Visit https://ebookfinal.

com to download the full version and

explore more ebooks

Stochastic Calculus of Variations for Jump

Processes 1st Edition Yasushi Ishikawa

_____ Click the link below to download _____

https://ebookfinal.com/download/stochastic-calculus-of-
variations-for-jump-processes-1st-edition-yasushi-
ishikawa/

Explore and download more ebooks at ebookfinal.com

Here are some suggested products you might be interested in.
Click the link to download

Levy processes and stochastic calculus 1st Edition David

Applebaum

https://ebookfinal.com/download/levy-processes-and-stochastic-
calculus-1st-edition-david-applebaum/

Applied calculus of variations for engineers Second

Edition. Edition Komzsik

https://ebookfinal.com/download/applied-calculus-of-variations-for-
engineers-second-edition-edition-komzsik/

Applied Calculus of Variations for Engineers Second

Edition Louis Komzsik

https://ebookfinal.com/download/applied-calculus-of-variations-for-
engineers-second-edition-louis-komzsik/

Stationary Stochastic Processes for Scientists and

Engineers 1st Edition Georg Lindgren

https://ebookfinal.com/download/stationary-stochastic-processes-for-
scientists-and-engineers-1st-edition-georg-lindgren/
Stochastic Processes and Models David Stirzaker

https://ebookfinal.com/download/stochastic-processes-and-models-david-
stirzaker/

Applied Stochastic Processes 1st Edition Ming Liao

(Author)

https://ebookfinal.com/download/applied-stochastic-processes-1st-
edition-ming-liao-author/

Stochastic processes selected papers of Hiroshi Tanaka 1st

Edition Hiroshi Tanaka

https://ebookfinal.com/download/stochastic-processes-selected-papers-
of-hiroshi-tanaka-1st-edition-hiroshi-tanaka/

Stochastic Analysis and Diffusion Processes 1st Edition

Gopinath Kallianpur

https://ebookfinal.com/download/stochastic-analysis-and-diffusion-
processes-1st-edition-gopinath-kallianpur/

Advanced Mathematics for Engineers with Applications in

Stochastic Processes 1st Edition Aliakbar Montazer
Haghighi
https://ebookfinal.com/download/advanced-mathematics-for-engineers-
with-applications-in-stochastic-processes-1st-edition-aliakbar-
montazer-haghighi/
Stochastic Calculus of Variations for Jump Processes 1st
Edition Yasushi Ishikawa Digital Instant Download
Author(s): Yasushi Ishikawa
ISBN(s): 9783110282016, 3110282011
Edition: 1
File Details: PDF, 1.35 MB
Year: 2013
Language: english
Yasushi Ishikawa
Stochastic Calculus of Variations for Jump Processes
De Gruyter Studies in
Mathematics

Edited by
Carsten Carstensen, Berlin, Germany
Nicola Fusco, Napoli, Italy
Fritz Gesztesy, Columbia, Missouri, USA
Niels Jacob, Swansea, United Kingdom
Karl-Hermann Neeb, Erlangen, Germany

Volume 54
Yasushi Ishikawa
Stochastic Calculus of
Variations for Jump
Processes
Mathematics Subject Classification 2010
60J25, 60J35, 60G51, 60H07

Author
Prof. Dr. Yasushi Ishikawa
Ehime University
Graduate School of Science and Engineering
Mathematics, Physics, and Earth Sciences
Bunkyo-cho 2-chome
790-8577 Matsuyama
Japan
[email protected]

ISBN 978-3-11-028180-4
e-ISBN 978-3-11-028200-9
Set-ISBN 978-3-11-028201-6

Library of Congress Cataloging-in-Publication Data

A CIP catalog record for this book has been applied for at the Library of Congress.

Bibliographic information published by the Deutsche Nationalbibliothek

The Deutsche Nationalbibliothek lists this publication in the Deutsche
Nationalbibliografie; detailed bibliographic data are available in the Internet at
http://dnb.dnb.de.

© 2013 Walter de Gruyter GmbH, Berlin/Boston

Typesetting: le-tex publishing services GmbH, Leipzig
Printing and binding: Hubert & Co. GmbH & Co. KG, Göttingen
♾ Printed on acid-free paper
Printed in Germany

www.degruyter.com
Preface
This book is a concise introduction to the stochastic calculus of variations (also
known as Malliavin calculus) for processes with jumps. It is written for researchers
and graduate students who are interested in Malliavin calculus for jump process-
es. In this book, “processes with jumps” include both pure jump processes and
jump-diffusions. The author has tried to provide many results on this topic in a self-
contained way; this also applies to stochastic differential equations (SDEs) “with
jumps”. This book also contains some applications of the stochastic calculus for
processes with jumps to control theory and mathematical finance.
The field of jump processes is quite wide-ranging nowadays, from the Lévy mea-
sure (jump measure) to SDEs with jumps. Recent developments in stochastic analy-
sis, especially Malliavin calculus with jumps in the 1990s and 2000s, have enabled
us to express various results in a compact form. Until now, these topics have been
rarely discussed in a monograph. Among the few books on this topic, we would like
to mention Bichteler–Gravereaux–Jacod (1987) and Bichteler (2002).
One objective of Malliavin calculus (of jump type) is to prove the existence of the
density function pt (x, y) of the transition probability of a jump Markov process Xt
probabilistically, especially the very important case where Xt is given by a (Itô, Mar-
cus, Stratonovich . . . ) SDE, cf. Léandre (1988). Furthermore, granting the existence
of the density, one may apply various methods to obtain the asymptotic behaviour of
pt (x, y) as t → 0 where x and y are fixed. The results are known to be different,
according to whether x ≠ y or x = y . We also describe this topic.
The starting point for this book was July 2009, when Prof. R. Schilling invited
me to the Technische Universität Dresden, Germany, to teach a short course on Malli-
avin’s calculus for jump processes. He suggested that I expand the manuscript, thus
creating a book. Prof. H. Kunita kindly read and commented on earlier drafts of the
manuscript. The author is deeply indebted to Professors R. Schilling, M. Kanda, H. Ku-
nita, J. Picard, R. Léandre, C. Geiss, F. Baumgartner, N. Privault and K. Taira.
This book is dedicated to the memory of the late Professor Paul Malliavin.

Matsuyama, December 2012 Yasushi Ishikawa

Contents
Preface v

0 Introduction 1

1 Lévy processes and Itô calculus 5

1.1 Poisson random measure and Lévy processes 5
1.1.1 Lévy processes 5
1.1.2 Examples of Lévy processes 8
1.1.3 Stochastic integral for a finite variation process 11
1.2 Basic materials to SDEs with jumps 13
1.2.1 Martingales and semimartingales 13
1.2.2 Stochastic integral with respect to semimartingales 15
1.2.3 Doléans’ exponential and Girsanov transformation 21
1.3 Itô processes with jumps 24

2 Perturbations and properties of the probability law 33

2.1 Integration-by-parts on Poisson space 33
2.1.1 Bismut’s method 35
2.1.2 Picard’s method 45
2.1.3 Some previous methods 51
2.2 Methods of finding the asymptotic bounds (I) 59
2.2.1 Markov chain approximation 59
2.2.2 Proof of Theorem 2.3 63
2.2.3 Proof of lemmas 70
2.3 Methods of finding the asymptotic bounds (II) 77
2.3.1 Polygonal geometry 77
2.3.2 Proof of Theorem 2.4 78
2.3.3 Example of Theorem 2.4 – easy cases 88
2.4 Summary of short time asymptotic bounds 96
2.4.1 Case that μ(dz) is absolutely continuous with respect to the
m-dimensional Lebesgue measure dz 96
2.4.2 Case that μ(dz) is singular with respect to dz 97
2.5 Auxiliary topics 99
2.5.1 Marcus’ canonical processes 99
2.5.2 Absolute continuity of the infinitely divisible laws 102
2.5.3 Chain movement approximation 107
2.5.4 Support theorem for canonical processes 109
viii Contents

3 Analysis of Wiener–Poisson functionals 114

3.1 Calculus of functionals on the Wiener space 114
3.1.1 Definition of the Malliavin–Shigekawa derivative Dt 116
3.1.2 Adjoint operator δ = D ∗ 120
3.2 Calculus of functionals on the Poisson space 122
3.2.1 One-dimensional case 122
3.2.2 Multidimensional case 125
3.2.3 Characterisation of the Poisson space 128
3.3 Sobolev space for functionals over the Wiener–Poisson space 132
3.3.1 The Wiener space 132
3.3.2 The Poisson Space 134
3.3.3 The Wiener–Poisson space 139
3.4 Relation with the Malliavin operator 148
3.5 Composition on the Wiener–Poisson space (I) – general theory 150
3.5.1 Composition with an element in S 150
3.5.2 Sufficient condition for the composition 156
3.6 Smoothness of the density for Itô processes 160
3.6.1 Preliminaries 160
3.6.2 Big perturbations 163
3.6.3 Concatenation (I) 167
3.6.4 Concatenation (II) – the case that (D) may fail 170
3.7 Composition on the Wiener–Poisson space (II) – Itô processes 178

4 Applications 181
4.1 Asymptotic expansion of the SDE 182
4.1.1 Analysis on the stochastic model 184
4.1.2 Asymptotic expansion of the density 205
4.1.3 Examples of asymptotic expansions 210
4.2 Optimal consumption problem 216
4.2.1 Setting of the optimal consumption 216
4.2.2 Viscosity solutions 220
4.2.3 Regularity of solutions 239
4.2.4 Optimal consumption 243
4.2.5 Historical sketch 246

Appendix 249

Bibliography 253

List of symbols 263

Index 264
0 Introduction
A theme of this book is to describe a close interplay between analysis and probability
via Malliavin calculus. Compared to other books on this subject, our focus is mainly
on jump processes, especially Lévy processes.
The concept of (abstract) Wiener space has been well known since the 1970s.
Since then, despite a genuine difficulty with respect to the definition of the (abstract)
Wiener space, many textbooks have been published on the stochastic calculus on the
Wiener space. It should be noted that it is directly connected to Itô’s theory of stochas-
tic differential equations, which Itô invented while inspired by the work of A. N. Kol-
mogorov. Already at this stage, a close relation between stochastic calculus and PDE
theory has been recognised through the transition probability pt (x, dy), whose den-
sity pt (x, y) is the fundamental solution to Kolmogorov’s backward equation.
Malliavin calculus started with the paper [150] by P. Malliavin (cf. [151]). One of
the motivations of his paper is the problem of hypoellipticity for operators associated
with stochastic differential equations of diffusion type. At the beginning, Malliavin’s
calculus was not very popular (except for his students, and a few researchers such as
Bismut, Jacod, Shigekawa, Watanabe and Stroock) due to its technical difficulties.
Malliavin’s paper was presented at the international symposium in Kyoto organ-
ised by Prof. K. Itô. At that time, a close relation began between P. Malliavin and the
Kyoto school of probability in Japan. The outcome was a series of works by Watan-
abe [204], Ikeda–Watanabe [79], Shigekawa [187], and others.
The relation between Malliavin calculus for diffusion processes and PDEs has
been deeply developed by Kusuoka and Stroock [128–130] and others.
On the other hand, Paul Lévy began his study on additive stochastic processes
(cf. [143]). The trajectories of his processes are continuous or discontinuous. The
additive processes he studied are now called Lévy processes. The discontinuous
Lévy processes have an infinitesimal generator of integro-differential type in the
semigroup theory in the sense of Hille–Yosida. Such integro-differential operators
have been studied in potential theory, by e.g. Ph. Courrège [42] and Bony–Courrège–
Priouret [32]. The latter paper is related to the boundary value problem associated
with integro-differential operators.
The theory developed following that of Fourier integral operators and pseudodif-
ferential operators (cf., e.g. [33]).
My first encounter with Malliavin calculus for jump processes was the paper by
Léandre [131], where he proves

p(t, x, dy) P (Xt ∈ dy|X0 = x)

= ∼ n(x, dy)(t → 0) ,
t t

if the jump process Xt can reach y by one single jump (y ≠ x ). Here, n(x, dy)
denotes the Lévy kernel. This result has been generalised to the case of n jumps,
n = 1, 2, . . . in [82], independently of the work by Picard.
2 Introduction

When I started my research in this field, I was inspired by the close relation be-
tween the theories of integro-differential operators and jump type Markov processes.
Consequently, my own research on the short time asymptotics of the transition den-
sity plays an important role in this monograph.
Later, Malliavin calculus has found new applications in the theory of finance.
The presentation of the contents follows the historical development of the theory.
The technical requirements of this book are usual undergraduate calculus, probabil-
ity and abstract measure theory.
Historically, the theory was started by Bismut. The approach by Bismut is based
on the Girsanov transform of the underlying probability measure. From an analytic
point of view, the main idea is to replace the Radon–Nikodým density function for the
perturbed law of the continuous trajectories by the one for discontinuous trajectories.
From an analytic point of view, the main idea is to replace the Radon–Nikodým den-
sity function in the Girsanov transform of measures induced by the perturbation of
the continuous trajectories by that induced by the perturbation of the discontinuous
trajectories. Subsequently, the theory was extended to cover singular Lévy measures
using perturbation methods (Picard).
Most textbooks on Malliavin calculus on the Wiener space (e.g. [151, 187]) adopt
a functional analytic approach, where the abstract Wiener space and the Malliavin
operator appear. I do not use such a setting in this book. This is partly because
such a setting is not very intuitive, and partly because the setting cannot directly be
transferred to the Poisson space from the Wiener space. This is also discussed in Sec-
tion 3.4.
In the spirit of potential theory and (nonlocal) integro-differential operators,
I have adopted the method of perturbations of trajectories on the Wiener–Poisson
space. This perspective fits well to the Markov chain approximation method used
in Sections 2.2, 2.3, and to the technique of path-decomposition used in Section 3.6.
Hence, it constitutes one of the main themes of this book.
In our approach, both in the Wiener space and in the Poisson space, the main
characters are the derivative operator Dt or the finite difference operator D̃u , and
their adjoints δ or δ̃. The derivative operator is defined to act on the random variable
F (ω) defined on a given probability space (Ω, F , P ).
In the Wiener space, Ω = C0 (T) is the space of continuous functions defined
on the interval T equipped with the topology given by the sup-norm. The Malliavin
derivative Dt F (ω) of F (ω) is then given in two ways, either as a functional derivative
or in terms of a chaos expansion, see Section 3.1.1 for details. The definition via chaos
expansion is quite appealing since it gives an elementary proof of the Clark–Ocone
formula, and since the definition can be carried over to the Poisson space in a natural
way; details are stated in Section 3.2.
Here is a short outline of all chapters.
 Introduction 3

Chapter 1. In Chapter 1, I briefly prepare basic materials which are needed for the
theory. Namely, I introduce Lévy processes, Poisson random measures, stochastic in-
tegrals, stochastic differential equations (SDE) with jumps driven by Lévy processes,
Itô processes, canonical processes, and so on. Some technical issues in the stochas-
tic analysis such as Girsanov transforms of measures, quadratic variation, and the
Doléans stochastic exponential are also discussed. The SDEs introduced in Section 1.3
are time independent, i.e. of autonomous (or “Markovian”) type.
In this chapter, technical details on materials concerning SDEs are often referred
to citations, as our focus is to expose basic elements for stochastic analysis briefly.
Especially, for materials and explanations of diffusion processes, Wiener processes,
stochastic integrals with respect to Wiener process, readers can refer to [110].

Chapter 2. The main subject in Chapter 2 is to relate the integration-by-parts proce-

dure in the Poisson space with the (classically) analytic object, that is, the transi-
tion density function. I present several methods and perturbations that lead to the
integration-by-parts formula on the Poisson space. I am particularly interested in the
Poisson space since such techniques on the Wiener space are already introduced in
the textbooks on Malliavin calculus. The integration-by-parts formula induces the ex-
istence of the smooth density of the probability law or the functional. Analytically,
“existence” and “smoothness” are two different subjects to attack. However, they are
obtained simultaneously in many cases. I present several upper and lower bounds of
the transition densities associated with jump processes. Then, I explain the methods
to find those bounds. The results stated here are adopted from several papers written
by R. Léandre, J. Picard and by myself in the 1990s.
One motivation for Sections 2.2, 2.3 and 2.4 is to provide the short time asymptotic
estimates for jump processes from the view point of analysis (PDE theory). Readers
will find sharp estimates of the densities which are closely related to the jumping
behaviour of the process. Here, the geometric perspectives as polygonal geometry
and chain movement approximation will come into play. Compared to the estima-
tion of the heat kernels associated with Dirichlet forms of jump type, such piecewise
methods for the solution to SDEs will give more precise upper and lower bounds for
transition densities. In Section 2.5, I provide some auxiliary materials.

Chapter 3. In Chapter 3, I study the Wiener, Poisson, and Wiener–Poisson space.

Here, I use stochastic analysis on the path space. In Section 3.1, I briefly review
stochastic calculus on the Wiener space using the Malliavin–Shigekawa’s perturba-
tion. I introduce the derivative operator D and its adjoint δ. In Section 3.2, I dis-
cuss stochastic calculus on the Poisson space using Picard’s perturbation D̃ . In Sec-
tion 3.3, I introduce perturbations on the Wiener–Poisson space, and define Sobolev
spaces on the Wiener–Poisson space based on the norms using these perturbations.
A Meyer’s type inequality for the adjoint operators on this space is explained in de-
4 Introduction

tail. This chapter is the main part in the theoretical aspect of stochastic analysis for
processes with jumps.
In Sections 3.5 (General theory) and 3.7 (Itô processes), I define the composi-
tion Φ ◦ F of a random variable F on the Wiener–Poisson space with a generalised
function Φ in the space S of tempered distributions, such as Φ(x) = (x − K)+ or
Φ(x) = δ(x). These results are mostly new. In Section 3.6, I investigate the smooth-
ness of the density of the processes defined on the Wiener–Poisson space as function-
als of Itô processes.

Chapter 4. Chapter 4 is devoted to applications of the material from the previous

chapters to problems in mathematical finance and optimal control. In Section 4.1, I
explain applications to asymptotic expansions using the composition of a Wiener–
Poisson functional with tempered distributions. In Section 4.1.1, I briefly repeat the
material on compositions of the type Φ ◦ F given in Sections 3.5, 3.7. Section 4.1.2
treats the asymptotic expansion of the density, which closely relates to the short time
asymptotics stated in Sections 2.2, 2.3. In Section 4.2, I give an application to the
optimal consumption problem associated to a jump-diffusion process.

I tried to make the content as self-contained as possible and to provide proofs to

all major statements (formulae, lemmas, propositions, . . . ). Nevertheless, in some in-
stances, I decided to refer the reader to papers and textbooks, mostly if the arguments
are lengthy or very technical. Sometimes, the proofs are postponed to the end of the
current section due to the length. For the readers’ convenience, I tried to provide
several examples. Also, I have repeated some central definitions and notations.
In composing each part of the book, principally, I acted as a “director”, governing
the content, composition, and whether I give the proof of the assertion or refer to the
reference article, for example. However, sometimes I interposed as an “actor” to write
into the text part, that is, as “we sketch the key idea here . . . ”, or “Indeed, . . . ”. In this
sense, this book can be regarded as a hybrid of a course book and a reading book.

How to use this book? Your choices are:

– If you are interested in the relation of Malliavin calculus with analysis, read Chap-
ters 1 and 2
– If you are interested in the basic theory of Malliavin calculus on the Wiener–
Poisson space, read Chapters 1 and 3
– If you are interested in the application of Malliavin calculus with analysis, read
Chapter 4 in reference with Chapter 1.

I hope this book will be useful as a textbook and as a resource for researchers in
probability and analysis.
1 Lévy processes and Itô calculus

Happy families are all alike; every unhappy family is unhappy in its own way.
Lev Tolstoy, Anna Karenina

In this chapter, we briefly prepare the basic concepts and mathematical tools which
are necessary for stochastic calculus with jumps throughout this book. We consider
Poisson processes, Lévy processes, and the Itô calculus associated with these pro-
cesses. Especially, we consider SDEs of Itô and canonical type.
We first introduce Lévy processes in Section 1.1. We provide basic materials to
SDEs with jumps in Section 1.2. Then, we introduce SDEs for Itô processes (Section 1.3)
in the subsequent section. Since the main objective of this article is to inquire into an-
alytic properties of the functionals on the Wiener–Poisson space, not all of the basic
results stated in this chapter are provided with full proofs.
Throughout this book, we shall denote the Itô process on the Poisson space by
xt or xt (x), and the canonical process by Yt . The expressions Xt , X(t) are used for
both cases of the above, or just in the sense of a general Itô process on the Wiener–
Poisson space. In the text, if we cite formula (l, m, n), we mean the formula (m, n)
in Chapter l.

1.1 Poisson random measure and Lévy processes

In this section, we recall the basic materials related to Poisson random measure, Lévy
processes and variation norms.

1.1.1 Lévy processes

We denote by (Ω, F , P ) a probability space where Ω is a set of trajectories defined on

T = [0, T ]. Here, T ≤ ∞ and we mean T = [0, +∞) in the case T = +∞. In most
cases, T is chosen finite. However, the infinite interval T = [0, +∞) may appear in
some cases. F = (Ft )t∈T is a family of σ -fields on Ω, where Ft denotes the minimal
σ -field, right continuous in t , for which each trajectory ω(s) is measurable up to
time t .

Definition 1.1. A Lévy process (z(t))t∈T on T is an m-dimensional stochastic process

defined on Ω such that1

1 Here and in what follows, a.s. denotes the abbreviation for ‘almost surely’. Similarly, a.e. stands for
‘almost every’ or ‘almost everywhere’.
6 Lévy processes and Itô calculus

(1) z(0) = 0 a.s.

(2) z(t) has independent increments (i.e. for 0 ≤ t0 < t1 < · · · < tn , ti ∈ T, the
random variables zti − zti−1 are independent)
(3) z(t) has stationary increments (i.e. the distribution of zt+h − zt depends on h,
but not on t )
(4) z(t) is stochastically continuous (i.e. for all t ∈ T \ {0} and all  > 0P (|z(t +
h) − z(t)| > ) → 0 as h → 0)
(5) z(t) has càdlàg (right continuous on T with left limits on T \ {0}) paths.

Here, m ≥ 1. In case m = 1, we also call z(t) a real-valued process.

We denote
Δz(t) = z(t) − z(t−) .

The same notation for Δ will be applied for processes Xt , Mt , xt , . . . which will appear
later.
We can associate the counting measure N to z(t) in the following way: for A ∈
B(Rm \ {0}), we put

N(t, A) = 1A (Δz(s)), t > 0 .
0≤s≤t

Note that this is a counting measure of jumps of z in A up to the time t . As the path
is càdlàg, for A ∈ B(Rm \ {0}) such that Ā ⊂ Rm \ {0}, we have N(t, A) < +∞ a.s.
A random measure on T × (Rm \ {0}) defined by

N((a, b] × A) = N(b, A) − N(a, A) ,

where a ≤ b and T = [0, T ], is called a Poisson random measure if it follows the Pois-
son distribution with mean measure E[N((a, b] × A)], and if for disjoint (a1 , b1 ] ×
A1 , . . ., (ar , br ] × Ar ∈ B(T × (Rm \ {0})), N((a1 , b1 ] × A1 ), . . ., N((ar , br ] × Ar )
are independent.

Proposition 1.1 (Lévy–Itô decomposition theorem, [180] Theorem I.42). Let z(t) be
a Lévy process. Then, z(t) admits the following representation

t  t 
z(t) = tc + σ W (t) + z Ñ(dsdz) + zN(dsdz) ,
0 |z|<1 0 |z|≥1

for a.e. ω for all t ∈ T. Here, c ∈ Rm , σ is an m × m-matrix, (W (t))t∈T , W (0) = 0

is an m-dimensional standard Wiener process, N(dtdz) is a Poisson random mea-
sure with the mean measure N̂(dtdz) = E[N(dtdz)], and Ñ(dtdz) = N(dtdz) −
t  t 
N̂(dtdz). Here, processes W (t) and t → ( 0 |z|<1 z Ñ(dsdz) + 0 |z|≥1 zN(dsdz))
are independent. Furthermore, this representation is unique.
 Poisson random measure and Lévy processes 7

By this proposition, N(., .) derived from z(t) defines a Poisson random measure
t 
on T × (Rm \ {0}). Here, we use the notation of stochastic integrals 0 zN(dsdz)
t 
and 0 zÑ(dsdz). The precise meaning of these integrals is postponed to Sec-
tion 1.2. However, it should be noted here that the Wiener process W (t) and the Pois-
son random measure N(dtdz) are adapted to the original filtration (Ft ) generated
by the Lévy process z(t).
We take the mean measure

μ(A) = E[N(1, A)], A ∈ B(Rm \ {0}) . (1.1)

This (deterministic) measure is called the Lévy measure associated to z or to N . Note

that μ enjoys 
(1 ∧ |z|2 )μ(dz) < +∞ . (1.2)
Rm \{0}

The compensated Poisson random measure associated to N is defined by

Ñ(dtdz) = N(dtdz) − dtμ(dz) .

In particular, if μ(dz) satisfies


|z|μ(dz) < +∞ ,
|z|≥1

then z(t) can be written in the compact form

t 
z(t) = tc  + σ W (t) + zÑ(dsdz) ,
0 Rm \{0}

where c  = c + |z|≥1 zμ(dz).
We remark that

E[Ñ((s, t] × A1 )Ñ((s, t] × A2 )] = N̂((s, t] × (A1 ∩ A2 )) = (t − s)μ(A1 ∩ A2 )

holds due to the independence property ([119] Proposition 2.1).

A measure μ on Rm \{0} is a Lévy measure associated to some Lévy process if and
only if it enjoys the property (1.2). Indeed, we have the following Lévy–Khintchine
representation.

Proposition 1.2.
(1) Let z be a Lévy process on Rm \ {0}. Then,

E[ei(ξ,z(t)) ] = etΨ (ξ) , ξ ∈ Rm , (1.3)

where

1
Ψ (ξ) = i(c, ξ) − (ξ, σ σ T ξ) + (ei(ξ,z) − 1 − i(ξ, z)1{|z|<1} )μ(dz) . (1.4)
2
8 Lévy processes and Itô calculus

Here, c ∈ Rm , σ σ T is a nonnegative matrix and μ is a measure which satis-

fies (1.2).
(2) Given c ∈ Rm , a matrix σ σ T ≥ 0 and a σ -finite measure μ on B(Rm \ {0})
satisfying (1.2), there exists a process z for which (1.3) and (1.4) hold. This process z
is a Lévy process.

For the proof, we use formulae

  1 T
E ei(ξ,W (t)) = e− 2 t(ξ,σ σ ξ) ,
  
 t  
i(ξ, 0 |z|<1 zÑ(dsdz)) i(ξ,z)
E e = exp t (e − 1 − i(ξ, z))μ(dz) ,
|z|<1

and
  
 t  
E ei(ξ, 0 |z|≥1 zN(dsdz)) = exp t (ei(ξ,z) − 1)μ(dz) .
|z|≥1

Please refer to Theorem 8.1 in [184], and Section 0 in [100]. In the above statement,
(a, b) denotes the inner product of a and b.
Let Dp = {t ∈ T; Δz(t) ≠ 0}. Then, it is a countable subset of T a.s. Let
A ⊂ Rm \ {0}. In case μ(A) < +∞, the process Dp  t → s≤t,Δz(s)∈A δ(s,Δz(s))
is called a Poisson counting measure associated to the Lévy process z(t) (or, the Lévy
measure μ(dz)) taking values in A. The function Dp  t → p(t) = Δz(t) is called
a Poisson point process.

1.1.2 Examples of Lévy processes

1. Poisson process
A Poisson process Nt with intensity λ > 0 is a nonnegative integer-valued process
defined on [0, +∞) which satisfies the following conditions:
(i) N0 = 0, ΔNt = Nt − Nt− is 0 or 1
(ii) For s < t , Nt − Ns is independent of Fs .
(iii) For all t1 , t2 and all s > 0, Nt1 +s − Nt1 has the same distribution as Nt2 +s −
Nt2 .
(iv)
1
P (Nt = k) = (λt)k e−λt , k = 0, 1, 2, . . . .
k!

In fact, the property (iv) follows from (i) to (iii), cf. [180] Theorem I.23. We put (iv)
for simplicity. The Lévy measure μ of the Poisson process is the point mass λδ{1} ,
and b = 0, σ = 0.
 Poisson random measure and Lévy processes 9

2. Compound Poisson process

Nt
Consider a compound Poisson process Yt = k=1 Yk , where (Yk ), k = 1, 2, . . .
are independent and identically distributed (i.i.d.) random variables with a com-
mon finite distribution μ on Rm \ {0} and Nt denotes a Poisson process with the
intensity λ > 0, independent of (Yi ). Then, Yt has a representation

t 
Yt = zN(dsdz) ,
0 Rm \{0}

where N(dsdz) denotes a Poisson random measure on T × (Rm \ {0}) with the
mean measure λdsμ(dz).
3. Stable process
A Lévy process such that its Lévy measure μ , given by

dz
μ(dz) = cα ,
|z|m+α

is called a symmetric stable process, where α ∈ (0, 2).

If the measure μ is given by

 z dz
μ(dz) = cα a ,
|z| |z|m+α

where a(.) is defined on S m−1 and a(·) ≥ 0, the process is called an asymmetric
stable process. In case m = 1, μ takes the form

dz
μ(dz) = (c− 1{z<0} + c+ 1{z>0} ) ,
|z|1+α

where c− ≥ 0, c+ ≥ 0.
4. Wiener process
A Wiener process (or Brownian motion) W (t) (on another probability space) such
that W (0) = 0 satisfies the conditions (1–5) of Definition 1.1. Hence, it is a (con-
tinuous) Lévy process.
1
A Wiener process has a scaling property that if c > 0, then c − 2 W (ct) is indistin-
guishable from W (t) in the sense of distribution.

Now, we proceed by presenting Itô’s formula.

Proposition 1.3. (Itô’s formula (I), [164] Theorems 9.4, 9.5)

(1) Let X(t) be a real-valued process given by

t 
X(t) = x + tc + σ W (t) + γ(z)Ñ(dsdz), t ≥ 0 ,
0 R\{0}
10 Lévy processes and Itô calculus


where γ(z) is such that R\{0} γ(z)2 μ(dz) < ∞. Let f : R → R be a function in
C 2 (R), and let
Y (t) = f (X(t)) .

Then, the process Y (t), t ≥ 0 is a real-valued stochastic process which satisfies

df df 1 d2 f
dY (t) = (X(t))c dt + (X(t))σ dW (t) + (X(t))σ 2 dt
dx dx 2 dx 2

df
+ f (X(t) + γ(z)) − f (X(t)) − (X(t))γ(z) μ(dz) dt
dx
R\{0}

+ [f (X(t) + γ(z)) − f (X(t))] Ñ(dt dz) .
R\{0}

(2) Let X(t) = (X 1 (t), . . ., X d (t)) be a d-dimensional process given by

t 
X(t) = x + tc + σ W (t) + γ(z)Ñ(dsdz) , t ≥ 0.
0

Here, c ∈ Rd , σ is a d × m-matrix, γ(z) = [γij (z)] is a d × m-matrix-

valued function such that the integral exists, W (t) = (W 1 (t), . . ., W d (t))T is
an m-dimensional standard Wiener process, and

Ñ(dtdz) = (N1 (dtdz1 ) − 1{|z1 |<1} μ(dz1 )dt, . . ., Nm (dtdzm )

− 1{|zm |<1} μ(dzm )dt),

where Nj ’s are independent Poisson random measures with Lévy measures μj , j =

1, . . ., m. That is, X i (t) is given by


m m t 

i
X (t) = xi + tci + σij Wj (t) + γij (z)Ñj (dsdzj ) , i = 1, . . ., d .
j=1 j=1 0 R\{0}

Let f : Rd → R be a function in C 2 (Rd ), and let

Y (t) = f (X(t)) .

Then, the process Y (t), t ≥ 0 is a real-valued stochastic process which satisfies

d
∂f d  m
∂f
dY (t) = (X(t))ci dt + (X(t))σij dWj (t)
i=1
∂xi i=1 j=1
∂xi

1  ∂2f
d
+ (X(t))(σ σ T )ij dt
2 i,j=1 ∂xi ∂xj
 Poisson random measure and Lévy processes 11


m  
+ f (X(t) + γ j (zj )) − f (X(t))
j=1 R\{0}

d
∂f 
− (X(t))γij (z) μj (dzj ) dt
i=1
∂xi

m   
+ f (X(t−) + γ j (z)) − f (X(t−)) Ñj (dtdzj ) .
j=1 R\{0}

Here, γ j denotes the j -th column of the matrix γ = [γij ].

For the precise meaning of the stochastic integrals with respect to dW (t) and
Ñ(dtdz), see Section 1.2.

Example 1.1. Let b = 0, γ(z) = 0, σ = 1 and f (x) = x 2 . Then, Itô’s formula leads
to 
1
W (t) dWt = (W (T )2 − T ) .
2
T

1.1.3 Stochastic integral for a finite variation process

A Lévy process z(t) is said to have a finite variation if the total variation

2n  
 t(k − 1) 
|z|t = sup z tk −z  (1.5)
 2n 2n 
n≥1 k=1

is finite a.s. on every compact interval of [0, +∞). If it is not so, the process is said to
have infinite variation.
We introduce the Blumenthal–Getoor index of the Lévy process z(t) by
⎧ ⎫
⎪
⎨  ⎪
⎬
β = inf δ ≥ 0; |z|δ μ(dz) < +∞ .
⎪
⎩ ⎪
⎭
|z|≤1

The index takes values in [0, 2]. It is known (cf. [37, 78]) that if β < 1, then z has
a finite variation path a.s., and if β > 1, then z has an infinite variation path a.s.
t
We shall define the stochastic integral s f (u, ω)dz(u, ω) first for the finite vari-
ation process and then for the infinite variation process, where f is a bounded, jointly
measurable function.
12 Lévy processes and Itô calculus

Definition of stochastic integral for finite variation (FV) processes

For a FV process z(t), we define
t
I(t, ω) = f (ω, u)dz(ω, u)
s

as a Lebesgue–Stieltjes integral (ω-wisely, a.s.), where f is bounded and jointly mea-

surable.
This is a Stieltjes integral of f by dz given ω-wise. In case that u → f (u) has
a continuous path a.s., it is called a Riemann–Stieltjes integral.
If z(t) is a FV process and if f (., ω) is differentiable a.s., then we have, in fact
the usual rule of the change of variables:
t
f (z(t)) − f (z(0)) = f  (z(s))dz(s)a.s.
0

For the integral using the infinite variation process, we need the predictable property
for u → f (u, .) and that of semimartingales. For these, see Section 1.2.2.
In the following sections of this chapter, we use the notion of a stochastic differ-
ential equation (SDE) with respect to the Lévy process z(t). The precise definition
and the properties of the solution are postponed to the next section.
Due to a recent development by T. Lyons [147, 148], there is a possibility to define
“stochastic integrals” ω-wisely by using the (iterated) Young integrals of processes of
finite or infinite variation not using the integration by semimartingales. The theory is
called the rough path theory, and it uses the notion of p -variation norm and spaces.
See also [50].
Indeed, similarly to (1.5), we can define the p -variation
  
2n 
 p
(p)
  tk t(k − 1) 
|z|t = sup z −z  (1.6)
n≥1 k=1
 2n 2n 

for p ≥ 1. All components in the series

(1, z 1 (t), z2 (t), . . .) (1.7)


of the iterated integrals zk (t) = 0≤u1 <...<uk <t dz(u1 )⊗. . .⊗dz(uk ), k = 1, 2, . . . are
measured by the p -variation norm. The series, viewed as multiplicative functionals
of t , are called rough paths. The stochastic integral
t
h(s)dz(s)
0

with respect to the integrator dzs of finite p -variation can be embedded into the the-
ory of integration using rough paths.
 Basic materials to SDEs with jumps 13

The space D
Let T = [0, T ], T < +∞. D = D(T) denotes the space of all functions defined on T
with values in Rm or Rd that are right continuous on [0, T ) and have left limits on
(0, T ] (càdlàg paths). We introduce a topology on D(T) by introducing the Skorohod
metric dT defined by
 
dT (f , g) = inf sup |f (t) − g(τ(t))| + |τ(t) − t| ,
τ t∈T

where τ moves over all strictly increasing, continuous mappings of T to T such that
τ(0) = 0, τ(T ) = T . The topological space (D(T), dT ) is called a Skorohod space.
The space (D(T), dT ) is separable, and by choosing an equivalent metric d◦T it is
complete ([26] Section 12).
D([0, +∞)) denotes the space of all càdlàg paths on [0, +∞). It is a Fréchet
space metrisable with the metric

1  
∞
d(f , g) = n
1 ∧ d◦[0,n] (f , g) ,
n=1
2

and the topological space (D([0, +∞)), d) is complete and separable.

If we adopt the sup-norm (as in [99])
∞
 
 1
d∞ (f , g) = 1 ∧ sup |f (s) − g(s)| ,
n=1
2n s∈[0,n]

the space (D([0, +∞)), d∞ ) is complete, but it is not separable. We will encounter
this space again in the next section and in Section 2.5.3.

1.2 Basic materials to SDEs with jumps

In this section, we study stochastic differential equation (SDE) with jumps. We begin
with the definitions of the martingale, semimartingale, and the stochastic integral.
We treat usual (Itô’s) SDEs and Marcus’ canonical SDEs. Solutions to these two SDEs
are different from each other. The “canonical” integral is introduced by Marcus [155],
and has been developed by Kurtz, Pardoux and Protter [125].

1.2.1 Martingales and semimartingales

Let (Ω, F , P ) be a probability space. A family (Ft )t∈T of sub σ -fields of F is called
a filtration if Fs ⊂ Ft for all s < t . A filtration (Ft )t∈T is said to satisfy the usual
conditions if F0 contains all null sets of F and if it is right continuous. Below, we
consider probability spaces equipped with filtrations which satisfy the usual condi-
tions.
14 Lévy processes and Itô calculus

A stochastic process (Xt )t∈T is said to be adapted if Xt is Ft -measurable for

all t . It is called progressively measurable if the function X. : [0, t] × Ω is B([0, t]) ⊗
Ft -measurable for all t ≥ 0.
An adapted process Mt having càdlàg (right continuous with left limit) paths is
called a martingale if it satisfies the following two conditions:

(1) Mt ∈ L1 (P ), t ∈ T (2.1)
(2) if s ≤ t then E[Mt |Fs ] = Ms , a.s., s, t ∈ T . (2.2)

In case that,

if s ≤ t then E[Mt |Fs ] ≥ Ms , a.s., s, t ∈ T ,

Xt is called a submartingale. In case that

if s ≤ t then E[Mt |Fs ] ≤ Ms , a.s., s, t ∈ T ,

Xt is called a supermartingale.
A random variable T : Ω → [0, +∞] is said to be a stopping time if the event
{T ≤ t} ∈ Ft for every t ∈ T. The set of all stopping times is denoted by T . Let
0 = T0 ≤ T1 ≤ . . . ≤ Tn ≤ . . . be a sequence of stopping times such that Tn →
+∞a.s . An adapted process Mt such that, for some sequence of stopping times as
above, Mt∧Tn is a martingale for any n is called a local martingale. A martingale is
a local martingale.
A process Xt is called a semimartingale if it can be written as

Xt = X0 + Mt + At ,

where Mt is a local martingale and At is an adapted càdlàg process of finite variation

(FV) a.s. In particular, an adapted FV process having a càdlàg path is a semimartin-
gale. A martingale is a semimartingale.
The one-dimensional standard Poisson process Nt is a semimartingale since it
can be written as
Nt = Ñt + t ,

where Ñt = Nt − t is a local martingale, even a martingale.

For a pure jump Lévy process z(t), the compensated Lévy process z̃(t) = z(t) −
t  
0 ds zμ(dz) is a local martingale if |z|μ(dz) < +∞.
A Lévy process is a semimartingale by the Lévy–Itô decomposition theorem.
If X is a martingale, then there exists a unique Y of X which is càdlàg. See the
Corollary in [180] Chap. I.2.
 Basic materials to SDEs with jumps 15

1.2.2 Stochastic integral with respect to semimartingales

t
We would like to define the stochastic (Itô) integral s h(s)dXu by

t t t
h(u)dXu = h(s)dMu + h(s)dAu ,
s s s

where h(u) is a (locally) bounded predictable process.

A process h is said to be predictable if it is measurable with respect to the
σ -field P on Ω × R+ . Here, P denotes the σ -field generated by adapted process-
es whose trajectories are left continuous with right limits.
As t → At is a finite variation process, the second term on the right-hand
side above is given as in the beginning of Chapter 1 (Section 1.1.3). We shall define
t
s h(s)dMu in what follows.
We denote by 0 = t0 ≤ t1 ≤ t2 ≤ . . . ≤ tn a sequence of times. An elementary
process h(t) is a process


n−1
h(t) = h0 1{0} (t) + hi 1(ti ,ti+1 ] (t) ,
i=0

where hi is Fti -measurable and |hi | < +∞ a.s. We denote by S the set of elementary
processes, endowed with the topology given by the uniform convergence in (t, ω).
We define the integral I(h) of an elementary process h ∈ S with respect to the mar-
tingale M having càdlàg path by


n
I(h) = h(0)M0 + hi (Mti+1 − Mti ) .
i=0

I(h) is called the stochastic integral of h with respect to M .

Then, I(h) has the following properties:
(1) If Mt = W (t) (Brownian motion) or Mt = Ñt (compensated Poisson process),
then I(h)(t) is a martingale. That is,

E[I(h)(t)|Fs ] = I(h)(s), s ≤ t .
t
(2) I 2 (h)(t) − h2 (s)d[M]s is a martingale.
0
t
(3) E[I 2 (h)(t)] = E[ 0 h2 (s)d[M]s ].

Here, [M] denotes the quadratic variation of M (see just below for the definition). For
the proof of (1), see [179] Proposition 2.5.7. It follows from (3) that h → I(h) extends
to an isometry from the space of elementary processes equipped with the norm on
16 Lévy processes and Itô calculus

(progressively measurable) adapted processes in L2 (Ω × [0, +∞), P × d[M]s ) into

L2 (Ω, F , P ).2 We state this more precisely in (i)–(iii) below.
(i) We denote by Λ the set

Λ = h ∈ L2 (Ω × [0, +∞), P × d[M]s );
there exists a sequence of elementary functions hn

such that hn → h in L2 (Ω × [0, +∞), P × d[M]s ) .

We can then define for h ∈ Λ, that is,

I(h) = lim I(hn ), It (h) = I(h · 1[0,t] ) .

n→+∞

We denote by D the space of adapted processes with càdlàg paths with the Sko-
rohod topology. It can be observed that the process I(h) for h ∈ S takes values
in D.
It is known (cf. [44]) that Λ contains all predictable processes h such that
∞
E[ 0 h2 (s)d[M]s ] < +∞. Hence, for h ∈ Λ, the previous properties (1–3)
for I(h) hold true.
(ii) More precisely, we first extend I(·) : S → D to I(·) : L → D, where L denotes the
space of adapted processes with càglàd paths (left continuous paths with right
limits) endowed with the topology given by the uniform convergence in proba-
bility on compact sets (ucp-topology, for short). Here, we say a sequence (hn )
converges to h in the ucp-topology if for each t > 0,

sup |hn (s) − h(s)| → 0

0≤s≤t

in probability.
For the proof of this extension, we use the fact that the elements in S are dense
in L in the ucp-topology, that bounded elements in L are dense in L, and that the
bounded elements in L can be approximated by the bounded elements in S in the
ucp-topology ([180] Theorems II.10, II.11.).
We then extend I(.) : L → D to I(.) : Λ → D. The map I is well-defined for each
h ∈ Λ.
(iii) In case Mt = W (t) or Mt = Ñt , we extend I(.) thus obtained to I(.) : L2 (Ω ×
[0, +∞), P × d[M]s ) → D by using the L2 -isometry (3) above. Here, we use the
fact that Λ is dense in L2 (Ω ×[0, +∞), P ×d[M]s ) and the bounded convergence
theorem.
To prove this statement, we first approximate an element in L2 (Ω × [0, +∞), P ×
d[M]s ) by a sequence of bounded adapted processes in the L2 -norm, and then

2 Two P s in the two L2 spaces are distinct. Here, we use the same symbol, supposing that no confu-
sion occurs.
 Basic materials to SDEs with jumps 17

we approximate the bounded adapted process by a sequence of elementary pro-

cesses in the L2 norm. For the precise argument, see [179] Proposition 2.5.3. In-
deed, integrals of predictable and adapted versions coincide with each other. The
extension I(h) does not depend on the choice of the sequences. To prove this, we
need the completeness of space of the square-integrable martingales.

The extension I : L2 (Ω×[0, +∞), P ×d[M]s ) → D thus obtained is called the stochas-
tic integral.

Quadratic variation process

We introduce the quadratic variation process that was previously mentioned. The
following result is due to P.-A. Meyer:

Theorem 1.1 ([180] Theorem III.11).

(1) Let X be a càdlàg supermartingale with X0 = 0 such that the class of random vari-
ables {Xτ ; τ ∈ T } is uniformly integrable. Then, there exists a unique, increasing,
predictable process A with A0 = 0 and a uniformly integrable martingale Mt such
that Xt = Mt − At .
(2) Let X be a càdlàg submartingale with X0 = 0 such that the class of random vari-
ables {Xτ ; τ ∈ T } is uniformly integrable. Then, there exists a unique, increasing,
predictable process A with A0 = 0 and a uniformly integrable martingale Mt such
that Xt = Mt + At .

Remark 1.1. We have a similar decomposition

Xt = Mt − At (resp. Xt = Mt + At )

without the above mentioned uniform integrability condition, but with replacing Mt
to be a local martingale ([180] Theorem III.16).

Let Mt be a square-integrable martingale, null at 0 and bounded in L2 . Then, by

Doob’s inequality,  
 
2 2
E sup Mt ≤ 4E M∞ < +∞ .
t

Hence, Mt2 is a submartingale which satisfies the above property. By Theorem 1.1,
there exists a unique, increasing, predictable process A with A0 = 0 such that

Xt = Mt + At .

Proposition 1.4 (cf. [182] Theorem IV.26). Let M be a square integrable martingale
such that M0 = 0. Then, there exists a unique increasing process [M], [M]0 = 0,
such that
(1) M 2 − [M] is an uniformly integrable martingale,
(2) Δ[M] = (ΔM)2 .
18 Lévy processes and Itô calculus

The process [M] is called the quadratic variation process of M . An intuitive mean-
ing of [M] is given by
 2
[M]t = lim Mtin − Mti−1
n
n→+∞
i

i
where tin =t∧ 2n .
By the two decompositions above, we see that At in Theorem 1.1(2) coincides with
the compensator of [M]t . Namely, the compensator is a predictable FV process, null
at 0, such that [M]t − At is a local martingale. We write it by Mt , and call it the
angle bracket of Mt .
The processes [M]t and Mt coincide if t → Mt is continuous a.s.
We can decompose
M = Mc + Md ,
where M c is the continuous part and M d is the purely discontinuous part. The
quadratic variation process [M]t can be decomposed into continuous and discontin-
uous parts by 
[M]t = [M c ]t + (ΔMs )2 .
0≤s≤t
Hence, we can decompose
[M] = [M]c + [M]d .
Here, [M]c = [M c ] and [M]d = [M d ], where [M d ] = 0≤s≤t (ΔMs )2 .
t
The property (2) of I(h) above implies that [I 2 (h)]t = 0 h2 (t)d[M]t .
For square integrable martingales M, N such that M0 = 0 and N0 = 0, the
quadratic covariance process [M, N] is given by
1
[M, N] = ([M + N] − [M − N]) . (2.3)
4
Using this notation, we have

[M, N] = [M, N]c + [M, N]d .

For a semimartingale X = X0 + M + A, we have [X] = [M] since [A] = 0. For two

semimartingales X, Y , we put
1
[X, Y ] = ([X + Y ] − [X − Y ])
4
as above. By the above definition (2.3), we have
t
[I(h), I(g)]t = h(s)g(s)d[M]s
0

(cf. the result of [44]).

In summary, we have the following properties of the stochastic integral I(h), h ∈
L2 (Ω × [0, +∞), P × d[M]s ).
 Basic materials to SDEs with jumps 19

Properties of the stochastic integral I(h):

(1) For constants α, β and h, g ∈ L2 (Ω × [0, +∞), P × d[M]s ),

I(αh + βg) = αI(h) + βI(g), a.e.

(2) For 0 ≤ t1 ≤ t2 ≤ t3

t3 t2 t3

h(t) dMt = h(t) dMt + h(t) dMt .
t1 t1 t2

(3) t → I(h)(t) is an adapted process.

(4) I(h)(0) = 0 a.s.
(5) If Mt = W (t) or Mt = Ñt , then t → I(h) is a martingale, and hence

E[I(h)(t)] = 0 , t > 0.

(6)
t t
2
[I(h)]t = |h(s)| d[M]s , [I(h), I(g)]t = h(s)g(s)d[M]s .
0 0

Below, up to the end of this subsection, (Ft ) denotes the filtration generated by the
Lévy process z(t) satisfying the usual conditions.
We can introduce the integral with respect to Ñ in terms of the z variable (Pois-
son random measure) by introducing that by elementary Poisson measures. See [119]
Section 2.1. Then, for
 
I(ϕ) = ϕ(z)Ñ((s, t] × dz) , I(ψ) = ψ(z)Ñ((s, t] × dz) ,

we have 
I(ϕ), I(ψ)t = (t − s) ϕ(z)ψ(z)μ(dz) .

Here, ϕ and ψ are measurable functions such that


(|ϕ(z)|2 + |ψ(z)|2 )μ(dz) < +∞ .

We can define an integral

 t
h(s, z)Ñ(dsdz)
0

by starting from simple predictable processes


h(t, z) = ψi (z)1[ti ,ti+1 ) (t) ,
i
20 Lévy processes and Itô calculus

and then approximating a measurable h(t, z) such that

⎡ ⎤
t 
⎢ ⎥
E⎣ |h(s, z)|2 dsμ(dz)⎦ < +∞ . (2.4)
0

Indeed, first we remark

⎡⎛ ⎞2 ⎤
t   t 
⎢⎜ ⎟  ⎥

⎢
E ⎣⎝ ⎥
ψ(z)Ñ(dr dz)⎠ Fs ⎦ = ψ2 (z)N̂(dr dz)
s s

= (t − s) ψ2 (z)μ(dz) , (2.5)

if ψ(z) is Fs -measurable.
For h(t, z) = i ψi (z)1[ti ,ti+1 ) (t), where ψi are Fti -measurable, we write
t  
h(s, z)Ñ(dsdz) = ψi (z)(Ñ(ti+1 ∧ tdz) − Ñ(ti ∧ tdz)). (2.6)
0 i

Then, by (2.5),
⎡ ⎤
.  t 
⎢ ⎥
⎣ h(s, z)Ñ(ds dz)⎦ = h2 (s, z)N̂(dsdz) .
0 t 0

This implies the L2 -isometry

⎡ 2 ⎤ ⎡ ⎤
t   t 

⎢  ⎥
 ⎢ ⎥
E⎢ ⎣ h(s, z)Ñ(dsdz) ⎥ = E⎣ h2 (s, z)N̂(ds dz)⎦ .
  ⎦
0  0

We denote by L2 (N̂) the set of all predictable functionals h(s, z) satisfying the
condition (2.4). The next assertion follows by the standard argument.

Proposition 1.5. Simple predictable processes h with the property (2.4) are dense in
L2 (N̂).

By this proposition, any element h in L2 (N̂) can be approximated by the stochas-

tic integral of the form (2.6) in L2 (N̂).
We have the following martingale representation theorem due to Kunita–Wata-
nabe. We assume m = 1 for simplicity.

Theorem 1.2 (Kunita–Watanabe representation theorem, cf. [113, 180] Theorem IV.43).
Let Mt be a locally square integrable martingale defined on (Ω, F , P ). Then, there exist
predictable, square integrable processes φ(t), ψ(t, z) such that
t t 
Mt = M0 + φ(s)dW (s) + ψ(t, z)dÑ(dsdz) .
0 0
 Basic materials to SDEs with jumps 21

In the above assertion, we take F = (Ft )t∈T , where Ft is the minimal sub σ -field
on which W (s) and Ñ((0, s] × E), s ≤ t are measurable for each E ⊂ Rm \ {0}.
Itô’s formula for the Lévy process (Proposition 1.3) can be extended to the follow-
ing form.

Theorem 1.3 (Itô’s formula (II), [180] Theorems II.32, II.33).

(1) Let X(t) be a real-valued semimartingale and let f be a C 2 function on R. Then,
f (X(t)) is a semimartingale, and it holds that

t t
1
f (X(t)) = f (X(0)) + f  (X(s−))dX(s) + f  (X(s−))d[X, X]cs
2
0 0

+ [f (X(s)) − f (X(s−)) − f  (X(s))ΔX(s)] .
0<s≤t

(2) Let X(t) = (X 1 (t), . . ., X d (t)) be a d-dimensional semimartingale, and let f :

Rd → R be a function in C 2 (Rd ). Then,

Y (t) = f (X(t))

is a semimartingale, and the following formula holds:

d t
∂f
Y (t) − Y (0) = (X(s−))dX i (s)
i=1
∂x i
0

d  t
1  ∂2f
+ (X(s−))d[X i , X j ]cs
2 i,j=1 ∂xi ∂xj
0
⎡ ⎤
 d
∂f
+ ⎣f (X(s)) − f (X(s−)) − (X(s−))ΔX (s)⎦ .
i

0<s≤t i=1
∂xi

1.2.3 Doléans’ exponential and Girsanov transformation

The above formula can be regarded as an equation with respect to the process X(t).
Such an equation is called a stochastic differential equation (SDE).
A typical example is Doléans–Dade (local martingale) exponential to the Doléans’
equation
t
Xt = 1 + Xs− dMs ,
0

where M is a local martingale, M0 = 0. The solution is

* +
1
E (M)t = exp Mt − [M c , M c ]t Πs≤t (1 + ΔMs ) exp(−ΔMs ) .
2
22 Lévy processes and Itô calculus

If M is a locally square integrable martingale such that ΔMt > −1 and if it holds
that ⎧ ⎛ ⎡ ⎤⎞⎫
⎪
⎨
⎜1 , ⎟
⎪
⎬
E exp ⎝ M c T + ⎣ f (ΔMt )⎦⎠ < +∞ , (2.7)
⎪
⎩ 2 ⎪
⎭
t≤T

then E (M)t is a martingale for t ∈ T. Here, f (x) = (1 + x) log(1 + x) − x and ϕ̃

denotes the compensator of ϕ. This result is due to Lépingle and Mémin [142] Theo-
rem III.1. In case ( t≤T,f (ΔMt )) = 0, this condition is called the Novikov condition.
We remark the use of the conditional quadratic variation in the condition.
In particular, if Mt = λW (t), λ ∈ R, then
 
λ2
E (λW )t = exp λW (t) − t .
2

Then, t → E (λW )t is again a local martingale.

Girsanov transformation of measures, martingale exponential (Change of variables)

The following Girsanov transformation of the underlying probability measures plays
an essential role in the perturbation of trajectories in Section 2.1.1. We assume the
processes are real-valued (cf. [142], see also [164] Theorem 12.21, [63] p. 149).

Theorem 1.4.
(1) Let θ(t, z), t ∈ T, z ∈ R \ {0} be a predictable process such that θ(t, z) < 1, and

 
| log(1 − θ(t, z))|2 + θ(t, z)2 dtμ(dz) < +∞ ,
T

and let u(t), t ∈ T be a predictable process such that T u(t)2 dt < +∞.
We put
- t t
1
Zt = exp u(s)dW (s) − u2 (s)ds
2
0 0
t 
+ log(1 − θ(s, z))Ñ(dsdz)
0
t  .
+ {(log(1 − θ(s, z)) + θ(s, z))}dsμ(dz) .
0
 Basic materials to SDEs with jumps 23

We further assume that it holds the Lépingle–Mémin condition:

⎡
 
⎢ 1
E ⎣ exp u2 (t)dt
2
T
⎤
 

⎥
+ (1 − θ(t, z)) log(1 − θ(t, z)) + θ(t, z) dtμ(dz) ⎦ < +∞ . (2.8)
T

Then, t → Zt is a martingale, and

E[ZT ] = 1 .

Hence Q(A) = E[1A .ZT ] defines a probability measure on (Ω, F ) such that
Q(A) = E[1A .Zt ] for A ∈ Ft . That is,

dQ
|F = Zt , t > 0.
dP t
(2) Let
Ñ1 (dtdz) = θ(t, z)dtμ(dz) + Ñ(dtdz)
and
dW1 (t) = u(t)dt + dW (t) .

Then, Ñ1 is a martingale counting measure with respect to Q, and W1 is a continu-

ous martingale with respect to Q.

Remark 1.2.
(1) To see Zt is a martingale, we use the Lépingle–Mémin result above. The condition
(2.7) is implied by our assumption (2.8). We put
t t 
U (t) = u(s)dW (s) − θ(s, z)Ñ(dsdz) .
0 0

Then, ΔU (t) = −θ(t, z) > −1. We can show by Itô’s formula with f (x) = ex
that
dZt = Zt− dU (t)
with Z0 = 1. Hence,
Zt = E (U )t
by the uniqueness of the Doléans’ exponential.
If we assume a uniformity condition in t , that is,

|u(t)| ≤ C and |θ(t, z)| ≤ C|z|, z ∈ supp μ

for some C > 0, then Zt is a positive martingale.

24 Lévy processes and Itô calculus

(2) In Section 2.1.1 below, we introduce a perturbation method using the Girsanov
transform of measures. Bismut [28] used the expression for Zt in terms of SDE
above, whereas Bass et al.[13] used the expression E (M)t .
(3) Under the assumption that (Ft ) satisfies the usual conditions, Ñ1 = Ñt |t=1 is not
necessarily a compensated Poisson random measure with respect to Q, and W1
is not necessarily a Brownian motion with respect to Q (cf. [24] Warning 3.9.20).

Let z(t) be a one-dimensional Lévy process with the Lévy measure μ(dz) :
t 
z(t) = 0 z Ñ(dsdz) being a martingale. Here, we assume that supp μ is compact.
We put θ(t, z) = 1 − eα(t)·z where α(t) is fixed below. Then,
log(1 − θ(t, z)) + θ(t, z) = α(t) · z + 1 − eα(t)·z
 
= − eα(t)·z − 1 − α(t) · z .

Furthermore, we choose u(t) ≡ 0. Then,

⎧ ⎫
⎨t 
⎪ t    ⎪
⎬
α(s)·z
Zt = exp α(s) · z Ñ(dsdz) − e − 1 − α(s) · z dsμ(dz)
⎪
⎩ ⎪
⎭
0 0

on Ft . We see u(t) and θ(t, z) satisfy the above mentioned condition for the unifor-
mity for a bounded α(t). Hence, putting a new measure dQ by dQ = Zt dP on Ft ,
t 
the process z1 (t) = 0 zÑ1 (dsdz) is a Lévy process which is a martingale under Q.
Here, we can choose α(t) to be some deterministic function. In particular,
dL
α(t) = (ḣ(t)) ,
dq
where L(q) denotes the Legendre transform of the Hamiltonian associated with the
process z(t):
H(p) = log E[ep·z(1) ] ,
and h(t) is an element in the Sobolev space W 1,p (T), p > 2. This setting is used in
the Bismut perturbation in Section 2.1.1.
From now until Section 2.5, we are mainly interested in the processes which are
obtained as a solution to the SDE driven by pure jump Lévy processes, and we shall
not consider those driven by Lévy processes having the diffusion part.

1.3 Itô processes with jumps

In this section, we introduce an Itô SDE driven by a Lévy process, and processes de-
fined by it. A solution to an Itô SDE is called an Itô process.
Let z(t) be a Lévy process, Rm -valued, with Lévy measure μ(dz) such that the
characteristic function ψt is given by
 
ψt (ξ) = E[ei(ξ,z(t)) ] = exp t ei(ξ,z) − 1 − i(ξ, z)1{|z|≤1} μ(dz) .
 Itô processes with jumps 25

We may write
t 
z(t) = (z1 (t), . . ., zm (t)) = z(N(dsdz) − 1{|z|≤1} .μ(dz)ds) ,
0 Rm \{0}

where N(dsdz) is a Poisson random measure on T×Rm \{0} with mean ds ×μ(dz).
We denote the index of the Lévy measure μ by β, that is,

β = inf{α > 0; |z|α μ(dz) < +∞} .
|z|≤1

We assume 
|z|2 μ(dz) < +∞ (3.1)
Rm \{0}

temporarily for simplicity.

We can write the SDE in a general formulation on Ω. That is, consider an SDE
t t t 
Xt = x + b(Xs− )ds + f (Xs− )dW (s) + g(Xs− , z)Ñ(dsdz) . (∗)
0 0 0 Rm \{0}

Here, we assume
|b(x)| ≤ K(1 + |x|), |f (x)| ≤ K(1 + |x|), |g(x, z)| ≤ K(z)(1 + |x|) ,

and
|b(x) − b(y)| ≤ L|x − y| ,
|f (x) − f (y)| ≤ L|x − y| ,
|g(x, z) − g(y, z)| ≤ L(z)|x − y| .

Here, K, L are positive constants, and K(z), L(z) are positive functions, satisfying

{K p (z) + Lp (z)}μ(dz) < +∞ ,
Rm \{0}

where p ≥ 2. In Section 4.1, we will use the SDE of this form as a financial model.

Theorem 1.5. Assume x is p -th integrable. Under the assumptions on b(x), f (x),
g(x, z) above, the SDE has a unique solution in Lp .

The proof depends on making a successive approximation (Xtn ) of the solution,

and on estimating
E[sup |Xtn+1 − Xtn |p ] .
t∈T
We omit the detail ([119] Section 3).
To state the notion of the weak solution of (∗), we prepare for the enlargement
(Ft ) of (Ft ), where (Ft ) denotes the original filtration generated by z(t). Here, (Ft )
satisfies the following properties:
26 Lévy processes and Itô calculus

(i) Ft ⊂ Ft for each t

(ii) (Ft ) satisfies the usual conditions
(iii) W (t) is a Brownian motion with respect to Ft
(iv) N(dtdz) is a Poisson random measure with respect to Ft
(v) x is F0 -measurable.

If we can find a process X̃t which is Ft -measurable for some (Ft ) as above, such
that X̃0 and x have the same distribution, and that X̃t satisfies (∗) for some Lévy
process z̃(t), then it is called a weak solution. A solution Xt to (∗) is called a strong
solution if it is an adapted process (to (Ft )) and if it is represented as a functional of
the integrator z(t) and x : X. = F (x, z(.)).
A strong solution is a weak solution. Few results are known for the existence of
strong solutions in case that z(t) is a general semimartingale. For the uniqueness
of the solution (in the strong and weak senses), see [180] Theorems V.6, V.7 (See al-
so [197]).

Definition 1.2. We say the pathwise uniqueness of the solution to (∗) holds if for any
two solutions X 1 , X 2 to (∗) defined on the same probability space and driven by (the
same Brownian motion and) the same Poisson random measure N(dtdz), it holds
that
P (sup |Xt1 − Xt2 | = 0) = 1 .
t∈T

There are several conditions known so as that the pathwise uniqueness holds for
the solution to the SDE (∗). See [12, 77].

Proposition 1.6 (Yamada–Watanabe type theorem).

(1) If the equation (∗) has a weak solution and the pathwise uniqueness holds, then it
has a strong solution.
(2) If the equation (∗) has two solutions X 1 , X 2 on two distinct probability spaces
(Ωi , F i , P i ), i = 1, 2, then there exist a probability space (Ω, F , P ) and two adapt-
ed processes X̃t1 , X̃t2 defined on (Ω, F , P ) such that the law of X̃ i coincides with
that of X i , i = 1, 2 and X̃ i , i = 1, 2 satisfy (∗) on (Ω, F , P ).

For the proof of this result, see [190] Theorem 137. The proof in [190] is rather
complex. In case that the SDE is driven only by the Wiener process, the result is called
“Barlow’s theorem”. See [180] Exercise IV.40 (p. 246 in Second edition).
In what follows, we consider, in particular, the following SDE on the Poisson
space with values in Rd of Itô type

t 
c
xt (x) = x + b(xs (x))ds + γ(xs− (x), Δz(s)) , x0 (x) = x . (3.2)
0 s≤t
 Itô processes with jumps 27

c
Here, denotes the compensated sum, that is,
⎧ ⎫

c ⎪
⎨  t  ⎪
⎬
γ(x, Δz(s)) = lim γ(x, Δz(s)) − ds γ(x, z)μ(dz)
⎪
→0 ⎩ ⎪
⎭
s≤t s≤t,|Δz(s)|≥ 0 |z|≥

(cf. [101] Chapter II). Functions γ(x, z) : Rd × Rm → Rd and b(x) : Rd → Rd are

C ∞ -functions whose derivatives of all orders are bounded, satisfying γ(x, 0) = 0.
Equivalently, we may write xt (x) as
t t 
xt (x) = x + b (xs (x))ds + γ(xs− (x), z)Ñ(ds dz)
0 0 |z|≤1

t 
+ γ(xs− (x), z)N(ds dz), (3.3)
0 |z|>1

where Ñ denotes the compensated Poisson random measure: Ñ(dsdz) = N(dsdz)−


dsμ(dz), b (x) = b(x) − |z|≥1 γ(x, z)μ(dz), where the integrability of γ(x, z)
with respect to 1{|z|≥1} dμ(z) is assumed. We abandon the restriction (3.1) hereafter,
and resume condition (3.2).
We assume that the function γ is the following form:
∂γ
γ(x, z) = (x, 0)z + γ̃(x, z) (3.4)
∂z
for some γ̃(x, z) = o(|z|) as z → 0. We assume further
(A.0) that there exists some 0 < β < 2 and positive C1 , C2 such that as ρ → 0,

C1 ρ 2−β I ≤ zzT μ(dz) ≤ C2 ρ 2−β I.
|z|≤ρ

(A.1)
(a) For any p ≥ 2 and any k ∈ Nd \ {0},
 
 ∂kγ
p

 
|γ(x, z)|p μ(dz) ≤ C(1 + |x|)p , sup  k (x, z) μ(dz) < +∞ .
x  ∂x 

(b) There exists δ > 0 such that

- ∗
.
∂γ ∂γ
inf y ∗ (x, 0) (x, 0) y; x ∈ R d
≥ δ|y|2
∂z ∂z
on Rm .

(A.2) We assume, for some C > 0,

 
 
inf det I + ∂γ (x, z) >C.
d
 ∂x 
x∈R ,z∈supp μ
28 Lévy processes and Itô calculus

Due to the previous result (Theorem 1.5), the SDE (3.3) has a unique solution xt (x).
Furthermore, the condition (A.2) guarantees the existence of the flow φst (x)(ω) :
Rd → Rd , xs (x) → xt (x) of diffeomorphisms for all 0 < s ≤ t .
Here, we say φst (x)(ω) : Rd → Rd , xs (x) → xt (x) is a flow of diffeomorphisms
if it is a bijection a.s. for which φst (x) and its inverse are smooth diffeomorphisms
a.s. for each s < t . We write ϕt (x) of xt (x) if s = 0.
We remark that at the jump time t = τ of xt (x),

xτ (x) = xτ− (x) + Δxτ (x) = xτ− (x) + γ(xτ− (x), Δz(τ)) .

Hence,
∂ ∂ ∂
ϕτ (x) = I + γ (xτ− (x), Δz(τ)) ϕτ− (x) ,
∂x ∂x ∂x
and this implies
−1 −1 −1
∂ ∂ ∂
ϕτ (x) = ϕτ− (x) I+ γ (xτ− (x), Δz(τ)) .
∂x ∂x ∂x

This describes the movements of the coordinate system induced by x → xτ (x).

We observe that in order for the map x → ϕt (x) to be a diffeomorphism a.s., it is
necessary that x → x + γ(x, z) is a diffeomorphism for each z ∈ supp μ . Other-
wise, we can not adequately connect the tangent space at xτ− (x) to that at xτ (x) =
xτ (x) + γ(xτ− (x), Δz(τ)). To this end, we assume (A.2).

Lp -estimates
The Lp -estimate for the solution of (3.2) is not easy in general. Here, we provide a sim-
ple one. Suppose xt (x) is given by

t 
xt (x) = x + bt + γ(z)Ñ(dr dz) .
0

Under the integrability of γ(z) with respect to 1{|z|≥1} .dμ(z), we have the following
Lp estimate for xt (x).

Proposition 1.7 (cf. [119] Cor. 2.12). For p ≥ 2, we have

  -
E sup |xs (x)|p ≤ Cp |x|p + |b|p t
0≤s≤t
  .
p/2
+t |γ(z)|2 μ(dz) +t |γ(z)|p μ(dz) ,

for some Cp > 0.

Indeed, the crucial condition (3.22) in [119] follows from (3.4) and (A.1).
 Itô processes with jumps 29

Jacobian
In order to inquire the flow property of xt (x), we need the derivative ∇xt (x) =
∂xt
∂x (x) of xt (x). It is known that under the conditions (A.0–A.2), the derivative sat-
isfies the following linear equation:

Proposition 1.8. Under the assumptions (A.0–A.2), the derivative ∇xt (x) satisfies the
following SDE:

t
∇xt (x) = I + ∇b (xs− (x))ds∇xs− (x)
0
t 
+ ∇γ(xs− (x), z)Ñ(dsdz)∇xs− (x)
0 |z|≤1

t 
+ ∇γ(xs− (x), z)N(dsdz)∇xs− (x) . (3.5)
0 |z|>1

Proof. We skip the proof for the differentiability since it is considerably long ([119]
Theorem 3.3). The SDE for which ∇xt (x) should satisfy is given as below.
Let e be a unit vector in Rd , and let
1
X  (t) = (xt (x + λe) − xt (x)), 0 < |λ| < 1 .
λ

Then, X  (t) satisfies the following SDE

t t  t 

X (t) = e + bλ (s)ds + γλ (s, z)Ñ(dsdz) + γλ (s, z)N(dsdz)
0 0 |z|≤1 0 |z|>1

where
1 
bλ (s) =(b (xs− (x + λe)) − b (xs− (x))) ,
λ
1
γλ (s, z) = (γ(xs− (x + λe), z) − γ(xs− (x), z)) .
λ

We let λ → 0. Then, by the chain rule,

lim bλ (s) = ∇b (xs− (x)) · ∇xs− (x)a.s.,

λ→0

and
lim γλ (s, z) = ∇γ(xs− (x), z) · ∇xs− (x)a.s.
λ→0
30 Lévy processes and Itô calculus

for each s . Hence, we have

t
∇xt (x) = I + ∇b (xs− (x))ds∇xs− (x)
0

t 
+ ∇γ(xs− (x), z)Ñ(dsdz)∇xs− (x)
0 |z|≤1

t 
+ ∇γ(xs− (x), z)N(dsdz)∇xs− (x).
0 |z|>1
The equation (3.5) is fundamental in considering the stochastic quadratic form
(an analogue of Malliavin matrix for the jump process). See Chap. 2.
The derivative ∇xt (x) depends on the choice of e ∈ Rd , and is indeed a di-
rectional derivative. We denote the Jacobian matrix by the same symbol, namely,
∇xt (x) = ∇x xt (x).
Here is an expression of det ∇xt (x) which will be useful in the analytic (volume)
estimates.
Let
t d
∂ 
At = b (xs (x))ds
i=1
∂xi
0
⎡ ⎤
t  d
⎣det(I + ∇γ(xs− (x), z)) − 1 − ∂
+ γ(xs− (x), z)⎦ dsμ(dz)
i=1
∂xi
0 |z|≤1
⎡ ⎤
t  d
⎣det(I + ∇γ(xs− (x), z)) − 1 − ∂
+ γ(xs− (x), z)⎦ N(ds, dz) ,
i=1
∂xi
0 |z|>1

t 
Mt = [det(I + ∇γ(xs− (x), z)) − 1] Ñ(dsdz) .
0 |z|≤1

We then have

Lemma 1.1.
det(∇xt (x)) = E (A)t · E (M)t . (3.6)
 Itô processes with jumps 31

Proof. By Itô’s formula,

t d
∂ 
det(∇xt (x)) = 1 + b (xs (x))ds det(∇xs (x))
i=1
∂x i
0
t 
/
+ det((I + ∇γ(xs− (x), z)) · ∇xs− (x))
0 |z|≤1
0
− det(∇xs− (x)) Ñ(dsdz)
t 
/
+ det((I + ∇γ(xs− (x), z)) · ∇xs− (x))
0 |z|>1
0
− det(∇xs− (x)) N(dsdz)
t 
/
+ det((I + ∇γ(xs− (x), z)) · ∇xs− (x)) − det(∇xs− (x))
0 |z|≤1

d
∂ 0
− γ(xs− (x), z) · det ∇xs− (x) dsμ(dz)
i=1
∂xi
t
= 1+ det(∇xs− (x))d(As + Ms ) .
0

Hence, we apply the Doléans’ exponential formula and Girsanov’s theorem (Theo-
rem 1.4) from Section 1.2.3. In view of [A, M]s = 0, [A + M] = [A] + [M] and we
have the assertion.
Using (3.6), it is possible to show the boundedness of E[supt∈T det(∇xt (x))−p ]
by a constant which depends on p > 1 and the coefficients of the SDE (3.2).

Inverse flow
Using the Jacobian ∇xt (x), we can show the existence of the inverse flow xt−1 (x) as
a representation of the flow property. By the inverse mapping theorem, this is due to
the (local) invertibility of the Jacobian ∇xt (x).
Associated to the above mentioned process ∇xt (x) = U (t):

t t 

U (t) = ∇b (xs− (x))U (s)ds + ∇γ(xs− (x), z)U (s)Ñ(dsdz)
0 0 |z|≤1

t 
+ ∇γ(xs− (x), z)U (s)N(dsdz) ,
0 |z|>1
Exploring the Variety of Random
Documents with Different Content
sufficiently early to see a herd of more than a thousand antelopes
that were going to the watering-place. My huntsman, also, who had
struck into another road, saw some hundred together; all the others
agreed that there were these thousand which I have mentioned. But
they soon dexterously divided to the right and the left on the
immeasurable level of this land, where there was merely low grass,
wild bamie and a quantity of basil, which latter was also met with on
all sides in the countries further up; and Suliman Kashef only shot
four, and my Sale not a single one. I myself could only see some
antelopes on the horizon, because it was already getting dusk, and I
stopped with Sabatier close to the vessels, in case some beast
should be scattered from the herd, but in vain. On this occasion,
also, I saw two lions at a distance.
At night the wind blew in coldly at the door and windows, and
even this morning the north-east wind was cool. At half-past six we
proceed E.N.E., and in a bend further to the right E. and E. by S.,
where we make a stronger evolution to the right. Eight o’clock.
Libàhn from S.E. by S. to S. We glide over shallows apparently
consisting of rubble-stone; the wind becomes strong and tosses the
waves. A quarter before nine, S.E. by S. to S., then still more to the
left, where we are soon thrown by the wind on the left shore, and
stop in E.S.E. Thibaut is with me, and they are calling for him; his
ship is full of water, and all the crew are summoned there: it is
fortunate that we are near land. Selim Capitan neglected to have the
vessels caulked at Khartùm, or to order at least gotrahm (instead of
tar) to be applied to the parts which we had stopped up with some
oakum.
At five minutes’ distance above, a large village deserted by
people; we are magnanimous enough on our side to keep the crew
from plundering it. It is slightly elevated: the same is also the case
with the shore, so that shallow lakes are formed right and left, at
present dry, and having vents to the water, which apparently are kept
open by human hands for the sustentation of the soil,—on which,
however, nothing is seen. A number of snail-shells are lying together
on the surface just as I have seen in other places, and it seems that
snails are eaten. We remain here on account of the accident to
Thibaut’s vessel, but the shores, à talus, do not allow us to bring it
on the dry land. Thermometer 17° and 24°.
 CHAPTER X.
VARIOUS SPECIES OF GRASSES. — FORMATION OF THE SHORES. —
WATERFOWLS. — AN ANTELOPE OF THE TETE SPECIES, NOW AT
BERLIN. — STRATA OF THE SHORE. — THE SOBÀT RIVER. THE MAIN
ROAD FOR THE NATIVES FROM THE HIGHLANDS TO THE PLAINS. —
OBSERVATIONS ON THE COURSE OF THE NILE AND SOBÀT. — A
THOUSAND ANTELOPES SEEN MOVING TOGETHER! — WILD
BUFFALOES, LIONS, AND HYÆNAS. — AFRICA, THE CRADLE OF THE
NEGRO RACE. — THE SHUDDER-EL-FAS: DESCRIPTION OF THIS SHRUB.
— ARNAUD’S CHARLATANRY. — OUR AUTHOR FEARED BY THE
FRENCHMEN. — ARNAUD AND SABATIER’S JOURNALS: THE
MARVELLOUS STORIES OF THE FORMER. — THIBAUT’S JEALOUSY. —
VISIT OF A SHEIKH OF THE SHILLUKS. — FEAR OF THE TURKS AT THESE
PEOPLE. — SULIMAN KASHEF PURSUED BY A LION.

18th March.—We sail at a quarter before seven o’clock with a

cold north-east wind S.E., and then S. by E. and S. The wind,
however, becomes too powerful; twice are we driven on oyster-beds
—that is, on those thorn muscles, as if over stones, and have reefed
sails to prevent the ships from going to pieces, their condition being
so bad. A quarter after seven. From S. by E. to the left. Visible sand-
banks in the curve seem to block up our road, but we managed to
pass by them on the right, with the assistance of the sails, close to
the left shore towards S.E., and away in the bend to E. by S. Half-
past seven. S.E. by E., and then a quarter before eight right round;
six huntsmen’s tokuls being near a sand-bank on the lower shore of
the projection of the left side of the river. We halt on the left by a
shallow island clothed with low verdure towards S.S.W., and intend
to stop here to-day and to-morrow to make observations, and the
most needful repairs to the vessels and sails.
Suliman Kashef shot yesterday evening, at a gazelle-hunt, a
large antelope, called by the Arabs Tete, in the foreleg, shattering it
to pieces; the animal fell twice, but made off at last on three legs.
Sabatier and I had chosen the left wing, and concealed ourselves
with one of my servants in the high grass: the former fired and
missed. The cracked earth displayed a magnificent soil: the grass,
standing thickly in tufts, reached up to our breast, and was a great
obstacle at the beginning of our rapid march. It was still green at the
bottom, and the present desiccation of the ground, on which we
remarked everywhere the traces of footsteps of wild beasts, and
their dung, might therefore have only taken place a short time. This
grass, narrow and three-edged, with cylindrical spikes, formed the
principal produce of the soil. Less common was the grass similar to
our species with flat two-edged leaves; it had knotty stalks like the
three-edged, but a couple of spikes grew together on each blade (I
have remarked this previously), which unfortunately were not yet at
maturity, and therefore very small. The third species of grass
consisted of slender reeds, cropped and sprouting anew, or trod on
the ground. I perceived, also, some bamie growing wild, and birds’
nests of grass hanging on it.
I had lost sight of my comrade; and although at the
commencement of my excursion I had seen the vessels sailing up
the river at my side, it soon became dark. Suliman Kashef, however,
had the sagacity on his return to the vessels, to order the reeds to be
set on fire as signals, so that luckily I found my way back, though
sinking every now and then up to the knees into the deep foot-prints
of hippopotami close to the river,—a further proof that the shores,
being only slightly elevated, form shallow lakes here at the rainy
season, which are not dried up so soon.
I had taken a short walk previously on the left shore. The very
same appearances of water remaining behind were visible, and I
found muscles on the dry ground, amongst which were the Erethria
ovata. Long traces of little deposits of earth, which, on closer
examination, I discovered to be dams against the high water on the
shore itself; and the alluvial reeds in conjunction with the muscles,
make me conjecture that the Sobàt ascends over its shores here, as
in many other places. Behind these low deposits lay an unlimited
stubble-field on the other side of the village which lies on a gently
ascending hill, elevated perhaps by the remains of clay walls, and
stretching far beyond the horizon. The better kind of tokuls have
frequently a roof, but the eaves only project inconsiderably: the
smaller ones have a round form of roof, low sleeping places and
reed-hedges being between them. Dinkas are said to dwell there; but
not a person, not a living creature, is to be seen. Thermometer,
sunrise, 17°; noon, 28°; sunset, 26°.
19th March.—We all dine together in the open air, after an
antelope-hunt. The island on which we are, is, properly speaking, a
large broad sand-bank, about a quarter of an hour long: its
somewhat elevated back is covered with verdure, and is connected
with the shore on the right at low water-mark. Purslane (Arabic Rigli)
is found very commonly upon it; we see also numerous birds fishing
in the many tongue-shaped segments of the upper part, and, in fact,
sharing among themselves the narrow lake on the high right shore,
close to which is a village, from whence the people have likewise
fled. These feathered occupants seem to remove very seldom from
this happy place. The antelopes presented themselves in great
numbers; but Suliman Kashef’s body-guard, though generally good
shots, did not know the huntsman’s custom of dividing and forming a
chain, so as to catch the herd in the middle. I had no inclination,
either, to join in such a surrounding of the game; for these Turks fire
as if they were shooting in files, and their guns carry far, and are
always recommended to the care of the supreme Allàh.
20th March.—Departure at a quarter after nine o’clock, with a
favourable north-east wind, without sails, S.S.W. and S.W. by W.,
where, on the right, behind the high shore, a village lies in the bend
to the left, and below it a broad sand-bank, on which some long-
legged water-fowls are wandering about. We leave at the right side
another sand-bank exactly similar to the former, throwing its shallows
far beyond the middle of the river, and halt, S.S.E. at the right shore
at half-past ten o’clock. Suliman Kashef’s halberdiers bring eight
antelopes, one of which I procure, being the largest of the Tete
species. This specimen is now in the Zoological Museum at Berlin as
a nova species.
 The shores have widened here, and fall off in an angle of 45° to
50°: though they appear on this account lower, yet it is plainly visible
by the steeper places, that they always become higher. It is only
below in the places where the river beats against, that the bluish clay
is seen: the remaining part of the shores has, apparently, merely
constituents of the same, as is the case in most places where the
high water has not washed away the crust of humus crumbling from
above and covering the base of the surface; for the original soil
discloses itself immediately under the covering of earth, as is seen in
precipices, and clefts in rocks caused by water. The river has also
thrown or deposited thick layers on the shores. We must not be
deceived here by observing various strata of earth mixed above and
below with sand; this is a later alluvial deposit.
A pure layer of clay is never to be seen, however, in these tracts
of strata, so far as I have remarked here and on the Nile. If it does
appear, it lies either as the foundation of the whole, below on the
banks of the water, as on the Nile; for all the ground there is alluvial
and earthy deposits, gained when the high water is drawn off; or it
rises, as in the Sobàt, with the talus of the shores to the surface,
which is covered with a crust of humus. The Sobàt dug a bed for
itself in firm clay-ground that resisted the water, and remained
tolerably constant in the trench opened by it, without having altered
its course, for no gohrs are seen on dry ground; but perhaps, in
some places, it has flowed over its bed, and formed channels. On
the contrary, the White Stream wallowed for a long time in the deep
slime of an emptied lake, before it threw up solid dams, on which
there are marshy forests, as on the old shores. This long valley-
basin lies also on a layer of clay.
The Sobàt may be considered as a further boundary of the
peninsula of Sennaar, and have given to the latter the name of
Gesira. Certainly it has been, like the Blue Nile, a main-road for the
tribes of the highlands of Ethiopia to the valleys of these countries;
and this must have been especially the case because it has no
accompanying marsh-lakes. Such nations could not have wound
down from the mountains of Bari and the highlands there, by reason
of the many marshes; for we are not to suppose that nomadic tribes
can provide themselves and families with a stock of provisions for a
long journey, or stow entire herds in their hewn-out trunks of trees
(canoes); and it is impossible that the cattle could have been driven
along the shore for their use.
The further I ascend the Sobàt, the plainer I perceive why the
right shore just behind Khartùm appears higher than the left, and
why I could not get rid of the idea that this oblique inclination of the
land was in opposition to the course and the mouth of the Nile, but
still might be explained. The deposit of particles of earth and sand
can only come from above, and will always try to level and equalise
the tracts of land which the Nile covers with showers of rain, brooks,
and rivulets. It is clear that the surface is elevated by that means,
and that, where these washed-away and liquid particles of earth
reach a stream like the White Nile, they are carried down by it,
without the other shore (the left side of the Nile here) deriving
naturally any advantages therefrom.
The high mountain chain of Fàzogl and Habesh mixed, as I
conjecture, its collective waters, owing to a breach in its partition-
walls, and their slime and morasses, and perhaps entire hills of
decayed and corrupted matter connected therewith, filled depths in
the lower valley on the side of the Nile up to the Delta—its most
famous memorial,—and levelled the mountains of the
neighbourhood, when Bertat, Dinka, and the country between the
Sobàt and Bari rivers might have shot up in indomitable strength like
artesian wells. Such catastrophes roll mountains and masses like a
brook does its little pebbles, and throw up the water released from
confinement in the cavities of heights which attract and collect it. A
flood of liquid earth rolled then far and wide from the mountains
without order and with numerous arms, but conformably to nature,
the heavy particles sank. The water itself washed away, smoothed
and levelled the ground. Therefore now we perceive those
immeasurable plains on the Sobàt, whereon beasts cannot hide
themselves, and which would be without shelter in the rainy season,
if there were not mountains and forests in the neighbourhood.
 Though it be mathematically proved that the great Nile runs in a
channel as upon an ass’s back, yet we find just the contrary in the
White Nile; but the Sobàt even displays that phenomenon, although
not at this moment, for its shores are emptied, except in the
lowermost grade. They lie and stretch higher than the adjacent land,
being heaped up by the waves of the river; they are, however,
generally narrow dams, only appearing wide in the places where
there are shallow lakes behind in distant connection, or overgrown
gohrs, the grass border of which more easily withstands that deep
washing away than these immeasurable plains, which might be
called beautiful from their splendid soil, if Ceres waved her golden
ears, and Pomona offered shade and fruit. They shew, indeed, but
little declination to the Nile, for which the Sobàt itself affords the best
standard, being stagnant, and its shores only increasing in height
here and there. The shores become higher, as on the great Nile
itself; the less precipitous ones (although this is only local) are
deceptive, as I have remarked several feet difference on the disrupt
shore, and still more on the return voyage. I cannot divest myself of
the idea that a lake has stood here also, or it may be that the surface
of the earth from the region above to this, has been laid flat by the
inundation, similar to the level fields of Egypt.
There is an incredible number of deer on the shores of the Sobàt,
for I can add from my own conviction, so far as my eyes and ears do
not deceive me, that I saw herds of antelopes at least a thousand
strong—the Turks say from three to four thousand. About evening
they shew themselves in immense lines on the bare horizon of the
steppe, stand still, and approach—their tread sounds, in truth, like
the evolutions of distant cavalry; at last, as soon as it is dark, they
separate in the little bushes on the margin of the shore, to descend
to the water. Hitherto I have not been able to seize this opportunity,
because no one would remain with me on account of the lions and
other savage beasts prowling about here, and it did not seem to me
exactly safe, by reason of my close acquaintance with the lion and
his just revenge, to lie alone behind a bush, and shoot some of the
animals at a few paces off. My cook, however, has promised to
accompany me on such sport, when we come again to a suitable
place.
We proceeded this afternoon at two o’clock with sails in a south-
easterly direction, and halted for the sake of antelope-hunting at
three o’clock, at the right shore, E.S.E., before the curve to the right.
Four o’clock: we sail to S.S.E., then a bend to the left, S.E. by E.,
where we stop again to hunt and also to remain. An antelope-herd of
about sixty head was standing shortly before at a little distance on
the shore, like a flock of goats, in harmless innocence and
anticipating nothing evil. Wild buffaloes, lions, and hyænas were
seen by several of our men: where are the holes and corners in
which these beasts lurk? there must be mountains. Thermometer,
sunrise, 15°; noon, 26°, 27°; from four to five o’clock, 28°; sunset
26½°.
21st March.—We sail from our place at half-past six o’clock, quite
quietly, without a drum being beat, and go E.S.E., with a cold north-
east wind. Half-past seven.—From E. Libàhn: again to the left,
E.S.E.; eight o’clock: and then again to the left at nine, N.N.E., where
we stop at the left shore for sport, pretending that it is owing to the
contrary winds.—Oh you Turks!
M. Arnaud now pays himself in skins, which he demands very
freely for “son Altesse,” but which he intends to sell in France at 500
francs the head. Eleven o’clock: we go with libàhn round a corner,
and come to N., whence the river winds again slowly to the right,
N.E. by N. and to E., where we are driven forward and indeed
without sails. I remark here that the lower border of the river is rocky,
with a layer of soft stone under the clay, from which it may originate,
and appearing to be limestone formation, until we break off the
deceitful exterior, and the apparent corrosion by the action of the
atmosphere turns out to be alluvial deposit,—the clay, however,
remaining sticking to the hands like sand. One o’clock: E.N.E. All the
vessels bear up! Four o’clock: three reïs had been sent forward with
the sandàl to sound the water, and the men were honourable enough
to express their conviction that there was still a watercourse: we
shall get now regularly fixed upon every sand-bank. I, for my part,
would like to make yet a good way, for I may stumble perhaps upon
a firmer foundation of stone or something else new. We proceed,
therefore, further,—against Mohammed Ali’s will, and certainly
against Ahmed Basha’s, who may be very much in want of vessels
and men at this moment, without any invasion on the part of England
or France. E. by N., and we squat again. A quarter before five
o’clock: S.E. by E. Water is on the right, but we cannot get there
because the bed of the river is elevated in the middle, and these
banks are magnificently larded with the spiry Conchylia, which would
deprive the vessels instantaneously of their beautiful caulking. The
reïs are sent out again in the sandàl; everyone is in doubt what is to
be done. Suliman Kashef and Selim Capitan want to return. An
island lies in view above. I should like to be there to make
observations, but that cannot be. Thermometer, 17°, 28° to 30°, 27°.
22nd March.—Suliman Kashef sails back for the sake of the
chase; two other vessels follow: obedience seems renounced. I go to
the island and shoot two antelopes, dark as it was. No artifice, such
as I have already related, was necessary in this neighbourhood.
23rd March.—We set out indeed at seven o’clock in the morning
for the return voyage, but stop soon again, notwithstanding the
favourable wind at the right shore, because deer are seen close to it.
The Sobàt and the Blue river might be conduits for the high land, like
the Tigris and Euphrates, Indus and Ganges were for the valley-land,
made subsequently fit for the nourishment of nations. Tradition and
history up to our time, teach us that Nature was not powerful
enough, and perhaps did not wish it, to form everlasting barriers
between nations, whether seas, rivers, or mountains; for the
destination of man is perfectability, which can only be attained by
mutual commercial intercourse.
Was Africa, therefore, although in the same latitude as other
countries,—for example, Arabia,—exclusively created for the black
species, who, so far as I have had experience of them, will never
leave the low grade of intelligence in which they have been for so
long a time, until they come into a closer and more continuous
association with whites? Anthropophagy, indeed, makes the Nile the
partition-wall between Asia and Africa, instead of the Isthmus of
Suez and the Red Sea. Our Sobàt (as also perhaps the White river,
unless the black people from below ascended here after the drying
up of that part assumed by me to have been a lake, which may be
almost taken for granted if we consider the affinity of languages from
the Dinka country to Bari,) drew down probably only a black race,—a
younger stem, I conjecture, than that of Caucasus,—between which
and the nations of the Nile there is a total difference in colour and
manners.
Habesh, like a second Cashmere, might be the cradle of white
men; no less than Arabia, lying opposite, which has nourished
perhaps the same species, but burnt by the sun;—and if a black race
shot up here instead of the white, the rivers of High Asia that
disembogue on this side were large enough to import white people.
The Red Sea and the Persian Gulf must, indeed, at the very earliest
time, have formed nearly insurmountable obstacles, if antediluvian
shipbuilding, even without Noah’s ark, had not brought about
cultivation, and caused the necessity of exertion and the desire of
emigration; and who will answer for it that the Red Sea was not at
one time a Tempe, so that an excursion to Africa might have been
made under the shade of rose and orange trees? The divulsed
coasts, the washed-away sharp rocks of the dangerous shallows,
and the submarine mountains, might easily indicate that there was
once a small irruption.
Habesh might have been the real mother country of Nubia and
Egypt, by means of the fecundation brought from Asia. Nuba
negroes mixed with the whites, and became masters of the country,
until they were forced back to their mountains in Kordofàn,—still
remaining a pure race, and nothing being left of the whole Egyptian
policy but circumcision. The black nation on the Nile might have
been separated from that in Habesh by a cordon; and both of these
tribes preferred, on cultivation increasing, a comfortable life in the
city to the free one in the mountains, until the Romans connected
again Habesh and Egypt, and built Axum.
The shores of the Sobàt have hitherto been without wood; but I
saw, upon the shore behind the island just left by us, a row of trees,
said to be Döbkers, and probably were so; for I recognised the little
solitary shrubs, in which I and my servants had groped about till late
at night, to be young shoots of Döbkers, which seem to have sprung
up this year at high-water; and also nebek and telle. I found water-
thistles on the shore above in unusual abundance, and collected
also that clay which appeared to me previously to be a layer of
stone. It was not quite free from fine sand, which the water had
washed on the shore, and exhibited, therefore, a rough, stone
surface, without being so yet. But where it lay piecemeal about the
shore it was as hard and black as stone saturated with water, and
dried or burnt by the sun, for that luminary has the power, with the
assistance of rain, of vitrifying rocks; and we have seen ourselves
such greenish ridges and blocks in the Nubian deserts.
According to Girard, the specimens I brought from the shores of
the Sobàt consist of a micaceous sand, dark-brown, ochrous clay,
chalky sand, and partly of a conglomerate composed of small
fragments of limestone baked together by the sun. The sand, where
it is pure, consists of several little yellowish grains of quartz, a small
portion of reddish feldspar, some brown iron-stone, little brown
tombac-mica, and a black mineral consisting of small grains, the
nature of which could not be exactly ascertained. These materials
indicate that the origin of the sand is derived from a mica-slate and
gneiss mountain not far distant; for if the sand were far from the
mountain whence it originated, it would not contain coloured mica.
An incredible number of potsherds were found below on the flat
shore which is deserted by the river. I thought to pick up pieces of
brick, but I soon discovered that they were fragments of that kind of
murhàka which the Shilluks bake from the mixed mud of the Nile,
owing to the want of stones. I did not discover any remains of
villages on the shore; they appear, however, to lie upon the high
gohrs, and the pools arising therefrom, as if upon the primitive shore.
These gohrs of the Sobàt running parallel, and lying actually twenty
feet higher,—at times, also, displaying higher dams or shores than
the river itself, indicate moreover that the river in its young days
flowed wider than it does now, without deviating from its direction, so
far as I have hitherto observed. I had an opportunity of convincing
myself of what I have stated above, at the village in the
neighbourhood of which I was shooting, and where I saw two
ostriches. Nevertheless these potsherds are testimonies either of a
previous and considerable population of the surrounding country,
and even the hill covered with shrubs above the potsherds might
have supported a village, or that more people dwell above than we
should have supposed; and the pieces of murhàkas may be a sign
that neither stones nor mountains are near at hand, and that the arts
did not flourish; for all these fragments were of rude origin. We have
never discovered traces of any finer vessels of terra-cotta, or pieces
of glass, enamel, or delft ware, such as we find in Egypt.
When the antelopes come to drink, for which they have room
enough,—for instance, yesterday evening—on the extensive plain,
they appear drawn up as if in line of battle, but run away like a flock
of sheep—yet not so close together—as soon as danger threatens.
These animals, so numerous here, might be easily tamed and fed by
the hand of man, if arms did not prevent such friendship; the lion and
hyæna, besides, keep them in constant fear and trembling, and take
a tenth part and more of them, in spite of all their innocence, as we
have seen by the remains of saraffes and gazelles. Perhaps, the
great lord of beasts, with whom I came so closely in contact that it
was only necessary for him to make a spring, to set his teeth in my
bit of Adam’s flesh, had made such a feast on venison before we
met.
It is surprising that I have not hitherto seen any mouse-holes or
mole-hills; the inundation and rains may be the cause of this, and we
have not gone from the river into the interior of the country. The
Schudder-el-Fas (axe-tree, mimosa sensitiva?) common in Bellet
Sudan, is seen on the shore itself,—a shrub spreading like brambles,
with small delicate acacia leaves, a similar sort of pods, flesh-
coloured button-flowers, and barbed thorns. If its twigs be touched
with an axe, or a knife, or merely knocked with the finger, its leaves
immediately close (from fear, as the Arabs say) as if they were
withered. I have not yet been able to procure ripe seeds of this
shrub.
M. Arnaud makes a section, but not where the water running
upon a shallower place allows a more accurate calculation to be
made, but there where the river is full,—namely, in these basins, in
which the water underneath is stagnant, whilst that on the surface
moves as slowly as we have seen the Atbara, which has the very
same deep basins. The public will therefore read of a mighty mass of
water, said to be carried by the Sobàt to the Nile: the breadth of the
river amounts here, as below, to one hundred and thirty mètres.
The section is therefore made, and we proceed at noon. The
strong wind breaks in two the sail-yard of the vessel commanded by
the Arnaut, Mohammed Aga, because Selim Capitan paid no
attention to it, though it was rotten,—and so was the whole vessel.
We land afterwards at our peninsula or sand-bank, where there is
a settlement of six tokuls. The poor inhabitants of these little straw
huts were perhaps those whom Thibaut observed when setting out
from hence, and who covered the whole island like a swarm of flies;
whereupon he slily retreated. I rather think that my European fellow-
travellers are afraid of my being an eyewitness of their doings and
acts, lest I should accuse them of falsehood, or laugh at them when
they exaggerate some small circumstance, or pretend to have seen
wonders, and gone through frightful adventures with courage: and
yet they give all this out of their journals to their attentive listeners. It
is really as good as a play to see them. Arnaud and Sabatier have
separate cabins in the same vessel; and, as soon as one has gone
on deck, ashore, or on board another vessel, the other tries to find
out the hiding-place of his journal, in order principally to read what is
said of him. If Arnaud abuses Sabatier, the latter gets in a rage; but
on our persuasion rests satisfied with threats, and sits down and
writes against Arnaud in his journal: then he leaves it carelessly lying
about, and goes away directly Arnaud enters, and the latter,
immediately his back is turned, seizes it and reads it. Thus they read
their notes and diaries in secret, and then I have the pleasure of
hearing all about it from Thibaut and Sabatier, who come on board
my bark.
Though Sabatier and Arnaud have related the history of their
hunting two little marafills, the latter has not hesitated to write down
that he came close upon two lions (whereas he only saw two
hyænas running at a distance, whom Sabatier observed at the same
time), because he is jealous, as his countrymen tell me, of my
chivalry with the lion. He describes, also, antelope, lion, and
crocodile hunts with similar Münchausen stories.
Our tokul inhabitants seem to have gained a little confidence,
although they do not shew themselves, for they have tied up a
sucking-calf close to their huts, which Suliman Kashef takes
immediately into his possession. We have here a proof that the
water is still falling, which I had not previously thought: this was the
case in the peninsula to the extent of a foot, and we shall be obliged
to go over the nearest oyster-bank, after we have remained here half
a day, or perhaps the whole, to make the Arnaut’s vessel again in
proper sailing trim. Thermometer 16°, 27°, 29°.
We find here, near the Sobàt, that the impregnation of iron oxyde,
which is frequently so strong, is only in the earth brought down from
above, in all the humus, and also in the coagulated layers of sand
under or between the humus, but not in the drifted sand. There is no
trace of it in the clay; it is therefore perhaps a primitive deposit,—the
lime or cement of the ancient world. From this appearance of iron
oxyde, which is not inferior to that on the upper part of the White
river, the Sobàt might provide Bellet Sudàn and Egypt with iron, if the
protecting hobgoblins of the mines be first driven away. Whether the
Sobàt leads to gohrs of gold would be worth investigation, but the
sand of this region has not the least appearance of it. For that very
reason Arnaud is projecting an immediate expedition up the Sobàt
this summer to the gold sources. Oh enviable Viceroy of Egypt,
around whom charlatans of every kind have drawn a line, within the
circle of which no one can penetrate but Italians of similar calibre,
may your eyes be closed against gold-dust! At noon we set out from
our halting-place, partly with libàhn and partly sailing, glide here and
there over oyster-banks, and soon stick fast again on the sand.
25th March.—Sate Mohammed and Sale lay down yesterday to
sleep close by, instead of going to shoot. I felt hungry, and had not
prepared anything: the cook gazed stupidly in the distance; and I
was obliged to eat soft biscuit and drink water. No sooner had I taken
this than I felt unwell, and continued so till this morning. Thibaut
seems to be possessed with the most trivial jealousy; he cannot
understand, like all the rest, what I have to write so continually. The
old monotony of the shores has hitherto continued. Only this morning
two groups of naked Dinkas shewed themselves on the shores,
without running away from us. They brought even goats and sheep
with them; but we sailed by, and halted at noon near the dhellèbs,
where several groups of fishermen’s huts stood here and there on
the shore. Thermometer 17° to 30°; sunset 28°.
26th March.—The wind is favourable; but Arnaud wants to renew
his sections, in order to make the world believe that the Sobàt gives
more water than the White river. He is fishing about, therefore, in the
deep places, and a fall has made its appearance, which was invisible
to other eyes. Yesterday evening I said to him, accidentally, that I
had an attack of fever the day before, accompanied with vomiting
and diarrhœa. He immediately complained of his health, and said
that he was suffering under tertian ague, and dropped a hint that
Khartùm was very unhealthy, in which I agreed, experience having
taught me the same; but I saw clearly what he meant—he intends to
play the invalid, and will certainly become worse in Khartùm itself, so
as to induce the Basha to let him go to Metemma, where he will find
more lucrative work in examining the ruins there than in composing
his map. He eats and drinks, however, like men in good health—is
active on his legs, even when there is nothing doing—shakes his
round head, talks to himself, lays his finger on his nose, and looks
first to the sky and then to the ground so quickly that his straw hat,
adorned as it is with a large knot of ribbons, in the Tyrolean style,
flaps up and down. A Turkish under-cap and this hat cover his partly
bald head, on which he has let a pigtail grow, to make up for the loss
of his hair at a future time. He stands still, and then sets off running:
we call him, but his profound meditations will not allow him to
answer. The Turks say, in short, that he is magnuhn (crazed). His
usual expression to the others, “Je le sais tout,” is worthy of a man of
such varied acquirements as he pretends to be; but he does not dare
say so to me. When the time comes that he puts on the appearance
of sickness, he orders Thibaut to be summoned, and makes the
latter understand, if he does not perceive directly the illness of his
lord and master, that he is unwell, and not inclined for work, and
therefore wants to have a little chat with him. Thibaut sits down very
quietly, yawns now and then, and answers “Oui Monsieur” to all his
remarks, and thus are the rays of these great minds reflected. But
Arnaud truly is an all-comprehensive genius! After he has spoken a
little about the way to make money, &c., it pleases him to fall into a
kind of somnambulism, which will be attributed subsequently to
feverish delirium—talks some hodgepodge about constellations and
declining spheres, which he means, however, to put in order, and
then all at once turns the conversation from stars to his property in
the moon, or to a royal princess, whose favours he has refused—all
in most beautiful harmony with what he tries, at other times, to
impose on the world. During all this scene Sabatier sits in his cabin,
and is nearly bursting with laughter. “Relata refero.”
We navigate part of to-day with a good wind, and may, therefore,
easily reach the Nile. The more I ascend the Sobàt, the more I am
convinced how the shore and land fall away towards the Nile; and
yet the former appeared to me high at first, without, however, being
much higher; but then we came from the Nile, which is shoreless.
The Sobàt on the tract navigated by us is like the shore of the Nile
near Kàhira, and the more ancient high shores in Nubia. The greater
gradation of its shores at a distance indicate a chain of mountains in
the neighbourhood, or an unusually strong falling away of the slope
in the highlands, which is certainly not far distant.
That the Sobàt should still retain its old channel when on the
point of discharging itself into the Nile, as it seems to do at the upper
part, is very improbable, because I saw to-day from the deck merely
a gohr on its left side, at the end of the forest above, and perhaps
flowing into it as usual at high water. Moreover, the alluvial soil at the
lower end of the Sobàt was contrary to the nature of such soft
deposits, and very certainly belonged to that which I have
considered a Nile lake, the reach of which might always elevate the
Sobàt sufficiently high. The undulating ground extending from the
river, and running parallel to it, displays still furrows caused by the
marshes having ebbed away, and the Sobàt, rushing over it
afterwards, flowed without opposition wherever it liked, and might
even have separated into several arms.
27th March.—Yesterday afternoon we arrived at the mouth of the
Sobàt, and remain here to-day to make observations; but if the
strong north wind do not change, we shall not be able to advance.
Yesterday evening another man was carried dead from our vessel,
who would not eat, and drank nothing but water from the Sobàt. A
species of marsh hyacinth, having scent, was found by Thibaut. It
has little bulbs, but spreads and increases by shoots from the root.
Arnaud, who is never at a loss for a name, declares that a beautiful
water-bird, with perfect web feet, which we remark, is the Phœnix.
The Shilluks came this morning from the opposite shore, and
asked us to barter with them. This afternoon a Sheikh came to us
(unless he was an impostor, as the one on our ascent was) and
invited us to come without fear to their city, which was close at hand,
and said that cows, poultry, &c. would be brought to us. He himself
presented us a cow, for which we gave him a ferda. At four o’clock
we were able to proceed with oars: we navigate to E. and E.N.E. The
shores are surprisingly elevated since we were here, and a
considerable drainage of water, probably the last one, seems
especially to have taken place within these few days. The Nile is not
broader here than the Sobàt. Dry grass-reeds at the right; large
hillocks of decayed ant-castles at the left. Five o’clock, N.E.; the foot
of an island on the left, behind which the river becomes broad, and a
group of Shilluks are stationed there on the ground, whilst further
below they have taken possession of the ant-hills, and stand in the
form of a pyramid. The Turks are dreadfully afraid of these people,
and say that they are “batalin” (bad), because they are aware that
the Shilluks know them well. Feïzulla Capitan has had a number of
cartridges made for his pistols, and Selim Capitan wears his thick
white woollen trousers, and not without cause.
Soon after five o’clock we stop, conformably to the wisdom of our
high dignitaries, at the uninhabited right shore, which is considerably
higher than the left, and belongs to the country of the Dinkas, whilst
the Shilluks appear to possess the mouth of the Sobàt, although not
protected by forts and guns. Suliman Kashef was pursued a few
days ago by a lion, notwithstanding he had four halberdiers with him,
but he happily escaped the danger. To shew his gratitude to Allah, he
gave every one of his crew twenty-five piasters.
 CHAPTER XI.
THE SHILLUKS, A VITIATED PEOPLE. — CAUSE OF THE VIOLENT RAINS IN
INNER AFRICA. — REFUSAL OF THE SULTAN OF THE SHILLUKS TO VISIT
THE VESSELS. — DESCRIPTION OF A SPECIES OF GRASS. — BARTER
WITH THE SHILLUKS. — CONQUEST OF THEIR COUNTRY NOT DIFFICULT.
— FORM OF THEIR BOATS. — AMBAK RAFTS. — IRON RARELY FOUND
AMONG THE EGYPTIAN ANTIQUITIES. — WORSHIP OF TREES BY THE
SHILLUKS: THEIR RELIGIOUS RITES. — STARS IN THE SOUTHERN
REGIONS OF AFRICA. — SHILLUK WOMEN: THEIR DRESS. — REFUSAL
OF THE MEN TO SELL THEIR ARMS. — THE BAGHÀRAS: THEIR DRESS,
ETC. — RE-APPEARANCE OF THE ISLAND PARKS, AND MOUNT
DEFAFAÙNGH. — ASCENT OF THIS MOUNTAIN, AND FULL DESCRIPTION
OF IT. — THE DINKAS: THEIR LOVE FOR OLD CUSTOMS. — DESERTION
OF TWO DINKA SOLDIERS, AND REFUSAL OF THEIR COUNTRYMEN TO
GIVE THEM UP. — SHEIKS SEIZED, AND DESERTERS RECOVERED.

28th March.—We navigated to the neighbouring island, and

passed by it on the left, N.E. by N., just as the sun was rising. Where
is the meat that was to be put to-day into the Egyptian flesh-pots?
The Shilluks will be delighted at having tricked the Turks, for they are
really an extremely vitiated people. According to nature, corruption of
morals cannot take place in the very earliest stage of life, and there
might have been, therefore, a previous cultivation here which has
now disappeared, and left only the bad parts behind. But how could
they have gone back to the naked state? Slight clothing is even
necessary here: a people that has once worn a fig-leaf never throws
it off, although a cynic here and there may choose to lead a free dog-
life.
We labour with oars, yet all is in vain against the north wind; the
river itself remains on the average N.W. as far as Khartùm. About
nine o’clock we halt at the right shore, and see herds and villages of
the Shilluks opposite to us; but we must make our appetites
disappear again—“Quid juvat adspectus, si non conceditur usus!”
Really it was very vexatious; and in addition thereto, the new moon
has appeared for the last two days, and yet my bold countryman
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.

Let us accompany you on the journey of exploring knowledge and

personal growth!

ebookfinal.com