Operational Calculus - A Theory of Hyperfunctions - Yosida

Applied Mathematical Sciences I Volume 55
Applied Mathematical Sciences
1. John: Partial Differential Equations, 4th ed.

2. Sirovich: Techniques of Asymptotic Analysis.
3. Hale: Theory of Functional Differential Equations, 2nd ed.
4. Percus: Combinatorial Methods.
5. von Mises/Friedrichs: Fluid Dynamics.
6. Freiberger/Grenander: A Short Course in Computational Probability and Statistics.
7. Pipkin: Lectures on Viscoelasticity Theory.
8. Giacaglia: Perturbation Methods in Non-Linear Systems.
9. Friedrichs: Spectral Theory of Operators in Hilbert Space.
10. Stroud: Numerical Quadrature and Solution of Ordinary Differential Equations.
11. Wolovich: Linear Multivariable Systems.
12. Berkovitz: Optimal Control Theory.
13. Bluman/Cole: Similarity Methods for Differential Equations.
14. Yoshizawa: Stability Theory and the Existence of Periodic Solutions and Almost
Periodic Solutions.
15. Braun: Differential Equations and Their Applications, 3rd ed.
16. Lefschetz: Applications of Algebraic Topology.
17. Collatz/Wetterling: Optimization Problems.
18. Grenander: Pattern Synthesis: Lectures in Pattern Theory, Vol I.
19. Marsden/McCracken: The Hopf Bifurcation and its Applications.
20. Driver: Ordinary and Delay Differential Equations.
21. Courant/Friedrichs: Supersonic Flow and Shock Waves.
22. Rouche/Habets/Laloy: Stability Theory by Liapunov's Direct Method.
23. Lamperti: Stochastic Processes: A Survey of the Mathematical Theory.
24. Grenander: Pattern Analysis: Lectures in Pattern Theory, Vol. II.
25. Davies: Integral Transforms and Their Applications.
26. Kushner/Clark: Stochastic Approximation Methods for Constrained and
Unconstrained Systems.
27. de Boor: A Practical Guide to Splines.
28. Keilson: Markov Chain Models-Rarity and Exponentiality.
29. de Veubeke: A Course in Elasticity.
30. Sniatycki: Geometric Quantization and Quantum Mechanics.
31. Reid: Sturmian Theory for Ordinary Differential Equations.
32. Meis/Markowitz: Numerical Solution of Partial Differential Equations.
33. Grenander: Regular Structures: Lectures in Pattern Theory, Vol. III.
34. Kevorkian/Cole: Perturbation Methods in Applied Mathematics.
35. Carr: Applications of Centre Manifold Theory.
(continued on inside back cover)

K. Yosida
Operational Calculus
A Theory of Hyperfunctions
Springer Science+Business Media, LLC

K. Yosida
3-24-4 Kajiwara
Kamakura 247
Japan
AMS Classification: 33A99, 44A40, 44A45
Library of Congress Cataloging in Publication Data

Yoshida Kosaku
Operational calculus.
(Applied mathematical sciences ; v. 55)
Bibliography: p.
Includes index.
1. Calculus, Operational. 2. Hyperfunctions. 1. Tîtle.
n. Series: Applied mathematical sciences (Springer-Verlag New York Inc.) ; v. 55.
QA432.Y6713 1984 515.7'2 84-10614
Tîtle ofthe original Japanese edition: Enzanshi-oo, Vniversity of

Tokyo Press, Tokyo, © 1982 by Kosaku Yosida.
© 1984 by Springer Science+Business Media New York

Originally published by Springer-Verlag New York Inc. in 1984
AlI rights reserved. No part of this book may be translated or
reproduced in any form without written permission from Springer-Verlag,
175 Fifth Avenue, New York, New York, 10010, V.S.A.
9 8 7 6 543 2
ISBN 978-0-387-96047-0 ISBN 978-1-4612-1118-1 (eBook)

DOI 10.l007/978-1-4612-1118-1
Preface
In the end of the last century, Oliver Heaviside inaugurated an

operational calculus in connection with his researches in electromagnetic
theory. In his operational calculus, the operator of differentiation was
denoted by the symbol "p". The explanation of this operator p as given
by him was difficult to understand and to use, and the range of the valid-
ity of his calculus remains unclear still now, although it was widely
noticed that his calculus gives correct results in general.
In the 1930s, Gustav Doetsch and many other mathematicians began to
strive for the mathematical foundation of Heaviside's operational calculus
by virtue of the Laplace transform
( e -pt f(t)dt.
However, the use of such integrals naturally confronts restrictions con-

cerning the growth behavior of the numerical function f(t) as t ~ ~.
At about the midcentury, Jan Mikusinski invented the theory of con-
volution quotients, based upon the Titchmarsh convolution theorem:
If f(t) and get) are continuous functions defined on [O,~)
such that the convolution f~ f(t-u)g(u)du = 0, then either

f(t) =0 or get) =0 must hold.
The convolution quotients include the operator of differentiation "s" and
related operators. Mikusinski's operational calculus gives a satisfactory
basis of Heaviside's operational calculus; it can be applied successfully
to linear ordinary differential equations with constant coefficients as
well as to the telegraph equation which includes both the wave and heat equa-
tions with constant coefficients.
v
vi Preface
The aim of the present book is to give a simplified exposition as well

as an extension of Mikusinski's operational calculus.
As for the simplification, I should like to mention two points 10
and 20 below.
0
1 • We give a plain proof of the Titchmarsh convolution theorem by
making use of the well-known Liouville Theorem in analytic function theory.
0
2 • For solving linear ordinary differential equations with constant
coefficients, we need not rely upon the Titchmarsh convolution theorem.
We need only a rather trivial theorem:
Let f(t) be continuous for o< t < GO

t
and f f(u)du - O.
O
Then f(t):: O.
As for the extension, I should like to mention the following point 30 •

0
3 • We define the general power (s-a)Y of the operator (s-a) (a
and yare complex numbers), by making use of the general binomial theorem
in analytic function theory:
GO
(l-z)Y L (~) (_z)k (convergent for Izi < 1).

k=O
Then, by virtue of the general power (s-a)Y, we can solve algebraically

the so-called Laplace's differential equation:
where both the a's and b's are complex numbers and a # O. This equa-
2
tion includes Bessel's, Laguerre's and confluent hypergeometric differen-
tial equations and the like. It is to be noted here that Henri Poincare
and Emile Picard inaugurated the treatment of such differential equations
by the Laplace transform combined with subtle contour integrations in the
complex plane.
0 0
The present book is a revised and enlarged (by 1 and 3 ) English
Edition of the author's book Operational Calculus, written in Japanese and
published by the University of Tokyo Press (1982). The English transla-
tion was done by the author.
The author gratefully acknowledges fine help from many friends;
Hiroshi Fujita, Heinz Gotze, Shuichi Okamoto, Shigetake Matsuura, KyUya
Masuda and Jan Mikusinski. Fujita kindly invited the author to write the
Japanese Edition for the U.T.P. Gotze of the Springer-Verlag kindly sug-
gested the English edition. The above mentioned 10 and 20 were obtained
Preface vii
by the author jointly with Matsuura and Okamoto, respectively. Masuda

0
was a fine critic on 3 during its maturity. Mikusinski's fine work
aroused the author's interest towards the operational calculus. To them
all, including the U.T.P. and the Springer-Verlag, Inc., I express my
sincere thanks.
Kosaku Yosida
Kamakura
July 1983
Contents
Page
Preface v
Part I. INTEGRATION OPERATOR h AND DIFFERENTIATION OPERATOR s

(CLASSES OF HYPERFUNCTIONS: C AND C )
H
Chapter I. INTRODUCTION OF THE OPERATOR h THROUGH THE
CONVOLUTION RING C
1. Convolution Ring 1
2. Operator of Integration h 3
Chapter II. INTRODUCTION OF THE OPERATOR s THROUGH THE RING CH 5

3. The Ring CH and the Identity Operator I = h/h 5
4. CH as a Class of Generalized Functions of
~yperfunctions 8
s. Operator of Differentiation s and Operator
of Scalar Multiplication [a) 9
6. The Theorem :-h,- = eat 12
S-taJ
Chapter III. LINEAR ORDINARY DIFFERENTIAL EQUATIONS WITH

CONSTANT COEFFICIENTS 14
7. The Conversion of the Initial Value Problem

for the Differential Equation into a
Hyperfunction Equation 14
8. The Polynomial Ring of Polynomials in s
has no Zero Factors 15
9. The Partial Fraction Decomposition of a
Rational Function of s 18
10. Hyperfunction Solution of the Ordinary
Differential Equation (The Operational
Calculus) 22
11. Boundary Value Problems for Ordinary
Differential Equations 29
Chapter IV. FRACTIONAL POWERS OF HYPERFUNCTIONS h, s AND __1__ 32

s-a
12. Euler's Integrals - The Gamma Function and
Beta Function 32
13. Fractional Powers of h, of (s-a) -1 ,and
of (s-a) 34
Chapter V. HYPERFUNCTIONS REPRESENTED BY INFINITE POWER

SERIES IN h 39
14. The Binomial Theorem 39

15. Bessel's Function In(t) 40
viii
Table o{ Contents ix
Page
CHAPTER V. (cont.)
16. Hyperfunctions Represented by Power Series
in h 42
PART II. LINEAR ORDINARY DIFFERENTIAL EQUATIONS WITH LINEAR

COEFFICIENTS (THE CLASS C/C OF HYPERFUNCTIONS) 47
CHAPTER VI. THE TITCHMARSH CONVOLUTION THEOREM AND THE

CLASS C/C 47
17. Proof of the Titchmarsh Convolution Theorem 47

18. The Class C/C of Hyperfunctions 50
CHAPTER VII. THE ALGEBRAIC DERIVATIVE APPLIED TO LAPLACE'S

DIFFERENTIAL EQUATION 53
19. The Algebraic Derivative 53
20. Laplace's Differential Equation 60
21. Supplements. I: Weierstrass' Polynomial
Approximation Theorem. II: Mikusinski's
Theorem of Moments 70
PART III. SHIFT OPERATOR exp(-As) AND DIFFUSION OPERATOR exp(_As l / 2) 74
CHAPTER VIII. EXPONENTIAL HYPERFUNCTIONS exp(-As) AND exp(_As l / 2) 74
22. Shift Operator exp(-As) = e -AS .

Function Space K = K[O,oo) 74
23. Hyperfunction-Valued Function f(A) and
Generalized Derivative d~ f(A) = f'(A) 79
AS
24. Exponential Hyperfunction exp(As) = e 83
25. Examples of Generalized Limit. Power Series
in e AS 86
26. ro e-AsfCA)dA = {f(t)} For {f(t)} 1/2
E C 92
_As l / 2
27. Exponential Hyperfunction exp(-As ) e 94
28. Logarithmic Hyperfunction wand Exponential
Hyperfunction exp(Aw) 99
PART IV. APPLICATIONS TO PARTIAL DIFFERENTIAL EQUATIONS 106
CHAPTER IX. ONE DIMENSIONAL WAVE EQUATION 108

29. Hyperfunction Equation of the Form
2
fllCA) - w f(A) = gCA), w E C/C 108
30. The Vibration of a String 113
31. D'Alembert's Method 118
32. The Vibration of an Infinitely Long String 122
CHAPTER X. TELEGRAPH EQUATION 124
33. The Hyperfunction Equation of the Telegraph

Equation 124
34. A Cable With Infinitely Small Loss 125
x Table of Contents
Page
CHAPTER X. (cont.)
35. Conductance Without Deformation 126
36. The Thomson Cable 128
37. Concrete Representations of exp(-A~) 132
38. A Cable Without Self-Induction 138
39. A Cable Without Leak-Conductance 140
40. The Case Where All the Four Parameters Are
Positive 143
CHAPTER XI. HEAT EQUATION 145
41. The Temperature of a Heat-Conducting Bar 145

42. An Infinitely Long Bar 147
43. A Bar Without an Outgoing Flow of Heat 149
44. The Temperature in a Bar with a Given Initial
Temperature 150
45. A Heat-Conducting Ring 153
46. Non-Insulated Heat Conduction 156
ANSWERS TO EXERCISES 157

FORMULAS AND TABLES 160
REFERENCES 166
PROPOSITIONS AND THEOREMS IN SECTIONS 167
INDEX 169
Part I
Integration Operator hand
Differentiation Operator s
(Classes of Hyperfunctions: C and CH )
Chapter I
Introduction of the Operator h
Through the Convolution Ring C
§l. CONVOLUTION RING
The totality of complex-valued continuous functions a(t), b(t), f(t)

and so forth defined on the interval [0,=) will playa particularly
important role in the operational calculus; we shall denote the class of
those functions by C[O,=) or simply by the letter C. The convolution
of two functions a = aCt) and b = bet) of C is defined by
(a*b)(t) = a*b(t) = I: a(t-u)b(u)du (0 ~ t < 00) , (1.1)
and we have
PROPOSITION 1. a*b belongs to C; i.e., a*b(t) is a continuous function

defined on [0,00).
PROOF: Let 0 be any positive number. We shall prove
lim a*b(t+o) = a*b(t).

0+0
In fact, we have
t+o It ft+o
o a(t+o-u)b(u)du = 0 a(t+o-u)b(u)du + t a(t+o-u)b(u)du
f
For fixed t > 0, it is clear that
lim II 0 = ft a(t-u)b(u)du = a*b(u).

0-'+0' 0
1
2 I. INTRODUCTION OF THE OPERATOR h THROUGH THE CONVOLUTION RING C
Also, for fixed t > 0 and 0 < 0 < 1, we have
t+ o
IIz,ol ~ t
J ABdu ~ o·AB, where
A= max la(s)l, B= max Ib(u)l.

O<s<t+l
== t~u~t+l
This proves lim IZ,o = 0, and so we have obtained

0++0
lim a*b(t+o) a*b(t) (for t > 0).
0++0
Similarly, we can prove lim a*b(t-o) a*b(t) for fixed t > O.

0++0
In the following, we shall denote a function in C by {f(t)} or
simply by f, while f(t) means the value at t of this function f.
And for the convolution a*b of two functions a and b of C, we
simply write ab so that
abet) = J: a(t-u)b(u)du (0 ~ t < GO). (1.1)'
For a,b € C, the sum a+b is defined by
(a+b)(t) aCt) + bet) (1. 2)

l
and (a+b) € C. For any a € C (the field of complex numbers) and
a € C, the scalar mUltiple aa is defined by
(aa)(t) = aa(t). (1. 3)
We have the following:
THEOREM 1.
a+b = b+a, (a+b)+c = a+(b+c), (1. 4)
a(b+c) = ab+ac, a(a+b) = aa + ab, (a+B)a = aa + Ba, (1.5)
(Greek letters denote complex numbers.)
a(ab) = (aa)b = a (ab) , (aB)a = a(Ba), (1. 6)
ab = ba, a(bc) = (ab)c, (a+b)c = ac + bc. (1. 7)
PROOF: We shall prove (1.7) only, since the rest are easy. By the sub-
stitution t-u = r, we obtain
(ab)(t) = J: a(t-u)b(u)du = _fO a(r)b(t-r)dr

I:
t
= b(t-r)a(r)dr = (ba) (t) .

2. Operator of Integration h 3
Next, by using a(bc) = (bc)a and changing the order of integration

of the iterated integral, we obtain
((a(bc))(t) ((bc)a)(t) = f:[f:- U

b(t-u-V)C(V)dv]a(U)dU
f:[f:- V
a(U)b(t-V-U)dU]C(V)dV = ((ab)c)(t).
REMARK 1.1. The content of Theorem 1 is expressed in the words of alge-

bra as follows: C is a ring with respect to addition and the multiplica-
tion (or the product) ab, and the multiplication is commutative, i.e.,
ab = ba. The zero of this ring C is the function f € C with f(t) = 0
on [O,~); we denote it by 0 and
f+O = f and fO = O. (1. 8)
EXAMPLES OF CONVOLUTION
EXAMPLE 1.1. For aCt) t and bet) = e t ,
abet)
t
tee -1)
-t + e t - 1.
EXAMPLE 1.2. For aCt) = 1 and bet)
{1}{sin 2t} = fo sin 2udu = -sin U cos UIU=t

t
u=o
+ f: 2
cos udu -sin t cos t + t - f: 2
sin udu.
= 21 1.
t - 2 Sln t cos t.
§2. OPERATOR OF INTEGRATION h
We shall denote the function {I} by h. Then, as in Example 1.2,

we have
(hf)(t) = f: f(u)du (f € C). (2.1)
so that the convolution product hf is equal to the indefinite integral

4 I. INTRODUCTION OF THE OPERATOR h THROUGH THE CONVOLUTION RING C
of f from 0 to t. Hence h represents the operator of integration.
REMARK 2.1. The introduction of the letter "h" is due to the historical
fact that Oliver Heaviside (1850-1925), the founder of the operational
calculus, used
H(t) .{ 1/2
1
o
t > 0
t
t
0
< O.
(2.2)
for the "unit function" in his calculus. Jan Mikusinski used the letter
"!" for 1 in his mathematical foundation of Heaviside's operational calcu-
lus. However, because of typographical reasons, the present book follows
the use of "h" due to Erdelyi [3].
PROPOSITION 2.
and by induction,
(n = 1,2, ... ). (2.3)
EXAMPLES. Verify the following equalities:
(eL) {t}2 {t}{t} = {61 3

t } =
224
(h) = h ,
2 6
(B) {t}3 {t}{t}2 = {1;0 t S} = (h )3 = h ,
(y) {e t }2 {et}{e t } = {te t },
(0) {e t }3 {e t He t }2 =
1 2 t
{-zy t e }.
EXERCISE 2. Simplify the following expressions:
(y) {l_t}{e t } + {e-t}{l+t} _ {l_t}{e- t } _ {et}{l+t},
(0) {1_t l / 2}{sin t} + {1+tl/ 2}{cos t} + {1_tl/ 2}{cos t} + {1+t l / 2}{sin t}.
Chapter II
Introduction of the Operator 5
Through the Ring CH
§3. THE RING CH AND THE IDENTITY OPERATOR I =~
PROPOSITION 3. Let
H = h n (n = 1,2, ... )}.

= {k: k
n
Then, for any k = h E Hand f E C, we have
kf = 0 implies f = 0, where 0 = {a} E C. (3.1)

PROOF: The equation
hf(t) = f: f(u)du =0 (0 ~ t < 00)
implies, by differentiation with respect to t, that
:t (hf) (t) = f(t) = 0 (0 ~ t < 00),
i.e., f = O. Hence
2
h f = h(hf) = 0 implies hf =0 and so f = O.
Therefore, by induction on n, we obtain (3.1).
As a corollary of Proposition 3, we obtain
PROPOSITION 4. Let k,k' E Hand f,f' E C. Then the following relation

:: of two "fractions" f and ~:, given by
f f'
k:: k' if and only if k'f = kf', (3.2)
is an equivalence relation. That is, we have
5
6 II. INTRODUCTION OF THE OPERATOR 5 THROUGH THE RING C
H
f f
k'" P (3.2) 1
f f'
implies f' f
k'" j(T j(T"'P (3.2)2
f f' f' f". . f f"

k::; j(T' j(T ::: kiT Imphes k '" kiT (3.2) 3
PROOF: (3.2)1 is evident from kf = kf. (3.2)2 is also evident since

k'f = kf' implies kf' = k'f and vice versa.
PROOF OF (3.2) 3: We have k' f = kf' and k"f' = k' f" by assumption.
Hence k"k'f = k"kf' and kk"f' = kk'f" and so, by (1.7), we have
k'k"f = k"k'f = k"kf' = kk'f" = k'kf".
Thus
k' (k"f - kf") = 0 which implies k"f - kf" = 0

by Proposition 3. The last equation means
REMARK 3.1. Since the equivalence relation holds, we can replace the above
'" by the equality so that
f f- = i f and only i f k'f = kf'. (3.2) ,

f
1HEOREM 2. We can define the sum and product of two "fractions" - l and
k
f l
2
- as follows.
k
2
(3.3)
(3.4)
PROOF: We have to show the following:
hold. That is, we have to prove that
follow from
This can be shown as in the case of ordinary numerical fractions.

3. The Ring CH and the Identity Operator I = hh 7
COROLLARY. By (3.3) and (3.4), the set
f € C and k € H} * (3.5)
is a ring whose "product" is commutative. Namely, the following (3.5)1-

(3.5)4 hold:
J
f f' f' f
k+j(T=j(T+P
lf + (~: + ~) = (f + ~:) + ~:;
f af 8f
(a+I!)j( = T + k
f fl _ f' f
K j(T - 1(T P
fff'
K\j(T k"
£:.) -_(!k j(T)K""
f'\fI'
,
{ f fl) fI' f f" fl fI'

( k + j(T j(iI = k j(iT + j(T fiT .
The proof is obtained as in the case of numerical fractions.
THE IDENTITY OPERATOR I. By (3.2)', we have
(k € H). (3.6)
n
h
Therefore, is the multiplicative unit of the ring C . Hence we
n H
h
n
h
shall denote --
n by or simply by 1 and call the uni t (or
h
identity) operator:
* Yosida, K. and Okamoto, S.: A note on Mikusinski's operational calculus,
Proc. Jap Acad., 56, Ser. A, No.1 (1980), 1-3.
8 II. INTRODUCTION OF THE OPERATOR 5 THROUGH THE RING C
H
(n 1,2, ... ). (3.7)
§4. CH AS A CLASS OF GENERALIZED FUNCTIONS OR HYPERFUNCTIONS
The ring C is a 5ubring of the ring C ' since we can identify

hnf . H
f = {f(t)} € C with --- € C , I.e.,
h
n H
hnf
f=--- (n=I,2, ... ). (4. I)
n
h
In fact, by (3.3)-(3.4), we have
and
hnf
Therefore, C contains ~ = f, which is a continuous function.
H
However, C is actually bigger than C as the following proposition shows.
H
PROPOSITION S. The unit operator I does not belong to C.
h
PROOF: We shall derive an absurdity from I = h = f where f € C. In
h hf . 2 2
fact, from h = 11 ' we obtaIn h = h f, so that
{t} = {t}f = f:(t-U)f(U)dU.
Hence, by differentiation, we obtain
{I} = {(t-t)f(t) + f: f(U)dU} = {f: f(U)dU}.
Thus, by further differentiation, we have
o= {ret)}, Le., f(t) = 0 for all t > o.

2 2 2
This yields a contradiction: {t} = h = h f = h 0 = O.
GENERALIZED FUNCTIONS OR HYPERFUNCTIONS. The equality I = ~ shows

that from a continuous function h € C we can obtain an element ~ € CH
which is not a continuous function. Hence I is a generalized function
5. Operator of Differentiation 5 and Scalar Multiplication [a] 9
* a hyperfunction ** ,obtained by the process of "dividing"

or, preferably,
elements of C by h. We shall further obtain, in §5, another hyperfunc-
tion s which is the operator of differentiation.
§5. OPERATOR OF DIFFERENTIATION s AND OPERATOR OF SCALAR MULTIPLICATION [a]

THEOREM 3. We put
O
(n 1,2, ... ; h I). (5.1)
s is the operator of differentiation in the following sense: If both f

and its derivative f' belong to C, then
f' = sf - [f(O)], where [f(O)] = s{f(O)} = {f 10)} , (5.2)
so that both sf and [f(O)] belong to CH.

PROOF: We have
sh = hs = h lh = I = 1. (5.3)
Hence, by the fundamental theorem of calculus, we have
hf' = {(hf')(t)} = {f: f' (U)dU} = {f(t) - f(O)}

(5.3) ,
{f(t)} - {f(O)} {f(t)} - h[f(O)],
and so, by (5.3),
shf' = (sh)f' = f' = sf - sh[f(O)] = sf - [f(O)].

THEOREM 4. We denote, for any complex number a,
(5.4)
Then [a] is the operator of scalar multiplication in the following sense:
f
For any --
n
€ C , we have
H (5.5)
h
*The adjective "generalized" is somewhat lengthy.

** Although this term "hyperfunction" has been used since 1959 by Professor
Mikio Sato for his generalization of the notion of functions, there should
be no serious confusion with my use due to the fact that Sato's generaliza-
tion is based upon the (generalized) boundary value of analytic functions.
10 II. INTRODUCTION OF THE OPERATOR s THROUGH THE RING C
. H
PROOF:
This means that [a] acts as a scalar multiplication by a in the ring

Cw
REMARK 5.1. Defining SO = I. we have
C =
H
{...!.n : f € C, (n = O.l .... )} = {snf : f € C. (n O.l .... )}.
h
COROLLARY OF THEOREM 3. If f.f'.~' •...• f(n) € C. then
fen) = sn f - sn-l [f(O)] _ sn-2 [fl (0)) _ ... _ s [f(n-2) (0)]

(5.6)
_ [f(n-I HO )].
PROOF: By (5.2). we have

~, sf ' - [fl (0)] s (sf - [f(O)]) - [fl (0)]
s2 f - s [f(O)] - [fl (0)) .
We have already established (5.6) for n = 1. Hence the general case of
(5.6) is proved by induction on n.
REMARK 5.2. When f € C = C[O.~) does not have its derivative ft be-
longing to C, we shall call sf = hf € C as the "generalized derivative"
of f. Thus. in this sense. f is infinitely differentiable in a generali-
zed sense so that the n-th generalized derivative of f is given by
Thus sn f is a hyperfunction (n 1.2 •... ). and
Further.
[a) = {a} = sial = aI

h
is also a hyperfunction. In the following. we shall write [a). [f(O)).

[f'(O)]. etc. simply by a. f(O). fl(O) etc .• in cases where no ambiguity
will result.
REMARK 5.3. Any polynomial in s given by
(5.7)
5. Operator o~ Di~~erentiation 5 and Scalar Multiplication [a] 11
belongs to CH, since CH is a ring and
(n = 0,1, ... ).
If
n-l
pes) = 0 n5n + an_Is +
r
(with n:: m),
then
n
a _
n l = ... = °m+l = 0 and
• (5.8)
a m = 13 m, a m-l 8 _ , ... ,a = 8 ,
mI O 0
PROOF: In the case n > m, we must have
and hence, by multiplying by hn , we obtain
n-m-l n-m n
= anI + (on_l h + + 0m+l h + (am-Bm)h + ... + (ao-Bo)h
anI + (On_l{l} + ... + Om+l{~:~:~~)!} + (am-Bm){~:~:~~)!}

+ ... + (ao-Bo){(~:~~!})
Therefore, by Proposition 5, we must have
anI = 0 and {a polynomial in t} = O.
Hence (5.8) is proved in the case n > m. The case n = m is easily

proved as above.
REMARK 5.4. In his book Operational Calculus, J. Mikusinski [5] introduced

the "fraction" of the form
f
g
(f,g € C and g F 0)
which is more general than our "fraction"
Mikusinski's standpoint was based upon the so-called Titchmarsh convolution

12 II. INTRODUCTION OF THE OPERATOR s THROUGH THE RING C
H
theorem which reads: If a.b € C be such that ab = O. then either

a =0 or b = O. By virtue of this theorem. we can define the "fraction"
or the convolution quotient
f
(f,g € C and g # 0)
g
similarly to our case of §3:
(f € C and n = 1,2 •... ).
Titchmarsh discovered the above theorem in 1926 by making use of deep

theorems on analytic functions. * An elementary proof was given in 1951
by Mikusiski jointly with C. Ryll-Nardzewski. ** It is elementary in the
sense that it does not appeal to analytic function theory, although it
takes about 9 pages. A plain proof *** will be given in §17 of the present
book by relying upon Liouville's theorem in analytic function theory. It
is to be noted here that before Chapter VI of the present book. we do not
need the Titchmarsh convolution theorem; we need only hyperfunctions of the
form
snf (n = 1.2, ... )

which were introduced by virtue of Proposition 3.
§6. THE THEOREM::-h:,.[ = {eCtt}

s- lCtJ
THEOREM 5. For any complex number ct, there exists one and only one solu-
tion f
n
of the equation
h
(s- [a))....!.
n
= (6.1)
h
.
glven by - f = {Ctt}
e € C.
n
h
PROOF: We have, by (5.2),
so that
*Edward Charles Titchmarsh (1899-1963): Introduction to the Theory of
Fourier Integrals, Oxford University Press (1948), p. 327-.
** J. Mikusinski [5], p. 15 - p. 23.
*** Kosaku Yosida-Shigetake Matsuura: A Note on Mikusinski's Proof of the
Titchmarsh Convolution Theorem, to be published in the Contemporary Mathe-
matics series of the Amer. Math. Society.
6.
I
The Theorem 5-[a] = {eat} 13
at
(s_[a]){e } [a]{eat } + [1] - [a]{e
at } [1] I,
(6.2)
{eat} (s- [a))
by the commutativity (3.5)4 of the product in the ring C .

H
The uniqueness of the solution ~ of (6.1) is proved as follows.
h
n
By multiplying on both sides of (6.1) by {eat}, we obtain
{eat } ((s- [a])-)

f = {e at }.
n
h
Hence, by the associativity (3.5)4 of the product in C ,

H
.
t hat 1S, --
f
n = {at}
e .
h
REMARK 6.1. Since (s-[a])-!
n
= (s-a) ~ = I has one and only one solution
h h
~ at
={e } in the ring C , we say that (s-a) admits the (multiplicative)
n H
h at
~nverse {e } in C . We shall denote this fact by
H
(s_[a))-l = ::-h
s-laJ
= _1_ = {eat}.
s-a
(6.3)
In the same way, the element {eat} € C ~ C has the (multiplicative)

H
inverse (s-[a]) = (s-a) € ~:
I
----- = (s-[a]) = (s-a). (6.4)
{eat}
n
COROLLARY. For any integer n > I, (s_a)n (s-[a]) € CH has the in-
verse in C given by
H
_1_=_...=.1_ _ tn-l at}

{ (n-l)! e (n = 1,2, ... ). (6.5)
(s_a)n (s_[a])n
PROOF: We have
and hence
1
---3 = s:a
1 1 at}{t at} -_ {Jt
-:---:2 = {e"..el u au du}
ea(t-u) yre
(s-a) (s-a) . 0 .
t 2
{eat J 0 friu} {2f
u
=
t at
e }
etc.
Chapter III
Linear Ordinary Differential Equations with
Constant Coefficients
§7. THE CONVERSION OF THE INITIAL VALUE PROBLEM FOR THE DIFFERENTIAL
EQUATION INTO A HYPERFUNCTION EQUATION
THEOREM 6. Let aO.al •...• a be complex numbers (an # 0) and f € C=

n
C[O.~). Consider the equation
together with the initial condition at t = 0:

_ (n-l)
YCO) = YO' y' (0) - Yl.···.y (0) = Yn - l · (7.2)
Then, (7.1)-(7.2) converts into the hyperfunction equation

n n-l
(ans + an_Is + ••• + als + aOI)y
n-1 n-2
=f + Bn_ls + Bn _2s + ••• + 6ls + BOI. (7.3)
(v = 0.1.2 •...• n-l)
by virtue of
y(k) = sky _ sk-ly(O) _ sk-2 y '(0) _ ... _ sy(k-2) (0) _ y(k-l)(O). (7.4)
where y(j)(O) = [y(j)(O)].

REMARK 7.1. The merit of the above converstion lies in the fact that the
initial condition (7.2) is absorbed in the equation (7.3). Since the terms
such as sky •s are hyperfunctions derived from the unknown function y € C.
we call (7.3) the hyperfunction equation.
14
8. The Polynomial Ring o~ Polynomials in s has No Zero Factors 15
EXAMPLE 7.1. The initial value problem
is converted into the hyperfunction equation
(7.5)
This is equivalent to
Multiplying by
i. e.,
THE SOLUTION OF THE HYPERFUNCTION EQUATION (7.3). Solving (7.3) for-

mally we obtain
y (7.6)
In the next section, we shall give the interpretation of the right hand
side of (7.6) as a function in C[O,~).
§8. THE POLYNOMIAL RING OF POLYNOMIALS IN s HAS NO ZERO FACTORS
Let
pes)
and
Q(s) = ~s
n
+ ~
m-l
s + ••• + ~Os
o (~ # 0)
m m-l m
be, respectively, polynomials of n-th and m-th degree.
PROPOSITION 6. The product P(s)Q(s) is a polynomial in s of (n+m)-th

degree.
PROOF: By the rules of the sum and product in the ring C , we obtain
H
16 III. LINEAR ORDINARY DIFFERENTIAL EQUATIONS WITH CONSTANT COEFFICIENTS
P(s)Q(s)
(8.1)
COROLLARY. Let pes) and Q(s) be polynomials in s. Then P(s)Q(s) 0*

implies that either pes) or Q(s) must be O.
REMARK 8.1. The totality of polynomials in s is a subring of the ring

CH. It is called the polynomial ring of s. The above Corollary is stated
as follows:
THEOREM 7. The polynomial ring of s has no zero factors. Precisely

the product of two non-zero polynomials pes) and Q(s) is not = O.
DEFINITION OF THE FRACTION OF TWO POLYNOMIALS IN s (THAT IS, THE

RATIONAL FUNCTIONS OF s).
PROPOSITION 7. Let, for two polynomials pes) and q(s), pes) F O. Then
we can define the "fraction" ~ by the rule of the equivalence rela-
pes)
tion ::::
(8.2)
because we can prove
~-~
pes) - pes) ,
~ _ ql (s) ql (s) _ n(s)
implies ------ - ~
pes) - PI (s) PI (s) p(s)'
~ _ ql (s) ql (s) q2(s)
pes) -Pl(s)' PI (s) ::: P2(s) implies
PROOF: We shall give the proof of the last relation (the transitive rule).
By assumption, we have
q(s)Pl (s) = P(s)ql (s),
ql (s)P2(s) = PI (s)q2(s).
Multiplying the first formula by P2(s), we obtain
o means that a n O. See Remark 5.3.

8. The Polynomial Ring of Polynomials in s Has No Zero Factors 17
Multiplying the second formula by pes), we obtain
Hence
q(s)PI (s)pz(s) = P(s)PI (s)qz(s), Le.,
PI(s)(q(s)PZ(S) - p(s)qZ(s)) = o.
Since PI(s) F0 as the denominator of the "fraction", we must have
q(s)pz(s) - p(s)qz(s) =0 (by Theorem 7).
Therefore
We can now replace ~ by the equality sign so that
~ = ql(s)
pes) PI(s) means q(s)PI(s) = P(s)ql(s). (8. Z) I
Therefore, as in the case of Theorem Z, we can prove
THEOREM 8. The sum and the product of the "fractions" above can be de-
fined by
~ + ql (5) = q(s)P I (s)+P(s)ql (5)
(8.3)
pes) PI (5) p(s)P I (5)
Moreover, the totality of the fractions is a ring with respect to the

above sum and product, and the product is commutative. We can also prove,
for scalar multiplication by complex numbers,
(~ ql (5)) _ aq(s) aql (5)

a pes) + PI (5) - pes) + ~ ,
(a+B)s.i&
pes)
= aq(s) + ~
pes) p(s)'
a(~ ql(S)) = aq(s)ql(s)
pes) PI (5) P(s)PI (5) .
REMARK 8.Z. We call the fraction ~~:~ (p(s) F 0) a rational function of

s. In particular, if pes) = sO = I, we shall identify ~ with q(s)
5
so that a polynomial q(s) may be considered as a rational function of s.
Therefore, we can use the following rule:
s ~ _ ql (s)q(s) ql(s)q(s)
ql( ) pes) - 0 pes)
s pes)
and so
pes) ~ = p(s)q(s) = q(s).

pes) pes)
§9. THE PARTIAL FRACTION DECOMPOSITION OF A RATIONAL FUNCTION OF s

PROPOSITION 8. Given a polynomial in s of n-th degree,
(lln + 0), (9.1)
we shall consider a polynomial in the complex variable z of n-th degree,
(9.2)
By a theorem of Carl Friedrich Gauss (1777-1855), p(z) admits a factori-

zation
k
L mi = n, (9.3)
i=l
where zl,z2, ... ,zk are distinct roots of the algebraic equation p(z) = O.
Then, by the relations between the roots zl,z2"" ,zk and the coefficients
lln,lln_l""'Ct o' we see that
m m
l 2 (9.3) ,
pes) = lln(s-zl) (s-z2)
THEOREM 9. Besides pes), consider another polynomial q(s) in s of

m-th degree with m < n. Then, we have the partial fraction decomposition
of ~.
pes) .
k m. C. I
~=
p (s) L P JU
(s_z.)u
(9.4)
j =1 u=l
J
with uniquely determined complex numbers C. 's. Therefore, we have, by
JU
(6.5),
k m. {tU-l z.t}
~=
pes) L P CJ' i (u-l)! e J (9.4) I
j=l j=l
which is an infinitely differentiable function of t > O.
PROOF: By (9.3)', we have

9. The Partial Fraction Decomposition of a Rational Function of s 19
where f(s) is a polynomial in 5 of (n-ml)-th degree. By (9.3), fez)

does not have (z-zl) as a factor so that f(zl) # 0 by the remainder
theorem. Hence, the complex number
is well-defined. Thus, if we write
~ = _..:Iq~(s~)_ _
pes) ml
(s-Zl) f(s)
then
so that, by the remainder theorem,
Since both q(z) and fez) are of degree < n, the degrees in s of
m
l
(s-zl) f(s) and f (s) are respectively nand < n-l. Hence, if
l
m - 1 > 0, we have
l
(s-Zl)fl (s)
m m -1 '
l
(s-zl) f(s) (s-zl) 1 f(s)
ml-l
where the degrees in s of (s-zl) f(s) and f 1 (s) are respectively
(n-l) and < (n-l).
Therefore, we are able to start with
and argue as above again. Repeating the same process, we can finally ob-
tain (9.4).
REMARK 9.1. The uniqueness of the Cjk's of (9.4) is proved as follows.

m
Multiplying l
(z-zl) by
k m. C.
~= LJ
L u=l JU
P (z) j=l (Z_Zj)u
we obtain
(9.S)
Next we obtain
(9.6)
C1,m - 2' (9.7)

1
and so forth.
EXAMPLE 9.1. Obtain a, b, and c in
ANSWER:
2
a = lim z (z+e) !
z~
2
z (z+a)
a
b lim z z+e e l' az+ae-e(z+a) _ a-e

- -2 - -2'
z+1~ az(z+a)
2
z+O z (z+a) az a
_ e-a
c lim (z+a) (z+e) - - 2 '
2
z+-a z (z+a) a
EXAMPLE 9.2. Obtain the partial fraction decomposition of
54 _ 5
-3- - 5 + 3"":""" .
5 -1 5 -1
ANSWER: We have
z abc il3 2 il3
-3- = z:T + z:w + --2 ' w=
-1 +
2
w
-1 -
2
z -1 z-w
and so
a = lim (z;1) z = lim -2-z- - =

z+l (z -1) z+l z +z+l
t,
z w2
b lim (z-w)z
3 3 lim 2 2 = T'
z+w (z _w ) z+w z +wz+w
2 z
(z-w ) z w
c = lim 2 lim
2 2 2 4 = 3'
z+w (z3_(w 2)3) z+w z +w z+w
9. The Partial Fraction Decomposition of a Rational Function of s 21
EXAMPLE 9.3. Let a and B be real numbers. Prove (i) and (ii) below.
(i)
2 2 (B + 0).
(s-a) +B
(ii)
s-a {eat sin Bt}.
2 2
(s-a) +B
PROOF:
I
2 2 = 2iB
1 (I I)
s-a-iB - s-a+iB
(s-a) +6
= -! {e(a+i6)t _ e(a- i 6)t}
2i6
s-a _ Y_1_ + __
2 2 - 2\s-a-i6 s-a+i6
1_)
(s-a) +B
EXAMPLE 9.4. By (i) of Example 9.3, we have
222
{~ft ea(t-T)sin 6(t-T)e aT sin 6TdT}
[(s-a) +6 ] 6 0
{_ea~ ft(COS Bt _ cos 6(t-2T))dT}

2B 0
= {ea~
26
(i sin Bt - t cos Bt)}.
As a remark to Theorem 9, we have
THEOREM 9'. Let pes) and q(s) be polynomials in s of degree nand

m respectively, with n < m. Then
~ = w(s) + res) (9.8)

pes) pes) ,
where w(s) is a polynomial in s of degree m-n and res) is a poly-

nomial in s of degree ~ n-l.
PROOF: We omi t the proof.
COROLLARY. The polynomial w(s) is an element of C and the fraction

H
~
pes)
is, by (9.4)', an infinitely differentiable function in C.
EXAMPLE 9.5.
3
5 5 1 I I
-2- = 5 + -2- = 5 + I(s:! + 5+1)
5 -1 5 -1
=5 + {e t e -t}
+
2 •
EXERCISES FOR §9. Prove the following formulas (a and 6 are real num-
bers) .
l I t
1. 2 = ~-6te - 3e t - e- t + 4e 2t },
(5-1) (5+1) (5-2)
2.
I (.!.. eat sinh(6t)} (6 + 0),
2 2 6
(s-a) -6
3.
s-a {eat cosh (6t)}. *
2 2
(s-a) -6
§10. HYPERFUNCTION SOLUTION OF THE ORDINARY DIFFERENTIAL EQUATION

(THE OPERATIONAL CALCULUS)
The initial value problem of §7:
any(n) + an_ly(n-l) + ... + aly' + aoy = f, (10.1)
_ ,_ (n-l)_ (10.2)
yeO) - YO' Y (0) - Yl,···,y (0) - Yn - l ,
was converted into the hyperfunction equation
(10.3)
(\I = 0,1,2, ... ,n-l).
Let us denote
(10.4)
Then we have the factorization
(10.5)
* sinh t sinh(t)
10. Hyper!unction Solution o! the Ordinary Differential Equation 23
and so, by (9.4)1,
I Ie~ PC
m. {tU-l e z.t}
J (10.6)
pes) Z .L J·u (u-l)! '
J"'l u.. l
and
(10.7)
+ ••• +
Therefore, by multiplying both sides of (10.3) by P(~)' we obtain
_ _I_f+tl&
y - pes) pes)
(l0.8)
= r
j=l
UfJl·
-
[Cju{(~~~~! eZjt}f + dju{(~~~~! e
Zjt
}].
REMARK 10.1. If f e: C[O,oo), then y z Yet) given by (10.8) is in

C[O,oo). In fact, we can prove that this yet) is n-times continuously
differentiable in t. Therefore, this yet) is the unique solution of
(10.1) - (10.2).
PROOF OF THE n-TIMES CONTINUOUS DIFFERENTIABILITY OF (10.8): Multiplying

both sides of (10.3) by hn , we obtain
Here
is n-times continuously differentiable in t, and so, y given by
y = -an -1 (an- lhy + a n- 2h 2y + ... + aoh ny) + a n-1 {F(t) } (10.9)
is once continuously differentiable in t (because, for example,

is, by the continuity of yet), once continuously differentiable;
furthermore, by (5.3)',
3
(h y)' = h2y 2
'" h (hy' + yeO)) = h3y' 2
+ h yeO)
.. h3y' + (a polynomial in t).)

Therefore, by differentiating (10.9) once, we obtain
y'
(10.9) ,
-1
+ (a polynomial in t) + an F'(t).
Repeating the above reasoning on (10.9)', we can prove that y, given by

(10.8), is n-times continuously differentiable in t whenever
f = {f(t)} E: C.
Thus we are now able to state:
THE OPERATIONAL METHOD OF SOLVING (10.1)-(10.2). Remembering the

formula
y' = sy - [YeO)],
we easily extend it to
2
y" = sy' - [y'(O)] = s(sy - [yeO)]) - [y'(O)] s Y - sy(O) - y'(O),
and finally to
By virtue of this formula, we obtain (10.3) from the initial value problem
(10.1) - (10.2). Hence we have pes) ((10.4)) and res) ((10.7)). There-
fore, we obtain the solution y of (10.1)-(10.2):
I !1&
y = pes) f + pes) •
Then, by making use of the partial fraction decomposition of

I
and ill2.
pes) pes)
together with the formula
I
UI
{t - Zjt}
(s_z.)u
=---,e,
(u-l).
J
we obtain a concrete expression of the solution yet) of (10.1)-(10.2).

EXAMPLE 10.1. Solve y' + ay = f, yeO) = a. By (7.4), we have
sy - a + ay = f
so that
= { e -at It0 eau f(u)du + ae -at} •

10. Hyperfunction Solution of the Ordinary Differential Equation 25
2
REMARK 10.2. The solution of y' - y {(2t-l)e t }, yeO) 2 is given by
because
2
-u u
e e du
It is to be noted that the above equation is difficult to solve by the cus-

tomary method of the Laplace transform, because
,.. At t2
J e (2t-l)e dt
O
diverges for any A.
EXAMPLE 10.2. Solve y' + ay = 8, yeO) = a. Since 8 means
{8} = [8]h = l!l-=~,

s s
*
the corresponding hyperfunction equation is
sy - a + ay = 1!l
s
= ~s'
Hence
2t 1
EXAMPLE 10.3. Solve y" - 4y e yeO) = I, y'(O) = 4' The correspond-
ing hyperfunction equation is
2 1 1
s Y- s - 4- 4y = s:2 '
that is,
2 1 1 1 s+2
(s -4)y = s:2 + 4 + s = 4 s:2 + s.
Hence, by (10.9),
*As was stated in §S, we shall write 8/s for [8]/s in case
confusion will not occur. We also write 1 for 1.
1 1 s+2 s 1 1 s
Y = 4" S2_ 4 s:2 + S2_ 4 = 4" (S_2)2 + S2_ 4
= i<1 te 2t } + I(s:2
1 1 1
+ s+2) =
1
{4" te
2t
+ cosh 2t}.
EXAMPLE 10.4. Let a,B be real numbers. Solve the system of linear or-
dinary differential equations
y' - ay - Bz = Beat, ZI + By - az 0,
{ yeO) = 0, z(O) = 1.
The corresponding system of hyperfunction equations is
sy - ay - Bz 8
= s:Q' sz - 1 + By - az = O.
We solve this system, obtaining
2 2
2B
y = -"="::':"'2-""'2'
z = _~(s:;,. -. ;;;;aL.)_-~8--:::-
(2-a) +B (s-a) ((s-a) 2+ 82)
Hence, by (i)-(ii) of Example 9.3, we have
y = {2eat sin Bt}

2(s-a) 1
z = 2 2 --=
s-a
(s-a) +8
EXAMPLE 10.5. Solve y' - y = 2, yell 1. Putting y(t+l) z(t), we
obtain
z' - z = 2, z(O) = 1.
The corresponding hyperfunction equation is
sz - 1 - z = ~s'
so that, by (6.5), we obtain
1 2 1
t } + 2(- 1
z = s:T + (s-l)s = {e s-l - -)
s
= {e t + 2e t _ 2} = {3e t - 2}.
t-l
Hence yet) = z(t-l) = {3e - 2}.
2
EXAMPLE 10.6. Solve y" + 2ay' + b y f(t), yeO) = k ' y'(O) = k l • Here
O
a,b,kO,k are real constants with a > 0 and b > O. The corresponding
l
hyperfunction equation is
10. Hyperfunction Solution of the Ordinary Differential Equation 27
and so
sk +k +2ak
f O l O
y 2 2 +
s +2as+b s2+2as+b 2
For the sake of convenience, we introduce
Ii b
2
- a
2
Then, if w 1 0, we have
f (s+a)kO+(akO+k l )
y (s+a+iw)(s+a-iw) + (s+a+iw)(s+a-iw)
-1 1_1_.__ ~)(f+ak +k )
2!W\s+a+lw s+a-1W 0 1
O (1
+k- ~+---
1 )
2 s+a+1W s+a-iw
iwt -iwt
-at e -e
e 2iw
-at ft akO+k l -at. -at

{~ 0 eaTsin w(t-T)f(T)dT + --w--- e Sln wt + kOe cos
REMARK 10.3. In the case w 0, i.e., b 2 = a2, we obtain
(s+a)k
O
y + --~:
(s+a) 2
{ te -at}{ f(t) } + { (akO+kl)te -at + kOe -at} .
The equation of Example 10.6 represents the motion of a particle of

unit mass. The left side of the equation is the sum of the acceleration,
the damping force and the elastic force, and the right side of the equa-
tion is the external force acting on the particle.
EXERCISES FOR §10. Solve the given differential equations (systems).
1. y" + 3y t + 2y 2te- t , yeO) y' (0) O.
2. y" + 3y' + 2y 2te -t , yell y' (1) 0 (Hint: Put y(t+l) z (t)) .
2t
3. y'" + y' e yeO) = y' (0) y" (0) O.
4. y' + z' -z t
= e 2y' + z' + 2z = cos tj x(O) yeO) = o.
5. x' y-z, y' = x+y, z' = x+z; x(O) = 1, yeO) 2, z(O) = 3.
6. x'-x-2y = t, -2x+y'-y = t, x(O) 2, yeO) 4.

7. x" + 3y' - 4x + 6y = 10 cos t,
x' + y" - 2x + 4y = 0,
x(O) = yeO) • 0, x'(O) = 4, y'(O) = 2.
EXAMPLE 10.7. CONCRETE EXAMPLES (ELECTRIC CIRCUITS). A circuit of self-
inductance L and resistance R has applied an electromotive force E.
If i(t) denotes the current, we
have, by Kirchhoff's law,
Li'(t) + Ri(t) = E,
where E is assumed to be
E = {EO}' Assuming the initial
condition of the current i(t)
to be i(O) = 0, we obtain
EO
Lsi + Ri = s
Hence
s(Ls+R)i EO
and so
i I Eo (I I)
EO s(Ls+R) = II 5 - s+R/L '
that is, R
i(t) = EO {I _ e-(R/L)t}.
R
EXAMPLE 10.8. Consider a circuit of self-inductance L, resistance R

and of capacitance C. Let the
charge at the electric pole of
the condenser be Q. Then
Li'(t) + Ri(t) + ~ = E,
Q'(t) = i(t).
Assuming the initial conditions i(O) o and Q(O) 0, we have
Lsi + Ri + %= E, sQ = i.
Eliminating Q from these two equations, we obtain, taking E = {EO}'
11. Boundary Value Problem$ for Ordinary Differential Equations 29
[EOl
E = --.
s
Thus
2
L«s+a) + \.I)
Hence, by (6.5), (i) of Example 9.3, and 2 of Exercises in §9, we obtain
EO 1 -at. ~
- - e Sln >'\.It (\.I > 0)
L Iii
EO -at
i(t) T te (\.I = 0)
EO 1 -at.
T - e slnh I=irt (\.I < 0).
,.':jj
§ll. BOUNDARY VALUE PROBLEMS FOR ORDINARY DIFFERENTIAL EQUATIONS

PROBLEM 11.1. Let a # 0 be a real number, to > 0 and f(t) E C[O,t O]'
the totality of complex-valued continuous functions defined on the closed
interval [O,t ]' Find the solution of the boundary value problem:
O
y"(t) + iy(t) = f(t) (0 ~ t ~ to)'

(11.1)
{ yeO) = y(t o) = 0,
ANSWER: Define the values of f for t ~ to by f(t) f(t O)' Then
f € C = C[O,~). We first obtain the solution of
2
y"(t) + ay(t) = f(t), yeO) = 0, y'(O) = 8. (11.2)
The hyperfunction equation for (11.2) is
and its solution y is given by

8 f
y=2'2+22"
s +a s +a
= ..L (-!.- __
2ia S-la
I )
s+ia
+ ~_I _ _ I )f
2IQ\s-ia s+ia
iat -e -iat} + _1 {iat

e -e -iat} f
=
~e
a 2i a 2i
= ~sin at} + ~ {f: sin a(t-U)f(U)dU},
that is,
yet) = ~sin at} + ~ {f: sin a(t-U)f(U)dU}. (11.3)
This yet) is a solution of (11.1) if y(t )

o
= O. To discuss this condi-
tion, we distinguish two cases:
i) the first case: sin ato ~ O. In this case, y(t ) o is equi-

O
valent to
(11.4)
8 is uniquely determined since sin ato ~ O. Thus, by taking this 8

in (11.3), we obtain the solution of (11.1).
ii) the second case: sin ato O. In this case, we must have
t
foo sin a(to-u)f(u) =0
from y(t o) = O. Thus, by

sin a(to-u) = sin atO'cos au - cos ato·sin au = -cos ato·sin au,
we must have
t
fo
o sin au·f(u)du = O. (11.5)
And if (11.5) holds (of course, under the condition sin ato = 0), yet)
given by (11.3) is a solution of (11.1) for all 8.
Therefore, under the condition sin ato = 0, if f does not satisfy
(11.5) then (11.1) has no solution.
REMARK 11.1. When f(t) = 0, (11.5) holds and so for those a for which
sin ato = 0, (11.1) has a solution.
Hence, those a ~ 0 with sin ato 0 are characterized by the
property that the equation
2
f'(t) + a y(t) = 0; yeO) (11.6)
2
admits a non-zero solution yet), i.e., the real number a ~ 0 is an
11. Boundary Value Problem~ ~or ordinary Differential Equations 31
eigenvalue of the boundary value problem (11.6) and the non-zero solution
yet) is the corresponding eigenfunction of (11.6).
PROBLEM 11.2. Is the following problem solvable?
2
y"(t) + a yet) = 0, yeO) = 0, y(2'11) = 1. (11. 7)
ANSWER: Any Y satisfying (11.7) is a solution of (11.2) with f(t) = O.
-1
Such a function is y = a 8 sin at. Hence, if (11.7) has a solution,
then 2a must not be an integer. And when 2a is not an integer, the
solution of (11.7) is given by
t _ sin at
y( ) - sin 2a'll
PROBLEM 11.3. Is the following problem solvable?
2
~'(t) - a yet) = 0, yeO) = 0, y(2'11) = 1. (11. 8)
ANSWER: We first solve the initial value problem
y"(t) - iy(t) = 0, yeO) = 0, y' (0) = B,

obtaining
This y satisfies y(2'11) = if

a sinh(at)
B = sinh(2a'll)' yet) sinh (2a'll)
EXERCISES FOR §ll. Determine the functions y which satisfy the given
differential equations and boundary conditions.
2
1. y" + a y = 0; yeO) 0, y' (2'11) 1.
(a is a real number).
2
2. y" - a y = 0; yeO) 0, y' (2'11) 1.
Chapter IV
Fractional Powers of Hyperfunctions
I
h, sand s - ex
§12. EULER'S INTEGRALS - THE GAMMA FUNCTION AND BETA FUNCTION

These functions are respectively defined by Euler's * integrals:
f(A) = ( t A- 1e- t dt (ReA>O)** (12.1)
B(A,~) = f: tA-1(1_t)~-ldt (Re A > 0, Re ~ > 0). (12.2)
THEOREM 10.
f(A+l) = Ar(A) , (12.3)
f (n) = (n-l)! (n = 1,2, ... ), (12.4)
B(A
,~
) = f(A)f(H)
f(A+~) , (12.5)
1
f(z) = Iif. (12.6)
PROOF: (12.3):
A
f( 1\') =( t A- 1e- t dt = [tT e -t]= + I1 [ t Ae -t dt = I1 f('1\+ 1) •
t=O 0
(12.4): Combine (12.3) with
* Leonhard Euler (1707-1783).

**t A- l = exp((A-l)log t) taking the principal value of log t for t > 0
so that ItA-II = exp((Re A-1)10g t). Hence the integrals in (12.1) and
in (12.2) are convergent.
32
12. Euler's Integrals - The Gamma Function and Beta Function 33
(12.5) :
f(A)f(U) =
roro t A-I e -t uIJ-l e -IJ dtdu.
JOJ
O
By the substitution
t = TV, U = (l-T)V (0 < T < 1, V > 0) (12.7)
with the Jacobian determinant
dTV d(l-T)V
aT"' at
D(t,u) _
D(T,V) -
IV' -v I= 'V,
T, (l-T)
dTV d(1-T)V
av-' dV
we obtain
f(A)f(lJ) I: dTr: dV{(TV)A-le-TV((l_T)V)IJ-le-(l-T)V v}
I:TA-l(l-T)IJ-ldTr: vA+IJ-le-Vdv = B(A,u)f(A+IJ).
Hence we have (12.5). *

(12.6): By (12.4) and (12.5),
f(l)f(l) f(l + 1) fl t- l /2(1_t)-1/2 dt

2 2 22 0
l
1.I 1 dt = II 2xdx
o (t(l-t) 0 12 2
t'x (I-x)
Io
I 2dx 2[ . ]X=l
- - - = arCS1n x x=o
lr_x2
11
1
so that f("2) = ±1iI. We take the plus sign since f(A) (A > 0) is
positive, being an integral of a positive-valued continuous function.
COROLLARY OF (12.6).
(12.8)
* We omit the proof that f(A) # 0 whenever Re A > O. More precisely,

it is proved in analytic function theory that l/f(A) (Re A > 0) can be
analytically continued to an entire function, i.e., to a regular analytic
function defined for every complex number A. See, e.g., E. Hille:
Analytic Function Theory, Vol. I, Ginn and Company (1959), 229-.
34 IV. FRACTIONAL POWERS OF HYPERFUNCTIONS h, s AND _I_
s-ex
PROOF: By the substitution t = T 2 , we have

2
l
ao
T
2 0 e- dT.
COROLLARY OF (12.5). If n is a positive integer,
1·2· ·(n-1)
B(A,n) (Re A > 0). (12.9)
1.(1.+1) (A+n-1)
PROOF:
B(A,l) = I: A1
t - (1_t)Odt = I.
Then, integrating by parts,
B(A,n) =
I°I t
A-I
(l-t)
n-1 [t A
dt = T(l-t)
n_1]t=1
t=O
+
0
tA
II T(n-1) (1-t)
n-2
dt
n-1 II A n-2 n-1

= -X- 0 t (l-t) dt = --1.- B(A+1,n-1).
Therefore,
n-1 n-2
B(A,n) = -X- r;r B(A+2,n-2)
= n-1 n-2 n-3 ... _1_ B(A+n-1 1)

A 1.+1 1.+2 A+n-2 '
(n-l) (n-2) .. ·1 1
1.(>.+1) ... CA+n-1) (B(A,l) = r)'
§13. FRACTIONAL POWERS OF h, OF (s_ex)-I, AND OF (s-ex)
FRACTIONAL POWER hI.. We already know (2.3), i.e.,
n~~~}
n n-1}
h = { (~-1)! =
(n = 1,2, ... ).
n
We can extend h to hI. with Re A > 0 through
(13.1)
because if A = an integer n > I, we have
hn = sh n+1 .
We shall call hI. (Re A > 0; hO = I) the fractional power of h.
PROPOSITION 9. For fEe, we have

-1
13. Fractional Powers of h, of (s-a) and of (s-a) 35
{h Af(t)} = {f(A)
1 Jt0 (t-u) A-I f(u)du } € C (Re A > 0). (13.2)
PROOF: By (13.1), we have
hAf = S{fCA+l)-l J: (t-U)Af(U)dU}
= a~ {fCA+l)-l J: (t-U)Af(U)dU} + {f(A+l)-l J:«-U)Af(U)dU}
fCA+l)-l«t-u)AfCU))u=t + I: f(A+l)-lA(t-u)A-lfCu)du
{f(A)-l I: (t_U)A-lf(U)dU}.
because A(t_u)A-lf(u) is integrable with respect to u uniformly in the

parameter t when t is restricted to within a finite closed interval
[O,t O]'
REMARK 13.1. hAf is called the fractional integration of f € C.
EXAMPLE 13.1. By f(1/2) = ~ and 1
(t_u)1/2-1 = ----
,Tt:U
(h l / 2f)Ct) = -l It 1_ f(u)du. (13.3)

~ a ,Tt:U
FRACTIONAL POWER OF (s_a)-l. We shall extend the natural number n

in
to A with Re A > a through
(13.4)
because if A = an integer n ~ I, we have
(s-a) -n = sh {fen) -1 t n-l eat }.

-1
We shall call (13.4) the fractional power of (s-a) .
Similarly as in Proposition 9, we have
PROPOSITION 9'. For f € C, we have
(13.5)
s-a
11iEOREM 11.
(5-a)-A(5-a)-~ = (5_a)-A-~, (Re A > 0, Re ~ > 0). (13.6)
tA-1 at} { t~-l at}

(5-a) -A (5-a) -~ = s {h f(X) e 5 h r(~) e
A1
5
2 { t - at}{
h l'(X) e h
t~-l
r (~) e
at}
2 2{ft (t_u)A-1 ea(t-u) ~-1 au }

5 h 0 r(A)r(~) u e du
1 { at ft A-I ~-1 }
r(A)r(~) e 0(t-u) u du.
By the substitution t-u = tw, we obtain
r(A)~(~) {f:(tW)A-1(t-tw)~-lteatdw}
1
rCA) r(~)
{tA+~-leat} f
1
wA-1(1_w)~-ldW
0
t A+~-l eat } (by (12.5))

{r (A+Il)
(s-a) -A-~ (by (13.4)).

COROLLARY. When a = 0, we have
5
-A (Re A > 0) (13.7)
50 that
(Re A > 0, Re ~ > 0; hO = I). (13.8)
PROOF: By (13.4) and (13.1),
A
5- = 5{f:r(A)-luA-1dU} = 5{r(A+1)-l t A} = h A.
FRACTIONAL POWER (s_a)A (Re A > 0). We shall prove

PROPOSITION 10. For 0 < Re A < I, we define
(13.9)
13. Fractional Powers of h, of (5_0.)-1 and of (5-0.) 37
Then we have
A -A (13.10)
(5-0.) (5-0.) = (0 < Re A < 1).
By (13.9) and (13.4),
(S_o.)A(S_o.)-A = f(~=~~~(A) {f: eo.(t-u) (t_U)-X UX-1 eo.U du}.
By the substitution t-u = tw,

t
A -A (s_o.){eo. } flo t-Aw-XtX-1(1_w)X-1t dw
(5-0.) (5-0.) = f(l-A)r(X)
I ') f(l- A)f(A) _1_ = 1.

= f(l-A)r(A) B(l-A,/\ f(l-A)f(A) f(1)
REMARK 13.1. We sometimes call sAf (0 < Re X < 1) the fractional dif-
ferentiation of fEe.
EXAMPLE 13.2. We have
I {2 2
fliit _u }
srs+a: = ...'a-if 0 e du (0. > 0). (13.11)
PROOF: Since
_1__ = (5+0.)-1/2
rs+a:
we have
1
_1_= h _1_ = {ft u- /2 e-o.UdU}
srs+a: rs+a: 0 Ii
{fo
AA 1
-
Ii Ia
v -1 _v 2
(-) e
v
·2 c;: dv
}
{~ Erf /at},
where
v2
Erf t = Erf(t) = ~ ft e- dV. (13.12)
Ii 0
This function Erf(t) often occurs in the theory of probability and is
termed the error function.
s-a
EXERCISES FOR §13. Prove the equality
I = {eat Erf /(it}. (13.13)

,Is (s-a) Ici
GENERAL POWER hY (Y IS A COMPLEX NUMBER). We can define the
general power hY as follows:
hY = hy +n = {f(y+n)-lt y+n - l } *
1
hn {f(n)-lt n - l } (13.14)
n being any integer ~ 1 such that Re(y+n) > 1.
REMARK 13.2. The above definition (13.14) is reasonable, i.e., it is

without contradiction. This means that if we take another integer m ~
such that Re(y+m) > I, then
(13.15)
PROOF: By the substitution t-u = tw, we obtain
I: (t_U)m-luy+n-l dU }
{ f(m)
-I
f(y+n)
-1
Io
I
t
m-l m-l y+n-l
w t (l-w)
y+n-l
t dW}
{f(m) -1 f(y+n) -1 B(m,y+n)t y+n+m-l}
(by (12.5)).
We shall further discuss the general power sY and the general power
(s-a)y in §l9.
*Here t y+n - l exp((y+n-l)log t) by taking the principal value of log t

for t > O.
Chapter V
Hyperfunctions Represented by
Infinite Power Series in h
§14. THE BINOMIAL THEOREM
THE DEFINITION OF (l+z)a. Let a be any complex number. For any complex
number z with Izi < 1, we define
(l+z) a = ea 10g(1+z) , where we take the branch of the

(14.1)
{ function 10g(1+z) determined by 10g(1+0) = o.
Then we have
d (1 )a ea 10g(1+z) a
dz +z = I+Z = e(a-1)10g(1+z)Nu __ uN(1+z)a-1. (14.2)
We can prove
THEOREM 12.
( l+z)a = ~l (a) k h
k z, were
(a) f(a+1)
k = f(a-k+1)k!
k=O (14.3)
{ = a(a-1)··· (a-k+1) a
k! (k = 1,2,3, ... ; (0) = 1),
and the series E~=o (~)zk is convergent for Izi < 1.

PROOF: 10g(1+z), determined by 10g(1+0) = 0, is analytic for Izi < 1.
k
Therefore, the Taylor series E;=o akz of the function (l+z)a is con-
vergent for Izi < 1 and represents the function (l+z)a. The coeffici-
ents a are given by differentiation:
k
~ = k~ (~ (l+Z)a) (k = 1,2, ... ; a O = 1),

dz z=O
so that, by (14.2),
39
40 V. HYPERFUNCTIONS REPRESENTED BY INFINITE POWER SERIES IN h
1 a-I a
al = IT(a(1+z) ) z=o = IT '
and so on.
EXAMPLES OF BINOMIAL EXPANSIONS (SERIES)
__1__ = 1 _ t + t 2 _ t 2 +
l+t
1
(1+t)2 =
1 1__ 2 1·3 3 1·3·5 t4
Ii+t = + zt 2.4 t + 2·4·6 t - 2·4'6·8
+ •••
§15. BESSEL'S FUNCTION In(t)
THEOREM 13. The Bessel function In(t) of order n is defined as a solu-

tion yet) of the Bessel * differential equation
1 n
2
y"(t) + t y' (t) + (1 - 2)y(t) = 0, (15.1)
t
and it can be obtained as a solution of (15.1) having the power series

expansion
yet) + •.. +
(15.2)
2n r(n+1) '
so that yet) = In(t):
I
In(t) = v=o r(n+v+1)r(v+1)
(_l)V (t)n+2V
2 (n = 0,1,2, ... ). (15.3)
PROOF: By the formal term-wise differentiation of (15.2), we have
n-2 n ••• +
y" n(n-1)C ot + (n+2)(n+1)C y +
2
n+2V-2
(n+2v) (n+2v-1)C t + •••
2v
n n
nCot - 2 + (2)
n+ C2t + •" + + •••
*Friedrich Wilhelm Bessel (1784-1846).

15. Bessel's Function J (t) 41
n
2
n
- ~ y
Hence, by putting this formal power series equal to zero, we obtain
(lJ = 0,1,2, ... ), (15.4)
since this is the coefficient of t n+2lJ - 2 equated with O. Thus
C -1 C
2lJ = (2n+2lJ)2lJ 2(lJ-l)
and so, combined with C = 1 , we obtain (15.3).
o 2n·f(n+l)
We next show that the formal power series given by (15.3) is absolutely
convergent for every t. To prove it, let k > 1 be a positive integer
O
such that
t 1
2k = is < 2·
O
Then, for any integer k > k ' we have
O
If(n+k+l)f(k+l)
k (-1) t n+2k
(2)
I
Hence
~ I (-1) lJ t n+ 2lJ I
lJ:k f(n+lJ+l)f(lJ+ l ) (2)
o
~_ 1!.2 1n kO-l
IT 11 t I
n+lJ '! . iJ 2" l.
I
11 t x (OO~ 02(k-kO+l)) •
lJ=l k=k O
The right hand side is convergent since roo o2(k-k o+l) is a geometric
series with 0 < 0 < 1/2. k=k O
Finally we recall for the reader the well known
oo n
PROPOSITION 11. If a power series pet) = rn=o ant in t is absolutely
convergent at a point t = to 1 0, then pet) is continuously differenti-
able at any t with It I < Itol and the derivative p'(t) is obtained
by termwise differentiation:
oo
k l
p'(t) = L kakt - at every t with It I < Itol. (15.5)
k=O
42 v. HYPERFUNCTIONS REPRESENTED BY INFINITE POWER SERIES IN h
Therefore. the above method of formal term-wise differentiation is

legitimate to obtain the Bessel function In(t). We will see in §20 that
In(t) can be obtained by virtue of the Operational Calculus.
§16. HYPERFUNCTIONS REPRESENTED BY POWER SERIES IN h

THEOREM 14. Let
n + •••
+'''+CLZ (16.1)
n
be a power series in complex variable z with complex coefficients

CLO.CL •••• • Assume that (16.1) is convergent at a certain value Zo # 0
l
of z. Then the power series in h:
CLOh
o+ CLlh + CLZh
Z+ ••• + CLnh
n
+ (16.2)
defines. by hn r(n)-ltn - l • a hyperfunction
+ '" + CL
n
Tn:n-r
n-l
+ ••• ) € C
w (16.3)
PROOF: Since (16.1) is convergent at z = Zo # O. there exists a positive

number M < ~ such that
suplCL zonl = M < ~. (16.4)

n~O n
Thus
(n 0.1.2 .... ).
Hence. for any to > O. there exists a natural number kO such that
_t_ /k 0 <lZ for o ~ t ~ to'
Izol 0
Then we have. for (n-l) ~ kO'
n
ICLnzo , t n- l I I
(n-l)l!zO' ~
~ M kOn-1 ( l ) 1 x IS n-k 0 ~ M kOn-1 ( 1 ) 1 (l) n-k 0

TZJ
t t
- k=l k TZ;;T TZQT - k=l r TZ;;T Z
Hence
16. Hyperfunction~ Repre~ented by Power Serie~ in h 43
is uniformly convergent for 0 ~ t < to'

Therefore,
belongs to C[O,~), and so (16.2) belongs to CH.

THEOREM 15. If a is a positive number, we have
(16.5)
PROOF: We have
and hence
Here the equality
I = (I + (ah)2)-1/2
II+(ah)2
is seen from
1 2 -1/2
--::"""';2"""1'""/>':<"2 = (l + (az) )
(l+(az) )
On the other hand, we have
EXAMPLE 16.1.
_1_=1+ z Z
+ z + '"
1-z
is convergent for Izi < 1. Hence we have
EXAMPLE 16. Z.
z3 z5
sin z = z - 3T + S! -
converges at every z. Hence
I h3 h
sin -s = sin h = h -3T+ST-
S
... = {I - tZ
3!Z! + S!4! -
t4
..1 (16.6)
EXAMPLE 16.3. Let a be a positive number. then
r22
(I s-+a-)- s)
n
(n 1 ,Z, ... ). (16.7)
PROOF: We shall make use of
~
(d+z - 1)
n
= n ~l. (-1)
k (n+Zk-1)!
z
n+k
Izl < 1. (16.8)
k=O zn+Zkk!(n+k)!
Prior to the proof of (16.8), we shall derive (16.7) from (16.8). Thus,
by Theorem 13,
n I (_l)k (n+Zk-1)! aZ(n+k)h n+Zk

k=O zn+Zkk!(n+k)!
= k aZ(n+k)tn+zk-l}
{n L (-1)
k=O zn+Zkk!(n+k)!
{n~n In(at)}.
Hence, in particular,
~
({ s-+a- - s)
n
= n
{=L (-1) k aZ(n+k)tn+Zk-1 } . (16.8)'
k=O zn+Zkk!(n+k)!
16. Hyperfunctions Represented by Power Series in h 45
Now we shall prove (16.8). Put
(n = 1,2, ... ). (16.9)
Then it is easy to see
(n+1) (,;r:;:z _ l)n

£~+l(z) 2,;r:;:z
{ (z-n /2 £ (z))' = 2
n -n/2-1 (1f+Z _ l)n
z
n ,;r:;:z
Hence
£ 1 (z)
=n+1
- z n/2+1( z-n/2£ ( z
))'.
n+ 1 n n
This, combined with £n+1(0) = 0, gives
f
n+l
(z) = ~ fZ zn/2+1(z- n/2 f (z))'dz.
nOn
(16.10)
(16.11)
Therefore, by (16.10) and (16.11) we obtain (16.8).

COROLLARY OF (16.7). We have
T22
(,I s-o\O- -s)
n
(n 0,1,2, ... ). (16.12)
,I!'""'22
s-+a-
PROOF: In (16.7) we substitute (n+l) and then differentiate with res-

-1
pect to a and multiply by ((n+1)a) .
EXAMPLE 16.4. Let a > O. Then
-a/s
e = S{J (2M)}. (16.13)
O
PROOF:
.. k '" k k+1
e
-ah L (_l)k (ah~ = s L (_l)k ~
k=O k. k=O k.
..
s{ L (-1) k ~
ktk}
= s{J
O
(2M)}.
k=O (k!)
EXAMPLE 16.5. Let 0 > O. Then
_e__
-0/5 = { _1_ cos 2M} .
(16.14)
IS ..r,rt
PROOF:
1 2 ah ~ k a kh k+1/ 2 ~ k a kt k- 1/ 2
h / e- = L (-1) k! L (-1) k!r(k+1/2)
k=O k=O
~ k a k t k- 1/ 2
L (-1)
k=O
k(k-1) ... 1(k-1/2)(k-3/2)'" (l/2)r(l/2)
~ k 22k a kt k-1/2
= L (-1)
k=O (2k)(2k-1)"'1';W
~ 1'::7 2k
= __1__ L (_l)k (2.at) ,
~ k=O (2k).
= s{ft __ 1_ cos 21cl1J dU}.

o IiiU
EXERCISES FOR §16. Prove 1-3 below.
1. ~ e-a/s = {1tTcX J 1 (2Icit)}

5
2. ..!.- eO / s = {_1__ cosh (2M)}

IS lift
3. ~ e- a / s = {J
O
(2M)}.
Part II
Linear Ordinary Differential Equations with
Linear Coefficients
(The Class CIC of Hyperfunctions)
Chapter VI
The Titchmarsh Convolution Theorem and the
Class CIC
§17. PROOF OF THE TITCHMARSH CONVOLUTION THEOREM *
THEOREM 16. (THE TITCHMARSH CONVOLUTION THEOREM). If the convolution

fg of two functions f and gEe is 0, then either f = 0 or g = 0
must be true.
PROOF: THE FIRST STEP (THE CASE f = g). We shall start with
I: f(t-T)f(T)dT = 0 for all t > O. (17.1)
Let T > 0 and B be real numbers. Put
~(t) = e-iBtf(t), where i = ;:T. (17.2)
Then, for any real number a, we obtain
(IT eau f(T_U)dU)2 = IT IT ea(U+V)f(T_u)f(T_v)du dv

-T -T -T
= I~T dv(I:=_v ea(u+V)f(T_U)f(T_V)dU)

(17.3)
(fu=-v )
I
+ T dv
-T -T
ea(U+V)f(T_u)f(T_v)du
= I a,B,l + I
a,B,Z
.
By the substitution
* Kosaku Yosida and Shigetake Matsuura: A note on Mikusinski's proof of

the Titchmarsh convolution theorem, to be published in the Contemporary
Mathematics Series of the Amer. Math. Society.
47
48 VI. THE T:J;TCHMARSH CONVOLUTION THEOREM AND THE CLASS C/C
u =T - (t-T), V = T-T with
2T ~ u+v ~ 0, T ~ u ~ -T, T ~ v > -T and
<l(u,v) = 1
<l(t,T) ,
we have, by (17.1) and (17.2),
Ia ,B,l f: T
a
dt(f: e (2T-t)f(t_T)f(T)dT)
f: T
ea (2T-t)-i Bt(f: f(t-T)f(T)dT)dt (17.4)
= O.
If a > 0, then eU(u+v) ~ in the integration domain of the integral
Hence
!I a ,B,2 1 ~ f~T dv(f~:-vlf(T-U)f(T-V) IdU)

(17.5)
T fT 2 o<
~ f-T -T If(T-u)f(T-v) Idudv ~ M with M< 00.
Therefore, by (17.3), (17.4) and (17.5), we have
If~T e(U+iB)Uf(T_U)dUI < M whenever U ~ O. (17.6)
Thus the Laplace integral
(17.7)
satisfies, for a = Re z ~ 0,
N = M+ fO If(T-u) Idu. (17.8)

-T
Hence, we easily prove that the entire analytic function (17.7) satisfies
(17.9)
Nl = max(N,K) with K = max T x If(T-u) I.
O<u<T
==
Therefore, by Liouville's theorem * in analytic function theory, we have
*Joseph Liouville (1809-1882).
17. Proof of the Titchmarsh Convolution Theorem 49
FT(z) =a constant for all z.
Thus, by differentiating FT(z) at z = 0, we obtain
f: unf(T-u)du = 0, (n • 1,2, .•. ),
and so, for every polynomial p(u) in u, we have
=0
fTo p(u)uf(T-u)du
* we obtain
Hence, by Weierstrass' polynomial approximation theorem,
f: q(u)uf(T-u)du = 0
for every continuous function q(u). This proves that uf(T-u) =0 for
o~ u ~ T, i.e., f(t) must vanish for 0 ~ t ~ T. As T > 0 was ar-
bitrarily chosen, we see that f(t) = O.
THE SECOND STEP (THE GENERAL CASE f ~ g). We start with
(fg)(t) = f: f(t-T)g(T)dT =0 for all t > O. (17.10)
We put, following J. Mikusinski [5), p. 22,
f 1 (t) = tf(t) and gl(t) = tg(t). (17.11)

Then
(17.12)
because
f:(t-T)f(t-T)g(T)dT + f: f(t-T)Tg(T)dT
=t f: f(t-T)g(T)dT =0 by (17.10).
Hence, by making use of Theorem 1 (§l) and (17.10), we obtain
o (fg l ) (£lg + fg l ) = (£g)(flg l ) + (£gl)(£gl)
(£gl)(fg l )·
*Karl Theodor Wilhelm Weierstrass (1815-1897). The theorem reads: Let

q(t) be any continuous function on the closed interval [O,T). Then, for
any number E > 0, there exists a polynomial pet) such that
max Iq(t)-p(t)1 < E. For a proof, see §2l of this book.
O~t~T
50 VI. THE TITCHMARSH CONVOLUTION THEOREM AND THE CLASS CIC
By virtue of the first step. we thus have fg O. i.e .•

l
f: f(t-t)tg(t)dt == O. (17.13)
Repeating the same reasoning. we have
f: f(t-t)tng(t)dt == 0 (n = 0.1.2 •... ). (17.14)
Hence. for every polynomial pet), we obtain
Ito f(t-t)p(t)g(t)dt == 0
and so. by Weierstrass' polynomial approximation theorem.

t
fo f(t-t)q(t)g(t)dt == 0
for every continuous function q(t).

This proves that
f(t-t)g(t) =0 for 0 ~ t ~ t < ~.
Therefore. either f(t) == 0 or get) == 0 must be true.
§18. THE CLASS CIC OF HYPERFUNCTIONS

The class C of hyperfunctions was introduced on the basis of
H
Proposition 3 (§3). Similarly. on the basis of Theorem 16 (§17). we can
introduce the class CIC of hyperfunctions. In fact. we can prove
PROPOSITION 12. If f and g are in C = C[O.oo) and g F O. then we
can define the "fraction" i = fig by the equivalence relation ::::
g
means (18.1)
because we can prove
implies
implies (18.1)3
18. The Class CIC of Hyperfunctions 51
PROOF: We shall prove (18.1)3 and omit the others. Thus we start with
Hence
and so
f
l
Since as the denominator of the fraction g-' we obtain by Theorem
1
16,
This proves the equivalence
Therefore, we can write
if and only if fg l = fIg and g # 0, &1 # O. (18.1)'
The totality CIC of fractions of the form gf constitutes a ring * with

the sum and product given by
! + f' = fg l + fl & (18.2)

g gr gg'
f f' ff'
ggr=ggr (18.3)
In the above, gg' # 0 by Theorem 16, since g # 0 and g' # O. This

ring CIC is commutative with respect to the product.
REMARK 18.1. As consequences of (18.1)', (18.2), and (18.3), we have the
following important results (i)-(iii):
(i) Fractions of the form 0 (g # 0) are mutually equal to one

&
another, and they represent the zero of the ring CIC so that
o f' f' o f' 0
i+~=&'
and g gr = g' (18.4)
(ii) Fractions of the form ~ (g # 0) are mutually equal to one

another, and they represent the (multiplicative) unit of the ring CIC
so that
Z. f' _ fl
g gr - gr (18.5)
*Similarly as CH.
52 VI. THE TITCHMARSH CONVOLUTION THEOREM AND THE CLASS C/C
(iii) A fraction of the form £g has a (multiplicative) inverse if

and only if f ~ 0, and, in this case,
£.8.=1
g r ' Le., 1/(£)
g
(£)-1
g
_ .8.
- f' (18.6)
Hence, in case f ~ 0, £
g and t are (multiplicative) inverses of each
other.
REMARK 18.2. f is considered as a fraction in CIC, the ring

Since
n
h
C is a subring of the ring CIC. Thus we may call fractions in CIC
H
hyperfunctions. In this way, the notion of operator or hyperfunction
given in Part I is generalized to a much bigger extent by CIC.
Chapter VII
The Algebraic Derivative Applied to Laplace's
Differential Equation
§19. THE ALGEBRAIC DERIVATIVE
Pierre Simon Laplace (1749 - 1827) in his treatise "Theorie analyti-

que des probabilites" of 1817 considered a differential equation which
now carries his name and which may be written as
(19.1)
where the a's and b's are given complex numbers with# O. a
Z
Laplace's method of integration of (19.1) was to use a transformation
yet) ~ Yeo) r: e-
ot
y(t)dt, (19.2)
which is now known as the Laplace transform.

As we have discussed in Chapter III, we could solve linear ordinary
differential equations with constant coefficients algebraically by making
use of hyperfunctions of the form
~ (p(s) and q(s) are polynomials in s)

pes)
belonging to the ring CH.
In view of the importance of the equation (19.1) in analysis, the
purpose of the next section, §20, is to show that we can solve (19.1)
algebraiclly by making use of the hyperfunctions
(s-aI) Y (a and yare complex numbers)
belonging to the ring CIC.
53
54 VII. THE ALGEBRAIC DERIVATIVE
To this aim, we first explain the notion of the algebraic derivative

D due to J. Mikusinski [5], and then we shall develop the calculus of D*
when applied to the fractional powers hY, (I-ah)Y and (s-aI)Y.
DEFINITION 19.1. The algebraic derivative D is a mapping of CIC into

CIC such that
f € C,
1
DDf_:=={-tf(t)} for
(19.3)
(Df) g - f(Dg) f
2 for g € CIC.
g
PROPOSITION 13. Definition (19.3) is consistent; that is,
Df = D i& = (D(fg))g - (fg)Dg (19.4)

g 2
g
implies D i _- 1
D _f
(19.5)
g gl
PROOF: We have
(19.6)
because
D(ff1) = {-t f: f(t-U)f 1 (U)dU}
{- f: (t-u)f(t-u)f 1 (u)du • f: f(t-u) (-U)f 1 (U)dU}
Hence,
D i& = (D(fg))g - fg(Dg) ((Df)g. f(Dg))g.- fg(Dg)

g 2 2
g g
= J..Q!2..&. = Df .
g
Next we have, by the commutativity of the ring C and (19.3)-(19.4),
(D(fg 1))ggl - fg 1 (0(ggl))

2
(ggl)
(Df)gl gg 1 • f(Ogl)ggl - fg 1 (Dg)gl - fg 1g(Dg 1)
2
(ggl)
*K. Yosida [10]: The algebraic derivative and Laplace's differential

equation, Proc. Japan Acad., 59, Ser. A, No.1 (1983), 1-4.
((O£)g - £(Og))gl g l + £gl ((Ogl)g - g(Ogl))

2 2
g gl
(Of)g - f(Og) D i.
g2 g
Similarly we have
On the other hand,
and so
proving that
PROPOSITION 14. We have
(19.7)
(19.8)
(19.9)
PROOF: We have
D[a] = 0; in particular, DI 0, (19.10)
because
D[a] D {a} = (O{a})h - {a}(Dh)

h h2
{-ta}h - {a}{-t}
2
h
We next prove (19.8). We have, for f and f € C,

1
D(f + f 1) {-t(f(t) + f (t))}
1 (19.11)
{-tf(t)} + {-tf1 (t)} Df + Of 1 .
56 VII • THE ALGEBRAIC DERIVATIVE
Since
f f1 f&l f 1&1
-+-=-+--,
& &1 &&1 &&1
we may restrict ourselves, in the proof of (19.8), to the case & = &1 = k.
Thus
D f+f' = (Df+Df') k - (Dk) (f+f')

k k2
(Df) k - (Dk) f (Df') k - (Dk) f'
2 + 2
k k
f f'
D k + D T
We then prove (19.9) in the case & = &1 k. Thus

2 2
f f ff 1 (D(ff 1))k - ff 1 (Dk )
Dk k
1 D- = ---"----.,---"---
2 4
k k
2 2
(Df)f 1k + f(Df )k 2k(Dk)ff
1 l
4 4
k k
(Df)k - (Dk)f f 1 f (Df 1)k - (Dk)f 1

k2 T +k k2
f f1 f ( f 1)
(D f)T + k DT
and so, by (19.5),
Hence (19.9) is proved.

Finally, by (19.10) and (19.9), we obtain (19.7). In fact,
D( [OJ~) (D [oni + [0] (D ~) [0] (D

f
g) .
PROPOSITION IS.
Ifa a = ~ E C/C and then

(19.12)
(Da)b - a(Db) D rna .
{ Db = b2 ~
PROOF: We have
a = a!b = ~~
-b n p =~
np (by (18.6)).
And
o ~ = (D(mq))np - mq(D(np))
np 2 2
n p
((Dm)q + m(Dq))np - mq((Dn)p + n(Op)

2 2
n p
((Dm)n - m(Dn))pq - mn((Dp)q - p(Dq))

2 2
n p
oa (Da)b - a(Db)
b b2
= ((Dm)n - m(Dn)) £ _ ~ (Dp)q - P(Dq)) 1(£)2

2 q n 2 q
n q
= ((Dm)n - m(Dn))pq - mn((Dp)q - p(Dq))

2 2
n p
The preceding formulas are given, roughly, in J. Mikusinski [1].

We shall now prove
THEOREM 17. ill For any complex number y, we have
DhY = -yhy+ 1, (19.13)

where, as was defined in (13.14),
hY = hy+n = {r(x+n)-l t x+n-1}
1
hn {r(n)-l t n-1} ,
(19.14)
n being any integer ~ 1 such that Re(y+n) > 1.
(ii) For complex numbers ~ and y, we have
(19.15)
where
(19.16)
(iii) We have
(19.17)
(19.18)
PROOF: (19.13). We have
O
(n = 0.1.2 •... ; h = I). (19.13)'
because
n n'.hn+1 1
n (tn-I) t _nh n+ .
Dh = D (n-1)! - (n-1)! = - (n-1)! =
In particular.
Dh O = DI =0 in accordance with (19.10).
Hence we have. by (19.14) (= (13.14)) and (19.12).
DhY = D hy+n = (Dhy+n)hn - hy+n(Dh n)

n 2n
h h
{_tr(y+n)-l t y+n-l}hn _ hy +n (_nh n+1)
2n
h
_(r(y+n)-lr(y+n+l)hy+n+lhn + nhy +2n +1

2n
h
_(y+n_n)hy +2n +1 y +1
2n -Yh
h
(19.15). Since the binomial expansion
converges for lazol < 1. the series
converges for every t. thanks to the factors ((k_1)!)-1 (see Theorem

14 in §16). Hence. by DI =' O.
(19.17). By (19.12), we have
D(s - aI)Y = D (I-ah)Y = (D(I-ah)Y)hY - (I-ah)Y(DhY)

2y
hY h
_ y(I_ah)Y-l(ah 2hY + (I-ah)hy +l )
- h 2y
_ y(I_ah)y-lhY+l y-l
- h2'Y = y(s-aI) .
d
REMARK 19.1. Thus the algebraic derivative D may be denoted by ds
when applied to "functions of s":
d
D = ds' (19.19)
For example, we have
(19.17)'
and (19.17).
REMARK 19.2. As a corollary of (19.18), we have
o
(s-aI) Y(s-aI) = (s-aI) y+o (19.20)
for any triple
{a,y,o}
of complex numbers a, y and O. Thus (13.6) is now completely general-
ized, and so we have
(19.20) 1
and
(19.20)"
Thus (13.8) is completely generalized.
PROOF OF (19.20); By (14.1), we obtain
(I-az)Y(I-az)o exp(y log(l-az))exp(o log(l-az))
exp((y+o)log(l-az))
(1-az) y+o.
This proves
(s-aI)Y(s-aI)O
PROOF OF (19.20)": We have
sY = h- Y, in particular SO hO I and Ds I, (19.20)'"

by (19.18) and (19.17) '.
§ZO. LAPLACE'S DIFFERENTIAL EQUATION
In the equation given by (19.1), we may assume that b 0 by taking

2
(t - :~)
as a new independent variable. Thus we shall discuss the differential
equation
aztY"(t) + (a 1t+b 1)yt (t) + (aot+bo)y(t) o

(ZO.l)
{ with a F 0
Z
in the domain t > O.
For the twice continuously differentiable solution yet) of (ZO.1),
given initial conditions yeO) and y'(O), we obtain, by (5.6) and (19.3),
the equation
Z
-aZD(s Y - s[y(O)] - [y' (0)))
- a 1D(sy - [yeO)]) + b 1 (sy - [yeO)]) - aoDy + bOY = O.
This is, by (19.9), (19.10) and (19.17)', the same as
Z
-(azs + a 1s + aO)Dy - (2a Zs + a 1 - b1s - bo)y
(ZO.Z)
+ (a Z-b 1) [y(O)] = O.
Therefore, we have
PROPOSITION 16. Assuming that the above solution yet) is ~ 0 and

satisfies the initial condition
(ZO.3)
we can convert (ZO.l) into

20. Laplace's Di{ferential Equation 61
(ZO.4)
We thus have
THEOREM 18. If the algebraic equation
(ZO.S)
has two distinct roots zl and zz, then (ZO.4) is solvable by a hyper-
function y € CIC:
(C is a non-zero constant) (ZO.6)
where complex numbers Y and Y are given from the partial fraction
l Z
expansion
(-ZaZ+bl)s - a l + bO
Z (ZO.7)
aZs + a l s + a OI
PROOF: We have, by (19.9) and (19.17),
Yl YZ
D(C(s - zlI) (s-zZI) )
Yl YZ Yl YZ
C((D(s-zlI) (s-zZI) + (s-zlI) (D(s-zZI) ))
Yl-l Y2 Y
l Y 2 -l
= C(yl(s-zlI) (s-zZI) + (s-zlI) YZ(S-ZZI) ).
Therefore, (ZO.6) satisfies
(-ZaZ+bl)s - a l + bO
Z
aZs + als + aOI
EXAMPLE ZO.l. (THE BESSEL DIFFERENTIAL EQUATION). The equation is
(ZO.8)
where a is a complex number. By the substitution
the above equation becomes
ty" (t) - (Za-l)y' (t) + ty(t) O. (ZO.9)
Here we have
so a - b Za. Hence the initial condition (ZO.3) becomes

Z l
yeO) = 0 if a f O. (ZO.lO)
The equation (ZO.4) for (ZO.9) is
Ql _ -(Za+l) -a-liZ -a-liZ

y sZ+I = ~ + --s:rr-- (ZO.l1)
Hence
satisfies (ZO.ll).
We have, in the case Re a ~ 0,
-a-liZ (-a-liZ) (-a-3/Z) ... (-a-k + 1/2)

( k ) k!
k
(Za+l) (Za+3) ... (Za+Zk-l) (-1)
k!Zk
_ 1:..!l....k (Za+l) (ZO+Z) (20+2k)
- k!Zk Zk(a+l) (a+k)
(-1)kr (Za+2k+l)r(a+l)
Zk .
Z r(k+l)r(Zo+l)r(o+k+l)
Thus we obtain
00 (-1) k t Zk Za
Ya (t) -- \
L r(k+l)r(a+k+l)
(-Z) + by taking
k=O (ZO .1Z)
c= r(2a+l)
za
r(0+1)2
If Re a > 1, then Ya(t) is twice continuously differentiable in

t for t ~ O. Hence the solution Ya(t) of (20.11) satisfying (20.10)
is also a solution of (ZO.9) for t ~ O. This means that, when t ~ 0
-2k+Za-l
and Re 0 > 1, the coefficients of t in the infinite series in
t given by
must vanish as analytic functions of 0 (k 0,1,2, ... ).

Therefore, since Ya(t) with Re a ~ 0 is twice continuously dif-

ferentiable in t > 0, we see, as in the case of Re a > 1, that the for-
mula
ty~(t) - (2a-l)y~(t) + tYa(t)
must vanish when t > 0, because the coefficients of t2k+2a-l all vanish
in the above formula.
Thus we have proved that, when Re a > 0 or a = 0, ya(t) given by
(20.12) is a solution of (20.9) at every t > 0 and satisfies (20.10).
In this way, we have obtained the Bessel function of the first kind
and of order a (Re a > 0 or a = 0)
oo}: (_l)k t 2k
J (t)
=t '" (t) _ (_) +a
_N
a ya = k=O r(k+l)r(a+k+l) 2 (20.13)
which satisfies the original Bessel differential equation (20.8) for

t > O.
REMARK 20.1. In order to obtain another solution ya(t) of (20.9) which

is linearly independent of y (t), we shall use the so-called D'Alembert
* a
method. We put
ya(t) = Ya(t)x(t), (20.14)
and from
{ t:~ (t) (2a-l)Y~(t) + tYa(t) 0,
ty~(t) - (2a-l)y~(t) + traCt) 0,
we obtain
tya(t)X"(t) + (2ty~(t) - (2a-l)ya (t))x'(t) = O. (20.15)
This is a linear ordinary differential equation of the first order for

the unknown x'(t). Hence, by knowing ya(t) ~ 0, we can obtain a solu-
tion x'(t) ~ 0 of (20.15). Therefore, we may take
where x(t) is not a constant since it is a primitive of x'(t) ~ O.

The same remark applies to the following Examples.
EXAMPLE 20.2. (THE CONFLUENT HYPERGEOMETRIC DIFFERENTIAL EQUATION). The

equation is
*Jean Ie Rond D'Alembert (1717-1783).
ty"(t) + (C-t)y'(t) - ay(t) = 0, (20.16)

where c and a are complex numbers. * Here we have
so (a 2-b l ) - c. Thus the initial condition (20.3) becomes
yeO) =0 if c # 1. (20.17)
The equation (20.4) for (20.16) is
Q[ _ (-2+c)s + 1 - a a-I
=--+
c - a -
(20.18)
y s2_ s S s -
Thus, C being a non-zero constant,

l
C hl-a-c+a+l(I_h)c-a-l
1
When Re(l-c) > 0 or l-c 0, we have
c-a-l k t k+l - c (a-c+l) (a-c+2) (a-c+k) t k+l - C

( k ) (-1) r(k+2-c) r(2-c) (2-c+l) (2-c+k-l) -k-!-
Thus, as in the case of Ya(t), we have the following result: If

Re(l-c) > 0 or (I-c) = 0,
then
l-c ~ (a-c+l) (a-c+2) ... (a-c+k) t k

y (t) = t L - (20.19)
c,a k=of(2-c) (2-c+l) '" (2-c+k-l) k!
is a solution of (20.16) for every t > 0 and satisfies (20.17).
EXAMPLE 20.3. (THE LAGUERRE** DIFFERENTIAL EQUATION). The equation is
*This equation is obtained from the Gauss differential equation
t(l-kt)y"(t) + (c-bt)y' (t) - ay(t) = O.
by confluence, i.e., by letting k ~ 0 and b ~ 1.

** Edmond Nicolas Laguerre (1834-1886).
tf'(t) - (t+a-l)y'(t) + (a+\)y(t) = O.

(20.20)
(a and \ are complex numbers).
so that it is essentially the same as the confluent hypergeometric dif-

ferential equation.
For the equation (20.20). we have
a 2 = 1. b 2 = O. a
l = -1. b l = I-a. a o = O. bo a+\.
so a -b = a. Hence the initial condition becomes

2 l
yeO) = 0 if a ~ O. (20.21)
The equation (20.4) for (20.20) becomes
Ql = (-2+1-a)s + -1 - a - \ +-
\
(20.22)
y /-s s s-1
Thus. for Re a > 0 or a = O.
(20.23)
is a solution of (20.22) satisfying (20.21). The proof is the same as in

Example 20.1.
-a
Moreover. t Ya.\ reduces to a polynomial in t if and only if
A = 0.1.2 •.... Furthermore. when
-1
\ =n and C = r(a+n+l)r(n+l) •
we obtain the n-th Laguerre polynomial of order a,
n k
~ n (-t)
L. (k)r(k+a+l)
k=O
(20.24)
because
n(n-l) ... (n-k+l) (a+n) (a+n-l) ... (a+k+l)r(a+k+l)

k! n(n-l) ... (n-k+l) ((n-k) !)r(a+k+l)
(n+a) 1
n-k IT
Usually, Ln(t) L~O)(t) is called the n-th Laguerre polynomial and

we have
A COMPLEMENT TO THEOREM 18. Consider the equation
.Ql.=-L+ 6 (a, 6 are complex numbers), (20.25)

y s-aI (s-aI) 2
in the case that the algebraic equation
has a double root a. In this case, we may obtain a solution y of

(20.25):
y = C(s - aI)Y9, where

(20.26)
y = Ta e- 6h € C/C (but not € C.)
Here Ta is the mapping* of C/C into C/C defined by
Ta ! = Taf = {eatf(t)} (when f,g € C and g ~ 0). (20.27)

g Tag {eatg(t)}
PROOF OF THE CONSISTENCY OF (20.27) WITH rIlf = {eatf(t)}: Starting with

rIlf = {eatf(t)} we have
rIl(fk) = {eat f: f(t-U)k(U)dU} (when f,k € C)
= {f: ea(t-U)f(t_U)eauk(U)dU} (20.28)
and
(20.29)
Thus
= g;
f fl
for a = g' b
(20.30)
(f,g,f 1,gl € C and g ~ 0, gl ~ 0 and fl ~ 0),
* Due to J. Mikusinski [5], p. 246ff; cf. §37 of the present book.

In this way, we have proved the above consistency.

Furthermore, we have
A f
for f and g € C such that y = g' (20.31)
In fact, we obtain, by (20.27) and (19.12),
T-a(Oy) = T-a(O!) = T-a (Of)g - f(Og)

g 2
g
= T-a((Of)g _ f(Og)) _ (T-a(Of)) (T-ag) - (T-af)(T-a(Og))

T- a g2 - (T-ag) (T-ag)
(O(T-af))(T-ag) - (T-af)(O(T-ag)) 0 T-af = O(T-ay).

(T-ag) (T-ag) T-ag
As a corollary of (20.31), we have

f
For y =- with f,g € C,
A
1
g
(20.32)
.Qi. = III is equivalent to
2
Y (s-aI)
A
THE PROOF IS EASY. In fact, we have
III 2 = (T- a ll)(T- a {eat })2

(s-aI)
and, vice versa,
= BHeat }2 = III ---I----".2 .

(s-aI)
Hence, motivated by (19.19), we shall take
(20.33)
because we have, by (19.19),
Dz = ~ e- Bls
ds
= e- Bls ~2 = ~
2' (20.34)
5 5
Therefore, by (20.33), we have obtained the proof of (20.26):
y = TaT-ay = Tae- Bh . (20.26)'
REMARK 20.1. We already know (16.13):
(16.13)'
so that, by (20.27),
{e
at J (2r'Bt)}
O
{eat}
(20.35)
Hence, by (15.3), we see that
Y€ CH c CIC, but y~ C since [JO(O)] # O.
EXAMPLE 20.4. For the equation
aty"(t) + Yet) = 0, (20.36)

we have
so a 2-b 1 = a. Hence the initial condition (20.3) for (20.36) becomes

yeO) =0 if a # O. (20.37)
Furthermore, (20.4) for (20.36) is given by
Ql _ -2as + -2a I
=-5-+2' (20.38)
y 52 5
Hence, by (20.35),
(20.39)
20. Laplace's Ditterential Equation 69
is a solution of (20.38) satisfying (20.37) when Re(2a-l) > o.

If Re a > f, then (20.39) is twice continuously differentiable in
t for t > O. Hence the solution (20.39) of (20.37) - (20.38) is, for
Re a > f, :
solution of (20.36) at every t ~ O.
Therefore, as in Example 20.1, we see that, for t > 0 and
1
Re a> 7' (20.39) is a solution of (20.36) satisfying (20.37).
EXERCISES FOR §20.

(a) Verify
(m = 1,2, ... ).
(6) Solve (b and c are real numbers)
t
ty"(t) + 2(b+l)y'(t) + (c + 4)y(t) = o.
(y) For complex numbers a. 's and b. 's, show that the equation
J J
a 3t Y"'(t) + (a 2t+b 2)y"(t) + (a l t+bl)y' (t) + (aot+bO)y(t) = 0
converts into
3 2
Dy(-a 3s - a 2s - als - a O)
+ y((-3a 3+b )s 2 + (-2a 2+b )s + (-al+b ))

2 l O
(0) Under the initial conditions
yeO) =0 if ((2a 3-b 2)s + (az-b l )) # 0,

{
y' (0) 0 if arb2 # 0,
show that the third order differential equation in (y) converts into
2
l2l = _C-_3_a~3:...+_b 2::,)::-s_ _
+_C....,-,...2_a.::,2+_b"",l,-)_s_-_a-,l:--+_b.. :::.,O
y 3 2
a 3s + a s + als + a
2 O
REMARK 20.2. The method of integration explained above in this section
can be extended to the integration of the n-th order linear differential
equation with linear coefficients:
§2l. SUPPLEMENTS. I: WEIERSTRASS' POLYNOMIAL APPROXIMATION THEOREM.

II: MIKUSINSKI'S THEOREM OF MOMENTS
I. The following formulation and proof of the theorem of Weierstrass

is due to Bernstein. *
THEOREM 19. Let f(t) be a complex-valued continuous function defined
on the closed interval [0,1]. Then, defining polynomials
(21.1)
we have
lim max If(t) - P (t) I
n
O. (21. 2)
n-- O~t~l
PROOF: Differentiating
(21.3)
with respect to t and multiplying by t, we obtain
n-l n
nt(t+u) = L p c tpun- p . (21.4)
p=O n p
Similarly, differentiating (2.13) twice with respect to t and multiplying

by t 2 , we obtain
n(n-l)t 2 (t+u) n-2 (21. 5)
Thus, if we put
(21.6)
we have
n n
L r (t) = 1, L p·r (t) = nt
p=O p p=o p
(21. 7)
{ I p(p-l)rp (t) = n(n-l)t .
p=o
2
Hence
* Sergei Natanovic Bernstein (1880-1968).

21. Supplements 71
n 2
L (p-nt) r (t)
p=O P
2 2 n n n
=n t L r (t) - 2nt L pr (t) + L p2r (t)
p=O P p=O P p=O P
2 2 2
= n t - 2nt.nt + (nt + n(n-1)t ) = nt(l-t),
i. e. J
n
2
L (p-nt) r (t) = nt(l - t).
p
(21.8)
p=O
By the uniform continuity of the function f(t), there exists, for

any E > 0, a 0 > 0 such that
If(t) - f(t') I < E whenever It - t'l < o. (21. 9)

Also by the continuity of f(t), we have
max If(t) I = M < +~. (21.10)

O~t~l
Therefore, by (21.7), we obtain
We have, by r (t) > 0, (21.7) and (21.9),

p
n
< E L r (t) = E.
p=O P
We also have, by (21.8) and (21.10),
Ll ~ 2M L r (t)
IIp-ntl>on - Ip-ntl>on p
2M n 2 < 2Mt(1-t) M
~ 22 L (p-nt) r (t)
- n 0 p=O P no 2 ~ 202n .
The extreme right term tends to 0 as n .. ~, and thus
lim max If(t) - P (t)! ~ E.

n-- O~t~l n-
Since E> 0 was arbitrarily chosen, we have proved (21.2).

We shall now state and prove II.
72 VII • THE ALGEBRAIC DERIVATIVE
THEOREM 20. Let a complex-valued continuous function wet) defined on

the closed interval [O.T]. 0 < T < ~. satisfy
(n = 1.2.3 •... ), (21.11)
where M is a constant. Then wet) =0 on [O.T].

PROOF: Let k and x be natural numbers. and put w(T-u) g(u). Then.
by (21.11). we have
(k 1.2.3 •... ; x = 1.2.3 •... ).

Hence
(-1) k-l -kx(T-t)

k •' e
IT e
kx(T-u) ( )d
g u u
I
0
~
- k=l
I k~
.
e-kx(T-t)M ~
-
M(exp(e-x(T-t)) - 1),
and so. for any £ > 0 with £ < T. the extreme right term tends to 0
as x ~ ~ uniformly in t on 0 ~ t ~ T-£. Thus
lim
~
L (_l)k-l -kx(T-t)
k! e
IT e
kx(T-u) ( )d
g u u =0
x~ k=l o (21.12)
i uniformly in t (0 < t ~ T-£).
Since we have
lim ~ ..!.- =
L k'
0
n~ k=n .
and
o< e-kx(T-t) < (0 ~ t ~ T; k 1.2 •... ).
we obtain
L (-1) k-l -kx(T-t)

e
IT e
kx(T-u) ( )d
g u u
k .' 0
k=l
I
T ~ k-l
';' (-1) kx(t-u) ( )d
L k' e g u u
o k=l .
Io
T
(1 - exp(-e
x(t-u)
))g(u)du.
21. Supplements 73
This implies, by (21.12),
i IT (1 -x(t-u)
lim - exp(-e )g(u)du o
x-- 0
(21.13)
uniformly in t (0 < t ~ T - e:).
Since the integrand
(1 - exp(_ex(t-u)))g(u)
is continuous and bounded on the domain defined by
o~ t ~ T - e:, o< u ~ T, 0 < x,
we can exchange the order of the limit and the integral in (21.13). Hence
o
Io T lim
x--
(1 - exp(-e
x(t-u)
))g(u)du
I: {~: t
t > u}
< u g(u)du = 0 g(u)du
It (0 < t < T - e:).
We have thus obtained
t
fo g(u)du o (0 ~ t ~ T),
because of the arbitrary choice of e: (0 < e: < T). This proves that the
continuous function get) vanishes for all t € [O,T], and so wet) also
vanishes for all t € [O,T].
REMARK 21.1. The above proof is adapted from J. Mikusinski [51, p. 18. J.
Mikusinski and R. Nardzewski applied Theorem 22 in an elementary proof of
the Titchmarsh convolution theorem.
Part III
Shift Operator exp( - AS) and Diffusion
Operator exp( - As 1/2)
Chapter VIII
Exponential Hyperfunctions exp( - AS) and
exp( - As1/2)
§22. SHIFT OPERATOR exp(-AS) = e- AS . FUNCTION SPACE K = K[O,oo)

Let A~ o. We call the function HA(t) defined by
0, 0 ~ t < A
H (t) ={ - (22.1)
A 1, 0 ~ A< t < ~.
Heaviside's unit function. * We then define
Io
t ~ 0, 0 < t < A
hl(A,t) = HA(u)du = = (22.2)
t-A, 0 ~ A~ t < ~,
a continuous function of t ~ 0, containing a parameter A~ o.

We shall define **
(22.3)
and give
PROPOSITION 17. For f E C, we have
0, 0 < t < A )
e
-AS}
{f(t) = t A
= (22.4)
fo-
S
{ f(u)du, 0 ~ A~ t < 00
PROOF: We have, by (22.2) and Theorem 3,
*We may give any finite value for H,(A).

** The reason why for the notation e-AA S will be explained in §24. (22.3)
was first introduced by J. Mikusinski in the form hA = s2{h (A,t)}. See
his book [5], p. 181. l
74
22. Shift Operator exp(-As) = e -AS 75
S2{I: hI (A,t-U)f(U)dU}
0, 0 ~ t < A
S,S
t-A
{
I o (t-U-A) f(u)du,
0, o< t < A
'r
{
because we have, for

f(u)du,
A < t,
t-A d It-A
S 0 (t-u-A)f(u)du = dt 0 (t-u-A)f(u)du
I
t -A
+0 = (t-u-A)f(u)!U=t_A + 0
I d~(t-U-A)f(u)dU.
REMARK 22.1. Since hI (O,t) {t}, we have
e-ASIA=o = s2{t} = s{l} (22.5)
so that e- os = I.
PROPOSITION 17'. For A > 0 and \.I > 0, we have
PROOF: We have to prove
Since the ring CIC is commutative, we have, similarly as in (22.4),

2 2
S {hl(A,t)}s {hl(\.I,t)}
t t A
s2 s 2{I hI (A,t-U)hl(\.I,U)dU} = s2 s2{I - (t-U-A)(U-\.I)dU}
o \.I
s2s{ (t-U-A) (u-\.I)IU=t-A + It-A(U_\.I)dU}
\.I
t
s2 s{I -A (u-\.I)du = s2{t_A_\.I}} when 0 ~ A+\.I < t,
\.I
and
1/2
76 VIII. EXPONENTIAL HYPERFUNCTIONS exp(-As) AND exp(-As )
because
hl(A,t-u)hl(~'u) > 0 only if both 0 < A < t-u and 0 < ~ < u
are satisfied.
REMARK 22.2. By (22.4) and by the operation of differentiation s, we

see that the operation of e->"s may be interpreted as follows:
~
o
J~--:-
o
_
That is, the operator e->"s gives a shift or translation of the graph of
{f(t)} to the right by length >... We shall thus represent e->"s{f(t)}
by the function
0,
{ f(t-A) ,
which might be discontinuous at t = >...

In order to introduce the above mentioned representation, we shall
consider the class K of functions. *
THE CLASS K OF FUNCTIONS. Let K = K[O,"') be the totality of

complex-valued functions f(t) defined on [0,"') and satisfying the
following conditions:
In every finite interval in [O,~), {f(t)} has only a
(22.7)
finite number of points at which {f(t)} is discontinuous;
For every t > 0, f:1f(u)ldU < "'. (22.8)
Thus, if f € K, the convolution product
hf = {l}{f(t)} = {I: f(U)dU}
*J. Mikusinski [5], p. 106.

22. Shift Operator exp(-Asl e
-AS
77
exists and hf E C. Moreover, for f and g € K, the convolution
(fg)(t) = {f: f(t-u)g(U)dU}

exists and fg E K. As in the case of C, K constitutes a commutative
ring. Of course, like HA(t), a function in K may have a jump at its
point of discontinuity. Thus for two functions f and g, we regard f
to be equal to g as functions of K if f(t) - get) vanishes at every
point to at which both f and g are continuous. Moreover, for f
and g € K, we define the sum of f and g through
(f + g)(t) = f(t) + get)
at every point t where both f and g are continuous. Finally, for

a
f € K and {get)} {fat f(u)du}, we have
sg = g' + sg(O) = {f(t)},
the equality in this equation meaning equality at every point t where
a
{f(t)} is continuous.
EXAMPLE 22.1. For {h (A,t)} of (22.2), we have
1
(22.9)
REMARK 22.3. By (22.3), we can rewrite (22.9) as
(22.9) ,
EXERCISES FOR §22. Verify the following equalities when a< A< v:
(i) s
-1
(e
-AS "S
- e-'" ) = {1,a, A< t
otherwise
< v}
(ii) s
-3
(e
REMARK 22.4. We shall identify

-AS
- e
-vs )
f = {f(t)} E K with hf E C so that

h H
J
the function {f(t)} € K may be regarded as a concrete representation
of the hyperfunction h: E C .
H
AN APPLICATION OF K TO PHYSICS. The so-called impulsive force in
mechanics can be given the following representation as a hyperfunction.
Let F and € be positive numbers. If an external force of strength
78 VIII. EXPONENTIAL HYPERFUNCTIONS exp(-As) AND eXP(_As 1/ 2 )
F
£ is applied to a particle with unit mass during the time interval between
t = ° and t £, then we say that, during this time interval, the impulse
f x = F = J£ f du
£
£
° £
is applied to the particle. We notice that the impulse is equal to the

convolution
where f (t) = {
£
f,0,
The so-called impulsive force is interpreted as a limit of the hyper-
function f£ when £ tends to 0. Thus
the impulsive force [F] = [F]I = s{F},
because, for any g € C,
(0 < £ < t)
so that
lim g*f {Fg(t)} = [F] times {get)}.

£+0 £
Thus, as £ + 0, the hyperfunction f£ tends to the hyperfunction

[F]I. Physicists call the delta function of Paul Adrian Maurice
Dirac (1902- ).
AN EXAMPLE OF THE IMPULSIVE FORCE. Consider the equation

2
y"(t) + 2ay' (t) + b yet) = f; yeO) = 0, y' (0) = 0,
where the external force f is given by a hyperfunction [F]I. As in

the Example 10.6, we have
2 2
s y + 2asy + b y = [F]I.
Hence
y
[F)I [F]I
s2+2as+b 2 (s+a-iw) (s+a+iw)
_ ill. (_I __ I )
- 2iw s+a-iw s+a+iw
23. Hyperfunction-valued Function f(A) 79
§23. HYPERFUNCTION-VALUED FUNCTION f(A) AND GENERALIZED DERIVATIVE

it f(A) = f' (A) .
OPERATOR-VALUED FUNCTION OR HYPERFUNCTION-VALUED FUNCTION. Let, for

each fixed value A of the parameter A, there be given an operator
O
(= hyperfunction) f(A O) E C/C. Then we shall call fCA) an operator-
valued function or a hyperfunction-valued function.
The shift operator
(A > 0)
is such an example, defined for A > 0 if we define
THE GENERALIZED DERIVATIVE. * f'CA) = dAd f(A) of an operator-valued

function f(A) is defined as follows:
Let, for a certain ~ F 0 of C/C,
~ f(A) be equal to {f(A,t)}

a
where {f(A,t)} is a complex-valued function of {A,t} in such a way

that, on the domain
both f(A,t) and ax f(A,t)

A () A
are continuous functions. Then we say that

the generalized derivative
f' (A) = d~ f(A)

exists for Al < A < A2 and is represented by
fl(A) = dAd f(A)

a
= ba {ax f(A,t)}. (23.1)
REMARK 23.1. The above definition of the generalized derivative is rea-

sonable, i.e., without contradiction. We have to prove the following:
If we assume that we have another couple
iF
c
0 of C/C and ?(A,t)
for which
* J. Mikusinski [5], p. 183.

1 2
80 VIII. EXPONENTIAL HYPERFUNCTIONS exp(-As) AND exp(_As / )
both ~ f{A) = {~{A,t)} and

a :: are
{ar f{A,t)}
{ contin:ous functions of {A,t} on V,
then we can prove that

aaA ca:::
~arf{A,t)} = ~ar f{A,t)}. (23.2)
PROOF: Multiplying the equality
E-<f{A,t)} = ~d{A,t)} = f{A)
by bd E C, we obtain
ad{f{A,t)} = bc{1(A,t)} = bdf(A);
that is,
t It
I° ad{t-u)f{A,U)du = ° bc{t-u)f{A,u)du.
A A
Differentiating with respect to A, we have
t It
I° ad{t-u)aA f(A,U)du = ° bc(t-u)ar f(A,U)du
a a A
i. e. ,
a A
ad{ar f{A,t)}
a ::
= bc{ar f{A,t)}.
Multiplying by b~' we obtain (23.2).

As an application of the above definition, we prove
PROPOSITION 18. For f{A) = s2{h (A,t)}, we obtain
l
fl (A) = -sf (A) (A > 0), frO) = 1. (23.3)
PROOF: We have
= h{hl(A,t)}
= h{O' ° ~ A } {O' ° A <
t-A, ° < A ~ t = {t-A)2~2, °

t < < t
< A~
Hence
I a 3
fl (A) = ~{ar =s
A
f{A,t)} {-hl(A,t)}
h
2
-s'S {hl{A,t)} = -Sf{A)
23. Hyperfunction-Va1ued Function f(A) 81
and
f(O) = s 2{hl(O,t)} = s 2{t} = s{l} =
REMARK 23.2. In the above, {f(A,t)} is defined and continuous on
{{A,t}; 0 ~ A, 0 ~ t} so that
f(A) = s 3 {f(A,t)}
"
is continuous wi th respect to A in generalized sense. This means that,

3
by multiplying f(A) by h # 0 of CIC, we obtain {f(A,t)} which is
continuous in A. Thus, for any fixed A ~O,
O
generalized lim f(A) * =~ {lim h 3 f(A)}
A~AO h A~AO
(23.4)
3" 3 "
=s {lim f(A,t)} =s {f(AO,t)} = f(AO)'
A~AO
Hence f(A) is continuous at every A ~ 0 in the generalized sense.

o
We next give a useful proposition regarding the generalized derivative.
**
PROPOSITION 19. (i). If f(A) has a generalized derivative for

A E (A l ,A 2), then, for any ~ # 0 of CIC, ~ f(A) has the generalized
derivative for A E (A ,A ) given by
l 2
~(~ f(A)) = ~ df(A) = ~ f' (A). (23.5)
dA P P dA P
(ii). If both f(A) and g(A) have generalized derivatives for
\ E (\1'\2)' then f(\) ± g(\) also have generalized derivatives for
A E (A l ,A 2) and
(f(A) ± g(A)) 1 fl (A) ± g' (A) • (23.6)

(iii). If both f(A) and g(A) have generalized derivatives for
A E (A l ,A 2), then f(A)g(A) also has a generalized derivative and
[f(A)g(A)) 1 = fl (A)g(A) + f(A)g' (A). (23.7)
*Let, for a certain ~ # 0 of CIC, ~ f(A) be equal to {f(A,t)} in such

a way that ~i~ f(A,t) exists and is in C[O,=) as a function of t.
Then ~{~i~o f(~,t)} is, by definition, the generalized limit of f(A)
as A ~ A ' That the value of the generalized limit of f(A) as A ~ A
O O
is independent of the choice of ~ shall be proved similarly as the case
a
of the generalized derivative; see §25.
** J. Mikusinski [5], p. 185-.
82 VIII. EXPONENTIAL HYPERFUNCTIONS exp(-As) AND exp(_As 1/ 2 )
(iv). f(A) has the generalized derivative f'CA) 0 for A E (AI,AZ)

i f and only if
f(A) is equal to a fixed hyperfunction c E CIC. (23.8)
PROOF: (i) Let both fi(A, t)} = £. f(A) (£.a -f 0) and a"
{"IT f(A,t)} be
a
continuous on V = {{A,t}; Al < A < A2 , 0 ~ t}:
a a"f(A, t) }.
f' (A) = j){"IT
Then, by
£q £.a (~f(A))
p
= £.a f(A) = {f(A,t)},
we see that ~ f(A) has a generalized derivative and
p
(; f(A))' =~ {;A f(A,t)} = ;(~ {a~ f(A,t)}) =~ f'CA).

"
(ii). Let {f(A,t)} = -b f(A) with -b -f 0 and "
{g(A,t)} = -d g(A)
d a a c
with C -f 0 such that
fl (A) = S{;A f(A,t)} and g'(A) a"g(A,t)}.

= dc {ar
Then we have
f(A)
a "
= ~f(A,t)} = bdI v
{f(A,t)}
with
v
f(A,t) = ((ad)(t-u)f(A,u)du
t" and f'(A) = bdI {aAa v
f(A,t)},
JO
and
c " v
g(A) = ~g(A,t)} = bdI {g(A,t)}
t
with v
g(A,t) "
= 0(bc) (t-u)g(A,u)du and g'(A) = b~ {;A g(A,t)}.
J
Therefore
I v v
f(A) ± g(A) = bd {f(A,t) ± g(A,t)}
and
(f(A) ± g(A))' = b~ {;A f(A,t) ± ;A g(A,t)} f'(A) + g'(A).
(iii). Let, as above,

I
f(A) = bd {f(A,t)}
v
and
I a
f'CA) = bd {"IT f(A,t)}
v
and
24. Exponential Hyperfunction exp(As) = e
AS
83
I
g(A) = bd {g(A,t)}
v
and g' (A) =
I
lid a v }
{ax g(A, t) .
Then
and so
(f(A)g(A))' = ---1---2 {;A Jt f(A,t-u)g(A,U)dU}

(bd) 0
=~ {Jt g(A,U);A f(A,t-u)du

(bd) 0
+ J: f(A,t-U);A g(A,u)du }
I a v v I v av
= ------2 {aA f(A,t)}{g(A,t)}+ ------2 {f(A,t)}{aA g(A,t)}
(bd) (bd)
= fl (A) g(A) + f(A) g' (A) .
(iv) Let f(A) =c E CIC for A E (A 1 ,A 2). Then
c =-ba with b E C, a E C.
Hence c = ab = hah {b(t)}, and so, taking the generalized derivative, we

obtain
dc h { d Le., fl (A) = O.
dA = ha ax bet)} = 0,
Conversely, let f'CA) O. Then we have

a ,., df(A,t) _ 0
f(A) =b {f(A,t)} and dA -.
Hence {f(A,t)} does not depend upon A, so that
f(A,t) = {f(t)}, i.e., f(A) = F{f(t)} E CIC.

§24. EXPONENTIAL HYPERFUNCT10N exp(As) e AS (_00 < A < 00).
We recall that the shift operator
was denoted by
exp(-As) = e -AS
84 VIII. EXPONENTIAL HYPERFUNCTIONS exp(-AS) AND exp(_As 1/ 2 )
-AS {hl(A,t)}
~
l e
e
os
= exp(-AS) =
= I.
h
2 (A ~ 0),
(24.1)
We shall define e AS (A > 0) as follows.

PROPOSITION 20. Since {hl(A,t)} # 0 for any A > 0, there exists, by
(18.6), the (multiplicative) inverse operator of e -AS given by
2
-AS -1 h I (24.2)
(e ) = {hl(A,t)} = e- AS € CIC (A ~ 0).
Hence, defining
(A > 0), (24.2) ,
we obtain
(A > 0) (24.3)
and furthermore
(_00 < A,IJ < 00). (24.4)
PROOF OF (24.4): When A ~ 0 and IJ ~ 0, (24.4) holds by (22.6) and

eOs I. Hence, by (24.3), we have (24.4) when A ~ 0 and IJ ~ O. There-
fore, when A ~ IJ ~ 0,
Thus, by (24.3),
when o~ A ~ IJ as well.
THEOREM 21. f(A) = exp(As) = e AS (_00 < A < 00) does not equal zero for
any value of A, and
Jf' (A) = sf(A) (_00 < A < 00), f(O) = I,
1 generalized lim f(A)

A"'A O
= f(A O) for _00 < A < 00.
O
(24.5)
Conversely, any hyperfunction-valued function f(A) satisfying (24.5) is

AS
equal to e
PROOF: That e AS # 0 is clear from (24.3). Next, for any A, we take
a with a> A. Then, by (24.4),
24. Exponential Hyperfunction exp(As) = e AS 85
Since a-A> 0,
de-(a-A)s - (a-A) s
-se (by (23.3)),
dCa-A)
i.e. ,
de-(a-A)s de-(a-A)s dCa-A) _ -(a-A)s

dA dCa-A) ~ - se
as
Thus, since e F 0, we can apply (i) of Proposition 19, obtaining
de AS _ as de-(a-A)s
"""'dr" - e dA
eaSse-(a-A)S = seAS
Similarly, for any A ' we take a with a > AO and, by (24.6),

O
(generalized e
-(a-A)s )
(by (23.4).
In addition, we know already that
OS
e ((23.3)).
Therefore, (24.5) is completely proved.
THE PROOF OF THE CONVERSE PART OF THE THEOREM. Assume that f(A) satis-
fies (24.5). We put
AS
f(A) - e = X(A)
and shall prove that X(A) 0 for every A. By (ii) of Proposition 19,
we have
x' (Al f' (A) - (e AS ) , sf(A) - seAS SX(A). (24.7)
Moreover,
s
X(O) = f(O) - eO ' =I - I O. (24.8)
Now we shall prove that X(A) =0 from (24.7)-(24.8). To this purpose,
take any real number ~ and put, as in Miksinski [5], p. 191,
YeA) = x(A)x(2~-A). (24.9)

By (iii) of Proposition 19, we obtain
y' (A) x' (A)x(2~-A) X(A)X' (2~-A)
sX(A)x(2~-A) sX(A)x(2~-A) O.
1 2
86 VIII. EXPONENTIAL HYPERFUNCTIONS exp(-As) AND eXP(_As / )
b
l
Hence, by (iv) of Proposition 19, there exists an element
a of CIC
l
such that
yeA) = x(A)x(2~-A) =~
a
for all A.
l
Hence, taking A = 0, we obtain
yeO) = x(0)x(2~) o since x(O) O.

b
This implies that -l= 0, whence
al
yeA) = x(A)x(2~-A) = 0 for all A. (24.10)
Therefore, by putting ~ A, we have
X(A)X(A) =0 for all A.

X(A) being in CIC, we have X(A) b and so
a
2
b
2" = 0, (24.11)
a
b E C and a E C being dependent on the parameter
We must have A.
a # 0 since it is the denominator of ~. Hence, by Theorem 16, a 2 # O.
2 a
Thus by (24.11), b = O. Again by Theorem 16, b = O.
This proves that X(A) = ~ = O.
a
REMARK 24.1. By virtue of the uniqueness, proved above, of the solution
f(A) of (24.5), we are justified in denoting f(A) by exp(As) = e AS
COROLLARY OF THEOREM 21. By taking the generalized derivative of f'CA)
sf(A), we obtain
2
fit (A) = (fl (A) ) , sf' (A) s f(A),
and, more generally,
AS
for f(A) e (24.12)
AS
§25. EXAMPLES OF GENERALIZED LIMIT. POWER SERIES IN e
The reader is referred to Remark 23.2.
PROPOSITION 21. Let

{f E CIC; n = 1,2, ... }
n
be a sequence of hyperfunctions. For a certain ~
a
# 0 of CIC, let there
25. Examples of Generalized Limit 87
exist a sequence of functions (n 1,2, ... ) and a function

{f(t)} € C in such a way that
~ f n = {fn(t)} and lim fn(t) = f(t) uniformly on

n-- (25.1)
every finite closed interval [O,T], 0< T <~.
Then we shall call
(25.2)
the generalized limit of the sequence {f € CIC; n = 1,2, ... }.

b n
PROOF OF THE CONSISTENCY OF THE ABOVE DEFINITION: For another -l
b v
a
1
# of °
CIC, let there exist a sequence of functions --l f = {f (t)} € C
v al n n V
V
(n = 1,2, ... ) and a function {f(t)} € C in such a way that lim fn(t) f(t)
n--
uniformly on every finite closed interval [O,T]. Then we can prove
a V
= ~l {f(t)}.
A
~ {f(t)}
1
In fact, we have
a l
V
f =~(f(t)} --b {f (t)},
nun 1 n
i.e. ,
Thus, by letting n ~ ~, we obtain
t It
I°(ab l ) (t-u)f(u)du = °(alb) (t-u)f(u)du,
A V
i.e. ,
Therefore, by multiplying by __1__ , we have

bb l
a A al V
b f = ~ f.
1
88 VIII. EXPONENTIAL HYPERFUNCTIONS exp(-AS) AND eXP(_As 1 / 2 )
EXAMPLES OF GENERALIZED LIMIT.

EXAMPLE 25.1. For f = {f(t)} E C. we have
fl = f. f2 f*f ..... fl = f*fl- l = fl-l*f (n 1.2 .... ; fO = I).
If f denotes fl. we obtain

n
generalized lim f = O. (25.3)
n-- n
PROOF: Choose T > 0 and let 0 < t ~ T. Then max If(t)1 ~ M for some
O~t~T
finite M > O. Assuming
(O<t~T; n=1.2 .....k) (25.4)
we obtain
(0 ~ t ~ T)
Hence (25.4) holds for all positive integer k.

Therefore. if we take a positive integer nO satisfying
MT/n o ~ 1/2.
then we obtain for k > nO'

n n
O
1~+l(t)1 ~ M °T , (MT)k+l-n o x ~
- nO' nO T
nO nO-l
<M T x (l)k+l-no (0 ~_ t <
__ T)
= nO! 2
This proves (25.3). In this example. we have made use of ab = I.

EXAMPLE 25.2. For f = {sin nt}. we have
n
Hence
generalized lim {sin nt} = s{lim l-C~S nt} = s{O} O. (25.5)

n.- n.-
EXAMPLE 25.3. For f
n
(~ - I) = s 2{h
n
n - h2} • we have
2 h 2
generalized lim (~ - I)
n
s {lim (-
n
- h )} = s 2 (-h)
2
= -I. (25.6)
n_
EXAMPLE 25.4. For fn = (ns - I) -1 = sin I _ =

nI
s:nI
generalized lim {nent } does not exist. (25.7)
* *
n-
nt
PROOF: Assume that # 0 of CIC exists such that {ne }
{f (t)} E C satisfies
n
lim f (t) = f(t) E C exists uniformly on [O,T]
n
n- (25.8)
for every positive T < w.
From this we shall derive a contradiction, as follows. (25.8) implies
that the sequence of functions
converges uniformly on every finite interval [O,T] as n + w. Hence,

by taking t =T > 0, we see that
a finite lim n IT b(T_u)enudu must exist.

n- 0
Thus there must exist a constant My such that

T
_ sup 1f b(T-u)enuduI ~ My < w. (25.9)
n-l,2,... 0
Therefore, by Mikusinski's theorem of moments (Theorem 20), we must have

b(T-u) = 0 for all u in [O,T], i.e., bet) :: 0 in [O,T].
Since T> 0 was arbitrarily chosen, we must have bet) :: 0 in [O,w).
This contradicts the assumption ~ # O.
a
EXAMPLE 25.5. Let {r(t)} € C satisfy f(t) > 0 and J~ feu) du = 1.
Then
generalized lim {nf(nt)} 1. (25.10)
n-
PROOF:
{nf(nt)} = s 2h 2{nf(nt)} = s 2h{[nt

0 f(v)dv }
On the other hand, we have, for every T > 0,

EXPONENTIAL HYPERFUNCTIONS exp(-As) AND exp(_As / 2)
1
90 VIII.
~:: f: f: du
u
f(v)dv =t uniformly on [O,T).
In fact, by f~ f(v)dv = 1,
nt
lim [t - It dt I f(V)dV]
n-- 0 0
This proves (25.10).

EXAMPLE 25.6. For any complex number B and any positive number A,
we have
-AS -1
(I - Be ) = s 2{ fB(A,t)} (25.11)
where the complex-valued function fB(A) = {fB(A,t)} is given as follows:
For T > A, let k be the positive integer satisfying
kA ~ T < (k+l)A. Then {fB(A,t)} = h2 + B{hl(A,t)} (25.12)
+ B2{hl(ZA,t)} + ... + Bk {hI (kA,t)}, where 0 ~ t < T.
PROOF: Let T ~ A. Then, for 0 ~ t ~ T,
(kA ~ T < (k+l)A; k = 1,2, ... ).
On the other hand,

2 -ZAs k -kAs
B e)
+ Be + ... +
+ •••
and
generalized lim Bne- nAs = s2{lim Bnh2e-nAs}

n-- n--
= s2{lim Bnhl(nA,t)} = s20 = 0,

n--
because, for 0 < t < T < (k+l)A,
hI (nA,t) = 0 if n> (k+l).

Therefore, we obtain
generalized lim (I + Be-As + ... + Bke- kAs )

k....
I + Be-As + + Bke- kAs + '" (25.13)
(A > 0).
REMARK 25.1. The existence of the multiplicative inverse (I _ Be-As)-l

-AS -AS
of (I - Be ) E CIC is proved directly from (I - Be ) # O. This
inequality is clear when B = OJ and if B # 0, then the equality
I = Be-As implies the contradiction B- l h 2 = h 2e- As = {hl(A,t)} (A> 0).
EXAMPLE 25.7. Let A> 0 and let f = {f(t)} € K[O,~) vanish for
A < t <~. Then
{get)} = (I_e-As)-lf
(25.14)
f + e-ASf + e-ZAsf +
is periodic with period A, i.e.,
get + A) = g(t).
PROOF: As was discussed in §22, we have
e
-kAS
f = {O' 0 ~ t < kA }
€ K.
f(t-kA), 0 < kA < t
REMARK 25.2. If g € K[O,~) is periodic with period A > 0, then g

can be represented as
-AS -1 -AS ) g € K[0 ,~)

g = (I - e ) f, where f = (I - e
(25.15)
{ and f(t) = 0 for A< t < ~.
A CONCRETE EXAMPLE OF (22.15). Let A> 0 and put

-AS
= {f(t)} = h 2 2
:}.
f - e (h +Ah)
t < A } _ ( t, 0 < t <

= {t} - {O' =
0 <
t+A-A, 0 < A~ t - \ 0, 0 < A<
Thus
f = -l.2 _(-l.2 lli)e -AS

+
S '
S S
(25.16)
{ -AS ) -1 f
g = (I - e
92 VIII. EXPONENTIAL HYPERFUNCTIONS exp(-As) AND exp(_As 1/ 2 )
gives an example of (22.15). (See the saw-toothed graph below.)
,) i
4+-,
1<"
5.
-AS
§26. f
00
Oe f(A)dA = {f(t)} FOR {f(t)} E C. Let F(A) be a hyperfunc-
tion-valued function (AI < A < A ). Then the generalized integral

2
A
2
fA F(A)dA is defined as follows. For a certain ~ I 0 of CIC,
l
~ F(A) = {F(A,t)}} is a complex-valued continuous function defined on
V = {{A,t}: Al < A< A
2
and 0 ~ t } such that
A
2
fA F(A,t)dA E C as a function of t. Then, by definition, we have
l
A A
f 2 F(A)dA = ~ {f 2 F(A,t)dA} (26.1)
Al Al
REMARK 26.1. The above definition of the §eneralized integral is reason-
able in the following sense: If another -l I 0 of CIC exists in such
a
~ b 1
l
a way that {F(A,t)} =--
a
F(A) is continuous on V and
l
{fA
2
~(A,t)dA} E C, then we have
Al
~ {fA 2 F(A,t)dA} = :1 {fA 2 ~(A,t)dA}. (26.2)
Al 1 Al
PROOF OF (26.2): From
a a ~
=b = ~1
A
F(A) {F(A,t)} {F(A,t)}

1
we obtain
26. f~ e-ASf(A)dA = {f(t)} For {f(t)} E C 93
o
and so, by integrating,
Thus
2
ab l {fA F(A,t)dA} = alb {fA 2 ~(A,t)dA}
Al Al
and, by multiplying by bIb' we obtain (26.2).

1
THEOREM 22. Let {f(t)} E C. Then
(26.3)
PROOF: For 0 ~ A< ~ and 0 ~ t < ~, we have
2
s {hI (A,t)f(A)}
Hence, by Theorem 2,
J: e-Asf(A)dA = s2{J: hI (A,t)f(A)dA}
s2{f:(t-A)f(A)dA}
s{(t-t)f(t) + f: a(~~A) f(A)dA + f:(O-A)f(A)dA}
s{f: f(A)dA} = {f(t) + f: f(A)dA} = {f(t)}.
~ -AS
REMARK 26.2. The above integral f e f(A)dA might be called the
O
generalized Laplace integral of f(A). In Heaviside's operational calcu-
Ius, * the operator of differentiation was denoted by p. The interpre-
tation of the operators like p as given by Heaviside is difficult to
justify, and the range of the validity of his calculus remains unclear
even now, although it was widely noticed that his calculus gives correct
*Oliver Heaviside: Electromagnetic Theory, ~ondon, 1-3 (1893-1899).

1 2
94 VIII. EXPONENTIAL HYFERFUNCTIONS exp(-As) AND eXP(_As / )
results in general. * Thus G. Doetsch ** and many other mathematicians

strove for the foundation of Heaviside's calculus by virtue of the theory
of the Laplace transform
( e-Ptf(t)dt or
However, the use of such integrals naturally confronts restriction due to

growth behavior of the numerical function f(t) as t ~~. See Remark
10.2.
2
§27. EXPONENTIAL HYPERFUNCTION exp(_As l / 2) = e- Asl / (_00 < A < 00).
PROPOSITION 22. The hyperfunction-valued function f(A) given by
i f(A,t) = ---e
A
2I~
-A 2/4t (A > 0, t > 0),
(27.1)
l f(A,O) o (A > 0),
satisfies
i
d
dA{f(A,t)}
(27.2)
generalized lim {f(A,t)} = I.
HO
h 1/2 {f(A,t)} = {It0 21f(t-u)

A -1/2 u -3/2 exp(-A 2/4u)du} .
By the substitution
we change integration variables, from u to o. Then
(t _ l;l(i + 4~~))-1/2(1~(i + 4~~))-3/2

(( 1 + 4~~t _ 1 )/(t 4~~)+ r l (t 4~~)r
1/\ +
3 2
/
*In 1921, Heaviside received the first Faraday Medal of the Institution
of Electrical Engineers of Great Britain.
** ..
Gustav Doetsch (1892-1977); Einfuhrung in Theorie und Anwendung der
Laplace-Transformation, Birkhauser Verlag (1958).
1/2 _As l / 2
27. Exponential Hyperfunction exp(-As ) =e (~ < A< ~) 95
2
du = - ( 80
A2 (1t + 740 )2) do.
Hence
2
h
1/2 (f(A,t)} =
fO - 211
A
=mA 80
2" (A2) _0
exp - 4t e do
00 20t A
_0 2
= 1"/2
00
and so, by f e do ((12.8)), we obtain
O
(27.3)
Thus
(27.4)
Hence the generalized derivative ~f(A,t)} is given by
d
ifif(A,t)} 5
3/2{ft ax
Cl - 1 exp(-A 2/4u)du }
o liiU
5
1/2 5 {ft -
-1 A exp(-A 2/4u)du }
2u
o I1fU
-5 1/2{ f(A,t) + fO A
~ exp(-A 2 }
/4u)du
o 211!u
Furthermore, by (27.4) and (13.3),
generalized lim f(A)

>.+0
lim {f(A,t)} s3/2{lim ft __1__ eXP (-A 2/4U)dU}

>.+0 >.+0 0 liiU
s3/2{ft __1__ dU} = s3/2. h1/2. h = s3/2· h 3/2 I.
o I1iU
96 VIII. EXPONENTIAL HYPERFUNCTIONS exp(-As) AND exp(_As 1 / 2 )
Therefore, by putting f(O) = I, we have
fICA) = _sl/2 f (A) (A > 0), f(O) = I, and so

(27.2) ,
{ generalized lim
A+A
O
DEFINITION 27.1. * We shall denote
1 2
e _As / {A 2}
~ exp(-A /4t) (A > 0),
2/wt 3
(27.5)
(A = 0).
Then, by (27.2)',
d 1/2 1/2 1/2
dA exp(-As ) = -s exp(-As ) (A > 0),
(27.6)
{ generalized lim exp(_As 1/ 2) = I = exp(_0.sl/2).
A+O
PROPOSITION 23. If ~ > A > 0, then
exp(-As 1/ 2)exp(- (~_A)sl/2) = exp(_~sl/2). (27.7)

PROOF: Putting f(A) exp(_As 1/ 2), we shall prove
f(A)f(~-A) = f(~).
By (iii) of Proposition 19 (§23), we have
d d
dA(f(A)f(~-A)) f'(A)f(~-A) + f(A)dA f(~-A)
_sl/2f(A)f(~_A) + f(A)sl/2f(~_A) 0
because
d d d(~-A)
dA f(~-A) = d(~-A) f(~-A)· ~ .
Hence, by (iv) of Proposition 19, we have
f(A)f(~-A) =a
b
E C/C
so that ab does not depend on A. Thus

f(A)f(~-A) = generalized lim f(A)f(~-A)
A+O
= If(~) = f(~).
* Due to J. Mikusinski [5], p. 220.
1/2 l/ 2
27. Exponential Hyperfunction exp(-As ) = e _As (~ < A< ~) 97
DEFINITION 27.2. We shall denote
exp(As l / 2) = exp(_Asl/2)-1 = ~I~~ (A ~ 0). (27.8)

exp( _As l / 2)
The existence of the (multiplicative) inverse exp(_As l / 2)-1 of

1/2
exp(-As ) for A > 0 is proved as follows:
Since s3/2 € C ((13.9)). we have. by (27.1). (27.4) and (27.5).
H
t
exp(-As 1/2 ) = s 3/2{I 1 exp(-A 2/4u)du} € C/C
--- (A > 0).
o IifU
2
Moreover. this expression shows that exp(_As l / ) F0 for every A > O.
Indeed. if it were otherwise. then by (27.4).
h3/2 exp(-As 1/2 ) = {It 11/2 exp(-A 2 }

/4u)du =0
o ('IIu)
for all t > O. We shall use Proposition 12.
Thus. if A > O. there exists the (multiplicative) inverse
exp(_As l / 2) in the ring C/C.

1/2
We shall denote. by using exp(-Os ) = I.
1/2 -1 _ 1/2
exp(-As ) = 1/2 - exp(As ) (A ~ 0).
exp (-As )
We have thus proved (27.8) and
(27.8)'
THEOREM 23. We have
1/2 1/2
exp(As )exp(~s ) = exp((A+~)S 1/2 )
(27.9)
(-~ < A. ~ < +~)
and
i
1/2
d
~Xp(AS ) = s 1/2 exp(As 1/2 ) (-~ < A < -).
(27.10)
1/2 1/2
generali zed lim exp(As ) = exp(AOs ) (-~ < A < ~).
A+A
O
O
We also have
n
d 1/2
--- exp(As )
dAn (27.10)'
(-~ < A <~; n = 1.2 •... )
98 VIII. EXPONENTIAL HYPERFUNCTIONS exp(-AS) AND eXp(_As 1/ 2 )
and
1/2 (_GO < A < 00).
exp(As ) ~ 0 (27.11)
PROOF: (27.9) is a consequence of Proposition 23 and (27.8)'. The reader
may follow the proof of (24.4).
(27.10) is also a consequence of (27.2)' and (27.9). The reader may
follow the proof of (24.5).
Finally, (27.11) is a consequence of (27.8)'.
REMARK 27.1. We can prove that any solution f(A) of
dfd~) = sl/2 f (A) (-GO < A< GO), f(O) = I,
1
(27.12)
generalized lim f(A) = f(A O)
A....A
O
must coincide with exp(As 1/ 2). The reader may follow the proof of the
converse part of Theorem 21 (§24).
COROLLARY OF PROPOSITION 22. We have
I exp(-As 1/2 ) = {I
-- --- exp(-A 2 },
/4t) (27.13)
IS lift
A} *
-I exp(-As 1/2 ) = {Cerf--- (27.14)
5 2It"
PROOF: (27.13) is already proved, by (27.1), (27.3), and (27.5). Next,

we obtain from (27.5)
5I exp(-As 1/2 ) h exp(-AS

1/2
)
{Io t A
2/1N3
eXP(-A 2/4U)dU}.
A We have thus
Then perform the change of the variable = ---.
proved (27.14). Uu
REMARK 27.2. For f(A) exp(-As 1/2 ), we have
f' (~ _sl/2 f (A),

fll(A) (f' (A))' _sl/2 f , (A) sl/2 s 1/2 f (A)
sf(A) .
* Cerf t 1 - Erf t, Erf t

28. LOgarithmic Hyperfunction w 99
Since s is the operator of differentiation with respect to t, exp(_As l/2 )

is intimately related to the heat equation
ZAA = Zt'
Thus exp(_As l/2 ) is sometimes called the heat operator or the diffusion
operator.
§28. LOGARITHMIC HYPERFUNCTION w AND EXPONENTIAL HYPERFUNCTION exp(Aw)

DEFINITION 28.1. Let a hyperfunction-valued function f(A) satisfy
f'CA) = wf(A) (-~ < A< ~), f(O) = I, (28.1)

where w € CIC. Then we call w a logarithmic hyperfunction and f(A) an
exponential hyperfunction.
EXAMPLES. sand sl/2 are logarithmic hyperfunctions.
PROPOSITION 24. Let w be a logarithmic hyperfunction. Then the solution

f(A) of (28.1) is uniquely determined, and we have
generalized lim f(A) = f(A O) (28.2)

A~AO
f(A) F a (-~ < A< ~), (28.3)
f(A)f(~) (-~ < A, ~ < ~). (28.4)
PROOF: The uniqueness of the solution f(A) of (28.1) is proved simi-

larly as in the converse part of Theorem 21 (§24). *
PROOF OF (28.2): The assumption that f(A) has the generalized derivative
f'(A) implies that f(A) is continuous in the generalized sense (see
Remark 23.2) so that (28.2) is true.
PROOF OF (28.3): Assume that f(A O) a for a certain A'
O
Then
f(A) = f(A+A ) is a solution of
O
= O.
A
~I(A) wf(A) (-~ < A < ~), f(O)

Hence the proof of the converse part of Theorem 21 can be modified to
=a = O.
A
prove that f(A) so that f(A)

PROOF OF (28.4): The proof of (27.9) can be modified to show that (28.4)
is true.
COROLLARY. Because of the uniqueness of the solution f(A) of (28.1),

we are justified in denoting it by exp(Aw).
* In the proof, we make use of (28.2).
100 VIII. EXPONENTIAL HYPERFUNCTIONS exp(-As) AND exp(_As l / 2
)
PROPOSITION 25. w = is = ;:r s E CIC is not a logarithmic hyperfunction.

PROOF: * Assume that there exists an f(A) which satisfies
f'(A) = is f(A) = ;:r s f(A) (-~ < A< ~), f(O) = I. (28.5)
Then there exist a closed interval [A l ,A 2] (with Al < 0 < A2) together
with an element ~ 1 0 of CIC such that
both {f(A,t)} = f(A) = * f(A) and a"

{'IT fCA,t)}
{ are continuous on V = {{A,t};
and so the generalized derivative
fl (A) = ba "
f'(A)
a"f(A,t)}
= ba {at is f(A)
= is 0a "f(A) = is ba {f(A,t)}.
"
Next let us put
Then, by f'CA) = is f(A), we have
JA 2 e-inwAf'(A)dA = is JA 2 e-inwAf(A)dA (n = 1,2, ... ). (28.6)

\ Al
-inwA 2
The left hand side is, by integrating by parts and using e
-inwA l
e , equal to
-inWA 2 " "

' ) - f(')) l'nw
JA 2 ' ,
e-1nw""f(')d'.
e (f( "2 "I
+
" "
Al
Hence, by (28.6), we have
and so
" "
f(A 2)-f(A l )
_~-=---:_ = ie
inwA . '"
JA 2 e-1nw"f(A)dA
2 (n 1,2, ... ) . (28.7)
s - nw
Al
* Adapted from J. Mikusinski [5], p. 414 ff.

28. Logarithmic Hyperfunction w 101
Since £P,z) f(\) ([P'z,t) [(I.'l,t)} € C[O,oo) and (s-nw)-l =

nwt
{e }, we see that the left hand side of (Z8.7) is in C[O,oo). Hence,
f(Az,t) - [(Al,t)1 fA z A
I s - nw ~_
Al
If(A,t) IdA (n = I,Z, ... ). (Z8. 7) I
Thus, letting
f(A 2 ,t) -s £(Al,t)}

g = (get)} {
k = (k(t)} = {rZlf(A,t) IdA},

Al
we obtain, from (Z8.7)I,
I~~~~I = h· If(Az,t~ - :~Al't)1 ~ h{k(t)};

that is,
t
nwu
g(t-u)e du
I f
~ 0 k(u)du (n = I, z, ... ). (Z8.8)
Therefore, by Theorem ZO (§Zl), we obtain the result
g(t-u) = 0 for all u with 0 < u < t,

i. e. J
g(u) = 0 for all u E [O,t].

Since t > 0 was chosen arbitrarily, we have proved that
A
o for all t> 0, i.e., f(A )

Z
Again, since Al and
A were arbitrary except for the restriction that
z
Al < 0 < A ' we must have the result that f(A) does not depend on A.
Z
This proves that f(A) = f(O) = I.
Hence fl(A) = 0 and so, by (Z8.5), f(A) = 0 (for all A), contra-
dicting f(O) = I.
PROPOSITION Z6. Let both wI and W E CIC be logarithmic hyperfunctions.
z
Then any linear combination
with real coefficients 01 and 0z is also a logarithmic hyperfunction.

Moreover, we have
102 VIII. EXPONENTIAL HYPERFUNCTIONS exp(-As) AND eXP(_As 1/ 2 )
(_00 < A < 00). (28.10)
PROOF: By applying (23.7) to
(28.11)
we obtain
f'CA) = ~1w1 exp(A~1w1)exp(A~2w2)
+ exp(A~1w1)~2w2 exp(A~2w2)
(~1w1 + ~2w2)f(A).
2
Furthermore, it is true that f(O) = r = r.
Therefore, by the uniqueness of the solution "f(A) of
" "f(O)
£1 (A) = (~lw1 + ~2w2)f(A) (_00 < A < 00)' = I,
we must have (28.10).
REMARK 28.1. As a Corollary of Proposition 26, we see that the operator

o E CIC is a logarithmic hyperfunction so that
f(A) =
is the unique solution of

fl(A) = Of (A) = 0 (_00 < A < 00), f(O) = I. (28.12)
EXAMPLES OF THE LOGARITHMIC HYPERFUNCTIONS

EXAMPLE 28.1. For any {ret)} E C, w = {f(t)} is a logarithmic hyper-
function.
PROOF: Let fO = I, fl = f, f2 = f*f, f3 = f*f 2 , ... ,f1 f*f1- 1 . Then,
as we already know in (25.4), we have, for any T> 0,
(O<t~T; n=1,2, ... )
Hence
00 An
exp(Af) = I + {f(A,t)} = L -. £1 E CH ~ CIC. (28.13)
n=O n. -
k An
is defined as the generalized lim E -0 ~
k-+ao n- n.
£1 E CH ~ CIC, and
-
28. Logarithmic Hyperfunction w 103
d eXPO,f) a ~ >.. n-l n

dU {ar f(>..,t)} = L
n=l
(n-l)! f
(28.14)
'" n
f· L ~ fn = f'exp(H) , exp(O'f) 1.
n=O n.
As an application of the above example, we have
PROPOSITION 27. Let a be a positive real number. Then
is a logarithmic hyperfunction and
(28.15)
PROOF: By (16.7), we have
and so, by (28.13),
(28.16)
On the other hand, we have, by (16.8) I,
;-z--z n+l { '"

(/s-+a- - s) = (_l)n(n+l) L (-1) k l
a
2+2k 2k-n
t
}
k=n 2 +2k-n(k_n)!(k+1)!'
We substitute this in (28.16) and exchange the order of summation to

obtain
= I - {>..Ik=O (_l)k 21+2ka 2k+2k!(k+1)!

tk k
L 2- n (k-n)!n! }
n=O
k!>..ntk-n
I - {>.. I (_l)k
2
a + t 2k k
(t+2A)k}
21+2k ! (k+1) !
k=O k
I _ { >.. aJ 1 (a / t 2+2At) } (by (15.3)).

/t 2+2At
104 VIII. EXPONENTIAL HYPERFUNCTIONS exp(-AS) AND eXP(_As 1/ 2 )
COROLLARY 1.
r'22
exp (A (s - Is- +a-)) = {J (a 1t 2+2At)} (28.17)
125 + a
2 0
PROOF: Differentiating (28.16) with respect to a and then dividing by

(-aA), we have
Then we substitute the expansion of JO(t) ((15.3)) into the right hand
side.
COROLLARY 2. Formulae (28.15) and (28.17) were obtained by making use

of binomial expansions with respect to a. Thus these formulae hold for
A~ 0 and for any complex number a F O. In particular, by substituting
ia = ~ a for a, we obtain
2 (28.18)
I - { A iaJ (ialt + 2At)},
l
/?:lli
I
exp(A(s _ s 2 _ ( 2 ))
= {J o(ia/t 2 + 2h)}. (28.19)
I s2 _ a
2
Finally, multiplying (28.15), (28.17), (28.18) and (28.19) by

exp(-As) and making use of (28.10) and of the shift effect of exp(-As),
we obtain
COROLLARY 3.
1"'22
exp(-Als~+a-) (28.20)
(28.21)
r 0, 0 < t < A
/T2 -AS
-I
exp(-A IS--A-) e
A iaJ (ia It 2 _A 2),
l/t 2
_A
2
1
(28.22)
28. Logarithmic Hyperfunction w
105
(28.23)
o~ \ < t
EXERCISES FOR §28.
Let w € CIC be a logarithmic hyperfunction and let f(\) be a

complex-valued continuous function defined for -~ < \ < ~ Prove that
\
z(\) Zo exp((\-\O)w) + J exp((\-IJ)w)f(lJ)dJl (28.24)
\0
is a solution of
dz(A) wz (\) (28.25)

~-
Part IV
Applications to
Partial Differential Equations
The purpose of this Part IV is to discuss mainly

2
Wave Equation: ZAA(A,t) = a Ztt(A,t),
Telegraph Equation: ZAA(A,t~ = Ztt(A,t) + 2KZ (A,t) + yZ(A,t),
t
Heat Equation: ZAA(A,t) =a Zt(A,t),
by converting these equations into hyperfunction equations.
The wave equation shall be discussed in the domain Al < A ~ A ,
2
t ~ O. By making use of (5.6), we obtain
2
s Z(A,t) = Ztt(A,t) + Zt(A,O) + SZ(A,O),
so that the initial condition with respect to t will be
Z(A,O) = $(A), Zt(A,O) = '!'(A).
Hence the hyperfunction equation for the solution Z(A) {Z(A,t)} of the
wave equation is
2 2
(i) Z"(A) - a s Z(A) = -a 2ljJ(A) Z
- a S$(A).
It is to be remarked that the right hand side of the equation is a linear

function of s.
The heat equation shall also be discussed in the domain Al < A < A '
Z
t ~ O. Then, by (5.6), we obtain
SZ(A) = Zt(A,t) + Z(A,O),
and hence, putting Z(A,O) = $(A), the hyperfunction equation for the solu-
tion Z(A) = {Z(A,t)} of the heat equation is
106
Applications to Partial Differential Equations 107
2 2
(ii) Z"eA) - (l sZeA) = -(l <P(A).
We are thus led to discuss the following problem. Let w E C/C be a

logarithmic hyperfunction. Consider a hyperfunction-valued function
f(A) = {f(A,t)} which satisfies
(iii) f"(A) - wf(A) = g(A)
where g(A) is a linear function of w, the coefficients of this linear

function being numerical valued functions of A.
The case w = s is the wave equation (Chapter IX) and the case
w = sl/2 is the heat equation (XI). We shall discuss the telegraph equa-
tion in Chapter X.
REMARK. In the case of the vibration of a string, let Z(A,t) be the
value of the vertical displacement of the point of the string (with the
abscissa A, at the instant t) from the equilibrium position of the
string (= the A axis). Then the customary assumption that the partial
derivative
a2
-2 Z(A,t)
n
is continuous at every point A of the string might be somewhat restric-
tive from the physical point of view. In this sense, our approach, using
the generalized derivative Z"(A), would be easier to handle as the reader
will see in due course in this Part IV (see, e.g., Remark 30.1 in the
Section 30 of this book).
Chapter IX
One Dimensional Wave Equation
2
§29. HYPERFUNCTION EQUATION OF THE FORM ~'(A) - Wf(A) = g(A), W€ CIC
PROPOSITION 28. Let W € CIC, kO € CIC and k € CIC. Further, let W
l
be a logarithmic hyperfunction and W F o. Then the solution of the equa-
tion
2
f"CA) - w f(A) = 0; f(O) = kO' f'(O) = kl , (_00< A <00) (29.1)
is uniquely determined from the A-initial condition * f(O) = kO and

f'(O) = k as follows:
l
-1 -1
2 (k O + W kl)exp(Aw)
(29.2)
-1 -1
+ 2 (k O - W kl)exp(-Aw) (_00 < A< 00).
PROOF: It is apparent that both

exp(AW) and exp(-AW)
are solutions of the equation
(29.1)'
and any linear combination
f l (A) = a expCAw) + b exp(-Aw) (29.3)
with coefficients a and b € CIC satisfies (29.1)' by Proposition 19

(§23). We shall determine a and b so that fleA) satisfies the A-
initial condition; i.e.,
*This is different from the initial condition with respect to t.
108
29. Hyperfunction Equation of the Form f"(A)-w 2f(A) g(A) 109
aw-bw k . (29.4)
l
Hence
and so, by virtue of the condition w # 0, we obtain the exitence of

-1 I
w w (see (18.6)) and
a = 2-1 (k O + w-1 k l ), b = 2- 1 -1
(k - w k ).
O l (29.S)
Next we shall prove the uniqueness of the solution of (29.1). Thus,

assume that there exists another solution f 2 (A) of (29.1). Then
f(A) = fleA) - f 2 (A)
is a solution of
0, f(O) o (_00 < A < (0). (29.6)
In order to show that rCA) must vanish, we take

v
fCA) (29.7)
where ~ denotes any real number. Then, by Proposition 19, we have
£1 CA) = £'1 (A.) f(211-A) - ff CA.) f' (2\.1-:\)

+ fl (A) fl (2~-A) - rCA) fll (2~-A)
2A A A 2A
W fCA) f(2~-A) - f(A)w f(2~-A)
O.
Therefore, by Proposition 19, we have the result that f(A) does not de-
pend on A. Hence, by (29.7) and reO) = fl(O) = 0, we have
f(A) = f(O) = fl(0)f(2~) + f(0)fl(2~) = 0 for all A.
Thus, by putting ~ = A in (29.7), we obtain
o i(A) = f'(A)f(A) + f(A)f'(A) = (f(A)2)"

Hence, again by Proposition 19, f(A)2 does not depend on A, and so
This implies, by Theorem 16 (§17), that
f(A) =0 (-'" < A < "') ,
We have thus proved the uniqueness of the solution of (29.1),

110 IX. ONE DIMENSIONAL WAVE EQUATION
PROPOSITION 29. Let w E CIC be a logarithmic hyperfunction and w # o.

Let, for Al and A2 satisfying Al < a < A , $(A) and ~(A) be
2
complex-valued continuous functions defined for A E [A ,A ]. Then, for
l 2
g(A) = -W~(A)
fOCAl = -(2w) -1
- ~(A),
r\
exp((\l-A)w)g(\l)d\l
(29.8)
(29.9)
{
_(2w)-1 f: 2
exp((A-\l)w)g(\l)d\l
is a solution of the hyperfunction equation
(29.10)
PROOF: We have, by calculating the generalized derivative of (29.9),
foCAl = 2~ {g(A) - I: 1
w exp((\l-A)w)g(\l)d\l
I:2
-
- g(A) + w eXP((A-\l)W)g(\l)d\l}
= ~I: w exp((\J-).)w)g(\J)d\J
1
+ I:2 w eXP((A-\l)w)g(\l)d\J}.
Hence
~O(A) = 2~ {-Wg(A) +
2
fA w exp((\l-A)w)g(\l)d\l
Al
- Wg(A) +
fA w exp(().-\l)w)g(\l)d\l
A2 2 }
2
g(A) + w fOCAl.
THEOREM 24. Let w E CIC be a logarithmic hyperfunction and w # O.

Let both k and k be E CIC. Let, further,
O l
g(A) = -W'(A) - ~(A).
Then the equation

29. Hyperfunction Equation of the Form f"(A)-w 2f(A) = g(A) 111
2
fll(A) - w f(A) = g(A)
(29.11)
{ f(O) = k ' fl (0) = k
O l
has a uniquely determined solution given by
f(A) = fo(A) + a exp(-Aw) + b exp(AW), (29.12)
(29.12)'
PROOF: By (29.10), it is not difficult to show that (29.12) is a solution

of
2
P'(A) - w f(A) = g(A).
To show that (29.12)-(29.12) I is a solution of (29.1), we shall verify the
A-intial condition. Thus
f' (0)
Finally, the uniqueness of the solution is proved as follows. Assume

that there exists another solution f of (29.1), and consider teA) =
f(A) - f(A), where f(A) is given by (29.12). Thus this teA) is a
solution of
2A A A
P' (A) - w fCA) = 0, f(O) f' (0) O. (29.11) I
Hence, as in the proof of Proposition 28, we prove that f(A) 0 for

all A.
REMARK 29.1. Formula (29.12) shows that the solution of (29.1) is given
by the sum of a partiCUlar solution fO(A) of the inhomogeneous equation
* We assume that Al < 0 < A .
2
(29.13)
(without mentioning A-initial condition) and a general solution
a exp(-AW) + b exp(Aw) of the homogeneous equation
2
f"(A) - w f(A) = O. (29.1) I
We then determine a E CIC and b E CIC in such a way that
f(A) = fO(A) + a exp(-Aw) + b exp(Aw)
satisfies the A-initial condition
f(O) = kO' fl (0) = k .

l
Hence the difficult part of solving (29.1) lies in obtaining a particular

solution fO(A) of (29.13).
There are cases where g(A) is a polynomial in A and in which we
can obtain fO(A) directly.
AN EXAMPLE.
y"(A) - s2y (A)
We substitute
into the equation and obtain
(2a
2 + 6a A) - s(a + alA + a A2 + a A3) = S(A 3 + 1).
3 O 2 3
k
Equating coefficients of A (k = 0, 1, 2, 3), we have
2
2a
2 - s a O = sl = 1, 6a 3 - s 2a l = 0, -s 2a 2 = 0, -s 2a 3 s;
that is,
6a 3 -61
-2 =-3
s s
Hence we have
3
-1 -6AI _A 1
=-+--+--
2 3 s
s s
EXERCISES FOR §29. Obtain a particular solution in 1 and 2 below.
1. y"CA) S2 Y (A) SA 3 .
2. y"(A) s2 yCA ) SA 4 .
§30. THE VIBRATION OF A STRING
A string of length A
O is streched along the horizontal A-axis
from point 0 to point AO' Let Z(A,t) be the value of the vertical
displacement of the point of the string with the abscissa A at the
instant t.
Then, as stated in the beginning of this Part IV, we have
(30. 1)
where the positive constant ~ = (p/P)1/2 is given by the tension P of

the string and p the mass per unit of length.
At t = 0, the string coincides with the A-axis, and, moreover, its
particles have no velocity, so that
Z(A,O) =0 and Zt(A,O) =0 (30.2)
Hence equation (30.1) converts into the hyperfunction equation
(30.3)
Further, let the left end A = 0 of the string move in a direction

perpendicular to the A-axis, with its motion defined by the function
vl(t); and similarly let the right end A = A of the string move per-
O
pendicular to the A-axis according to v 2 (t). Thus
Z(O,t) = vI(t) (0 < t < 00). (30.4)
Therefore, we wish to solve (30.3) under the conditions
Z(O) = VI EC (~CIC) and Z(A O) = v 2 E C (~CIC). (30.4) ,

PROPOSITION 30. The (hyper)function zeAl satisfying equation (30.3)

with condition (30.4) I can uniquely be represented as the sum
(30.5)
i. e. ,
e -a.AS _ e-a.( 2A O-A)s

zeAl
I _ e- 2a.AOs vI
(0 ~ A < (30.5)'
e-a.(AO-A)S _ e-a.(AO+A)S
+
_ e-2CXAOS
so that zl (A) and Z2(A) satisfy the same equation (30.3) and the
i
condi tions
vI' zl (A O) O·,
zl (0)
z2 (0) 0, Z2(A O) v2*
respectively.
PROOF: The condition that the solution
zeAl a exp(a.As) + b exp(-a.AS) (a,b E CIC)
of (30.3) satisfies (30.4)' is given by
i
z(O) = vI = a + b,
(30.6)
Z(A O) = v2 = ae a.Aos + be-CXAOs
This is solved by
- 2CXAOs -CXAOs
-e vI + e v2
a =
I _ e-2CXAOS
(30.7)
-CXAOs
vI - e v2
b
_ e-2CXAOS
Here the existence of the inverse (I - e- 2a.AOS)-1 is given by Remark

25.1. We have thus proved that (30.5)' is a solution of (30.3)-(30.4)'.
The uniqueness of this solution (30.5)' is shown as follows. Let
f(A) be a solution of (30.3). Then f(A) satisfies the A-initial con-
dition flO) = kO' f'(O) = kl . Moreover, in
* The function zl(A) represents the motion of a string whose right end is
held fast, and the function Z2(A) represents the motion of a string
whose left end is held fast.
X(A) = a l exp(aAs) + b l exp(-aAs),
we can uniquely choose al and bl from C/C in such a way that
x(O) = al + bl = kO' x'(O) = alas - bIas = k l ,

since the determinant
1 1 I -- -2 as i O.
I as -as
Thus, by Proposition 28, f(A) must be of the form X(A). This fact insures
that the (30.5)', obtained above, is the unique solution.
GRAPHICAL INTERPRETATION OF (30.5)'. How does the point of the

string with abscissa A move when t increases? We shall consider this
for the solution Zl(A) in the case
a
1
= 2' AO 10 and VI (t)
= f'
4,
0 < t < 4
=
4 < t < 00
(30.8)
(Mikusinski [5], p. 203).
Then, for A = 2, we have
-s -9s
e - e
zl (2) = =----l::..,O:'"""s- VI' (30.9)
I - e
(See the following three graphs.)
?-
11
;C"
5'-'1,,(1)1, J,----t-:-,--.'-I,-,(-')I....
J?s--
O r
The graph of e -AS {vl(t)} is given by shifting the graph of {vl(t)} to

the right by the distance 1. Similarly, the graph e -9s{ vl(t) } and the
graph of e -sf vl(t) } - e -9s{ vl(t) } are given also.
Further, since
+ e
-20s + e
-30s
+
we have
-lOs -20s -30s -s -9s

Zl(2)=(I+e +e +e +···)(e -e )v .
l
Hence the graph of zl(2) is given by the sum of the graphs which are ob-
tained, respectively, by shifting the graph of
fee -s - e -9s )v l }
to the right by the distance 0, the distance 10, the distance, 20, etc.
o 1 9 11 13 15 1920
REMARK 30.1. zl(A) = {zl(A,t)} is a hyperfunction solution of (30.3).

Hence if {vl(t)} is not smooth, as in (30.8), then it is natural that
the graph of {zl(2,t)} has many "corners".
GRAPHICAL INTERPRETATION OF (30.S)'. We shall consider the shape

of the string at the instant t = to < 10. We do this for the case
(30.8). Then
x (exp (- t AS) - eXP(-(lO - ~ s) )V l

nIo{eXP(-(lOn + t A)s) - exp(-(lOn + 10 - ~)s)}vl
by making use of the addition theorem (28.4).
For this zl(A), the terms in which the coefficients of (-s) are
larger than 10 vanish if 0 ~ t < 10 because of the property of the
shift operator e -AS given in Remark 22.2. Thus, to examine zl (A) when
o < to < 10 and A < A ~ 10, we have only to examine its term
O
[ex p (-} AS) - eXP(-(lO -} A)S)]{Vl(t)} ~ xl(A,t) - X2 (A,t)
for t ~ to and A < 10. By (30.8) and Remark 22.2, we have
{
(0 ~ t < } A)
1 1 (l2 A < 1
exp(- 2 As){vl(t)} - 2 A t < 2 /..+4)
1
(2 A+4 < t) ~
and
(0 <
~
t < 10- l2 A)
(10- }A ~ t < 14- 1)

1
(14- 2 A ~ t).
1
The graph of xl(A,t O) ~ exp(- 2 As){vl(t)}t~t is represented by a con-
o A
tinuous line and the graph of X2 (A,t O) ~ exp(-(lO - 2)s) x {vI (t)}t-t
is represented by a dotted line. - 0
D ., J
OJ .:.: -t) 2i)-2I u 10
f
0
i)\---,,--i.
!--------"--'
21",-8 20-21010
A
Subtracting, graphically, the dotted line from the continuous line, we

obtain the shpae of the string at the instant to and for A ~ A ~ 10.
O
§3l. D'ALEMBERT'S METHOD

We shall treat the vibration of a strong which is held fast at its
two ends A= 0 and A= A ;
O
Z(O,t) = 0, Z(AO,t) = 0 (0 ~ t < 00), (31.1)
with, furthermore, its initial shape Z(A,O) at the instant t 0 given

by a real-valued continuous function ~(A); i.e.,
zCA,O) = Ijl(A) (31.2)
As was suggested by Jean Ie Rond D'Alembert (1717-1783), we shall make

use of
PROPOSITION 31. * The generalized derivative with respect to A of
(31.3)
satisfies
Zi(A) =aszl(A) -a[~(A)], where [~(A)] S{~(A)} SIjl(A). (31.4)
PROOF: Since
we have
zi (A) = s {ax
aA t
f
a aA + cHex)dT T} = as{¢>(A + t - cHA) }
ex)
COROLLARY 1. For
(31.3) I
we have
ZZ(A) -aSZ 2 (A) + a[~(A)]. (31.4) ,
COROLLARY 2. The function
(31. 5)
*J. Mikusinski [5), p. 214.

31. D'A1embert's Method 119
satisfies the hyperfunction equation
zlt{>.) = a. 2s 2z{>.) 2
- a. s[4>{>')] (31.6)
which corresponds to the string vibration equation and initial conditions

as follows:
PROOF: Since
we have
z"{>.) = 21 o.s{zi{>') - zZ{>'))
= 21 a. 2s 2 (zl (>.) + z2{>')) - 21 2

20. s[4>{>')]
= 0.2 s 2z {>.) _ 0.2 s [4>(>')].
Moreover, z{>',O) = 4>{A) is clear and, if z{>.) is twice continuously

differentiable with respect to t, then, by (5.6),
Ztt{>.,t) = s 2 Z(A,t) - S[Z{A,O)] - [Zt{A,O)].
REMARK 31.1. In order that
satisfies the >'-boundary condition (31.1), we must have
(O < t < +00 ) (31.7)
This condition (31.7) is surely satisfied if
(31. 8)
Thus if 4>{A) is an odd function defined for all real A and, moreover,
is periodic with period 2A > 0, then 4>(>') satisfies (31.8). Such is
O
the case, for example, if 4>(>') = sin(w>./>.O).
Therefore, if we take $(A) as above, then z(A) given by (31.5)

is a hyperfunction solution of (31.6) satisfying (31.1)-(31.2).
REMARK 31.2. Any solution z(A) of (31.6) satisfying the A-boundary

condition
~(O) = 0, 0,
must be given by
with ~(A) as above.
PROOF: Let z(A) be such a solution and let
Z(A) = Z(A) - Z(A).
Then we have
(31.9)
Therefore, as in the proof of Proposition 30, the solution of (31.9) is

uniquely determined so that z(A) must vanish, where ~(A) = z(A).
THE STRING VIBRATION WITH GIVEN INITIAL VELOCITY. We shall deter-
mine the motion of a string under the initial condition
Z(A,O) = 0, Zt(A,O) = ~(A) (31.10)
and the boundary condition
Z(O,t) = 0, (0 ~ t < eo), (31.11)
assuming that no external force acts upon the string. The corresponding
hyperfunction equation is, by (5.6),
2 2 2
z"(A) = ex s z(A) - ex [~(A)]. (31.12)
Hence if z(A) = {z(A,t)} satisfies (31.10), (31.11) and (31.12), then
YeA) = sz(A)
satisfies
2
Y"(A) = ex s\(A) - ex2s[~(A)],
(31.13)
{ yeO) = y(A
o) = o.
Thus, as in the case of (31.5), we obtain
31. D'A1embert's Method 121
so that
(31.14)
would be a candidate for the unique hyperfunction solution of (31.12) under

conditions (31.10) and (31.11). As in the case of (31.7), we see that
(31.14) satisfies (31.10) and (31.11) if ~(A) satisfies
-~(-A) = ~(A) = ~(A + 2A O)' (31.8)'
D'ALEMBERT'S WAVE FORMULA. We have thus obtained the following result.

If ~(A) and ~(A) are respectively twice and once continuously differ-
entiable, then, under the condition (31.8)-(31.8)', the so-called
D'A1embert's formula
A+t/ex}
Z(A,t) = ~ {Ij>(A - ~) + Ij>(A + !') + ex ~(u)du
ex JA-t/ex
gives the solution of
EXAMPLE 31.1. If ~(A) = sin A~ A, the function

o
Z(A,t) = i{Sin(:o (A - ~)) + Sin(:o (A + ~))}
= COS(A: ~)SinUo A)
is the solution of
2
{ zH (A,t)
z(O,t)
=
0,
ex Ztt(A,t),
Z(AO,t) = 0, Z(A,O) = $(A), Zt(A,O) O.
EXAMPLE 31. 2. If ~(A) = sin to A, the function
Z(A,t) = ~2 It0 (~(A + ~)

ex
+ ~(A - !.))dT
ex
= sin ~~ J: cos(:o ~)dT = ex: o sin ~~ sin(A: ~)

is the solution of
J ZAA(A,t) = a 2Ztt (A,t),

1 Z(O,t) = 0, Z(AO,t) = 0, Z(A,O)
§32. THE VIBRATION OF AN INFINITELY LONG STRING
Consider the uniquely determined hyperfunction solution (see (30.5) ')

-aAS -a(2A O-A)S
e - e
zl(A) = ~---=-....,.......--- vI (a > 0) (32.1)
-2aAos
- e
of the equation
(32.2)
satisfying
~Zl(A'O) 0, (zl)t(A,O) = 0 (0 ~ A < AO),

(32.3)
zl (O,t) {vI (t)}, zl(AO,t) 0 (0 < t) .
We shall examine zl(A) for 0 < t < aA ' Consider the decomposi-
O
tion of zl (A) :
Then, by the property of the shift operator, we see that
-a(2A O+A)S
both e {vl(t)} and
vanish when
(32.4)
In fact, the conditions that
hold when (32.4) is satisfied.

Thus, if the string is very long, i.e., if A is very large, we
O
may put r(A){vl(t)} = 0 so that we have
32. The Vibration of an Infinitely Long String 123
0, o< t < aA
(32.5)
={ vl(t-aA), o~ aA < t.
Therefore, it can be seen that at the instant t the front of the

wave moving from the initial point A = 0 of the string at t =0 reaches
the point A = t/a. Thus the wave moves with the velocity
l/a = IPTP.
Such is the motion of a string which is said -- though not quite correctly --
to be infinitely long.
t'l(t-a,O
aA
The above graph describes the case of

t, 0 < t < 4
vI (t) ={ 4, 4 < t < 00.
Chapter X
Telegraph Equation
§33. THE HYPERFUNCTION EQUATION OF THE TELEGRAPH EQUATION

We shall discuss the voltage and the current in a long cable con-
sisting of two parallel wires stretched along the A-axis. Let U(A,t)
and I(A,t) respectively be the voltage and the current at the point of
the cable with coordinate A at the instant t. Then the following rela-
tions hold between the functions U and I:
(33.1)
where R denotes resistance, L self-inductance, G leak-conductance and

C capacitance; these quantities are measured per unit length of the cable.
Suppose that at the initial instant t = 0, no current flows through
the cable and the voltage is equal to zero everywhere:
U(A,O) 0, I(A,O) = O. (33.2)
Hence
U (A) = sU(A), It (A) sl(A)

t
and so, by (33.1),
{ U' (A) -LsI (A) - RI (A) - (Ls + R)I(A)

(33.3)
I I (A) -CsU(A) GU(A) -(Cs + G)U(A).
This implies
U"(A) = -(Ls+R)I'(A) = (Ls+R)(Cs+G)U(A).
We have thus obtained the equation for the voltage U(A):
U"(A) = (LCs 2 + (RC+LG)s + RG)U(A). (33.4)

124
34. A Cable With Infinitely Small Loss 125
Similarly, we obtain the equation for the current I(A):
I"(A) = (Cs+G)(Ls+R)I(A) = (LCs 2 + (RC+LG)s + RG)I(A). (33.4)'
These two equations (33.4) and (33.4)' are of the same form, and we call
them telegraph equations.
REMARK 33.1. The telegraph equation reduces to the wave equation if

LC = a 2 , (RC + LG) = 0 and RG = O. Similarly, the telegraph equation
reduces to the heat equation if LC = 0, (RC + LG) = a 2 and RG = O. We
shall give, in the following sections, the concrete representation of the
hyperfunction solution of the telegraph equation by classifying the equa-
tion according to the relationship among the parameters L, C, G and R.
§34. A CABLE WITH INFINITELY SMALL LOSS
Conductance never occurs without losses. They are due to the resis-
tance R and the leak-conductance G of the current. However, if the
quantities Rand G are very small, self-inductance L and capacity
C being large, we may assume approximately that
R =0 and G 0 (furthermore L> 0 and C> 0).
We shall call such a cable as with infinitely small loss or rather simply
as a cable without loss.
The telegraph equation is then reduced to
(L> 0, C> 0). (34.1)
We further assume that the cable is considerably long, and that at the
left end A=0 of the cable an electromotive force given by E = {E(t)}
is applied.
Thus we are confronted with (34.1) subject to the A-boundary condi-
tion
U(O) E = {E(t)}, (34.2)
where A > 0 is considerably large.

O
Therefore, as in the case of zi(A) of §32, we can take
U(Al = exp(-/LC As){E(t)} (34.3)
and so
U(A,t) (34.4)
126 x. TELEGRAPH EQUATION
Thus the wave of voltage moves with the velocity l/ILCA. Similarly we
have, by making use of (33.3), (34.3), and R = 0,
I (A) ~ U'(A) = IC7L exp(-ILCAS){E(t)}; (34.5)
Ls
that is,
ICA,t) (34.6)
O<ILCA<t.
§35. CONDUCTANCE WITHOUT DEFORMATION

It is said that O. Heaviside was the first to point out that in the
more general case
R G (L > 0, C > 0) (35.1)

r=C
both the voltage and the current are transferred in a way similar to the
case of the preceding section, the only difference being the fact that the
wave amplitudes are exponentially damped as the waves move forward.
Indeed, from (35.1), the telegraph equation
U"CA) - (Ls+R) (CS+G)U(A) 0
is written, by putting L yC and R = yG with y > 0, as
2
U" (A) - y(Cs+G) U(A) 0;
that is,
2
U"CA) - (o.s+8) U(A) = 0 (0. = ILC, 8 IRG) . (35.2)
As in the preceding section, we assume
U(O) = E = {E(t)}, U(A )

O
=0 (35.3)
and that the cable is considerably long. Thus we may prove as in the case
of zl (A) of §32, that
U(A) {U(A,t)} = exp(- (o.S+8)A){E(t)}. (35.4)
PROOF: The condition that the general solution of (35.2),
U(A) (a,b E CIC) ,
satisfies the A-boundary condition (35.3) is given by
(o.S+8)A -(o.S+8)A
a+b O + be O O·,
E {E(t)}, ae
35. Conductance Without Deformation 127
that is, by
-2(as+B)AO
-e ·E E
a = -2(as+B)>"0
, b
-2(as+B)>..0
I-e I-e
Hence
-(as+B)>" -(as+B) (2)''0->'')
e 'E-e 'E
U(>")
I _ e- 2 (as+6)>"0
(35.5)
-(as+6) (2)''0+>'') -(as+6) (2)''0->'')

e 'E - e ·E
rCA)E
- 2(as+6)A0
I - e
In the numerator of r(>..)E, both the two terms
-B(2)''0+>'') -a(2)''0+>'')s
e ·e {E(t)},
-6(2)'' ->..) -a(2)'' ->..)s

e O.e 0 {E(t)}
are zero if
(35.6)
because of the presence of shift operators
and
In fact, (35.6) implies
Thus when >"0 + ~ we obtain r(>..)E = O. Hence we have proved (35.4) for
an infinitely long cable satisfying (35.1) and (35.3).
As a Corollary of (35.4), we have, by making use of a = ILC, 6 = /RIT
and exp(-(as+6))>" = e-B>"e-a>..s,
0, 0 ~ t < ILC >..

U(>..,t) = /RIT- (35.7)
{ e-A RG E(t-ILeA), 0 ~ 1Le>.. < t
Hence the amplitude e ->..IRG of the wave decays exponentially as >.. + ~.

Similarly, by virtue of (33.3) we obtain
I (A) _ U'(A) = as+B e-(as+B)A{E(t)}

Ls+R Ls+R
Here we have made use of (35.1) to the effect that
(Ls+R)1C7[ = ILCs + fRG = as + B.
Hence we have
o o<t<ILCA
I(A,t) = ' =
{ IClL e-fRGAE(t - ILCA), °~ ICLA < t (35.8)
Thus, for I(A,t) .

as well, the amp11tude =- e -fRGA
yC/L of the wave decays
exponentially as A~ w.
REMARK 35.1. The uniqueness of the solution given in (35.5) is proved as

in the case of the solution given in (30.5)'.
§36. THE THOMSON * CABLE
This is the case
L =G= ° (R> 0, C > 0).

In this case, under the assumption of considerably long cable, Thomson
calculated, in the middle of the 19th century, the propagation of current
through a cable on the sea bottom across the Atlantic Ocean.
The telegraph equation is, when L = G = 0, of the form
U" (A) = RCsU(A). (36.1)
Let the initial voltage at t = ° be zero:
U(A,O) = ° (36.2)
Here 1.. is the length of the cable, which is sufficiently long. Let the
0
electromotive force E = {E(t)} applied at the initial point of the cable
be constant in time:
U(O,t) (36.3)
Thus putting ;RC = a > 0, we have to solve the equation
(a = ;RC > 0) (36.4)
under conditions (36.2) and (36.3).
* Willi~m Thomson (later Lord Kelvin, 1824-1907).

PROPOSITION 32. The hyperfunction solution of (36.4) combined with (36.2)-

(36.3) is uniquely determined and is given by
(e- aAIS _ e- a ( 2A O-A)IS)E

U(A) (36.5)
-2aA OIS
I - e
PROOF: For a € CIC and b € CIC,
aAIS
U(A) = ae- + be aAIS (36.6)
is the general solution of (36.4). We shall determine a and b so

that (36.6) satisfies (36.3):
r U(O) = U(O,t) = a + b = E,
It U(A ) -aAOIS aAOIS
(36.7)
O = U(AO,t) ae + be O.
By solving (36.7), we obtain
E
a
I - e- 2aA olS
(36.8)
-e -2 aA o1SE
b
I -e -2aAOIS
Therefore, (36.5) is proved.

The uniqueness of the solution UeA) is proved as follows.
Let F(A) be a solution of (36.1). Then F(A) satisfies the A-
initial condition F(O) = k O' F' (0) = kl · Moreover, in
we can uniquely choose al and b

l
from CIC in such a way that
x(O) = al + bl = k o' x'(O) (36.9)
since the determinant
I -a~ a~ I = 2alS .
is nonzero. Thus, by Theorem 24, F(A) must be of the form X(A). This
fact assures that (36.5), obtained above, is the unique solution.
PROPOSITION 33. The unique solution UeA) given by (36.5) is expanded as

follows.
130 X. TELEGRAPH EQUATION
U(A) Ee-OA15 + rCA),

(36.10)
'" -0(2\1A O+A)15 -O(2VA O-A)15 )
{ rCA) = E( L (e - e ) .
\1=1
PROOF: Expanding similarly as in (25.13), we have
-20A 15 -1) '" -208A 15

(I -e 0 =1+Le 0
\1=1
Hence, from (36.5), we obtain (36.10).

In order to estimate rCA) when A ~ "', we shall make use of
O
PROPOSITION 34. * If A > 0 and 0 < t < "', we have
2
- - ) ~ fA,
f(A,t) = ->..- exp (A ( 3 II... .
6>..2) ="2/:;:3-3 (36.11)
2i;t3" 4t A 1fe
PROOF: Since
a
at f(A,t) = (A 2-6t) ->..- exp(>..
- 4t2) ' (36.12)
8M
the function f(>..,t) has a maximum at t = >..2/ 6 .
Moreover, by
lim f(A,t) 0 and lim f(A,t) = 0, we see that f(>..,t) attains, for
t-l-O t-+«> >..2 2
>.. > 0, its largest value f(A, 6) at t = A /6.
REMARK 36.1. (36.11) may be written as
(36.11) ,
in the sense of hyperfunctions; that is, the left hand term is the hyper-
function represented by
{_A ex~ _ ~~ )}
2Q
and the right hand term is the hyperfunction represented by
t~/;3} .
Hence (36.11)' describes (36.11).
Now we sestimate rCA) by virtue of (36.11)'. We obtain, for
o< >.. ~ AO' an estimate of r(A):
*Mikusinski [5], p. 225.

2
Ir(A) I < IEI*[ 313]tL
1/0.
(2\1A +A)
2
+ Iii
(2\1A -A)
2 }
O O
< [3/3
3 2 ~
1 ] IEI*{l)
1fe a2A~ \1=1 (2\1_1)2
(K is a positive constant),
where lEI means the absolute value of the function E, and * denotes
convolution. More precisely, we have obtained
(0 < A ~ AO' 0 < t),

(36.13)
Therefore, we have
PROPOSITION 35. In (36.13), the error Ir(A,t)1 tends to 0 as A

O
tends to 00. Thus, for the infinitely long Thomson cable with a = IRe
and E = {EO}' we may take
U(A) (36.14)
That is, by virtue of (27.14),
(36.15)
Thus, by L = G = 0, (33.1), and (36.14), we obtain
I(>..)=-.!.U'(A)
R
(36.16)
that is, by (27.13),
I(A,t) = EO { ~ exp(- R~~2)} (36.17)
EXERCISES FOR §36. Prove the following inequalities.
(a) 22
1
~ as1 (a > 0)
s +a
2
*Since r:::
Cerf t = (2/t'1f) f t e
.. -t
dt, the initial condition lim U(A,t) o
(when A > 0) holds.
HO
(8) ..l. e -AIS < _1_ 12 (A > 0).

IS = rnAS
§37. CONCRETE REPRESENTATIONS OF exp (-A/as+S), exp( -AI as 2+Bs) AND

exp(_A((s+a)2_ S2)1/2)
For the telegraph equation
UtI(A) = (Ls + R)(Cs + G)U(A),
we have the following cases, excepting those of §34, §35 and §36.
Case (i) L= ° (R> 0, C> 0, G> 0):

exp(-A(RCs + RG)l/2).
Case (ii) G= ° (L> 0, R> 0, C> 0):

2
exp(-A(LCs + RCs)1/2).
Case (iii) L > 0, R > 0, C > 0, G > 0:
exp(-A(LC[(s+a)2 _ S2])1/2), where
a = i (~ g),
+ B= i (f - g)
since
(Ls+R) (Cs+G) (f + g) + s + i(f - g))

Lc(i
x (i(f ~) s - i(f - ~))

+ +
= LC1 (CR+LCs)(LG+LCs) = (R+Ls) (G+Cs).

Therefore, the corresponding exponential hyperfunctions come in as
solutions of the telegraph equation.
In this section, we shall make use of the Mikusinski mapping Ta of
C/C into C/C (introduced in §20) to obtain concrete representation of
the above exponential hyperfunctions as functions of t.
We already know (see §20) the following:
(37.1)
(37.2)
37. Concrete Representations 133
a So a
T = I....s. (p,q E C/C) , (37.3)
p Tap
Ta(ab) = (Taa)·(T~) (a,b E C/C), (37.4)
T:[Y] = [y]; in particular, Tal = I, and

(37.S)
{ T ([y]a) = [y]Taa (a E C/C).
We have further that
a a I I
T 5 = T ill = {eat} = 5 - [a] 5 - a. (37.6)
and, more generally,
n
Tas = (s_a)n (n = ' ' ' , -3, -2, -1,0,1,2,3,"')' (37.7)
PROOF: If n > 0, we have
by (37.6) and (37.4). If n < 0, we have
If n = 0, we have
o
5 •
If R(s) is a rational function of s, then
rClR(s) = R(s - a). (37.8)
PROOF: We have
R(s) = ~ where Q(s) and pes) are polynomials in s.

P (5)'
Thus
TaR(s) - rClQ(s) _ Q(s-a) - R(s-a) ,

- ~P(s) - pes-a) -
since
~Q(s) = Q(s - a),

as may be seen from (37.4), (37.S) and (37.7).
(37.7) is generalized as follows.
Ta...!.... = _ I _ (_00 < A < 00). (37.9)

I. A
s (s-a)
PROOF: Let n be a natural number such that A+n > 1. Then
and so
r(A+n)-ltA+n-Ieat
r(n)-ltn-Ie at
_----"-,1,-- ( I )-1 I
= (s_a)A+n (s_a)n = (s_a)A
by (10.20).
EXAMPLE 37.1.
(37.10)
and
T I
a /"T2 a r--;:; a r--;:;
s~-B~ = (T ys-B)(T ys+B)
(37.11)
= Is-B-a Is+B-a = / (s-a) 2 - B2 .
If f(A) has the generalized derivative f'(A), then
(37.12)
PROOF: There exists an element a E C/C with a # 0 such that af(A)

{fl(A,t)} is continuously differentiable with respect to A and
f'(A) = a
-1 {
axa fl(A,t)}. Hence, by (37.4),
Taf(A) = reta-l{f l (A,t)}

Thus
= rCt(a-l{it fl(A,t)})
= retf' (A) .
As a Corollary, we have
~exp(Aw) = exp(A~w). (37.13)
PROOF: We have
(~exp(Aw))1 ~(exp(Aw))1 = Ta(w exp(Aw))
~w'Taexp(Aw) ,
Taexp(O'w) = Tal = I.
Hence, by Proposition 24, we must have (37.13).
APPLICATIONS OF THE MAPPING Ta .
(37.14)
by (27.5).
~ e-A.'s=a T
a
~ e- AIS = {~ exp(at _ ~:)} (37.15)
by (27.3).
For any real number a and any positive number B, we have
2
r B)} -2a6( a ) -2BIs+a
{Erf( avt + It =
1
5e I 2 -1 e +
I
5' (37.16)
s+a
and
(37.17)
Proof: For (37.16): Writing
f(t) =--2 Jalt B/ 1t e _u 2du

+
Ii 0
we have
Hence, by (27.3) and (27.5),

Furthermore,
ff = sf - s [1]
by
f(+O) lim f(t) 1;

HO
that is,
ff I
f =5 + 5'
proving that (37.16) is valid.

Similarly, writing
2 fait-B/it _u Z
get) = -- e du Erf(alt - ~ ),
Ii 0 It
we have
Z
{g'(t)} e za6(- -
a - + 1) e -2B/s+a ,
Z
/ s+a
whence follows (37.17) by virtue of g(+O) = -1.

From (37.16) and (37.17), we easily obtain
.!.
s
2
e-Z6/s+a = {.!.. e-zaBErf(alt
Z
_ .1..)
It
(37.18)
1 ZB ( B) eZaB+e- ZaB }
- - e a Erf alt + -- - -'----;::-'---
Z It 2
and
a e-ZB.;;;;z = H e-ZaB Erf( alt - ~)

s;;:;.z (37.19)
ZaB -e -zaB}
B) - e
+ -1 e ZaB Erf( alt + -- Z •
Z It
Next we have
s -
by (37.6) and (37.11). Hence, by (28.15) and (28.17), we obtain
and
exp(A(s-1 (s_a)2+ 62)) {ea (A+t)J (6

o
I t 2+2At)}. (37.21)
/ (s_a)2 + 62
Replacing, in formulas (37.20)-(37.21), a by -a and B by ia,

we obtain
e -aA - { A e -a(A+t) laJ

. .;-Z-- }
(lar't-+2At) (37.22)
l
/ t 2+2At
and
exp(A(s-~)) -a(A+ t) ;-:r-- }

{e Jo(iar' t~+2At) . (37.23)
/ s2+2as
MUltiplying (37.20), (37.21), (37.22), and (37.23) by the shift opera-

-AS replaces t by (t-A) and so we have
tor e
(37.24)
O~'<t},
o~"J,
exp (-AI (s-a) 2+62)
(37.25)
2
/ (s-a) 2+8
-aA -AS
e e
(37.26)
o~
and
o< t < >.

(37.27)
Therefore, we have obtained concrete representations of the cases

(i) and (ii) mentioned at the beginning of this section, by (37.14) and
(37.26) respectively. As for the case (iii), we obtain the following:
a>. ->.s
e e
lJ'
o< t < >.,
(37.28)
o<
>. < t
which can be derived by replacing 6 in (37.24) by i6.
§38. A CABLE WITHOUT SELF-INDUCTION
This is the case
L = 0 (R> 0, C > 0, G > 0).
The hyperfunction equation of the cable is
U"(>.) = (RCs + RG)U(>,) (38.1)
with the A-boundary condition
O. (38.2)
Thus, putting
a = IRe, y = RG/RC = G/C, (38.3)
the solution will be given by
U(>') = a exp (-a>.lS+Y) + b exp (a>.lS+Y) ,

{ a E CIC, b E CIC.
Hence, by (38.2), we have
a+b = E, a exp(-a>'ol5+-¥) + b exp(a>'ol5+-¥) O.
and so
38. A Cable Without Self-Induction 139
a ;; E b
I-e
- 2aA
olS+Y'
U(A)
In
-aAIS -a(2A O-A)1S

TYU(A) e - e E
I-e
-2aA
olS
we let 1. ~~. Then as in the case of the infinitely long cable (§36),
0
we may take
e -aA!s+y E ,.
that is, when E = {EO}'

[Eol ( )
U(A) = ---5- exp -RCA(/s + G/C) . (38.5)
Therefore, by (37.18), we have, when
U(A,t) = TEO [ e - A,lRG Erf( IGt7C - Z(/RC/t)

A )
(38.6)
Next, for leA), we obtain, by (33.3),
-1 E/RCS + RG
I(A) = TUfCA) = Rs exp(-A/RCs + RG)
(38.7)
E ( C + G ) exp(-A/RCs + RG),
O
IRCs+RG sIRCs+RG
whence, by (37.14), (37.15) and (37.19),

140 X. TELEGRAPH EQUATION
I(A,t) ~ -EO
"C/R
&
(G
exp - - t - -RC A
C 4t
2)
+ EO
IG7R 2" [-A/RG
e Erf( IGt/C - I(/RC/t)
A ) (38.8)
§39. A CABLE WITHOUT-LEAK CONDUCTANCE

This is the case
G=a (L > 0, C > 0, R > 0).
In many practical calculations, the leakage of the current is so

small that it can be neglected. Thus the equation is
2
U"(A) = (LCs + RCs)U(A), U(O) = E, (39.1)
under, of course, the assumption that at the initial instant t a there

is no voltage and no current on the cable.
The general solution of the hyperfunction equation (39.1) is
UCA) = a exp( -AI LCs 2 + RCs) + b eXP(AI LCs

2
+ RCs) ,
(39.2)
a € CIC, b € CIC.
We shall be concerned with such a hyperfunction solution U(A) which
can be represented as a numerical-valued function U(A,t) of A> a and
t ~ O. This situation is essential, since otherwise the U(A) cannot be
calculated numerically.
Assuming that the cable is infinitely long, we have
PROPOSITION 36. If the solution (39.2) can be represented as a numerical-

valued function U(A,t) in the domain
a < A <~, a< t < ~,
then the coefficient b must be O.
PROOF: * By the binomial expansion (14.3), we have
*Mikusinski [5], p. 258.

39. A Cable Without Leak-Conductance 141
_
2 R)1/2
(LCs + RCS)l / 2 ILC s ( I + Ls
(39.3)
lLC ILC R
s +--+ f = ILC s + CX + f,
2L
where
r'cR
cx=--,
u'L
By (39.2), (39.3) and (28.10), we have
~ -ACX -Af ~ ACX Af

U(A) - a exp(-ArLC s)e e = b exp(ArLC s)e e . (30.4)
Putting
a = 3.,
p
b = ~ (P,q,Pl,ql E C),
PI
-ACX -H
and then nul tip lying (39.4) by PPle e , we obtain
-CXA -Af -2ACX -2Af ~

PPle e U(A) - qple e exp(-ArLC s)
(39.5)
Since f is a numerical-valued continuous function of t,

-CXA -H
PPle e U(A) is also a numerical-valued continuous function of A > 0
and t ~ O. Furthermore, since exp(-AILC s) is a shift operator,
~ -2ACX -2Af
exp(-ArLC s)(qple e )
is a numerical-valued function in K[O,oo) when A > O. Thus, by (39.5),
J exp(AILC s)pql is a numerical-valued function
1 of A and t in the domain 0 < A< 00, 0 < t.

(39.6)
On the other hand, for any A > 0, and A > 0 with AD > A > 0,
O
we have, denoting
{get)},
that
exp(\1LC s)pql exp(-(\O-\)ILC s)exp(\OILC s)pql
exp(-(\O-\)fU: s){g(t)} (39.7)
=0 when 0 < t < (\0-\)fU:,
since exp(-(\o-\)fU: s) with (\0-\) > 0 is a shift operator.

Hence, by the assumption that the cable is infinitely long, we may
put
exp(\fU: s)pql =0
by letting \0 ~~. Thus
b exp(\1CL s) = __1__ exp(\lC[ s)pql =0 for \ > O.

PPl
But, as an exponential hyperfunction, exp(\fU: s) € CIC does not vanish. *

This proves that
b = O.
Therefore, combined with (39.2) and the A-initial condition

V(O) = E, we obtain
V(\) = E exp (-\1 LCs 2 + RCs). (39.8)

If E is of the form E = {EO}, then
V(\)
EO
= -- exp(-\
I LCs
2
+ RCs), (39.9)
s
and, by (33.3),
_ V' (A) __ EO/s ILCs 2 I 2

1(\) = Ls+R Ls+R + RCs exp(-\ LCs + RCs)
(39.10)
EOC exp(-\/LCS 2 + RCs).
2
Ir.Cs + RCs
Hence, by putting
Ct' = .!. ~
2 L '
we see that
*See Proposition 24 (§28).

40. The Case Where All the Four Parameters are Positive 143
EO / 2
U(A) =--
s
exp(-AILC s +2a's)
= EOh exp(-AILC ~s2+2a's).
Therefore, by (37.26),
(39.11) *
when 0 ~ ILC A < t.

Similarly, by (37.27) and (39.10),
~ -a't . / 2 2
I(A,t) = EO .C/L e JO(la ' t -LCA )
when 0 ~ ILC A < t. Moreover, both U(A,t) and I(A,t) vanish when
o< t < ILC A.
§40. THE CASE WHERE ALL THE FOUR PARAMETERS ARE POSITIVE
This is the case
L > 0, R > 0, C > 0, G > O.
Introducing, as in §37, the notation
the hyperfunction equation of the cable is
(40.1)
As in the preceding section §39, we assume that the cable is infinitely

long. Then the solution U(A) of (40.1), which can be represented as a
numerical-valued function U(A,t) in the domain 0 < A< A ' 0 ~ t < w,
O
is given by
(40.2)
By (33.3), we have
* Compare with that on pg. 250 of J. Mikusinski, [5].

144
X. TELEGRAPH EQUATION
I (A) - -U' CA)

- i:"5""+'R
so that
(40.3)
by making use of
!C7T(I + CGS) x s(Ls+R) (Cs + G) (Ls + R)

(s+a)
2 2
- B.
ILC LC
Therefore, by (37.24) and (40.2),
(40.4) *
when 0 ~ ILC A < t.

Also, by (27.25) and (40.3),
at 2
I(A,t) = EO IC7i::((e- J (iB/ t 2 _L0 )
o
(40.5)
+ gItILC Ae-atJo(iB! t 2_LCA 2)dt)
when 0 < lLC A < t. Moreover, both U(A,t) and I(A,t) vanish when
o~ t < ILC A.
REMARK 40.1. Formulas (40.4) and (40.5) tell us that, with the growth of
A, the greater the number a the faster the voltage and the current de-
crease. Thus a is called a damping coefficient. The number eis
called as the deformation coefficient, since if e = 0 then as A in-
creases the wave is only damped and not deformed at all.
* Compare with that on pg. 260 of J. Mikusinski [5].

Chapter XI
Heat Equation
§4l. THE TEMPERATURE OF A HEAT-CONDUCTING BAR

Let us imagine that a bar of lengthAO is placed along the A-axis,
the abcissa of the left end of the bar being A = 0 and the right end
A A ' Let k denote the heat conductivity, c the specific heat, and
O
p the mass density of the bar. Furthermore, let the lateral surface of
the bar be perfectly insulated so that heat can flow in and flow out only
through the ends of the bar. If we denote by z(A,t) the temperature
at the point of the bar at abcissa A at the instant t, then the heat
equation in the bar is
(41.1)
Assume that at the initial instant t ° the temperature throughout

the bar is 0, i.e.,
Z(A,O) = 0 (41. 2)
and that the temperatures at the ends of the bar are given by
(0 ~ t < 00). (41. 3)
By (41.2), we have Zt = sz so that the hyperfunction equation of

(41.1) can be written
2
Z"(A) = a sz(A). (41.4)
In the general solution of (41.4), given by
zeAl ae -aA IS +
be +aA IS (a and b € CIC) ,
145
146 XI. HEAT EQUATION
we shall choose a and b in such a way that (41.3) is satisfied:
a+b = vI
-CXA/S CXAOIS (41.5)
ae + be
Hence
-cxAolS -zcxAolS -cxAolS
vl-vZe -e vI + e V
z
a = ~--='---- b (41.6)
-zCXAOIS -zcxAOIS
I - e I - e
and so
-cxAIS -CX(ZAO-A)IS
.::.e -......;:e _
Z(A) vI
-CX(AO-A)IS -CX(AO+A)IS
e - e
+ .:::------::....-.---- v Z (41. 7)
-ZcxAolS
I - e
Furthermore, the fact that the hyperfunction solution of (41.3)-(41.4) is

uniquely determined by (41.7) shall be proved as in Proposition 30.
Now we have the series expansion
~ (-CX(ZVAO+A)IS -CX(Z(V+l)AO-A)IS)
zl (A) = vI' L e - e ,
v=O (41. 8)
Zz (A) v Z·
~ (-CX((ZV+l)AO-A)1S
L e - e
-cx((Zv+l) AO+A) IS) ,
v=o
since
-ZcxAolS -zcxAovlS
+ e + ••. + e + ...
-zcxAolS
I-e
Hence, by
l exp(_As l / Z)
s
= {cerf~}
zit
= {~ ro
Ii J A/Z/t
e-VZdV}
'
(Z7.13)
[Ell
we have, when vI =--
s
42. An Infinitely Long Bar 147
El {
~
L-
2 fa (2(V+l)A O-A)/2/t e _V 2
dv
}
v=O.r,r a(2VA +A) 12/t (41.9)
o
(0 ~ A < A and 0 < t < ~).
O
[E 2 ]
Similarly we have, when v 2 = {E 2} = ---s--
(41.10)
(0 < A < A and 0 < t < ~).

O
Thus
lim zl(A,t) 0,
t+O (41.11)
Hence the initial condition (at t = 0) and the boundary conditions (at
A=0 and atA = A ) are satisfied as above, although Z(A,t) is not
O
defined as a function of (A,t) at t = O. The same remark applies to
z2(A,t).
§42. AN INFINITELY LONG BAR

We shall prove the following proposition concerning an infinitely
long bar.
PROPOSITION 37. Let a hyperfunction solution Z(A) of
(a > 0 and 0 ~ A< (0) (42.1)
be given by
Z(A) = ae -aAlS +
b eaAvS (a,b € CIC) (42.2)
in such a way that it satisfies
Z(A,O) = 0 (0 < A < (0), z(O,t) vI (t) € e[O,oo) (42.3)
together with
IZ(A,t)1 < met) (0 ~ A < 00, 0 ~ t < (0) . (42.4)
Here Z(A,t) is a numerical-valued function in the domain
o< A< 00, o< t < 00.

Then we must have
-etA IS
Z(A) = vIe Zl(A). (42.5)
PROOF: From (42.2), we have
b = Z(A) e -etA IS - ae -2etAIS (42.6)

(0 < A < 00 and 0 < t < 00) (42.7)
by (36.11)'. Hence, by (42.4), we obtain
lim Z(A,t)e-etAIS = 0, lim ae- 2etAIS = O.

A->OO A-+-O
Thus, by (42.6), we must have b O. Therefore, by (42.2) and z(O,t)

vl(t), we obtain (42.5).
COROLLARY. I f we further assume that
[E l )
v = {E }
l s
we obtain, by (27.14),
2
Z(A,t) = El{cerf ~} = E {--
2 [ }
e _v dv. (42.8)
2vT 1 Iii etA/21t
Conversely, it is easy to show that (42.8) satisfies the assumptions

of Proposition 37 if we define
Z(A,O) = lim Z(A,t) = O.

tfO
AN EXAMPLE OF THE APPLICATION OF (42.8). Let us imagine a long bar

of silver with initial temperature O· Centigrade; this bar is constantly
heated at one end (A = 0) up to the temperature 100· Centigrade. This
example is found in J. Mikusinski [5), p. 229.
We have the following physical constants of silver:
specific heat c = 0.055

conducti vity k 1. 01
density p 10.5
and so
43. A Bar Without an Outgoing Flow of Heat 149
a =10.055 x 10.5/1.01 0.756 (approximately).
Hence
§43. A BAR WITHOUT AN OUTGOING FLOW OF HEAT

We shall solve the equation
(0 < A < A '

O
0 < t < ~) (43.1)
subject to the conditions
Z(A,O) = 0 (0 ~ A ~ A )
O
(43.2)
{ z(O,t) = vI (t), ZA(AO,t) o (0 ~ t < ~).
Here ZA(Ao,t) denotes the amount of heat flowing through the point A
O
at the instant t. Thus the assumption ZA(AO,t) =0 means that the
right end of the bar, like the whole of its lateral surface, is perfectly
insulated from its environment.
The general solution of (43.1) is
(a,b E C/C);
and a and b shall be determined from the boundary condition
r z(O) a+b = vI'
l ZA (AO) -aalS e
-ClAOIS
+ balS e
aAOIS
O.
That is,
-2ClA IS
O
vI vIe
a b
-2ClA IS
O -2ClA OIS
I+e I+e
Hence we obtain
Z(A)
Therefore, by using
00
I + L (_l)V
v=l
we obtain the solution zeAl, expanded into a series. The formula obtained
in this way will differ from (41.8) only in signs of the terms of the in-
finite series. We shall not write down the details.
§44. THE TEMPERATURE IN A BAR WITH A GIVEN INITIAL TEMPERATURE

FOURIER'S METHOD. Imagine that the temperature distribution in a
bar of length Ao is given at the initial instant t = 0 by
zeA,O)
so that
Zt(A,t) = sZ(A,t) - [z(A,O)) = sZ(A,t) - <P(A),
where <P(A) is the scalar mUltiplication operator.

Thus we are concerned with solving the hyperfunction equation
(44.1)
under the boundary condition
Z(O) = 0, (44.2)
REMARK 44.1. In order to apply Proposition 29 to obtain a (special) hyper-

function solution of (44.1), we shall extend the numerical-valued continu-
ous function <P(A) to a numerical-valued continuous function defined for
A ~ 0 also.
EXAMPLE 44.1. We take the case
¢CA) = (44.3)
where <P(A) is naturally extended to A ~ 0 also. Then the equation is
2
z"(A) - a sZ(A) -a 2. mrA
sln- (a > 0). (44.4)
Ao
Since the right hand term vanishes for A 0 and A A ' we shall tenta-
O
tively substitute
')
Z (A
. --A--
= a SIn nTrA (a E CIC) (44.5)
o
44. The Temperature in a Bar With a Given Initial Temperature 151
for zeAl in (44.4). Then
2 2
n 1! 2 2
-a - - aa s -a
2 -
Aa
so that we obtain
I
a = -----2--2 = {exp(-n 2 B2 t)}
s+n B
Hence we have a solution of (44.3) satisfying (44.2):
(44.6)
PROPOSITION 38. The solution of (44.2)-(44.3) is uniquely determined by

(44.6) .
PROOF: Assume that there exists another solution ~(A) of (44.2)-(44.3).

Then
weAl zeAl - ~(A),
where zeAl is given by (44.6), satisfies
2
w"(A) - a SW(A) = 0, w(O) = 0, W(A )
O
o.
To prove that this weAl is 0, we put
WCA) = Ca, b E C/C)

and choose a and b in such a way that W satisfies the boundary condi-
tion
-aA /5 aA O/5
W(O) a+b = 0, ae O + be O.
Then a b = 0, because the determinant
is not zero. For, i f
-2aA /5
O
I = e
or, more generally,

(n = 1,2,3, ... ). (44.7)
This contradicts
[3
e ->./S ~;Z 1(6/rre ) h. 3] (44.8)
Indeed, for any f(t) E C[O,~), we have, by (44.7)-(44.8),
-20>' nlS I
If(t)j = e
I a {f(t)}
~ {3 1(6/rre 3) 1 2 ft!f(U)!dU} (n 1,2, ... ).

(20)'On) a
This is absurd, as is seen by letting n ~ 00.
A CONSEQUENCE OF EXAMPLE 44.1. Let ~(>.) be given by

~
,1,(')
~ A = I;'L ~n
•
sin nrr>.
-->.-- (44.9)
n=l a
where
(44.10)
Then the equation

~
2 I;' . nrr>.
z"eA) - isz(>.) -0 L ~ Sin -->.--,
n=l n a
(44.11 )
{
z(O) 0, zeA o) a
is solved by
2 2
{L
00
z(>.) = {z(>.,t)} = ~n exp(- n rr 2 t) sin

2
~rr>. }. (44.12)
n=l 0 >'0 a
PROOF: By virtue of (44.10) and
a~ ex p(- n~~~ t) ~ 1, ISin ~I ~ 1, Icos ~rro>'1 ~ 1, (44.13)

o A
O
a
the right-hand side of (44.12) and its term-wise differentiations
and
are all uniformly convergent on A ~ AO' 0 < t. 0 ~

2
Therefore, (44.12) and its differentiations a/aA, a /aA 2 and a/at
are obtained by term-wise differentiations of the right-hand side of (44.12).
Hence, by Example 44.1, we easily prove that (44.12) is a solution of
(44.11). Moreover, the uniqueness of this solution (44.12) is proved
similarly as in Proposition 38.
REMARK 44.2. The solution of the heat equation by the above Fourier's
method is customary in many textbooks, so we shall give only one other
example of the method, in the next section.
REMARK 44.3. For the equation (44.1)-(44.2), we may obtain a hyperfunc-

tion solution by making use of Proposition 29 as it was stated in Remark
44.1. However, it would sometimes be difficult to prove the smoothness
(or the differentiability) of the hyperfunction solution. For this reason,
Fourier's method is sometimes easier to handle.
§45. A HEAT-CONDUCTING RING
We shall discuss the temperature of a heat-conducting ring, given the

temperature distribution at the initial instant t = O. It will be as-
sumed that for this ring there is no external flow of heat in or out.
If the length of the ring is AO' we can regard it in discussion as
a bar of length AO and consider the equation
(45.1)
where the function ~(A), given in the interval [O,Aol and thought of
as periodic with period AO' represents the initial distribution of tem-
perature.
We thus assume that ~(A) is expressed as
~(A) = -f
CL
+ L
00
n=l
(
CL
n
cos 2~1!A
0
+ B sin 2n1!A)
n AO
(45.2)
so that it will be natural to assume that the hyperfunction solution zeAl

of (45.1)-(45.2) is given by
zeAl 2n1!A
cos ---- +
. ----
b SIn 2n1!A) (45.3)
A n A
o O
where an bn E C/C
and (n = 0,1,2, ... ).
In order to follow the reasoning of the preceding section, we shall
assume that
(45.4)
Then we obtain
CLn
a = ---;;2;""""'2' b
n
(45.5)
n s+4n 6
so that
zeAl {z(A,t)}
(45.6)
PROPOSITION 39. (45.6) is the uniquely determined solution of (45.1) which

is periodic in A with period AO'
PROOF: Assume the contrary and let ~(A) F {z(A,t)} be another solution
of (45.1) which is periodic with period 1. '
0
Then
zeAl = zeAl - zeAl Fa

is a solution of
-zIICA) = CL 2sZ(A)
- (45.7)
and satisfies
(n = ±1,±2, ... ).
As a solution of (45.7), we must have
-(')
Z A = ae -CLAIS + beCLAIS (a,b E C/C).
By the periodicity of Z(A), we have
Z(0) = z(A O) = z(2A O),
i.e. ,
-CLAOIS CLAOIS
z(O) a + b = z(A ) ae + be , (45.8)
O
and
-2CLA IS 2CLA IS
a + b = Z (21. ) O + be O (45.9)
z(O) ae
0
hold.
Since Z(A) i 0, we must not have abO. Let a i O. Then,

from (45.8),
because we have
for y > 0 and for y < O. (45.10)
(The proof of (45.10) is as follows: If y > 0, then eYIS =I implies
(see (27.5)), contrary to the Proposition 5 (§4).)

Now, as above, we have, from (45.9),
-2ClA IS
~ = .:,I_---::e:- O _
a 2ClA lS
O
e - 1
and hence
-ClAOIS -2ClA IS -ClAOIS -ClAOIS

1 - e O
1 - e 1 - e (1 + e )
ClAOIS 2ClA OlS ClAOIS ClAOIS
e - I e - I e - I (e + I)
proving, via (45.10), that
-ClAOIS
1 =1 + e and so
ClA/S
e + 1
This implies
1 =
contrary to (45.10).*
*The above "uniqueness proof" may be compared with that given by J.

Mikusinski [5], p. 245-246.
§46. NON-INSULATED HEAT CONDUCTION

So far, we have assumed that the bar is perfectly insulated from
the environment. In this section, we assume that the heat loss at each
point of the bar is proportional to the temperature difference at each
point from the temperature of the environment, assumed to be constant at
O· Centigrade. We then have
(0.>0,6>0).
Thus, assuming that the temperature of the bar at the initial instant
t 0 is O· Centigrade, we have the equation
(46.1)
with the boundary condition
z(O) = v, (46.2)
This equation is the same as the telegraph equation we have studied

in §38, the cable without self-induction, so we do not go into detail.
However, for the convenience of the reader, we shall consider the
case of an infinitely long bar for which v = {w} = ~
s
(46.3)
that is,
Z(A,t) = wfe-AIB Erf (IBt _ ~)

2\. 0. 21t"
(46.4)
A
- e ,!6 Erf(/Bf + ~) + 2 cos h(A!B)}.
0. 21t"
Answers to Exercises
§2.
(a) {t}
2
(6) {t /2}
(y) {-4 sin t - t}
(0) {2 sin t - 2 cos t + 2}
§9.
1. 1 -1 l' { (z-l) -1'(Z-~)}

lim (z+I)(z-2) =:2' 1m 2 -
z-+1 z-+1 (z-l) (z+l) (z-2) 2(z-l)
I" z(z-l) -1 1 -1
1m1 2(z-l) (z+l) (z-2) = 4' lim
z-+ z-+-1 (z-l) 2(z_2) = 12'
lim 1 1
z-+2 (z-I)2(z+l) =3
Hence I 1 [-6 -3 -1 4 ]
(5-1)2(5+1)(5-2) = 12 (5_1)2 + s::T + s.T + s:2
2. I
(s_a)2_ 62 =
-1 (I
ze (s-a)+6
I)
- (s-a)-6
3. s-a 1(I I)
( s-a ) 2- 62 = 2" s-a+6 + s-a-6
§10.
-t 2
1. e (t - 2t + 1)
2 -t
2. yet) (t-l) e
3. Yet)
1
-2"+TQe
1 2t 2 cos t _ .!.
+ 5" 5 51' n t
157
158 ANSWERS TO EXERCISES
1 t 11 4t 3 5 .
4. Yet) - 2 + e - 34 e - 17 cos t + 17 S1n t,
2 t 22 4t 4 1 .
z (t) - 3 e + 51 e + 17 cos t - 17 S1n t
t t t t t
S. x(t) = 2 - e , yet) = -2 + 4e - te , z (t) = -2 + Se - te
28 3t - e -t 1 t 28 3t 1
+ e -t - "9 t
6. x(t) = 9' e - "9 - 3' yet) = 9" e -3
2t
7. x(t) 2e - 2 cos t - 3t sin t
yet) sin t + t cos t - 2t sin t
§ 11.
1. Solvable if and only if cos 2~a F 0; and then
- 1 .
Y(t) - a cos 2~a S1n at.
at -at
e - e
2. Yet)
§13.
at
= _e_ Erf(ICit)
fci'
§16.
-a/s ~ n n+2 {~ n n+l }

1. 2' e
I L (_l)n ~ = L (_l)n _a~t~~
s n=O n. n=O n! (n+l)!
= {(ItTci) L
n=O
I a/s '"' a nhn+1/ 2
2. -- e
IS
= L n'.
n=O
'"' nn-l/2 }
{ n~o n(n-l)"'1(n-~/~)(n-3/2)' .. (1/2)r(1/2)
Answers to Exercises 159
00 22n n n-1/2 }
- {
n~o (2n)~2~-1)' .. 1
{_l_ cosh (2M)}
11ft
3. See (16.13).
§20.
m m 00 mk
(eL) Since e -8/s = e -8h = I + L\ (_8)k _h,
k=l k.
m 00
De- 8/s = DI + L (-8l k ~ (Dh mk )

k=l k.
o + I (_8)k ~k. (-mk hmk+l )

k=l
00 k-l hm(k-l) m1
k~l (-8) (k-l)! m8 h +
-8/s m mB
e m+ 1
s
(8) Ql - (-2+2b+2)s + c 2bs + c

y 52 + 1/4 52 + 1/4
b + ci b - ci
s + i/2 + s - i/2 '
so that
y = C(s + i/2)b+ci(s _ i/2)b-ci (i !=T) .
.@..
1.
2.
§36.
(eL) _1_ = {.!. sin at} < .!.

2 2 a a
5 +a
The function on the right-hand side takes its maximum value
at
Formulas and Tables
I. Special Functions
(§ 12)
f(h1) = Af(A), fen) = (n-1)! (n = 1,2,"')
B(A,V) = f(A)f(V) J1 t A- 1(1_t)V-1 dt (Re A > 0, Re V > 0) (§ 12)
r:
r(A+V) 0
v2
f(1/2) = IW = 2 e- dv
2
Erf t = --
IW
It
0
e- v 2dv (§ 13)
v2
Cerf t = ~[ e- dV = 1 - Erf t
Ii t
~ n+2k
J (t) = L (_l)k n
t (n = 0,1,2, .. ·) (§lS, §20)
n k=O 2 +2k k !(k+n)!
~ n+2k
J (it) = in Ln
_
~-;:;i-t (n = 0,1,2,"')
n k=O 2 +2k k ! (k+n)!
k
La(t) = L (n+a)
00
1:!l- (§20)
n k=O n-k k!
(l+z)Y = exp(Y log(l+z)) for complex numbers Y and z
= L (r)Zk (convergent for Izl < 1) (§l4)

k=O
160
Formulas and Tables 161
II. Formulas of the Operational Calculus
C = C[O ,00) (§ 1)
CH = {i; h
f E C and k = 1,2, ... } (§ 3)
C/C: convolution quotients !g (f,g E C wi th g -F 0) (§I8)
{aCt)} + {bet)} = {aCt) + bet)},
{a(t)}{b(t)} = {f: a(t-U)b(U)dU},

a{f(t)} = {af(t)}
h{f(t)} = {f: f(U)dU}
s{f(t)} = {f' (t)} + [f(O)], [f(O)] = s{f(O)} (§S)
sn f = fen) + sn-l [f(O)] + .•• + s [f(n-2) (0)] + [f(n-1) (0)]
Ta{f(t)} = {eatf(t)}, TaR(s) R(s-a)
Ta £. = Tab (§20, §37)

a Taa
Df = {-tf(t)}, D(fg) (Df) g + f(Dg)

D £. _ (Db)a - b(Da) (§ 19)
a - a2
( e -AS f(A) dA = { }
f(t) (§26)
a
ax e
-AS
= -se
-As
(§23)
a _As 1/ 2 1/2 _As 1/ 2

axe -s e , (§26)
III. Tables of Hyperfunctions ? C/C

f = h = {I}, I = i (f(t) * 0) (§S)
I n t n-1}
n=h = { (n-I)! (n = 1,2, ... ) (§ 6)
s
I
"""I = h = rCA)
A {t A- 1} (Re A > 0) (§13)
s
162 FORMULAS AND TABLES
hy +n f( +n)-l t y+n-1
hY =-----
n = Y -1 n -1 for complex numbers Y
h fen) t
(integer n ~ such that Re(y+n) > 1) (§ 19)
D(I-ah)y = Y(I_ah)y-1 ah 2. where
(l-ah)Y = I (~)(_a)khk
k=O
(§19)
D(s-al)Y = D (I-ah) Y = y(s_al)Y-1 (§ 19)

hY
.l. = h1/2= {_1_} (§ 13)

IS ffi
_1_ = {eat} (§6)
s-a
A1
-1- = {t- --- eat} (Re A > 0) (§ 13)
(s-a) A f (A)
1 {I/iite -at}
,;s.;:a=
(§ 13)
_1_ =
s .;s.;:a
{...!.. Erf
Iii"
Iat} (a > 0) (§ 13)
sY = h- Y• in particular sO = hO = 1 and Ds = 1 (§ 19)
(s-al)Y(s-al)O for complex numbers Y and 0

(§ 19)
= (s-al) Y+o
hYh o = hY+o (§19)
I at
2 2 { Be sin 8t}
(s-a) +8
[(s-a) +13 ]
222 {;:~ [i sin Bt - t cos Bt]}
s-a (§9)
(s_a)2+ 82
1
2 2
(s-a) -13
= {i eat sinh Bt}
s-~ 2 = {eat cosh Bt}
(s-a) -13
I
2 2 1/2 {JO(iat)}
( 5 -a )
(§16)
((5
2+a)
2 1/2 -5) n = {nan
-t- In(at) } (n = 1,2, ... )
(n = 0,1,2, ... )
~ e-A/s = {1(t7A) J (21Xt)}

5
1
(§ 16)
- I e -A/s = {1
- cos 2Yl\t r=}
IS m
-I e AS
IS
e
-AS (§ 22)
exp(-A.fS) ={ A__ eXP(-A 2/4t)}

2;';;3
1.... exp(-A.fS) = {_1_ 2

eXP(-A /4t)} (§27)
IS m
.!..5 exp( -A.fS) = {Cerf(A/21t)}
2 2 1/2
exp(A[s-(S +a) ])
2 2 1/2
exp(A[s-(S +a) ])
(s2+i)1/2
(§28)
2 2 1/2
exp(A[s-(S -a) ])
164 FORMULAS AND TABLES
2 1/2 -a),
exp(),[s-(s +2as) j) = e -
exp(),[s-(s 2+2as) 1/2 j)

(s2+ 2as )1/2
2 2 1/2
exp(-),(s +a) )
(§28)
2 2 1/2
exp(-),(s -a) )
o~
o<
2 1/2
exp(-),(s +2as) )
2 1/2
(s +2as)
2 2 1/2
exp (-), [(s-a) +8] )
2 2 1/2
exp(-),[(s-a) +6]
((s_a)2+ B2)1/2
)
o ~ , , J
(There are also those obtained by substituting ~ B for B in the

last two formulas above.)
(A > 0)
(§ 36)
I
IS e -AIS ~ [= I1]h
(v2/~e) (A > 0)
References
1. Berg, E. J.: Heaviside's Operational Calculus as Applied to Engineer-

ing and Physics, McGraw-Hill Book Company (1936).
2. Doetsch, Gustav: Einfuhrung in Theorie und Anwendung der Laplace-
Transformation, Birkhauser Verlag (1958).
3. Erdelyi, Arthur: Operational Calculus and Generalized Functions,
Holt, Reinhart, and Winston (1962).
4. Heaviside, Oliver: Electromagnetic Theory, I-III, London (1893-1899).
5. Mikusinski, Jan: Operational Calculus, Pergamon Press (1959).
6. Mikusinski, Jan: The Bochner Integral, Academic Press (1978).
7. Krabbe, Gregers: Operational Calculus, Springer-Verlag (1970).
8. Okamoto, Shuichi: A simplied derivation of Mikusinski's operational
calculus, Proc. Jap Acad. ~, Ser. A, No.1 (1979), 1-5.
9. Yosida, Kosaku-Okamoto, Shuichi: A note on Mikusinski's operational
calculus, Proc. Jap. Acad. ~, Ser. A, No.1 (1980), 1-3.
10. Yosida, Kosaku: The algebraic derivative and Laplace's differential
equation, Proc. Jap. Acad. 59, Ser. A, No.1 (1983), 1-4.
11. Yosida, Kosaku-Matsuura, Shigetake: A note on Mikusinski's proof
of the Titchmarsh convolution theorem, to be published in the Con-
temporary Mathematics Series of the Amer. Math. Soc.
166
Propositions and Theorems in Sections
Proposition Section Theorem Section
1 1
2 2 2 3
3 3 3 5
4 3 4 5
5 4 5 6
6 7 6 7
7 8 7 8
8 9 8 8
9,9' 13,13 9,9' 9,9
10 13 10 12
11 15 11 13
12 18 12 14
13 19 13 15
14 19 14 16
15 19 15 16
16 20 16 17
17,17' 22 ,22 17 19
18 23 18 20
19 23 19 21
20 24 20 21
21 25 21 24
22 27 22 26
23 27 23 27
24 28 24 29
25 28
26 28
27 28
28 29
29 29
30 30
31 31
32 36
167
168 PROPOSITIONS AND THEOREMS IN SECTIONS
Proposition Section
33 36
34 36
35 36
36 39
37 42
38 44
39 45
Index
Algebraic derivative, 53-59 Fractional integration, 35

Fractional powers, 34-38
Bessel differential equation,
61-63 Gamma function, 32
Bessel functions, 40-46
Gauss differential equation, 64
Beta function, 32
General powers, 38
Binomial theorem, 39
Generalized continuity, 81
Boundary value problems, 29-31
Generalized derivative, 10, 79
Confluent hypergeometric differ- Generalized functions, 8-9

ential equation, 63-64 Generalized integral, 92
Convolution, 1, 76 Generalized Laplace integral, 93
Convolution quotient, 12
Convolution ring, 3 Heat equation, 106, 145-156
Convolution theorem, 11-12, 47 Heaviside's unit function, 74

Hyperfunction, 8-9, 50-51
D'Alembert method, 63, 118-123
Impulsive force, 77-78
Delta function, 78
Differentiation operator, 9, 76 Integration, 3
Inverse (see "multiplicative inverse")
Electric circuits, 28-29 Initial conditions, 14, 108
Entire function, 33 Initial value problem for ODEs, 14,
Error function, 37 22
Euler's integrals, 32
Laguerre polynomials, 65-66
Exponential hyperfunctions, 74-
76, 83-86, 94-105 Laguerre differential equation, 64-
66
Fourier's method, 150, 152, 153 Laplace transform, 25, 53

Logarithmic hyperfunction, 99, 102-105
Fractional differentiation, 37
169
170 INDEX
Moments, Mikusinskits theorem of,

70, 72-73, 89
Multiplicative inverse, 13, 52, 84
Multiplicative unit, 7, 51
Partial fraction decomposition, 18
Rational functions, 17-18
Scalar multiplication, 9-10

Shift operator, 74-76, 83-86
Telegraph equation, 106, 124-144

Titchmarsh convolution theorem, 12,
47
Thomson cable, 128
Uniqueness of solution, 23, 84, 98,

99, 108, 114, 120, 129, 146,
147, 151, 154
Wave equation, 106, 108-123

Weierstrass' theorem, 70-71
Zero factors, 15-16

Applied Mathematical Sciences
cant. from page ii
36. Bengtsson/Ghil/Kallen: Dynamic Meterology: Data Assimilation Methods.

37. Saperstone: Semidynamical Systems in Infinite Dimensional Spaces.
38. Lichtenberg/Lieberman: Regular and Stochastic Motion.
39. Piccinini/StampacchialVidossich: Ordinary Differential Equations in R".
40. Naylor/Sell: Linear Operator Theory in Engineering and Science.
41. Sparrow: The Lorenz Equations: Bifurcations, Chaos, and Strange Attractors.
42. Guckenheimer/Holmes: Nonlinear Oscillations, Dynamical Systems and
Bifurcations of Vector Fields.
43. OckendonlTayler: Inviscid Fluid Flows.
44. Pazy: Semigroups of Linear Operators and' Applications to Partial Differential Equations.
45. Glashoff/Gustafson: Linear Optimization and Approximation: An Introduction to the
Theoretical Analysis and Numerical Treatment of Semi-Infinite Programs.
46. Wilcox: Scattering Theory for Diffraction Gratings.
47. Hale et al.: An Introduction to Infinite Dimensional Dynamical Systems - Geometric
Theory.
48. Murray: Asymptotic Analysis.
49. Ladyzhenskaya: The Boundary-Value Problems of Mathematical Physics.
50. Wilcox: Sound Propagation in Stratified Fluids.
51. Golubitsky/Schaeffer: Bifurcation and Groups in Bifurcation Theory, Vol. I.
52. Chipot: Variational Inequalities and Flow in Porous Media.
53. Majda: Compressible Fluid Flow and Systems of Conservation Laws in Several Space
Variables.
54. Wasow: Linear Turning Point Theory.
55. Yosida: Operational Calculus: A Theory of Hyperfunctions.

Operational Calculus - A Theory of Hyperfunctions - Yosida

Uploaded by

Document Information

Original Title

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

Operational Calculus - A Theory of Hyperfunctions - Yosida

Uploaded by

Copyright:

Available Formats

Applied Mathematical Sciences I Volume 55

Applied Mathematical Sciences

1. John: Partial Differential Equations, 4th ed.

(continued on inside back cover)

Springer Science+Business Media, LLC

AMS Classification: 33A99, 44A40, 44A45

Library of Congress Cataloging in Publication Data

Tîtle ofthe original Japanese edition: Enzanshi-oo, Vniversity of

© 1984 by Springer Science+Business Media New York

ISBN 978-0-387-96047-0 ISBN 978-1-4612-1118-1 (eBook)

In the end of the last century, Oliver Heaviside inaugurated an

However, the use of such integrals naturally confronts restrictions con-

If f(t) and get) are continuous functions defined on [O,~)

such that the convolution f~ f(t-u)g(u)du = 0, then either

The aim of the present book is to give a simplified exposition as well

Let f(t) be continuous for o< t < GO

As for the extension, I should like to mention the following point 30 •

(l-z)Y L (~) (_z)k (convergent for Izi < 1).

Then, by virtue of the general power (s-a)Y, we can solve algebraically

by the author jointly with Matsuura and Okamoto, respectively. Masuda

Part I. INTEGRATION OPERATOR h AND DIFFERENTIATION OPERATOR s

Chapter II. INTRODUCTION OF THE OPERATOR s THROUGH THE RING CH 5

Chapter III. LINEAR ORDINARY DIFFERENTIAL EQUATIONS WITH

7. The Conversion of the Initial Value Problem

Chapter IV. FRACTIONAL POWERS OF HYPERFUNCTIONS h, s AND __1__ 32

Chapter V. HYPERFUNCTIONS REPRESENTED BY INFINITE POWER

14. The Binomial Theorem 39

PART II. LINEAR ORDINARY DIFFERENTIAL EQUATIONS WITH LINEAR

CHAPTER VI. THE TITCHMARSH CONVOLUTION THEOREM AND THE

17. Proof of the Titchmarsh Convolution Theorem 47

CHAPTER VII. THE ALGEBRAIC DERIVATIVE APPLIED TO LAPLACE'S

PART III. SHIFT OPERATOR exp(-As) AND DIFFUSION OPERATOR exp(_As l / 2) 74

CHAPTER VIII. EXPONENTIAL HYPERFUNCTIONS exp(-As) AND exp(_As l / 2) 74

22. Shift Operator exp(-As) = e -AS .

CHAPTER IX. ONE DIMENSIONAL WAVE EQUATION 108

33. The Hyperfunction Equation of the Telegraph

CHAPTER XI. HEAT EQUATION 145

41. The Temperature of a Heat-Conducting Bar 145

ANSWERS TO EXERCISES 157

§l. CONVOLUTION RING

The totality of complex-valued continuous functions a(t), b(t), f(t)

(a*b)(t) = a*b(t) = I: a(t-u)b(u)du (0 ~ t < 00) , (1.1)

PROPOSITION 1. a*b belongs to C; i.e., a*b(t) is a continuous function

PROOF: Let 0 be any positive number. We shall prove

lim a*b(t+o) = a*b(t).

For fixed t > 0, it is clear that

lim II 0 = ft a(t-u)b(u)du = a*b(u).

Also, for fixed t > 0 and 0 < 0 < 1, we have

A= max la(s)l, B= max Ib(u)l.

This proves lim IZ,o = 0, and so we have obtained

Similarly, we can prove lim a*b(t-o) a*b(t) for fixed t > O.

abet) = J: a(t-u)b(u)du (0 ~ t < GO). (1.1)'

For a,b € C, the sum a+b is defined by

(a+b)(t) aCt) + bet) (1. 2)

(aa)(t) = aa(t). (1. 3)

We have the following:

a+b = b+a, (a+b)+c = a+(b+c), (1. 4)

a(b+c) = ab+ac, a(a+b) = aa + ab, (a+B)a = aa + Ba, (1.5)

(Greek letters denote complex numbers.)

a(ab) = (aa)b = a (ab) , (aB)a = a(Ba), (1. 6)

ab = ba, a(bc) = (ab)c, (a+b)c = ac + bc. (1. 7)

(ab)(t) = J: a(t-u)b(u)du = _fO a(r)b(t-r)dr

= b(t-r)a(r)dr = (ba) (t) .

Chapter IV. FRACTIONAL POWERS OF HYPERFUNCTIONS h, s AND 1 32

(ab)(t) = ab(t) = I: a(t-u)b(u)du (0 ~ t < 00) , (1.1)

PROPOSITION 1. ab belongs to C; i.e., ab(t) is a continuous function

lim ab(t+o) = ab(t).

Similarly, we can prove lim ab(t-o) ab(t) for fixed t > O.