Post-Hartree–Fock methods

Post-Hartree–Fock methods
Name	Post-Hartree–Fock methods
Classification	Ab initio quantum chemistry methods
Founded in	Mid-20th century
Key people	John C. Slater, Clemens C. J. Roothaan, Per-Olov Löwdin, Isaiah Shavitt
Related methods	Hartree–Fock method, Density functional theory, Quantum Monte Carlo

Post-Hartree–Fock methods are a family of advanced ab initio quantum chemistry methods designed to provide more accurate solutions to the Schrödinger equation than the standard Hartree–Fock method. These methods systematically account for electron correlation, a critical quantum mechanical effect neglected in the Hartree–Fock approximation, which is essential for accurately predicting molecular properties, reaction energies, and spectroscopic data. The development of these techniques was pioneered by theoretical chemists and physicists including John C. Slater, Clemens C. J. Roothaan, and Per-Olov Löwdin, and they form the cornerstone of modern high-accuracy computational chemistry.

Overview and theoretical foundation

The theoretical foundation of post-Hartree–Fock methods rests on the Born–Oppenheimer approximation and the use of a Slater determinant as an initial wavefunction ansatz. The central goal is to recover the correlation energy, defined as the difference between the exact non-relativistic energy of a system and the energy obtained from the Hartree–Fock method. This is achieved by expanding the wavefunction in a basis set, often composed of Gaussian-type orbitals, and systematically improving upon the single-determinant description. Key mathematical frameworks include configuration interaction, coupled cluster theory, and Møller–Plesset perturbation theory, which were significantly advanced by the work of Isaiah Shavitt and others at institutions like the University of Florida and the University of Uppsala.

Single-reference methods

Single-reference methods assume the Hartree–Fock method wavefunction is a qualitatively good starting point. The most prominent families are configuration interaction, particularly full configuration interaction and coupled cluster theory, with the CCSD(T) method often termed the "gold standard" for single-reference calculations. Møller–Plesset perturbation theory, especially at the MP2 and MP4 levels, provides a computationally efficient alternative. These methods are implemented in widely used software packages such as Gaussian (software), NWChem, and Psi (computational chemistry), which originated from collaborative efforts at the Pacific Northwest National Laboratory and other research centers.

Multireference methods

Multireference methods are necessary when single-determinant descriptions fail, such as in systems with degenerate energy levels, bond dissociation, or transition metal complexes. These approaches, including complete active space SCF and multireference configuration interaction, construct wavefunctions from multiple Slater determinants. The development of methods like CASPT2 by Björn O. Roos and colleagues at Lund University has been instrumental for studying photochemical processes investigated at institutions like the University of California, Berkeley. The MRCI method is particularly important for accurate potential energy surface calculations.

Perturbative approaches

Perturbative approaches add correlation corrections to a reference wavefunction using Rayleigh–Schrödinger perturbation theory. The most common is Møller–Plesset perturbation theory, which uses the Fock operator as the unperturbed Hamiltonian. Higher-order variants like MP4 improve accuracy but increase cost. Perturbative corrections are also applied to multireference wavefunctions in methods like CASPT2 and NEVPT2, developed by theorists including C. David Sherrill at the Georgia Institute of Technology. These methods are crucial for benchmarking within projects like the GMTKN55 database.

Computational cost and scaling

The computational cost, measured by scaling with system size, is a major limitation. While Hartree–Fock method scales as O(N⁴), post-Hartree–Fock methods are far more expensive: MP2 scales as O(N⁵), CCSD as O(N⁶), and CCSD(T) as O(N⁷). Full configuration interaction has factorial scaling, making it prohibitive for all but the smallest systems. This drives the use of high-performance computing resources at facilities like the Texas Advanced Computing Center and the development of local correlation methods to extend applicability to larger molecules like those studied in biochemistry.

Applications and limitations

These methods are applied to predict precise spectroscopic constants, reaction enthalpies, barrier heights, and intermolecular forces, supporting fields from atmospheric chemistry to drug design. They are routinely used to benchmark density functional theory functionals. However, their high computational cost limits application to molecules with roughly 10-50 atoms, even with modern supercomputers like those at the Argonne National Laboratory. They also typically neglect relativistic quantum chemistry effects, which are important for systems containing heavy elements like those studied at the University of Tokyo.

Category:Computational chemistry Category:Quantum chemistry Category:Ab initio quantum chemistry methods