Strong diamagnetism for general domains in R and ...helffer/petersbourg07.pdf · Strong...
Transcript of Strong diamagnetism for general domains in R and ...helffer/petersbourg07.pdf · Strong...
Strong diamagnetism for general
domains in R3 and applications
to superconductivity
Bernard HelfferMathematiques -
Univ Paris Sud- UMR CNRS 8628(After S. Fournais and B. Helffer)
Conference in St-Petersburg
June 2007
– Typeset by FoilTEX –
Main goals
We consider the Neumann Laplacian with constantmagnetic field on a regular domain. Let B be thestrength of the magnetic field, and let λ1(B) be thefirst eigenvalue of the magnetic Neumann Laplacianon the domain. It is proved that B 7→ λ1(B) ismonotone increasing for large B.
This result was proved by Fournais-Helffer in the caseof dimension 2 (first under a generic assumption,one year later in full generality). Our purpose (thisis again a common work with S. Fournais) is toshow here how one can prove the same result indimension 3 (but under generic assumptions). Theproof depends heavily on the two term asymptoticsof λ1(B) obtained by Pan and Helffer-Morame in2002.
If time permits,
we will discuss also applications of this monotonicity
for the identification of the critical fields insuperconductivity
and
present similar questions in the context of the theoryof liquid crystals.
Three models with parameters.
Model 1The spectral analysis is based in particular on theanalysis of the family
H(ξ) = D2t + (t+ ξ)2 , (1)
on the half-line (Neumann at 0) whose lowesteigenvalue µ(ξ) admits a unique minimum at ξ0 < 0.
We have to keep in mind two universal constantsattached to the problem on R
+.
The first one is
Θ0 = µ(ξ0) . (2)
It corresponds to the bottom of the spectrum of theNeumann realization in R2
+ (with B = 1).
Note thatΘ0 ∈]0, 1[ .
The second constant is
δ0 =1
2µ′′(ξ0) , (3)
Model 2The second model is quite specific of the problem indimension 3. We look in x1 > 0 to
L(ϑ,−i∂t) = −∂2x1−∂2
x2+(−i∂t+cosϑx1+sinϑx2)
2 .
By Partial Fourier transform, we arrive to :
L(ϑ, τ) = −∂2x1
− ∂2x2
+ (τ + cosϑx1 + sinϑx2)2 ,
in x1 > 0 and with Neumann condition on x1 = 0.
It is enough to consider the variation with respect toϑ ∈ [0, π
2 ].
The bottom the spectrum is given by :
ς(ϑ) := inf Spec (L(ϑ,−i∂t)) = infτ
(inf Spec (L(ϑ, τ))) .
We first observe the following lemma :
Lemma a.If ϑ ∈]0, π
2 ], then Spec (L(ϑ, τ)) is independent of τ .
This is trivial by translation in the x2 variable.
One can then show that the function ϑ 7→ ς(ϑ) iscontinuous on ]0, π
2 [ .This is based on the analysis of the essential spectrumof
L(ϑ) := D2x1
+D2x2
+ (x1 cosϑ+ x2 sinϑ)2 .
and that the bottom of the spectrum of this operatorcorresponds to an eigenvalue.We then show easily that
ς(0) = Θ0 < 1 .
andς(π
2) = 1 .
Finally, one shows that ϑ 7→ ς(ϑ) is monotonicallyincreasing and that
ς(ϑ) = Θ0 + α1|ϑ| + O(ϑ2) , (4)
with
α1 =
√µ′′(ξ0)
2. (5)
Model 3 : Montgomery’s model.
When the assumptions are not satisfied, and thatthe magnetic field B vanishes. Other models shouldbe considered. An interesting case is the case whenB vanishes along a line. This model was proposedby Montgomery in connection with subriemanniangeometry but this model appears also in the analysisof the dimension 3 case.
More precisely, we meet the following family(depending on ρ) of quartic oscillators :
D2t + (t2 − ρ)2 . (6)
Denoting by ν(ρ) the lowest eigenvalue, Kwek-Panhave shown that there exists a unique minimum ofν(ρ) leading to a new universal constant
ν0 = infρ∈R
ν(ρ) . (7)
Main results
Here we will describe the results of Helffer-Morame,Lu-Pan, Pan and the recent paper of Fournais-Helffer.
Let Ω ⊂ R3 be a bounded open set with smoothboundary, let β ∈ R3 be a unit vector, and define F
to be a vector field such that
curl F = β, in Ω, F ·N = 0 on ∂Ω,(8)
where N(x) is the unit interior normal vector to ∂Ω.Define QB to be the closed quadratic form
W 1,2(Ω) ∋ u 7→ QB(u) :=
∫
Ω
|(−i∇ +BF)u(x)|2 dx.
(9)
Let H(B) be the self-adjoint operator associated toQB.
In other words, H(B) is the differential operator(−i∇ +BF)2 with domain
u ∈W 2,2(Ω) : N · ∇u|∂Ω = 0 .
The operator H(B) has compact resolvent and weintroduce
λ1(B) := inf Spec H(B) . (10)
We will prove (under a generic assumption onthe domain Ω) that the mapping B 7→ λ1(B) ismonotonically increasing for sufficiently large valuesof B.
We will work under the following geometricassumption.
“Generic” Assumptions = G-Ass.We assume that the set of boundary points where βis tangent to ∂Ω, i.e.
Γ := x ∈ ∂Ω∣∣β ·N(x) = 0, (11)
is a regular submanifold of ∂Ω :
grad ′(β ·N)(x) 6= 0 , ∀x ∈ Γ . (12)
We finally assume that the set of points where β istangent to Γ is finite.
These assumptions are rather generic and for instancesatisfied for ellipsoids.
We will need the known two-term asymptotics of thegroundstate energy of H(B). The following resultwas proved by Helffer-Morame (the correspondingupper bound was also given by Pan).
Theorem 1If Ω and β satisfy G-Assumptions, then asB → +∞
λ1(B) = Θ0B + γ0B23 + O(B
23−η), (13)
for some η > 0.Here γ0 is defined by
γ0 := infx∈Γ
γ0(x), (14)
where
γ0(x) := 2−2/3ν0δ1/30 |kn(x)|2/3
(δ0+(1−δ0)|T (x)·β|2
)1/3
.
(15)Here T (x) is the oriented, unit tangent vector to Γat the point x and
kn(x) = | grad ′(β ·N)(x)| .
The new result obtained in collaboration withS. Fournais is the
Theorem 2Let Ω ⊂ R3 and β satisfying G- Assumptions
Let Γ1, . . . ,Γn be the collection of disjoint smoothcurves making up Γ. We assume that, for all j thereexists xj ∈ Γj such that γ0(xj) > γ0.
Then the directional derivativesλ′1,± := limt→0±
λ1(B+t)−λ(B)t ,
exist.
Moreover
limB→∞
λ′1,+(B) = limB→∞
λ′1,−(B) = Θ0. (16)
Localization estimatesWe start by recalling the decay of a groundstate inthe direction normal to the boundary. We will oftenuse the notation
t(x) := dist (x, ∂Ω). (17)
Now, if φ ∈ C∞0 (Ω), i.e. has support away from the
boundary, a simple integration by parts implies that
QB(φ) ≥ B‖φ‖22. (18)
It is a consequence of this elementary inequality(and the fact that Θ0 < 1) that groundstates areexponentially localized near the boundary.
Theorem 3There exist constants C, a1 > 0, B0 > 0 such that
∫Ωe2a1B1/2t(x)
(|ψB(x)|2
+B−1|(−i∇ +BF)ψB(x)|2)dx
≤ C ‖ψB‖22,
(19)
for all B ≥ B0, and all groundstates ψB of theoperator H(B).
We will mainly use this localization result in thefollowing form.
Corollary 4For all n ∈ N, there exists Cn > 0 and Bn ≥ 0 suchthat, ∀B ≥ Bn,
∫t(x)n|ψB(x)|2 dx ≤ Cn B
−n/2‖ψB‖22 .
We work in tubular neighborhoods of the boundaryas follows. For ǫ > 0, define
B(∂Ω, ǫ) = x ∈ Ω : t(x) ≤ ǫ. (20)
For sufficiently small ǫ0 we have that, for all x ∈B(∂Ω, 2ǫ0), there exists a unique point y(x) ∈ ∂Ωsuch that t(x) = dist (x, y(x)).
Define, for y ∈ ∂Ω, the function ϑ(y) ∈ [−π/2, π/2]by
sinϑ(y) := −β ·N(y). (21)
We extend ϑ to the tubular neighborhood B(∂Ω, 2ǫ0)by ϑ(x) := ϑ(y(x)).
In order to obtain localization estimates in thevariable normal to Γ, we use the following operatorinequality (due to Helffer-Morame).
Theorem 5Let B0 be chosen such that B
−3/80 = ǫ0 and define,
for B ≥ B0, C > 0 and x ∈ Ω,
WB(x) :=
B − CB1/4, t(x) ≥ 2B−3/8,
Bς(ϑ(x)) − CB1/4, t(x) < 2B−3/8.
(22)
Then, for C large enough
H(B) ≥WB, (23)
(in the sense of quadratic forms) for all B ≥ B0.
We use this energy estimate to prove Agmon typeestimates on the boundary.
Theorem 6Suppose that Ω ⊂ R3 and β satisfy G-Assumptions.Define for x ∈ ∂Ω,
dΓ(x) := dist (x,Γ),
and extend dΓ to a tubular neighborhood of theboundary by dΓ(x) := dΓ(y(x)).Then there exist constants C, a2 > 0, B0 ≥ 0, suchthat
∫
B(∂Ω,ǫ0)
e2a2B1/2dΓ(x)3/2|ψB(x)|2 dx ≤ C‖ψB‖
22,
(24)
for all B ≥ B0 and all groundstates ψB of H(B).
We have the following easy consequence.
Corollary 7Suppose that Ω ⊂ R
3 satisfies G-Assumptionsrelatively to β. Then for all n ∈ N there existsCn > 0 such that
∫
B(∂Ω,ǫ0)
dΓ(x)n|ψB(x)|2 dx ≤ CnB−n/3‖ψB‖
22,
(25)
for all B > 0 and all groundstates ψB of H(B).
Consider now the set MΓ ⊂ Γ where the functionγ0 is minimal,
MΓ := x ∈ Γ : γ0 = γ0. (26)
For simplicity, we asume that Γ is connected.
Theorem 8Suppose that Ω ⊂ R3 satisfies G-Assumptionsrelatively to β and let δ > 0. Then for all N > 0there exists CN such that if ψB is a groundstate ofH(B), then
∫
x∈Ω : dist (x,MΓ)≥δ
|ψB(x)|2 dx ≤ CNB−N ,
(27)
for all B > 0.
Proposition 9Let dΓ be the function defined in Theorem 6 . Lets0 ∈ Γ and define, for ǫ > 0,
Ω(ǫ, s0)= x ∈ Ω : dist (x,Γ) < ǫ and dist (x, s0) > ǫ.
Then, if ǫ is sufficiently small, there exists a functionφ ∈ C∞(Ω) such that
A := F + ∇φ ,
and satisfies
|A(x)| ≤ C(t(x) + dΓ(x)2
),
for all x ∈ Ω(ǫ, s0).
[Proof of Prop. 9 ]We use adapted coordinates (r, s, t) near Γ (N = 1).Γ is parametrized by arc-length as
|Γ|
2πS
1 ∋ s 7→ Γ(s) ∈ ∂Ω.
Given x ∈ Ω, close to Γ, there is a uniquepoint Γ(s(x)) ∈ Γ such that dist ∂Ω(y(x),Γ) =dist ∂Ω(y(x),Γ(s(x))), where dist ∂Ω denotes thegeodesic distance on the boundary. The coordinates(r, s, t) associated to the point x now satisfy
|r| = dist ∂Ω(y(x),Γ), s = s(x), t = dist (x, ∂Ω).
Notice that dΓ(x) ∼ |r(x)|, so we may replace dΓ
by r in the proposition.
Let A1dr + A2ds+ A3dt be the magnetic one-formωA = A · dx pulled-back (or pushed forward) to thenew coordinates (r, s, t). Also write
dωA = B12dr ∧ ds+ B13dr ∧ dt+ B23ds ∧ dt.
Clearly, Bij = ∂iAj − ∂jAi, for i < j. Here weidentify (1, 2, 3) with (r, s, t) for the derivatives.
The magnetic field β corresponds to the magnetictwo-form via the Hodge-map. In particular, since βis tangent to ∂Ω at Γ we get that
B12(0, s, 0) = 0. (28)
We now find a particular solution A such thatcurl A = B .
We make the Ansatz
A1 = −
∫ t
0
B13(r, s, τ) dτ, (29)
A2 = −
∫ t
0
B23(r, s, τ) dτ +
∫ r
0
B12(ρ, s, 0) dρ,
(30)
A3 = 0. (31)
One verifies by inspection that with these choices
|A| ≤ C(r2 + t). (32)
Transporting this A back to the original coordinatesgives an A with
curl A = 1, |A(x)| ≤ C(t(x) + dΓ(x)2).
Since Ω(ǫ, s0) is simply connected (for sufficiently
small ǫ) A is gauge equivalent to F and theproposition is proved.
Monotonicity
We now prove how one can derive the monotonicityresult from the known asymptotics of the groundstateenergy and localization estimates for the groundstateitself.
Based on these estimates the proof of Theorem isvery similar to the two-dimensional case.
Proof of Theorem 2For simplicity, we assume that Γ is connected.Applying analytic perturbation theory to H(B) weget the first part.
Let s0 ∈ Γ be a point with γ(s0) > γ0. Let A bethe vector potential defined in Proposition 9.
Let QB the quadratic form
W 1,2(Ω) ∋ u 7→ QB(u) =
∫
Ω
| − i∇u+BAu|2dx ,
and H(B) be the associated operator.
Then H(B) and H(B) are unitarily equivalent:
H(B) = eiBφH(B)e−iBφ, for some φ independentof B.With ψ+
1 ( · ;β) being a suitable choice of normalizedgroundstate, we get (by analytic perturbation theoryapplied to H(B) and the explicit relation between
H(B) and H(B),
λ′1,+(B)
= 〈Aψ+1 ( · ;B) , pB bA
ψ+1 ( · ;B)〉
+〈pB bAψ+
1 ( · ;B) , Aψ+1 ( · ;B)〉 .
(33)
We now obtain for any b > 0,
λ′1,+(B) =QB+b(ψ
+1 ( · ;B)) − QB(ψ+
1 ( · ;B))
b(34)
− b
∫
Ω
|A|2 |ψ+1 (x;B)|2 dx
≥λ1(B + b) − λ1(B)
b− b
∫
Ω
|A|2 |ψ+1 (x;B)|2 dx .
(35)
We choose b := MB23−η, with η from (13) and
M > 0 (to be taken arbitrarily large in the end).Then, using (13), (34) becomes
λ′1,+(B) ≥ Θ0 + γ0B−1/3(1+b/B)2/3−1
b/B
−CM−1 − b∫Ω|A|2 |ψ+
1 (x;B)|2 dx ,(36)
for some constant C independent of M,B.
If we can prove that
B23
∫
Ω
|A|2 |ψ+1 (x;B)|2 dx ≤ C, (37)
for some constant C independent of B, then we cantake the limit B → ∞ in (36) and obtain
lim infB→∞
λ′1,+(B) ≥ Θ0 − CM−1. (38)
Since M was arbitrary this implies the lower boundfor λ′1,+(B). Applying the same argument to thederivative from the left, λ′1,−(B), we get (theinequality gets turned since b < 0)
lim supB→∞
λ′1,−(B) ≤ Θ0. (39)
Since, by perturbation theory, λ′1,+(B) ≤ λ′1,−(B)for all B, we get (16).
Thus it remains to prove (37).
By Proposition 9 we can estimate
∫Ω|A|2 |ψ+
1 (x;B)|2 dx≤ C
∫Ω(ǫ,s0
)(t2 + r4)|ψ+1 (x;B)|2 dx
+‖A‖2∞
∫Ω\Ω
(ǫ,s0
) |ψ+1 (x;B)|2 dx .
Combining Corollaries 4 and 7 and Theorem 8,we therefore find the existence of a constant C > 0such that :
∫
Ω
|A|2 |ψ+1 (x;B)|2 dx ≤ C B−1 , (40)
which is stronger than the needed estimate (37).
Ginzburg-Landau functional
The Ginzburg-Landau functional is given by
Eκ,H[ψ,A] =∫Ω
|∇κHAψ|
2 − κ2|ψ|2 + κ2
2 |ψ|4
+κ2H2| curl A − β|2dx ,
withΩ simply connected,(ψ,A) ∈W 1,2(Ω; C) ×W 1,2(Ω; R3),β = (0, 0, 1)and where∇A = (∇ + iA).
We fix the choice of gauge by imposing that
div A = 0 in Ω , A · ν = 0 on ∂Ω .
Terminology for the minimizers
The pair (0,F) is called the Normal State.
A minimizer (ψ,A) for which ψ never vanishes willbe called SuperConducting State.
In the other cases, one will speak about Mixed State.
The general question is to determine the topology ofthe subset in R
+ × R+ of the (κ,H) corresponding
to minimizers belonging to each of these threesituations.
Theorem 10 (Lu-Pan-Fournais-Helffer)
There exists κ0 such that, ∀κ ≥ κ0, (0,F) is a globalminimizer of Eκ,H iff λ1(κH) < κ2 .
Remark.
This makes our analysis of the monotonicity of λ1
(for B large), which implies, for κ large, the existenceof a unique H such that
λ1(κH) = κ2 .
particularly interesting.
Same questions in the theory of Liquid
crystals
A similar arises in the theory of Liquid crystals. Thesimples question is to consider the case when themagnetic vector field is only of constant module !Typically one is interested in the case when
β(x) = B (cos τx3, sin τx3, 0) ,
as B → +∞. There are partial results for thismodel obtained by Almog and Pan.
The energy for this model can be written as
E [ψ,n] =
∫
Ω
|∇qnψ|
2 − k2|ψ|2 +
k2
2|ψ|4
+K1 | div n|2 + K2 |n · curl n + τ |2
+K3 |n× curl n|2
dx ,
where :
• Ω is the region occupied by the liquid crystal,
• ψ is a complex-valued function called the order
parameter,
• n is a real vector field of unit length called director
field,
• q is a real number called wave number,
• τ is a real number measuring the chiral pitch insome liquid crystal materials,
• K1, K2 and K3 are positive constants called theelastic coefficients,
and
• k is a positive constant which depends on thematerial and on the temperature.
References
[Ag] S. Agmon. Lectures on exponential decay of
solutions of second order elliptic equations.
Math. Notes, T. 29, Princeton University Press(1982).
[BaPhTa] P. Bauman, D. Phillips, and Q. Tang.Stable nucleation for the Ginzburg-Landausystem with an applied magnetic field. Arch.Rational Mech. Anal. 142, p. 1-43 (1998).
[BeSt] A. Bernoff and P. Sternberg. Onset ofsuperconductivity in decreasing fields for generaldomains. J. Math. Phys. 39, p. 1272-1284(1998).
[BoHe] C. Bolley and B. Helffer. An application ofsemi-classical analysis to the asymptotic studyof the supercooling field of a superconductingmaterial. Ann. Inst. H. Poincare (SectionPhysique Theorique) 58 (2), p. 169-233 (1993).
[Bon1] V. Bonnaillie. Analyse mathematique de lasupraconductivite dans un domaine a coins :methodes semi-classiques et numeriques. Thesede Doctorat, Universite Paris 11 (2003).
[Bon2] V. Bonnaillie. On the fundamental state fora Schrodinger operator with magnetic fields indomains with corners. Asymptotic Anal. 41 (3-4), p. 215-258, (2005).
[BonDa] V. Bonnaillie and M. Dauge. Asymptoticsfor the fundamental state of the Schrodingeroperator with magnetic field near a corner.(2004).
[BonFo] V. Bonnaillie-Noel and S. Fournais.Preprint 2007.
[CFKS] H.L. Cycon, R.G. Froese, W. Kirsch, andB. Simon. Schrodinger Operators. Springer-Verlag, Berlin 1987.
[DaHe] M. Dauge and B. Helffer. Eigenvalues
variation I, Neumann problem for Sturm-Liouville operators. J. Differential Equations104 (2), p. 243-262 (1993).
[DiSj] M. Dimassi and J. Sjostrand. SpectralAsymptotics in the semi-classical limit. LondonMathematical Society. LLocalizationecture NoteSeries 268. Cambridge University Press (1999).
[FoHel1] S. Fournais and B. Helffer. Energyasymptotics for type II superconductors. Calc.Var. Partial Differ. Equ. 24 (3) (2005), p. 341-376.
[FoHel2] S. Fournais and B. Helffer. Accurateeigenvalue asymptotics for Neumann magneticLaplacians. Ann. Inst. Fourier 56 (1) (2006),p. 1-67.
[FoHel3] S. Fournais and B. Helffer. On the thirdcritical field in Ginzburg-Landau theory. Comm.in Math. Physics 266 (1) (2006), p. 153-196.
[FoHel4] S. Fournais and B. Helffer. Strongdiamagnetism for general domains andapplications. To appear in Ann. Inst. Fourier(2007).
[FoHel5] S. Fournais and B. Helffer. Optimaluniform elliptic estimates for the Ginzburg-Landau System. Submitted (2006).
[FoHel6] S. Fournais and B. Helffer. Onthe Ginzburg-Landau critical field in threedimensions. In preparation (2007).
[GiPh] T. Giorgi and D. Phillips. The breakdown ofsuperconductivity due to strong fields for theGinzburg-Landau model. SIAM J. Math. Anal.30 (1999), no. 2, 341–359 (electronic).
[Hel] B. Helffer. Introduction to the semiclassical
analysis for the Schrodinger operator and
applications. Springer lecture Notes in Math.1336 (1988).
[HeMo1] B. Helffer and A. Mohamed. Semiclassicalanalysis for the ground state energy of aSchrodinger operator with magnetic wells. J.Funct. Anal. 138 (1), p. 40-81 (1996).
[HeMo2] B. Helffer and A. Morame. Magneticbottles in connection with superconductivity. J.Funct. Anal. 185 (2), p. 604-680 (2001).
[HeMo3] B. Helffer and A. Morame. Magneticbottles for the Neumann problem : curvatureeffect in the case of dimension 3 (General case).Ann. Sci. Ecole Norm. Sup. 37, p. 105-170(2004).
[HePa] B. Helffer and X. Pan. Upper criticalfield and location of surface nucleation ofsuperconductivity. Ann. Inst. H. Poincare(Section Analyse non lineaire) 20 (1), p. 145-181 (2003).
[HeSj] B. Helffer and J. Sjostrand. Multiple wells
in the semiclassical limit I. Comm. PartialDifferential Equations 9 (4), p. 337-408 (1984).
[LuPa1] K. Lu and X-B. Pan. Estimates of the uppercritical field for the Ginzburg-Landau equationsof superconductivity. Physica D 127, p. 73-104(1999).
[LuPa2] K. Lu and X-B. Pan. Eigenvalue problems ofGinzburg-Landau operator in bounded domains.J. Math. Phys. 40 (6), p. 2647-2670, June 1999.
[LuPa3] K. Lu and X-B. Pan. Gauge invarianteigenvalue problems on R
2 and R2+. Trans.
Amer. Math. Soc. 352 (3), p. 1247-1276 (2000).
[LuPa4] K. Lu and X-B. Pan. Surface nucleationof superconductivity in 3-dimension. J. ofDifferential Equations 168 (2), p. 386-452(2000).
[Pan] X-B. Pan. Surface superconductivity in
applied magnetic fields above HC3 Comm.Math. Phys. 228, p. 327-370 (2002).
[PiFeSt] M. del Pino, P.L. Felmer, and P. Sternberg.Boundary concentration for eigenvalue problemsrelated to the onset of superconductivity.Comm. Math. Phys. 210, p. 413-446 (2000).
[SaSe] E. Sandier, S. Serfaty. Important seriesof contributions.... including a recent book inBirkhauser.
[S-JSaTh] D. Saint-James, G. Sarma, E.J. Thomas.Type II Superconductivity. Pergamon, Oxford1969.
[St] P. Sternberg. On the Normal/SuperconductingPhase Transition in the Presence of LargeMagnetic Fields. In Connectivity andSuperconductivity, J. Berger and J. RubinsteinEditors. Lect. Notes in Physics 63, p. 188-199(1999).
[TiTi] D. R. Tilley and J. Tilley: Superfluidity
and superconductivity. 3rd edition. Instituteof Physics Publishing, Bristol and Philadelphia1990.
[Ti] M. Tinkham, Introduction to
Superconductivity. McGraw-Hill Inc., NewYork, 1975.