38.

1 Restricted Closed Shell Hartree-Fock

The so-called closed-shell restricted Hartree-Fock method is essentially the Hartree-Fock approach
implemented for systems with an even number n of electrons, with each orbital j populated by 2
electrons with opposite spins (one with spin , described by
j
, and the other one with spin ,
described by

j
). It is called restricted to indicate that the spin states are restricted to be either
or , and closed shell to indicate that each shell
j
is full with 2 electrons. The system is described
by the Slater determinant,
| = |
1

2
...
n

= |
1

1
...
n/2

n/2

(117)
where
1
=
1
,
2
=

1
, ...,
n1
=
n/2
,
n
=

n/2
.
The energy of this closed-shell restricted Hartree-Fock wave function is computed, according
to Eq. (101), by replacing the sums over n spin-orbitals
j
by sums over n/2 spin-orbitals with
spin ,
j
and n/2 spin-orbitals with spin ,

j
, as follows:

E =

n/2
j=1

j
|

h
(1)
|
j
+
1
2

n/2
j=1

n/2
k=1

k
|
e
2
r
12
|
j

k

j

k
|
e
2
r
12
|
k

+
1
2

n/2
j=1

n/2
k=1

k
|
e
2
r
12
|
j

X
X
X
X
X
X
X
XX

k
|
e
2
r
12
|

n/2
j=1

j
|

h
(1)
|

j
+
1
2

n/2
j=1

n
k=1

k
|
e
2
r
12
|

X
X
X
X
X
X
X
XX

k
|
e
2
r
12
|
k

+
1
2

n/2
j=1

n/2
k=1

k
|
e
2
r
12
|

k
|
e
2
r
12
|

j
,
(118)
where we can cross out the terms that cancel upon integration over the spin variable to obtain:

E = 2

n/2
j=1

j
|

h
(1)
|
j
+

n/2
j=1

n/2
k=1
2
j

k
|
e
2
r
12
|
j

k

j

k
|
e
2
r
12
|
k

j
. (119)
Analogously to the general case, we minimize

E with respect to
j
, subject to the constraint of
orthonormal orbitals by applying the Lagrange multiplier method for the following functional:
L(
1
, ...,
n
) =

E +

jk
[
j
|
k

jk
], (120)
where
jk
are the Lagrange multipliers. Varying the spin orbitals
j
in an innitesimal amount
j
,
with respect to expansion coefcients c
kj
, we obtain:
L(
1
, ...,
n
) =

E +
n

j=1
n

k=1

jk
[
j
|
k
+
j
|
k
]. (121)
Varying the spatial orbitals
j
in an innitesimal amount
j
with respect to the expansion coef-
cients c
j
, we obtain:

E = 2

n/2
j=1

j
|

h
(1)
|
j
+

n/2
j=1

n/2
k=1
2
j

k
|
e
2
r
12
|
j

k

j

k
|
e
2
r
12
|
k

n/2
j=1

n/2
k=1
2
j

k
|
e
2
r
12
|
j

k

j

k
|
e
2
r
12
|
k

+2

n/2
j=1

j
|

h
(1)
|
j
+

n/2
j=1

n/2
k=1
2
j

k
|
e
2
r
12
|
j

k

j

k
|
e
2
r
12
|
k

n/2
j=1

n/2
k=1
2
j

k
|
e
2
r
12
|
j

k

j

k
|
e
2
r
12
|
k

(122)
96
which gives

E = 2

n/2
j=1

j
|

h
(1)
|
j
+ 2

n/2
j=1

n/2
k=1
2
j

k
|
e
2
r
12
|
j

k

j

k
|
e
2
r
12
|
k

j
+c.c.
(123)
Substituting Eq. (123) into Eq. (121), we obtain:
L = 2
n/2

j=1

j
|
_
_
h
(1)
|
j
+
n/2

k=1
2
k
|
e
2
r
12
|
k
|
j

k
|
e
2
r
12

P
12
|
k
|
j
+
jk
|
k

_
_
+c.c. = 0,
(124)
which is satised when
_
_
h
(1)
+
n/2

k=1

k
|
e
2
r
12
[2

P
12
]|
k

_
_
|
j
=
jk
|
k
. (125)
To write Eq. (125) in the canonical eigenvalue form, we change the basis set according to the
unitary transformation,
|

j
=

k
|
k

kj
, (126)
with

= 1. Considering that the Lagrange multipliers matrix is Hermitian (since the func-
tional L is real), it is always possible to nd a that diagonalizes according to the similarity
transformation:

(127)
Such a transformation denes the set of canonical orbitals |

j
for which

f
(1)
rhf
|

j
=

jj
|

j
, (128)
for j = 1 n/2, where

f
(1)
rhf
is the closed-shell restricted Hartree-Fock operator,

f
(1)
rhf
=

h
(1)
+

V
rhf
1
, (129)
where

V
rhf
j
is the restricted Hartree-Fock potential describing the interactions between electrons
of the same spin, as follows:

V
rhf
1
= 2J
1
(r
1
)

X
1
(r
1
), (130)
where J
1
(r
1
) is the Coulombic mean-eld potential due to the presence of other electrons of the
same spin,
J
1
(r
1
) =
n/2

k=1

k
(2)|
e
2
r
12
|

k
(2). (131)
Analogously,

X
1
(r
1
) is dened as the exchange operator,

X
1
(r
1
) =
n/2

k=1

k
(2)|
e
2
r
12

P
12
|

k
(2), (132)
97
where the permutation operator

P
12
interchanges electrons 1 and 2.
Equation (128) denes a self-consistent eld (SCF) problem since

f
(1)
rhf
depends on
j
through

V
rhf
1
. To solve this SCF problem, we rst obtain approximate solutions
j
by approximating

f
(1)
rhf
by

h
j
(i.e., neglecting the electron-electron interactions introduced by J
1
and

X
1
), or by using a
semiempirical method (like the H uckel method described in these lectures). The resulting approx-
imate solutions
j
are then used to compute J
1
and

X
1
, giving a better approximation to

f
(1)
rhf
that
can be used to obtain improved functions
j
. The process is repeated until convergence.
In practice, the restricted Hartree-Fock eigenvalue problem, introduced by Eq. (128), is solved
with a set of matrix equations, obtained by substituting

j
by a linear combination of atomic
orbitals, analogous Eq. (95): |

j
=

n/2
k=1
C
kj
|
k
. Making the substitution and multiplying from
the left with
k
|, we obtain:
n/2

k=1

k
|

f
(1)
rhf
|
k
C
kj
=
jj
n/2

k=1

k
|
j
C
kj
, (133)
or, in matrix form,
F
rhf
C = SC

(134)
where we have introduced the overlap matrix S
jk
=
j
|
k
, the restricted Hartree Fock matrix
F
rhf
jk
=
j
|

f
(1)
rhf
|
k
, (135)
and the matrix of column eigenvectors C
kj
dening the canonical orbitals

j
, with energies
jj
, in
the basis of atomic orbitals |
k
.
The electronic density (r) of the system with 2 electrons per orbital, populating the lowest n/2
states (i.e., closed-shell Hartree-Fock density) can be computed, as follows:
(r) = 2

n/2
k=1

(r)

k
(r)
= 2

lm

l
(r)
m
(r)

n/2
k=1
C

kl
C
mk
= 2

lm

l
(r)
m
(r)P
lm
(136)
where
P
lm
=
n/2

k=1
C

kl
C
mk
(137)
dene the elements of the electronic density matrix P.
The elements of the density matrix, P
lm
, are thus computed from the solution of the eigenvalue
problem, introduced by Eq. (134). The resulting elements of the density matrix P
lm
can be used to
compute not only the density, according to Eq. (136), but also the restricted Hartree-Fock matrix,
according to Eq. (135) since

f
(1)
rhf
=

h
(1)
+

V
rhf
1
, with

V
rhf
1
= 2J
1
(r
1
)

X(r
1
), where
J
1
(r
1
) =

n/2
k=1
_
dr
2

k
(r
2
)
e
2
r
12

k
(r
2
)
=

n/2
k=1
_
dr
2

n/2
l=1

n/2
m=1
C

km

m
(r
2
)
e
2
r
12
C
lk

l
(r
2
)
=

n/2
l=1

n/2
m=1

n/2
k=1
C

km
C
lk
_
dr
2

m
(r
2
)
e
2
r
12

l
(r
2
)
=

n/2
l=1

n/2
m=1
P
ml
_
dr
2

m
(r
2
)
e
2
r
12

l
(r
2
)
(138)
98
and

X
1
(r
1
) =

n/2
k=1

k
|
e
2
r
jk

P
kj
|

k
,
=

n/2
l=1

n/2
m=1
P
ml
_
dr
2

m
(r
2
)
e
2
r
12

P
12

l
(r
2
).
(139)
Therefore,

f
(1)
rhf
(r
1
) =

h
(1)
+
n/2

l=1
n/2

m=1
P
ml
_
2
m
(r
2
)|
e
2
r
12
|
l
(r
2
)
m
(r
2
)|
e
2
r
12

P
12
|
l
(r
2
)
_
, (140)
and
F
rhf
jk
=
j
|

f
(1)
rhf
|
k
,
= H
core
jk
+G
jk
,
(141)
with
H
core
jk
=
j
(1)|

h
(1)
|
k
(1), (142)
and
G
jk
=
n/2

l=1
n/2

m=1
P
ml
_
2
j
(r
1
)
m
(r
2
)|
e
2
r
12
|
l
(r
2
)
k
(r
1
)
j
(r
1
)
m
(r
2
)|
e
2
r
12

P
12
|
l
(r
2
)
k
(r
1
)
_
.
(143)
To solve Eq. (134), we rst diagonalize the overlap matrix by computing the matrix X = S
1/2
that transforms the overlap matrix into the identity matrix, as follows: X

SX = 1. Then, we
introduce the matrix

C, as follows:
C = X

C, (144)
that, according to Eq. (134), satises the eigenvalue problem:
F
rhf
X

C = SX

(145)
or

F
rhf

C =

C

(146)
with

F
rhf
= X

F
rhf
X. (147)
These equations allow for the implementation of the self-consistent-eld restricted Hartree-
Fock (SCF RHF) method, for a xed nuclear conguration, as follows:
1. Calculate the matrix elements S
jk
, H
core
jk
and the 2-electron integrals
j
(r
1
)
m
(r
2
)|
e
2
r
12
|
l
(r
2
)
k
(r
1
)
and
j
(r
1
)
m
(r
2
)|
e
2
r
12
|
l
(r
1
)
k
(r
2
).
2. Diagonalize S
jk
to obtain X
jk
3. Obtain an approximate density matrix P
jk
, according to Eq. (137), by solving Eq. (134) with

f
(1)

h
(1)
, or

f
(1)

h
(1)
EH
, where

h
(1)
EH
is the semiempirical extended-H uckel Hamiltonian
of the system.
99
4. Compute the matrix elements G
jk
, according to Eq. (143), using P
jk
and the 2-electron
integrals.
5. Compute the Fock matrix F
rhf
jk
, according to Eq. (141), by using H
core
jk
, P
jk
and the 2-
electron integrals.
6. Compute the transformed Fock matrix

F
rhf
jk
by using F
rhf
jk
and X
jk
, according to Eq. (147).
7. Obtain

C and

by solving Eq. (146).

8. Calculate C by using Eq. (144).
9. Compute a new density matrix P, according to Eq. (137), based on C obtained in (8).
10. If P has changed more than a given tolerance, relative to the previous iteration, go to (4).
Otherwise, the SCF calculation has converged and the solution is given by the current eigen-
vectors C and eigenvalues .
Conguration Interaction
Improvement over the one-determinant trial wave function can be achieved by using a trial
wave function that involves a linear combination of Slater determinants. This method is known as
conguration interaction. The energy correction over the Hartree-Fock energy,
E
cor
= E E
HF
,
is known as correlation energy.
100
39 Quantum Mechanics/Molecular Mechanics Methods
A practical approach for describing the electronic structure of a molecular fragment in a complex
molecular environment is the so-called quantum mechanics/molecular mechanics (QM/MM) ap-
proach, where the environment (represented by sticks in the gure) is described in terms of a sum
of classical potentials V
MM
described by a molecular mechanics (MM) force eld, while the molec-
ular fragment (R
1
in the Figure below) is described by quantum chemistry (QM) methods, as for
example by a Slater determinant obtained according to the Hartree-Fock method. The interaction
between the molecular fragment and the environment is usually dened as the sum of the elec-
trostatic interaction between the atomic charges of the environment and the nuclear and electronic
charges in molecular fragment. If the fragment is covalently bound to the environment, the bond is
broken and the covalency is completed with a link atom (usually a H atom).
The interaction between the fragment and the environment is included as an additional term in
the one-electron core Hamiltonian

h
(1)
, as follows:

h
(i)
=

2
2m
e

2
r
i

j=1
z
j
e
2
r
ji
+
N

j=1
N

=1
z
j
z
j
e
2
r
jj

+

h
(i)
QM/MM
, (148)
with

h
(i)
QM/MM
=
Ne

k=1
z
k
e
2
r
ki
+
N

j=1
Ne

k=1
z
k
z
j
e
2
r
ki
+
Ne

k=1
V
vdw
(r
ki
), (149)
where we have considered a molecular fragment with N nuclei embedded in an environment with
N
e
electrostatic potential atomic charges. The third term in Eq. (149) is a van der Waals potential
that accounts for the interaction of the electrons in the molecular fragment with electrons in the
environment that are not explicitly considered. Implementing the Hartree-Fock method with the
one-electron core-Hamiltonian, introduced by Eq. (148), we obtain the Hartree Fock energy E
QM
of the molecular fragment in the electrostatic eld of the surrounding environment yielding the
total QM/MM energy of the system, as follows:
E
QM/MM
= E
QM
+V
MM
. (150)
101
In this simple form, the molecular fragment is polarized by the surrounding environment. Polariza-
tion of the environment due to the distribution of charges in the molecular fragment can be included
by using a polarizable molecular mechanics force eld, or a moving domain QM/MM approach,
where the charges in the environment are obtained self-consistently.
Another QM/MM approach is the ONIOM methodology, as implemented in Gaussian, where
the total energy is computed, as follows:
E
ONIOM
= E
QM
+V
system
MM
V
fragment
MM
, (151)
where V
system
MM
and V
fragment
MM
are the energies of the complete system and the fragment as described
by the molecular mechanics force eld.
The force elds are usually parametrized to match the experimental or ab-initio ground state
potential energy surfaces as a function of nuclear coordinates. The following section illustrates the
parametrization of the potential energy surface for diatomic molecules.
40 Empirical Parametrization of Diatomic Molecules
The main features of chemical bonding by electron pairs are properly described by the HL model
of H
2
. According to such model, the covalent bond is described by a singlet state,
1

HL
= N
1
[(1)(2) (1)(2)][
A
(1)
B
(2) +
A
(2)
B
(1)],
with energy
1
E
+
=<
1

HL
|H|
1

HL
>=
J +K
1 +S
2
,
where H = h(1) +h(2) +e
2
/r
12
, with
h(1) =

2
2m

2
1

e
2
r
1A

e
2
r
1B
,
h(2) =

2
2m

2
2

e
2
r
2A

e
2
r
2B
,
J =<
A
(1)
B
(2)|H|
A
(1)
B
(2) > Coulomb integral
K =<
A
(1)
B
(2)|H|
A
(2)
B
(1) > Exchange integral
S
2
=<
A
(1)
B
(2)|
A
(2)
B
(1) > .
Similarly, the triplet state is described as follows,
3

HL
= N
3
[
A
(1)
B
(2)
B
(1)
A
(2)]
_

_
[(1)(2) +(1)(2)]
(1)(2)
(1)(2)
,
and has energy
3
E

=
(J K)
(1 S
2
)
.
The energies of the singlet and triplet states are parametrized by the internuclear H-H distance and
can be represented by the following diagram,
102
-
6
.. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. ..
Energy
0
1
E
+
3
E

H-H bond length

R
0
The energies
1
E and
3
E can be approximated by the following analytical functions:
1
E
+
D
_
e
2a(RR
0
)
2e
a(RR
0
)

M(R),
3
E

D
2
_
e
2a(RR
0
)
+ 2e
a(RR
0
)

(R).
Parameters D and a can be obtained by tting M(R) to the actual (experimental or ab-initio) ground
state potential energy surface. Such parametrization allows us to express the Coulombic and Reso-
nance integrals J and K in terms of available experimental (or ab initio) data as follows,
J
1
2
[(M +M

) +S
2
(M M

)],
K
1
2
[(M M

) +S
2
(M +M

)].
This parametrization of Hamiltonian matrix elements illustrates another example of semi-empirical
parametrization that can be implemented by using readily available experimental information (re-
member that in the previous section we described the semiempirical parametrization of the H uckel
model according to the absorption spectrum of the molecule).
The covalent nature of the chemical bond signicantly changes when one of the two atoms in
the molecule is substituted by an atom of different electronegativity. Under those circumstances,
the wave function should include ionic terms, e.g.,
1

ion
A
=

N
A
(1)
A
(2)[(1)(2) (1)(2)],
and
1

ion
B
=

N
B
(1)
B
(2)[(1)(2) (1)(2)].
103
The complete wave function (with both covalent and ionic terms) can be described as follows,
= C
1

1
+C
2

2
, where the covalent wave function is

1
= [(1)(2) (1)(2)](
A
(1)
B
(2) +
A
(2)
B
),
and the ionic wave function is

2
= [(1)(2) (1)(2)][
A
(1)
A
(2)
1
+
B
(1)
B
(2)(1
1
)],
where the parameter
1
is determined by the relative electronegativity of the two atoms. For exam-
ple, consider the HF molecule. For such molecule
1
=1, A represents the F atom, and B represents
the H atom (i.e., due to the electronegativity difference between the two atoms, the predominant
ionic conguration is H
+
F

). Therefore, the ground state energy E

g
is obtained as the lowest
eigenvalue of the secular equation,

H
11
E H
12
H
12
H
22
E

= 0. (152)
Here we have neglected S
12
, assuming that such approximation can be partially corrected according
to the parametrization of H
12
. The semiempirical parametrization strategy can be represented by
the following diagram:
.. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. ..
-
6
.. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. ..
.. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. ..
?
6
?
6
6
?
Energy
R
HF
H
.
F
.
H
+
F

D
HF
H
11
H
22
E
g

D
HF
0
This diagram represents the following curves:
H
11
=

M =

D[e
2a(RR
0
)
2e
a(RR
0
)
] is a covalent state represented by a Morse poten-
tial

M.
H
22
= I EA
332
R
+ Ae
bR
+ CR
9
, is the potential energy surface of the ionic state,
where the difference between the H ionization energy and the F electron afnity, I-EA, corresponds
to the energy of forming the ion pair H
+
F

. The term
332
R
is the Coulombic interaction and
Ae
bR
+CR
9
is the short range repulsive potential.
The ground state potential energy surface E
g
= M = D[e
2a(RR
0
)
2e
a(RR
0
)
] is represented
by a Morse potential M. Parameters D and R
0
can be obtained from the experimental bond-energy
104
and bond-length. The parameter a can be adjusted to reproduce the vibrational frequency of the
diatomic molecule. The parameter

D
HF
=

D
HH
D
FF
and = 0.05

A. Parameters A and C are

adjusted so that the minimum energy of H
22
corresponds to the H-F bond-length (i.e., the sum of
ionic radii of H and F). This empirical parametrization allows us to solve Eq. (66) for H
12
,
H
12
=
_
(H
11
M)(H
22
M),
and obtain the Hamiltonian matrix elements in terms of empirical parameters.
Conclusion: Potential energy surfaces parametrized by a few empirical parameters are able to
describe bonding properties of molecules associated with atoms of different electronegativity.
Dipole Moment
The dipole moment is one of the most important properties of molecules and can be computed as
follows,

g
=<
g
| |
g
>,
where
=

i
er
i
+

j
ez
j
R
j
.
The rst term of this equation involves electronic coordinates r
i
and the second term involves
nuclear coordinates R
j
.
For example, the dipole moment of HF can be computed as follows,

g
= C
2
1
<
1
| |
1
>
. .
0
+C
2
2
<
2
| |
2
>
. .
eR
0
+2C
1
C
2
<
1
| |
2
>
. .
0
,
since
1
represents a covalent state and the overlap between
1
and
2
is assumed to be negligible.
The dipole moment is usually reported in Debye units, where 4.803 Debye is the dipole moment
of two charges of 1 a.u. with opposite sign and separated by 1

A, from each other.
40.1 Exercise 56
Evaluate the dipole moment for HF using the following parameters for the semiempirical model of
HF potential energy surfaces (energies are expressed in kcal/mol, and distances in

A),
D=134;

D=61; R
0
=0.92; a=2.27;
A=640; b=2.5; C=20; I=313; EA=83.
Polarization
The electric eld of an external charge z located at coordinate R
0
along the axis of the molecule
does not affect the energy of the covalent state H
11
, but affects the energy of the ionic state H
22
as
105