Universität Konstanz
Pseudodifferenzial operators with non-regular operator-valued symbols
Bienvenido Barraza Martínez Robert Denk
Jairo Hernández Monzón
Konstanzer Schriften in Mathematik Nr. 307, September 2012
ISSN 1430-3558
© Fachbereich Mathematik und Statistik Universität Konstanz
Fach D 197, 78457 Konstanz, Germany
Konstanzer Online-Publikations-System (KOPS) URL: http://nbn-resolving.de/urn:nbn:de:bsz:352-202597
OPERATOR-VALUED SYMBOLS
BIENVENIDO BARRAZA MART´INEZ, ROBERT DENK, AND JAIRO HERN ´ANDEZ MONZ ´ON
Abstract. In this paper, we consider pseudodifferential operators with operator- valued symbols and their mapping properties, without assumptions on the un- derlying Banach spaceE. We show that, under suitable parabolicity assump- tions, theWpk(Rn, E)-realization of the operator generates an analytic semi- group. An application to non-autonomous pseudodifferential Cauchy problems gives the existence and uniqueness of a classical solution. Our approach is based on oscillatory integrals and kernel estimates for them.
1. Introduction
In this paper, we consider pseudodifferential operators with operator-valued sym- bols, their mapping properties, generation of an analytic semigroup and the ap- plication to non-autonomous vector-valued evolution equations. Operator-valued symbols and vector-valued function spaces appear in a natural way in several appli- cations, e.g. in models of coagulation-fragmentation processes where an additional parameter (the cluster size) appears in the model, see [Am00]. Another example is given by boundary value problems in cylindrical domains which can be treated by a Fourier multiplier approach which leads to operator-valued symbols, too (see [NS11]). Therefore, during the last decade, the investigation of vector-valued func- tion spaces and differential equations with operator-valued coefficients has gained increasing interest.
Our work is motivated by two directions of research: On one hand, operator- valued Fourier multipliers inarbitrary Banach spaces and their mapping properties have been considered in [Am97]. Based on these Fourier multiplier results, the generation of an analytic semigroup for differential operators could be shown in [Am01]. On the other hand, under an additional geometric assumption on the Ba- nach space (to be a UMD space), Weis could establish in [We01] a vector-valued Mikhlin type theorem (see also [NNH02], [GW03]). This was the basis for a large number of results on maximal regularity for differential and pseudodifferential op- erators in UMD spaces. As references for operator-valued differential boundary value problems in UMD spaces and the connection toR-sectoriality and maximal regularity, we mention [DHP03] and [KW04]. For vector-valued pseudodifferential operators in UMD spaces, we refer to [DK07], [PS06] and the references therein.
Date: August 31, 2012.
2010Mathematics Subject Classification. 35S05, 47D06, 35R20.
Key words and phrases. Pseudodifferential operators, operator-valued symbols, generation of analytic semigroup.
1
The restriction to UMD spaces, however, excludes natural state spaces asL1, and therefore the present paper deals withpseudodifferential operators with operator- valued symbols in arbitrary Banach spaces. We consider operator-valued symbols ain the standard H¨ormander class with limited smoothness both in the variablex and in the covariableξand with positive orderm >0. Similar operators were also considered in [Ki01],[Ki03] with additional assumptions on the symbol (e.g., order m >1, existence of a homogeneous principal part, infinite smoothness inξ).
One of the main results in the present paper states that, under suitable parabol- icity assumptions, theWpk(Rn, E)-realization of the symbolagenerates an analytic semigroup in Wpk(Rn, E), see Theorem 4.2 and Corollary 4.3 below. An appli- cation to non-autonomous pseudodifferential Cauchy problems gives the existence and uniqueness of a classical solution (Theorem 5.3). Our approach is based on oscillatory integrals and careful kernel estimates for them, the technical key result being Lemma 3.2. For the case of constant coefficients, results on vector-valued pseudodifferential operators were obtained in our paper [BDH12].
2. Vector-valued pseudo-differential operators In the following, we set |x| := Pn
j=1|xj|2)1/2, |x, µ| := (|x|2+µ2)1/2, hxi:=
(1 +|x|2)1/2, and hx, µi := (1 +|x|2 +µ2)1/2 for x = (x1, . . . , xn) ∈ Rn and µ∈R. Throughout this paper, (E,k · k) denotes an arbitrary Banach space, and for locally convex spacesX, Y we writeL(X, Y) for the space of all continuous linear operators fromX toY, and setL(X) :=L(X, X). LetCb∞(Rn, E) be the space of allu:Rn→Esuch that∂αuis bounded and continuous for allα∈Nn0 where we use standard multi-index notation. The spaceCb∞(Rn, E) is endowed with the locally convex topology given by the seminorms kukk := max|α|≤ksupx∈Rnk∂αu(x)kE, k∈N0. The Schwartz space of rapidly decreasingE-valued functions is denoted by S(Rn, E), and fork ∈N0 and 1≤p <∞, we write Wpk(Rn, E) for the E-valued Sobolev space endowed with the norm
kukWk
p(Rn,E):= X
|α|≤k
k∂αukpLp(Rn,E)
1/p
.
Here k · kLp(Rn,E) stands for the norm in the Lebesgue-Bochner space Lp(Rn, E).
We setRn+:=Rn×[0,∞),Dα:= (−i)|α|∂α andd¯(ξ, y) := (2π)−nd(ξ, y).
We start with the definition of the symbol class (which is a non-smooth parameter- dependent version of the standard H¨ormander classS1,0m) and the related pseudo- differential operators. In the following, for n ∈ N and m ∈ R, we set ρn :=
(n+ 1, ifnis odd,
n+ 2, ifnis even. andρn,m:=
([n+m] + 1, if [n+m] is odd, [n+m] + 2, if [n+m] is even.
Definition 2.1. Let m∈R,ν ∈[0,∞], andr∈N0.
a) We defineSm,ν,r :=Sm,ν,r(Rnx×Rn+, L(E))as the set of all functionsa:Rnx× Rn+→L(E)such that
(i) x7→a(x, ξ, µ)∈Cr(Rn, L(E))for all(ξ, µ)∈Rn+, (ii) (ξ, µ)7→a(x, ξ, µ)∈C[ν](Rn+, L(E))for all x∈Rn, and (iii) |a|(ν,r)m <∞ where
|a|(ν,r)m := sup
x∈Rn
sup
(ξ,µ)∈Rn+
max
|α|+k≤νmax
|β|≤rhξ, µi−m+|α|+k
∂xβ∂ξα∂µka(x, ξ, µ) L(E).
b) Fora∈Sm,ν,rwithν ≥ρn, the pseudo-differential operatorop(a) :=a(x, D, µ) is defined by
[a(x, D, µ)u](x) := Os− Z Z
R2n
eiξ·ya(x, ξ, µ)u(x−y)d¯(ξ, y) (x∈Rn)
foru∈Cb∞(Rnx, E).
For the convenience of the reader, we state the definition and some properties of the oscillatory integral in the appendix.
In the following, we will also study parameter-dependent double symbols a = a(x, y, ξ, µ) witha∈Sm,ν,r(R2nx,y×Rn+, L(E)) withν ≥ρn andr > ρn,m. Here we set
(1) [a(x, y, D, µ)u](x) := Os− Z Z
R2n
eiξ·ya(x, x−y, ξ, µ)u(x−y)d¯(ξ, y) (x∈Rn)
foru∈Cb∞(Rn, E). In the particular case wherea(x, y, ξ, µ) does not depend onx (dual symbol), we will writea(y, D, µ).
Remark 2.2. a) Forν≥ρn, Lemma A.4 shows that the oscillatory integral in b) exists and due to Lemma A.5 too, that a(x, D, µ)∈L(Cb∞(Rn, E), Cbr(Rn, E))for every fixedµ∈[0,∞).
b) Forν ≥ρn andr > ρn,m, the oscillatory integral in (1) exists. Note that for n+m≤0 the conditionr > ρn,m is satisfied for allr∈N.
The following result is taken from [Ki03], Remark 2.2.6.
Lemma 2.3. Let ai ∈ Smi,ν,r, i = 1,2, with m1, m2 ∈ R, ν ≥ ρn and r ≥ ρ2n,m1+m2. Then a1(x, D, µ)a2(y, D, µ) = op a1(x, ξ, µ)a2(y, ξ, µ)
.
Definition 2.4. a) Let m, ν ∈(0,∞)and r∈N0, and leta∈ Sm,ν,r. Then ais called parameter-elliptic if there are constants κ > 0 and ω ≥0 such that for all (x, ξ, µ)∈Rn×Rn×[0,∞)with|ξ, µ| ≥ω we have
(i) a(x, ξ, µ) :E→E is bijective, and (ii) ka(x, ξ, µ)−1kL(E)≤κhξ, µi−m.
The set of alla∈Sm,ν,r satisfying (i)–(ii) will be denoted byEκ,ωm,ν,r:=Eκ,ωm,ν,r(Rnx× Rn+, L(E)).
b) Letm∈(0,∞)andr, ν∈N0, and let a∈Sm,ν,r(Rnx×Rnξ, L(E))be a symbol independent of µ∈(0,∞). Thenais called parabolic with constants κ >0,ω≥0 if (x, ξ, µ)7→a(x, ξ) +µmeiθ belongs toEκ,ωm,ν,r for all|θ| ≤ π2.
For a symbol a ∈Eκ,ωm,ν,r, we define a smooth version of the inverse symbol in the following way: Letψ ∈C∞(Rn+1) with 0≤ψ≤1,ψ(ξ, µ) = 0 for |ξ, µ| ≤ 12 andψ(ξ, µ) = 1 for|ξ, µ| ≥1 be a zero extinction function. Forω0> ω we set
a#(x, ξ, µ) :=
(ψ(2ωξ
0,2ωµ
0)a(x, ξ, µ)−1, ifx∈Rn,|ξ, µ| ≥ω0, 0, ifx∈Rn,|ξ, µ| ≤ω0. Thena#(x, ξ, µ) = 0 if|ξ, µ| ≤ω0anda#(x, ξ, µ) =a(x, ξ, µ)−1 if|ξ, µ| ≥2ω0.
3. Key estimates for parameter-elliptic symbols
In this section, we will prove key estimates for the inverse of parameter-elliptic symbols and mapping properties for the corresponding operators. We start with a result which shows that the inverse of a parameter-elliptic symbol is again in the calculus.
Lemma 3.1. Let m, ν∈(0,∞),r∈N0, and letA ⊂Sm,ν,r be bounded, i.e. there exists a constant K > 0 such that |a|(ν,r)m ≤K (a ∈ A). Assume moreover that A ⊂Eκ,ωm,ν,r. Then there exists a constantC=C(K, κ, ν, r)>0 such that
k∂xβ∂ξα∂kµa#(x, ξ, µ)kL(E)≤Chξ, µi−m−|α|−k
for all x∈Rn,(ξ, µ)∈Rn+,|α|+k≤ν,|β| ≤r, and a∈ A, i.e., {a# :a∈ A} ⊂ S−m,ν,r is bounded.
Proof. Note that{hξ, µim+|α|+kk∂βx∂ξα∂kµa#(x, ξ, µ)kL(E):x∈Rn,|ξ, µ| ≤2ω0,|α|+
k≤ν,|β| ≤r} is bounded by continuity and compactness. Therefore, it suffices to consider the case|ξ, µ| ≥2ω0.
Letx∈Rn,|ξ, µ| ≥2ω0, α, β ∈Nn0 \ {0}, k∈N0 with |β| ≤rand |α|+k≤ν.
Then∂xβ∂ξα∂µka(x, ξ, µ)−1 is a finite linear combination of terms of the form (2) a(x, ξ, µ)−1
∂xβ(1)∂ξα(1)∂µk(1)a(x, ξ, µ)
a(x, ξ, µ)−1. . . . . .
∂xβ(p)∂ξα(p)∂µk(p)a(x, ξ, µ)
a(x, ξ, µ)−1
with 1≤p≤ |α|+|β|+kand withα(i), β(i)∈Nn0,k(i)∈N(i= 1, . . . , p) satisfying Pp
i=1α(i)=α,Pp
i=1β(i)=β,Pp
i=1k(i)=k. The norm of the operator in (2) can be estimated by
ka(x, ξ, µ)−1kp+1L(E)
p
Y
i=1
∂βx(i)∂ξα(i)∂µk(i)a(x, ξ, µ) L(E)
≤κp+1hξ, µi−m(p+1)
p
Y
i=1
|a|(ν,r)m hξ, µim−|α(i)|−k(i)
=κp+1 |a|(ν,r)m p
hξ, µi−m−|α|−k. Summing up over all terms of the form (2), we obtain
k∂xβ∂αξ∂µka−1(x, ξ, µ)kL(E)≤Chξ, µi−m−|α|−k
with a constantC=C(K, κ, ν, r).
From now on we set χε(ξ, y) := χ(εξ)χ(εy) for ε > 0 and ξ, y ∈ Rn, where χ∈S(Rn) withχ(0) = 1. The following result is the key estimate for parameter- dependent symbols of negative order.
Lemma 3.2. Let b ∈ S−m,ν,r with m > 0, ν ≥n+ 1, r ∈N0 and ω0 >0. For ε∈(0,1) define
Kε(x, y, µ) :=
Z
Rn
eiξ·yχε(ξ, y)b(x−y, ξ, µ)dξ (x, y, µ)∈R2n×[ω0,∞) . a) There exists a constantC such that
kKε(x, y, µ)kL(E)≤C|b|(ν,r)−mµ−m+n |µy|θ0+|µy|θ1
|µy|n(1 +|µy|) (x, y, µ)∈R2n×[ω0,∞)
whereθ0:= 12min{m,1}andθ1:= 12. Moreover,
(3) kKε(x,·, µ)kL1(Rn,L(E))≤C|b|(ν,r)−mµ−m (x, µ)∈Rn×[ω0,∞) holds with a constantC independent of ε,xandµ.
b) There exists a strongly measurable function K: R2n×[ω0,∞)→L(E) with Kε(x, y, µ) → K(x, y, µ) (ε & 0) pointwise, and the estimate (3) holds with Kε being replaced byK.
Proof. a) Substitutingξ7→µξ, we have Kε(x, y, µ) =µn
Z
Rn
eiµξ·yχε(µξ, y)b(x−y, µξ, µ)dξ.
We fix γ ∈ Nn0 with |γ| = n+i, i ∈ {0,1}. By assumption, b(x−y,·, µ) ∈ C|γ|(Rn, L(E)) and therefore
Z
Rn
Dξγ
χε(µξ, y)b(x−y, µξ, µ) dξ= 0.
With this and partial integration, we see (µy)γKε(x, y, µ) =µn
Z
Rn
(eiµξ·y−1)Dξγ
χε(µξ, y)b(x−y, µξ, µ) dξ.
To estimate the integrand, we apply the Leibniz rule noting that forα+β =γwe have
Dαξχε(µξ, y) =
Dαξχ(εµξ)
|χ(εy)| ≤cαµ|α|hµξi−|α|
by Lemma A.3. This and
(4) kDξβ[b(x−y, µξ, µ)]kL(E)=µ|β|k(Dξβb)(x−y, µξ, µ)kL(E)
≤µ|β||b|(ν,r)−mhµξ, µi−m−|β|
gives
Dγξ
χε(µξ, y)b(x−y, µξ, µ) L(E)
≤ X
α+β=γ
cαβµ|α|hµξi−|α||µ||β|hµξ, µi−m−|β||b|(ν,r)−m
≤cγµ|γ|hµξi−m−|γ||b|(ν,r)−m
where we usedhµξ, µi−m−|β|≤ hµξi−m−|β|. Now we apply the elementary estimate
|eiµξy−1| ≤2|µy|θ|ξ|θ valid for allθ∈(0,1) and obtain k(µy)γKε(x, y, µ)kL(E)≤Cµn+|γ||µy|θ|b|(ν,r)−m
Z
Rn
|ξ|θhµξi−m−|γ|dξ
=Cµn−m|µy|θ|b|(ν,r)−mI(θ,|γ|) withI(θ,|γ|) :=R
Rn|ξ|θ(µ−2+|ξ|2)−m/2−|γ|/2dξ.
For|γ|=n+iwithi∈ {0,1}andθ0:=12min{m,1},θ1:= 12, we haveθi∈(0,1) and
I(θi,|γ|) = Z
|ξ|≤1
|ξ|θi(µ−2+|ξ|2)−m/2−|γ|/2dξ +
Z
|ξ|≥1
|ξ|θi(µ−2+|ξ|2)−m/2−|γ|/2dξ
≤ω0−m−|γ|
Z
|ξ|≤1
|ξ|θidξ+ Z
|ξ|≥1
|ξ|θi−m−|γ|dξ≤C <∞
due toθi>0 andθi−m− |γ|<−n. Therefore, for allγwith|γ|=n+i,i∈ {0,1}, we have
k(µy)γKε(x, y, µ)kL(E)≤Cµn−m|µy|θi|b|(ν,r)−m and consequently
|µy|n+ikKε(x, y, µ)kL(E)≤n(n+i)/2 X
|γ|=n+i
k(µy)γKε(x, y, µ)kL(E)
≤Cµn−m|µy|θi|b|(ν,r)−m.
Summing up these inequalities fori= 0 andi= 1, we obtain the first assertion in a). Since the function
y7→µn |µy|θ0+|µy|θ1
|µy|n(1 +|µy|)
belongs toL1(Rn) and itsL1-norm does not depend onµ, we obtain inequality (3).
b) Letε, ε0∈(0,1), ξ, y∈Rn andµ∈[ω0,∞). From the proof of a) we see that (5)
(µy)γ(Kε(x, y, µ)−Kε0(x, y, µ)
=µn Z
Rn
(eiµξ·y−1)Dξγ
χε(µξ, y)−χε0(µξ, y)
b(x−y, µξ, µ) dξ.
From Lemma A.3 we know that Dαξ χε(µξ, y)−χε0(µξ, y)
→0 for ε, ε0 & 0 for all α∈ Nn0 and all µ, ξ, y. Therefore the integrand in (5) converges pointwise to zero for ε, ε0 & 0. By a), we haveKε(x,·, µ)∈ L1(Rn, L(E)) with a dominating function independent of ε, and by dominated convergence we see that for fixed (x, y, µ)∈ R2n×[ω0,∞) we get kKε(x, y, µ)−Kε0(x, y, µ)kL(E) → 0 (ε, ε0 & 0).
Therefore there exists a strongly measurable function K: R2n×[ω0,∞) →L(E) such thatKε→K(ε&0) pointwise. By dominated convergence again, inequality
(3) holds forK(x,·, µ) instead ofKε(x,·, µ).
Theorem 3.3. Letm >0,1≤p <∞,r, k, ν∈N0withr≥ρn,−m+k+ 1,ν ≥ρn, andω0>0. Let b∈S−m,ν,r. Then b(y, D, µ)∈L(Wpk(Rn, E))and we have
kb(y, D, µ)kL(Wk
p(Rn,E))≤Cµ−m|b|(ν,r)−m (µ∈[ω0,∞)) with a constantC independent of bandµ.
Proof. (i) Consider first the casek= 0. Foru∈S(Rn, E) we have (b(y, D, µ)u)(x) = Os−
Z Z
R2n
eiξ·yb(x−y, ξ, µ)u(x−y)d¯(ξ, y)
= lim
ε&0
Z
Rn
Z
Rn
eiξ·yχε(ξ, y)b(x−y, ξ, µ)u(x−y)d¯ξd¯y
= lim
ε&0
Z
Rn
Kε(x, y, µ)u(x−y)d¯y with Kε(x, y, µ) := R
Rneiξ·yχε(ξ, y)b(x−y, ξ, µ)d¯ξ. By Lemma 3.2, Kε(x,·, µ) ∈ L1(Rn, L(E)) withε-independent dominating function, andKεconverges pointwise
to a strongly measurable function K for ε & 0. By the dominated convergence theorem, we have
(b(y, D, µ)u)(x) = Z
Rn
K(x, y, µ)u(x−y)d¯y= [K(x,·, µ)∗u](x)
where ∗ stands for the standard convolution. Because of theL1-estimate of K in Lemma 3.2, we obtain forµ∈[ω0,∞)
kb(y, D, µ)ukLp(Rn,E)≤CkK(x,·, µ)kL1(Rn,L(E))kukLp(Rn,E)
≤Cµ−m|b|(ν,r)−mkukLp(Rn,E).
(ii) Now letk∈N. By Lemma A.5, forβ ∈Nn0 with|β| ≤kwe write
∂xβ[b(y, D, µ)u]
(x) = Os− Z Z
R2n
eiξ·y∂xβ
b(x−y, ξ, µ)u(x−y) d¯(ξ, y)
= X
β0≤β
cββ0Os− Z Z
R2n
eiξ·y(∂xβ0b)(x−y, ξ, µ)(∂xβ−β0u)(x−y)d¯(ξ, y).
From∂xβ0b∈Sm,ν,r−|β0|and part (i) of the proof we obtain k∂xβ[b(y, D, µ)u]kLp(Rn,E)≤Cµ−m|b|(ν,r)−m X
|γ|≤k
k∂γukLp(Rn,E)
≤Cµ−m|b|(ν,r)−mkukWk p(Rn,E).
Remark 3.4. The last result shows that the mapping b 7→ b(y, D, µ), S−m,ν,r → L(Wpk(Rn, E))is continuous with norm not greater than a constant timesµ−mfor µ∈[ω0,∞).
Corollary 3.5. In the situation of Lemma 3.1, we havea#(y, D, µ)∈L(Wpk(Rn, E)) and
sup
a∈A, µ≥0
ka#(y, D, µ)kL(Wk
p(Rn,E))<∞ for allk∈N0 withr≥ρn,−m+k+ 1.
Proof. This follows forµ≥ω0 immediately from Lemma 3.1 and Remark 3.4 while forµ≤ω0 we havea#(y, D, µ) = 0 for alla∈ A.
4. Generation of an analytic semigroup
We will need the following slight generalization of Lemma 3.2.
Lemma 4.1. Letm >0,m1, m2∈Rwithm1+m2=−m,ν, r∈N0withν≥n+1, ω0>0. Let b1∈Sm1,ν,r andb2∈Sm2,ν,r. For ε∈(0,1)define
Kε(x, y, µ) :=
Z
Rn
eiξ·yχε(ξ, y)b(x, y, ξ, µ)dξ with b(x, y, ξ, µ) := R1
0 b1(x−τ y, ξ, µ)b2(x−y, ξ, µ)dτ. Then the assertions of Lemma 3.2 hold with |b|(ν,r)−m being replaced by |b1|(ν,r)m1 |b2|(ν,r)m2 .
Proof. In the proof of Lemma 3.2, we replace inequality (4) by (6) kDξβ[b(x, y, µξ, µ)]kL(E)≤µ|β||b1|(ν,r)m
1 |b2|(ν,r)m
2 hµξ, µi−m−|β|. This can be seen by
kDβξ[b(x, y, µξ, µ)]kL(E)
≤µ|β| X
β0≤β
cββ0
Z 1
0
(Dξβ0b1)(x−τ y, µξ, µ)(Dξβ−β0b2)(x−y, µξ, µ)dτ L(E)
≤µ|β| X
β0≤β
cββ0|b1|(ν,r)m1 hµξ, µim1−|β0||b2|(ν,r)m2 hµξ, µim2−(|β|−|β0|)
≤µ|β||b1|(ν,r)m
1 |b2|(ν,r)m
2 hµξ, µi−m−|β|.
With (6) instead of (4) we can follow the proof of Lemma 3.2 almost literally.
Throughout the remainder of this section, we fix k ∈N0, m > 0, 1≤ p <∞, r, ν ∈N0 withr≥ρn,m+k+ 1 andν ≥n+ 2. We will consider theWpk(Rn, E)- realizationAk of a(x, D, µ), i.e. the restriction of a(x, D, µ) to Dmax(Ak) :={u∈ Wpk(Rn, E) :a(x, D, µ)u∈Wpk(Rn, E)}.
Theorem 4.2. Let a ∈ Sm,ν,r be parameter-elliptic, i.e., a ∈ Eκ,ωm,ν,r for some κ > 0 and ω ≥ 0. Then there exist constants M1 > 0 and µ1 >0 such that the Wpk-realizationAk of a(x, D, µ)is invertible for µ≥µ1 and
kA−1k kL(Wk
p(Rn,E))≤ M1
1 +µm (µ≥µ1).
Proof. Letω0> ω, µ≥2ω0, andu∈S(Rn, E). From Lemma 2.3 we know that a(x, D, µ)a#(y, D, µ)u= [a(x, ξ, µ)a#(y, ξ, µ)](D)u= I+p(x, y, D, µ)
u with the double symbol
p(x, y, ξ, µ) := a(x, ξ, µ)−a(y, ξ, µ)
a(y, ξ, µ)−1. (i) Letx∈Rn. By a Taylor expansion we obtain
p(x, y, D, µ)u (x)
= Os− Z Z
R2n
eiξ·y a(x, ξ, µ)−a(x−y, ξ, µ)
a(x−y, ξ, µ)−1u(x−y)d¯(ξ, y)
= Os− Z Z
R2n
eiξ·y
n
X
j=1
yj Z 1
0
(∂xja)(x−τ y, ξ, µ)a(x−y, ξ, µ)−1dτ u(x−y)d¯(ξ, y)
=
n
X
j=1
Os− Z Z
R2n
eiξ·y Z 1
0
Dξj
(∂xja)(x−τ y, ξ, µ)a(x−y, ξ, µ)−1
dτ u(x−y)d¯(ξ, y)
=
n
X
j=1 ε&0lim
Z
Rn
Kε(j,1)(x, y, µ) +Kε(j,2)(x, y, µ)
u(x−y)d¯y with
Kε(j,m)(x, y, µ) :=
Z
Rn
eiξ·yχε(ξ, y)b(j,m)(x, y, ξ, µ)d¯ξ,
b(j,1)(x, y, ξ, µ) :=
Z 1
0
(Dξj∂xja)(x−τ y, ξ, µ)a(x−y, ξ, µ)−1dτ, b(j,2)(x, y, ξ, µ) :=
Z 1
0
(∂xja)(x−τ y, ξ, µ)Dξj[a(x−y, ξ, µ)−1]dτ.
We haveDξj∂xja∈Sm−1,r−1,ν−1 and Dξja−1 ∈S−m−1,r,ν−1. By Lemma 4.1 we obtain forj= 1, . . . , nandm= 1,2 the existence of a strongly measurable function K(j,m) withKε(j,m)→K(j,m)(ε&0) pointwise and
kK(j,m)(x,·, µ)kL1(Rn,L(E)) ≤Cµ−1|a|(ν,r)m |a#|(ν,r)−m forµ≥2ω0. Therefore,
kp(x, y, D, µ)ukLp(Rn,E)≤
n
X
j=1 2
X
m=1
k[K(j,m)(x,·, µ)∗u]kLp(Rn,E)
≤Cµ−1|a|(ν,r)m |a#|(ν,r)−mkukLp(Rn,E).
(ii) Now letβ ∈Nn0 with|β| ≤k. In the same way as in the proof of part (ii) of Theorem 3.3, we write
(∂βxp(x, y, D, µ)u)(x)
= X
β0≤β
cββ0Os− Z Z
R2n
eiξ·y(∂xβ0p)(x, y, ξ, µ)(∂xβ−β0u)(x−y)d¯(ξ, y).
Replacing in (i)pby∂βx0panduby∂xβ−β0u, we obtain k∂xβ[p(x, y, D, µ)u]kLp(Rn,E)≤C X
β0≤β
k(∂xβ0p)(x, y, D, µ)(∂xβ−β0u)kLp(Rn,E)
≤C X
β0≤β
µ−1|a|(ν,r)m |a#|(ν,r)−m X
|γ|≤k
k∂xγukLp(Rn,E)
and therefore kp(x, y, D, µ)ukWk
p(Rn,E) ≤Cµ−1|a|(ν,r)m |a#|(ν,r)−mkukWk
p(Rn,E) forµ≥ 2ω0 with a constant C independent of a and µ. Taking µ ≥ µ1 with µ1 suf- ficiently large, we obtain kp(x, y, D, µ)kL(Wpk(Rn,E)) ≤ 12. By a Neumann series argument, we see thatI+p(x, y, D, µ)∈L(Wpk(Rn, E)) is invertible and therefore a(x, D, µ)a#(y, D, µ)(I+p(x, y, D, µ))−1= idWk
p(Rn,E)forµ≥µ1. This shows that Ak is surjective.
(iii) In the same way, we see thata#(x, D, µ)a(y, D, µ) =I+ ˜p(x, y, D, µ) with
˜
p(x, y, ξ, µ) := (a(x, ξ, µ)−1−a(y, ξ, µ)−1)a(y, ξ, µ), and that kp(x, y, D, µ)k˜ L(Wk
p(Rn,E)) ≤ 12 for sufficiently large µ. Therefore, (1 +
˜
p(x, y, D, µ))−1a#(x, D, µ)a(y, D, µ) = idDmax(Ak). This shows thatAk is injektive, and we obtain the invertibility ofAk. Moreover, we have for sufficiently largeµ
kA−1k kL(Wk
p(Rn,E))≤ ka#(y, D, µ)kL(Wp
k(Rn,E))k(I+p(x, y, D, µ))−1kL(Wk p(Rn,E))
≤ 2M0
1 +µm
forµ≥µ1 withµ1sufficiently large, whereM0 does not depend onµ.
In the following, we set Σϑ,R:={λ∈C:|λ| ≥R,|argλ| ≤ϑ}forϑ∈[0, π) and R >0.
Corollary 4.3. Let a∈Sm,ν,r(Rnx×Rnξ, L(E))be parabolic with constants κ >0 and ω ≥ 0, and let Ak denote the Wpk(Rn, E)-realization of a(x, D). Then there are constantsM2>0,ϑ∈(π2, π)andR >0 such that for the resolvent setρ(−Ak) of −Ak we have Σϑ,R⊂ρ(−Ak),
(7) k(λ+Ak)−1kL(E)≤ M2
1 +|λ| (λ∈Σϑ,R).
Therefore, −Ak:Wpk(Rn, E)⊃ D(Ak)→ Wpk(Rn, E) generates an analytic semi- group onWpk(Rn, E).
Proof. By Definition 2.4 b), the symbola(x, ξ)+µmeiθis parameter-elliptic for allθ withθ∈Σπ
2,R. By a standard continuity argument, we see that the set of anglesθ where parameter-ellipticity holds is open. Therefore, we may assume that the above symbol is parameter-elliptic for allθ ∈Σϑ,R with ϑ > π2. Now the invertibility of λ+Ak,λ=µmeiθ and the estimate (7) follows directly from Theorem 4.2.
Remark 4.4. The proof of Theorem 4.2 shows that the constants µ1 and M1 in Theorem 4.2 can be chosen independently of a for alla∈ A where A ⊂Sm,ν,r is bounded andA ⊂Em,ν,rκ,ω for fixedκ >0andω≥0. In the same way, the constants ϑ, M2, and R in Corollary 4.3 can be chosen independently of a for all a ∈ A0 where A0 ⊂Sm,ν,r(Rnx×Rξn, L(E)) is bounded and where all a∈ A0 are parabolic with the same constantsκ >0 andω≥0.
5. An application
Let T > 0, J := [0, T] a closed interval in R and t ∈ J. In the following t in a(t,·)∈Sm,ν,r R2n, L(E)
and ina(t, x, D) denotes only a parameter. Moreover we will consider in this section a family A := {a(t,·) :t∈J} ⊂ Em,ν,rκ,ω for fixed κ >0 andω≥0, such that
J 3t7−→a(t,·)∈Sm,ν,r R2n, L(E)
is a H¨older continuous function relative to the topology in the space of the symbols.
We will use the results of the previous sections to study the existence and uniqueness of solutions for the Cauchy problem
(8)
∂tu+Ak(t)u=f(t), t∈J\ {0}, u(0) =u0,
inWpk(Rn, E). There,Ak(t),t∈J, is theWpk(Rn, E)-realization ofa(t, x, D) and f : [0, T]→Wpk(Rn, E) is a given H¨older continuous function. A function
u∈C1(]0, T], E)∩C([0, T], E)
is calledclassical solutionof (8) in [0, T], ifu(t)∈D(Ak(t)) andu0(t)+Ak(t)u(t) = f(t) for allt∈]0, T], andu(0) =u0. More precisely, we will use Corollary 4.3, The- orem 2.5.1 of Chapter IV in [Am95] and Lemma 5.2 to obtain results of existence and uniqueness of solutions of the Cauchy problem (8) inWpk(Rn, E) if 1≤p <∞.
In the following ˚Bp,qs (Rn, E),s∈R,p, q∈[1,∞], denote theE-valued homogeneous Besov spaces of ordersand parameterspandq(see [Am97] for their definition and
properties).
Lemma 5.1. Let m ∈R, ν, r ∈N with ν ≥2n+ 1 and a∈ Sm,ν,r R2n, L(E) . Then
a(x, D) : ˚Bs+mp,q (Rn, E)→B˚p,qs (Rn, E)
is a bounded and linear map for all p, q ∈ [1,∞] and all s ∈ R with 0 ≤ s < r.
Moreover it holds the following estimate:
ka(x, D)kL(B˚s+mp,q (Rn,E),B˚p,qs (Rn,E))≤c|a|(2n+1,r)m . Proof. See [Ki01], Theorem 3.5.
Lemma 5.2. Let m ∈ R, ν ≥ 2n+ 1, r ≥ ρn,m+k+ 1, p ∈ [1,∞) and A :=
{a(t,·) ; t∈J} ⊂Sm,ν,r with Ak(t)the Wpk(Rn, E)-realization ofa(t, x, D). Then (9) B˚k+mp,1 (Rn, E),→D(Ak(t)),→d Wpk(Rn, E).
Proof. It’s clear thatD(Ak(t)),→Wpk(Rn, E) (D(Ak(t)) endowed with the graph norm) and it’s also known that
(10) S(Rn, E),→B˚p,1k+m(Rn, E),→B˚p,1k (Rn, E),→Wpk(Rn, E), 1≤p <∞.
From this and Lemma 5.1 it follows that u, a(t, x, D)u ∈ Wpk(Rn, E) whenever u∈S(Rn, E). Therefore S(Rn, E)⊂D(Ak(t)). But S(Rn, E),→d Wpk(Rn, E) for 1≤p <∞and then we haveD(Ak(t)),→d Wpk(Rn, E).
On the other hand, (10) and Lemma 5.1 imply that ˚Bp,1k+m(Rn, E)⊂D(Ak(t)) and kukD(Ak(t))=kukWk
p(Rn,E)+ka(t, x, D)ukWk p(Rn,E)
≤c1kukB˚p,1k+m(Rn,E)+c2ka(t, x, D)ukB˚kp,1(Rn,E)
≤c1kukB˚p,1k+m(Rn,E)+c3kukB˚p,1k+m(Rn,E)
=ckukB˚k+mp,1 (Rn,E). So we have ˚Bk+mp,1 (Rn, E),→D(Ak(t)).
Theorem 5.3. Letk∈N0,α, m∈R+ with0< α <1,p∈[1,∞),ν ≥2n+ 1and r ≥ρn,m+k+ 1. Furthermore for fixed κ >0 andω ≥0, letA :={a(t,·) ;t ∈ J} ⊂Eκ,ωm,ν,r be a family of parabolic symbols, such that
(11) t7−→a(t,·)∈Cα J, Sm,ν,r(R2n, L(E)) and such that there is a constantM with
(12) k(λ+Ak(t))−1kL( ˚Bk
p,∞(Rn,E),B˚p,∞k+m(Rn,E))≤M,
for all t ∈ J and λ in the sector Σϑ,R of Corollary 4.3. There Ak(t) is the Wpk(Rn, E)-realization ofa(t, x, D).
If u0 ∈ Wpk(Rn, E) and f ∈ Cσ J, Wpk(Rn, E)
for some σ ∈ ]0,1[, then the Cauchy problem (8) has a unique solution
u∈C J, Wpk(Rn, E)
∩C1 J\ {0}, Wpk(Rn, E) .
Proof. First we claim that D(Ak(t)) ,→ B˚p,∞k+m(Rn, E). In fact, if u ∈ D(Ak(t)), then u, a(t, x, D)u ∈ Wpk(Rn, E) ⊂ B˚p,∞k (Rn, E). For λ0 ∈ Σϑ,R fixed and v :=
(λ0+Ak(t))u∈B˚p,∞k (Rn, E), (12) implies thatu∈B˚p,∞k+m(Rn, E) and kukB˚p,∞k+m(Rn,E)≤MkvkB˚p,∞k (Rn,E)
≤ckvk˜ Wk p(Rn,E)
≤c˜
|λ0|kukWk
p(Rn,E)+kAk(t)ukWk p(Rn,E)
≤ckukD(Ak(t)),
i.e. D(Ak(t)),→B˚p,∞k+m(Rn, E). From this, (9), Corollary 4.3 and Theorem 2.5.1 of Chapter IV in [Am95], we obtain the desired result in similar way to the proof of Theorem 4.3 in [BDH12].
Appendix A. Appendix: Oscillatory integrals
For the definition of pseudo-differential operators we needed the theory of (vector- valued) oscillatory integrals. Therefore, we summarize below some definitions and results which can be found in [Ba09] and [Ku81].
Definition A.1. Let a:R2n→E be a strongly measurable function. We say that ais integrable in the oscillatory sense if for each χ∈S(R2n)with χ(0,0) = 1 the limit
Os− Z Z
R2n
e−iξ·ya(ξ, y)d¯(ξ, y) := lim
ε&0
Z Z
R2n
e−iξ·yχ(εξ, εy)a(ξ, y)d¯(ξ, y) exists and does not depend on the choice of χ.
Definition A.2. Let m ∈ R, τ ∈ [0,∞), ν, ρ ∈ N0 ∪ {∞}. Then the space A(m,ν,ρ)τ (Rnξ ×Rny, E) of amplitude functions consists of all a: Rnξ ×Rny → E for which all derivatives∂ξα∂βya with|α| ≤ν,|β| ≤ρare continuous and for which all norms|a|m,`,`0 with `, `0∈N0,`≤ν,`0≤ρare finite. Here
|a|m,`,`0 := max
|α|≤`max
|β|≤`0 sup
ξ∈Rn
sup
y∈Rn
hξi−m−|β|hyi−τk∂ξα∂yβa(ξ, y)kE.
Lemma A.3. Forχ∈S(Rn)withχ(0) = 1the following assertions hold.
(i) χ(εx)→1 (ε&0) uniformly on all compact subsets ofRn, (ii) ∂xαχ(εx)→0 (ε&0) uniformly on Rn for allα∈Nn0\ {0},
(iii) for all α∈Nn0 there existscα>0 such that for all ε∈(0,1) we have
|∂xαχ(εx)| ≤cαεσhxi−(|α|−σ) (σ∈[0,|α|], x∈Rn).
Proof. See [Ku81], Lemma 6.3.
Lemma A.4. Let m∈R,τ ∈[0,∞),`, `0 ∈Nwith n+τ <2`, m+n <2`0. Let furthermoreν, ρ∈N0∪{∞}with2`≤ν,2`0≤ρ. Then fora∈ A(m,ν,ρ)τ (Rnξ×Rny, E) the oscillatory integral Os−
Z Z
R2n
e−iξ·ya(ξ, y)d¯(ξ, y)exists, and
Os−
Z Z
R2n
e−iξ·ya(ξ, y)d¯(ξ, y)
E≤c|a|m,2`,2`0
holds with a constantc not depending ona.
Proof. See [Ba09], Satz 1.3.1.
Lemma A.5. With m, τ, ν, ρ, `, and `0 as in Lemma A.4, let N ∈ N0 and let a:Rnx×Rnξ ×Rny →E be a function with ∂xγa(x,·,·)∈A(m,ν,ρ)τ (Rnξ,Rny, E) for all x∈Rnand|γ| ≤N. Suppose further that for everyx∈Rnthere is a neighbourhood Uxofxsuch that the set{(∂xγa)(x0,·,·) :x0 ∈Ux}is bounded inA(m,ν,ρ)τ (Rnξ,Rny, E).
Then for all|γ| ≤N we have
∂xγh Os−
Z Z
R2n
e−iξ·ya(x, ξ, y)d¯(ξ, y)i
= Os− Z Z
R2n
e−iξ·y∂xγa(x, ξ, y)d¯(ξ, y).
Proof. See [Ba09], Satz 1.3.3.
References
[Am95] H. Amann: Linear and quasilinear parabolic problems, Vol. I. Birkh¨auser Verlag, Basel (1995).
[Am97] H. Amann: Operator-valued fourier multipliers, vector-valued besov spaces, and appli- cations.Math. Nachr.186(1997), 5-56.
[Am00] H. Amann Coagulation-Fragmentation Processes, Arch. Rot. Mech. Anal. 151, 339-366 (2000).
[Am01] H. Amann, Elliptic Operators with Infinite Dimensional State Spaces, J. Evo. Equ. 1, 143-188 (2001).
[Ba09] B. Barraza Mart´ınez, Pseudodifferentialoperatoren mit nichtregul¨aren banachraumwer- tigen Symbolen, Dissertation, Johannes Gutenberg-Universit¨at Mainz, (2009).
[BDH12] B. Barraza Mart´ınez, R. Denk, J. Hern´andez Monz´on, Analytic semigroups of pseudo- differential operators on vector-valued Sobolev spaces, Preprint Konstanzer Schriften in Mathematik, Nr. 303, Mai 2012.
[DHP03] R. Denk, M. Hieber, J. Pr¨uss, R-Boundedness, Fourier multipliers and problems of elliptic and parabolic type, Mem. Amer. Math. Soc. 166 (2003).
[DK07] R. Denk, T. Krainer,R-boundedness, pseudodifferential operators, and maximal regu- larity for some classes of partial differential operators, Manuscripta Math. 124, 319-342 (2007).
[GW03] M. Girardi and L. Weis, Operator-Valued Fourier Multiplier Theorems on Besov Spaces, Math. Nachr. 251, 34-51 (2003).
[Ki01] Chr. Kiehn, Analytic Semigroup of Pseudodifferential Operators onLp(Rn, E), Preprint Reihe des Fachbereichs Mathematik Nr 10, Johannes Gutenberg-Universit¨at, (2001).
[Ki03] Chr. Kiehn, Analytic Semigroups of Pseudodifferential Operators on Vector-Valued Function Spaces, Shaker Verlag, Aachen (2003).
[Ku81] H. Kumano-go, Pseudo-Differential Operators, MIT Press, Cambridge, MA (1981).
[KW04] P. C. Kunstmann, L. Weis, Maximal Lp-regularity for parabolic equations, Fourier multiplier theorems andH∞-functional calculus, in Da Prato et al., editor, “Functional Analytic Methods for Evolution equations”, Lecture Notes in Math. 1855, Springer, 2004.
[NS11] T. Nau, J. Saal,R-sectoriality of cylindrical boundary value problems, in J. Escher et al., editor, “Parabolic Problems: The Herbert Amann Festschrift”, Progr. Nonlinear Differential Equations Appl. 80, Birkhuser/Springer, 479-506 (2011).
[NNH02] A. Noll, R. Haller, and H. Heck, Mikhlin’s Theorem for Operator-Valued Multipliers in nVariables, Math. Nach. 244, 110-130 (2002).
[PS06] P. Portal and ˇZ. ˇStrkalj, Pseudodifferential operators on Bochner Spaces and an appli- cation, Math. Z., 253, 805-819 (2006).
[We01] L. Weis, Operator-Valued Fourier Multiplier Theorems and Maximal Lp-Regularity, Math. Ann. 319, 735-758 (2001).
Universidad del Norte, Departamento de Matem´aticas, Km 5 Via a Puerto Colombia, Barranquilla, Colombia
E-mail address:bbarraza@uninorte.edu.co
Universit¨at Konstanz, Fachbereich Mathematik, 78457 Konstanz, Germany E-mail address:robert.denk@uni-konstanz.de
Universidad del Norte, Departamento de Matem´aticas, Km 5 Via a Puerto Colombia, Barranquilla, Colombia
E-mail address:jahernan@uninorte.edu.co