EMBEDDING THEOREMS IN BANACH-VALUED B-SPACES AND MAXIMAL B-REGULAR DIFFERENTIAL-OPERATOR EQUATIONS
VELI B. SHAKHMUROV
Received 28 September 2004; Revised 8 November 2005; Accepted 4 May 2006
The embedding theorems in anisotropic Besov-Lions type spacesBlp,θ(Rn;E0,E) are stud- ied; hereE0 and Eare two Banach spaces. The most regular spacesEαare found such that the mixed differential operatorsDαare bounded fromBlp,θ(Rn;E0,E) toBsq,θ(Rn;Eα), whereEαare interpolation spaces betweenE0andEdepending onα=(α1,α2,. . .,αn) and l=(l1,l2,. . .,ln). By using these results the separability of anisotropic differential-operator equations with dependent coefficients in principal part and the maximal B-regularity of parabolic Cauchy problem are obtained. In applications, the infinite systems of the quasielliptic partial differential equations and the parabolic Cauchy problems are stud- ied.
Copyright © 2006 Veli B. Shakhmurov. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
1. Introduction
Embedding theorems in function spaces have been studied in [8,35,37,38]. A com- prehensive introduction to the theory of embedding of function spaces and historical references may be also found in [37]. In abstract function spaces embedding theorems have been investigated in [4,5,10,17,21,27,34,40]. Lions and Peetre [21] showed that if
u∈L2
0,T;H0
, u(m)∈L2(0,T;H), (1.1)
then
u(i)∈L2
0,T;H,H0
i/m
, i=1, 2,. . .,m−1, (1.2) whereH0,H are Hilbert spaces,H0is continuously and densely embedded inH, where [H0,H]θare interpolation spaces betweenH0andH for 0≤θ≤1. The similar questions for anisotropic Sobolev spacesWpl(Ω;H0,H),Ω⊂Rnand for corresponding weighted
Hindawi Publishing Corporation Journal of Inequalities and Applications Volume 2006, Article ID 16192, Pages1–22 DOI10.1155/JIA/2006/16192
spaces have been investigated in [28–31] and [23,24], respectively. Embedding theorems in Banach-valued Besov spaces have been studied in [4,5,27,32]. The solvability and spectrum of boundary value problems for elliptic differential-operator equations (DOE’s) have been refined in [3–7,13,28–33,39,40]. A comprehensive introduction to DOE’s and historical references may be found in [15,18,40]. In these works, Hilbert-valued function spaces essentially have been considered. The maximalLpregularity and Fredholmness of partial elliptic equations in smooth regions have been studied, for example, in [1,2,20]
and for nonsmooth domains studied, for example, in [16,26]. For DOE’s the similar problems have been investigated in [13,28–32,36,39,40].
LetE0,Ebe Banach spaces such thatE0is continuously and densely embedded inE.
In the present paper,E-valued Besov spacesBl+sp,θ(Rn;E0,E)=Bsp,θ(Rn;E0)∩Bl+sp,θ(Rn;E) are introduced and called Besov-Lions type spaces. The most regular interpolation classEα
betweenE0 and Eis found such that the appropriate mixed differential operatorsDα are bounded fromBl+sp,q(Rn;E0,E) toBsp,q(Rn;Eα). By applying these results the maximal regularity of certain class of anisotropic partial DOE with varying coefficients in Banach- valued Besov spaces is derived.
The paper is organized as follows.Section 2collects notations and definitions.Section 3presents the embedding theorems in Besov-Lions type spaces
Bs+lp,qRn;E0,E. (1.3)
Section 4contains applications of the underlying embedding theorem to vector-valued function spaces. Section 5is devoted to the maximal regularity (in Bsp,q(Rn;E)) of the certain class of anisotropic DOE with variable coefficients in principal part. Then by us- ing these results the maximalB-regularity of the parabolic Cauchy problem is shown. In Section 6these DOE are applied to BVP’s and Cauchy problem for the finite and infinite systems of quasielliptic and parabolic PDEs, respectively.
2. Notations and definitions
LetEbe a Banach space. LetLp(Ω;E) denote the space of all strongly measurableE-valued functions that are defined onΩ⊂Rnwith the norm
fLp(Ω;E)= f(x)Epdx
1/ p
, 1≤p <∞, fL∞(Ω;E)=ess sup
x∈Ω
f(x)E, x=
x1,x2,. . .,xn
.
(2.1)
The Banach spaceEis said to be aζ-convex space (see [9,11,12,19]) if there exists onE×Ea symmetric real-valued functionζ(u,v) which is convex with respect to each of the variables, and satisfies the conditions
ζ(0, 0)>0, ζ(u,v)≤ u+v, foru ≤1≤ v. (2.2)
Aζ-convex spaceEis often called a UMD-space and written asE∈UMD. It is shown in [9] that the Hilbert operator
(H f)(x)=lim
ε→0
|x−y|>ε
f(y)
x−yd y (2.3)
is bounded inLp(R;E), p∈(1,∞) for those and only those spacesE, which possess the property of UMD spaces. The UMD spaces include, for example,Lp,lp spaces and the Lorentz spacesLpq,p,q∈(1,∞).
Let C be the set of complex numbers and let Sϕ=
λ;λ∈C,|argλ−π| ≤π−ϕ∪ {0}, 0< ϕ≤π. (2.4) A linear operatorAis said to be aϕ-positive in a Banach spaceE, with boundM >0 if D(A) is dense onEand
(A−λI)−1L(E)≤M1 +|λ|−1
(2.5) withλ∈Sϕ,ϕ∈(0,π],Iis identity operator inE, andL(E) is the space of all bounded linear operators inE. SometimesA+λIwill be written asA+λand denoted byAλ. It is known [37, Section 1.15.1] that there exist fractional powersAθof the positive operator A. LetE(Aθ) denote the spaceD(Aθ) with the graphical norm
uE(Aθ)=
up+Aθup1/ p, 1≤p <∞,−∞< θ <∞. (2.6) LetE0andEbe two Banach spaces. By (E0,E)σ,p, 0< σ <1, 1≤p≤ ∞we will denote the interpolation spaces obtained from{E0,E}by theK-method (see, e.g., [37, Section 1.3.1] or [10]).
LetS(Rn;E) denote a Schwartz class, that is, the space of allE-valued rapidly decreasing smooth functionsϕonRn.E=C will be denoted byS(Rn). LetS(Rn;E) denote the space ofE-valued tempered distributions, that is, the space of continuous linear operators from S(Rn) toE.
Letα=(α1,α2,. . .,αn),αiare integers. AnE-values generalized functionDαf is called a generalized derivative in the sense of Schwartz distributions of the generalized function
f ∈S(Rn,E) if the equality
Dαf,ϕ=(−1)|α|f,Dαϕ (2.7) holds for allϕ∈S(Rn).
By using (2.7) the following relations FDαxf=
iξ1
α1
,. . .,iξn
αnf, DαξF(f)=F−ixn
α1
,. . .,−ixn
αn
f (2.8) are obtained for all f ∈S(Rn;E).
LetL∗θ(E) denote the space of allE-valued function spaces such that uL∗θ(E)=
∞
0
u(t)θEdt t
1/θ
<∞, 1≤θ <∞, uL∗∞(E)= sup
0<t<∞
u(t)E. (2.9)
Lets=(s1,s2,. . .,sn) andsk>0. LetFdenote the Fourier transform. Fourier-analytic rep- resentation ofE-valued Besov space onRnis defined as
Bsp,θRn;E=
u∈SRn;E,uBsp,θ(Rn;E)
= F−1
n k=1
tκk−sk1 +ξkκk
e−t|ξ|2Fu
L∗θ(Lp(Rn;E))
,
p∈(1,∞), θ∈[1,∞],κk> sk
.
(2.10)
It should be noted that the norm of Besov space do not depend onκk. Sometimes we will writeuBsp,θin place ofuBsp,θ(Rn;E).
Letl=(l1,l2,. . .,ln),s=(s1,s2,. . .,sn), wherelkare integers andskare positive numbers.
LetWlBsp,θ(Rn;E) denote anE-valued Sobolev-Besov space of all functionsu∈Bsp,θ(Rn;E) such that they have the generalized derivativesDlkku=∂lku/∂xklk∈Bsp,θ(Rn;E),k=1, 2,. . .,n with the norm
uWlBsp,θ(Rn;E)= uBsp,θ(Rn;E)+ n k=1
DlkkuBs
p,θ(Rn;E)<∞. (2.11) LetE0is continuously and densely embedded intoE.WlBsp,θ(Rn;E0,E) denotes a space of all functionsu∈Bsp,θ(Rn;E0)∩WlBsp,θ(Rn;E) with the norm
uWlBsp,θ= uWlBsp,θ(Rn;E0,E)= uBsp,θ(Rn;E0)+ n k=1
Dklku
Bsp,θ(Rn;E)<∞. (2.12) Letl=(l1,l2,. . .,ln),s=(s1,s2,. . .,sn), whereskare real numbers andlkare positive num- bers.Bl+sp,θ(Rn;E0,E) denotes a space of all functionsu∈Bsp,θ(Rn;E0)∩Bl+sp,θ(Rn;E) with the norm
uBs+lp,θ(Rn;E0,E)= uBsp,θ(Rn;E0)+uBl+sp,θ(Rn;E). (2.13) ForE0=Ethe spaceBl+sp,θ(Rn;E0,E) will be denoted byBl+sp,θ(Rn;E).
Letmbe a positive integer.C(Ω;E) andCm(Ω;E) will denote the spaces of allE-valued bounded continuous andm-times continuously differentiable functions onΩ, respec- tively. We set
Cb(Ω;E)=
u∈C(Ω;E), lim
|x|→∞u(x) exists
. (2.14)
LetE1andE2be two Banach spaces. A functionΨ∈Cm(Rn;L(E1,E2)) is called a multi- plier fromBsp,θ(Rn;E1) toBsq,θ(Rn;E2) forp∈(1,∞) andq∈[1,∞] if the mapu→Ku= F−1Ψ(ξ)Fu,u∈S(Rn;E1), is well defined and extends to a bounded linear operator
K:Bsp,θRn;E1
−→Bsq,θRn;E2
. (2.15)
The set of all multipliers fromBsp,θ(Rn;E1) toBsq,θ(Rn;E2) will be denoted byMq,θp,θ(s,E1, E2).E1=E2=Ewill be denoted byMq,θp,θ(s,E). The multipliers and operator-valued mul- tipliers in Banach-valued function spaces were studied, for example, by [25], [37, Section 2.2.2.], and [4,11,12,14,22], respectively.
Let
Hk=
Ψh∈Mq,θp,θs,E1,E2
,h=
h1h2,. . .,hn∈K (2.16) be a collection of multipliers inMq,θp,θ(s,E1,E2). We say thatHkis a uniform collection of multipliers if there exists a constantM0>0, independent onh∈K, such that
F−1ΨhFuBs
p,θ(Rn;E2)≤M0uBq,θs (Rn;E1) (2.17) for allh∈Kandu∈S(Rn;E1).
Letβ=(β1,β2,. . .,βn) be multiindexes. We also define Vn=
ξ=
ξ1,ξ2,. . .,ξn∈Rn,ξi=0,i=1, 2,. . .,n, Un=
β:|β| ≤n, ξβ=ξ1β1ξ2β2,. . .,ξnβn, ν= 1 p−
1
q. (2.18)
Definition 2.1. A Banach spaceEsatisfies aB-multiplier condition with respect to p,q, θ, ands(or with respect to p,θ, ands for the case of p=q) whenΨ∈Cn(Rn;L(E)), 1≤p≤q≤ ∞,β∈Un, andξ∈Vnif the estimate
ξ1β1+νξ2β2+ν,. . .,ξnβn+νDβΨ(ξ)L(E)≤C (2.19)
impliesΨ∈Mq,θp,θ(s,E).
Remark 2.2. Definition 2.1is a combined restriction toE,p,q,θ, ands. This condition is sufficient for our main aim. Nevertheless, it is well known that there are Banach spaces satisfying theB-multiplier condition for isotropic case andp=q, for example, the UMD spaces (see [4,14]).
A Banach spaceEis said to have a local unconditional structure (l.u.st.) if there exists a constantC <∞such that for any finite-dimensional subspaceE0ofEthere exists a finite- dimensional spaceFwith an unconditional basis such that the natural embeddingE0⊂E factors asABwithB:E0→F,A:F→E, andAB ≤C. All Banach lattices (e.g.,Lp, Lp,q, Orlicz spaces,C[0, 1]) have l.u.st.
The expressionuE1∼uE2means that there exist the positive constantsC1andC2
such that
C1uE1≤ uE2≤C2uE1 (2.20) for allu∈E1∩E2.
Letα1,α2,. . .,αnbe nonnegative and letl1,l2,. . .,lnbe positive integers and let 1≤p≤q≤ ∞, 1≤θ≤ ∞, |α:.l| =
n k=1
αk
lk, κ= n k=1
αk+ 1/ p−1/q lk , Dα=Dα11Dα22,. . .,Dnαn= ∂|α|
∂x1α1∂x2α2,. . .,∂xαnn, |α| = n k=!
αk.
(2.21)
Consider in general, the anisotropic differential-operator equation (L+λ)u=
|α:.l|=1
aα(x)Dαu+Aλ(x)u+
|α:.l|<1
Aα(x)Dαu= f (2.22) inBsp,θ(Rn;E), whereaαare complex-valued functions andA(x),Aα(x) are possibly un- bounded operators in a Banach spaceE, here the domain definitionD(A)=D(A(x)) of operatorA(x) does not depend onx. Forl1=l2=,. . .,=lnwe obtain isotropic equations containing the elliptic class of DOE.
The function belonging to spaceBs+lp,θ(Rn;E(A),E) and satisfying (2.22) a.e. onRn is said to be a solution of (2.22) onRn.
Definition 2.3. The problem (2.22) is said to be aB-separable (orBsp,θ(Rn;E)-separable) if the problem (2.22) for all f ∈Bsp,θ(Rn;E) has a unique solutionu∈Bs+lp,θ(Rn;E(A),E) and
AuBsp,θ(Rn;E)+
|α:l|=1
DαuBs p,θ
Rn;E≤CfBsp,θ(Rn;E). (2.23)
Consider the following parabolic Cauchy problem
∂u(y,x)
∂y + (L+λ)u(y,x)=f(y,x), u(0,x)=0, y∈R+,x∈Rn, (2.24) whereLis a realization differential operator inBsp,θ(Rn;E) generated by problem (2.22), that is,
D(L)=Bs+lp,θRn;E(A),E, Lu=
|α:.l|=1
aα(x)Dαu+A(x)u+
|α:.l|<1
Aα(x)Dαu. (2.25) We say that the parabolic Cauchy problem (2.24) is said to be a maximalB-regular, if for all f ∈Bsp,θ(Rn+1+ ;E) there exists a unique solutionusatisfying (2.24) almost every- where onRn+1+ and there exists a positive constantCindependent on f, such that it has the estimate
∂u(y,x)
∂y
Bsp,θ(Rn+1+ ;E)+LuBsp,θ(Rn+1+ ;E)≤CfBsp,θ(Rn+1+ ;E). (2.26) 3. Embedding theorems
In this section we prove the boundedness of the mixed differential operatorsDα in the Besov-Lions type spaces.
Lemma 3.1. LetAbe a positive operator in a Banach spaceE, letbbe a positive number, r=(r1,r2,. . .,rn),α=(α1,α2,. . .,αn), andl=(l1,l2,. . .,ln), whereϕ∈(0,π],rk∈[0,b],lk
are positive andαk,k=1, 2,. . .,n, are nonnegative integers such thatκ= |(α+r) :l| ≤1.
For 0< h≤h0<∞and 0≤μ≤1−κthe operator-function
Ψ(ξ)=Ψh,μ(ξ)=ξ1r1ξ2r2,. . .,ξnrn(iξ)αA1−κ−μh−μA+η(ξ)−1 (3.1) is a bounded operator inEuniformly with respect toξandh, that is, there is a constantCμ such that
Ψh,μ(ξ)L(E)≤Cμ (3.2)
for allξ∈Rn, where
η=η(ξ)= n k=1
ξklk+h−1. (3.3)
Proof. Since−η(ξ)∈S(ϕ), for allϕ∈(0,π] andAis aϕ-positive inE, then the operator A+η(ξ) is invertiable inE. Let
u=h−μA+η(ξ)−1f . (3.4)
Then
Ψ(ξ)fE=(hA)1−κ−μuEh−(1−μ)h1/l1ξ1α1+r1,. . .,h1/lnξnαn+rn. (3.5) Using the moment inequality for powers of positive operators, we get a constantCμde- pending only onμsuch that
Ψ(ξ)fE≤Cμh−(1−μ)hAu1−κ−μuκ+μh1/l1ξ1α1+r1,. . .,h1/lnξnαn+rn. (3.6) Now, we apply the Young inequality, which states thatab≤ak1/k1+bk2/k2for any positive real numbersa,bandk1,k2with 1/k1+ 1/k2=1 to the product
hAu1−κ−μ
uκ+μh1/l1ξ1α1+r1,. . .,h1/lnξnαn+rn
(3.7) withk1=1/(1−κ−μ),k2=1/(κ+μ) to get
Ψ(ξ)fE≤Cμh−(1−μ)(1−κ−μ)hAu
+(κ+μ)h1/l1ξ1(α1+r1)/(κ+μ),. . .,h1/lnξn(αn+rn)/(κ+μ)u . (3.8) Since
n i=1
αi+ri
(κ+μ)= 1 κ+μ
n i=1
αi+ri
li = κ
κ+μ≤1, (3.9)
there exists a constantM0independent onξ, such that ξ1(α1+r1)/(κ+μ),. . .,ξn(αn+rn)/(κ+μ)≤M0
1 +
n k=1
ξklk
(3.10) for allξ∈Rn. Substituting this on the inequality (3.8) and absorbing the constant coeffi- cients inCμ, we obtain
ψ(ξ)f≤Cμ
hμ
Au+
n k=1
ξklku
+h−(1−μ)u
. (3.11)
Substituting the value ofuwe get ψ(ξ)f≤Cμhμ
AA+η(ξ)−1f+ n k=1
ξklkA+η(ξ)−1f
+h−(1−μ)
A+η(ξ)−1f.
(3.12)
By using the properties of the positive operatorAfor allf ∈Ewe obtain from (3.12)
Ψ(ξ)fE≤CμfE. (3.13)
Lemma 3.2. LetEbe a UMD space with l.u.st.,p∈(1,∞),θ∈[1,∞] and let for allk,j∈ (1,n)
sk
lk+sk+ sj
lj+sj ≤1. (3.14)
Then the spacesBl+sp,θ(Rn;E) andWlBsp,θ(Rn;E) are coincided.
Proof. In the first step we show that the continuous embeddingWlBsp,θ(Rn;E)⊂Bl+sp,θ(Rn; E) holds, that is, there is a positive constantCsuch that
uBl+sp,θ(Rn;E)≤CuWlBsp,θ(Rn;E) (3.15) for allu∈WlBsp,θ(Rn;E). For this aim by using the Fourier-analytic definition of anE- valued Besov space and the space WlBsp,θ(Rn;E) it is sufficient to prove the following estimate:
F−1
n k=1
tκk−lk−sk1 +ξkκk
e−t|ξ|2Fu
Lθp
≤CF−1 n k=1
tκk−sk1 +ξkκk
e−t|ξ|2Fυ
Lθp
, (3.16) where
Lθ p=L∗θLp
Rn;E, υ=F−1
1 + n k=1
ξklk
Fu. (3.17)
To see this, it is sufficient to show that the function φ(ξ)=
n k=1
1 +ξklk+sk+δn
k=1
1 +ξksk+δ−1 1 +
n k=1
ξklk−1
, δ >0 (3.18) is Fourier multiplier inLp(Rn;E). It is clear to see that forβ∈Unandξ∈Vn
ξ1β1ξ2β2,. . .,ξnβnDβφ(ξ)L(E)≤C. (3.19) Then in view of [41, Proposition 3] we obtain that the functionφis Fourier multiplier in Lp(Rn;E).
In the second step we prove that the embeddingBl+sp,θ(Rn;E)⊂WlBsp,θ(Rn;E) is contin- uous. In a similar way as in the first step we show that forsk/(lk+sk) +sj/(lj+sj)≤1 the function
ψ(ξ)= n
k=1
1 +ξksk+δ1 + n k=1
ξklk n
k=1
1 +ξklk+sk+δ
−1
(3.20) is Fourier multiplier inLp(Rn;E). So, we obtain for allu∈Bl+sp,θ(Rn;E) the estimate
F−1
n k=1
tκk−sk1 +ξkκk 1 +
n k=1
ξklk
e−t|ξ|2Fu Lθp
≤CF−1 n k=1
tκk−lk−sk1 +ξkκk
e−t|ξ|2Fu
Lθp
.
(3.21)
It implies the second embedding. This completes the prove ofLemma 3.2.
Theorem 3.3. Suppose the following conditions hold:
(1)Eis a UMD space with l.u.st. satisfying theB-multiplier condition with respect to p,q∈(1,∞),θ∈[1,∞], ands=(s1,s2,. . .,sn), whereskare positive numbers;
(2)α=(α1,α2,. . .,αn),l=(l1,l2,. . .,ln), whereαkare nonnegative,lkare positive integers, andsk such thatsk/(lk+sk) +sj/(lj+sj)≤1 fork,j=1, 2,. . .,nand 0≤μ≤1−κ,κ=
|(α+ 1/ p−1/q) :l|;
(3)Ais aϕ-positive operator inE, whereϕ∈(0,π] and 0< h≤h0<∞. Then the following embedding
DαBl+sp,θRn;E(A),E⊂Bq,θs Rn;EA1−κ−μ (3.22) is continuous and there exists a positive constantCμdepending only onμ, such that
DαuBs
q,θ(Rn;E(A1−κ−μ))≤CμhμuBl+s
p,θ(Rn;E(A),E)+h−(1−μ)uBsp,θ(Rn;E)
(3.23)
for allu∈Bl+sp,θ(Rn;E(A),E).