Professional Documents
Culture Documents
Operational Calculus
A Theory of Hyperfunctions
9 8 7 6 543 2
( e -pt f(t)dt.
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
Kosaku Yosida
Kamakura
July 1983
Contents
Page
Preface v
1. Convolution Ring 1
2. Operator of Integration h 3
viii
Table o{ Contents ix
Page
CHAPTER V. (cont.)
16. Hyperfunctions Represented by Power Series
in h 42
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
and we have
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
1
2 I. INTRODUCTION OF THE OPERATOR h THROUGH THE CONVOLUTION RING C
t+ o
IIz,ol ~ t
J ABdu ~ o·AB, where
THEOREM 1.
PROOF: We shall prove (1.7) only, since the rest are easy. By the sub-
stitution t-u = r, we obtain
f:[f:- V
a(U)b(t-V-U)dU]C(V)dV = ((ab)c)(t).
EXAMPLES OF CONVOLUTION
abet)
t
tee -1)
-t + e t - 1.
= 21 1.
t - 2 Sln t cos t.
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,
(0) {e t }3 {e t He t }2 =
1 2 t
{-zy t e }.
(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
PROPOSITION 3. Let
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
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
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
(3.3)
(3.4)
follow from
f € C and k € H} * (3.5)
J
f f' f' f
k+j(T=j(T+P
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'
,
(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
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
O
(n 1,2, ... ; h I). (5.1)
sh = hs = h lh = I = 1. (5.3)
(5.4)
f
For any --
n
€ C , we have
H (5.5)
h
PROOF:
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
Further.
(5.7)
5. Operator o~ Di~~erentiation 5 and Scalar Multiplication [a] 11
(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
n-m-l n-m n
= anI + (on_l h + + 0m+l h + (am-Bm)h + ... + (ao-Bo)h
f
(f,g € C and g # 0)
g
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))
.
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)
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
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
by virtue of
y(k) = sky _ sk-ly(O) _ sk-2 y '(0) _ ... _ sy(k-2) (0) _ y(k-l)(O). (7.4)
14
8. The Polynomial Ring o~ Polynomials in s has No Zero Factors 15
(7.5)
This is equivalent to
Multiplying by
i. e.,
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,~).
Let
pes)
and
Q(s) = ~s
n
+ ~
m-l
s + ••• + ~Os
o (~ # 0)
m m-l m
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)
(8.2)
~-~
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
ql (s)P2(s) = PI (s)q2(s).
Hence
PI(s)(q(s)PZ(S) - p(s)qZ(s)) = o.
Since PI(s) F0 as the denominator of the "fraction", we must have
Therefore
~ = ql(s)
pes) PI(s) means q(s)PI(s) = P(s)ql(s). (8. Z) I
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)
s ~ _ ql (s)q(s) ql(s)q(s)
ql( ) pes) - 0 pes)
s pes)
and so
(9.2)
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)
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.
~ = _..:Iq~(s~)_ _
pes) ml
(s-Zl) f(s)
then
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).
we obtain
(9.S)
20 III. LINEAR ORDINARY DIFFERENTIAL EQUATIONS WITH CONSTANT COEFFICIENTS
Next we obtain
(9.6)
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
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
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
222
{~ft ea(t-T)sin 6(t-T)e aT sin 6TdT}
[(s-a) +6 ] 6 0
= {ea~
26
(i sin Bt - t cos Bt)}.
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
_ ,_ (n-l)_ (10.2)
yeO) - YO' Y (0) - Yl,···,y (0) - Yn - l ,
(10.3)
Let us denote
(10.4)
(10.5)
* sinh t sinh(t)
10. Hyper!unction Solution o! the Ordinary Differential Equation 23
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)
+ ••• +
_ _I_f+tl&
y - pes) pes)
(l0.8)
= r
j=l
UfJl·
-
[Cju{(~~~~! eZjt}f + dju{(~~~~! e
Zjt
}].
Here
y'
(10.9) ,
-1
+ (a polynomial in t) + an F'(t).
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) •
I
UI
{t - Zjt}
(s_z.)u
=---,e,
(u-l).
J
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
,.. 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
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.
26 III. LINEAR ORDINARY DIFFERENTIAL EQUATIONS WITH CONSTANT COEFFICIENTS
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)
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
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
(s+a)k
O
y + --~:
(s+a) 2
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.
28 III. LINEAR ORDINARY DIFFERENTIAL EQUATIONS WITH CONSTANT COEFFICIENTS
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
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)
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
= ..L (-!.- __
2ia S-la
I )
s+ia
+ ~_I _ _ I )f
2IQ\s-ia s+ia
that is,
(11.4)
ii) the second case: sin ato O. In this case, we must have
t
foo sin a(to-u)f(u) =0
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
2
~'(t) - a yet) = 0, yeO) = 0, y(2'11) = 1. (11. 8)
ANSWER: We first solve the initial value problem
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
THEOREM 10.
f(A+l) = Ar(A) , (12.3)
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
(12.5) :
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
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)
l
ao
T
2 0 e- dT.
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
Therefore,
n-1 n-2
B(A,n) = -X- r;r B(A+2,n-2)
(n-l) (n-2) .. ·1 1
1.(>.+1) ... CA+n-1) (B(A,l) = r)'
n~~~}
n n-1}
h = { (~-1)! =
(n = 1,2, ... ).
n
We can extend h to hI. with Re A > 0 through
(13.1)
hn = sh n+1 .
{h Af(t)} = {f(A)
1 Jt0 (t-u) A-I f(u)du } € C (Re A > 0). (13.2)
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}.
(13.4)
(13.5)
36 IV. FRACTIONAL POWERS OF HYPERFUNCTIONS h, s AND _I_
s-a
11iEOREM 11.
A1
5
2 { t - at}{
h l'(X) e h
t~-l
r (~) e
at}
1 { at ft A-I ~-1 }
r(A)r(~) e 0(t-u) u du.
r(A)~(~) {f:(tW)A-1(t-tw)~-lteatdw}
1
rCA) r(~)
{tA+~-leat} f
1
wA-1(1_w)~-ldW
0
5
-A (Re A > 0) (13.7)
50 that
A
5- = 5{f:r(A)-luA-1dU} = 5{r(A+1)-l t A} = h A.
(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).
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.
38 IV. FRACTIONAL POWERS OF HYPERFUNCTIONS h, s AND _I_
s-a
hY = hy +n = {f(y+n)-lt y+n - l } *
1
hn {f(n)-lt n - l } (13.14)
(13.15)
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.
THE DEFINITION OF (l+z)a. Let a be any complex number. For any complex
number z with Izi < 1, we define
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),
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.
__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
+ •••
1 n
2
y"(t) + t y' (t) + (1 - 2)y(t) = 0, (15.1)
t
yet) + •.. +
(15.2)
2n r(n+1) '
I
In(t) = v=o r(n+v+1)r(v+1)
(_l)V (t)n+2V
2 (n = 0,1,2, ... ). (15.3)
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 + •" + + •••
2
n
- ~ y
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
CLOh
o+ CLlh + CLZh
Z+ ••• + CLnh
n
+ (16.2)
+ '" + CL
n
Tn:n-r
n-l
+ ••• ) € C
w (16.3)
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' ~
Hence
16. Hyperfunction~ Repre~ented by Power Serie~ in h 43
(16.5)
PROOF: We have
and hence
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
44 V. HYPERFUNCTIONS REPRESENTED BY INFINITE POWER SERIES IN h
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! -
I h3 h
sin -s = sin h = h -3T+ST-
S
... = {I - tZ
3!Z! + S!4! -
t4
..1 (16.6)
r22
(I s-+a-)- s)
n
(n 1 ,Z, ... ). (16.7)
~
(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~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
{ (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
f
n+l
(z) = ~ fZ zn/2+1(z- n/2 f (z))'dz.
nOn
(16.10)
(16.11)
-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!)
46 V. HYPERFUNCTIONS REPRESENTED BY INFINITE POWER SERIES IN h
_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).
= I a,B,l + I
a,B,Z
.
By the substitution
47
48 VI. THE T:J;TCHMARSH CONVOLUTION THEOREM AND THE CLASS C/C
<l(u,v) = 1
<l(t,T) ,
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
(17.7)
satisfies, for a = Re z ~ 0,
(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
=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.
(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).
(£gl)(fg l )·
f: f(t-t)tg(t)dt == O. (17.13)
Ito f(t-t)p(t)g(t)dt == 0
fo f(t-t)q(t)g(t)dt == 0
means (18.1)
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,
Z. f' _ fl
g gr - gr (18.5)
*Similarly as CH.
52 VI. THE TITCHMARSH CONVOLUTION THEOREM AND THE CLASS C/C
£.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.
(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)
53
54 VII. THE ALGEBRAIC DERIVATIVE
1
DDf_:=={-tf(t)} for
(19.3)
(Df) g - f(Dg) f
2 for g € CIC.
g
implies D i _- 1
D _f
(19.5)
g gl
PROOF: We have
(19.6)
because
Hence,
= J..Q!2..&. = Df .
g
Similarly we have
and so
proving that
(19.7)
(19.8)
(19.9)
PROOF: We have
D[a] = 0; in particular, DI 0, (19.10)
because
{-ta}h - {a}{-t}
2
h
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
f f1 f ( f 1)
(D f)T + k DT
PROPOSITION IS.
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
oa (Da)b - a(Db)
b b2
1
hn {r(n)-l t n-1} ,
(19.14)
n being any integer ~ 1 such that Re(y+n) > 1.
(19.15)
where
(19.16)
(iii) We have
(19.17)
58 VII. THE ALGEBRAIC DERIVATIVE
(19.18)
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.
_ 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)
(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
exp((y+o)log(l-az))
(1-az) y+o.
This proves
60 VII. THE ALGEBRAIC DERIVATIVE
(s-aI)Y(s-aI)O
(t - :~)
as a new independent variable. Thus we shall discuss the differential
equation
Z
-aZD(s Y - s[y(O)] - [y' (0)))
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
(ZO.3)
(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:
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
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) ).
(-ZaZ+bl)s - a l + bO
Z
aZs + als + aOI
EXAMPLE ZO.l. (THE BESSEL DIFFERENTIAL EQUATION). The equation is
(ZO.8)
Here we have
62 VII. THE ALGEBRAIC DERIVATIVE
Hence
satisfies (ZO.ll).
We have, in the case Re a ~ 0,
(-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
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
and from
we obtain
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 -
C hl-a-c+a+l(I_h)c-a-l
1
then
a 2 = 1. b 2 = O. a
l = -1. b l = I-a. a o = O. bo a+\.
Ql = (-2+1-a)s + -1 - a - \ +-
\
(20.22)
y /-s s s-1
(20.23)
because
(n+a) 1
n-k IT
66 VII. THE ALGEBRAIC DERIVATIVE
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),
1
g
(20.32)
.Qi. = III is equivalent to
2
Y (s-aI)
A
(20.33)
Dz = ~ e- Bls
ds
= e- Bls ~2 = ~
2' (20.34)
5 5
(16.13)'
so that, by (20.27),
{e
at J (2r'Bt)}
O
{eat}
(20.35)
Ql _ -2as + -2a I
=-5-+2' (20.38)
y 52 5
Hence, by (20.35),
(20.39)
20. Laplace's Ditterential Equation 69
(m = 1,2, ... ).
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)
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:
70 VII. THE ALGEBRAIC DERIVATIVE
(21.1)
we have
lim max If(t) - P (t) I
n
O. (21. 2)
n-- O~t~l
PROOF: Differentiating
(21.3)
n-l n
nt(t+u) = L p c tpun- p . (21.4)
p=O n p
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
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
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 .
~
- 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
we obtain
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
i IT (1 -x(t-u)
lim - exp(-e )g(u)du o
x-- 0
(21.13)
uniformly in t (0 < t ~ T - e:).
(1 - exp(_ex(t-u)))g(u)
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:).
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)
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 < ~,
(22.3)
and give
0, 0 < t < A )
e
-AS}
{f(t) = t A
= (22.4)
fo-
S
{ f(u)du, 0 ~ A~ t < 00
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
{
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.
so that e- os = I.
PROPOSITION 17'. For A > 0 and \.I > 0, we have
because
hl(A,t-u)hl(~'u) > 0 only if both 0 < A < t-u and 0 < ~ < u
are satisfied.
~
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) ,
(22.9)
(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
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
£
£
° £
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
_ ill. (_I __ I )
- 2iw s+a-iw s+a+iw
23. Hyperfunction-valued Function f(A) 79
(A > 0)
iF
c
0 of C/C and ?(A,t)
for which
by bd E C, we obtain
that is,
t It
I° ad{t-u)f{A,U)du = ° bc{t-u)f{A,u)du.
A A
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)}.
PROOF: We have
= h{hl(A,t)}
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)}
"
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)} = 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
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
+ 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.
dc h { d Le., fl (A) = O.
dA = ha ax bet)} = 0,
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)
2
-AS -1 h I (24.2)
(e ) = {hl(A,t)} = e- AS € CIC (A ~ 0).
Hence, defining
we obtain
(A > 0) (24.3)
and furthermore
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
Since a-A> 0,
de-(a-A)s - (a-A) s
-se (by (23.3)),
dCa-A)
i.e. ,
de AS _ as de-(a-A)s
"""'dr" - e dA
eaSse-(a-A)S = seAS
(by (23.4).
OS
e ((23.3)).
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,
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
AS
for f(A) e (24.12)
AS
§25. EXAMPLES OF GENERALIZED LIMIT. POWER SERIES IN e
(25.2)
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. ,
t It
I°(ab l ) (t-u)f(u)du = °(alb) (t-u)f(u)du,
A V
i.e. ,
PROOF: Choose T > 0 and let 0 < t ~ T. Then max If(t)1 ~ M for some
O~t~T
finite M > O. Assuming
we obtain
(0 ~ t ~ T)
Hence
2 h 2
generalized lim (~ - I)
n
s {lim (-
n
- h )} = s 2 (-h)
2
= -I. (25.6)
n_
* *
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
~:: 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
-AS -1
(I - Be ) = s 2{ fB(A,t)} (25.11)
+ •••
and
Therefore, we obtain
(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 +
get + A) = g(t).
e
-kAS
f = {O' 0 ~ t < kA }
€ K.
f(t-kA), 0 < kA < t
:}.
f - e (h +Ah)
Thus
,) i
4+-,
1<"
5.
-AS
§26. f
00
a a ~
=b = ~1
A
we obtain
26. f~ e-ASf(A)dA = {f(t)} For {f(t)} E C 93
o
Thus
2
ab l {fA F(A,t)dA} = alb {fA 2 ~(A,t)dA}
Al Al
(26.3)
2
s {hI (A,t)f(A)}
Hence, by Theorem 2,
s2{f:(t-A)f(A)dA}
~ -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
( e-Ptf(t)dt or
2
§27. EXPONENTIAL HYPERFUNCTION exp(_As l / 2) = e- Asl / (_00 < A < 00).
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
PROOF: We have, by (13.3),
By the substitution
*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)
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
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
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)· ~ .
f(A)f(~-A) =a
b
E C/C
(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
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"
{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
sf(A) .
ZAA = Zt'
Thus exp(_As l/2 ) is sometimes called the heat operator or the diffusion
operator.
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
fl (A) = ba "
f'(A)
a"f(A,t)}
= ba {at is f(A)
= is 0a "f(A) = is ba {f(A,t)}.
"
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
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
t
nwu
g(t-u)e du
I f
~ 0 k(u)du (n = I, z, ... ). (Z8.8)
(28.11)
we obtain
+ 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,
f(A) =
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
(28.15)
(28.16)
I - {>.. I (_l)k
2
a + t 2k k
(t+2A)k}
21+2k ! (k+1) !
k=O k
COROLLARY 1.
r'22
exp (A (s - Is- +a-)) = {J (a 1t 2+2At)} (28.17)
125 + a
2 0
Then we substitute the expansion of JO(t) ((15.3)) into the right hand
side.
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
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
is a solution of
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).
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).
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)
(29.1)'
and any linear combination
108
29. Hyperfunction Equation of the Form f"(A)-w 2f(A) g(A) 109
aw-bw k . (29.4)
l
Hence
a = 2-1 (k O + w-1 k l ), b = 2- 1 -1
(k - w k ).
O l (29.S)
is a solution of
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
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
(29.10)
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.
2
fll(A) - w f(A) = g(A)
(29.11)
{ f(O) = k ' fl (0) = k
O l
(29.12)'
f' (0)
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
112 IX. ONE DIMENSIONAL WAVE EQUATION
(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
AN EXAMPLE.
We substitute
(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
30. The Vibration of a String 113
1. y"CA) S2 Y (A) SA 3 .
2. y"(A) s2 yCA ) SA 4 .
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)
(30.3)
(30.5)
i. e. ,
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.
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
* 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.
30. The Vibration of a String 115
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.
a
1
= 2' AO 10 and VI (t)
= f'
4,
0 < t < 4
=
4 < t < 00
(30.8)
-s -9s
e - e
zl (2) = =----l::..,O:'"""s- VI' (30.9)
I - e
?-
11
;C"
5'-'1,,(1)1, J,----t-:-,--.'-I,-,(-')I....
J?s--
O r
116 IX. ONE DIMENSIONAL WAVE EQUATION
+ e
-20s + e
-30s
+
we have
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
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
{
(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)
D ., J
OJ .:.: -t) 2i)-2I u 10
f
0
i)\---,,--i.
!--------"--'
21",-8 20-21010
A
(31.3)
satisfies
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
(31. 5)
zlt{>.) = a. 2s 2z{>.) 2
- a. s[4>{>')] (31.6)
PROOF: Since
we have
(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).
120 IX. ONE DIMENSIONAL WAVE EQUATION
~(O) = 0, 0,
must be given by
Then we have
(31.9)
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)
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)
A+t/ex}
Z(A,t) = ~ {Ij>(A - ~) + Ij>(A + !') + ex ~(u)du
ex JA-t/ex
gives the solution of
= 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.
of the equation
(32.2)
satisfying
We shall examine zl(A) for 0 < t < aA ' Consider the decomposi-
O
tion of zl (A) :
-a(2A O+A)S
both e {vl(t)} and
vanish when
(32.4)
0, o< t < aA
(32.5)
={ vl(t-aA), o~ aA < t.
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
(33.1)
Hence
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):
These two equations (33.4) and (33.4)' are of the same form, and we call
them telegraph equations.
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
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
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
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.
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)
and that the cable is considerably long. Thus we may prove as in the case
of zl (A) of §32, that
(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)
-B(2)''0+>'') -a(2)''0+>'')s
e ·e {E(t)},
are zero if
(35.6)
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,
Hence we have
o o<t<ILCA
I(A,t) = ' =
{ IClL e-fRGAE(t - ILCA), °~ ICLA < t (35.8)
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)
aAIS
U(A) = ae- + be aAIS (36.6)
r U(O) = U(O,t) = a + b = E,
It U(A ) -aAOIS aAOIS
(36.7)
O = U(AO,t) ae + be O.
E
a
I - e- 2aA olS
(36.8)
-e -2 aA o1SE
b
I -e -2aAOIS
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.
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.
(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):
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
Therefore, we have
U(A) (36.14)
(36.15)
I(>..)=-.!.U'(A)
R
(36.16)
(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
132 x. TELEGRAPH EQUATION
we have the following cases, excepting those of §34, §35 and §36.
a = i (~ g),
+ B= i (f - g)
since
(37.1)
(37.2)
37. Concrete Representations 133
a So a
T = I....s. (p,q E C/C) , (37.3)
p Tap
a a I I
T 5 = T ill = {eat} = 5 - [a] 5 - a. (37.6)
n
Tas = (s_a)n (n = ' ' ' , -3, -2, -1,0,1,2,3,"')' (37.7)
If n = 0, we have
o
5 •
PROOF: We have
Thus
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 .
(37.12)
f'(A) = a
-1 {
axa fl(A,t)}. Hence, by (37.4),
= rCt(a-l{it fl(A,t)})
= retf' (A) .
37. Concrete Representations 135
As a Corollary, we have
PROOF: We have
~w'Taexp(Aw) ,
Taexp(O'w) = Tal = I.
Hence, by Proposition 24, we must have (37.13).
(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)
Furthermore,
ff = sf - s [1]
by
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
.!.
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
Next we have
37. Concrete Representations 137
s -
and
and
(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~
138 x. TELEGRAPH EQUATION
and
a>. ->.s
e e
lJ'
o< t < >.,
(37.28)
o<
>. < t
O. (38.2)
Thus, putting
and so
38. A Cable Without Self-Induction 139
a ;; E b
I-e
- 2aA
olS+Y'
U(A)
In
e -aA!s+y E ,.
-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
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)
2
U"(A) = (LCs + RCs)U(A), U(O) = E, (39.1)
_
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
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
~ -2ACX -2Af
exp(-ArLC s)(qple e )
On the other hand, for any A > 0, and A > 0 with AD > A > 0,
O
we have, denoting
{get)},
that
142 x. TELEGRAPH EQUATION
exp(\fU: s)pql =0
by letting \0 ~~. Thus
b = O.
V(\)
EO
= -- exp(-\
I LCs
2
+ RCs), (39.9)
s
and, by (33.3),
Hence, by putting
Ct' = .!. ~
2 L '
we see that
EO / 2
U(A) =--
s
exp(-AILC s +2a's)
Therefore, by (37.26),
(39.11) *
~ -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.
(40.1)
(40.2)
By (33.3), we have
so that
(40.3)
by making use of
(40.4) *
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.
(41.1)
Z(A,O) = 0 (41. 2)
and that the temperatures at the ends of the bar are given by
zeAl ae -aA IS +
be +aA IS (a and b € CIC) ,
145
146 XI. HEAT EQUATION
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
~ (-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)
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).
be given by
Z(A) = ae -aAlS +
b eaAvS (a,b € CIC) (42.2)
together with
-etA IS
Z(A) = vIe Zl(A). (42.5)
[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
and so
43. A Bar Without an Outgoing Flow of Heat 149
Hence
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
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
150 XI. HEAT EQUATION
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.
zeA,O)
so that
(44.1)
Z(O) = 0, (44.2)
¢CA) = (44.3)
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
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)
2
w"(A) - a SW(A) = 0, w(O) = 0, W(A )
O
o.
To prove that this weAl is 0, we put
-2aA /5
O
I = e
This contradicts
[3
e ->./S ~;Z 1(6/rre ) h. 3] (44.8)
-20>' nlS I
If(t)j = e
I a {f(t)}
,1,(')
~ A = I;'L ~n
•
sin nrr>.
-->.-- (44.9)
n=l a
where
(44.10)
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
and
45. A Heat-Conducting Ring 153
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.
(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)
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
154 XI. HEAT EQUATION
(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)
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
and satisfies
(n = ±1,±2, ... ).
-(')
Z A = ae -CLAIS + beCLAIS (a,b E C/C).
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.
45. A Heat-Conducting Ring 155
because we have
and hence
This implies
1 =
contrary to (45.10).*
(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)
z(O) = v, (46.2)
(46.3)
that is,
§2.
(a) {t}
2
(6) {t /2}
(y) {-4 sin t - t}
(0) {2 sin t - 2 cos t + 2}
§9.
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 .
Y(t) - a cos 2~a S1n at.
at -at
e - e
2. Yet)
§13.
at
= _e_ Erf(ICit)
fci'
§16.
= {(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
-8/s m mB
e m+ 1
s
b + ci b - ci
s + i/2 + s - i/2 '
so that
.@..
1.
2.
§36.
at
Formulas and Tables
I. Special Functions
(§ 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!
160
Formulas and Tables 161
C = C[O ,00) (§ 1)
CH = {i; h
f E C and k = 1,2, ... } (§ 3)
( e -AS f(A) dA = { }
f(t) (§26)
a
ax e
-AS
= -se
-As
(§23)
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
(l-ah)Y = I (~)(_a)khk
k=O
(§19)
1 {I/iite -at}
,;s.;:a=
(§ 13)
_1_ =
s .;s.;:a
{...!.. Erf
Iii"
Iat} (a > 0) (§ 13)
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
Formulas and Tables 163
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, ... )
- I e -A/s = {1
- cos 2Yl\t r=}
IS m
-I e AS
IS
e
-AS (§ 22)
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 -
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
Formulas and Tables 165
(A > 0)
(§ 36)
I
IS e -AIS ~ [= I1]h
(v2/~e) (A > 0)
References
166
Propositions and Theorems in Sections
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
Euler's integrals, 32
Laguerre polynomials, 65-66
Exponential hyperfunctions, 74-
76, 83-86, 94-105 Laguerre differential equation, 64-
66
169
170 INDEX