| configuration interaction | |
|---|---|
| Name | Configuration interaction |
| Developer | C. C. J. Roothaan; developed further by H. J. Monkhorst, B. O. Roos, and others |
| Released | 20th century |
| Programming language | Fortran, C++ |
| Operating system | Cross-platform |
| Genre | Quantum chemistry, many-body method |
configuration interaction
Configuration interaction is a post-Hartree–Fock electronic structure method in Quantum Physics and quantum chemistry that represents a correlated wavefunction as a linear combination of Slater determinants or configuration state functions. It systematically improves on mean-field descriptions by including excited electronic configurations, and is widely used for accurate calculations of molecular energies, spectroscopic properties, and potential energy surfaces.
Configuration interaction (CI) emerged from mid-20th century efforts to incorporate electron correlation beyond the independent-particle approximation. Early theoretical foundations trace to work on atomic and molecular structure by C. C. J. Roothaan and the formulation of many-electron wavefunctions using determinants attributed to John C. Slater. The systematic CI expansion was developed and applied in the context of quantum chemistry by researchers including Per-Olov Löwdin, H. J. Monkhorst, and Bengt O. Roos. CI became a benchmark method during the rise of computational approaches in the 1960s–1990s, influencing software packages such as Gaussian, GAMESS (US), MOLPRO, and CFOUR. The method's historical importance lies in establishing standards for correlation treatments and motivating developments in coupled cluster theory and multireference approaches.
In CI the exact nonrelativistic electronic wavefunction Ψ is approximated as Ψ_CI = Σ_I c_I Φ_I, where {Φ_I} are antisymmetric determinants built from a single-particle basis (usually molecular orbitals from a Hartree–Fock calculation) and c_I are variational coefficients determined by diagonalizing the electronic Hamiltonian H. Matrix elements ⟨Φ_I|H|Φ_J⟩ are evaluated using Slater–Condon rules and one- and two-electron integrals computed in a chosen atomic orbital basis such as a Gaussian-type orbital basis set or Slater-type orbital. CI can be formulated equivalently in second quantization using creation and annihilation operators and occupation-number representation, connecting to concepts in many-body perturbation theory and Fock space.
Practical CI uses truncation schemes to limit the determinant space. Common truncations include: - CIS (configuration interaction with singles) includes single excitations and is frequently applied for low-order excited-state properties and in the context of TDHF benchmarking. - CID (doubles) and CISD (singles and doubles) include two-electron excitations and are standard for ground-state correlation corrections. - MRCI (multireference configuration interaction) constructs the expansion from a multiconfigurational reference such as CASSCF and is essential for bond-breaking and near-degeneracy problems; notable implementations include MRCI variants by Bengt O. Roos's group. Higher excitations (triples, quadruples) lead to full CI in the limit of a complete basis, which yields the exact nonrelativistic solution within the finite orbital space.
CI implementations rely on efficient generation and storage of determinants, sparse Hamiltonian techniques, and iterative eigensolvers like the Davidson algorithm for large configuration spaces. Integral transformation from atomic to molecular orbital bases and use of symmetry (point group symmetry, spin-adaptation) reduce computational cost. Modern codes exploit distributed-memory parallelization, tensor contractions, and direct CI algorithms that avoid storing two-electron integrals. Selected CI approaches (e.g., CIPSI, adaptive sampling) and stochastic CI variants reduce the active determinant set using perturbative selection criteria. Implementations are present in production programs such as MOLPRO, ORCA, Psi4, and Q-Chem.
CI is applied to calculate accurate potential energy surfaces, excitation energies, ionization potentials, and spectroscopic constants for atoms and small-to-medium molecules. MRCI and high-level truncated CI provide benchmark data for validating methods like coupled cluster theory and density functional theory. CI has been used in studies of photochemistry, conical intersections, and transition metal complexes where static correlation is important; it underpins simulations compared against experimental techniques such as photoelectron spectroscopy, electronic spectroscopy, and ultrafast spectroscopy.
Although variational and systematically improvable, truncated CI methods (e.g., CISD) are not size-consistent or size-extensive, causing errors that grow with system size when separate fragments are treated. Full CI is size-consistent but computationally intractable beyond small basis sets due to exponential scaling of the determinant space. CI also depends on the quality of the underlying orbital basis; orbital optimization via methods like CASSCF or use of natural orbitals can improve convergence. Accuracy can be augmented by perturbative corrections (e.g., Davidson corrections) and combination with complete basis set extrapolation techniques, but for large systems coupled cluster methods (e.g., CCSD(T)) are often preferred for balanced accuracy and size extensivity.
CI is closely related to other many-body approaches. Coupled cluster theory and configuration interaction can be connected via similarity-transformed Hamiltonians and excitation operators; truncated CC methods often yield better size-extensivity for comparable excitation levels. Multireference perturbation theories (e.g., CASPT2), density matrix renormalization group (DMRG) approaches, and tensor network states provide alternative ways to capture static and dynamic correlation, particularly for large active spaces. CI concepts extend to quantum computing algorithms for quantum chemistry, such as the variational quantum eigensolver (VQE) where electronic states are expanded in qubit-representable configurations, and to nuclear structure via configuration interaction shell-model calculations in nuclear physics.
Category:Quantum chemistry methods Category:Computational chemistry