Fundamentals of Computational Chemistry!

A Guide for Organic Chemists!

Hui-Wen Shih!
MacMillan Group Meeting!
June 17, 2009!
 References

! Lead references

Cramer, Christopher J. Essentials of Computational Chemistry. 2nd ed. West Sussex: Wiley, 2004.

Ramachandran, K. I., Deepa, G., Namboori, K. Computational Chemistry and Molecular Modeling. Coimbatore: Springer, 2008

! General overviews
Grant, Guy H., Richards, W. G. Computational Chemistry. Oxford: Oxford University Press, 1995.

Bachrach, Steven M. Computational Organic Chemistry. Hoboken: Wiley, 2007.

Anslyn, Eric V., Dougherty, D.A. "Advanced Concepts in Electronic Structure Theory." Modern Physical Organic Chemistry.
Sausalito: University Science Books, 2006.
 Outline

Introduction

Molecular Mechanics

Physical Chemistry Review

Basis Sets

Hartree-Fock

Ab Initio Beyond Hartree-Fock

Semi-Empirical Methods

Density Functional Theory

Solvation Methods

Example
“To take advantage of readily accessible theoretical tools, and to
understand the results reported by theoretical collaborators (or
competitors), even the wettest of wet chemists can benefit from
some familiarity with theoretical chemistry.”!

- Cramer, Essentials of Computational Chemistry!

 The Basics

! Computational Chemistry ! The use of computer models to solve chemical problems, including

! Molecular energies and structures

! Geometry optimization

! Excited states, transition states, reaction pathways

! Bonds and orbitals

! Multipoles, polarizability, atomic charges, electrostatic potential, magnetic properties

! Vibrational Frequencies

! Spectroscopy: IR, Raman, NMR, Circular Dichroism

 Advantages and Costs

! Benefits of combining experiment with theory

! Interpretation of ambiguous or conflicting results

! Optimization of design or progress of experimental program

! Prediction of properties difficult or dangerous to obtain via experimentation

! A few caveats...

! Models are approximations

! Computational cost = computer resources + time

! Number of terms (length)

! How terms scale (Nm)

! In general, more accurate approximations require more time

! Therefore, choice of calculation depends on time availability and desired accuracy

 Molecular Mechanics

! Classical treatment of molecules

! Nuclei and electrons are treated as unified atom-like particles

! Atom-like particles are spherical balls

! Bonds are springs

! Interactions are treated as potential functions derived from classical mechanics

! The sum of interactions determines the conformation

! Depends on

! Force field: set of functions and constants to find potential energy

! Parameter set: reference points and force constants (from ab initio calculations or expt)

! Minimizing algorithm

! Each method is parameterized against a specific class of molecules

! Extensively used in biochemistry

 Molecular Mechanics: Commonly Used Force Fields

! MMn

! Gold standard for force fields

! parameterized to fit values from electron diffraction

! Bond stretching represented by Hooke's law expansion

k
U(r) = " # F(r)dr = " # "kx(dx) $ (l " l0 )2 [1" k'(l " l0 ) " k"(l " l0 ) " ...]
2
Hooke's Law Taylor expansion
harmonic oscillator

! ! MM2 can be unstable and is largely superseded by MM3

! MM3 has 153 atom types for organics and biomolecules

! MM4 is use for hydrocarbons, alcohols, ethers and carbohydrates

! Note: MM2 and MM3 in Chem3D is modified, as is MM2* and MM3* in MacroModel

! Modifications may be followed by date, e.g. MM3(94)

 Molecular Mechanics: Commonly Used Force Fields

! UFF (Universal Force Field)

! Range: General

! Covers most of the periodic table using 126 atom types

! AMBER

! Range: Biomolecules

! United atom approach (H's coalesced into heavy atoms)

! CHARMM (Chemistry at Harvard Macromolecular Mechanics)

! Range: Biomolecules

! Parameterized by experimental data

! MMFF (Merck Molecular Force Field)

! Range: Organics and biomolecules

! Focuses on condensed-phase processes in molecular dynamics

! Designed for conformational energies and nonbonding interactions of pharma compounds

 Physical Chemistry Review

! Quantum Mechanics

! The Hamiltonian
H" = E"
! For a general case:

!
! 2 N #1 2 !2 n 2
H = # $ M A % A + $ e Z A Z B rAB #
2 #1
$ % # $ $ e 2
Z r #1
A Ai + $ e 2 #1
rij
2 A A<B
2m i A i i< j

KE of Nuclei Nuclear-nuclear KE of electrons Nuclear-electron Electron-electron

Repulsion attraction repulsion

! ! Where
upper case terms refer to nuclei
lower case terms refer to electrons
Z = nuclear charge
e = electron charge
r = distance
 Common Types of Basis Sets

! Minimal basis sets

! One basis function per atomic orbital

! Carbon has 5 basis functions: 1s, 2s, 2px, 2py, 2pz

! STO-nG basis set uses a linear combination of n GTOs to describe an STO

! Carbon has 5n primitive functions

! STO-nG basis sets are used for "fast and dirty" calculations

! STO-3G: optimum combination of speed and accuracy

 Common Types of Basis Sets

! Pople split-valence basis sets

! More basis functions lead to better description of orbital wavefunction

! Described in the form x-yz+G(p,d)

! Polarization functions (p and d type functions) are added to describe s and p distortion

! Diffuse functions (+) are added for delocalized charges

! Carbon
1s 2s, 2px, 2py, 2pz 2s, 2px, 2py, 2pz

core orbital inner valence orbital outer valence orbital other total total basis
x gaussians y gaussians z gaussians +, p, d primitives functions
3-21G 3 2 1 16 9

6-31G 6 3 1 22 9

6-31G(d) 6 3 1 6 28 15

6-31+G 6 3 1 4 26 13

6-311G 6 3 1+1 26 13
 Common Types of Basis Sets

! Correlation consistent basis sets

! Maximize electron correlation energy

! correlation-consistent polarized split-valence N-zeta: cc-pVNZ

! As N increases, polarization functions also increase

! aug- designates addition of diffuse functions

! Other split valence basis sets

! Generically named

DZV (valence double-zeta)

QZVPP (valence quadruple-zeta plus polarization)

! Double, triple,... zeta basis sets

! Multiple basis functions for each AO

! Notation: w / x // y / z

E from w theory using x basis set

using geometry optimized with y theory using z basis set
 Hartree-Fock Theory

! HF SCF procedure

choose choose
molecular guess initial construct and solve HF equation
basis set density matrix
geometry

repeat with
construct new density matrix new density
matrix
choose new N
geometry
new density matrix
sufficiently similar to old?
N
Y

optimized? Y need to optimize geometry?

Y N

output data output data

 Strengths of Hartree-Fock Theory

! Strengths

! HF limit result is upper bound of true energy for system

! Acceptable results for minimum energy structures

! Using double-zeta polarized basis sets...

! Errors in heavy atom bond length: ~0.03 A

! Errors in heavy atom-hydrogen bond length: ~0.015 A

! Errors in bond angles: ~1.5o

! Dihedral angles are generally good

! H-bonds are generally good

 Weaknesses of Hartree-Fock Theory

! Weaknesses

! Bond interactions tend to be overestimated

! Bond lengths are usually too short

! Failure to account for electron correlation

! Should not be used for reaction energies

! Should not be used for structures with dispersion

! Heavier atom predictions lack inclusion of relativistic effects

! Unoccupied orbitals are not optimized

 Ab Initio Methods Beyond Hartree-Fock

! Correlated calculations: Add corrections for explicit electron-electron repulsion to HF

! Improved accuracy over HF

! Time-consuming, thus only used for small systems

! Variational theory may not apply

! Methods

! Moller-Plesset perturbation theory (MPx)

! Use of perturbation to account for electron correlation

! Multi-Configurations Self Consistent Field (MCSCF)

! Orbitals are optimized for a combination of configurations
configurations: molecular spin states and occupation numbers of orbitals

! Configuration Interaction (CI)

! Use of CASSCF (Complete Active Space SCF) with inclusion of specified excited states
e.g. CISD includes single and double excited states

! Coupled Cluster theory (CC)

! Includes excitation operator to generate specified excited states
! CCSD(T) is the "gold standard" of quantum chemistry
 Semi-Empirical Methods

! HF modified with empirical parameters

! Typically faster than ab initio calculations
! Accurate within parameter set
! Approximations
! Eliminates core electrons from calculation
! Uses a minimum number of basis sets
! Reduces the number of two electron integrals (Coulomb and exchange)
! Applications
! Geometry optimization, heats of formation, IP, charge distribution
! Structure-activity relationships
! Semi-empirical methods are only approximate!
 Semi-Empirical Methods

! Neglect of Differential Overlap (NDO)

! Complete Neglect of Differential Overlap (CNDO)

! Ignores all integrals involving different atomic orbitals

! Modified Neglect of Diatomic Overlap (MNDO)

! Parameterizes using spectroscopic data and includes multipole-multipole interactions

! Austin Model 1 (AM1)

! MNDO-based with modified nuclear-nuclear repulsion to mimic VDW interactions

! Reparameterized for hydrogen bonds,dipole moments, ionization potentials, and

molecular geometries

! Parametric Method 3 (PM3)

! MNDO-based parameterized for a large number of molecular properties

! Handles hydrogen-bonding well but can overestimate H-H attraction

! Slightly more accurate than AM1

! AM1 and PM3 are widely available and remains in use

 Density Functional Theory

! Energy for a system can be solved from the electron density

! E = Eelectron KE + Enuclear-electron attraction+ Eelectron-electron repulsion+ Enuclear-nuclear repulsion + Exc

! Exc (exchange-correlation) is an operator with an unknown form

! DFT is exact, but Exc is approximate

! Exc is usually separated into an exchange functional and a correlational functional

! Functionals are specified in method name

! Self-consistent field method similar to HF, solving for Kohn-Sham orbitals to obtain density

! Scales as N3, where N = #basis functions

! In general, computational time similar to HF, but accuracy similar to MP2

! Parameter methods exist for better fit to experimental data

! Nomenclature
exchange number of correlation
functional parameters functional
 Common Functionals for Exchange and Correlation

! Exchange

! Becke (B): Semi-empirical, extremely popular

! Other empirical functionals: CAM, FT97, O, PW, mPW, X...

! Non-empirical functionals: B86, P, PBE, mPBE...

! Correlation

! Lee, Yang and Parr (LYP): extremely popular

! PW91, !1, VWN, LG...

! P86 should be avoided

! Parameters

! Becke optimized for 3 parameters

! 1 parameter models also exist

! Parameterized (hybrid) methods are generally more robust

! B3LYP is the most popular and shows excellent results

 Performance Comparison

! Test set: Energies of 148 molecules including first and second row atoms

level of theory mean absolute error (kcal/mol) max error (kcal/mol)

MNDO 9.3 116.7

AM1 7.8 58.2

PM3 7.0 32.2

HF/6-311+G(3df,2p) 148.3 344.1

B3LYP/6-311+G(3df,2p) 3.1 20.2

B3PW91/6-311+G(3df,2p) 3.5 21.8

Scheiner et.al. J. Comput. Chem., 1997, 18, 775

 Performance Comparison

! Test set: Energies of 108 molecules including first and second row atoms

level of theory mean absolute error (kcal/mol)

HF/6-31G(d) 150.6

HF/6-311+G(3df,2p) 144.4

MP2/6-31G(d) 38.1

MP2/6-311+G(3df,2p) 9.7

QCISD/6-31G(d) 13.5

BLYP/6-3G(d,p) 7.0

BPW91/6-31G(d,p) 7.4

B3LYP/6-31G(d) 6.8

Curtiss et.al. J. Chem. Phys. 1998, 109, 42.

 Condensed Phases

! Changes in dipole result in changes to the potential energy surface

! Models are closer to experimental reality

! Highly desired for biochemical problems

protein folding, molecular conformations, protein binding, structure-activity relationship

! Computational cost is major concern with large system

! Models differ in treatment of reaction space, solvent molecules and boundaries

! Unit cell is usually created so solute is completely solvated and boundaries are continuous

! Implicit models treat solvent as a continuous medium

! Explicit models treat solvent as individual molecules, generally highly simplified

! Mixed models using both implicit and explicit treatment of solvation also exist
 Models for Condensed Phases

! Implicit Models

! Arbitrary cavity

! Define molecular cavity within 3D grid

! Assign dielectric constant at gridpoints on surface

! Allow solvent charge distribution to interact to determine electrostatic potential

! Determine energy from optimized interaction

! For biomolecules, solvated electrostatic potential is often desired result

! Boundary element approach

! Solve charge-charge interactions

! Molecular surface defined with charge density

! Conductor-like Screening Model (COSMO)

! If solvent is an ideal conductor, no potential exists

! Charge develops on solute surface
! Solvation energy is calculated from charge distribution on molecule and solvent charge
! Applicable to semi-empirical, HF, DFT...