Computational Chemistry is the use of mathematical models, algorithms and computer simulations to investigate chemical systems, molecular structures, reactions and properties. It applies quantum mechanics, classical mechanics, statistical mechanics, numerical analysis and computational algorithms to questions that can be difficult to answer through theory or experiment alone. Its subjects include molecular structure, electronic structure, chemical bonding, spectroscopy, thermodynamics, reaction mechanisms, molecular dynamics, materials and intermolecular interactions.
Methods range from highly accurate quantum-mechanical calculations on relatively small systems to approximate and classical approaches capable of treating very large molecular assemblies. Computational Chemistry is closely related to theoretical chemistry but places particular emphasis on calculations that can be executed numerically and systematically. Its results can be used to predict molecular properties, interpret experiments, screen compounds, investigate reaction pathways and understand chemical behaviour at scales or conditions that may be difficult to study directly.
Quantum chemistry is a major component of Computational Chemistry because electronic structure determines many molecular properties. The Schrödinger equation provides the fundamental quantum-mechanical description, but exact solutions become computationally infeasible for many-electron systems as molecular size increases. Approximation methods therefore include Hartree-Fock theory, density functional theory, post-Hartree-Fock methods and semi-empirical approaches. Hartree-Fock represents a many-electron wavefunction through a mean-field approximation, while density functional theory reformulates the problem around electron density and requires an approximation for the exchange-correlation functional.
Correlation methods such as coupled-cluster and configuration-interaction approaches can provide high accuracy for suitable systems but generally have substantially greater computational cost. Basis sets represent molecular orbitals using finite mathematical functions, and the choice of basis set affects both accuracy and computational requirements. The appropriate electronic-structure method depends on molecular size, chemical environment, desired accuracy and computational resources.
Molecular mechanics provides an alternative approach for systems too large for many quantum-mechanical calculations. It represents atoms and bonds using classical force fields in which the potential energy is expressed through terms describing bond stretching, angle bending, torsional motion, electrostatics and non-bonded interactions. Molecular dynamics integrates equations of motion to simulate how atoms move over time, allowing investigation of conformational behaviour, diffusion, solvation and other properties.
Monte Carlo methods instead generate configurations according to statistical distributions and can be useful for equilibrium properties. Force fields are parameterised models and their accuracy depends on the chemical systems and conditions for which they were developed. Molecular mechanics generally cannot describe electronic rearrangement or bond breaking and formation directly because its standard representation assumes a fixed bonding topology. Hybrid quantum-mechanical/molecular-mechanical methods address this limitation by treating a chemically important region quantum mechanically while representing the surrounding environment with a lower-cost model.
Computational Chemistry also includes computational approaches to chemical reactions and thermodynamics. Reaction-path methods seek structures corresponding to transition states or minimum-energy pathways between reactants and products. Activation energies and reaction profiles can provide information about kinetic feasibility, although computed values depend on the electronic-structure method, treatment of solvent, temperature and other modelling assumptions. Free-energy calculations account for entropic and thermal effects that are not captured adequately by a simple comparison of potential energies.
Methods such as umbrella sampling, thermodynamic integration and free-energy perturbation can estimate differences in free energy between molecular states, while enhanced-sampling methods address the difficulty of exploring rare or high-barrier events. Solvent effects may be treated explicitly using molecular simulations or approximately using continuum models. Accurate reaction modelling therefore requires attention to both electronic structure and statistical-mechanical contributions.
Computational Chemistry is extensively used in drug discovery, materials research, catalysis, spectroscopy and molecular engineering. Virtual screening can evaluate large collections of candidate molecules using docking, molecular descriptors, machine-learning models or other computational filters before experimental testing. Quantum calculations can estimate electronic properties, vibrational frequencies, charge distributions and reaction energetics. Materials simulations can investigate crystal structures, defects, surfaces, mechanical properties and electronic behaviour. Catalysis studies can examine adsorption, reaction intermediates and transition states on catalysts.
Computational spectroscopy can predict quantities that can be compared with experimental spectra to assist molecular identification. These applications do not eliminate experimental chemistry because models inevitably simplify real systems, but they can reduce the number of hypotheses requiring laboratory investigation and provide molecular-level interpretations that may be difficult to obtain directly from measurements.
Accuracy and computational cost form a central trade-off in Computational Chemistry. High-level quantum methods can provide excellent results for suitable systems but may scale steeply with system size, limiting their application to large molecules. Density functional theory offers a widely used compromise but different functionals can produce different results for particular chemical phenomena. Molecular mechanics permits much larger simulations but depends on empirical or semi-empirical parameterisation.
Sampling is another limitation: a simulation may be numerically accurate for the configurations it visits while failing to explore relevant regions of molecular configuration space. Finite simulation times can therefore produce biased estimates of equilibrium or kinetic quantities. Boundary conditions, temperature control, pressure control, solvent treatment and initial configurations can all affect molecular simulations. Reliable computational chemistry consequently requires method selection based on the chemical question rather than assuming that one computational approach is universally superior.
Computational Chemistry overlaps with Computational Physics, Computational Biology, Materials Science and Scientific Computing. Computational Physics provides many of the numerical and physical techniques used in molecular simulation, while Computational Chemistry specialises them for chemical systems and questions. Computational Biology applies computational methods to biological organisation, and its molecular-level studies may use computational chemistry techniques for proteins, ligands and biomolecular interactions. Materials science uses both chemistry and physics calculations to investigate solids and interfaces.
Quantum chemistry differs from computational chemistry in emphasis: quantum chemistry focuses on applying quantum theory to chemical systems, whereas computational chemistry encompasses quantum methods as well as classical, statistical and hybrid computational approaches. Cheminformatics is more strongly concerned with chemical information, molecular databases, representations and data-driven analysis, although it frequently intersects with computational chemistry.
Validation is essential when computational chemistry predictions are used to support scientific conclusions. Computed properties can be compared against experimental measurements or high-level benchmark calculations, while convergence studies examine sensitivity to basis sets, numerical grids, system size and other computational parameters. Reproducibility requires documenting molecular structures, computational methods, software versions, basis sets, force fields, parameters and relevant simulation settings. Different models may agree on qualitative behaviour while differing quantitatively, and apparent agreement with experiment can sometimes arise from compensating errors.
Machine-learning approaches introduce additional considerations concerning training data, extrapolation and uncertainty. Computational Chemistry is therefore most reliable when predictions are interpreted within the known domain of validity of the chosen model and supported by appropriate validation. Its broader significance is that computation can provide controlled molecular-level experiments, quantitative predictions and mechanistic information that complements laboratory chemistry and theoretical analysis.
