Quantum Chemistry II

Electronic Structure Class II

Albeiro Restrepo
June 1, 2009
Contents
1 The variational method 2
2 Probelm 1.19 from Szabo and Ostlund 4
List of Tables
List of Figures
1
1 THE VARIATIONAL METHOD 2
1 The variational method
We are about to begin the study of one major topic in quantum mechanics. Like so many
things we have seen so far in this class, without knowing it, you already have lots of
experience working with the variational method. We will bring about the formalism that
will take you away from the mere operational, sometimes unaware consumer of quantum
ideas you already are, to a (hopefully) rational, understanding individual, at least for this
particular topic. The foundations, like all truly great ides are so disarmingly simple that
I shiver when I think about the power this method brings to our hands: it will enable us
to calculate many quantities with accuracies that rival the experimental results without
settign foot in the lab, all you need is the power of your brain or to reduce all this knowl-
edge to computer code, something people have already done for us! Some derivations will
require you to exercise your math muscle, as there is a fair bit of algebra, calculus and
dierential equations involved.
This is the situation: the BornOppenheimer approximation left us with a monster, a
second order dierential equation in partial derivatives with lots of terms; the equation
can not be solved analytically because the electron coordinates can not be separated.
What is a student of nature to do? drop the case and focus on easier things? some do,
yet some dont. We will attempt approximate solutions. The pragmatic way if we do not
know the answer to a problem? we invent the solution and then see how it works. This is
exactly what the variational method does, it will let us invent solutions to the quantum
problem, but more importantly, it will give us a criterium to decide which one out of the
many (innite in principle) proposed solutions, is the best one; pretty amazing!
Lets consider a case for which the Hamiltonian has a solution space V. We now construct
|
t
, a normalized trial solution for the associated Schrodinger equation H|
t
= E
t
|
t
.
Naturally, |
t
must satisfy the conditions imposed by V, namely, it must belong to V,
therefore it can be written as a linear combination of the vectors of some basis set of V.
Please be alert, this is a tricky argument for if we know the basis vectors of V, the problem
is already solved, all we need to know is what H does to every vector in the basis; what
I am requiring here is that you write a linear combination of the vectors of an unknown
basis set, this is not too much of a problem however because the existence of the set is
not questioned, the basis set must exist since V is a vector space, we just do not know its
identity. If |
t
is normalized and belongs in V, it follows that
|
t
=

i
c
i
|i ,
t
|
t
= 1 =

i
|c
i
|
2
= 1 (1)
where {|i} is a basis set for V, the solution space of H; each of the vectors in the basis set
having an energy eigenvalue given by the individual Schrodinger equation H|i =
i
|i.
Lets label the energy eigenvalues in increasing order, that is, consider
0

1

2
.
We are now in position to state and prove the variational theorem:
1 THE VARIATIONAL METHOD 3
Theorem 1.1 The energy obtained with any trial function is an upper bound of the true
ground state energy of the system.
Proof Consider a normalized trial function |
t
for H, the Hamiltonian associated to the
system:
H|
t
= E
t
|
t
=E
t
=
t
|H|
t

now, lets use equation 1 to write the trial function as a linear combination of the vectors
of some basis:
E
t
=
_
n

i
c

i
i |
_
H
_
_
n

j
c
j
|j
_
_
=
n

i
n

j
c

i
c
j
i |H| j =
n

i
n

j
c

i
c
j
i |
j
| j
E
t
=
n

i
n

j
c

i
c
j

j
i | j =
n

i
n

j
c

i
c
j

ij
=E
t
=
n

i
|c
i
|
2

i
Lets now compare the sum of all the |c
i
|
2

i
terms with a sum where all the magnitude
of the coecients are multiplied by the same
0
, which is the smallest of all possible
eigenvalues, obviously, the rst sum is larger; lets also use the normalization condition
for the wave function as stated to the right of equation 1:
n

i
|c
i
|
2

i

n

i
|c
i
|
2

0
=E
t

0
n

i
|c
i
|
2
=E
t

0
which o course completes the proof because the energy obtained with any trial function
is larger (or equal in the best possible scenario) than the true energy of the ground state
of the system.
The variational theorem just exposed carries some very deep implications: I already
mentioned that it provides a criterium for the quality of two trial functions: the lower
the energy, the better the function
1
. Let me remind you that the proof of the variational
theorem rests in the ability of the wavefunction to be written as a linear combination of
the kets of some existant, yet unknown basis set. Some important consequences of this
fact:
1. We are not restricted to use linear combinations as trial functions, any algebraic
function that can itself be written as a linear combination or as a power series
would work. See the solved example below.
1
This is true only in the sense that it is a better function energywise; it can not be proved however
that the trial function that minimizes the variational energy is the best function for all properties.
2 PROBELM 1.19 FROM SZABO AND OSTLUND 4
2. Since one does not know the basis set (should it be known, the problem would
be solved), there is freedom not only to invent the solution but also to invent the
basis, as long as the kets of the invented basis themselves can be written as linear
combinations of the kets of a true basis.
2 Probelm 1.19 from Szabo and Ostlund
I already sent you this example, however I am reattaching it here for the sake of com-
pleteness. Information needed:
H =
1
2

1
r

2
f(r) =
1
r
2
d
dr
_
r
2
d
dr
_
f(r)
_

0
dr r
2m
e
r
2
=
(2m)!

2
2m+1
m!
m+1/2
_

0
dr r
2m+1
e
r
2
=
m!
2
m+1
| = Ne
r
2
First, lets normalize the trial function:
| = 1 = N
2
_
2
0
d
_

0
sin d
_

0
_
e
r
2
_

e
r
2
r
2
dr
_
2
0
d
_

0
sin d = 2 (cos )|

0
= 2 ([1 1]) = 4
This is the angular part of the total integral, it will reapear frequently, so lets keep in
mind. Withe the angular part evaluated, we obtain
| = 1 = 4N
2
_

0
_
e
r
2
_

e
r
2
r
2
dr = 4N
2
_

0
r
2
e
2r
2
dr
for m = 1, and the 2 factor in the exponential, this integral becomes
_

0
r
2
e
2r
2
dr =
(2m)!

2
2m+1
m! (2)
m+1/2
2 PROBELM 1.19 FROM SZABO AND OSTLUND 5
=
2!

2
3
1! (2)
3/2
=
2

8 (2)
3/2
=

4 (2)
3/2
=1 = 4N
2
_

4 (2)
3/2
_
=1 = N
2
_

2
_
3/2
=N
2
=
_
2

_
3/2
The energy for the trial function is calculated from Schrodingers equation:
= |H| =
_
2
0
d
_

0
sin d
_

0

(H) r
2
dr
= 4
_

0
_
Ne
r
2
_

_
H
_
Ne
r
2
__
r
2
dr = 4N
2
_

0
e
r
2
_
H
_
e
r
2
__
r
2
dr
= = 4N
2
_

0
e
r
2
_
H
_
e
r
2
__
r
2
dr (2)
lets evaluate the eect of the Hamiltonian on the trial function:
H
_
e
r
2
_
=
_

1
2

1
r
_
e
r
2
=
1
2
_
1
r
2
d
dr
_
r
2
d
dr
_
e
r
2
_

e
r
2
r
=
1
2
_
1
r
2
d
dr
_
r
2
2re
r
2
_
_

e
r
2
r
=
_
1
r
2
d
dr
_
r
3
e
r
2
_
_

e
r
2
r
=
_
1
r
2
_
3r
2
e
r
2
2rr
3
e
r
2
_
_

e
r
2
r
=H
_
e
r
2
_
=
_
3 2
2
r
2

1
r
_
e
r
2
Lets plug this result into equation 2:
= 4N
2
_

0
e
r
2
_
H
_
e
r
2
__
r
2
dr
= 4N
2
_

0
e
r
2
_
3 2
2
r
2

1
r
_
e
r
2
r
2
dr
= 4N
2
_
3
_

0
r
2
e
2r
2
dr 2
2
_

0
r
4
e
2r
2
dr
_

0
re
2r
2
dr
_
= 4N
2
_
3I
1
2
2
I
2
I
3
_
(3)
I
1
was actually evaluated during the normalization procedure; we found
2 PROBELM 1.19 FROM SZABO AND OSTLUND 6
I
1
=
_

0
r
2
e
2r
2
dr =

4 (2)
3/2
evaluation of I
2
:
I
2
=
_

0
r
4
e
2r
2
dr =
(2m)!

2
2m+1
m! (2)
m+1/2
, m = 2
I
2
=
4!

2
5
2! (2)
5/2
=
24

32 2 (2)
5/2
=
3

8 (2)
5/2
evaluation of I
3
:
I
3
=
_

0
re
2r
2
dr =
m!
2 (2)
m+1
, m = 0 =I
3
=
1
4
Now replace I
1
, I
2
, I
3
and N
2
into equation 3:
= 4
_
2

_
3/2
_
3
_

4 (2)
3/2
_
2
2
_
3

8 (2)
5/2
_

1
4
_
=
4
_
2

2
_

3/2

3/2
_
_
3
_
_
_

4
_
2

2
_

3/2
_
_
_
2
2
_
_
_
3

8
_
4

2
_

5/2
_
_
_

1
4
_
_
= 8
_
_

_
_

3/2
___

2
_
3
8
1/2

__

2
_
3
16
1/2

1
4
_
= 3
3
2
2
_
_

_
_

1/2
= =
3
2
2
_
_

_
_

1/2
is the energy for every wavefunction of the form | = Ne
r
2
. The variational principle
tells us that no matter what the function is, the obtained energy is always an upper bound
to the true energy of the system, lets then nd the lowest possible energy for this family
of functions by equating to zero the derivative of the energy with respect to the parameter
:

=
3
2

1

1/2
_
_

_
_
= 0 =
3
2
=
1

1/2
_
_

_
_
2 PROBELM 1.19 FROM SZABO AND OSTLUND 7
=
1/2
=
2
3
_
_

_
_
= =
8
9
the variational energy is then
=
3
2
2
_
_

_
_

1/2
=
3
2
_
8
9
_
2
_
_

_
_

8
9
=
4
3
2

16
9
2
=
4
3

8
3
= =
4
3
0.42 Hartrees
The true energy for the Hydrogen atom is -0.5 Hartrees, is obviously greather than
that. Any other value for would yield an even larger energy (try it and see for yourself
please!!).