Spin-Coupled Generalized Valence Bond (SCGVB)

• SCGVB is a spin-adapted nonorthogonal valence-bond ansatz that represents an N-electron wavefunction as a linear combination of spin-coupled structures using localized, chemically meaningful orbitals.
• It employs a generalized secular equation (HC = SCE) to optimize energy, navigating the challenges of nonorthogonal determinants and complex orbital overlaps.
• Recent developments integrate quantum computing approaches and alternative formulations, highlighting SCGVB’s efficiency in describing bond reorganization and strong electron correlation.

Searching arXiv for SCGVB and directly related work to ground the article. Spin-coupled generalized valence bond (SCGVB) is a spin-adapted nonorthogonal valence-bond ansatz in which an $N$-electron wavefunction in a fixed $(S,M)$ sector is written as a linear combination of spin-coupled structures, each structure being an antisymmetrized product of generally nonorthogonal spatial orbitals and a spin eigenfunction. After expansion of the spin functions into a primitive spin basis, the ansatz becomes a structured linear combination of nonorthogonal Slater determinants, and energy minimization yields the generalized secular equation $HC=SCE$ rather than an ordinary eigenvalue problem (Gabrielly, 12 Mar 2026). This formal architecture underlies recent work on quantum estimators for SCGVB matrix elements (Gabrielly, 12 Mar 2026), motivates nonorthogonal Jordan–Wigner encodings for valence-bond methods (Marruzzo et al., 16 Sep 2025), and provides the reference point against which more restricted perfect-pairing, wave-packet, and orthogonal-orbital spin-coupled variants are best understood (Ando, 2017, Marti-Dafcik et al., 2024, Wang et al., 2020).

1. Formal wavefunction architecture

At the SCGVB level, the wavefunction is written as a linear combination of spin-coupled structures,

$$\Psi^{\mathrm{SCGVB}}=\mathcal N \sum_{k=1}^{f_N^S} c_{S,k}\,\psi^N_{S,M;k} =\mathcal N \sum_{k=1}^{f_N^S} c_{S,k}\,\hat A\,\phi_1\phi_2\cdots\phi_N\,\Theta^N_{S,M;k},$$

where $f_N^S$ is the number of linearly independent $N$-electron spin eigenfunctions for total spin $S$, $c_{S,k}$ are spin-coupling coefficients, $\{\phi_i\}$ are generally nonorthogonal spatial orbitals, and $\hat A$ is the antisymmetrizer (Gabrielly, 12 Mar 2026). The spin functions are expanded in a primitive spin basis according to

$(S,M)$0

and the coefficients are recombined as

$(S,M)$1

Substitution into the SCGVB ansatz yields a determinant expansion in which the spin-adapted state becomes a structured sum of nonorthogonal determinants $(S,M)$2 (Gabrielly, 12 Mar 2026).

Because the determinant basis is nonorthogonal, SCGVB does not reduce to an ordinary Hermitian eigenvalue problem in an orthonormal basis. The wavefunction is determined by minimizing

$(S,M)$3

which yields the generalized secular equation

$(S,M)$4

Equivalently, for a basis $(S,M)$5 of nonorthogonal structures,

$(S,M)$6

and one solves $(S,M)$7 (Gabrielly, 12 Mar 2026). The normalization factor can be written using Löwdin rules as $(S,M)$8 with

$(S,M)$9

where $HC=SCE$0 is the overlap matrix between the spin orbitals entering determinants $HC=SCE$1 and $HC=SCE$2 (Gabrielly, 12 Mar 2026).

This architecture makes two features central. First, SCGVB is explicitly spin-adapted rather than spin-contaminated. Second, its compactness depends on chemically meaningful localized orbitals and spin-coupling coefficients rather than on a large orthogonal determinant expansion. A plausible implication is that SCGVB’s efficiency is tied less to determinant count than to the structure imposed by spin adaptation and locality.

2. Spin coupling, nonorthogonality, and relation to neighboring formalisms

SCGVB is often discussed alongside generalized valence bond, perfect-pairing, and broader spin-coupled valence-bond approaches, but the comparison literature distinguishes these levels of flexibility rather than treating them as identical. In the modern SCGVB formulation summarized above, several independent spin eigenfunctions can contribute through the coefficients $HC=SCE$3, and the orbital set is generally nonorthogonal (Gabrielly, 12 Mar 2026). By contrast, the wave-packet valence-bond model of “Localized electron wave packet description of chemical bond and excitation: Floating and breathing Gaussian with valence-bond coupling” is explicitly a minimal perfect-pairing VB model with wave-packet orbitals, not a full spin-coupled expansion (Ando, 2017). Likewise, “Describing Strong Correlation with Block-Correlated Coupled Cluster Theory” is built on GVB perfect pairing, defined in the paper as the antisymmetric product of strongly orthogonal geminals and its special case GVB-PP, rather than on a general SCGVB reference (Wang et al., 2020). “Spin-coupled molecular orbitals: chemical intuition meets quantum chemistry” is closely aligned in chemical content, because it uses localized orbitals and explicit spin eigenfunctions, but within each contributing configuration the occupied orbitals are orthonormal and the total wavefunction is a linear combination of different CSFs built from different orbital sets (Marti-Dafcik et al., 2024).

Aspect	SCGVB	Related restricted or reformulated approaches
Spin structure	Linear combination of spin-coupled structures with coefficients $HC=SCE$4	One perfect-pairing spin function in WP-VB; strongly orthogonal singlet geminals in GVB-BCCC; explicit spin eigenfunctions in spin-coupled MO CSFs
Orbital character	Generally nonorthogonal spatial orbitals	One floating-and-breathing Gaussian per electron in WP-VB; orthogonal occupied orbitals within each CSF in spin-coupled MO
Variational problem	Generalized secular equation $HC=SCE$5	Perfect-pairing VB optimization in WP-VB; block-correlated CC equations on a GVB-PP reference in GVB-BCCC

These distinctions matter because they determine what “spin coupling” means operationally. In SCGVB, spin adaptation is not restricted to a single pairing pattern. In the wave-packet model, only one VB spin-coupling pattern is used, and the conclusion explicitly describes the method as involving “non-orthogonal valence-bond spin-coupling” rather than a fully general spin-coupled treatment (Ando, 2017). In GVB-BCCC, spin purity is preserved through singlet geminals and spin-complete block states, but there are no explicit general spin-coupling coefficients among many open-shell electrons (Wang et al., 2020). In spin-coupled MO theory, the nonorthogonality is shifted from orbitals within a single VB-type function to the state-interaction problem between distinct CSFs (Marti-Dafcik et al., 2024).

3. Chemical interpretation and bonding patterns

The chemical appeal of SCGVB and related spin-coupled descriptions lies in the fact that bonding is represented in terms of localized electrons, spin coupling, and distinct valence patterns rather than only through canonical delocalized MOs. The quantum-estimator study on $HC=SCE$6 makes this especially explicit. For the singlet sector of square and rectangular $HC=SCE$7, the SCGVB wavefunction has two spin-coupled structures. At the near-square geometry $HC=SCE$8 Å, the coefficients and Coulson–Chirgwin weights are

$HC=SCE$9

As one H–H distance increases, one structure becomes dominant, progressing through $$\Psi^{\mathrm{SCGVB}}=\mathcal N \sum_{k=1}^{f_N^S} c_{S,k}\,\psi^N_{S,M;k} =\mathcal N \sum_{k=1}^{f_N^S} c_{S,k}\,\hat A\,\phi_1\phi_2\cdots\phi_N\,\Theta^N_{S,M;k},$$0 at $$\Psi^{\mathrm{SCGVB}}=\mathcal N \sum_{k=1}^{f_N^S} c_{S,k}\,\psi^N_{S,M;k} =\mathcal N \sum_{k=1}^{f_N^S} c_{S,k}\,\hat A\,\phi_1\phi_2\cdots\phi_N\,\Theta^N_{S,M;k},$$1, $$\Psi^{\mathrm{SCGVB}}=\mathcal N \sum_{k=1}^{f_N^S} c_{S,k}\,\psi^N_{S,M;k} =\mathcal N \sum_{k=1}^{f_N^S} c_{S,k}\,\hat A\,\phi_1\phi_2\cdots\phi_N\,\Theta^N_{S,M;k},$$2 at $$\Psi^{\mathrm{SCGVB}}=\mathcal N \sum_{k=1}^{f_N^S} c_{S,k}\,\psi^N_{S,M;k} =\mathcal N \sum_{k=1}^{f_N^S} c_{S,k}\,\hat A\,\phi_1\phi_2\cdots\phi_N\,\Theta^N_{S,M;k},$$3, $$\Psi^{\mathrm{SCGVB}}=\mathcal N \sum_{k=1}^{f_N^S} c_{S,k}\,\psi^N_{S,M;k} =\mathcal N \sum_{k=1}^{f_N^S} c_{S,k}\,\hat A\,\phi_1\phi_2\cdots\phi_N\,\Theta^N_{S,M;k},$$4 at $$\Psi^{\mathrm{SCGVB}}=\mathcal N \sum_{k=1}^{f_N^S} c_{S,k}\,\psi^N_{S,M;k} =\mathcal N \sum_{k=1}^{f_N^S} c_{S,k}\,\hat A\,\phi_1\phi_2\cdots\phi_N\,\Theta^N_{S,M;k},$$5, and $$\Psi^{\mathrm{SCGVB}}=\mathcal N \sum_{k=1}^{f_N^S} c_{S,k}\,\psi^N_{S,M;k} =\mathcal N \sum_{k=1}^{f_N^S} c_{S,k}\,\hat A\,\phi_1\phi_2\cdots\phi_N\,\Theta^N_{S,M;k},$$6 at $$\Psi^{\mathrm{SCGVB}}=\mathcal N \sum_{k=1}^{f_N^S} c_{S,k}\,\psi^N_{S,M;k} =\mathcal N \sum_{k=1}^{f_N^S} c_{S,k}\,\hat A\,\phi_1\phi_2\cdots\phi_N\,\Theta^N_{S,M;k},$$7 (Gabrielly, 12 Mar 2026). The stated interpretation is bond reorganization and dissociation from a symmetry-enforced equal superposition toward separated $$\Psi^{\mathrm{SCGVB}}=\mathcal N \sum_{k=1}^{f_N^S} c_{S,k}\,\psi^N_{S,M;k} =\mathcal N \sum_{k=1}^{f_N^S} c_{S,k}\,\hat A\,\phi_1\phi_2\cdots\phi_N\,\Theta^N_{S,M;k},$$8 fragments.

The broader spin-coupled MO formulation presents an analogous picture for bond breaking in several archetypical systems. In $$\Psi^{\mathrm{SCGVB}}=\mathcal N \sum_{k=1}^{f_N^S} c_{S,k}\,\psi^N_{S,M;k} =\mathcal N \sum_{k=1}^{f_N^S} c_{S,k}\,\hat A\,\phi_1\phi_2\cdots\phi_N\,\Theta^N_{S,M;k},$$9, the state evolves from a delocalized MO closed shell toward a localized singlet diradical. In $f_N^S$0, sequential bond breaking is represented by a small number of valence configurations built from localized open-shell orbitals and explicit spin couplings, and the paper states that four CSFs recover $f_N^S$1 overlap for $f_N^S$2 while three CSFs give $f_N^S$3 overlap for $f_N^S$4 (Marti-Dafcik et al., 2024). The same framework describes dissociation limits in terms of high-spin fragment states coupled back to the required total molecular spin, which is a distinctly spin-coupled reading of strong correlation.

The wave-packet analogue on LiH shows the same qualitative logic in a more compressed form. Its localized nonorthogonal electron packets separate into two contracted Li $f_N^S$5 core packets, one H $f_N^S$6-like packet near the proton, and one labile Li $f_N^S$7-like packet. In the singlet ground state the Li $f_N^S$8 packet lies near H, whereas in the triplet state it is pushed behind Li relative to H (Ando, 2017). This preserves the central SCGVB intuition that singlet versus triplet coupling changes the optimal spatial localization pattern and therefore the bond character, even though the actual spin function is only perfect-pairing.

4. Matrix elements, orthogonalization, and structure weights

The dominant computational obstacle in SCGVB is the evaluation of overlap and off-diagonal Hamiltonian matrix elements between nonorthogonal structures. In orthogonal determinant methods, overlaps are trivial and Hamiltonian matrix elements can be reduced efficiently by Slater–Condon rules. In SCGVB, by contrast, one must evaluate matrix elements between determinants built from different nonorthogonal orbitals and different spin-coupled structures, and the paper on quantum estimators repeatedly identifies this as the major obstacle (Gabrielly, 12 Mar 2026).

For classical reference calculations, the nonorthogonal valence-bond basis can be transformed by Löwdin symmetric orthogonalization. If $f_N^S$9 and the orthonormality condition is

$N$0

the unique Hermitian choice is

$N$1

The coefficients transform as

$N$2

the orthogonalized Hamiltonian is

$N$3

and the generalized problem becomes the ordinary eigenproblem

$N$4

(Gabrielly, 12 Mar 2026). This does not remove the need to construct the original $N$5 and $N$6 matrices, but it regularizes the final diagonalization step.

Chemically interpretable postprocessing is typically done in terms of structure weights. For a nonorthogonal expansion

$N$7

the Chirgwin–Coulson weight is

$N$8

In an orthonormal basis this reduces to $N$9, but in the nonorthogonal case it includes interference from overlaps and is not strictly probabilistic; the paper notes that CC weights can become negative or exceed one when overlap effects are strong (Gabrielly, 12 Mar 2026). Löwdin weights and inverse weights are also given, the latter enforcing $S$0 (Gabrielly, 12 Mar 2026). The prominence of these weight definitions reflects a core SCGVB feature: chemically relevant interpretation depends on the metric of the nonorthogonal basis and cannot be reduced to squared expansion coefficients alone.

5. Quantum formulations of SCGVB

Recent quantum-computing work has focused on the specific algebraic burden created by SCGVB nonorthogonality rather than on preparing the full SCGVB state directly on hardware. The paper “Extension of the Jordan-Wigner mapping to nonorthogonal spin orbitals for quantum computing application to valence bond approaches” addresses the first obstacle by formulating a nonorthogonal Jordan–Wigner-type mapping in which the mixed anticommutation relation becomes

$S$1

with $S$2 the overlap matrix of nonorthogonal spin orbitals (Marruzzo et al., 16 Sep 2025). A key point is that the adjoint $S$3 is not a true annihilation operator in the usual occupation-number sense; applying it to an occupation vector produces a linear combination of occupation vectors weighted by orbital overlaps. The paper therefore introduces a biorthogonal partner basis and operators $S$4 satisfying

$S$5

which permits a Hamiltonian encoding with standard Jordan–Wigner-looking strings for the biorthogonal annihilators (Marruzzo et al., 16 Sep 2025). The same paper explicitly presents an $S$6 SCGVB example with two spin-coupling modes in the singlet sector and recovers chemically sensible Chirgwin–Coulson weights, including $S$7 at the square geometry $S$8 (Marruzzo et al., 16 Sep 2025).

The later measurement-driven framework “Compactifying the Electronic Wavefunction II: Quantum Estimators for Spin-Coupled Generalized Valence Bond Wavefunctions” keeps essentially all SCGVB algebra classical and uses quantum hardware only for determinant-level overlap and Hamiltonian primitives (Gabrielly, 12 Mar 2026). Determinant overlaps are rewritten as vacuum expectation values,

$S$9

and determinant Hamiltonian elements as

$c_{S,k}$0

After nonorthogonal Jordan–Wigner mapping and Pauli expansion, the determinant-overlap estimator reconstructs overlaps from all-zero computational-basis outcomes, while the Pauli-Grouped Hamiltonian Estimator measures grouped Pauli strings through local Clifford rotations and computational-basis sampling. The paper emphasizes that the circuits use a single system register, no ancilla qubits, no controlled operations, no entangling gates, and constant depth / shallow measurement circuits (Gabrielly, 12 Mar 2026). For square $c_{S,k}$1 the reported resource summary is 4 spatial orbitals, 8 spin-orbitals, 8 qubits, 86,906 total circuits, average circuit depth 2, maximum depth 4, and shots per circuit of $c_{S,k}$2 (Gabrielly, 12 Mar 2026).

These developments do not constitute a full quantum SCGVB algorithm in the modern orbital-variational sense. Rather, they separate the classical construction of the nonorthogonal spin-coupled problem from the measurement task on the quantum register. This suggests a hybrid route in which quantum hardware acts as a measurement engine for the hardest nonorthogonal matrix elements.

6. Scope, limitations, and likely developments

Several limitations recur across the literature. The measurement-driven SCGVB backend does not claim a formal quantum speedup; SCGVB algebra, determinant construction, Pauli expansion, and grouping remain heavily classical, and the paper reports that preliminary tests on larger systems such as $c_{S,k}$3 became computationally prohibitive at the preprocessing stage, exceeding practical multi-day supercomputing times (Gabrielly, 12 Mar 2026). The nonorthogonal Jordan–Wigner work explicitly solves only one of two major obstacles: a suitable mapping for nonorthogonal orbitals, not the construction of a chemically faithful VB/SCGVB ansatz with orbital optimization and direct spin-adapted state preparation (Marruzzo et al., 16 Sep 2025).

The wave-packet valence-bond model demonstrates how much of the SCGVB philosophy can survive in an extremely compact form, but it omits several ingredients of full SCGVB: no general spin coupling, no multiple VB structures or resonance expansion, one spherical Gaussian per electron, no explicit higher angular functions, excited states obtained by one-particle quantization on mean-field packet potentials, and mean-field/semiquantal dynamics rather than full many-electron antisymmetrized nonorthogonal wave-packet propagation (Ando, 2017). It is therefore best viewed as a wave-packet-based minimalist cousin of SCGVB rather than as SCGVB itself.

The post-reference correlation literature points in a different direction. GVB-BCCC is not a post-SCGVB method formally, because it is built on GVB-PP/APSG rather than on a general spin-coupled valence-bond reference, yet it shows that a localized, spin-pure, bond-pair reference can support an effective dynamic-correlation layer. The paper states that GVB-BCCC2b can provide highly comparable results as the density matrix renormalization group method for potential energy surfaces along simultaneous dissociation of all 12 C–C bonds in tridecane, even though the GVB reference function is qualitatively wrong for the studied processes (Wang et al., 2020). A plausible implication is that an analogous post-treatment built on a genuinely general SCGVB reference could be attractive, provided the nonorthogonal matrix-element bottleneck can be controlled.

Taken together, the current literature presents SCGVB as a compact, chemically transparent, spin-adapted nonorthogonal valence-bond framework whose main difficulty is not conceptual but algebraic. Its strength is the direct representation of localized electrons and explicit spin coupling; its burden is the metric structure of nonorthogonal orbitals and the resulting generalized secular problem. Recent work has therefore moved along three complementary lines: reduction to simplified but chemically intuitive analogues such as wave-packet perfect-pairing models (Ando, 2017), reformulation in orthogonal-orbital but nonorthogonal-state language (Marti-Dafcik et al., 2024), and hybrid classical–quantum machinery for overlap and Hamiltonian estimation in the full nonorthogonal setting (Marruzzo et al., 16 Sep 2025, Gabrielly, 12 Mar 2026).