abrikosov lattice solutions of the ginzburg-landau ... · used extensively in di erential geometry....
TRANSCRIPT
-
Abrikosov Lattice Solutions of the Ginzburg-LandauEquations of Superconductivity
by
Tim Tzaneteas
A thesis submitted in conformity with the requirementsfor the degree of Doctor of PhilosophyGraduate Department of Mathematics
University of Toronto
Copyright c© 2010 by Tim Tzaneteas
-
Abstract
Abrikosov Lattice Solutions of the Ginzburg-Landau Equations of Superconductivity
Tim Tzaneteas
Doctor of Philosophy
Graduate Department of Mathematics
University of Toronto
2010
In this thesis we study the Ginzburg-Landau equations of superconductivity, which are
among the basic nonlinear partial differential equations of Theoretical and Mathemat-
ical Physics. These equations also have geometric interest as equations for the section
and connection of certain principal bundles and are related to Seiberg-Witten equations
used extensively in Differential Geometry. In 1957, Abrokosov suggested that for suffi-
ciently high magnetic fields there exist solutions for which all physical quantities have
the periodicity of a lattice, with the magnetic field penetrating the superconductor at the
vertices of the lattice (Abrikosov lattice solutions). The corresponding phenomenon was
confirmed experimentally and is among the most interesting aspects of superconductivity
and is discussed in every book on the subject. In 2003, Abrikosov was awarded the Nobel
Prize in Physics for this discovery.
Building on the previous results in the subject we prove the existence of such lattices
in the case where each lattice cell contains a single quantum of magnetic flux, and in the
general case reduce the problem to an n-dimensional problem, where n is the number of
quanta of flux. We prove that for Type II superconductors, these solutions are stable,
and in the case n = 1, we show that as the external magnetic field approaches the critical
value at which superconductivity first appears, the lattice which minimizes the average
free energy per lattice cell is the triangular lattice.
ii
-
Contents
1 Introduction 1
2 Existence of Abrikosov Lattice Solutions 10
2.1 Abrikosov Lattice States . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2 Fixing the Gauge and Rescaling . . . . . . . . . . . . . . . . . . . . . . . 16
2.3 Asymptotics of solutions to (2.4) . . . . . . . . . . . . . . . . . . . . . . 19
2.4 Reformulation of the problem . . . . . . . . . . . . . . . . . . . . . . . . 22
2.5 Reduction to a finite-dimensional problem . . . . . . . . . . . . . . . . . 24
2.6 Bifurcation theorem for n = 1 . . . . . . . . . . . . . . . . . . . . . . . . 27
2.7 Proof of Theorem 8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.8 Bifurcation with Symmetry . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.9 The Operators Ln and M . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.10 Fixing the Gauge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3 Stability of Abrikosov Lattice Solutions 38
3.1 Gorkov-Eliashberg-Schmidt equations . . . . . . . . . . . . . . . . . . . . 38
3.2 Linearized stability of static solutions . . . . . . . . . . . . . . . . . . . . 40
3.3 Review of existence results . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.4 Explicit form of L and its properties . . . . . . . . . . . . . . . . . . . . 42
3.5 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.6 Gauge Fixing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
iii
-
3.7 Complexification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.8 Proof of Theorem 23 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.9 Bloch Decomposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.10 Proof of Theorem 24 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
Bibliography 63
iv
-
Chapter 1
Introduction
This thesis deals with the Ginzburg-Landau equations for superconductivity on R2
∆AΨ = κ2(|Ψ|2 − 1)Ψ, (1.1a)
curl∗ curlA = Im(Ψ̄∇AΨ). (1.1b)
Here Ψ : R2 → C and A : R2 → R2. ∇A := ∇− iA and ∆A = −∇∗A∇A are the covariant
gradient and covariant Laplacian. curl and curl∗ are given by curlA = ∂x1A2 − ∂x2A1
and curl∗ f = (∂x2f,−∂x1f). κ is a positive constant.
The Ginzburg-Landau equations first arose in the Ginzburg-Landau model of super-
conductivity, introduced by Ginzburg and Landau in 1950 [19] (reviews can be found
in any text on superconductivity, e.g. [37, 42, 43]). It gives a macroscopic description
of a superconducting material in terms of a complex-valued order parameter Ψ where
ns = |Ψ|2 gives the local density of (Cooper pairs of) superconducting electrons, and a
vector field A where B = curlA is the magnetic field. The vector quantity J = Im(Ψ̄∇AΨ)
is the superconducting current. The parameter κ depends on the material properties of
the superconductor. For the Ginzburg-Landau equations on R2 the underlying geometry
is a superconductor that fills all space but is homogeneous in one direction, in which case
the original equations on R3 reduce to equations on R2.
1
-
Chapter 1. Introduction 2
The equations also arise in particle physics as the Abelian-Higgs model, which is the
simplest and perhaps most important ingredient of the standard model. Here Ψ and A
are the Higgs and U(1) gauge (electromagnetic) fields.
The Ginzburg-Landau equations are the Euler-Lagrange equations of the critical
points of the Ginzburg-Landau energy functional,
E(Ψ, A) = 12
∫R2|∇AΨ|2 + | curlA−H|2 +
κ2
2(1− |Ψ|2)2.
For superconductors, the functional E represents the difference in the Helmholtz free
energy of the superconducting and normal states, near the transition temperature, in the
presence of an external magnetic field H. Alternatively, taking H = 0, one can view E
as the difference in the Gibbs free energy under the constraint that the average magnetic
flux, given by
Φ =
∫R2
curlA,
is fixed. For particle physics, the functional represents the energy of a static configuration
in the U(1) Yang-Mills-Higgs classical gauge theory.
The Ginzburg-Landau equations admit several symmetries, i.e., transformations which
map solutions to solutions. The most important of these is gauge symmetry: for any suf-
ficiently regular function η : R2 → R, Ψ 7→ eiηΨ, A 7→ A + ∇η. Pairs (Ψ, A) that are
gauge equivalent represent the same physical state. There is also the translation sym-
metry: for any t ∈ R2, Ψ 7→ Ψ(x + t), A 7→ A(x + t), and the rotation and reflection
symmetry: for any R ∈ O(2), Ψ 7→ Ψ(Rx), A 7→ R−1A(Rx). This infinite dimensional
symmetry group is one of the most interesting aspects that arise in the analysis of the
Ginzburg-Landau equations.
An important property the Ginzburg-Landau equations is the quantization of mag-
netic flux. Finite energy states (Ψ, A) are classified by their topological degree (the
winding number of Ψ at ∞),
deg(Ψ) := degree
(Ψ
|Ψ|
∣∣∣∣|x|=R
: S1 → S1),
-
Chapter 1. Introduction 3
for R � 1. This is well defined because it clear from the definition of E that for a state
to have finite energy, |Ψ| → 1 as |x| → ∞. For each such state we have the quantization
of magnetic flux,
Φ = 2π deg(Ψ) ∈ 2πZ,
which by Stokes theorem and the requirement that |Ψ| → 1 and |∇AΨ| → 0 as |x| → ∞.
For Abrikosov lattice states (see below) the energy is infinite, but the flux quantization
still holds for each lattice cell because of gauge-periodic boundary conditions.
The simplest solutions to the Ginzburg-Landau equations are the trivial ones corre-
sponding to physically homogeneous states:
1. The perfect superconductor solution: Ψ ≡ 1 and A ≡ 0 (so the magnetic field
B = 0,
2. The normal metal solution, where Ψ = 0 and the magnetic field B is constant
(as well, of course, as any gauge transformation of one of these solutions). The perfect
superconductor is a solution only when the magnetic flux Φ = 0. On the other hand,
there is a normal solution for any value of Φ, where Φ is then the strength of the applied
external magnetic field.
Solving the Ginzburg-Landau equations near a flat interface between the normal and
superconducting states shows that (in the units used here) the magnetic field varies
on a length scale of 1, the penetration depth, while the order parameter varies on a
length scale of 1/mκ, the coherence length, where mκ = min(κ√
2, 2) [14]. The two
length scales coincide when κ2 = 1/2. Considering a flat interface between the normal
and superconducting states, one can show easily that at this point the surface tension
changes sign from positive for κ2 < 1/2 to negative for κ2 > 1/2. This critical value
κ2 = 1/2 separates superconductors into two classes with different properties:
1. κ2 < 1/2: Type-I superconductors exhibit first-order (discontinuous) phase transi-
tions from the normal state to the superconducting state.
-
Chapter 1. Introduction 4
2. κ2 > 1/2: Type-II superconductors exhibit second-order (continuous) phase tran-
sitions and the formation of vortex lattices.
One of the most interesting mathematical and physical phenomenon connected with
Ginzburg-Landau equations is the presence of vortices in solutions. Roughly speaking, a
vortex is a spatially localized structure in the solution, around which the order parameter
has a nontrivial winding. In a superconductor, a vortex represents a localized defect
where the normal state intrudes, and magnetic flux penetrates [20]. In the last decade
or so, vortex solutions have become the object of intense mathematical study in several
directions. One direction is to consider the singular limit (extreme Type II) κ → ∞
(on a bounded domain), in which vortices become point defects whose locations are
determined by some reduced finite-dimensional problem (see, for example, the books
of Bethuel-Brezis-Hélein [11] for a model problem without magnetic field and Serfaty-
Sandier [38]).
In the self-dual case κ2 = 1/2, vortices effectively become non-interacting, and there
is a rich multivortex solution family. Bogomolnyi [12] found the topological energy lower
bound,
E(Ψ, A)|κ2=1/2 ≥ π| deg(Ψ)|,
and showed that this bound is saturated (and hence the Ginzburg-Landau equations are
solved) when certain first-order equations are satisfied. The mathematical implications of
this self-duality, and consequent reduction to a first-order equations, were worked out by
Taubes [41] who showed that for a given degree n, the family of solutions modulo gauge
transformations is 2|n|-dimensional, and the 2|n| parameters describe the locations of the
zeros of the scalar field, i.e., the vortex centres. A review of this theory can be found in
the book of Jaffe and Taubes [24].
A model for a vortex is given, for each degree n ∈ Z, by a “radially symmetric” (or
more precisely equivariant) solution of the Ginzburg-Landau equations of the form
Ψ(n)(x) = fn(r)einθ, A(n)(x) = an(r)∇(nθ),
-
Chapter 1. Introduction 5
where (r, θ) are the polar coordinates of x ∈ R2. Note that deg(Ψ(n) = n. The
pair (Ψ(n), A(n) is called the n-vortex (magnetic or Abrikosov in superconductivity, and
Nielsen-Olesen or Nambu string in particle physics). For superconductors, this is a mixed
state with the normal phase residing at the point where the vortex vanishes [25]. The
existence of such solutions of the Ginzburg-Landau equations was already noticed by
Abrikosov [1]. The n-vortex solution exhibits the length scales discussed above. Indeed,
the asymptotics for the field components of the n-vortex are [24, 34]
J (n)(x) = nβnK1(r)[1 + o(e−mκr)]Jx̂
B(n)(r) = nβnK1(r)[1− 12r +O(1/r2)]
|1− fn(r)| ≤ ce−mκr
|f ′n(r)| ≤ ce−mκr,
as |x| → ∞, where J (n) = Im(Ψ(n)∇A(n)Ψ(n)) is the n-vortex supercurrent, B(n) = curlA(n)
is the n-vortex magnetic field, βn > 0 is a constant, and K1 is the modified Bessel function
of order 1 of the second kind. The length scale of Ψ(n) is 1/mκ. Since K1(r) behaves like
e−r/√r for large r, we see that the length scale for J (n) and B(n) is 1.
The n-vortex is a critical point of the Ginzburg-Landau energy E , and the second
variation of the energy,
L(n) := Hess E(Ψ(n), A(n)),
is the linearized operator for the Ginzburg-Landau equations around the n-vortex, acting
on the space X = L2(R2,C)⊕L2(R2; R2) [29]. The symmetry group of E , which is infinite
dimensional due to gauge transformations, gives rise to an infinite-dimensional subspace
L(n) ⊂ X, which we denote here by Zsym. We say the n-vortex is (linearly) stable if for
some c > 0,
L(n)∣∣Z⊥sym
≥ c,
and unstable if L(n) has a negative eigenvalue. By this definition, a stable state is a local
energy minimizer which is a strict minimizer in directions orthogonal to the infinitesimal
-
Chapter 1. Introduction 6
symmetry transformations. An unstable state is an energy saddle point. The basic result
on vortex stability is the following [21]:
1. For Type-I superconductors, all n-vortices are stable.
2. For Type-II superconductors, the 1-vortices are stable, while n-vortices with |n| ≥ 2
are unstable.
This stability behaviour was long conjectured [24], based on numerical computations [35],
leading to a “vortex interaction” picture in which intervortex interactions are always
attractive in the Type-I case, but become repulsive for like-signed vortices in the Type-II
case [33]. This result agrees with the fact, mentioned above, that the surface tension is
positive for κ2 < 1/2 and negative for κ2 > 1/2, so the vortices try to minimize their
“surface” for κ2 < 1/2 and maximize it for κ2 > 1/2.
In 1957, Abrikosov predicted the existence of states of Type-II superconductors ex-
hibiting vortices arrayed in a lattice pattern, now called Abrikosov lattices, within the
Ginzburg-Landau theory [1]. (Due to a calculation error, Abrikosov concluded that the
lattice which gives the minimum energy is the square lattice. The error was corrected by
Kleiner et al. [27] who showed that it is, in fact, the triangular (also known as the hexago-
nal) lattice which minimizes the energy.) These lattice were later observed experimentally
and have since played an important role in experimental work in superconductivity as
well as having been the study of theoretical works (of the more mathematical studies,
we mention the articles of Eilenberger [18] and Lasher [28]). In 2003, Abrikosov received
the Nobel Prize for this discovery.
The rigorous investigation of Abrikosov solutions began soon after their discovery.
Odeh [32] proved the existence of nontrivial minimizers and obtained a result concerning
the bifurcation of solutions at the critical field strength. Barany et al. [10] investigated
this bifurcation for certain lattices using equivariant bifurcation theory, and Takáč [40]
adapted these results to study the zeros of the bifurcating solutions. Further results were
-
Chapter 1. Introduction 7
first proved in [15]. Except for a variational result of [32] (see also [15]), work done by
both physicists and mathematicians has followed the general strategy of [1].
Among related results is a relation of the Ginzburg-Landau minimization problem,
for a fixed, finite domain and for increasing Ginzburg-Landau parameter κ2 and external
magnetic field, to the Abrikosov lattice variational problem [4, 6]. Boundaries between
superconducting, normal, and mixed phases have also been found [16, 15].
In Chapter 2 of this thesis we combine and extend the previous techniques to give
a complete and self-contained proof of the existence of Abrikosov lattice solutions. To
formulate our results we mention that lattices L ⊂ R2 are characterized by the area |L|
of the fundamental lattice cell ΩL (for details see Chapter 2). We will prove the following
result, whose precise formulation will be given below (Theorem 8).
Theorem 1. Let L be a lattice with∣∣|ΩL| − 2π
κ2
∣∣� 1.(I) If |ΩL| > 2π
κ2, then there exists an L-lattice solution, (ΨL, AL). If |ΩL| ≤ 2π
κ2, then
there is no L-lattice solution in a neighbourhood of the branch of normal solutions.
(II) The solution (ΨL, AL) is close to the branch of normal solutions and is unique, up
to symmetry, in a neighbourhood of this branch.
(III) The solution (ΨL, AL) is real analytic in |ΩL| in a neighbourhood of 2πκ2
.
(IV) The lattice shape for which the average energy per lattice cell is minimized ap-
proaches the triangular lattice as |ΩL| → 2πκ2
.
Remark 2.
(a) [32, 16] showed that for all |ΩL| > 2πκ2
there exists a global minimizer of EΩL .
(b) [32, 10] proved results related to the first part of (I).
(c) [28] proved partial results on (IV), which is generalization of an earlier result of
[27] showing that the triangular lattice gives the minimum energy for the linearized
problem.
-
Chapter 1. Introduction 8
All the rigorous results above deal with Abrikosov lattices with one quantum of mag-
netic flux per lattice cell. Partial results for higher magnetic fluxes were proven in [13, 7]
In Chapter 3 of the thesis we present our results concerning the stability of Abrikosov
lattice solutions within the framework of the Gorkov-Eliashberg-Schmidt time-dependent
Ginzburg-Landau equations. In particular we consider the linearized stability in terms of
the Hessian of the Ginzburg-Landau equations. More precisely we consider the quadratic
form induced by E ′′GL on two classes of perturbations and say that the solution is linearly
stable if this quadratic form is positive.
We recall that the underlying geometry of the superconductor of the equations on
R2 is a superconductor that fills all space but is homogeneous in one direction, and so,
vortex solutions correspond to vortex lines in the superconductor. In general, therefore,
therefore the perturbations we consider are three-dimensional. For Type-II superconduc-
tors, however, vortices repel and therefore any solution with vortex lines that are not
straight will be unstable, as the forces acting on the vortex lines will balance out only
when the lines are straight. Limiting ourselves then to solutions of straight vortex lines,
the homogeneity once again reduces the problem to a problem on R2.
We first consider perturbations that exhibit the same double periodicity as the lattice
solutions themselves. For Type-II superconductors we prove the following result showing
that all Abrikosov lattice solutions found in Theorem 1 that are sufficiently close to the
normal solution are linearly stable under such perturbations.
Theorem 3. If κ2 > 12, then for L such that |ΩL| is sufficiently close to 2π
κ2, then
〈v, E ′′GL(ΨL, AL)v
〉> 0
for all L-lattice states v ⊥ ZL, where ZL is the infinite subspace of zero mode arising
from the gauge symmetry.
We also consider perturbations that have finite total energy. In this case we prove
that there exists a constant SLκ , for which we give an explicit expression, that determines
-
Chapter 1. Introduction 9
the instability of the Abrikosov lattice solutions, i.e., the Abrikosov solution is linearly
stable if and only if SLκ > 0 (we refer to Theorem 24 for a precise formulation of this
result). We are confident that we will be able to determine the sign of this constant for
Type-I and Type-II superconductors in the near future. Experimentally it is known that
triangular Abrikosov lattices exist for Type-II superconductors. Hence we expect that
SLκ > 0 in this case.
-
Chapter 2
Existence of Abrikosov Lattice
Solutions
The main goal of this chapter is to prove Theorems 1.
2.1 Abrikosov Lattice States
We begin by giving a mathematical definition of an Abrikosov lattice state and by dis-
cussing their basic properties. We recall that intuitively an Abrikosov lattice represents
a superconductor whose physical properties are doubly periodic.
2.1.1 Lattices in R2
We first define lattices and introducing the terminology we will be using. A (Bravais)
lattice L is a subset of R2 with the following properties.
• L is discrete, i.e., it has no finite limit points.
• L is a subgroup of R2 as an additive group.
• L is not contained in any proper vector subspace of R2.
10
-
Chapter 2. Existence of Abrikosov Lattice Solutions 11
These properties imply that L is the set of points of the form
L = {m1t(1) +m2t(2) : m1,m2 ∈ Z }
for some linearly independent vectors t(1), t(2) ∈ R2, called a basis of L.
A cell of the lattice L is any parallelogram whose sides are elements of the lattice,
i.e., a set Ω of the form
Ω = { x+ pt+ p′t′ : 0 ≤ p, p′ ≤ 1 },
where x ∈ R2 and t, t′ ∈ L are linearly independent. We denote the area of Ω by |Ω| and
remark that |Ω| = |t ∧ t′|, where the wedge product is defined by x ∧ y = x1y2 − x2y1.
Any lattice has a non-zero minimal cell area and we denote this by |L|, which we
consider as a measure of the size of the lattice.
To define the shape of the lattice, we identify R2 with C via the map x 7→ x1 + ix2
and view L ⊂ C. It is well-known (see [5]) that any lattice L ⊆ C can be given a basis
t(1), t(2) such that the ratio τ = t(2)
t(1)satisfies the inequalities:
• |τ | ≥ 1.
• Im τ > 0.
• −12< Re τ ≤ 1
2, and Re τ ≥ 0 if |τ | = 1.
(In effect, this means τ is in the fundamental domain of the modular group acting on the
upper halfspace.) Although the basis is not unique, the value of τ is, and we take τ as a
measure of the shape of the lattice L.
Now, given a function f on R2 that is L-periodic, i.e., f(x + t) = f(x) for all t ∈ L,
we define the average per lattice cell of f , 〈f〉L, to be
〈f〉L :=1
|Ω|
∫Ω
f(x) d2x,
where Ω is any cell of L. It can easily be checked the right hand side of this definition is
independent of the choice of Ω.
-
Chapter 2. Existence of Abrikosov Lattice Solutions 12
2.1.2 Abrikosov Lattice State
An Abrikosov lattice state is a pair (Ψ, A) ∈ H1loc(R2; C) × H1loc(R2; R2) for which there
exists a lattice L and a family of functions gt ∈ H2loc(R2; R), t ∈ L, such thatΨ(x+ t) = eigt(x)Ψ(x),
A(x+ t) = A(x) +∇gt(x).(2.1)
(We will refer to such pairs as (Abrikosov) L-lattice states when it necessary to make
L explicit.) A lattice state is therefore a state whose translations are mutually gauge
equivalent. As gauge equivalent states represent the same physical state, this definition
captures the idea of a superconductor whose physical properties are doubly periodic. In
particular we note that the superconducting charge density |Ψ|2, the superconducting
current Im(Ψ̄∇AΨ), and the magnetic field curlA are all L-periodic as can be easily
verified.
2.1.3 Symmetries
We define a symmetry of Abrikosov lattices to be a group action on the set of Abrikosov
lattices that preserves the property of being a solution of the Ginzburg-Landau equations.
In particular we do not impose the requirement that a symmetry preserve the underlying
lattice. We list the most important symmetries.
1. Gauge symmetry: given η ∈ H2loc(R2; R), we define the map Tη to be
(TηΨ(x), TηA(x)) = (eiη(x)Ψ(x), A(x) +∇η(x)).
We note that Tη maps Abrikosov L-lattices to Abrikosov L-lattices.
2. Translation symmetry: given y ∈ R2, we define the map Ty to be
(TyΨ(x), TyA(x)) = (Ψ(x+ y), A(x+ y)).
Again, Ty maps Abrikosov L-lattices to Abrikosov L-lattices.
-
Chapter 2. Existence of Abrikosov Lattice Solutions 13
3. Rotation and reflection symmetry: given R ∈ O(R2), the group of orthogonal
matrices on R2, we define the map TU to be
(TRΨ(x), TRA(x)) = (Ψ(Rx), R−1A(Rx)).
In this case TU maps Abrikosov L-lattices to Abrikosov R−1L-lattices.
2.1.4 Energy of Abrikosov Lattices
Because the total energy of Abrikosov lattices is infinite we will instead consider the
average energy per cell, EL, defined to be
EL(Ψ, A) :=〈
1
2|∇AΨ|2 +
κ2
4(1− |Ψ|2)2 + | curlA|2
〉L.
We note that all the symmetries preserve the average energy per cell.
We begin by proving that the energy functional is well-defined for Abrikosov lattices.
Proposition 4. For any Abrikosov L-lattice (Ψ, A), EL(Ψ, A)
-
Chapter 2. Existence of Abrikosov Lattice Solutions 14
In the case where Ψ is non-zero on the boundary of a cell Ω with |Ω| = |L|, we can
define ϕ on ∂Ω by the relation Ψ = |Ψ|eiϕ. Then we have Im(Ψ̄∇AΨ) = |Ψ|2(∇ϕ − A),
which implies that ∇ϕ− A is L-periodic. Using Stoke’s theroem we then calculate
〈curlA〉L =1
|Ω|
∮∂Ω
A · ds = 1|Ω|
∮∂Ω
∇ϕ · ds = 2πn|L|
,
where the final step follows from the fact that Ψ is single-valued. This argument demon-
strates the relation between the average magnetic flux and the index of Ψ, and therefore
between the flux and the number of zeros of Ψ per cell.
For the general case, however, we need a more indirect proof. We begin with the
following lemma.
Lemma 6. Let (Ψ, A) be an Abrikosov L-lattice such that Ψ 6≡ 0. For t, t′ ∈ L, define
Kt,t′ by the formula
Kt,t′ = gt(x+ t′)− gt(x)− gt′(x+ t) + gt′(x),
where x is such that Ψ(x) 6= 0. Then Kt,t′ is independent of the choice of x and there
exists n ∈ Z such that
Kt,t′ =2πn
|L|t ∧ t′.
Proof. Fix t and t′ ∈ L. Using (2.1), we have the relationsΨ(x+ t+ t′) = eigt(x+t
′)eigt′ (x)Ψ(x),
Ψ(x+ t+ t′) = eigt′ (x+t)eigt(x)Ψ(x).
Therefore for any x such that Ψ(x) 6= 0, we must have gt(x+t′)−gt(x)−gt′(x+t)+gt′(x) =
2πnt,t′(x), for some nt,t′(x) ∈ Z. From (2.1) we also have the relations thatA(x+ t+ t′) = A(x) +∇gt(x+ t′) +∇gt′(x),
A(x+ t+ t′) = A(x) +∇gt′(x+ t) +∇gt(x).
Therefore we have ∇(gt(x+ t′)− gt(x)− gt′(x+ t) + gt′(x)) = 0, which means that nt,t′(x)
is independent of x and therefore Kt,t′ = 2πnt,t′(x) is well-defined.
-
Chapter 2. Existence of Abrikosov Lattice Solutions 15
We now claim that Kt,t′ is an anti-symmetric bilinear form on L taking values in 2πZ.
This can easily be checked using (2.1) where necessary. Now since t∧ t′ is the signed area
of the cell with sides t and t′, 2π|L|t ∧ t′ ∈ 2πZ, and one can easily check that this defines
anti-symmetric bilinear form on L. But the group of such forms is isomorphic to Z, so
there exists n ∈ Z such that
Kt,t′ =2πn
|L|t ∧ t′,
and the lemma is proven.
We now complete the proof of quantization of flux.
Proof of Proposition 5. Since Ψ 6≡ 0 we can find x such that Ψ(x) 6= 0. Let Ω be the cell
with the bottom-left corner at x and sides parallel to vectors t and t′ from L, labelled so
that |Ω| = t ∧ t′. Using (2.1), a simple calculation gives that∫Ω
curlA dx =
∮∂Ω
A · ds = gt(x+ t′)− gt(x)− gx+t′(t) + gt′(x) = Kt,t′ .
Applying the previous lemma we therefore have
〈curlA〉L =1
|Ω|
∫Ω
curlA dx =1
|t ∧ t′|2πn
|L|t ∧ t′ = 2πn
|L|,
which proves the proposition.
One can check that the gauge and translation symmetries preserve the average flux
per cell but for R ∈ O(R2), we have
〈curlTRA〉R−1L = (detR) 〈curlA〉L .
We note that under the constant 〈A〉L = b, the flux quantization leads to a relation
between the parameter b and the size of the lattice |L|.
2.1.6 Reduction to a single cell
An important property of lattice states is that they are defined by their restriction to a
single cell and can be reconstructed from this restriction using the lattice translations.
-
Chapter 2. Existence of Abrikosov Lattice Solutions 16
2.2 Fixing the Gauge and Rescaling
In this section we fix the gauge for solutions, (Ψ, A), of (1.1) and then rescale them to
eliminate the dependence of the size of the lattice on b. Our space will then depend only
on the number of quanta of flux and the shape of the lattice.
The symmetries allows one to fix solutions to be of a desired form. We first use the
rotation and reflection symmetry to assume that b > 0 and that the lattice L has a basis
of the form
t(1) = r
10
, t(2) = r Re τ
Im τ
,where r > 0. From now we let Ω be the cell with side t(1) and t(2). We note that
|Ω| = r2 Im τ.
We will also use the following preposition to fix the gauge symmetry. This was
introduced by [32] and proved in [40]. We provide an alternate proof in Section 2.10.
Proposition 7. Let (Ψ, A) be an L-lattice state, and let b be the average magnetic flux
per cell. Then there is a L-lattice state (φ,Ab0+a) that is gauge-equivalent to a translation
of (Ψ, A), such that Ab0(x) =b2Jx, where Jx = x⊥ := (−x2, x1), and φ and a satisfy the
following conditions:
(i) a is doubly periodic with respect to L: a(x+ t) = a(x) for all t ∈ L;
(ii) a has mean zero: 〈a〉L = 0;
(iii) a is divergence-free: div a = 0;
(iv) φ(x+ t) = eib2t∧xφ(x), for t = t(1), t(2).
Suppose now that we have a L-lattice state (Ψ, A). Let b be the average magnetic
flux per cell of the state and n the quanta of flux per cell. From the quantization of the
-
Chapter 2. Existence of Abrikosov Lattice Solutions 17
flux, we know that
b =2πn
|Ω|=
2πn
r2 Im τ,
We set σ :=(nb
) 12 . The last two relations give σ =
(Im τ2π
) 12 r. We now define the rescaling
(ψ, α) to be
(ψ(x), α(x)) := (σΨ(σx), σA(σx)).
Let Lτ be the lattice spanned by t(1) and t(2) as above with r = rτ , where
rτ :=
(2π
Im τ
) 12
. (2.2)
We let Ωτ be the cell with sides t(1) and t(2). We note that |Ωτ | = 2πn. We summarize
the effects of the rescaling above:
(i) (ψ, α) is a Lτ -lattice state.
(ii) 1|Ωτ |ELτ (ψ,A) = Eλ(ψ, α), where λ =κ2nb
and
Eλ(ψ, α) =κ4
2πλ2
∫Ωτ
(|∇αψ|2 + | curlα|2 +
κ2
2(|ψ|2 − λ
κ2)2)dx. (2.3)
(iii) Ψ and A solve the Ginzburg-Landau equations if and only if ψ and a solve the
rescaled Ginzburg-Landau equations
(−∆α − λ)ψ = −κ2|ψ|2ψ, (2.4a)
curl∗ curlα = Im{ψ̄∇αψ} (2.4b)
for λ = κ2nb
. The latter equations are valid on Ωτ with the boundary conditions
given in the next statement.
(iv) If (Ψ, A) is of the form described in Proposition 7, then
α = An0 + a, where An0 (x) :=
n
2Jx, (2.5)
where ψ and a satisfy
-
Chapter 2. Existence of Abrikosov Lattice Solutions 18
(a) a is double periodic with respect to Lτ ,
(b) 〈a〉Lτ = 0,
(c) div a = 0,
(d) ψ(x+ t) = ein2t∧xψ(x) for t = t
(1)τ , t
(2)τ .
We now introduce the spaces Hn(τ) and~H (τ) as follows. We define the Hilbert space
Ln(τ) to be the closure under the L2-norm of the space of all smooth ψ on Ωτ satisfying
the quasiperiodic boundary condition (d) in part (iv) above. Hn(τ) is then the space of
all ψ ∈ Ln(τ) whose (weak) partial derivatives up to order 2 are square-integrable.
Similarly, we define the Hilbert space ~L (τ) to be the closure of the space of all smooth
a on Ωτ that satisfy periodic boundary conditions, have mean zero, and are divergence
free, and ~H (τ) is then the subspace of ~L (τ) consisting of those elements whose partial
derivatives up to order 2 are square-integrable.
Our problem then is, for each n = 1, 2, . . ., find (ψ, a) ∈ Hn(τ) × ~H (τ) such that
(ψ,An0 + a) solves the rescaled Ginzburg-Landau equations (2.4), and among these find
the one that minimizes the average energy Eλ.
We will prove the following theorem for the case n = 1
Theorem 8.
(I) For every b sufficiently close to but less than the critical value b = κ2 and ev-
ery lattice shape τ , there exists an Lτ−lattice solution, (Ψτλ, Aτλ) of the rescaled
Ginzburg-Landau equations with one quantum of flux per cell.
(II) This solution is unique, up to the symmetries, in a neighbourhood of the normal
solution.
(III) The family of these solutions is real analytic in b in a neighbourhood of bc.
-
Chapter 2. Existence of Abrikosov Lattice Solutions 19
(IV) If κ2 > 1/2, then the global minimizer Lb of the average energy per cell, E(L) ≡1|Ω|EΩL(Ψ
Lb , A
Lb ), approaches the Ltriangular as b→ bc in the sense that the shape τb
approaches τtriangular = eiπ/3 in C.
2.3 Asymptotics of solutions to (2.4)
In this section, assuming that equations (2.4), have a family, (ψ�, a�, λ�), of solutions
depending on a small parameter � > 0, we establish some asymptotic properties of such
a family. These properties will be needed below. Most of the results of this section were
first stated in [1] (see also [13]). The main result of this section is the following
Proposition 9. If the equations (2.4) have a family, (ψ�, a�, λ�), �→ 0, of solutions of
the form
ψ� = �ψ0 +O(�3), a� = �
2a1 +O(�4), λ� = n+ �
2λ1 +O(�4), (2.6)
then ψ0 and a1 satisfy the equations
−∆An0ψ0 = nψ0, and curl a1 = H −1
2|ψ0|2, with H :=
1
2〈|ψ0|2〉, (2.7)
where 〈f〉 stands for the average of a function f over the lattice cell Ωτ . Furthermore,
we have
Eλ�(ψ�, α�) =κ2
2+n2κ4
λ2− κ
4λ21�4
2λ2((κ2 − 1
2)β(ψ0) +
12
) +O(�6), (2.8)where α� := A
n0 + a� and β(ψ0) is the Abrikosov function, which is defined by
β(ψ0) :=
∫Ωτ|ψ0|4(∫
Ωτ|ψ0|2
)2 . (2.9)Proof. Plugging (2.6) into (2.4) and taking �→ 0 gives the first equation in (2.7) and
curl∗ curl a1 = Im(ψ̄0∇An0ψ0). (2.10)
We show now that
Im(ψ0∇An0ψ0) = −1
2curl∗ |ψ0|2. (2.11)
-
Chapter 2. Existence of Abrikosov Lattice Solutions 20
(Recall, that for a scalar function, f(x) ∈ R, curl∗ f = (∂2f,−∂1f) is a vector.) It is
easy to see (see (2.47), Section 2.9) that ψ0 satisfies the first order equation((∇An0 )1 + i(∇An0 )2
)ψ0 = 0. (2.12)
Multiplying this relation by ψ̄0, we obtain ψ̄0(∇An0 )1ψ0 + iψ̄0(∇An0 )2ψ0 = 0. Taking
imaginary and real parts of this equation gives
Im ψ̄0(∇An0 )1ψ0 = −Re ψ̄0(∇An0 )2ψ0 = −∂x2ψ0
and
Im ψ̄0(∇An0 )2ψ0 = Re ψ̄0(∇An0 )1ψ0 = ∂x1ψ0,
which, in turn, gives (2.11).
The equations (2.10) and (2.11) give the second equation in (2.7), with H a constant
of integration. H has to be chosen so that∫ τ
Ωcurl a1 = 0, which gives the third equation
in (2.7).
Lemma 10.
(−λ1 +H)〈|ψ0|2〉+(κ2 − 1
2
)〈|ψ0|4〉 = 0 (2.13)
and
Eλ�(ψ�, α�) =κ2
2+n2κ4
λ2− κ
4λ12λ2
�4〈|ψ0|2〉+O(�6). (2.14)
Proof. Now we prove (2.13). We multiply the equation (2.4a) scalarly (in L2(Ωτ )) by
ψ0, use that the operator −∆A is self-adjoint and (−∆A − n)ψ0 = 0, substitute the
expansions (2.6) and take � = 0, to obtain
−λ1∫
Ωτ|ψ0|2 + 2i
∫Ωτψ̄0a1 · ∇An0ψ0 + κ
2
∫Ωτ|ψ0|4 = 0. (2.15)
This expression implies that the imaginary part of the second term on the l.h.s. of
(2.15) is zero. (We arrive at the same conclusion by integrating by parts and using that
div a1 = 0.) Therefore
2i
∫Ωτψ̄0a1 · ∇An0ψ0 = −2
∫Ωτa1 · Im(ψ̄0∇An0ψ0) = −2
∫Ωτa1 · curl∗ curl a1.
-
Chapter 2. Existence of Abrikosov Lattice Solutions 21
Integrating the last term by parts, we obtain 2i∫
Ωτψ̄0a1 ·∇An0ψ0 = −2
∫Ωτ
(curl a1)2. Using
this equation and the second equation in (2.7), we obtain
2i
∫Ωτψ̄0a1 · ∇An0ψ0 = −
1
2
∫Ωτ|ψ0|4 +H
∫Ωτ|ψ0|2. (2.16)
This equation together with (2.15) gives (2.13).
Now, we prove the statement (2.14) about the Ginzburg-Landau energy. Multiplying
(2.4a) scalarly by ψ and integrating by parts gives∫Ωτ|∇αψ|2 = κ2
∫Ωτ
(λ|ψ|2 − κ2|ψ|4
).
Substituting this into the expression for the energy and using that |Ωτ | = 2π, we find
Eλ(ψ, α) =κ4
λ2〈 λ
2
2κ2− κ
2
2|ψ|4 + | curlα|2〉, (2.17)
where, recall, 〈f〉 := 1|Ωτ |∫
Ωτf . Using the expansions (2.7) and the facts that curlAn0 = n
and 〈curl a1〉 = 0 gives
Eλ�(ψ�, α�) =κ2
2+n2κ4
λ2+κ4
λ2�4(−κ
2
2〈|ψ0|4〉+ 〈| curl a1|2〉
)+O(�6). (2.18)
Next, using the second equation in (2.7) in the form
curl a1 = −1
2|ψ0|2 +
1
2〈|ψ0|2〉 (2.19)
and substituting it into (2.18), we obtain
Eλ�(ψ�, α�) =κ2
2+n2κ4
λ2+
κ4
2λ2�4(−(κ2 − 1
2)〈|ψ0|4〉 −
1
4〈|ψ0|2〉2
)+O(�6). (2.20)
Finally, using (2.13) and the definition H := 12〈|ψ0|2〉 gives (2.14).
Eqn (2.13), together with the definitions (2.9) and H := 12〈|ψτ0 |2〉 (see (2.7)), implies
λ1〈|ψ0|2〉 =((κ2 − 1
2)β + 1
2
)〈|ψ0|2〉2 . We solve this equation for 〈|ψ0|2〉 to obtain
〈|ψ0|2〉 =λ1
(κ2 − 12)β + 1
2
. (2.21)
This equation together with (2.14) yields (2.8).
-
Chapter 2. Existence of Abrikosov Lattice Solutions 22
2.4 Reformulation of the problem
In this section we reduce two equations (2.4) for ψ and α to a single equation for ψ.
Substituting α = An0 + a, we rewrite (2.4) as
(Ln − λ)ψ + 2ia · ∇An0ψ + |a|2ψ + κ2|ψ|2ψ = 0, (2.22a)
(M + |ψ|2)a− Im(ψ̄∇An0ψ) = 0, (2.22b)
where
Ln := −∆An0 and M := curl∗ curl . (2.23)
The operators Ln and M are elementary and well studied. Their properties that will
be used below are summarized in the following theorems, whose proofs may be found in
Section 2.9.
Theorem 11. Ln is a self-adjoint operator on Hn(τ) with spectrum σ(Ln) = { (2k+1)n :
k = 0, 1, 2, . . . }, each eigenvalue being of (complex) multiplicity n. The lowest eigenvalue
is given explicitly as
null(Ln − n) =
{ein2x2(x1+ix2)
∞∑k=−∞
ckeki√
2π Im τ(x1+ix2) | ck+n = einπτe2kiπτck
}
The infinite series converges for all x ∈ R2 since Im τ > 0 and therefore the ck decay
exponentially as k → ±∞.
Theorem 12. M is a strictly positive operator on ~H (τ) with discrete spectrum.
We first solve the second equation (2.22b) for a in terms of ψ, which we rewrite as
(M + |ψ|2)a− Im(ψ̄∇An0ψ) = 0, (2.24)
using the fact that M is a strictly positive operator on ~H (τ). A naive answer is
a(ψ) = (M + |ψ|2)−1 Im(ψ̄∇An0ψ). (2.25)
-
Chapter 2. Existence of Abrikosov Lattice Solutions 23
A direct derivation of (2.25) is, however, surprisingly subtle, since one has to supplement
(2.24) by the two equation div Ja = 0, 〈Ja〉 = 0, where Ja := Im(ψ̄∇An0 +aψ). We show that
these equations hold for any solution (ψ,An0 + a) of the first Ginzburg-Landau equation.
Differentiating the equation Eλ(eisχψ,An0 +a+ s∇χ) = Eλ(ψ,An0 +a) with respect to s at
s = 0, we obtain ∂ψEλ(ψ,An0 +a)iχψ+∂aEλ(ψ,An0 +a)∇χ = 0. Since ∂ψEλ(ψ,An0 +a) = 0,
this gives
0 =
∫Ω
(Ma− Ja) · ∇χ =∫
Ω
div Jaχ. (2.26)
Since the last equation holds for any χ ∈ H1(Ωτ ,R), we conclude that div Ja = 0.
Choosing χ = h · x, ∀h ∈ R2, in the first equation in (2.26), we find 〈Ja〉 = 0.
Now, (2.24) can be rewritten as a fixed point problem a = M−1Ja, which has a unique
solution in ~H (τ). The latter can be rewritten as (2.25).
We collect the elementary properties of the map a in the following proposition, where
we identify Hn(τ) with a real Banach space using ψ ↔−→ψ := (Reψ, Imψ).
Proposition 13. The unique solution, a(ψ), of (2.22b) maps Hn(τ) to~H (τ) and has
the following properties:
(a) a(·) is analytic as a map between real Banach spaces.
(b) a(0) = 0.
(c) For any α ∈ R, a(eiαψ) = a(ψ).
Proof. The only statement that does not follow immediately from the definition of a is
(a). It is clear that Im(ψ̄∇An0ψ) is real-analytic as it is a polynomial in ψ and ∇ψ, and
their complex conjugates. We also note that (M − z)−1 is complex-analytic in z on the
resolvent set of M , and therefore, (M + |ψ|2)−1 is analytic. (a) now follows.
Now we substitute the expression (2.25) for a into (2.22a) to get a single equation
F (λ, ψ) = 0, where the map F : R×Hn(τ)→ Ln(τ) is defined as
F (λ, ψ) = (Ln − λ)ψ + 2ia(ψ) · ∇An0ψ + |a(ψ)|2ψ + κ2|ψ|2ψ. (2.27)
-
Chapter 2. Existence of Abrikosov Lattice Solutions 24
The following proposition lists some properties of F .
Proposition 14.
(a) F is analytic as a map between real Banach spaces,
(b) for all λ, F (λ, 0) = 0,
(c) for all λ, DψF (λ, 0) = Ln − λ,
(d) for all α ∈ R, F (λ, eiαψ) = eiαF (λ, ψ).
(e) for all ψ, 〈ψ, F (λ, ψ)〉 ∈ R.
Proof. The first property follows from the definition of F and the corresponding ana-
lyticity of a(ψ). (b) through (d) are straightforward calculations. For (e), we calculate
that
〈ψ, F (λ, ψ)〉 = 〈ψ, (Ln − λ)ψ〉+ 2i∫
Ωτψ̄a(ψ) · ∇ψ + 2
∫Ωτ
(a(ψ) · An0 )|ψ|2
+
∫Ωτ|a(ψ)|2|ψ|2 + κ2
∫Ωτ|ψ|4.
The final three terms are clearly real and so is the first because Ln − λ is self-adjoint.
For the second term we calculate the complex conjugate and see that
2i
∫Ωτψ̄a(ψ) · ∇ψ = −2i
∫Ωτψa(ψ) · ∇ψ̄ = 2i
∫Ωτ
(∇ψ · a(ψ))ψ̄,
where we have integrated by parts and used the fact that the boundary terms vanish due
to the periodicity of the integrand and that div a(ψ) = 0. Thus this term is also real and
(e) is established.
2.5 Reduction to a finite-dimensional problem
In this section we reduce the problem of solving the equation F (λ, ψ) = 0 to a finite
dimensional problem. We address the latter in the next section. We use the standard
-
Chapter 2. Existence of Abrikosov Lattice Solutions 25
method of Lyapunov-Schmidt reduction. Let X := Hn(τ) and Y := Ln(τ) and let
K = null(Ln − n). We let P be the Riesz projection onto K, that is,
P := − 12πi
∮γ
(Ln − z)−1 dz, (2.28)
where γ ⊆ C is a contour around 0 that contains no other points of the spectrum of
Ln. This is possible since 0 is an isolated eigenvalue of Ln. P is a bounded, orthogonal
projection, and if we let Z := nullP , then Y = K ⊕ Z. We also let Q := I − P , and so
Q is a projection onto Z.
The equation F (λ, ψ) = 0 is therefore equivalent to the pair of equations
PF (λ, Pψ +Qψ) = 0, (2.29)
QF (λ, Pψ +Qψ) = 0. (2.30)
We will now solve (2.30) for w = Qψ in terms of λ and v = Pψ. To do this, we
introduce the map G : R × K × Z → Z to be G(λ, v, w) := QF (λ, v + w). Applying
the Implicit Function Theorem to G, we obtain a real-analytic function w : R×K → Z,
defined on a neighbourhood of (n, 0), such that w = w(λ, v) is a unique solution to
G(λ, v, w) = 0, for (λ, v) in that neighbourhood. We substitute this function into (2.29)
and see that the latter equation in a neighbourhood of (n, 0) is equivalent to the equation
γ(λ, v) := PF (λ, v + w(λ, v)) = 0 (2.31)
(the bifurcation equation). Note that γ : R × K → C. We have shown that in a
neighbourhood of (n, 0) in R × X, (λ, ψ) solves F (λ, ψ) = 0 if and only if (λ, v), with
v = Pψ, solves (2.31). Moreover, the solution ψ of F (λ, ψ) = 0 can be reconstructed
from the solution v of (2.31) according to the formula
ψ = v + w(λ, v), (2.32)
We note that w and γ inherit the symmetry of the original equation:
-
Chapter 2. Existence of Abrikosov Lattice Solutions 26
Lemma 15. For every α ∈ R, w(λ, eiαv) = eiαw(λ, v) and γ(λ, eiαv) = eiαγ(λ, v).
Proof. We first check that w(λ, eiαv) = eiαw(λ, v). We note that by definition of w,
G(λ, eiαv, w(λ, eiαv)) = 0, but by the symmetry of F , we also haveG(λ, eiαv, eiαw(λ, v)) =
eiαG(λ, v, w(λ, v)) = 0. The uniqueness of w then implies that w(λ, eiαv) = eiαw(λ, v).
We can now verify that
γ(λ, eiαv) = PF (λ, eiαv + w(λ, eiαv)) = eiαPF (λ, v + w(λ, v))〉 = eiαγ(λ, v).
We will also need the following property of w below.
Lemma 16.
w(λ, v) = o(‖v‖3).
Proof. Since w is real-analytic and w(λ, eiαv) = eiαw(λ, v), we have that w(λ, 0) = 0 and
Dvvw(λ, 0) = 0. Now we differentiate the relation QF (λ, v + w(λ, v)) = 0 with respect
to v to obtain
QDψF (λ, v + w(λ, v))(I +Dvw(λ, v)) = 0.
At v = 0, we then have Q(Ln − λ)(I + Dvw(λ, 0)) = 0, and since Q commutes with
Ln, we then have (Ln − λ)Dvw(λ, 0)) = 0. But for λ in a neighbourhood of 1, (Ln − λ)
is invertible on the range of Dvw(λ, 0) and therefore Dvw(λ, 0) = 0. The lemma now
follows.
Solving the bifurcation equation (2.31) is a subtle problem unless n = 1. The latter
case is tackled in section 2.6.
We conclude this section with mentioning an approach to finding solutions to the bifur-
cation equation (2.31) for any n. For a fixed n, we define the first reduced energy Eλ(ψ) :=
Eλ(ψ,A), where A = An0 + a, with An0 (x) := n2x⊥ and a(ψ) = (M + |ψ|2)−1 Im(ψ̄∇An0ψ)
(see (2.5) and (2.25)). Critical points of this energy solve the equation F (λ, ψ) = 0.
Next, we introduce the finite dimensional effective Ginzburg-Landau energy
eλ(v) := Eλ(v + w(λ, v)).
-
Chapter 2. Existence of Abrikosov Lattice Solutions 27
It is a straightforward to show that
(i) eλ(v) has a critical point v0 iff Eλ(u) has a critical point u0 = v0 + w(λ, v0);
(ii) Critical points, v0, of eλ(v) solve the equation (2.29);
(iii) eλ(v) is gauge invariant, eλ(eiαv) = eλ(v). One can further find the leading
behaviour of eλ(v) in v.
2.6 Bifurcation theorem for n = 1
In this section we look at the case n = 1, and look for solutions near the trivial solution.
For convenience we drop the (super)index n = 1 from the notation. We will see that as
b = κ2
λdecreases past the critical value b = κ2, a branch of non-trivial solutions bifurcates
from the trivial solution. More precisely, we have the following result.
Theorem 17. For every τ there exists a branch, (λs, ψs, αs), s ∈ C with |s|2 < � for
some � > 0, of nontrivial solutions of the rescaled Ginzburg-Landau equations (2.4),
unique (apart from the trivial solution (1, 0, A0)) in a sufficiently small neighbourhood of
(1, 0, A0) in R×H (τ)× ~H (τ), and s.t.λs = 1 + gλ(|s|2),
ψs = sψ0 + sgψ(|s|2),
αs = A0 + gA(|s|2),
where (L− 1)ψ0 = 0, gψ is orthogonal to null(L− 1), gλ : [0, �)→ R, gψ : [0, �)→H (τ),
and gA : [0, �) → ~H (τ) are real-analytic functions such that gλ(0) = 0, gψ(0) = 0,
gA(0) = 0 and g′λ(0) > 0. Moreover,
g′λ(0) =
(κ2 − 1
2
) ∫Ωτ|ψ0|4∫
Ωτ|ψ0|2
+1
4π
∫Ωτ|ψ0|2. (2.33)
Proof. The proof of this theorem is a slight modification of a standard result from the
bifurcation theory. It can be found in Section 2.8, Theorem 21, whose hypotheses are
-
Chapter 2. Existence of Abrikosov Lattice Solutions 28
satisfied by F as shown above (see also [32, 10]). The latter theorem gives us a neigh-
bourhood of (1, 0) in R×H (τ) such that the only non-trivial solutions are given byλs = 1 + gλ(|s|2),
ψs = sψ0 + sgψ(|s|2).
Recall that a(ψ) is defined in (2.25). We now define g̃A(s) = a(ψs), which is real-analytic
and satisfies g̃A(−t) = a(−ψt) = g̃A(t), and therefore is really a function of t2, gA(t2).
Hence As = A0 + gA(|s|2).
Finally, (2.33) follows from (2.13) with n = 1 and the relation |Ωτ | = 2π.
Theorem 17 implies (I) - (III) of Theorem 8.
Finally, we mention
Lemma 18. Recall that Im τ > 0. Let (λs, ψs, αs) be the solution branch constructed
above and let mτ =
(√
Im τ)−1
1 Re τ0 Im τ
. Then (λs, ψ̃s, α̃s), where the functions (ψ̃s, α̃s) are defined on aτ -independent square lattice and are given by
ψ̃s(x) = ψs(mτx),
α̃s(x) = Mtταs(mτx),
(2.34)
depend R-analytically on τ .
We sketch the proof of this lemma. The transformation above maps functions on a
lattice of the shape τ into functions on a τ -independent square lattice, but leads to a
slightly more complicated expression for the Ginzburg-Landau equations. Namely, let
Uτψ(x) := ψ(mτx) and Vτa(x) := mtτa(mτx). Applying Uτ and Vτ to the equations
(2.22), we conclude that (ψ̃s, α̃s) satisfy the equations
(Lnτ − λ)ψ + 2i(mtτ )−1a · (mtτ )−1∇An0ψ + |(mtτ )−1a|2ψ + κ2|ψ|2ψ = 0, (2.35a)
-
Chapter 2. Existence of Abrikosov Lattice Solutions 29
(Mτ + |ψ|2)a+ F̃ aτ (ψ) = 0, (2.35b)
where
Lnτ := −Uτ∆An0U−1τ and Mτ := Vτ curl
∗ curlV −1τ . (2.36)
Here we used that VτAn0 = A
n0 and Uτ∇ψ = (mtτ )−1Uτψ. (The latter relation is a
straightforward computation and the former one follows from the facts that for any
matrix m, (mx)⊥ = (detm)(mt)−1x⊥, and that in our case, detmτ = 1.) Note that
the gauge in the periodicity condition will still depend on Im τ . These complications,
however, are inessential and the same techniques as above can be applied in this case.
The important point here is to observe that the function ψ0, constructed in Section B, the
function w(λ, sψ0), where w(λ, v) is the solution of (2.30), and the bifurcation equation
(2.31) depend on τ real-analytically. We leave the details of the proof to the interested
reader.
2.7 Proof of Theorem 8
In this section, we continue with the case n = 1 and prove Theorem 8, which, as was
mentioned above, is a precise restatement of Theorem 1 of Introduction. Theorem 17
implies, after rescaling to the original variables, the statements (I)-(III) of Theorem 8.
It remains to prove the statement (IV).
We fix a lattice shape τ and denote the functions ψ0, ψs and αs given in Theorem
17 by ψτ0 , ψτs and α
τs , respectively. Recall that b =
κ2
λ. Since the function gλ(|s|2) given
in Theorem 17 obeys gλ(0) = 0 and g′λ(0) 6= 0, the function bs = κ2(1 + gλ(|s|2))−1 =:
κ2 +gb(|s|2) can be inverted to obtain |s| = s(b). Absorbing ŝ = s|s| into ψτ0 , we can define
the family (ψτs(b), ατs(b), b
τs(b)) of Lτ -periodic solutions of the Ginzburg-Landau equations
parameterized by average magnetic flux b. Clearly, ψτs(b), ατs(b), b
τs(b) are analytic in b. We
note the relation between the new perturbation parameter µ := κ2−b and the bifurcation
-
Chapter 2. Existence of Abrikosov Lattice Solutions 30
parameter |s|2 :
µ =gλ(|s|2)λ
κ2 = g′λ(0)κ2|s|2 +O(|s|4). (2.37)
Using the real-analyticity of the function gb, gτψ, and g
τA, we can express λ(µ) := κ
2/b,
ψτ (µ) := ψτs(b) and ατ (µ) := ατs(b) as
λ(µ) = 1 +1
κ2µ+O(µ2) (2.38)
ψτ (µ) = µ1/2ψτ0 + µ3/2ψτ1 +O(µ
5/2) (2.39)
ατ (µ) = A0 + µaτ1 +O(µ
2). (2.40)
We identify the expansions (2.38) - (2.40) with the expansion (2.6) of Proposition 9 with
� = µ1/2 (so that λ1 =1κ2
).
Now, we define the energy of the state (ψτ (µ), ατ (µ), λ(µ)):
Eb(τ) := Eλ(µ)(ψτ (µ), ατ (µ)). (2.41)
The equation (2.8) of this proposition together with (2.41) gives for n = 1
Eb(τ) =κ2
2+κ4
λ2− κ
4λ21µ2
2λ2((κ2 − 1
2)β(τ) + 1
2
) +O(µ3), (2.42)where β(τ) ≡ β(ψτ0 ) is of the Abrikosov function,
β(τ) :=
∫Ωτ|ψτ0 |4(∫
Ωτ|ψτ0 |2
)2 . (2.43)Here, recall, ψτ0 is a non-zero element in the nullspace of the operator L
n − 1 acting
on Hn(τ). Since the nullspace is a one-dimensional complex subspace, β depends only
on τ . The next result establishes a relation between the minimizers of the energy and
Abrikosov function.
Theorem 19. In the case κ > 1√2, the minimizers, τb, of τ 7→ Eb(τ) are related to the
minimizer, τ∗, of β(τ), as τb − τ∗ = O(µ1/2), In particular, τb → τ∗ as b→ κ2.
Proof. To prove the theorem we note that Eb(τ) is of the form Eb(τ) = e0 + e1µ +
e2(τ)µ2 + O(µ3). The first two terms are constant in τ , so we consider Ẽb(τ) = e2(τ) +
-
Chapter 2. Existence of Abrikosov Lattice Solutions 31
O(µ). τb is also the minimizer of τ 7→ Ẽb(τ) and τ∗, of e2(τ). We have the expansions
Ẽb(τ∗)− Ẽb(τb) = 12Ẽ′′
b (τb)(τ∗ − τb)2 +O((τ∗ − τb)3) and Ẽb(τ∗)− Ẽb(τb) = −12e′′2(τb)(τ∗ −
τb)2 +O((τ∗ − τb)3) +O(µ), which imply the desired result. That concludes the proof of
the theorem.
The following result was discovered numerically in the physics literature and proven
in [3] using earlier result of [31]:
Theorem 20. The function β(τ) has exactly two critical points, τ = eiπ/3 and τ = eiπ/2.
The first is minimum, whereas the second is a maximum.
Theorems 19, 20 imply the remaining, (IV), statement of Theorem 8.
2.8 Bifurcation with Symmetry
In this section we present a variant of a standard result in Bifurcation Theory.
Theorem 21. Let X and Y be complex Hilbert spaces, with X a dense subset of Y , and
consider a map F : R × X → Y that is analytic as a map between real Banach spaces.
Suppose that for some λ0 ∈ R, the following conditions are satisfied:
1. F (λ, 0) = 0 for all λ ∈ R,
2. DψF (λ0, 0) is self-adjoint and has an isolated eigenvalue at 0 of (geometric) mul-
tiplicity 1,
3. For non-zero v ∈ nullDψF (λ0, 0), 〈v,Dλ,ψF (λ0, 0)v〉 6= 0,
4. For all α ∈ R, F (λ, eiαψ) = eiαF (λ, ψ).
5. For all ψ ∈ X, 〈ψ, F (λ, ψ)〉 ∈ R.
-
Chapter 2. Existence of Abrikosov Lattice Solutions 32
Then (λ0, 0) is a bifurcation point of the equation F (λ, ψ) = 0. In fact, there is a family
of non-trivial solutions, (λ, ψ), unique in a neighbourhood of (λ0, 0) in R ×X, and this
family has the form λ = φλ(|s|2),
ψ = sv + sφψ(|s|2),
for s ∈ C with |s| < �, for some � > 0. Here v ∈ nullDψF (λ0, 0), and φλ : [0, �) → R
and φψ : [0, �)→ X are unique real-analytic functions, such that φλ(0) = λ0, φψ(0) = 0.
Proof. The analysis of Section 2.5 reduces the problem to the one of solving the bifurca-
tion equation (2.31). Since the projection P , defined there, is rank one and self-adjoint,
we have
Pψ =1
‖v‖2〈v, ψ〉v, with v ∈ nullDψF (λ0, 0). (2.44)
We can therefore view the function γ in the bifurcation equation (2.31) as a map γ :
R× C→ C, where
γ(λ, s) = 〈v, F (λ, sv0 + w(λ, sv)〉.
We now look for non-trivial solutions of this equation, by using the Implicit Function
Theorem to solve for λ in terms of s. Note that if γ(λ, t) = 0, then γ(λ, eiαt) = 0 for all
α, and conversely, if γ(λ, s) = 0, then γ(λ, |s|) = 0. So we need only to find solutions
of γ(λ, t) = 0 for t ∈ R. We now show that γ(λ, t) ∈ R. Since the projection Q is
self-adjoint, and since Qw(λ, v) = w(λ, v) we have
〈w(λ, tv), F (λ, tv + w(λ, tv)〉 = 〈w(λ, tv), QF (λ, tv + w(λ, tv)〉 = 0.
Therefore, for t 6= 0,
〈v, F (λ, tv + Φ(λ, tv))〉 = t−1〈tv + w(λ, tv), F (λ, tv + w(λ, tv))〉,
and this is real by condition (5) of the theorem. Thus we can restrict γ to a function
γ0 : R× R→ R.
-
Chapter 2. Existence of Abrikosov Lattice Solutions 33
By a standard application of the Implicit Function Theorem to t−1γ0(λ, t) = 0, in
which (1)-(3) are used (see for example [8]), there is � > 0 and a real-analytic function
φ̃λ : (−�, �) → R such that φ̃λ(0) = λ0 and if γ0(λ, t) = 0 with |t| < �, then either t = 0
or λ = φλ(t). Recalling that γ(λ, eiαt) = eiαγ(λ, t), we have shown that if γ(λ, s) = 0
and |s| < �, then either s = 0 or λ = φλ(|s|).
We also note that by the symmetry, φ̃λ(−t) = φ̃λ(|t|) = φ̃λ(t), so φ̃λ is an even real-
analytic function, and therefore must in fact be a function solely of |t|2. We therefore set
φλ(t) = φ̃λ(√t), and so φλ is real-analytic.
We now define φψ : (−�, �)→ R to be
φψ(t) =
t−1w(φλ(t), tv) t 6= 0,
0 t = 0,
(2.45)
φψ is also real-analytic and satisfies sφψ(|s|2) = w(φλ(|s|2), sv) for any s ∈ C with |s|2 < �.
Now we know that there is a neighbourhood of (λ0, 0) in R × nullDψF (λ0, 0) such
that in that neighbourhood F (λ, ψ) = 0 if and only if γ(λ, s) = 0 where Pψ = sv. By
taking a smaller neighbourhood if necessary, we have proven that F (λ, ψ) = 0 in that
neighbourhood if and only if either s = 0 or λ = φλ(|s|2). If s = 0, we have ψ =
sv + sφψ(|s|2) = 0 which gives the trivial solution. In the other case, ψ = sv + sφψ(|s|2)
and that completes the proof of the theorem.
2.9 The Operators Ln and M
In this section we prove Theorems 12 and 11. The proofs below are standard.
Proof of Theorem 12. The fact that M is positive follows immediately from its definition.
We note that its being strictly positive is the result of restricting its domain to elements
having mean zero.
-
Chapter 2. Existence of Abrikosov Lattice Solutions 34
Proof of Theorem 11. First, we note that Ln is clearly a positive self-adjoint operator. To
see that it has discrete spectrum, we first note that the inclusion H2 ↪→ L2 is compact for
bounded domains in R2 with Lipschitz boundary (which certainly includes lattice cells).
Then for any z in the resolvent set of Ln, (Ln− z)−1 : L2 → H2 is bounded and therefore
(Ln − z)−1 : L2 → L2 is compact.
In fact we find the spectrum of Ln explicitly. We introduce the harmonic oscillator
annihilation and creation operators, αn and (αn)∗, with
αn := (∇An0 )1 + i(∇An0 )2 = ∂x1 + i∂x2 +n
2x1 +
in
2x2. (2.46)
One can verify that these operators satisfy the following relations:
1. [αn, (αn)∗] = 2 curlAn0 = 2n;
2. Ln − n = (αn)∗αn.
As for the harmonic oscillator (see for example [22]), this gives the explicit information
about σ(L) as stated in the theorem. Furthermore, the second and the third properties
imply
null(Ln − n) = nullαn. (2.47)
We can now prove the following.
Proposition 22. nullLn − n is given by
null(Ln − n) = { ein2x2(x1+ix2)
∞∑k=−∞
ckeki√
2π Im τ(x1+ix2) | ck+n = einπτe2kiπτck} (2.48)
and therefore, in particular, dimC nullLn = n.
Proof. We find nullαn. A simple calculation gives the following operator equation
en4|x|2αne−
n4|x|2 = ∂x1 + i∂x2 .
-
Chapter 2. Existence of Abrikosov Lattice Solutions 35
This immediately proves that ψ ∈ nullαn if and only if ξ = en4 |x|2ψ satisfies ∂x1ξ+i∂x2ξ =
0. We now identify x ∈ R2 with z = x1 + ix2 ∈ C and see that this means that ξ is
analytic. We therefore define the entire function Θ to be
Θ(z) = e−n(rτ )2
4π2z2ξ
(rτz
π
).
The quasiperiodicity of ψ transfers to Θ as follows.
Θ(z + π) = Θ(z),
Θ(z + πτ) = e−2inze−inπτΘ(z).
To complete the proof, we now need to show that the space of the analytic functions
which satisfy these relations form a vector space of dimension n. It is easy to verify that
the first relation ensures that Θ have a absolutely convergent Fourier expansion of the
form
Θ(z) =∞∑
k=−∞
cke2kiz.
We recall that Im τ > 0 and therefore this series does converge. The second relation, on
the other hand, leads to relation for the coefficients of the expansion. Namely, we have
ck+n = einπτe2kiπτck
And that means such functions are determined solely by the values of c0, . . . , cn−1 and
therefore form an |n|-dimensional vector space.
This completes the proof of Theorem 11.
2.10 Fixing the Gauge
We provide here an alternate proof of Proposition 7, largely based on ideas in [18]. We
begin by defining the function B : R→ R to be
B(ζ) =1
r
∫ r0
curlA(ξ, ζ) dξ.
-
Chapter 2. Existence of Abrikosov Lattice Solutions 36
It is clear that b = 1rτ2
∫ rτ20
B(ζ) dζ. A calculation shows that B(ζ + rτ2) = B(ζ).
We now define P = (P1, P2) : R2 → R2 to be
P1(x) = bx2 −∫ x2
0
B(ζ) dζ,
P2(x) =
∫ x1τ1τ2x2
curlA(ξ, x2) dξ +τ ∧ xτ2
B(x2).
A calculation shows that P is doubly-periodic with respect to L.
We now define η′ : R2 → R to be
η′(x) =b
2x1x2 −
∫ x10
A1(ξ, 0) dξ −∫ x2
0
A2(x1, ζ)− P2(x1, ζ) dζ.
η′ satisfies
∇η = −A+ A0 + P.
Now let η′′ be a doubly-periodic solution of the equation ∆η′′ = − divP . Also let
C = (C1, C2) be given by
C = − 1|Ω|
∫Ω
P +∇η dx,
where Ω is any fundamental cell, and set η′′′ = C1x1 + C2x2.
We claim that η = η′+η′′+η′′′ is such that A+∇η satisfies (i) - (iii) of the proposition.
We first note that A+∇η = A−A+A0 +P+∇η′′+C. By the above, A′ = P+∇η′′+C is
periodic. We also calculate that divA′ = divP+∆η′′ = 0. Finally∫A′ =
∫P+∇η−C =
0.
All that remains is to prove (iv). This will follow from a gauge transformation and
translation of the state. We note that
A0(x+ t) + A′(x+ t) = A0(x) + A
′(x) +b
2
−t2t1
.This means that A0(x+ t)+A
′(x+ t) = A0(x)+A′(x)+∇gt(x), where gt(x) = b2t∧x+Ct
for some constant Ct. To establish (iv), we need to have it so that Ct = 0 for t = r, rτ .
-
Chapter 2. Existence of Abrikosov Lattice Solutions 37
First let l be such that r ∧ l = −Crb
and rτ ∧ l = −Crτb
. This l exists as it is the solution
to the matrix equation 0 r−rτ2 rτ1
l1
l2
= −Crb−Crτ
b
,and the determinant of the matrix is just r2τ2, which is non-zero because (r, 0) and rτ
form a basis of the lattice. Let ζ(x) = b2l ∧ x. A straight forward calculation then shows
that eiζ(x)ψ(x+ l) satisfies (iv) and that A(x+ l) +∇ζ(x) still satisfies (i) through (iii).
This proves the proposition.
-
Chapter 3
Stability of Abrikosov Lattice
Solutions
In this chapter we study the stability of the Abrikosov lattice solutions of the Ginzburg-
Landau equations found in Chapter 2 under different classes of perturbations.
3.1 Gorkov-Eliashberg-Schmidt equations
We begin by discussing the framework of the stability problem, i.e., by discussing the
time-dependent Ginzburg-Landau equations of superconductivity (the Gorkov-Eliashberg-
Schmidt equations): γ∂tΦΨ = ∆AΨ + κ
2(1− |Ψ|2)Ψ,
σ(A+∇Φ) = − curl∗ curlA+ Im(Ψ∇AΨ),(3.1)
where Ψ : R2 × R+ → C is the order parameter, A : R2 × R+ → R2,Φ : R2 × R+ → R
are the vector and scalar potentials, respectively, γ is a complex number, and σ is a real
number. ∂tΦ is defined to be ∂tΦ := ∂t + iΦ.
The second equation is based on Ampère’s law (a Maxwell equation with −∂tE ne-
38
-
Chapter 3. Stability of Abrikosov Lattice Solutions 39
glected) and Faraday’s law,
curlB = J and curlE + ∂tB = 0,
where B, E and J are the magnetic field, the electric field and the total current. To
this one adds the constitutive relations J = Js + Jn, where Js = Im(Ψ∇AΨ), Jn are
the superconducting and normal current, respectively, and Jn = σE, where σ is the
conductivity coefficient (in general, the conductivity tensor), and the expression of the
magnetic and electric fields in terms of the vector and magnetic potentials: B = curlA
and E = −∇Φ− ∂tA.
Multiplying the second equation in (3.1) by σ−1 and taking div of the result we obtain
∆Φ = −∂t divA+ σ−1 div[Im(Ψ∇AΨ)]. (3.2)
This gives an equation for Φ, which can be easily solved. The solution is determined up
to a harmonic function on R2. We fix the solution (up to a constant) by demanding that
Φ is bounded. In what follows we always assume that Φ is a bounded solution of the
equation (3.2) and, in particular, is a function of Ψ and A, and we do not list it among
unknowns.
The Gorkov-Eliashberg-Schmidt equations (3.1) admit several symmetries, that is,
transformations which map solutions to solutions.
Gauge symmetry: for any sufficiently regular function η : R2 → R,
Ψ 7→ eiηΨ, A 7→ A+∇η, Φ 7→ Φ− ∂tη;
Translation symmetry: for any h ∈ R2,
Ψ(x, t) 7→ Ψ(x+ h, t), A(x, t) 7→ A(x+ h, t), Φ(x, t) 7→ Φ(x+ h, t);
Rotation and reflection symmetry: for any R ∈ O(2)
Ψ(x, t) 7→ Ψ(Rx, t), A(x, t) 7→ R−1A(Rx, t), Φ(x, t) 7→ Φ(Rx, t).
-
Chapter 3. Stability of Abrikosov Lattice Solutions 40
In order to keep the notation as simple as possible we consider only the case γ = σ = 1.
In this case (3.1) is the L2−gradient flow, in the sense that ∂tΨ + iΦΨ∂tA+∇Φ
= −E ′Ω(Ψ, A), (3.3)for the Ginzburg-Landau energy functional
EΩ(Ψ, A) :=1
2
∫Ω
{|∇AΨ|2 + (curlA)2 +
κ2
2(1− |Ψ|2)2
}, (3.4)
with either Ω = R2 or Ω ⊂ R2, and with appropriate boundary conditions. (Note that
the right hand side of (3.1) is the L2−gradient of EΩ(Ψ, A).)
3.2 Linearized stability of static solutions
We wish to study the stability of certain static solutions (Ψ∗, A∗,Φ∗) of (3.1). The
latter asserts that solutions of (3.1) with initial conditions close, in a certain norm, to
(Ψ∗, A∗,Φ∗) (with possibly modified parameters) remain close as t → ∞. Proving such
a result is an open problem. We address the simpler problem of linearized stability.
Linearized stability can be stated as follows. Let L := E ′′GL(Ψ∗, A∗), the Hessian of
the energy. We consider the operator L on a suitable space of w = (ξ, α) with the inner
product
〈w,w′〉 =∫
Re ξξ′ + α · α′. (3.5)
L is a real-linear operator satisfying (see [21]) 〈w′, Lw〉 = 〈Lw′, w〉 in the inner product
(3.5). We say that the solution u∗ = (Ψ∗, A∗) is linearly stable if and only if
nullL = Z, (3.6)
〈w,Lw〉 > 0, ∀w ⊥ Z. (3.7)
Here Z is the span of the zero modes of L (see below for identification of these modes).
-
Chapter 3. Stability of Abrikosov Lattice Solutions 41
To explain the meaning of this definition, we linearize (3.1) around the solution
(Ψ∗, A∗,Φ∗) to obtain the real-linear equation (∂t + iφ)ξ∂tα +∇φ
= −L ξ
α
. (3.8)Linearized stability should imply that any solution of (3.8), with an initial condition
w0 ⊥ Z satisfies ‖w‖ → 0, as t → ∞. Indeed, differentiating the expression ‖w‖2 with
respect to time and using (3.8) for ∂tw, we obtain
1
2∂t‖w‖2 = −〈w,Lw〉.
Equations (3.6) and (3.7) imply that the right hand side is positive, which indicates that
the desired result might be true. In fact, this statement would be true if we had the
stronger bound 〈w,Lw〉 ≥ c‖w‖2, c > 0.
3.3 Review of existence results
We now turn to the more specific case of the Abrikosov lattice solutions found in the
first chapter, u� = (ψ�, a�) and consider the rescaled Ginzburg-Landau equations (2.4).
We recall the properties of these Abrikosov lattice solutions:
1. For any lattice shape τ , there exists �0 > 0 and a family (λe, ψ�, a�) ∈ R ×
C∞(R2; C) × C∞(R2; R2), defined for � ∈ [0, �0) such that for all � ∈ [0, �0),
u� = (ψ�, a�) is a solution of ∆aψ + λ�ψ − κ2|ψ|2ψ− curl∗ curl a+ Im(ψ̄∇aψ)
= 0.2. Set t(1) and t(2) to be
t(1) = r
10
, t(2) = r Re τ
Im τ
, where r = ( 2πIm τ
) 12
, (3.9)
-
Chapter 3. Stability of Abrikosov Lattice Solutions 42
and set Ω to be
Ω = {p1t(1) + p2t(2) : 0 ≤ p1, p2 < 1}.
Then ψ�(x + t(j)) = e
i2t(j)∧xψ�(x) for j = 1, 2, and a�(x) = a0(x) + α�(x) with
a0(x) =12Jx and α� is such that α�(x + t
(j)) = α(x) for j = 1, 2, divα� = 0 and
〈α�〉Ω = 0.
3. The map � → (λ�, ψ�, α�) is a real analytic map from [0, �0) to R × L2(Ω; C) ×
L2(Ω; R2) and λ� = 1 + �
2λ1 + o(�4),
ψ� = �ψ0 + �3ψ1 + o(�
5),
α� = �2a1 + o(�
4).
Moreover we have the relations
〈Im(ψ̄0∇a0ψ0) · (curl∗ curl)−1 Im(ψ̄0∇a0ψ0)
〉=
1
4
〈|ψ0|4
〉− 1
4
〈|ψ0|2
〉2.
λ1〈|ψ0|2
〉=
(κ2 − 1
2
)〈|ψ0|4
〉+
1
2
〈|ψ0|2
〉2.
3.4 Explicit form of L and its properties
The Hessian L� = E ′′GL(ψ�, a�), or more accurately, the linearization of the rescaled
Ginzburg-Landau equations, is explicitly given by
L�v =
−∆a�ξ − λ�ξ + 2κ2|ψ�|2ξ + κ2ψ2� ξ̄ + 2iα · ∇a�ψ� + iψ� divαcurl∗ curlα + |ψ�|2α− Im(ψ̄�∇a�ξ + ξ̄∇a�ψ�)
, (3.10)where v = (ξ, α).
It is clear that L is a real linear operator and one can easily show that it is symmetric,
i.e.,
〈v, L�v′〉 = 〈L�v, v′〉
with respect to the inner product (3.5).
-
Chapter 3. Stability of Abrikosov Lattice Solutions 43
L also has zero modes arising from the gauge and translation symmetries of the
Ginzburg-Landau equations. Explicitly the translation modes Ti, i = 1, 2, are given by
Ti =
(∇a�ψ�)i−(curl a�)e⊥i
,and the gauge modes Gγ, γ sufficiently regular, are given by
Gγ =
iγψ�∇γ
. (3.11)3.5 Results
We will first study the simpler problem of stability of u� under perturbations that exhibit
the same type of double periodicity as the solutions. We therefore define our space of
perturbations Hper to be the set of v = (ξ, α) ∈ L2loc(R2; C× R2) satisfying ξ(x+ t(j)) =
ei2t(j)∧xξ(x) for j = 1, 2 and α(x+ t(j)) = α(x) for j = 1, 2.
We denote by Lper� the operator L� acting on Hper. We will prove the following result
for this operator.
Theorem 23. Let Z� be the (infinite) subspace spanned by the gauge modes Gγ for
γ ∈ H2loc(R2; R) such that γ(x+ t(i)) = γ(x). If κ2 > 12 , then for � sufficiently small
infv⊥Z�,‖v‖=1
〈v, Lper� v〉 > 0.
We will also consider a more natural class of perturbations by removing the require-
ment that the perturbations have the same double periodicity as the solutions. We define
the space of perturbations H to simply be the set of v = (ξ, α) ∈ L2(R2; C× R2).
We now consider L� on H. To state our result we define the functions φkσ1 , k ∈
[0, 2π)2, σ1 = ±1, to be
φkσ1(x) = e−(k1+ik2)x2e−
12x22+
i2x1x2
∞∑m=−∞
cmeimπr
(x1+ix2),
-
Chapter 3. Stability of Abrikosov Lattice Solutions 44
where the cm ∈ C satisfy
cm+2 = e−rτ2(k1+ik2)eiπτeimπτcm,
and c0 = 1, c1 = 0 for σ1 = 1 and c0 = 0, c1 = 1 for σ1 = −1. We note that the series
converges because Im τ > 0 and thus the cm decay exponentially as m→ ±∞.
We now define the stability constants Sκ,τ to be
Sκ,τ = infk,σ
(2κ2 − 1)[−β(τ)〈|φ01|2
〉 〈|φkσ1|2
〉+ 2
〈|φ01|2|φkσ1|2
〉+ σ2 Re
〈φ̄201φ
2kσ1
〉]
− 2k · Im〈φ̄kσ1∇φkσ1
〉, (3.12)
where σ = (σ1, σ2).
Our main result is the following.
Theorem 24. Fix κ and τ . Let Z� be the (infinite) subspace spanned by the gauge modes
Gγ for γ ∈ H2(R2; R). Then for sufficiently small �,
infv⊥Z�,‖v‖=1
〈v, L�v〉 > 0 (resp. < 0) if and only if Sκ,τ > 0 (resp. < 0) .
We are confident that there are relations between the expression in the definition of
S for different φ (due to symmetries discussed and used in the proof below) and that we
can compute these numbers. We plan to do this in the near future.
The rest of this chapter is devoted to the proof of these two theorems.
3.6 Gauge Fixing
To deal with the infinite dimensional kernel of L�, we follow [21] and first restrict the
class of perturbations to the space of those v = (ξ, α) that are orthogonal to the gauge
zero-modes (3.11). After integration by parts this leads to the condition that
Im(ψ̄�ξ) = divα. (3.13)
-
Chapter 3. Stability of Abrikosov Lattice Solutions 45
We now consider a modified quadratic form L̃� defined by〈v, L̃�v
〉= 〈v, L�v〉 +
∫(Im(ψ̄�ξ)− divα)2.
It is clearly true that L̃� agrees with L� on the subspace defined by the gauge condition
(3.13), but it has the advantage of shifting the essential spectrum away of zero, as can
be seen below. It is straightforward to show that L̃� is explicitly given by
L̃�v =
−∆a�ξ − λ�ξ + (2κ2 + 12)|ψ�|2ξ + (κ2 − 12)ψ2� ξ̄ + 2iα · ∇a�ψ�−∆α + |ψ�|2α− 2 Im(ξ̄∇a�ψ�)
. (3.14)The goal is now to show that
〈v, L̃�v
〉> 0 for all v.
3.7 Complexification
It will be convenient to complexify the spaces Hper and H. We do this explicitly only
for the latter space as the process is the same. We first identify α : R2 → R2 with the
function αC = α1 − iα2 : R2 → C. In order to express all quantities in terms of the
complexified field we introduce the operator ∂aC , defined to be
∂aC = ∂x1 − i∂x2 − iaC.
We also introduce the conjugate of an operator: given an operator A, Ā denotes the
operator CAC where C denotes complex conjugation.
Straightforward calculations show that
2iα · ∇a�ψ� = −i(∂∗aC�ψ�)αC + i(∂aC�ψ�)ᾱ
C.
Also
− Im(ξ̄∇a�ψ�)C =i
2(∂∗aC�ψ�)ξ +
i
2(∂aC� 1ψ�)ξ̄.
We also note that −∂aC0 and −∂∗aC0
are respectively the creation and annihilation op-
erators associated to ∆a0 . In particular we have
∂∗aC0ψ0 = 0.
-
Chapter 3. Stability of Abrikosov Lattice Solutions 46
There is also the important fact that
Im(ψ̄0∇a0ψ0)C = −i
2ψ̄0(∂aC0ψ0).
We now drop the C superscript from the notation and define the complex Hilbert space
K consisting of v = (ξ, φ, α, ω) where α and ω are divergence-free and have mean-zero.
H is embedded in K via the injection
ξα
→
ξ
ξ̄
α
ᾱ
,
and through this embedding L̃� induces an operator on K, which we denote by K� and
which is given by
K� =
−∆a� − λ� + (2κ2 + 12)|ψ�|2 (κ2 − 1
2)ψ2� −i(∂∗a�ψ�) i(∂a�ψ�)
(κ2 − 12)ψ̄2� −∆a� − λ� + (2κ2 + 12)|ψ�|
2 −i(∂a�ψ�) i(∂∗a�ψ�)
i(∂∗a�ψ�) i(∂a�ψ�) −∆ + |ψ�|2 0
−i(∂a�ψ�) −i(∂∗a�ψ�) 0 −∆ + |ψ�|2
.
We will need the following simple relation between L� and its complexification
〈v, L�v〉 =〈vC, K�v
C〉 ,where vC is the the vector corresponding to v in the embedding of H given above.
3.8 Proof of Theorem 23
We now focus on the case of periodic perturbations and prove Theorem 23. From the
above it is clearly sufficient to show that for all v ∈ Kper,
〈v,Kper� v〉 > 0.
-
Chapter 3. Stability of Abrikosov Lattice Solutions 47
We first prove the following lemma about the unperturbed operator Kper0 , which is
explicitly given by
Kper0 =
−∆a0 − 1 0 0 0
0 −∆a0 − 1 0 0
0 0 −∆ 0
0 0 0 −∆
.
Lemma 25. Kper0 is a positive self-adjoint operator with discrete spectrum. It has a zero
eigenvalue of multiplicity 2 and the kernel is spanned by the elements v1 = (ψ0, 0, 0, 0)
and v2 = (0, ψ̄0, 0, 0).
Proof. This operator is simply the operators Ln with n = 1 and M , which we studied
previously in section 2.9, with the difference that M no longer acts on divergence-free
mean-zero fields.
We now expand Kper� in powers of �. We have Kper� = K
per0 + �W
per0 + �
2W per1 + o(�3).
By standard perturbation theory (see e.g. [26, 36, 23]), the spectrum of Kper� consists of
eigenvalues of the same total multiplicities in an �−neighbourhood of the eigenvalues of
Kper0 . Thus it suffices to determine the two lowest eigenvalues of Kper� .
To do this we use the Feshbach-Schur map argument (see e.g. [22, 9]) with the
projection P given by orthogonal projection onto nullKper0 . This argument implies that
λ ∈ σ(Kper� ) if and only if λ ∈ σ(FP (λ)), where
FP (λ) := PKpere P − PKper� P̄ (P̄Kper� P̄ − λ)−1P̄Kper� P, (3.15)
and P̄ = 1− P . It is straightforward to show that
‖W per� ‖ . 1. (3.16)
We know that σ(P̄Kper0 P̄ ) ⊂ [ν0,∞) for some ν0 > 0 and therefore, by standard pertur-
bation theory we have that
σ(P̄Kper� P̄ ) ⊂ [c,∞), (3.17)
-
Chapter 3. Stability of Abrikosov Lattice Solutions 48
with c = ν0 + O(�). We now write Wper� = W
per0 + �W
per1 + o(�
2). We note that we have
the relation Kper0 P = PKper0 = 0 and the fact that by (3.16), ‖PKper� P̄‖ = O(�) and by
(3.17), ‖(P̄Kper� P̄ − λ)−1‖ . 1, provided λ < c, so assuming that λ = O(�) we have
F(λ) = �PW per0 P + �2(PWper1 P − PW
per0 P̄ (P̄K
per0 P̄ )
−1P̄W per0 P ) +O(�3).
We are thus led to considering the lower order operators in FP (λ).
For the order � operator, W per0 is explicitly given by
W per0 =
0 0 0 i(∂a0ψ0)
0 0 −i(∂a0ψ0) 0
0 i(∂a0ψ0) 0 0
−i(∂a0ψ0) 0 0 0
. (3.18)
Here we have used the fact that ∂∗a0ψ0 = 0.
Lemma 26. PW per0 P = 0.
Proof. A simple calculation shows that
W per0 v1 =
0
0
0
−iψ0(∂a0ψ0)
, W per0 v2 =
0
0
iψ̄0(∂a0ψ0)
0
.
It is clear then that 〈vj,W per0 vk〉 = 0 for j, k = 1, 2, and the lemma follows.
We now to turn to the �2 order operator, which we will represent as a 2 × 2 matrix
using the basis β = {v1, v2}. We first note that W per1 is explicitly given by
W per1 =
A1 00 A2
, (3.19)where
A1 =
−λ1 + (2κ2 + 12)|ψ0|2 − ia1∂∗a0 + iā1∂a0 (κ2 − 12)ψ20(κ2 − 1
2)ψ̄20 −λ1 + (2κ2 + 12)|ψ0|
2 + iā1∂∗a0 − ia1∂a0
,
-
Chapter 3. Stability of Abrikosov Lattice Solutions 49
and
A2 =
|ψ0|2 00 |ψ0|2
.
Lemma 27. In terms of the basis β, PW per1 P − PWper0 P̄ (P̄K
per0 P̄ )
−1P̄W per0 P is repre-
sented by the matrix
〈|ψ0|2
〉β(τ)
κ2 − 12 + 1β(τ) κ2 − 12κ2 − 1
2κ2 − 1
2+ 1
β(τ)
.
Proof. We begin with PW per1 P , which is represented by the matrix M1 given by
M1jk =〈vj,W per1 vk〉‖vj‖2
.
Using the fact that ∂∗a0ψ0 = 0, we calculate that
W per1 v1 =
−λ1ψ0 + (2κ2 + 12)|ψ0|2ψ0 + iā1∂a0ψ0
(κ2 − 12)|ψ0|2ψ̄0
0
0
,
W per1 v2 =
(κ2 − 12)|ψ0|2ψ0
−λ1ψ̄0 + (2κ2 + 12)|ψ0|2ψ̄0 − ia1∂a0ψ0
0
0
.
We now note that iā1∂a0ψ0 = 2ia1 · ∇a0ψ0, and therefore
〈iā1ψ̄0∂a0ψ0
〉=〈2ia1 · ψ̄0∇a0ψ0
〉= −2
〈a1 · Im(ψ̄0∇a0ψ0)
〉= −1
2
〈|ψ0|4
〉+
1
2
〈|ψ0|2
〉2.
-
Chapter 3. Stability of Abrikosov Lattice Solutions 50
We now calculate that〈ψ0,−λ1ψ0 + (2κ2 +
1
2)|ψ0|2ψ0 + iā1∂a0ψ0
〉=
〈−λ1|ψ0|2 + (2κ2 +
1
2)|ψ0|4 + iā1ψ̄0∂a0ψ0
〉= −λ1
〈|ψ0|2
〉+ (2κ2 +
1
2)〈|ψ0|4
〉− 1
2
〈|ψ0|4
〉+
1
2
〈|ψ0|2
〉2= −
(κ2 − 1
2
)〈|ψ0|4
〉− 1
2
〈|ψ0|2
〉2+ (2κ2 +
1
2)〈|ψ0|4
〉− 1
2
〈|ψ0|4
〉+
1
2
〈|ψ0|2
〉2= (κ2 +
1
2)〈|ψ0|4
〉.
By similar calculations we can see that M1 is
〈|ψ0|2
〉β(τ)
κ2 + 12 κ2 − 12κ2 − 1
2κ2 + 1
2
.We now turn to PW per0 P̄ (P̄K
per0 P̄ )
−1P̄W per0 P and we let M2 be its matrix represen-
tation. We note that P̄W per0 P = Wper0 P , and therefore using the above we calculate
that
(P̄Kper0 P̄ )−1P̄W per0 v1 =
0
0
0
i∆−1ψ0(∂a0ψ0)
, (P̄Kper0 P̄ )
−1P̄W per0 v2 =
0
0
−i∆−1ψ̄0(∂a0ψ0)
0
.
We then have
W per0 P̄ (P̄Kper0 P̄ )
−1P̄W per0 v1 =
−(∂a0ψ0)∆−1ψ0(∂a0ψ0)
0
0
0
,
W per0 P̄ (P̄Kper0 P̄ )
−1P̄W per0 v2 =
0
−(∂a0ψ0)∆−1ψ̄0(∂a0ψ0)
0
0
.
-
Chapter 3. Stability of Abrikosov Lattice Solutions 51
Now we know that
a1 = ((curl∗ curl)−1 Im(ψ̄0∇a0ψ0))C =
i
2∆−1ψ̄0(∂a0ψ0).
Therefore
〈ψ0,−(∂a0ψ0)∆−1ψ0(∂a0ψ0)
〉=〈−2iψ̄0(∂a0ψ0)ā1
〉=〈|ψ0|4
〉−〈|ψ0|2
〉2.
After a similar calculation, it follows that M2 is
〈|ψ0|2
〉(β(τ)− 1)
1 00 1
.The lemma now follows as PW per1 P −PW
per0 P̄ (P̄K
per0 P̄ )
−1P̄W per0 P is represented by
the matrix M1 −M2.
It can be easily checked that eigenvalues of the �2 order operator are
µ1 =〈|ψ0|2
〉 (β(τ)(2κ2 − 1) + 1
), µ2 =
〈|ψ0|2
〉.
Since κ2 > 12
it is clear that both eigenvalues are strictly positive. The Feschbach-Schur
operator now implies that the two lowest eigenvalues of Kper� are positive. We have seen
above that the other eigenvalue are greater or equal to some c > 0, and it therefore
follows that Kper� is a positive operator, and the proof of Theorem 23 is complete.
3.9 Bloch Decomposition
We now turn to the operator K�. The analysis here will be very similar to the Bloch
theory (or Floquet theory) of Schrödinger operators with periodic potentials (see [36, 17],
as well as [39, 30] for the application of this theory to other equations). The basic idea
of the analysis is to decompose the space K as the direct integral of spaces on a compact
domain in such a way that the operator K� likewise is decomposed as the direct integral
of operators on these spaces.
-
Chapter 3