Back to the page

In the print window, choose “Save as PDF”.

doubtly doubtly.in

Computational Tools Notes FE Sem 2 Computational Chemistry | MU

Introduction to Computational Chemistry · Updated 27 Oct 2026 · Read online: www.doubtly.in/mu-fe-computational-chemistry-tools-notes/

A computational chemist has a toolbox of methods, each suited to a different size of problem and a different level of accuracy. This module introduces the five main tools named in the syllabus: molecular mechanics, ab initio calculations, semi-empirical methods, density functional theory (DFT) and molecular dynamics. For each tool you should know the basic idea, the key equation, what it is good at, its limitations and typical examples. The module maps to course outcome CO2, and comparison questions between these methods are very common in exams.

Summary

Molecular mechanics (MM): Molecular mechanics ignores electrons completely and treats a molecule as a set of balls (atoms) joined by springs (bonds). The energy of a geometry, called the steric energy, is the sum of simple terms: bond stretching, angle bending, torsion (twisting about bonds), van der Waals interactions and electrostatic interactions, \(E = E_{\text{str}} + E_{\text{bend}} + E_{\text{tors}} + E_{\text{vdW}} + E_{\text{elec}}\). The constants in these terms (force constants, reference bond lengths, atomic charges) form a force field such as MM2, MM3, MMFF94, AMBER, CHARMM or OPLS. Because each term is a simple formula, MM is extremely fast and can handle proteins, DNA and polymers with thousands to millions of atoms. Its limitations: it cannot describe bond breaking or forming (in normal force fields), cannot give electronic properties such as orbitals or spectra, and works well only for molecules similar to those used to make the parameters. Module 4 covers MM in detail.

Ab initio calculations: Ab initio means "from the beginning". These methods solve the electronic Schrödinger equation \(\hat{H}\Psi = E\Psi\) using only fundamental constants (the masses and charges of electrons and nuclei, Planck's constant) and no experimental data. Using the Born-Oppenheimer approximation the nuclei are held fixed and only the electrons are treated quantum mechanically. The simplest ab initio method is Hartree-Fock (HF), in which each electron moves in the average field of all the others; the wave function is a Slater determinant of molecular orbitals, which are built from a basis set of atom-centred functions. HF misses the correlation between the instantaneous motions of electrons, so post-HF methods add electron correlation: Møller-Plesset perturbation theory (MP2), configuration interaction (CI) and coupled cluster (CCSD, CCSD(T)). Advantages: no empirical parameters, results can be improved systematically by using a bigger basis set and a higher method, and CCSD(T) with a large basis set is the "gold standard" for small molecules. Disadvantage: high cost, HF scales as N⁴, MP2 as N⁵ and CCSD(T) as N⁷, so only small and medium molecules can be treated. Module 6 explains HF and basis sets.

Semi-empirical methods: Semi-empirical methods use the same framework as Hartree-Fock but make it much cheaper by (1) treating only the valence electrons, using a minimal basis set, (2) neglecting many of the difficult electron repulsion integrals (the zero differential overlap, ZDO, approximation), and (3) replacing other integrals with parameters chosen so that the results match experimental data such as heats of formation and geometries. The simplest is the Hückel method for π electrons in conjugated hydrocarbons, where the energy of each π molecular orbital is written in terms of two parameters, the Coulomb integral α and the resonance integral β (both negative). Extended Hückel includes all valence electrons. Further methods in history are CNDO, INDO and the NDDO-based MNDO, AM1, PM3, PM6 and PM7. They are 100 to 1000 times faster than ab initio methods and handle hundreds to thousands of atoms, but their accuracy is unreliable for molecules unlike those in the training set.

Density functional theory (DFT): DFT uses the electron density \(\rho(\mathbf{r})\), a function of only three coordinates, instead of the many-electron wave function, which depends on 3n coordinates for n electrons. The first Hohenberg-Kohn theorem (1964) proves that the ground-state electron density determines all ground-state properties, including the energy, so the energy is a functional of the density, \(E[\rho]\). The second theorem gives a variational principle: the correct density gives the lowest energy. Kohn and Sham (1965) made this practical by introducing a fictitious system of non-interacting electrons in orbitals that have the same density as the real system. The total energy is \(E[\rho] = T_s[\rho] + E_{\text{ne}}[\rho] + J[\rho] + E_{\text{xc}}[\rho]\), where the exchange-correlation energy \(E_{\text{xc}}\) contains all the difficult quantum effects. The exact form of \(E_{\text{xc}}\) is unknown, so approximate functionals are used: LDA (depends only on density), GGA such as BLYP and PBE (also uses the density gradient), meta-GGA, and hybrid functionals such as B3LYP and PBE0 (mix in some exact HF exchange). DFT includes electron correlation at a cost similar to HF (formally about N³ to N⁴), so it is the most widely used method for molecules of 10 to a few hundred atoms. Weaknesses: no systematic way to improve functionals, poor description of dispersion (van der Waals) forces unless a correction such as D3 is added, and some errors for charge transfer and band gaps.

Molecular dynamics (MD): Molecular dynamics simulates how atoms move with time. The force on each atom is calculated from the potential energy, usually from a force field, \(\mathbf{F}_i = -\nabla_i V\), and Newton's second law \(\mathbf{F}_i = m_i\mathbf{a}_i\) is solved numerically in small time steps of about 1 to 2 femtoseconds using an integrator such as the Verlet or velocity Verlet algorithm. Repeating this millions of times produces a trajectory, a movie of atomic positions and velocities. Temperature is related to the average kinetic energy, \(\langle KE\rangle = \tfrac{3}{2}Nk_BT\) for N atoms. Simulations are run in ensembles: NVE (constant number, volume, energy), NVT (constant temperature, using a thermostat) and NPT (constant pressure and temperature, using a barostat). Periodic boundary conditions let a small box represent a bulk liquid or solid. MD gives diffusion coefficients, density, radial distribution functions, protein folding and drug binding behaviour, and conformational changes. When forces come from quantum calculations it is called ab initio MD. Monte Carlo simulation is a related tool that samples random arrangements instead of following time.

Choosing a tool: For a protein or polymer, use MM and MD. For a quick estimate of a large organic molecule, use a semi-empirical method. For most molecules up to a few hundred atoms where good energies and geometries are needed, use DFT. For small molecules where very high accuracy is needed, use correlated ab initio methods such as CCSD(T). Hybrid QM/MM methods treat a small reactive region quantum mechanically and the rest with MM; this idea won the 2013 Nobel Prize.

Mind map

Tools of Computational Chemistry

  • Molecular mechanics
    • Balls and springs
    • Force field: MM2, AMBER, CHARMM
    • Very fast, very large systems
    • No bond breaking, no electrons
  • Ab initio
    • Schrödinger equation, no parameters
    • HF: mean field
    • Post-HF: MP2, CI, CCSD(T)
    • Accurate but costly (N⁴ to N⁷)
  • Semi-empirical
    • Valence electrons, minimal basis
    • ZDO approximation
    • Parameters from experiment
    • Hückel, CNDO, MNDO, AM1, PM3, PM7
  • DFT
    • Electron density ρ(r)
    • Hohenberg-Kohn theorems
    • Kohn-Sham orbitals
    • Functionals: LDA, GGA, B3LYP
  • Molecular dynamics
    • Newton's laws, F = −∇V
    • Verlet integration, 1-2 fs steps
    • Ensembles NVE, NVT, NPT
    • Trajectory, temperature, diffusion

How a molecular dynamics simulation works

  1. Set up the system: starting coordinates of all atoms (for example a protein in a box of water) and a force field.
  2. Apply periodic boundary conditions so that atoms leaving one side of the box enter from the other side.
  3. Minimize the energy briefly to remove bad contacts.
  4. Assign random starting velocities corresponding to the chosen temperature (Maxwell-Boltzmann distribution).
  5. Calculate the force on every atom from the force field: \(\mathbf{F}_i = -\partial V/\partial \mathbf{r}_i\).
  6. Find accelerations \(\mathbf{a}_i = \mathbf{F}_i/m_i\) and update positions and velocities over a time step Δt (about 1-2 fs) with the Verlet algorithm.
  7. Repeat steps 5 and 6 millions of times, using a thermostat and barostat to keep T and P constant if required.
  8. Equilibrate first, then save positions and velocities at intervals (the trajectory) during the production run.
  9. Analyse the trajectory: average energy, temperature, density, RMSD, radial distribution function, diffusion coefficient.

The Hückel method in brief

For a conjugated molecule each carbon contributes one 2p orbital. The π molecular orbitals are linear combinations of these, and the energies come from a secular determinant in which every diagonal element is α − E, every element for two bonded carbons is β, and all other elements are zero. For ethylene \(\begin{vmatrix} \alpha - E & \beta \\ \beta & \alpha - E\end{vmatrix} = 0\) gives \(E = \alpha \pm \beta\). Since β is negative, \(\alpha + \beta\) is the bonding (lower) orbital and \(\alpha - \beta\) the antibonding orbital. For 1,3-butadiene the four energies are \(\alpha \pm 1.618\beta\) and \(\alpha \pm 0.618\beta\).

Important definitions

TermMeaning
Force fieldThe set of energy functions and parameters used in molecular mechanics to calculate the energy of a molecule from its geometry.
Ab initio methodA quantum method that solves the Schrödinger equation from first principles without experimental parameters.
Hartree-Fock methodThe basic ab initio method in which each electron moves in the average field of the others; the wave function is a single Slater determinant.
Electron correlationThe tendency of electrons to avoid each other instantaneously, which HF neglects; post-HF and DFT methods include it.
Semi-empirical methodA simplified HF-type method that neglects many integrals and uses parameters fitted to experimental data.
ZDO approximationZero differential overlap: products of different basis functions on the same electron are set to zero, removing most integrals.
Electron density ρ(r)The number of electrons per unit volume at point r; it integrates to the total number of electrons.
FunctionalA rule that turns a whole function into a number; the DFT energy is a functional of the density, E[ρ].
Exchange-correlation functionalThe part of the DFT energy that contains exchange and correlation effects; it must be approximated (LDA, GGA, hybrid).
Molecular dynamicsSimulation of the motion of atoms with time by solving Newton's equations of motion.
Time stepThe small time interval Δt over which positions and velocities are updated in MD, typically 1-2 fs.
TrajectoryThe record of positions and velocities of all atoms at successive times in an MD simulation.
EnsembleThe set of conditions held constant in a simulation, such as NVE, NVT or NPT.
Periodic boundary conditionsReplicating the simulation box in all directions so a small system behaves like bulk matter.

Formulas

FormulaMeaning
\( E_{\text{MM}} = E_{\text{str}} + E_{\text{bend}} + E_{\text{tors}} + E_{\text{vdW}} + E_{\text{elec}} \)Molecular mechanics steric energy.
\( \hat{H}\Psi = E\Psi \)Time-independent Schrödinger equation solved by ab initio methods.
\( \Psi_{\text{total}} \approx \psi_{\text{el}}(\mathbf{r};\mathbf{R})\,\chi_{\text{nuc}}(\mathbf{R}) \)Born-Oppenheimer approximation: electronic and nuclear motions are separated.
\( E_{\text{corr}} = E_{\text{exact}} - E_{\text{HF}} \)Correlation energy (always negative).
\( E = \alpha \pm \beta \)Hückel π orbital energies of ethylene; α, β negative.
\( E[\rho] = T_s[\rho] + E_{\text{ne}}[\rho] + J[\rho] + E_{\text{xc}}[\rho] \)Kohn-Sham DFT energy: non-interacting kinetic energy, nucleus-electron attraction, classical electron repulsion, exchange-correlation.
\( \left[-\tfrac{1}{2}\nabla^2 + v_{\text{eff}}(\mathbf{r})\right]\phi_i = \varepsilon_i\phi_i,\quad \rho = \sum_i |\phi_i|^2 \)Kohn-Sham equations (atomic units).
\( \mathbf{F}_i = m_i\mathbf{a}_i = -\nabla_i V \)Newton's equation used in MD; force is minus the gradient of the potential.
\( x(t+\Delta t) = x(t) + v(t)\Delta t + \tfrac{1}{2}a(t)\Delta t^2 \)Velocity Verlet position update.
\( v(t+\Delta t) = v(t) + \tfrac{1}{2}[a(t) + a(t+\Delta t)]\Delta t \)Velocity Verlet velocity update.
\( x(t+\Delta t) = 2x(t) - x(t-\Delta t) + a(t)\Delta t^2 \)Original (position) Verlet algorithm.
\( \langle KE\rangle = \tfrac{3}{2}Nk_BT \)Temperature from average kinetic energy of N atoms; \(k_B = 1.381\times10^{-23}\) J/K.

Key points to remember

Solved examples

Example 1. How many time steps are needed for a 1 ns MD simulation with a 2 fs time step?

Number of steps \(= \dfrac{1\times10^{-9}}{2\times10^{-15}} = 5\times10^{5}\) steps (five lakh). Each step needs a full force calculation, which is why long simulations are expensive.

Example 2. In an MD run with Δt = 1 fs, an atom has x = 1.000 Å, v = 0.010 Å/fs and a = −0.002 Å/fs². Find the new position. If the new acceleration is −0.0022 Å/fs², find the new velocity (velocity Verlet).

\(x(t+\Delta t) = 1.000 + 0.010\times1 + \tfrac{1}{2}(-0.002)(1)^2 = 1.000 + 0.010 - 0.001 = 1.009\) Å. \(v(t+\Delta t) = 0.010 + \tfrac{1}{2}(-0.002 - 0.0022)(1) = 0.010 - 0.0021 = 0.0079\) Å/fs.

Example 3. A simulation box contains 100 argon atoms with total kinetic energy 8.0 × 10⁻¹⁹ J. Find the temperature.

\(T = \dfrac{2\,KE}{3Nk_B} = \dfrac{2\times8.0\times10^{-19}}{3\times100\times1.381\times10^{-23}} = 386\) K.

Example 4. Using Hückel theory, find the π energies of ethylene and the total π electron energy.

Secular determinant: \((\alpha - E)^2 - \beta^2 = 0\), so \(E = \alpha \pm \beta\). Both π electrons go into the lower orbital \(\alpha + \beta\) (β is negative), so \(E_\pi = 2(\alpha + \beta) = 2\alpha + 2\beta\).

Example 5. The Hückel energies of 1,3-butadiene are α ± 1.618β and α ± 0.618β. Find its total π energy and delocalisation energy.

The four π electrons fill the two lowest orbitals: \(E_\pi = 2(\alpha + 1.618\beta) + 2(\alpha + 0.618\beta) = 4\alpha + 4.472\beta\). Two isolated double bonds (two ethylenes) would have \(2(2\alpha + 2\beta) = 4\alpha + 4\beta\). Delocalisation energy \(= (4\alpha + 4.472\beta) - (4\alpha + 4\beta) = 0.472\beta\). Since β is negative, butadiene is stabilized by delocalisation.

Example 6. A DFT calculation (scaling N³) and an HF calculation (scaling N⁴) each take 10 minutes for a molecule. Estimate the times for a molecule twice as large.

DFT: \(10\times2^3 = 80\) minutes. HF: \(10\times2^4 = 160\) minutes.

Important questions and answers

Q1. Compare molecular mechanics, semi-empirical, ab initio and DFT methods. (8 marks)

Basis: MM uses classical mechanics (balls and springs); semi-empirical, ab initio and DFT use quantum mechanics. Electrons: MM ignores electrons; semi-empirical treats valence electrons only; ab initio and DFT treat all electrons. Parameters: MM uses force field parameters; semi-empirical uses parameters fitted to experiment; ab initio uses none; DFT uses an approximate exchange-correlation functional. Cost: MM very low (about N²), semi-empirical low, DFT moderate (about N³), ab initio high (HF N⁴, CCSD(T) N⁷). System size: MM up to millions of atoms, semi-empirical thousands, DFT hundreds, high-level ab initio tens. Outputs: MM gives geometry and steric energy only; quantum methods also give orbitals, charges, spectra and reaction energies, including bond breaking. Examples: MM2, AMBER; AM1, PM3; HF, MP2, CCSD(T); B3LYP, PBE.

Q2. What are ab initio calculations? State their advantages and limitations. (5 marks)

Ab initio calculations solve the electronic Schrödinger equation using only fundamental physical constants, with the Born-Oppenheimer approximation and a chosen basis set, without any experimental parameters. Hartree-Fock is the simplest; MP2, CI and coupled cluster include electron correlation. Advantages: no empirical data needed, can be applied to any element or unusual molecule, results improve systematically with better method and larger basis set, very high accuracy possible. Limitations: computationally expensive and the cost rises steeply (N⁴ to N⁷), so only small and medium molecules can be treated; HF itself neglects correlation and gives errors in bond energies.

Q3. Explain semi-empirical methods with examples. (5 marks)

Semi-empirical methods are simplified Hartree-Fock-type methods. Simplifications: only valence electrons are treated, a minimal basis set is used, and the zero differential overlap approximation removes most two-electron integrals. The remaining integrals are replaced by parameters adjusted to reproduce experimental heats of formation, geometries, dipole moments and ionization energies. Examples: Hückel (π electrons only), extended Hückel, CNDO, INDO, MNDO, AM1, PM3, PM6, PM7. Advantages: very fast, can treat large molecules, give reasonable heats of formation for organic molecules. Limitations: unreliable for molecules unlike the parameterization set, poor for transition metals and some types of bonding, and cannot be systematically improved.

Q4. Explain the basic principle of density functional theory. (8 marks)

DFT calculates the energy from the electron density ρ(r) rather than from the many-electron wave function. Hohenberg-Kohn theorem 1: the ground-state density uniquely determines the external potential and hence all ground-state properties; the energy is a functional E[ρ]. Theorem 2: the true ground-state density minimizes the energy functional (variational principle). Kohn-Sham approach: a fictitious system of non-interacting electrons is chosen to have the same density as the real system; its orbitals satisfy \([-\tfrac12\nabla^2 + v_{\text{eff}}]\phi_i = \varepsilon_i\phi_i\), and \(\rho = \sum|\phi_i|^2\). The energy is \(E = T_s + E_{\text{ne}} + J + E_{\text{xc}}\). The equations are solved self-consistently, like HF. Since \(E_{\text{xc}}\) is unknown, approximations are used: LDA, GGA (BLYP, PBE), hybrids (B3LYP, PBE0). Advantages: includes electron correlation, accuracy close to MP2 or better at about HF cost, so it is the most used method for molecules and solids. Limitations: no systematic improvement, dispersion needs correction, some errors in band gaps and charge transfer.

Q5. Explain molecular dynamics simulation. What information does it provide? (7 marks)

MD follows the motion of atoms with time by solving Newton's equations. Forces are calculated from a potential (usually a force field) as \(\mathbf{F} = -\nabla V\); accelerations are \(\mathbf{a} = \mathbf{F}/m\); positions and velocities are updated over a small time step (1-2 fs) using the Verlet algorithm; the cycle is repeated millions of times. Initial velocities are chosen for the desired temperature; thermostats and barostats maintain T and P in NVT or NPT ensembles; periodic boundary conditions mimic bulk matter. Information obtained: time evolution of structure, average energy, temperature and pressure, density, diffusion coefficients, radial distribution functions, conformational changes, protein folding, ligand binding and the behaviour of liquids, membranes and polymers.

Q6. Why is a time step of about 1-2 fs used in molecular dynamics? (3 marks)

The fastest motions in molecules are bond vibrations involving hydrogen, with periods of about 10 fs. To follow these accurately the time step must be much smaller than the period; a larger step makes the integration unstable and the energy blows up. A smaller step would be accurate but would need many more steps for the same simulation time.

Q7. Write a short note on the Hückel method. (5 marks)

Hückel theory treats only the π electrons of planar conjugated molecules. Each carbon contributes one p orbital, and π MOs are LCAOs of these. Approximations: overlap between different atoms is set to zero; the Coulomb integral α is the same for all carbons; the resonance integral β is the same for bonded neighbours and zero for non-neighbours. The secular determinant has α − E on the diagonal and β for bonded pairs. Ethylene gives E = α ± β; butadiene gives α ± 1.618β and α ± 0.618β, with delocalisation energy 0.472β. Benzene gives a delocalisation energy of 2β, explaining its extra stability. It is the simplest semi-empirical method and gives useful qualitative insight.

Exam tips

FAQs

Is DFT an ab initio method?

Opinions differ. DFT is based on exact theorems and uses no experimental data in principle, but the exchange-correlation functionals are approximate and some contain fitted parameters. Many books list it separately from ab initio wave function methods, as the syllabus does.

Why does molecular mechanics not need electrons?

The force field parameters already include the average effect of electrons on bond lengths, angles and interactions. This makes MM fast, but it also means MM cannot describe processes where electrons rearrange, such as bond breaking.

What is B3LYP?

A hybrid DFT functional combining Becke's three-parameter exchange mixing (including some exact HF exchange) with the Lee-Yang-Parr correlation functional. It has been one of the most widely used functionals for organic molecules.

Can molecular dynamics show a chemical reaction?

Ordinary force-field MD cannot break or form bonds. Reactive force fields (such as ReaxFF), QM/MM MD or ab initio MD, where forces come from quantum calculations, are used to simulate reactions.

These notes follow the University of Mumbai first-year engineering syllabus (NEP 2020, Introduction to Computational Chemistry) and were prepared with AI help and checked.

Practise this on Doubtly (free)