Miskolc Mathematical Notes HU e-ISSN 1787-2413 Vol. 20 (2019), No. 1, pp. 365–379 DOI: 10.18514/MMN.2019.2089
SOME PROPERTIES OF GREEN’S FUNCTIONS OF SHILOV-TYPE PARABOLIC SYSTEMS
V. LITOVCHENKO AND G. UNGURYAN Received 15 September, 2016
Abstract. For Shilov-type parabolic systems with nonnegative genus and coefficients of bounded smoothness, the properties of differentiability of Green’s function with respect to spatial variables are studied.
2010Mathematics Subject Classification: 35K40; 35A08
Keywords: Cauchy problem, Shilov-type parabolic systems, Green’s function
1. INTRODUCTION
Developing the idea advanced by Ya.I. Zhitomirskii in [10] concerning the descrip- tion of the systems of partial differential equations parabolically steady to a change in coefficients, the following wide class of parabolic systems with variable coefficients was defined in [7]:
@tu.tIx/D fP0.tIi @x/CP1.t; xIi @x/gu.tIx/; .tIx/2˘.0IT : (1.1) Here,uWDcol.u1; :::; um/; ˘.0IT WD f.tIx/W t2.0IT ; x2Rng, andP0.tIi @x/and P1.t; xIi @x/are matrix differentiable expressions of the orders, respectively, pand p1,p > p10;with coefficients depending on the time variablet. In this case, the coefficients of the expressionP1can depend also on the spatial variablex. It is also assumed that the corresponding system
@tu.tIx/DP0.tIi @x/u.tIx/; .tIx/2˘.0IT ; (1.2) is uniformly parabolic by Shilov with parabolicity indexh,0 < hp, nonnegative genus;and reduced orderp0[2]. In other words, there exist constantsı0> 0and ı10such that, for allx2Rnandt2Œ0IT ;the following inequality holds:
.tIx/WD max
j2Nm
Rej.tIx/ ı0kxkhCı1:
Here,j are roots of the equation det.P0.tI/ E/D0; 2Cn; E is the identity matrix of the orderm,k kis the Euclid norm inRn, andNmWD f1; 2; : : : ; mg.
We recall also that the reduced order p0 of a system is called the exact power order of growth of a function.I /in the complex spaceCn(for a parabolic system,
c 2019 Miskolc University Press
we have always p0 > 1). In addition, the genus of a system parabolic by Shilov is called the largest index such that the function.I /in the domainG.K/WD Œ0IT fxCiy2CnW kyk K.1C kxk/gwith someK > 0satisfies the estimate
.tIxCiy/ ıkxkhCı1; ı> 0; ı10:
It is known that1 .p0 h/1:
The following condition must be satisfied for (1.1):
(A)W0p1< h n.1 h=p0/ .m 1/.p h/:
It is obvious that the class of systems (1.1) envelops completely the class of Shilov- type parabolic systems with nonnegative genus. In this connection, it was proposed in [10] to call such systems Shilov-type parabolic systems with nonnegative genus and variable coefficients. To a certain extent, these systems solve the problem of proper extension of the notion of parabolicity by Shilov to the linear systems of equations with variable coefficients, since the Shilov class of systems is not parabolically stable to a change in its coefficients [4].
Let us consider systems (1.1) under the condition that their coefficients are the complex-valued functions that are continuous in the variable t, infinitely differen- tiable with respect to the variable x; and bounded together with their derivatives in a ball ˘Œ0IT : For these systems, Green’s functionZ.t; xI; /, 0 < t T, fx; g Rn, was constructed, and its main properties such as the smoothness and the behavior in vicinities of infinitely remote spatial points were studied [7]. On the basis of these results, the theory of the Cauchy problem was developed in [5,8,9]
for such systems in spaces of the typeS by Gelfand and Shilov [3]. In particular, the proper solvability of the Cauchy problem with generalized initial data like the Gevrey ultradistributions was established, the form of a representation of classical solutions with generalized limiting values on the input hyperplane was obtained, their qualitat- ive properties were studied, and the sets of generalized input functions, for which the corresponding solutions are elements of the Schwartz spaceS or some space of the typeS;were described.
In order to get similar results under weaker conditions imposed on the coeffi- cients of system (1.1), it is necessary firstly to clarify the properties of corresponding Green’s function Z. In the present work, we will study properties of the function Z.t; xI; /with respect to spatial variablesxand under the minimum conditions of smoothness imposed on the coefficients of this system.
2. AUXILIARY INFORMATION
Let the differential expressionsP0andP1of system (1.1) have the structure P0.tIi @x/D X
jkjCp
A0;k.t /@kx; P1.t; xIi @x/D X
jkjCp1
A1;k.tIx/@kx;
whereA0;k.t /WDijkjC
alj0;k.t /m
l;jD1,A1;k.tIx/WDijkjC
alj1;k.tIx/m
l;jD1are mat- rix coefficients,iis the imaginary unity, andjkjCWDk1C: : :Ckn; k2ZnC.
By G.t; I /, 0 < t T, we denote Green’s function of system (1.2). It is known that G.t; I /DF
t./
.t; I /, where F Œis the Fourier transformation operator, andt./is a matriciant of the corresponding Fourier duality of the system.
The following proposition is proper [5,6]:
8T > 09ı > 08k2ZnC9c > 08t2.IT 8 2Œ0IT /8 fx; g RnW j@kxG.t; Ix /j c.t /
nCjkjCC
h e ı
kx k .t /˛
11˛
; (2.1)
where WD.m 1/.p h/and˛WD=p0,j.alj/ml;jD1j WD max
fl;jgNm
jaljj.
Here, we consider systems (1.1), which satisfy, in addition to condition (A), the following condition:
(B): the coefficientsalj0;k.t /,alj1;k.tIx/are continuous in the variablet uni- formly with respect tox, differentiable with respect to the variablex up to the order˛inclusively, and bounded together with their deri- vatives by complex-valued functions in a ball˘Œ0IT .
In [7], Green’s function of system (1.1) was constructed in the form
Z.t; xI; /DG.t; Ix /CW .t; xI; /; .t; xI; /2˘T2; (2.2) where˘T2 WD f.t; xI; /j0 < tT;fx; g Rng,
W .t; xI; /WD
t
Z
dˇ Z
Rn
G.t; ˇIx y/˚.ˇ; yI; /dy; (2.3) and
˚.t; xII/D
1
X
lD1
Kl.t; xI; /; (2.4)
where
K1.t; xI; /WDP1.t; xIi @x/G.; Ix /;
Kl.t; xI; /WD
t
Z
dˇ Z
Rn
K1.t; xIˇ; y/Kl 1.ˇ; yI; /dy; l > 1: (2.5) In this case, it was established that condition (A) and the boundedness of the coef- ficients of system (1.1) ensure the absolute uniform convergence of the functional series (2.4) for allfx; g Rn,t2.IT ;and 2Œ0; T /. Moreover, its sum˚ and the iterated kernelsKl satisfy the estimates
j˚.t; xI; /j c1.t /˛0 .1C˛ n/e ı1
kx k .t /˛
11˛
; (2.6)
jKl.t; xI; /j c0l 0
@
l 1
Y
jD1
c.j"/B.˛0; j˛0/ 1 A
.t /l˛0 .1C˛ n/e ı.1 .l 1/"/
kx k .t /˛
1
1 ˛
; "2.0I1/; (2.7) with the estimating constants independent oft; ; x;and. Here˛0WD1C˛ n .nC p1C /= h > 0andB.;/is the Euler beta-function.
We note that estimates (2.1) and (2.6) forfx; g Rnand0 < tT guarantee the absolute convergence of the integral, by which the potentialW is determined.
Thus, the matrix functionZ.t; xI; /is properly determined by formula (2.2) on the whole set˘T2.
Completing this item, we present the following estimates from [1], which will be of importance in what follows:
e
ı (
kx yk .t ˇ/˛
1
1 ˛
C ky k
.ˇ /˛
1
1 ˛
)
e ı
kx k .t /˛
11˛
I (2.8)
Z
Rn
e
ı (
kx yk .t ˇ/˛
11˛ C
ky k .ˇ /˛
11˛ )
dy
..t ˇ/.ˇ //˛ n
c"e ı.1 "/
kx k .t /˛
1
1 ˛
.t /˛ n ; ı > 0 (2.9)
(here,fx; y; g Rn,ˇ2.It /,0 < tT,"2.0I1/;andı > 0, and the quantity c"depends only on").
3. PROPERTIES OF GREEN’S FUNCTION
First, we estimate the derivatives of the iterated kernelsKl.
According to representation (2.5), the smoothness of the kernelK1.t; xI; /in the spatial variablesx and is determined, respectively, by the smoothness of the coefficients of system (1.1) and the functionG.t; Ix /. Therefore, there exist the derivatives@r@qxK1forfr; qg ZnC,jqjC˛, and the following equality holds:
@r@qxK1.t; xI; /D X
jkjCp1
q
X
lD0
Cql
@lxA1;k.tIx/ @k.x /CrCq lG.t; Ix /
; whereCql is a binomial coefficient. From whence, with regard for condition (B) and estimate (2.1) forfr; qg ZnC,jqjC˛,.t; xI; /2˘T2;we get
j@r@qxK1.t; xI; /j cr;q.t /
nCp1CCjrCqjC
h e ı
kx k .t /˛
11˛
(3.1)
(here, the estimating constants are independent oft; ; x;and).
Forl > 1;we will use the representation Kl.t; xI; /D
t1
Z
dˇ Z
Rn
K1.t; xIˇ; C/Kl 1.ˇ; CI; /d
C
t
Z
t1
dˇ Z
Rn
K1.t; xIˇ; x ´/Kl 1.ˇ; x ´I; /d´; t1WDtC 2 : According to it,
@r@qxKl.t; xI; /D X
jr1jCjrjC
Crr1
t1
Z
dˇ Z
Rn
@r1@qxK1.t; xIˇ; C/
@r r 1Kl 1.ˇ; CI; /
dC X
jq1jCjqjC
Cqq1
t
Z
t1
dˇ Z
Rn
@qx1K1.t; xIˇ; x ´/
@r@q qx 1Kl 1.ˇ; x ´I; /
d´; jqjC˛; .t; xI; /2˘T2: (3.2) Hence, the estimation ofj@r@qxKl.t; xI; /jis reduced to that of the expressions
j@r@qxK1.t; xI; C/j; j@qxK1.t; xI; x ´/j; j@rKl 1.t; CI; /j; j@r@qxKl 1.t; x ´I; /j:
In view of the boundedness of@qxalj1;k.tIx/,jqjC˛, and estimate (2.1), for all fq; rg 2ZnC,jqjC˛;fx; ; g 2Rn; t2.IT ;and2Œ0IT /;we have
ˇˇ@r@qxK1.t; xI; C/ˇ
ˇm X
jkjCp1
X
jq1jCjqjC
Cqq1ˇ
ˇ@qx1A1;k.tIx/ˇ ˇ
ˇ
ˇ@k.x /CrCq q1G.t; Ix /ˇ
ˇcr;q.t /
nCp1CCjrCqjC
h e ı k
x k .t /˛
11˛ I (3.3) ˇˇ@qxK1.t; xI; x /ˇ
ˇD ˇ ˇ
ˇ@qx X
jkjCp1
A1;k.tIx/@kxG.t; I/ˇ ˇ ˇm
ˇ ˇ
ˇ@qxA1;0.tIx/
ˇ ˇ ˇ
ˇ ˇ
ˇG.t; I/ ˇ ˇ
ˇbcq.t / nChe ı k
k .t /˛
11˛
cq.t /
nCp1C
h e ı k
k .t /˛
11˛ : (3.4)
We now estimate the expressionˇ
ˇ@rKl.t; CI; /ˇ ˇ:Since
@rK1.t; CI; /D X
jkjCp1
@rA1;k.tIC/@kG.t; I/; .t; xI; /2˘T2; (3.5) we have, according to condition (A), that the iterated kernels Kl.t; CI; /are differentiable with respect to the variable only to the order˛. This fact and (3.2) imply that@qxKl.t; xI; /,jqj˛, is also a function differentiable with respect to only to this order˛.
Representation (3.5) and estimate (2.1) yield ˇˇ@rK1.t; CI; /ˇ
ˇc1;r.t /
nCp1C
h e ı k
k .t /˛
11˛
: (3.6)
We note that
@rK2.t; CI; /D@r
t
Z
dˇ Z
Rn
K1.t; CIˇ; y/K1.ˇ; yI; /dy
! : Let us change the variable of integration in the last integral by the formulayD´C: In view of estimates (3.6) and (2.9) and the equalities
t
Z
..t ˇ/.ˇ //˛0 1dˇD.t /2˛0 1B.˛0; ˛0/ (3.7) and
@rK1.t; CI; ´C/D@rK1.t; . ´/CI; / ˇ ˇ ˇ
D´C; we get
ˇˇ@rK2.t; CI; /ˇ
ˇm X
jr1jCjrjC
Crr1
t
Z
dˇ Z
Rn
ˇ ˇ
ˇ@r1K1.t; CIˇ; ´C/ ˇ ˇ ˇ
ˇ ˇ
ˇ@r r 1K1.ˇ; ´CI; / ˇ ˇ
ˇd´m X
jr1jCjrjC
Crr1c1;r1c1;.r r1/
t
Z
.t ˇ/.ˇ / nCp1hC Z
Rn
e ı
k ´k .t ˇ/˛
11˛
C .ˇk´ /˛k 11˛ d´dˇ
c2;r."/B.˛0; ˛0/.t /˛0 nCp1hCe ı.1 "/ k
k .t /˛
11˛
; "2.0I1/:
By reasoning analogously step by step, we arrive at the inequality ˇˇ@rKl.t; CI; /ˇ
ˇcl;r."/l 1Y
jD1
B.˛0; j˛0/
.t /.l 1/˛0
nCp1C h
e ı.1 .l 1/"/ .tk /˛k 11˛
; (3.8)
which is satisfied for all f; g Rn, jrjC ˛, 0 < t T, "2.0I1/; and l2Nn f1gand, hence, until the existence of such numberl, for which
ˇˇ@rKl.t; CI; /ˇ
ˇcl;r."/lY1
jD1
B.˛0; j˛0/
e ı.1 .l 1/"/ k
k .t /˛
11˛ (3.9) (here, the quantitiescl;r."/ > 0do not depend on the variablest; ; ;and, which vary in the above-indicated way).
Since
@r@qxKl.t; xI; C/D@r@qxKl.t; xI; /ˇ ˇ ˇ
DC;
@r@qxKl.t; x ´I; /D@r@qyKl.t; yI; / ˇ ˇ ˇy
Dx ´;
the expressions @r@qxKl.t; xI; C/; @r@qxKl.t; x ´I; / and @r@qxKl.t; xI; / are of the same type. Therefore, with regard for representation (3.2) and estimates (3.3), (3.4), (3.8), and (2.9), we have
ˇˇ@r@qxK2.t; xI; /ˇ
ˇm2jrCqjC X
jr1jCjrjC
cr1;qc1;.r r1/
t1
Z
.t ˇ/
nCp1CCjr1CqjC h
.ˇ /
nCp1C h
Z
Rn
e ı
kx k .t ˇ/˛
11˛
C .ˇkk /˛11˛ ddˇ
C X
jq1jCjqjC
cq1cr;.q q1/
t
Z
t1
.ˇ /
nCp1CCjrCq q1jC
h .t ˇ/ nCp1hC Z
Rn
e ı k´k
.t ˇ/˛
11˛
C kx.ˇ ´ /˛k11˛ d´dˇ
!
m2jrCqjCc"e ı.1 "/ k
x k .t /˛
11˛
.t / ˛ n X
jr1jCjrjC
cr1;qc1;.r r1/
t1
Z
.t ˇ/˛ n
nCp1CCjr1CqjC
h .ˇ /˛0 1dˇC X
jq1jCjqjC
cq1cr;.q q1/
t
Z
t1
.t ˇ/˛0 1.ˇ /˛ n
nCp1CCjrCq q1jC
h dˇ
;jrjC˛;jqjC˛; "2.0I1/:
In view of the estimates
t1
Z
.t ˇ/˛ n
nCp1CCjr1CqjC
h .ˇ /˛0 1dˇ 2jr1C
qjC
h .t /2˛0 1Cjr1C
qjC h
B.˛0; ˛0/ and
t
Z
t1
.t ˇ/˛0 1.ˇ /˛ n
nCp1CCjrCq q1jC
h dˇ
2j
rCq q1jC
h .t /2˛0 1C
jrCq q1jC h
B.˛0; ˛0/;
we get the inequality ˇˇ@r@qxK2.t; xI; /ˇ
ˇc2;r;q.t /2˛0 1C˛ nC
jrCqjC h
B.˛0; ˛0/e ı.1 "/ .tkx /˛k 11˛
: By continuing stepwise the process of estimation, we obtain
ˇ
ˇ@r@qxKl.t; xI; /ˇ
ˇcl;"r;q.t /l˛0 1C˛ nC
jrCqjC h
e ı.1 .l 1/"/ .tkx /˛k11˛l 1Y
jD1
B.˛0; j˛0/
; (3.10) for alljrjC˛;jqjC˛;fx; g Rn; 0 < tT; "2.0I1/;andl2Nnf1g.
Let us pass to the estimation of the expressionˇ
ˇ@r@qxKl.t; xI; /ˇ
ˇ, which will be suitable for the establishment of the differentiability of the matrix function˚ with respect to the spatial variables. Directly from (3.10), we arrive at the existence of a numberlsuch that
ˇˇ@r@qxKl.t; xI; /ˇ
ˇclr;q;"e ı.1 .l 1/"/ k
x k .t /˛
11˛l
1
Y
jD1
B.˛0; j˛0/ : Let us setlCWDmaxfl; lg; l WDminfl; lg, wherelis the corresponding number from (3.9),"WDr1lC; ıWDı 1 r1
; r> 2; T0WDmaxf1; Tg, and
c0WD max
l2Nl
Cnf1g
n
c1;r; cl;r."/l 1Y
jD1
B.˛0; j˛0/
; cr;q; cl;"r;ql 1Y
jD1
B.˛0; j˛0/o
;
cWDc0.T0/lC l :Then (3.8) and (3.10) imply that, for allfx; ; g Rn,0 <
tT,jrjC˛;andjqjC˛; ˇ
ˇ@r@qxKlC.t; xI; /ˇ
ˇce ı k
x k .t /˛
11˛
; ˇˇ@rKlC.t; CI; /ˇ
ˇce ı .tkk /˛
11˛ : In view of this result, estimate (2.8), the equality
Z
Rn
e ı0 .tkxˇ/˛yk
11˛ dy .t ˇ/˛ n D
Z
Rn
e ı0k´k
1 1 ˛
d´DWbE <C1;
representation (3.2), and inequalities (3.3) and (3.4), we obtain ˇˇ@rKlCC1.t; CI; /ˇ
ˇ X
jr1jCjrjC
Crr1
t
Z
dˇ Z
Rn
ˇˇ@r1K1.t; CIˇ; ´C/@r r 1
KlC.ˇ; ´CI; /ˇ
ˇd´mc2 X
jr1jCjrjC
Crr1
t
Z
.t ˇ/˛0 1
Z
Rn
e ı
k ´k .t ˇ/˛
11˛
C .ˇk´ /˛k 11˛ e
ı r
k ´k .t ˇ/˛
11˛ d´
.t ˇ/n˛dˇ mcr0bEc2B.˛0; 1/.t /˛0e ı k
k .t /˛
11˛
; cr0WD X
jr1jCjrjC
Crr1I
ˇ
ˇ@r@qxKlCC1.t; xI; /ˇ
ˇ X
jr1jCjrjC
Crr1
t1
Z
dˇ Z
Rn
ˇ
ˇ@r1@qxK1.t; xIˇ; C/
@r r 1KlC.ˇ; CI; /ˇ
ˇdC X
jq1jCjqjC
Cqq1
t
Z
t1
dˇ Z
Rn
ˇˇ@qx1K1.t; xIˇ; x ´/
@r@q qx 1KlC.ˇ; x ´I; /ˇ ˇd´
mc2 X
jr1jCjrjC
Crr1
t1
Z
.t ˇ/˛0 1Cjr1C
qjC h
Z
Rn
e ı
kx k .t ˇ/˛
11˛
C .ˇk /˛k 11˛ e
ı r
kx k .t ˇ/˛
11˛ d .t ˇ/˛ ndˇ
C X
jq1jCjqjC
Cqq1
t
Z
t1
t ˇ˛0 1Z
Rn
e ı k´k
.t ˇ/˛
11˛
C k.ˇx ´ /˛k11˛
e rı k
´k .t ˇ/˛
11˛ d´
.t ˇ/˛ ndˇ
!
mc2Eeb ı k
x k .t /˛
11˛
t
Z
.t ˇ/˛0 1dˇ
X
jr1jCjrjC
Crr1.t t1/ jr1C
qjC h
! Ccq
!
mc2Eeb ı .tkx /˛k 11˛
.t /˛0
B.˛0; 1/
2j
rCqjC
h X
jr1jCjrjC
Crr1.t / jr1C
qjC h
! Ccq
!
mcr;q0 c2E.2Tb 0/j
rCqjC
h B.˛0; 1/.t /˛0 j
rCqjC h e ı k
x k .t /˛
11˛
; cr;q0 WDcrCcq: Applying the method of induction, we can verify firstly the validity of the estimate
ˇ
ˇ@rKlCCl.t; CI; /ˇ ˇ
c.mc0rcbE.t /˛0/le ı .tk /˛k
11˛ l 1 Y
jD1
B.˛0; 1Cj˛0/
!
; (3.11) and, hence, the estimate
ˇˇ@r@qxKlCCl.t; xI; /ˇ ˇc
mcr;q0 cE.2Tb 0/j
rCqjC h
l
.t /l˛0 j
rCqjC h
e ı k
x k .t /˛
11˛ l 1 Y
jD1
B.˛0; 1Cj˛0/
!
; (3.12) forjrjC˛,jqjC˛,.t; xI; /2˘T2 andl2Nnf1g.
The following propositions hold true.
Lemma 1. The matrix function˚.t; xI; /on the set˘T2 is a function differenti- able with respect to each of the spatial variablesxandto the order˛inclusively, and their derivatives satisfy the following estimates:
ˇˇ@r@qx˚.t; xI; /ˇ
ˇc1.t /˛0 .1C˛ nCj
rCqjC
h /e ı .tkx /˛k 11˛
; (3.13) ˇˇ@r˚.t; CI; /ˇ
ˇc2.t /˛0 .1C˛ n/e ı .tk /˛k 11˛
;f; g Rn (3.14) (here, the estimating constantsc1; c2;andıare independent oft; ; x; ; ).
Proof. In any way, let us fix a point.x0I0/fromR2n;and let us consider a ball Kı.x
0I0/with radiusı > 0;which is centered at the point.x0I0/;in this space. Then, in view of structure (2.4) of the function˚ and the differentiability of the iterated kernels Kl with respect to spatial variables onR2n to the order ˛ inclusively, we can conclude that, in order to prove the differentiability of the matrix function˚ at the point .x0I0/ to the indicated order, it is necessary only to prove the uniform convergence of the formally differentiated series (2.4) in the variablesxand on the setKı.x
0I0/,ı > 0(at every fixedt and; 0 < t T):
1
X
lD1
@r@qxKl.t; xI; /; jrjC˛;jqjC˛: (3.15) Directly from estimates (3.10) and (3.12) and the equality
l 1
Y
jD0
B.˛0; 1Cj˛0/D .˛0/l
.1Cl˛0/;
where ./is the Euler gamma-function, forfr; qg ZnC,jrjC˛,jqjC˛, and .t; xI; /2˘T2;we have
ˇ ˇ ˇ
1
X
lD1
@r@qxKl.t; xI; / ˇ ˇ ˇ
lC
X
lD1
ˇ ˇ
ˇ@r@qxKl.t; xI; / ˇ ˇ ˇC
1
X
lDlCC1
ˇ ˇ
ˇ@r@qxKl.t; xI; / ˇ ˇ ˇ
c
lC
X
lD1
.t /l˛0 .1C˛ nCj
rCqjC h /
C
1
X
lD1
mcr;q0 cbE.2T0/j
rCqjC h l
.t /l˛0 j
rCqjC h
l 1Y
jD1
B.˛0; 1Cj˛0/
!
e ı .tkx /˛k 11˛
c1.t /˛0 1C˛ nC
jrCqjC h
e ı .tkx /˛k 11˛
:
From whence, we get the uniform convergence of series (3.15) inxandand, hence, the validity of estimates (3.13).
Due to the corresponding estimates (3.8) and (3.11), we can verify analogously the validity of estimate (3.14).
The lemma is proved.
Lemma 2. The volumetric potentialW .t; xI; /on the set˘T2 is a function dif- ferentiable with respect to each of the spatial variablesxand to, respectively, the
orders˛Cp1and˛inclusively. In this case,
@r@qxW .t; xI; /D
r
X
lD0
Crl
t1
Z
dˇ Z
Rn
@l@qxG.t; ˇIx y /@r l ˚.ˇ; yCI; /dy
C
t
Z
t1
dˇ Z
Rn
@qxG.t; ˇIx y/@r˚.ˇ; yI; /dy; jqjCp1;jrjC˛; (3.16)
@r@qxW .t; xI; /D
r
X
lD0
Crl
t1
Z
dˇ Z
Rn
@l@qxG.t; ˇIx y /@r l ˚.ˇ; yCI; /dy
C
t
Z
t1
dˇ Z
Rn
@kG.t; ˇI/@r@q kx ˚.ˇ; x I; /d; jrjC˛; (3.17) jkjCDp1; p1<jqjC˛Cp1:
Proof. ForjqjCp1andjrjC˛;we use the representation W .t; xI; /D
t1
Z
dˇ Z
Rn
G.t; ˇIx y /˚.ˇ; yCI; /dy
C
t
Z
t1
dˇ Z
Rn
G.t; ˇIx y/˚.ˇ; yI; /dy:
From whence, by the formal differentiation under the sign of integral, we obtain equality (3.16). Hence, in order to substantiate the validity of equality (3.16), it is sufficient to prove the uniform convergence of the following integrals in the variables xandonR2n:
Ir;l;q1 .t1; xI; /WD
t1
Z
dˇ Z
Rn
j@l@qxG.t; ˇIx y /jj@r l ˚.ˇ; yCI; /jdy;
jljC jrjCII2r;q.t; xIt1; /WD
t
Z
t1
dˇ Z
Rn
j@qxG.t; ˇIx y/jj@r˚.ˇ; yI; /jdy:
(3.18)
This convergence becomes obvious, if we take condition (A) and the following es- timates into account forfx; g Rnand0 < tT:
Ir;l;q1 .t1; xI; /cc2Eeb ı .tkx /˛k 11˛
.t t1/
nCCjlCqjC h
t1
Z
.ˇ /˛0 1dˇ;jljC jrjCI (3.19)
I2r;q.t; xIt1; / cc1Eeb ı .tkx /˛k
11˛
.t1 /
nCp1CCjrjC h
t
Z
t1
.t ˇ/˛0 1Cp1 j
qjC
h dˇ: (3.20) These estimates follow directly from (2.1), (3.13), and (3.14).
We now prove the validity of formula (3.17). For this purpose, we fix anyk2ZnC such thatjkjCDp1. Then, according to (3.16) forp1<jqjC˛Cp1andjrjC
˛;we have
@r@qxW .t; xI; /D
r
X
lD0
Crl@q kx
t1
Z
dˇ Z
Rn
@l@kxG.t; ˇIx y /@r l ˚.ˇ; yCI; /
dyC@q kx
t
Z
t1
dˇ Z
Rn
@kG.t; ˇI/@r˚.ˇ; x I; /d; .t; xI; /2˘T2: Hence, it remains to substantiate the possibility to introduce the operation@q kx under the signs of the corresponding integrals. In other words, we should prove the uniform convergence inxandof the following integrals onR2nfor0 < tT:
t1
Z
dˇ Z
Rn
@l@qxG.t; ˇIx y /˚.ˇ; yCI; /dy;
t
Z
t1
dˇ Z
Rn
@kG.t; ˇI/@r@q kx ˚.ˇ; x I; /d:
By reasoning similarly to the case of integrals (3.18) and using estimates (2.1), (3.13), and (3.14), we get the necessary convergence of the indicated integrals.
The lemma is proved.
The main result can be formulated as the following proposition.
Theorem 1. Let system(1.1)satisfy conditions (A) and (B). Then the correspond- ing function Z.t; xI; / defined by equality (2.2) is a function differentiable with respect to each of the spatial variablesx and on the set˘T2 to, respectively, the orders˛Cp1and˛inclusively, and
9ı > 08fr; qg ZnC;jqjC˛Cp1;jrjC˛;9c > 08.t; xI; /2˘T2 W j@r@qxZ.t; xI; /j c.t /
nCjrCqjCC
h e ı k
x k .t /˛
11˛
I (3.21)
j@kZ.t; xCI; /j ck.t /ˇk nChe ı1 kxk
.t /˛
11˛
; (3.22)
where jkjC˛, 0 < t T, fx; g Rn, ˇk WD
0; kD0;
˛0; k¤0 (here, the estimating constants are independent oft,,x;and).
Proof. With regard for structure (2.2) and the infinite differentiability of the func- tion G.t; I/ with respect to the variable , the smoothness of the function Z.t; xI; / in the variables x and becomes obvious directly from the assertion of Lemma2.
LetjqjCp1andjrjC˛. Then, according to (3.16), we get j@r@qxZ.t; xI; /j j@rx CqG.t; Ix /j C
r
X
lD0
CrlIr;l;q1 .t1; xI; /CIr;q2 .t; xIt1; /:
From whence, by using estimates (2.1), (3.19), and (3.20), we obtain assertion (3.21).
In a similar way, by using formula (3.17), we verify the validity of assertion (3.21) also forp1<jqjC˛andjrjC˛.
Then, according to estimates (2.1) and (3.14), we have Yk.t; xI; /WDˇ
ˇ ˇ
t
Z
dˇ Z
Rn
G.t; ˇIx /@k˚.ˇ; CI; /d ˇ ˇ ˇ
cc2 t
Z
.t ˇ/˛0Cp1h 1.ˇ /˛0 1 Z
Rn
exp n
ı0
nkx k .t ˇ/˛
11˛
C kk .ˇ /˛
11˛oo dydˇ
..t ˇ/.ˇ /˛ n; ı0WDminfı; ıg;jkjC˛: Using estimate (2.9) and equality (3.7), we get
Yk.t; xI; /c".t /˛0 nChe ı0.1 "/
kxk .t /˛
11˛
; "2.0I1/;