Equivariant Cellular Sheaves for Molecular Electronic Structure: Bridging Sheaf Cohomology and E(3)-Equivariant Hamiltonian Learning

Krishna Harish

Source-linked AI summary

Equivariant Cellular Sheaves for Molecular Electronic Structure: Bridging Sheaf Cohomology and E(3)-Equivariant Hamiltonian Learning

Krishna Harish

arXiv:2608.23571v1Jun 14, 2026cs.LGphysics.chem-ph

TL;DR

Molecular electronic-structure learning has not connected equivariant Hamiltonian prediction with cellular-sheaf topology. The paper identifies shifted localized Hamiltonians with equivariant sheaf Laplacians, proving cohomological and expressivity results and validating the construction numerically, while limiting claims to the current evaluation scope.

Problem
Equivariant Hamiltonian prediction and cellular-sheaf topology have developed separately, leaving their structural connection unformalized for electronic structure.
Method
The paper builds a molecular cell complex and represents a positive-semidefinite shifted localized Hamiltonian as an E(3)- and permutation-equivariant cellular-sheaf Laplacian.
Results
The embedding and equivariance are exact to machine precision, cohomology reproduces non-bonding counts across eleven molecules, and the model improves accuracy and rotation robustness over a coordinate baseline.
Takeaways & Limitations
Sheaf cohomology provides chemical invariants for non-bonding orbitals and cycle structure while ECSN strictly extends equivariant message-passing and CW networks.
Takeaways & Limitations
The evaluation omits large-scale QM9, MD17, and self-consistent Hamiltonian benchmarks, leaving proposed benefits of ring 2-cells and PSD-local parameterization untested.

Abstract

from arXiv · show

Equivariant message-passing networks are the standard model for molecular property and interatomic-potential prediction, and recent work predicts the electronic Hamiltonian itself in an E(3)-equivariant way. Separately, topological deep learning has extended graph networks to cellular sheaves. Our central observation is structural: in a localized atomic-orbital basis, the molecular single-particle Hamiltonian, after a constant shift that makes it positive semidefinite, is the Laplacian of a cellular sheaf on a regular cell complex built from the molecule. Making the restriction maps O(3)-steerable two-center kernels from bond geometry recovers the Slater-Koster form as a special case and yields an E(3)- and permutation-equivariant operator. Three consequences follow. First, the zeroth sheaf cohomology H^0 = ker L is a topological invariant equal to the non-bonding (zero-mode) orbitals, recovering the classical alternant non-bonding-orbital count as a lower bound. Second, the Hodge 1-Laplacian lets higher cells (rings) carry cycle and delocalization information through H^1. Third, the model strictly generalizes E(3)-equivariant message-passing networks and CW networks, and inherits the anti-oversmoothing of non-trivial sheaf diffusion. We prove equivariance, expressivity, and cohomological-correspondence results for the Equivariant Cellular Sheaf Networks, and validate them numerically: the Hamiltonian-to-sheaf embedding is exact to machine precision, the cohomology dimension reproduces non-bonding-orbital counts across eleven conjugated molecules, the sheaf Laplacian is O(3)-equivariant to machine precision, and the equivariant model attains lower error and rotation generalization on a directional electronic target. Our contribution is this sheaf-theoretic formalization and its invariants, not equivariant Hamiltonian prediction itself.

1 Introduction

The paper connects equivariant electronic-structure learning with cellular sheaves by identifying shifted localized Hamiltonians with equivariant sheaf Laplacians. This formalization yields cohomological chemical invariants, a strict expressivity extension, and numerical validation.

A shifted localized-orbital Hamiltonian becomes a cellular-sheaf Laplacian on a molecular cell complex, connecting geometry, chemical topology, and electronic spectra.The resulting model is called the Equivariant Cellular Sheaf Network (ECSN).
Under a positive-semidefinite energy shift and per-bond factorization, the construction contains Slater–Koster tight binding as a special case.
O(3)-steerable stalks and restriction maps make the sheaf Laplacian E(3)- and permutation-equivariant.
Zeroth sheaf cohomology counts non-bonding orbitals and, for alternant systems, is lower bounded by sublattice imbalance.
ECSN strictly generalizes E(3)-equivariant message-passing and cellular networks while inheriting anti-oversmoothing from non-trivial sheaf diffusion.
Numerical validation finds machine-precision embedding and equivariance, correct cohomological counts across eleven molecules, and improved accuracy and rotation robustness over a coordinate baseline.
The paper’s novelty is the sheaf-theoretic formalization and its invariants, rather than equivariant Hamiltonian prediction itself.

2 Related Work

Prior work established equivariant molecular networks, equivariant Hamiltonian prediction, and topological or sheaf-based learning separately. This paper positions ECSN as a new reinterpretation that combines these directions for electronic structure.

Equivariant chemistry networks predict invariant or covariant atomic-graph targets, but do not use higher cells or sheaf structure.
PhiSNet, DeepH, and QHNet predict localized Hamiltonian matrices equivariantly; this paper newly reinterprets the predicted operator as a sheaf Laplacian.
Existing sheaf and cellular-complex methods are not O(3)-equivariant and are not applied to electronic structure.
Related molecular topology work uses density critical-point complexes or persistence-based descriptors, whereas this paper uses chemical cells and sheaf cohomology.

3 Background

The background constructs molecular regular cell complexes, defines cellular sheaves and their Hodge Laplacians, and introduces O(3)-steerable stalks and restriction maps for orbital representations.

3.1 Regular cell complexes from molecules: Atoms form 0-cells, cutoff pairs form unordered 1-cells, and chosen cycle-basis faces form 2-cells with oriented signed incidences.
3.2 Cellular sheaves and the sheaf Laplacian: A cellular sheaf assigns a finite-dimensional inner-product stalk to every cell and a linear restriction map to every incidence.
3.2 Cellular sheaves and the sheaf Laplacian: The degree-k Hodge–sheaf Laplacian is Lk = (δk)ᵀδk + δk−1(δk−1)ᵀ, with LF = L0 = (δ0)ᵀδ0 for vertices.
3.2 Cellular sheaves and the sheaf Laplacian: LF is symmetric positive semidefinite, and its kernel equals H^0(X; F), the space of global sections.
3.3 O(3)-equivariant sheaves: O(3) irreducible stalks model atomic orbitals by angular momentum, while direction-dependent steerable maps are built from Clebsch–Gordan contractions of spherical harmonics.
3.3 O(3)-equivariant sheaves: The steerable kernels use learnable weights and radial scalars depending on bond length.

4 Mathematical Framework

The framework identifies a PSD-shifted localized-orbital Hamiltonian with a cellular sheaf Laplacian and makes the construction E(3)- and permutation-equivariant. Its cohomology gives chemically meaningful non-bonding and cycle invariants.

Hamiltonian–sheaf correspondence: A localized Hamiltonian with sparse bond-localized blocks becomes a sheaf Laplacian after a PSD energy shift.The construction may use augmented stalks and satisfies LF = H − ErefI.
Hamiltonian–sheaf correspondence: Per-bond SVD factorization matches hopping blocks, while augmented self-incidences match residual on-site terms without violating positive semidefiniteness.The residual construction yields LF = H − ErefI under the PSD assumption.
Equivariance: O(3)-steerable restriction maps make the sheaf Laplacian E(3)- and permutation-equivariant, and equivariance passes to diffusion layers and invariant or covariant readouts.Translations act trivially on stalks, while rotations, reflections, and atom relabelings conjugate the operator appropriately.
Sheaf cohomology: Choosing Eref at the non-bonding level makes ker LF equal to the eigenspace of H at Eref and makes its dimension deformation-stable.Different reference energies produce sheaves related by stalk-wise orthogonal gauge transformations.
Sheaf cohomology: For alternant systems, dim H0 is lower bounded by the sublattice imbalance, recovering the classical non-bonding π-orbital count.The bound follows from the nullity of the off-diagonal bipartite Hamiltonian.
Sheaf cohomology: Benzene has dim H0 = 0, whereas cyclobutadiene and trimethylenemethane each have dim H0 = 2.The benzene result is consistent with equal bipartite sublattices and its closed-shell spectrum.
Higher cells: Adding ring 2-cells yields L1 = δ0(δ0)⊤ + (δ1)⊤δ1, whose harmonic space H1 captures independent cycles and non-trivial restriction-map closure failures.For the trivial sheaf, dim H1 equals the first Betti number; non-trivial sheaves can signal holonomy or frustration.

5 Proposed Method: Equivariant Cellular Sheaf Networks

ECSN learns geometry-conditioned steerable restriction maps, assembles their sheaf Laplacian, diffuses stalk features equivariantly, and supports invariant, covariant, and topological readouts.

Complex and sheaf construction: ECSN constructs a regular cell complex from molecular coordinates and species, then assigns atomic stalks and geometry-conditioned steerable restriction maps.The maps depend on invariant features and bond geometry through radial networks.
Operator assembly: The learned restriction maps are assembled into the sheaf Laplacian and optionally normalized as D−1/2LFD−1/2.Here D is the diagonal matrix formed from the vertex blocks of LF.
Sheaf diffusion: Equivariant sheaf diffusion uses gated nonlinearities acting only on O(3) norms, preserving the representation structure during feature propagation.Within-stalk and channel mixing use separate equivariance-preserving transformations.
Readout heads: Invariant readouts predict energies and gaps, covariant readouts output dipoles, forces, or Hamiltonian blocks, and topological readouts report non-bonding and cycle dimensions.The spectral gap is used to estimate the HOMO–LUMO gap, while the Hamiltonian target is recovered as a structured special case.

6 Theoretical Properties

ECSN has a strict expressivity hierarchy over standard MPNNs and CW networks, while non-trivial sheaf diffusion preserves structure that trivial diffusion smooths away.

Expressivity hierarchy: Scalar ECSN diffusion recovers CW-network updates and standard MPNN message passing on the molecular 1-skeleton.These reductions follow by restricting stalks and maps to scalar forms.
Expressivity hierarchy: Restricting to the 1-skeleton with diagonal steerable blocks recovers an E(3)-equivariant tensor-field MPNN.
Expressivity hierarchy: ECSN is strictly more expressive because non-trivial sheaves can have H0(X; F) = 0, unlike trivial-sheaf models whose kernels always contain constant sections.Such sheaves can distinguish complexes that trivial-sheaf models cannot.
Anti-oversmoothing: Trivial sheaf diffusion collapses features onto its constant-containing kernel, whereas trivial agreement spaces on cycles prevent this oversmoothing.For non-trivial sheaves, the Dirichlet energy is bounded below on the orthogonal complement of ker LF.
Complexity: With maximum stalk dimension d, assembling and applying LF over T layers costs O(T|E|d2) time and O(|E|d2) memory.An optional spectral readout adds an O(Nd)-dimensional sparse eigenproblem.

7 Experiments

Experiments validate the Hamiltonian-to-sheaf correspondence, cohomological counts, and O(3) equivariance to machine precision. On a synthetic directional target, the equivariant model is more accurate and robust to unseen rotations, while large-scale benchmarks remain unevaluated.

E1: Hamiltonian embedding: The scalar-stalk Hamiltonian-to-sheaf reconstruction is exact, with maxmol ∥LF − ˜H∥F = 0 across eleven π-conjugated molecules.For multi-orbital benzene systems, reconstruction error is at most 8.4 × 10−15.
E2: Cohomology counts: Computed dim H0 reproduces known non-bonding π-orbital counts across eleven conjugated molecules, including zero, one, and two zero-mode cases.The bipartite lower bound holds for every bipartite system and is tight for odd alternants and trimethylenemethane.
E3: Equivariance: Across 200 random rotations, the relative equivariance error is 8.1 × 10−16, and reflection error is exactly 0.Replacing steerable maps with raw-bond-vector maps raises the error to 0.58.
E4: Learning benefit: At N = 160, the equivariant model is 58% more accurate on rotated inputs than the coordinate model for the HOMO–LUMO gap target.Its MAE decreases from 0.32 to 0.11 as N grows from 20 to 160, while the coordinate model plateaus near 0.22 and reaches 0.27–0.30 on rotated molecules.
Scope of evaluation: The evaluation omits large-scale QM9, MD17, and self-consistent Hamiltonian benchmarks, leaving ring-cell H1 benefits and full Hamiltonian-regression data efficiency as predictions.The paper explicitly states these are predictions, not results.

8 Limitations

The framework has reference-energy, gauge, cycle-basis, computational-cost, and single-particle scope constraints. Its directional-target evaluation also compares only the stated coordinate baseline and setting.

Reference energy: A poor reference energy Eref can move the chemically meaningful kernel after the required PSD shift.Choosing Eref at or below the spectrum is required; selecting the non-bonding level makes the kernel chemically meaningful.
Gauge freedom: Restriction maps are identifiable only up to a stalk-wise orthogonal gauge, making the Laplacian better posed for supervision than individual maps.The Laplacian is gauge-invariant, whereas direct supervision on individual restriction maps is complicated.
Cycle representation: Cycle bases for 2-cells are non-unique, so results should use a fixed canonical choice such as the smallest set of smallest rings.
Evaluation setting: The directional-gap comparison uses no rotation augmentation and contrasts the equivariant sheaf model with a coordinate model that degrades on rotated test molecules.
Cost and scope: Sheaf assembly costs d2 relative to scalar GNNs, and genuinely correlated multireference systems remain outside the single-particle model's scope.The stated assumption is single-particle mean-field electronic structure; many-body extensions on higher cells would be required for correlated systems.

9 Conclusion

The paper formalizes localized-orbital Hamiltonians as equivariant cellular-sheaf Laplacians, linking electronic structure with molecular topology. Its consequences include cohomological invariants, strict expressivity generalization, and anti-oversmoothing, while the novelty is the formalization rather than Hamiltonian prediction itself.

The localized-orbital electronic Hamiltonian becomes an E(3)- and permutation-equivariant sheaf Laplacian, with Slater–Koster tight binding as a special case.
Sheaf cohomology supplies topological invariants that count non-bonding orbitals and detect cycle structure in molecular cell complexes.
ECSN strictly generalizes equivariant MPNNs and CW networks while inheriting the anti-oversmoothing guarantee of non-trivial sheaf diffusion.
The framework unifies topological deep learning with equivariant electronic-structure learning and presents sheaf cohomology as a language for chemical-bonding topology.
The contribution is the sheaf-theoretic formalization and its invariants, not equivariant Hamiltonian prediction itself.