quantization
MingChe Chang
Department of Physics, National Taiwan Normal University, Taipei, Taiwan
(Dated: March 7, 2013)
(1)
(2)
The index can be a set of quantum numbers, e.g., = (n, l, m). Whenever required, we
will label, from low energy to high energy, = 1, 2, .
Orthogonality:
h  i = .
Completeness:
 ih  = 1.
(3)
(4)
The summation runs over all possible eigenstates, and 1 is the identity matrix in onebody
Hilbert space H(1) .
B. N particles, noninteracting
N
X
H (1) (ri );
(5)
i=1
H(r1 , r2 , ) = E(r1 , r2 , ).
(6)
(Note: For simplicity, unless necessary, we often write r as r.) Since there is no interaction
between particles, the Schrodinger equation is separable. Assume
(r1 , r2 , , rN ) = (r1 )(r2 ) (rN ),
(7)
(8)
then
(9)
with eigenvalues
E1 ,2 , = 1 + 2 + + N .
(10)
Notice that we have used a round bracket  ) to represent a product state. The angular
bracket  i is reserved for later use. Orthogonality:
Completeness:
01 ,02 , ,0N 1 ,2 , ,N = 01 1 02 2 0N N
X
1 2 N )(1 2 N  = 1.
(11)
(12)
1 2 N
The summation runs over all possible manybody eigenstates {1 2 N }, and 1 is the
(1)
identity matrix in manybody Hilbert space H(N ) = H
H(1){z
H(1)}.

N times
(13)
(14)
In order for the eigenstates to satisfy this permutation symmetry, we need to symmetrize
the state in Eq. (9).
For bosons,
1 X
P 1 (r1 )2 (r2 ) ,
B
(r
,
r
,
)
=
1
2
1 ,2 ,
N ! allP
(15)
where the summation runs over all possible permutations. Even though we have inserted a
(16)
Both summations in Eqs. (15) and (16) can be written in the following form,
1 (r1 ) 1 (r2 )
B/F
(r
,
r
,
)
=
2 (r1 ) 2 (r2 ) .
1 ,2 , 1 2
N!
..
...
(17)
For fermions, Eq. (16) is equal to the usual determinant   , in which half of the terms
have negative signs. For bosons, Eq. (15) is equal to the socalled permanent  + , in
which all of the terms have positive signs. For fermions, F is equal to zero if i = j for
any two states. This is the Pauli exclusion principle.
D. Normalization
In a bracket notation, Eqs. (15) and (16) (now free from the coordinate representation)
are written as,
1 X
1 , 2 , }
(1)P P 1 i2 i
N ! allP
1 X
(1)P P1 iP2 i ,
=
N ! allP
(18)
1 X
hP10 P1 ihP20 P2 i ,
N ! P,P 0
(19)
in which hP10 P1 i = P10 P1 ... etc. If all of the particles live in different states, then after
using
P10 = P1
(20)
P20 = P2
..
. ,
we would have N ! terms (all of them being one) in the summation. Therefore,
{1 , 2 , 1 , 2 , } = 1.
(21)
However, if there are n1 particles live in state 1, n2 particles live in state 2, ... etc (recall
that 1, 2 are the label of oneparticle eigenstate from low energy to high energy), then there
4
are n1 of the Pi states that are the same (= 1). Therefore, there are n1 ! different ways to
match (using Pi0 = Pi ) members in that n1 group, n2 ! different ways to match members
in the n2 group ... etc. As a result (pause and think),
{1 , 2 , 1 , 2 , } = n1 !n2 ! .
(22)
That is, in general this ket state is not normalized. The normalized state should be defined
as
1
1 , 2 , },
n1 ! n2 !
h1 , 2 , 1 , 2 , i = 1.
1 , 2 , i =
(23)
Fermions
Fermions do not have such a counting problem since no two particles can live in the same
state. We can identify the normalized state 1 , 2 , i as 1 , 2 , } and
h1 , 2 , 1 , 2 , i
1 X
0
(1)P +P hP10 P1 ihP20 P2 i ,
=
N ! P,P 0
(24)
= 1.
The factor (1)P
0 +P
In the formulation of second quantization, operators are written in creation and annihilation operators.
(25)
(26)
(27)
= (1)i1 1 , 2 , , , , N }.
{z}
ith
For annihilation operator a , if is the same as one of the i , but not the same as
others, then
ai 1 , 2 , , N }
(28)
= (1)i1 1 , 2 , , (no i ), N }.
For bosons, if there are multiple coincidence of i s with , then
a 1 , 2 , , N }
N
X
=
hi i 1 , 2 , , (no i ), N }.
 {z }
i=1
(29)
(30)
 n1 , n2 , , n` + 1, i.
 {z }
{z} {z}
Notice that the number of slots for  i could be infinite, if there are infinite numbers of
singleparticle eigenstates. Comparing Eqs. (27) and (30), we have
a` n1 , n2 , , n` , i =
n` + 1n1 , n2 , , n` + 1, i.
(31)
Similarly, one can have the annihilation operator that maps a N particle state to a
(N 1)particle state,
a` n1 , n2 , , n` , i n1 , n2 , , n` 1, i.
(32)
(33)
{n0 }
Xp
{n0 }
n` n1 , n2 , , n` 1, i.
The ket {n0 }i is an abbreviation of n01 , n02 , i. We have inserted a completeness relation,
and used a Hermitian conjugate operation to get the first equation.
By applying the creation operator repeatedly to the vacuum state 0, 0, i (abbreviated
as 0i), one can reach any of the manybody state n1 , n2 , i as follows,
n1 n2
1
1
n1 , n2 , i =
a1
a2
0i.
n1 ! n2 !
(34)
Fermions
The rules are simpler for fermions since the occupation number ni ( i) can only be 0 or 1.
If the state ` is empty (n` = 0), then
a` n1 , n2 , , 0, i = (1) n1 , n2 , , 1, i,
where =
P`1
i=1
(35)
(36)
One cannot add two fermions to the same state because of the exclusion principle.
On the other hand, for the annihilation operator, one has
a` n1 , n2 , , 1, i = (1) n1 , n2 , , 0, i,
a` n1 , n2 , , 0, i = 0,
where =
P`1
i=1
ni .
7
(37)
Also, a manybody state can be reached by applying the creation operator to the vacuum
state (compare with Eq. (34)),
n1 n2
0i.
n1 , n2 , i = a1
a2
(38)
F Si =
ak ak 0i.
(39)
k<kF
B. Commutation relations
Recall that
a 1 , 2 , , N } = , 1 , 2 , , N }.
Therefore,
a a 1 , 2 , , N } = , , 1 , 2 , , N },
(41)
a a 1 , 2 , , N } = , , 1 , 2 , , N }.
The two states on the RHS are identical for bosons, but differ by a sign for fermions. Since
such a connection applies to any manybody state 1 , 2 , , N }, we can remove the ket
and simply write, for bosons,
a a = a a
or
[a , a ] = 0
(42)
(43)
or
{a , a } = 0
(44)
(45)
Notice that fermion creation operators do not commute, but anticommute with each other.
Also, if = , then one has
= 0,
(46)
P`1
i=1
(48)
a a n1 , n2 , , 0, i = 0.
One can also consider the possibility of n` = 1 and write down another two equations. Both
cases lead to a a + a a = 1, the identity operator. Its not difficult to show that, if the
annihilation and creation operators act on different states, and , then a a + a a = 0.
Therefore, in general, for fermions,
{a , a } = .
(49)
Its left as an exercise to show that, for bosons, similar argument leads to
[a , a ] = .
(50)
(51)
C. Change of basis
ih
i.
(52)
(53)
1 ,
2, ,
N }.
h
ia 
X
h
ia .
(54)
X
h
ia .
(55)
(56)
{a , a} = for fermions
That is, the canonical commutation relations remain invariant under a unitary transformation.
The ket states in previous Section are representation free. We will project such ket states
to specific basis, such as coordinate basis, or momentum basis. Recall that the singleparticle
wave function is
(r) = hri.
10
(57)
We can change the basis to rbasis using the unitary transformation (Eq. (54)), then
(r) =
hria =
(r)a .
We have rewritten ar as (r). Such an operator that creates (or annihilates) a particle at
a particular point in space is called a field operator. The inverse transformation is
Z
Z
(59)
For example, the quantum state i of a particle in an empty box with volume V0 can be
(60)
This is nothing but the Fourier expansion. Its inverse transformation is,
Z
(k) =
d3 rhrki (r)
(61)
Z
1
=
d3 reikr (r).
V0
Solid state theorists prefer to move the 1/ V0 factor in Eq. (61) to Eq. (60). That is,
1 X ikr
e
ak ,
V0 k
Z
(k) =
d3 reikr (r).
(r) =
(62)
We have assumed the system is inside a box with a finite volume, so the momentum k is
quantized. If V0 approaches infinity, then one can replace the summation with an integral,
Z
1 X
d3 k
.
(63)
V0 k
(2)3
Consider another example: the hydrogen atom. The quantum numbers for a spinless
electron are = (n, l, m), and the field operator is
(r) =
nlm (r)anlm .
(64)
nlm
(65)
With the help of Eq. (59), the particlenumber operator in the coordinate representation
becomes
N =
a a
Z
d3 rd3 r0
(66)
X
hrihr0 i (r)(r0 )
d3 r (r)(r).
We have used a completeness relation to remove the summation, and used the orthogonality relation hrr0 i = (r r0 ). Naturally, one can define the particledensity operator
as
(r) = (r)(r).
(67)
This looks like the usual particle density in quantum mechanics. But beware that this is an
operator relation, not a numerical relation.
To deal with manybody systems, we need an operator A that acts on N body states
(1)
the operator for all of the particles. The latter operates in a much larger, N particle Hilbert
space H(N ) = H(1) H(1) H(1) .
(1)
The operator Ai
Ai
(68)
(1)
A ih,
(1)
(69)
(1)
N
X
(1)
Ai .
(70)
i=1
More precisely, it is
A1 i 2 i N i = A(1) 1 i 2 i N i
+ 1 i A(1) 2 i N i
(71)
+ 1 i 2 i A(1) N i .
(1)
= a
=
(72)
N
X
(1)i1 i 1 i (no i i) N i
i=1
a a 1 i
2 i N i.
Notice that the summation here should be interpreted like the series in Eq. (72). The
P
(1)
sign (1) is for bosons/fermions. In general, when A = A ih, each term in the
summation can be identified as a a using the procedure above. As a result,
X (1)
A a a .
A=
(73)
Firstly, we need to know the first quantized form of density. For a single particle, it is
(1) (r) = (r r).
(74)
Notice that r is an operator, while r is just an ordinary vector. One can easily verify that
h(1) (r)i = (r)(r). The Fourier transform of the density operator is
Z
(1)
(q) =
dveiqr (1) (r)
= eiqr .
13
(75)
We have used a simpler notation dv for d3 r. For many particles (but still in first quantized
form),
(r) =
N
X
(ri r).
(76)
i=1
eiqri .
(77)
(1)
a a .
(78)
a. Coordinate representation
(79)
Z
dv
(80)
= (r)(r)
This is the same as Eq. (67), and is similar to the usual expression (r) = (r)(r). The
difference is that (r) is a oneparticle wave function, while (r) is a field operator.
b. Momentum representation
(81)
00
The matrix element hr0 eiqr r00 i = eiqr (r0 r00 ). Also, hrki = eikr / V0 , so we get
(1)
(82)
Therefore,
X
(q) =
(1)
(83)
k0 k00
ak0 ak0 +q .
k0
p2
+ V (1) (r).
2m
(84)
N 2
X
pi
(1)
H=
+ V (ri ) .
2m
i=1
(85)
a. Energyeigenstate representation
H = .
(87)
(1)
H a a =
a a .
(88)
b. Coordinate representation
~2 2
(r r0 ) + V (1) (r)(r r0 ).
2m
15
(89)
Z
~2 2
(1)
=
dv (r)
+ V (r) (r).
2m
(90)
,
t
~2 2
(1)
+ V (r) (r, t) = i~ .
2m
t
(91)
(92)
Even though this looks exactly the same as the Schrodinger for a single particle, it is actually
an equation for the field operator. It is as if we have promoted the singleparticle wave
function to a field operator. That is why the present formulation is called the second
quantization.
c. Momentum representation
The Hamiltonian matrix in the momentum basis is (recall that hrki = eikr / V0 )
Z
~2 k 2
(1) 0
0
hkH k i =
hkk i + dvhkriV (1) (r)hrk0 i
2m
Z
~2 k 2
1
0
=
kk0 +
dvV (1) (r)ei(kk )r .
2m
V0
(93)
The second integral is actually the Fourier transform of V (1) (r): V (1) (k k0 ). The second
quantized Hamiltonian is
H =
XX
hkH (1) k0 iak ak0
k
k0
X ~2 k 2
k
2m
ak ak +
(94)
1 X X (1)
V (k k0 )ak ak0 .
V0 k k0
(95)
Some useful operators in first and second quantized form are summarized in Table 1.
16
ri
i (
2nd quantization
(r)(r)
r)
iq
ri
ie
(q)
current density j(r)
1
2m
j(q)
1
2m
pi (ri
i [
iV
(1) (
ri
V (1) (q)
V (q)
magnetic moment density m(r)
r) + (ri r)
pi ]
P iqri
ie
i
+ eiqri p
i p
ri
i i (
V (1) (q)ak+q ak
(r)(r) [ is a spinor]
r)
iq
ri
i i e
m(q)
~ P (2k + q) a ak+q
k
k
2m
iq
ri
ie
r)
ak ak+q
k k+q
B. Twobody operator
Twobody operators act on two particles at a time. A typical example is the interaction
potential V (ri rj ). In general, the operator for the whole system can be written as
A=
1 X (2)
A .
2 i6=j ij
(96)
Aij =
(2)
(97)
0 0
(2)
A0 0 i0 ihh 0 ,
0 0
in which and label quantum states of particle i; 0 and 0 label quantum states of
particle j. Notice that we have switched the subscripts , 0 of A(2) . Some textbooks prefer
not to switch them.
17
(2)
Aij 1 i N i
X
=
hi ih 0 j i1 i i 0 i N i
i<j
= a a0
(98)
X
(1)i1 (1)j2 i 0 j
i6=j
1 i (no i i) (no j i) N i
= a a0 a 0 a 1 i 2 i N i.
That is, i0 ihh 0  can be identified with a a0 a 0 a . The same is true if i > j. In
(2)
1 X (2)
A0 0 a a0 a 0 a .
2 0 0
(99)
a. Coordinate representation
(100)
Z
dvdv 0 (r) (r0 )V (2) (r r0 )(r0 )(r).
(101)
(102)
b. Momentum representation
(103)
V =
(104)
1 X (2)
0
V (q)eiq(rr ) ,
V0 q
(105)
1 X (2)
V (q)ak1 +q ak2 q ak2 ak1 .
2V0 k ,k ,q
1
(106)
One can visualize the item in the summation as follows: two incoming particles with momenta k1 and k2 are scattered by the potential and become outgoing particles with momenta
k1 + q and k2 q. The total momentum is conserved (elastic scattering), but there is a
transfer of momentum q from particle 2 to particle 1.
V. GENERAL DISCUSSION
A. Spin degree of freedom
From now on, we will focus only on systems of electrons (or fermions). Noninteracting
(or weakly interacting) electrons are referred to as electron gas. When interaction plays an
important role, we will call them as electron liquid. For spinful electrons in an empty box,
the Hamiltonian in coordinate representation is
XZ
~2 2
(1)
H =
dv s (r)
+ Vs (r) s (r)
2m
s
Z
1X
(2)
+
dvdv 0 s (r)s0 (r0 )Vss0 (r r0 )s0 (r0 )s (r),
2 s,s0
19
(107)
(1)
in which we have allowed for spindependent external potential Vs (r) and electron inter(2)
X ~2 k 2
aks aks +
1 X (1)
V (q)ak+q,s aks
V0 kqs s
2m
ks
1 X
(2)
V 0 (q)ak+q,s ak0 q,s0 ak0 s0 aks .
+
2V0 ks,k0 s0 ,q ss
(108)
The subscripts (k, s) are good quantum numbers for (noninteracting) electrons in an empty
box, which has freeparticle energy 0k = ~2 k 2 /2m. For electrons in a lattice without spinorbit coupling, the proper label for quantum states is (n, k, s), where n is the band index,
and k is the quasimomentum for Bloch electron. Also, the freeparticle energy 0k has to be
replaced by Bloch energy nk .
B. Tightbinding model
In the tightbinding model, the electrons hop from one atom to another. The operator
anks for a Bloch state has to be replaced by al , where l contains information such as band
index n, lattice site R, and spin s. The Wannier function (sometimes atomic orbitals are
used instead) is
hrli = hrn, R, si = wn (r R)s ,
where s is a spinor,
1
+ = ,
0
0
= .
1
(109)
(110)
s s0 = ss0 ,
s s = 122 .
(111)
(1)
(112)
lms
(1)
p2
+ V (1) (r)mi,
assuming V (1) is a spinindependent
where l = (n, R), and Hlm = hl 2m
potential. We have contracted the spin degree of freedom in the matrix elements.
20
1 X (2)
V
a a am0 al0 ,
2 lmm0 l0 lm m 0 l0 l m
(113)
where
(2)
Vlm m 0 l0
Z
=
dv1 dv2 wn l (r1 Rl )wn m (r2 Rm )V (2) (r1 r2 )
(114)
1
2
(115)
(2)
R1 R2 R20 R10
ss0
in which we have assumed that the onebody and twobody potentials are spinindependent.
The simplest possible model (with interaction) is to keep only the nearestneighbor hopping in the first term, and only the onsite energy in the second term. That is,
H = t
aRs aR0 s + U
<RR0 >s
= t
aR aR aR aR
(116)
aRs aR0 s + U
<RR0 >s
nR nR ,
where nRs is the occupationnumber operator. Notice that there are only two parameters:
the hopping amplitude t and the onsite energy U . This is the famous Hubbard model
proposed by Hubbard more than 50 years ago to study narrowband materials, such as
transition metal oxides. It is also considered as the underlying model for high temperature
superconductors. Despite enormous effort from numerous researchers, the phase diagrams
for the Hubbard model in dimension two and higher remain inconclusive. (see the article by
J. Quintanilla and C. Hooley in Phys. World, June, 2009)
21
C. What do we calculate
Most of the time we are interested in obtaining some of the following properties:
. Ground state energy
. Phase diagram (comparing lowest energies of different phases)
. Other ground state properties, such as charge order or spin order (correlation function),
density of states ... etc.
. Lowlying excitations (quasiparticles, collective excitations)
. Linear response function (conductivity, susceptibility ... etc)
. Other quantities of experimental interest.
Of course, whether the result is satisfactory or not depends on whether, at the first place,
the simplified model captures the essential ingredients of the phenomena we intend to study.
D. How do we calculate
The Hilbert space H(N ) of a manybody system is mind bogglingly big. In classical
mechanics, if the solution space of a particle has dimension d, then the solution space of the
whole system is N d. However, in quantum mechanics, the latter is dN . This causes major
problem for analytical and numerical calculations, and many different methods have been
proposed:
. Mean field approximation (MFA)
. Equation of motion (EOM) method
. Perturbation expansion using Greens function (diagram expansion)
. Variational method
. Density functional theory (DFT)
. Quantum Monte Carlo (QMC) method
. Density matrix renormalization group (DMRG)
22
Homework:
1. Under an unitary transformation, an orthonormalized set {i} is transformed to another orthonormalized set {
i}. Show that the anticommutation relation (for fermions)
is invariant under the unitary transformation.
{a , a } = {a , a} = .
2. Show that, if
~2 2
(1)
+ V (r) (r),
dv (r)
2m
H=
~2 2
(1)
+ V (r) (r, t) = i~ .
2m
t
3. Start from
j(1) (r) =
1
[
p(r r) + (r r)
p] ,
2m
first find out its Fourier transform j(1) (q), then show that (see Table 1)
j(q) =
~ X
(2k + q) ak ak+q .
2m k
References
[1] Chap 6 of R.P. Feynman, Statistical Mechanics, a set of lectures, AddisonWesley Publishing
Company, 1990.
[2] Chap 1 of H. Bruss and K. Flensberg, Manybody quantum theory in condensed matter physics,
Oxford University Press, 2004.
23