| configuration interaction | |
|---|---|
| Name | Configuration interaction |
| Field | Quantum chemistry |
| Introduced | 1929 |
| Developers | C. A. Coulson; Hermann Weyl; later developed by John Pople, Per-Olov Löwdin |
| Related | Hartree–Fock method, Post-Hartree–Fock methods |
configuration interaction
Configuration interaction (CI) is a post-Hartree–Fock electronic structure technique in quantum mechanics used to describe correlated many-electron wavefunctions by linear combinations of Slater determinant configurations. CI systematically improves on single-determinant approximations to better predict molecular energies, spectra, and properties, and is foundational to modern computational quantum chemistry and materials simulation.
Configuration interaction sits within the family of post-Hartree–Fock methods alongside Møller–Plesset perturbation theory and coupled cluster theory. It addresses electron correlation—the interaction beyond mean-field models—by constructing a variational expansion of the exact electronic wavefunction in a basis of excited determinants derived from a reference, typically the Hartree–Fock determinant. CI underpins accurate predictions for molecular orbital energies, ionization potentials, excited states, and transition moments used in interpreting spectroscopy and guiding experimental chemical physics studies. Historically connected to developments by C. A. Coulson and Per-Olov Löwdin, CI remains a touchstone for benchmarking advanced methods like density functional theory approximations and quantum Monte Carlo.
The CI wavefunction is Psi_CI = Σ_I c_I |Φ_I⟩ where |Φ_I⟩ are Slater determinants or configuration state functions generated by single, double, triple, ... excitations relative to a reference determinant. The CI secular equation arises from the variational principle, leading to a matrix eigenvalue problem Hc = Ec with Hamiltonian matrix elements ⟨Φ_I|Ĥ|Φ_J⟩ evaluated using second quantization and Slater–Condon rules. Truncation schemes—configuration interaction singles (CIS), configuration interaction doubles (CID), CISD, CISDT, and full CI (FCI)—represent systematic approximations converging to the exact non-relativistic solution within a finite one-electron basis set. Orbital choice (canonical vs. localized) and use of symmetry-adapted configuration state functions influence sparsity and size of the CI Hamiltonian; group theory via point-group symmetry often reduces computational cost.
Practical CI requires efficient generation and diagonalization of large sparse Hamiltonians. Algorithms include direct CI builds, iterative eigensolvers such as the Davidson algorithm and Lanczos algorithm, and modern selected CI schemes like CIPSI (Configuration Interaction by Perturbatively Selected Iterative method) and Heat-bath CI. Implementations appear in software packages including Gaussian, GAMESS, Molpro, NWChem, Psi4, and research codes from groups at Lawrence Berkeley National Laboratory and Argonne National Laboratory. Recent efforts combine CI with tensor factorization (e.g., DMRG) and quantum computing proposals using variational quantum eigensolvers to target FCI for small molecules on devices developed by IBM Quantum and Google Quantum AI.
CI methods accurately predict electronic excited states, potential energy surfaces, and spectroscopic observables used by experimentalists in photoelectron spectroscopy, UV–vis spectroscopy, and photoemission. CI-derived transition dipole moments inform modeling of nonlinear optics and photochemistry relevant to renewable energy materials and photovoltaic research. In transition metal and actinide chemistry—areas with strong static correlation—multi-reference CI techniques yield insights into catalytic cycles studied at institutions like Max Planck Society laboratories and university research groups. CI remains a benchmark for validating density functional theory functionals and for interpreting spectra from facilities such as the European Synchrotron Radiation Facility.
The principal limitation of CI is combinatorial scaling: FCI scales factorially with system size, making it intractable beyond small molecules or minimal basis sets. Truncated CI (e.g., CISD) is not size-extensive, causing errors in thermochemical limits; remedies include size-consistent corrections and use of coupled cluster as an alternative. Approximations to manage cost include active space selection (CASSCF/CASCI), selected CI, perturbative corrections (e.g., CIPSI second-order correction), and exploitation of sparsity and locality. High-performance computing infrastructures at national labs and supercomputing centers are routinely used to push CI calculations, while algorithmic innovations target parallelism, low-rank tensor methods, and machine-learning-guided selection.
Multi-reference CI (MRCI) extends the CI idea to active-space references, combining a complete active space self-consistent field (CASSCF) reference with CI expansions to capture both static and dynamic correlation. Variants include MR-CISD, internally contracted MRCI, and hybrid methods linking CI to multireference perturbation theory (e.g., CASPT2). DMRG and selected CI provide alternative routes to approximate FCI in large active spaces; embedding techniques like density functional embedding and quantum embedding integrate CI fragments into larger systems, enabling studies of materials and condensed-phase problems.
Access to CI-level computational science intersects with equity in research capacity: high computational cost concentrates capability in well-funded institutions, disadvantaging scholars in resource-poor regions. Open-source packages (Psi4, NWChem) and community data initiatives help democratize access to CI tools and benchmark datasets. Transparent reporting, reproducible workflows, and training programs at universities and organizations such as the Royal Society and American Chemical Society promote inclusive participation. Equitable allocation of supercomputing resources and support for community-maintained software are vital to ensure diverse researchers can apply CI methods to problems in climate chemistry, energy justice, and public-health-related molecular design.