Introduction to Computational Chemistry

Computational (chemistry education)

and/or

(Computational chemistry) education

– First one: Use computational tools to help
increase student understanding of material
already covered in various courses

– Second one: Teach students about

computational chemistry (molecular modeling)
itself, in both courses and research projects
1

Guiding Questions
1. What is the role and purpose of molecular
modeling?
2. What is the fundamental mathematical
expression that needs to be solved? What
are the terms, what is their significance, and
what variations can be used?
3. What are the pros/cons of approximations
that can be used in the calculations? How
does the choice of approximation affect the
results, computing time, etc.?
2

Guiding Questions
4. There are four different methods commonly
employed: molecular mechanics, semi-
empirical, ab initio, and density functional
theory (DFT). What are these methods?
How do they differ?
5. What are the fundamental units of measure
commonly used by computational chemists?
6. What are some of the computer codes and
platforms used to do computational
chemistry? What are their pros/cons?
3

1
 Why Computational Chemistry??
• In 1929, P.A.M. Dirac wrote:
– “The underlying physical laws necessary
for the mathematical theory of . . . the whole
of chemistry are thus completely known,
and the difficulty is only that the exact
application of these laws leads to equations
much too complicated to be soluble.”

– Dirac didn’t have access to digital

computers, but we do!
4

Molecular Modeling Defined

• Provides information that is complementary to
experimental data on the structures, properties,
and reactions of substances
• Mainly based on one algorithm: Schrödinger’s
Equation
• Until recently, required the use of high
performance computers (architecture)
• Modern PC’s are now of sufficient power to
run many molecular modeling codes
• Everyone now has access to this tool!!
5

Chemistry Today: A Different View

Old Way New Way
Desired
Properties
e

Sim
lat

ret
e

Me
rr

ula
erp
Co

asu

te
t
In

re

Structural Synthesize
Compounds
Design
Build

2
 Some Molecular Modeling Codes
• CAChe (Computer Aided Chemistry)
• Spartan
• HyperChem
• PC Model
• Chem3D
• Gaussian
• Sybyl
• MOPAC (Molecular Orbital Package)
• GAMESS (General Atomic and Molecular
Electronic Structure System)
7

1. Molecular Mechanics
• Apply classical mechanics to molecules
→ No electrons, no orbital interactions!!
– Atoms are spheres with element dependent mass
– Bonds are springs that obey Hooke’s Law:
F = -kx
where k is the force constant (for a specified bond
type between certain atoms)
– Other types of springs represent bond angles,
dihedral angles, etc.
8

Molecular Mechanics: Components

• Bond stretching (l)
• Bond Angle bending (θ)
• Dihedral Angle rotation (Ф)
• Van der Waals forces H
O O
• Hydrogen bonding
• Electrostatic interactions
• Others
9

3
 Equations: Bond Stretching
• Use the Harmonic Oscillator Approximation:
ks
E HOA =
2
( l − l0 )
2

ks = force constant
l0 = equilibrium bond length

– Could include higher order terms:

E HOA =
ks
2 [ ]
(l − l0 )2 1 − k ' (l − l0 ) − k '' (l − l0 )2 − k ''' (l − l0 )3 − ...
– MM2 (cubic); MM3 (quartic)
10

Equations: Bond Angle Bending

• Mathematically similar to stretching:

1
k (θ − θ0 )
2
Eθ =
2 θ
where θ0 is the equilibrium angle

– Again, cubic and quartic terms give better fit with

experimental values, but may not be included
because of computational cost

11

Equations: Dihedral Angle Rotation

• Use a sum of periodic functions:
Etorsion = 0.5 V1 (1 + cos φ ) + 0.5 V2 (1 + cos 2φ ) +
0.5 V3 (1 + cos 3φ ) + .....?
where Vn is the dihedral force constant,
n is the periodicity, and φ is the dihedral angle

– Various contributions to the energy may have

different periodicities
• Contributions may include dipole-dipole
interactions, hyperconjugation, van der Waals, H-
bonding, etc.
12

4
 Equations: Van der Waals & H-bonding
• Lennard-Jones, or “6-12” potential:
A B A = repulsive term
EvdW = 12 − 6
r r B = attractive term

• Hydrogen Bonding A B
E HB = 12
− 10
r r
– Often handled in the van der Waals and
electrostatic terms
– Sometimes, explicitly placed in a separate term
• Called a “10-12” potential: Attractive region
decays more rapidly with distance
13

Equations: Electrostatic Energy

• Based on Coulomb’s Law: qa qb
Eelectro =
ε ab rab
qn = (partial) charge on atom
ε = dielectric constant
r = interatomic distance
Atom Charges:
1. Could be fixed values (easiest)
2. Values could also be calculated
• Related to electronegativity of atom, as well as
those atoms connected to it
14

Molecular Mechanics: Overall Energy

• Also called the “steric” energy in MM
– Summation of all the terms:

Esteric = Estretch + Ebend + Etorsion + EvdW

+ E H − bonding + Eelectro + Eother
– The collection of all the functional forms and the
associated constants is called a force field
– BEWARE: “Energies” reported by MM are often
meaningless – are not externally referenced
• May be useful for conformers of same molecule
15

5
 Parameters
~100 elements: N(N+1)/2 = 5050 single bonds
– Multiple bonds: Define atom hybridizations
~300 atoms types: → 45,150 force constants!
• Need 300 partial atomic charges; 45,150 values of
ε (electrostatics); 45,150 values for A & B (van
der Waals); 45,150 values of l0, just to cover bond
stretching!
• Also need kθ, θ0, Vn, and φ values for all possible
angles, A & B values, etc., etc.
• To be thorough, would need ~108 parameters
gathered from experimental data!!

16

Molecular Mechanics
• Advantages: Very fast, excellent structural
results (for compounds with good parameters
available), computationally inexpensive (can
be applied to large molecules)
– Geometry optimization: Move all atoms until
sum of all forces on each = 0

• Disadvantages: ~80% of known compounds

do not have parameters available
– No orbital information, can’t look at reactions or
transition states, can’t predict reactivity, etc.
17

2. Ab Initio (Hartree-Fock) Method

• Based on Schrödinger’s Equation: ĤΨ = EΨ
• Ĥ is the “Hamiltonian operator”
• E is the energy of the atom or molecule
• Ψ is the wavefunction, which is what we want
h2 2 h2 2 e 2 Zk
H=− ∑i
∇ −
2me i ∑ k
∇ −
2mk k ∑∑ i k
rik
2 2
e e Z k Zl
+ ∑r +∑
i< j ij k< l
rkl
18

6
 Approximations Used
1. Born-Oppenheimer: Compared to e-’s,
nuclei are stationary
• Electrons move in a field of fixed nuclei
2. Hartree-Fock: Separate Ψ (many-electron
wavefunction) into a series of one-electron
spin orbitals
3. LCAO (Linear Combination of Atomic
Orbitals): MO’s expressed as linear
combinations of single electron atomic
orbitals, represented by basis functions
19

MO Construction
N
• Individual MO is defined as:
φ= ∑a ϕ
i= 1
i i

Basis set = set of N functions φi, each associated

with a molecular orbital expansion coefficient ai

• Variational Principle:
– We know that the set of all Ei will have a lowest
energy value (ground state)
– Try different ai’s until energy minimum located
• At this point, we have found the ground state
20

Roothaan-Hall Equations
→ Used to find the set of molecular orbital
expansion coefficients that minimize the energy
• The solution process is iterative
Process overview:
1. Select a basis set (φi) and a molecular geometry
2. Guess a set of ai’s defining the Fock matrix
3. Solve Roothaan-Hall equations to get new ai’s
and a new Fock matrix
4. Repeat until a self-consistent solution is found
(HF-SCF Method)
• This energy minimum is the ground state
21

7
 Ab Initio Method
• Advantages: Ψ is often “close enough” so
properties can be calculated via application of
the appropriate operator
– Useful results can be obtained
– Can serve as a stepping stone to more advanced
treatments

• Disadvantages: e-/e- repulsion is over-

estimated; Computationally expensive;
Correlation energy is significant; Difficult to
model reactions involving bond dissociation
22

Difficulties of HF Method
• N4 total integrals need to be evaluated
N = number of basis functions used
Computationally expensive
– Limited to study of smaller molecules

• Theoretical limitations
– HF treatment: Each electron experiences the
others as an average distribution; no instantaneous
e-/e- interaction is included
• Energies are often far from experimental values
– One electron nature of the operators used
23

Higher Level Methods

• Called electron correlation, or post-SCF
methods
– Deal with the electron correlation problem
• CI (Configuration Interaction)
– Extreme computational cost, but provides the
most complete treatment of a molecular system
possible
• Møller-Plesset Perturbation Theory – MP2,
MP3, etc.
• Computational cost is again very high
24

8
 Overcoming Limitations of HF-SCF Method
1. Since several approximations are made,
introduce further simplifications to increase
speed and accuracy
– Replace some integrals with parameterized
values that will reproduce experimental results
→ Constitutes “Semiempirical” techniques
2. Don’t worry about Ψ at all
– Instead, focus on the electron density
→ Density Functional Theory (DFT)
25

3. Semiempirical Methods
• Simplified Hartree-Fock:
– Want to avoid the computational time needed to
solve N4 integrals

– Introduce approximations that make HF theory

more tractable AND somehow better account for
correlation energy
• Improved chemical accuracy

– Different ways of doing this define the various

semi-empirical methods

26

Simplifications
1. Only look at valence electrons
– Core electrons subsumed into nucleus

2. Neglect certain integrals

3. Parameterize other integrals using
experimental data
– Different methods are parameterized to
reproduce different properties

4. Use a minimal basis set

5. Employ a noniterative solution process
27

9
 Semiempirical Methods
– So, the Hamiltonian is changed
• Calculated energy can be above OR below the
exact energy (no Variational Principle anymore)
– Parameterization is used to reproduce
experimental data
– Usually, molecular geometry and energy (∆fH)
are the desired quantities
• ZINDO/S parameterized to reproduce UV-Vis
spectra
• TNDO parameterized to reproduce NMR chemical
shifts
28

Some Semiempirical Methods

Pariser-Parr-Pople MO Theory (PPP)
Extended Hückel MO Theory (EHMO)
Complete Neglect of Differential Overlap (CNDO)
Intermediate Neglect of Diff. Overlap (INDO)
Modified INDO (MINDO)
Modified Neglect of Diatomic Overlap (MNDO)
Austin Model 1 (AM1)
Most popular
Parametric Method 3 (PM3)
More recent: SAM1, PM5
• Different Hamiltonians, mainly differing in how
the e-/e- repulsion is handled, are used in each
29

Semiempirical
Advantages: Very fast, handles large molecules,
get good qualitative and sometimes ~quantitative
results
• Often works well for solution phase (water)
studies
Disadvantages: Not all atoms have parameters
available
• Similarity between molecule and parameterization
set
• Limited to ground state equilibrium geometries
– Can’t calculate properties not addressed in
parameterization process
30

10
 4. Density Functional Theory
• Instead of Ψ, look at electron density
– HF-SCF Theory is 4N dimensional
– Electron density is 3N dimensional (N = # e-’s)
• Get rid of one-dimension (spin)
– Hohenberg & Kohn (1964)
“The ground state energy E of an N-electron system
is a functional of the electronic density ρ, and E is
a minimum when evaluated with the exact ground
state density”
• Also
∫
N = ρ ( r ) dr
31

What is a “Functional”?
• The energy depends on a complicated
function, which is the e- density in 3-D space

– A function whose argument is also a function is

called a functional

– The energy is a unique functional of ρ(r)

– A functional enables a function to be mapped to a

number

32

DFT Process
• The energy is minimized with respect to
variations in ρ, subject to the constraint of
charge conservation:

∫
N = ρ ( r ) dr
– Electron density is easier than dealing with Ψ
– Instead of attempting to calculate the full N-
electron wavefunction, DFT only attempts to
calculate the total electronic energy and the
overall electron density distribution
33

11
 Problem
→ We have no idea of what the function describing
the electron density looks like!
• Kohn and Sham (1965): Look at specific
components of the energy functional:
E ( ρ) = Tni ( ρ) + Vne ( ρ) + Vee ( ρ) + ∆T ( ρ) + ∆Vee ( ρ)
where Tni = kinetic energy of non - interacting electrons,
Vne = nuclear - electron attraction, Vee = classical
electron - electron repulsion, ∆T = correction to the
kinetic energy from electron interaction, and ∆Vee =
sum of all non - classical corrections to the electron -
electron replusion energy
34

Solution
• The first three terms from the previous slide
have been seen before
– The last two terms are the troublesome ones
– They are usually combined together to form what
is called the exchange-correlation energy Exc
• This term includes effects of quantum mechanical
exchange and correlation

→ The correlation energy is what was missing

from HF theory, and was a major limitation

35

Approaches to Exc
• Terminology of density functionals:
1. Local: Simple dependence of Exc on ρ(r)
2. Nonlocal, or gradient corrected: Exc depends
on both ρ(r) and the gradient of ρ(r), ∇ ρ(r)

Variational approach:
• Overall energy minimum corresponds to the
exact ground state electron density (all other
densities give higher energy)
Further constraint:
• Number of electrons is fixed
∫
N = ρ ( r ) dr
36

12
 Differences Between HF-SCF and DFT
• DFT contains no approximations: It is exact
– All we need to know is Exc as a function of ρ
• As we will see, we must approximate Exc

• HF is a deliberately approximate theory so that

we can solve the equations exactly

• So, with DFT our theory is exact and the

equations are solved approximately, while
with HF our theory is approximate but we can
solve the equations exactly
37

Some Common Functionals

B3LYP: Becke, Lee, Parr, and Yang; uses a
hybrid exchange correlation energy functional
B3PW91: Becke, Perdew, and Wang; The PW91
correlation functional replaces LYP
PBE-96: Perdew, Burke, and Ernzerhof; an
updated correlation functional
VWN#5: Vosko, Wilk, and Nusair; another
correlation functional
BP86: Becke and Perdew; uses the Becke 1986
exchange energy and the Perdew 1996 GC
correlation functional
38

Density Functional Theory

Advantages: Scales as N3 (HF was N4)
• With few exceptions, DFT is the most cost-
effective method to achieve a given level of
accuracy
• Electron correlation included with less expense
Disadvantages: All e- are included
• Costs still large for molecules of interest
• With Ψ’s apply the correct operator to determine
the molecular property of interest
• With DFT, must know how the molecular property
depends on the electron density
39

13
 Brief Comparison of Methods
Things to consider:
Semi-
MM HF DFT
Emp.
Results Good qual. Better
Fast, large
Advantages can be ~ and ~quant. accuracy
molecules
quantitative results than HF
Parameters e- correlation Parameters Limited to
Dis-
may not be limits may not be smaller
advantages available accuracy available systems
Least
Expense is Inexpensive Expense is
Expense expensive
quite high technique high
technique
40

Units in Computational Chemistry

• Bohr: Atomic unit of Length (a0)
– Equal to the radius of the first Bohr orbit for a
hydrogen atom
• 5.29 x 10-11m (0.0529 nm, 52.9pm, 0.529Å)
• Hartree: Atomic unit of Energy
– Equal to twice the energy of a ground state
hydrogen atom
• 627.51 kcal/mole
• 2625.5 kJ/mole
• 27.211 eV
• 219474.6 cm-1
41

Lab Exercises
• Goal: Develop familiarity with the CAChe
software
• Drawing and viewing molecules
• Running various calculations
• Determining bond angles and distances

• Remember:
– We have a wide variety of software available
• We will explore all of them this evening
• You may want to try some of the lab exercises
using one of the other available programs
42

14