ANNALES
DE LA FACULTÉ DES SCIENCES
Mathématiques
JAMELBENAMEUR, RIDHASELMI
Study of Anisotropic MHD system in Anisotropic Sobolev spaces
Tome XVII, no1 (2008), p. 1-22.
<http://afst.cedram.org/item?id=AFST_2008_6_17_1_1_0>
© Université Paul Sabatier, Toulouse, 2008, tous droits réservés.
L’accès aux articles de la revue « Annales de la faculté des sci- ences de Toulouse Mathématiques » (http://afst.cedram.org/), implique l’accord avec les conditions générales d’utilisation (http://afst.cedram.
org/legal/). Toute reproduction en tout ou partie cet article sous quelque forme que ce soit pour tout usage autre que l’utilisation à fin strictement personnelle du copiste est constitutive d’une infraction pénale. Toute copie ou impression de ce fichier doit contenir la présente mention de copyright.
cedram
Article mis en ligne dans le cadre du
Annales de la Facult´e des Sciences de Toulouse Vol. XVII, n◦1, 2008 pp. 1–22
Study of Anisotropic MHD system in Anisotropic Sobolev spaces
(∗)Jamel Ben Ameur(1), Ridha Selmi(2)
ABSTRACT. — Three-dimensional anisotropic magneto-hydrodynamical system is investigated in the whole spaceR3. Existence and uniqueness results are proved in the anisotropic Sobolev space H0,s for s > 1/2.
Asymptotic behavior of the solution when the Rossby number goes to zero is studied. The proofs, where the incompressibility condition is crucial, use the energy method, an appropriate dyadic decomposition of the frequency space, product laws in anisotropic Sobolev spaces and Strichartz-type estimates.
R´ESUM ´E.— On ´etudie un syst`eme magneto-hydro-dynamique tridimen- sionnel dans le cas de l’espace entier R3. On d´emontre l’existence et l’unicit´e de la solution pour des donn´ees initiales dans les espaces de Sobolev anisotropes ;H0,s,s >1/2. On ´etudie le comportement asymp- totique de la solution lorsque le nombre de Rossby tends vers z´ero. Les preuves se basent essentiellement sur des m´ethodes d’energie, une d´ecompo- sition ad´equate de l’espace de fr´equences, les lois produit dans les espaces de Sobolev anisotropes et une estimation de type Stichartz. La condition d’incompressibilit´e joue un rˆole crucial dans les d´emonstrations.
(∗) Re¸cu le 17 janvier 2006, accept´e le 27 septembre 2006
(1) Facult´e des Sciences de Bizerte, Tunisie.
jamel.BenAmeur@fsb.rnu.tn
(2) Institut Sup´erieur d’Informatique, l’Ariana 2080, Tunisie.
Ridha.Selmi@isi.rnu.tn
1. Introduction
The purpose of this paper is to study the incompressibleM HD model with anisotropic diffusion in the limit of small Rossby number, namely the following system
(M HDνεh)
∂tu−νh∆hu+u· ∇u−b· ∇b+1
εcurlb×e3
+1
εu×e3=−∇p inR+×R3
∂tb−ηh∆hb+u· ∇b−b· ∇u +1
εcurl (u×e3) = 0 in R+×R3 divu= 0 in R+×R3
divb= 0 in R+×R3 (u, b)|t=0= (u0, b0),
where the velocity fieldu,the induced magnetic perturbationband the pres- surepare unknown functions of timet and space variablex= (x1, x2, x3), e3is the third vector of the cartesian coordinate system,νhis the anisotropic dynamic viscosity,ηhis the anisotropic magnetic diffusivity, ∆hdenotes the horizontal Laplace operator defined by ∆h =∂12+∂22 andε is the Rossby number which is the ratio between the fluid’s typical velocity to the earth rotation velocity around the axise3.
The above system is a simplified version of a general one given in [4] to modelise the magneto-hydrodynamical process in the earth’s core which is believed to support a self-excited dynamo process generating the earth’s magnetic field.
Our first motivation follows from Taylor Proudman Theorem [10], that the three-dimensional fluid is inclined to behave as a two-dimensional one. The second motivation follows from the fact that large turbulent eddies move preferably in horizontal layers and do rarely move vertically, which makes vertical turbulent diffusion negligible compared to the horizontal one (see [3] and references therein). So, it makes sense to consider an anisotropic diffusive process. Without loss of generality,−e3is supposed to be the time independent component of the earth magnetic fieldB defined in [4] by
B=−e3+θb,
where θ denotes the Reynolds’ number. According to [5], the magnetic Prandtl number Pm defined as the ratio of the dynamic viscosity by the
magnetic viscosity is equal to 10−6in the case of the earth’s core. However, to simplify the mathematical computation we will takeνh=ηh. Moreover, we remark that
curl (u×e3) =∂3u and thanks to the divergence free condition,
curlb×e3=∂3b− ∇b3.
Since the proofs are based in taking the L2-scaler product either in (u, b) or in their associated dyadic blocs (∆qvu,∆qvb) (see section 2 for notation), terms like ∇f, for any function f, have no contribution in the L2-energy estimate. Thus, we can reformulate the (M HDνεh) system to obtain the following noted also (M HDενh)
(M HDενh)
∂tu−νh∆hu+u· ∇u−b· ∇b+1 ε∂3b+1
εu×e3
=−∇P inR+×R3
∂tb−νh∆hb+u· ∇b−b· ∇u+1
ε∂3u= 0 in R+×R3 divu= 0 in R+×R3
divb= 0 in R+×R3 (u, b)|t=0= (u0, b0), where
P =p+b3.
This system can be written in the following abstract form
(Sε)
∂tU+Q(U, U) +a2,h(D)U+Lε(U) = t(−∇p,0) in R+×R3 divu= 0 in R+×R3
divb= 0 in R+×R3 U|t=0=U0
where
U = u
b
, the quadratic term Qis defined by
Q(U, U) =
u· ∇u−b· ∇b u· ∇b−b· ∇u , the viscous term is
a2,h(D)U =−νh∆hU and the linear perturbationLε is given by
Lε(U) =1
εL(U) := 1 ε
∂3b+u×e3
∂3u .
We note thatLε is a skew-symmetric linear operator. This skew-symmetry is an important property for the existence results since the perturbation disappears in the energy estimate.
In the case of anisotropic diffusion process, existence, uniqueness and asymp- totic behavior of some magneto hydrodynamical systems have been studied by several authors ([1], [2], [11]). According to our knowledge, no paper is concerned by the anisotropic case. Nevertheless, anisotropic Navier-Stokes equation was studied in [3], where it was proved the local existence for ar- bitrary initial data and global existence for small initial data inH0,s(R3) for s >1/2 and uniqueness for s > 3/2. In [8], the author closed the gap between existence and uniqueness and proved that uniqueness holds when existence does; that is fors >1/2.
In this paper, dealing with all coupled nonlinearities and singular per- turbation, we establish local well-posedness (existence and uniqueness) for arbitrary initial data, global existence for small initial data inH0,s(R3) for s >1/2 and we investigate the asymptotic behavior of the solution as the Rossby number εtends to zero.
Precisely, we establish global existence for small initial data and local existence of solution on uniform time for arbitrary initial data. Namely, solution given by the following theorems.
Theorem 1.1. — Let s > 1/2 be a real number and U0 = (u0, b0) ∈ H0,s(R3) such that divu0 = 0 and divb0 = 0. There exists a positif time T such that for all ε > 0, there exists a unique solution Uε of (M HDενh) satisfying Uε ∈ L∞T (H0,s(R3)) and ∇hUε ∈ L2T(H0,s(R3)). Moreover, Uε satisfies the following energy estimate
Uε(t, .)2H0,s(R3)+νh
t 0
∇hUε(τ, .)2H0,s(R3)dτ U02H0,s. (1.1) Furthermore, there exists a constantcsuch that ifU0H0,s(R3)cνh, then the solution is global.
To prove Theorem 1.1, we will rearrange the nonlinear terms in a suit- able way to apply Lemma 2.2. The proof, where the divergence free con- dition plies a crucial role, uses the energy method, an appropriate dyadic decomposition of the frequency space, compactness argument and classical analysis.
The following theorem gives an uniqueness result.
Theorem 1.2. — LetUεandVεbe two solutions of(M HDενh)on[0, T] belonging to
L∞T (H0,s(R3))∩L2T(H1,s(R3)),
wheres >1/2. IfUε andVε have the same initial data, thenUε=Vε. To prove Theorem 1.2, we will estimate theH0,−1/2 norm of the differ- enceUε − Vεinstead of theH0,1/2one. This will be done via Lemma 3.3 and standard arguments.
Once the existence and the uniqueness results are proved we turn to the asymptotic behavior of the solution when the initial data belongs toH0,s(R3), s >2. This will be the aim of the following theorem.
Theorem 1.3. — Lets >2,U0= (u0, b0)∈H0,s(R3)such that divu0= 0 and divb0= 0. Let (Uε)be the family of solution given by Theorem 1.1, then
Uε −−−→
ε−→0 0 in L2T(C−3/2(R3)). (1.2) That is, for χ inD(R),URε =χ(|∇h|/R))Uε andU˜Rε =Uε−URε, it holds
lim sup
ε→0 URεL4T(L∞) −−−−−→
R−→+∞0 (1.3)
lim sup
ε→0 U˜RεL2T(H1−η,0∩H0,s) −−−−−→
R−→+∞ 0, ∀s < sandη >0. (1.4) To prove Theorem 1.3, we will use a Strichartz type inequality and stan- dard Fourier analysis results. In fact, dispersive effects are of great impor- tance in the study of nonlinear partial differential equations, since they yield decay estimates on waves when the domain is the whole space R3, chosen here essentially for mathematical convenience.
2. Proof of existence results
2.1. Anisotropic Littlewood-Paley theory
To introduce the anisotropic Sobolev spaces, we use an anisotropic dyadic decomposition of the frequency space and we introduce, for any functiona, the following operators of localization in Fourier space :
∆hja = F−1(ϕ(2−j|ξh|)F(a)) for j∈Z,
∆vqa = F−1(ϕ(2−q|ξ3|)F(a)) for q∈N,
∆v−1a = F−1(ϑ(|ξ3|)F(a))
and
∆vqa=0 for q−2.
The functionsϕandϑrepresent a dyadic partition of unity inR. This means that these functions are regular non-negative and satisfy
supp (ϑ)⊂B(0,4/3), supp (ϕ)⊂ C(0,3/4,8/3) and for allt∈R,
ϑ(t) +
q0
ϕ(2−qt) = 1.
Here,
C(0,3/4,8/3) ={ξ, ξ∈R3, 3/4|ξ|8/3}. We define, also, for any functionu, the operatorsSqvuandSjhuby
Sqvu=
qq−1
∆vqu
and
Sjhu=
jj−1
∆hju.
By this way, we note that we consider an homogeneous decomposition in the horizontal variable and an inhomogeneous one in the vertical.
We define the corresponding Sobolev spaces by the following.
Definition 2.1. — Lets ands be two real numbers and ua tempered distribution. Let
uHs,s =
j,q
22(js+qs)∆hj∆vqu2L2 1/2
.
The space Hs,s(R3)is the closure of D(R3)for the above semi-norm.
The interest of the dyadic decomposition lies in the fact that any vertical derivative of a function localized in vertical frequencies of size 2q acts as a multiplication by 2q. We refer to [8] for a precise construction of anisotropic Sobolev spaces.
2.2. Global existence result for small initial data
In this section, the following lemma, a detailed proof of which is given in [3], will be useful to estimate the nonlinear part.
Lemma 2.2. — For any real numbers s0>1/2 andss0a constantC exists such that for any divergence free vectors fieldsaandb, we have
|(∆vq(a· ∇b)|∆vqb)L2| Cdq2−2qsbH1/2,s
aH1/2,s0∇hbH0,s
+aH1/2,s∇hbH0,s0 + ∇haH0,s0bH1/2,s+∇haH0,sbH1/2,s
where the positive numbersdq verify
q∈Z
dq = 1.
To prove Theorem 1.1, apply the operator ∆vq to (Sε), denote Uq = (∆vquε,∆vqbε) and use an L2 energy estimate to obtain
1 2
d
dtUqε(t)L2+νh∇hUqε(t)L2 ∆vq(uε· ∇uε)|uεq
L2 + ∆vq(uε· ∇bε)|bεq
L2 + ∆vq(bε· ∇bε)|uεq
L2
+
∆vq(bε· ∇uε)|bεq
L2.
In order to estimate the left hand side of the above inequality, rearrange the nonlinearity as
(bε· ∇bε|uε)L2+ (bε· ∇uε|bε)L2 =
bε· ∇(uε+bε)|(uε+bε)
L2
− (bε· ∇bε|bε)L2−(bε· ∇uε|uε)L2. Note that the Hs,s norm of any vector field is the Eucldiean norm of the Hs,s norms of its components. That is why
uε,bεUε,
∇huε,∇hbε∇hUε and so on.
Use Lemma 2.2, in the cases=s0, to deduce that 1
2 d
dtUqε2L2+νh∇hUqε2L2Cdq2−2qs0Uε2H1/2,s0∇hUεH1/2,s0. (2.1) Multiply by 22qs0 and take the sum overqto obtain
1 2
d
dtUε2H0,s0 +νh∇hUε2H0,s0 Uε2H1/2,s0∇hUεH1/2,s0.
An interpolation inequality yields
Uε2H1/2,s0 UεH0,s0∇hUεH0,s0.
So, one concludes that ifU0H0,s0 is sufficiently small thenUε(t)2H0,s0 is a decreasing function. More precisely, if
U0H0,s0 νh
2C then for any positive timet, one gets
Uε(t)2H0,s0+νh
t 0
∇hUε(τ)2H0,s0dτ U02H0,s0. (2.2) Finally, a standard compactness argument gives the global existence result.
2.3. Local existence result for arbitrary initial data
Now, we turn to the local existence result of (M HDενh). Decompose the frequency space in order to investigate the high vertical frequencies then the low vertical-high horizontal ones.
For the high vertical frequencies term, one multiplies inequality (2.1) by 22qs0 then takes the sum for q greater than n−1 in order to obtain after integration in time
(Id−Snv)UεL∞T(H0,s0) + 2νh∇h(Id−Snv)Uε2L2T(H0,s0)
C(Id−Snv)Uε2H0,s0 (2.3)
+ C
T 0
Uε(t)2H1/2,s0∇hUε(t)H0,s0dt.
To study the low vertical-high horizontal frequency, letmandntow cutoff integers such that mnand defineUm,nε by
Um,nε =(Id−Smh)SnvUε.
Denote, respectively, byQh andQv the horizontal and the vertical part of the nonlinear term defined by
Qh(U, U) =
uh· ∇hu−bh· ∇hb uh· ∇hb−bh· ∇hu
and
Qv(U, U) =
u3·∂3u−b3·∂3b u3·∂3b−b3∂3u·
.
By standardL2 energy estimate, one has 1
2 d
dtUm,nε 2L2+νh∇hUm,nε 2L2 =
(Id−Smh)Snv(Qh(Uε, Uε))|Um,nε
L2
+
(Id−Smh)Snv(Qv(Uε, Uε))|Um,nε
L2, where
(Id−Smh)Snv(Qh(Uε, Uε))|Um,nε
L2 =
(Id−Smh)Snv(uεh· ∇huε)|uεm,n
L2
−
(Id−Smh)Snv(bεh· ∇hbε)|uεm,n
L2
+
(Id−Smh)Snv(uεh· ∇hbε)|bεm,n
L2
−
(Id−Smh)Snv(bεh· ∇huε)|bεm,n
L2
and
(Id−Smh)Snv(Qv(Uε, Uε))|Um,nε
L2 =
(Id−Smh)Snv(uε3·∂3uε)|uεm,n
L2
−
(Id−Smh)Snv(bε3·∂3bε)|uεm,n
L2
+
(Id−Smh)Snv(uε3·∂3bε)|bεm,n
L2
−
(Id−Smh)Snv(bε3·∂3uε)|bεm,n
L2. By product law in Sobolev spaces, one infers that
(Id−Shm)Snv(Qh(Uε, Uε)(., x3))H−1/2(R2xh)
CUhε(., x3)H1/2(R2xh)∇hUε(., x3)L2(R2xh). Sinces0>1/2, one obtains
(Id−Shm)Snv(Qh(Uε, Uε))L2
Rx3;H−1/2(R2)CUhεH1/2,s0∇hUεH0,s0. By Cauchy-Schwarz inequality, one has
(Id−Smh)Snv(Qh(Uε, Uε))|Um,nε
L2(R3)CUε2H1/2,s0∇hUεH0,s0. To estimateQv(Uε, Uε), use the divergence free condition; that is
∂3uε3=−divhuεh and
∂3bε3=−divhbεh.
These equalities allow one to reformulate the vertical nonlinearity as Qv(Uε, Uε) =Qv1(Uε, Uε) +Qv2(Uε, Uε),
where
Qv1(Uε, Uε) =
∂3(uε3uε)−∂3(bε3bε)
∂3(uε3bε)−∂3(bε3uε)
and
Qv2(Uε, Uε) =
uεdivhuεh−bεdivhbεh bεdivhuεh−uεdivhbεh
. The nonlinear term,
(Id−Smv)Snv(Qv2(Uε, Uε))|Um,nε
L2 can be estimated exactly as the term
(Id−Smv)Svn(Qh(Uε, Uε))|Um,nε
L2. To study the other term, one observes that
Id−Smh = 2−m(Id−Shm)(χ1(D)∂1+χ2(D)∂2),
where χ1 and χ2 are homogeneous functions of degree 0. Then, it is clear that, whenmn,
(Id−Smv)Snv(Qv1(Uε, Uε))|Um,nε
L2 2n−mU3εUεL2(R3)∇hUm,nε L2(R3)
CUε2H1/2,s0∇hUm,nε H0,s0
CUε2H1/2,s0∇hUεH0,s0. Finally, by integration in time, one has
Um,nε 2L∞T(H0,s0) + 2νh∇hUm,nε 2L2T(H0,s0)
CUm,nε (0)2H0,s0 (2.4)
+ C
T 0
Uε(t)2H1/2,s0∇hUε(t)H0,s0dt.
To conclude the proof of the local part, one begins by observing that since U0 belongs toL2(R3) the basicL2 energy estimate holds
Uε(t)2L2+ 2νh
t 0
∇hUε(τ)2L2dτ U0ε2L2. (2.5) Thus, it is obvious that
SnvSmhUε(t)2H0,s0+2νh
t 0
∇hSnvSmhUε(τ)2H0,s0dτ C22m+2ns0TU0ε2L2. Using an interpolation inequality, estimations (2.3) and (2.4) one infers that
(2νh)1/2Uε2L4
T(H1/2,s0) +2νh∇hUε2L2
T(H0,s0)C
(Id−Snv)U0ε2H0,s0
+Snv(Id−Smh)U0ε2H0,s0+ 22m+2ns0TU02L2
+ T
0
Uε(t)2H1/2,s0∇hUε(t)H0,s0dt .
It follows that νh1/2Uε2L4
T(H1/2,s0)+νh∇hUε2L2
T(H0,s0)C
(Id−Snv)U0ε2H0,s0+ Snv(Id−Smh)U0ε2H0,s0 + 22m+2ns0TU02L2+ C
νhUε(t)4L4
T(H1/2,s0) . Choose, in this order, an integer n, then an integer m greater than nand finally a strictly positive real numberT0such that
(Id−Snv)U0ε2H0,s0 η, Snv(Id−Smh)U0ε2H0,s0 η and
22m+2ns0T0U0ε2L2 η to obtain, for anyT T0,
νh1/2Uε2L4
T(H1/2,s0)+νh∇hUε2L2
T(H0,s0)Cη+ C νh
Uε(t)4L4
T(H1/2,s0). So, for any positive real numberη, such that
0< η < νh2 4C2 a positive real numberT1 exists, such that
νh1/2Uε2L4
T1(H1/2,s0)+νh∇hUε2L2T
1(H0,s0)η.
That proves Theorem 1.1.
3. Proof of uniqueness result
In this section we denote byxthe quantity (1 +|x|2)1/2and by Λ3
the operator defined by
Λ3=(1−∂32)1/2,
that is, the operator of multiplication byξ3in the frequency space. Clearly, Λ3is an isometry fromHs,s toHs,s−1 for all real numberssands.
The following lemmas will be useful in the sequel. The first one and the second one are proved in [7], the third one is proved in [8].
Lemma 3.1. — Lets, t, s, t∈Randα∈[0,1]. Iff ∈Hs,s∩Ht,t then f belongs to Hαs+(1−αt),αs+(1−αt) and
f gHαs+(1−αt),αs+(1−αt) fαHs,sg1H−t,tα.
Lemma 3.2. — Let s, t <1,s+t >0 and s, t <1/2, s+t >0. Iff belongs to Hs,s and g belongs toHt,t thenf g belongs to Hs+t−1,s+t−1/2 and there exists a constant C such that
f gHs+t−1,s+t−1/2 CfHs,sgHt,t.
Lemma 3.3. — Lets, t <1,s+t >0ands >1/2. Iff belongs toHs,s andg belongs toHs,−1/2thenf g belongs toHs+t−1,−1/2 and there exists a constant C such that
f gHs+t−1,−1/2CfHs,sgHt,−1/2.
The originality of Lemma 3.3, as said in [8], is to give a product law for the anisotropic Sobolev spaces where the regularities in the vertical direction are supercritical for one of the terms and subcritical for the other.
3.1. Proof of Theorem 1.2
Let Uε = t(uε, bε) and Vε = t(vε, cε) be two solutions of (M HDενh) satisfying Theorem 1.1 and having the same initial data. It is explained, in Remark 2, that this makes sense.
Denote byWε= t(wε, βε) the difference Wε=Uε−Vε. One has
∂tWε+νh∇hWε+Q(Wε, Wε+ 2Uε) +1 εL(Wε)
= t(−∇pε,0) in R+×R3 divwε = 0 in R+×R3 divβε= 0 in R+×R3.
(3.1)
Without loss of generality, assume that s < 1. Suppose that Uε and Vε belong toCT0(H0,r) for allr < s, a fact on which we will return in Remark 1. By this assumption, one is able to multiply equation (3.1) by Λ−31Wε then integrate on (ε, t)×R3, letεtends to zero and use the continuity in time ofWεH0,−1/2 to obtain
Wε2H0,−1/2 + 2νh
t 0
∇hWε2H0,−1/2dτ
= −2 t
0
Q(Wε, Wε+ 2Uε)Λ−31Wεdxdτ, (3.2) where
Q(Wε, Wε+ 2Uε) =
uε· ∇wε+wε· ∇vε−bε· ∇βε−βε· ∇cε uε· ∇βε+wε· ∇cε−bε· ∇wε−βε· ∇vε
.
Here, note that for every multi-indexsand sufficiently smooth functionU,
one has
∂αU Lε(∂αUε) = 0.
To estimate the left hand side of equation (3.2), one restricts itself to the two first terms, the same holds for the others since that
uε, bεUε and so on.
Following [8], decompose the nonlinearity as follows
(uε· ∇wε+wε· ∇vε)·Λ−31wεdx= 4 i=1
Li, where
L1 =
uεh· ∇hwε·Λ−31wεdx, L2 =
uε3·∂3wε·Λ−31wεdx, L3 =
wεh· ∇hvε·Λ−31wεdx and
L4=
w3ε·∂3vε·Λ−31wεdx.
To estimateL1, note that
|L1|uεh· ∇hwεH−1/2,−1/2Λ−31wεH1/2,1/2,
apply Lemma 3.3 to the first factor, use properties of Λ3 and Lemma 3.1 for the second to obtain
|L1|CUεH1/2,sWε1/2H0,−1/2Wε3/2H1,−1/2. The same holds forL3, just begin by
|L3|wεh· ∇hvεH−3/4,−1/2Λ−31wεH3/4,1/2
to obtain
|L3|CVεH1/2,sWε1/2H0,−1/2Wε3/2H1,−1/2. ForL4, remark that
|L4|wε3∂3vεH−1/2,(2s−3)/4Λ−31wεH1/2,(3−2s)/4,