You are on page 1of 10

On surface defects and their impact on the

superconducting phase in quantum wires


arXiv:1712.07650v1 [math-ph] 20 Dec 2017

1
Joachim Kerner
Department of Mathematics and Computer Science
FernUniversität in Hagen
58084 Hagen
Germany

Abstract
In this paper we are interested in understanding the impact of surface defects on
the superconducting phase in quantum wires. Based on previous results [Ker], we es-
tablish a simple mathematical model in order to incorporate such surface effects. For
a system of non-interacting pairs, we will prove the destruction of the superconduct-
ing phase in the bulk. Finally, taking repulsive interactions between the pairs into
account, we will show that the superconducting phase is recovered for pair densities
larger than a critical one given the density of the surface defects is not too small.

1
E-mail address: Joachim.Kerner@fernuni-hagen.de
1 Introduction
In this paper we are interested in establishing a mathematical model which allows to
understand the impact of surface defects on the superconducting phase in a simple quantum
wire, namely the half-line R+ = [0, ∞).
As understood by Bardeen, Schrieffer and Cooper [Coo56, BCS57], the superconducting
phase in (type-I) superconductors results from a coherent behaviour of pairs of electrons
(Cooper pairs) similar to the one occurring in Bose gases (Bose-Einstein condensation)
[MR04]. Discovered first by Onnes, the most striking experimental feature of supercon-
ductors is a vanishing of the electrical resistance below some critical temperature [Onn91].
In a superconductor, the formation of a Cooper pair is a result of the interaction of the
two electrons with the lattice constituting the solid (electron-phonon-electron interaction).
Due to the negative binding energy of each formed pair, the many-particle ground state
of the superconductor (which itself is formed of pairs only) is separated from the excited
states by a finite energy gap ∆ [BCS57, MR04]. This energy gap, on the other hand, is one
of the most important features that distinguishes the superconducting from the normal
conducting phase. It is therefore not surprising that explaining the formation of such a
gap was the main objective of Coopers ground-breaking work [Coo56].
Superconductivity as described above is a bulk phenomenon. However, in solid state
physics it has long become clear that surface effects play an important role in various
situations [dG64] and, in particular, affect the superconducting behaviour of metals [SB70,
BKV96, KT00, Bel03, Nar17]. Hence, starting from [Ker] where rigorous result regarding
the superconducting phase in quantum wires where obtained, we will construct a simple
mathematical model that allows to incorporate surface defects which are small compared to
the bulk. After establishing the model, we will investigate condensation of pairs in the bulk:
In a first result we show that the ground-state of the bulk does not remain macroscopically
occupied after taking the surface defects into account. Hence, from a physical point of
view, it becomes favourable for the pairs to accumulate in the surface defects. In a second
step we introduce repulsive interactions between the pairs and establish the existence of
different regimes, some in which condensation prevails and some in which it does not. Most
importantly, given the density of the surface defects is not too small, there exists a critical
pair density such that the pairs condense in the bulk for densities larger than this critical
one.

2 Formulation of the model


We consider the quantum wire which is modelled by the half-line R+ = [0, ∞). On this
quantum wire we place, as in [Ker], a system of two interacting electrons (with same spin)
whose Hamiltonian shall formally be given by

∂2 ∂2
Hp = − − + vb (|x − y|) (2.1)
∂x2 ∂y 2

2
with a binding-potential vb : R+ → R defined as
(
0 if 0 ≤ x ≤ d ,
vb (x) := (2.2)
∞ else .

Due to the chosen binding potential, the two electrons form a pair whose spatial extension
is characterised by the parameter d > 0. In [Ker] a mathematically rigorous realisation of
(2.1) was obtained via the construction of a suitable quadratic form on the Hilbert space
L2a (Ω) := {ϕ ∈ L2 (Ω)| ϕ(x, y) = −ϕ(y, x)} with Ω := {(x, y) ∈ R2+ | |x − y| ≤ d}.
Now, in order to incorporate (localised) surface effects we extend our Hilbert space.
More explicitly, we shall be working on the direct sum

H = L2a (Ω) ⊕ ℓ2 (N) (2.3)

which means that we couple the (continuous) quantum wire to a discrete graph which is
supposed to model surface defects. From a physical point of view this seems reasonable in
a regime where the surface defects are relatively small compared to the bulk.
Consequently, the Hamiltonian of free pair (meaning without surface-bulk interactions)
is then be given by
H0 = Hp ⊕ L(γ) , (2.4)
L(γ) being the (weighted) graph Laplacian, i.e.,
X
(L(γ)f )(n) := γnm (f (m) − f (n)) (2.5)
m

where γ := (γn,m ∈ R+ ) = γ T is the associated edge weight matrix. Since our graph is
given by N, one actually sets γmn = δ|n−m|,1en with (en )n∈N ⊂ R+ .

Remark 2.1. Since the distance between two neighbouring surface defects varies from wire
to wire, it might also seem reasonable to regard the edge weights as random variables which
leads one to consider the random Hamiltonian (see [Sto01] for a general introduction to
this field)
H0 (ω) = Hp ⊕ L(γω ) (2.6)
with {en (ω) ∈ [e1 , e2 ]} being i.i.d. random variables with 0 < e1 < e2 < ∞.

3 On the superconducting phase without surface pair


interactions
In order to study the effect of the surface defects on the superconducting phase we need to
investigate the condensation phenomenon for pairs of electrons similar to [Ker]. Of course,
if one wishes to describe the dynamics of a pair one would like to add a non-diagonal
interaction term to (2.4) which describes the coupling between the bulk and the surface

3
defects. However, since we are interested in quantum statistical properties only, we will
simplify the discussion in this paper by modelling the interaction as a diagonal operator.
The coupling between the surface and the bulk is then realised through the “heat bath”
[Sch06].
More explicitly, we consider the (effective) one-pair operator

Hα := Hp ⊕ (L(γ) − α1) (3.1)

with α ≥ 0 some constant, the “surface tension”. Now, since we want to investigate Bose-
Einstein condensation of pairs one has to employ a suitable thermodynamic limit [Sch06].
For this, the half-line R+ is replaced by the interval [0, L] and one considers the restriction

HαL := Hp |L2 (ΩL ) ⊕ (L(γ) − α1)|Cn(L) (3.2)

of (3.1) defined on
HL = L2 (ΩL ) ⊕ Cn(L) (3.3)
where ΩL := {(x, y) ∈ Ω| 0 ≤ x, y ≤ L} and n(L) ∈ N refers to the number of surface
defects up to length L of the wire.
Since HαL is a direct sum of two operators one has σ(HαL ) = σ(Hp |L2 (ΩL ) ) ∪ σ((L(γ) −
α1)|Cn(L) ). As a consequence, HαL has also purely discrete spectrum (see [Ker] for a discus-
sion of Hp |L2 (ΩL ) ). In the following, the eigenvalues of HpL := Hp |L2 (ΩL ) shall be denoted by
{En = En (L) : n ∈ N0 } and the ones of (L(γ) − α1)|Cn(L) by {λj = λj (L) : j ∈ N}, both
counted with multiplicity.
We have the following statement.

Proposition 3.1. For all sequences of edge weights (en )n∈N ⊂ R+ one has

inf σ(HαL ) = −α . (3.4)

Furthermore,
2π 2
lim E0 (L) = E0 < 0.93 · . (3.5)
L→∞ d2
Proof. The first equation follows directly from the fact that zero is the lowest eigenvalue
to the discrete Laplacian associated with the constant eigenfunction (1, 1, 1, ...)T ∈ Cn(L) .
The second part of the statement was proven in [Theorem 2.3,[Ker]] and [Lemma 3.1,[Ker]].

In order to investigate condensation of pairs we will work, as customary in statistical


mechanics, in the grand canonical ensemble [Ver11, Sch06]. The corresponding Gibbs state
is L
L TrFb (e−β(Γ(Hα )−µN̂ ) [ · ])
ωβ,µ ( · ) := , (3.6)
Z(β, µ)

4
where β = T1 is the inverse temperature, µ ∈ (−∞, µmax ) the chemical potential (with µmax
L
specified later) and Z(β, µ) = TrFb (e−β(Γ(Hα )−µN̂ ) ) the partition function. Furthermore, Fb
is the bosonic Fock space over HL ,
n(L) ∞
X X
N̂ = a∗j aj + a∗n an (3.7)
j=1 n=0

the number operator and


n(L) ∞
X X
Γ(HαL ) = (λj − α)a∗j aj + En a∗n an (3.8)
j=1 n=0

the second quantisation of HαL


[MR04, BHE08]. Furthermore, if Φ ∈ HL is a state with
associated number operator nΦ = a∗Φ aΦ , the particle density in this state is given by
L
ωβ,µ (nΦ )
. (3.9)
L
More specifically, for any element of the form ϕn ⊕ 0 with ϕn being the n-th eigenstate of
Hp |L2 (ΩL ) with corresponding eigenvalue En = En (L) one has [Sch06], setting nϕn := nϕn ⊕0 ,
L
ωβ,µ (nϕn ) 1 1
= β(E (L)−µ)
(3.10)
L Le n −1
and the same formula for any element of the form 0 ⊕ fj with fj being the j-th eigenstate
of (L(γ) − α1)|Cn(L) with corresponding eigenvalue λj .
As shown in [Ker], the state ϕ0 is macroscopically occupied in the thermodynamic
limit (meaning that the quotient (3.10) is larger than zero in the thermodynamic limit,
see the following definition) given the underlying Hilbert space is L2 (ΩL ) only. Since this
macroscopic occupation of ϕ0 constitutes the superconducting phase as outlined in [Ker],
we arrive at the following definition.
Definition 3.2 (Superconducting phase in the bulk). Let any pair density ρ > 0 and a
Gibbs state be given. Furthermore, let (µL ) ∈ (−∞, µmax ) be a converging sequence of
chemical potentials such that
 
n(L) ∞
1 X
L
X
L
ρ = lim  ωβ,µ (nλj ) + ωβ,µ (nϕn ) (3.11)
L→∞ L L L
j=1 n=0

holds.
Then, if
L
ωβ,µ L
(nϕ0 )
lim >0 (3.12)
L→∞ L
we say that the state ϕ0 is macroscopically occupied in the thermodynamic limit and that
the superconducting phase in the bulk exists.
If, on the other hand, the limit (3.12) is zero, we say the superconducting phase in the
bulk has been destroyed.

5
Remark 3.3. Note that choosing a sequence chemicals according to (3.11) differs somewhat
from the standard thermodynamic limit where equality has to hold for all values of L. Our
definition of the thermodynamic limit is closer to Einstein’s original considerations [Ein25]
(“Einstein’s argument”) and the one treated in textbooks when discussing Bose-Einstein
condensation. From a physical perspective, however, both limits are expected to yield the
same result.

With this definition at hand we can now establish the following result.

Theorem 3.4 (Destruction of the superconducting phase I). For the Gibbs state associated
with the Hamiltonian HαL , the superconducting phase in the bulk is destroyed for all values
ρ > 0 of the pair density.

Proof. Since µmax = inf σ(HαL ) in the non-interacting case [Sch06], Proposition 3.1 implies
that (µL ) ⊂ (−∞, −α). The result then readily follows with (3.10) taking Proposition 3.1
into account.

4 On the superconducting phase with surface pair in-


teractions
In the previous section we have seen that the superconducting phase in the bulk is destroyed
through the presence of surface defects (see Theorem 3.4). However, since the defects are
imagined relatively small when compared to the bulk, (repulsive) interactions between the
pairs should be taken into account for large pair surface densities.
In order to account for those interactions, we pursue a (quasi) mean-field approach.
More explicitly, the first term on the right-hand side of (3.8) (i.e., the part of the Hamil-
tonian associated with the discrete graph) is replaced by
n(L)
X
λj a∗j aj − αNs + λρs (µL , L)Ns := hM F (ρs , α, λ) , (4.1)
j=1

Pn(L)
where Ns := j=1 a∗j aj is the associated number operator and
L
ωβ,µL
(Ns )
ρs (µL , L) := (4.2)
n(L)

the pair density on Cn(L) . Furthermore, λ > 0 is the interaction strength associated with
the repulsive interactions between the pairs in the surface defects.
2
Remark 4.1. Note that the interaction term in the standard mean-field approach is λ N̂V
with V the associated volume, see [MV99, Ver11].

6
Due to the nature of the interaction part in (4.1) we can rewrite it as
n(L)
X
hM F (ρs , α, λ) = (λj − α + λρs (µL , L)) a∗j aj (4.3)
j=1

which means that the corresponding eigenvalues {λj } are effectively shifted by λρs (µL , L)−
α. Accordingly, the problem has reduced to an effective non-interacting “particle” model
which, in particular, means that the chemical potential satisfies

µL < min{λρs (µL , L) − α, E0 (L)} , (4.4)

taking into account that the lowest eigenvalue of the Laplacian is zero. Hence, µmax (L) =
min{λρs (µL , L) − α, E0 (L)}. Furthermore, the sequences of the chemical potentials (µL )
and the surface pair densities (ρs (L)) have to be chosen in a way such that
 
n(L) ∞
1 X 1 X 1
ρ = lim  −α+λρ (µ ]
+ −µ
 (4.5)
L→∞ L eβ[(λj s L ,L))−µL − 1 eβ(E n L) − 1
j=1 n=0

holds with
n(L)
1 X 1
ρs (µL , L) = β[(λ −α+λρs L ,L))−µL ] − 1

. (4.6)
n(L) j=1 e j

Theorem 4.2 (Surface pair density). Let (µL ) ∈ (−∞, E0 (L)) be a convergent sequence
of chemical potentials with limit value µ < E0 . Then
∞ Z
1X ∞
 
n(L) 1
lim ρs (µL , L) = ρ − 2π 2 n2
dx . (4.7)
L→∞ L π n=1 0 eβ d2 eβ(x2 −µ) − 1

Proof. Starting from (4.5), the statement


√ follows directly from formula (3.4) in [Ker] setting
~ = 1, me = 1/2 and replacing d by d/ 2. Note here that this formula holds true also for
a sequence of chemical potentials by uniform convergence.
L
Now, if we assume that limL→∞ n(L) =: δ with 0 ≤ δ ≤ ∞ then

!

1X 1
Z
lim ρs (µL , L) = δ ρ− 2π 2 n2
dx
L→∞ π n=1 0 eβ d2 eβ(x2 −µ) − 1 (4.8)
=: ρs (µ, δ) ,
for µL → µ < E0 . From a physical point of view it is now very interesting to write ρs as

ρs (µ, δ) = δ(ρ − ρexc ) , (4.9)

where ρexc = ρexc (µ) equals the second term in brackets in (4.8). Note that ρexc is the
density of pairs occupying all excited states in the bulk in the thermodynamic limit.

7
Lemma 4.3 (Destruction of the superconducting phase II). If
E0 + α
δ(ρ − ρexc ) < , (4.10)
λ
then the superconducting phase in the bulk is destroyed.
Proof. By relation (4.4) and the assumptions we conclude

µ < E0 − ǫ (4.11)

for the limit point of (µL ) for some constant ǫ > 0. By Proposition 3.1 we get
1 1
lim (L)−µ
=0 (4.12)
L→∞ L eβ(E 0 L) − 1

which yields the statement .


We immediately obtain the following statement which is interesting from a physical
point of view.
Corollary 4.4. Assume that δ = 0. Then, for all values λ > 0 of the interaction strength,
the superconducting phase in the bulk is destroyed.
More generally, if
E0 + α
δ·ρ< (4.13)
λ
holds, the superconducting phase in the bulk is destroyed.
Remark 4.5. Corollary 4.4 means that the superconducting phase in the bulk is destroyed
for arbitrarily large repulsive (quasi) mean-field interactions if the density of surface defects
is large, i.e., of order larger than L.
In addition, Lemma 4.3 implies that the superconducting phase in the bulk is destroyed
for arbitrarily large pair densities ρ > 0 given the interaction strength λ > 0 is small
enough or the “surface tension” α ≥ 0 large enough.
Finally, we obtain the following result.
Theorem 4.6 (Reconstruction of the superconducting phase). If δ > 0 then there is a
critical pair density ρcrit = ρcrit (β, α, λ) > 0 such that for all densities ρ > ρcrit there exists
a superconducting phase in the bulk.
Proof. In a first step one chooses ρ, for given values β, α, λ > 0, so large that
E0 + α
ρs (E0 , δ) > . (4.14)
λ
Then, using the same reasoning as in the proof of Lemma 4.3, one concludes that

µ ≤ E0 (4.15)

8
for the limit point of the sequence of chemical potentials, taking Proposition 3.1 into
account.
Finally, using (4.14) and (4.15) in (4.6), we conclude the existence of a constant C > 0
such that
|ρs (µL , L)| < C (4.16)
for all L ≥ L0 . This, however, contradicts (4.8) when choosing a large enough pair den-
sity ρ > 0. As a consequence, the state ϕ0 has to be macroscopically occupied in the
thermodynamic limit and its contribution to (4.8) has to be taken into account.

References
[BCS57] J. Bardeen, L. N. Cooper, and J. R. Schrieffer, Theory of superconductivity, Phys. Rev.
108 (1957), 1175–1204.

[Bel03] L. V. Belevtsov, Interplay between surface effects and the magnetic properties in poly-
crystalline superconductors, Journal of Low Temperature Physics 131 (2003), no. 1,
37–49.

[BHE08] J. Blank, M. Havliček, and P. Exner, Hilbert space operators in quantum physics,
Springer, 2008.

[BKV96] L. Burlachkov, A. E. Koshelev, and V. M. Vinokur, Transport properties of high-


temperature superconductors: Surface vs bulk effect, Phys. Rev. B 54 (1996), 6750–6757.

[Coo56] Leon N. Cooper, Bound electron pairs in a degenerate Fermi gas, Phys. Rev. 104
(1956), 1189–1190.

[dG64] P. G. de Gennes, Boundary effects in superconductors, Rev. Mod. Phys. 36 (1964),


225–237.

[Ein25] A. Einstein, Sitzber. Kgl. Preuss. Akadm. Wiss. (1925), 3.

[Ker] J. Kerner, On bound electron pairs in a quantum wire, preprint, arXiv:1708.03753.

[KT00] S. Kashiwaya and Y. Tanaka, Tunnelling effects on surface bound states in unconven-
tional superconductors, Reports on Progress in Physics 63 (2000), no. 10, 1641.

[MR04] P. A. Martin and F. Rothen, Many-Body Problems and Quantum Field Theory, Springer
Verlag, 2004.

[MV99] T. Michoel and A. Verbeure, Non-extensive Bose-Einstein condensation model, Journal


of Mathematical Physics 40 (1999), no. 3, 1268–1279.

[Nar17] A. V. Narlikar, The Oxford Handbook of Small Superconductors, Oxford University


Press, 2017.

9
[Onn91] H. K. Onnes, Through measurement to knowledge: The selected papers of Heike Kamer-
lingh Onnes 1853–1926, pp. 267–272, Springer Netherlands, Dordrecht, 1991.

[SB70] S. T. Sekula and J. H. Barrett, Surface effects and low frequency losses in hard super-
conductors, Applied Physics Letters 17 (1970), no. 5, 204–205.

[Sch06] F. Schwabl, Statistical mechanics, Springer, 2006.

[Sto01] P. Stollmann, Caught by disorder: bound states in random media, vol. 20, Springer
Science & Business Media, 2001.

[Ver11] A. Verbeure, Many-body boson systems, Theoretical and Mathematical Physics,


Springer London, Ltd., London, 2011.

10

You might also like