EE5 in 2008 Lecture notes

Univ. of Iceland Hannes Jonsson

IIa. Systems of many electrons
Exchange of particle labels
Consider a system of n identical particles. By identical particles we mean that all
intrinsic properties of the particles are the same, such as mass, spin, charge, etc.
Figure II.1 A pairwise permutation of the labels on identical particles.
In classical mechanics we can in principle follow the trajectory of each individual
particle. They are, therefore, distinguishable even though they are identical. In quantum
mechanics the particles are not distinguishable if they are close enough or if they interact
strongly enough. The quantum mechanical wave function must reect this fact.
When the particles are indistinguishable, the act of labeling the particles is an arbi-
trary operation without physical signicance. Therefore, all observables must be unaected
by interchange of particle labels. The operators are said to be symmetric under interchange
of labels. The Hamiltonian, for example, is symmetric, since the intrinsic properties of all
the particles are the same. Let (1, 2, . . . , n) be a solution to the Schrodinger equation,
(Here n represents all the coordinates, both spatial and spin, of the particle labeled with
n). Let P
ij
be an operator that permutes (or exchanges) the labels i and j:
P
ij
(1, 2, . . . , i, . . . , j, . . . , n) = (1, 2, . . . , j, . . . , i, . . . , n).
This means that the function P
ij
depends on the coordinates of particle j in the same way
that depends on the coordinates of particle i. Since H is symmetric under interchange
of labels:
H(P
ij
) = P
ij
H, i.e., [P
ij
, H] = 0.
Therefore, P
ij
is also a solution of the Schrodinger equation with the same eigenvalue as
.
Exchange Degeneracy: There are n! dierent permutations of n lables and all the n!
wave functions have the same energy. Some linear combination of the n! functions gives the
proper description of the system. Since [P
ij
, H] = 0, i.e. P
ij
is a constant of the motion,
21
the linear combination that describes the system initially, is the proper linear combination
for all time.
The density (i.e. probability distribution) must be unaected by P
|P
ij
|
2
= ||
2
but that is not the case with the wavefunction, which is not an observable. Therefore, the
functions and P
ij
can dier in phase
P
ij
= e
i

where is a real number. Applying P

ij
again will undo the permutation, so we must have
P
ij
(P
ij
) = P
2
ij
=
that is
(e
i
)
2
= 1.
There are only two distinct solutions = 0 and = corresponding to
P
ij
=
_
+ symmetric under interchange
antisymmetric under interchange .
It turns out that both solutions are found in nature. For some particles, called Fermions,
the wavefunction should be antisymmetric ( = ). Examples are electrons, protons
and neutrons. For other particles, called Bosons, the wavefunction should be symmetric
( = 0). Examples are photons (particles with integral spin).
The behavior of the wavefunction when composite particles (not elementary), such as
atoms, are interchanged can be deduced by counting the number of Fermions. For example:
When two
4
He atoms are interchanged
2 Electrons: Pick up factor (1)(1) = +1
2 Protons: Pick up factor (1)(1) = +1
2 Neutrons: Pick up factor (1)(1) = +1 .
So the wavefunction should not change sign, i.e.
4
He atoms behave as Bosons.
When two
3
He atoms are interchanged:
2 Electrons: Pick up factor (-1)(-1) = +1
2 Protons: Pick up factor (-1)(-1) = +1
1 Neutron: Pick up factor (-1) .
So the wavefunction should cange sign,
3
He atoms are Fermions.
Since hydrogen atoms are very light, one often needs to use a quantum mechanical
description of the motion of the nuclei as well as the elecrons. The wave function changes
sign if we interchange labels of two electrons (electrons are Fermions) and it changes sign
if we interchange the lables of two nuclei (the nuclei consist of one proton which is a
22
Fermion), but if we interchange two H atoms (both electron and proton), then we pick up
a factor of (1)(1) = +1, i.e., the wavefunction does not change sign. Hydrogen atoms
behave as Bosons.
When a stationary state wave function is being constructed for a system, for example
to evaluate possible values of the energy, it is not enough to just solve for eigenfunctions
of the Hamiltonian operator, one also needs to build in the right symmetry with respect to
the exchange of labels on the particles. For example, if a given a function (1, 2, . . . , n) is
found to satisfy the Schrodinger equation but does not have the right exchange symmetry,
it is necessary to construct a linear combination of permutations of the labels to obtain
a function that is symmetric under the exchange of labels when dealing with Bosons or
antisymmetric when dealing with Fermions. Let P be any one of the n! permutations. It
can be constructed from a sequence of pairwise permutations. Let the number of pairwise
permutations required be r
P
. For example:
(1, 2, 3)
(1, 3, 2) r = 1 odd
P
13
(2, 3, 1) r = 2 even
P
12
(2, 1, 3)
_
r = 3
or r = 1
odd
The normalized symmetric linear combination appropriate for Bosons is

+
=
1

n!

p
P
p
(1, 2, . . . , n) .
The index p runs over all the n! permutations. Similarly, the normalized antisymmetric
linear combination appropriate for Fermions is:

=
1

n!

p
(1)
r
p
P
p
(1, 2, . . . , n) .
The dierence in the energy that is obtained from and that obtained from

is called
the exchange energy. That is
E
exch
= <

|H|

> < |H| > .

We can dene operators that perform the required linear combination. The sym-
metrizer, S, is
S
1

n!

P
23
and the antisymmetrizer, A, is
A
1

n!

(1)
r
p
P .
Then
+
= S and

= A.
Non-Interacting Electrons: (the simplest system containing several electrons)
The total Hamiltonian of the system is a sum of terms, each acting only on a single
electron:
H = H
1
+H
2
+. . . H
n
.
Since the Hamiltonian separates, a solution to the Schrodinger equation can be found that
is a product of one electron functions:
(1, 2, . . . , n) =
a
(1)
b
(2) . . .
0
(n) .
The one electron functions are called spin-orbitals and depend both on spin and spatial
coordinates and the product function, (1, 2, . . . , n), is called a Hartree Product. The
properly antisymmetrized many electron wavefunction is:

(1, 2, . . . , n) = A (1, 2, . . . , n)
=
1

n!

a
(1)
b
(1) . . .
0
(1)

a
(2)
b
(2) . . .
0
(2)
.
.
.
.
.
.
.
.
.
.
.
.

a
(n)
b
(n) . . .
0
(n)

.
Applying some general rules about determinants, we can see that this wavefunction has
the required properties: When the determinant is expanded out there are n! terms half
of which have a minus sign. A permutation P
ij
corresponds to interchanging the rows i
and j, which causes the determinant to change sign. Therefore the antisymmetry is built
in. We will frequently be dealing with determinantal wave functions in this section. It is
therefore essential to have a convenient short hand notation for such functions. We will
use the notation

(1, 2, . . . , n) = |
a
(1)
b
(2) . . .
0
(n) > .
The pointed bracket will therefore have a dierent meaning in this section than it had
previously.
If two spin orbitals are equal, for example
a
=
b
then two columns are the same and
the determinant vanishes, = 0. That is, two electrons cannot be in the same spin-orbital.
This is referred to as Pauli exclusion.
Assuming the hamiltonian does not couple spatial coordinates and spin, the spin-
orbitals can be written as a product of spatial and spin functions. For example, if electron
1 is in spatial orbital a and has spin up, the spin-orbital is

a
(1) =
I
(r
1
)(
1
) .
24
Since we have used the symbols of the ket notation to mean a determinantal wave function
in this section, we will be explicitly dealing with functions rather than kets. To deal with
the spin, we therefore imagine having functions () for spin up and () for spin down
and introduce, purely for convenience, a hypothetical variable, . Eventually, we will
always integrate over expressions involving the functions and and the important thing
to remember is that they are orthonormal
_
d

() () = 0
and
_
d

() () = 1.
The Hartree product is truly an independent electron wavefunction. But the antisym-
metrized wavefunction obtained by constructing the Slater determinant has introduced
some, limited correlation between any pair of lelectrons that have parallel spin. This is
called the exchange correlation (not to be confused with the correlation energy which is
the dierence between the Hartree-Fock estimate of the energy and the exact value of the
energy). The eect can be seen by looking at the electron density in conguration space.
Consider a two electron Slater determinant:

(1, 2) = A
a
(1)
b
(2) = |
a
(1)
b
(2) > .
Example a. First, assume the two electrons have opposite spins and occupy dierent
spatial orbitals

a
(1) =
I
(r
1
)(
1
)
and

b
(2) =
II
(r
2
) (
2
) .
( means spin up and spin down). By expanding the determinant, we get:

(1, 2) =
1
2
(
I
(r
1
)(
1
)
II
(r
2
)(
2
)
I
(r
2
)(
2
)
II
(r
1
)(
1
)) .
The simultaneous probability of nding one of the two electron near a point r
s
and the
other near a point r
s
(that is, one of the electrons is in an interval between r
s
and r
s
+dr
and the other electron in an interval between r
t
and r
t
+dr is obtained from the magnitude
squared of the wave function |(1, 2)|
2
by integrating out the spin variables:
P(r
1
, r
2
)dr
1
dr
2
=
_
d
1
_
d
2
|(1, 2)|
2
dr
1
dr
2
=
1
2
_
|
I
(r
1
)|
2
|
II
(r
2
)|
2
+|
I
(r
2
)|
2
|
II
(r
1
)|
2
_
dr
1
dr
2
.
By setting r
1
= r
s
and r
2
= r
t
, the rst term is the product of the probability of nding
electron 1 within r
s
and (r
s
+dr) times the probability of nding electron 2 within r
t
and
25
(r
t
+dr) as electron 1 occupies orbital
I
and electron 2 occupies orbital
II
. The second
term has electron 2 occupying
I
and electron 1 occupying
II
. Since the electrons are
indistinguishable, the correct probability is the average of the two terms. Therefore, two
electrons with opposite spin are fully uncorrelated at this level of theory.
Example b. Secondly, assume the two electrons have the same spin, say spin down. Then

a
(1) =
I
(r
1
)(
1
)

b
(2) =
II
(r
2
)(
2
) .
Then, the probability distribution becomes:
P(r
1
, r
2
) =
1
2
_
|
I
(r
1
)|
2
|
II
(r
2
)|
2
+|
I
(r
2
)|
2
|
II
(r
1
)|
2
2Re
_

I
(r
1
)
II
(r
1
)

II
(r
2
)
I
(r
2
)
__
.
The extra term that has appeared as compared with example a introduces correlation in
the distribution of the two electrons. In particular P(r
1
= r, r
2
= r) = 0, i.e., two electrons
with parallel spin cannot be found at the same point in space. A Fermi hole is said to exist
around each electron, a region from which other electrons of the same spin are excluded.
The presence of the Fermi hole is an important consideration. It results, for example, in a
substantial dierence in the accuracy of the Hartree-Fock approximation for singlet states
and for triplet states, since the presence of the extra Fermi hole in the triplet tends to keep
the electrons apart and, thereby, lower the Coulomb repulsion between them.
Calculation of Matrix Elements using Slater Determinants:
Let (1, 2, 3, . . . , N) be a normalized Slater determinant
(1, 2, . . . , N) =
1

N!

i
(1)
j
(1) . . .
k
(1)

i
(2)
j
(2) . . .
k
(2)
.
.
.
.
.
.
.
.
.
.
.
.

i
(N)
j
(N) . . .
k
(N)

where the
i
are spin-orbitals. We will use the shorthand notation
(1, 2, . . . , N) = |
i

j
. . .
k
>
where it is understood that the electron labels are in the order 1, 2, . . . , N. Because of the
antisymmetry we have, for example:
| . . .
m
. . .
n
. . . >= | . . .
n
. . .
m
. . . > .
Various calculations in quantum mechanics involve the evaluation of matrix elements of an
operator with the wavefunction. It, therefore, becomes important to develop expressions
for matrix elements of various types of operators when the wavefunction is of the form of
26
a Slater determinant. Given an operator O and two N-electron determinants |K > and
|L >, the problem now is to evaluate the matrix element < K|O|L >. We need to obtain
an expression involving integrals over the individual spin-orbitals. In particular, we need
to evaluate the energy, < H >=< K|H|L >.
There are two types of operators that are relevant:
One electron operators are of the form
O
1
=
N

i=1
h(i)
where h(i) only involves the i th electron and the summation index, i, runs over all
electrons. For example
h(i) =
h
2
2m

2
i
.
Two electron operators are of the form
O
2
=
N

i=1
N

j>i
v(i, j)

i<j
v(i, j)
where v(i, j) is an operator that depends on the coordinates of both electrons, i and j. An
important example is the Coulomb interaction e
2
/r
ij
and we will be using that example
in the following discussion.
Regarding the determinants |K > and |L >, there are three dierent situations:
Case 1: The two determinants are equal, i.e.,
|L >= |K >= | . . .
m

n
. . . > .
Case 2: The two determinants dier by one spin-orbital: For example, replacing
m
by

p
in |L >
|K >= | . . .
m

n
. . . >
|L >= | . . .
p

n
. . . > .
Case 3: The two determinants dier by two spin-orbitals: For example, replacing
m
by

p
and
n
by
q
in |L >
|K >= | . . .
m

n
. . . >
|L >= | . . .
p

q
. . . > .
It can easily be shown that the matrix elements of O
1
are zero if |K > and |L > dier
by more than one spin-orbital, and that the matrix elements of O
2
are zero if |K > and
|L > dier by more than two spin-orbitals.
It is essential to use some convenient shorthand notation for the various integrals.
Unfortunately, there are two dierent notations in common use, that are sometimes referred
27
to as chemists notation and physicists notation. We will use the former and denote
integrals over spin-orbitals with square brackets, [ ]. The denitions are as follows:
[i|h|j]
_
dx
1

i
(1)h(r
1
)
j
(1)
and
[ij|k]
_
dx
1
_
dx
2

i
(1)
j
(1)
1
r
12

k
(2)

(2)
The integration variable x
i
denotes both the spatial and spin coordinates of electron i
(note, we assume the Hamiltonian does not depend on spin and therefore denote the
variable there as r
1
).
The following table gives the matrix elements of one-electron operators for determinant
wave functions in terms of the integrals over spin-orbitals.
Table 1 : Matrix elements of one electron operators of the form O
1
=
N

i=1
h(i)
Case 1: |K >= | . . . mn. . . >
< K|O
1
|K >=
N

m
[m|h|m]
Case 2 : |K >= | . . . mn. . . >
|L >= | . . . pn. . . >
< K|O
1
|L >= [m|h|p]
Case 3 : |K >= | . . . mn. . . >
|L >= | . . . pq . . . >
< K|O
1
|L >= 0
28
The following table gives the matrix elements of the two-electron operator representing
Coulomb interaction of N electrons in terms of the spin-orbitals of the determinants.
Table 2 : Matrix elements of two electron operators of the form O
2
=
N

i=1
N

j>i
r
1
ij
Case 1: |K >= | . . . nm. . . >
< K|O
2
|K >=
1
2
N

m
N

n
[mm|nn] [mn|nm]
Case 2 : |K >= | . . . mn. . . >
|L >= | . . . pn. . . >
< K|O
2
|L >=
N

n
[mp|nn] [mn|np]
Case 3 : |K >= | . . . mn. . . >
|L >= | . . . pq . . . >
< K|O
2
|L >= [mp|nq] [mq|np]
To illustrate how the rules in the table come about, calculations of some of the matrix
elements are presented below for some of the possible choices of Slater determinants.
Example 1: The simplest case is when the same Slater determinant appears on the left
and right hand side of the matrix element (case 1)
| >= |
1

2
>
1.a For a one-electron operator:
< |h(1)| > =
_
dx
1
_
dx
2
1

2
_

1
(x
1
)
2
(x
2
)
2
(x
1
)
1
(x
2
)
_

h(r
1
)
1

2
_

1
(x
1
)
2
(x
2
)
2
(x
1
)
1
(x
2
)
_
=
1
2
_
dx
1
_
dx
2
_

1
(x
1
)

2
(x
2
)h(r
1
)
1
(x
1
)
2
(x
2
)
+

2
(x
1
)

1
(x
2
)h(r
1
)
2
(x
1
)
1
(x
2
)

1
(x
1
)

2
(x
2
)h(r
1
)
2
(x
1
)
1
(x
2
)

2
(x
1
)

1
(x
2
)h(r
1
)
1
(x
1
)
2
(x
2
)
_
.
29
Since the operator h(1) does not involve coordinates of electron 2 we can easily carry out
the x
2
integration. In the rst two terms, x
2
only appears in the same spin-orbital within
each term. Therefore using the fact that both
1
and
2
are normalized, the integration
over x
2
gives 1 in the rst two terms. However, in the last two terms, x
2
appears in both
1
and
2
within each term. The integration over x
2
then gives zero because the spin-orbitals
are orthogonal. We are left with
< |h(1)| > =
1
2
_
dx
1

1
(x
1
)h(r
1
)
1
(x
1
)
+
1
2
_
dx
1

2
(x
1
)h(r
1
)

1
(x
2
)
=
1
2
2

m
[m|h|m].
Similarly
< |h(2)| >=
1
2
2

m
[m|h|m].
Adding the two gives case 1 in the table for the one-electron operator, O
1
.
1.b For the two-electron operator O
2
= 1/r
12
:
< |
1
r
12
| > =
1
2
_
dx
1
_
dx
2
_

1
(x
1
)

2
(x
2
)
1
r
12

1
(x
1
)
2
(x
2
)
+

2
(x
1
)

1
(x
2
)
1
r
12

2
(x
1
)
1
(x
2
)

1
(x
1
)

2
(x
2
)
1
r
12

2
(x
1
)
1
(x
2
)

2
(x
1
)

1
(x
2
)
1
r
12

1
(x
1
)
2
(x
2
)
_
Since 1/r
12
= 1/r
21
, the indices can be interchanged on the dummy integration variables.
Therefore, the rst term is the same as the second and the third term is the same as the
fourth.
< |
1
r
12
| > = [11|22] [12|21].
This illustrates case 1 in the table for the two-electron operator, O
2
.
Example 2: Now use determinants that dier by one spin-orbital (case 3). Again, the
Slater determinants are constructed by using two spin-orbitals but dierent spin-orbitals
are chosen for the two determinants:
|
A
>= |
1

2
>
30
and
|
B
>= |
3

4
> .
We clearly have
<
A
|h(1)|
B
>= 0
because the x
2
integration gives zero in all terms,
0 =
_
dx
2

i
(x
2
)
j
(x
2
) when i = j.
This illustrates case 3 in the table for the one electron operator, O
1
.
The two-electron matrix element is, by denition,
<
A
|
1
r
12
|
B
>= [13|24] [14|23].
The energy of a determinantal wavefunction:
Using the tables, we can readily nd the expectation value of the total energy in a
single determinant state |K > of N electrons
< H > =< K|H|K >=< K|O
1
+ O
2
|K >
=
N

m
[m|h|m] +
1
2
N

m
N

n
[mm|nn] [mn|nm]
The sums run over all the occupied spin-orbitals. The one electron operator h inclues the
kinetic energy of the electron and the interaction with all the nuclei
h(i) =
1
2

2
i

A
Z
A
r
iA
The summation index A runs over all nuclei that electron i interacts with. Here the electron
charge has been set to one, e = 1, and Z
A
is the nuclear charge in units of the electron
charge, e. Each occupied spin-orbital
i
in |K > contributes [i|h|i] to the energy and every
unique pair of occupied spin-orbitals
i
,
j
contributes < ij||ij >. Note that
1
2
N

m
N

n
[mm|nn] [mn|nm] =

n>m
[mm|nn] [mn|nm]
since
[mn|nm] = [nm|mn].
and terms with n = m cancel out.
31