2|x| + 2|y| 2t  
|x| + |y | 2t
+ exp + 2 Erfc √ + .
2c c c 2t c

For results on sticky Brownian motion see Harrison and Lemoine (1981), Amir
(1991), Chitashvili (1997), and Warren (1997).

9. Brownian motion sticky at 0 and killed elastically at 0

Speed measure: m(dx) = 2 dx + 2c ε{0} (dx), c > 0.

Scale function: s(x) = x.
Killing measure: κ(dx) = γ ε{0} (dx), γ > 0.
1 d2 f 1 1
Generator: Gf = , x 6= 0, Gf (0) = f 00 (0+) = f 00 (0−).
2 dx2 2 2
Domain: D = {f : f, Gf ∈ Cb (R), c f 00 (0+) = f 0 (0+) − f 0 (0−) − γf (0)}.
Green function:

−1
 √
−x 2α 2 c+ √  y√2α
w e − √ sh(x 2α e , 0 ≥ x ≥ y,
α
2



 √ √
Gα (x, y) = wα−1 e−x 2α ey 2α , x ≥ 0 ≥ y,
√ 2 c+ √ √

  
wα−1 e−x 2α sh(y 2α) + ey 2α
x ≥ y ≥ 0,

 √ ,
2

1  −|x−y|√2α −(|x|+|y|)√2α  1 −(|x|+|y|)√2α

= √ e −e + e .
2 2 w
√
Wronskian: wα = γ + 2 2α + 2α c.
Transition density w.r.t. m:
1   (x − y)2  
(|x| + |y|)2 
p(t; x, y) = √ exp − − exp −
2 2t 2t 2t
2 √
 +1 (|x| + |y|) ( + 1) t |x| + | y | (1 +  ) t 
  
+ exp + Erfc √ + √
4c 1− 2c2 2t c 2
√
 −1

(|x| + |y|) ( − 1)2 t  
|x| + | y | (1 −  ) t 
+ exp + Erfc √ + √ ,
4c 1+ 2c2 2t c 2
√
where Υ = 1 − γc. The epression for the density is valid for all non-negative
values on c and γ; also when γc > 1. We remark that the densities in No. 7
and No. 8 can be obtained from the density above by letting c → 0 and γ → 0,
respectively.
 BRIEFLY ON SOME DIFFUSIONS 129

10. Reflecting Brownian motion on [0, ∞) killed elastically at 0

Speed measure: m(dx) = 2 dx.

Scale function: s(x) = x.
Killing measure: k(dx) = γ ε{0} (dx), γ > 0.
1 d2 f 1
Generator: Gf = , x > 0, Gf (0) = f 00 (0+).
2 dx2 2
Domain: D = {f : f, Gf ∈ Cb ([0, ∞)), f 0 (0+) = γf (0)}.
Green function:
√  √ √ 
Gα (x, y) = wα−1 e−x 2α
ch(y 2α) + √ sh(y 2α) , x ≥ y ≥ 0.
2
√
Wronskian: wα = 2α + γ.
Transition density w.r.t. m:
1   (x − y)2  
(x + y)2 
p(t; x, y) = √ exp − + exp −
2 2t 2t 2t
 2t  
x+y+ t

− exp γ(x + y) + Erfc √ , x ≥ 0, y ≥ 0.
2 2 2t

11. Reflecting Brownian motion on [0, ∞) sticky at 0 and

killed elastically at 0

Speed measure: m(dx) = 2 dx + 2c ε{0} (dx), c > 0.

Scale function: s(x) = x.
Killing measure: κ(dx) = γ ε{0} (dx), γ > 0.
1 d2 f 1
Generator: Gf = , x > 0, Gf (0) = f 00 (0+).
2 dx2 2
Domain: D = {f : f, Gf ∈ Cb ([0, ∞)), c f 00 (0+) = f 0 (0+) − γf (0)}.
Green function:
√  √ 2 c+ √ 
Gα (x, y) = wα−1 e−x 2α ch(y 2α) + √ sh(y 2α) , x ≥ y ≥ 0.
2
√
Wronskian: wα = γ + 2α + 2αc.
Transition density w.r.t. m:
1   (x − y)2  
(x + y)2 
p(t; x, y) = √ exp − − exp −
2 2t 2t 2t
√ 
 +1

(x + y)2 ( + 1)2 t  
x + y (1 +  ) t
+ exp + Erfc √ + √
4c 1− 8c2 2t 2c 2
√ 
 −1

(x + y)2 ( − 1)2 t  
x + y (1 −  ) t
+ exp + Erfc √ + √ ,
4c 1+ 8c2 2t 2c 2
√
where Υ = 1 − 4γc and x ∧ y ≥ 0. The epression for the density is valid for all
non-negative values on c and γ; also when 4γc > 1.
130 APPENDIX 1

12. Brownian motion skew at zero

This process is usually indexed by one parameter β ∈ (0, 1) (called the skew-
ness parameter of the process). Skew Brownian motion behaves like ordinary
Brownian motion when off the origin but if, e.g., β > 1/2, it has more tendency
to move up than down from the origin. In particular, letting {ω(t) : t ≥ 0} de-
note a skew Brownian motion starting at 0 we have for every t > 0 the following
property
P0 (ω(t) ≥ 0) = 1 − P0 (ω(t) ≤ 0) = β.

Speed measure:

4β dx, x > 0,
m(dx) =
4(1 − β) dx, x < 0.

Scale function:  x
, x ≥ 0,

2
s(x) = x
, x ≤ 0.
2(1 − )


1 d2 f 1 1
Generator: Gf = , x 6= 0, Gf (0) = f 00 (0+) = f 00 (0−).
2 dx2 2 2
Domain: D = {f : f, Gf ∈ Cb (R), (1 − β)f 0 (0−) = βf 0 (0+)}.
Green function:

−1 1 − 2
  √ √ 
−x 2α
√
w sh(x 2α) + e ey 2α , 0 ≥ x ≥ y,
α 1−



 √ √
Gα (x, y) = wα−1 e−x e , 2α y 2α
x ≥ 0 ≥ y,
 w−1 e−x 2α 1 − 2 sh(y √2α) + ey 2α ,

 √  √ 
x ≥ y ≥ 0,

α

√ √ √
e−|x−y| 2 − e−(|x|+|y|) 2 e−(|x|+|y|) 2
= √ + √ .
2 2 (1 + (2 − 1) sgn(x ∧ y)) 2 2

√
Wronskian: wα = 2 2α.
Transition density w.r.t. m:

e−(x−y) =2t − e−(|x|+|y|) =2t e−(|x|+|y|) =2t

2 2 2

p(t; x, y) = √ + √ .
2 2t(1 + (2 − 1) sgn(x ∧ y)) 2 2t

For further results concerning skew Brownian motion see Walsh (1978), Harrison
and Shepp (1981), and Barlow (1988).
 BRIEFLY ON SOME DIFFUSIONS 131

13. Brownian motion skew at 0, sticky at 0 and

killed elastically at 0

Speed measure: m(dx) = 2(1 + (2β − 1) sgn x) dx + 2c ε{0} (dx), c > 0.

x
Scale function: s(x) = .
1 + (2 − 1) sgn x
Killing measure: κ(dx) = γ ε{0} (dx), γ > 0.
1 d2 f 1 1
Generator: Gf = , x 6= 0, Gf (0) = f 00 (0+) = f 00 (0−).
2 dx2 2 2
Domain:

D = {f : f, Gf ∈ Cb (R), c f 00 (0+) = 2βf 0 (0+) − 2(1 − β)f 0 (0−) − γf (0)}.

Green function:
 
w
 √ √  √
wα−1 2 − √ sh(x 2α) + e−x 2α ey 2α , 0 ≥ x ≥ y,
2(1 − ) 2



√ √

Gα (x, y) = wα e−1 −x 2α
ey 2α , x ≥ 0 ≥ y,

 √  w
 w−1 e−x 2α
 √ √ 
− 2 sh(y 2α) + ey 2α , x ≥ y ≥ 0,

√
α
2 2
√ √ √
e−|x−y| 2 − e−(|x|+|y|) 2 e−(|x|+|y|) 2
= √ + .
2 2 (1 + (2 − 1) sgn(x ∧ y)) w

√
Wronskian: wα = γ + 2 2α + 2αc.
Transition density w.r.t. m:
1  2 2

p(t; x, y) = √ e−(x−y) /2t − e−(|x|+|y|) /2t
2 2t(1 + (2 − 1) sgn(x ∧ y))
√
 +1

(|x| + |y|) ( + 1)2 t  
|x| + | y | (1 +  ) t 
+ exp + Erfc √ + √
4c 1− 2c2 2t c 2
2 √
 −1 (|x| + |y|) ( − 1) t |x| + |y | (1 −  ) t 
  
+ exp + 2 Erfc √ + √ ,
4c 1+ 2c 2t c 2
√
where Υ = 1 − γc. The epression for the density is valid for all non-negative
values on c and γ; also when γc > 1.

14. Brownian motion with drift

Speed measure: m(dx) = 2e2µx dx.

1
Scale function: s(x) = (1 − e−2µx ).
2
1 d2 f df
Generator: Gf = +µ .
2 dx2 dx
132 APPENDIX 1

Domain: D = {f : f, Gf ∈ Cb (R)}.
Green function:
√ √
2α+µ2 +µ)x ( 2α+µ2 −µ)y
Gα (x, y) = wα−1 e−( e , x ≥ y.
p
Wronskian: wα = 2 2α + µ2 .
Transition density w.r.t. m:
1 2 t (x − y)2
 
p(t; x, y) = √ exp −µ(x + y) − − .
2 2t 2 2t

Spectral representation of the transition density:

Z ∞
1 2 2
p(t; x, y) = e−µ(y+x) e−(µ +λ )t/2 cos(λ(x − y)) dλ .
2 0

Absolute continuity:
dP(xµ) 2 t
 
= exp ω(t) − x) − Px -a.s.,
dPx c
Ft+ 2

where P(xµ) and Px are the measures associated to Brownian motion with drift
and Brownian motion, respectively.

15. Brownian motion with alternating drift

Speed measure: m(dx) = 2e2µ|x| dx.

1
Scale function: s(x) = (1 − e−2µ|x| ).
2 sgn x
Generator:
1 d2 f df 1 1
Gf = + µ sgn x , x 6= 0, Gf (0) = f 00 (0+) = f 00 (0−).
2 dx2 dx 2 2
Domain: D = {f : f, Gf ∈ Cb (R)}.
Positive recurrence: if µ < 0 the process is positively recurrent. The stationary
probability measure is
m̂(dx) = |µ|e2µ|x| dx.

Green function:
 −(Υ −µ)x
 

wα−1 e−(Υ −µ)x + e − e(Υ +µ)x e(Υ +µ)y , 0 ≥ x ≥ y,





Gα (x, y) = wα−1 e−(Υ +µ)x e(Υ +µ)y , x ≥ 0 ≥ y,


  
 w−1 e−(Υ +µ)x e(Υ −µ)y +

e(Υ −µ)y − e−(Υ +µ)y , x ≥ y ≥ 0,


α 

√
1 −µ(|x|+|y|)

−|x−y| 2α+µ2 2 −(|x|+|y|)√2α+µ2 
= e e − e ,
2 2 + 2 w
p
 BRIEFLY ON SOME DIFFUSIONS 133
p
where Υ = 2α + µ2 .
p
Wronskian: wα = 2( 2α + µ2 + µ).
Transition density w.r.t. m:

1 2 (x − y)2   |x| + |y | + t
  
p(t; x, y) = √ exp −µ(|x| + |y|) − t − − Erfc √ .
2 2t 2 2t 4 2t

Absolute continuity:

dP(x±) 2 t
 
= exp µ(|ω(t)| − |x|) − µ `(t, 0) − Px -a.s.,
dPx c
Ft+ 2

where P(x±) and Px are the measures associated to Brownian motion with alter-
nating drift and Brownian motion, respectively, and `(t, 0) is the local time in
the canonical framework with respect to Lebesgue measure, that is,

1 t
Z
`(t, 0) := lim 1I(−ε,ε) (ωs )ds.
ε↓0 2" 0

Notice that if {ω(t) : t ≥ 0} is a Brownian motion with alternating drift (also

called a bang-bang Brownian motion) then {|ω(t)| : t ≥ 0} is a reflected Brown-
ian motion with drift (see No. 16), see also Graversen and Shiryaev (2000).

16. Brownian motion on [0, ∞) with drift and reflected at 0

Speed measure: m(dx) = 2e2µx dx.

1
Scale function: s(x) = (1 − e−2µx ).
2
1 d2 f df 1
Generator: Gf = +µ , x > 0, Gf (0) = f 00 (0+) + µf 0 (0+).
2 dx2 dx 2
Domain: D = {f : f, Gf ∈ Cb ([0, ∞)), f 0 (0+) = 0}.
Positive recurrence: if µ < 0 the process is positively recurrent. The stationary
probability measure is
m̂(dx) = 2|µ|e2µx dx.

Green function:
 +  (Υ −µ)y  −  −(Υ +µ)y
 
Gα (x, y) = wα−1 e−(Υ +µ)x e + e , x ≥ y,
2 2
p
where Υ = 2α + µ2 .
p
Wronskian: wα = 2α + µ2 + µ.
134 APPENDIX 1

Transition density w.r.t. m:

1 2 (x − y)2  (x + y)2 
   
p(t; x, y) = √ exp −µ(x + y) − t exp − + exp −
2 2t 2 2t 2t
 x + y + t
 
− Erfc √ , x ≥ 0, y ≥ 0.
2 2t

Absolute continuity (Kinkladze (1982)):

dP(xµ)  `(t; 0) 2 t
 
= exp µ(ω(t) − x) − − Px -a.s.,
dPx Ft+c 2 2

where P(xµ) and Px are the measures associated to reflected Brownian motion
with drift and reflected Brownian motion, respectively, and `(t, 0) is the local
time in the canonical framework with respect to Lebesgue measure, that is,

1 t
Z
`(t, 0) := lim 1I[0,ε) (ωs )ds.
ε↓0 " 0

17. Brownian motion on (0, ∞) with drift and killed at 0

Speed measure: m(dx) = 2e2µx dx.

1
Scale function: s(x) = (1 − e−2µx ).
2
1 d2 f df
Generator: Gf = +µ , x > 0.
2 dx2 dx
Domain: D = {f : f, Gf ∈ Cb ((0, ∞)), f (0+) = 0}.
Probability that the process drifts to ∞:
   1 − e−2µx , µ ≥ 0,
Px lim ω(t) = ∞ =
t→∞ 0, µ ≤ 0.

Green function:
√  √ √ 
2α+µ2 +µ)x 2 2
Gα (x, y) = wα−1 e−( e( 2α+µ −µ)y − e−( 2α+µ +µ)y , x ≥ y.

p
Wronskian: wα = 2 2α + µ2 .
Transition density w.r.t. m:
1 2 (x − y)2  (x + y)2 
   
p(t; x, y) = √ exp −µ(x + y) − t exp − − exp − .
2 2t 2 2t 2t

Spectral representation of the transition density:

1 −µ(x+y) ∞ −(µ2 +λ2 )t/2
Z
p(t; x, y) = e e sin(λx) sin(λy) dλ.
 0
 BRIEFLY ON SOME DIFFUSIONS 135

18. Brownian motion with drift and killed elastically at 0

Speed measure: m(dx) = 2e2µx dx.

1
Scale function: s(x) = (1 − e−2µx ).
2
Killing measure: k(dx) = γε{0} (dx), γ > 0.
1 d2 f df
Generator: Gf = +µ , x 6= 0,
2 dx2 dx
1 1
Gf (0) = f 00 (0+) + µf 0 (0+) = f 00 (0−) + µf 0 (0−).
2 2
Domain: D = {f : f, Gf ∈ Cb (R), f 0 (0+) − f 0 (0−) = γf (0)}.
Green function:
   
−1 −(Υ +µ)x (Υ −µ)x
w 1 + e − e e(Υ −µ)y , 0 ≥ x ≥ y,
α 2 2




Gα (x, y) = wα−1 e−(Υ +µ)x e(Υ −µ)y , x ≥ 0 ≥ y,

   
 w−1 e−(Υ +µ)x 1 +

e(Υ −µ)y − e−(Υ +µ)y , x ≥ y ≥ 0,
α 2 2
√ √
1 −µ(x+y)

−|x−y| 2α+µ2 2α+µ2

= 2 e e − e−(|x|+|y|) ,
2 2 + w
p

p
where Υ = 2α + µ2 .
p
Wronskian: wα = 2 2α + µ2 + γ.
Transition density w.r.t. m:

2 t 1 (x − y)2 
  
p(t; x, y) = exp −µ(x + y) − exp −
√
2 2 2t 2t

|x| + |y | 2t  
2|x| + 2|y| + t 
− exp + Erfc √ .
8 2 8 2 2t

19. Brownian motion on [0, ∞) with drift and killed elastically at 0

Speed measure: m(dx) = 2e2µx dx.

1
Scale function: s(x) = (1 − e−2µx ).
2
Killing measure: k(dx) = γε{0} (dx), γ > 0.
1 d2 f df
Generator: Gf = +µ , x > 0,
2 dx2 dx
1
Gf (0) = f 00 (0+) + µf 0 (0+).
2
Domain: D = {f : f, Gf ∈ Cb ([0, ∞)), f 0 (0+) = γf (0)}.
136 APPENDIX 1

Green function:
Gα (x, y) = wα−1 e−(Υ +µ)x

1 +   (Υ −µ)y  1 +   −(Υ +µ)y 
× + e + − e , x ≥ y ≥ 0,
2 2 2 2
p
where Υ = 2α + µ2 .
p
Wronskian: wα = 2α + µ2 + γ + µ.
Transition density w.r.t. m:
1 2 (x − y)2  (x + y)2 
   
p(t; x, y) = √ exp −µ(y + x) − t exp − + exp −
2 2t 2 2t 2t
+
 2t  
x + y + t + t

− exp γ(x + y + µt) + Erfc √ , x ≥ 0, y ≥ 0.
2 2 2t

20. Geometric (or exponential) Brownian motions

These diffusions are usually indexed by two parameters σ 2 > 0 and µ ∈ R. Set
 1
ν= 2− .
 2
2
Speed measure: m(dx) = 2 x2ν−1 dx.

Scale function: ( x−2
− , ν 6= 0,
s(x) = 2
ln x, ν = 0.
Boundary classification: for all values of ν both boundary points 0 and ∞ are
natural. Moreover
if ν < 0, then lim ω(t) = 0 a.s.,
t→∞
if ν > 0, then lim ω(t) = ∞ a.s.,
t→∞
if ν = 0, then lim inf ω(t) = 0, lim sup ω(t) = +∞ a.s.
t→∞ t→∞
1 d2 f df
Generator: Gf = σ 2 x2 2 + µ x , x > 0.
2 dx dx
Domain: D = {f : f, Gf ∈ Cb ((0, ∞))}.
Green function:
√ 2 2
√ 2 2
Gα (x, y) = wα−1 x− ν +2α/σ −ν y ν +2α/σ −ν , x ≥ y > 0.
p
Wronskian: wα = 2 ν 2 + 2α/σ 2 .
Transition density w.r.t. m:
 2 2
| |   t (ln y − ln x)2

p(t; x, y) = √ (xy)−ν exp − − 2 .
2 2t 2 2 t
Let {ω(t) : t ≥ 0} be a Brownian motion started at 0. Then for x > 0
{x exp((µ − σ 2 /2)t + σ ω(t)) : t ≥ 0}
is a geometric Brownian motion started at x and having the parameters µ, σ 2 .
 BRIEFLY ON SOME DIFFUSIONS 137

21. Bessel processes

Bessel processes are indexed by a parameter ν ∈ R. Recall from No. IV.34 that
ν = n/2 − 1, where n is the “dimension” parameter, i.e, n = 2ν + 2. The state
space of a Bessel process is [0, ∞) or (0, ∞) depending on the value of ν and the
boundary condition at 0. The transition density of a Bessel process is computed
in Molchanov (1967).
Speed measure:
mν (dx) = 2x2ν+1 dx.

Scale function: ( x−2

− , ν 6= 0,
sν (x) = 2
log x, ν = 0.

Boundary classification: for all values of ν, the boundary point ∞ is natural.

The nature of the boundary point 0 depends on the value of ν as follows:
if ν ≥ 0, then 0 is entrance-not-exit,
if −1 < ν < 0, then 0 is non-singular,
if ν ≤ −1, then 0 is exit-not-entrance.
In the non-singular case the boundary condition at 0 is usually reflection or
killing.
1 d2 f 2 + 1 d f
Generator: Gf = + , x > 0.
2 dx2 2x dx
Domain: if ν ≥ 0 or if −1 < ν < 0 and 0 is reflecting,

D = {f : f, Gf ∈ Cb ([0, ∞)), f + (0+) = 0};

if ν ≤ −1 or if −1 < ν < 0 and 0 is killing,

D = {f : f, Gf ∈ Cb ([0, ∞)), f (0) = 0}.

Green function:
if ν ≥ 0 or if −1 < ν < 0 and 0 is reflecting,
√ √
Gα (x, y) = wα−1 x−ν Kν (x 2α) y −ν Iν (y 2α), x ≥ y > 0,
 √ ν √
−1 2 1
Gα (0, x) = wα x−ν Kν (x 2α), x > 0;
2 ( + 1)

if ν ≤ −1 or if −1 < ν < 0 and 0 is killing,

√ √
Gα (x, y) = wα−1 x−ν Kν (x 2α) y −ν I−ν (y 2α), x ≥ y > 0.
138 APPENDIX 1

Wronskian: wα = 1.
Transition density w.r.t. m:
if ν ≥ 0 or if −1 < ν < 0 and 0 is reflecting,
 2
1 x + y2 xy
  
p(t; x, y) = (xy)−ν exp − Iν ,
2t 2t t
 2
1 y
p(t; 0, y) =  +1  +1 exp − ;
2 t ( + 1) 2t

if ν ≤ −1 or if −1 < ν < 0 and 0 is killing,

 2
1 x + y2 xy
  
p(t; x, y) = (xy)−ν exp − I|ν| .
2t 2t t

Absolute continuity:
 2
 − 2
Z t
dP(xµ) (ω(t))µ−ν ds

= exp − 2 P(xν ) -a.s. on {H0 > t},
dP(xν ) c
Ft+ xµ−ν 2 0
! (s)

where x > 0 and P(ν ) is the measure associated to a Bessel process of order ν.
Notice that if ν ≥ 0 the condition “on {H0 > t}” can be omitted. Putting here
ν = −1/2 gives the density w.r.t. the Wiener measure. For a proof in the case
µ ≥ 0 and ν ≥ 0 see Yor (1980). The general case is obtained from No. IV.43
using the chain rule.

22. Special case of No. 21: 3-dimensional Bessel process, (i.e., ν = 1/2)

Speed measure: m(dx) = 2x2 dx.

1
Scale function: s(x) = − .
x
Boundary classification: ∞ is natural and 0 is entrance-not-exit.
1 d2 f 1 d f
Generator: Gf = + , x > 0.
2 dx2 x dx
Domain: D = {f : f, Gf ∈ Cb ([0, ∞)), f + (0+) = 0}.
Green function:
1 √
1 √ √
Gα (x, y) = wα−1 e−x 2α
ey 2α
− e−y 2α


, x ≥ y > 0,
x y
1 √
− 2αx
Gα (0, x) = e , x > 0.
x
√
Wronskian: wα = 2 2α.
Transition density w.r.t. m:
1 1   (x − y)2  
(x + y)2 
p(t; x, y) = √ exp − − exp − ,
2 2t xy 2t 2t
 2
1 y
p(t; 0, y) = √ 3 exp − .
2t 2t
 BRIEFLY ON SOME DIFFUSIONS 139

Absolute continuity:

dP(1/2)
x !(t)
= Px -a.s. on {H0 > t},
dPx c
Ft+ x

where x > 0, P(1/2) is the measure associated to a three-dimensional Bessel

process, and P is the measure associated to a Brownian motion.

23. Squared Bessel processes

Squared Bessel processes are indexed by a parameter ν ∈ R. The state space of

these diffusions is [0, ∞) or (0, ∞) depending on the value of ν and the boundary
condition at 0.
Let {ω(t) : t ≥ 0} be a Bessel process of index (or order) ν. Then {ω(t)2 : t ≥ 0}
is a squared Bessel process of order ν.
Speed measure:
x
mν (dx) = dx.
2
Scale function: ( x−
− , ν 6= 0,
sν (x) = 
log x, ν = 0.

Boundary classification: for all values of ν, the boundary point ∞ is natural.

The nature of the boundary point 0 depends on the value of ν as follows:
if ν ≥ 0, then 0 is entrance-not-exit,
if −1 < ν < 0, then 0 is non-singular,
if ν ≤ −1, then 0 is exit-not-entrance.
d2 f df
Generator: Gf = 2x + (2ν + 2) , x > 0 .
dx2 dx
Domain: if ν ≥ 0 or if −1 < ν < 0 and 0 is reflecting,

D = {f : f, Gf ∈ Cb ([0, ∞)), f + (0+) = 0},

if ν ≤ −1 or if −1 < ν < 0 and 0 is killing,

D = {f : f, Gf ∈ Cb ([0, ∞)), f (0) = 0} .

Green function:
if ν ≥ 0 or if −1 < ν < 0 and 0 is reflecting,
√
Gα (x, y) = wα−1 x−ν/2 Kν ( 2αx) y −ν/2 Iν ( 2αy),
p
x ≥ y > 0,
 √ ν √
2 1
Gα (0, x) = wα−1 x−ν/2 Kν ( 2αx), x > 0;
2 ( + 1)
140 APPENDIX 1

if ν ≤ −1 or if −1 < ν < 0 and 0 is killing,

√
Gα (x, y) = wα−1 x−ν/2 Kν ( 2αx) y −ν/2 I−ν ( 2αy),
p
x ≥ y > 0.

Wronskian: wα = 1/2.
Transition density w.r.t. m:
if ν ≥ 0 or if −1 < ν < 0 and 0 reflecting,
  √xy 
1 
x+y
p(t; x, y) = (xy)−ν/2 exp − Iν ,
t 2t t
1 
y

p(t; 0, y) =   +1 exp − ;
2 t ( + 1) 2t

if ν ≤ −1 or if −1 < ν < 0 and 0 is killing,

 √xy 
1 
x+y

p(t; x, y) = (xy)−ν/2 exp − I|ν| .
t 2t t

24. Ornstein–Uhlenbeck process

Ornstein–Uhlenbeck processes are usually indexed by one parameter γ ∈ R. For

γ = 0 we have Brownian motion. The behavior of Ornstein–Uhlenbeck processes
is very different depending on whether γ > 0 or γ < 0.
Speed measure: mγ (dx) = 2 exp(−γx2 ) dx .
Z x
Scale function: sγ (x) = exp(γy 2 ) dy .
0
1 d2 f df
Generator: Gf = − γx .
2 dx2 dx
Domain: D = {f : f, Gf ∈ Cb (R)}.
Doob’s transformation: Let Px denote the measure associated to Brownian
motion, and for γ 6= 0 introduce

e2 t − 1
τt := .
2
Then, under the measure Px , the process

{e−γt ω(τt ) : t ≥ 0}

is identical in law to an Ornstein–Uhlenbeck process of index γ started at x.

Absolute continuity:

dP(xγ ) t 2 t
 Z 
2 2
= exp − (ω (t) − x ) + − ω 2 (s) ds Px -a.s.,
dPx c
Ft+ 2 2 2 0
 BRIEFLY ON SOME DIFFUSIONS 141

where P(γ ) is the measure associated to an Ornstein–Uhlenbeck process of index

γ and P is the Wiener measure. By the chain rule,
dP(x−γ )
= exp γ(ω 2 (t) − x2 ) − γ t P(xγ ) -a.s.


(γ )
dPx Ft+ c

(A) The case γ > 0.

The process is positively recurrent. Its stationary probability measure is
√
m̂(dx) = √ exp −γx2 dx.



Green function:
 2
x
Gα (x, y) = wα−1 exp
p
D−α/γ (x 2γ)
2
 2
y p
× exp D−α/γ (−y 2γ), x ≥ y.
2
2√ 
Wronskian: wα = .
( = )
Transition density w.r.t. m :
(x2 + y2 ) (x2 + y2 ) ch( t) − 2xy 
√
1 
t
p(t; x, y) = exp + − .
2 2 sh( t) 2 2 2 sh( t)
p

(B) The case γ < 0.

Green function:
| |x2
  p
Gα (x, y) = wα−1 exp − D−α/|γ|−1 (x 2|γ|)
2
| |y 2
  p
× exp − D−α/|γ|−1 (−y 2|γ|), x ≥ y.
2
2 | |
p
Wronskian: wα = .
( =| | + 1)
The transition density w.r.t. m has the same form as for γ > 0 :
1 | |t | |(x2 + y 2 ) (x2 + y2 )| | ch(| |t) − 2xy| | 
p
| |

p(t; x, y) = p exp − − − .
2 2 sh(| |t) 2 2 2 sh(| |t)

For γ < 0 the Ornstein–Uhlenbeck process is transient and, by symmetry,

   
P0 lim ω(t) = +∞ = P0 lim ω(t) = −∞ = 1/2.
t→∞ t→∞

In Part II Section 7 we consider the Ornstein–Uhlenbeck process indexed by two

parameters σ > 0 and θ > 0 with the generator
d2 f df
Gf = σ 2 θ 2 − θx , θ > 0.
dx dx
The one parameter model √ in the case γ > 0 is obtained from the two parameter
model by setting σ = 1/ 2γ and θ = γ.
142 APPENDIX 1

25. Radial Ornstein–Uhlenbeck process

Let {ω (i) (t) : t ≥ 0}, i = 1, 2, . . . n, be independent Ornstein–Uhlenbeck pro-

cesses with the same parameter γ. Then the process
n
nX
2 1/2 o
ω (k) (t) : t≥0
k=1

is called radial Ornstein-Uhlenbeck process with parameters ν = n/2 − 1 and γ.

For γ = 0 we have a Bessel process of order ν. Below we let, as in the case of
Bessel processes, ν be an arbitrary real number.


Speed measure: m(dx) = 2x2ν+1 exp −γx2 dx.
Z x
Scale function: s(x) = y −2ν−1 exp(γy 2 ) dy.

Doob’s transformation: let Px denote the measure associated to a Bessel process

of order ν started at x, and for γ 6= 0 introduce

e2 t − 1
τt := .
2
Then, under the measure Px , the process

{e−γt ω(τt ) : t ≥ 0}

is identical in law to a radial Ornstein–Uhlenbeck process with the parameters

ν and γ started at x.
Boundary classification: for all values of ν, the boundary point ∞ is natural.
The nature of the boundary point 0 depends on the value of ν as follows:
if ν ≥ 0, then 0 is entrance-not-exit,
if −1 < ν < 0, then 0 is non-singular,
if ν ≤ −1, then 0 is exit-not-entrance.
In the non-singular case we take below the boundary condition at 0 to be reflec-
tion or killing.
1 d2 f  2 + 1 
df
Generator: Gf = 2 + − γx , x > 0.
2 dx 2x dx
Domain:
if ν ≥ 0 or if −1 < ν < 0 and 0 is reflecting,

D = {f : f, Gf ∈ Cb ([0, ∞)), f + (0+) = 0};

if ν ≤ −1 or if −1 < ν < 0 and 0 is killing,

D = {f : f, Gf ∈ Cb ([0, ∞)), f (0) = 0}.

 BRIEFLY ON SOME DIFFUSIONS 143

Absolute continuity: for µ ≥ 0, ν ∈ R, and x > 0

2 −  2
Z t
dP(xµ) (ω(t))µ−ν ds
 
= exp γ(µ−ν)t− 2 P(xν ) -a.s. on {H0 > t},
dP(xν ) c
Ft+ xµ−ν 2 0
! (s)

where P(xµ) and P(xν ) are the measures associated to the radial Ornstein–Uhlen-
beck processes of order µ and ν, respectively. Notice that if ν ≥ 0 the condition
“on {H0 > t}” can be omitted. Putting here ν = −1/2 gives the density w.r.t.
the Ornstein–Uhlenbeck measure.
(A) The case γ > 0.
For ν ≥ 0, or if −1 < ν < 0 and 0 is a reflecting boundary the process is
positively recurrent, and its stationary probability measure is

2  +1 x2 +1  
m̂(dx) = exp −γx2 dx.
( + 1)

For ν = 0 and γ = 1/2 this is the so-called Rayleigh distribution, for ν = 1/2 and
γ = 1/2 we have the Maxwell distribution and, in general, for ν = n/2 − 1, n =
1, 2, . . . , and γ = 1/2 the distribution of the square root of the χ2 -distributed
random variable with n degrees of freedom.
Green function:
if ν ≥ 0 or if −1 < ν < 0 and 0 is reflecting,
 2
x
Gα (x, y) = wα−1 x−ν−1 exp W− α + ν+1 , ν γx2


2 2γ 2 2
 2
−ν−1 y
M− α + ν+1 , ν γy 2 ,


×y exp x ≥ y.
2 2γ 2 2

2 (1 +  )
Wronskian: wα = .
( =2 )
If ν ≤ −1 or if −1 < ν < 0 and 0 is killing,
 2
x
Gα (x, y) = wα−1 x−ν−1 exp W− α + ν+1 ,− ν γx2


2 2γ 2 2
 2
y
× y −ν−1 exp M− α + ν+1 ,− ν γy 2 ,


x ≥ y.
2 2γ 2 2

2 (1 − 
)
Wronskian: wα = .
2 −
Transition density w.r.t. m:
if ν ≥ 0 or if −1 < ν < 0 and 0 is reflecting,

e ( +1)t e− t (x2 + y2 ) xy
   
p(t; x, y) = exp − I ,
2x y sh( t) 2 sh( t) ν
sh( t)
 +1 e ( +1)t 
e− t y2

p(t; 0, y) =  +1 exp − ;
2 ( + 1)(sh( t)) +1 2 sh( t)
144 APPENDIX 1

if ν ≤ −1 or if −1 < ν < 0 and 0 is killing,

e ( +1)t e− t (x2 + y2 ) xy
   
p(t; x, y) =   exp − I−ν .
2x y sh( t) 2 sh( t) sh( t)

(B) The case γ < 0.

Green function:
if ν ≥ 0 or if −1 < ν < 0 and 0 is reflecting,

| | x2
 
Gα (x, y) = wα−1 x−ν−1 exp − W− α − ν+1 , ν |γ|x2


2 2|γ| 2 2

| | y2
 
× y −ν−1 exp − M− α − ν+1 , ν |γ|y 2 ,


x ≥ y.
2 2|γ| 2 2

2| (1 +  )

|
Wronskian: wα = .
2| | +  + 1
If ν ≤ −1 or if −1 < ν < 0 and 0 is killing,

| | x2
 
Gα (x, y) = wα−1 x−ν−1 exp − W− α − ν+1 , |ν| |γ|x2


2 2|γ| 2 2

| | y2
 
−ν−1
M− α − ν+1 , |ν| |γ|y 2 ,


×y exp − x ≥ y.
2 2|γ| 2 2

2| | (1 − 
)
Wronskian: wα = .
2| |
+1
Transition density w.r.t. m has the same form as for γ > 0.
In Part II Section 8 the radial Ornstein–Uhlenbeck process is parameterized to
have the generator

d2 f
 2
 (2 + 1) df

Gf = σ 2 θ + − θx , θ > 0.
dx2 x dx

26. Squared radial Ornstein–Uhlenbeck process

Let {ω(t) : t ≥ 0} be a radial Ornstein–Uhlenbeck process with the parameters

ν and γ. Then the process o
 2
ω (t) : t ≥ 0

is called squared radial Ornstein–Uhlenbeck process with the parameters ν and

γ. For γ = 0 we have a squared Bessel process of order ν.
1

Speed measure: m(dx) = xν exp −γ x dx.
2
Z x
Scale function: s(x) = y −ν−1 exp(γ y) dy.
 BRIEFLY ON SOME DIFFUSIONS 145

Time transformation: let Px denote the measure associated to a squared Bessel

process of order ν started at x, and for γ 6= 0 introduce
e t−1
τt := .

Then, under the measure Px , the process

{e−γt ω(τt ) : t ≥ 0}

is identical in law to a squared radial Ornstein–Uhlenbeck process with the

parameters ν and γ started at x.
Boundary classification: For all values of ν and γ, the boundary point ∞ is
natural. The nature of the boundary point 0 depends on the value of ν as
follows:
if ν ≥ 0, then 0 is entrance-not-exit,
if −1 < ν < 0, then 0 is non-singular,
if ν ≤ −1, then 0 is exit-not-entrance.
In the non-singular case we take below the boundary condition at 0 to be reflec-
tion or killing.
d2 f df
 
Generator: Gf = 2x 2 + 2ν + 2 − 2γx .
dx dx
Domain:
if ν ≥ 0 or if −1 < ν < 0 and 0 is reflecting,

D = {f : f, Gf ∈ Cb ([0, ∞)), f + (0+) = 0};

if ν ≤ −1 or if −1 < ν < 0 and 0 is killing,

D = {f : f, Gf ∈ Cb ([0, ∞)), f (0) = 0}.

(A) The case γ > 0.

For ν ≥ 0, or if −1 < ν < 0 and 0 is a reflecting boundary the process is
positively recurrent, and its stationary probability measure is
 +1 x  
m̂(dx) = exp −γ x dx,
( + 1)
i.e., the gamma–distribution with the parameters γ and ν + 1.
Green function:
if ν ≥ 0 or if −1 < ν < 0 and 0 is reflecting,
x
 
Gα (x, y) = wα−1 x−(ν+1)/2 exp


W− α + ν+1 , ν γx
2 2γ 2 2

y
 
× y −(ν+1)/2 exp


M− α + ν+1 , ν γy , x ≥ y.
2 2γ 2 2
146 APPENDIX 1
(1 +  )
Wronskian: wα = .
( =2 )
If ν ≤ −1 or if −1 < ν < 0 and 0 is killing,
x
 
Gα (x, y) = wα−1 x−(ν+1)/2 exp


W− α + ν+1 ,− ν γx
2 2γ 2 2

y
 
× y −(ν+1)/2 exp


M− α + ν+1 ,− ν γy , x ≥ y.
2 2γ 2 2

(1 −  )

Wronskian: wα = .
2 −
Transition density w.r.t. m:
if ν ≥ 0 or if −1 < ν < 0 and 0 is reflecting,

e ( +1)t
  √xy 
e− t (x + y)

p(t; x, y) = =2 =2 exp − Iν ,
x y sh( t) 2 sh( t) sh( t)
 +1 e ( +1)t 
e− t y

p(t; 0, y) =   +1 exp − ;
2 ( + 1)(sh( t)) 2 sh( t)

if ν ≤ −1 or if −1 < ν < 0 and 0 is killing,

e ( +1)t e− t (x2 + y2 )
   √xy 
p(t; x, y) = =2 =2 exp − I−ν .
x y sh( t) 2 sh( t) sh( t)

B) The case γ < 0.

Green function:
if ν ≥ 0 or if −1 < ν < 0 and 0 is reflecting,
x
 
Gα (x, y) = wα−1 x−(ν+1)/2 exp


W− α − ν+1 , ν −γ x
2 2|γ| 2 2

y
 
× y −(ν+1)/2 exp


M− α − ν+1 , ν −γ y , x ≥ y.
2 2|γ| 2 2

| | (1 +  )

Wronskian: wα = .
2| | +  + 1
If ν ≤ −1 or if −1 < ν < 0 and 0 is killing,
x
 
Gα (x, y) = wα−1 x−(ν+1)/2 exp


W− α − ν+1 , |ν| −γ x
2 2|γ| 2 2

y
 
× y −(ν+1)/2 exp


M− α − ν+1 , |ν| −γ y , x ≥ y.
2 2|γ| 2 2

| | (1 −  )

Wronskian: wα = .
2| | + 1
Transition density w.r.t. m has the same form as for γ > 0.
 BRIEFLY ON SOME DIFFUSIONS 147

27. Constant elasticity of variance (CEV) process

The CEV process is a weak solution of the SDE

dSt = µSt dt + σSt1+β dWt , S0 = x > 0.

considered on (0, +∞) up to H0 := inf{t : St = 0} (if finite). The parameters β

and µ are arbitrary real numbers. For β = 0 we recover the geometric Brown-
ian motion treated in No. 20 and for β = −1/2 the squared radial Ornstein-
Uhlenbeck process with the parameter ν = −1, see No. 26. We assume now
that β 6= 0.
2 


Speed measure: m(dx) = 2 x−2(1+β) exp − 2 x−2β dx.
 
Z x
 −2β
 
Scale function: s(x) = exp 2 y dy.

Boundary classification: The nature of the boundary point ∞ depends on the
value of β as follows:
if β < 0, then ∞ is natural,
if β > 0, then ∞ is entrance-not-exit.
The nature of the boundary point 0 depends on the values of β as follows:
if β < −1/2, then 0 is non-singular,
if −1/2 ≤ β < 0, then 0 is exit-not-entrance,
if β > 0, then 0 is natural.
In the non-singular case, 0 is taken to be a killing boundary.
1 2 2+2β d2 f df
Generator: Gf = σ x + µx , x > 0.
2 dx2 dx
Domain:
if β > 0,
D = {f : f, Gf ∈ Cb ((0, ∞)), lim f 0 (x) = 0},
x→∞

if β < −1/2 and 0 is a killing boundary, or −1/2 ≤ β < 0,

D = {f : f, Gf ∈ Cb ((0, ∞)), f (0+) = 0}.

Transience and recurrence:

if β < 0 and µ ≤ 0, then H0 < ∞ a.s.,
if β < 0 and µ > 0, then lim St = +∞ with positive probability,
t→∞

if β > 0 and µ ≤ 0, then lim St = 0 a.s.,

t→∞

if β > 0 and µ > 0, then the process is positively recurrent.

148 APPENDIX 1

Green function:
If β < 0, µ 6= 0,
||(x−2 + y−2 )
 
Gα (x, y) = wα−1 exp 2 (xy)β+1/2
2 | |
||x−2 ||y −2
   
× Wn,m M , x ≥ y,
2 | | n,m
2 | |
where (and also below)

1 1 1
 = sign(µβ), n= + − , m= .
2 4 2| | 4| |
2|| (2m + 1)
Wronskian: wα = 2 .
 (m − n + 1=2)
If β > 0, µ 6= 0,
||(x−2 + y−2 )
 
Gα (x, y) = wα−1 exp (xy)β+1/2
22 | |
| | x − 2 ||y −2
   
× Mn,m 2 Wn,m 2 , x ≥ y.
 | |  | |
2|| (2m + 1)
Wronskian: wα = 2 .
 (m − n + 1=2)
If β < 0, µ = 0,
x− √
 − √ 
√ y
 
Gα (x, y) = wα−1 xy K1/2|β| 2α I1/2|β| 2α , x ≥ y.
| | | |

Wronskian: wα = |β|.
If β > 0, µ = 0,
x− √
 − √ 
√ y
 
Gα (x, y) = wα−1 xy I1/2|β| 2α K1/2|β| 2α , x ≥ y.
| | | |
Wronskian: wα = |β|.
Transition density w.r.t. m:
if β 6= 0, µ 6= 0,
|| e( +1=2)t √ e− t (x−2 + y−2 ) x− y−
   
p(t; x, y) = xy exp − 2 I1/2|β| 2 ,
2 sh(| |t) 2 sh( t)  sh( t)

if β 6= 0, µ = 0,
(x−2 + y−2 ) 
√
xy 
x y
 − − 
p(t; x, y) = exp − I .
2| |t 22 2 t 1/2|β|
2 2 t

CEV processes were introduced by Cox (1975). We refer also to the papers
Emanuel and MacBeth (1982), Goldenberg (1991) and Davydov and Linetsky
(2001).
 Part II

TABLES OF DISTRIBUTIONS
OF FUNCTIONALS OF BROWNIAN MOTION
AND RELATED PROCESSES

INTRODUCTION

The tables are divided into nine sections:

1. Brownian motion 159
2. Brownian motion with drift 256
3. Reflecting Brownian motion 339
4. Bessel process of non-negative order 379
5. Bessel process of order 1/2 435
6. Bessel process of order zero 513
7. Ornstein–Uhlenbeck process 528
8. Radial Ornstein–Uhlenbeck process 571
9. Geometric (exponential) Brownian motion 612

Let X denote any of these processes. In the tables we consider functionals of X

up to the following stopping times:
1. τ — exponentially distributed random “time” independent of X.
2. Hz := inf{s : Xs = z} — first hitting time of the point z.
3. Ha,b := inf{s : Xs 6∈ (a, b)} — first exit time from the interval (a, b).
4. %(t, z) := inf{s : `(s, z) = t} — inverse local time at the point z
5. θv := min{t : sup Xs − inf Xs = v} — first range time at the level v.
0≤s≤t 0≤s≤t

Recall that considering functionals of X stopped at the independent exponential

time τ corresponds to taking the Laplace transform of the functional with respect
to the time parameter. Hence, inverting the transform gives the distribution at a
fixed time t. All of the above mentioned stopping times are not treated for every
process (see the list of contents of the tables on p. vii). In particular, formulae for
the first range time are computed only for Brownian motion.
The aim has been to give a “systematic” presentation. The structure of the
tables is displayed inside the individual sections via a triple numbering system. The
first number characterizes the stopping time. The second indicates a functional
or a group of functionals. There is a list of these below the reference numbers
indicated. The third number refers to the Laplace transform or to the distribution
itself; usually several special cases are considered.

© Springer Basel 2015 151

A.N. Borodin, P. Salminen, Handbook of Brownian Motion - Facts and Formulae,
Probability and Its Applications, DOI 10.1007/978-3-0348-8163-0_7
152 INTRODUCTION

Roughly speaking, Brownian motion is the most computationally tractable diffu-

sion. The structure of the tables is therefore determined by the first section, which
also contains more formulae than the other sections. In particular, this means that,
e.g., formula 1.3.2 for a Radial Ornstein–Uhlenbeck process in Section 8 has as its
counterpart formula 1.3.2 for a Brownian motion in Section 1. On the other hand,
formula 1.3.3 in Section 1 has no counterpart in Section 8. This strive toward a
systematic presentation has also lead to, e.g., there not being a Subsection 3 in
Section 3. In Section 6 there is no Subsection 1 because the formulae in Subsec-
tion 1 in Section 4 (when taking the parameter ν equal to 0) are also valid in this
case. In Section 1 some notation is also introduced that will be used in Sections 2
and 5.
Many formulae contain so called special functions. Concerning these there is
Appendix 2 at the end of the tables. For more complete information on special
functions we refer to Erdelyi et al. (1953), Abramowitz and Stegun (1970), and
Lebedev (1972) .
It is important to notice that the local times in the tables are densities of the
occupation time measure with respect to Lebesgue measure:

1 t
Z
`(t, x) = lim 1I[x,x+ε) (Xs ) ds,
ε→0 " 0

where X is any of the diffusions in the tables. Letting L(t, x) be the local time
with respect to the speed measure of X and m(x) the density of the speed measure
with respect to Lebesgue measure, then (see No. II.24)

(a) `(t, x) = m(x) L(t, x).

1. List of functionals

The number of the functional on this list corresponds to the second number in
the triple numbering system used in the tables. Here X is any of the processes and
T is any of the stopping times considered.

1. sup Xs
0≤s≤T

2. inf Xs
0≤s≤T

3. `(T, r)
Z T

4. 1I[r,∞) Xs ds
0
 1. LIST OF FUNCTIONALS 153
Z T

5. 1I(−∞,r] Xs ds
0

Z T

Z T

6. 1I(−∞,r) Xs ds, 1I[r,∞) Xs ds
0 0

Z T

7. 1I[r,u] Xs ds
0

Z T Z T
8. Xs ds or |Xs |ds
0 0

Z T
9. Xs2 ds
0

Z T
10. e2βXs ds
0

11. sup `(T, y)

y∈R

12. Ȟ(T ):= inf{s<T : Xs = sup Xs } or Ĥ(T ):= inf{s<T : Xs = inf Xs }

0≤s≤T 0≤s≤T

13. Ȟ(T ), sup Xs

0≤s≤T

14. Ĥ(T ), inf Xs

0≤s≤T

15. inf Xs , sup Xs

0≤s≤T 0≤s≤T

16. `(T, r), sup Xs

0≤s≤T

17. `(T, r), inf Xs

0≤s≤T

18. `(T, r), `(T, u)

T
ds
Z
19.
0
Xs

T
ds
Z
20.
0
Xs2

T T
ds
Z Z
21. , Xs2 ds
0
Xs2 0
154 INTRODUCTION
T T
ds ds
Z Z
22. ,
0
Xs2 0
Xs
Z T



23. p1I(−∞,r) Xs + q1I[r,∞) Xs ds, sup Xs
0 0≤s≤T

Z T



24. p1I(−∞,r) Xs + q1I[r,∞) Xs ds, inf Xs
0 0≤s≤T

25. `(T, r), inf Xs , sup Xs

0≤s≤T 0≤s≤T

Z T



26. p1I(−∞,r) Xs + q1I[r,∞) Xs ds, inf Xs , sup Xs
0 0≤s≤T 0≤s≤T

Z T

Z T

27. 1I(−∞,r) Xs ds, 1I[r,∞) Xs ds, `(T, r)
0 0

28. Ĥ(T ), Ȟ(T ), inf Xs , sup Xs

0≤s≤T 0≤s≤T

Z T

Z T

29. 1I(−∞,r) Xs ds, 1I[r,∞) Xs ds, inf Xs , sup Xs , `(T, r)
0 0 0≤s≤T 0≤s≤T

Z T Z T
30. e2βXs ds, eβXs ds
0 0

`(s, u) − α−1 `(s, r) , α > 0

31. sup
0≤s≤T

2. Comments and references

Many formulae in the tables have a general formulation which often can be found
in Part I, especially in Chapters II, IV, V, and VI. In particular, we have included
formulae based on the results in No.’s II.19, 25, 26, and IV.44. However, the main
tool has been the Feynman–Kac formula which for Brownian motion and Brownian
motion with drift is discussed in Chapter VI; see also No. II.24. Although Chapter
VI is devoted to Brownian motion, the solutions therein are given in terms of the
Green functions, indicating the form of the solution for an arbitrary diffusion. It
is explained in No. II.11 how to compute the Green function of a given diffusion.
It should be possible for the reader to check a great number of the formulae in
Sections 1 and 2 by solving the corresponding differential problems in Chapter VI
or, whichever is easier, by checking that the solution is as claimed. In the “Expo-
nential stopping” subsections one must also have some familiarity with inverting
Laplace transforms. Hereby, Kamke (1943) and Erdelyi and et al. (1954) are good
references. In Appendix 3 we have collected a number of “key” Laplace transforms
 2. COMMENTS AND REFERENCES 155

mainly taken from Erdelyi and et al. (1954), and Appendix 4 concerning second
order differential equations is extracted mainly from Kamke (1943).
Below we give more detailed comments concerning the particular functionals in
the list above. In the papers referred to one can usually find a special case of the
considered functional or further developments.
Functionals 1 and 2. Formulae associated to these can be derived using
standard diffusion theory. Notice that, e.g.,

Px ( inf Xt < a) = Px (Ha < Hc ),

0≤t≤Hc

Px ( inf Xt < a) = Px (Ha < τ ) = Ex e−λHa ,

0≤t≤τ

and see No. II.4 and 10 for these expressions for general regular diffusions. We
refer also to No. VI.10 and 4, respectively, which when choosing b = +∞, f ≡ 0,
γi = 0, i = 1, 2, . . . k and F ≡ 1 lead to the same solutions. A general form of

Px ( inf Xs > a)
0≤s≤%(t,z)

is given in No. VI.18.

Functional 3. Recall the formula (a) in No. II.27:


Px (L(ζ, x) > α) = exp − ,
G0 (x; x)

where ζ is the life time and G0 is the Green function of the considered diffusion.
Using this, the strong Markov property and the relation `(t, x) = m(x) L(t, x) give


(b) Pz (`(ζ, x) > α) = Pz (Hx < ∞) exp − ,
m(x)G0 (x; x)

where m is the density of the speed measure with respect to the Lebesgue measure
(when it exists). Choosing in (b) ζ = τ or Hc or Ha,b and computing the cor-
responding Green function give many of the formulae displayed in the tables. In
this way we may also find the distribution of local time of a diffusion stopped at
%(τ, x) and this can be inverted with respect to λ. For the approach via differential
equations see No. VI.4, 9, 16, 18, and 22.
Functionals 4–7. Deriving distributions of these is a typical application of the
Feynman–Kac formula. Also here it is possible to formulate general results for a
large class of diffusions. Indeed, consider, e.g., the problem of finding the joint
distribution of the occupation times
Z t Z τ
A− A+



(c) t := 1I(−∞,0) Xs ds, t := 1I(0,∞) Xs ds,
0 0

where X is a regular diffusion on the interval I = (l, r) containing 0. We assume

that the boundaries l and r are entrance-not-exit or natural and that there is no
156 INTRODUCTION

killing. Moreover suppose that the infinitesimal generator is given by a smooth

ordinary differential operator, denoted G, as in No. II.9. From No. VI.4 we have
 Z r
−


E0 exp −α1 A+ τ − α A
2 τ = λ G̃λ (0, y) m(dy)
l
(d) Z 0 Z r

 
= ϕ̃λ (0) ψ̃λ (y) m(dy) + ψ̃λ (0) ϕ̃λ (y) m(dy) .
w~ l 0

Due to the assumptions on the boundary behavior, the functions ϕ̃λ and ψ̃λ can
be expressed (cf. No. II.10) in terms of the fundamental solutions ϕλ and ψλ
associated to X as follows
c1 ϕλ+α2 (x) + c2 ψλ+α2 (x), x ≤ 0,

ϕ̃λ (x) =
ϕλ+α1 (x), x ≥ 0,
(e)
x ≤ 0,

ψλ+α2 (x),
ψ̃λ (x) =
c3 ϕλ+α1 (x) + c4 ψλ+α1 (x), x ≥ 0,

where the constants c1 , c2 , c3 and c4 are determined by the demand that ϕ̃λ and
ψ̃λ are in C 1 (I). Using the fact that ϕλ and ψλ are solutions of the ODE Gu = λu
and computing the Wronskian w̃λ we obtain

−

   

E0 exp −α1 A+ τ − α2 Aτ = F (0) − G(0) /(F (0) − G(0)),
+ 2 + 1

where
0
F (0) := ψλ+α2
(0)/ψλ+α2 (0) and G(0) := ϕ0λ+α1 (0)/ϕλ+α1 (0).

For this formula (with a proof based on excursion theory) see Truman and Williams
(1990). To find the corresponding formula when τ is replaced with Hz , z > 0 we
can proceed as explained in No. VI.9. This gives using the notation above


−
E0 exp −α1 A+ Hz − α2 AHz = ψ̃0 (0)/ψ̃0 (z).

We refer to Pitman and Yor (1999) for an alternative derivation of this formula
and connections to excursion theory. See Knight (1969), (1978), and Jeulin and
Yor (1981) for formulae for Brownian motion. Further, we remark that the arcsine
law of Lévy (1939) is included in 1.1.4.4. and the formula 2.1.4.4. is due to Takàcs
(1996).
Functional 8. To analyze this functional using the Feynman–Kac formula one
should be able to solve the ODE

Gu(x) − α |x|u(x) = λu(x),

where α > 0, λ ≥ 0, and G is the differential operator associated to the considered

diffusion. In Appendix 4 the fundamental solutions are given for Brownian motion
in Eq. 2, for Brownian motion with drift in Eq. 9, and for Ornstein–Uhlenbeck
 2. COMMENTS AND REFERENCES 157

processes in Eq. 18. Notice also that Eq. 15.a when multiplied by x2 gives the
solutions for geometric Brownian motion, and Eq. 15 when multiplied by x gives
the ODE associated to squared Bessel processes. Especially, for the formula 1.1.8.5
we refer to Shepp (1982), where the functional is considered for Brownian bridge;
see also Rice (1982) and Johnson and Killeen (1983). The formula 9.1.8.4(1) is first
presented in Dufresne (1990), see also Yor (1992c).
Functional 9. As in the previous case one can find the needed differential
equations and their solutions in Appendix 4. The Cameron–Martin formula is
1.1.9.3; see Cameron and Martin (1945). Special cases of the general formula in
No. IV.44 due to Pitman and Yor (1982) are displayed here at many spots. See also
Wenocur (1990). Furthermore, we refer to Neveu (1968) for additional references
and results on quadratic functionals. For more recent works, see Ikeda, Kusuoka
and Manabe (1995) and Jeanblanc, Pitman and Yor (1997).
Functional 10. Formulae associated to this functional are only displayed for
Brownian motion, Brownian motion with drift and three-dimensional Bessel pro-
cess. In these cases the computations can be done studing the functional 8 for
geometric Brownian motion. For further results, see Yor (1992a,b).
Functional 11. For Brownian motion, see No. VI.8, 13, 14, 21, and 23. These
results are due to Borodin (1982, 1989), see also Csáki, Földes and Salminen (1987)
and Eisenbaum (1990). For 1.1.11.4 see Csáki and Földes (1986).
Functionals 12–14. These are special cases of the general formula in No. II.19,
see Csáki, Földes and Salminen (1987). See also Lévy (1948), Vincze (1957), Shepp
(1979), Imhof (1984), Louchard (1984), and Pitman and Yor (1996).
Functional 15. The formula 1.1.15.4 is due to Lévy. For 1.1.15.8 see Feller
(1951) and for 1.1.15.8(2) Kac, Kiefer and Wolfowitz (1955). From the latter one we
can deduce the distribution of the range of a Brownian bridge. For an explanation
that this distribution is the same as the distribution of the maximum of a three-
dimensional Bessel bridge, see No. IV.24. We refer further to Doob (1949) for the
probability that Brownian motion stays between two linear boundaries.
Functionals 16–17. See Chapter VI.
Functional 18. Joint distributions of (stopped) local times can be calculated
using the Ray–Knight theorems; see Chapter V.
Functionals 19–27, 29, 30. These are typical applications of the Feynman–
Kac formula with a side condition, see Chapter VI. In particular, we have for the
functional 27 with the notations (and assumptions) as in (c) and (d)
 Z r
+ −


E0 exp −α1 Aτ − α2 Aτ − γ `(τ, 0) = λ G̃λ (0, y) m(dy),
l

and the Green function G̃λ is expressed in terms of ϕ̃λ and ψ̃λ satisfying (e).
The constants c1 , c2 , c3 and c4 are determined by the demand that ϕ̃λ and ψ̃λ are
continuous and satisfy
2
ϕ̃0λ (0+) − ϕ̃0λ (0−) = 2 ϕ̃λ (0),
a (0)
158 INTRODUCTION

and
2
ψ̃λ0 (0+) − ψ̃λ0 (0−) = 2 ψ̃λ (0),
a (0)
respectively, where a2 is the infinitesimal variance of X. See Karatzas and Shreve
(1984), where the case with Brownian motion is discussed. A general formula can
also be derived when X is stopped at the first hitting time Hz , cf. Pitman and Yor
(1999).
Functional 28. The joint distributions of the functionals 12–14 can be obtained
from the formula in No. II.19 by conditioning.
Functional 31. For this, see Borodin (2000).
1 W τ ∼ Exp(λ), independent of W

1. BROWNIAN MOTION

1. Exponential stopping


1.0.1 Ee−ατ =
+

1.0.2 P(τ > t) = e−λt

2 iβx
1.0.3 Ex eiβWτ = e
2 + 2

2
1.0.4 Ex eiβWt = eiβx−β t/2

√
 √
1.0.5 Px (Wτ ∈ dz) = √ e−|z−x| 2λ
dz
2

1 −(z−x)2 /2t
1.0.6 Px Wt ∈ dz = √ e dz
2t
√

2 −γx
1.1.1 Ex exp −γ sup Ws = √ e
0≤s≤τ + 2

√
Px sup Ws ≥ y = e(x−y) 2λ


1.1.2
0≤s≤τ
x≤y

2 p
Ex exp −γ sup Ws = e−γx+γ t/2 Erfc(γ t/2)


1.1.3
0≤s≤t

y−x

 
1.1.4 Px sup Ws ≥ y = Erfc √
0≤s≤t 2t
x≤y

2 √
√ e−γ(x∨z)−|z−x| 2λ dz


1.1.5 Ex exp −γ sup Ws ; Wτ ∈ dz =
0≤s≤τ + 2 2

√
 √
e(z+x−2y) 2λ


1.1.6 Px sup Ws ≥ y, Wτ ∈ dz = √ dz, x∨z ≤y
0≤s≤τ 2

© Springer Basel 2015 159

A.N. Borodin, P. Salminen, Handbook of Brownian Motion - Facts and Formulae,
Probability and Its Applications, DOI 10.1007/978-3-0348-8163-0_8
160 1. BROWNIAN MOTION

1 W τ ∼ Exp(λ), independent of W



1.1.7 Ex exp −γ sup Ws ; Wt ∈ dz
0≤s≤t

√

1 −(z−x)2 /2t 2

|z − x| t

= e−γ(x∨z) √ e − eγ|z−x|/2+γ t/8 Erfc √ + √ dz
2t 4 2t 2 2

1 −(|z−y|+y−x)2 /2t
1.1.8 Px sup Ws ≥ y, Wt ∈ dz = √ e dz
0≤s≤t 2t
x≤y

√

2 γx
1.2.1 Ex exp γ inf Ws = √ e
0≤s≤τ + 2

√
inf Ws ≤ y = e(y−x) 2λ


1.2.2 Px
0≤s≤τ
y≤x

2 p
Ex exp γ inf Ws = eγx+γ t/2 Erfc(γ t/2)


1.2.3
0≤s≤t

x−y

 
1.2.4 Px inf Ws ≤ y = Erfc √
0≤s≤t 2t
y≤x

2 √
√ eγ(x∧z)−|z−x| 2λ dz


1.2.5 Ex exp γ inf Ws ; Wτ ∈ dz =
0≤s≤τ + 2 2
√
 √
e(2y−z−x) 2λ


1.2.6 Px inf Ws ≤ y, Wτ ∈ dz = √ dz, y ≤x∧z
0≤s≤τ 2



1.2.7 Ex exp γ inf Ws ; Wt ∈ dz
0≤s≤t

√

1 −(z−x)2 /2t 2

|z − x| t

= eγ(x∧z) √ e − eγ|z−x|/2+γ t/8 Erfc √ + √ dz
2t 4 2t 2 2

1 −(|z−y|−y+x)2 /2t
1.2.8 Px inf Ws ≤ y, Wt ∈ dz = √ e dz
0≤s≤t 2t
y≤x

√
1.3.1 Ex e−γ`(τ,r) = 1 − √ e−|r−x| 2λ
+ 2
 1. EXPONENTIAL STOPPING 161

1 W τ ∼ Exp(λ), independent of W

√ √
1.3.2 Px `(τ, r) ∈ dy = 2λe−(y+|r−x|) 2λ dy
0<y

√
Px `(τ, r) = 0 = 1 − e−|r−x| 2λ


(1)

√
|r − x| |r − x| t
  2
 
1.3.3 Ex e−γ`(t,r) = 1 − Erfc √ + e|r−x|γ+γ t/2
Erfc √ + √
2t 2t 2
√
2 2
e−(y+|r−x|) /2t


1.3.4 Px `(t, r) ∈ dy = √ dy
t
0<y

|r − x|

 
(1) Px `(t, r) = 0 = 1 − Erfc √
2t
√
 √
=2 −(|z−r|+|r−x|)√2λ
p
Ex e−γ`(τ,r) ; Wτ ∈ dz = e−|z−x| 2λ

1.3.5 √ dz − √ e dz
2 + 2

√
Px `(τ, r) ∈ dy, Wτ ∈ dz = λe−(|z−r|+|r−x|+y) 2λ dydz


1.3.6
0<y

√ √
 √
 √
e−|z−x| 2λ
e−(|z−r|+|r−x|) 2λ


(1) Px `(τ, r) = 0, Wτ ∈ dz = √ dz − √ dz
2 2

1 −(z−x)2 /2t
Ex e−γ`(t,r) ; Wt ∈ dz =

1.3.7 √ e dz
2t
√
t |z − r| + |r − x|
 
− exp (|z − r| + |r − x|)γ + γ 2 t/2 Erfc


√ + √ dz
2 2 2t

1.3.8 Px `(t, r) ∈ dy, Wt ∈ dz
0<y
1 2
y + |z − r| + |r − x| e−(|z−r|+|r−x|+y) /2t dydz


= √
t 2t

1 2 1 −(|z−r|+|r−x|)2/2t
Px `(t, r) = 0, Wt ∈ dz = √ e−(z−x) /2tdz −


(1) √ e dz
2t 2t
162 1. BROWNIAN MOTION

1 W τ ∼ Exp(λ), independent of W

 Z τ

1.4.1 Ex exp −γ 1I[r,∞) Ws ds
0

√
 √
 
x≤r =1− 1− √ e(x−r) 2λ
+
√
   √
 
r≤x = − − √ e(r−x) 2λ+2γ
+ + +

Z τ

1.4.2 Px 1I[r,∞) Ws ds ∈ dy
0

√
 (x−r)√2λ−λy
x≤r = √ e dy
y
√
x−r 
  2
r≤x = λe−λy Erf √ dy + √ e−λy−(x−r) /2y dy
2y y

Z τ  √
1I[r,∞) Ws ds = 0 = 1 − e(x−r) 2λ


(1) Px
x≤r 0

 Z t


1.4.3 Ex exp −γ 1I[r,∞) Ws ds
0

1 1
Z t
r−x
  2
x≤r = Erf √ + e−γt eγs−(r−x) /2s ds
2t  s (t − s )
p
0

x−r 1 t 1

 Z
2
−γt
r≤x =e Erf √ + e−γs−(x−r) /2s ds
2t  0 s(t − s)
p

t

x=r = e−γt/2 I0
2

 1I (v)
[0;t] −(r−x)2 /2(t−v)
Z t  pv(t − v) e
  dv, x ≤ r


1.4.4 Px 1I[r,∞) Ws ds ∈ dv =
0
1I[0;t] (v) −(x−r)2 /2v
e dv, r≤x

 v (t − v )
 p

t
r−x
Z
  
(1) Px 1I[r,∞) Ws ds = 0 = 1 − Erfc √
0 2t
x≤r
 1. EXPONENTIAL STOPPING 163

1 W τ ∼ Exp(λ), independent of W
t
x−r
Z
  
(2) Px 1I[r,∞) Ws ds = t = 1 − Erfc √
0 2t
r≤x

n  Z τ
 o
1.4.5 Ex exp −γ 1I[r,∞) Ws ds ; Wτ ∈ dz =: λQ(4)
λ (z)dz
0

 √ 
2 1  (z+x−2r)√2λ
x≤r = √ e−|z−x| 2λ
dz + λ √ √ − √ e dz
2 2 + 2 + 2 2
z≤r
2 √ √
x≤r = √ √ e(x−r) 2λ+(r−z) 2λ+2γ dz
2 + 2 + 2
r≤z
2 √ √
r≤x = √ √ e(r−x) 2λ+2γ+(z−r) 2λ dz
2 + 2 + 2
z≤r
√

 √ 
2 2  √ 
r≤x = √ e−|z−x| 2λ+2γ
+ 1− √ √ e(2r−x−z) 2λ+2γ dz
2 + 2 2 + 2 + 2
r≤z

Z τ

1.4.6 Px 1I[r,∞) Ws ds ∈ dy, Wτ ∈ dz =: λBλ(4) (y, z)dydz
0

√ √

2 p  √
x≤r =λ √ e−λy − 2λ Erfc( λy) e(z+x−2r) 2λ dydz
y
z≤r

√
 2 (x−r)√2λ−λy−(z−r)2 /2y √ 3 (z+x−2r)√2λ z−r p
  
x≤r = √ e − 2λ e Erfc √ + λy dydz
y 2y
r≤z

√
 2 (z−r)√2λ−λy−(x−r)2/2y √ 3 (z+x−2r)√2λ x−r p
  
r≤x = √ e − 2λ e Erfc √ + λy dydz
y 2y
z≤r


1 −λy −(z−x)2 /2y 2
+ e−(2r−z−x) /2y


r≤x =λ √ e e
2y
r≤z
√ √ 
z + x − 2r p

− 2λe(z+x−2r) 2λ Erfc √ + λy dydz
2y

τ √
 √ √
Z 
e−|z−x| 2λ
− e(z+x−2r) 2λ



(1) Px 1I[r,∞) Ws ds = 0, Wτ ∈ dz = √ dz
0 2
x≤r
z≤r
164 1. BROWNIAN MOTION

1 W τ ∼ Exp(λ), independent of W

n  Z t

 o
1.4.7 Ex exp −γ 1I[r,∞) Ws ds ; Wt ∈ dz =: Vx(4) (t, z)dz
0

1 2 2
e−(z−x) /2t − e−(2r−x−z) /2t dz


x<r = √
2t
z<r
2r − x − z  1 − e− t   1 −(2r−x−z)2 /2t 
+ ∗ 3=2 e dz
2 t3=2 t

(z − r)(r − x)  1 − e− t   e− t −(z−r)2 /2t   1 −(r−x)2 /2t 

x<r = ∗ 3=2 e ∗ 3=2 e dz
(2)3=2 t3=2 t t
r<z

(x − r)(r − z )  1 − e− t   e− t −(x−r)2 /2t   1 −(r−z)2 /2t 

r<x = ∗ 3=2 e ∗ 3=2 e dz
(2)3=2 t3=2 t t
z<r

e− t −(z−x)2 /2t 2
− e−(x+z−2r) /2t dz


r<x = √ e
2t
r<z
x + z − 2r 1 − e− t
  − t
e
 
−(x+z−2r)2 /2t
+ ∗ e dz
2 t3=2 t3=2
1 − e− t
x=r = dz
(2)1=2 t3=2
z=r
z − r 1 − e− t
  − t
e
 2

x=r = 3= 2 ∗ 3=2 e−(z−r) /2t dz
2 t t
r<z
r − z 1 − e− t 1
   2

x=r = 3= 2 ∗ 3=2 e−(r−z) /2t dz
2 t t
z<r

Z t


1.4.8 Px 1I[r,∞) Ws ds ∈ dv, Wt ∈ dz =: Fx(4) t, v, z dvdz
v<t 0

Z ∞


= Bx(27) t − v, v, y, z dy dvdz for Bx(27) u, v, y, z see 1.27.6
0

√
(2r − x − z ) t − v 
(2r − x − z )2 
x≤r = √ exp − dvdz
t2 v 2(t − v)
z≤r
√

1 (2r − x − z )2  −(2r−x−z)2 /2t 
(2r − x − z ) v 
+ √ 3=2 − √ 5=2 e Erfc p dvdz
2t 2t 2t(t − v)
 1. EXPONENTIAL STOPPING 165

1 W τ ∼ Exp(λ), independent of W
√ √

(z − r) v (r − x) t − v  
(z − r)2 (r − x)2 
x≤r = 2 √ + 2 √ exp − − dvdz
t t − v t v 2v 2(t − v)
r≤z
√ √

1 (2r − x − z )2  2

(z − r) t − v (r − x) v 
+ √ 3=2 − √ 5=2 e−(2r−x−z) /2t Erfc √ +p dvdz
2t 2t 2tv 2t(t − v)

√ √

(x − r ) v (r − z ) t − v  
(x − r)2 (r − z )2 
r≤x = 2√ + 2√ exp − − dvdz
t t−v t v 2v 2(t − v)
z≤r
√ √

1 (2r − x − z )2  2

(x − r) t − v (r − z ) v 
+ √ 3=2 − √ 5=2 e−(2r−x−z) /2t Erfc √ +p dvdz
2t 2t 2tv 2t(t − v)

√
(x + z − 2r) v 
(x + z − 2r)2 
r≤x = √ exp − dvdz
t2 t − v 2v
r≤z
√

1 (x + z − 2r)2  −(x+z−2r)2 /2t 
(x + z − 2r) t − v 
+ √ 3=2 − √ 5=2 e Erfc √ dvdz
2t 2t 2tv

Z t

1  −(z−x)2 /2t −(2r−x−z)2 /2t 
(1) Px 1I[r,∞) Ws ds = 0, Wt ∈ dz = √ e −e dz
0 2t
x≤r
z≤r
Z t

1  −(z−x)2 /2t −(x+z−2r)2 /2t 
(2) Px 1I[r,∞) Ws ds = t, Wt ∈ dz = √ e −e dz
0 2t
r≤x
r≤z
Z t

1
(3) Px 1I[x,∞) Ws ds ∈ dv Wt = x = 1I[0,t] (v)dv
0
t

 Z τ

1.5.1 Ex exp −γ 1I(−∞,r] Ws ds
0
√
   √
 
x≤r = − − √ e(x−r) 2λ+2γ
+ + +
√
 √
 
r≤x =1− 1− √ e(r−x) 2λ
+

Z τ

1.5.2 Px 1I(−∞,r] Ws ds ∈ dy
0
√
r−x 
  2
x≤r = λe−λy Erf √ dy + √ e−λy−(r−x) /2y dy
2y y
√
 (r−x)√2λ−λy
r≤x = √ e dy
y
166 1. BROWNIAN MOTION

1 W τ ∼ Exp(λ), independent of W

Z τ  √
1I(−∞,r] Ws ds = 0 = 1 − e(r−x) 2λ


(1) Px
r≤x 0
 Z t


1.5.3 Ex exp −γ 1I(−∞,r] Ws ds
0

r−x 1 t 1
  Z
2
x≤r = e−γt Erf √ + e−γs−(r−x) /2s ds
2t  s(t − s)
p
0

1 1
Z t
x−r
  2
r≤x = Erf √ + e−γt eγs−(x−r) /2s ds
2t  s (t − s )
p
0

t

x=r = e−γt/2 I0
2
 1I (v)
[0;t] −(r−x)2 /2v
Z t   pv(t − v) e
 dv, x≤r


1.5.4 Px 1I(−∞,r] Ws ds ∈ dv =
0
1I[0;t] (v) −(x−r)2 /2(t−v)
e dv, r ≤ x

 v(t − v)
 p

t
x−r
Z
  
(1) Px 1I(−∞,r] Ws ds = 0 = 1 − Erfc √
0 2t
r≤x
t
r−x
Z
  
(2) Px 1I(−∞,r] Ws ds = t = 1 − Erfc √
0 2t
x≤r

n  Z τ
 o
1.5.5 Ex exp −γ 1I(−∞,r] Ws ds ; Wτ ∈ dz =: λQ(5)
λ (z)dz
0

√

 √ 
2 2  √ 
x≤r = √ e−|z−x| 2λ+2γ + 1 − √ √ e(z+x−2r) 2λ+2γ dz
2 + 2 2 + 2 + 2
z≤r

2 √ √
x≤r = √ √ e(x−r) 2λ+2γ+(r−z) 2λ dz
2 + 2 + 2
r≤z

2 √ √
r≤x = √ √ e(r−x) 2λ+(z−r) 2λ+2γ dz
2 + 2 + 2
z≤r

 √ 
2 1  (2r−x−z)√2λ
r≤x = √ e− 2λ|z−x|
dz + λ √ √ − √ e dz
2 2 + 2 + 2 2
r≤z
 1. EXPONENTIAL STOPPING 167

1 W τ ∼ Exp(λ), independent of W

Z τ

1.5.6 Px 1I(−∞,r] Ws ds ∈ dy, Wτ ∈ dz =: λBλ(5) (y, z)dydz
0


1 −λy −(z−x)2 /2y 2
+ e−(2r−x−z) /2y


x≤r =λ √ e e
2y
z≤r
√ √ 
2r − x − z p 
− 2λe(2r−x−z) 2λ Erfc √ + λy dydz
2y

√
 2 (r−z)√2λ−λy−(r−x)2 /2y √ 3 (2r−x−z)√2λ r−x p
  
x≤r = √ e − 2λ e Erfc √ + λy dydz
y 2y
r≤z

√
 2 (r−x)√2λ−λy−(z−r)2 /2y √ 3 (2r−x−z)√2λ r−z p
  
r≤x = √ e − 2λ e Erfc √ + λy dydz
y 2y
z≤r

√ √

2 p  √
r≤x =λ √ e−λy − 2λ Erfc( λy) e(2r−x−z) 2λ dydz
y
r≤z
τ √
 √ √
Z 
e−|z−x| 2λ
− e(2r−x−z) 2λ



(1) Px 1I(−∞,r] Ws ds = 0, Wτ ∈ dz = √ dz
0 2
r≤x
r≤z
n  Z t

 o
1.5.7 Ex exp −γ 1I(−∞,r] Ws ds ; Wt ∈ dz =: Vx(5) (t, z)dz
0

1 −γt −(x−z)2 /2t 2

− e−(2r−x−z) /2t dz


x<r = √ e e
2t
z<r
2r − x − z  1 − e− t   e− t −(2r−x−z)2 /2t 
+ ∗ 3=2 e dz
2 t3=2 t

(z − r)(r − x)  1 − e− t   e− t −(r−x)2 /2t   1 −(z−r)2 /2t 

x<r = ∗ 3=2 e ∗ 3=2 e dz
(2)3=2 t3=2 t t
r<z

(x − r)(r − z )  1 − e− t   e− t −(r−z)2 /2t   1 −(x−r)2 /2t 

r<x = ∗ 3=2 e ∗ 3=2 e dz
(2)3=2 t3=2 t t
z<r

1  −(x−z)2 /2t 2

r<x = √ e − e−(x+z−2r) /2t dz
2t
r<z
168 1. BROWNIAN MOTION

1 W τ ∼ Exp(λ), independent of W
x + z − 2r 1 − e− t 1
   2

+ 3= 2 ∗ 3=2 e−(x+z−2r) /2t dz
2 t t

1 − e− t
x=r = dz
(2)1=2 t3=2
z=r
z − r 1 − e− t 1
   2

x=r = 3= 2 ∗ 3=2 e−(z−r) /2t dz
2 t t
r<z
r − z 1 − e− t
  − t
e
 
−(r−z)2 /2t
x=r = ∗ e dz
2 t3=2 t3=2
z<r

Z t


1.5.8 Px 1I(−∞,r] Ws ds ∈ dv, Wt ∈ dz =: Fx(5) t, v, z dvdz
v<t 0

Z ∞


= Bx(27) v, t − v, y, z dy dvdz for Bx(27) u, v, y, z see 1.27.6
0

√
(2r − x − z ) v 
(2r − x − z )2 
x≤r = 2√ exp − dvdz
t t−v 2v
z≤r
√

1 (2r − x − z )2  −(2r−x−z)2 /2t 
(2r − x − z ) t − v 
+ √ 3=2 − √ 5=2 e Erfc √ dvdz
2t 2t 2tv

√ √

(z − r ) t − v (r − x) v  
(z − r)2 (r − x)2 
x≤r = 2√ + 2√ exp − − dvdz
t v t t−v 2(t − v) 2v
r≤z
√ √

1 (2r − x − z )2  2

(z − r) v (r − x) t − v 
+ √ 3=2 − √ 5=2 e−(2r−x−z) /2t Erfc p + √ dvdz
2t 2t 2t(t − v) 2tv

√ √

(x − r ) t − v (r − z ) v  
(x − r)2 (r − z )2 
r≤x = 2√ + 2√ exp − − dvdz
t v t t−v 2(t − v) 2v
z≤r
√ √

1 (2r − x − z )2  2

(x − r) v (r − z ) t − v 
+ √ 3=2 − √ 5=2 e−(2r−x−z) /2t Erfc p + √ dvdz
2t 2t 2t(t − v) 2tv

√
(x + z − 2r) t − v 
(x + z − 2r)2 
r≤x = √ exp − dvdz
t2 v 2(t − v)
r≤z
√

1 (x + z − 2r)2  −(x+z−2r)2 /2t 
(x + z − 2r) v 
+ √ 3=2 − √ 5=2 e Erfc p dvdz
2t 2t 2t(t − v)
 1. EXPONENTIAL STOPPING 169

1 W τ ∼ Exp(λ), independent of W
Z t

1  −(z−x)2 /2t −(x+z−2r)2 /2t
(1) Px 1I(−∞,r] Ws ds = 0, Wt ∈ dz = √ e −e dz
0 2t
r≤x
r≤z
Z t

1  −(z−x)2 /2t −(2r−x−z)2 /2t
(2) Px 1I(−∞,r] Ws ds = t, Wt ∈ dz = √ e −e dz
0 2t
x≤r
z≤r
Z t

1
(3) Px 1I(−∞,x] Ws ds ∈ dv Wt = x = 1I[0,t] (v)dv
0
t

 Z τ



1.6.1 Ex exp − p1I(−∞,r) Ws + q1I[r,∞) Ws ds
0
   √
 
x≤r = − − e(x−r) 2λ+2p
+p +p ( + p)( + q)
p

   √
 
r≤x = − − e(r−x) 2λ+2q
+q +q ( + p)( + q)
p

Z τ

Z τ

1.6.2 Px 1I(−∞,r) Ws ds ∈ du, 1I[r,∞) Ws ds ∈ dv
0 0

 2
√ e−λ(u+v)−(r−x) /2u dudv, x≤r
 uv

=
 √ e−λ(u+v)−(x−r)2 /2v dudv, r≤x
 uv

τ Z τ
r−x
Z   
1I(−∞,r) Ws ds ∈ du, 1I[r,∞) Ws ds = 0 = λe−λu Erf √



(1) Px du
0 0 2u
x≤r

τ Z τ
x−r
Z   
1I(−∞,r) Ws ds = 0, 1I[r,∞) Ws ds ∈ dv = λe−λv Erf √



(2) Px dv
0 0 2v
r≤x

 Z t



1.6.3 Ex exp − p1I(−∞,r) Ws + q1I[r,∞) Ws ds
0

1 1
Z t
r−x
  2
x≤r = e−pt Erf √ + e−qt e−(p−q)s−(r−x) /2s ds
2t  s (t s )
p
0 −

1 1
Z t
x−r
  2
r≤x = e−qt Erf √ + e−pt e−(q−p)s−(x−r) /2s ds
2t  s (t s )
p
0 −
170 1. BROWNIAN MOTION

1 W τ ∼ Exp(λ), independent of W

|p − q |t
 
x=r = e−(p+q)t/2 I0
2

Z t



1.6.4 Px p1I(−∞,r) Ws + q1I[r,∞) Ws ds ∈ dv
0

1I((q∧p)t;(q∨p)t) (v) 2
x≤r = e−(r−x) |p−q|/2|v−qt| dv
 |pt − v||v − qt|
p

1I((q∧p)t;(q∨p)t) (v) 2
r≤x = e−(x−r) |p−q|/2|pt−v| dv
 |pt − v||v − qt|
p

t
r−x
Z


  
(1) Px p1I(−∞,r) Ws + q1I[r,∞) Ws ds = pt = 1 − Erfc √
0 2t
x≤r
t
x−r
Z


  
(2) Px p1I(−∞,r) Ws + q1I[r,∞) Ws ds = qt = 1 − Erfc √
0 2t
r≤x

n  Z τ


 o
1.6.5 Ex exp − p1I(−∞,r) Ws + q1I[r,∞) Ws ds ; Wτ ∈ dz =: λQ(6)
λ (z)dz
0

 √
x≤r = √ e−|z−x| 2λ+2p dz
2 + 2p
z≤r

2 1  √
+λ √ √ −√ e(z+x−2r) 2λ+2p dz
2 + 2p + 2 + 2q 2 + 2p
2 √ √
x≤r =√ √ e(x−r) 2λ+2p+(r−z) 2λ+2q dz
2 + 2p + 2 + 2q
r≤z
2 √ √
r≤x = √ √ e(r−x) 2λ+2q+(z−r) 2λ+2p dz
2 + 2p + 2 + 2q
z≤r

 √
r≤x = √ e−|z−x| 2λ+2q
dz
2 + 2q
r≤z

2 1  √
+λ √ √ −√ e(2r−x−z) 2λ+2q dz
2 + 2p + 2 + 2q 2 + 2q

Z τ

Z τ

1.6.6 Px 1I(−∞,r) Ws ds ∈ du, 1I[r,∞) Ws ds ∈ dv, Wτ ∈ dz
0 0
 1. EXPONENTIAL STOPPING 171

1 W τ ∼ Exp(λ), independent of W

= λe−λ(u+v) Fx(4) (u + v, v, z)dudvdz for Fx(4) (t, v, z) see 1.4.8

Z τ

Z τ

(1) Px 1I(−∞,r) Ws ds ∈ du, 1I[r,∞) Ws ds = 0, Wτ ∈ dz
0 0

1 −(z−x)2 /2u 1 −(2r−x−z)2 /2u 
x≤r = λe−λu √ e −√ e dudz
2u 2u
z≤r
Z τ

Z τ

(2) Px 1I(−∞,r) Ws ds = 0, 1I[r,∞) Ws ds ∈ dv, Wτ ∈ dz
0 0

1 −(z−x)2 /2v 1 −(x+z−2r)2 /2v 
r≤x = λe−λv √ e −√ e dvdz
2v 2v
r≤z

n  Z t


 o
1.6.7 Ex exp − p1I(−∞,r) Ws + q1I[r,∞) Ws ds ; Wt ∈ dz =: Vx(6) (t, z)dz
0

e−pt −(z−x)2 /2t 2

− e−(2r−x−z) /2t dz


x<r = √ e
2t
z<r
2r − x − z  |e−pt − e−qt |   e−pt −(2r−x−z)2 /2t 
+ ∗ 3=2 e dz
2 |p − q |t3=2 t

(z − r)(r − x) |e−pt − e−qt |   e−qt −(z−r)2 /2t   e−pt −(r−x)2 /2t 
x<r = ∗ 3=2 e ∗ 3=2 e dz
(2)3=2 |p − q |t3=2 t t
r<z

(x − r)(r − z ) |e−pt − e−qt |   e−qt −(x−r)2 /2t   e−pt −(r−z)2 /2t 

r<x = ∗ 3=2 e ∗ 3=2 e dz
(2)3=2 |p − q |t3=2 t t
z<r

e−qt −(z−x)2 /2t 2

− e−(x+z−2r) /2t dz


r<x = √ e
2t
r<z
x + z − 2r |e−pt − e−qt |
  −qt
e
 2

+ 3= 2 ∗ 3=2 e−(x+z−2r) /2t dz
2 |p − q |t t
|e−pt − e−qt |
x=r = dz
(2)1=2 |p − q|t3=2
z=r

z − r |e−pt − e−qt |
  −qt
e
 
−(z−r)2 /2t
x=r = ∗ e dz
2 |p − q |t3=2 t3=2
r<z
172 1. BROWNIAN MOTION

1 W τ ∼ Exp(λ), independent of W

r − z |e−pt − e−qt |
  −pt
e
 
−(r−z)2 /2t
x=r = ∗ e
2 |p − q |t3=2 t3=2
z<r

Z t



1.6.8 Px p1I(−∞,r) Ws + q1I[r,∞) Ws ds ∈ dv, Wt ∈ dz
0

 1 
v − pt


(4)
F t, , z dvdz, pt < v < qt, for Fx(4) t, v, z see 1.4.8
 |p − q | q−p
x



= 1 
v − qt


Fx(5) t, , z dvdz, qt < v < pt, for Fx(5) t, v, z see 1.5.8


 | p − q | p−q

0, otherwise

Z t



(1) Px p1I(−∞,r) Ws + q1I[r,∞) Ws ds = pt, Wt ∈ dz
x≤r 0

1 −(z−x)2 /2t 1 2
z≤r = √ e dz − √ e−(2r−x−z) /2t dz
2t 2t

Z t



(2) Px p1I(−∞,r) Ws + q1I[r,∞) Ws ds = qt, Wt ∈ dz
r≤x 0

1 −(z−x)2 /2t 1 2
r≤z = √ e dz − √ e−(x+z−2r) /2t dz
2t 2t

 Z τ

1.7.1 Ex exp −γ 1I[r,u] Ws ds
0

√
sh((u − r) ( + )=2)e(x−r) 2
p
x≤r =1− √ √ √
 + (  + sh((u − r) ( + )=2) +  ch((u − r) ( + )=2))
p p

√
  ch((2x − u − r) ( + )=2))
p
r≤x = + √ √
+ ( + )(  + sh((u − r) ( + )=2) +  ch((u − r) ( + )=2))
p p
x≤u
√
sh((u − r) ( + )=2)e(u−x) 2
p
u≤x =1− √ √ √
 + (  + sh((u − r) ( + )=2) +  ch((u − r) ( + )=2))
p p

Z t

1.7.4 Px 1I[r,u] Ws ds ∈ dv
0
 1. EXPONENTIAL STOPPING 173

1 W τ ∼ Exp(λ), independent of W

∞
u−r
Z


x≤r =2 ht−v (0, y + r − x) ecv 0, 0, , 0, y dy dv
0
2

∞
u−r u−r
Z


+2 ht−v (1, y + r − x) escv , , 0, y dy dv
0
2 2

∞
u + r − 2x u − r
Z


r≤x =2 ht−v (0, y) eccv , , 0, y dy dv
0
2 2
x≤u

∞
u + r − 2x u − r u − r

Z
+2 ht−v (1, y) ecscv , , , y dy dv
0
2 2 2

∞
u−r
Z


u≤x =2 ht−v (0, y + x − u) ecv 0, 0, , 0, y dy dv
0
2

∞
u−r u−r
Z


+2 ht−v (1, y + x − u) escv , , 0, y dy dv
0
2 2

r−x
  
1 − Erfc √ , x≤r
2t

Z t
 

(1) Px 1I[r,u] Ws ds = 0 = 0, r≤x≤u
0  1 − Erfc x√− u , u ≤ x

  
2t

u + r − 2x u − r

Z t

(2) Px 1I[r,u] Ws ds = t = 1 − cc , r≤x≤u
2 2
et
0

2 t3 
 Z t  
1.8.3 Ex exp −γ Ws ds = exp −γxt +
0
6

√
Z t 
3 
3(y − xt)2 
1.8.4 Px Ws ds ∈ dy = √ exp − dy
0 2t3 2t3

τ
 Ai 21=3 −2=3 ( + |z |)
n  Z  o

1.8.5 E0 exp −γ |Ws |ds ; Wτ ∈ dz = − dz
(2 )1=3 Ai0 21=3 −2=3 


0

n  Z t  o
1 
(z − x)2 (z + x)t 2 t3 
1.8.7 Ex exp −γ Ws ds ; Wt ∈ dz = √ exp − − + dz
0 2t 2t 2 24
174 1. BROWNIAN MOTION

1 W τ ∼ Exp(λ), independent of W

Ai 0k + (2 )1=3 |z | e( =2) t
n  Z t  o ∞
2 1/3 0
X k
(1) E0 exp −γ |Ws |ds ; Wt ∈ dz = − dz
0 22=3 −1=3 0 Ai( 0 ) k k
k=1

√
Z t 
3 
(z − x)2 3(2y − (z + x)t)2 
1.8.8 Ex Ws ds ∈ dy, Wt ∈ dz = exp − − dydz
0
t2 2t 2t3
 2Z t  2

1 x sh(t )

1.9.3 Ex exp − Ws2 ds = exp −
2 ch(t ) 2 ch(t )
p
0

2 t
 Z 
2
(1) Ex exp −βWt − Ws2 ds
2 0
1 x2 ( sh(t ) + 2 ch(t ))
 
=p exp −
ch(t ) + 2 −1 sh(t ) 2(ch(t ) + 2 −1 sh(t ))

n  2Z t  o
(2) Ex exp − Ws2 ds ; 0 < inf Ws
2 0 0≤s≤t
0<x
 2  x√
1 x sh(t )
 
= exp − Erf p
ch(t ) 2 ch(t ) sh(2t )
p

Z t 
1.9.4 Px Ws2 ds ∈ dy = ecy (0, 1/2, t, 0, x2 /2)dy
0

√ ∞
Z t 
2 X (−1)k (k + 1=2) 
(4k + 1)t 
(1) P0 Ws2 ds < y = √ Erfc √
0
 k! 2 2y
k=0

n  2Z τ  o
1.9.5 Ex exp − Ws2 ds ; Wτ ∈ dz
2 0

  (1=2 + = ) √ √
√ D−1/2−λ/γ (−x 2γ)D−1/2−λ/γ (z 2γ)dz, x ≤ z



=
 (1= 2 + = ) √ √
√ D−1/2−λ/γ (x 2γ)D−1/2−λ/γ (−z 2γ)dz, z ≤ x




n  2Z t  o
1.9.7 Ex exp − Ws2 ds ; Wt ∈ dz
2 0
(x2 + z 2 ) ch(t ) − 2xz 
√ 
= exp − dz
2 sh(t ) 2 sh(t )
p
 1. EXPONENTIAL STOPPING 175

1 W τ ∼ Exp(λ), independent of W

Z t 
1.9.8 Px Ws2 ds ∈ dy, Wt ∈ dz = (2π)−1/2 eey (1/2, t, (x2 + z 2 )/2, −xz)dydz
0

Z t 
(1) P0 Ws2 ds < y Wt = 0
0
∞ √
X t (k + 1=2) 4k + 1 −(4k+1)2 t2/16y

(4k + 1)2 t2 
= √ 3=2 e K1/4
y k! 16y
k=0

Z τ 
1.10.2 Px e2βWs ds ∈ dy
0

√ √
e2 x
 2 x
 ( =| | 2) e
 
β 6= 0 = √ √ exp −βx − 2 M1/2,√λ/|β|√2 dy
| | 2y (1 + 2=| |) 4 y 2 2y

√
2| | e2 x 1 e2 x 
Z t    
1.10.4 Px e2βWs ds ∈ dy = √ exp −βx − 2 m2β 2 t − , 2 dy
0
y 4 y 2 4 y
β 6= 0

n  Z τ  o
1.10.5 Ex exp −γ e2βWs ds ; Wτ ∈ dz
0

√ √
2 2 βx  √ 2 βz 

 I√2λ/β
 e K 2λ/β e dz, x ≤ z
0<β = √ √
 2 K√ 2 βx √ 2 βz
 
e I 2λ/β e dz, z ≤ x

2λ/β

√ √
2 2 2
  
 | | K√2λ/|β| | | eβx I√2λ/|β| | | eβz dz, x ≤ z

β<0 = √ √
 2 I√
 2 βx  √
e K
2 βz 
e dz, z ≤ x
2λ/|β|
| | 2λ/|β|
| | | |

 2 x 2 z  (x+z) 
e +e
τ
 e
Z 
1.10.6 Px e2βWs ds ∈ dy, Wτ ∈ dz = exp − 2 I√2λ/|β| 2 y dydz
0
| |y 2 y
β 6= 0

√ √
2 βx 2 βz 
n  Z t  o
1.10.7 Ex exp −γ e2βWs ds ; Wt ∈ dz = |β| kiβ 2 t/2 e , e dz
0
| | | |
β 6= 0
 2 x 2 z  (x+z) 
e +e
t
e
Z
| |

1.10.8 Px e2βWs ds ∈ dy, Wt ∈ dz = exp − i 2 y dydz
2y 2 2y β 2 t/2

β 6= 0 0
176 1. BROWNIAN MOTION

1 W τ ∼ Exp(λ), independent of W

√
h 2I1 (h =2)
p


1.11.2 Px sup `(τ, y) ≥ h =
sh2 (h =2)I0 (h =2)
p p
y∈R

∞
X 1 2 2
e−2j0,k t/h


1.11.4 P sup `(t, y) < h = 4
y∈R sin2 j0;k
k=1
∞ 
4tkJ1 (k) J (k) J 2 (k)
+ 1 − 12
X  2 2 2
+4 2 − 1 e−2π k t/h
h J0 (k) kJ0 (k) J0 (k)
k=1

√

1.12.1 Ex e−γ Ȟ(τ ) = Ex e−γ Ĥ(τ ) = √
+

√


e−v
1.12.2 Px Ȟ(τ ) ∈ dv = Px Ĥ(τ ) ∈ dv = √ dv
v

t
 
1.12.3 Ex e−γ Ȟ(t) = Ex e−γ Ĥ(t) = e−γt/2 I0
2

2  √ 

v
1.12.4 Px Ȟ(t) < v = Px Ĥ(t) < v = arctg √ , 0≤v≤t
 t−v

√
 2 √

√ √ e(x−z) 2λ+2γ
dz, x≤z
+ + 


−γ Ȟ(τ )

1.12.5 Ex e ; Wτ ∈ dz = √
 √  2 √ e(z−x) 2λ dz,
√
z≤x

+ + 

√
 2 √

√ 2λ
dz, √ e(x−z) x≤z
+ + 


−γ Ĥ(τ )

(1) Ex e ; Wτ ∈ dz = √
 √  2 √ e(z−x) 2λ+2γ dz,
√
z≤x

+ + 


1.12.6 Px Ȟ(τ ) ∈ dv, Wτ ∈ dz

√
 2 −λv−(z−x)2 /2v √ √
z−x √
  
x≤z = √ e − λ3/2 2e(z−x) 2λ Erfc √ + λv dvdz
v 2v
√ √ √

 2 −λv−(x−z)√2λ √

z≤x = √ e − λ3/2 2e−(x−z) 2λ Erfc λv dvdz
v

(1) Px Ĥ(τ ) ∈ dv, Wτ ∈ dz
 1. EXPONENTIAL STOPPING 177

1 W τ ∼ Exp(λ), independent of W

√ √ √

 2 −λv−(z−x)√2λ √

x≤z = √ e − λ3/2 2e−(z−x) 2λ Erfc λv dvdz
v

√
 2 −λv−(x−z)2 /2v √ √
x−z √
  
z≤x = √ e − λ3/2 2e(x−z) 2λ Erfc √ + λv dvdz
v 2v

Ex e−γ Ȟ(t) ; Wt ∈ dz

1.12.7

1 − e− t   z − x −γt−(z−x)2 /2t 
x≤z = √ ∗ √ 3=2 e dz
2 t3=2 2t

1 − e− t   x − z −(x−z)2 /2t 
z≤x = √ ∗ √ 3=2 e dz
2 t3=2 2t

Ex e−γ Ĥ(t) ; Wt ∈ dz

(1)

1 − e− t   z − x −(z−x)2 /2t 
x≤z = √ ∗ √ 3=2 e dz
2 t3=2 2t

1 − e− t   x − z −γt−(x−z)2 /2t 
z≤x = √ ∗ √ 3=2 e dz
2 t3=2 2t

∞
(y − x)(y − z ) (y − x)2 (y − z )2 
Z 


1.12.8 Px Ȟ(t)∈dv, Wt ∈dz = 3=2 exp − − dy dvdz
x∨z  (v (t − v ))
2v 2(t − v)
v<t

x∧z
(x − y)(z − y) (x − y)2 (z − y)2 
Z 


(1) Px Ĥ(t)∈dv, Wt ∈dz = exp − − dy dvdz
−∞ (v(t − v))3=2 2v 2(t − v)
v<t

√ √
Ex e−γ Ȟ(τ ) ; sup Ws ∈ dy = 2λe−(y−x) 2λ+2γ dy

1.13.1
0≤s≤τ
x<y

√

2(y − x) 
(y − x)2 
1.13.2 Px Ȟ(τ ) ∈ dv, sup Ws ∈ dy = √ exp −λv − dvdy
0≤s≤τ 2v3=2 2v
x<y


y−x

(y − x)2  1
Ex e−γ Ȟ(t) ; sup Ws ∈ dy =

1.13.3 3= 2 exp −γt − ∗ √ dy
0≤s≤t t 2t t
x<y

1I[0;t] (v)(y − x) 
(y − x)2 
1.13.4 Px Ȟ(t) ∈ dv, sup Ws ∈ dy = 3=2 exp − dvdy
0≤s≤t v (t − v)1=2 2v
x<y
178 1. BROWNIAN MOTION

1 W τ ∼ Exp(λ), independent of W

√ √
Ex e−γ Ȟ(τ ) ; sup Ws ∈ dy, Wτ ∈ dz = 2λe(x−y) 2λ+2γ e(z−y) 2λ dydz

1.13.5
0≤s≤τ
x<y
z<y


1.13.6 Px Ȟ(τ ) ∈ dv, sup Ws ∈ dy, Wτ ∈ dz
0≤s≤τ
x<y
z<y
2(y − x) 
(y − x)2  −(y−z)√2λ
= √ exp −λv − e dvdydz
2v3=2 2v

Ex e−γ Ȟ(t) ; sup Ws ∈ dy, Wt ∈ dz


1.13.7
0≤s≤t
x<y
z<y
(y − x)(y − z )  1 
(y − x)2   1 
(y − z )2 
= exp −γt − ∗ exp − dydz
 t3=2 2t t3=2 2t

1.13.8 Px Ȟ(t) ∈ dv, sup Ws ∈ dy, Wt ∈ dz
0≤s≤t
v<t

(y − x)(y − z ) 
(y − x)2 (y − z )2 
= exp − − dy dvdz, x∨z <y
(v(t − v))3=2 2v 2(t − v)

√ √
Ex e−γ Ĥ(τ ) ; inf Ws ∈ dy = 2λe−(x−y) 2λ+2γ dy

1.14.1
0≤s≤τ
y<x

√

2(x − y) 
(x − y)2 
1.14.2 Px Ĥ(τ ) ∈ dv, inf Ws ∈ dy = √
3= 2 exp −λv − dvdy
0≤s≤τ 2v 2v
y<x


x−y

(x − y)2  1
Ex e−γ Ĥ(t) ; inf Ws ∈ dy =

1.14.3 exp −γt − ∗ √ dy
0≤s≤t t3=2 2t t
y<x

1I[0;t] (v)(x − y) 
(x − y)2 
1.14.4 Px Ĥ(t) ∈ dv, inf Ws ∈ dy = 3=2 1= 2 exp − dvdy
0≤s≤t v (t − v) 2v
y<x

√ √
Ex e−γ Ĥ(τ ) ; inf Ws ∈ dy, Wτ ∈ dz = 2λe(y−x) 2λ+2γ e(y−z) 2λ dydz

1.14.5
0≤s≤τ
y<x
y<z
 1. EXPONENTIAL STOPPING 179

1 W τ ∼ Exp(λ), independent of W


1.14.6 Px Ĥ(τ ) ∈ dv, inf Ws ∈ dy, Wτ ∈ dz
0≤s≤τ
y<x
y<z
2(x − y) 
(x − y)2  −(z−y)√2λ
= √ exp −λv − e dvdydz
2v3=2 2v
Ex e−γ Ĥ(t) ; inf Ws ∈ dy, Wt ∈ dz

1.14.7
0≤s≤t
y<x
y<z
(x − y)(z − y)  1 
(x − y)2   1 
(z − y)2 
= exp −γt − ∗ exp − dydz
 t3=2 2t t3=2 2t

1.14.8 Px Ĥ(t) ∈ dv, inf Ws ∈ dy, Wt ∈ dz
0≤s≤t
v<t
(x − y)(z − y) 
(x − y)2 (z − y)2 
= exp − − dy dvdz, y <x∧z
(v(t − v))3=2 2v 2(t − v)

ch((b + a − 2x) =2)

p


1.15.2 Px a < inf Ws , sup Ws < b = 1 −
ch((b − a) =2)
p
0≤s≤τ 0≤s≤τ

sh2 (yp=2)
(1) Px sup Ws − inf Ws < y = 2 p
0≤s≤τ 0≤s≤τ ch (y =2)

1.15.4 Px a < inf Ws , sup Ws < b
0≤s≤t 0≤s≤t

∞ Z b
1 X 2 2
e−(z−x+2k(b−a)) /2t − e−(z+x−2a+2k(b−a)) /2t dz


= √
2t a
k=−∞
∞
ky
X  
(−1)k k Erfc √


(1) Px sup Ws − inf Ws < y = 1 + 4
0≤s≤t 0≤s≤t 2t
k=1

1.15.6 Px a < inf Ws , sup Ws < b, Wτ ∈ dz
0≤s≤τ 0≤s≤τ

√ √ √
 ch((b − a − |z − x|) 2) − ch((b + a − z − x) 2)


= √ √ dz, a < x ∧ z, x ∨ z < b
2 sh((b − a) 2)

(1) Px sup Ws − inf Ws < y, Wτ ∈ dz for |z − x| < y
0≤s≤τ 0≤s≤τ
180 1. BROWNIAN MOTION

1 W τ ∼ Exp(λ), independent of W

 sh (y − |z − x|)√2
ch(y√2
√


= 2 √ √ − (y − |z − x|) ch |z − x| 2λ dz
sh (y 2) 2

1.15.8 Px a < inf Ws , sup Ws < b , Wt ∈ dz
0≤s≤t 0≤s≤t

∞
1 X −(z−x+2k(b−a))2 /2t 2
− e−(z+x−2a+2k(b−a)) /2t dz, a < x ∧ z,


= √ e
2t x∨z <b
k=−∞

(1) Px sup |Ws | ≥ b, Wt ∈ dz
0≤s≤t

∞
1 X −(z−x+4kb)2 /2t 2
− e−(z+x+2b+4kb) /2t dz,


= √ e −b < x ∧ z,
2t x∨z <b
k=−∞

(2) Px sup Ws − inf Ws < y, Wt ∈ dz for |z − x| < y
0≤s≤t 0≤s≤t

∞
1 X 2k  2
= √ 2k + 1 − (y − |z − x|)(|z − x| + 2ky) e−(|z−x|+2ky) /2t dz
2t t
k=−∞

√
Ex e−γ`(τ,r) ; sup Ws < b = 1 − e−(b−x) 2λ

1.16.1
0≤s≤τ
r<b
√
(1 − e−(b−r) 2 ) √
−|x−r| 2λ
√

−(2b−r−x) 2λ
−√ √ e − e
2 + (1 − e−2(b−r) 2 )

1.16.2 Px `(τ, r) ∈ dy, sup Ws < b
0≤s≤τ
r<b

√ √ √ √
2(e−|x−r| 2 − e−(2b−r−x) 2 ) 
y 2

0<y = √ √ exp − √ dy
(1 + e−(b−r) 2 )(1 − e−2(b−r) 2 ) 1 − e−2(b−r) 2
√ √
√
e−|x−r| 2 − e−(2b−r−x) 2
Px `(τ, r) = 0, sup Ws < b = 1 − e−(b−x) 2λ −


(1) √
0≤s≤τ 1 + e−(b−r) 2
 1. EXPONENTIAL STOPPING 181

1 W τ ∼ Exp(λ), independent of W


1 
1.16.4 Px `(t, r) ∈ dy, sup Ws < b r∨x≤b
2 0≤s≤t

0<y = (est (−1, 2, b − r, y + |x − r| − 2b + 2r, y) + est (−1, 2, b − r, y + b − x, y)

− est (−1, 2, b − r, y + |x − r| − b + r, y) − est (−1, 2, b − r, y + r − x, y))dy

(1) Px `(t, r) = 0, sup Ws < b r≤x≤b
0≤s≤t
∞ Z b
1 X 2 2
e−(z−x+2k(b−r)) /2t − e−(z+x−2r+2k(b−r)) /2t dz


= √
2t rk=−∞

√ n
 √ √
Ex e−γ`(τ,r) ; sup Ws < b,Wτ ∈ dz = √ e−|z−x| 2λ − e(x+z−2b) 2λ

1.16.5
0≤s≤τ 2
r<b
z<b
√ √
(e−|x−r| 2 − e−(2b−x−r) 2 ) −|z−r|√2λ √
o
−(2b−z−r) 2λ
− √ √ e − e dz
2 + (1 − e−2(b−r) 2 )
√
y 2

 
1.16.6 Px `(τ, r) ∈ dy, sup Ws < b, Wτ ∈ dz = exp − √

r<b 0≤s≤τ 1 − e−2(b−r) 2

z<b
√ √ √ √
(e−|x−r| 2 − e−(2b−r−x) 2 )(e−|z−r| 2 − e−(2b−z−r) 2 )
0<y × √ dydz
(1 − e−2(b−r) 2 )2
√
 √ √
n
e−|z−x| 2λ − e(x+z−2b) 2λ


(1) Px `(τ, r) = 0, sup Ws < b, Wτ ∈ dz = √
0≤s≤τ 2
√ √
−|x−r | 2
e − e−(2b−x−r) 2
√ √
o
− √ e−|z−r| 2λ
− e−(2b−z−r) 2λ
dz
1 − e−2(b−r) 2


1 
1.16.8 Px `(t, r) ∈ dy, sup Ws < b, Wt ∈ dz =: Fx(16) (t, y, z)dydz
2 0≤s≤t

for r ∨ x ∨ z < b
1
0<y = (est (0, 2, b − r, y + |x − r| + |z − r| − 2b + 2r, y)
2
+ est (0, 2, b − r, y + 2b − x − z, y) − est (0, 2, b − r, y + |x − r| + r − z, y)

− est (0, 2, b − r, y + |z − r| + r − x, y))dydz

182 1. BROWNIAN MOTION

1 W τ ∼ Exp(λ), independent of W

(1) Px `(t, r) = 0, sup Ws < b, Wt ∈ dz r < x ∧ z, x ∨ z < b
0≤s≤t
∞
1 X −(z−x+2k(b−r))2 /2t 2
− e−(z+x−2r+2k(b−r)) /2t dz


= √ e
2t
k=−∞

√
Ex e−γ`(τ,r) ; a < inf Ws = 1 − e−(x−a) 2λ

1.17.1
0≤s≤τ
a<r
√
(1 − e−(r−a) 2 ) √
−|x−r| 2λ
√

−(x+r−2a) 2λ
−√ √ e − e
2 + (1 − e−2(r−a) 2 )

1.17.2 Px `(τ, r) ∈ dy, a < inf Ws
0≤s≤τ
a<r
√ √ √ √
2(e−|x−r| 2 − e−(x+r−2a) 2 ) 
y 2

0<y = √ √ exp − √ dy
(1 + e−(r−a) 2 )(1 − e−2(r−a) 2 ) 1 − e−2(r−a) 2
√ √
√
e−|x−r| 2 − e−(x+r−2a) 2
Px `(τ, r) = 0, a < inf Ws = 1 − e−(x−a) 2λ −


(1) √
0≤s≤τ 1 + e−(r−a) 2


1 
1.17.4 Px `(t, r) ∈ dy, a < inf Ws a<x∧z∧r
2 0≤s≤t

0<y = (est (−1, 2, r − a, y + |x − r| − 2r + 2a, y) + est (−1, 2, r − a, y + x − a, y)

− est (−1, 2, r − a, y + |x − r| − r + a, y) − est (−1, 2, r − a, y + x − r, y))dy

(1) Px `(t, r) = 0, a < inf Ws a≤x≤r
0≤s≤t
∞ Z r
1 X 2 2
e−(z−x+2k(r−a)) /2t − e−(z+x−2a+2k(r−a)) /2t dz


= √
2t a
k=−∞

√
 √ √
n
Ex e−γ`(τ,r) ; a < inf Ws , Wτ ∈ dz = e−|z−x| 2λ − e(2a−x−z) 2λ


1.17.5 √
0≤s≤τ 2
a<r
a<z
√ √
(e−|x−r| 2 − e−(x+r−2a) 2 ) −|z−r|√2λ √
o
−(z+r−2a) 2λ
− √ √ e − e dz
2 + (1 − e−2(r−a) 2 )
 1. EXPONENTIAL STOPPING 183

1 W τ ∼ Exp(λ), independent of W

√
y 2

 
1.17.6 Px `(τ, r) ∈ dy, a < inf Ws , Wτ ∈ dz = exp − √

a<r
0≤s≤τ 1 − e−2(r−a) 2
a<z
√ √ √ √
(e−|x−r| 2 − e−(x+r−2a) 2 )(e−|z−r| 2 − e−(z+r−2a) 2 )
0<y × √ dydz
(1 − e−2(r−a) 2 )2
√
 √ √
n
e−|z−x| 2λ − e(2a−x−z) 2λ


(1) Px `(τ, r) = 0, a < inf Ws , Wτ ∈ dz = √
0≤s≤τ 2
√ √
e−|x−r| 2 − e−(x+r−2a) 2 −|z−r|√2λ √
o
−(z+r−2a) 2λ
− √ e − e dz
1 − e−2(r−a) 2


1 
(17)
1.17.8 Px `(t, r) ∈ dy, a < inf Ws , Wt ∈ dz =: Fa,x (t, y, z)dydz
2 0≤s≤t

for a<x∧z∧r
1
0<y = (est (0, 2, r − a, y + |x − r| + |z − r| − 2r + 2a, y)
2
+ est (0, 2, r − a, y + x + z − 2a, y) − est (0, 2, r − a, y + |x − r| + z − r, y)

− est (0, 2, r − a, y + |z − r| + x − r, y))dydz

(1) Px `(t, r) = 0, a < inf Ws , Wt ∈ dz a < x ∧ z, x ∨ z < r
0≤s≤t
∞
1 X −(z−x+2k(r−a))2 /2t 2
− e−(z+x−2a+2k(r−a)) /2t dz


= √ e
2t
k=−∞

1.18.1 Ex e−γ`(τ,r)−η`(τ,u) = 1

√ √

√ √ √

√
2 +  1 − e−|u−r| 2 e−|r−x| 2 +  2 + 1 − e−|u−r| 2 e−|u−x| 2
− √ √ √
( 2 + )( 2 + ) − e−2|u−r| 2

Ex e−γ`(τ,r)−η`(τ,u) Wτ = z

1.18.5

√ √ √
√
( 2 + )e−|r−x| 2 − e−(|u−x|+|u−r|) 2 e−(|z−r|−|z−x|) 2
=1− √ √ √
( 2 + )( 2 + ) − e−2|u−r| 2
184 1. BROWNIAN MOTION

1 W τ ∼ Exp(λ), independent of W

√ √ √ √
 ( 2 + )e−|u−x| 2 − e−(|r−x|+|u−r|) 2 e−(|z−u|−|z−x|) 2


− √ √ √
( 2 + )( 2 + ) − e−2|u−r| 2

Ex e−γ`(τ,r) `(τ, u) = g, Wτ = z

(1)

for r < u ≤ x ∧ z or x ∨ z ≤ u < r

 √ √


−1
= exp −gγe−2|r−u| 2λ 1 + √ 1 − e−2|r−u| 2λ ,
2
for x ∧ z ≤ r < u ≤ x ∨ z or x ∧ z ≤ u < r ≤ x ∨ z
√


−1
= 1 + √ 1 − e−2|r−u| 2λ
2
 √ √


−1
× exp −gγe−2|r−u| 2λ 1 + √ 1 − e−2|r−u| 2λ
2

1.18.6 Px `(τ, r) ∈ dy `(τ, u) = g, Wτ = z

for r < u ≤ x ∧ z or x ∨ z ≤ u < r

√  √ ye|r−u|√2 + ge−|r−u|√2
  √
g 2gy 
=√ √ exp − √ √ I1 √ dy
2y sh(|r − u| 2) 2 sh(|r − u| 2) sh(|r − u| 2)

for x ∧ z ≤ r < u ≤ x ∨ z or x ∧ z ≤ u < r ≤ x ∨ z

√ √  √ ye|r−u|√2 + ge−|r−u|√2
  √
e|r−u| 2 2gy 
=√ √ exp − √ √ I0 √ dy
2 sh(|r − u| 2) 2 sh(|r − u| 2) sh(|r − u| 2)

√
g 2
  
(1) Px `(τ, r) = 0 `(τ, u) = g, Wτ = z = exp − 2|r−u|√2
e −1

for r < u ≤ x ∧ z or x ∨ z ≤ u < r

(18)
1.18.8 Px `(t, r) ∈ dy, `(t, u) ∈ dg, Wt ∈ dz =: Fx,t (y, g, z)dydgdz

for r < u ≤ x ∧ z or x ∨ z ≤ u < r

√
g √
= √ ist (1, |r − u|, |z + x − 2u| + (y + g)/2, (y + g)/2, gy/2)dydgdz
2 y
 1. EXPONENTIAL STOPPING 185

1 W τ ∼ Exp(λ), independent of W

for x ∧ z ≤ r < u ≤ x ∨ z or x ∧ z ≤ u < r ≤ x ∨ z

1 √
= ist (0, |r − u|, |z − x| − |r − u| + (y + g)/2, (y + g)/2, gy/2)dydgdz
2

τ  x√2 sh(yp=2) 
ds x
Z 
1.19.2 Px ∈ dy, 0 < inf Ws = 2 exp − dy
Ws ch (y =2) ch(y =2)
p p
0 0≤s≤τ
0<x

t
ds
Z 
1.19.4 Px ∈ dy, 0 < inf Ws = x ect (0, 2, y/2, 0, x)dy
0
Ws 0≤s≤t
0<x

τ
ds
n  Z  o
1.19.5 Ex exp −γ ; 0 < inf Ws , Wτ ∈ dz =: λQ(19)
λ (z)dz
0
Ws 0≤s≤τ
0<x

√ √ √ √

x≤z = √ Γ(1 + γ/ 2λ)M−γ/√2λ,1/2 (2x 2λ)W−γ/√2λ,1/2 (2z 2λ)dz
2
√ √ √ √

z≤x = √ Γ(1 + γ/ 2λ)W−γ/√2λ,1/2 (2x 2λ)M−γ/√2λ,1/2 (2z 2λ)dz
2

τ
ds
Z 
1.19.6 Px ∈ dy, 0 < inf Ws , Wτ ∈ dz
0
Ws 0≤s≤τ
0<x

√  (x + z )√2 ch(yp=2)   √
 2xz 2 2xz 
= exp − I1 dydz
sh(y =2) sh(y =2) sh(y =2)
p p p

t
ds
Z  √ √
1.19.8 Px ∈ dy, 0 < inf Ws , Wt ∈ dz = xz ist (1, y/2, 0, x + z, xz)dydz
0
Ws 0≤s≤t
0<x

 2Z t
ds
n  o
1.20.3 Ex exp − 2 ; 0 < inf Ws
2 0
Ws 0≤s≤t
0<x

(2t)1=4 (3=4 + 1 + 4 2 =4) −x2 /4t  2

x
p
= √ e M √ 2
x ( 2 + 1=4 + 1) −1/4, 4γ +1/4 2t
p

Z t
ds

2(2t)1=4 −y/8−x2 /4t 
1 x2 
1.20.4 Px ∈ dy, 0 < inf W = √ e m , dy
0
Ws2 0≤s≤t
s
x 2y
4 4t
0<x
186 1. BROWNIAN MOTION

1 W τ ∼ Exp(λ), independent of W

 2Z τ
ds
n  o
1.20.5 Ex exp − ; 0 < inf W , W ∈ dz
2 0
Ws2 0≤s≤τ
s τ
0<x

 √ √
√

 2λ xzI√γ 2 +1/4 x 2λ K√γ 2 +1/4 z 2λ dz, 0<x≤z
= √
√

 2λ√xzK√ 2 x 2λ I√ 2 z 2λ dz, z≤x
γ +1/4 γ +1/4

τ
ds
Z 
1.20.6 Px ∈ dy, 0 < inf Ws , Wτ ∈ dz
0
Ws2 0≤s≤τ
0<x

√ √ √
= λ xze−y/8 kiy/2 (x 2λ, z 2λ)dydz

 2Z t
ds
n  o
1.20.7 Ex exp − ; 0 < inf W , W ∈ dz
2 0
Ws2 0≤s≤t
s t
0<x

√
xz −(x2 +z2 )/2t √ xz
 
= e I γ 2 +1/4 dz
t t

t
ds
Z 
1.20.8 Px ∈ dy, 0 < inf Ws , Wt ∈ dz
0
Ws2 0≤s≤t
0<x

√
xz 2
+z 2 )/2t−y/8 xz

= e−(x iy/2 dydz
2t t

p2 q2 Ws2

n  Z t  o
1.21.3 Ex exp − + ds ; 0 < inf W
0
2Ws2 2 0≤s≤t
s
0<x

(2 sh(tq))1=4 (3=4 + 4p2 + 1=4)  2

x q ch(2tq)
 2 
x q
p 
= √ exp − M− 1 ,√4p2 +1/4
x(q ch(tq))1= 4 2
(1 + p + 1=4) 2 sh(2tq) sh(2tq)
p
4

 2Z t t
q ds
n  Z o
2
(1) Ex exp − Ws ds ; ∈ dy, 0 < inf Ws
2 0 0
Ws2 0≤s≤t

2(2 sh(tq))1=4 
y x2 q ch(2tq)
 
1 x2 q

0<x = √ exp − − m , dy
x(q ch(tq))1=4 8 2 sh(2tq) 2y
4 2 sh(2tq)
 1. EXPONENTIAL STOPPING 187

1 W τ ∼ Exp(λ), independent of W

p2 q2 Ws2

n  Z τ  o
1.21.5 Ex exp − 2 + ds ; 0 < inf Ws , Wτ ∈ dz =: λQ(21)
λ (z)dz
0
2Ws 2 0≤s≤τ
0<x

 ((1 + p2 + 1=4 + =q)=2)

p
x≤z = √ M−λ/2q,√4p2 +1/4 (qx2 )W−λ/2q,√4p2 +1/4 (qz 2 )dz
q (1 + p2 + 1=4) xz
p

 ((1 + p2 + 1=4 + =q)=2)

p
z≤x = √ W−λ/2q,√4p2 +1/4 (qx2 )M−λ/2q,√4p2 +1/4 (qz 2 )dz
q (1 + p2 + 1=4) xz
p

p2 q2 Ws2

n  Z t  o
1.21.7 Ex exp − + ds ; 0 < inf W , W ∈ dz
0
2Ws2 2 0≤s≤t
s t
0<x

√
q xz

(x2 + z 2 )q ch(tq)  √ 
xzq

= exp − I p2 +1/4 dz
sh(tq) 2 sh(tq) sh(tq)

 2Z t t
q ds
n  Z o
(1) Ex exp − Ws2 ds ; ∈ dy, 0 < inf Ws , Wt ∈ dz
2 0 0
Ws2 0≤s≤t

√
q xz

y (x2 + z 2 )q ch(tq)  
xzq

= exp − − iy/2 dydz
2 sh(tq) 8 2 sh(tq) sh(tq)

 2Z t  Z t
p ds
n o
2
(2) Ex exp − ; W ds ∈ dy, 0 < inf W , W ∈ dz
2 0
Ws2 0
s
0≤s≤t
s t

√ p
= xz isy ( p2 + 1/4, t, 0, (x2 + z 2 )/2, xz/2)dydz

t t
ds
Z Z 
1.21.8 Px Ws2 ds ∈ dg, ∈ dy, 0 < inf Ws , Wt ∈ dz
0 0
Ws2 0≤s≤t
0<x

√
= 2−1 xze−y/8 eig (y/2, t, (x2 + z 2 )/2, xz)dgdydz

 2Z τ  Z τ o √ sh(yp=2)
p ds ds
n
1.22.1 Ex exp − ; ∈ dy, 0 < inf Ws = √
2 Ws2 Ws 2 ch(y =2)
p
0 0 0≤s≤τ
(1)

√ √ √
(3=2 + p2 + 1=4) x 2 ch(y 2) 2x 2 
p   
0<x × exp − √ M √ 2 √ dy
(1 + 4p2 + 1) sh(y 2) −1, p +1/4 sh(y 2)
p
188 1. BROWNIAN MOTION

1 W τ ∼ Exp(λ), independent of W

τ τ
ds ds
Z Z 
1.22.2 Px ∈ dg, ∈ dy, 0 < inf Ws
0
Ws2 0
Ws 0≤s≤τ
0<x
√ √ √ √
 sh(y =2) g x 2 ch(y 2) x 2
p    
= √ exp − − √ mg/2 1, √ dgdy
2 2 ch(y =2) 8 sh(y 2) sh(y 2)
p

p2
τ
q

n  Z  o
1.22.5 Ex exp − + ds ; 0 < inf W , W ∈ dz =: λQ(22)
λ (z)dz
0
2Ws2 Ws 0≤s≤τ
s τ
0<x
√ √
 ( p2 + 1=4 + 1=2 + q= 2)
p
= √
2 (2 p2 + 1=4 + 1)
p

√ √
 M−q/√2λ,√p2 +1/4 (2x 2λ)W−q/√2λ,√p2 +1/4 (2z 2λ)dz,

x≤z
× √ √
W √ √ 2 (2x 2λ)M √ √ 2 (2z 2λ)dz, z≤x
−q/ 2λ, p +1/4 −q/ 2λ, p +1/4

 2Z τ  Z τ
p ds ds
n o
(1) Ex exp − 2 ; ∈ dy, 0 < inf Ws , Wτ ∈ dz
2 0
Ws 0
Ws 0≤s≤τ

√  (x + z )√2 ch(yp=2)   √
 2xz 2 2xz 
= exp − I√4p2 +1 dydz
sh(y =2) sh(y =2) sh(y =2)
p p p

τ τ
ds ds
Z Z 
1.22.6 Px ∈ dg, ∈ dy, 0 < inf Ws , Wτ ∈ dz
0
Ws2 0
Ws 0≤s≤τ
0<x
√ √  √
 2xz g (x + z ) 2 ch(y =2)  2 2xz 
 p
= exp − − ig/8 dgdydz
8 sh(y =2) 8 sh(y =2) sh(y =2)
p p p

 2Z t  Z t
p ds ds
n o
1.22.7 Ex exp − 2 ; ∈ dy, 0 < inf Ws , Wt ∈ dz
2 0
Ws 0
Ws 0≤s≤t
(1)

√ p √
0<x = xz ist ( 4p2 + 1, y/2, 0, x + z, xz)dydz

t t
ds ds
Z Z 
1.22.8 Px ∈ dg, ∈ dy, 0 < inf Ws , Wt ∈ dz
0
Ws2 0
Ws 0≤s≤t
0<x

√ √
= 8−1 xze−g/8 eit (g/8, y/2, x + z, 2 xz)dgdydz
 1. EXPONENTIAL STOPPING 189

1 W τ ∼ Exp(λ), independent of W

n  Z τ


 o
1.23.1 Ex exp − p1I(−∞,r) Ws + q1I[r,∞) Ws ds ; sup Ws < b
0 0≤s≤τ
r<b

 (p2 + 2(q − p) ch((b − r)q ))e(x−r)p

x≤r = −
 + p ( + p)q (p sh((b − r)q ) + q ch((b − r)q ))


 2(q − p) + (p2 − p q )e(r−b)q sh((b − x)q )

r≤x = 1 − e(x−b)Υq +
+q ( + q)p (p sh((b − r)q ) + q ch((b − r)q ))
x≤b

n  Z τ


 o
1.23.5 Ex exp − p1I(−∞,r) Ws + q1I[r,∞) Ws ds ; sup Ws < b, Wτ ∈ dz
0 0≤s≤τ
r<b

z<b =: λQ(23)
λ (z)dz

 −|z−x|Υp (p sh((b − r)q ) − q ch((b − r)q ))e(z+x−2r)p

x≤r = e dz + dz
p p (p sh((b − r)q ) + q ch((b − r)q ))
z≤r

2 sh((b − z )q )e(x−r)p

x≤r = dz
p sh((b − r)q ) + q ch((b − r)q )
r≤z

2 sh((b − x)q )e(z−r)p

r≤x = dz
p sh((b − r)q ) + q ch((b − r)q )
z≤r

 −|z−x|Υq

2(q − p ) sh((b − z )q ) sh((b − x)q )e(r−b)q 
r≤x = e − e(x+z−2b)Υq + dz
q p sh((b − r)q ) + q ch((b − r)q )
r≤z

n  Z τ


 o
1.24.1 Ex exp − p1I(−∞,r) Ws + q1I[r,∞) Ws ds ; a < inf Ws
0 0≤s≤τ
a<r

  2(q − p) + (p q − q2 )e(a−r)p sh((x − a)p )

(a−x)Υp


a≤x = 1−e −
+p ( + p)q (q sh((r − a)p ) + p ch((r − a)p ))
x≤r

√
Υs := 2λ + 2s
190 1. BROWNIAN MOTION

1 W τ ∼ Exp(λ), independent of W

 (q2 − 2(q − p) ch((r − a)p ))e(r−x)q

r≤x = −
+q p ( + q)(q sh((r − a)p ) + p ch((r − a)p ))

n  Z τ


 o
1.24.5 Ex exp − p1I(−∞,r) Ws + q1I[r,∞) Ws ds ; a < inf Ws , Wτ ∈ dz
0 0≤s≤τ
a<r

a<z =: λQ(24)
a,λ (z)dz

 −|z−x|Υp

2(p − q ) sh((z − a)p ) sh((x − a)p )e(a−r)p 
x≤r = e − e(2a−z−x)Υp + dz
p q sh((r − a)p ) + p ch((r − a)p )
z≤r

2 sh((x − a)p )e(r−z)q

x≤r = dz
q sh((r − a)p ) + p ch((r − a)p )
r≤z

2 sh((z − a)p )e(r−x)q

r≤x = dz
q sh((r − a)p ) + p ch((r − a)p )
z≤r

 −|z−x|Υq (q sh((r − a)p ) − p ch((r − a)p ))e(2r−x−z)q

r≤x = e dz + dz
q q (q sh((r − a)p ) + p ch((r − a)p ))
r≤z

Ex e−γ`(τ,r) ; a < inf Ws , sup Ws < b


1.25.1
0≤s≤τ 0≤s≤τ
r<b

ch((r + a − 2x) =2)

p
a≤x =1−
ch((r − a) =2)
p
x≤r
√ √
2 sh((b − a) =2) sh((b − r) =2) sh((x − a) 2)
p p
+ √ √ √
( sh((b − r) 2) sh((r − a) 2) + =2 sh((b − a) 2)) ch((r − a) =2)
p p

ch((b + r − 2x) =2)

p
r≤x =1−
ch((b − r) =2)
p
x≤b
√ √
2 sh((b − a) =2) sh((r − a) =2) sh((b − x) 2)
p p
+ √ √ √
( sh((b − r) 2) sh((r − a) 2) + =2 sh((b − a) 2)) ch((b − r) =2)
p p

√
Υs := 2λ + 2s
 1. EXPONENTIAL STOPPING 191

1 W τ ∼ Exp(λ), independent of W

1.25.2 Px `(τ, r) ∈ dy, a < inf Ws , sup Ws < b
0≤s≤τ 0≤s≤τ
r<b

√
y =2 sh((b − a) 2)
 p 
= exp − √ √
sh((b − r) 2) sh((r − a) 2)

√
=2 sh((b − a) =2) sh((x − a) 2)
 p p
 ch((b − r) =2) ch((r − a) =2) sh((r − a)√2) dy, a ≤ x ≤ r

 p p
× √
=2 sh((b − a) =2) sh((b − x) 2)
p p
dy, r ≤ x ≤ b

 √
sh((b − r) 2) ch((b − r) =2) ch((r − a) =2)
 p p

(1) Px `(τ, r) = 0, sup Ws < b
0≤s≤τ

ch((r + a − 2x) =2)

 p
 1 − ch((r − a)p=2) , a≤x≤r


=
ch((b + r − 2x) =2)
p
 1− r≤x≤b

 ,
ch((b − r) =2)
p

Ex e−γ`(τ,r) ; a < inf Ws , sup Ws < b, Wτ ∈ dz =: λQ(25)


1.25.5 a,λ (z)dz
0≤s≤τ 0≤s≤τ
r<b
z<b
√ √ √
 ch((r − a − |z − x|) 2) − ch((r + a − x − z ) 2)


z≤r = √ √ dz
2 sh((r − a) 2)
x≤r
√ √ √
 sh((b − r) 2) sh((z − a) 2) sh((x − a) 2)
+ √ √ √ √ dz
( sh((b − r) 2) sh((r − a) 2) + =2 sh((b − a) 2)) sh((r − a) 2)
p

√ √
 sh((b − x) 2) sh((z − a) 2)
z≤r = √ √ √ dz
( sh((b − r) 2) sh((r − a) 2) + =2 sh((b − a) 2))
p
r≤x

√ √
 sh((b − z ) 2) sh((x − a) 2)
r≤z = √ √ √ dz
( sh((b − r) 2) sh((r − a) 2) + =2 sh((b − a) 2))
p
x≤r

√ √ √
 ch((b − r − |z − x|) 2) − ch((b + r − z − x) 2)


r≤z = √ √ dz
2 sh((b − r) 2)
r≤x
√ √ √
 sh((b − z ) 2) sh((r − a) 2) sh((b − x) 2)
+ √ √ √ √ dz
( sh((b − r) 2) sh((r − a) 2) + =2 sh((b − a) 2)) sh((b − r) 2)
p
192 1. BROWNIAN MOTION

1 W τ ∼ Exp(λ), independent of W

(25)
1.25.6 Px `(τ, r) ∈ dy, a < inf Ws , sup Ws < b, Wτ ∈ dz =: λBa,λ (y, z)dydz
0≤s≤τ 0≤s≤τ
r<b
z<b
√ √ √
 sh((z − a) 2) sh((x − a) 2) y =2 sh((b − a) 2)
 p 
z≤r = √ exp − √ √ dydz
sh2 ((r − a) 2) sh((b − r) 2) sh((r − a) 2)
x≤r
√ √ √
 sh((b − x) 2) sh((z − a) 2) y =2 sh((b − a) 2)
 p 
z≤r = √ √ exp − √ √ dydz
sh((b − r) 2) sh((r − a) 2) sh((b − r) 2) sh((r − a) 2)
r≤x
√ √ √
 sh((b − z ) 2) sh((x − a) 2) y =2 sh((b − a) 2)
 p 
r≤z = √ √ exp − √ √ dydz
sh((b − r) 2) sh((r − a) 2) sh((b − r) 2) sh((r − a) 2)
x≤r
√ √ √
 sh((b − z ) 2) sh((b − x) 2) y =2 sh((b − a) 2)
 p 
r≤z = 2 √ exp − √ √ dydz
sh ((b − r) 2) sh((b − r) 2) sh((r − a) 2)
r≤x

(1) Px `(τ, r) = 0, a < inf Ws , sup Ws < b, Wτ ∈ dz
0≤s≤τ 0≤s≤τ

√ √ √
 ch((r − a − |z − x|) 2) − ch((r + a − x − z ) 2)


z≤r = √ √ dz
2 sh((r − a) 2)
x≤r
√ √ √
 ch (b − r − |z − x|) 2 − ch (b + r − z − x) 2




r≤z = √ √
dz
2 sh (b − r) 2
r≤x

(25)
1.25.8 Px `(t, r) ∈ dy, a < inf Ws , sup Ws < b, Wt ∈ dz =: Fa,x (t, y, z)dydz
0≤s≤t 0≤s≤t

for a < r < b, a<z<b

z≤r = est (0, 0, b − r, 0, y/2) ∗ essst (z − a, x − a, r − a, y/2)dydz

x≤r
z≤r = esst (b − x, b − r, 0, y/2) ∗ esst (z − a, r − a, 0, y/2)dydz
r≤x
r≤z = esst (b − z, b − r, 0, y/2) ∗ esst (x − a, r − a, 0, y/2)dydz
x≤r
r≤z = essst (b − z, b − x, b − r, y/2) ∗ est (0, 0, r − a, 0, y/2)dydz
r≤x
 1. EXPONENTIAL STOPPING 193

1 W τ ∼ Exp(λ), independent of W

n  Z τ



1.26.1 Ex exp − p1I(−∞,r) Ws + q1I[r,∞) Ws ds ;
r<b 0
o
a<r a < inf Ws , sup Ws < b
0≤s≤τ 0≤s≤τ

 ch((r + a − 2x)p =2)

 
a≤x = 1−
+p ch((r − a)p =2)
x≤r
2(p th((b − r)q =2) + q th((r − a)p =2)) sh((x − a)p )
+
p q (q cth((b − r)q ) + p cth((r − a)p )) sh((r − a)p )



ch((b + r − 2x)q =2) 
r≤x = 1−
+q ch((b − r)q =2)
x≤b
2(p th((b − r)q =2) + q th((r − a)p =2)) sh((b − x)q )
+
p q (q cth((b − r)q ) + p cth((r − a)p )) sh((b − r)q )

n  Z τ



1.26.5 Ex exp − p1I(−∞,r) Ws + q1I[r,∞) Ws ds ; a < inf Ws ,
0 0≤s≤τ
r<b
o
a<r sup Ws < b, Wτ ∈ dz =: λQ(26)
a,λ (z)dz
0≤s≤τ

 −|z−x|Υp
− e−(z+x−2a)Υp dz


z≤r = e
p
x≤r
2(p th((b − r)q ) − q ) sh((z − a)p ) sh((x − a)p )e(a−r)p
+ dz
p (p th((b − r)q ) ch((r − a)p ) + q sh((r − a)p ))

2 sh((z − a)p ) sh((b − x)q )

z≤r = dz
p sh((b − r)q ) ch((r − a)p ) + q ch((b − r)q ) sh((r − a)p )
r≤x

2 sh((b − z )q ) sh((x − a)p )

r≤z = dz
p sh((b − r)q ) ch((r − a)p ) + q ch((b − r)q ) sh((r − a)p )
x≤r

 −|z−x|Υq
− e(x+z−2b)Υq dz


r≤z = e
q
r≤x
2(q th((r − a)p ) − p ) sh((b − z )q ) sh((b − x)q )e(r−b)q
+ dz
q (p sh((b − r)q ) + q ch((b − r)q ) th((r − a)p ))

√
Υs := 2λ + 2s
194 1. BROWNIAN MOTION

1 W τ ∼ Exp(λ), independent of W

n  Z τ


 1 o
1.27.1 Ex exp − p1I(−∞,r) Ws + q1I[r,∞) Ws ds ; `(τ, r) ∈ dy
0
2

2 2 √  √ √
x≤r = √ +√ e(x−r) 2λ+2p e−y( 2λ+2p+ 2λ+2q) dy
2 + 2p 2 + 2 q

2 2  (r−x)√2λ+2q −y(√2λ+2p+√2λ+2q)
r≤x = √ +√ e e dy
2 + 2p 2 + 2 q

n  Z τ


 o
(1) Ex exp − p1I(−∞,r) Ws + q1I[r,∞) Ws ds ; `(τ, r) = 0
0
 √
1 − e(x−r) 2λ+2p


x≤r =
+p
 √
1 − e(r−x) 2λ+2q


r≤x =
+q

Z τ τ
1
Z 



1.27.2 Px 1I(−∞,r) Ws ds ∈ du, 1I[r,∞) Ws ds ∈ dv, `(τ, r) ∈ dy dy
0 0
2

 y + r − x y −λ(u+v)−(y+r−x)2 /2u−y2 /2v

 
x≤r = √ + e dudv
 uv u v

e−(u+v) d −(y+r−x)2 /2u−y2 /2v

or =− √ e dudv
 uv dy

 y y + x − r −λ(u+v)−y2 /2u−(y+x−r)2 /2v

 
r≤x = √ + e dudv
 uv u v

e−(u+v) d −y2 /2u−(y+x−r)2 /2v

or =− √ e dudv
 uv dy

Z τ

Z τ

(1) Px 1I(−∞,r) Ws ds ∈ du, 1I[r,∞) Ws ds = 0, `(τ, r) = 0
x≤r 0 0

r−x
  
= λe−λu 1 − Erfc √ du
2u

Z τ

Z τ

(2) Px 1I(−∞,r) Ws ds = 0, 1I[r,∞) Ws ds ∈ dv, `(τ, r) = 0
x≤r 0 0

x−r
  
= λe−λv 1 − Erfc √ dv
2v
 1. EXPONENTIAL STOPPING 195

1 W τ ∼ Exp(λ), independent of W

Z t


1 
1.27.4 Px p1I(−∞,r) Ws + q1I[r,∞) Ws ds ∈ dv, `(t, r) ∈ dy dy
0
2

for (p ∧ q)t ≤ v ≤ (p ∨ q)t

 2
|p − q |((y + r − x)|pt − v | + y |v − qt|) y |p − q| (y + r − x)2 |p − q|

x≤r = exp − − dv
|pt − v|3=2 |v − qt|3=2 2|pt − v| 2|v − qt|

 2
1 d y |p − q| (y + r − x)2 |p − q|

or =− exp − − dv
|pt − v|1=2 |v − qt|1=2 dy 2|pt − v| 2|v − qt|

|p − q |(y |pt − v | + (y + x − r)|v − qt|)


(y + x − r)2 |p − q| y2 |p − q| 
r≤x = exp − − dv
|pt − v|3=2 |v − qt|3=2 2|pt − v| 2|v − qt|

1 d

(y + x − r)2 |p − q| y2 |p − q| 
or =− exp − − dv
|pt − v|1=2 |v − qt|1=2 dy 2|pt − v| 2|v − qt|

t
r−x
Z


  
(1) Px p1I(−∞,r) Ws + q1I[r,∞) Ws ds=pt, `(t, r) = 0 =1 − Erfc √
0 2t
x≤r

t
x−r
Z


  
(2) Px p1I(−∞,r) Ws + q1I[r,∞) Ws ds=qt, `(t, r) = 0 =1 − Erfc √
0 2t
r≤x
τ

 1
n  Z o


1.27.5 Ex exp − p1I(−∞,r) Ws + q1I[r,∞) Ws ds ; `(τ, r) ∈ dy, Wτ ∈ dz
0
2

√ √
x≤r = 2λe−y 2λ+2q −(y+2r−x−z) 2λ+2p
e dydz
z≤r
√ √
x≤r = 2λe−(y+z−r) 2λ+2q −(y+r−x) 2λ+2p
e dydz
r≤z
√ √
r≤x = 2λe−(y+x−r) 2λ+2q −(y+r−z) 2λ+2p
e dydz
z≤r
√ √
r≤x = 2λe−(y+x+z−2r) 2λ+2q −y 2λ+2p
e dydz
r≤z

n  Z τ


 o
(1) Ex exp − p1I(−∞,r) Ws + q1I[r,∞) Ws ds ; `(τ, r) = 0, Wτ ∈ dz
0
196 1. BROWNIAN MOTION

1 W τ ∼ Exp(λ), independent of W

 √ √
 
x≤r = √ e−|z−x| 2λ+2p − e(z+x−2r) 2λ+2p dz
2 + 2p
z≤r
 √ √
 
r≤x = √ e−|z−x| 2λ+2q − e(2r−x−z) 2λ+2q dz
2 + 2q
r≤z

Z τ τ
1
Z 



1.27.6 Px 1I(−∞,r) Ws ds ∈ du, 1I[r,∞) Ws ds ∈ dv, `(τ, r) ∈ dy, Wτ ∈ dz
0 0
2

=: λe−λ(u+v) Bx(27) (u, v, y, z)dudvdydz

(y + 2r − x − z )y 
(y + 2r − x − z )2 y2 
x≤r = λe−λ(u+v) exp − − dudvdydz
u3=2 v3=2 2u 2v
z≤r

(y + r − x)(y + z − r) 
(y + r − x)2 (y + z − r)2 
x≤r = λe−λ(u+v) exp − − dudvdydz
u3=2 v3=2 2u 2v
r≤z

(y + r − z )(y + x − r) 
(y + r − z )2 (y + x − r)2 
r≤x = λe−λ(u+v) exp − − dudvdydz
u3=2 v3=2 2u 2v
z≤r

 2
y(y + x + z − 2r) y (y + x + z − 2r)2 
r≤x = λe−λ(u+v) 3= 2 3= 2 exp − − dudvdydz
u v 2u 2v
r≤z

Z τ

Z τ

(1) Px 1I(−∞,r) Ws ds ∈ du, 1I[r,∞) Ws ds = 0, `(τ, r) = 0, Wτ ∈ dz
0 0

1 −(z−x)2 /2u 1 −(2r−x−z)2 /2u 
x≤r = λe−λu √ e −√ e dudz
2u 2u
z≤r
Z τ

Z τ

(2) Px 1I(−∞,r) Ws ds = 0, 1I[r,∞) Ws ds ∈ dv, `(τ, r) = 0, Wτ ∈ dz
0 0

1 −(z−x)2 /2v 1 −(x+z−2r)2 /2v 
r≤x = λe−λv √ e −√ e dvdz
2v 2v
r≤z

Z t


1 
1.27.8 Px p1I(−∞,r) Ws + q1I[r,∞) Ws ds ∈ dv, `(t, r) ∈ dy, Wt ∈ dz
0
2
1 
|qt − v | |pt − v |

= Bx(27) , , y, z 1I((q∧p)t,(q∨p)t) (v) dvdydz
|p − q | |p − q | |p − q |
 1. EXPONENTIAL STOPPING 197

1 W τ ∼ Exp(λ), independent of W

for Bx(27) u, v, y, z see 1.27.6

Z t



(1) Px p1I(−∞,r) Ws + q1I[r,∞) Ws ds = pt, `(t, r) = 0, Wt ∈ dz
x≤r 0

1 −(z−x)2 /2t 1 2
z≤r = √ e dz − √ e−(2r−x−z) /2t dz
2t 2t

Z t



(2) Px p1I(−∞,r) Ws + q1I[r,∞) Ws ds = qt, `(t, r) = 0, Wt ∈ dz
r≤x 0

1 −(z−x)2 /2t 1 2
r≤z = √ e dz − √ e−(x+z−2r) /2t dz
2t 2t

Ex e−γ Ĥ(τ )−ηȞ(τ ) ; inf Ws ∈ da, sup Ws ∈ db


1.28.1
0≤s≤τ 0≤s≤τ
a<x
x<b
√
2(2 + 2) sh((b − x) 2 + 2 + 2)
n p
= √ √
sh((b − a) 2 + 2 + 2) sh((b − a) 2 + 2)

√
2(2 + 2 ) sh((x − a) 2 + 2 + 2) o sh((b − a) =2)
p p
+ √ √ dadb
sh((b − a) 2 + 2 + 2) sh((b − a) 2 + 2 ) ch((b − a) =2)
p

1.28.2 Px Ĥ(τ ) ∈ du, Ȟ(τ ) ∈ dv, inf Ws ∈ da, sup Ws ∈ db
0≤s≤τ 0≤s≤τ

√ p
u<v = 2λe−λv th((b − a) λ/2) ssu (b − x, b − a) sv−u (b − a) dudvdadb
√ p
v<u = 2λe−λu th((b − a) λ/2) ssv (x − a, b − a) su−v (b − a) dudvdadb

1.28.4 Px Ĥ(t) ∈ du, Ȟ(t) ∈ dv, inf Ws ∈ da, sup Ws ∈ db
0≤s≤t 0≤s≤t

for 0 < u ∧ v < u ∨ v < t, a<x<b

u<v = 2 ssu (b − x, b − a) sv−u (b − a) sct−v ((b − a)/2, (b − a)/2)dudvdadb

v<u = 2 ssv (x − a, b − a) su−v (b − a) sct−u ((b − a)/2, (b − a)/2)dudvdadb

198 1. BROWNIAN MOTION

1 W τ ∼ Exp(λ), independent of W

Ex e−γ Ĥ(τ )−ηȞ(τ ) ; inf Ws ∈ da, sup Ws ∈ db, Wτ ∈ dz


1.28.5
0≤s≤τ 0≤s≤τ
a<x
x<b
√ √ √
2 2 + 2 sh((b − x) 2 + 2 + 2) sh((z − a) 2)
= √ √ √ dadbdz
sh((b − a) 2 + 2 + 2) sh((b − a) 2 + 2) sh((b − a) 2)
√ √ √
2 2 + 2 sh((x − a) 2 + 2 + 2) sh((b − z ) 2)
+ √ √ √ dadbdz
sh((b − a) 2 + 2 + 2) sh((b − a) 2 + 2 ) sh((b − a) 2)

1.28.6 Px Ĥ(τ ) ∈ du, Ȟ(τ ) ∈ dv, inf Ws ∈ da, sup Ws ∈ db, Wτ ∈ dz
0≤s≤τ 0≤s≤τ

√
sh((z − a) 2)
u<v = 2λe−λv √ ss (b − x, b − a) sv−u (b − a) dudvdadbdz
sh((b − a) 2) u

√
sh((b − z ) 2)
v<u = 2λe−λu √ ss (x − a, b − a) su−v (b − a) dudvdadbdz
sh((b − a) 2) v

1.28.8 Px Ĥ(t) ∈ du, Ȟ(t) ∈ dv, inf Ws ∈ da, sup Ws ∈ db, Wt ∈ dz
0≤s≤t 0≤s≤t

for 0 < u ∧ v < u ∨ v < t, a<x∧z <x∨z <b

u<v = 2 ssu (b − x, b − a) sv−u (b − a) sst−v (z − a, b − a)dudvdadbdz

v<u = 2 ssv (x − a, b − a) su−v (b − a) sst−u (b − z, b − a)dudvdadbdz

n  Z τ



1.29.1 Ex exp − p1I(−∞,r) Ws + q1I[r,∞) Ws ds ; a < inf Ws ,
0 0≤s≤τ

1 o
sup Ws < b, `(τ, r) ∈ dy
0≤s≤τ 2

yq ch((b − r)q ) yp ch((r − a)p )

 
= 2λ exp − −
sh((b − r)q ) sh((r − a)p )


sh((x − a)p )
 sh((r − a) ) dy, a ≤ x ≤ r

sh((b − r)q =2) sh((r − a)p =2) 
p
× +
q ch((b − r)q =2) p ch((r − a)p =2)  sh((b − x )q )
dy, r ≤ x ≤ b
sh((b − r)q )

 1. EXPONENTIAL STOPPING 199

1 W τ ∼ Exp(λ), independent of W

n  Z τ



(1) Ex exp − p1I(−∞,r) Ws + q1I[r,∞) Ws ds ; a < inf Ws ,
0 0≤s≤τ
o
sup Ws < b, `(τ, r) = 0
0≤s≤τ



ch((2x − r − a)p =2) 
  + p 1 − ch((r − a) =2)
 , a≤x≤r
p
=
  1 − ch((b + r − 2x)q =2) ,
 
r≤x≤b

+q ch((b − r)q =2)

n  Z τ



1.29.5 Ex exp − p1I(−∞,r) Ws + q1I[r,∞) Ws ds ; a < inf Ws ,
0 0≤s≤τ

1 o
sup Ws < b, `(τ, r) ∈ dy, Wτ ∈ dz =: λQ(29)
λ (y, z)dydz
0≤s≤τ 2

yq ch((b − r)q ) yp ch((r − a)p )

 
= 2λ exp − −
sh((b − r)q ) sh((r − a)p )

sh((x − a)p ) sh((z − a)p )


dydz, a ≤ x ≤ r, a ≤ z ≤ r
sh2 ((r − a)p )




sh((b − x)q ) sh((z − a)p )



dydz, a≤z≤r≤x≤b

sh((b − r)q ) sh((r − a)p )


×
sh((b − z )q ) sh((x − a)p )
dydz, a≤x≤r≤z≤b

sh((b − r)q ) sh((r − a)p )




sh((b − x)q ) sh((b − z )q )



dydz, r ≤ x ≤ b, r ≤ z ≤ b
sh2 ((b − r)q )



n  Z τ



(1) Ex exp − p1I(−∞,r) Ws + q1I[r,∞) Ws ds ; a < inf Ws ,
0 0≤s≤τ
o
sup Ws < b, `(τ, r) = 0, Wτ ∈ dz
0≤s≤τ

(ch((r − a − |z − x|)p ) − ch((r + a − z − x)p ))

x≤r = dz
p sh((r − a)p )
z≤r
(ch((b − r − |z − x|)q ) − ch((b + r − z − x)q ))
r≤x = dz
q sh((b − r)q )
r≤z

√
Υs := 2λ + 2s
200 1. BROWNIAN MOTION

1 W τ ∼ Exp(λ), independent of W

Z τ

Z τ

1.29.6 Px 1I(−∞,r) Ws ds ∈ du, 1I[r,∞) Ws ds ∈ dv, a < inf Ws ,
0 0 0≤s≤τ

1 
sup Ws < b, `(τ, r) ∈ dy, Wτ ∈ dz
0≤s≤τ 2

=: λe−λ(u+v) Bx(29) (u, v, y, z)dudvdydz

z≤r = λe−λ(u+v) 2 esssu (x − a, z − a, r − a, y) esv (0, 0, b − r, 0, y)dudvdydz

x≤r

z≤r = λe−λ(u+v) 2 essu (z − a, r − a, 0, y) essv (b − x, b − r, 0, y)dudvdydz

r≤x

r≤z = λe−λ(u+v) 2 essu (x − a, r − a, 0, y) essv (b − z, b − r, 0, y)dudvdydz

x≤r

r≤z = λe−λ(u+v) 2 esu (0, 0, r − a, 0, y) esssv (b − z, b − x, b − r, y)dudvdydz

r≤x

Z τ

Z τ

(1) Px 1I(−∞,r) Ws ds ∈ du, 1I[r,∞) Ws ds = 0, a < inf Ws ,
0 0 0≤s≤τ

sup Ws < b, `(τ, r) = 0, Wτ ∈ dz
0≤s≤τ

for a < x ∧ z ≤ x ∨ z < r

= λe−λu (csu (r − a − |z − x|, r − a) − csu (r + a − z − x, r − a))dudz

Z τ

Z τ

(2) Px 1I(−∞,r) Ws ds = 0, 1I[r,∞) Ws ds ∈ dv, a < inf Ws ,
0 0 0≤s≤τ

sup Ws < b, `(τ, r) = 0, Wτ ∈ dz
0≤s≤τ

for r < x ∧ z ≤ x ∨ z < b

= λe−λv (csv (b − r − |z − x|, b − r) − csv (b + r − z − x, b − r))dvdz

 1. EXPONENTIAL STOPPING 201

1 W τ ∼ Exp(λ), independent of W

Z t



1.29.8 Px p1I(−∞,r) Ws + q1I[r,∞) Ws ds ∈ dv, a < inf Ws , sup Ws < b,
0 0≤s≤t 0≤s≤t

1 
`(t, r) ∈ dy, Wt ∈ dz
2

1 
|qt − v | |pt − v |

= Bx(29) , , y, z 1I((q∧p)t,(q∨p)t) (v) dvdydz
|p − q | |p − q | |p − q |

for Bx(29) u, v, y, z see 1.29.6

Z t

Z t

(1) Px 1I(−∞,r) Ws ds = t, 1I[r,∞) Ws ds = 0, a < inf Ws ,
0 0 0≤s≤t

sup Ws < b, `(t, r) = 0, Wt ∈ dz
0≤s≤t

for a < x ∧ z ≤ x ∨ z < r

= cst (r − a − |z − x|, r − a)dz − cst (r + a − z − x, r − a)dz

Z t

Z t

(2) Px 1I(−∞,r) Ws ds = 0, 1I[r,∞) Ws ds = t, a < inf Ws ,
0 0 0≤s≤t

sup Ws < b, `(t, r) = 0, Wt ∈ dz
0≤s≤t

for r < x ∧ z ≤ x ∨ z < b

= cst (b − r − |z − x|, b − r)dz − cst (b + r − z − x, b − r)dz

 2Z τ √
p
 Z τ  p ch(yp| |=2) ( 2=| |)
n o p
1.30.1 Ex exp − e2βWs ds ; eβWs ds ∈ dy = √
2 | | sh(yp| |=2) (1 + 8=| |)
p
(1) 0 0

x pe x ch(yp| |) 2pe x 
  
β 6= 0 × exp − − M1/2,√2λ/|β| dy
2 | | sh(yp| |) | | sh(yp| |)
202 1. BROWNIAN MOTION

1 W τ ∼ Exp(λ), independent of W

 2Z t  Z t
p
n o
1.30.3 Ex exp − e2βWs ds ; eβWs ds ∈ dy
2 0 0
(1)

2 pp ch(yp| |=2) 
x pe x ch(yp| |)
 
1 pe x

β 6= 0 = exp − − mβ 2 t/2 − , dy
2 | | sh(yp| |=2) 2 | | sh(yp| |) 2 | | sh(yp| |)
p

 Z τ 2
p 2βWs
n   o
1.30.5 Ex exp − e + qeβWs ds ; Wτ ∈ dz =: λQ(30)
λ (z)dz
0
2
β 6= 0
√
 ( 2=| | + 1=2 + q=p| |)
= e−β(x+z)/2 √
p (2 2=| | + 1)

2p βx 2p βz



M−q/p|β|,√2λ/|β| W−q/p|β|,√2λ/|β| |β|
(
|β| e e dz, βx ≤ βz
× 2p

2p


W−q/p|β|,√2λ/|β| |β| eβx M−q/p|β|,√2λ/|β| |β| eβz dz, βz ≤ βx

 2Z τ τ
p
n  Z o
(1) Ex exp − e2βWs ds ; eβWs ds ∈ dy, Wτ ∈ dz =: λBλ(30) (y, z)dydz
2 0 0

p

p(e x + e z ) ch(yp| |=2)
 
2pe (x+z)=2 
= exp − I2√2λ/|β| dydz
sh(yp| |=2) | | sh(yp| |=2) | | sh(yp| |=2)

Z τ Z τ 
2βWs
1.30.6 Px e ds ∈ dg, eβWs ds ∈ dy, Wτ ∈ dz
0 0

√

2 2 y| | 1 βx
1 
β 6= 0 = λ isg , , 0, e + eβz , eβ(x+z)/2 dgdydz
| | 2 | | | |

 2Z t  Z t
p
n o
1.30.7 Ex exp − e2βWs ds ; eβWs ds ∈ dy, Wt ∈ dz =: Fx,t
(30)
(y, z)dydz
2 0 0
(1)

2p 
p(e x + e z ) ch(yp| |=2)
 
2pe (x+z)=2 
β 6= 0 = exp − iβ 2 t/8 dydz
8 sh(yp| |=2) | | sh(yp| |=2) | | sh(yp| |=2)
 1. EXPONENTIAL STOPPING 203

1 W τ ∼ Exp(λ), independent of W

Z t Z t 
1.30.8 Px e2βWs ds ∈ dg, eβWs ds ∈ dy, Wt ∈ dz
0 0

2  2
t y| | 1 βx
2 
β 6= 0 = eig , , e + eβz , eβ(x+z)/2 dgdydz
8 8 2 | | | |

√ √ √


1.31.1 Ex/√2λ exp −γ 2λ sup `(s, u/ 2λ) − α−1 `(s, r/ 2λ)
0≤s≤τ
0<α

2 − 2e−2|u−r| −  e|u−r|−|x−r| +  e−|u−x|

=1−
2 1 − e−2|u−r| + 

 √ √
√  h
1.31.2 Px/√2λ sup `(s, u/ 2λ) − α−1 `(s, r/ 2λ) > h/ 2λ = e|u−r|−|x−r|
0≤s≤τ
0<α


i h
 
−
e|u−r|−|x−r| − e−|u−x| exp −
2 1 − e−2|u−r| 2 1 − e−2|u−r|

n √ √ √

√ o
1.31.5 Ex/√2λ exp −γ 2λ sup `(s, u/ 2λ) − α−1 `(s, r/ 2λ) ; Wτ ∈ dz/ 2λ
0≤s≤τ
0<α

1 e−|z−u| 2 − 2e−2|u−r| −  e|u−r|−|x−r| +  e−|u−x|



= e−|z−x| dz − dz
2 2 2 1 − e−2|u−r| + 

 √ √
√ √ 
1.31.6 Px/√2λ sup `(s, u/ 2λ) − α−1 `(s, r/ 2λ) > h/ 2λ, Wτ ∈ dz/ 2λ
0≤s≤τ
0<α

1 h |u−r|−|x−r| 

e|u−r|−|x−r| − e−|u−x|

i
= e −
2 2 1 − e−2|u−r|

h
 
× exp − |z − u| −
dz
2 1 − e−2|u−r|

q

Υ := (α − 1)2 + 4α 1 − e−2|u−r| + 1 − α
204 1. BROWNIAN MOTION

1 W Hz = min{s : Ws = z}

2. Stopping at first hitting time

√
2.0.1 Ex e−αHz = e−|x−z| 2α

|x − z |

(x − z )2 
2.0.2 Px Hz ∈ dt = √ 3=2 exp − dt
2t 2t

∞
1
Z
sup Ws = e−γx − γ(x − z)e−γz e−γy dy


2.1.1 Ex exp −γ
0≤s≤Hz x−z
y
z≤x

x≤z
(
1,
y−x


2.1.2 Px sup Ws < y =
, z≤x≤y
0≤s≤Hz
y−z

 √
 e−(z−x) 2α , x≤z≤y
√
e−αHz ; sup Ws < y =

 sh((y − x)√2 ) ,
2.1.4 Ex
0≤s≤Hz z≤x≤y
sh((y − z ) 2 )

( |x − z |

(x − z )2 
√ exp − dt, x ≤ z ≤ y
2t3=2 2t


(1) Px sup Ws < y, Hz ∈ dt =
0≤s≤Hz sst (y − x, y − z)dt, z≤x≤y

∞
1
Z
γx γz
e−γy dy


2.2.1 Ex exp γ inf Ws = e − γ(z − x)e
0≤s≤Hz z−x
y
x≤z

( x−y
, y≤x≤z
z−y


2.2.2 Px inf Ws > y =
0≤s≤Hz 1, y≤z≤x

√
sh((x − y) 2 )

√ , y≤x≤z
sh((z − y) 2 )

Ex e−αHz ; inf Ws > y =

2.2.4
0≤s≤Hz  −(x−z)√2α
e , y≤z≤x

sst (x − y, z − y)dt, y≤x≤z

(
(x − z )2 


(1) Px inf Ws > y, Hz ∈ dt = |x − z |

0≤s≤Hz √ exp − dt, y ≤ z ≤ x
2t3=2 2t
 2. STOPPING AT FIRST HITTING TIME 205

1 W Hz = min{s : Ws = z}


 1, x ≤ z ≤ r, r≤z≤x
 1 + 2 |r − x|


Ex e−γ`(Hz ,r) , r∧z ≤x≤r∨z
2.3.1 = 1 + 2 |z − r |
1


, z ≤ r ≤ x, x≤r≤z

1 + 2 |z − r |


x ≤ z ≤ r, r ≤ z ≤ x

 0,
 |x − z | y

  
exp − dy, z∧r ≤x≤r∨z
2(r − z )2 2|r − z |



2.3.2 Px ` Hz , r ∈ dy =
1 y

  
exp − dy, z ≤ r ≤ x, x ≤ r ≤ z

2|r − z | 2|r − z |


 1, x ≤ z ≤ r, r≤z≤x



 |r − x|
(1) Px ` H z , r = 0 = , z∧r ≤x≤r∨z
 |r − z |

0, z ≤ r ≤ x, x≤r≤z

2.3.3 Ex e−αHz −γ`(Hz ,r)

1, x ≤ z ≤ r, r≤z≤x

√ √
−2|x−r | 2

 2 + 1−e



√

√ √
−2|z −r | 2

, r∧z ≤x≤r∨z
= e−|z−x| 2α
2 + 1−e
√
2



√ √
, z ≤ r ≤ x, x≤r≤z

2 + 1 − e−2|z−r| 2


√
y 2
 
Ex e−αHz ; ` Hz , r ∈ dy = exp −


2.3.4 √
−2|r −z | 2
1−e

0, x ≤ z ≤ r, r≤z≤x
√ √

 2 sh(|x − z | 2 )



√ dy, z∧r ≤x≤r∨z
× 2 sh2 (|r − z | 2 )
√
2 √


e−|r−x| 2α dy,

 √ z ≤ r ≤ x, x≤r≤z
2 sh(|r − z | 2 )


 −|z−x|√2α
 e , x ≤ z ≤ r, r≤z≤x
√
sh(|x − r| 2 )


 −αHz

(1) Ex e ; ` Hz , r = 0 = √ , r∧z ≤x≤r∨z



sh(|z − r| 2 )
0, z ≤ r ≤ x, x≤r≤z
206 1. BROWNIAN MOTION

1 W Hz = min{s : Ws = z}


1

(2) Px Hz ∈ dt, ` Hz , r ∈ dy
2
for z ∧ r ≤ x ≤ r ∨ z
1
= (est (1, 2, |r − z|, y − |x − z|, y) − est (1, 2, |r − z|, y + |x − z|, y))dtdy
2
for z ≤ r ≤ x, x≤r≤z

= est (1, 1, |r − z|, y + |x − r|, y)dtdy

 Z Hz

2.4.1 Ex exp −γ 1I[r,∞) Ws ds
0

x≤z

1,
 1 + (r − x)√2



√ , z≤x≤r
z≤r = 1 + (r − z ) 2
1 √


√ e−(x−r) 2γ , r≤x

1 + (r − z ) 2


 1
√ , x≤r

 ch(( z − r) 2 )
√

ch((x − r) 2 )

r≤z = √ , r≤x≤z

 ch((z − r) 2 )
 −(x−z)√2γ

e , z≤x

Z Hz

2.4.2 Px 1I[r,∞) Ws ds ∈ dy
0

x≤z

0,
  √y 
x−z 1 1

 
y/2(r−z)2
− z≤x≤r

√ e Erfc √ dy,
 (r − z )2 2y 2(r − z )

(r − z ) 2


z<r = 1 2
√ e−(x−r) /2y dy−
(r − z ) 2y



√
1 x−r y x−r y 

   

−

exp + Erfc √ + √ dy, r≤x
2(r − z )2 r−z 2(r − z )2 2y (r − z ) 2


 ccy (0, z − r)dy, x≤r

r<z = ccy (x − r, z − r)dy, r≤x≤z
 √x − z e−(x−z)2 /2y dy,

 z≤x
2y3=2
 2. STOPPING AT FIRST HITTING TIME 207

1 W Hz = min{s : Ws = z}

x≤z

1,
Hz
r−x
Z  



(1) Px 1I[r,∞) Ws ds = 0 = , z≤x≤r
z≤r 0  r−z

0, r≤x

Z Hz
 Z ∞
2.4.4 Px 1I[r,∞) Ws ds ∈ dv, Hz ∈ dt = Bz(27) (t − v, v, y)dy dvdt
0 0

(27)
for Bz (u, v, y) see 2.27.2

Hz
ht (1, z − x)dt, x≤z
Z  


(1) Px 1I[r,∞) Ws ds = 0, Hz ∈ dt =
0 sst (r − x, r − z)dt, z ≤ x ≤ r

 Z Hz

2.5.1 Ex exp −γ 1I(−∞,r] Ws ds
0

 −(z−x)√2γ
e , x≤z
 ch((r − x)√2 )



√ , z≤x≤r
z≤r = ch((r − z ) 2 )
1


r≤x

√ ,
ch((r − z ) 2 )


 1 √
√ e−(r−x) 2γ , x≤r
 1 + (z − r) √2



r≤z = 1 + (x − r ) 2
√ , r≤x≤z
 1 + (z − r ) 2


z≤x

1,

Z Hz

2.5.2 Px 1I(−∞,r] Ws ds ∈ dy
0

 z − x −(z−x)2 /2y
√ e dy, x ≤ z

 2y3=2
z<r = ccy (r − x, r − z)dy, z≤x≤r


ccy (0, r − z)dy, r≤x
 1 −(r−x) 2
/2y
√ e dy−
 (z − r) 2y


√
1 y 

r−x y r−x

   
− exp + Erfc + √ dy, x≤r

√
2(z − r)2 z−r 2(z − r)2 2y (z − r) 2

r<z = √
 z−x
 
1 1 2
 y 
√ − ey/2(z−r) Erfc √ dy, r≤x≤z
(z − r)2 2y 2(z − r) (z − r) 2





z≤x

0,
208 1. BROWNIAN MOTION

1 W Hz = min{s : Ws = z}

x≤r

0,
Hz
x−r
Z  



(1) Px 1I(−∞,r] Ws ds = 0 = , r≤x≤z
r≤z 0  z−r

1, z≤x

Z Hz
 Z ∞
2.5.4 Px 1I(−∞,r) Ws ds ∈ dv, Hz ∈ dt = Bz(27) (v, t − v, y)dy dvdt
0 0

for Bz(27) (u, v, y) see 2.27.2

Hz
sst (x − r, z − r)dt, r ≤ x ≤ z
Z  


(1) Px 1I(−∞,r) Ws ds = 0, Hz ∈ dt =
0 ht (1, x − z)dt, z≤x

 Z Hz



2.6.1 Ex exp − p1I(−∞,r) Ws + q1I[r,∞) Ws ds =: Q(6)
z (2p, 2q)
0
 −(z−x)√2p
 e , x≤z
√ √ √ √
 p ch((r − x) 2p) + q sh((r − x) 2p)



√ √ √ √ , z≤x≤r
z≤r = p ch((r − z ) 2p) + q sh((r − z ) 2p)
√ −(x−r) 2q √
pe



√ √ , r≤x

 √ √
p ch((r − z ) 2p) + q sh((r − z ) 2p)
√ −(r−x)√2p
qe

√ √ √ , x≤r
q ch(( z r ) 2q) + √p sh((z − r) 2q)

−



 √ √ √ √
r≤z = q ch((x − r) 2q) + p sh((x − r) 2q)
√ √ √ √ , r≤x≤z


 q ch((z − r) 2q) + p sh((z − r) 2q)
 √
e−(x−z) 2q ,

z≤x

Z Hz

Z Hz

2.6.2 Px 1I(−∞,r) Ws ds ∈ du, 1I[r,∞) Ws ds ∈ dv
0 0

Z ∞
= Bz(27) (u, v, y)dy dudv for Bz(27) (u, v, y) see 2.27.2
0

Z Hz

Z Hz

(1) Px 1I(−∞,r) Ws ds ∈ du, 1I[r,∞) Ws ds = 0
0 0

hu (1, z − x)du, x≤z


=
ssu (r − x, r − z)du, z≤x≤r
 2. STOPPING AT FIRST HITTING TIME 209

1 W Hz = min{s : Ws = z}

Z Hz

Z Hz

(2) Px 1I(−∞,r) Ws ds = 0, 1I[r,∞) Ws ds ∈ dv
0 0

ssv (x − r, z − r)dv, r≤x≤z


=
hv (1, x − z)dv, z≤x

 Z Hz

2.7.1 Ex exp −γ 1I[r,u] Ws ds
r<u 0

x≤z

1,
√ √ √

(r − x) 2 sh((u − r) 2 ) + ch((u − r) 2 )



 √ √ √
 (r − z ) 2 sh((u − r) 2 ) + ch((u − r) 2 , z≤x≤r
)


√
z≤r = ch((u − x) 2 )
√ √ √ , r≤x≤u
(r − z ) 2 sh((u − r) 2 ) + ch((u − r) 2 )



1



, u≤x

√ √ √
(r − z ) 2 sh((u − r) 2 ) + ch((u − r) 2 )


 1
√ √ √ , x≤r
(z − u) 2 sh((u − r) 2 ) + ch((u − r) 2 )



 √
ch((x − r) 2 )


r≤x≤u

√ √ √ ,
(z − u) 2 sh((u − r) 2 ) + ch((u − r) 2 )

u≤z = √ √ √
 (x − u) 2 sh((u − r) 2 ) + ch((u − r) 2 )
u≤x≤z

√ √ √ ,
(z − u) 2 sh((u − r) 2 ) + ch((u − r) 2 )





z≤x

1,

Z Hz

2.7.2 Px 1I[r,u] Ws ds ∈ dy =: Bz(7) (y)dy
r<u 0


(x − z )
 (r − z ) rsy (0, u − r, u − r, r − z)dy, z ≤ x ≤ r


= rs (0, u − r, u − x, r − z))dy, r≤x≤u
 y

rsy (0, u − r, 0, r − z)dy, u≤x




 rsy (0, u − r, 0, z − u)dy, x≤r

= rs y (0, u − r, x − r, z − u)dy, r≤x≤u
 (z − x) rsy (0, u − r, u − r, z − u)dy, u ≤ x ≤ z


(z − u)
210 1. BROWNIAN MOTION

1 W Hz = min{s : Ws = z}

 Z H0 
|Ws |ds = 32/3 Γ(2/3) Ai (2γ)1/3 |x|


2.8.1 Ex exp −γ
(1) 0

Z H0  
2 1/3 |x| −2|x|3 /9y
2.8.2 Px |Ws |ds ∈ dy = e dy
0
9y4 (1=3)
(1)

√
 D−1=2− = (− x 2 )

√ , x≤z
2 = (−z 2 )
Hz   D
−1=2−
 Z
2.9.3 Ex exp −αHz − Ws2 ds = √
2 D = (x 2 )
0  −1=2− z≤x

√ ,
= (z 2 )

D−1=2−

 2Z
|x| 3=2
 2
x ch(t )
n H0  o 
2.9.4 Ex exp − Ws2 ds ; H0 ∈ dt = √ exp − dt
2 0 2 sh3=2 (t ) 2 sh(t )
(1)

3 3 x2

H0
| x|
Z 
(2) Px Ws2 ds ∈ dy, H0 ∈ dt = √ esy , , t, 0, dydt
0 2 2 2 2

√
I 2 =| | | 2| e x
 √ 

, (z − x)β ≥ 0
 √
 I√2 =| | | 2| e z


 Z Hz  
2βWs
2.10.3 Ex exp −αHz − γ e ds = √
2 e x
 K√

0  2 = | | | |
 , (z − x)β ≤ 0
 √
K 2 =| | | 2| e z

 √

∞
2 1
Z 
e−λz P0 sup ` Hz , y < h dz = 1 +



2.11.2 λ √ −1
0 y∈R h I0 ( 2h)

∞
X exp(−zj02;k =2h)
(1) P0 sup ` Hz , y < h = 8
y∈R j03;k J1 (j0;k )
z≥0 k=1

∞
1
Z
Ex e−γ Ȟ(Hz ) = Ex e−γ Ĥ(Hz ) = sh(|x − z| 2γ)
p
2.12.1 √ dh
|x−z|
h sh(h 2 )

Z ∞
h−1 ssu (|x − z|, h)dh du



2.12.2 Px Ȟ(Hz ) ∈ du = Px Ĥ(Hz ) ∈ du =
|x−z|
 2. STOPPING AT FIRST HITTING TIME 211

1 W Hz = min{s : Ws = z}

2.12.4 Px Ȟ(Hz ) ∈ du, Hz ∈ dt = Px Ĥ(Hz ) ∈ du, Hz ∈ dt
u<t
Z ∞
= ssu (|x − z|, h) st−u (h)dh dudt
|x−z|

√
sh((x − z ) 2 )
Ex e−γ Ȟ(Hz ) ; sup Ws ∈ dy =

2.13.1 √ dy
0≤s≤Hz (y − z ) sh((y − z ) 2 )
z<x
x<y

sup Ws ∈ dy = (y − z)−1 ssu (x − z, y − z)dudy

2.13.2 Px Ȟ(Hz ) ∈ du,
0≤s≤Hz
z<x
x<y
√ √
 −γ Ȟ(Hz )−αHz 2 sh((x − z ) 2 + 2 )
2.13.4 Ex e ; sup Ws ∈ dy = √ √ dy
0≤s≤Hz sh((y − z ) 2 ) sh((y − z ) 2 + 2 )
z<x
x<y


(1) Px Ȟ(Hz ) ∈ du, Hz ∈ dt, sup Ws ∈ dy
0≤s≤Hz
u<t

= ssu (x − z, y − z) st−u (y − z)dudtdy

√
sh((z − x) 2 )
Ex e−γ Ĥ(Hz ) ;

2.14.1 inf Ws ∈ dy = √ dy
0≤s≤Hz (z − y) sh((z − y) 2 )
x<z
y<x

Ws ∈ dy = (z − y)−1 ssu (z − x, z − y)dudy

2.14.2 Px Ĥ(Hz ) ∈ du, inf
0≤s≤Hz
x<z
y<x
√ √
2 sh((z − x) 2 + 2 )
Ex e−γ Ĥ(Hz )−αHz ;

2.14.4 inf Ws ∈ dy = √ √ dy
0≤s≤Hz sh((z − y) 2 ) sh((z − y) 2 + 2 )
x<z
y<x


(1) Px Ĥ(Hz ) ∈ du, Hz ∈ dt, inf Ws ∈ dy
0≤s≤Hz
u<t

= ssu (z − x, z − y) st−u (z − y)dudtdy

212 1. BROWNIAN MOTION

1 W Hz = min{s : Ws = z}
 x−a

 z − a, a≤x≤z
2.15.2 Px a < inf Ws , sup Ws < b =
0≤s≤Hz 0≤s≤Hz  b − x, z≤x≤b
b−z

 sh((x − a)√2 )
 sh((z − a)√2 ) , a ≤ x ≤ z

Ex e−αHz ; a <

2.15.4 inf Ws , sup Ws < b = √
0≤s≤Hz 0≤s≤Hz  sh((b − x)√2 ) , z ≤ x ≤ b

sh((b − z ) 2 )

sst (x − a, z − a)dt, a ≤ x ≤ z



(1) Px Hz ∈ dt, a < inf Ws , sup Ws < b =
0≤s≤t 0≤s≤t sst (b − x, b − z)dt, z ≤ x ≤ b

Ex e−γ`(Hz ,r) ; sup Ws < b


2.16.1
0≤s≤Hz
r<b
z<r
 b−x
 b − z + 2 (b − r)(r − z ) ,
 r≤x≤b
= (b − r)(x − z )
 r−x +

, z≤x≤r
r−z (r − z ) b − z + 2 (b − r)(r − z )

2.16.2 Px `(Hz , r) ∈ dy, sup Ws < b
0≤s≤Hz
r<b
z<r
b−x (b − z )y 
 
 2(b − r)(r − z ) exp − 2(b − r)(r − z ) dy,
 r≤x≤b
=
 x − z exp − (b − z )y
 
dy, z≤x≤r

2(r − z )2 2(b − r)(r − z )
(

0, r≤x≤b
(1) Px `(Hz , r) = 0, sup Ws < b = r−x
, z≤x≤r
0≤s≤Hz
r−z

Ex e−γ`(Hz ,r) ; a <


2.17.1 inf Ws
0≤s≤Hz
a<r
r<z
x−r (z − x)(r − a)

+
, r≤x≤z
z−r (z − r) z − a + 2 (z − r)(r − a)


=
x−a
 , a≤x≤r
z − a + 2 (z − r)(r − a)

 2. STOPPING AT FIRST HITTING TIME 213

1 W Hz = min{s : Ws = z}

2.17.2 Px `(Hz , r) ∈ dy, a < inf Ws
0≤s≤Hz
a<r
r<z
z−x (z − a)y 
 
 2(z − r)2 exp − 2(z − r)(r − a) dy,
 r≤x≤z
=
x−a

(z − a)y 
exp − dy, a≤x≤r

2(z − r)(r − a) 2(z − r)(r − a)


( x−r
, r≤x≤z
z−r


(1) Px `(Hz , r) = 0, a < inf Ws =
0≤s≤Hz 0, a≤x≤r

2.18.1 Ex exp −γ`(Hz , r) − η`(Hz , u)
r<u

 1, x≤z
1 + 2 (r − x) + 2(u − x) + 4 (u − r)(r − x)




 1 + 2 (r − z ) + 2(u − z ) + 4 , z≤x≤r
(u − r)(r − z )


z<r = 1 + 2(u − x)
, r≤x≤u
1 + 2 (r z ) + 2 (u − z ) + 4 (u − r)(r − z )


 −
1



, u≤x

1 + 2 (r − z ) + 2(u − z ) + 4 (u − r)(r − z )

 1
, x≤r
1 + 2 (z r ) + 2  (z u) + 4 (z − u)(u − r)



 − −
1 + 2 ( x − r)


r≤x≤u

,
1 + 2 ( ) + 2 ( )+4 (z − u)(u − r)

u<z = z − r  z − u
 1 + 2 (x − r ) + 2  ( x − u)+4 (x − u)(u − r)
, u≤x≤z

1 + 2 ( z r ) + 2  ( z u ) +4 (z − u)(u − r)




 − −
1, z≤x


(18)
2.18.2 Px `(Hz , r) ∈ dy, `(Hz , u) ∈ dg =: Bx,z (y, g)dydg
r<u
 x−z
pr dydg, z≤x≤r
 r−z


z<r = u−x x−r
pr dydg + p dydg, r ≤ x ≤ u


 u − r u −r u
pu dydg, u≤x

qr dydg, x≤r

 u−x x−r


q dydg + q dydg, r ≤ x ≤ u
u<z = u−r r u−r u
 z−x


q dydg, u≤x≤z
z−u u
214 1. BROWNIAN MOTION

1 W Hz = min{s : Ws = z}

where
√   √yg 
y y+g y

pr := √ exp − − I
4(u − r)(r − z ) g 2(u − r) 2(r − z ) 1 u − r
  √yg 
1 
y+g g
qr := exp − − I
4(z − u)(u − r) 2(u − r) 2(z − u) 0 u − r
  √yg 
1 
y+g y
pu := exp − − I
4(u − r)(r − z ) 2(u − r) 2(r − z ) 0 u − r
√   √yg 
g y+g g

qu := √ exp − − I
4(z − u)(u − r) y 2(u − r) 2(z − u) 1 u − r

(1) Px `(Hz , r) ∈ dy, `(Hz , u) = 0

x−z y(u − z )
  
 2(r − z )2 exp − 2(u − r)(r − z ) dy,
 z≤x≤r
=
u−x y (u − z )
 
exp − dy, r ≤ x ≤ u

2(u − r)(r − z ) 2(u − r)(r − z )


(2) Px `(Hz , r) = 0, `(Hz , u) ∈ dg

x−r g(z − r)
  
 2(z − u)(u − r) exp − 2(z − u)(u − r) dg,
 r≤x≤u
=
 z − x exp − g(z − r)
 
dg, u≤x≤z

2(z − u) 2 2(z − u)(u − r)

 r−x
, z≤x≤r
r−z


(3) Px `(Hz , r) = 0, `(Hz , u) = 0 =
 x − u, u≤x≤z
z−u

 1=2 √
x I1 (2 2 x)
Hz  z 1=2 I (2√2 z ) , 0 < x ≤ z
ds
Z 
1
n   o
2.19.1 Ex exp −γ ; 0 < inf Ws = √
0
Ws 0≤s≤Hz  x1=2 K1 (2 2 x)
√ , z≤x
z 1=2 K1 (2 2 z )


H0
ds
 Z  p p
2.19.1 Ex exp −γ =2 2γxK1 (2 2γx) 0≤x
0
Ws
(1)

Z H0
ds

2x
2.19.2 Px ∈ dy = 2 e−2x/y dy 0≤x
0
Ws y
(1)
 2. STOPPING AT FIRST HITTING TIME 215

1 W Hz = min{s : Ws = z}

  √
x 1/2+ 1/4+γ 2
, 0<x≤z

2 Hz
ds z
n  Z  o 
2.20.1 Ex exp − ; 0 < inf W = √
2 0
Ws2 0≤s≤Hz
s
  x 1/2− 1/4+γ 2
, z≤x


z

Hz
ds
Z 
2.20.4 Px ∈ dy, 0 < inf Ws
0
Ws2 0≤s≤Hz

 1=2
x ln(z=x) 2
 1=2 √ 3=2 e−y/8 e− ln (z/x)/2y dy, 0 < x ≤ z
z 2y

= 1=2
 x √ln(x=z ) e−y/8 e− ln2 (x/z)/2y dy, z ≤ x

z =2 2y3=2
1

p2 q2 Ws2

n  Z Hz  o
2.21.1 Ex exp − + ds ; 0 < inf W
0
2Ws2 2 0≤s≤Hz
s

 x1=2 I√4p2 +1=4 (qx2 =2)

, 0<x≤z
z 1=2 I√4p2 +1=4 (qz 2 =2)




=
x1=2 K√ 2 (qx2 =2)
 1=2 √4p +1=4 2 , z ≤ x


z K 4p2 +1=4 (qz =2)


p2
Hz
q

n  Z  o
2.22.1 Ex exp − + ds ; 0 < inf W
0
2Ws2 Ws 0≤s≤Hz
s

 x1=2 I√ √
4p2 +1 (2 2qx)
 z 1=2 I√ 2 (2√2qz ) , 0 < x ≤ z



4p +1
= 1=2 K√ √
x (2 2qx)
 1=2 √4p +1 √
 2
, z≤x

z K 4p2 +1 (2 2qz )


n  Z Hz


 1
o
2.27.1 Ex exp − p1I(−∞,r) Ws + q1I[r,∞) Ws ds ; ` Hz , r ∈ dy
0
2

=: Q(27)
z (2p, 2q, y)dy

√ √
2p sh((x − z ) 2p) p p p

z≤r = √ exp −y 2p cth((r − z) 2p) − y 2q dy
sh2 ((r − z ) 2p)
x≤r
√
2p p p p

z≤r = √ exp −y 2p cth((r − z) 2p) − (y + x − r) 2q dy
sh((r − z ) 2p)
r≤x
216 1. BROWNIAN MOTION

1 W Hz = min{s : Ws = z}

√
2q p p p

r≤z = √ exp −y 2q cth((z − r) 2q) − (y + r − x) 2p dy
sh((z − r) 2q)
x≤r
√ √
2q sh((z − x) 2q) p p p

r≤z = 2 √ exp −y 2q cth((z − r) 2q) − y 2p dy
sh ((z − r) 2q)
r≤x

n  Z Hz



o
(1) Ex exp − p1I(−∞,r) Ws + q1I[r,∞) Ws ds ; ` Hz , r = 0
0

 −(z−x)√2p
e , x≤z
√

sh((r − x) 2p)



 sh((r − z )√2p) ,

 z≤x≤r
= √
sh((x − r) 2q)
 √ , r≤x≤z
sh((z − r) 2q)




 √
e−(x−z) 2q

, z≤x

Z Hz Hz
1
Z 




2.27.2 Px 1I(−∞,r) Ws ds ∈ du, 1I[r,∞) Ws ds ∈ dv, ` Hz , r ∈ dy
0 0
2

=: Bz(27) (u, v, y)dudvdy

z≤r = escsu (x − z, o, r − z, y) hv (1, y)dudvdy

x≤r

z≤r = esu (1, 1, r − z, 0, y) hv (1, y + x − r)dudvdy

r≤x

r≤z = hu (1, y + r − x) esv (1, 1, z − r, 0, y)dudvdy

x≤r

r≤z = hu (1, y) escsv (z − x, 0, z − r, y)dudvdy

r≤x

Z Hz

Z Hz


(1) Px 1I(−∞,r) Ws ds ∈ du, 1I[r,∞) Ws ds = 0, ` Hz , r = 0
0 0

hu (1, z − x)du, x≤z


=
ssu (r − x, r − z)du, z≤x≤r

Z Hz

Z Hz


(2) Px 1I(−∞,r) Ws ds = 0, 1I[r,∞) Ws ds ∈ dv, ` Hz , r = 0
0 0
 2. STOPPING AT FIRST HITTING TIME 217

1 W Hz = min{s : Ws = z}

ssv (x − r, z − r)dv, r≤x≤z


=
hv (1, x − z)dv, z≤x

p2
 Z Hz

2.30.1 Ex exp − e2βWs + qeβWs ds
0
2
 − x=2
e M−q=p| 2p e x

|;0 | |
2p e z
, (z − x)β ≥ 0

 e− z=2 M−q=p|

|;0 | |

= 2p e x

 e− x=2 W−q=p| |;0 | |
2p e z
, (z − x)β ≤ 0

 − z=2

e W−q=p| |;0 | |



2.31.1 Ex exp − sup `(s, u) − α−1 `(s, r)
2 0≤s≤Hz
0<α

for z ≤ x ∧ r ∧ u or x ∨ r ∨ u ≤ z

( + |r − z | − |u − z |)(|u − z | ∧ |x − z |)
=1−
2 (|u − z | ∧ |r − z |)|u − r| +  + |r − z | − |u − z |

2(|u − z | ∧ |r − z |)|u − r| − ( + |r − z | − |u − z |)|u − z |)(|x − z | ∧ |r − z |)

−
(2 (|u − z | ∧ |r − z |)|u − r| +  + |r − z | − |u − z |)(|u − z | ∧ |r − z |)

 
`(s, u) − α−1 `(s, r) > h


2.31.2 Px sup
0≤s≤Hz
0<α

for z ≤ x ∧ r ∧ u or x ∨ r ∨ u ≤ z


( + |r − z | − |u − z |)h h |x − z | ∧ |r − z |
= exp −
4(|u − z | ∧ |r − z |)|u − r| |u − z | ∧ |r − z |

 + |r − z | − |u − z | (|x − z | ∧ |r − z |)|u − z |
 i
− − |u − z| ∧ |x − z|
2(|u − z | ∧ |r − z |)|u − r| |u − z | ∧ |r − z |

p
Υ := (|r − z| − α|u − z|)2 + 4α(|u − z| ∧ |r − z|)|u − r|
218 1. BROWNIAN MOTION

1 W H = Ha,b = min{s : Ws ∈
/ (a, b)}

3. Stopping at first exit time

√ √
sh (b − x) 2 + sh (x − a) 2 ch (b + a − 2x) =2


p

3.0.1 Ex e−αH = √
=
sh (b − a) 2 ch (b − a) =2
p

b + a − 2x b − a

3.0.2 Px (H ∈ dt) = sst (b − x, b − a)dt + sst (x − a, b − a)dt = cct , dt
2 2

b − x iβa x − a iβb
3.0.3 Ex eiβWH = e + e
b−a b−a

b−x
3.0.4 Px W H = a =
b−a
(a)

x−a
3.0.4 Px W H = b =
b−a
(b)

√
sh (b − x) 2


−αH

3.0.5 Ex e ; WH = a = √
sh (b − a) 2


(a)

√
sh (x − a) 2


Ex e−αH ; WH = b =

3.0.5 √
sh (b − a) 2


(b)

3.0.6 Px H ∈ dt, WH = a = sst (b − x, b − a)dt
(a)

3.0.6 Px H ∈ dt, WH = b = sst (x − a, b − a)dt
(b)

(b − y)(x − a)
3.1.6 Px sup Ws ≥ y, WH = a = , a≤x≤y≤b
0≤s≤H (b − a)(y − a)

Ex e−αH ; sup Ws ≥ y, WH = a

3.1.8
0≤s≤H
(1)
 3. STOPPING AT FIRST EXIT TIME 219

1 W H = Ha,b = min{s : Ws ∈
/ (a, b)}

√ √
sh (b − y) 2 sh (x − a) 2



= √
√
, a≤x≤y≤b
sh (b − a) 2 sh (y − a) 2

3.1.8 Px sup Ws ≥ y, WH = a, H ∈ dt
0≤s≤H
(2)

= sst (b − y, b − a) ∗ sst (x − a, y − a)dt, a≤x≤y≤b

(b − x)(y − a)
3.2.6 Px inf Ws ≤ y, WH = b = , a≤y≤x≤b
0≤s≤H (b − y)(b − a)

Ex e−αH ; inf

3.2.8 Ws ≤ y, WH = b
0≤s≤H
(1)

√ √
sh (b − x) 2 sh (y − a) 2



= √
√
, a≤y≤x≤b
sh (b − y) 2 sh (b − a) 2

3.2.8 Px inf Ws ≤ y, WH = b, H ∈ dt
0≤s≤H
(2)

= sst (b − x, b − y) ∗ sst (y − a, b − a)dt, a≤y≤x≤b

b−a+2 (x − r)(r − a)

 b−a+2 , r≤x≤b
(b − r)(r − a)

−γ`(H,r)
3.3.1 Ex e =
 b−a+2
 (b − r)(r − x)
, a≤x≤r
b−a+2 (b − r)(r − a)

 (b − x)(b − a) (b − a)y 
 

 2(b − r)2 (r − a) exp − 2(b − r)(r − a) dy, r≤x≤b
3.3.2 Px `(H, r) ∈ dy =
 (x − a)(b − a) exp − (b − a)y
 
dy, a≤x≤r

2(b − r)(r − a)2 2(b − r)(r − a)

 x−r
, r≤x≤b
b−r


(1) Px `(H, r) = 0 = r − x
 , a≤x≤r
r−a
√ √ √ √
2 sh((b − r) 2 ) + 2 sh((r − a) 2 )
3.3.3 Er e−αH−γ`(H,r) = √ √ √ √ =: L
2 sh((b − a) 2 ) + 2 sh((b − r) 2 ) sh((r − a) 2 )
220 1. BROWNIAN MOTION

1 W H = Ha,b = min{s : Ws ∈
/ (a, b)}

 sh((x − r)√2 ) sh((b − x) 2

√
)
 sh((b − r)√2 + √ L, r≤x≤b
) sh((b − r) 2 )

−αH−γ`(H,r)
Ex e = √ √
 sh((r − x)√2
 ) sh((x − a) 2
+ √
)
L, a≤x≤r
sh((r − a) 2 ) sh((r − a) 2 )

√ √

2 sh((b − a) 2 )y 
Ex e−αH ; `(H, r) ∈ dy = exp −

3.3.4 √ √
2 sh((b − r) 2 ) sh((r − a) 2 )

√ √ √ √
 2 sh((b − x) 2 )[sh((b − r) 2 ) + sh((r − a) 2 )]
√ √ dy, r≤x≤b
2 sh2 ((b − r) 2 ) sh((r − a) 2 )


× √ √ √ √
2 sh((x − a) 2 )[sh((b − r) 2 ) + sh((r − a) 2 )]
dy, a≤x≤r

√ √
2 sh((b − r) 2 ) sh2 ((r − a) 2 )


 sh((x − r)√2 )
 sh((b − r)√2 , r≤x≤b
)

Ex e−αH ; `(H, r) = 0 =

(1) √
 sh((r − x)√2
 )
, a≤x≤r
sh((r − a) 2 )

b−x

 b − a + 2 (b − r)(r − a) , r ≤ x ≤ b

Ex e−γ`(H,r) ; WH = a =: Q(3)

3.3.5 a (2γ) =
(a)  b − x + 2 (b − r)(r − x) , a ≤ x ≤ r

b − a + 2 (b − r)(r − a)

x − a + 2 (x − r)(r − a)

, r≤x≤b
b − a + 2 (b − r)(r − a)


Ex e−γ`(H,r) ; WH
 (3)
3.3.5 = b =: Qb (2γ) = x−a
(b)
 , a≤x≤r
b − a + 2 (b − r)(r − a)


3.3.6 Px `(H, r) ∈ dy, WH = a =: Ba(3) (y)dy
(a)

b−x (b − a)y 
 
 2(b − r)(r − a) exp − 2(b − r)(r − a) dy,
 r≤x≤b
=
 x − a exp − (b − a)y
 
dy, a≤x≤r

2(r − a)2 2(b − r)(r − a)

(

0, r≤x≤b
(1) Px `(H, r) = 0, WH =a = r−x
, a≤x≤r
r−a
 3. STOPPING AT FIRST EXIT TIME 221

1 W H = Ha,b = min{s : Ws ∈
/ (a, b)}

3.3.6 Px `(H, r) ∈ dy, WH = b =: Bb(3) (y)dy
(b)
b−x (b − a)y 
 
 2(b − r)2 exp − 2(b − r)(r − a) dy,
 r≤x≤b
=
x−a

(b − a)y 
exp − dy, a≤x≤r

2(b − r)(r − a) 2(b − r)(r − a)


( x−r
, r≤x≤b
b−r


(1) Px `(H, r) = 0, WH =b =
0, a≤x≤r

Ex e−αH−γ`(H,r) ; WH = a =: Va(3) (2α)


3.3.7
(a)

√ √
 2 sh((b − x) 2 )
√ √ √ √ , r≤x≤b
2 sh((b − a) 2 ) + 2 sh((b − r) 2 ) sh((r − a) 2 )


= √ √ √ √
 2 sh((b − x) 2 ) + 2 sh((b − r) 2 ) sh((r − x) 2 )
√ √ √ √ , a≤x≤r
2 sh((b − a) 2 ) + 2 sh((b − r) 2 ) sh((r − a) 2 )


Ex e−αH−γ`(H,r) ; WH = b =: Vb(3) (2α)


3.3.7
(b)

√ √ √ √
 2 sh((x − a) 2 ) + 2 sh((x − r) 2 ) sh((r − a) 2 )
√ √ √ √ , r≤x≤b
2 sh((b − a) 2 ) + 2 sh((b − r) 2 ) sh((r − a) 2 )


= √ √
2 sh((x − a) 2 )
, a≤x≤r

 √ √ √ √
2 sh((b − a) 2 ) + 2 sh((b − r) 2 ) sh((r − a) 2 )
√ √

2 sh((b − a) 2 )y 
Ex e−αH ; `(H, r) ∈ dy, WH = a =exp −

3.3.8 √ √
2 sh((b − r) 2 ) sh((r − a) 2 )
(a)

√ √
 2 sh((b − x) 2 )
 2 sh((b − r)√2 ) sh((r − a)√2 ) dy,
 r≤x≤b
× √ √
 2 sh((x − a√
 ) 2 )
dy, a≤x≤r
2 sh2 ((r − a) 2 )


 0, r≤x≤b
√
Ex e−αH ; `(H, r) = 0, WH = a = sh((r − x) 2 )

(1)
√ , a≤x≤r
sh((r − a) 2 )

222 1. BROWNIAN MOTION

1 W H = Ha,b = min{s : Ws ∈
/ (a, b)}

√ √

2 sh((b − a) 2 )y 
Ex e−αH ; `(H, r) ∈ dy, WH = b =exp −

3.3.8 √ √
2 sh((b − r) 2 ) sh((r − a) 2 )
(b)

 √2 sh((b − x)√2 )
 2 sh2 ((b − r)√2 ) dy,
 r≤x≤b
× √ √
2 sh((x − a) 2 )
dy, a≤x≤r

√ √
2 sh((b − r) 2 ) sh((r − a) 2 )


√
 sh((x − r)√ 2 ) ,

r≤x≤b
Ex e−αH ; `(H, r) = 0, WH sh((b − r) 2 )

(1) =b =
0, a≤x≤r


 Z H

3.4.1 Ex exp −γ 1I[r,b] Ws ds
0
√ √
r−x (x − a)(sh((b − r) 2 ) + 2 (r − a))

+ √ √ √ , a≤x≤r
r − a (r − a)(sh((b − r) 2 ) + 2 (r − a) ch((b − r) 2 ))




√
sh((x − r) 2 )


= √
 sh((b − r) 2 )
√ √ √
sh((b − x) 2 )(sh((b − r) 2 ) + 2 (r − a))



+ √ √ √ √ , r≤x≤b

sh((b − r) 2 )(sh((b − r) 2 ) + 2 (r − a) ch((b − r) 2 ))

n  Z H
 o
3.4.5 Ex exp −γ 1I[r,b] Ws ds ; WH = a =: Q(4)
a (2γ)
(a) 0

√ √ √
sh((b − r) 2 ) + ( r − x) 2 ch((b − r) 2 )

 sh((b − r)√2 √ √ , a≤x≤r
) + ( r − a) 2 ch((b − r) 2 )

= √
sh((b − x) 2 )
√ √ √ , r≤x≤b

sh((b − r) 2 ) + ( r − a) 2 ch((b − r) 2 )


n  Z H
 o
3.4.5 Ex exp −γ 1I[r,b] Ws ds ; WH = b =: Q(4)
b (2γ)
(b) 0

√
(x − a) 2

 sh((b − r)√2 ) + (r − a)√2 ch((b − r)√2 ) ,
 a≤x≤r
= √ √ √
 sh((x − r)√2 ) + (r − a)√2 ch((x − r)√ 2 ) ,

r≤x≤b
sh((b − r) 2 ) + (r − a) 2 ch((b − r) 2 )
 3. STOPPING AT FIRST EXIT TIME 223

1 W H = Ha,b = min{s : Ws ∈
/ (a, b)}

Z H

3.4.6 Px 1I[r,b] Ws ds ∈ dy, WH = a =: Ba(4) (y)dy
(a) 0

( x−a
rc (0, b − r, b − r, r − a)dy, a ≤ x ≤ r
= r−a y
rcy (0, b − r, b − x, r − a)dy, r≤x≤b
( r−x
H
Z  , a≤x≤r
r−a


(1) Px 1I[r,b] Ws ds = 0, WH =a =
0 0, r≤x≤b

Z H

3.4.6 Px 1I[r,b] Ws ds ∈ dy, WH = b =: Bb(4) (y)dy
(b) 0


 (x − a)rc
 e y (1, b − r, 0, r − a)dy, a≤x≤r
= ssy (x − r, b − r)dy

+(r − a) ssy (b − x, b − r) ∗ rce y (1, b − r, 0, r − a)dy, r ≤ x ≤ b


Z H

(1) Px 1I[r,b] Ws ds = 0, WH = b = 0
0

Z H
 Z ∞
3.4.8 Px 1I[r,b] Ws ds ∈ dv, H ∈ dt, WH = a = Ba(27) (t − v, v, y)dy dvdt
(a) 0 0

for Ba(27) (u, v, y) see 3.27.6(a)

Z H
 Z ∞
3.4.8 Px 1I[r,b] Ws ds ∈ dv, H ∈ dt, WH = b = Bb(27) (t − v, v, y)dy dvdt
(b) 0 0

for Bb(27) (u, v, y) see 3.27.6(b)

 Z H

3.5.1 Ex exp −γ 1I[a,r] Ws ds
0
√
sh((r − x) 2 )

√
sh((r − a) 2 )




√ √ √
sh((x − a) 2 )(sh((r − a) 2 ) + (b − r) 2 )


= + √ √ √ √ , a≤x≤r
 sh((r − a) 2 )(sh((r − a) 2 ) + (b − r) 2 ch((r − a) 2 ))
√ √
 x−r + (b − x)(sh((r − a) 2 ) + 2 (b − r))



√ √ √ , r≤x≤b

b−r (b − r)(sh((r − a) 2 ) + (b − r) 2 ch((r − a) 2 ))
224 1. BROWNIAN MOTION

1 W H = Ha,b = min{s : Ws ∈
/ (a, b)}

n  Z H
 o
3.5.5 Ex exp −γ 1I[a,r] Ws ds , WH = a =: Q(5)
a (2γ)
(a) 0

√ √ √
sh((r − x) 2 ) + 2 (b − r) ch((r − x) 2 )

 sh((r − a)√2 ) + √2 (b − r) ch((r − a)√2 , a≤x≤r
)

= √
(b − x) 2
√ √ √ , r≤x≤b

sh((r − a) 2 ) + (b − r) 2 ch((r − a) 2 )


n  Z H
 o
3.5.5 Ex exp −γ 1I[a,r] Ws ds ; WH = b =: Q(5)
b (2γ)
(b) 0

√
sh((x − a) 2 )

 sh((r − a)√2 √ √ , a≤x≤r
) + (b − r ) 2 ch((r − a) 2 )

= √ √ √
 sh((r − a) √2
 ) + (x − r ) 2
√
ch((r − a) 2
√
)
, r≤x≤b
sh((r − a) 2 ) + (b − r ) 2 ch((r − a) 2 )

Z H

3.5.6 Px 1I[a,r] Ws ds ∈ dy, WH = a =: Ba(5) (y)dy
(a) 0


 ssy (r − x, r − a)dy

= +(b − r) ssy (x − a, r − a) ∗ rc e y (1, r − a, 0, b − r)dy, a ≤ x ≤ r

(b − x)rc
e y (1, r − a, 0, b − r)dy, r≤x≤b


Z H

(1) Px 1I[a,r] Ws ds = 0, WH = a = 0
0

Z H

3.5.6 Px 1I[a,r] Ws ds ∈ dy, WH = b =: Bb(5) (y)dy
(b) 0

rcy (0, r − a, x − a, b − r)dy, a≤x≤r

(
= b−x
rc (0, r − a, r − a, b − r)dy, r ≤ x ≤ b
b−r y

(
Z H
 0, a≤x≤r
(1) Px 1I[a,r] Ws ds = 0, WH = b = x−r
, r≤x≤b
0 b−r
 3. STOPPING AT FIRST EXIT TIME 225

1 W H = Ha,b = min{s : Ws ∈
/ (a, b)}

Z H
 Z ∞
3.5.8 Px 1I[a,r] Ws ds ∈ dv, H ∈ dt, WH = a = Ba(27) (v, t − v, y)dy dvdt
(a) 0 0

for Ba(27) (u, v, y) see 3.27.6(a)

Z H
 Z ∞
3.5.8 Px 1I[a,r] Ws ds ∈ dv, H ∈ dt, WH = b = Bb(27) (v, t − v, y)dy dvdt
(b) 0 0

for Bb(27) (u, v, y) see 3.27.6(b)

 Z H



3.6.1 Ex exp − p1I[a,r] Ws + q1I[r,b] Ws ds
0

√ √
sh((x − r) 2q) sh((b − x) 2q)

 sh((b − r)√2q) + sh((b − r)√2q) Q,
 r≤x≤b
= √ √
 sh((r − x)√2p) + sh((x − a)√ 2p) Q,

a≤x≤r
sh((r − a) 2p) sh((r − a) 2p)
where

√ √ √ √
p sh((b − r) 2q) + q sh((r − a) 2p)
Q := √ √ √ √ √ √
q ch((b − r) 2q) sh((r − a) 2p) + p sh((b − r) 2q) ch((r − a) 2p)

n  Z H


 o
3.6.5 Ex exp − p1I[a,r] Ws + q1I[r,b] Ws ds ; WH = a =: Q(6)
a (2p, 2q)
(a) 0
√ √ √ √ √ √
q ch((b − r) 2q) sh((r − x) 2p) + p sh((b − r) 2q) ch((r − x) 2p)
a≤x = √ √ √ √ √ √
q ch((b − r) 2q) sh((r − a) 2p) + p sh((b − r) 2q) ch((r − a) 2p)
x≤r

√ √
p sh((b − x) 2q)
r≤x = √ √ √ √ √ √
q ch((b − r) 2q) sh((r − a) 2p) + p sh((b − r) 2q) ch((r − a) 2p)
x≤b
n  Z H


 o
3.6.5 Ex exp − p1I[a,r] Ws + q1I[r,b] Ws ds ; WH = b =: Q(6)
b (2p, 2q)
(b) 0

√ √
q sh((x − a) 2p)
a≤x = √ √ √ √ √ √
q ch((b − r) 2q) sh((r − a) 2p) + p sh((b − r) 2q) ch((r − a) 2p)
x≤r
226 1. BROWNIAN MOTION

1 W H = Ha,b = min{s : Ws ∈
/ (a, b)}

√ √ √ √ √ √
q ch((x − r) 2q) sh((r − a) 2p) + p sh((x − r) 2q) ch((r − a) 2p)
r≤x = √ √ √ √ √ √
q ch((b − r) 2q) sh((r − a) 2p) + p sh((b − r) 2q) ch((r − a) 2p)
x≤b

Z H

Z H

3.6.6 Px 1I[a,r) Ws ds ∈ du, 1I[r,b] Ws ds ∈ dv, WH = a
(a) 0 0
Z ∞
= Ba(27) (u, v, y)dy dudv
0

for Ba(27) (u, v, y) see 3.27.6(a)

Z H

Z H

3.6.6 Px 1I[a,r) Ws ds ∈ du, 1I[r,b] Ws ds ∈ dv, WH = b
(b) 0 0
Z ∞
= Bb(27) (u, v, y)dy dudv
0

for Bb(27) (u, v, y) see 3.27.6(b)

 Z H

3.7.1 Er exp −γ 1I[r,u] Ws ds =: Qr
a≤r 0

u≤b
√ √ √ √
sh((u − r) 2 ) + (b − u) 2 ch((u − r) 2 ) + (r − a) 2
= √ √ √
(1 + (b − u)(r − a)2 ) sh((u − r) 2 ) + (b − u + r − a) 2 ch((u − r) 2 )

 Z H

Eu exp −γ 1I[r,u] Ws ds =: Qu
0

√ √ √ √
sh((u − r) 2 ) + (b − u) 2 + (r − a) 2 ch((u − r) 2 )
= √ √ √
(1 + (b − u)(r − a)2 ) sh((u − r) 2 ) + (b − u + r − a) 2 ch((u − r) 2 )

 Z H

Ex exp −γ 1I[r,u] Ws ds
0
 r−x x−a
+ Q , a≤x≤r
r−a r−a r


√ √

sh((u − x) 2 ) sh((x − r) 2 )


= √ Qr + √ Q , r≤x≤u
 sh((u − r) 2 ) sh((u − r) 2 ) u
 x − u + b − xQ ,



u≤x≤b
b−u b−u u
 3. STOPPING AT FIRST EXIT TIME 227

1 W H = Ha,b = min{s : Ws ∈
/ (a, b)}

n  Z H
 o
3.7.5 Ex exp −γ 1I[r,u] Ws ds ; WH = a =: Q(7)
a (2γ)
(a) 0

√ √ √
(1 + (b − u)(r − x)2 ) sh((u − r) 2 ) + (b − u + r − x) 2 ch((u − r) 2 )
a≤x = √ √ √
(1 + (b − u)(r − a)2 ) sh((u − r) 2 ) + (b − u + r − a) 2 ch((u − r) 2 )
x≤r

√ √ √
sh((u − x) 2 ) + (b − u) 2 ch((u − x) 2 )
r≤x = √ √ √
(1 + (b − u)(r − a)2 ) sh((u − r) 2 ) + (b − u + r − a) 2 ch((u − r) 2 )
x≤u

√
(b − x) 2
u≤x = √ √ √
(1 + (b − u)(r − a)2 ) sh((u − r) 2 ) + (b − u + r − a) 2 ch((u − r) 2 )
x≤b

n  Z H
 o
3.7.5 Ex exp −γ 1I[r,u] Ws ds ; WH = b =: Q(7)
b (2γ)
(b) 0

√
(x − a) 2
a≤x = √ √ √
(1 + (b − u)(r − a)2 ) sh((u − r) 2 ) + (b − u + r − a) 2 ch((u − r) 2 )
x≤r

√ √ √
sh((x − r) 2 ) + (r − a) 2 ch((x − r) 2 )
r≤x = √ √ √
(1 + (b − u)(r − a)2 ) sh((u − r) 2 ) + (b − u + r − a) 2 ch((u − r) 2 )
x≤u

√ √ √
(1 + (x − u)(r − a)2 ) sh((u − r) 2 ) + (x − u + r − a) 2 ch((u − r) 2 )
u≤x = √ √ √
(1 + (b − u)(r − a)2 ) sh((u − r) 2 ) + (b − u + r − a) 2 ch((u − r) 2 )
x≤b

 Z H 
3.8.1 Ex exp −γ Ws ds
0

xbS1=3 ( 23 b3=2 2 ; 23 x3=2 2 ) + xaS1=3 ( 23 x3=2 2 ; 23 a3=2 2 )

√ √ √ √
=
abS1=3 ( 23 b3=2 2 ; 23 a3=2 2 )
√ √

Z H o xS ( 2 b3=2 √2 ; 23 x3=2 2 )
√
1=3 3
n  
3.8.5 Ex exp −γ Ws ds ; WH = a =
aS1=3 ( 23 b3=2 2 ; 23 a3=2 2 )
√ √
(a) 0

Z H o xS ( 2 x3=2 √2 ; 23 a3=2 2 )
√
1=3 3
n  
3.8.5 Ex exp −γ Ws ds ; WH = b =
bS1=3 ( 23 b3=2 2 ; 23 a3=2 2 )
√ √
(b) 0
228 1. BROWNIAN MOTION

1 W H = Ha,b = min{s : Ws ∈
/ (a, b)}
2 H
 Z 
3.9.3 Ex exp −αH − Ws2 ds
2 0

e(a −x ) =2 S ( 12 + ; b 2 ; x 2 ) + e(b −x ) =2 S ( 12 + ; x 2 ; a 2 )
2 2 √ √ 2 2 √ √
=
S ( 12 + ; b 2 ; a 2 )
√ √

2
Z H o ea2 =2 S ( 1 + ; b√2 ; x√2 )
Ws2 ds ; WH = a = x2 =2 21
n  
3.9.7 Ex exp −αH − √ √
2 0 e S( 2 + ; b 2 ; a 2 )
(a)

eb =2 S ( 12 + ; x 2 ; a 2 )
√ √
2 H 2
n  Z  o
3.9.7 Ex exp −αH − Ws2 ds ; WH = b = x2 =2 1 √ √
2 0 e S( 2 + ; b 2 ; a 2 )
(b)
√
S0 2 e b ; √2 e x + S √2 e x ; √2 e a
  
H 0 | |
 Z 
2βWs | | | | | |
3.10.1 Ex exp −γ e ds = √ √
S0 | 2| e b ; | 2| e a

0

√ √
o S√2 =| 2 e b; 2 e x

n  Z H  | | | | |
3.10.7 Ex exp −αH − γ e2βWs ds ; WH =a = √
2 e √
S√2 =| b; 2 e a

(a) 0 | | | | |

√ √
S√2 =| | | 2| e x ; | 2| e a

n  Z H  o
2βWs
3.10.7 Ex exp −αH − γ e ds ; WH =b = √ √
S√2 =| | | 2| e b ; | 2| e a

(b) 0

√
b−a
sh((x − a) 2 )
Z
Ex e−γ Ȟ(H) ; WH = a =

3.12.5 √ dy
x−a
y sh(y 2 )
(a)

√
b−a
sh((b − x) 2 )
Z
−γ Ĥ(H)

3.12.5 Ex e ; WH = b = √ dy
b−x
y sh(y 2 )
(b)

Z b−a
y −1 ssu (x − a, y)dy du


3.12.6 Px Ȟ(H) ∈ du, WH = a =
(a) x−a

Z b−a
y −1 ssu (b − x, y)dy du


3.12.6 Px Ĥ(H) ∈ du, WH = b =
(b) b−x

b−a √ √
2 sh((x − a) 2 + 2 )
Z
−αH−γ Ȟ(H)

3.12.7 Ex e ; WH = a = √ √ dy
x−a sh(y 2 ) sh(y 2 + 2 )
(a)
 3. STOPPING AT FIRST EXIT TIME 229

1 W H = Ha,b = min{s : Ws ∈
/ (a, b)}

b−a √ √
2 sh((b − x) 2 + 2 )
Z
−αH−γ Ĥ(H)

3.12.7 Ex e ; WH = b = √ √ dy
b−x sh(y 2 ) sh(y 2 + 2 )
(b)


Z b−a
3.12.8 Px Ȟ(H) ∈ du, H ∈ dt, WH = a = ssu (x − a, y) st−u (y)dy dudt
(a) x−a

u<t


Z b−a
3.12.8 Px Ĥ(H) ∈ du, H ∈ dt, WH = b = ssu (b − x, y) st−u (y)dy dudt
(b) b−x

u<t

3.18.1 Ex exp −γ`(H, r) − η`(H, u)
r<u
b − a + 2 (b − r)(r − x) + 2(b − u)(u − x) + 4 (b − u)(u − r)(r − x)
a≤x =
b − a + 2 (b − r)(r − a) + 2(b − u)(u − a) + 4 (b − u)(u − r)(r − a)
x≤r
b − a + 2 (x − r)(r − a) + 2(b − u)(u − x)
r≤x =
b − a + 2 (b − r)(r − a) + 2(b − u)(u − a) + 4 (b − u)(u − r)(r − a)
x≤u
b − a + 2 (x − r)(r − a) + 2(x − u)(u − a) + 4 (x − u)(u − r)(r − a)
u≤x =
b − a + 2 (b − r)(r − a) + 2(b − u)(u − a) + 4 (b − u)(u − r)(r − a)
x≤b



3.18.5 Ex exp −γ`(H, r) − η`(H, u) ; WH = a =: Q(18)
a (2γ, 2η)
(a)
b − x + 2 (b − r)(r − x) + 2(b − u)(u − x) + 4 (b − u)(u − r)(r − x)
a≤x =
b − a + 2 (b − r)(r − a) + 2(b − u)(u − a) + 4 (b − u)(u − r)(r − a)
x≤r
b − x + 2(b − u)(u − x)
r≤x =
b − a + 2 (b − r)(r − a) + 2(b − u)(u − a) + 4 (b − u)(u − r)(r − a)
x≤u
b−x
u≤x =
b − a + 2 (b − r)(r − a) + 2(b − u)(u − a) + 4 (b − u)(u − r)(r − a)
x≤b



3.18.5 Ex exp −γ`(H, r) − η`(H, u) ; WH = b =: Q(18)
b (2γ, 2η)
(b)
x−a
a≤x =
b − a + 2 (b − r)(r − a) + 2(b − u)(u − a) + 4 (b − u)(u − r)(r − a)
x≤r
x − a + 2 (x − r)(r − a)
r≤x =
b − a + 2 (b − r)(r − a) + 2(b − u)(u − a) + 4 (b − u)(u − r)(r − a)
x≤u
230 1. BROWNIAN MOTION

1 W H = Ha,b = min{s : Ws ∈
/ (a, b)}

x − a + 2 (x − r)(r − a) + 2(x − u)(u − a) + 4 (x − u)(u − r)(r − a)

u≤x =
b − a + 2 (b − r)(r − a) + 2(b − u)(u − a) + 4 (b − u)(u − r)(r − a)
x≤b

(18)
3.18.6 Px `(H, r) ∈ dy, `(H, u) ∈ dg, WH = a =: Bx,a (y, g)dydg
(a)
 x−a
p dydg, a≤x≤r
 ur − a r


−x x−r

= p dydg + p dydg, r ≤ x ≤ u
u −r r u −r u
 b−x



p dydg, u≤x
b−u u
where
√   √yg 
y y+g y g

pr := √ exp − − − I
4(u − r)(r − a) g 2(u − r) 2(r − a) 2(b − u) 1 u − r
  √yg 
1 
y+g y g
pu := exp − − − I
4(u − r)(r − a) 2(u − r) 2(r − a) 2(b − u) 0 u − r

(1) Px `(H, r) ∈ dy, `(H, u) = 0, WH = a

x−a y (u − a)
  
 2(r − a)2 exp − 2(u − r)(r − a) dy,
 a≤x≤r
=
u−x y(u − a)
 
exp − dy, r ≤ x ≤ u

2(u − r)(r − a) 2(u − r)(r − a)


r−x
(2) Px `(H, r) = 0, `(H, u) = 0, WH = a = , a≤x≤r
r−a

(18)
3.18.6 Px `(H, r) ∈ dy, `(H, u) ∈ dg, WH = b =: Bx,b (y, g)dydg
(b)
 x−a
q dydg, x≤r
 ur − a r


−x x−r

u<b = qr dydg + q dydg, r ≤ x ≤ u
 u−r u−r u
 b−x


q dydg, u≤x≤b
b−u u
where
√
1 
y+g y g
  yg 
qr := exp − − − I
4(b − u)(u − r) 2(u − r) 2(r − a) 2(b − u) 0 u − r
√   √yg 
g y+g y g

qu := √ exp − − − I
4(b − u)(u − r) y 2(u − r) 2(r − a) 2(b − u) 1 u − r
 3. STOPPING AT FIRST EXIT TIME 231

1 W H = Ha,b = min{s : Ws ∈
/ (a, b)}


(1) Px `(H, r) = 0, `(H, u) ∈ dg, WH = b
x−r g(b − r)
  
 2(b − u)(u − r) exp − 2(b − u)(u − r) dg,
 r≤x≤u
=
 b − x exp − g(b − r)
 
dg, u≤x≤b

2(b − u) 2 2(b − u)(u − r)

x−u
(2) Px `(H, r) = 0, `(H, u) = 0, WH = b = , u≤x≤b
b−u

√ √
xS (2 2 b; 2 2 x)
H
ds
Z
; WH = a = 1 √
n   o
3.19.5 Ex exp −γ √ 0<a≤x≤b
0
Ws aS1 (2 2 b; 2 2 a)
(a)

√ √
xS (2 2 x; 2 2 a)
H
ds
Z
; WH = b = 1 √
n   o
3.19.5 Ex exp −γ √ 0<a≤x≤b
0
Ws bS1 (2 2 b; 2 2 a)
(b)
 2Z H o x1=2 (b=x)√1=4+ √
− (x=b) 1=4+
ds
n  2 2

3.20.5 Ex exp − ; WH = a = √ √
2 0
Ws2 a1=2 (b=a) 1=4+ 2
− (a=b) 1=4+
2

(a)
0<a
 2Z H o x1=2 (x=a)√1=4+ √
− (a=x) 1=4+
ds
n  2 2

3.20.5 Ex exp − ; WH = b = √ √
2 0
Ws2 b1=2 (b=a) 1=4+ 2
− (a=b) 1=4+
2

(b)
0<a

p2 q2 Ws2

n  Z H  o
3.21.5 Ex exp − + ds ; W = a
0
2Ws2 2 H
(a)
√
x p2 +1=4+1=2 S√4p2 +1=4 (qb2 =2; qx2 =2)
= √ , 0<a≤x≤b
a p2 +1=4+1=2 S√4p2 +1=4 (qb2 =2; qa2 =2)

p2 q2 Ws2

n  Z H  o
3.21.5 Ex exp − + ds ; W = b
0
2Ws2 2 H
(b)
√
x p2 +1=4+1=2 S√4p2 +1=4 (qx2 =2; qa2 =2)
= √ , 0<a≤x≤b
b p2 +1=4+1=2 S√4p2 +1=4 (qb2 =2; qa2 =2)

p2
H
q

n  Z  o
3.22.5 Ex exp − 2 + ds ; WH = a
0
2Ws Ws
(a)
√ √ √
x p2 +1=4+1=2 S√4p2 +1 (2 2qb; 2 2qx)
= √ √ √ , 0<a≤x≤b
a p2 +1=4+1=2 S√4p2 +1 (2 2qb; 2 2qa)
232 1. BROWNIAN MOTION

1 W H = Ha,b = min{s : Ws ∈
/ (a, b)}

p2
H
q

n  Z  o
3.22.5 Ex exp − + ds ; W = b
0
2Ws2 Ws H
(b)
√ √ √
x p2 +1=4+1=2 S√4p2 +1 (2 2qx; 2 2qa)
= √ √ √ , 0<a≤x≤b
b p2 +1=4+1=2 S√4p2 +1 (2 2qb; 2 2qa)

n  Z H


 1
o
3.27.1 Ex exp − p1I[a,r) Ws + q1I[r,b] Ws ds ; `(H, r ∈ dy
0
2
p p p p

= exp −y( 2q cth((b − r) 2q) + 2p cth((r − a) 2p))
 √2 sh((b − x)√2q)[√p sh((b − r)√2q) + √q sh((r − a)√2p)]
√ √ dy, r ≤ x ≤ b
sh2 ((b − r) 2q) sh((r − a) 2p)


× √ √ √ √ √ √
 2 sh((x − a) 2p)[ p sh((

√
b − r) 2q) + q sh((r − a) 2p)]
√ dy, a ≤ x ≤ r
2
sh((b − r) 2q) sh ((r − a) 2p)

n  Z H


 1
o
3.27.5 Ex exp − p1I[a,r) Ws + q1I[r,b] Ws ds ; `(H, r ∈ dy, WH = a
0
2
(a)
p p p p

= exp −y( 2q cth((b − r) 2q) + 2p cth((r − a) 2p))

√ √
2p sh((b − x) 2q)

 sh((b − r)√2q) sh((r − a)√2p) dy, r ≤ x ≤ b

× √ √
 2p sh((x − a√) 2p) dy, a≤x≤r
sh2 ((r − a) 2p)


n  Z H



o
(1) Ex exp − p1I[a,r) Ws + q1I[r,b] Ws ds ; ` Hz , r = 0, WH = a
0
√
sh((r − x) 2p)
= √ , a≤x≤r
sh((r − a) 2p)

n  Z H


 1
o
3.27.5 Ex exp − p1I[a,r) Ws + q1I[r,b] Ws ds ; ` Hz , r ∈ dy, WH = b
0
2
(b)
p p p p

= exp −y( 2q cth((b − r) 2q) + 2p cth((r − a) 2p))

√ √
2q sh((b − x) 2q)

√ dy, r≤x≤b
sh2 ((b − r) 2q)


× √ √
2q sh((x − a) 2p)
√ √ dy, a ≤ x ≤ r

sh((b − r) 2q) sh((r − a) 2p)

 3. STOPPING AT FIRST EXIT TIME 233

1 W H = Ha,b = min{s : Ws ∈
/ (a, b)}

n  Z H



o
(1) Ex exp − p1I[a,r) Ws + q1I[r,b] Ws ds ; ` H, r = 0, WH = b
0

√
sh((x − r) 2q)
= √ , r≤x≤b
sh((b − r) 2q)

Z H H
1
Z 



3.27.6 Px 1I[a,r) Ws ds ∈ du, 1I[r,b] Ws ds ∈ dv, `(H, r) ∈ dy, WH = a
0 0
2
(a)

=: Ba(27) (u, v, y)dudvdy

escsu (x − a, 0, r − a, y)dudvdy, a≤x≤r


=
esu (1, 1, r − a, 0, y) essv (b − x, b − r, 0, y)dudvdy, r ≤ x ≤ b

Z H H
1
Z 



3.27.6 Px 1I[a,r) Ws ds ∈ du, 1I[r,b] Ws ds ∈ dv, `(H, r) ∈ dy, WH = b
0 0
2
(b)

=: Bb(27) (u, v, y)dudvdy

essu (x − a, r − a, 0, y) esv (1, 1, b − r, 0, y)dudvdy, a≤x≤r


=
esu (0, 0, r − a, 0, y) escsv (b − x, 0, b − r, y)dudvdy, r ≤ x ≤ b

p2
n  Z H   o
3.30.5 Ex exp − e2βWs + qeβWs ds ; WH = a
0
2
(a)

exp − | p | e x S p|q + 12 ; 1; |2p| e b ; |2p| e x

   
|
=
exp − | p | e a S p|q + 12 ; 1; |2p| e b ; |2p| e a
   
|

p2
n  Z H   o
3.30.5 Ex exp − e2βWs + qeβWs ds ; WH = b
0
2
(b)

exp − | p | e x S p|q + 12 ; 1; |2p| e x ; |2p| e a

   
|
=
exp − | p | e b S p|q + 12 ; 1; |2p| e b ; |2p| e a
   
|
234 1. BROWNIAN MOTION

1 W H = Ha,b = min{s : Ws ∈
/ (a, b)}

n 

3.31.5 Ex(b−a)+a exp − sup `(s, u(b − a) + a) − α−1 `(s, r(b − a) + a) ;
2(b − a) 0≤s≤H
(a)

o
(1 − u)( + r − r2 − u + u2 )(u ∧ x − ux)
0<α WH = a = 1 − x −
2 (u ∧ r − ur)|u − r| +  + r − r2 − u + u2

(1 − u) 2(u ∧ r − ur)|u − r| − ( + r − r2 − u + u2 )(u − u2 ))(x ∧ r − xr)

−
(2 (u ∧ r − ur)|u − r| +  + r − r2 − u + u2 )(u ∧ r − ur)

n 

3.31.5 Ex(b−a)+a exp − sup `(s, u(b − a) + a) − α−1 `(s, r(b − a) + a) ;
2(b − a) 0≤s≤H
(b)

u( + r − r2 − u + u2 )(u ∧ x − ux)

o
0<α WH = b = x −
2 (u ∧ r − ur)|u − r| +  + r − r2 − u + u2

u 2(u ∧ r − ur)|u − r| − ( + r − r2 − u + u2 )(u − u2 ))(x ∧ r − xr)

−
(2 (u ∧ r − ur)|u − r| +  + r − r2 − u + u2 )(u ∧ r − ur)


`(s, u(b − a) + a) − α−1 `(s, r(b − a) + a) > (b − a)h,


3.31.6 Px(b−a)+a sup
0≤s≤H
(a)

 
( + r − r2 − u + u2 )h h x ∧ r − xr
0<α WH = a = (1 − u) exp −
4(u ∧ r − ur)|u − r| u ∧ r − ur

 + r − r2 − u + u2 (x ∧ r − xr)(u − u2 )
 i
− − u ∧ x + ux
2(u ∧ r − ur)|u − r| u ∧ r − ur


`(s, u(b − a) + a) − α−1 `(s, r(b − a) + a) > (b − a)h,


3.31.6 Px(b−a)+a sup
0≤s≤H
(b)

 
( + r − r2 − u + u2 )h h x ∧ r − xr
0<α WH = b = u exp −
4(u ∧ r − ur)|u − r| u ∧ r − ur

 + r − r2 − u + u2 (x ∧ r − xr)(u − u2 )
 i
− − u ∧ x + ux
2(u ∧ r − ur)|u − r| u ∧ r − ur

p
Υ := (r − r2 − αu + αu2 )2 + 4α(u ∧ r − ur)|u − r|
 4. STOPPING AT INVERSE LOCAL TIME 235

1 W % = %(v, z) = min{s : `(s, z) = v}

4. Stopping at inverse local time

√
4.0.1 Ex e−α% = e−(|z−x|+v) 2α

| z − x| + v (|z − x| + v)2
4.0.2 Px % ∈ dt = √ 3=2 exp(− )dt
2t 2t

Ex exp −γ sup Ws = e−γz 2vγK1 ( 2vγ)

p p
4.1.1
0≤s≤%
x≤z
v
  
exp − , x≤z
2(y − z )




4.1.2 Px sup Ws < y =
 y − x exp − v
 
0≤s≤%  , z≤x≤y
y−z 2(y − z )

Ex e−α% ; sup Ws < y


4.1.4
0≤s≤%

√ √
√ v e(y−z) 2
  
 exp −(z − x) 2α − 2 sh((y − z )√2
√ , x≤z≤y

)

= √ √ (y−z)√2
 sh((y − x)√2 ) exp − √v e
  
 √ , z≤x≤y
sh((y − z ) 2 ) 2 sh((y − z ) 2 )

( v v

est 0, 0, y − z, z − x + , dt, x ≤ z ≤ y
(1) Px sup Ws < y, % ∈ dt =

2 2
v v

0≤s≤% esst y − x, y − z, , dt, z≤x≤y
2 2

p p
Ex exp γ inf Ws = eγz 2vγK1 ( 2vγ)


4.2.1
0≤s≤%
z≤x
 x−y v
  

 z − y exp − 2(z − y) , y≤x≤z
4.2.2 Px inf Ws > y =
 exp − v
 
0≤s≤%  , y≤z≤x
2(z − y)

Ex e−α% ; inf Ws > y


4.2.4
0≤s≤%
236 1. BROWNIAN MOTION

1 W % = %(v, z) = min{s : `(s, z) = v}

√ √ √
sh((x − y) 2 ) v e(z−y) 2
  
 sh((z − y) 2 ) exp − 2 sh((z − y)√2
√ √ , y≤x≤z

)

= √ (z−y)√2
√
 exp −(x − z) 2α − √v e
  
 √ , y≤z≤x
2 sh((z − y) 2 )

( v v

esst x − y, z − y, , dt, y≤x≤z
(1) Px inf Ws > y, % ∈ dt =

2 2
v v

0≤s≤% est 0, 0, z − y, x − z + , dt, y ≤ z ≤ x
2 2
v
  
exp − , x ≤ z ≤ r, r ≤ z ≤ x


 1 + 2 |r − z |
1 + 2 | r − x| v

  
4.3.1 Ex e−γ`(%,r) = exp − , r∧z ≤x≤r∨z
 1 + 2 |r − z | 1 + 2 |r − z |
1 v

  

 exp − , z ≤ r ≤ x, x ≤ r ≤ z
1 + 2 |r − z | 1 + 2 |r − z |

4.3.2 Px `(%, r) ∈ dy

√ √

1 v

(v + y)   vy 
√ exp − I dy, x ≤ z ≤ r, r ≤ z ≤ x
2|r − z | y 2|r − z | 1
|r − z |




√ √
1 (v + y) v vy

  h  

exp − |r − x| √ I1

2(r − z )2 2|r − z | y |r − z |


=  √vy i
+|z − x|I dy, z∧r ≤x≤r∨z

|r − z |

 0


√
1 (v + y) vy

    

exp − I dy, z ≤ r ≤ x, x ≤ r ≤ z

2|r − z | 2|r − z | 0 |r − z |


v
  
exp − , x ≤ z ≤ r, r≤z≤x
2|r − z |





| r − x| v
 
(1) Px `(%, r) = 0 = exp − , z∧r ≤x≤r∨z
 |r − z |

 2| r − z |
z ≤ r ≤ x, x≤r≤z

0,

√ √ √
v 2 ( + 2 )
 
4.3.3 Ex e−α%−γ`(%,r) =: Vz(3) (2α) = exp −|z − x| 2α − √ √

2 + 1 − e−2|r−z| 2

1, x ≤ z ≤ r, r≤z≤x

√ √
−2|r −x| 2

 2 + 1−e




√ √
−2|r −z | 2

, r∧z ≤x≤r∨z
× 2 + 1−e
√
2



 √ √
, z ≤ r ≤ x, x≤r≤z

2 + 1 − e−2|r−z| 2
 4. STOPPING AT INVERSE LOCAL TIME 237

1 W % = %(v, z) = min{s : `(s, z) = v}

√

(v + y) 2 
Ex e−α% ; `(%, r) ∈ dy =: Fz(3) (2α, y)dy = exp −

4.3.4 √
−2|r −z | 2
1−e

√ √ √ √
e−|z−x| 2 v 2 vy
  
 √ √ √ I1 √ dy, x ≤ z ≤ r, r ≤ z ≤ x
2 sh(|r − z | 2 ) y sh(|r − z | 2 )



√ √ √

√ v 2 vy

 h  

 √ 2 √ sh(|r − x| 2α) √ I1 √
2 sh (|r − z | 2 ) y sh(|r − z | 2 )


× √
√ 
2 vy i
+ sh(|z − x| 2α)I0 √ dy, z ∧ r ≤ x ≤ r ∨ z

sh(|r − z | 2 )




√ −|r−x|√2 √

e 2 vy

  

√ √ I0 √ dy, z ≤ r ≤ x, x ≤ r ≤ z

2 sh(|r − z | 2 ) sh(|r − z | 2 )


Ex e−α% ; `(%, r) = 0

(1)

 −|z−x|√2α
e , x ≤ z ≤ r, r≤z≤x
√ √
v 2 sh(|r − x| 2 )
 

= exp − √ √ , r∧z ≤x≤r∨z
1 − e−2|r−z| 2  sh(|r − z | 2 )

0, z ≤ r ≤ x, x≤r≤z

(2) Px `(%, r) ∈ dy, % ∈ dt =: Fz(3) (y, t)dydt

for x ≤ z ≤ r or r ≤ z ≤ x
√
v √
= ist (1, |r − z|, |z − x| + (y + v)/2, (y + v)/2, vy/2)dydt
2√y

for z ≤ r ≤ x or x ≤ r ≤ z
1 √
= is (0, |r − z|, |r − x| + (y + v)/2, (y + v)/2, vy/2)dydt
2 t

 Z %

4.4.1 Ex exp −γ 1I[r,∞) Ws ds
0
√
v 2
  
exp − √ , x≤z
2(1 + (r − z ) 2 )




√ √
1 + ( r − x) 2 v 2

  
z≤r = √ exp − √ , z≤x≤r
 1 + (r − z ) 2 2(1 + (r − z ) 2 )
√
1 v 2


  √ 
√ exp (r − x) 2γ − √ , r≤x

1 + (r − z ) 2 2(1 + (r − z ) 2 )

238 1. BROWNIAN MOTION

1 W % = %(v, z) = min{s : `(s, z) = v}

√
v e(z−r) 2 
√
1
 
√ exp − √ √ , x≤r
ch((z − r) 2 )

2 ch((z − r) 2 )




√ √
v e(z−r) 2 
√
ch((x − r) 2 )

 
r≤z = √ exp − √ √ , r≤x≤z
 ch((z − r) 2 ) 2 ch((z − r) 2 )
√

√ (z−r) 2
 exp (z − x)√2γ − √v e

  

, z≤x

√
2 ch((z − r) 2 )

Z %

4.4.2 Px 1I[r,∞) Ws ds ∈ dy
0

√ Z ∞√  2  √ 
v

v u u 2vu
z≤r = √ exp − u exp − − I du dy
2 (r − z )y3=2 2(r − z ) 0 2y r − z 1 r − z
x≤z

Z ∞  2  √ 
1 
v − 2x + 2r u u 2vu
z≤r = √ exp − u exp − − I0 du dy
2(r − z )y3=2 2(r − z ) x−r
2y r − z r − z
r≤x

r≤z = ecy (0, 1, z − r, v/2, v/2)dy

x≤r

r≤x = eccy (x − r, z − r, v/2, v/2)dy

x≤z

r≤z = ecy (0, 0, z − r, v/2 + x − z, v/2)dy

z≤x

v
  
exp − , x≤z
2(r − z )


Z %  
r−x v

 
(1) Px 1I[r,∞) Ws ds = 0 = exp − , z≤x≤r
z≤r 0 

 r−z 2(r − z )
0, r≤x


 Z %

4.5.1 Ex exp −γ 1I(−∞,r] Ws ds
0

√
v e(r−z) 2 
√
√
 
 exp (x − z) 2γ − √ √ , x≤z
2 ch((r − z ) 2 )




√ √
v e(r−z) 2 
√
ch((r − x) 2 )

 
z≤r = √ exp − √ √ , z≤x≤r
 ch((r − z ) 2 ) 2 ch((r − z ) 2 )
√

v e(r−z) 2 
√
1

 

√ exp − √ , r≤x

√
ch((r − z ) 2 ) 2 ch((r − z ) 2 )

 4. STOPPING AT INVERSE LOCAL TIME 239

1 W % = %(v, z) = min{s : `(s, z) = v}

√
1 √ v 2
  
√ exp (x − r) 2γ − √ , x≤r
1 + (z − r ) 2 2(1 + (z − r) 2 )




√ √
1 + (x − r ) 2 v 2

  
r≤z = √ exp − √ , r≤x≤z
 1 + ( z − r ) 2 2(1 + ( z − r) 2 )
√
v 2

  

 exp −

√ , z≤x
2(1 + (z − r) 2 )

Z %

4.5.2 Px 1I(−∞,r] Ws ds ∈ dy
0

z≤r = ecy (0, 0, r − z, v/2 + z − x, v/2)dy

x≤z

z≤x = eccy (r − x, r − z, v/2, v/2)dy

x≤r

z≤r = ecy (0, 1, r − z, v/2, v/2)dy

r≤x
Z ∞  2  √ 
1 
v − 2r + 2x u u 2vu
r≤z = √ exp − u exp − − I du dy
2(z − r)y3=2 2(z − r) r−x
2y z − r 0 z − r
x≤r
√ Z ∞√  2  √ 
v

v u u 2vu
r≤z = √ exp − u exp − − I du dy
2 (z − r)y3=2 2(z − r) 0 2y z − r 1 z − r
z≤x

 0, x≤r
x r v

−
Z %   
exp − , r≤x≤z
 
z−r 2(z − r)


(1) Px 1I(−∞,r] Ws ds = 0 =
 exp − v
r≤z 0 
  
, z≤x

2(z − r)

 Z %



4.6.1 Ex exp − p1I(−∞,r) Ws + q1I[r,∞) Ws ds =: Q(6)
z (2p, 2q)
0
√
v p( p + q)e(r−z) 2p
 √ √ √ 
z≤r = exp − √ √ √ √ √
2( p ch((r − z ) 2p) + q sh((r − z ) 2p))
 −(z−x)√2p
 e , x≤z
√ √ √ √
 p ch((r − x) 2p) + q sh((r − x) 2p)



√ √ √ √ , z≤x≤r
× p ch((r − z ) 2p) + q sh((r − z ) 2p)
√
√ −(x−r) 2q
pe



√ √ , r≤x

 √ √
p ch((r − z ) 2p) + q sh((r − z ) 2p)
√
v q( p + q)e(z−r) 2q
 √ √ √ 
r≤z = exp − √ √ √ √ √
2( q ch((z − r) 2q) + p sh((z − r) 2q))
240 1. BROWNIAN MOTION

1 W % = %(v, z) = min{s : `(s, z) = v}

√ −(r−x)√2p
qe

√ √ √√ , x≤r
q ch((z − r) 2q) + p sh((z − r) 2q)




 √ √ √ √
× q ch((x − r) 2q) + p sh((x − r) 2q)
√ √ √ √ , r≤x≤z


 q ch((z − r) 2q) + p sh((z − r) 2q)
 √
e−(x−z) 2q

, z≤x

Z %

Z %

4.6.2 Px 1I(−∞,r) Ws ds ∈ du, 1I[r,∞) Ws ds ∈ dg
0 0
Z ∞
= Bz(27) (u, g, y)dy dudg for Bz(27) (u, g, y) see 4.27.2
0
 Z %

4.7.1 Ex exp −γ 1I[r,u] Ws , ds
0

√ √
 sh((u − r) 2 )
v 
z≤r = exp − √ √ √ √
2((r − z ) 2 sh((u − r) 2 ) + ch((u − r) 2 ))
 (r − x)√2 √
sh((u − r) 2 ) + ch((u − r) 2
√
)
√ √ √ , z≤x≤r
(r − z ) 2 sh((u − r) 2 ) + ch((u − r) 2 )



√
ch((u − x) 2 )


× √ √ √ , r≤x≤u
 (r − z ) 2 sh((u − r) 2 ) + ch((u − r) 2 )
1



 √ √ √ , u≤x
(r − z ) 2 sh((u − r) 2 ) + ch((u − r) 2 )
√ √
 sh((u − r) 2 )
v 
u≤z = exp − √ √
√ √
2((z − u) 2 sh((u − r) 2 ) + ch((u − r) 2 ))
 1
√ √ √ , x≤r

 (z − u) 2 sh((u − r) 2 ) + ch((u − r) 2 )
√

ch((x − r) 2 )


× √ √ √ , r≤x≤u
(z − u) 2 sh((u − r) 2 ) + ch((u − r) 2 )
√ √ √

(x − u) 2 sh((u − r) 2 ) + ch((u − r) 2 )



 √ √ √ , u≤x≤z
(z − u) 2 sh((u − r) 2 ) + ch((u − r) 2 )

 Z  %(v,0) 
(6 )1=3 (2=3)v 
|Ws |ds = 32/3 Γ(2/3) Ai (2γ)1/3 |x| exp −


4.8.1 Ex exp −γ
0
(1=3)
2 %
 Z 
4.9.3 Ex exp −α% − Ws2 ds
2 0
√
D− 1 − α (−x 2 ) √

2 γ
 v  
exp − 1 , x≤z

√ √ √
 D− 1 − α (−z 2 ) ( 2 + )D− 1 − α (−z 2 )D− 1 − α (z 2 )



2 γ 2 γ 2 γ
= √
 D− 1 − α (x 2 )  √
v  
2 γ
 D 1 α (z √2 ) exp − ( 1 + )D 1 α (−z √2 )D 1 α (z √2 ) , z≤x



− − 2 γ 2 − − − − 2 γ 2 γ
 4. STOPPING AT INVERSE LOCAL TIME 241

1 W % = %(v, z) = min{s : `(s, z) = v}

%
v| |
 Z   
4.10.3 Ex exp −α% − γ e2βWs ds = exp − √ √
2I√2 =| | | 2| e z K√2 =| | | 2| e z
 
0
√
I 2 =| | | 2| e x
 √ 

, (z − x)β ≥ 0
 √
 I√2 =| | | 2| e z



× √
K√2 =| | | 2| e x



 , (z − x)β ≤ 0
 √
K 2 =| | | 2| e z

 √

Z ∞
e−λz P0 sup ` %, y < h dz



4.11.2 λ
0 y∈R
√
v

h
v 2  I0 ( 2v) i
= 1− 1− + √ − 1 1I[v,∞) (h)
h h h I0 ( 2h)

v
X exp(−zj02;k =2h)J0 (j0;k v=h)

∞ p



(1) P0 sup ` %, y < h = 8 1 − 1I[0,h] (v)
y∈R h j03;k J1 (j0;k )
z≥0 k=1


v
2
(2) P0 sup ` %(v, 0), y < h = 1 − 1I[0,h] (v)
y∈R h

√
sh((x − z )+ 2 ) 
v

Ex e−γ Ȟ(%) ; sup Ws ∈ dy =

4.13.1 √ exp − dy
0≤s≤% (y − z ) sh((y − z ) 2 ) 2(y − z )

√ √ √
2 2 e−(y−z) 2  
v 2
 
v

+ √ √ exp − √ − exp −
1 − 2(y − z ) 2 − e2(z−y) 2 1 − e2(z−y) 2 2(y − z )

 √
 e−(z−x) 2γ dy, x≤z
× sh((y − x)√2 )
√ dy, z ≤ x ≤ y
sh((y − z ) 2 )


√
sh((z − x)+ 2 ) 
v

Ex e−γ Ĥ(%) ; inf Ws ∈ dy =

4.14.1 √ exp − dy
0≤s≤% (z − y) sh((z − y) 2 ) 2(z − y)

√ √ √
2 2 e−(z−y) 2  
v 2
 
v

+ √ 2(y z )√
2 exp − 2(y z )√
2 − exp −
1 − 2(z − y) 2 − e − 1−e − 2(z − y)

√
 sh((x − y)√2 ) dy, y ≤ x ≤ z


× sh((z − y√) 2 )
 −(x−z) 2γ
e dy, z≤x
242 1. BROWNIAN MOTION

1 W % = %(v, z) = min{s : `(s, z) = v}

4.15.2 Px a < inf Ws , sup Ws < b
0≤s≤% 0≤s≤%

x−a (b − a)v 
 
 z − a exp − 2(z − a)(b − z ) ,
 a≤x≤z
=
 b − x exp − (b − a)v
 
, z≤x≤b

b−z 2(z − a)(b − z )

b+x bv
  
 b + z exp − (b + z )(b − z ) ,


−b ≤ x ≤ z
(1) Px sup |Ws | ≤ b =
 b − x exp − bv
 
0≤s≤%  , z≤x≤b
b−z (b + z )(b − z )

Ex e−α% ; a < inf Ws , sup Ws < b


4.15.4
0≤s≤% 0≤s≤%

 sh((x − a)√2 )  √
v 2 sh((b − a) 2 )
√ 
 sh((z − a)√2 exp − √ √ , a≤x≤z
) 2 sh((b − z ) 2 ) sh((z − a) 2 )

= √ √ √
 sh((b − x)√2 ) v 2 sh((b − a) 2 )
 
exp − , z≤x≤b

√ √
sh((b − z ) 2 ) 2 sh((b − z ) 2 ) sh((z − a) 2 )

Ex e−α% ; sup |Ws | ≤ b


(1)
0≤s≤%

 sh((b + x)√2 )  √
v 2
√
sh((2b 2 ) 
 sh((b + z )√2 exp − √ √ , −b ≤ x ≤ z
) ch(2b 2 ) − ch(2z 2 )

= √ √ √
 sh((b − x)√2 ) v 2 sh(2b 2 )
 
exp − , z≤x≤b

√ √
sh((b − z ) 2 ) ch(2b 2 ) − ch(2z 2 )

Ex e−γ`(%,r) ; sup Ws < b


4.16.1
0≤s≤%
r<b
z<r
b−x

 b − z + 2 (b − r)(r − z ) , r ≤ x ≤ b
v(1 + 2 (b − r))
 
= exp −
 b − x + 2 (b − r)(r − x) , z ≤ x ≤ r
2(b − z + 2 (b − r)(r − z )) 
b − z + 2 (b − r)(r − z )


v (b − z )y 
4.16.2 Px `(%, r) ∈ dy, sup Ws < b = exp − −
0≤s≤% 2(r − z ) 2(b − r)(r − z )
r<b
z<r
 4. STOPPING AT INVERSE LOCAL TIME 243

1 W % = %(v, z) = min{s : `(s, z) = v}

 √vy 
 b−x
 2(b − r)(r − z ) I0 r − z dy,
 r≤x≤b
× √  √vy   √vy i
1 h
v
(r − x) √ I1 + (x − z)I0 dy, z ≤ x ≤ r

2(r − z )2 y r−z r−z


0, r≤x≤b
(
r−x v


(1) Px `(%, r) = 0, sup Ws < b =  
exp − , z≤x≤r
0≤s≤%
r−z 2(r − z )

Ex e−γ`(%,r) ; a < inf Ws


4.17.1
0≤s≤%
a<r
r<z
x − a + 2 (x − r)(r − a)

, r≤x≤z
v(1 + 2 (r − a)) z − a + 2 (z − r)(r − a)
 

= exp −
2(z − a + 2 (z − r)(r − a))  x−a
 , a≤x≤r
z − a + 2 (z − r)(r − a)


v (z − a)y 
4.17.2 Px `(%, r) ∈ dy, a < inf Ws = exp − −
0≤s≤% 2(z − r) 2(z − r)(r − a)
a<r
r<z

√  √vy   √vy i
1 v
 h
 2(z − r)2 (x − r) √y I1 z − r + (z − x)I0 z − r dy, r ≤ x ≤ z

×  √vy 
x−a
I0 dy, a≤x≤r

2(z − r)(r − a) z−r


( x−r 
v

exp − , r≤x≤z
z−r 2(z − r)


(1) Px `(%, r) = 0, a < inf Ws =
0≤s≤%
0, a≤x≤r

4.18.1 Ex exp −γ`(%, r) − η`(%, u)
r<u

v( +  + 2 (u − r))
 
z<r = exp −
1 + 2 (r − z ) + 2(u − z ) + 4 (u − r)(r − z )
 1 + 2 (r − x) + 2(u − x) + 4 (u − r)(r − x)
, z≤x≤r
1 + 2 (r − z ) + 2(u − z ) + 4 (u − r)(r − z )



1 + 2(u − x)


× , r≤x≤u
 1 + 2 (r − z ) + 2(u − z ) + 4 (u − r)(r − z )
1



 , u<x
1 + 2 (r − z ) + 2(u − z ) + 4 (u − r)(r − z )
244 1. BROWNIAN MOTION

1 W % = %(v, z) = min{s : `(s, z) = v}

v( +  + 2 (u − r))
 
u<z = exp −
1 + 2 (z − r) + 2(z − u) + 4 (z − u)(u − r)
 1
, x<r


 1 + 2 (z − r ) + 2  ( z − u) + 4 (z − u)(u − r)
1 + 2 (x − r )


× 1 + 2 (z − r) + 2(z − u) + 4 , r≤x≤u
 (z − u)(u − r)
 1 + 2 (x − r) + 2(x − u) + 4 (x − u)(u − r)



, u≤x≤z
1 + 2 (z − r) + 2(z − u) + 4 (z − u)(u − r)

%
ds
n  Z  o
4.19.1 Ex exp −γ ; 0 < inf Ws
0
Ws 0≤s≤%

√
x1=2 I1 (2 2 x) v
  
1= 2 √ exp − √ √ , 0<x≤z
z I1 (2 2 z ) 4zI1 (2 2 z )K1 (2 2 z )


= 1=2 √
 x K1 (2√2 x) exp − v
 
, z≤x

1= 2 √ √
z K1 (2 2 z ) 4zI1 (2 2 z )K1 (2 2 z )
 2Z %
ds
n  o
4.20.1 Ex exp − ; 0 < inf W
2 0
Ws2 0≤s≤%
s

√
2 + 1=4 
 
x 1/2+ γ 2 +1/4 v
 p
exp − , 0<x≤z

z z


=   √
 x 1/2− γ 2 +1/4  p
v 2 + 1=4 
exp − z≤x

 ,
z z

%
ds
Z 
4.20.4 Px ∈ dy, 0 < inf Ws
0
Ws2 0≤s≤%
0<y

 1=2
x (ln(z=x) + v=z ) −y/8 −(ln(z/x)+v/z)2 /2y
√ e e dy, 0 < x ≤ z
z 1=2 2y3=2


= 1=2
 x (ln(√
 x=z ) + v=z ) −y/8 −(ln(x/z)+v/z)2 /2y
e e dy, z ≤ x
z 1=2 2y3=2

p2 q2 Ws2

n  Z %  o
4.21.1 Ex exp − 2 + ds ; 0 < inf Ws
0
2Ws 2 0≤s≤%
0<x
 √xI√ 2
4p2 +1=4 (qx =2) v
 
√ 2 exp − √ , x≤z
(qz 2 =2)K√ (qz 2 =2)

 zI√ (qz =2) zI

4p2 +1=4 4p2 +1=4 4p2 +1=4

= √ √
 xK 4p2 +1=4 (qx2 =2) 
v

exp − , z≤x

√ √
zK 4p2 +1=4 (qz 2 =2) zI√4p2 +1=4 (qz 2 =2)K√4p2 +1=4 (qz 2 =2)

 4. STOPPING AT INVERSE LOCAL TIME 245

1 W % = %(v, z) = min{s : `(s, z) = v}

p2
%
q

n  Z  o
4.22.1 Ex exp − + ds ; 0 < inf W
0
2Ws2 Ws 0≤s≤%
s
0<x
 √xI√ √
4p2 +1 (2 2qx) v
 
 √ √ exp − √ √ √ , x≤z
 zI√ (2 2qz ) 4p2 +1 (2 2qz )K 4p2 +1 (2 2qz )
4zI √

4p2 +1

= √ √ √
 xK 4p2 +1 (2 2qx) 
v

√ exp − √ √ √ , z≤x

√ √
zK 4p2 +1 (2 2qz ) 4zI 4p2 +1 (2 2qz )K√4p2 +1 (2 2qz )


n  Z %


 o
4.23.1 Ex exp − p1I(−∞,r) Ws + q1I[r,b] Ws ds ; sup Ws < b
0 0≤s≤%
r<b
z<r

√ √
v p( p th((b − r) 2q) + q)e(r−z) 2p
 √ √ √ 
= exp − √ √ √ √ √ √
2( p th((b − r) 2q) ch((r − z ) 2p) + q sh((r − z ) 2p))

√ √
 p sh((b − x) 2q)
√ √ √ √ √ √ , r≤x
p sh((b − r) 2q) ch((r − z ) 2p) + q ch((b − r) 2q) sh((r − z ) 2p)


x≤b



√ √ √ √ √ √
p sh((b − r) 2q) ch((r − x) 2p) + q ch((b − r) 2q) sh((r − x) 2p)

× √ √ √ √ √ √ , z≤x


 p sh((b − r) 2q) ch((r − z ) 2p) + q ch((b − r) 2q) sh((r − z ) 2p) x≤r

 √
e−(z−x) 2p

, x<z

n  Z %


 o
4.24.1 Ex exp − p1I[a,r) Ws + q1I[r,∞) Ws ds ; a < inf Ws
0 0≤s≤%
a<r
r<z

√ √
v q( p + q th((r − a) 2p))e(z−r) 2q
 √ √ √ 
= exp − √ √ √ √ √ √
2( p sh((z − r) 2q) + q ch((z − r) 2q) th((r − a) 2p))

 −(z−x)√2p
 e , z<x
√ √ √ √ √ √
p sh((x r ) 2q ) ch(( r a ) 2p ) + q ch(( x r ) 2q ) sh((r a ) 2p )

− − − −


√ √ √ √ , r≤x

 √ √
× p sh((z − r) 2q) ch((r − a) 2p) + q ch((z − r) 2q) sh((r − a) 2p) x ≤ z
√ √
q sh((x − a) 2p)



 √p sh((z − r) 2q) ch((r − a) 2p) + √q ch((z − r)√2q) sh((r − a)√2p) , a ≤ x
√ √


x≤r
246 1. BROWNIAN MOTION

1 W % = %(v, z) = min{s : `(s, z) = v}

n  Z %



o
4.27.1 Ex exp − p1I(−∞,r) Ws + q1I[r,∞) Ws ds ; ` %, r ∈ dy
0

=: Q(27)
z (2p, 2q, y)dy

 v√p y√q (v + y)√p ch((r − z )√2p) 

z≤r = exp − √ − √ − √ √
2 2 2 sh((r − z ) 2p)
√ √
pve−(z−x) 2p
√
2pvy
  
√ √ √ I1 √ dy, x≤z
sh((r − z ) 2p)

2 sh((r − z ) 2p) y



√

√ √
p v 2pvy


 h √  
√ sh((r − x) 2p) I √

√
2 sh2 ((r − z ) 2p)
√
y 1
sh((r − z ) 2p)


× √
√ 
2pvy i
+ sh((x − z) 2p)I0 dy, z≤x≤r

√
sh((r − z ) 2p)




√ −(x−r) 2q √ √

pe 2pvy

  

√ √ I0 √ dy, r≤x

2 sh((r − z ) 2p) sh((r − z ) 2p)


 v√q y√p (v + y)√q ch((z − r)√2q) 

r≤z = exp − √ − √ − √ √
2 2 2 sh((z − r) 2q)

√ −(r−x)√2p √
qe 2qvy
  
√ √ I0 √ dy, x≤r
sh((z − r) 2q)

2 sh((z − r) 2q)



√ √ √

q √ v 2qvy

 h  

 √ 2 √ sh((x − r) 2q) √ I1 √
2 sh ((z − r) 2q) y sh((z − r) 2q)


× √
√ 
2qvy i
+ sh((z − x) 2q)I0 √ dy, r≤x≤z

sh((z − r) 2q)




√ √
qve−(x−z) 2q
√

2qvy

  

 √ √ √ I1 √ dy, z≤x
2 sh((z − r) 2q) y sh((z − r) 2q)


n  Z %



o
(1) Ex exp − p1I(−∞,r) Ws + q1I[r,∞) Ws ds ; ` %, r = 0
0
 √
√
 e−(z−x) 2p , x≤z
v pe(r−z) 2p  sh((r − x)√2p)
 √ 

z≤r = exp − √ √ √ , z≤x≤r
2 sh((r − z ) 2p) 
 sh((r − z ) 2p)
0, r≤x


x≤r

0,
√ √

v qe(z−r) 2q  sh((x − r)√2q)
 √ 

r≤z = exp − √ , r≤x≤z
√
2 sh((z − r) 2q)  sh((z − r) 2q)
 −(x−z)√2q

e , z≤x
 4. STOPPING AT INVERSE LOCAL TIME 247

1 W % = %(v, z) = min{s : `(s, z) = v}

Z %

Z %


4.27.2 Px 1I(−∞,r) Ws ds ∈ du, 1I[r,∞) Ws ds ∈ dg, ` %, r ∈ dy
0 0

=: Bz(27) (u, g, y)dudgdy

√
v √
x≤z = isu (1, r − z, z − x + v/2, (y + v)/2, vy/2) hg (1, y/2)dudgdy
2√y
z≤r
1 √
z≤r = is (0, r − z, v/2, (y + v)/2, vy/2) hg (1, x − r + y/2)dudgdy
2 u
r≤x
1 √
x≤r = h (1, r − x + y/2) isg (0, z − r, v/2, (y + v)/2, vy/2)dudgdy
2 u
r≤z
√
v √
r≤z = hu (1, y/2) isg (1, z − r, x − z + v/2, (y + v)/2, vy/2)dudgdy
2√y
z≤x

Z %

Z %


(1) Px 1I(−∞,r) Ws ds ∈ du, 1I[r,∞) Ws ds = 0, ` %, r = 0
0 0

esu (0, 0, r − z, z − x + v/2, v/2)du, x≤z


=
essu (r − x, r − z, v/2, v/2)du, z≤x≤r

Z %

Z %


(2) Px 1I(−∞,r) Ws ds = 0, 1I[r,∞) Ws ds ∈ dg, ` %, r = 0
0 0

essg (x − r, z − r, v/2, v/2)dg, r≤x≤z


=
esg (0, 0, z − r, x − z + v/2, v/2)dg, z≤x

 Z % 2
p 2βWs
 
4.30.1 Ex exp − e + qeβWs ds
0
2

pve z
 
= exp −
(1=2 + q=p| |)M−q=p| |;0 |2p| e z W−q=p| |;0 |2p| e z

e
 − x=2 M 2p e x

−q=p| |;0 | |
z=2 M−q=p| |;0 2p e z
, (z − x)β ≥ 0

 e−


| |
× x=2 W 2p x

 e− −q=p| |;0 | | e
z=2 W−q=p| |;0 2p e z
, (z − x)β ≤ 0


e
 −
| |
248 1. BROWNIAN MOTION

1 W θv = min{t : sup0≤s≤t Ws − inf 0≤s≤t Ws = v}

5. Stopping at first range time

1
5.0.1 Ex e−αθv =
ch2 v =2
p

∞
4v X 2 2
5.0.2 Px (θv ∈ dt) = √ (−1)k−1 k 2 e−k v /2t dt
t 2t
k=1
2
5.0.3 Ex eiβWθv = 2 2 eiβx (βv sin(βv) + cos(βv) − 1)
v

|z − x|
5.0.4 Px Wθv ∈ dz = dz, |z − x| ≤ v
v2
√ √
2 sh |z − x| 2


Ex e−αθv ; Wθv ∈ dz =

5.0.5 √
dz, |z − x| ≤ v
sh2 v 2


5.0.6 Px θv ∈ dt, Wθv ∈ dz |z − x| ≤ v
√ ∞
2 X 2
k 1 − t−1 (|x − z| + 2kv)2 e−(|x−z|+2kv) /2t dtdz


= √
t t
k=−∞

Ex e−γ`(θv ,r) ; Wθv ∈ dz


5.3.5 |z − x| ≤ v
|z − x|

 (v + 2 (v − |z − r|)|r − z |)2 dz,
 z∧r ≤x≤z∨r
=
 |z − x| + 2 |z − r||r − x| dz,

z∧x≤r ≤z∨x
(v + 2 (v − |z − r|)|r − z |)2

vy

 
5.3.6 Px `(θv , r) ∈ dy, Wθv ∈ dz = exp −
2(v − |z − r|)|z − r|
|z − x|y

 4(v − |z − r|)2 (z − r)2 dydz,
 z∧r ≤x≤z∨r
×
 2(v − |z − r|)|x − r| + (v − |z − x|)y dydz,

z∧x≤r ≤z∨x
4(v − |z − r|)3 |z − r|

Ex e−αθv −γ`(θv ,r) ; Wθv ∈ dz


5.3.7 |z − x| ≤ v

√ √
2 sh(|z − x| 2 )

√
dz, z ∧ r ≤ x ≤ z ∨ r
sh((v − |z − r|) 2 ) sh(|r − z | 2 ) 2
√ √ √
√ 2

sh(v 2 ) +



= √ √ √ √
2 sh(|z − x| 2 ) + 2 sh(|z − r| 2 ) sh(|r − x| 2 )

dz, z ∧ x ≤ r ≤ z ∨ x

√
sh(v 2 ) + √ 2 sh((v − |z − r|) 2 ) sh(|r − z | 2 ) 2
√ √ √



 5. STOPPING AT FIRST RANGE TIME 249

1 W θv = min{t : sup0≤s≤t Ws − inf 0≤s≤t Ws = v}

Ex e−αθv ; `(θv , r) ∈ dy, Wθv ∈ dz


5.3.8
√ √

2 sh(v 2 )y 
= exp − √ √
2 sh((v − |z − r|) 2 ) sh(|z − r| 2 )
√ √
 2 sh(|z − x| 2 )y
√ √ dydz, z∧r ≤x≤z∨r
2 sh2 ((v − |z − r|) 2 ) sh2 (|z − r| 2 )



√

sh(|x − r| 2 )


× 2 √ √ dydz
 sh ((v − |z − r|) 2 ) sh(|z − r| 2 )
 √ √
2 sh((v − |z − x|) 2 )y



 + 3 √ √ dydz, z ∧ x ≤ r ≤ z ∨ x
2 sh ((v − |z − r|) 2 ) sh(|z − r| 2 )

n  Z θv
 o
5.4.5 Ex exp −γ 1I[r,∞) Ws ds ; Wθv ∈ dz
0


2 (x − z )
 (sh((v + z − r)√2 ) + (r − z )√2 ch((v + z − r)√2 ))2 dz,
 z≤x≤r
= √ √ √
2 sh((x − r) 2 ) + 2 (r − z ) ch((x − r) 2 )
√ √ √ dz, r ≤x≤z+v

(sh((v + z − r) 2 ) + (r − z ) 2 ch((v + z − r) 2 ))2


 √ √ √
2 sh((z − r) 2 ) + 2 (r − x) ch((z − r) 2 )
 (sh((z − r)√2 ) + (v − z + r)√2 ch((z − r)√2 ))2 dz,
 z−v ≤x≤r
= √ √
2 sh((z − x) 2 )
√ √ √ dz, r ≤x≤z+v

(sh((z − r) 2 ) + (v − z + r) 2 ch((z − r) 2 ))2


n  Z θv
 o
5.5.5 Ex exp −γ 1I(−∞,r] Ws ds , Wθv ∈ dz
0

√ √
2 sh((x − z ) 2 )

 (sh((r − z )√2 ) + (v + z − r)√2 ch((r − z )√2 ))2 dz,
 z≤x≤r
= √ √ √
2 sh((r − z ) 2 ) + 2 (x − r) ch((r − z ) 2 )
√ √ √ dz, r ≤x≤z+v

(sh((r − z ) 2 ) + (v + z − r) 2 ch((r − z ) 2 ))2


√ √ √
2 sh((r − x) 2 ) + 2 (z − r) ch((r − x) 2 )

√ √ √ dz, z−v ≤x≤r
(sh((v − z + r) 2 ) + (z − r) 2 ch((v − z + r) 2 ))2


=
2 (z − x )
√ √ √ dz, r≤x≤z

(sh((v − z + r) 2 ) + (z − r) 2 ch((v − z + r) 2 ))2


n  Z θv


 o
5.6.5 Ex exp − p1I(−∞,r] Ws + q1I[r,∞) Ws ds ; Wθv ∈ dz
0
250 1. BROWNIAN MOTION

1 W θv = min{t : sup0≤s≤t Ws − inf 0≤s≤t Ws = v}

√ √
q 2p sh((x − z ) 2p)
z≤x = √ √ √ √ dz
( q ch((v + z − r) 2q) sh((r − z ) 2p) + √p sh((v + z − r) 2q) ch((r − z ) 2p))2
√
x≤r

√ √ √ √ √ √
q 2p ch((x − r) 2q) sh((r − z ) 2p) + p 2q sh((x − r) 2q) ch((r − z ) 2p)
r≤x = √ √ √ √ dz
( q ch((v + z − r) 2q) sh((r − z ) 2p) + √p sh((v + z − r) 2q) ch((r − z ) 2p))2
√
x≤b

where b = z + v, a = z − v

√ √ √ √ √ √
q 2p ch((z − r) 2q) sh((r − x) 2p) + p 2q sh((z − r) 2q) ch((r − x) 2p)
a≤x = √ √ √ √ dz
( q ch((z − r) 2q) sh((v − z + r) 2p) + √p sh((z − r) 2q) ch((v − z + r) 2p))2
√
x≤r

√ √
p 2q sh((z − x) 2q)
r≤x = √ √ √ √ dz
(√q ch((z − r) 2q) sh((v − z + r) 2p) + √p sh((z − r) 2q) ch((v − z + r) 2p))2
x≤z

n  Z θv
 o
5.7.5 Ex exp −γ 1I[r,u] Ws ds ; Wθv ∈ dz
0

(x − z )2
z≤x = √ √ √ dz
((1 + (v + z − u)(r − z )2 ) sh((u − r) 2 ) + (v − u + r) 2 ch((u − r) 2 ))2
x≤r

√ √ √
2 sh((x − r) 2 ) + (r − z )2 ch((x − r) 2 )
r≤x = √ √ √ dz
((1 + (v + z − u)(r − z )2 ) sh((u − r) 2 ) + (v − u + r) 2 ch((u − r) 2 ))2
x≤u

√ √ √
(1 + (x − u)(r − z )2 ) 2 sh((u − r) 2 ) + (x − u + r − z )2 ch((u − r) 2 )
u≤x = √ √ √ dz
((1 + (v + z − u)(r − z )2 ) sh((u − r) 2 ) + (v − u + r) 2 ch((u − r) 2 ))2
x≤b

where b = z + v, a = z − v

√ √ √
(1 + (z − u)(r − x)2 ) 2 sh((u − r) 2 ) + (z − u + r − x)2 ch((u − r) 2 )
a≤x = √ √ √ dz
((1 + (z − u)(v − z + r)2 ) sh((u − r) 2 ) + (v − u + r) 2 ch((u − r) 2 ))2
x≤r

√ √ √
2 sh((u − x) 2 ) + (z − u)2 ch((u − x) 2 )
r≤x = √ √ √ dz
((1 + (z − u)(v − z + r)2 ) sh((u − r) 2 ) + (v − u + r) 2 ch((u − r) 2 ))2
x≤u

(z − x)2
u≤x = √ √ √ dz
((1 + (z − u)(v − z + r)2 ) sh((u − r) 2 ) + (v − u + r) 2 ch((u − r) 2 ))2
x≤z
 5. STOPPING AT FIRST RANGE TIME 251

1 W θv = min{t : sup0≤s≤t Ws − inf 0≤s≤t Ws = v}

n  Z θv  o
5.8.5 Ex exp −γ Ws ds ; Wθv ∈ dz
0

35=3 xS1=3 ( 23 x3=2 2 ; 23 z 3=2 2 )

 √ √

 4z (z + v)2 S 2 ( 2 (z + v)3=2 2 ; 2 z 3=2 √2 ) dz,


 √ z ≤x≤z+v
1=3 3 3
=
35=3 xS1=3 ( 23 z 3=2 2 ; 23 x3=2 2 )
√ √

dz, z−v ≤x≤z
4z (z − v)2 S12=3 ( 23 z 3=2 2 ; 23 (z − v)3=2 2 )
√ √



n  Z θv  o
5.9.5 Ex exp −2γ Ws2 ds ; Wθv ∈ dz
0

 2xS1=4 (x2 √ ; z 2 √ )
 z (z + v)2 1=4 S 2 ((z + v)2 √ ; z 2 √ ) dz,

 z ≤x≤z+v
1=4
=
2xS1=4 (z 2 √ ; x2 √ )
z−v ≤x≤z

√ dz,
z (z − v)2 1=4 S12=4 (z 2 ; (z − v)2 )
√



n  Z θv  o
5.10.5 Ex exp −γ e2βWs ds ; Wθv ∈ dz
0
√ √
S0 | 2| e x ; | 2| e z
 

 dz, z ≤x≤z+v
 √ √
 S02 | 2| e (z+v) ; | 2| e z



= √
2 e z ; √2 e x
S0


 | | | |
 dz, z−v ≤x≤z
 √ √
S02 | 2| e z ; | 2| e (z−v)



∞ √
h
Z
−λv


5.11.2 λ e Px sup `(θv , y) ≥ h dv = √ √ √
0 y∈R 2I0 ( 2h)I1 ( 2h)

∞ 

X 1 2
−j0,k v/2h 1 2
−j1,k v/2h

5.11.4 P sup `(θv , y) < h = e − e
y∈R J12 (j0;k ) J02 (j1;k )
k=1

sh(v =2)
p
Ex e−γ Ȟ(θv ) ; Wθv < x =

5.13.1 √
v 2 ch(v =2)
p

Px Ȟ(θv ) ∈ du, Wθv < x = v −1 scu (v/2, v/2)du

5.13.2

√
sh(v ( + )=2)
p
Ex e−αθv −γ Ȟ(θv ) ; Wθv < x =

5.13.3 √ √
+ sh(v 2 ) ch(v ( + )=2)
p

5.13.4 Px Ȟ(θv ) ∈ du, θv ∈ dt, Wθv < x = scu (v/2, v/2) st−u (v)dudt
u<t
252 1. BROWNIAN MOTION

1 W θv = min{t : sup0≤s≤t Ws − inf 0≤s≤t Ws = v}

√
sh((x − z ) 2 )
Ex e−γ Ȟ(θv ) ; Wθv ∈ dz =

5.13.5 √ dz
v sh(v 2 )
z<x
Px Ȟ(θv ) ∈ du, Wθv ∈ dz = v −1 ssu (x − z, v)dudz


5.13.6
z<x
√ √
2 sh((x − z ) 2 + 2 )
Ex e−αθv −γ Ȟ(θv ) ; Wθv ∈ dz =

5.13.7 √ √ dz
sh(v 2 ) sh(v 2 + 2 )
z<x


5.13.8 Px Ȟ(θv ) ∈ du, θv ∈ dt, Wθv ∈ dz = ssu (x − z, v) st−u (v)dudtdz
u<t
sh(v =2)
p
Ex e−γ Ĥ(θv ) ; Wθv > x =

5.14.1 √
v 2 ch(v =2)
p

Px Ĥ(θv ) ∈ du, Wθv > x = v −1 scu (v/2, v/2)du

5.14.2

√
sh(v ( + )=2)
p
−αθv −γ Ĥ(θv )

5.14.3 Ex e ; W θv > x = √ √
+ sh(v 2 ) ch(v ( + )=2)
p

5.14.4 Px Ĥ(θv ) ∈ du, θv ∈ dt, Wθv > x = scu (v/2, v/2) st−u (v)dudt
u<t

√
sh((z − x) 2 )
Ex e−γ Ĥ(θv ) ; Wθv ∈ dz =

5.14.5 √ dz
v sh(v 2 )
x<z
Px Ĥ(θv ) ∈ du, Wθv ∈ dz = v −1 ssu (z − x, v)dudz


5.14.6
x<z
√ √
2 sh((z − x) 2 + 2 )
Ex e−αθv −γ Ĥ(θv ) ; Wθv ∈ dz =

5.14.7 √ √ dz
sh(v 2 ) sh(v 2 + 2 )
x<z


5.14.8 Px Ĥ(θv ) ∈ du, θv ∈ dt, Wθv ∈ dz = ssu (z − x, v) st−u (v)dudtdz
u<t



5.18.5 Ex exp −γ`(θv , r) − η`(θv , u) ; Wθv ∈ dz

x−z
z≤x = dz
(v + 2 (v + z − r)(r − z ) + 2(v + z − u)(u − z ) + 4 (v + z − u)(u − r)(r − z ))2
x≤r
x − z + 2 (x − r)(r − z )
r≤x = dz
(v + 2 (v + z − r)(r − z ) + 2(v + z − u)(u − z ) + 4 (v + z − u)(u − r)(r − z ))2
x≤u
 5. STOPPING AT FIRST RANGE TIME 253

1 W θv = min{t : sup0≤s≤t Ws − inf 0≤s≤t Ws = v}

x − z + 2 (x − r)(r − z ) + 2(x − u)(u − z ) + 4 (x − u)(u − r)(r − z )

u≤x = dz
(v + 2 (v + z − r)(r − z ) + 2(v + z − u)(u − z ) + 4 (v + z − u)(u − r)(r − z ))2
x≤b

where b = z + v, a = z − v

z − x + 2 (z − r)(r − x) + 2(z − u)(u − x) + 4 (z − u)(u − r)(r − x)

a≤x = dz
(v + 2 (z − r)(v + r − z ) + 2(z − u)(v + u − z ) + 4 (z − u)(u − r)(v + r − z ))2
x≤r
z − x + 2(z − u)(u − x)
r≤x = dz
(v + 2 (z − r)(v + r − z ) + 2(z − u)(v + u − z ) + 4 (z − u)(u − r)(v + r − z ))2
x≤u
z−x
u≤x = dz
(v + 2 (z − r)(v + r − z ) + 2(z − u)(v + u − z ) + 4 (z − u)(u − r)(v + r − z ))2
x≤z

5.18.6 Px `(θv , r) ∈ dy, `(θv , u) ∈ dg, Wθv ∈ dz
z≤x
 x−z
p dydgdz, z≤x≤r
r−z r



 u−x

x−r
= pr dydgdz + p dydgdz, r ≤ x ≤ u
u − r u −r u
 z + v − x + 2(x − u) pu dydgdz,

 

u≤x≤z+v

z+v−u g

where

√   √yg 
yg y+g y g

pr := exp − − − I
8(z + v − u)2 (u − r)(r − z ) 2(u − r) 2(r − z ) 2(z + v − u) 1 u − r

  √yg 
g y+g y g

pu := exp − − − I
8(z + v − u)2 (u − r)(r − z ) 2(u − r) 2(r − z ) 2(z + v − u) 0 u − r

x≤z Px `(θv , r) ∈ dy, `(θv , u) ∈ dg, Wθv ∈ dz

2(r − x) x + v − z 
+ q dydgdz, z−v ≤x≤r
y r+v−z r




u−x x−r

= qr dydgdz + q dydgdz, r≤x≤u
 u − r u −r u
 z−x


q dydgdz, u≤x≤z
z−u u


where
254 1. BROWNIAN MOTION

1 W θv = min{t : sup0≤s≤t Ws − inf 0≤s≤t Ws = v}

  √yg 
y y+g y g

qr := exp − − − I
8(z − u)(u − r)(v + r − z )2 2(u − r) 2(r + v − z ) 2(z − u) 0 u − r
√   √yg 
yg y+g y g

qu := exp − − − I
8(z − u)(u − r)(v + r − z )2 2(u − r) 2(r + v − z ) 2(z − u) 1 u − r

θv
ds
n  Z  o
5.19.5 Ex exp −γ ; Wθv ∈ dz
0
Ws

√ √
 xS1 (2 2 x; 2 2 z )
 16z (z + v)2 S 2 (2 2 (z + v); 2√2 z ) dz,
 p 0<z ≤x≤z+v
= 1
√ √
xS 1 (2 2 z; 2 2 x)
dz, 0<z−v ≤x≤z

2 √
16z (z − v)2 S1 (2 2 z; 2 2 (z − v))
 p

 2Z θv
ds
n  o
5.20.5 Ex exp − 2 ; Wθv ∈ dz
2 0
Ws

√ √ √
2 1=4 + 2 x (x=z ) 1=4+ 2 − (z=x) 1=4+ 2
 p

 √z (z + v) ((z + v)=z ) 1=4+ 2 − (z=(z + v)) 1=4+ 2
2 dz, z ≤ x ≤ z + v

√ √



= √ √ √
2 1=4 + 2 x (z=x) 1=4+ 2 − (x=z ) 1=4+ 2
p


√ √
dz, z − v ≤ x ≤ z

z (z − v) (z=(z − v)) 1=4+ 2 − ((z − v)=z ) 1=4+ 2 2
√


p2 q2 Ws2

n  Z θv  o
5.21.5 Ex exp − 2 + ds ; Wθv ∈ dz
0
2Ws 2
√
2(2x=qz ) 1=4+p2 (x=z )1=2 S√4p2 +1=4 (qx2 =2; qz 2 =2)

0<z ≤x≤z+v

 √ dz,
2
 (z + v)2 1=4+p2 +1 S√ (q(z + v)2 =2; qz 2 =2)



2 4p +1=4
= √

 2(2x=qz ) 1=4+p2 (x=z )1=2 S√4p2 +1=4 (qz 2 =2; qx2 =2)
 (z − v)2√1=4+p2 +1 S√ dz, 0<z−v ≤x≤z

2

(qz 2 =2; q(z − v)2 =2)

4p2 +1=4

p2
θv
q

n  Z  o
5.22.5 Ex exp − 2 + ds ; Wθv ∈ dz
0
2Ws Ws
 −3√4p2 +1 √ √ √
2 (x=z ) 1=4+p2 +1=2 S√4p2 +1 (2 2qx; 2 2qz )
z ≤x≤z+v

 √ √ √ dz,
2
 2q 4p2 +1 (z + v) 4p2 +1+1 S√ (2 2q(z + v); 2 2qz )

 p

2 4p +1
= √ √ √ √

 2−3 4p2 +1 (x=z ) 1=4+p2 +1=2 S√4p2 +1 (2 2qz; 2 2qx)
 2q√4p2 +1 (z − v)√4p2 +1+1 S√ dz, z−v ≤x≤z

2 √

(2 2qz; 2 2q(z − v))
 p
4p2 +1
 5. STOPPING AT FIRST RANGE TIME 255

1 W θv = min{t : sup0≤s≤t Ws − inf 0≤s≤t Ws = v}

n  Z θv


 1
o
5.27.5 Ex exp − p1I(−∞,r) Ws + q1I[r,∞) Ws ds ; `(θv , r ∈ dy, Wθv ∈ dz
0
2

p p p p

z≤x = exp −y 2q cth((v + z − r) 2q) − y 2p cth((r − z) 2p)

√
 2√pq sh((x − r) 2q)
√ √ dydz
sh2 ((v + z − r) 2q) sh((r − z ) 2p)




√ √
2yq 2p sh((v + z − x) 2q)


× + 3 √ √ dydz, r ≤ x ≤ z + v
 sh ((v + z − r) 2q) sh((r − z ) 2p)
√ √
2yq 2p sh((x − z ) 2p)



dydz, z≤x≤r

√ √
sh2 ((v + z − r) 2q) sh2 ((r − z ) 2p)


p p p p

x≤z = exp −y 2q cth((z − r) 2q) − y 2p cth((v + r − z) 2p)

√ √
 2yp 2q sh((z − x) 2q)
√ √ dydz, r≤x≤z
sh2 ((z − r) 2q) sh2 ((v + r − z ) 2p)



√

2√pq sh((r − x) 2p)


× √ √ dydz
 sh((z − r) 2q) sh2 ((v + r − z ) 2p)
√ √
2yp 2q sh((v + x − z ) 2p)



+ dydz, z − v ≤ x ≤ r

√ 3 √
sh((z − r) 2q) sh ((v + r − z ) 2p)


Z θv θv
1
Z 


5.27.6 Px 1I(−∞,r) (Ws )ds ∈ du, 1I[r,∞) (Ws )ds ∈ dg, `(θv , r ∈ dy, Wθv ∈ dz
0 0
2

esu (1, 1, r − z, 0, y) escsg (x − r, o, v + z − r, y)


y



+ esg (2, 3, v + z − r, x − z − v, y)

2

z≤x = y

− esg (2, 3, v + z − r, v + z − x, y) dudgdydz, r ≤x≤z+v
2





y esg (2, 2, v + z − r, 0, y) escsu (x − z, 0, r − z, y)dudgdydz, z≤x≤r

y esu (2, 2, v + r − z, 0, y) escsg (z − x, 0, z − r, y)dudgdydz, r≤x≤z




 esg (1, 1, z − r, 0, y) escsu (r − x, 0, v + r − z, y)


x≤z = y
+ esu (2, 3, v + r − z, z − x − v, y)

 2
y



− esu (2, 3, v + r − z, v + x − z, y) dudgdydz, z−v ≤ x ≤ r

2
2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

2. BROWNIAN MOTION WITH DRIFT

1. Exponential stopping

(µ) 2
1.0.3 Ex eiβWτ = eiβx
2 − 2i  + 2

(µ) 2
1.0.4 Ex eiβWt = eiβ(x+µt)−β t/2

√
 2λ+µ2
1.0.5 Px (Wτ(µ) ∈ dz) = eµ(z−x)−|z−x| dz
2 + 2
p

1 −(z−µt−x)2 /2t
Px Wt(µ) ∈ dz =


1.0.6 √ e dz
2t

2 + 2 −  −γx
p
Ex exp −γ sup Ws(µ) =


1.1.1 e
+ 2 + 2 − 
p
0≤s≤τ

√ 2
Px sup Ws(µ) ≥ y = eµ(y−x)+(x−y) 2λ+µ


1.1.2
0≤s≤τ
x≤y

Ex exp −γ sup Ws(µ)

1.1.3
0≤s≤t

√  √
−  −γx+γ(γ−2µ)t/2

( − ) t  −γx

 t

= e Erfc √ − e Erfc √
− 2 2 − 2 2

|| − 
Ex exp −γ sup Ws(µ) = e−γx


(1)
0≤s<∞ + || − 

√  √ 

1 
y−x  t 1 
y−x  t
1.1.4 Px sup Ws(µ) ≥ y = Erfc √ − √ + e2µ(y−x) Erfc √ + √
0≤s≤t 2 2t 2 2 2t 2
x≤y

© Springer Basel 2015 256

A.N. Borodin, P. Salminen, Handbook of Brownian Motion - Facts and Formulae,
Probability and Its Applications, DOI 10.1007/978-3-0348-8163-0_9
 1. EXPONENTIAL STOPPING 257

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

sup Ws(µ) ≥ y = e2µ(y−x) ,

(1) Px x≤y
0≤s<∞
x≤y
µ<0

Ex exp −γ sup Ws(µ) ; Wτ(µ) ∈ dz



1.1.5
0≤s≤τ
√
2 µ(z−x)−γ(x∨z)−|z−x| 2λ+µ2
= e dz
+ 2 2 + 2
p

√
 2λ+µ2
Px sup Ws(µ) ≥ y, Wτ(µ) ∈ dz = eµ(z−x)+(z+x−2y)


1.1.6 dz
2 + 2
p
0≤s≤τ
x≤y
z≤y
2
Ex exp −γ sup Ws(µ) ; Wt(µ) ∈ dz = eµ(z−x)−µ t/2−γ(x∨z)


1.1.7
0≤s≤t

√ 

1 2 2

|z − x| t
× √ e−(z−x) /2t − eγ|z−x|/2+γ t/8 Erfc √ + √ dz
2t 4 2t 2 2

1 µ(z−x)−µ2 t/2−(|z−y|+y−x)2 /2t

Px sup Ws(µ) ≥ y, Wt(µ) ∈ dz =


1.1.8 √ e dz
0≤s≤t 2t
x≤y

 + 2 + 2 γx
p
Ex exp γ inf Ws(µ) =


1.2.1 e
+  + 2 + 2
p
0≤s≤τ

√ 2
inf Ws(µ) ≤ y = eµ(y−x)+(y−x) 2λ+µ


1.2.2 Px
0≤s≤τ
y≤x

Ex exp γ inf Ws(µ)

1.2.3
0≤s≤t

√ √
+  γx+γ(γ+2µ)t/2 
( + ) t  

 t

= e Erfc √ + eγx Erfc − √
+ 2 2 + 2 2

 + || γx
Ws(µ) =


(1) Ex exp γ inf e
0≤s<∞ +  + | |

√  √ 

1 
x−y  t 1 
x−y  t
1.2.4 Px inf Ws(µ) ≤ y = Erfc √ + √ + e2µ(y−x) Erfc √ − √
0≤s≤t 2 2t 2 2 2t 2
y≤x
258 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

Ws(µ) ≤ y = e2µ(y−x)


(1) Px inf
0≤s<∞
y≤x
µ>0

Ex exp γ inf Ws(µ) ; Wτ(µ) ∈ dz



1.2.5
0≤s≤τ
√
2 µ(z−x)+γ(x∧z)−|z−x| 2λ+µ2
= e dz
+ 2 2 + 2
p

√
 2λ+µ2
inf Ws(µ) ≤ y, Wτ(µ) ∈ dz = eµ(z−x)+(2y−z−x)


1.2.6 Px dz
2 + 2
p
0≤s≤τ
y≤x
y≤z

2
Ex exp γ inf Ws(µ) ; Wt(µ) ∈ dz = eµ(z−x)−µ t/2+γ(x∧z)


1.2.7
0≤s≤t

√ 

1 2 2

|z − x| t
× √ e−(z−x) /2t − eγ|z−x|/2+γ t/8 Erfc √ + √ dz
2t 4 2t 2 2

1 µ(z−x)−µ2 t/2−(|z−y|+x−y)2 /2t

inf Ws(µ) ≤ y, Wt(µ) ∈ dz =


1.2.8 Px √ e dz
0≤s≤t 2t
y≤x

√
2λ+µ2
1.3.1 Ex e−γ`(τ,r) = 1 − 2e
µ(r−x)−|r−x|
+ 2 + 
p

p √ 2
1.3.2 Px `(τ, r) ∈ dy = 2λ + µ2 eµ(r−x)−(y+|r−x|) 2λ+µ dy
0<y

√ 2
Px `(τ, r) = 0 = 1 − eµ(r−x)−|r−x| 2λ+µ


(1)

√ 
h
1 
|r − x| || t
1.3.3 Ex e−γ`(t,r) = 1 − eµ(r−x) e|µ||r−x| Erfc √ + √
2 − | | 2t 2

√
1 
| r − x| | | t
i
+ e−|µ||r−x| Erfc √ − √
+ | | 2t 2
2 √
| r − x| t
2
 
−µ2 )t/2
+ 2 eµ(r−x)+|r−x|γ+(γ Erfc √ + √
− 2 2t 2
 1. EXPONENTIAL STOPPING 259

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

(1) Ex e−γ`(∞,r) = 1 − eµ(r−x)−|µ||r−x|

+ | |

√
2 2
t/2−(y+|r−x|)2 /2t
eµ(r−x)−µ


1.3.4 Px `(t, r) ∈ dy = √ dy
t
0<y

√
|| µ(r−x) y + |r − x| | | t
h  
+ e e−|µ|(y+|r−x|) Erfc √ − √
2 2t 2
√
y + | r − x | | | t
 i
−e|µ|(y+|r−x|) Erfc √ + √ dy
2t 2
√ 
1 h 
|r − x| || t
Px `(t, r) = 0 = 1 − eµ(r−x) e|µ||r−x| Erfc √


(1) + √
2 2t 2
√ i
| r − x| || t

+e−|µ||r−x| Erfc √ − √
2t 2

Px `(∞, r) ∈ dy = |µ|eµ(r−x)−|µ||r−x|−|µ|y dy


(2)
0<y

Px `(∞, r) = 0 = 1 − eµ(r−x)−|µ||r−x|


(3)

√
 2λ+µ2
Ex e−γ`(τ,r) ; Wτ(µ) ∈ dz = eµ(z−x)−|z−x|

1.3.5 dz
2 + 2
p

√
 2λ+µ2
−p eµ(z−x)−(|z−r|+|r−x|) dz
2 + 2 ( + 2  + 2 )
p

√ 2
Px `(τ, r) ∈ dy, Wτ(µ) ∈ dz = λeµ(z−x)−(|z−r|+|r−x|+y) 2λ+µ dydz


1.3.6
0<y

√
 2λ+µ2
Px `(τ, r) = 0, Wτ(µ) ∈ dz = eµ(z−x)−|z−x|


(1) dz
2 + 2
p

√
 2λ+µ2
−p eµ(z−x)−(|z−r|+|r−x|) dz
2 + 2
260 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

1 −(z−µt−x)2 /2t
Ex e−γ`(t,r) ; Wt(µ) ∈ dz =

1.3.7 √ e dz
2t
√
t | z − r | + |r − x|
 
µ(z−x)+(|z−r|+|r−x|)γ+(γ 2−µ2 )t/2
− e Erfc √ + √ dz
2 2 2t

Px `(t, r) ∈ dy, Wt(µ) ∈ dz

1.3.8
0<y
1 2 2
y + |z − r| + |r − x| eµ(z−x)−µ t/2−(|z−r|+|r−x|+y) /2t dydz


= √
t 2t

Px `(t, r) = 0, Wt(µ) ∈ dz


(1)

1 −(z−µt−x)2 /2t 1 2 2
= √ e dz − √ eµ(z−x)−µ t/2−(|z−r|+|r−x|) /2t dz
2t 2t

 Z τ

1.4.1 Ex exp −γ 1I[r,∞) Ws(µ) ds
0

√
( 2  + 2 +  2 + )
p
2
x≤r =1− 2 2 e(x−r)( 2λ+µ −µ)
( + )( 2 + 2 +  + 2 +  )
p p

√
 ( 2 + 2 − )
p
(r−x)( 2λ+µ2 +2γ+µ)
r≤x = + e
+ ( + )( 2 + 2 + 2 + 2 + 2 )
p p

Z τ 
1I[r,∞) Ws(µ) ds ∈ dy


1.4.2 Px
0
√
e−y−(r−x)( 2+2−) e− y=2
 √y 

 2

x≤r = √ √ + √ Erfc − √ dy
 + 2 =2 + = 2 y 2 2
p

√  √ 

1 
x−r  y 1 
x−r  y
r≤x = λe−λy 1 − Erfc √ + √ − e2µ(r−x) Erfc √ − √ dy
2 2y 2 2 2y 2
 −(x−r+y)2 =2y √ 
e−y e  x−r  y

+p √ √ + √ e2µ(r−x) Erfc √ − √ dy
 + 2 =2 + = 2 y 2 2y 2

Z τ  √ 2
1I[r,∞) Ws(µ) ds = 0 = 1 − e(x−r)( 2λ+µ −µ)


(1) Px
x≤r 0
 1. EXPONENTIAL STOPPING 261

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

2 + 2 +  −2µ(x−r)
 p
 1− 2 + 2 − 
e , x≤r

 Z ∞

 p
1.4.3 Ex exp −γ 1I[r,∞) Ws(µ) ds = √
(r−x)( 2 +2+)
(1) 0  − 2ep

 , r≤x
µ<0 2 + 2 − 

Z t 
1I[r,∞) Ws(µ) ds ∈ dy = Bx(6) (t − y, y)1I(0,t) (y)dy


1.4.4 Px
0

for Bx(6) (u, v) see 1.6.2

Z t 
1I[r,∞) Ws(µ) ds = 0


(1) Px
x≤r 0

√ √
1 
r−x  t

1 
r−x  t

= 1 − Erfc √ − √ − e2µ(r−x) Erfc √ + √
2 2t 2 2 2t 2

Z t 
1I[r,∞) Ws(µ) ds = t


(2) Px
r≤x 0

√ √
1 
x−r  t

1 
x−r  t

=1− Erfc √ + √ − e2µ(r−x) Erfc √ − √
2 2t 2 2 2t 2

Z ∞ 
1I[r,∞) Ws(µ) ds ∈ dy


(3) Px
µ<0 0

√  √ y 
 2 2
x≤r = − √ e2(r−x)µ−µ y/2 dy − µ2 e2(r−x)µ Erfc − √ dy
y 2

√ √ 
 2 x−r  y
2

r≤x = − √ e−(x−r+µy) /2y dy − µ2 e2(r−x)µ Erfc √ − √ dy
y 2y 2

Z ∞ 
1I[r,∞) Ws(µ) ds = 0 = 1 − e(x−r)(|µ|−µ)


(4) Px
x≤r 0

n  Z τ
 o
1.4.5 Ex exp −γ 1I[r,∞) Ws(µ) ds ; Wτ(µ) ∈ dz = λeµ(z−x) Q(4)
λ+µ2 /2 (z)dz
0

for Q(4)
λ (z) see 1.1.4.5
262 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

Z τ 
1I[r,∞) Ws(µ) ds ∈ dy, Wτ(µ) ∈ dz = λeµ(z−x) Bλ+µ

(4)
1.4.6 Px 2 /2 (y, z)dydz
0

for Bλ(4) (y, z) see 1.1.4.6

Z τ 
1I[r,∞) Ws(µ) ds = 0, Wτ(µ) ∈ dz


(1) Px
x≤r 0
√ √
 2λ+µ2 2λ+µ2
eµ(z−x)−|z−x| − eµ(z−x)+(z+x−2r)


z≤r =p dz
2 + 2

n  Z t

 o 2
1.4.7 Ex exp −γ 1I[r,∞) Ws(µ) ds ; Wt(µ) ∈ dz = eµ(z−x)−µ t/2 Vx(4) (t, z)dz
0

(4)
for Vx (t, z) see 1.1.4.7

Z t  2
1I[r,∞) Ws(µ) ds ∈ dv, Wt(µ) ∈ dz = eµ(z−x)−µ t/2 Fx(4) (t, v, z)dvdz


1.4.8 Px
v<t 0

for Fx(4) (t, v, z) see 1.1.4.8

Z t 
1I[r,∞) Ws(µ) ds = 0, Wt(µ) ∈ dz


(1) Px
x≤r 0

1 −(z−µt−x)2 /2t 1 2 2
z≤r = √ e dz − √ eµ(z−x)−µ t/2−(2r−x−z) /2t dz
2t 2t

Z t 
1I[r,∞) Ws(µ) ds = t, Wt(µ) ∈ dz


(2) Px
r≤x 0

1 −(z−µt−x)2 /2t 1 2 2
r≤z = √ e dz − √ eµ(z−x)−µ t/2−(x+z−2r) /2t dz
2t 2t

 Z τ

1.5.1 Ex exp −γ 1I(−∞,r] Ws(µ) ds
0

√
 ( 2 + 2 + )
p
(x−r)( 2λ+µ2 +2γ−µ)
x≤r = + e
+ ( + )( 2 + 2 + 2 + 2 + 2 )
p p

√
( 2  + 2 +  2 − )
p
(r−x)( 2λ+µ2 +µ)
r≤x =1− e
( + )( 2 + 2 + 2 + 2 + 2 )
p p
 1. EXPONENTIAL STOPPING 263

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

Z τ 
1I(−∞,r] Ws(µ) ds ∈ dy


1.5.2 Px
0

√  √ 

1 
r−x  y 1 
r−x  y
x≤r = λe−λy 1 − Erfc √ − √ − e2µ(r−x) Erfc √ + √ dy
2 2y 2 2 2y 2
 −(r−x−y)2 =2y √ 
e−y e  r−x  y

+p √ √ − √ e2µ(r−x) Erfc √ + √ dy
 + 2 =2 − = 2 y 2 2y 2
√
e−y−(x−r)( 2+2+) e− y=2
 √y 

 2

r≤x = p 2 √ √ − √ Erfc √ dy
 +  =2 − = 2 y 2 2

Z τ  √ 2
1I(−∞,r] Ws(µ) ds = 0 = 1 − e(r−x)( 2λ+µ +µ)


(1) Px
r≤x 0
√
2e(x−r)( 2 +2−)

, x≤r

2 + 2 + 
 Z ∞ 

  p
(µ)
1.5.3 Ex exp −γ 1I(−∞,r] Ws ds =
2
 1 − p2 +  −  e2µ(r−x) , r ≤ x
p
(1) 0 

µ>0 2 + 2 + 

Z t 
1I(−∞,r] Ws(µ) ds ∈ dy = Bx(6) (y, t − y)1I(0,t) (y)dy


1.5.4 Px
0

for Bx(6) (u, v) see 1.6.2

Z t 
1I(−∞,r] Ws(µ) ds = 0


(1) Px
r≤x 0
√ √
1 
x−r  t

1 
x−r  t

=1− Erfc √ + √ − e2µ(r−x) Erfc √ − √
2 2t 2 2 2t 2

Z t 
1I(−∞,r] Ws(µ) ds = t


(2) Px
x≤r 0
√ √
1 
r−x  t

1 
r−x  t

=1− Erfc √ − √ − e2µ(r−x) Erfc √ + √
2 2t 2 2 2t 2

Z ∞ 
1I(−∞,r] Ws(µ) ds ∈ dy


(3) Px
µ>0 0
√ √ 
 2 −(r−x−µy)2 /2y r−x  y

x≤r = √ e dy − µ2 e2(r−x)µ Erfc √ + √ dy
y 2y 2
264 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

√  √ y 
 2 2
r≤x = √ e2(r−x)µ−µ y/2 dy − µ2 e2(r−x)µ Erfc √ dy
y 2

n  Z τ
 o
1.5.5 Ex exp −γ 1I(−∞,r] Ws(µ) ds ; Wτ(µ) ∈ dz = λeµ(z−x) Q(5)
λ+µ2 /2 (z)dz
0

for Q(5)
λ (z) see 1.1.5.5

Z τ 
1I(−∞,r] Ws(µ) ds ∈ dy, Wτ(µ) ∈ dz = λeµ(z−x) Bλ+µ

(5)
1.5.6 Px 2 /2 (y, z)dydz
0

for Bλ(5) (y, z) see 1.1.5.6

Z τ 
1I(−∞,r] Ws(µ) ds = 0, Wτ(µ) ∈ dz


(1) Px
r<x 0
√ √
 2λ+µ2 2λ+µ2
eµ(z−x)−|z−x| − eµ(z−x)+(2r−x−z)


r≤z =p dz
2 + 2

n  Z t
 o 2
1.5.7 Ex exp −γ 1I(−∞,r] Ws(µ) ds ; Wt(µ) ∈ dz = eµ(z−x)−µ t/2 Vx(5) (t, z)dz
0

(5)
for Vx (t, z) see 1.1.5.7

Z t  2
1I(−∞,r] Ws(µ) ds ∈ dv, Wt(µ) ∈dz = eµ(z−x)−µ t/2 Fx(5) (t, v, z)dvdz


1.5.8 Px
v<t 0

for Fx(5) (t, v, z) see 1.1.5.8

Z t 
1I(−∞,r] Ws(µ) ds = 0, Wt(µ) ∈ dz


(1) Px
r≤x 0

1 −(z−µt−x)2 /2t 1 2 2
z≤r = √ e dz − √ eµ(z−x)−µ t/2−(2r−x−z) /2t dz
2t 2t
Z t 
1I(−∞,r] Ws(µ) ds = t, Wt(µ) ∈ dz


(2) Px
x≤r 0

1 −(z−µt−x)2 /2t 1 2 2
r≤z = √ e dz − √ eµ(z−x)−µ t/2−(x+z−2r) /2t dz
2t 2t

 Z τ


p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds


1.6.1 Ex exp −
0
 1. EXPONENTIAL STOPPING 265

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

√ √


−(r−x)( 2λ+2p+µ2−µ)

2e−(r−x)( 2+2p+2−)
x≤r = 1−e +
+p ( 2 + 2q + 2 − )( 2 + 2p + 2 + )
p p

√ √


−(x−r)( 2λ+2q+µ2+µ)

2e−(x−r)( 2+2q+2+)
r≤x = 1−e + p
+q ( 2 + 2q + 2 − )( 2 + 2p + 2 + )
p

Z τ Z τ 
1I(−∞,r) Ws(µ) ds ∈ du, 1I[r,∞) Ws(µ) ds ∈ dv



1.6.2 Px
0 0

=: λe−λ(u+v) Bx(6) (u, v)dudv

 −(r−x−u)2 =2u √ 
e  r−x  u

x≤r = λe−λ(u+v) √ − √ e2µ(r−x) Erfc √ + √
u 2 2u 2

 −2 v=2  √ 
e   v
× √ + √ Erfc − √ dudv
v 2 2

 −2 u=2  √ 
e   u
r≤x = λe−λ(u+v) √ − √ Erfc √
u 2 2

 −(x−r+v)2 =2v √ 
e  x−r  v

× √ + √ e2µ(r−x) Erfc √ − √ dudv
v 2 2v 2

Z τ Z τ 
1I(−∞,r) Ws(µ) ds ∈ du, 1I[r,∞) Ws(µ) ds = 0



(1) Px
x≤r 0 0

√  √ 

1 
r−x  u 1 
r−x  u
= λe−λu 1 − Erfc √ − √ − e2µ(r−x) Erfc √ + √ du
2 2u 2 2 2u 2

Z τ Z τ 
1I(−∞,r) Ws(µ) ds = 0, 1I[r,∞) Ws(µ) ds ∈ dv



(2) Px
r≤x 0 0

√  √ 

1 
x−r  v 1 
x−r  v
= λe−λv 1 − Erfc √ + √ − e2µ(r−x) Erfc √ − √ dv
2 2v 2 2 2v 2

Z t 
p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds ∈ dv




1.6.4 Px
0
1 
|qt − v | |pt − v |


= Bx(6) , 1I((q∧p)t,(q∨p)t) (v) dv for Bx(6) u, v see 1.6.2
|p − q | |p − q | | p − q |
266 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

Z t 
p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds = pt




(1) Px
x≤r 0

√ √
1 
r−x  t

1 
r−x  t

=1− Erfc √ − √ − e2µ(r−x) Erfc √ + √
2 2t 2 2 2t 2

Z t 
p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds = qt




(2) Px
r≤x 0

√ √
1 
x−r  t

1 
x−r  t

= 1 − Erfc √ + √ − e2µ(r−x) Erfc √ − √
2 2t 2 2 2t 2

n  Z τ

 o
p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds ; Wτ(µ) ∈ dz


1.6.5 Ex exp −
0

= λeµ(z−x) Q(6)
λ+µ2 /2 (z)dz, for Q(6)
λ (z) see 1.1.6.5

Z τ Z τ 
1I(−∞,r) Ws(µ) ds ∈ du, 1I[r,∞) Ws(µ) ds ∈ dv, Wτ(µ) ∈ dz



1.6.6 Px
0 0

2
= λeµ(z−x) e−(λ+µ /2)(u+v)
Fx(4) (u + v, v, z)dudvdz

for Fx(4) (t, v, z) see 1.1.4.8

n  Z t

 o
Ex exp − p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds ; Wt(µ) ∈ dz


1.6.7
0
2
= eµ(z−x)−µ t/2
Vx(6) (t, z)dz, for Vx(6) (t, z) see 1.1.6.7

Z t 
p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds ∈ dv, Wt(µ) ∈ dz




1.6.8 Px
0

v − pt
  
(4)
F t, , z dvdz, pt < v < qt
x q−p


1


µ(z−x)−µ2 t/2 v − qt
 
= e
|p − q |  Fx(5) t,
p−q
, z dvdz, qt < v < pt


0, otherwise


for Fx(4) t, v, z see 1.1.4.8, for Fx(5) t, v, z see 1.1.5.8
 1. EXPONENTIAL STOPPING 267

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

Z t 
p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds = pt, Wt(µ) ∈ dz




(1) Px
x≤r 0

1 −(z−µt−x)2 /2t 1 2 2
z≤r = √ e dz − √ eµ(z−x)−µ t/2−(2r−x−z) /2t dz
2t 2t

Z t 
p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds = qt, Wt(µ) ∈ dz




(2) Px
r≤x 0

1 −(z−µt−x)2 /2t 1 2 2
r≤z = √ e dz − √ eµ(z−x)−µ t/2−(x+z−2r) /2t dz
2t 2t

 Z τ

1.7.1 Ex exp −γ 1I[r,u] Ws(µ) ds
0

√
Sr 2λ+µ2 −µ)
x≤r =1− e(x−r)(
+

 (1 − Sr ) sh((u − x) )e(r−x) (1 − Su ) sh((x − r) )e(u−x)

r≤x = + +
+ + sh((u − r) ) + sh((u − r) )
x≤u

√
Su 2λ+µ2 +µ)
u≤x =1− e(u−x)(
+

where

( − )( − 2 + 2 ) sh((u − r) ) + ( 2 + 2 + ) (e(u−r)λ − e(u−r) )

p p
Sr := 
2(2 + 2 + ) sh((u − r) ) + 2 2 + 2 ch((u − r) )
p

( + )( − 2 + 2 ) sh((u − r) ) + ( 2 + 2 − ) (e(u−r)λ − e(r−u) )

p p
Su :=
2(2 + 2 + ) sh((u − r) ) + 2 2 + 2 ch((u − r) )
p

( √
τ 2λ+µ2 −µ)
Z
(µ)

 1 − e(x−r)( , x≤r
1.7.2 Px 1I[r,u] Ws ds = 0 = √
(u−x)( 2λ+µ2 +µ)
(1) 0 1−e , u≤x

p
Υs := 2s + 2γ + µ2
268 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

 Z ∞


1.7.3 Ex exp −γ 1I[r,u] Ws(µ) ds
(1) 0

x≤r = 1 − e−(r−x)(|µ|−µ) + Qr e−(r−x)(|µ|−µ)

sh((u − x) 2 + 2 ) sh((x − r) 2 + 2 )
p p
r≤x = Qr e(r−x)µ 2 + Qu e(u−x)µ
sh((u − r) 2 +  ) sh((u − r) 2 + 2 )
p p
x≤u

u≤x = 1 − e−(x−u)(|µ|+µ) + Qu e−(x−u)(|µ|+µ)

where
√
(|| − ) +) 2 √
2 + 2 cth((u − r) 2 + 2 ) + || + (||sh(( +2 e(u−r)µ
p p

u−r) 2 +2 )
Qr :=
2( + 2 ) + 2|| 2 + 2 cth((u − r) 2 + 2 )
p p

√
(|| + ) 2 + 2 cth((u − r) 2 + 2 ) + || + (||−) 2 √+2 e(r−u)µ
p p

sh((u−r) 2 +2 )
Qu :=
2( + 2 ) + 2|| 2 + 2 cth((u − r) 2 + 2 )
p p

Z ∞ 
1I[r,u] Ws(µ) ds ∈ dv


1.7.4 Px
(1) 0

µ>0
Z v
(u−r)µ−µ2 v/2 2
x≤r = µe e−µ y/2
esv−y (1, 1, u − r, 0, µy)dy dv
0

Z v
2 2
r≤x = µe(u−x)µ−µ v/2
e−µ y/2
escsv−y (u − x, 0, u − r, µy)
x≤u 0

+µ essv−y (x − r, u − r, 0, µy) + escsv−y (x − r, u − r, u − r, µy) dy dv

Z v
2 2
u≤x = µe2(u−x)µ−µ v/2
e−µ y/2
µ esv−y (0, 0, u − r, 0, µy)
0


+ escsv−y (u − r, u − r, u − r, µy) dy dv

∞   1 − e−(r−x)(|µ|−µ) ,
Z x≤r
1I[r,u] Ws(µ) ds = 0 =


(2) Px
0 1 − e−(x−u)(|µ|+µ) , u≤x

t2 2 t3 
 Z t  
1.8.3 Ex exp −γ Ws(µ) ds = exp −γxt − γµ +
0
2 6
 1. EXPONENTIAL STOPPING 269

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

√
Z t 
3 
3(y − xt − t2 =2)2 
1.8.4 Px Ws(µ) ds ∈ dy = √ exp − dy
0 2t3 2t3

 Z τ ez Ai 21=3 −2=3 ( + 2 =2 + |z |)

n  o

1.8.5 E0 exp −γ |Ws(µ) |ds ; Wτ(µ) ∈ dz =− dz
(2 )1=3 Ai0 21=3 −2=3 ( + 2 =2)


0

n  Z t  o
1.8.7 Ex exp −γ Ws(µ) ds ; Wt(µ) ∈ dz
0

1 
(z − t − x)2 (z + x)t 2 t3 
= √ exp − − + dz
2t 2t 2 24

n  Z t  o
(1) E0 exp −γ |Ws(µ) |ds ; Wt(µ) ∈ dz
0

2
∞
X
Ai 0k + 21=3 1=3 |z |

= −eµz−µ t/2
γ 1/3 exp 2−1/3 γ 2/3 tαk0 dz
22=3 0k Ai( 0k )
k=1

Z t 
1.8.8 Ex Ws(µ) ds ∈ dy, Wt(µ) ∈ dz
0

√
3 
(z − t − x)2 3(2y − (z + x)t)2 
= exp − − dydz
t2 2t 2t3

∞ √  ( 0 )3 t3   ( 0 )2 t2 
Z t
(µ) (µ)
 X
(9=y)1=3 t k k
(1) P0 |Ws |ds < y Wt =0 = exp Ai √
0
1= 6
2 | k|0 27y2 (3y 2)4=3
k=1

 2Z t  2
e−x  t (x2 2 − 2 ) sh(t ) − 2x
 
1.9.3 Ex exp − (Ws(µ) )2 ds = p exp − −
2 0 ch(t ) 2 2 ch(t )

n  2Z τ  o
1.9.5 Ex exp − (Ws(µ) )2 ds ; Wτ(µ) ∈ dz
2 0

 1 + 2+2

2 2 p p
x≤z = √ eµ(z−x) D− 1 − 2λ+µ2 (−x 2γ)D− 1 − 2λ+µ2 (z 2γ)dz
 2 2γ 2 2γ

 1 + 2+2

2 2 p p
z≤x = √ eµ(z−x) D− 1 − 2λ+µ2 (x 2γ)D− 1 − 2λ+µ2 (−z 2γ)dz
 2 2γ 2 2γ
270 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

n  2Z t  o
1.9.7 Ex exp − (Ws(µ) )2 ds ; Wt(µ) ∈ dz
2 0
2 t (x2 + z 2 ) ch(t ) − 2xz
√  
= exp µ(z − x) − − dz
2 sh(t ) 2 2 sh(t )
p

Z t 
1.9.8 Px (Ws(µ) )2 ds ∈ dy, Wt(µ) ∈ dz
0

2
= (2π)−1/2 eµ(z−x)−µ t/2
eey (1/2, t, (x2 + z 2 )/2, −xz)dydz

(2 2 y)=2 (=2 + 2 + 2 =2| |)

Z τ  p
2βWs(µ)
1.10.2 Px e ds ∈ dy = √
| | 2y (1 + 2 + 2 =| |)
p
0

e2 x
 2 x
e
 
β 6= 0 × exp −(β + µ)x − 2 M1/2−µ/2β,√2λ+µ2 /2|β| dy
4 y 2 2y

√ √
 Z ∞
(µ)

2( 2 =2| |)||=| | −µx 2 βx 
1.10.3 Ex exp −γ e2βWs ds = e K|µ|/|β| e
 0
(||=| |) | |
<0

Z t 
(µ)
1.10.4 Px e2βWs ds ∈ dy µ/2β > −1
0

√
2| |(2 2 y)=2 
2 t e2 x
 
 1 e2 x 
β 6= 0 = √ exp −(β + µ)x − − 2 m2β 2 t − , 2 dy
y 2 4 y 2 2 4 y

Z ∞
(µ)

(2 2 y)−||=| | 
e2 x

(1) Px e2βWs ds ∈ dy = exp −2µx − 2 dy
 0
y (||=| |) 2 y
<0

n  Z τ  o
(µ)
1.10.5 Ex exp −γ e2βWs ds ; Wτ(µ) ∈ dz
0

√ √
2 2 βx  √ 2 βz 

 I√2λ+µ2 /β
 e K 2λ+µ2 /β e dz, x ≤ z
0<β =eµ(z−x) √ √
 2 K√ 2 βx √ 2 βz
 
e I 2λ+µ2 /β e dz, z ≤ x

2λ+µ2 /β
 1. EXPONENTIAL STOPPING 271

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

√ √
2 2 2
  
 | | K√2λ+µ2 /|β| | | eβx I√2λ+µ2 /|β| | | eβz dz, x ≤ z

β<0 =eµ(z−x) √ √
 2 I√

2
2 βx  √
e K 2
2 βz 
e dz, z ≤ x
2λ+µ /|β|
| | 2λ+µ /|β| | | | |

Z τ 
(µ)
1.10.6 Px e2βWs ds ∈ dy, Wτ(µ) ∈ dz
β 6= 0 0

e2 x + e2 z √
 (x+z) 
 e
 
= exp µ(z − x) − 2 I 2λ+µ2 /|β| 2y dydz
| |y 2 y

n  Z t  o
(µ)
1.10.7 Ex exp −γ e2βWs ds ; Wt(µ) ∈ dz
0

√ √
2 2 2 
β 6= 0 = |β| eµ(z−x)−µ t/2
kiβ 2 t/2 eβx , eβz dz
| | | |

Z t 
(µ)
1.10.8 Px e2βWs ds ∈ dy, Wt(µ) ∈ dz
0

2 t e2 x + e2 z
 (x+z) 
| | e
 
β 6= 0 = exp µ(z − x) − − 2 iβ 2 t/2 2y dydz
2y 2 2 y

h2 (2 + 2 )
1.11.2 Px sup `(τ, y) ≥ h = 2
8 sh h 2 + 2 =2
p

y∈R

3+ √ 2
M 32 − √  2 ; 3; h 2 + 2 
2 2 2+2 ; 3; h 2 + 
M p p

2 2+
× 1+ √ 2
+ M 1 − √  2

M 2 2 2+2 ; 1; h 2 +  2 2 2+2 ; 1; h 2 + 
p p

2 + 2 − 
p
1.12.1 Ex e−γ Ȟµ (τ ) =
2 + 2 + 2 − 
p

2 + 2 + 
p
(1) Ex e−γ Ĥµ (τ ) =
2 + 2 + 2 + 
p

 −2 v=2  √ 
p e   v
Px Ȟµ (τ ) ∈ dv = ( 2λ + µ2 − µ)e−λv √


1.12.2 + Erfc − √ dv
2v 2 2

 −2 v=2  √ 
p e   v
Px Ĥµ (τ ) ∈ dv = ( 2λ + µ2 + µ)e−λv √


(1) − Erfc √ dv
2v 2 2
272 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

 √   −2 t=2  √ 
e− t− t=2 e− t  t e   t
 2

1.12.3 Ex e−γ Ȟµ (t) = √ + √ Erfc − √ ∗ √ − √ Erfc √

t 2 2 t 2 2

 √   −2 t=2  √ 
e− t− t=2 e− t  t e   t
 2

(1) Ex e−γ Ĥµ (t) = √ − √ Erfc √ ∗ √ + √ Erfc − √

t 2 2 t 2 2

| | − 
(2) Ex e−γ Ȟµ (∞) =
2 + 2 − 
p

| | + 
(3) Ex e−γ Ĥµ (∞) =
2 + 2 + 
p

1.12.4 Px Ȟµ (t) ∈ dv 0≤v≤t

 √  −2 (t−v)=2  √

e− v=2   v e   t−v
 2 
= √ + √ Erfc − √ − √ Erfc √ dv
v 2 2 (t − v) 2 2
p

(1) Px Ĥµ (t) ∈ dv 0≤v≤t

 √  −2 (t−v)=2  √

e− v=2   v e   t−v
 2 
= √ − √ Erfc √ + √ Erfc − √ dv
v 2 2 (t − v) 2 2
p

 −2 v=2  √ 

e   v
(2) Px Ȟµ (∞) ∈ dv = (|µ| − µ) √ + Erfc − √ dv
2v 2 2

 −2 v=2  √ 

e   v
(3) Px Ĥµ (∞) ∈ dv = (|µ| + µ) √ − Erfc √ dv
2v 2 2
√
2e(x−z)( 2+2+2 −)

 dz x ≤ z
2 + 2 + 2 + 2 + 2

 p p
Ex e−γ Ȟµ (τ ) ; Wτ(µ)

1.12.5 ∈ dz = √
(z−x)( 2+2 +)
 p 2e

dz z ≤ x

2 + 2 + 2 + 2 + 2
p

√
2e(x−z)( 2+2 −)

 dz x ≤ z
2 + 2 + 2 + 2 + 2

 p p
Ex e−γ Ĥµ (τ ) ; Wτ(µ)

(1) ∈ dz = √
(z−x)( 2+2+2 +)
 p2e

dz z ≤ x

2 + 2 + 2 + 2 + 2
p
 1. EXPONENTIAL STOPPING 273

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

Ex e−γ Ȟµ (t) ; Wt(µ) ∈ dz


1.12.7

1 − e− t   z − x −γt−(z−x)2 /2t 
2

x≤z = eµ(z−x)−µ t/2
∗ √ 3=2 e
√ dz
2 t3=2 2t
2

1 − e− t   x − z −(x−z)2 /2t 
z≤x = eµ(z−x)−µ t/2 √ ∗ √ 3=2 e dz
2 t3=2 2t

Ex e−γ Ĥµ (t) ; Wt(µ) ∈ dz


(1)

1 − e− t   z − x −(z−x)2 /2t 
2

x≤z = eµ(z−x)−µ t/2
∗ √ 3=2 e
√ dz
2 t3=2 2t
2

1 − e− t   x − z −γt−(x−z)2 /2t 
z≤x = eµ(z−x)−µ t/2 √ ∗ √ 3=2 e dz
2 t3=2 2t

Px Ȟµ (t) ∈ dv, Wt(µ) ∈ dz

1.12.8
v<t

∞
(y − x)(y − z ) (y − x)2 (y − z )2 
Z 
2
= eµ(z−x)−µ t/2
3= 2 exp − − dy dvdz
x∨z (v(t − v)) 2v 2(t − v)

Px Ĥµ (t) ∈ dv, Wt(µ) ∈ dz

(1)
v<t
x∧z
(x − y)(z − y) (x − y)2 (z − y)2 
Z 
2
= eµ(z−x)−µ t/2
exp − − dy dvdz
−∞ (v(t − v))3=2 2v 2(t − v)

√
2 −(y−x)( 2λ+2γ+µ2 −µ)
Ex e−γ Ȟµ (τ ) ; sup Ws(µ) ∈ dy =

1.13.1 e dy
 + 2 + 2
p
0≤s≤τ
x<y

Px Ȟµ (τ ) ∈ dv, sup Ws(µ) ∈ dy

1.13.2
0≤s≤τ
x<y
2(y − x) 2
(y − x)2 

= √ exp µ(y − x) − λ + v − dvdy
2( 2 + 2 + )v3=2 2 2v
p

Ex e−γ Ȟµ (t) ; sup Ws(µ) ∈ dy


1.13.3
0≤s≤t
x<y
274 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

 √ 
y − x µ(y−x)−γt−µ2 t/2 (y − x)2   e− t=2   t
  2

= √ e exp − ∗ √ − √ Erfc √ dy
t3=2 2t t 2 2

√ 2
Ex e−γ Ȟµ (∞) ; sup Ws(µ) ∈ dy = (|µ| − µ)e−(y−x)( 2γ+µ −µ) dy

(1)
0≤s<∞
x<y

Px Ȟµ (t) ∈ dv, sup Ws(µ) ∈ dy

1.13.4 0≤v≤t
0≤s≤t
x<y
 √
y−x (y − x − v)2  e− (t−v)=2   t−v
 2 
= √ exp − − √ Erfc √ dvdy
v3=2 2v (t − v) 2 2
p

(|| − )(y − x) 
(y − x − v)2 
(1) Px Ȟµ (∞) ∈ dv, sup Ws(µ) ∈ dy = √ 3=2 exp − dvdy
0≤s<∞ 2v 2v
x<y

Ex e−γ Ȟµ (τ ) ; sup Ws(µ) ∈ dy, Wτ(µ) ∈ dz


1.13.5 x∨z <y
0≤s≤τ

√ √
2λ+2γ+µ2 (z−y) 2λ+µ2
= 2λeµ(z−x) e(x−y) e dydz

Px Ȟµ (τ ) ∈ dv, sup Ws(µ) ∈ dy, Wτ(µ) ∈ dz

1.13.6 x∨z <y
0≤s≤τ

2(y − x) 
(y − x)2 2
p 
2 dvdydz
= √ exp µ(z − x) − − λ + v − (y − z) 2λ + µ
2v3=2 2v 2

Ex e−γ Ȟµ (t) ; sup Ws(µ) ∈ dy, Wt(µ) ∈ dz


1.13.7 x∨z <y
0≤s≤t


y−x

(y − x − t)2   y − z 
(y − z + t)2 
= 3= 2 exp −γt − ∗ 3=2 exp − dydz
t 2t t 2t

Px Ȟµ (t) ∈ dv, sup Ws(µ) ∈ dy, Wt(µ) ∈ dz

1.13.8 x∨z <y
0≤s≤t
v<t

(y − x)(y − z ) 
(y − x − v)2 (y − z + (t − v))2 
= 3= 2 exp − − dy dvdz
(v(t − v)) 2v 2(t − v)
 1. EXPONENTIAL STOPPING 275

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

√
2 −(x−y)( 2λ+2γ+µ2 +µ)
Ex e−γ Ĥµ (τ ) ; inf Ws(µ) ∈ dy =

1.14.1 e dy
2 + 2 − 
p
0≤s≤τ
y<x

Px Ĥµ (τ ) ∈ dv, inf Ws(µ) ∈ dy

1.14.2
0≤s≤τ
y<x
2(x − y) 2
(x − y)2 

= √ exp µ(y − x) − λ + v − dvdy
2( 2 + 2 − )v3=2 2 2v
p

Ex e−γ Ĥµ (t) ; inf Ws(µ) ∈ dy


1.14.3
0≤s≤t
y<x
 √ 
x − y µ(y−x)−γt−µ2 t/2 (x − y)2   e− t=2   t
  2

= √ e exp − ∗ √ + √ Erfc −√ dy
t3=2 2t t 2 2

√
2γ+µ2 +µ)
Ex e−γ Ĥµ (∞) ; inf Ws(µ) ∈ dy = (|µ| + µ)e−(x−y)(

(1) dy
0≤s<∞
y<x

Px Ĥµ (t) ∈ dv, inf Ws(µ) ∈ dy

1.14.4 0≤v≤t
0≤s≤t
y<x
 √
x−y (x − y + v)2  e− (t−v)=2   t−v
 2 
= √ exp − + √ Erfc − √ dvdy
v3=2 2v (t − v) 2 2
p

(|| + )(x − y) 
(x − y + v)2 
(1) Px Ĥµ (∞) ∈ dv, inf Ws(µ) ∈ dy = √ 3=2 exp − dvdy
0≤s<∞ 2v 2v
y<x

Ex e−γ Ĥµ (τ ) ; inf Ws(µ) ∈ dy, Wτ(µ) ∈ dz


1.14.5 y <x∧z
0≤s≤τ

√ √
2λ+2γ+µ2 (y−z) 2λ+µ2
= 2λeµ(z−x) e(y−x) e dydz

Px Ĥµ (τ ) ∈ dv, inf Ws(µ) ∈ dy, Wτ(µ) ∈ dz

1.14.6 y <x∧z
0≤s≤τ

2(x − y) 
(x − y)2 2
p 
= √
3= 2 exp µ(z − x) − − λ+ v − (z − y) 2λ + µ2 dvdydz
2v 2v 2
276 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

Ex e−γ Ĥµ (t) ; inf Ws(µ) ∈ dy, Wt(µ) ∈ dz


1.14.7 y <x∧z
0≤s≤t


x−y

(x − y + t)2   z − y 
(z − y − t)2 
= exp −γt − ∗ exp − dydz
t3=2 2t t3=2 2t

Px Ĥµ (t) ∈ dv, inf Ws(µ) ∈ dy, Wt(µ) ∈ dz

1.14.8 y <x∧z
0≤s≤t
v<t
(x − y)(z − y) 
(x − y + v)2 (z − y − (t − v))2 
= exp − − dy dvdz
(v(t − v))3=2 2v 2(t − v)

Px a < inf Ws(µ) ; sup Ws(µ) < b

1.15.2
0≤s≤τ 0≤s≤τ

e(a−x) sh (b − x) 2 + 2 + e(b−x) sh (x − a) 2 + 2
p
p

=1−
sh (b − a) 2 + 2
p

Px sup Ws(µ) − inf Ws(µ) < y

(1)
0≤s≤τ 0≤s≤τ

2 2 + 2  sh2 (y( 2 + 2 + )=2) sh2 (y( 2 + 2 − )=2) 

p p p
=1− 2 p +
sh (y 2 + 2 ) 2 + 2 +  2 + 2 − 
p p

Px a < inf Ws(µ) ; sup Ws(µ) < b

1.15.4
0≤s≤t 0≤s≤t

∞ Z b
e− t=2 X
2
2 2
eµ(z−x) e−(z−x+2k(b−a)) /2t − e−(z+x−2a+2k(b−a)) /2t dz


= √
2t a k=−∞

Px sup Ws(µ) − inf Ws(µ) < y

(1)
0≤s≤t 0≤s≤t

∞ √ √
X h
1
  (2k + 1)y  t  
(2k − 1)y  t 
= e−2kµy − kµy Erf √ − √ − Erf √ − √
2 2t 2 2t 2
k=−∞

n  (2k + 1)y √t 

2
 √
(2k − 1)y  t 
+k 2 − 2kµy + µ t Erf √ − √ + Erf √ − √
2t 2 2t 2
 1. EXPONENTIAL STOPPING 277

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

√ √ √

2ky  t o k 2t n  
2ky  t 2 
−2 Erf √ − √ + √ 2 exp − √ − √
2t 2  2t 2
√ √
 
(2k + 1)y  t 2   
(2k − 1)y  t 2 oi
− exp − √ − √ − exp − √ − √
2t 2 2t 2

Px a < inf Ws(µ) ; sup Ws(µ) < b, Wτ(µ) ∈ dz

1.15.6
0≤s≤τ 0≤s≤τ

 ch (b − a − |z − x|) 2 + 2 − ch (b + a − z − x) 2 + 2
p
p


= dz, a < x ∧ z,
e(x−z) 2 + 2 sh (b − a) 2 + 2
p p

x∨z <b

Px sup Ws(µ) − inf Ws(µ) < y, Wτ(µ) ∈ dz

(1) for |z − x| < y
0≤s≤τ 0≤s≤τ

e(z−x)
ch(y 2 + 2

 p
p
= sh (y − |z − x|) 2λ + µ2
sh2 (y 2 + 2 ) 2 + 2
p p

p

−(y − |z − x|) ch |z − x| 2λ + µ2 dz

Px a < inf Ws(µ) ; sup Ws(µ) < b , Wt(µ) ∈ dz

1.15.8
0≤s≤t 0≤s≤t

for a < x ∧ z ≤ x ∨ z < b

∞
1 µ(z−x)−µ2 t/2 X −(z−x+2k(b−a))2 /2t 2
− e−(z+x−2a+2k(b−a)) /2t dz


= √ e e
2t
k=−∞

Px sup Ws(µ) − inf Ws(µ) < y, Wt(µ) ∈ dz

(1) for |z − x| < y
0≤s≤t 0≤s≤t

∞ 
e(z−x)− t=2 X 2k(y − |z − x|)(|z − x| + 2ky) −(|z−x|+2ky)2/2t
2

= √ 2k + 1 − e dz
2t t
k=−∞

√ 2
Ex e−γ`(τ,r) ; sup Ws(µ) < b = 1 − e(b−x)(µ− 2λ+µ )

1.16.1
0≤s≤τ
r<b
√
(e(r−x) − e(b−x)−(b−r) 2+2 ) −|x−r|√2λ+µ2 √
−(2b−r−x) 2λ+µ2


−p √ e − e
2 + 2 + (1 − e− 2(b−r ) 2+ 2
)
278 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

y 2  + 2
 p 
Px `(τ, r) ∈ dy, sup Ws(µ) < b = exp −


1.16.2 √
r<b 0≤s≤τ 1 − e−2(b−r) 2+2
0<y

√ √ √
2 + 2 (e(r−x) − e(b−x)−(b−r) 2+2 )(e−|x−r| 2+2 − e−(2b−r−x) 2+2 )
p
× √ dy
1 − e−2(b−r) 2+2 2

√ 2
Px `(τ, r) = 0, sup Ws(µ) < b = 1 − e(b−x)(µ− 2λ+µ )


(1)
0≤s≤τ

√ √ √
(e(r−x) − e(b−x)−(b−r) 2+2 )(e−|x−r| 2+2 − e−(2b−r−x) 2+2 )
− √
1 − e−2(b−r) 2+2
Ex e−γ`(∞,r) ; sup Ws(µ) < b

1.16.3 r∨x≤b
0≤s<∞
(1)
µ<0
(e(|x−r|+r−x) − e2(b−x) )
= 1 − eµ(|x−r|+r−x) −
(1 − e2(b−r) ) − 

Px `(∞, r) ∈ dy, sup Ws(µ) < b

1.16.4 r∨x≤b
0≤s<∞
(1)
µ<0
||(e(|x−r|+r−x) − e2(b−x) ) y | |
 
= exp − dy
1 − e2(b−r) 1−e 2 (b−r )

Px `(∞, r) = 0, sup Ws(µ) < b = 1 − eµ(|r−x|+r−x)

(2)
0≤s<∞

Ex e−γ`(τ,r) ; sup Ws(µ) < b, Wτ(µ) ∈ dz


1.16.5
0≤s≤τ

√ √

n
µ(z−x) −|z−x| 2λ+µ2 −(2b−z−x) 2λ+µ2
= e e − e
2 + 2
p

√ √ √ √
e−|x−r| 2+2 − e−(2b−x−r) 2+2 e−|z−r| 2+2 − e−(2b−z−r) 2+2 o



− √ dz
2 + 2 + 1 − e−2(b−r) 2+2
p

Px `(τ, r) ∈ dy, sup Ws(µ) < b, Wτ(µ) ∈ dz

1.16.6
0≤s≤τ
 1. EXPONENTIAL STOPPING 279

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

y 2 + 2
 p 
= λ exp µ(z − x) − √
1 − e−2(b−r) 2+2
√ √ √ √
(e−|x−r| 2+2 − e−(2b−r−x) 2+2 )(e−|z−r| 2+2 − e−(2b−z−r) 2+2 )
× √ dydz
1 − e−2(b−r) 2+2 2


1 
1.16.8 Px `(t, r) ∈ dy, sup Ws(µ) < b, Wt(µ) ∈ dz
2 0≤s≤t

2
= eµ(z−x)−µ t/2
Fx(16) (t, y, z)dydz, for Fx(16) (t, y, z) see 1.1.16.8

Px `(t, r) = 0, sup Ws(µ) < b, Wt(µ) ∈ dz

(1) r < x ∧ z, x ∨ z < b
0≤s≤t
∞
2 1 X −(z−x+2k(b−r))2 /2t 2
= eµ(z−x)−µ t/2
− e−(z+x−2r+2k(b−r)) /2t dz


√ e
2t
k=−∞

√ 2
Ex e−γ`(τ,r) , a < inf Ws(µ) = 1 − e(a−x)(µ+ 2λ+µ )

1.17.1
0≤s≤τ
a<r
√
(e(r−x) − e(a−x)−(r−a) 2+2 ) −|x−r|√2λ+µ2 √
−(x+r−2a) 2λ+µ2


−p √ e − e
2
2 +  + (1 − e − 2(r−a ) 2+2
)

y 2 + 2
 p 
Px `(τ, r) ∈ dy, a < inf Ws(µ) = exp −


1.17.2 √
a<r
0≤s≤τ 1 − e−2(r−a) 2+2

√ √ √
2 + 2 (e(r−x) − e(a−x)−(r−a) 2+2 )(e−|x−r| 2+2 − e−(x+r−2a) 2+2 )
p
× √ dy
1 − e−2(r−a) 2+2 2

√ 2
Px `(τ, r) = 0, a < inf Ws(µ) = 1 − e(a−x)(µ+ 2λ+µ )


(1)
0≤s≤τ
a<r
√
(e(r−x) − e(a−x)−(r−a) 2+2 )
− √ √ √
1 − e−2(r−a) 2+2 e−|x−r| 2+2 − e−(x+r−2a) 2+2

Ex e−γ`(∞,r) ; a < Ws(µ)


1.17.3 inf a≤r∧x
0≤s<∞
(1)
(e−(|x−r|+x−r) − e−2(x−a) )
µ>0 = 1 − e−µ(|x−r|+x−r) +
(1 − e−2(r−a) ) + 
280 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

Ws(µ)


1.17.4 Px `(∞, r) ∈ dy, a < inf a≤r∧x
0≤s<∞
(1)
(e−(|x−r|+x−r) − e−2(x−a) ) y
 
µ>0 = exp − dy
1 − e−2(r−a) 1 − e−2(r−a)

Ws(µ) = 1 − e−µ(|x−r|+x−r)


(2) Px `(∞, r) = 0, a < inf
0≤s<∞
a<r

Ex e−γ`(τ,r) ; a < inf Ws(µ) , Wτ(µ) ∈ dz


1.17.5
0≤s≤τ

√ √

n 2 2
= 2 eµ(z−x) e−|z−x| 2λ+µ − e−(z+x−2a) 2λ+µ
2 + 
p

√ √ √ √
e−|x−r| 2+2 − e−(x+r−2a) 2+2 e−|z−r| 2+2 − e−(z+r−2a) 2+2 o



− √ dz
2 + 2 + 1 − e−2(r−a) 2+2
p

Px `(τ, r) ∈ dy, a < inf Ws(µ) , Wτ(µ) ∈ dz

1.17.6
0≤s≤τ

y 2 + 2
 p 
= λ exp µ(z − x) − √
1 − e−2(r−a) 2+2
√ √ √ √
(e−|x−r| 2+2 − e−(x+r−2a) 2+2 )(e−|z−r| 2+2 − e−(z+r−2a) 2+2 )
× √ dydz
1 − e−2(r−a) 2+2 2


1 
1.17.8 Px `(t, r) ∈ dy, a < inf Ws(µ) , Wt(µ) ∈ dz
2 0≤s≤t

2
= eµ(z−x)−µ t/2 (17)
Fa,x (t, y, z)dydz, (17)
for Fa,x (t, y, z) see 1.1.17.8

Px `(t, r) = 0, a < inf Ws(µ) , Wt(µ) ∈ dz

(1) a < x ∧ z, x ∨ z < r
0≤s≤t
∞
2 1 X −(z−x+2k(r−a))2 /2t 2
= eµ(z−x)−µ t/2
− e−(z+x−2a+2k(r−a)) /2t dz


√ e
2t
k=−∞

e(r−x) (0 + ) − e(u−x)−|u−r|0 e−|r−x|0

1.18.1 Ex e−γ`(τ,r)−η`(τ,u) = 1 −
(0 + )(0 + ) − e−2|u−r|0
 1. EXPONENTIAL STOPPING 281

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

 e(u−x) (0 + ) − e(r−x)−|u−r|0 e−|u−x|0

−
(0 + )(0 + ) − e−2|u−r|0

e(r−x) (|| + ) − e(u−x)−|u−r||| e−|r−x|||

1.18.3 Ex e−γ`(∞,r)−η`(∞,u) = 1 −
(|| + )(|| + ) − e−2|u−r|||
(1)
 e(u−x) (|| + ) − e(r−x)−|u−r||| e−|u−x|||


−
(|| + )(|| + ) − e−2|u−r|||

1.18.4 Px `(∞, r) ∈ dy, `(∞, u) ∈ dg
(1)

for x ≤ r < u, µ > 0 or u < r ≤ x, µ < 0

2 e|u−r|||
  ||√yg 
||(y + g )

= exp − I dydg
2 sh(|u − r|||) 1 − e−2|u−r||| 0 sh(|u − r|||)

for r < u ≤ x, µ > 0 or x ≤ u < r, µ < 0

2 e−2|u−x||| g
√   ||√yg 
||(y + g )

= √ exp − I dydg
2 sh(|u − r|||) y 1 − e−2|u−r||| 1 sh(|u − r|||)

for r < x < u, µ > 0 or u < x < r, µ < 0

2 e|u−x|||
 ||√yg 
||(y + g )
 h
= exp − 2 u r  sh(|u − x||µ|)I0
2 sh(|u − r|||) 1−e − | − || | sh(|u − r|||)

√
|| yg
p  i
+ sh(|x − r||µ|) g/yI1 dydg
sh(|u − r|||)


(2) Px `(∞, r) = 0, `(∞, u) ∈ dg

for r < x < u, µ > 0 or u < x < r, µ < 0

sh(|x − r|||) 
g | |

= |µ|e|u−x||µ| exp − dg
sh(|u − r|||) 1 − e−2|||u−r|

for r < u ≤ x, µ > 0 or x ≤ u < r, µ < 0

g||
 
= |µ|e−2|µ||u−x| exp − −2|||u−r |
dg
1−e
282 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

for r < u ≤ x, µ > 0 or x ≤ u < r, µ < 0

Px `(∞, r) = 0, `(∞, u) = 0 = 1 − e−2|µ||u−x|

(3)

Ex e−γ`(τ,r)−η`(τ,u) Wτ(µ) = z

1.18.5

(0 + )e−|r−x|0 − e−(|u−x|+|u−r|)0 e−(|z−r|−|z−x|)0

=1−
(0 + )(0 + ) − e−2|u−r|0
 (0 + )e−|u−x|0 − e−(|r−x|+|u−r|)0 e−(|z−u|−|z−x|)0


−
(0 + )(0 + ) − e−2|u−r|0

Ex e−γ`(τ,r) `(τ, u) = g, Wτ(µ) = z


(1)

for r < u ≤ x ∧ z or x ∨ z ≤ u < r

g 0 e−2|r−u|0
 
= exp − 2 r u 
,
0 + 1 − e − | − | 0

for x ∧ z ≤ r < u ≤ x ∨ z or x ∧ z ≤ u < r ≤ x ∨ z

0 g 0 e−2|r−u|0
 
= 2 r u  exp − 2 r u 
0 + (1 − e− | − | 0) 0 + 1 − e − | − |


0

Px `(τ, r) ∈ dy `(τ, u) = g, Wτ(µ) = z

1.18.6

for r < u ≤ x ∧ z or x ∨ z ≤ u < r

√
0 g   ye|r−u|0 + ge−|r−u|0
   √gy 
0 0
= √ exp − I1 dy
2 y sh(|r − u|0 ) 2 sh(|r − u|0 ) sh(|r − u|0 )
for x ∧ z ≤ r < u ≤ x ∨ z or x ∧ z ≤ u < r ≤ x ∨ z

  ye|r−u|0 + ge−|r−u|0
   √gy 

0 e|r−u|0 0 0
= exp − I0 dy
2 sh(|r − u|0 ) 2 sh(|r − u|0 ) sh(|r − u|0 )

g
Px `(τ, r) = 0 `(τ, u) = g, Wτ(µ) = z = exp − 2|r−u|00

 
(1)
e −1

for r < u ≤ x ∧ z or x ∨ z ≤ u < r

p
Υ0 := 2λ + µ2
 1. EXPONENTIAL STOPPING 283

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

2
Px `(t, r) ∈ dy, `(t, u) ∈ dg, Wt(µ) ∈ dz = eµ(z−x)−µ t/2 Fx,t

(18)
1.18.8 (y, g, z)dydgdz

(18)
for Fx,t (y, g, z) see 1.1.18.8

Z τ
ds

(2 + 2 )
1.19.2 Px ()
∈ dy, 0 < inf Ws(µ) = 2 p
0 Ws 0≤s≤τ sh (y 2 + 2 =2)
0<x

x

2x 
× exp − dy
( 2 + 2 cth(y 2 + 2 =2) − )2 2 + 2 cth(y 2 + 2 =2) − 
p p p p

τ
ds
n  Z  o
(µ) (µ)
1.19.5 Ex exp −γ ; 0 < inf W , W ∈ dz
0 Ws() 0≤s≤τ
s τ
0<x

= λeµ(z−x) Q(19)
λ+µ2 /2 (z)dz, for Q(19)
λ (z) see 1.1.19.5

 (2 + 2 )xz
τ
ds
Z  p
(µ) (µ)
1.19.6 Px ∈ dy, 0 < inf W , W ∈ dz =
Ws() s τ
sh(y 2 + 2 =2)
p
0 0≤s≤τ
0<x

(x + z ) 2 + 2 ch(y 2 + 2 =2)   2 (2 + 2 )xz 

 p p p
× exp µ(z − x) − I1 dydz
sh(y 2 + 2 =2) sh(y 2 + 2 =2)
p p

t
ds
Z 
1.19.8 Px ∈ dy, 0 < inf Ws(µ) , Wt(µ) ∈ dz
0 Ws() 0≤s≤t
0<x
2 √ √
= eµ(z−x)−µ t/2
xz ist (1, y/2, 0, x + z, xz)dydz

 2Z τ
ds
n  o
(µ) (µ)
1.20.5 Ex exp − 2 ; 0 < inf W , W ∈ dz
2 Ws() s τ


0 0≤s≤τ
0<x
√
= 2λeµ(z−x) xzI√γ 2 +1/4 x 2λ + µ2 K√γ 2 +1/4 z 2λ + µ2 dz
p
p

x≤z
√
= 2λeµ(z−x) xzK√γ 2 +1/4 x 2λ + µ2 I√γ 2 +1/4 z 2λ + µ2 dz
p
p

z≤x

τ
ds
Z 
(µ) (µ)
1.20.6 Px
∈ dy, 0 < inf Ws , Wτ ∈ dz
0 Ws() 2 0≤s≤τ
0<x
√ p p
= λ xzeµ(z−x)−y/8 kiy/2 (x 2λ + µ2 , z 2λ + µ2 )dydz
284 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

 2Z t
ds
n  o
1.20.7 Ex exp − ()
2
; 0 < inf Ws(µ) , Wt(µ) ∈ dz
2 0 Ws 0≤s≤t
0<x

√
xz µ(z−x)−µ2 t/2 −(x2 +z2 )/2t √ xz
 
= e e I γ 2 +1/4 dz
t t

t
ds
Z 
(µ) (µ)
1.20.8 Px
∈ dy, 0 < inf Ws , Wt ∈ dz
0 Ws() 2 0≤s≤t
0<x

√
xz µ(z−x)−µ2 t/2−(x2 +z2 )/2t−y/8 xz
 
= e iy/2 dydz
t t

p2 q2 (Ws() )2
n  Z τ   o
(µ) (µ)
1.21.5 Ex exp − + ds ; 0 < inf W , W ∈ dz
2 Ws() 2 2 s τ


0 0≤s≤τ
0<x
= λeµ(z−x) Q(21)
λ+µ2 /2 (z)dz, for Q(21)
λ (z) see 1.1.21.5

p2 q2 (Ws() )2
n  Z t   o
(µ) (µ)
1.21.7 Ex exp − () 2
+ ds ; 0 < inf W , W ∈ dz
0 2 Ws


2 0≤s≤t
s t
0<x
√
q xz 2 t (x2 + z 2 )q ch(tq) √ xzq
   
= exp µ(z − x) − − I p2 +1/4 dz
sh(tq) 2 2 sh(tq) sh(tq)

t t
ds
Z Z 
(µ) 2 (µ) (µ)
1.21.8 Px (Ws ) ds ∈ dg,
∈ dy, 0 < inf Ws , Wt ∈ dz
0 0 Ws() 2 0≤s≤t
0<x
√ 2
= 2−1 xzeµ(z−x)−µ t/2−y/8 eig (y/2, t, (x2 + z 2 )/2, xz)dgdydz

p2
 Z τ
q
n   o
(µ) (µ)
1.22.5 Ex exp − + ds ; 0 < inf W , W ∈ dz
2 Ws() 2 Ws() s τ


0 0≤s≤τ
0<x
= λeµ(z−x) Q(22)
λ+µ2 /2 (z)dz, for Q(22)
λ (y, z) see 1.1.22.5

τ τ
ds ds
Z Z 
(µ) (µ)
1.22.6 Px
∈ dg, ∈ dy, 0 < inf W , W ∈ dz
0 Ws() 2 0 Ws
()
0≤s≤τ
s τ
0<x
 (2 + 2 )xz g (x + z ) 2 + 2 ch(y 2 + 2 =2)
p  p p 
= exp µ(z − x) − −
8 sh(y 2 + 2 =2) 8 sh(y 2 + 2 =2)
p p

2 (2 + 2 )xz 
 p
× ig/8 dgdydz
sh(y 2 + 2 =2)
p
 1. EXPONENTIAL STOPPING 285

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

 2Z t  Z t
p ds ds
n o
1.22.7 Ex exp − ()
2
; ()
∈ dy, 0 < inf Ws(µ) , Wt(µ) ∈ dz
2 0 Ws 0 Ws 0≤s≤t
(1)
√ 2 p √
0<x = xz eµ(z−x)−µ t/2
ist ( 4p2 + 1, y/2, 0, x + z, xz)dydz

t Z t
ds ds
Z 
1.22.8 Px () 2
∈ dg, ()
∈ dy, 0 < inf Ws(µ) , Wt(µ) ∈ dz
Ws 0 Ws


0 0≤s≤t

√ 2 √
= 8−1 xzeµ(z−x)−µ t/2−g/8 eit (g/8, y/2, x + z, 2 xz)dgdydz
n Z τ

 o
p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds ; sup Ws(µ) < b


1.23.1 Ex exp
0 0≤s≤τ
r<b

x≤r =
+p
 (p − q)e(r−x) ( + q cth((b − r)q )) sh((b − r)q ) − ( + p)q e(b−x) e(x−r)p


+
( + p)( + q)(p sh((b − r)q ) + q ch((b − r)q ))

 e(b−x) sh((x − r)q )

 
r≤x = 1−
+q sh((b − r)q )
x≤b
 (q − p)e(r−x) (p − ) sh((b − r)q ) − ( + p)q e(b−x) sh((b − x)q )


+
( + p)( + q)(p sh((b − r)q ) + q ch((b − r)q )) sh((b − r)q )

n  Z ∞
 o
1.23.3 Ex exp −q 1I[r,∞) Ws(µ) ds ; sup Ws(µ) < b r∨x≤b
0 0≤s<∞
(1)
µ<0
2||e2(r−x) sh((b − r) 2 + 2q)
p
2µ(r−x)
x≤r =1−e +
2 + 2q ch((b − r) 2 + 2q) + || sh((b − r) 2 + 2q)
p p p

2||e(r−x) sh((b − x) 2 + 2q)

p
r≤x =p 2
 + 2q ch((b − r) 2 + 2q) + || sh((b − r) 2 + 2q)
p p
x≤b

Z ∞ 
1I[r,∞) Ws(µ) ds ∈ dy, sup Ws(µ) < b


1.23.4 Px r∨x≤b
0 0≤s<∞
(1)
µ<0
2
x≤r = 2e2µ(r−x)−µ y/2
rcy (0, b − r, b − r, 1/|µ|)dy

p
Υs := 2λ + µ2 + 2s
286 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

2
r≤x = 2eµ(r−x)−µ y/2
rcy (0, b − r, b − x, 1/|µ|)dy
x≤b

n  Z τ


p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds ;


1.23.5 Ex exp −
r<b 0
o
z<b sup Ws(µ) < b, Wτ(µ) ∈ dz = λeµ(z−x) Q(23)
λ+µ2 /2 (z)dz
0≤s≤τ

for Q(23)
λ (z) see 1.1.23.5

n  Z τ

 o
p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds ; a < inf Ws(µ)


1.24.1 Ex exp −
0 0≤s≤τ
a<r
 e(a−x) sh((r − x)p )
 
a≤x = 1−
+p sh((r − a)p )
x≤r
 (p − q)e(r−x) (q + ) sh((r − a)p ) − ( + q)p e(a−x) sh((x − a)p )


+
( + p)( + q)(q sh((r − a)p ) + p ch((r − a)p )) sh((r − a)p )

r≤x =
+q
 (q − p)e(r−x) (p cth((r − a)p ) − ) sh((r − a)p ) − ( + q)p e(a−x) )e(r−x)q
+
( + p)( + q)(q sh((r − a)p ) + p ch((r − a)p ))

n  Z ∞
 o
1.24.3 Ex exp −p 1I(−∞,r) Ws(µ) ds ; a < inf Ws(µ) a≤r∧x
0 0≤s<∞
(1)
µ>0
2e(r−x) sh((x − a) 2 + 2p)
p
a≤x =
2 + 2p ch((r − a) 2 + 2p) +  sh((r − a) 2 + 2p)
p p p
x≤r
2e2(r−x) sh((r − a) 2 + 2p)
p
r≤x = 1 − e2µ(r−x) +
2 + 2p ch((r − a) 2 + 2p) +  sh((r − a) 2 + 2p)
p p p

Z ∞ 
1I(−∞,r) Ws(µ) ds ∈ dy, a < Ws(µ)


1.24.4 Px inf
0 0≤s<∞
(1)
µ>0
2
a≤x = 2eµ(r−x)−µ y/2
rcy (0, r − a, x − a, 1/|µ|)dy
x≤r

p
Υs := 2λ + µ2 + 2s
 1. EXPONENTIAL STOPPING 287

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

2
r≤x = 2e2µ(r−x)−µ y/2
rcy (0, r − a, r − a, 1/|µ|)dy

n  Z τ


p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds ;


1.24.5 Ex exp −
a<r 0
o
a<z a < inf Ws(µ) , Wτ(µ) ∈ dz = λeµ(z−x) Q(24)
a,λ+µ2 /2 (z)dz
0≤s≤τ

for Q(24)
a,λ (z) see 1.1.24.5

Ex e−γ`(τ,r) ; a < inf Ws(µ) , sup Ws(µ) < b


1.25.1
0≤s≤τ 0≤s≤τ
r<b

e(a−x) sh((r − x)0 ) + e(r−x) sh((x − a)0 )

a≤x =1−
sh((r − a)0 )
x≤r
(e(r−x) sh((b − a)0 ) − e(b−x) sh((r − a)0 ) − e(a−x) sh((b − r)0 )) sh((x − a)0 )
+
0−1 (2 sh((b − r)0 ) sh((r − a)0 ) + 0 sh((b − a)0 )) sh((r − a)0 )

e(r−x) sh((b − x)0 ) + e(b−x) sh((x − r)0 )

r≤x =1−
sh((b − r)0 )
x≤b
(e(r−x) sh((b − a)0 ) − e(b−x) sh((r − a)0 ) − e(a−x) sh((b − r)0 )) sh((b − x)0 )
+
0−1 (2 sh((b − r)0 ) sh((r − a)0 ) + 0 sh((b − a)0 )) sh((b − r)0 )

Px `(τ, r) ∈ dy, a < inf Ws(µ) , sup Ws(µ) < b

1.25.2
0≤s≤τ 0≤s≤τ
r<b

y0 sh((b − a)0 )

 
= exp −
2 sh((b − r)0 ) sh((r − a)0 )

  (er sh((b − a) ) − eb sh((r − a) ) − ea sh((b − r) )) sh((x − a) )
0 0 0 0 0
dy,
2 ex sh((b − r)0 ) sh2 ((r − a)0 )





 a≤x≤r
× r sh((b − a)0 ) − eb sh((r − a)0 ) − ea sh((b − r)0 )) sh((b − x)0 )
0 (e
dy,
2ex sh2 ((b − r)0 ) sh((r − a)0 )






r≤x≤b

p
Υ0 := 2λ + µ2
288 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

Px `(τ, r) = 0, sup Ws(µ) < b

(1)
0≤s≤τ

 e(a−x) sh((r − x)0 ) + e(r−x) sh((x − a)0 )

 1−

sh((r − a) )
, a≤x≤r
0
= (r−x) sh (b − x)0 + e(b−x) sh (x − r)0

e


 1−

, r≤x≤b
sh (b − r)0

Ex e−γ`(τ,r) ; a < inf Ws(µ) , sup Ws(µ) < b, Wτ(µ) ∈ dz


1.25.5
0≤s≤τ 0≤s≤τ
r<b
z<b
= λeµ(z−x) Q(25)
a,λ+µ2 /2 (z)dz, for Q(25)
a,λ (z) see 1.1.25.5

Px `(τ, r) ∈ dy, a < inf Ws(µ) , sup Ws(µ) < b, Wτ(µ) ∈ dz

1.25.6
0≤s≤τ 0≤s≤τ
r<b
z<b
= λeµ(z−x) Ba,λ+µ
(25)
2 /2 (y, z)dydz,
(25)
for Ba,λ (y, z) see 1.1.25.6

Px `(τ, r) = 0, a < inf Ws(µ) , sup Ws(µ) < b, Wτ(µ) ∈ dz

(1)
0≤s≤τ 0≤s≤τ

e(z−x) ch((r − a − |z − x|)0 ) − ch((r + a − x − z )0 )

z≤r = dz
0 sh((r − a)0 )
x≤r

e(z−x) ch((b − r − |z − x|)0 ) − ch((b + r − z − x)0 )

r≤z = dz
0 sh((b − r)0 )
r≤x

Px `(t, r) ∈ dy, a < inf Ws(µ) , sup Ws(µ) < b, Wt(µ) ∈ dz

1.25.8
0≤s≤t 0≤s≤t

2
= eµ(z−x)−µ t/2 (25)
Fa,x (t, y, z)dydz, (25)
for Fa,x (t, y, z) see 1.1.25.8

p
Υ0 := 2λ + µ2
 1. EXPONENTIAL STOPPING 289

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

n  Z τ


p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds ;


1.26.1 Ex exp −
r<b 0
o
a<r a < inf Ws(µ) , sup Ws(µ) < b
0≤s≤τ 0≤s≤τ

 e(a−x) sh((r − x)p )

 
a≤x = 1−
+p sh((r − a)p )
x≤r

(p − q)e(r−x) (q cth((b − r)q ) + ) sh((x − a)p )

+
( + p)( + q)(p cth((r − a)p ) + q cth((b − r)q )) sh((r − a)p )

(( + p)q e(b−x) sh((r − a)p ) + ( + q)p e(a−x) sh((b − r)q )) sh((x − a)p )
−
( + p)( + q)(p cth((r − a)p ) + q cth((b − r)q )) sh((b − r)q ) sh2 ((r − a)p )

 e(b−x) sh((x − r)q )

 
r≤x = 1−
+q sh((b − r)q )
x≤b

(q − p)e(r−x) (p cth((r − a)p ) − ) sh((b − x)q )

+
( + p)( + q)(p cth((r − a)p ) + q cth((b − r)q )) sh((b − r)q )

(( + p)q e(b−x) sh((r − a)p ) + ( + q)p e(a−x) sh((b − r)q )) sh((b − x)q )
−
( + p)( + q)(p cth((r − a)p ) + q cth((b − r)q )) sh2 ((b − r)q ) sh((r − a)p )

n  Z τ


p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds ;


1.26.5 Ex exp −
r<b 0
o
a<r a < inf Ws(µ) , sup Ws(µ) < b, Wτ(µ) ∈ dz
0≤s≤τ 0≤s≤τ

= λeµ(z−x) Q(26)
a,λ+µ2 /2 (z)dz, for Q(26)
a,λ (z) see 1.1.26.5

n  Z τ

 1 o
p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds ; `(τ, r) ∈ dy dy


1.27.1 Ex exp −
0
2

√ √

2e(r−x) 2e(r−x) 
(x−r−y) 2λ+µ2+2p−y 2λ+µ2+2q
x≤r = + e
2 + 2 + 2p +  2 + 2 + 2q − 
p p

p
Υs := 2λ + µ2 + 2s
290 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

√ √ √
2e(r−x) e(x−r) 2+2+2p d −y


2λ+µ2+2p+ 2λ+µ2+2q
or =− p 2 2 e
( 2 +  + 2p + )( 2 +  + 2q − ) dy
p

√ √

2e(r−x) 2e(r−x)  2 2
r≤x = 2 +p 2 e(r−x−y) 2λ+µ +2q−y 2λ+µ +2p
2 +  + 2p +  2 +  + 2q − 
p

√ √ √
2e(r−x) e(r−x) 2+2+2q d −y


2λ+µ2+2p+ 2λ+µ2+2q
or =− p e
( 2 + 2 + 2p + )( 2 + 2 + 2q − ) dy
p

n  Z τ

 o
p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds ; `(τ, r) = 0


(1) Ex exp −
0

√

 
2λ+µ2 +2p−µ)
x≤r = 1 − e(x−r)(
+p
√

 
2λ+µ2 +2q+µ)
r≤x = 1 − e(r−x)(
+q

Z τ τ
1
Z 
1I(−∞,r) Ws(µ) ds ∈ du, 1I[r,∞) Ws(µ) ds ∈ dv, `(τ, r) ∈ dy dy



1.27.2 Px
0 0
2

=: λe−λ(u+v) Bx(27) (u, v, y)dudv

 −y2 =2v−2 v=2 √ 

d e  y  v

x≤r = −λe−λ(u+v) eµ(r−x) √ + √ e−µy Erfc √ − √
dy v 2 2v 2

 −(y+r−x)2 =2u−2 u=2 √ 

e  y+r−x  u

× √ − √ eµ(y+r−x) Erfc √ + √ dudv
u 2 2u 2

yu + (y + r − x)v µ(r−x)−(y+r−x)2 /2u−y2 /2v−µ2 t/2


or = λe−λ(u+v) e
v3=2 u3=2

√ 
y y+r−x  u

2µ(r−x)−(µv−y)2 /2v
−√ e Erfc √ + √
2v3=2 2u 2

√ 
(y + r − x) 2µ(r−x)−(y+r−x+µu)2 /2u y  v

+ √ e Erfc √ − √ dudv
2u3=2 2v 2

 −y2 =2u−2 u=2 √ 

d e  y  u

r≤x = −λe−λ(u+v) eµ(r−x) √ − √ eµy Erfc √ + √
dy u 2 2u 2
 1. EXPONENTIAL STOPPING 291

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

 −(y+x−r)2 =2v−2 v=2 √ 

e  y+x−r  v

× √ + √ e−µ(y+x−r) Erfc √ − √ dudv
v 2 2v 2


(y + x − r)u + yv 2
/2u−(y+x−r)2 /2v−µ2 t/2
or = λe−λ(u+v) eµ(r−x)−y
v3=2 u3=2

√ 
(y + x − r) 2µ(r−x)−(y+x−r−µv)2 /2v y  u

− √ e Erfc √ + √
2v3=2 2u 2

√ 
y y+x−r  v

2µ(r−x)−(y+µu)2 /2u
+√ e Erfc √ − √ dudv
2u3=2 2v 2

Z τ Z τ 
(µ)
1I[r,∞) Ws(µ) ds = 0, `(τ, r) = 0



(1) Px 1I(−∞,r) Ws ds ∈ du,
x≤r 0 0
√ √

1 r−x  u
 
1 
r−x  u

= λe−λu 1 − Erfc √ − √ − e2µ(r−x) Erfc √ + √ du
2 2t 2 2 2u 2

Z τ Z τ 
(µ)
1I[r,∞) Ws(µ) ds ∈ dv, `(τ, r) = 0



(2) Px 1I(−∞,r) Ws ds = 0,
x≤r 0 0
√ √

1 x−r  v
 
1 
x−r  v

= λe−λv 1 − Erfc √ + √ − e2µ(r−x) Erfc √ − √ dv
2 2t 2 2 2v 2

n  Z ∞
 1 o
1.27.3 Ex exp −q 1I[r,∞) Ws(µ) ds ; `(∞, r) ∈ dy
0
2
(1)
µ<0
( √ 2
2|µ|e2µ(r−x)−y( µ +2q−µ) dy, x≤r
= √ 2 √ 2
2|µ|e(r−x)( µ +2q+µ)−y( µ +2q−µ) dy, r ≤ x

n  Z ∞
 o
(2) Ex exp −q 1I[r,∞) Ws(µ) ds ; `(∞, r) = 0 = 1 − eµ(|r−x|+r−x)
µ<0 0

n  Z ∞
 1 o
(3) Ex exp −p 1I(−∞,r) Ws(µ) ds ; `(∞, r) ∈ dy
0
2
µ>0
( √ √
µ2+2p−µ)−y( µ2+2p+µ)
2µe(x−r)( dy, x ≤ r
= √
µ2+2p+µ)
2µe−2µ(x−r)−y( dy, r≤x
292 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

n  Z ∞
 o
(4) Ex exp −p 1I(−∞,r) Ws(µ) ds ; `(∞, r) = 0 = 1 − e−µ(|x−r|+x−r)
µ>0 0

Z t
1 
p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds ∈ dv, `(t, r) ∈ dy




1.27.4 Px
0
2

1 
|qt − v | |pt − v |

= Bx(27) , ,y 1I((q∧p)t,(q∨p)t) (v) dvdy
|p − q | |p − q | |p − q |

for Bx(27) u, v, y see 1.27.2

Z t 
p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds = pt, `(t, r) = 0




(1) Px
x≤r 0

√ √
1 
r−x  t

1 
r−x  t

=1− Erfc √ − √ − e2µ(r−x) Erfc √ + √
2 2t 2 2 2t 2

Z t 
p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds = qt, `(t, r) = 0




(2) Px
r≤x 0

√ √
1 
x−r  t

1 
x−r  t

=1− Erfc √ + √ − e2µ(r−x) Erfc √ − √
2 2t 2 2 2t 2

∞
Z
1 
1I[r,∞) Ws(µ) ds ∈ du, `(∞, r) ∈ dy


(3) Px
0
2
µ<0
2||y 2µ(r−x) −(y−µu)2 /2u

√ e e dudy, x≤r
2u3=2


=
 2|√
 |(y + x − r) 2µ(r−x) −(y+x−r−µu)2 /2u
e e dudy, r ≤ x
2u3=2

Z ∞ 
1I[r,∞) Ws(µ) ds = 0, `(∞, r) = 0 = 1 − eµ(|r−x|+r−x)


(4) Px
µ<0 0

∞
Z
1 
1I(−∞,r) Ws(µ) ds ∈ du, `(∞, r) ∈ dy


(5) Px
0
2
µ>0
 1. EXPONENTIAL STOPPING 293

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

 2(y + r − x) 2
 √ 3=2 e2µ(r−x) e−(y+r−x+µu) /2u dudy, x ≤ r
=
2u
 √ 2y e−2µ(x−r) e−(y+µu)2 /2u dudy, r≤x
2u3=2
Z ∞ 
1I(−∞,r) Ws(µ) ds = 0, `(∞, r) = 0 = 1 − e−µ(|x−r|+x−r)


(6) Px
µ>0 0

n  Z τ

 1 o
p1I(−∞,r) Ws(µ) +q1I[r,∞) Ws(µ) ds ; `(τ, r) ∈ dy, Wτ(µ) ∈ dz


1.27.5 Ex exp −
0
2

√ √
2λ+µ2 +2q −(y+2r−x−z) 2λ+µ2 +2p
x≤r = 2λeµ(z−x) e−y e dydz
z≤r
√ √
2λ+µ2 +2q −(y+r−x) 2λ+µ2 +2p
x≤r = 2λeµ(z−x) e−(y+z−r) e dydz
r≤z
√ √
2λ+µ2 +2q −(y+r−z) 2λ+µ2 +2p
r≤x = 2λeµ(z−x) e−(y+x−r) e dydz
z≤r
√ √
2λ+µ2 +2q −y 2λ+µ2 +2p
r≤x = 2λeµ(z−x) e−(y+x+z−2r) e dydz
r≤z

n  Z τ

 o
p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds ; `(τ, r) = 0, Wτ(µ) ∈ dz


(1) Ex exp −
0

√ √

 
µ(z−x) −|z−x| 2λ+µ2 +2p (z+x−2r) 2λ+µ2 +2p
x≤r = e e − e dz
2 + 2 + 2p
p
z≤r

√ √

 
µ(z−x) −|z−x| 2λ+µ2 +2q (2r−x−z) 2λ+µ2 +2q
r≤x = e e − e dz
2 + 2 + 2q
p
r≤z

Z τ Z τ
1I(−∞,r) Ws(µ) ds ∈ du, 1I[r,∞) Ws(µ) ds ∈ dv,



1.27.6 Px
0 0

1 
`(τ, r) ∈ dy, Wτ(µ) ∈ dz = λeµ(z−x) Bλ+µ
(27)
2 /2 (u, v, y, z)dudvdydz
2

for Bλ(27) (u, v, y, z) see 1.1.27.6

294 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

Z t
1 
p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds ∈ dv, `(t, r) ∈ dy, Wt(µ) ∈ dz




1.27.8 Px
0
2
1 2

|qt − v | |pt − v |

= eµ(z−x)−µ t/2
Bx(27) , , y, z 1I((q∧p)t,(q∨p)t) (v) dvdydz
|p − q | |p − q | | p − q |

for Fx(27) (u, v, y, z) see 1.1.27.6

Ex e−γ Ĥµ (τ )−ηȞµ (τ ) ; inf Ws(µ) ∈ da, sup Ws(µ) ∈ db


1.28.1
0≤s≤τ 0≤s≤τ
a<x
x<b
 sh((b − x) + ) e(b−x) (0 ch((b − a)0 ) −  sh((b − a)0 )) − e(a−x) 0
 
= dadb
sh((b − a) + ) sh((b − a) ) sh((b − a)0 )
 sh((x − a) + ) e(a−x) (0 ch((b − a)0 ) +  sh((b − a)0 )) − e(b−x) 0
 
+ dadb
sh((b − a) + ) sh((b − a) ) sh((b − a)0 )

Px Ĥ(τ ) ∈ du, Ȟ(τ ) ∈ dv, inf Ws(µ) ∈ da, sup Ws(µ) ∈ db

1.28.2
0≤s≤τ 0≤s≤τ
 p2 + 2 eb ch((b − a)p2 + 2 ) − ea

µb
u<v = − µe
sh((b − a) 2 + 2 )
p

2
×e−µx−(λ+µ /2)v
ssu (b − x, b − a) sv−u (b − a) dudvdadb

 p2 + 2 ea ch((b − a)p2 + 2 ) − eb


µa
v<u = + µe
sh((b − a) 2 + 2 )
p

2
×e−µx−(λ+µ /2)u
ssv (x − a, b − a) su−v (b − a) dudvdadb

Ex e−γ Ĥµ (τ )−ηȞµ (τ ) ; inf Ws(µ) ∈ da, sup Ws(µ) ∈ db, Wτ(µ) ∈ dz

1.28.5
0≤s≤τ 0≤s≤τ
a<x
x<b
2e(z−x)  sh((b − x) + ) sh((z − a)0 )
= dadbdz
sh((b − a) + ) sh((b − a) ) sh((b − a)0 )
2e(z−x)  sh((x − a) + ) sh((b − z )0 )
+ dadbdz
sh((b − a) + ) sh((b − a) ) sh((b − a)0 )

p
Υs := 2λ + µ2 + 2s
 1. EXPONENTIAL STOPPING 295

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

Px Ĥ(τ ) ∈ du, Ȟ(τ ) ∈ dv, inf Ws(µ) ∈ da, sup Ws(µ) ∈ db, Wτ(µ) ∈ dz


1.28.6
0≤s≤τ 0≤s≤τ

µ(z−x)−(λ+µ2/2)v sh((z − a) 2 + 2 )
p
u<v = 2λe
sh((b − a) 2 + 2 )
p

× ssu (b − x, b − a) sv−u (b − a) dudvdadbdz

2 sh((b − z ) 2 + 2 )
p
2
v<u = eµ(z−x)−(λ+µ /2)u
sh((b − a) 2 + 2 )
p

× ssv (x − a, b − a) su−v (b − a) dudvdadbdz

Px Ĥµ (t) ∈ du, Ȟµ (t) ∈ dv, inf Ws(µ) ∈ da, sup Ws(µ) ∈ db, Wt(µ) ∈ dz


1.28.8
0≤s≤t 0≤s≤t

for 0 < u ∧ v < u ∨ v < t, a<x∧z <x∨z <b

2
u<v = 2eµ(z−x)−µ t/2
ssu (b − x, b − a) sv−u (b − a) sst−v (z − a, b − a) dudvdadbdz

2
v<u = 2eµ(z−x)−µ t/2
ssv (x − a, b − a) su−v (b − a) sst−u (b − z, b − a)dudvdadbdz

n  Z τ


p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds ; a < inf Ws(µ) ,


1.29.1 Ex exp −
0 0≤s≤τ

1 o
sup Ws(µ) < b, `(τ, r) ∈ dy eµ(x−r)
0≤s≤τ 2

  ch((b − r) ) − e(b−r)
 ch((r − a) ) − e(a−r)

q q p p (p − q)

=λ + +
( + q) sh((b − r)q ) ( + p) sh((r − a)p ) ( + p)( + q)


sh((x − a)p )
 sh((r − a) ) dy, a ≤ x ≤ r
yq ch((b − r)q ) yp ch((r − a)p )

p

× exp − −
sh((b − r)q ) sh((r − a)p )  sh((b − x )q )
dy, r ≤ x ≤ b
sh((b − r)q )


p
Υs := 2λ + µ2 + 2s
296 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

n  Z τ


p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds ; a < inf Ws(µ) ,


(1) Ex exp −
0 0≤s≤τ
o
sup Ws(µ) < b, `(τ, r) = 0
0≤s≤τ

 e(a−x) sh((r − x)p ) + e(r−x) sh((x − a)p )

  
 +p 1−

sh((r − a)p )
, a≤x≤r
= (r−x) sh((b − x)q ) + e(b−x) sh((x − r)q )
  1− e
 
, r≤x≤b

+q sh((b − r)q )

∞
n  Z
 1 o
1.29.3 Ex exp −q 1I[r,∞) Ws(µ) ds ; sup Ws(µ) < b, `(∞, r) ∈ dy
0 0≤s<∞ 2
(1)
µ<0
y 2 + 2q ch((b − r) 2 + 2q)
 p p 
x≤r = 2|µ| exp 2µ(r − x) − + yµ dy
sh((b − r) 2 + 2q)
p

2|| sh((b − x) 2 + 2q) y 2 + 2q ch((b − r) 2 + 2q)

p  p  p
r≤x = exp µ(r − x + y) − dy
sh((b − r) 2 + 2q) sh((b − r) 2 + 2q)
p p
x≤b

∞
n  Z
 1 o
(2) Ex exp −p 1I(−∞,r) Ws(µ) ds ; a < inf Ws(µ) , `(∞, r) ∈ dy
0 0≤s<∞ 2
µ>0

2 sh((x − a) 2 + 2p) y 2 + 2p ch((r − a) 2 + 2p)

p  p  p
a≤x = exp µ(r − x − y) − dy
sh((r − a) 2 + 2p) sh((r − a) 2 + 2p)
p p
x≤r

y 2 + 2p ch((r − a) 2 + 2p)
 p p 
r≤x = 2µ exp 2µ(r − x) − 2 − yµ dy
sh((r − a)  + 2p)
p

∞
Z
1 
1I[r,∞) Ws(µ) ds ∈ du, sup Ws(µ) < b, `(∞, r) ∈ dy


1.29.4 Px
0 0≤s<∞ 2
(1)
µ<0
2
(
2|µ|e2µ(r−x)+µy e−µ u/2
esu (0, 0, b − r, 0, y)dudy, x≤r
= 2
µ(r−x)+µy −µ u/2
2|µ|e e essu (b − x, b − r, 0, y)dudy, r≤x≤b

∞
Z
1 
1I(−∞,r) Ws(µ) ds ∈ du, a < Ws(µ) , `(∞, r) ∈ dy


(2) Px inf
0 0≤s<∞ 2
µ>0
 1. EXPONENTIAL STOPPING 297

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

2
(
2µeµ(r−x)−µy e−µ u/2
essu (x − a, r − a, 0, y)dudy, a ≤ x ≤ r
=
2µ(r−x)−µy −µ2 u/2
2µe e esu (0, 0, r − a, 0, y)dudy, r≤x

n  Z τ


p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds ; a < inf Ws(µ) ,


1.29.5 Ex exp −
0 0≤s≤τ

1 o
sup Ws(µ) < b, `(τ, r) ∈ dy, Wτ(µ) ∈ dz = λeµ(z−x) Q(29)
λ+µ2 /2 (y, z)dydz
0≤s≤τ 2

for Q(29)
λ (y, z) see 1.1.29.5

n  Z τ


p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds ; a < inf Ws(µ) ,


(1) Ex exp −
0 0≤s≤τ
o
sup Ws(µ) < b, `(τ, r) = 0, Wτ(µ) ∈ dz
0≤s≤τ

(ch((r − a − |z − x|) 2 + 2 + 2p) − ch((r + a − z − x) 2 + 2 + 2p))

p p
x≤r = dz
e(x−z) 2 + 2 + 2p sh((r − a) 2 + 2 + 2p)
p p
z≤r

(ch((b − r − |z − x|) 2 + 2 + 2q) − ch((b + r − z − x) 2 + 2 + 2q))

p p
r≤x = dz
e(x−z) 2 + 2 + 2q sh((b − r) 2 + 2 + 2q)
p p
r≤z

Z τ Z τ
(µ)
1I[r,∞) Ws(µ) ds ∈ dv, a < inf Ws(µ) ,



1.29.6 Px 1I(−∞,r) Ws ds ∈ du,
0 0 0≤s≤τ

1 
sup Ws(µ) < b, `(τ, r) ∈ dy, Wτ(µ) ∈ dz
0≤s≤τ 2

2
= λeµ(z−x)−(λ+µ /2)(u+v)
Bx(29) (u, v, y, z)dudvdydz

for Bx(29) (u, v, y, z) see 1.1.29.6

Z t
p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds ∈ dv, a < inf Ws(µ) ,




1.29.8 Px
0 0≤s≤t

1 
sup Ws(µ) < b, `(t, r) ∈ dy, Wt(µ) ∈ dz
0≤s≤t 2
298 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

1 2

|qt − v | |pt − v |

= eµ(z−x)−µ t/2
Bx(29) , , y, z 1I((q∧p)t,(q∨p)t) (v) dvdydz
|p − q | |p − q | |p − q |

for Bx(29) (u, v, y, z) see 1.1.29.6

 2Z τ τ
p
n  Z o
2βWs(µ) (µ)
1.30.1 Ex exp − e ds ; eβWs ds ∈ dy
2 0 0
(1)
p ch(yp| |=2)
1/2−µ/β
(= + 2 + 2 =| |)
 p
β 6= 0 =λ
| | sh(yp| |=2) (1 + 2 2 + 2 =| |)
p

x pe x ch(yp| |) 2pe x 
  
× exp −µx − − M1/2−µ/β,√2λ+µ2 /|β| dy
2 | | sh(yp| |) | | sh(yp| |)

 2Z t
p ch(yp| |=2) 1/2−µ/β
 Z t
p
n (µ) (µ)
o  
1.30.3 Ex exp − e2βWs ds ; eβWs ds ∈ dy =
2 0 0
| | sh(yp| |=2)
(1)

2 
x 2 t pe x ch(yp| |)
 
 1 pe x

β 6= 0 × exp −µx − − − mβ 2 t/2 − , dy
2 2 2 | | sh(yp| |) 2 | | sh(yp| |)

p2
 Z ∞  
(µ) (µ)
(2) Ex exp − e2βWs + qeβWs ds
0
2

||
+ p|q | + 12  2p |µ|/|β|−1/2
 
 | |

2p 
<0 = 
2||
 e−(µ+β/2)x W−q/p|β|,|µ|/|β| eβx
| | | |
| |

 2Z ∞ ∞
p
n  Z o
2βWs(µ) (µ)
(3) Ex exp − e ds ; eβWs ds ∈ dy
2 0 0

 2 p
 1+2|µ|/|β| 
pe x ch(yp| |=2)

<0 = exp −2µx − dy
2 (2||=| |) | | sh(yp| |=2) | | sh(yp| |=2)

Z ∞ Z ∞ 
(µ) (µ)
1.30.4 Px e2βWs ds ∈ dg, eβWs ds ∈ dy
(1) 0 0
 1. EXPONENTIAL STOPPING 299

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

 | |1−2||=| | e−2x 2|| y| | e x 


<0 = eeg 1 + , , , 0 dgdy
2 (2||=| |) | | 2 | |

 Z τ 2
p 2βWs(µ)
n (µ)
  o
1.30.5 Ex exp − e + qeβWs ds ; Wτ(µ) ∈ dz
0
2

= λeµ(z−x) Q(30)
λ+µ2 /2 (z)dz, for Q(30)
λ (z) see 1.1.30.5

 2Z τ τ
p
n (µ)
 Z (µ)
o
(1) Ex exp − e2βWs ds ; eβWs ds ∈ dy, Wτ(µ) ∈ dz
2 0 0

= λeµ(z−x) Bλ+µ
(30)
2 /2 (y, z)dydz, for Bλ(30) (y, z) see 1.1.30.5(1)

Z τ Z τ 
(µ) (µ)
1.30.6 Px e2βWs ds ∈ dg, eβWs ds ∈ dy, Wτ(µ) ∈ dz
0 0

2 2 + 2 y| | 1 βx
1
 p 
β 6= 0 = λeµ(z−x) isg , , 0, e + eβz , eβ(x+z)/2 dgdydz
| | 2 | | | |

 2Z t  Z t
p
n (µ) (µ)
o
1.30.7 Ex exp − e2βWs ds ; eβWs ds ∈ dy, Wt(µ) ∈ dz
2 0 0
(1)
2
= eµ(z−x)−µ t/2 (30)
Fx,t (y, z)dydz, (30)
for Fx,t (y, z) see 1.1.30.7(1)

Z t Z t 
2βWs(µ) (µ)
1.30.8 Px e ds ∈ dg, eβWs ds ∈ dy, Wt(µ) ∈ dz
0 0

2 2
 2
t y| | 1 βx
2 
β 6= 0 = eµ(z−x)−µ t/2
eig , , e + eβz , eβ(x+z)/2 dgdydz
8 8 2 | | | |

u r
 p


Ex/√2λ+µ2 exp −γ 2λ + µ2 sup ` s, p − α−1 ` s, p


1.31.1 2 2
0≤s≤τ 2 +  2 + 
0<α

2 − 2e−2|u−r| −  e|u−r|−|x−r| +  e−|u−x| (u−x)µ/√2λ+µ2

=1− e
2 1 − e−2|u−r| + 


300 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws τ ∼ Exp(λ), independent of W (µ)

u r h
 
1.31.2 Px/√2λ+µ2 sup ` s, p −1




− α ` s, >
2 + 2 2 + 2 2 + 2
p p
0≤s≤τ


h
i
0<α = e|u−r|−|x−r| − 2 u r
e|u−r|−|x−r| − e−|u−x|
2 1−e− | − |

(u − x) h
 
× exp p − 2 u r
2 +  2 2 1−e− | − |

u
r


 
1.31.3 Ex/|µ| exp −γ|µ| sup ` s, − α−1 ` s,
0≤s<∞ | | ||
(1)
2 − 2e−2|u−r| −  e|u−r|−|x−r| +  e−|u−x| (u−x)µ/|µ|


=1− e
2 1 − e−2|u−r| + 

u
r

h h
   
1.31.4 Px/|µ| sup ` s, − α−1 ` s, > = exp −
| | || || 2 1 − e−2|u−r|


0≤s<∞
(1)


h i
×e(u−x)µ/|µ| e|u−r|−|x−r| −
e|u−r|−|x−r| − e−|u−x|


0<α 2 u r
2 1−e− | − |

u r
n  p


Ex/√2λ+µ2 exp −γ 2λ + µ2 sup ` s, p −1


1.31.5 − α ` s, ;
2 + 2 2 + 2
p
0≤s≤τ
0<α

√
dz 
o 2

Wτ(µ) ∈ p = e(z−x)µ/ 2λ+µ e−|z−x|
2 + 2 2 + 2

 e−|z−u| 2 − 2e−2|u−r| −  e|u−r|−|x−r| +  e−|u−x| 



− dz
(2 + 2 ) 2 1 − e−2|u−r| + 

u r h

Px/√2λ+µ2 − α−1 ` s, p




1.31.6 sup ` s, p 2 2 >p ,
0≤s≤τ 2 +  2 +  2 + 2
0<α

dz  ( z − x ) h
  
Wτ(µ) ∈ = exp − |z − u| −
2 + 2 2  + 2 2 +  2 2 1 e−2|u−r |
p p

−


h  i
× e|u−r|−|x−r| −
e|u−r|−|x−r| − e−|u−x| dz
2 1 − e−2|u−r|

q

Υ := (α − 1)2 + 4α 1 − e−2|u−r| + 1 − α
 2. STOPPING AT FIRST HITTING TIME 301

2 Ws(µ) = µs + Ws Hz = min{s : Ws(µ) = z}

2. Stopping at first hitting time

√
2α+µ2
2.0.1 Ex e−αHz = Ex {e−αHz ; Hz < ∞} = eµ(z−x)−|z−x|

|z − x|

(z − x − t)2 
2.0.2 Px Hz ∈ dt = √ 3=2 exp − dt
2t 2t

Px Hz = ∞ = 1 − eµ(z−x)−|µ||z−x|


(1)

e(µ−|µ|)(z−x) , x≤z
(
(µ)
sh( ( ))


2.1.2 Px sup Ws < y, Hz < ∞ = |  | y − x
eµ(z−x) , z≤x≤y
z<y 0≤s≤Hz
sh(||(y − z ))

sup Ws(µ) = ∞ = 1 − e−(µ+|µ|)(x−z)

(1) Px
0≤s≤Hz
z≤x
 √ 2
 e(z−x)(µ− 2α+µ ) , x≤z≤y
Ex e−αHz ; sup Ws(µ)

2.1.4 < y = sh((y − x) 2 + 2 )
p
0≤s≤Hz eµ(z−x) , z ≤ x ≤ y
sh((y − z ) 2 + 2 )
 p

2
sup Ws(µ) < y, Hz ∈ dt = eµ(z−x)−µ t/2 sst (y − x, y − z)dt


(1) Px
0≤s≤Hz
z≤x
x≤y
sh(||(x − y))


 eµ(z−x) , y≤x≤z
2.2.2 Px inf Ws(µ) > y, Hz < ∞ = sh(||(z − y))
0≤s≤Hz  −(µ+|µ|)(x−z)
e , y≤z≤x

Ws(µ) = −∞ = 1 − e(µ−|µ|)(z−x)


(1) Px inf
0≤s≤Hz
x≤z
sh((x − y) 2 + 2 ) µ(z−x)
 p

2 e , y≤x≤z
Ex e−αHz ; inf Ws(µ) > y = sh((z − y) √2 +  )
  p
2.2.4
0≤s≤Hz  (z−x)(µ+ 2α+µ2 )
e , y≤z≤x
2
Px inf Ws(µ) > y, Hz ∈ dt = eµ(z−x)−µ t/2 sst (x − y, z − y)dt


(1)
0≤s≤Hz
z≤x
y≤x
302 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws Hz = min{s : Ws(µ) = z}

Ex e−γ`(Hz ,r) ; Hz < ∞


2.3.1

 1, r ≤ z ≤ x, x≤z≤r
 | | + 1−e −2|||x−r |





, z ∧ r ≤ x ≤ z ∨ r
= eµ(z−x)−|µ||z−x| || + 1 − e−2|||z−r|
| |




, z ≤ r ≤ x, x≤r≤z
1 − e−2|||z−r|

| | +


 −γ`(Hz ,r)
(1) Ex e ; Hz = ∞ for µ(z − r) < 0

r ≤ z ≤ x, x ≤ z ≤ r

 0,
 || 1 − e−2|||x−z|





, z∧r ≤x≤z∨r
= || + 1 − e−2|||z−r|
−2|||x−r | − e−2|||x−z |
 1 − e−2|µ||x−r| + || e




, z ≤ r ≤ x, x ≤ r ≤ z

|| + 1 − e−2|||z−r|

||y

 
2.3.2 Px `(Hz , r) ∈ dy, Hz < ∞ = exp µ(z − x) − −2|||z −r |
1−e

0, x ≤ z ≤ r, r ≤ z ≤ x

 || sh(|||z − x|)



dy, r ∧ z ≤ x ≤ r ∨ z
× 2 sh2 (|||z − r|)
 ||e−|||r−x| dy,


z ≤ r ≤ x, x ≤ r ≤ z

2 sh(|||z − r|)

(1) Px `(Hz , r) ∈ dy, Hz = ∞ for µ(z − r) < 0

0, x ≤ z ≤ r, r ≤ z ≤ x

−2|||x−z |

 1 e  y


| | − | |

  
exp − dy, r∧z ≤x≤r∨z

= 1 − e−2|||z−r| 1 − e−2|||z−r|
 || e−2|||x−r| − e−2|||x−z| ||y

 

exp − dy, x ≤ r ≤ z
1 − e−2|||z−r| 1 − e−2|||z−r|


z≤r≤x

p
Ex e−αHz −γ`(Hz ,r) = exp µ(z − x) − |z − x| 2α + µ2


2.3.3

1, x ≤ z ≤ r, r ≤ z ≤ x

√

2 x r 2 + 
 2 + 2 + 1 − e

2
 p
− | − |



√
, r ∧ z ≤ x ≤ r ∨ z
× 2 + 2 + 1 − e−2|z−r| 2 +2
p

2 + 2

 p


, z ≤ r ≤ x, x ≤ r ≤ z

 √
2 +  + 1 − e−2|z−r| 2 +2
2
 p
 2. STOPPING AT FIRST HITTING TIME 303

2 Ws(µ) = µs + Ws Hz = min{s : Ws(µ) = z}

y 2 + 2
 p 
Ex e−αHz ; ` Hz , r ∈ dy = exp µ(z − x) −


2.3.4 √
1 − e−2|r−z| 2 +2
x ≤ z ≤ r, r ≤ z ≤ x

 0,
2 +  2 sh( x z 2 +  2 )
 p p
| − |


dy, z∧r ≤x≤r∨z

× 2 sh2 (|r − z | 2 + 2 )
p

 p
2 + 2 √
−|r−x| 2α+µ2

e dy, z ≤ r ≤ x, x ≤ r ≤ z

2 sh(|r − z | 2 + 2 )
 p

Ex e−αHz ; ` Hz , r = 0


(1)

 −|z−x|√2α+µ2
 e , x ≤ z ≤ r, r ≤ z ≤ x
| x − r | 2 + 2 )

sh(
 p
µ(z−x)
=e
2 , r∧z ≤x≤r∨z
 sh(|z − r| 2 +  )
p


0, z ≤ r ≤ x, x ≤ r ≤ z


1

(2) Px Hz ∈ dt, ` Hz , r ∈ dy
2

1
2
 2 (est (1, 2, |r − z|, y − |x − z|, y)

= eµ(z−x)−µ t/2 − est (1, 2, |r − z|, y + |x − z|, y))dtdy, z ∧ r ≤ x ≤ r ∨ z


est (1, 1, |r − z|, y + |x − r|, y)dtdy, z ≤ r ≤ x, x ≤ r ≤ z

n  Z Hz
 o
2.4.1 Ex exp −γ 1I[r,∞) Ws(µ) ds ; Hz < ∞ = eµ(z−x) Q(6) 2 2
z (µ , µ + 2γ)
0

for Q(6)
z (2p, 2q) see 1.2.6.1

n  Z Hz
 o
(1) Ex exp −γ 1I[r,∞) Ws(µ) ds ; Hz = ∞
µ<0 0

r≤z

e2(x−r)|| ( 2 + 2 ch((z − r) 2 + 2 ) − || sh((z − r) 2 + 2 ))

 p p p
1 − , x≤r
2 + 2 ch((z − r) 2 + 2 ) + || sh((z − r) 2 + 2 )

 p p p



2||e(x−r)|| sh((z − x) 2 + 2 )
p
=
, r≤x≤z
2 + 2 ch((z − r) 2 + 2 ) + || sh((z − r) 2 + 2 )

 p p p



0, z≤x
304 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws Hz = min{s : Ws(µ) = z}

Z Hz 
1I[r,∞) Ws(µ) ds ∈ dy, Hz < ∞ eµ(x−z)


2.4.2 Px
0
 −1 (x−r)|µ|−µ2 y/2

 |µ| e e y (1, z − r, 0, 1/|µ|)dy,
rc x≤r
 2
 e−µ y/2 ss (x − r, z − r)

y
r<z = −1




 +|µ| ssy (z − x, z − r) ∗ rc e y (1, z − r, 0, 1/|µ|) dy, r ≤ x ≤ z

 −µ2 y/2
e hy (1, x − z)dy, z≤x

0, x≤z


 −2 y=2
|| sh((x − z )||) e || ch((r − z )||)



√ −

sh 2 ((r − z )||) 2 y 2 sh((r − z )||)




2 y


    || ch((r − z )||)√y 
× exp Erfc √ dy, z≤x≤r

z<r = 2
2 sh ((r − z )||) 2 sh((r − z )||)
||e− y=2−(x−r) =2y || ch((r − z )||) (x − r)|| ch((r − z )||)

 2 2 

√ dy − exp

2 y sh(( r z )  ) 2 sh 2 (( r z )  ) sh((r − z )||)

− | | − | |



2y  ch(( r z )  )√y 

 x r

− | | − | |
  
+ 2 Erfc √ + √ dy, r≤x

2 sh ((r − z )||) 2y 2 sh((r − z )||)


Z Hz 
1I[r,∞) Ws(µ) ds ∈ dy, Hz = ∞


(1) Px
µ<0 0

2
(
2e2(x−r)|µ|−µ y/2
rcy (0, z − r, z − r, 1/|µ|)dy, x ≤ r
= 2
2e(x−r)|µ|−µ y/2
rcy (0, z − r, z − x, 1/|µ|)dy, r≤x≤z

Z Hz 
1I[r,∞) Ws(µ) ds ∈ dv, Hz ∈ dt


2.4.4 Px
0
Z ∞
2
= eµ(z−x)−µ t/2 Bz(27) (t − v, v, y)dy dvdt, for Bz(27) (u, v, y) see 1.2.27.2
0

Z Hz 
1I[r,∞) Ws(µ) ds = 0, Hz ∈ dt


(1) Px
0

2
(
eµ(z−x)−µ t/2
ht (1, z − x)dt, x≤z
= 2
eµ(z−x)−µ t/2
sst (r − x, r − z)dt, z ≤ x ≤ r

n  Z Hz
 o
2.5.1 Ex exp −γ 1I(−∞,r] Ws(µ) ds ; Hz < ∞ = eµ(z−x) Q(6) 2 2
z (µ + 2γ, µ )
0

for Q(6)
z (2p, 2q) see 1.2.6.1
 2. STOPPING AT FIRST HITTING TIME 305

2 Ws(µ) = µs + Ws Hz = min{s : Ws(µ) = z}

n  Z Hz
 o
(1) Ex exp −γ 1I(−∞,r] Ws(µ) ds ; Hz = ∞
0

x≤z

 0,
2||e(r−x)|| sh((x − z ) 2 + 2 )

 p

, z≤x≤r
 p
z≤r = 2 +  ch((r − z ) 2 + 2 ) + || sh((r − z ) 2 + 2 )
2
p p

e2(r−x)|| ( 2 + 2 ch((r − z ) 2 + 2 ) − || sh((r − z ) 2 + 2 ))


 p p p

1− , r≤x

2 + 2 ch((r − z ) 2 + 2 ) + || sh((r − z ) 2 + 2 )
 p p p

Z Hz 
1I(−∞,r] Ws(µ) ds ∈ dy, Hz < ∞ eµ(x−z)


2.5.2 Px
0

 −1 (r−x)|µ|−µ2 y/2

 |µ| e e y (1, r − z, 0, 1/|µ|)dy,
rc r≤x
 2
 e−µ y/2 ss (r − x, r − z)

y
z<r = −1




 +|µ| ssy (x − z, r − z) ∗ rc e y (1, r − z, 0, 1/|µ|) dy, z ≤ x ≤ r

 −µ2 y/2
e hy (1, z − x)dy, x≤z

0, z≤x


 −2 y=2
|| sh((z − x)||) e || ch((z − r)||)



√ −

sh 2 ((z − r)||) 2 y 2 sh((z − r)||)




2 y


    || ch((z − r)||)√y 
× exp Erfc √ dy, r ≤ x ≤ z

r<z = 2
2 sh ((z − r)||) 2 sh((z − r)||)
||e− y=2−(r−x) =2y || ch((z − r)||) (r − x)|| ch((z − r)||)

 2 2 

√ dy − exp

2y sh((z − r)||) 2 sh2 ((z − r)||) sh((z − r)||)




2 r − x || ch((z − r)||) y
√ 

 y

  
+ 2 Erfc √ + √ dy, x ≤ r

2 sh ((z − r)||) 2y 2 sh((z − r)||)


Z Hz 
1I(−∞,r] Ws(µ) ds ∈ dy, Hz = ∞


(1) Px
µ>0 0

2
(
2e(r−x)|µ|−µ y/2
rcy (0, r − z, x − z, 1/|µ|)dy, z≤x≤r
= 2
2e2(r−x)|µ|−µ y/2
rcy (0, r − z, r − z, 1/|µ|)dy, r ≤ x

0, x≤r

Hz
sh((x r )  )
Z 
− | |
 
Ws(µ) ds = 0 = eµ(z−x)


(2) Px 1I(−∞,r] , r≤x≤z
0
sh((z − r)||)
r≤z


1, z≤x
µ<0
306 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws Hz = min{s : Ws(µ) = z}

Z Hz 
1I(−∞,r) Ws(µ) ds ∈ dv, Hz ∈ dt


2.5.4 Px
0

Z ∞
µ(z−x)−µ2 t/2
=e Bz(27) (v, t − v, y)dy dvdt, for Bz(27) (u, v, y) see 1.2.27.2
0

Z Hz 
1I(−∞,r) Ws(µ) ds = 0, Hz ∈ dt


(1) Px
0

2
(
eµ(z−x)−µ t/2
sst (x − r, z − r)dt, r ≤ x ≤ z
= 2
eµ(z−x)−µ t/2
ht (1, x − z)dt, z≤x

 Z Hz


p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds


2.6.1 Ex exp −
0

µ(z−x) 2 2
=e Q(6)
z (2p + µ , 2q + µ ), for Q(6)
z (2p, 2q) see 1.2.6.1

Z Hz Z Hz 
1I(−∞,r) Ws(µ) ds ∈ du, 1I[r,∞) Ws(µ) ds ∈ dv



2.6.2 Px
0 0

Z ∞
2
= eµ(z−x)−µ (u+v)/2
Bz(27) (u, v, y)dy dudv, for Bz(27) (u, v, y) see 1.2.27.2
0

Z Hz Z Hz 
(µ)
1I[r,∞) Ws(µ) ds = 0



(1) Px 1I(−∞,r) Ws ds ∈ du,
0 0

2
(
eµ(z−x)−µ u/2
hu (1, z − x)du, x≤z
= 2
µ(z−x)−µ u/2
e ssu (r − x, r − z)du, z≤x≤r

Z Hz Z Hz 
1I(−∞,r) Ws(µ) ds = 0, 1I[r,∞) Ws(µ) ds ∈ dv



(2) Px
0 0

2
(
eµ(z−x)−µ v/2
ssv (x − r, z − r)dv, r≤x≤z
= 2
eµ(z−x)−µ v/2
hv (1, x − z)dv, z≤x

n  Z Hz
 o
2.7.1 Ex exp −γ 1I[r,u] Ws(µ) ds ; Hz < ∞ eµ(x−z)+|µ||z−x|
r<u 0
 2. STOPPING AT FIRST HITTING TIME 307

2 Ws(µ) = µs + Ws Hz = min{s : Ws(µ) = z}


 1, x≤z
( 2 + (1 − e2(x−r)|| )) sh((u − r) ) + || ch((u − r) )



, z≤x≤r

 (2 + (1 − e2(z−r)|| )) sh((u − r) ) + || ch((u − r) )



z≤r = e(x−r)|| (2 sh((u − x) ) + || ch((u − x) ))


 ( 2 + (1 − e2(z−r)|| )) sh((u − r) ) + || ch((u − r) ) , r ≤ x ≤ u

|| e||(u−r)



, u≤x

(2 + (1 − e2(z−r)|| )) sh((u − r) ) + || ch((u − r) )


|| e||(u−r)

2 2( u z )  , x≤r
( + (1 − e )) sh((u − r) ) + || ch((u − r) )



 − | |

e(u−x)|| (2 sh((x − r) ) + || ch((x − r) ))





 , r≤x≤u
u≤z = ( + (1 − e2(u−z)|| )) sh((u − r) ) + || ch((u − r) )
2
(2 + (1 − e2(u−x)|| )) sh((u − r) ) + || ch((u − r) )


, u≤x≤z

(2 + (1 − e2(u−z)|| )) sh((u − r) ) + || ch((u − r) )





1, z≤x


n  Z Hz
 o
(1) Ex exp −γ 1I[r,u] Ws(µ) ds ; Hz = ∞
r<u 0

for µ > 0, z≤r

2|| e(u−x)|| sh((x − z )||)

z≤x = ( )
2
( e r −z | | + 2 sh((r − z )||)) sh((u − r) ) + || e(r−z)|| ch((u − r) )
x≤r
2e(u−x)|| [2 ch((r − z )||) sh((x − r) ) + || sh((r − z )||) ch((x − r) )]
r≤x =
(2 e(r−z)|| + 2 sh((r − z )||)) sh((u − r) ) + || e(r−z)|| ch((u − r) )
x≤u
e2(u−x+z−r)|| [( (1 − e2(z−r)|| ) − 2 ) sh((u − r) ) + || ch((u − r) )]
u≤x =1−
(2 + (1 − e2(z−r)|| )) sh((u − r) ) + || ch((u − r) )
for µ < 0, u≤z

e2(u−z+x−r)|| [( (1 − e2(u−z)|| ) − 2 ) sh((u − r) ) + || ch((u − r) )]

x≤r =1−
(2 + (1 − e2(u−z)|| )) sh((u − r) ) + || ch((u − r) )
2e(x−r)|| [2 ch((z − u)||) sh((u − x) ) + || sh((z − u)||) ch((u − x) )]
r≤x =
(2 e(z−u)|| + 2 sh((z − u)||)) sh((u − r) ) + || e(z−u)|| ch((u − r) )
x≤u
2|| e(x−r)|| sh((z − x)||)
u≤x = ( )
2
( e z −u || + 2 sh((z − u)||)) sh((u − r) ) + || e(z−u)|| ch((u − r) )
x≤z

p
Υ := µ2 + 2γ
308 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws Hz = min{s : Ws(µ) = z}

H0
Ai 21=3 −2=3 (2 =2 + |x|)
n  Z  o

2.8.1 Ex exp −γ |Ws(µ) |ds ; H0 < ∞ = e−µx
Ai (2 )−2=3 2


0
 e(z−x) D √
 − 1 − 2α+µ2
(−x 2 )
2 2γ
√ , x≤z

D− 1 − 2α+µ2 (−z 2 )


2 Hz
 Z  

(µ) 2 2 2γ
2.9.3 Ex exp −αHz − (Ws ) ds = √
2 0 
 e(z−x) D− 1 − 2α+µ2 (x 2 )
2 2γ
z≤x

√ ,
D− 1 − 2α+µ2 (z 2 )



2 2γ

n  2Z
|x| 3=2 e−x− t=2
 2
x ch(t )
H0  o 2 
2.9.4 Ex exp − (Ws(µ) )2 ds ; H0 ∈ dt = √ 3= 2 exp − dt
2 0 2 sh (t ) 2 sh(t )
(1)

3 3 x2

H0
| x|
Z  2
(2) Px (Ws(µ) )2 ds ∈ dy, H0 ∈ dt = √ e−µx−µ t/2 esy , , t, 0, dydt
0 2 2 2 2
√
I√2 +2 =| | | 2| e x
 
 µ(z−x) z x
e
2 e z , ≥
 √
I

√
Z Hz 

2 +  2 =| |
 (µ)
  | |
2.10.3 Ex exp −αHz − γ e2βWs ds = √
K√2 +2 =| | | 2| e x

0 z x

 µ(z−x)
e , ≤
 √
K√2 +2 =| | | 2| e z



Z ∞
e−λz P0 sup ` Hz , y < h, Hz < ∞ dz



2.11.2 λ
0 y∈R

||e−h=2 ||e||h=2
=1− +
( − 2) sh(h||=2) ( − 2) sh(h||=2)M 12 + 2−  ; 1; h||

||

||e(z−x) sh((x − z ) 2 + 2 )
p
Ex e−γ Ȟµ (Hz ) ; sup Ws(µ) ∈ dy =

2.13.1 dy
sh(||(y − z )) sh((y − z ) 2 + 2 )
p
0≤s≤Hz
z<x
2 ssu (x − z; y − z )
Px Ȟµ (Hz ) ∈ du, sup Ws(µ) ∈ dy = |µ|eµ(z−x)−µ u/2


2.13.2 dudy
0≤s≤Hz sh(||(y − z ))
z<x
x<y

Ex e−αHz−γ Ȟµ (Hz ) ; sup Ws(µ) ∈ dy


2.13.4 z<x<y
0≤s≤Hz

2 + 2 e(z−x) sh((x − z ) 2 + 2 + 2 )
p p
= dy
sh((y − z ) 2 + 2 ) sh((y − z ) 2 + 2 + 2 )
p p

sup Ws(µ) ∈ dy


(1) Px Ȟµ (Hz ) ∈ du, Hz ∈ dt,
0≤s≤Hz
u<t
 2. STOPPING AT FIRST HITTING TIME 309

2 Ws(µ) = µs + Ws Hz = min{s : Ws(µ) = z}

2
= eµ(z−x)−µ t/2
ssu (x − z, y − z) st−u (y − z)dudtdy

||e(z−x) sh((z − x) 2 + 2 )
p
Ex e−γ Ĥµ (Hz ) ; Ws(µ) ∈ dy =

2.14.1 inf dy
sh(||(z − y)) sh((z − y) 2 + 2 )
p
0≤s≤Hz
x<z
2 ssu (z − x; z − y)
Ws(µ) ∈ dy = |µ|eµ(z−x)−µ u/2


2.14.2 Px Ĥµ (Hz ) ∈ du, inf dudy
0≤s≤Hz sh(||(z − y))
x<z
y<x

Ex e−αHz−γ Ĥµ (Hz ) ; Ws(µ) ∈ dy


2.14.4 inf y<x<z
0≤s≤Hz

2 + 2 e(z−x) sh((z − x) 2 + 2 + 2 )
p p
= dy
sh((z − y) 2 + 2 ) sh((z − y) 2 + 2 + 2 )
p p

Ws(µ) ∈ dy


(1) Px Ĥµ (Hz ) ∈ du, Hz ∈ dt, inf
0≤s≤Hz
u<t
2
= eµ(z−x)−µ t/2
ssu (z − x, z − y) st−u (z − y)dudtdy

sh((x − a)||)

 sh((z − a)||) , a ≤ x ≤ z

Ws(µ) , sup Ws(µ) < b = eµ(z−x)


2.15.2 Px a < inf
0≤s≤Hz 0≤s≤Hz  sh((b − x)||) , z ≤ x ≤ b

sh((b − z )||)

Ex e−αHz ; a < Ws(µ) , sup Ws(µ) < b


2.15.4 inf
0≤s≤Hz 0≤s≤Hz

sh((x − a) 2 + 2 )
 p
 , a≤x≤z
 sh((z − a) 2 + 2 )
 p
= eµ(z−x)
sh((b − x) 2 + 2 )
p
, z≤x≤b


sh((b − z ) 2 + 2 )
 p

Px Hz ∈ dt, a < inf Ws(µ) , sup Ws(µ) < b

(1)
0≤s≤t 0≤s≤t

sst (x − a, z − a)dt, a ≤ x ≤ z

µ(z−x)−µ2 t/2
=e
sst (b − x, b − z)dt, z≤x≤b

Ex e−γ`(Hz ,r) ; sup Ws(µ) < b


2.16.1
0≤s≤Hz
r<b
z<r
310 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws Hz = min{s : Ws(µ) = z}

|| sh((b − x)||)


 || sh((b − z )||) + 2 sh((b − r)||) sh((r − z )||) , r ≤ x ≤ b

= eµ(z−x)
 || sh((b − x)||) + 2 sh((b − r)||) sh((r − x)||)
, z≤x≤r
|| sh((b − z )||) + 2 sh((b − r)||) sh((r − z )||)


|| sh((b − z )||)y

 
Px `(Hz , r) ∈ dy, sup Ws(µ) < b = exp −


2.16.2
0≤s≤Hz 2 sh((b − r)||) sh((r − z )||)
r<b
z<r

|| sh((b − x)||)


 2 sh((b − r)||) sh((r − z )||) dy, r ≤ x ≤ b

×eµ(z−x)
 || sh((x − z )||) dy, z≤x≤r
2 sh2 ((r − z )||)


0, r≤x≤b
(
(µ)
sh((r − x)||)


(1) Px `(Hz , r) = 0, sup Ws <b =
eµ(z−x) , z≤x≤r
0≤s≤Hz
sh((r − z )||)

Ex e−γ`(Hz ,r) ; a < Ws(µ)


2.17.1 inf
0≤s≤Hz
a<r
r<z
|| sh((x − a)||) + 2 sh((x − r)||) sh((r − a)||)

 || sh((z − a)||) + 2 sh((z − r)||) sh((r − a)||) , r ≤ x ≤ z

µ(z−x)
=e
|| sh((x − a)||)
, a≤x≤r

|| sh((z − a)||) + 2 sh((z − r)||) sh((r − a)||)


|| sh((z − a)||)y

 
Ws(µ) = exp −


2.17.2 Px `(Hz , r) ∈ dy, a < inf
0≤s≤Hz 2 sh((z − r)||) sh((r − a)||)
a<r
r<z
|| sh((z − x)||)

dy, r≤x≤z
2 sh2 ((z − r)||)


µ(z−x)
×e
|| sh((x − a)||)
dy, a ≤ x ≤ r

2 sh((z − r)||) sh((r − a)||)


( sh((x − r)||)
eµ(z−x) , r≤x≤z
(µ) sh((z − r)||)


(1) Px `(Hz , r) = 0, a < inf Ws =
0≤s≤Hz
0, a≤x≤r
 2. STOPPING AT FIRST HITTING TIME 311

2 Ws(µ) = µs + Ws Hz = min{s : Ws(µ) = z}

Ex exp −γ`(Hz , r) − η`(Hz , u) ; Hz < ∞ eµ(x−z)+|µ||z−x|



2.18.1
r<u

for z < r
2 + || (1 − e2(x−r)|| ) + ||(1 − e2(x−u)|| ) + (1 − e2(r−u)|| )(1 − e2(x−r)|| )
z≤x = 2
 + || (1 − e2(z−r)|| ) + ||(1 − e2(z−u)|| ) + (1 − e2(r−u)|| )(1 − e2(z−r)|| )
x≤r

2 + ||(1 − e2(x−u)|| )
r≤x = 2
 + || (1 − e2(z−r)|| ) + ||(1 − e2(z−u)|| ) + (1 − e2(r−u)|| )(1 − e2(z−r)|| )
x≤u

2
u≤x = 2 2(z r )  2(z u)|| ) +  (1 − e2(r−u)|| )(1 − e2(z−r)|| )
 + || (1 − e − | | ) + ||(1 − e −

for u < z
2
x≤r = 2
 + || (1 − e2(r−z)|| ) + ||(1 − e2(u−z)|| ) + (1 − e2(u−z)|| )(1 − e2(r−u)|| )

2 + || (1 − e2(r−x)|| )
r≤x = 2
 + || (1 − e2(r−z)|| ) + ||(1 − e2(u−z)|| ) + (1 − e2(u−z)|| )(1 − e2(r−u)|| )
x≤u

2 + || (1 − e2(r−x)|| ) + ||(1 − e2(u−x)|| ) + (1 − e2(u−x)|| )(1 − e2(r−u)|| )

u≤x = 2
 + || (1 − e2(r−z)|| ) + ||(1 − e2(u−z)|| ) + (1 − e2(u−z)|| )(1 − e2(r−u)|| )
x≤z



(1) Ex exp −γ`(Hz , r) − η`(Hz , u) ; Hz = ∞
r<u

for µ > 0, z<r

2 (1 − e2(z−x)|| )
z≤x = 2
 + || (1 − e2(z−r)|| ) + ||(1 − e2(z−u)|| ) + (1 − e2(r−u)|| )(1 − e2(z−r)|| )
x≤r

2 (1 − e2(z−x)|| ) + || (1 − e2(z−r)|| )(1 − e2(r−x)|| )

r≤x = 2
 + || (1 − e2(z−r)|| ) + ||(1 − e2(z−u)|| ) + (1 − e2(r−u)|| )(1 − e2(z−r)|| )
x≤u

u≤x = 1 − e2(z−x)|µ|
2 + || (e2(r−z)|| − 1) + ||(e2(u−z)|| − 1) + (e2(u−r)|| − 1)(e2(r−z)|| − 1)
× 2
 + || (1 − e2(z−r)|| ) + ||(1 − e2(z−u)|| ) + (1 − e2(r−u)|| )(1 − e2(z−r)|| )
312 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws Hz = min{s : Ws(µ) = z}

for µ < 0, u<z

x≤r = 1 − e2(x−z)|µ|

2 + || (e2(z−r)|| − 1) + ||(e2(z−u)|| − 1) + (e2(z−u)|| − 1)(e2(u−r)|| − 1)

× 2
 + || (1 − e2(r−z)|| ) + ||(1 − e2(u−z)|| ) + (1 − e2(u−z)|| )(1 − e2(r−u)|| )

2 (1 − e2(x−z)|| ) + || (1 − e2(u−z)|| )(1 − e2(x−u)|| )

r≤x = 2
 + || (1 − e2(r−z)|| ) + ||(1 − e2(u−z)|| ) + (1 − e2(u−z)|| )(1 − e2(r−u)|| )
x≤u

2 (1 − e2(x−z)|| )
u≤x = 2
 + || (1 − e2(r−z)|| ) + ||(1 − e2(u−z)|| ) + (1 − e2(u−z)|| )(1 − e2(r−u)|| )
x≤z

Hz
ds
n  Z  o
(µ)
2.19.1 Ex exp −γ ; 0 < inf W , H < ∞
Ws() s z
0 0≤s≤Hz

 −x
e M− =||;1=2 (2||x)
 e−z M
 , 0<x≤z
=||;1=2 (2||z )

−
= −x W
e =||;1=2 (2||x)
 −z − z≤x

,
=||;1=2 (2||z )

e W−

 2Z Hz
ds
n  o
2.20.1 Ex exp − ()
2
; 0 < inf Ws(µ) , Hz < ∞
2 0 Ws 0≤s≤Hz

 e−x √xI√
2 +1=4 (||x)

 e−z √zI√ 2
 , 0<x≤z
+1=4 (||z )


= √
 e−x xK√ 2 +1=4 (||x)
, z≤x

 −z √ √
e zK 2 +1=4 (||z )


p2 q2 (Ws() )2
n  Z Hz    o
2.21.1 Ex exp − ()
2
+ ds ; 0 < inf Ws(µ) , Hz < ∞
0 2 Ws 2 0≤s≤Hz

 e−x √zM √ (qx2 )

−2 =4q; 4p2 +1=4
√ , 0<x≤z
(qz 2 )

 e−z xM 2 √ 2


− =4q; 4p +1=4
= √
 e−x zW−2 =4q;√4p2 +1=4 (qx2 )
, z≤x

 −z √
e xW−2 =4q;√4p2 +1=4 (qz 2 )

 2. STOPPING AT FIRST HITTING TIME 313

2 Ws(µ) = µs + Ws Hz = min{s : Ws(µ) = z}

p2
Hz 
q
n  Z   o
2.22.1 Ex exp − ()
2
+ () ds ; 0 < inf Ws(µ) , Hz < ∞
0 2 Ws Ws 0≤s≤Hz

 e−x M √ (2||x)
−q=||; p2 +1=4

z , 0<x≤z
 e M √ (2||z )

−
−q=||; p2 +1=4

=
 e−x W−q=||;√p2 +1=4 (2||x)
, z≤x

 −z
e W−q=||;√p2 +1=4 (2||z )


n  Z Hz

 1 o
p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds ; ` Hz , r ∈ dy



2.27.1 Ex exp −
0
2

= eµ(z−x) Q(27) 2 2
z (2p + µ , 2q + µ , y)dy, for Q(27)
z (2p, 2q) see 1.2.27.1

n  Z Hz

 o
p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds ; ` Hz , r = 0



(1) Ex exp −
0

sh((r − x) 2p + 2 )
 p
 sh((r − z )p2p + 2 ) , z ≤ x ≤ r


= eµ(z−x)
sh((x − r) 2q + 2 )
p
, r≤x≤z


sh((z − r) 2q + 2 )
 p

Z Hz Hz
1
Z 
(µ)
1I[r,∞) Ws(µ) ds ∈ dv, ` Hz , r ∈ dy




2.27.2 Px 1I(−∞,r) Ws ds ∈ du,
0 0
2

2
= eµ(z−x)−µ (u+v)/2
Bz(27) (u, v, y)dudvdy, for Bz(27) (u, v, y) see 1.2.27.2

p2
 Z Hz   
(µ) (µ)
2.30.1 Ex exp − e2βWs + qeβWs ds
0
2
 −(+ =2)x
e M−q=p| |;||=| 2p e x

| | |
2p e z
, (z − x)β ≥ 0

 e−(+ =2)z M−q=p| |;||=|


| | |
= 2p e x

 e−(+ =2)x W−q=p| |;||=| | | |
2p e z
, (z − x)β ≤ 0

 −(+ =2)z

e W−q=p| |;||=| | | |
314 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws Hz = min{s : Ws(µ) = z}

u
r


n   o
2.31.1 Ez+ |µ|
x exp −γ sup ` s, z + − α−1 ` s, z + ; Hz < ∞ exµ/|µ|
0≤s≤Hz | | ||
0<α
for 0 ≤ x ∧ r ∧ u or x ∨ r ∨ u ≤ 0
e−|u|  − e−2|r| + e−2|u| e−|x−u| − e−|x+u|



= e−|x| −
2 1 − e−2(|u|∧|r|) 1 − e−2|u−r| + ||( − e−2|r| + e−2|u| )

2 1− e−2(|u|∧|r|) 1 − e−2|u−r| − ( − e−2|r| + e−2|u| )(1− e−2|u| ) e−|x−r| − e−|x+r|

− |u|
e 2 1 − e−2(|u|∧|r|) 1 − e−2|u−r| + ||( − e−2|r| + e−2|u| ) e−|u−r| − e−|u+r|

u
r

(x−u)µ/|µ|
 
2.31.1 x exp −γ
Ez+ |µ| sup ` s, z + − α−1 ` s, z + e
0≤s≤Hz || ||
(1)

for 0 ≤ x ∧ r ∧ u or x ∨ r ∨ u ≤ 0
 − e−2|r| + e−2|u| e−|x−u| − e−|x+u|



0<α = e(x−u)µ/|µ| −
2 1 − e |u|∧|r|) 1 − e−2|u−r| + ||( − e−2|r| + e−2|u| )
−2(

2 1− e−2(|u|∧|r|) 1 − e−2|u−r| − ( − e−2|r| + e−2|u| )(1− e−2|u| ) e−|x−r| − e−|x+r|

−
2 1 − e−2(|u|∧|r|) 1 − e−2|u−r| + ||( − e−2|r| + e−2|u| ) e−|u−r| − e−|u+r|

u
r


 
2.31.2 Pz+ |µ|
x sup ` s, z + − α−1 ` s, z + > h; Hz < ∞
0≤s≤Hz || ||
0<α
for 0 ≤ x ∧ r ∧ u or x ∨ r ∨ u ≤ 0
||  − e−2|r| + e−2|u| h
h −|x−r| −|x+r|
x e −e


= exp − |u| − −
|| 2 1 − e−2(|u|∧|r|) 1 − e−2|u−r| e−|u−r| − e−|u+r|

 − e−2|r| + e−2|u|
 −|x−r| −|x+r|
(e −e )(1− e−2|u| ) −|u−x| −|x+u|i
− −e +e
2 1 − e−2(|u|∧|r|) 1 − e−2|u−r| e−|u−r| − e−|u+r|

u
r


 
2.31.2 Pz+ |µ|
x sup ` s, z + − α−1 ` s, z + >h
0≤s≤Hz || ||
(1)

for 0 ≤ x ∧ r ∧ u or x ∨ r ∨ u ≤ 0
||  − e−2|r| + e−2|u| h
h −|x−r| −|x+r|
(u − x) e −e


0<α = exp −
| | 2 1−e − 2(|u |∧|r |) 1−e−2|u −r | e−|u−r| − e−|u+r|

 − e−2|r| + e−2|u|
 −|x−r| −|x+r|
(e −e )(1− e−2|u| ) −|u−x| −|x+u|i
− −e +e
2 1 − e−2(|u|∧|r|) 1 − e−2|u−r| e−|u−r| − e−|u+r|

q
2


Υ := 1− e−2|r| − α+ αe−2|u| + 4α 1− e−2(|u|∧|r|) 1− e−2|u−r| + 1− α
 3. STOPPING AT FIRST EXIT TIME 315

2 Ws(µ) = µs + Ws H = Ha,b = min{s : Ws(µ) ∈

/ (a, b)}

3. Stopping at first exit time

e(a−x) sh (b − x) 2 + 2 + e(b−x) sh (x − a) 2 + 2
p
p

3.0.1 Ex e−αH =
sh (b − a) 2 + 2
p

2
Px (H ∈ dt) = e−µ t/2
 µ(a−x)
3.0.2 e sst (b − x, b − a)dt + eµ(b−x) sst (x − a, b − a)dt

(µ) sh((b − x)||) µ(a−x)+iβa sh((x − a)||) µ(b−x)+iβb

3.0.3 Ex eiβWH = e + e
sh((b − a)||) sh((b − a)||)
(µ) sh((b − x)||)
= a = eµ(a−x)


3.0.4 Px W H
sh((b − a)||)
(a)

(µ) sh((x − a)||)

= b = eµ(b−x)


3.0.4 Px W H
sh((b − a)||)
(b)
sh (b − x) 2 + 2

p
(µ)
Ex e−αH ; WH = a = eµ(a−x)

3.0.5
sh (b − a) 2 + 2
p

(a)

e(b−x) sh (x − a) 2 + 2

p
(µ)
Ex e−αH ; WH

3.0.5 =b =
sh (b − a) 2 + 2
p

(b)
(µ) 2
= a = eµ(a−x)−µ t/2 sst (b − x, b − a)dt


3.0.6 Px H ∈ dt, WH
(a)

(µ) 2
= b = eµ(b−x)−µ t/2 sst (x − a, b − a)dt


3.0.6 Px H ∈ dt, WH
(b)

(µ) sh((b − y)||) sh((x − a)||)

sup Ws(µ) ≥ y, WH = a = eµ(a−x)


3.1.6 Px
0≤s≤H sh((b − a)||) sh((y − a)||)
x≤y

(µ)
Ex e−αH ; sup Ws(µ) ≥ y, WH

3.1.8 =a
0≤s≤H
(1)
µ(a−x) sh (b − y) 2 + 2 sh (x − a) 2 + 2
p
p

=e
, a≤x≤y≤b
sh (b − a) 2 + 2 sh (y − a) 2 + 2
p
p
316 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws H = Ha,b = min{s : Ws(µ) ∈

/ (a, b)}

(µ)
sup Ws(µ) ≥ y, WH


(2) Px = a, H ∈ dt
0≤s≤H
2
= eµ(a−x)−µ t/2
{sst (b − y, b − a) ∗ sst (x − a, y − a)}dt, a≤x≤y≤b

(µ)
Ws(µ) ≤ y, WH


3.2.6 Px inf =b
0≤s≤H

sh((b − x)||) sh((y − a)||)

= eµ(b−x) , a≤y≤x≤b
sh((b − y)||) sh((b − a)||)

(µ)
Ex e−αH ; inf Ws(µ) ≤ y, WH

3.2.8 =b
0≤s≤H
(1)
sh (b − x) 2 + 2 sh (y − a) 2 + 2
p
p

= eµ(b−x)
, a≤y≤x≤b
sh (b − y) 2 + 2 sh (b − a) 2 + 2
p
p

(µ)
Ws(µ) ≤ y, WH


(2) Px inf = b, H ∈ dt
0≤s≤H
2
= eµ(b−x)−µ t/2
{sst (b − x, b − y) ∗ sst (y − a, b − a)}dt, a≤y≤x≤b

|| sh((b − a)||)

3.3.1 Er e−γ`(H,r) = =: Lµ
|| sh((b − a)||) + 2 sh((b − r)||) sh((r − a)||)

sh((x − r)||) sh((b − x)||)


 eµ(b−x) sh((b − r)||) + eµ(r−x) sh((b − r)||) Lµ ,
 r≤x≤b
Ex e−γ`(H,r) =
 eµ(a−x) sh((r − x)||) + eµ(r−x) sh((x − a)||) Lµ ,

a≤x≤r
sh((r − a)||) sh((r − a)||)

|| sh((b − a)||)y

 
3.3.2 Px `(H, r) ∈ dy = exp µ(r − x) −
2 sh((b − r)||) sh((r − a)||)

|| sh((b − x)||) sh((b − a)||)


 2 sh2 ((b − r)||) sh((r − a)||) dy, r ≤ x ≤ b

×
 || sh((x − a)||) sh(( b − a)||)
dy, a ≤ x ≤ r
2 sh((b − r)||) sh2 ((r − a)||)


sh((x − r)||)


 eµ(b−x) sh((b − r)||) ,
 r≤x≤b
(1) Px `(H, r) = 0 =
 eµ(a−x) sh((r − x)||) ,

a≤x≤r
sh((r − a)||)
 3. STOPPING AT FIRST EXIT TIME 317

2 Ws(µ) = µs + Ws H = Ha,b = min{s : Ws(µ) ∈

/ (a, b)}

 [e(a−r) sh((b − r) ) + e(b−r) sh((r − a) )]

3.3.3 Er e−αH−γ`(H,r) = =: Lµ
 sh((b − a) ) + 2 sh((b − r) ) sh((r − a) )

Ex e−αH−γ`(H,r)
sh((x − r) ) sh((b − x) )

 eµ(b−x) sh((b − r)

)
+ eµ(r−x)
sh((b − r)
L , r≤x≤b
) µ
=
 eµ(a−x) sh((r − x)
 )
+ eµ(r−x)
sh((x − a) )
L , a≤x≤r
sh((r − a) ) sh((r − a) ) µ

 sh((b − a) )y
 
Ex e−αH ; `(H, r) ∈ dy = exp −

3.3.4
2 sh((b − r) ) sh((r − a) )

 sh((b − x) )[e(a−x) sh((b − r) ) + e(b−x) sh((r − a) )]


dy, r ≤ x ≤ b
2 sh2 ((b − r) ) sh((r − a) )


× (a−x) sh((b − r) ) + e(b−x) sh((r − a) )]
  sh((x − a) )[e

dy, a ≤ x ≤ r
2 sh((b − r) ) sh2 ((r − a) )

sh((x − r) )

 eµ(b−x) sh((b − r) ) ,
 r≤x≤b
Ex e−αH ; `(H, r) = 0 =

(1)
 eµ(a−x) sh((r − x) ) ,

a≤x≤r
sh((r − a) )

(µ)
Ex e−γ`(H,r) ; WH = a = eµ(a−x) Va(3) (µ2 )

3.3.5
(a)

for Va(3) (2α) see 1.3.3.7

(µ)
Ex e−γ`(H,r) ; WH = b = eµ(b−x) Vb(3) (µ2 )

3.3.5
(b)

for Vb(3) (2α) see 1.3.3.7

|| sh((b − a)||)y

 
(µ)


3.3.6 Px `(H, r) ∈ dy, WH = a = exp µ(a − x) −
2 sh((b − r)||) sh((r − a)||)
(a)
|| sh((b − x)||)

 2 sh((b − r)||) sh((r − a)||) dy, r ≤ x ≤ b

×
 || sh((x − a)||)
dy, a≤x≤r
2 sh2 ((r − a)||)


p
Υα := 2α + µ2
318 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws H = Ha,b = min{s : Ws(µ) ∈

/ (a, b)}

0, r≤x≤b
(
(µ)
µ(a−x) sh((r − x)||)


(1) Px `(H, r) = 0, WH =a =
e , a≤x≤r
sh((r − a)||)

|| sh((b − a)||)y

 
(µ)


3.3.6 Px `(H, r) ∈ dy, WH = b = exp µ(b − x) −
2 sh((b − r)||) sh((r − a)||)
(b)
|| sh((b − x)||)

dy, r≤x≤b
2 sh2 ((b − r)||)


×
|| sh((x − a)||)
dy, a ≤ x ≤ r

2 sh((b − r)||) sh((r − a)||)


( sh((x − r)||)
eµ(b−x) , r≤x≤b
(µ)
sh((b − r)||)


(1) Px `(H, r) = 0, WH = b =
0, a≤x≤r

(µ)
Ex e−αH−γ`(H,r) ; WH = a = eµ(a−x) Va(3) (2α + µ2 )

3.3.7
(a)

for Va(3) (2α) see 1.3.3.7

(µ)
Ex e−αH−γ`(H,r) ; WH = b = eµ(b−x) Vb(3) (2α + µ2 )

3.3.7
(b)

for Vb(3) (2α) see 1.3.3.7

 sh((b − a) )y
 
(µ)
Ex e−αH ; `(H, r) ∈ dy, WH

3.3.8 = a = exp −
2 sh((b − r) ) sh((r − a) )
(a)
 sh((b − x) )

 2 sh((b − r) ) sh((r − a) ) dy, r ≤ x ≤ b

µ(a−x)
×e
  sh(( x − a) )
dy, a≤x≤r
2 sh2 ((r − a) )


0, r≤x≤b
(
−αH (µ)
sh((r − x) )

(1) Ex e ; `(H, r) = 0, WH = a =
eµ(a−x) , a≤x≤r
sh((r − a) )

 sh((b − a) )y
 
(µ)
Ex e−αH ; `(H, r) ∈ dy, WH

3.3.8 = b = exp −
2 sh((b − r) ) sh((r − a) )
(b)
 3. STOPPING AT FIRST EXIT TIME 319

2 Ws(µ) = µs + Ws H = Ha,b = min{s : Ws(µ) ∈

/ (a, b)}

 sh((b − x) )

dy, r≤x≤b
2 sh2 ((b − r) )


×eµ(b−x)
 sh((x − a) )
dy, a ≤ x ≤ r

2 sh((b − r) ) sh((r − a) )


( sh((x − r) )
eµ(b−x) , r≤x≤b
−αH (µ)
sh((b − r) )

(1) Ex e ; `(H, r) = 0, WH = b =
0, a≤x≤r

 Z H

3.4.1 Er exp −γ 1I[r,b] Ws(µ) ds =: Qµ
0

||e(a−r) sh((b − r) ) +  e(b−r) sh((r − a)||)

=
 ch((b − r) ) sh((r − a)||) + || sh((b − r) ) ch((r − a)||)

 Z H

Ex exp −γ 1I[r,b] Ws(µ) ds
0

sh((x − r) ) sh((b − x) )


 eµ(b−x) sh((b − r) ) + eµ(r−x) sh((b − r) ) Qµ ,
 r≤x≤b
=
 eµ(a−x) sh((r − x)||) + eµ(r−x) sh((x − a)||) Qµ , a ≤ x ≤ r

sh((r − a)||) sh((r − a)||)
n  Z H  o
(µ)
1I[r,b] Ws(µ) ds ; WH = a = eµ(a−x) Q(6) 2 2


3.4.5 Ex exp −γ a (µ , µ + 2γ)
(a) 0

for Q(6)
a (2p, 2q) see 1.3.6.5
n  Z H

 (µ)
o
3.4.5 Ex exp −γ 1I[r,b] Ws(µ) ds ; WH = b = eµ(b−x) Q(6) 2 2
b (µ , µ + 2γ)
(b) 0

for Q(6)
b (2p, 2q) see 1.3.6.5

Z H 
(µ)
1I[r,b] Ws(µ) ds ∈ dy, WH


3.4.6 Px =a
(a) 0

sh((x − a)||)e(a−x)− y=2

2

a≤x = rc (0, b − r, b − r, th((r − a)|µ|)/|µ|)dy

sh((r − a)||) ch((r − a)||) y
x≤r

p
Υs := 2s + µ2
320 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws H = Ha,b = min{s : Ws(µ) ∈

/ (a, b)}

e(a−x)− y=2
2

r≤x = rc (0, b − r, b − x, th((r − a)|µ|)/|µ|)dy

ch((r − a)||) y
x≤b

  e(a−x) sh((r − x)||) ,


H
Z a≤x≤r
(µ)
1I[r,b] Ws(µ) ds = 0, WH sh((r − a)||)


(1) Px =a =
0 0, r≤x≤b


Z H 
(µ)
1I[r,b] Ws(µ) ds ∈ dy, WH


3.4.6 Px =b
(b) 0

sh((x − a)||) µ(b−x)−µ2 y/2

a≤x = e e y (1, b − r, 0, th((r − a)|µ|)/|µ|)dy
rc
|| ch((r − a)||)
x≤r
2

r≤x = eµ(b−x)−µ y/2
ssy (x − r, b − r)
x≤b
sh((r − a)||) 
+ ssy (b − x, b − r) ∗ rc
e y (1, b − r, 0, th((r − a)|µ|)/|µ|) dy
|| ch((r − a)||)

 Z H

3.5.1 Er exp −γ 1I[a,r] Ws(µ) ds =: Qµ
0

e(a−r) 2 + 2 sh((b − r)||) + e(b−r) || sh((r − a) 2 + 2 )

p p
=
2 + 2 sh((b − r)||) ch((r − a) 2 + 2 ) + || ch((b − r)||) sh((r − a) 2 + 2 )
p p p

 Z H

Ex exp −γ 1I[a,r] Ws(µ) ds
0

µ(b−x) sh((x − r )||) sh((b − x)||)


e + eµ(r−x) Q , r≤x≤b
sh((b − r)||) sh((b − r)||) µ

= 2 2
 eµ(a−x) sh((r − x)p2 +  ) + eµ(r−x) sh((x − a)p 2 +  ) Qµ , a ≤ x ≤ r
p p

sh((r − a) 2 + 2 ) sh((r − a) 2 + 2 )
n  Z H  o
(µ)
1I[a,r] Ws(µ) ds ; WH = a = eµ(a−x) Q(6) 2 2


3.5.5 Ex exp −γ a (µ + 2γ, µ )
(a) 0

for Q(6)
a (2p, 2q) see 1.3.6.5
n  Z H

 (µ)
o
3.5.5 Ex exp −γ 1I[a,r] Ws(µ) ds ; WH = b = eµ(b−x) Q(6) 2 2
b (µ + 2γ, µ )
(b) 0

for Q(6)
b (2p, 2q) see 1.3.6.5
 3. STOPPING AT FIRST EXIT TIME 321

2 Ws(µ) = µs + Ws H = Ha,b = min{s : Ws(µ) ∈

/ (a, b)}

Z H 
(µ)
1I[a,r] Ws(µ) ds ∈ dy, WH


3.5.6 Px =a
(a) 0

2

a≤x = eµ(a−x)−µ y/2
ssy (r − x, r − a)
x≤r
sh((b − r)||)
+ ssy (x − a, r − a) ∗ rc
e y (1, r − a, 0, th((b − r)|µ|)/|µ|)dy
|| ch((b − r)||)
sh((b − x)||) µ(a−x)−µ2 y/2 
r≤x = e e y (1, r − a, 0, th((b − r)|µ|)/|µ|) dy
rc
|| ch((b − r)||)
x≤b

Z H 
(µ)
1I[a,r] Ws(µ) ds ∈ dy, WH


3.5.6 Px =b
(b) 0

e(b−x)− y=2
2

a≤x = rc (0, r − a, x − a, th((b − r)|µ|)/|µ|)dy

ch((b − r)||) y
x≤r
sh((b − x)||)e(b−x)− y=2
2

r≤x = rc (0, r − a, r − a, th((b − r)|µ|)/|µ|)dy

sh((b − r)||) ch((b − r)||) y
x≤b

Z H  0, a≤x≤r
(µ) (µ)
= b = e(b−x) sh((x − r)||)


(1) Px 1I[a,r] Ws ds = 0, WH
, r≤x≤b
0
sh((b − r)||)


n  Z H
 o
(µ)
3.5.7 Ex exp −αH − γ 1I[a,r] Ws(µ) ds ; WH =a
(a) 0

= eµ(a−x) Q(6) 2
a (2α + µ + 2γ, 2α + µ ),
2
for Q(6)
a (2p, 2q) see 1.3.6.5
n  Z H
 o
(µ)
3.5.7 Ex exp −αH − γ 1I[a,r] Ws(µ) ds ; WH =b
(b) 0

= eµ(b−x) Q(6) 2 2
b (2α + µ + 2γ, 2α + µ ), for Q(6)
b (2p, 2q) see 1.3.6.5

 Z H


p1I[a,r] Ws(µ) + q1I[r,b] Ws(µ) ds =: Q


3.6.1 Er exp −
0
e(a−r) p sh((b − r)q ) + e(b−r) q sh((r − a)p )
=
q ch((b − r)q ) sh((r − a)p ) + p sh((b − r)q ) ch((r − a)p )

n  Z H

o
p1I[a,r] Ws(µ) + q1I[r,b] Ws(µ) ds


Ex exp −
0
322 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws H = Ha,b = min{s : Ws(µ) ∈

/ (a, b)}

sh((x − r)q ) sh((b − x)q )
 eµ(b−x) sh((b − r) ) + eµ(r−x) sh((b − r) ) Q,
 r≤x≤b
q q
=
 µ(a−x) sh((r − x)p ) µ(r−x) sh((x − a)p )
e +e Q, a≤x≤r
sh((r − a)p ) sh((r − a)p )


n  Z H

 o
(µ)
p1I[a,r] Ws(µ) + q1I[r,b] Ws(µ) ds ; WH


3.6.5 Ex exp − =a
(a) 0

= eµ(a−x) Q(6) 2 2
a (2p + µ , 2q + µ ), for Q(6)
a (2p, 2q) see 1.3.6.5
n  Z H  o
(µ)
p1I[a,r] Ws(µ) + q1I[r,b] Ws(µ) ds ; WH




3.6.5 Ex exp − =b
(b) 0

= eµ(b−x) Q(6) 2 2
b (2p + µ , 2q + µ ), for Q(6)
b (2p, 2q) see 1.3.6.5

Z H Z H 
(µ) (µ)
1I[r,b] Ws(µ) ds ∈ dv, WH



3.6.6 Px 1I[a,r) Ws ds ∈ du, =a
(a) 0 0
Z ∞
2
u/2−µ2 v/2
= eµ(a−x)−µ Ba(27) (u, v, y)dy dudv
0

for Ba(27) (u, v, y) see 1.3.27.6(a)

Z H Z H 
(µ)
1I[a,r) Ws(µ) ds ∈ du, 1I[r,b] Ws(µ) ds ∈ dv, WH



3.6.6 Px =b
(b) 0 0
Z ∞
2
u/2−µ2 v/2
= eµ(b−x)−µ Bb(27) (u, v, y)dy dudv
0

(27)
for Bb (u, v, y) see 1.3.27.6(b)

 Z H 
3.8.1 Ex exp −γ Ws(µ) ds
0

e(a−x)  2b  2x S1=3 ( 31  3b ; 31  3x ) + e(b−x)  2x  2a S1=3 ( 31  3x ; 31  3a )

=
 2b  2a S1=3 ( 31  3b ; 31  3a )

H o e(a−x)  2 S1=3 ( 1  3 ; 1  3 )
x x
Z
3 b 3
n  
(µ)
3.8.5 Ex exp −γ Ws(µ) ds ; WH =a =
0  2a S1=3 ( 31  3b ; 31  3a )
(a)

H o e(b−x)  2 S1=3 ( 1  3 ; 1  3 )
x x 3 a
Z
3
n  
(µ)
3.8.5 Ex exp −γ Ws(µ) ds ; WH =b =
0  2b S1=3 ( 31  3b ; 31  3a )
(b)
 3. STOPPING AT FIRST EXIT TIME 323

2 Ws(µ) = µs + Ws H = Ha,b = min{s : Ws(µ) ∈

/ (a, b)}

 2Z H 
3.9.1 Ex exp − (Ws(µ) )2 ds
2 0

ea+(a −x ) =2 S 12 + 2 ; b 2 ; x 2 + eb+(b −x ) =2 S 12 + 2 ; x 2 ; a 2
2 2 2 √ √
2 2 2 √ √

=
ex S 12 + 2 ; b 2 ; a 2
2 √ √

2 H
n  Z  o
(µ)
3.9.7 Ex exp −αH − (Ws(µ) )2 ds ; WH =a
2 0
(a)

S 12 + 2 2+ ; b 2 ; x 2
2 √ √

2
−x2 )γ/2
= eµ(a−x)+(a
S 12 + 2 2+ ; b 2 ; a 2
2 √ √

2 H
n  Z  o
(µ)
3.9.7 Ex exp −αH − (Ws(µ) )2 ds ; WH =b
2 0
(b)

S 12 + 2 2+ ; x 2 ; a 2
2 √ √

2
−x2 )γ/2
= eµ(b−x)+(b
S 12 + 2 2+ ; b 2 ; a 2
2 √ √

 Z H 
(µ)
3.10.1 Ex exp −γ e2βWs ds
0
√
2 e b ; √2 e x + e(−|| sign )b S
√
e(−|| sign )a S||=| 2 e x ; √2 e a
 
| | | | | ||=| | | | | |
= √
2 √
2
e(  sign )x b
S||=| | | | e ; | | e a

−| |

√ √
o e(−α =| |)a S Υα | 2| e b ; | 2| e x

n  Z H 
2βWs(µ) (µ) |β|
3.10.7 Ex exp −αH −γ e ds ;WH = a = √ √
e(−α =| |)x S Υα | 2| e b ; | 2| e a

(a) 0
|β|

√ √
o e(−α =| |)b S Υα | 2| e x ; | 2| e a

n  Z H 
2βWs(µ) (µ) |β|
3.10.7 Ex exp −αH −γ e ds ;WH = b = √ √
e(−α =| |)x S Υα | 2| e b ; | 2| e a

(b) 0
|β|

||e(a−x) sh((x − a) 2 + 2 )
Z b−a p
(µ)
Ex e−γ Ȟµ (H) ; WH

3.12.5 =a = dy
sh(||y) sh(y 2 + 2 )
p
(a) x−a

p
Υs := 2s + µ2
324 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws H = Ha,b = min{s : Ws(µ) ∈

/ (a, b)}

||e(b−x) sh((b − x) 2 + 2 )
Z b−a p
−γ Ĥµ (H) (µ)

3.12.5 Ex e ; WH = b = dy
sh(||y) sh(y 2 + 2 )
p
(b) b−x

b−a
ssu (x − a; y)
Z
(µ) 2
= a = |µ|eµ(a−x)−µ u/2


3.12.6 Px Ȟµ (H) ∈ du, WH dy du
x−a
sh(||y)
(a)
b−a
ssu (b − x; y)
Z
(µ) 2
= b = |µ|eµ(b−x)−µ u/2


3.12.6 Px Ĥµ (H) ∈ du, WH dy du
b−x
sh(||y)
(b)

b−a p
2 + 2 ea sh((x − a) 2 + 2 + 2 )
Z p
(µ)
Ex e−αH−γ Ȟµ (H) ; WH

3.12.7 =a = dy
e sh(y 2 + 2 ) sh(y 2 + 2 + 2 )
x
p p
(a) x−a

b−a p
2 + 2 eb sh((b − x) 2 + 2 + 2 )
Z p
−αH−γ Ĥµ (H) (µ)

3.12.7 Ex e ; WH = b = dy
ex sh(y 2 + 2 ) sh(y 2 + 2 + 2 )
p p
(b) b−x

(µ)


3.12.8 Px Ȟµ (H) ∈ du, H ∈ dt, WH =a u<t
(a)
Z b−a
2
= eµ(a−x)−µ t/2
ssu (x − a, y) st−u (y)dy dudt
x−a

(µ)


3.12.8 Px Ĥµ (H) ∈ du, H ∈ dt, WH =b u<t
(b)
Z b−a
µ(b−x)−µ2 t/2
=e ssu (b − x, y) st−u (y)dy dudt
b−x

(µ)


3.18.5 Ex exp −γ`(H, r) − η`(H, u) ; WH =a
(a)
2 s(b; x) + 2|| s(b; r)s(r; x) + 2||s(b; u)s(u; x) + 4 s(b; u)s(u; r)s(r; x)
a≤x = eµ(a−x) 2
 s(b; a) + 2|| s(b; r)s(r; a) + 2||s(b; u)s(u; a) + 4 s(b; u)s(u; r)s(r; a)
x≤r
2 s(b; x) + 2||s(b; u)s(u; x)
r≤x = eµ(a−x) 2
 s(b; a) + 2|| s(b; r)s(r; a) + 2||s(b; u)s(u; a) + 4 s(b; u)s(u; r)s(r; a)
x≤u
2 s(b; x)
u≤x = eµ(a−x) 2
 s(b; a) + 2|| s(b; r)s(r; a) + 2||s(b; u)s(u; a) + 4 s(b; u)s(u; r)s(r; a)
x≤b

s(y, x) := sh((y − x)|µ|)

 3. STOPPING AT FIRST EXIT TIME 325

2 Ws(µ) = µs + Ws H = Ha,b = min{s : Ws(µ) ∈

/ (a, b)}

(µ)


3.18.5 Ex exp −γ`(H, r) − η`(H, u) ; WH =b
(b)
2 s(x; a)
a≤x = eµ(b−x) 2
 s(b; a) + 2|| s(b; r)s(r; a) + 2||s(b; u)s(u; a) + 4 s(b; u)s(u; r)s(r; a)
x≤r
2 s(x; a) + 2|| s(x; r)s(r; a)
r≤x = eµ(b−x) 2
 s(b; a) + 2|| s(b; r)s(r; a) + 2||s(b; u)s(u; a) + 4 s(b; u)s(u; r)s(r; a)
x≤u
2 s(x; a) + 2|| s(x; r)s(r; a) + 2||s(x; u)s(u; a) + 4 s(x; u)s(u; r)s(r; a)
u≤x = eµ(b−x) 2
 s(b; a) + 2|| s(b; r)s(r; a) + 2||s(b; u)s(u; a) + 4 s(b; u)s(u; r)s(r; a)
x≤b

xe−(+||)x S (1 + =||; 2; 2||b; 2||x)

H
ds
n  Z 
(µ)
3.19.5 Ex exp −γ ()
; WH = a = −(+||)a
0 Ws ae S (1 + =||; 2; 2||b; 2||a)
(a)
0<a

xe−(+||)x S (1 + =||; 2; 2||x; 2||a)

H
ds
n  Z 
(µ)
3.19.5 Ex exp −γ ()
; WH = b = −(+||)b
0 Ws be S (1 + =||; 2; 2||b; 2||a)
(b)
0<a
√
n  2Z H
ds

(µ)
e−x x 2+1=4 S√ 2+1=4 (||b; ||x)
3.20.5 Ex exp − ; W = a = √
2 Ws() 2 H
e−a a 2+1=4 S√ 2+1=4 (||b; ||a)


(a) 0

0<a
√
n  2Z H
ds

(µ)
e−x x 2+1=4 S√ 2+1=4 (||x; ||a)
3.20.5 Ex exp −
; WH = b = √
2 0 Ws() 2 e−b b 2+1=4 S√ 2+1=4 (||b; ||a)
(b)
0<a

p2 q2 (Ws() )2
n  Z H  
(µ)
3.21.5 Ex exp − ()
2
+ ds ; WH =a
0 2 Ws 2
(a)
√ √ √
e−x−qx =2 x 4p2+1=4q S 4q + 4p4 +1 + 12 ; 4p2 +1 + 1; qb2 ; qx2
2
 2 2 2


0<a = √ √ √ , a≤x≤b
e−a−qa2 =2 a 4p2+1=4q S 4q + 4p4 +1 + 12 ; 4p2 +1 + 1; qb2 ; qa2
2 2
 2 

p2 q2 (Ws() )2
n  Z H  
(µ)
3.21.5 Ex exp − ()
2
+ ds ; WH =b
0 2 Ws 2
(b)

s(y, x) := sh((y − x)|µ|)

326 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws H = Ha,b = min{s : Ws(µ) ∈

/ (a, b)}

√ √ √
e−x−qx =2 x 4p2+1=4q S 4q + 4p4 +1 + 12 ; 4p2 +1 + 1; qx2 ; qa2
2
 2 2 2


0<a = √ √ √ , a≤x≤b
e−b−qb2 =2 b 4p2+1=4q S 4q + 4p4 +1 + 12 ; 4p2 +1 + 1; qb2 ; qa2
2 2
 2 

p2
H
q
n  Z  
(µ)
3.22.5 Ex exp − ()
2
+ () ds ; WH =a
(a) 0 2 Ws Ws

√
e−(+||)x x p2+1=4+1=2 S p2 + 1=4 + 12 + |q | ; 8p + 1 + 1; 2||b; 2||x
p √

0<a = √
e−(+||)a a p2+1=4+1=2 S p2 + 1=4 + 12 + |q | ; 8p + 1 + 1; 2||b; 2||a
p √

p2
H
q
n  Z  
(µ)
3.22.5 Ex exp − ()
2
+ () ds ; WH =b
(b) 0 2 Ws Ws

√
e−(+||)x x p2+1=4+1=2 S p2 + 1=4 + 12 + |q | ; 8p + 1 + 1; 2||x; 2||a
p √

0<a = √
e−(+||)b b p2+1=4+1=2 S p2 + 1=4 + 12 + |q | ; 8p + 1 + 1; 2||b; 2||a
p √

n  Z H

 1 o
p1I[a,r) Ws(µ) + q1I[r,b] Ws(µ) ds ; `(H, r) ∈ dy, WH = a


3.27.5 Ex exp −
0
2
(a)
2q + 2 ch((b − r) 2q + 2 ) 2p + 2 ch((r − a) 2p + 2 ) 
 p p p p 
= exp −y +
sh((b − r) 2q + 2 ) sh((r − a) 2p + 2 )
p p

2p + 2 sh((b − x) 2q + 2 )
 p p
 sh((b − r) 2q + 2 ) sh((r − a)p2p + 2 ) dy, r ≤ x ≤ b

 p
µ(a−x)
×e
2p + 2 sh((x − a) 2p + 2 )
p p
a≤x≤r

dy,
sh2 ((r − a) 2p + 2 )

 p

n  Z H

 o
p1I[a,r) Ws(µ) + q1I[r,b] Ws(µ) ds ; `(H, r) = 0, WH = a


(1) Ex exp −
0

µ(a−x) sh((r − x) 2p + 2 )
p
=e , a≤x≤r
sh((r − a) 2p + 2 )
p

n  Z H

 1 o
p1I[a,r) Ws(µ) + q1I[r,b] Ws(µ) ds ; `(H, r) ∈ dy, WH = b


3.27.5 Ex exp −
0
2
(b)
2q + 2 ch((b − r) 2q + 2 ) 2p + 2 ch((r − a) 2p + 2 ) 
 p p p p 
= exp −y +
sh((b − r) 2q + 2 ) sh((r − a) 2p + 2 )
p p
 3. STOPPING AT FIRST EXIT TIME 327

2 Ws(µ) = µs + Ws H = Ha,b = min{s : Ws(µ) ∈

/ (a, b)}

2q + 2 sh((b − x) 2q + 2 )
 p p
dy, r≤x≤b
sh2 ((b − r) 2q + 2 )

 p

×eµ(b−x)
2q + 2 sh((x − a) 2p + 2 )
p p
dy, a ≤ x ≤ r


sh((b − r) 2q + 2 ) sh((r − a) 2p + 2 )
 p p

n  Z H

 o
p1I[a,r) Ws(µ) + q1I[r,b] Ws(µ) ds ; `(H, r) = 0, WH = b


(1) Ex exp −
0

sh((x − r) 2q + 2 )
p
= eµ(b−x) , r≤x≤b
sh((b − r) 2q + 2 )
p

Z H H
1
Z 
(µ) (µ)
1I[r,b] Ws(µ) ds ∈ dv, `(H, r ∈ dy, WH




3.27.6 Px 1I[a,r) Ws ds ∈ du, =a
0 0
2
(a)
2
u/2−µ2 v/2
= eµ(a−x)−µ Ba(27) (u, v, y)dudvdy

for Ba(27) (u, v, y) see 1.3.27.6(a)

Z H H
1
Z 
(µ) (µ)
1I[r,b] Ws(µ) ds ∈ dv, ` H, r ∈ dy, WH




3.27.6 Px 1I[a,r) Ws ds ∈ du, =b
0 0
2
(b)
2
u/2−µ2 v/2
= eµ(b−x)−µ Bb(27) (u, v, y)dudvdy

for Bb(27) (u, v, y) see 1.3.27.6(b)

p2
n  Z H  
(µ) (µ) (µ)
3.30.5 Ex exp − e2βWs + qeβWs ds ; WH =a
0
2
(a)

e(|| sign −)x− |p| e x S | |

exp ||
+ p|q | + 12 ; 2|||| + 1; |2p| e b ; |2p| e x
   

=
e(|| sign −)a exp − | p | e a S |||| + p|q | + 12 ; 2|||| + 1; |2p| e b ; |2p| e a
   

p2
n  Z H  
(µ) (µ) (µ)
3.30.5 Ex exp − e2βWs + qeβWs ds ; WH =b
0
2
(b)

e(|| sign −)x exp || 2||

− | p | e x S | | + p|q | + 12 ; | | + 1; |2p| e x ; |2p| e a
   

=
e(|| sign −)b exp − | p | e b S |||| + p|q | + 12 ; 2|||| + 1; |2p| e b ; |2p| e a
   
328 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws H = Ha,b = min{s : Ws(µ) ∈

/ (a, b)}

u+a
r + a


 
3.31.2 P x+a sup ` s, − α−1 ` s, >h ∆ = b−a
|µ|
0≤s≤H | | ||
0<α a , b
|µ| |µ|


(u − x) ||( + sh r sh(
− r) − sh u sh(
− u))h h sh(x ∧ r) sh(
− x ∨ r)
=exp −
|| 4 sh |u − r| sh(u ∧ r) sh(
− u ∨ r) sh(u ∧ r) sh(
− u ∨ r)

 + sh r sh(
− r) − sh u sh(
− u) sh(x ∧ r) sh(
− x ∨ r) sh u sh(
− u)


−
2 sh |u − r| sh(u ∧ r) sh(
− u ∨ r) sh(u ∧ r) sh(
− u ∨ r) sh

sh(x ∧ u) sh(
− x ∨ u) i
−
sh

u+a
r + a

a
  xµ
(µ)
3.31.6 P x+a sup ` s, − α−1 ` s, > h, WH = e |µ|
|µ|
0≤s≤H | | || a , b ||
(a) a , b
|µ| |µ|
|µ| |µ|

sh(
− u) 
||( + sh r sh(
− r) − sh u sh(
− u))h sh(x ∧ r) sh(
− x ∨ r)
h
0<α = exp −
sh
4 sh |u − r| sh(u ∧ r) sh(
− u ∨ r) sh(u ∧ r) sh(
− u ∨ r)

 + sh r sh(
− r) − sh u sh(
− u) sh(x ∧ r) sh(
− x ∨ r) sh u sh(
− u)


−
2 sh |u − r| sh(u ∧ r) sh(
− u ∨ r) sh(u ∧ r) sh(
− u ∨ r) sh

sh(x ∧ u) sh(
− x ∨ u) i
−
sh

u+a
r + a

b
  xµ
(µ)
3.31.6 P x+a sup ` s, − α−1 ` s, > h, WH = e |µ|
|µ|
0≤s≤H | | || a , b ||
(b) a , b
|µ| |µ|
|µ| |µ|

sh u 

 ||( + sh r sh(
− r) − sh u sh(
− u))h h sh(x ∧ r) sh(
− x ∨ r)
0<α = exp −
sh
|| 4 sh |u − r| sh(u ∧ r) sh(
− u ∨ r) sh(u ∧ r) sh(
− u ∨ r)

 + sh r sh(
− r) − sh u sh(
− u) sh(x ∧ r) sh(
− x ∨ r) sh u sh(
− u)


−
2 sh |u − r| sh(u ∧ r) sh(
− u ∨ r) sh(u ∧ r) sh(
− u ∨ r) sh

sh(x ∧ u) sh(
− x ∨ u) i
−
sh

(sh r sh(
− r) − sh u sh(
− u))2 + 4 sh
sh |u − r| sh(u ∧ r) sh(
− u ∨ r)

p
Υ :=
 4. STOPPING AT INVERSE LOCAL TIME 329

2 Ws(µ) = µs + Ws % = %(v, z) = min{s : `(s, z) = v}

4. Stopping at inverse local time

√ 2
Ex e−α% = Ex e−α% ; % < ∞ = eµ(z−x)−(|x−z|+v) 2α+µ

4.0.1

|x − z | + v 2 t (|x − z | + v)2
 
4.0.2 Px % ∈ dt = √ 3=2 exp µ(z − x) − − dt
2t 2 2t

Px % = ∞ = 1 − eµ(z−x)−|µ|(|z−x|+v)


(1)

Px sup Ws(µ) < y, % < ∞

4.1.2
0≤s≤%
z<y
| | v
  
 exp (µ − |µ|)(z − x) − , x≤z
1 − e−2||(y−z)

=
 sh(||(y − x)) exp µ(z − x) − ||v
 
, z≤x≤y
1 − e−2||(y−z)

sh(||(y − z ))

Ex e−α% ; sup Ws(µ) < y


4.1.4
0≤s≤%

v 2 + 2
  p p 
2
 exp µ(z − x) − (z − x) 2α + µ − √ , x≤z≤y

1 − e−2(y−z) 2 +2

=
sh((y − x) 2 + 2 ) v 2 + 2
p  p 
exp µ(z − x) − z≤x≤y


2
√ ,
sh((y − z ) 2 +  ) 1 − e−2(y−z) 2 +
 p
2

Px sup Ws(µ) < y, % ∈ dt

(1)
0≤s≤%

est 0, 0, y − z, z − x + v/2, v/2 dt, x ≤ z ≤ y

µ(z−x)−µ2 t/2
=e

esst y − x, y − z, v/2, v/2 dt, z≤x≤y

inf Ws(µ) > y, % < ∞

4.2.2 Px
0≤s≤%
y<z
sh(||(x − y)) ||v
  
exp µ(z − x) − , y≤x≤z
sh(||(z − y)) 1 − e−2||(z−y)


=
||v
 
 exp (µ + |µ|)(z − x) −

, z≤x
1 − e−2||(z−y)
330 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws % = %(v, z) = min{s : `(s, z) = v}

Ex e−α% ; inf Ws(µ) > y


4.2.4
0≤s≤%

sh((x − y) 2 + 2 ) v 2 + 2
 p  p 

 sh((z − y) 2 + 2 ) exp µ(z − x) − √ , y≤x≤z
1 − e−2(z−y) 2 +2
 p
=
v 2 + 2
 p 
p
 exp µ(z − x) − (x − z) 2α + µ2 − y≤z≤x

 √ ,
1 − e−2(z−y) 2 +2

inf Ws(µ) > y, % ∈ dt

(1) Px
0≤s≤%

esst x − y, z − y, v/2, v/2 dt, y≤x≤z

µ(z−x)−µ2 t/2
=e

est 0, 0, z − y, x − z + v/2, v/2 dt, y ≤ z ≤ x

Ex e−γ`(%,r) ; % < ∞ = eµ(z−x) Vz(3) (µ2 )


4.3.1

for Vz(3) (2α) see 1.4.3.3

Ex e−γ`(%,r) ; % = ∞

(1) for µ(z − r) < 0
 ||  ||(|| + )v
 
1 − exp − , x ≤ z ≤ r, r ≤ z ≤ x
 || + || + (1 − e−2|||z−r| )


||e−2|||z−x| || + (1 − e−2|||x−r| )

||




 −
|| + (|| + ) || + (1 − e−2|||z−r| )





||(|| + )v

  
= × exp − , r∧z ≤x≤r∨z
 || + (1 − e−2|||z−r| )
|| + (1 − e−2|||x−r| ) ||e−2|||z−x|



−

 + ( + ) | + (1 − e−2|||z−r| )






 | | | | |
||(|| + )v

  
× exp − z ≤ r ≤ x, x ≤ r ≤ z

,
|| + (1 − e−2|||z−r| )


Px `(%, r) ∈ dy, % < ∞ = eµ(z−x) Fz(3) (µ2 , y)dy

4.3.2

for Fz(3) (2α, y) see 1.4.3.4

(1) Px `(%, r) = 0, % < ∞
 −|µ||z−x|
e , x ≤ z ≤ r, r≤z≤x
||v

sh(|||x − r|)
 
= exp µ(z − x) − , z∧r ≤x≤r∨z
1 − e−2|||z−r| 
 sh(|||z − r|)
0, z ≤ r ≤ x, x≤r≤z
 4. STOPPING AT INVERSE LOCAL TIME 331

2 Ws(µ) = µs + Ws % = %(v, z) = min{s : `(s, z) = v}

4.3.3 Ex e−α%−γ`(%,r) = eµ(z−x) Vz(3) (2α + µ2 )

for Vz(3) (2α) see 1.4.3.3

Ex e−α% ; `(%, r) ∈ dy = eµ(z−x) Fz(3) (2α + µ2 , y)dy


4.3.4

for Fz(3) (2α, y) see 1.4.3.4

v 2 + 2
 p 
−α%

(1) Ex e ; `(%, r) = 0 = exp µ(z − x) − √
1 − e−2|z−r| 2 +2
 −|z−x|√2α+µ2
 e , x ≤ z ≤ r, r ≤ z ≤ x
2

× sh(|x − r|p2 +  ) , r ∧ z ≤ x ≤ r ∨ z
 p

2
 sh(|z − r| 2 +  )


0, z ≤ r ≤ x, x ≤ r ≤ z

2
Px `(%, r) ∈ dy, % ∈ dt = eµ(z−x)−µ t/2 Fz(3) (y, t)dydt


(2)

for Fz(3) (y, t) see 1.4.3.4

n  Z %
 o
4.4.1 Ex exp −γ 1I[r,∞) Ws(µ) ds ; % < ∞ = eµ(z−x) Q(6) 2 2
z (µ , µ + 2γ)
0

for Q(6)
z (2p, 2q) see 1.4.6.1

Z % 
1I[r,∞) Ws(µ) ds = 0, % < ∞


4.4.2 Px
(1) 0

 −(z−x)|µ|
e , x≤z
v||e||(r−z) sh((r − x)||)
  

z≤r = exp µ(z − x) − , z≤x≤r
sh((r − z )||)  sh((r − z )||)
0, r≤x
n  Z %  o
1I(−∞,r] Ws(µ) ds ; % < ∞ = eµ(z−x) Q(6) 2 2


4.5.1 Ex exp −γ z (µ + 2γ, µ )
0

for Q(6)
z (2p, 2q) see 1.4.6.1
 %
Z 
1I(−∞,r] Ws(µ) ds = 0, % < ∞


4.5.2 Px
(1) 0
332 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws % = %(v, z) = min{s : `(s, z) = v}


 0, x≤r
v||e||(z−r)
  sh((x − r)||)

r≤z = exp µ(z − x) − , r≤x≤z
sh((z − r)||) 

sh((z − r)||)
 −(x−z)|µ|
e , z≤x

 Z %


p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds


4.6.1 Ex exp −
0

= eµ(z−x) Q(6) 2 2
z (2p + µ , 2q + µ ), for Q(6)
z (2p, 2q) see 1.4.6.1

Z % Z % 
1I(−∞,r) Ws(µ) ds ∈ du, 1I[r,∞) Ws(µ) ds ∈ dg



4.6.2 Px
0 0
Z ∞
2
u/2−µ2 g/2
= eµ(z−x)−µ Bz(27) (u, g, y)dy dudg
0

(27)
for Bz (u, g, y) see 1.4.27.2

n  Z %
 o
4.7.1 Ex exp −γ 1I[r,u] Ws(µ) ds ; % < ∞ eµ(x−z)+|µ||z−x|
r<u 0

v||[(2 + ) sh((u − r) ) + || ch((u − r) )]

 
z≤r = exp − 2
( + (1 − e2(z−r)|| )) sh((u − r) ) + || ch((u − r) )

 1, x≤z
 2
( + (1 − e2(x−r)|| )) sh((u − r) ) + || ch((u − r) )


, z≤x≤r

 (2 + (1 − e2(z−r)|| )) sh((u − r) ) + || ch((u − r) )



× e(x−r)|| (2 sh((u − x) ) + || ch((u − x) ))
2 + (1 − e2(z−r)|| )) sh((u − r) ) + || ch((u − r) , r≤x≤u
( )




|| e||(u−r)



, u≤x

(2 + (1 − e2(z−r)|| )) sh((u − r) ) + || ch((u − r) )


v||[(2 + ) sh((u − r) ) + || ch((u − r) )]

 
u≤z = exp − 2 2(u z ) 
( + (1 − e − | | )) sh((u − r) ) + || ch((u − r) )

|| e||(u−r)

, x≤r
(2 + (1 − e2(u−z)|| )) sh((u − r) ) + || ch((u − r) )




e(u−x)|| (2 sh((x − r) ) + || ch((x − r) ))




, r≤x≤u
( + (1 − e2(u−z)|| )) sh((u − r) ) + || ch((u − r)

× 2 )
(2 + (1 − e2(u−x)|| )) sh((u − r) ) + || ch((u − r) )


, u≤x≤z

(2 + (1 − e2(u−z)|| )) sh((u − r) ) + || ch((u − r) )





1, z≤x


p
Υs := 2s + µ2
 4. STOPPING AT INVERSE LOCAL TIME 333

2 Ws(µ) = µs + Ws % = %(v, z) = min{s : `(s, z) = v}

n  Z %(v,0)  o
4.8.1 Ex exp −γ |Ws(µ) |ds ; %(v, 0) < ∞
0
(2 )1=3 Ai0 (2 )−2=3 2 v  Ai 21=3 −2=3 (2 =2 + |x|)



= exp −µx −
Ai (2 )−2=3 2 Ai (2 )−2=3 2

2 %
 Z 
4.9.3 Ex exp −α% − (Ws(µ) )2 ds
2 0

√
D− 1 − 2α+µ2 (−x 2 ) −1
2 2γ
 √
v  (1=2 + (2 + 2 )=2 ) 
x≤z = √ exp µ(z − x) − √ √
D− 1 − 2α+µ2 (−z 2 ) D− 1 − 2α+µ2 (−z 2 )D− 1 − 2α+µ2 (z 2 )
2 2γ 2 2γ 2 2γ

√
D− 1 − 2α+µ2 (x 2 ) −1
2 2γ
 v 
√
(1=2 + (2 + 2 )=2 ) 
z≤x = √ exp µ(z − x) − √ √
D− 1 − 2α+µ2 (z 2 ) D− 1 − 2α+µ2 (−z 2 )D− 1 − 2α+µ2 (z 2 )
2 2γ 2 2γ 2 2γ

 Z % 
(µ)
4.10.3 Ex exp −α% − γ e2βWs ds eµ(x−z)
0
√
I  α =| 2 e x

v| | z−x

|
 
| |
 exp −  , ≥0
 √ √ √
2 e z 2 e z K 2 e z

 I  α =| 2I
 
 α =| | | |  α =| | | |

| | |
= √
2 x
 Kα =| |  √| | e 

v| | z−x

  
2 exp − √
2
√
2
 , ≤0
z z z

K =| | e 2I =| | e K  =| | e
 
α | | α | | α | |

∞
sh((h − v)||=2) 
Z
e−λz P0 sup ` %, y < h, % < ∞ dz =



4.11.2 λ sh((h − v)|µ|/2)
0 y∈R sh2 (h||=2)

||e−(h−v)=2 ||e||(h−v)=2 M 12 + 2−  ; 1; v ||


||
− + 1I[v,∞) (h)
( − 2) ( − 2)M 12 + 2−  ; 1; h||

||

sh2 ((h − v)||=2)

(1) P0 sup ` %(v, 0), y < h, %(v, 0) < ∞ = 1I[v,∞) (h)
y∈R sh2 (h||=2)

Px a < inf Ws(µ) , sup Ws(µ) < b

4.15.2
0≤s≤% 0≤s≤%

sh(||(b − x)) || sh(||(b − a))v

  
 sh(||(b − z )) exp µ(z − x) − 2 sh(||(b − z )) sh(||(z − a)) ,
 z≤x≤b
=
 sh(||(x − a)) exp µ(z − x) − || sh(||(b − a))v
 
, a≤x≤z

sh(||(z − a)) 2 sh(||(b − z )) sh(||(z − a))
334 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws % = %(v, z) = min{s : `(s, z) = v}

Px sup |Ws(µ) | ≤ b


(1)
0≤s≤%

sh(||(b − x)) || sh(2||b)v

  
 sh(||(b − z )) exp µ(z − x) − ch(2||b) − ch(2||z ) , z ≤ x ≤ b

=
 sh(||(b + x)) || sh(2||b)v
 
exp µ(z − x) − , −b ≤ x ≤ z
sh(||(b + z )) ch(2||b) − ch(2||z )


Ex e−α% ; a < inf Ws(µ) , sup Ws(µ) < b


4.15.4
0≤s≤% 0≤s≤%

v 2 + 2 sh((b − a) 2 + 2 )
 p p 
= exp µ(z − x) −
2 sh((b − z ) 2 + 2 ) sh((z − a) 2 + 2 )
p p

sh((x − a) 2 + 2 )
 p
 , −b ≤ x ≤ z
 sh((z − a) 2 + 2 )
 p
×
sh((b − x) 2 + 2 )
p
, z≤x≤b


sh((b − z ) 2 + 2 )
 p

Ex e−α% ; sup |Ws(µ) | ≤ b eµ(x−z)


(1)
0≤s≤%

sh((b + x) 2 + 2 ) v 2 + 2 sh((2b 2 + 2 ) 
 p  p p

 sh((b + z ) 2 + 2 ) exp − , −b ≤ x ≤ z
ch(2b 2 + 2 ) − ch(2z 2 + 2 )
 p p p
=
sh((b − x) 2 + 2 ) v 2 + 2 sh(2b 2 + 2 ) 
p  p p
− , z≤x≤b

 exp
sh((b − z ) 2 + 2 ) ch(2b 2 + 2 ) − ch(2z 2 + 2 )
 p p p

Ex e−γ`(%,r) ; sup Ws(µ) < b, % < ∞


4.16.1
0≤s≤%
r<b
z<r
||e(r−z)|| (||e(b−r)|| + 2 sh((b − r)||))v
 
= exp µ(z − x) −
2(|| sh((b − z )||) + 2 sh((b − r)||) sh((r − z )||))

|| sh((b − x)||)


 || sh((b − z )||) + 2 sh((b − r)||) sh((r − z )||) ,
 r≤x≤b
×
 || sh((b − x)||) + 2 sh((b − r)||) sh((r − x)||) ,

z≤x≤r
|| sh((b − z )||) + 2 sh((b − r)||) sh((r − z )||)

Px `(%, r) ∈ dy, sup Ws(µ) < b, % < ∞

4.16.2
0≤s≤%
r<b
z<r
 4. STOPPING AT INVERSE LOCAL TIME 335

2 Ws(µ) = µs + Ws % = %(v, z) = min{s : `(s, z) = v}

||(e(r−z)|| sh((b − r)||)v + sh((b − z )||)y )

 
= exp µ(z − x) −
2 sh((b − r)||) sh((r − z )||)
 ||√vy 
|| sh((b − x)||)

I dy, r≤x≤b
2 sh((b − r)||) sh((r − z )||) 0
sh((r − z )||)



√

 ||√vy 
|| v

 h
× 2 sh((r − x)|µ|) √ I1
 2 sh (( r − z )|  | ) y sh((r − z )||)


  ||√vy i
+ sh((x − z)|µ|)I0 dy, z ≤ x ≤ r

sh((r − z )||)


Px `(%, r) = 0, sup Ws(µ) < b, % < ∞

(1)
0≤s≤%

0, r≤x≤b
(
= sh((r − x)||) 
| | v

exp µ(z − x) − , z≤x≤r
sh((r − z )||) 2 sh((r − z )||)

Ex e−γ`(%,r) ; a < inf Ws(µ) , % < ∞


4.17.1
0≤s≤%
a<r
r<z
||e(z−r)|| (||e(r−a)|| + 2 sh((r − a)||))v
 
= exp µ(z − x) −
2(|| sh((z − a)||) + 2 sh((z − r)||) sh((r − a)||))

|| sh((x − a)||) + 2 sh((x − r)||) sh((r − a)||)


 || sh((z − a)||) + 2 sh((z − r)||) sh((r − a)||) ,
 r≤x≤z
×
|| sh((x − a)||)
, a≤x≤r

|| sh((z − a)||) + 2 sh((z − r)||) sh((r − a)||)


Px `(%, r) ∈ dy, a < inf Ws(µ) , % < ∞

4.17.2
0≤s≤%
a<r
r<z
||(e(z−r)|| sh((r − a)||)v + sh((z − a)||)y )
 
= exp µ(z − x) −
2 sh((z − r)||) sh((r − a)||)

√  ||√vy 
|| v
 h
2 sh((x − r)|µ|) √ I1
2 sh ((z − r)||) y sh((z − r)||)





  ||√vy i
× + sh((z − x)|µ|)I0 dy, r ≤ x ≤ z
 sh((z − r)||)
 ||√vy 
|| sh((x − a)||)



I dy, a≤x≤r

2 sh((z − r)||) sh((r − a)||) 0 sh((z − r)||)

336 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws % = %(v, z) = min{s : `(s, z) = v}

Px `(%, r) = 0, a < inf Ws(µ) , % < ∞

(1)
0≤s≤%

( sh((x − r)||) 
| | v

exp µ(z − x) − , r≤x≤z
= sh((z − r)||) 2 sh((z − r)||)
0, a≤x≤r

Ex exp −γ`(%, r) − η`(%, u) ; % < ∞ eµ(x−z)+|µ||z−x|



4.18.1
r<u
−v ||(2 + || + || +  (1 − e2(r−u)|| ))
 
z<r =exp 2
 + || (1 − e2(z−r)|| )+ ||(1 − e2(z−u)|| )+ (1 − e2(r−u)|| )(1 − e2(z−r)|| )


 1, x≤z
2 + || (1 − e2(x−r)|| ) + ||(1 − e2(x−u)|| ) + (1 − e2(r−u)|| )(1 − e2(x−r)|| )



,

2 + || (1 − e2(z−r)|| ) + ||(1 − e2(z−u)|| ) + (1 − e2(r−u)|| )(1 − e2(z−r)|| )





z≤x≤r



2 + ||(1 − e2(x−u)|| )


× ,

 2 + || (1 − e2(z−r)|| ) + ||(1 − e2(z−u)|| ) + (1 − e2(r−u)|| )(1 − e2(z−r)|| )
r≤x≤u



2




,
2 + || (1 − e2(z−r)|| ) + ||(1 − e2(z−u)|| ) + (1 − e2(r−u)|| )(1 − e2(z−r)|| )





u≤x


−v ||(2 + || + || +  (1 − e2(r−u)|| ))

 
u<z =exp 2
 + || (1 − e2(r−z)|| )+ ||(1 − e2(u−z)|| )+ (1 − e2(u−z)|| )(1 − e2(r−u)|| )

2

,
2 + || (1 − e2(r−z)|| ) + ||(1 − e2(u−z)|| ) + (1 − e2(u−z)|| )(1 − e2(r−u)|| )





x≤r



2
 + || (1 − e 2( r x )  )

− | |


,

2 + || (1 − e2(r−z)|| ) + ||(1 − e2(u−z)|| ) + (1 − e2(u−z)|| )(1 − e2(r−u)|| )



× r≤x≤u
2 + || (1 − e2(r−x)|| ) + ||(1 − e2(u−x)|| ) + (1 − e2(u−x)|| )(1 − e2(r−u)|| )




,
2 + || (1 − e2(r−z)|| ) + ||(1 − e2(u−z)|| ) + (1 − e2(u−z)|| )(1 − e2(r−u)|| )





u≤x≤z





z≤x

1,

%
ds
n  Z  o
4.19.1 Ex exp −γ ()
; 0 < inf Ws(µ) , % < ∞
0 Ws 0≤s≤%
 4. STOPPING AT INVERSE LOCAL TIME 337

2 Ws(µ) = µs + Ws % = %(v, z) = min{s : `(s, z) = v}

||v
 
= exp µ(z − x) −
(1 + =||)M− =||;1=2 (2||z )W− =||;1=2 (2||z )

− =||;1=2 (2||x)
 M
, 0<x≤z
− =||;1=2 (2||z )
 M

×
 W− =||;1=2 (2||x) , z ≤ x

W− =||;1=2 (2||z )

 2Z %
ds
n  o
4.20.1 Ex exp − ()
2
; 0 < inf Ws(µ) , % < ∞
2 0 Ws 0≤s≤%

 e−x √xI√
+1=4 (||x) v
 
 e−z √zI√ +1=4 (||z ) exp − 2zI√ +1=4 (||z )K√ +1=4 (||z ) , 0 < x ≤ z



= √
 e−x xK√ +1=4 (||x) 
v

exp − √ , z≤x

 −z √ √
e zK +1=4 (||z ) 2zI +1=4 (||z )K√ +1=4 (||z )


p2 q2 (Ws() )2
n  Z %   o
4.21.1 Ex exp − ()
2
+ ds ; 0 < inf Ws(µ) , % < ∞
0 2 Ws 2 0≤s≤%
0<x

qzv (1 + p2 + 1=4) −1 ((1 + p2 + 1=4 + 2 =2q)=2)

 p p 
= exp −
M−2 =4q;√4p2 +1=4 (qz 2 )W−2 =4q;√4p2 +1=4 (qz 2 )

 e−x √zM √ (qx2 )

−2 =4q; 4p2 +1=4
√ , 0<x≤z
(qz 2 )

 e−z xM 2 √ 2


− =4q; 4p +1=4
× √
 e−x zW−2 =4q;√4p2 +1=4 (qx2 )
, z≤x

 −z √
e xW−2 =4q;√4p2 +1=4 (qz 2 )


p2
 Z %
q
n   o
4.22.1 Ex exp − ()
2
+ () ds ; 0 < inf Ws(µ) , % < ∞
0<x 0 2 Ws Ws 0≤s≤%

||v (1 + 4p2 + 1) −1 (1=2 + p2 + 1=4 + q=||)

 p p 
= exp −
M−q=||;√p2 +1=4 (2||z )W−q=||;√p2 +1=4 (2||z )

 e−x M √ (2||x)
−q=||; p2 +1=4

 e−z M , 0<x≤z
√ (2||z )

−q=||; p2 +1=4

×
 e−x W−q=||;√p2 +1=4 (2||x)
, z≤x

 −z
e W−q=||;√p2 +1=4 (2||z )

338 2. BROWNIAN MOTION WITH DRIFT

2 Ws(µ) = µs + Ws % = %(v, z) = min{s : `(s, z) = v}

n  Z %

 o
p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds ; ` %, r ∈ dy



4.27.1 Ex exp −
0

= eµ(z−x) Q(27) 2 2
z (2p + µ , 2q + µ )dy, for Q(27)
z (2p, 2q) see 1.4.27.1

n  Z %

 o
p1I(−∞,r) Ws(µ) + q1I[r,∞) Ws(µ) ds ; ` %, r = 0



(1) Ex exp −
0

 −(z−x)√2p+µ2
e √ , x≤z
v 2p + 2 e(r−z) 2p+2 2

sh((r − x) 2p +  )
 p  p
z≤r = exp µ(z − x) −
2 sh((r − z ) 2p + 2 )  2 , z≤x≤r
 sh((r − z ) 2p +  )
p p

0, r≤x

 0, x≤r
√
sh((x − r) 2q + 2 )

v 2q + 2 e(z−r) 2q+
 p
2 
 p
r≤z = exp µ(z − x) − , r≤x≤z
2 sh((z − r) 2q + 2 ) sh((z − r ) 2q +  2 )
p p


 √
2q+µ2
e−(x−z) z≤x

,

Z % Z % 
1I(−∞,r) Ws(µ) ds ∈ du, 1I[r,∞) Ws(µ) ds ∈ dg, ` %, r ∈ dy




4.27.2 Px
0 0

2
u/2−µ2 g/2
= eµ(z−x)−µ Bz(27) (u, g, y)dudgdy

for Bz(27) (u, g, y) see 1.4.27.2

 Z % 2
p 2βWs(µ)
n (µ)
  o
4.30.1 Ex exp − e + qeβWs ds ; % < ∞
0
2

pv (1 + 2||=| |)e z
 
= exp −
(1=2 + ||=| | + q=p| |)M−q=p| |;||=| | |2p| e z W−q=p| |;||=| 2p e z



| | |

 −(+
e =2)x M 2p e x

−q=p| |;||=| | | |
2p e z
, (z − x)β ≥ 0

 e−(+ =2)z M−q=p| |;||=|


| | |
× 2p e x

 e−(+ =2)x W
−q=p| |;||=| | | |
2p e z
, (z − x)β ≤ 0

 −(+

=2)z W−q=p| |;||=|
e | | |
3 |W | τ ∼ Exp(λ), independent of W

3. REFLECTING BROWNIAN MOTION

1. Exponential stopping

√ √
2 √
Ex e−β|Wτ | = 2λe−βx − βe− 2λx


1.0.3 , 0≤x
2 − 2
√
 √ √
e−|z−x| 2λ
+ e−(z+x) 2λ


1.0.4 Px (|Wτ | ∈ dz) = √ dz, 0 ≤ z, 0≤x
2
√ √
1 2

t x

1 2

t x

1.0.5 Ex e−β|Wt | = eβx+β t/2
Erfc √ +√ + e−βx+β t/2
Erfc √ −√
2 2 2t 2 2 2t

1 2 2
e−(z−x) /2t + e−(z+x) /2t dz



1.0.6 Px |Wt | ∈ dz = √
2t
√
ch(x 2)

1.1.2 Px sup |Ws | ≥ y = √ , 0≤x≤y
0≤s≤τ ch(y 2)

1.1.4 Px sup |Ws | ≥ y = cc
e t (x, y)
0≤s≤t
∞ Z y
1 X 2 2
e−(z−x+4ky) /2t − e−(z−x+2y+4ky) /2t dz


=1− √
2t −y k=−∞

1.1.6 Px sup |Ws | < y, |Wτ | ∈ dz
0≤s≤τ
√ √ √
 sh((y − |z − x|) 2) + sh((y − x − z ) 2)


= √ √ dz, x∨z <y
2 ch(y 2)

1.1.8 Px sup |Ws | < y, |Wt | ∈ dz
0≤s≤t
∞
1 X 2 2
(−1)k e−(z−x+2ky) /2t + e−(z+x+2ky) /2t dz,


= √ x∨z <y
2t
k=−∞

√ √
e−|r−x| 2 + e−(x+r) 2


−γ`(τ,r)
1.3.1 Ex e =1− √ √

2 + 1 + e−2r 2

© Springer Basel 2015 339

A.N. Borodin, P. Salminen, Handbook of Brownian Motion - Facts and Formulae,
Probability and Its Applications, DOI 10.1007/978-3-0348-8163-0_10
340 3. REFLECTING BROWNIAN MOTION

3 |W | τ ∼ Exp(λ), independent of W

√ √ √
√ √

2 e−|r−x| 2 + e−(x+r) 2 
y er 2

1.3.2 Px `(τ, r) ∈ dy = √

2 exp − √ √ dy
1 + e−2r 2 2 ch(r 2)

√ √

e−|r−x| 2 + e−(x+r) 2
(1) Px `(τ, r) = 0 = 1 − √
1 + e−2r 2
√
 1 − ch(x√ 2) ,

0≤x≤r
= ch(r 2)
√
1 − e−(x−r) 2λ , r ≤ x


1.3.3 Ex e−γ`(t,r)
√ ∞
y(k+1)=2 −(|r−x|+2rl)2 /4y
k
2X k+1
X  
|r − x| + 2rl

=1+ √ (−γ) k! e D−k−2 √
 l!(k − l)! y
k=0 l=0

x + r + 2rl
2
 
+e−(x+r+2rl) /4y
D−k−2 √
y


1 
1.3.4 Px `(t, r) ∈ dy
2
= ecy (−1, 2, r, y + |x − r| − 2r, y)dy + ecy (−1, 2, r, y + x − r, y)dy

1 − cc
e t (x, r), 0≤x≤r
(


(1) Px `(t, r) = 0 = x−r
 
1 − Erfc √ , r≤x
2t
√
 √ √
n
Ex e−γ`(τ,r) ; |Wτ | ∈ dz = e−|z−x| 2λ + e−(z+x) 2λ

1.3.5 √
2
√ √
e−|x−r| 2 + e−(x+r) 2 −|z−r|√2λ √
o


−(z+r) 2λ
− √ √
e + e dz
2 + 1 + e−2r 2

√ √
y er 2

 
1.3.6 Px `(τ, r) ∈ dy, |Wτ | ∈ dz = exp − √ √
2 ch(r 2)
√ √ √ √
 e−|x−r| 2 + e−(x+r) 2 e−|z−r| 2 + e−(z+r) 2



× dydz
1 + e−2r 2 2
√

√
 √ √
n
e−|z−x| 2λ + e−(z+x) 2λ


(1) Px `(τ, r) = 0, |Wτ | ∈ dz = √
2
 1. EXPONENTIAL STOPPING 341

3 |W | τ ∼ Exp(λ), independent of W

√ √
e−|x−r| 2 + e−(x+r) 2 −|z−r|√2λ √
o
−(z+r) 2λ
− √ e + e dz
1 + e−2r 2

1 
1.3.8 Px `(t, r) ∈ dy, |Wt | ∈ dz
2
1
= (ecy (0, 2, r, y + |x − r| + |z − r| − 2r, y) + ecy (0, 2, r, y + |x − r| + z − r, y)
2
+ ecy (0, 2, r, y + |z − r| + x − r, y) + ecy (0, 2, r, y + x + z, y))dydz

x + z + y=2 −(x+z+y/2)2 /2t

(1) Px `(t, 0) ∈ dy, |Wt | ∈ dz = √ e dydz
t 2t


(2) Px `(t, r) = 0, |Wt | ∈ dz
( 1 −(z−x)2 /2t 1 2
√ e dz − √ e−(z+x−2r) /2t dz, r < x ∧ z
= 2t 2t
sct (r − |z − x|, r)dz + sct (r − z − x, r)dz, x∨z <r

 Z τ

1.4.1 Ex exp −γ 1I[r,∞) |Ws | ds
0
√
ch(x 2)
0≤x =1− √ √ √√ √
 + (  sh(r 2) +  + ch(r 2))
x≤r
√ √ √
  sh(r 2)e−(x−r) 2+2
r≤x = + √ √ √ √
+ ( + )(  sh(r 2) +  + ch(r 2))

Z τ

1.4.2 Px 1I[r,∞) |Ws | ds ∈ dy
0

√ √ √ √
ch(x 2) n  th(r 2) ey th (r 2) √
p 2 o
0≤x = e−λy √ √ + √ Erfc( λy th(r 2λ)) dy
ch(r 2) y ch2 (r 2)
x≤r

√
n
−λy

x−r

th(r 2) −(x−r)2 /2y
r≤x = λe 1 − Erfc √ + √ e
2y y
1 √ √ 2
√ 
x−r p √ o
+ 2 √ e(x−r) 2λ th(r 2λ)+λy th (r 2λ)Erfc √ + λy th(r 2λ) dy
ch (r 2) 2y

√
  1 − ch(x√ 2) , 0 ≤ x ≤ r

Z τ
ch(r 2)


(1) Px 1I[r,∞) |Ws | ds = 0 =
0 
0, r≤x
342 3. REFLECTING BROWNIAN MOTION

3 |W | τ ∼ Exp(λ), independent of W

Z t
 Z ∞
1.4.4 Px 1I[r,∞) |Ws | ds ∈ dv = Bx(27) (t − v, v, y)dy 1I(0,t) (v)dv
0 0

for Bx(27) (u, v, y) see 1.27.2

 t
Z


(1) Px 1I[r,∞) |Ws | ds = 0 = 1 − cce t (x, r)
x≤r 0
t
x−r
Z
  
(2) Px 1I[r,∞) |Ws | ds = t = 1 − Erfc √
0 2t
r≤x

n  Z τ
 o
1.4.5 Ex exp −γ 1I[r,∞) |Ws | ds ; |Wτ | ∈ dz 0≤x∧z
0
√ √ √ √ √ √

 √ √
2e−r 2 (  + − ) ch(x 2) ch(z 2) 
z≤r = √ e−|z−x| 2λ
+ e−(z+x) 2λ
− √ √ √ √ dz
2  sh(r 2) +  + ch(r 2)
x≤r
√ √ √
2 ch(z 2)e−(x−r) 2+2
z≤r = √ √ √ √ dz
 sh(r 2) +  + ch(r 2)
r≤x
√ √ √
2 ch(x 2)e−(z−r) 2+2
r≤z = √ √ √ √ dz
 sh(r 2) +  + ch(r 2)
x≤r

 √
r≤z = √ e−|z−x| 2λ+2γ dz
2 + 2
r≤x
√ √ √ √ √
(  + ch(r 2) −  sh(r 2))e−(z+x−2r) 2+2
+ √ √ √ √ √ dz
2 + 2 (  sh(r 2) +  + ch(r 2))

Z τ

1.4.6 Px 1I[r,∞) |Ws | ds ∈ dy, |Wτ | ∈ dz
0

√ √ √
 ch(x 2) ch(z 2) 2 −λy
z≤r = √ √ e
ch2 (r 2) y
x≤r
√ √ 2
√ p √ 
− 2λ th(r 2λ)e−λy/ ch (r 2λ) Erfc( λy th(r 2λ)) dydz

√ √
 ch(z 2) 2 −λy−(x−r)2 /2y √ √
z≤r = √ √ e − 2λ th(r 2λ)
ch(r 2) y
r≤x
√ √ 2
√ 
x−r p √ 
×e(x−r) 2λ th(r 2λ)−λy/ ch (r 2λ) Erfc √ + λy th(r 2λ) dydz
2y
 1. EXPONENTIAL STOPPING 343

3 |W | τ ∼ Exp(λ), independent of W

√ √
 ch(x 2) 2 −λy−(z−r)2 /2y √ √
r≤z = √ √ e − 2λ th(r 2λ)
ch(r 2) y
x≤r
√ √ √
2λ th(r 2λ)−λy/ ch2 (r 2λ)

z−r p √ 
×e(z−r) Erfc √ + λy th(r 2λ) dydz
2y

 2 2 √ √
e−λy−(z−x) /2y + e−λy−(z+x−2r) /2y dydz − λ 2λ th(r 2λ)


r≤z = √
2y
r≤x
√ √ √
2λ th(r 2λ)−λy/ ch2 (r 2λ)

z + x − 2r p √ 
×e(z+x−2r) Erfc √ + λy th(r 2λ) dydz
2y

Z τ

(1) Px 1I[r,∞) |Ws | ds = 0, |Wτ | ∈ dz
z≤r 0

√ √ √
(sh((r − |z − x|) 2) + sh((r − x − z ) 2))
x≤r = √ √ dz
2 ch(r 2)

Z t
 Z ∞
1.4.8 Px 1I[r,∞) |Ws | ds ∈ dv, |Wt | ∈ dz = Bx(27) (t − v, v, y, z)dy dvdz
v<t 0 0

for Bx(27) (u, v, y, z) see 1.27.6

Z t

(1) Px 1I[r,∞) |Ws | ds = 0, |Wt | ∈ dz
x≤r 0

z≤r = (sct (r − |z − x|, r) + sct (r − z − x, r))dz

  −(z−x)2 =2t
e−(z+x−2r) =2t
Z t
e
2 


(2) Px 1I[r,∞) |Ws | ds = t, |Wt | ∈ dz = √ − √ dz
0 2t 2t
r≤x
r≤z

 Z τ

1.5.1 Ex exp −γ 1I[0,r] |Ws | ds
0
√ √
  ch(x 2 + 2 )
0≤x = + √ √ √ √
+ ( + )(  ch(r 2 + 2 ) +  + sh(r 2 + 2 ))
x≤r

√ √
sh(r 2 + 2 )e−(x−r) 2
r≤x =1− √ √ √ √ √
 + (  ch(r 2 + 2 ) +  + sh(r 2 + 2 ))
344 3. REFLECTING BROWNIAN MOTION

3 |W | τ ∼ Exp(λ), independent of W

Z τ

1.5.2 Px 1I[0,r] |Ws | ds ∈ dy
0

√ √
0≤x = λe−λy dy + e−λy (rsy (0, r, x, 1/ 2λ) − 2λ rsy (−2, r, x, 1/ 2λ))dy
x≤r
r≤x = λe−λy−(x−r) dy
√ √ √

+e−λy−(x−r) 2λ
rsy (0, r, r, 1/ 2λ) − 2λ rsy (−2, r, r, 1/ 2λ) dy

Z τ
  0, 0≤x≤r
(1) Px 1I[0,r] |Ws | ds = 0 = √
0 1 − e−(x−r) 2λ , r ≤ x

Z t
 Z ∞
1.5.4 Px 1I[0,r] |Ws | ds ∈ dv = Bx(27) (v, t − v, y)dy 1I(0,t) (v)dv
0 0

for Bx(27) (u, v, y) see 1.27.2

Z t

(1) Px 1I[0,r] |Ws | ds = t = 1 − cc
e t (x, r)
x≤r 0

t
x−r
Z
  
(2) Px 1I[0,r] |Ws | ds = 0 = 1 − Erfc √
0 2t
r≤x

n  Z τ
 o
1.5.5 Ex exp −γ 1I[0,r] |Ws | ds ; |Wτ | ∈ dz
0≤x 0

0≤z
 √ √
e−|z−x| 2λ+2γ
+ e−(z+x) 2λ+2γ


x≤r = √ dz
2 + 2
z≤r
√ √ √ √ √
2(  + − )e−r 2+2 ch(x 2 + 2 ) ch(z 2 + 2 )
+ √ √ √ √ √
dz
2 + 2  ch(r 2 + 2 ) +  + sh(r 2 + 2 )

√ √ √
 2 ch(x 2 + 2 )e−(z−r) 2
x≤r = √ √ √ √ dz
 ch(r 2 + 2 ) +  + sh(r 2 + 2 )
r≤z

√ √ √
 2 ch(z 2 + 2 )e−(x−r) 2
r≤x = √ √ √ √ dz
 ch(r 2 + 2 ) +  + sh(r 2 + 2 )
z≤r
 1. EXPONENTIAL STOPPING 345

3 |W | τ ∼ Exp(λ), independent of W

√
 √
r≤x = √ e−|z−x| 2λ
dz
2
r≤z
√ √ √ √ √ √
(  ch(r 2 + 2 ) −  + sh(r 2 + 2 ))e−(z+x−2r) 2
+ √ √ √ √ √ dz
2(  ch(r 2 + 2 ) +  + sh(r 2 + 2 ))

Z τ

1.5.6 Px 1I[0,r] |Ws | ds ∈ dy, |Wτ | ∈ dz
0

√ √ √
x≤r = 2λe−λy−(z−r) 2λ
rsy (0, r, x, 1/ 2λ)dydz
r≤z
√ √ √
r≤x = 2λe−λy−(x−r) 2λ
rsy (0, r, z, 1/ 2λ)dydz
z≤r

Z τ

(1) Px 1I[0,r] |Ws | ds = 0, |Wτ | ∈ dz
r≤z 0
√
 √ √
e−|z−x| 2λ
− e(2r−x−z) 2λ


r≤x = √ dz
2

Z t
 Z ∞
1.5.8 Px 1I[0,r] |Ws | ds ∈ dv, |Wt | ∈ dz = Bx(27) (v, t − v, y, z)dy dvdz
v<t 0 0

for Bx(27) (u, v, y, z) see 1.27.6

Z t

(1) Px 1I[0,r] |Ws | ds = t, |Wt | ∈ dz
x≤r 0

z≤r = (sct (r − |z − x|, r) + sct (r − z − x, r))dz

  −(z−x)2 =2t
e−(z+x−2r) =2t
Z t
e
2 


(2) Px 1I[0,r] |Ws | ds = 0, |Wt | ∈ dz = √ − √ dz
0 2t 2t
r≤x
r≤z

 Z τ



1.6.1 Ex exp − p1I[0,r) |Ws | + q1I[r,∞) |Ws | ds
0

√
 (p − q) ch(x 2 + 2p)
0≤x = + √ √ √ √ √
+p ( + p)  + q(  + q ch(r 2 + 2p) +  + p sh(r 2 + 2p))
x≤r
346 3. REFLECTING BROWNIAN MOTION

3 |W | τ ∼ Exp(λ), independent of W

√ √
 (q − p) sh(r 2 + 2p)e−(x−r) 2+2q
r≤x = +√ √ √ √ √
+q  + p( + q)(  + q ch(r 2 + 2p) +  + p sh(r 2 + 2p))

Z τ

Z τ

1.6.2 Px 1I[0,r] |Ws | ds ∈ du, 1I[r,∞) |Ws | ds ∈ dv
0 0

Z ∞
= λe−λ(u+v) Bx(27) (u, v, y)dy dudv for Bx(27) (u, v, y) see 1.27.2
0

Z τ

Z τ

(1) Px 1I[0,r] |Ws | ds ∈ du, 1I[r,∞) |Ws | ds = 0
x≤r 0 0

= λe−λu (1 − cc
e u (x, r))du

τ τ
x−r
Z Z   
1I[r,∞) |Ws | ds ∈ dv = λe−λv Erf √



(2) Px 1I[0,r] |Ws | ds = 0, dv
0 0 2v
r≤x

Z t



1.6.4 Px p1I[0,r) |Ws | + q1I[r,∞) |Ws | ds ∈ dv
0

∞
1 |qt − v | |pt − v |
Z  
= Bx(27) , ,y dy 1I((q∧p)t,(q∨p)t) (v) dv
|p − q | 0
|p − q | |p − q |


for Bx(27) u, v, y see 1.27.2

Z t



(1) Px p1I[0,r) |Ws | + q1I[r,∞) |Ws | ds = pt = 1 − cc
e t (x, r)
x≤r 0

t
x−r
Z


  
(2) Px p1I[0,r) |Ws | + q1I[r,∞) |Ws | ds = qt = 1 − Erfc √
0 2t
r≤x

n  Z τ


 o
1.6.5 Ex exp − p1I[0,r) |Ws | + q1I[r,∞) |Ws | ds ; |Wτ | ∈ dz
0

 √ √
e−|z−x| 2λ+2p + e−(z+x) 2λ+2p dz


x≤r = √
2 + 2p
z≤r
 1. EXPONENTIAL STOPPING 347

3 |W | τ ∼ Exp(λ), independent of W

√ √ √ √ √
2(  + p −  + q)e−r 2+2p ch(x 2 + 2p) ch(z 2 + 2p)
+ √ √ √ √ √ dz
2 + 2p(  + p sh(r 2 + 2p) +  + q ch(r 2 + 2p))

√ √ √
2 ch(x 2 + 2p)e−(z−r) 2+2q
x≤r = √ √ √ √ dz
 + p sh(r 2 + 2p) +  + q ch(r 2 + 2p)
r≤z

√ √ √
2 ch(z 2 + 2p)e−(x−r) 2+2q
r≤x =√ √ √ √ dz
 + p sh(r 2 + 2p) +  + q ch(r 2 + 2p)
z≤r

 √
r≤x = √ e−|z−x| 2λ+2q
dz
2 + 2q
r≤z

√ √ √ √ √
(  + q ch(r 2 + 2p) −  + p sh(r 2 + 2p))e−(z+x−2r) 2+2q
+ √ √ √ √ √ dz
2 + 2q(  + p sh(r 2 + 2p) +  + q ch(r 2 + 2p))

Z τ

Z τ

1.6.6 Px 1I[0,r] |Ws | ds ∈ du, 1I[r,∞) |Ws | ds ∈ dv, |Wτ | ∈ dz
0 0

Z ∞
−λ(u+v)
= λe Bx(27) (u, v, y, z)dy dudvdz
0

for Bx(27) (u, v, y, z) see 1.27.6

Z τ

Z τ

(1) Px 1I[0,r] |Ws | ds ∈ du, 1I[r,∞) |Ws | ds = 0, |Wτ | ∈ dz
x≤r 0 0

= λe−λu (scu (r − |z − x|, r) + scu (r − z − x, r))dudz

Z τ

Z τ

(2) Px 1I[0,r] |Ws | ds = 0, 1I[r,∞) |Ws | ds ∈ dv, |Wτ | ∈ dz
r≤x 0 0

 −λv −(z−x)2 /2v 2

− e−(z+x−2r) /2v dvdz


= √ e e
2v

Z t



1.6.8 Px p1I[0,r) |Ws | + q1I[r,∞) |Ws | ds ∈ dv, |Wt | ∈ dz
0
348 3. REFLECTING BROWNIAN MOTION

3 |W | τ ∼ Exp(λ), independent of W

∞
1 |qt − v | |pt − v |
Z  
= Bx(27) , , y, z dy 1I((q∧p)t,(q∨p)t) (v) dvdz
|p − q | 0
|p − q | |p − q |


for Bx(27) u, v, y, z see 1.27.6

Z t



(1) Px p1I[0,r) |Ws | + q1I[r,∞) |Ws | ds = pt, |Wt | ∈ dz
x≤r 0

= (sct (r − |z − x|, r) + sct (r − z − x, r))dz

Z t



(2) Px p1I[0,r) |Ws | + q1I[r,∞) |Ws | ds = qt, |Wt | ∈ dz
r≤x 0

1 2 2
e−(z−x) /2t − e−(z+x−2r) /2t dz


= √
2t

 Z τ

1.7.1 Er exp −γ 1I[r,u] |Ws | ds =: Sr
0
√
1 + √+ th((u − r) ( + )=2)
 p 

=1− √ √ √ √
( + ) 1 + √+ (1 + th(r 2)) cth((u − r) 2 + 2 ) + + th(r 2)
 

 Z τ

Eu exp −γ 1I[r,u] |Ws | ds =: Su
0
√ √
1 + √+ th(r 2) th((u − r) ( + )=2)
 p 

=1− √ √ √ √
( + ) 1 + √+ (1 + th(r 2)) cth((u − r) 2 + 2 ) + + th(r 2)
 

 Z τ

Ex exp −γ 1I[r,u] |Ws | ds
0

√
ch(x 2)
0≤x = 1 − (1 − Sr ) √
ch(r 2)
x≤r
√ √
 
sh((u − x) 2 + 2 ) 
sh((x − r) 2 + 2 )
r≤x = + Sr − √ + Su − √
+  + sh((u − r) 2 + 2 )  + sh((u − r) 2 + 2 )
x≤u
√
u≤x = 1 − (1 − Su )e(u−x) 2λ

n  Z τ  o
1.8.5 Ex exp −γ |Ws |ds ; |Wτ | ∈ dz
0
 1. EXPONENTIAL STOPPING 349

3 |W | τ ∼ Exp(λ), independent of W

√ √ √ √
(3 )1=3  + zC−2=3 23 2 ( + x)3=2 ; 23 2 3=2 K1=3 23 2 ( + z )3=2
   

x≤z = √ √ dz
 + xK2=3 23 2 3=2
 

√ √ √ √
(3 )1=3  + xC−2=3 23 2 ( + z )3=2 ; 23 2 3=2 K1=3 23 2 ( + x)3=2
   

z≤x = √  √ dz
 + zK2=3 23 2 3=2


τ
2 Ai 21=3 −2=3 ( + z )
n  Z  o

(1) E0 exp −γ |Ws |ds ; |Wτ | ∈ dz = −
dz
0 (2 )1=3 Ai 21=3 −2=3 
0

n  Z t  o
1.8.7 E0 exp −γ |Ws |ds ; |Wt | ∈ dz
0
∞
X
Ai 0k + (2 )1=3 z

= −2 γ 1/3 exp 2−1/3 γ 2/3 tαk0 2=3 0 dz
2 k Ai( k )
0
k=1

∞ √  ( 0 )3 t3   ( 0 )2 t2 
Z t
(9=y)1=3 t k Ai √k 4=3
 X
1.8.8 P0 |Ws |ds < y |Wt | = 0 = exp
0
1= 6
2 | k|0 27y2 (3 2y)
k=1

 2Z t  2

1 x sh(t )

1.9.3 Ex exp − |Ws |2 ds = exp −
2 ch(t ) 2 ch(t )
p
0

2 t
 Z 
(1) Ex exp −β|Wt |2 − |Ws |2 ds
2 0

1 x2 ( sh(t ) + 2 ch(t ))
 
= exp −
ch(t ) + 2 −1 sh(t ) 2(ch(t ) + 2 −1 sh(t ))
p

Z t 
1.9.4 Px |Ws |2 ds ∈ dy = ecy (0, 1/2, t, 0, x2 /2)dy
0

n  2Z τ  o
1.9.5 Ex exp − |Ws |2 ds ; |Wτ | ∈ dz
2 0

 1 + 

2 p p  p
x≤z = √ D− 1 − λ (−x 2γ) + D− 1 − λ (x 2γ) D− 1 − λ (z 2γ)dz
 2 γ 2 γ 2 γ

 1 + 

2 p p  p
z≤x = √ D− 1 − λ (−z 2γ) + D− 1 − λ (z 2γ) D− 1 − λ (x 2γ)dz
 2 γ 2 γ 2 γ
350 3. REFLECTING BROWNIAN MOTION

3 |W | τ ∼ Exp(λ), independent of W

n  2Z t  o
1.9.7 Ex exp − |Ws |2 ds ; |Wt | ∈ dz
2 0

(x2 + z 2 ) ch(t )   xz 
√
xz
  
= exp − exp + exp − dz
2 sh(t ) 2 sh(t ) sh(t ) sh(t )
p

Z t 
1.9.8 Px |Ws |2 ds ∈ dy, |Wt | ∈ dz
0

= (2π)−1/2 (eey (1/2, t, (x2 + z 2 )/2, −xz) + eey (1/2, t, (x2 + z 2 )/2, xz))dydz

n  Z τ  o
1.10.5 E0 exp −γ e2β|Ws | ds ; |Wτ | ∈ dz
0
√ √
2K√2=| 2 e z

| | |
β>0 =− √ √
2
 dz, z>0
K√
0
2=| | | |

√ √
2I√2=| 2 e z

| | |
β<0 = √ √
2
 dz, z>0
I√
0
2=| | | |

=2

p Z h p
1.11.2 P0 sup `(τ, y) ≥ h = I0 (v λ/2)dv
sh(h =2)I0 (h =2)
p p
y∈[0,∞) 0

1.11.4 P0 sup `(t, y) < h
y∈[0,∞)

e−2j0,k t=h
∞ n
−2 2 k2 t=h2
2 2 j0.k πk
ke
X Z Z o
=2 J0 (v)dv − (−1) J0 (v)dv
j0;k J1 (j0;k ) sin j0;k 0
kJ0 (k) 0
k=1

√ √ √
2 ch(x 2 + 2 ) sh(y 2)
Ex e−γ Ȟ(τ ) ; sup |Ws | ∈ dy =

1.13.1 √ √ dy
0≤s≤τ ch(y 2 + 2 ) ch(y 2)
x<y

√ √

2 sh(y 2) −λv
1.13.2 Px Ȟ(τ ) ∈ dv; sup |Ws | ∈ dy = √ e ccv (x, y)dvdy
0≤s≤τ ch(y 2)

Ex e−γ Ȟ(t) ; sup |Ws | ∈ dy = 2 e−γt cct (x, y) ∗ sct (x, y)dy


1.13.3
0≤s≤t
 1. EXPONENTIAL STOPPING 351

3 |W | τ ∼ Exp(λ), independent of W

1.13.4 Px Ȟ(t) ∈ dv; sup |Ws | ∈ dy = 21I[0,t] (v) ccv (x, y) sct−v (x, y)dvdy
0≤s≤t

√ √
2 ch(x 2 + 2 ) ch(z 2)
Ex e−γ Ȟ(τ ) ; sup |Ws | ∈ dy, |Wτ | ∈ dz =

1.13.5 √ √ dydz
0≤s≤τ ch(y 2 + 2 ) ch(y 2)
x<y
z<y

√
2 ccv (x; y) ch(z 2)


1.13.6 Px Ȟ(τ ) ∈ dv; sup |Ws | ∈ dy, |Wτ | ∈ dz = √ dvdydz
0≤s≤τ ev ch(y 2)

Ex e−γ Ȟ(t) ; sup |Ws | ∈ dy, |Wt | ∈ dz = 2 e−γt cct (x, y) ∗ cct (z, y)dydz


1.13.7
0≤s≤t

1.13.8 Px Ȟ(t) ∈ dv, sup |Ws | ∈ dy, |Wt | ∈ dz = 2 ccv (x, y) cct−v (z, y)dvdydz
0≤s≤t
v<t

Ex e−γ`(τ,r) ; sup |Ws | < b


1.16.1
0≤s≤τ
r<b
√ √
( sh((b − r) 2) + =2) ch(x 2)
p
0≤x =1− √ √ √
sh((b − r) 2) ch(r 2) + =2 ch(b 2)
p
x≤r

√ √ √ √
sh((x − r) 2) ( sh((b − r) 2) + =2) ch(r 2) sh((b − x) 2)
p
r≤x =1− √ − √ √ √ √
sh((b − r) 2) ( sh((b − r) 2) ch(r 2) + =2 ch(b 2)) sh((b − r) 2)
p
x≤b

√
y =2 ch(b 2)
 p 


1.16.2 Px `(τ, r) ∈ dy, sup |Ws | < b = exp − √ √
0≤s≤τ sh((b − r) 2) ch(r 2)
r<b
√ √ √
=2(ch(b 2) − ch(r 2)) ch(x 2)
 p
√ √ dy, 0≤x≤r
sh((b − r) 2) ch2 (r 2)


× √ √ √
=2(ch(b 2) − ch(r 2)) sh((b − x) 2)
p
r≤x≤b

√ √ dy,
sh2 ((b − r) 2) ch(r 2)


√
 ch(x 2)
1 − √ , 0≤x≤r
ch(r 2)




(1) Px `(τ, r) = 0, sup |Ws | < b =
x) =2)
 1 − ch((b + r − 2p
p
0≤s≤τ
, r≤x≤b

sh((b − r) =2)
352 3. REFLECTING BROWNIAN MOTION

3 |W | τ ∼ Exp(λ), independent of W

Ex e−γ`(τ,r) ; sup |Ws | < b, |Wτ | ∈ dz


1.16.5
0≤s≤τ
r<b
z<b
√ √ √
 sh((r − |z − x|) 2) + sh((r − x − z ) 2)


z≤r = √ √ dz
2 ch(r 2)
x≤r
√ √ √
 sh((b − r) 2) ch(z 2) ch(x 2)
− √ √ √ √ dz
( sh((b − r) 2) ch(r 2) + =2 ch(b 2)) ch(r 2)
p

√ √
 sh((b − x) 2) ch(z 2)
z≤r = √ √ √ dz
sh((b − r) 2) ch(r 2) + =2 ch(b 2)
p
r≤x

√ √
 sh((b − z ) 2) ch(x 2)
r≤z = √ √ √ dz
sh((b − r) 2) ch(r 2) + =2 ch(b 2)
p
x≤r

√ √ √
 ch (b − r − |z − x|) 2 − ch (b + r − z − x) 2




r≤z = √ √
dz
2 sh (b − r) 2
r≤x
√ √ √
 sh((b − z ) 2) ch(r 2) sh((b − x) 2)
− √ √ √ √ dz
( sh((b − r) 2) ch(r 2) + =2 ch(b 2)) sh((b − r) 2)
p

1.16.6 Px `(τ, r) ∈ dy, sup |Ws | < b, |Wτ | ∈ dz
0≤s≤τ
r<b
z<b
√ √ √
 ch(z 2) ch(x 2) y =2 ch(b 2)
 p 
z≤r = 2 √ exp − √ √ dydz
ch (r 2) sh((b − r) 2) ch(r 2)
x≤r

√ √ √
 sh((b − x) 2) ch(z 2) y =2 ch(b 2)
 p 
z≤r = √ √ exp − √ √ dydz
sh((b − r) 2) ch(r 2) sh((b − r) 2) ch(r 2)
r≤x

√ √ √
 sh((b − z ) 2) ch(x 2) y =2 ch(b 2)
 p 
r≤z = √ √ exp − √ √ dydz
sh((b − r) 2) ch(r 2) sh((b − r) 2) ch(r 2)
x≤r

√ √ √
 sh((b − z ) 2) sh((b − x) 2) y =2 ch(b 2)
 p 
r≤z = 2 √ exp − √ √ dydz
sh ((b − r) 2) sh((b − r) 2) ch(r 2)
r≤x

(1) Px `(τ, r) = 0, sup |Ws | < b, |Wτ | ∈ dz
0≤s≤τ
 1. EXPONENTIAL STOPPING 353

3 |W | τ ∼ Exp(λ), independent of W

√ √ √
 sh((r − |z − x|) 2) + sh((r − x − z ) 2)


z≤r = √ √ dz
2 ch(r 2)
x≤r
√ √ √
 ch (b − r − |z − x|) 2 − ch (b + r − z − x) 2




r≤z = √ √
dz
2 sh (b − r) 2
r≤x


1.16.8 Px `(t, r) ∈ dy, sup |Ws | < b, |Wt | ∈ dz
0≤s≤t
0<y

z≤r = est (0, 0, b − r, 0, y/2) ∗ eccct (z, x, r, y/2)dydz

x≤r

z≤r = esst (b − x, b − r, 0, y/2) ∗ ecct (z, r, 0, y/2)dydz

r≤x

r≤z = esst (b − z, b − r, 0, y/2) ∗ ecct (x, r, 0, y/2)dydz

x≤r

r≤z = essst (b − z, b − x, b − r, y/2) ∗ ect (0, 0, r, 0, y/2)dydz

r≤x

(1) Px `(t, r) = 0, sup |Ws | < b, |Wt | ∈ dz
0≤s≤t

z≤r = sct (r − |z − x|, r)dz + sct (r − x − z, r)dz

x≤r

r≤z = cst (b − r − |z − x|, b − r)dz − cst (b + r − z − x, b − r)dz

r≤x
√ √ √ √

1.18.1 Ex/√2λ exp −γ 2λ `(τ, r/ 2λ) − η 2λ `(τ, u/ 2λ)

(1 + (1 + e−2u ) − (er−u + e−r−u ))(e−|x−r| + e−x−r )

r<u =1−
1 + (1 + e−2r ) + (1 + e−2u ) + (1 + e−2r )(1 − e2(r−u) )

(1 + (1 + e−2r ) − (er−u + e−r−u ))(e−|u−x| + e−u−x )

−
1 + (1 + e−2r ) + (1 + e−2u ) + (1 + e−2r )(1 − e2(r−u) )

n  Z τ


 o
1.23.1 Ex exp − p1I[0,r) |Ws | + q1I[r,∞) |Ws | ds ; sup |Ws | < b
0 0≤s≤τ
r<b
2 2(p2 + 2(q − p) ch((b − r)q )) ch(xp )
0≤x = − 2
p2 p q (p sh((b − r)q ) sh(rp ) + q ch((b − r)q ) ch(rp ))
x≤r
354 3. REFLECTING BROWNIAN MOTION

3 |W | τ ∼ Exp(λ), independent of W

2  sh((x − r)q )  2(p q + 2(q − p) sh((b − r)q ) th(rp )) sh((b − x)q )

r≤x = 1 − +
q2 sh((b − r)q ) p q2 (p th(rp ) + q cth((b − r)q )) sh2 ((b − r)q )
x≤b

n  Z τ



1.23.5 Ex exp − p1I[0,r) |Ws | + q1I[r,∞) |Ws | ds ;
r<b 0
o
z<b sup |Ws | < b, |Wτ | ∈ dz
0≤s≤τ

 −|z−x|Υp
+ e−(z+x)Υp dz


z≤r = e
p
x≤r

2(q − p th((b − r)q )) ch(zp ) ch(xp )e−rp

+ dz
p (p th((b − r)q ) sh(rp ) + q ch(rp ))
2 ch(zp ) sh((b − x)q )
z≤r = dz
p sh((b − r)q ) sh(rp ) + q ch((b − r)q ) ch(rp )
r≤x
2 sh((b − z )q ) ch(xp )
r≤z = dz
p sh((b − r)q ) sh(rp ) + q ch((b − r)q ) ch(rp )
x≤r
 −|z−x|Υq
− e(x+z−2b)Υq dz


r≤z = e
q
r≤x

2(p − p th(rp )) sh((b − z )q ) sh((b − x)q )e(r−b)q

+ dz
p (p sh((b − r)q ) th(rp ) + q ch((b − r)q ))

n  Z τ


 1 o
1.27.1 Ex exp − p1I[0,r) |Ws | + q1I[r,∞) |Ws | ds ; `(τ, r) ∈ dy
0
2


2 sh(rp ) 2  
yp sh(rp ) ch(xp )

x≤r = e−yΥq + exp − dy
p ch(rp ) q ch(rp ) ch(rp )

2 sh(rp ) 2  
yp sh(rp )

r≤x = e−(y+x−r)Υq + exp − dy
p ch(rp ) q ch(rp )

n  Z τ


 o
(1) Ex exp − p1I[0,r) |Ws | + q1I[r,∞) |Ws | ds ; `(τ, r) = 0
0

√
Υs := 2λ + 2s
 1. EXPONENTIAL STOPPING 355

3 |W | τ ∼ Exp(λ), independent of W

 ch(xp ) 
 
1− , x≤r
+p ch(rp )


=
 

1 − e(r−x)Υq , r≤x
+q


Z τ τ
1
Z 



1.27.2 Px 1I[0,r) |Ws | ds ∈ du, 1I[r,∞) |Ws | ds ∈ dv, `(τ, r) ∈ dy
0 0
2

=: λe−λ(u+v) Bx(27) (u, v, y)dudvdy

x≤r = λe−λ(u+v) 2(ecscu (x, r, r, y) hv (1, y) + eccu (x, r, 0, y) hv (0, y))dudvdy

r≤x = λe−λ(u+v) 2(escu (r, r, 0, y) hv (1, y + x − r)

+ ecu (0, 0, r, 0, y) hv (0, y + x − r))dudvdy

Z τ

Z τ

(1) Px 1I[0,r] |Ws | ds ∈ du, 1I[r,∞) |Ws | ds = 0, `(τ, r) = 0
x≤r 0 0

= λe−λu (1 − cce u (x, r))du

 τ
Z


Z τ


(2) Px 1I[0,r] |Ws | ds = 0, 1I[r,∞) |Ws | ds ∈ dv, `(τ, r) = 0
r≤x 0 0

x−r
 
= λe−λv Erf √ dv
2v

Z t


1 
1.27.4 Px p1I[0,r) |Ws | + q1I[r,∞) |Ws | ds ∈ dv, `(t, r) ∈ dy
0
2

1 
|qt − v | |pt − v |

= Bx(27) , ,y 1I((q∧p)t,(q∨p)t) (v) dvdy
|p − q | |p − q | |p − q |


for Bx(27) u, v, y see 1.27.2

Z t



(1) Px p1I[0,r) |Ws | + q1I[r,∞) |Ws | ds = pt, `(t, r) = 0 = 1 − cc
e t (x, r)
x≤r 0

t
x−r
Z


  
(2) Px p1I[0,r) |Ws | + q1I[r,∞) |Ws | ds = qt, `(t, r) = 0 = Erf √
0 2t
r≤x
356 3. REFLECTING BROWNIAN MOTION

3 |W | τ ∼ Exp(λ), independent of W

n  Z τ


 1 o
1.27.5 Ex exp − p1I[0,r) |Ws | + q1I[r,∞) |Ws | ds ; `(τ, r) ∈ dy, |Wτ | ∈ dz
0
2

yp sh(rp ) ch(xp ) ch(zp )

 
x≤r = 2λe−yΥq exp − dydz
ch(rp ) ch2 (rp )
z≤r
yp sh(rp ) ch(xp )
 
x≤r = 2λe−(y+z−r)Υq exp − dydz
ch(rp ) ch(rp )
r≤z
yp sh(rp ) ch(zp )
 
r≤x = 2λe−(y+x−r)Υq exp − dydz
ch(rp ) ch(rp )
z≤r
yp sh(rp )
 
r≤x = 2λe−(y+x+z−2r)Υq exp − dydz
ch(rp )
r≤z

n  Z τ


 o
(1) Ex exp − p1I[0,r) |Ws | + q1I[r,∞) |Ws | ds ; `(τ, r) = 0, |Wτ | ∈ dz
0

(sh((r − |z − x|)p ) + sh((r − z − x)p ))

x≤r = dz
p ch(rp )
z≤r
 −|z−x|Υq 
r≤x = e dz − e(2r−x−z)Υq dz
q q
r≤z

Z τ τ
1
Z 



1.27.6 Px 1I[0,r) |Ws | ds ∈ du, 1I[r,∞) |Ws | ds ∈ dv, `(τ, r) ∈ dy, |Wτ | ∈ dz
0 0
2

=: λe−λ(u+v) Bx(27) (u, v, y, z)dudvdydz

x≤r = λe−λ(u+v) 2 ecccu (x, z, r, y) hv (1, y)dudvdydz

z≤r
x≤r = λe−λ(u+v) 2 eccu (x, r, 0, y) hv (1, y + z − r)dudvdydz
r≤z
r≤x = λe−λ(u+v) 2 eccu (z, r, 0, y) hv (1, y + x − r)dudvdydz
z≤r
r≤x = λe−λ(u+v) 2 ecu (0, 0, r, 0, y) hv (1, y + x + z − 2r)dudvdydz
r≤z

√
Υs := 2λ + 2s
 1. EXPONENTIAL STOPPING 357

3 |W | τ ∼ Exp(λ), independent of W

Z τ

Z τ

(1) Px 1I[0,r) |Ws | ds ∈ du, 1I[r,∞) |Ws | ds = 0, `(τ, r) = 0, |Wτ | ∈ dz
x≤r 0 0

= λe−λu (scu (r − |z − x|, r) + scu (r − z − x, r))dudz

Z τ

Z τ


(2) Px 1I[0,r) |Ws | ds = 0, 1I[r,∞) |Ws | ds ∈ dv, `(τ, r) = 0, |Wτ | ∈ dz
r≤x 0 0

 2 2
e−λv e−(z−x) /2v
− e−(z+x−2r) /2v


= √ dvdz
2v

Z t


1 
1.27.8 Px p1I[0,r) |Ws | + q1I[r,∞) |Ws | ds ∈ dv, `(t, r) ∈ dy, |Wt | ∈ dz
0
2
1 
|qt − v | |pt − v |

= Bx(27) , , y, z 1I((q∧p)t,(q∨p)t) (v) dvdydz
|p − q | |p − q | |p − q |


for Bx(27) u, v, y, z see 1.27.6

Z t



(1) Px p1I[0,r) |Ws | + q1I[r,∞) |Ws | ds = pt, `(t, r) = 0, |Wt | ∈ dz
x≤r 0

= (sct (r − |z − x|, r) + sct (r − z − x, r))dz

 t
Z




(2) Px p1I[0,r) |Ws | + q1I[r,∞) |Ws | ds = qt, `(t, r) = 0, |Wt | ∈ dz
r≤x 0

1 2 2
e−(z−x) /2t − e−(z+x−2r) /2t dz


= √
2t

n  Z τ



1.29.1 Ex exp − p1I[0,r) |Ws | + q1I[r,∞) |Ws | ds ;
0
1 o
r<b sup |Ws | < b, `(τ, r) ∈ dy
0≤s≤τ 2

yq ch((b − r)q ) yp sh(rp )

 
= 2λ exp − −
sh((b − r)q ) ch(rp )


ch(xp )
 ch(r ) dy, 0≤x≤r

ch((b − r)q ) − 1 sh(rp ) 
p
× +
q sh((b − r)q ) p ch(rp ) 
 sh((b − x)q ) dy, r ≤ x ≤ b
sh((b − r)q )
358 3. REFLECTING BROWNIAN MOTION

3 |W | τ ∼ Exp(λ), independent of W

n  Z τ



(1) Ex exp − p1I[0,r) |Ws | + q1I[r,∞) |Ws | ds ;
0
o
sup |Ws | < b, `(τ, r) = 0
0≤s≤τ



ch(xp ) 
  + p 1 − ch(r ) ,
 0≤x≤r
p
=
 

sh((b − x)q ) + sh((x − r)q ) 
1− , r≤x≤b
+q sh((b − r)q )


n  Z τ



1.29.5 Ex exp − p1I[0,r) |Ws | + q1I[r,∞) |Ws | ds ;
0
1 o
r<b sup |Ws | < b, `(τ, r) ∈ dy, |Wτ | ∈ dz
0≤s≤τ 2
yq ch((b − r)q ) yp sh(rp )
 
= 2λ exp − −
sh((b − r)q ) ch(rp )
ch(xp ) ch(zp )

dydz, 0 ≤ x ≤ r, 0 ≤ z ≤ r
ch2 (rp )




sh((b − x)q ) ch(zp )



 sh((b − r) ) ch(r ) dydz, 0≤z≤r≤x≤b


q p
×
 sh((b − z )q ) ch(xp )
dydz, 0≤x≤r≤z≤b
sh((b − r)q ) ch(rp )




 sh((b − x2)q ) sh((b − z )q ) dydz, r ≤ x ≤ b, r ≤ z ≤ b




sh ((b − r)q )

n  Z τ



(1) Ex exp − p1I[0,r) |Ws | + q1I[r,∞) |Ws | ds ;
0
o
sup |Ws | < b, `(τ, r) = 0, |Wτ | ∈ dz
0≤s≤τ

(sh((r − |z − x|)p ) + sh((r − z − x)p ))


dz, x∨z ≤r
 ch(r )

p p

=
 (ch((b − r − |z − x|)q ) − ch((b + r − z − x)q )) dz,

r ≤x∧z
q sh((b − r)q )

Z τ

Z τ

1.29.6 Px 1I[0,r) |Ws | ds ∈ du, 1I[r,∞) |Ws | ds ∈ dv,
0 0
1 
sup |Ws | < b, `(τ, r) ∈ dy, |Wτ | ∈ dz
0≤s≤τ 2

√
Υs := 2λ + 2s
 1. EXPONENTIAL STOPPING 359

3 |W | τ ∼ Exp(λ), independent of W

z≤r = 2λe−λ(u+v) ecccu (x, z, r, y) esv (0, 0, b − r, 0, y)dudvdydz

x≤r

z≤r = 2λe−λ(u+v) eccu (z, r, 0, y) essv (b − x, b − r, 0, y)dudvdydz

r≤x

r≤z = 2λe−λ(u+v) eccu (x, r, 0, y) essv (b − z, b − r, 0, y)dudvdydz

x≤r

r≤z = 2λe−λ(u+v) ecu (0, 0, r, 0, y) esssv (b − z, b − x, b − r, y)dudvdydz

r≤x

Z t



1.29.8 Px p1I[0,r) |Ws | + q1I[r,∞) |Ws | ds ∈ dv,
0
1 
r<b sup |Ws | < b, `(t, r) ∈ dy, |Wt | ∈ dz
0≤s≤t 2

for (p ∧ q)t ≤ v ≤ (p ∨ q)t

2
x≤r = es|pt−v|/|p−q| (0, 0, b − r, 0, y) eccc|v−qt|/|p−q| (x, z, r, y)dvdydz
|p − q |
z≤r
2
z≤r = ess|pt−v|/|p−q| (b − x, b − r, 0, y) ecc|v−qt|/|p−q| (z, r, 0, y)dvdydz
|p − q |
r≤x
2
r≤z = ess|pt−v|/|p−q| (b − z, b − r, 0, y) ecc|v−qt|/|p−q| (x, r, 0, y)dvdydz
|p − q |
x≤r
2
r≤x = esss|pt−v|/|p−q| (b − x, b − z, b − r, y) ec|v−qt|/|p−q| (0, 0, r, 0, y)dvdydz
|p − q |
r≤z
Z t

Z t

(1) Px 1I[0,r) |Ws | ds = t, 1I[r,∞) |Ws | ds = 0
0 0

sup |Ws | < b, `(t, r) = 0, |Wt | ∈ dz
0≤s≤t

= sct (r − |z − x|, r)dz + sct (r − z − x, r)dz, 0≤x∧z ≤x∨z <r

Z t

Z t

(2) Px 1I[0,r) |Ws | ds = 0, 1I[r,∞) |Ws | ds = t
0 0

sup |Ws | < b, `(t, r) = 0, |Wt | ∈ dz
0≤s≤t

= cst (b − r − |z − x|, b − r)dz − cst (b + r − z − x, b − r)dz, r <x∧z

x∨z <b
360 3. REFLECTING BROWNIAN MOTION

3 |W | τ ∼ Exp(λ), independent of W

√ √ √


1.31.1 Ex/√2λ exp −γ 2λ sup `(s, u/ 2λ) − α−1 `(s, r/ 2λ)
0≤s≤τ
0<α
 + e−2r − e−2u e−|x−u| + e−x−u



= 1−
2 1 + e 2(u∧r) 1 − e−2|u−r| +  + e−2r − e−2u
−

2 1 + e−2(u∧r) 1 − e−2|u−r| − ( + e−2r − e−2u )(1 + e−2u ) e−|x−r| + e−x−r

−
2 1 + e−2(u∧r) 1 − e−2|u−r| +  + e−2r − e−2u e−|u−r| + e−u−r

 √ √
√ 
1.31.2 Px/√2λ sup `(s, u/ 2λ) − α−1 `(s, r/ 2λ) > h/ 2λ
0≤s≤τ
0<α
 + e−2r − e−2u h h −|x−r| −x−r
e +e


= exp −
2 1 + e−2(u∧r) 1 − e−2|u−r| e−|u−r| + e−u−r

 + e−2r − e−2u )(1 + e−2u )

 −|x−r| −x−r
(e +e i
− 2(u r ) 2 u r u r u r − e−|u−x| − e−x−u
2 1+e− ∧ 1−e− | − | e−| − | +e



− −

n √ √ √

√ o
1.31.5 Ex/√2λ exp −γ 2λ sup `(s, u/ 2λ) − α−1 `(s, r/ 2λ) ; |Wτ | ∈ dz/ 2λ
0≤s≤τ
0<α
1 −|z−x| −z−x

= e +e dz
2
 + e−2r − e−2u e−|x−u| + e−x−u




n
− e−|z−u| + e−z−u
2 2 1 + e−2(u∧r) 1 − e−2|u−r| +  + e−2r − e−2u

2 1 + e−2(u∧r) 1 − e−2|u−r| − ( + e−2r − e−2u )(1 + e−2u ) e−|x−r| + e−x−r o

− dz
2 1 + e−2(u∧r) 1 − e−2|u−r| +  + e−2r − e−2u e−|u−r| + e−u−r

 √ √
√ √ 
1.31.6 Px/√2λ sup `(s, u/ 2λ) − α−1 `(s, r/ 2λ) > h/ 2λ, |Wτ | ∈ dz/ 2λ
0≤s≤τ
0<α
1 −|z−u| −z−u
 + e−2r − e−2u h h −|x−r| −x−r
e +e


= e +e exp −
2 2 1 + e−2(u∧r) 1 − e−2|u−r| e−|u−r| + e−u−r

 + e−2r − e−2u )(1 + e−2u )

 −|x−r| −x−r
(e +e i
− 2(u r ) 2 u r u r u r − e−|u−x| − e−x−u dz
2 1+e− ∧ 1−e− | − | e−| − | +e



− −

q
2


Υ := 1 + e−2r − α − αe−2u + 4α 1 + e−2(u∧r) 1 − e−2|u−r| + 1 − α
 2. STOPPING AT FIRST HITTING TIME 361

3 |W | Hz = min{s : |Ws | = z}

2. Stopping at first hitting time

√
 ch(x√2 ) ,

0≤x≤z
2.0.1 Ex e−αHz = ch(z 2 )
 −(x−z)√2α
e , z≤x

cct (x, z)dt, 0≤x≤z

(
(x − z )2 


2.0.2 Px Hz ∈ dt = x−z

√ exp − dt, z ≤ x
z≤b 2t3=2 2t

0≤x≤z
(
1,
y−x


2.1.2 Px sup |Ws | < y =
, z≤x≤y
z≤y 0≤s≤Hz
y−z

 ch(x√2 )
 −αHz  ch(z √2 ) ,
 0≤x≤z
2.1.4 Ex e ; sup |Ws | < y = √
z≤y 0≤s≤Hz  sh((y − x)√2 ) , 0 ≤ z ≤ x ≤ y

sh((y − z ) 2 )

0≤x≤z


cct (x, z)dt,
(1) Px sup |Ws | < y, Hz ∈ dt =
0≤s≤Hz sst (y − x, y − z)dt, z ≤ x ≤ y

 1, 0 ≤ x ≤ z ≤ r, r ≤ z ≤ x
 1 + 2 |x − r |


Ex e−γ`(Hz ,r) , z∧r ≤x≤z∨r
2.3.1 = 1 + 2 |z − r |
1


, 0 ≤ x ≤ r ≤ z, z ≤ r ≤ x

1 + 2 |z − r |


x ≤ z ≤ r, r ≤ z ≤ x

 0,
z x y

| − |
  
exp − dy, z ∧ r ≤ x ≤ z ∨ r

2(z − r)2 2| z − r |


2.3.2 Px `(Hz , r) ∈ dy =
1 y

  
exp − dy, x ≤ r ≤ z, z ≤ r ≤ x

2|z − r| 2| z − r |


1, 0 ≤ x ≤ z ≤ r, r ≤ z ≤ x

|x − r |



(1) Px `(Hz , r) = 0 = , z∧r ≤x≤z∨r
 |z − r |

0, 0 ≤ x ≤ r ≤ z, z ≤ r ≤ x
362 3. REFLECTING BROWNIAN MOTION

3 |W | Hz = min{s : |Ws | = z}

2.3.3 Ex e−αHz −γ`(Hz ,r)

√
ch(x 2 )

 ch(z 2 ) ,

 √ 0≤x≤z

 √2 e(r−x)√2 + 2 sh((r − x)√2

)

= √ √ √ , z≤x≤r
2 e(r−z) 2 + 2 sh((r − z ) 2 )
z≤r

√ √

2 e−(x−r) 2



 √ √ , r ≤ x,
 √
2 e(r−z) 2 + 2 sh((r − z ) 2 )

√ √

2 ch(x 2 )
√ √ √ √ , 0≤x≤r
2 ch(z 2 ) + 2 sh((z − r) 2 ) ch(r 2 )




√ √ √ √
2 ch(x 2 ) + 2 sh((x − r) 2 ) ch(r 2 )

r≤z = √ √ √ √ , r≤x≤z

 2 ch(z 2 ) + 2 sh((z − r) 2 ) ch(r 2 )
√


e−(x−z) 2α

, z ≤ x,

Ex e−αHz ; `(Hz , r) ∈ dy

2.3.4

0, 0≤x≤z
 √ √

 2 sh((x − z ) 2 )

√
y 2
 √ dy, z ≤ x ≤ r

z≤r = exp − √ 2 sh2 ((r − z ) 2 )
1 − e−2(r−z) 2  √ −(x−r ) 2
√
 2 e


√ dy, r≤x

2 sh((r − z ) 2 )

0, z≤x
√ √

 2 sh((z − x) 2 )

√ √

y 2 ch(z 2 )

√ dy, r≤x≤z

r≤z = exp − √ √ 2 sh2 ((z − r) 2 )
2 sh((z − r) 2 ) ch(r 2 )  √ √

 2 ch(x 2 )
 √ √ dy, 0 ≤ x ≤ r
2 sh((z − r) 2 ) ch(r 2 )


√
ch(x 2 )

 ch(z 2 ) ,

 √ 0≤x≤z≤r



 −(x−z) 2α √

(1) Ex e−αHz ; `(Hz , r) = 0 = e

 , r≤z≤x
√
 sh(| x − r | 2 )
 √ , z∧r ≤x≤z∨r
sh(|z − r| 2 )





0, 0 ≤ x ≤ r ≤ z, z ≤ r ≤ x

 Z Hz

2.4.1 Ex exp −γ 1I[r,∞) |Ws | ds
0
 2. STOPPING AT FIRST HITTING TIME 363

3 |W | Hz = min{s : |Ws | = z}

 1, 0≤x≤z
 r−x x−z


+ √ , z≤x≤r
z≤r = r−z (r − z )(1 + (r − z ) 2 )

1 √
√ e−(x−r) 2γ ,

r≤x

1 + (r − z ) 2


 1
√ , 0≤x≤r

 ch(( z − r) 2 )
√

ch((x − r) 2 )

r≤z = √ , r≤x≤z

 ch((z − r) 2 )
 (z−x)√2γ

e , z≤x

Z Hz

2.4.2 Px 1I[r,∞) |Ws | ds ∈ dy
0
0≤x≤z

0,
  √y 
x−z 1 1

 
y/2(r−z)2
− z≤x≤r

√ e Erfc √ dy,
(r − z )2 2y 2(r − z )

(r − z ) 2



z<r = 1 2
√ e−(x−r) /2y dy−
(r − z ) 2y



√
1 x−r y x−r y 

 


− exp + Erfc √ + √ dy, r ≤ x

2(r − z )2 r−z 2(r − z )2 2y (r − z ) 2




 ccy (0, z − r)dy, 0≤x≤r

r<z = ccy (x − r, z − r)dy, r≤x≤z
 x − z

(x − z )2 
√ exp − dy, z ≤ x

2y3=2 2y


0≤x≤z

1,
Hz
r−x
Z  



(1) Px 1I[r,∞) |Ws | ds = 0 = , z≤x≤r
z≤r 0  r−z

0, r≤x

Z Hz

2.4.4 Px 1I[r,∞) |Ws | ds ∈ dv, Hz ∈ dt
0
Z ∞
= Bz(27) (t − v, v, y)dy dvdt, for Bz(27) (u, v, y) see 2.27.2
0

Z Hz

(1) Px 1I[r,∞) |Ws | ds = 0, Hz ∈ dt
0

0≤x≤z

cct (x, z)dt,
=
sst (r − x, r − z)dt, z ≤ x ≤ r
364 3. REFLECTING BROWNIAN MOTION

3 |W | Hz = min{s : |Ws | = z}

 Z Hz

2.5.1 Ex exp −γ 1I[0,r] |Ws | ds
0
 ch(x√2 )
√ , 0≤x≤z
ch(z 2 )



√
ch((r − x) 2 )


z≤r = √ , z≤x≤r
 ch((r − z ) 2 )
1



 √ , r≤x
ch((r − z ) 2 )
√
 ch(x 2 )
√ √ √ , 0≤x≤r
 ch(r√2 ) + √2 (z − r) sh(r √2 )



r≤z = ch(r 2 ) + 2 (x − r) sh(r 2 )
√ √ √ , r≤x≤z
 ch(r 2 ) + 2 (z − r) sh(r 2 )



1, z≤x

Z Hz

2.5.2 Px 1I[0,r] |Ws | ds ∈ dy
0

 ccy (x, z)dy,
 0≤x<z
z<r = ccy (r − x, r − z)dy, z < x ≤ r

ccy (0, r − z)dy, r≤x


rs (0, r, x, z − r)dy, 0≤x≤r


 y
z−x

r<z = rsy (0, r, r, z − r))dy, r ≤ x ≤ z
 z−r

0, z≤x

0≤x≤r

0,
Hz
x−r
Z  



(1) Px 1I[0,r] |Ws | ds = 0 = , r≤x≤z
r≤z 0  z−r

1, z≤x

Z Hz

2.5.4 Px 1I[0,r] |Ws | ds ∈ dv, Hz ∈ dt
0
Z ∞
= Bz(27) (v, t − v, y)dy dvdt, for Bz(27) (u, v, y) see 2.27.2
0

Z Hz

(1) Px 1I[0,r] |Ws | ds = 0, Hz ∈ dt
0
 2. STOPPING AT FIRST HITTING TIME 365

3 |W | Hz = min{s : |Ws | = z}

sst (x − r, z − r)dt, r ≤ x ≤ z

=
ht (1, x − z)dt, z≤x

 Z Hz



2.6.1 Ex exp − p1I[0,r) |Ws | + q1I[r,∞) |Ws | ds
0

 ch(x√2p)
√ , 0≤x≤z
ch(z 2p)



 √p ch((r − x)√2p) + √q sh((r − x)√2p)


z≤r = √ √ √ √ , z≤x≤r
 p ch((r − z ) 2p) + q sh((r − z ) 2p)
 √
√ −(x−r) 2q
pe


, r≤x

 √ √ √ √
p ch((r − z ) 2p) + q sh((r − z ) 2p)

√ √
q ch(x 2p)

√ √ √ √ √ √ , 0≤x≤r
q ch(r 2p) ch((z − r) 2q) + p sh(r 2p) sh((z − r) 2q)




√ √ √ √ √ √
q ch(r 2p) ch((x − r) 2q) + p sh(r 2p) sh((x − r) 2q)

r≤z = √ √ √ √ √ √ , r≤x≤z

 q ch(r 2p) ch((z − r) 2q) + p sh(r 2p) sh((z − r) 2q)
√


e−(x−z) 2q
z≤x

,

Z Hz

Z Hz

2.6.2 Px 1I[0,r) |Ws | ds ∈ du, 1I[r,∞) |Ws | ds ∈ dv
0 0

Z ∞
= Bz(27) (u, v, y)dy dudv, for Bz(27) (u, v, y) see 2.27.2
0

Z Hz

Z Hz

(1) Px 1I[0,r) |Ws | ds ∈ du, 1I[r,∞) |Ws | ds = 0
0 0

0≤x≤z

ccu (x, z)du,
=
ssu (r − x, r − z)du, z≤x≤r

Z Hz

Z Hz

(2) Px 1I[0,r) |Ws | ds = 0, 1I[r,∞) |Ws | ds ∈ dv
0 0

ssv (x − r, z − r)dv, r≤x≤z


=
hv (1, x − z)dv, z≤x
366 3. REFLECTING BROWNIAN MOTION

3 |W | Hz = min{s : |Ws | = z}

 Z Hz

2.7.1 Ex exp −γ 1I[r,u] |Ws | ds
r<u 0

0≤x≤z

1,
√ √ √

(r − x) 2 sh((u − r) 2 ) + ch((u − r) 2 )



 √ √ √
(r − z ) 2 sh((u − r) 2 ) + ch((u − r) 2 , z≤x≤r
)


√
z≤r = ch((u − x) 2 )
√ √ √ , r≤x≤u
(r − z ) 2 sh((u − r) 2 ) + ch((u − r) 2 )



1



, u≤x

√ √ √
(r − z ) 2 sh((u − r) 2 ) + ch((u − r) 2 )


 1
√ √ √ , 0≤x≤r
(z − u) 2 sh((u − r) 2 ) + ch((u − r) 2 )



 √
ch((x − r) 2 )


r≤x≤u

√ √ √ ,
= (z − u) √2 sh((u − r) 2 ) + ch((u − r) 2 )

u≤z √ √
(x − u) 2 sh((u − r) 2 ) + ch((u − r) 2 )
, u≤x≤z

√ √ √
(z − u) 2 sh((u − r) 2 ) + ch((u − r) 2 )





z≤x

1,

Z Hz

2.7.2 Px 1I[r,u] |Ws | ds ∈ dy = Bz(7) (y)dy, for Bz(7) (y) see 1.2.7.2
r<u 0

xI−1=3 ( 23 x3=2 2 )
√ √
, 0≤x≤z
( 2 z 3=2 2 )
Hz √zI

 √
−1=3
Z
3
 
2.8.1 Ex exp −γ |Ws |ds =
xK ( 2 x3=2 2 )
√ √
0 √ 1=3 23 3=2 √ ,

 z≤x
zK1=3 ( 3 z 2 )

Z H0  
2 1/3 |x| −2|x|3 /9y
2.8.2 Px |Ws |ds ∈ dy = e dy
0
9y4 (1=3)
(1)

√
xI−1=4 (x2 )
 √
√ , √ 0≤x≤z
zI−1=4 (z 2 )

 Hz Z  

2
2.9.1 Ex exp −2γ |Ws | ds = √
xK (x2 )
√
0  √ 1=4 2 √ , z≤x

zK1=4 (z )


 2Z
x 3=2
 2
x ch(t )
n H0  o 
2.9.4 Ex exp − |Ws |2 ds ; H0 ∈ dt = √ 3= 2 exp − dt
2 0 2 sh (t ) 2 sh(t )
(1)
3 3 x2

H0
x
Z 
(2) Px |Ws |2 ds ∈ dy, H0 ∈ dt = √ esy , , t, 0, dydt
0 2 2 2 2
 2. STOPPING AT FIRST HITTING TIME 367

3 |W | Hz = min{s : |Ws | = z}

√
C 2 =| 2 ; √2 e x
 √ 
| | | | |
2 ; √2 e z , 0≤x≤z

 √
C√
 
 2 =|



 | | | | |
√
 K√ 2 e x
 
Hz 
2 =|
 Z  | | |
2.10.3 Ex exp −αHz − γ e2β|Ws | ds = √
2 e z , z ≤ x, β>0
K√2 =| | | |

0 


√
I√2 =| | | 2| e x

 



 √ , z ≤ x, β<0
I 2 =| | | 2| e z

 √

∞
1
Z
e−λz P0


2.11.2 λ sup `(Hz , y) < h dz = 1 − √
0 y∈[0,z) I0 ( 2h)

e−j0,k z=2h
∞
X 2


(1) P0 sup `(Hz , y) < h = 2
y∈[0,z) j0;k J1 (j0;k )
k=1

2.18.1 Ex exp −γ`(Hz , r) − η`(Hz , u)
r<u

 1, 0≤x≤z
1 + 2 (r − x) + 2(u − x) + 4 (u − r)(r − x)



 1 + 2 (r − z ) + 2(u − z ) + 4 (u − r)(r − z ) ,


 z≤x≤r
z<r = 1 + 2(u − x)
, r≤x≤u
1 + 2 (r − z ) + 2(u − z ) + 4 (u − r)(r − z )



1



, u≤x

1 + 2 (r − z ) + 2(u − z ) + 4 (u − r)(r − z )

 1
, 0≤x≤r
1 + 2 (z − r) + 2(z − u) + 4 (z − u)(u − r)




1 + 2 (x − r)


r≤x≤u

,
1 + 2 ( ) + 2 (z − u) + 4 (z − u)(u − r)

u<z = z − r
 1 + 2 (x − r) + 2(x − u) + 4 (x − u)(u − r)
, u≤x≤z

1 + 2 (z − r) + 2(z − u) + 4 (z − u)(u − r)





1, z≤x


(18)
2.18.2 Px `(Hz , r) ∈ dy, `(Hz , u) ∈ dg = Bx,z (y, g)dydg
r<u
(18)
for Bx,z (y, g) see 1.2.18.2

n  Z Hz


 1
o
2.27.1 Ex exp − p1I[0,r) |Ws | + q1I[r,∞) |Ws | ds ; ` Hz , r ∈ dy
0
2
368 3. REFLECTING BROWNIAN MOTION

3 |W | Hz = min{s : |Ws | = z}

√ √ √ √
2p sh((x − z ) 2p) 
2p ch((r − z ) 2p) p 
z≤r = 2 √ exp −y √ − y 2q dy
sh ((r − z ) 2p) sh((r − z ) 2q)
x≤r

√ √ √
2p 
2p ch((r − z ) 2p) p 
z≤r = √ exp −y √ − (y + x − r) 2q dy
sh((r − z ) 2p) sh((r − z ) 2q)
r≤x

√ √  √ √ √ √
2q ch(x 2p) 2q ch((z − r) 2q) 2p sh(r 2p) 
r≤z = √ √ exp −y √ + √ dy
ch(r 2p) sh((z − r) 2q) sh((z − r) 2q) ch(r 2p)
x≤r

√ √ √ √ √ √
2q sh((z − x) 2q)  
2q ch((z − r) 2q) 2p sh(r 2p) 
r≤z = √ exp −y √ + √ dy
sh2 ((z − r) 2q) sh((z − r) 2q) ch(r 2p)
r≤x

n  Z Hz



o
(1) Ex exp − p1I[0,r) |Ws | + q1I[r,∞) |Ws | ds ; ` Hz , r = 0
0
√
ch(x 2p)

√ , 0≤x≤z
ch(z 2p)




√
sh((r − x) 2p)


= √ , z≤x≤r
 sh((r − z ) 2p)
√
 sh((x − r)√2q) , r ≤ x ≤ z




sh((z − r) 2q)

Z Hz Hz
1
Z 




2.27.2 Px 1I[0,r) |Ws | ds ∈ du, 1I[r,∞) |Ws | ds ∈ dv, ` Hz , r ∈ dy
0 0
2

=: Bz(27) (u, v, y)dudvdy

z≤r = escsu (x − z, 0, r − z, y) hv (1, y)dudvdy

x≤r

z≤r = esu (1, 1, r − z, 0, y) hv (1, y + x − r)dudvdy

r≤x

r≤z = eccu (x, r, 0, y) esv (1, 1, z − r, 0, y)dudvdy

x≤r

r≤z = ecu (0, 0, r, 0, y) escsv (z − x, 0, z − r, y)dudvdy

r≤x

Z Hz

Z Hz


(1) Px 1I[0,r) |Ws | ds ∈ du, 1I[r,∞) |Ws | ds = 0, ` Hz , r = 0
0 0
 2. STOPPING AT FIRST HITTING TIME 369

3 |W | Hz = min{s : |Ws | = z}

0≤x≤z

ccu (x, z)du,
=
ssu (r − x, r − z)du, z ≤ x ≤ r

Z Hz

Z Hz


(2) Px 1I[0,r) |Ws | ds = 0, 1I[r,∞) |Ws | ds ∈ dv, ` Hz , r = 0
0 0

ssv (x − r, z − r)dv, r ≤ x ≤ z

=
hv (1, x − z)dv, z≤x



2.31.1 Ex exp − sup `(s, u) − α−1 `(s, r)
2 0≤s≤Hz
0<α

for z ≤ x ∧ r ∧ u or x ∨ r ∨ u ≤ z

( + |r − z | − |u − z |)(|u − z | ∧ |x − z |)
=1−
2 (|u − z | ∧ |r − z |)|u − r| +  + |r − z | − |u − z |

2(|u − z | ∧ |r − z |)|u − r| − ( + |r − z | − |u − z |)|u − z |)(|x − z | ∧ |r − z |)

−
(2 (|u − z | ∧ |r − z |)|u − r| +  + |r − z | − |u − z |)(|u − z | ∧ |r − z |)

 
`(s, u) − α−1 `(s, r) > h


2.31.2 Px sup
0≤s≤Hz
0<α

for z ≤ x ∧ r ∧ u or x ∨ r ∨ u ≤ z


( + |r − z | − |u − z |)h h |x − z | ∧ |r − z |
= exp −
4(|u − z | ∧ |r − z |)|u − r| |u − z | ∧ |r − z |

 + |r − z | − |u − z | (|x − z | ∧ |r − z |)|u − z |
 i
− − |u − z| ∧ |x − z|
2(|u − z | ∧ |r − z |)|u − r| |u − z | ∧ |r − z |

p
Υ := (|r − z| − α|u − z|)2 + 4α(|u − z| ∧ |r − z|)|u − r|
370 3. REFLECTING BROWNIAN MOTION

3 |W | % = %(v, z) = min{s : `(s, z) = v}

4. Stopping at inverse local time

√ √ √
ch(x 2 ) v ez 2
  
 ch(z √2 ) exp − √2 ch(z √2 ) , 0≤x≤z


4.0.1 Ex e−α% = √ √
 e−(x−z)√2α exp − √v ez√2
  
, z≤x
2 ch(z 2 )


0≤x≤z


ecct (x, z, v/2, v/2)dt,
4.0.2 Px % ∈ dt =
ect (0, 0, z, x − z + v/2, v/2)dt, z ≤ x

Ex exp −γ sup |Ws | = e−γz 2vγK1 ( 2vγ),

p p
4.1.1 0≤x≤z
0≤s≤%
x≤z

v
  
 exp − 2(y − z ) ,


0≤x≤z
4.1.2 Px sup |Ws | < y =
 y − x exp − v
 
z≤y 0≤s≤%  , 0≤z≤x≤y
y−z 2(y − z )

Ex e−α% ; sup |Ws | < y


4.1.4
0≤s≤%
z≤y

 ch(x√2 )  √ √
v 2 ch(y 2 )

 ch(z √2 ) exp − 2 ch(z √2 ) sh((y − z )√2 ) ,
 0≤x≤z≤y
= √ √ √
 sh((y − x) 2 ) 
v 2 ch(y 2 )

√ exp − √ √ , 0≤z≤x≤y
sh((y − z ) 2 ) 2 ch(z 2 ) sh((y − z ) 2 )


v
  
exp − , x ≤ r ≤ z, r≤z≤x


 1 + 2 |z − r |
1 + 2 | r − x| v

  
4.3.1 Ex e−γ`(%,r) = exp − , r∧z ≤x≤r∨z
 1 + 2 |z − r | 1+2 |z − r |
1 v

  

 exp − , z ≤ r ≤ x, x≤r≤z
1 + 2 |z − r | 1+2 |z − r |

4.3.2 Px `(%, r) ∈ dy
 4. STOPPING AT INVERSE LOCAL TIME 371

3 |W | % = %(v, z) = min{s : `(s, z) = v}

√ √

v

(v + y)   vy 
√ exp − I dy, x ≤ z ≤ r, r ≤ z ≤ x
2|r − z | y 2|r − z | 1
|r − z |




√ √
1 (v + y) v vy

  h  

exp − |r − x| √ I1

2(r − z )2 2|r − z | y |r − z |


=  √vy i
+|x − z|I dy, z∧r ≤x≤r∨z

|r − z |

 0


√
1 (v + y) vy

    

exp − I dy, z ≤ r ≤ x, x ≤ r ≤ z

2|r − z | 2|r − z | 0 |r − z |


v
  
exp − , x ≤ z ≤ r, r ≤ z ≤ x
2|r − z |





| r − x| v
 
(1) Px `(%, r) = 0 = exp − , z∧r ≤x≤r∨z


 |r − z | 2| r − z |
z ≤ r ≤ x, x ≤ r ≤ z

0,

4.3.3 Ex e−α%−γ`(%,r)
√ √ √
v( er 2 + 2 ch(r 2 ))
 
z≤r = exp − √ √ √ √
ch(z 2 ) 2 e(r−z) 2 + 2 sh((r − z ) 2 )

√
ch(x 2 )

√ , 0≤x≤z
ch(z 2 )




 √2 e(r−x)√2 + 2 sh((r − x)√2

)

× √ (r−z)√2 √ , z≤x≤r
 2 e + 2 sh((r − z ) 2 )
√ √

2 e−(x−r) 2



 √ √ , r ≤ x,
 √
2 e(r−z) 2 + 2 sh((r − z ) 2 )
√ √ √ √
ve(z−r) 2 ( er 2 + 2 ch(r 2 ))
 
r≤z = exp − √ √ √ √
2 ch(z 2 ) + 2 sh((z − r) 2 ) ch(r 2 )
√ √

2 ch(x 2 )
√ √ √ √ , 0≤x≤r
2 ch(z 2 ) + 2 sh((z − r) 2 ) ch(r 2 )




√ √ √ √
2 ch(x 2 ) + 2 sh((x − r) 2 ) ch(r 2 )

× √ √ √ √ , r≤x≤z

 2 ch(z 2 ) + 2 sh((z − r) 2 ) ch(r 2 )
√


e−(x−z) 2α

, z ≤ x,

Ex e−α% ; `(%, r) ∈ dy

4.3.4
√ √ √ (r−z)√2

ch(r 2 )v e y

z≤r = exp − √ √ √ −√ √
2 sh((r − z ) 2 ) ch(z 2 ) 2 sh((r − z ) 2 )
372 3. REFLECTING BROWNIAN MOTION

3 |W | % = %(v, z) = min{s : `(s, z) = v}

√ √ √
ch(x 2 ) v 2 vy
  
√ √ √ √ I1 √ dy, 0≤x≤z
2 sh((r − z ) 2 ) ch(z 2 ) y sh((r − z ) 2 )




√ √ √

√ v 2 vy

 h  

 √ 2 √ sh((r − x) 2α) √ I1 √
2 sh ((r − z ) 2 ) y sh((r − z ) 2 )


× √
√ 
2 vy i
+ sh((x − z) 2α)I0 √ dy, z≤x≤r

sh((r − z ) 2 )




√ −(x−r)√2 √

e 2 vy

  

√ √ I0 √ dy, r≤x

2 sh((r − z ) 2 ) sh((r − z ) 2 )


√ (z−r)√2 √ √

e v ch(z 2 )y 
r≤z = exp − √ √ − √ √ √
2 sh((z − r) 2 ) 2 sh((z − r) 2 ) ch(r 2 )
√ √ √
e−(x−z) 2 v 2 vy
  
 √ √ √ I1 √ dy, z≤x
2 sh((z − r) 2 ) y sh((z − r) 2 )




√ √
√ qv 2 vy

 h  

√ √ sh((x − r) 2α) I √

2
2 sh ((z − r) 2 ) y 1 sh((z − r) 2 )


× √
√ 
2 vy i
+ sh((z − x) 2α)I0 dy, r≤x≤z

√
sh((z − r) 2 )




√ √ √

ch(x 2 ) 2 vy

  

√ √ √ I0 √ dy, 0≤x≤r

2 sh((z − r) 2 ) ch(r 2 ) sh((z − r) 2 )


 v√2 e(|r−z|+z)√2 + e−r√2


−α%

(1) Ex e ; `(%, r) = 0 = exp − √ √
4 ch(z 2 ) sh(|z − r| 2 )
√
ch(x 2 )

√ , 0≤x≤z≤r
ch(z 2 )





 √
e−(x−z) 2α , r≤z≤x

× √
 sh(|x − r| 2 )
 √ , z∧r ≤x≤z∨r
sh(|z − r| 2 )





0, 0 ≤ x ≤ r ≤ z, z ≤ r ≤ x

 Z %

4.4.1 Ex exp −γ 1I[r,∞) |Ws | ds
0

√
v 2
  
exp − √ , 0≤x≤r
2(1 + (r − z ) 2 )




√ √
1 + ( r − x) 2 v 2

  
z≤r = √ exp − √ , z≤x≤r
 1 + ( r − z ) 2 2(1 + ( r − z) 2 )
√
1 v 2


  √ 
√ exp −(x − r) 2γ − √ , r≤x

1 + (r − z ) 2 2(1 + (r − z ) 2 )

 4. STOPPING AT INVERSE LOCAL TIME 373

3 |W | % = %(v, z) = min{s : `(s, z) = v}

√
1 v 2
  
√ exp − √ , 0≤x≤r
ch((z − r) 2 exp(2(r − z ) 2 ) + 1




√ √
ch((x − r) 2 v 2

  
r≤z = √ exp − √ , r≤x≤z
 ch(( z − r ) 2 exp(2(r − z) 2 ) + 1
√
v 2

 exp (z − x)√2γ −

  
√ , z≤x

exp(2(r − z ) 2 ) + 1

n Z % o
−α%


4.4.4 Ex e ; 1I[r,∞) |Ws | ds = 0
(1) 0

 ch(x√2 )
 ch(z √2 ) ,
 0≤x≤z
√ √
v 2 ch(r 2 )
 
 √
z≤r = exp − √ √ sh((r − x) 2 )
2 ch(z 2 ) sh((z − r) 2 )  √ , z≤x≤r
 sh((r − z ) 2 )


0, r≤x

 Z %

4.5.1 Ex exp −γ 1I[0,r] |Ws | ds
0

 ch(x√2 )
√ , 0≤x≤z
 ch(z 2 )


√ √ √
v 2 sh(r 2 ) ch((r − x) 2 )
 

z≤r = exp − √ √ √ , z≤x≤r
2 ch(z 2 ) ch((r − z ) 2 )  ch((r − z ) 2 )
1



 √ , r≤x
ch((r − z ) 2 )

√ √
v 2
th(r 2 )
 
r≤z = exp − √ √
2(1 + 2 (z − r) th(r 2 ))

√
 ch(x 2 )
√ √ √ , 0≤x≤r
 ch(r√2 ) + √2 (z − r) sh(r √2 )



× ch(r √2 ) + √2 (x − r) sh(r√2 )
, r≤x≤z
 ch(r 2 ) + 2 (z − r) sh(r 2 )



1 z≤x

n Z % o
−α%


4.5.4 Ex e ; 1I[0,r] |Ws | ds = 0
(1) 0
374 3. REFLECTING BROWNIAN MOTION

3 |W | % = %(v, z) = min{s : `(s, z) = v}

 0, x≤r

2 (√
z r )  sh((x − r)√2 )
√
v 2 e
 − 
r≤z = exp − √ √ , r≤x≤z
2 sh((z − r) 2 )  sh((z − r) 2 )

 −(x−z) 2α√
e , z≤x

 Z %



4.6.1 Ex exp − p1I[0,r) |Ws | + q1I[r,∞) |Ws | ds
0
√ √ √ √ √
 v p( p sh(r 2p) + q ch(r 2p)) 
z≤r = exp − √√ √ √ √
ch(z 2p)( 2p ch((r − z ) 2p) + 2q sh((r − z ) 2p))

 ch(x√2p)
√ , 0≤x≤z
ch(z 2p)



 √p ch((r − x)√2p) + √q sh((r − x)√2p)


× √p ch((r − z )√2p) + √q sh((r − z )√2p) , z ≤ x ≤ r

√ −(x−z)√2q

pe


, r≤x

 √ √ √ √
p ch((r − z ) 2p) + q sh((r − z ) 2p)

√ √ √
v q( p sh(r 2p) + q ch(r 2p))e−(z−r) 2q
 √ √ √ 
r≤z = exp − √ √ √ √ √ √
2q ch(r 2p) ch((z − r) 2q) + 2p sh((z − r) 2q) sh(r 2p)

√ √
q ch(x 2p)

√ √ √
√ √ √ , x≤r
q ch(r 2p) ch((z − r) 2q) + p sh(r 2p) sh((z − r) 2q)




√ √ √ √ √ √
q ch(r 2p) ch((x − r) 2q) + p sh(r 2p) sh((x − r) 2q)

× √ √ √ √ √ √ , r≤x≤z

 q ch(r 2p) ch((z − r) 2q) + p sh(r 2p) sh((z − r) 2q)
√


e−(x−z) 2q
z≤x

,

 Z %

4.7.1 Ex exp −γ 1I[r,u] |Ws | ds
r<u 0

√ √
v 2 sh((u − r) 2 )
 
z≤r = exp − √ √ √
2((r − x) 2 sh((u − r) 2 ) + ch((u − r) 2 ))
0≤x≤z

1,
√ √ √

(r x ) 2 sh((u r ) 2 ) + ch((u r ) 2 )

− − −


(r − z )√2 sh((u − r)√2 ) + ch((u − r)√2 ) , z ≤ x ≤ r



√
× ch((u − x) 2 )
√ √ √ , r≤x≤u
(r − z ) 2 sh(u − r) 2 ) + ch((u − r) 2 )



1


, u≤x

√ √ √
(r − z ) 2 sh(u − r) 2 ) + ch((u − r) 2 )

 4. STOPPING AT INVERSE LOCAL TIME 375

3 |W | % = %(v, z) = min{s : `(s, z) = v}

√ √
v 2 sh((u − r) 2 )
 
u≤z = exp − √ √ √
2((z − u) 2 sh((u − r) 2 ) + ch((u − r) 2 ))
 1
√ √ √ , 0≤x≤r
(z − u) 2 sh((u − r) 2 ) + ch((u − r) 2 )



 √
ch((x − r) 2 )


r≤x≤u

√ √ √ ,
× (z − u) √2 sh((u − r) 2 ) + ch((u − r) 2 )

√ √
(x − u)√2 sh((u − r) 2 ) + ch((u − r) 2 )

, u≤x≤z

√ √
(z − u) 2 sh((u − r) 2 ) + ch((u − r) 2 )




z≤x

1,

 Z % 
4.8.1 Ex exp −γ |Ws |ds
0
xI−1=3 ( 23 x3=2 2 )
√ √
, 0≤x≤z
zI−1=3 ( 23 z 3=2 2 )
 √ √
3z −1 v
  

= exp −
4I−1=3 ( 23 z 3=2 2 )K1=3 ( 23 z 3=2 2 ) 
√ √ √ 2 3=2 √
√xK1=3 ( 3 x √ 2 ) , z ≤ x
 2
zK1=3 ( 3 z 3=2 2 )

 Z % 
4.9.1 Ex exp −2γ |Ws |2 ds
0
√
xI−1=4 (x2 )
 √
√ , √ 0≤x≤z
zI−1=4 (z 2 )

v
 

= exp − √ √ √
zI−1=4 (z 2 )K1=4 (z 2 )  xK (x2 )
√
 √ 1=4 2 √ , z≤x
zK1=4 (z )


 Z % 
4.10.1 Ex exp −γ e2β|Ws | ds
0
  √ 2 √2 x 
C0 | | ; | | e
, 0 ≤ x ≤ z

√
2
  √2 √2

K1 v  C0 | | ; | | e z

 
| |
β>0 = exp − √
2 ; √2 e z K
√
2 e z √
2C0 K0 | 2| e x
  
| | | | 0 | |


, z≤x
 √
K0 | 2| e z



  √2 √2 x 
C0 | | ; | | e
, 0 ≤ x ≤ z
 √ √
 C0 | 2| ; | 2| e z
√
2

I1 v
 


| |
β<0 = exp − √
2 ; √2 e z I
 √
2 e z √
2C0 I0 | 2| e x
 
| | | | 0 | |


  √2 z  , z≤x


I0 | | e
376 3. REFLECTING BROWNIAN MOTION

3 |W | % = %(v, z) = min{s : `(s, z) = v}

∞ √
v
I0 ( 2v)
Z  
−λz


4.11.2 λ e P0 sup `(%, y) < h dz = 1 − 1− √ 1I (v)
0 y∈[0,∞) h I0 ( 2h) [0,h]

v
X e−j0,k z=2h J0 (j0;k v=h)

∞ 2 p


(1) P0 sup `(%, y) < h = 2 1 − 1I[0,h] (v)
y∈[0,∞) h j0;k J1 (j0;k )
k=1

v

(2) P0 sup `(%(v, 0), y) < h = 1 − 1I (v)
y∈[0,∞) h [0,h]

√
sh((x − z )+ 2 ) 
v

Ex e−γ Ȟ(%) ; sup |Ws | ∈ dy =

4.13.1 √ exp − dy
0≤s≤% (y − z ) sh((y − z ) 2 ) 2(y − z )
√ √
v 2 ch(y 2 ) v 
exp − exp −
  
− √ √
2 sh((z − y) 2 ) ch(z 2 ) 2(y − z )
+ √ √ √ √
sh((y − z ) 2 ) ch(z 2 ) − (y − z ) 2 ch(y 2 )

(√ √
2γ ch(x 2γ)dy, 0≤x≤z
√ √ √
× 2 sh((y − x) 2 ) ch(z 2 )
√ dy, z ≤ x ≤ y
sh((y − z ) 2 )

4.18.1 Ex exp −γ`(%, r) − η`(%, u)
r<u
v( +  + 2 (u − r))
 
z≤r = exp −
1 + 2 (r − z ) + 2(u − z ) + 4 (u − r)(r − z )

 1, 0≤x≤z
1 + 2 (r − x) + 2(u − x) + 4 (u − r)(r − x)



 , z≤x≤r
1 + 2 (r − z ) + 2(u − z ) + 4 (u − r)(r − z )



× 1 + 2(u − x)
, r≤x≤u
1 + 2 (r − z ) + 2(u − z ) + 4 (u − r)(r − z )



1



, u≤x

1 + 2 (r − z ) + 2(u − z ) + 4 (u − r)(r − z )


v( +  + 2 (u − r))
 
u≤z = exp −
1 + 2 (z − r) + 2(z − u) + 4 (z − u)(u − r)
 1
 1 + 2 (z − r) + 2(z − u) + 4 (z − u)(u − r) , 0 ≤ x ≤ r



1 + 2 (x − r))


, r≤x≤u

× 1 + 2 (z − r) + 2(z − u) + 4 (z − u)(u − r)

 1 + 2 (x − r) + 2(x − u) + 4 (x − u)(u − r)
, u≤x≤z

1 + 2 (z − r) + 2(z − u) + 4 (z − u)(u − r)





1, z≤x

 4. STOPPING AT INVERSE LOCAL TIME 377

3 |W | % = %(v, z) = min{s : `(s, z) = v}

n  Z %



o
4.27.1 Ex exp − p1I[0,r) |Ws | + q1I[r,∞) |Ws | ds ; ` %, r ∈ dy
0

 v√p sh(z √2p) (y + v)√p ch((r − z )√2p) y√q 

z≤r = exp − √ √ − √ √ − √
2 ch(z 2p) 2 sh((r − z ) 2p) 2

√ √ √
pv ch(x 2p) 2pvy
  
√ √ √ √ I1 √ dy, 0≤x≤z
2 sh((r − z ) 2p) ch(z 2p) y sh((r − z ) 2p)




√ √ √

p √ v 2pvy

 h  

 √ 2 √ sh((r − x) 2p) √ I1 √
2 sh ((r − z ) 2p) y sh((r − z ) 2p)


× √
√ 
2pvy i
+ sh((x − z) 2p)I0 √ dy, z≤x≤r

sh((r − z ) 2p)




 √ −(x−r) 2q √ √
pe 2pvy

  

 √ √ I0 √ dy, r≤x
2 sh((r − z ) 2p) sh((r − z ) 2p)


 v√q (v + y)√q ch((z − r)√2q) y√p sh(r√2p) 

r≤z = exp − √ − √ √ − √ √
2 2 sh((z − r) 2q) 2 ch(r 2p)

√ √ √
q ch(x 2p) 2qvy
  
√ √ √ I0 √ dy, 0≤x≤r
2 sh((z − r) 2q) ch(r 2p) sh((z − r) 2q)




√ √ √

q √ v 2qvy

 h  

 √ 2 √ sh((x − r) 2q) √ I1 √
2 sh ((z − r) 2q) y sh((z − r) 2q)


× √
√ 
2qvy i
 + sh((z − x) 2q)I0 √ dy, r≤x≤z
sh((z − r) 2q)




√ √
qve−(x−z) 2q
√

2qvy

  
z≤x

√ √ √ I1 √ dy,
sh((z − r) 2q)

2 sh((z − r) 2q) y


n  Z %



o
(1) Ex exp − p1I[0,r) |Ws | + q1I[r,∞) |Ws | ds ; ` %, r = 0
0

 ch(x√2p)
 ch(z √2p) ,
 0≤x≤z
√ √
v p ch(r 2p)  √
 
z≤r = exp − √ √ √ sh((r − x) 2p)
2 sh((r − z ) 2p) ch(z 2p)  √ , z≤x≤r
 sh((r − z ) 2p)


0, r≤x
378 3. REFLECTING BROWNIAN MOTION

3 |W | % = %(v, z) = min{s : `(s, z) = v}

0≤x≤r

0,
√ √

v qe(z−r) 2q  sh((x − r)√2q)
 √ 

r≤z = exp − √ , r≤x≤z
 sh((z − r√) 2q)
√
2 sh((z − r) 2q) 
 −(x−z) 2q
e , z≤x

Z % %
1
Z 




4.27.2 Px 1I[0,r) |Ws | ds ∈ du, 1I[r,∞) |Ws | ds ∈ dg, ` %, r ∈ dy
0 0
2

√
v √
x≤z = √ (is (1, r − z, 0, y + v, vy) ∗ eccu (x, z, 0, v)) hg (1, y)dudgdy
y u
z≤r

√
z≤r = (isu (0, r − z, 0, y + v, vy) ∗ ecu (0, 0, z, 0, v)) hg (1, x − r + y)dudgdy
r≤x

√
x≤r = eccu (x, r, 0, y) isg (0, z − r, v, y + v, vy)dudgdy
r≤z

√
v √
r≤z = √ ec (0, 0, r, 0, y) isg (1, z − r, x − z + v, y + v, vy)dudgdy
y u
z≤x

Z %

Z %


(1) Px 1I[0,r) |Ws | ds ∈ du, 1I[r,∞) |Ws | ds = 0, ` %, r = 0
0 0

eccu (x, z, 0, v/2) ∗ esu (0, 0, r − z, 0, v/2)du, 0≤x≤z


=
ecu (0, 0, z, 0, v/2) ∗ essu (r − x, r − z, 0, v/2)du, z≤x≤r

Z %

Z %


(2) Px 1I[0,r) |Ws | ds = 0, 1I[r,∞) |Ws | ds ∈ dg, ` %, r = 0
0 0

essg (x − r, z − r, v/2, v/2)dg, r≤x≤z


=
esg (0, 0, z − r, x − z + v/2, v/2)dg, z≤x
4 Rs(n) ν ≥ 0, n = 2ν + 2 ≥ 2, τ ∼ Exp(λ), independent of R(n)

4. BESSEL PROCESS OF ORDER ν

1. Exponential stopping

( √ √
(n)
2λz ν+1 x−ν Iν (x 2λ)Kν (z 2λ)dz, 0 ≤ x ≤ z
1.0.5 Px (Rτ ∈ dz) = √ √
2λz ν+1 x−ν Kν (x 2λ)Iν (z 2λ)dz, z ≤ x
p p
or = 2λz ν+1 x−ν Iν ((z + x − |z − x|) λ/2)Kν ((z + x + |z − x|) λ/2)dz

2 2
Px Rt(n) ∈ dz = z ν+1 x−ν t−1 e−(x +z )/2t Iν (xz/t)dz


1.0.6

x− I (x√2)
1.1.2 Px sup Rs(n) ≥ y = − √ , 0≤x≤y
0≤s≤τ y I (y 2)

∞
X 2x− J (j;k x=y) −jν,k 2
t/2y 2
Px sup Rs(n) ≥ y = 1 −


1.1.4 − J
e , 0≤x≤y
0≤s≤t j ;k y (j
 +1 ;k )
k=1

Px sup Rs(n) ≥ y, Rτ(n) ∈ dz

1.1.6
0≤s≤τ

2z  +1 x− √ √ √
= √ I (x 2λ)Iν (z 2λ)Kν (y 2λ)dz, x∨z ≤y
I (y 2) ν

x− K (x√2)
1.2.2 Px inf Rs(n) ≤ y = − √
0≤s≤τ y K (y 2)
y≤x

x−2
1.2.4 Px inf Rs(n) ≤ y = −2 y≤x
0≤s<∞ y
(1)

inf Rs(n) ≤ y, Rτ(n) ∈ dz

1.2.6 Px
0≤s≤τ

2z  +1 x− √ √ √
= √ K (x 2λ)Kν (z 2λ)Iν (y 2λ)dz, y ≤x∧z
K (y 2) ν

√ √
 2 r +1 x− I (x 2)K (r 2)
 1 − 1 + 2 rI (r√2)K (r√2) , 0 ≤ x ≤ r

 
1.3.1 Ex e−γ`(τ,r) =  +1 x− K (x√2)I (r√2)
 1− 2 r
, r≤x

√ √
1 + 2 rI (r 2)K (r 2)

© Springer Basel 2015 379

A.N. Borodin, P. Salminen, Handbook of Brownian Motion - Facts and Formulae,
Probability and Its Applications, DOI 10.1007/978-3-0348-8163-0_11
380 4. BESSEL PROCESS OF ORDER 

4 Rs(n) ν ≥ 0, n = 2ν + 2 ≥ 2, τ ∼ Exp(λ), independent of R(n)

 r −1 x− I (x√2)

√  2I 2 (r√2)K (r√2) dy, 0 ≤ x ≤ r
yI−1 (r 2)

 

 
1.3.2 Px `(τ, r) ∈ dy = exp − √
 −1 x− K (x√2)
2rK (r 2)   r √ √ dy, r ≤ x
2I (r 2)K2 (r 2)
√
 x− I (x 2)
1 − √ , 0≤x≤r
r− I (r 2)



(1) Px `(τ, r) = 0 = −
√
 1 − x K (x√ 2) , r ≤ x

r− K (r 2)

 r
 1 −  + r, 0≤x≤r
1.3.3 Ex e−γ`(∞,r) = 2 +1 2
 1− r x −
, r≤x
(1) + r

 
 e−νy/r dy, 0≤x≤r
r


1.3.4 Px `(∞, r) ∈ dy = 2
 x −
e−νy/r dy, r ≤ x
(1) r1−2

x−2
(2) Px `(∞, r) = 0 = 1 − −2
r
r≤x

√ √
Ex e−γ`(τ,r) , Rτ(n) ∈ dz = (2λ)ν+1 z 2ν+1 Fν (x 2λ, z 2λ)dz

1.3.5
√ √ √ √
2 (2rz )2 +1 F (r 2; z 2)F (x 2; r 2)
− √ √ dz
1 + 2 r2 +1 (2) F (r 2; r 2)

Px `(τ, r) ∈ dy, Rτ(n) ∈ dz

1.3.6

√ √ √ √
F (r 2; z 2)F (x 2; r 2) y
 
= √ √ exp − √ √ dydz
r2 +1 z −2 −1 F2 (r 2; r 2) 2(2) r2 +1 F (r 2; r 2)

√ √
Px `(τ, r) = 0, Rτ(n) ∈ dz = (2λ)ν+1 z 2ν+1 Fν (x 2λ, z 2λ)dz


(1)

√ √ √ √
F (r 2; z 2)F (x 2; r 2)
−(2λ)ν+1 z 2ν+1 √ √ dz
F (r 2; r 2)
 1. EXPONENTIAL STOPPING 381

4 Rs(n) ν ≥ 0, n = 2ν + 2 ≥ 2, τ ∼ Exp(λ), independent of R(n)

 Z τ

1.4.1 Ex exp −γ 1I[r,∞) Rs(n) ds
0

√ √
( + )−1=2 r x− I (x 2)K +1 (r 2 + 2 )
0≤x =1− √ √ √ √ √ √
I +1 (r 2)K (r 2 + 2 ) +  + K +1 (r 2 + 2 )I (r 2)
x≤r

√ √ √
 ( + )−1 r x− K (x 2 + 2 )I +1 (r 2)
r≤x = + √ √ √ √ √ √
+ I +1 (r 2)K (r 2 + 2 ) +  + K +1 (r 2 + 2 )I (r 2)

√
  1 − x− I (x√ 2) , 0 ≤ x ≤ r

Z τ
(n)
r− I (r 2)


1.4.2 Px 1I[r,∞) Rs ds = 0 =
(1) 0 
0, r≤x

Z τ   0, 0≤x≤r
√
x− K (x 2)
(n)


(2) Px 1I[r,∞) Rs ds = τ =
 1 − −  √ , r≤x
0
r K (r 2)

n  Z τ
 o
1.4.5 Ex exp −γ 1I[r,∞) Rs(n) ds , Rτ(n) ∈ dz
0

√ √ √ √
z≤r = (2λ)ν+1 z 2ν+1 Fν (x 2λ, z 2λ)dz + 2λz ν+1 x−ν Iν (x 2λ)Iν (z 2λ)
x≤r
√ √ √ √ √ √
K +1 (r 2)K (r 2 + 2 ) −  + K +1 (r 2 + 2 )K (r 2)
× √ √ √ √ √ √ dz
I +1 (r 2)K (r 2 + 2 ) +  + K +1 (r 2 + 2 )I (r 2)

√ √ √
 2r−1 z  +1 x− K (x 2 + 2 )I (z 2)
z≤r = √ √ √ √ √ √ dz
I +1 (r 2)K (r 2 + 2 ) +  + K +1 (r 2 + 2 )I (r 2)
r≤x

√ √ √
 2r−1 z  +1 x− I (x 2)K (z 2 + 2 )
r≤z = √ √ √ √ √ √ dz
I +1 (r 2)K (r 2 + 2 ) +  + K +1 (r 2 + 2 )I (r 2)
x≤r

p p
r≤z = 2λ(2λ + 2γ)ν z 2ν+1 Fν (x 2λ + 2γ, z 2λ + 2γ)dz
r≤x
p p
+2λz ν+1 x−ν Kν (x 2λ + 2γ)Kν (z 2λ + 2γ)
√ √ √ √ √ √
 + I +1 (r 2 + 2 )I (r 2) − I +1 (r 2)I (r 2 + 2 )
×√ √ √ √ √ √ dz
I +1 (r 2)K (r 2 + 2 ) +  + K +1 (r 2 + 2 )I (r 2)
382 4. BESSEL PROCESS OF ORDER 

4 Rs(n) ν ≥ 0, n = 2ν + 2 ≥ 2, τ ∼ Exp(λ), independent of R(n)

 Z τ

1.5.1 Ex exp −γ 1I[0,r] Rs(n) ds
0

√ √ √
 ( + )−1 r x− I (x 2 + 2 )K +1 (r 2)
0≤x = + √ √ √ √ √ √
+  + I +1 (r 2 + 2 )K (r 2) + K +1 (r 2)I (r 2 + 2 )
x≤r
√ √
( + )−1=2 r x− K (x 2)I +1 (r 2 + 2 )
r≤x =1− √ √ √ √ √ √
 + I +1 (r 2 + 2 )K (r 2) + K +1 (r 2)I (r 2 + 2 )


Z τ   0, 0≤x≤r
(n)
 √
x K (x 2)


1.5.2 Px 1I[0,r] Rs ds = 0 =
 1 − −  √ , r ≤ x
−

(1) 0
r K (r 2)

√
2x− I (x 2 )

√ √ , x≤r
∞
2 r1− I−1 (r 2 )
 Z  

Rs(n) ds =


1.5.3 Ex exp −γ 1I[0,r] −2
√
(1) 0  1 − x I +1 (r√ 2 ) , r ≤ x

r−2 I−1 (r 2 )

n  Z τ
 o
1.5.5 Ex exp −γ 1I[0,r] Rs(n) ds , Rτ(n) ∈ dz
0

p p
z≤r = 2λ(2λ + 2γ)ν+1 z 2ν+1 Fν (x 2λ + 2γ, z 2λ + 2γ)dz
x≤r
p p
+2λz ν+1 x−ν Iν (x 2λ + 2γ)Iν (z 2λ + 2γ)
√ √ √ √ √ √
 + K +1 (r 2 + 2 )K (r 2) − K +1 (r 2)K (r 2 + 2 )
× √ √ √ √ √ √ dz
 + I +1 (r 2 + 2 )K (r 2) + K +1 (r 2)I (r 2 + 2 )

√ √ √
 2r−1 z  +1 x− K (x 2)I (z 2 + 2 )
z≤r =√ √ √ √ √ √ dz
 + I +1 (r 2 + 2 )K (r 2) + K +1 (r 2)I (r 2 + 2 )
r≤x

√ √ √
 2r−1 z  +1 x− I (x 2 + 2 )K (z 2)
r≤z = √ √ √ √ √ √ dz
 + I +1 (r 2 + 2 )K (r 2) + K +1 (r 2)I (r 2 + 2 )
x≤r
√ √ √ √
r≤z = (2λ)ν+1 z 2ν+1 Fν (x 2λ, z 2λ)dz + 2λz ν+1 x−ν Kν (x 2λ)Kν (z 2λ)
r≤x
√ √ √ √ √ √
I +1 (r 2)I (r 2 + 2 ) −  + I +1 (r 2 + 2 )I (r 2)
×√ √ √ √ √ √ dz
 + I +1 (r 2 + 2 )K (r 2) + K +1 (r 2)I (r 2 + 2 )
 1. EXPONENTIAL STOPPING 383

4 Rs(n) ν ≥ 0, n = 2ν + 2 ≥ 2, τ ∼ Exp(λ), independent of R(n)

 Z τ


p1I[0,r) Rs(n) + q1I[r,∞) Rs(n) ds


1.6.1 Ex exp −
0
 4(q − p)r x− I (xp )K +1 (rq )
0≤x = − 2
+p p q (p I +1 (rp )K (rq ) + q K +1 (rq )I (rp ))
x≤r

 4(q − p)r x− K (xq )I +1 (rp )

r≤x = + 2
+q p q (p I +1 (rp )K (rq ) + q K +1 (rq )I (rp ))

n  Z τ

 o
p1I[0,r) Rs(n) + q1I[r,∞) Rs(n) ds , Rτ(n) ∈ dz


1.6.5 Ex exp −
0

z≤r = 2λΥp2ν z 2ν+1 Fν (xΥp , zΥp )dz

x≤r
2z  +1 I (xp )I (zp )(p K +1 (rp )K (rq ) − q K +1 (rq )K (rp ))
+ dz
x (p I +1 (rp )K (rq ) + q K +1 (rq )I (rp ))

2r−1 z  +1 x− K (xq )I (zp )

z≤r = dz
p I +1 (rp )K (rq ) + q K +1 (rq )I (rp )
r≤x
2r−1 z  +1 x− I (xp )K (zq )
r≤z = dz
p I +1 (rp )K (rq ) + q K +1 (rq )I (rp )
x≤r

r≤z = 2λΥq2ν z 2ν+1 Fν (xΥq , zΥq )dz

r≤x
2z  +1 K (xq )K (zq )(q I +1 (rq )I (rp ) − p I +1 (rp )I (rq ))
+ dz
x (p I +1 (rp )K (rq ) + q K +1 (rq )I (rp ))

 2Z t  2

1 x sh(t )

1.9.3 Ex exp − (Rs(n) )2 ds = exp −
2 0
(ch(t )) +1 2 ch(t )
2 t
 Z 
(1) Ex exp −β(Rt(n) )2 − (Rs(n) )2 ds
2 0
1 x2 ( sh(t ) + 2 ch(t ))
 
= exp −
(ch(t ) + 2 −1 sh(t )) +1 2(ch(t ) + 2 −1 sh(t ))
 t
Z

2 
1.9.4 Px Rs(n) ds ∈ dy = ecy (0, ν + 1, t, 0, x2 /2)dy
0

√
Υs := 2λ + 2s
384 4. BESSEL PROCESS OF ORDER 

4 Rs(n) ν ≥ 0, n = 2ν + 2 ≥ 2, τ ∼ Exp(λ), independent of R(n)

∞
Z t
(n)

2 
ν+1
X (−1)k ( + k + 1) 
( + 2k + 1)t 
(1) P0 Rs ds < y = 2 Erfc √
0
( + 1) k! 2y
k=0

n  2Z τ  o
1.9.5 Ex exp − (Rs(n) )2 ds ; Rτ(n) ∈ dz
2 0
0<x
 (1=2 + =2 + =2 )
x≤z = M−λ/2γ,ν/2 (γx2 )W−λ/2γ,ν/2 (γz 2 )dz
(1 +  )x +1 z −
 (1=2 + =2 + =2 )
z≤x = W−λ/2γ,ν/2 (γx2 )M−λ/2γ,ν/2 (γz 2 )dz
(1 +  )x +1 z −

n  2Z t  o
1.9.7 Ex exp − (Rs(n) )2 ds ; Rt(n) ∈ dz
2 0
z  +1 x− (x2 + z 2 ) ch(t )   xz 

= exp − Iν dz
sh(t ) 2 sh(t ) sh(t )

z  +1
Z t
2 
1.9.8 Px Rs(n) ds ∈ dy, Rt(n) ∈ dz =  isy (ν, t, 0, (x2 + z 2 )/2, xz/2)dydz
0
x

∞ √ √ √
2x− I (x 2 + 2 )I +1 (y 2)
Z
1.12.1 Ex e−γ Ȟ(τ ) = √ √ dy
x y− I (y 2 + 2 )I (y 2)

x √ √ √
2x− K (x 2 + 2 )K +1 (y 2)
Z
(1) Ex e−γ Ĥ(τ ) = √ √ dy
0 y− K (y 2 + 2 )K (y 2)

√ √
∞
2z  +1 x− I (x 2 + 2 )I (z 2)
Z
Ex e−γ Ȟ(τ ) ; Rτ(n) ∈ dz =

1.12.5 √ √ dy dz
x∨z yI (y 2 + 2 )I (y 2)

√ √
x∧z
2z  +1 x− K (x 2 + 2 )K (z 2)
Z
−γ Ĥ(τ ) (n)

(1) Ex e ; Rτ ∈ dz = √ √ dy dz
0 yK (y 2 + 2 )K (y 2)

√ √ √
2x− I (x 2 + 2 )I +1 (y 2)
Ex e−γ Ȟ(τ ) ; sup Rs(n) ∈ dy =

1.13.1 √ √ dy
0≤s≤τ y− I (y 2 + 2 )I (y 2)
x<y

Px Ȟ(τ ) ∈ dv, sup Rs(n) ∈ dy

1.13.2
0≤s≤τ
x<y
√ √ ∞
2I +1 (y 2) X x− j;k J (j;k x=y) −λv−jν,k
2
v/2y 2
= √ 2  e dvdy
I (y 2) y J +1 (j;k )
−
k=1
 1. EXPONENTIAL STOPPING 385

4 Rs(n) ν ≥ 0, n = 2ν + 2 ≥ 2, τ ∼ Exp(λ), independent of R(n)

Ex e−γ Ȟ(t) ; sup Rs(n) ∈ dy


1.13.3
0≤s≤t
x<y
∞ ∞
X x− j;k J (j;k x=y) −γt−jν,k
2
t/2y 2
X 2 2
=2 3  e ∗ e−jν,k t/2y dy
y J +1 (j;k )
−
k=1 k=1

Px Ȟ(t) ∈ dv, sup Rs(n) ∈ dy

1.13.4
0≤s≤t
x<y
∞ ∞
X x− j;k J (j;k x=y) −jν,k
2
v/2y 2
X 2 2
= 21I[0,t] (v) e e−jν,k (t−v)/2y dvdy
y3− J +1 (j;k )
k=1 k=1

Ex e−γ Ȟ(τ ) ; sup Rs(n) ∈ dy, Rτ(n) ∈ dz


1.13.5 x∨z <y
0≤s≤τ
√ √
2z  +1 x− I (x 2 + 2 )I (z 2)
= √ √ dydz
yI (y 2 + 2 )I (y 2)

Px Ȟ(τ ) ∈ dv; sup Rs(n) ∈ dy, Rτ(n) ∈ dz

1.13.6 x∨z <y
0≤s≤τ
∞ √
X j;k J (j;k x=y) −λv−jν,k
2
v/2y 2 I (z 2)
= 2λz ν+1 x−ν e √ dvdydz
y3 J +1 (j;k ) I (y 2)
k=1

Ex e−γ Ȟ(t) ; sup Rs(n) ∈ dy, Rt(n) ∈ dz


1.13.7 x∨z <y
0≤s≤t
∞ ∞
2z  +1 X j;k J (j;k x=y) −γt−jν,k
2
t/2y 2
X j;k J (j;k x=y) −jν,k
2
t/2y 2
= e ∗ e dydz
y 5 x J +1 (j;k ) J +1 (j;k )
k=1 k=1

Px Ȟ(t) ∈ dv, sup Rs(n) ∈ dy, Rt(n) ∈ dz

1.13.8 x∨z <y
0≤s≤t
v<t
∞ ∞
2z  +1 X j;k J (j;k x=y) −jν,k
2
v/2y 2
X j;k J (j;k z=y) −jν,k
2
(t−v)/2y 2
= e e dvdydz
y 5 x J +1 (j;k ) J +1 (j;k )
k=1 k=1

√ √ √
 −γ Ĥ(τ ) (n) 2x− K (x 2 + 2 )K +1 (y 2)
1.14.1 Ex e ; inf Rs ∈ dy = √ √ dy
0≤s≤τ y− K (y 2 + 2 )K (y 2)
y<x
386 4. BESSEL PROCESS OF ORDER 

4 Rs(n) ν ≥ 0, n = 2ν + 2 ≥ 2, τ ∼ Exp(λ), independent of R(n)

√
2x− K (x 2 )
Ex e−γ Ĥ(∞) ; inf Rs(n) ∈ dy =

1.14.3 √ dy
0≤s<∞ y1− K (y 2 )
(1)
y<x

Ex e−γ Ĥ(τ ) ; inf Rs(n) ∈ dy, Rτ(n) ∈ dz


1.14.5 0<y <x∧z
0≤s≤τ
√ √
2z  +1 x− K (x 2 + 2 )K (z 2)
= √ √ dydz
yK (y 2 + 2 )K (y 2)

√ √ √ √
S (b 2; x 2)+ S (x 2; a 2)
Px a < inf Rs(n) , sup Rs(n) < b =1 −


1.15.2 √ √
0≤s≤τ 0≤s≤τ S (b 2; a 2)

Px a < inf Rs(n) , sup Rs(n) < b, Rτ(n) ∈ dz

1.15.6
0≤s≤τ 0≤s≤τ

√ √
(2) +1 z 2 +1 S (b 2; (z + x + |z − x|) =2)S ((z + x − |z − x|) =2; a 2)
p p
= √ √ dz
S (b 2; a 2)
√ √ √ √
 S (b 2; z 2)S (x 2; a 2)
 (2λ)ν+1 z 2ν+1
 √ √
S (b 2; a 2)
dz, a ≤ x ≤ z
or = √ √ √ √
 (2λ)ν+1 z 2ν+1 S (b 2; x √
 2)S (z 2; a 2)
√ dz, z ≤ x ≤ b
S (b 2; a 2)

Ex e−γ`(τ,r) ; sup Rs(n) < b


1.16.1
0≤s≤τ
r<b
√ √ √
(2 S (b 2; r 2) + (2)− r−2 −1 )x− r I (x 2)
0≤x =1− √ √ √ √
2 S (b 2; r 2)I (r 2) + (2)− r− −1 b− I (b 2)
x≤r
√ √
S (x 2; r 2)
r≤x =1− √ √
S (b 2; r 2)
x≤b
√ √ √ √ √
(2 S (b 2; r 2) + (2)− r−2 −1 )I (r 2)S (b 2; x 2)
− √ √ √ √ √ √
(2 S (b 2; r 2)I (r 2) + (2)− r− −1 b− I (b 2))S (b 2; r 2)

√
y(2)− r− −1 b− I (b 2)
 
Px `(τ, r) ∈ dy, sup Rs(n) < b = exp −


1.16.2 √ √ √
0≤s≤τ 2S (b 2; r 2)I (r 2)
r<b
 (b− I (b 2) − r− I (r 2))x− I (x√2)
√ √
  
√ √ √ dy, 0≤x≤r
2(2) rS (b 2; r 2)I2 (r 2)


× −
√ −
√ √ √
 (b I (b 2) − r I√ (r 2√
 ))S (b 2; x 2)
√ dy, r ≤ x ≤ b
2(2) r +1 S2 (b 2; r 2)I (r 2)
 1. EXPONENTIAL STOPPING 387

4 Rs(n) ν ≥ 0, n = 2ν + 2 ≥ 2, τ ∼ Exp(λ), independent of R(n)

Px `(τ, r) = 0, sup Rs(n) < b

(1)
0≤s≤τ
√
 x− I (x 2)
 1 − r− I (r√2) ,
 0≤x≤r

= √ √ √ √
 1 − S (b 2; x 2√) + S√(x 2; r 2) , r ≤ x ≤ b

S (b 2; r 2)

Ex e−γ`(τ,r) ; sup Rs(n) < b, Rτ(n) ∈ dz


1.16.5 r∨z <b
0≤s≤τ
√
(2) +1 z 2 +1 S (r 2; (z + x + |z − x|) =2)I ((z + x − |z − x|) =2)
p p
z≤r = √ dz
((z + x − |z − x|)=2) r− I (r 2)
x≤r
√ √ √ √
2z  +1 r−1 x− S (b 2; r 2)I (z 2)I (x 2)
+ √ √ √ √ √ dz
(2 S (b 2; r 2)I (r 2) + (2)− r− −1 b− I (b 2))I (r 2)
√ √ √
2z  +1 r− −1 S (b 2; x 2)I (z 2)
z≤r = √ √ √ √ dz
2 S (b 2; r 2)I (r 2) + (2)− r− −1 b− I (b 2)
r≤x

√ √ √
2z 2 +1 r− −1 x− S (b 2; z 2)I (x 2)
r≤z = √ √ √ √ dz
2 S (b 2; r 2)I (r 2) + (2)− r− −1 b− I (b 2)
x≤r

√ √
(2) +1 z 2 +1 S (b 2; (z + x + |z − x|) =2)S ((z + x − |z − x|) =2; r 2)
p p
r≤z = √ √ dz
S (b 2; r 2)
r≤x
√ √ √ √ √
2z 2 +1 r−2 −1 S (b 2; z 2)I (r 2)S (b 2; x 2)
+ √ √ √ √ √ √ dz
(2 S (b 2; r 2)I (r 2) + (2)− r− −1 b− I (b 2))S (b 2; r 2)

Px `(τ, r) ∈ dy, sup Rs(n) < b, Rτ(n) ∈ dz

1.16.6 r∨z <b
0≤s≤τ

√ √ √
z  +1 r−1 x− I (z 2)I (x 2) y(2)− r− −1 b− I (b 2)
 
z≤r = √ exp − √ √ √ dydz
I2 (r 2) 2S (b 2; r 2)I (r 2)
x≤r

√ √ √ √
z  +1 r− −1 S (b 2; x 2)I (z 2) y(2)− r− −1 b− I (b 2)
 
z≤r = √ √ √ exp − √ √ √ dydz
S (b 2; r 2)I (r 2) 2S (b 2; r 2)I (r 2)
r≤x

√ √ √ √
z 2 +1 S (b 2; z 2)I (x 2) y(2)− r− −1 b− I (b 2)
 
r≤z =  +1  √ √ √ exp − √ √ √ dydz
r x S (b 2; r 2)I (r 2) 2S (b 2; r 2)I (r 2)
x≤r
388 4. BESSEL PROCESS OF ORDER 

4 Rs(n) ν ≥ 0, n = 2ν + 2 ≥ 2, τ ∼ Exp(λ), independent of R(n)

√ √ √ √ √
S (b 2; z 2)S (b 2; x 2) y(2)− r− −1 b− I (b 2)
 
r≤z = √ √ exp − √ √ √ dydz
z −2 −1 r2 +1 S2 (b 2; r 2) 2S (b 2; r 2)I (r 2)
r≤x

Px `(τ, r) = 0, sup Rs(n) < b, Rτ(n) ∈ dz

(1)
0≤s≤τ

√
(2) +1 z 2 +1 S (r 2; (z + x + |z − x|) =2)I ((z + x − |z − x|) =2)
p p
z≤r = √ dz
((z + x − |z − x|)=2) r− I (r 2)
x≤r

√ √
(2) +1 z 2 +1 S (b 2; (z + x + |z − x|) =2)S ((z + x − |z − x|) =2; r 2)
p p
r≤z = √ √ dz
S (b 2; r 2)
r≤x

Ex e−γ`(τ,r) ; a < inf Rs(n)


1.17.1
0≤s≤τ
a<r

√ √ √ √
S (r 2; x 2) + S (x 2; a 2)
a≤x =1− √ √
S (r 2; a 2)
x≤r
√ √ √ √
(a− K (a 2) − r− K (r 2))S (x 2; a 2)
+  +1
√ √ √ √ √ √
(2(2) r K (r 2)S (r 2; a 2) + a− K (a 2))S (r 2; a 2)

√
x− K (x 2)
r≤x =1− √
r− K (r 2)
√ √ √
(a− K (a 2) − r− K (r 2))x− K (x 2)
+ √ √ √ √ √
(2(2) r +1 K (r 2)S (r 2; a 2) + a− K (a 2))r− K (r 2)

√
y(2)− r− −1 a− K (a 2)
 
Px `(τ, r) ∈ dy, a < inf Rs(n) = exp −


1.17.2 √ √ √
0≤s≤τ 2K (r 2)S (r 2; a 2)
a<r

 (a− K (a√2) − r− K (r√2))S (x√2; a√2)

  
√ √ √ dy, a ≤ x ≤ r
2(2) r +1 K (r 2)S2 (r 2; a 2)


× −
√ −
√ −
√
 (a K (a 2) − r √ K (r 2√))x √K (x 2) dy,

r≤x
2(2) rK2 (r 2)S (r 2; a 2)

Px `(τ, r) = 0, a < inf Rs(n)

(1)
0≤s≤τ
 1. EXPONENTIAL STOPPING 389

4 Rs(n) ν ≥ 0, n = 2ν + 2 ≥ 2, τ ∼ Exp(λ), independent of R(n)

√ √ √ √
 S (r 2; x 2) + S (x 2; a 2)
 1−
 √ √
S (r 2; a 2)
, a≤x≤r

= −
√
 1 − x K (x√ 2) ,

r≤x
r  K (r 2)
−

Ex e−γ`(∞,r) ; a < Rs(n)


1.17.3 inf
0≤s<∞
(1)
a<r
a−2 − x−2

 a−2 +  −1 r (a−2 − r−2 ) ,
 a≤x≤r
= −2
 1− x (a−2 − r−2 )x−2
+ −2 , r≤x≤b

r−2 (a +  −1 r (a−2 − r−2 ))r−2

Rs(n) =


1.17.4 Px `(∞, r) ∈ dy, a < inf
0≤s<∞
(1)
a<r
 (a−2 − x−2 ) a−2 y
  
 r(a−2 − r−2 ) exp − r(a−2 − r−2 ) dy, a ≤ x ≤ r

= −2
 x a−2 y
 
exp − dy, r ≤ x.

r1−2 r(a−2 − r−2 )

Ex e−γ`(τ,r) ; a < inf Rs(n) , Rτ(n) ∈ dz


1.17.5 a<r∧z
0≤s≤τ

√ √
(2) +1 z 2 +1 S (r 2; (z + x + |z − x|) =2)S ((z + x − |z − x|) =2; a 2)
p p
z≤r = √ √ dz
S (r 2; a 2)
x≤r
√ √ √ √ √
2z 2 +1 r− K (r 2)S (z 2; a 2)S (x 2; a 2)
+ √ √ √ √ √ √ dz
(2 r +1 K (r 2)S (r 2; a 2) + (2)− a− K (a 2))S (r 2; a 2)

√ √ √
2z 2 +1 x− K (x 2)S (z 2; a 2)
z≤r = √ √ √ √ dz
2 r +1 K (r 2)S (r 2; a 2) + (2)− a− K (a 2)
r≤x

√ √ √
2z  +1 K (z 2)S (x 2; a 2)
r≤z = +1
√ √ √ √ dz
2 r K (r 2)S (r 2; a 2) + (2)− a− K (a 2)

x≤r

√
(2) +1 z 2 +1 K ((z + x + |z − x|) =2)S ((z + x − |z − x|) =2; r 2)
p p
r≤z = √ dz
((z + x + |z − x|)=2) r− K (r 2)
r≤x
390 4. BESSEL PROCESS OF ORDER 

4 Rs(n) ν ≥ 0, n = 2ν + 2 ≥ 2, τ ∼ Exp(λ), independent of R(n)

√ √ √ √
2z  +1 x− K (z 2)S (r 2; a 2)K (x 2)
+ √ √ √ √ √ dz
(2 r +1 K (r 2)S (r 2; a 2) + (2)− a− K (a 2))r− K (r 2)

Px `(τ, r) ∈ dy, a < inf Rs(n) , Rτ(n) ∈ dz

1.17.6 a<r∧z
0≤s≤τ

√ √ √ √ √
S (z 2; a 2)S (x 2; a 2) y(2)− r− −1 a− K (a 2)
 
z≤r = √ √ exp − √ √ √ dydz
z −2 −1 r2 +1 S2 (r 2; a 2) 2K (r 2)S (r 2; a 2)
x≤r

√ √ √ √
z 2 +1 x− K (x 2)S (z 2; a 2) y(2)− r− −1 a− K (a 2)
 
z≤r = √ √ √ exp − √ √ √ dydz
r +1 K (r 2)S (r 2; a 2) 2K (r 2)S (r 2; a 2)
r≤x

√ √ √ √
z  +1 K (z 2)S (x 2; a 2) y(2)− r− −1 a− K (a 2)
 
r≤z =  +1 √ √ √ exp − √ √ √ dydz
r K (r 2)S (r 2; a 2) 2K (r 2)S (r 2; a 2)
x≤r

√ √ √
z  +1 x− K (z 2)K (x 2) y(2)− r− −1 a− K (a 2)
 
r≤z = √ exp − √ √ √ dydz
rK2 (r 2) 2K (r 2)S (r 2; a 2)
r≤x

Px `(τ, r) = 0, a < inf Rs(n) , Rτ(n) ∈ dz

(1)
0≤s≤τ

√ √
(2) +1 z 2 +1 S (r 2; (z + x + |z − x|) =2)S ((z + x − |z − x|) =2; a 2)
p p
z≤r = √ √ dz
S (r 2; a 2)
x≤r

√
(2) +1 z 2 +1 K ((z + x + |z − x|) =2)S ((z + x − |z − x|) =2; r 2)
p p
r≤z = √ dz
((z + x + |z − x|)=2) r− K (r 2)
r≤x

r2+1 r−2 + x−2 − |r−2 − x−2 |

1.18.3 Ex exp −γ`(∞, r) − η`(∞, u) = 1 −
2 ru 1 − (r=u)2 + r + u +  2



(1)

 u2+1 + ur u2 − r2 x−2 + u−2 − |u−2 − x−2 |

r<u −
2 ru 1 − (r=u)2 + r + u +  2

1.18.4 Px `(∞, r) ∈ dy, `(∞, u) ∈ dg
(1)
 1. EXPONENTIAL STOPPING 391

4 Rs(n) ν ≥ 0, n = 2ν + 2 ≥ 2, τ ∼ Exp(λ), independent of R(n)

   (ru)−−1=2 √yg 
 2 r−2−1 r−2−1 (uy + rg)

x≤r = exp − I0 dydg
u(r−2 − u−2 ) u(r−2 − u−2 ) r−2 − u−2
r<u

  (ru)−−1=2 √yg 
2 r−2−1 (uy + rg) x−2 − u−2
 h
r<x = exp − I
u(r−2 − u−2 )2 u(r−2 − u−2 ) r2+1 0
r−2 − u−2
x<u

  (ru)−−1=2 yg i
r−2 − x−2 g
√ √
+ √ I dydg
(ru)+1=2 y 1
r−2 − u−2

 2 x−2 r−−1=2 g r−2−1 (uy + rg)

   (ru)−−1=2 √yg 
√ 
r<u = √ exp − I1 dydg
u−+3=2 (r−2 − u−2 ) y u(r−2 − u−2 ) r−2 − u−2
u≤x

(2) Px `(∞, r) = 0, `(∞, u) ∈ dg

 −2 −2 −2

x (r − x ) r−2 g
 
 u1−2 (r−2 − u−2 ) exp − u(r−2 − u−2 ) dg, r ≤ x ≤ u

= −2
 x r−2 g
 
exp − −2 −2 dg, u≤x

u1−2 u(r − u )

x−2
(3) Px `(∞, r) = 0, `(∞, u) = 0 = 1 − −2 , u≤x
u

τ  (2)1=4−=2 (2 + 1) sh +3=2 (yp=2)

ds
Z
1.19.2 Px ∈ dy =
Rs(n) 2x +1=2 ch +3=2 (y =2)
p
0<x 0

 √ √ √
x 2 ch(y 2) 2x 2 
 
× exp − √ M−ν−3/2,ν √ dy
sh(y 2) sh(y 2)

τ
ds
n  Z  o
(n)
1.19.5 Ex exp −γ ; R ∈ dz
0 Rs(n) τ
0<x
√ √
 ( + 1=2 + = 2)z  +1=2 √ √
x≤z = √
 +1= 2 M−γ/√2λ,ν (2x 2λ)W−γ/√2λ,ν (2z 2λ)dz
2 (2 + 1)x

√ √
 ( + 1=2 + = 2)z  +1=2 √ √
z≤x = √
 +1= 2 W−γ/√2λ,ν (2x 2λ)M−γ/√2λ,ν (2z 2λ)dz
2 (2 + 1)x
392 4. BESSEL PROCESS OF ORDER 

4 Rs(n) ν ≥ 0, n = 2ν + 2 ≥ 2, τ ∼ Exp(λ), independent of R(n)

τ
ds
Z 
1.19.6 Px ∈ dy, Rτ(n) ∈ dz
0 Rs(n)
0<x
√  (x + z )√2 ch(yp=2)   √
 2z  +1 2 2xz 
=  exp − I2ν dydz
x sh(y =2) sh(y =2) sh(y =2)
p p p

z  +1
t
ds √
Z 
1.19.8 Px (n)
∈ dy, Rt(n) ∈ dz =  ist (2ν, y/2, 0, x + z, xz)dydz
0 Rs x
0<x

 2Z t
ds

1.20.3 Ex exp − (n)
2
2 0 Rs
0<x
(2t)( +1)=2 (1 + =2 +  2 + 2 =2) −x2 /4t  2
x
p
= e M √ 2 2
x +1 (1 +  2 + 2 ) −ν/2−1/2, ν +γ /2 2t
p

Z t
ds

2(2t)( +1)=2 −ν 2 y/2−x2 /4t 
 + 1 x2

1.20.4 Px (n)
2
∈ dy = e m , dy
0 Rs x +1 2y
2 4t
0<x

 2Z τ
ds
n  o
(n)
1.20.5 Ex exp −
; Rτ ∈ dz
2 0 Rsn 2
( )

√
√

 2λx−ν z ν+1 I√ν 2 +γ 2 x 2λ K√ν 2 +γ 2 z 2λ dz, 0 < x ≤ z


= √
√

 2λx−ν z ν+1 K√ 2 2 x 2λ I√ 2 2 z 2λ dz, z ≤ x
ν +γ ν +γ

z  +1 −ν 2 y/2 √ √
τ
ds
Z 
(n)
1.20.6 Px
∈ dy, Rτ ∈ dz = e kiy/2 (x 2λ, z 2λ)dydz
0 Rsn 2
( )
x 
0<x

 2Z
z  +1
t
ds xz
n  o 2 2
 
1.20.7 Ex exp −
; Rt
(n)
∈ dz =  e−(x +z )/2t I√ν 2 +γ 2 dz
2 0 Rsn 2
( )
x t t

z  +1 −(x2 +z2 )/2t−ν 2 y/2

t
ds xz

Z 
1.20.8 Px (n)
2
∈ dy, Rt(n) ∈ dz = e iy/2 dydz
0 Rs 2tx t
0<x

 Z t
p2 q2 Rs(n) 2   (2 sh(tq))( +1)=2


1.21.3 Ex exp − (n) 2
+ ds =  +1
2 Rs 2 x (q ch(tq))( +1)=2


0<x 0
 2  2 
(1 + =2 +  2 + p2 =2) x q ch(2tq) x q
p 
× exp − M √
(1 +  2 + p2 ) 2 sh(2tq) ν 2+p2
sh(2tq)
p
− ν+1
2 , 2
 1. EXPONENTIAL STOPPING 393

4 Rs(n) ν ≥ 0, n = 2ν + 2 ≥ 2, τ ∼ Exp(λ), independent of R(n)

 2Z t t
q ds
n Z

2  o
(1) Ex exp − Rs(n) ds ; (n)
2
∈ dy
2 0 0 Rs
2(2 sh(tq))( +1) = 2  2
 y x2 q ch(2tq)
 
+1 x2 q

0<x =  +1 ( +1) = 2 exp − − m2y , dy
x (q ch(tq)) 2 2 sh(2tq) 2 2 sh(2tq)

 Z τ
p2 q2 Rs(n) 2   (n)
n
o
1.21.5 Ex exp − (n)
2
+ ds ; Rτ ∈ dz
0 2 Rs 2
0<x

 ((1 + p2 +  2 + =q)=2)
p
x≤z = M √ (qx2 )W−λ/2q,√p2 +ν 2 /2 (qz 2 )dz
q (1 + p2 +  2 )x +1 z − −λ/2q, p2 +ν 2 /2
p

 ((1 + p2 +  2 + =q)=2)
p
z≤x = W √ (qx2 )M−λ/2q,√p2 +ν 2 /2 (qz 2 )dz
q (1 + p2 +  2 )x +1 z − −λ/2q, p2 +ν 2 /2
p

 Z t
p2 q2 Rs(n) 2   (n)
n
o
1.21.7 Ex exp − (n)
2
+ ds ; Rt ∈ dz
0 2 Rs 2

qz  +1 (x2 + z 2 )q ch(tq)  √ xzq

  
=  exp − I p2 +ν 2 dz
x sh(tq) 2 sh(tq) sh(tq)

 2Z t t
q ds
n Z

2  (n)
o
(1) Ex exp − Rs(n) ds ;
∈ dy, Rt ∈ dz
2 0 0 Rsn 2 ( )

qz  +1
 2
 y (x2 + z 2 )q ch(tq)  
xzq

=  exp − − iy/2 dydz
2x sh(tq) 2 2 sh(tq) sh(tq)

 2Z t t
p ds
n  Z
2 o
(2) Ex exp − (n)
2
; Rs(n) ds ∈ dy, Rt(n) ∈ dz
0 Rs 2 0

z  +1 p
=  isy ( p2 + ν 2 , t, 0, (x2 + z 2 )/2, xz/2)dydz
x

t t
ds
Z
2
Z 
1.21.8 Px Rs(n)
ds ∈ dg, (n)
2
∈ dy, Rt(n) ∈ dz
0 0 Rs
0<x
z  +1 −ν 2 y/2
= e eig (y/2, t, (x2 + z 2 )/2, xz)dgdydz
2x

 2Z τ τ o (2)1=4−=2 sh +3=2 (yp=2)

p ds ds
n  Z
1.22.1 Ex exp − (n)
2
; ∈ dy =
2 Rs Rs(n) 2x +1=2 ch +3=2 (y =2)
p
(1) 0 0
394 4. BESSEL PROCESS OF ORDER 

4 Rs(n) ν ≥ 0, n = 2ν + 2 ≥ 2, τ ∼ Exp(λ), independent of R(n)

√ √ √
(2 +  +  2 + p2 ) x 2 ch(y 2) √ 2 2 2x √2 dy
p    
0<x × exp − √ M
(1 + 2  2 + p2 ) sh(y 2) 3
−ν− 2 , ν +p sh(y 2)
p

τ τ  (2)1=4−=2 sh +3=2 (yp=2)

ds ds
Z Z
1.22.2 Px (n)
2
∈ dg, ∈ dy =
Rs Rs(n) 4x +1=2 ch +3=2 (y =2)
p
0<x 0 0
 2 √ √ √
 g x 2 ch(y 2) 3 x 2 
 
× exp − − √ mg/2 ν + , √ dgdy
2 sh(y 2) 2 sh(y 2)

p2
τ
q

n  Z  o
(n)
1.22.5 Ex exp − 2 + n ds ; R ∈ dz
2 Rs(n) Rs
( ) τ


0

√ √
 (  2 + p2 + 1=2 + q= 2)z  +1=2
p
0<x = √
2 (2  2 + p2 + 1)x +1=2
p

√ √
 M−q/√2λ,√ν 2 +p2 (2x 2λ)W−q/√2λ,√ν 2 +p2 (2z 2λ)dz, x ≤ z


× √ √
 W √ √ 2 2 (2x 2λ)M √ √ 2 2 (2z 2λ)dz, z ≤ x
−q/ 2λ, ν +p −q/ 2λ, ν +p

 2Z τ τ
p ds ds
n  Z o
(1) Ex exp − (n)
2
; ∈ dy, Rτ(n) ∈ dz
2 0 Rs 0 Rs(n)

√  (x + z )√2 ch(yp=2)   √
 2z  +1 √ 2 2xz 
=  exp − I2 ν 2 +p2 dydz
x sh(y =2) sh(y =2) sh(y =2)
p p p

τ τ
ds ds
Z Z 
1.22.6 Px (n)
2
∈ dg, ∈ dy, Rτ(n) ∈ dz
0 Rs 0 Rs(n)
0<x
√ √  √
 2z  +1
 2 (x + z ) 2 ch(y =2) 
 g 2 2xz 
p
=  exp − − ig/8 dgdydz
8x sh(y =2) 2 sh(y =2) sh(y =2)
p p p

 2Z t
p ds ds
n  o
(n)
1.22.7 Ex exp −
; ∈ dy, Rt ∈ dz
2 0 Rsn 2 Rs(n)
( )
(1)
z  +1 p √
0<x =  ist ( 4p2 + 4ν 2 , y/2, 0, x + z, xz)dydz
x

t t
ds ds
Z Z 
1.22.8 Px (n)
2
∈ dg, ∈ dy, Rt(n) ∈ dz
0 Rs 0 Rs(n)
0<x
z  +1 −ν 2 g/2 √
= e eit (g/8, y/2, x + z, 2 xz)dgdydz
8x
 1. EXPONENTIAL STOPPING 395

4 Rs(n) ν ≥ 0, n = 2ν + 2 ≥ 2, τ ∼ Exp(λ), independent of R(n)

n  Z τ

 o
p1I[0,r) Rs(n) + q1I[r,∞) Rs(n) ds ; sup Rs(n) < b


1.23.1 Ex exp −
0 0≤s≤τ
r<b
2(2(p − q)C (rq ; bq ) − (rq ) −2 −1
p2 )r x− I (xp )
2
x≤r = 2+ 2
p p q (p S (bq ; rq )I +1 (rp ) + q C (rq ; bq )I (rp ))

2  S (xq ; rq )
 4p−1 q−2 (q − p)I +1 (rp )S (bq ; xq )
r≤x = 2 1 − +
q S (bq ; rq ) p S (bq ; rq )I +1 (rp ) + q C (rq ; bq )I (rp )
x≤b
2r−2 −1 q−2 −2 I (rp )S (bq ; xq )
−
(p S (bq ; rq )I +1 (rp ) + q C (rq ; bq )I (rp ))S (bq ; rq )

n  Z τ


p1I[0,r) Rs(n) + q1I[r,∞) Rs(n) ds ;


1.23.5 Ex exp −
r<b 0
o
z<b sup Rs(n) < b, Rτ(n) ∈ dz
0≤s≤τ

2p2 z 2 +1 S (rp ; (z + x + |z − x|)p =2)I ((z + x − |z − x|)p =2)

z≤r = dz
((z + x − |z − x|)=2) r− I (rp )
x≤r
2z  +1 r−1 x− S (bq ; rq )I (zp )I (xp )
+ dz
(p S (bq ; rq )I +1 (rp ) + q C (rq ; bq )I (rp ))I (rp )

2z  +1 r− −1 I (zp )S (bq ; xq )

z≤r = dz
p S (bq ; rq )I +1 (rp ) + q C (rq ; bq )I (rp )
r≤x
2z 2 +1 r− −1 x− S (bq ; zq )I (xp )
r≤z = dz
p S (bq ; rq )I +1 (rp ) + q C (rq ; bq )I (rp )
x≤r
2q2 z 2 +1 S (bq ; (z + x + |z − x|)q =2)S ((z + x − |z − x|)q =2; rq )
r≤z = dz
S (bq ; rq )
r≤x
2z 2 +1 r−2 −1 I (rp )S (bq ; zq )S (bq ; xq )
+ dz
(p S (bq ; rq )I +1 (rp ) + q C (rq ; bq )I (rp ))S (bq ; rq )

n  Z τ

 o
p1I[0,r) Rs(n) + q1I[r,∞) Rs(n) ds ; a < inf Rs(n)


1.24.1 Ex exp −
0 0≤s≤τ
a<r
2  S (rp ; xp )
 4p−2 q−1 (p − q)K +1 (rq )S (xp ; ap )
a≤x = 1− +
p2 S (rp ; ap ) p K (rq )C (rp ; ap ) + q K +1 (rq )S (rp ; ap )
x≤r

√
Υs := 2λ + 2s
396 4. BESSEL PROCESS OF ORDER 

4 Rs(n) ν ≥ 0, n = 2ν + 2 ≥ 2, τ ∼ Exp(λ), independent of R(n)

2r−2 −1 p−2 −2 K (rq )S (xp ; ap )

−
(p K (rq )C (rp ; ap ) + q K +1 (rq )S (rp ; ap ))S (rp ; ap )

2 2(2(q − p)C (rp ; ap ) − (rp )−2 −1 q2 )r x− K (xq )
r≤x = 2 +
q p q2 (p K (rq )C (rp ; ap ) + q K +1 (rq )S (rp ; ap ))

n  Z ∞
 o
1.24.3 Ex exp −p 1I[0,r) Rs(n) ds ; a < inf Rs(n)
0 0≤s<∞
(1)
a<r
√ √ √
2a 2pS (x 2p; a 2p)

 C (a√2p; r√2p) ,
 a≤x≤r
−1
= −2
√ √ √
 1− x 2a 2px−2 S (r 2p; a 2p)
+ √ √ , r≤x

r−2 r−2 C−1 (a 2p; r 2p)

n  Z τ


p1I[0,r) Rs(n) + q1I[r,∞) Rs(n) ds ;


1.24.5 Ex exp −
a<r 0
o
a<z a < inf Rs(n) , Rτ(n) ∈ dz
0≤s≤τ

2p2 z 2 +1 S (rp ; (z + x + |z − x|)p =2)S ((z + x − |z − x|)p =2; ap )

z≤r = dz
S (rp ; ap )
x≤r

2z 2 +1 r−2 −1 K (rq )S (zp ; ap )S (xp ; ap )

+ dz
(p K (rq )C (rp ; ap ) + q K +1 (rq )S (rp ; ap ))S (rp ; ap )

2z 2 +1 r− −1 x− S (zp ; ap )K (xq )

z≤r = dz
p K (rq )C (rp ; ap ) + q K +1 (rq )S (rp ; ap )
r≤x

2z  +1 r− −1 K (zq )S (xp ; ap )

r≤z = dz
p K (rq )C (rp ; ap ) + q K +1 (rq )S (rp ; ap )
x≤r

2q2 z 2 +1 K ((z + x + |z − x|)q =2)S ((z + x − |z − x|)q =2; rq )

r≤z = dz
((z + x + |z − x|)=2) r− K (rq )
r≤x
2z  +1 r−1 x− S (rp ; ap )K (zq )K (xq )
+ dz
(p K (rq )C (rp ; ap ) + q K +1 (rq )S (rp ; ap ))K (rq )

√
Υs := 2λ + 2s
 1. EXPONENTIAL STOPPING 397

4 Rs(n) ν ≥ 0, n = 2ν + 2 ≥ 2, τ ∼ Exp(λ), independent of R(n)

Ex e−γ`(τ,r) ; a < inf Rs(n) , sup Rs(n) < b


1.25.1 a<r<b
0≤s≤τ 0≤s≤τ

√ √ √ √
S (r 2; x 2) + S (x 2; a 2)
a≤x =1− √ √
S (r 2; a 2)
x≤r
√ √ √ √ √ √ √ √
(S (b 2; a 2) − S (b 2; r 2) − S (r 2; a 2))S (x 2; a 2)
+ √ √ √ √ √ √ √ √
(2(2) r2 +1 S (b 2; r 2)S (r 2; a 2) + S (b 2; a 2))S (r 2; a 2)

√ √ √ √
S (b 2; x 2) + S (x 2; r 2)
r≤x =1− √ √
S (b 2; r 2)
x≤b
√ √ √ √ √ √ √ √
(S (b 2; a 2) − S (b 2; r 2) − S (r 2; a 2))S (b 2; x 2)
+ √ √ √ √ √ √ √ √
(2(2) r2 +1 S (b 2; r 2)S (r 2; a 2) + S (b 2; a 2))S (b 2; r 2)

Px `(τ, r) ∈ dy, a < inf Rs(n) , sup Rs(n) < b

1.25.2
0≤s≤τ 0≤s≤τ

√ √
y(2)− r−2 −1 S (b 2; a 2)
 
= exp − √ √ √ √
2S (b 2; r 2)S (r 2; a 2)

√ √ √ √ √ √ √ √
(S (b 2; a 2) − S (b 2; r 2) − S (r 2; a 2))S (x 2; a 2)

√ √ √ √ dy,
2(2) r2 +1 S (b 2; r 2)S2 (r 2; a 2)






 a≤x≤r
× √ √ √ √ √ √ √ √
(S (b 2; a 2) − S  (b 2; r 2) − S ( r 2; a 2))S  (b 2; x 2)
√ √ √ √ dy,

2(2) r2 +1 S2 (b 2; r 2)S (r 2; a 2)





r≤x≤b


Px `(τ, r) = 0, a < inf Rs(n) , sup Rs(n) < b

(1)
0≤s≤τ 0≤s≤τ

√ √ √ √
 S (r 2; x 2) + S (x 2; a 2)
1 − √ √ , a≤x≤r
S (r 2; a 2)



= √ √ √ √
 1 − S (b 2; x 2√) + S√(x 2; r 2) ,

r≤x≤b
S (b 2; r 2)

Ex e−γ`(τ,r) ; a < inf Rs(n) , sup Rs(n) < b, Rτ(n) ∈ dz


1.25.5 a<r<b
0≤s≤τ 0≤s≤τ
398 4. BESSEL PROCESS OF ORDER 

4 Rs(n) ν ≥ 0, n = 2ν + 2 ≥ 2, τ ∼ Exp(λ), independent of R(n)

√ √
(2) +1 z 2 +1 S (r 2; (z + x + |z − x|) =2)S ((z + x − |z − x|) =2; a 2)
p p
z≤r = √ √ dz
S (r 2; a 2)
x≤r

√ √ √ √ √ √
(2) +1 z 2 +1 S (b 2; r 2)S (z 2; a 2)S (x 2; a 2)dz
+ √ √ √ √ √ √ √ √
(2 (2) r2 +1 S (b 2; r 2)S (r 2; a 2) + S (b 2; a 2))S (r 2; a 2)


√ √ √ √
(2) +1 z 2 +1 S (b 2; x 2)S (z 2; a 2)
z≤r = √ √ √ √ √ √ dz
2 (2) r2 +1 S (b 2; r 2)S (r 2; a 2) + S (b 2; a 2)
r≤x

√ √ √ √
(2) +1 z 2 +1 S (b 2; z 2)S (x 2; a 2)
r≤z = √ √ √ √ √ √ dz
2 (2) r  +1 S (b 2; r 2)S (r 2; a 2) + S (b 2; a 2)
 2
x≤r

√ √
(2) +1 z 2 +1 S (b 2; (z + x + |z − x|) =2)S ((z + x − |z − x|) =2; r 2)
p p
r≤z = √ √ dz
S (b 2; r 2)
r≤x

√ √ √ √ √ √
(2) +1 z 2 +1 S (b 2; z 2)S (r 2; a 2)S (b 2; x 2)dz
+ √ √ √ √ √ √ √ √
(2 (2) r2 +1 S (b 2; r 2)S (r 2; a 2) + S (b 2; a 2))S (b 2; r 2)


Px `(τ, r) ∈ dy, a < inf Rs(n) , sup Rs(n) < b, Rτ(n) ∈ dz

1.25.6
0≤s≤τ 0≤s≤τ
a<r
r<b
√ √
y(2)− r−2 −1 S (b 2; a 2)
 
z≤r = exp − √ √ √ √
2S (b 2; r 2)S (r 2; a 2)

 z 2 +1 S (z √2; a√2)S (x√2; a√2)

 
√ √ dydz, 0 ≤ x ≤ r
r2 +1 S2 (r 2; a 2)


× √ √ √ √
2 +1 S (b 2; x 2)S (z 2; a 2)
 z

√ √ √ √ dydz, r ≤ x
r  +1 S (b 2; r 2)S (r 2; a 2)
2

√ √
y(2)− r−2 −1 S (b 2; a 2)
 
r≤z = exp − √ √ √ √
2S (b 2; r 2)S (r 2; a 2)

 z 2 +1 S (b√2; z √2)S (x√2; a√2)

 
 r2 +1 S (b√2; r√2)S (r√2; a√2) dydz, 0 ≤ x ≤ r

 
× 2 +1 S (b√2; z √2)S (b√2; x√2)
 z

√ √ dydz, r ≤ x
r2 +1 S2 (b 2; r 2)
 1. EXPONENTIAL STOPPING 399

4 Rs(n) ν ≥ 0, n = 2ν + 2 ≥ 2, τ ∼ Exp(λ), independent of R(n)

Px `(τ, r) = 0, a < inf Rs(n) , sup Rs(n) < b, Rτ(n) ∈ dz

(1)
0≤s≤τ 0≤s≤τ

√ √
(2) +1 z 2 +1 S (r 2; (z + x + |z − x|) =2)S ((z + x − |z − x|) =2; a 2)
p p
z≤r = √ √ dz
S (r 2; a 2)
x≤r

√ √
(2) +1 z 2 +1 S (b 2; (z + x + |z − x|) =2)S ((z + x − |z − x|) =2; r 2)
p p
r≤z = √ √ dz
S (b 2; r 2)
r≤x

n  Z τ


p1I[0,r) Rs(n) + q1I[r,∞) Rs(n) ds ;


1.26.1 Ex exp −
a<r 0
o
r<b a < inf Rs(n) , sup Rs(n) < b
0≤s≤τ 0≤s≤τ

 2r−2 −1  −2 −2 S (b ; r )S (x ; a )S −1 (r ; a )

2  S (rp ; xp ) p  q q  p p  p p
a≤x = 1 − −
p2 S (rp ; ap ) p S (bq ; rq )C (rp ; ap )+ q C (rq ; bq )S (rp ; ap )
x≤r
2(2(p − q)C (rq ; bq ) − (rq )−2 −1 p2 )S (xp ; ap )
+ 2
p q (p S (bq ; rq )C (rp ; ap ) + q C (rq ; bq )S (rp ; ap ))

 2r−2 −1  −2 −2 S (r ; a )S (b ; x )S −1 (b ; r )

2  S (xq ; rq ) q  p p  q q  q q
r≤x = 1 − −
q2 S (bq ; rq ) p S (bq ; rq )C (rp ; ap )+ q C (rq ; bq )S (rp ; ap )
x≤b
2(2(q − p)C (rp ; ap ) − (rp )−2 −1 q2 )S (bq ; xq )
+
p q2 (p S (bq ; rq )C (rp ; ap ) + q C (rq ; bq )S (rp ; ap ))

n  Z τ


p1I[0,r) Rs(n) + q1I[r,∞) Rs(n) ds ;


1.26.5 Ex exp −
r<b 0
o
a<r a < inf Rs(n) , sup Rs(n) < b, Rτ(n) ∈ dz
0≤s≤τ 0≤s≤τ

2p2 z 2 +1 S (rp ; (z + x + |z − x|)p =2)S ((z + x − |z − x|)p =2; ap )

z≤r = dz
S (rp ; ap )
x≤r
2z 2 +1 r−2 −1 S (bq ; rq )S (zp ; ap )S (xp ; ap )
+ dz
(p S (bq ; rq )C (rp ; ap ) + q C (rq ; bq )S (rp ; ap ))S (rp ; ap )

√
Υs := 2λ + 2s
400 4. BESSEL PROCESS OF ORDER 

4 Rs(n) ν ≥ 0, n = 2ν + 2 ≥ 2, τ ∼ Exp(λ), independent of R(n)

2z 2 +1 r−2 −1 S (zp ; ap )S (bq ; xq )

z≤r = dz
p S (bq ; rq )C (rp ; ap ) + q C (rq ; bq )S (rp ; ap )
r≤x
2z 2 +1 r−2 −1 S (bq ; zq )S (xp ; ap )
r≤z = dz
p S (bq ; rq )C (rp ; ap ) + q C (rq ; bq )S (rp ; ap )
x≤r

2q2 z 2 +1 S (bq ; (z + x + |z − x|)q =2)S ((z + x − |z − x|)q =2; rq )

r≤z = dz
S (bq ; rq )
r≤x
2z 2 +1 r−2 −1 S (rp ; ap )S (bq ; zq )S (bq ; xq )
+ dz
(p S (bq ; rq )C (rp ; ap ) + q C (rq ; bq )S (rp ; ap ))S (bq ; rq )

n  Z τ

 o
p1I[0,r) Rs(n) + q1I[r,∞) Rs(n) ds ; `(τ, r) ∈ dy


1.27.1 Ex exp −
0

yp I +1 (rp ) yq K +1 (rq )

 
= exp − −
2I (rp ) 2K (rq )

 
r I (xp ) I +1 (rp ) K +1 (rq )
 
 x I (r )  I (r ) +  K (r ) dy, 0 ≤ x ≤ r

 p p  p q  q
× 
 r K (xq ) I +1 (rp ) + K +1 (rq ) dy, r ≤ x
 

x K (rq ) p I (rp ) q K (rq )

n  Z τ

 o
p1I[0,r) Rs(n) + q1I[r,∞) Rs(n) ds ; `(τ, r) = 0


(1) Ex exp −
0
 x− I (xp )
  
 +p 1−

r− I (rp )
, 0≤x≤r
=
  1− x− K (xq )
 
, r≤x

+q r− K (rq )

n  Z ∞
 o
1.27.3 Ex exp −p 1I[0,r) Rs(n) ds ; `(∞, r) ∈ dy
(1) 0

  −1 √
r I (x 2p)
√ √
y pI +1 (r 2p) √ dy, 0 ≤ x ≤ r
y x I (r 2p)
  

= exp − − √ √
−2
r 2I (r 2p)   x dy, r≤x
r1−2

√
Υs := 2λ + 2s
 1. EXPONENTIAL STOPPING 401

4 Rs(n) ν ≥ 0, n = 2ν + 2 ≥ 2, τ ∼ Exp(λ), independent of R(n)

n  Z τ

 o
Ex exp − p1I[0,r) Rs(n) + q1I[r,∞) Rs(n) ds ; `(τ, r) ∈ dy, Rτ(n) ∈ dz


1.27.5
0
z  +1 I (xp )I (zp ) yp I +1 (rp ) yq K +1 (rq )
 
z≤r = exp − − dydz
x rI2 (rp ) 2I (rp ) 2K (rq )
x≤r
z  +1 I (zp )K (xq ) yp I +1 (rp ) yq K +1 (rq )
 
z≤r =  exp − − dydz
x rI (rp )K (rq ) 2I (rp ) 2K (rq )
r≤x
z  +1 I (xp )K (zq ) yp I +1 (rp ) yq K +1 (rq )
 
r≤z = exp − − dydz
x rI (rp )K (rq ) 2I (rp ) 2K (rq )
x≤r
z  +1 K (xq )K (zq ) yp I +1 (rp ) yq K +1 (rq )
 
r≤z = exp − − dydz
x rK2 (rq ) 2I (rp ) 2K (rq )
r≤x
n  Z τ

 o
p1I[0,r) Rs(n) + q1I[r,∞) Rs(n) ds ; `(τ, r) = 0, Rτ(n) ∈ dz


(1) Ex exp −
0
2z  +1 n I (xp )I (zp )K (rp )
o
z≤r = Fν (xΥp , zΥp ) − dz
x I (rp )
x≤r
2z  +1 n K (xq )K (zq )I (rq )
o
r≤z = Fν (xΥq , zΥq ) − dz
x K (rq )
r≤x

Ex e−γ Ĥ(τ )−ηȞ(τ ) ; inf Rs(n) ∈ da, sup Rs(n) ∈ db dadb

 
1.28.1
0≤s≤τ 0≤s≤τ
a<x
x<b
√ √ √ √ √
(2 +2)− S (b 2 +2 +2; x 2 +2 +2) 2b2+1 C (b 2; a 2) − (2)−


= √ √ √ √ √ √
(ba)2+1 S (b 2 +2 +2;a 2 +2 +2)S (b 2 +2;a 2 +2)S (b 2;a 2)

√ √ √ √ √
(2 +2 )− S (x 2 +2 +2; a 2 +2 +2) 2a2+1 C (a 2; b 2) − (2)−


+ √ √ √ √ √ √
(ba)2+1 S (b 2 +2 +2;a 2 +2 +2)S (b 2 +2 ;a 2 +2 )S (b 2;a 2)

Ex e−γ Ĥ(τ )−ηȞ(τ ) ; inf Rs(n) ∈ da, sup Rs(n) ∈ db, Rτ(n) ∈ dz dadbdz
 
1.28.5
0≤s≤τ 0≤s≤τ
a<x
x<b
√ √ √ √
2(2 + 2)− (ba)−2 −1 z 2 +1 S (b 2 + 2 + 2;x 2 + 2 + 2)S (z 2;a 2)
= √ √ √ √ √ √
S (b 2 + 2 + 2; a 2 + 2 + 2)S (b 2 + 2; a 2 + 2)S (b 2; a 2)

√ √ √ √
2(2 + 2 )− (ba)−2 −1 z 2 +1 S (x 2 + 2 + 2; a 2 + 2 + 2)S (b 2; z 2)
+ √ √ √ √ √ √
S (b 2 + 2 + 2; a 2 + 2 + 2)S (b 2 + 2 ; a 2 + 2 )S (z 2; a 2)

√
Υs := 2λ + 2s
402 4. BESSEL PROCESS OF ORDER 

4 Rs(n) ν ≥ 0, n = 2ν + 2 ≥ 2, τ ∼ Exp(λ), independent of R(n)

n  Z τ


p1I[0,r) Rs(n) + q1I[r,∞) Rs(n) ds ; a < inf Rs(n) ,


1.29.1 Ex exp −
0 0≤s≤τ
o
0≤a sup Rs(n) < b, `(τ, r) ∈ dy
0≤s≤τ

C (rq ; bq ) − (rq )−2 −1 C (rp ; ap ) − (rp )−2 −1

 
=λ +
q S (bq ; rq ) p S (rp ; ap )
S (xp ; ap )

 S (r ; a ) dy, a≤x≤r
yq C (rq ; bq ) yp C (rp ; ap )

 p p

× exp − −
2S (bq ; rq ) 2S (rp ; ap )    q q ) dy,
S (b ; x
r≤x≤b
S (bq ; rq )

n  Z τ


p1I[0,r) Rs(n) + q1I[r,∞) Rs(n) ds ; a < inf Rs(n) ,


(1) Ex exp −
0 0≤s≤τ
o
sup Rs(n) < b, `(τ, r) = 0
0≤s≤τ

 2  S (rp ; xp ) + S (xp ; ap )


 2 1 −

S (rp ; ap )
, a≤x≤r
p
=
 2 1 − S (bq ; xq ) + S (xq ; rq ) ,
 
r≤x≤b

q2 S (bq ; rq )

n  Z ∞
 o
(2) Ex exp −p 1I[0,r) Rs(n) ds ; a < inf Rs(n) , `(∞, r) ∈ dy
0 0≤s<∞

√ √
S (x 2p; a 2p)
  √

√ √ √
y pC (r 2p; a 2p)
√ dy, a ≤ x ≤ r
y rS (r 2p; a 2p)
  
= exp − − √ √ √
2
r 2S (r 2p; a 2p)   x
−
dy, r≤x
r1−2

n  Z ∞
 o
(3) Ex exp −p 1I[0,r) Rs(n) ds ; a < inf Rs(n) , `(∞, r) = 0
0 0≤s<∞

a≤x≤r
(
0,
= x−2
1 − −2 , r≤x
r

√
Υs := 2λ + 2s
 1. EXPONENTIAL STOPPING 403

4 Rs(n) ν ≥ 0, n = 2ν + 2 ≥ 2, τ ∼ Exp(λ), independent of R(n)

n  Z τ


p1I[0,r) Rs(n) + q1I[r,∞) Rs(n) ds ; a < inf Rs(n) ,


1.29.5 Ex exp −
0 0≤s≤τ
o
0≤a sup Rs(n) < b, `(τ, r) ∈ dy, Wτ ∈ dz
0≤s≤τ

z 2 +1 yq C (rq ; bq ) yp C (rp ; ap )

 
= exp − −
r2 +1 2S (bq ; rq ) 2S (rp ; ap )

 S (xp ; ap )S (zp ; ap )

dydz, a ≤ x ≤ r, a ≤ z ≤ r
S2 (rp ; ap )



S (bq ; xq )S (zp ; ap )



dydz, a≤z≤r≤x≤b

S (bq ; rq )S (rp ; ap )


×
S (bq ; zq )S (xp ; ap )
dydz, a≤x≤r≤z≤b

S (bq ; rq )S (rp ; ap )




S (bq ; xq )S (bq ; zq )



dydz, r ≤ x ≤ b, r ≤ z ≤ b
S2 (bq ; rq )


n  Z τ


p1I[0,r) Rs(n) + q1I[r,∞) Rs(n) ds ; a < inf Rs(n) ,


(1) Ex exp −
0 0≤s≤τ

o
sup Rs(n) < b, `(τ, r) = 0, Wτ ∈ dz
0≤s≤τ

2z 2 +1 p2 S (rp ; (z + x + |z − x|)p =2)S ((z + x − |z − x|)p =2; ap )

x≤r = dz
S (rp ; ap )
z≤r
2z 2 +1 q2 S (bq ; z + x + |z − x|)q =2)S ((z + x − |z − x|)q =2; rq )
r≤x = dz
S (bq ; rq )
r≤z

√
Υs := 2λ + 2s
404 4. BESSEL PROCESS OF ORDER 

4 Rs(n) ν > 0, n = 2ν + 2 > 2 Hz = min{s : Rs(n) = z}

2. Stopping at first hitting time

 x− I (x√2 )

 z − I (z √2 ) , x ≤ z


2.0.1 Ex e−αHz = −
√
 x K (x√ 2 ) , z ≤ x

z − K (z 2 )

∞

X j;k x− J (j;k x=z ) −jν,k
2
t/2z 2
2.0.2 Px Hz ∈ dt = e dt
z 2− J +1 (j;k )
x≤z k=1

x≤z
(
0,
−2


(1) Px H z = ∞ = x
1− , z≤x
z −2

1, x≤z
(
sup Rs(n) x−2 − y−2


2.1.2 Px <y =
, z≤x≤y
z≤y 0≤s≤Hz
z −2 − y−2

 x− I (x√2 )

 z − I (z √2 ) ,
 x≤z

Ex e−αHz ; sup Rs(n)

2.1.4 <y = √ √
z≤y 0≤s≤Hz  S (y √2 ; x√2 ) , z ≤ x ≤ y

S (y 2 ; z 2 )

 −2
 y − x−2

 y−2 − z −2 , y ≤ x ≤ z
2.2.2 Px inf Rs(n) > y, Hz < ∞ = −2
y≤z
0≤s≤Hz  x

, z≤x
z −2
x−2


 1 − y−2 , z ≤ y ≤ x

(1) Px inf Rs(n) > y, Hz = ∞ = −2
z≤y
0≤s≤Hz  1− x

, y≤z≤x
z −2

 S (x √ 2 ; y √ 2 )

 S (z √2 ; y√2 ) ,
 y≤x≤z
 −αHz (n) 
2.2.4 Ex e ; inf Rs >y = √
− K (x 2 )
0≤s≤Hz x 
y≤z , z≤x

√
z − K (z 2 )


Ex e−γ`(Hz ,r) ; Hz < ∞


2.3.1
 2. STOPPING AT FIRST HITTING TIME 405

4 Rs(n) ν > 0, n = 2ν + 2 > 2 Hz = min{s : Rs(n) = z}

 x−2
−2
, z≤x
 z 2 +1



+r (r−2 − x−2 )

r≤z = , r≤x≤z
  + r2 +1 (r−2 − z −2 )



, 0≤x≤r

 + r2 +1 (r−2 − z −2 )


 x−2
, r≤x
 z −2 + r(z −2 − r−2 )



z≤r = x−2 + r(x−2 − r−2 )
−2 +
, z≤x≤r
 z


 r(z −2 − r−2 )
1, 0≤x≤z

Ex e−γ`(Hz ,r) ; Hz = ∞

(1)
z≤r
 x−2 x−2 (z −2 − r−2 )
1 − 2 + , r≤x
r  r (z −2 + r(z −2 − r−2 ))
2

 − −

2 2

=  (z − −x − )
−2 + r (z −2 − r −2 )
, z≤x≤r



 z
0, 0≤x≤z

2.3.2 Px `(Hz , r) ∈ dy, Hz < ∞

0, 0≤x≤z

 r−2 −1 (z −2 − x−2 ) z −2 y

  

exp − −2 dy, z ≤ x ≤ r
z≤r = (z −2 − r−2 )2 r(z − r−2 )
x−2 z −2 y

  
exp − −2 r≤x

dy,
r(z −2 − r−2 ) r (z − r−2 )


z≤x

0,
  (x−2 − z −2 )

y

  
exp − 2 +1 −2 dy, r ≤ x ≤ z
r≤z = r2 +1 (r−2 − z −2 )2 r (r − z −2 )
 y

  
exp − 2 +1 −2 dy, 0 ≤ x ≤ r

r2 +1 (r−2 − z −2 ) r (r − z −2 )


(1) Px `(Hz , r) ∈ dy, Hz = ∞
0, 0≤x≤z

  (z −2 − x−2 ) z −2 y

  
exp − −2 dy, z ≤ x ≤ r

z≤r = r(z −2 − r−2 ) r(z − r−2 )
 x−2 exp − z −2 y

  
r≤x

dy,
r1−2 r(z −2 − r−2 )
406 4. BESSEL PROCESS OF ORDER 

4 Rs(n) ν > 0, n = 2ν + 2 > 2 Hz = min{s : Rs(n) = z}

 −2
x
, r≤z≤x
z −2




1, 0 ≤ x ≤ z ≤ r,





(2) Px ` Hz , r = 0, Hz < ∞ = | x −2 − r −2 |
, z∧r ≤x≤r∨z
|z −2 − r−2 |





z ≤ r ≤ x, 0 ≤ x ≤ r ≤ z

0,

x−2
(3) Px ` Hz , r = 0, Hz = ∞ = 1 − −2 , r≤x
r
z≤r

2.3.3 Ex e−αHz −γ` Hz ,r

 x− K (x√2 )

√
− K (z 2 )
, z≤x
z




√ √ √ √

r +1 (2 ) S (x 2 ; r 2 )I (r 2
 −
x I (x 2 ) + 2 )

r≤z = √ √ √ √ , r≤x≤z
 z − I (z 2 ) + 2 r +1 (2 ) S (z 2 ; r 2 )I (r 2 )
√
x− I (x 2 )




√ √ √ √ , 0≤x≤r
r +1 (2 ) S (z 2 ; r 2 )I (r 2

z − I (z 2 ) + 2  )

 x− I (x√2 )

√
− I (z 2 )
, x≤z
z




√ √ √ √

r +1 (2 ) S (r 2 ; x 2 )K (r 2
 −
x K (x 2 ) + 2 )

z≤r = √ √ √ √ , z≤x≤r
 z − K (z 2 ) + 2 r +1 (2 ) S (r 2 ; z 2 )K (r 2 )
√
x− K (x 2 )




√ √ √ √ , r≤x
r +1 (2 ) S (r 2 ; z 2 )K (r 2

z − K (z 2 ) + 2 )

Ex e−αHz ; ` Hz , r ∈ dy


2.3.4

 (2 x)− r− −1 I (x√2 )

 2S (z √2 ; r√2 )I (r√2 ) dy, 0 ≤ x ≤ r

  1 √   

(2 z ) r
− − − I (z 2 )y
 √ √
r≤z = exp − √ √ √ (2 )− S (z 2 ; x 2 )
2S (z 2 ; r 2 )I (r 2 )  √ √ dy, r≤x≤z
 2r2 +1 S2 (z 2 ; r 2 )


0, z≤x
 (2 x)− r− −1 K (x√2 )
 2S (r√2 ; z √2 )K (r√2 ) dy, r ≤ x

  1 √   

(2 z ) r
− − − K (z 2 )y

 √ √
z≤r = exp − √ √ √ (2 )− S ( x 2 ; z 2 )
2S (r 2 ; z 2 )K (r 2 )  2 +1 2 √ √ dy, z ≤ x ≤ r
 2r

 S (r 2 ; z 2 )
0, 0≤x≤z
 2. STOPPING AT FIRST HITTING TIME 407

4 Rs(n) ν > 0, n = 2ν + 2 > 2 Hz = min{s : Rs(n) = z}

0≤x≤r≤z z≤r≤x

0,
√ √

|S (x 2 ; r 2 )|



√ √ , r∧z ≤x≤r∨z

|S (z 2 ; r 2 )|




√
z − K (x 2 )
 −αHz

(1) Ex e ; ` Hz , r = 0 =
√ , r≤z≤x
z − K (z 2 )



√

z − I (x 2 )



√ , 0≤x≤z≤r

z − I (z 2 )


 Z Hz

1I[r,∞) Rs(n) ds ; Hz < ∞

2.4.1 Ex exp −γ
0

1, 0≤x≤z

√ √
 x−2 K+1 (r 2 ) − r−2 K−1 (r 2 )



√ √ , z≤x≤r
z≤r = z −2 K+1 (r 2 ) − r−2 K−1 (r 2 )
√
2 (2 )−1=2 r−−1 x− K (x 2 )



√ √ , r≤x
z −2 K+1 (r 2 ) − r−2 K−1 (r 2

)

√
(r 2 )−2 −1

√ √ , 0<x≤r
C (r 2 ; z 2 )




√ √
C (r 2 ; x 2 )


r≤z = √ √ , r≤x≤z
 C (r 2 ; z 2 )
−
√
 x K (x√ 2 ) ,



z≤x

z − K (z 2 )

 Z Hz

1I[r,∞) Rs(n) ds ; Hz = ∞ = 0

(1) Ex exp −γ
0

1, 0≤x≤z

 −2
− r−2
Hz 
x
Z 
(n)


2.4.2 Px 1I[r,∞) Rs ds = 0 = −2 − r −2
, z≤x≤r
(1) 0  z

0, r≤x
z≤r

 Z Hz

1I[0,r] Rs(n) ds ; Hz < ∞

2.5.1 Ex exp −γ
0
√
2 (2 )−1=2 r−−1 x− I (x 2 )

√ √ , 0≤x≤r
r−2 I (r 2 ) − z −2 I+1 (r 2 )

 −2 −1 √



√
r I−1 (r 2 ) − x−2 I+1 (r 2

r≤z = )
√ √ , r≤x≤z
 r−2 I−1 (r 2 ) − z −2 I+1 (r 2 )
 −2 ,

x



z≤x
z −2
408 4. BESSEL PROCESS OF ORDER 

4 Rs(n) ν > 0, n = 2ν + 2 > 2 Hz = min{s : Rs(n) = z}

 x− I (x√2 )

√
− I (z 2 )
, 0≤x≤z
z




√ √

 −1
x C−1 (x 2 ; r 2 )

z≤r = √ √ , z≤x≤r
 z −1 C−1 (z 2 ; r 2 )
(2 )−+1=2 x−2



, r≤x

√ √
z −1 C−1 (z 2 ; r 2 )


 Z Hz

1I[0,r] Rs(n) ds ; Hz = ∞

(1) Ex exp −γ
z≤r 0

0, 0≤x≤z

√ √ √
 2 2 S (x 2 ; z 2 )



√ √ , z≤x≤r
= z −1 C−1 (z 2 ; r 2 )
√ √
 1 − 2 x−2 r2 +1 C√ (r 2 ; z 2 )


, r≤x

√
z −1 C−1 (z 2 ; r 2 )

0, x≤r

Hz  −2

2
r −x
Z  −
Rs(n) ds = 0 =


2.5.2 Px 1I[0,r] −2 − z −2
, r≤x≤z
(2) 0  r

1, z≤x
r≤z

n  Z Hz

 o
p1I[0,r) Rs(n) + q1I[r,∞) Rs(n) ds ; Hz < ∞


2.6.1 Ex exp −
0

√ √ −2 −1 √
 q(r 2q) (x=r)− I (x 2p)
√ √ √ √ √ √ √√ , 0≤x≤r
pS (z 2q; r 2q)I +1 (r 2p) + qC (r 2q; z 2q)I (r 2p)




√ √ √ √ √ √ √ √
pS (x 2q; r 2q)I +1 (r 2p) + qC (r 2q; x 2q)I (r 2p)


r≤z = √ √ √ √ √ √ √ √ , r≤x≤z
 pS (z 2q; r 2q)I +1 (r 2p) + qC (r 2q; z 2q)I (r 2p)
√
x− K (x 2q)



z≤x

√ ,
z − K (z 2q)


x− I (x√2p)

z − I (z √2p) ,

 0≤x≤z
√pK (r√2q)C (r√2p; x√2p) + √qK (r√2q)S (r√2p; x√2p)


  +1 
z≤r = √pK (r√2q)C (r√2p; z √2p) + √qK (r√2q)S (r√2p; z √2p) , z ≤ x ≤ r
   +1 
√ √ −2 −1 √
− K (x 2q )
p (r 2p ) (x=r )




, r≤x

√ √ √ √ √ √ √ √
pK (r 2q)C (r 2p; z 2p) + qK+1 (r 2q)S (r 2p; z 2p)

x I2=3 ( 23 x3=2 2 )
 − √
, 0≤x≤z
( 2 z 3=2 2 )
Hz  z − I

 √
2=3 3
n  Z  o
2.8.1 Ex exp −γ Rs(n) ds ; Hz < ∞ =
( 2 x3=2 2 )
 √
x K
−
0  − 2=3 23 3=2 √ ,

 z≤x
z K2=3 ( 3 z 2 )
 2. STOPPING AT FIRST HITTING TIME 409

4 Rs(n) ν > 0, n = 2ν + 2 > 2 Hz = min{s : Rs(n) = z}

x I=2 (x2 =2)

 −
 z − I (z 2 =2) , 0 ≤ x ≤ z

 2Z Hz 
=2
n  o
2.9.1 Ex exp − (Rs(n) )2 ds ; Hz < ∞ =
2 0  x− K=2 (x2 =2)
, z≤x
z K=2 (z 2 =2)
 −


 − −1 2
 x M− =2 ;=2 ( x )
 z − −1 M 2 , 0≤x≤z
2 Hz =2 ;=2 ( z )
 Z  
−
2.9.3 Ex exp −αHz − (Rs(n) )2 ds = − −1 W 2
2 0 x =2 ;=2 ( x )
 − −1 − z≤x

2 ,
=2 ;=2 ( z )

z W−

√ √
2y−1−2 S (x 2 ; z 2 )
Ex e−γ Ȟ(Hz ) ; sup Rs(n) ∈ dy =

2.13.1 √ √ dy
0≤s≤Hz (z 2 − y−2 )S (y 2 ; z 2 )
−
z<x
x<y

Ex e−γ Ȟ(Hz )−ηHz ; sup Rs(n) ∈ dy


2.13.4 z<x<y
0≤s≤Hz

√ √
(2)− y−1−2 S (x 2 + 2; z 2 + 2)
= √ √ √ √ dy
S (y 2; z 2)S (y 2 + 2; z 2 + 2)

√ √
2y−1−2 S (z 2 ; x 2 )
Ex e−γ Ĥ(Hz ) ; Rs(n) ∈ dy =

2.14.1 inf √ √ dy
0≤s≤Hz (y−2 − z−2 )S (z 2 ; y 2 )
x<z
y<x

Ex e−γ Ĥ(Hz )−ηHz ; Rs(n) ∈ dy


2.14.4 inf y<x<z
0≤s≤Hz

√ √
(2)− y−1−2 S (z 2 + 2; x 2 + 2)
= √ √ √ √ dy
S (z 2; y 2)S (z 2 + 2; y 2 + 2)



2.18.1 Ex exp −γ`(Hz , r) − η`(Hz , u) ; Hz < ∞
r<u

for z < r
 2 x−2 +  r(x−2 − r−2 ) + u(x−2 − u−2 ) + u(r2+1 x−2 − r)(r−2 − u−2 )
z≤x = 2 −2
 z +  r(z−2 − r−2 ) + u(z−2 − u−2 ) + u(r2+1 z−2 − r)(r−2 − u−2 )
x≤r
 2 x−2 + u(x−2 − u−2 )
r≤x = 2 −2
 z +  r(z−2 − r−2 ) + u(z−2 − u−2 ) + u(r2+1 z−2 − r)(r−2 − u−2 )
x≤u
410 4. BESSEL PROCESS OF ORDER 

4 Rs(n) ν > 0, n = 2ν + 2 > 2 Hz = min{s : Rs(n) = z}

 2 x−2
u≤x = 2 −2
 z +  r(z−2 − r−2 ) + u(z−2 − u−2 ) + u(r2+1 z−2 − r)(r−2 − u−2 )

for u < z

2
0≤x = 2 2 +1 2 2 +1
 +  (r − r  
z ) + (u − u
−  z−2 ) + (r − r2+1 u−2 )(u − u2+1 z−2 )
x≤r
 2 +  r2+1 (r−2 − x−2 )
r≤x = 2 2 +1
 +  (r − r  z ) + (u − u2+1 z−2 ) + (r − r2+1 u−2 )(u − u2+1 z−2 )
−2
x≤u
 2 +  (r − r2+1 x−2 ) + (u − u2+1 x−2 ) + (r − r2+1 u−2 )(u − u2+1 x−2 )
u≤x = 2
 +  (r − r2+1 z−2 ) + (u − u2+1 z−2 ) + (r − r2+1 u−2 )(u − u2+1 z−2 )
x≤z
x−2
z≤x =
z−2



(1) Ex exp −γ`(Hz , r) − η`(Hz , u) ; Hz = ∞
z<r
r<u
 2 (z−2 − x−2 )
z≤x = 2 −2
 z +  r(z−2 − r−2 ) + u(z−2 − u−2 ) + u(r2+1 z−2 − r)(r−2 − u−2 )
x≤r

 2 (z−2 − x−2 ) +  r2+1 (z−2 − r−2 )(r−2 − x−2 )

r≤x = 2 −2
 z +  r(z−2 − r−2 ) + u(z−2 − u−2 ) + u(r2+1 z−2 − r)(r−2 − u−2 )
x≤u

u2+1 (z−2 − u−2 )(u−2 − x−2 ) +  r2+1 (z−2 − r−2 )(r−2 − x−2 )
u≤x = 2 −2
 z +  r(z−2 − r−2 ) + u(z−2 − u−2 ) + u(r2+1 z−2 − r)(r−2 − u−2 )

 2 (z−2 − x−2 ) + (ru)2+1 (z−2 − r−2 )(r−2 − u−2 )(u−2 − x−2 )

+ 2 −2
 z +  r(z−2 − r−2 ) + u(z−2 − u−2 ) + u(r2+1 z−2 − r)(r−2 − u−2 )

Px r−2ν−1 `(Hz , r) ∈ dy, u−2ν−1 `(Hz , u) ∈ dg, Hz < ∞

2.18.2
r<u
 −2 −2
z −x
pr dydg, z≤x≤r
 z−2 − r−2



z<r = 2
x −u
−  −2 2
r −x
− −2
p dydg + −2 −2 pu dydg, r ≤ x ≤ u


 r−2 − u−2 r r −u
u≤x

pu dydg,
 2. STOPPING AT FIRST HITTING TIME 411

4 Rs(n) ν > 0, n = 2ν + 2 > 2 Hz = min{s : Rs(n) = z}


qr dydg, x≤r
 x−2 − u−2 r−2 − x−2



q dydg + −2 −2 qu dydg, r ≤ x ≤ u
u<z = r−2 − u−2 r r −u

 x− 2 − z−2
u≤x≤z

q dydg,
u−2 − z−2 u


where
 2 (uzr2 )2 y=g  (y + g(u=r)2 ) y
   √yg 
p 
pr := exp − − I
(u2 − r2 )(r2 − z 2 ) r−2 − u−2 z−2 − r−2 1 r−2 − u−2

 2 (zru2 )2
   √yg 
 (y + g) g

qr := 2 exp − − I
(z − u2 )(u2 − r2 ) r−2 − u−2 u−2 − z−2 0 r−2 − u−2

 2 (uzr2 )2  (y + g(u=r)2 )

   √yg 
y

pu := exp − − I
(u2 − r2 )(r2 − z 2 ) r−2 − u−2 z−2 − r−2 0 r−2 − u−2

 2 (zru2 )2 g=y  (y + g) g

   √yg 
p 
qu := 2 2 2 2 exp − −2 −2 − −2 −2 I1 −2 −2
(z − u )(u − r ) r −u u −z r −u

Px r−2ν−1 `(Hz , r) ∈ dy, `(Hz , u) = 0, Hz < ∞

(1)

 −2 −2
 (z − x ) y(z−2 − u−2 )
 
 (z−2 − r−2 )2 exp − (z−2 − r−2 )(r−2 − u−2 ) dy,
 z≤x≤r
=
 (x−2 − u−2 ) y(z−2 − u−2 )
 
exp − −2 −2 −2 −2 dy, r ≤ x ≤ u

2 2
(z − r )(r − u ) 2 2 (z − r )(r − u )

− − − −

Px `(Hz , r) = 0, u−2ν−1 `(Hz , u) ∈ dg, Hz < ∞

(2)

 (r−2 − x−2 ) g(r−2 − z−2 )

  
 (r−2 − u−2 )(u−2 − z−2 ) exp − (r−2 − u−2 )(u−2 − z−2 ) dg, r ≤ x ≤ u

= −2 −2
  (x − z ) exp − g(r−2 − z−2 )
 
dg, u≤x≤z

2
(u − z )
− −2 2 −2 − 2
(r − u )(u − z )
− 2 −2

 −2 −2
 x −r z≤x≤r

 z−2 − r−2 ,
(3) Px `(Hz , r) = 0, `(Hz , u) = 0, Hz < ∞ = −2 −2
 u −x ,

u≤x≤z
u − z 2
−2 −

Px r−2ν−1 `(Hz , r) ∈ dy, u−2ν−1 `(Hz , u) ∈ dg, Hz = ∞

(4)
412 4. BESSEL PROCESS OF ORDER 

4 Rs(n) ν > 0, n = 2ν + 2 > 2 Hz = min{s : Rs(n) = z}

 z−2 − x−2
−2 −2 r
p∞ dydg, z≤x≤r
 z−2 − r −2



2 2

x −u r −x
− −
z<r = −2 − u−2 r
p∞ dydg + −2 −2 p∞u dydg, r ≤ x ≤ u
 r r −u
2
 u ∞



p dydg, u≤x
x2 u
where

 2 u2  (y + g(u=r)2 )
   √yg 
y

p∞
r := exp − − I
r−2 − u−2 r−2 − u−2 z−2 − r−2 0 r−2 − u−2
 2 u2 g=y  (y + g(u=r)2 ) y
   √yg 
p 
p∞ := exp − − I
u r−2 − u−2 r−2 − u−2 z−2 − r−2 1 r−2 − u−2

(5) Px `(Hz , r) = 0, `(Hz , u) ∈ dg, Hz = ∞
z<r
 (r−2 − x−2 ) gr−2
  
 u(r−2 − u−2 ) exp − u(r−2 − u−2 ) dg, r ≤ x ≤ u

=
 x−2 gr−2
 
exp − dg, u≤x
u1−2 u(r−2 − u−2 )


u2
(6) Px `(Hz , r) = 0, `(Hz , u) = 0, Hz = ∞ = 1 − 2 , u≤x
x

 − √
x I2 (2 2 x)
Hz  z − I (2√2 z ) , 0 < x ≤ z
ds
Z o 
2
n  
2.19.1 Ex exp −γ ; Hz < ∞ =
Rs(n) −
√
0  x K2 (2√2 x) , z ≤ x

z − K2 (2 2 z )

 − −1=2 √
x M− =√2 ; (2x 2 )
√ , 0<x≤z

 z − −1=2 M √
; (2z 2 )

Hz
ds
 
− = 2
 Z
2.19.3 Ex exp −αHz − γ =
Rs(n)
√
0  x− −1=2 W− =√2 ; (2x 2 )
√ , z≤x

 − −1=2
z W− =√2 ; (2z 2 )


  √
x −ν+ ν 2 +γ 2
, 0<x≤z

2 Hz
ds z
n  Z  o 
2.20.1 Ex exp − (n)
2
; Hz < ∞ = √
2 Rs
 x
0   −ν− ν 2 +γ 2
, z≤x

z
 2. STOPPING AT FIRST HITTING TIME 413

4 Rs(n) ν > 0, n = 2ν + 2 > 2 Hz = min{s : Rs(n) = z}

 x− I√ √

 2+ 2 x 2

 z − I√ 2 2 z 2 , 0 < x ≤ z

 √

2 Hz
ds
   +
 Z
2.20.3 Ex exp −αHz − =
Rs(n) 2
√

2 0


 −
 x K√  2 + 2 x 2
 − √
 √
, z ≤ x
z K 2+ 2 z 2

 
z ln(z=x) 2 2
  √ 3=2 e−ν y/2 e− ln (z/x)/2y dy, 0 < x ≤ z
x 2y
Hz
ds
Z  
2.20.4 Px
∈ dy =
0 Rs(n) 2 
 z √ln(x=z ) e−ν 2 y/2 e− ln2 (x/z)/2y dy, z ≤ x
0<y

x 2y3=2


 Z Hz 
p2 q2 Rs(n) 2  
n
o
2.21.1 Ex exp − + ds ; H < ∞
2 Rs(n) 2 2

z
0

 x− I√ 2
 2 +p2 =2 (qx =2)
 z − I√ 2 2 (qz 2 =2) ,


 0<x≤z
 + p =2
=
 x− K√ 2 +p2 =2 (qx2 =2)
, z≤x

z K  2 +p2 =2 (qz 2 =2)
 − √


Hz 
p2 q2 Rs(n) 2  
 Z

2.21.3 Ex exp −αHz −
+ ds
0 2 Rs(n) 2 2
 x− −1 M √ (qx2 )
− =2q; p2 + 2 =2
1 , 0<x≤z
M− =2q; p2 + 2 =2 (qz 2 )

 z  √

 − −

=
 x− −1 W− =2q;√p2 + 2 =2 (qx2 )
, z≤x

 − −1
z W− =2q;√p2 + 2 =2 (qz 2 )


p2
Hz
q

n  Z  o
2.22.1 Ex exp − + ds ; H < ∞
2 Rs(n) 2 Rs(n)

z
0

 x− I √ √
2  2 +p2 (2 2qx)
 z − I √ 2 2 (2√2qz ) ,


 0<x≤z
2  +p
= √
 x− K2√ 2 +p2 (2 2qx)
, z≤x

 − √ √
z K2  2 +p2 (2 2qz )

414 4. BESSEL PROCESS OF ORDER 

4 Rs(n) ν > 0, n = 2ν + 2 > 2 Hz = min{s : Rs(n) = z}

p2
Hz
q

 Z 
2.22.3 Ex exp −αHz − (n)
2
+ (n) ds
0 2 Rs Rs

 x− −1=2 M √ √ √
−q= 2 ;  2 +p2
(2x 2 )
 z − −1=2 M √ √ 2 2 (2z 2
 √ , 0<x≤z
)


−q= 2 ;  +p
= − −1=2 W
√
 x
 √ √
−q= 2 ;  2 +p2
(2x 2 )
 − −1=2
 √ , z≤x
z W−q=√2 ;√ 2 +p2 (2z 2 )

n  Z Hz

 o
p1I(−∞,r) Rs(n) + q1I[r,∞) Rs(n) ds ; ` Hz , r ∈ dy



2.27.1 Ex exp −
0

  √pC (r√2p; z √2p) √ √

qK +1 (r 2q) 

z≤r = exp −y √ √ √ + √ √
2S (r 2p; z 2p) 2K (r 2q)

√ √
(2p)− S (x 2p; z 2p)

 2r2 +1 S 2 (r√2p; z √2p) dy,
 z≤x≤r

× √
(2p)− x− K (x 2q)
√ √ √ dy, r ≤ x

 +1
2r K (r 2q)S (r 2p; z 2p)


  √qC (r√2q; z √2q) √ √

pI +1 (r 2p) 

r≤z = exp −y √ √ √ + √ √
2S (z 2q; r 2q) 2I (r 2p)

√
(2q)− x− I (x 2p)

 2r +1 I (r√2p)S (z √2q; r√2q) dy, 0 ≤ x ≤ r

 
× √ √
− S (z 2q; x 2q )
(2q ) 
√ √ dy, r≤x≤z

2r 2 +1 2
S (z 2q; r 2q)


n  Z Hz

 o
p1I(−∞,r) Rs(n) + q1I[r,∞) Rs(n) ds ; ` Hz , r = 0



(1) Ex exp −
0

√ √
S (r 2p; x 2p)

 S (r√2p; z √2p) , z ≤ x ≤ r


= √ √
 (x√2q; r√2q) , r ≤ x ≤ z
 S
S (z 2q; r 2q)
 3. STOPPING AT FIRST EXIT TIME 415

4 Rs(n) ν > 0, n = 2ν + 2 > 2 H = Ha,b = min{s : Rs(n) ∈

/ (a, b)}

3. Stopping at first exit time

√ √ √ √
S (b 2 ; x 2 ) + S (x 2 ; a 2 )
3.0.1 Ex e−αH = √ √
S (b 2 ; a 2 )

(n) x−2 − b−2 −βa a−2 − x−2 −βb

3.0.3 Ex e−βRH = e + −2 −2 e
a−2 − b−2 a −b

(n)

x−2 − b−2
3.0.4 Px R H = a = −2 −2
a −b
(a)

(n)

a−2 − x−2
3.0.4 Px R H = b = −2 −2
a −b
(b)

√ √
(n) S (b 2 ; x 2 )
Ex e−αH ; RH

3.0.5 =a = √ √
S (b 2 ; a 2 )
(a)

√ √
(n) S (x 2 ; a 2 )
Ex e−αH ; RH

3.0.5 =b = √ √
S (b 2 ; a 2 )
(b)

(n)

(y−2 − b−2 )(a−2 − x−2 )
3.1.6 Px sup Rs(n) ≥ y, RH = a = −2 −2 −2 , a≤x≤y≤b
0≤s≤H (a −b )(a − y −2 )

(n)
Ex e−αH ; sup Rs(n) ≥ y, RH

3.1.8 =a
0≤s≤H
(1)
√ √ √ √
S (b 2 ; y 2 )S (x 2 ; a 2 )
= √ √ √ √ , a≤x≤y≤b
S (b 2 ; a 2 )S (y 2 ; a 2 )

(n)

(x−2 − b−2 )(a−2 − y−2 )
3.2.6 Px inf Rs(n) ≤ y, RH = b = −2 −2 −2 −2 , a≤y≤x≤b
0≤s≤H (y −b )(a −b )

(n)
Ex e−αH ; inf Rs(n) ≤ y, RH

3.2.8 =b
0≤s≤H
(1)
√ √ √ √
S (b 2 ; x 2 )S (y 2 ; a 2 )
= √ √ √ √ , a≤y≤x≤b
S (b 2 ; y 2 )S (b 2 ; a 2 )
416 4. BESSEL PROCESS OF ORDER 

4 Rs(n) ν > 0, n = 2ν + 2 > 2 H = Ha,b = min{s : Rs(n) ∈

/ (a, b)}

3.3.1 Ex e−γ`(H,r)

 (a−2 − b−2 ) + r2 +1 (r−2 − x−2 )(a−2 − r−2 )


  (a−2 − b−2 ) + , r≤x≤b
r2 +1 (r−2 − b−2 )(a−2 − r−2 )

=
  (a−2 − b−2 ) + r2 +1 (r−2 − b−2 )(x−2 − r−2 )
, a≤x≤r
 (a−2 − b−2 ) + r2 +1 (r−2 − b−2 )(a−2 − r−2 )


 (a−2 − b−2 )y

 
3.3.2 Px `(H, r) ∈ dy = exp − 2 +1 −2 −2 −2
r (r −b )(a − r−2 )

 (x−2 − b−2 )(a−2 − b−2 )


 r2 +1 (r−2 − b−2 )2 (a−2 − r−2 ) dy, r ≤ x ≤ b

× −2 − x−2 )(a−2 − b−2 )
  (a

dy, a ≤ x ≤ r
r 2 +1 (r−2 − b−2 )(a−2 − r−2 )2

 −2
 r − x−2

 r−2 − b−2 , r ≤ x ≤ b
(1) Px `(H, r) = 0 = −2 − r −2
 x

, a≤x≤r
a 2 − r−2
−

3.3.3 Er e−αH−γ`(H,r)

√ √ √ √
S (b 2 ; r 2 ) + S (r 2 ; a 2 )
= √ √ √ √ √ √ =: L
S (b 2 ; a 2 ) + 2 r2 +1 (2 ) S (b 2 ; r 2 )S (r 2 ; a 2 )

 S (x √ 2 ;r 2
√
) S (b 2
√
;x 2
√
)

 S (b√2 √ + √ √ L, r ≤ x ≤ b
;r 2 ) S (b 2 ;r 2 )

−αH−γ`(H,r) 
Ex e = √ √ √ √
 (r √2
 S ;x 2
√
) S (x 2
+ √
;a 2
√
)
L, a ≤ x ≤ r
S (r 2 ;a 2 ) S (r 2 ;a 2 )

√ √

(2 )− S (b 2 ; a 2 )y 
Ex e−αH ; `(H, r) ∈ dy = exp − 2 +1

3.3.4 √ √ √ √
2r S (b 2 ; r 2 )S (r 2 ; a 2 )

(2 )− S (b√2 ; x√2 )S (b√2 ; r√2 ) + S (r√2 ; a√2 )

  
√ √ √ √ dy, r ≤ x ≤ b
2r2 +1 S2 (b 2 ; r 2 )S (r 2 ; a 2 )


× −
√ √ √ √ √ √
(2 ) S (x 2 ; a 2 √) S (b√ 2 ; r 2√) + S√  (r 2 ; a 2 )
 
dy, a ≤ x ≤ r

2r2 +1 S (b 2 ; r 2 )S2 (r 2 ; a 2 )
 3. STOPPING AT FIRST EXIT TIME 417

4 Rs(n) ν > 0, n = 2ν + 2 > 2 H = Ha,b = min{s : Rs(n) ∈

/ (a, b)}

 S (x√2 √
;r 2 )

 S (b√2 √ , r≤x≤b
;r 2 )

 −αH 
(1) Ex e ; `(H, r) = 0 = √ √
 S (r 2 ;x 2 )
√ √ , a≤x≤r
S (r 2 ;a 2 )


(n)
Ex e−γ`(H,r) ; RH

3.3.5 =a
(a)

 (x−2 − b−2 )

  (a−2 − b−2 ) + , r≤x≤b
r2 +1 (r−2 − b−2 )(a−2 − r−2 )

=
  (x−2 − b−2 ) + r2 +1 (r−2 − b−2 )(x−2 − r−2 )
, a≤x≤r
 (a−2 − b−2 ) + r2 +1 (r−2 − b−2 )(a−2 − r−2 )


(n)
Ex e−γ`(H,r) ; RH

3.3.5 =b
(b)

 (a−2 − x−2 ) + r2 +1 (r−2 − x−2 )(a−2 − r−2 )


  (a−2 − b−2 ) + , r≤x≤b
r2 +1 (r−2 − b−2 )(a−2 − r−2 )

= 2 2
 (a − −x − )
, a≤x≤r

 (a−2 − b−2 ) + r2 +1 (r−2 − b−2 )(a−2 − r−2 )


 (a−2 − b−2 )y
 
(n)


3.3.6 Px `(H, r) ∈ dy, RH = a = exp − 2 +1 −2 −2 −2 2
r (r −b )(a −r − )
(a)

 (x−2 − b−2 )

 r2 +1 (r−2 − b−2 )(a−2 − r−2 ) dy, r ≤ x ≤ b

× −2 − x−2 )
  (a

dy, a≤x≤r
r 2 +1 (a−2 − r−2 )2

0, r≤x≤b
(
(n)
x−2 − r−2


(1) Px `(H, r) = 0, RH =a =
, a≤x≤r
a−2 − r−2

 (a−2 − b−2 )y
 
(n)


3.3.6 Px `(H, r) ∈ dy, RH = b = exp − 2 +1 −2 −2 −2 2
r (r −b )(a −r − )
(b)
418 4. BESSEL PROCESS OF ORDER 

4 Rs(n) ν > 0, n = 2ν + 2 > 2 H = Ha,b = min{s : Rs(n) ∈

/ (a, b)}

 (x−2 − b−2 )

 r2 +1 (r−2 − b−2 )2 dy,
 r≤x≤b
×
 (a−2 − x−2 )
dy, a ≤ x ≤ r

r2 +1 (r−2 − b−2 )(a−2 − r−2 )


( r−2 − x−2
, r≤x≤b
(n)
r−2 − b−2


(1) Px `(H, r) = 0, RH =b =
0, a≤x≤r

(n)
Ex e−αH−γ`(H,r) ; RH

3.3.7 =a
(a)

√ √
 S (b 2 ; x 2 )
 S (b√2 ; a√2 ) + 2 √ √ √ √ , r≤x≤b
r2 +1 (2 ) S (b 2 ; r 2 )S (r 2 ; a 2 )


= √ √ √ √ √ √
  (b√2 ; x√2 ) + 2
S r2 +1 (2 ) S (b 2 ; r 2 )S (r 2 ; x 2 )
, a≤x≤r

√ √ √ √
S (b 2 ; a 2 ) + 2 r2 +1 (2 ) S (b 2 ; r 2 )S (r 2 ; a 2 )

(n)
Ex e−αH−γ`(H,r) ; RH

3.3.7 =b
(b)

 S (x√2 ; a√2 ) + 2 √ √ √ √
r2 +1 (2 ) S (x 2 ; r 2 )S (r 2 ; a 2 )

 S (b√2 ; a√2 ) + 2 √ √ √ √ , r≤x≤b
r2 +1 (2 ) S (b 2 ; r 2 )S (r 2 ; a 2 )


= √ √
S (x 2 ; a 2 )
, a≤x≤r

√ √ √ √ √ √
S (b 2 ; a 2 ) + 2 r2 +1 (2 ) S (b 2 ; r 2 )S (r 2 ; a 2 )


(n)
Ex e−αH ; `(H, r) ∈ dy, RH

3.3.8 =a
(a)
√ √

(2 )− S (b 2 ; a 2 )y 
= exp − 2 +1 √ √ √ √
2r S (b 2 ; r 2 )S (r 2 ; a 2 )

√ √
 (2 )− S (b 2 ; x 2 )
 2r2 +1 S (b 2 ; r 2 )S (r√2 ; a√2 ) dy, r ≤ x ≤ b
 √ √
 
× √ √
− S (x 2 ; a 2 )
(2 ) 
dy, a≤x≤r

√ √
2r2 +1 S2 (r 2 ; a 2 )



 0, r≤x≤b
(n) √ √
Ex e−αH ; `(H, r) = 0, RH S (r 2 ; x 2 )

(1) =a =
√ √ , a≤x≤r
S (r 2 ; a 2 )

 3. STOPPING AT FIRST EXIT TIME 419

4 Rs(n) ν > 0, n = 2ν + 2 > 2 H = Ha,b = min{s : Rs(n) ∈

/ (a, b)}

(n)
Ex e−αH ; `(H, r) ∈ dy, RH

3.3.8 =b
(b)

√ √

(2 )− S (b 2 ; a 2 )y 
= exp − 2 +1 √ √ √ √
2r S (b 2 ; r 2 )S (r 2 ; a 2 )

 (2 )− S (b√2 ; x√2 )


 2r2 +1 S 2 (b√2 ; r√2 ) dy,
 r≤x≤b

× √ √
(2 )− S (x 2 ; a 2 )
dy, a ≤ x ≤ r

√ √ √ √
2r2 +1 S (b 2 ; r 2 )S (r 2 ; a 2 )


√ √
 S (x√ 2 ; r√ 2 ) , r ≤ x ≤ b

(n)
Ex e−αH ; `(H, r) = 0, RH S (b 2 ; r 2 )

(1) =b =
0, a≤x≤r


 Z H

3.4.1 Er exp −γ 1I[r,b] Rs(n) ds
0

√ √
2S (b 2 ; r 2 ) + (2 )− (a−2 − r−2 )
= √ √ √ √ √ =: Q
2S (b 2 ; r 2 ) + 2 r2 +1 (a−2 − r−2 )C (r 2 ; b 2 )

 Z H

Ex exp −γ 1I[r,b] Rs(n) ds
0

 −2
x − r−2 a−2 − x−2
 −2 2 + −2 Q, a≤x≤r
a −r  a − r−2

−
= √ √ √ √
 S (x√ 2 ; r√ 2 ) + S (b √2 ; x√ 2 ) Q, r ≤ x ≤ b

S (b 2 ; r 2 ) S (b 2 ; r 2 )

( x−2 − r−2
Z H
, a≤x≤r
a−2 − r−2

(n)


3.4.2 Px 1I[r,b] Rs ds = 0 =
(1) 0 0, r≤x≤b

n  Z H
 (n) o
3.4.5 Ex exp −γ 1I[r,b] Rs(n) ds ; RH =a
(a) 0
420 4. BESSEL PROCESS OF ORDER 

4 Rs(n) ν > 0, n = 2ν + 2 > 2 H = Ha,b = min{s : Rs(n) ∈

/ (a, b)}

√ √ √ √ √
2S (b 2 ; r 2 ) + 2 r2 +1 (x−2 − r−2 )C (r 2 ; b 2 )

√ √ √ √ √ , a≤x≤r
2S (b 2 ; r 2 ) + 2 r2 +1 (a−2 − r−2 )C (r 2 ; b 2 )


= √ √
2S (b 2 ; x 2 )
√ √ √ √ √ , r≤x≤b

2S (b 2 ; r 2 ) + 2 r2 +1 (a−2 − r−2 )C (r 2 ; b 2 )


n  Z H
 (n) o
3.4.5 Ex exp −γ 1I[r,b] Rs(n) ds ; RH =b
(b) 0

(2 )− (a−2 − x−2 )


 2S (b√2 ; r√2 ) + √2 r2 +1 (a−2 − r−2 )C (r√2 ; b√2 ) , a ≤ x ≤ r

 
= √ √ √ 2 +1 (a−2 − r−2 )C (r√2 ; x√2 )
2S  (x 2 ; r 2 ) + 2 r
√ √ √ √ √ , r≤x≤b

2S (b 2 ; r 2 ) + 2 r2 +1 (a−2 − r−2 )C (r 2 ; b 2 )


( x−2 − r−2
Z H
, a≤x≤r
a−2 − r−2

(n) (n)


3.4.6 Px 1I[r,b] Rs ds = 0, RH =a =
(1) 0 0, r≤x≤b

 Z H

3.5.1 Er exp −γ 1I[a,r] Rs(n) ds
0

√ √
2S (r 2 ; a 2 ) + (2 )− (r−2 − b−2 )
= √ √ √ √ √ =: Q
2S (r 2 ; a 2 ) + (r−2 − b−2 ) 2 r2 +1 C (r 2 ; a 2 )

 Z H

Ex exp −γ 1I[a,r] Rs(n) ds
0

√ √ √ √
S (r 2 ; x 2 ) S (x 2 ; a 2 )

√ √ + √ √ Q, a ≤ x ≤ r
S (r 2 ; a 2 ) S (r 2 ; a 2 )


=
 r−2 − x−2 x−2 − b−2
2 2 + −2 −2 Q, r≤x≤b
r −b r −b

− −

0, a≤x≤r
(
Z H 
r−2 − x−2
(n)


3.5.2 Px 1I[a,r] Rs ds = 0 =
, r≤x≤b
(1) 0
r−2 − b−2

n  Z H
 (n) o
3.5.5 Ex exp −γ 1I[a,r] Rs(n) ds ; RH =a
(a) 0
 3. STOPPING AT FIRST EXIT TIME 421

4 Rs(n) ν > 0, n = 2ν + 2 > 2 H = Ha,b = min{s : Rs(n) ∈

/ (a, b)}

√ √ √ √ √
2S (r 2 ; x 2 ) + (r−2 − b−2 ) 2 r2 +1 C (r 2 ; x 2 )

 2S (r√2 ; a√2 ) + (r−2 − b−2 )√2 r2 +1 C (r√2 ; a√2 , a≤x≤r
)

 
=
(x−2 − b−2 )(2 )−
√ √ √ √ √ , r≤x≤b

2S (r 2 ; a 2 ) + (r−2 − b−2 ) 2 r2 +1 C (r 2 ; a 2 )


n  Z H

 (n) o
3.5.5 Ex exp −γ 1I[a,r] Rs(n) ds ; RH =b
(b) 0

√ √
2S (x 2 ; a 2 )

 2S (r√2 ; a√2 ) + (r−2 − b−2 )√2 r2 +1 C (r√2 ; a√2 , a≤x≤r
)

 
= √ √ √ √ √
−2 − x−2 ) 2 r 2 +1 C (r 2 ; a 2
2S  (r 2 ; a 2 ) + ( r  )
√ √ √ √ √ , r≤x≤b

2S (r 2 ; a 2 ) + (r−2 − b−2 ) 2 r2 +1 C (r 2 ; a 2 )


0, a≤x≤r
(
Z H 
(n)
r−2 − x−2
(n)


3.5.6 Px 1I[a,r] Rs ds = 0, RH = b =
, r≤x≤b
(1) 0
r−2 − b−2

 Z H
 
p1I[a,r] Rs(n) + q1I[r,b] Rs(n) ds =: Q


3.6.1 Er exp −
0

√ √ √ √
r−2 −1 2− −1=2 (p− S (b 2q; r 2q) + q− S (r 2p; a 2p))
= √ √ √ √ √ √ √ √ √ √
pS (b 2q; r 2q)C (r 2p; a 2p) + qC (r 2q; b 2q)S (r 2p; a 2p)

 Z H
 
p1I[a,r] Rs(n) + q1I[r,b] Rs(n) ds


Ex exp −
0

√ √ √ √
S (x 2q; r 2q) S (b 2q; x 2q)

 S (b√2q; r√2q) + S (b√2q; r√2q) Q, r ≤ x ≤ b

 
= √ √ √ √
S (r 2p; x 2 p) S (x 2p; a 2p)
√ √ + √ √ Q, a ≤ x ≤ r

S (r 2p; a 2p) S (r 2p; a 2p)


n  Z H
  (n) o
p1I[a,r] Rs(n) + q1I[r,b] Rs(n) ds ; RH


3.6.5 Ex exp − =a
(a) 0
√ √ √ √ √ √ √ √ √ √
pS (b 2q; r 2q)C (r 2p; x 2p) + qC (r 2q; b 2q)S (r 2p; x 2p)
a≤x = √ √ √ √ √ √ √ √ √ √
pS (b 2q; r 2q)C (r 2p; a 2p) + qC (r 2q; b 2q)S (r 2p; a 2p)
x≤r
√ √
r−2 −1 2− −1=2 p− S (b 2q; x 2q)
r≤x = √ √ √
√ √ √ √ √ √ √
pS (b 2q; r 2q)C (r 2p; a 2p) + qC (r 2q; b 2q)S (r 2p; a 2p)
x≤b
422 4. BESSEL PROCESS OF ORDER 

4 Rs(n) ν > 0, n = 2ν + 2 > 2 H = Ha,b = min{s : Rs(n) ∈

/ (a, b)}
n  Z H
  (n) o
p1I[a,r] Rs(n) + q1I[r,b] Rs(n) ds ; RH


3.6.5 Ex exp − =b
(b) 0
√ √
r−2 −1 2− −1=2 q− S (x 2p; a 2p)
a≤x = √ √ √ √ √ √ √ √ √ √
pS (b 2q; r 2q)C (r 2p; a 2p) + qC (r 2q; b 2q)S (r 2p; a 2p)
x≤r

√ √ √ √ √ √ √ √ √ √
pS (x 2q; r 2q)C (r 2p; a 2p) + qC (r 2q; x 2q)S (r 2p; a 2p)
r≤x = √ √ √ √ √ √ √ √ √ √
pS (b 2q; r 2q)C (r 2p; a 2p) + qC (r 2q; b 2q)S (r 2p; a 2p)
x≤b

2 3=2 √2 ; 23 x3=2 2 )
√
2=3 ( 3 b
n  Z H  o S
(n)
3.8.5 Ex exp −γ Rs(n) ds ; RH =a =
S2=3 ( 23 b3=2 2 ; 23 a3=2 2 )
√ √
(a) 0

2 3=2 √2 ; 23 a3=2 2 )
√
2=3 ( 3 x
n  Z H  o S
(n)
3.8.5 Ex exp −γ Rs(n) ds ; RH =b =
S2=3 ( 23 b3=2 2 ; 23 a3=2 2 )
√ √
(b) 0

Z H  S (b2 √ ; x2 √ ) + S (x2 √ ; a2 √ )
=2 =2

3.9.1 Ex exp −2γ (Rs(n) )2 ds = √ √
0
S=2 (b2 ; a2 )

Z H o S (b2 √ ; x2 √ )
=2
n  
(n)
3.9.5 Ex exp −2γ (Rs(n) )2 ds ; RH =a = √ √
0
S=2 (b2 ; a2 )
(a)

Z H o S (x2 √ ; a2 √ )
=2
n  
(n)
3.9.5 Ex exp −2γ (Rs(n) )2 ds ; RH =b = √ √
0
S=2 (b2 ; a2 )
(b)

√ √
b
2y−1−2 S (x 2 + 2 ; a 2 + 2 )
Z
(n)
Ex e−γ Ȟ(H) ; RH

3.12.5 =a = √ √ dy
x
(a−2 − y−2 )S (y 2 + 2 ; a 2 + 2 )
(a)

√ √
x
2y−1−2 S (b 2 + 2 ; x 2 + 2 )
Z
(n)
Ex e−γ Ĥ(H) ; RH

3.12.5 =b = √ √ dy
a
(y−2 − b−2 )S (b 2 + 2 ; y 2 + 2 )
(b)

√ √
b
(2 )− y−1−2 S (x 2 + 2 ; a 2 + 2 )
Z
−αH−γ Ȟ(H) (n)

3.12.7 Ex e ; RH = a = √ √ √ √ dy
x S (y 2 ; a 2 )S (y 2 + 2 ; a 2 + 2 )
(a)

√ √
x
(2 )− y−1−2 S (b 2 + 2 ; x 2 + 2 )
Z
−αH−γ Ĥ(H) (n)

3.12.7 Ex e ; RH = b = √ √ √ √ dy
a S (b 2 ; y 2 )S (b 2 + 2 ; y 2 + 2 )
(b)
 3. STOPPING AT FIRST EXIT TIME 423

4 Rs(n) ν > 0, n = 2ν + 2 > 2 H = Ha,b = min{s : Rs(n) ∈

/ (a, b)}


3.18.1 Ex exp −γ`(H, r) − η`(H, u)
r≤u

 2 (b; a) +  r2 +1 (b; r)(r; x) + u2 +1 (b; u)(u; x) + (ru)2 +1 (b; u)(u; r)(r; x)
a≤x = 2
 (b; a) +  r2 +1 (b; r)(r; a) + u2 +1 (b; u)(u; a) + (ru)2 +1 (b; u)(u; r)(r; a)
x≤r

 2 (b; a) +  r2 +1 (x; r)(r; a) + u2 +1 (b; u)(u; x)

r≤x = 2
 (b; a) +  r  +1 (b; r)(r; a) + u2 +1 (b; u)(u; a) + (ru)2 +1 (b; u)(u; r)(r; a)
2
x≤u

 2 (b; a) +  r2 +1 (x; r)(r; a) + u2 +1 (x; u)(u; a) + (ru)2 +1 (x; u)(u; r)(r; a)
u≤x =
 2 (b; a) +  r2 +1 (b; r)(r; a) + u2 +1 (b; u)(u; a) + (ru)2 +1 (b; u)(u; r)(r; a)
x≤b


3.18.5 Ex exp −γ`(H, r) − η`(H, u) ; WH = a
(a)
r≤u
 2 (b; x) +  r2 +1 (b; r)(r; x) + u2 +1 (b; u)(u; x) + (ru)2 +1 (b; u)(u; r)(r; x)
a≤x = 2
 (b; a) +  r2 +1 (b; r)(r; a) + u2 +1 (b; u)(u; a) + (ru)2 +1 (b; u)(u; r)(r; a)
x≤r

 2 (b; x) + u2 +1 (b; u)(u; x)

r≤x = 2
 (b; a) +  r2 +1 (b; r)(r; a) + u2 +1 (b; u)(u; a) + (ru)2 +1 (b; u)(u; r)(r; a)
x≤u

 2 (b; x)
u≤x = 2 2 +1
 (b; a) +  r  (b; r)(r; a) + u +1 (b; u)(u; a) + (ru)2 +1 (b; u)(u; r)(r; a)
2
x≤b


3.18.5 Ex exp −γ`(H, r) − η`(H, u) ; WH = b
(b)
r≤u
 2 (x; a)
a≤x = 2
 (b; a) +  r2 +1 (b; r)(r; a) + u2 +1 (b; u)(u; a) + (ru)2 +1 (b; u)(u; r)(r; a)
x≤r

 2 (x; a) +  r2 +1 (x; r)(r; a)

r≤x = 2 2 +1
 (b; a) +  r  (b; r)(r; a) + u2 +1 (b; u)(u; a) + (ru)2 +1 (b; u)(u; r)(r; a)
x≤u

 2 (x; a) +  r2 +1 (x; r)(r; a) + u2 +1 (x; u)(u; a) + (ru)2 +1 (x; u)(u; r)(r; a)
u≤x =
 2 (b; a) +  r2 +1 (b; r)(r; a) + u2 +1 (b; u)(u; a) + (ru)2 +1 (b; u)(u; r)(r; a)
x≤b

(y, x) := x−2ν − y −2ν

424 4. BESSEL PROCESS OF ORDER 

4 Rs(n) ν > 0, n = 2ν + 2 > 2 H = Ha,b = min{s : Rs(n) ∈

/ (a, b)}

√ √
S (2 2 b; 2 2 x)
H
ds
Z
= a = 2 √
n   o
(n)
3.19.5 Ex exp −γ (n)
; RH √
0 Rs S2 (2 2 b; 2 2 a)
(a)

√ √
S2 (2 2 x; 2 2 a)
H
ds
n  Z  o
(n)
3.19.5 Ex exp −γ ; R = b = √ √
0 Rs(n) H S2 (2 2 b; 2 2 a)
(b)

 2Z H o x− (b=x)√ 2 + √
− (x=b)  +
ds
n  2 2 2

(n)
3.20.5 Ex exp − (n)
2
; RH =a = √ √
2 0 Rs a− (b=a)  2 + 2
− (a=b)  +
2 2

(a)

 2Z H o x− (x=a)√ 2 + √
− (a=x)  +
ds
n  2 2 2

(n)
3.20.5 Ex exp − ; R = b = √ √
2 Rs(n) 2 H
b− (b=a)  2 + − (a=b)  +

2 2 2

(b) 0

 Z H
p2 q2 Rs(n) 2  
n
o
(n)
3.21.5 Ex exp − + ds ; R = a
2 Rs(n) 2 2 H


(a) 0

√
x  2 +p2 − S√ 2 +p2 =2 (qb2 =2; qx2 =2)
= √
a  2 +p2 − S√ 2 +p2 =2 (qb2 =2; qa2 =2)

 Z H
p2 q2 Rs(n) 2  
n
o
(n)
3.21.5 Ex exp − + ds ; R = b
2 Rs(n) 2 2 H


(b) 0

√
x  2 +p2 − S√ 2 +p2 =2 (qx2 =2; qa2 =2)
= √
b  2 +p2 − S√ 2 +p2 =2 (qb2 =2; qa2 =2)

p2
H
q

n  Z  o
(n)
3.22.5 Ex exp − + ds ; R = a
2 Rs(n) 2 Rs(n) H


(a) 0
√ √ √
x  2 +p2 − S2√ 2 +p2 (2 2qb; 2 2qx)
= √ √ √
a  2 +p2 − S2√ 2 +p2 (2 2qb; 2 2qa)

p2
H
q

n  Z  o
(n)
3.22.5 Ex exp − + ds ; R = b
2 Rs(n) 2 Rs(n) H


(b) 0
√ √ √
x  2 +p2 − S2√ 2 +p2 (2 2qx; 2 2qa)
= √ √ √
b  2 +p2 − S2√ 2 +p2 (2 2qb; 2 2qa)
 3. STOPPING AT FIRST EXIT TIME 425

4 Rs(n) ν > 0, n = 2ν + 2 > 2 H = Ha,b = min{s : Rs(n) ∈

/ (a, b)}

n  Z H

 o
p1I[a,r) Rs(n) + q1I[r,b] Rs(n) ds ; `(H, r ∈ dy



3.27.1 Ex exp −
0
  √pC (r√2p; a√2p) √ √ √
qC (r 2q; b 2q) 

= exp −y √ √ √ + √ √ √
2S (r 2p; a 2p) 2S (b 2q; r 2q)

 S (b√2q; x√2q)q S (b√2q; r√2q) + p S (r√2p; a√2p)

  
√ √ √ √ dy, r ≤ x ≤ b
2(2pq) r2 +1 S2 (b 2q; r 2q)S (r 2p; a 2p)


× √ √   √ √  √ √
 S (x 2p; a 2p) q S (b√2q; r√2q) + p √S (r √2p; a 2p) dy, a ≤ x ≤ r


2(2pq) r2 +1 S (b 2q; r 2q)S2 (r 2p; a 2p)

n  Z H

 o
(n)
p1I[a,r) Rs(n) + q1I[r,b] Rs(n) ds ; `(H, r ∈ dy, RH



3.27.5 Ex exp − =a
(a) 0
  √pC (r√2p; a√2p) √ √ √
qC (r 2q; b 2q) 

= exp −y √ √ √ + √ √ √
2S (r 2p; a 2p) 2S (b 2q; r 2q)

√ √
(2p)− S (b 2q; x 2q)

 2r2 +1 S (b√2q; r√2q)S (r√2p; a√2p) dy, r ≤ x ≤ b

 
× −
√ √
 (2p) S (x√2p; a√2p) dy,

a≤x≤r
2r2 +1 S2 (r 2p; a 2p)

n  Z H

 o
(n)
p1I[a,r) Rs(n) + q1I[r,b] Rs(n) ds ; ` Hz , r = 0, RH



(1) Ex exp − =a
0
√ √
S (r 2p; x 2p)
= √ √ , a≤x≤r
S (r 2p; a 2p)

n  Z H

 o
(n)
p1I[a,r) Rs(n) + q1I[r,b] Rs(n) ds ; ` Hz , r ∈ dy, RH



3.27.5 Ex exp − =b
(b) 0
  √pC (r√2p; a√2p) √ √ √
qC (r 2q; b 2q) 

= exp −y √ √ √ + √ √ √
2S (r 2p; a 2p) 2S (b 2q; r 2q)

√ √
(2q)− S (b 2q; x 2q)

 2r2 +1 S 2 (b√2q; r√2q) dy,
 r≤x≤b

× √ √
(2q)− S (x 2p; a 2p)
√ √ √ √ dy, a ≤ x ≤ r

2r2 +1 S (b 2q; r 2q)S (r 2p; a 2p)


n  Z H

 o
(n)
p1I[a,r) Rs(n) + q1I[r,b] Rs(n) ds ; ` H, r = 0, RH



(1) Ex exp − =b
0
√ √
S (x 2q; r 2q)
= √ √ , r≤x≤b
S (b 2q; r 2q)
426 4. BESSEL PROCESS OF ORDER 

4 Rs(n) ν > 0, n = 2ν + 2 > 2 % = %(v, z) = min{s : `(s, z) = v}

4. Stopping at inverse local time

 x− I (x√2 ) v

 
 z − I (z √2 ) exp − 2zI (z √2 )K (z √2 ) , 0 ≤ x ≤ z

  
4.0.1 Ex e−α% = √
− K (x 2 )
x  v
 
exp − , z≤x

√ √ √
z − K (z 2 ) 2zI (z 2 )K (z 2 )


1 − e−νv/z , 0≤x≤z
(
x−2 −νv/z


4.0.2 Px % = ∞ =
1 − −2 e , z≤x
(1) z

Px sup Rs(n) < y

4.1.2
0≤s≤%
z≤y
v
  
 exp − z 2 +1 (z −2 − y−2 ) ,
 0≤x≤z
= −2 − y −2
 x v
 
exp − 2 +1 −2 , z≤x≤y

z −2 −y −2 z (z −y −2 )

Ex e−α% ; sup Rs(n) < y


4.1.4
0≤s≤%
z≤y

 x− I (x√2 ) √
y− (2 )− I (y 2 )v

 
 z − I (z √2 ) exp − 2z  +1 I (z √2 )S (y√2 ; z √2 ) ,
 0≤x≤z
  
= √ √ √
 S (y √2 ; x√2 ) exp − y− (2 )− I (y 2 )v
 
, z≤x≤y

√ √ √
S (y 2 ; z 2 ) 2z  +1 I (z 2 )S (y 2 ; z 2 )

inf Rs(n) > y, % < ∞

4.2.2 Px
0≤s≤%
0≤y
 −2
y − x−2 y−2 v
 
 y−2 − z −2 exp − z (y−2 − z −2 ) ,
 y≤x≤z
= −2
 x y−2 v
 
exp − −2 , z≤x

z −2 z (y −z − 2 )

inf Rs(n) > y, % = ∞

(1) Px
0≤s≤%
 4. STOPPING AT INVERSE LOCAL TIME 427

4 Rs(n) ν > 0, n = 2ν + 2 > 2 % = %(v, z) = min{s : `(s, z) = v}

x−2

1 − −2 , z≤y≤x
y




x−2 y−2 v

   
= 1 − −2 exp − −2 2 , y≤x≤z
 y z (y −z − )
−2  −2
x−2 y−2 v
 1− x x

   

− −2 − −2 exp − −2 , y≤z≤x

y − 2 z y z (y −z −2 )

Ex e−α% ; inf Rs(n) > y


4.2.4
0≤s≤%
y≤z
 S (x√2 ; y√2 ) √
y− (2 )− K (y 2 )v

 
 S (z √2 ; y√2 ) exp − 2z  +1 K (z √2 )S (z √2 ; y√2 ) ,
 y≤x≤z
  
= √ √
 x− K (x 2 ) 
y− (2 )− K (y 2 )v

√ exp − √ √ √ , z≤x
z − K (z 2 ) 2z  +1 K (z 2 )S (z 2 ; y 2 )


Ex e−γ`(%,r) ; % < ∞

4.3.1

z −2 −1 ( + r)v
  
exp − , 0≤x≤z
z −2 + r(z −2 − r−2 )




 −2
x + r(x−2 − r−2 ) z −2 −1 ( + r)v
  
z≤r = exp − , z≤x≤r
 z −2 + r(z −2 − r−2 ) z −2 + r(z −2 − r−2 )
x−2 z −2 −1 ( + r)v

  

exp − −2 , r≤x

z 2  + r(z 2 −r 2 ) z + r(z −2 − r−2 )

− − −

  ( + r)v
  
exp − , 0≤x≤r
 + r2 +1 (r−2 − z −2 ) z ( + r2 +1 (r−2 − z −2 )




 + r2 +1 (r−2 − x−2 )  ( + r)v

  
r≤z =
  + r 2 +1 (r−2 − z −2 ) exp − z ( + r2 +1 (r−2 − z −2 ) , r≤x≤z
−2  ( + r)v
 x

  

exp − , z≤x

z −2 z ( + r2 +1 (r−2 − z −2 )

Ex e−γ`(%,r) ; % = ∞

(1)
 z−2−1 ( + r)v
   
1 − exp − −2 2 2 , 0≤x≤z
+ r z + r(z − r )

− −



x−2 + r(x−2 − r−2 ) z−2−1 ( + r)v



   
  + r 1 − z−2 + r(z−2 − r−2 ) exp − z−2 + r(z−2 − r−2 ) , z ≤ x



z≤r = x≤r
 r 2+1 x−2 x −2
1− −
2( + r)(z−2 + r(z−2 − r−2 ))

+ r




z−2−1 ( + r)v

  

× exp − −2 , r≤x

z + r(z − r ) 2 2

− −
428 4. BESSEL PROCESS OF ORDER 

4 Rs(n) ν > 0, n = 2ν + 2 > 2 % = %(v, z) = min{s : `(s, z) = v}

  ( + r)v
   
1 − exp − , 0≤x≤r
+ r z ( + r2+1 (r−2 − z−2 )




 + r2+1 (r−2 − x−2 )  ( + r)v

   


1 − exp − , r≤x≤z

+ r z ( + r2+1 (r−2 − z−2 )


r≤z =
 + r2+1 (r−2 − x−2 ) x−2 + r2+1 (r−2 − z−2 )
−

+ r + r




 ( + r )v

  

× exp − , z≤x

z ( + r2+1 (r−2 − z−2 )


4.3.2 Px `(%, r) ∈ dy, % < ∞

 (v=z + y=r)
 
z≤r = exp − 2 −2 2
z (z −r − )
 1 = 2 √  2 (rz )− −1=2 √vy 
 (rz ) v
 − −
 (z −2 − r−2 ) √y I1 dy, 0≤x≤z
z −2 − r−2




h −2 √  2 (rz )− −1=2 √vy 
 x − r−2 v



√ I1

 −2 2 2
(z −r  ) (rz ) +1= 2 y z −2 − r−2

−
×
 −2ν −2ν −2ν−1
 2 (rz )− −1=2 √vy i
+(z − x )r I dy, z ≤ x ≤ r
z −2 − r−2



 0

2 (rz )− −1=2 vy


x−2


  √ 
I dy, r≤x

r(z −2 − r−2 ) 0 z −2 − r−2


 (v=z + y=r)
 
r≤z = exp − 2 −2
r (r − z −2 )
 2
x (rz )
− − − 1=2 √v  2 (rz )− −1=2 √vy 
2 2 √ I dy, z≤x
z (r   − z 2 ) y 1 r−2 − z −2



 − − −

√
 r−2 − x−2 v 2 (rz )− −1=2 vy
√

 h  

I

(rz ) +1=2 √y 1
 −2
(r − z −2 )2 r−2 − z −2

×  2 (rz )− −1=2 √vy i
+(x−2ν − z −2ν )r−2ν−1 I0

dy, r ≤ x ≤ z
r−2 − z −2




 2 (rz )− −1=2 √vy 





I0 dy, 0≤x≤r

r 2 +1 (r 2  −z 2 ) r−2 − z −2

− −

(2) Px `(%, r) = 0, % < ∞

v
  
exp − , 0≤x≤z
z 2 +1 (z −2 − r−2 )



(x−2 − r−2 )

z≤r = v
 
2  2 exp − 2 +1 −2 2 , z≤x≤r


 (z − −r − ) z (z −r − )

0, r≤x
 4. STOPPING AT INVERSE LOCAL TIME 429

4 Rs(n) ν > 0, n = 2ν + 2 > 2 % = %(v, z) = min{s : `(s, z) = v}

 x−2 
r−2 v

exp − , z≤x
 z −2 z (r−2 − z −2 )



r≤z = (r−2 − x−2 ) 
r−2 v

2 2 exp − −2 , r≤x≤z
  − z −2 )
 (r −z ) z (r

 − −

0, 0≤x≤r

(3) Px `(%, r) = 0, % = ∞
 x−2
1 − −2 , z≤r≤x
r


x−2 r−2 v


   
= 1 − −2 exp − −2 , r≤x≤z
r z (r − z −2 )
 1 − x−2 − x−2 − x−2 exp − r−2 v



    
, r≤z≤x
r−2 z −2 r−2 z (r−2 − z −2 )

4.3.3 Ex e−α%−γ`(%,r)

√ √

(1 + 2 rI (r 2 )K (r 2 ))v 
z≤r = exp − √ √
 +1  √ √ √
2zI (z 2 )(K (z 2 ) + 2 r 2 z ) S (r 2 ; z 2 )K (r 2 )

√
 x− I (x 2 )
√ , 0≤x≤z
z − I (z 2 )



√ √ √ √

x− K (x 2 ) + 2 r +1 (2 ) S (r 2 ; x 2 )K (r 2 )


× √ √ √ √ , z≤x≤r
 z − K (z 2 ) + 2 r +1 (2 ) S (r 2 ; z 2 )K (r 2 )
√
x− K (x 2 )




√ √ √ √ , r≤x
z − K (z 2 ) + 2 r +1 (2 ) S (r 2 ; z 2 )K (r 2

)
√ √

(1 + 2 rI (r 2 )K (r 2 ))v 
r≤z = exp − √ √ √ √ √
2zK (z 2 ) I (z 2 ) + 2 r +1 (2 z ) S (z 2 ; r 2 )I (r 2 )

√
 x− K (x 2 )
√ , z≤x
z − K (z 2 )



√ √ √ √

x− I (x 2 ) + 2 r +1 (2 ) S (x 2 ; r 2 )I (r 2 )


× √ √ √ √ , r≤x≤z
 z − I (z 2 ) + 2 r +1 (2 ) S (z 2 ; r 2 )I (r 2 )
√
x− I (x 2 )




√ √ √ √ , 0≤x≤r
z − I (z 2 ) + 2 r +1 (2 ) S (z 2 ; r 2 )I (r 2

)

Ex e−α% ; `(%, r) ∈ dy

4.3.4

√ √

(2 r)− I (r 2 )v (2 z )− K (z 2 )y 
z≤r = exp −  +1 √ √ √ −  +1 √ √ √
2z S (r 2 ; z 2 )I (z 2 ) 2r S (r 2 ; z 2 )K (r 2 )
430 4. BESSEL PROCESS OF ORDER 

4 Rs(n) ν > 0, n = 2ν + 2 > 2 % = %(v, z) = min{s : `(s, z) = v}

√ √  (2 rz )− √vy 
(2 xr)− I (x 2 ) v

 √ √ √ √ I1 √ √ √ dy, 0≤x≤z
2I (z 2 )S (r 2 ; z 2 ) yrz S (r 2 ; z 2 ) rz



√
(2 rz )− vy 
√

(2 rz )− √ √ v

 h 

 2S 2 (r√2 ; z √2 ) Sν (r 2α, x 2α) √yrz I1 S (r√2 ; z √2 )√rz


 
× √ √   (2 rz )− √vy i
z
+Sν (x 2α, z 2α)  +1 I0 dy, z≤x≤r

√ √ √
r S (r 2 ; z 2 ) rz




√  (2 rz )− √vy 
(2 x)− K (x 2 )




 √ √ √ I √ √ √ dy, r≤x
2r +1 K (r 2 )S (r 2 ; z 2 ) 0
S (r 2 ; z 2 ) rz


√ √

(2 r)− K (r 2 )v (2 z )− I (z 2 )y 
r≤z = exp −  +1 √ √ √ −  +1 √ √ √
2z S (z 2 ; r 2 )K (z 2 ) 2r S (z 2 ; r 2 )I (r 2 )

√ √  (2 rz )− √vy 
(2 xr)− K (x 2 ) v

 √ √ √ √ I1 √ √ √ dy, z≤x
2K (z 2 )S (z 2 ; r 2 ) yrz S (z 2 ; r 2 ) rz



√
(2 rz )− vy 
√

(2 rz )− √ √ v

 h 

 2S 2 (z √2 ; r√2 ) Sν (x 2α, r 2α) √yrz I1 S (z √2 ; r√2 )√rz


 
× √ √   (2 rz )− √vy i
z
+Sν (z 2α, x 2α)  +1 I0 dy, r≤x≤z

√ √ √
r S (z 2 ; r 2 ) rz




√  (2 rz )− √vy 
(2 x)− I (x 2 )




 √ √ √ I √ √ √ dy, 0≤x≤r
2r +1 I (r 2 )S (z 2 ; r 2 ) 0
S (z 2 ; r 2 ) rz


Ex e−α% ; ` %, r = 0


(1)

0, r≤x
√ √ √

 S (r 2 ; x 2 ) (2 r)− I (r 2 )v

  

√ √ exp −  +1 √ √ √ , z≤x≤r
z≤r = S (r 2 ; z 2 ) 2z S (r 2 ; z 2 )I (z 2 )
√ √
 I (x√2 ) exp − (2 r)− I (r 2 )v

  

 √ √ √ , 0≤x≤z
I (z 2 ) 2z  +1 S (r 2 ; z 2 )I (z 2 )

0, 0≤x≤r
√ √ √

 S (x 2 ; r 2 ) (2 r)− K (r 2 )v

  

√ √ exp −  +1 √ √ √ , r≤x≤z
r≤z = S (z 2 ; r 2 ) 2z K (z 2 )S (z 2 ; r 2 )
√ √
 K (x√2 ) exp − (2 r)− K (r 2 )v

  

 √ √ √ , z≤x
K (z 2 ) 2z  +1 K (z 2 )S (z 2 ; r 2 )

 Z %

1I[r,∞) Rs(n) ds ; % < ∞

4.4.1 Ex exp −γ
0

√
z −2 −1 K +1 (r 2 )v
 
z≤r = exp − √ √
z K+1 (r 2 ) − r−2 K−1 (r 2 )
−2
 4. STOPPING AT INVERSE LOCAL TIME 431

4 Rs(n) ν > 0, n = 2ν + 2 > 2 % = %(v, z) = min{s : `(s, z) = v}

1, 0≤x≤z

√ √
 x−2 K+1 (r 2 ) − r−2 K−1 (r 2 )



2  √ 2 √ , z≤x≤r
× z K+1 (r 2 ) − r K−1 (r 2
− − )
√
2 (2 )−1=2 r−−1 x− K (x 2 )



√ √ , r≤x
z −2 K+1 (r 2 ) − r−2 K−1 (r 2

)
 √ −2 −1
(r 2 )
√ √ , 0<x≤r
C  (r 2 ; z 2 )



√

√ √
(2 r)− K +1 (r 2 )v C (r 2 ; x 2 )
  

r≤z = exp −  +1 √ √ √ √ √ , r≤x≤z
2z K (z 2 )C (r 2 ; z 2 )  C (r 2 ; z 2 )
−
√
 x K (x√ 2 ) , z ≤ x




z − K (z 2 )

Z % 
1I[r,∞) Rs(n) ds = 0, % < ∞


4.4.2 Px
(1) 0

z −2 −1 v
  
exp − −2 2  , 0≤x≤z
z −r


 −
2 2 2 1

= x − −r − z − − v
 
−2 − r −2
exp − −2 , z≤x≤r
z z − r−2



r≤x

0,

 Z %

1I[0,r] Rs(n) ds ; % < ∞

4.5.1 Ex exp −γ
0

√

(2 r)− I−1 (r 2 )v 
z≤r = exp −  √ √ √
2z I (z 2 )C−1 (x 2 ; r 2 )
 x− I (x√2 )

 z − I (z √2 ) ,

 0≤x≤z
√ √

 −1
x C−1 (x 2 ; r 2 )

× √ √ , z≤x≤r
 z −1 C−1 (z 2 ; r 2 )
(2 )−+1=2 x−2



, r≤x

√ √
z −1 C−1 (z 2 ; r 2 )


√
r−2 I−1 (r 2 )v
 
r≤z = exp − −2
√ √
z (r I−1 (r 2 ) − z −2 I+1 (r 2 ))
√
2 (2 )−1=2 r−−1 x− I (x 2 )

√ √ , 0≤x≤r
r−2 I−1 (r 2 ) − z −2 I+1 (r 2 )




√ √
r−2 I−1 (r 2 ) − x−2 I+1 (r 2

)

× √ √ , r≤x≤z
 r−2 I−1 (r 2 ) − z −2 I+1 (r 2 )
x−2




, z≤x
z −2

432 4. BESSEL PROCESS OF ORDER 

4 Rs(n) ν > 0, n = 2ν + 2 > 2 % = %(v, z) = min{s : `(s, z) = v}

0, 0≤x≤r

 −2
− x−2
%  
r
Z
(n)


4.5.2 Px 1I[0,r] Rs ds = 0 = −2 − z −2
, r≤x≤z
(1) 0  r

1, z≤x
r≤z

n  Z %

 o
p1I[0,r) Rs(n) + q1I[r,∞) Rs(n) ds ; % < ∞


4.6.1 Ex exp −
0

(2qrz )− K−1 (z 2q)√pI+1 (r 2p)K (r 2q)v

 √ √ √
r≤z = exp − √ √ √ √ √ √
2z ( pS (z 2q; r 2q)I+1 (r 2p)+ √qC (r 2q; z 2q)I (r 2p))
√

(2qrz )− K−1 (z 2q)√qI (r 2p)K+1 (r 2q)v

√ √ √ 
− √ √ √ √ √ √ √ √
2z ( pS (z 2q; r 2q)I+1 (r 2p)+ qC (r 2q; z 2q)I (r 2p))
√ √ −2 −1 √
 q(r 2q) (x=r)− I (x 2p)
√ √ √ √ √ √ √ √ , 0≤x≤r
pS (z 2q; r 2q)I+1 (r 2p) + qC (r 2q; z 2q)I (r 2p)




√ √ √ √ √ √ √ √
pS (x 2q; r 2q)I+1 (r 2p) + qC (r 2q; x 2q)I (r 2p)


× √ √ √ √ √ √ √ √ , r≤x≤z
 pS (z 2q; r 2q)I+1 (r 2p) + qC (r 2q; z 2q)I (r 2p)

 √
x K (x 2q)
−


z≤x

√ ,
z − K (z 2q)


(2prz )− I−1 (z 2p)√pI+1 (r 2p)K (r 2q)v

 √ √ √
z≤r = exp − √ √ √ √ √ √
2z ( pK (r 2q)C (r 2p; z 2p)+ √qK+1 (r 2q)S (r 2p; z 2p))
√

(2prz )− I−1 (z 2p)√qI (r 2p)K+1 (r 2q)v

√ √ √ 
− √ √ √ √ √ √ √ √
2z ( pK (r 2q)C (r 2p; z 2p)+ qK+1 (r 2q)S (r 2p; z 2p))
√
 x− I (x 2p)
√ , 0≤x≤z
z − I (z 2p)



√ √ √ √ √ √ √ √

pK (r 2q)C (r 2p; x 2p) + qK+1 (r 2q)S (r 2p; x 2p)


× √ √ √ √ √ √ √ √ , z≤x≤r
 pK (r 2q)C (r 2p; z 2p) + qK+1 (r 2q)S (r 2p; z 2p)
√ √ −2 −1 √
p(r 2p) (x=r)− K (x 2q)



, r≤x

√ √ √ √ √ √ √ √
pK (r 2q)C (r 2p; z 2p) + qK+1 (r 2q)S (r 2p; z 2p)


n  Z %  o
4.8.1 Ex exp −γ Rs(n) ds ; % < ∞
0
z I2=3 ( 23 x3=2 2 )
  √
, 0≤x≤z
( 2 z 3=2 2 )
  x I
 √
3v

2=3 3

= exp −
4zI2=3 ( 23 z 3=2 2 )K2=3 ( 23 z 3=2 2 ) 
√ √
z  K2=3 ( 23 x3=2 2 )
√
, z≤x
x K2=3 ( 23 z 3=2 2 )
  √


n  Z %  o
4.9.1 Ex exp −2γ (Rs(n) )2 ds ; % < ∞
0
 4. STOPPING AT INVERSE LOCAL TIME 433

4 Rs(n) ν > 0, n = 2ν + 2 > 2 % = %(v, z) = min{s : `(s, z) = v}

x I=2 (x2 )
 − √
v
 
 z − I (z 2 √ ) exp − zI (z 2 √ )K (z 2 √ ) , 0 ≤ x ≤ z


=2 =2 =2
= − K 2√ )
x (x v
 − =2 2 √ exp −
 
, z≤x

√ √
z K=2 (z ) zI=2 (z 2 )K=2 (z 2 )


%
ds
n  Z  o
4.19.1 Ex exp −γ (n)
; %<∞
0 Rs
 √
x I2 (2 2 x)
− v
  
 √ exp − √ √ , 0<x≤z
z I2 (2 2 z ) 4zI2 (2 2 z )K2 (2 2 z )

 −
= √
x− K2 (2 2 x) v
 
√ exp − √ √ , z≤x

z − K2 (2 2 z ) 4zI2 (2 2 z )K2 (2 2 z )


 2Z %
ds
n  o
4.20.1 Ex exp −
; %<∞
2 0 Rsn 2
( )

√  p 2
v  + 2
 
x −ν+ ν 2 +γ 2 
exp − , 0<x≤z

z z


= √  p 2
  x −ν− ν 2 +γ 2 v  + 2
 
 exp − , z≤x
z z

 
z (ln(z=x) + v=z ) −ν 2 y/2−(ln(z/x)+v/z)2 /2y
√ e dy, x ≤ z
x 2y3=2
%
ds
Z  
4.20.4 Px ∈ dy =
Rs(n) 2 
 z (ln(√
x=z ) + v=z ) −ν 2 y/2−(ln(x/z)+v/z)2 /2y


0<y 0 e dy, z ≤ x

x 2y3=2

 Z %
p2 q2 Rs(n) 2  
n
o
4.21.1 Ex exp − (n)
2
+ ds ; % < ∞
0 2 Rs 2
0<x
 x− I√ 2
 2 +p2 =2 (qx =2) v
 
 z − I√ 2 2 (qz 2 =2) exp − , x≤z
zI√ 2 +p2 =2 (qz 2 =2)K√ 2 +p2 =2 (qz 2 =2)



 + p =2
=
 x− K√ 2 +p2 =2 (qx2 =2) 
v

exp − , z≤x

z K  2 +p2 =2 (qz 2 =2) zI√ 2 +p2 =2 (qz 2 =2)K√ 2 +p2 =2 (qz 2 =2)
 − √


p2
%
q

n  Z  o
4.22.1 Ex exp − (n)
2
+ (n) ds ; % < ∞
0<x 0 2 Rs Rs

 x− I √ √
2  2 +p2 (2 2qx) v
 

 z − I √ √ exp − √ √ , x≤z
(2 2qz ) 4zI
2  2 +p2 (2 2qz )K2  2 +p2 (2 2qz )
√ √


2  2 +p2
= √
 x− K2√ 2 +p2 (2 2qx) 
v

exp − , z≤x

 − √ √ √ √
z K2  2 +p2 (2 2qz ) 4zI2√ 2 +p2 (2 2qz )K2√ 2 +p2 (2 2qz )

434 4. BESSEL PROCESS OF ORDER 

4 Rs(n) ν > 0, n = 2ν + 2 > 2 % = %(v, z) = min{s : `(s, z) = v}

n  Z %

 o
p1I[0,r) Rs(n) + q1I[r,∞) Rs(n) ds ; `(%, r) ∈ dy


4.27.1 Ex exp −
0

√ √ √ √ √ √

(2pr)− I (r 2p)v qK+1 (r 2q)y pC (r 2p; z 2p)y 
z≤r = exp − +1 √ √ √ − √ √ − √ √ √
2z S (r 2p; z 2p)I (z 2p) 2K (r 2q) 2S (r 2p; z 2p)
√ √  (2prz )− √vy 
 (2pxr)− I (x 2p) v
√ √ √ √
 2I (z 2p)S (r 2p; z 2p) yrz I √ √ √ dy, 0 ≤ x ≤ z
1
S (r 2p; z 2p) rz



√  (2prz )− √vy 
(2prz )− v


 h √ √
 2S 2 (r√2p; z √2p) Sν (r 2p, x 2p) √yrz I1 S (r√2p; z √2p)√rz


 
×   (2prz )− √vy i
√ √ z
dy, z ≤ x ≤ r

+Sν (x 2p, z 2p)  +1 I0 √ √ √
S (r 2p; z 2p) rz

r



√
(2px)− K (x 2q) (2prz )− vy
√



  

 +1 √ √ √ I0 √ √ √ dy, r≤x
2r K (r 2q)S (r 2p; z 2p) S (r 2p; z 2p) rz

√ √ √ √ √ √

(2qr)− K (r 2q)v qC (r 2q; z 2q)y pI+1 (r 2p)y 
r≤z = exp −  +1 √ √ √ − √ √ √ − √ √
2z S (z 2q; r 2q)K (z 2q) 2S (z 2q; r 2q) 2I (r 2p)
√ √  (2qrz )− √vy 
 (2qxr)− K (x 2q) v
√ √ √ √ I √ √ √ dy, z≤x
2K (z 2q)S (z 2q; r 2q) yrz 1
S (z 2q; r 2q) rz




√  (2qrz )− √vy 
(2qrz )− v


 h √ √
 2S 2 (z √2q; r√2q) Sν (x 2q, r 2q) √yrz I1 S (z √2q; r√2q)√rz


 
×   (2qrz )− √vy i
√ √ z
dy, r ≤ x ≤ z

+Sν (z 2q, x 2q)  +1 I0 √ √ √
S (z 2q; r 2q) rz

r



√
(2qx)− I (x 2p) (2qrz )− vy
√



  
√ √ √ I √ √ √ dy, 0≤x≤r
2r +1 I (r 2p)S (z 2q; r 2q) 0 S (z 2q; r 2q) rz


n  Z %

 o
p1I[0,r) Rs(n) + q1I[r,∞) Rs(n) ds ; ` %, r = 0



(1) Ex exp −
0

0, r≤x

√ √  √
 S (r 2p; x 2p) (2pr) I (r 2p)v

−

  
√ √ exp −  +1 √ √ √ , z≤x≤r
z≤r = S (r 2p; z 2p) 2z S (r 2p; z 2p)I (z 2p)
√ √
 x− I (x√ 2p) exp − (2pr)− I (r 2p)v

  
0≤x≤z

√ √ √ ,
z − I (z 2p) 2z  +1 S (r 2p; z 2p)I (z 2p)

0, 0≤x≤r

√ √ √
 S (x 2q; r 2q) (2qr)− K (r 2q)v

  

√ √ exp −  +1 √ √ √ , r≤x≤z
r≤z = S (z 2q; r 2q) 2z K (z 2q)S (z 2q; r 2q)
√ √
 x− K (x√ 2q) exp − (2qr)− K (r 2q)v

  
z≤x

√ √ √ ,
z − K (z 2q) 2z  +1 K (z 2q)S (z 2q; r 2q)
 P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)

5. BESSEL PROCESS OF ORDER 1/2

1. Exponential stopping

(3)   √
Ex e−βRτ = −βx
e−x 2λ
− e−βx


1.0.3 e +
 − 2 =2 x( − 2 =2)2

√
t x t
(3)
n
2
 
1.0.4 Ex e−βRt = e−βx+β 1− t/2
+ (βt + x)e2βx Erfc √ + √
x 2t 2
√ o
x t

+(βt − x) Erfc √ − √
2t 2
√
z  −|z−x|√2λ √

1.0.5 Px (Rτ(3) ∈ dz) = √ e − e−(z+x) 2λ dz, 0 ≤ z, 0≤x
x 2

z 2 2
e−(z−x) /2t − e−(z+x) /2t dz



1.0.6 Px Rt(3) ∈ dz = √
x 2t

√
(3) y sh(x 2)


1.1.2 Px sup Rs ≥y = √
0≤s≤τ x sh(y 2)
x≤y

y
1.1.4 Px sup Rs(3) ≥ y = ss
e (x, y)
0≤s≤t x t
x≤y

1.1.6 Px sup Rs(3) < y, Rτ(3) ∈ dz
0≤s≤τ
√ √ √
z  ch (y − |z − x|) 2 − ch (y − x − z ) 2




= √ √
dz, x∨z <y
x 2 sh y 2


1.1.8 Px sup Rs(3) < y, Rt(3) ∈ dz
0≤s≤t
∞
z X 2 2
e−(z−x+2ky) /2t
− e−(z+x+2ky) /2t


= √ dz, x∨z <y
x 2t
k=−∞

y √
1.2.2 Px inf Rs(3) ≤ y = e−(x−y) 2λ
0≤s≤τ x
y≤x

y 
x−y

1.2.4 Px inf Rs(3) ≤ y = Erfc √
0≤s≤t x 2t
y≤x

© Springer Basel 2015 435

A.N. Borodin, P. Salminen, Handbook of Brownian Motion - Facts and Formulae,
Probability and Its Applications, DOI 10.1007/978-3-0348-8163-0_12
436 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)

y
(1) Px Rs(3) ≤ y =
inf
0≤s<∞ x
y≤x

z √2 (y−z−x)√2λ √
1.2.6 Px inf Rs(3) ≤ y, Rτ(3) ∈ dz = e sh(y 2λ)dz, y ≤x∧z
0≤s≤τ x

z 2 2
e−(z+x−2y) /2t
− e−(z+x) /2t



1.2.8 Px inf Rs(3) ≤ y, Rt(3) ∈ dz = √ dz
0≤s≤t x 2t
y≤x
√ √
r e−|x−r| 2 − e−(x+r) 2


1.3.1 Ex e−γ`(τ,r) = 1 − √ √


x 2 + 1 − e−2r 2

 √ r√2  r√2 e−|x−r|√2 − e−(x+r)√2

e y
1.3.2 Px `(τ, r) ∈ dy = exp − √ √ dy
x 1 − e−2r 2 2
√

2 sh(r 2)

√ √
r e−|x−r| 2 − e−(x+r) 2




(1) Px `(τ, r) = 0 = 1 − √

x 1 − e−2r 2

1.3.3 Ex e−γ`(t,r) = 1
√ ∞
(−1)l t(k+1)=2  −(|r−x|+2rl)2 /4t
k
r 2X |r − x| + 2rl
X  
+ √ (−γ)k+1 k! e D−k−2 √
x  l!(k − l)! t
k=0 l=0
x + r + 2rl
2
 
−e−(x+r+2rl) /4t
D−k−2 √
t

r(x + r − |x − r|)
(1) Ex e−γ`(∞,r) = 1 −
x(1 + 2r )


1 
1.3.4 Px `(t, r) ∈ dy
2
r
= (esy (−1, 2, r, y + |x − r| − 2r, y) − esy (−1, 2, r, y + x − r, y))dy
x

 r
 1 − sse (x, r), 0≤x≤r

x t 
(1) Px `(t, r) = 0 =
 1 − r Erfc x√− r , r ≤ x

x 2t
 1. EXPONENTIAL STOPPING 437
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)

x + r − |x − r| −y/2r
(2) Px `(∞, r) ∈ dy = e dy
4xr

x + r − |x − r |
(3) Px `(∞, r) = 0 = 1 −
2x
√ n
z  √ √
Ex e−γ`(τ,r) ; Rτ(3) ∈ dz = √ e−|z−x| 2λ − e−(z+x) 2λ

1.3.5
x 2
√ √
(e−|x−r| 2 − e−(x+r) 2 ) −|z−r|√2λ √
o
−(z+r) 2λ
− √ √ e − e dz
2 + (1 − e−2r 2 )

√
y 2

 
1.3.6 Px `(τ, r) ∈ dy, Rτ(3) ∈ dz = exp − √

0<y
1 − e−2r 2
√ √ √ √
z(e−|x−r| 2 − e−(x+r) 2 )(e−|z−r| 2 − e−(z+r) 2 )
× dydz
x 1 − e−2r 2 2
√

z √ n √ √
(1) Px `(τ, r) = 0, Rτ(3) ∈ dz = √ e−|z−x| 2λ − e−(z+x) 2λ
x 2
√ √
e−|x−r| 2 − e−(x+r) 2 −|z−r|√2λ √
o
−(z+r) 2λ
− √ e − e dz
1 − e−2r 2


1 
1.3.8 Px `(t, r) ∈ dy, Rt(3) ∈ dz
2

z
= (esy (0, 2, r, y + |x − r| + |z − r| − 2r, y) − esy (0, 2, r, y + |x − r| + z − r, y)
2x
− esy (0, 2, r, y + |z − r| + x − r, y) + esy (0, 2, r, y + x + z, y))dydz

 Z τ

1.4.1 Ex exp −γ 1I[r,∞) Rs(3) ds
0
√ √
(1 + r 2 + 2 ) sh(x 2)
0≤x =1− √
√ √ √
x( + )( 2 ch(r 2) + 2 + 2 sh(r 2))
x≤r

√ √ √ √
 (r 2 ch(r 2) − sh(r 2))e−(x−r) 2+2
r≤x = + √ √ √ √
+ x( + )( 2 ch(r 2) + 2 + 2 sh(r 2))
438 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)
Z τ

1.4.2 Px 1I[r,∞) Rs(3) ds ∈ dy
0

√ √ √ √
sh(x 2) n r 2 ch(r 2) − sh(r 2)
n
1
0≤x = e−λy √ rλ + √ √ √
x sh(r 2) 2 sh(r 2) y
x≤r
√
√ √ ey cth (r 2) √
2 p oo
− λ th(r 2λ) − √ √ Erfc( λy th(r 2λ)) dy
sh(r 2) ch(r 2)
√ √ √
r 2 ch(r 2) − sh(r 2) e−(x−r) =2y √ x−r
n n 2  
r≤x = e−λy λ + √ √ √ − λ cth(r 2λ)Erfc √
x 2 sh(r 2) y 2y
√ (x−r)√2 cth(r√2)+y cth2 (r√2) √
e x−r
 p oo
+ √ √ Erfc √ + λy cth(r 2λ) dy
sh(r 2) ch(r 2) 2y
√
  1 − r sh(x√2) , 0 ≤ x ≤ r

Z τ
x sh(r 2)
(3)


(1) Px 1I[r,∞) Rs ds = 0 =
0 
0, r≤x

Z t
 Z ∞
1.4.4 Px 1I[r,∞) Rs(3) ds ∈ dv = Bx(27) (t − v, v, y)dy 1I(0,t) (v)dv
0 0

for Bx(27) (u, v, y) see 1.27.2

Z t
r


(1) Px 1I[r,∞) Rs(3) ds = 0 = 1 − ss e t (x, r)
0
x
x≤r
t
r x−r
Z
  
(2) Px 1I[r,∞) Rs(3) ds = t = 1 − Erfc √
0
x 2t
r≤x

n  Z τ
 o
1.4.5 Ex exp −γ 1I[r,∞) Rs(3) ds ; Rτ(3) ∈ dz
0
√
z  √ √

x≤r = √ e−|z−x| 2λ − e−(z+x) 2λ dz
x 2
z≤r
√ √ √ √ √ √
z 2(  + − ) sh(z 2) sh(x 2)e−r 2
− √ √ √ √ dz
x(  + sh(r 2) +  ch(r 2))

√ √ √
z 2 sh(x 2)e(r−z) 2+2
x≤r = √ √ √ √ dz
x(  + sh(r 2) +  ch(r 2))
r≤z

√ √ √
z 2 sh(z 2)e(r−x) 2+2
r≤x = √ √ √ √ dz
x(  + sh(r 2) +  ch(r 2))
z≤r
 1. EXPONENTIAL STOPPING 439
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)

z √ z √
r≤x = √ e−|z−x| 2λ+2γ dz − √ e(2r−x−z) 2λ+2γ dz
x 2 + 2 x 2 + 2
r≤z
√ √ √
z 2 sh(r 2)e(2r−x−z) 2+2
+ √ √ √ √ dz
x(  + sh(r 2) +  ch(r 2))

Z τ

1.4.6 Px 1I[r,∞) Rs(3) ds ∈ dy, Rτ(3) ∈ dz
0

√ √ √
z sh(x 2) sh(z 2) 2 −λy
z≤r = √ √ e
x sh2 (r 2) y
x≤r

√ √ 2
√ p √ 
− 2λ cth(r 2λ)e−λy/ sh (r 2λ) Erfc( λy cth(r 2λ)) dydz

√ √
z sh(z 2) 2 −λy−(x−r)2 /2y √ √
z≤r = √ √ e − 2λ cth(r 2λ)
x sh(r 2) y
r≤x

√ √ √
2λ cth(r 2λ)−λy/ sh2 (r 2λ)

x−r p √ 
×e(x−r) Erfc √ + λy cth(r 2λ) dydz
2y

√ √
z sh(x 2) 2 −λy−(z−r)2 /2y √ √
r≤z = √ √ e − 2λ cth(r 2λ)
x sh(r 2) y
x≤r

√ √ √
2λ cth(r 2λ)−λy/ sh2 (r 2λ)

z−r p √ 
×e(z−r) Erfc √ + λy cth(r 2λ) dydz
2y

√ √
z 2 2 z 2
e−λy−(z−x) /2y + e−λy−(z+x−2r) /2y dydz −


r≤z = √ cth(r 2λ)
x 2y x
r≤x

√ √ √
2λ cth(r 2λ)−λy/sh2 (r 2λ)

z + x − 2r p √ 
×e(z+x−2r) Erfc √ + λy cth(r 2λ) dydz
2y

Z τ

(1) Px 1I[r,∞) Rs(3) ds = 0, Rτ(3) ∈ dz
x≤r 0
√ √ √
z  ch (r − |z − x|) 2 − ch (r − x − z ) 2




= √ √
dz, z≤r
x 2 sh r 2
440 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)

Z t
 Z ∞
1.4.8 Px 1I[r,∞) Rs(3) ds ∈ dv, Rt(3) ∈ dz = Bx(27) (t − v, v, y, z)dy dvdz
v<t 0 0

for Bx(27) (u, v, y, z) see 1.27.6

Z t

(1) Px 1I[r,∞) Rs(3) ds = 0, Rt(3) ∈ dz
x≤r 0

z
z≤r = (cst (r − |z − x|, r) − cst (r − z − x, r))dz
x
  −(z−x)2 =2t
ze−(z+x−2r) =2t
Z t
ze
2 


(2) Px 1I[r,∞) Rs(3) ds = t, Rt(3) ∈ dz = √ − √ dz
0 x 2t x 2t
r≤x
r≤z

 Z τ

1.5.1 Ex exp −γ 1I[0,r] Rs(3) ds
0

√ √
 (1 + r 2) sh(x 2 + 2 )
0≤x = + √ √ √ √
+ x( + )( 2 sh(r 2 + 2 ) + 2 + 2 ch(r 2 + 2 ))
x≤r
√ √ √ √
re(r−x) 2 (1 + r 2) sh(r 2 + 2 )e(r−x) 2
r≤x =1− + √ √ √ √
x( + ) x( + )( 2 sh(r 2 + 2 ) + 2 + 2 ch(r 2 + 2 ))

Z τ

1.5.2 Px 1I[0,r] Rs(3) ds ∈ dy
0

√ √ √
1 + r 2 −λy
0≤x = λe−λy dy + √ e (rcy (0, r, x, 1/ 2λ) − 2λ rcy (−2, r, x, 1/ 2λ))dy
x 2
x≤r

√ √
r −λy−(x−r)√2λ 1 + r 2 −λy−(x−r)√2λ
r≤x = e dy + √ e rcy (0, r, r, 1/ 2λ)
x x 2

√

−2λ rcy (−2, r, r, 1/ 2λ) dy

(
Z τ
 0, 0≤x≤r
r √
(3)
(1) Px 1I[0,r] Rs ds = 0 =
1 − e(r−x) 2λ , r≤x
0
x
 1. EXPONENTIAL STOPPING 441
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)

√
sh(x 2 )

√ √ , 0≤x≤r
∞
x 2 ch(r 2 )
 Z  



1.5.3 Ex exp −γ 1I[0,r] Rs(3) ds = √
(1) 0  1 − r + √sh(r 2√) , r ≤ x

x x 2 ch(r 2 )

Z t
 Z ∞
(3)
1.5.4 Px 1I[0,r] Rs ds ∈ dv = Bx(27) (v, t − v, y)dy 1I(0,t) (v)dv
0 0

for Bx(27) (u, v, y) see 1.27.2

t
r
Z

(1) Px 1I[0,r] Rs(3) ds = t = 1 − ss
e (x, r)
0
x t
x≤r

t
r x−r
Z
  
(2) Px 1I[0,r] Rs(3) ds = 0 = 1 − Erfc √
0
x 2t
r≤x

 1
Z ∞  scy (x, r)dy, 0 ≤ x ≤ r
x


(3)
(3) Px 1I[0,r] Rs ds ∈ dy =
0  1 sc (r, r)dy, r ≤ x
x y

n  Z τ
 o
1.5.5 Ex exp −γ 1I[0,r] Rs(3) ds ; Rτ(3) ∈ dz
0

z √ z √
x≤r = √ e−|z−x| 2λ+2γ dz − √ e−(z+x) 2λ+2γ dz
x 2 + 2 x 2 + 2
z≤r

√ √ √ √ √
z 2(  + − ) sh(z 2 + 2 ) sh(x 2 + 2 )e−r 2+2
+ √ √ √ √ √ dz
x 2 + 2 (  sh(r 2 + 2 ) +  + ch(r 2 + 2 ))

√ √ √
z 2 sh(x 2 + 2 )e(r−z) 2
x≤r = √√ √ √ dz
x(  sh(r 2 + 2 ) +  + ch(r 2 + 2 ))
r≤z

√ √ √
z 2 sh(z 2 + 2 )e(r−x) 2
r≤x = √ √ √ √ dz
x(  sh(r 2 + 2 ) +  + ch(r 2 + 2 ))
z≤r
442 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)
√ √
z  −|z−x|√2λ z  √
r≤x = √ e dz − √ e(2r−x−z) 2λ dz
x 2 x 2
r≤z
√ √ √
z 2 sh(r 2 + 2 )e(2r−x−z) 2
+ √ √ √ √ dz
x(  sh(r 2 + 2 ) +  + ch(r 2 + 2 ))

Z τ

1.5.6 Px 1I[0,r] Rs(3) ds ∈ dy, Rτ(3) ∈ dz
0

√ √
z 2 −λy−(z−r)√2λ
x≤r = e rcy (0, r, x, 1/ 2λ)dydz
x
r≤z
√ √
z 2 −λy−(x−r)√2λ
r≤x = e rcy (0, r, z, 1/ 2λ)dydz
x
z≤r

Z t
 Z ∞
(3) (3)
1.5.8 Px 1I[0,r] Rs ds ∈ dv, Rt ∈ dz = Bx(27) (v, t − v, y, z)dy dvdz
v<t 0 0

for Bx(27) (u, v, y, z) see 1.27.6

Z t

(1) Px 1I[0,r] Rs(3) ds = t, Rt(3) ∈ dz
x≤r 0

z
z≤r = (cst (r − |z − x|, r) − cst (r − z − x, r))dz
x
  −(z−x)2 =2t
ze−(z+x−2r) =2t
Z t
ze
2 


(2) Px 1I[0,r] Rs(3) ds = 0, Rt(3) ∈ dz = √ − √ dz
0 x 2t x 2t
r≤x
r≤z

 Z τ



1.6.1 Ex exp − p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds
0

√ √
 (p − q)(1 + r 2 + 2q) sh(x 2 + 2p)
0≤x = + √ √ √ √
+p x( + p)( + q)( 2 + 2p ch(r 2 + 2p) + 2 + 2q sh(r 2 + 2p))
x≤r

√ √ √ √
 (q − p)(r 2 + 2p ch(r 2 + 2p) − sh(r 2 + 2p))e(r−x) 2+2q
r≤x = + √ √ √ √
+q x( + p)( + q)( 2 + 2p ch(r 2 + 2p) + 2 + 2q sh(r 2 + 2p))

Z τ

Z τ

1.6.2 Px 1I[0,r] Rs(3) ds ∈ du, 1I[r,∞) Rs(3) ds ∈ dv
0 0
 1. EXPONENTIAL STOPPING 443
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)

Z ∞
= λe−λ(u+v) Bx(27) (u, v, y)dy dudv for Bx(27) (u, v, y) see 1.27.2
0

Z τ

Z τ

(1) Px 1I[0,r] Rs(3) ds ∈ du, 1I[r,∞) Rs(3) ds = 0
x≤r 0 0

r
= λe−λu (1 − ss
e (x, r))du
x u

Z τ

Z τ

(3)
(2) Px 1I[0,r] Rs ds = 0, 1I[r,∞) Rs(3) ds ∈ dv
r≤x 0 0

r x−r
  
= λe−λv 1 − Erfc √ dv
x 2v

Z t



1.6.4 Px p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds ∈ dv
0
∞
1 |qt − v | |pt − v |
Z  
= Bx(27) , ,y dy 1I((q∧p)t,(q∨p)t) (v) dv
|p − q | 0
|p − q | |p − q |


for Bx(27) u, v, y see 1.27.2

t
r
Z



(1) Px p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds = pt = 1 − ss
e (x, r)
0
x t
x≤r
t
r x−r
Z


  
(2) Px p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds = qt = 1 − Erfc √
0
x 2t
r≤x

τ
z
n  Z


 o
1.6.5 Ex exp − p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds ; Rτ(3) ∈ dz = Q(24) (z)dz
0
x 0,λ
(24)
for Qa,λ (z) see 1.1.24.5

Z τ

Z τ

(3)
1.6.6 Px 1I[0,r] Rs ds ∈ du, 1I[r,∞) Rs(3) ds ∈ dv, Rτ(3) ∈ dz
0 0

Z ∞
= λe−λ(u+v) Bx(27) (u, v, y, z)dy dudvdz
0

(27)
for Bx (u, v, y, z) see 1.27.6
444 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)

Z τ

Z τ

(1) Px 1I[0,r] Rs(3) ds ∈ du, 1I[r,∞) Rs(3) ds = 0, Rτ(3) ∈ dz
x≤r 0 0

z
= λe−λu (csu (r − |z − x|, r) − csu (r − z − x, r))dudz
x

Z τ

Z τ

(2) Px 1I[0,r] Rs(3) ds = 0, 1I[r,∞) Rs(3) ds ∈ dv, Rτ(3) ∈ dz
r≤x 0 0

z −λv −(z−x) 2 2
/2v
− e−(z+x−2r) /2v


= √ e e dvdz
x 2v

Z t



1.6.8 Px p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds ∈ dv, Rt(3) ∈ dz
0

∞
1 |qt − v | |pt − v |
Z  
= Bx(27) , , y, z dy 1I((q∧p)t,(q∨p)t) (v) dvdz
|p − q | 0
|p − q | |p − q |


for Bx(27) u, v, y, z see 1.27.6

Z t



(1) Px p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds = pt, Rt(3) ∈ dz
x≤r 0

z
= (cst (r − |z − x|, r) − cst (r − z − x, r))dz
x
Z t



(2) Px p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds = qt, Rt(3) ∈ dz
r≤x 0

z 2 2
e−(z−x) /2t
− e−(z+x−2r) /2t


= √ dz
x 2t
 Z τ

1.7.1 Er exp −γ 1I[r,u] Rs(3) ds =: Sr
0

2 r 2 + 0 +  (r0 + 1) th((u − r) =2) sh((u − r) ) − 0  (u − r)

=1−
r  2 + 0  (1 + cth(r0 )) cth((u − r) ) + 02 cth(r0 ) sh((u − r) )
2

 Z τ

Eu exp −γ 1I[r,u] Rs(3) ds =: Su
0
2 u 2 − 0 cth(r0 ) +  (u0 cth(r0 ) − 1) th((u − r) =2)


=1−
u 2  2 + 0  (1 + cth(r0 )) cth((u − r) ) + 02 cth(r0 )

2 0 (u − r) cth(r0 )
−
u  2 + 0  (1 + cth(r0 )) cth((u − r) ) + 02 cth(r0 ) sh((u − r) )


 1. EXPONENTIAL STOPPING 445
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)

 Z τ

Ex exp −γ 1I[r,u] Rs(3) ds
0
r sh(x0 )
0≤x = 1 − (1 − Sr )
x sh(r0 )
x≤r
 r



sh((u − x) ) u  

sh((x − r) )
r≤x = + Sr − + Su −
+ x + sh((u − r) ) x + sh((u − r) )
x≤u

u
u≤x = 1 − (1 − Su )e(u−x)Υ0
x

 Z ∞

1.7.3 Ex exp −γ 1I[r,u] Rs(3) ds
(1) 0

1
0≤x = √ √ √
r 2 sh((u − r) 2 ) + ch((u − r) 2 )
x≤r
√
r sh((u − x) 2 )
r≤x = √ √√ √
x(r 2 sh((u − r) 2 ) + ch((u − r) 2 )) sh((u − r) 2 )
x≤u
√ √ √ √
(sh((u − r) 2 ) + r 2 ch((u − r) 2 )) sh((x − r) 2 )
+
√ √ √ √ √
x 2 (r 2 sh((u − r) 2 ) + ch((u − r) 2 )) sh((u − r) 2 )
√ √ √
u

(sh((u − r) 2 ) + r 2 ch((u − r) 2 )) 
u≤x =1− 1− √ √ √ √
x u 2 (r 2 sh((u − r) 2 ) + ch((u − r) 2 ))

Z ∞

1.7.4 Px 1I[r,u] Rs(3) ds ∈ dy
(1) 0

0≤x e y (0, u − r, 0, r)dy

= rs
x≤r
r 1
r≤x = ss (u − x, u − r) ∗ rs
e y (0, u − r, 0, r)dy + ss (x − r, u − r)
x y 2rx y
x≤u

∗(1 − 2 rsy (−2, u − r, u − r, r) + 2r2 rsy (0, u − r, u − r, r))dy

1
u≤x = (1 − 2 rsy (−2, u − r, u − r, r) + 2r2 rsy (0, u − r, u − r, r))dy
2rx

√
Υs := 2λ + 2s
446 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)

n  Z τ  o
1.8.5 Ex exp −γ Rs(3) ds ; Rτ(3) ∈ dz
0

√ √ √ √ √
8z ( + x)  + zS1=3 23 2 ( + x)3=2 ; 23 2 3=2 K1=3 23 2 ( + z )3=2
   

x≤z = √ dz
(3 )5=3 xK1=3 23 2 3=2
 

√ √ √ √ √
8z ( + z )  + xS1=3 23 2 ( + z )3=2 ; 23 2 3=2 K1=3 23 2 ( + x)3=2
   

z≤x =  √ dz
(3 )5=3 xK1=3 23 2 3=2


n  Z τ  o 2z Ai 21=3 −2=3 ( + z )

(1) E0 exp −γ Rs(3) ds ; Rτ(3) ∈ dz = dz
22=3 Ai 21=3 −2=3 


0

z Ai 0k + (2 )1=3 z e( =2) t
n  Z t  o X∞
2 1/3 0
k
1.8.7 E0 exp −γ Rs(3) ds ; Rt(3) ∈ dz = −2=3
dz
0
0
(2 ) Ai ( k )
(1) k=1

 2Z t  2

1 x sh(t )

1.9.3 Ex exp − (Rs(3) )2 ds = exp −
2 0 (ch(t ))3=2 2 ch(t )

2 t
 Z 
(1) Ex exp −β(Rt(3) )2 − (Rs(3) )2 ds
2 0
1 x2 ( sh(t ) + 2 ch(t ))
 
= −1 sh(t ))3= 2 exp − 1
(ch(t ) + 2 2(ch(t ) + 2 − sh(t ))

Z t
2 
1.9.4 Px Rs(3) ds ∈ dy = ecy (0, 3/2, t, 0, x2 /2)dy
0

√ ∞
Z t
2 
4 2 X (−1)k (k + 3=2) 
(4k + 3)t 
(1) P0 Rs(3) ds < y = √ Erfc √
0
 k! 2 2y
k=0

n  2Z τ  o
1.9.5 Ex exp − (Rs(3) )2 ds ; Rτ(3) ∈ dz
2 0

z 1 + 

2 p p
p
x≤z = √ D− 1 − λ (−x 2γ) − D− 1 − λ (x 2γ) D− 1 − λ (z 2γ)dz
x  2 γ 2 γ 2 γ

z 1 + 

2 p p
p
z≤x = √ D− 1 − λ (−z 2γ) − D− 1 − λ (z 2γ) D− 1 − λ (x 2γ)dz
x  2 γ 2 γ 2 γ
 1. EXPONENTIAL STOPPING 447
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)

n  2Z t  o
1.9.7 Ex exp − (Rs(3) )2 ds ; Rt(3) ∈ dz
2 0

(x2 + z 2 ) ch(t )   xz 
√
z  
xz

= exp − exp − exp − dz
x 2 sh(t ) 2 sh(t ) sh(t ) sh(t )
p

Z t
2 
1.9.8 Px Rs(3) ds ∈ dy, Rt(3) ∈ dz
0
z
= √ (eey (1/2, t, (x2 + z 2 )/2, −xz) − eey (1/2, t, (x2 + z 2 )/2, xz))dydz
x 2
√
S0 2 ; √2 e x

 Z ∞ 
2βRs(3) | |
| |
1.10.3 Ex exp −γ e ds = √  , β<0
(1) 0 x| |I0 | 2|

n  Z τ  o
(3)
1.10.5 Ex exp −γ e2βRs ds ; Rτ(3) ∈ dz
β 6= 0 0

for (z − x)β ≥ 0
√
√ √ 2 z
2 βx 2  K√2=| | | | e 
 
√ √
2zex sign 2  2  2λ/|β|
√
= 2 S e , √ dz
x| | 2λ/|β| | | | | K√2=| | | 2|

for (z − x)β ≤ 0
√
√ √ 2 x
2 βz 2  K√2=| | | | e 
 
√ √
2zez sign 2  2  2λ/|β|
√
= 2 S e , √ dz
x| | 2λ/|β| | | | | K√2=| | | 2|

√
h 


1.11.2 P0 sup ` τ, y > h = √
2 sh(h =2)I0 (h =2)
p p
y∈(0,∞)

K0 (h =2)
nZ h p p p Z h p o
× I0 (v λ/2)K0 (v λ/2)dv − I02 (v λ/2)dv
I0 (h =2) 0
p
0

√ √ √




1 − e−x 2+2
1.12.1 Ex e−γ Ĥ(τ ) = √ + 1− √ √
+ + x 2 + 2

e−v n x√ 1 − e−x2 =2v √


 
x
o
1.12.2 Px Ĥ(τ ) ∈ dv = √ + √ − √ 1 − Erfc √ dv
x v 2v 2 2v
448 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)

√
−γ Ĥ(∞) 1 − e−x 2
1.12.3 Ex e = √
x 2
(1)

√

1I[0;t] (v) n x 1 − e−x =2v 1 − Erfc x= 2v o
2

1.12.4 Px Ĥ(t) ∈ dv = + √ − p dv
x  (t − v )v 2v 2(t − v)
p

1  2

1 − e−x /2v dv


(1) Px Ĥ(∞) ∈ dv = √
x 2v

Ex e−γ Ĥ(τ ) ; Rτ(3) ∈ dz


1.12.5 x∨z <y
√
z 2 √ √ √ 

x≤z = √ √ e(x−z) 2λ − e−x 2λ+2γ−z 2λ dz
x(  + + )
√
z 2 √ √ √ 

z≤x = √ √ e(z−x) 2λ+2γ − e−x 2λ+2γ−z 2λ dz
x(  + + )

√ √ √
sh(x 2 + 2 )(y 2 cth(y 2) − 1)
Ex e−γ Ȟ(τ ) ; sup Rs(3) ∈ dy =

1.13.1 √ dy
0≤s≤τ x sh(y 2 + 2 )
x<y

e−v √ √
1.13.2 Px Ȟ(τ ) ∈ dv, sup Rs(3) ∈ dy = (y 2λ cth(y 2λ) − 1) ssv (x, y)dvdy
0≤s≤τ x
x<y

Ex e−γ Ȟ(t) ; sup Rs(3) ∈ dy = e−γt sst (x, y) ∗ (2y cst (y, y) − 1)dy


1.13.3
0≤s≤t
x<y

1
1.13.4 Px Ȟ(t) ∈ dv, sup Rs(3) ∈ dy = 1I[0,t] (v) ssv (x, y)(2y cst−v (y, y) − 1)dvdy
0≤s≤t x
x<y
√ √
2z sh(x 2 + 2 ) sh(z 2)
Ex e−γ Ȟ(τ ) ; sup Rs(3) ∈ dy, Rτ(3) ∈ dz =

1.13.5 √ √ dydz
0≤s≤τ x sh(y 2 + 2 ) sh(y 2)
x<y
z<y

2z ssv (x; y) sh(z √2)
1.13.6 Px Ȟ(τ ) ∈ dv; sup Rs(3) ∈ dy, Rτ(3) ∈ dz = √ dvdydz
0≤s≤τ xev sh(y 2)
x<y
z<y

2z −γt
Ex e−γ Ȟ(t) ; sup Rs(3) ∈ dy, Rt(3) ∈ dz =


1.13.7 e sst (x, y) ∗ sst (z, y)dydz
0≤s≤t x
x<y
z<y
 1. EXPONENTIAL STOPPING 449
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)

1.13.8 Px Ȟ(t) ∈ dv, sup Rs(3) ∈ dy, Rt(3) ∈ dz = 2 ssv (x, y) sst−v (z, y)dvdydz
0≤s≤t
v<t
1 √ √
Ex e−γ Ĥ(τ ) ; inf Rs(3) ∈ dy = (y 2λ + 1)e−(x−y) 2λ+2γ) dy

1.14.1
0≤s≤τ x
y<x

(y√2 + 1)(x − y) 
(x − y)2 
1.14.2 Px Ĥ(τ ) ∈ dv, inf Rs(3) ∈ dy = √ exp −λv − dvdy
0≤s≤τ x 2v3=2 2v
y<x

√
(x − y)e− t−(x−y) =2t  y 2
2 
Ex e−γ Ĥ(t) ; inf Rs(3) ∈ dy =

1.14.3 √
3= 2 ∗ √ + 1 dy
0≤s≤t x 2t t
y<x

1 √
Ex e−γ Ĥ(∞) ; inf Rs(3) ∈ dy = e−(x−y) 2γ

(1) dy
0≤s<∞ x

1.14.4 Px Ĥ(t) ∈ dv, inf Rs(3) ∈ dy
0≤s≤t
y<x
√
1I[0;t] (v)(x − y) 
(x − y)2  y 2 
= √ exp − + 1 dvdy
x 2v3=2 2v (t − v)
p

x−y

(x − y)2 
(1) Px Ĥ(∞) ∈ dv, inf Rs(3) ∈ dy = √ exp − dvdy
0≤s<∞ x 2v3=2 2v
y<x

2z (y−x)√2λ+2γ (y−z)√2λ

Ex e−γ Ĥ(τ ) ; inf Rs(3) ∈ dy, Rτ(3) ∈ dz =

1.14.5 e e dydz
0≤s≤τ x
y<x
y<z


1.14.6 Px Ĥ(τ ) ∈ dv; inf Rs(3) ∈ dy, Rτ(3) ∈ dz
0≤s≤τ
y<x
2z (x − y) 
(x − y)2  −(z−y)√2λ
y<z = √ exp −λv − e dvdydz
x 2v3=2 2v

Ex e−γ Ĥ(t) ; inf Rs(3) ∈ dy, Rt(3) ∈ dz


1.14.7 y <x∧z
0≤s≤t

z (x − y)(z − y) 1
 
(x − y)2   1 
(z − y)2 
= exp −γt − ∗ exp − dydz
x t3=2 2t t3=2 2t


1.14.8 Px Ĥ(t) ∈ dv, inf Rs(3) ∈ dy, Rt(3) ∈ dz
0≤s≤t
v<t
450 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)

z (x − y)(z − y)

(x − y)2 (z − y)2 
= exp − − dvdydz, y <x∧z
xv3=2 (t − v)3=2 2v 2(t − v)
√ √
a sh (b − x) 2 + b sh (x − a) 2



(3) (3)


1.15.2 Px a < inf Rs , sup Rs < b =1− √

0≤s≤τ 0≤s≤τ x sh (b − a) 2

a b
1.15.4 Px a < inf Rs(3) , sup Rs(3) < b = 1 − ss
e (b − x, b − a) − ss
e (x − a, b − a)
0≤s≤t 0≤s≤t x t x t

1.15.6 Px a < inf Rs(3) , sup Rs(3) < b, Rτ(3) ∈ dz
0≤s≤τ 0≤s≤τ

√ √ √
z  ch (b − a − |z − x|) 2 − ch (a + b − x − z ) 2




= √ √
dz, a < x ∧ z,
x 2 sh (b − a) 2 x∨z <b

1.15.8 Px a < inf Rs(3) , sup Rs(3) < b , Rt(3) ∈ dz
0≤s≤t 0≤s≤t

∞
z X 2 2
e−(z−x+2k(b−a)) /2t
− e−(z+x−2a+2k(b−a)) /2t


= √ dz, a<x∧z
x 2t x∨z <b
k=−∞

Ex e−γ`(τ,r) ; sup Rs(3) < b


1.16.1
0≤s≤τ
r<b
√ √
( r sh((b − r) 2) + b =2) sh(x 2)
p
0≤x =1− √ √ √
x( sh((b − r) 2) sh(r 2) + =2 sh(b 2))
p
x≤r

√ √ √ √
b sh((x − r) 2) ( r sh((b − r) 2) + b =2) sh(r 2) sh((b − x) 2)
p
r≤x = 1− √ − √ √ √ √
x sh((b − r) 2) x( sh((b − r) 2) sh(r 2)+ =2 sh(b 2)) sh((b − r) 2)
p
x≤b
√
y =2 sh(b 2)
 p 


1.16.2 Px `(τ, r) ∈ dy, sup Rs(3) < b = exp − √ √
0≤s≤τ sh((b − r) 2) sh(r 2)
r<b
√ √ √
=2(r sh(b 2) − b sh(r 2)) sh(x 2)
 p
√ √ dy, 0≤x≤r
x sh((b − r) 2) sh2 (r 2)


× √ √ √
=2(r sh(b 2) − b sh(r 2)) sh((b − x) 2)
p
r≤x≤b

√ √ dy,
x sh2 ((b − r) 2) sh(r 2)

 1. EXPONENTIAL STOPPING 451
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)

(1) Px `(τ, r) = 0, sup Rs(3) < b
0≤s≤τ
√
r sh(x 2)

 1 − √ , 0≤x≤r
x sh(r 2)


= √
√

r sh (b − x) 2 + b sh (x − r) 2
 1− , r≤x≤b

 √

x sh (b − r) 2

z
Ex e−γ`(τ,r) ; sup Rs(3) < b, Rτ(3) ∈ dz = Q(25)

1.16.5 (z)dz
0≤s≤τ x 0,λ
r<b
z<b for Q(25)
a,λ (z) see 1.1.25.5

z (25)
1.16.6 Px `(τ, r) ∈ dy, sup Rs(3) < b, Rτ(3) ∈ dz = B0,λ (y, z)dydz
0≤s≤τ x
r<b
(25)
z<b for Ba,λ (y, z) see 1.1.25.6

(1) Px `(τ, r) = 0, sup Rs(3) < b, Rτ(3) ∈ dz
0≤s≤τ
√ √ √
z  ch((r − |z − x|) 2) − ch((r − x − z ) 2)


z≤r = √ √ dz
x 2 sh(r 2)
x≤r
√ √ √
z  ch (b − r − |z − x|) 2 − ch (b + r − z − x) 2




r≤z = √ √
dz
x 2 sh (b − r) 2
r≤x

z (25)
1.16.8 Px `(t, r) ∈ dy, sup Rs(3) < b, Rt(3) ∈ dz = F0,x (t, y, z)dydz
0≤s≤t x
r<b
(25)
z<b for Fa,x (t, y, z) see 1.1.25.8

a √
Ex e−γ`(τ,r) ; a < inf Rs(3) = 1 − e−(x−a) 2λ

1.17.1
0≤s≤τ x
a<r
√
(r − ae−(r−a) 2 ) √
−|x−r| 2λ
√

−(x+r−2a) 2λ
a≤x − √ √
e − e
x 2 + (1 − e−2(r−a) 2 )

√
y 2

 
1.17.2 Px `(τ, r) ∈ dy, a < inf Rs(3) = exp − √

a<r
0≤s≤τ 1 − e−2(r−a) 2
a≤x
452 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)

√ √ √ √
2(r − ae−(r−a) 2 )(e−|x−r| 2 − e−(x+r−2a) 2 )
0<y × dy
x 1 − e−2(r−a) 2 2
√

√ √
a sh (r − x) 2 + r sh (x − a) 2





(1) Px `(τ, r) = 0, a < inf Rs(3) = 1 − √

0≤s≤τ x sh (r − a) 2

x + r − |x − r| x + r − 2a − |x − r|
Ex e−γ`(∞,r) ; a <

1.17.3 inf Rs(3) = 1 − +
0≤s<∞ 2x 2x(1 + 2 (r − a))
(1)
a≤x

(x + r − 2a − |x − r|) −y/2(r−a)
1.17.4 Px `(∞, r) ∈ dy, a < inf Rs(3) = e dy
0≤s<∞ 4x(r − a)
(1)

r
(2) Px `(∞, r) = 0, a < infRs(3) = 1 −
0≤s<∞ x
r≤x
a<r
√ n
z  √ √
Ex e−γ`(τ,r) ; a < inf Rs(3) , Rτ(3) ∈ dz = √ e−|z−x| 2λ − e−(z+x−2a) 2λ

1.17.5
0≤s≤τ x 2
a<r
a<z
√ √
z (e−|x−r| 2 − e−(x+r−2a) 2 ) −|z−r|√2λ √
o
−(z+r−2a) 2λ
− √ √ e − e dz
x 2 + (1 − e−2(r−a) 2 )

√
y 2

 
1.17.6 Px `(τ, r) ∈ dy, a < inf Rs(3) , Rτ(3) ∈ dz = exp − √
−2(r −a) 2
a<r
0≤s≤τ 1−e
a<z
√ √ √ √
z(e−|x−r| 2 − e−(x+r−2a) 2 )(e−|z−r| 2 − e−(z+r−2a) 2 )
0<y × √
dydz
x 1 − e−2(r−a) 2

z √  √ √
(1) Px `(τ, r) = 0, a < inf Rs(3) , Rτ(3) ∈ dz = √ e−|z−x| 2λ − e−(z+x−2a) 2λ
0≤s≤τ x 2

√ √
z e−|x−r| 2 − e−(x+r−2a) 2 −|z−r|√2λ √



−(z+r−2a) 2λ
− √
e − e dz
x 1 − e−2(r−a) 2

z (17)
1.17.8 Px `(t, r) ∈ dy, a < inf Rs(3) , Rt(3) ∈ dz = Fa,x (t, y, z)dydz
0≤s≤t x
a<r
(17)
a<z for Fa,x (t, y, z) see 1.1.17.8
 1. EXPONENTIAL STOPPING 453
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)

√ √ √ √

1.18.1 Ex/√2λ exp −γ 2λ `(τ, r/ 2λ) − η 2λ `(τ, u/ 2λ)

(r + r(1 − e−2u ) − u(er−u − e−r−u ))(e−|x−r| − e−x−r )

r<u =1−
x(1 + (1 − e−2r ) + (1 − e−2u ) + (1 − e−2r )(1 − e2(r−u) ))

(u + u(1 − e−2r ) − r(er−u − e−r−u ))(e−|u−x| − e−u−x )

−
x(1 + (1 − e−2r ) + (1 − e−2u ) + (1 − e−2r )(1 − e2(r−u) ))

1.18.3 Ex exp −γ`(∞, r) − η`(∞, u)
(1)
r (x + r − |x − r|) + (u + 2 r(u − r))(x + u − |x − u|)
r<u =1−
x(4 r(u − r) + 2r + 2u + 1)

1.18.4 Px `(∞, r) ∈ dy, `(∞, u) ∈ dg
(1)
  √yg 
1 
uy + rg
x≤r = exp − I dydg
4r(u − r) 2r(u − r) 0 u − r
r<u

 √yg  √ √
1 
uy + rg
h g  yg i
r<x = exp − (u − x)I + (x − r) √ I dydg
4x(u − r)2 2r(u − r) 0
u−r y 1
u−r
x<u

√   √yg 
g uy + rg

r<u = √ exp − I dydg
4x(u − r) y 2r(u − r) 1 u − r
u≤x

x−r g
  
exp − dg, r ≤ x ≤ u
2x(u − r) 2(u − r)




(2) Px `(∞, r) = 0, `(∞, u) ∈ dg = 1
 exp − g
 
 dg, u≤x
2x 2(u − r)

u
(3) Px `(∞, r) = 0, `(∞, u) = 0 = 1 − , u≤x
x

τ
ds
Z 
1.19.2 Px ∈ dy
0 Rs(3)
0<x
√ √ √
sh2 (y =2) x 2 ch(y 2) 2 2x 
p   
= exp − √ M √ dy
2x ch2 (y =2) sh(y 2) −2,1/2
sh(y 2)
p
454 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)

τ
ds
n  Z  o
1.19.5 Ex exp −γ ; Rτ(3) ∈ dz
0 Rs(3)
0<x

√ √ √ √
 (1 + = 2)z
x≤z = √ M−γ/√2λ,1/2 (2x 2λ)W−γ/√2λ,1/2 (2z 2λ)dz
2x
√ √ √ √
 (1 + = 2)z
z≤x = √ W−γ/√2λ,1/2 (2x 2λ)M−γ/√2λ,1/2 (2z 2λ)dz
2x

τ
ds
Z 
1.19.6 Px ∈ dy, Rτ(3) ∈ dz
0 Rs(3)
0<x

√  (x + z )√2 ch(yp=2)   √
 2z 3=2 2 2xz 
= 1=2 exp − I1 dydz
x sh(y =2) sh(y =2) sh(y =2)
p p p

z 3=2
t
ds √
Z 
1.19.8 Px ∈ dy, Rt(3) ∈ dz = 1=2 ist (1, y/2, 0, x + z, xz)dydz
0 Rs(3)
x
0<x

 2Z t
ds (2t)3=4 (5=4+ 1 + 4 2 =4) −x2 /4t  2
x
 p
1.20.3 Ex exp − = e M √ 2
(3)
2
2 Rs x3=2 ( 2 + 1=4 + 1) −3/4, 4γ +1/4 2t
p
0<x 0

Z t
ds

2(2t)3=4 −y/8−x2 /4t 
3 x2 
1.20.4 Px ∈ dy = e m , dy
Rs(3) 2 x3=2 4 4t

2y
0<x 0

 2Z τ
ds
n  o
1.20.5 Ex exp − (3)
2
; Rτ(3) ∈ dz
2 0 Rs

√
2z z √
√


 √ I√γ 2 +1/4 x 2λ K√γ 2 +1/4 z 2λ dz, 0 < x ≤ z
x

= √
2
 √ z K√ 2
z √
√

x 2λ I√γ 2 +1/4 z 2λ dz, z ≤ x

x γ +1/4

z 3=2 √ √
τ
ds
Z 
1.20.6 Px (3) 2
∈ dy, R (3)
∈ dz = 1=2 e−y/8 kiy/2 (x 2λ, z 2λ)dydz
Rs τ x


0<x 0

 2Z t √
ds z z −(x2 +z2 )/2t √ xz
n  o  
1.20.7 Ex exp − (3)
2
; Rt(3) ∈ dz = √ e I γ 2 +1/4 dz
2 0 Rs t x t
0<x
 1. EXPONENTIAL STOPPING 455
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)

z 3=2 −(x2 +z2 )/2t−y/8

t
ds xz

Z 
1.20.8 Px (3)
2
∈ dy, Rt(3) ∈ dz = 1= 2 e iy/2 dydz
0 Rs 2tx t
0<x

 Z t
p2 q2 Rs(3) 2  


1.21.3 Ex exp −
+ ds
0<x 0 2 Rs(3) 2 2

(2 sh(tq))3=4 (5=4 + 4p2 + 1=4)  2

x q ch(2tq)
 2 
x q
p 
= 3=2 exp − M √ 2
x (q ch(tq)) 3= 4 2
(1 + p + 1=4) 2 sh(2tq ) − 3
sh(2tq)
p
4 , 4p +1/4

 2Z t t
q ds
n Z

2  o
1.21.3 Ex exp − Rs(3) ds ;
∈ dy
2 0 0 Rs(3) 2
(1)

2(2 sh(tq))3=4 
y x2 q ch(2tq)
 
3 x2 q

0<x = 3=2 exp − − m , dy
x (q ch(tq))3=4 8 2 sh(2tq) 2y
4 2 sh(2tq)

 Z τ
p2 q2 Rs(3) 2   (3) z
n
o
1.21.5 Ex exp − + ds ; R ∈ dz = Q(21) (z)dz
2 Rs(3) 2 2 τ x λ


0<x 0

for Q(21)
λ (z) see 1.1.21.5

 Z t
p2 q2 Rs(3) 2   (3)
n
o
1.21.7 Ex exp − + ds ; R ∈ dz
2 Rs(3) 2 2 t


0<x 0

qz 3=2 (x2 + z 2 )q ch(tq)  √ xzq

  
= 1=2 exp − I p2 +1/4 dz
x sh(tq) 2 sh(tq) sh(tq)

 2Z t t
q ds
n Z

2  (3)
o
(1) Ex exp − Rs(3) ds ;
∈ dy, Rt ∈ dz
2 0 0 Rs(3) 2

qz 3=2 y (x2 + z 2 )q ch(tq)  xzq

  
= exp − − i dydz
2x1=2 sh(tq) 8 2 sh(tq) y/2
sh(tq)

 2Z t  Z t
p ds
n o
(3) 2

(3)
(2) Ex exp − ; R ds ∈ dy, R ∈ dz
2 Rs(3) 2 s t


0 0

z 3=2 p
= 1=2 isy ( p2 + 1/4, t, 0, (x2 + z 2 )/2, xz/2)dydz
x
456 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)

t t
ds
Z
2
Z 
(3)
1.21.8 Px Rs(3) ds ∈ dg,
∈ dy, Rt ∈ dz
0 0 Rs(3) 2
0<x
z 3=2
= 1=2 e−y/8 eig (y/2, t, (x2 + z 2 )/2, xz)dgdydz
2x

 2Z
p
τ
ds
 Z τ
ds sh2 (y =2)
n o p
1.22.1 Ex exp − ; ∈ dy =
2 Rs(3) 2 0 Rs
(3)
2x ch2 (y =2)

p
(1) 0
√ √ √
(5=2 + p2 + 1=4) x 2 ch(y 2) 2x 2 
p   
0<x × exp − √ M √ 2 √ dy
(1 + 4p2 + 1) sh(y 2) −2, p +1/4 sh(y 2)
p

τ τ
ds ds
Z Z 
1.22.2 Px
∈ dg, ∈ dy
0 Rs(3) 2 0 Rs(3)
0<x
√ √ √
sh2 (y =2) g x 2 ch(y 2) x 2
p    
= exp − − √ m 2, √ dgdy
4x ch2 (y =2) 8 sh(y 2) g/2
sh(y 2)
p

p2
τ
q
z
n  Z  o
1.22.5 Ex exp − + ds ; R (3)
∈ dz = Q(22) (z)dz
2 Rs(3) 2 Rs(3) τ x λ


0<x 0

for Q(22)
λ (z) see 1.1.22.5

 2Z τ  Z τ
p ds ds
n o
(3)
(1) Ex exp − ; ∈ dy, R ∈ dz
2 Rs(3) 2 0 Rs
(3) τ


0

√  (x + z )√2 ch(yp=2)   √
 2z 3=2 2 2xz 
= 1=2 exp − I√4p2 +1 dydz
x sh(y =2) sh(y =2) sh(y =2)
p p p

τ τ
ds ds
Z Z 
1.22.6 Px
∈ dg, ∈ dy, Rτ(3) ∈ dz
0 Rs(3) 2 0 Rs(3)
0<x
√ √  √
 2z 3=2 g (x + z ) 2 ch(y =2)  2 2xz 
 p
= 1 2 exp − − i dgdydz
=
8x sh(y =2) 8 sh(y =2) g/8
sh(y =2)
p p p

 2Z t  Z t
p ds ds
n o
(3)
1.22.7 Ex exp − 2 ; ∈ dy, R ∈ dz
2 0 Rs(3)


0 Rs
(3) t
(1)
z 3=2 p √
0<x = 1=2 ist ( 4p2 + 1, y/2, 0, x + z, xz)dydz
x
 1. EXPONENTIAL STOPPING 457
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)

t t
ds ds
Z Z 
1.22.8 Px
∈ dg, ∈ dy, Rt(3) ∈ dz
0 Rs(3) 2 0 Rs(3)
0<x
z 3=2 √
= e−g/8 eit (g/8, y/2, x + z, 2 xz)dgdydz
8x1=2
n  Z τ


 o
1.23.1 Ex exp − p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds ; sup Rs(3) < b
0 0≤s≤τ
r<b
2 4((q − p)(rq ch((b − r)q ) + sh((b − r)q )) + bq ) sh(xp )
0≤x = −
p2 xp2 q2 (p sh((b − r)q ) ch(rp ) + q ch((b − r)q ) sh(rp ))
x≤r
b sh((x − r)q )
2  
r≤x = 1−
q2 x sh((b − r)q )
x≤b
4((q − p)(rp cth(rp ) − 1) sh((b − r)q ) − bp ) sh((b − x)q )
+
xp2 q2 (p cth(rp ) + q cth((b − r)q )) sh2 ((b − r)q )

n  Z τ


 o
1.23.5 Ex exp − p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds ; sup Rs(3) < b, Rτ(3) ∈ dz
0 0≤s≤τ
r<b
z (26)
z<b = Q (z)dz, for Q(26)
a,λ (z) see 1.1.26.5
x 0,λ

n  Z τ


 o
1.24.1 Ex exp − p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds ; a < inf Rs(3)
0 0≤s≤τ
a<r

 2 2(q − p)(1 + r ) + a 2 ( −  )e(a−r)p
sh((x − a) )

2 
a (a−x)Υp q q p q p
a≤x = 2 1− e −
p x xp q (q sh((r − a)p ) + p ch((r − a)p ))
x≤r

2 2(2(q − p)(rp ch((r − a)p ) − sh((r − a)p )) − aq2 p )e(r−x)q

r≤x = −
q2 xp q (q sh((r − a)p ) + p ch((r − a)p ))

n  Z ∞
 o
1.24.3 Ex exp −p 1I[0,r) Rs(3) ds ; a < inf Rs(3)
0 0≤s<∞
(1)
√
sh((x − a) 2p)

 x√2p ch((r − a)√2p) ,
 a≤x≤r
= √
 1 − r + √sh((r − a) 2√
 p)
, r≤x
x x 2p ch((r − a) 2p)

√
Υs := 2λ + 2s
458 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)

Z ∞

1.24.4 Px 1I[0,r) Rs(3) ds ∈ dy, a < inf Rs(3)
0 0≤s<∞
(1)
 1
 scy (x − a, r − a)dy, a ≤ x ≤ r
= x
 1 sc (r − a, r − a)dy, r ≤ x
x y

n  Z τ


 o
1.24.5 Ex exp − p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds ; a < inf Rs(3) , Rτ(3) ∈ dz
0 0≤s≤τ
a<r
z (24)
a<z = Q (z)dz, for Q(24)
a,λ (z) see 1.1.24.5
x a,λ

Ex e−γ`(τ,r) ; a < inf Rs(3) , sup Rs(3) < b


1.25.1
0≤s≤τ 0≤s≤τ
r<b

√ √
a sh((r − x) 2) + r sh((x − a) 2)
a≤x =1− √
x sh((r − a) 2)
x≤r
√ √ √ √
=2(r sh((b − a) 2) − b sh((r − a) 2) − a sh((b − r) 2)) sh((x − a) 2)
p
+ √ √ √ √
x( sh((b − r) 2) sh((r − a) 2) + =2 sh((b − a) 2)) sh((r − a) 2)
p

√ √
r sh((b − x) 2) + b sh((x − r) 2)
r≤x =1− √
x sh((b − r) 2)
x≤b
√ √ √ √
=2(r sh((b − a) 2) − b sh((r − a) 2) − a sh((b − r) 2)) sh((b − x) 2)
p
+ √ √ √ √
x( sh((b − r) 2) sh((r − a) 2) + =2 sh((b − a) 2)) sh((b − r) 2)
p

1.25.2 Px `(τ, r) ∈ dy, a < inf Rs(3) , sup Rs(3) < b
0≤s≤τ 0≤s≤τ
r<b

√ √
y  sh((b − a) 2)
 
= exp − √ √ √
2 sh((b − r) 2) sh((r − a) 2)
√ √ √ √ √
(r sh((b − a) 2) − b sh((r − a) 2) − a sh((b − r) 2)) sh((x − a) 2)

√ √ √ dy,
x 2 sh((b − r) 2) sh2 ((r − a) 2)






 a≤x≤r
× √ √ √ √ √
(r sh((b − a) 2) − b sh((r − a) 2) − a sh((b − r) 2)) sh((b − x) 2)
√ √ √ dy,

x 2 sh2 ((b − r) 2) sh((r − a) 2)





r≤x≤b

 1. EXPONENTIAL STOPPING 459
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)

(1) Px `(τ, r) = 0, a < inf Rs(3) , sup Rs(3) < b
0≤s≤τ 0≤s≤τ

√ √
a sh((r − x) 2) + r sh((x − a) 2)

 1 − √ , a≤x≤r
x sh((r − a) 2)


= √
√

r sh (b − x) 2 + b sh (x − r) 2
 1− , r≤x≤b

 √

x sh (b − r) 2

z
Ex e−γ`(τ,r) ; a < inf Rs(3) , sup Rs(3) < b, Rτ(3) ∈ dz = Q(25)

1.25.5 (z)dz
0≤s≤τ 0≤s≤τ x a,λ
r<b

z<b for Q(25)

a,λ (z) see 1.1.25.5

1.25.6 Px `(τ, r) ∈ dy, a < inf Rs(3) , sup Rs(3) < b, Rτ(3) ∈ dz
0≤s≤τ 0≤s≤τ
r<b
z (25) (25)
z<b = B (y, z)dydz, for Ba,λ (y, z) see 1.1.25.6
x a,λ

(1) Px `(τ, r) = 0, a < inf Rs(3) , sup Rs(3) < b, Rτ(3) ∈ dz
0≤s≤τ 0≤s≤τ

√ √ √
z  ch((r − a − |z − x|) 2) − ch((r + a − x − z ) 2)


z≤r = √ √ dz
x 2 sh((r − a) 2)
x≤r

√ √ √
z  ch((b − r − |z − x|) 2) − ch((b + r − z − x) 2)


r≤z = √ √ dz
x 2 sh((b − r) 2)
r≤x

1.25.8 Px `(t, r) ∈ dy, a < inf Rs(3) , sup Rs(3) < b, Rt(3) ∈ dz
0≤s≤t 0≤s≤t
r<b
z (25) (25)
z<b = Fa,x (t, y, z)dydz, for Fa,x (t, y, z) see 1.1.25.8
x

n  Z τ



1.26.1 Ex exp − p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds ;
a<r 0
o
r<b a < inf Rs(3) , sup Rs(3) < b
0≤s≤τ 0≤s≤τ
460 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)

 a sh((r − x)p )
 
a≤x = 1−
+p x sh((r − a)p )
x≤r
(p − q)(rq cth((b − r)q ) + 1) sh((x − a)p )
+
x( + p)( + q)(p cth((r − a)p ) + q cth((b − r)q )) sh((r − a)p )

(( + p)q b sh((r − a)p ) + ( + q)p a sh((b − r)q )) sh((x − a)p )

−
x( + p)( + q)(p cth((r − a)p ) + q cth((b − r)q )) sh((b − r)q ) sh2 ((r − a)p )

 b sh((x − r)q )
 
r≤x = 1−
+q x sh((b − r)q )
x≤b
(q − p)b(p cth((r − a)p ) − 1) sh((b − x)q )
+
x( + p)( + q)(p cth((r − a)p ) + q cth((b − r)q )) sh((b − r)q )

(( + p)q b sh((r − a)p ) + ( + q)p a sh((b − r)q )) sh((b − x)q )

−
x( + p)( + q)(p cth((r − a)p ) + q cth((b − r)q )) sh2 ((b − r)q ) sh((r − a)p )
n  Z τ



1.26.5 Ex exp − p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds ; a < inf Rs(3) ,
0 0≤s≤τ
a<r
z
o
r<b sup Rs(3) < b, Rτ(3) ∈ dz = Q(26) (z)dz
0≤s≤τ x a,λ

for Q(26)
a,λ (z) see 1.1.26.5

n  Z τ


 1 o
1.27.1 Ex exp − p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds ; `(τ, r) ∈ dy
0
2
2 r ch(rp )

1 r 1  
yp ch(rp ) sh(xp )

x≤r = e−yΥq − + + 2 exp − dy
x p sh(rp ) p2 q q sh(rp ) sh(rp )
2 
r ch(rp ) 1 r 1 
yp ch(rp )

r≤x = e−(y+x−r)Υq − + + 2 exp − dy
x p sh(rp ) p2 q q sh(rp )

n  Z τ


 o
(1) Ex exp − p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds ; `(τ, r) = 0
0
r sh(xp )

 
,1− x≤r
+p x sh(rp )


=
  1 − r e(r−x)Υq ,
 
r≤x

+q x

√
Υs := 2λ + 2s
 1. EXPONENTIAL STOPPING 461
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)

Z τ τ
1
Z 



1.27.2 Px 1I[0,r) Rs(3) ds ∈ du, 1I[r,∞) Rs(3) ds ∈ dv, `(τ, r) ∈ dy
0 0
2

=: λe−λ(u+v) Bx(27) (u, v, y)dudvdy

2
x≤r = λe−λ(u+v) ((r escsu (x, r, r, y) − ess
f u (x, r, 0, y)) hv (1, y)
x
+ essu (x, r, 0, y)(r hv (0, y) + hv (−1, y)))dudvdy
2
r≤x = λe−λ(u+v) ((r ecsu (r, r, 0, y) − esu (−2, 0, r, 0, y)) hv (1, y + x − r)
x
+ esu (0, 0, r, 0, y)(r hv (0, y + x − r) + hv (−1, y + x − r)))dudvdy

Z τ

Z τ

(1) Px 1I[0,r] Rs(3) ds ∈ du, 1I[r,∞) Rs(3) ds = 0, `(τ, r) = 0
x≤r 0 0

r
= λe−λu


1 − ss
e u (x, r) du
x

Z τ

Z τ

(3)
(2) Px 1I[0,r] Rs ds = 0, 1I[r,∞) Rs(3) ds ∈ dv, `(τ, r) = 0
r≤x 0 0

r x−r
  
= λe−λv 1 − Erfc √ dv
x 2v

n  Z ∞
 o
1.27.3 Ex exp −p 1I[0,r) Rs(3) ds ; `(∞, r) ∈ dy
(1) 0
 sh(x√2p)  y√p ch(r√2p) 
 2x sh(r√2p) exp − √2 sh(r√2p) dy, 0 ≤ x ≤ r

=  y√p ch(r√2p) 
 1
exp − √ √ dy, r≤x
2x 2 sh(r 2p)


Z t


1 
1.27.4 Px p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds ∈ dv, `(t, r) ∈ dy
0
2

1 
|qt − v | |pt − v |

= Bx(27) , ,y 1I((q∧p)t,(q∨p)t) (v) dvdy
|p − q | |p − q | | p − q |


for Bx(27) u, v, y see 1.27.2

t
r
Z



(1) Px p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds = pt, `(t, r) = 0 = 1 − ss
e (x, r)
0
x t
x≤r
462 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)

t
r x−r
Z


  
(2) Px p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds = qt, `(t, r) = 0 = 1 − Erfc √
0
x 2t
r≤x

Z ∞

(3) Px 1I[0,r) Rs(3) ds ∈ du, `(∞, r) ∈ dy
0
 1
ess (x, r, 0, y/2)dudy, 0 ≤ x ≤ r
2x u

=
 1 es (0, 0, r, 0, y/2)dudy, r ≤ x
2x u
n  Z τ


 1 o
1.27.5 Ex exp − p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds ; `(τ, r) ∈ dy, Rτ(3) ∈ dz
0
2

2z 
yp ch(rp ) sh(xp ) sh(zp )

x≤r = e−yΥq exp − dydz
x sh(rp ) sh2 (rp )
z≤r
2z 
yp ch(rp ) sh(xp )

x≤r = e−(y+z−r)Υq exp − dydz
x sh(rp ) sh(rp )
r≤z
2z 
yp ch(rp ) sh(zp )

r≤x = e−(y+x−r)Υq exp − dydz
x sh(rp ) sh(rp )
z≤r
2z 
yp ch(rp )

r≤x = e−(y+x+z−2r)Υq exp − dydz
x sh(rp )
r≤z

n  Z τ


 o
(1) Ex exp − p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds ; `(τ, r) = 0, Rτ(3) ∈ dz
0

z(ch((r − |z − x|)p ) − ch((r − z − x)p ))

x≤r = dz
xp sh(rp )
z≤r
z −|z−x|Υq z (2r−x−z)Υq
r≤x = e dz − e dz
xq xq
r≤z

Z τ τ
1
Z 
(3)



1.27.6 Px 1I[0,r) Rs ds ∈ du, 1I[r,∞) Rs(3) ds ∈ dv, `(τ, r) ∈ dy, Rτ(3) ∈ dz
0 0
2

=: λe−λ(u+v) Bx(27) (u, v, y, z)dudvdydz

√
Υs := 2λ + 2s
 1. EXPONENTIAL STOPPING 463
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)

2z
x≤r = λe−λ(u+v) esssu (x, z, r, y) hv (1, y)dudvdydz
x
z≤r
2z
x≤r = λe−λ(u+v) essu (x, r, 0, y) hv (1, y + z − r)dudvdydz
x
r≤z
2z
r≤x = λe−λ(u+v) essu (z, r, 0, y) hv (1, y + x − r)dudvdydz
x
z≤r
2z
r≤x = λe−λ(u+v) esu (0, 0, r, 0, y) hv (1, y + x + z − 2r)dudvdydz
x
r≤z

Z τ

Z τ

(3)
(1) Px 1I[0,r) Rs ds ∈ du, 1I[r,∞) Rs(3) ds = 0, `(τ, r) = 0, Rτ(3) ∈ dz
x≤r 0 0

z
= λe−λu (csu (r − |z − x|, r) − csu (r − z − x, r))dudz
x

Z τ

Z τ

(3)
(2) Px 1I[0,r) Rs ds = 0, 1I[r,∞) Rs(3) ds ∈ dv, `(τ, r) = 0, Rτ(3) ∈ dz
r≤x 0 0

z 2 2
= λe−λv e−(z−x) /2v
− e−(z+x−2r) /2v


√ dvdz
x 2v

Z t


1 
1.27.8 Px p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds ∈ dv, `(t, r) ∈ dy, Rt(3) ∈ dz
0
2

1 
|qt − v | |pt − v |

= Bx(27) , , y, z 1I((q∧p)t,(q∨p)t) (v) dvdydz
|p − q | |p − q | |p − q |


for Bx(27) u, v, y, z see 1.27.6

Z t



(1) Px p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds = pt, `(t, r) = 0, Rt(3) ∈ dz
x≤r 0

z
z≤r = (cst (r − |z − x|, r) − cst (r − z − x, r))dz
x

Z t



(2) Px p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds = qt, `(t, r) = 0, Rt(3) ∈ dz
r≤x 0
464 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)

z 2 z 2
r≤z = √ e−(z−x) /2t dz − √ e−(x+z−2r) /2t dz
x 2t x 2t

Ex e−γ Ĥ(τ )−ηȞ(τ ) ; inf Rs(3) ∈ da, sup Rs(3) ∈ db dadb

 
1.28.1
0≤s≤τ 0≤s≤τ
a<x
x<b
√ √ √ √ √ √
2 + 2 sh((b − x) 2 + 2 + 2)(b 2 ch((b − a) 2) − sh((b − a) 2) − a 2)
= √ √ √
x sh((b − a) 2 + 2 + 2) sh((b − a) 2 + 2) sh((b − a) 2)

√ √ √ √ √ √
2 + 2 sh((x − a) 2 + 2 + 2)(a 2 ch((b − a) 2) + sh((b − a) 2) − b 2)
+ √ √ √
x sh((b − a) 2 + 2 + 2) sh((b − a) 2 + 2 ) sh((b − a) 2)

1.28.2 Px Ĥ(τ ) ∈ du, Ȟ(τ ) ∈ dv, inf Rs(3) ∈ da, sup Rs(3) ∈ db
0≤s≤τ 0≤s≤τ

√ √ √ √
b 2 ch((b − a) 2) − sh((b − a) 2) − a 2
u<v = √
x sh((b − a) 2)

×e−λv ssu (b − x, b − a) sv−u (b − a) dudvdadb

√ √ √ √
a 2 ch((b − a) 2) + sh((b − a) 2) − b 2
v<u = √
x sh((b − a) 2)

×e−λu ssv (x − a, b − a) su−v (b − a) dudvdadb

1.28.4 Px Ĥ(t) ∈ du, Ȟ(t) ∈ dv, inf Rs(3) ∈ da, sup Rs(3) ∈ db
0≤s≤t 0≤s≤t

for 0 < u ∧ v < u ∨ v < t, a<x<b

1
u<v = ssu (b − x, b − a) sv−u (b − a)
x


× 2b cst−v (b − a, b − a) − 1 − 2a cst−v (0, b − a) dudvdadb

1
v<u = ssv (x − a, b − a) su−v (b − a)
x


× 2a cst−u (b − a, b − a) + 1 − 2b cst−u (0, b − a) dudvdadb
 1. EXPONENTIAL STOPPING 465
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)

Ex e−γ Ĥ(τ )−ηȞ(τ ) ; inf Rs(3) ∈ da, sup Rs(3) ∈ db, Rτ(3) ∈ dz

1.28.5
0≤s≤τ 0≤s≤τ
a<x
x<b
√ √ √
2z 2 + 2 sh((b − x) 2 + 2 + 2) sh((z − a) 2)
= √ √ √ dadbdz
x sh((b − a) 2 + 2 + 2) sh((b − a) 2 + 2) sh((b − a) 2)
√ √ √
2z 2 + 2 sh((x − a) 2 + 2 + 2) sh((b − z ) 2)
+ √ √ √ dadbdz
x sh((b − a) 2 + 2 + 2) sh((b − a) 2 + 2 ) sh((b − a) 2)

1.28.6 Px Ĥ(τ ) ∈ du, Ȟ(τ ) ∈ dv, inf Rs(3) ∈ da, sup Rs(3) ∈ db, Rτ(3) ∈ dz
0≤s≤τ 0≤s≤τ
√
2z sh((z − a) 2)
u<v = e−λv √ ss (b − x, b − a) sv−u (b − a) dudvdadbdz
x sh((b − a) 2) u
√
2z −λu sh((b − z ) 2)
v<u = e √ ss (x − a, b − a) su−v (b − a) dudvdadbdz
x sh((b − a) 2) v

1.28.8 Px Ĥ(t) ∈ du, Ȟ(t) ∈ dv, inf Rs(3) ∈ da, sup Rs(3) ∈ db, Rt(3) ∈ dz
0≤s≤t 0≤s≤t

for 0 < u ∧ v < u ∨ v < t, a<x∧z <x∨z <b

2z
u<v = ssu (b − x, b − a) sv−u (b − a) sst−v (z − a, b − a)dudvdadbdz
x

2z
v<u = ssv (x − a, b − a) su−v (b − a) sst−u (b − z, b − a)dudvdadbdz
x

n  Z τ



1.29.1 Ex exp − p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds ; a < inf Rs(3) ,
0 0≤s≤τ

1 o
sup Rs(3) < b, `(τ, r) ∈ dy
0≤s≤τ 2
r ch((b − r)q ) − b r ch((r − a)p ) − a 2(p − q)
 
= 2λ + +
xq sh((b − r)q ) xp sh((r − a)p ) xp2 q2

sh((x − a)p )
 sh((r − a) ) dy, a≤x≤r
yq ch((b − r)q ) yp ch((r − a)p )

p

× exp − −
sh((b − r)q ) sh((r − a)p )   sh((b − x)q ) dy, r≤x≤b
sh((b − r)q )

√
Υs := 2λ + 2s
466 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)

n  Z τ



(1) Ex exp − p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds ; a < inf Rs(3) ,
0 0≤s≤τ
o
sup Rs(3) < b, `(τ, r) = 0
0≤s≤τ

 a sh((r − x)p ) + r sh((x − a)p )

  
 +p 1−

x sh((r − a)p )
, a≤x≤r
=
  1 − r sh((b − x)q ) + b sh((x − r)q ) , r ≤ x ≤ b
 

+q x sh((b − r)q )

n  Z ∞
 o
1.29.3 Ex exp −p 1I[0,r) Rs(3) ds ; a < inf Rs(3) , `(∞, r) ∈ dy
0 0≤s<∞
(1)

 sh((x − a)√2p)  y√p ch((r − a)√2p) 

 2x sh((r − a)√2p) exp − √2 sh((r − a)√2p) dy, a ≤ x ≤ r

=  √ √
 1 exp − y√ p ch((r − a)√ 2p) dy,

r≤x

2x 2 sh((r − a) 2p)

Z ∞

1.29.4 Px 1I[0,r) Rs(3) ds ∈ du, a < inf Rs(3) , `(∞, r) ∈ dy
0 0≤s<∞
(1)
 1
ess (x − a, r − a, 0, y/2)dudy, a ≤ x ≤ r
2x u

= 1
 es (0, 0, r − a, 0, y/2)dudy, r≤x
2x u
n  Z τ



1.29.5 Ex exp − p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds ; a < inf Rs(3) ,
0 0≤s≤τ

1 o
z (29)
sup Rs(3) < b, `(τ, r) ∈ dy, Rτ(3) ∈ dz = Q (y, z)dydz
0≤s≤τ 2 x λ

for Q(29)
λ (y, z) see 1.1.29.5

n  Z τ



(1) Ex exp − p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds ; a < inf Rs(3) ,
0 0≤s≤τ
o
sup Rs(3) < b, `(τ, r) = 0, Rτ(3) ∈ dz
0≤s≤τ

√
Υs := 2λ + 2s
 1. EXPONENTIAL STOPPING 467
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)

z (ch((r − a − |z − x|)p ) − ch((r + a − z − x)p ))

x≤r =
xp sh((r − a)p )
z≤r
z (ch((b − r − |z − x|)q ) − ch((b + r − z − x)q ))
r≤x =
xq sh((b − r)q )
r≤z
Z τ

Z τ

1.29.6 Px 1I[0,r) Rs(3) ds ∈ du, 1I[r,∞) Rs(3) ds ∈ dv, a < inf Rs(3) ,
0 0 0≤s≤τ

`(; r)

ze −(u+v )
sup Rs(3) < b, ∈ dy, Rτ(3) ∈ dz = Bx(29) (u, v, y, z)dudvdydz
0≤s≤τ 2 x

for Bx(29) (u, v, y, z) see 1.1.29.6

Z t



1.29.8 Px p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds ∈ dv, a < inf Rs(3) , sup Rs(3) < b,
0 0≤s≤t 0≤s≤t
 z 1I
`(t; r) ((q∧p)t;(q∨p)t) (v ) (29) |qt − v | |pt − v |
 
∈ dy, Rt(3) ∈ dz = Bx , , y, z dvdydz
2 x| p − q | |p − q | |p − q |

for Bx(29) (u, v, y, z) see 1.1.29.6

 pe x  1 q 2p 2pe x 
 Z ∞ 2   exp − S + ; 1; ;
p 2βRs(3) (3) | | 2 p| | | | | |
1.30.3 Ex exp − e + qeβRs ds = 1 q 2p 
0
2 x| |M + ; 1;
β<0 2 p| | | |

 Z τ 2 o |β| 1+2√2λ/|β|
p 2βRs(3)
n  
βRs(3) (3)
1.30.5 Ex exp − e + qe ds ; Rτ ∈ dz
0
2 2p
z √
p z
 
for (z − x)β ≥ 0 = ex sign β 2λ−p/|β| exp − eβx −
px | | 2
 2p
e z
2p  W−q=p| |; 2=|

√ √ √

q 2 1 2 2 2p | | |
×S + + , + 1, eβx ,  2p  dz
p| | | | 2 | | | | | |
W−q=p| |; 2=| | √
| |
z √
p x
 
for (z − x)β ≤ 0 = ez sign β 2λ−p/|β| exp − eβz −
px | | 2
 2p
e x
2p βz 2p  W−q=p| |; 2=|

√ √ √

q 2 1 2 2 | | |
×S + + , + 1, e ,  2p  dz
p| | | | 2 | | | | | |
W √
−q=p| |; 2=| | | |

√ √ √


1.31.1 Ex/√2λ exp −γ 2λ sup `(s, u/ 2λ) − α−1 `(s, r/ 2λ)
0≤s≤τ
0<α
468 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 τ ∼ Exp(λ), independent of R(3)
u  − e−2r + e−2u e−|x−u| − e−x−u



= 1−
x 2 1 − e−2(u∧r) 1 − e−2|u−r| +  − e−2r + e−2u

u 2 1 − e−2(u∧r) 1 − e−2|u−r| − ( − e−2r + e−2u )(1 − e−2u ) e−|x−r| − e−x−r

−
x 2 1 − e−2(u∧r) 1 − e−2|u−r| +  − e−2r + e−2u e−|u−r| − e−u−r

 √ √
√ 
1.31.2 Px/√2λ sup `(s, u/ 2λ) − α−1 `(s, r/ 2λ) > h/ 2λ
0≤s≤τ
0<α
u  − e−2r + e−2u h h −|x−r| −x−r
e −e


= exp −
x 2 1 − e−2(u∧r) 1 − e−2|u−r| e−|u−r| − e−u−r

 − e−2r + e−2u )(1 − e−2u )

 −|x−r| −x−r
(e −e −|u−x| −x−u
i
− − e + e
2 1 − e−2(u∧r) 1 − e−2|u−r| e−|u−r| − e−u−r


1
  p
(r − u)2 +4 (u ∧ r)|u − r| + r − u
(1) Px sup `(s, u)− `(s, r) >h = exp −
0≤s<∞ 4(u ∧ r)|u − r|h−1
u x∧r (r − u)2 + 4 (u ∧ r)|u − r| + r − u  (x ∧ r)|u|
h p i
0<α × − −u∧x
x u∧r 2(u ∧ r)|u − r| u∧r

n √ √ √

√ o
1.31.5 Ex/√2λ exp −γ 2λ sup `(s, u/ 2λ) − α−1 `(s, r/ 2λ) ; Rτ(3) ∈ dz/ 2λ
0≤s≤τ
0<α
z −|z−x|
− e−z−x dz


= e
2x
z  − e−2r + e−2u e−|x−u| − e−x−u



−|z−u| −z−u

n
− e −e
2x 2 1 − e−2(u∧r) 1 − e−2|u−r| +  − e−2r + e−2u

2 1 − e−2(u∧r) 1 − e−2|u−r| − ( − e−2r + e−2u )(1 − e−2u ) e−|x−r| − e−x−r o

− dz
2 1 − e−2(u∧r) 1 − e−2|u−r| +  − e−2r + e−2u e−|u−r| − e−u−r

 √ √
√ √ 
1.31.6 Px/√2λ sup `(s, u/ 2λ) − α−1 `(s, r/ 2λ) > h/ 2λ, Rτ(3) ∈ dz/ 2λ
0≤s≤τ
0<α
z −|z−u|  − e−2r + e−2u h h −|x−r| −x−r
e −e


−z−u


= e −e exp − 2( ) 2
2x 2 1−e− u ∧ r 1−e− |u−r | e−|u−r| − e−u−r

 − e−2r + e−2u )(1 − e−2u )

 −|x−r| −x−r
(e −e
i
− 2(u r ) 2 u r u r u r − e−|u−x| + e−x−u dz
2 1−e− ∧ 1−e− | − | e−| − | −e



− −

q
2


Υ := 1 − e−2r − α + αe−2u + 4α 1 − e−2(u∧r) 1 − e−2|u−r| + 1 − α
 2. STOPPING AT FIRST HITTING TIME 469
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 Hz = min{s : Rs(3) = z}

2. Stopping at first hitting time

 z √2
 sh(z 2 ) , x=0<z

 √
o  √
z sh(x 2 )
n 
2.0.1 Ex e−αHz ; Hz < ∞ = √ , 0<x≤z
 sh(z 2 )
 x
 z −(x−z) 2α
√


e , z≤x
x

z s (z), x=0<z

 t

 z sst (x, z)dt,

0<x≤z
2.0.2 Px Hz ∈ dt = x

 2
 z√(x − z ) exp − (x − z ) dt, z ≤ x
 
x 2t3=2 2t
(
0, 0≤x≤z
z


(1) Px H z = ∞ =
1− , z≤x
x

1, 0≤x≤z
(
z (y − x )


2.1.2 Px sup Rs(3) < y, Hz < ∞ =
, z≤x≤y
z≤y 0≤s≤Hz
x(y − z )

 z sh(x√2 )
 x sh(z √2 ) ,
 0≤x≤z
Ex e−αHz ; sup Rs(3)

2.1.4 < y} = √
z≤y 0≤s≤Hz  z sh((y − x)√2 ) , z ≤ x ≤ y

x sh((y − z ) 2 )


z st (z), x=0<z
 z


(1) Px sup Rs(3) < y, Hz ∈ dt) = ss (x, z)dt, 0<x≤z
x t
z≤y 0≤s≤Hz
 z ss (y − x, y − z)dt, z ≤ x ≤ y


x t
 z (x − y )

 x(z − y) , y ≤ x ≤ z
2.2.2 Px inf Rs(3) > y, Hz < ∞ =
0≤s≤Hz  z, y≤z≤x
0≤y x

( y
1− , z≤y≤x
(1) Px inf Rs(3)


> y, Hz = ∞ = x
z
0≤s≤Hz 1− , y≤z≤x
x
470 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 Hz = min{s : Rs(3) = z}

 z sh((x − y)√2 )
√ , y≤x≤z
x sh((z − y) 2 )

Ex e−αHz ;

2.2.4 inf Rs(3) >y =
0≤s≤Hz  z −(x−z) 2α√
0≤y e , y≤z≤x
x

z
 x sst (x − y, z − y)dt,

y≤x≤z
(1) Px inf
0≤s≤Hz
Rs(3) > y, Hz ∈ dt = z (x − z ) 
(x − z )2 
 √ exp − dt, y ≤ z ≤ x
x 2t3=2 2t

Ex e−γ`(Hz ,r) ; Hz < ∞


2.3.1


 1, 0≤x≤z
 z (1 + 2 (r − x))


, z≤x≤r
z≤r = x(1 + 2 (r − z ))

 z
, r≤x

x(1 + 2 (r − z ))


z
0≤x≤r

,

 z + 2 ( z − r)r
z (x + 2 (x − r)r)


r≤z = , r≤x≤z
 x(z + 2 (z − r)r)
 z


, z≤x
x
 r r−z
1− + , r≤x

 x x(1 + 2 (r − z ))
x−z
 −γ`(Hz ,r) 
(1) Ex e ; Hz = ∞ = , z≤x≤r
z≤r


 x (1 + 2 (r − z ))
0, 0≤x≤z

2.3.2 Px `(Hz , r) ∈ dy, Hz < ∞

0≤x≤z

 0,
 z (x − z ) y

  
exp − dy, z ≤ x ≤ r
z≤r = 2x(r − z )2 2(r − z )
z y

  
exp − dy, r≤x

2x(r − z ) 2(r − z )


z zy
  
exp − dy, 0≤x≤r
 2r(z − r) 2r(z − r)



r≤z z (z − x) zy
 
=

 2x ( z − r )2 exp − 2r(z − r) dy, r ≤ x ≤ z

z≤x

0,
 2. STOPPING AT FIRST HITTING TIME 471
P3
1/2
5 Rs(3) = (k)
k=1 (Ws )2 Hz = min{s : Rs(3) = z}


 0, 0≤x≤z
x z y

−
  
exp − dy, z ≤ x ≤ r

2x(r − z ) 2(r − z )



(1) Px ` Hz , r ∈ dy, Hz = ∞ =
z≤r  1 exp − y

  
dy, r≤x

2x 2(r − z )
 z
, r≤z≤x


 x




 1, 0 ≤ x ≤ z ≤ r,
(2) Px ` Hz , r = 0, Hz < ∞ = z | r − x|
, z∧r ≤x≤r∨z
 x|r − z |




0, z ≤ r ≤ x, 0 ≤ x ≤ r ≤ z

r
(3) Px ` Hz , r = 0, Hz = ∞ = 1 − , r≤x
x
z≤r

Ex e−αHz −γ`(Hz ,r) ; Hz < ∞


2.3.3

 z sh(x√2 )
√ , 0≤x≤z
x sh(z 2 )




 z √2 e(r−x)√2 + 2 sh((r − x)√2 )



z≤r = √ √ √
, z≤x≤r
 x 2 e(r−z) 2 + 2 sh((r − z ) 2 )
√ √

z 2 e−(x−r) 2




, r≤x

 √ √ √
x 2 e(r−z) 2 + 2 sh((r − z ) 2 )

√ √
z 2 sh(x 2 )

√ √ √ √
, 0≤x≤r
x 2 sh(z 2 ) + 2 sh((z − r) 2 ) sh(r 2 )




 √
 √ √ √
z 2 sh(x 2 ) + 2 sh((x − r) 2 ) sh(r 2 )


r≤z = √ √ √ √
, r≤x≤z

 x 2 sh(z 2 ) + 2 sh((z − r) 2 ) sh(r 2 )
 z e−(x−z)√2α ,



z≤x
x

Ex e−αHz ; `(Hz , r) ∈ dy, Hz < ∞


2.3.4

0, 0≤x≤z
 √ √

 z 2 sh((x − z ) 2 )

√
y 2
 √ dy, z≤x≤r

z≤r = exp − √ 2x sh2 ((r − z ) 2 )
1 − e−2(r−z) 2  √ √
−(x−r ) 2
 z 2 e


√ dy, r≤x

2x sh((r − z ) 2 )
472 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 Hz = min{s : Rs(3) = z}

0, z≤x
√ √

 z 2 sh((z − x) 2 )

√ √

y 2 sh(z 2 )

√ dy, r≤x≤z

r≤z = exp − √ √ 2x sh2 ((z − r) 2 )
2 sh((z − r) 2 ) sh(r 2 )  √ √

 z 2 sh(x 2 )
 √ √ dy, 0 ≤ x ≤ r
2x sh((z − r) 2 ) sh(r 2 )


Ex e−αHz ; ` Hz , r = 0


(1)


0, 0≤x≤r
 z sh((x − r)√2 )



r≤z = √ , r≤x≤z
x sh((z − r) 2 )
 z
 √
 e−(x−z) 2α ,

z≤x
x
 z sh(x√2 )
√ , 0≤x≤z
 x sh(z 2 )√



z≤r = z sh((r − x) 2 )
√ , z≤x≤r
 sh((r − z ) 2 )
x



0, r≤x

n  Z Hz
 o
2.4.1 Ex exp −γ 1I[r,∞) Rs(3) ds ; Hz < ∞
0

1, 0≤x≤z

√
 z (1 + (r − x) 2 )



√ , z≤x≤r
z≤r = x(1 + (r − z ) 2 )
√
 ze−(x−r) √ 2



, r≤x
x(1 + (r − z ) 2 )
√
 z 2
√ √ √ , 0≤x≤r
r 2 ch((z − r) 2 ) + sh((z − r) 2 )



√ √ √

z r 2 ch((x − r) 2 ) + sh((x − r) 2 )


r≤z = √ √ √
, r≤x≤z
 x r 2 ch((z − r) 2 ) + sh((z − r) 2 )
 z −(x−z)√2γ



e , z≤x
x

n  Z Hz
 o
(1) Ex exp −γ 1I[r,∞) Rs(3) ds ; Hz = ∞ = 0
0

Z Hz

2.4.2 Px 1I[r,∞) Rs(3) ds ∈ dy, Hz < ∞
0
 2. STOPPING AT FIRST HITTING TIME 473
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 Hz = min{s : Rs(3) = z}


0, 0≤x≤z
 √y 

z (x − z) 1 1

  2
y/2(r−z)

 x(r − z )2 √2y − 2(r − z ) e Erfc dy, z≤x≤r
 √
(r − z ) 2


z<r = z 2
√ e−(x−r) /2y dy−


 x ( r − z ) 2y
√
z x−r y x−r y 

   

−

exp + Erfc √ + √ dy, r ≤ x
2x(r − z ) 2 r−z 2(r − z )2 2y (r − z ) 2


 e y (1, z − r, 0, r)dy,
z rc 0≤x≤r
z

r<z = x (ssy (x − r, b − r) + r ssy (z − x, z − r) ∗ rc
e y (1, z − r, 0, r))dy, r ≤ x ≤ z

 z
 h (1, x − z)dy, z≤x
x y

1, x≤z

z (r − x)
Z Hz
 

(3)
(1) Px 1I[r,∞) Rs ds = 0 = , z≤x≤r
z≤r 0  x(r − z )

0, r≤x

Z Hz

2.4.4 Px 1I[r,∞) Rs(3) ds ∈ dv, Hz ∈ dt
0
Z ∞
= Bz(27) (t − v, v, y)dy dvdt, for Bz(27) (u, v, y) see 2.27.2
0

Z Hz

(1) Px 1I[r,∞) Rs(3) ds = 0, Hz ∈ dt
0

( z
ss (x, z)dt, 0≤x≤z
= x t
z
ss (r − x, r − z)dt, z ≤ x ≤ r
x t

n  Z Hz
 o
2.5.1 Ex exp −γ 1I[0,r] Rs(3) ds ; Hz < ∞
0

 z sh(x√2 )
√ , 0≤x≤z
x sh(z 2 )



√

z ch((r − x) 2 )

z≤r = √ , z≤x≤r
 x ch((r − z ) 2
 )

 z
√ , r≤x
x ch((r − z ) 2 )

474 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 Hz = min{s : Rs(3) = z}
√
 z sh(x 2 )
√ √ √
, 0≤x≤r
x sh(r 2 ) + 2 (z − r) ch(r 2 )



√ √ √

z sh(r 2 ) + 2 (x − r) ch(r 2 )


r≤z = √ √ √
, r≤x≤z

 x sh(r 2 ) + 2 (z − r) ch(r 2 )
 z


, z≤x
x

  Z Hz

(1) Ex exp −γ 1I[0,r] Rs(3) ds ; Hz = ∞
z≤r 0

0, 0≤x≤z

√
sh((x − z ) 2 )



z≤x≤r

√ √ ,
= x 2 ch((r − z ) 2 )
√
 1 − r + √sh((r − z ) 2√) ,


r≤x

x x 2 ch((r − z ) 2 )

Z Hz

2.5.2 Px 1I[0,r] Rs(3) ds ∈ dy, Hz < ∞
0
 z
 x rcy (0, r, x, z − r)dy, 0≤x≤r
r<z =
 z (z − x) rcy (0, r, r, z − r)dy, r ≤ x ≤ z
x(z − r)
 −1
 zx ssy (x, z)dy,
 0≤x<z
z<r = zx−1 ccy (r − x, r − z)dy, z < x ≤ r

 −1
zx ccy (0, r − z)dy, r≤x

0≤x≤z

 0,
Z Hz 1
 
Rs(3) ds ∈ dy, Hz = ∞ = x scy (x − z, r − z)dy, z ≤ x ≤ r


(1) Px 1I[0,r]
z≤r 0 1


sc (r − z, r − z)dy, r ≤ x
x y

 0, x≤r
Z Hz
  z (x − r)

(2) Px 1I[0,r] (3)
Rs ds = 0 = , r≤x≤z
0
x(z − r)
r≤z


 −1
zx , z≤x

Z Hz

2.5.4 Px 1I[0,r] Rs(3) ds ∈ dv, Hz ∈ dt
0
Z ∞
= Bz(27) (v, t − v, y)dy dvdt, for Bz(27) (u, v, y) see 2.27.2
0
 2. STOPPING AT FIRST HITTING TIME 475
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 Hz = min{s : Rs(3) = z}

Z Hz

(1) Px 1I[0,r] Rs(3) ds = 0, Hz ∈ dt
0

( z
ss (x − r, z − r)dt, r ≤ x ≤ z
= x t
z
h (1, x − z)dt, z≤x
x t

n  Z Hz


 o
2.6.1 Ex exp − p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds ; Hz < ∞
0
 z sh(x√2p)
√ , 0≤x≤z
x sh(z 2p)



 z √p ch((r − x)√2p) + √q sh((r − x)√2p)



z≤r = √ √ √ √
, z ≤ x ≤ r
 x p ch((r − z ) 2p) + q sh((r − z ) 2p)
√
z pe−(x−r) 2q
√



√
, r ≤ x

 √ √ √
x p ch((r − z ) 2p) + q sh((r − z ) 2p)
√ √
 z q sh(x 2p)
√ √ √ √ √ √
, 0 ≤ x ≤ r
 x √ q sh(r√ 2p) ch((z − r)√2q) + √p ch(r√2p) sh((z − r) √2q)





r≤z = z √q sh(r √2p) ch((x − r)√2q) + √p ch(r √2p) sh((x − r)√2q)
, r ≤ x ≤ z

 x q sh(r 2p) ch((z − r) 2q) + p ch(r 2p) sh((z − r) 2q)
 −1 −(x−z)√2q


zx e , z≤x

Z Hz

Z Hz

2.6.2 Px 1I[0,r) Rs(3) ds ∈ du, 1I[r,∞) Rs(3) ds ∈ dv
0 0
Z ∞
= Bz(27) (u, v, y)dy dudv, for Bz(27) (u, v, y) see 2.27.2
0

Z Hz

Z Hz

(1) Px 1I[0,r) Rs(3) ds ∈ du, 1I[r,∞) Rs(3) ds = 0
0 0

( z
ss (x, z)du, 0≤x≤z
= x u
z
ss (r − x, r − z)du, z ≤ x ≤ r
x u

Z Hz

Z Hz

(2) Px 1I[0,r) Rs(3) ds = 0, 1I[r,∞) Rs(3) ds ∈ dv
0 0

( z
ss (x − r, z − r)dv, r≤x≤z
= x v
z
h (1, x − z)dv, z≤x
x v
476 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 Hz = min{s : Rs(3) = z}

n  Z Hz
 o
2.7.1 Ex exp −γ 1I[r,u] Rs(3) ds , Hz < ∞
r<u 0


1, 0≤x≤z
√ √ √

z ((r − x) 2 sh((u − r) 2 ) + ch((u − r) 2 ))



 x((r − z )√2 sh((u − r)√2 ) + ch((u − r)√2 , z≤x≤r

))


z≤r √
= z ch((u − x) 2 )
√ √ √ , r≤x≤u
x((r − z ) 2 sh((u − r) 2 ) + ch((u − r) 2 ))



z



√ √ √ , u≤x

x((r − z ) 2 sh((u − r) 2 ) + ch((u − r) 2 ))


√
 z 2
√ √ √ , 0≤x≤r
(1 + (z − u)r2 ) sh((u − r) 2 ) + (r + z − u) 2 ch((u − r) 2 )



√ √ √

z (sh((x − r) 2 ) + r 2 ch((x − r) 2 ))



√ √ √ , r≤x≤u

x ((1 + ( z u) r2 ) sh((u − r) 2 ) + (r + z − u) 2 ch((u − r) 2 ))

u≤z = −
√ √ √
 z ((1 + (x − u)r2 ) sh((u − r)√2 ) + (r + x − u)√2 ch((u − r)√2 ))

, u≤x≤z
x((1 + (z − u)r2 ) sh((u − r) 2 ) + (r + z − u) 2 ch((u − r) 2 ))



 z



, z≤x
x

 √
2 2
 √
 z + xI 13  3 ( + x)3=2


√ , 0≤x≤z
 x√ + zI 1 23 2 (

+ z )3=2
Z Hz 
  
3
2.8.3 Ex exp −αHz − γ Rs(3) ds =  √
√ 2 2 + x)3=2
 z + xK 13  3 (

0 

 x√ + zK 1 2√2 ( , z≤x
+ z )3=2


3 3

√
zI1=4 (x2 )
 √
√ , √ 0≤x≤z
xI1=4 (z 2 )

n  Hz Z  o 

(3) 2
2.9.1 Ex exp −2γ (Rs ) ds ; Hz < ∞ = √
zK (x2 )
√
0  √ 1=4 2 √ , z≤x

xK1=4 (z )


 Z Hz 
2.9.3 Ex exp −αHz − 2γ (Rs(3) )2 ds
0
√ √
z D− 1 − α (−x 2 ) − D− 1 − α (x 2 )


 2 γ 2 γ
 x D− 1 − α (−z 2 ) − D− 1 − α (z 2 ) , 0 ≤ x ≤ z


 √ √

2 γ 2 γ
= √
 zD− 1 − α (x 2 )
2 γ
z≤x

 xD 1 α (z √2 ) ,


− −2 γ
 2. STOPPING AT FIRST HITTING TIME 477
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 Hz = min{s : Rs(3) = z}

√ √
zS 2 =| | | 2| e x ; | 2|
 √ 

, 0 ≤ x ≤ z

√ √
xS√2 =| | | 2| e z ; | 2|






√
2

 zK√ x
2 =| | | | e
 
 Z Hz  
(3)
2βRs
2.10.3 Ex exp −αHz − γ e ds = √  , z ≤ x, β > 0
0 

 xK√2 =| | | 2| e z
√
zI√2 =| | | 2| e x

 



 √  , z ≤ x, β < 0
xI 2 =| | | 2| e z

 √

∞

8z X exp(−zj02;k =2h)
2.11.2 Px sup `(Hz , y) < h, Hz < ∞ =
y∈[z,∞) x j03;k J1 (j0;k )
z<x k=1

√
z sh((x − z ) 2 )
Ex e−γ Ȟ(Hz ) ; sup Rs(3) ∈ dy =

2.13.1 √ dy, x<y
0≤s≤Hz x(y − z ) sh((y − z ) 2 )
z<x

z
2.13.2 Px Ȟ(Hz ) ∈ du, sup Rs(3) ∈ dy = ss (x − z, y − z)dudy
0≤s≤Hz x (y − z ) u
z<x
x<y

√ √
z 2 sh((x − z ) 2 + 2 )
Ex e−αHz−γ Ȟ(Hz ) ; sup Rs(3) ∈ dy =

2.13.4 √ √ dy
0≤s≤Hz x sh((y − z ) 2 ) sh((y − z ) 2 + 2 )
z<x
x<y

(1) Px Ȟ(Hz ) ∈ du, Hz ∈ dt, sup Rs(3) ∈ dy
0≤s≤Hz
u<t
z
= ss (x − z, y − z) st−u (y − z)dudtdy
x u
√
z sh((z − x) 2 )
Ex e−γ Ĥ(Hz ) ;

2.14.1 inf Rs(3) ∈ dy = √ dy, y<x
0≤s≤Hz x(z − y) sh((z − y) 2 )
x<z

z
2.14.2 Px Ĥ(Hz ) ∈ du, inf Rs(3) ∈ dy = ss (z − x, z − y)dudy
0≤s≤Hz x (z − y ) u
y<x
x<z
√ √
 −αHz−γ Ĥ(Hz ) (3) z 2 sh((z − x) 2 + 2 )
2.14.4 Ex e ; inf Rs ∈ dy = √ √ dy
0≤s≤Hz x sh((z − y) 2 ) sh((z − y) 2 + 2 )
y<x
x<z
478 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 Hz = min{s : Rs(3) = z}

(1) Px Ĥ(Hz ) ∈ du, Hz ∈ dt, inf Rs(3) ∈ dy
0≤s≤Hz
u<t
z
= ss (z − x, z − y) st−u (z − y)dudtdy
x u

 z (x − a )


 x(z − a) , a≤x≤z
2.15.2 Px a < inf Rs(3) , sup Rs(3) <b =
0≤s≤Hz 0≤s≤Hz  z (b − x) ,

z≤x≤b
x(b − z )

Ex e−αHz , a <

2.15.4 inf Rs(3) , sup Rs(3) < b
0≤s≤Hz 0≤s≤Hz
 z sh((x − a)√2 )
 x sh((z − a)√2 , −b ≤ x ≤ z
)

= √
 z sh((b − x) 2 )
√ , z≤x≤b
x sh((b − z ) 2 )


(1) Px Hz ∈ dt, a < inf Rs(3) , sup Rs(3) < b
0≤s≤t 0≤s≤t

( z
ss (x − a, z − a)dt, a ≤ x ≤ z
= x t
z
ss (b − x, b − z)dt, z ≤ x ≤ b
x t



2.18.1 Ex exp −γ`(Hz , r) − η`(Hz , u) ; Hz < ∞
r<u
1, 0≤x≤z

z (1 + 2 (r x ) + 2  (u x ) + 4  ( u r )(r x ))

− − − −


 x(1 + 2 (r − z ) + 2(u − z ) + 4 (u − r)(r − z )) , z ≤ x ≤ r



z<r = z (1 + 2(u − x))
, r≤x≤u
x (1 + 2 ( r − z ) + 2 (u − z ) + 4  (u − r)(r − z ))



z



 , u≤x
x(1 + 2 (r − z ) + 2(u − z ) + 4 (u − r)(r − z ))
z
0≤x≤r

,

 z + 2 (z − r)r + 2(z − u)u + 4 (z − u)(u − r)r
z (x + 2 (x − r)r)



, r≤x≤u

x(z + 2 (z − r)r + 2(z − u)u + 4 (z − u)(u − r)r)


u<z =
z (x + 2 (x − r)r + 2(x − u)u + 4 (x − u)(u − r)r)
, u≤x≤z

x(z + 2 (z − r)r + 2(z − u)u + 4 (z − u)(u − r)r)




 z


, z≤x
x
 2. STOPPING AT FIRST HITTING TIME 479
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 Hz = min{s : Rs(3) = z}



(1) Ex exp −γ`(Hz , r) − η`(Hz , u) ; Hz = ∞
r<u

 0, 0≤x≤z
x−z


, z≤x≤r

 x(1 + 2 (r − z ) + 2(u − z ) + 4 (u − r)(r − z ))



z<r = x − z + 2 (x − r)(r − z )
, r≤x≤u
x (1 + 2 (r − z ) + 2(u − z ) + 4 (u − r)(r − z ))



 1 − z + 2 r(r − z ) + 2u(u − z ) + 4 u(u − r)(r − z ) , u ≤ x




x(1 + 2 (r − z ) + 2(u − z ) + 4 (u − r)(r − z ))

2.18.2 Px `(Hz , r) ∈ dy, `(Hz , u) ∈ dg, Hz < ∞
r<u
z (18) (18)
z<r = Bx,z (y, g)dydg, for Bx,z (y, g) see 1.2.18.2
x

0≤x≤r

qr dydg,
 r(u − x) u(x − r)



q dydg + q dydg, r ≤ x ≤ u
u<z = x(u − r) r x(u − r) u
 u(z − x) qu dydg,


u≤x≤z

x(z − u)

where
√
z

uy (z − r)g   yg 
qr := exp − − I0
4r(z − u)(u − r) 2r(u − r) 2(z − u)(u − r) u−r
√  √yg 
z g 
uy (z − r)g 
qu := √ exp − − I1
4u(z − u)(u − r) y 2r(u − r) 2(z − u)(u − r) u−r

(1) Px `(Hz , r) ∈ dy, `(Hz , u) = 0, Hz < ∞

z (x − z ) y(u − z )
  
 2x(r − z )2 exp − 2(u − r)(r − z ) dy,
 z≤x≤r
=
z (u − x) y (u − z )
 
exp − dy, r ≤ x ≤ u

2x(u − r)(r − z ) 2(u − r)(r − z )


(2) Px `(Hz , r) = 0, `(Hz , u) ∈ dg, Hz < ∞

z (x − r ) g(z − r)
  
 2x(z − u)(u − r) exp − 2(z − u)(u − r) dg,
 r≤x≤u
=
 z (z − x) exp − g(z − r)
 
dg, u≤x≤z

2x(z − u) 2 2(z − u)(u − r)
480 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 Hz = min{s : Rs(3) = z}

 z (r − x)


 x(r − z ) , z ≤ x ≤ r
(3) Px `(Hz , r) = 0, `(Hz , u) = 0, Hz < ∞ =
 z (x − u) , u ≤ x ≤ z

x(z − u)

(4) Px `(Hz , r) ∈ dy, `(Hz , u) ∈ dg, Hz = ∞
 r(x − z )
p∞
r dydg, z≤x≤r
 x(r − z )




z<r = r(u − x) ∞ u(x − r) ∞
p dydg + p dydg, r ≤ x ≤ u
 x (u − r) r x(u − r) u
 u p∞ dydg,



u≤x
x u
where
  √yg 
1 
y+g y
p∞
r := exp − − I
4r(u − r) 2(u − r) 2(r − z ) 0 u − r
√   √yg 
g y+g y

p∞ := √ exp − − I
u 4u(u − r) y 2(u − r) 2(r − z ) 1 u − r

(5) Px `(Hz , r) = 0, `(Hz , u) ∈ dg, Hz = ∞

x−r g
  
 2x(u − r) exp − 2(u − r) dg, r ≤ x ≤ u

=
 1 exp − g
 
 dg, u≤x
2x 2(u − r)

u
(6) Px `(Hz , r) = 0, `(Hz , u) = 0, Hz = ∞ = 1 − , u≤x
x

 −1=2 √
x I1 (2 2 x)
Hz  z −1=2 I (2√2 z ) , 0 < x ≤ z
ds
Z o 
1
n  
2.19.1 Ex exp −γ ; Hz < ∞ = √
−1=2 K (2 2 x)
0 Rs(3)  x 1
, z≤x

√
z −1=2 K1 (2 2 z )

√
zM− =√2 ;1=2 (2x 2 )

√ , 0<x≤z

;1=2 (2z 2 )
 xM √

Hz
ds
 
− = 2
 Z
2.19.3 Ex exp −αHz − γ = √
Rs(3) zW ;1=2 (2x 2 )
√
0 − = 2

√ , z≤x

xW− =√2 ;1=2 (2z 2 )


 2. STOPPING AT FIRST HITTING TIME 481
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 Hz = min{s : Rs(3) = z}

  √
x −1/2+ γ 2 +1/4
, 0<x≤z

2 Hz
ds z
n  Z  
o 
2.20.1 Ex exp − (3)
2
; Hz < ∞ =   √ 2
2 Rs
 x
0  −1/2− γ +1/4
, z≤x

z

√ √ √
zI 2 +1=4 x 2


 √ √ √
, 0<x≤z
xI +1=4 z 2

2 Hz
ds
Z 
 2
 
2.20.3 Ex exp −αHz − =
Rs(3) 2
√ √
2 0


 zK√ 2 +1=4 x 2


z≤x

 √ √ √
,
xK 2 +1=4 z 2


 1=2
z ln(z=x) 2
 1=2 √ 3=2 e−y/8 e− ln (z/x)/2y dy, 0 < x ≤ z
Hz
ds x 2y
Z  
2.20.4 Px
∈ dy =
0 Rs(3) 2 1=2
 z √ln(x=z ) e−y/8 e− ln2 (x/z)/2y dy, z ≤ x
0<y

x1=2 2y3=2

 Z Hz 
p2 q2 Rs(3) 2  
n
o
2.21.1 Ex exp − 2 + ds ; H < ∞
0 2 Rs(3)


2 z

 x−1=2 I√ 2
4p2 +1=4 (qx =2)
 z −1=2 I√ 2

2 , 0<x≤z
4p +1=4 (qz =2)


=
x−1=2 K√ 2 (qx2 =2)
 −1=2 √4p +1=4 2 ,

z≤x

K 4p2 +1=4 (qz =2)

z

Hz 
p2 q2 Rs(3) 2  
 Z

2.21.3 Ex exp −αHz − + ds
2 Rs(3) 2 2


0

 x−3=2 M √ (qx2 )
− =2q; 4p2 +1=4
 z −3=2 M , 0<x≤z
(qz 2 )

√


− =2q; 4p2 +1=4
=
 x−3=2 W− =2q;√4p2 +1=4 (qx2 )
, z≤x

 −3=2
W− =2q;√4p2 +1=4 (qx2 )

z

p2
Hz
q

n  Z  o
2.22.1 Ex exp − 2 + ds ; H z < ∞
2 Rs(3) Rs
(3)


0
482 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 Hz = min{s : Rs(3) = z}

 x−1=2 I√ √
4p2 +1 (2 2qx)

 z −1=2 I√ 2 (2√2qz ) , 0 < x ≤ z




4p +1
= √
 x−1=2 K√4p2 +1 (2 2qx)
, z≤x

 −1=2 √ √
K 4p2 +1 (2 2qz )

z

p2
Hz
q

 Z 
2.22.3 Ex exp −αHz − (3)
2
+ (3) ds
0 2 Rs Rs
 x−1 M √ √ √
−q= 2 ; p2 +1=4
(2x 2 )
 z −1 M √ √ 2
 √ , 0<x≤z
(2z 2 )

−q= 2 ; p +1=4

= √
 x−1 W−q=√2 ;√p2 +1=4 (2x 2 )
, z≤x

 −1
 √
z W−q=√2 ;√p2 +1=4 (2z 2 )

n  Z Hz


 1
o
2.27.1 Ex exp − p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds ; ` Hz , r ∈ dy
0
2

 p p p 
z≤r = exp −y 2p cth((r − z) 2p) − y 2q
 √ √
z 2p sh((x − z ) 2p)
2 √ dy, z≤x≤r
x sh ((r − z ) 2p)


× √
z 2p √
√ e−(x−r) 2q dy, r ≤ x

x sh((r − z ) 2p)


p p p p

r≤z = exp −y( 2q cth((z − r) 2q) + 2p cth(r 2p))
√ √
z 2q sh(x 2p)

 x sh((z − r)√2q) sh(r√2p) dy, 0 ≤ x ≤ r

× √ √
 z 2q sh((z − x√) 2q) dy, r≤x≤z
2

x sh ((z − r) 2q)

n  Z Hz



o
(1) Ex exp − p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds ; ` Hz , r = 0
0

 z sh(x√2p)
√ , 0≤x≤z
x sh(z 2p)



√

 z sh((r − x)√2p) , z ≤ x ≤ r




= x sh((r − z ) 2p)
√
 z sh((x − r) 2q)
√ , r≤x≤z

x sh((z − r) 2q)



 z −(x−z) 2q


 √
e , z≤x
x
 2. STOPPING AT FIRST HITTING TIME 483
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 Hz = min{s : Rs(3) = z}

Z Hz Hz
1
Z 




2.27.2 Px 1I[0,r) Rs(3) ds ∈ du, 1I[r,∞) Rs(3) ds ∈ dv, ` Hz , r ∈ dy
0 0
2

=: Bz(27) (u, v, y)dudvdy

z
z≤r = escsu (x − z, 0, r − z, y) hv (1, y)dudvdy
x
x≤r
z
z≤r = es (1, 1, r − z, 0, y) hv (1, y + x − r)dudvdy
x u
r≤x
z
r≤z = essu (x, r, 0, y) esv (1, 1, z − r, 0, y)dudvdy
x
x≤r
z
r≤z = es (0, 0, r, 0, y) escsv (z − x, 0, z − r, y)dudvdy
x u
r≤x

Z Hz

Z Hz


(3)
(1) Px 1I[0,r) Rs ds ∈ du, 1I[r,∞) Rs(3) ds = 0, ` Hz , r = 0
0 0
( z
ss (x, z)du, 0≤x≤z
= x u
z
ss (r − x, r − z)du, z ≤ x ≤ r
x u

Z Hz

Z Hz


(2) Px 1I[0,r) Rs(3) ds = 0, 1I[r,∞) Rs(3) ds ∈ dv, ` Hz , r = 0
0 0
( z
ss (x − r, z − r)dv, r≤x≤z
= x v
z
h (1, x − z)dv, z≤x
x v

p2
n  Z Hz    o
(3) (3)
2.30.1 Ex exp − e2βRs + qeβRs ds ; Hz < ∞
0
2
 1 2p x ; 2p
 p q
− e x S

z exp + ; 1; e

2 p

 | | | | | | | |
0≤x≤z

 1 2 ,
z ; 2p
 p q p


z
x exp − e S + ; 1; e


2 p| | | |

| | | |


2p x


 ze− x=2 M−q=p| |;0

e
  

| |
=  2p , z ≤ x, β<0

 xe− z=2 M−q=p| |;0 e z

 | |
 2p


ze− x=2 W−q=p| |;0 e x

 

| |


, z ≤ x, β>0
 xe− z=2 W−q=p| |;0 2p e z



 
| |
484 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 Hz = min{s : Rs(3) = z}

n 
 o
2.31.1 Ex exp − sup `(s, u) − α−1 `(s, r) ; Hz < ∞ z ≤x∧r∧u
2 0≤s≤Hz
(1)
z z (1 + r − z − u + z )(u ∧ x − z )
0<α = −
x x(2 (u ∧ r − z )|u − r| + 1 + r − z − u + z )

z 2(u ∧ r − z )|u − r| − (1 + r − z − u + z )(u − z ))(x ∧ r − z )

−
x(2 (u ∧ r − z )|u − r| + 1 + r − z − u + z )(u ∧ r − z )

n 
 o
2.31.1 Ex exp − sup `(s, u) − α−1 `(s, r) ; Hz = ∞ z ≤x∧r∧u
2 0≤s≤Hz
(2)
z (u − z )(1 + r − z − u + z )(u ∧ x − z )
0<α =1− −
x x(2 (u ∧ r − z )|u − r| + 1 + r − z − u + z )

(u − z ) 2(u ∧ r − z )|u − r| − (1 + r − z − u + z )(u − z ))(x ∧ r − z )

−
x(2 (u ∧ r − z )|u − r| + 1 + r − z − u + z )(u ∧ r − z )



2.31.1 Exz exp − sup `(s, uz) − α−1 `(s, rz) x∨r∨u≤z
2z 0≤s≤Hz
(3)
u(2 + r − r2 − u + u2 )(u ∧ x − ux)
0<α =1−
x(2 (u ∧ r − ur)|u − r| + 2 + r − r2 − u + u2 )

u 2(u ∧ r − ur)|u − r| − (2 + r − r2 − u + u2 )(u − u2 ))(x ∧ r − xr)

−
x(2 (u ∧ r − ur)|u − r| + 2 + r − r2 − u + u2 )(u ∧ r − ur)

 
`(s, u) − α−1 `(s, r) > h, Hz < ∞


2.31.2 Px sup z ≤x∧r∧u
0≤s≤Hz
(1)
z ( + r − z − u + z )h h x ∧ r − z
exp − 1

0<α =
x 4(u ∧ r − z )|u − r| u ∧ r−z

 + r − z − u + z (x ∧ r − z )|u − z |
− 1
 i
−u∧x+z
2(u ∧ r − z )|u − r| u ∧ r−z

p
Υ1 := (r − z − αu + αz)2 + 4α(u ∧ r − z)|u − r|

p
Υ2 := (r − r2 − αu + αu2 )2 + 4α(u ∧ r − ur)|u − r|
 2. STOPPING AT FIRST HITTING TIME 485
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 Hz = min{s : Rs(3) = z}

 
`(s, u) − α−1 `(s, r) > h, Hz = ∞


2.31.2 Px sup z ≤x∧r∧u
0≤s≤Hz
(2)
(u − z ) ( + r − z − u + z )h h x ∧ r − z
exp − 1

0<α =
x 4(u ∧ r − z )|u − r| u ∧ r−z

 + r − z − u + z (x ∧ r − z )|u − z |
− 1
 i
−u∧x+z
2(u ∧ r − z )|u − r| u ∧ r−z

 
`(s, uz) − α−1 `(s, rz) > hz


2.31.2 Pxz sup x∨r∨u≤z
0≤s≤Hz
(3)
u ( + r − r2 − u + u2 )h h x ∧ r − xr
exp − 2

0<α =
x 4(u ∧ r − ur)|u − r| u ∧ r − ur

 + r − r2 − u + u2 (x ∧ r − xr)(u − u2 )
− 2
 i
− u ∧ x + ux
2(u ∧ r − ur)|u − r| u ∧ r − ur

p
Υ1 := (r − z − αu + αz)2 + 4α(u ∧ r − z)|u − r|

p
Υ2 := (r − r2 − αu + αu2 )2 + 4α(u ∧ r − ur)|u − r|
486 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 H = Ha,b = min{s : Rs(3) ∈
/ (a, b)}

3. Stopping at first exit time

√ √
a sh (b − x) 2 + b sh (x − a) 2



3.0.1 Ex e−αH = √
, 0≤a≤x≤b
x sh (b − a) 2

a b
3.0.2 Px (H ∈ dt) = ss (b − x, b − a)dt + sst (x − a, b − a)dt
x t x

(3)a(b − x) −βa b(x − a) −βb

3.0.3 Ex e−βRH = e + e
x(b − a) x(b − a)
(3)

a(b − x)
3.0.4 Px R H =a =
x(b − a)
(a)
(3)

b(x − a)
3.0.4 Px R H =b =
x(b − a)
(b)

√
a sh (b − x) 2


Ex e−αH ; RH
 (3)
3.0.5 =a = √
x sh (b − a) 2


(a)

√
b sh (x − a) 2


−αH
 (3)
3.0.5 Ex e ; RH =b = √

x sh (b − a) 2
(b)
(3)

a
3.0.6 Px H ∈ dt, RH = a = sst (b − x, b − a)dt
x
(a)
(3)

b
3.0.6 Px H ∈ dt, RH = b = sst (x − a, b − a)dt
x
(b)

(3)

a(b − y)(x − a)
3.1.6 Px sup Rs(3) ≥ y, RH =a = , a≤x≤y≤b
0≤s≤H x(b − a)(y − a)

Ex e−αH ;
 (3)
3.1.8 sup Rs(3) ≥ y, RH =a
0≤s≤H
(1)
√ √
a sh (b − y) 2 sh (x − a) 2



= √
√
, a≤x≤y≤b
x sh (b − a) 2 sh (y − a) 2

(3)


3.1.8 Px sup Rs(3) ≥ y, RH = a, H ∈ dt
0≤s≤H
(2)
 3. STOPPING AT FIRST EXIT TIME 487
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 H = Ha,b = min{s : Rs(3) ∈
/ (a, b)}

a
= ss (b − y, b − a) ∗ sst (x − a, y − a)dt, a≤x≤y≤b
x t

(3)

b(b − x)(y − a)
3.2.6 Px inf Rs(3) ≤ y, RH =b = , a≤y≤x≤b
0≤s≤H x(b − y)(b − a)

Ex e−αH ;
 (3)
3.2.8 inf Rs(3) ≤ y, RH =b
0≤s≤H
(1)
√ √
b sh (b − x) 2 sh (y − a) 2



= √
√
, a ≤ y ≤ x ≤ b
x sh (b − y) 2 sh (b − a) 2
(3)


3.2.8 Px inf Rs(3) ≤ y, RH = b, H ∈ dt
0≤s≤H
(2)
b
= ss (b − x, b − y) ∗ sst (y − a, b − a)dt, a≤y≤x≤b
x t

b(x − r) r(b − a)(b − x)


 x(b − r) + x(b − r)(b − a + 2 (b − r)(r − a)) , r ≤ x ≤ b

3.3.1 Ex e−γ`(H,r) =
 a(r − x ) +
 r(x − a)(b − a)
, a≤x≤r
x(r − a) x(r − a)(b − a + 2 (b − r)(r − a))

r(b − x)(b − a) (b − a)y 

 
exp − dy, r ≤ x ≤ b
2x(b − r)2 (r − a) 2(b − r)(r − a)




3.3.2 Px `(H, r) ∈ dy =
 r(x − a)(b − a) exp − (b − a)y
 
dy, a ≤ x ≤ r

2x(b − r)(r − a)2 2(b − r)(r − a)

 b(x − r)


 x(b − r) , r ≤ x ≤ b
(1) Px `(H, r) = 0 =
 a(r − x) , a ≤ x ≤ r

x(r − a)

3.3.3 Er e−αH−γ`(H,r)
√ √ √
2 a sh((b − r) 2 ) + b sh((r − a) 2 )


= √ √ √ √
=: L
r 2 sh((b − a) 2 ) + 2 sh((b − r) 2 ) sh((r − a) 2 )

 b sh((x − r)√2 √
) r sh((b − x) 2 )
 x sh((b − r)√2 + √ L, r ≤ x ≤ b
) x sh((b − r) 2 )

−αH−γ`(H,r)
Ex e = √ √
 a sh((r − x)√2
 ) r sh((x − a) 2
+ √
)
L, a ≤ x ≤ r
x sh((r − a) 2 ) x sh((r − a) 2 )
488 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 H = Ha,b = min{s : Rs(3) ∈
/ (a, b)}

√ √

2 sh((b − a) 2 )y 
Ex e−αH ; `(H, r) ∈ dy = exp −

3.3.4 √ √
2 sh((b − r) 2 ) sh((r − a) 2 )

√ √ √ √
 2 sh((b − x) 2 )[a sh((b − r) 2 ) + b sh((r − a) 2 )]
√ √ dy, r ≤ x ≤ b
2x sh2 ((b − r) 2 ) sh((r − a) 2 )


× √ √ √ √
2 sh((x − a) 2 )[a sh((b − r) 2 ) + b sh((r − a) 2 )]
dy, a ≤ x ≤ r

√ √
2x sh((b − r) 2 ) sh2 ((r − a) 2 )


 b sh((x − r)√2 )
 x sh((b − r)√2 , r≤x≤b
)

Ex e−αH ; `(H, r) = 0 =

(1) √
 a sh((r − x)√2
 )
, a≤x≤r
x sh((r − a) 2 )

a
Ex e−γ`(H,r) ; RH
 (3)
3.3.5 = a = Q(3) (2γ), for Q(3)
a (2γ) see 1.3.3.5
x a
(a)

b
Ex e−γ`(H,r) ; RH
 (3)
3.3.5 = b = Q(3) (2γ), for Q(3)
b (2γ) see 1.3.3.5
x b
(b)

(3)

a
3.3.6 Px `(H, r) ∈ dy, RH = a = Ba(3) (y)dy, for Ba(3) (y) see 1.3.3.6
x
(a)

0, r≤x≤b
(
a(r − x)
(3)


(1) Px `(H, r) = 0, RH =a =
, a≤x≤r
x(r − a)

(3)

b
3.3.6 Px `(H, r) ∈ dy, RH = b = Bb(3) (y)dy, for Bb(3) (y) see 1.3.3.6
x
(b)
( b(x − r)
, r≤x≤b
x(b − r)
(3)


(1) Px `(H, r) = 0, RH =b =
0, a≤x≤r

a
Ex e−αH−γ`(H,r) ; RH
 (3)
3.3.7 = a = Va(3) (2α), for Va(3) (2α) see 1.3.3.7
x
(a)

b
Ex e−αH−γ`(H,r) ; RH
 (3)
3.3.7 = b = Vb(3) (2α), for Vb(3) (2α) see 1.3.3.7
x
(b)
 3. STOPPING AT FIRST EXIT TIME 489
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 H = Ha,b = min{s : Rs(3) ∈
/ (a, b)}

√ √

2 sh((b − a) 2 )y 
Ex e−αH ; `(H, r) ∈ dy, RH
 (3)
3.3.8 = a = exp − √ √
2 sh((b − r) 2 ) sh((r − a) 2 )
(a)
√ √
 a 2 sh((b − x) 2 )
 2x sh((b − r)√2 ) sh((r − a)√2 ) dy, r ≤ x ≤ b

× √ √
 a 2 sh((x − a√
 ) 2 )
dy, a≤x≤r
2x sh2 ((r − a) 2 )

 0, r≤x≤b
 −αH √
a sh((r − x) 2 )
(3)
(1) Ex e ; `(H, r) = 0, RH =a =
√ , a≤x≤r
x sh((r − a) 2 )


√ √

2 sh((b − a) 2 )y 
Ex e−αH ; `(H, r) ∈ dy, RH
 (3)
3.3.8 = b = exp − √ √
2 sh((b − r) 2 ) sh((r − a) 2 )
(b)
 b√2 sh((b − x)√2 )
 2x sh2 ((b − r)√2 ) dy,
 r≤x≤b
× √ √
b 2 sh((x − a) 2 )
dy, a ≤ x ≤ r

√ √
2x sh((b − r) 2 ) sh((r − a) 2 )


√
 b sh((x − r)√2 ) , r ≤ x ≤ b


Ex e−αH ; `(H, r) = 0, RH x sh((b − r) 2 )

 (3)
(1) =b =
0, a≤x≤r


 Z H

3.4.1 Ex exp −γ 1I[r,b] Rs(3) ds
0
√ √
a(r − x ) (x − a)(a sh((b − r) 2 ) + b 2 (r − a))

+ √ √ √ , a≤x≤r
x(r − a) x(r − a)(sh((b − r) 2 ) + 2 (r − a) ch((b − r) 2 ))




√
b sh((x − r) 2 )


= √
 sh((b − r) 2 )
x
√ √ √
sh((b − x) 2 )(a sh((b − r) 2 ) + b 2 (r − a))



 + √ √ √ √ , r≤x≤b

x sh((b − r) 2 )(sh((b − r) 2 ) + 2 (r − a) ch((b − r) 2 ))

Z H

3.4.2 Px 1I[r,b] Rs(3) ds ∈ dy
0

x−a a

rcy (0, b − r, b − r, r − a) + brc
e y (1, b − r, 0, r − a) dy, x ≤ r

 x r − a
= b ssy (x − r, b − r)dy + 1 ssy (b − x, b − r)


x x



∗ a rcy (0, b − r, b − r, r − a) + b(r − a)rc
e y (1, b − r, 0, r − a) dy, r ≤ x
490 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 H = Ha,b = min{s : Rs(3) ∈
/ (a, b)}

( a (r − x )
H
Z , a≤x≤r
x (r − a)


(3)
(1) Px 1I[r,b] Rs ds = 0 =
0 0, r≤x≤b

H
a
n  Z

 (3)
o
3.4.5 Ex exp −γ 1I[r,b] Rs(3) ds ; RH = a = Q(4) (2γ)
0
x a
(a)

for Q(4)
a (2γ) see 1.3.4.5

H
b
n  Z

 (3)
o
3.4.5 Ex exp −γ 1I[r,b] Rs(3) ds ; RH = b = Q(4) (2γ)
0
x b
(b)

for Q(4)
b (2γ) see 1.3.4.5

H
a
Z

(3)
3.4.6 Px 1I[r,b] Rs(3) ds ∈ dy, RH = a = Ba(4) (y)dy
0
x
(a)

for Ba(4) (y) see 1.3.4.6

( a(r − x)
H
Z , a≤x≤r
x(r − a)


(3) (3)
(1) Px 1I[r,b] Rs ds = 0, RH =a =
0 0, r≤x≤b

H
b
Z

(3)
3.4.6 Px 1I[r,b] Rs(3) ds ∈ dy, RH = b = Bb(4) (y)dy
0
x
(b)

for Bb(4) (y) see 1.3.4.6

 Z H

3.5.1 Ex exp −γ 1I[a,r] Rs(3) ds
0

√
a sh((r − x) 2 )

√
x sh((r − a) 2 )




√ √ √
sh((x − a) 2 )(b sh((r − a) 2 ) + a 2 (b − r))


= + √ √ √ √ , a≤x≤r
 x sh((r − a) 2 )(sh((r − a) 2 ) + 2 (b − r) ch((r − a) 2 ))
√ √
 b(x − r) + (b − x)(b sh((r − a) 2 ) + a 2 (b − r))



√ √ √ , r≤x≤b

x(b − r) x(b − r)(sh((r − a) 2 ) + 2 (b − r) ch((r − a) 2 ))
 3. STOPPING AT FIRST EXIT TIME 491
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 H = Ha,b = min{s : Rs(3) ∈
/ (a, b)}

Z H

3.5.2 Px 1I[a,r] Rs(3) ds ∈ dy
0

a 1
ssy (r − x, r − a)dy + ssy (x − a, r − a)
x x




= ∗ b rcy (0, r − a, r − a, b − r) + a(b − r)rc
e y (1, r − a, 0, b − r) dy, x ≤ r
b−x b



rc (0, r − a, r − a, b − r) + arc
e y (1, r − a, 0, b − r) dy, r ≤ x
x b−r y

0, a≤x≤r
(
Z H 
b(x − r)
(3)


(1) Px 1I[a,r] Rs ds = 0 =
, r≤x≤b
0 x(b − r)

H
a
n  Z

 (3)
o
3.5.5 Ex exp −γ 1I[a,r] Rs(3) ds ; RH = a = Q(5) (2γ)
0
x a
(a)

for Q(5)
a (2γ) see 1.3.5.5
Z H
b
n 
 o
(3)
3.5.5 Ex exp −γ 1I[a,r] Rs(3) ds ; RH = b = Q(5)
b (2γ)
0
x
(b)

for Q(5)
b (2γ) see 1.3.5.5
 H
a
Z 

(3)
3.5.6 Px 1I[a,r] Rs(3) ds ∈ dy, RH = a = Ba(5) (y)dy
0
x
(a)

for Ba(5) (y) see 1.3.5.6

Z H

(3)
(1) Px 1I[a,r] Rs(3) ds = 0, RH =a =0
0

H
b
Z

(3)
3.5.6 Px 1I[a,r] Rs(3) ds ∈ dy, RH = b = Bb(5) (y)dy
0
x
(b)

for Bb(5) (y) see 1.3.5.6

0, a≤x≤r
(
Z H 
b(x − r)
(3)

(3)
(1) Px 1I[a,r] Rs ds = 0, RH = b =
, r≤x≤b
0 x(b − r)
492 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 H = Ha,b = min{s : Rs(3) ∈
/ (a, b)}

 Z H

 
3.6.1 Er exp − p1I[a,r] Rs(3) + q1I[r,b] Rs(3) ds =: Q
0
√ √ √ √
a p sh((b − r) 2q) + b q sh((r − a) 2p)
= √ √ √ √ √ √
r( q ch((b − r) 2q) sh((r − a) 2p) + p sh((b − r) 2q) ch((r − a) 2p))

 Z H

 
Ex exp − p1I[a,r] Rs(3) + q1I[r,b] Rs(3) ds
0

√ √
b sh((x − r) 2q) r sh((b − x) 2q)

 x sh((b − r) 2q) x sh((b − r)√2q) Q, r ≤ x ≤ b
 √ +
= √ √
 a sh((r − x)√2p) + r sh((x − a)√2p) Q, a ≤ x ≤ r

x sh((r − a) 2p) x sh((r − a) 2p)

H
a
n  Z

  (3) o
3.6.5 Ex exp − p1I[a,r] Rs(3) + q1I[r,b] Rs(3) ds ; RH = a = Q(6) (2p, 2q)
0
x a
(a)

for Q(6)
a (2p, 2q) see 1.3.6.5

H
b
n  Z

  (3) o
3.6.5 Ex exp − p1I[a,r] Rs(3) + q1I[r,b] Rs(3) ds ; RH = b = Q(6) (2p, 2q)
0
x b
(b)

for Q(6)
b (2p, 2q) see 1.3.6.5

 Z H

3.7.1 Er exp −γ 1I[r,u] Rs(3) ds
a≤r 0

u≤b
√ √ √ √
a sh((u − r) 2 ) + a(b − u) 2 ch((u − r) 2 ) + b(r − a) 2
= √ √ √ =: Qr
r((1 + (b − u)(r − a)2 ) sh((u − r) 2 ) + (b − u + r − a) 2 ch((u − r) 2 ))

 Z H

Eu exp −γ 1I[r,u] Rs(3) ds
0

√ √ √ √
b sh((u − r) 2 ) + a(b − u) 2 + b(r − a) 2 ch((u − r) 2 )
= √ √ √ =: Qu
u((1 + (b − u)(r − a)2 ) sh((u − r) 2 ) + (b − u + r − a) 2 ch((u − r) 2 ))

 Z H

Ex exp −γ 1I[r,u] Rs(3) ds
0
 3. STOPPING AT FIRST EXIT TIME 493
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 H = Ha,b = min{s : Rs(3) ∈
/ (a, b)}

a(r − x) r(x − a) r

+ Q , a≤x≤r
x(r − a) x(r − a)




√ √
r sh((u − x) 2 ) r u sh((x − r) 2 ) u


= √ Q + √ Q , r≤x≤u
 x sh((u − r) 2 ) x sh((u − r) 2 )
 b(x − u) + u(b − x) Qu ,



u≤x≤b

x(b − u) x(b − u)

H
a
n  Z

 (3)
o
3.7.5 Ex exp −γ 1I[r,u] Rs(3) ds ; RH = a = Q(7) (2γ)
0
x a
(a)

for Q(7)
a (2γ) see 1.3.7.5

H
b
n  Z

 (3)
o
3.7.5 Ex exp −γ 1I[r,u] Rs(3) ds ; RH = b = Q(7) (2γ)
0
x b
(b)

for Q(7)
b (2γ) see 1.3.7.5

H  S 1 ( 2 b3=2 √2 ; 2 x3=2 √2 ) + S 1 ( 2 x3=2 √2 ; 2 a3=2 √2 )

3 3 3 3 3
 Z
3.8.1 Ex exp −γ Rs(3) ds = 3
S1=3 ( 23 b3=2 2 ; 23 a3=2 2 )
√ √
0

Z H o S ( 2 b3=2 √2 ; 23 x3=2 2 )
√
1=3 3
n  
(3)
3.8.5 Ex exp −γ Rs(3) ds ; RH =a =
S1=3 ( 23 b3=2 2 ; 23 a3=2 2 )
√ √
(a) 0

Z H o S ( 2 x3=2 √2 ; 23 a3=2 2 )
√
1=3 3
n  
(3)
3.8.5 Ex exp −γ Rs(3) ds ; RH =b =
S1=3 ( 23 b3=2 2 ; 23 a3=2 2 )
√ √
(b) 0

Z H  S (b2 √ ; x2 √ ) + S (x2 √ ; a2 √ )
1=4 1=4

3.9.1 Ex exp −2γ (Rs(3) )2 ds = √ √
0
S1=4 (b2 ; a2 )

Z H o S (b2 √ ; x2 √ )
1=4
n  
3.9.5 Ex exp −2γ (Rs(3) )2 ds ; RH
(3)
=a = √ √
0
S1=4 (b2 ; a2 )
(a)

Z H o S (x2 √ ; a2 √ )
1=4
n  
3.9.5 Ex exp −2γ (Rs(3) )2 ds ; RH
(3)
=b = √ √
0
S1=4 (b2 ; a2 )
(b)
494 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 H = Ha,b = min{s : Rs(3) ∈
/ (a, b)}
√
aS0 2 e b ; √2 e x + bS √2 e x ; √2 e a
  
H 0 | |
 Z 
2βRs(3) | | | | | |
3.10.1 Ex exp −γ e ds = √ √
2
xS0 | | e ; | 2| e a
b

0

√
o aS√2 =| 2 e b ; √2 e x

n  Z H 
2βRs(3) (3) | | | | |
3.10.7 Ex exp −αH − γ e ds ; RH =a = √
2 e b ; √2 e a
xS√2 =|

(a) 0 | | | | |

√ √
bS√2 =| | | 2| e x ; | 2| e a

n  Z H  o
2βRs(3) (3)
3.10.7 Ex exp −αH − γ e ds ; RH =b = √ √
xS√2 =| | | 2| e b ; | 2| e a

(b) 0

√
a sh((x − a) 2 )
Z b
−γ Ȟ(H)
 (3)
3.12.5 Ex e ; RH = a = √ dy
x
xy sh(y 2 )
(a)

√
b sh((b − x) 2 )
Z x
Ex e−γ Ĥ(H) ; RH
 (3)
3.12.5 =b = √ dy
a
xy sh(y 2 )
(b)
b
a
Z
(3)


3.12.6 Px Ȟ(H) ∈ du, RH =a = ss (x − a, y)dy du
x
xy u
(a)
x
b
Z
(3)


3.12.6 Px Ĥ(H) ∈ du, RH = b = ss (b − x, y)dy du
a
xy u
(b)
√ √
a 2 sh((x − a) 2 + 2 )
Z b
−αH−γ Ȟ(H)
 (3)
3.12.7 Ex e ; RH = a = √ √ dy
x x sh(y 2 ) sh(y 2 + 2 )
(a)
√ √
b 2 sh((b − x) 2 + 2 )
Z x
Ex e−αH−γ Ĥ(H) ; RH
 (3)
3.12.7 =b = √ √ dy
a x sh(y 2 ) sh(y 2 + 2 )
(b)

a
Z b
(3)
3.12.8 Px Ȟ(H) ∈ du, H ∈ dt, RH =a = ssu (x − a, y) st−u (y)dy dudt
x x
(a)
u<t
x
b
Z
(3)


3.12.8 Px Ĥ(H) ∈ du, H ∈ dt, RH =b = ssu (b − x, y) st−u (y)dy dudt
x a
(b)
u<t

3.18.1 Ex exp −γ`(H, r) − η`(H, u)
r≤u

a (x − a) a − b + 2a (b − r) + 2a(b − u) + 4a (b − u)(u − r)

a≤x = −
x x((b − a) + 2 (b − r)(r − a) + 2(b − u)(u − a) + 4 (b − u)(u − r)(r − a))
x≤r
 3. STOPPING AT FIRST EXIT TIME 495
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 H = Ha,b = min{s : Rs(3) ∈
/ (a, b)}

x(b − a) + 2 b(x − r)(r − a) + 2a(b − u)(u − x)

r≤x =
x((b − a) + 2 (b − r)(r − a) + 2(b − u)(u − a) + 4 (b − u)(u − r)(r − a))
x≤u

b (b − x) b − a + 2b(u − a) + 2b (r − a) + 4b (u − r)(r − a)

u≤x = −
x x((b − a) + 2 (b − r)(r − a) + 2(b − u)(u − a) + 4 (b − u)(u − r)(r − a))
x≤b


(3) a
3.18.5 Ex exp −γ`(H, r) − η`(H, u) ; RH = a = Q(18) (2γ, 2η)
x a
(a)

for Q(18)
a (2γ, 2η) see 1.3.18.5


(3) b
3.18.5 Ex exp −γ`(H, r) − η`(H, u) ; RH = b = Q(18) (2γ, 2η)
x b
(b)

for Q(18)
b (2γ, 2η) see 1.3.18.5

(3)

a (18)
3.18.6 Px `(H, r) ∈ dy, `(H, u) ∈ dv, RH = a = Bx,a (y, v)dydv
x
(a)
(18)
for Bx,a (y, v) see 1.1.18.6
(3)

b (18)
3.18.6 Px `(H, r) ∈ dy, `(H, u) ∈ dv, RH = b = Bx,b (y, v)dydv
x
(b)
(18)
for Bx,b (y, v) see 1.1.18.6

√ √
S (2 2 b; 2 2 x)
H
ds
Z
=a = 1 √
n   o
(3)
3.19.5 Ex exp −γ ; RH √
0 Rs(3)
S1 (2 2 b; 2 2 a)
(a)

√ √
S (2 2 x; 2 2 a)
H
ds
Z
=b = 1 √
n   o
(3)
3.19.5 Ex exp −γ ; RH √
0 Rs(3)
S1 (2 2 b; 2 2 a)
(b)

 2Z H o √a (b=x)√ 2 +1=4 − (x=b)√ 2 +1=4

ds
n 
(3)
3.20.5 Ex exp − ; R = a = √ √ √
2 Rs(3) 2 H
x (b=a) 2 +1=4 − (a=b) 2 +1=4



(a) 0
496 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 H = Ha,b = min{s : Rs(3) ∈
/ (a, b)}

 2Z H o √b (x=a)√ 2 +1=4 − (a=x)√ 2 +1=4

ds
n 
(3)
3.20.5 Ex exp − ; R = b = √ √ √
2 Rs(3) 2 H
x (b=a) 2 +1=4 − (a=b) 2 +1=4



(b) 0

 Z H
p2 q2 Rs(3) 2  
n
o
(3)
3.21.5 Ex exp − + ds ; R = a
2 Rs(3) 2 2 H


(a) 0
√
x p2 +1=4−1=2 S√4p2 +1=4 (qb2 =2; qx2 =2)
= √
a p2 +1=4−1=2 S√4p2 +1=4 (qb2 =2; qa2 =2)

 Z H
p2 q2 Rs(3) 2  
n
o
(3)
3.21.5 Ex exp − (3)
2
+ ds ; RH =b
0 2 Rs 2
(b)
√
x p2 +1=4−1=2 S√4p2 +1=4 (qx2 =2; qa2 =2)
= √
b p2 +1=4−1=2 S√4p2 +1=4 (qb2 =2; qa2 =2)

p2
H
q

n  Z  o
(3)
3.22.5 Ex exp − + ds ; R = a
2 Rs(3) 2 Rs(3) H


(a) 0

√ √ √
x p2 +1=4−1=2 S√4p2 +1 (2 2qb; 2 2qx)
= √ √ √
a p2 +1=4−1=2 S√4p2 +1 (2 2qb; 2 2qa)

p2
H
q

n  Z  o
(3)
3.22.5 Ex exp − (3)
2
+ (3) ds ; RH =b
(b) 0 2 Rs Rs

√ √ √
x p2 +1=4−1=2 S√4p2 +1 (2 2qx; 2 2qa)
= √ √ √
b p2 +1=4−1=2 S√4p2 +1 (2 2qb; 2 2qa)

n  Z H


 1
o
3.27.1 Ex exp − p1I[a,r) Rs(3) + q1I[r,b] Rs(3) ds ; `(H, r ∈ dy
0
2
p p p p

= exp −y( 2q cth((b − r) 2q) + 2p cth((r − a) 2p))

 √2 sh((b − x)√2q)[a√p sh((b − r)√2q) + b√q sh((r − a)√2p)]

√ √ dy, r ≤ x ≤ b
x sh2 ((b − r) 2q) sh((r − a) 2p)


× √ √ √ √ √ √
 2 sh((x − a) 2p)[a p sh((

√
b − r) 2q) + b q sh((r − a) 2p)]
√ dy, a ≤ x ≤ r
2
x sh((b − r) 2q) sh ((r − a) 2p)
 3. STOPPING AT FIRST EXIT TIME 497
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 H = Ha,b = min{s : Rs(3) ∈
/ (a, b)}

n  Z H


 1
(3)
o
3.27.5 Ex exp − p1I[a,r) Rs(3) + q1I[r,b] Rs(3) ds ; `(H, r ∈ dy, RH =a
0
2
(a)
p p p p

= exp −y( 2q cth((b − r) 2q) + 2p cth((r − a) 2p))

√ √
a 2p sh((b − x) 2q)

 x sh((b − r)√2q) sh((r − a)√2p) dy, r ≤ x ≤ b

× √ √
 a 2p sh((x − a√) 2p) dy, a≤x≤r
2

x sh ((r − a) 2p)

n  Z H



o
(3)
(1) Ex exp − p1I[a,r) Rs(3) + q1I[r,b] Rs(3) ds ; ` Hz , r = 0, RH =a
0
√
a sh((r − x) 2p)
= √ , a≤x≤r
x sh((r − a) 2p)

n  Z H


 1
(3)
o
3.27.5 Ex exp − p1I[a,r) Rs(3) + q1I[r,b] Rs(3) ds ; ` Hz , r ∈ dy, RH =b
0
2
(b)
p p p p

= exp −y( 2q cth((b − r) 2q) + 2p cth((r − a) 2p))

 √ √
b 2q sh((b − x) 2q)
√ dy, r≤x≤b
x sh2 ((b − r) 2q)


× √ √
b 2q sh((x − a) 2p)
√ √ dy, a ≤ x ≤ r

x sh((b − r) 2q) sh((r − a) 2p)


n  Z H



o
(3)
(1) Ex exp − p1I[a,r) Rs(3) + q1I[r,b] Rs(3) ds ; ` H, r = 0, RH =b
0
√
b sh((x − r) 2q)
= √ , r≤x≤b
x sh((b − r) 2q)

Z H H
1
Z 


(3)
3.27.6 Px 1I[a,r) Rs(3) ds ∈ du, 1I[r,b] Rs(3) ds ∈ dv, `(H, r) ∈ dy, RH =a
0 0
2
(a)
a
= Ba(27) (u, v, y)dudvdy, for Ba(27) (u, v, y) see 1.3.27.6
x

Z H H
1
Z 


(3)
3.27.6 Px 1I[a,r) Rs(3) ds ∈ du, 1I[r,b] Rs(3) ds ∈ dv, `(H, r) ∈ dy, RH =b
0 0
2
(b)
b
= Bb(27) (u, v, y)dudvdy, for Bb(27) (u, v, y) see 1.3.27.6
x
498 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 H = Ha,b = min{s : Rs(3) ∈
/ (a, b)}

n 

3.31.5 Ex(b−a)+a exp − sup `(s, u(b − a) + a) − α−1 `(s, r(b − a) + a) ;
2(b − a) 0≤s≤H
(a)

a(1 − x) a(1 − u)( + r − r2 − u + u2 )(u ∧ x − ux)

o
(3)
0<α RH =a = −
x(b − a)+ a (x(b − a)+ a)(2 (u ∧ r − ur)|u − r| +  + r − r2 − u + u2)

a(1 − u) 2(u ∧ r − ur)|u − r| − ( + r − r2 − u + u2 )(u − u2 ))(x ∧ r − xr)

−
(x(b − a) + a)(2 (u ∧ r − ur)|u − r| +  + r − r2 − u + u2 )(u ∧ r − ur)

n 

3.31.5 Ex(b−a)+a exp − sup `(s, u(b − a) + a) − α−1 `(s, r(b − a) + a) ;
2(b − a) 0≤s≤H
(b)

bx bu( + r − r2 − u + u2 )(u ∧ x − ux)

o
(3)
0<α RH =b = −
x(b − a)+ a (x(b − a)+ a)(2 (u ∧ r − ur)|u − r| +  + r − r2 − u + u2)

bu 2(u ∧ r − ur)|u − r| − ( + r − r2 − u + u2 )(u − u2 ))(x ∧ r − xr)

−
(x(b − a) + a)(2 (u ∧ r − ur)|u − r| +  + r − r2 − u + u2 )(u ∧ r − ur)


`(s, u(b − a) + a) − α−1 `(s, r(b − a) + a) > (b − a)h,


3.31.6 Px(b−a)+a sup
0≤s≤H
(a)

(3)

a(1 − u)

( + r − r2 − u + u2 )h h x ∧ r − xr
0<α RH =a = exp −
x(b − a) + a 4(u ∧ r − ur)|u − r| u ∧ r − ur

 + r − r2 − u + u2 (x ∧ r − xr)(u − u2 )
 i
− − u ∧ x + ux
2(u ∧ r − ur)|u − r| u ∧ r − ur


`(s, u(b − a) + a) − α−1 `(s, r(b − a) + a) > (b − a)h,


3.31.6 Px(b−a)+a sup
0≤s≤H
(b)

(3)

bu

( + r − r2 − u + u2 )h h x ∧ r − xr
0<α RH =b = exp −
x(b − a) + a 4(u ∧ r − ur)|u − r| u ∧ r − ur

 + r − r2 − u + u2 (x ∧ r − xr)(u − u2 )
 i
− − u ∧ x + ux
2(u ∧ r − ur)|u − r| u ∧ r − ur

p
Υ := (r − r2 − αu + αu2 )2 + 4α(u ∧ r − ur)|u − r|
 4. STOPPING AT INVERSE LOCAL TIME 499
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 % = %(v, z) = min{s : `(s, z) = v}

4. Stopping at inverse local time

√ √
z sh(x 2 ) v 2
  
√ exp − √ , 0≤x≤z
x sh(z 2 ) 1 − e−2z 2


Ex e−α% ; % < ∞ =

4.0.1 √
 z exp −(x − z)√2α − v 2
 
, z≤x

√
x 1 − e−2z 2

( z
esst (x, z, v/2, v/2)dt, 0≤x≤z
4.0.2 Px % ∈ dt =

x
z
es (0, 0, z, x − z + v/2, v/2)dt, z ≤ x
x t

v
  
 1 − exp − , 0≤x≤z

2z
(1) Px % = ∞ =
 1 − z exp − v , z ≤ x
 
x 2z
√ √ 
2v 2v
Ex exp −γ sup Rs(3) ); % < ∞ = e−γz √ K1 √

4.1.1
0≤s≤% z z
x≤z

vy
  

 exp − 2z (y − z ) , 0≤x≤z

4.1.2 Px sup Rs(3) < y, % < ∞ =
 z (y − x) exp − vy
 
0≤s≤% , z≤x≤y
z≤y

x(y − z ) 2z (y − z )

Ex e−α% ; sup Rs(3) < y, % < ∞


4.1.4
0≤s≤%
z≤y

 z sh(x√2 )  √ √
v 2 sh(y 2 )

 x sh(z √2 ) exp − 2 sh(z √2 ) sh((y − z )√2 ) ,
 0≤x≤z
= √ √ √
 z sh((y − x)√2 ) exp − v 2 sh(y 2 )
 
, z≤x≤y

√ √
x sh((y − z ) 2 ) 2 sh(z 2 ) sh((y − z ) 2 )

z (x − y) v
  
exp − , y≤x≤z
x(z − y) 2(z − y)




4.2.2 Px inf Rs(3) > y, % < ∞ =
 z exp − v
 
0≤s≤%
0≤y
 , y≤z≤x
x 2(z − y)
500 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 % = %(v, z) = min{s : `(s, z) = v}

(1) Px inf Rs(3) > y, % = ∞
0≤s≤%

 y
1− , z≤y≤x
x 


y v

   

1 − 1 − exp − , y≤x≤z
= x 2(z − y)
 1 − y − z − y exp − v

  

, y≤z≤x

x x 2(z − y)

Ex e−α% ;

4.2.4 inf Rs(3) > y
0≤s≤%
0≤y

√ √
z sh((x − y) 2 ) v 2
  
√ exp − √ , y≤x≤z
x sh((z − y) 2 ) 1 − e−2(z−y) 2


= √
 z exp −(x − z)√2α − v 2
 
y≤z≤x

√ ,
x 1 − e 2(z−y) 2
−

(z
esst (x − y, z − y, v/2, v/2)dt, x≤z
> y, % ∈ dt = xz
(3)


(1) Px inf Rs
0≤x
0≤s≤% es (0, 0, z − y, x − z + v/2, v/2)dt, z ≤ x
x t

Ex e−γ`(%,r) ; % < ∞

4.3.1

v(1 + 2r )
  
exp − , 0≤x≤z


 2z (1 + 2 (r − z ))
z (1 + 2 (r − x)) v(1 + 2r )

  
z≤r = exp − , z≤x≤r
 x (1 + 2 (r − z )) 2z (1 + 2 (r − z ))
z v(1 + 2r )

  

 exp − , r≤x
x(1 + 2 (r − z )) 2z (1 + 2 (r − z ))

z v(1 + 2r )
  
exp − , 0≤x≤r
z + 2 (z − r)r 2(z + 2 (z − r)r)



z (x + 2 (x − r)r) v(1 + 2r )

  
r≤z = exp − , r≤x≤z
 x (z + 2 (z − r)r) 2(z + 2 (z − r)r)
 z exp − v(1 + 2r )

  

, z≤x
x 2(z + 2 (z − r)r)

Ex e−γ`(%,r) ; % = ∞

(1)
 4. STOPPING AT INVERSE LOCAL TIME 501
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 % = %(v, z) = min{s : `(s, z) = v}
 1  
v(1 + 2r )

1 − exp − , 0≤x≤z
1+2 r 2z (1 + 2 (r − z ))



1 z (1 + 2 (r − x)) v(1 + 2r )

   

1− exp − , z≤x≤r

1+2 r x(1 + 2 (r − z )) 2z (1 + 2 (r − z ))


z≤r =
 x + 2 r(x − r)
1+2 r




z v(1 + 2r )

  

− exp − , r≤x
x(1 + 2 r)(1 + 2 (r − z )) 2z (1 + 2 (r − z ))

 1  
v(1 + 2r )

1 − exp − , 0≤x≤r


 1+2 r 2(z + 2 (z − r)r)
x + 2 (x − r)r v(1 + 2r )

   
r≤z = 1 − exp − , r≤x≤z
 x (1 + 2 r ) 2(z + 2 (z − r)r)
 x + 2 (x − r)r − z + 2 (z − r)r exp − v(1 + 2r )

  

, z≤x
x(1 + 2 r) x(1 + 2 r) 2(z + 2 (z − r)r)

4.3.2 Px ` %, r ∈ dy, % < ∞
√ √

1 v

(v + y)   vy 
√ exp − I dy, 0 ≤ x ≤ z,
2(r − z ) y 2(r − z ) 1 (r − z )




√  √vy 
z (v + y) h v

 

exp − (r − x) √ I1

2x(r − z )2 2(r − z ) y (r − z )


=  √vy i
z≤r +(x − z)I0 dy, z≤x≤r

(r − z )




  √vy 
z (v + y)

 

exp − I dy, r≤x

2x(r − z ) 2(r − z ) 0 (r − z )


√ √

z v

(vr + yz )   vy 
√ exp − I dy, z≤x
2x(z − r) y 2(z − r)r 1
(z − r)




√ √
z (vr + yz ) v vy 

  h 

exp − (x − r) I

√
2x(z − r)2 2(z − r)r y 1
(z − r)


=  √vy i
r≤z +(z − x)I0 dy, r≤x≤z

(z − r)




√
z (vr + yz ) vy 

   

exp − I0 dy, 0≤x≤r

2r(z − r) 2(z − r)r (z − r)


(2) Px ` %, r = 0, % < ∞

vr
  
exp − , 0≤x≤z
2z (r − z )




z≤r z (r − x) vr
 
= exp − , z≤x≤r


 x(r − z ) 2z (r − z )
r≤x

0,
502 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 % = %(v, z) = min{s : `(s, z) = v}

0≤x≤r

 0,
 z (x − r) v

  
exp − , r≤x≤z
r≤z = x(z − r) 2(z − r)
 z exp − v

  
, z≤x

x 2(z − r)
 r
1− , z≤r≤x
x


r v

   


Px `(%, r) = 0, % = ∞ = 1− 1 − exp − , r≤x≤z
(3) x 2(z − r)
 1 − r − z − r exp − v

  

, r≤z≤x

x x 2(z − r)

4.3.3 Ex e−α%−γ`(%,r)

√ √√
v er 2 + 2 sh(r 2 )


z≤r = exp − √ √ √ √
sh(z 2 ) 2 e(r−z) 2 + 2 sh((r − z ) 2 )

√
 z sh(x 2 )
√ , 0≤x≤z
x sh(z 2 )




√ √ √
z 2 e(r−x) 2 + 2 sh((r − x) 2

)



× √ √ √
, z≤x≤r
 x 2 e(r−z) 2 + 2 sh((r − z ) 2 )
√ √

z 2 e−(x−r) 2




, r≤x

√ √ √
2 e(r−z) 2 + 2 sh((r − z ) 2 )

x
√ √ √ √
ve(z−r) 2 er 2 + 2 sh(r 2 )


r≤z = exp − √ √ √ √
2 sh(z 2 ) + 2 sh((z − r) 2 ) sh(r 2 )

√ √
z 2 sh(x 2 )

√ √ √ √
, 0≤x≤r
x 2 sh(z 2 ) + 2 sh((z − r) 2 ) sh(r 2 )




√ √ √ √

z 2 sh(x 2 ) + 2 sh((x − r) 2 ) sh(r 2 )


× √ √ √ √
, r≤x≤z

 x 2 sh(z 2 ) + 2 sh((z − r) 2 ) sh(r 2 )
z −(x−z) 2α
√



 e , z≤x
x

Ex e−α% ; `(%, r) ∈ dy

4.3.4

√ √ √ √

2 sh(r 2 )v 2 e(r−z) 2 y 
z≤r = exp − √ √ − √
2 sh((r − z ) 2 ) sh(z 2 ) 2 sh((r − z ) 2 )
 4. STOPPING AT INVERSE LOCAL TIME 503
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 % = %(v, z) = min{s : `(s, z) = v}

√ √ √
z sh(x 2 ) 2 v 2 vy
  
√ √ √ 1 I √ dy, 0≤x≤z
2x sh(z 2 ) sh((r − z ) 2 ) y sh((r − z ) 2 )




√ √ √

z 2 √ v 2 vy

 h  

√ sh((r − x) 2α) I √
2x sh2 ((r − z ) 2 )
√

y 1
sh((r − z ) 2 )


× √
√ 
2 vy
i
+ sh((x − z) 2α)I0 √ dy, z≤x≤r

sh((r − z ) 2 )




√ √ √

 z 2 e−(x−r) 2 2 vy

  
√ I0 √ dy, r≤x

2x sh((r − z ) 2 ) sh((r − z ) 2 )


 √ (z−r)√2 √ √
2 e v 2 sh(z 2 )y 
r≤z = exp − √ − √ √
2 sh((z − r) 2 ) 2 sh((z − r) 2 ) sh(r 2 )

 √ √ √
z 2 ve−(x−z) 2 2 vy
 
 √ √ I1 √ dy, z≤x
2x sh((z − r) 2 ) y sh((z − r) 2 )



√

√
z 2 √ 2 vy

 h qv  

sh((x − r) 2α) I1

2 √ √
2x sh ((z − r) 2 ) y sh((z − r) 2 )


× √
√ 
2 vy
i
+ sh((z − x) 2α)I0 r≤x≤z

 √ dy,
sh((z − r) 2 )



√ √ √

z 2 sh(x 2 ) 2 vy

  

√ √ I √ dy, 0≤x≤r

2x sh((z − r) 2 ) sh(r 2 ) 0 sh((z − r) 2 )


Ex e−α% ; `(%, r) = 0

(1)

 z sh(x√2 )
 x sh(z √2 ) ,
 0≤x≤z
√ √
v 2 sh(r 2 )
 
 √
z≤r = exp − √ √ z sh((r − x) 2 )
2 sh(z 2 ) sh((r − z ) 2 )  √ , z≤x≤r
 x sh((r − z ) 2 )


0, r≤x
 z √
e−(x−z) 2α , z≤x
 x

√ √

v 2 e (z − r ) 2 √
z sh((x − r) 2 )
 
r≤z = exp − √ , r≤x≤z
2 sh((z − r) 2 )  √
 x sh((z − r) 2 )

0, 0≤x≤r


n  Z %
 o
4.4.1 Ex exp −γ 1I[r,∞) Rs(3) ds ; % < ∞
0
504 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 % = %(v, z) = min{s : `(s, z) = v}

1, 0≤x≤z

√
 z (1 + (r − x) 2 )

√

v(1 + 2 r)

√ , z≤x≤r

z≤r = exp − √ x(1 + (r − z ) 2 )
2z (1 + (r − z ) 2 )  √
 ze−(x−r) √ 2


, r≤x
x(1 + (r − z ) 2 )

√ √
v( 2 + 2r )e(z−r) 2
 
r≤z = exp − √ √ √
2 r 2 ch((z − r) 2 ) + sh((z − r) 2 )

√
 z 2
√ √ √ , 0≤x≤r
r 2 ch((z − r) 2 ) + sh((z − r) 2 )



√ √ √

z r 2 ch((x − r) 2 ) + sh((x − r) 2 )


× √ √ √
, r≤x≤z
 x r 2 ch((z − r) 2 ) + sh((z − r) 2 )
z −(x−z)√2γ



e , z≤x

x

v
  
exp − , 0≤x≤z
2(r − z )


Z %  
r−x v
(3)

 
4.4.2 Px 1I[r,∞) Rs ds = 0 = exp − , z≤x≤r
(1) 0 

 r−z 2(r − z )
0, r≤x

z≤r

n Z % o
−α%


4.4.4 Ex e ; 1I[r,∞) Rs(3) ds = 0
(1) 0

 z sh(x√2 )
 x sh(z √2 ) ,
 0≤x≤z
√ √
v 2 sh(r 2 )
 
 √
z≤r = exp − √ √ z sh((r − x) 2 )
sh(z 2 ) sh((r − z ) 2 )  √ , z≤x≤r
 x sh((r − z ) 2 )


0, r≤x

n  Z %
 o
4.5.1 Ex exp −γ 1I[0,r] Rs(3) ds ; % < ∞
0
 z sh(x√2 )
√ , 0≤x≤z
 x sh(z 2 )


√ √ √
v 2 ch(r 2 )

z ch((r − x) 2 )
 
z≤r = exp − √ √ √ , z≤x≤r
2 sh(z 2 ) ch((r − z ) 2 )  x ch((r − z ) 2 )
z



√ , r≤x
x ch((r − z ) 2 )

 4. STOPPING AT INVERSE LOCAL TIME 505
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 % = %(v, z) = min{s : `(s, z) = v}

√ √
v 2 ch(r 2 )
 
r≤z = exp − √√ √
2(sh(r 2 ) + 2 (z − r) ch(r 2 )
√
 z sh(x 2 )
√ √ √
, 0≤x≤r
x sh(r 2 ) + 2 (z − r) ch(r 2 )



√ √ √

z sh(r 2 ) + 2 (x − r) ch(r 2 )


× √ √ √
, r≤x≤z

 x sh(r 2 ) + 2 (z − r) ch(r 2 )
z


, z≤x

x

n  Z %
 o
(1) Ex exp −γ 1I[0,r] Rs(3) ds ; % = ∞
0

√ √ √

sh(x 2 )  
v 2 ch(r 2 )

√ √ 1 − exp − √ √ , 0≤x≤z
x 2 ch(r 2 ) 2 sh(z 2 ) ch((r − z ) 2 )



√ √ √

sh(x 2 ) sh(z 2 ) ch((r − x) 2 )



√ √ − √ √ √

x 2 ch( r 2 ) x 2 ch(r 2 ) ch((r − z ) 2 )




√ √
v 2 ch(r 2 )

  
z≤r = × exp − √ √ , z≤x≤r
 2 sh( z 2 ) ch((r − z ) 2 )
√ √ √ √
sh(r 2 ) + 2 (x − r) ch(r 2 ) sh(z 2 )



√ √ − √ √ √

x 2 ch( r 2 ) x 2 ch( r 2 ) ch((r − z ) 2 )




√ √
v 2 ch(r 2 )

  

× exp − √ √ , r≤x

2 sh(z 2 ) ch((r − z ) 2 )


√ √ √

sh(x 2 )  
v 2 ch(r 2 )

√ √ 1 − exp − √ √ √ ,0≤x≤r
x 2 ch(r 2 ) 2(sh(r 2 ) + 2 (z − r) ch(r 2 )



√ √ √

sh(r 2 ) + 2 (x − r) ch(r 2 )



√ √

x 2 ch(r 2 )




√ √
v 2 ch(r 2 )

   
r≤z = × 1 − exp − √ √ √ , r≤x≤z
 2(sh(r 2 ) + 2 (z − r) ch(r 2 )
√ √ √ √ √ √
sh(r 2 ) + 2 (x − r) ch(r 2 ) sh(r 2 ) + 2 (z − r) ch(r 2 )



√ √ − √ √

x 2 ch(r 2 ) x 2 ch(r 2 )




√ √
v 2 ch(r 2 )

  

× exp − √ √ √ , z≤x

2(sh(r 2 ) + 2 (z − r) ch(r 2 )


0≤x≤r

 0,
z (x − r) v

%
Z   
− r≤x≤z
 
exp ,
x(z − r) 2(z − r)
(3)


4.5.2 Px 1I[0,r] Rs ds = 0 =
 z exp − v
(1) 0 
  
, z≤x

r≤z x 2(z − r)
506 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 % = %(v, z) = min{s : `(s, z) = v}

n Z % o
Ex e−α% ;


4.5.4 1I[0,r] Rs(3) ds = 0
(1) 0


0, 0≤x≤r
√

√ √2 (z−r)  z sh((x − r) 2 )

v e
 
r≤z = exp − √ √ √ , r≤x≤z
2 sh((z − r) 2 )  x sh((z − r) 2 )
 z √
 e−(x−z) 2α ,

z≤x
x

 Z %



4.6.1 Ex exp − p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds
0

√ √ √ √ √
v p p ch(r 2p) + q sh(r 2p)


z≤r = exp − √ √ √ √ √

sh(z 2p) 2p ch((r − z ) 2p) + 2q sh((r − z ) 2p)

√
 z sh(x 2p)
√ , 0≤x≤z
x sh(z 2p)



 √ √ √ √

z p ch((r − x) 2p) + q sh((r − x) 2p)


× √ √ √ √
, z≤x≤r
 x p ch((r − z ) 2p) + q sh((r − z ) 2p)
√
z pe−(x−r) 2q
√



r≤x

√ √ √ √
,
x p ch((r − z ) 2p) + q sh((r − z ) 2p)


√ √ √
v q p ch(r 2p) + q sh(r 2p) e(z−r) 2q
 √ √ √

r≤z = exp − √ √ √ √ √ √
2q sh(r 2p) ch((z − r) 2q) + 2p ch(r 2p) sh((z − r) 2q)

√ √
 z q sh(x 2p)
√ √ √ √ √ √
, 0 ≤ x ≤ r
x q sh(r 2p) ch((z − r) 2q) + p ch(r 2p) sh((z − r) 2q)



√ √ √ √


 √ √
× z √q sh(r √2p) ch((x − r)√2q) + √p ch(r √2p) sh((x − r)√2q)
, r ≤ x ≤ z

 x q sh(r 2p) ch((z − r) 2q) + p ch(r 2p) sh((z − r) 2q)
 z −(x−z)√2q


e , z≤x
x

n  Z %
 o
4.7.1 Ex exp −γ 1I[r,u] Rs(3) ds ; % < ∞
r<u 0

√ √ √
v r 2 sh((u − r) 2 ) + ch((u − r) 2 )
 

z≤r = exp − √ √ √
2z ((r − z ) 2 sh((u − r) 2 ) + ch((u − r) 2 ))
 4. STOPPING AT INVERSE LOCAL TIME 507
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 % = %(v, z) = min{s : `(s, z) = v}

1, 0≤x≤z
√ √ √

z ((r − x) 2 sh((u − r) 2 ) + ch((u − r) 2 ))



√ √ √ , z≤x≤r

x((r − z ) 2 sh((u − r) 2 ) + ch((u − r) 2 ))



× √
z ch((u − x) 2 )
√ √ √ , r≤x≤u
x((r − z ) 2 sh((u − r) 2 ) + ch((u − r) 2 ))



z



√ √ √ , u≤x

x((r − z ) 2 sh((u − r) 2 ) + ch((u − r) 2 ))


√ √ √
v r2 sh((u − r) 2 ) + 2 ch((u − r) 2 )


u≤z = exp − √ √ √
2((1 + (z − u)r2 ) sh((u − r) 2 ) + (r + z − u) 2 ch((u − r) 2 ))
√
 z 2
√ √ √ , 0≤x≤r
(1 + (z − u)r2 ) sh((u − r) 2 ) + (r + z − u) 2 ch((u − r) 2 )



√ √ √

z (sh((x − r) 2 ) + r 2 ch((x − r) 2 ))



√ √ √ , r≤x≤u

x ((1 + ( z u) r2 ) sh((u − r) 2 ) + (r + z − u) 2 ch((u − r) 2 ))

× −
√ √ √
 z ((1 + (x − u)r2 ) sh((u − r) 2 ) + (r + x − u) 2 ch((u − r) 2 ))
√ √ √ , u≤x≤z

x((1 + (z − u)r2 ) sh((u − r) 2 ) + (r + z − u) 2 ch((u − r) 2 ))



z



, z≤x
x

n  Z %  o
4.8.1 Ex exp −γ Rs(3) ds ; % < ∞
0
zI1=3 ( 23 x3=2 2 )
 √ √
, 0≤x≤z
xI1=3 ( 23 z 3=2 2 )
 √ √
 −1
3z v 

= exp −
4I1=3 ( 23 z 3=2 2 )K1=3 ( 23 z 3=2 2 ) 
√ √
zK ( 2 x3=2 2 )
√ √
 √ 1=3 32 3=2 √ ,
 z≤x
xK1=3 ( 3 z 2 )

n  Z %  o
4.9.1 Ex exp −2γ (Rs(3) )2 ds ; % < ∞
0
√
zI1=4 (x2 )
 √
√ √ , 0≤x≤z
xI1=4 (z 2 )

v
 

= exp − √ √ √
zI1=4 (z 2 )K1=4 (z 2 )  zK (x2 )
√
 √ 1=4 2 √ , z≤x
xK1=4 (z )


√
v 2−1 K√2 = | 2|

 Z %   
(3)
4.10.3 Ex exp −α% − γ e2βRs ds =exp − √
2 √ 
2
√
2
S√2 = | | e z; | | K√2 = | | e z

0

√ √
zS 2 = | 2| e x ; | 2|
 √ 

, 0 ≤ x ≤ z
 √ √
 xS√2 = | 2| e z ; | 2|



β>0 × √
zK√2 = | 2| e x



, z≤x
 √
xK√2 = | 2| e z


508 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 % = %(v, z) = min{s : `(s, z) = v}

√
z sh((x − z )+ 2 ) yv
 
Ex e−γ Ȟ(%) ; sup Rs(3) ∈ dy =

4.13.1 √ exp − dy
0≤s≤% x(y − z ) sh((y − z ) 2 ) 2z (y − z )

√ √
√ v 2 sh(y 2 ) yv 
 
z2 2 exp − exp −
  
− √ √
2 sh((y − z ) 2 ) sh(z 2 ) 2z (y − z )
+ √ √ √ √
x(y sh((y − z ) 2 ) sh(z 2 ) − z (z − y) 2 sh(y 2 ))

√
sh(x 2γ)dy, 0≤x≤z
(
√ √
× sh((y − x) 2 ) sh(z 2 )
√ dy, z ≤ x ≤ y
sh((y − z ) 2 )

Ex e−γ Ĥ(%) ; inf Rs(3) ∈ dy, % < ∞


4.14.1
0≤s≤%

√
z sh((z − x)+ 2 ) v
 
= √ exp − dy
x(z − y) sh((z − y) 2 ) 2(z − y)

√ √ √
2 2 e−(z−y) 2  
v 2
 
v

+ √ 2(y z )√
2 exp − 2(y z )√
2 − exp −
1 − 2(z − y) 2 − e − 1−e − 2(z − y)

 z sh((x − y)√2 )
√ dy, y ≤ x ≤ z
× x sh((z − y√) 2 )

 z −(x−z) 2γ
e dy, z≤x
x

Ex e−γ Ĥ(%) ; inf Rs(3) ∈ dy, % = ∞


(1)
0≤s≤%

√
sh((z − x)+ 2 )  
v

= √ 1 − exp − dy
x sh((z − y) 2 ) 2(z − y)
√
√ v 2 v  
2(z − y) 2 exp − exp −
   
− 2(y z )√
2
v 1−e 2(z − y)

−
 
+ 1 − exp − − √ √
2(z − y) 1 − 2(z − y) 2 − e2(y−z) 2

 −(z−y)√2 √
 e sh((x − y) 2 )
 √ dy, y ≤ x ≤ z
× x sh((z − y) 2 )
 1 e−(x−y) 2γ dy,
√
z≤x

x
 4. STOPPING AT INVERSE LOCAL TIME 509
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 % = %(v, z) = min{s : `(s, z) = v}

4.15.2 Px a < inf Rs(3) , sup Rs(3) < b
0≤s≤% 0≤s≤%

z (x − a) (b − a)v 
 
 x(z − a) exp − 2(z − a)(b − z ) ,
 a≤x≤z
=
 z (b − x) exp − (b − a)v
 
, z≤x≤b

x(b − z ) 2(z − a)(b − z )

Ex e−α% ; a < inf Rs(3) , sup Rs(3) < b


4.15.4
0≤s≤% 0≤s≤%

 z sh((x − a)√2 )  √ √
v 2 sh((b − a) 2 )

 x sh((z − a)√2 exp − √ √ , −b ≤ x ≤ z
) 2 sh((b − z ) 2 ) sh((z − a) 2 )

= √ √ √
 z sh((b − x)√2 ) v 2 sh((b − a) 2 )
 
exp − , z≤x≤b

√ √
x sh((b − z ) 2 ) 2 sh((b − z ) 2 ) sh((z − a) 2 )



4.18.1 Ex exp −γ`(%, r) − η`(%, u) ; % < ∞
r<u
v(1 + 2 r + 2u + 4 (u − r)r)
 
z<r = exp −
2z (1 + 2 (r − z ) + 2(u − z ) + 4 (u − r)(r − z ))

1, 0≤x≤z

z (1 + 2 (r − x) + 2(u − x) + 4 (u − r)(r − x))



, z≤x≤r

x(1 + 2 (r − z ) + 2(u − z ) + 4 (u − r)(r − z ))



× z (1 + 2(u − x))
, r≤x≤u
x(1 + 2 (r − z ) + 2(u − z ) + 4 (u − r)(r − z ))



z



 , u≤x
x(1 + 2 (r − z ) + 2(u − z ) + 4 (u − r)(r − z ))

v(1 + 2 r + 2u + 4 (u − r)r)

 
u<z = exp −
2(z + 2 (z − r)r + 2(z − u)u + 4 (z − u)(u − r)r)

z
0≤x≤r

,

 x(z + 2 (z − r)r + 2(z − u)u + 4 (z − u)(u − r)r)
z (x + 2 (x − r)r)



, r≤x≤u

x(1 + 2 (z − r)r + 2(z − u)u + 4 (z − u)(u − r)r)


u<z ×
z (x + 2 (x − r)r + 2(x − u)u + 4 (x − u)(u − r)r)
, u≤x≤z

x(z + 2 (z − r)r + 2(z − u)u + 4 (z − u)(u − r)r)




z


, z≤x

x
510 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 % = %(v, z) = min{s : `(s, z) = v}

%
ds
n  Z  o
4.19.1 Ex exp −γ ; % < ∞
0 Rs(3)

 −1=2 √
x I1 (2 2 x) v
 
 z −1=2 I (2√2 z ) exp − 4zI (2√2 z )K (2√2 z ) , 0 < x ≤ z

1 1 1
= √
−1=2 K (2 2 x)
x 1 v
 
exp − , z≤x

√ √ √
z −1=2 K1 (2 2 z ) 4zI1 (2 2 z )K1 (2 2 z )


 2Z %
ds
n  o
4.20.1 Ex exp − (3)
2
; %<∞
2 0 Rs

√
2 + 1=4 
 
x −1/2+ γ 2 +1/4 v
 p
exp − , 0<x≤z

z z


=   √
v 2 + 1=4 
 x
 −1/2− γ 2 +1/4  p
exp − z≤x

,
z z

%
ds
Z 
4.20.4 Px
∈ dy
0 Rs(3) 2
0<y
 1=2
z (ln(z=x) + v=z ) −y/8 −(ln(z/x)+v/z)2 /2y
√ e e dy, 0 < x ≤ z
x1=2 2y3=2


= 1=2
 z (ln(√
 x=z ) + v=z ) −y/8 −(ln(x/z)+v/z)2 /2y
e e dy, z ≤ x
x1=2 2y3=2

 Z %
p2 q2 Rs(3) 2  
n
o
4.21.1 Ex exp − (3) 2
+ ds ; % < ∞
0 2 Rs


2
0<x
 √zI√ 2
4p2 +1=4 (qx =2) v
 
√ √ 2 exp − , x≤z
zI√4p2 +1=4 (qz 2 =2)K√4p2 +1=4 (qz 2 =2)

4p +1=4 (qz =2)
 xI 2


= √ √
 zK 4p2 +1=4 (qx2 =2) 
v

exp − √ , z≤x

√ √ 2 2 2
xK 4p2 +1=4 (qz =2) zI 4p2 +1=4 (qz =2)K 4p2 +1=4 (qz =2)

√

p2
%
q

n  Z  o
4.22.1 Ex exp − + ds ; % < ∞
2 Rs(3) 2 Rs(3)


0<x 0

 √zI√ √
4p 2 +1 (2 2qx) v
 
 √xI√
 √ exp − √ √
√ , x≤z
(2 2qz ) 4p2 +1 (2 2qz )K 4p2 +1 (2 2qz )
4zI √

4p2 +1

= √ √ √
 zK 4p2 +1 (2 2qx) 
v

√ exp − √ √ √ , z≤x

√ √
xK 4p2 +1 (2 2qz ) 4zI 4p2 +1 (2 2qz )K√4p2 +1 (2 2qz )

 4. STOPPING AT INVERSE LOCAL TIME 511
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 % = %(v, z) = min{s : `(s, z) = v}

n  Z %


 o
4.27.1 Ex exp − p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds ; `(%, r) ∈ dy
0

 v√p ch(z √2p) (y + v)√p ch((r − z )√2p) y√q 

z≤r = exp − √ − √ √ − √
sh(z 2p) 2 sh((r − z ) 2p) 2
√ √ √
z sh(x 2p) 2pv 2pvy
  
√ √ √ I √ dy, 0≤x≤z
2x sh(z 2p) sh((r − z ) 2p) y 1 sh((r − z ) 2p)



√ √ √

z 2p √ v 2pvy

 h  

 2x sh2 ((r − z )√2p) sh((r − x) 2p) √y I1 sh((r − z )√2p)


× √
√ 
2pvy i
+ sh((x − z) 2p)I0 √ dy, z≤x≤r

sh((r − z ) 2p)




√ √ √
 z 2pe−(x−r√ ) 2q 

2pvy

 
r≤x

I0 √ dy,
2x sh((r − z ) 2p) sh((r − z ) 2p)

 v√q (y + v)√q ch((z − r)√2q) y√p ch(r√2p) 

r≤z = exp − √ − √ √ − √ √
2 2 sh((z − r) 2q) 2 sh((r 2p)
√ √ √
z 2q sh(x 2p) 2qvy
  
√ √ I √ dy, 0≤x≤r
2x sh((z − r) 2q) sh(r 2p) 0
sh((z − r) 2q)



√ √

z 2q √ qv 2qvy

 h  

 2x sh2 ((z − r)√2q) sh((x − r) 2q) y I1 sh((z − r)√2q)


× √
√ 
2qvy i
 + sh((z − x) 2q)I √ dy, r≤x≤z
0
sh((z − r) 2q)




√ √
−(x−z ) 2q
√
 z 2qve √ √ I1 2qvy

  

dy, z≤x

√
2x sh((z − r) 2q) y sh((z − r) 2q)

n  Z %


 o
(1) Ex exp − p1I[0,r) Rs(3) + q1I[r,∞) Rs(3) ds ; `(%, r) = 0
0

 z sh(x√2p)
 x sh(z √2p) ,
 0≤x≤z
√ √
v 2p sh(r 2p)
 
 √
z≤r = exp − √ √ z sh((r − x) 2p)
2 sh(z 2p) sh((r − z ) 2p)  √ , z≤x≤r
 sh((r − z ) 2p)
x


0, r≤x

 0, 0≤x≤r

√ (z r ) 2q  z sh((x − r)√2q)
√
v 2qe
 − 
r≤z = exp − √ , r≤x≤z
√
2 sh((z − r) 2q)  x sh((z − r) 2q)
 z −(x−z) 2q
 √
e , z≤x
x
512 5. BESSEL PROCESS OF ORDER 1=2
P3
1/2
5 Rs(3) = k=1 (Ws
(k)
)2 % = %(v, z) = min{s : `(s, z) = v}

Z % %
1
Z 




4.27.2 Px 1I[0,r) Rs(3) ds ∈ du, 1I[r,∞) Rs(3) ds ∈ dg, ` %, r ∈ dy
0 0
2

√
z v √
x≤z = √ (is (1, r − z, 0, y + v, vy) ∗ essu (x, z, 0, v)) hg (1, y)dudgdy
x y u
z≤r
z √
z≤r = (isu (0, r − z, 0, y + v, vy) ∗ esu (0, 0, z, 0, v)) hg (1, x − r + y)dudgdy
x
r≤x
z √
x≤r = essu (x, r, 0, y) isg (0, z − r, v, y + v, vy)dudgdy
x
r≤z
√
z v √
r≤z = √ esu (0, 0, r, 0, y) isg (1, z − r, x − z + v, y + v, vy)dudgdy
x y
z≤x

Z %

Z %


(3)
(1) Px 1I[0,r) Rs ds ∈ du, 1I[r,∞) Rs(3) ds = 0, ` %, r = 0
0 0

( z
essu (x, z, 0, v/2) ∗ esu (0, 0, r − z, 0, v/2)du, 0≤x≤z
= x
z
es (0, 0, z, 0, v/2) ∗ essu (r − x, r − z, 0, v/2)du, z ≤ x ≤ r
x u

Z %

Z %


(2) Px 1I[0,r) Rs(3) ds = 0, 1I[r,∞) Rs(3) ds ∈ dg, ` %, r = 0
0 0

( z
essg (x − r, z − r, v/2, v/2)dg, r≤x≤z
= x
z
es (0, 0, z − r, x − z + v/2, v/2)dg, z≤x
x g

 Z % 2
p 2βRs(3)
n (3)
  o
4.30.1 Ex exp − e + qeβRs ds ; % < ∞
0
2
1 q 2p 
vU + ; 1;
2 p
 
β>0 = exp −  p  1 q 2p  1 q 2p 2p 
2 exp − e z U + ; 1; e z S + ; 1; e z ;
2 p 2 p
 1 2p x ; 2p
 p q
z exp − e x S + ; 1; e
 
2 p


 1 2p , 0≤x≤z
z ; 2p
 p q


x exp − e z S + ; 1; e


2 p


×  2p
ze− x=2 W−q=p ;0 e x




, z≤x

 2p

xe− z=2 W−q=p ;0 e z




1/2
6 Rs(2) = (Ws(1) )2 + (Ws(2) )2 Hz = min{s : Rs(2) = z}

6. BESSEL PROCESS OF ORDER ZERO

2. Stopping at first hitting time

 I (x√2 )
0
 I (z √2 ) ,
 x≤z
0
2.0.1 Ex e−αHz = √
 0 (x√2 ) ,
 K
z≤x
K0 (z 2 )

1, x≤z
(
ln(y=x)
(2)


2.1.2 Px sup Rs < y =
, z≤x≤y
z≤y 0≤s≤Hz
ln(y=z )
( ln(x=y)
, y≤x≤z
ln(z=y)


2.2.2 Px inf Rs(2) >y =
0≤s≤Hz
y≤z 1, z≤x

 1, x ≤ z ≤ r, r≤z≤x
 1 + 2 r| ln(r=x)|


Ex e−γ`(Hz ,r) = , r∧z ≤x≤r∨z
2.3.1 1 + 2 r| ln(r=z )|
1


, z ≤ r ≤ x, x≤r≤z

1 + 2 r| ln(r=z )|



2.3.2 Px `(Hz , r) ∈ dy

x ≤ z ≤ r, r≤z≤x

 0,
 | ln(x=z )| y

  
2 exp − dy, z∧r ≤x≤r∨z
= 2r ln (z=r) 2r| ln(z=r)|
1 y

  
exp − dy, z ≤ r ≤ x, x≤r≤z

2r| ln(z=r)| 2r| ln(z=r)|


1, x ≤ z ≤ r, r≤z≤x

| ln(r=x)|




(1) Px `(Hz , r) = 0 = , z∧r ≤x≤r∨z

 | ln(r=z )|
0, z ≤ r ≤ x, x≤r≤z

 Z Hz

2.4.1 Ex exp −γ 1I[r,∞) Rs(2) ds
0
 K0 (x√2 )
√ , z≤x
K0 (z 2 )



√ √
C0 (r 2 ; x 2 )


r≤z = √ √ , r≤x≤z
 C0 (r 2 ; z 2 )
1



 √ √ √ , 0≤x≤r
r 2 C0 (r 2 ; z 2 )

© Springer Basel 2015 513

A.N. Borodin, P. Salminen, Handbook of Brownian Motion - Facts and Formulae,
Probability and Its Applications, DOI 10.1007/978-3-0348-8163-0_13
514 6. BESSEL PROCESS OF ORDER ZERO

1/2
6 Rs(2) = (Ws(1) )2 + (Ws(2) )2 Hz = min{s : Rs(2) = z}

√
 K0 (x 2 )
√ √ √ , r≤x
 K0 (r√2 ) + ln(r=z )r √2 K1 (r √2 )



z≤r = K0 (r 2 ) + ln(r=x)r 2 K1 (r 2 )
√ √ √ , z≤x≤r
 K0 (r 2 ) + ln(r=z )r 2 K1 (r 2 )



1, 0≤x≤z

1, x≤z

Z Hz  
ln(r=x)


(2)
2.4.2 Px 1I[r,∞) Rs ds = 0 = , z≤x≤r
z≤r 0  ln(r=z )

0, r≤x

 Z Hz

2.5.1 Ex exp −γ 1I[0,r] Rs(2) ds
0

√
 I0 (x 2 )
√ √ √ , 0≤x≤r
 I0 (r√2 ) + ln(z=r)r √2 I1 (r √2 )



r≤z = I0 (r 2 ) + ln(x=r)r 2 I1 (r 2 )
√ √ √ , r≤x≤z
 I0 (r 2 ) + ln(z=r)r 2 I1 (r 2 )



1, z≤x

 I0 (x√2 )
√ , 0≤x≤z
I0 (z 2 )



√ √
C0 (r 2 ; x 2 )


z≤r = √ √ , z≤x≤r
 C0 (r 2 ; z 2 )


 1
 √ √ √ , r≤x
r 2 C0 (r 2 ; z 2 )

0, x≤r

Hz
ln(x=r)
Z  

(2)


2.5.2 Px 1I[0,r] Rs ds = 0 = , r≤x≤z
r≤z 0  ln(z=r)

1, z≤x

 Z Hz



2.6.1 Ex exp − p1I[0,r) Rs(2) + q1I[r,∞) Rs(2) ds
0
√ √
 (r 2)−1 I0 (x 2p)
√ √ √ √ √ √ √ √ , 0≤x≤r
pS0 (z 2q; r 2q)I1 (r 2p) + qC0 (r 2q; z 2q)I0 (r 2p)




√ √ √ √ √ √ √ √
pS (x 2q; r 2q)I1 (r 2p) + qC0 (r 2q; x 2q)I0 (r 2p)

r≤z = √ 0 √ √ √ √ √ √ √ , r≤x≤z
 pS0 (z 2q; r 2q)I1 (r 2p) + qC0 (r 2q; z 2q)I0 (r 2p)
 √
 K0 (x 2q)


√ , z≤x
K0 (z 2q)

 2. STOPPING AT FIRST HITTING TIME 515

1/2
6 Rs(2) = (Ws(1) )2 + (Ws(2) )2 Hz = min{s : Rs(2) = z}
 I (x√2p)
0
√ , 0≤x≤z
I0 (z 2p)



√ √ √ √ √ √ √ √

 pK (r 2q)C (r 2p; x 2p) + qK (r 2q)S (r 2p; x 2p)

0 0 1 0
z≤r = √pK (r√2q)C (r√2p; z √2p) + √qK (r√2q)S (r√2p; z √2p) , z ≤ x ≤ r
0 0 1 0
√

√
(r 2)−1 K0 (x 2q)



, r≤x

√ √ √ √ √ √ √ √
pK0 (r 2q)C0 (r 2p; z 2p) + qK1 (r 2q)S0 (r 2p; z 2p)

I0 ( 23 x3=2 2 )
 √
  I0 ( 2 z 3=2 2 ) , 0 ≤ x ≤ z
√

HzZ 
3

(2)
2.8.1 Ex exp −γ Rs ds =
K0 ( 23 x3=2 2 )
√
0 
, z≤x
K0 ( 23 z 3=2 2 )
√



I0 (x2 =2)

 2Z Hz  I (z 2 =2) ,
 x≤z
0

2.9.1 Ex exp − (Rs(2) )2 ds = 2
2 0  0 (x =2) ,
K
z≤x

2
K0 (z =2)

√ √
S0 (x 2 ; z 2 )
Ex e−γ Ȟ(Hz ) ; sup Rs(2) ∈ dy =

2.13.1 √ √ dy
0≤s≤Hz y ln(y=z )S0 (y 2 ; z 2 )
z<x
x<y

√ √
S0 (z 2 ; x 2 )
Ex e−γ Ĥ(Hz ) ;

2.14.1 inf Rs(2) ∈ dy = √ √ dy
0≤s≤Hz y ln(z=y)S0 (z 2 ; y 2 )
x<z
y<x

2.18.1 Ex exp −γ`(Hz , r) − η`(Hz , u)
r<u

 1, x≤z
1 + 2 r ln(r=x) + 2u ln(u=x) + 4 ru ln(u=r) ln(r=x)



 1 + 2 r ln(r=z ) + 2u ln(u=z ) + 4 ru ln(u=r) ln(r=z ) ,


 z≤x≤r
z<r = 1 + 2u ln(u=x)
, r≤x≤u
1 + 2 r ln(r=z ) + 2u ln(u=z ) + 4 ru ln(u=r) ln(r=z )



1



, u≤x

1 + 2 r ln(r=z ) + 2u ln(u=z ) + 4 ru ln(u=r) ln(r=z )

 1
, x≤r
1 + 2 r ln(z=r) + 2u ln(z=u) + 4 ru ln(z=u) ln(u=r)




1 + 2 r ln(x=r)


r≤x≤u

,
1 + 2 r ln(z=r) + 2u ln(z=u) + 4 ru ln(z=u) ln(u=r)

u<z =
 1 + 2 r ln(x=r) + 2u ln(x=u) + 4 ru ln(x=u) ln(u=r)
, u≤x≤z

1 + 2 r ln(z=r) + 2u ln(z=u) + 4 ru ln(z=u) ln(u=r)





1, z≤x

516 6. BESSEL PROCESS OF ORDER ZERO

1/2
6 Rs(2) = (Ws(1) )2 + (Ws(2) )2 Hz = min{s : Rs(2) = z}


1 1 
2.18.2 Px `(Hz , r) ∈ dy, `(Hz , u) ∈ dg
r u
r<u
 ln(x=z )
pr dydg, z≤x≤r
 ln(r=z )



z<r = ln(u=x) ln(x=r)
pr dydg + p dydg, r ≤ x ≤ u


 ln(u=r ) ln(u=r) u

pu dydg, u≤x

0≤x≤r

qr dydg,
 ln(u=x) ln(x=r)



q dydg + q dydg, r ≤ x ≤ u
u<z = ln(u=r) r ln(u=r) u
 ln(z=x) qu dydg,


u≤x≤z

ln(z=u)
where
√   √yg 
y y+g y

pr := exp −
√ − I
4 ln(u=r) ln(r=z ) g 2 ln(u=r) 2 ln(r=z ) 1 ln(u=r)
  √yg 
1 
y+g g
qr := exp − − I
4 ln(z=u) ln(u=r) 2 ln(u=r) 2 ln(z=u) 0 ln(u=r)
  √yg 
1 
y+g y
pu := exp − − I
4 ln(u=r) ln(r=z ) 2 ln(u=r) 2 ln(r=z ) 0 ln(u=r)
√   √yg 
g y+g g

qu := √ exp − − I
4 ln(z=u) ln(u=r) y 2 ln(u=r) 2 ln(z=u) 1 ln(u=r)


1 
(1) Px `(Hz , r) ∈ dy, `(Hz , u) = 0
r

ln(x=z ) y ln(u=z )
  
2 exp − dy, z≤x≤r
2 ln (r=z ) 2 ln(u=r) ln(r=z )


=
ln(u=x) 
y ln(u=z )

exp − dy, r ≤ x ≤ u

2 ln(u=r) ln(r=z ) 2 ln(u=r) ln(r=z )



1 
(2) Px `(Hz , r) = 0, `(Hz , u) ∈ dg
u

ln(x=r) g ln(z=r)
  
 2 ln(z=u) ln(u=r) exp − 2 ln(z=u) ln(u=r) dg,
 r≤x≤u
=
 ln(2z=x) exp − g ln(z=r)
 
 dg, u≤x≤z
2 ln (z=u) 2 ln(z=u) ln(u=r)
 2. STOPPING AT FIRST HITTING TIME 517

1/2
6 Rs(2) = (Ws(1) )2 + (Ws(2) )2 Hz = min{s : Rs(2) = z}

 ln(r=x)


 ln(r=z ) , z ≤ x ≤ r
(3) Px `(Hz , r) = 0, `(Hz , u) = 0 =
 ln(x=u)
, u≤x≤z
ln(z=u)

√
I0 (2 2 x)

Z Hz  I (2√2 z ) , 0 < x ≤ z
ds
 
0

2.19.1 Ex exp −γ = √
R  0 (2√2 x) , z ≤ x
K
(2)
0 s 
K0 (2 2 z )

x
  |γ|
 2Z Hz
ds , 0<x≤z
z
 

2.20.1 Ex exp − =  |γ|
2 Rs(2) 2  z


0 
, z≤x
x


ln(z=x) 2
 √ 3=2 e− ln (z/x)/2y dy, 0 < x ≤ z
Hz
ds 2y
Z  
2.20.4 Px
∈ dy =
0 Rs(2) 2  √ln(x=z ) e− ln2 (x/z)/2y dy, z ≤ x
0<y

2y3=2

2
 I|p|=2 (qx =2) , 0 < x ≤ z

 Z Hz 
p 2 q 2     I
|p|=2 ( qz 2 =2)
2.21.1 Ex exp − (2)
2
+ (Rs(2) )2 ds = 2
2 Rs 2 K (qx =2)
0  |p|=2 2 , z ≤ x

K|p|=2 (qz =2)


p2 q2
 Z Hz   
2.21.3 Ex exp −αHz − (2)
2
+ (Rs(2) )2 ds
0 2 Rs 2

2
zM− =2q;|p|=2 (qx )


2 , 0<x≤z
 xM (qz )

− =2q;|p|=2
= 2
zW − =2q;|p|=2 (qx )
z≤x

,
=2q;|p|=2 (qz 2 )

xW−


√
I2|p| (2 2qx)

 I (2√2qz ) ,
 0<x≤z
p2
Hz 
q

2|p|
 Z  
2.22.1 Ex exp − 2 + ds = √
2 Rs R (2)
K (2 2qx)
s
 2|p| √
(2)

0
z≤x

,
K2|p| (2 2qz )


√ √
zM−q=√2 ;|p| (2 x 2 )
√ , 0<x

√
xM−q=√2 ;|p| (2z 2 ) x≤z

p2
Z Hz
q
   
2.22.3 Ex exp −αHz −
+ ds = √
2 Rs(2) 2 Rs(2)
√
0  zW−q=√2 ;|p| (2x 2 )
√ , z≤x

√
xW−q=√2 ;|p| (2z 2 )

518 6. BESSEL PROCESS OF ORDER ZERO

1/2
6 Rs(2) = (Ws(1) )2 + (Ws(2) )2 H = Ha,b = min{s : Rs(2) ∈
/ (a, b)}

3. Stopping at first exit time

√ √ √ √
S (b 2 ; x 2 ) + S0 (x 2 ; a 2 )
3.0.1 Ex e −αH
= 0 √ √
S0 (b 2 ; a 2 )

(2) ln(b=x) −βa ln(x=a) −βb

3.0.3 Ex e−βRH = e + e
ln(b=a) ln(b=a)

(2)

ln(b=x)
3.0.4 Px R H =a =
ln(b=a)
(a)
(2)

ln(x=a)
3.0.4 Px R H =b =
ln(b=a)
(b)

(2)

ln(b=y) ln(x=a)
3.1.6 Px sup Rs(2) ≥ y, RH =a = , a≤x≤y≤b
0≤s≤H ln(b=a) ln(y=a)

(2)

ln(b=x) ln(y=a)
3.2.6 Px inf Rs(2) ≤ y, RH =b = , a≤y≤x≤b
0≤s≤H ln(b=y) ln(b=a)

ln(b=a) + 2 r ln(x=r) ln(r=a)


 ln(b=a) + 2 , r≤x≤b
r ln(b=r) ln(r=a)

3.3.1 Ex e−γ`(H,r) =
 ln(b=a) + 2
 r ln(b=r) ln(r=x)
, a≤x≤r
ln(b=a) + 2 r ln(b=r) ln(r=a)


ln(b=x) ln(b=a)
 2r ln2 (b=r) ln(r=a) dy, r ≤ x ≤ b


ln(b=a)y 
3.3.2 Px `(H, r) ∈ dy = exp −
2r ln(b=r) ln(r=a) 
 ln(x=a) ln(2b=a) dy, a ≤ x ≤ r
2r ln(b=r) ln (r=a)

 ln(x=r)


 ln(b=r) , r ≤ x ≤ b
(1) Px `(H, r) = 0 =
 ln(r=x) , a ≤ x ≤ r

ln(r=a)

ln(b=x)

 ln(b=a) + 2 r ln(b=r) ln(r=a) , r ≤ x ≤ b

Ex e−γ`(H,r) ; RH
 (2)
3.3.5 =a =
(a)  ln(b=x) + 2 r ln(b=r) ln(r=x) , a ≤ x ≤ r

ln(b=a) + 2 r ln(b=r) ln(r=a)
 3. STOPPING AT FIRST EXIT TIME 519

1/2
6 Rs(2) = (Ws(1) )2 + (Ws(2) )2 H = Ha,b = min{s : Rs(2) ∈
/ (a, b)}

ln(x=a) + 2 r ln(x=r) ln(r=a)


 −γ`(H,r)  ln(b=a) + 2 r ln(b=r) ln(r=a) , r ≤ x ≤ b

(2)
3.3.5 Ex e ; RH = b =
ln(x=a)
(b) , a≤x≤r

ln(b=a) + 2 r ln(b=r) ln(r=a)


(2)


3.3.6 Px `(H, r) ∈ dy, RH =a
(a)
ln(b=x)

 2r ln(b=r) ln(r=a) dy, r ≤ x ≤ b

ln(b=a)y 
= exp −
 ln(2x=a) dy,
2r ln(b=r) ln(r=a) 
a≤x≤r
2r ln (r=a)

0, r≤x≤b
(
ln(r=x)
(2)


(1) Px `(H, r) = 0, RH =a =
, a≤x≤r
ln(r=a)

(2)


3.3.6 Px `(H, r) ∈ dy, RH =b
(b)
ln(b=x)

dy, r≤x≤b

ln(b=a)y 

2r ln2 (b=r)
= exp −
2r ln(b=r) ln(r=a)  ln(x=a)
dy, a ≤ x ≤ r
2r ln(b=r) ln(r=a)


( ln(x=r)
, r≤x≤b
ln(b=r)
(2)


(1) Px `(H, r) = 0, RH =b =
0, a≤x≤r

 Z H

3.4.1 Er exp −γ 1I[r,b] Rs(2) ds
0

√ √
S0 (b 2 ; r 2 ) + ln(r=a)
= √ √ √ √ √ =: Q
S0 (b 2 ; r 2 ) + ln(r=a)r 2 C0 (r 2 ; b 2 )

 Z H

Ex exp −γ 1I[r,b] Rs(2) ds
0


ln(r=x) ln(x=a)
 ln(r=a) + ln(r=a) Q,
 a≤x≤r
= √ √ √ √
 S0 (x√ 2 ; r√ 2 ) + S0 (b √2 ; x√ 2 ) Q, r ≤ x ≤ b

S0 (b 2 ; r 2 ) S0 (b 2 ; r 2 )
520 6. BESSEL PROCESS OF ORDER ZERO

1/2
6 Rs(2) = (Ws(1) )2 + (Ws(2) )2 H = Ha,b = min{s : Rs(2) ∈
/ (a, b)}

( ln(r=x)
H
Z , a≤x≤r
ln(r=a)


(2)
3.4.2 Px 1I[r,b] Rs ds = 0 =
(1) 0 0, r≤x≤b

n  Z H
 o
(2)
3.4.5 Ex exp −γ 1I[r,b] Rs(2) ds ; RH =a
(a) 0

√ √ √ √ √
S0 (b 2 ; r 2 ) + ln(r=x)r 2 C0 (r 2 ; b 2 )

 S (b 2 ; r√2
√ √ √ √ , a≤x≤r
) + ln(r=a)r 2 C0 (r 2 ; b 2 )

0
= √ √
S0 (b 2 ; x 2 )
√ √ √ √ √ , r≤x≤b

S0 (b 2 ; r 2 ) + ln(r=a)r 2 C0 (r 2 ; b 2 )


n  Z H
 o
(2)
3.4.5 Ex exp −γ 1I[r,b] Rs(2) ds ; RH =b
(b) 0


ln(x=a)
 S (b√2 ; r√2 ) + ln(r=a)r√2 C (r√2 ; b√2 ) , a ≤ x ≤ r

0 0
= √ √ √ √ √
S0 ( x 2 ; r 2 ) + ln( r=a )r 2 C 0 (r 2 ; x 2 )
√ √ √ √ √ , r≤x≤b

S0 (b 2 ; r 2 ) + ln(r=a)r 2 C0 (r 2 ; b 2 )


( ln(r=x)
H
Z , a≤x≤r
ln(r=a)


(2) (2)
3.4.6 Px 1I[r,b] Rs ds = 0, RH =a =
(1) 0 0, r≤x≤b

 Z H

3.5.1 Er exp −γ 1I[a,r] Rs(2) ds
0

√ √
S0 (r 2 ; a 2 ) + ln(b=r)
= √ √ √ √ √ =: Q
S0 (r 2 ; a 2 ) + ln(b=r)r 2 C0 (r 2 ; a 2 )

 Z H

Ex exp −γ 1I[a,r] Rs(2) ds
0

√ √ √ √
S0 (r 2 ; x 2 ) S0 (x 2 ; a 2 )

√ √ + √ √ Q, a ≤ x ≤ r
S0 (r 2 ; a 2 ) S0 (r 2 ; a 2 )


=
 ln(x=r) + ln(b=x) Q,

r≤x≤b
ln(b=r) ln(b=r)
 3. STOPPING AT FIRST EXIT TIME 521

1/2
6 Rs(2) = (Ws(1) )2 + (Ws(2) )2 H = Ha,b = min{s : Rs(2) ∈
/ (a, b)}

0, a≤x≤r
(
Z H 
ln(x=r)
(2)


3.5.2 Px 1I[a,r] Rs ds = 0 =
, r≤x≤b
(1) 0 ln(b=r)

n  Z H
 o
(2)
3.5.5 Ex exp −γ 1I[a,r] Rs(2) ds ; RH =a
(a) 0

√ √ √ √ √
S0 (r 2 ; x 2 ) + ln(b=r)r 2 C0 (r 2 ; x 2 )

√ √ √ √ √ , a≤x≤r
S0 (r 2 ; a 2 ) + ln(b=r)r 2 C0 (r 2 ; a 2 )


=
ln(b=x)
√ √ √ √ √ , r≤x≤b

S0 (r 2 ; a 2 ) + ln(b=r)r 2 C0 (r 2 ; a 2 )


n  Z H
 o
(2)
3.5.5 Ex exp −γ 1I[a,r] Rs(2) ds ; RH =b
(b) 0
√ √
S0 (x 2 ; a 2 )

 S (r √ 2 ; a √ 2 √ √ √ , a≤x≤r
) + ln(b=r)r 2 C0 (r 2 ; a 2 )

0
= √ √ √ √ √
 S0 (r 2 ; a 2 ) + ln(x=r)r 2 C0 (r 2 ; a 2 )
√ √ √ √ √ , r≤x≤b
S0 (r 2 ; a 2 ) + ln(b=r)r 2 C0 (r 2 ; a 2 )


0, a≤x≤r
(
Z H 
ln(x=r)
(2)

(2)
3.5.6 Px 1I[a,r] Rs ds = 0, RH = b =
, r≤x≤b
(1) 0 ln(b=r)

H o S ( 2 b3=2 √2 ; 23 x3=2 2 )
√
0 3
n  Z 
(2)
3.8.5 Ex exp −γ Rs(2) ds ; RH =a =
S0 ( 32 b3=2 2 ; 23 a3=2 2 )
√ √
(a) 0

S0 ( 23 x3=2 2 ; 23 a3=2 2 )
n  Z H  o √ √
(2) (2)
3.8.5 Ex exp −γ Rs ds ; RH =b =
S0 ( 23 b3=2 2 ; 23 a3=2 2 )
√ √
(b) 0

 2Z
S (b2 =2; x2 =2)
H
=a = 0 2
n  o
3.9.5 Ex exp − (Rs(2) )2 ds ; RH
(2)

2 0
S0 (b =2; a2 =2)
(a)

 2Z
S (x2 =2; a2 =2)
H
=b = 0 2
n  o
3.9.5 Ex exp − (Rs(2) )2 ds ; RH
(2)

2 0
S0 (b =2; a2 =2)
(b)
√ √
S0 (x 2 ; a 2 )
Z b
−γ Ȟ(H)
 (2)
3.12.5 Ex e ; RH = a = √ √ dy
x
y ln(y=a)S0 (y 2 ; a 2 )
(a)
522 6. BESSEL PROCESS OF ORDER ZERO

1/2
6 Rs(2) = (Ws(1) )2 + (Ws(2) )2 H = Ha,b = min{s : Rs(2) ∈
/ (a, b)}

√ √
S0 (b 2 ; x 2 )
Z x
Ex e−γ Ĥ(H) ; RH
 (2)
3.12.5 =b = √ √ dy
a
y ln(b=y)S0 (b 2 ; y 2 )
(b)

3.18.1 Ex exp −γ`(H, r) − η`(H, u)
r≤u
ln(b=a) + 2 r ln(b=r) ln(r=x) + 2u ln(b=u) ln(u=x) + 4 ru ln(b=u) ln(u=r) ln(r=x)
a≤x =
ln(b=a) + 2 r ln(b=r) ln(r=a) + 2u ln(b=u) ln(u=a) + 4 ru ln(b=u) ln(u=r) ln(r=a)
x≤r

ln(b=a) + 2 r ln(x=r) ln(r=a) + 2u ln(b=u) ln(u=x)

r≤x =
ln(b=a) + 2 r ln(b=r) ln(r=a) + 2u ln(b=u) ln(u=a) + 4 ru ln(b=u) ln(u=r) ln(r=a)
x≤u

ln(b=a) + 2 r ln(x=r) ln(r=a) + 2u ln(x=u) ln(u=a) + 4 ru ln(x=u) ln(u=r) ln(r=a)
u≤x =
ln(b=a) + 2 r ln(b=r) ln(r=a) + 2u ln(b=u) ln(u=a) + 4 ru ln(b=u) ln(u=r) ln(r=a)
x≤b

(2)
3.18.5 Ex exp −γ`(H, r) − η`(H, u) ; RH =a
(a)
ln(b=x) + 2 r ln(b=r) ln(r=x) + 2u ln(b=u) ln(u=x) + 4 ru ln(b=u) ln(u=r) ln(r=x)
a≤x =
ln(b=a) + 2 r ln(b=r) ln(r=a) + 2u ln(b=u) ln(u=a) + 4 ru ln(b=u) ln(u=r) ln(r=a)
x≤r

ln(b=x) + 2u ln(b=u) ln(u=x)

r≤x =
ln(b=a) + 2 r ln(b=r) ln(r=a) + 2u ln(b=u) ln(u=a) + 4 ru ln(b=u) ln(u=r) ln(r=a)
x≤u

ln(b=x)
u≤x =
ln(b=a) + 2 r ln(b=r) ln(r=a) + 2u ln(b=u) ln(u=a) + 4 ru ln(b=u) ln(u=r) ln(r=a)
x≤b

(2)
3.18.5 Ex exp −γ`(H, r) − η`(H, u) ; RH =b
(b)
ln(x=a)
a≤x =
ln(b=a) + 2 r ln(b=r) ln(r=a) + 2u ln(b=u) ln(u=a) + 4 ru ln(b=u) ln(u=r) ln(r=a)
x≤r

ln(x=a) + 2 r ln(x=r) ln(r=a)

r≤x =
ln(b=a) + 2 r ln(b=r) ln(r=a) + 2u ln(b=u) ln(u=a) + 4 ru ln(b=u) ln(u=r) ln(r=a)
x≤u

ln(x=a) + 2 r ln(x=r) ln(r=a) + 2u ln(x=u) ln(u=a) + 4 ru ln(x=u) ln(u=r) ln(r=a)
u≤x =
ln(b=a) + 2 r ln(b=r) ln(r=a) + 2u ln(b=u) ln(u=a) + 4 ru ln(b=u) ln(u=r) ln(r=a)
x≤b
 3. STOPPING AT FIRST EXIT TIME 523

1/2
6 Rs(2) = (Ws(1) )2 + (Ws(2) )2 H = Ha,b = min{s : Rs(2) ∈
/ (a, b)}


1 1 (2)

3.18.6 Px `(H, r) ∈ dy, `(H, u) ∈ dg, RH =a
r u
(a)
 ln(x=a)
p dydg, a≤x≤r


 ln(r=a) r
ln(u=x) ln(x=r)


= pr dydg + p dydg, r ≤ x ≤ u
 ln(u=r ) ln(u=r) u
 ln(b=x) p dydg,



u≤x≤b
ln(b=u) u
where
√   √yg 
y y+g y g

pr := exp −√ − − I
4 ln(u=r) ln(r=a) g 2 ln(u=r) 2 ln(r=a) 2 ln(b=u) 1 ln(u=r)
  √yg 
1 
y+g y g
pu := exp − − − I
4 ln(u=r) ln(r=a) 2 ln(u=r) 2 ln(r=a) 2 ln(b=u) 0 ln(u=r)


1 (2)

(1) Px `(H, r) ∈ dy, `(H, u) = 0, RH =a
r

ln(x=a) y ln(u=a)
  
exp − dy, a≤x≤r
2 ln2 (r=a) 2 ln(u=r) ln(r=a)


=
ln(u=x) 
y ln(u=a)

exp − dy, r ≤ x ≤ u

2 ln(u=r) ln(r=a) 2 ln(u=r) ln(r=a)


(2)

ln(r=x)
(2) Px `(H, r) = 0, `(H, u) = 0, RH =a = , a≤x≤r
ln(r=a)


1 1 (2)

3.18.6 Px `(H, r) ∈ dy, `(H, u) ∈ dg, RH =b
r u
(b)
 ln(x=a)
q dydg, a≤x≤r


 ln(r=a) r
ln(u=x) ln(x=r)


= qr dydg + q dydg, r ≤ x ≤ u
 ln(u=r ) ln(u=r) u
 ln(b=x) q dydg,



u≤x≤b
ln(b=u) u
where

√
1 
y+g y g
  yg 
qr := exp − − − I
4 ln(b=u) ln(u=r) 2 ln(u=r) 2 ln(r=a) 2 ln(b=u) 0 ln(u=r)
524 6. BESSEL PROCESS OF ORDER ZERO

1/2
6 Rs(2) = (Ws(1) )2 + (Ws(2) )2 H = Ha,b = min{s : Rs(2) ∈ / (a, b)}
√ √
g y+g y g yg 
  
qu := √ exp − − − I1
4 ln(b=u) ln(u=r) y 2 ln(u=r) 2 ln(r=a) 2 ln(b=u) ln(u=r)


1 (2)

(1) Px `(H, r) = 0, `(H, u) ∈ dg, RH =b
u

ln(x=r) g ln(b=r)
  
 2 ln(b=u) ln(u=r) exp − 2 ln(b=u) ln(u=r) dg,
 r≤x≤u
=
 ln(2b=x) exp − g ln(b=r)
 
 dg, u≤x≤b
2 ln (b=u) 2 ln(b=u) ln(u=r)

(2)

ln(x=u)
(2) Px `(H, r) = 0, `(H, u) = 0, RH =b = , u≤x≤b
ln(b=u)

√ √
S0 (2 2 b; 2 2 x)
H
ds
n  Z  o
(2)
3.19.5 Ex exp −γ ; R = a = √ √
0 Rs(2) H S0 (2 2 b; 2 2 a)
(a)

√ √
S (2 2 x; 2 2 a)
H
ds
Z
=b = 0 √
n   o
(2)
3.19.5 Ex exp −γ ; RH √
0 Rs(2)
S0 (2 2 b; 2 2 a)
(b)

 2Z
n H
ds

(2)
o
(b=x)| | − (x=b)| |
3.20.5 Ex exp − (2) 2
; R = a = , a≤x≤b
2 0 Rs

H (b=a)| | − (a=b)| |
(a)

 2Z
n H
ds

(2)
o
(x=a)| | − (a=x)| |
3.20.5 Ex exp − (2)
2
; RH =b = , a≤x≤b
2 0 Rs (b=a)| | − (a=b)| |
(b)

 Z H
p2 qRs(2) 2   o x|p| S 2 2
|p|=2 (qb =2; qx =2)
n

(2)
3.21.5 Ex exp − (2) 2
+ ds ; R = a = |p| 2 2
0 2 Rs


2 H a S|p|=2 (qb =2; qa =2)
(a)

 Z H
p2 qRs(2) 2   o x|p| S 2 2
|p|=2 (qx =2; qa =2)
n

(2)
3.21.5 Ex exp − (2)
2
+ ds ; RH = b = |p|
0 2 Rs 2 b S|p|=2 (qb2 =2; qa2 =2)
(b)

o x|p| S (2√2qb; 2√2qx)

p2
H
q
 Z
2|p|
n  
(2)
3.22.5 Ex exp − + ds ; R = a = |p| √ √
2 Rs(2) 2 Rs(2) H a S2|p| (2 2qb; 2 2qa)


(a) 0

o x|p| S (2√2qx; 2√2qa)

p2
H
q
 Z
2|p|
n  
(2)
3.22.5 Ex exp − + ds ; R = b = |p| √ √
2 Rs(2) 2 Rs(2) H b S2|p| (2 2qb; 2 2qa)


(b) 0
 4. STOPPING AT INVERSE LOCAL TIME 525

1/2
6 Rs(2) = (Ws(1) )2 + (Ws(2) )2 % = %(v, z) = min{s : `(s, z) = v}

4. Stopping at inverse local time

 I (x√2 ) v
0
 
 I (z √2 ) exp − 2zI (z √2 )K (z √2 ) , x ≤ z

0 0 0
4.0.1 Ex e−α% = √
K 0 ( x 2 ) 
v

exp − , z≤x

√ √ √
K0 (z 2 ) 2zI0 (z 2 )K0 (z 2 )


v
  

 exp − 2z ln(y=z ) ,
 x≤z
4.1.2 Px sup Rs(2) < y =
 ln(y=x) exp − v
 
0≤s≤% , z≤x≤y
z≤y

ln(y=z ) 2z ln(y=z )
ln(x=y) v
  
exp − , y≤x≤z
ln(z=y) 2z ln(z=y)




4.2.2 Px inf Rs(2) > y =
v
 
0≤s≤%  exp −
y≤z
 , z≤x
2z ln(z=y)

4.3.1 Ex e−γ`(%,r)
rv
  
exp − , x ≤ z ≤ r, r≤z≤x
z (1 + 2 r| ln(r=z )|)



1 + 2 r| ln(r=x)| rv

  
= exp − , r∧z ≤x≤r∨z
 1 + 2 r| ln(r=z )| z (1 + 2 r| ln(r=z )|)
1 rv

  

 exp − , z ≤ r ≤ x, x ≤ r ≤ z
1 + 2 r| ln(r=z )| z (1 + 2 r| ln(r=z )|)

4.3.2 P `(%, r) ∈ dy
√ √
 1 v

v=z + y=r
  vy 
√ exp − I1 √ dy, r≤z≤x
2| ln(r=z )| yrz 2| ln(r=z )| | ln(r=z )| rz

x≤z≤r



√ √

1 v=z + y=r v vy

  h  

exp − | ln(r/x)| √ I1 √
2 ln2 (r=z )

2| ln(r=z )| yrz | ln(r=z )| rz

= √
 1  vy i
+| ln(x/z)| I0 √ dy, z∧r ≤x≤r∨z

r | ln(r=z )| rz



√

1 v=z + y=r vy

    

 exp − I √ dy, z ≤ r ≤ x, x ≤ r ≤ z
2r| ln(r=z )| 2| ln(r=z )| 0 | ln(r=z )| rz

v
  
exp − , x ≤ z ≤ r, r ≤ z ≤ x
2z | ln(r=z )|





| ln(r=x)| v
 
(1) Px `(%, r) = 0 = exp − , z∧r ≤x≤r∨z


 | ln(r=z )| 2 z | ln( r=z )|
z ≤ r ≤ x, x ≤ r ≤ z

0,
 Z % 


4.4.1 Ex exp −γ 1I[r,∞) Rs(2) ds
0
526 6. BESSEL PROCESS OF ORDER ZERO

1/2
6 Rs(2) = (Ws(1) )2 + (Ws(2) )2 % = %(v, z) = min{s : `(s, z) = v}
√ √
K0 (x 2
) )v K1 (r 2
  
√exp − √ √ √ , z≤x
K0 (z 2 ) 2zC0 (r 2 ; z 2 )K0 (z 2 )


 
√ √ √
C0 (r 2 ; x 2 ) K1 (r 2 )v

  
r≤z = √ √ exp − √ √ √ , r≤x≤z
 C0 (r 2 ; z 2 ) 2zC0 (r 2 ; z 2 )K0 (z 2 )
√
1 K1 (r 2 )v

  

 √ √ √ exp − √ √ √ , 0≤x≤r

r 2 C0 (r 2 ; z 2 ) 2zC0 (r 2 ; z 2 )K0 (z 2 )
√ √
r 2 K1 (r 2 )v
 
z≤r = exp − √ √ √
2z (K0 (r 2 ) + ln(r=z )r 2 K1 (r 2 ))
√
 K0 (x 2 )
√ √ √ , r≤x
 K0 (r√2 ) + ln(r=z )r √2 K1 (r √2 )



× K0 (r √2 ) + ln(r=x)r√2 K1 (r√2 )
, z≤x≤r
 K0 (r 2 ) + ln(r=z )r 2 K1 (r 2 )



1, 0≤x≤z

v
  
exp − , x≤z
Z% 2z ln(r=z )


 
ln(r=x) v


Rs(2) ds = 0 =
 
4.4.2 Px 1I[r,∞) exp − , z≤x≤r
z≤r 0


 ln(r=z ) 2 z ln( r=z )
r≤x

0,

 Z %

4.5.1 Ex exp −γ 1I[0,r] Rs(2) ds
0
√ √
r 2 I1 (r 2 )v
 
r≤z = exp − √ √ √
2z (I0 (r 2 ) + ln(z=r)r 2 I1 (r 2 ))
√
 I0 (x 2 )
√ √ √ , 0≤x≤r
I0 (r 2 ) + ln(z=r)r 2 I1 (r 2 )



√ √ √

× I0 (r 2 ) + ln(x=r)r 2 I1 (r 2 )
√ √ √ , r≤x≤z



 I0 (r 2 ) + ln(z=r)r 2 I1 (r 2 )
1, z≤x
√ √
I0 (x 2 ) I1 (r 2 )v
  
√ exp − √ √ √ , 0≤x≤z
I0 (z 2 ) 2zC0 (r 2 ; z 2 )I0 (z 2 )




√ √ √
C0 (r 2 ; x 2 ) I1 (r 2 )v

  
z≤r = √ √ exp − √ √ √ , z≤x≤r
 C0 (r 2 ; z 2 ) 2zC0 (r 2 ; z 2 )I0 (z 2 )
√
1 I1 (r 2 )v

  

 √ √ √ exp − √ √ √ , r≤x

r 2 C0 (r 2 ; z 2 ) 2zC0 (r 2 ; z 2 )I0 (z 2 )

x≤r

0,
Z%

 ln(x=r) v
  
 
exp − , r≤x≤z

ln(z=r) 2z ln(z=r)


4.5.2 Px 1I[0,r] Rs(2) ds = 0 =
r≤z v

  
0  exp − , z≤x

2z ln(z=r)
 4. STOPPING AT INVERSE LOCAL TIME 527

1/2
6 Rs(2) = (Ws(1) )2 + (Ws(2) )2 % = %(v, z) = min{s : `(s, z) = v}


4.18.1 Ex exp −γ`(%, r) − η`(%, u) r<u

( r + u + 2 ru ln(u=r))v 
z<r = exp −
z (1 + 2 r ln(r=z ) + 2u ln(u=z ) + 4 ru ln(u=r) ln(r=z ))

 1,
 x≤z

 1 + 2 r ln(r=x ) + 2 u ln(u=x ) + 4 ru ln(u=r ) ln(r=x )
 1 + 2 r ln(r=z ) + 2u ln(u=z ) + 4 ru ln(u=r) ln(r=z ) , z ≤ x ≤ r



× 1 + 2u ln(u=x)
, r≤x≤u
1 + 2 r ln(r=z ) + 2 u ln(u=z ) + 4 ru ln(u=r) ln(r=z )



1



, u≤x

1 + 2 r ln(r=z ) + 2u ln(u=z ) + 4 ru ln(u=r) ln(r=z )



( r + u + 2 ru ln(u=r))v 
u<z = exp −
z (1 + 2 r ln(z=r) + 2u ln(z=u) + 4 ru ln(z=u) ln(u=r))
 1
, x≤r
1 + 2 r ln(z=r ) + 2 u ln(z=u ) + 4 ru ln(z=u) ln(u=r)




1 + 2 r ln(x=r)


, r≤x≤u

× 1 + 2 r ln(z=r) + 2u ln(z=u) + 4 ru ln(z=u) ln(u=r)

 1 + 2 r ln(x=r) + 2u ln(x=u) + 4 ru ln(x=u) ln(u=r)
, u≤x≤z

1 + 2 r ln(z=r) + 2u ln(z=u) + 4 ru ln(z=u) ln(u=r)





1, z≤x


 2Z %
x|γ| z−|γ| e−v|γ|/z , 0 < x ≤ z
ds
 
4.20.1 Ex exp − =
2 Rs(2) 2 z |γ| x−|γ| e−v|γ|/z ,


0 z≤x

(ln(z=x) + v=z ) −(ln(z/x)+v/z)2 /2y
√ e dy, 0 < x ≤ z
Z %
ds
 
2y3=2

4.20.4 Px ∈ dy =
Rs(2) 2  (ln(√ x=z ) + v=z ) −(ln(x/z)+v/z)2 /2y


0<y 0 e dy, z ≤ x

2y3=2
p2 q2 (Rs(2) )2
 Z %  
4.21.1 Ex exp − (2)
2
+ ds
0 2 Rs 2
0<x
I|p|=2 (qx2 =2) v
  

2 exp − 2 2 , x≤z
|p|=2 (qz =2) zI|p|=2 (qz =2)K|p|=2 (qz =2)
I

=
K (qx2 =2) v
 |p|=2 2
 
exp − , z≤x

K|p|=2 (qz =2) zI|p|=2 (qz 2 =2)K|p|=2 (qz 2 =2)


p2
 Z %
q
 
4.22.1 Ex exp − (2)
2
+ (2) ds
0<x 0 2 Rs Rs
√
I2|p| (2 2qx) v
  
 I (2√2qz ) exp − 4zI (2√2qz )K (2√2qz ) , x ≤ z


2|p| 2|p| 2|p|
= √
K (2 2qx ) v
 2|p| √
 
exp − , z≤x

√ √
K2|p| (2 2qz ) 4zI2|p| (2 2qz )K2|p| (2 2qz )

7 Us = σe−θs We2θs −1 σ, θ > 0 τ ∼ Exp(λ), independent of U

7. ORNSTEIN–UHLENBECK PROCESS

1. Exponential stopping

 (=) (x2 −z2 )/4σ2


√ e D−λ/θ (− σx )D−λ/θ ( σz )dz, x ≤ z
 2


1.0.5 Px (Uτ ∈ dz) =
  (√
=) (x2 −z2 )/4σ2
e D−λ/θ ( σx )D−λ/θ (− σz )dz, z ≤ x
 2


 (=) (x2 −z2 )/4σ2

D−λ/θ − z+x−|z−x| z+x+|z−x|



or = √ e D−λ/θ dz
 2 2σ 2σ

1 
(z − xe−t )2 
1.0.6 Px Ut ∈ dz = exp − dz
(22 (1 − e−2t ))1=2 22 (1 − e−2t )

ex =4 D−= (− x )
2 2

1.1.2 Px sup Us ≥ y = y2 =42

0≤s≤τ e D−= (− y )
x≤y
|x|

 
1.1.4 Px sup Us < 0 = Erf x≤0
 2(e2t − 1)
p
0≤s≤t
(1)


1.1.6 Px sup Us ≥ y, Uτ ∈ dz x∨z ≤y
0≤s≤τ

 (=) 2 2 2
= √ e(x −z )/4σ D−λ/θ (− σx )D−λ/θ (− σz )D−λ/θ ( σy )dz
 2D−= (− y )

1 
(|z | − xe−t )2 
1.1.8 Px sup Us ≥ 0, Ut ∈ dz = exp − dz
0≤s≤t (22 (1 − e−2t ))1=2 22 (1 − e−2t )
x≤0

ex =4 D−= ( x )
2 2

1.2.2 Px inf Us ≤ y = y2 =42

0≤s≤τ e D−= ( y )
y≤x
x

 
1.2.4 Px 0 < inf Us = Erf 0≤x
 2(e2t − 1)
p
0≤s≤t
(1)


1.2.6 Px inf Us ≤ y, Uτ ∈ dz y ≤x∧z
0≤s≤τ

 (=) 2 2 2
= √ y e(x −z )/4σ D−λ/θ ( σx )D−λ/θ ( σz )D−λ/θ (− σy )dz
 2D−= (  )

1 
(|z | + xe−t )2 
1.2.8 Px inf Us ≤ 0, Ut ∈ dz = exp − 2 dz
0≤s≤t 2
(2 (1 − e− 2t ))1= 2 2 (1 − e−2t )
0≤x

© Springer Basel 2015 528

A.N. Borodin, P. Salminen, Handbook of Brownian Motion - Facts and Formulae,
Probability and Its Applications, DOI 10.1007/978-3-0348-8163-0_14
 1. EXPONENTIAL STOPPING 529

7 Us = σe−θs We2θs −1 σ, θ > 0 τ ∼ Exp(λ), independent of U

(=)e(x −r )=4 D−= (− x )D−= ( r )

 2 2 2

 1−
√
 2 + (=)D−= (− r )D−= ( r )
, x≤r


1.3.1 Ex e−γ`(τ,r) =
 1 − (= )e(x −r )=4 D−= ( x )D−= (− r )
2 2 2

√ , r≤x

 2 + (=)D−= (− r )D−= ( r )
√
 2y

 
1.3.2 Px `(τ, r) ∈ dy = exp −
(=)D−= (− r )D−= ( r )
 √ (x2 −r2 )=42
 2e D−= (− x )
 (=)D2 (− r )D−= ( r ) dy, x ≤ r



−=  
× √ x
  2e (x 2
− r 2
)= 4 2
D−= (  )
r dy, r≤x
(=)D−= (− r )D−

2
= (  )



ex =4 D−= (− x )
 2 2

1 − r2 =42 , x≤r
D−= (− r )

e




(1) Px `(τ, r) = 0 =
ex =4 D−= ( x )
2 2

1 − r2 =42 , r≤x

e D−= ( r )


Ex e−γ`(τ,r) ; Uτ ∈ dz

1.3.5

√
e−(z +r )=2 F (  ; r ; z )F (  ; x ; r )
2 2 2
  −z2 /2σ2 λ x z


= √ e F θ, σ, σ dz − √ dz
 2 22 2 +  2 e−r2 =22 F (  ; r ; r )

√ 2
F (  ; r ; z )F (  ; x ; r )


 2er =2 y
2 

1.3.6 Px `(τ, r) ∈ dy, Uτ ∈ dz = (z2−r2 )=22 2  r r exp − √ dydz

e F (; ; ) 
F (  ;  ; r )
r

(1) Px `(τ, r) = 0, Uτ ∈ dz

√
√
 e−z =2 F (  ; r ; z )F (  ; x ; r )
2 2
  −z2 /2σ2 λ x z


= √ e F θ, σ, σ dz − √ dz
 2  2F (  ; r ; r )

 Z τ



1.6.1 Ex exp − p1I(−∞,r) Us + q1I[r,∞) Us ds
0

( + p)−1 (q − p)e(x −r )=4 D− λ+p (− x )D− θ+λ+q ( r )

2 2 2

x≤r = − θ θ
+p ( + p)D− θ+λ+p (− r )D− λ+q ( r ) + ( + q)D− θ+λ+q ( r )D− λ+p (− r )
θ θ θ θ
530 7. ORNSTEIN{UHLENBECK PROCESS

7 Us = σe−θs We2θs −1 σ, θ > 0 τ ∼ Exp(λ), independent of U

( + q)−1 (q − p)e(x −r )=4 D− λ+q ( x )D− θ+λ+p (− r )

2 2 2

r≤x = + θ θ
+q ( + p)D− θ+λ+p (− r )D− λ+q ( r ) + ( + q)D− θ+λ+q ( r )D− λ+p (− r )
θ θ θ θ

x ex =4 D
−= (−  )
Z τ


 2 2

1.6.2 Px p1I(−∞,r) Us + q1I[r,∞) Us ds = pτ = 1 − r2 =42

0 e D−= (− r )
(1)
x≤r

τ
ex =4 D−= ( x )
Z  2 2




(2) Px p1I(−∞,r) Us + q1I[r,∞) Us ds = qτ = 1 − r2 =42
0 e D−= ( r )
r≤x

n  Z τ


 o
1.6.5 Ex exp − p1I(−∞,r) Us + q1I[r,∞) Us ds ; Uτ ∈ dz
0

√
  −z2 /2σ2 λ+p x z


z≤r = √ e F θ , σ, σ dz
 2
x≤r
 (( + p)=) (x2 −z2 )/4σ2
+ √ e D− λ+p (− σx )D− λ+p (− σz )
 2 θ θ

( + p)D− θ+λ+p ( r )D− λ+q ( r ) − ( + q)D− θ+λ+q ( r )D− λ+p ( r )

× θ θ θ θ
dz
( + p)D− θ+λ+p (− r )D− λ+q ( r ) + ( + q)D− θ+λ+q ( r )D− λ+p (− r )
θ θ θ θ

−1 e(x −z )=4 D− λ+q ( x )D− λ+p (− z )

2 2 2

z≤r = θ θ
dz
( + p)D− θ+λ+p (− r )D− λ+q ( r ) + ( + q)D− θ+λ+q ( r )D− λ+p (− r )
r≤x θ θ θ θ

−1 e(x −z )=4 D− λ+p (− x )D− λ+q ( z )

2 2 2

r≤z = θ θ
dz
( + p)D− θ+λ+p (− r )D− λ+q ( r ) + ( + q)D− θ+λ+q ( r )D− λ+p (− r )
x≤r θ θ θ θ

√
  −z2 /2σ2 λ+q x z


r≤z = √ e F θ , σ, σ dz
 2
r≤x
 (( + q)=) (x2 −z2 )/4σ2
+ √ e D− λ+q ( σx )D− λ+q ( σz )
 2 θ θ

( + q)D− θ+λ+q (− r )D− λ+p (− r ) − ( + p)D− θ+λ+p (− r )D− λ+q (− r )

× θ θ θ θ
dz
( + p)D− θ+λ+p (− r )D− λ+q ( r ) + ( + q)D− θ+λ+q ( r )D− λ+p (− r )
θ θ θ θ
 1. EXPONENTIAL STOPPING 531

7 Us = σe−θs We2θs −1 σ, θ > 0 τ ∼ Exp(λ), independent of U

 Z t  
x 2 2
2

1 − e−θt + 2 2θt + 1 − 2 − e−θt


1.8.3 Ex exp −γ Us ds = exp −
0
 2

Z t 
1.8.4 Px Us ds ∈ dy
0

 y − x 1 − e−t 2 



= exp −

dy
22 2t + 1 − 2 − e−t 2

2

1=2
 2 2t + 1 − 2 − e−t

Z τ

n   o
1.8.5 Ex exp − Us ds ; Uτ ∈ dz
 0

  − 2

x≤z = √ 2 2 2
e(x −z )/4σ Dγ 2 −λ/θ − σx − 2γ Dγ 2 −λ/θ

z


+ 2γ dz
 2 σ

  − 2

z≤x = √ 2 2 2
e(x −z )/4σ Dγ 2 −λ/θ x
+ 2γ Dγ 2 −λ/θ − σz − 2γ dz



 2 σ

Z t

n   o
1.8.7 Ex exp − Us ds ; Ut ∈ dz
 0

e t+ (z−x)= (z − xe−t + 2 (1 − e−t ))2 

2 
= exp − dz
(22 (1 − e−2t ))1=2 22 (1 − e−2t )

n  Z t  o
(1) Ex exp −γ Us ds Ut = z
0


(x + z ) 2 2

= exp − th(θt/2) + 2 θt − 2 th(θt/2)
 

Z t 
1.8.8 Px Us ds ∈ dy Ut = z
0

 y − (x + z ) th(t=2)
2 

=
1=2 exp − dy
2 (t − 2 th(t=2)) 42 t − 2 th(t=2)

√ t=2
( 2 − 1) t 2  e x2 ( 2 − 1) sh(t )
 Z  
1.9.3 Ex exp − 2 Us ds = p exp − 2
4 0 sh(t )+ ch(t ) 4 (sh(t )+ ch(t ))
1≤γ
532 7. ORNSTEIN{UHLENBECK PROCESS

7 Us = σe−θs We2θs −1 σ, θ > 0 τ ∼ Exp(λ), independent of U

( 2 − 1) t 2  et=2
n  Z o √
(1) Ex exp − U ds ; 0 < inf U =
42 s s
sh(t ) + ch(t )
p
0 0≤s≤t
0<x

x2 ( 2 − 1) sh(t ) x
   
× exp − 2 Erf
4 (sh(t ) + ch(t )) 2 sh(t )(sh(t ) + ch(t ))
p

( 2 − 1) τ 2 
n  Z o
1.9.5 Ex exp − U ds ; U ∈ dz
4 2 0
s τ
1≤γ
 ( 12 − 21 +  ) (x2−z2 )/4σ2 √
x γ
√
z γ
x≤z = √ e D− 1 + 1 − λ (− σ )D− 1 + 1 − λ ( σ )dz
 2 2 2γ θγ 2 2γ θγ

 ( 12 − 21 +  ) (x2−z2 )/4σ2 √
x γ
√
z γ
z≤x = √ e D− 1 + 1 − λ ( σ )D− 1 + 1 − λ (− σ )dz
 2 2 2γ θγ 2 2γ θγ

( 2 − 1) t 2 
n  Z o
1.9.7 Ex exp − U ds ; U ∈ dz
4 2 0
s t

et=2+(x −z )=4 (x2 + z 2 ) ch(t ) − 2xz 

√ 2 2 2 
= exp − dz
2  sh(t ) 42 sh(t )
p

Z t 
1.9.8 Px Us2 ds ∈ dy, Ut ∈ dz
0
1 2
−z 2 −θy)/4σ 2

1 x2 + z 2 xz

= √ etθ/2+(x eey/2θσ2 , t, , − dydz
4 3=2 3 2 42 22

Z t  2
(1) P0 Us2 ds ∈ dy Ut = 0 e−θt/2+θy/4σ
0

√ p ∞ √
 1 − e−2t X (k + 1=2) −(4k+1)2 t2 σ2 θ/8y (4k + 1)t  

= e D3/2 √ dy
(2)1=4 y5=4 k! 2y
k=0

e(x −y )=4 D−(+ )= (− x )D−(+)= (− y )

2 2 2

Ex e−γ Ȟ(τ ) ; sup Us ∈ dy =


1.13.1 dy
0≤s≤τ D−(+ )= (− y )D−= (− y )
x<y

@

1.13.4 P Ȟ(t) ∈ dv, sup Us < y
@y x 0≤s≤t
y=0
x<0
 1. EXPONENTIAL STOPPING 533

7 Us = σe−θs We2θs −1 σ, θ > 0 τ ∼ Exp(λ), independent of U

|x|ev−t=2+x =4 1I[0;t] (v) x2 ch(v)

2 2  
= exp − dv
22 sh3=2 (v) sh1=2 ((t − v)) 42 sh(v)

Ex e−γ Ȟ(τ ) ; sup Us ∈ dy, Uτ ∈ dz


1.13.5 x∨z <y
0≤s≤τ

e(x −z )=4 D−(+ )= (− x )D−= (− z )

2 2 2

= dydz
2 D−(+ )= (− y )D−= (− y )

@

1.13.8 Px Ȟ(t) ∈ dv, sup Us < y, Ut ∈ dz x∨z ≤0
@y 0≤s≤t
y=0
(1)
xzet=2 1I[0;t] (v)  2 2
x −z x2 ch(v) z 2 ch((t − v))

= exp − − dvdz
24 sh3=2 (v) sh3=2 ((t − v)) 4 2 42 sh(v) 42 sh((t − v))

e(x −y )=4 D−(+ )= ( x )D−(+)= ( y )

2 2 2

Ex e−γ Ĥ(τ ) ; inf Us ∈ dy =


1.14.1 dy
0≤s≤τ D−(+ )= ( y )D−= ( y )
y<x

@

1.14.4 P Ĥ(t) ∈ dv, inf Us < y
@y x 0≤s≤t y=0
0<x
xev−t=2+x =4 1I[0;t] (v) x2 ch(v)
2 2  
= 3= 2 1= 2 exp − 2 dv
22 sh (v) sh ((t − v)) 4 sh(v)

Ex e−γ Ĥ(τ ) ; inf Us ∈ dy, Uτ ∈ dz


1.14.5 y <x∧z
0≤s≤τ

e(x −z )=4 D−(+ )= ( x )D−= ( z )

2 2 2

= dydz
2 D−(+ )= ( y )D−= ( y )

@

1.14.8 Px Ȟ(t) ∈ dv, inf Us < y, Ut ∈ dz 0≤x∧z
@y 0≤s≤t y=0
(1)
xzet=2 1I[0;t] (v)  2 2
x −z x2 ch(v) z 2 ch((t − v))

= exp − − dvdz
24 sh3=2 (v) sh3=2 ((t − v)) 4 2 42 sh(v) 42 sh((t − v))

S(  ; b ; x ) + S(  ; x ; a )
Px a < inf Us , sup Us < b = 1 −    b a  


1.15.2
0≤s≤τ 0≤s≤τ S(  ;  ;  )
534 7. ORNSTEIN{UHLENBECK PROCESS

7 Us = σe−θs We2θs −1 σ, θ > 0 τ ∼ Exp(λ), independent of U

1.15.6 Px a < inf Us , sup Us < b, Uτ ∈ dz
0≤s≤τ 0≤s≤τ

 e−z =2 S (  ; b ; z+x+2|z−x| )S (  ; z+x−|z−x| ; a )

√ 2 2

= √ 2  dz
 b a
 2S (  ;  ;  )

Ex e−γ`(τ,r) ; sup Us < b


1.16.1
0≤s≤τ
r<b
√ √
e−r =2 S (  ; b ; r ) +  2)ex =4 D−= (− x )
(
2 2 2 2

x≤r = 1 − √ −r2 =42  b r   √

e S (  ;  ;  )D−= (− r ) +  2eb2 =42 D−= (− b )

S (  ; b ; x ) + S (  ; x ; r )
r≤x =1−
S (  ; b ; r )
x≤b
√
 2 eb =4 D−= (− b ) − er =4 D−= (− r ) S (  ; b ; x )
2 2 2 2

+ √ √
e−r2 =42 S (  ; b ; r )D−= (− r ) +  2eb2 =42 D−= (− b ) S (  ; b ; r )

 y√2e(b2 +r2 )=42 D b 

−= (−  )
1.16.2 Px `(τ, r) ∈ dy, sup Us < b = exp − √  b r r
0≤s≤τ S (  ;  ;  )D−= (−  )
r<b
 √
 2 eb =4 D−= (− b ) − er =4 D−= (− r ) ex =4 D−= (− x )
2 2 2 2
2 2

dy, x ≤ r

√
S (  ; b ; r )D−2 (− r )


= 

× √
 2 eb =4 D−= (− b ) − er =4 D−= (− r ) S (  ; b ; x )
2 2 2 2


dy, r≤x≤b

√ 2  b r −r2 =42
S (  ;  ;  )e D−= (− r )



ex =4 D−= (− x )
 2 2

1 − , x≤r
er2 =42 D (− r )


−= 


(1) Px `(τ, r) = 0, sup Us < b =  ; b ; x ) + S(  ; x ; r )
0≤s≤τ 
1 − S (       , r≤x≤b
S (  ; b ; r )


Ex e−γ`(τ,r) ; sup Us < b, Uτ ∈ dz


1.16.5 r∨z <b
0≤s≤τ

 e−z =2 S (  ; r ; z+x+2|z−x| )e(z+x−|z−x|) =4 D−= (− z+x−|z−x| )

√ 2 2 2 2

z≤r = √ 2 2 dz
 2e r = 4 2 r
D−= (−  )
x≤r
√
 e(x −z −r )=4 S (  ; b ; r )D−= (− z )D−= (− x )
2 2 2 2

+ √ −r2 =42  b r √ dz
e S (  ;  ;  )D−= (− r ) +  2eb2 =42 D−= (− b ) D−= (− r )


 1. EXPONENTIAL STOPPING 535

7 Us = σe−θs We2θs −1 σ, θ > 0 τ ∼ Exp(λ), independent of U

√
 e−z =4 S (  ; b ; x )D−= (− z )
2 2

z≤r = √ √ dz
e−r2 =42 S (  ; b ; r )D−= (− r ) +  2eb2 =42 D−= (− b )
r≤x

√
 e−z =2 S (  ; b ; z )ex =4 D−= (− x )
2 2 2 2

r≤z = √ −r2 =42  b r    r √ dz

e S (  ;  ;  )D−= (−  ) +  2eb2 =42 D−= (− b )
x≤r

 e−z =2 S (  ; b ; z+x+2|z−x| )S (  ; z+x−|z−x| ; r )

√ 2 2

r≤z = √ 2 2 dz
 2S (  ; b ; r )
r≤x
√
 e−z =2 S (  ; b ; z )er =4 D−= (− r )S (  ; b ; x )
2 2 2 2

+ √ √ dz
e−r2 =42 S (  ; b ; r )D−= (− r ) +  2eb2 =42 D−= (− b ) S (  ; b ; r )

1.16.6 Px `(τ, r) ∈ dy, sup Us < b, Uτ ∈ dz r∨z <b
0≤s≤τ

D−= (− z )D−= (− x )  y√2e(b2 +r2 )=42 D b 

−= (−  )
z≤r = (z2 −x2 )=42 2 r exp − √  b r r dydz
e D−= (−  ) S (  ;  ;  )D−= (−  )
x≤r

e(r −z )=4 S (  ; b ; x )D−= (− z )  y√2e(b2 +r2 )=42 D b 

−= (−  )
2 2 2

z≤r = exp − √ dydz

S (  ; b ; r )D−= (− r ) S (  ; b ; r )D−= (− r )
r≤x

e(r +x )=4 S (  ; b ; z )D−= (− x )  y√2e(b2 +r2 )=42 D b 

−= (−  )
2 2 2

r≤z =  b r r exp − √  b r r dydz

ez =2 S (  ;  ;  )D−= (−  )
2 2
S (  ;  ;  )D−= (−  )
x≤r

e(r −z )=2 S (  ; b ; z )S (  ; b ; x )  y√2e(b2 +r2 )=42 D b 

−= (−  )
2 2 2

r≤z = exp − √ dydz

S 2 (  ; b ; r ) S (  ; b ; r )D−= (− r )
r≤x

(1) Px `(τ, r) = 0, sup Us < b, Uτ ∈ dz
0≤s≤τ

 e−z =2 S (  ; r ; z+x+2|z−x| )e(z+x−|z−x|) =4 D−= (− z+x−|z−x| )

√ 2 2 2 2

z≤r = √ 2 2 dz
 2er =4 D−= (−  )
2 r
x≤r
2 e−z =2 S (  ; b ; z+x+2|z−x| )S (  ; z+x−|z−x| ; r )
√ 2 2

r≤z = √ 2  dz
 b r
 2S (  ;  ;  )
r≤x
536 7. ORNSTEIN{UHLENBECK PROCESS

7 Us = σe−θs We2θs −1 σ, θ > 0 τ ∼ Exp(λ), independent of U

Ex e−γ`(τ,r) ; a < inf Us


1.17.1
0≤s≤τ
a<r
S (  ; r ; x ) + S (  ; x ; a )
a≤x =1−
S (  ; r ; a )
x≤r
√
 2 ea =4 D−= ( a ) − er =4 D−= ( r ) S (  ; x ; a )
2 2 2 2

+ √ √
e−r2 =42 D−= ( r )S (  ; r ; a ) +  2ea2 =42 D−= ( a ) S (  ; r ; a )

√ √
e−r =2 S (  ; r ; a ) +  2)ex =4 D−= ( x )
(
2 2 2 2

r≤x = 1 − √ −r2 =42 √

e r
D−= (  )S (  ;  ;  ) +  2ea2 =42 D−= ( a )
 r a

 y√2e(a2 +r2 )=42 D a 

−= (  )
1.17.2 Px `(τ, r) ∈ dy, a < inf Us = exp − √ r  r a
0≤s≤τ D−= (  )S (  ;  ;  )
a<r
 √
 2 ea =4 D−= ( a ) − er =4 D−= ( r ) S (  ; x ; a )
2 2 2 2

dy, a≤x≤r

√
e−r2 =42 D−= ( r )S 2 (  ; r ; a )



× √
 2 ea =4 D−= ( a ) − er =4 D−= ( r ) ex =4 D−= ( x )
2 2 2 2
2 2


√ 2 r  r a dy, r ≤ x
D− = (  )S (  ;  ;  )



S (  ; r ; x ) + S (  ; x ; a )

 1− S(  ; r ; a )
, a≤x≤r



  
(1) Px `(τ, r) = 0, a < inf Us =
0≤s≤τ ex2 =42 D
 1 − 2 2 −= r ,
 (x)
 r≤x
er =4 D−= (  )

Ex e−γ`(τ,r) ; a < inf Us , Uτ ∈ dz


1.17.5 a<r∧z
0≤s≤τ

 e−z =2 S (  ; r ; z+x+2|z−x| )S (  ; z+x−|z−x| ; a )

√ 2 2

z≤r = √ 2  dz
 2S (  ; r ; a )
x≤r
√
 e−z =2 er =4 D−= ( r )S (  ; z ; a )S (  ; x ; a )
2 2 2 2

+ √ √ dz
e−r =42 D−= ( r )S (  ; r ; a ) +  2ea2 =42 D−= ( a ) S (  ; r ; a )
2

√
 e−z =2 ex =4 D−= ( x )S (  ; z ; a )
2 2 2 2

z≤r = √ −r2 =42 √ dz

e D−= ( r )S (  ; r ; a ) +  2ea2 =42 D−= ( a )
r≤x

√
 e−z =4 D−= ( z )S (  ; x ; a )
2 2

r≤z = √ √ dz
e−r =4 D−= ( r )S (  ; r ; a ) +  2ea2 =42 D−= ( a )
2 2
x≤r
 1. EXPONENTIAL STOPPING 537

7 Us = σe−θs We2θs −1 σ, θ > 0 τ ∼ Exp(λ), independent of U

 e−z =2 e(z+x+|z−x|) =4 D−= ( z+x+2|z−x| )S (  ; z+x−|z−x| ; r )

√ 2 2 2 2

r≤z = √ 2 2  dz
r
 2er =4 D−= (  )
2
r≤x
√
 e(x −z −r )=4 D−= ( z )S (  ; r ; a )D−= ( x )
2 2 2 2

+ √ −r2 =42 √ dz
e D−= ( r )S (  ; r ; a ) +  2ea2 =42 D−= ( a ) D−= ( r )

1.17.6 Px `(τ, r) ∈ dy, a < inf Us , Uτ ∈ dz a<r∧z
0≤s≤τ

e(r −z )=4 S (  ; z ; a )S (  ; x ; a )  y√2e(a2 +r2 )=42 D a 

−= (  )
2 2 2

z≤r = exp − √ dydz

S 2 (  ; r ; a ) D−= ( r )S (  ; r ; a )
x≤r
e(x +r )=4 D−= ( x )S (  ; z ; a )  y√2e(a2 +r2 )=42 D a 
−= (  )
2 2 2

z≤r = exp − √ dydz

ez2 =22 D−= ( r )S (  ; r ; a ) D−= ( r )S (  ; r ; a )
r≤x

e(r −z )=4 D−= ( z )S (  ; x ; a )  y√2e(a2 +r2 )=42 D a 

−= (  )
2 2 2

r≤z = r  r a exp − √ r  r a dydz

D−= (  )S (  ;  ;  ) D−= (  )S (  ;  ;  )
x≤r
e(x −z )=4 D−= ( z )D−= ( x )  y√2e(a2 +r2 )=42 D a 
−= (  )
2 2 2

r≤z = 2 r exp − √ dydz

D− = (  ) D−= ( r )S (  ; r ; a )
r≤x

(1) Px `(τ, r) = 0, a < inf Us , Uτ ∈ dz
0≤s≤τ

 e−z =2 S (  ; r ; z+x+2|z−x| )S (  ; z+x−|z−x| ; a )

√ 2 2

z≤r = √ 2  dz
 r a
 2S (  ;  ;  )
x≤r
 e−z =2 e(z+x+|z−x|) =4 D−= ( z+x+2|z−x| )S (  ; z+x−|z−x| ; r )
√ 2 2 2 2

r≤z = √ 2 2  dz
 2er =42 D−= ( r )
r≤x

n  Z t 2 2   o
1.20.3 Ex exp − 2 ds ; 0 < inf Us
0
Us 0≤s≤t
0<x

(42 sh(t))1=4 3+ 1 + 2
 q 
x2 e−t
 2 −t 
4 16 4 x e
 
= exp − 2 M 1 q 1 γ2
√ −t=4 q 1 2 8 sh(t) 42 sh(t)
xe 4 + +1

− 4 , 16 + 4

t
ds
Z 
1.20.4 Px ∈ dy, 0 < inf Us
0
Us2 0≤s≤t
0<x
538 7. ORNSTEIN{UHLENBECK PROCESS

7 Us = σe−θs We2θs −1 σ, θ > 0 τ ∼ Exp(λ), independent of U

(42 )5=4 sh1=4 (t) (t − 2 y) x2 e−t 1 x2 e−t 

  
= 1= 2 exp − 2 m4σ2 θy , 2 dy
x 4 8 sh(t) 4 8 sh(t)

1 1 
q
 16 + 4 + 4 + 2
2

n  Z τ 2 2   o
1.20.5 Ex exp − ds ; 0 < inf Us , Uτ ∈ dz =
Us2  1 + 2 + 1 √xz
q 
0≤s≤τ
0<x 0 4

 2   2 
 x z
M θ−2λ q 1 γ 2 2 W θ−2λ q 1 γ 2 dz, 0 ≤ x ≤ z
2  22

, + , +
2 2 2

×e(x −z )/4σ
4θ 16 4 4θ 16 4
 2   2 
x z
W M θ−2λ q 1 γ 2 dz, z ≤ x

2 2 22
q
θ−2λ 1 γ2
4θ , 16 + 4 4θ , 16 + 4

n  Z t 2 2   o
1.20.7 Ex exp − ds ; 0 < inf U , U ∈ dz
0
Us2 0≤s≤t
s t
0<x
√
xz t x2 − z 2 (x2 + z 2 ) ch(t) √ xz
   
= exp + − I dz
22 sh(t) 2 42 42 sh(t) 1
4 +γ
2
22 sh(t)

t
ds
Z 
1.20.8 Px ∈ dy, 0 < inf Us , Ut ∈ dz
0
Us2 0≤s≤t
0<x

√
 xz t 2 y x2 − z 2 (x2 + z 2 ) ch(t) xz
   
= exp − + − i dydz
2 sh(t) 2 4 42 42 sh(t) 22 sh(t)
σ 2 θy

 Z t 2 2
p   (q2 − 1) 2
n   o
1.21.3 Ex exp − 2 + 2 Us ds ; 0 < inf Us
0
Us 4 0≤s≤t
0<x
(42 sh(tq))1=4 et=2 x2 (2(q2 − 1) sh2 (tq) + q2 )
 
= 1= 4 exp − 2
(sh(tq) + q ch(tq)) 8 sh(tq)(sh(tq) + q ch(tq))

3+
1 p2
q
16 + 4
 
4 x2 q 2
 
× √ q  M 1 q 1 p2
x 1 + p2 + 1 − 4 , 16 + 4 42 sh(tq)(sh(tq) + q ch(tq))
4

(q2 − 1)
Z t  Z t
ds
n  o
2
(1) Ex exp − 2 Us ds ; 2 ∈ dy, 0 < inf Us
4 0 0
Us 0≤s≤t

(42 )5=4 sh1=4 (tq)et=2 1 x2 q 2

 
= 1=2 m ,
x (sh(tq) + q ch(tq))1=4 4 82 sh(tq)(sh(tq) + q ch(tq))
4σ 2 θy

x2 (2(q2 − 1) sh2 (tq) + q2 )

 2
 y

× exp − − 2 dy
4 8 sh(tq)(sh(tq) + q ch(tq))
 1. EXPONENTIAL STOPPING 539

7 Us = σe−θs We2θs −1 σ, θ > 0 τ ∼ Exp(λ), independent of U

 Z τ 2 2
p   (q2 − 1)Us2
n   o
1.21.5 Ex exp − 2 + 2 ds ; 0 < inf Us , Uτ ∈ dz
0
Us 4 0≤s≤τ
0<x
1 p2 1 2−
q
 16 + 4 + 2 + 4q

2 2 2
q>1 = e(x −z )/4σ
1
√
2
q 4 + p + 1 xz
q

 2  2
qx qz

 M θ−2λ , 1 + p2 22 W θ−2λ , 1 + p2 22 dz,

 q q 0≤x≤z
4θq 16 4 4θq 16 4
×  2  2
qx qz
 W θ−2λ 1 p2 M dz, z≤x

22 22
 q q
2
θ−2λ 1 p
4θq , 16 + 4 4θq , 16 + 4

 Z t 2 2
p   (q2 − 1) 2
n   o
1.21.7 Ex exp − + U ds ; 0 < inf U , U ∈ dz
0
Us2 42 s
0≤s≤t
s t
0<x
√
q xz t x2 − z 2 (x2 + z 2 )q ch(tq) √ xzq
   
= exp + − I dz
22 sh(tq) 2 42 42 sh(tq) 1
4 +p
2
22 sh(tq)

(q2 − 1)
Z t  Z t
ds
n  o
2
(1) Ex exp − 2 Us ds ; 2 ∈ dy, 0 < inf Us , Ut ∈ dz
4 0 0
Us 0≤s≤t

√
q xz t 2 y x2 − z 2 (x2 + z 2 )q ch(tq) xzq
   
= exp − + 2 − 2 iσ2 θy 2 dydz
2 sh(tq) 2 4 4 4 sh(tq) 2 sh(tq)

p2  2 
n  Z t  Z t o
2
(2) Ex exp − ds ; U ds ∈ dy, 0 < inf U , U ∈ dz
0
Us2 0
s
0≤s≤t
s t

√
xz t x2 − z 2 − y x2 + z 2 xz
  q 
1
= 4 2 exp + 2 is + p2 , t, 0, 2 , 2 dydz
4  2 4 4  4 
2
y/2σ θ 4

t t
ds
Z Z 
1.21.8 Px Us2 ds
2 ∈ dy, 0 < inf Us , Ut ∈ dz
∈ dg,
0 0
Us 0≤s≤t
0<x
√
xz t 2 y x2 − z 2 − g x2 + z 2 xz
   
2
= 2 exp − + ei σ θy, t, , dgdydz
42 42  22 
g
4  2 4 2σ 2 θ

n  Z τ


 o
1.23.1 Ex exp − p1I(−∞,r) Us + q1I[r,∞) Us ds ; sup Us < b
0 0≤s≤τ
r<b

( + p)−1 (p − q)C ( +q r b

 ;  ;  ) − ( + p) 2=e
r2 =22
D λ+p (− x )ex2 =42
p
 − 
x≤r = + θ
+p ( + q) S ( + q ; b ; r ) + p D− θ+λ+p (− r ) + C ( + q ; r ; b )D− λ+p (− r ) er2 =42


θ θ
540 7. ORNSTEIN{UHLENBECK PROCESS

7 Us = σe−θs We2θs −1 σ, θ > 0 τ ∼ Exp(λ), independent of U


 S ( + q x r 
 ; ; )
r≤x = 1−
+q S ( + q b r
 ; ; )
x≤b
( + q)−1 2=er =2 D− λ+p (− r )S ( + q b x
 ; ; )
p 2 2

− θ

S ( + q b r  +p r  +q r b r  +q b r
 ;  ;  )  D− θ+λ+p (−  ) + C (  ;  ;  )D− λ+p (−  ) S (  ;  ;  )


θ θ

(q − p)D− θ+λ+p (− r )S ( + q b x

 ; ; )
θ
+
( + q) S ( + q ; b ; r ) + p D− θ+λ+p (− r ) + C ( + q ; r ; b )D− λ+p (− r )


θ θ

n  Z τ


 o
1.23.5 Ex exp − p1I(−∞,r) Us + q1I[r,∞) Us ds ; sup Us < b, Uτ ∈ dz
0 0≤s≤τ
r<b
z<b

 e−z =2 S ( + p ; r ; z+x+2|z−x| )e(z+x−|z−x|) =4 D− λ+p (− z+x−|z−x| )

√ 2 2 2 2
2
z≤r = √ θ
dz
 2er2 =42 D− λ+p (− r )
x≤r θ

e(x −z )=4 S ( +q b r z x

 ;  ;  )D− λ+p (−  )D− λ+p (−  )
2 2 2

θ θ
+ dz
 S ( +q b r +p
 ;  ;  )  D− θ+λ+p (− r ) + C ( + q ; r ; b )D− λ+p (− r ) D− λ+p (− r )


θ θ θ

e(r −z )=4 D− λ+p (− z )S ( + q b x

 ; ; )
2 2 2

z≤r = θ

dz
 S ( + q b r  +p r  +q r b r
 ;  ;  )  D− θ+λ+p (−  ) + C (  ;  ;  )D− λ+p (−  )
r≤x θ θ

e−z =2 e(x +r )=4 S ( + q b z x

 ;  ;  )D− λ+p (−  )
2 2 2 2 2

r≤z = θ

dz
 S ( + q b r  +p
 ;  ;  )  D− θ+λ+p (− r ) + C ( + q ; r ; b )D− λ+p (− r )
x≤r θ θ

 e−z =2 S ( +
√ q b z+x+|z−x| )S ( +q ; z+x−|z−x| ; r )
 ; √;
2 2

r≤z = 2  2  dz
 2S ( + q b r
 ; ; )
r≤x
e(r −z )=2 D− λ+q (− r )S ( +q b z +q b x
 ;  ;  )S (  ;  ;  )
2 2 2

θ
+ dz
 S ( +q b r +p
 ;  ;  )  D− θ+λ+p (− r ) + C ( + q ; r ; b )D− λ+p (− r ) S ( + q ; b ; r )


θ θ

n  Z τ


 o
1.24.1 Ex exp − p1I(−∞,r) Us + q1I[r,∞) Us ds ; a < inf Us
0 0≤s≤τ
a<r


 S ( + p ; r ; x ) 
a≤x = 1−
+p S ( + p ; r ; a )
x≤r
 1. EXPONENTIAL STOPPING 541

7 Us = σe−θs We2θs −1 σ, θ > 0 τ ∼ Exp(λ), independent of U

(p − q)D− θ+λ+p ( r )S ( + q x a

 ; ; )
θ
+
( + p) D− λ+q ( r )C ( + q ; r ; a ) + + q D− θ+λ+q ( r )S ( + p ; r ; a )


θ θ

( + p)−1 2=er =2 D− λ+q ( r )S ( + p ; x ; a )

p 2 2

− θ

D− λ+q ( r )C ( + p ; r ; a ) + +q

 D− θ+λ+q ( r )S ( + p ; r ; a ) S ( + p ; r ; a )


θ θ

( + p)−1 ((q − p)C ( + p ; r ; a ) − ( + q) 2=er =2 )D− λ+q ( x )ex =4
p 2 2 2 2

r≤x = + θ
+q ( + q) D− λ+q ( r )C ( + p ; r ; a ) + + q D− θ+λ+q ( r )S ( + p ; r ; a ) er2 =42


θ θ

n  Z τ


 o
1.24.5 Ex exp − p1I(−∞,r) Us + q1I[r,∞) Us ds ; a < inf Us , Uτ ∈ dz
0 0≤s≤τ
a<r
a<z

 e−z =2 S ( + p ; r ; z+x+2|z−x| )S ( + p ; z+x−|z−x| ; a )

√ 2 2

z≤r = √ 2  dz
 + p
 2S (  ; r ; a )
x≤r
e(r −z )=2 D− λ+q ( r )S ( + p ; z ; a )S ( + p ; x ; a )
2 2 2

θ
+ dz
 D− λ+q ( r )C ( + p ; r ; a ) + +q
 D− θ+λ+q ( r )S ( + p ; r ; a ) S ( + p ; r ; a )


θ θ

e−z =2 e(x +r )=4 S ( + p ; z ; a )D− λ+q ( x )

2 2 2 2 2

z≤r = θ

dz
 D− λ+q ( r )C ( + p ; r ; a ) + + q
 D− θ+λ+q ( r )S ( + p ; r ; a )
r≤x θ θ

e(r −z )=4 D− λ+q ( z )S ( + p ; x ; a )

2 2 2

r≤z = θ
+p r a
dz
 D− λ+q ( r )C ( + p ; r ; a ) + +

qD r
θ+λ+q (  )S (  ;  ;  )
x≤r θ
− θ

 e−z =2 e(z+x+|z−x|) =4 D− λ+q ( z+x+2|z−x| )S ( + q z+x−|z−x| ; r )

√
 ;
2 2 2 2
2 
r≤z = √ θ
dz
 2er2 =42 D− λ+q ( r )
r≤x θ

e(x −z )=4 S ( + p ; r ; a )D− λ+q ( z )D− λ+q ( x )

2 2 2

θ θ
+ dz
 D− λ+q ( r )C ( + p ; r ; a ) + + q
 D− θ+λ+q ( r )S ( + p ; r ; a ) D− λ+q ( r )


θ θ θ

Ex e−γ`(τ,r) ; a < inf Us , sup Us < b


1.25.1 a<r<b
0≤s≤τ 0≤s≤τ

S (  ; r ; x ) + S (  ; x ; a )
a≤x =1−
S (  ; r ; a )
x≤r
542 7. ORNSTEIN{UHLENBECK PROCESS

7 Us = σe−θs We2θs −1 σ, θ > 0 τ ∼ Exp(λ), independent of U

√
 2 S (  ; b ; a ) − S (  ; b ; r ) − S (  ; r ; a ) S (  ; x ; a )


+ √ √
e−r2 =22 S (  ; b ; r )S (  ; r ; a ) +  2S (  ; b ; a ) S (  ; r ; a )

S (  ; b ; x ) + S (  ; x ; r )
r≤x =1−
S (  ; b ; r )
x≤b
√
 2 S (  ; b ; a ) − S (  ; b ; r ) − S (  ; r ; a ) S (  ; b ; x )


+ √ √
e−r2 =22 S (  ; b ; r )S (  ; r ; a ) +  2S (  ; b ; a ) S (  ; b ; r )

1.25.2 Px `(τ, r) ∈ dy, a < inf Us , sup Us < b
0≤s≤τ 0≤s≤τ

 y√2er2 =22 S (  ; b ; a ) 
= exp − √   
S (  ; b ; r )S (  ; r ; a )

 √
 2 S (  ; b ; a ) − S (  ; b ; r ) − S (  ; r ; a ) S (  ; x ; a )


 √ −r2 =22  b r 2  r a dy, a ≤ x ≤ r
e S (  ;  ;  )S (  ;  ;  )


× √
 2 S (  ; b ; a ) − S (  ; b ; r ) − S (  ; r ; a ) S (  ; b ; x )



√ −r2 =22 2  b r dy, r ≤ x ≤ b
S (  ;  ;  )S (  ; r ; a )

e


(1) Px `(τ, r) = 0, a < inf Us , sup Us < b
0≤s≤τ 0≤s≤τ

S (  ; r ; x ) + S (  ; x ; a )

 1− S (  ; r ; a )
, a≤x≤r


=
S(  ; b ; x ) + S(  ; x ; r )
 1 −    b r    , r ≤ x ≤ b


S(  ;  ;  )

Ex e−γ`(τ,r) ; a < inf Us , sup Us < b, Uτ ∈ dz


1.25.5 a<r<b
0≤s≤τ 0≤s≤τ

 e−z =2 S (  ; r ; z+x+2|z−x| )S (  ; z+x−|z−x| ; a )

√ 2 2

z≤r = √ 2  dz
 r a
 2S (  ;  ;  )
x≤r
√
 e−z =2 S (  ; b ; r )S (  ; z ; a )S (  ; x ; a )
2 2

+ √ −r2 =22  b r √ dz
e S (  ;  ;  )S (  ; r ; a ) +  2S (  ; b ; a ) S (  ; r ; a )

√
 e−z =2 S (  ; b ; x )S (  ; z ; a )
2 2

z≤r = √ √ dz
e−r =22 S (  ; b ; r )S (  ; r ; a ) +  2S (  ; b ; a )
2
r≤x
 1. EXPONENTIAL STOPPING 543

7 Us = σe−θs We2θs −1 σ, θ > 0 τ ∼ Exp(λ), independent of U

√
 e−z =2 S (  ; b ; z )S (  ; x ; a )
2 2

r≤z = √ √ dz
e−r =22 S (  ; b ; r )S (  ; r ; a ) +  2S (  ; b ; a )
2
x≤r

 e−z =2 S (  ; b ; z+x+2|z−x| )S (  ; z+x−|z−x| ; r )

√ 2 2

r≤z = √ 2  dz
 b r
 2S (  ;  ;  )
r≤x
√
 e−z =2 S (  ; b ; z )S (  ; r ; a )S (  ; b ; x )
2 2

+ √ √ dz
e−r 2 =2 2
S (  ; b ; r )S (  ; r ; a ) +  2S (  ; b ; a ) S (  ; b ; r )

1.25.6 Px `(τ, r) ∈ dy, a < inf Us , sup Us < b, Uτ ∈ dz
0≤s≤τ 0≤s≤τ
a<r
r<b
 −z2 =22  z a  x a
e S (  ;  ;  )S (  ;  ;  )  y√2er2 =22 S (  ; b ; a ) 
exp − √    dydz, x ≤ r
e−r2 =22 S 2 (  ; r ; a ) S (  ; b ; r )S (  ; r ; a )



z≤r =  y√2er2 =22 S (  ; b ; a ) 
 e−z2 =22 S (  ; b ; x )S (  ; z ; a )    dydz, r ≤ x
exp − √

e−r2 =22 S (  ; b ; r )S (  ; r ; a ) S (  ; b ; r )S (  ; r ; a )


 −z2 =22  b z
e S (  ;  ;  )S (  ; x ; a )  y√2er2 =22 S (  ; b ; a ) 
  
 e−r2 =22 S (  ; b ; r )S (  ; r ; a ) exp − √S (  ; b ; r )S (  ; r ; a ) dydz, x ≤ r


           
r≤z = √ r2 =22  b a 
−z 2 =2 2 S (  ; b ; z )S (  ; b ; x ) 2e S(  ;  ;  )
 e       exp − y
 
dydz, r ≤ x

2 =2 2 2  b r √
e − r S (; ; ) 
S (  ;  ;  )S (  ; r ; a )
b r

(1) Px `(τ, r) = 0, a < inf Us , sup Us < b, Uτ ∈ dz
0≤s≤τ 0≤s≤τ

 e−z =2 S (  ; r ; z+x+2|z−x| )S (  ; z+x−|z−x| ; a )

√ 2 2

z≤r = √ 2  dz
 2S (  ; r ; a )
x≤r

 e−z =2 S (  ; b ; z+x+2|z−x| )S (  ; z+x−|z−x| ; r )

√ 2 2

r≤z = √ 2  dz
 b r
 2S (  ;  ;  )
r≤x

n  Z τ


 o
1.26.1 Ex exp − p1I(−∞,r) Us + q1I[r,∞) Us ds ; a < inf Us , sup Us < b
0 0≤s≤τ 0≤s≤τ
a<r
r<b
544 7. ORNSTEIN{UHLENBECK PROCESS

7 Us = σe−θs We2θs −1 σ, θ > 0 τ ∼ Exp(λ), independent of U


 S ( + p ; r ; x ) 
a≤x = 1−
+p S ( + p ; r ; a )
x≤r
( + p)−1 2=er =2 S ( + q b r  +p x a
 ;  ;  )S (  ; 
;  )
p 2 2

−
S ( + q b r  +p r a  +q r b  +p r a  +p r a
 ;  ;  )C (  ;  ;  ) + C (  ;  ;  )S (  ;  ;  ) S (  ;  ;  )
((p − q)C ( + q r b
 ;  ;  ) − ( + p) 2=e
r =2 )S ( +p ; x ; a )
p 2 2

+   
( + p)( + q) S ( + q ; b ; r )C ( + p ; r ; a ) + C ( + q ; r ; b )S ( + p ; r ; a )


 S ( + q x r 
 ; ; )
r≤x = 1− +
+q S (  q ; b ; r )

x≤b
( + q)−1 2=er =2 S ( + p ; r ; a )S ( + q b x
 ; 
;  )
p 2 2

−  +
S (  ;  ;  )C (  ;  ;  ) + C (  ;  ;  )S (  ;  ; a ) S ( +
q b r  + p r a  + q r b  + p r q b r
 ; ; )
((q − p)C ( + p ; r ; a ) − ( + q) 2=er =2 )S ( + q b x
 ; ; )
p 2 2

+  +
( + p)( + q) S (  ;  ;  )C (  ;  ;  ) + C (  ;  ;  )S ( + p ; r ; a )
q b r  +p r a +q r b

n  Z τ



1.26.5 Ex exp − p1I(−∞,r) Us + q1I[r,∞) Us ds ;
r<b 0
o
a<r a < inf Us , sup Us < b, Uτ ∈ dz
0≤s≤τ 0≤s≤τ

 e−z =2 S ( + p ; r ; z+x+2|z−x| )S ( + p ; z+x−|z−x| ; a )

√ 2 2

z≤r = √ 2  dz
 + p r
 2S (  ;  ;  ) a
x≤r
e(r −z )=2 S ( + q b r  +p z a  +p x a
 ;  ;  )S (  ;  ;  )S (  ; 
;  )
2 2 2

+ dz
 +
 S (  ;  ;  )C (  ;  ;  ) + C (  ;  ;  )S (  ; r ; a ) S ( + p ; r ; a )
q b r  + p r a  + q r b  + p

e(r −z )=2 S ( + p ; z ; a )S ( + q b x

 ; ; )
2 2 2

z≤r = +p r a
dz
 S ( + q b r + p r a + q r b
 ;  ;  )C (  ;  ;  ) + C (  ;  ;  )S (  ;  ;  )
r≤x

e(r −z )=2 S ( + q b z  +p x a
 ;  ;  )S (  ;  ;  )
2 2 2

r≤z =
dz
 +
 S (  ;  ;  )C (  ;  ;  ) + C (  ; r ; b )S ( + p ; r ; a )
q b r  + p r a  + q
x≤r

 e−z =2 S ( +
√ q b z+x+|z−x| )S ( +q ; z+x−|z−x| ; r )
 ; √;
2 2

r≤z = 2  2  dz
 2S ( + q; b; r)
  
r≤x
e(r −z )=2 S ( + p ; r ; a )S ( + q b z  +q b x
 ;  ;  )S (  ; 
;  )
2 2 2

+ +p b r dz
 S ( + q b r + p r a + q r b  +p r a
 ;  ;  )C (  ;  ;  ) + C (  ;  ;  )S (  ;  ;  ) S (  ;  ;  )
 1. EXPONENTIAL STOPPING 545

7 Us = σe−θs We2θs −1 σ, θ > 0 τ ∼ Exp(λ), independent of U

n  Z τ


 o
1.27.1 Ex exp − p1I(−∞,r) Us + q1I[r,∞) Us ds ; `(τ, r) ∈ dy
0

 y( + p)D θ+λ+p (− r ) y( + q)D θ+λ+q ( r ) 

−  − 
= exp − θ
− θ
D− λ+p (− r ) D− λ+q ( r )
θ θ

ex =4 D− λ+p (− x )  D− θ+λ+p (− r ) D− θ+λ+q ( r ) 

 2 2

r + D λ+q ( r ) dy, x ≤ r
θ θ θ

 er2 =42 D λ+p (− r ) λ+p (−  )

D

 − θ  − θ
− θ

× x r r
ex =4 D− λ+q (  ) D− θ+λ+p (−  ) D− θ+λ+q (  ) 
2 2 

 θ θ
+ θ
dy, r ≤ x
D− λ+p (− r ) D− λ+q ( r )

D− λ+q ( r )
 r2 =42
e

θ θ θ

n  Z τ


 o
(1) Ex exp − p1I(−∞,r) Us + q1I[r,∞) Us ds ; `(τ, r) = 0
0
ex =4 D− λ+p (− x ) 
 2 2


1 − r2 =42 θ
, x≤r

 +p D− λ+p (− r )

e


θ
=
ex =4 D− λ+q ( r ) 
2 2

 

1 − r2 =42 θ
, r≤x

+q e D− λ+q ( r )


θ

n  Z τ


 o
1.27.5 Ex exp − p1I(−∞,r) Us + q1I[r,∞) Us ds ; `(τ, r) ∈ dy, Uτ ∈ dz
0

 y( + p)D θ+λ+p (− r ) y( + q)D θ+λ+q ( r ) 

−  − 
z≤r = exp − θ
− θ
D− λ+p (− r ) D− λ+q ( r )
θ θ

ex =4 D− λ+p (− x )D− λ+p (− z )

2 2


θ θ
dydz, x ≤ r

ez2 =42 D2 λ+p (− r )




− θ
×
ex =4 D− λ+p (− z )D− λ+q ( x )
2 2


θ θ
 ez2 =42 D λ+p (− r )D λ+q ( r ) dydz, r≤x



−  − θ
 θ

 y( + p)D θ+λ+p (− r ) y( + q)D θ+λ+q ( r ) 

−  − 
r≤z = exp − θ
− θ
D− λ+p (− r ) D− λ+q ( r )
θ θ

ex =4 D− λ+p (− x )D− λ+q ( z )

2 2


θ θ
dydz, x ≤ r

 ez2 =42 D− λ+p (− r )D− λ+q ( r )



θ θ
×
ex =4 D− λ+q ( x )D− λ+q ( z )
2 2

 θ θ
dydz, r≤x
ez2 =42 D2 λ+q ( r )




− θ
546 7. ORNSTEIN{UHLENBECK PROCESS

7 Us = σe−θs We2θs −1 σ, θ > 0 τ ∼ Exp(λ), independent of U

n  Z τ


 o
(1) Ex exp − p1I(−∞,r) Us + q1I[r,∞) Us ds ; `(τ, r) = 0, Uτ ∈ dz
0
√
 e−z =2
2 2 n ( + p )D− λ+p (− x )D− λ+p (− z )D− λ+p ( r ) o
λ+p x z


z≤r = √ F θ , σ, σ − θ θ θ
dz
 2 e(−x2−z2 )=42 D− λ+p (− r )
x≤r θ

√
 e−z =2
2 2 n ( + q )D− λ+q ( x )D− λ+q ( z )D− λ+q (− r ) o
λ+q x z


r≤z = √ F θ , σ, σ − θ θ θ
dz
 2 e(−x2 −z2 )=42 D− λ+q ( r )
r≤x θ

Ex e−γ Ȟ(τ )−ηĤ(τ ) ; inf Us ∈ da, sup Us ∈ db


1.28.1
0≤s≤τ 0≤s≤τ
a<x
√ √ √

2e(b +a )=2 S ( + + ; b ; x ) e−b =2 C (  ; b ; a ) − 2= 
2 2 2 2 2

x<b = dadb
2 S ( + + ; b ; a )S ( +  ; b ; a )S (  ; b ; a )
√

√ √ √

2e(b +a )=2 S ( + + ; x ; a ) e−a =2 C (  ; a ; b ) − 2= 
2 2 2 2 2

+ dadb
2 S ( + + ; b ; a )S ( + ; b ; a )S (  ; b ; a )
√

Ex e−γ Ȟ(τ )−ηĤ(τ ) ; inf Us ∈ da, sup Us ∈ db, Uτ ∈ dz


1.28.5
0≤s≤τ 0≤s≤τ
a<x
√
 2e(b +a −z )=2 S ( + + ; b ; x )S (  ; z ; a )
2 2 2 2

x<b = dadbdz
3 S ( + + ; b ; a )S ( +  ; b ; a )S (  ; b ; a )
√

√
 2e(b +a −z )=2 S ( + + ; x ; a )S (  ; b ; z )
2 2 2 2

+ 3√ dadbdz
 S ( + + ; b ; a )S ( + ; b ; a )S (  ; b ; a )

n  Z τ



1.29.1 Ex exp − p1I(−∞,r) Us + q1I[r,∞) Us ds ; a < inf Us ,
0 0≤s≤τ
o
sup Us < b, `(τ, r) ∈ dy
0≤s≤τ
 C ( +q ; r ; b ) − p2=er2 =22 C ( + p ; r ; a ) − 2=er =2 
p 2 2

= λθσ    +
 + q
( + q)S (  ; b ; r )  + p
( + p)S (  ; r ; a )

  +p x a
 S (  ;  ;  ) dy, a ≤ x ≤ r
 S (  +p ; r ; a )
 yC ( +q ; r ; b ) yC ( +p ; r ; a ) 
   −      
× exp −  + q b r  + p r a  +q ; b ; x )
S(  ;  ;  ) S(  ;  ;  )  S (    dy, r ≤ x ≤ b
 +
S (  q ; b ; r )

 1. EXPONENTIAL STOPPING 547

7 Us = σe−θs We2θs −1 σ, θ > 0 τ ∼ Exp(λ), independent of U

n  Z τ



(1) Ex exp − p1I(−∞,r) Us + q1I[r,∞) Us ds ; a < inf Us ,
0 0≤s≤τ
o
sup Us < b, `(τ, r) = 0
0≤s≤τ

 S ( + p ; r ; x ) + S ( + p ; x ; a ) 
 
 +p 1− S ( + p ; r ; a )
, a≤x≤r


=
S ( + q b x  +q x r 
 ;  ;  ) + S(  ;  ;  ) , r ≤ x ≤ b
  1−
 
S ( + q b r

+q  ; ; )

n  Z τ



1.29.5 Ex exp − p1I(−∞,r) Us + q1I[r,∞) Us ds ; a < inf Us ,
0 0≤s≤τ
o
sup Us < b, `(τ, r) ∈ dy, Uτ ∈ dz
0≤s≤τ

 yC ( +q ; r ; b ) yC ( +p ; r ; a ) 

= λ exp −      
S ( + q ; b ; r ) − S (  +p ; r ; a )
     

e−z =2 S ( + p ; z ; a )S ( + p ; x ; a )

 2 2

dydz, a ≤ x ≤ r, a ≤ z ≤ r
e−r2 =22 S 2 ( + p ; r ; a )





e−z =2 S ( + q b x  +p z a

 ;  ;  )S (  ;  ;  ) dydz,
2 2



a≤z≤r≤x≤b

e−r =2 S ( + q b r  +p r a

 ;  ;  )S (  ;  ;  )
2 2


×
e−z =2 S ( + q b z  +p x a
 ;  ;  )S (  ;  ;  ) dydz,
2 2

a≤x≤r≤z≤b

e−r2 =22 S ( + q b r  +p r a
 ;  ;  )S (  ;  ;  )





e−z =2 S ( + q b z  +q b x

 ;  ;  )S (  ;  ;  ) dydz,

 2 2


r 2 =2 2 2 +q b r
r ≤ x ≤ b, r ≤ z ≤ b
e S (  ; ; )

−

n  Z τ



(1) Ex exp − p1I(−∞,r) Us + q1I[r,∞) Us ds ; a < inf Us ,
0 0≤s≤τ
o
sup Us < b, `(τ, r) = 0, Uτ ∈ dz
0≤s≤τ

 e−z =2 S ( + p ; r ; z+x+2|z−x| )S ( + p ; z+x−|z−x| ; a )

√ 2 2

z≤r = √ 2  dz
 2S ( + p ; r ; a )
x≤r

 e−z =2 S ( +
√ q b z+x+|z−x| )S ( +q ; z+x−|z−x| ; r )
 ; √;
2 2

r≤z = 2  2  dz
 2S ( + q; b; r)
  
r≤x
548 7. ORNSTEIN{UHLENBECK PROCESS

7 Us = σe−θs We2θs −1 σ, θ > 0 Hz = min{s : Us = z}

2. Stopping at first hitting time

 x2 =42 x
e D− = (−  )
 ez2 =42 D

z , x≤z
= (−  )

−
2.0.1 Ex e−αHz = x
x = 4
e D = (  )
2 2
−

z , z≤x

ez2 =42 D− = (  )


| x|

 
2.0.2 Px H0 > t = Erf
 2(e2t − 1)
p
(1)
Er d x ; z




2.1.2 Px sup Us ≥ y =
, z≤x≤y
0≤s≤Hz Er d y ; z
z≤y
D− = (− x )
 x2 =42
e
, x≤z

D− = (− z )
 z2 =42
e

Ex e−αHz ; sup Us ≤ y,

2.1.4 =
z≤y 0≤s≤Hz  S (  ; y ; x )
, z≤x≤y
S (  ; y ; z )



Er d z ; x


2.2.2 Px inf Us ≤ y =
, y ≤ x ≤ z
0≤s≤Hz Er d z ; y
y≤z
 S( ; x ; y )
   y≤x≤z
 S( ; z ; y ) ,


 −αHz   
2.2.4 Ex e ; inf Us ≥ y =
ex =4 D (x)
2 2

 z2 =42 − = z , z ≤ x
0≤s≤Hz
y≤z


e D− = (  )

1, x ≤ z ≤ r, r ≤ z ≤ x

√
 e−r2 =22 | Er d x ; r
|

1 +

 √
√ 2  



, r∧z ≤x≤r∨z

2.3.1 Ex e−γ`(Hz ,r) = 
1 +  2 e
√ −r 2 =2 2
| Er d z ; r |
1




, z ≤ r ≤ x, x ≤ r ≤ z
 √
1 + √2 e−r2 =22 | Er d z ; r |



2.3.2 Px `(Hz , r) ∈ dy

0, x ≤ z ≤ r, r ≤ z ≤ x

z x

  2=| Er d  ;  |

 p

 p2=er2 =22 y 
dy, r∧z ≤x≤r∨z
= exp − z r e−r2 =22 Er d2 z ; r
| Er d  ;  |


√ r2 =22
 √ 2e




dy, z ≤ r ≤ x, x ≤ r ≤ z

| Er d z ; r |
 2. STOPPING AT FIRST HITTING TIME 549

7 Us = σe−θs We2θs −1 σ, θ > 0 Hz = min{s : Us = z}


 1, x ≤ z ≤ r, r≤z≤x
x r



 | Er d  ; 
| ,
(1) Px `(Hz , r) = 0 = z r r∧z ≤x≤r∨z

 | Er d ;  |
z ≤ r ≤ x, x≤r≤z

0,

2.3.3 Ex e−αHz −γ`(Hz ,r)

ex =4 D− = ( x )
 2 2

z
√
 r r z , r≤x
ez2 =42 D− = (  ) + √2 e−r =4 D− = (  )S (  ;  ;  )

 2 2



√
ex =4 D− = ( x ) + √2 e−r =4 D− = ( r )S (  ; r ; x )


 2 2 2 2

z≤r = √ , z≤x≤r


 ez2 =42 D− = ( z ) + √2 e−r2 =42 D− = ( r )S (  ; r ; z )
x

x2 =42 D
 e − = (−  )



, x≤z

e z 2 =4 2 z
D− = (−  )

ex =4 D− = ( x )
 2 2

, z≤x
ez2 =42 D− = ( z )





√
ex =4 D− = (− x ) +  D− = (− r )S (  ; x ; r )
√ −r =4

e

 2 2 2 2
√ 2

r≤z =  , r≤x≤z

 ez2 =42 D− = (− z ) +  2 e 2
D− = (− r )S (  ; z ; r )
√ −r =4
2

ex =4 D− = (− x )

2 2



, x≤r

√
ez =4 D− = (− z ) + √2 e−r2 =42 D− = (− r )S (  ; r ; z )

 2 2

Ex e−αHz ; `(Hz , r) ∈ dy

2.3.4

 √ (x2 +r2 )=42

 2e D− = ( x )
√ dy, r ≤ x
S (  ;  ;  )D− = ( r )
r z


 √2e(z2 +r2 )=42 D z 

− = (  )y

√ r2 =22
z≤r = exp − √
D− = ( r )S (  ; r ; z )  2e S (  ; x ; z )
√ 2 dy, z ≤ x ≤ r
S (  ; r ; z )





0, x≤z

0, z≤x

 √ r2 =22 z x
 2e S ( ; ; )

 e(z2 +r2 )=42 D z     dy, r ≤ x ≤ z

− = (−  )y

√ 2
r≤z = exp − p r z r S (  ; z ; r )
=2D− = (−  )S (  ;  ;  ) 
 √2e(x2 +r2 )=42 D− = (− x )


 √  dy, x ≤ r
S (  ; z ; r )D− = (− r )
550 7. ORNSTEIN{UHLENBECK PROCESS

7 Us = σe−θs We2θs −1 σ, θ > 0 Hz = min{s : Us = z}

0 r≤x

r x
 S(  ;  ;  ) ,



z≤x≤r

S (  ;  ; z )
r
 −αHz
(1) Ex e ; `(Hz , r) = 0 =
x2 =42 D x
 e − = (−  )
z≤r



, x≤z
ez =4 D− = (− z )
 2 2

0 x≤r

S (  ; x ; r )



r≤x≤z

,
S (  ; z ; r )
 −αHz 
r≤z Ex e ; `(Hz , r) = 0 =
(x)
 x2 =42
e D
 2 2 − = z ,



z = 4 z≤x
e D− = (  )

 Z Hz

2.4.1 Ex exp −γ 1I[r,∞) Us ds
0
1, x≤z

√

r2 =22 D ( r ) + √ Er d r ; x
D ( r)

e − =    −(+ )= 

√ 2


r , z≤x≤r

r 2 =2 2 r ) + √ Er d r ; z
D

z≤r = e D − = (  2   −(+ )= (  )
e(x +r )=4 D− = ( x )

 2 2 2


 √ , r≤x
D− = ( r ) + √2 Er d r ; z D−(+ )= ( r )
 r2 =22

e

2=er =2
 p 2 2

x≤r
C (  ; r ; z )




 C( ; r ; x )


   , r≤x≤z
r≤z = C (  ; r ; z )

 ex =4 D x

 2 2 − = (  ) , z ≤ x
 2 2


e z = 4 D− = ( z )


 1, x≤z
Z Hz   Er d r ; x

 
, z ≤ x ≤ r


2.4.2 Px 1I[r,∞) Us ds = 0 =
0 Er d r ; z
(1)


r≤x

z≤r 0,

 Z Hz

2.5.1 Ex exp −γ 1I(−∞,r] Us ds
0
D− = (− x )
 x2 =42
e
, x≤z
z 2 =4 2
D− = (− z )

e



C (  ; r ; x )



z≤r = , z≤x≤r

 C (  ; r ; z )
r2 =22
 2=e

 p

r≤x

C (  ; r ; z )
 2. STOPPING AT FIRST HITTING TIME 551

7 Us = σe−θs We2θs −1 σ, θ > 0 Hz = min{s : Us = z}

e(x +r )=4 D− = (− x )
 2 2 2
 √ , x≤r
er2 =22 D− = (− r ) + √2 Er d z ; r D−(+ )= (− r )






√

er =2 D− = (− r ) + √2 Er d x ; r D−(+ )= (− r )
2 2
r≤z =


√ , r≤x≤z
r2 =22 D− = (− r ) + √ Er d z ; r
D−(+ )= (− r )

e

   

 2




1, z≤x


0, x≤r
 Er d x ; r


Z Hz 
 
, r ≤ x ≤ z


2.5.2 Px 1I(−∞,r] Us ds = 0 = z; r
0 Er d  
(1)


z≤x

r≤z 1,

 Z Hz



2.6.1 Ex exp − p1I(−∞,r] Us + q1I[r,∞) Us ds
0

ex =4 D x
−p= (−  )
2 2

, x≤z

ez 2 =4 2
D−p= (−  ) z




D−q= ( r )C ( p ; r ; x ) + q D−(+q)= ( r )S ( p ; r ; x )



z≤r = , z≤x≤r

 D−q= ( r )C ( p ; r ; z ) + q D−(+q)= ( r )S ( p ; r ; z )
2=e(x +r )=4 D−q= ( x )

 p 2 2 2

r≤x

,
D−q= (  )C (  ;  ;  ) +  D−(+q)= ( r )S ( p ; r ; z )
r p r z q



2=e(x +r )=4 D−p= (− x )

2 2 2
 p
, x≤r
C ( q ; r ; z )D−p= (− r ) + p S ( q ; z ; r )D−(+p)= (− r )





 C ( q ; r ; x )D r p q x r r

   −p= (−  ) +  S (  ;  ;  )D−(+p)= (−  ) , r ≤ x ≤ z

r≤z = q p q
C (  ; r ; z )D−p= (− r ) +  S (  ; z ; r )D−(+p)= (− r )


ex =4 D (x)
2 2

 z2 =42 −q= z ,


z≤x

D−q= (  )

e

D− 1 + 1 − α (− x  )
 x2 =42 √
 e
2 2γ θγ
, x≤z

 ez2 =42 D− 1 + 1 − α (− z  )

 √

( 2 − 1) Hz 2 
 Z 
2 2γ θγ
2.9.3 Ex exp −αHz − Us ds =
42 D− 1 + 1 − α ( x  )
√
0 ex 2
= 4 2
1≤γ

 2 2γ θγ
z√ , z≤x


 ez2 =42 D
−1+ 1 − α (  )

2 2γ θγ

( 2 − 1) H0 2 
n  Z o
2.9.4 Ex exp − 2 Us ds ; H0 ∈ dt
4 0
(1)
|x| 3=2 t x2 x2 ch(t )
 
= √ exp + − dt
2  sh3=2 (t ) 2 42 42 sh(t )
552 7. ORNSTEIN{UHLENBECK PROCESS

7 Us = σe−θs We2θs −1 σ, θ > 0 Hz = min{s : Us = z}

Z H0 
(2) Px Us2 ds ∈ dy, H0 ∈ dt
0

|x| t x2 − y 3 3 x2

 
= √ exp + 2 esy/2σ2 θ , , t, 0, 2 dydt
2(22 )3=2 2 4 2 2 4 
√
2 ey =2 S  ; x ; z
2 2

−γ Ȟ(Hz )

2.13.1 Ex e ; sup Us ∈ dy =
dy
  Er d y ; z S  ; y ; z
√

0≤s≤Hz
z<x
x<y
√
2 ey =2 S + ; x ; z
2 2

−αHz−γ Ȟ(Hz )

2.13.4 Ex e ; sup Us ∈ dy = √ y z
+ y z
dy
0≤s≤Hz  S  ;  ;  S  ; ; 
z<x
x<y
√
2 ey =2 S  ; z ; x
2 2

−γ Ĥ(Hz )

2.14.1 Ex e ; inf Us ∈ dy = √
dy
  Er d z ; y S  ; z ; y


0≤s≤Hz
x<z
y<x
√
2 ey =2 S + ; z ; x
2 2

−αHz−γ Ĥ(Hz )

2.14.4 Ex e ; inf Us ∈ dy = √ z y
+ z y
dy
0≤s≤Hz  S  ;  ;  S  ; ; 
x<z
y<x

Hz
ds
n  Z  o
2.20.3 Ex exp −αHz − p2 σ 2 θ 2 ; 0 < inf Us
0
Us 0≤s≤Hz

 −1=2 x2 =42
x e M θ−2α ; 1 √1+4p2 x2

 22
4θ 4
z2
, 0<x≤z

 z −1=2 ez =4 M θ−2α ; 1 √1+4p2

 2 2
22

4θ 4
=
x−1=2 ex =4 W θ−2α ; 1 √1+4p2
2 2 x2

22


4θ 4
z2
, z≤x

 z −1=2 ez2 =42 W

√

; 1+4p2
θ−2α 1
4θ 4
22

p2 2  (q2 − 1) 2
n  Z Hz    o
2.21.3 Ex exp −αHz − + U ds ; 0 < inf U
0
Us2 42 s
0≤s≤Hz
s


x−1=2 ex =4 M θ−2α ; 1 √1+4p2
2 2 qx2


 4θq 4
22

qz2
, 0<x≤z
 z −1=2 ez2 =42 M θ−2α ; 1 √1+4p2


22

4θq 4
=
x−1=2 ex =4 W θ−2α ; 1 √1+4p2
2 2 qx2

22


4θq 4


 z −1=2 ez2 =42 W θ−2α 1 √

 qz2
, z≤x
; 1+4p24θq 4
22
 2. STOPPING AT FIRST HITTING TIME 553

7 Us = σe−θs We2θs −1 σ, θ > 0 Hz = min{s : Us = z}

n  Z Hz



o
2.27.1 Ex exp − p1I(−∞,r) Us + q1I[r,∞) Us ds ; ` Hz , r ∈ dy
0

 C ( p ; r ; z ) qD r 
   + −( +q )= (  )

z≤r = exp −yθσ p r z r
S(  ;  ;  ) D−q= (  )

 √ r2 =22 p x z
 2e S( ; ; )
 √ 2 p r  z   dy, z≤x≤r
S (  ;  ;  )


× √ (x2 +r2 )=42
 √ 2e
 D−q= ( x )
dy, r ≤ x
D−q= ( r )S ( p ; r ; z )


 C ( q ; r ; z ) pD r 
   + −( +p)= (−  )

r≤z = exp −yθσ q z r r
S(  ;  ;  ) D−p= (−  )

 √ (x2 +r2 )=42

 2e D−p= (− x )
 √D r )S ( q ; z ; r ) dy, x ≤ r

(

−p= −    
× √ r2 =22 q z x
 2e S ( ; ; )
√ 2 q z  r   dy,

r≤x≤z

S (  ;  ;  )


n  Z Hz



o
(1) Ex exp − p1I(−∞,r) Us + q1I[r,∞) Us ds ; ` Hz , r = 0
0

ex =4 D−p= (− x )
 2 2

, x≤z≤r
ez2 =42 D−p= (− z )





S ( p ; r ; x )




, z≤x≤r
S ( p ; r ; z )



= q
S (  ; x ; r )
, r≤x≤z
S ( q ; z ; r )






ex =4 D (x)

 2 2

 2 2 −q= z ,



z = 4 r≤z≤x
e D−q= (  )
554 7. ORNSTEIN{UHLENBECK PROCESS

7 Us = σe−θs We2θs −1 σ, θ > 0 H = Ha,b = min{s : Us ∈

/ (a, b)}

3. Stopping at first exit time

S (  ; b ; x ) + S (  ; x ; a )
3.0.1 Ex e−αH = , a≤x≤b
S (  ; b ; a )

Er d b ; x −µa Er d x ; a −µb



3.0.3 Ex e−µUH =
e +
e , a≤x≤b
Er d b ; a Er d b ; a

Er d b ; x


3.0.4 Px UH = a =
, a≤x≤b
Er d b ; a
(a)

Er d x ; a


3.0.4 Px UH = b =
, a≤x≤b
Er d b ; a
(b)

S( ; b ; x )
Ex e−αH ; UH = a =  b a ,

3.0.5 a≤x≤b
S(  ;  ;  )
(a)

S( ; x ; a )
Ex e−αH ; UH = b =  b a ,

3.0.5 a≤x≤b
S(  ;  ;  )
(b)

Er d b ; y Er d x ; a



3.1.6 Px sup Us ≥ y, UH = a =
, a≤x≤y≤b
Er d b ; a Er d y ; a


0≤s≤H

Ex e−αH ; sup Us ≥ y, UH = a

3.1.8
0≤s≤H
(1)
S (  ; b ; y )S (  ; x ; a )
= , a≤x≤y≤b
S (  ; b ; a )S (  ; y ; a )

Er d b ; x Er d y ; a





3.2.6 Px inf Us ≤ y, UH = b =
, a≤y≤x≤b
Er d b ; y Er d b ; a


0≤s≤H

Ex e−αH ; inf

3.2.8 Us ≤ y, UH = b
0≤s≤H
(1)
S (  ; b ; x )S (  ; y ; a )
= , a≤y≤x≤b
S (  ; b ; y )S (  ; b ; a )
 3. STOPPING AT FIRST EXIT TIME 555

7 Us = σe−θs We2θs −1 σ, θ > 0 H = Ha,b = min{s : Us ∈

/ (a, b)}
√
Er d b ; a + √2 e−r =2 Er d x ; r Er d r ; a

2 2


, r≤x≤b

√
 Er d b ; a
+ √ e−r2 =22 Er d b ; r
Er d r ; a




   2    
3.3.1 Ex e−γ`(H,r) = √
Er d b ; a + √2 e−r =2 Er d b ; r Er d r ; x

2 2




√
, a ≤ x ≤ r

Er d b ; a + √2 e−r2 =22 Er d b ; r Er d r ; a



 √2er2 =22 Er d b ; a
y 
3.3.2


Px `(H, r) ∈ dy = exp − √  

 Er d b ; r Er d r ; a

 √ r2 =22
 2e Er d b ; x Er d b ; a



√ dy, r ≤ x ≤ b
 Er d2 b ; r Er d r ; a





× √ r2 =22
2e
  √ Er d x ; a Er d b ; a




dy, a ≤ x ≤ r

 Er d b ; r Er d2 r ; a

Er d x ; r



, r ≤ x ≤ b

 Er d b ; r


(1) Px `(H, r) = 0 =
Er d r ; x



, a ≤ x ≤ r

Er d r ; a



S (  ; b ; r ) + S (  ; r ; a )
3.3.3 Er e−α m −γ`(H,r) = √ =: L
S (  ;  ;  ) + √2 e−r2 =22 S (  ; b ; r )S (  ; r ; a )
b a

S (  ; x ; r ) S (  ; b ; x )

 S ( ; b ; r ) + S ( ; b ; r ) L, r ≤ x ≤ b


Ex e−α m −γ`(H,r) =      
S (  ; r ; x ) S (  ; x ; a )
L, a ≤ x ≤ r

+
S (  ; r ; a ) S (  ; r ; a )



 √2er2 =22 S ( ; b ; a )y 
3.3.4 Ex e−αH ; `(H, r) ∈ dy = exp − √
   
S (  ; b ; r )S (  ; r ; a )
 √ r2 =22
 2e S (  ; b ; x ) S (  ; b ; r ) + S (  ; r ; a )
h i

dy, r ≤ x ≤ b

√ 2
S (  ; b ; r )S (  ; r ; a )



× √ 2
 2er =2 S (  ; x ; a ) S (  ; b ; r ) + S (  ; r ; a )
2
h i


√ dy, a ≤ x ≤ r

S (  ; b ; r )S 2 (  ; r ; a )


 S( ; x ; r )
  
 S( ; b ; r ) , r ≤ x ≤ b

(1)
 −αH
Ex e ; `(H, r) = 0 =   
( r x
 S  ; ), a ≤ x ≤ r
;
S (  ;  ; a )
r

556 7. ORNSTEIN{UHLENBECK PROCESS

7 Us = σe−θs We2θs −1 σ, θ > 0 H = Ha,b = min{s : Us ∈

/ (a, b)}

Ex e−γ`(H,r) ; UH = a

3.3.5
(a)
Er d b ; x


√
, r ≤ x ≤ b
 Er d b ; a + √2 e−r2 =22 Er d b r

 ;  Er d r ; a




= √
 Er d b ; x + √2 e−r =2 Er d b r

 ;  Er d r ; x

2 2

, a ≤ x ≤ r

√
Er d b ; a + √2 e−r2 =22 Er d b r
d r ; a

 ;  Er


Ex e−γ`(H,r) ; UH = b

3.3.5
(b)
√
Er d x ; a + √2 e−r =2 Er d x r

 ;  Er d r ; a

2 2

, r ≤ x ≤ b

 Er d b ; a
+ √√ e−r2 =22 Er d b r
d r ; a

 ;  Er

=    2
Er d x ; a




, a ≤ x ≤ r
 √
Er d b ; a + √2 e−r2 =22 Er d b r
d r ; a

 ;  Er


 √2er2 =22 Er d b ; a
y 
3.3.6


Px `(H, r) ∈ dy, UH = a = exp − √  

 Er d b ; r Er d r ; a


(a)
√ 2
 2er =2 Er d b ; x
 2

  Er d b ; r Er d r ; a
dy, r ≤ x ≤ b
√



×    
√ r2 =22
2e
  √
x
Er d  ; a



dy, a≤x≤r

 Er d2 r ; a


 0, r≤x≤b
Er d r x

; 
,


(1) Px `(H, r) = 0, UH =a =
r a a≤x≤r
Er d ; 


 √2er2 =22 Er d b ; a
y 
3.3.6


Px `(H, r) ∈ dy, UH = b = exp − √  

 Er d b ; r Er d r ; a


(b)
 √ r2 =22
 2e Er d b ; x


√ dy, r≤x≤b
 Er d2 b ; r




× √ r2 =22
 √ 2e Er d x ; a




dy, a ≤ x ≤ r

 Er d b ; r Er d r ; a

Er d x ; r



, r ≤ x ≤ b
Er d b ; r


(1) Px `(H, r) = 0, UH = b =
0, a≤x≤r

 3. STOPPING AT FIRST EXIT TIME 557

7 Us = σe−θs We2θs −1 σ, θ > 0 H = Ha,b = min{s : Us ∈

/ (a, b)}

Ex e−α m −γ`(H,r) ; UH = a

3.3.7
(a)
 S (  ; b ; x )
√
 , r≤x≤b
b a
 S (  ;  ;  ) + √2 e−r2 =22 S (  ; b ; r )S (  ; r ; a )



= √
 S (  ; b ; x ) + √2 e−r =2 S (  ; b ; r )S (  ; r ; x )
2 2

, a≤x≤r

√
S (  ; b ; a ) + √2 e−r2 =22 S (  ; b ; r )S (  ; r ; a )



Ex e−α m −γ`(H,r) ; UH = b

3.3.7
(b)
√
S (  ; x ; a ) + √2 e−r =2 S (  ; x ; r )S (  ; r ; a )
 2 2

 S ( ; b ; a ) + √ e−r2 =22 S ( ; b ; r )S ( ; r ; a ) , r ≤ x ≤ b


 √

=     2      
S ( ; x; a)
  

, a≤x≤r
 √
S (  ; b ; a ) + √2 e−r2 =22 S (  ; b ; r )S (  ; r ; a )



Ex e−αH ; `(H, r) ∈ dy, UH = a


3.3.8
(a)
 √ r2 =22
 √ 2e S (  ; b ; x )
√
  2er =2 S ( ; b ; a )y 
2 2  S ( ; b ; r )S ( ; r ; a ) dy, r ≤ x ≤ b
= exp − √         
S (  ; b ; r )S (  ; r ; a )  √ r2 =22
2e S (  ; x ; a )
  √

dy, a ≤ x ≤ r
S (  ; r ; a )
2


 0, r≤x≤b
Ex e−αH ; `(H, r) = 0, UH S (  ; r ; x )

(1) =a =
, a≤x≤r
S (  ; r ; a )


Ex e−αH ; `(H, r) ∈ dy, UH = b


3.3.8
(b)
 √ r2 =22
  √2e S (  ; b ; x )
√ dy, r ≤ x ≤ b
  2er =2 S ( ; b ; a )y 
2 2
  
 S (  ; b ; r )
2
= exp − √
S (  ; b ; r )S (  ; r ; a )  √ r2 =22
S (  ; x ; a )
 √ 2e

dy, a ≤ x ≤ r
S (  ;  ;  )S (  ; r ; a )
b r

x r
 S(  ;  ;  ) , r ≤ x ≤ b

S (  ; b ; r )
 −αH
(1) Ex e ; `(H, r) = 0, UH = b =
0, a≤x≤r

558 7. ORNSTEIN{UHLENBECK PROCESS

7 Us = σe−θs We2θs −1 σ, θ > 0 H = Ha,b = min{s : Us ∈

/ (a, b)}

 Z H

3.4.1 Er exp −γ 1I[r,b] Us ds
0

S (  ; b ; r ) + Er d r ; a


= √ =: U
S (  ;  ;  ) + √2 e−r2 =22 Er d r ; a C (  ; r ; b )
b r

Er d r ; x Er d x ; a



+
U, a ≤ x ≤ r

  Er d r ; a Er d r ; a

H
 Z 

Ex exp −γ 1I[r,b] Us ds =
S( ; x ; r ) S( ; b ; x )
0   b r +  b r U, r ≤ x ≤ b


S(  ;  ;  ) S(  ;  ;  )

r x

  Er d  ; 
, a ≤ x ≤ r

Z H
Er d r ; a


3.4.2 Px 1I[r,b] Us ds = 0 =
(1) 0 
0, r≤x≤b

n  Z H
 o
3.4.5 Ex exp −γ 1I[r,b] Us ds ; UH = a
(a) 0

√

S (  ; b ; r ) + √
 −r2 =22
e Er d r ; x C (  ; r ; b )


 √
2 , a≤x≤r
 S( ; b ; r ) +  −r2 =22 Er d r ; a C (  ; r ; b )

2e


√
  
=
 S (  ; b ; x )
, r≤x≤b

√
S (  ; b ; r ) +  −r2 =22 Er d r ; a C (  ; r ; b )

2e


√

n  Z H
 o
3.4.5 Ex exp −γ 1I[r,b] Us ds ; UH = b
(b) 0

Er d x ; a


√
 −r2 =22 , a≤x≤r
 S (  ; b ; r ) + Er d r ; a C (  ; r ; b )

2e


 √

= x r
√

√ e−r =2 Er d r ; a C ( ; r ; x )
 S(  ;  ;  ) +
2 2

 2     
√
 −r2 =22 , r≤x≤b
S (  ; b ; r ) + Er d r ; a C (  ; r ; b )

2e


√

r x

  Er d  ; 
, a ≤ x ≤ r

Z H
Er d r ; a


3.4.6 Px 1I[r,b] Us ds = 0, UH =a =
(1) 0 
0, r≤x≤b

 Z H

3.5.1 Er exp −γ 1I[a,r] Us ds
0
 3. STOPPING AT FIRST EXIT TIME 559

7 Us = σe−θs We2θs −1 σ, θ > 0 H = Ha,b = min{s : Us ∈

/ (a, b)}

S (  ; r ; a ) + Er d b ; r


= √ =: U
S (  ;  ;  ) + √2 e−r2 =22 Er d b ; r C (  ; r ; a )
r a

S (  ; r ; x ) S (  ; x ; a )

+ U, a ≤ x ≤ r

  S (  ; r ; a ) S (  ; r ; a )

 Z H 
Ex exp −γ 1I[a,r] Us ds =
Er d x ; r Er d b ; x



0

+
U, r ≤ x ≤ b

Er d b ; r Er d b ; r




Z H   0, a≤x≤r
Er d x ; r


3.5.2 Px 1I[a,r] Us ds = 0 =



, r ≤ x ≤ b
(1) 0
Er d b ; r


n  Z H
 o
3.5.5 Ex exp −γ 1I[a,r] Us ds ; UH = a
(a) 0

√

S (  ; r ; x ) + √
 −r2 =22
e Er d b ; r C (  ; r ; x )


 √
2 , a≤x≤r
 S( ; r ; a ) +  −r2 =22 Er d b ; r C (  ; r ; a )

2e


√
  
=
Er d b ; x



, r≤x≤b

√
S (  ; r ; a ) +  −r2 =22 Er d b ; r C (  ; r ; a )

2e


√

n  Z H
 o
3.5.5 Ex exp −γ 1I[a,r] Us ds ; UH = b
(b) 0

 S (  ; x ; a )
√
 −r2 =22 , a≤x≤r
 S (  ; r ; a ) + 2e Er d b ; r C (  ; r ; a )



 √

= √

 S (  ; r ; a ) + √ e−r =2 Er d x ; r C ( ; r ; a )
2 2

2     
, r≤x≤b
 √
S (  ; r ; a ) +  r 2 =2 2 b
Er d  ;  C (  ; r ; a )
r

√ e
 −


2


Z H  0, a≤x≤r
Er d x ; r


3.5.6 Px 1I[a,r] Us ds = 0, UH = b =



, r ≤ x ≤ b
(1) 0
Er d b ; r


 Z H



3.6.1 Er exp − p1I[a,r] Us + q1I[r,b] Us ds =: U
0

√
2er =2 (S ( q ; b ; r ) + S ( p ; r ; a ))
2 2

=
 S ( q ; b ; r )C ( p ; r ; a ) + C ( q ; r ; b )S ( p ; r ; a )
√

560 7. ORNSTEIN{UHLENBECK PROCESS

7 Us = σe−θs We2θs −1 σ, θ > 0 H = Ha,b = min{s : Us ∈

/ (a, b)}

 Z H

 
Ex exp − p1I[a,r] Us + q1I[r,b] Us ds
0
 q x r
S (  ;  ;  ) S ( q ; b ; x )
 S ( q ; b ; r ) + S ( q ; b ; r ) U, r ≤ x ≤ b


=      
S ( p ; r ; x ) S ( p ; x ; a )
 p r a + p r a U, a ≤ x ≤ r

S(  ;  ;  ) S(  ;  ;  )


n  Z H

  o
3.6.5 Ex exp − p1I[a,r] Us + q1I[r,b] Us ds ; UH = a
(a) 0
√
2er =2 S ( q ; b ; x )
 2 2

, r≤x≤b
(S ( q ; b ; r )C ( p ; r ; a ) + C ( q ; r ; b )S ( p ; r ; a ))
√



=
S ( q ; b ; r )C ( p ; r ; x ) + C ( q ; r ; b )S ( p ; r ; x )
 q b r

a≤x≤r
S (  ;  ;  )C ( p ; r ; a ) + C ( q ; r ; b )S ( p ; r ; a )


n  Z H

  o
3.6.5 Ex exp − p1I[a,r] Us + q1I[r,b] Us ds ; UH = b
(b) 0
 S ( q ; x ; r )C ( p ; r ; a ) + C ( q ; r ; x )S ( p ; r ; a )
            r≤x≤b
 S ( q ; b ; r )C ( p ; r ; a ) + C ( q ; r ; b )S ( p ; r ; a ) ,


           
= √
2er =2 S ( p ; x ; a )
2 2

, a≤x≤r
(S ( q ; b ; r )C ( p ; r ; a ) + C ( q ; r ; b )S ( p ; r ; a ))
√



( 2 − 1) H 2 
 Z
3.9.1 Ex exp − Us ds
42 0
1≤γ
S 12 − 21 ; b  ; x  + eb ( −1)=4 S 21 − 21 ; x  ; a 
√ √
√ √

ea ( −1)=4 2
2 2 2

=
ex2 ( −1)=42 S 12 − 21 ; b  ; a 
√ √

( 2 − 1) H 2 
n  Z o
3.9.7 Ex exp −αH − U ds ; U = a γ≥1
42 0
s H
(a)

S 12 − 21 +  ; b  ; x 
√ √

ea ( −1)=4 2
2

= 2
ex ( −1)=42 S 12 − 21 +  ; b  ; a 
√ √

( 2 − 1) H 2 
n  Z o
3.9.7 Ex exp −αH − U ds ; U = b γ≥1
42 0
s H
(b)
S 12 − 21 +  ; x  ; a 
√ √

eb ( −1)=4 2
2

=
S 12 − 21 +  ; b  ; a 
√ √

ex2 ( −1)=4 2
 3. STOPPING AT FIRST EXIT TIME 561

7 Us = σe−θs We2θs −1 σ, θ > 0 H = Ha,b = min{s : Us ∈

/ (a, b)}

√
b
2 ey =2 S  ; x ; a
Z 2 2

Ex e−γ Ȟ(H) ; UH = a =

3.12.5
dy
  Er d y ; a S  ; y ; a
√

(a) x

√
x
2 ey =2 S  ; b ; x
Z 2 2

−γ Ĥ(H)

3.12.5 Ex e ; UH = b =
dy
  Er d b ; y S  ; b ; y
√

(b) a

√
b
2 ey =2 S + ; x ; a
Z 2 2

−αH−γ Ȟ(H)

3.12.7 Ex e ; UH = a = √ y a
+ y a
dy
x  S  ;  ;  S  ; ; 
(a)

√
x
2 ey =2 S + ; b ; x
Z 2 2

−αH−γ Ĥ(H)

3.12.7 Ex e ; UH = b = √ b y + b y
dy
 S  ;  ;  S  ; ; 


(b) a

H
ds
n  Z  o
2 2
3.20.5 Ex exp −γ σ θ ; U = a
0
Us2 H
(a)
√
x +1=4+1=2 S 1 p4 2 + 1 +
2 1 ; 1 p4 2 + 1 + 1; b2 ; x2 )
= √ 2 4 4 2 42 42 , 0<a≤x≤b
a +1=4+1=2 S 1 p4 2 + 1 + 1 ; 1 p4 2 + 1 + 1; b2 ; a2 )
4 4 2 42 42
H
ds
n  Z  o
3.20.5 Ex exp −γ 2 σ 2 θ ; U = b
0
Us2 H
(b)
√
x +1=4+1=2 S 1 p4 2 + 1 + 1 ; 1 p4 2 + 1 + 1; x2 ; a2 )
2

= √ 2 4 4 2 42 42 , 0<a≤x≤b

b +1=4+1=2 S 1 p4 2 + 1 + 1 ; 1 p4 2 + 1 + 1; b22 ; a22 )
4 4 2 4 4

p2 2  (q2 − 1) 2
n  Z H   o
3.21.5 Ex exp − + U ds ; U = a
0
Us2 42 s H
(a)
√
x p2 +1=4+1=2 ex (1−q)=4 S 14 4p2 + 1 + 2q4−q 1 ; 12 4p2 + 1 + 1; 4qb2 ; 4qx2 )
2 2 p p 2 2

0<a = √2
a p +1=4+1=2 ea2 (1−q)=42 S 14 4p2 + 1 + 2q4−q 1 ; 12 4p2 + 1 + 1; 4qb2 ; 4qa2 )
p p 2 2

p2 2  (q2 − 1) 2
n  Z H   o
3.21.5 Ex exp − + U ds ; U = b
0
Us2 42 s H
(b)
√
x p2 +1=4+1=2 ex (1−q)=4 S 14 4p2 + 1 + 2q4−q 1 ; 12 4p2 + 1 + 1; 4qx2 ; 4qa2 )
2 2 p p 2 2

0<a = √2
b p +1=4+1=2 eb2 (1−q)=42 S 14 4p2 + 1 + 2q4−q 1 ; 12 4p2 + 1 + 1; 4qb2 ; 4qa2 )
p p 2 2
562 7. ORNSTEIN{UHLENBECK PROCESS

7 Us = σe−θs We2θs −1 σ, θ > 0 H = Ha,b = min{s : Us ∈

/ (a, b)}

n  Z H



o
3.27.1 Ex exp − p1I[a,r) Us + q1I[r,b] Us ds ; `(H, r ∈ dy
0

 C ( p ; r ; a ) C ( q ; r ; b ) 
   +   

= exp −yθσ
S ( p ; r ; a ) S ( q ; b ; r )

 √ r2 =2 q b x h q b r
 2e S (  ;  ;  ) S (  ;  ;  ) + S ( p ; r ; a )
i

dy, r ≤ x ≤ b

√ 2 q b r
S (  ;  ;  )S ( p ; r ; a )



× √ 2 2
 2er = S ( p ; x ; a ) S ( q ; b ; r ) + S ( p ; r ; a )
h i


dy, a ≤ x ≤ r
S ( q ; b ; r )S 2 ( p ; r ; a )
√



n  Z H



o
3.27.5 Ex exp − p1I[a,r) Us + q1I[r,b] Us ds ; `(H, r ∈ dy, UH = a
(a) 0

√ 2 2
 2er = S ( q ; b ; x )

 S ( q ; b ; r )S ( p ; r ; a ) dy, r ≤ x
√

 C ( p ; r ; a ) C ( q ; r ; b ) 


            x≤b
= exp −yθσ +
S ( p ; r ; a ) S ( q ; b ; r )  
√ r2 =2 p x a
2e S ( ; ; )
√ 2 p r  a  dy, a≤x

S (  ;  ;  )


x≤r

n  Z H



o
(1) Ex exp − p1I[a,r) Us + q1I[r,b] Us ds ; ` H, r = 0, UH = a
0
S ( p ; r ; x )
= , a≤x≤r
S ( p ; r ; a )

n  Z H



o
3.27.5 Ex exp − p1I[a,r) Us + q1I[r,b] Us ds ; ` H, r ∈ dy, UH = b
(b) 0
 √ r2 =2 q b x
 2e S(  ;  ;  )
 √ 2 q b r dy, r≤x
 C ( p ; r ; a ) C ( q ; r ; b )  S (  ;  ;  )


   +   
 x≤b
= exp −yθσ
S ( p ; r ; a ) S ( q ; b ; r )  √ r2 =2 p x a
 2e S(  ;  ;  )
dy, a ≤ x

S ( q ; b ; r )S ( p ; r ; a )
√

x≤r

n  Z H



o
(1) Ex exp − p1I[a,r) Us + q1I[r,b] Us ds ; ` H, r = 0, UH = b
0

S ( q ; x ; r )
= , r≤x≤b
S ( q ; b ; r )
 4. STOPPING AT INVERSE LOCAL TIME 563

7 Us = σe−θs We2θs −1 σ, θ > 0 % = %(v, z) = min{s : `(s, z) = v}

4. Stopping at inverse local time

√
D− = (− x )
 x2 =42
e 
 2v

exp − , x≤z
 ez2 =42 D z ( =)D− = (− z )D− = ( z )

− = (−  )

4.0.1 Ex e−α% = x √
x2 =42
 e 2 2 D− = (  ) exp −  2v
  
z z , z≤x

e z = 4 z
D− = (  ) ( =)D− = (−  )D− = (  )

√
 2ez =2 v
  2 2 
 exp − √ y z , x≤z
 Er d(  ;  )




4.1.2 Px sup Us < y =
Er d y ; x
√
 2ez =2 v

 2 2 
0≤s≤%
z≤y


y z
exp − √ y z , z≤x≤y
Er d  ;   Er d(  ;  )


Ex e−α% ; sup Us ≤ y

4.1.4
0≤s≤%
z≤y

 x2 =42
D− = (− x )  √2e(z2 +y2 )=42 D y 
e − = (−  )v
exp − √ y , x≤z
 ez2 =42 D z D− = (− z )S (  ;  ; z )

− = (−  )

= √ (z2 +y2 )=42
y x
 S (  ; y ;  ) exp −  √2e
  D− = (− y )v 
, z≤x≤y
D− = (− z )S (  ; y ; z )

S (  ;  ; z )

Er d x ; y
√
 2ez =2 v

 2 2 
exp − , y≤x≤z

 Er d z ; y  Er d( z ; y )
√



4.2.2 Px inf Us ≤ y = √
 2ez =2 v
2 2
0≤s≤%  
y≤z

 exp − √ y , z≤x
 Er d( z ;  )


Ex e−α% ; inf Us ≥ y

4.2.4
0≤s≤%
y≤z

S (  ; x ; y )  √2e(z2 +y2 )=42 D y 

− = (  )v

 S ( ; z ; y ) exp − √S ( ; z ; y )D z , y≤x≤z
   − = (  )


  
= x2 =42 x √ (z2 +y2 )=42 D y 
 e 2 2 D− = (  ) exp −  √2e − = (  )v
 
z y z , z≤x

ez =4 D− = (  )z S (  ;  ;  )D− = (  )

ez =2 v
 2 2 
4.3.1 Ex e−γ`(%,r) = exp − √

er2 =22 +  2 | Er d( r ; z )|
√
564 7. ORNSTEIN{UHLENBECK PROCESS

7 Us = σe−θs We2θs −1 σ, θ > 0 % = %(v, z) = min{s : `(s, z) = v}

1, x ≤ z ≤ r, r ≤ z ≤ x

√
 e−r2 =22 | Er d x ; r
|

1 +

 √
√ 2  



, r∧z ≤x≤r∨z

× 1 + √ e−r2 =22 | Er d z ; r |
 2
1




, z ≤ r ≤ x, x ≤ r ≤ z
 √
1 + √2 e−r2 =22 | Er d z ; r |



√
 2(ez =2 v + er =2 y)
 2 2 2 2 


4.3.2 Px `(%, r) ∈ dy = exp − √ r z
| Er d(  ;  )|

 √ 2 2 2 √  2√2e(z2 +r2 )=42 √vy 

 2e(z +r )=4 v
√ I √ dy, r≤z≤x
| Er d( r ; z )| y | Er d( r ; z )|


 √ 1
x≤z≤r



√ 2  2√2e(z2 +r2 )=42 √vy 

√
 2er =4 z 2 /4σ 2 v

 2 h
 r x
| Erfid( σ , σ )|e √ I1 √

| Er d( r ; z )|
 √
 Er d2 ( r ; z ) y

× √
 2 2
 2 2e(z2 +r2 )=42 √vy i
+| Erfid( σx , σz )|er /4σ I0 dy, z ∧ r ≤ x ≤ r ∨ z

√
| Er d( r ; z )|





√ √
 2 2e(z +r )=4 √vy 
r2 =22

 √ 2e r z I0

 2 2 2

√ dy, z ≤ r ≤ x, x ≤ r ≤ z
| Er d( r ; z )|

| Er d(  ;  )|

(1) Px `(%, r) = 0
√
 2ez =2 v
  2 2 
exp − √ r z , x ≤ z ≤ r, r ≤ z ≤ x
| Er d(  ;  )|





x ; r
| √ z2 =22 
= | Er d  
exp − √ 2e v

, r∧z ≤x≤r∨z
| Er d z ; r | | Er d( r ; z )|





0, z ≤ r ≤ x, x ≤ r ≤ z

4.3.3 Ex e−α%−γ`(%,r)

√
 ( ( =2) + D− = (− r )D− = ( r ))v 
z≤r = exp − z z
√
 r r z
D− = (−  )(D− = (  ) + √2 e−(z +r )=4 D− = (  )S (  ;  ;  ))
2 2 2

ex =4 D− = ( x )
 2 2

z
√
 r r z , r≤x
ez2 =42 D− = (  ) + √2 e−r =4 D− = (  )S (  ;  ;  )

 2 2



√
ex =4 D− = ( x ) + √2 e−r =4 D− = ( r )S (  ; r ; x )


 2 2 2 2

× √ , z≤x≤r


 ez2 =42 D− = ( z ) + √2 e−r2 =42 D− = ( r )S (  ; r ; z )
ex =4 D− = (− x )

2 2



, 0≤x≤z

ez2 =42 D− = (− z )

 4. STOPPING AT INVERSE LOCAL TIME 565

7 Us = σe−θs We2θs −1 σ, θ > 0 % = %(v, z) = min{s : `(s, z) = v}

√
 ( ( =2) + D− = (− r )D− = ( r ))v 
r≤z = exp − √
D− = ( z )(D− = (− z ) + √2 e−(z2 +r2 )=42 S (  ; z ; r )D− = (− r ))

ex =4 D− = ( x )
 2 2

, z≤x
ez2 =42 D− = ( z )





√
ex =4 D− = (− x ) + 
√ −r =4 D− = (− r )S (  ; x ; r )

e

 2 2 2 2
√ 2

×  , r≤x≤z

 ez2 =42 D− = (− z ) +  2 e
√ −r =4
2 2
D− = (− r )S (  ; z ; r )
ex =4 D− = (− x )

2 2



, x≤r

√
ez2 =42 D− = (− z ) + √2 e−r2 =42 D− = (− r )S (  ; z ; r )



Ex e−α% ; `(%, r) ∈ dy

4.3.4

 √2e(z2 +r2 )=42 D r √

 2e(z +r )=4 D− = ( z )y 
− = (−  )v
2 2 2

z≤r = exp − √ − √
S (  ; r ; z )D− = (− z ) D− = ( r )S (  ; r ; z )

 √ (x2 +r2 )=42 √

 2e D− = (− x ) √v  2 2e(z2 +r2 )=42 √vy 
√ √ I1 √ dy, 0 ≤ x ≤ z
D− = (− z )S (  ; r ; z ) S (  ; r ; z )

y




√ √ √
 2 2e(z2 +r2 )=42 √vy 
 2e(z +r )=4

v

 2 2 2 h
 α r x
 √S 2 ( ; r ; z ) S( θ , σ , σ ) √y I1 √
S (  ; r ; z )


  
× √
 2 2e(z2 +r2 )=42 √vy i
er =4
2 2

+S( αθ , σx , σz ) z2 =42 I0

√ dy, z≤x≤r
S (  ; r ; z )

e




√ (x2 +r2 )=42 √
D− = ( x )  2 2e(z2 +r2 )=42 √vy 

 √ 2e



I dy, r≤x

√
D− = ( r )S (  ; r ; z ) 0 S (  ; r ; z )

 √2e(z2 +r2 )=42 D r √

 2e(z +r )=4 D− = (− z )y 
− = (  )v
2 2 2

r≤z = exp − √ z r z − √
S (  ;  ;  )D− = (  ) r
D− = (−  )S (  ; z ; r )

 √ (x2 +r2 )=42 √

 2e D− = ( x ) √v  2 2e(z2 +r2 )=42 √vy 
√ I √ dy, z ≤ x
D− = ( z )S (  ; z ; r ) S (  ; z ; r )
√

y 1




√ √  2√2e(z2 +r2 )=42 √vy 
 2e(z +r )=4

v

 2 2 2 h
 α x r
√ S( , , ) I √
S 2 (  ; z ; r ) S (  ; z ; r )
√

y 1

 θ σ σ
× √ (z2 +r2 )=42 √ i
er =4 2 2e vy
2 2 
+S( αθ , σz , σx ) z2 =42 I0

 √ z r dy, r≤x≤z
e S ( ; ; )

  



√ (x2 +r2 )=42 √
D− = (− x )  2 2e(z2 +r2 )=42 √vy 

 √ 2e



I dy, 0≤x≤r

√
D− = (− r )S (  ; z ; r ) 0
S (  ; z ; r )
566 7. ORNSTEIN{UHLENBECK PROCESS

7 Us = σe−θs We2θs −1 σ, θ > 0 % = %(v, z) = min{s : `(s, z) = v}

Ex e−α% ; `(%, r) = 0

(1)


 0, r≤x
r ; x)  √2e(z2 +r2 )=42 D r )v 

S ( ; (

   exp − √ − = − 


, z≤x≤r
S (  ; r ; z ) D− = (− z )S (  ; r ; z )

z≤r =
√ (z2 +r2 )=42
 ex2 =42 D− = (− x ) D− = (− r )v 
 exp −  √2e

 
, x≤z

D− = (− z ) z
D− = (−  )S (  ; r ; z )
 z2 =42

e


 0, x≤r
x; r )  √2e(z2 +r2 )=42 D r )v 

S ( ; (

   exp − √ − = 


, r≤x≤z
S (  ; z ; r ) S (  ; z ; r )D− = ( z )

r≤z =
 ex =4 D− = ( x )  √2e(z2 +r2 )=42 D r 
− = (  )v

 2 2

exp − √ , z≤x
D− = ( z ) S (  ; z ; r )D− = ( z )
 z2 =42

e

 Z %

4.4.1 Ex exp −γ 1I[r,∞) Us ds
0

ez =2 D−(+ )= ( r )v

 2 2 
z≤r = exp − r
√
 r z r
er2 =22 D− = (  ) + √2 Er d(  ;  )D−(+ )= (  )

1, x≤z

√

r = 2 r  r x
D− = (  ) +  2 Er d  ;  D−(+ )= (  ) r

e
 2 2


 √
, z≤x≤r

√
r 2 =2 2
D− = ( r ) + √2 Er d r ; z D−(+ )= ( r )

e


×
e(x +r )=4 D− = ( x )

 2 2 2



r
√
 r z
r , r≤x
er2 =22 D− = (  ) + √2 Er d  ;  D−(+ )= (  )



2=er =2  √2e(z2 +r2 )=42 D r 

−( + )= (  )v
 p 2 2

r z exp − √ r z z , x≤r
C(  ;  ;  ) C (  ;  ;  )D− = (  )




C (  ; r ; x )  √2e(z2 +r2 )=42 D r 

−( + )= (  )v


r≤z = r z exp − √ r z z , r≤x≤z

 C(  ;  ;  ) C (  ;  ;  )D− = (  )
 ex2 =42 D− = ( x )  √2e(z2 +r2 )=42 D r 
−( + )= (  )v




z = 4 z exp − √ r z z , z≤x
e D− = (  ) C (  ;  ;  )D− = (  )
 2 2
 4. STOPPING AT INVERSE LOCAL TIME 567

7 Us = σe−θs We2θs −1 σ, θ > 0 % = %(v, z) = min{s : `(s, z) = v}

Z %

4.4.2 Px 1I[r,∞) Us ds = 0
(1) 0

√
 2ez =2 v
  2 2 
exp − √ , x≤z
 Er d( r ; z )





√
z≤r = Er d r ; x  2ez =2 v
2 2

 
r z
exp − √ r z , z≤x≤r
Er d  ;   Er d(  ;  )





0, r≤x

 Z %

4.5.1 Ex exp −γ 1I(−∞,r] Us ds
0

 x2 =42
D− = (− x )  √2e(z2 +r2 )=42 D r 
e −( + )= (−  )v

 ez2 =42 D z exp − √ r z z , x≤z
− = (−  ) C (  ;  ;  )D− = (−  )



 √2e(z2 +r2 )=42 D

C (  ; r ; x ) r 
−( + )= (−  )v


z≤r = r z exp − √ r z z , z≤x≤r
 C(  ;  ;  )
 C (  ;  ;  )D− = (−  )
√
  2e(z2 +r2 )=42 D r 
 2=er2 =22

−( + )= (−  )v

 p
r z exp − √ r z z , r≤x

C(  ;  ;  ) C (  ;  ;  )D− = (−  )


ez =2 D−(+ )= (− r )v

 2 2 
r≤z = exp − r
√
 z r r
er2 =42 D− = (−  ) + √2 Er d(  ;  )D−(+ )= (−  )

e(x +r )=4 D− = (− x )
 2 2 2

r
√
 z r
r , x≤r
er2 =22 D− = (−  ) + √2 Er d  ;  D−(+ )= (−  )





√

er =2 D− = (− r ) + √2 Er d x ; r D−(+ )= (− r )
2 2
×


√ , r≤x≤z
er2 =22 D− = (− r ) + √2 Er d z ; r D−(+ )= (− r )








1, z≤x

Z %

4.5.2 Px 1I(−∞,r] Us ds = 0
(1) 0

0, x≤r

x; r
√ z2 =22 

Er d  2e v

 
exp − √

 
, r≤x≤z
Er d z ; r  Er d( z ; r )

r≤z =
√
 2ez =2 v

  2 2 

 exp − √

, z≤x
z r
 Er d(  ;  )
568 7. ORNSTEIN{UHLENBECK PROCESS

7 Us = σe−θs We2θs −1 σ, θ > 0 % = %(v, z) = min{s : `(s, z) = v}

 Z %



4.6.1 Ex exp − p1I(−∞,r] Us + q1I[r,∞) Us ds
0

 √ (r2+z2 )=42
 2e (pD−(+p)= (− r )D−q= ( r ) + qD−p= (− r )D−(+q)= ( r ))v

z≤r = exp − −2 √
 D−p= (− z )(D−q= ( r )C ( p ; r ; z ) + qD−(+q)= ( r )S ( p ; r ; z ))

ex =4 D−p= (− x )
2 2

, x≤z

ez2 =42 D−p= (− z )




D−q= ( r )C ( p ; r ; x ) + q r p r x

 D−(+q)= (  )S (  ;  ;  ) ,


× z≤x≤r

 D−q= ( r )C ( p ; r ; z ) + qD r p r z
 −(+q)= (  )S (  ;  ;  )
2=e(x +r )=4 D−q= ( x )

 p 2 2 2

r≤x

,
D−q= (  )C ( p ; r ; z ) + q D−(+q)= ( r )S ( p ; r ; z )
r



 √ (r2+z2 )=42
 2e (pD−(+p)= (− r )D−q= ( r ) + qD−p= (− r )D−(+q)= ( r ))v

r≤z = exp − √ −2
 D−q= ( z )(pS ( q ; z ; r )D−(+p)= (− r ) + C ( q ; r ; z )D−p= (− r ))

2=e(x +r )=4 D−p= (− x )

2 2 2
 p
, x≤r
q
C (  ;  ;  )D−p= (−  ) +  S ( q ; z ; r )D−(+p)= (− r )
p

r z r




C ( q ; r ; x )D−p= (− r ) + p S ( q ; x ; r )D−(+p)= (− r )



× , r≤x≤z

 C ( q ; r ; z )D−p= (− r ) + p S ( q ; z ; r )D−(+p)= (− r )
ex =4 D−q= ( x )
2 2



, z≤x

ez2 =42 D−q= ( z )



( 2 − 1) % 2 
 Z
4.9.3 Ex exp −α% − Us ds
42 0
1≤γ
x√

ex =4 D− 1 + √ −1 1 − 1 +
v
2 2
−   2
 
2 2 
1 α
− θγ
2 2γ
x≤z = z √ exp − z √
z √
ez2 =42 D− 1 + D− 1 + D− 1 +




1 α
− θγ −  1 α
− θγ  1 α
− θγ − 
2 2γ 2 2γ 2 2γ

x√

ex =4 D− 1 +  2 −1 12 − 21 +  v
√
2 2



1 α
− θγ
2 2γ
z≤x = z √
exp −
D− 1 + 1 − α z  D− 1 + 1 − α − z 
√
√

ez2 =42 D− 1 + 1 α
− θγ 
2 2γ 2 2γ θγ 2 2γ θγ

%
ds
n  Z  o
4.20.1 Ex exp −γσ 2 θ ; 0 < inf U
0
Us2 0≤s≤%
s

zv 12 1 + 4 + 1
 √ 

= exp − 1 √1 + 4 + M 1 1 √ 1
z2
W 1 1 √ z2

4 4 ; 1+4 22
4 4
; 1+4 22
4 4
 4. STOPPING AT INVERSE LOCAL TIME 569

7 Us = σe−θs We2θs −1 σ, θ > 0 % = %(v, z) = min{s : `(s, z) = v}

 −1=2 x2 =42
x e M 1 ; 1 √1+4 x2

22
4 4
0<x≤z

z2
,

 z −1=2 ez2 =42 M 1 ; 1 √1+4

22

4 4
×
x −1=2 ex2 =4 2 W
1 ; 1 √1+4
x2

22


 −1=2 z2 =42 4 4 √ z2
, z≤x


z e W 1 ; 1 1+4 22
4 4

 Z % 2
p  (q2 − 1) 2
n   o
4.21.1 Ex exp − + U ds ; 0 < inf U
0
Us2 42 s
0≤s≤%
s

qzv 12 1 + 4p + 1
 √

= exp − 1 √1 + 4p + 2q−1 M 1 1 √ qz2 qz2
; 1+4p 22 W 1 ; 1 √1+4p 22




4 4q 4q 4 4q 4

x−1=2 ex =4 M 1 ; 1 √1+4p 2qx2

 2 2 2

 4q 4
2
, 0<x≤z

z −1=2 ez2 =42 M 1 ; 1 √1+4p 2qz2




4q 4
×
x−1=2 ex =4 W 1 ; 1 √1+4p 2qx2
2 2 2


 4q 4
z≤x

2
,
z −1=2 ez2 =42 W 1 ; 1 √1+4p 2qz2



4q 4

n  Z %


 o
4.27.1 Ex exp − p1I(−∞,r] Us + q1I[r,∞) Us ds ; `(%, r) ∈ dy
0

√
 2e(z +r )=4 D−p= (− r )v yC ( p ; r ; z ) qyD−(+q)= ( r )
 2 2 2 
z≤r = exp − √ − −
S ( p ; r ; z )D−p= (− z ) S ( p ; r ; z ) D−q= ( r )

 √ (x2 +r2 )=42 √

 2e D−p= (− x ) √v  2 2e(z2 +r2 )=42 √vy 
I dy, 0 ≤ x ≤ z
D−p= (− z )S ( p ; r ; z ) S ( p ; r ; z )
 √ √ √
y 1




 √2e(z2 +r2 )=42 h √  2√2e(z2 +r2 )=42 √vy 

v

 p r x
S( , , ) I
S 2 ( p ; r ; z ) S ( p ; r ; z )
√ √ √

y 1

 θ σ σ
× √ (z2 +r2 )=42 √ i
er =4 2 2e vy
2 2 
+S( pθ , σx , σz ) z2 =42 I0

 √ p r z dy, z≤x≤r
e S ( ; ; )

  



√ (x2 +r2 )=42 √
D−q= ( x )  2 2e(z2 +r2 )=42 √vy 

 √ 2e



I dy, r≤x
D−q= ( r )S ( p ; r ; z ) S ( p ; r ; z )

0 √
570 7. ORNSTEIN{UHLENBECK PROCESS

7 Us = σe−θs We2θs −1 σ, θ > 0 % = %(v, z) = min{s : `(s, z) = v}

√
 2e(z +r )=4 D−q= ( r )v yC ( q ; r ; z ) pyD−(+p)= (− r )
 2 2 2 
r≤z = exp − √ − −
S ( q ; z ; r )D−q= ( z ) S ( q ; z ; r ) D−p= (− r )

 √ (x2 +r2 )=42 √

 2e D−q= ( x ) √v  2 2e(z2 +r2 )=42 √vy 
I dy, z ≤ x
D−q= ( z )S ( q ; z ; r ) S ( q ; z ; r )
 √ √ √
y 1




√ √  2√2e(z2 +r2 )=42 √vy 
 2e(z +r )=4

v

 2 2 2 h
 q x r
S( , , ) I
S 2 ( q ; z ; r ) S ( q ; z ; r )
√ √ √

y 1

 θ σ σ
× √ (z2 +r2 )=42 √ i
er =4 2 2e vy
2 2 
+S( θq , σz , σx ) z2 =42 I0

 √ q z r dy, r≤x≤z
e S ( ; ; )

  



√ (x2 +r2 )=42 √
D−p= (− x )  2 2e(z2 +r2 )=42 √vy 

 √ 2e



q I dy, 0≤x≤r
S ( q ; z ; r )

√
D−p= (− r )S (  ; z ; r ) 0

n  Z %


 o
(1) Ex exp − p1I(−∞,r] Us + q1I[r,∞) Us ds ; `(%, r) = 0
0

 0, r≤x

p r x √ (z2 +r2 )=42 r
 S(  ;  ;  )  2e D−p= (−  )v

  

exp − √ , z≤x≤r
z≤r = S ( p ; r ; z ) D−p= (− z )S ( p ; r ; z )
√
ex =4 D (− x )   2e(z2 +r2 )=42 D r 
−p= (−  )v
 2 2

 z2 =42 −p= z exp − √



, x≤z

D−p= (−  ) z p r
D−p= (−  )S (  ;  ;  ) z

e


 0, x≤r
q; x; r )  √2e(z2 +r2 )=42 D r )v 

S ( (

    −q= 


exp − √ , r≤x≤z
r≤z = S ( q ; z ; r ) S ( q ; z ; r )D−q= ( z )
√ (z2 +r2 )=42
 ex2 =42 D−q= ( x ) D−q= ( r )v 
 exp −  √2e

 
, z≤x

 z2 =42
D−q= ( z ) q
S (  ; z ; r )D−q= ( z )

e
8 Q(sn) = σe−θs Re(n2θs
)
−1
n ≥ 2, σ, θ > 0 τ ∼ Exp(λ), independent of Q(n)

8. RADIAL ORNSTEIN–UHLENBECK PROCESS

1. Exponential stopping

1.0.2 P(τ > t) = e−αt

1.0.5 Px (Q(τn) ∈ dz)

 2 z n−1 e−z =2

2 2
λ n x2 λ n z2



M 2θ , 2 , 2σ 2 U 2θ , 2 , 2σ 2 dz, 0≤x≤z
2n=2 n2 n




= 
n−1 −z2 =22
  2 z ne
λ n x2 λ n z2



U 2θ , 2 , 2σ 2 M 2θ , 2 , 2σ 2 dz, z≤x

2n=2 2 n

Px Q(tn) ∈ dz


1.0.6
z n=2 ent=2
 2 2
x −z (x2 + z 2 ) ch(t)  
xz

= 2 ( n 2) = 2 exp 2 − 2 I n−2 2 dz
2 x − sh(t) 4 4 sh(t) 2 2 sh(t)

M 2 ; n2 ; 2x22


1.1.2 Px sup Q(sn) ≥ y = 2
, 0≤x≤y
0≤s≤τ M 2 ; n2 ; 2y2

Px sup Q(sn) ≥ y, Q(τn) ∈ dz

1.1.6 x∨z ≤y
0≤s≤τ

 
n−1 e−z2 =22 M  ; n ; x2

= 2 z 2 2 22 M λ n z2

λ n y
2

2θ , 2 , 2σ 2 U 2θ , 2 , 2σ 2 dz
2n=2 n2 n M 2 ; n2 ; 2y2

2

U 2 ; n2 ; 2x22


1.2.2 Px inf Q(sn) ≤ y = 2
, 0≤y≤x
0≤s≤τ U 2 ; n2 ; 2y2

inf Q(sn) ≤ y, Q(τn) ∈ dz

1.2.6 Px y ≤x∧z
0≤s≤τ

 
n−1 e−z2 =22 U  ; n ; x2

= 2 z 2 2 22 U λ n z2

λ n y
2

2θ , 2 , 2σ 2 M 2θ , 2 , 2σ 2 dz
2n=2 n2 n U 2 ; n2 ; 2y2

2

1.3.1 Ex e−γ`(τ,r)

© Springer Basel 2015 571

A.N. Borodin, P. Salminen, Handbook of Brownian Motion - Facts and Formulae,
Probability and Its Applications, DOI 10.1007/978-3-0348-8163-0_15
572 8. RADIAL ORNSTEIN{UHLENBECK PROCESS

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 τ ∼ Exp(λ), independent of Q(n)

rn−1 M 2 ; n2 ; 2x2 U  n r2

2 ; 2 ; 22
2

x≤r =1− n
n r2 =22 + rn−1 
M  n r2
 n r2

2n=2 2  e 2 2 ; 2 ; 22 U 2 ; 2 ; 22

rn−1 U 2 ; n2 ; 2x2 M  n r2

2 ; 2 ; 22
2

r≤x =1− n
n r2 =22 + rn−1 
M  n r2
 n r2

2n=2 2  e 2 2 ; 2 ; 22 U 2 ; 2 ; 22

2n=2 n2 n er =2 y

2 2 


1.3.2 Px `(τ, r) ∈ dy = exp − n−1   n r  n r
r 2 M 2 ; 2 ; 22 U 2 ; 2 ; 22

2
2

2n=2 n2 n er =2 M 2 ; n2 ; 2x2


2 2 2

 rn−1  M 2  ; n ; r2 U  ; n ; r2
dy, 0 ≤ x ≤ r





2 2 2 22 2 2 22
×
2n=2 n
n r2 =22
 e U 2 ; n2 ; 2x2
 2

2

 n r2
2  n r2
dy, r ≤ x

 n−1 


r 2 M 2 ; 2 ; 22 U 2 ; 2 ; 22

M 2 ; n2 ; 2x2
 2

1 − 2
, 0≤x≤r

M 2 ; n2 ; 2r2




(1) Px `(τ, r) = 0 =
U 2 ; n2 ; 2x2
2


 1− 2
, r≤x

U 2 ; n2 ; 2r2


z n−1 e−z =2

2 2
λ n x2 z2
Ex e−γ`(τ,r) , Q(τn) ∈ dz =


1.3.5 F 2θ , 2 , 2σ 2 , 2σ 2 dz
2n=2 n

z n−1 e−z =2 F 2 ; n2 ; 2r2 ; 2z2 F  n x2 r 2

2 ; 2 ; 22 ; 22  dz
2 2 2 2

−  n r2 r2

2n=2 n 2n=2 r1−n n er2 =22 + F 2 ; 2 ; 22 ; 22


2n=2 r1−n n er =2 y

 2 2 

Px `(τ, r) ∈ dy, Q(τn) ∈ dz = exp −

1.3.6  n r r
F 2 ; 2 ; 22 ; 22
2 2

z n−1 e−z =2 F 2 ; n2 ; 2r2 ; 2z2 F 2 ; n2 ; 2x2 ; 2r2

2 2 2 2
2 2

× dydz
rn−1 e−r2 =22 F 2 2 ; n2 ; 2r2 ; 2r2
2 2

z n−1 e−z2 =22 λ n x2 z2

Px `(τ, r) = 0, Q(τn) ∈ dz =


(1) F 2θ , 2 , 2σ 2 , 2σ 2 dz
2n=2 n

z n−1 e−z =2 F 2 ; n2 ; 2r2 ; 2z2 F 2 ; n2 ; 2x2 ; 2r2

2 2 2 2
2 2

− dz
2n=2 n F 2 ; n2 ; 2r2 ; 2r2
2 2

 1. EXPONENTIAL STOPPING 573

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 τ ∼ Exp(λ), independent of Q(n)
 Z τ


p1I[0,r) Q(sn) + q1I[r,∞) Q(sn) ds


1.6.1 Ex exp −
0

0≤x =
+p
x≤r
(p − q)n2−1 ( + p)−2 M 2+p ; n2 ; 2x2 U +2q+2 n+2 r
 ; 2 ; 22
2
2

+ n(+q) +q+2 ; n+2 ; r2
M +p ; n ; r2

M +2p+2 n+2 r +q n r
 ; 2 ; 22 U 2 ; 2 ; 22 + 2(+p) U
2
2

2 2 22 2 2 22

r≤x =
+q
(q − p)( + q)−1 ( + p)−1 U 2+q ; n2 ; 2x2 M +2p+2  n+2 r
 ; 2 ; 22
2
2

+ n ( + q )
M 2 ; 2 ; 22 U 2 ; 2 ; 22 + 2(+p) U 2 ; 2 ; 2r2 M 2+p ; n2 ; 2r2
+p+2 n+2 r +q n r +q+2 n+2
2
2
2
2

n  Z τ

 o
p1I[0,r) Q(sn) + q1I[r,∞) Q(sn) ds , Q(τn) ∈ dz dz


1.6.5 Ex exp −
0

z n−1 e−z =2 S 2+p ; n2 ; 2r2 ; (z+x+4|z2−x|) M 2+p ; n2 ; (z+x−|z−x|)2

2 2 2 2

z≤r = 4 2

n= 2 n  +
2  M 2 ; 2 ; 22 p n r 2

x≤r
M +p ; n ; x2

+ 2 2 22
M +p ; n ; r2

2 2 22

n( + p)−1 z n−1 r−n e(r −z )=2 U 2+q ; n+2 r +q n z

2 ; 22 M 2 ; 2 ; 22
2 2 2 2
2

× n(+q) +q+2 ; n+2 ; r2
M +p ; n ; r2

M +2p+2 n+2 r +q n r
 ; 2 ; 22 U 2 ; 2 ; 22 + 2(+p) U
2
2

2 2 22 2 2 22

n( + p)−1 z n−1 r−n e(r −z )=2 U 2+q ; n2 ; 2x2 M 2+p ; n2 ; 2z2
2 2 2 2
2

z≤r =  n+2 r2
+q n r2
n(+q) +q+2 ; n+2 ; r2
M +p ; n ; r2

r≤x M +2p+2
 ; 2 ; 22 U 2 ; 2 ; 22 + 2(+p) U 2 2 22 2 2 22

n( + p)−1 z n−1 r−n e(r −z )=2 U 2+q ; n2 ; 2z2 M 2+p ; n2 ; 2x2
2 2 2 2
2

r≤z = +p+2 r2
+q n r2
n(+q) +q+2 ; n+2 ; r2
M +p ; n ; r2

x≤r M 2 ; n+22 ; 22 U 2 ; 2 ; 22 + 2(+p) U 2 2 22 2 2 22

z n−1 e−z =2 U 2+q ; n2 ; (z+x+4|z2−x|) S 2+q ; n2 ; (z+x−|z−x|)2 ; r2

2 2 2

r≤z = 42 22

2n=2 n U 2+q ; n2 ; 2r2
2

r≤x
U +q ; n ; x2

+ 2 2 22
U +q ; n ; r2

2 2 22

n( + p)−1 z n−1 r−n e(r −z )=2 M 2+p ; n+2 r +p n z

2 ; 22 U 2 ; 2 ; 22
2 2 2 2
2

× n(+q) +q+2 ; n+2 ; r2
M +p ; n ; r2

M +2p+2 n+2 r2 +q n r2
 ; 2 ; 22 U 2 ; 2 ; 22 + 2(+p) U



2 2 22 2 2 22
574 8. RADIAL ORNSTEIN{UHLENBECK PROCESS

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 τ ∼ Exp(λ), independent of Q(n)

( 2 − 1) t (n) 2 
 Z
1.9.3 Ex exp − (Qs ) ds
42 0
1≤γ

n=2 etn=2 
x2 ( 2 − 1) sh(t )

= exp −
(sh(t ) + ch(t ))n=2 42 (sh(t ) + ch(t ))

( 2 − 1) τ (n)
2  (n)

n  Z o
1.9.5 Ex exp − (Q ds ; Q ∈ dz
42 0
s τ
0<x

 n− n + 
 2  2
4
4 2 (x2−z 2 )/4σ 2 x z
x≤z = n xn=2 z 1−n=2 e M W dz
 22 22
θn−2λ n−2 θn−2λ n−2
4θγ , 4 4θγ , 4
2

 n− n + 
 2  2
4
4 2 (x2−z 2 )/4σ 2 x z
z≤x = n xn=2 z 1−n=2 e W θn−2λ , n−2 2 M θn−2λ , n−2 2 dz
 2 4θγ 4 2 4θγ 4 2

( 2 − 1) t (n) 2  (n)

n  Z o
1.9.7 Ex exp − 2 (Qs ) ds ; Qt ∈ dz
4 0
1≤γ

x1−n=2 z n=2 tn x2 − z 2 (x2 + z 2 ) ch(t )  xz

  
= exp + − I dz
22 sh(t ) 2 42 42 sh(t ) 22 sh(t )
n−2
2

Z t
2 
1.9.8 Px Q(sn) ds ∈ dy, Q(tn) ∈ dz
0

x(z=x)n=2 tn x2 − z 2 − y n−2 x2 + z 2 xz

 
= 4 2 exp + 2 isy/2σ2 θ , t, 0, , dydz
4  2 4 2 42  42 

yM 2+ ; n2 ; 2x2 M +2  n+2 y 2

2 ; 2 ; 22 dy
2


Ex e−γ Ȟ(τ ) ; sup Q(sn) ∈ dy =

1.13.1
2 nM 2+ ; n2 ; 2y2 M 2 ; n2 ; 2y2
2
2

0≤s≤τ
x<y

Ex e−γ Ȟ(τ ) ; sup Q(sn) ∈ dy, Q(τn) ∈ dz


1.13.5 x∨z <y
0≤s≤τ

y1−n z n−1 e(y −z )=2 M 2+ ; n2 ; 2x2 M 2 ; n2 ; 2z2

2 2 2 2
2

= dydz
2 M 2+ ; n2 ; 2y2 M 2 ; n2 ; 2y2
2
2

 1. EXPONENTIAL STOPPING 575

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 τ ∼ Exp(λ), independent of Q(n)

yU 2+ ; n2 ; 2x2 U +2  n+2 y 2

2 ; 2 ; 22 dy
2


Ex e−γ Ĥ(τ ) ; inf Q(sn) ∈ dy =

1.14.1
22 U 2+ ; n2 ; 2y2 U 2 ; n2 ; 2y2
2
2

0≤s≤τ
y<x

Ex e−γ Ĥ(τ ) ; inf Q(sn) ∈ dy, Q(τn) ∈ dz


1.14.5 0<y <x∧z
0≤s≤τ

y1−n z n−1 e(y −z )=2 U 2+ ; n2 ; 2x2 U 2 ; n2 ; 2z2

2 2 2 2
2

= dydz
2 U 2+ ; n2 ; 2y2 U 2 ; n2 ; 2y2
2
2

S 2 ; n2 ; 2b2 ; 2x2 + S 2 ; n2 ; 2x2 ; 2a2

2 2
2 2

(n) (n)


1.15.2 Px a < inf Qs , sup Qs < b =1 −
S 2 ; n2 ; 2b2 ; 2a2
2 2

0≤s≤τ 0≤s≤τ

Px a < inf Q(sn) , sup Q(sn) < b, Q(τn) ∈ dz

1.15.6
0≤s≤τ 0≤s≤τ

 n−1 −z2 =22  n b2 z2
 n x2 a2

z e S 2 ; 2 ; 22 ; 22 S 2 ; 2 ; 22 ; 22
dz, a ≤ x ≤ z

2n=2 n S 2 ; n2 ; 2b2 ; 2a2

 2 2


=
 z n−1 e−z =2 S 2 ; n2 ; 2b2 ; 2x2 S 2 ; n2 ; 2z2 ; 2a2
2 2 2 2
2 2


dz, z ≤ x ≤ b
2n=2 n S 2 ; n2 ; 2b2 ; 2a2

 2 2

 yr1−n 2n=2 n er2 =22 M  ; n ; b2


1.16.2 Px `(τ, r) ∈ dy, sup Qs (n)


< b = exp − 2 2 22
S 2 ; n2 ; 2b2 ; 2r2 M 2 ; n2 ; 2r2
2 2
2

0≤s≤τ
r<b

2  M 2 ; n2 ; 2b2 − M 2 ; n2 ; 2r2 M 2 ; n2 ; 2x2

 n=2 n 2
2

2

dy, 0≤x≤r

rn−1 e−r2 =22 S 2 ; n2 ; 2b2 ; 2r2 M 2 2 ; n2 ; 2r2

 2 2
2


×
2n=2 n M 2 ; n2 ; 2b2 − M 2 ; n2 ; 2r2 S 2 ; n2 ; 2b2 ; 2x2
2
2

2 2


dy, r ≤ x ≤ b

rn−1 e−r2 =22 S 2 2 ; n2 ; 2b2 ; 2r2 M 2 ; n2 ; 2r2

 2 2
2

Px `(τ, r) = 0, sup Q(sn) < b

(1)
0≤s≤τ

M 2 ; n2 ; 2x2
 2

1 − 2
, 0≤x≤r

M 2 ; n2 ; 2r2



=
S  ; n; b 2 ; x 2 + S  ; n; x 2 ; r 2
2 2
2 2

 1 − 2 2 2 2 n b2 2r2 2
2 2 , r ≤ x ≤ b



S 2 ; 2 ; 22 ; 22
576 8. RADIAL ORNSTEIN{UHLENBECK PROCESS

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 τ ∼ Exp(λ), independent of Q(n)

Px `(τ, r) ∈ dy, sup Q(sn) < b, Q(τn) ∈ dz

1.16.6 r∨z <b
0≤s≤τ

 yr1−n 2n=2 n er2 =22 M  ; n ; b2


z n−1 e−z =2
2 2

= exp − 2 2 22
rn−1 e−r =2
2 2
S 2 ; n2 ; 2b2 ; 2r2 M 2 ; n2 ; 2r2
2 2
2

M 2 ; n2 ; 2z2 M 2 ; n2 ; 2x2

 2
2

 dydz, 0 ≤ x ≤ r, z≤r
M 2 2 ; n2 ; 2r2

 2




 n b2 x2
 n z2


S 2 ; 2 ; 22 ; 22 M 2 ; 2 ; 22 dydz,



z≤r≤x

 n b2 r2
 n r2


S 2 ; 2 ; 22 ; 22 M 2 ; 2 ; 22


×
S 2 ; n2 ; 2b2 ; 2z2 M 2 ; n2 ; 2x2
2 2
2


2
dydz, 0≤x≤r≤z

S 2 ; n2 ; 2b2 ; 2r2 M 2 ; n2 ; 2r2


 2 2




S 2 ; n2 ; 2b2 ; 2z2 S 2 ; n2 ; 2b2 ; 2x2
2 2
2 2




r ≤ x, r≤z

dydz,
S 2 2 ; n2 ; 2b2 ; 2r2

2 2


Px `(τ, r) = 0, sup Q(sn) < b, Q(τn) ∈ dz

(1)
0≤s≤τ

z n−1 e−z =2 S 2 ; n2 ; 2r2 ; (z+x+4|z2−x|) )M 2 ; n2 ; (z+x−|z−x|)2

2 2 2 2

z≤r = 4 2
dz
2n=2 n M 2 ; n2 ; 2r2
2

x≤r

z n−1 e−z =2 S 2 ; n2 ; 2b2 ; (z+x+4|z2−x|) )S 2 ; n2 ; (z+x−|z−x|)2 ; r2

2 2 2 2

r≤z = 4 2 22 dz
n= 2 n  n b
2  S 2 ; 2 ; 22 ; 22
2 r 2

r≤x

 yr1−n 2n=2 n er2 =22 U  ; n ; a2


1.17.2 Px `(τ, r) ∈ dy, a < inf Q(sn) = exp −

2 2 22
U 2 ; n2 ; 2r2 S 2 ; n2 ; 2r2 ; 2a2
2
2 2

0≤s≤τ
a<r

2  U 2 ; n2 ; 2a2 − U 2 ; n2 ; 2r2 S 2 ; n2 ; 2x2 ; 2a2

 n=2 n 2
2

2 2

dy, a ≤ x ≤ r

rn−1 e−r2 =22 U 2 ; n2 ; 2r2 S 2 2 ; n2 ; 2r2 ; 2a2

 2
2 2


×
2n=2 n U 2 ; n2 ; 2a2 − U 2 ; n2 ; 2r2 U 2 ; n2 ; 2x2
2
2

2


dy, r≤x

rn−1 e−r2 =22 U 2 2 ; n2 ; 2r2 S 2 ; n2 ; 2r2 ; 2a2

 2
2 2

Px `(τ, r) = 0, a < inf Q(sn)

(1)
0≤s≤τ
 1. EXPONENTIAL STOPPING 577

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 τ ∼ Exp(λ), independent of Q(n)

S 2 ; n2 ; 2r2 ; 2x2 + S 2 ; n2 ; 2x2 ; 2a2

 2 2
2 2

 1− , a≤x≤r

S 2 ; n2 ; 2r2 ; 2a2

 2 2

=
U 2 ; n2 ; 2x2
2


 1− 2
, r≤x

U 2 ; n2 ; 2r2


Px `(τ, r) ∈ dy, a < inf Q(sn) , Q(τn) ∈ dz

1.17.6 a<r∧z
0≤s≤τ

 yr1−n 2n=2 n er2 =22 U  ; n ; a2


z n−1 e−z =2
2 2

= exp − 2 2 22
rn−1 e−r =2
2 2
U 2 ; n2 ; 2r2 S 2 ; n2 ; 2r2 ; 2a2
2
2 2

S 2 ; n2 ; 2z2 ; 2a2 S 2 ; n2 ; 2x2 ; 2a2

 2 2
2 2

 dydz, 0 ≤ x ≤ r, z ≤ r
S 2 2 ; n2 ; 2r2 ; 2a2

 2 2




 n x2
 n z 2 a2


U 2 ; 2 ; 22 S 2 ; 2 ; 22 ; 22 dydz,



z≤r≤x

 n r 2
 n r 2 a2


U 2 ; 2 ; 22 S 2 ; 2 ; 22 ; 22


×
U 2 ; n2 ; 2z2 S 2 ; n2 ; 2x2 ; 2a2
2
2 2


2
dydz, 0≤x≤r≤z

U 2 ; n2 ; 2r2 S 2 ; n2 ; 2r2 ; 2a2


 2
2



U 2 ; n2 ; 2z2 U 2 ; n2 ; 2x2
2
2




r ≤ x, r ≤ z

dydz,
U 2 2 ; n2 ; 2r2

2


Px `(τ, r) = 0, a < inf Q(sn) , Q(τn) ∈ dz

(1)
0≤s≤τ

z n−1 e−z =2 S 2 ; n2 ; 2r2 ; (z+x+4|z2−x|) )S 2 ; n2 ; (z+x−|z−x|)2 ; a2

2 2 2 2

z≤r = 4 2 22 dz
2n=2 n S 2 ; n2 ; 2r2 ; 2a2
2 2

x≤r

z n−1 e−z =2 U 2 ; n2 ; (z+x+4|z2−x|) )S 2 ; n2 ; (z+x−|z−x|)2 ; r2

2 2 2

r≤z = 42 22 dz

2n=2 n U 2 ; n2 ; 2r2
2

r≤x

 Z t 2 2   
x2 e−t

1.20.3 Ex exp − (n)
2
ds = (4σ 2 etθ sh(tθ))n/4 exp − 2
0 Qs 8 sh(t)
0<x

n + (n−2)2 + 2 + 1
 q 
 2 −t 
4 16 4 2 x e
× M 2 sh(t)
4
q
n=2 (n−2)2 2
2 2
q (n−2)
4 + +1
γ
x
n

− 4 , 16 + 4
578 8. RADIAL ORNSTEIN{UHLENBECK PROCESS

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 τ ∼ Exp(λ), independent of Q(n)

t
ds
Z 
1.20.4 Px
∈ dy
0 Q(sn) 2
0<x

 shn=4 (t)etn=4 (n − 2)2 2 y x2 e−t n x2 e−t

   
= n=2 2 −1−n=4 exp − − 2 m4σ2 θy , 2 dy
x (4 ) 4 8 sh(t) 4 8 sh(t)

n  Z τ 2 2   o
(n)
1.20.5 Ex exp − ds ; Q ∈ dz
Q(sn) 2 τ


0

(n−2)2 + 2 + 1 + 2−n
q


16 4 2 4 2 2 2 2
= q e(x −z ) /4σ
(n −2)2 2 n=2 1 n=2
 4 + +1 x z

−

 2   2 
 x z
M
 nθ−2λ , (n−2) + γ 2 22 W dz, 0 ≤ x ≤ z
22
q q
 2 (n−2)2 2
4θ 16 4
nθ−2λ
4θ , 16 + γ4
×  2   2 
x z
W M nθ−2λ q (n−2)2 γ 2 dz, z ≤ x

2  2 22
q
nθ−2λ (n−2)2 γ2
4θ , 16 + 4 4θ , 16 + 4

n  Z t 2 2  
(n)
o
1.20.7 Ex exp − 2 ds ; Q ∈ dz
Q(sn) t


0

x(z=x)n=2 tn x2 − z 2 (x2 + z 2 ) ch(t) q xz

   
= exp + − I dz
22 sh(t) 2 42 42 sh(t) (n−2)2
4 +γ 2 2 2 sh(t )

x(z=x)n=2
t
ds
Z 
1.20.8 Px (n)
2
∈ dy, Q(tn) ∈ dz =
0 Qs 2 sh(t)
0<x

tn (n − 2)2 2 y x2 − z 2 (x2 + z 2 ) ch(t) xz

   
× exp − + − i dydz
2 4 4 2 42 sh(t) 22 sh(t)
σ 2 θy

 Z t 2 2
p   (q2 − 1) (n)
2
 
1.21.3 Ex exp − (n)
2
+ 2 Qs ds
0 Qs 4
0<x

(42 sh(tq))n=4 etn=2 x2 (2(q2 − 1) sh2 (tq) + q2 )

 
= exp −
(sh(tq) + q ch(tq))n=4 82 sh(tq)(sh(tq) + q ch(tq))

n + (n−2)2 + p2 + 1
 q 
4 16 4 2 x2 q 2
 
× M
42 sh(tq)(sh(tq) + q ch(tq))
q
xn=2 (n−2)2 + p2 + 1 2 2
q (n−2)
−n + p4

4,
4 16
 1. EXPONENTIAL STOPPING 579

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 τ ∼ Exp(λ), independent of Q(n)

(q2 − 1)
Z t Z t
ds
n 
 o
(n) 2
(1) Ex exp − 2 Q ds ; (n)
2
∈ dy
4 0
s
0 Qs
0<x

(42 )1+n=4 shn=4 (tq)etn=2 n x2 q 2

 
= m ,
xn=2 (sh(tq) + q ch(tq))n=4 4 82 sh(tq)(sh(tq) + q ch(tq))
4σ 2 θy

(n − 2)2 2 y x2 (2(q2 − 1) sh2 (tq) + q2 )

 
× exp − − dy
4 82 sh(tq)(sh(tq) + q ch(tq))

 Z τ 2 2
p   (q2 − 1) (n) 2
n   o
(n)
1.21.5 Ex exp − + (Q ) ds ; Q ∈ dz
Q(sn) 2 42 s τ


0<x 0

(n−2)2 + p2 + 1 + 2−n
q


16 4 2 4q 2 2 2
= e(x −z )/4σ
(n−2) + p2 + 1 xn=2 z 1−n=2
q
q
2

4

 2  2
qx qz

M
 nθ−2λ , (n−2) + p 22 W dz, 0 ≤ x ≤ z
22

 q
2 2
q
2 2
nθ−2λ (n−2) p
4θq 16 4 4θq , 16 + 4
×  2  2
qx qz
 W nθ−2λ q (n−2)2 p2 M nθ−2λ q (n−2)2 p2 dz, z ≤ x

2 2 22

4θq , 16 + 4 4θq , 16 + 4

 Z t 2 2
p   (q2 − 1) (n) 2
n   o
(n)
1.21.7 Ex exp − + (Q ) ds ; Q ∈ dz
Q(sn) 2 42 s t


0

qx(z=x)n=2 tn x2 − z 2 (x2 + z 2 )q ch(tq) q xzq

   
= 2 exp + 2 − 2 I (n−2)2 2 2 dz
2 sh(tq) 2 4 4 sh(tq) 4 +p 2 sh(tq)

(q2 − 1) qx(z=x)n=2
Z t

 t ds
n  Z o
(n) 2 (n)
(1) Ex exp − 2 Q ds ; (n) 2
∈ dy, Q ∈ dz =
4 0
s
0 Qs

t 2 sh(tq)

tn (n − 2)2 2 y x2 − z 2 (x2 + z 2 )q ch(tq) xzq

   
× exp − + − i dydz
2 4 42 42 sh(tq) 22 sh(tq)
σ 2 θy

p2  2  x(z=x)n=2
n  Z t  Z t o
(n) 2 (n)


(2) Ex exp − (n) 2
ds ; Q ds ∈ dy, Q ∈ dz =
0 Qs


0
s t 44 2

tn x2 − z 2 − y x2 + z 2 xz
  q 
(n−2)2
× exp + 2 isy/2σ2 θ + p2 , t, 0, 2 , 2 dydz
2 4 4 4  4 
580 8. RADIAL ORNSTEIN{UHLENBECK PROCESS

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 τ ∼ Exp(λ), independent of Q(n)

z n=2
t t
ds
Z
2
Z 
1.21.8 Px Q(sn) ds ∈ dg, (n)
2
∈ dy, Q(tn) ∈ dz = 2 (n−2)=2
0 0 Qs 4 x
0<x

tn (n − 2)2 2 y x2 − z 2 − g x2 + z 2 xz
   
× exp − + 2 ei g2 σ 2 θy, t, 2 , 2 dgdydz
2 4 4 2σ θ 4  2 

Px `(τ, r) ∈ dy, a < inf Q(sn) , sup Q(sn) < b

1.25.2
0≤s≤τ 0≤s≤τ

 yr1−n 2n=2 n er2 =22 S  ; n ; b2 ; a2


= θr1−n 2n/2 σ n er
2
/2σ 2
exp − 2 2 22 22
S 2 ; n2 ; 2b2 ; 2r2 S 2 ; n2 ; 2r2 ; 2a2
2 2
2 2

S 2 ; n2 ; 2b2 ; 2a2 − S 2 ; n2 ; 2b2 ; 2r2 − S 2 ; n2 ; 2r2 ; 2a2

2 2
2 2
2 2

×
S 2 ; n2 ; 2b2 ; 2r2 S 2 ; n2 ; 2r2 ; 2a2
2 2
2 2

S 2 ; n2 ; 2x2 ; 2a2

 2 2

 S  ; n ; r2 ; a2
dy, a ≤ x ≤ r



2 2 22 22
×
S  n b2 x2

2 ; 2 ; 22 ; 22 dy, r ≤ x ≤ b


S 2 ; n2 ; 2b2 ; 2r2

 2 2

Px `(τ, r) = 0, a < inf Q(sn) , sup Q(sn) < b

(1)
0≤s≤τ 0≤s≤τ

S 2 ; n2 ; 2r2 ; 2x2 + S 2 ; n2 ; 2x2 ; 2a2

 2 2
2 2

1 − , a≤x≤r

S 2 ; n2 ; 2r2 ; 2a2

 2 2


=
S 2 ; n2 ; 2b2 ; 2x2 + S 2 ; n2 ; 2x2 ; 2r2
2 2
2 2


 1− , r≤x≤b

S 2 ; n2 ; 2b2 ; 2r2

2 2

Ex e−γ`(τ,r) ; a < inf Q(sn) , sup Q(sn) < b, Q(τn) ∈ dz


1.25.5 a<r<b
0≤s≤τ 0≤s≤τ

z n−1 e−z =2 S 2 ; n2 ; 2r2 ; (z+x+4|z2−x|) )S 2 ; n2 ; (z+x−|z−x|)2 ; a2

2 2 2 2

z≤r = 4 2 22 dz
n= 2 n  n r
2  S 2 ; 2 ; 22 ; 22
2 a 2

x≤r
 1. EXPONENTIAL STOPPING 581

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 τ ∼ Exp(λ), independent of Q(n)

S  n x2 a2

+ 2 ; 2 ; 22 ; 22
S  n r 2 a2

2 ; 2 ; 22 ; 22

z n−1 e−z =2 S  ; n ; b 2 ; r 2 S  ; n ; 2z2 ; 2a2

2 2 2 2
2 2

× n−1 −r2 =22  n b2 r2 2
 2 2n r22 a2 2
 2 n= 2
dz
r e S 2 ; 2 ; 22 ; 22 S 2 ; 2 ; 22 ; 22 + 2 2 n S 2 ; n2 ; 2b2 ; 2a2
2

z n−1 e−z =2 S  ; n ; b 2 ; x 2 S  ; n ; 2z2 ; 2a2

2 2 2 2
2 2

z ≤ r = n−1 −r2 =22  n b2 r2 2
 2 2n r22 a2 2
 2 n= 2
dz
r e S 2 ; 2 ; 22 ; 22 S 2 ; 2 ; 22 ; 22 + 2 2 n S 2 ; n2 ; 2b2 ; 2a2
2

r≤x

z n−1 e−z =2 S  ; n ; b 2 ; z 2 S  ; n ; 2x2 ; 2a2

2 2 2 2
2 2

r ≤ z = n−1 −r2 =22  n b2 r2
2 2 2n r22 a2
2 2 n= 2
dz
r e S 2 ; 2 ; 22 ; 22 S 2 ; 2 ; 22 ; 22 + 2 2 n S 2 ; n2 ; 2b2 ; 2a2
2

x≤r

z n−1 e−z =2 S 2 ; n2 ; 2b2 ; (z+x+4|z2−x|) S 2 ; n2 ; (z+x−|z−x|)2 ; r2

2 2 2 2

r≤z = 4 2 22 dz
2n=2 n S 2 ; n2 ; 2b2 ; 2r2
2 2

r≤x
S n b2 x2

+ 2 ; 2 ; 22 ; 22
S n b2 r2

2 ; 2 ; 22 ; 22
z n−1 e−z =2 S 2 ; n2 ; 2b2 ; 2z2 S 2 ; n2 ; 2r2 ; 2a2
2 2 2 2
2 2

× n−1 −r2 =22  n b2 r2
 n r2 a2
2
dz

r e S 2 ; 2 ; 22 ; 22 S 2 ; 2 ; 22 ; 22 + 2n=2 n S 2 ; n2 ; 2b2 ; 2a2
2

Px `(τ, r) ∈ dy, a < inf Q(sn) , sup Q(sn) < b, Q(τn) ∈ dz

1.25.6
0≤s≤τ 0≤s≤τ
a<r
r<b

 yr1−n 2n=2 n er2 =22 S  ; n ; b2 ; a2


z n−1 e−z =2
2 2

= n−1 −r2 =22 exp − 2 2 22 22

r e S 2 ; n2 ; 2b2 ; 2r2 S 2 ; n2 ; 2r2 ; 2a2
2 2
2 2

S 2 ; n2 ; 2z2 ; 2a2 S 2 ; n2 ; 2x2 ; 2a2

 2 2
2 2

 dydz, 0 ≤ x ≤ r, z ≤ r
S 2 2 ; n2 ; 2r2 ; 2a2

 2 2




 n b2 x2
 n z2 a2


S 2 ; 2 ; 22 ; 22 S 2 ; 2 ; 22 ; 22 dydz,



z≤r≤x

 n b2 r2
 n r2 a2


S 2 ; 2 ; 22 ; 22 S 2 ; 2 ; 22 ; 22


×
S 2 ; n2 ; 2b2 ; 2z2 S 2 ; n2 ; 2x2 ; 2a2
2 2
2 2


2
dydz, 0≤x≤r≤z

S 2 ; n2 ; 2b2 ; 2r2 S 2 ; n2 ; 2r2 ; 2a2


 2 2
2



S 2 ; n2 ; 2b2 ; 2z2 S 2 ; n2 ; 2b2 ; 2x2
2 2
2 2




r ≤ x, r ≤ z

dydz,
S 2 2 ; n2 ; 2b2 ; 2r2

2 2


582 8. RADIAL ORNSTEIN{UHLENBECK PROCESS

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 τ ∼ Exp(λ), independent of Q(n)

Px `(τ, r) = 0, a < inf Q(sn) , sup Q(sn) < b, Q(τn) ∈ dz

(1)
0≤s≤τ 0≤s≤τ

z n−1 e−z =2 S 2 ; n2 ; 2r2 ; (z+x+4|z2−x|) )S 2 ; n2 ; (z+x−|z−x|)2 ; a2

2 2 2 2

z≤r = 42 22 dz

2n=2 n S 2 ; n2 ; 2r2 ; 2a2
2 2

x≤r

z n−1 e−z =2 S 2 ; n2 ; 2b2 ; (z+x+4|z2−x|) )S 2 ; n2 ; (z+x−|z−x|)2 ; r2

2 2 2 2

r≤z = 4 2 22 dz
2n=2 n S 2 ; n2 ; 2b2 ; 2r2
2 2

r≤x

n  Z τ


p1I[0,r) Q(sn) + q1I[r,∞) Q(sn) ds ;


1.26.1 Ex exp −
a<r 0
o
r<b a < inf Q(sn) , sup Q(sn) < b
0≤s≤τ 0≤s≤τ

S +p ; n ; r2 ; x2
 S 2+p ; n2 ; 2x2 ; 2a2

2 2

 2 2 22 22

a≤x = 1− +p ; n ; r2 ; a2 −
+p S S 2+p ; n2 ; 2r2 ; 2a2

2 2

x≤r 2 2 22 22

( + p)−1 2n=2 (=r)n S +q ; n ; b2 ; r2

× +q n b2 r2
+p n r2 a2
2 2 22 22
S 2 ; 2 ; 22 ; 22 C 2 ; 2 ; 22 ; 22 + C +q ; n ; r2 ; b2
S +p ; n ; r2 ; a2

2 2 22 22 2 2 22 22
√

+ q 2 2
n
( + q)−1 ( + p)−1 (p − q)C 2 ; n2 ; 2r2 ; 2b2 −  r 2 er =2 S 2+p ; n2 ; 2x2 ; 2a2
2 2

 2 2


+
S 2+q ; n2 ; 2b2 ; 2r2 C 2+p ; n2 ; 2r2 ; 2a2 + C 2+q ; n2 ; 2r2 ; 2b2 S 2+p ; n2 ; 2r2 ; 2a2
2 2
2 2
2 2
2 2

S +q ; n ; x2 ; r2
 S 2+q ; n2 ; 2b2 ; 2x2

2 2

 2 2 22 22

r≤x = 1− +q ; n ; b2 ; r2 −
+q S S 2+q ; n2 ; 2b2 ; 2r2

2 2

x≤b 2 2 22 22

( + q)−1 2n=2 (=r)n S +p ; n ; r2 ; a2

× +q n b2 r2
+p n r2 a2
2 2 22 22
S 2 ; 2 ; 22 ; 22 C 2 ; 2 ; 22 ; 22 + C +q ; n ; r2 ; b2
S +p ; n ; r2 ; a2

2 2 22 22 2 2 22 22
√

2 n
( + p)−1 ( + q)−1 (q − p)C 2 ; n2 ; 2r2 ; 2a2 ) −  r 2 er =2 S 2+q ; n2 ; 2b2 ; 2x2 )
2 2 2
 2 2


+
S 2+q ; n2 ; 2b2 ; 2r2 C 2+p ; n2 ; 2r2 ; 2a2 + C 2+q ; n2 ; 2r2 ; 2b2 S 2+p ; n2 ; 2r2 ; 2a2
2 2
2 2
2 2
2 2

n  Z τ


p1I[0,r) Q(sn) + q1I[r,∞) Q(sn) ds ;


1.26.5 Ex exp −
r<b 0
o
a<r a < inf Q(sn) , sup Q(sn) < b, Q(τn) ∈ dz
0≤s≤τ 0≤s≤τ
 1. EXPONENTIAL STOPPING 583

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 τ ∼ Exp(λ), independent of Q(n)

z n−1 e−z =2 S 2+p ; n2 ; 2r2 ; (z+x+4|z2−x|) S 2+p ; n2 ; (z+x−|z−x|)2 ; a2

2 2 2 2

z≤r = 42 22 dz

2n=2 n S 2+p ; n2 ; 2r2 ; 2a2
2 2

x≤r

S +p ; n ; x2 ; a2

+ 2 2 22 22
S +p ; n ; r2 ; a2

2 2 22 22

(r)−1 (z=r)n−1 e(r −z )=2 S +q ; n ; b 2 ; r 2 S +p ; n ; z 2 ; a 2

2 2 2 2 2
2 2

× +q n b2 r2
+p n r2 a2
2 2 2+q n2 r2 b22
2 2+p n2 r2 a2
dz
S 2 ; 2 ; 22 ; 22 C 2 ; 2 ; 22 ; 22 + C 2 ; 2 ; 22 ; 22 S 2 ; 2 ; 22 ; 22

(r)−1 (z=r)n−1 e(r −z )=2 S 2+p ; n2 ; 2z2 ; 2a2 S 2+q ; n2 ; 2b2 ; 2x2
2 2 2 2 2
2 2

z≤r = 2
dz
S 2 ; n2 ; 2b2 ; 2r2 C 2+p ; n2 ; 2r2 ; 2a2 + C 2+q ; n2 ; 2r2 ; 2b2 S 2+p ; n2 ; 2r2 ; 2a2
+q 2 2
2 2
2 2
2

r≤x

(r)−1 (z=r)n−1 e(r −z )=2 S +q ; n ; b 2 ; z 2 S +p ; n ; x 2 ; a 2

2 2 2 2 2
2 2

r≤z = +q n b2 r2
+p n r2 a2
2 2 2+q n2 r2 b22
2 2+p n2 r2 a2
dz
S 2 ; 2 ; 22 ; 22 C 2 ; 2 ; 22 ; 22 + C 2 ; 2 ; 22 ; 22 S 2 ; 2 ; 22 ; 22
x≤r

z n−1 e−z =2 S 2+q ; n2 ; 2b2 ; (z+x+4|z2−x|) S 2+q ; n2 ; (z+x−|z−x|)2 ; r2

2 2 2 2

r≤z = 42 22 dz

2n=2 n S 2+q ; n2 ; 2b2 ; 2r2
2 2

r≤x

S +q ; n ; b2 ; x2

+ 2 2 22 22
S +q ; n ; b2 ; r2

2 2 22 22

(r)−1 (z=r)n−1 e(r −z )=2 S 2+p ; n2 ; 2r2 ; 2a2 S 2+q ; n2 ; 2b2 ; 2z2
2 2 2 2 2
2 2

× 2
dz
S 2 ; n2 ; 2b2 ; 2r2 C 2+p ; n2 ; 2r2 ; 2a2 + C 2+q ; n2 ; 2r2 ; 2b2 S 2+p ; n2 ; 2r2 ; 2a2
+q 2 2
2 2
2 2
2

n  Z τ

 o
p1I[0,r) Q(sn) + q1I[r,∞) Q(sn) ds ; `(τ, r) ∈ dy


1.27.1 Ex exp −
0

 yr( + p)M +p+2 ; n+2 ; r2 ) yr( + q)U +q+2 ; n+2 ; r2 ) 

0≤x = exp − 2 2 22 − 2 2 22
+p n r
nM 2 ; 2 ; 22 )
2 +q n r
2U 2 ; 2 ; 22 )
2

rM 2+p ; n2 ; 2x2  M +2p+2  n+2 r2

 ; 2 ; 22 + U
+q+2 ; n+2 ; r2

 2

 2 2 22 dy, x ≤ r
 M  ; n ; r2 +p ; n ; r2 2 +q ; n ; r2


nM U



2 2 2 2 2 2 2 2 2 2 22
×  + q +p+2  + +2
n x
rU 2 ; 2 ; 22 M 2 ; 2 ; 22 n +2 r U 2  ; n+2
 q ; 2r2 
2
 2
2

2

+ dy, r ≤ x

U 2+q ; n2 ; 2r2 nM 2+p ; n2 ; 2r2 2U 2+q ; n2 ; 2r22

 2
2


584 8. RADIAL ORNSTEIN{UHLENBECK PROCESS

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 τ ∼ Exp(λ), independent of Q(n)
n  Z τ

 o
p1I[0,r) Q(sn) + q1I[r,∞) Q(sn) ds ; `(τ, r) = 0


(1) Ex exp −
0

+p ; n ; x2 ) 
M


2 2 22

1−
+p ; n ; r2 ) , 0≤x≤r

+p

M

2 2 22

= +
U 2q ; n2 ; 2x2 ) 
 2

 
1− , r≤x

+q U 2+q ; n2 ; 2r2 )

 2

n  Z τ

 o
p1I[0,r) Q(sn) + q1I[r,∞) Q(sn) ds ; `(τ, r) ∈ dy, Q(τn) ∈ dz


1.27.5 Ex exp −
0

 yr( + p)M +p+2 ; n+2 ; r2
yr( + q)U +q+2 ; n+2 ; r2


= exp − 2 2 22 − 2 2 22
nM 2+p ; n2 ; 2r2 2U 2+q ; n2 ; 2r22
2

z n−1 e−z =2 M 2+p ; n2 ; 2x2 )M 2+p ; n2 ; 2z2 )

 2 2 2 2

dydz z∨x≤r
rn−1 e−r2 =22 M 2 2+p ; n2 ; 2r2 )



 2


z n−1 e−z =2 M +p ; n ; z2 )U +q ; n ; x2 )

 2 2
2 2 22 2 2 22


+p ; n ; r2 )U +q ; n ; r2 ) dydz z≤r≤x

rn−1 e−r2 =22 M


2 2 22 2 2 22

×
z n−1 e−z =2 M 2+p ; n2 ; 2x2 )U 2+q ; n2 ; 2z2 )
2 2 2 2

dydz x≤r≤z

rn−1 e−r2 =22 M 2+p ; n2 ; 2r2 )U 2+q ; n2 ; 2r2 )


 2 2


z n−1 e−z =2 U 2+q ; n2 ; 2x2 )U 2+q ; n2 ; 2z2 )

2 2

 2 2

r ≤ z, r ≤ x

rn−1 e−r2 =22 U 2 2+q ; n2 ; 2r2 )

2


n  Z τ

 o
p1I[0,r) Q(sn) + q1I[r,∞) Q(sn) ds ; `(τ, r) = 0, Q(τn) ∈ dz


(1) Ex exp −
0

z n−1 e−z =2 S 2+p ; n2 ; 2r2 ; (z+x+4|z2−x|) M 2+p ; n2 ; (z+x−|z−x|)2

2 2 2 2

x≤r = 42
2n=2 n M 2+p ; n2 ; 2r2
2

z≤r
z n−1 e−z =2 U 2+q ; n2 ; (z+x+4|z2−x|) S 2+q ; n2 ; (z+x−|z−x|)2 ; r2

2 2 2

r≤x = 4 2 22
2n=2 n U 2+q ; n2 ; 2r2
2

r≤z

Ex e−γ Ĥ(τ )−ηȞ(τ ) ; inf Q(sn) ∈ da, sup Q(sn) ∈ db dadb

 
1.28.1
0≤s≤τ 0≤s≤τ
a<x
x<b
2n=2 n−4 bea =2 S +2+ ; n2 ; 2b2 ; 2x2 ) C 2 ; n2 ; 2b2 ; 2a2 − 2n=2 n b−n eb =2
2 2 2 2 2 2
2 2

=
an−1 S +2+ ; n2 ; 2b2 ; 2a2 )S 2+ ; n2 ; 2b2 ; 2a2 )S 2 ; n2 ; 2b2 ; 2a2
2 2 2 2 2 2

 1. EXPONENTIAL STOPPING 585

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 τ ∼ Exp(λ), independent of Q(n)

2n=2 n−4 aeb =2 S +2+ ; n2 ; 2x2 ; 2a2 ) C 2 ; n2 ; 2a2 ; 2b2 − 2n=2 n a−n ea =2
2 2 2 2 2 2
2 2

+
bn−1 S +2+ ; n2 ; 2b2 ; 2a2 )S 2+ ; n2 ; 2b2 ; 2a2 )S 2 ; n2 ; 2b2 ; 2a2
2 2 2 2 2 2

Ex e−γ Ĥ(τ )−ηȞ(τ ) ; inf Q(sn) ∈ da, sup Q(sn) ∈ db, Q(τn) ∈ dz dadbdz
 
1.28.5
0≤s≤τ 0≤s≤τ
a<x
x<b
2n=2 n−4 (ab)1−n z n−1 e(b +a −z )=2 S +2+ ; n2 ; 2b2 ; 2x2 )S 2 ; n2 ; 2z2 ; 2a2
2 2 2 2 2 2 2 2

=
S + 2+eta ; n2 ; 2b2 ; 2a2 )S 2+ ; n2 ; 2b2 ; 2a2 )S 2 ; n2 ; 2b2 ; 2a2
2 2 2 2 2 2

2n=2 n−4 (ab)1−n z n−1 e(b +a −z )=2 S +2+ ; n2 ; 2x2 ; 2a2 )S 2 ; n2 ; 2b2 ; 2z2
2 2 2 2 2 2 2 2

+
S +2+ ; n2 ; 2b2 ; 2a2 )S 2+ ; n2 ; 2b2 ; 2a2 )S 2 ; n2 ; 2b2 ; 2a2
2 2 2 2 2 2

n  Z τ


p1I[0,r) Q(sn) + q1I[r,∞) Q(sn) ds ; a < inf Q(sn) ,


1.29.1 Ex exp −
0 0≤s≤τ
o
0≤a sup Q(sn) < b, `(τ, r) ∈ dy
0≤s≤τ

 C +q ; n ; r2 ; b2
− √2
n er2 =22 C +p ; n ; r2 ; a2
− √2
n er2 =22 

= 2 2 22 22 r + 2 2 22 22 r
( + q)S 2+q ; n2 ; 2b22 ; 2r22 ( + p)S 2+p ; n2 ; 2r22 ; 2a22

 yrC +q ; n ; r2 ; b2
yrC +p ; n ; r2 ; a2


×λθr exp − 2 2 22 22 − 2 2 22 22
S 2+q ; n2 ; 2b2 ; 2r2 S 2+p ; n2 ; 2r2 ; 2a2 )
2 2
2 2

S 2+p ; n2 ; 2x2 ; 2a2

 2 2

 S +p ; n ; r2 ; a2
dy, a ≤ x ≤ r



2 2 22 22
× +q ; n ; b2 ; x2


 S 2 2 22 22 dy, r ≤ x ≤ b
S 2+q ; n2 ; 2b2 ; 2r2 )

 2 2

n  Z τ


p1I[0,r) Q(sn) + q1I[r,∞) Q(sn) ds ; a < inf Q(sn) ,


(1) Ex exp −
0 0≤s≤τ
o
sup Q(sn) < b, `(τ, r) = 0
0≤s≤τ

S 2+p ; n2 ; 2r2 ; 2x2 + S 2+p ; n2 ; 2x2 ; 2a2 

2 2
2 2


 
1− , a≤x≤r

+p S 2+p ; n2 ; 2r2 ; 2a2

 2 2


=
S 2+q ; n2 ; 2b2 ; 2x2 + S 2+q ; n2 ; 2x2 ; 2r2 
2 2
2 2

  
1− , r≤x≤b

+q S 2+q ; n2 ; 2b2 ; 2r2

 2 2

586 8. RADIAL ORNSTEIN{UHLENBECK PROCESS

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 τ ∼ Exp(λ), independent of Q(n)

n  Z τ


p1I[0,r) Q(sn) + q1I[r,∞) Q(sn) ds ; a < inf Q(sn) ,


1.29.5 Ex exp −
0 0≤s≤τ
o
0≤a sup Q(sn) < b, `(τ, r) ∈ dy, Q(τn) ∈ dz
0≤s≤τ

 yrC +q ; n ; r2 ; b2
yrC +p ; n ; r2 ; a2


z n−1 e−z =2
2 2

= n−1 −r2 =22 exp − 2 2 22 22 − 2 2 22 22

r e S 2+q ; n2 ; 2b2 ; 2r2 S 2+p ; n2 ; 2r2 ; 2a2
2 2
2 2

S 2+p ; n2 ; 2x2 ; 2a2 S 2+p ; n2 ; 2z2 ; 2a2

 2 2
2 2

dydz, a ≤ x ≤ r, a ≤ z ≤ r
S 2 2+p ; n2 ; 2r2 ; 2a2



 2 2



+q ; n ; b2 ; x2
S +p ; n ; z2 ; a2


S

2 2 22 22 2 2 22 22


+q ; n ; b2 ; r2
S +p ; n ; r2 ; a2
dydz, a≤z≤r≤x≤b


S

2 2 22 22 2 2 22 22

×
S 2+q ; n2 ; 2b2 ; 2z2 S 2+p ; n2 ; 2x2 ; 2a2
2 2
2 2


2
dydz, a≤x≤r≤z≤b

S 2+q ; n2 ; 2b2 ; 2r2 S 2+p ; n2 ; 2r2 ; 2a2


 2 2
2


S 2+q ; n2 ; 2b2 ; 2x2 S 2+q ; n2 ; 2b2 ; 2z2

2 2
2 2




r ≤ x ≤ b, r ≤ z ≤ b

dydz,
S 2 2+q ; n2 ; 2b2 ; 2r2

2 2



n  Z τ


p1I[0,r) Q(sn) + q1I[r,∞) Q(sn) ds ; a < inf Q(sn) ,


(1) Ex exp −
0 0≤s≤τ
o
sup Q(sn) < b, `(τ, r) = 0, Q(τn) ∈ dz
0≤s≤τ

z n−1 e−z =2 S 2+p ; n2 ; 2r2 ; (z+x+4|z2−x|) S 2+p ; n2 ; (z+x−|z−x|)2 ; a2

2 2 2 2

x≤r = 4 2 22 dz
2n=2 n S 2+p ; n2 ; 2r2 ; 2a2
2 2

z≤r

z n−1 e−z =2 S 2+q ; n2 ; 2b2 ; (z+x+4|z2−x|) S 2+q ; n2 ; (z+x−|z−x|)2 ; r2

2 2 2 2

r≤x = 42 22 dz

2n=2 n S 2+q ; n2 ; 2b2 ; 2r2
2 2

r≤z
 2. STOPPING AT FIRST HITTING TIME 587

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 Hz = min{s : Q(sn) = z}

2. Stopping at first hitting time

M 2 ; n2 ; 2x2
 2

 M ; n ; z2
, x ≤ z



2 2 22
2.0.1 Ex e−αHz =
U 2 ; n2 ; 2x2
2


2
, z≤x

U 2 ; n2 ; 2z2



Ed n; y ; x

2.1.2 Px sup Qs (n)


<y = 2   z≤x≤y
Ed n; y ; z
,
0≤s≤Hz 2  

M 2 ; n2 ; 2x2
 2

 M ; n ; z22
, x≤z



2 2 2
Ex e−αHz ; sup Q(sn)

2.1.4 <y =
S 2 ; n2 ; 2y2 ; 2x2
2 2

0≤s≤Hz
z≤y

2
, z≤x≤y

S 2 ; n2 ; 2y2 ; 2z2

 2

Ed n2 ; x ; y


2.2.2 Px inf Q(sn) > y =
, y≤x≤z
0≤s≤Hz Ed n2 ; z ; y

S 2 ; n2 ; 2x2 ; 2y2
 2 2

 S ; n ; z2 ; y2
, y ≤ x ≤ z



2.2.4 Ex e−αHz ;

inf Q(sn) >y = 2 2 22 22
U n x2

2 ; 2 ; 22 ,
0≤s≤Hz
y≤z

z≤x

U 2 ; n2 ; 2z2

2



2.3.1 Ex e−γ`(Hz ,r)

1, x ≤ z ≤ r, r≤z≤x

n ; x ; r
|
 2n=2 n er =2 + rn−1 | Ed

 2 2
2  


n ; z ; r
| , r∧z ≤x≤r∨z
= 2n=2 n er =2 + rn−1 | Ed
2 2
2  
2n=2 n er =2

 2 2

n ; z ; r
| , 0 ≤ x ≤ r ≤ z, z≤r≤x

2n=2 n er =2 + rn−1 | Ed

2 2
2  

2n=2 n er =2 y
 2 2 


2.3.2 Px `(Hz , r) ∈ dy = exp − n−1
r | Ed n2 ; z ; r |


588 8. RADIAL ORNSTEIN{UHLENBECK PROCESS

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 Hz = min{s : Q(sn) = z}

0, 0 ≤ x ≤ z ≤ r, r ≤ z ≤ x

2 2 n x z

n=
 2  e n r 2
=  2
| Ed 2 ;  ;  |




dy, r ∧ z ≤ x ≤ r ∨ z
× rn−1 Ed2 n2 ; z ; r

n=2 n r2 =22
 2  e




dy, 0 ≤ x ≤ r ≤ z, z ≤ r ≤ x

r | Ed n2 ; z ; r |
n −1


1, 0 ≤ x ≤ z ≤ r, r≤z≤x
| Ed n r x


2; ;


| ,



(2) Px ` H z , r = 0 = n r z z∧r ≤x≤r∨z

 | Ed 2 ;  ;  |
0 ≤ x ≤ r ≤ z, z≤r≤x

0,

2.3.3 Ex e−αHz −γ` Hz ,r

0≤x
U 2 ; n2 ; 2x2
 2


2
, z≤x
U 2 ; n2 ; 2z2





 2n=2 n er2 =22 M n x2
n−1 S n x2 r 2
n r2


2 ; 2 ; 22 + r 2 ; 2 ; 22 ; 22 M 2 ; 2 ; 22 , r ≤ x

r≤z = n z2 n z2 r2 n r2

2n=2 n er2 =22 M 2 ; 2 ; 22 + r
n−1 S
2 ; 2 ; 22 ; 22 M 2 ; 2 ; 22





 x≤z
n x
2n=2 n er =2 M 2 ; 2 ; 22

 2 2 2



n z2
n z2 r2
n r2
, x ≤ r

2n=2 n er2 =22 M 2 ; 2 ; 22 + r
n−1 S
2 ; 2 ; 22 ; 22 M 2 ; 2 ; 22


M 2 ; n2 ; 2x2
 2


2
, x≤z
M 2 ; n2 ; 2z2





 2n=2 n er2 =22 U ; n ; x2
+ rn−1 S ; n ; r2 ; x2
U ; n ; r2



z≤r = 2 2 22 2 2 22 22 2 2 22 , z≤x
 2n=2  n er2 =22 U ; n ; z2
+ rn−1 S ; n ; r 2 ; z 2
U n r2



 2 2 2 2 2 2 2 2 2 2 2 ; 2 ; 22 x≤r
2n=2 n er =2 U 2 ; n2 ; 2x2

 2 2 2



2
, r≤x
2n=2 n er2 =22 U 2 ; n2 ; 2z2 + rn−1 S 2 ; n2 ; 2r2 ; 2z2 U 2 ; n2 ; 2r2

 2
2 2

Ex e−αHz ; ` Hz , r ∈ dy


2.3.4

2n=2 n er =2 M 2 ; n2 ; 2z2 y 

 2 2 2

r≤z = exp − n−1

r S 2 ; n2 ; 2z2 ; 2r2 M 2 ; n2 ; 2r2
2 2
2

2n=2 n er =2 M 2 ; n2 ; 2x2

 2 2 2

n r2
dy, 0 ≤ x ≤ r

n−1 n z2 r2
 r S 2 ; 2 ; 22 ; 22 M 2 ; 2 ; 22





× 2n=2 n er2 =22 S 2 ; n2 ; z22 ; x22

2
2 dy, r≤x≤z
rn−1 S 2 2 ; n2 ; 2z2 ; 2r2

 2 2




0, z≤x
 2. STOPPING AT FIRST HITTING TIME 589

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 Hz = min{s : Q(sn) = z}

2n=2 n er =2 U 2 ; n2 ; 2z2 y 

 2 2 2

z≤r = exp − n−1

r S 2 ; n2 ; 2r2 ; 2z2 U 2 ; n2 ; 2r2
2 2
2

2n=2 n er =2 U 2 ; n2 ; 2x2

 2 2 2

n r2
dy, r ≤ x

n−1 n r2 z2

 r S 2 ; 2 ; 22 ; 22 U 2 ; 2 ; 22




× 2n=2 n er2 =22 S 2 ; n2 ; x22 ; z22

2 2 dy, z≤x≤r
n−1 S 2 n ; r2 ; z2


r ;

2 2 2 22


 2

0, 0≤x≤z

0, 0≤x≤r≤z z≤r≤x

n x r

S ; ; ;
 2 2

| 2 2 22 22 | , r ∧ z ≤ x ≤ r ∨ z



|S 2 ; n z2 r2


2 ; 22 ; 22 |





Ex e−αHz ; ` Hz , r = 0 = U 2 ; n2 ; 2x2
2



(1)
2
, r≤z≤x
U 2 ; n2 ; 2z2





 M 2 ; n2 ; 2x2

 2



2
, 0≤x≤z≤r
M 2 ; n2 ; 2z2


 Z Hz

2.4.1 Ex exp −γ 1I[r,∞) Q(sn) ds
0≤x 0

 1, x≤z
1+n=2  n er2 =22 U n r2
n r x
+2 ; n+2 ; r2


n
 2 2 ; 2 ; 22 + r Ed 2 ;  ;  U

2 2 22


+2 ; n+2 ; r2
, z≤x

21+n=2 n er2 =22 U 2 ; n2 ; 2r2 + rn Ed n2 ; r ; z U
2


z≤r = 2 2 22 x≤r
21+n=2 n er =2 U 2 ; n2 ; 2x2

 2 2 2



+2 ; n+2 ; r2
, r ≤ x

U 2 ; n2 ; 2r2 + rn Ed n2 ; r ; z U
 1+n=2 n r2 =22
2

 e
2


2 2 22

 n=2 n −n r2 =22
2  r e
n r2 z2
, 0 < x ≤ r

C 2 ; 2 ; 22 ; 22





 C ; n ; r2 ; x2



r≤z = 2 2 22 22 , r ≤ x ≤ z
C n r2 z2


 2 ; 2 ; 22 ; 22
U 2 ; n2 ; 2x2

 2



2
, z≤x

U 2 ; n2 ; 2z2




1, 0≤x≤z
 Ed n ; r ; x


Z Hz 
2  
,

Q(sn) ds = 0 =


2.4.2 Px 1I[r,∞) z≤x≤r
0 Ed n2 ; r ; z
(1)


r≤x

z≤r 0,
590 8. RADIAL ORNSTEIN{UHLENBECK PROCESS

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 Hz = min{s : Q(sn) = z}

 Z Hz

2.5.1 Ex exp −γ 1I[0,r] Q(sn) ds
0≤x 0

n2n=2 n er =2 M 2 ; n2 ; 2x2

 2 2 2

n r2
n z r
+2 ; n+2 ; r2
, x ≤ r

n=2 n r2 =22 M n
 n2  e 2 ; 2 ; 22 + r Ed 2 ;  ;  M

2 2 22



r≤z = n2n=2 n er =2 M 2 ; n2 ; 2r2 + rn Ed n2 ; x ; r M
2 2 2

+2 ; n+2 ; r2

2 2 22
n2n=2 n er2 =22 M 2 ; n2 ; 2r2 + rn Ed n2 ; z ; r M +2 ; n+2 ; r2
, r ≤ x

 2


2 2 22 x≤z




1, z≤x

M 2 ; n2 ; 2x2
 2

, 0≤x≤z

M 2 ; n2 ; 2z2

 2




C 2 ; n2 ; 2r2 ; 2x2
2 2



z≤r = 2
, z≤x≤r
C 2 ; n2 ; 2r2 ; 2z2
2



 2n=2 n r−n er2 =22


2
, r≤x

C 2 ; n2 ; 2r2 ; 2z2

 2


 0, x≤r
Z Hz   Ed n ; x ; r

(n) 2  
, r ≤ x ≤ z


2.5.2 Px 1I[0,r] Qs ds = 0 =
0 Ed n2 ; z ; r
(1)


z≤x

r≤z 1,

 Z Hz


p1I[0,r) Q(sn) + q1I[r,∞) Q(sn) ds


2.6.1 Ex exp −
0≤x 0

2n=2 n r−n er =2 M 2p ; n2 ; 2x2

 2 2 2

p S q ; n ; z2 ; r2 M p+2 ; n+2 ; r2 + C q ; n ; r2 ; z2
M p n r2
, x ≤ r

2 ; 2 ; 22




n 2 2 22 22 2 2 22 2 2 22 22



 p S q ; n ; x2 ; r2
M p+2 ; n+2 ; r2
+ C q ; n ; r2 ; x2
M p n r2


2 ; 2 ; 22 , r ≤ x

r≤z = n 2 2 22 22 2 2 22 2 2 22 22
p S q ; n ; z2 ; r2
M p+2 ; n+2 ; r2
+ C q ; n ; r2 ; z2
M p n r2

 n

 2 2 2 2
2 2 2 2 2 2 2 2 22 22 2 ; 2 ; 22 x≤z
q n x

U ; ;
 2

 2q n2 2z22
,


 z≤x
U 2 ; 2 ; 22

M 2p ; n2 ; 2x2
 2

2
, x≤z
M 2p ; n2 ; 2z2






 2U q ; n ; r2
C p ; n ; r2 ; x2
+ qU q+2 ; n ; r2
S p n r 2 x2


2 ; 2 ; 22 ; 22 , z ≤ x

z≤r = 2 2 22 2 2 22 22 2 2 22
2U q ; n ; r2
C p ; n ; r2 ; z2
+ qU q+2 ; n ; r2
S p n r2 z2



 2 2 2 2 2 2 2 2 2 2 2 2 22 2 ; 2 ; 22 ; 22 x≤r
21+n=2 n r−n er =2 U 2q ; n2 ; 2x2

 2 2 2



2U 2q ; n2 ; 2r22 C 2p ; n2 ; 2r22 ; 2z22 + qU q+2 n r2
p n r2 z2
, r ≤ x

2 ; 2 ; 22 S 2 ; 2 ; 22 ; 22



 2. STOPPING AT FIRST HITTING TIME 591

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 Hz = min{s : Q(sn) = z}

 x2 
z n=2 ex =4 M n ; n−2

2 2

2 2

 4γ 4
 z2  , x ≤ z



 xn=2 ez =4 M n ; n−2

 2 2
( 2 − 1) 22
 Z Hz  
(n) 2 4γ 4
2.9.1 Ex exp − 2 (Q ) ds = 2
4 0
s
 z n=2 ex2 =42 W n n−2 x

0<x


 4γ
; 4
2
2 , z ≤ x
 xn=2 ez2 =42 W n−2  z 2 




n ;
4γ 4 22

2n=2 n−2 y1−n ey =2 S 2 ; n2 ; 2x2 ; 2z2

2 2 2 2

−γ Ȟ(Hz ) (n)

2.13.1 Ex e ; sup Qs ∈ dy = dy
Ed n2 ; y ; z S 2 ; n2 ; 2y2 ; 2z22

2

0≤s≤Hz
z<x
x<y

Ex e−αHz−γ Ȟ(Hz ) ; sup Q(sn) ∈ dy


2.13.4 z<x<y
0≤s≤Hz

2n=2 n−2 y1−n ey =2 S 2+ ; n2 ; 2x2 ; 2z2

2 2 2 2

=
dy
S 2 ; n2 ; 2y2 ; 2z2 S 2+ ; n2 ; 2y2 ; 2z2
2
2 2 2

2n=2 n−2 y1−n ey =2 S 2 ; n2 ; 2z2 ; 2x2

2 2 2 2

Ex e−γ Ĥ(Hz ) ; Q(sn) ∈ dy =

2.14.1 inf dy
Ed n2 ; z ; y S 2 ; n2 ; 2z22 ; 2y2

2

0≤s≤Hz
x<z
y<x

Ex e−αHz−γ Ĥ(Hz ) ; Q(sn) ∈ dy


2.14.4 inf y<x<z
0≤s≤Hz

2n=2 n−2 y1−n ey =2 S 2+ ; n2 ; 2z2 ; 2x2

2 2 2 2

=
dy
S 2 ; n2 ; 2z2 ; 2y2 S 2+ ; n2 ; 2z2 ; 2y2
2
2 2 2

 x2 
n=2 ex2 =42 M q

z
+ 4 2 2

n; (n−2)2 γ 2

 4 16
 z2  , x ≤ z



x n=2 ez2 =42 M q


+ γ 2 2

Hz n; 2 2
ds
 Z  
 (n−2)

Ex exp −γ 2 σ 2 θ
4 16 4
2.20.1 =
Q(sn) 2  x2 


z n=2 ex =4 W n q (n−2)2 γ 2
0 2 2
0<x

+ 4 2 2

;


 4 16
 2 , z ≤ x

 xn=2 ez2 =42 W q (n−2)2 2 z 2




n;
4 16
+ γ4 2
592 8. RADIAL ORNSTEIN{UHLENBECK PROCESS

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 Hz = min{s : Q(sn) = z}

p2  2  (q2 − 1) (n) 2
 Z Hz   
2.21.3 Ex exp −αHz − (n)
2
+ (Q ) ds
0 Qs 42 s

 qx2 
x−n=2 ex =4 M nθ−2α q (n−2)2 p2
 2 2

; + 4 2 2


 4θq 16
 qz 2  , 0 < x ≤ z



z −n=2 ez2 =42 M nθ−2α q (n−2)2 p2


+ 4 2 2

;


4θq 16
=  qx2 
x−n=2 ex =4 W nθ−2α q (n−2)2 p2
 2 2

+ 4 2 2

;


 4θq 16
 qz 2  , z≤x


 z −n=2 ez2 =42 W


+ p42 2 2
 q
; (n−2)
nθ−2α 2
4θq 16

n  Z Hz

 o
p1I[0,r) Q(sn) + q1I[r,∞) Q(sn) ds ; ` Hz , r ∈ dy



2.27.1 Ex exp −
0

C p n r2 z2
q+2 n+2 r2

2 ; 2 ; 22 ; 22 + qU 2 ; 2 ; 22

z≤r = exp −yθr p n r2 z2
S 2 ; 2 ; 22 ; 22 2U 2q ; n2 ; 2r22

 n=2 n r2 =22 p n x2 z2

2  e S 2 ; 2 ; 22 ; 22
dy, z≤x≤r

n 1 2 p
r S 2 ; n2 ; 2r2 ; 2z2

 2 2

 −
×
2n=2 n er =2 U q ; n ; x 2
2 2 2

 n−1 q n r2
p2 n2 r22 z2
dy, r ≤ x




r U 2 ; 2 ; 22 S 2 ; 2 ; 22 ; 22

C q n r2 z2
p+2 n+2 r2

2 ; 2 ; 22 ; 22 + pM 2 ; 2 ; 22

r≤z = exp −yθr q n z2 r2

S 2 ; 2 ; 22 ; 22 nM 2p ; n2 ; 2r2
2

2n=2 n er =2 M 2p ; n2 ; 2x2

 2 2 2

 rn−1 M p ; n ; r2 S q ; n ; z2 ; r2
dy, 0 ≤ x ≤ r




2 2 22 2 2 22 22
×
 2n=2  n er2 =22 S q ; n ; z2 ; x2

2 2 22 22 dy,

r≤x≤z

rn−1 S 2 2q ; n2 ; 2z2 ; 2r2

 2 2

n  Z Hz

 o
p1I(−∞,r) Q(sn) + q1I[r,∞) Q(sn) ds ; ` Hz , r = 0



(1) Ex exp −
0

S 2p ; n2 ; 2r2 ; 2x2

 2 2

 S p ; n ; r2 ; z2
, z ≤ x ≤ r



2 2 22 22
=
 S 2q ; n2 ; 2x2 ; 2r2
2 2


2
, r≤x≤z
S 2q ; n2 ; 2z2 ; 2r2

 2
 3. STOPPING AT FIRST EXIT TIME 593

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 H = Ha,b = min{s : Q(sn) ∈
/ (a, b)}

3. Stopping at first exit time

S 2 ; n2 ; 2b2 ; 2x2 + S 2 ; n2 ; 2x2 ; 2a2

2 2
2 2

3.0.1 Ex e−αH =
S 2 ; n2 ; 2b2 ; 2a2
2 2

n; b ; x

Ed Ed n; x; a

3.0.3 Ex e −βQH
(n)
=2   −βa + 2   −βb
n; b ; a
e
Ed Ed n; b ; a
e
2   2  

Ed n2 ; b ; x


3.0.4 Px Q(Hn) =a =
Ed n2 ; b ; a


(a)

Ed n2 ; x ; a


3.0.4 Px Q(Hn) = b =
Ed n2 ; b ; a


(b)
S ; n; b 2 ; x 2
2 2

Ex e−αH ; Q(Hn) = a = 2 n2 2b2 2a2


3.0.5
S 2 ; 2 ; 22 ; 22
(a)

S ; n; x 2 ; a 2
2 2

−αH (n)
= b = 2 n2 2b2 2a2


3.0.5 Ex e ; QH
S 2 ; 2 ; 22 ; 22
(b)

Ed n2 ; b ; y Ed n; x; a



3.1.6 Px sup Q(sn) ≥ y, Q(Hn) = a = 2   a≤x≤y≤b
Ed n2 ; b ; a Ed

n; y ; a
,
0≤s≤H 2  

Ex e−αH ; sup Q(sn) ≥ y, Q(Hn) = a


3.1.8
0≤s≤H
(1)
S 2 ; n2 ; 2b2 ; 2y2 S 2 ; n2 ; 2x2 ; 2a2
2 2
2 2

= 2
, a≤x≤y≤b
S 2 ; n2 ; 2b2 ; 2a2 S 2 ; n2 ; 2y2 ; 2a2
2 2
2

Ed n2 ; b ; x Ed

n; y ; a

3.2.6 Px infQ(sn) ≤ y, Q(Hn) = b = 2   a≤y≤x≤b
Ed n2 ; b ; y Ed

n; b ; a
,
0≤s≤H 2  

Ex e−αH ; inf Q(sn) ≤ y, Q(Hn) = b


3.2.8
0≤s≤H
(1)
S ; n; b 2 ; x 2 S ; n; y 2 ; a 2
2 2
2 2

= 2 n2 2b2 2y2
2 n2 2b2 2a2
, a≤y≤x≤b
S 2 ; 2 ; 22 ; 22 S 2 ; 2 ; 22 ; 22
594 8. RADIAL ORNSTEIN{UHLENBECK PROCESS

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 H = Ha,b = min{s : Q(sn) ∈
/ (a, b)}

3.3.1 Ex e−γ`(H,r)

 n=2 n r2 =22
2  e Ed n; b ; a
+ rn−1 Ed n ; x ; r
Ed n; r ; a

2   2   2   r≤x≤b
 2n=2 n er2 =22 Ed n; b ; a
+ rn−1 Ed n ; b ; r
Ed n; r ; a
,


= 2   2   2  
n=2 n r2 =22 n; b ; a
+ rn−1 Ed n ; b ; r
Ed n; r ; x

 2  e 2 2 Ed 2   2   2  

n; r ; a
, a≤x≤r

2n=2 n er =2 Ed n; b ; a
+ rn−1 Ed n ; b ; r
Ed
2   2   2  

 2n=2 n er2 =22 Ed n ; b ; a
y 

3.3.2


Px `(H, r) ∈ dy = exp − n−1 2  

r Ed n2 ; b ; r Ed n2 ; r ; a

Ed n2 ; b ; x Ed n2 ; b ; a
 n=2 n r2 =22
2  e



dy, r ≤ x ≤ b
rn−1 Ed2 n2 ; b ; r Ed n2 ; r ; a





×
n=2 n r2 =22 Ed n ; x ; a
Ed n ; b ; a

 2  e 2
  2
  dy, a ≤ x ≤ r


rn−1 Ed n2 ; b ; r Ed2 n2 ; r ; a

Ed n2 ; x ; r



, r ≤ x ≤ b

 Ed n2 ; b ; r


(1) Px `(H, r) = 0 =
Ed n2 ; r ; x



, a ≤ x ≤ r

Ed n2 ; r ; a



3.3.3 Er e−αH−γ`(H,r) =: L

2n=2 n er =2 S 2 ; n2 ; 2b2 ; 2r2 + S 2 ; n2 ; 2r2 ; 2a2

2 2
2 2

2 2
 

= n=2 n r2 =22
2  e S 2 ; n2 ; 2b2 ; 2a2 + rn−1 S 2 ; n2 ; 2b2 ; 2r2 S 2 ; n2 ; 2r2 ; 2a2
2 2
2 2
2 2

S n x2 r 2
n b2 x2

2 ; 2 ; 22 ; 22 + S 2 ; 2 ; 22 ; 22 L,

r≤x≤b

n b2 r2
S 2 ; n2 ; 2b2 ; 2r2

 S
2 ; 2 ; 22 ; 22
 2 2

−αH−γ`(H,r)
Ex e =
 S n r 2 x2
n x2 a2

 2 ; 2 ; 22 ; 22 + S 2 ; 2 ; 22 ; 22 L, a≤x≤r
n r 2 a2 S 2 ; n2 ; 2r2 ; 2a2

S 2 ; 2 ; 22 ; 22

2 2

2n=2 n er =2 S 2 ; n2 ; 2b2 ; 2a2 y 

 2 2 2 2

Ex e−αH ; `(H, r) ∈ dy = exp − n−1


3.3.4
r S 2 ; n2 ; 2b2 ; 2r2 S 2 ; n2 ; 2r2 ; 2a2
2 2
2 2

S 2 ; n2 ; 2b2 ; 2x2 S 2 ; n2 ; 2b2 ; 2r2 + S 2 ; n2 ; 2r2 ; 2a2

2 2
2 2
2 2

 h i

 2−n=2 −n e−r2 =22 rn−1 S 2 ; n ; b2 ; r2
S ; n ; r2 ; a2
dy, r ≤ x ≤ b



2 2 22 22 2 2 22 22
×
S 2 ; 2 ; 22 ; 22 S 2 ; 2 ; 22 ; 22 + S 2 ; n2 ; 2r2 ; 2a2
n x a n b r
2 2
2 2
2 2

h i


2
dy, a≤x≤r

r S 2 ; n2 ; 2b2 ; 2r2 S 2 2 ; n2 ; 2r2 ; 2a2
 −n=2 −n −r2 =22 n−1
2

 e
2 2
2
 3. STOPPING AT FIRST EXIT TIME 595

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 H = Ha,b = min{s : Q(sn) ∈
/ (a, b)}

S 2 ; n2 ; 2x2 ; 2r2
 2 2

 S ; n ; b2 ; r2
, r ≤ x ≤ b



2 2 22 22
Ex e−αH ; `(H, r) = 0 =

(1)
S 2 ; n2 ; 2r2 ; 2x2
2 2


2
, a≤x≤r

S 2 ; n2 ; 2r2 ; 2a2

 2

Ex e−γ`(H,r) ; Q(Hn) = a

3.3.5
(a)

2n=2 n er =2 Ed n2 ; b ; x
 2 2

 2n=2 n er2 =22 Ed n ; b ; a + rn−1 Ed n ; b ; r
Ed


n; r ; a
, r≤x≤b
= 2   2   2  
 2n=2  n er2 =22 Ed n ; b ; x
+ rn−1 Ed n ; b ; r
Ed n; r ; x

2  
2  
2  

n; r ; a
, a≤x≤r

2n=2 n er2 =22 Ed n2 ; b ; a + rn−1 Ed n2 ; b ; r Ed

2  

Ex e−γ`(H,r) ; Q(Hn) = b

3.3.5
(b)

Ed n2 ; x ; a + rn−1 Ed n2 ; x ; r Ed n; r ; a

 n=2 n r2 =22
2  e



2   r≤x≤b
 2n=2 n er2 =22 Ed n ; b ; a + rn−1 Ed n ; b ; r
Ed n; r ; a
,



= 2   2   2  
n= 2
2  e n r = 2 n
Ed 2 ;  ; a
x
2 2


n; r ; a
, a≤x≤r

2n=2 n er2 =22 Ed n2 ; b ; a + rn−1 Ed n2 ; b ; r Ed



2  

 2n=2 n er2 =22 Ed n ; b ; a
y 

3.3.6 Px `(H, r) ∈ dy, Q(Hn) = a = exp − n−1

2  

r Ed n2 ; b ; r Ed n2 ; r ; a


(a)

2n=2 n er =2 Ed n2 ; b ; x
 2 2

 rn−1 Ed n ; b ; r Ed n ; r ; a
dy, r ≤ x ≤ b



2   2  
× n=2  n er2 =22 Ed n ; x ; a

  2 2
  dy, a≤x≤r
rn−1 Ed n2 ; r ; a 2





 0, r≤x≤b
(1) Px `(H, r) = 0, QH (n)
=a = Ed n2 ; r ; x



, a ≤ x ≤ r
Ed n2 ; r ; a


 2n=2 n er2 =22 Ed n ; b ; a
y 

3.3.6 Px `(H, r) ∈ dy, QH (n)


= b = exp − n−1 2  

r Ed n2 ; b ; r Ed n2 ; r ; a


(b)
596 8. RADIAL ORNSTEIN{UHLENBECK PROCESS

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 H = Ha,b = min{s : Q(sn) ∈
/ (a, b)}

Ed n2 ; b ; x
 n=2 n r2 =22
2  e


dy, r≤x≤b
rn−1 Ed n2 ; b ; r 2




×
n=2 n r2 =22 Ed n ; x ; a

 2  e 2  
dy, a ≤ x ≤ r


rn−1 Ed n2 ; b ; r Ed n2 ; r ; a

n x r

 Ed 2 ;  ; 
, r ≤ x ≤ b


Px `(H, r) = 0, Q(Hn) Ed n2 ; b ; r


(1) =b =
0, a≤x≤r


Ex e−αH−γ`(H,r) ; Q(Hn) = a

3.3.7
(a)

2n=2 n er =2 S ; n ; b 2 ; x 2 + rn−1 S 2 ; n2 ; 2b2 ; 2r2 S 2 ; n2 ; 2r2 ; 2x2

2 2 2 2
2 2
2 2

a≤x = n=2 n r2 =22 2 n2 2b2 2a2

2  e S 2 ; 2 ; 22 ; 22 + rn−1 S 2 ; n2 ; 2b2 ; 2r2 S 2 ; n2 ; 2r2 ; 2a2
2 2
2 2

x≤r

2n=2 n er =2 S 2 ; n2 ; 2b2 ; 2x2

2 2 2 2

r≤x = n=2
2  en r 2 =2 2
S 2 ; 2 ; 2b2 ; 2a2 + rn−1 S 2 ; n2 ; 2b2 ; 2r2 S 2 ; n2 ; 2r2 ; 2a2
n 2 2
2 2
2 2

x≤b

Ex e−αH−γ`(H,r) ; Q(Hn) = b

3.3.7
(b)

2n=2 n er =2 S 2 ; n2 ; 2x2 ; 2a2

2 2 2 2

a≤x =
2n=2 n er =2 S 2 ; 2 ; 2b2 ; 2a2 + rn−1 S 2 ; n2 ; 2b2 ; 2r2 S 2 ; n2 ; 2r2 ; 2a2
n
2 2 2 2
2 2
2 2

x≤r

2n=2 n er =2 S ; n ; x 2 ; a 2 + rn−1 S 2 ; n2 ; 2x2 ; 2r2 S 2 ; n2 ; 2r2 ; 2a2

2 2 2 2
2 2
2 2

r≤x = n=2 n r2 =22 2 n2 2b2 2a2

2  e S 2 ; 2 ; 22 ; 22 + rn−1 S 2 ; n2 ; 2b2 ; 2r2 S 2 ; n2 ; 2r2 ; 2a2
2 2
2 2

x≤b

Ex e−αH ; `(H, r) ∈ dy, Q(Hn) = a


3.3.8
(a)
2n=2 n er =2 S 2 ; n2 ; 2b2 ; 2a2 y 
 2 2 2 2

= exp − n−1
r S 2 ; n2 ; 2b2 ; 2r2 S 2 ; n2 ; 2r2 ; 2a2
2 2
2 2

2n=2 n er =2 S 2 ; n2 ; 2b2 ; 2x2

 2 2 2 2

 rn−1 S ; n ; b2 ; r2 S ; n ; r2 ; a2
dy, r ≤ x ≤ b




2 2 22 22 2 2 22 22
×
n=2
2  e n r = 2 S 2 ; 2 ; 22 ; 2a2
n x
2 2 2 2


dy, a≤x≤r

rn−1 S 2 2 ; n2 ; 2r2 ; 2a2

 2 2

 3. STOPPING AT FIRST EXIT TIME 597

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 H = Ha,b = min{s : Q(sn) ∈
/ (a, b)}

 0, r≤x≤b


S 2 ; n2 ; 2r2 ; 2x2
−αH (n) 2 2


(1) Ex e ; `(H, r) = 0, QH =a =
2
, a≤x≤r
S 2 ; n2 ; 2r2 ; 2a2
 2

Ex e−αH ; `(H, r) ∈ dy, Q(Hn) = b


3.3.8
(b)
2n=2 n er =2 S 2 ; n2 ; 2b2 ; 2a2 y 
 2 2 2 2

= exp − n−1
r S 2 ; n2 ; 2b2 ; 2r2 S 2 ; n2 ; 2r2 ; 2a2
2 2
2 2

S 2 ; n2 ; 2b2 ; 2x2
 n=2 n r2 =22
2  e
2 2

dy, r≤x≤b

rn−1 S 2 2 ; n2 ; 2b2 ; 2r2

 2 2


×
2n=2 n er =2 S 2 ; n2 ; 2x2 ; 2a2
2 2 2 2


2
dy, a≤x≤r

r S 2 ; n2 ; 2b2 ; 2r2 S 2 ; n2 ; 2r2 ; 2a2
 n−1

2 2
2

n x2 r2

 S 2 ; 2 ; 22 ; 22 ,

r≤x≤b
Ex e−αH ; `(H, r) = 0, Q(Hn) S 2 ; n2 ; 2b2 ; 2r2
 2 2

(1) =b =

0, a≤x≤r

 Z H

3.4.1 Er exp −γ 1I[r,b] Q(sn) ds
0

2n=2 n er =2 S 2 ; n2 ; 2b2 ; 2r2 + Ed n2 ; r ; a

2 2

2 2



= n=2 n r2 =22 2
=: Q
2  e S 2 ; n2 ; 2b2 ; 2r2 + rn Ed n2 ; r ; a C 2 ; n2 ; 2r2 ; 2b2
2 2

2

 Z H

Ex exp −γ 1I[r,b] Q(sn) ds
0

 Ed n ; r ; x
Ed n ; x ; a

2   2   a≤x≤r
 Ed n2 ; r ; a
+ Ed n2 ; r ; a
Q,


=
S 2 ; n2 ; 2x2 ; 2r2 S 2 ; n2 ; 2b2 ; 2x2
2 2
2 2


+ 2
Q, r≤x≤b
S 2 ; n2 ; 2b2 ; 2r2 S 2 ; n2 ; 2b2 ; 2r2

 2 2
2

n r x

  Ed 2 ;  ; 
, a ≤ x ≤ r

Z H
Q(sn) ds = 0 = Ed n2 ; r ; a


3.4.2 Px 1I[r,b]
(1) 0 
0, r≤x≤b
598 8. RADIAL ORNSTEIN{UHLENBECK PROCESS

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 H = Ha,b = min{s : Q(sn) ∈
/ (a, b)}

n  Z H
 o
3.4.5 Ex exp −γ 1I[r,b] Q(sn) ds ; Q(Hn) = a
(a) 0

2n=2 n er =2 S ; n ; b 2 ; r 2 + rn Ed n ; r ; x
C n r2 b2

2 ; 2 ; 22 ; 22
2 2 2 2

a≤x = n=2 n r2 =22 2 n2 2b2 2r2
n 2  
n ; r ; a
C n r2 b2

x≤r
2  e S 2 ; 2 ; 22 ; 22 + r Ed 2   2 ; 2 ; 22 ; 22

2n=2 n er =2 S 2 ; n2 ; 2b2 ; 2x2

2 2 2 2

r≤x = n=2 n r2 =22
2  e S 2 ; n2 ; 2b2 ; 2r2 + rn Ed n2 ; r ; a C 2 ; n2 ; 2r2 ; 2b2
2 2

2 2

x≤b

n  Z H
 o
3.4.5 Ex exp −γ 1I[r,b] Q(sn) ds ; Q(Hn) = b
(b) 0

2n=2 n er =2 Ed n2 ; x ; a
2 2

a≤x = n=2 n r2 =22
2  e S 2 ; n2 ; 2b2 ; 2r2 + rn Ed n2 ; r ; a C 2 ; n2 ; 2r2 ; 2b2
2 2

2 2

x≤r

2n=2 n er =2 S ; n ; x 2 ; r 2 + rn Ed n ; r ; a
C n r2 x2

2 ; 2 ; 22 ; 22
2 2 2 2

r≤x = n=2 n r2 =22 2 n2 2b2 2r2
n 2  
n; r ; a C n r2 b2

2  e S 2 ; 2 ; 22 ; 22 + r Ed 2 ; 2 ; 22 ; 22


x≤b 2  

n r x

  Ed 2 ;  ; 
, a ≤ x ≤ r

Z H
Q(sn) ds = 0, Q(Hn) = a = Ed n2 ; r ; a


3.4.6 Px 1I[r,b]
(1) 0 
0, r≤x≤b

 Z H

3.5.1 Er exp −γ 1I[a,r] Q(sn) ds
0

2n=2 n er =2 S 2 ; n2 ; 2r2 ; 2a2 + Ed n2 ; b ; r

2 2

2 2



= n=2 n r2 =22 2
=: Q
2  e S 2 ; n2 ; 2r2 ; 2a2 + rn Ed n2 ; b ; r C 2 ; n2 ; 2r2 ; 2a2
2 2

2

 Z H

Ex exp −γ 1I[a,r] Q(sn) ds
0

S 2 ; n2 ; 2r2 ; 2x2 S 2 ; n2 ; 2x2 ; 2a2

 2 2
2 2

2
+ 2
Q, a≤x≤r
S 2 ; n2 ; 2r2 ; 2a2 S 2 ; n2 ; 2r2 ; 2a2


 2 2

=
 Ed n ; x ; r
Ed n ; b ; x

 2  
+ 2  
Q, r≤x≤b
Ed n2 ; b ; r Ed n2 ; b ; r



Z H   0, a≤x≤r
1I[a,r] Q(sn) ds = 0 = Ed n2 ; x ; r


3.5.2 Px



, r ≤ x ≤ b
(1) 0
Ed n2 ; b ; r

 3. STOPPING AT FIRST EXIT TIME 599

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 H = Ha,b = min{s : Q(sn) ∈
/ (a, b)}

n  Z H
 o
3.5.5 Ex exp −γ 1I[a,r] Q(sn) ds ; Q(Hn) = a
(a) 0

2n=2 n er =2 S ; n ; r 2 ; x 2 + rn Ed n ; b ; r
C n r 2 x2

2 ; 2 ; 22 ; 22
2 2 2 2

a≤x = n=2 n r2 =22 2 n2 2r2 2a2
n 2  
n; b ; r C n r 2 a2

2  e S 2 ; 2 ; 22 ; 22 + r Ed 2 ; 2 ; 22 ; 22


x≤r 2  

2n=2 n er =2 Ed n2 ; b ; x
2 2

r≤x = n=2 n r2 =22
2  e S 2 ; n2 ; 2r2 ; 2a2 + rn Ed n2 ; b ; r C 2 ; n2 ; 2r2 ; 2a2
2 2

2 2

x≤b

n  Z H
 o
3.5.5 Ex exp −γ 1I[a,r] Q(sn) ds ; Q(Hn) = b
(b) 0

2n=2 n er =2 S 2 ; n2 ; 2x2 ; 2a2

2 2 2 2

a≤x = n=2 n r2 =22

2  e S 2 ; n2 ; 2r2 ; 2a2 + rn Ed n2 ; b ; r C 2 ; n2 ; 2r2 ; 2a2
2 2

2 2

x≤r

2n=2 n er =2 S ; n ; r 2 ; a 2 + rn Ed n ; x ; r
C n r 2 a2

2 ; 2 ; 22 ; 22
2 2 2 2

r≤x = n=2 n r2 =22 2 n2 2r2 2a2
n 2  
n; b ; r C n r 2 a2

2  e S 2 ; 2 ; 22 ; 22 + r Ed 2 ; 2 ; 22 ; 22


x≤b 2  


Z H   0, a≤x≤r
Q(sn) ds = 0, Q(Hn) = b = Ed n2 ; x ; r


3.5.6 Px 1I[a,r]



, r ≤ x ≤ b
(1) 0
Ed n2 ; b ; r


 Z H
 
p1I[a,r] Q(sn) + q1I[r,b] Q(sn) ds =: Q


3.6.1 Er exp −
0

2n=2 n r−n er =2 S 2q ; n2 ; 2b2 ; 2r2 + S 2p ; n2 ; 2r2 ; 2a2

2 2
2 2

2 2
 

=
S 2q ; n2 ; 2b2 ; 2r2 C 2p ; n2 ; 2r2 ; 2a2 + C 2q ; n2 ; 2r2 ; 2b2 S 2p ; n2 ; 2r2 ; 2a2
2 2
2 2
2 2
2 2

 Z H
 
p1I[a,r] Q(sn) + q1I[r,b] Q(sn) ds


Ex exp −
0

S 2q ; n2 ; 2x2 ; 2r2 S 2q ; n2 ; 2b2 ; 2x2

 2 2
2 2

+ 2
Q, r≤x≤b

 S q ; n ; b2 ; r2 S 2q ; n2 ; 2b2 ; 2r2


2
2 2 22 22
=
S 2p ; n2 ; 2r2 ; 2x2 S 2p ; n2 ; 2x2 ; 2a2
2 2
2 2


+
Q, a ≤ x ≤ r

S 2p ; n2 ; 2r2 ; 2a2 S 2p ; n2 ; 2r2 ; 2a2

 2 2
2 2
600 8. RADIAL ORNSTEIN{UHLENBECK PROCESS

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 H = Ha,b = min{s : Q(sn) ∈
/ (a, b)}

n  Z H
  o
p1I[a,r] Q(sn) + q1I[r,b] Q(sn) ds ; Q(Hn) = a


3.6.5 Ex exp −
(a) 0

q n b2 r2
p n r 2 x2
q n r2 b2
p n r 2 x2

a≤x =
S 2 ; 2 ; 22 ; 22 C 2 ; 2 ; 22 ; 22 + C 2 ; 2 ; 22 ; 22 S 2 ; 2 ; 22 ; 22
q n b2 r2
p n r 2 a2
q n r2 b2
p n r 2 a2

x≤r
S 2 ; 2 ; 22 ; 22 C 2 ; 2 ; 22 ; 22 + C 2 ; 2 ; 22 ; 22 S 2 ; 2 ; 22 ; 22

2n=2 n r−n er =2 S 2q ; n2 ; 2b2 ; 2x2

2 2 2 2

r≤x =
S 2q ; n2 ; 2b2 ; 2r2 C 2p ; n2 ; 2r2 ; 2a2 + C 2q ; n2 ; 2r2 ; 2b2 S 2p ; n2 ; 2r2 ; 2a2
2 2
2 2
2 2
2 2

x≤b

n  Z H
  o
p1I[a,r] Q(sn) + q1I[r,b] Q(sn) ds ; Q(Hn) = b


3.6.5 Ex exp −
(b) 0

2n=2 n r−n er =2 S 2p ; n2 ; 2x2 ; 2a2

2 2 2 2

a≤x = r2
C p ; n ; r2
S 2q ; n2 ; 2b2 ; 22 a2 q n r2 b2
p n r2 a2

2 2 22 ; 22 + C 2 ; 2 ; 22 ; 22 S 2 ; 2 ; 22 ; 22
2

x≤r

q n x2 r 2
q n r 2 a2
q n r 2 x2
q n r 2 a2

r≤x =
S 2 ; 2 ; 22 ; 22 C 2 ; 2 ; 22 ; 22 + C 2 ; 2 ; 22 ; 22 S 2 ; 2 ; 22 ; 22
q n b2 r2
q n r 2 a2
q n r2 b2
q n r 2 a2

x≤b
S 2 ; 2 ; 22 ; 22 C 2 ; 2 ; 22 ; 22 + C 2 ; 2 ; 22 ; 22 S 2 ; 2 ; 22 ; 22

( 2 − 1)
n  Z H  o
(n) 2 (n)
3.9.5 Ex exp − (Q ) ds ; Q = a
42 0
s H
(a)
ex (1− )=4 S n4 − 4nq ; n2 ; 2b2 ; x22
2 2 2 2

=
ea2 (1− )=42 S n4 − 4nq ; n2 ; 2b2 ; a22
2 2

( 2 − 1)
n  Z H  o
3.9.5 Ex exp − 2 (Q(sn) )2 ds ; Q(Hn) = b
4 0
(b)
ex (1− )=4 S n4 − 4nq ; n2 ; x22 ; a22
2 2 2 2

= 2
eb (1− )=42 S n4 − 4nq ; n2 ; 2b2 ; a22
2 2

2n=2 n−2 y1−n ey =2 S 2+ ; n2 ; 2x2 ; 2a2

Z b 2 2 2 2

Ex e−αH−γ Ȟ(H) ; Q(Hn) = a =


3.12.7
dy
S 2 ; n2 ; 2y2 ; 2a2 S 2+ ; n2 ; 2y2 ; 2a2
2
2 2 2
(a) x

2n=2 n−2 y1−n ey =2 S 2+ ; n2 ; 2b2 ; 2x2

Z x 2 2 2 2

−αH−γ Ĥ(H) (n)

3.12.7 Ex e ; QH = b =
dy
S 2 ; n2 ; 2b2 ; 2y2 S 2+ ; n2 ; 2b2 ; 2y2
2
2 2 2

(b) a
 3. STOPPING AT FIRST EXIT TIME 601

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 H = Ha,b = min{s : Q(sn) ∈
/ (a, b)}

H
ds
n  Z  o
3.20.5 2 2
Ex exp −p σ θ (n)
2
; Q(Hn) = a
0 Qs
(a)
√
x1−n=2+ (n−2)2+4p2 =2 S (n−2)2 + p2 + 2−n ; (n−2)2 + p2 + 1; b2 ; x2
q q 
16 4 4 4 22 22
= √
1 n=2+ (n 2)2+4p2 =2 (n−2)2 p 2−n (n−2)2 2 b2 a2
q q
a S 16 + 4 + 4 ; 4 + p + 1; 22 ; 22
2

− −

H
ds
n  Z  o
3.20.5 Ex exp −p2 σ 2 θ (n)
2
; Q(Hn) = b
0 Qs
(b)
√
x1−n=2+ (n−2)2+4p2 =2 S (n−2)2 + p2 + 2−n ; (n−2)2 + p2 + 1; x2 ; a2
q q 
16 4 4 4 22 22
= √
b1−n=2+ (n−2) +4p =2 S (n 2) p 2−n (n 2) 2 b2 a2
q q
16 + 4 + 4 ; 4 + p + 1; 22 ; 22
2 2
2

2 2 − −

p2  2  (q2 − 1) (n) 2
H o n=2−1
x
n  Z  
(n)
3.21.5 Ex exp − (n)
2
+ (Q ) ds ; Q = a
0 Qs 42 s H an=2−1
(a)
√
x (n−2)2+4p2 =2 ex (1−q)=4 S (n−2)2 + p2 + 2q−n ; (n−2)2 + p2 + 1; b2 q ; x2 q
2 2
q q 
16 4 4q 4 22 22
= √
(n 2)2+4p2 =2 a2 (1−q )=4 2 (n−2)2 p 2q −n (n−2)2 2 b2 q a2 q
q q
a e S 16 + 4 + 4q ; 4 + p + 1; 22 ; 22
 2

−

p2  2  (q2 − 1) (n) 2
H o n=2−1
x
n  Z  
(n)
3.21.5 Ex exp − + (Q ) ds ; Q = b
Q(sn) 2 42 s H bn=2−1


(b) 0

√
x (n−2)2+4p2 =2 ex (1−q)=4 S (n−2)2 + p2 + 2q−n ; (n−2)2 + p2 + 1; x2 q ; a2 q
2 2
q q 
16 4 4q 4 22 22
= √
(n 2)2+4p2 =2 b2 (1−q )=4 2 (n−2)2 p 2q−n (n−2)2 2 b2 q a2 q
q q
b e S 16 + 4 + 4q ; 4 + p + 1; 22 ; 22
2

−

n  Z H

 o
p1I[a,r) Q(sn) + q1I[r,b] Q(sn) ds ; `(H, r ∈ dy



3.27.1 Ex exp −
0

C p n r 2 a2
q n r2 b2

2 ; 2 ; 22 ; 22 + C 2 ; 2 ; 22 ; 22

n/2 n r 2 /2σ 2
= θ2 σ e exp −yθr p n r 2 a2

S 2 ; 2 ; 22 ; 22 S 2q ; n2 ; 2b2 ; 2r2
2 2

S 2q ; n2 ; 2b2 ; 2x2 S 2q ; n2 ; 2b2 ; 2r2 + S 2p ; n2 ; 2r2 ; 2a2

2 2
2 2
2 2

 h i

dy, r ≤ x ≤ b

rn−1 S 2 2q ; n2 ; 2b2 ; 2r2 S 2p ; n2 ; 2r2 ; 2a2

 2 2
2 2


×
S 2p ; n2 ; 2x2 ; 2a2 S 2q ; n2 ; 2b2 ; 2r2 + S 2p ; n2 ; 2r2 ; 2a2
2 2
2 2
2 2

h i


dy, a ≤ x ≤ r

rn−1 S 2q ; n2 ; 2b2 ; 2r2 S 2 2p ; n2 ; 2r2 ; 2a2

 2 2
2 2

602 8. RADIAL ORNSTEIN{UHLENBECK PROCESS

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 H = Ha,b = min{s : Q(sn) ∈
/ (a, b)}

n  Z H

 o
p1I[a,r) Q(sn) + q1I[r,b] Q(sn) ds ; `(H, r ∈ dy, Q(Hn) = a



3.27.5 Ex exp −
(a) 0

C p n r 2 a2
q n r2 b2

2 ; 2 ; 22 ; 22 + C 2 ; 2 ; 22 ; 22

= exp −yθr p n r 2 a2
S 2 ; 2 ; 22 ; 22 S 2q ; n2 ; 2b2 ; 2r2

2 2

2n=2 n er =2 S 2q ; n2 ; 2b2 ; 2x2

 2 2 2 2

 rn−1 S q ; n ; b2 ; r2
S p ; n ; r2 ; a2
dy, r ≤ x ≤ b



2 2 22 22 2 2 22 22
× p ; n ; x2 ; a2

 2n=2  n er 2
= 2 2
S 2 2 22 22 dy,

a≤x≤r

rn−1 S 2 2p ; n2 ; 2r2 ; 2a2

 2 2

n  Z H

 o
p1I[a,r) Q(sn) + q1I[r,b] Q(sn) ds ; ` Hz , r = 0, Q(Hn) = a



(1) Ex exp −
0

S p n r 2 x2

= 2 ; 2 ; 22 ; 22 , a≤x≤r
S p n r 2 a2

2 ; 2 ; 22 ; 22

n  Z H

 o
p1I[a,r) Q(sn) + q1I[r,b] Q(sn) ds ; ` Hz , r ∈ dy, Q(Hn) = b



3.27.5 Ex exp −
(b) 0

C p n r 2 a2
q n r2 b2

2 ; 2 ; 22 ; 22 + C 2 ; 2 ; 22 ; 22

= exp −yθr p n r 2 a2
q n b2 r2

S 2 ; 2 ; 22 ; 22 S 2 ; 2 ; 22 ; 22
 n=2 n r2 =22 q n b2 x2

2  e S 2 ; 2 ; 22 ; 22
dy, r≤x≤b

n 1 2 q
r S 2 ; n2 ; 2b2 ; 2r2

 2 2

 −
×
2n=2 n er =2 S p ; n ; x 2 ; a 2
2 2 2 2

 n−1 q n b2 r2
2 2p 2n r22 a2
dy, a ≤ x ≤ r




r S 2 ; 2 ; 22 ; 22 S 2 ; 2 ; 22 ; 22

n  Z H

 o
p1I[a,r) Q(sn) + q1I[r,b] Q(sn) ds ; ` H, r = 0, Q(Hn) = b



(1) Ex exp −
0

S q n x2 r2

= 2 ; 2 ; 22 ; 22 , r≤x≤b
S q n b2 r2

2 ; 2 ; 22 ; 22
 4. STOPPING AT INVERSE LOCAL TIME 603

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 % = %(v, z) = min{s : `(s, z) = v}

4. Stopping at inverse local time

M 2 ; n2 ; 2x2 2n=2 n2 n z 1−n ez =2 v

 2

2 2 
exp − , x≤z

 M ; n ; z2 n z2 n z2

2 M 2 ; 2 ; 22 U 2 ; 2 ; 22





2 2 22
4.0.1 Ex e−α% =
U 2 ; n2 ; 2x2 2n=2 n2 n z 1−n ez =2 v
2

2 2 
0≤x

exp − , z≤x

U 2 ; n2 ; 2z2 n z n z

2 M 2 ; 2 ; 22 U 2 ; 2 ; 22
 2

2
2

2n=2 n ez =2 v
  2 2 
 exp − n 1 n y z
, 0≤x≤z
z Ed 2 ;  ; 

 −
Px sup Q(sn)


4.1.2 <y =
Ed n2 ; y ; x 2n=2 n ez =2 v
2 2

 
0≤s≤%
z≤y

exp − , z≤x≤y
Ed n2 ; y ; z z n−1 Ed n2 ; y ; z



Ex e−α% ; sup Q(sn) < y


4.1.4
0≤s≤%
z≤y

M 2 ; n2 ; 2x2
2
 2n=2 n z 1−n ez2 =22 M ; n ; y2
v 
exp − 2 2 22 0≤x≤z
n ; y2 ; z2
,

 M ; n ; z22 n ; z2
S

M ; ;


2 2 2 2 2 22 2 2 22 22
=
S ; n; y ; x  2n=2 n z 1−n ez2 =22 M ; n ; y2
v 
2 2

 2 n2 2y22 2z22
exp − 2 2 22


, z≤x≤y

M 2 ; n2 ; 2z2 S 2 ; n2 ; 2y2 ; 2z2

S 2 ; 2 ; 22 ; 22
2
2 2

Ed n2 ; x ; y 2n=2 n ez =2 v
2 2

 
y exp − y , y≤x≤z
Ed n2 ; z ;  z n−1 Ed n2 ; z ; 





inf Q(sn)


4.2.2 Px >y =
 exp − 2n=2 n ez =2 v
,
2 2
0≤s≤%  
0≤y

n z y z≤x
z n−1 Ed 2 ;  ; 


Ex e−α% ; inf Q(sn) > y


4.2.4
0≤s≤%
y≤z

S 2 ; n2 ; 2x2 ; 2y2  2n=2 n z 1−n ez2 =22 U ; n ; y2
v 

 2 2

2 2 22
 S ; n ; z2 ; y2
exp − U ; n ; z2
S ; n ; z2 ; y2
, y ≤ x ≤ z



2 2 22 22 2 2 22 2 2 22 22
=
n x n=2 n
2  z e 1 n z = 2 U 2 ; n2 ; 2y2 v 
2

U 2 ; 2 ; 22
2
2 2
  −

2
exp − z≤x

,
U 2 ; n2 ; 2z2 U 2 ; n2 ; 2z2 S 2 ; n2 ; 2z2 ; 2y2

2

 2
2
604 8. RADIAL ORNSTEIN{UHLENBECK PROCESS

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 % = %(v, z) = min{s : `(s, z) = v}

2n=2 n rn−1 z 1−n ez =2 v

 2 2 
4.3.1 Ex e−γ`(%,r) = exp − n=2 n z2 =22
2  e + rn−1 | Ed n2 ; z ; r |

1, x ≤ z ≤ r, r≤z≤x

n ; x ; r
|
 2n=2 n er =2 + rn−1 | Ed

 2 2
2  


n ; z ; r
| , r∧z ≤x≤r∨z
× 2n=2 n er =22 + rn−1 | Ed
2
2  
2n=2 n er =2

 2 2

n ; z ; r
| , 0 ≤ x ≤ r ≤ z, z≤r≤x

2n=2 n er =2 + rn−1 | Ed

2 2
2  

 2n=2 n z 1−n ez2 =22 v + r1−n er2 =22 y


4.3.2 Px `(%, r) ∈ dy = exp −
| Ed n r z

2; ;  |
0≤x

√  21+n=2 n e(r2+z2 )=42 √vy 

2n=2 n e(r +z )=4 v
 2 2 2

I dy,
(rz )(n−1)=2 | Ed n2 ; r ; z | y √
(rz )(n−1)=2 | Ed n2 ; r ; z |





1



x ≤ z ≤ r, r ≤ z ≤ x








2n=2 n e(r +z )=4
√v  21+n=2 n e(r2+z2 )=42 √vy 

 2 2 2 h

n r x
Ed 2 σ σ √y I1
, ,

(rz )(n−1)=2 Ed2 n2 ; r ; z (rz )(n−1)=2 | Ed n2 ; r ; z |








×
(1−n)=2 er2 =42  21+n=2 n e(r2+z2 )=42 √vy i
n x z r
+ Ed 2 σ σ z (1−n)=2 ez 2 =4 2 I0
, , dy,
(rz )(n−1)=2 | Ed n2 ; r ; z |







r∧z ≤x≤r∨z








 21+n=2 n e(r2+z2 )=42 √vy 
2n=2 n er =2

 2 2

I dy,

rn−1 | Ed n2 ; r ; z | (rz )(n−1)=2 | Ed n2 ; r ; z |


0





z ≤ r ≤ x, x ≤ r ≤ z

(1) Px `(%, r) = 0

2n=2 n ez =2 v
  2 2 
exp − n r z , 0 ≤ x ≤ z ≤ r, r ≤ z ≤ x
z n−1 | Ed 2 ;  ;  |






z≤r = | Ed n r x
2 ;  ; 
| exp − 2  e
n=2 n z2 =22 v 


n r z
, r∧z ≤x≤r∨z
| Ed 2 ;  ;  | z n−1 | Ed n2 ; r ; z |





0, 0 ≤ x ≤ r ≤ z, z ≤ r ≤ x

4.3.3 Ex e−α%−γ`(%,r)
 4. STOPPING AT INVERSE LOCAL TIME 605

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 % = %(v, z) = min{s : `(s, z) = v}

2 2n 2n z 1−n e(r +z )=2 (n=2) −1 ( =2)M −1 2 ; n2 ; 2z2 v

 2 2 2 2

r≤z = exp − n=2 n r2 =22
2  e U 2 ; n2 ; 2z2 + rn−1 S 2 ; n2 ; 2r2 ; 2z2 U 2 ; n2 ; 2r2
2
2 2
2

  2n=2 n z 1−n rn−1 ez2 =22 M ; n ; r2
U ; n ; r2
M −1 ; n ; z2
v 

× exp − n=2 n r2 =22 2 2 22 2 2 22 2 2 22
2  e U 2 ; 2 ; 22 + r S 2 ; 2 ; 22 ; 22 U 2 ; n2 ; 2r2
n z n 1 n r z
2
2 2
2

−

U 2 ; n2 ; 2x2
 2


2
, z≤x
U 2 ; n2 ; 2z2





2n=2 n er =2 M 2 ; n2 ; 2x2 + rn−1 S 2 ; n2 ; 2x2 ; 2r2 M 2 ; n2 ; 2r2

 2 2 2
2 2
2


× 2
, r≤x
2n=2 n er2 =22 M 2 ; n2 ; 2z2 + rn−1 S 2 ; n2 ; 2z2 ; 2r2 M 2 ; n2 ; 2r2
2
2 2



 x≤z
2n=2 n er =2 M 2 ; n2 ; 2x2

 2 2 2



2
, x≤r
2n=2 n er2 =22 M 2 ; n2 ; 2z2 + rn−1 S 2 ; n2 ; 2z2 ; 2r2 M 2 ; n2 ; 2r2

 2
2 2

2 2n 2n z 1−n e(r +z )=2 (n=2) −1 ( =2)U −1 2 ; n2 ; 2z2 v

 2 2 2 2

z≤r = exp − n=2 n r2 =22
2  e M 2 ; n2 ; 2z2 + rn−1 S 2 ; n2 ; 2z2 ; 2r2 M 2 ; n2 ; 2r2
2
2 2
2

  2n=2 n z 1−n rn−1 ez2 =22 M ; n ; r2
U ; n ; r2
U −1 ; n ; z2
v 

× exp − n=2 n r2 =22 2 2 22 2 2 22 2 2 22
2  e M 2 ; n2 ; 2z2 + rn−1 S 2 ; n2 ; 2z2 ; 2r2 M 2 ; n2 ; 2r2
2
2 2
2

M 2 ; n2 ; 2x2
 2


2
, x≤z
M 2 ; n2 ; 2z2





2n=2 n er =2 U 2 ; n2 ; 2x2 + rn−1 S 2 ; n2 ; 2r2 ; 2x2 U 2 ; n2 ; 2r2

 2 2 2
2 2
2


× 2
, z≤x
2n=2 n er2 =22 U 2 ; n2 ; 2z2 + rn−1 S 2 ; n2 ; 2r2 ; 2z2 U 2 ; n2 ; 2r2
2
2 2



 x≤r
2n=2 n er =2 U 2 ; n2 ; 2x2

 2 2 2



2
, r≤x
2n=2 n er2 =22 U 2 ; n2 ; 2z2 + rn−1 S 2 ; n2 ; 2r2 ; 2z2 U 2 ; n2 ; 2r2

 2
2 2

Ex e−α% ; `(%, r) ∈ dy

4.3.4
0≤x

 2n=2 n z 1−n ez2 =22 M ; n ; r2
v 2n=2 n r1−n er2 =22 U ; n ; z2
y 

r≤z = exp − 2 2 22 − 2 2 22
S 2 ; n2 ; 2r2 ; 2z2 M 2 ; n2 ; 2z2 S 2 ; n2 ; 2r2 ; 2z2 U 2 ; n2 ; 2r2
2 2
2
2 2
2

2n=2 n e(r +z )=4

2 2 2

×
(rz )(n−1)=2 S 2 ; n2 ; 2r22 ; 2z22


606 8. RADIAL ORNSTEIN{UHLENBECK PROCESS

8 Q(sn) = σe−θs Re(n2θs )

−1
n ≥ 2, σ, θ > 0 % = %(v, z) = min{s : `(s, z) = v}
n x2
√ 1+n=2  n e(r2+z2 )=42 √vy 
 M 2 ; 2 ; 222
√v I1 2
 
dy, x≤z

M 2 ; n2 ; 2z2 y S 2 ; n2 ; 2r2 ; 2z2 (rz )(n−1)=2

 2 2





S 2 ; n2 ; 2r2 ; 2x2 √v  21+n=2 n e(r2+z2 )=42 √vy 

 2 2



n r2 z2
√y I1 S dy

n r2 z2
(n−1)=2
 S 2 ; 2 ; 22 ; 22 2 ; 2 ; 22 ; 22 (rz )






S 2 ; n2 ; 2x2 ; 2z2 r(1−n)=2 er2 =42  21+n=2 n e(r2+z2 )=42 √vy 
2 2

×
+ 2 I0 dy,
S 2 ; n2 ; 2r2 ; 2z2 z (1−n)=2 ez =4 S 2 ; n2 ; 2r2 ; 2z2 (rz )(n−1)=2

 2 2
2 2 2




z≤x≤r








 r(1−n)=2 er2 =42 U ; n ; x22
 21+n=2 n e(r2+z2 )=42 √vy 


 (1−n)=2 z2 =42 2 n2 2r2
I0

dy, r ≤ x

S 2 ; n2 ; 2r2 ; 2z2 (rz )(n−1)=2

z e U 2 ; 2 ; 22
2 2

 2n=2 n z 1−n ez2 =22 U ; n ; r2
v 2n=2 n r1−n er2 =22 M ; n ; z2
y 

z≤r = exp − 2 2 22 − 2 2 22
S 2 ; n2 ; 2z2 ; 2r2 U 2 ; n2 ; 2z2 S 2 ; n2 ; 2z2 ; 2r2 M 2 ; n2 ; 2r2
2 2
2
2 2
2

2n=2 n e(r +z )=4

2 2 2

×
(rz ) n−1)=2 S 2 ; n2 ; 2z22 ; 2r22
(

U 2 ; n2 ; 2x2 √v  21+n=2 n e(r2+z2 )=42 √vy 

 2

2
√ I1 dy, z≤x

U 2 ; n2 ; 2z2 y S 2 ; n2 ; 2z2 ; 2r2 (rz )(n−1)=2

 2 2






S 2 ; n2 ; 2x2 ; 2r2 √v  21+n=2 n e(r2+z2 )=42 √vy 

 2 2



2
√ I1 dy

S 2 ; n2 ; 2z2 ; 2r2 y S 2 ; n2 ; 2z2 ; 2r2 (rz )(n−1)=2

 2 2 2






S ; n ; z 2 ; x 2 r(1−n)=2 er2 =42  21+n=2 n e(r2+z2 )=42 √vy 
2 2

×
+ 2 n2 2z2 2r2
(1−n)=2 z2 =42 I0 dy,
S 2 ; 2 ; 22 ; 22 z e S 2 ; n2 ; 2z2 ; 2r2 (rz )(n−1)=2

 2 2




r≤x≤z








r(1−n)=2 er =4 M 2 ; n2 ; 2x2



 2 2  21+n=2 n e(r2+z2 )=42 √vy 
2

I dy, x≤r

z (1−n)=2 ez2 =42 M 2 ; n2 ; 2r2 S 2 ; n2 ; 2z2 ; 2r2 (rz )(n−1)=2
 0
 2
2 2

Ex e−α% ; ` %, r = 0


(1)

0, r≤x


n r x n= 2 n 1 n z = 2 n r
 S 2 ; 2 ; 22 ; 22
exp − 2  z e M 2 ; 2 ; 22 v
2 2
2

 2 2

  −
, z≤x≤r

= S 2 ; n2 ; 2r2 ; 2z2 S 2 ; n2 ; 2r2 ; 2z2 M 2 ; n2 ; 2z2
2 2 2 2
2

z≤r
M 2 ; n2 ; 2x2  2n=2 n z 1−n ez2 =22 M ; n ; r2
v 

 2

2 2 22


 exp − , 0≤x≤z
M 2 ; n2 ; 2z2 S 2 ; n2 ; 2r2 ; 2z2 M 2 ; n2 ; 2z2
 2
2 2
2

 4. STOPPING AT INVERSE LOCAL TIME 607

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 % = %(v, z) = min{s : `(s, z) = v}

0, 0≤x≤r


S ; n ; x 2
; r 2

 2n= 2  n z 1−n ez 2 =2 2 U ; n ; r v
2



2 2 22 22 exp − 2 2 22
 
, r≤x≤z

= S 2 ; n2 ; 2z2 ; 2r2 U 2 ; n2 ; 2z2 S 2 ; n2 ; 2z2 ; 2r2

2 2
2
2 2

r≤z
U 2 ; n2 ; 2x2  2n=2 n z 1−n ez2 =22 U ; n ; r2
v 

 2

2 2 22


 exp − , z≤x
U 2 ; n2 ; 2z2 U 2 ; n2 ; 2z2 S 2 ; n2 ; 2z2 ; 2r2
 2
2
2 2

 Z %

4.4.1 Ex exp −γ 1I[r,∞) Q(sn) ds
0≤x 0

 2n=2 n z 1−n rn ez =2 U +2  n+2 r2

2 ; 2 ; 22 v
 2 2 
z≤r = exp − 1+n=2 n r2 =22 n r n r z +2 n+2 r
2  e U 2 ; 2 ; 22 + rn Ed 2 ;  ;  U 2 ; 2 ; 22
2

2


 1, x≤z
1+n=2  n er2 =22 U n r2
n r x
+2 ; n+2 ; r2


n
 2 2 ; 2 ; 22 + r Ed 2 ;  ;  U

2 2 22


+2 ; n+2 ; r2
, z≤x

× 21+n=2 n er =2 U 2 ; n2 ; 2r2 + rn Ed n2 ; r ; z U
2 2 2


2 2 22 x≤r
21+n=2 n er =2 U 2 ; n2 ; 2x2

 2 2 2



+2 ; n+2 ; r2
, r≤x

U 2 ; n2 ; 2r2 + rn Ed n2 ; r ; z U
 1+n=2 n r2 =22
2

 e
2


2 2 22

 n=2 n −n r2 =22
2  r e
, x≤r
 C 2 ; n2 ; r22 ; z22





 2  2 
 2n=2 n z 1−n ez2 =22 U +2 ; n+2 ; r2
v  C 2 ; n2 ; 2r2 ; 2x2

 2 2

2  2 2 2
r≤x
r≤z = exp − 2
,
2U 2 ; n2 ; 2z22 C 2 ; n2 ; 2r22 ; 2z22 C 2 ; n2 ; 2r2 ; 2z2


2

 x≤z
U 2 ; n2 ; 2x2


 2


2
, z≤x

U 2 ; n2 ; 2z2



2n=2 n ez =2 v
  2 2 
exp − n r z
, x≤z
z n−1 Ed 2 ;  ; 




Z %  
n
Qs ds = 0 = Ed 2 ;  ; x

r

2n=2 n ez =2 v
(n)


2 2
4.4.2 Px 1I[r,∞)
 
n r z exp − n r z
, z≤x
(1) 0 


 Ed 2 ;  ;  z n−1 Ed 2 ;  ;  x≤r
0≤x

0, r≤x


 Z %

4.5.1 Ex exp −γ 1I[0,r] Q(sn) ds
0

 2n=2 n z 1−n ez2 =22 M +2 ; n+2 ; r2
v 

z≤r = exp − 2 2 22
nM 2 ; n2 ; 2z2 C 2 ; n2 ; 2r2 ; 2z2
2
2 2

608 8. RADIAL ORNSTEIN{UHLENBECK PROCESS

8 Q(sn) = σe−θs Re(n2θs )

−1
n ≥ 2, σ, θ > 0 % = %(v, z) = min{s : `(s, z) = v}

M 2 ; n2 ; 2x2
 2

2
, 0≤x≤z

M 2 ; n2 ; 2z2






C 2 ; n2 ; 2r2 ; 2x2
2 2



× 2
, z≤x≤r
C 2 ; n2 ; 2r2 ; 2z2
2



n=2 n −n r2 =22
 2  r 2e 2
, r ≤ x




C 2 ; n2 ; 2r2 ; 2z2


 2n=2 n z 1−n rn ez =2 M +2  n+2 r2

2 ; 2 ; 22 v
 2 2 
r≤z = exp −
n2n=2 n er2 =22 M 2 ; n2 ; 2r2 + rn Ed n2 ; z ; r M +2 n+2 r
2 ; 2 ; 22
2

2

n2n=2 n er =2 M 2 ; n2 ; 2x2

 2 2 2

n r2 n z r
+2 n+2 r2
, x ≤ r

n=2 n r2 =22 M n
 n2  e 2 ; 2 ; 22 + r Ed 2 ;  ;  M 2 ; 2 ; 22





× n2n=2 n er2 =22 M 2 ; n2 ; 2r22 + rn Ed n2 ; x ; r M +2  n+2 r2

2 ; 2 ; 22 , r ≤ x

n2n=2 n er2 =22 M 2 ; n2 ; 2r2 + rn Ed n2 ; z ; r M +2  n+2 r2

2 ; 2 ; 22

 2


x≤z




1, z≤x

 0, x≤r
n x r n= 2 n z = 2
 Ed 2 ;  ; 
exp − 2  e

v


 2 2 
Z %  
, r≤x
n z r z n−1 Ed n2 ; z ; r
1I[0,r] Q(sn) ds = 0 = Ed 2 ;  ; 


4.5.2 Px x≤z
(1) 0
n=2  n ez2 =22 v 

 2

 

0≤x  exp − n−1
, z≤x
Ed n2 ; z ; r

z

 Z %


p1I[0,r) Q(sn) + q1I[r,∞) Q(sn) ds


4.6.1 Ex exp −
0

p2n=2 n z 1−n ez =2 M p+2  n+2 r2

2 ; 2 ; 22 U
q n r2
−1 q ; n ; z2
v
2 ; 2 ; 22 U
 2 2

r≤z = exp − 2 2 22

pS 2q ; n2 ; 2z2 ; 2r2 M p+2  n+2 r2

2 ; 2 ; 22 + nC
q ; n ; r 2 ; z 2
M p ; n ; r 2

2 2

2 2 2 2 2 2 2 2 22

qn2n=2−1 n z 1−n ez =2 U q+2  n+2 r p n r

2 ; 2 ; 22 M 2 ; 2 ; 22 U
−1 q ; n ; z
2 2 22 v
2 2 2
2
2

− +2
pS 2 ; n2 ; 2z2 ; 2r2 M 2 ; 2 ; 2r2 + nC 2 ; n2 ; 2r2 ; 2z2 M 2p ; n2 ; 2r2
q p  n+2 q
2 2
2
2 2
2

2n=2 n r−n er =2 M 2p ; n2 ; 2x2

 2 2 2

p q n z2 r2
p+2 n+2 r2
q n r2 z2
p n r2
, x ≤ r

n S 2 ; 2 ; 22 ; 22 M 2 ; 2 ; 22 + C 2 ; 2 ; 22 ; 22 M 2 ; 2 ; 22





 p S q ; n ; x2 ; r2
M p+2 ; n+2 ; r2
+ C q ; n ; r2 ; x2
M p n r2


2 ; 2 ; 22 , r ≤ x

× n 2 2 22 22 2 2 22 2 2 22 22
p S q ; n ; z2 ; r2
M p+2 ; n+2 ; r2
+ C q ; n ; r2 ; z2
M p n r2



 n 2 2 2 2 2 2 2 2 2 2 2 2 22 22 2 ; 2 ; 22 x≤z
q n x

U ; ;
 2

 2q n2 2z22
,


 z≤x
U 2 ; 2 ; 22
 4. STOPPING AT INVERSE LOCAL TIME 609

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 % = %(v, z) = min{s : `(s, z) = v}

pn−1 21+n=2 n z 1−n ez =2 M p+2  n+2 r2
q n r2
−1 p ; n ; z2
v

2 ; 2 ; 22 U 2 ; 2 ; 22 M
 2 2

z≤r = exp − 2 2 22

2U 2q ; n2 ; 2r22 C 2p ; n2 ; 2r22 ; 2z22 + qU q+2  ; n ; r 2
S p ; n ; r 2 ; z 2




2 2 2 2 2 2 22 22

q2n=2 n z 1−n ez =2 U q+2  n+2 r p n r

2 ; 2 ; 22 M 2 ; 2 ; 22 M
−1 p ; n ; z
2 2 22 v
2 2 2
2
 2

−
2U 2q ; n2 ; 2r22 C 2p ; n2 ; 2r22 ; 2z22 + qU q+2 n r2 p n r2 z2
2 ; 2 ; 22 S 2 ; 2 ; 22 ; 22

M 2p ; n2 ; 2x2
 2

2
, x≤z
M 2p ; n2 ; 2z2






 2U q ; n ; r2
C p ; n ; r2 ; x2
+ qU q+2 ; n ; r2
S p n r 2 x2


2 ; 2 ; 22 ; 22 , z ≤ x

× 2 2 22 2 2 22 22 2 2 22
2U q ; n ; r2
C p ; n ; r2 ; z2
+ qU q+2 ; n ; r2
S p n r2 z2



 2 2 22 2 2 22 22 2 2 22 2 ; 2 ; 22 ; 22 x≤r
2 1+ n= 2 n n r = 2 U 2q ; n2 ; 2x2

 r e
2 2 2

−



2U 2q ; n2 ; 2r22 C 2p ; n2 ; 2r22 ; 2z22 + qU q+2  n r2
p n r2 z2
, r ≤ x

2 ; 2 ; 22 S 2 ; 2 ; 22 ; 22


n
−1 n − n

( 2 − 1)  zv
Z %  
2 4 4
 
(n) 2
4.9.1 Ex exp − (Q ) ds = exp −  z2   z2 
42 s
1≤γ 0 M n ; n−2 2 W n ; n−2
4γ 4 2 4γ 4 22
 x2 
−n=2 ex2 =4 2 M

x n ; n−2
22 , 0 < x ≤ z


 4γ 4

z2 


 z −n=2 ez2 =42 M n ; n−2

 
2  2

4γ 4
× 2
x
x−n=2 ex =4 W n ; n−2
 2 2

2  2


4γ 4

, z≤x
 z −n=2 ez2 =42 W n−2  z 2 




n ;
4γ 4 2 2

%
ds
 Z 
4.20.1 Ex exp −γ 2 σ 2 θ
Q(sn) 2


0

(n−2)2 + 2 + 1
q
zv
 
4
= exp −  z2   z2 
(n−2)2 + 2 − n + 1 M q
q
W

16 4 4 2 + γ 2 2 + γ 2 2
q
n; (n−2)2 2 n; (n−2)2 2
4 16 4 4 16 4

x2 
x−n=2 ex =4 M n q (n−2)2 γ 2
 
2 2

+ 4 2 2

;

 4 16
 z2  , 0<x≤z



z −n=2 ez2 =42 M n q (n−2)2 γ 2


+ 4 2 2

;


4 16
×
x2 
x−n=2 ex =4 W n
 2 2

22
 q
; + (n−2)2 γ2


 4 16 4

  z2  , z≤x
z −n=2 ez2 =42 W n q (n−2)2 γ 2


+ 4 2 2


4
; 16
610 8. RADIAL ORNSTEIN{UHLENBECK PROCESS

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 % = %(v, z) = min{s : `(s, z) = v}

 Z % 2 2
p   (q2 − 1) (n) 2
 
4.21.1 Ex exp − + (Q ) ds
Q(sn) 2 42 s


0<x 0

(n−2)2 + p2 + 1
q
qzv
  
4
= exp −  qz 2   qz 2 
(n−2)2 + p2 + 2q−n M q
q
W n q (n−2)2 p2
 
16 4 4q n ; (n−2)2
+ p2 2 2
; + 22
4q 16 4 4q 16 4

 qx2 
x−n=2 ex =4 M n
 2 2

22
q
; (n−2)2
+

 p2
 4q 16 4
 qz 2  , 0<x≤z



z −n=2 ez2 =42 M n


+ p42 2 2
 q
; (n−2)2


4q 16
×  qx2 
x−n=2 ex =4 W n q (n−2)2 p2
 2 2

22

; +


 4q 16 4
 qz 2  , z≤x


−n=2 ez2 =42 W

z

22

 q
n
4q
; (n−2)2
16
+ p2
4

n  Z %

 o
p1I[0,r) Q(sn) + q1I[r,∞) Q(sn) ds ; `(%, r) ∈ dy


4.27.1 Ex exp −
0≤x 0

 2n=2 n z 1−n ez2 =22 M p n r2

z≤r = exp − 2 ; 2 ; 22 v
S 2p ; n2 ; 2r2 ; 2z2 M p n z2

2 ; 2 ; 22
2 2

rqU q+2  ; n+2 ; r2
y rC 2p ; n2 ; 2r2 ; 2z2 y  2n=2 n e(r +z )=4
2 2
2 2 2

− 2 2 2 2
−
2U 2q ; n2 ; 2r2
2
S 2p ; n2 ; 2r2 ; 2z2
2 2
(rz )(n−1)=2 S 2p ; n2 ; 2r2 ; 2z22
2

M p n x2
√  21+n=2 n e(r2+z2 )=42 √vy 

2 ; 2 ; 22 v

p ; n ; z2
√y I1 S p ; n ; r2 ; z2
(rz )(n−1)=2 dy, x≤z


M

2 2 22 2 2 22 22





p n r2 x2
√  21+n=2 n e(r2+z2 )=42 √vy 

S 2 ; 2 ; 22 ; 22 v



p n r2 z2
√y 1 S p ; n ; r2 ; z2
(rz )(n−1)=2 dy
I

S 2 ; 2 ; 22 ; 22


2 2 22 22





p n x2 z2
(1−n)=2 r2 =42 
21+n=2 n e(r +z )=4 vy 
2√
× S 2 ; 2 ; 22 ; 22 r e
2 2

+ p n r2 z2 I
(1−n)=2 ez2 =42 0 S p ; n ; r2 ; z2 (rz )(n−1)=2 dy,
S 2 ; 2 ; 22 ; 22 z




2 2 22 22



z≤x≤r








r(1−n)=2 er =4 U q n x2


21+n=2 n e(r +z )=4 vy 
2√
2 ; 2 ; 22 I
2 2

 2 2

dy, r≤x

z (1−n)=2 ez2 =42 U q n r2 S 2p ; n2 ; 2r2 ; 2z2 (rz )(n−1)=2
 0
2 ; 2 ; 22

2 2

 4. STOPPING AT INVERSE LOCAL TIME 611

8 Q(sn) = σe−θs Re(n2θs

)
−1
n ≥ 2, σ, θ > 0 % = %(v, z) = min{s : `(s, z) = v}

 2n=2 n z 1−n ez2 =22 U q n r2

r≤z = exp − 2 ; 2 ; 22 v
S 2q ; n2 ; 2z2 ; 2r2 U q n z2

2 ; 2 ; 22
2 2

q n r2 z2
p+2 n+2 r2

rC 2 ; 2 ; 22 ; 22 y − rpM 2 ; 2 ; 22 y 2n=2 n e(r +z )=4
2 2 2

− q ; n ; z2 ; r2
p n r q
S nM 2 ; 2 ; 22
2

(rz )(n−1)=2 S 2 ; 2 ; 22 ; 2r22
n z 2

2 2 22 22

U q n x2
√  21+n=2 n e(r2+z2 )=42 √vy 

2 ; 2 ; 22 v

I dy, z≤x

q n z √
y S 2q ; n2 ; 2z2 ; 2r2 (rz )(n−1)=2
1

U 2 ; 2 ; 22

 2
2 2





q n x2 r 2
√  21+n=2 n e(r2+z2 )=42 √vy 

S 2 ; 2 ; 22 ; 22 v



q ; n ; z2 ; r2
√y I1 S q ; n ; z2 ; r2
(rz )(n−1)=2 dy

S


2 2 22 22 2 2 22 22





q n z2 x2
(1−n)=2 r2 =42 
21+n=2 n e(r +z )=4 vy 
2√
× S 2 ; 2 ; 22 ; 22 r e
2 2

+ I
q n z2 r2
(1−n)=2 ez2 =42 0 S q ; n ; z2 ; r2
(rz )(n−1)=2 dy,
S 2 ; 2 ; 22 ; 22 z


2 2 22 22



r≤x≤z








r(1−n)=2 er =4 M p n x2


21+n=2 n e(r +z )=4 vy 
2√
2 ; 2 ; 22 I
2 2

 2 2

p n r2
0 S q ; n ; z2 ; r2
(rz )(n−1)=2 dy, x≤r

z (1−n)=2 ez2 =42 M

2 ; 2 ; 22

2 2 22 22

n  Z %

 o
p1I[0,r) Q(sn) + q1I[r,∞) Q(sn) ds ; ` %, r = 0



(1) Ex exp −
0

0, r≤x

p p

n r x n= 2 n 1 n z = 2 n r
 S 2 ; 2 ; 22 ; 22
exp − 2  z e M 2 ; 2 ; 22 v
2 2
2

 2 2

  −
, z≤x≤r

p n r
S 2 ; 2 ; 22 ; 22
2 z 2
S 2 ; 2 ; 22 ; 22 M 2p ; n2 ; 2z2
p n r 2 z 2
2

z≤r =
 M 2p ; n2 ; 2x2  2n=2 n z 1−n ez2 =22 M p ; n ; r2
v 

 2

2 2 22

2
exp − , 0≤x≤z
M 2p ; n2 ; 2z2 S 2p ; n2 ; 2r2 ; 2z2 M 2p ; n2 ; 2z2

 2 2
2

0, 0≤x≤r

q −n ez 2 =2 2 U q ; n ; r 2 v 

S ; n ; x 2
; r 2

 2n= 2  n z 1



2 2 22 22 exp − 2 2 22
 
, r≤x≤z

S 2q ; n2 ; 2z2 ; 2r2 U 2q ; n2 ; 2z2 S 2q ; n2 ; 2z2 ; 2r2

2 2
2
2 2

r≤z =
U 2q ; n2 ; 2x2  2n=2 n z 1−n ez2 =22 U q ; n ; r2
v 

 2

2 2 22


2
exp − , z≤x
U 2q ; n2 ; 2z2 U 2q ; n2 ; 2z2 S 2q ; n2 ; 2z2 ; 2r2

 2
2 2

 2
9 Vs = eσ νs+σWs
σ>0 τ ∼ Exp(λ), independent of V

9. GEOMETRIC BROWNIAN MOTION

1. Exponential stopping

 √ν 2+2λ/σ2−ν
 x


 z2 p 2 + 2=2 z
 dz, x ≤ z
1.0.5 Px (Vτ ∈ dz) = √
 z
   ν 2+2λ/σ2+ν
dz, z ≤ x

z2  2 + 2=2 x
 p

1 2 2 2
e−(ln(z/x)−νσ t) /2σ t dz


1.0.6 Px Vt ∈ dz = √
z 2t

 √ν 2+2λ/σ2−ν
x

1.1.2 Px sup Vs ≥ y =
0≤s≤τ y
x≤y

√ √

1 
ln(y=x)  t  1  y 2ν 
ln(y=x)  t 
1.1.4 Px sup Vs ≥ y = Erfc √ − √ + Erfc √ + √
0≤s≤t 2  2t 2 2 x  2t 2
x≤y

 x |ν|−ν
(1) Px sup Vs ≥ y =
0≤s<∞ y
x≤y
 ν  √ν 2+2λ/σ2

z zx
1.1.6 Px sup Vs ≥ y, Vτ ∈ dz = 2 p 2 dz
0≤s≤τ z  + 2= 2 x y2
x≤y
z≤y

 −1
 z −ν 2 σ 2 t/2 − ln2 (z/x)/2σ 2 t
y≤z

 x √2t e e dz,
1.1.8 Px sup Vs ≥ y, Vt ∈ dz = −1
x≤y 0≤s≤t  z √ e−ν 2 σ2 t/2 e− ln2 (y2 /zx)/2σ2 t dz, z ≤ y

x  2t

 √ν 2+2λ/σ2+ν
y

1.2.2 Px inf Vs ≤ y =
0≤s≤τ x
y≤x

√ √

1 
ln(x=y)  t  1  y 2ν 
ln(x=y)  t 
1.2.4 Px inf Vs ≤ y = Erfc √ + √ + Erfc √ − √
0≤s≤t 2  2t 2 2 x  2t 2
y≤x

© Springer Basel 2015 612

A.N. Borodin, P. Salminen, Handbook of Brownian Motion - Facts and Formulae,
Probability and Its Applications, DOI 10.1007/978-3-0348-8163-0_16
 1. EXPONENTIAL STOPPING 613
2
9 Vs = eσ νs+σWs
σ>0 τ ∼ Exp(λ), independent of V

 y |ν|+ν
(1) Px inf Vs ≤ y =
0≤s<∞ x
y≤x
 ν  2 √ν 2+2λ/σ2

z y 
1.2.6 Px inf Vs ≤ y, Vτ ∈ dz = 2 p 2 dz
0≤s≤τ z  + 2=2 x xz
y≤x
y≤z

 −1
 z −ν 2 σ 2 t/2 − ln2 (zx/y 2 )/2σ 2 t
dz, y ≤ z

 x √2t e e
1.2.8 Px inf Vs ≤ y, Vt ∈ dz =
0≤s≤t  z −1 −ν 2 σ2 t/2 − ln2 (x/z)/2σ2 t
y≤x √ e e dz, z≤y
x  2t


 √ν 2+2λ/σ2−ν
x

 1− √ , x≤r

+ r 2  2 + 2 r

1.3.1 Ex e−γ`(τ,r) = √
r
   ν 2+2λ/σ2+ν
 1−

√ , r≤x
+ r 2  2 + 2 x
 √ √
r  2  2 + 2  x  ν 2+2λ/σ2−ν

 yr√ 2 2+2 r dy, x ≤ r


1.3.2 Px `(τ, r) ∈ dy = e √ √
0<y  r 2  2 + 2  r  ν 2+2λ/σ2+ν

 √ dy, r ≤ x
eyr  2 2+2 x
  √ν 2+2λ/σ2−ν
x

 1− r , x≤r


(1) Px `(τ, r) = 0 = √
 1− r
   ν 2+2λ/σ2+ν
, r≤x

x

 ν h | || ln(r=x)| √ 
r e | ln(r=x)| | |  t

1.3.3 Ex e−γ`(t,r) = 1 − Erfc √ + √
2 x − | | 2 r  2t 2

√ 
e−| || ln(r=x)| | ln(r=x)| | | t

+ Erfc √ − √
+ | | 2 r  2t 2

√
2 e| ln(r=x)| = r | ln(r=x)| t
2  i
2
σ 2−γ 2 /σ 2 r 2 )t/2
+ 2 −  2 4 r2 e−(ν Erfc √ + √
 2t r 2

x
  |ν|−ν
 1−

+ | |r r
, x≤r
(1) Ex e−γ`(∞,r) =
r
  |ν|+ν
 1− , r≤x
+ | |r x
614 9. GEOMETRIC BROWNIAN MOTION
2
9 Vs = eσ νs+σWs
σ>0 τ ∼ Exp(λ), independent of V

r√2  r ν −ν 2 σ2 t/2−(σ2 ry+| ln(r/x)|)2 /2σ2 t

1.3.4 Px `(t, r) ∈ dy = √ e dy
t x
0<y

√ 
| | 2 r
 2
r  ry + | ln(r=x)| | | t
 ν h
−|ν|(σ 2 ry+| ln(r/x)|)
+ e Erfc √ − √
2 x  2t 2

 2 √ i
2  ry + | ln(r=x)| | | t
−e|ν|(σ ry+| ln(r/x)|)
Erfc √ + √ dy
 2t 2
√ 

1  r ν h |ν|| ln(r/x)| 
| ln(r=x)| | | t
(1) Px `(t, r) = 0 = 1 − e Erfc √ + √
2 x  2t 2
√
| ln(r=x)| | | t
 i
+e−|ν|| ln(r/x)| Erfc √ − √
 2t 2

x
  |ν|−ν 2
2

 |ν|σ r r
 e−|ν|σ ry dy, x ≤ r
(2) Px `(∞, r) ∈ dy =
 |ν|σ 2 r r |ν|+ν e−|ν|σ2 ry dy, r ≤ x
 
0<y

x

x
  |ν|−ν

 1− r , x≤r

(3) Px `(∞, r) = 0 =
 1 − r |ν|+ν , r ≤ x
  
x

r
σ 2 ry + | ln(z/r)| + | ln(r/x)|



1.3.8 Px `(t, r) ∈ dy, Vt ∈ dz = √
zt 2t
0<y
z
 ν 2 2 2 2 2
× e−ν σ t/2−(| ln(z/r)|+| ln(r/x)|+σ ry) /2σ t dydz
x

1 2 2 2
e−(ln(z/x)−νσ t) /2σ t dz


(1) Px `(t, r) = 0, Vt ∈ dz = √
z 2t

1  z ν −ν 2 σ2 t/2−(| ln(z/r)|+| ln(r/x)|)2 /2σ2 t

− √ e dz
z 2t x

 2 + 2 =2 +  x 2|ν|
 p  
1 −
 2 + 2 =2 −  r
, x≤r

∞
 Z  p


1.4.3 Ex exp −γ 1I[r,∞) Vs ds = √
(1) 0  2| | r
  ν 2+2γ/σ2+ν
, r≤x

2 2
 + 2 = −  x
p
ν<0
 1. EXPONENTIAL STOPPING 615
2
9 Vs = eσ νs+σWs
σ>0 τ ∼ Exp(λ), independent of V

Z t

1.4.4 Px 1I[r,∞) Vs ds ∈ dy = Bx(6) (t − y, y)1I(0,t) (y)dy
0

for Bx(6) (u, v) see 1.6.2

Z t

(1) Px 1I[r,∞) Vs ds = 0
x≤r 0
√ √
1 
ln(r=x)  t  1  r 2ν 
ln(r=x)  t 
=1− Erfc √ − √ − Erfc √ + √
2  2t 2 2 x  2t 2

Z t

(2) Px 1I[r,∞) Vs ds = t
r≤x 0
√ √
1 
ln(x=r)  t  1  r 2ν 
ln(x=r)  t 
=1− Erfc √ + √ − Erfc √ − √
2 2t 2 2 x 2t 2

Z ∞

(3) Px 1I[r,∞) Vs ds ∈ dy dy
ν<0 0

√  | | √ y 
|  | 2  x 2|ν| 2
σ 2 y/2 x
 2|ν|
x≤r = √ e−ν − ν 2 σ2 Erfc √
t r r 2
√ √
|  | 2  x |ν| 2
σ 2 y/2−ln2 (x/r)/2σ 2 y x
 2|ν| 
ln(x=r)  y 
r≤x = √ e−ν − ν 2 σ2 Erfc √ − √
t r r  2y 2

∞
x
Z
  |ν|−ν
(4) Px 1I[r,∞) Vs ds = 0 = 1 −
0
r
x≤r
 √ν 2+2γ/σ2−ν
2 x

, x≤r

 Z ∞
   2 + 2 =2 +  r
 p
1.5.3 Ex exp −γ 1I(0,r) Vs ds =
2 2
 1 − p + 2 = −  r
p  2ν
(1) 0 
, r≤x

ν>0  2 + 2 =2 +  x

Z t

1.5.4 Px 1I(0,r) Vs ds ∈ dy = Bx(6) (y, t − y)1I(0,t) (y)dy
0

for Bx(6) (u, v) see 1.6.2

Z t

(1) Px 1I(0,r) Vs ds = 0
r≤x 0
616 9. GEOMETRIC BROWNIAN MOTION
2
9 Vs = eσ νs+σWs
σ>0 τ ∼ Exp(λ), independent of V
√ √
1 
ln(x=r)  t  1  r 2ν 
ln(x=r)  t 
=1− Erfc √ + √ − Erfc √ − √
2  2t 2 2 x  2t 2

Z t

(2) Px 1I(0,r) Vs ds = t
x≤r 0
√ √
1 
ln(r=x)  t  1  r 2ν 
ln(r=x)  t 
=1− Erfc √ − √ − Erfc √ + √
2 2t 2 2 x 2t 2

Z ∞

(3) Px 1I(0,r) Vs ds ∈ dy dy
ν>0 0

√   √
 2 r ν −ν 2 σ2 y/2−ln2 (r/x)/2σ2 y r ln(r=x)  y 
 2ν 
x≤r = √ e − ν 2 σ2 Erfc √ + √
t x x  2y 2

√    √y 
 2 r 2ν −ν 2 σ2 y/2 r
 2ν
r≤x = √ e − ν 2 σ2 Erfc √
t x x 2

 Z τ



1.6.1 Ex exp − p1I(0,r) Vs + q1I[r,∞) Vs ds
0

√
(q − p)(  2 + 2( + q)=2 +  )(x=r)  2+2(+p)=2 −

p
x≤r = −
 + p ( + p)( + q)(  2 + 2( + q)=2 +  2 + 2( + p)=2 )
p p

√
 (q − p)(  2 + 2( + p)=2 −  )(r=x)  2+2(+q)=2 +
p
r≤x = +
+q ( + p)( + q)(  2 + 2( + q)=2 +  2 + 2( + p)=2 )
p p

Z τ

Z τ

1.6.2 Px 1I(0,r) Vs ds ∈ du, 1I[r,∞) Vs ds ∈ dv
0 0

=: λe−λ(u+v) Bx(6) (u, v)dudv

√ 
r 1  2 2 v   v
 ν    
x≤r = λe−λ(u+v) √ exp − + √ Erfc − √
x v 2 2 2

√

1 
ln2 (r=x)  2 2 u    r ν 
ln(r=x)  u 
× √ exp − − − √ Erfc √ + √ dudv
u 22 u 2 2 x  2u 2

 √ 
r 1  2 2 u   u
 ν   
r≤x = λe−λ(u+v) √ exp − − √ Erfc √
x u 2 2 2
 1. EXPONENTIAL STOPPING 617
2
9 Vs = eσ νs+σWs
σ>0 τ ∼ Exp(λ), independent of V

√

1 
ln2 (x=r)  2 2 v    r ν 
ln(x=r)  v 
× √ exp − − + √ Erfc √ − √ dudv
v 22 v 2 2 x  2v 2

Z τ

Z τ

(1) Px 1I(0,r) Vs ds ∈ du, 1I[r,∞) Vs ds = 0
x≤r 0 0

√ √

1 
ln(r=x)  t  1  r 2ν 
ln(r=x)  t 
= λe−λu 1 − Erfc √ − √ − Erfc √ + √ du
2  2t 2 2 x  2t 2

Z τ

Z τ

(2) Px 1I(0,r) Vs ds = 0, 1I[r,∞) Vs ds ∈ dv
r≤x 0 0

√ √

1 
ln(x=r)  t  1  r 2ν 
ln(x=r)  t 
= λe−λv 1 − Erfc √ + √ − Erfc √ − √ dv
2  2t 2 2 x  2t 2

Z t



1.6.4 Px p1I(0,r) Vs + q1I[r,∞) Vs ds ∈ dv,
0
1 
|qt − v | |pt − v |


= Bx(6) , 1I((q∧p)t,(q∨p)t) (v) dv for Bx(6) u, v see 1.6.2
|p − q | | p − q | |p − q |

Z t



(1) Px p1I(0,r) Vs + q1I[r,∞) Vs ds = pt
x≤r 0

√ √
1 
ln(r=x)  t  1  r 2ν 
ln(r=x)  t 
=1− Erfc √ − √ − Erfc √ + √
2  2t 2 2 x  2t 2

Z t



(2) Px p1I(0,r) Vs + q1I[r,∞) Vs ds = qt
r≤x 0

√ √
1 
ln(x=r)  t  1  r 2ν 
ln(x=r)  t 
= 1 − Erfc √ + √ − Erfc √ − √
2  2t 2 2 x  2t 2

n  Z τ


 o
1.6.5 Ex exp − p1I(0,r) Vs + q1I[r,∞) Vs ds ; Vτ ∈ dz
0

 z
 ν √
−| ln(z/x)| ν 2+2(λ+p)/σ 2
x≤r = e dz
z2  2 + 2( + p)=2 x
p
z≤r
 618 9. GEOMETRIC BROWNIAN MOTION
2
9 Vs = eσ νs+σWs
σ>0 τ ∼ Exp(λ), independent of V

 √ν 2+2(λ+p)/σ2

2z−1 −1 (z=x) z−1 −1 (z=x) zx
+ − dz
 2 2 + 2 + 2p +  2 2 + 2 + 2q  2 2 + 2 + 2p r2
p p p

√ √
2z−1 −2 (z=x) x
  ν 2+2(λ+p)/σ 2  r  ν 2+2(λ+q)/σ 2
x≤r =p 2 dz
 + 2( + p)=2 +  2 + 2( + q)=2 r z
p
r≤z
√ √
2z−1 −2 (z=x) r
  ν 2+2(λ+q)/σ 2 z  ν 2+2(λ+p)/σ 2
r≤x = p 2 dz
 + 2( + p)=2 +  2 + 2( + q)=2 x r
p
z≤r

 z
 ν √
−| ln(z/x)| ν 2+2(λ+q)/σ 2
r≤x = e dz
 2 + 2( + q)=2 x
p
r≤z
 2 √ν 2+2(λ+q)/σ2

2z−1 −1 (z=x) z−1 −2 (z=x) r
+ − dz
 2 2 + 2 + 2p +  2 2 + 2 + 2q  2 2 + 2 + 2q xz
p p p

2−1 y−1=2 (  2 + 2=2 +  )

Z τ  p
1.8.2 Px Vs ds ∈ dy = +1=2 −1=2 p 2
0 x 2 (2  + 2=2 + 1)

2

2x 
×e−x/σ y M1/2−ν,√ν 2+2λ/σ2 dy
2 y

 √
 Z ∞ 
2| |+1 | | x| | 2 2 x
1.8.3 Ex exp −γ Vs ds = 2| | K2|ν|
0  (2| |) 
ν<0

2+1 y−1=2 1 x 
Z t  
2 2 2
1.8.4 Px Vs ds ∈ dy = +1=2 +1=2 e−ν σ t/2−x/σ y mσ2 t/2 ν − , 2 dy
0 2 x 2  y
−1 < ν

Z ∞ 
22| | x2| | 2
(1) Px Vs ds ∈ dy = 4| | 2| |+1 e−2x/σ y dy
0  y (2| |)
ν<0

n  Z τ  o
1.8.5 Ex exp −γ Vs ds ; Vτ ∈ dz
0

 √  √ 
4z −1 2 2 x √ 2 2 z

  2 I2√ν 2+2λ/σ2 K2 ν 2+2λ/σ2 dz, x ≤ z
x   

= −1 √ √
 4z K √ 2 2 2 x √ 2 2 z
   
I2 ν 2+2λ/σ2 dz, z ≤ x

x 2 2 ν +2λ/σ2  
 1. EXPONENTIAL STOPPING 619
2
9 Vs = eσ νs+σWs
σ>0 τ ∼ Exp(λ), independent of V

 √ 
Z τ 
2z −1 2 4 xz
1.8.6 Px Vs ds ∈ dy, Vτ ∈ dz = 2  e−2(x+z)/σ y I2√ν 2+2λ/σ2 dydz
0
 yx 2 y

 √ √
z −1 2 2 x 2 2 z
n  Z t  o 2 2
1.8.7 Ex exp −γ Vs ds ; Vt ∈ dz =  e−ν σ t/2 kiσ2 t/8 , dz
0
2x  

 √ 
z −1 −ν 2 σ2 t/2−2(x+z)/σ2 y 4 xz
Z t 
1.8.8 Ex Vs ds ∈ dy, Vt ∈ dz = e iσ2 t/8 dydz
0
4yx  2 y

h2 2 (2 +  2 2 )
1.11.2 Px sup (y`(τ, y)) ≥ h = √
8 sh2 h 2 +  2 2 =2


0<y<∞

3 + √  √ √
M 2 2

2 2 2+ 2 2 ; 3; h 2 +   M 32 − √  2 2 ; 3; h 2 +  2 2 


2 2+ 
× 1 + √  √
2 2
+ M 1 − √  √
2 2

M 2 2 2+ 2 2 ; 1; h 2 +   2 2 2+ 2 2 ; 1; h 2 +  

 2 + 2=2 − 
p
1.12.1 Ex e−γ Ȟ(τ ) =
 + 2( + )=2 − 
2
p

 2 + 2=2 + 
p
−γ Ĥ(τ )
(1) Ex e =
 2 + 2( + )=2 + 
p

(p 2 + 2=2 −  )  e− 2 2 v=2 √ 

  v

1.12.2 Px Ȟ(τ ) ∈ dv = √ + Erfc − √ dv
ev 2v 2 2

(p 2 + 2=2 +  )  e− 2 2 v=2  √ 

  v
(1) Px Ĥ(τ ) ∈ dv = √ − Erfc √ dv
ev 2v 2 2

| | − 
1.12.3 Ex e−γ Ȟ(∞) =
 2 + 2 =2 − 
p
(1)
| | + 
(2) Ex e−γ Ĥ(∞) =
 2 + 2 =2 + 
p

1.12.4 Px Ȟ(t) ∈ dv 0≤v≤t

√  − 2 2 (t−v)=2  √
e−  v=2   v e   t − v
 2 2  
= √ + √ Erfc − √ − √ Erfc √ dv
v 2 2  (t − v ) 2 2
p

(1) Px Ĥ(t) ∈ dv 0≤v≤t
620 9. GEOMETRIC BROWNIAN MOTION
2
9 Vs = eσ νs+σWs
σ>0 τ ∼ Exp(λ), independent of V

 √  − 2 2 (t−v)=2 √
e−  v=2   v e   t − v
 2 2  
= √ − √ Erfc √ + √ Erfc − √ dv
v 2 2 (t − v) 2 2
p

 − 2 2 v=2 √ 
e   v


(2) Px Ȟ(∞) ∈ dv = σ(|ν| − ν) √ + Erfc − √ dv
2v 2 2
 − 2 2 v=2  √ 

e   v
(3) Px Ĥ(∞) ∈ dv = σ(|ν| + ν) √ − Erfc √ dv
2v 2 2

√
2 x=z  +2(+ )= −
 2 2

 √ dz x ≤ z
z(  2 2 + 2 + 2 +  2 2 + 2)

 p
Ex e−γ Ȟ(τ ) ; Vτ ∈ dz =

1.12.5
√
2 z=x  +2= +
 2 2

 √ dz z ≤ x
z(  2 + 2 + 2 +  2 2 + 2)
2
 p

√
2 x=z  +2= −
 2 2

 √ dz x ≤ z
z(  2 + 2 + 2 +  2 2 + 2)
2

 p
Ex e−γ Ĥ(τ ) ; Vτ ∈ dz =

(1)
√
2 z=x  +2(+ )= +
 2 2

 √ dz z ≤ x
z(  2 2 + 2 + 2 +  2 2 + 2)
 p

 √ν 2+2(λ+γ)/σ2 −ν
 2 + 2=2 −  x
p
−γ Ȟ(τ )

1.13.1 Ex e ; sup Vs ∈ dy = dy
0≤s≤τ y y
x<y


1.13.2 Px Ȟ(τ ) ∈ dv, sup Vs ∈ dy
0≤s≤τ
x<y

2y−1 ln(y=x) 
 2 2 v ln2 (y=x) 
=  2√ √ 2 2 exp −λv − − dvdy
x  2(   + 2 + )v3=2 2 22 v

 √ν 2+2γ/σ2 −ν
 −γ Ȟ(∞) | | −  x
1.13.3 Ex e ; sup Vs ∈ dy = dy
0≤s<∞ y y
(1)
x<y


1.13.4 Px Ȟ(t) ∈ dv, sup Vs ∈ dy 0≤v≤t
0≤s≤t
x<y
√
ln(y=x) (ln(y=x) − 2 v)2 e−  (t−v)=2   t − v
 2 2  
= √ exp − 2 − √ Erfc √ dvdy
y2 v3=2 2 v (t − v) 2 2
p

(| | −  ) ln(y=x) 
(ln(y=x) − 2 v)2 
(1) Px Ȟ(∞) ∈ dv, sup Vs ∈ dy = √ exp − dvdy
0≤s<∞ y 2v3=2 22 v
x<y
 1. EXPONENTIAL STOPPING 621
2
9 Vs = eσ νs+σWs
σ>0 τ ∼ Exp(λ), independent of V

Ex e−γ Ȟ(τ ) ; sup Vs ∈ dy, Vτ ∈ dz


1.13.5 x∨z <y
0≤s≤τ

√ √
2  z ν  x  ν 2+2(λ+γ)/σ2  z  ν 2+2λ/σ2
= dydz
yz2 x y y

1.13.6 Px Ȟ(τ ) ∈ dv, sup Vs ∈ dy, Vτ ∈ dz x∨z <y
0≤s≤τ

√
2 ln(y=x)  z ν  z  ν 2+2λ/σ 2 
ln2 (y=x)  2 2 v

= √ exp − − λv − dvdydz
yz3 2v3=2 x y 22 v 2

1.13.8 Px Ȟ(t) ∈ dv, sup Vs ∈ dy, Vt ∈ dz x∨z <y
0≤s≤t
v<t

ln(y=x) ln(y=z ) 
(ln(y=x) − 2 v)2 (ln(y=z ) + 2 (t − v))2 
= exp − − dvdydz
yz4 (v(t − v))3=2 2 2 v 22 (t − v)

 √ν 2+2(λ+γ)/σ2+ν
 2 + 2=2 +  y
p
−γ Ĥ(τ )

1.14.1 Ex e ; inf Vs ∈ dy = dy
0≤s≤τ y x
y<x

1.14.2 Px Ĥ(τ ) ∈ dv, inf Vs ∈ dy
0≤s≤τ
y<x
2y−1 ln(x=y) 
 2 2 v ln2 (x=y) 
=  2√ √ 2 2 exp −λv − − dvdy
x  2(   + 2 − )v3=2 2 22 v

 √ν 2+2γ/σ2 +ν
 −γ Ĥ(∞) | | +  y
1.14.3 Ex e ; inf Vs ∈ dy = dy
0≤s<∞ y x
(1)
y<x

1.14.4 Px Ĥ(t) ∈ dv, inf Vs ∈ dy 0≤v≤t
0≤s≤t
y<x
√
ln(x=y) (ln(x=y)+ 2 v)2 e−  (t−v)=2   t − v
 2 2  
= 2√ exp − + √ Erfc − √ dvdy
y v3=2 22 v (t − v) 2 2
p

(| | +  ) ln(x=y) 
(ln(x=y) + 2 v)2 
(1) Px Ĥ(∞) ∈ dv, inf Vs ∈ dy = √ exp − dvdy
0≤s<∞ y 2v3=2 22 v
y<x
622 9. GEOMETRIC BROWNIAN MOTION
2
9 Vs = eσ νs+σWs
σ>0 τ ∼ Exp(λ), independent of V

Ex e−γ Ĥ(τ ) ; inf Vs ∈ dy, Vτ ∈ dz


1.14.5 y <x∧z
0≤s≤τ

√ √
2  z ν  y  ν 2+2(λ+γ)/σ2  y  ν 2+2λ/σ2
= dydz
yz2 x x z

1.14.6 Px Ĥ(τ ) ∈ dv, inf Vs ∈ dy, Vτ ∈ dz y <x∧z
0≤s≤τ

√
2 ln(x=y)  z ν  y  ν 2+2λ/σ 2 
ln2 (x=y)  2 2 v

= √ exp − − λv − dvdydz
yz3 2v3=2 x z 22 v 2

1.14.8 Px Ĥ(t) ∈ dv, inf Vs ∈ dy, Vt ∈ dz y <x∧z
0≤s≤t
v<t
ln(x=y) ln(z=y) 
(ln(x=y) + 2 v)2 (ln(z=y) − 2 (t − v))2 
= exp − − dvdydz
4
yz (v(t − v)) 3= 2 22 v 22 (t − v)


1.15.2 Px a < inf Vs ; sup Vs < b = 1
0≤s≤τ 0≤s≤τ

√ √ √ √
a
 b
 2+2=2 − x
 2+2=2 + b
 x
 2+2=2 − a
 2+2=2
   
x x b x a x
− √ √
b
 2+2=2 − a
 2+2=2
a b

e− 2 2 t=2
1.15.4 Px a < inf Vs ; sup Vs < b = √
0≤s≤t 0≤s≤t  2t

∞ Z b
X z −1 −(ln(z/x)+2k ln(b/a))2 /2σ2 t 2 2 2

×  e − e−(ln(zx/a )+2k ln(b/a)) /2σ t dz
a
x
k=−∞

1.15.6 Px a < inf Vs ; sup Vs < b, Vτ ∈ dz a<x∧z <x∨z <b
0≤s≤τ 0≤s≤τ

2z −1 ch (ln(b=a) − | ln(z=x)|)  2 + 2=2 − ch ln(ba=zx)  2 + 2=2

p
p


= √ √ √ dz
x   2 2 + 2 (b=a)  2+2=2 − (a=b)  2+2=2


1.15.8 Px a < inf Vs ; sup Vs < b , Vt ∈ dz dz a<x∧z ≤x∨z <b
0≤s≤t 0≤s≤t
∞
z −1 e−  t=2 X −(ln(z/x)+2k ln(b/a))2 /2σ2 t
2 2
2 2 2

=  √ e − e−(ln(zx/a )+2k ln(b/a)) /2σ t
x  2t
k=−∞
 1. EXPONENTIAL STOPPING 623
2
9 Vs = eσ νs+σWs σ>0 τ ∼ Exp(λ), independent of V
Z τ
ds

1.19.2 Px ∈ dy
0
Vs
0<x
2(2 y=2)− (  2 + 2=2 −  ) −1/σ2 xy 2 
p 
= −1=2 √ e M √ 2 dy
x 2y (2  2 + 2=2 + 1) 1/2+ν, ν +2λ/σ 2  2 xy
p

 √ 
 Z ∞
ds

2+1  x− 2 2
1.19.3 Ex exp −γ = 2 K2ν √
0
Vs  (2 )  x
0<ν

(2 y=2)− 1 1 
t
ds
Z  2 2 2

1.19.4 Px ∈ dy = √ −1=2 e−ν σ t/2−1/σ xy mσ2 t/2 −ν − , 2 dy
0
Vs 2yx 2  xy
ν<1

Z ∞
ds

22 x−2 2
(1) Px ∈ dy = 4 2+1 e−2/σ xy dy
0
Vs  y (2 )
0<ν

τ
ds
n  Z  o
1.19.5 Ex exp −γ ; Vτ ∈ dz
0
Vs
0<x
 √   √
4z −1 2 2 2 2
 
  2 K2√ν 2+2λ/σ2 √ I2√ν 2+2λ/σ2 √ dz, x≤z
x   x  z

=  1  √   √
 4z I √ 2 2 2 2 2
− 
√ K2√ν 2+2λ/σ2 √ dz, z≤x

x 2 2 ν +2λ/σ2  x  z

Z τ
ds

2z −1 2

4 
1.19.6 Px ∈ dy, Vτ ∈ dz = 2  e−2(x+z)/σ xzy I2√ν 2+2λ/σ2 2 √ dydz
0
Vs  yx  y xz
0<x

 √ √
z −1 2 2 2 2 
t
ds
n  Z  o 2 2
1.19.7 Ex exp −γ ; Vt ∈ dz =  e−ν σ t/2 kiσ2 t/8 √ , √ dz
0
Vs 2x  x  z
0<x

z −1 −ν 2 σ2 t/2−2(x+z)/σ2 xzy 4 

t
ds
Z  
1.19.8 Px ∈ dy, Vt ∈ dz = e i √ dydz
Vs 4yx 2 y xz
σ 2 t/8

0<x 0

(22 2 y)=2 −1=2 (  2 + 2=2 =2| | + =2 )

Z τ  p
1.20.2 Px Vs2β ds ∈ dy =
x+ (  2 + 2=2 =| | + 1)
p
0
 2 
2β 2 2 x
×e−x /4σ β y M1/2−ν/2β,√ν 2+2λ/σ2 /2|β| dy
22 2 y
 2 ∞
2( =2| |)| |=| | x− x
Z   
1.20.3 Ex exp − Vs2β ds = K|ν|/|β|
 2 0 (| |=| |) | |
<0
624 9. GEOMETRIC BROWNIAN MOTION
2
9 Vs = eσ νs+σWs
σ>0 τ ∼ Exp(λ), independent of V

Z t 
1.20.4 Px Vs2β ds ∈ dy
0

 (22 2 )=2 +1=2 2

σ 2 t/2−x2β /4σ 2 β 2 y

 1 x2 
>−2 = e−ν m2σ2 β 2 t − , dy
x+ y1=2−=2 2 2 42 2 y

Z ∞ 
(22 2 y)−| |=| | −2ν −x2β /2σ2 β 2 y
(1) Px Vs2β ds ∈ dy = x e dy
 0
y (| |=| |)
<0

n  2Z τ  o
1.20.5 Ex exp − Vs2β ds ; Vτ ∈ dz
2 0

2z −1 x z
    
 x 2 | | I√ν 2+2λ/σ2 /|β| | | K√ν 2+2λ/σ2 /|β| | | dz,
 (z − x)β ≥ 0
= −1
 2z x z
   

K √ 2 I√ 2 dz, (z − x)β ≤ 0
x  2 | | ν +2λ/σ /|β|  | |
2 ν +2λ/σ /|β|  | |
2

Z τ 
1.20.6 Px Vs2β ds ∈ dy, Vτ ∈ dz
0

z −1 x z
2β 2β 2 2
 
= 2  e−(x +z )/2σ β y I√ν 2+2λ/σ2 /|β| 2 2 dydz
 yx | |  y

n  2Z t
z −1 | | −ν 2 σ2 t/2 x z
 o  
1.20.7 Ex exp − Vs2β ds ; Vt ∈ dz = e ki , dz
2 0 x | | | |
σ 2 β 2 t/2

Z t 
1.20.8 Px Vs2β ds ∈ dy, Vt ∈ dz
0

z −1 | | −ν 2 σ2 t/2−(x2β+z2β )/2σ2 β 2 y x z

 
= e i dydz
2yx 2 2 y
σ 2 β 2 t/2

n  2Z τ τ
p
 Z o
2β
1.21.1 Ex exp − Vs ds ; Vsβ ds ∈ dy
2 0 0
(1)
 (  2 + 2=2 =| | + = ) p ch(yp| |=2) 1/2−ν/β
p  
= + =2 p 2
x (2  + 2=2 =| | + 1) | | sh(yp| |=2)

px ch(yp| |) 2px
   
× exp − M 1 ν √ν 2+2λ/σ2 dy
| | sh(yp| |) 2−β , |β|
| | sh(yp| |)
 1. EXPONENTIAL STOPPING 625
2
9 Vs = eσ νs+σWs
σ>0 τ ∼ Exp(λ), independent of V

 2Z t
2 2 p ch(yp| |=2) 1/2−ν/β
 Z t
p
n o  
2β
1.21.3 Ex exp − Vs ds ; Vsβ ds ∈ dy =
2 0 0
2 | | sh(yp| |=2)
(1)

 2 2
   t px ch(yp| |)  1 px
  
>−2 ×x−(ν+β/2) exp − − mσ2 β 2 t/2 − , dy
2 | | sh(yp| |) 2 | | sh(yp| |)

p2
 Z ∞

(2) Ex exp − Vs2β + qVsβ ds
0
2

| |
 + pq| | + 12  2p |ν|/|β|−1/2 −(ν+β/2) 2px 


| |
<0 = x W−q/pσ|β|,|ν|/|β|
(2| |=| |) | | | |

 2Z ∞ ∞
p
n  Z o
(3) Ex exp − Vs2β ds ; Vsβ ds ∈ dy
2 0 0

 2 2 x−2 p px ch(yp| |=2)

 1+2|ν|/|β|  
<0 = exp − dy
2 (2| |=| |) | | sh(yp| |=2) | | sh(yp| |=2)

Z ∞ Z ∞ 
1.21.4 Px Vs2β ds ∈ dg, Vsβ ds ∈ dy
(1) 0 0

 (| |)1−2| |=| | x−2 2| | y| | x

<0 = eeg 1 + , , , 0 dgdy
2 (2| |=| |) | | 2 | |
√
o z− =2  2+2=2 1 q
+ +
 
p2
τ
2 p
n  Z 
| | | |
V 2β + qVsβ ds ; Vτ ∈ dz =


1.21.5 Ex exp −
2 s pzx =2 ( | 2 |  2 + 2=2 + 1)
p
0

 z 2pxβ

2pz β


√ √ (z − x)β ≥ 0
 x M− q , ν 2+2λ/σ2
 σ|β| W q
− pσ|β| ,
ν 2+2λ/σ 2 σ|β| dz,
pσ|β| |β| |β|
× z
2pxβ 2pz β



  W q √ν 2+2λ/σ2
 M √ dz, (z − x)β ≤ 0
x − pσ|β| , |β|
σ|β| q
− pσ|β| ,
ν 2+2λ/σ 2
|β|
σ|β|

 2Z
pz −1
τ τ
p
n  Z o
(1) Ex exp − Vs2β ds ; Vsβ ds ∈ dy, Vτ ∈ dz = 
2 0 0
x  sh(yp| |=2)


p(x + z ) ch(yp| |=2)
 
2p(xz ) =2 
× exp − I2√ν 2+2λ/σ2 /|β| dydz
| | sh(yp| |=2) | | sh(yp| |=2)
626 9. GEOMETRIC BROWNIAN MOTION
2
9 Vs = eσ νs+σWs
σ>0 τ ∼ Exp(λ), independent of V

Z τ Z τ 
1.21.6 Px Vs2β ds ∈ dg, Vsβ ds ∈ dy, Vτ ∈ dz
0 0

x + z (xz ) =2
 p 2
 z ν
 
2  + 2=2 y| | 
= isg , , 0, , dgdydz
z x | | 2 | | | |

 2Z t
 2 pz −1
 Z t
p
n o
2β
1.21.7 Ex exp − Vs ds ; Vsβ ds ∈ dy, Vt ∈ dz = 
2 0 0
8x sh(yp| |=2)
(1)

2
σ 2 t/2

p(x + z ) ch(yp| |=2)
 
2p(xz ) =2 
×e−ν exp − iσ2 β 2 t/8 dydz
| | sh(yp| |=2) | | sh(yp| |=2)

Z t Z t 
1.21.8 Px Vs2β ds ∈ dg, Vsβ ds ∈ dy, Vt ∈ dz
0 0

 2 z −1 −ν 2 σ2 t/2
 2 2
 t y| | x + z 2(xz ) =2

= e ei , , , dgdydz
8x g
8 2 | | | |

n  Z τ


 1 o
1.27.5 Ex exp − p1I(0,r) Vs + q1I[r,∞) Vs ds ; `(τ, r) ∈ dy, Vτ ∈ dz
0
2
√ √ √ν 2+2(λ+p)/σ2
2r  z ν −yσr( ν 2 σ 2+2λ+2p+ ν 2 σ 2+2λ+2q)

zx
x≤r = e dydz
z x r2
z≤r
√ √ √
2r  z ν  2+2(+q)=2
ν 2 σ 2+2λ+2p+ ν 2 σ 2+2λ+2q) (r=z )
x≤r = e−yσr( √ dydz
z x (x=r)  2+2(+p)=2
r≤z
√ √ √
2r  z ν  2+2(+q)=2
−yσr( ν 2 σ 2+2λ+2p+ ν 2 σ 2+2λ+2q) (r=x)
r≤x = e √ dydz
z x (z=r)  2+2(+p)=2
z≤r
√ √  2 √ν 2+2(λ+q)/σ2
2r  z ν −yσr( ν 2 σ 2+2λ+2p+ ν 2 σ 2+2λ+2q) r
r≤x = e dydz
z x xz
r≤z

n  Z τ


 o
(1) Ex exp − p1I(0,r) Vs + q1I[r,∞) Vs ds ; `(τ, r) = 0, Vτ ∈ dz
0
√  √ν 2+2(λ+p)/σ2 
z zx
 ν 
−| ln(z/x)| ν 2+2(λ+p)/σ 2
x≤r = p 2 2 e − 2 dz
z   + 2 + 2p x r
z≤r
√  2 √ν 2+2(λ+q)/σ2 
 z r
 ν 
−| ln(z/x)| ν 2+2(λ+q)/σ 2
r≤x = p 2 2 e − dz
z   + 2 + 2q x zx
r≤z
 1. EXPONENTIAL STOPPING 627
2
9 Vs = eσ νs+σWs
σ>0 τ ∼ Exp(λ), independent of V

 
1.31.2 Px sup (`(s, u) − α−1 `(s, r)) > h
0≤s≤τ
0<α

 √ν 2+2λ/σ2+ν
u  h2 r  2 + 2=2
p
exp − 
 
r≤u = √
x 2 − 2(r=u)2  2+2=2
 √ 2 √ √
x 2 ν +2λ/σ2  (x=r)2  +2= − (x=u)2  +2=
2 2 2 2

− √ , x≤r

 r 2 − 2(r=u)2  2+2=2




√
×  1 − (x=u)2  2+2=2


1 − √ , r≤x≤u
− 2(r=u)2  +2=

2 2 2





1, u≤x

 √ν 2+2λ/σ2−ν
x  h2 r  2 + 2=2
p
exp − 
 
u≤r = √
u 2 − 2(u=r)2  +2=
2 2

1, x≤u

 √
2  2+2= 2

 1 −  1 − (u=x) √



u≤x≤r

,
× 2 − 2(u=r)2  2+2=2
 2√ν 2+2λ/σ2  (r=x)2√ 2+2=2 − (u=x)2√ 2+2=2


 r




 − √ , r≤x
x 2 − 2(u=r)2  2+2=2

x
  ν+|ν| sign(u−r)
(2) Px sup (`(s, u) − α−1 `(s, r)) > h
0≤s<∞ u

x 0 (x=r)2| | − (x=u)2| |
  2|ν|

− , x≤r
2 − 2(r=u)2| |

 r


0 h2 r| |
 
r≤u = exp − 0 1 − (x=u)2| |


2 − 2(r=u)2| |  1− , r≤x≤u


 2 − 2(r=u)2| |
1, u≤x



 1, x≤u
 1 ( u=x)2| |


0

−
 h2 r| |

= exp − 0
1− , u≤x≤r
  
u≤r
2 − 2(u=r)2| |  2 − 2(u=r)2| |
 r
2|ν| 0 (r=x)2| | − (u=x)2| |
  


− , r≤x

x 2 − 2(u=r)2| |

q √ 2 2
Υλ := (1 − αru−1 )2 + 4αru−1 (1 − ((r ∧ u)/(r ∨ u))2 ν +2λ/σ ) + 1 − α
628 9. GEOMETRIC BROWNIAN MOTION
2
9 Vs = eσ νs+σWs
σ>0 Hz = min{s : Vs = z}

2. Stopping at first hitting time

  √ 2
x ν +2α/σ 2−ν
, x≤z

z


2.0.1 Ex e−αHz = Ex {e−αHz ; Hz < ∞} =  √
 z
 ν 2+2α/σ 2+ν
, z≤x

x

  t ln2 (z=x)
 2 2
| ln(z=x)| z ν

  
2.0.2 Px Hz ∈ dt = √ 3=2 exp − − dt
 2t x 2 2 2 t

x
  |ν|−ν

 1− z , x≤z

(1) Px H z = ∞ =
 1 − z |ν|+ν , z ≤ x
  
x

 x
  |ν|−ν

 z , x≤z
2.1.2 Px sup Vs < y, Hz < ∞ =  ν | | | |
z<y 0≤s≤Hz  z (y=x) − (x=y) ,

z≤x≤y
x (y=z ) − (z=y)  |
| | |

z

 ν+|ν|
(1) Px sup Vs = ∞ = 1 −
0≤s≤Hz x
z≤x
  √ν 2+2α/σ2−ν
x
 , x≤z
 z


z≤y

 −αHz
2.1.4 Ex e ; sup Vs < y = √ √
 ν y  +2 = − x  +2 =
2 2 2 2

 z



0≤s≤Hz x √ y √

, z≤x

 x

y
 2+2 =2 − z
 2+2 =2
z y x≤y

2 2 ln(y/x) ln(y/z)

sup Vs < y, Hz ∈ dt = (z/x)ν e−ν σ t/2 sst


(1) Px σ , σ dt
0≤s≤Hz
z≤x
x≤y
| | | |
 z (x=y) − (y=x)
  ν

 x (z=y)| | − (y=z )| | , y ≤ x ≤ z
2.2.2 Px inf Vs > y, Hz < ∞ =  |ν|+ν
0≤s≤Hz  z

, y≤z≤x
x

x

 |ν|−ν
(1) Px inf Vs = 0 = 1 −
0≤s≤Hz z
x≤z
 2. STOPPING AT FIRST HITTING TIME 629
2
9 Vs = eσ νs+σWs
σ>0 Hz = min{s : Vs = z}

   x
√ 2+2 =2 √

y
 2+2 =2
z ν y − x √
y≤x

 √ ,
x z  2+2 = 2 y
 2+2 =2


x≤z


y − z

Ex e−αHz ; inf Vs > y =

2.2.4 √
0≤s≤Hz
z

   ν 2+2α/σ2+ν
, y≤z

 x

z≤x
2 2 ln(x/y) ln(z/y)

Vs > y, Hz ∈ dt = (z/x)ν e−ν σ t/2 sst


(1) Px inf σ , σ dt
0≤s≤Hz
z≤x
y≤x

Ex e−γ`(Hz ,r) ; Hz < ∞


2.3.1

 1, r ≤ z ≤ x, x ≤ z ≤ r
 | |r2 + 1 − e−2| || ln(x=r)|



z
 ν 

, z ∧ r ≤ x ≤ z ∨ r
= e−|ν|| ln(z/x)| | |r2 + 1 − e−2| || ln(z=r)|
x
| |r 2




, z ≤ r ≤ x, x ≤ r ≤ z
| |r 2 + 1 − e−2| || ln(z=r)|



Ex e−γ`(Hz ,r) ; Hz = ∞

(1) for ν ln(z/r) < 0

0, r ≤ z ≤ x, x ≤ z ≤ r

2 −2| || ln(x=z )|

 r 1 e


| | −


 | |r2 + 1 − e−2| || ln(z=r)|
,

 z∧r ≤x≤z∨r
=
 1 − e−2|ν|| ln(x/r)|

| |r 2 e−2| || ln(x=r)| − e−2| || ln(x=z)|




+ , z ≤ r ≤ x, x ≤ r ≤ z

| |r 2 + 1 − e−2| || ln(z=r)|


 z ν | |yr 2
 
2.3.2 Px `(Hz , r) ∈ dy, Hz < ∞ = exp −
x 1 − e−2| || ln(z=r)|

0, x ≤ z ≤ r, r ≤ z ≤ x

 | |r2 sh(| || ln(z=x)|)



dy, r ∧ z ≤ x ≤ r ∨ z
× 2 sh2 (| || ln(z=r)|)
 | |r2 e−| || ln(r=x)| dy,



z ≤ r ≤ x, x ≤ r ≤ z
2 sh(| || ln(z=r)|)

(1) Px `(Hz , r) ∈ dy, Hz = ∞ for ν ln(z/r) < 0
630 9. GEOMETRIC BROWNIAN MOTION
2
9 Vs = eσ νs+σWs
σ>0 Hz = min{s : Vs = z}

0, x ≤ z ≤ r, r ≤ z ≤ x

2 2  ln( x=z )

 | |r 1 − e | |yr 2

 − | || |

 

exp − dy, r∧z ≤x≤r∨z
= 1 − e−2| || ln(z=r)| 1 − e−2| || ln(z=r)|
| |r 2 e−2| || ln(x=r)| − e−2| || ln(x=z)| | |yr 2

 

exp − dy, x ≤ r ≤ z

1−e 2  ln(z=r ) 1−e 2  ln(z=r )

 − | || | − | || |
z≤r≤x

z
 ν √ 2 2
2.3.3 Ex e−αHz −γ`(Hz ,r) = e−| ln(z/x)| ν +2α/σ
x
x ≤ z ≤ r, r ≤ z ≤ x

1,
√

 √ 2 2

−2| ln(x=r )|  2+2 = 2
r   + 2 + 1 e


−


√
, r ∧ z ≤ x ≤ r ∨ z

√
× r  2 2 + 2 + 1 − e−2| ln(z=r)|  2+2 =2
√
r  2 2 + 2



√
, z ≤ r ≤ x, x ≤ r ≤ z

 √
r  2 2 + 2 + 1 − e−2| ln(z=r)|  2+2 =2


 Z Hz



2.6.1 Ex exp − p1I(0,r) Vs + q1I[r,∞) Vs ds
0
x
 Υp−ν
x≤z =
z
z≤r
 ν  (r=x)p + (x=r)p
+  (r=x)p − (x=r)p

z p q
z≤x =
x p (r=z )p + (z=r)p + q (r=z )p − (z=r)p



x≤r
z
 ν
2p (r=x)q
z≤r =
x p (r=z )p + (z=r)p + q (r=z )p − (z=r)p



r≤x
z
 ν
2q (x=r)p
x≤r =
x q (z=r)q + (r=z )q + p (z=r)q − (r=z )q



r≤z
 ν  (x=r)q + (r=x)q
+  (x=r)q − (r=x)q

z q p
r≤x =
x q (z=r)q + (r=z )q + p (z=r)q − (r=z )q



x≤z
z
 Υq+ν
r≤z =
x
z≤x
√
2+2 = 2 (2 2 x= )
  ν I √
z 2 
 √ , x≤z
2  +2 =2 (2 2 z= )
 x I√ 2
Z Hz 
  
2.8.3 Ex exp −αHz − γ Vs ds =   √
0  z ν K2√ 2+2 =2 (2 2 x=)
√ , z≤x

x K2√ 2+2 =2 (2 2 z=)



p
Υs := ν 2 + 2s/σ 2
 2. STOPPING AT FIRST HITTING TIME 631
2
9 Vs = eσ νs+σWs
σ>0 Hz = min{s : Vs = z}

Ex e−γ Ȟ(Hz ) ; sup Vs ∈ dy


2.13.1 z<x<y
0≤s≤Hz

√ √
(x=z )  2+2 =2 − (z=x)  2+2 =2
| |(z=x)


= √ √
dy
y (y=z )| | − (z=y)| | (y=z )  2+2 =2 − (z=y)  2+2 =2

ln(x=z) ; ln(y=z)

z −ν σ u/2 | | ssu
 ν 2 2
2.13.2 Px Ȟ(Hz ) ∈ du, sup Vs ∈ dy = e

 
dudy
0≤s≤Hz x y (y=z )| | − (z=y)| |
z<x
x<y

Ex e−αHz−γ Ȟ(Hz ) ; sup Vs ∈ dy


2.13.4 z<x<y
0≤s≤Hz

√ √
y−1  2 + 2 =2 (z=x) (x=z )  2+2( + )=2 − (z=x)  2+2( + )=2
p

= √ √ √ √
dy
(y=z )  2+2 =2 − (z=y)  2+2 =2 (y=z )  2+2( + )=2 − (z=y)  2+2( + )=2

(1) Px Ȟ(Hz ) ∈ du, Hz ∈ dt, sup Vs ∈ dy
0≤s≤Hz
u<t

z
 ν
1 −ν 2 σ2 t/2 ln(x/z) ln(y/z)

st−u ln(y/z)


= e ssu , σ dudtdy
x y σ σ

Ex e−γ Ĥ(Hz ) ;

2.14.1 inf Vs ∈ dy y<x<z
0≤s≤Hz
x<z
√ √
(z=x)  2+2 =2 − (x=z )  2+2 =2
| |(z=x)


= √ √
dy
y (z=y)| | − (y=z )| | (z=y)  2+2 =2 − (y=z )  2+2 =2

ln(z=x) ; ln(z=y)

z −ν σ u/2 | | ssu
 ν 2 2
2.14.2 Px Ĥ(Hz ) ∈ du, inf Vs ∈ dy = e

 
dudy
0≤s≤Hz x y (z=y)| | − (y=z )| |
x<z
y<x

Ex e−αHz−γ Ĥ(Hz ) ;

2.14.4 inf Vs ∈ dy y<x<z
0≤s≤Hz

√ √
y−1  2 + 2 =2 (z=x) (z=x)  2+2( + )=2 − (x=z )  2+2( + )=2
p

= √ √ √ √
dy
(z=y)  2+2 =2 − (y=z )  2+2 =2 (z=y)  2+2( + )=2 − (y=z )  2+2( + )=2


632 9. GEOMETRIC BROWNIAN MOTION
2
9 Vs = eσ νs+σWs
σ>0 Hz = min{s : Vs = z}


(1) Px Ĥ(Hz ) ∈ du, Hz ∈ dt, inf Vs ∈ dy
0≤s≤Hz
u<t
z
 ν
1 −ν 2 σ2 t/2 ln(z/x) ln(z/y)

st−u ln(z/y)


= e ssu , σ dudtdy
x y σ σ

  z ν (x=a)| | − (a=x)| |



, a ≤ x ≤ z

 x (z=a)| | − (a=z )| |

2.15.2 Px a < inf Vs , sup Vs < b =  ν
 z (b=x)| | − (x=b)| |


0≤s≤Hz 0≤s≤Hz

, z ≤ x ≤ b

x (b=z )| | − (z=b)| |

√ √
2  2+2 =2 (2 2 = x)
  ν K √
z
√ √ , x≤z

2  +2 =2 (2 2 = z )
 x K√ 2

Hz
ds
 Z  
2.19.3 Ex exp −αHz − γ =   √ √ √
0
Vs  z ν I2  2+2 =2 (2 2 = x)
√ √ , z≤x

x I2√ 2+2 =2 (2 2 = z )



2 Hz
 Z 
2.20.3 Ex exp −αHz − Vs2β ds
2 0

 2+2 =2 =| | ( x = | |)
   ν I√
z
 , (z − x)β ≥ 0
 +2 =2 =| | ( z = | |)
 x I 2 √


=   √
 z ν K  2+2 =2 =| | ( x =| |)
, (z − x)β ≤ 0

x K√ 2+2 =2 =| | ( z =| |)



p2
 Z Hz

2.21.3 Ex exp −αHz − Vs2β + qVsβ ds
0
2
   ν z =2 M √ (2px =| |)
z −q=p | |;  2+2 =2 =| |
 x x =2 M
 , (z − x)β ≥ 0
√ (2pz =| |)

|;  2+2 = 2 =| |

−q=p |
= = 2
 
z ν
 z W −q=p |
√
|;  2+2 = 2 =| |
(2px =| |)
, (z − x)β ≤ 0

x x =2 W−q=p| (2pz =| |)

√

|;  2+2 = 2 =| |
 3. STOPPING AT FIRST EXIT TIME 633
2
9 Vs = eσ νs+σWs
σ>0 H = Ha,b = min{s : Vs ∈
/ (a, b)}

3. Stopping at first exit time

 a ν (b=x)| | − (x=b)| |
3.0.4 Px V H = a =
x (b=a)| | − (a=b)| |
(a)

 b ν (x=a)| | − (x=y)| |
3.0.4 Px V H = b =
x (b=a)| | − (a=b)| |
(b)

√ √
a (b=x)  2+2 =2 − (x=b)  2+2 =2
 ν
−αH

3.0.5 Ex e ; VH =a = √ √
x (b=a)  2+2 =2 − (a=b)  2+2 =2
(a)

√ √
b (x=a)  2+2 =2 − (a=x)  2+2 =2
 ν
−αH

3.0.5 Ex e ; VH =b = √ √
x (b=a)  2+2 =2 − (a=b)  2+2 =2
(b)

2 2 ln(b/x) ln(b/a)

Px H ∈ dt, VH = a = (a/x)ν e−ν σ t/2 sst


3.0.6 σ , σ dt
(a)
2 2 ln(x/a) ln(b/a)

Px H ∈ dt, VH = b = (b/x)ν e−ν σ t/2 sst


3.0.6 σ , σ dt
(b)

3.1.6 Px sup Vs ≥ y, VH = a
0≤s≤H

 ν (b=y)| | − (y=b)| |
(x=a)| | − (x=y)| |

a
=
, a≤x≤y≤b
x (b=a)| | − (a=b)| | (y=a)| | − (a=y)| |

Ex e−αH ; sup Vs ≥ y, VH = a

3.1.8 a≤x≤y≤b
0≤s≤H
(1)

 ν (b=y)√ 2+2 =2 − (y=b)√ 2+2 =2
(x=a)√ 2+2 =2 − (a=x)√ 2+2 =2

a
= √ √ √ √
x (b=a)  2+2 =2 − (a=b)  2+2 =2 (y=a)  2+2 =2 − (a=y)  2+2 =2

3.2.6 Px inf Vs ≤ y, VH = b
0≤s≤H

 ν (b=x)| | − (x=b)| |
(y=a)| | − (a=y)| |

b
=
, a≤y≤x≤b
x (b=y)| | − (y=b)| | (b=a)| | − (a=b)| |


634 9. GEOMETRIC BROWNIAN MOTION
2
9 Vs = eσ νs+σWs
σ>0 H = Ha,b = min{s : Vs ∈
/ (a, b)}

Ex e−αH ; inf

3.2.8 Vs ≤ y, VH = b a≤y≤x≤b
0≤s≤H
(1)

 ν (b=x)√ 2+2 =2 − (x=b)√ 2+2 =2
(y=a)√ 2+2 =2 − (a=y)√ 2+2 =2

b
= √ √ √ √
x (b=y)  2+2 =2 − (y=b)  2+2 =2 (b=a)  2+2 =2 − (a=b)  2+2 =2

(b=a)| | − (a=b)| |
| |


3.3.1 Er e−γ`(H,r) =
=:L
| | (b=a)| | − (a=b)| | + r2 (b=r)| | − (r=b)| | (r=a)| | − (a=r)| |

  b ν (x=r)| | − (r=x)| |  r ν (b=x)| | − (x=b)| |

+ L, r ≤ x
 x (b=r)| | − (r=b)| | x (b=r)| | − (r=b)| |


x≤b
Ex e−γ`(H,r) =  ν | | − (x=r )| | | | − (x=y )| |
a (r=x ) r
 ν
(x=a )
L, a ≤ x

+
x (r=a)| | − (a=r)| | x (r=a)| | − (a=r)| |


x≤r

 r ν | |yr 2 (b=a)| | − (a=b)| |



3.3.2 Px `(H, r) ∈ dy = exp −
x (b=r)| | − (r=b)| | (r=a)| | − (a=r)| |

| |r 2 (b=x)| | − (x=b)| | (b=a)| | − (a=b)| |




dy, r ≤ x ≤ b
(b=r)| | − (r=b)| | 2 (r=a)| | − (a=r)| |





×
| |r 2 (x=a)| | − (x=y )| | (b=a)| | − (a=b)| |



dy, a ≤ x ≤ r


2

(b=r)| | − (r=b)| | (r=a)| | − (a=r)| |


b (x=r)| | − (r=x)| |
  ν
 x (b=r)| | − (r=b)| | , r ≤ x ≤ b



(1) Px `(H, r) = 0 =  ν | | | |
 a (r=x) − (x=r) , a ≤ x ≤ r

x (r=a) − (a=r)  |
| | |

x
 ν
Ex e−γ`(H,r) ; VH = a

3.3.5
a
(a)

| | (b=x)| | − (x=b)| |


 | | (b=a)| | − (a=b)| | + 2 (b=r)| | − (r=b)| |
(r=a)| | − (a=r)| |
, r ≤ x ≤ b



r
=
| | (b=x)| | − (x=b)| | + r2 (b=r)| | − (r=b)| | (r=x)| | − (x=r)| |






, a ≤ x ≤ r
| | (b=a)| | − (a=b)| | + r2 (b=r)| | − (r=b)| | (r=a)| | − (a=r)| |



x
 ν
Ex e−γ`(H,r) ; VH = b

3.3.5
b
(b)
 3. STOPPING AT FIRST EXIT TIME 635
2
9 Vs = eσ νs+σWs
σ>0 H = Ha,b = min{s : Vs ∈
/ (a, b)}

| | (x=a)| | − (a=x)| | + r2 (x=r)| | − (r=x)| | (r=a)| | − (a=r)| |





 | | (b=a)| | − (a=b)| | +

| | − (r=b)| | (r=a)| | − (a=r )| |
, r≤x≤b
2 (b=r )




= r
| | (x=a)| | − (a=x)| |




, a≤x≤r
| | (b=a)| | − (a=b)| | + r2 (b=r)| | − (r=b)| | (r=a)| | − (a=r)| |



| |yr 2(b=a)| | − (a=b)| |





3.3.6 Px `(H, r) ∈ dy, VH = a = exp −
(b=r)| | − (r=b)| | (r=a)| | − (a=r)| |



(a)
a | |r 2 (b=x)| | − (x=b)| |
  ν


dy, r ≤ x ≤ b
(b=r)| | − (r=b)| | (r=a)| | − (a=r)| |

 x


×  ν
a | |r2 (x=a)| | − (x=y)| |


dy, a≤x≤r


2
x (r=a)| | − (a=r)| |





 0, r≤x≤b
(1) Px `(H, r) = 0, VH = a =  a ν (r=x)| | − (x=r)| |
, a≤x≤r
x (r=a)| | − (a=r)| |


| |yr 2(b=a)| | − (a=b)| |





3.3.6 Px `(H, r) ∈ dy, VH = b = exp −
(b=r)| | − (r=b)| | (r=a)| | − (a=r)| |



(b)
b | |r2 (b=x)| | − (x=b)| |
  ν

dy, r≤x≤b
(b=r)| | − (r=b)| | 2

 x


×  
| |r 2 (x=a)| | − (x=y )| |
 b
ν




dy, a ≤ x ≤ r
(b=r)| | − (r=b)| | (r=a)| | − (a=r)| |

x

(  b ν (x=r)| | − (r=x)| |
, r≤x≤b
x (b=r)| | − (r=b)| |


(1) Px `(H, r) = 0, VH = b =
0, a≤x≤r
 Z H
x ν
n


 o 
3.6.5 Ex exp − p1I[a,r] Vs + q1I[r,b] Vs ds ; VH = a
0
a
(a)
q (b=r)q + (r=b)q (r=x)p − (x=r)p + p (b=r)q − (r=b)q (r=x)p + (x=r)p





a≤x =
q (b=r)q + (r=b)q (r=a)p − (a=r)p + p (b=r)q − (r=b)q (r=a)p + (a=r)p





x≤r

2p (b=x)q − (x=b)q

r≤x =
q (b=r)q + (r=b)q (r=a)p − (a=r)p + p (b=r)q − (r=b)q (r=a)p + (a=r)p





x≤b

p
Υs := ν 2 + 2s/σ 2
636 9. GEOMETRIC BROWNIAN MOTION
2
9 Vs = eσ νs+σWs
σ>0 H = Ha,b = min{s : Vs ∈
/ (a, b)}

H
x
n  Z


 o ν
3.6.5 Ex exp − p1I[a,r] Vs + q1I[r,b] Vs ds ; VH = b
0
b
(b)
2q (x=a)p − (a=x)p


a≤x =
q (b=r) q + (r=b) q (r=a) p − (a=r)p + p (b=r)q − (r=b)q (r=a)p + (a=r)p
  





x≤r

q (x=r)q + (r=x)q (r=a)p − (a=r)p + p (x=r)q − (r=x)q (r=a)p + (a=r)p

r≤x =
q (b=r)q + (r=b)q (r=a)p − (a=r)p + p (b=r)q − (r=b)q (r=a)p + (a=r)p





x≤b

 Z H 
3.8.3 Ex exp −αH − γ Vs ds
0

√
a−α Sα 2 2 b ; 2 2 x + b−α Sα 2 2 x ; 2 2 a
√
√ √

= √
x−α Sα 2 2 b ; 2 2 a
√

√
o xα− S 2 2 b 2√2 x

H
; 
Z
α
√
n  
3.8.7 Ex exp −αH − γ Vs ds ; VH =a = 2 2 b √
aα− Sα  ; 2 2 a


(a) 0

√
Z H
xα− Sα 2 2 x ; 2√2 a


n   o
3.8.7 Ex exp −αH − γ Vs ds ; VH =b = √
(b) 0 bα− Sα 2 2 b 2√2 a

; 

Ex e−γ Ȟ(H) ; VH = a

3.12.5
(a)

√ √
b
(x=a)  2+2 =2 − (a=x)  2+2 =2
| |(a=x)
Z

= √ √
dy
y (y=a)| | − (a=y)| | (y=a)  2+2 =2 − (a=y)  2+2 =2


x

Ex e−γ Ĥ(H) ; VH = b

3.12.5
(b)

√ √
x
(b=x)  2+2 =2 − (x=b)  2+2 =2
| |(b=x)
Z

= √ √
dy
y (b=y)| | − (y=b)| | (b=y)  2+2 =2 − (y=b)  2+2 =2


a

p
Υs := ν 2 + 2s/σ 2
 3. STOPPING AT FIRST EXIT TIME 637
2
9 Vs = eσ νs+σWs
σ>0 H = Ha,b = min{s : Vs ∈
/ (a, b)}

 a ν −ν 2 σ2 u/2
Z b ln(x=a) ; ln(y=a)

| | ssu
3.12.6 Px Ȟ(H) ∈ du, VH =a = e  
dy du
x x y (y=a)| | − (a=y)| |
(a)

 b ν −ν 2 σ2 u/2
Z x ln(b=x) ; ln(b=y)

| | ssu
3.12.6 Px Ĥ(H) ∈ du, VH = b = e  
dy du
x a y (b=y)| | − (y=b)| |
(b)

3.12.8 Px Ȟ(H) ∈ du, H ∈ dt, VH = a u<t
(a)

1
Z b
a
 ν
−ν 2 σ 2 t/2 ln(x/a) ln(y/a)

st−u ln(y/a)


= e ssu , σ dy dudt
x x
y σ σ

3.12.8 Px Ĥ(H) ∈ du, H ∈ dt, VH = b u<t
(b)

1
Z x
b
 ν 2 2 ln(b/x) ln(b/y)

e−ν σ t/2 st−u ln(b/y)


= ssu , σ dy dudt
x a
y σ σ

H o aα+ S2α 2√√2 √

; 2√2x
ds
Z

 b
n  
3.19.7 Ex exp −αH − γ ; VH = a = 2√√2 √
Vs xα+ S2α ; 2√2a


(a) 0  b

H bα+ S2α 2√√2 ; 2√√2

ds
Z
 x  a
n   o
3.19.7 Ex exp −αH − γ ; VH = b = 2√√2 ; 2√√2

Vs x−α+ S2α
(b) 0  b  a

o xα sign − Sα =| bβ xβ

2 | | ; | |
n  Z H  |
3.20.7 Ex exp −αH − Vs2β ds ; VH =a =
2 0 aα sign − Sα =| |
bβ aβ

| | ; | |
(a)

o xα sign − S =| xβ aβ

2 | | ; | |
n  Z H 
α |
3.20.7 Ex exp −αH − Vs2β ds ; VH = b =
2 0 bα sign − Sα =| |
bβ aβ

| | ; | |
(b)

p
Υs := ν 2 + 2s/σ 2
638 9. GEOMETRIC BROWNIAN MOTION
2
9 Vs = eσ νs+σWs
σ>0 H = Ha,b = min{s : Vs ∈
/ (a, b)}

p2
n  Z H
 o
3.21.7 Ex exp −αH − Vs2β + qVsβ ds ; VH = a
0
2
(a)

xα sign − expS | α| + pq| | + 12 ; 2|α| + 1; 2pb

− px| ; 2px
β
β β

| | | | |
=
aα sign − exp − pa| | S | α| + pq| | + 12 ; 2|α| + 1; 2pb
β 2paβ

;
β

| | | |

p2
n  Z H
 o
3.21.7 Ex exp −αH − Vs2β + qVsβ ds ; VH = b
0
2
(b)

xα sign − expS | α| + pq| | + 12 ; 2|α| + 1; 2px

− px| ; 2pa
β
β β

| | | | |
=
bα sign − exp − pb| | S | α| + pq| | + 12 ; 2|α| + 1; 2pb
β 2paβ

;
β

| | | |

p
Υs := ν 2 + 2s/σ 2
 4. STOPPING AT INVERSE LOCAL TIME 639
2
9 Vs = eσ νs+σWs
σ>0 % = %(v, z) = min{s : `(s, z) = v}

4. Stopping at inverse local time

  √ 2 √ 2
x ν +2α/σ 2−ν 2 2
e−vzσ ν +2α/σ , x ≤ z

z


4.0.1 Ex e−α% =  √
 z ν 2+2α/σ 2+ν 2
√ 2 2
e−vzσ ν +2α/σ , z ≤ x


x

| ln(x=z )| + vz2  z ν  2 2
  t (| ln(x=z )| + vz2 )2

4.0.2 Px % ∈ dt = √
3= 2 exp − − 2 dt
 2t x 2 2 t

x
  |ν|−ν 2
1 − e−|ν|vzσ , x ≤ z
z




(1) Px % = ∞ =
 1 − z |ν|+ν e−|ν|vzσ2 , z ≤ x
  
x

4.1.2 Px sup Vs < y, % < ∞
0≤s≤%
z<y
x | |vz 2
  |ν|−ν  
exp − , x≤z
z 1 − (z=y)2| |


=  ν | | | | 2 
 z (y=x) − (x=y) exp − | |vz

z≤x≤y

,
|
x (y=z ) − (z=y)
| |  | 1 − (z=y)2| |

Ex e−α% ; sup Vs < y


4.1.4
0≤s≤%

 √ν 2+2α/σ2−ν
x vz2  2 + 2 =2
 p 

 z exp − √ , x≤z
1 − (z=y)2  2+2 =2


z≤y
=   √ √
z ν (y=x)  +2 = − (x=y)  +2 =
2 2 2 2
vz2  2 + 2 =2
 p 

√ √ exp − √ , z≤x

x (y=z )  2+2 =2 − (z=y)  2+2 =2

1 − (z=y)2  2+2 =2

x≤y

4.2.2 Px inf Vs > y, % < ∞
0≤s≤%
y<z
| |vz 2
  z ν (x=y)| | − (y=x)| |  
 x (z=y)| | − (y=z )| | exp − 1 − (y=z )2| | , y ≤ x ≤ z

=  |ν|+ν
 z | |vz 2
 
exp − y≤z≤x

2  ,
x 1 − (y=z ) | |
640 9. GEOMETRIC BROWNIAN MOTION
2
9 Vs = eσ νs+σWs
σ>0 % = %(v, z) = min{s : `(s, z) = v}

Ex e−α% ; inf Vs > y


4.2.4
0≤s≤%

√ √
z (x=y)  2+2 =2 − (y=x)  2+2 =2 vz2  2 + 2 =2
 ν  p 
√ √ exp − √ , y≤x
 x (z=y)  2+2 =2 − (y=z )  2+2 =2

1 − (y=z )2  2+2 =2

x≤z

=  √ 2
z 2
ν +2α/σ +ν vz2  2 + 2 =2
 p 

exp − √ , y≤z

x

1 − (y=z )2  2+2 =2

z≤x

vz2 | |( + | |r2 )
 
Ex e−γ`(%,r) ; % < ∞ = exp −

4.3.1
| |r 2 + 1 − e−2| ln(r=z)|| |


 1, r ≤ z ≤ x, x ≤ z ≤ r
 | |r2 + 1 − e−2| || ln(x=r)|



z
 ν 

, z ∧ r ≤ x ≤ z ∨ r
= e−|ν|| ln(z/x)| | |r2 + 1 − e−2| || ln(z=r)|
x
| |r 2




, z ≤ r ≤ x, x ≤ r ≤ z
| |r 2 + 1 − e−2| || ln(z=r)|



 z ν 
(vz + yr)2 | | 
4.3.2 Px `(%, r) ∈ dy, % < ∞ = exp −
x 1 − e−2| ln(r=z)|| |

√  | |2 √vzyr 
| |r 2 e−| ln(z=x)|| | vz

√ I1 dy, x ≤ z ≤ r, r≤z≤x
2 sh(| ln(r=z )|| |) yr sh(| ln(r=z )|| |)




√  | |2 √vzyr 
| |r 2 vz

 h

2 sh(| ln(r/x)||ν|) √ I1

2 sh (| ln(r=z )|| |) yr sh(| ln(r=z )|| |)


×  | |2 √vzyr i
dy, z ∧ r ≤ x ≤ r∨z

+ sh(| ln(z/x)||ν|)I0
sh(| ln(r=z )|| |)




| |r 2 e−| ln(r=x)|| | | | 2 vzyr
√



  
 I dy, z ≤ r ≤ x, x≤r≤z
sh(| ln(r=z )|| |) 0 sh(| ln(r=z )|| |)

(1) Px `(%, r) = 0, % < ∞

 −|ν|| ln(z/x)|
e , x ≤ z ≤ r, r ≤ z ≤ x
z | |vz 2 sh(| || ln(x=r)|)
 ν  

= exp − , z∧r ≤x≤r∨z
x 1 − e−2| || ln(z=r)| 
 sh(| || ln(z=r)|)
0, z ≤ r ≤ x, x ≤ r ≤ z
 4. STOPPING AT INVERSE LOCAL TIME 641
2
9 Vs = eσ νs+σWs
σ>0 % = %(v, z) = min{s : `(s, z) = v}

vz2
 Z %   
4.8.3 Ex exp −α% − γ Vs ds = exp − √  √ 
2 2
4I2  2+2 =2  z K2√ 2+2 =2 2 2 z
 
0 √

√
2  2+2 =2 (2 2 x= )
  ν I √
z
 √ , x≤z
2  +2 =2 (2 2 z= )
 x I√ 2


×   √
z ν K2  2+2 =2 (2 2 x=)
 √
√ , z≤x

x K2√ 2+2 =2 (2 2 z=)



Ex e−α% ; a < inf Vs , sup Vs < b


4.15.4
0≤s≤% 0≤s≤%

√ √
vz2  2 + 2 =2 (b=a)  2+2 =2 − (a=b)  2+2 =2
 p 

= exp − √ √ √ √
2 (b=z )  2+2 =2 − (z=b)  2+2 =2 (z=a)  2+2 =2 − (a=z )  2+2 =2

√ √
z (x=a)  2+2 =2 − (a=x)  2+2 =2
  ν
 x (z=a)√ 2+2
 √ , −b ≤ x ≤ z
=2 − (a=z )  2+2 =2

×   √ √
 2+2 =2 − (x=b)  2+2 =2
 z (b=x)√
 ν
√ , z≤x≤b

x (b=z )  2+2 =2 − (z=b)  2+2 =2

vz2
%
ds
 Z   
4.19.3 Ex exp −α% − γ = exp − 2√√2
 √ 
2 2
Vs 4I2√ 2+2 =2  z K2  2+2 =2 √z
 
0 √

√ √
2+2 = 2 (2 2 = x)
  ν K √
z 2 
√ √ , x≤z

2  +2 =2 (2 2 = z )
 x K√ 2


×   √ √ √
 z ν I2  2+2 =2 (2 2 = x)
√ √ , z≤x

x I2√ 2+2 =2 (2 2 = z )



2 %
vz2 | |
 Z   
4.20.3 Ex exp −α% − Vs2β ds = exp −
2 0 2I √ν 2+2α/σ2 zβ
√ zβ

| | K ν 2+2α/σ2 | |
|β| |β|

 2+2 =2 =| | ( x = | |)
   ν I√
z
 , (z − x)β ≥ 0
 +2 =2 =| | ( z = | |)
 x I√ 2


×   √
 z ν K  2+2 =2 =| | ( x =| |)
, (z − x)β ≤ 0

x K√ 2+2 =2 =| | ( z =| |)


642 9. GEOMETRIC BROWNIAN MOTION
2
9 Vs = eσ νs+σWs
σ>0 % = %(v, z) = min{s : `(s, z) = v}

p2
 Z %

4.21.3 Ex exp −α% − Vs2β + qVsβ ds
0
2
0<x
 √ √
 pvz +1  2  2 2+2 + 1 −1  2 2+2 + 12 + pq|
 
| | | |

|
= exp − 2pzβ
W 2pzβ

M q √
2 2 | |
√
ν 2 σ 2+2α | |
− pσ|β| ; ν σ +2α
σ|β|
− q
pσ|β|
; σ|β|

  ν z =2 M √ (2px =| |)
z −q=p | |;  2+2 =2 =| |

=2 M , (z − x)β ≥ 0
 x x √ (2pz =| |)

−q=p | |;  2+2 = 2 =| |

× =2 W
  ν z
z
√
−q=p | |;  2+2 = 2 =| |
(2px =| |)
, (z − x)β ≤ 0

=2 W (2pz =| |)

x x √

−q=p | |;  2+2 = 2 =| |
 APPENDIX 2

SPECIAL FUNCTIONS

1. Hyperbolic functions

1 ∞ x2k+1
sh x := (ex − e−x ) =
P
2 k=0
(2k + 1)!

1 ∞ x2k
ch x := (ex + e−x ) =
P
2 k=0
2k!

sh x
th x :=
ch x
ch x
cth x :=
sh x
2. Gamma function
Z ∞
Γ(x) := ux−1 e−u du, Re x > 0
0
1
Γ(x + 1) = xΓ(x), ' x as x → 0
(x)

Γ(x)Γ(1 − x) =
sin(x)

22x−1 1

Γ(2x) = √ Γ(x)Γ x +
 2
√
Γ(ax + b) ' 2πe−ax (ax)ax+b−1/2 as x → ∞, a>0
√
Γ(n + 1) = n!, Γ(n + 1/2) = 1 · 3 · . . . · (2n − 1)2−n π, n = 1, 2, . . .
√ √ √
Γ(1/2) = π, Γ(3/2) = π/2, Γ(−1/2) = −2 π

3. Bessel functions
P∞ (−1)k (x=2) +2k
Jν (x) :=
k=0
k! ( + k + 1)

1

Yν (x) := J (x) cos(νπ) − J−ν (x)
sin() ν

Yν0 (x)Jν (x) − Yν (x)Jν0 (x) = Yν (x)Jν+1 (x) − Yν+1 (x)Jν (x) = 2/(πx)

0 < jν,1 < jν,2 < . . . – positive zeros of Jν (x) when ν ≥ 0

643
644 APPENDIX 2

4. Modified Bessel functions

∞ (x=2) +2k
= i−ν Jν (ix)
P
Iν (x) :=
k=0
k ! ( + k + 1)



Kν (x) := I−ν (x) − Iν (x)
2 sin()

are linearly independent solutions of the Bessel equation.

x2 Y 00 (x) + xY 0 (x) − (x2 + ν 2 )Y (x) = 0, x>0

Integral representations: for x > 0

Z 1
(x=2)
Iν (x) = √ (1 − t2 )ν−1/2 ext dt, Re ν > −1/2
 ( + 1=2) −1

√
(x=2) ∞ 2
Z
Kν (x) = (t − 1)ν−1/2 e−xt dt, Re ν > −1/2
( + 1=2) 1

Properties:

I−ν (x) = Iν (x) ν = 1, 2, . . .

K−ν (x) = Kν (x)

Iν0 (x)Kν (x) − Iν (x)Kν0 (x) = Iν (x)Kν+1 (x) + Iν+1 (x)Kν (x) = 1/x

0
0
x−ν Iν (x) = x−ν Iν+1 (x), xν Iν (x) = xν Iν−1 (x)

0
0
x−ν Kν (x) = −x−ν Kν+1 (x), xν Kν (x) = −xν Kν−1 (x)

2 2
Iν−1 (x) − Iν+1 (x) = Iν (x), Kν+1 (x) − Kν−1 (x) = Kν (x)
x x
1  x ν
Iν (x) ' as x → 0, ν 6= −1, −2, . . .
( + 1) 2

( )  x −ν
Kν (x) ' , ν > 0, K0 (x) ' − ln x as x → 0
2 2
1 x
Iν (x) ' √ e as x → ∞
2x
√

Kν (x) ' √ e−x as x → ∞
2x
√ ν √ λ
√ x/ν

1  √
Iν √ ν γe '√ 1= 4 √ √ √ e(ν+x) λ+γ as ν → ∞
λ 2( + )  + +
 SPECIAL FUNCTIONS 645
√ √ −ν √λ
√  √

e−(ν+x)
x/ν

λ+γ
Kν √ ν γe ' √ √ √ as ν → ∞
λ ( + )1=4 2 + +

Special cases:
√ √ √
2  2
I 12 (x) = √ sh x; K 12 (x) = K− 12 (x) = √ e−x ; I− 12 (x) = √ ch x;
x 2x x
√ √
2 
I 32 (x) = √ (x ch x − sh x); K 32 (x) = √ e−x (x + 1).
x3=2 2x3=2

5. Airy function
√
1√  2
2
 x 2

Ai(x) := x I−1/3 x3/2 − I1/3 x3/2 = √ K1/3 x3/2
3 3 3  3 3

. . . αk < · · · < α2 < α1 < 0 – zeros of the Airy function Ai(x)

. . . αk0 < · · · < α20 < α10 < 0 – zeros of the derivative of the Airy function Ai0 (x)

6. Hermite polynomials and related functions

2 dn −x2/2
P (−1)k 2−k n! n−2k

Hen (x) := (−1)n ex /2 n e = x
dx 0≤k≤n/2
k!(n − 2k)!

√ 1 2

v

hy (n, v) := L−1 (2γ)n/2−1/2 e−v −v
2γ

/2y
= √
(n+1)=2 e Hen √ , 0<v
γ
2y y

7. Binomial series
∞ ( + 1) ∞ (−1)k (− + k )
(1 + x)µ = xk = xk ,
P P
|x| < 1
k=0
( + 1 − k) k! k=0
(−) k!

8. Error functions

2
Z x
2 2 P ∞ (−1)k x2k+1 2 2 P
∞ 2k x2k+1
Erf(x) := √ e−v dv = √ = √ e−x
 0  k=0 k!(2k + 1)  k=0
1 · 3 · 5 · · · (2k + 1)

2 ∞
Z
2
Erfc(x) := √ e−v dv = 1 − Erf(x)
 x

x2k+1
Z x
2 2 2 P
∞
Erfi(x) := √ ev dv = √
 0  k=0 k!(2k + 1)

x y
   
Erfid(x, y) := Erfi √ − Erfi √
2 2
1 2
Erfc(x) ' √ e−x as x → ∞
x
646 APPENDIX 2

9. Parabolic cylinder functions

2 √ n 1  ∞  ( + 2) · · · ( + 2k − 2) 
D−ν (x) := e−x /4 2−ν/2 π x2k
P
1+
(( + 1)=2) k=1
(2k)!
√ 
x 2 ∞ ( + 1)( + 3) · · · ( + 2k − 1) o
x2k
P
− 1+
(=2) k=1
(2k + 1)!
D
e −ν (x) := D−ν (−x)

are linearly independent solutions of the differential equation

 2
x 2 − 1 
Y 00 (x) − + Y (x) = 0, x ∈ R.
4 2
Integral representation: for x ∈ R

1 −x2 /4 ∞ ν−1 −xt−t2 /2

Z
D−ν (x) = e t e dt, Re ν > 0
( ) 0

Properties:

0 x x
D−ν (x) = − D−ν (x) − νD−ν−1 (x) = D−ν (x) − D−ν+1 (x)
2 2
2
0 2 2
0 2
ex /4 D−ν (x) = −νex /4 D−ν−1 (x), e−x /4 D−ν (x) = −e−x /4 D−ν+1 (x)
2
0 2
x1−ν e−x /4 D−ν (x) = −x−ν e−x /4 D−ν+2 (x)
√
0 0 2
D−ν (−x)D−ν (x) + D−ν (−x)D−ν (x) = −
( )
1
√ √ √
lim 2α/4θ Γ + D−α/2θ (x 2θ) = πe−x α , x∈R
θ↓0 4 2

Special cases:
√
e−x =4 dn x2/2 x


2
2
Dn (x)=e−x /4Hen (x), D−n−1 (x)= √ e Erfc √ , n = 0, 1, . . .
(−1)n 2n! dxn 2
√  2  2
x x √ x
D−1/2 (x) = √ K1/4 , D−1/2 (−x) − D−1/2 (x) = πxI1/4 , x≥0
2 4 4
 2  2 
2  x 3/2  x x
D−3/2 (x) = √ − K1/4 + K3/4 , x≥0
 2 4 4
 2  2  2 
1  x 5/2  x x x
D3/2 (x) = √ 2K1/4 + 3K3/4 − K5/4 , x≥0
 2 4 4 4
 2
√ x
D−1/2 (−x) + D−1/2 (x) = πxI−1/4 , x≥0
4
 SPECIAL FUNCTIONS 647

10. Kummer and Whittaker functions

P∞ a(a + 1) : : : (a + k − 1)xk
M (a, b, x) := 1 +
k=1
b(b + 1) : : : (b + k − 1)k!

(1 − b) (b − 1)
U (a, b, x) := M (a, b, x) + x1−b M (1 + a − b, 2 − b, x)
(1 + a − b) (a)

are linearly independent solutions of the Kummer equation

xY 00 (x) + (b − x)Y 0 (x) − aY (x) = 0, x > 0.

Kummer transformations:

M (a, b, x) = ex M (b − a, b, −x), U (a, b, x) = x1−b U (1 + a − b, 2 − b, x)

x1−b M (1 + a − b, 2 − b, x) = x1−b ex M (1 − a, 2 − b, −x)

e−x U (b − a, b, x) = x1−b e−x U (1 − a, 2 − b, x), x>0

Whittaker functions:

Mn,m (x) := xm+1/2 e−x/2 M (m − n + 1/2, 2m + 1, x)

Wn,m (x) := xm+1/2 e−x/2 U (m − n + 1/2, 2m + 1, x)

Integral representations:

(b)
Z 1
M (a, b, x) = ext ta−1 (1 − t)b−a−1 dt, Re b > Re a > 0
(a) (b − a) 0
1 ∞
Z
U (a, b, x) = e−xt ta−1 (1 + t)b−a−1 dt, Re a > 0, Re x > 0
(a) 0

Properties:

(−2m) (2m)
Wn,m (x) = M (x) + M (x)
(1=2 − m − n) n,m (1=2 + m − n) n,−m

@ @ (b) −b x
M (a, b, x) U (a, b, x) − M (a, b, x) U (a, b, x) = x e
@x @x (a)

0 0 (2m + 1)
Mn,m (x)Wn,m (x) − Mn,m (x)Wn,m (x) =
(m − n + 1=2)
@ a @
M (a, b, x) = M (a + 1, b + 1, x), U (a, b, x) = −aU (a + 1, b + 1, x)
@x b @x
@ b
(x M (a, b + 1, x)) = b xb−1 M (a, b, x)
@x
648 APPENDIX 2
@ b
(x U (a, b + 1, x)) = (b − a)xb−1 U (a, b, x)
@x
@ −x a − b −x
(e M (a, b, x)) = e M (a, b + 1, x)
@x b
@ −x
(e U (a, b, x)) = −e−x U (a, b + 1, x)
@x
@ b −x
(x e M (a + 1, b + 1, x)) = bxb−1 e−x M (a, b, x)
@x
@ b −x
(x e U (a + 1, b + 1, x)) = −xb−1 e−x U (a, b, x)
@x
a √
, b + 1, θx = 2b Γ(b + 1)(xa)−b/2 Ib ( xa)


lim M
θ↓0 4
a
a
√
lim θb Γ U , b + 1, θx = 21−b (a/x)b/2 Kb ( xa)
θ↓0 4 4

Special cases:
√
M0,m (2x) = 22m Γ(m + 1) 2xIm (x), x≥0
p
W0,m (2x) = 2x/πKm (x), x≥0

(=2) ν/2−1/4 √

M1/4−ν/2,1/4 (x2/2) = √ 2 x D−ν (−x) − D−ν (x) , x≥0
4 
(( + 1)=2) ν/2−1/4 √

M1/4−ν/2,−1/4 (x2/2) = √ 2 x D−ν (−x) + D−ν (x) , x≥0
2 
√
W1/4−ν/2,1/4 (x2/2) = W1/4−ν/2,−1/4 (x2/2) = 2ν/2−1/4 xD−ν (x), x≥0
√ √
x 2 x 2
M−1/4,1/4 (x2 ) = ex /2 Erf(x), M1/4,1/4 (x2 ) = e−x /2 Erfi(x)
2 2
√ 2 √ 2
W−1/4,1/4 (x2 ) = πx ex /2 Erfc(x), W1/4,1/4 (x2 ) = x e−x /2 , x≥0

11. Hypergeometric functions

For −1 < x < 1

∞
P ( + 1) : : : ( + k − 1) ( + 1) : : : ( + k − 1)xk
F (α, β, γ, x) := 1 +
k=1
( + 1) : : : ( + k − 1)k!

for 0 < x < 2

1
G(α, β, γ, x) := F (α, β, α + β + 1 − γ, 1 − x), Re(α + β + 1 − γ) > 0
( + +1− )

are linearly independent solutions of the hypergeometric differential equation

 SPECIAL FUNCTIONS 649

x(1 − x)Y 00 (x) + (γ − (α + β + 1)x)Y 0 (x) − αβY (x) = 0, 0 < x < 1.

Integral representations: for −1 < x < 1

( ) 1 Z
F (α, β, γ, x) = tα−1 (1 − t)γ−α−1 (1 − tx)−β dt, Re γ > Re α > 0
( ) ( − ) 0

for 0 < x < ∞

1 ∞ Z
G(α, β, γ, x) = tα−1 (1+t)γ−α−1 (1+tx)−β dt, Re(β +1−γ) > 0
( ) ( +1− ) 0
Re α > 0

Properties:

@ @ ( )x− (1 − x) − − −1
F (α, β, γ, x) G(α, β, γ, x) − F (α, β, γ, x) G(α, β, γ, x) =
@x @x ( ) ( )
@
F (α, β, γ, x) = F (α + 1, β + 1, γ + 1, x)
@x

@
G(α, β, γ, x) = −αβG(α + 1, β + 1, γ + 1, x)
@x
@
( − )
(1 − x)α F (α, β, γ, x) = (1 − x)α−1 F (α + 1, β, γ + 1, x)
@x

@

(1 − x)α G(α, β, γ, x) = −α(1 − x)α−1 G(α + 1, β, γ + 1, x)
@x
@ γ

x (1 − x)1+α+β−γ F (α + 1, β + 1, γ + 1, x) = γxγ−1 (1 − x)α+β−γ F (α, β, γ, x)
@x

@ γ
x −1 (1 − x) + −
x (1 − x)1+α+β−γ G(a + 1, b + 1, γ + 1, x) = − G(α, β, γ, x)
@x 1+ + −

F (α, β, γ, x) = (1 − x)γ−α−β F (γ − α, γ − β, γ, x), 0≤x<1

G(α, β, γ, x) = x1−γ G(α + 1 − γ, β + 1 − γ, 2 − γ, x), 0<x≤1

x ∞
Z
F (α, β, γ, 1/x) = e−xt tβ−1 M (α, γ, t) dt, 1 ≤ x, Re β > 0
( ) 0

x ∞
Z
G(α, β, γ, 1/x) = e−xt tβ−1 U (α, γ, t) dt, 0 ≤ x, Re β > 0
( ) ( +1− ) 0

(1 − )F ( ; ; ; x) ( − 1)x1− F ( + 1 − ; + 1 − ; 2 − ; x)
G(α, β, γ, x) = +
( +1− ) ( +1− ) ( ) ( )
0<x<1
x
x

lim F α, β, γ, = M (α, γ, x), lim Γ(β + 1 − γ)G α, β, γ, = U (α, γ, x)
β→∞ β→∞
650 APPENDIX 2

Special cases:

( ) ( − − ) x−
F (α, β, γ, 1)= , F (α, β, β, x) = (1 − x)−α , G(α, β, β + 1, x)=
( − ) ( − ) ( )

(1 − ) 1
G(α, β, γ, 0)= , G(α, β, γ, 1)=
( +1− ) ( +1− ) ( + +1− )

12. Legendre functions

Legendre function of the first kind

for −1 < x < 1

1 1 + x q/2

1 − x

Pepq (x) := F − p, 1 + p, 1 − q, , Re q < 1
(1 − q) 1 − x 2
for 1 < x

2q xq−p−1 p−q+1 p−q+2 1

Ppq (x) := F , , 1 − q, 1 − 2 , Re q < 1
(x2 − 1)q=2 (1 − q) 2 2 x

Legendre function of the second kind, divided by eiqπ , for 1 < x

√
e qp (x) :=  (p + q + 1)(x2 − 1)q=2 p+q+1 p+q+2 3 1

Q F , ,p + , 2
2p+1 (p + 3=2) xp+q+1 2 2 2 x

are linearly independent solutions of the Legendre equation

q2

(1 − x2 )Y 00 (x) − 2xY 0 (x) + p(1 + p) − Y (x) = 0.
1 − x2

Integral representations: for −1 < x < 1

2−p (1 − x2 )−q=2 ∞
Z
Pepq (x) = (x + ch t)q−p−1 sh1+2p t dt, Re q < Re(−p) < 1
(−p − q) (1 + p) 0
for 1 < x
−q 2 √
q=2  (p + q + 1) ∞
e qp (x) = 2 (x − 1)
Z √
Q (x + x2 − 1 ch t)−q−p−1 sh2q t dt
(p − q + 1) (q + 1=2) 0
Re(p ± q + 1) > 0

Properties:

d q e qp (x) − Ppq (x) d Q

e q (x) = (1 + p + q)
P (x)Q , 1<x
dx p dx p (1 + p − q)(x2 − 1)

d eq d 2
P (−x)Pepq (x) − Pepq (−x) Pepq (x) = , −1 < x < 1
dx p dx (−p − q) (1 + p − q)(1 − x2 )

e −q (p − q + 1) e q q q
Q p (x) = (p + q + 1) Qp (x), Ppq (x) = P−p−1 (x), Pepq (x) = Pe−p−1 (x)
 SPECIAL FUNCTIONS 651

2q (1 − x2 )p=2
lim(x2 − 1)q/2 Ppq (x) = , Pep−p (x) = p
x↓1 (1 − q) 2 (p + 1)

−p −p 2p (p) −p −p 2p (p) x
Pe−p (−x) + Pe−p (x) = , Pe1−p (−x) − Pe1−p (x) =
(2p)(1 − x2 )p=2 (2p − 1)(1 − x2 )p=2

d
 P q (x) 
p P q+1 (x) d
 Qeq (x)  e q+1 (x)
Q
2 q=2 = 2 p (q+1)=2 , 2
p
q=2 = − 2 p (q+1)=2 , 1<x
dx (x − 1) (x − 1) dx (x − 1) (x − 1)
d

(x2 − 1)q/2 Ppq (x) = (p + q)(p − q + 1)(x2 − 1)(q−1)/2 Ppq−1 (x), 1<x
dx
d

(1 − x2 )q/2 Pepq (x) = (p + q)(p − q + 1)(1 − x2 )(q−1)/2 Pepq−1 (x), −1 < x < 1
dx
d

(x2 − 1)q/2 Q
e qp (x) = −(p + q)(p − q + 1)(x2 − 1)(q−1)/2 Q
e q−1 (x), 1<x
dx p

z

e qp ch z = Kq (z)
lim pq Pp−q ch lim p−q Q



= Iq (z),
p→∞ p p→∞ p

Special cases:
√ √
−1/2 2 2 sh(x(p + 1=2)) 1/2 2 ch(x(p + 1=2))
Pp (ch x) = √ , Pp (ch x) = √
(2p + 1) sh1=2 x  sh1=2 x
√ √
−1/2 2e−x(p+1=2) 1/2 e−x(p+1=2)
Qp (ch x) = , Qp (ch x) = √
(2p + 1) sh1=2 x 2 sh1=2 x
13. Theta functions of imaginary argument
√

ch(v 2 )  1 ∞ 2
csy (v, t) := L−1 e−(v+t+2kt) /2y
P
√ √ = √
γ 2 sh(t 2 ) 2y k=−∞
√

sh(v 2 )  1 ∞ 2
scy (v, t) := L−1 (−1)k e−(t−v+2kt) /2y
P
√ √ = √
γ 2 ch(t 2 ) 2y k=−∞
√

sh(v 2 )  ∞ t − v + 2kt −(t−v+2kt)2/2y
ssy (v, t) := L−1
P
√ = √ e , v<t
γ sh(t 2 ) k=−∞ 2y3=2
√

sh(v 2 )  ∞ 
|t − v + 2kt|

e y (v, t) := L−1
P
ss √ = sign(t − v + 2kt) Erfc √
γ sh(t 2 ) k=−∞
2y
√

ch(v 2 )  ∞ v + t + 2kt 2
ccy (v, t) := L−1 (−1)k √ 3=2 e−(v+t+2kt) /2y ,
P
√ = v<t
γ ch(t 2 ) k=−∞ 2y
√

ch(v 2 )  ∞ 
|v + t + 2kt|

e y (v, t) := L−1 (−1)k sign(v + t + 2kt) Erfc
P
cc √ = √
γ ch(t 2 ) k=−∞
2y
√ √

2 
2 ∞ 2 2
L−1 ((2k + 1)2 t2 − y)e−(2k+1) t /2y
P
sy (t) := √ = √ , 0<t
γ sh(t 2 ) y5=2 k=0
652 APPENDIX 2

14. Special Inverse Laplace transforms


(2 )=2 −z√2γ 
cy (µ, ν, t, z) := L−1 √ e t>0
γ ch (t 2 )
∞ (−1)k ( + k )e−(t+z+2kt)2=4y 
t + z + 2kt

= 2ν
P
√ D √ , ν ≥ 0, νt + z > 0
k=0 2y1+=2 ( )k! µ+1
y

(2 )=2 −z√2γ 
sy (µ, ν, t, z) := L−1 √ e t>0
γ sh (t 2 )

( + k)e−(t+z+2kt) =4y t + z + 2kt

∞ 2  
= 2ν
P
√ D √ , ν ≥ 0, νt + z > 0
k=0 2y1+=2 ( )k! µ+1
y
√

(2 )=2 e−x 2 
ry (µ, t, x, z) := L−1 √ √ z 6= −1, t>0
γ z sh(t 2 ) + ch(t 2 )
√
2(z − 1)k e−(x+t+2kt) =4y x + t + 2kt
∞ 2  
P
= √ (k +1) 1+=2 Dµ+1 √ , x+t>0
k=0 (z + 1) y y
√

(2 )=2 e−x 2 
rc
e y (µ, t, x, z) := L−1 √ √ √ z 6= 0, t>0
γ sh(t 2 ) + z 2 ch(t 2 )

P∞ (−1)k k ! Pk (−1)l
= k k +1 c (µ − k − 1, k + 1, t, x + kt − 2lt), x > −t
k=0
2 z l=0
(k − l)!l! y
√

(2 )=2 sh(x 2 ) 
rcy (µ, t, x, z) := L−1 √ √ √ t>0
γ sh(t 2 ) + z 2 ch(t 2 )
1 1
e y (µ, t, −x, z) − rc
= rc e y (µ, t, x, z), x<t
2 2
√

(2 )=2 e−x 2 
e y (µ, t, x, z) := L−1
rs √ √ √ z 6= 0, t>0
γ z 2 sh(t 2 ) + ch(t 2 )

P∞ (−1)k k ! P k 1
= k k +1 s (µ − k − 1, k + 1, t, x + kt − 2lt), x > −t
k=0
2 z l=0
( k − l)!l! y
√

(2 )=2 ch(x 2 ) 
rsy (µ, t, x, z) := L−1 √ √ √ t>0
γ z 2 sh(t 2 ) + ch(t 2 )

1 1
e y (µ, t, −x, z) + rs
= rs e y (µ, t, x, z), x<t
2 2
√ √

(2 )=2  √
z 2 sh(t 2 )

ecy (µ, ν, t, x, z) := L−1 √ exp −x 2γ − √ t>0
γ ch (t 2 ) ch(t 2 )
P∞ zk
= c (µ + k, ν + k, t, x + z + kt), ν ≥ 0, νt + x + z > 0
k=0
k! y
 SPECIAL FUNCTIONS 653
√ √

(2 )=2  √
z 2 ch(t 2 )

esy (µ, ν, t, x, z) := L−1 √ exp −x 2γ − √ t>0
γ sh (t 2 ) sh(t 2 )
P∞ (−z )k
= s (µ + k, ν + k, t, x + z + kt), ν ≥ 0, νt + x + z > 0
k=0
k! y
√  √ √ √

2 ν
z 2 ch(t 2 ) x 2

eey (ν, t, z, x) := L−1 √ exp − √ − √ t>0
γ sh(t 2 ) sh(t 2 ) sh(t 2 )
P ∞ (−x)l
∞ (−z )k P
= s (ν + k + l, ν + k + l, t, z + kt), ν ≥ 0, νt + z > 0
k=0
k! l=0 l! y

For t + x + z − v > 0, t>0

√ √ √

ch(v 2 )  √
z 2 sh(t 2 )

eccy (v, t, x, z) := L−1 √ exp −x 2γ − √
γ ch(t 2 ) ch(t 2 )
1 1
= ec (0, 1, t, x − v, z) + ecy (0, 1, t, x + v, z)
2 y 2
√ √ √

sh(v 2 )  √
z 2 ch(t 2 )

essy (v, t, x, z) := L−1 √ exp −x 2γ − √
γ sh(t 2 ) sh(t 2 )
1 1
= es (0, 1, t, x − v, z) − esy (0, 1, t, x + v, z)
2 y 2
√ √ √

sh(v 2 )  √
z 2 sh(t 2 )

escy (v, t, x, z) := L−1 √ √ exp −x 2γ − √
γ 2 ch(t 2 ) ch(t 2 )
1 1
= ec (−1, 1, t, x − v, z) − ecy (−1, 1, t, x + v, z)
2 y 2
√ √ √

ch(v 2 )  √
z 2 ch(t 2 )

ecsy (v, t, x, z) := L−1 √ √ exp −x 2γ − √
γ 2 sh(t 2 ) sh(t 2 )
1 1
= es (−1, 1, t, x − v, z) + esy (−1, 1, t, x + v, z)
2 y 2
√ √ √

ch(v 2 )  √
z 2 sh(t 2 )

f y (v, t, x, z) := L−1
ecc √ exp −x 2γ − √
γ 2 ch(t 2 ) ch(t 2 )
1 1
= ec (−2, 1, t, x − v, z) + ecy (−2, 1, t, x + v, z)
2 y 2
√ √ √

sh(v 2 )  √
z 2 ch(t 2 )

f y (v, t, x, z) := L−1
ess √ exp −x 2γ − √
γ 2 sh(t 2 ) sh(t 2 )
1 1
= es (−2, 1, t, x − v, z) − esy (−2, 1, t, x + v, z)
2 y 2
For 2t + z − u − v > 0, t>0
√ √  √ √

ch(u 2 ) ch(v 2 ) z 2 sh(t 2 )

ecccy (u, v, t, z) := L−1 √ exp − √
γ ch2 (t 2 ) ch(t 2 )
654 APPENDIX 2
1 1
= ec (0, 2, t, −u − v, z) + ecy (0, 2, t, u + v, z)
4 y 4
1 1
+ ecy (0, 2, t, −u + v, z) + ec (0, 2, t, u − v, z)
4 4 y
√ √  √ √

sh(u 2 ) sh(v 2 ) z 2 ch(t 2 )

esssy (u, v, t, z) := L−1 √ exp − √
γ sh2 (t 2 ) sh(t 2 )
1 1
= es (0, 2, t, −u − v, z) + esy (0, 2, t, u + v, z)
4 y 4
1 1
− esy (0, 2, t, −u + v, z) − es (0, 2, t, u − v, z)
4 4 y
√ √  √ √

ch(u 2 ) sh(v 2 ) z 2 sh(t 2 )

ecscy (u, v, t, z) := L−1 √ 2 √ exp − √
γ 2 ch (t 2 ) ch(t 2 )
1 1
= ec (−1, 2, t, −u − v, z) − ecy (−1, 2, t, u + v, z)
4 y 4
1 1
− ecy (−1, 2, t, −u + v, z) + ec (−1, 2, t, u − v, z)
4 4 y
√ √ √  √ √
2 sh(u 2 ) ch(v 2 ) z 2 ch(t 2 )

escsy (u, v, t, z) := L−1 √ exp − √
γ sh2 (t 2 ) sh(t 2 )
1 1
= es (1, 2, t, −u − v, z) − esy (1, 2, t, u + v, z)
4 y 4
1 1
+ esy (1, 2, t, −u + v, z) − es (1, 2, t, u − v, z)
4 4 y
For ν ≥ −1, t + νt + r + z > 0, t>0
√ √ √   √ 

2  √
z 2 ch(t 2 ) 2x 2
isy (ν, t, r, z, x) := L−1 √ exp −r 2γ − √ Iν √
γ sh(t 2 ) sh(t 2 ) sh(t 2 )
∞
P x +2l
= es (1 + ν + 2l, 1 + ν + 2l, t, r, z)
l=0
( + l + 1)l! y

For Re z > 0
2=4y Z ∞
√ (z) = ze√
2
iy (z) := L−1 e−z ch u−u /4y sh u sin(πu/2y)du


I
γ γ
 y 0

For µ ≥ η

kiy (µ, η) := L−1

γ K√γ (µ)I√γ (η)

e =4y ∞ K1 (2 + 2 + 2 ch u)1=2 −u2/4y

2 Z

= √ e sh u sin(πu/2y)du
 y 0 2 + 2 + 2 ch u 1=2


 SPECIAL FUNCTIONS 655

For µ ≤ η kiy (µ, η) := kiy (η, µ)

For Re µ > −3/2, Re z > 0

 (√ +  + 1=2) 
my (µ, z) := L−1 √ M−µ,√γ (2z)
γ (2 + 1)

8z 3=2  + 32 e =4y Z ∞ −z ch(2u)−u2/y 3

2
u



M − µ, , 2z sh2 u sh(2u) sin


= √ e du
 2y 0 2 y

For ν > 0, t+z >0

√  √ √   √

2 z 2 ch(t 2 ) x 2

eiy (ν, t, z, x) := L−1 √ exp − √ iν √
γ sh(t 2 ) sh(t 2 ) sh(t 2 )

xe =4 ∞ −u2/4ν

2 Z
= √ e eey (2, t, z, x ch u) sh u sin(πu/2ν)du
  0

Properties:
√
my (0, z) = 2πz iy (z)

2 (2z )− −z
lim e−(µ+1/2) y
my (µ, z) = e , −3/2 < µ < −1/2
y→∞ (−2 − 1)
 2
√ √
1  + 2 
  
L−1 ki (µ γ, η γ) = exp − iz
γ z
2y 4y 2y
√
L−1


γ isr ( γ, t, 0, z, x) = eir (y, t, z, 2x)

15. Two parameter functions connected to the Bessel functions

Sν (x, y) := (xy)−ν Iν (x)Kν (y) − Kν (x)Iν (y)

Cν (x, y) := (xy)−ν Iν+1 (x)Kν (y) + Kν+1 (x)Iν (y)

Fν (x, z) := (xz)−ν Iν ((x + z − |x − z|)/2)Kν ((x + z + |x − z|)/2)

(xz)−ν Iν (x)Kν (z), x≤z


=
(xz)−ν Kν (x)Iν (z), z≤x

Properties:
Cν (x, y) as a function of y and Sν (x, y), Fν (x, z) as functions of both variables
satisfy the Bessel equation
2 + 1
(B) Z 00 (x) + Z 0 (x) − Z(x) = 0.
x

Sν (x, x) = 0, Cν (x, x) = x−1−2ν , Sν (x, y) = −Sν (y, x), Fν (x, z) = Fν (z, x)

656 APPENDIX 2
@ @ @
S (x, y) = Cν (x, y), S (x, y) = −Cν (y, x), C (x, y) = −xySν+1 (x, y)
@x ν @y ν @y ν
@ 2ν+1 @ @
F (z + 0, z) − Fν (z − 0, z) = −z −1−2ν


x Cν (x, y) = x2ν+1 Sν (x, y),
@x @x ν @x
Sν (q, p)Sν (r, z) + Sν (q, r)Sν (z, p) = Sν (q, z)Sν (r, p)

Sν (q, r)Cν (r, p) + Cν (r, q)Sν (r, p) = r−1−2ν Sν (q, p)

2νSν (x, y) + xCν (x, y) = y −1 Cν−1 (y, x)

lim S0 (xθ, yθ) = ln(x/y)

θ↓0

1 −2ν
lim lim θ2ν Sν (xθ, yθ) = (y − x−2ν ), ν 6= 0
θ↓0 θ↓0 2
√
√ √ √
e−(x+y)  √
lim ν(ν 2 γ)ν Sν √λ ν γ ex/ν , ν γ ey/ν = √
λ
sh((x − y) λ + γ)
ν→∞ +
√ √ √

lim (ν 2 γ)ν λ+1/2 Cν √λ ν γ ex/ν , ν γ ey/ν
ν→∞
√
e−(x+y)  √ √ √ √

= √ λ + γ ch((x − y) λ + γ) − λ sh((x − y) λ + γ)
+
Special cases:
S−1/2 (x, y) = sh(x − y), C−1/2 (x, y) = ch(x − y)
1 1
S1/2 (x, y) = sh(x − y), C1/2 (x, y) = 2 (x ch(x − y) − sh(x − y))
xy x y
1 1 −|x−z|
F−1/2 (x, z) = (e−|x−z| + e−(x+z) ), F1/2 (x, z) = (e − e−(x+z) )
2 2xz
16. Two parameter functions connected
to the parabolic cylinder functions
( ) 2
+y 2 )/4


S(ν, x, y) := e(x D−ν (−x)D−ν (y) − D−ν (x)D−ν (−y) , ν>0

( + 1) 2
+y 2 )/4


C(ν, x, y) := e(x D−ν−1 (−x)D−ν (y) + D−ν−1 (x)D−ν (−y)

( ) 2
+y 2 )/4
F (ν, x, z) := e(x D−ν (−(x + z − |x − z|)/2)D−ν ((x + z + |x − z|)/2)
  ( ) 2 2
e(x +z )/4 D−ν (−x)D−ν (z), x ≤ z


=
 ( ) (x2 +z2 )/4
e D−ν (−z)D−ν (x), z ≤ x

Properties:
C(ν, x, y) as a function of y and S(ν, x, y), F (ν, x, y) as functions of both variables
x, y satisfy the Weber equation
(W) Z 00 (x) − xZ 0 (x) − νZ(x) = 0.
√
2 2
S(ν, x, x) = 0, S(ν, x, y) = −S(ν, y, x), C(ν, x, x) = √ ex /2

 SPECIAL FUNCTIONS 657
@ @
S(ν, x, y) = C(ν, x, y), S(ν, x, y) = −C(ν, y, x)
@x @y
@ @ −1 −x2/2 2
C(ν, x, y) = e−x /2 S(ν, x, y)


C(ν, x, y) = −νS(ν + 1, x, y), ν e
@y @x
√
@ @ 2 2
F (ν, z + 0, z) − F (ν, z − 0, z) = − √ ez /2
@x @x 

S(ν, q, p)S(ν, r, z) + S(ν, q, r)S(ν, z, p) = S(ν, q, z)S(ν, r, p)

√
2 2
S(ν, q, r)C(ν, r, p) + C(ν, r, q)S(ν, r, p) = √ er /2
S(ν, q, p)

√ √ √
2 √
lim √ S , x 2θ, y 2θ = √ sh((x − y) α)
θ↓0  2 
√
√ √
2 √
lim C , x 2θ, y 2θ = √ ch((x − y) α)
θ↓0 2 

Special cases:
x y
   
S(0, x, y) := lim S(ν, x, y) = Erfi √ − Erfi √ =: Erfid(x, y)
ν→0 2 2
 2 2 √
xy (x2 +y 2 )/4 x y 2 2
S(1/2, x, y) = √ e S1/4 , , C(0, x, y) = √ ex /2
2 2 4 4 

Properties:
Erfid(x, y) as a function of both variables satisfies the equation (W) when ν = 0.

17. Two parameter functions connected to the Kummer functions

(a)

S(a, b, x, y) := M (a, b, x)U (a, b, y) − U (a, b, x)M (a, b, y)
(b)
(a + 1)

C(a, b, x, y) := M (a + 1, b + 1, x)U (a, b, y) + bU (a + 1, b + 1, x)M (a, b, y)
(b + 1)
(a)

 (b) M (a, b, x)U (a, b, z), 0 ≤ x ≤ z

F (a, b, x, z) :=
 (a) M (a, b, z)U (a, b, x), z ≤ x

(b)
Properties:
C(a, b, x, y) as a function of y and S(a, b, x, y), F (a, b, x, y) as functions of both
variables x, y satisfy the Kummer equation

(K) xZ 00 (x) + (b − x)Z 0 (x) − aZ(x) = 0.

S(a, b, x, x) = 0, S(a, b, x, y) = −S(a, b, y, x), C(a, b, x, x) = x−b ex

658 APPENDIX 2
@ @
S(a, b, x, y) = C(a, b, x, y), S(a, b, x, y) = −C(a, b, y, x)
@x @y

@
C(a, b, x, y) = −aS(a + 1, b + 1, x, y)
@y

@ −1 b −x
a x e C(a, b, x, y) = xb−1 e−x S(a, b, x, y)


@x
@ @
F (a, b, z + 0, z) − F (a, b, z − 0, z) = −z −b ez
@x @x

S(a, b, q, p)S(a, b, r, z) + S(a, b, q, r)S(a, b, z, p) = S(a, b, q, z)S(a, b, r, p)

S(a, b, q, r)C(a, b, r, p) + C(a, b, r, q)S(a, b, r, p) = r−b er S(a, b, q, p)

√ √
lim (θ/α)ν S , ν + 1, xθ, yθ = 2Sν ( αx, αy)
θ↓0 4

1 √ √
lim (θ/α)ν+1 C , ν + 1, xθ, yθ = √ Cν ( αx, αy)
θ↓0 4 x

Special cases: for x > 0, y > 0

Z x=√2
x2 y 2 x2 y 2
    2
S 0, b, , := lim S a, b, , = 2 √ u1−2b eu du =: Ed(b, x, y)
2 2 a→0 2 2 y= 2

a 1 x2 y2 √ 1
 
S ν + , 2ν + 1, 2x, 2y = 2−2ν ex+y Sν (x, y)


S , , , = πS(a, x, y),
2 2 2 2 2
√
Ed(1/2, x, y) = π Erfid(x, y), C(0, b, x, y) = x−b ex

Properties:
√ √
Ed(b, 2x, 2y) as a function of variables x, y satisfies the equation (K) when
a = 0.
 APPENDIX 3

INVERSE LAPLACE TRANSFORMS

General formulae
Z ∞
L−1 e−γy f (y)dy,


γ F (γ) =: f (y), where F (γ) = Re γ ≥ σ ≥ 0
0
1 c+i∞ γy
Z
0. f (y) = e F (γ)dγ, c>σ
2i c−i∞

1 y

a. L−1 F (αγ + β) = e−βy/α f


γ , α>0

b. L−1 e−βγ F (γ) = f (y − β)1I[β,∞) (y),

γ β>0

Z ∞ 
1
c. L−1 F (x)dx = f (y)
γ y

Z y
F( )
 
d. L−1
γ = f (x)dx
0

e. L−1 γF (γ) − f (+0) = f 0 (y)

γ

Zy Z y
f. L−1
γ F1 (γ)F2 (γ) = f1 (x)f2 (y − x)dx = f1 (y − x)f2 (x)dx = f1 (y) ∗ f2 (y)
0 0

√
1 ∞
Z
2
g. L−1 F ( γ) = √ xe−x /4y
f (x)dx
γ 2 y3=2 0

√
√
2 ∞ x

Z
2
i. L−1 γ µ F ( γ) = √ e−x /8y D2µ+1 √ f (x)dx
γ (2y)+1 0 2y

L−1 −1


j. γ Lη F1 (γ + η)F2 (η) = f1 (y)f2 (g − y)1I[0,g] (y)

|y − qg | |y − pg | 1I[(p∧q)g;(p∨q)g] (y )
   
k. L−1 −1


γ Lη F1 (pγ + η)F2 (qγ + η) = f1
|p − q |
f2
|q − p| |q − p|

−1 −1
l. L−1


γ Lη Lλ F1 (γ + η + λ)F2 (η + λ)F3 (λ)

= f1 (y)f2 (g − y)f3 (t − g)1I[0,g] (y)1I[y,t] (g)

659
660 APPENDIX 3

1 
1. L−1 = e−βy
γ +
√ √
√
2. L−1 e− α γ
= √ 3=2 e−α/4y , α>0
γ 2 y
√ √ √
1 
y

3. L−1 γ e− α γ
= √ − e−α/4y , α>0
γ y5=2 4 2
√ √
√  
4. L−1 γ e− α γ
= √ − 3 e−α/4y , α>0
γ 4  y5=2 2y

1 √ √  1
5. L−1 √ e− e−α/4y ,
α γ
= √ α≥0
γ y
√ 

1 −√α√γ 
6. L−1 e = Erfc √ , α≥0
γ 2 y
√ √ 

1 −√α√γ  2 y −α/4y √
7. L−1 e = √ e − α Erfc , α≥0
γ 3=2  2√y
√ √
 √ √  2  
8. L−1 γ µ/2 e− α γ = √ =2+1 e−α/8y Dµ+1 √ , α>0
γ  (2y) 2y
√ √ √ √ √ √
e− e y+ √ e y− √
   
9. L−1 = Erfc √ + βy + Erfc √ − βy
γ − 2 2 y 2 2 y

1 √ 1 2
 2 √  
10. L−1 √ e−α γ = √ e−α /4y − βeαβ+β y Erfc √ + β y , α≥0
γ + y 2 y

1 √  2
 √ 
11. L−1 √ √ e−α γ
= eαβ+β y
Erfc √ +β y , α≥0
γ ( + ) 2 y
 √    2
 √ 
12. L−1 √ e−α γ
= Erfc √ − eαβ+β y
Erfc √ +β y , α≥0
γ ( + ) 2 y 2 y
1 e− y − e− y
 
13. L−1 √ √ = √
γ + + + 2 ( − )y3=2
√
+ 
14. L−1


γ √ − 1 = α I0 (αy) + I1 (αy) , α>0
−
√
√
15. L−1


eα/γ − 1 = √ I1 (2 αy), α>0
γ y

1 
y
µ/2 √
16. L−1
γ +1 eα/γ
= Iµ (2 αy), α > 0 µ > −1
√
| | e− y−
 2− 2   µ/2 p
17. L−1 2− 2 2 − 2
 = y + Iµ (|β| y 2 −α2 )1I(α,∞) (y), α≥0
γ
+
p p

√
21=2− √ (y − ) −1
18. L−1 D−2ν (2 αγ) = 1I (y), α > 0, ν > 0
γ ( )(y + ) +1=2 (α,∞)
√
√
19. L−1



Erfc αγ = √ 1I(α,∞) (y), α>0
γ y y −
 INVERSE LAPLACE TRANSFORMS 661
√
√ √
20. L−1 e−αγ −



παγ Erfc αγ = 1I(α,∞) (y), α>0
γ 2y3=2
√ √ √


21. L−1
γ γe−2α γ Erfc β + γ α+β > 0
n
+  2 2
= + 2√ e−β −(α+β) /(y−1)
y(y − 1)3=2 y y − 1
 2 1  2  √ o
y
+ √ 5=2 − √ 3=2 e−α /y Erfc p +√ 1I[1,∞) (y)
y 2 y y(y − 1) y−1
√ 1
22. L−1 y 1/2−µ e−α/2y Wµ−1/2,ν (α/y),


γ µ−1 K2ν (2 αγ) = √ α>0
γ 2
√ 1 −α/2y
23. L−1 γ−1/2 K2ν (2 αγ) =


e Kν (α/2y), α>0
γ 2√y
√ =2
24. L−1 e−α/4y ,


γ ν/2 Kν ( αγ) = α > 0, ν ≥ 0
γ (2y) +1
 =2 √
e
 (2 + 1) µ−1/2 √
25. L−1  M−µ,ν = y I2ν (2 αy), µ + ν > −1/2
γ ( +  + 1=2)

26. L−1 Γ(γ − β)W−γ,β+1/2 (4α) = α−β (sh(y/2))2β e−2α cth(y/2) ,

γ α>0≥β
√

1 +

 √ √
27. L−1 exp Iν = Iν ( 2αy)Iν ( 2βy), ν > −1
γ 2
 2 2  
√ √
1 x +z xz
28. L−1 I (x 2γ)K (z 2γ) = exp − Iν , 0∨x≤z
γ ν ν
2y 2y y
ey=2 (x2 + z 2 ) ch y + 2xz 

29. L−1


Γ(γ)D (x)D (z) = √ exp −
γ −γ −γ
2 sh y 4 sh y

√ −x2 /4 x
 
γ
30. L−1 −1 γ/2 1


γ 2 Γ + D (x) = π e Erfc , x≥0
γ 2 2 −γ
2(e2y − 1)
p

xey=2
 2
D− (x) x ch y
  
31. L−1 =3= √
2 exp − , x≥0
γ D− (0) 2  sh y 4 sh y

32. L−1


γ Γ(1/2 + ν + γ)M−γ,ν (x2 )W−γ,ν (z 2 ) 0≤x≤z

(2 + 1)xz 
(x2 + z 2 ) ch(y=2)   xz 
= exp − I2ν , ν > −1/2
2 sh(y=2) 2 sh(y=2) sh(y=2)
33. L−1


γ Γ(γ)M (γ, ν + 1, x2 )U (γ, ν + 1, z 2 ) 0≤x≤z

( + 1)e(+1)y=2  2 2
x +z (x2 + z 2 ) ch(y=2)   xz 
=   exp − Iν , ν > −1
2x z sh(y=2) 2 2 sh(y=2) sh(y=2)

1  p
34. L−1 −1
γ Lη = e−by−ag I0 (2 (ab − c)yg)
 + a + b + c

+b 1  p p
35. L−1 −1
γ Lη − = e−by−ag (ab − c)g/y I1 (2 (ab − c)yg)
 + a + b + c  + a
 APPENDIX 4

DIFFERENTIAL EQUATIONS AND THEIR SOLUTIONS

Below ψ and ϕ denote a nonnegative increasing and decreasing, respectively,

solutions of the differential equations, w = ψ 0 ϕ − ψϕ0 is the Wronskian,
λ and q are given positive parameters.

1 00
1. Y (x) − λY (x) = 0, x∈R
2
√ √ √
ψ(x) = ex 2λ
, ϕ(x) = e−x 2λ
, w = 2 2λ

1 00
2. Y (x) − (λ + γx)Y (x) = 0, x>0
2
√
√ 2 2

ψ(x) = λ + γxI1/3 (λ + γx)3/2 , w = 3γ/2, γ>0
3
√
√ 2 2 √
(λ + γx)3/2 = π 3 Ai 21/3 γ −2/3 (λ + γx)



ϕ(x) = λ + γxK1/3
3

1 00 2

3. Y (x) − λ + x2 Y (x) = 0, x∈R
2 2
√ √
ψ(x) = D−1/2−λ/γ (−x 2γ), ϕ(x) = D−1/2−λ/γ (x 2γ)

2√
w= , γ>0
(1=2 + = )

1 00 2 − 2−2

4. Y (x) − λ + Y (x) = 0, x>0
2 2x2
√ √
√ √
1
ψ(x) = xIγ x 2λ , ϕ(x) = xKγ x 2λ , w = 1, γ≥
2

1 00 2 − 2−2
a. Y (x) − Y (x) = 0, x>0
2 2x2
1
ψ(x) = x1/2+γ , ϕ(x) = x1/2−γ , w = 2γ, γ≥
2

1 00 p2 − 2−2 q 2 x2

5. Y (x) − λ + + Y (x) = 0, x>0
2 2x2 2
ψ(x) = x−1/2 M−λ/2q,p/2 (qx2 ), ϕ(x) = x−1/2 W−λ/2q,p/2 (qx2 )
662
 DIFFERENTIAL EQUATIONS 663
2q (1 + p) 1
w= , p≥
((1 + p + =q)=2) 2

1 00 p2 − 2−2 q2 x2

a. Y (x) − + Y (x) = 0, x>0
2 2x2 2
√ √ 1
ψ(x) = xIp/2 (qx2/2), ϕ(x) = xKp/2 (qx2/2), w = 2, p≥
2

1 00 p2 − 2−2 q

6. Y (x) − λ + + Y (x) = 0, x>0
2 2x2 x
√ √
ψ(x) = M−q/√2λ,p (2x 2λ), ϕ(x) = W−q/√2λ,p (2x 2λ)
√
2 2 (2p + 1) 1
w= √ , p≥
(p + 1=2 + q= 2) 2

1 00 p2 − 2−2 q

a. Y (x) − 2 + Y (x) = 0, x>0
2 2x x
√ √ √ √ 1 1
ψ(x) = xI2p (2 2qx), ϕ(x) = xK2p (2 2qx), w= , p≥
2 2

1 00
7. Y (x) − (λ + γe2βx )Y (x) = 0, x∈R
2
√ √
2 
2 
ψ(x) = I√2λ/|β| eβx , ϕ(x) = K√2λ/|β| eβx , w=β
| | | |

1 00 p2

8. Y (x) − λ + e2βx + qeβx Y (x) = 0, x∈R
2 2
√
−βx/2 √ 2p βx

2p (2 2=| | + 1)
ψ(x) = e M−q/p|β|, 2λ/|β| |β| e , w= √
( 2=| | + 1=2 + q=p| |)
2p βx
ϕ(x) = e−βx/2 W−q/p|β|,√2λ/|β|


|β| e , p>0

1 00
9. Y (x) + µY 0 (x) − (λ + f (x))Y (x) = 0, x∈R
2
solution Yµ,λ (x) of this equation satisfies the relation

Yµ,λ (x) = e−µx Y0,λ+µ2/2 (x)

1 00 1
10. Y (x) + Y 0 (x) − (λ + f (x))Y (x) = 0, x>0
2 x
664 APPENDIX 4

solution of this equation is x−1 Y0,λ (x), where Y0,λ (x) is the solution of the
equation 9 for µ = 0

1 2
 
11. Y 00 (x) + Y 0 (x) − 1 + 2 Y (x) = 0, x>0
x x

1
ψ(x) = Iν (x), ϕ(x) = Kν (x), w= , ν>0
x

1 00 2 + 1 0
12. Y (x) + Y (x) − λY (x) = 0, x>0
2 2x
√ √ 1
ψ(x) = x−ν Iν (x 2λ), ϕ(x) = x−ν Kν (x 2λ), w = 2 +1 , ν > −1
x

1 00 2 + 1 0
a. Y (x) + Y (x) = 0, x>0
2 2x
2
ψ(x) = 1, ϕ(x) = x−2ν , w = 2 +1 , ν>0
x

1 00 1
b. Y (x) + Y 0 (x) = 0, x>1
2 2x
1
ψ(x) = ln x, ϕ(x) = 1, w=
x

1 00 1
c. Y (x) + Y 0 (x) − λY (x) = 0, x>0
2 x
√
1 √ 1 √
2 2
ψ(x) = sh(x 2λ), ϕ(x) = e−x 2λ
, w=
x x x2

1 00 2 + 1 0
13. Y (x) + Y (x) − γxY (x) = 0, x>0
2 2x
2 3/2 √
2 3/2 √

ψ(x) = x−ν I2ν/3 x 2γ , ϕ(x) = x−ν K2ν/3 x 2γ
3 3
3
w = x−2ν−1 , ν > −1
2

1 00 2 + 1 0 2

14. Y (x) + Y (x) − λ + 2 Y (x) = 0, x>0
2 2x 2x
√
√

ψ(x) = x−ν I√ν 2+γ 2 x 2λ , ϕ(x) = x−ν K√ν 2+γ 2 x 2λ
 DIFFERENTIAL EQUATIONS 665

w = x−2ν−1 , ν≥0

1 00 2 + 1 0 2
a. Y (x) + Y (x) − 2 Y (x) = 0, x>0
2 2x 2x
√ 2 2 √ 2 2 + 2
p
ν 2+γ 2−ν
ψ(x) = x ν +γ −ν , ϕ(x) = x −
, w= , ν≥0
x2 +1

1 00 2 + 1 0 p2 q 2 x2

15. Y (x) + Y (x) − λ + 2 + Y (x) = 0, x>0
2 2x 2x 2

2q (1 + p2 +  2 )x−2 −1
p
ψ(x) = x−ν−1 M−λ/2q,√p2+ν 2 /2 (qx2 ), w=
((1 + p2 +  2 + =q)=2)
p

ϕ(x) = x−ν−1 W−λ/2q,√p2+ν 2 /2 (qx2 ), ν≥0

1 00 2 + 1 0 p2 q 2 x2

a. Y (x) + Y (x) − + Y (x) = 0, x>0
2 2x 2x 2 2

ψ(x) = x−ν I√ν 2+p2 /2 (qx2/2), w = 2x−2ν−1

ϕ(x) = x−ν K√ν 2+p2 /2 (qx2/2), ν≥0

1 00 2 + 1 0 p2 q

16. Y (x) + Y (x) − λ + 2 + Y (x) = 0, x>0
2 2x 2x x
√
1 √ 2x1−2 2 (2  2 + p2 + 1)
p
ψ(x) = +1=2 M−q/ √ √ (2x 2λ), w= p 2 2 √
x 2λ, ν 2+p2 (  + p + 1=2 + q= 2)

1 √
ϕ(x) = +1=2 W−q/√2λ,√ν 2+p2 (2x 2λ), ν≥0
x

1 00 2 + 1 0 p2 q

a. Y (x) + Y (x) − + Y (x) = 0, x>0
2 2x 2x2 x
√
ψ(x) = x−ν I2√ν 2+p2 (2 2qx), w = 2−1 x−2ν−1

√
ϕ(x) = x−ν K2√ν 2+p2 (2 2qx), ν≥0

 2
x 2 − 1 
17. Y 00 (x) − + Y (x) = 0, x∈R
4 2
666 APPENDIX 4
√
2
ψ(x) = D−ν (−x), ϕ(x) = D−ν (x), w= ν>0
( )


( 2 − 1)x2 
18. σ 2 Y 00 (x) − xY 0 (x) − λ + Y (x) = 0, x∈R
42
√ √
2 2 x
2 2 x

ψ(x) = ex /4σ D− 1 + − , ϕ(x) = ex /4σ D− 1 +
 
1
2 2γ −λ
γ 2
1
2γ −λ
γ

√
2 x 2
/2σ 2
w= 1 − 1 + 
e , σ > 0, γ≥1
 2 2

a. σ 2 Y 00 (x) − xY 0 (x) = 0, x∈R

Z x=√2 √
2 2 2 2 2
ψ(x) = √ ev dv, ϕ(x) = 1, w= √ ex /2σ , σ>0
 0 

x

19. σ 2 Y 00 (x) − xY 0 (x) − λ + Y (x) = 0, x∈R


2 2 x
2 2 x

ψ(x) = ex /4σ Dγ 2 −λ − − 2γ , ϕ(x) = ex /4σ Dγ 2 −λ + 2γ
 
√
2 x2/2σ 2
√
w=
e , σ > 0, λ>γ≥0
 − 2


(p2 − 2−2 )2 (q2 − 1)x2 
20. σ 2 Y 00 (x) − xY 0 (x) − λ + + Y (x) = 0, x>0
x2 42

qx2 q (p + 1)ex =2

  2 2
2 2
ψ(x) = x−1/2 ex /4σ M 1−2λ , p 2 , w=
2 2 p+1 2−1
2 + 4q
 
4q 2

qx2 1
2 2
 
ϕ(x) = x−1/2 ex /4σ W 1−2λ , p , p≥ , q≥1
4q 2 22 2

 2
 (2 + 1) p2 2 (q2 − 1) 2
  
21. σ 2 Y 00 (x)+ −x Y 0 (x)− λ+ 2 + x Y (x) = 0, x>0
x x 42

qx2
 
ψ(x) = x−ν−1 ex /4σ M ν+1−λ , 1 √ν 2+p2
2 2

2q 2 2 2

qx2
 
ϕ(x) = x−ν−1 ex /4σ W ν+1−λ , 1 √ν 2+p2
2 2

2q 2 22
 DIFFERENTIAL EQUATIONS 667

qex =2  2 + p2 + 1
2 2 p

w= , q ≥ 1, ν≥0
2 x2+1 1 2 2 1 −−1
2  + p + 2 + 2q
 p

1 00 
2 + 1 
22. Y (x) + − γx Y 0 (x) − λY (x) = 0, x>0
2 2x
for γ > 0, ν > −1
λ λ



ψ(x) = M 2γ , ν + 1, γx2 , ϕ(x) = U 2γ , ν + 1, γx2

2 ( + 1)x−2 −1 γx2
w= e
(=2 ) 
for γ > 0, ν < 0

ψ(x) = x−2ν M λ
ϕ(x) = x−2ν U λ
2

2


2γ −ν, 1−ν, γx , 2γ −ν, 1−ν, γx

2 (1 −  )x−2 −1 γx2
w= e
(=2 −  ) 
for γ < 0, ν > −1
2 2
λ λ



ψ(x) = eγx M ν +1− 2γ , ν +1, −γx2 , ϕ(x) = eγx U ν +1− 2γ , ν +1, −γx2

2 ( + 1)x−2 −1 −γx2
w= e
( + 1 − =2 )| |
for γ < 0, ν < 0
2 2
ψ(x) = x−2ν eγx M 1− 2γ
λ
, 1−ν, −γx2 , ϕ(x) = x−2ν eγx U 1− 2γ
λ



, 1−ν, −γx2

2 (1 −  )x−2 −1 −γx2
w= e
(1 − =2 )| |

1 00 1
23. Y (x) + ρ cth(x)Y 0 (x) − (µ2 − ρ2 )Y (x) = 0, x>0
2 2
P−−1+1 =2
=2 (ch x)
e −1=2 (ch x)
Q −1=2 1 1
ψ(x) = , ϕ(x) = , ω= , µ≥ρ>−
sh−1=2 x sh−1=2 x sh2 x 2

1 00
24. Y (x) − ρ th(x)Y 0 (x) − 2α(ρ + α)Y (x) = 0, x>0
2
ψ(x) = chρ x Peρ−2α−ρ (− th x) + Peρ−2α−ρ (th x) ,


α > 0, α + ρ ≥ 0

2 ch2 x
ϕ(x) = chρ x Peρ−2α−ρ (th x), ω=
(2 ) (2 + 2 + 1)

1 00
25. Y (x) + cth x − ρ th x)Y 0 (x) − 2α(ρ + α − 1)Y (x) = 0, x>0
2
668 APPENDIX 4
ch x e1−2α−ρ

ψ(x) = P (− th x) − Peρ1−2α−ρ (th x) , α > 0, α + ρ ≥ 1
sh x ρ
ch x e1−2α−ρ 2 ch2 x
ϕ(x) = P (th x), ω= 2
sh x ρ sh x (2 − 1) (2 + 2)

1 00 1
1

th x Y 0 (x) − 2α(β − γ)Y (x) = 0


26. Y (x) + γ− cth x − β − α −
2 2 2
1 1
F α, β, γ, th2 x , G α, β, γ, th2 x



ψ(x) = ϕ(x) =
ch2 x ch2 x
2 ( ) ch2( − −1=2) x
w= , α > 0, β ≥ γ > 0
( ) ( ) sh2 −1 x

For the differential equations in 27-29 it is assumed that αβ > 0, α + β ≥ 0,

α + β + 1 > γ > 0, and α, β can be complex conjugate.

27. x(1 − x)Y 00 (x) + (γ − (α + β + 1)x)Y 0 (x) − αβY (x) = 0, 0<x<1

( )(1 − x) − − −1
ψ(x) = F (α, β, γ, x), ϕ(x) = G(α, β, γ, x), ω=
x ( ) ( )

+ −1 ex
 
28. Y 00 (x) + 1 − γ + Y 0 (x) −
Y (x) = 0, x∈R
ex + 1 ex + 1 2

1  
1 
ψ(x) = G α, β, γ, x , ϕ(x) = F α, β, γ, x
e +1 e +1

( ) (ex + 1) + −1
w=
( ) ( ) ex( + − )

e −x
 
29. Y 00 (x) + − α − β Y 0 (x) − Y (x) = 0, x>0
1 − e−x 1 − e−x

ψ(x) = F α, β, γ, 1 − e−x , ϕ(x) = G α, β, γ, 1 − e−x

( ) ex( + − )
w=
( ) ( ) (1 − e−x )
 APPENDIX 5

FORMULAE FOR n-FOLD DIFFERENTIATION

For the n times differentiable real valued function f introduce

dn
(f (x))(n) := f (n) (x) := f (x), n = 1, 2, 3, . . . , f (0) (x) := f (x).
dxn

1. Leibnitz’s formula for the nth derivative of the product of two functions:
n

(n) X n!
f (x)g(x) = f (k) (x) g (n−k) (x).
k ! (n − k ) !
k=0

2. de Bruno’s formula for the nth derivative of the composition of two functions:
n n

(n) X X Y 1  f (j ) (x) kj
F (f (x)) = n! F (m) (f (x)) .
kj ! j!
m=1 k1 +2k2 +···+nkn =n j=1
k1 +k2 +···+kn =m

3. The nth derivative of the ratio of two functions:

f (x) f (n) (x) (−1)m m ! n ! f (n−l) (x)

 (n) X
= +
g(x) g(x) (n − l) ! gm+1 (x)
1≤m≤l≤n
l
X Y 1  g(j ) (x) kj
× .
kj ! j!
k1 +2k2 +···+lkl =l j=1
k1 +k2 +···+kl =m

4. The (n + 1)th derivative of the inverse function (see Bödewadt (1942)): Let F
be a given smooth function with the inverse f. Then

1  F (j+1) (f (x)) kj

n n
(n+1)
X (−1)m (n + m) ! X Y
f (x) = .
(F 0 (f (x)))n+m+1 kj ! (j + 1) !
m=1 k1 +2k2 +···+nkn =n j=1
k1 +k2 +···+kn =m

The second sum in the formulae above is taken over all combinations of nonnegative
integers k1 , k2 , . . . such that the indicated equalities hold.
5. The nth derivative of the composition of two functions (see Adams (1947)):
n m

(n) X X (−1)l
(n) m−l
F (f (x)) = (−1)m F (m) (f (x)) f l (x) f (x).
l ! (m − l ) !
m=1 l=1

669
670 APPENDIX 5

6. Examples:

n−1
1


(n) X (−1)n (n − 1) ! n! F (n−k) (1=x)
F =
x k ! (n − 1 − k ) ! ( n − k ) ! x2n−k
k=0

n−1

(n) X (−1)n (n − 1) ! n ! an−k ea=x
ea/x =
k ! (n − 1 − k) ! (n − k) ! x2n−k
k=0

(n) X n! F (n−k) (x2 )
F (x2 ) =
k ! (n − 2k) ! (2x)2k−n
0≤k≤n/2

2
(n) 2 X n!
eax = eax an−k (2x)n−2k
k ! (n − 2k) !
0≤k≤n/2

n−1 √
√
(n) X (−1)k (n − 1 + k) ! F (n−k) ( x)
F ( x) = √ n+k
k ! (n − 1 − k ) ! (2 x)
k=0

c + dx
(n)
(cb − ad) bn−1
= (−1)n n !
a + bx (a + bx)n+1
√
(cb − ad) bn−k−1
n−1

c+d x
(n) X (−1)n (n − k) (n − 1 + k) !
√ = √ √ .
a+b x k! (a + b x)n−k+1 (2 x)n+k
k=0
 BIBLIOGRAPHY

Abramowitz, M., and Stegun I.A., Mathematical Functions, Dover Publications, Inc., New York,
1970 (9th printing).
Adams, E.P., Smithsonian Mathematical Formulae and Tables of Elliptic Functions, Smithsonian
Institution, Washington, 1947.
Alili, L., Dufresne, D., and Yor, M., Sur l'identite de Bougerol pour les fonctionnelles expo-
nentielles du mouvement brownien avec drift, in \Exponential Functionals and Principal
Values related to Brownian Motion; A collection of research papers", Revista Matematica
Iberoamericana, Madrid, 1997, pp. 3{14.
Alili, L., and Gruet J-C., An explanation of a generalized Bougerol's identity in terms of hyper-
bolic Brownian motion, in \Exponential Functionals and Principal Values related to Brown-
ian Motion; A collection of research papers", Revista Matematica Iberoamericana, Madrid,
1997, pp. 15{33.
Amir, M., Sticky Brownian motion as the strong limit of a sequence of random walks, Stoch.
Proc. Appl. 39 (1991), 221{237.
Bachelier, L., Theorie de la speculation, Ann. Sci. E cole Norm. Sup. 17 (1900), 21{86.
Balkema, A.A., and Chung, K.L., Paul Levy's way to his local time, in \Seminar on Stochastic
Processes, 1990", Birkhauser Verlag, Boston, Basel, and Berlin, 1991, pp. 5{14.
Barlow, M., One-dimensional stochastic di erential equation with no strong solution, J. London
Math. Soc. 26 (1982), 335{345.
Barlow, M., Skew Brownian motion and a one-dimensional stochastic di erential equation, Stoch-
astics 25 (1988), 1{2.
Bass, R.F., Probabilistic Techniques in Analysis, Springer Verlag, Berlin, Heidelberg, and New
York, 1995.
Bass, R.F., and Khoshnevisan, D., Laws of the iterated logarithm for local times of the empirical
processes, Ann. Probab. 23 (1995), 388{399.
Bertoin, J., Excursions of a BES0 (d) and its drift term (0 < d < 1), Probab. Th. Rel. Fields 84
(1990), 251{265.
Bertoin, J., An extension of Pitman's theorem for spectrally positive Levy processes, Ann. Probab.
20 (1992), no. 3, 1464{1483.

Bertoin, J., and Pitman, J., Path transformations connecting Brownian bridge, excursion and
meander, Bull. Sci. Math. 118 (1994), 147{166.
Biane, Ph., Relations entre pont brownien et du mouvement Brownien reel, Ann. Inst. H. Poincare
22 (1986), 1{7.

Biane, Ph., Decompositions of Brownian trajectories and some applications, in \Probability and
Statistics; Rencontres Franco{Chinoises en Probabilites et Statistiques. Proceedings of the
Wuhan meeting", World Scienti c, 1993, pp. 51{76.
Biane, Ph., and Yor, M., Valeurs principales associees aux temps locaux browniens, Bull. Sci.
Math. 111 (1987), 23{101.
Biane, Ph., and Yor, M., Quelques precisions sur le meandre Brownien, Bull. Sci. Math. 112
(1988), 101{109.
Bismut, J.M., Last exit decomposition and regularity at the boundary of transition probabilities,
Z. Wahrscheinlichkeitstheorie verw. Gebiete 69 (1985), 65{98.

© Springer Basel 2015 671

A.N. Borodin, P. Salminen, Handbook of Brownian Motion - Facts and Formulae,
Probability and Its Applications, DOI 10.1007/978-3-0348-8163-0
672 BIBLIOGRAPHY

Blumenthal, R.M., Excursions of Markov Processes, Birkhauser, Boston, Basel, and Berlin, 1992.
Blumenthal, R.M., and Getoor, R.K., Markov Processes and Potential Theory, Academic Press,
New York and London, 1968.
Bodewadt, U.T., Die Kettenregel fur Hohere Ableitungen, Math. Zeitschrift 48 (1942), 735{746.
Borodin, A.N., Distribution of integral functionals of Brownian motion, Zapiski Nauchn. Semin.
LOMI 119 (1982), 19{38; English transl. in J. Soviet Math.
Borodin, A.N., On distribution of supremum of increments of Brownian local time, Zapiski
Nauchn. Semin. LOMI 142 (1985), 6{24; English transl. in J. Soviet Math.
Borodin, A.N., Brownian local time, Uspekhi Mat. Nauk 44 (1989a), no. 2, 7{48; English transl.
in Russian Math. Surveys.
Borodin, A.N., Distributions of functionals of the Brownian local time I and II, Th. Probab.
Appl. 34 (1989b), 385{401, 576{590.
Borodin, A.N., On distribution of functionals of the Brownian motion stopped at the moment
inverse to local time, Zapiski Nauchn. Semin. LOMI 184 (1990), 37{61; English transl. in
J. Soviet Math.
Borodin, A.N., On distribution of Brownian motion stopped at the moment inverse to the range
of the process, Zapiski Nauchn. Semin. POMI 260 (1999), 50{72; English transl. in J. Math.
Sciences.
Borodin, A.N., On distributions of supremum of linear combinations of local times of random
processes, Trans. St. Petersburg Math. Soc. 8 (2000), 5{28; English transl. in Transl., Ser 2,
AMS Russian.
Borodin, A.N., Stochastic Processes, Lan', St. Petersburg, 2013. (Russian)
Borodin, A.N., and Salminen, P., Some exponential integral functionals of BM() and BES(3),
Zapiski Nauchn. Semin. LOMI 311 (2004), 51{78; English transl. in J. of Math. Sciences
133, (2006), no. 3, 1231{1248.

Bougerol, Ph., Exemple du theoremes locaux sur les groupes resolubles, Ann. Inst. H. Poincare
19 (1983), 369{391.

Bouleau, N., and Yor, M., Sur la variation quadratique des temps locaux de certaines semimartin-
gales, C.R. Acad. Sci. Paris, Serie I 292 (1981), 491{494.
Breiman, L., Probability, Addison-Wesley Publ. Co, Reading, Mass., 1968.
Bruno de, F., Note sur une nouvelle formule de calcul di erentiel, Quarterly J. Pure Appl. Math
1 (1857), 359{360.

Bulinski, A.V., and Shiryaev, A.N., Theory of Stochastic Processes, Fizmatlit, Moscow, 2005.
(Russian)
Cameron, R.H., and Martin W.T., Evaluations of various Wiener integrals by use of certain
Sturm{Liouville di erential equation, Bull. Amer. Math. Soc. 51 (1945), 73{90.
Chaumont, L., Hobson, D., and Yor, M., Some consequences of the cyclic exchangeability property
for exponential functionals of Levy processes, in \Seminaire de Probabilites, XXXV", Lec-
ture Notes in Mathematics, vol. 1755, Springer-Verlag, Berlin, Heidelberg, and New York,
2001, pp. 334{347.
Chitashvili, R., On the nonexistence of a strong solution in the boundary problem for a sticky
Brownian motion, Proc. A. Razmadze Math. Inst. 115 (1997), 17{31.
Chong, K.S., Cowan, R., and Holst, L., The ruin problem and cover times of asymmetric random
walks and Brownian motions, Adv. Appl. Probab. 32 (2000).
 BIBLIOGRAPHY 673

Chung, K.L., Excursion in Brownian motion, Arkiv for matematik 14 (1976), 155{177.
Chung, K.L., A cluster of great formulas, Acta Math. Acad. Sci. Hungar. 39 (1982a), 65{67.
Chung, K.L., Lectures from Markov Processes to Brownian Motion, Springer-Verlag, Berlin, Hei-
delberg, and New York, 1982b.
Chung, K.L., Reminiscences of some of Paul Levy's ideas in Brownian motion and in Markov
chains, in \Seminar on Stochastic Processes, 1988", Birkhauser Verlag, Boston, Basel, and
Berlin, 1989, pp. 99{108.
Chung, K.L., Green, Brown, and Probability, World Scienti c, Singapore, 1995.
Chung, K.L., Probability and Doob, Amer. Math. Monthly 105(1) (1998), 28{35.
Chung, K.L., and Getoor, R., The condenser problem, Ann. Probab. 5(1) (1977), 82{86.
Chung, K.L., and Walsh, J.B., Markov processes, Brownian Motion, and Time Symmetry, 2nd
edition, Springer-Verlag, Berlin, Heidelberg, and New York, 2005.
Chung, K.L., and Williams, R., Introduction to Stochastic Integration, 2nd edition, Birkhauser,
Boston, 1990.
Comtet, A., Monthus, C., and Yor, M., Exponential functionals of Brownian Motion and disor-
dered systems, J. Appl. Probab. 35 (1998), 255{271.
Cootner, P.H. (ed.), The Random Character of Stock Market Prices, The M.I.T. Press, Cam-
bridge, MA., 1964.
Cox, J.C., Notes on option pricing I: Constant elasticity of variance di usions, Working paper,
Stanford University, 1975, reprinted in Journal of Portfolio Management 22 (1996), 15{17.
Cox, J.C., Ingersoll, J.E., and Ross, S., A theory of the term structure of interest rates, Econo-
metrica 53 (1985), 385{407.
Csaki, E., and Foldes, A., How small are the increments of the local time of a Wiener process,
Ann. Probab. 14 (1986), 533{546.
Csaki, E., Foldes, A., and Salminen, P., On the joint distribution of the maximum and its location
for a linear di usion, Ann. Inst. H. Poincare 23 (1987), 179{194.
Csaki, E., and Salminen, P., On additive functionals of di usion processes, Studia Sci. Math.
Hungar. 31 (1996), 47{62.
Darling, D.A., and Siegert, A.J.F., The rst passage problem for a continuous Markov process,
Ann. Math. Statistics 24 (1953), 624{639.
Davydov, D., and Linetsky, V., Pricing and hedging path-dependent options under the SEV
process, Management Science 47 (2001), no. 7, 949{965.
Decamps, M., Schepper de, A., Goovaerts, M. and Schoutens, W., A note on some new perpetu-
ities, Scandinavian Actuarial J. 4 (2005), 261{270.
Dellacherie, C., and Meyer, P.A., Probabilites et Potentiel I{IV, Hermann, Paris, 1976, 1980,
1983, 1987.
Donati-Martin, C., Matsumoto, H., and Yor, M., On positive and negative moments of the integral
of geometric Brownian motion, Stat. Probab. Letters 49 (2000a), 45{52.
Donati-Martin, C., Matsumoto, H., and Yor, M., On striking identities about the exponential func-
tionals of the Brownian bridge and Brownian motion, Periodica Math. Hung. 41 (2000b),
103{119.
Donati-Martin, C., Matsumoto, H., and Yor, M., The law of geometric Brownian motion and its
integral, revisited; application to conditional moments, in \Mathematical Finance, Bachelier
674 BIBLIOGRAPHY

Congress, 2000(Paris),, Springer Verlag, Berlin, Heidelberg, and New York, 2002, pp. 221{
243.
Doney, R., and Yor, M., On the formula of Takacs for Brownian motion with drift, J. Appl.
Probab. 35 (1998), 272{280.
Doob, J.L., Heuristic approach to the Kolmogorov{Smirnov theorems, Ann. Math. Stat. 20
(1949), 393{403.
Doob, J.L., Stochastic Processes, Wiley and Sons, New York, 1953.
Doob, J.L., Wiener's work in probability theory, Bull. Amer. Math. Soc. 72(1), part II (1966),
69{72.
Doob, J.L., Classical Potential Theory and Its Probabilistic Counterpart, Springer-Verlag, Berlin,
Heidelberg, and New York, 1984.
Dufresne, D., The distribution of a perpetuity, with applications to risk theory and pension fund-
ing, Scand. Actuarial J. 1-2 (1990), 39{79.
Dufresne, D., Laguerre series for Asian and other options, Report No. 69, Center for Actuarial
Studies, University of Melbourne, Australia (1999).
Durrett, R., Brownian Motion and Martingales in Analysis, Wadsworth Inc., Belmont, California,
1984.
Durrett, R., Stochastic Calculus: A Practical Introduction, CRC Press, Boca Raton, Florida,
1996.
Durrett, R., and Iglehart, D.L., Functionals of Brownian meander and Brownian excursion, Ann.
Probab. 5 (1977), 130{135.
Dynkin, E.B., Markov Processes I, II, Springer-Verlag, Berlin, Heidelberg, and New York, 1965.
Einstein, A., On the movement of small particles suspended in a stationary liquid demanded by
the molecular-kinetic theory of heat, Ann. Physik 17 (1905), 549{560.
Eisenbaum, N., Un theoreme de Ray{Knight lie au supremum des temps locaux browniens,
Probab. Th. Rel. Fields 87 (1990), 79{95.
Emanuel, D., and MacBeth, J., Further results on the constant elasticity of variance call option
pricing model, J. Financial and Quant. Anal. 17 (1982), 533{554.
Engelbert, H.J., and Schmidt, W., On one-dimensional stochastic di erential equations with gen-
eralized drift, in \Stochastic Di erential Systems", Lecture notes in Control and Information
Sciences, Vol. 69, Springer-Verlag, Berlin, Heidelberg, and New York, 1984, pp. 143{155.
Engelbert, H.J., and Schmidt, W., On solutions of stochastic di erential equations without drift,
Z. Wahrscheinlichkeitstheorie verw. Gebiete 68 (1985), 287{317.
Engelbert, H.J., and Schmidt, W., Strong Markov continuous local martingales and solutions
of one-dimensional stochastic di erential equations (Part III), Math. Nachr. 151 (1991),
149{197.
Engelbert, H.J., and Senf, T., On functionals of Wiener process with drift and exponential local
martingales, Dozzi, M., Engelbert, H.J., and Nualart, D. (ed.), Stochastic processes and
related topics. Proc. Wintersch. Stochastic Processes, Optim. Control, Georgenthal/Ger.
1990, N 61 in Math. Res., Academic Verlag, Berlin, 1991, pp. 45{58.
Erdelyi, A., et al., Higher Transcendental Functions, I, II, III, McGraw-Hill, New York, 1953.
Erdelyi, A., et al., Tables of Integral Transforms, I, II, McGraw-Hill, New York, 1954.
Ethier, S.N., and Kurtz, T.G., Markov Processes. Characterization and Convergence, Wiley and
Sons, New York, 1986.
 BIBLIOGRAPHY 675

Feller, W., Zur Theorie der stochastischen Prozesse, Math. Ann. 113 (1936), 113{160.
Feller, W., The asymptotic distribution of the range of sums of independent random variables,
Ann. Math. Stat. 22 (1951), no. 3, 427{432.
Fitzsimmons, P.J., and Pitman, J.W., Kac's moment formula and the Feynman{Kac formula for
additive functionals of a Markov process, Stoch. Proc. Appl. 79 (1999), 117-134.
Follmer, H., Protter, P., and Shiryaev, A.N., Quadratic variation and an extension of It^o's
formula, Bernoulli 1 (1995), 149{169.
Follmer, H., and Protter, P., On It^o's formula for multidimensional Brownian motion, Probab.
Theory Relat. Fiedds 116 (2000), 1{20.
Freedman, D., Brownian Motion and Di usion, Holden-Day, San Francisco, 1971.
Geman, D., and Horowitz, J., Occupation densities, Ann. Probab. 8 (1980), 1{67.
Geman, H., and Yor, M., Bessel processes, Asian options and perpetuities, Mathematical Finance
3 (1993), 349{375.

Getoor, R.K., The Brownian escape process, Ann. Probab. 7 (1979a), 864{867.
Getoor, R.K., Excursions of a Markov process, Ann. Probab. 8 (1979b), 244{266.
Getoor, R.K., Transience and recurrence of Markov processes, in \Seminaire de Probabilites, XIV,
1978/79", Lecture Notes in Mathematics, vol. 784, Springer Verlag, Berlin, Heidelberg, and
New York, 1980, pp. 397{409.
Getoor, R.K., and Sharpe, M., Excursions of Brownian motion and Bessel processes, Z. Wahr-
scheinlichkeitstheorie verw. Gebiete. 47 (1979), 83{106.
Getoor, R.K., and Sharpe, M., Excursions of dual processes, Adv. in Math. 45 (1982), 259{309.
Gihman, I.I., and Skorohod, A.V., Theory of Stochastic Processes I, II, III, Springer-Verlag,
Berlin, Heidelberg and New York, 1974, 1975, 1979.
Girsanov, I.V., On transforming a certain class of stochastic processes by absolutely continuous
substitution of measures, Th. Probab. Appl. 5 (1960), 285{301.
Goldenberg, D., A uni ed method for pricing options on di usion processes, J. Financial Econom.
29 (1991), 3{34.

Graversen, S.-E., and Shiryaev, A.N., An extension of P. Levy's distributional properties to the
case of a Brownian motion, Bernoulli 6 (2000), 615{620.
Gruet, J.-C., Windings of hyperbolic Brownian motion, Yor, M. (ed.), Exponential Functionals
and Principal Values related to Brownian Motion, Revista Matematica Iberoamericana,
Madrid, 1997, pp. 35{72.
Harrison, J.M., Brownian Motion and Stochastic Flow Systems, Wiley, New York, 1985.
Harrison, J.M., and Lemoine, A.J., Sticky Brownian motion as the limit of storage processes, J.
Appl. Probab. 18 (1981), 216{226.
Harrison, J.M., and Shepp, L.A., On skew Brownian motion, Ann. Probab. 9 (1981), 309{313.
Ikeda, N., Kusuoka, S., and Manabe, S., Levy's stochastic area formula for Gaussian processes,
Comm. Pure App. Math. XLVII (1994), 329{360.
Ikeda, N., and Watanabe, S., Stochastic Di erential Equations and Di usion Processes, North-
Holland Publ. Co. and Kodansha Ltd, Amsterdam, Oxford, New York, and Tokyo, 1981.
Imhof, J.-P., Density factorization for Brownian motion, meander and the three-dimensional
Bessel process, and applications, J. Appl. Probab. 21 (1984), 500{510.
676 BIBLIOGRAPHY

Imhof, J.-P., On the time spent above a level by Brownian motion with negative drift, Adv. Appl.
Probab. 18 (1986), 1017{1018.
It^o, K., Stochastic integral, Proc. Imperial Acad. Tokyo 20 (1944), 519{524.
It^o, K., Poisson point processes attached to Markov processes, in \Proc. of the sixth Berkeley
Symposium in Mathematical Statistics and Probability, vol. 3", University of California,
Berkeley, 1970, pp. 225{239.
It^o, K., and McKean, Jr., H.P., Di usion Processes and Their Sample Paths, Springer Verlag,
Berlin, Heidelberg, and New York, 1974.
Jacobsen, M., Splitting times for Markov processes and a generalized Markov property for di u-
sions, Z. Wahrscheinlichkeitstheorie verw. Gebiete 30 (1974), 27{43.
Jacobsen, M., Laplace and the origin of the Ornstein{Uhlenbeck process, Bernoulli 2 (1996),
271{286.
Jeanblanc, M., Pitman, J., and Yor, M., The Feynman{Kac formula and decomposition of Brown-
ian paths, Comput. Appl. Math. 16 (1997), 27{52.
Jeulin, T., Semi-martingales et Grossissement d'une Filtration, Lecture Notes in Mathematics,
vol. 833, Springer-Verlag, Berlin, Heidelberg, and New York, 1980.
Jeulin, T., Applications de la theorie du grossissement de ltrations a l'etude des temps locaux du
mouvement Browniens, in \Grossissement de ltrations: exemples et applications", Lecture
Notes in Mathematics, vol. 1118, Springer Verlag, Berlin, Heidelberg, and New York, 1985,
pp. 197{304.
Jeulin, T., and Yor, M., Inegalite de Hardy, semimartingales, et faux-amis, in \Seminaire de
Probabilites, XII, 1977/78", Lecture Notes in Mathematics, vol. 721, Springer-Verlag, Berlin,
Heidelberg, and New York, 1979, pp. 332{359.
Jeulin, T., and Yor, M., Sur les distributions de certaines fonctionnelles du mouvement brown-
ien, in \Seminaire de Probabilites, XV, 1979/80", Lecture Notes in Mathematics, vol. 850,
Springer-Verlag, Berlin, Heidelberg, and New York, 1981, pp. 210{226.
Johnson, B., and Killeen, T., An explicit formula for the c.d.f. of the L1 norm of the Brownian
bridge, Ann. Probab. 11 (1983), 807{808.
Kac, M., On distribution of certain Wiener functionals, Trans. Amer. Math. Soc. 65 (1949),
1{13.
Kac, M., On some connection between probability theory and di erential and integral equations,
in \Proc. of the second Berkeley Symposium in Mathematical Statistics and Probability",
University of California, Berkeley and Los Angeles, 1951, pp. 189{215.
Kac, M., Wiener and integration in function space, Bull. Amer. Math. Soc. 72(1), part II
(1966), 52{68.
Kac, M., Kiefer, J., and Wolfowitz, J., On tests of normality and other tests of goodness of t
based on distance methods, Ann. Math. Stat. 18 (1955), 189{211.
Kamke, E., Di erentialgleichungen, Losungsmethoden und Losungen, Akademische Verlagsge-
sellschaft Becker & Erler Kom.-Ges., Leipzig, 1943.
Karatzas, I., and Shreve, S.E., Trivariate density of Brownian motion, its local and occupation
times, with application to stochastic control, Ann. Probab. 12 (1984), 819{828.
Karatzas, I., and Shreve, S.E., Brownian Motion and Stochastic Calculus, 2nd edition, Springer-
Verlag, Berlin, Heidelberg, and New York, 1991.
Karlin, S., and Taylor, H.M., A second course in stochastic processes, Academic Press, Boston,
San Diego, and New York, 1981.
 BIBLIOGRAPHY 677

Kennedy, D., The distribution of the maximum of Brownian excursion, J. Appl. Probab. 13
(1976), 371{376.
Kesten, H., An iterated logarithm law for local times, Duke Math. J. 32 (1965), 447{456.
Kesten, H., The in uence of Mark Kac on probability theory, Ann. Probab. 14 (1986), 1103{1128.
Khas'minskii, R. Z., Probability distribution for functionals of the path of a random process of
di usion type, Dokl. Akad. Nauk SSSR 104 (1955), no. 1, 22{25. (Russian)
Khoshnevisan, D., Salminen, P., and Yor, M., A note on a.s. niteness of perpetual integral
functionals of di usions, Elect. Comm. in Probab. 11 (2006), 108{117.
Kinkladze, G.N., A note on the structure of processes the measure of which is absolutely contin-
uous with respect to the Wiener process modulus measure, Stochastics 8 (1982), 39{84.
Knight, F.B., Random walks and a sojourn density process of Brownian motion, Trans. Amer.
Math. Soc. 109 (1963), 56{86.
Knight, F.B., Brownian local times and taboo processes, Trans. Amer. Math. Soc. 143 (1969),
173{185.
Knight, F.B., On the sojourn times of killed Brownian motion, in \Seminaire de Probabilites,
XII, 1976/77", Lecture Notes in Mathematics, vol. 649, Springer-Verlag, Berlin, Heidelberg,
and New York, 1978, pp. 428{445.
Knight, F.B., Essentials of Brownian Motion and Di usion, Math. Surv. No.18, Amer. Math.
Soc., Providence and Rhode Island, 1981.
Kolmogorov, A.N., Uber die analytischen Methoden in der Wahrscheinlichkeitsrechnung, Math.
Ann. 104 (1931), 415{458.
Kramkov, D.O., and Mordecki, E., Integral options, Th. Probab. Appl. 39 (1994), 162{171.
Kunita, H., and Watanabe, T., Markov processes and Martin boundaries, I, Illinois J. Math. 9
(1965), 485{526.
Lai, T. L., Space-time processes, parabolic functions and one-dimensional di usions, Trans. Amer.
Math. Soc. 175 (1973), 409{438.
Lamperti, J., Stochastic processes; a survey of the mathematical theory, Springer-Verlag, New
York, Heidelberg, Berlin, 1977.
Langer, H., and Schenk, W., Generalised second-order di erential operators, corresponding gap
di usions and superharmonic transformations, Math. Nachr. 148 (1990), 7{45.
Lebedev, N.N., Special functions and their applications, Dover, New York, 1972.
Le Gall, J.F., Applications du temps local aux equations di erentielles stochastiques unidimen-
sionelles, in \Seminaire de Probabilites, XVII, 1981/82", Lecture Notes in Mathematics,
vol. 986, Springer Verlag, Berlin, Heidelberg, and New York, 1983, pp. 15{31.
Lehoczky, J.P., Formulas for stopped di usion processes with stopping times based on the maxi-
mum, Ann. Probab. 5 (1977), no. 4, 601{607.
Lenglart, E., Lepingle, D., and Pratelli, M., Une presentation uni ee des inegalites en theorie des
martingales, in \Seminaire de Probabilites, XIV, 1978/79", Lecture Notes in Mathematics,
vol. 784, Springer-Verlag, Berlin, Heidelberg, and New York, 1980, pp. 26{48.
Levy, P., Sur certains processus stochastiques homogenes, Compositio Math. 7 (1939), 283{339.
Levy, P., Processus stochastiques et mouvement brownien, Gauthier-Villars, Paris, 1948.
Levy, P., Construction du processus de W.Feller et H.P.McKean en partant du mouvement
Brownien, in \Probability and Statistics: The Harald Cramer Volume", Wiley and Sons,
New York, 1959, pp. 162{174.
678 BIBLIOGRAPHY

Lindvall, T., On coupling of di usion processes, J. Appl. Probab. 20 (1983), 82{93.

Liptser, R.S. and Shiryaev, A.N., Statistics of Random Processes I, II, Springer-Verlag, Berlin,
Heidelberg and New York, 1977, 1978.
Loeve, M., Paul Levy, 1886-1971, Ann. Probab. 1 (1973), 1{18.
Louchard, G., Kac's formula, Levy's local time and Brownian excursion, J. Appl. Probab. 21
(1984), 479{499.
McGill, P., Markov properties of di usion local time: a martingale approach, Adv. Appl. Probab.
14 (1982), 789{810.

McKean, H.P., Stochastic Integrals, Academic Press, New York, 1969.

McKean, H.P., Brownian local time, Adv. Math. 15 (1975), 91{111.
Mandl, P., Analytical treatment of one-dimensional Markov processes, Springer-Verlag, Berlin,
Heidelberg, and New York, 1968.
Matsumoto, H., and Yor, M., On Bougerol and Dufresne's identities for exponential Brownian
functionals, Proc. Japan Acad., Ser A 74 (1998), 152{155.
Matsumoto, H., and Yor, M., A version of Pitman's 2M{X theorem for geometric Brownian
motion, C. R. Acad. Sc. Paris Series I 328 (1999), 1067{1074.
Matsumoto, H., and Yor, M., An analogue of Pitman's 2M{X theorem for exponential Wiener
functionals, Part I: A time inversion approach, Nagoya Math. J. 159 (2000), 1{42.
Matsumoto, H., and Yor, M., An analogue of Pitman's 2M{X theorem for exponential Wiener
functionals, Part II: The role of the generalized inverse Gaussian laws, Nagoya Math. J.
162 (2001), 65{86.

Meyer, P.A., Flot d'une equation di erentielle stochastique, in \Seminaire de Probabilites, XV,
1979/80", Lecture Notes in Mathematics, vol. 850, Springer-Verlag, Berlin, Heidelberg, and
New York, 1981, pp. 103{117.
Meyer, P.A., Smythe, R.T., and Walsh, J.B., Birth and death of Markov processes, in \Proc.
Sixth Berkeley Symposium III", University of California Press, 1971, pp. 295{305.
Molchanov, S.A., Martin boundaries for invariant Markov process on a solvable group, Th.
Probab. Appl. 12 (1967), 310{314.
Nagasawa, M., Time reversions of Markov processes, Nagoya Math. J. 24 (1964), 177{204.
Nelson, E., Dynamical theories of Brownian motion, Princeton University Press, Princeton, NJ,
1967.
Neveu, J., Processus aleatoires gaussiens, Les Presses de l'Univ. de Montreal, Montreal, 1968.
ksendal, B., Stochastic Di erential Equations, an introduction with applications, 5th edition,
Springer-Verlag, Berlin, Heidelberg, and New York, 1998.
Paulsen, J., Risk theory in a stochastic economic environment, Stoch. Proc. Appl. 46(2) (1993),
327{362.
Perkins, E., The exact Hausdor measure of the level sets of Brownian motion, Z. Wahrschein-
lichkeitstheorie verw. Gebiete 58 (1981), 373{388.
Perkins, E., Local time is a semimartingale, Z. Wahrscheinlichkeitstheorie verw. Gebiete 60
(1982), 79{117.
Perkins, E., Local time and pathwise uniqueness for stochastic di erential equations, in \Semi-
naire de probabilite XVI", Lecture notes in Mathematics., vol. 920, Springer Verlag, Berlin,
Heidelberg, and New York, 1982b, pp. 201{208.
 BIBLIOGRAPHY 679

Pitman, J.W., One-dimensional Brownian motion and three-dimensional Bessel process, Adv.
Appl. Probab. 7 (1975), 511{526.
Pitman, J.W., The SDE solved by local times of a Brownian excursion or bridge derived from
the height pro le of a random tree or forest, Ann. Probab. 27 (1999), 261{283.
Pitman, J.W., and Yor, M., Bessel processes and in nitely divisible laws, in \Stochastic integrals",
Lecture notes in Mathematics., vol. 851, Springer-Verlag, Berlin, Heidelberg, and New York,
1981, pp. 285{370.
Pitman, J.W., and Yor, M., A decomposition of Bessel bridges, Z. Wahrscheinlichkeitstheorie
verw. Gebiete 59 (1982), 425{457.
Pitman, J.W., and Yor, M., Decomposition at the maximum for excursions and bridges of one-
dimensional di usions, in \It^o's Stochastic Calculus and Probability Theory", Springer-
Verlag, Berlin, Heidelberg, and New York, 1996.
Pitman, J.W., and Yor, M., Laplace transforms related to excursions of a one-dimensional dif-
fusions, Bernoulli 5 (1999a), 249{255.
Pitman, J.W., and Yor, M., The law of the maximum of a Bessel bridge, Elect. J. Probab. 4
(1999b), 1{35.
Rauscher, B., Some remarks on Pitman's theorem, in \Seminaire de probabilite XXXI", Lecture
notes in Mathematics., vol. 1655, Springer-Verlag, Berlin, Heidelberg, and New York, 1997,
pp. 266{271.
Ray, D.B., Sojourn times of a di usion process, Ill. J. Math. 7 (1963), 615{630.
Revuz, D., and Yor, M., Continuous Martingales and Brownian Motion, 3rd edition, Springer-
Verlag, Berlin, Heidelberg, and New York, 2001.
Rice, S.O., The integral of the absolute value of the pinned Wiener process { calculation of its
probability density by numerical integration, Ann. Probab. 10 (1982), no. 1, 240{243.
Roberts, A.W., and Varberg, D.E., Convex Functions, Academic Press, New York, 1973.
Rogers, L.C.G., Characterizing all di usions with the 2M{X property, Ann. Probab. 9 (1981),
no. 4, 561{572.
Rogers, L.C.G., A guided tour through excursions, Bull. London Math. Soc. 12 (1989), no. 4,
305{341.
Rogers, L.C.G., and Pitman, J., Markov functions, Ann. Probab. 9 (1981), no. 4, 573{582.
Rogers, L.C.J., and Williams, D., Di usions, Markov Processes, and Martingales, Volume 2: It^o
Calculus, 2nd edition, Cambridge University Press, Cambridge, 2000.
Rogers, L.C.J., and Williams, D., Di usions, Markov Processes, and Martingales, Volume 1:
Foundations, 2nd edition, Wiley and Sons, New York, 1994.
Rosler, U., The tail - eld of time-homogeneous one-dimensional di usion processes, Ann. Probab.
7 (1979), no. 5, 847{857.

Saisho, T., and Tanemura, H., Pitman type theorem for one-dimensional di usion process, Tokyo
J. Math. 13 (1990), no. 2, 429{440.
Salminen, P., Mixing Markovian laws; with an application to path decompositions, Stochastics 9
(1983), 223{231.
Salminen, P., One-dimensional di usions and their exit spaces, Math. Scand. 54 (1984), 209{220.
Salminen, P., Optimal stopping of one-dimensional di usions, Math. Nachr. 124 (1985), 85{101.
Salminen, P., On the distribution of supremum of di usion local time, Statistics & Probability
Letters 18 (1993), 219{225.
680 BIBLIOGRAPHY

Salminen, P., A pointwise limit theorem for the transition density of a linear di usion, in \Proc. of
the fth Finnish{Russian Symposium on Probability Theory and Mathematical Statistics",
Frontiers in Pure and Applied Probability, VSP and TVP, Utrecht and Moscow, 1996.
Salminen, P., On last exit decompositions of linear di usions, Studia Sci. Math. Hungar. 33
(1997), 251{262.
Salminen, P., and Vallois, P., On rst range times of linear di usions, J. Theor. Probab. 18
(2005), 567{593.
Salminen, P., and Yor, M., Perpetual integral functionals as hitting and occupation times, Elect.
J. Probab. 10 (2005), 371{419.
Salminen, P., and Yor, M., Properties of perpetual integral functionals of Brownian motion with
drift, Ann. I. H. Poincare, PR 41 (2005), 335{347.
Schwartz, L., Quelques re exions et souvenirs sur Paul Levy, in \Colloque Paul Levy sur les
processus stochastiques", vol. 157{158, 1988, pp. 13{28.
Sharpe. M., Some transformations of di usions by time reversal, Ann. Probab. 8 (1980), 1157{
1162.
Sharpe, M.J., General theory of Markov processes, Academic Press, New York, 1988.
Shepp, L.A., The joint density of the maximum and its location for a Wiener process with drift,
J. Appl. Probab. 16 (1979), 423{427.
Shepp, L.A., On the integral of the absolute value of the pinned Wiener process, Ann. Probab.
10 (1982), 234{239.

Shepp, L.A., and Shiryaev, A.N., The Russian option: reduced regret, Ann. Appl. Probab. 3
(1993), 631{640.
Shepp, L.A., and Shiryaev, A.N., A new look at pricing of the \Russian" option, Theory Probab.
Appl. 39 (1994), 103{119.
Shiga, T., and Watanabe, S., Bessel di usions as a one-parameter family of di usion processes,
Z. Wahrscheinlichkeitstheorie verw. Gebiete 27 (1973), 37{46.
Shiryaev, A.N., Kolmogorov: life and creative activities, Ann. Probab. 17 (1989), 866{944.
Stroock, D.W., and Varadhan, S.R.S., Di usion processes with continuous coecients I, II,
Comm. Pure Appl. Math. 22 (1969), 345{400, 479{530..
Stroock, D.W., and Varadhan, S.R.S., Multidimensional Di usion Processes, Springer-Verlag,
Berlin, Heidelberg, and New York, 1979.
Takacs, L., On the generalization of the arc-sine law, Ann. Appl. Probab. 6 (1996), 1035{1039.
Taqqu, M.S., Bachelier and his times: A conversation with Bernard Bru, Finance and Stochastics
5 (2001), 3{32.

Taylor, H.M., A stopped Brownian motion formula, Ann. Probab. 3 (1975), no. 2, 234{246.
Taylor, S.J., and Wendel, J.G., The exact Hausdor measure of the zero set of a stable process,
Z. Wahrscheinlichkeitstheorie verw. Gebiete 6 (1966), 170{180.
Trotter, H.F., A property of Brownian motion paths, Ill. J. Math. 2 (1958), 425{433.
Truman, A., and Williams, D., A generalized arc-sine law and Nelson's stochastic mechanics
of one-dimensional time-homogeneous di usions, in \Proc. Int. Conf., Evanston/IL 1989",
Prog. Probab. vol. 22, Birkhauser Verlag, Boston/MA, 1990, pp. 117{135.
Vallois, P., Di usion arretee au premier instant ou l'amplitude atteint un niveau donne, Stoch.
and Stoch. Reports 43 (1993), 93{115.
 BIBLIOGRAPHY 681

Vallois, P., Decomposing the Brownian path via the range process, Stoch. Proc. Appl. 55 (1995),
211{226.
Vervaat, W., A relation between brownian bridge and brownian excursion, Ann. Probab. 7(1)
(1979), 143{149.
Vincze, I., Einige zweidimensionale Verteilungs- und Grenzverteilungssatze in der Theorie der
geordneten Stichproben, Publ. Math. Inst. Hungar. Acad. Sci. 11 (1957), 183{203.
Volkonskij, V.A., Random time changes in strong Markov processes, Th. Probab. Appl. 3 (1958),
no. 2, 332{350.
Walsh, J.B., A di usion with discontinuous local time, Asterisque 52-53 (1978), 37{46.
Walsh, J.B., Downcrossings and the Markov property of local time, Asterisque 52-53 (1978),
89{115.
Warren, J., Branching processes, the Ray{Knight theorem, and sticky Brownian motion, in
\Seminaire de Probabilites, XXXI", Lecture Notes in Mathematics, vol. 1665, Springer-
Verlag, Berlin, Heidelberg, and New York, 1997, pp. 1{15.
Wenocur, M., Ornstein{Uhlenbeck process with quadratic killing, J. Appl. Probab. 23 (1990),
707{712.
Wiener, N., Di erential space, J. Math. Phys. 2 (1923), 131{174.
Williams, D., Path decomposition and continuity of local time for one-dimensional di usions I,
Proc. London Math. Soc. 28 (1974), 738{768.
Williams, D., Levy's downcrossing theorem, Z. Wahrscheinlichkeitstheorie verw. Gebiete 40
(1977), 157{158.
Williams, D., Di usions, Markov Processes, and Martingales, I: Foundations., Wiley and Sons,
New York, 1979.
Yamada, T., and Watanabe, S., On the uniqueness of solutions of stochastic di erential equations,
J. Math. Kyoto Univ. 11 (1971), 155{167.
Yor, M., Un exemple de processus qui n'est pas une semi-martingale., Asterisque 52-53 (1978),
219{221.
Yor, M., Loi de l'indice du lacet Brownien et distribution de Hartman{Watson., Z. Wahrschein-
lichkeitstheorie verw. Gebiete 53 (1980), 71{95.
Yor, M., Some Aspects of Brownian Motion. Part I: Some Special Functionals, Birkhauser, Basel,
Boston, and Berlin, 1992a.
Yor, M., On some exponential functionals of Brownian motion, Adv. Appl. Probab. 24 (1992b),
509{531.
Yor, M., On certain exponential functionals of real-valued Brownian motion, J. Appl. Probab.
29 (1992c), 202{208.

Yor, M., Local times and excursions for Brownian motion: a concise introduction, Lecciones en
Matematicas, Universidad Central de Venezuela, 1995.
Yor, M., Some Aspects of Brownian Motion. Part II: Some Recent Martingale Problems, Birk-
hauser, Basel, Boston, and Berlin, 1997a.
Yor, M. (ed.), Exponential Functionals and Principal Values related to Brownian Motion; A
collection of research papers, Revista Matematica Iberoamericana, Madrid, 1997b.
Zvonkin, A. K., A transformation of the phase space of a di usion process that moves the drift,
Math. USSR Sbornik 22 (1974), 129{149. (Russian)
 SUBJECT INDEX

additive functional II.21 Brownian motion, definition IV.1,

– continuous II.21 A1.1
– natural II.21 – absorbed A1.4
– representing measure II.23 – alternating drift A1.15
– strong Markov II.21 – killed A1.3,6
Airy function A2.5 – killed elastically A1.7
arcsine law IV.18 – laws of iterated logarithms IV.5
backward Kolmogorov equation II.12 – local properties of sample paths
Bessel functions A2.3 IV.6
Bessel process, definition IV.36, 39, – martingale characterization IV.3
A1.21, 22 – reflection principle IV.2
– absolute continuity IV.45 – scaling property IV.2
– laws of iterated logarithms IV.43 – skew A1.12
– last exit time IV.46 – sticky A1.8
– scaling property IV.42 – strong law of large numbers IV.4
– time inversion IV.42 – time inversion IV.2
Blumenthal 0–1 law II.2 Brownian motion with drift, definition
Bougerol’s identity IV.52 IV.27, A1.14
– laws of iterated logarithms IV.29
boundary classification II.6
– last exit time IV.31
– absorbing
– scaling property IV.29
– attractive
– strong law of large numbers IV.29
– elastic
– time inversion IV.29
– entrance
Burkholder–Davis–Gundy inequality
– exit
I.22
– killing
Cameron–Martin–Girsanov
– natural transformation of measure III.26
– non–singular Cameron–Martin formula IV.47
– reflecting canonical realization I.12
– sticky cemetery state I.6
– trap Chapman–Kolmogorov equation I.6
Brownian bridge, definition IV.20 CIR-processes IV.39
– as Doob’s h–transform IV.23 coffin state I.6
– local time IV.25 compensator I.21
Brownian local time IV.7 completion of a σ–algebra I.3
– laws of iterated logarithms IV.10 – universal I.14
– modulus of continuity IV.10 continuous
– scaling IV.10 – almost surely I.1
– stopped at exponential time V.9 – in mean square I.2
– stopped at first exit time V.8 – in probability I.2
– stopped at first hitting time V.6 – stochastically I.2
– stopped at first range time V.7 de Bruno’s formula A5.2
– stopped at inverse local time V.10 diffusion II.1
Brownian meander IV.19 – absolute continuity III.26,27,28

© Springer Basel 2015 682

A.N. Borodin, P. Salminen, Handbook of Brownian Motion - Facts and Formulae,
Probability and Its Applications, DOI 10.1007/978-3-0348-8163-0
 SUBJECT INDEX 683

– communicating class II.3 – weak II.7

– linear II.1 infinitesimal parameters II.9
– recurrent, null–, positively II.12 – killing rate
– regular II.3 – mean
– standard II.9 – variance
– transient II.12 initial distribution I.12
Doob’s h–transform II.31 invariant function II.28
Doob–Kolmogorov inequality I.20 invariant measure II.12
Doob–Meyer decomposition I.21 inverse local time process II.14
Dufresne’s identity IV.50 – Lévy–Khintchine representation
equilibrium measure II.12 II.14
error functions A2.8 isometry III.4
excessive function II.28 Itô formula III.7
– α–excessive II.28 Itô integral III.1
– minimal II.30 Itô measure IV.12
excessive transform II.30 – Itô’s description IV.15
excursion bridge IV.15 – Williams’ description IV.16
excursion length IV.11 Itô representation theorem III.7
excursion process IV.11 Jeulin’s lemma IV.34
excursion space IV.11 Kac moment formula II.25
excursion straddling t IV.18 kernel I.6
exponential Brownian motion IV.48, killed Brownian motion A1.3, 6
A1.20 – elastically A1.7, 9
Feller process I.10 killing measure II.4
– semigroup I.10 Kolmogorov’s criterion for continuity
Feller–Dynkin process I.11 I.2
– semigroup I.11 Kummer functions A2.10
Feynman–Kac formula II.24, VI.1 LCCB space I.1
filtered probability space I.3 Lp -inequalities I.20
filtration I.3 last exit time II.20
– natural I.4 level sets IV.6
– right continuous I.3 Lévy–Khintchine representation II.14
finite dimensional distributions I.1 Lévy process, increasing II.14
first range time V.7, VI.22 Lévy’s area formula IV.46
fundamental solutions II.11 Lévy’s martingale characterization
gamma function A2.2 IV.3
geometric Brownian motion, definition lifetime I.12
IV.48, A.20 local maxima and minima IV.6
Green function II.11 local time II.13
Hermite polynomials A2.6 localization argument III.5
hitting time II.10 Markov process I.6
hyperbolic functions A2.1 – conservative I.6
Hölder continuity IV.6 – honest I.6
independent increments II.14, IV.1 – normal I.6
infinitesimal generator I.8 – time-homogeneous I.6
– of a diffusion II.7 Markov property I.13
– strong I.8 – strong I.13, 14
684 SUBJECT INDEX

Martin boundary II.30 Ray–Knight theorems for Brownian

martingale, super–, sub– I.15 motion with drift
– bounded in L1 I.16 – stopped at exponential time V.15
– class D I.21 – stopped at first exit time V.14
– Lp –inequalities I.20 – stopped at first hitting time V.12,
– local I.17 13
– maximal inequalities I.19 – stopped at inverse local time V.16
– stopped I.17 – total local time V.11
martingale convergence I.16 Ray–Knight theorems for reflecting
martingale representation theorems Brownian motion V.19
III.7 Ray–Knight theorems for diffusion local
maximum of excursion bridge IV.16 time V.2, 3, 4
maximum up to time t and its location reflecting Brownian motion IV.7,
II.19 A1.2, 5
Maxwell distribution A1.25 – killed elastically A1.10
minimal set-up III.14 reflecting Brownian motion with drift
modified Bessel functions A2.4 A1.14
modulus of continuity IV.6, 10 reflection principle IV.2
n–dimensional Bessel process IV.36 resolvent I.7
natural scale II.4 – equation I.7
nowhere differentiability IV.6 sample path I.1
optional σ-field III.2 scale function II.4
optional stopping I.18 semimartingale I.23
Ornstein–Uhlenbeck process A1.24 shift operator I.13
parabolic cylinder functions A2.9 simple predictable process III.2
passage time II.17 skew Brownian motion A1.12
path decomposition IV.33 Skorokhod space I.12
pathwise uniqueness III.18 speed measure II.4
perfect set IV.6 squared Bessel process V.5, A1.23
perpetual integral functional IV.34 squared radial Ornstein–Uhlenbeck
pinned Brownian motion IV.20 processes A1.26
Pitman’s theorem IV.40 state space I.1
points of increase and decrease IV.6 – LCCB I.1
Polish space I.1 – Polish I.1
predictable σ-field III.2 stationary increments II.14
radial Ornstein–Uhlenbeck processes stationary measure II.12
A1.25 stationary process I.1
random time change, substitution sticky Brownian motion A1.8,9
II.16 stochastic process I.1
Rayleigh distribution A1.25 – adapted I.4
Ray–Knight theorems for Bessel process – equivalent I.1
– stopped at exponential time V.21 – identical in law I.1
– stopped at first exit time V.20 – indistinguishable I.1
– stopped at first hitting time V.19 – measurable I.4
– total local time V.17, 18 – modifications I.1
Ray–Knight theorems for Brownian – predictable III.2
motion, see Brownian local time – progressively measurable I.4
 SUBJECT INDEX 685

– standard I.9
– versions I.1
– uniformly integrable I.16
stopping time I.5
strong solution III.14
strong uniqueness III.14
subordinator II.14
tail σ-field II.36
Tanaka formula III.12
theta functions A2.11
tied-down Brownian motion IV.20
time reversed diffusion II.33
trajectory I.1
transition function I.6
– Feller I.10
– Feller–Dynkin I.11
– jointly measurable I.6
transition density II.4
uniqueness in law III.18
universal completions I.14
usual conditions I.2
usual augmentation I.3
weak solution III.17
Vervaat’s path transformation IV.24
Whittaker functions A2.10
Williams’ path decomposition IV.33
Wronskian II.11