5.

61 Physical Chemistry Lecture #28 1

MOLECULAR ORBITAL THEORY­ PART I

At this point, we have nearly completed our crash­course introduction to
quantum mechanics and we’re finally ready to deal with molecules. Hooray!
To begin with, we are going to treat what is absolutely the simplest
molecule we can imagine: H+2 . This simple molecule will allow us to work
out the basic ideas of what will become molecular orbital (MO) theory.

We set up our coordinate r

system as shown at right, e­
with the electron positioned rB rA
at r, and the two nuclei
positioned at points RA and
RB, at a distance R from one HA HB
another. The Hamiltonian is RA R RB
easy to write down: H2+ Coordinates
∇2 ∇2 1 1 1
Ĥ = − 1 ∇r2 − A − B − − +
2 2M A 2MB ˆ − r̂
R ˆ − r̂
R ˆ −R
R ˆ
A B A B

Electron HA HB e­­HA e­­HB HA­HB

Kinetic Kinetic Kinetic Attraction Attraction Repulsion
Energy Energy Energy
Now, just as was the case for atoms, we would like a picture where we can
separate the electronic motion from the nuclear motion. For helium, we
did this by noting that the nucleus was much heavier than the electrons
and so we could approximate the center of mass coordinates of the
system by placing the nucleus at the origin. For molecules, we will make a
similar approximation, called the Born­Oppenheimer approximation.
Here, we note again that the nuclei are much heavier than the electrons.
As a result, they will move much more slowly than the light electrons.
Thus, from the point of view of the electrons, the nuclei are almost
sitting still and so the moving electrons see a static field that arises
from fixed nuclei. A useful analogy here is that of gnats flying around on
the back of an elephant. The elephant may be moving, but from the gnats’
point of view, the elephant is always more or less sitting still. The
electrons are like the gnats and the nucleus is like the elephant.
5.61 Physical Chemistry Lecture #28 2

The result is that, if we are interested in the electrons, we can to a good

approximation fix the nuclear positions – RA and RB – and just look at the
motion of the electrons in a molecule. This is the B­Oppenheimer
approximation, which is sometimes also called the clamped­nucleus
approximation, for obvious reasons. Once the nuclei are clamped, we can
make two simplifications of our Hamiltonian. First, we can neglect the
kinetic energies of the nuclei because they are not moving. Second, because
the nuclei are fixed, we can replace the operators R ˆ and R ˆ with the
A B
numbers RA and RB. Thus, our Hamiltonian reduces to
∇2 1 1 1
el ( )
Hˆ R A ,R B = − r −
2
− +
R A − r̂ R B − r̂ R A − R B
where the last term is now just a number – the electrostatic repulsion
between two protons at a fixed separation. The second and third terms
depends only on the position of the electron, r, and not its momentum, so
we immediately identify those as a potential and write:
∇2 1
( ) 2
RA , RB
Hˆ el R A ,R B = − r +Veff ( r̂ ) +
RA − RB
This Hamiltonian now only contains operators for the electrons (hence the
subscript “el”), but the eigenvalues of this Hamiltonian depend on the
distance, R, between the two nuclei. For example, the figure below shows
the difference between the effective potentials the electron “feels” when
the nuclei are close together versus far apart:
R Small R Large

Veff(r)

Likewise, because the electron feels a different potential at each bond

distance R, the wavefunction will also depend on R. In the same limits as
above, we will have:
ψel(r)

R Small R Large
5.61 Physical Chemistry Lecture #28 3

Finally, because the electron eigenfunction, ψel, depends on R then the

eigenenergy of the electron, Eel(R), will also depend on the bond length.
Mechanically, then, what we have to do is solve for the electronic
eigenstates, ψel, and their associated eigenvalues, Eel(R), at many different
fixed values of R. The way that these eigenvalues change with R will tell us
about how the energy of the molecule changes as we stretch or shrink the
bond. This is the central idea of the Born­Oppenheimer approximation, and
it is really very fundamental to how chemists think about molecules. We
think about classical point­like nuclei clamped at various different positions,
with the quantum mechanical electrons whizzing about and gluing the nuclei
together. When the nuclei move, the energy of the system changes because
the energies of the electronic orbitals change as well. There are certain
situations where this approximation breaks down, but for the most part the
Born­Oppenheimer picture provides an extremely useful and accurate way to
think about chemistry.

How are we going to solve for these eigenstates? It should be clear that
looking for exact solutions is going to lead to problems in general. Even for
H2+ the solutions are extremely complicated and for anything more complex
than H2+exact solutions are impossible. So we have to resort to
approximations again. The first thing we note is that if we look closely at
our intuitive picture of the H2+ eigenstates above, we recognize that these
molecular eigenstates look very much like the sum of the 1s atomic orbitals
for the two hydrogen atoms. That is, we note that to a good approximation
we should be able to write:
ψ el ( r ) ≈ c11s A ( r ) + c2 1sB ( r )
where c1 and c2 are constants. In the common jargon, the function on the
left is called a molecular orbital (MO), whereas the functions on the right
are called atomic orbitals (AOs). If we write our MOs as sums of AOs, we
are using what is called the linear combination of atomic orbitals (LCAO)
approach. The challenge, in general, is to determine the “best” choices for c1
and c2 – for H2+ it looks like the best choice for the ground state will be
c1=c2. But how can we be sure this is really the best we can do? And what
about if we want something other than the ground state? Or if we want to
describe a more complicated molecule like HeH+2?
5.61 Physical Chemistry Lecture #28 4

THE VARIATIONAL PRINCIPLE

In order to make further progress, we will use the Variational Principle
to predict a better estimate of the ground state energy. This method is
very general and its use in physical chemistry is widespread. Assume you
have a Hamiltonian (such as the Helium atom) but you don’t know the
ground state energy E0 and or ground state eigenfunction φ0.
Ĥφ0 = E0φ0 ⇒ Ĥ = ∫ φ0*Ĥ φ0dτ = ∫ φ0*E0φ0dτ = E0
Now, assume we have a guess, ψ , at the ground state wavefunction, which we
will call the trial wavefunction. Compute the average value of the energy for
the trial wavefunction:
∫ψ *Ĥψ dτ = ψ * Hˆψ dτ (if ψ normalized)
Eavg = ∫
∫ψ *ψ dτ
the Variational Theorem tells us that Eavg ≥ E0 for any choice of the trial
function ψ! This make physical sense, because the ground state energy is,
by definition, the lowest possible energy, so it would be nonsense for the
average energy to be lower.

SIDEBAR: PROOF OF VARIATIONAL THEOREM

Expand ψ (assumed normalized) as linear combination of the unknown
eigenstates, φn , of the Hamiltonian:
ψ = ∑ anφn
n

Note that in practice you will not know these eigenstates. The important
point is that no matter what function you choose you can always expand it in
terms of the infinite set of orthonormal eigenstates of Ĥ .

∫ψ *ψ dτ = ∑ an am ∫ φn *φm dτ = ∑ an amδ mn = ∑ an
* * 2
=1
n ,m n ,m n

Eavg = ∫ψ *Ĥψ dτ = ∑ an*am ∫ φn *Ĥ φm dτ = ∑ an*am ∫ φn *Emφm dτ

n ,m n ,m

= ∑ an*am Emδ mn = ∑ an En
2

n ,m n
Now, subtracting the ground state energy from the average
5.61 Physical Chemistry Lecture #28 5

E0 = 1• E0 = ∑ an E0
2

⇒ Eavg − E0 = ∑ an En − ∑ an E0 = ∑ an
2 2 2
( En − E0 ) ≥ 0 since En ≥ E0
n n n
Where, in the last line we have noted that the sum of terms that are non­
negative must also be non­negative. It is important to note that the
equals sign is only obtained if an=0 for all states that have En>E0. In this
situation, ψ actually is the ground state of the system (or at least one
ground state, if the ground state is degenerate).

The variational method does two things for us. First, it gives us a means of
comparing two different wavefunctions and telling which one is closer to the
ground state – the one that has a lower average energy is the better
approximation. Second, if we include parameters in our wavefunction
variation gives us a means of optimizing the parameters in the following way.
Assume that ψ depends on some parameters c – such as is the case for our
LCAO wavefunction above. We’ll put the parameters in brackets ­ψ[c] – in
order to differentiate them from things like positions that are inside of
parenthesis ­ψ(r).Then the average energy will depend on these parameters:

(c) = ∫
*
ψ [ c ] Ĥψ [c ] dτ
Eavg
∫ψ [c ] *ψ [c ] dτ
Note that, using the variational principle, the best parameters will be the
ones that minimize the energy. Thus, the best parameters will satisfy
∂ ∫ ψ [ c ] Hψ [c ] dτ
*
∂Eave ( c ) ˆ
= =0
∂ci ∂ci ∫ ψ [c ]*ψ [c ] dτ
Thus, we can solve for the optimal parameters without knowing anything
about the exact eigenstates!

Let us apply this in the particular case of our LCAO­MO treatment of H2+.
Our trial wavefunction is:
ψ el [ c ] = c11s A + c2 1sB
5.61 Physical Chemistry Lecture #28 6

where c=(c1 c2). We want to determine the best values of c1 and c2 and
the variational theorem tells us we need to minimize the average energy
to find the right values. First, we compute the average energy. The
numerator gives:
∫ψ el Ĥ elψ el dτ = ∫ ( c11sA + c21sB ) Ĥ ( c11sA + c21sB ) dτ
* *

= c1*c1 ∫ 1s A Ĥ el 1s A dτ + c1* c2 ∫ 1s A Ĥ el 1sB dτ + c2* c1 ∫ 1sB Ĥ el 1s A dτ + c2* c2 ∫ 1sB Ĥ el 1sB dτ

≡H11 ≡H12 ≡H21 ≡H22

≡ c1* H11c1 + c1* H12 c2 + c2* H 21c1 + c2* H 22 c2
while the normalization integral gives:
∫ψ ψ el dτ = ∫ ( c11s A + c2 1sB ) ( c11s A + c2 1sB ) dτ
* *
el

= c1*c1 ∫ 1s A1s A dτ + c1*c2 ∫ 1s A1sB dτ + c2* c1 ∫ 1sB 1s A dτ + c2* c2 ∫ 1sB 1sB dτ

≡S11 ≡S12 ≡S21 ≡S22

≡ c1* S11c1 + c1* S12 c2 + c2* S21c1 + c2* S22 c2

So that the average energy takes the form:

c * H c + c * H c + c2* H 21c1 + c2* H 22 c2
Eavg ( c ) = 1 * 11 1 1 * 12 2
c1 S11c1 + c1 S12 c2 + c2* S21c1 + c2* S22 c2
We note that there are some simplifications we could have made to this
formula: for example, since our 1s functions are normalized S11=S22=1. By
not making these simplifications, our final expressions will be a little more
general and that will help us use them in more situations.

Now, we want to minimize this average energy with respect to c1 and c2.
Taking the derivative with respect to c1 and setting it equal to zero [Note:
when dealing with complex numbers and taking derivatives one must treat
variables and their complex conjugates as independent variables. Thus d/dc1
has no effect on c1*]:
∂Eavg c1* H11 + c2* H 21
=0= *
∂c1 c1 S11c1 + c1* S12 c2 + c2* S21c1 + c2* S22 c2
c1* H11c1 + c1* H12 c2 + c2* H 21c1 + c2* H 22 c2
− 2 (c*
1 11S + c2* S21 )
(c *
S c + c S c + c2 S21c1 + c2 S22 c2 )
1 11 1
*
1 12 2
* *
5.61 Physical Chemistry Lecture #28 7

c1* H11c1 + c1* H12 c2 + c2* H 21c1 + c2* H 22 c2 *

⇒ 0 = ( c1* H11 + c2* H 21 ) − * * * * ( c1 S11 + c2* S21 )
c1 S11c1 + c1 S12 c2 + c2 S21c1 + c2 S22 c2
⇒ 0 = ( c1* H11 + c2* H 21 ) − Eavg ( c1* S11 + c2* S21 )
Applying the same procedure to c2:
∂Eavg c1* H12 + c2* H 22
=0= *
∂c2 c1 S11c1 + c1* S12 c2 + c2* S21c1 + c2* S22 c2
c1* H11c1 + c1* H12 c2 + c2* H 21c1 + c2* H 22 c2
− 2 (c*
1 12S + c2* S22 )
(c *
S c + c S c + c2 S21c1 + c2 S22 c2 )
1 11 1
*
1 12 2
* *

c1* H11c1 + c1* H12 c2 + c2* H 21c1 + c2* H 22 c2 *

⇒ 0 = ( c H12 + c2 H 22 ) −
1
* *
* * * * ( c1 S12 + c2* S22 )
c1 S11c1 + c1 S12 c2 + c2 S21c1 + c2 S22 c2
⇒ 0 = ( c1* H12 + c2* H 22 ) − Eavg ( c1* S12 + c2* S22 )
We notice that the expressions above look strikingly like matrix­vector
operations. We can make this explicit if we define the Hamiltonian matrix:
⎛ 1s A Ĥ el 1sB dτ ⎞
H≡⎜
⎛ H11 H12 ⎞ ⎜ 1s A Ĥ el 1s A dτ ∫ ∫ ⎟
⎟≡⎜ ⎟
H
⎝ 21 H 22 ⎠
∫
⎜ 1sB Ĥ el 1s A dτ
⎝ ∫
1sB Ĥ el 1sB dτ ⎟
⎠
and the Overlap matrix:
⎛ 1s A 1sB dτ ⎞
S≡⎜
⎛ S11 S12 ⎞ ⎜ 1s A 1s A dτ ∫ ⎟∫
⎟≡⎜ ⎟
⎝ S21 S22 ⎠ ⎜ 1sB 1s A dτ
⎝ ∫
1sB 1sB dτ ⎟
⎠∫
Then the best values of c1 and c2 satisfy the matrix eigenvalue equation:
⎛H H12 ⎞ ⎛S S12 ⎞
(c 1
*
)
c2* ⎜ 11
⎝ H 21
⎟
H 22 ⎠
(
= Eavg c1* )
c2* ⎜ 11
⎝ S21 S22 ⎟⎠

Which means that:

∂Eavg
= 0 ⇔ c† iH = Eavg c† iS Eq. 1
∂c
This equation doesn’t look so familiar yet, so we need to massage it a bit.
First, it turns out that if we had taken the derivatives with respect to c1*
and c2* instead of c1 and c2, we would have gotten a slightly different
equation:
⎛ H11 H12 ⎞ ⎛ c1 ⎞ ⎛S S12 ⎞ ⎛ c1 ⎞
⎜H ⎟ ⎜ ⎟ = Eavg ⎜ 11 ⎟⎜ ⎟
⎝ 21 H 22 ⎠ ⎝ c2 ⎠ ⎝ S21 S22 ⎠ ⎝ c2 ⎠
or
5.61 Physical Chemistry Lecture #28 8

∂Eavg
= 0 ⇔ H ic = Eavg Sic Eq. 2
∂c *
Taking the derivatives with respect to c1* and c2* is mathematically
equivalent to taking the derivatives with respect to c1 and c2 [because we
can’t change c1 without changing its complex conjugate, and vice versa].
Thus, the two matrix equations (Eqs. 1&2) above are precisely equivalent, and
the second version is a little more familiar. We can make it even more
familiar if we think about the limit where 1sA and 1sB are orthogonal (e.g.
when the atoms are very, very far apart). Then we would have for the
Overlap matrix:
⎛ 1s 1s dτ 1s A 1sB dτ ⎞ ⎛ 1 0 ⎞
S≡ ⎜ ∫
A A
⎟= ∫
⎜⎜ ⎟ ⎜ ⎟=1
⎝ ∫
1sB 1s A dτ
⎠ ∫
1sB 1sB dτ ⎟ ⎝ 0 1 ⎠

Thus, in an orthonormal basis the overlap matrix is just the identity matrix
and we can write Eq. 2 as:
∂Eavg
= 0 ⇔ H ic = Eavg c
∂c *
Now this equation is in a form where we certainly recognize it: this is an
eigenvalue equation. Because of its close relationship with the standard
eigenvalue equation, Eq. 2 is usually called a Generalized Eigenvalue
Equation.

In any case, we see the quite powerful result that the Variational theorem
allows us to turn operator algebra into matrix algebra. Looking for the
lowest energy LCAO state is equivalent to looking for the lowest eigenvalue
of the Hamiltonian matrix H. Further, looking for the best c1 and c2 is
equivalent to finding the lowest eigenvector of H.

Let’s go ahead and apply what we’ve learned to obtain the MO coefficients c1
and c2 for H2+. At this point we make use of several simplifications. The
off­diagonal matrix elements of H are identical because the Hamiltonian is
Hermitian and the orbitals are real:
*
∫ 1s A Hˆ el 1sB dτ = (∫ 1s B
*
Hˆ el 1s A dτ ) = ∫ 1s Hˆ 1s dτ ≡ V
B el A 12

Meanwhile, the diagonal elements are also equal, but for a different reason.
The diagonal elements are the average energies of the states 1sA and 1sB. If
these energies were different, it would imply that having the electron on one
5.61 Physical Chemistry Lecture #28 9

side of H2+ was more energetically favorable than having it on the other,
which would be very puzzling. So we conclude

∫ 1s A Ĥ el 1s A dτ ∫
= 1sB Ĥ el 1sB dτ ≡ ε
Finally, we remember that 1sA and 1sB are normalized, so that

∫ 1s 1s
A A dτ ∫
= 1sB 1sB dτ = 1
and because the 1s orbitals are real, the off­diagonal elements of S are also
the same:

∫ ∫
S12 = 1s A 1sB dτ = 1sB 1s A dτ = S21 .
Incorporating all these simplifications gives us the generalized eigenvalue
equation:
⎛ ε V12 ⎞ ⎛ c1 ⎞ ⎛ 1 S12 ⎞ ⎛ c1 ⎞
⎜V ⎟ ⎜ ⎟ = Eavg ⎜ ⎟⎜ ⎟
⎝ 12 ε ⎠ ⎝ c2 ⎠ ⎝ S21 1 ⎠ ⎝ c2 ⎠
All your favorite mathematical programs (Matlab, Mathematica, Maple,
MathCad…) are capable of solving for the generalized eigenvalues and
eigenvectors, and for more complicated cases we suggest you use them.
However, this case is simple enough that we can solve it by guess­and test.
Based on our physical intuition above, we guess that the correct eigenvector
will have c1=c2. Plugging this in, we find:
⎛ ε V12 ⎞ ⎛ c1 ⎞ ⎛ 1 S12 ⎞ ⎛ c1 ⎞
⎜ ⎟ ⎜ ⎟ = Eavg ⎜ ⎟⎜ ⎟
⎝ V12 ε ⎠ ⎝ c1 ⎠ ⎝ S21 1 ⎠ ⎝ c1 ⎠
⎛ (ε + V12 ) c1 ⎞ ⎛ (1 + S12 ) c1 ⎞
⇒⎜ ⎟ = Eavg ⎜ ⎟
⎝ (ε + V12 ) c1 ⎠ ⎝ (1 + S12 ) c1 ⎠
ε + V12
⇒ Eavg = ≡ Eσ
1 + S12
Thus, our guess is correct and one of the eigenvectors of this matrix has
c1=c2. This eigenvector is the σ­bonding state of H2+, and we can write down
the associated orbital as:
ψ elσ = c11s A + c2 1sB = c11s A + c11sB ∝ 1s A + 1sB
where in the last expression we have noted that c1 is just a normalization
constant. In freshman chemistry, we taught you that the σ­bonding orbital
existed, and this is where it comes from.

We can also get the σ∗­antibonding orbital from the variational procedure.
Since the matrix is a 2x2 it has two unique eigenvalues: the lowest one
(which we just found above) is bonding and the other is antibonding. We can
5.61 Physical Chemistry Lecture #28 10

again guess the form of the antibonding eigenvector, since we know it has
the characteristic shape +/­, so that we guess the solution is c1=­c2:
⎛ ε V12 ⎞ ⎛ c1 ⎞ ⎛ 1 S12 ⎞ ⎛ c1 ⎞
⎜ ⎟⎜ ⎟ = Eavg ⎜ ⎟⎜ ⎟
⎝ V12 ε ⎠ ⎝ − c1 ⎠ ⎝ S21 1 ⎠ ⎝ − c1 ⎠
⎛ (ε − V12 ) c1 ⎞ ⎛ (1 − S12 ) c1 ⎞
⇒⎜ ⎟ = Eavg ⎜ ⎟
⎝ − (ε − V12 ) c1 ⎠ ⎝ − (1 − S12 ) c1 ⎠
ε − V12
⇒ Eavg = = Eσ *
1 − S12
so, indeed the other eigenvector
has c1=­c2. The corresponding
antibonding orbital is given by: ψσ(r)
ψ elσ * = c11s A + c2 1sB = c11s A − c11sB ∝ 1s A − 1sB

where we note again that c1 is

just a normalization constant.
Given these forms for the
bonding and antibonding orbitals, ψσ(r)
we can draw a simple picture for
the H2+ MOs (see right).

We can incorporate the energies obtained above into a simple MO diagram of

H2+:
ε − V12
Eσ∗=
1 − S12

E1sA=ε E1sB=ε

ε + V12
Eσ=
1 + S12
On the left and right, we draw the energies of the atomic orbitals (1sA and
1sB) that make up our molecular orbitals (σ and σ*) in the center. We note
5.61 Physical Chemistry Lecture #28 11

that when the atoms come together the energy of the bonding and
antibonding orbitals are shifted by different amounts:
ε − V12 ε − V12 ε (1 − S12 ) ε S12 − V12
Eσ * − E1s = −ε = − =
1 − S12 1 − S12 1 − S12 1 − S12
ε + V12 ε (1 + S12 ) ε + V12 ε S12 − V12
E1s − Eσ * = ε − = − =
1 + S12 1 + S12 1 + S12 1 + S12
Now, S12 is the overlap between two 1s orbitals. Since these orbitals are
never negative, S12 must be a positive number. Thus, the first denominator
is greater than the second, from which we conclude
ε S − V12 ε S12 − V12
Eσ * − E1s = 12 > = E1s − Eσ *
1 − S12 1 + S12
Thus, the antibonding orbital is destabilized more than the bonding orbital is
stabilized. This conclusion need not hold for all diatomic molecules, but it is
a good rule of thumb. This effect is called overlap repulsion. Note that in
the special case where the overlap between the orbitals is negligible, S12
goes to zero and the two orbitals are shifted by equal amounts. However,
when is S12 nonzero there are two effects that shift the energies: the
physical interaction between the atoms and the fact that the 1sA and 1sB
orbitals are not orthogonal. When we diagonalize H, we account for both
these effects, and the orthogonality constraint pushes the orbitals upwards
in energy.