Sharp asymptotic estimates for eigenvalues of Aharonov ... · Veronica Felli University of...
Transcript of Sharp asymptotic estimates for eigenvalues of Aharonov ... · Veronica Felli University of...
Sharp asymptotic estimates for eigenvalues ofAharonov-Bohm operators with varying poles
Veronica Felli
University of Milano-Bicocca
Workshop in Nonlinear PDEs
joint works with L. AbatangeloarXiv:1504.00252, arXiv:1505.05280
Veronica Felli AB-operators with varying poles September 10, 2015 1 / 27
Introduction The Aharonov-Bohm operator
The Aharonov-Bohm operator
For a = (a1, a2) ∈ R2 and γ ∈ R \ Z, we consider the vector potential
Aa(x1, x2) = γ
(−(x2 − a2)
(x1 − a1)2 + (x2 − a2)2,
x1 − a1
(x1 − a1)2 + (x2 − a2)2
),
(x1, x2) ∈ R2 \ a. Aa generates the Aharonov-Bohm magnetic field inR2 with pole a and circulation γ.
The AB magnetic field is produced by an infinitely long thin solenoidintersecting perpendicularly the plane (x1, x2) at the point a, as the radiusof the solenoid goes to zero and the magnetic flux remains constantlyequal to γ. Negletting the irrelevant coordinate along the solenoid, theproblem becomes 2-dimensional.
Veronica Felli AB-operators with varying poles September 10, 2015 2 / 27
Introduction The Aharonov-Bohm operator
The Aharonov-Bohm operator
For a = (a1, a2) ∈ R2 and γ ∈ R \ Z, we consider the vector potential
Aa(x1, x2) = γ
(−(x2 − a2)
(x1 − a1)2 + (x2 − a2)2,
x1 − a1
(x1 − a1)2 + (x2 − a2)2
),
(x1, x2) ∈ R2 \ a. Aa generates the Aharonov-Bohm magnetic field inR2 with pole a and circulation γ.
Aharonov-Bohm effect [Y. Aharonov, D. Bohm, Phys. Rev. (1959)]
The AB magnetic field is a δ-like magnetic field: a quantum particle movingin R2 \ a is affected by the magnetic potential, despite being confined toa region in which the magnetic field is zero.
Veronica Felli AB-operators with varying poles September 10, 2015 2 / 27
Introduction The magnetic eigenvalue problem
The magnetic eigenvalue problem
Let us consider the Schrodinger operator with AB vector potential
(i∇+Aa)2u = −∆u+ 2iAa · ∇u+ |Aa|2u.
In a bounded, open and simply connected domain Ω ⊂ R2, ∀ a ∈ Ω, weconsider the eigenvalue problem
(i∇+Aa)2u = λu, in Ω,
u = 0, on ∂Ω,(Ea)
in a weak sense.
Veronica Felli AB-operators with varying poles September 10, 2015 3 / 27
Introduction The magnetic eigenvalue problem
The functional setting
For every a ∈ Ω, H1,a(Ω,C) is the completion ofu ∈ H1(Ω,C) ∩ C∞(Ω,C) : u vanishes in a neighborhood of a w.r.t.
‖u‖H1,a(Ω,C) =(‖∇u‖2L2(Ω,C2) + ‖u‖2L2(Ω,C) +
∥∥ u|x−a|
∥∥2
L2(Ω,C)
)1/2,
which is equivalent to(‖(i∇+Aa)u‖2L2(Ω,C2) + ‖u‖2L2(Ω,C)
)1/2.
Hardy type inequality [A. Laptev, T. Weidl (1999)]∫Dr(a)
|(i∇+Aa)u|2 dx ≥(
minj∈Z|j − γ|
)2 ∫Dr(a)
|u(x)|2
|x− a|2dx,
for all r > 0, a ∈ R2 and u ∈ H1,a(Dr(a),C)
Let H1,a0 (Ω,C) =
u ∈ H1
0 (Ω,C) : u|x−a| ∈ L
2(Ω,C)
.
Veronica Felli AB-operators with varying poles September 10, 2015 4 / 27
Introduction The magnetic eigenvalue problem
The functional setting
For every a ∈ Ω, H1,a(Ω,C) is the completion ofu ∈ H1(Ω,C) ∩ C∞(Ω,C) : u vanishes in a neighborhood of a w.r.t.
‖u‖H1,a(Ω,C) =(‖∇u‖2L2(Ω,C2) + ‖u‖2L2(Ω,C) +
∥∥ u|x−a|
∥∥2
L2(Ω,C)
)1/2,
which is equivalent to(‖(i∇+Aa)u‖2L2(Ω,C2) + ‖u‖2L2(Ω,C)
)1/2.
Hardy type inequality [A. Laptev, T. Weidl (1999)]∫Dr(a)
|(i∇+Aa)u|2 dx ≥(
minj∈Z|j − γ|
)2 ∫Dr(a)
|u(x)|2
|x− a|2dx,
for all r > 0, a ∈ R2 and u ∈ H1,a(Dr(a),C)
Let H1,a0 (Ω,C) =
u ∈ H1
0 (Ω,C) : u|x−a| ∈ L
2(Ω,C)
.
Veronica Felli AB-operators with varying poles September 10, 2015 4 / 27
Introduction The magnetic eigenvalue problem
The functional setting
For every a ∈ Ω, H1,a(Ω,C) is the completion ofu ∈ H1(Ω,C) ∩ C∞(Ω,C) : u vanishes in a neighborhood of a w.r.t.
‖u‖H1,a(Ω,C) =(‖∇u‖2L2(Ω,C2) + ‖u‖2L2(Ω,C) +
∥∥ u|x−a|
∥∥2
L2(Ω,C)
)1/2,
which is equivalent to(‖(i∇+Aa)u‖2L2(Ω,C2) + ‖u‖2L2(Ω,C)
)1/2.
Hardy type inequality [A. Laptev, T. Weidl (1999)]∫Dr(a)
|(i∇+Aa)u|2 dx ≥(
minj∈Z|j − γ|
)2 ∫Dr(a)
|u(x)|2
|x− a|2dx,
for all r > 0, a ∈ R2 and u ∈ H1,a(Dr(a),C)
Let H1,a0 (Ω,C) =
u ∈ H1
0 (Ω,C) : u|x−a| ∈ L
2(Ω,C)
.
Veronica Felli AB-operators with varying poles September 10, 2015 4 / 27
Introduction The magnetic eigenvalue problem
The eigenvalues
Classical spectral theory (Ea) admits a sequence of real divergingeigenvalues λakk≥1 with finite multiplicity
λa1 ≤ λa2 ≤ · · · ≤ λaj ≤ . . . (repeated according to their multiplicity)
Problem:
study the behavior of the function
a 7→ λaj
in a neighborhood of a fixed point b ∈ Ω.
Up to a translation, it is not restrictive to consider b = 0 ∈ Ω.
Veronica Felli AB-operators with varying poles September 10, 2015 5 / 27
Introduction The magnetic eigenvalue problem
The eigenvalues
Classical spectral theory (Ea) admits a sequence of real divergingeigenvalues λakk≥1 with finite multiplicity
λa1 ≤ λa2 ≤ · · · ≤ λaj ≤ . . . (repeated according to their multiplicity)
Problem:
study the behavior of the function
a 7→ λaj
in a neighborhood of a fixed point b ∈ Ω.
Up to a translation, it is not restrictive to consider b = 0 ∈ Ω.
Veronica Felli AB-operators with varying poles September 10, 2015 5 / 27
Introduction The magnetic eigenvalue problem
The case γ = 12 , half-integer circulation
Aa(x) = A0(x− a), where A0(x1, x2) =1
2
(− x2
x21 + x2
2
,x1
x21 + x2
2
).
Bonnaillie-Noel, Helffer, Hoffmann-Ostenhof [J. Phys. A (2009)]Noris, Terracini [Indiana Univ. Math. J. (2010)]:nodal domains of eigenfunctions are related to spectral minimalpartitions of the Dirichlet laplacian with points of odd multiplicity
Bonnaillie-Noel, Noris, Nys, Terracini [Analysis and PDE (2014)]Lena [J. Math. Phys. (2015)], Noris, Nys, Terracini [CMP, to appear]Noris, Terracini [Indiana Univ. Math. J. (2010)]:a strong connection between nodal properties of eigenfunctions andthe critical points of the map a 7→ λaj .
Veronica Felli AB-operators with varying poles September 10, 2015 6 / 27
Introduction The magnetic eigenvalue problem
The case γ = 12 , half-integer circulation
Aa(x) = A0(x− a), where A0(x1, x2) =1
2
(− x2
x21 + x2
2
,x1
x21 + x2
2
).
Bonnaillie-Noel, Helffer, Hoffmann-Ostenhof [J. Phys. A (2009)]Noris, Terracini [Indiana Univ. Math. J. (2010)]:nodal domains of eigenfunctions are related to spectral minimalpartitions of the Dirichlet laplacian with points of odd multiplicity
Bonnaillie-Noel, Noris, Nys, Terracini [Analysis and PDE (2014)]Lena [J. Math. Phys. (2015)], Noris, Nys, Terracini [CMP, to appear]Noris, Terracini [Indiana Univ. Math. J. (2010)]:a strong connection between nodal properties of eigenfunctions andthe critical points of the map a 7→ λaj .
Veronica Felli AB-operators with varying poles September 10, 2015 6 / 27
The function a 7→ λaj Regularity
Properties of the map a 7→ λaj
Bonnaillie-Noel, Noris, Nys, Terracini (2014), Lena (2015):
∀ j ≥ 1, the map a 7→ λaj is continuos in Ω and admits a continous
extension on Ω (as a→ ∂Ω, λaj → j-th eigenvalue of −∆ in Ω)
⇓
since it constant on ∂Ω, the map a 7→ λaj has an extremal point in Ω.
Improved regularity for simple eigenvalues: if ∃n0 ≥ 1 s.t.
λ0n0
is simple,
then the function a 7→ λan0is analytic in a neighborhood of 0.
Question:
what is the leading term in the asymptotic expansion of the map a 7→ λan0?
Veronica Felli AB-operators with varying poles September 10, 2015 7 / 27
The function a 7→ λaj Regularity
Properties of the map a 7→ λaj
Bonnaillie-Noel, Noris, Nys, Terracini (2014), Lena (2015):
∀ j ≥ 1, the map a 7→ λaj is continuos in Ω and admits a continous
extension on Ω (as a→ ∂Ω, λaj → j-th eigenvalue of −∆ in Ω)
⇓
since it constant on ∂Ω, the map a 7→ λaj has an extremal point in Ω.
Improved regularity for simple eigenvalues: if ∃n0 ≥ 1 s.t.
λ0n0
is simple,
then the function a 7→ λan0is analytic in a neighborhood of 0.
Question:
what is the leading term in the asymptotic expansion of the map a 7→ λan0?
Veronica Felli AB-operators with varying poles September 10, 2015 7 / 27
The function a 7→ λaj Regularity
Properties of the map a 7→ λaj
Bonnaillie-Noel, Noris, Nys, Terracini (2014), Lena (2015):
∀ j ≥ 1, the map a 7→ λaj is continuos in Ω and admits a continous
extension on Ω (as a→ ∂Ω, λaj → j-th eigenvalue of −∆ in Ω)
⇓
since it constant on ∂Ω, the map a 7→ λaj has an extremal point in Ω.
Improved regularity for simple eigenvalues: if ∃n0 ≥ 1 s.t.
λ0n0
is simple,
then the function a 7→ λan0is analytic in a neighborhood of 0.
Question:
what is the leading term in the asymptotic expansion of the map a 7→ λan0?
Veronica Felli AB-operators with varying poles September 10, 2015 7 / 27
The function a 7→ λaj Rate of convergence of λa → λ0
Let us assume that ∃n0 ≥ 1 such that λ0n0
is simple and denote λ0 = λ0n0
and, for any a ∈ Ω, λa = λan0. Thus λa → λ0.
Let ϕ0 ∈ H1,00 (Ω,C) be an eigenfunction of (i∇+A0)2 associated to λ0
(i∇+A0)2ϕ0 = λ0ϕ0, in Ω,
ϕ0 = 0, on ∂Ω,such that
∫Ω|ϕ0(x)|2 dx = 1.
Theorem [F., Ferrero, Terracini (2011)]
There exists k ∈ N odd s.t.
r−k/2ϕ0(r(cos t, sin t))→ β1eit2√π
cos(k2 t)
+ β2eit2√π
sin(k2 t)
as r → 0+
in C1,τ ([0, 2π],C), with β1, β2 ∈ C, (β1, β2) 6= (0, 0).
[Helffer-Hoffmann-Ostenhof-Hoffmann-Ostenhof-Owen (1999)]: e−it2ϕ0(r(cos t, sin t))
is a multiple of a real-valued function and therefore either β1 = 0 or β2β1
is real.
Veronica Felli AB-operators with varying poles September 10, 2015 8 / 27
The function a 7→ λaj Rate of convergence of λa → λ0
ϕ0 has at 0 a zero or order k2 for some
odd k ∈ Nϕ0 has got exactly k nodal linesmeeting at 0 and dividing the wholeangle into k equal parts.
0
Noris, Terracini (2010), Bonnaillie-Noel, Noris, Nys, Terracini (2014):The behavior of the eigenvalue λa is strongly related to the structure ofthe nodal lines of the associated eigenfunction.
rate of convergence of λa to λ0 ←→ order of vanishing of ϕ0 at 0
Veronica Felli AB-operators with varying poles September 10, 2015 9 / 27
The function a 7→ λaj Rate of convergence of λa → λ0
ϕ0 has at 0 a zero or order k2 for some
odd k ∈ Nϕ0 has got exactly k nodal linesmeeting at 0 and dividing the wholeangle into k equal parts.
0
Noris, Terracini (2010), Bonnaillie-Noel, Noris, Nys, Terracini (2014):The behavior of the eigenvalue λa is strongly related to the structure ofthe nodal lines of the associated eigenfunction.
rate of convergence of λa to λ0 ←→ order of vanishing of ϕ0 at 0
Veronica Felli AB-operators with varying poles September 10, 2015 9 / 27
The function a 7→ λaj Rate of convergence of λa → λ0
Theorem [Bonnaillie-Noel, Noris, Nys, Terracini (2014)]
If k ≥ 3, then
|λa − λ0| ≤ C|a|k+12 as a→ 0 for a constant C > 0 independent of a
the Taylor polynomials of the function a 7→ λ0 − λa with center 0 anddegree less or equal than k−1
2 vanish.
The control |a|k+12 is not the optimal one.
Our goal:
to establish the exact order of the asymptotic expansion of λa − λ0
proving that along a suitable direction the exact order is |a|k, where kis the number of nodal lines of ϕ0 at 0
to find the leading term in the Taylor expansion of the eigenvaluefunction, detecting its exact coefficients.
Veronica Felli AB-operators with varying poles September 10, 2015 10 / 27
The function a 7→ λaj Rate of convergence of λa → λ0
Theorem [Bonnaillie-Noel, Noris, Nys, Terracini (2014)]
If k ≥ 3, then
|λa − λ0| ≤ C|a|k+12 as a→ 0 for a constant C > 0 independent of a
the Taylor polynomials of the function a 7→ λ0 − λa with center 0 anddegree less or equal than k−1
2 vanish.
The control |a|k+12 is not the optimal one.
Our goal:
to establish the exact order of the asymptotic expansion of λa − λ0
proving that along a suitable direction the exact order is |a|k, where kis the number of nodal lines of ϕ0 at 0
to find the leading term in the Taylor expansion of the eigenvaluefunction, detecting its exact coefficients.
Veronica Felli AB-operators with varying poles September 10, 2015 10 / 27
The function a 7→ λaj Rate of convergence of λa → λ0
Theorem [Bonnaillie-Noel, Noris, Nys, Terracini (2014)]
If k ≥ 3, then
|λa − λ0| ≤ C|a|k+12 as a→ 0 for a constant C > 0 independent of a
the Taylor polynomials of the function a 7→ λ0 − λa with center 0 anddegree less or equal than k−1
2 vanish.
The control |a|k+12 is not the optimal one.
Our goal:
to establish the exact order of the asymptotic expansion of λa − λ0
proving that along a suitable direction the exact order is |a|k, where kis the number of nodal lines of ϕ0 at 0
to find the leading term in the Taylor expansion of the eigenvaluefunction, detecting its exact coefficients.
Veronica Felli AB-operators with varying poles September 10, 2015 10 / 27
The function a 7→ λaj Rate of convergence of λa → λ0
Theorem [Bonnaillie-Noel, Noris, Nys, Terracini (2014)]
If k ≥ 3, then
|λa − λ0| ≤ C|a|k+12 as a→ 0 for a constant C > 0 independent of a
the Taylor polynomials of the function a 7→ λ0 − λa with center 0 anddegree less or equal than k−1
2 vanish.
The control |a|k+12 is not the optimal one.
Our goal:
to establish the exact order of the asymptotic expansion of λa − λ0
proving that along a suitable direction the exact order is |a|k, where kis the number of nodal lines of ϕ0 at 0
to find the leading term in the Taylor expansion of the eigenvaluefunction, detecting its exact coefficients.
Veronica Felli AB-operators with varying poles September 10, 2015 10 / 27
Sharp asymptotics Moving along directions of nodal lines
Moving along directions of nodal lines
Theorem (F.-Abatangelo, 2015)
Let r be the half-line tangent to a nodalline of eigenfunction ϕ0 associated to λ0
ending at 0. Then, as a→ 0 with a ∈ r,
λ0 − λa|a|k
→ −4|β1|2 + |β2|2
πmk.
0 ar
Here
(β1, β2) 6= (0, 0) is s.t.
r−k2ϕ0(r(cos t, sin t))→ β1
eit2√π
cos(k2 t)
+ β2eit2√π
sin(k2 t)
as r → 0+
mk < 0 is a negative constant depending only on k.
Veronica Felli AB-operators with varying poles September 10, 2015 11 / 27
Sharp asymptotics Moving along directions of nodal lines
Moving along directions of nodal lines
Theorem (F.-Abatangelo, 2015)
Let r be the half-line tangent to a nodalline of eigenfunction ϕ0 associated to λ0
ending at 0. Then, as a→ 0 with a ∈ r,
λ0 − λa|a|k
→ −4|β1|2 + |β2|2
πmk.
0 ar
Here
(β1, β2) 6= (0, 0) is s.t.
r−k2ϕ0(r(cos t, sin t))→ β1
eit2√π
cos(k2 t)
+ β2eit2√π
sin(k2 t)
as r → 0+
mk < 0 is a negative constant depending only on k.
Veronica Felli AB-operators with varying poles September 10, 2015 11 / 27
Sharp asymptotics Moving along directions of nodal lines
The constant mk
For every odd k ∈ N, ψk(r cos t, r sin t) = rk/2 sin(k2 t), r ≥ 0, t ∈ [0, 2π],
is the unique (up to a multiplicative constant) function which is
vanishing on s0 = [0,+∞)×0harmonic on R2 \ s0 and homogeneous of degree k/2
Let s := [1,+∞)×0 and R2+ = (x1, x2) ∈ R2 : x2 > 0).
Let D1,2s (R2
+) be the completion of C∞c (R2+ \ s) w.r.t. (
∫R2+|∇u|2 dx)1/2.
mk = minu∈D1,2
s (R2+)Jk(u)
where Jk(u) = 12
∫R2+|∇u(x)|2 dx−
∫ 10 u(x1, 0)∂ψk∂x2
(x1, 0) dx1.
mk = Jk(wk) = −12
∫R2+|∇wk(x)|2 dx
= −12
∫ 10∂+ψk∂x2
(x1, 0)wk(x1, 0) dx1 < 0
Veronica Felli AB-operators with varying poles September 10, 2015 12 / 27
Sharp asymptotics Moving along directions of nodal lines
The constant mk
For every odd k ∈ N, ψk(r cos t, r sin t) = rk/2 sin(k2 t), r ≥ 0, t ∈ [0, 2π],
is the unique (up to a multiplicative constant) function which is
vanishing on s0 = [0,+∞)×0harmonic on R2 \ s0 and homogeneous of degree k/2
Let s := [1,+∞)×0 and R2+ = (x1, x2) ∈ R2 : x2 > 0).
Let D1,2s (R2
+) be the completion of C∞c (R2+ \ s) w.r.t. (
∫R2+|∇u|2 dx)1/2.
mk = minu∈D1,2
s (R2+)Jk(u)
where Jk(u) = 12
∫R2+|∇u(x)|2 dx−
∫ 10 u(x1, 0)∂ψk∂x2
(x1, 0) dx1.
mk = Jk(wk) = −12
∫R2+|∇wk(x)|2 dx
= −12
∫ 10∂+ψk∂x2
(x1, 0)wk(x1, 0) dx1 < 0
Veronica Felli AB-operators with varying poles September 10, 2015 12 / 27
Sharp asymptotics Moving along directions of nodal lines
The constant mk
For every odd k ∈ N, ψk(r cos t, r sin t) = rk/2 sin(k2 t), r ≥ 0, t ∈ [0, 2π],
is the unique (up to a multiplicative constant) function which is
vanishing on s0 = [0,+∞)×0harmonic on R2 \ s0 and homogeneous of degree k/2
Let s := [1,+∞)×0 and R2+ = (x1, x2) ∈ R2 : x2 > 0).
Let D1,2s (R2
+) be the completion of C∞c (R2+ \ s) w.r.t. (
∫R2+|∇u|2 dx)1/2.
mk = minu∈D1,2
s (R2+)Jk(u)
where Jk(u) = 12
∫R2+|∇u(x)|2 dx−
∫ 10 u(x1, 0)∂ψk∂x2
(x1, 0) dx1.
mk = Jk(wk) = −12
∫R2+|∇wk(x)|2 dx
= −12
∫ 10∂+ψk∂x2
(x1, 0)wk(x1, 0) dx1 < 0
Veronica Felli AB-operators with varying poles September 10, 2015 12 / 27
Sharp asymptotics Moving along directions of nodal lines
Consequences
Analyticity of a 7→ λa
λ0 − λa|a|k
→ 4|β1|2 + |β2|2
πmk
as a→ 0 along the half-line opposite to thetangent to a nodal line of ϕ0.
0a
⇓
The restriction of the function λ0 − λa on the straight line tangent to anodal line changes sign at 0: is positive on the side of the nodal line andnegative on the opposite side
⇓
if λ0 is simple, then 0 cannot be an extremal point of the map a 7→ λa.
Veronica Felli AB-operators with varying poles September 10, 2015 13 / 27
Sharp asymptotics Moving along directions of nodal lines
Consequences
Analyticity of a 7→ λa
λ0 − λa|a|k
→ 4|β1|2 + |β2|2
πmk
as a→ 0 along the half-line opposite to thetangent to a nodal line of ϕ0.
0a
⇓
The restriction of the function λ0 − λa on the straight line tangent to anodal line changes sign at 0: is positive on the side of the nodal line andnegative on the opposite side
⇓
if λ0 is simple, then 0 cannot be an extremal point of the map a 7→ λa.
Veronica Felli AB-operators with varying poles September 10, 2015 13 / 27
Sharp asymptotics Moving along directions of nodal lines
Consequences
Analyticity of a 7→ λa
λ0 − λa|a|k
→ 4|β1|2 + |β2|2
πmk
as a→ 0 along the half-line opposite to thetangent to a nodal line of ϕ0.
0a
⇓
The restriction of the function λ0 − λa on the straight line tangent to anodal line changes sign at 0: is positive on the side of the nodal line andnegative on the opposite side
⇓
if λ0 is simple, then 0 cannot be an extremal point of the map a 7→ λa.
Veronica Felli AB-operators with varying poles September 10, 2015 13 / 27
Sharp asymptotics Moving along directions of nodal lines
Consequences
Proposition [Bonnaillie-Noel, Noris, Nys, Terracini (2014)]
Fix any j ∈ N. If 0 is an extremal point of a 7→ λaj , then
(i) either λ0j is not simple
(ii) or the eigenfunction of (i∇+A0)2 associated to λ0j has at 0 a zero
of order k/2 with k ≥ 3 odd.
Our theorem we can exclude alternative (ii): If 0 is an extremalpoint of a 7→ λaj =⇒ λ0
j is not simple.
if λ0 is simple and k ≥ 3, 0 is a saddle point forthe map a 7→ λa.
if λ0 is simple and k = 1, the gradient of the mapa 7→ λa in 0 is different from zero, then 0 is not astationary point, a fortiori not even an extremalpoint (Noris, Terracini (2010)).
Veronica Felli AB-operators with varying poles September 10, 2015 14 / 27
Sharp asymptotics Moving along directions of nodal lines
Consequences
Proposition [Bonnaillie-Noel, Noris, Nys, Terracini (2014)]
Fix any j ∈ N. If 0 is an extremal point of a 7→ λaj , then
(i) either λ0j is not simple
(ii) or the eigenfunction of (i∇+A0)2 associated to λ0j has at 0 a zero
of order k/2 with k ≥ 3 odd.
Our theorem we can exclude alternative (ii): If 0 is an extremalpoint of a 7→ λaj =⇒ λ0
j is not simple.
if λ0 is simple and k ≥ 3, 0 is a saddle point forthe map a 7→ λa.
if λ0 is simple and k = 1, the gradient of the mapa 7→ λa in 0 is different from zero, then 0 is not astationary point, a fortiori not even an extremalpoint (Noris, Terracini (2010)).
Veronica Felli AB-operators with varying poles September 10, 2015 14 / 27
Sharp asymptotics Moving along directions of nodal lines
Consequences
Proposition [Bonnaillie-Noel, Noris, Nys, Terracini (2014)]
Fix any j ∈ N. If 0 is an extremal point of a 7→ λaj , then
(i) either λ0j is not simple
(ii) or the eigenfunction of (i∇+A0)2 associated to λ0j has at 0 a zero
of order k/2 with k ≥ 3 odd.
Our theorem we can exclude alternative (ii): If 0 is an extremalpoint of a 7→ λaj =⇒ λ0
j is not simple.
if λ0 is simple and k ≥ 3, 0 is a saddle point forthe map a 7→ λa.
if λ0 is simple and k = 1, the gradient of the mapa 7→ λa in 0 is different from zero, then 0 is not astationary point, a fortiori not even an extremalpoint (Noris, Terracini (2010)).
Veronica Felli AB-operators with varying poles September 10, 2015 14 / 27
Sharp asymptotics Idea of the proof
Rotate the axes in such a way that the positive x1-axis is tangent to oneof the k nodal lines of ϕ0 ending at 0 it is not restrictive to assumethat β1 = 0.
For every b ∈ R2, we define θb : R2 \ b → [0, 2π)so that θb(b+ r(cos t, sin t)) = t ∀ r > 0, t ∈ [0, 2π).
∇(θb2
)= Ab, e−i
θb2 (i∇+Ab)
2u = −∆(e−i
θb2 u)
b
x
θb(x)
For all a ∈ Ω, let ϕa ∈ H1,a0 (Ω,C) \ 0 be an eigenfunction of (Ea)
associated to the eigenvalue λa such that∫Ω|ϕa(x)|2 dx = 1 and
∫Ωei2
(θ0−θa)(x)ϕa(x)ϕ0(x) dx is positive real.
We have that ϕa → ϕ0 in H1(Ω,C) and in C2loc(Ω \ 0,C) and
(i∇+Aa)ϕa → (i∇+A0)ϕ0 in L2(Ω,C).
Veronica Felli AB-operators with varying poles September 10, 2015 15 / 27
Sharp asymptotics Idea of the proof
Rotate the axes in such a way that the positive x1-axis is tangent to oneof the k nodal lines of ϕ0 ending at 0 it is not restrictive to assumethat β1 = 0.
For every b ∈ R2, we define θb : R2 \ b → [0, 2π)so that θb(b+ r(cos t, sin t)) = t ∀ r > 0, t ∈ [0, 2π).
∇(θb2
)= Ab, e−i
θb2 (i∇+Ab)
2u = −∆(e−i
θb2 u)
b
x
θb(x)
For all a ∈ Ω, let ϕa ∈ H1,a0 (Ω,C) \ 0 be an eigenfunction of (Ea)
associated to the eigenvalue λa such that∫Ω|ϕa(x)|2 dx = 1 and
∫Ωei2
(θ0−θa)(x)ϕa(x)ϕ0(x) dx is positive real.
We have that ϕa → ϕ0 in H1(Ω,C) and in C2loc(Ω \ 0,C) and
(i∇+Aa)ϕa → (i∇+A0)ϕ0 in L2(Ω,C).
Veronica Felli AB-operators with varying poles September 10, 2015 15 / 27
Sharp asymptotics Idea of the proof
Blow-up
Theorem
ϕa(|a|x)
|a|k/2→ β2√
πΨk as a = (|a|, 0)→ 0,
in H1,e(DR,C) ∀R > 1, a.e. and in C2loc(R2 \ e,C), where e = (1, 0).
Ψk is the unique function in H1,eloc (R2) satisfying
(i∇+Ae)2Ψk = 0, in R2,
and ∫R2
|(i∇+Ae)(Ψk − ei2θeψk)|2 < +∞.
If wk ∈ D1,2s (R2
+) minimizes the functional Jk, Ψk = eiθe2 (ψk + wk).
Veronica Felli AB-operators with varying poles September 10, 2015 16 / 27
Sharp asymptotics Idea of the proof
Blow-up
Theorem
ϕa(|a|x)
|a|k/2→ β2√
πΨk as a = (|a|, 0)→ 0,
in H1,e(DR,C) ∀R > 1, a.e. and in C2loc(R2 \ e,C), where e = (1, 0).
Ψk is the unique function in H1,eloc (R2) satisfying
(i∇+Ae)2Ψk = 0, in R2,
and ∫R2
|(i∇+Ae)(Ψk − ei2θeψk)|2 < +∞.
If wk ∈ D1,2s (R2
+) minimizes the functional Jk, Ψk = eiθe2 (ψk + wk).
Veronica Felli AB-operators with varying poles September 10, 2015 16 / 27
Sharp asymptotics Idea of the proof
Upper bound for λ0 − λa: the Rayleigh quotient for λ0
For all 1 ≤ j ≤ n0 and a ∈ Ω, let ϕaj ∈ H1,a0 (Ω,C) \ 0 be an
eigenfunction of problem (Ea) associated to the eigenvalue λaj such that∫Ω|ϕaj (x)|2 dx = 1 and
∫Ωϕaj (x)ϕa` (x) dx = 0 if j 6= `.
For j = n0, we choose ϕan0= ϕa.
Courant-Fisher minimax characterization of the eigenvalue λ0, we havethat
λ0 =min
max
u∈F\0
∫Ω |(i∇+A0)u|2dx∫
Ω |u|2 dx: F is a subspace of H1,0
0 (Ω,C),dimF = n0
.
Veronica Felli AB-operators with varying poles September 10, 2015 17 / 27
Sharp asymptotics Idea of the proof
Upper bound for λ0 − λa: the Rayleigh quotient for λ0
For all 1 ≤ j ≤ n0 and a ∈ Ω, let ϕaj ∈ H1,a0 (Ω,C) \ 0 be an
eigenfunction of problem (Ea) associated to the eigenvalue λaj such that∫Ω|ϕaj (x)|2 dx = 1 and
∫Ωϕaj (x)ϕa` (x) dx = 0 if j 6= `.
For j = n0, we choose ϕan0= ϕa.
Courant-Fisher minimax characterization of the eigenvalue λ0, we havethat
λ0 =min
max
u∈F\0
∫Ω |(i∇+A0)u|2dx∫
Ω |u|2 dx: F is a subspace of H1,0
0 (Ω,C),dimF = n0
.
Veronica Felli AB-operators with varying poles September 10, 2015 17 / 27
Sharp asymptotics Idea of the proof
For R > 2 and a = (|a|, 0) with |a| small, define vj,R,a as follows:
vj,R,a =
vextj,R,a, in Ω \DR|a|,
vintj,R,a, in DR|a|,j = 1, . . . , n0,
vextj,R,a := ei2
(θ0−θa)ϕaj in Ω\DR|a|, (i∇+A0)2vextj,R,a = λajvextj,R,a in Ω\DR|a|,
(i∇+A0)2vintj,R,a = 0, in DR|a|,
vintj,R,a = ei2
(θ0−θa)ϕaj , on ∂DR|a|.
dim(
spanv1,R,a, . . . , vn0,R,a)
= n0 After a Gram-Schmidtnormalization, use vj,R,a as test functions in the Rayleigh quotient
⇓
λ0−λa ≤∫DR|a|
∣∣(i∇+A0)vintn0,R,a
∣∣2dx−∫DR|a|
∣∣(i∇+Aa)ϕa∣∣2dx+o(|a|k)
as |a| → 0.
Veronica Felli AB-operators with varying poles September 10, 2015 18 / 27
Sharp asymptotics Idea of the proof
For R > 2 and a = (|a|, 0) with |a| small, define vj,R,a as follows:
vj,R,a =
vextj,R,a, in Ω \DR|a|,
vintj,R,a, in DR|a|,j = 1, . . . , n0,
vextj,R,a := ei2
(θ0−θa)ϕaj in Ω\DR|a|, (i∇+A0)2vextj,R,a = λajvextj,R,a in Ω\DR|a|,
(i∇+A0)2vintj,R,a = 0, in DR|a|,
vintj,R,a = ei2
(θ0−θa)ϕaj , on ∂DR|a|.
dim(
spanv1,R,a, . . . , vn0,R,a)
= n0
After a Gram-Schmidtnormalization, use vj,R,a as test functions in the Rayleigh quotient
⇓
λ0−λa ≤∫DR|a|
∣∣(i∇+A0)vintn0,R,a
∣∣2dx−∫DR|a|
∣∣(i∇+Aa)ϕa∣∣2dx+o(|a|k)
as |a| → 0.
Veronica Felli AB-operators with varying poles September 10, 2015 18 / 27
Sharp asymptotics Idea of the proof
For R > 2 and a = (|a|, 0) with |a| small, define vj,R,a as follows:
vj,R,a =
vextj,R,a, in Ω \DR|a|,
vintj,R,a, in DR|a|,j = 1, . . . , n0,
vextj,R,a := ei2
(θ0−θa)ϕaj in Ω\DR|a|, (i∇+A0)2vextj,R,a = λajvextj,R,a in Ω\DR|a|,
(i∇+A0)2vintj,R,a = 0, in DR|a|,
vintj,R,a = ei2
(θ0−θa)ϕaj , on ∂DR|a|.
dim(
spanv1,R,a, . . . , vn0,R,a)
= n0 After a Gram-Schmidtnormalization, use vj,R,a as test functions in the Rayleigh quotient
⇓
λ0−λa ≤∫DR|a|
∣∣(i∇+A0)vintn0,R,a
∣∣2dx−∫DR|a|
∣∣(i∇+Aa)ϕa∣∣2dx+o(|a|k)
as |a| → 0.
Veronica Felli AB-operators with varying poles September 10, 2015 18 / 27
Sharp asymptotics Idea of the proof
Blow-up and let R→ +∞
lim sup|a|→0
λ0 − λa|a|k
≤ |β2|2
πk√π(ξ(1)−
√π)
where
ξ(r) := 1√π
∫ 2π
0e−
i2θe(r cos t,r sin t)Ψk(r cos t, r sin t) sin
(k2 t)dt.
In a similar way we obtain a lower bound for λ0 − λa estimating theRayleigh quotient for λa
lim inf|a|→0
λ0 − λa|a|k
≥ |β2|2
πk√π(ξ(1)−
√π)
Hence
lim|a|→0
λ0 − λa|a|k
=|β2|2
πk√π(ξ(1)−
√π)
Veronica Felli AB-operators with varying poles September 10, 2015 19 / 27
Sharp asymptotics Idea of the proof
ξ(1)−√π =
1√π
∫ 2π
0wk(cos t, sin t) sin
(k2 t)dt = ω(1)
where
ω(r) :=1√π
∫ 2π
0wk(r cos t, r sin t) sin
(k2 t)dt.
wk ∈ D1,2s (R2
+) minimizes Jk . . .
ξ(1)−√π = ω(1) = − 4
k√πmk > 0
Veronica Felli AB-operators with varying poles September 10, 2015 20 / 27
Sharp asymptotics Idea of the proof
ξ(1)−√π =
1√π
∫ 2π
0wk(cos t, sin t) sin
(k2 t)dt = ω(1)
where
ω(r) :=1√π
∫ 2π
0wk(r cos t, r sin t) sin
(k2 t)dt.
wk ∈ D1,2s (R2
+) minimizes Jk . . .
ξ(1)−√π = ω(1) = − 4
k√πmk > 0
Veronica Felli AB-operators with varying poles September 10, 2015 20 / 27
Sharp asymptotics Idea of the proof
λ0 − λa|a|k
→ |β2|2
πk√π(ξ(1)−
√π) = −4
|β2|2
πmk
as a→ 0 tangentially to any of the k nodal lines of ϕ0.
The Taylor polynomials of the function a 7→ λ0 − λa with center 0 anddegree strictly smaller than k vanish, since they vanish on the kindependent directions corresponding to the nodal lines of ϕ0
⇓
λ0 − λa = P (a) + o(|a|k), as |a| → 0+,
for some
P 6≡ 0, P (a) = P (a1, a2) =
k∑j=0
cjak−j1 aj2
homogeneous polynomial of degree k.
Veronica Felli AB-operators with varying poles September 10, 2015 21 / 27
Sharp asymptotics Idea of the proof
λ0 − λa|a|k
→ |β2|2
πk√π(ξ(1)−
√π) = −4
|β2|2
πmk
as a→ 0 tangentially to any of the k nodal lines of ϕ0.
The Taylor polynomials of the function a 7→ λ0 − λa with center 0 anddegree strictly smaller than k vanish, since they vanish on the kindependent directions corresponding to the nodal lines of ϕ0
⇓
λ0 − λa = P (a) + o(|a|k), as |a| → 0+,
for some
P 6≡ 0, P (a) = P (a1, a2) =
k∑j=0
cjak−j1 aj2
homogeneous polynomial of degree k.
Veronica Felli AB-operators with varying poles September 10, 2015 21 / 27
Sharp asymptotics Moving along any direction
Moving along any direction
For α ∈ [0, 2π), p = (cosα, sinα)and a = |a|p = |a|(cosα, sinα)
ϕa(|a|x)
|a|k/2→ β2√
πΨp as a = |a|p→ 0,
in H1,p(DR,C) for every R > 1 andin C2
loc(R2 \ p,C).
0 α
a r
Ψp is the unique function in H1,ploc (R2,C) satisfying
(i∇+Ap)2Ψp = 0, in R2,∫R2\Dr
∣∣(i∇+Ap)(Ψp − ei2
(θp−θp0 )ei2θ0ψk)
∣∣2 dx < +∞, ∀ r > 1.
θp(p+r(cos t, sin t)) = t, θp0 (r(cos t, sin t)) = t, ∀ r > 0, t ∈ [α, α+ 2π)
Veronica Felli AB-operators with varying poles September 10, 2015 22 / 27
Sharp asymptotics Moving along any direction
Moving along any direction
For α ∈ [0, 2π), p = (cosα, sinα)and a = |a|p = |a|(cosα, sinα)
ϕa(|a|x)
|a|k/2→ β2√
πΨp as a = |a|p→ 0,
in H1,p(DR,C) for every R > 1 andin C2
loc(R2 \ p,C).
0 α
a r
Ψp is the unique function in H1,ploc (R2,C) satisfying
(i∇+Ap)2Ψp = 0, in R2,∫R2\Dr
∣∣(i∇+Ap)(Ψp − ei2
(θp−θp0 )ei2θ0ψk)
∣∣2 dx < +∞, ∀ r > 1.
θp(p+r(cos t, sin t)) = t, θp0 (r(cos t, sin t)) = t, ∀ r > 0, t ∈ [α, α+ 2π)
Veronica Felli AB-operators with varying poles September 10, 2015 22 / 27
Sharp asymptotics Moving along any direction
For α ∈ [0, 2π), p = (cosα, sinα) and a = |a|p = |a|(cosα, sinα)
lim|a|→0
λ0 − λa|a|k
=|β2|2
πk√πf(α),
wheref : [0, 2π)→ R, f(α) = (ξp(1)−
√π),
ξp(r) :=1√π
∫ 2π
0e−
i2
(θp−θp0 )(r cos t,r sin t)Ψp(r cos t, r sin t)e−it2 sin
(k2 t)dt.
We already know that f(0) = ξe(1)−√π = ξ(1)−
√π = − 4
k√πmk > 0
Veronica Felli AB-operators with varying poles September 10, 2015 23 / 27
Sharp asymptotics Moving along any direction
For α ∈ [0, 2π), p = (cosα, sinα) and a = |a|p = |a|(cosα, sinα)
lim|a|→0
λ0 − λa|a|k
=|β2|2
πk√πf(α),
wheref : [0, 2π)→ R, f(α) = (ξp(1)−
√π),
ξp(r) :=1√π
∫ 2π
0e−
i2
(θp−θp0 )(r cos t,r sin t)Ψp(r cos t, r sin t)e−it2 sin
(k2 t)dt.
We already know that f(0) = ξe(1)−√π = ξ(1)−
√π = − 4
k√πmk > 0
Veronica Felli AB-operators with varying poles September 10, 2015 23 / 27
Sharp asymptotics Moving along any direction
Symmetry properties of f(α)
Rotation of 2πk :
R1(x1, x2) =
(cos 2π
k − sin 2πk
sin 2πk cos 2π
k
)(x1
x2
)Reflexion through the x1-axis:
R2(x) = R2(x1, x2) = (x1,−x2),
Problem:
how does the limit profile Ψp change when the above trasformationsact on p?
how does the coefficient ξp(1) and hence f(α) change when theabove trasformations act on p?
Veronica Felli AB-operators with varying poles September 10, 2015 24 / 27
Sharp asymptotics Moving along any direction
Symmetry properties of f(α)
Rotation of 2πk :
R1(x1, x2) =
(cos 2π
k − sin 2πk
sin 2πk cos 2π
k
)(x1
x2
)Reflexion through the x1-axis:
R2(x) = R2(x1, x2) = (x1,−x2),
Problem:
how does the limit profile Ψp change when the above trasformationsact on p?
how does the coefficient ξp(1) and hence f(α) change when theabove trasformations act on p?
Veronica Felli AB-operators with varying poles September 10, 2015 24 / 27
Sharp asymptotics Moving along any direction
The function ψk (to which the limit profile is asymptotic at ∞ up tosuitable phases) is invariant (up to phases) under the actions R1,R2:
ei2
(θ0R1)(ψk R1) = −eiπk e
i2θ0ψk, ψk R2 = ψk.
⇓
ΨR−11 (p) = −e−i
πk(Ψp R1
), ΨR2(p) = −eiθR2(p)
(Ψp R2
).
⇓
ξR−11 (p)(1) = ξp(1), ξR2(p)(1) = ξp(1),
⇓
f(α) = f(α− 2π
k
)f(α) = f(2π − α)
Veronica Felli AB-operators with varying poles September 10, 2015 25 / 27
Sharp asymptotics Moving along any direction
The function ψk (to which the limit profile is asymptotic at ∞ up tosuitable phases) is invariant (up to phases) under the actions R1,R2:
ei2
(θ0R1)(ψk R1) = −eiπk e
i2θ0ψk, ψk R2 = ψk.
⇓
ΨR−11 (p) = −e−i
πk(Ψp R1
), ΨR2(p) = −eiθR2(p)
(Ψp R2
).
⇓
ξR−11 (p)(1) = ξp(1), ξR2(p)(1) = ξp(1),
⇓
f(α) = f(α− 2π
k
)f(α) = f(2π − α)
Veronica Felli AB-operators with varying poles September 10, 2015 25 / 27
Sharp asymptotics Moving along any direction
The function ψk (to which the limit profile is asymptotic at ∞ up tosuitable phases) is invariant (up to phases) under the actions R1,R2:
ei2
(θ0R1)(ψk R1) = −eiπk e
i2θ0ψk, ψk R2 = ψk.
⇓
ΨR−11 (p) = −e−i
πk(Ψp R1
), ΨR2(p) = −eiθR2(p)
(Ψp R2
).
⇓
ξR−11 (p)(1) = ξp(1), ξR2(p)(1) = ξp(1),
⇓
f(α) = f(α− 2π
k
)f(α) = f(2π − α)
Veronica Felli AB-operators with varying poles September 10, 2015 25 / 27
Sharp asymptotics Moving along any direction
The function ψk (to which the limit profile is asymptotic at ∞ up tosuitable phases) is invariant (up to phases) under the actions R1,R2:
ei2
(θ0R1)(ψk R1) = −eiπk e
i2θ0ψk, ψk R2 = ψk.
⇓
ΨR−11 (p) = −e−i
πk(Ψp R1
), ΨR2(p) = −eiθR2(p)
(Ψp R2
).
⇓
ξR−11 (p)(1) = ξp(1), ξR2(p)(1) = ξp(1),
⇓
f(α) = f(α− 2π
k
)f(α) = f(2π − α)
Veronica Felli AB-operators with varying poles September 10, 2015 25 / 27
Sharp asymptotics Moving along any direction
g : R→ R, g(α) := P (cosα, sinα)
g(α) =∑k
j=0 cj(cosα)k−j(sinα)j
g(α) = |β2|2√πkf(α) ⇒ g(α) = g
(α+ 2π
k
), g(α) = g(2π − α) ∀α ∈ R
c0 = g(0) = −4 |β2|2
π mk > 0
⇓
g(α) = c0 cos(kα) for all α ∈ R.
Veronica Felli AB-operators with varying poles September 10, 2015 26 / 27
Sharp asymptotics Moving along any direction
Theorem [F.-Abatangelo (2015)]
If α ∈ [0, 2π), then
λ0 − λa|a|k
→ C0 cos(kα) as a→ 0 with a = |a|(cosα, sinα),
with C0 = −4 |β2|2
π mk.
The leading term in the Taylor expansion of a 7→ λ0 − λa is then given by
P (|a|(cosα, sinα)) = C0|a|k cos(kα).
Hence P (a1, a2) = C0<((a1 + i a2)k
)and then
the polynomial P is harmonic.
Veronica Felli AB-operators with varying poles September 10, 2015 27 / 27
Sharp asymptotics Moving along any direction
Theorem [F.-Abatangelo (2015)]
If α ∈ [0, 2π), then
λ0 − λa|a|k
→ C0 cos(kα) as a→ 0 with a = |a|(cosα, sinα),
with C0 = −4 |β2|2
π mk.
The leading term in the Taylor expansion of a 7→ λ0 − λa is then given by
P (|a|(cosα, sinα)) = C0|a|k cos(kα).
Hence P (a1, a2) = C0<((a1 + i a2)k
)and then
the polynomial P is harmonic.
Veronica Felli AB-operators with varying poles September 10, 2015 27 / 27