Song, Song, Mo, and Wu (J. Comput. Chem. 30, 399, 2009) published an efficient algorithm for analytic energy gradients of valence bond (VB) self-consistent field wave functions built from nonorthogonal localized orbitals, and their concluding discussion left a four-part future-work agenda: establishing performance in regimes with many Slater determinants, improving orbital-optimizer convergence, deriving analytic second derivatives (with respect to both orbital coefficients and nuclear coordinates), and demonstrating large-system applicability. The analytic second derivative of the VB energy with respect to nuclear coordinates — the geometric Hessian needed for harmonic vibrational frequencies and Hessian-based stationary-point characterization — has not been derived or implemented for multistructure VB wave functions with nonorthogonal orbitals. The open problem is to derive closed-form second nuclear derivatives of the VBSCF energy over nonorthogonal orbitals, including the coupled response of the optimized orbital and structure coefficients.

Background

Valence bond (VB) theory describes molecules as superpositions of chemically interpretable electronic structures (VB structures), each written as a spin-adapted combination of Slater determinants (antisymmetrized products of one-electron spin-orbitals) built from localized, generally mutually nonorthogonal orbitals. In the VB self-consistent field (VBSCF) method, both the coefficients of the VB structures and the nonorthogonal orbital shapes are varied to minimize the energy. Because the orbitals overlap, the Slater–Condon rules (the standard formulas for Hamiltonian matrix elements between determinants built from orthogonal orbitals) of molecular-orbital (MO) theory do not apply: Hamiltonian matrix elements between determinants must be evaluated through the inverse of the orbital-overlap matrix or, following Song, Song, Mo, and Wu (2009), through a transition density matrix (the density matrix connecting two different determinants) that yields a Hartree–Fock-like expression in which each determinant pair contributes a Fock-matrix-like build (a Fock matrix is the effective one-electron Hamiltonian matrix built from Coulomb and exchange contributions). That 2008/2009 algorithm achieved analytic energy gradients with respect to orbital coefficients at a cost scaling of m4m^4 (mm being the number of basis functions), comparable to Hartree–Fock, plus analytic gradients with respect to nuclear coordinates, and it was claimed to outperform earlier gradient work such as Dijkstra and van Lenthe (2000) and the Groningen VBSCF optimizer (van Lenthe, Verbeek, and Pulay, 1991), which used an augmented Newton–Raphson (aNR) orbital-optimization scheme in the TURTLE program. The paper's own concluding discussion listed future work: the algorithm builds one Fock-like matrix per determinant pair, so its efficiency relative to aNR-type methods in regimes with many determinants (many determinant pairs relative to the basis size, as noted in that concluding discussion) was unestablished; the optimizer used Broyden–Fletcher–Goldfarb–Shanno (BFGS, a quasi-Newton method) with a unit initial diagonal Hessian, with suggested but untested convergence enhancements such as DIIS (direct inversion in the iterative subspace, an extrapolation accelerator); only first derivatives were derived, and analytic second derivatives with respect to both orbital coefficients and nuclear coordinates were left as future work; and large-system, integral-direct (recomputing two-electron integrals on the fly instead of storing them) behavior was untested. A 2012 comment by van Lenthe, Broer-Braam, and Rashid further disputed the timing comparisons, noting that Broer and Nieuwpoort (1988) had already used a Fock matrix for matrix elements between different determinants, with a reply by Wu and Mo. Subsequent work by the same group replaced the per-pair Fock builds with an N-body reduced-density-matrix (RDM, the electron-density matrices of fixed particle rank obtained by integrating out the other electrons) formalism based on an enhanced Wick's theorem for nonorthogonal orbitals (Chen, Chen, and Wu, 2013), implemented a Hessian-based Newton–Raphson VBSCF optimizer using the analytic orbital Hessian and an automatic formula/code generator (AFCG, a tool that automatically derives and codes the matrix-element formulas between internally contracted excited configurations; Chen, Chen, and Wu, 2014), and applied Cholesky decomposition (a factorization technique that here approximates the two-electron integral matrix) and low-rank techniques — resolution of the identity and chain of spheres for exchange, which approximate integral transformations with controlled accuracy — to push full-orbital-optimization VB beyond 100 atoms (2025). In MO-based methods, the geometric Hessian — the matrix of second derivatives of the energy with respect to nuclear coordinates, obtained analytically since the 1980s for Hartree–Fock, density-functional, and complete-active-space self-consistent field (CASSCF, an MO multiconfigurational ansatz with orthogonal orbitals) methods — underlies harmonic vibrational frequency computations and the characterization of minima and transition states. For classical multistructure VB wave functions with nonorthogonal orbitals, only analytic first nuclear derivatives exist; the analytic geometric Hessian remains unavailable.

Problem Statement

Derive the closed-form analytic second derivatives of the valence bond self-consistent field (VBSCF) energy with respect to nuclear coordinates — the geometric Hessian — for multistructure VB wave functions whose localized orbitals are mutually nonorthogonal. The existing transition-density-matrix and reduced-density-matrix formalisms currently yield analytic first derivatives (Song, Song, Mo, and Wu, 2009; Chen, Chen, and Wu, 2013); any equivalent nonorthogonal-orbital formalism is admissible. The derivation must include the first and second derivatives of the nonorthogonal orbital-overlap matrix and of the one- and two-electron integrals with respect to nuclear displacements, and, because the orbital coefficients and VB structure coefficients are variationally optimized, it must include the coupled response (relaxed) contributions — the nonorthogonal-orbital analogue of the Z-vector treatment of multi-configurational self-consistent-field (MCSCF) second derivatives, in which a Lagrange-multiplier vector (the Z-vector) folds the first-order response of the variational parameters into a relaxed density — so that the full Hessian at a stationary point is obtained. Implementation in a VB program, agreement of the analytic Hessian elements with finite differences of the existing analytic gradients, and harmonic vibrational frequency computations at the VBSCF (or breathing-orbital VB) level would be natural demonstrations of correctness rather than mandatory acceptance requirements. Extension to post-VBSCF correlated methods — breathing-orbital VB (BOVB, a VB variant in which the orbitals of each VB structure relax independently), VB configuration interaction (VBCI), and second-order VB perturbation theory (VBPT2) — is a natural continuation but is not required for the core answer.

References

  1. Lingchun Song, Jinshuai Song, Yirong Mo, Wei Wu, 'An efficient algorithm for energy gradients and orbital optimization in valence bond theory', Journal of Computational Chemistry 30, 399-406 (2009). DOI: 10.1002/jcc.21065, https://doi.org/10.1002/jcc.21065
  2. J. H. van Lenthe, H. B. Broer-Braam, Z. Rashid, 'On the efficiency of VBSCF algorithms, a comment on "An efficient algorithm for energy gradients and orbital optimization in valence bond theory"', Journal of Computational Chemistry 33, 911-913 (2012). DOI: 10.1002/jcc.22924, https://doi.org/10.1002/jcc.22924
  3. Wei Wu, Yirong Mo, 'Reply to comment on the paper "An efficient Algorithm for Energy Gradients and Orbital Optimization in Valence Bond Theory"', Journal of Computational Chemistry 33, 914-915 (2012). DOI: 10.1002/jcc.22923, https://doi.org/10.1002/jcc.22923
  4. ZhenHua Chen, QianEr Zhang, Wei Wu, 'A new algorithm for inactive orbital optimization in valence bond theory', Science in China Series B: Chemistry 52, 1879-1884 (2009). DOI: 10.1007/s11426-009-0265-Y, https://doi.org/10.1007/s11426-009-0265-Y
  5. Zhenhua Chen, Xun Chen, Wei Wu, 'Nonorthogonal orbital based N-body reduced density matrices and their applications to valence bond theory. I. Hamiltonian matrix elements between internally contracted excited valence bond wave functions', The Journal of Chemical Physics 138, 164119 (2013). DOI: 10.1063/1.4801631, https://doi.org/10.1063/1.4801631
  6. Zhenhua Chen, Xun Chen, Wei Wu, 'Nonorthogonal orbital based N-body reduced density matrices and their applications to valence bond theory. II. An efficient algorithm for matrix elements and analytical energy gradients in VBSCF method', The Journal of Chemical Physics 138, 164120 (2013). DOI: 10.1063/1.4801632, https://doi.org/10.1063/1.4801632
  7. Jinshuai Song, Zhenhua Chen, Sason Shaik, Wei Wu, 'An efficient algorithm for complete active space valence bond self-consistent field calculation', Journal of Computational Chemistry 34, 38-48 (2013). DOI: 10.1002/jcc.23103, https://doi.org/10.1002/jcc.23103
  8. Xun Chen, Zhenhua Chen, Wei Wu, 'Nonorthogonal orbital based N-body reduced density matrices and their applications to valence bond theory. IV. The automatic implementation of the Hessian based VBSCF method', The Journal of Chemical Physics 141, 194113 (2014). DOI: 10.1063/1.4901729, https://doi.org/10.1063/1.4901729
  9. Zhenhua Chen, Fuming Ying, Xun Chen, Jinshuai Song, Peifeng Su, Lingchun Song, Yirong Mo, Qianer Zhang, Wei Wu, 'XMVB 2.0: A new version of Xiamen valence bond program', International Journal of Quantum Chemistry 115, 731-737 (2015). DOI: 10.1002/qua.24855, https://doi.org/10.1002/qua.24855
  10. Xiping Gong, Zhenhua Chen, Wei Wu, 'The application of Cholesky decomposition in valence bond calculation', Journal of Computational Chemistry 37, 2157-2162 (2016). DOI: 10.1002/jcc.24442, https://doi.org/10.1002/jcc.24442
  11. Chenru Ji, Fuming Ying, Peifeng Su, Chen Zhou, Wei Wu, 'Implementation of molecular symmetry in valence bond calculation', Journal of the Chinese Chemical Society 70, 341-348 (2023). DOI: 10.1002/jccs.202200476, https://doi.org/10.1002/jccs.202200476
  12. Zhenhua Chen, Wei Wu, 'Ab initio valence bond theory: A brief history, recent developments, and near future', The Journal of Chemical Physics 153, 090902 (2020). DOI: 10.1063/5.0019480, https://doi.org/10.1063/5.0019480
  13. Zhenhua Chen, Jinshuai Song, Xun Chen, Chen Zhou, Wei Wu, 'N-Body Reduced Density Matrix-Based Valence Bond Theory and Its Applications in Diabatic Electronic-Structure Computations', Accounts of Chemical Research 54, 3895-3905 (2021). DOI: 10.1021/acs.accounts.1c00421, https://doi.org/10.1021/acs.accounts.1c00421
  14. Jeppe Olsen, 'Novel methods for configuration interaction and orbital optimization for wave functions containing non-orthogonal orbitals with applications to the chromium dimer and trimer', The Journal of Chemical Physics 143, 114102 (2015). DOI: 10.1063/1.4929724, https://doi.org/10.1063/1.4929724
  15. Chenru Ji, Yueyang Zhang, Fuming Ying, Chen Zhou, Wei Wu, 'Low-Rank Algorithms for Ab Initio Valence Bond Approaches', Journal of Chemical Theory and Computation 21, 2462-2471 (2025). DOI: 10.1021/acs.jctc.4c01787, https://doi.org/10.1021/acs.jctc.4c01787
  16. Chen Zhou, Xun Wu, Fuming Ying, Wei Wu, 'Ab Initio Valence Bond Theory for Strongly Correlated Systems', Journal of Chemical Theory and Computation 21, 6243-6257 (2025). DOI: 10.1021/acs.jctc.5c00596, https://doi.org/10.1021/acs.jctc.5c00596
  17. Miao Guo, Xun Wu, Wei Wu, Chen Zhou, 'Ab Initio Valence Bond Molecular Dynamics: A Study of SN2 Reaction Mechanisms', The Journal of Physical Chemistry A 129, 2361-2370 (2025). DOI: 10.1021/acs.jpca.4c08431, https://doi.org/10.1021/acs.jpca.4c08431
  18. R. Broer, W. C. Nieuwpoort, 'Broken orbital symmetry and the description of valence hole states in the tetrahedral [CrO4]2- anion', Theoretica Chimica Acta 73, 405-418 (1988). DOI: 10.1007/BF00527744, https://doi.org/10.1007/BF00527744
  19. Fokke Dijkstra, Joop H. van Lenthe, 'Gradients in valence bond theory', The Journal of Chemical Physics 113, 2100-2108 (2000). DOI: 10.1063/1.482021, https://doi.org/10.1063/1.482021
  20. J. H. Van Lenthe, J. Verbeek, P. Pulay, 'Convergence and efficiency of the valence bond self-consistent field method', Molecular Physics 73, 1159-1170 (1991). DOI: 10.1080/00268979100101841, https://doi.org/10.1080/00268979100101841