Analytical Nuclear Gradients for State-Averaged Configuration Interaction Singles Variants: Application to Conical Intersections

Takashi Tsuchimochi

arxiv: 2602.15700 · v2 · submitted 2026-02-17 · ⚛️ physics.chem-ph

Analytical Nuclear Gradients for State-Averaged Configuration Interaction Singles Variants: Application to Conical Intersections

Takashi Tsuchimochi This is my paper

Pith reviewed 2026-05-15 21:46 UTC · model grok-4.3

classification ⚛️ physics.chem-ph

keywords analytical nuclear gradientsstate-averaged CISconical intersectionsminimum-energy crossing pointsorbital optimizationconfiguration interaction singlesgeometry optimizationexcited states

0 comments

The pith

Analytical nuclear gradients for state-averaged CIS variants enable efficient minimum-energy conical intersection searches at mean-field cost.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper derives analytical nuclear gradients for state-averaged orbital-optimized configuration interaction singles (SACIS) and its spin-projected extension (SAECIS). These gradients support geometry optimization and minimum-energy conical intersection (MECX) searches within a low-cost CIS-based framework. For twisted-pyramidalized ethylene, both methods reproduce the correct conical intersection topology, unlike conventional CIS. Benchmarks on twelve MECXs show that both reproduce geometries with mean RMSDs below 0.1 Å relative to high-level references. SACIS captures the degeneracy through variational orbital relaxation that alleviates Hartree-Fock bias and incorporates static correlation via localization effects, making spin projection non-essential for most cases.

Core claim

State-averaged orbital-optimized CIS captures the essential degeneracy at conical intersections through variational orbital relaxation, which alleviates ground-state Hartree-Fock orbital bias and effectively incorporates static correlation through localization effects. The Lagrangian formulation of the analytical nuclear gradients, with explicit removal of null-space contributions in the coupled perturbed equations, ensures numerically stable gradients that enable black-box, qualitatively reliable descriptions of these intersections at mean-field computational cost.

What carries the argument

The Lagrangian approach to analytical nuclear gradients for SACIS and SAECIS, which removes null-space contributions in the coupled perturbed equations to ensure numerical stability.

Load-bearing premise

The state-averaged orbital optimization and Lagrangian gradient formulation remain numerically stable and qualitatively correct for conical intersections even when higher double-excitation character is present, without post-hoc adjustments.

What would settle it

A benchmark calculation on one of the twelve MECXs or a new system with strong double-excitation character where SACIS or SAECIS yields an RMSD exceeding 0.1 Å or fails to locate a degenerate point.

read the original abstract

We derive analytical nuclear gradients for state-averaged orbital-optimized configuration interaction singles (SACIS) and its spin-projected extension (SAECIS), enabling efficient geometry optimization and minimum-energy conical intersection (MECX) searches within a low-cost CIS-based framework. The formulation employs a Lagrangian approach and explicitly removes null-space contributions in the coupled perturbed equations to ensure numerically stable gradients. For twisted-pyramidalized ethylene, both SACIS and SAECIS qualitatively reproduce the correct conical intersection topology, in sharp contrast to conventional CIS and ECIS. Benchmark calculations on twelve MECXs demonstrate that both methods reproduce geometries with mean RMSDs below 0.1~{\AA} relative to high-level reference methods. SACIS captures the essential degeneracy through variational orbital relaxation, which alleviates ground-state Hartree--Fock (HF) orbital bias and effectively incorporates static correlation through localization effects; notably, spin projection is found to be non-essential for the qualitative description of these intersections. Overall, SACIS and SAECIS provide qualitatively reliable CX descriptions at mean-field computational cost in a black-box manner. Given their comparable accuracy and the additional overhead associated with spin projection, SACIS offers a more favorable cost-performance balance for general applications, whereas SAECIS may become advantageous when higher excited states with significant double-excitation character are involved.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

This derives analytical gradients for SACIS/SAECIS that let you locate conical intersections reliably at low cost, with benchmarks showing close geometry matches to higher methods.

read the letter

The key point is that the authors worked out analytical nuclear gradients for state-averaged orbital-optimized CIS and its spin-projected version. They use a Lagrangian setup and strip out null-space terms in the coupled-perturbed equations to keep the gradients stable during optimizations and MECX searches. For twisted ethylene this recovers the correct intersection topology where plain CIS fails, and across twelve benchmark cases the optimized geometries sit within 0.1 Å RMSD of high-level references on average. SACIS alone handles the degeneracy through orbital relaxation, and the extra spin projection turns out to be unnecessary for most of these examples, which keeps the cost down.

Referee Report

1 major / 0 minor

Summary. The manuscript derives analytical nuclear gradients for state-averaged orbital-optimized configuration interaction singles (SACIS) and its spin-projected extension (SAECIS) using a Lagrangian approach that explicitly removes null-space contributions from the coupled-perturbed equations for numerical stability. These gradients are applied to geometry optimizations and minimum-energy conical intersection (MECX) searches. Benchmarks on twelve MECXs show mean RMSDs below 0.1 Å relative to high-level references, with SACIS reproducing correct topologies for cases like twisted-pyramidalized ethylene through variational orbital relaxation.

Significance. If validated, this provides a low-cost, black-box method for exploring conical intersections in systems where multireference methods are too expensive. The explicit handling of orbital relaxation and degeneracy in the gradient formulation is a notable technical contribution, supported by the reported benchmark accuracies.

major comments (1)

The Lagrangian formulation removes null-space contributions to ensure stability. At conical intersections, the two states are degenerate, so the null space includes relative orbital rotations between them. Projecting these out may affect components in the branching plane. The ethylene example reproduces correct topology, but no direct comparison of gradients or geometries with versus without the projection is provided, which is needed to confirm the approximation does not alter the effective nuclear gradients.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the positive assessment of our work and the constructive comment on the stability of the Lagrangian formulation. We address the concern below and will revise the manuscript accordingly.

read point-by-point responses

Referee: The Lagrangian formulation removes null-space contributions to ensure stability. At conical intersections, the two states are degenerate, so the null space includes relative orbital rotations between them. Projecting these out may affect components in the branching plane. The ethylene example reproduces correct topology, but no direct comparison of gradients or geometries with versus without the projection is provided, which is needed to confirm the approximation does not alter the effective nuclear gradients.

Authors: We appreciate the referee's careful reading and the valid point raised about the null-space projection at degeneracies. In the Lagrangian approach, the null-space components correspond to relative orbital rotations between the degenerate states; these directions have zero eigenvalues in the coupled-perturbed equations and therefore do not contribute to the physical nuclear gradients or to the branching-plane vectors. Removing them is strictly a numerical stabilization step that leaves the effective gradient unchanged. The fact that both SACIS and SAECIS recover the correct conical-intersection topology for twisted-pyramidalized ethylene (in contrast to standard CIS) and yield mean RMSDs below 0.1 Å across the twelve-molecule benchmark set already indicates that the projection preserves the essential physics. Nevertheless, we acknowledge that an explicit side-by-side comparison was not provided in the original manuscript. In the revised version we will add a short supplementary analysis for the ethylene MECX that reports the gradient norms and optimized geometries obtained with and without the projection; the differences are expected to be negligible (<0.01 Å RMSD) and the branching-plane vectors identical within numerical precision. This addition will directly address the referee's request. revision: yes

Circularity Check

0 steps flagged

No circularity: Lagrangian gradient derivation is independent of fitted data or self-citations

full rationale

The paper presents an explicit derivation of analytical nuclear gradients for SACIS/SAECIS using a Lagrangian formulation with null-space removal in the coupled-perturbed equations. No step reduces a claimed prediction or result to a fitted parameter, self-defined quantity, or load-bearing self-citation whose validity depends on the present work. Benchmarks compare to external high-level references rather than internal fits, and the topology reproduction for ethylene is shown via direct computation rather than by construction. The derivation chain is self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on standard quantum chemistry assumptions plus the new gradient derivation; no free parameters or invented entities are introduced beyond the existing SACIS framework.

axioms (2)

domain assumption Born-Oppenheimer approximation for separating nuclear and electronic degrees of freedom
Invoked implicitly for nuclear gradients in electronic structure calculations.
domain assumption Validity of state-averaged orbital optimization for capturing static correlation at conical intersections
Central to why SACIS succeeds where HF-based CIS fails.

pith-pipeline@v0.9.0 · 5539 in / 1189 out tokens · 42774 ms · 2026-05-15T21:46:59.231633+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

The formulation employs a Lagrangian approach and explicitly removes null-space contributions in the coupled perturbed equations to ensure numerically stable gradients.

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

TD$\Delta$SCF: Time-Dependent Density Functional Theory with a Non-Aufbau Reference for near-degenerate states
physics.chem-ph 2026-03 unverdicted novelty 7.0

TDΔSCF performs TDDFT linear response on a non-Aufbau ΔSCF reference to describe near-degenerate singlet states with less functional sensitivity and fewer spurious states than collinear SF-TDDFT.