Вы находитесь на странице: 1из 47

Notes on many body theory of Bose and Fermi gases at low

temperatures
Vctor Romero-Rochn
Instituto de Fsica, Universidad Nacional Autonoma de Mexico,
Apdo. Postal 20-364, Mexico D. F. 01000, Mexico.
(Dated: February 1, 2011)
Abstract
This is a brief course of basic ideas on second quantization and on Bose and Fermi gases. We
will review the formalism of second quantization rst. Then, we discuss the role of the scattering
length in calculations of thermodynamic properties at low temperatures. We shall study a Bose gas
within the Bogolubov approximation and will analyze the superuid behavior of the gas. We will
describe the formation of vortices and study its description via the Gross-Pitaevskii equation. For
Fermi gases we will study the Bardeen-Cooper-Schriefer (BCS) formalism to describe a gas with
attractive interactions. Additionally, we will describe the molecular BEC - atomic BCS crossover
that occur in a Fermi gas by externally varying the scattering length. Finally, we will give a brief
description of the thermodynamics of these gases when they are conned or trapped by an external
inhomogenous potential
This notes are based, essentially, on the books by Fetter and Walecka (Quantum Theory on Many-
Particle Systems), the Landau and Lifshitz collection (Quantum Mechanics, Statistical Physics I
and II) , the monography by Pethick and Smith (Bose-Einstein Condensation in dilute gases), the
classical article of Leggett on the BEC-BCS crossover, and in personal articles and notes.
1
I. SECOND QUANTIZATION
A. Quantum Mechanics of a system of N particles (Many-Body
The system. Consider a system (gas) of N interacting particles, each with mass m and
spin s, whose Hamiltonian is,
H =
N

i=1
p
2
i
2m
+

i<j
u([r
i
r
j
[) +
N

i=1
V
ext
(r
i
) (1)
where u([r
i
r
j
[) is the interaction potential between particles i and j, the same for all pairs.
Note,

i<j
=
1
2
N

i=1
N

j=1
but i ,= j.
u(r)
r
FIG. 1. Typical interatomic potential u(r).
The term V
ext
(r
i
) is the conning potential, the same for all particles. We shall consider
rst the conning to be a box of rigid walls with volume V . Later we shall consider a
magnetic or optical trap. The potential of the box of rigid walls is,
V
ext
(r) =
_
_
_
0 if r V
if r / V
(2)
2
NOTE: Operators should be written with its hat always, i.e.

A. However, if the context
makes it obvious we will not write them up.
Quantum Mechanics. Schr odinger equation for the state of the system,
i

t
[(t) = H [(t) (3)
[(t) state of the N-particle system ... we need a basis set to calculate it:
[(t) =

{m}
C
{m}
(t) [m (4)
where [m , m is a complete basis of the Hilbert space of the N-particle system.
m denotes the set of quantum numbers needed to specify one N-particle state (there are
as many numbers as degrees of freedom, including spin). The basis satisfy,
m[n =
{m},{n}
orthonormality (5)
and

{m}
[mm[ =

1 completeness (6)
with

1 the unit operator in the Hilbert space.
With the basis, Sch odinger equation implies a set of equations for the complex coecients
C
{m}
(t), (show!)
i

t
C
{m}
(t) =

{n}
m[nC
{n}
(t) (7)
one equation for each coecient C
{m}
(t)
First Problem: Which basis is good? How do we nd it?
B. The basis of the ideal gas.
An ideal gas is u([r[) 0, i.e. no interactions among the particles. We assume that the
particles have xed total spin s. The Hamiltonian is
H
0
=
N

i=1
p
2
i
2m
+
N

i=1
V
ext
(r
i
) (8)
A good basis is that of the product of N one-particle wavefunctions or state. Let [

k, m
be the basis of the momentum p and the projection S
z
of the spin of one particle,
p [

k, m =

k [

k, m
S
z
[

k, m = m[

k, m (9)
3
where m = s, s + 1, . . . , s 1, s. The wavefunction should be zero in the walls because
of the external potential V
ext
, eq.(2), however, it can be shown (later) that in the thermo-
dynamic limit N , V , but N/V = constant, a basis with periodic boundary
conditions gives the same results,

k,m
(r, ) r, [

k, m =

k[r[m
=
1

V
e
i

kr

m
(10)
with

k =
2
L
(n
x
, n
y
, n
z
) and n
x
= 0, 1, 2, . . . ; n
y
= 0, 1, 2, . . . ; n
z
= 0, 1, 2, . . . (11)
and where we have assumed a box of volume V = L
3
. The quantity
m
is the projection of
spin state [m > on a given basis [ >.
NOTE: From now on we shall write k (

k, m) to specify all four quantum numbers of


one-particle, and also r (r, ) to specify both the spatial and the spin coordinates.
Clearly, a product of one-particle states is an eigenstate of H
0
,
(r
1
, r
2
, . . . , r
N
) =
k
1
(r
1
)
k
2
(r
2
)
k
N
(r
N
) (12)
show:
H
0
(r
1
, r
2
, . . . , r
N
) =
_
N

i=1

2
k
2
i
2m
_
(r
1
, r
2
, . . . , r
N
), (13)
the ground state being k
1
= k
2
= = k
N
= 0.
C. Indistinguishability, symmetry of N-particle wavefunctions
There is a fundamental problem, however. Quantum mechanically, particles with the
same mass and same spin are indistinguishable. This gives rise to two types of particles
bosons and fermions.
Symmetry of exchange. If particles are indistinguishable, the state must be the same if
two particles are interchanged, r
i
r
j
. This gives rise to two cases, (show)
Symmetric wavefunctions, bosons:
(r
1
, r
2
, . . . , r
i
, . . . , r
j
, . . . r
N
) = (r
1
, r
2
, . . . , r
j
, . . . , r
i
, . . . r
N
) (14)
4
Antisymmetric wavefunctions, fermions:
(r
1
, r
2
, . . . , r
i
, . . . , r
j
, . . . r
N
) = (r
1
, r
2
, . . . , r
j
, . . . , r
i
, . . . r
N
). (15)
Note that a change of sign of the wavefunction gives the same state (make sure you under-
stand this!).
An important consequence is that, since the linear superposition of states is again a state,
then for a given system of indistinguishable particle, all of their wavefunctions are symmetric
or all are antisymmetric (show).
It can be shown (using relativistic quantum mechanics) that symmetric wavefunctions
correspond to particles with integer spin s = 0, 1, 2, . . . and antisymmetric ones to those
with half-integer spin s = 1/2, 3/2, . . . .
D. Ideal fermions.
It is an exercise to show that for the ideal gas of fermions, the energy eigenfunctions,
H
0

(F)
k
1
,k
2
,...,k
N
(r
1
, r
2
, . . . , r
N
)
(F)
=
_
N

i=1

2
k
2
i
2m
_

(F)
k
1
,k
2
,...,k
N
(r
1
, r
2
, . . . , r
N
), (16)
are given by the (normalized) Slater determinant,

(F)
k
1
,k
2
,...,k
N
(r
1
, r
2
, . . . , r
N
) =
1

N!

k
1
(r
1
)
k
2
(r
1
) . . .
k
N
(r
1
)

k
2
(r
2
)
k
2
(r
2
) . . .
k
N
(r
2
)
.
.
.
.
.
.
.
.
.
.
.
.

k
N
(r
N
)
k
2
(r
N
) . . .
k
N
(r
N
)

(17)
This guarantees that if we interchange r
i
r
j
, the wavefunction changes sign. In addition,
if two values of k are the same, k
i
= k
j
, (i.e. the four quantum numbers are the same) or
two values of r are the same, i.e. r
i
= r
j
, then the wavefunction vanishes; that is, there is
zero probability for two values to be the same. This is the Pauli Exclusion Principle. We
dene the number of particles in state k as n
k
. For fermions this is, n
k
= 1 if k is part of
the set (k
1
, k
2
, . . . , k
N
) and it is n
k
= 0 if it is not part of the set.
Note that the subindex i of k
i
is only to dierentiate the values of k, there is no further
meaning. We could have used k
a
, k
b
, etc.
5
Basis for N fermions interacting or not. The basis for a system of N fermions is all
wavefunctions
(F)
k
1
,k
2
,...,k
N
(r
1
, r
2
, . . . , r
N
) for all values of dierent (k
1
, k
2
, . . . , k
N
), of the type
(17).
E. Ideal bosons.
Since the symmetric wavefunction does not put any restriction on the values of k one has
the energy eigenfunctions,
H
0

(B)
k
1
,k
2
,...,k
N
(r
1
, r
2
, . . . , r
N
) =
_
N

i=1

2
k
2
i
2m
_

(B)
k
1
,k
2
,...,k
N
(r
1
, r
2
, . . . , r
N
), (18)
with the normalized wavefunctions (show)

(B)
k
1
,k
2
,...,k
N
(r
1
, r
2
, . . . , r
N
) =
_
n
k
1
!n
k
2
! . . .
N!

perm

k
1
(r
1
)
k
2
(r
1
) . . .
k
N
(r
N
) (19)
where

perm
is the sum over all permutations of dierent values of k
1
, k
2
, ... , and n
k
1
is the
number of times the value k
1
appeared, n
k
2
is the number of times the value k
2
appeared etc.
This is very important to understand. Since there is no exclusion principle in this case,
it is possible to have repeated values of some k, for instance, one could have k
1
= k
15
= k
235
,
that is, the number of particles in state k
1
is n
k
1
= 3. And the same for other values of k.
It is clear, however, that the following sum is obeyed,

k
n
k
= N (20)
where the sum runs over all values of k. There is no problem since if one value of k does
not appear in the state, say, k

, then n
k
= 0 and it does not contribute to the sum.
Regarding the normalization of the state (19), we see the following: For N particles with
n
k
1
in state k
1
, n
k
2
in state k
2
, ..., n
k
j
in state k
j
, the number of permutations of the dierent
values of k is, (show)
N!
n
k
1
!n
k
2
!...
(21)
Since 0! = 1 we can include all values of k in the denominator, only those n
k
that are
dierent from zero contribute.
Basis for N bosons. The basis for a system of N fermions is all wavefunctions

(B)
k
1
,k
2
,...,k
N
(r
1
, r
2
, . . . , r
N
) for all values of dierent or repeated (k
1
, k
2
, . . . , k
N
), of the type
(19).
6
F. Interacting fermions or bosons
If the system is ideal, the basis (17) for fermions or (19) for bosons is the basis that
diagonalizes the Hamiltonian H
0
. If the particles interact, the above basis do not diagonalize
H, eq.(1), but they can be used to construct the corresponding eigenfuncions:
H
{M}
(r
1
, r
2
, . . . r
N
) = E
{M}

{M}
(r
1
, r
2
, . . . r
N
) (22)
where,

{M}
(r
1
, r
2
, . . . r
N
) =

k
1
,k
2
,...,k
N
C
{M}
(k
1
, k
2
, . . . , k
N
)
(F)
k
1
,k
2
,...,k
N
(r
1
, r
2
, . . . , r
N
) (23)
if the system is composed of fermions, and analogous for bosons. The sum is over all allowed
values of (k
1
, k
2
, . . . , k
N
). The notation M refers again to the set of quantum numbers
needed to specify one eigenstate of H. The problem of the interacting N-particle system is
reduced to calculate the coecients C
{M}
(k
1
, k
2
, . . . , k
N
) ... looks dicult, it is! Specially
because not even the eigenfunctions
(F)
k
1
,k
2
,...,k
N
(r
1
, r
2
, . . . , r
N
) or
(F)
k
1
,k
2
,...,k
N
(r
1
, r
2
, . . . , r
N
)
are easy to calculate, neither is easy to make the sum in (23). An alternative procedure is
based on looking at the number of particles n
k
instead on the wavefunctions. This is the
technique of Second Quantization.
G. The occupation number representation.
The diculty above is certainly due to the indistinguishability and, thus, the need to
introduce symmetrized (bosons) and antisymmetrized (fermions) wavefunctions. It turns
out that it is much easier if the N-particle state is characterized by the number of particles
in each of the one-particle states k, i.e. by the set n
k
for all k. This is called the occupation
number of state k.
We dene the number-particle states (Fock space) as follows. First, choose an order for
all the one-particle states, k
0
, k
1
, k
2
, . . . , k

(e.g. ground states, rst excited, etc). This is


unimportant for bosons but crucial for fermions. Then we write,
[n
k
0
, n
k
1
, . . . , n
k
= [n
k
0
[n
k
1
[n
k
(24)
which reads the state with n
k
1
particles in the state of k
1
of one-particle, n
k
2
particles in the
state of k
2
of one-particle, etc. If the particles are fermions, then n
k
j
= 0 or 1 only; if they
7
are bosons, n
k
j
= 0, 1, 2, 3, ..., , i.e. any value. This set of states obey the usual conditions:
n

k
0
, n

k
1
, . . . , n

k
[n
k
0
, n
k
1
, . . . , n
k
=
n

k
0
n
k
1

k
1
n
k
1

n

k
n
k
orthogonality (25)

n
k
0
,n
k
1
,n
k
2
,...,n
k
[n
k
0
, n
k
1
, n
k
2
, ..., n
k
n
k
0
, n
k
1
, n
k
2
, ..., n
k
[ =

1 completeness (26)
where the sum is over all values of the occupation numbers.
From now on, we shall only describe the case of bosons. Notice that there is a
Hilbert space [n
k
> for each value of k. We introduce two operators for each value of k, b
k
and b

k
, with the following commutations rules:
_
b
k
, b

_
=
kk

[b
k
, b
k
] =
_
b

k
, b

_
= 0. (27)
Show, using these commutations properties that
b

k
b
k
[n
k
= n
k
[n
k
n
k
= 0, 1, 2, . . . ,
b
k
[n
k
=

n
k
[n
k
1 (28)
b

k
[n
k
=

n
k
+ 1 [n
k
+ 1
Since b
k
generates a state with one less particles, it is called the annihilation operator; in
the same note, b

k
is called the creation operator since it generates a particle with one more
particles. For obvious reasons, the product b

k
b
k
is called the number-operator.
The occupation number basis, eqs.(24)-(26), is as good a basis as the bosons basis

k
1
,k
2
,...,k
N
(r
1
, r
2
, . . . , r
N
) (where we have dropped the superindex (B)).

k
1
,k
2
,...,k
N
(r
1
, r
2
, . . . , r
N
) =
_
n
k
1
!n
k
2
! . . .
N!

perm

k
1
(r
1
)
k
2
(r
1
) . . .
k
N
(r
N
) (29)
As a matter of fact, this basis and the number occupation are completely equivalent as we
now indicate.
First, we note that the wavefunction
k
1
,k
2
,...,k
N
(r
1
, r
2
, . . . , r
N
) can be read in the following
way: The probability amplitude of nding N particles with coordinates (spatial and spin)
r
1
, r
2
, . . . , r
N
, given that n
k
1
particles are in the state k
1
of one particle, n
k
2
particles are in
the state k
2
, and so on, with the restriction that

k
n
k
= N. That is, we can rewrite,

n
k
0
,n
k
1
,...,n
k
(r
1
, r
2
, . . . , r
N
) =
_
n
k
0
!n
k
1
! . . . n
k
!
N!

perm

k
1
(r
1
)
k
2
(r
1
) . . .
k
N
(r
N
). (30)
8
One must read this carefully: The set (k
1
, k
2
, . . . , k
N
) is well dened for given values of
the k
j
. Of this set there can be several k
j
that are equal to each other. This denes the
numbers n
k
. There will be many (innite!) values of k that do not appear in the set; the
corresponding numbers n
k
are zero. Thus, given a particular set of occupations numbers
n
k
, k, with

k
n
k
= N it is clear that the state
n
k
0
,n
k
1
,...,n
k
(r
1
, r
2
, . . . , r
N
) is uniquely
dened.
Hence, an arbitrary wavefunction [(t) > of the system can be expanded in the basis

n
k
0
,n
k
1
,...,n
k
(r
1
, r
2
, . . . , r
N
),
(r
1
, r
2
, . . . , r
n
; t) = r
1
, r
2
, . . . , r
N
[(t)
=

n
k
0
,n
k
1
,...,n
k
C(n
k
0
, n
k
1
, . . . , n
k
; t)
n
k
0
,n
k
1
,...,n
k
(r
1
, r
2
, . . . , r
N
) (31)
where the sum is restricted,

n
k
0
,n
k
1
,...,n
k
sum over values of n

k
s with restriction

k
n
k
= N. (32)
H. Crucial step. Second Quantization.
The important point is that if we expand the state [(t) > in the occupation number
basis one gets,
[(t) =

n
k
0
,n
k
1
,...,n
k
C(n
k
0
, n
k
1
, . . . , n
k
; t) [n
k
0
, n
k
1
, . . . , n
k
(33)
where C(n
k
0
, n
k
1
, . . . , n
k
; t) are the same coecients in eqs. (31) and (33). The way to
verify this is by showing that both expansions lead to the same (linear) dierential equations
for such coecients, if the Hamiltonian when acting in the expansion (33), i.e. in the Fock
space, is expressed as

H =

kk

[h
0
[ kb

b
k
+
1
2

k
1
k
2
k
3
k
4
k
1
k
2
[u[ k
3
k
4
b

k
1
b

k
2
b
k
3
b
k
4
(34)
where
h
0
=
p
2
i
2m
+ V
ext
(r) (35)
is the Hamiltonian of one particle, and u = u([r
i
r
j
[) is the two-particle interaction
potential. The matrix elements are,
k

[h
0
[ k =
_
dr

k
(r)
_

2
2m

2
+ V
ext
(r)
_

k
(r) (36)
9
and
k
1
k
2
[u[ k
3
k
4
=
_
dr
_
dr

k
1
(r)

k
2
(r

)u([r r

[)
k
3
(r)
k
4
(r

). (37)
The demonstration that expansions (31) and (33) have the same coecients is laborious
but straightforward. Considering Schr odinger equation (3) for [(t) >, we suggest the
following steps:
First, use expansion (31) for the wavefunction (r
1
, r
2
, . . . , r
n
; t) = r
1
, r
2
, . . . , r
N
[(t),
and nd the equation that the coecients C(n
k
0
, n
k
1
, . . . , n
k
; t) satisfy. This is the most
dicult part. See Chapter 1 of Fetter and Walecka for suggestions.
Second, use expansion (33) for the state [(t) >, and use the form (34) for the Hamil-
tonian. Verify that you nd the same equation for C(n
k
0
, n
k
1
, . . . , n
k
; t) as before.
The procedure of using the occupation number expansion (33) for the state [(t) >,
together with the form of the Hamiltonian

H, eq.(34), is called Second Quantization. The
name is because the procedure uses the expectation values of the h
0
and u(r), eqs.(36) and
(37), which are already quantized (i.e. in rst quantization). To work in this scheme
proves to be much better, simpler and systematic than in the standard expansion (31),
but they are completely equivalent. Moreover, the treatment allows also for doing statistical
physics calculations that would be almost impossible in the other scheme.
I. An alternative form of Second Quantization. Field Theory.
Consider a system of bosons of spin s = 0. This makes the next considerations easier to
write down. Dene eld operators as the following linear superposition

(r) =

k
(r) b

(r) =

k
(r) b

k
(38)
They are called eld operators, because are operators that depend on the coordinate
position r. It is easy to check that obey the commutation rules:
_

(r),

(r

)
_
=

k
(r)

k
(r

) = (r r

)
_

(r),

(r

)
_
=
_

(r),

(r

)
_
= 0. (39)
10
Show that the Hamiltonian

H in Second Quantization, eq.(34), can be written as,

H =
_
d
3
r

(r)
_

2
2m

2
+ V
ext
(r)
_

(r)+
1
2
_
d
3
r
_
d
3
r

(r)

(r

)u([rr

[)

(r

(r).
(40)
Note that once we have written this form, it becomes very general, independent of the original
expansion (38). Moreover, it allows for nding the time evolution of the eld operators
themselves, using Heisenberg equation,
i

t

(r, t) =
_

(r, t),

H
_
. (41)
This will be used when discussing the Gross-Pitaevskii equation.
J. Comments
For purposes of presentation, and because they will be useful, we used the eigenstates
of momentum (with periodic boundary conditions) for the spatial part of the one-particle
states:

k
(r) =
1

V
e
i

kr
(42)
with

k = (2/L)(m
x
, m
y
, m
z
), m
i
= 0, 1, 2, . . . . However, it should be clear that we
could use any other complete basis of the spatial part (i.e. a complete set of three quantum
numbers corresponding to three commuting observables). One would then dene creation
and annihilation operators for particles in those one-particle states. It should also be clear
that we can consider arbitrary conning potentials V
ext
(r), such as the actual traps for cold
atoms; there is a subtlety, however, when taking the thermodynamic limit in these cases.
We will limit ourselves to the case of a box of rigid walls in this review and will later indicate
how the presence of an inhomogeneous trap can be included.
Once one has the formalism of Second Quantization, one can then proceed with tech-
niques of Green functions and Feynmann diagrams, that allow for systematic exact expan-
sions. These are dicult to calculate, except when there is a small parameter, then one can
realize a perturbation series scheme. To develop these techniques one requires, at least, a
one semester course. Excellent books for this are Fetter and Walecka (Quantum Theory of
Many-Particle Systems) and Landau and Lifshitz (Statistical Physics, Part II).
11
In this brief lectures we will use the Second Quantized fomalism to study, at the simplest
level, a weakly interacting Bose gas in the Bogolubov approximation, superuidity and the
solution of a linear vortex using the Gross-Pitaevskii equation.
12
A demonstration of the equality of expansions (31) and (33) .
We have the basis in the usual way,

n
k
0
,n
k
1
,...,n
k
(r
1
, r
2
, . . . , r
N
) =
_
n
k
0
!n
k
1
! . . . n
k
!
N!

perm

k
1
(r
1
)
k
2
(r
2
) . . .
k
N
(r
N
). (43)
where the set n (n
k
0
, n
k
1
, . . . , n
k
) is a collection of given occupation numbers with the
restriction

k
n
k
= N. (44)
We rewrite as follows for simplicity,

{n}
(r
1
, r
2
, . . . , r
N
) =
_
n
0
!n
1
! . . . n

!
N!

perm

(r
1
)

(r
2
) . . .

(r
N
). (45)
where we assume n

,= 0, n

,= 0, ... n

,= 0, all other n
k
= 0 for k ,= ,= ,= , = .
For one particle basis,
_
dr

(r)

=
,
. (46)
Let us check orthonormality of
{n}
(r
1
, r
2
, . . . , r
N
). First,
_
dr
1
dr
2
. . . dr
N

{n}
(r
1
, r
2
, . . . , r
N
)
{n

}
(r
1
, r
2
, . . . , r
N
) = 0 if n , = n

. (47)
If n = n

, we nd
_
dr
1
dr
2
. . . dr
N

{n}
(r
1
, r
2
, . . . , r
N
)
{n}
(r
1
, r
2
, . . . , r
N
) =
n
0
!n
1
! . . . n

!
N!

perm

perm

_
dr
1
dr
2
. . . dr
N

(r
1
)

(r
2
) . . .

(r
N
)

(r
1
)

(r
2
) . . .

(r
N
) (48)
where the set (

, . . . ,

) is the same as (, , . . . , ) but just a dierent permutation.


Fix one permutation of (, , . . . , ). Then, there is only one term in

perm

that gives
a contribution dierent from zero, i.e.
_
dr
1
dr
2
. . . dr
N

{n}
(r
1
, r
2
, . . . , r
N
)
{n}
(r
1
, r
2
, . . . , r
N
) =
n
0
!n
1
! . . . n

!
N!

perm
_
dr
1
dr
2
. . . dr
N

(r
1
)

(r
2
) . . .

(r
N
)

(r
1
)

(r
2
) . . .

(r
N
)
=
n
0
!n
1
! . . . n

!
N!

perm
= 1. (49)
13
where we used the important result, i.e. the total number of permutations is,

perm
=
N!
n
0
!n
1
! . . . n

!
(50)
Let us now consider the expansion,
(r
1
, r
2
, . . . , r
n
; t) =

{n

}
C(n

0
, n

1
, . . . , n

; t)
{n

}
(r
1
, r
2
, . . . , r
N
). (51)
Use it in Schr odinger equation to give,
i

t
C(n
0
, n
1
, . . . , n

; t) = (52)

{n

}
_
dr
1
dr
2
. . . dr
N
_

{n}
(r
1
, r
2
, . . . , r
N
)

H
{n

}
(r
1
, r
2
, . . . , r
N
)
_
C(n

0
, n

1
, . . . , n

; t)
where the set n in the left is a given set. For deniteness, let us take it again the
set (n

, n

, . . . , n

) all other zero. The sum in the right-hand-side is over all the sets of
occupation number that

k
n

k
= N.
We do the ideal case in detail, i.e. the kinetic energy term. You do the interaction term!
That is, we take

H =
N

i=1

T
i
=
N

i=1
p
2
i
2m
(53)
We call this a one-body operator because it involves operators of one particle only, i.e.
p
i
. It has matrix elements,
_
dr
i

(r
i
)

T
i

(r
i
) = [T[
=

2
k
2

2m

,
(54)
Notice that the right-hand-side of the rst line is independent of the value of i.
14
So, Schr odinger equation is,
i

t
C(n
0
, n
1
, . . . , n

; t) = (55)
N

i=1

{n

}
_
dr
1
dr
2
. . . dr
N
_

{n}
(r
1
, r
2
, . . . , r
N
)

T
i

{n

}
(r
1
, r
2
, . . . , r
N
)
_
C(n

0
, n

1
, . . . , n

; t)
=

{n

}
_
n
0
!n
1
! . . . n

!
N!
_
n

0
!n

1
! . . . n

!
N!
C(n

0
, n

1
, . . . , n

; t)
N

i=1

perm

perm

_
dr
1
dr
2
. . . dr
N

(r
1
)

(r
2
) . . .

(r
N
)T
i

(r
1
)

(r
2
) . . .

(r
N
) (56)
We rst note that the contribution of T
i
only comes with an integral of one

of the left
with one of the right. Therefore, there are two cases.
Case I The subindex of the left is equal to the subindex

of the right. i.e. =

.
Therefore, due to the orthonormality of the one-particle states , it follows that n

= n.
This means that of all the sum

{n

}
only survives the set of occupation numbers equal to
the xed one n.
This has a contribution in the right-hand-side of Schr odinger equation (56) equal to
n
0
!n
1
! . . . n

!
N!
C(n
0
, n
1
, . . . , n

; t)
N

i=1

perm

perm

_
dr
1
dr
2
. . . dr
N

(r
1
)

(r
2
) . . .

(r
N
)T
i

(r
1
)

(r
2
) . . .

(r
N
)
Fix a permutation of (, , . . . , ). Then, there can only be one permutation of (

, . . . ,

) =
(, , . . . , ),
n
0
!n
1
! . . . n

!
N!
C(n
0
, n
1
, . . . , n

; t)
N

i=1

perm

perm

_
dr
1
dr
2
. . . dr
N

(r
1
)

(r
2
) . . .

(r
N
)T
i

(r
1
)

(r
2
) . . .

(r
N
)
=
n
0
!n
1
! . . . n

!
N!
C(n
0
, n
1
, . . . , n

; t)
N

i=1

perm
_
dr
1
dr
2
. . . dr
N

(r
1
)

(r
2
) . . .

(r
N
)T
i

(r
1
)

(r
2
) . . .

(r
N
) (57)
15
Now it comes the crucial result:
N

i=1

perm
_
dr
1
dr
2
. . . dr
N

(r
1
)

(r
2
) . . .

(r
N
)T
i

(r
1
)

(r
2
) . . .

(r
N
)
=

perm

[T[ =

[T[

perm
=

[T[
N!
n
0
!n
1
! . . . n

!
(58)
We can sum over all values of because if n

= 0, it does not contribute.


Replacing, we get the contribution from Case I:
n
0
!n
1
! . . . n

!
N!
C(n
0
, n
1
, . . . , n

; t)
N

i=1

perm

perm

_
dr
1
dr
2
. . . dr
N

(r
1
)

(r
2
) . . .

(r
N
)T
i

(r
1
)

(r
2
) . . .

(r
N
)
=

[T[ n

C(n
0
, n
1
, . . . , n

; t). (59)
Case II. The subindex of the left is not equal to the subindex of the right of

(r
i
)T
i

(r
i
). Due to the orthonormality of the one-particle states , it follows that
n

= n

if ,= and ,= . But n

= n

1 and n

= n

+ 1. This means that of all the


sum

{n

}
only survives those sets of occupation numbers satisfying these requirments.
One gets the contribution,
_
n
0
! . . . n

!
N!

=
_
n
0
! . . . (n

1)! . . . (n

+ 1)! . . . n

!
N!
C(n
0
, . . . , n

1, . . . , n

+ 1, . . . , n

; t)
N

i=1

perm

perm

_
dr
1
dr
2
. . . dr
N

(r
1
)

(r
2
) . . .

(r
i
) . . .

(r
N
)T
i

(r
1
)

(r
2
) . . .

(r
i
) . . .

(r
N
)
Again, x a permutation (, , . . . , , . . . , ). There can only be one term of

perm

that
gives a contribution dierent from zero. Then, by running the sum over i = 1 to N, we get
16
a number of terms dierent from zero only equal to n

,
_
n
0
! . . . n

!
N!

=
_
n
0
! . . . (n

1)! . . . (n

+ 1)! . . . n

!
N!
C(n
0
, . . . , n

1, . . . , n

+ 1, . . . , n

; t)
N

i=1

perm

perm

_
dr
1
dr
2
. . . dr
N

(r
1
)

(r
2
) . . .

(r
i
) . . .

(r
N
)T
i

(r
1
)

(r
2
) . . .

(r
i
) . . .

(r
N
)
=
_
n
0
! . . . n

!
N!

=
_
n
0
! . . . (n

1)! . . . (n

+ 1)! . . . n

!
N!

C(n
0
, . . . , n

1, . . . , n

+ 1, . . . , n

; t)

perm
n

[T[
=
_
n
0
! . . . n

!
N!

=
_
n
0
! . . . (n

1)! . . . (n

+ 1)! . . . n

!
N!

C(n
0
, . . . , n

1, . . . , n

+ 1, . . . , n

; t)n

[T[
N!
n
0
!n
1
! . . . n

!
=

=
[T[

+ 1C(n
0
, . . . , n

1, . . . , n

+ 1, . . . , n

; t). (60)
We put it together in Schr odinger equation,
i

t
C(n
0
, n
1
, . . . , n

; t) =

[T[ n

C(n
0
, n
1
, . . . , n

; t)
+

=
[T[

+ 1 C(n
0
, . . . , n

1, . . . , n

+ 1, . . . , n

; t). (61)
If we now take the expansion
[(t) =

{n

}
C(n

0
, n

1
, . . . , n

; t)[n

0
, n

1
, . . . , n

(62)
and use it into Schr odinger equation
i

t
[(t) =

H[(t) (63)
with

H =

[T[

(64)
we obtain easily the same result as before (61). Simply, multiply Schrodinger equation on
the left by n
0
, n
1
, . . . , n

[, use the rules of the creation and annihilation operators and in


two steps one gets (61).
17
II. AN INTERACTING BOSE GAS AND THE SCATTERING LENGTH
The physical problem we want to address is to nd the macroscopic, thermodynamic and
dynamic, properties of an interacting Bose gas at low temperatures. We shall consider the
system given by Hamiltonian

H given by eq.(34), bosons with spin s = 0. At low temper-
atures the formalism of Second Quantization is particularly useful. At high temperatures,
the gas behaves classically.
From the point of view of statistical physics, the description of the thermodynamic states
of the system requires the knowledge of the equilibrium density matrix:
Canonical Ensemble. In this ensemble, the system is studied under xed temperatureT,
volume V and number of particles N,

C
=
e


H
Tr e


H
, (65)
where = 1/kT, and the average of any physical quantity

A can be calculated,

A =
1
Tr e


H
Tr
_
e


H

A
_
(66)
and for the trace one can use the occupation number basis, e.g.
Tr
_
e


H

A
_
=

n
k
0
,n
k
1
,...,n
k
n
k
0
, n
k
1
, . . . , n
k
[e


H

A[n
k
0
, n
k
1
, . . . , n
k
(67)
A BIG problem appears and it is that the sum over the occupation numbers is restricted by
the condition

k
n
k
= N. This problem is solved in the Grand Canonical ensemble.
Grand Canonical Ensemble. In this case, the system is studied under xed temperatureT,
volume V and chemical potential . The number of particles is found afterwards. The den-
sity matrix is,
=
e


N

H
Tr e


N

H
, (68)
with

N the particle-number operator,

N =

k
b

k
b
k
. (69)
Again, the average of any quantity is

A =
1
Tr e


N

H
Tr
_
e


N

H

A
_
(70)
18
but now the trace is
Tr
_
e


N

H

A
_
=

n
k
0
,n
k
1
,...,n
k
n
k
0
, n
k
1
, . . . , n
k
[e


N

H

A[n
k
0
, n
k
1
, . . . , n
k
(71)
with NO restriction on the sum, that is, one sums over all possible occupation numbers,
from 0 to . In particular, the average number of particles is

N =
1
Tr e


N

H
Tr
_
e


N

H

N
_
. (72)
This number depends on T, V and ,

N =

N(T, V, ). Therefore, if we equal the actual


number of particles N to its Grand Canonical average, N =

N, we can solve for the value


of , i.e. = (T, V, N), and use this esemble to calculate all the properties.
The relation to thermodynamics is via the Grand Potential = (T, V, ),
(T, V, ) = kT ln Tr e


N

H
=

H TS

N
= pV (73)
where S is the entropy, p the pressure, and the internal energy is E =

H.
As one can imagine, this procedure is still very hard to realize for an interacting system.
However, if the temperatures are very low, we can rst consider the case T = 0. This
situation, although prohibited to be achieved by the Third Law of Thermodynamics, can
nevertheless be studied and used as a basis to calculate the properties of the system at very
low temperatures. At zero temperature, the entropy is also zero, S = 0, and this means that
the N-particle system occupies only one state, which is its N-particle ground state. In this
case, the averages in (73) are in the ground state, denoted by [0,

A = 0[

A[0. (74)
... this is still hard to do! but with physical and sensible assumptions one can nd the
ground state approximately. To begin, we shall see rst that the interacting part of the
Hamiltonian can be simplied, at low temperatures, by the introduction of the scattring
length.
19
A. The scattering length
Let us go back to Hamiltonian (34) and calculate explicitely the dierent terms. Here,
we use k

k, the 3D wavevector:

H =

kk

[h
0
[ kb

b
k
+
1
2

k
1
k
2
k
3
k
4
k
1
k
2
[u[ k
3
k
4
b

k
1
b

k
2
b
k
3
b
k
4
. (75)
where (show)
k

[h
0
[ k =
_
d
3
r

k
(r)
_

2
2m

2
+ V
ext
(r)
_

k
(r)
=

2
k
2
2m

kk
(76)
and
k
1
k
2
[u[ k
3
k
4
=
_
d
3
r
1
_
d
3
r
2

k
1
(r
1
)

k
2
(r
2
)u([r
1
r
2
[)
k
3
(
1
)
k
4
(r
2
).
=
k
1
+k
2
,k
3
+k
4
1
V
_
d
3
r e
i(k
1
k
3
)r)
u(r)
=
k
1
+k
2
,k
3
+k
4
1
V

U(k
1
k
3
) (77)
where in the last line we have dened the Fourier transform of the interaction potential
u(r). This matrix element can be interpreted as given the collision of two particles via the
interaction potential, and the delta function indicates conservation of momentum.
To evaluate the above integrals and for further use, we list the properties of the basis

k
(r)

k
(r) =
1

V
e
ikr
_
d
3
r e
i(kk

)r
= V
k,k

k
e
ik(rr

)
= V (r r

). (78)
Substituting the above matrix elements in the Hamiltonian (75), and after some rear-
rengements, one nds the useful form, (show)

H =

2
k
2
2m
b

k
b
k
+
1
2V

kk

U(q)b

k+q
b

q
b
k
b
k
. (79)
As mentioned before, the interaction term can be physically interpreted as the collision
of two particles, with momenta k and k

before the collision (i.e. they are annihilated at


20
the collision), scattered into two particles with momenta k +q and k

q after the collision


(i.e. created at the collision). The collision is mediated by the interaction term

U(q)
dependent on the transferred momentum q.
B. Two-particle collisions at low energies.
We are working under the assumption of low temperatures, that means that the kinetic
energies of the particles is small, i.e.
2
k
2
/2m small, which implies k small. This sug-
gests considering collisions in the limit k 0. We shall see now that this has a profound
implication on the form of

U(q), which in turn simplies enormously the problem.
Consider the collision of two identical particles of mass m at low energies, interacting via
the potential u([r
1
r
2
[).
FIGURE
The scattering is best studied by separating the motion in center of mass R = (r
1
+r
2
)/2
and relative r = r
1
r
2
coordinates. The center of mass motion is that of a free particle and
we ignore it. The collision occurs in the relative motion and corresponds of the scattering of
a particle of mass m/2 by a center of force u(r). We assume that the situation is stationary,
that is, a current of particles with incident wavefunction e
ikr
is scattered into a wavefunction

sc
(r). The goal is to determine such a scattered wavefunction. The total wafunction is found
by solving the (reduced one-particle) Schr odinger equation,


2
2m/2

2
(r) + u(r)(r) = E(r) (80)
A detailed analysis of this problem in terms of the partial waves expansion can be found in
any book in Quantum Mechanics, we urge you to study it! Here, because of our purposes,
we will briey present the scheme of scattering in kspace.
Propose the (unnormalized) wavefunction,
(r) = e
ikr
+
sc
(r) (81)
Dene Fourier transforms (in a large box of volume V ) of a function f(r)

f(k) =
_
V
d
3
r e
ikr
f(r) and f(r) =
1
V

k
e
ikr

f(k) (82)
21
with
_
V
d
3
r e
i(kk

)r
= V
k,k
and

k
e
ik(rr

)
= V (r r

). (83)
The values of k are the same for periodic boundary conditions,

k =
2
L
(n
x
, n
y
, n
z
) and n
x
= 0, 1, 2, . . . ; n
y
= 0, 1, 2, . . . ; n
z
= 0, 1, 2, . . . (84)
Scattering in k-space. Take the Fourier transform of both sides of Schrondiger equation
(80), work a little bit, and arrive to the following result (careful! the vector k is xed by the
incident wave, so take the Fourier transforms using k

and k

):

2
k
2
m

(k

) +
1
V

U(k

(k

) = E
k

(k

) (85)
where we hace labeled the energy E
k
, since the solution corresponds to a given value of k
(see below). Now, use

(k

) = V
k,k
+

sc
(k

) (86)
and one gets an equation for

sc
(k

),

2
k
2
m
_
V
k,k
+

sc
(k

)
_
+

U(k

k)+
1
V

U(k

sc
(k

) = E
k
_
V
k,k
+

sc
(k

)
_
(87)
We now search for solutions that represent that far from the region of the collision, the
particle is free, and by conservation of energy, the energy E
k
of the above equation is just
the kinetic energy of the incident particle,
E
k
=

2
k
2
m
. (88)
This yields the equation for the scattered amplitude

sc
(k

),
_

2
k
2
m


2
k
2
m
_

sc
(k

) =

U(k

k) +
1
V

U(k

sc
(k

). (89)
We now dene the scattering T matrix, as the solution of the above equation. That is,
dene rst,

sc
(k

) =
1
E
k
E
k
+ i
T(k

, k; E
k
) (90)
where 0
+
is a quantity needed to get outgoing waves in the scattered solution, when
the inverse Fourier transform is taken. Substitution of (90) into (89) gives the equation for
T,
T(k

, k; E
k
) =

U(k

k) +
1
V

U(k

)
1
E
k
E
k
+ i
T(k

, k; E
k
)). (91)
22
This is called the Lippman-Schwinger equation. It is a self-consisten (i.e. integral) equation
for T. Very hard to solve for an arbitrary potential u(r), but it is in a form that allows for a
perturbation analysis, as we will use below. Our goal in this notes is to make the connection
with the scattering length.
Scattering in r-space. Let us return to problem of scattering in real space r. We will give
just the necessary informatio here, you should look at a good book in Quantum Mechanics
for the details. We write now all the vectors to make it clearer. The Schr odinger equation
for the relative coordinate is,


2
2m/2

2
(r) + u([r[)(r) = E(r) (92)
and the (unnormalized) wavefunction is proposed as,
(r) = e
i

kr
+
sc
(r). (93)
Consider spherical coordinates r = (r, , ). Because of the radial symmetry of the potential
u(r), r [r[, we choose the incident wavevector along z,

k = k z, and therefore the scattering


must be independent of , i.e the scattered wavefunction depends only r and ,
sc
(r) =

sc
(r, ).
In general, the problem is very hard for arbitrary potentials. However, all interatomic
potentials are expected to vanish, u(r) 0, for r r
0
, r
0
being called the potential range
and typically is of the order of few angstroms. Hence, we can search for solutions at large
distances from where the collision takes place, i.e. for r r
0
, as outgoing spherical waves
with a coecient

sc
(r, ) f()
e
ikr
r
for r r
0
. (94)
CAREFUL! Here kr = [

k[[r[. The coecient f() is called the scattering amplitude, and


bears all the information of the collision; as you can imagine, it is intimately related to the
scattering matrix T,
f() =
m
4
2
T(

k
0
, k z; E
k
) with [

k
0
[ = k and

k
0
z = k
0
cos . (95)
Here we will only look at the relation as k 0.
In spherical coordinates, one proposes a solution
(r, ) =

l=0
A
l
P
l
(cos )R
lk
(r) (96)
23
where l = 0, 1, 2, . . . are the angular momentum quantum numbers, and P
l
(cos ) the Leg-
endre polinomials. Schrodinger equation (92) for R
lk
(r) looks as
1
r
2
d
dr
_
r
2
d
dr
_
R
lk
(r) +
_
k
2

l(l + 1)
r
2

m

2
u(r)
_
R
lk
(r) = 0. (97)
This procedure yields the value of f():
f() =

l=0
(2l + 1)f
l
P
l
(cos ). (98)
A very important quantity is the scattering cross section , which gives us a measure of
the total probability of a particle to be scattered. It may also be seen as the eective
area that one particle shows to the other in the collisions. The larger this is, the more the
particles interact and the more are scattered:
= 2
_

0
[f()[
2
sin d. (99)
The relevant approximation for our purposes is to look for solutions at low energy k 0.
In this case, we can consider only l = 0, this is called swave scattering, and neglect all
other values of l (the problem becomes independent of !). One can then show that, for
r r
0
, one obtains the following results,
f()
1
2ik
_
e
2i
0
1
_
(100)
where
0
is called the s wave phase-shift. Its value and its sign depend very strongly on
the specic form of the potential u(r). That is, all the information of the details of the
potential on the collision, is contained in
0
. We shall see it in a moment.
The scattering length. The scattering length a, in general, is dened as,
tan
0
= ka. (101)
In the limit of small k, that is [ka[ 1, one can clearly approximate,
tan
0
sin
0

0
ka. (102)
The s-scattered wave can be written as,

sc
(r) = f()
e
ikr
r

1
2ik
_
e
2i
0
1
_
e
ikr
r

sin
0
kr
e
i(kr+
0
)
(103)
24
and nally, with the approximation [ka[ 1, as

sc
(r)
a
r
e
ik(ra)
. (104)
There is a very nice physical rationalization of this formula regarding the sign of a (said to
me by Vanderlei Bagnato): The scattered wave is a simple spherical wave, that for large
r, appears a being emitted from a center not at r = 0 but at r = a. Thats why appears
shifted. So, if a > 0, it appears to have been scattered at a point a which is closer to the
observation point, i.e. in front of the origin. Thus, the net or eective interaction appears
repulsive. On the contrary, if a < 0, the wave appears to be emitted at r = [a[, that is, as
if it have been scattered at a point behind the center! i.e. as if it had spent more time in
the collision region. In this case, the net or eective interaction is attractive. The sign of a
is fundamental for the stability or not of a BEC condensate, it needs to be positive a > 0
(... while for the creation of BCS states in fermions, it must be negative).
Connection of a with T. Now, the connection with the matrix T, and therefore with
the Fourier transform of the potential, which is what we need to continue with our Many
Body problem.
Recall equation (90) relating the Fourier transform of the scattered wave to the matrix
T. We take the inverse Fourier transform and obtain,

sc
(r) =
1
V

e
ik

r
1
E
k
E
k
+ i
T(k

, k; E
k
) (105)
Now, take the limit of a very large box V (thermodynamic limit),

V
(2)
3
_
d
3
k as V (106)
to get

sc
(r) =
1
(2)
3
_
d
3
k

e
ik

r
1
E
k
E
k
+ i
T(k

, k; E
k
). (107)
Consider now a cold, low energy collision, E
k
0 and k 0. Although it can be made
more rigorous, we can make the assumption, inside the integral, that,
T(k

, k 0; E
k
0) T(0, 0; 0), (108)
25
such that in the strict limit k 0, one obtains,

sc
(r) =
1
(2)
3
_
d
3
k

e
ik

r
1
E
k
E
k
+ i
T(k

, k; E
k
)

1
(2)
3
m

2
T(0, 0; 0)
_
d
3
k

e
ik

r
k
2

1
(2)
3
m

2
T(0, 0; 0)
_
2
2
r
_
=
m
4
2
T(0, 0; 0)
1
r
(109)
Comparing with the equation (104) of
sc
(r) in the strict limit k 0, we arrive at the
fundamental result,
T(0, 0; 0) =
4
2
a
m
. (110)
Connection of a with

U(

k). Go back now to the Lippman-Schwinger equation and take


the limit k = 0 and k

= 0, one obtains,
T(0, 0; 0) =

U(0) +
1
V

U(k

)
1
E
k
+ i
T(k

, 0; 0)). (111)
This is still an exact equation. One can use it now for a perturbation series. To lowest order,
one neglects the second term and obtains the most used result in these studies, namely
T(0, 0; 0)

U(0), (112)
namely

U(0)
4
2
a
m
. (113)
Notice that

U(0) =
_
d
3
r u(r)
4
2
a
m
. (114)
This form is also used, specially when working in the eld operator version.
To next order, assuming again that only values of k 0 are relevant, one can write
T(0, 0, 0)

U(0) +
1
V

U(0)T(0, 0; 0)

1
E
k
+ i
4
2
a
m


U(0)
1
V

U(0)
4
2
a
m

1
E
k
+ i
(115)
To the same order of validity, we can solve for

U(0) yielding,

U(0) U
0
+
U
2
0
V

1
E
k
+ i
(116)
26
where we have introduced a common notation,
U
0
=
4
2
a
m
(117)
The last sum in (116), when coneverted to an integral, is actually divergent but this form
will be used to precisely cancel another divergence! We shall keep these two forms, eq(113)
and (116) for future calculations.
27
III. A DILUTED INTERACTING BOSE GAS AT VERY LOW TEMPERATURES.
BOGOLUBOV APPROXIMATION
We are nally ready to start the study of the Bose gas at low temperatures! We return
to the Hamiltonian (79) and take the appropriate form for k 0, i.e. for low temperatures,
using

U(q)

U(0), and from the discussion of the previous section we know its value as a
function of the scattering length a. Thus, we shall use the following form of the Hamiltonian
from now on,

H =

2
k
2
2m
b

k
b
k
+
1
2V

U(0)

kk

q
b

k+q
b

q
b
k
b
k
. (118)
We consider repulsive interactions, that is, positive scattering length a > 0. If negative, the
gas becomes unstable and the phenomenon of condensation does not occur.
A. Ideal Bose-Einstein Condensation (BEC).
The physical idea to approach the interacting gas is rooted in the study of the ideal gas.
In this case, one neglects all interactions and nds,

H
0
=

0
k
b

k
b
k
. (119)
where

0
k
=

2
k
2
2m
. (120)
Note that the many-body Hamiltonian reduces to a sum of one-body Hamiltonians, one
for each value of k. The solution to this problem is known exactly, because the number-
occupation basis is the basis of b

k
b
k
. The solution is,

H
0
[n
k
0
, n
k
1
, . . . , n
k
= E
{n
k
0
,n
k
1
,...,n
k
}
n
k
[n
k
0
, n
k
1
, . . . , n
k

=
_

0
k
n
k
_
[n
k
0
, n
k
1
, . . . , n
k
. (121)
If the system has N xed number of particles, the occupation numbers n
k
satisy

k
n
k
= N.
Thus, the ground state is n
k
= 0 for all values of k ,= 0, that is n
0
= N and E
N,0,0,...
= 0:
all the particles are in the ground state k = 0 of one particle.
28
For nite temperatures, we use the grand canonical ensemble and one nds the following
results (see e.g. Landau and Lifshitz, Statistical Physics I): the grand potential is

0
(T, V, ) = kT
V

3
T
g
5/2
() (122)
from which one can nd all thermodynamics. In particular, the mean number of particles is
N =

k
b

k
b
k

k
n
k

k
1
e
(
k
)
1
=
V

3
T
g
3/2
(). (123)
In the last line we took the thermodynamic limit, eq.(106). This last expression gives
N = N(V, T, ), which can be inverted for xed N to yield = (T, N, V ).
In the above formulae we used the de Broglie wavelength,

T
=
h

2mkT
(124)
and the Bose functions,
g
n
() =
1
(n)
_
dx
x
n1
e
x
1
. (125)
Look at the Eq.(123). Because g
3/2
(0) 2.61 is nite, this implies that for xed particle
density N/V , there is a characteristic temperature T
c
, called the BEC temperature, at which
the chemical potential vanishes (T
c
, N, V ) = 0,
N
V
=
_
2mkT
c
h
2
_
3/2
g
3/2
(0). (126)
For lower values of the temperature, T T
c
, the one-particle ground state k = 0 begins
to be macroscopically occupied, i.e. BEC sets in. One nds that the condensate, i.e. the
number of particles in the many body ground state, N
0
= n
k
0
=0
, is given by
N
0
(T) = N
_
1
_
T
T
c
_
3/2
_
. (127)
It is important to notice that the occupation of the one-particle ground state is the same as
the occupation of the ground state of the Nparticle system. This changes with interactions!
29
B. Interacting gas.
Once we consider interactions among the particles, the behavior is expected to change
but not so much. In particular, the BEC phenomenon,i.e. the macroscopic occupation of
the one-particle ground state, is expected to still be true for very low temperatures, specially
if the interaction is not very strong. As we will see below, weak interactions means,
Na
3
V
1. (128)
Namely, the separation between the particles must be much larger than the scattering length,
therefore, one can always be in this limit by diluting the gas. The mathematical idea is to
try to make the correct approximations, such that the interacting Hamiltonian (118) looks
like an ideal one, as (119). You will see.
Bogolubov reasoned in the following way. First, we look at zero temperature T = 0 to
nd the ground state of the interacting gas. Clearly, because of the interacting term, the
occupation number state are no longer eigenstates of

H, therefore, the N-particle ground
state cannot be n
k
0
= N
0
= N and n
k
= 0 for k ,= 0. Nevertheless, we expect that the
in the true ground state, there should be many particles in the one-particle ground state
k = 0. We call this number of particles, still, the condensate. The main assumption is that
N
0
N, i.e. is a nite fraction of the total number of particles. In the true many-body
ground state, there are particles in the excited one-particle states k ,= 0. We now show how
to include those states.
We recall rst, that b

k
0
b
k
0
[n
k
0
, n
k
1
, . . . , n
k
= N
0
[n
k
0
, n
k
1
, . . . , n
k
. But since this num-
ber is expected to b N
0
n
k
for all k ,= 0, we make the approximation of replacing, in the
Hamiltonian

H eq.(118), the operators b

k
0
and b
k
0
by a number, i.e.
b

k
0
N
1/2
0
b
k
0
N
1/2
0
. (129)
This is justied, for when taking matrix elements of

H in the number-occupation states
[n
k
0
, n
k
1
, . . . , n
k
, we will nd factors like

N
0
1 and/or

N
0
2. But since N
0
N,
we make a very small error by approximating them by

N
0
. With this idea in mind, we
30
can rst write the sum of the interaction term as follows, (show)

kk

q
b

k+q
b

q
b
k
b
k
= b

0
b

0
b
0
b
0
+ 4b

0
b
0

k=0
b

k
b
k
+ b
0
b
0

k=0
b

k
b

k
+ b

0
b

k=0
b
k
b
k
+
2b
0

k=0

q=k
b

q
b

kq
b
k
+ 2b

k=0

q=k
b

k+q
b

q
b
k
+
2

k=0

q=0
b

k
b

q
b
q
b
k
+

q=0

k=q

=k=q
b

k
b
k

q
b
k
b
kq
. (130)
We now approximate b

k
0
N
1/2
0
and b
k
0
N
1/2
0
as discussed above and implement the
second important approximation, which is to neglect terms with three or four b
k
and b

k
for
k ,= 0. This is because those imply terms proportional to products of numbers of particles
n
k
in states with k ,= 0, but we are assuming those occupation numbers are smaller than
N
0
. With these approximations the Hamiltonian (118) becomes,

H =

k
_

0
k
+
2N
0

U(0)
V
_
b

k
b
k
+
N
2
0
2V
+
N
0

U(0)
2V

k=0
_
b

k
b

k
+ b
k
b
k
_
. (131)
There is a small but important point. We note that the number of particles within this
approximation is,
N = N
0
+

k=0
b

k
b
k
, (132)
such that, at the same level of approximation as the Hamiltonian (i.e. neglecting products
of n
k
)
N
2
N
2
0
+ 2N
0

k=0
b

k
b
k
. (133)
Hence, implementing this approximation in the Hamiltonian (131), we obtain Bogolubovs
Hamiltonian, correct to the approximation used,

H =
N
2

U(0)
2V
+

k
_

0
k
+
N

U(0)
V
_
b

k
b
k
+
N

U(0)
2V

k=0
_
b

k
b

k
+ b
k
b
k
_
. (134)
The rst term is a constant that depends on the number of particles only. The second one
may be seen as that of an ideal gas with an eective one-particle energy that also depends on
the density of the gas. If we kept only these terms, we would get the ground state of the gas
in the so-called Hartree-Fock contributions. The last term is very interesting, it is quadratic
in the creation and annihilation operators but not in the form b

b, thus they are not number


operators. The interpretation of b
k
b
k
is that of a collision of two particles, one of momentum
31
k with another of momentum k, scattered into two particles with momentum zero. The
other, b

k
b

k
, is the collision of two particles of momentum zero scattered into two particles
of momentum k and k. Thus, the N-particle ground state does not correspond to all
particles with momentum k = 0, but also includes these possibilities of scattering. The
solution was also provided by Bogolubov.
C. The Bogolubov transformation.
Since the Hamiltonian (134) is quadratic in the creation and annihilation operators, one
can make a linear transfomation to make it look as an ideal type of Hamiltonian. Consider
the following transformation in terms of new operators
k
and
k
,

k
= u
k
b
k
+ v
k
b

k
= u
k
b
k
+ v
k
b

k
, (135)
where u
k
and v
k
are real coecients to be determined, with the property u
k
= u
k
and
v
k
= v
k
. These operators are boson operators if they obey the commutation rules,
[
k
,
k
] =
k,k
. (136)
Show that this is true if,
u
2
k
v
2
k
= 1. (137)
With this considerations, the inverse transformation is
b
k
= u
k

k
v
k

k
b
k
= u
k

k
v
k

k
. (138)
32
Substitution into

H, eq.(134), yields, with previous rearrangement and denitions, (do it!)

H =
N
2

U(0)
2V
+

k>0
__

0
k
+
N

U(0)
V
_
_
b

k
b
k
+ b

k
b

k
_
+
N

U(0)
V
_
b

k
b

k
+ b
k
b
k
_
_
=
N
2

U(0)
2V
+

k>0
_

k
_
b

k
b
k
+ b

k
b

k
_
+
N

U(0)
V
_
b

k
b

k
+ b
k
b
k
_
_
=
N
2

U(0)
2V
+

k>0
_
_

k
+

k
_
_
(u
2
k
+ v
2
k
)
k

2N

U(0)
V
u
k
v
k
_
+ 2v
2
k

2N

U(0)
V
u
k
v
k
+
_

k
+
k

k
_
_
2
k
u
k
v
k
+
N

U(0)
V
(u
2
k
+ v
2
k
)
__
(139)
where

k
=
0
k
+
N

U(0)
V
. (140)
The above Hamiltonian becomes an ideal-type of Hamiltonian (i.e. proportional to terms

k
) if the last term proportional to

k
+
k

k
does not contribute. This can be done
by choosing the values of u
k
and v
k
such that the coecient that multiplies the last term is
zero, i.e.
2
k
u
k
v
k
+
N

U(0)
V
(u
2
k
+ v
2
k
) = 0 k. (141)
The resulting values of u
k
and v
k
are,
u
2
k
=
1
2
_
_

k
_

0 2
k
+ 2
N

U(0)
V

0
k
+ 1
_
_
v
2
k
=
1
2
_
_

k
_

0 2
k
+ 2
N

U(0)
V

0
k
1
_
_
(142)
D. Elementary excitations and ground state energy.
The Hamiltonian

H can now be written in term of the transformed operators, (do it!)

H =
N
2

U(0)
2V
+

k>0
_
_

0 2
k
+ 2
N

U(0)
V

0
k

0
k

N

U(0)
V
_
_
+

k=0

0 2
k
+ 2
N

U(0)
V

0
k

k
. (143)
33
This has a lot of profound information. The term in the second line, indicates that, in
addition to the rst contribution, the Hamiltonian is that of an ideal gas, not of particles,
but of quasi-particles that are created and annihilated by the operators
k
and

k
, with
momentum k and energy
k
, given by

k
=

0 2
k
+ 2
N

U(0)
V

0
k
. (144)
In the next section we will see that this spectrum leads to the phenomenon of superuidity.
These quasiparticles are also called elementary excitations. If we see again the form of the
operators
k
and

k
, eq.(135), one nds that the their eigenstates are (quantum) superpo-
sitions of occupation number states with dierent number of particles in the states k and
k, thus the excitations are collective excitations of the whole gas. However, we can nd
occupation-number states for the operators
k
and

k
, call them n
()
k
with the same rules
as for creation-annihilation operators,

n
()
k
=
_
n
()
k

n
()
k
1

n
()
k
=
_
n
()
k
+ 1

n
()
k
+ 1 . (145)
Therefore, the states of the N-particle system, within this approximation, are the occupation
number states [n
()
k
1
, n
()
k
2
, . . . , n
()
k
but for the quasiparticles. Clearly, the ground state
corresponds to n
()
k
= 0 for all k ,= 0. We call this state [0. Thus, the energy of the ground
state of the Nparticle system, E
0
= 0[

H[0 is the rst line of (143):


E
0
=
N
2

U(0)
2V
+

k>0
_
_

0 2
k
+ 2
N

U(0)
V

0
k

0
k

N

U(0)
V
_
_
. (146)
With the results of last section on the scattering length, we can evaluate this expression
explicitly. If we consider

U(0) as a constant, i.e. to lowest order, the second term of E
0
diverges; this divergence can be canceled if we take next order in the rst term of E
0
, see
eqs.(113) and (116):
E
0
=
N
2
2V
_
U
0
+
U
2
0
2V

k
1

0
k
_
+
1
2

k
_
_

0 2
k
+ 2
NU
0
V

0
k

0
k

NU
0
V
_
(147)
where
U
0
=
4
2
a
m
. (148)
34
It is now a classic calculation that you must do for E
0
. Change the sums to integrals using
the thermodynamic limit (106), and show that the result is,
E
0
=
N
2
U
0
2V
_
1 +
128
15
1/2
_
Na
3
V
_
1/2
_
. (149)
This result was originally found by Yang and Lee. To lowest order is the Gross-Pitaevskii
(Hartree-Fock) ground state energy of a uniform Bose gas. The rst correction indicates
that for this to be small, Na
3
/V 1, i.e. weakly interacting. To obtain higher order
corrections the procedure is highly non-trivial and there are many articles devoted to it (see
e.g. Yukalovs review).
35
IV. SUPERFLUIDITY
One of the most celebrated successes of Bogolubov theory is that it shows that a weakly
interacting gas shows the phenomenon of supeuidity at low temperatures. The reason is
due to the form of the energy spectrum of the elementary excitations,

k
=
_

0 2
k
+ 2
NU
0
V

0
k
. (150)
In this section, we review briey Landaus criterion of superuidity and verify that these
excitations satisfy it.
A. Landaus criterion for superuidity.
Superuid is a quantum phenomenon characterized by several properties of which, we
shall only mention two: a) the capability of owing through narrow capillary tubes without
viscosity; and b) the creation of quantized vortices when the uid rotates.
superfluid
velocity v
capillary
FIG. 2. A superuid owing through a capillary with velocity v.
Let us look rst at the property of owing through narrow capillaries without viscosity.
Consider a uid at T = 0 and assume it is moving with velocity v along a capillary tube.
36
Because of interactions with the walls, there is the possibility of generating elementary
excitations in the uid. The kinetic energy of the gas will then be diminished by transfering
to the creation of those excitaions and the uid will slow down; that is, the excitations will
generate viscosity. Let us see how this can happen.
We thus rst assume that the uid is moving with velocity v. Now let us make a change
of reference of frame to one with velocity v. In this, the uid appears static and the walls
of the capillary move with velocity v. Because of interactions with the walls, consider
an excitation is created with energy (p) and momentum p = k. Thus, in this frame the
energy of the uid is E

= (p) and the momentum of the uid is



P

= p. Now, let us return
to the frame in which the walls are at rest. The energy of the uid in this frame is
E = E

+
(

P

+ Mv)
2
2M
= (p) + p v +
1
2
Mv
2
+
p
2
2M
(p) + p v +
1
2
Mv
2
(151)
where M is the mass of the uid, assumed macroscopic, thus allowing to neglect the last
term of the second line. The term Mv
2
/2 is the kinetic energy of the uid in the absence
of the excitation. Thus, the term (p) + p v is the change of energy of the uid after the
creation of the excitation. This must be negative, for the dissipation of energy must reduce
the energy of the uid, i.e.
(p) + p v < 0. (152)
The rst term of this expression is positive, therefore, for a given value of p, the minimum
value of [v[ for this condition occurs when p and v are antiparallel, Therefore, the condition
to create an excitation is
(p) < pv. (153)
In words, the energy of the excitation must be smaller than the momentum of the excitation
times the velocity of the ow. If this is not satised, the excitation cannot be created and
therefore the uid will not slow down, i.e. it will not have viscosity. It will be a superuid.
37
B. Ideal gas.
The ideal gas cannot be a superuid. This is surprising at rst sight, but it was the
observation of Landau, which in turn explained that a BEC in an ideal gas and the transition
to superuidity, although related, are not the same thing.
The reason for an ideal gas not to be a superuid is that the elementary excitations out of
the N-particle ground state [0 = [0, 0, . . . , 0, are free particles with the energy spectrum,
(p) =
p
2
2m
(154)
In gure 3 we can see that, no matter how small the velocity v of the ow is, there will always
excitations with > pv and, therefore, the kinetic energy of the uid will be transferred to
the excitation and the uid will slow down, i.e. it will have viscosity.
pv

p
p
FIG. 3. Elementary excitation spectrum of an ideal gas, (p) = p
2
/2m. v is the velocity of the ow
along the capillary.
38
C. Bogolubov interacting gas.
An interacting gas can show superuid motion. This follows from the results of Bogol-
ubov (previous section). The ground state of the Nparticle gas is that of no elementary
excitations. But now the elementary excitations have the following spectrum, see eq.(155),
(p) =

_
p
2
2m
_
2
+
NU
0
V
p
2
m
. (155)
Now, if the momentum is very small the rst term can become negligible with respect to
the second one, yielding,
(p)
_
NU
0
mV
_
1/2
p
=
_
4
2
a
m
2
_
1/2
p
= pc
s
(156)
where in the last line we have identied the speed of sound of the uid as
c
s
=

4a
m
. (157)
Notice that it depends on the uid density = N/V and on the value of the scattering
length a.
Thus, for small momenta, the spectrum of the excitations is
(p) = pc
s
for pc
s

p
2
2m
. (158)
This is the spectrum of acoustic phonons (sound waves) with speed c
s
(this speed can also
be calculated from pure thermodynamic arguments, check it!). That is, the elementary
excitations of a uid of interacting bosons are sound waves, in contrast to the free particle
excitations of an ideal one.
As shown in gure 4, if the velocity of ow of the uid v is smaller than the speed of
sound c
s
, according to Landaus criterion, the uid cannot transfer its kinetic energy to the
creation of excitation and, therefore, it will ow with no viscosity. It behaves as a superuid.
One would think that the limit is the speed of sound of the uid but it turns out it is
not that simple. A more careful analysis shows that creation of vortices induces an eective
viscosity, and the lost of superuidity occurs at a much lower ow velocity than the speed
of sound.
39

p
p
pv
pc
s

FIG. 4. Elementary excitation spectrum of a interacting gas at low momentum p, (p) pc
s
, where
c
s
is the speed of sound. v is the velocity of the ow along the capillary.
V. VORTICES IN SUPERFLUIDS. GROSS-PITAEVSKII EQUATION.
A phenomenological approach to the hydrodynamics of superuids is through the so-
called two-uid model, due to Landau and Tisza. This is inspired on the fact that at nite
temperature, but below a critical temperature, the uid is not completely superuid but
only a part of it. It behaves as if one part were normal, with viscosity, and the other
as superuid with no viscosity. One must be careful not to believe that these parts are
mixed, in the usual sense, and that could be separated by some physical procedure. In this
way, the two-uid model is incorrect. One must think instead that the uid is quantum and
that it shows properties of both type of uids simultaneously. The model gives a lot insight
anyway. An important one is that one can assign a macroscopic velocity v
n
to the normal
part and a velocity v
s
to the superuid part v
s
. With the use of the Many-Body theory,
within the Gross-Pitaevskii approximation, one can reach the very important conclusion
that superuid ow is irrotational, which in turn implies the possibility of vortices in the
uid. We shall briey review this here and use the Gross-Pitaevskii equation to analyze the
40
structure of a single vortex.
A. The Gross-Pitaevskii equation and the wavefunction of the condensate.
Let us return to the eld-operator version of the second quantized Hamiltonian, eq.(40)

H =
_
d
3
r

(r)
_

2
2m

2
+ V
ext
(r)
_

(r)+
1
2
_
d
3
r
_
d
3
r

(r)

(r

)u([rr

[)

(r

(r),
(159)
with Heisengerg equation for the evolution of the elds
i

t

(r, t) =
_

(r, t),

H
_
. (160)
For low temperatures, one can assume that the most important one-particle wavefunction
are those for small k, which change very slowly at microscopic scales, i.e. they have few
nodes in lengths of the order of V
1/3
, the size of the system. Within these assumptions one
can approximate the true interatomic potential u(r) by a contact potential,
u([r[) U
0
(r) (161)
where U
0
= 4
2
a/m. We can make this identication remembering one of our previous
results,

U(0) =
_
d
3
r u(r) (162)
with

U(0) U
0
to lowest order. With this approximation the Hamiltonian looks as,

H =
_
d
3
r

(r)
_

2
2m

2
+ V
ext
(r)
_

(r) +
1
2
U
0
_
d
3
r
_
d
3
r

(r)

(r

)(r r

(r

(r)

H =
_
d
3
r

(r)
_

2
2m

2
+ V
ext
(r)
_

(r) +
1
2
U
0
_
d
3
r

(r)

(r)

(r)

(r) (163)
Heisenberg equation for

(r, t) is
i

t

(r, t) =
_

2
2m

2
+ V
ext
(r)
_

(r, t) + U
0

(r, t)(r, t)

(r, t). (164)


Recall that the elds are dened as,

(r) =

k
(r)b
k
,

(r) =

k
(r)b

k
(165)
41
such that, Heisenberg equation is an equation for the creation and annihilation operators b

k
and b
k
.
We now consider the uid to be at zero temperature T = 0. In this case, we expect that
most of the uid is in the ground state of the condensate of non-interacting particles, i.e.
those with k = 0 and wavefunction
0
(k). We do know that there are particles out of the
condensate. However, the approximation here (in the spirit of Hartree-Fock) is that all the
particles have momentum k = 0, i.e. are in the condensate, but that the one-particle is not

0
(r) but a function

0
(r) to be determined. In other words, we can approximate in this
case,

(r, t)

0
(r, t)b
0

(r, t)

0
(r, t)b

0
(166)
Notice that the action of the eld operators in the occupation number states is

(r, t)[n
0
, n
1
, . . . , n

0
(r, t)

n
0
[n
0
1, n
1
, . . . , n

(r, t)[n
0
, n
1
, . . . , n

0
(r, t)

n
0
+ 1[n
0
1, n
1
, . . . , n

. (167)
But the approximation is that n
0
N and N is macroscopic, N . Thus, the eld
operators can be considered to be diagonal in the number states, and the 1 changes be
neglected

(r, t)[n
0
, n
1
, . . . , n

0
(r, t)

n
0
[n
0
, n
1
, . . . , n

(r, t)[n
0
, n
1
, . . . , n

0
(r, t)

n
0
[n
0
1, n
1
, . . . , n

. (168)
If we now make a further identication

0
(r, t) =

n
0

0
(r, t) (169)
we can write,

(r, t)
0
(r, t)b
0

(r, t)

0
(r, t)b

0
. (170)
We call
0
(r, t) the wavefunction of the condensate.
If we now substitute the eld operators (170) in the Heisenberg equation, we nd an
equation for
0
(r, t) (do it!)
i

t

0
(r, t) =
_

2
2m

2
+ V
ext
(r)
_

0
(r, t) + U
0

0
(r, t)
0
(r, t)
0
(r, t). (171)
42
This is the Gross-Pitaevskii equation. It is an equation for the wavefunction of the con-
densate. Note that although it looks a bit like a Schrodinger equation, it is not, essentially
because it is non-linear in
0
(r, t). This latter properties gives it a tremendous mathemat-
ical richness, with its concomitant predictions of physical situations. We shall explore here
the solution of a vortex line. But before, we should see some properties of the wavefunction
of the condensate.
B. The wavefunction of the condensate and the superuid velocity.
We can imagine situations where the external eld V
ext
(r) does not only represents a static
external eld, but that it can also represent a gradient eld or even a time-dependent one.
That is, one can consider very general situations. In this case, we can set the wavefunction
in the following form,

0
(r, t) =
_
n
0
(r, t
1

V
e
i(r,t)
. (172)
That is,

0
(r, t)

0
(r, t) =
n
0
(r, t)
V
(173)
is the density of particles in the condensate, and we allow it to vary both in space and time.
The quantity (r) (sorry for the notation!) is the phase of the condensate wavefunction, a
quantity of fundamental importance.
Consider a general situation, for instance, one in which the condensate is owing. In this
case we can calculate the associated quantum mechanical current,

j(r, t) =
i
2m
(
0
(r, t)

0
(r, t)

0
(r, t)
0
(r, t))
=

m
n
0
V
(r, t) (174)
This current can interpreted as the actual current of the condensate ow and identify
with it the ow velocity, j = (n
0
/V )v
s
,
v
s
(r) =

m
(r). (175)
Since at T = 0 the uid is all superuid (and in this approximation, all the uid is in the
condensate), we identify v
s
as the superuid ow velocity. The motion is potential ow and,
therefore, irrotational
v
s
= 0. (176)
43
C. Quantized vortices.
Within the two-uid model, one can consider the situation in which the uid is set into
macroscopic rotation. In this case, naively, one could expect that only the normal part would
rotate and the superuid, because of its absence of viscosity, would not ow. This is also in
agreement that the superuid ow is irrotational (as we see below). Experiments in rotating
buckets with superuid Helium showed that this was not the case. It was found that the
whole uid behaved as it were rotating. That is, angular momentum was also transferred to
the superuid part. The explanation of this apparent contradiction is the creation of vortex
lines, which are singularities in the velocity distribution.
Suppose an arbitrary ow given v
s
(r, t). Calculate, the circulation C of the ow around
an arbitrary closed circuit (,
( =
_
C
v(r) d

l. (177)
But by Stokes theorem, this is equal to
_
C
v(r) d

l =
_
S
(v
s
(r)) d

S. (178)
By the condition v
s
= 0, the circulation ( vanishes. If the uid were set to rotate around
an axis z, integration around such an axis, will give circulation zero, and by symmetry, the
velocity v
s
should be zero, i.e. v
s
= 0, implying no ow of the superuid.
However, there is one instance in which this is not true. Consider a line in the uid where
the velocity is singular, that is, where the velocity diverges. The velocity is not dened there,
and therefore, one cannot apply Stokes theorem in that point. But we can apply the concept
of circulation. That is, while v
s
= 0 is zero everywhere, except along the line, we can
integrate the velocity around a contour that is crossed by the singularity and obtain a value
dierent from zero. That is,
_
C
v(r) d

l = 2K (179)
where K is a value to be determined but independent of of the contour, as long as it encloses
the singularity.
Let us consider a straight vortex lament centered at r = 0. The streamlines are circles
in the planes perpendicular to the vortex. Choose the contour (, as a circle at r along to
44
vortex filament
velocity flow
FIG. 5. Flow around a lament vortex line.
the velocity eld, v
s
(r) = v
s
(r)

,
_
C
v(r) d

l =
_
2
0
v
s
(r)

rd)
= r v
s
2
= 2K. (180)
That is, the velocity eld around the vortex line is,
v
s
=
K
r

. (181)
We see that it diverges as r 0 and that, except at r = 0, v
s
= 0.
On the other hand, from the quantum mechanical expression, eq.(175, the circulation
around any contour around the vortex line is,
_
C
v(r) d

l =

m
_
C
(r) d

l
=

m
(182)
where is the change of the phase around the contour. Because the wavefunction must
be single valued, this means that = 2. Therefore, K = /m, that is, the circulation
45
around the vortex is quantized with this value,
_
C
v d

l =

m
. (183)
There are many more aspects of vortices, of course. An important point is that the ele-
mentary excitations of the normal part of the uid can be scattered by the vortex laments,
thus transferring momentum and energy between the superuid and the normal part, giving
rise to viscosity and thus eventual loss of superuidity.
D. Gross-Pitaevskii solution of a vortex lament.
Let us briey present the form of the condensate wavefunction around a vortex lament.
This was originally and independently solved by Gross and Pitaevskii (1961). The Gross-
Pitaesvkii equation for the condensate wavefunction, in the absence of external elds, is
i

t

0
(r, t) =

2
2m

0
(r, t) + U
0

0
(r, t)
0
(r, t)
0
(r, t). (184)
We propose rst a stationary solution with cylindrical symmetry r = (r, , z) (r
2
= x
2
+y
2
),

0
(r, t) = e
i
0
t/
(r, ) (185)
where
(
r, ) is a function to be detemined, depending only on r and by symmetry, and
0
is the eigenenergy of one particle in the state
0
(r), the proposed wavefunction for particles
in the condensate. This energy can be calculated as the change in energy E
0
of the N-
particle ground state due to the addition of one particle, that is
0
is the chemical potential

=
E
0
N
. (186)
Using the result of Bogolubov at lowest order, eq.(149), E
0
N
2
U
0
/2V , one gets
0

NU
0
/V = U
0
. The Gross-Pitaevskii equation becomes,

2
2m

2
(r, ) + U
0
[(r, )[
2
(r, ) U
0
(r, ) = 0. (187)
The solution is proposed as,
(r, ) =

e
i
f(r/r
0
) (188)
where the phase , besides being the cylindrical polar coordinate must be the phase that
gives rise to the circulation ( = /m; this can be checked to be satised. The function f
46
must satisfy that as r f 1, such that [

(r, )(r, )[ the uniform density of


the gas. The characteristic value r
0
is given by,
r
0
=

2mU
0

. (189)
Dening the adimensional variable = r/r
0
, the Gross-Pitaevskii equation becomes an
equation for ,
1

d
d
_

df
d
_

2
f + f f
3
= 0. (190)
This is a non-linear equation that must be solved numerically, g. 6 shows a sketch of the
solution. One nds that f 0 linearly in , as 0, and f 1 as . Thus, as
r 0 ( 0), the probability of nding a particle at the vortex becomes zero.
0 1 2 3 4 5

0.5
1.0
f ()
FIG. 6. Radial wavefunction for a quantum lament vortex (sketch).
The parameter r
0
determines the size of the lament vortex,
r
0
=
1

8
1

a

1

1/3
(a
1/3
)
1/2

1

1/3
. (191)
That is, since in this approximation the gas is diluted a
3
1, the size of the lament is
much larger than the mean separation between atoms.
47

Вам также может понравиться