ANNALES SOCIETAT1S MATHEMATICAE P O I O N A E Series I COM MENT ATIO N ES M A T H E M ' T I C A E XXVII (1988) ROCZNIKI POLSKIEGO TOWARZ YSTW A MA TEM AT YC ZN EG O
Séria I PRACE MATEM 'T Y C Z N E XXVII (1988)
Z. Do m a n s k i and Z. Ro j e k (Warsaw)
Application of Fisher-Riesz-Kupradze method to solving the second Fourier problem
Abstract. In this work we construct the solution o f the second Fourier problem using Fisher-Riesz-Kupradze method.
1. Introduction. Let Q be a bounded domain in Euclidean space Е ъ the boundary o f which (*) is a closed surface S of class C 2,A. W e shall denote the points of Q as x = (x ,, x 2, x 3), у = (у,, y 2, y3), . the point o f the surface S as £ = (<^, £2, £3), d = (Vu Ч2, Чз)> ••• The symbol |x-y| denotes the Eu
clidean distance between points x and y: QT = Q x(0, 7); ST = S x(0, T);
= t^i (//), ot2(r}), x3(»/)] is the inner unit normal to surface S at //.
Let у be a function defined in the domain Q and let this function have partial derivatives dx/dxj, j = 1, 2, 3, uniformly continuous in Q. (x) will be the function defined on the surface S by
( X ) ( r j ) = lim x(x).
xeQ
De f i n i t i o n 1. Let a function /< defined on surface S be given and let (x) = ц. We define the derivatives Dxn , j = 1, 2, 3, of the function ц (see [3],
p. 15-20) by the formula
The function /r for which Dx p , j = 1, 2, 3, exist will be said to be differentiable in Hugoniot-Hadamard sense, simply in H -H sense.
О For the definition o f the surface o f class C 2,A see [3 ], p. 96-98, where L 2(B, X) = C 2 X.
2 — Prace M atem atyczne 27.2
216 Z. D o m a n s k i and Z. R o j e k
W e know that (see [3 ], p. 19)
(2 ) I o C j D ^ f i = 0 .
j = i
Instead o f using Dx. we shall use Dj. If the function g defined on ST has in H -H sense derivatives of order l ^ 0 with respect to space variables satisfying Holder condition with exponent h with respect to these variables, 0 ^ h < 1, and if the function g satisfies the Holder condition with exponent x, 0 ^ x ^ 1, with respect to time variable, we shall say that g e C l,h,x(ST).
II. Properties of the Poisson-Weierstrass integral and heat potentials. We know (see [5], p. 529) that for the equation
(3) A u - ~ = g
we have the fundamental formula
t
(4) ju(y, 0) v ( x - y , t)dy— J$g(y, x ) v { x - y , t - x ) d y d x -
(2 0 Q
1 r du
— f — v(x — rj,t — x) dS„dx + b'sdn„
f dv
+ f f “ (»7, t) — (x - rj, t - x) dSndx
os d n ti
Г0 for хфй, t > 0,
■=<U u (£, 0 for (^, î)eSt , lu(x, t) for (x, t )e Q T, where
(5) ',<X’ ,) = ( 4 J p eXP( “ ^ ) f O r , > 0 '
The integrals in (4) play the main role when initial value problems are concerned with (3). Some properties o f these integrals will be given.
De f i n i t i o n 2. The Poisson-Weierstrass integral with density q> is (see [11], p. 21-23; [9], p. 282-283) the integral o f the form
(6) I ( x , t) = $ (p (y )v {x -y , t)dy.
Q
Th e o r e m 1. I f the function (p belongs to Cq(Q), then the Poissonr-Weier- strass integral /(£, t) for set ST belongs to the class C 2,0;1(ST).
W e can prove this theorem using the classical method.
Application o f Fisher-Riesz-Kupradze method 217
De f i n i t i o n 3. The heat volumen potential with density g is (see [11], p.
23-34) the integral of the form
t
(7) J{x, t) = t) v { x - y , t-x )d y d x .
о h
Th e o r e m 2. I f g(-, t) eCq(Q) for every t e(0, T) and g(x, )e C ((0 , T)) for every xeQ, then J( -, •) e C 2,0:1 (5r ).
The proof of this theorem is analogous to the proof o f Theorem 1.
De f i n i t i o n 4. The heat potential of the simple layer with density ф is (see [10], p. 81-112) the integral of the form
t
(8) U (x, t) = j j i /ф1,t)v (x — rj, t — x)dSndx.
o s
Th e o r e m 3. I f the function ф is defined on ST and if ф (•, t) is differentiable in H -H sense at any value x e(0, T) and if the derivatives D j ф , j = 1,2,3, are bounded and integrable on ST, then the derivatives D jU , j = 1 ,2 ,3 , exist and satisfy the Holder condition with respect to space variables with an exponent arbitrarily smaller than 1.
P r o o f. If x<£S, then
dU r dv(x — rj,t — x)
(x, t) = J jФ{г}, т )--------dSndx
Яу . ^ J Яу
V X j O S U X j
Using formula (1), we obtain ÔU{x, t) = -
os
OS ° 4 j
dv (9) 75—(*, 0 = - т)Я;1*ЙчЛт-ЯаД»/Ж»/, T) — dS^dx.
V X j O S O S a n ti
Taking into account the equality £ ak (rj) = 1 and that £ ak(r])Djak(rj) = 0, we have
k= 1
i
3k= 1
( 10 )
Dj0 И = £ a*
DJ( af c
Ф»)•
k= 1
3
Using the equality £ ak Dk ^ v ) = 0 (see (2)), we get k= 1
3
£ a7l ° k (<** Ф и) - ф и ° к a * ] = о.
k= 1
(И)
218 Z. D o m a r i s k i and Z. R o je k
ij/DjV = Djiij/i^ — vDiiJ/: therefore from (10) and (11) we get
t t 3
(12) — ( |if/Dj vdSn dx = f {(Dji/s — c/iji/j X Dkcck) vdSndx +
as a s к = l
t 3 3
+ J I l aJ X ° k ( a k Ф » ) - x a k d j (a * Ф М d s n d x •
bs fc= 1 k=1
From the Stokes theorem and from the fact that S is a closed surface we see that the last integral in (12) vanishes. Therefore we can write (9) in the form
dU ' 3 1 dv
(13) — (x, t) = U ( D j t - o L j t X DkoLk)vdS4d x - \\^(г})ф(г1, x) — dS4dx.
dX; os k= i os a n n
The first of these integrals is the heat potential o f the simple layer with
3
density Djij/ — cCjift X and the second o f these integrals is the heat
к — 1
potential o f the double layer with density а/ф. Using the known boundary property of the heat potential of the double layer (see [11], p. 13-21), we get
ÔU
Therefore
(14)
dU
d x j
lim ^— (x, t) = (^ — )(£, t) = f f ( D , > - a 7> X DkoLk)vdSvdx -
OS k = 1
- jS aljф Y-dSчdx + faj(Ç)ф(Çt t) = l s- W} + ф.
0 S dnn
D j U = l j - X j У « „ h - l W j - a j £ Щ Wt).
k = 1 k = 1
The heat potential of the simple layer f and the heat potential of the double layer Wj satisfy the Holder condition (see [11], p. 6-10, 18-21) with respect to the space variable with an arbitrary exponent smaller than 1 with the previously given assumptions about function ф, therefore (14) we see that Dj U satisfies the Holder condition with the same exponent.
Th e o r e m 4. I f the density ф is a bounded integrable function in ST and satisfies the Holder condition with respect to the time variable t with exponent x, 0 < x ^ 1, and the condition
lim ф(п, t) = ф(р, 0) = 0, t - o +
then the heat potential of the simple layer U(Ç, t) satisfies the Holder condition with respect to the variable t with exponent x.
The proof o f this theorem is easy.
Application of Fisher-Riesz-Kupradze method 219
III. Problem. Using Fisher-Riesz-Kupradze method, we shall solve the following Fourier problem:
We seek the function u(x, t) satisfying in QT the equation
(15) du
Ли— — = g
dt У
and the following conditions:
(16)
(17)
lim u{x, t) = tp(x), l-*0 '
lim — du(x, t) = ф(£, t).
x -*£, dïlç x e Q
We assume that S is of class С 2Д,
g{-, t)e C l{ Q ) for any f e ( 0, T), g{x, -)e C ((0 , T)) for any x gQ, (peCl(Q),
il/eCl ’0:X(ST) and 0) = 0.
S o lu t io n o f th e p ro b le m . Using the basic formulas (4) we can write the solution (if it exists) in the form
t
(18) u(x, t) = j’ (p (y )v (x - y , t)dy— Ц д {у , z ) v ( x - y , t - z ) d y d z -
b on
1 1 dv
- J $Ф(г}, z ) v ( x - r j , t - T ) d S 4dT+ JJ/to, t — z)dSndz,
0s 0 s
where the known function / (g, z) is the solution of the functional equation
(19) i i f ( g , z ) ^ - ( x - r i , t - z ) d S t,dz = F ( x , t ) ( x f Ü , t e(0, T)).
os <4
In (19) we denoted
t
(20) F (x , t) = - (< p (y )v (x - y , t)d y + f f g (y , z ) v ( x - y , t-z )d y d z +
b о n
t
+ I \ф(т], z )v (x —rj, t — z)dSndz.
o s
Th e o r e m 5. The functional equation (19) has exactly one solution be
longing to class C l,h:/(ST).
P r o o f. Let us assume that (19) has got a solution. Using the properties of the*heat potential o f the double layer (see [11], p. 15), we get the integral
220 Z. D o m a n s k i and Z. R o j e k
equation
(2 1 ) 1 dv
- V ( L t ) + JJ/fo, t t - x ) d S ndx = F d , t).
Л Я Я
OS u n r\
The integral equation (21) is a Volterra equation with a continuous right side; it has exactly one continuous solution (see [12], p. 129-135). Taking into account the assumptions, we infer on the basis of Theorems 1-4 that the function F(Ç, t ) e C 1,h’x(ST) and F ( Ç , 0) = 0, W e shall show that the solution of the integral equation (19) is o f class C 1,h;X(ST).
W e shall use the following lemmas.
Le m m a 1. I f a function F is continuous and bounded on ST, it satisfies the Holder condition with exponent A with respect to variable t and F(Ç, 0) = 0 for ÇeS, then the solution of the integral equation (21) satisfies the Holder condition with the same exponent A with respect to variable t.
We get the proof o f Lemma 1 by showing that any o f the terms o f the Neumann series (which is the solution o f (21)) satisfies the Holder condition with respect to variable t with exponent A, and that the sum o f the Neumann series satisfies the Holder condition with respect to variable t with expo
nent A.
Lemma 2. I f the function F ( - , ' t ) e C 1,h{S) for any te(0, T), then the solution f { - , t) of the integral equation (21) belongs to class C l,h(S) for any t e {0, T).
Lemma 2 is proved by application o f a localization principle based on the appropriate partition of unity (see [1], p. 63-88).
The function F(£, t) satisfies assumptions o f Lemmas 1 and 2, therefore on the basis o f these lemmas we conclude that the solution o f the integral equation (21) is o f class C 1,h:X(ST). W e shall show that the solution / o f the integral equation (21) is a solution o f the functional equation (19).
Let us suppose that the function / is not a solution o f the functional equation (19). Let us denote
where the function / (rj, x) is a solution o f the integral equation (21). W e have Ф(х, t) f it ) for хфй. It is easy to check that
0 S u n f)
(22)
(23) И т Ф ( х , t) = 0 ,
Application of Fisher-Riesz-Kupradze method 221
(24) НшФ(х, 0 = 0,
x$Q
(25) lim Ф(х, t) = 0.
Ijcf -►+ GO
Looking for the solution o f problems (22), (23), (24), (25) in the form o f the heat potential of the double layer with continuous and bounded density p
* dv
(26) Ф{х, t) = j jp(rj, t) — ( x - r j , t~z)dSqdz
o s d n n
and using condition (24) and the property of the heat potential of the double layer, we get an integral equation o f the form
1 dv
(27) ~ ? p ( £ , f) + J j> (? 7 , t) — - t - z ) d S vdz = 0.
0 S a n n
The equation (27) has exactly one solution p(rj, т) = 0. Therefore, Ф{х, t) = 0. The contradiction we arrived at ends the proof of Theorem 5.
Because the solution o f (19) belongs to class C 1,h,x(ST) the Lapunov- Taubers theorem is valid for the function и given by formula (18) (see [2], corollary, p. 142). Taking the above into account, it is easy to check that the function и is a solution o f problems (15), (16), (17).
C o n s t r u c t io n o f th e s o lu tio n . Let S1 be an arbitrary closed Lapunov’s surface which is the boundary of a domain Q1 which includes Q in its interior.
Let us write
Q\ = Q 1 x (0, T ); SlT = S1 x (0, T).
Let us take a countable dense set o f points {(£*, rf) } , k , i = 1, 2, ..., on the surface Sj and let us consider the set of the functions
ГКО rtpk , 4 J for T < t h
(28) r { C - r j , f -т) = < dn„
[ 0 for f ^ t< T, k, i = 1, 2, ...
Ordering the set (28) in a certain way, we write (29) Г (i ki - 4, t, . - t) = r t((/, t), 7 = 1 ,2 ,...,
and consider the space L 2(ST) with the norm
||ф|| T)\2dS„dx)1/2.
os
Le m m a 3. The set of functions {T j( -, •)}, j = 1, 2, ..., is linearly indepen
dent.
222 Z. D o m a r ï s k i and Z. R o je k
Lemma 4. The set of functions {T j { -, •){, j = 1 , 2 , . . . , is complete in the domain L 2{ST).
The proofs o f Lemmas 3 and 4 are similar to those in [7], [8].
Subjecting the set {//•(*, •)] to the process of orthonormalization, we get the set \cûj(-, •)] o f orthonormal functions. The elements of the set [to/-, •){, 7 = 1 ,2 ,. . ., are linear combinations o f the elements of the set {ГД-, •)) and
vice versa (see [4], p. 72-73), i.e.,
CO 0 = £ A k j T k ( t i , t); Гj(rj, r ) = £ Bkjcok(rj, i ) .
k = 1 k = 1
We denote Fourier coefficients of the function / ( •, •) with respect to the set
\a)j(\ •);, j = 1, 2, ..., by
T
Ф] = ï I f(r i, ?)Mj(th ^dS^dx, j = 1 ,2 ,...
0s
Putting x = c J, Л ; t = t(j into equation (19), multiplying this equation by Ar and taking the sum from r equal one to r equal j, we get
T
Jо s
JJ/to, t) £ Arjr r(tt, r)dS„dx = £ Arj F(£kr, g r = 1
and therefore
Ф,
= I
A , j F ( p , t,J.r = 1
Because F (x , r) is a given function the coefficients Arj are the normalization constants, therefore all Fourier coefficients Ф, are determined. Since f e L 2{ST), we have
(30)
Let us write
(31)
lim Ц/- X Фко)к|| = 0.
fc= l
/ * ( > 7 , 0 = X Ф к ю к (>1, т), l
(32) и * / * , /) = f t) —dv(.y- ; ; , t — x)dSndx — F( x, t) .
о s dn„
Th e o r e m 6. For any x e Q and for any t from the interval (0 , T) and for any £ > 0 there exists a N 0 such that for N > N 0 we have
(33) \u(x, t) — uN(x, t)\ < 8.
P r o o f. W e have
dv (34) \u(x, t ) - u N(x, 01 ^ JJ|f i n , t) - f N(r],t)|
os dn.■ (x tj, t -t) dSqdx.
Application of Fisher-Riesz-Kupradze method 223
Using the estimate dv
dn,( x - r j , t -t) ^ M [ ( t - z ) vab- 2vT \
where 0 < v < 1, a = inf|x — rj\ and M is a positive constant, we get
f/ e S
(35)
И
os dvdn( x - r ) , X -t) dS„dx ^ (mes S ) M T l ~vr6 — 2v
(1 — v) = M\.
On the basis of (30) we can choose an N such that
(36) IJ If ( 4 , T ) - / w(4, t)|2dSndx W f - r w 1 S)
os M i
Using the Schwarz inequality in (34) and taking into account (35) and (36), we get (33).
On the basis of Theorem 6 we get the solution of problems (15), (16), (17)
1 N dv
(37) u{x, t ) = lim J j ( X 4>kcok(ri,T)) — ( x - r i , t - T ) d S 4d T - F ( x , t ) . N ~*OD Cl S k= 1 Xln„
References
[1 ] M. S. A g r a n о v i с. M. I. Vi s i k, Èlipticeskie zadaci s paramétrant i paraboliceskie zadaci obscego vida, Usp. Mat. Nauk 19 (3) (117) (1964) (in Russian) ^
[2 ] Z. D o m a n s k i , A. P i s k o r e k i Z. R o j e k , О primenenii metod,■ 'Fisera-Rissa-Kupradze dlja resenija perroj zadaci Fare, Ann. Soc. Math. Polon. Ser. 1, Comment. Math. 16 (1972),
137-147 (in Russian).
[3 ] N. M. G i u n t e r , Teoria potencjalu, P W N , Warszawa 1957.
[4 ] W . K o l o d z i e j , Wybrane rozdzialy analizy matematycznej, P W N , Warszawa 1970.
[5 ] M. K r z y z a n s k i , Râwnania rôzniczkowe czqstkowe rzçdu drugiego, T. 1, P W N , Warszawa 1957.
[6 ] V. D. K u p r a d z e , Metody potenciala v teorii uprugosti, Moskwa 1963 (in Russian).
[7 ] —, Ob odnom met ode priblizennogo resenija predeX nych zadac matematiceskoj fiziki, 2.V.M . i M.F. 6 (1964) (in Russian).
[8 ] —, О priblizennom resenii zadac matematiceskoj fiziki, Usp. Mat. Nauk 22 (2) (134) (1967), 59-107 (in Russian).
[9 ] A. P i s k o r e k . Râwnania calkowe. Element y teorii i zastosowania, W N T , Warszawa 1980.
[10] —, 0 pewnych wlasnosciach pochodnych potencjalow cieplnych w teorii râwnania przewod- nictwa, Biuletyn W A T nr 29, Warszawa 1957.
[11] W. P o g o r z e l s k i , Wlasnosci potencjalow cieplnych w teorii râwnania przewodnictwa, ibidem nr 29, Warszawa 1957.
[12] —, Râwnania calkowe i ich zastosowania, T. 2, P W N , Warszawa 1958.