Landau's Quasiparticle Theory

Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

Laudaus Quasi-particle Theory and Many-body Problem for Fermions

Final Report for the Course: Topics in Theoretical Materials


Student: Yi-Chun Chen
Advisor: Mei-Yin Chou
I. INTRODUCTION
The many-body problems have always been big tasks of physicists. These problems cover topics from the
solar system, such as motions of planets, to the microscopic world, like the strongly interacting fermions. No
general solution exists to the many-body problems. However, there is an great number of approximations which
explain the phenomena in many limiting cases. Here we discuss one of the great approximations: The Landaus
theory for the interacting fermions. Landau proposed the theory for liquid
3
He at very low temperatures
in 1957. Soon it was realized that this approximation could be applied to other interacting fermion systems
(Fermi Liquid). In Landaus theory, He assumed that there is a one-to-one correspondence in occupation
numbers between the Fermi liquid and ideal fermion gas at very low temperature. So the properties of Fermi
liquid greatly resemble to ideal fermion gas. Thus we could apply the quasi-particle viewpoint to describe
the interacting fermions. On the other hand, Landaus theory also stimulated the development of many-body
physics. Greens function method provides good verication of Landaus theory in microscopic viewpoint. The
perturbation expansion for Greens function leads to eective Greens function for many-body system and
eective mass and life time for quasi-particle. It has been fty years since Landau rst announced his theory.
Theories for interacting fermions at low temperature are well-established. There are many excellent textbooks:
(1) P. Nozieres, Theory of interacting Fermi systems (Bejamin, New York 1964).
(2) L.D. Landau and E.M. Lifshitz, Statistical Physics, Part 2 (Pergamon, Oxford, 1980).
(3) G. Baym and C. Pethick, Landau Fermi-liquid theory (Wiley, Ney York 1991).
(4) R.D. Mattuck, A guide to Feynman diagrams in the many-body problem (McGraw-Hill, 1976)
In this article, we introduce the Landau quasi-particle theory and Greens function viewpoint. For detail
mathematical derivations please refer (1)P. Nozieres, Theory of interacting Fermi systems.
II. LANDAUS PHENOMENOLOGICAL THEORY
Landau began from the idea of a system of quasiparticles-excitations which behave like particles, with some
modications due to interactions between particles. He started with Boltzmann equation for collision:
n
t
+
n
x

p

n
p

x
= I(n), (1)
where n is the quasi-particle number, and p are the energy and momentum of the quasi-particle, respectively,
and I is the collision term. Landau tried to construct a momentum ux by multiplying the equation be p
i
and
integrating,

p
i
nd +

p
i

n
x

p

n
p

d = 0 (2)
, where d = g[d
3
p/(2)
3
] with degeneracy g (for example, 2 for liquid
3
He). The integral of collision term is
zero because of conservation of momentum in the collisions. Continuing manipulation and integration by parts,
the equation can be written as the form

p
i
nd +

x

p
i

p
n + nd


n
x
i
d = 0 (3)
We hope to be able to write this equation as the form

p
i
nd +

ik
x
k
= 0 (4)
2
where
ik
is the momentum ux and the equation stands for the conservation of total momentum. This could
happen only if the /x
i
of the third term in the right hand side of (3) be taken outside the integral. That is
to say, if we have
E =

(p)n(p)d (5)
then


n
x
i
d =
E
x
i
(6)
Hence we get the form (4). So Landau made the (5) as an assumption that, or more generally,
E/n(p) = (p). (7)
That is to say, the quasi-particle energy (p) is obtained by varying the energy with respect to quasi-particle
number. We can see that this assumption was led by the assumption of conservation of quasi-particle momentum.
Now the energy E is a functional of the occupation number n(p) of all of the quasi-particles, i,e,
E = E (n(p
1
), n(p
2
), n(p
3
), ...) . (8)
How does this functional form of energy work? Here we introduce Landaus another assumption: the weakly
excited states of a Fermi liquid greatly resemble those of a weakly excited ideal Fermi gas. So there is a one-
to-one correspondence between them in the number of states. This may be physically realized by adiabatically
switching on the two-particle interaction. When we start from an ideal gas system and switch on the interaction,
there will be some variation of occupation number n(p) as FIG.1.
Figure1: Variation of Occupation Number
Note that variation happen near the Fermi surface. After variation, i.e., switching on the interaction, the
occupation number changes and energy also change,
E

= E (n

(p
1
), n

(p
2
), n

(p
3
), ...) (9)
and we can expand the dierence in energyE = E

E to the second order of n,


E =

0
(p)n(p) +
1
2

p,p

f(p, p

)n(p)n(p

) (10)
, where the second derivative coecient f(p, p

) is dened as the interaction between quasi-particles. The energy


of quasiparticle can be also written as
(p) = E/n(p) =
0
(p) +

f(p, p

)n(p

) (11)
In Landaus theory, we want to describe the collections of interacting fermions as independent quasi-particles.
So the eect of interactions must be absorbed somewhere. So now we are going to explore the relation between
quasi-particle eective mass m

and interaction f(p, p

). Based on the assumption of Galilean invariance, we


consider the Fermi surface is displaced as FIG.2
3
Figure2: Displacement of Fermi Surface
Then the energy per unit volume is increased by
E = q
2
/2m, (12)
where is the density. On the other hand, E can be computed through Eq(10). First we replace the summation
by the integral, then the rst order of variation can be written as

0
p
n(p) =
1
(2)
2

1
1
dcos()p
2
F

qcos()
0
(p
F
+ p)
2
2m

dp = q
2
/2m

,
(13)
where the in the integral is shown in FIG.2. The second order of variation can be rewritten as
1
2
p
4
F
(2)
4

1
1
d cos()

1
1
d cos(

q cos()
0
dp

q cos(

)
0
dpf(p, p

),
(14)
with |p| = |p

| = p
F
. We further expand the f(p, p

) in terms of spherical functions, i.e,


f(q, q

) =

i
f
i
P
i
(cos()), (15)
and integral the P
1
component. Finally, we get
q
2
/2m = q
2
/2m

+ 1/2(p
F
/6
2
)f
1
q
2
(16)
Dene F
1
= (p
F
m

/2
2
)f
1
, we have
m

= m(1 + F
1
/3) (17)
In the theory, variations happen near the Fermi surface, i.e., Fermi sphere for ideal gas, so we only consider
the f
1
component of interaction. With Eq(17), we can describe the Fermi liquid as particle-like system. For
example, the specic heat
c
v
=
m

m
c
0
v
, (18)
where c
0
v
is the specic heat for ideal Fermi gas with bare particle mass m. So the specic heat of interacting
fermions at very low temperature will be proportional to temperature T with some modication in mass. For
instance, the specic heat of
3
He. FIG.3 shows this proportionality.
Figure3: Specic Heat of
3
Helium [Greywall(1983)]
4
III. GREENS FUNCTION VERIFICATION
Quasi-particle method is a phenomenological theory, discussing the variation of occupation numbers near the
Fermi surface. Greens function approach provides us microscopic verication to this phenomenological theory.
For details in the verication, such as the rederivation of Eq(17) through Greens function, please refer (1) P.
Nozieres, Theory of Interacting Fermi Systems or (2) E. N. Economou, Greens Functions in Quantum Physics,
Berlin, 1983. Here we just give some examples and discuss the physics meanings of quasi-particle and Greens
function method.
For free Fermion particles, they satisfy the Paulis exclusion principle and Hamiltonian H
0
=
2
/2m. The
occupation in k-space is shown as FIG.4(a). When the system is excited, particles will be removed from states of
low energy, lower than Fermi energy, to states of higher energy, higher than Fermi energy. Then the occupation
will behave as FIG.4(b). However, it is inconvenient to draw all particles when most of them werent transferred
in forming the excited states. So we can treat the occupation in the FIG.4(a) as ground state and just record
the changes from it. To draw this, we remove the lled Fermi sea from the FIG.4(a) and (b), yielding FIG.4(c)
and (d). This is called particle-hole description. Note that the Fermi sea is physically still present but only
removed from the drawing of the system.
Figure4: Occupation in k space from ordinary picture and particle-hole picture
In the particle-hole picture, if a hole occupy the state
k
, it means that a particle of energy
k
is removed
from the system. Hence the hole energy is negative and we have

hole
k
=
k
. (19)
Thus the time-dependent wave function for a hole in state
k
is
(t)
hole
=
k
e
i(
k
)t
=
k
e
i
k
(t)
(20)
where
k
<
F
. The second equal in Eq(20) shows that the hole may be viewed as a particle moving backward
in time. This backward in time viewpoint helps us to draw Feynman diagram for interactions. For instance,
consider there an extra particle, with energy
k
>
F
, shot into the non-interacting fermions system but under
potential V (r). The component of V (r) in K space are
V
kl
=

k
(r)V (r, p)
l
(r)d
3
r, (21)
Then there are four possibilities shown in FIG.5 mean: (a) scattering of a particle from state
l
to
k
, (b) the
potential scatters a particle out of state
l
, where
l
<
F
(note the direction of l component in the gure.b
) into state
k
, where
k
>
F
, etc. Here the backward in time concept helps us to visualize the interaction
acting on holes.
5
Figure5: Four Possibilities for External Potential in Diagrams
If the nal momentum of the extra particle is label by k > k
F
, and the potential is, in special case, such that
V
km
and V
mk
(
m
<
F
) are large, then the Greens function for this particle can be computed through:
or
G
+
(k, ) =
1
[G
+
(k, )]
1
V
km
V
mk
G

0
(m, )
=
1
(
k
+ i)
|V
km
|
2
(
m
i)
, (22)
where G
+
0
(k, ) is the Greens function for state
k
after Fourier transform, and equals to 1/(
k
+ i). On
the other hand, G

0
(k, ) stands for the Greens function of hole and equals to 1/( +
k
+ i). We get the
eective Greens function by Eq(22). By nding the poles of Eq(22), we get the eective energy for this extra
particle
=

k
+
m
2

1
2

(
k

m
)
2
+ 4|V
km
|
2
(23)
That is how Greens function method works for the Fermi system. Including the interactions,
real particle + coat or cloud of interactions = quasi particle (24)
and the perturbation series in diagrams help us to calculate the second term in Eq(24). Back to our topic: the
interacting Fermions. Consider the interaction V (|r
i
r
j
|), and its component V
klmn
are dened as
V
klmn
=

d
3
r

d
3
r

k
(r)

l
(r

)V (|r r

|)
m
(r)
n
(r

) = V
lknm
(25)
or
For Hartree-Fock approximation, the interaction components could be approximated as
V
klmn
=
mk

nl
V
klkl
+
ml

nk
V
kllk
+ small terms (26)
that is to say, two kinds of interactions(a)(b) in the FIG.6 are important
6
Figure6: Diagram for Hartree-Fock Approximation
,and they correspond to
[iG
+
0
(k, )]
2

l<k
F
[
i
2
V
klkl
] and [iG
+
0
(k, )]
2

l<k
F
[iV
klkl
], (27)
respectively. Note that there is a restriction l < k
F
in the summation. This could be realized as at time t
1
in
the FIG.6, the interaction kicks out a particle for existed state
l
to create a particle-hole pair. So the state
l
must be contained in Fermi surface. Finally, we can get the eective Greens function by partial sum
Figure7: Partial Sum for Hartree-Fock Approximation
Translate the diagram into our language
G
+
(k, ) =
1

k

l<k
F
(V
klkl
V
kllk
) + i
(28)
This leads to quasi particle with energy

k
and life time
k

k
=
k
+

l<k
F
(V
klkl
V
kllk
) and
k
= . (29)
The result is same as what we derived in the class. As we saw in previous examples, Greens function derivation
is useful for variation near the Fermi surface. We could avoid the diculties to deal with particles which are
far below the Fermi level, and use Greens function to describe the holes and particles. This viewpoint meets
Landaus theory. For example, the particle-hole picture in Green function corresponds to the number of state
variation in Landaus theory.
There are also other verications for Landaus theory. For example, the renormalization group theory, which
concludes that the second term in the R.H.S. of Eq(10) are marginal and describes Laudaus theory as a xed
point theory in microscopic viewpoint. For further details, see R. Shankar, Renormalization-group approach
to interacting fermions, Rev. Mod. Phys. 66, 129192 (1994).

You might also like