Download ebooks file Nonclassical Linear Volterra Equations of the First Kind A. S. Apartsyn all chapters

Visit https://ebookultra.
com to download the full version and

explore more ebooks
Nonclassical Linear Volterra Equations of the First

Kind A. S. Apartsyn
_____ Click the link below to download _____

https://ebookultra.com/download/nonclassical-linear-
volterra-equations-of-the-first-kind-a-s-apartsyn/
Explore and download more ebooks at ebookultra.com

Here are some recommended products that might interest you.
You can download now and explore!
Differential Equations Linear Nonlinear Ordinary Partial

1st edition Edition King A. C.
https://ebookultra.com/download/differential-equations-linear-
nonlinear-ordinary-partial-1st-edition-edition-king-a-c/
ebookultra.com
Optimal Solution of Nonlinear Equations First Edition

Krzysztof A. Sikorski
https://ebookultra.com/download/optimal-solution-of-nonlinear-
equations-first-edition-krzysztof-a-sikorski/
ebookultra.com
A first course in the numerical analysis of differential

equations 2nd Edition Arieh Iserles
https://ebookultra.com/download/a-first-course-in-the-numerical-
analysis-of-differential-equations-2nd-edition-arieh-iserles/
ebookultra.com
A first course in integral equations 2ed. Edition Wazwaz
https://ebookultra.com/download/a-first-course-in-integral-
equations-2ed-edition-wazwaz/
ebookultra.com
Loving Out Loud The Power of a Kind Word 1st Edition Robyn
Spizman
https://ebookultra.com/download/loving-out-loud-the-power-of-a-kind-
word-1st-edition-robyn-spizman/
ebookultra.com
A First Course in Linear Algebra With Concurrent Examples

Hamilton
https://ebookultra.com/download/a-first-course-in-linear-algebra-with-
concurrent-examples-hamilton/
ebookultra.com
Linear Partial Differential Equations for Scientists and

Engineers Fourth Edition Tyn Myint-U
https://ebookultra.com/download/linear-partial-differential-equations-
for-scientists-and-engineers-fourth-edition-tyn-myint-u/
ebookultra.com
Decomposition Analysis Method in Linear and Nonlinear

Differential Equations 1st Edition Kansari Haldar (Author)
https://ebookultra.com/download/decomposition-analysis-method-in-
linear-and-nonlinear-differential-equations-1st-edition-kansari-
haldar-author/
ebookultra.com
Linear Integrated Circuits 2nd Edition S. Salivahanan
https://ebookultra.com/download/linear-integrated-circuits-2nd-
edition-s-salivahanan/
ebookultra.com
Nonclassical Linear Volterra Equations of the First Kind
A. S. Apartsyn Digital Instant Download
Author(s): A. S. Apartsyn, Anatoly S. Apartsyn
ISBN(s): 9789067643757, 9067643750
Edition: Reprint 2010
File Details: PDF, 33.57 MB
Year: 2003
Language: english
INVERSE AND ILL-POSED PROBLEMS SERIES
Nonclassical Linear Volterra Equations
of the First Kind
Also available in the Inverse and Ill-Posed Problems Series:
Poorly Visible Media in X-ray Tomography Identification Problems ofWave Phenomena -
D.S.Anikonov,V.G. Nazarov, and I.V. Prokhorov Theory and Numerics
Dynamical Inverse Problems of Distributed S.I. Kobonikh/n and A. Lorenzi
Systems Inverse Problems of Electromagnetic
V.l. Maksimov Geophysical Fields
Theory of Linear Ill-Posed Problems and its P.S. Martyshko
Applications Composite Type Equations and Inverse Problems
VK. lvanov,V.V.Vasin andV.RTanana A.I. Koznonov
Ill-Posed Internal Boundary Value Problems for Inverse Problems ofVibrational Spectroscopy
the Biharmonic Equation A.G.Yagola, I.V. Kochikov, G.M. Kuramshina andYuA
MA. Atakhodzhaev Pentin
Investigation Methods for Inverse Problems Elements of the Theory of Inverse Problems
V.G. Romanov AM Denisov
Operator Theory. Nonclassical Problems Volterra Equations and Inverse Problems
S.G. Pyotkoy A. L Bugngeim
Inverse Problems for Partial Differential Small Parameter Method in Multidimensional
Equations Inverse Problems
Yu.Yo. e/ov A.S. arashkov
Method of Spectral Mappings in the Inverse Regularization, Uniqueness and Existence of
Problem Theory Volterra Equations of the First Kind
V.Yurko A. Asanov
Theory of Linear Optimization Methods for Solution of Nonlinear Operator
/./. Eremin Equations
Integral Geometry and Inverse Problems for V.P.Tanana
Kinetic Equations Inverse and Ill-Posed Sources Problems
AKh.Amirov Yu.E.Anikonov, B.A. Bubnov and G.N. Erokhin
Computer Modelling in Tomography and Methods for Solving Operator Equations
Ill-Posed Problems ν.ΡΤαπαηα
MM. Lavrent'ev, SM. Zerkal and O.ETrofimov Nonclassical and Inverse Problems for
An Introduction to Identification Problems Pseudoparabolic Equations
via Functional Analysis A.Asanov and E.R.Atamanov
A. Lorenzi Formulas in Inverse and Ill-Posed Problems
Coefficient Inverse Problems for Parabolic Type Yu.EAmkonoy
Equations and Their Application Inverse Logarithmic Potential Problem
P.G. Danilaev V.G. Cherednichenko
Inverse Problems for Kinetic and Other Multidimensional Inverse and Ill-Posed
Evolution Equations Problems for Differential Equations
Yu.E Anikonov YuEAn/konov
Inverse Problems of Wave Processes Ill-Posed Problems with A Priori Information
A.S. Blagoveshchenskii V.V.Vasin andA.LAgeev
Uniqueness Problems for Degenerating Integral Geometry ofTensor Fields
Equations and Nonclassical Problems V.A. Sharafutdinov
S.P. Shishatskii, A. Asanov and ER Atamanov Inverse Problems for Maxwell's Equations
Uniqueness Questions in Reconstruction of V.G. Romanov and S.I. Kabanikhin
Multidimensional Tomography-Type
Projection Data
V.P. Golubyatnikov
Monte Carlo Method for Solving Inverse
Problems of Radiation Transfer
V.S.Antyufeev
Introduction to the Theory of Inverse Problems
A. L Bukhgeim
INVERSE AND ILL-POSED PROBLEMS SERIES
Nonclassical Linear
Vo I terra Equations
of the First Kind
A.S. Apartsyn
III MB?Ill
UTRECHT · BOSTON
2003
VSP Tel: +31306925790
P.O. Box 346 Fax: +3130 693 2081
3700 AH Zeist [email protected]
The Netherlands www.vsppub.com
© VSP 2003
First published in 2003
ISBN 90-6764-375-0
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system,
or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or
otherwise, without the prior permission of the copyright owner.
Printed in The Netherlands by Ridderprint bv, Ridderkerk.

Contents
Introduction 1
Chapter 1. Classical Volterra equations of the first kind 5

1.1. Classification of integral Volterra equations of the fist kind . . . . 5
1.2. The Gronwall-Bellman lemma 10
1.3. A difference analog of the Gronwall-Bellman lemma 14
1.4. Self-regularization 17
1.5. Two-parametric (a, fo)-regularization 19
1.6. Inequalities with isotone operators 32
1.7. Inequalities with interchangeable isotone operators 39
1.8. Unimprovable estimates of solutions of multidimensional integral
inequalities 43
1.9. The well-posedness of a two-dimensional Volterra equation of the
first kind 49
1.10. Unimprovable estimates of solutions of two-dimensional differ-
ence inequalities 51
Chapter 2. Volterra equations of the first kind with two

variable integration limits. The case a(<o) < to 57
2.1. Problem statement 57
2.2. The method of steps 58
2.3. Illustrative examples 62
2.4. The existence and uniqueness theorem 65
2.5. An estimate of the solution stability 68
2.6. The study of a special problem of mathematical programming . . 70
viii A. S. Apartsyn. Nonclassical Linear Volterra Equations ...
2.7. A numerical solution of the test example 79

2.8. A geometrical illustration of the reduction by unity in the order
of convergence 81
2.9. A theorem on the convergence of the quadrature method (the
general case) 84
2.10. Some numerical results 90
2.11. On self-regularization 96
Chapter 3. Volterra equations of the first kind with two

variable limits of integration. The case α(ίο) = to 99
3.1. Problem statement 99
3.2. Solution of the simplest test equation 100
3.3. Existence and uniqueness theorem (the general case) 102
3.4. Estimation of the solution stability 104
3.5. Some generalizations of the Gronwall-Bellman inequality 107
3.6. Numerical solution of a test example 114
3.7. The proof of convergence for the quadrature method (the general
case) 118
3.8. Some numerical results 124
3.9. Self-regularization (the case of a disturbance in the right-hand side) 126
3.10. Stability of a numerical solution with respect to disturbances of
a(<) 128
3.11. Multidimensional Volterra equations of the first kind related
to the modelling of nonlinear dynamic systems using the
Volterra series 130
Bibliography 155
Index 167
Introduction
Year 1996 marked the hundredth anniversary of the publications by Vito

Volterra that laid the foundations for the theory of integral equations later
named after him. Particularly, in Volterra (1896), he showed that a solution
φ(ί) of the integral equation of the first kind
{ ) de = /(*), te[io,n (o.i)

•/ίο
subject to the corresponding smoothness of the input data, that is, the
kernel K(t,s) and the right-hand side f(t] and provided that K(t,t] ^
0 Vi € [iojT1]) can be represented in the form of an explicit inversion
formula containing a resolvent kernel.
Thousands of publications in the last century were devoted to the theory
and numerical methods for solving equation (0.1) and its generalizations.
An overview of the state of the art in this field and a survey of the related
articles is given by Brunner and van der Houwen (1986). We also note a
survey given by Tsaluk (1977) that contains a more complete list of works
by Russian mathematicians.
A new impulse to the study of Volterra equations of the first kind was
given by the development of the general theory of ill-posed problems. A great
contribution to this theory was made by such outstanding mathematicians
as A. N. Tikhonov, M. M. Lavrent'ev, V. K. Ivanov, and their pupils and
followers. The Fredholm equation of the first kind is a typical representative
of ill-posed problems in the classical Hadamard sense. Since equation (0.1)
is a particular case of the Fredholm equation of the first kind, the concepts of
regularization methods of ill-posed problems taking into account the specific
character of the Volterra operators are also applicable to (0.1). Among the
works devoted to this aspect, we point out Sergeev (1971), Apartsyn and
Bakushinsky (1972), Denisov (1975), and Magnitsky (1975). In the sequel,
2 A. S. Apartsyn. Nonclassical Linear Volterra Equations ...
equation (0.1) will be called classical (or standard) in view of numerous

publications devoted to it.
In 1897, Volterra considered the integral equation of the first kind (see
Volterra, 1897)
/'
ft
Ja(
te[t0,T\, (0.2)
with to = 0? o-(t) = qt, Q < q < 1, and showed that (0.2) is equivalent to
some integro-functional equation of the second kind whose solution, in turn,
could be found using the results obtained in (Volterra, 1896).
The paper Brunner (1997) devoted to the hundredth anniversary of the
introduction of Volterra equations of the first kind contains a list of works
published in early 20th century that are related to (0.2).
In 1977, V.M. Glushkov (see Glushkov, 1977) proposed a two-sector
macroeconomic model described by integral relations with Volterra oper-
ators of the type (0.2). The function a(t) representing the dynamics of
extinction or substitution of the elements in a (developing) system, together
with <£>(£), was assumed to be sought as a control element optimizing some
quality criterion of the system operation. An important a priori property
of α(ί) in such models is its nondecreasing due to the irreversibility of the
process of extinction in the developing system.
In a series of publications by V. M. Glushkov and his colleagues, the ideas
of Glushkov (1977) were further developed in terms of both mathematical
methods and applications (in biology, ecology, medicine, engineering, etc.).
This subject is considered in greater detail in Glushkov, Ivanov, and Yanenko
(1983), and Yatsenko (1991).
Since the problem of identification of the functional parameters appear-
ing in the models of the Glushkov type is reduced to equations of the type
(0.2), these equations drew even greater attention of researchers. For exam-
ple, some numerical methods for solving (0.2) were considered by Yatsenko
(1991) (see also Naubetova and Yatsenko (1988)).
It should be noted that the generalization of results obtained for (0.1)
to the case of (0.2) turned out to be rather nontrivial. Moreover, the math-
ematical apparatus for investigating (0.2) depends to a great extent on the
following assumption (the same s in models of the Glushkov type):
a(t) <t W € [ f 0 , T ] , (0.3)
or, for the initial point t — to,
a(t0) = to. (0.4)
Introduction 3
To gain a better understanding of the subject, it is recommended to

consider three elementary test equations assuming that K(t, s) = 1; α(ί) =
ί - 1 (the case (0.3)), α(ί) = ί/2 (the case (0.4), i0 = 0). Suppose that the
corresponding solvability conditions for (0.1), (0.2) in C^0)T] are satisfied.
Then their solutions can be written as follows. The solution of (0.1) is
/'(*), ί€[ί 0 ,Τ]; (0.5)
the solution of (0.2) with the condition (0.3) is

fc-l
t-i}+(Po(t-k), te[t0 + k-l,to + k), (0.6)
i=0
where <po(t) is a given function on the prehistory ί 6 [ίο — 1, ίο], k =
l, ... , [Τ - to] + 1; the solution of (0.2) with the condition (0.4), t0 = 0,
is oo 1
ie[0 T] (0 7)
*(*) = Σ */V'
t=0
' -
Formulas (0.5)-(0.7) show that (0.2) is much more complicated than
(0.1) and, in addition, demonstrate how the structure of the solution of (0.2)
depends on the properties of α(ΐ). Thus, for satisfying (0.3), it is necessary
to additionally specify the solution on the prehistory, and the series (0.6)
must be finite. For (0.4), all the information on the solution is contained in
the equation (0.2) itself, but in general the series representing the solution
is infinite.
In the sequel, equations of the type (0.2) will be called nonclassical.
As far as the author knows, the publications devoted to (0.2) are rather
few in number. Some of them are mentioned in Brunner (1997), while
others, related mainly to the problem of numerical solution of (0.2), are
not widely available. In general, equations of the type (0.2) are less studied
than equations of the type (0.1).
The purpose of the present monograph is to bridge this gap.
The method for investigating equations (0.2) that is used in this book
is based on generalizations of the Gronwall-Bellmann (G.-B.) inequalities
and their discrete analogs.
As far as some generalizations of the G.-B. inequality are also of interest
in the case of constant lower limits of integration, these generalizations re-
lated to the solution of classical multidimensional Volterra equations of the
first kind are considered in Chapter 1 of the monograph.
4 A. S. Apartsyn. Nonclassical Linear Vex/terra Equations ...
Equations (0.2) with the condition (0.3) are analyzed in Chapter 2 and
those with the condition (0.4) are considered in Chapter 3.
We have restricted ourselves to the linear Volterra equations of the first
kind with variable lower limit in order to focus our attention on specific
problems arising in transition from (0.1) to (0.2), though the extension of
some results to the equation of the type
t
K(t, 5, φ(β}} as = /(i), t e [ίο,Τ] (0.2')
o(t)
presents no significant problem. Systems of equations of the type (0.2) for

which the corresponding generalizations are not as trivial as in the classical
case for detK(t, t) ^ 0 Vi 6 [to,T] are not considered either.
The final section of the book describes a class of linear multidimensional
Volterra equations of the first kind that arise in the identification problems
of nonlinear dynamic systems. "Nonclassical" nature of these equations is
also due to the variable limits of integration.
Conceptually these equations are close to the problems of computer to-
mography as the existence and uniqueness of their solutions in the required
class of functions are provided by a special choice of integration domains. As
detailed description of such equations would considerably increase the vol-
ume of the book, we consider only the main concepts and refer the reader
to the corresponding publications for details.
The author hopes that the monograph will attract the attention of spe-
cialists in computer mathematics, inverse and ill-posed problems, mathe-
matical modelling, and will be of interest for young researchers. For the
latter, the subject studied here can serve as a large testing ground for their
potential in the prospective field of applied mathematics.
The author greatly appreciates the assistance of Prof. N. A. Sidorov
for useful discussions of individual results and the monograph as a whole.
The author wishes to thank E.V. Markova for her numerical results that
contributed greatly to the contents of this book. The author also highly
appreciates the skillful assistance of I.D. Guseva.
Financial support from the Russian Foundation for Basic Research (grant
No. 99-01-00116) is gratefully acknowledged.
Chapter 1.
Classical Volterra equations of
the first kind
1.1. CLASSIFICATION OF INTEGRAL VOLTERRA

EQUATIONS OF THE FIST KIND
Consider an integral Fredholm equation of the first kind
rT
= K(t,s)tp(s)ds = f ( t ) , te[i 0 ,T]. (1.1.1)
•/ίο
>*s.
It is well known that if the kernel K(t, s] is continuously differentiable

with respect to the first argument and continuous in the second argument on
the square D = {t, s/to < s, t < T}, then, even if a solution φ(ί] of (1.1.1)
exists in the space £*[ί0)Γ] of continuous functions on [to,T] (further denoted
by C for brevity), it is a priori unstable with respect to disturbances in the
right-hand side of (1.1.1) measured in the metric of C or LI (Tikhonov and
Arsenin, 1986). Thus, (1.1-1) is a typical ill-posed problem in the classical
sense of Hadamard.
It is necessary to mention the decisive role of the kernel smoothness in
providing the ill-posedness of (1.1.1). Indeed, consider two extreme cases
of specifying K(t, s). Let K(t,s) Ξ const on D . Then equation (1.1.1)
has infinitely many solutions in C for f ( t ) = const, whereas it is α priori
unsolvable for f ( t ) ^ const. The other example is the case K(t, s) = S(t — s ) ,
where δ(ν] is Dirac's delta function. Let (Φ,Ρ) denote the pair consisting of
the domain of solutions and the domain of right-hand sides of the operator
equation Αφ = f . Since for any continuous function φ(ί] on [to,T] we have
Γ
Jto
(hence, the Fredholm operator with such a kernel plays the role of the iden-
tity E in (7), the Fredholm equation of the first kind
1
t <—t Γ·*[LQ, Τ !
J. J, (Λ 1 1\
(L.l.o)
considered on the pair ((7, C) is well-posed as A""1 = E and φ(ί) = f ( t ) .

Equality (1.1.2) is a consequence of the following properties of the "func-
tion" S(t>):
\ oo, ι/ = 0
and oo
<S(i/)di/ = l. (1.1.5)
/ •oo
By virtue of (1.1.4), the kernel of equation (1.1.3) has a discontinuity of

the second kind at every point of the diagonal s = t of the square Q It is
the presence of singularities of K(t, s} for s = t that provides the possibility
of correction of the original problem, the "degree" of ill-posedness and the
type of kernel singularity being closely related.
From this standpoint, we consider the case where
-K { , ,
^ 0, s>t
and
min \K(t,t)\=k^Q. (1.1.7)
t€[to,T\
If (1.1.6) is satisfied, then equation (1.1.1) turns into a classical Volterra
equation of the first kind
Υφ = f K(t,s)<p(8)ds = f ( t ) , t 6 [ίο, (1-1-8)

JtQ
^**
and condition (1.1.7) means that the kernel K(t, s) has a discontinuity of
the first kind at every point of the diagonal s = t of the square D .
It is easy to see that if
K't(t, s) Ε ΟΔ, Δ = {t, s/t0 < s < t < T}, (1.1.9)
and
/'(*) 6 qto,T], (i.i.io)
/(*o)=0, (1.1.11)
then the inverse of the operator V considered as an operator that maps Ο^,Τ]
°(1) Δ fl)
into C[t0}T] = [f(t)/f(t) e C'LV]' /(*o) — 0} is bounded and therefore the
problem of solving (1.1.8) is well-posed on the pair (C, C ).

Indeed, applying the operator
1
a4> (1-1-12)
to both sides of (1.1.8), we obtain the Volterra equation of the second kind
(E - ψ)φ = ψ(ί] + Γ ^^-φ^) ds = D/, (1.1.13)

Λο ^\1^)
which is equivalent to (1.1.8) because the operator T> is linear and condition
(1.1.11) holds. Therefore, from (1.1.9) it follows that the (unique) solution
φ of (1.1.13) is continuous, and (1.1.10) is a solution of (1.1.8):
φ= (E-WrlVf. (1.1.14)
Hence,
V~l = (E - W}~l-D. (1.1.15)
We need only show that
II V-1 || o ( i ) <oo. (1.1.16)

C
(t0,T]-*C[t0,T]
We have
<\\(E-Wrl\L\\V\\0(l} . (1.1.17)
, < C
As follows from (1.1.7),

II. (D <!/*, (1.1.18)
C
and
00
'11
II (E - W}' \\c_>c < IIw* \\c-+c< °°> i1·1·19)
i=0
the convergence in norm of the Neumann series being ensured by the prop-
erty of generalized contraction in C of the operator
(1.1.20)
t0
Finally, (1.1.16) follows from (1.1.17) (1.1.19).
Thus, if (1.1.7) and (1.1.9)-(1.1.11) hold, then the integral Volterra equa-
o(l)
tion of the first kind (1.1.8) is well-posed on (C, C ), so that
ν-ι=ΒΤ>, (1.1.21)
where Β = (Ε — W}~1 is bounded in C and D is a (linear) first-order

differential operator. We emphasize that, if the discontinuity of the second
kind of the kernel K(t, s) = S(t — s) for s = t provides the well-posedness
of (1.1.1) on the pair (C,C), then, in the case of finite discontinuity of the
kernel (1.1.6), (1.1.7) on the diagonal of D, the well-posedness of (1.1.1) is
achieved by restricting the space of images and endowing it with a stronger
topology than that of C, i.e., with the topology of C^.
The representation (1.1.21) of the operator V~l as the composition of
some bounded operator acting from C to C and a differential operator is
characteristic of integral Volterra equations of the first kind and provides a
basis for the following definition.
Definition 1.1.1. Integral equation (1.1.1) belongs to the type

ξ Ε [Ο,οο], if the representation (1.1.21) is valid, where Β is some bounded
one-to-one operator from C into C and D is a (linear) differential operator
of order ξ.
Definition 1.1.2. If integral equation (1.1.1) belongs to the type

then the number ξ is referred to as the degree of instability of solution
(1.1.1). If two integral equations of the first kind belong to the types V(fi)
and ^(£2)5 respectively, then we say that the solution instability of the
first problem is stronger (weaker) than that of the second one if ξι > £2
(ξι < £2)· With ξ\ = £2 > the problems are equivalent in terms of their
solution instability.
According to Definition 1.1.1, equation (1.1.2) belongs to the type V(0)

and the classical Volterra equation of the first kind (1.1.8) with conditions
(1.1.7), (1.1.9)-(1.1.11) belongs to the type V(l).
Suppose that conditions (1.1.6) and (1.1-7) hold for the derivative of the
kernel with respect to t of some order p > 0 rather than for the kernel
K(t,s) itself:
.
0, s > ί;
min \K$(t,s}\s=t\=kp^Q. (1.1.7')
ο(ρ+1) Λ
Then (1.1.1) is well-posed on the pair (C,C ), C[t^T] = {/(*)//(*) £
C
[t0+T}' / (l/) (*o) = 0, i/ = O^p}, so that (1.1.1) belongs to the type V(p + 1).
An important example of the kernel A"(i, s) satisfying (1.1.6'), (1.1.7') is
(t ~ s)P/p\, tQ<s<t<T-
'"' - - - '
0, 5 > t.
In this case, the derivative of order (p+ 1) in the right-hand side of (1.1.1) is
a solution of (1.1.1) (or the equivalent equation (1.1.8) with the right-hand
side /ω (ί)).
In particular, if ρ = 0, then (1.1.6'), (1.1.7') coincide with (1.1.6),
(1.1.7), and the kernel (1.1.22)
K(t,s) =
0, s >t
corresponds to the problem of single differentiation.

The Abel equation is an example of equation (1.1.1) of type
to which the simplest problem of computer tomography is reduced when

the reconstructed characteristic has the property of axial symmetry in the
selected section (Preobrazhensky and Pikalov, 1982). In this case,
where Γ(·) is the gamma function, άι~α/ dtl~a is the operator of partial
differentiation of order 1 - a (Samko et al. (1987)). Equation (1.1.23) is
ο (λ) ο (λ)
well posed on the pair (C, Lip ), X = 1 — α + β, 0 < /9 < α, where Ιιίρ[ίθ)Γι
is the Banach space of functions satisfying the Holder condition with the
index λ and norm (see Kantorovich and Akilov, 1977)
/(ί) II ο (Λ) =|| /(ί) ||qto,T] 4- sup " ν , ~Λ, ; Αλ """, /(ίο) = 0.
LiP[t0,T] [ί,ί+Δί]<ΐ[ίο,Τ] ' *^
Consider the equation

r
^ /ΤΛ/Λ/Ο^ /· <r? ·*
t > 0. (1.1.24)
Jo
as an example of the Volterra equation of the fist kind of the type V(oo).
An inverse problem of heat conduction that consists in determining the
temperature ip(t) on the boundary χ = 0 of the half-space χ > 0 from the
temperature measurements /(ζο,ί) at a distance χ > 0 from the boundary
(a is the heat conductivity coefficient; initial temperature distribution in the
half-space is zero) is reduced to (1.1.24).
Since the kernel of equation (1.1.24)
is such that
K<f>(t,8)\e=t =0
for ρ as large as desired, which contradicts (1.1.7'), it follows that (1.1.24)

α priori does not belong to the type V(£) with ξ < oo. Unlike the problems
discussed above, such equations are naturally called strongly ill-posed. A
more detailed description of the class Υ(ξ), ξ < oo of (weakly ill-posed)
equations of the first kind can be found in Apartsyn (1981, 1987, 1988).
1.2. THE GRONWALL-BELLMAN LEMMA

Inequalities play a special role in the history of mathematics. As noted
in the preface to the classical monograph of Bellman and Beckenbach (1965),
". . . the main results of mathematics are most often expressed by inequalities
rather than by equalities" . It is difficult to obtain important results in math-
ematics without using the fundamental inequalities like the triangle inequal-
ity, the inequalities of Hadamard, Bessel, Holder, Cauchy-Bunyakowsky,
etc. The Gronwall-Bellman (G.-B.) inequality stands out among them and
is known as a fundamental result of the stability theory (Bellman and Beck-
enbach, 1965). G.-B. inequality and its modifications and generalizations
are used for studying the issues of existence, uniqueness and stability of
solutions of different classes of ordinary differential equations, differential
equations in partial derivatives as well as integral and integro-differential
equations.
A widely used simple version of G.-B. inequality and a proof obtained
by Bellman (1943) will be given below.
Lemma of Gronwall—Bellman. If functions g(t) andu(t) are nonneg-

ative for t > ίο and c is a nonnegative constant, then from the inequality
ft
u(t)<c + g(s)u(s)ds, > i0 (1-2.1)
ΛΟ
it follows that
u(t) < c exp ( i g(s) ds\ i > i0. (1.2.2)
Proof. Prom inequality (1.2.1) we have
c +. f"Jtog(s)u(s)ds
Λ , , , ^ »(*)· (L2·3)
Integrating (1.2.3) from ίο to i, we obtain
l n (v c + / g(s}u(s)ds] - l n c < / g ( s ] d s . (1.2.4)

Jt0 ' Jto
Hence,
/ g(s)u(s) ds < c exp ( / p(5)d5), (1.2.5)
V
îo Jto '
and inequality (1.2.2) follows from (1.2.1) and (1.2.5). D
In many applications, a simplified version of (1.2.1) with g(i) = m > 0
is used. If c > 0 and m > 0, then the inequality
u(t) <c + m i u(s}ds, t>tQ (1.2.1')
implies that
u(t) <cem(t-to\ t>tQ. (1.2.2')
Let us describe the application of the Gronwall-Bellman Lemma for

estimating the stability of solution of the classical integral Volterra equation
of the first kind (1.1.8). It is required to estimate || V~l \\ 0 (i) under the
c -+c
assumptions (1.1.7), (1.1.9)-(1.1.11).
Prom equation (1.1.13) equivalent to (1.1.8) we can obtain the estimate
ί6[ί τ] (L2 6)
°' · '
We put
(1.2.7)
(1.2.8)
Making (1.2.6) stronger, by virtue of (1.1.7) we have
Then, applying the Gronwall-Bellman Lemma to (1.2.1'), (1.2.2') with c =

j m = Κι/k, from (1.2.9) we readily obtain
Klk ' ί-ί 1
lΙί/^ί/Ή
^£/1 (> l l <
^ f/Λ
l i ^ / /k"\p
rblC Ι ο; · /V ρ
^ [>„
I uy * τ
J. ]Ι . Πl J. .^
2 .101
XL/I
Hence,
< (Fi/fc)e Klfc ^-^J. (1.2.11)
Furthermore,
|| o(1 ) , (1.2.12)
and (1.2.12) implies
< (I/k)eKlk~1F-i°l (1.2.13)
Note that the estimate (1.1.17) was obtained in Section 1.1. We now show
that (1.2.13) follows from (1.1.7).
By virtue of (1.1.18) and (1.1.19), it suffices to estimate
00
i wl
where the operator W is determined by the formula (1.1.20).
Set
Using (1.1.7) and (1.2.7), we successively obtain
<£(*-*<>) II <«>(«) l
!«(«)! < II «W !<%..„;
MIT
\ rv /
^-Γ^
ϊ·
II <*>(*) Hq to ,T]'
which implies
/ T,\ ΓΤ ^ ( (i ^ is)
Since ψΐ = Wl(po, from (1.2.15) we obtain the estimate
(1 2 16)
(f i^· ··
By virtue of (1.1.19) and (1.2.16),
00
v \i Ιτ< 4 \i
^ °'
c[to,r]->qto,T] v /
i=0
Finally, substitution of (1.1.18) and (1.2.17) into (1.1.17) gives the result
equivalent to (1.2.13).
It is evident that, on one hand, the use of the G.-B. inequality is the
shortest way to obtain the estimate of solution stability for equation (1.1.8);
on the other hand, the method based on representing the operator V~l in
the form (1.1.5) is more universal. In what follows, we shall use it to obtain
(in a certain sense) unimprovable estimates of solutions of Volterra equations
of the first kind that are more general than (1.1.18).
1.3. A DIFFERENCE ANALOG

OF THE GRONWALL-BELLMAN LEMMA
Above we have demonstrated the role of the Gronwall-Bellman Lemma for
estimating the stability of a solution of the classical Volterra equation of
the first kind (1.1.8). Difference (discrete) analogs of the Gronwall-Bellman
Lemma play a similar role in studying numerical methods for solving (1.1.8).
We now formulate the simplest of these statements, which will be used below.
Lemma 1.3.1. If
i-l
Ui < c + m^Piij, z' = 2, 3, . . . , MI < c, (1.3.1)
3=1
where c and m are nonnegative constants, then
U i ^ c i l + m)'-1, « = 1,2,... . (1.3.2)
Proof. The proof is by induction. For i = 1, the assertion is true. Let

(1.3.2) be valid for i = n. We need to show that (1.3.2) is valid for i = n + 1.
We have n n
3=1 3=1
n-1 Ti-2
^(l+m) - ) = c(l + m) (l + m ^ ( l + m)J
J 1 2
3=1 3=1
··· = c(l + m) n ~ 1 (l +m) =c(l + m) n . G
A more general case is proved similarly: if (1.3.1) is satisfied, but u\ < ci,
then
1 + τη)^2, < = 2,3, . . . . (1.3.2')
In inequalities of the type (1.3.1) which appear in the study of numerical
methods, c, ra and i are usually interrelated. To be more precise,
C = CI/I Q , m = rai/i, ih < d,
where h is the mesh size and ci, α,, τηχ, d are constants independent of h.
By virtue of (1.3.2),
«i < ci/i a (l + mi/i)^'1 < ci/i a e mid . (1.3.3)

If Ui is the absolute value of the error of the numerical solution at the

ith point, then (1.3.3) implies the convergence of the numerical method, α
being the order of convergence.
Taking this fact into account, we now investigate the convergence of the
simplest quadrature method for numerical solution of (1.1.8).
Introducing a mesh of points ti = to + z/ι, i = 1, n, nh = Τ — to in [tQ, Τ]
and approximating the integral
/"*·'
/ K(tits)<p(s)ds
Jto
by the quadrature formula of the right-hand rectangles, we write a mesh

analog of (1.1.8) which is a triangular system of linear algebraic equations
(SLAB):
, i = l^. (1.3.4)
3=1
By virtue of the condition (1.1.7), the system (1.3.4) has a unique solu-
tion for any right-hand side, which will be denoted by <ph = (φ^, ...,</?£).
According to Gavurin (1971), <f>h will be called the carcass of the approxi-
mate solution of integral equation (1.1.8), and the vector (ψ(ί\), . . . ,v?(f n ))
will be called the carcass of the exact solution φ(ί] of (1.1.8). Let Ch and
l
be the spaces of mesh functions with the norms
ν
·**»·»" . 1 J II ϊ
l<i<n h
respectively. Then (1.3.4) can be represented in the operator form
' = /*, (1-3.5)
where VhtC(Ch,Ch ).
We introduce the vector of errors
- V?}, i -
It is evident that
Vhe = -τ(φ], (1.3.6)
where
τ}(φ)= (1.3.7)
t0
is an error of the quadrature formula of right-hand rectangles on the segment

[*o,*i].
Subtracting the (i — l)th row from the iih row in (1.3.5) and passing to
the estimate by absolute value in the obtained equality, we have
sAi < ηι, V Ι·^*'^') ~ Kfa-l'tj^ (1.3.8)

t-i \ —
— /' ,^ , | r ^/, , \i
i = 2,n.
First, we need to estimate \τ^(φ) — Γ^_ί . By virtue of (1.3.7),
t0
t-l
+ (1.3.9)
Applying the estimate for the standard error of the quadrature formula of
right-hand rectangles (see Mysovskikh, 1962), from (1.3.9) we obtain
h2
Η max
2 t,
(1.3.10)
(here we assume that the corresponding derivatives are continuous).
Returning to (1.3.8) and using the previous notation, we have
i-l
(1.3.11)
where
max
(1.3.12)
In addition, we readily verify that |ε^| < Rh. Now, according to

Lemma 1.3.1, we can use inequality (1.3.3) with ci = R, α = 1, τη\ = k~lKi,
d = T — to, such that
~ (1.3.13)
which proves the convergence of order h of such a quadrature method.
If the kernel of the sought solution is smoother, i.e.,
then the second-order convergence of the quadrature method involving the

formula of middle rectangles is proved in a similar way.
Remark 1.3.1. It is wrong to assume that the use of a quadrature for-

mula of higher order of accuracy necessarily improves the rate of convergence
of the quadrature method. For example, it was shown by Linz (1969) (see
also Apartsyn, 1973), Simpson, and Gregori that formulas of order greater
than unity produce unstable finite-difference schemes.
Since late sixties, a great number of works have been devoted to the
analysis and construction of convergent numerical methods for solving equa-
tion (1.1.8). A review of these works can be found in (Brunner and
Houwen, 1986).
Remark 1.3.2. The publications of Apartsyn (1981), (1987),(1992e)

deal with a general scheme of constructing numerical methods for solving
(1.1.8) on the basis of a discrete analog of (1.1.15), i.e., the representations
Vh~l = (Eh - WhrlVh, h<h0,
Wh € C(Ch,Ch], Vh 6 C(Ch ,Ch), ({Dh},{Wh}}, and the notion of a

coordinated pair of families ({Έ^}, {W/i}) that was formulated in terms of
the theory of discrete convergence of operators (Vainikko, 1976, 1979).
1.4. SELF-REGULARIZATION
So far it has been assumed that the input data in (1.1.8) are known exactly.
However, in practice, they always have noises. For example, suppose that
the right-hand side f(t] is replaced by a function f ( t ) such that
\\f(t)-f(t}\\C[to,T]<o, (1Α1)
where δ is a given number (sometimes it characterizes the accuracy of mea-

suring instruments).
Though equation (1.1.8) itself with the right-hand side /(i) is not
solvable because of possible violation of conditions of the type (1.1.10) and
(or) (1.1.11), by virtue of (1.L7), with J(t) substituted for f ( t ) ,
(1-4-2)
the system (1.3.4) still has a unique solution. We denote it by (ph = {φ%},
i = l,n, and introduce the error vector εΛ = {έ^} = {<p(ti) — $}, i = l,n.
We now estimate its norm in Ch> Since eh satisfies the system (SLAB)
Vheh = -f h (Y>), (1.4.3)
where
it follows that the calculations performed in Section 1.3 to estimate ε^ yield
|| e \\Ch= max | ε | < c^ + c2 Ξ </>(<*, Λ). (1-4.4)

1<ι<η η,
Here
ci = Rek~lK^T~^\ c2 = 2Ar 1 eT 1Jfl < r - to >.
Assume that the mesh size h is chosen to be a function of δ determined by
the formula
h(8) = c357; c3 = const > 0, 0 < 7 < 1. (1.4.5)
Then from (1.4.4) if follows that ψ(δ,ίι(δ)) -> 0 with δ -» 0 and, moreover,
II £h llcfc— >· 0 as 5 -> 0. It means that the problem of reconstructing
the carcass of the approximate solution of (1.1.8) considered on the pair
(Cfi,Ch] is stable with respect to disturbances in the right-hand side of
(1.1.8) that are measured in the metric of C^^.
It is natural to choose the parameter 7 in (1.4.5) so that the rate of
convergence of the function ψ(δ, h(S)) to zero is maximized. This value
of 7 will be referred to as quasi-optimal (q.o.). It is easy to see that
7q.0. = 1/2. In this case hq.0(S) = ο3δ^·°· χ δτ and || eh«·0· ||cfcq.0 .= O(S*).
A majorant estimate of the type (1.4.4) for the quadrature of middle
rectangles has the form
\\eh\\Ch<c1h2+c2S-=^l(6,h) (1.4.4')
and therefore
A stabilizing effect of discretization as applied to the problem of numer-

ical differentiation is well known (for example, see Bakhvalov, 1973; Demi-
dovich, 1967; Kalitkin, 1978; Apartsyn and Ten Men Van, 1972). For inte-
gral equations of the first kind, this effect was first pointed out by Tikhonov
and Dmitriev (1968), Dmitriev and Zakharov (1968) and was called a "self-
regularization effect" . This term emphasizes the absence of any additional,
artificially introduced regularization parameter in such regularization meth-
ods (a type of the parameter a in A. N. Tikhonov's method).
Remark 1.4.1. If the error in the right-hand side of (1.1.8) is measured

in the metric of Z/2, then, as shown in Apartsyn and Bakushinsky (1972),
the self-regularization of the quadrature method (1.4.2) is guaranteed by the
estimate || ehq-°· \\ch 0 = Ο ( δ & ) under the assumption that hq.0.(S) χ δ*.
Remark 1.4.2. Of importance is the case of kernel disturbance in

equation (1.1.8) such that the deviation of the given function A"(i, 5) from
the original function K(t,s), for example, in the metric of CA, is equal to
a given value δ:
As follows from Apartsyn (1982), if^the conditions (1.3.14) are satisfied

after replacing K, fci, ho by K, k\, HQ , then the above asymptotic
estimates hold.
Remark 1.4.3. In the publications of Vainikko and Hyamarik (1985,

1988), the self-regularization property was studied for approximation meth-
ods of projection type that are based on making the original integral equa-
tion finite-dimensional (see also Natterer, 1977; Richter, 1978; Grebennikov,
1984, 1986; Voronin and Tsetsokho, 1981).
1.5. TWO-PARAMETRIC (a, #)-REGULARIZATION

In Denisov (1975), a regularization scheme known as the Lavrent'ev method
(see Lavrent'ev, 1959, 1962) applied to operator equations in Hubert spaces
was studied in the case of disturbances of the type (1.4.1). This procedure
is based on transition from (1.1.8) to the equation of the second kind
αφ(ί) + f K(t,sMs)ds = f ( t ) , t e [ίο, Τ], (1.5.1)

Jto
or, in the operator form,

. (1.5.1')
In this case, the condition
sign α = sign K(t, t)
is assumed to be satisfied (from (1.1.7) and the continuity of the kernel in

the domain Δ, it follows that K(t,t) is of constant sign). Without loss of
generality, we assume that α > 0.
The main advantage of (1.5.1), as compared, for example, to the method
of A.N. Tikhonov, is the preservation of the Volterra type for equation
(1.5.1), whereas the Euler equation for the functional of A.N. Tikhonov
includes the operator V*V, which is not a Volterra operator.
Let
= 0, (1.5.2)
where (<f>(t), t € [to, T], is still an exact solution of (1.1.8). Then, according
to Denisov (1975), the estimate
||£(f) -<p a (i)||q to , T] < dla + d2-, d^dz = const (1.5.3)
is valid, where fia(t) is a solution of (1.5.1) (the case 6 = 0 was considered

earlier in Blondel, 1971). Condition (1.5.2) is essential for obtaining (1.5.3)
as φα(ίο) = 0 Va > 0, 6 = 0 and therefore the uniform convergence of
ψα(ί) to φ(ί) with respect to t α priori does not hold if <f>(to) / 0. The
following simple example illustrates the emergence of the boundary layer in
the neighborhood of the point ί0· Consider the equation
/ φ(8) as = t - t0, te[t0,T], (1.5.4)

ΛΟ
whose solution is φ(ί] = 1. A regularized version of (1.5.4)
αφ(ί)+ f <p(s)ds = t-tQ, te[t0,T], (1.5.5)

•/ίο
t-t n
has a solution ψα(ί) = 1 — e <* such that for fixed t* = to + ε, ε > 0, we
have \\<p(t) - </?a(i)||qt.,r]a30 0, but \\ψ(ί) - ψα(ί)\\ο^η = 1.
To transfer the boundary value into the right-hand side of (1.5.1), it is
necessary to add the term αφ(ίο). Thus, unlike (1.5.5), the equation
ft
αφ(ί)+ I (p(s)ds = t-t0 + a, t e [i0,T], (1.5.6)
Jto
has a solution identical to the solution of (1.5.4). However, such a proce-

dure assumes that (p(to) is α priori known, which is rather problematic in
practice.
In the previous section we considered a self-regularization method that
had no such drawback. To improve its noise- proof property, we can introduce
a combined procedure: discretization together with α-regular izat ion.
For example, using the quadrature formula of middle rectangles as a
basis, consider a system of linear algebraic equations following Apartsyn
(1979):
i
(L5 7)
' -
=
i = l, n, nh = T — io> ti = t -\- ih, ij_i ίο + (i ~ 0)^·
We denote the solution of (1.5.7) by $Q>h = {φα'\}, 1
i = T~n and
2
introduce the vector ea'h = {ε°^λ} = {<p(t f _i) - φα'\_}, i = T~n. Our
1 2 l
2 2
α>Λ
objective is to estimate ||ε ||(7Λ. Since
i-l
(1.5.8)
where
rti
ri ~l
(φ)= i) (1-5.9)
Jto
is a residual term of the quadrature formula of middle rectangles, subtracting

(1.5.7) from (1.5.8) we have
i-l
•Σ 3=1
.ι), < = 1,η. (1.5.10)

After finding the difference between neighboring rows of the triangular
system (1.5.10), we turn to estimating absolute values. Taking into account
(1.2.7) and (1.3.14), we obtain
.7=1
_
1_ i = 2,n. (1.5.11)
If conditions (1.3.15) are also satisfied, then, using a standard esti-

mate of the residual term of the quadrature formula of middle rectangles
(Mysovskikh, 1962) it is easy to find that
\Γΐ(φ) — Γί-ι(φ)\ < c/ι3, c = const, i = 2,n.
Since
Φι = max |y'(t)|,
' x /(
from (1.5.11) it follows that
v,h \+ ?δ+ ch* +

a+
» = 2,n. (1.5.12)
In addition, for i = 1, from (1.5.10) it immediately follows that
|< r . v(1-5.13)
y
I '- α 4-
If tp(to) = 0, then
f '~ α
otherwise,
· *=
We denote the right-hand sides of (1.5.14) and (1.5.14') by Γ0 and ΓΙ,

respectively. In what follows, we will need the estimate |e|' |. Assuming
that i = 2 in (1.5.12), we get
if ψ(ίο) = 0; otherwise, we have
j-α,/ΐι ^ ρ , α + h KI . ,,
ε3 < Γ 0 + — ΓΙ. (1.5.15)
2 α + ηκι
For definiteness, let <^(<ο) = 0. Consider the difference equation
α + h?Ki α+ __ „ .
- , i = 2,n, μι = Γ 0 . (1.5.16)
α +, nk\
..
It is evident that
Ι3*Λ I < W , i = M- (1-5.17)
Finding the difference μϊ - μί_ι, we reduce (1.5.16) to the Cauchy prob-

lem for a homogeneous second-order difference equation:
- i-2 = 0, t = 3~ίΤ; (1.5.18)

α 4- /ι«ι / α+
(1.5.19)
It is well known (see Godunov and Ryabenky, 1977) that a solution of

(1.5.18), (1.5.19) has the form
& = li\\~l + £2\ϊ-1 , * = 17n, (1.5.20)
where AI and 2 are the roots of the characteristic equation
Λ2 , α+ ^ a
λ2 - 1 + -ττ^ λ + -ΓΓ- = °>
V ot + hki ) a + hki
and
^ _ μιΛ 2 -μι ^ _ μι - \\μ\
λ —Α λ —λ
24 A. S. Aparisyn. Nonclassical Linear VbJterra Equations ...
Simple computations show that

ι ±
λΐ 2 =
'
-2; (1.5.22)
1
τ
2 ί
Γρ / 2α
2
2
Formulas (1.5.21)-(1.5.24) completely determine the solution of the
problem (1.5.18), (1.5.19).
As A2 > 0 and £2 < 0 Va > 0, from (1.5.20) we have
1
, « = ϊ7η. (1.5.25)
It is evident that V(<*,/i) = 1 + O(«) V/ι > 0; therefore, by (1.5.21),
ΛΙ = 1 + O(h], and (1.5.23) implies that t\ < const TQ. Thus, taking into
account the inequality i < (Τ — ίο)/^> from (1.5.25) we obtain
P-i < ΓΟ · const, i = l,n.
Consequently, since
_ 26 + ch3 +
°~
we get
^ .
μ·ΐ < -η- > ί = l,ra, ci, C2, 03 = const .
α
By virtue of (1.5.17),
^. -—-- , Tι_= 1, η.
< n « ; o(1.5.26)
^
a + ΛΛΙ
Thus, we have obtained the desired error estimate of the α-regularized
carcass of the approximate solution of equation (1.1.8) for the case v?(io) = 0.
Similarly, using inequalities (1.5.14') and (1.5.15') instead of (1.5.14) and

(1.5.15), we establish the estimate
for the case φ(ίο) ^ 0.

The dependencies h(5) and α(δ) maximizing the rate of convergence of

the right-hand sides of (1.5.26), (1.5.27) to zero as δ —>· 0 can naturally be
called quasi-optimally concordant (q.o.c.). It is easily seen that if <p(to) = 0,
then /iq.o.c.(i) x £3, aq.0.c.(5) x <^, and ||£Q<'-0-<:"/l<1-0-c-|kq.0.c. = O(^); if
<f>(t0) φ 0, then Aq.o.c.(<5) x £3, aq.0.c.(<J) x 5, and ||eaq-0-c"hll-0-c-||cAq.0.c. =
0(δ\).
The results obtained are summed up by the following theorem.
Theorem 1.5.1. Let conditions (1.3.14), (1.3.15), and (1.4.1) be satis-

fied. Then, for α-regularized carcass of the approximate solution of equation
(1.1.8) that satisfies the system (1.5.7) with α —>· 0, the estimate
), z = M, (1.5.28)
2
is valid if
h(6) X I, α(ί) χ ί § (1.5.29)
in the case φ(ίο) = 0, and
h(S) χ ii, a(J) x<5 (1.5.30)
in the case φ(ίο) / 0.
Remark 1.5.1. In the case of a disturbance of the kernel (1.1.8) in the

metric of C& characterized by
\\K(t,8)-K(t,8}\\c±<8, (1.5.31)
as was shown in Apartsyn (1982), the asymptotic estimate (1.5.28), where
<£>a'h is a solution of the system
i=T
j_!<^_!=/z, ^' (1.5.32)
is also preserved under the condition that (1.5.29), (1.5.30) hold and there
exists ho > 0 such that
min \K(t,t-h)\ = ~k\ > 0.
h<t<T
0<h<h0
However, in general, for fixed δ, kernel disturbances have stronger effect on
the carcass errors than disturbances in the right-hand side because of larger
values of the corresponding constants.
Remark 1.5.2. Quasi-optimal asymptotic formulas (1.5.29), (1.5.30)

do not provide the way to choose α and h for fixed δ as they contain the
constants characterizing the exact solution (or, equivalently, the exact kernel
and the right-hand side of (1.1.8)).
Remark 1.5.3. The estimates (1.5.29), (1.5.30) show that the param-
eters α and h are not equivalent because, regardless of the value of <£(io)>
α(δ) = o(h(6)) and the introduction of additional parameter o; into a self-
regularization difference scheme does not improve the asymptotics of the car-
cass error (but can degrade it if, for example, we assume that h(o) = ο(α(δ}).
Therefore, the parameter h can naturally be interpreted as the basic regu-
larization parameter, whereas α as an auxiliary one.
Let us now show that the obtained estimates and conclusions based on
them are not due to imperfection of the technique applied but reflect the
essence of the matter. For this purpose following Apartsyn (1982), (1985)
we construct examples from which the unimprovability of these estimates
follows since they are achieved as strict equalities.
Let us go back to the example (1.5.4), for which the discrete analog of
(1.5.5) given by
αφ·ι + ϊ ι φ · ι = ih, i = T/r , nh = T (1.5.5Λ)

2 ^ 2
3=1
has a solution
Since ψ(ί) = 1, we have
and therefore
a +h
Hence, the condition α = o(h) is necessary for the α-regularized carcass to
converge in the norm of Ch·
We illustrate the possibility of improving (1.5.28) and (1.5.29) by the
following example.
Consider the equation

ί £3
tp(s)ds = —, 0 < ί < 1 , (1.5.33)
whose exact solution is the function φ(ί) = t2 (the second power guarantees
a nonzero error of the quadrature formula of middle rectangles, which is
important for our purpose). We introduce a saw-type disturbance of the
right-hand side:
Then equation (1.5.10) for the error of the regularized carcass can be written
as
+ £ = (-l)M- + o(t-i) , < = ΜΪ, (1.5.34)
since
fih
Then (1.5.34) implies the recursion formulas
_
,' » = 2,n,
' ' (1.5.35)
V /
ah2
- -,
a+h
Next, we obtain the explicit formula ε0' i = ip(6,a,h,i) from (1.5.35),
l
~2
(1.5.36) and make sure that the optimization of the function ψ with re-
spect to the parameters α(δ) and h(6) results in the estimates of (1.5.28),
(1.5.29).
Set
a (-1Y2S-&
α+η α ...ι.
+h '
From (1.5.35) it follows that
-,· - ξί_,·), i = 2^· (1-5.37)

Since for 0 < λ < 1 we have

i-l
3=0
, it is easy to verify that
i-2
j=0 12
i-2
We need only calculate Applying the rule of summation by
j=o
parts, we have
j=0 j=0 ^ i/= jf=0 i/=0
According to the introduced notation,

1-1.7-1 _ , »-l / / \ j\
t Λ
a +h
ι—· 1-1
j=0 i/=0
h j=o h h h.
and therefore
j=0
Thus, (1.5.37) is transformed as follows:

h* , ah2
/ χ i-l χ
, a -l
a +h 2a + h a +h
2ahi-2a(a Ι-
12 a +h a +h
i — 3,n. (1.5.38)
We need only maximize the rate of convergence of -φ to zero as δ —ϊ Ο
using α(ί) and /ι(<5). Let α(5) χ <J", Λ(ί) χ <Γ, 0 < /9,7 < 1. Since * < 1/Λ
and
as α —>· 0, we need to find the value of μ = max min{27,1 — min(/3,7), β}.

0< ,j<l
It is evident that μ = 1/3 and corresponds to the values β = 2/3, 7 = 1/3,
which proves that (1.5.28), (1.5.29) are unimprovable.
That (1.5.28) and (1.5.30) are unimprovable is established in a similar
way if we consider, for example, the equation
r
Jo
= t+-, i € [0,1], (1.5.39)
with it's exact solution (p(i) = 1 -4- £2 and the same disturbance of the right-
hand side.
Thus, we have proved that the introduction of an additional regular-
ization parameter α into a discrete analog of the integral equation does
not improve the asymptotics of the carcass error and does not change the
asymptotics of the quasi-optimal mesh size, as compared to the pure self-
regularization method. But despite that the parameter α is "asymptotically
useless", the availability of an additional regularization parameter for fixed
5 can considerably improve the accuracy of numerical solution, the effect
of introducing α being greater with larger values of δ. We shall verify the
above statement using the test equation
cos(t — s)</?(s) ds = 1 — cos i, 0 < t < 1, (1.5.40)

/
Jo
whose exact solution is φ(ί) = t.
Table 1.5.1 gives the results of calculations for the pure self-regularization
method. For each level of <5, a saw-type disturbance was introduced into the
right-hand side of (1.5.33) following the above procedure, and the value of
hopi with the corresponding value of ||eft°pt \\ch were determined using the
Fibonacci method (15 tests).
Table 1.5.1.
δ /i0pt
iiôptii
llt "Chovt
ΙΟ-2 0,60238 0,02215
3
ίο- 0,28026 0,00408
10~4 0,13226 0,00081
10~5 0,06092 0,00017
10~6 0,02875 0,00003
It is evident that theoretical estimates for quasi-optimal values of ftq.0.

and ehci ·°· also hold for actually optimal values. (With reduction of δ by a
factor of 1000, hopt diminishes by a factor of 10 and the corresponding error
diminishes at least by a factor of 100, so that /iopt ~ δ* and ||εΛορί||<2Λ =
Table 1.5.2 presents the results for the (a, /i)-algorithm. For fixed δ
and hopi, a one-parametric optimization of the carcass error with respect
to α was also performed using the Fibonacci method by virtue of unimodal
nature of He^Hch as a function of a.
Table 1.5.2.
δ ôpt «opt
\\jrOiopt, hopt \\
H fc nc>ioDt
ίο-2 0,60238 0,02424 0,00907
io-3 0,28026 0,003548 0,00225
ΗΓ4 0,13226 0,00063 0,00048
10~5 0,06092 0,00027 0,00011
10~6 0,02875 0,00002 0,00002
The availability of the second parameter would allow one to reduce the
error of the numerical solution almost to one third with δ = 10~2, and only
by a factor of 1.5 with δ = 10~6. The validity of the same asymptotics
as those theoretically obtained for ftq.o.c., q.o.c., and ||£Qi-°'c'»Vo.c. ||Cfc ^ is
evident.
To illustrate the effect of the kernel disturbance on the carcass error, we
consider another test example borrowed from Keech (1977):
/ (l + t-e)¥>(e)ds = l-cost + sint, t €[0,10]. (1.5.41)

JQ
The function φ(ί) = cost is an exact solution of (1.5.41). Similarly to the

above example, a saw-type disturbance of the kernel
was introduced for fixed 5, and optimization with respect to α was performed
using the Fibonacci method for fixed hi = l/(i + 1), i — 1,2,
Table 1.5.3.
δ ôpt «opt
paoptôptU^
ίο-32 0,5 1,5-HT 3 0,05125

icr 0,25 1,7-10- 5 0,01064
io-4 0,125 1,9-10- 7 0,00225
1(T5 0,066 2, 1 · 10~9 0,00048
In this example, we have φ(0) = ί j£ Q, which accounts for the rapid

decreasing of aopt (δ) preventing the occurrence of a boundary layer. In this
case, the asymptotics of hopt(S) and ||εα°ρ*>Λ°ρ*||£Λ are the same as in
the case of disturbances in the right-hand side.
Remark 1.5.4. The numerical results discussed above were obtained

by Slyusar (1983).
We now consider the possibility of further refinement of the numerical
solution of (1.1.8) using some iteration procedures.
The carcass ψΗι·°· of the approximate solution of the original equation
found using the procedure of self-regularizing discretization (more precisely,
for the case of disturbance in the right-hand side) will be called primary.
Assuming that it is a nonzero approximation, consider the following
iteration process discussed in Apartsyn (1972)
, φ^·°· =^1-0-, (1.5.42)
which is actually a simple iteration method applied (for fixed a) to equation

(1.5.7) multiplied by the regularizing factor μ to provide the transition of
the matrix contraction from iteration to iteration.
Following the general ideology of the principle of the residual in ill-posed
problems (Morozov, 1967, 1968), which consists in conforming the residual
norm of the approximate solution to the accuracy of the input data, we call
the vector φ*' q ° a secondary carcass of the approximate solution if it
satisfies the equality
~a,hq
"•q.o.
= <y. (1.5.43)
Generally speaking, equation (1.5.43) has infinitely many solutions (pairs

(a,n)). On the other hand, for α = 0 we have A/"(0,n) = 0. Then the
32 A. S. Apartsyn. Nonclassical Linear Voiterra Equations ...
minimum α such that (1.5.43) is still solvable for η is denoted by ares. and the
corresponding η is called nres.· Table 1.5.4 gives the results of calculations
for the model equation
ft
I ei [0,1] (1.5.44)
Jo
with the exact solution φ(ί) = 1.
Table 1.5.4
δ ores. ^res.A/"(ares.,nres.) lie**

II
0
· l\\Ch.
|l
-'Aq.o.
Μ ~<lrea. )Ίς.Ο. Μ
litres. \\Ch(1.0.
ΗΓ1 1 3 0,10075 0,3427 0,2148

ίο-2 2-3 7 0,01103 0,0956 0,0839
3
io- 2-7 17 0,00104 0,0012 0,0010
1(T4 2-10 30 0,00010 0,0008 0,0007
Remark 1.5.5. Table 1.5.4 was borrowed from Bulatov (1982).
Remark 1.5.6. Throughout the section, the quadrature formula of mid-

dle rectangles has been used as a basis. But the main results are also valid
for other convergent quadrature methods if there are no disturbances of the
input data. We note the methods of the Runge-Kutta type (Hoog and Weiss,
1973a, 1973b) and a method discussed in Linz (1971) that has second-order
convergence and is intended for the case of strongly oscillating kernels. This
method was successfully applied for solving a real inverse seismic problem
(Apartsyn, Guseva, Ten Men Yan and Shulbaev, 1978; Apartsyn, Guseva,
and Ten Men Yan, 1982).
1.6. INEQUALITIES WITH ISOTONE OPERATORS

Here we consider some generalizations of the classical G.-B. inequality that
plays a major role in studying multidimensional Volterra equations of the
first kind. These generalizations, in turn, can be considered from a single
standpoint if one uses elements of the theory of partially ordered spaces and
monotonous operators.
Chapter 1. Classical VoJterra equations of the first kind 33
For convenience, we give the necessary statements and definitions.
Definition 1.6.1. The set 1C of the linear space Λ is called a cone

if αχ 6 1C for every element χ ζ 1C and any α > 0. A cone 1C is called
convex if χ (Ξ fC, y 6 1C implies χ + y G 1C.
The set A" of the Euclidean space Rn that consists of all vectors
χ = (χι,...,χη) such that Xi > 0, i — l,n and the set C+{a^ of all
continuous nonnegative functions on [a, b] are important examples of convex
cones.
Definition 1.6.2. Let /C be a convex cone in A. If χ £ /C, then we

put χ > 0 and the element χ will be called nonnegative. If x, y e A and
z — y € /C, then we put x > y and say that a; is more than or equal to y.
The space Λ is said to be partially ordered by the cone 1C .
The relation > introduced above has the following properties:
1) χ > χ (reflexivity) ;
2) x >y /\y > z ==>· χ > z (transitivity);
3 ) : r > y / \ a > 0 => ax > ay (positive homogeneity) ;
4) x\>yif\X2> yi =>· x\ + x-2 > y\ + yi (additivity).
The relation < (less or equal) is used along with the relation >. The
inequality y < χ for x, y e A means that χ — y € /C.
Definition 1.6.3. An operator L : A -» A is called isotone (with

respect to the cone /C) if χ — y E fC implies L(x — y) 6 /C.
Definition 1.6.4. An operator L : A -> A is called antitone (with

respect to the cone /C) if χ — y € /C implies L(y - x) & 1C.
Example 1.6.1. An integral Volterra operator in (1.1.8) with contin-
uous nonnegative kernel A"(i, s) in the domain Δ is isotone with respect to
the cone C+μ^; the same operator with nonpositive kernel K(t, s) in Δ
is antitone.
Example 1.6.2. An integral Volterra operator V : Οπη ->·
rti rtn
V(p== I ··· ΙΙ # ( < ι , . . . , £ n ; s i , . . . , s n ) ^ ( s i , . . . , 5 n ) d s i · · · dsn,
./αϊ
αϊ Jan
TTn = {o>i < ti < Tj, i = l,n}, with continuous nonnegative kernel in the
domain Ωη = {αϊ < Si < ti < Ti, i = l,n} is isotone with respect to the
cone C+πη ; the same operator with nonpositive kernel in the domain Ωη
is antitone.
Remark 1.6.1. Prom the definition of an isotone (antitone) operator

and properties of the relation > , it directly follows that the sum and super-
position of isotone (antitone) operators are also isotone (antitone) operators.
The next statement is of special importance for our objectives.
Lemma 1.6.1. Let A be a Banach space, V a linear operator bounded

on Λ and isotone with respect to the cone 1C , and let E — V have bounded
oo
inverse operator (E — V}~1 = £) Vn (E is the identity operator in A).
n=0
Then solutions of the inequality
(1.6.1)
are majorized by a (unique) solution of the equation
ψ = / + νψ, (1.6.2)
that is
φ<ψ. (1.6.3)
P r o o f . Subtracting (1.6.1) from (1.6.2), we get
φ-φ>ν(φ-φ}. (1.6.4)
Denoting ψ — φ = η, we have
η > Υη.
Hence,
η = νη + ξ, (1.6.5)
where £ > 0 ( £ e / C ) .
As follows from the assumption,
00
i=0
Since the operator V is isotone, V*£ € /C V i = 1, 2, · · · and (E + V -\ ----- l·

Vi + · · ·)ξ € Κ, (see Remark 1.6.1). Thus, η 6 /C, i.e., η > 0, which yields
(1.6.3). .
Remark 1.6.2. The estimate (1.6.3) is unimprovable in the sense that

it becomes an equality if (1.6.1) is so.
Let us illustrate the assertion of Lemma 1.6.1 taking inequality (1.2.1)
as an example.
The Volterra operator from (1.8.8) with K(t,s) = g(s) > 0 is isotone
in (7[ί0)τ] with respect to the cone C+[tQjT^ therefore, solutions of the
inequality
w(i) < c + i g(s}u(s) ds, t 6 [ί0, Τ]
Jto
are majorized by the solution of the equation
w(t}=c+ f g ( s ) w ( s ) d s , ie[t 0 ,T]. (1.6.7)

Jt0
We find this solution using the method of successive approximations or,
equivalently, by constructing the Neumann series
oo
i=0
Let
-ί
/ g(s)wi-i(s)ds, » = 1,2,...; w0(t) = c; te[t0,T].
to
Jto
Since
, » = 1,2,...,
Z·
from (1.6.8) we immediately obtain
w(t) = c ^ — '— = ci
i=0
which coincides with the G.-B. estimate (1.2.2).
Thus, the classical G.-B. inequality is unimprovable. This conclusion is
naturally valid for (1.2.1'), (1.2.2'). The solution of (1.6.7) and, moreover,
of the equation
w(t)=c + m w(s)ds, te[t0,T], (1.6.10)

Jto
can also be found using some other techniques due to the simplicity of both
equations. We demonstrate these techniques using (1.6.10) as an example.
Differentiation of (1.6.10) reduces it to an equivalent Cauchy problem
w'(t) = mtu(i), t E [ΐ0,Τ], (1.6.11)

tu(to) = c, (1.6.12)
whose solution is obvious:
which coincides with (1.6.9) for g(s) = m.

It is also reasonable to use the method of undetermined coefficients.
Assume that oo
*>)*. (L6·13)
Substituting (1.6.13) into (1.6.10) and equating the coefficients of the same
powers of t — ίο, we obtain the recursion formulas
= —r—, ζ = 1,2,..., CQ = c, (1.6.14)
which implies
and (1.6.13) again yields the same result.

Consider now the two-dimensional inequality
Γ^Ι /"^2
ti(ii,i 2 ) < c + m / I u(si,s 2 )dsids 2 ,
Jo Jo
c, m > 0, 0 < ti < Ti, » = 1,2 (1.6.15)
(for simplicity, the lower limits in (1.6.15) are equal to zero without loss of
generality).
For (1.6.15), the Vendroff estimate is known (see Bellman and Becken-
bach, 1965):
u(ii,i 2 )<ce m i l i 2 . (1.6.16)
Since the Volterra operator
fl ft2
Vu = m I n(si,S2)dsids2 (1.6.17)
JQ Jo
is isotone with respect to the cone C+n2 (see Example 1.6.2), according to
Lemma 1.6.1, the solution of the equation
ti rti
I w(si,S2)dsids2, * ι , < 2 € π 2 , (1.6.18)
/ ./ο
is an unimprovable estimate of solutions of inequality (1.6.15).
Since oo
w(t1,t2) = ^Vic, ί ι , ί 2 € π 2 , (1.6.19)
t=0
where V is the operator (1.6.17) and
Vc-c ( ., )2 , i-1,2,...,
we conclude that (1.6.19) implies
—, ίι,ί 2 € 7Γ2. (1.6.20)

i=o
Compare (1.6.20) to the VendrofF estimate (1.6.16). Since
i=o ·
we have
<ce m i l i 2
which is a natural consequence of the fact that (1.6.20) is unimprovable.
According to the method of undetermined coefficients, the same formula
(1.6.20) can be obtained by substituting the series
oo
(1.6.21)
i=0
into (1.6.18), which yields the recursion formula
-, ι = ι, ζ , . . . , CQ = c,
Ζ"
whence
T^c, (1.6.22)
and substitution of (1.6.22) into (1.6.21) gives (1.6.20).

A different representation could be obtained for the series (1.6,20) if we

turn from the integral equation (1.6.18) to its differential analog.
Applying the operator &tddt to both sides of (1.6.18), we obtain
< ί 2 (ίι,ί 2 ) = ™™(ίι,ί 2 ), (1.6.23)

and (1.6.28) implies
w(Q,t2)=w(ti,Q)=c. (1.6.24)
The general solution of equation (1.6.23) is well known (for example, see
Vladimirov et al, 1974):
! / _ \
/ι (β) Jo 2i^mt2(ti - s) ds+
/ \ J
rt-z / _ \ / \ /
+ / f2(S)J0(2i^rnt1(t2-s))ds + /i(0) + /2(0) J0
Jo \ / \ / \
(1.6.25)
where Jo is the zeroth order Bessel function, i is the imaginary unit, /i and
/2 are arbitrary functions of class C^.
Hence, taking into account (1.6.24), we find that the function
w(ti,t2) = cJo2iVmiii 2 Y (1.6.26)
is a solution of the Goursat problem (1.6.23), (1.6.24).

Due to the equivalence of (1.6.18) and (1.6.23), (1.6.24), the right-hand
sides of (1.6.20), (1.6.26) must coincide.
Indeed, by virtue of the known decomposition of the zeroth order Bessel
function (Nikiforov and Uvarov, 1978)
with ζ = 2i,/mtit2, (1.6.27) yields
0. / , . .
2tVmtlt2\ = ^ -2 2 *(fc!l 2- ' (1.6.28)
/ V
Jfc=0 ''
and the series in (1.6.20) and (1.6.28) coincide as i2 = — 1.

When we turn to more general two-dimensional integral inequalities, we
shall have to use both the isotone property of integral operators and their
interchangeability.
1.7. INEQUALITIES WITH INTERCHANGEABLE

ISOTONE OPERATORS
Let the operator V in Lemma 1.6.1 be the sum of n isotone operators V^,
i = l,n:
V = V1 + "- + Vn, (1.7.1)
such that the solution of equation (1.6.5) is
oo
fc=0
Assume that the operators Vi, i = l, n commute with one another. Then,
using the combinatorial formula of Vilenkin (1969)
k_ y> fe! ^ in _
ι 4 ί ι 1 · · . 7_ 1
(1.7.2) can be rewritten in the form

oo
>-^ Λ!
Σ Σ -ΤΓΓ
In particular, for η = 2 we have

oo fc
k=0 i=0
where Clk is the number of combinations of i unordered elements from k

possibilities.
Consider the integral inequality
ti rti
u(s, t<z) as + 7Ti2 / w(ii,s)ds, ί ι , < 2 € π 2 , (1.7.6)
/ Jo
whose VendrofT estimate (see Bellman and Beckenbach (1965), p. 215) has
the form
u(ti,i 2 ) <ce miil+m2i2+mim2tlt2 . (1-7.7)
As isotone operators
/·<! ftl
= m\ I u(s,<2)ds, Vzu = m2 I u(ii,s)ds (1.7.8)
7o Jo
Discovering Diverse Content Through
Random Scribd Documents
man in evening dress, with a ledger before him. The new arrival shut
the door carefully after him; it clicked to, on a spring-lock. Advancing
to the desk, he announced, "Number Twenty-one, sir," and stood
respectfully waiting. The big man looked up, showing the number 1
startlingly white on his velvet mask. His eyes, of a curious hard blue,
scanned Rogers attentively. At a sign from him, Rogers removed his
mask. Having verified his identity with care, the President said, "Very
well, Number Twenty-one," and made an entry in the ledger. The
voice was hard and metallic, like his eyes. The close scrutiny from
behind the immovable black mask seemed to make Rogers uneasy;
he shifted his feet, and his eyes fell. Number One made a sign of
dismissal, and Rogers, with a faint sigh as though of relief, replaced
his mask and left the room. As he came out, the next comer passed
in in his place.
The room in which the Society met was a large one, made by
knocking the two largest of the first-floor rooms into one. It was
furnished in the standardised taste of twentieth-century suburbia and
brilliantly lighted. A gramophone in one corner blared out a jazz tune,
to which about ten couples of masked men and women were
dancing, some in evening dress and others in tweeds and jumpers.
In one corner of the room was an American bar. Rogers went up and
asked the masked man in charge for a double whisky. He consumed
it slowly, leaning on the bar. The room filled. Presently somebody
moved across to the gramophone and stopped it. He looked round.
Number One had appeared on the threshold. A tall woman in black
stood beside him. The mask, embroidered with a white 2, covered
hair and face completely; only her fine bearing and her white arms
and bosom and the dark eyes shining through the eye-slits
proclaimed her a woman of power and physical attraction.
"Ladies and gentlemen." Number One was standing at the upper end
of the room. The woman sat beside him; her eyes were cast down
and betrayed nothing, but her hands were clenched on the arms of
the chair and her whole figure seemed tensely aware.
"Ladies and gentlemen. Our numbers are two short to-night." The
masks moved; eyes were turned, seeking and counting. "I need not
inform you of the disastrous failure of our plan for securing the plans
of the Court-Windlesham helicopter. Our courageous and devoted
comrades, Number Fifteen and Number Forty-eight, were betrayed
and taken by the police."
An uneasy murmur rose among the company.
"It may have occurred to some of you that even the well-known
steadfastness of these comrades might give way under examination.
There is no cause for alarm. The usual orders have been issued,
and I have this evening received the report that their tongues have
been effectually silenced. You will, I am sure, be glad to know that
these two brave men have been spared the ordeal of so great a
temptation to dishonour, and that they will not be called upon to face
a public trial and the rigours of a long imprisonment."
A hiss of intaken breath moved across the assembled members like
the wind over a barley-field.
"Their dependants will be discreetly compensated in the usual
manner. I call upon Numbers Twelve and Thirty-four to undertake
this agreeable task. They will attend me in my office for their
instructions after the meeting. Will the Numbers I have named kindly
signify that they are able and willing to perform this duty?"
Two hands were raised in salute. The President continued, looking at
his watch:
"Ladies and gentlemen, please take your partners for the next
dance."
The gramophone struck up again. Rogers turned to a girl near him in
a red dress. She nodded, and they slipped into the movement of a
fox-trot. The couples gyrated solemnly and in silence. Their shadows
were flung against the blinds as they turned and stepped to and fro.
"What has happened?" breathed the girl in a whisper, scarcely
moving her lips. "I'm frightened, aren't you? I feel as if something
awful was going to happen."
"It does take one a bit short, the President's way of doing things,"
agreed Rogers, "but it's safer like that."
"Those poor men——"
A dancer, turning and following on their heels, touched Rogers on
the shoulder.
"No talking, please," he said. His eyes gleamed sternly; he twirled
his partner into the middle of the crowd and was gone. The girl
shuddered.
The gramophone stopped. There was a burst of clapping. The
dancers again clustered before the President's seat.
"Ladies and gentlemen. You may wonder why this extraordinary
meeting has been called. The reason is a serious one. The failure of
our recent attempt was no accident. The police were not on the
premises that night by chance. We have a traitor among us."
Partners who had been standing close together fell distrustfully
apart. Each member seemed to shrink, as a snail shrinks from the
touch of a finger.
"You will remember the disappointing outcome of the Dinglewood
affair," went on the President, in his harsh voice. "You may recall
other smaller matters which have not turned out satisfactorily. All
these troubles have been traced to their origin. I am happy to say
that our minds can now be easy. The offender has been discovered
and will be removed. There will be no more mistakes. The misguided
member who introduced the traitor to our Society will be placed in a
position where his lack of caution will have no further ill-effects.
There is no cause for alarm."
Every eye roved about the company, searching for the traitor and his
unfortunate sponsor. Somewhere beneath the black masks a face
must have turned white; somewhere under the stifling velvet there
must have been a brow sweating, not with the heat of the dance. But
the masks hid everything.
"Ladies and gentlemen, please take your partners for the next
dance."
The gramophone struck into an old and half-forgotten tune: "There
ain't nobody loves me." The girl in red was claimed by a tall mask in
evening dress. A hand laid on Roger's arm made him start. A small,
plump woman in a green jumper slipped a cold hand into his. The
dance went on.
When it stopped, amid the usual applause, everyone stood,
detached, stiffened in expectation. The President's voice was raised
again.
"Ladies and gentlemen, please behave naturally. This is a dance, not
a public meeting."
Rogers led his partner to a chair and fetched her an ice. As he
stooped over her, he noticed the hurried rise and fall of her bosom.
"Ladies and gentlemen." The endless interval was over. "You will no
doubt wish to be immediately relieved from suspense. I will name the
persons involved. Number Thirty-seven!"
A man sprang up with a fearful, strangled cry.
"Silence!"
The wretch choked and gasped.
"I never—I swear I never—I'm innocent."
"Silence. You have failed in discretion. You will be dealt with. If you
have anything to say in defence of your folly, I will hear it later. Sit
down."
Number Thirty-seven sank down upon a chair. He pushed his
handkerchief under the mask to wipe his face. Two tall men closed in
upon him. The rest fell back, feeling the recoil of humanity from one
stricken by mortal disease.
The gramophone struck up.
"Ladies and gentlemen, I will now name the traitor. Number Twenty-
one, stand forward."
Rogers stepped forward. The concentrated fear and loathing of forty-
eight pairs of eyes burned upon him. The miserable Jukes set up a
fresh wail.
"Oh, my God! Oh, my God!"
"Silence! Number Twenty-one, take off your mask."
The traitor pulled the thick covering from his face. The intense hatred
of the eyes devoured him.
"Number Thirty-seven, this man was introduced here by you, under
the name of Joseph Rogers, formerly second footman in the service
of the Duke of Denver, dismissed for pilfering. Did you take steps to
verify that statement?"
"I did—I did! As God's my witness, it was all straight. I had him
identified by two of the servants. I made enquiries. The tale was
straight—I'll swear it was."
The President consulted a paper before him, then he looked at his
watch again.
"Ladies and gentlemen, please take your partners...."
Number Twenty-one, his arms twisted behind him and bound, and
his wrists hand-cuffed, stood motionless, while the dance of doom
circled about him. The clapping, as it ended, sounded like the
clapping of the men and women who sat, thirsty-lipped, beneath the
guillotine.
"Number Twenty-one, your name has been given as Joseph Rogers,
footman, dismissed for theft. Is that your real name?"
"No."
"What is your name?"
"Peter Death Bredon Wimsey."
"We thought you were dead."
"Naturally. You were intended to think so."
"What has become of the genuine Joseph Rogers?"
"He died abroad. I took his place. I may say that no real blame
attaches to your people for not having realised who I was. I not only
took Roger's place; I was Rogers. Even when I was alone, I walked
like Rogers, I sat like Rogers, I read Rogers's books, and wore
Rogers's clothes. In the end, I almost thought Rogers's thoughts.
The only way to keep up a successful impersonation is never to
relax."
"I see. The robbery of your own flat was arranged?"
"Obviously."
"The robbery of the Dowager Duchess, your mother, was connived
at by you?"
"It was. It was a very ugly tiara—no real loss to anybody with decent
taste. May I smoke, by the way?"
"You may not. Ladies and gentlemen...."
The dance was like the mechanical jigging of puppets. Limbs jerked,
feet faltered. The prisoner watched with an air of critical detachment.
"Numbers Fifteen, Twenty-two and Forty-nine. You have watched the
prisoner. Has he made any attempts to communicate with anybody?"
"None." Number Twenty-two was the spokesman. "His letters and
parcels have been opened, his telephone tapped, and his
movements followed. His water-pipes have been under observation
for Morse signals."
"You are sure of what you say?"
"Absolutely."
"Prisoner, have you been alone in this adventure? Speak the truth,
or things will be made somewhat more unpleasant for you than they
might otherwise be."
"I have been alone. I have taken no unnecessary risks."
"It may be so. It will, however, be as well that steps should be taken
to silence the man at Scotland Yard—what is his name?—Parker.
Also the prisoner's manservant, Mervyn Bunter, and possibly also his
mother and sister. The brother is a stupid oaf, and not, I think, likely
to have been taken into the prisoner's confidence. A precautionary
watch will, I think, meet the necessities of his case."
The prisoner appeared, for the first time, to be moved.
"Sir, I assure you that my mother and sister know nothing which
could possibly bring danger on the Society."
"You should have thought of their situation earlier. Ladies and
gentlemen, please take——"
"No—no!" Flesh and blood could endure the mockery no longer. "No!
Finish with him. Get it over. Break up the meeting. It's dangerous.
The police——"
"Silence!"
The President glanced round at the crowd. It had a dangerous look
about it. He gave way.
"Very well. Take the prisoner away and silence him. He will receive
Number 4 treatment. And be sure you explain it to him carefully first."
"Ah!"
The eyes expressed a wolfish satisfaction. Strong hands gripped
Wimsey's arms.
"One moment—for God's sake let me die decently."
"You should have thought this over earlier. Take him away. Ladies
and gentlemen, be satisfied—he will not die quickly."
"Stop! Wait!" cried Wimsey desperately. "I have something to say. I
don't ask for life—only for a quick death. I—I have something to sell."
"To sell?"
"Yes."
"We make no bargains with traitors."
"No—but listen! Do you think I have not thought of this? I am not so
mad. I have left a letter."
"Ah! now it is coming. A letter. To whom?"
"To the police. If I do not return to-morrow——"
"Well?"
"The letter will be opened."
"Sir," broke in Number Fifteen. "This is bluff. The prisoner has not
sent any letter. He has been strictly watched for many months."
"Ah! but listen. I left the letter before I came to Lambeth."
"Then it can contain no information of value."
"Oh, but it does."
"What?"
"The combination of my safe."
"Indeed? Has this man's safe been searched?"
"Yes, sir."
"What did it contain?"
"No information of importance, sir. An outline of our organisation—
the name of this house—nothing that cannot be altered and covered
before morning."
Wimsey smiled.
"Did you investigate the inner compartment of the safe?"
There was a pause.
"You hear what he says," snapped the President sharply. "Did you
find this inner compartment?"
"There was no inner compartment, sir. He is trying to bluff."
"I hate to contradict you," said Wimsey, with an effort at his ordinary
pleasant tone, "but I really think you must have overlooked the inner
compartment."
"Well," said the President, "and what do you say is in this inner
compartment, if it does exist?"
"The names of every member of this Society, with their addresses,
photographs, and finger-prints."
"What?"
The eyes round him now were ugly with fear. Wimsey kept his face
steadily turned towards the President.
"How do you say you have contrived to get this information?"
"Well, I have been doing a little detective work on my own, you
know."
"But you have been watched."
"True. The finger-prints of my watchers adorn the first page of the
collection."
"This statement can be proved?"
"Certainly. I will prove it. The name of Number Fifty, for example——"
"Stop!"
A fierce muttering arose. The President silenced it with a gesture.
"If you mention names here, you will certainly have no hope of
mercy. There is a fifth treatment—kept specially for people who
mention names. Bring the prisoner to my office. Keep the dance
going."
The President took an automatic from his hip-pocket and faced his
tightly fettered prisoner across the desk.
"Now speak!" he said.
"I should put that thing away, if I were you," said Wimsey
contemptuously. "It would be a much pleasanter form of death than
treatment Number 5, and I might be tempted to ask for it."
"Ingenious," said the President, "but a little too ingenious. Now, be
quick; tell me what you know."
"Will you spare me if I tell you?"
"I make no promises. Be quick."
Wimsey shrugged his bound and aching shoulders.
"Certainly. I will tell you what I know. Stop me when you have heard
enough."
He leaned forward and spoke low. Overhead the noise of the
gramophone and the shuffling of feet bore witness that the dance
was going on. Stray passers-by crossing the Heath noted that the
people in the lonely house were making a night of it again.
"Well," said Wimsey, "am I to go on?"

From beneath the mask the President's voice sounded as though he
were grimly smiling.
"My lord," he said, "your story fills me with regret that you are not, in
fact, a member of our Society. Wit, courage, and industry are
valuable to an association like ours. I fear I cannot persuade you?
No—I supposed not."
He touched a bell on his desk.
"Ask the members kindly to proceed to the supper-room," he said to
the mask who entered.
The "supper-room" was on the ground-floor, shuttered and curtained.
Down its centre ran a long, bare table, with chairs set about it.
"A Barmecide feast, I see," said Wimsey pleasantly. It was the first
time he had seen this room. At the far end, a trap-door in the floor
gaped ominously.
The President took the head of the table.
"Ladies and gentlemen," he began, as usual—and the foolish
courtesy had never sounded so sinister—"I will not conceal from you
the seriousness of the situation. The prisoner has recited to me more
than twenty names and addresses which were thought to be
unknown, except to their owners and to me. There has been great
carelessness"—his voice rang harshly—"which will have to be
looked into. Finger-prints have been obtained—he has shown me
the photographs of some of them. How our investigators came to
overlook the inner door of this safe is a matter which calls for
enquiry."
"Don't blame them," put in Wimsey. "It was meant to be overlooked,
you know. I made it like that on purpose."
The President went on, without seeming to notice the interruption.
"The prisoner informs me that the book with the names and
addresses is to be found in this inner compartment, together with
certain letters and papers stolen from the houses of members, and
numerous objects bearing authentic finger-prints. I believe him to be
telling the truth. He offers the combination of the safe in exchange
for a quick death. I think the offer should be accepted. What is your
opinion, ladies and gentlemen?"
"The combination is known already," said Number Twenty-two.
"Imbecile! This man has told us, and has proved to me, that he is
Lord Peter Wimsey. Do you think he will have forgotten to alter the
combination? And then there is the secret of the inner door. If he
disappears to-night and the police enter his house——"
"I say," said a woman's rich voice, "that the promise should be given
and the information used—and quickly. Time is getting short."
A murmur of agreement went round the table.
"You hear," said the President, addressing Wimsey. "The Society
offers you the privilege of a quick death in return for the combination
of the safe and the secret of the inner door."
"I have your word for it?"
"You have."
"Thank you. And my mother and sister?"
"If you in your turn will give us your word—you are a man of honour
—that these women know nothing that could harm us, they shall be
spared."
"Thank you, sir. You may rest assured, upon my honour, that they
know nothing. I should not think of burdening any woman with such
dangerous secrets—particularly those who are dear to me."
"Very well. It is agreed—yes?"
The murmur of assent was given, though with less readiness than
before.
"Then I am willing to give you the information you want. The word of
the combination is UNRELIABILITY."
"And the inner door?"
"In anticipation of the visit of the police, the inner door—which might
have presented difficulties—is open."
"Good! You understand that if the police interfere with our
messenger——"
"That would not help me, would it?"
"It is a risk," said the President thoughtfully, "but a risk which I think
we must take. Carry the prisoner down to the cellar. He can amuse
himself by contemplating apparatus Number 5. In the meantime,
Numbers Twelve and Forty-six——"
"No, no!"
A sullen mutter of dissent arose and swelled threateningly.
"No," said a tall man with a voice like treacle. "No—why should any
members be put in possession of this evidence? We have found one
traitor among us to-night and more than one fool. How are we to
know that Numbers Twelve and Forty-six are not fools and traitors
also?"
The two men turned savagely upon the speaker, but a girl's voice
struck into the discussion, high and agitated.
"Hear, hear! That's right, I say. How about us? We ain't going to have
our names read by somebody we don't know nothing about. I've had
enough of this. They might sell the 'ole lot of us to the narks."
"I agree," said another member. "Nobody ought to be trusted,
nobody at all."
The President shrugged his shoulders.
"Then what, ladies and gentlemen, do you suggest?"
There was a pause. Then the same girl shrilled out again:
"I say Mr. President oughter go himself. He's the only one as knows
all the names. It won't be no cop to him. Why should we take all the
risk and trouble and him sit at home and collar the money? Let him
go himself, that's what I say."
A long rustle of approbation went round the table.
"I second that motion," said a stout man who wore a bunch of gold
seals at his fob. Wimsey smiled as he looked at the seals; it was that
trifling vanity which had led him directly to the name and address of
the stout man, and he felt a certain affection for the trinkets on that
account.
The President looked round.
"It is the wish of the meeting, then, that I should go?" he said, in an
ominous voice.
Forty-five hands were raised in approbation. Only the woman known
as Number Two remained motionless and silent, her strong white
hands clenched on the arm of the chair.
The President rolled his eyes slowly round the threatening ring till
they rested upon her.
"Am I to take it that this vote is unanimous?" he enquired.
The woman raised her head.
"Don't go," she gasped faintly.
"You hear," said the President, in a faintly derisive tone. "This lady
says, don't go."
"I submit that what Number Two says is neither here nor there," said
the man with the treacly voice. "Our own ladies might not like us to
be going, if they were in madam's privileged position." His voice was
an insult.
"Hear, hear!" cried another man. "This is a democratic society, this is.
We don't want no privileged classes."
"Very well," said the President. "You hear, Number Two. The feeling
of the meeting is against you. Have you any reasons to put forward
in favour of your opinion?"
"A hundred. The President is the head and soul of our Society. If
anything should happen to him—where should we be? You"—she
swept the company magnificently with her eyes—"you have all
blundered. We have your carelessness to thank for all this. Do you
think we should be safe for five minutes if the President were not
here to repair your follies?"
"Something in that," said a man who had not hitherto spoken.
"Pardon my suggesting," said Wimsey maliciously, "that, as the lady
appears to be in a position peculiarly favourable for the reception of
the President's confidences, the contents of my modest volume will
probably be no news to her. Why should not Number Two go
herself?"
"Because I say she must not," said the President sternly, checking
the quick reply that rose to his companion's lips. "If it is the will of the
meeting, I will go. Give me the key of the house."
One of the men extracted it from Wimsey's jacket-pocket and
handed it over.
"Is the house watched?" he demanded of Wimsey.
"No."
"That is the truth?"
"It is the truth."
The President turned at the door.
"If I have not returned in two hours' time," he said, "act for the best to
save yourselves, and do what you like with the prisoner. Number
Two will give orders in my absence."
He left the room. Number Two rose from her seat with a gesture of
command.
"Ladies and gentlemen. Supper is now considered over. Start the
dancing again."
Down in the cellar the time passed slowly, in the contemplation of

apparatus Number 5. The miserable Jukes, alternately wailing and
raving, at length shrieked himself into exhaustion. The four members
guarding the prisoners whispered together from time to time.
"An hour and a half since the President left," said one.
Wimsey glanced up. Then he returned to his examination of the
room. There were many curious things in it, which he wanted to
memorise.
Presently the trap-door was flung open. "Bring him up!" cried a voice.
Wimsey rose immediately, and his face was rather pale.
The members of the gang were again seated round the table.
Number Two occupied the President's chair, and her eyes fastened
on Wimsey's face with a tigerish fury, but when she spoke it was with
a self-control which roused his admiration.
"The President has been two hours gone," she said. "What has
happened to him? Traitor twice over—what has happened to him?"
"How should I know?" said Wimsey. "Perhaps he has looked after
Number One and gone while the going was good!"
She sprang up with a little cry of rage, and came close to him.
"Beast! liar!" she said, and struck him on the mouth. "You know he
would never do that. He is faithful to his friends. What have you done
with him? Speak—or I will make you speak. You two, there—bring
the irons. He shall speak!"
"I can only form a guess, madame," replied Wimsey, "and I shall not
guess any the better for being stimulated with hot irons, like
Pantaloon at the circus. Calm yourself, and I will tell you what I think.
I think—indeed, I greatly fear—that Monsieur le Président in his
hurry to examine the interesting exhibits in my safe may, quite
inadvertently, no doubt, have let the door of the inner compartment
close behind him. In which case——"
He raised his eyebrows, his shoulders being too sore for shrugging,
and gazed at her with a limpid and innocent regret.
"What do you mean?"
Wimsey glanced round the circle.
"I think," he said, "I had better begin from the beginning by explaining
to you the mechanism of my safe. It is rather a nice safe," he added
plaintively. "I invented the idea myself—not the principle of its
working, of course; that is a matter for scientists—but just the idea of
the thing.
"The combination I gave you is perfectly correct as far as it goes. It is
a three-alphabet thirteen-letter lock by Bunn & Fishett—a very good
one of its kind. It opens the outer door, leading into the ordinary
strong-room, where I keep my cash and my Froth Blower's cuff-links
and all that. But there is an inner compartment with two doors, which
open in quite a different manner. The outermost of these two inner
doors is merely a thin steel skin, painted to look like the back of the
safe and fitting closely, so as not to betray any join. It lies in the
same plane as the wall of the room, you understand, so that if you
were to measure the outside and the inside of the safe you would
discover no discrepancy. It opens outwards with an ordinary key,
and, as I truly assured the President, it was left open when I quitted
my flat."
"Do you think," said the woman sneeringly, "that the President is so
simple as to be caught in a so obvious trap? He will have wedged
open that inner door undoubtedly."
"Undoubtedly, madame. But the sole purpose of that outer inner
door, if I may so express myself, is to appear to be the only inner
door. But hidden behind the hinge of that door is another door, a
sliding panel, set so closely in the thickness of the wall that you
would hardly see it unless you knew it was there. This door was also
left open. Our revered Number One had nothing to do but to walk
straight through into the inner compartment of the safe, which, by the
way, is built into the chimney of the old basement kitchen, which runs
up the house at that point. I hope I make myself clear?"
"Yes, yes—get on. Make your story short."
Wimsey bowed, and, speaking with even greater deliberation than
ever, resumed:
"Now, this interesting list of the Society's activities, which I have had
the honour of compiling, is written in a very large book—bigger,
even, than Monsieur le Président's ledger which he uses downstairs.
(I trust, by the way, madame, that you have borne in mind the
necessity of putting that ledger in a safe place. Apart from the risk of
investigation by some officious policeman, it would be inadvisable
that any junior member of the Society should get hold of it. The
feeling of the meeting would, I fancy, be opposed to such an
occurrence.)"
"It is secure," she answered hastily. "Mon dieu! get on with your
story."
"Thank you—you have relieved my mind. Very good. This big book
lies on a steel shelf at the back of the inner compartment. Just a
moment. I have not described this inner compartment to you. It is six
feet high, three feet wide, and three feet deep. One can stand up in it
quite comfortably, unless one is very tall. It suits me nicely—as you
may see, I am not more than five feet eight and a half. The President
has the advantage of me in height; he might be a little cramped, but
there would be room for him to squat if he grew tired of standing. By
the way, I don't know if you know it, but you have tied me up rather
tightly."
"I would have you tied till your bones were locked together. Beat him,
you! He is trying to gain time."
"If you beat me," said Wimsey, "I'm damned if I'll speak at all. Control
yourself, madame; it does not do to move hastily when your king is in
check."
"Get on!' she cried again, stamping with rage.
"Where was I? Ah! the inner compartment. As I say, it is a little snug
—the more so that it is not ventilated in any way. Did I mention that
the book lay on a steel shelf?"
"You did."
"Yes. The steel shelf is balanced on a very delicate concealed
spring. When the weight of the book—a heavy one, as I said—is
lifted, the shelf rises almost imperceptibly. In rising it makes an
electrical contact. Imagine to yourself, madame; our revered
President steps in—propping the false door open behind him—he
sees the book—quickly he snatches it up. To make sure that it is the
right one, he opens it—he studies the pages. He looks about for the
other objects I have mentioned, which bear the marks of finger-
prints. And silently, but very, very quickly—you can imagine it, can
you not?—the secret panel, released by the rising of the shelf, leaps
across like a panther behind him. Rather a trite simile, but apt, don't
you think?"
"My God! oh, my God!" Her hand went up as though to tear the
choking mask from her face. "You—you devil—devil! What is the
word that opens the inner door? Quick! I will have it torn out of you—
the word!"
"It is not a hard word to remember, madame—though it has been
forgotten before now. Do you recollect, when you were a child, being
told the tale of 'Ali Baba and the Forty Thieves'? When I had that
door made, my mind reverted, with rather a pretty touch of
sentimentality, in my opinion, to the happy hours of my childhood.
The words that open the door are—'Open Sesame'."
"Ah! How long can a man live in this devil's trap of yours?"
"Oh," said Wimsey cheerfully, "I should think he might hold out a few
hours if he kept cool and didn't use up the available oxygen by
shouting and hammering. If we went there at once, I dare say we
should find him fairly all right."
"I shall go myself. Take this man and—do your worst with him. Don't
finish him till I come back. I want to see him die!"
"One moment," said Wimsey, unmoved by this amiable wish. "I think
you had better take me with you."
"Why—why?"
"Because, you see, I'm the only person who can open the door."
"But you have given me the word. Was that a lie?"
"No—the word's all right. But, you see, it's one of these new-style
electric doors. In fact, it's really the very latest thing in doors. I'm
rather proud of it. It opens to the words 'Open Sesame' all right—but
to my voice only."
"Your voice? I will choke your voice with my own hands. What do you
mean—your voice only?"
"Just what I say. Don't clutch my throat like that, or you may alter my
voice so that the door won't recognise it. That's better. It's apt to be
rather pernickety about voices. It got stuck up for a week once, when
I had a cold and could only implore it in a hoarse whisper. Even in
the ordinary way, I sometimes have to try several times before I hit
on the exact right intonation."
She turned and appealed to a short, thick-set man standing beside
her.
"Is this true? Is it possible?"
"Perfectly, ma'am, I'm afraid," said the man civilly. From his voice
Wimsey took him to be a superior workman of some kind—probably
an engineer.
"Is it an electrical device? Do you understand it?"
"Yes, ma'am. It will have a microphone arrangement somewhere,
which converts the sound into a series of vibrations controlling an
electric needle. When the needle has traced the correct pattern, the
circuit is completed and the door opens. The same thing can be
done by light vibrations equally easily."
"Couldn't you open it with tools?"
"In time, yes, ma'am. But only by smashing the mechanism, which is
probably well protected."
"You may take that for granted," interjected Wimsey reassuringly.
She put her hands to her head.
"I'm afraid we're done in," said the engineer, with a kind of respect in
his tone for a good job of work.
"No—wait! Somebody must know—the workmen who made this
thing?"
"In Germany," said Wimsey briefly.
"Or—yes, yes, I have it—a gramophone. This—this—he—shall be
made to say the word for us. Quick—how can it be done?"
"Not possible, ma'am. Where should we get the apparatus at half-
past three on a Sunday morning? The poor gentleman would be
dead long before——"
There was a silence, during which the sounds of the wakening day
came through the shuttered windows. A motor-horn sounded
distantly.
"I give in," she said. "We must let him go. Take the ropes off him. You
will free him, won't you?" she went on, turning piteously to Wimsey.
"Devil as you are, you are not such a devil as that! You will go
straight back and save him!"
"Let him go, nothing!" broke in one of the men. "He doesn't go to
peach to the police, my lady, don't you think it. The President's done
in, that's all, and we'd all better make tracks while we can. It's all up,
boys. Chuck this fellow down the cellar and fasten him in, so he can't
make a row and wake the place up. I'm going to destroy the ledgers.
You can see it done if you don't trust me. And you, Thirty, you know
where the switch is. Give us a quarter of an hour to clear, and then
you can blow the place to glory."
"No! You can't go—you can't leave him to die—your President—your
leader—my—I won't let it happen. Set this devil free. Help me, one
of you, with the ropes——"
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!
ebookultra.com

Download ebooks file Nonclassical Linear Volterra Equations of the First Kind A. S. Apartsyn all chapters

Uploaded by

Copyright:

Available Formats

Download ebooks file Nonclassical Linear Volterra Equations of the First Kind A. S. Apartsyn all chapters

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Download ebooks file Nonclassical Linear Volterra Equations of the First Kind A. S. Apartsyn all chapters

Uploaded by

Copyright:

Available Formats

Visit https://ebookultra.

com to download the full version and

Nonclassical Linear Volterra Equations of the First

_____ Click the link below to download _____

Explore and download more ebooks at ebookultra.com

Differential Equations Linear Nonlinear Ordinary Partial

Optimal Solution of Nonlinear Equations First Edition

A first course in the numerical analysis of differential

A first course in integral equations 2ed. Edition Wazwaz

A First Course in Linear Algebra With Concurrent Examples

Linear Partial Differential Equations for Scientists and

Decomposition Analysis Method in Linear and Nonlinear

Linear Integrated Circuits 2nd Edition S. Salivahanan

First published in 2003

Printed in The Netherlands by Ridderprint bv, Ridderkerk.

Chapter 1. Classical Volterra equations of the first kind 5

Chapter 2. Volterra equations of the first kind with two

2.7. A numerical solution of the test example 79

Chapter 3. Volterra equations of the first kind with two

Year 1996 marked the hundredth anniversary of the publications by Vito

{ ) de = /(*), te[io,n (o.i)

equation (0.1) will be called classical (or standard) in view of numerous

To gain a better understanding of the subject, it is recommended to

/'(*), ί€[ί 0 ,Τ]; (0.5)

the solution of (0.2) with the condition (0.3) is

presents no significant problem. Systems of equations of the type (0.2) for

1.1. CLASSIFICATION OF INTEGRAL VOLTERRA

It is well known that if the kernel K(t, s] is continuously differentiable

considered on the pair ((7, C) is well-posed as A""1 = E and φ(ί) = f ( t ) .

By virtue of (1.1.4), the kernel of equation (1.1.3) has a discontinuity of

Υφ = f K(t,s)<p(8)ds = f ( t ) , t 6 [ίο, (1-1-8)

It is easy to see that if

K't(t, s) Ε ΟΔ, Δ = {t, s/t0 < s < t < T}, (1.1.9)

problem of solving (1.1.8) is well-posed on the pair (C, C ).

(E - ψ)φ = ψ(ί] + Γ ^^-φ^) ds = D/, (1.1.13)

II V-1 || o ( i ) <oo. (1.1.16)

As follows from (1.1.7),

where Β = (Ε — W}~1 is bounded in C and D is a (linear) first-order

Definition 1.1.1. Integral equation (1.1.1) belongs to the type

Definition 1.1.2. If integral equation (1.1.1) belongs to the type

According to Definition 1.1.1, equation (1.1.2) belongs to the type V(0)

min \K$(t,s}\s=t\=kp^Q. (1.1.7')

corresponds to the problem of single differentiation.

to which the simplest problem of computer tomography is reduced when

Consider the equation

for ρ as large as desired, which contradicts (1.1.7'), it follows that (1.1.24)

1.2. THE GRONWALL-BELLMAN LEMMA

Lemma of Gronwall—Bellman. If functions g(t) andu(t) are nonneg-

Proof. Prom inequality (1.2.1) we have

Integrating (1.2.3) from ίο to i, we obtain

l n (v c + / g(s}u(s)ds] - l n c < / g ( s ] d s . (1.2.4)

u(t) <c + m i u(s}ds, t>tQ (1.2.1')

Let us describe the application of the Gronwall-Bellman Lemma for

Making (1.2.6) stronger, by virtue of (1.1.7) we have

Then, applying the Gronwall-Bellman Lemma to (1.2.1'), (1.2.2') with c =

and (1.2.12) implies

< (I/k)eKlk~1F-i°l (1.2.13)

Using (1.1.7) and (1.2.7), we successively obtain

!«(«)! < II «W !<%..„;

Since ψΐ = Wl(po, from (1.2.15) we obtain the estimate

_ Click the link below to download _