Superconductivity and

Quantum Coherence
Gil Lonzarich Lent Term 2012

Acknowledgements: Christoph Bergemann,

John Waldram, David Khmelnitskii, …
and, importantly, former students

• 12 Lectures: Wednesday & Friday 11-12 am, Mott Seminar Room

• Three Supervisions, each with one examples sheet
• Questions and suggestions are welcome

Complete versions will be made available on the course web site:

www-qm.phy.cam.ac.uk/teaching.php
1
“Superconductivity, once called one of the best understood
many-body phenomenon in physics, became again 100 years
after its discovery a problem full of questions, mysteries and
challenges.”

X.-G. Wen, MIT

“Quantum Field Theory of Many Body Systems”,
Oxford University Press, 2009

2
 Literature:
JF Annett: Superconductivity, Superfluids and
Condensates
JR Waldram: Superconductivity of Metals
and Cuprates

AJ Leggett: Quantum Liquids – Bose Condensation &

Cooper Pairing in Condensed-Matter Systems
R Feynman: Lectures on Physics Volume III
A Altland & B Simons: Condensed Matter Field Theory
CJ Pethick & H Smith: Bose-Einstein Condensation
in Dilute Gases
M Tinkham: Introduction to Superconductivity
VV Schmidt: The Physics of Superconductors
GE Volovik: The Universe in a Helium Droplet
3
 Outline:

• Ginzburg-Landau (GL) Theory of the

Superconducting State Phenomenological
• Applications of Superconductivity GL Theory
• Bose-Einstein Condensates
• Superfluidity in 4He

• Quantum Coherence and Bardeen- Microscopic

Cooper-Schrieffer (BCS) Theory BCS Theory

• Unconventional Superconductivity New

in Advanced Materials Developments

4
 Lecture 1:
• Historical overview
• Macroscopic manifestation of superconductivity: ρ,
χ, C/T
• Meissner effect and levitation
• Type-I and type-II superconductivity
• Superconductivity as an ordered state –
introduction to the Ginzburg-Landau theory

Literature: Waldram ch. 4

(or equivalent chapters in Annett, Leggett,
Schmidt, or Tinkham)

5
 Timeline:
1898-1908 Liquefaction of H2 & 4He - the crucial steps
Dewar & Onnes
1911 Superconductivity in mercury
1925 Prediction of Bose-Einstein condensation
1927-38 Superfluidity in 4He
1933 Meissner effect
1950 Ginzburg-Landau theory
1952-57 Superconducting vortices
1957 BCS & Bogoliubov theory
1962-64 Josephson effect and SQUIDs
1971 Superfluidity in 3He
1970s-now Unconventional superconductors, including
high temperature superconductors
1990s-now Cold atomic gases, Topological phases, …

6
 Examples of Superconductors
Hg first superconductor Tc = 4.1 K
Nb highest Tc amongst the elements 9.3 K
NbTi superconducting magnets to ~ 9 T 10 K

Nb3Sn superconducting magnets to ~ 20 T 24.5 K

MgB2 high-Tc s-wave superconductor 39 K

CeCu2Si2 Unconventional spin-singlet 0.5 K

Cs3C60 superconductors on border of up to ~40 K
Iron Pnictides Mott transitions and/or up to ~60 K
Copper Oxides antiferromagnetism up to ~160 K

Sr2RuO4 spin-triplet superconductors on

0.5-1.5 K
UGe2 the border of ferromagnetism

many additional important examples exist

7
 Superconducting elements:

www.webelements.com - examples sheet

 Basic experimental facts:
• The resistivity of a superconductor drops to zero below some
transition temperature Tc
• Immediate corollary: can’t change the magnetic field inside a
superconductor

#B
= ! curl % " ! $ curl J = 0, since $ = 0
#t

B=0 B

Switch on external B:

zero field cooled

9
What if we cool a superconductor in a magnetic field and then
switch the field off – do we get something like a permanent
magnet?

B
Experimentally, this does not work
– even when field cooled, the
superconductor expels the field!

field cooled

B
This is known as the Meissner effect.
Superconductivity arises through a
thermodynamic phase transition (state
depends only on final conditions, e.g., T
and B).

field cooled
10
The Meissner effect leads to the stunning
levitation effects that underlie many of the
proposed technological applications of
superconductivity (see examples sheet).

The superconducting state is destroyed

above a critical field Hc
Ideal magnetisation curve…

Hc1 Hc Hc2
H
…and so-called type-II superconductivity
M (which we’ll discuss later)
B

NB: These curves apply for a

magnetic field along a long rod.

11
 The electronic specific heat around the superconducting
transition temperature Tc:

• exponential low-T
behaviour indicative of
energy gap
(explained by BCS)

• power-law behaviour at exponential in simple

superconductors
low-T in unconventional
superconductors
(to be discussed later)

• matching areas means entropy is continuous at Tc: consistent

with second order phase transition

12
From the form of C/T we find that the entropy S vs. temperature
has the following form:
S
superconducting
state

normal state

T
Tc
The superconducting state has lower entropy than the normal
state and is therefore the more ordered state. A general theory
based on just a few reasonable assumptions about the order
parameter is remarkably powerful. It describes not just
conventional superconductors but also the high-Tc
superconductors, superfluids, and Bose-Einstein condensates.
This is known as Ginzburg-Landau theory.
13
 Free energy near a second order
phase transition:
For a second order phase transition, the order parameter
vanishes continuously at Tc. In the conventional description,
known as the Landau model, one assumes that sufficiently close
to Tc the free energy density relative to that of the normal state
can be expanded in a Taylor series in the order parameter, ψ

f (" ) = #" 2 + " 4

!
(! > 0)
2

This assumes that the order parameter is real and that the free
energy density is an even function of the order parameter.

Where is the free energy minimum?

14
Free energy curves:
α> 0 α< 0
f f

Picture credits: A. J. Schofield

ψ −ψ 0 ψ0 ψ

The phase transition takes place at α(Tc) = 0. Thus, a power

series expansion of α(T ) around Tc may be expected to have
the following leading form:

" = a(T ! Tc ) (a > 0)

This is consistent, in particular, with a specific heat jump that

characterizes most superconductors (examples sheet).
15
This is appropriate, e.g., for ferromagnetism where ψ in the
uniform magnetization along a given axis. In the Ginzburg-Landau
(GL) theory, however, ψ is assumed to be complex rather than real
as is the case for a macroscopic wave function. We will see how a
complex order parameter arises naturally from a microscopic
theory. The assumptions in the GL theory are:
2 4
• ψ can be complex-valued "2 # " , "4 # "

• ψ can vary in space – but this carries an energy penalty

proportional to 2
!!
"
• Crucially, ψ couples to the electromagnetic field in the same
way as for an ordinary wavefunction (Feynman III, ch. 21)

" # " ! iqA / h

Here, A is the magnetic vector potential and q is the relevant
charge, which is found to be q = –2e.
16
 This provides the first clue that superconductivity has got
something to do with electron pairs.

A final part in the free energy is the relevant magnetic field

energy density BM2/2µ0, where BM=B-BE is due to currents in the
superconductor and BE is due to external sources. (Note that
when the material is introduced the total field energy density
changes from BE2/2µ0 to B2/2µ0, but the part BMBE/µ0 is taken up
by the external sources (Waldram, Ch.6)).

So finally we arrive at the Ginzburg-Landau free energy density:

$ 4 1
2 1
f ="# + # + ( %ih& + 2eA )# 2+ ( B % BE )2
2 2m 2µo

We have written the free energy so that the gradient term

involve an effective mass m = 2me , which is consistent with
! q = –2e. This represents an effective field theory unifying
matter field ψ & gauge field A (recall B=curlA) in the static limit.
17
 Last Lecture (#1):
The basic superconducting state is characterised by

"=0 B = 0+

and by some internal order (yet to be fully identified).

!
+The condition B = 0 ceases to be valid at high fields in so-
called type-II superconductors where vortex lines enter the
superconductor ( lecture 3).

We introduced the phenomenological Ginzburg-Landau theory

based on a complex-valued order parameter. The theory will be
used to describe the above properties of the superconducting
state and, in particular, the spatial variation of the order
parameter near surfaces or in a magnetic field.
1
 Lecture 2:
• Critical field Bc
• Ginzburg-Landau equations – what happens if we try to
minimise the Ginzburg-Landau free energy?
• London equations – field expulsion and resistance loss
• Penetration depth and coherence length as the two
characteristic length scales of superconductivity
• Gauge transformations and U(1) symmetry breaking

• Literature: Waldram chs. 2 & 4

(or equivalent chapters in Annett, Schmidt, or Tinkham)

2
 The Critical Field Bc
Recall the Ginzburg-Landau (GL) free energy density
2
$ 4 1 2 (B % BE )2
f = "# + # + %ih&# + 2eA# +
2 2m 2µ0
First we look for homogeneous solutions for ψ deep inside the bulk
of the superconductor (sc). Consider the two cases:
!
normal state (α > 0): " 0 = 0, B = BE , f =0
#$ $ 2 BE2
sc state (α < 0): "0 = & 0, B = 0, f =- +
% 2% 2µ 0
The sc state is !
thus evidently stable (f < 0 ) below a critical field

µ 0# 2
! Bc "
$
3
 Minimise GL Free Energy
Close to surfaces and in the presence of vortices (next lecture)
where the magnetic field is incompletely screened and the
supercurrent density is finite we need to look for spatially
inhomogeneous solutions.

To find the most probable configurations of the fields ψ(r) and

A(r), we minimise the free energy density w.r.t. these fields that
in general can vary in a complicated way with position in the
superconductor.

We use the usual shortcut of taking the derivatives w.r.t. ψ and

ψ* rather than Re ψ and Im ψ, but the end result is of course
the same. The parameters of the GL model other than the ψ
and A fields are held constant. The most probable configuration
will be assumed to be the equilibrium average configuration.

4
(i) The functional derivative w.r.t. ψ : To see what key
mathematical steps are needed, consider first the simpler free
energy density for a real order parameter ψ
2$ 4 2
f = "# + # + % &#
2
3 2
"f = 2#$"$ + 2%$ "$ + &" '$
The problem is how to handle the last term: to first order in δψ
!
2 2 2
! " #$ = #( $ + "$ ) % #$ = 2#$ & #"$
Next note that
" # (("$ )%$ ) = ("2$ )%$ + "$ # "%$
! 1442443
gives surface contribution
Thus, in the bulk
!
"f = 2"#[$%&2# + (' + (# 2 )#] = 0
1444442444443
=0
! if δf = 0 for any variation δψ
5
 Similarly, the functional derivative of the GL free energy density
w.r.t ψ* gives:
1 2 2
("ih# + 2eA) $ + (% + & $ )$ = 0
2m
First Ginzburg-Landau equation

(ii) The functional

! derivative w.r.t. A : this requires
manipulations similar to that above, but now with
# " (# " A $ !A) = # " B $ !A + B $ # " !A
123 123
14424B
43 µ0 J
gives surface contribution

Thus, in the bulk we find: setting ψ = ns1/2 exp(iθ)

ieh 4e2 "2ehns % 2eA (

Js = (" * #" $ "#"*) $ A" * " = ' #$ + *
m m m & h )
Second Ginzburg-Landau equation
6
! !
 Comments on GL Equations
The first GL equation has the form of the free-particle Schrödinger
equation – were it not for the non-linear (ψ *ψ ψ ) term. This means
we expect ψ to behave in many respects like a macroscopic
wavefunction, but without certain properties associated with linearity:
superposition and normalisation. Also, A now includes the screening
field, which yields B=0 deep inside the sc (next slide).

Without the A-term it is also known as the

Gross-Pitaevskii equation in the context of
Veff
Bose-Einstein condensates. It can be
inferred empirically as arising from an
effective potential Veff ∝ ψ*ψ (lecture 7)

The second GL equation is identical to the

usual quantum-mechanical definition of current.

The GL equations contain the essentials of superconductivity: we

consider in this lecture the Meissner effect and zero resistance.
7
 Meissner Effect
Let’s make the London approximation and assume that the
amplitude |ψ| is fixed near the equilibrium value Im ψ
|ψ|2 = ns = – α/β and that ψ = |ψ|exp(iθ).

θ
Then from the curl of the 2nd GL equation: Reψ

" 4e2ns
curl Js = B (since curl gradθ =0 and curl A=B)
m
also
µ 0 curl Js = curl curl B = (grad div B – ∇2B) (since µ0 Js= curlB)
123
0
!
B m
Thus, " 2B = #=
#2 4µ0e2ns

2nd London Equation

(see below, p. 13, for 1st London equation)
! 8
 Penetration Depth
The last equation describes magnetic screening: B decays
exponentially inside a superconductor, with characteristic length
scale, λ , known as the penetration depth. λ is typically of the
order of 1000 Å (examples sheet).

B B
inside the
superconductor
outside

λ
distance

Note that the value of λ depends only on the superfluid density

ns and on the effective mass m. As T → 0, ns → ne /2 so that in
this limit λ is purely determined by normal state properties. At
Tc , ns → 0, so that λ diverges and screening breaks down.
9
 Gauge Transformations
Thus, the GL theory describes the Meissner effect. What about
zero resistance itself? Obviously, the screening currents
associated with the Meissner effect are persistent and must
therefore flow without resistance. To make a general case for
zero resistance and for later reference it is instructive to
reconsider the problem of gauge transformations.

Recall the forms of the electric and magnetic fields in terms of the
scalar and vector potentials φ and A:

E = "#$ " !A / !t B = curl A

We can easily see that the observable fields E and B remain

unchanged under the gauge transformation

A # A + "! % # % " !$ / !t

10
where χ(r,t) is an arbitrary differentiable and single-valued
function.

"2ehns % 2eA (
The supercurrent Js = ' #$ + *
m & h )

which is an observable, remains unchanged if the phase θ of the

sc wavefunction !
simultaneously transforms as
Im ψ θ + δθ
2e
" $" # ! = θ + δθ θ
h
Re ψ

Thus, a gauge transformation evidently corresponds to a

rotation in the internal space of the matter field at each point
(r,t).

11
 Time Dependence of Phase
The time dependence of ψ cannot be inferred from the GL theory
and is in general complicated. However, London suggested that
the time derivative of θ might be expected to have a contribution
of the form
#" 2e
= !
#t h
a result familiar from quantum mechanics, and consistent with
the gauge transformations
!& !& 2e !$ !$
# " and % #%"
!t !t h !t !t
The relations between J, θ and φ allows us to write

'Js 2ehns & 2e 2e 'A #

=) $ ( * + !
't m % h h 't "

12
 !A
and, from % = "#$ " , thus obtain
!t

!Js 4e2ns
= E
!t m

1st London equation

(see p. 8, for 2nd London equation)

This is an acceleration equation – an electric field is only needed

to kick-start a supercurrent, but not to sustain it.

Historically, the two London equations predate the GL treatment,

and both of course predate the microscopic BCS theory.

13
 Gauge Symmetry Breaking
J, φ and A are affected by χ only through its time and space
derivatives. Thus, we have the freedom to add a global
constant, δθ =–2eχ / ћ, to the phase θ of ψ without changing the
the current and the potentials – a property known as global
gauge symmetry. As this corresponds to a multiplication of ψ
with exp(iδθ) and since these c-numbers form the multiplicative
group U(1) this symmetry is referred to as U(1) (gauge)
symmetry. free energy

A particular choice of ψ = ψ0 in the GL

theory breaks global (not local) U(1) ψο
symmetry. This is visualised by means
of the “mexican hat” (or “bottom of
wine bottle”) potential shown right.
Broken global U(1)
(Qualifications will be given in lecture 12.) symmetry 14
 Coherence Length
The GL free energy has already provided us with one
characteristic length scale, the penetration depth λ, which is the
scale over which magnetic fields vary in a sc.

From the non-linear Schrödinger equation (p. 6 with α = −|α|), we

can define a second length scale, the GL coherence length ξ,

2 2 & 2 h2
" # $ +$ % $ $ = 0 "=
' 2m '

This gives the scale over which |ψ| itself will vary.

!
Unlike λ, the coherence length ξ cannot easily be related to
normal state properties since it depends crucially on α. ξ
diverges at Tc and T → 0 values can vary greatly from material
to material (10 Å to 10,000 Å).
15
 Appendix: Boundary Conditions
(i) SC – vacuum interface:
ψ
n means normal
(" ih!n + 2eAn )# = 0 component i

outside
This implies that there is no current
ii
flow across the surface and that the
iii
full value of ψ continues right up to the
superconductor
surface (a thin film will normally have
the same Tc as the bulk).
distance
(ii) SC − normal metal interface:
"ih$
("ih# n + 2eAn )$ = , where l is a characteristic length
l
over which superconductivity is induced in the normal metal.
(iii) SC − ferromagnet interface:
! ψ = 0 at the interface
for usual case where ferromagnetism destroys superconductivity.
(see e.g., Waldram pp 45-46)
16
 Last lecture (#2):
The Ginzburg-Landau (GL) theory is shown to be consistent
with the dissipationless flow of current and screening of the
magnetic field (Meissner effect).

Two length scales λ and ξ : the penetration depth, λ , describes

the extent of screening currents and magnetic field penetration;
the GL coherence length, ξ , the scale over which the GL
wavefunction ψ(r) typically changes. From the GL equations we
find:

B m 2
$2"# = 2
"#
" 2B = , ! = , " = , ! <0
!2 4 µ0e2 ns 2 2 m !
!

We have encountered the critical field Bc below which the

homogeneous sc state is favoured over the non-sc state. Next
we will describe the inhomogeneous vortex state that is of both
fundamental and technological interest.
1
 Lecture 3:
• Flux quantisation
• The isolated vortex line – a topological defect in the
superconducting phase – and its energy
• The lower critical field Bc1 at which vortices start to enter the
superconductor
• Type-I versus type-II superconductors
• The vortex lattice and the ideal type-II magnetisation curve
• The upper critical field Bc2 where sc is destroyed in a 2nd order
phase transition (spilling over into next lecture)

• Literature: Waldram chs. 2, 4 & 5 or Annett ch. 4; plus

Tinkham ch. 5

2
 Flux Quantisation
Consider a sc with a hole,
threaded by magnetic field lines
with total flux φ: λ

Recall from the last lecture that Js

Js " ( !# / 2e) + A . Far away φ
from the surface, Js → 0 and
thus
"! Integration
A=# path
2e
We can now relate A to the flux
φ through the hole (including
the penetration depth):

!= ' B " d 2r = ' A " ds = # 2 e ' $% " ds = # 2e " &%

3
The phase difference Δθ must be an integer multiple of 2π to
keep ψ(r) the same, i.e., if ψ(r) (and not solely |ψ(r)|2) is
essentially observable. This means that the flux φ is quantised:

h
! = interger x ! , ! = = 2.06783376(5) x 10 "15 Vs
0 0 2e

This is an example of a
macroscopic quantum
phenomenon, topological
invariance and exact science.
Measurements of flux
quantisation and other related
experiments confirm, as stated
in lecture 1, that the relevant
charge is twice the charge of an Early observation of
electron. flux quantisation
(Deaver & Fairbank, 1961)
4
 The Vortex Line

Do we need a hole in the sc

λ
to support a flux quantum?

Picture credits: DAMTP; D. A. Weitz

It is enough to have a Js
region where ψ vanishes.
Consider a normal region
Vortex
of radius of order ξ – the Closed
vortex core.
ξ core
path

This is an example of a
topological defect. Such
defects appear in many areas
ranging from spin systems
(see (a)) & liquid crystals
(see (b)) to cosmic strings.
(a) Spin vortex (b) Director Vortex in
in a 2D System a Liquid Crystal
5
 Free Energy of a Vortex Line (λ >> ξ)
The free energy cost ΔFv of introducing a vortex line per unit
length is ΔFv = F(with vortex) - F(uniform), where F is the
appropriate integral of the GL free energy density f over space.
Recall 2
2" 4 1 2 (B $ BE )
f = # |! | + | ! | + | $ i %! + 2eA ! | +
2 2m 2µ0
µ
constant except 0 "2 J 2 if n is constant*
in narrow core s s will give flux
2 quantum
drop
1
$ f v = f ( with vortex ) " f (u nifor m) = [#2 ( µ0 J s )2 + B 2 " 2 B ! BE ]
! 2 µ0
In the vortex state the kinetic energy (proportional to Js2)
increases, but the field energy BM2/2µ0 decreases. At sufficiently
high BE ≡ Bc1, ΔFv becomes negative and thus vortices form
spontaneously. To find Bc1 we need to find Js(r) for a vortex with
total flux φ0. (Higher flux quanta are energetically unfavourable.)
*The kinetic energy density in this form together with the magnetic
field energy density is known as the London free energy density. 6
 Currents & Fields for a Vortex Line (λ>>ξ)
For a vortex line with a narrow core at r = 0 and of total flux φ0
the London screening equation can be written in the form
B % #2$2B = " 0 ! 2 D (r )
! curlcurlB
We give an approximate solution (see appendix for exact treat-
ment). First integrate the above over a circle of finite radius r
2 2 2 2 2
$ d r B + " $ d r curlcurlB = $ d r B + 2#r" µ 0 J s = ! 0
µ0 Js
For r>>λ, Js is exponentially small (appendix), so that the term
proportional to Js drops out and the flux enclosed by the circle
reduces to φ0 as required. For r << λ, on the other hand, the
integral over B drops out and the current density is
#0
µ Js $
2!"2r
0
*B r -0 &+ #
also with = -µ 0 J s & B(+ ) ' 0, B ' ) ( dr µ 0 J s ' ln $ !
*r + 2,+2 % r " 7
We can now evaluate the free energy density Δfv and thus the
free energy per unit length ΔFv of the vortex

2
)Fv = ! d r )f v ! Bc1
$ $2' 2 BB E '0 '0
0
# ! 2& rdr ( " )= (( " BE )
% 8& 2 µ 0 $ 4 r 2 µ0 µ0 4&$ 2

The 2D integral of the term proportional to B is set equal to the

flux and the weaker second order term in B has been dropped.
In the above η = ln (λ/ξ); a more accurate analysis gives η = 1.

~2λ
Vortex lattice
for B~Bc1

8
 The Critical Field Bc1
Vortices form spontaneously if BE exceeds a critical field given
approximately by

#0
Bc1 =
4!"2

• Bc < Bc1: The sample becomes completely normal before

vortices can enter – type-I superconductivity
• Bc1 < Bc: Vortices enter at Bc1 and the Meissner flux expulsion
is no longer perfect. The free energy considerations above hold
for every single vortex, so vortices will enter until vortex-
vortex repulsion becomes appreciable (we’ll discuss this later).
The vortices then form a vortex lattice, and due to its
favourable free energy, sc persists up to an upper critical field
Bc2 > Bc. This behaviour is called type-II superconductivity.

9
 Type-I vs. Type-II – the Role of κ = λ/ξ

We can rewrite Bc given in the last lecture in a form similar to

Bc1, namely
2$
Bc = 0
4!"#

In terms of the dimensionless parameter κ = λ/ξ the ratio of Bc

to Bc1 is
Bc / Bc1 = 2!

10
 Comments on Type-I and Type-II
• The superconducting elements are type-I except niobium and
perhaps some of the elements at high pressure (e.g., hcp Fe)
• Pure compounds with low electronic effective masses and low
Tc tend to be type-I.
• Metals with high electronic effective masses or high Tc and
impure metals tend to be type-II.

Some Examples (T 0 K)
Compound Tc [K] ξ [nm] λ [nm] κ = λ /ξ

Al 1.18 1600 50 0.03 type-I

Pb 7.19 83 39 0.47 type-I
Nb 9.25 40 44 1.1 border
Nb3Sn 18.2 3.6 124 34 type-II
YBa2Cu3O7-δ 90 (1.5) (130) (87) type-II
2
m
" = ! =
2m ! 4 µ0 e2 ns 11
 Magnetisation of a Type-II Superconductor
Model for a vortex lattice in a long superconducting cylinder
with BE along the cylinder axis
• Array of circular vortices with inner radius ξ & outer radius rB<λ;
• B varies slowly (logarithmically) and can be taken to be
uniform (see examples sheet);
• The flux per vortex is assumed to be one flux quantum and
define rB by φ0 = πrB2 B, or rB = !0 / " B
• The space-average free energy density is ~2rB

rB 2#rdr f v $02 + rB ( ( B % BE )2
- , ln ) && +
2 ) ! 2µ 0
! # rB 2 2
#rB 4#µ 0" * '
We now minimise the r.h.s. w.r.t. B to get
the most probable field defined by
( &# +
#
* where M is the
B ! BE " 0 ln * 0 -=B +µ M
- E magnetisation and η
8 $% 2 * 2 2B - 0
) $' , is of order unity
12
In a more accurate calculation (lecture 4) η = 1. Thus, the mag-
netisation goes to zero, and sc is destroyed, at a critical field

#0
Bc 2 = ; hence Bc1Bc 2 = Bc 2
2!" 2

At Bc2, the vortex cores start to overlap and sc is quenched. The

magnetisation vanishes continuously (in fact, linearly), as
expected for a 2nd order phase transition.

Meissner state
-µ 0 M M = !BE / µ0 #0 Bc2
M%$ 2
ln ( )
“vortex rush”, 8"µ 0 ! B
M is vertical

2nd order,
M is linear in B
Bc1 Bc Bc2 BE
13
 Vortex Imaging
Can map the magnetic field
directly with a scanning
SQUID magnetometer (we’ll
see later what a SQUID is…)

Picture credits: J. R. Kirtley

For this to work, need low
fields just above Bc1 to get
well separated vortex cores.

At higher applied fields,

vortices form a well defined
lattice, and we can image it
as well – either in real space
or even in reciprocal space,
using scattering techniques
(next slide).

14
 Picture credits: Essmann/Träuble; Birmingham SANS group
Vortex Lattice Imaging

Bitter decoration: Deposit fine

Fe powder on surface and take
electron microscope picture.

Small angle neutron scattering: The

Vortex lattice “Bragg-scatters” the
neutrons. [B is parallel to neutron
beam; small angle because vortices
are widely spaced.]
Vortex lattices are usually hexagonal but in some exceptional
cases square or even more exotic lattice symmetries have been
observed.
15
Appendix: Full Vortex Field and Current Distributions
!2 B !B
In cylindrical coordinates "2 B = + so that we consider
!r 2 r!r
B’’ + B’ /u + "B = 0, B' = !B/ ! u, B'' = !B'/ !u, u = r/# & " = ±1
This has Bessel function solutions of zero order if η = 1 and
modified Bessel function solutions of zero order if η = -1. The
latter applies for the London screening equation and for the
boundary conditions of p. 7 we have
%0 r "
B= K (
2 0 #
), ! du uK 0 (u) = 1
2$# 0
$B #0 r
µ Js = % = K1( )
2"!3
0 $r !
Here K0(u) is a modified Bessel function of zero order (the
solution that vanishes at ∞) and K1(u) = -∂K0(u)/∂u. Note that

1 # "u
K 1 (u ) ! for u << 1, but ! e for u >> 1
u 2u
16
 Last lecture (#3):
Flux quantisation in multiples of φ0 – a very intriguing property.
Vortices in type-II superconductors and the vortex lattice lead
to a very intricate magnetisation curve:
Meissner state Vortex state
-µ 0 M ! µ0M = BE #0 Bc 2
% µ0 M $ 2
ln ( )
“vortex rush” 8!" B
M vertical

2nd order
M linear
Bc1 Bc Bc2 BE

In the vortex state we still have a finite sc order parameter ψ

across most of the material – but no perfect Meissner
diamagnetism. One obvious question to ask:

is the resistance still zero?

1
 Lecture 4:

• Nucleation of superconductivity at Bc2

• Critical fields – a summary
• The B-T phase diagram for type-I and type-II sc
• Vortex-current and vortex-vortex interactions – interpretation
in terms of the “Magnus effect” and topological charge
• Vortex pinning, flux flow, and finite resistance
• Critical currents
• Experimental observations – hysteresis and vortex liquid state
• A few words on applications: superconducting wires and
superconducting high field magnets

• Literature: Waldram chs. 4, 5 & 18, plus Schmidt ch.3

2
 An Improved Model for Bc2
At Bc2 we have a 2nd order phase transition into the normal
state. Since |ψ| goes to zero continuously at Bc2, we consider
the linearised GL equation, i.e., the Schrödinger equation
1
("ih! + 2eA)2$ + #$ = 0
2m
kz
This leads to Landau eigenvalues satisfying

& 1# h2k z2
$ n + !h(c + +' = 0
% 2" 2m

where n is an integer and ωc = 2eB/m.

Bc2 is the maximum field where this
condition is satisfied (n = kz = 0)

1 |$ | m #0
h%c + $ = 0 or Bc2 = = Landau
2 eh 2!" 2 Tubes
as given in lecture 3 (with η = 1). in k-space 3
In summary, when λ >> ξ the critical fields are given by:

#0 #0
Bc2 = , Bc1 = and Bc1Bc2 = Bc 2
2!$ 2 4!"2

2 m 2 h2 |! |
$ = # = ns = ! = a(T-Tc )
4 µ 0 e2ns 2m | ! | "

A quantum of flux φ0 is distributed over a circle of radius of order

λ at Bc1 and of order ξ at Bc2 (at Bc2 the vortex cores essentially
touch).
The temperature dependences of the critical fields depend on
α = a(T – Tc) near Tc. Thus, the fields go to zero linearly as T
goes to Tc. The GL model involves a low order expansion in ψ
and in (T – Tc) and thus cannot describe temperature
dependences in the low temperature limit. A more complete
description of the low T behaviour requires a microscopic theory.
4
 B-T Phase Diagrams

Type-I Type-II

B B
2nd order

1st order
Vortex state

Meissner state
Tc Meissner state
T Tc T
2nd order

5
 Critical Current in Type-I Superconductors
Consider first the simplest case of a conventional type-I super-
conductor. There is an intrinsic limit to how much supercurrent
can be sustained: the field produced by the current must be
less than or of the order of Bc. From the figure below the critical
current Ic is given by
µ0 Ic 2! r Bc
" Bc or Ic " r λ
2! r µ0
Since the current is within the penetration depth λ of
the surface, the critical current density Jc is thus

Ic B Silsbee’s rule for the

Jc # = c B
2"r! µ0! depairing current density
if λ << r.
Js
A more detailed analysis gives the same qualitative
result.

6
 Critical Current in Type-II Superconductors
Vortex-Current Interactions
In a type-II superconductor we need to consider the effects of a
steady current on vortex lines.
The interaction between the supercurrent and the magnetic field
leads to a Lorentz force on the supercurrent. For an isolated
vortex and homogeneous background transport current Jtr, so
that Js = Jvortex + Jtr, the total force per unit length
B
along the vortex reduces to Jvortex
) f
2 2
f = $ Js " B d r = Jtr " $ B d r = Jtr " ( # 0 B )
Js
This force is analogous to the Magnus force Js
on J
! a rotating football:
tr

7
It can be shown that the form of the interaction remains the
same in a more general situation. Jtr is then the current density
at the vortex core due to all current sources other than the
vortex itself.1

One interesting consequence of the Lorentz force is that two

vortices in a type-II superconductor repel each other, as can be
seen from their current patterns:2

B
f

Jvortex
Jtr

1 O. Narayan, J. Phys. A 36, L373 (2003)

2 J (r) = φ K (r/λ) / (2π µ λ3) (appendix, lecture 3), so that the
tr 0 1 0
force is f = φ 0 K1(r/λ) / (2π µ0λ3) per unit length.
2

8
 Comment on Topological Charge
We may think of a vortex as a topological defect in the
superconducting phase. In general topological defects interact
and their interaction can be thought of as arising from their
effective charge, the topological charge. In the case of a

Picture credits: M. Srinivasarao

superconductor, the charge is the winding number, Δθ/2π, of the
phase around the vortex core. Defects with charges of the same
sign usually repel each other (e.g., two superconducting
vortices), while defects with charges of opposite sign attract.

Equal
charges:
energy penalty
repulsion

Opposite
charges:
annihilation Two interacting topological
attraction defects in a liquid crystal
9
 Dissipation Due to Vortex Motion
Reconsider the single vortex in a homogeneous transport current
Jtr. If the Lorentz force acting on the vortex leads to vortex
motion at velocity v, this implies that the vortex is subject to a
power input per unit volume of vortex of characteristic radius rB
fv "0
P= 2
= Jtr 2
v = Jtr B v
!r B !r B Lorentz force
per unit volume

The power is supplied by an electric field ε defined by

P= ε Jtr and from above, ε = Bv
which implies that the wire is resistive, with
resistivity B
! Bv v Jvortex
" = = f
Jtr Jtr

The key point is that vortex motion Js

leads to dissipation, and the zero
resistance state breaks down! Js
Jtr
10
 Vortex Flow in the Presence of
a Transport Current Jtr
BE

vortex
v
Jtr
Defect
(possible
pinning
centre)

Jtr

BE

The critical current for dissipationless flow in the vortex state

is dependent on vortex pinning by defects! (Appendix 1)
11
 Hysteresis Due to Vortex Pinning
Recall the ideal magnetisation curve:
-µ0M

BE

Vortex pinning has a profound influence on the kind of M(BE)

curve we actually observe in experiment – strong hysteresis
(Appendix 2):
-µ0M

ΔM
BE
zero field
cooled

12
The critical lines Bc1 and Bc2 vs. T do not provide the
whole story for the B-T phase diagram:

Picture credit: Cavendish Laboratory,

A.P. Mackenzie, S.R. Julian et al.
Magnetic Field (Tesla)

Irreversibility Line for

a Cuprate Superconductor
(Tl2Ba2CuO6, 10mK to 16K)

Temperature (K)

13
 Irreversibility Line

Picture credits: Cavendish Laboratory, J. R. Cooper

In the cuprates in particular the critical current Jc is found to
vanish at fields far below Bc2. The zero resistance state with
finite Jc is found only below a new critical field Birr. In the figure
the line Birr(T), below which the resistivity vanishes, is known as
the irreversibility line. It is
particularly noticeable in the Vortex
cuprates probably because Liquid
of the high transition
temperature, low supercon-
ducting coherence length |ψ | = 0
and low dimensionality. The
region below Bc2 and above
Birr may be thought of as a
vortex liquid. A description
J > 0 Jc = 0
of this state requires a c
|ψ | > 0
consideration of the effects | ψ | > 0
of fluctuations in the GL
model, which is beyond the Vortex
scope of this course. Lattice
14
 Magnetic Field – Temperature Phase
Bc2(T) Diagram of Cuprate Superconductors

Picture credit: E.M. Forgan et al.

Birr(T)

Irreversibility
Line

Bc1(T)

15
Some Applications of Type-II Superconductors
The use of superconducting wires in DC magnets cooled by liquid
helium is now at an advanced stage. Applications include:

• Magnetic resonance Imaging (MRI) systems– a few Tesla

• Magnetic cooling to nK – less than 10T
• Research magnets – up to 25T (2012)

MRI
Use NbTi up to 9T – ductile metal, severe
cold working to introduce pinning centres.

Use Nb3Sn up to 22T (45T for hybrids) –

brittle material, needs to be wound in
precursor form and then fired as a solenoid.

ATLAS solenoid
(CERN)
For higher fields, to 25T and beyond, high-Tc
materials are used, but this is still an active
area of R&D. Some main challenges: to
reduce brittleness, improve effectiveness of
pinning and reduce risk of thermal runaway.
16
 Appendix 1: Vortex Pinning
The key to the zero resistance state in high fields is the
seemingly unexciting issue of vortex pinning.
Vortices can be pinned by various types of defects in the crystal,
such as dislocations or grain boundaries. At a defect |ψ| may be
suppressed anyway, so having the vortex core at the defect
gives you “two normal regions for the price of one” !
energetically favourable. This is a complex problem involving a
number of other factors. It is enough to stress that the
effectiveness of vortex pinning is crucial to attaining high critical
currents and low dissipation.

The Lorentz force per volume goes as Jtr B, i.e., it rises as B

increases. Once it exceeds the pinning force density of the
vortex lattice, the material will lose its zero resistance.
Generally, we expect the critical value of Jtr when this happens
to scale as
1
Jc !
B
17
 Picture credits: J. Petermann / A. M. Campbell & J. E. Evetts
This is roughly consistent with experiment except at low fields
where not all pinning centres carry a vortex core and at high
fields near Bc2 when Jc clearly vanishes. Also, we get an
enhancement of Jc at intermediate fields when the average
pinning centre separation matches the average separation of the
vortices. This enhancement in Jc is known as the peak effect
(see figure below).
Jc can be made quite high by
metallurgical processes.
Paradoxically, defects are good.*
Typical values that can be
achieved are 5x105 A/cm2 (NbTi
enhancement
at 4 T and 4.2 K) and 106 A/cm2
(YBCO films at 10 T and 77 K).

*Also note that empirically

impurities tend to reduce ξ and
hence increase Bc2.

18
 Appendix 2: The Bean Model

The Bean Model offers a crude description of the hysteresis loop

of a thin slab of a type-II superconductor. The model assumes:

• |∂B/∂x| = µ0|Js| is equal to the critical value µ0Jc (taken to be a

constant) and B = BE at the surface;

• when ramping up BE (blue lines in figure on p20) the flux

enters at the surface and when ramping down (red lines) the
flux leaves near the surface.

There is a critical value of BE where the field B penetrates to the

centre of the slab (near the third blue line from bottom). When
the field BE is reduced to zero the flux remains trapped inside the
superconductor (bottom red line).

19
 The Bean Model

End Type-II Superconductor

B
ramp up ramp up
Ramp down

ramp down

x
Start
End
ramp up
ramp down
20
 Last lecture (#4):
We completed the discussion of the B-T phase diagram of type-I
and type-II superconductors. In contrast to type-I, the type-II
state has finite resistance unless vortices are pinned by defects.

Jvortex
Jtr

• Homogeneous state → Meissner effect → type-I sc

• Inhomogeneities at isolated points → vortices → type-II sc
• Inhomogeneities at weak links → Josephson effect, SQUIDs

1
 Lecture 5:

• Weak links and the Josephson phase relation

• Josephson critical current
• DC and AC Josephson effect with voltage source
(current source given in an appendix)
• Gauge-invariant phase
• Quantum interference for weak links
• The DC SQUID
• Applications of SQUIDS
• Other applications of Josephson phenomena:
frequency mixers and voltage standards

• Literature: Waldram chs 6 & 18

2
 Weak Links
ψ
V
"1 = "2 =
SC 1 SC 2
i! 1 i! 2
|" | e |" | e
d

• Bulk superconductors 1 & 2 are separated by a very thin

region of normal metal or insulator.

• A simple model of a 1D superconducting weak link, d<<ξ,

and strong link, d>>ξ, is given in the appendix. Here we
consider a more general phenomenological approach

• Phase difference ϕ = θ1 − θ2 evolves as (lecture 2)

!" / !t = 2eV / h

3
 • In GL free energy density and in expression for current we
replace
$1 " $2
"#$ by % exp(i& ) " 1
d
• The current through the link is then of the form

! I = IJ sin ϕ

In a weak link (i.e., d<<ξ for 1D sc link) the current is

periodic in ϕ with period 2π.

• The current I is consistent with a free energy term of the

form ΔF = -F0 cos ϕ, where F0 = ћIJ /(2e). Proof:

Power = "#F / "t = F sin $ "$ / "t = IV as required.

0

! 4
 Voltage-Biased Josephson Weak Links
The Josephson Current-Phase Equation
Consider the resistively shunted junction (RSJ):
Is IJ

V O
I
R
In
The total current with bias voltage V is
V
I = Is + In = I J sin ! +
R
Since V = (ћ/2e)∂ϕ /∂t, we can rewrite I in terms of the phase ϕ
alone
h !"
I = I J sin " +
2eR !t
This is a strange circuit equation unlike any known in conventional
circuit theory and it leads to remarkable I-V characteristics. IJ is
known as the Josephson critical current of the weak link and is a
constant that depends on the microscopic details of the junctions.
Typical values of IJ are in the range 10-6 A to 10-2 A. 5
 The Josephson Current-Voltage Relation
From the phase-voltage relation V = (ћ/2e)∂ϕ /∂t, we can write ϕ
as an integral over V(t)
2e t
" = "0 + # V(t )dt
h 0
where ϕ0 is a constant. Thus, the current-voltage form of the
Josephson equation becomes
$ 2e t ' V
! I = I J sin &&" 0 + # V(t )dt )) +
% h 0 ( R

Consider first the case V = 0. Then In = 0 and

!
I = I s = I J sin ! 0
Current flows without an applied voltage, i.e., Is is indeed a
supercurrent flowing through the weak link. This is the
DC-Josephson effect.
6
 AC-Josephson Effect
The main surprise comes when we apply a finite voltage. Consider
first a DC voltage V. This leads to a time-dependent phase
2eV
! = !0 + t
h
and thus to an oscillatory component in the current

V 2eV
I = I J sin ($ 0 + # J t ) + , where # J = = 2"V / !
R h 0
"J V
f = = = (4.8359... x 108 Hz/1µV) V is the Josephson frequency.
J 2# $0
I
Remarkably, a DC applied voltage
drives an oscillating DC super- |Is| · IJ In = V/R
! current at a frequency that is (1/φ0)
per unit of voltage applied. This is V
the AC-Josephson Effect. The DC I-V
characteristic of a RSJ weak link is
given on the right.
7
 Combined DC and AC Applied Voltages
Now include both a DC and an AC voltage V = V0 + VRF cos(! t )
RF
so that
! JRF
" = " 0 + ! J0 t + sin(! RF t )
! RF
where ! J 0 = 2eV0 / h and ! JRF = 2eVRF / h .

Substituting V and ϕ in I = IJ sinϕ + V/R, we find after some

manipulations, using well known harmonic expansions*

+ ) ! JRF &
I = I J * J" ' $ sin [# + (! + " ! ) t ]
0 J0 RF
'! $
" = ,+ ( RF %
side band frequencies
V0 VRF
+ + cos ! RF t
R R
$ $
* sin(" sin x ) = # J! (" ) sin(!x ), cos(" sin x ) = # J! (" ) cos(!x )
! = %$ odd ! = %$ even
8
 Shapiro Spikes
The AC voltage generates a current response at the sideband
frequencies ωJ0 +νωRF. The DC part of the current is just V0/R
unless the the Josephson frequency matches a multiple of the
AC frequency, ωJ0 + νωRF =0. In that case we generate a DC
supercurrent |Is| = IJ Jν(ωJRF /ωRF). This is known as the inverse
AC Josephson effect. Even though there is a quantum
interpretation – the RF photons supply the energy needed to lift
a pair across the junction – the effect is really more subtle. In
I
particular the DC tends to zero as the RF power is increased
(since Jν ! 0 for all ν). In = V0/R
In the DC I-V characteristic
the supercurrent appears V0
at the so-called
Shapiro spikes as shown right. h# RF
"V =
0 =f $
RF 0
2e
ultra sharp spikes
(parts in 109)
! 9
 Gauge Invariant Phase
So far we have ignored the effect of the coupling of the change
to the vector potential. This coupling requires that we look for a
gauge invariant form of the phase ϕ. Recall that to obtain a
gauge invariant current we required (lecture 2)

2eA
!# " !# +
h
By integration we arrive at a gauge invariant generalization of
the phase ϕ

& 2e ) 2
" = (#1 $ #2 ) % (#1 $ #2 ) $ ( + - A , ds
' h * 1

This has major consequences for a wide weak link and for two
weak links in an applied field. Here we consider two weak links
used
! in the design of a SQUID.
10
 Macroscopic Quantum Interference Between Two Weak Links:
Matter Field Interferometer

Ia a

Itot flux φ
(1) (2)

Ib b

• Phase change ϕ12 from (1) to (2) is given in two ways

2e
$12 (path a) = $ a # ! A " ds ϕa is phase change
h path a across junction a,
2e & ϕb is phase change
$12 (path b) = $ b # ! A " ds across junction b
h path b

• Since ϕ12 (path a) = ϕ12 (path b), we get

2e 2e %
"a # "b = # & A $ ds = # % = #2'
h h %0
11
 Ia a

Itot flux φ
(1) (2)

Ib b

• Define ϕa + ϕb = 2ϕave , so that

e e
" a = " ave # ! " b = " ave + !
h h
• The total current Ia +Ib is then

e e
Itot = I J sin(! ave * " ) + I J sin(! ave + " )
h h
) " &
= 2I J cos' # $ sin !
ave
' " $
( 0%
12
• The critical Josephson current for the pair of links is
!
Ic = 2I J | cos(" )|
!0
• Ic oscillates with φ with period equal to the flux quantum φ0.
Ic

φ
φ0
• Analogy to interference from a pair of Young slits, but now for
matter waves instead of light waves.

• Superconducting Quantum Interference Device or SQUID:

high-sensitivity measurements of magnetic fields, voltages
and currents in the fT, fV and fA ranges, respectively.
13
 SQUID Applications
The device is highly sensitive: under ideal conditions one can
measure a change of 10-6 φ0/√Hz. SQUIDs are used as precision
magnetometers in the examples below:

Picture credits: J. R. Kirtley; 4-D

Neuroimaging; Quantum Design
Magneto-
encephalography
measures tiny magnetic
fields (fT range) created
by active areas in the
brain

Scanning SQUID
microscopy
for exotic experiments Magnetic properties
on high-Tcs measurement system
(more later…) for susceptibility
measurements etc. – can
detect moments down to
~10-13 Am2 14
 Other Applications of the Josephson Effect

The exactness of the Josephson frequency-voltage relation

ωJ = 2eV/h has led to the adoption of Josephson junction arrays
as the primary voltage standard: an incident RF field is tuned to
match the Shapiro steps of the array. Frequency can be
measured highly accurately, and the Shapiro steps are extremely
!
sharp, giving a relative voltage uncertainty of 1 part in 109.
This is one out of several quantum standards that have
revolutionized metrology, the quantum Hall effect resistance
standard being another prominent example.
Some other applications include microwave detectors and
frequency mixers – exploiting the strong nonlinearity and the
sideband generation of the weak link, respectively. A lot of
research effort is presently going into developing super-
conducting transistors and quantum computers using
superconducting qubits based on circulating currents and
enclosed flux in weak link circuits or non-analytic anyons
in exotic pairing states (more later …).
15
 Appendix 1: Short Quasi-1D Superconducting
Weak Link
ψ(0) = 1 " (d ) = e # i!
Assume β =−α : x
0 d
• for a weak link, 0<x<d<<ξ , the first GL equation
2
# 2" ' '+" (1 ! " ) = 0
reduces to ψ” ≅ 0 → ψ = a + bx so that from boundary
x
conditions " =1# (1 # e # i! )
d
sin !
The current is proportional to $ Im (" * " #) =
d
• For a strong link, 0 < x < d >> ξ , the first GL equation gives

ψ (1 - |ψ|2) ≅ 0 → ψ = exp(-iϕx/d)

(instead of sin ! )
!
so that $ Im (" * " #) =
d d 16
 Appendix 2: Simplified Treatment of Combined DC
and AC Applied Voltages
! JRF
• IS =I J sin [! J 0 t + sin (! RF t )] if ϕ0 = 0
! RF
A
B
= I J [sin A cos B + cos A sin B]
B+…
• The term (cos A) B is proportional to
cos(! J 0 t ) sin(! RF t )
1 1
= sin[(! RF + ! J 0 ) t ] + sin[(! RF " ! J 0 ) t ]
2 2
2eV
becomes DC if ! RF = ! J 0 =
h
• Thus, we will get a DC Josephson effect when
h
V = ! RF
2e
17
 Appendix 3: Current Biased Weak Links
In practice, it is usually the current rather than the voltage
that is controlled in a weak link. We need to invert the
Josephson phase equation. Suppose first that we apply a
DC current I0 to the RSJ. Then from p. 5 the phase ϕ is
given by
h #!
= I0 " I J sin !
2eR #t
If |I0| ≤ IJ, the phase reaches an equilibrium value given by
∂ ϕ /∂ t = 0, i.e., I0 = IJ sin ϕequil . Note that ∂ ϕ /∂ t = 0
means V = 0. If |I0| > IJ such an equilibrium is not possible
and ϕ keeps changing with time.
|I0| > IJ:
rolling, rolling,
rolling…
|I0| · IJ:
equilibrium ϕ
ϕ
18
It can be shown that for a given I0 the phase ϕ satisfies
" V ! t
I0 tan = I J + 0 tan J
2 R 2
where V0 is the mean voltage across the link
defined by I0
I0 = I 2 + V 2
/ R2 IJ
J 0

This gives the current-bias I-V

characteristic shown on the right. V0
– IJ

The I-V characteristic under current bias and under voltage bias
are therefore quite different.
Generally, the weak link equations have to be integrated
numerically.

19
 Last lecture (#5):

We discussed the peculiar properties of superconducting weak

links. The Josephson phase relation Is = IJ sinϕ led to remarkable
non-linear I-V phenomena with important applications.
We have now covered some of the major consequences of the
Ginzburg-Landau theory of the superconducting state. A
fundamental quantity that emerge is the flux quantum φ0 that
enters the study of vortices and the Josephson effect

φ0 = h/2e = V/fJ = 2.067 833 636 10-15 Volt/Hz

where fJ is the Josephson frequency and V the junction voltage.

Recall that the starting point was quite general, and did not
involve microscopic details. It may come as no surprise therefore
that we can apply at least some of what we have learned to the
related phenomena of superfluidity.
1
 Lecture 6:

• Phenomenology of superfluidity
• The superfluid order parameter
• Superflow in response to pressure and temperature gradients
– the Josephson-Anderson phase relation and the Euler
equation
• Two fluid model and the fountain effect
• Flow quantisation and vortices
• Excitations: phonons & rotons
• Landau’s critical velocity

• Literature: Annett ch. 2, Pethick & Smith chs. 9 & 10

2
 Superfluidity
Examples of superfluidity:
• Liquid 4He below 2.17 K
• Liquid 3He below 2 mK
• Cold atoms in the nK range
• neutron stars (probably)
• hydrogen under extreme pressures (possibly)

In this lecture we will concentrate on 4He – we’ll discuss the more

intricate case of liquid 3He later.
4He
What’s so special about He?
4

Consider the zero point energy:

h2 ! 2
E ZP # 3" ( )
2m d
d

3
For liquid 4He, d = 3.8 Å → EZP = 1.2 meV in this
approximation, close to the depth of the attractive potential
between two 4He atoms. Zero point motion prevents 4He from

Picture credits: after Buckingham & Fairbank

solidifying except at high pressures. It remains a quantum fluid
down to the lowest temperatures investigated.

Just below 4.2 K 4He liquefies into a normal

quantum fluid known as He-I. An early sign of
superfluidity is the so-called λ transition in the
specific heat at 2.17 K.

Below this transition, the liquid can flow

through capillaries without an applied pressure
difference. i.e., the viscosity is zero, η = 0. This
is a superfluid state known as He-II.

p1 p2 = p1
Tλ = 2.17 K
4
 The Superfluid Order Parameter
The specific heat anomaly indicates that He-II is an ordered state
as is the superconducting state and the vanishing viscosity in He-
II is similar to the vanishing resistivity in a superconductor. In
analogy with superconductivity, we therefore assume that the
order parameter is again a complex field ψ (r, t).

At first sight the Ginzburg-Landau description seems simpler than

in charged systems because the vector potential drops out.
However, the coherence length turns out to be so small (ξ(0) of
the order of an Å) that spontaneous fluctuations in the order
parameter, ignored in this course, are essential for
understanding, e.g., the λ heat capacity anomaly near Tc.

Importantly, the form of the current-phase relation remains

unchanged, i.e., the superfluid mass current in given by

h h
Js = (# * $# % #$# *) = " s $! = " s v s , # = " s e i!
2im m

where ρs & vs are the superfluid mass density & velocity. 5

 h
vs = !" (where m = m4He )
m

For superconductivity, we took the spatial and time derivatives of

this to arrive at the screening and supercurrent acceleration
equations, respectively. Let’s try the same here. Taking the
spatial derivative (the curl) we get the superfluid equivalent of
the 2nd London equation

curl v s = 0

i.e., the superflow is irrotational, a strange property.

As regards the time derivative, we replace the electric potential

energy qV with the chemical potential µ as follows:
#!
recall that for a supercondu ctor h = "qV (q = "2e)
#t
#!
thus for a superfluid h = "µ
#t
6
This is the Anderson-Josephson phase evolution relation, which
yields the time evolution of vs as
$v s $ h $!
m = m ( #! ) = #(h ) = "#µ
$t $t m $t

The gradient of µ can be expressed in terms of p, T and vs via (i)

the Gibbs-Duhem relation for an equilibrium system
V S
dµ = dp ! dT
N N

and (ii) the Bernoulli effect for a system in dynamic flow

1 2
p" p+ !v s
2
N is the number of particles, V is
the volume, S is the entropy, p is
the pressure and T is the
temperature.
7
This leads to the superfluid equivalent of the 1st London equation

#v s !(v s2 ) !p
=" " + $ !T
#t 2 %

where σ is the entropy per unit mass. It is known as Euler’s

equation, which is the Navier-Stokes equation in the limit of zero
viscosity.

Historically Helium-II has been described in terms of the two-

fluid model: a superposition of a pure superfluid with zero
entropy and zero viscosity and a normal fluid that gives rise to σ
in the above equation and η (not included).

8
 Two Fluid Model
A key experiment supporting the two-
fluid model shows that a torsional

Picture credits: Andronikashvili/S. Yuan

oscillator in He-II has a T-dependent ρ /ρ
frequency, and hence a T-dependent s
effective rotating mass. This can be
viewed as arising from the effect of a
normal-fluid boundary layer dragged by ρ /ρ
n
the oscillator. The normal-fluid
concentration ρn is 0 at T=0K and grows
to total concentration ρ=ρs+ρn at Tλ:
T(K)
Density Velocity Entropy Viscosity
Density
Superfluid ρs vs 0 0
Normal fluid ρn vn σρ η

The normal fluid corresponds to a system of elementary

excitations. This model has some peculiar thermo-mechanical
effects. We consider one example below: the fountain effect.
9
 Fountain Effect

Picture credits: J. Allen / Virtual Helium Web Site

In a nearly blocked porous plug (figures below) only the
superfluid can flow while the heat flow is blocked. This
is enough to lift the superfluid against gravity and force it
out at the top of the vessel, as seen in the spectacular
fountain effect. It is an unusual example of osmosis.

10
 Flow Quantization
Recall the problem of flux quantization
in a superconducting ring. The
corresponding problem in a superfluid
leads to quantization of circulation, κ.
Since the superfluid is neutral, the vs
vector potential does not enter and
using an analysis similar to that in
lecture 2, we find:
h h
v
! s " ds = ! # $ " ds = %$
m m
circulation κ 2π n

h Integration
=n
m path
quantum of circulation κ0
h
$ = n$ 0 $0 = # 1.0 " 10! 7 m2 / s for 4 He
m
n is an integer 11
 Superfluid Vortices
Next, what about vortices? Let’s consider superflow around a
vortex core with circulation equal to one quantum κ0. The
superfluid velocity vs is then given by

" 0 = # v s $ ds = v s (r )2!r
"0
v s (r ) =
2! r vs

There is no screening (effectively λ → ∞) and the superfluid

vortex core is of the order of an Å and thus very small compared
with that of conventional superconductors.

Recall that in a superconductor vortices can be induced by

applying a magnetic field. In a superfluid vortices can be
generated by rotating the cryostat.
12
 Elementary Excitations of Liquid 4He
The frequency spectrum of elementary excitations observed by
neutron scattering is shown in the figure.

Picture credits: H. W. Jackson

There are three main features rotons
which we’ll discuss in turn: phonons

ANGULAR FREQUENCY
ck
ωk
(i) the linear phonon regime
where ω =c k;
cmink
(ii) the peculiar roton region;

(iii) and the Landau critical

velocity cmin .

13
 Phonon and Rotons

• Phonons: quanta of density fluctuations of the fluid, with sound

Picture credits: hunter@MonkeyView

velocity c = 238 m/s in the low-T limit. Phonons are the sole
low energy excitations – the single particle parabola, ω =
ћk2/2m, is absent (the proof will be given in the next lecture).

• Rotons: the spectrum ω vs. k bends over

and goes through a pronounced minimum
near 2π/d. Note that in the extended zone
scheme of a periodic solid ω ~ 0 near this
same wavevector. Detailed calculations of
the flow of 4He atoms show that rotons
may be usefully described as microscopic
Roton
smoke rings (figure right).

14
 Dissipationless Flow
Now that we know the spectrum of the low-lying excitations, let’s
see whether we can excite them in the simplest possible way: by
dragging an object through the superfluid at velocity v. This

Picture credits: H. W. Jackson

induces a potential U(r, t) = U(r–vt), with 1D Fourier transform

i#t " ikx

Uk ,# = !! dxdt U ( x " vt )e e let s = x ! vt , ds = dx

= ! ds U(s)e " iks ! dte i#t e " ikvt = 2%U k $ (# " vk )

This means that the dragged object ck

ANGULAR FREQUENCY
can only create excitations that
conserve momentum & energy, ω = ωk
vk. If this line does not intersect the
dispersion curve, in particular for v < cmink
cmin (lower dashed line in the figure)
vk
no excitations can be created and the
flow is dissipationless.
15
 Landau Critical Velocity
A finite cmin, the Landau critical velocity, is necessary for
superfluidity. (Note that for a free particle parabola ω = ћk2/2m,
cmin vanishes and superfluidity is not possible). For He-II, cmin is
less than the sound velocity c, because of the roton minimum. A
further reduction of cmin comes from the creation of vortices
under certain flow conditions. Note that cmin is the critical flow
velocity of the superfluid beyond which the viscosity becomes
finite.
This may seem unfamiliar, but similar ideas arise in other

Picture credits: B. Samuelson

contexts from the propagation of an electron in an
electromagnetic vacuum to subsonic/supersonic flight.

BOOM!
v < c – silent v≥c– 16
 Appendix 1: Fountain Effect
Both the superfluid and normal fluid components carry particle
currents, but only the normal component carries heat current.

What happens if we connect two reservoirs as shown below?

T1 T2
p1 r p2

Assume steady state

#v s "p
=0=! + % "T or p1 ! p 2 = $% (T1 ! T2 )
#t $

A temperature difference implies a pressure difference.

17
Consider two extremes:

Picture credits: J. Allen / Virtual Helium Web Site

• A single large vessel of He-II: r → ∞. Bulk He-II is a very
good thermal conductor and effectively acts as a thermal
short so that rp never has a chance to build up.
• A nearly blocked capillary (or porous plug, figures below):
r → 0. Only the superfluid can flow while the heat flow is
blocked. This is enough to lift the superfluid against gravity
and force it out at the top of the vessel, as seen in the
spectacular fountain effect. It is an unusual example of
osmosis.

18
 Appendix 2: Vortices in Rotating Superfluid
But how many vortices do we get in equilibrium? Rotating a
cylinder of radius r at angular frequency ω induces a circulation

# = ! v s " ds = % r 2$ r = n # 0

Picture credits: E. Thuneberg

round the outside which must correspond to n = 2πr2ω /κ0
vortices in the fluid, each with a single quantum of circulation.
The superposition of vortices (figure) simulates bulk rotation,
and we get a parabolic meniscus just like in a rotating bucket of
water.
n vortices

In turbulent superflow it is
also possible to generate
vortex rings and tangles.
The study of their dynamics
ωr is an active field of research.

2r 19
 Appendix 3: Second Sound
Besides conventional density fluctuations we also have in
He-II, second sound: the motion of the normal fluid and

Picture credits: hunter@MonkeyView

superfluid yields a system with two coupled degrees of
freedom. In ordinary sound ρn and ρs oscillate in phase. In
second sound ρn and ρs oscillate in antiphase with a velocity
different from that of ordinary sound. Second sound may be
described as a wave in the phonon gas at finite
temperatures (and thus does not correspond to an
elementary excitation of the superfluid in the usual sense).

ρ ρ
normal fluid: ρn excess
normal fluid: ρn
entropy
superfluid: ρs superfluid: ρs transport

First sound: density wave Second sound: thermal wave

20
 Last lecture (#6):
We discussed superfluid 4He and its properties in analogy with
superconductors:
•  zero viscosity vs. zero resistance
•  phase evolution driven by Δµ = m(Δp/ρ – σΔT) vs. voltage ΔV
•  irrotational flow vs. Meissner screening
•  Euler’s equation vs. 1st London equation
•  flow quantization vs. flux quantisation
plus a number of effects in superfluids such as: the fountain
effect, phonons and rotons, critical velocity.

Today we will deal with a simpler superfluid formed

by a system of dilute atoms at ultra-low
temperatures. The weak effective interaction
among the atoms allows us to develop a simple
microscopic theory of macroscopic quantum
coherence and superfluidity.

1
 Lecture 7:

I. Bose-Einstein Condensates (BEC)

A.  Bose-Einstein statistics
B.  Ultra-Cold Atomic Gases
C.  BEC with Weak Interactions

II. Coherent States and 2nd Quantization

A.  Ladder Operators for
Simple Harmonic Oscillator
B.  Coherent States

Literature: Annett chs. 1 & 5, Pethick & Smith chs. 2, 4, 6 & 8

2
 I. Bose-Einstein Condensates (BEC)
IA. Bose-Einstein Statistics

The average occupation number of a single-particle microstate of

energy ε for a gas of bosons in equilibrium with a bath at
temperature T and chemical potential µ is given by the Bose-
Einstein distribution function

1
f (ε ) =
e(ε − µ ) / kBT − 1

For f (ε) to remain positive and non-singular, µ must be lower than

the lowest lying energy level, which we set to zero. If g(ε) is the
density of states, the particle density in the system can be
expressed as
∞
n = n0 + n n = n0 + ∫ f (ε )g(ε )dε
ε →0
where n0 is the density of particles in the ground state, namely
the Bose condensate density.
3
For a 3D gas of spinless atoms of
mass m f(ε) Density of
m3 / 2 ε f(ε)g(ε)
States g(ε)
g(ε ) =
2 π 2 3

and so for µ → 0 Energy ε

3 /2
m3 / 2 3 / 2
∞
u du & mkBT #
n = n0 + (kBT ) ∫ u = n0 + 2.612 $$ !!
2 3 2
2π  0e −1 % 2π  "

2.315
As the temperature rises, so does nn, and the ground
state density n0 = n – nn → 0 at T → Tc given by

2 /3
2π 2 & n #
Tc = $ !
kB m % 2.612 "
4
Liquid 4He: inserting the value of n, corresponding to an atomic
separation of 0.36 nm, and of m we get Tc ≅ 3.1 K, close to the
λ-point. He-II may be thought of as a strongly interacting
Bose-Einstein condensate (BEC) in which the interactions
among atoms reduces the condensate fraction by at least an
order of magnitude from that of the ideal (non-interacting) BEC.
The interactions lead to superfluidity and rotons.

Systems of Dilute Cold Atoms: typically alkali atoms of spin 1

or 2 with densities corresponding to atomic separations in the
range 100-1000 nm and Tc < 1 µK. The interactions are
reduced to the point where the condensate fraction is close to
the ideal value for a BEC and the effects of interactions can be
described with simple models. Even in these systems
interactions are not negligible and lead to superfluidity as in
liquid 4He. (Rotons do not seem to arise in the limit of weak
interactions.)

5
 IB. Ultra Cold Atomic Gases
The central ideas behind cooling atom systems to sub-µK

Picture credits: K. Y. M. Wong / Physics2000

temperatures are explained intuitively on
www.colorado.edu/physics/2000/bec/evap_cool.html.

•  Cooling with a beam of photons

of frequency matched to an
atomic resonance.

•  Laser and Magnetic trap: need

bosons with strong atomic
transitions and with magnetic
moments, e.g., atoms with odd
number of nucleons such
as 7Li, 23Na, 87Rb.

•  Evaporation cooling in a
laser and magnetic trap.

6
How can we verify BEC? Look at k-state distribution by

Picture credits: Physics2000 (Weiman/Cornell) / W. Ketterle

1.  switching off the magnetic trap;
2.  taking a picture of the atom cloud a short time t0 later.
Distance moved from the centre is kt0/m, so the picture directly
shows us the momentum distribution and the macroscopic n0.
The final (low T) distribution is not a delta function in k-space
because the finite trap size gives rise to a spread in k due to the
Heisenberg uncertainty principle.

Evidence of vortex lattice

in cold cold atomic gas
The existence of vortex lattices means that even cold atomic
gases are superfluids with parallels to liquid helium. 7
 IC. BEC with Weak Interactions
In dilute atomic gases the effective two-particle interaction is
modelled by a weak localized potential energy

V (r − r' ) = gδ (r − r' )

where g is related to the s-scattering amplitude.

In the Hartree approximation an atom feels the mean potential
produced by all other atoms so that the one-particle eigenstates
and eigenvalues are given by
( 2 2 %
&− ∇ + gn(r )#ϕ i = ε i ϕ i
&' 2m #$
where
n(r ) = ∑ f (ε j ) | ϕ j (r ) |2
j
This is a self-consistent pair of equations for ϕi and εi known as
the Gross-Pitaevskii equations. The solutions, including the
effects of the magnetic trap, are obtained numerically and show
a form of BEC with a finite critical temperature. 8
Comments:

•  Some of the most interesting aspects of the BEC arise from

interactions. In particular, interactions change the one-
particle spectrum from εk ∼ k2 to εk ∼ k at small k (next
lecture). This leads to a finite Landau critical velocity and
hence to superfluidity.

•  Strong interactions as in He-II appear to be needed to

produce the roton part of the excitation spectrum.

•  The Gross-Pitaevskii equation for ϕ0(r) is analogous to the

first Ginzburg-Landau equation for ψ0(r). However, ϕ0(r) is
the single-particle ground state while ψ0(r) represents the
superfluid order parameter. To understand the order
parameter more fully we need to introduce the creation-
annihilation operator formalism and coherent states.

9
 II. Coherent States and 2nd Quantization

IIA. Ladder Operators for Simple Harmonic Oscillator

Picture credits: R. G. Griffin & K. Nelson (M.I.T.)

Recall that the quantum harmonic oscillator
can be completely described in terms of the
ground state ϕ0 (a gaussian) and the ladder
operators a+ and a defined by
a+ϕn = n + 1 ϕn +1 aϕn = n ϕn −1
ϕ3(x)
It follows that
a+a ϕn = n ϕn aa+ ϕn = (n + 1) ϕn ϕ2(x)

n̂ ϕ1(x)
[a,a+] = 1 and [a,a] = [a+,a+] = 0
ϕ0(x)
The Hamiltonian operator and eigenstates
are then given by
1 1
ˆ = ω (n
H ˆ+ ) ϕn = (a+ )nϕ0
2 n!
10
 Ladder Operators and Momentum-Position Operators

In terms of the dimensionless momentum and position variables

1 mω
P = p, X = x, H = ω(P 2 + X 2 )
2mω 2
the ladder operators arise from the transformation

A = X + iP A* = X − iP

which corresponds to a rotation in the complex plane.

P
A
ˆ= X
a= A ˆ + iP
ˆ

X ˆ* = X
a+ = A ˆ − iP
ˆ

A*
11
Consider the eigenvalue equation for the lowering operator, a

aα = α α a = α a α = α = α e iθ = X
ˆ +i P
ˆ

While the eigenstates of P̂ have zero uncertainty in P but infinite

in X, and vice versa for X̂ , the eigenstates of a have equal
minimal uncertainties (standard deviations) in both X and P, i.e.,

P P

P̂
α Fixed radius
α
Δθ
equal to ½ for all α

θ

X X
X̂

12
 IIB. Coherent States for a Harmonic Oscillator

Coherent states were introduced in the 1920s in an effort to link

the quantum and classical pictures in the limit of large n. Such
states have applications in quantum optics to describe coherent
photon states and, as we shall see later, to describe the related
problem of coherent boson states. Here we consider the
definition of coherent states of a quantum harmonic oscillator. A
coherent state is a wavepacket made up of all the ϕn , which
produces the narrowest spread for a given average energy
< n̂ > ω . In general, a coherent state (known a Glauber state in
quantum optics) is defined by a wavepacket of the form

∞ α nϕ n α a+ −|α|2 / 2
α =C ∑ = Ce 0, C =e
n =0 n!

where α = |α|eiθ is a complex number and |0> = ϕ0.

13
Energy eigenstate of a
harmonic oscillator for
n = 7: the amplitude is a
stationary function of x.

Coherent state of a harmonic

oscillator for < n̂ > = 7: the
amplitude of wave packet of
fixed width (independent of
< n̂ > ) moves back & forth
between the classical limits at
the harmonic oscillator
frequency. The width
vanishes relative to the
classical limits for large
< n̂ > . 14
Summary of Properties of Coherent States:

•  a|α> = α|α>, thus a coherent state is an eigenstate of the

lowering operator with eigenvalue α = |α|eiθ .

•  The probability distribution over the states n is Poisson

2n −|α|2
Pn = | α | e / n!
ˆ > = var n = Δn2 = |α |2
<n so that α = ˆ > e iθ
<n

•  Substituting α = |α|eiθ in the definition of |α> we find

− i∂ | α > / ∂θ = n
ˆ|α > or n
ˆ = − i∂ / ∂θ

Thus, θ and n̂ are conjugate quantities. In analogy to

momentum and position we expect

Δn Δθ ≥ 1 / 2
15
•  To see how the phase relations between ϕn states are
preserved in a coherent state go to
www.phys.uri.edu/∼yoon/qhomain.html.

•  Coherent states are not orthogonal and form an over-complete

set (note that the operator a is not hermitian).

Derivation of Properties of Coherent States:

•  Normalization factor C: from the orthogonality of the energy

eigenstates, we have
2n
2
∞α 2 |α|2
< α |α > = C ∑ =C e =1
n = 0 n!
−|α|2 / 2
∴ C =e

16
• Exponential form:
αa +
∞ αn + n
e ϕ0 = ∑ (a ) ϕ 0 = C −1 α
n = 0 n!

since (a+ )n ϕ0 = n! ϕ n .

• |α> is an eigenvector of a with eigenvalue α:

∞ αn ∞ αn
aα =C ∑ aϕ n = C ∑ nϕ n −1
n = 0 n! n = 0 n!

∞ α n −1 ∞ αm
= Cα ∑ ϕ n −1 = Cα ∑ ϕm = α α
n =1 (n − 1)! m = 0 m!

17
• |α> is Poisson distributed over the energy eigenstates ϕn:
| α |2n −|α|2
[Probability that oscillator is in state ϕn] = e
n!
λn − λ
[Poisson distribution] = e
n!

where λ is the average of n and the variance of n (the two are

equal for a Poisson distribution). Thus, Δn = < nˆ>
ˆ > = | α |2
<n Δn = | α | Δn / < n
ˆ > = 1/ < n
ˆ>

This allows us to write

ˆ > e iθ
α = <n

a form analogous to that of the order parameter in the

Ginzburg-Landau model.

18
• The number-phase uncertainty relation: using α = | α |eiθ
rewrite |α> as n
∞ α e inθ
|α > = C ∑ ϕn
n =0 n!
The derivative ∂ /(i∂θ) pulls down a factor of n for a general
term n in the series and thus has the same effect as n̂,
since n
ˆϕ n = nϕ n . Thus, n̂ and θ are conjugate quantities

n
ˆ = ∂ / i∂θ and thus expect ΔnΔθ ≥ 1/2

This is analogous to the relation between momentum and position

operators p̂/ = ∂ /(i∂x), which leads to the usual Heisenberg
uncertainty relation (Δp/)Δx ≥ ½. For a coherent state with
lα l=Δn>>1, we showed explicitly that ΔnΔθ = ½ (p. 12).
Since Δn = < n ˆ > the phase uncertainty vanishes for large < n̂ > .
The general proof of the uncertainty relation ΔnΔθ ≥ ½ must take
account of the fact that unlike the position variable, the angle θ
is only defined in a finite range (0 to 2π).
19
•  Consider the eigenvalue equation for the lowering operator, a
aα = α α a = α a α = α = α e iθ = X
ˆ +i P
ˆ
While the eigenstates of P̂ have zero uncertainty in P but infinite
in X, and vice versa for X̂ , the eigenstates of a have equal
minimal uncertainties (standard deviations) in both X and P, i.e.,
2 2
H ˆ2 + P
ˆ / ω = X ˆ2 = X
ˆ ˆ
+ P + ΔX 2 + ΔP 2 = n
ˆ +1/2
2 2 2
n ˆ
ˆ = X ˆ
+ P = α , ΔX 2 + ΔP 2 = 1/2, or ΔP = ΔX = 1 / 2
2
Let nˆ = α >> 1, then α Δθ = n ˆ Δθ = ΔnΔθ = 1 / 2, since
here n ˆ= X ˆ2 + Pˆ2 , so that Δn2 = (2 X
ˆΔX )2 + (2P
ˆΔP )2 = n
ˆ
(n. b., ΔX , ΔP, Δn & Δθ represent standard deviations).
P P

P̂
α α Fixed radius
Δθ
equal to ½ for all α

θ

X X
X̂ 20
 Appendix: Cooling Ultra Cold Atomic Gases

Picture credits: K. Y. M. Wong / Physics2000

Stage I: Laser (Doppler) cooling:
the photon frequency is tuned just
below atomic transition so that
only fast atoms are slowed down
(by Δp = ω/c).

Stage II: Magnetic trapping of the atomic

magnetic dipole moments and evaporation
cooling to lose the highest energy atoms
(how a cup of tea cools).
21
 Last lecture (#7):

We discussed Bose-Einstein condensates and considered

examples in dilute atomic gases at ultra-low temperatures.
Even in these dilute systems, residual interactions cannot be
ignored. They lead to modifications of the spectrum of low-lying
excitations, to finite Landau critical velocities and to
superfluidity.

To go further we need a more microscopic description. This

leads us to coherent states for superfluids and the formalism of
second quantization. We made a start last time and in this
lecture we will develop a Hamiltonian in second quantized form
that is central to the study of interacting Bose as well as Fermi
systems, and introduce the Bogoliubov theory.

1
 Lecture 8:
I. Creation and Annihilation Operators and Coherent States for
Bose Systems
IA. Generalization of Harmonic Oscillator Formalism
IB. Coherent States for a Bose System

II. Second Quantization Forms of Quantum Fields and Hamiltonian

IIA. Kinetic Energy & Quantum Field Operators
IIB. Potential Energy, Density Operators & Hamiltonian

III. Introduction to the Bogoliubov Theory

Appendix A: Examples of States of a Bose System in the

Schrödinger and Occupation Number
Representations
Appendix B: Commutation Rules for Fermions

Literature: Annett chs. 2 & 5, Pethick & Smith chs. 8 & 10,
Waldram chs. 2, 9 & appendix
2
I. Creation and Annihilation Operators and Coherent States for
Bose Systems
IA. Generalization of Harmonic Oscillator Formalism

• Photons: the electromagnetic field can be represented in terms

of modes with wavevector k, each of which behaves as an
oscillator of frequency ω = ck. For each mode we introduce ak
and ak+ which annihilate and create, respectively, a photon of
wavevector k. For a given k, the commutation rules for these
operators are the same as for the lowering and raising
operators of a harmonic oscillator. The annihilation and creation
operators for different k’s all commute.

• More General Bose Systems: the above treatment for massless

Bose particles (photons) extends to massive Bose particles
when the boson number is not conserved, as in a grand
canonical ensemble. All of the properties of the creation and
annihilation operators carry over. The form of the single-
particle energy spectrum and the (now finite) value of the
chemical potential are the main changes.
3
 • The quantum states are described in terms of the occupation
numbers in basis states of definite one-particle occupation
| nk1 , nk2 ,... > ! | n1, n2 ,... >
+
The creation operator ak increases ni by unity and multiplies
i
the state by in + 1

ak+ | n1, n2 , ..., ni , ... > = ni + 1 | n1, n2 , ..., ni + 1, ... >

i

Similarly, ak reduces ni by unity and multiplies the state by ni

i

aki | n1 , n2 ,..., ni ,... > = ni | n1 , n2 ,..., ni " 1,... >

# a+ aki | n1 , n2 ,..., ni ,... > = (ni " 1) + 1 ni | n1 , n2 ,..., ni ,... >

k i
ni
)
Thus, nk = ak+ ak is the occupation number operator.

All other relations follow in a similar way by analogy to the

! simple harmonic oscillator model.
4
• The commutation rules for bosons can be summarized as
follows:
[ak,ak’+] = δkk’, [ak+,ak’+] = [ak,ak’] = 0

• The occupation number formalism is known as second

quantization. It automatically includes particle indistinguish-
ability. The statistics of the particles is contained in the
commutation rules. So far, we have considered only bosons
that have the same commutation rules as the ladder operators
of a harmonic oscillator. For fermions the commutation rule

[A,B] = AB - BA

is everywhere replaced by the anticommutation rule

{A,B} = AB + BA

This crucial change incorporates the Pauli exclusion principle.

5
 IB. Coherent States for a Bose System

With the above generalizations we arrive at coherent states of a

boson system of the form

#$|" ki |2 / 2 % (
| " k0 , " k1 ,... > =e exp'$ " ki a+ * | vac >
& ki )

Now!|vac> is the vacuum state with no bosons present.

In an ideal laser, for example, one of the modes of the

electromagnetic field, say ks, has macroscopic occupation,
)
< nks > = | " ks |2 , while the others may be ignored.

The application of this coherent state concept to superfluids will

be discussed in the next lecture.
!

6
II. 2nd Quantized Forms of Quantum Field & Hamiltonian
IIA. Kinetic Energy & Quantum Field Operators
In order to use the 2nd quantization formalism, we need to write
the Hamiltonian in terms of creation and annihilation operators
(as we did for the simple harmonic oscillator with the
corresponding ladder operators). The kinetic energy operator is
) )
T = # " k nk
k
)
n
Where k = ak+ak is the number operator and εk is the one-particle
energy corresponding to the one-particle basis state ϕk. The
)
expectation value!of the kinetic energy is thus Σk εk < k >.
n

To
! write the potential energy in 2nd quantized form it is useful to
introduce the Fourier transforms of the operators ak & ak , i.e.,
+
) )
the quantum fields " + (r ) & "(r ) in a volume V
!

)+ )
" (r ) = 1 % e#ik$r ak+ "(r ) = 1 % eik$r ak
V k V k
)+ ! )
" (r ) creates a particle in a position eigenstate and (r )
"
annihilates a particle in a position eigenstate. 7
!
 IIB. Potential Energy, Density Operators & Hamiltonian
For a binary interaction, g(r), dependent only on the separation r
of the interacting particles we may write the potential energy of
the system of identical particles as

V = 1 $ g(ri " r j ) = 1 $ gq ( %*q %q " N ), where

2 2 V
i# j q

gq = # drg(r )e"iqr , $q = 1 # drn(r )e"iqr & n(r) = & %(r-ri )

V V V i
! g is the Fourier transform of g(r) and ρ is the Fourier
Here q q
transform of the particle density n(r). To find the 2nd quantized
! form of the potential energy we replace n(r) by the operator
) ) )
n(r ) = " + (r ) "(r )
) )
so that #q+ = 1 $ ak+q+
ak and #q = 1 $ ak+ %&q ak %
V k V k%

+
The operator ak + q ak annihilates a particle at k and creates a
particle at k+q. This increases the momentum by h q, but does
! not change the particle number. The sum of such operators over
all k represents a collective mode, i.e., a density fluctuation. 8
From the above density operators and the commutation rules we
find that the potential energy operator in 2nd quantized form is

) 1 )
)+ )
V = $ gq ("q "q # ) = 1 $ gq ak+q
N +
ak+ %#q ak %ak
2 V 2V
q kk %q

We have used two permutations of the single-particle operators,

!one of which leads to a constant term that cancels the) self-
interaction correction term (the term proportional to N / V ). If we
use anticommutation rules the sign changes )that arise in each
permutation cancel out so that this form of V holds for both
bosons and fermions. In general we must also! include spin
indices. Note that if g(r) is a contact interaction

g(r) = g! (r ) then gq = g
!
For simplicity we shall use the latter approximation that is
expected to be valid for scattering processes in the limit of small
momentum transfer h q. More general forms of the interaction
will be introduced in the final lectures. 9
Thus, the 2nd quantized form of our model Hamiltonian is

) g
H = % " k ak+ ak + +
% ak+q ak+ #$q ak #ak
2V
k kk #q

k’ k’ – q
!
k k+q
Including spin indices, this is relevant to, e.g., (i) the Gross-
Pitaevskii and Bogoliubov theories for superfluids (bosons, g > 0),
(ii) the BCS and Bogoliubov theories for conventional supercon-
ductors (fermions, g < 0 near kF), & (iii) the Hubbard model for
strongly correlated electron systems (e.g., high Tc cuprates).

The kinetic energy is diagonal in the single-particle operators,

whereas the potential energy is diagonal in the density operators.
One might expect that a superposition of such operators may be
needed to diagonalize the full Hamiltonian. This leads us to the
Bogoliubov theory.
10
 III. Introduction to the Bogoliubov Theory

If g = 0 the ground state is the simplest BEC state in which all

particles are in the same single-particle eigenstate, which we
take to be ϕ0 (i.e., the k = 0 state). The density of particles n0
= N0/V in state ϕ0 is then just the total density n = N/V. For
non-zero but small g we show that the ground state of the Bose
system can be approximated by a coherent state (for which, )
recall, the total number
) of particles is not fixed). N0 =< 0 > is
N
now less than N=< N > but still macroscopic (N0 >>1).

Since N0 >> 1, the operators a0+ and a0 effectively

! commute
(the difference
! between N0 and N0+1 being ignorable) and thus
become c-numbers, i.e., both can be replaced by N0

a0+ ! s " N0 ! s , a0 ! s " N0 ! s

In this limit lΨs > is thus approximately an eigenstate of the

annihilation operator, and hence approximately a coherent state.
11
This means the density operator can be approximated as
)
"q+ # 1 (aq+ a0 + a0+ a$q ) = n0 (aq+ + a$q )
V
We have kept terms that dominate as N0 → ∞. This is an
example of the random phase approximation. Interestingly, a
density creation operator in this case reduces to an equal super-
!
position of a particle creation and a particle annihilation operator.

With the above density operator the Hamiltonian reduces to a

quadratic form in the single-particle operators* that can be
diagonalized by a linear transformation from the starting Bose
operators (ak, ak+) to new Bose operators (αk, αk+). The latter
correspond to density operators at small k, single-particle
operators at high k and superpositions of both in general.

An excited state can be described in terms of elementary

excitations with number density αk+αk and energy Ek. The goal
is to find the new Bose operators (αk, αk+) and new excitation
energy Ek in terms of the parameters of the Hamiltonian, εk & g.
This is done via the Bogoliubov transformation.
*We shall use another but equivalent approach in the next lecture. 12
 Appendix A: Examples of States of a Bose System in the
Schrödinger and Occupation Number Representations

Let {ϕk(r)} represent a basis set of one-electron orbitals (ignoring

spin for now) and let |vac> ≡ |0,0,0,…> be the vacuum state with
no particles. We build up states starting with |vac> via creation
operators as follows:

ak+ | vac > = 1 ! " k (r ) = | 0,0,..., (nk = 1),0,0,... >

(The prefactor nk + 1 = 1, since nk = 0 in | vac >)
1
(ak+ )2 | vac > = 2 " [! k (ra )! k (rb ) + ! k (rb )! k (ra )]
2!
where ra and rb are coordinates of particles a and b, respectively.
1
ak+'ak+ | vac > = 1 " [! k (ra )! k '(rb ) + ! k (rb )! k '(ra )]
2!
ak+ ak [(ak+ )2 | vac >] = (2 " 1) + 1 ! 2 [(ak+ )2 | vac >]
2 13
 • A more complicated example (now including spin in general)
might look as follows:
| 0,2,0,0,1,0,1,0,0,... >
# + &2 # + & # + &
% a ( % a ( %a (
n1! n2! ... $ k2 ' $ k5 ' $ k7 '
= ) "2(a)"2(b)"5(c )"7 (d ) = | vac >
N! permutations 2! 1! 1!

where n2 = 2, n5 = 1, n7 = 1 (all rest zero) and N = 4. In this

! example we have four identical bosons, two in state 2, one in
state 5 and one in state 7. ϕ2(a), for example, represents a
one-particle state in which particle a is in state 2 of definite
orbit and spin. The sum is over all permutations of the particles
(see Waldram, appendix).

14
 Appendix B: Commutation Rules for Fermions
For a system of fermions the many-particle wavefunctions are
antisymmetric and are described by Slater determinants. The
label k now includes both orbital and spin quantum numbers.
Consider, for example, two fermion states with k’ ≠ k

1 ! k (a) ! k (b)
ak+'ak+ | vac > =
2! ! k '(a) ! k '(b)

By convention, the new state k’ is fed from below in the Slater

determinant. Now inverting the creation operators, we get

1 " k '(a) " k '(b)

ak+ ak+' | vac > = = ! ak+'ak+ | vac >
2! " k (a) " k (b)

Thus, the operators for k ≠ k’ anticommute. Proceeding in this

way, we find that the commutation rules for bosons are replaced
by analogous anticommutation rules for fermions
{ak+ , ak '} = ! kk ' , {ak+ , ak+'} = {ak , ak '} = 0
15
The difference in the commutation rules arises from the symmetry
properties of the many-particle wavefunctions. The wavefunctions
are symmetric for bosons and antisymmetric for fermions.
To demonstrate the commutation rule for ak+ and ak consider the
effects of these operators on state k

Empty state ≡ |0> Occupied state ≡ |1>

εk
Energy = 0 Energy = εk

ak |0> = 0 ak+|1> = 0
where |0> means the state is unoccupied (empty) and |1> means
it is occupied. Next note that ak+ak|0> = 0|0> and ak+ak|1> =
1|1>. This means that ak+ak is the number operator n̂k . Also note
that akak+|0> = 1|0> and akak+|1> = 0|1>. This means
akak+ = 1 - n̂k , i.e., {ak,ak+}=1 as claimed.

To conclude, creation and annihilation operators anticommute

except for products of ak+ and ak’ for the same state, k’ = k. Note
that the factors nk and nk + 1 that appear in the definitions of
ak and ak+ for bosons can only equal zero or one for fermions.
16
“I remember that when someone had started to teach me
about creation and annihilation operators, that this operator
creates an electron, I said ‘How do you create an electron? It
disagrees with conservation of charge’.”
R.P. Feynman, Nobel Lecture

“In nonrelativistic quantum mechanics the use of second

quantization for electron states is purely a matter of notation.”
W.A. Harrison, Solid State Theory

“…the Schrödinger representation has an unnecessary

perversity. The particles are actually indistinguishable, but in
the Schrödinger representation we first label the particles as
though they were distinguishable and then promptly remove
the distinguishability by antisymmetrizing the function. It is
this antisymmetrization which introduces the algebraic
complexity. The occupation number representation, which
never labels the particles, is much simpler and more rational in
this respect,…”
J.R. Waldram, Superconductivity of Metals and Cuprates
17
 Important reminders:
I) The first supervision for the course will held at the following
times (in groups of 4 or 5):
1) 14.00, Friday, 17 February, Mott Seminar Room
2) 15.00, Friday, 17 February, Mott Seminar Room
3) 14.00, Wednesday, 22 February, Committee Room (Room 213,
Bragg Building)
II) Because I must attend a scientific meeting on superconductivity
in the USA, I would like to hold the last two lectures the week
of 5 March, that is:
1) Lecture 11 will be held at 11.00 on Wednesday, 7 March, Mott
Seminar Room
instead of Friday, 24 February
2) Lecture 12 will be held at 11.00 on Friday, 9 March, Mott
Seminar Room instead of Wednesday, 29 February
I hope that these changes work for you.
If there are any questions, please see me after the lecture.
18
 Last lecture (#8):
We developed the 2nd quantized form of the Hamiltonian central to
the microscopic theory of superfluidity and superconductivity, and
other problems in Bose and Fermi systems
" N̂% g
+ 1
Ĥ = ∑ ε ak ak + ∑ $ gq ρ̂q ρ̂q − '' → ∑ ε ak+ ak +
$ + +
∑ ak+q ak+ '−q ak 'ak
2 # V& 2V
k k q k k kk 'q

ρ̂q+ = 1 ∑ ak+q
+ a so that if N = Vn >> 1
k 0 0 k’ k’ – q
V k
ρ̂q+ ≈ 1 (aq+ a0 + a0+a−q ) = n0 (aq+ + a−q ) k k+q
V
The microscopic theory is based on the trial wave-function method.
The trial ground state wavefunction is taken to be a coherent state:
a BEC state for superfluids such as He-II and atomic gases and a
BCS state for conventional superconductors and for superfluids such
as liquid 3He. The elementary excitations above the ground states
are then found via the Bogoliubov theory. We treat superfluidity in
this lecture and then turn to superconductivity. 1
 Lecture 9:

I. Bogoliubov Theory
IA. Effective Hamiltonian
IB. Bogoliubov Transformation
IC. Elementary Excitations and Superfluidity
ID. Condensate Depletion

II. Connection to the Ginzburg-Landau Model

IIA. Order Parameter ψs(r)
IIB. Ginzburg-Landau Free Energy

Literature: Annett chs. 2 & 5, Pethick & Smith chs. 8 & 10,
Waldram chs. 2 & 9

2
 I. Bogoliubov Theory
IA. Effective Hamiltonian
Recall that in the limit of macroscopic occupation in the k=0 state
(i.e., N0 → ∞) the leading terms in the potential energy operator
will be those with the highest (even) powers of a0 ≈ a+ ≈ N 0 .
0
Retaining terms 2 and 4 order in these quantities we arrive at
nd th

a reduced or effective Hamiltonian of the form

2
1 " + + + + % N0
g
Ĥeff = ∑ $ (εk + 2n g)(ak ak + a−k a−k ) + n g(ak a−k + ak a−k ) ' +
2 # 0 0 & 2V
k ≠0
1 " + + + + % N 2g
= ∑ $ (εk + n g)(ak ak + a−k a−k ) + n g(ak a−k + ak a−k ) ' +
2 # 0 0 & 2V
k ≠0
We have used the boson commutation rules, the relation
ˆ=N
N ˆ0 + ∑ ak+ a , so that ˆ2 ≈ N
N ˆ02 + 2N
ˆ0 ∑ ak+ a
k k
k ≠0 k ≠0
and replaced the number operators by real numbers N & N0. The
Hartree-Fock chemical potential and energy spectrum are

µ HF = ng ≈ n0g, ξ k =εkHF − µ HF = εk + n0g

3
The anomalous aa and a+a+ terms in the effective Hamiltonian
give crucially important corrections to Hartree-Fock theory for a
boson system.

The effective Hamiltonian is a quadratic form that can be

diagonalized by a linear transformation from the Bose operators
(ak, ak+) to new Bose operators (αk, αk+) so that

ˆeff = E zp + ∑ E k α k+α k
H
k ≠0

where Ezp is the zero point energy and Ek is the excitation energy
spectrum of interest here. (For a discussion of Ezp see, e.g.,
Pethick & Smith p. 212.)

An excited state of the Bose system may be described in terms of

elementary excitations with number density αk+αk and energy Ek.
We need to find the new boson operators and then Ek in terms
of εk & g.
4
 IB. Bogoliubov Transformation

We might expect αk+ to be a linear superposition of a single-

particle creation operator ak+ and a density-fluctuation creation
operator ρk+. In our model, ρk+ is just a superposition of ak+ and
a-k and thus we assume that αk+ is of the form

α k+ = uk ak+ + vk a−k

where the coefficients uk and vk can be taken to be real and must

be adjusted to diagonalize Ĥeff .

This is the Bogoliubov transformation. As can be confirmed by

direct substitution in Ĥeff the solutions for these coefficients and
for the excitation energy spectrum are as given below (see
examples sheet 3 and, e.g., Pethick & Smith §8.1):

5
 2 1 ⎛⎜ ξ k ⎞ 2 1 ⎛⎜ ξ k ⎞
uk = ⎜ + 1⎟⎟ vk = ⎜ − 1⎟⎟
2 ⎝ E k ⎠ 2 ⎝ E k ⎠
2
E k2 = ξ k − (n0 g)2 , ξ k = ε k + n0 g
where ξk is the excitation energy measured from the chemical
potential in the Hartree-Fock approximation (p3). The excitation
energies in the Bogoliubov theory at low and high k are then

⎧ n0 g
⎪⎪ k ε k << n0 g
E k = ⎨ m
⎪ 2k 2
⎪⎩ ε k >> n0 g
2m

Thus, at low k (i.e., εk << n0 g) the spectrum is linear and the

elementary excitations are collective density fluctuations (uk ≈ vk)
with sound velocity c = n0 g / m . At high k (i.e., εk >> n0 g) the
spectrum is quadratic and the excitations are single-particle
excitations (uk ≈ 1, vk ≈ 0) with energy εk.
6
 IC. Elementary Excitations and Superfluidity

For finite g the model gives the spectrum Ek shown in the figure
below. The Landau critical velocity is finite and equal to the speed
of sound. The depression of the density of states of elementary
excitations due to the repulsive interaction turns a BEC into a
superfluid.
α k+ = uk ak+ + vk a−k
In this model the low k excitations
correspond to Goldstone modes Ek
that arise because of the breaking
of a continuous symmetry (see p.
14). Lattice vibrations in solids and free
spin waves in magnetic systems uk particle
arise for similar reasons. Though a
major improvement over Hartree-
Fock the Bogoliubov theory does vk
not predict the existence of rotons
expected to arise in the presence of k
strong interactions, as in He-II.
7
 ID. Condensate Depletion
Interactions reduce the condensate fraction n0/n in the one-particle
state ϕ0. This effect can be calculated from

ˆ = ∑ a+ ak = N
N ˆ0 + ∑ a+ ak
k k
k k ≠0
and rewriting ak+ and ak in terms of αk+ and αk of the Bogoliubov
transformation

ak+ = ukαk+ - vkα-k ak = ukαk – vkα-k+

In the product ak+ak the term vk2α-kα-k+ = vk2α-k+α-k + vk2

leads to a correction in N̂ even in the absence of thermal
excitations. Thus, in the ground state

N < N̂ > N0 1 # mc &3

= = + ∑ vk2 , or n = n0 + 1 %% ((
V V V V k ≠0 3π 2 $  '

8
The condensate fraction n0/n is thus less than unity unless the
speed of sound c → 0. In atomic gases n0/n is above 90%,
whereas in He-II it is of the order of or less than 10%.
n0

Single-particle
momentum distribution n - n0

Note that n0 should not be confused with ns in the two-fluid model

or in the Ginzburg-Landau model. In particular, the superfluid
current density is given by nsvs and not by n0vs. The particles N –
N0 should not be thought of as being in excited incoherent states.
They are bound up in a coherent ground state wavefunction along
with N0 and are thus “dragged” with N0 to produce the total
superfluid current at T = 0K (see, e.g., Waldram §2.2 and §9.4).
9
 II. Connection to Ginzburg-Landau Model
IIA. Order Parameter ψs(r)

In the absence of interactions all particles condense in the same

single-particle state is ⋅r
e
ϕ s (r ) =
V
where s is non-zero if the BEC carries a current. The many-
particle state for fixed particle number N is

|Ψs> = ϕs(r1)… ϕs(rN)

More generally we assume |Ψs> can be approximated by a trial

many-body state satisfying
ψˆ(r ) |Ψ s > = ψ s (r ) |Ψ s >
i.e., we assume that |Ψs> is a coherent state. Recall that a
coherent state is an eigenstate of the annihilation operator, so
that the latter can essentially be replaced by a c-number.
10
The eigenvalue, ψs(r), of ψˆ(r ) for this many-body state,
represents the macroscopic wave-function or the order
parameter in the Ginzburg-Landau model; ψs(r) reduces to the
single-particle state ϕs(r) only in the absence of interactions.

The reasons for choosing a coherent state for the trial

wavefunction are that

•  it gives the correct state in the absence of interactions;

•  it involves a superposition of states of different number of

particles as is appropriate for a grand canonical ensemble;

•  and it allows us to evaluate the Hamiltonian in a

+
straightforward way since the operators ψˆ (r ) and ψˆ(r )
reduce to the c-numbers ψs*(r) and ψs(r) at each r.

Recall that even a crude trial wavefunction can produce

accurate energies. We look for internal consistency of the
theory and consistency with experiment.
11
 IIB. Ginzburg-Landau Free Energy

To obtain the Ginzburg-Landau model we evaluate a free energy

for the superfluid defined as
ˆ − µN
Fs = < Ψs | H ˆ |Ψs >

From the relations between (ψ(r), ψ+(r)) and (ak, ak+) we easily
confirm that for εk = 2k 2 / 2m the Hamiltonian can be written as

2k 2 + g +
Ĥ = ∑ ak ak + ∑ ak+q ak+ '−q ak 'ak
k 2m 2V kk 'q
" 2 %
 2 g
= ∫ dr $ ∇ψ̂(r) + ψ̂ + (r)ψ̂ + (r)ψ̂(r)ψ̂(r)'
$ 2m 2 '
# &
2 + 2
The kinetic energy factor | ∇ψ̂ | is equivalent to −ψ̂ ∇ ψ̂
via integration by parts as in lecture 2.

Using the fact that |Ψs> is a coherent state, we then find

12
 2
⎛ 2  2 g 4 ⎞
Fs = E s − µN = ∫ dr − µ | ψ s | +
⎜ | ∇ψ s | + | ψ s | ⎟
⎜
⎝ 2m 2 ⎟
⎠
where µ is the chemical potential equal to ng in the Hatree-Fock
approximation. The integrand is essentially the Ginzburg-Landau
free energy density as a function of the order parameter ψs(r) (at
T->0K). This provides a connection between the macroscopic
parameters of the Ginzburg-Landau and the microscopic
parameters of the underlying Hamiltonian, i.e., in this model the
Ginzburg-Landau parameters defined in the first lecture are
α=-ng, m=m and β=g.

Finally, the dynamics of the order parameter may be described as

in the Gross-Pitaevskii theory (going beyond Hatree-Fock) by
2 2 2
−( / 2m)∇ ψ s + g ψ s ψ s = iψ s (Pethick & Smith ch. 7)

This predicts that small oscillations δψs about the equilibrium

state ψ0 , i.e., Goldstone modes at small k, correspond to the
low-lying excitations in the Bogoliubov theory.
13
 Goldstone and Higgs Modes
Goldstone modes correspond to fluctuations of the order
parameter transverse to the equilibrium ψο
in the complex
plane, and are gapless as k goes to 0. Higgs modes
correspond to fluctuations parallel to ψο
and have an energy
gap in their spectrum. The Goldstone modes for a superfluid
are collisionless sound waves as obtained in the Bogoliubov
theory.
free energy

ψο

Broken U(1) symmetry

14
 Last lecture (#9):

We found that the low-lying elementary excitations of a weakly

interacting BEC are density fluctuations with a linear frequency
spectrum ωk = c k at low k. This leads to a finite Landau critical
velocity and superfluidity. We also made contact between
microscopic theory and the phenomenological Ginzburg-Landau
model. The order parameter is a coherent state (similar to that
for a laser) involving a superposition of many-particle wave-
functions with different numbers of particles. When g = 0 the
order parameter reduces to the single-particle wavefunction that
is macroscopically populated as in the ideal BEC.

We now turn to the Bardeen-Cooper-Schrieffer (BCS) theory of

superconductivity. Some of the mathematical techniques will
look similar to that of the last lecture, but there are very
important technical and conceptual differences.

1
Lecture 10: The Bardeen-Cooper-Schrieffer (BCS)
Theory Of Superconductivity

•  BEC-BCS Crossover
•  Cooper Pairs
•  The BCS Wavefunction
•  Mean-Field Approximation and Bogoliubov
Transformation
•  The Energy Gap Equation
•  The Bogoliubov Quasiparticles
•  Experimental Support for the BCS Model
•  Connection to Ginzburg-Landau Theory

Literature: Annett ch. 6, Waldram ch. 7 &

Schmidt ch. 6

2
 BEC-BCS Crossover
Consider a BEC where each boson is a molecule of two tightly
bound fermions, and then slowly reduce the attractive interaction
between the two fermions. What we get is a gradual crossover

Picture credits: M. Greiner

from the BEC to the BCS state.

(real space) (wavevector space)

Superfluidity (or superconductivity if the particles are charged)

can survive this crossover, but the microscopic nature of the BCS
state is very different from that of the BEC state.
In the BEC limit the binding of the fermions occurs at tempera-
tures above the BEC condensation temperature. In the BCS limit
the opposite is the case. The fermion pairs are known as Cooper
pairs. They are not true bosons in that their creation and
annihilation operators do not obey the Bose commutation rules.
3
 Cooper Pairs
The Cooper-pair wavefunction extends over distances much larger
than the inter-particle spacing, so that our usual notion of
molecules undergoing a Bose-Einstein condensation is an
oversimplification. We need a new approach starting with fermions
occupying states within and near the Fermi surface.

–k’
k k↑ k’↑
Filled
Fermi
–k sea -k↓ -k’↓
k’
Region of Attraction

Surprisingly, important insights can be gained by considering a very

simple model: a homogeneous gas of fermions with a weak
attractive interaction g = -|g| for low-energy transitions, |δε| < εc ,
around the Fermi surface at low temperatures. Sums over k will
thus be restricted in this lecture to a k-space shell kF-kc< k< kF+kc
where kF is the Fermi wavevector and kc = εc/(vF).
4
We show that the Fermi liquid state is unstable to a formation of
a coherent state of Cooper pairs. This is the BCS state that is
characterized by an energy gap in the spectrum of elementary
excitations. The origin of the attractive interaction will be
discussed in the next lecture.

We know from elementary quantum mechanics that two

particles in 3D cannot form a bound state unless the attractive
potential is strong enough. However, Cooper pairs can bind for
any finite |g| if g is negative. It turns out that the bound state
is stabilized by the presence of the filled Fermi sea (examples).
Since only states near the Fermi surface will be affected by g it
is useful to consider a description of the Cooper state in k-
space. The transitions which are produced by g are as shown in
the figure (p. 4) for the case where the total momentum of the
Cooper state is zero. g produces repeated scattering from a
state (k,-k) to a state (k’,-k’) in the shaded shell near the Fermi
surface. Since the scattering is s-wave the spin state must be a
singlet. 5
 The BCS Wavefunction
Cooper’s finding that a bound state of pairs lowers the energy of
the Fermi system suggests that the true ground will involve some
coherent state of Cooper pairs. Of special interest now is the pair
creation operator
a+ a+
k + q↑ − k ↓

which creates a pair of fermions at (k + q,-k) with total

momentum q and total spin zero. This is the analogue of the
+
operator ak + q ak , which we introduced in the last two lectures. In
this lecture a+ and a satisfy the commutation rules for fermions.
BCS conjectured that the ground state should be a coherent state
built up from q=0 pair creation operators

exp(s a+ a+ ) = 1 + s a+ a+ + ... (note that (a+ )2 = 0)

k k ↑ −k ↓ k k ↑ −k ↓

This can be normalized by replacing sk by vk and the first term by

uk, where sk
1
u = v =
k 2 k 2
1+ s k
1+ s k
6
Including all k we arrive at the BCS wave-function
2 2
Ψ BCS = ∏ (u + v a+ a+ ) vac , u +v =1
k k k ↑ −k ↓ k k
k
This is a trial wavefunction defined by the variational parameters
uk and vk. These are to be determined by minimizing the energy

E = Ψ ˆΨ
H
BCS BCS BCS

It is sufficient to work with the reduced Hamiltonian*

-k↓ -k’↓
+ g g
Ĥ → ∑ ε ak ak − ∑ ak+ '↑a−k
+
'↓
a−k↓ak↑
V
k k k,k '
k↑ k’↑
Instead of minimizing EBCS, we apply a mean-field approximation
and the Bogoliubov transformation similar to that used in the last
lecture for small g. This is the weak coupling BCS theory.
*The usual factor of 1/2 in the interaction term does not apply because the scat-
tering is between different ‘species’ of particles (i.e., particles of different spin). 7
 Mean-Field Approximation and
Bogoliubov Transformation
The mean-field approximation in this context replaces an operator
ˆB
A ˆ by
ˆB
A ˆ → <A
ˆ>B
ˆ+ A
ˆ<B
ˆ> −< A
ˆ >< B
ˆ>

This assumes that for the chosen trial wavefunction < δAˆ δBˆ>
can be ignored, where δAˆ= Aˆ−< A ˆ > and similarly for δB̂ .

There are three types of relevant replacements for the four fermion
operator (dropping the constant terms)

a4+ a3+ a2a1 → 〈 a4+ a3+ 〉 a2a1 + a4+ a3+ 〈 a2a1 〉 Bogoliubov

− a4+ a2a3+ a1 → − 〈 a4+ a2 〉 a3+ a1 − a4+ a2 〈 a3+ a1 〉 Exchange

a4+ a1 a3+ a2 → 〈 a4+ a1 〉 a3+ a2 + a4+ a1 〈 a3+ a2 〉 Direct

For the BCS model we consider only the first replacement.

8
Applying this approximation to Ĥ we obtain*
Ĥ → ∑ ξ ak+σ akσ − ∑ (Δ ak↑
+ +
a−k↓ + Δ * a−k↓ak↑)
kσ k k
' * ' *
' + * ) ξk −Δ , ) ak↑ ,
= ∑ ) ak↑, a−k↓, ) , ) , , ξ = ε -µ
k
( + ) −Δ * −ξ , ) a+ , k k
( k + ( −k↓ +
where µ is the chemical potential & Δ is a c-number defined by
g
Δ=
V ∑ < a− k ↓ak ↑ >
k
Ĥ can be diagonalized by a Bogoliubov transformation
from the bare fermion operators (akσ , ak+σ ) to new fermion
operators (α kσ , α k+σ ) .

*For simplicity the single particle energy is measured from the chemical potential and only terms
explicitly dependent on the field operators (a and a+) have been retained. A more complete
averaging of products of creation and annihilation operators in the potential energy also leads to
Hartree-Fock terms. This is an application of Wick’s theorem.
See also lecture 9. 9
The required linear transformation can be expressed in the
normalized form
+ + 2
a = uk α + vkα = (α + sk α ) / 1 + s k , σ =↑ or ↓
kσ kσ − k −σ kσ − k −σ

As in the last lecture, we can show by substitution that Ĥ has the

diagonal form*

Ĥ → ∑ Ek α k+σ α kσ
kσ
if sk and Ek satisfy

s = (E k − ξ ) / Δ , s = −s
k k −k k
2 This is the positive solution as appropriate
E = ξ2 + Δ for excitations above the ground state.
k k

Δ is the BCS energy gap function (page 9), which is taken to be

real and positive except for non s-wave pairing (lectures 11 & 12).
*For simplicity only terms explicitly dependent on the field operators (α and α+) have been
retained. 10
 The BCS Energy Gap Δ(T)

Δ(T) vs T is obtained by rewriting in the equation for Δ (page 9) the
original fermion operators in terms of the new fermion operators
and setting <αk+αk> equal to the Fermi function for excitations

f (Ek ) = 1 , Ek = ξ 2 + Δ(T )2 > 0

exp(Ek / k B T ) +1 k
This readily yields the self-consistent BCS gap equation

g 1 − 2f (Ek ) εc tanh[ ξ 2 + Δ(T )2 / (2k T )]dξ

Δ= ∑ Δ → 1 = N(0) g ∫ B
V 2Ek
k 0 ξ 2 + Δ(T )2

N(ξ) is the single-particle density of states per spin per volume

assumed to be a constant N(0) within εc of the Fermi level. A
simple numerical analysis gives Δ(T) (next page), where

⎛ 1 ⎞
Δ (T = 0) ≈ 2 ε c exp⎜⎜ − ⎟ ≈ 1.76 k Tc
⎟ if N( 0 ) g << 1
⎝ N(0) g ⎠
B
11
Δ
Cel /T

Δ(0) = 1.76 kBTc exponential ΔCel = 1.43 Cel,n

form at Tc
at low T

T T
Tc Tc

From f(Ek), one can also calculate the entropy and thus the heat
capacity. The numerical result is shown in the figure above right.
The ratios Δ(0)/(kBTc) and ΔCel /Cel,n as well as other properties are
in striking agreement with experiment in conventional supercon-
ductors, namely, s-wave spin-singlet superconductors in the weak
coupling limit.

12
 The Bogoliubov Quasiparticles
+
The new operator α kσ creates a fermion with wavevector and spin
+ +
(k,σ). Well above the Fermi surface α kσ reduces to akσ and so
adds a bare fermion with (k,σ). Well below the Fermi surface
α k+σ reduces to a−k −σ and so removes a bare fermion
with (-k,-σ).

At intermediate energies, however, α k+σ is quite strange. It creates

a fermion with well-defined wavevector and spin, but indeterminate
particle number. In particular, the new fermion has zero charge at
µ even if the bare fermions carry a charge. This is a remarkable
example of a fermionic excitation in which spin and charge are
separated. Recall that the corresponding boson operator in the last
lecture was interpreted as the creation operator for a density
fluctuation in a BEC.

13
The figures shows the excitation energy spectrum Ek and the
Bogoliubov fermion weights vk and uk as a function ξk = vFk near
the Fermi level in the BCS model.

2 2
Ek = ξk + Δ

ξ k = εk − µ α k+σ = u k ak+σ − v ∗k a− k −σ

E/Δ
Ek
-ξ k ξk 2 2
v u
holes electrons k k

ξk/Δ
ξk/Δ

14
 Connection to Ginzburg-Landau Theory:
Order Parameter & Coherence Length
The Ginzburg-Landau theory was derived from the BCS model by
Gorkov. The derivation is more subtle than for a BEC. The order
parameter turns out to be proportional to the expectation of the
pair annihilation operator in real space (cf. lecture 9 for the BEC)

Ψ s (r) = ψ̂ (r) ψ̂ (r)

↓ ↑

For the uniform state this can be expressed in terms of

<a a > , which is related to Δ (T) (see p. 9).
−k ↓ k ↑
The T=0 K superconducting coherence length ξ can be interpreted
as the characteristic size of the Cooper-pair state. This may be
estimated via the uncertainty principle δkδx ~ 1. If Δ ~ vF δk &
δx ~ ξ, we find ξ ~ vF/Δ. The BCS model gives a similar result

vF v F
ξ ≈ ≈ 0.18
πΔ kBTc
15
 Appendix: Further Experimental Support for the BCS Model
Isotope effect: for a given N(εF)|g|, Tc is

Picture credits: Hyperphysics@GSU / I. Giaever

expected to scale as εc, which is proportional to
the Debye frequency ωD for the electron-phonon
interaction (next lecture). ωD can be varied by
changing the isotopic mass (atomic number A).
Lighter isotopes have higher ωD and
hence have higher
Tc, assuming that
other factors are
equal (figure left).

Tunneling conductivity provides a direct

measure of the density of states of
elementary excitations and thus the BCS
energy gap (figures below and right).
E E
eV
k g(E)
16
The energy gap Δ can also be determined from the heat capacity and thermal
conductivity, which contain contributions from the Bogoliubov quasiparticles at
low T < Tc as well as from phonons.

Picture credits: C. A. Bryant / E. Schuberth

electrons
Cel (log-scale!)

phonons
(Tc (Sn) = 3.7 K)

The sharp change in the thermal

conductivity of a pure type-I supercon-
ductor on crossing the critical field Bc can
be used to produce high-performance
heat switches used in laboratory- and
space-based applications.
Tc/T » Δ/kBT
17
 Last lecture (#10):

We presented the BCS theory of superconductivity for a weak

attractive interaction that is isotropic, spin independent and finite
only in a thin shell around the Fermi surface. The ground state is
a coherent state made up of spin-singlet Cooper pairs and the
excited states have an isotropic energy gap function. The order
parameter can be taken to be
Ψ s (r ) =< ψˆ (r ) ψˆ (r ) >
↓ ↑
or equivalently <a-k ↓ ak ↑ > , which defines the gap function.

We now consider the possible origins of the attractive interaction

and generalize the BCS theory to describe complex materials
such as the high-temperature superconductors MgB2 and the
copper-oxide and iron-pnictide compounds.

1
 Lecture 11: Anisotropic Superfluidity and
Superconductivity

I.  Induced Interactions

IA. Polarizer-Analyzer model
IB. Phonon Mediated Electron-Electron Interaction

II. Generalization of BCS Theory

IIA. Singlet Energy Gap Equation
IIB. Singlet Equation for Tc and Applications

Appendix I. Further Generalization of BCS Theory

Appendix II. Spin-Triplet Pairing States

Literature: Annett ch 7; Waldram selections from chs 7, 11-17

2
 I. Induced Interactions
IA. Polarizer –Analyzer Model

Interactions between particles can usually be described as

induced interactions in terms of the polarizer-analyzer
model. For example, the electron-electron interaction in the
electromagnetic vacuum may be described as follows. One
electron, the polarizer, induces a change in the
electromagnetic field and the second electron, the analyzer,
samples this charge.

One can think of the interactions between fermionic

quasiparticles in a metallic vacuum in a similar way, but the
induced interactions now include effects from all of the other
particles in the system.

3
Basic analyser-polarizer model: let g be the interaction between
a particle and a field and χ(r,t) the impulse response for the field,
then the interaction potential between two particles induced by a
test particle moving at velocity u is

V (r , t ) = ∫ V (r − r ' , t − t ' )δ (r '−ut ' )dr ' dt '

g g
V (r − r ' , t − t ' ) = − g 2 χ (r − r ' , t − t ' )
χ

is the impulse interaction

In general, we include the dependence of the interaction on

spin, the initial states and the momentum and energy transfers.
Nearby in space and time we expect the dominant interaction to
be repulsive. To get a bound state the quasi-particles must take
advantage of attractions that appear at a finite separations in
time or space, i.e., by avoidance in time or space.

4
An example of attraction by time avoidance is the retarded
interaction between charges moving in a deformable lattice, i.e.,
the induced interaction arising from the virtual emission and
absorption of phonons. The figure below shows the form of the
interaction potential V(r,t) versus r at a moment in time due to
the screened Coulomb interaction in the Thomas-Fermi model plus
the retarded induced interaction due to a deformable positively
charged medium (jellium). An attractive tail appears if the test
charge is moving (figure right).

Retarded positively charged

screening cloud

5
Examples of attraction by space avoidance include the van der
Waals attraction in liquid 3He and the induced spin-spin
interactions in liquid 3He and in nearly magnetic metals. On the
border of a Mott transition in the cuprates, for example, the
relevant spin-spin interaction is known as the superexchange
interaction acting between nearest neighbours in a square lattice.
The figures below show the forms of the triplet and singlet
interaction potentials for carriers on the border of ferromagnetism
and anti-ferromagnetism, respectively, for homogeneous media.
Note that due to the Pauli principle there is effectively a hard core
repulsion in the triplet as well as in the singlet case.

6
 IB. Phonon Mediated Electron-Electron Interaction

We consider explicitly the form of the overall interaction in a

simple model for jellium, for which the real part of the Fourier
transform of the impulse interaction is given by

2 ⎤
e2 1 ⎡ ωq
Vqω = ⎢1 − ⎥
2 2 2
ε 0 q + k ⎢ ωq − ω ⎥2
TF ⎣ ⎦

where kTF is the Thomas-Fermi wavevector and ωq is the

phonon frequency. The first term represents the
instantaneous screened Coulomb repulsion while the second
term gives the electron-phonon mediated attraction.* The
overall interaction is attractive in the frequency range 0 < ω <
ωq. This is the origin of the attractive term near the Fermi
level in the BCS model.
*We replace ω in Vqω by the energy transfer in a scattering process so that the scattering
matrix elements depend on the initial momentum as well as the momentum transfer.
7
 The form of the phonon mediated interaction potential can be
obtained from second order perturbation theory in which the
initial and final states are represented by (k,-k) and (k’,-k’) ,
respectively, and the intermediate states involve a virtual
excitation (a phonon) as in the following diagrams

-k’ k’
k’=k+q -k’
g g

time
-q q
g g
-k k
k ε − k = ε k , ω− q = ωq & g − q = gq -k

ind ⎛ 1 / 2 1 / 2 ⎞
Vqω = 2 g q ⎜ + ⎟ g q
⎜ 2ε k' − (ε k' + ε k + ωq ) 2ε k − (ε k' + ε k + ωq ) ⎟
⎝ ⎠
2 g 2q ωq
=− , where ω = ε k' − ε k
 ω2 − ω 2
q 8
 Coulomb repulsion
V qω V qω
BCS
Theory

0 ω
0 ω

− | g|
Net
attraction
€ εc/ 
ωq Phonon frequency
at wavevector q

Because Vqω is in general a dynamical interaction, a full treatment

requires the use of many-body Green functions or of Lagrangians
and the path integral methods. We consider a simplified scheme
where the interaction is represented in terms of weak coupling
matrix elements that depend only on q and ω=εk+q-εk. The
strong coupling extension of the BCS theory will not be developed
but will be reviewed below (see, e.g., Waldram ch 11).
9
 II. Generalization of BCS Theory
IIA. Singlet Gap Equation
The self-consistent gap equation can be generalized in a similar
way. For example, for spin-singlet pairing the expression for Δ
given in lecture 10
⎛ g ⎞ (1 − 2f (E k '))
Δ = − ∑ ⎜ ⎟ Δ
V 2E k '
generalizes to k ' ⎝ ⎠

(1 − 2f (E k ' ))
Δk = − ∑ Vkk'Δk'
k' 2E k '
where Vkk’ stands for the appropriate scattering amplitude*

e.g., Vkk' = Vqω , where q = k'−k & ω = (ε k' − ε k ) / 

Note that Δk can depend on both the magnitude and direction of k
near the Fermi surface. In conventional singlet superconductors Δk
has the same sign over the Fermi surface. In unconventional
superconductors Δk changes in sign over the Fermi surface.
*This applies also to the spin dependent interaction to be discussed in lecture 12
if Vkk’ stands for the spin-singlet amplitude (see e.g., Annett pp 154-156). 10
The anisotropy of Δk near the Fermi surface is usually

Picture credits: C. Bergemann, Cavendish Laboratory

illustrated as in the following figure in which the k-space gap
at the Fermi surface is a relative measure of the energy gap
function (i.e., k - space gap = Δ / v ) .
k F
–

+ + +

–
s-wave d-wave (dx2-y2)
conventional (BCS) high-Tc
superconductors superconductors

If the starting Hamiltonian has rotational symmetry then the

labels s and d stand for angular momentum.
11
 IIB. Singlet Equation for Tc & Applications
The equation for Tc can also be generalized. For example, for
electron-phonon mediated superconductivity the expression for Tc
given in lecture 10
1.13 ε c ⎛ 1 ⎞
Tc ≈ exp⎜ −
⎜ ⎟
⎟
k ⎝ N (0 ) g ⎠
B

generalizes, in the Eliashberg-McMillan strong-coupling theory, to

θD ⎛ 1.04 (1 + λ ) ⎞
Tc ≈ exp⎜ − ⎟
1.45 ⎜ λ − µ *(1 + 0.62λ ) ⎟
⎝ ⎠

where θD is a Debye temperature, λ is the electron-phonon mass

renormalization factor and µ* arises from the effect of the direct
Coulomb repulsion, corrected for screening and recoil.
The McMillan formula is found to be in reasonable accord with
experiment in a number of s-wave superconductors from Hg to
the high-temperature superconductor MgB2, Tc=39 K.
12
 Fermi Surface Experiments: A. Carrington et al.

Graphite-like
boron sheets

Boron σ p-orbital
Fermi surface sheet
MgB2: Crystal Structure and Fermi Surface

Boron σ p-orbital

13

Top: www.ncnr.nist.gov/staff/taner/mgb2/cchoose.shtml
Bottom: www.physik.tu-freiberg.de/~kortus/
 CaFe2As2: Temperature-Pressure
Phase Diagram and Fermi Surfaces

David Tompsett, Quantum Matter Group

Tetragonal
Temperature

“Collapsed”
AFM Tetragonal
Orthorhombic

Pressure

AFM = antiferromagnetic metal

14
 Appendix I. Further Generalization of BCS Theory
Appendix IA. Anisotropic BCS Hamiltonian
The generalization of the BCS Hamiltonian of lecture 10 is thus
ˆ = ∑ ε ka+ akσ +
H ∑ Vkαβγδ a + a+ a a
kσ 'k k'α − k' β − kγ kδ
kσ kk'αβγδ
-kγ -k’β

kδ k’α
where α, β, γ, δ are spin indices.

The generalization of the gap equation of lecture 10 is then

Δαβ
k
= V αβγδ
∑ kk' < a− k'γ ak'δ >
For systems with separate singlet and triplet pairing states, it is
helpful to write the four gap components Δkαβ in terms of a scalar
Δk and a 3D vector dk = (dkx,dky,dkz)
15
⎛ Δ↑↑ Δ↑↓ ⎞ ⎛ 0 Δk ⎞ ⎛ -d x + id y Δk + d z ⎞
⎜ k k ⎟ = ⎜⎜ ⎟⎟ singlet, ⎜ k k k ⎟ triplet
⎜ Δ↓↑ ↓↓ ⎟
Δk ⎠ ⎝ − Δk 0 ⎠ ⎜ − Δ + d z d x + id y ⎟
⎝ k ⎝ k k k k ⎠

The scalar Δk and the components of the vector dk can be found in

terms of the starting gap functions Δk↑↑, Δk↓↓ and Δk↑↓ by inversion.

In this scheme the excitation energies are found to satisfy

Ek2 = ξk2 + | Δk |2 singlet, Ek2 = ξk2 + | dk |2 ± | dk × dk

*| triplet

In superfluid 3He the last term vanishes so that |dk| plays the
same role as |Δ| in the isotropic case. Note that we take the
positive solution for Ek for the elementary excitations.

16
 Appendix IB. s, p and d Gap Functions

Picture credits: C. Bergemann, Cavendish Laboratory

The anisotropy of Δk or dk near the Fermi surface is usually
illustrated as follows

+ i –

+ – + + +

– –i –
s-wave p-wave (py) (px+ipy) d-wave (dx2-y2)
original BCS (dk shown) high-Tcs

If the starting Hamiltonian has rotational symmetry then the

labels s, p and d stand for angular momentum. Pairing can arise
in the presence of short range repulsion either by time avoidance
in an s-state or by space avoidance in a non-s or finite angular
momentum state in which the pair wavefunction vanishes when
the interacting particles get too close.
17
Non-s-wave states tend to be important in strongly correlated
electron systems in which low Fermi velocities reduce the relative
effectiveness of the time avoidance mechanism for pairing.

In the absence of rotational invariance, we label gap functions

according to the irreducible representations of the symmetry
group of the lattice. Thus, s-wave corresponds to the simple
representation invariant under all the relevant symmetry
operations, while dx2-y2 corresponds to the B1 representation of
the tetragonal group D4 of the lattice relevant to the cuprates.
Other representations are of course possible in general and the
lowest energy representation is to be determined by experiment
and interpreted by theory.

18
 Appendix II. Spin-Triplet Pairing States
Appendix IIA. Spin-Triplet Superfluidity in 3He
The temperature-pressure-magnetic field phase diagram of
superfluid 3He shows that there are two basic superfluid phases,
A and B, with different gap functions dk.

Picture credits: E. Thuneberg

The A phase has point nodes (at opposite poles) while the B
phase is fully gapped, but the direction of dk rotates around the
Fermi surface.
19
 3He-A 3He-B

Picture credits: J.F. Annett

Point nodes Fully gapped, but
at north and south poles direction of dk varies
(the arrows in the figures give the directions of dk)

Some aspects of superfluid hydrodynamics are similar to those of

He-II, however, the complex nature of the order parameter leads
to non-trivial topological defects.

20
 Appendix IIB. Spin-Triplet Superconductivity
in Sr2RuO4 and UGe2

Picture credits: C. Bergemann, Cavendish Laboratory

Sr2RuO4 becomes superconducting below 1.5 K and is believed
to order in a state similar to the A phase of 3He in 2D. The
crystal structure is similar to that of the La2CuO4 family of the
cuprates and has a quasi-2D Fermi surface with three sheets
corresponding to the one-electron orbitals dxy, dxz and dyz.

RuO6
octahedra

Sr

dxz,yz
Crystal structure Fermi surface dxy
Sr2RuO4
21
Superconductivity is normally destroyed by ferromagnetic order,

Picture credits: S. S. Saxena et al., Cavendish Laboratory

but there are some exceptions. Perhaps the most surprising is
UGe2 where the magnetic electrons are itinerant and form
distinct majority and minority spin Fermi surfaces. Since the
states (k,↑) and (-k,↓) are not degenerate spin-singlet pairing is
energetically unfavourable. Pairing is expected to be in a spin-
triplet state, but the detailed nature of the order parameter is
still unknown.

22
 Last lecture (#11):

We considered the origin of the attractive interaction in the BCS

theory and more general interactions that give rise to anisotropic
energy gap functions. The electron-phonon coupling gives rise to
a time-avoidance mechanism for s-wave pairing that accounts for
superconductivity for low-Tc metals such as Hg (Tc~4 K) as well
as high-Tc layered compounds such as MgB2 (Tc~40 K).
Non-s-wave or unconventional forms of superconductivity arise in
strongly-correlated electron systems where the space-avoidance
mechanism for pairing can be important. The unconventional
superconductors may be characterized by the symmetry
properties of the energy gap function.
In this lecture we consider unconventional superconductivity in
the high-Tc copper oxide systems having energy gap functions of
d-symmetry. We review how this has been inferred
experimentally and introduce an exhange pairing mechanism that
can lead to d-wave superconductivity and to high-Tc.
1
 Fermi Surface Experiments: A. Carrington et al.

Graphite-like
boron sheets

Boron σ p-orbital
Fermi surface sheet
MgB2: Crystal Structure and Fermi Surface

Boron σ p-orbital

www.ncnr.nist.gov/staff/taner/mgb2/cchoose.shtml
Bottom: www.physik.tu-freiberg.de/~kortus/
 Lecture 12:

I.  Superconductivity in the High-Tc Cuprates

IA. Introduction
IB. Measurements of Magnitude of Energy Gap Function over
Fermi surface
IC. Phase-Sensitive Measurements of the Energy Gap Function

II.  Possible Mechanism for Superconductivity on the Border of

the Mott transition
II.  Some Guidelines in Search for High Tc Superconductivity
Appendices
Literature:
Review of measurements of gap function:
C.C. Tsuei and J.R. Kirtley, Rev. of Mod. Phys. 72 (2000) 969
Discussions of pseudogap and pairing mechanisms:
P.W. Anderson et al., J. Phys.: Condens. Matter 16 (2004) R755;
Waldram chs 13, 17.
3
 I. Superconductivity in the High-Tc Cuprates
IA. Introduction

Picture credits: C. Bergemann, Cavendish Laboratory

These compounds have in common CuO2 planes as illustrated
below separated by planes of other atoms (not shown) that act
as spacers and as electron reservoirs to tune the concentration
of carriers in the CuO2 planes. The electronic properties are
modelled in terms of a single tight-binding band made up of d-
orbitals, dx2-y2, in a square lattice (quasi 2D) near to half filling.

Cu

one localized electron per Cu site mobile holes

antiferromagnetic Mott insulator high-Tc superconductor

LaxSr2-xCuO4 YBa2Cu3O7-δ
HgBa2Ca2Cu3O8+δ

Tc up to ∼40 K Tc up to ∼90 K Tc up to ∼160 K!
4
As illustrated below, a Mott insulating state at half filling

Picture credits: Panagopoulos et al., Cavendish Laboratory

transforms into a d-wave superconducting state with a
maximum Tc at around 0.2 holes per CuO2 square. The dx2-y2
symmetry of the superconducting state has been established by
novel experimental techniques.
300 K
CuO6
Antiferromagnetic octahedra

parent compound La2CuO4

Mott Insulator

Crystal structure for the

Temperature

La

Pseudo-
Gap State

Fermi Liquid
State

d-Wave
Superconductor

0
Carrier Concentration (Hole/Cu) 40%
5
 Fermi Surface and d-Wave Gap in the Cuprates
(Brillouin zone dimensions in units of inverse lattice constant)

Δk
+ (π,π) +

–
d-wave (dx2-y2)

(1 − 2f (E k ' ))
Δk = − ∑ Vkk'Δk'
k' 2E k '

E k = ξk2 + | Δk |2 Δ k = Δ(cos k x − cos ky )

6
 IB. Measurements of Magnitude of Gap Function

i –
+

+ – + + +

–i –
–
s-wave d-wave (dx2-y2)
BCS p-wave (py)1 p-wave (px+ipy)1 high-Tc’s

Gap nodes drastically change the density of low-lying excited states

in a superconductor and hence the temperature dependences of
thermal and transport properties. These properties depend on
Fermi surface averages involving |Δk| and hence cannot in general
determine the locations of the nodes on the Fermi surface, nor
prove that Δk change sign.1
1 For p-wave pairing Δk is replaced by dk defined in an appendix of lecture 11. 7
 Density of excitations Nexc(ε) for s-wave and d-wave
superconducting gaps

ε Ek Nexc(ε)
s-wave
pairing N(ε)
ξk -Δ/kBT
Cel ~ e
δk⊥ ε
Δ

ε € Nexc(ε)
Ek d-wave
δk|| pairing
Nodal Cel ~ T2
point
δk⊥ ε
The number
€ of k-states within a circle of radius |δk|~ε is
proportional to |δk|2 ~ε2, so that Nexc(ε) ~ ε for ε<<Δ.
8
€
To determine the locations of the gap nodes on the Fermi surface
we need more microscopic probes such as angle-resolved
photoemission (ARPES):
Δ k = Δ(cos k x − cos ky )
photons electrons –
ϑ →

Picture credits: H. Ding

ϑ

+ +

–
d-wave (dx2-y2)
high-Tc’s

Bulk experiments in the cuprates reveal the existence of lines of

nodes consistent with a d-wave pairing state in quasi 2D. The
angle-resolved photoemission experiments show that the nodal
lines (points in the Fermi surface plane shown) are along the
diagonal of the Brillouin zone of a square lattice. This is
consistent with a gap function with dx2-y2 symmetry. 9
 IC. Phase Sensitive Measurements of Gap Function
To show that the gap function Δk actually changes sign as
expected for the dx2-y2 gap state we need to look for interference
phenomena. In the corner junction geometry shown below the
positive and negative lobes of the d-wave gap function lead to
cancellation between the two currents shown:

Picture credits: Tsuei & Kirtley

d-wave B

s-wave

Measurements as a function of applied magnetic field, that can

frustrate the cancellation, provide phase sensitive confirmation of
the dx2-y2 symmetry of the gap function in the cuprates. 10
 II. Possible Mechanism for Superconductivity
on the Border of a Mott Transition

Picture credits: Monthoux, Pines & Lonzarich Nature 2007

The electron-phonon interaction, which plays a vital role in
conventional superconductors, may not account fully for
superconductivity in the cuprates. A natural explanation for
the d-wave gap symmetry and for the magnitude and doping
dependence of Tc has been given in terms of an effective spin-
spin interaction (t-J model) between the carriers on the border
of a Mott transition and antiferromagnetism.
Exchange Interaction d-Wave
Cooper State
(probability
distribution
in real space)

11
As in the original BCS theory we consider an effective pairing
Hamiltonian but now for the exchange interaction rather than the
electron-phonon interaction. For a single tight-binding band near
half filling on the border of the Mott transition we consider

Ĥeff = −teff ∑ (c +jσ ciσ + hc) + Jeff ∑ Si ⋅ S j

<ij>,σ <ij>

where <ij> denotes sum over nearest neighbour (nn) pairs of

sites, teff is the effective nn hopping matrix element, Jeff (>0) is
the effective exchange constant, ciσ destroys an electron of spin
index σ on atomic site i and Si is the electron spin on site i.
The last term can be expressed in terms of the field operators via
Siz = 1 (ni ↑ − ni ↓ ), niσ = c + c
2 iσ iσ
Si ± = Six ± iSiy , Si − = c + c , Si + = c + c
i↓ i↑ i↑ i↓
and Heff can be rewritten in terms of the k-space field operators
using
ciσ = 1 ∑ akσ exp(ikr )
Ns i
k 12
For simplicity consider first the Ising model in which Si.Sj is
replaced by Siz.Sjz. Keeping only spin-singlet components as in the
original BCS model we obtain
Ĥeff → ∑ εk ak +a + V a+ a+ a a
k ∑ k ' k k '↑ −k '↓ −k↓ k↑ -k↓ -k’↓
k k,k '
where εk = −teff γ k , Vk ' k = −Jeff γ k '−k / 4Ns
and γ k = ∑ exp(−ikri ) = 2(cos k x a + cos ky a) k↑ k’↑

i=nn
γk is the structure factor, which is central to the lattice model and
defines the forms of both the kinetic and potential energies.
Replacing the factor of ¼ by ¾ in Vk’k as is appropriate for the
Heisenberg vs. the Ising interaction for the singlet state*, the
singlet gap function at T=0 is thus given by (see page 6)
3
J
4 eff
γ k − k ' Δk ' 2 2
Δk = ∑ , E = (−teff γ k − µ) + Δ
Ns 2E k ' k k
k'

* <Si.Sj> = ¾ in the singlet state. To derive Δk more formally rewrite Si.Sj in

+ + + +
terms of Sz, S- & S+ and define the singlet state via c j ↓ci ↑ - c j ↑ci ↓ .
13
To find the possible solutions for the gap function we rewrite the
structure factor (using its exponential form or otherwise) as

γ = 1γ γ +η η where η = cos(kx a) − cos(ky a)

k − k' 4 k k' k k' k

Using the fact that γk/2 and ηk are normalized and orthogonal
(the k-space averages of γk2/4 and ηk2 are unity, and the k-space
average of γkηk vanishes) we obtain the following solutions to the
gap equation

Δ =Δ η d-wave & Δ = 1 Δs γ extended s-wave

k k k 2 k

The d-wave solution is favoured and the magnitude of the T=0

energy gap Δ is given by

3J
4 eff
ηk2
1= ∑
Ns
k 2 (−teff γ k − µ )2 + Δ2ηk
2

14
This result is consistent with the physical picture in page 11
and the gap function typically observed in the cuprates (pages
6-10). In a more realistic treatment we would replace the
energy band spectrum (–teff γk) by the Hartree-Fock spectrum
as discussed in a previous lecture. The Hatree-Fock contri-
bution in this case (i.e., in the presence of nonlocal inter-
actions) can lead to unconventional particle-hole as well as
particle-particle instabilities.

The above effective Hamiltonian can be derived approximately

from the single-band Hubbard model, defined by a nn hopping
matrix element t and a repulsive interaction energy U for sites
occupied by two electrons (which are necessarily of opposite
spin). For transitions involving energies much less than U and
for U>>t the Hubbard model can be replaced by the projected
t-J model, in which states with double occupancy are projected
out and electrons are coupled via a nn kinetic superexchange
interaction with exchange constant J = 4t2/U. J is positive and
thus antiparallel spin electrons on neighbouring sites attract
and can form singlet pairs. The effects of superexange and the
15
the projection treated in Gutzwiller’s mean field approximation
then lead to an effective Hamiltonian of the form

Ĥeff = −gt t ∑ (c +jσ ciσ + hc) + gs J ∑ Si ⋅ S j

<ij>,σ <ij>
where gt = 2x / (1 + x), gs = 4 / (1 + x)2 & x = 1- < ni↑ + ni↓ >

Thus teff=gtt & Jeff=gsJ, which tend to t and J respectively as x → 1 ,

i.e., in the empty band limit where double occupancy vanishes
even without projection. Crucially teff decreases with decreasing x
and vanishes as x → 1, i.e., the effective bandwidth goes to zero
(hopping is suppressed) as the Mott transition is approached.

The dependences of Δ and gt Δ on doping in this model are

shown on page 17 and compared with the phase diagram of a
cuprate superconductor Bi2212. Δ represents the energy scale
for the formation of spin singlets. These are frozen and cannot
lead to superconductivity in the limit x=0. kBTc appears to follow
gt Δ, which vanishes as required at low x and becomes
comparable to Δ at high x where the singlets are fully mobile. 16
 t-J Model
0.3 Bi2212
in MFA

After: Anderson, et al., J. Phys.: Condens. Matter 16

t/J = 5

Energy / meV
Energy/3J 60

(2004) R755 and references cited therein

J~102 meV ARPES
0.2 Gap
Δ
40

0.1 20
gtΔ
kBTc

0 0
0 0.1 0.2 0 0.1 0.2
Doping x Doping x

The predictions of the t-J model in the mean field approxi-

mation are consistent with the symmetry of the gap function
and the orders of magnitude of the gap and Tc in terms of
realistic estimates of t and J. However, this is only a possible
beginning to the description of the cuprates and in particular
of an understanding the so-called pseudogap regime (x<0.2). 17
 III. Some Guidelines in Search for High Tc Superconductivity

•  Fermi surface near to real or virtual Brillouin zone boundary;

e.g., arising from lattice and/or antiferromagnetic order or
virtual order;

•  strong lattice and/or antiferromagnetic fluctuations;

•  high energy scales; e.g., high t, θD and/or J (= 4t2/U);

•  quasi-2D lattice structure, e.g., tetragonal lattice with weak

but non-vanishing c-axis coupling;

•  high superfluid density ns and low carrier mass m, i.e., high

‘phase stiffness’ proportional to ns/m and hence weak
superconducting fluctuations.

18
 Appendix I: Pseudogap state in Underdoped Cuprates

The Fermi surface undergoes a ‘Cheshire cat’ disappearance on

going from the conventional metallic state at high T to the
pseudogap state and finally to the superconducting state at low T:
the surface first shrinks to Fermi arcs and finally to Fermi points

Picture credits: M. R.Norman / Disney

at the nodes of the superconducting gap function (figure right).
The normal state in the underdoped regime in particular remains
quite mysterious.
π
(π,π)
Fermi surface:T>TPG
–

(π,π) Fermi arc:T<TPG

+ +
Fermi point:
T<Tc
–

d-wave (dx2-y2)
π

NS = Normal state, PG = Pseudogap state, SC = Superconducting state 19

 Appendix II: Anderson’s Kinetic Superexchange Due to Virtual
Hopping Between Nondegenerate Atomic Orbitals in Sites i & j

Consider the single-band Hubbard model characterized by nearest

neighbour transition amplitude t and on-site repulsion energy U.
In the infinite U limit at half filling of the band the atomic sites
are singly occupied and the two spin configurations shown below
for adjacent sites are degenerate.
Virtual hopping
leading to inter-
mediate states of
double occupancy
with energy U. This
lowers the energy
by 2t2/U if t<<U.

i j i j
In second order in perturbation theory for finite t/U this
degeneracy is lifted by
4t 2 ⎛ 1 ⎞ t
Energy = ⎜ Si ⋅ S j − ⎟ if << 1
U ⎝ 4 ⎠ U
20
 Appendix IIIA: Collective Modes in
Superfluids and Superconductors

In the Bogoliubov theory for bosons the low-lying elementary

excitations of a superfluid were found to be collective density
fluctuations or phonons (lecture 9). These can be viewed as
Goldstone modes corresponding to fluctuations in the phase of
the Ginzburg-Landau order parameter. In the long wavelength
limit they reduce to rotations of the direction of ψ in the complex
plane at fixed amplitude at the minimum of the “mexican hat”
shaped free energy well shown below:
Goldstone
mode Free energy
density

Im ψ

Re ψ

21
On the other hand, we found that in the BCS-Bogoliubov theory
for fermions with attractive interactions the low-lying excitations
were Bogoliubov fermions with an energy gap above the ground
state. What happened to the Goldstone mode expected in the
Ginzburg-Landau model?

The collective excitations were lost in the approximations leading

up to the BCS model and the Bogoliubov transformation in the
case of fermions.

A more accurate analysis that effectively takes account of

interactions among the Bogoliubov fermions leads to collective
modes. For a neutral BCS-Bogoliubov superfluid, we obtain
gapless density fluctuations or a Goldstone mode as in the BEC
state. For a superconductor, however, this mode is effectively
strongly gapped by the long-range Coulomb interaction. The low-
lying excitations are thus essentially the Bogoliubov fermions as
in the BCS-Bogoliubov theory of lecture 10.

22
But what about the amplitude fluctuations in the Ginzburg-
Landau model, i.e., oscillations around the minimum of the
“mexican hat” well?

Goldstone
mode

Higgs
mode

These oscillations can be thought of as arising from a

superposition of Bogoliubov fermions at the Fermi surface with
zero excitation momentum. The amplitude modes are bosons (at
low k) and can be associated with the Higgs particle in the
simplest model of the Anderson-Higgs mechanism for the
spontaneous generation of mass.
23
 Appendix IIIB: Model for the Anderson-Higgs Mechanism

A simple idealized model for the Anderson-Higgs mechanism can

be described as follows:
•  Above Tc and with no coupling of the gauge field A to the GL
field ψ we have two massless gauge bosons (A) and two
massive scalar bosons (ψ ) with <ψ > = 0.
•  Below Tc with no coupling of the gauge field to the GL field we
have again two massless gauge bosons (A), but now one
massless Goldstone boson (fluctuation in phase of ψ ) and one
massive scalar boson (fluctuation in amplitude of ψ ).
•  When we couple ψ to A in the broken symmetry state below
Tc we have three massive bosons# and one massive scalar
boson – the Higgs boson.

# Longitudinal and transverse oscillations at plasma frequencies exist

not only below Tc, but also above Tc in a metallic vacuum. The
difference is that the transverse electromagnetic field has spectral
intensity extending down to low frequencies above Tc, but not below
Tc in the Meissner phase. 24
This is a mechanism by which gauge fields acquire mass:

“the gauge particle ‘eats’ a Goldstone boson and thereby

becomes massive”.

This general idea has been applied to the more complex problem
of the weak interaction which is mediated by the W-bosons.

Essentially, the initially massless W-gauge particles become

massive below a symmetry breaking phase transition through a
generalized form of the Anderson-Higgs mechanism. This
symmetry breaking transition is analogous in some sense to
superconductivity with a high transition temperature:

Tc ∼ 1016 K !

For further elementary discussions of the Anderson-Higgs Mechanisms see e.g.,

G.M. Caughlan and T.E. Todd, chs 20, 24, University Press
25