Simple Variation Method for the Hydrogen Molecule

1. Calculation of the energy levels of wavefunctions

In our treatment we will be required to calculate the energy associated with an electronic
wavefunction. This will be introduced first.

The Schroedinger equation in time-independent form is

^
H Ψ = EΨ [1]

(Ĥ…Hamilton operator, E…energy eigenvalue)

which upon multiplication from the left with the conjugate complex of Ψ

^
Ψ * H Ψ = Ψ * EΨ [2]

and integration over all space

∫ Ψ H Ψdτ = ∫ Ψ EΨdτ = E ∫ Ψ Ψdτ

* * *
[3]

allows the calculation of the energy level E associated with the wavefunction Ψ

∫ Ψ H Ψdτ
*

E= [4]
∫ Ψ Ψ dτ
*

If we know the exact form of the wavefunction Ψ, we can use the relation

∫ Ψ Ψ dτ = 1
*
[5]

since Ψ*Ψ corresponds to the probability of finding the electron wave at a certain point in
space and the probability of finding it anywhere, expressed as the integral over all space,
must be 1.

Equation [4] reduces then to

^
E = ∫ Ψ * H Ψdτ [6]

which is termed the expectation value of the Hamiltonian. We will use this relation in the
lecture and in the problem sheets to derive the energy levels of simple molecular orbitals
(MOs). For our further treatment here we will use equation [4].
2. Definition of the wavefunction of the MO

The simplest approach is to construct the wavefunction of the molecular orbitals for H2
from the atomic wavefunctions of the hydrogen atoms. As with all theories we have to
compare our results with experimentally determined values. If the model predicts these
quantities in a satisfactory manner we accept it as a good representation of reality. If it
does not we have to refine the model or, if it can not be refined in a meaningful way,
abandon it. Although we will see that the simple treatment does not give satisfactory
results it is stated here that a satisfactory refinement within the LCAO (linear
combination of atomic orbitals) framework is possible. This would, however, go beyond
the intentions of this course and therefore we will not explore this.

We define the wavefunction of the molecular orbital as a linear combination of the two 1s
orbitals of the hydrogen atoms with the protons a and b at the respective centres:

ψ = c1ψ 1sa + c2ψ 2 sb [7]

c1 and c2 are the coefficients we need to determine and ψ1sa and ψ1sb are the hydrogen 1s
wavefunctions (the exact form of ψ1sa and ψ1sb is obtained by solving Schroedinger’s
equation for the hydrogen atom which can be done analytically – you will have seen the
solutions in the properties of matter course). Since the hydrogen molecule is symmetric
one would already expect here that the shape of the MOs have to be symmetric and that
the coefficients c1 and c2 should have equal values.

In principle, we have an infinite number of combinations of c1 and c2 to chose from. Each

choice will result in a wavefunction with an associated energy eigenvalue. The specific
solution we are looking for must fulfill certain requirements. Most importantly it should
accurately predict measurable quantities, such as bond length and bond energy. Before
reading on, ask yourself what requirements the solution has to fulfill regarding the energy
values associated with it?

Let us assume for the time being that we can “guess” the correct wavefunction by picking
the right values for c1 and c2. The energy level of the wavefunction should then
correspond to the bonding energy of the hydrogen molecule and assume a minimum
value. After all, hydrogen is a fairly stable molecule and it exists – if a more stable state
were to exist, we would not be able to observe any hydrogen.
Therefore, when we carry out our “guess” for c1 and c2 we have to aim for a minimum in
the resulting energy level of the wavefunction. This is the best we can do without
introducing more components (e.g., higher lying orbitals) into our trial wavefunction
given in [7]. Any other combination of c1 and c2 that does not lead to a minimum in
energy must be worse and therefore be dismissed. This is the whole principle of the
variation method.

When we look for appropriate values for the coefficients c1 and c2 we will therefore look
for those solutions that lead to a minimum in the associated energy eigenvalue. How are
we going to describe this in a mathematical fashion?
3. Finding the minimum in energy associated with the wavefunction

A minimum in energy is obtained when the first derivative of the energy expression
equals zero. Hence the following conditions must be fulfilled

∂E ∂E
= 0 and =0
∂c1 ∂c 2

To simplify the following treatment we introduce these relations

ψ 1sa ≡ ψ a ψ 1sb ≡ ψ b
^ ^
H aa = ∫ψ a Hψ a dτ H bb = ∫ψ b Hψ b dτ
* *

^ ^
H ab = ∫ψ a Hψ b dτ H ba = ∫ψ a Hψ b dτ
* *

S = ∫ ψ a ψ b dτ
*

We further realise that, since ψa = ψb (both are hydrogen 1s wavefunctions)

Haa = Hbb and Hab = Hba

This will simplify the analytical treatment of the solution. In the case where different
wavefunctions are to be combined (e.g., 1s with a 2s or 2p wavefunction), the solution
strategy will be the same. The arising equations may, however, not be as easily solvable
requiring more sophisticated mathematical treatment and we will not embark on this here.
In the lectures we will see that we can still make use of such combinations in an
illustrative way and derive meaningful results that help us understand bonding in
molecules other than H2.

Let us first express the energy eigenvalues of the MOs by substituting the MO
wavefunction defined in [7] into [4]

∫ (c1ψ a + c2ψ b ) H(c1ψ a + c2ψ b )dτ

* *

E=
∫ (c ψ
*
1 a + c 2ψ b* )(c1ψ a + c2ψ b )dτ
[8]

now we make use of the fact that the Hamilton operator is a linear operator meaning that

Ĥ(c·Ψ) = c·ĤΨ (c…constant) and

Ĥ(Ψ1 + Ψ2) = ĤΨ1 + ĤΨ2
With this we can evaluate [8] (try to do it!)

c12 H aa + 2c1c 2 H ab + c 22 H bb
E=
c12 + 2c1c2 S + c 22
[9]

We need to form the partial derivatives of [9] with respect to c1 and c2. Let

N = c12 H aa + 2c1c 2 H ab + c 22 H bb D = c12 + 2c1c2 S + c 22

∂N ∂D
D −N
∂E ∂c1 ∂c1 1 ∂N E ∂D 2c1 H aa + 2c 2 H ab E (2c1 + 2c 2 S )
= 2
= − = −
∂c1 D D ∂c1 D ∂c1 D D

and setting the partial derivative zero, to obtain the minimum

∂E
= 0 = c1 ( H aa − E ) + c 2 ( H ab − SE ) [10a]
∂c1

in analogy we obtain for the partial derivative with respect to c2 (try to do it!)

∂E
= 0 = c1 ( H ab − SE ) + c 2 ( H bb − E ) [10b]
∂c 2

If we try to solve this system of simultaneous equations in the traditional manner we will
only obtain the trivial solutions c1 = c2 = 0. These are not meaningful solutions since they
do not describe a suitable wavefunction. Only if the determinant of the coefficients itself
is zero can we obtain non-trivial solutions which, however, must be associated with
specific values of E. Ask yourself at this point if this is unexpected?

What do you know about wavefunctions that fulfill Schroedinger’s equation? They are
solutions associated with a specific energy eigenvalue, i.e., an electron wave in the
electric field of a nucleus can not assume any energy value but only discrete values.
Recall the results of the particle in a box treatment from your properties of matter course
which show this clearly! Here it is the same. If we want to attribute any deliberate energy
value to our MO wavefunction, [10a] and [10b] essentially tell you that this is not
possible.

It is therefore not surprising that the resulting wavefunctions for molecular orbitals will
also be associated with very specific energy eigenvalues as indicated above.
4. Finding the coefficients c1 and c2 associated with the wavefunction

The further strategy is the following: we solve for the energy eigenvalues by setting the
determinant of the coefficients zero, substitute the resulting energy eigenvalues back into
[10] and evaluate c1 and c2. Solving (note that: Haa = Hbb)

H aa − E H ab − SE
=0
H ab − SE H aa − E

gives

( H aa − E ) 2 − ( H ab − SE ) 2 = 0 [11]

In this case the resulting equation is quadratic in E. Try solving this for E! The results
will be

H aa + H ab H aa − H ab
E1 = and E2 = [12]
1+ S 1− S

Substitution of these energy eigenvalues back into [10] allows solving for the ratio of c1
and c2 (try to do this!)

c1
= ±1 [13]
c2

We obtain two wavefunctions that fulfill our requirement that the associated energy
eigenvalues assume a minimum value and as we had expected the wavefunction will be
symmetric. Substituting the values for c1/c2 into [7]

ψ 1 = c1 (ψ 1sa + ψ 1sb ) and ψ 2 = c1 (ψ 1sa − ψ 1sb ) [14]

The remaining unknown constant c1 can now be evaluated using a normalisation

condition since the probability of finding the electron anywhere in each of the molecular
orbitals must be 1. Mathematically this translates into (compare [5])

∫ψ
2
dτ = 1

and for the MO wavefunctions ψ1, ψ2 given in [14]

c12 ( ∫ψ 12sa dτ ± 2 ∫ψ 1saψ 1sb dτ + ∫ψ 12sb dτ ) = 1 [15]

since ψ1sa and ψ1sb are pure hydrogen 1s wavefunctions

∫ψ dτ = ∫ψ 12sb dτ =1
2
1sa

and

∫ψ ψ 1sb dτ = S
1sa

S is the so-called overlap integral, a scalar quantity that can be determined. With this the
value for c1 is

1
c1 = [16]
2 ± 2S

And the final wavefunctions of the MOs are

1
ψ1 = (ψ 1sa + ψ 1sb ) [17a]
2 + 2S

1
ψ2 = (ψ 1sa − ψ 1sb ) [17b]
2 − 2S

We will discuss in the lecture what these two wavefunctions “look” like and what the
associated energy values mean. Based on this we will generalise the principle of the
LCAO method to other two-atom systems and simple multi-atom systems.

It should be mentioned here that the simple treatment using a linear combination of
hydrogen 1s wavefunctions does not give a satisfactory result. We will discuss the
reasons for this in the lecture, but essentially our simplified MO approach neglects
electron-electron repulsions. We can get a better numerical approximation if we allow the
contribution of higher lying but vacant orbitals such as 2s and 2p.