Σ

ΣE = - (qa +qb)Σab
ab
Sab + ΣΣR
k<l
QkQl
kl
+ ΣΣ ΣΣ
occ

r
. unocc.

s
-
occ. unocc.

s r
2(Σab csbΣab)2
cra
Er - E
s

Closed-Shell Repulsion
Coulomb Interaction FrontierOrbitalInteraction

Chem 1140; Molecular Modeling
• Molecular Mechanics
• Semiempirical QM Modeling
• CaCHE

   
 A. Modeling Software

Overview:
http://cmm.info.nih.gov/software.html

- CaCHe: Structure building. Extended

MM2(87) energy minimization, extended
Hückel, MOPAC, ZINDO MO programs.
Orbital electron density and electrostatic
maps. Prediction of octanol/water
partition coefficient and water solubility
   
 B. Introduction to Molecular Modeling

Objectives:
- Computer graphics visualization of molecules
("Dreiding models")
- Matching (overlays, docking) of molecules
- Use of empirical force-fields to determine molecular
properties as well as interatomic distances.
- Correlate molecular properties with an electronic
structure from ab initio quantum mechanics or
semiempirical quantum mechanics.
- Gain information on dynamic molecular movements.
- Use computer-assisted design for molecular
recognition in organic, bioorganic, and medicinal
 chemistry and material
  science.
   
Schematic of molecular force field expression. Diagonal terms refer to interactions that can be expressed as
functions of a single internal coordinate, whereas cross terms introduce coupled interactions involving two or more
coordinates.
Bond stretch:

   
Dihedral potentials

The traditional expression for the dihedral potential energy is a cosine series:
 Computational Options

Single point calculation. The energy of the system at a specified configuration.

   
In local methods forfinding a minimum effort is directedtoward a nearby local minimum, and steps that
mayincrease the functi on are not pe
rmitted.

Descen
t struct
ure of loca
l minimization algor
ithms:
The local minima problem
The search for the global minimum1

Generally, we are most interested in the preferred (most stable) conformation of our molecules and the conformers
which are s ignificantly populated at r oom temperature (within approx. 3 kcal/mol of the global minimum). A
number of conformational search techniques have been developed to answer these questions.

Structures of local and global minimization algorithms:

1 Saunders, M.et al.J. Am. Chem. Soc. 1990, 112, 1419

Σ - Grid search; For very simple structures, this can be do ne manually. (MM2/3: angle driver to create
reasonable starting geometries for conformational searching).

H
H

H H

This systematic proced ure providesguaranteedcoverage of all regions of spaceif a sufcient numberof
start
ing pointsis selected(atleast3 perdihedr al angle).
Problems: an extraordinary large number of start ingstructuresis generated,which renders he
t problem very
computer-time intensive, espec ially for flexible structures.For example, for polypeptidesof n residues, a rough
range of 10n to 25n reasonable starting pointsby coarse subdomain part
ition of back
bone and sidechains have tobe
investigated.
If the numberof values of for each torsion is reduced,the resolution with which spaceis coveredmay be
inadeq uatetoprovidea start inggeometry in the vicinity of each minimum .
Newer molecular modeling packages offer several force fields. Choose wisely!

MM2: Organic molecules. Last release: 1987.1

MM3: Organic molecules. Last release: 1994 (The MM3* on Macromodel is the MM3/91 release and is
limited to fewer functionalities).
AMBER: Proteins and nucleic acids, carbohydrates.2
OPLS: Peptides. The united-atom simulation treats certain atom groups as one atom, with one atom type.3
BIO+: (CHARMM) Biological macromolecules.4

1 Allinger, N. L. et al.J. Am. Chem. Soc. 1992, 114, 1.

2 Kollman, P. A. et al.J. Comp. Chem. 1986, 7, 230.
3 Jorgensen, W. L. et al. J. Am. Chem. Soc. 1991, 113, 2810.
4 Karplus, M.et al.J. Comp. Chem. 1983, 4, 187.
 (1 a.u. = 627.510 kcal/mol)

Force Field Eq. (kcal/mol) Ax. (kcal/mol) (kcal/mol)

MM3/92 9.10 10.87 1.77

Chem3D (MM2) 6.90 8.67 1.77
CaCHe (MM2?) 6.89 8.68 1.79
MM2* 6.89 8.67 1.78
MM3* 9.10 10.87 1.77
Amber* 1.84 2.87 1.03
OPLS* 0.78 2.57 1.79
Sybyl (tripos) 1.30 2.70 1.40
Sybyl (Spartan) 1.30 2.70 1.40
MMFF (Spartan) 0.70 2.07 1.37
Exp. 1.74
MNDO -36.19 -35.16 1.03
AM1 -43.70 -42.28 1.42
PM3 -36.87 -35.75 1.12
STO-3G E(HF) -270.063379 a.u. -270.060444 a.u. 1.84
6-31G* E(HF) -273.2436640 a.u. -273.2399991 a.u. 2.30
6-31+G* E(HF) -273.245892 a.u. -273.242122 a.u. 2.37
6-311+G** E(HF) -273.308789 a.u. -273.305906 a.u. 1.81
 
DFT/LSDA/VWN  
-272.6182101 a.u. -272.6149537 a.u. 2.04
The ab initio calculation at the 6-311+G** E(HF) level required 1,800 x more time than MM2.
 C. Semiempirical Methods

Molecular mechanics methods are based on classical

concepts and require computer time roughly proportional to
the square of the number of atoms. In contrast,
semiempirical methods at the Hartree-Fock level use a
combination of quantum chemical models and experimentally
determined parameters to strive for accuracy and scale up as
(4N)3. Finally, ab initio quantum mechanics proceeds as
(10N)4 (for glucose: 1:1,500:6,000,000).

There are important differences between classical molecular

mechanics and quantum mechanical methods. MM methods
are extremely fast and able to handle very large systems.
Particularly for hydrocarbons, they are also as accurate as
  best ab initio methods.
the   But they are only parametrized for
ground state systems and common bonding situations
(functional groups). MM methods are unable to anticipate the
making and breaking of most bonds, the electronic properties
Use molecular orbital methods to compute:

- bond orders
- dipole moments
- ionization potentials
- vibrational frequencies and IR, UV spectra
- MO energies
- partial charges
- potential energy maps
- reaction pathways
- transition states
- heats of formation
- mapping bond breaking and bond formation

The transition between semiempirical and ab initio

 
quantum chemistry  is not always obvious. Basis sets, for
example, are empirical in nature, as are effective core
potentials.
Approximate solutions of the Schrödinger equation:

H =E 1

The MO-LCAO approximation:

Σ
i
=
∑
C
ν i
Φ
ν
 
ν

The overlap between two atomic orbitals   and   defines the elements of a matrix S: 

 
The  Hartree­Fock  method  introduces  a on e­electron  Hamiltonian,  F,  which  is  the  sum  of  the  kinetic  energy 
operator for an electron, a potential that a single electron would feel coming from the fixed nuclei, and an average 
of the effects of all the other N­1 electrons: F i = E i.  
The Fock equations can be most easily be solved by solving the matrix equations: FC = SCE. 
 
The following steps are then common to all molecular orbital programs: 
 
1. Calculate the integrals needed to form the Fock matrix F. Guess  i's. 
2. Calculate the overlap matrix S. 
3. Diagonalize S.  S = Φ Φ d τ

4. Diagonalize F for the MO eigenvalues E. 
φ
μ ν ∫ μ ν
φ
μ ν

4. Solve the matrix equation to obtain the MO coefficients C. 

                                                 
1 "The underlying physical laws neces
saryfor themathematicaltheory of a large part of physicsand thewhole of chemistryare
thus o
cmpletelyknown, andthe difficultyis onlythattheexactapplication of the
se laws leadsto equationsmuchtoocomplicat ed
to be sol P. A. M.Proc. Roy. Soc. (London) 1929, 123, 714.
uble". Dirac,
Semiempirical methods:

The neglect of differential overlap approximation says that S << 1 in many situations and that the two electron
integrals can be ignored or approximated in various ways (one-center terms). A second basic approximation is the
neglect of diatomic differential overlap: Σ
A

Σ
Φ  (two­center terms). 
B

ν
= Φ
A

μ
Φ
ν
A

δ
AB

 
The  greatest  simplification  is  achieved  with  complete  neglect  of  differential  overlap  (CNDO),  f.  ex.  CNDO/S. 1 
Because  INDO  calculations  take  very  little  more  time  than  CNDO  calculations,  they  are re placing  the  CNDO 
models. 

The intermediate neglect of differential overlap (INDO) schemes keep all of the one-center terms above: INDO/S1
(spectroscopy of transition metal complexes; it was used on the reaction center of photosynthetic bacteria).
MINDO/3 was de veloped in 1975 to reproduce such diverse experimental properties as molecular geometries,
heats of formation, dipole moments, and ionization potentials. All quantities in th e Fock matrix and the energy
expression were treated as free parameters, and the resonance integral was chosen to be a pair parameter rather than
the average of two atomic parameters. H, B, C, N, O, F, Si, P, S, and Cl are parametrized.

                                                 
1 Mataga, N.; Ishimoto, K.Z. Physik. Chem. (Frankfurt)1957, 13, 140.
MNDO, 1 AM1,2 and PM33 are all NDDO techniques developed by Dewar and his coworkers and use essentially
all the same model functions. MNDO has been parametrized for H, Li, B, C, N, O, F, Al , Si, P, S, Cl, Zn, Ge, Br,
Sn, I, Hg, and Pb. It uses monoatomic parameters and is computationally quite efficient but failed for biologically
interesting molecules. AM1 has been parametrized for H, B, C, N, O, F, Al, Si, P, S, Cl, Zn, Ge, Br, I, and Hg, and
PM3 for H, Be, C, N, O, F, Mg, Al, Si, P, S, Cl, Zn, Ga, Ge, As, Se, Br, Cd, In, Sn, Sb, Te, I, Hg, Tl, Pb, and Bi
(the third- and fourth-row main-group elements have only recently been added). PM3 and AM1 use a different
equation for the core-core repulsion term (this is a problem for AM1 with P, Al, and N). PM3 differs from AM1 in
that the former treats the one-center, two-electron integrals as pure parameters, as opposed to being derived from
atomic spectroscopy. Spartan’s PM3(tm) model is parametrized for Ti, Zr, Hf, Ta, Cr, Mo, W, Co, R h, Ni, and Pd,
used in conjunction with PM3 for non-transition metals.

AM1:
- seems to be most accurate for calculations of dipole moments for elements that it has been parametrized
for.
- reproduces hydrogen-bonds slightly better and
- gives energies that are still too large for activation barriers.
- reproduces ground-state geometries quite accurately.
- is poor in reproducing lon-pair lon-pair repulsion.
- has an average error of heats of formation of C, H, N, O compounds of ca. 40%.

1 Dewar, M.J. S. et al.J. Comp. Chem. 1981, 2, 433.

2 AustinModel 1; Dewar, M.J. S. et al.J. Comp.Chem. 1981, 2, 433.
3 c Method Number 3. Stewart,J. J. P. J. Comput.Chem. 1989, 10, 221. This representsthethird parametri
Parametri zation of
MNDO (AM1 beingthe second).
 Characterizations of the wave function

In a se mi-empirical calculation, the energy, the first derivatives of the energy and the wave function for the
molecule are computed. Other molecular properties are basically an average over the wave function of certain
operators describing the properties.
Among the visualizable properties that are generally computed are:

Total electron density represents the probability of an electron being at a particular point in space.

Spin density defines the excess probability of finding spin-up over spin-down electrons at a point in space. The
spin density at the position of a nucleus is the a prime determinant of ESR spectra.

Orbital plots are used to vi sualize molecular orbitals, especially frontier orbitals, for the analysis of organic
reactions.

Electrostatic potentials can be used to predict initial attack positions of ions during a reaction. The potential of a
positive probe charge is calculated as a function of the distance to the molecule, the nuclear charge, and the
electronic density.

Atomic partial charges give a crude idea for the distribution of charges in a molecule.

IR, UV/VIS absorption spectra.

Kondru, R. K.; Wipf, P.; Beratan, D. N., "Theory assisted determination of absolute stereochemistry for
complex natural products via computation of molar rotation angles." J. Am. Chem. Soc. 1998, 120, 2204-
2205.

Wipf, P.; Kerekes, A. D., "Structure reassignment of the fungal metabolite TAEMC161 as the phytotoxin
viridiol." Journal of Natural Products 2003, 66, 716-718.