Semiempirical

Molecular Orbital Theory

Slater-determinantal wave func:ons

and Hartree-Fock theory

Video III.iv
 Constructing a 1-Electron Wave Function
The units of the wave function are such that its square is electron per
volume. As electrons are quantum particles with non-point distributions,
sometimes we say density or probability density instead of electron per
volume (especially when there is more than one electron, since they are
indistinguishable as quantum particles)
For instance, a valid wave function in cartesian coordinates for one electron
might be:

2Z 5 / 2 % 2 2 2 ( $Z x 2 +y 2 +z 2 / 3
"( x, y, z; Z ) = ' 6 $ Z x + y + z * ye
81 # & )

normalization radial phase cartesian ensures

factor factor directionality square
(if any) integrability
 Constructing a 1-Electron Wave Function
A valid wave function in cartesian coordinates for one electron might be:

2Z 5 / 2 % 2 2 2 ( $Z x 2 +y 2 +z 2 / 3
"( x, y, z; Z ) = ' 6 $ Z x + y + z * ye
81 # & )

normalization radial phase cartesian ensures

factor factor directionality square
(if any) integrability

This permits us to compute the

probability of finding the electron
x1 x x 2 within a particular cartesian volume
x2 y2 z2 2 element (normalization factors are
P y1 y y 2 =
x1 y1 z1
dx dy dz determined by requiring that P = 1
when all limits are infinite, i.e.,
z1 z z 2
integration over all space)
 Constructing a 1-Electron Wave Function
To permit additional flexibility, we may take our wave function to be a linear
combination of some set of common basis functions, e.g., atomic orbitals
(LCAO). Thus
N
( r ) = a ( r)
i
i=1

For example, consider the wave
function for an electron in a CH bond. C
It could be represented by s and p
functions on the atomic positions, or s
functions along the bond axis, or any
other fashion convenient. H
 Constructing a 1-Electron Wave Function
To optimize the coefficients in our LCAO expansion, we use the variational
principle, which says that

a i i H a j j dr
a a H
i j i j dr a a i j H ij

i j ij ij
E= = =

a a i j i j dr a a i j S ij
a i i a j j dr

ij ij
i j

which is minimized by one root E of the secular equation (the

other roots are excited states)each value of E that satisfies the

secular equation determines all of the ai values and thus the shape
of the molecular orbital wave function (MO)

H11 ES11 H12 ES12 H1N ES1N

H21 ES21 H22 ES22 H2 N ES2 N
= 0

HN1 ESN1 H N2 ES N2 HNN ESNN
 What Are These Integrals H?
The electronic Hamiltonian includes kinetic energy, nuclear attraction, and, if
there is more than one electron, electron-electron repulsion

nuclei electrons 2
1 Zk
H ij = i 2 j i
2
rk
j + i rn
j
k n

The final term is problematic. Solving for all electrons at once is a

many-body problem that has not been solved even for classical
particles. An approximation is to ignore the correlated motion of the
electrons, and treat each electron as independent, but even then, if
each MO depends on all of the other MOs, how can we determine
even one of them? The Hartree-Fock approach accomplishes this
for a many-electron wave function expressed as an
antisymmetrized product of one-electron MOs (a so-called Slater
determinant)
Choose a basis set

Choose a molecular geometry q(0) The Hartree-

Compute and store all overlap,
Fock procedure
one-electron, and two-electron Guess initial density matrix P(0)
integrals

Construct and solve Hartree-

Fock secular equation

Construct density matrix from

F11 ES11 F12 ES12 F1N ES1N
Replace P(n1) with P(n) occupied MOs F21 ES21 F22 ES22 F2 N ES2 N
= 0

Is new density matrix P(n) FN1 ES N1 FN2 ES N2 FNN ESNN
no
sufficiently similar to old
Choose new geometry density matrix P(n1) ?
1 nuclei 1
according to optimization
algorithm yes
F = 2 Zk
2 k rk
Optimize molecular geometry?
1
no
yes no
+ P ( ) ( )
2
Does the current geometry occupied
Output data for
satisfy the optimization
criteria? unoptimized geometry P = 2 ai ai
i
yes 1
( ) = (1) (1) r12
( 2) (2)dr (1)dr (2)
Output data for optimized
geometry One MO per root E
 Cement Your Understanding

1 nuclei 1
F = 2 Zk
2 k rk
1
+ P ( ) ( )
2

Try to write out, neatly and carefully, in full

mathematical notation, every generic integral in the
Fock matrix for Slater-type orbital basis functions with
all functions and operators expressed in Cartesian
coordinates (which is how they are most typically
evaluated in actual practice)
 Semiempirical
Molecular Orbital Theory

Semiempirical simplica:ons
of the Fock matrix (early)

Video III.v
Choose a basis set

Choose a molecular geometry q(0) The Hartree-

Compute and store all overlap,
Fock procedure
one-electron, and two-electron Guess initial density matrix P(0)
integrals

Construct and solve Hartree-

Fock secular equation

Construct density matrix from

F11 ES11 F12 ES12 F1N ES1N
Replace P(n1) with P(n) occupied MOs F21 ES21 F22 ES22 F2 N ES2 N
= 0

Is new density matrix P(n) FN1 ES N1 FN2 ES N2 FNN ESNN
no
sufficiently similar to old
Choose new geometry density matrix P(n1) ?
1 nuclei 1
according to optimization
algorithm yes
F = 2 Zk
2 k rk
Optimize molecular geometry?
1
no
yes no
+ P ( ) ( )
2
Does the current geometry occupied
Output data for
satisfy the optimization
criteria? unoptimized geometry P = 2 ai ai
i
yes 1
( ) = (1) (1) r12
( 2) (2)dr (1)dr (2)
Output data for optimized
geometry One MO per root E
 nuclei occupied
1 2 1
F = Zk P = 2 ai ai
2 k rk i
1
+ P ( ) ( )

2 i = a
i

F11 ES11 F12 ES12 F1N ES1N

F21 ES21 F22 ES22 F2N ES2N
= 0

FN1 ES N1 FN2 ES N2 FNN ESNN

Consider only rst- and second-row atoms; consider a minimal basis set

1
( ) = (1) (1) r ( 2) ( 2) dr1dr2
12
 nuclei occupied
1 2 1
F = Zk P = 2 ai ai
2 k rk i
1
+ P ( ) ( )

2 i = a i

1
( ) = (1) (1) r ( 2) ( 2) dr1dr2
12
Things to consider in a semiempirical model:

1) What shall I choose for a basis set?
2) What shall I do for overlap integrals (in the secular determinant)?
3) What shall I do for Fock matrix elements?
4) As part of (3), what shall I do about electron repulsion integrals?

Essen:ally all semiempirical models adopt a minimal basis set. The AO basis
func:ons cover only the valence s, p, d, and f orbitals (p, d, and f only as needed),
one func:on per orbital (so, 1 s, 3 p, 5d, and 7f orbitals). The basis func:ons
themselves are taken to be Slater orbitals. n +1/ 2

STO ( r, , ;,n,l,m) =
( 2) n 1 r m
e Yl ( , )
1/ 2 r
[(2n)!]
 nuclei occupied
1 2 1
F = Zk P = 2 ai ai
2 k rk i
1
+ P ( ) ( )

2 i = a i

1
( ) = (1) (1) r ( 2) ( 2) dr1dr2
12

General choice for overlap integrals, S:

The overlap between any two STOs can be computed analy:cally (i.e., it is a simple
arithme:c func:on of posi:on and values of , n, l, and m). Overlap integrals nd
use in compu:ng some Fock matrix elements.

However, when forming the secular determinant, the simplica:on S = is

adopted.

n +1/ 2

STO ( r, , ;,n,l,m) =
( 2)
r n 1e rYlm ( , )
1/ 2
[(2n)!]
 nuclei occupied
1 2 1
F = Zk P = 2 ai ai
2 k rk i
1
+ P ( ) ( )

2 i = a
i

1
( ) = (1) (1) r ( 2) ( 2) dr1dr2
12

There are 3 types of Fock matrix integrals, F:

A Fock matrix element can be diagonal (i.e., = ), or not. When the matrix
element is o-diagonal, it can be monatomic (i.e., , but and are on the
same atom), or it can be diatomic (the only remaining possibility)

As every Fock matrix element includes ERIs, rst lets consider the ERIs (well get to
one-electron integrals later)
 COMPLETE NEGLECT OF DIFFERENTIAL OVERLAP
1
Reducing the N scaling problem:
4
( ) = (1) (1)
r
( 2) ( 2) dr dr 1 2
12

CNDO: ( | ) = ( | ) only ERIs not equal to zero are Coulomb

integrals between 2 AO basis func:ons, which
may or may not be on the same atom

( | ) = AB is a number or func:on that depends only on

the atoms A and B on which and are
found, respec:vely.
AA = IPA EAA interac:on between electrons on same atom
taken as dierence between ioniza:on
poten:al and electron anity
one extra e-e repulsion
AA + BB IP on atom M
AB =
2 + rAB ( AA + BB )
M + M M + M
func:on above has proper limits at rAB = 0
(since then A must be B, the same atom)
and at rAB very large ( just 1/rAB, which is EA
Coulombs law in atomic units)
 COMPLETE NEGLECT OF DIFFERENTIAL OVERLAP
1
Reducing the N scaling problem:
4
( ) = (1) (1)
r
( 2) ( 2) dr dr
1 2
12

CNDO: ( | ) = ( | ) only ERIs not equal to zero are Coulomb

integrals between 2 AO basis func:ons, which
may or may not be on the same atom

( | ) = AB is a number or func:on that depends only on

the atoms A and B on which and are
found, respec:vely.

no dis4nc4on in 2-electron repulsions:

!CC !NN
H H
H H H : : H
C !CC C
!NN
N N
H : H :
H H H H
 Semiempirical
Molecular Orbital Theory

Semiempirical simplica:ons
of the Fock matrix (subsequent)

Video III.vi
 COMPLETE NEGLECT OF DIFFERENTIAL OVERLAP
1
Reducing the N scaling problem:
4
( ) = (1) (1)
r
( 2) ( 2) dr dr
1 2
12

CNDO: ( | ) = ( | ) only ERIs not equal to zero are Coulomb

integrals between 2 AO basis func:ons, which
may or may not be on the same atom

( | ) = AB is a number or func:on that depends only on

the atoms A and B on which and are
found, respec:vely.

no dis4nc4on in 2-electron repulsions:

!CC
H H
C !CC C

H H
 INTERMEDIATE NEGLECT OF DIFFERENTIAL OVERLAP
1
Reducing the N scaling problem:
4
( ) = (1) (1)
r
( 2) ( 2) dr dr 1 2
12

INDO: ( | ) = ( | ) s:ll taken to be true unless and are on the

same atom
( ss ss) = G ss
For an atom with only s and p valence
( ss pp) = G sp func:ons, 5 unique integrals are kept (they are
simply values, in energy units, that can be
( pp pp) = G pp
determined from spectroscopy or taken as
( pp p p) = G pp free parameters). Note that p refers to a p
orbital dierent from a reference one, e.g.,
( sp sp) = L sp
( pp | pp ) might be (pxpx | pypy ), i.e., the
through-space repulsive interac:on between
electrons in two perpendicular p orbitals

This extra exibility in the atomic ERIs tends to improve rela:ve energies
for electronic states that dier by occupa:on of orbitals on single-centers.
That is, for instance, excited states that are well characterized as one-
electron local excita:ons. INDO s:ll nds substan:al use for spectroscopy.
 NEGLECT OF DIATOMIC DIFFERENTIAL OVERLAP
1
Reducing the N scaling problem:
4
( ) = (1) (1)
r
( 2) ( 2) dr dr
1 2
12

NDDO: ( | ) = A X B Y ( | ) so, the only integrals that survive are those

where both and are on the same atom,
( ss ss) = Gss and and are on the same atom, but
neither the two atoms need to be the same,
( ) sp
ss pp = G nor do the two func:ons on the individual
( pp pp) = Gpp atoms need to be the same. For monatomic
ERIs, the INDO-like values are retained. For
( pp p p) = Gpp diatomic ERIs, there are 100 possible integrals
( sp sp) = Lsp for s and p basis func:ons (try and verify
that)

The diatomic integrals are not easily evaluated for Slater type orbitals.

Instead, NDDO methods model the integrals as classical mul:pole
interac:ons between the two atoms. Whether the mul:pole is a point
charge (ss), a dipole (sp), or a quadrupole (pp or pp) depends on the
nature of the orbitals, and the orienta:on of the higher mul:poles does,
too. The magnitudes of the dipoles and quadrupoles depend on the
exponents of the Slater basis func:ons.
 ONE-ELECTRON INTEGRALS
nuclei
1 2 1
Zk
2 k
rk

Consider diagonal term, = :

1 2 1 nuclei
1
Zk Zk
2 rk k k
rk
nuclei

= U Z ( s s )
k k k
k k

where U is a parameter that should be nearly equal to the ioniza:on poten:al of

an electron in atomic orbital and the alrac:on to a nucleus at the posi:ons of
other atoms k is Zk :mes the repulsion with an electron in an s orbital on the same
atom.
 ONE-ELECTRON INTEGRALS
nuclei
1 2 1
Zk
2 k
rk

Consider o-diagonal term, , both on same atom

1 2 1 nuclei
1
Zk Zk
2 rk k k
rk
nuclei

= Z ( s s )
k k k
k k

where the rst two terms are zero for s and p orbitals by symmetry considera:ons
and the alrac:on to a nucleus at the posi:ons of other atoms k is Zk :mes the
repulsion with an electron in an s orbital on the same atom.
 ONE-ELECTRON INTEGRALS
nuclei
1 2 1
Zk
2 k
rk

Consider o-diagonal term, , each on a dierent atom

( + )S

=
2
where is our old friend from Hckel theory, the resonance integral (a numerical
parameter) and S is the overlap matrix computed from the Slater type orbitals.
Note that the values depend on both the atomic number and the orbital (i.e.,
carbon has a dierent value for an s orbital than for a p orbital)

What would a table of parameters look like for an NDDO model?

 Semiempirical
Molecular Orbital Theory

Semiempirical parameteriza:ons
and modern usage

Video III.vii
 Big Personali:es

John Pople Michael Dewar

 FINAL MOLECULAR-MECHANICS-LIKE TOUCHUP
First adopted in MNDO, later in AM1 and PM3, etc.

VN ( A,B) = Z A Z B sA sA sBsB ( )
Z A Z B 4 b A,i ( rAB c A,i )
2
b B,i ( rAB c B,i )
2

+
rAB i=1

aA,ie + aB,ie

where the rst term says that nuclei repel each other propor:onal to the degree
that their s electrons repel one another, but the following sum eec:vely creates
gaussian shaped ripples, posi:ve or nega:ve, in the poten:al energy surface
surrounding the atoms. Such an approach permits a ne tuning of atomic
separa:ons that must be considered rather ad hoc, but not necessarily
inconsistent with semiempirical principles. (Later extended by Jorgensen and co-
workers using Pairwise Distance Directed Gaussians (PDDG) for MNDO and
PM3.)
 Chemical Intui:on vs Maths
Observables Occurrences 2 1 / 2

Z =
( )
calci, j expt i, j
wi2
i j

Michael Dewar Jimmy Stewart

MNDO MNDO
AM1 AM1
PM3
AMPAC MOPAC
Errors in Computed Heats of Forma:on (kcal/mol)
 MODERN USAGE? SPECIFIC RANGE PARAMETERS
Parameter AM1 AM1-SRP
H H
O
C C O + H + H2
H H C
Us 52.03 49.85 H H2O H H

Up 39.61 40.34
!s 15.72 16.91
!p 7.72 9.19
O
Us 97.83 99.18
Up 78.26 80.76
!s 29.27 29.00
!p 29.27 29.25

Source "Erxn De(CH) De(HH)

AM1 28.0 81.4 109.4

AM1-SRP 4.9 104.4 109.4
Expt. 5.1 104.4 109.5