Vol. 20 (2019), No. 2, pp. 1013–1019 DOI: 10.18514/MMN.2019.3040
INTERMEDIATE REGULARITY RESULTS FOR THE SOLUTION OF A HIGH ORDER PARABOLIC EQUATION
AREZKI KHELOUFI Received 19 September, 2019
Abstract. In this work we give new intermediate regularity results for the solution of the follow- ing2m-th order parabolic equation
@tuC. 1/m
n
X
iD1
@2mxi uD0;
wheremis a positive integer, subject to Dirichlet condition on the lateral boundary of a cyl- indrical domain and to a non-homogeneous initial Cauchy data.
2010Mathematics Subject Classification: 35K05; 35K55 Keywords: regularity results, parabolic equations, Sobolev spaces
1. INTRODUCTION
Let˝be an open bounded set ofRnwith boundary andQthe cylinderRC˝ with lateral boundary˙ DRC :We assume that˝ is of classC2m:Consider in Qthe following boundary value problem:
8 ˆˆ ˆˆ
<
ˆˆ ˆˆ :
@tuC. 1/mPn
iD1@2mxi uD0 inQ;
@juD0 on˙,j D0; 1; :::; m 1;
u .0; x/Du0.x/,x2˝;
(1.1)
wheremis a positive integer and@jis the derivative of orderj throughout the normal vectoron˙:
Classical results on the resolution of Problem (1.1) when the initial Cauchy data u0belongs toL2.˝/or to the usual Sobolev spaceH0m.˝/can be found in [6], see also [1], [2], [8] and [9] and the references therein.
Our interest in this work is the regularity of the solutionuof (1.1) in terms of the regularity of the initial datau0. More precisely, we are interested on the question of the regularity of the solution of (1.1) whenu0is ”between”H0m.˝/andL2.˝/ :The
c 2019 Miskolc University Press
results obtained here complement those obtained for the heat equation, i.e.mD1, in [4].
The organization of this paper is as follows. In Section 2, we begin by preliminar- ies where we define the basic functional spaces, in which we will work and we give some of their properties needed for our study. Then, we recall a classical result for Problem (1.1) and we prove a fundamental lemma which will allow us to prove our main result in Section 3.
2. PRELIMINARIES
2.1. Function spaces
In this subsection, we recall the definitions of the basic functional spaces, in which we will work. We will need some anisotropic Sobolev spaces (see [6]), which we recall in the following definitions
Hr;s Rn Dn
u2L2 Rn Wh
1C2r=2
C 1C2s=2i
bu2L2 Rno wherebuis the Fourier transform ofuandr; sare two non-negative numbers. We put
Hr;s.˝/D˚
uj˝ Wu2Hr;s Rn ;
with˝ is an open subset ofRn. For0r1 and a positive integerm;we recall that the spaceHr;2mr.Q/, whereQDRC˝;can be defined by
Hr;2mr.Q/DL2 RC; H2mr.˝/
\Hr RC; L2.˝/
and the spaceH0r.˝/is defined by
H0r.˝/D fu2Hr.˝/IuD0on g for 12< r1, is the boundary of˝:
H0r.˝/DH
1 2
00.˝/
forrD12;
H0r.˝/DHr.˝/
for0r < 12.
We will need also some interpolation spaces in Hilbert spaces (see [7] and [3]), which we recall in the following definition.
LetX,Y be two Hilbert spaces with
X Y continuously.
Here, we give one of the usual methods, namely, that of Lions-Peetre [7] which allow us to build spaces
ŒX; Y 0 < < 1;
”intermediate” betweenX andY:
Definition 1. The space ŒX; Y 0 < < 1; is a sub-space of Y consisting of elementsawhich can be written in the form
aD Z 1
0
u .t /dt
t (2.1)
with
tu .t /2L2.X /,t 1u .t /2L2.Y / : (2.2) This space is endowed with the norm
a7 !inf
"
Z 1 0
t2ju .t /jX2
dt t
12 C
Z 1 0
t2. 1/ju .t /j2Y
dt t
12# I the inf is taken with respect touverifying (2.1) and (2.2).
HereL2.V / denotes the space of functions fromt > 0 with values inV, which are square integrable for the Haar measuredt =t:
Example 1. Hr;s.˝/can also be defined as a real interpolation space between Hr=.1 /;s=.1 /.˝/andL2.˝/ ; 20; 1Œ ;(see [10])
Hr;s.˝/Dh
Hr=.1 /;s=.1 /.˝/ ; L2.˝/i
: (2.3)
In this work, we consider the casesD2mr,D1 r;
Hr;2mr.˝/D
H1;2m.˝/ ; L2.˝/
1 r 8r20; 1Œ : (2.4) The spaceHr;2mr.˝/is well defined by Relationship (2.4) because the right hand side term of (2.4) is well defined as an interpolation space between two well defined spacesH1;2m.˝/andL2.˝/ :
Example2. The usual Sobolev spacesHs.˝/(s0) can be defined by interpol- ation
Hs.˝/Dh
Hs=.1 /.˝/ ; L2.˝/
i
; 20; 1Œ :
Hereafter some interpolation theory properties on spacesŒ:; : (see Triebel [10]), needed for proving our main result in the following section.
Theorem 1([10]). LetA0,A1be two Hilbert spaces with A0A1continuously.
Then
L2.A0/ ; L2.A1/
DL2.ŒA0; A1/ ; 0 < < 1.
A direct consequence of Theorem1is
Corollary 1. For each0r1and any positive integerm;we have L2 RC; H2m.˝/
; L2.RC; Hm.˝//
1 r D L2 RC;
H2m.˝/ ; Hm.˝/
1 r
D L2
RC; Hm.1Cr/.˝/
: with˝an open bounded set ofRn:
2.2. Lemmas
The next result is well known ( cf. Lions and Magenes [6]).
Lemma 1. 1) For givenu0inH0m.˝/ ;Problem (1.1) has a unique solutionuin H1;2m.Q/defined by
H1;2m.Q/DL2 RC; H2m.˝/
\H1 RC; L2.˝/
: 2) For givenu0inL2.˝/, Problem (1.1) has a unique weak solutionuin
L2 RC; H0m.˝/
\H1 RC; H m.˝/
\L1 RC; L2.˝/
:
We will need the following lemma for proving our main result in the following section.
Lemma 2. Letu02L2.˝/. Then the solutionuof Problem (1.1) associated to u0is inH12 RC; L2.˝/
:Moreover, there exists a positive constantC (independent ofu0) such that
kukH12.RC;L2.˝//Cku0kL2.˝/:
Proof. Consider a sequence of spectral elements.k; 'k/ ; k2Nof the Dirichlet problem for the operator. 1/mPn
iD1@2mx
i
8 ˆˆ ˆˆ
<
ˆˆ ˆˆ :
. 1/mPn
iD1@2mxi 'k Dk'k
'k2H0m.˝/
k'kkL2.˝/D1:
The sequence.'k/k2Nis a basis ofL2.˝/. Ifu02L2.˝/we may write u0.x/D X
k2N
ak'k.x/
withku0kL22.˝/DP
k2Nak2:The solution associated tou0is u .t; x/DX
k2N
akexp. kt / 'k.x/ : Noteeuthe extension ofutoR, i.e.,
eu .t; x/DX
akexpj ktj'k.x/ :
By the Fourier transform
beu .; x/DCX akk
2C2k'k.x/ ; whereC is a constant, from which
beu .; :/
2
DC2X
a2k 2k 2C2k2
and by elementary calculations, we check easily that Z
Rjj beu .; :/
2
d DC0X ak2; whereC0is a constant. Consequently
eu2H12 R; L2.˝/
; then
u2H12 RC; L2.˝/
by restriction ofeutot > 0:
3. MAIN RESULT
In the sequel, we will assume thatu02H0rm.˝/ ; 0r1:ThusH0rm.˝/is the interpolation space of order1 r betweenH0m.˝/andL2.˝/ :Indeed, it suffices to takeD1 randsDrmin Example2. We look for the regularity ofuin term ofr:Our main result in this work is
Theorem 2. For given u0 in H0rm.˝/ 0r 1; Problem (1.1) has a unique weak solutionuinH1C2r;m.1Cr/.Q/ :
Proof. Letu02H0rm.˝/ 0r1thenu02L2.˝/and consequently (1.1) ad- mits a unique weak solution (see Lemma1)uinL2 RC; H0m.˝/
:In order to show that this solution is inL2
RC; Hm.1Cr/.˝/
it suffices to interpolate the operator S which associatesutou0. IndeedS Wu07!uis linear continuous fromH0m.˝/
toL2 RC; H2m.˝/
and fromL2.˝/toL2.RC; Hm.˝// :By interpolation, it is linear continuous from
H0m.˝/ ; L2.˝/
1 r into
L2 RC; H2m.˝/
; L2 RC; Hm.˝/
1 r: Thanks to Corollary1,S is linear continuous from
H0rm.˝/ intoL2
RC; Hm.1Cr/.˝/
:
We can interpolate againSfor proving thatu2HrC21 RC; L2.˝/
:Indeed,Sis lin- ear continuous fromL2.˝/intoH12 RC; L2.˝/
(see Lemma2) and fromH0m.˝/
intoH1 RC; L2.˝/
(see Theorem1). By interpolation, it is linear continuous from H0m.˝/ ; L2.˝/
1 r into h
H12 RC; L2.˝/
; H1 RC; L2.˝/i
1 r: But, (see Triebel [10])
h
H12 RC; L2.˝/
; H1 RC; L2.˝/i
1 rDHrC21 RC; L2.˝/
: Then,Sis linear continuous fromH0rm.˝/intoHrC21 RC; L2.˝/
:This ends the
proof of Theorem2.
Remark1. Note that the anisotropic Hilbert spaces used in this work may be ex- tended to the case of parabolic Sobolev spaces built onLp; p¤2:An idea for this extension can be found in [5].
ACKNOWLEDGEMENT
The author is grateful to anonymous reviewers for carefully reading this paper and for their valuable comments and suggestions, which have improved the paper.
REFERENCES
[1] S. Gala, Q. Liu, and M. A. Ragusa, “Logarithmically improved regularity criterion for the nematic liquid crystal flows inB11;1space.”Computers and Mathematics with Applications, vol. 65, pp.
1738–1745, 2013, doi:10.1016/j.camwa.2013.04.003.
[2] V. A. Galaktionov, “On regularity of a boundary point for higher-order parabolic equations: to- wards Petrovskii-type criterion by blow-up approach.”Nonlinear Differ. Equ. Appl., vol. 16, no. 5, pp. 597–655, 2009, doi:10.1007/s00030-009-0025-x.
[3] P. Grisvard, “P. Grisvard, Caract´erisation de quelques espaces d’interpolation.”Arch. Rational Mech. Anal., vol. 25, no. 1, pp. 40–63, 1967, doi:10.1007/BF00281421.
[4] A. Kheloufi and B. Sadallah, “On the regularity of the heat equation solution: Two ap- proaches.”Applied Mathematics and Computation, vol. 218, no. 5, pp. 1623–1633, 2011, doi:
10.1016/j.amc.2011.06.042.
[5] J. L. Lions, “Quelques remarques sur les probl`emes de Dirichlet et de Neumann.”S´eminaire Jean Leray, no. 6, pp. 1–18, 1961-1962.
[6] J. L. Lions and E. Magenes,Probl`emes aux Limites Non Homog`enes et Applications 1, 2. Paris:
Dunod, 1968.
[7] J. L. Lions and J. Peetre, “Sur une classe d’espaces d’interpolation.”Publ. Math. de l’I. H. E. S., no. 19, pp. 5–68, 1964, doi:10.1007/BF02684796.
[8] M. A. Ragusa, “Cauchy-Dirichlet problem associated to divergence form parabolic equations.”
Communications in Contemporary Mathematics, vol. 6, no. 3, pp. 377–393, 2004.
[9] B. Sadallah, “Spectral properties of even order derivative operators.”Appl. Math. E-Notes, pp.
10–15, 2003.
[10] H. Triebel,Interpolation Theory, Function Spaces, Differential Operators. Amsterdam - New York - Oxford: North-Holland Publishing Company, 1978.
Author’s address
Arezki Kheloufi
Bejaia University, Department of Technology, Lab. of Applied Mathematics, 06000 Bejaia, Algeria E-mail address:arezkinet2000@yahoo.fr