Path-ordered linked product approximation to the global electronic overlap matrix

Bing Gu; Yujuan Xie

REVIEW 2 major objections 5 minor 1 cited by

Path-ordered linked product approximation to the global electronic overlap matrix

T0 review · 2 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The global electronic overlap matrix can be assembled as a path-ordered product of nearest-neighbor overlap links, cutting the cost of exact nonadiabatic dynamics.

desk verdict A clean, parameter-free approximation that cuts the main bottleneck of LDR overlap matrix construction; validation is thin but the idea is sound and deserves serious review. read the letter →

arxiv 2501.05003 v1 pith:HMTOURPR submitted 2025-01-09 physics.chem-ph

Yujuan Xie , Bing Gu This is my paper · ORCID

classification physics.chem-ph

keywords globalelectronicoverlapmatrixlocaldiabaticrepresentationconicalintersectiondynamicsgeometricphasepath-orderedproductnearest-neighborproton-coupledelectrontransfernonadiabatic

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper claims that the global many-electron overlap matrix—the object that encodes all beyond-Born-Oppenheimer effects in the local diabatic representation—can be reconstructed from nearest-neighbor electronic overlaps alone. Specifically, the overlap between two distant nuclear geometries is approximated by a path-ordered product of overlap matrices along a shortest grid path connecting them. This drops the electronic-structure cost from $O(n^{2d})$ to $O(d n^d)$ for a $d$-dimensional grid with $n$ points per dimension. In a numerical test on a two-dimensional proton-coupled electron transfer model with a conical intersection, the approximate overlap matrix visibly differs from the exact one for long-range pairs, yet the resulting population dynamics, proton position, and geometric-phase node are almost exactly reproduced. If the approximation is general, it removes the main computational bottleneck of an exact, singularity-free nonadiabatic dynamics framework.

What carries the argument

The machinery is the path-ordered linked product: each 'link' is a nearest-neighbor overlap matrix $L_{n,\pm j} \equiv A_{n,n\pm e_j}$, and the global overlap is built by multiplying links along a path. A recursive construction, Eq. (15), assembles the $d$-dimensional global overlap matrix from one-dimensional links, so only links require electronic-structure calculations. The path-ordering operator $P_\gamma$ and the insertion of electronic projection operators $\hat{P}_n$ plus neglected complements $\hat{Q}_n$ supply the formal justification, while the path dependence is tested by comparing two shortest paths.

What would settle it

A decisive test would be to compute the exact global overlap matrix for a model with a third adiabatic state that approaches an intermediate geometry along a link path, then compare the linked-product approximation and the resulting conical-intersection dynamics. If the population curves or the geometric-phase node change measurably when the third state is included, the two-state truncation that makes Eq. (10) practical is not generally safe.

Watch

Extended reading notes

Core claim

The central discovery is Eq. (10): $A_{mn} \approx P_\gamma \prod_{k=0}^{L-1} A_{\gamma_k,\gamma_{k+1}}$, where $A_{mn}$ is the overlap matrix between adiabatic electronic states at geometries $R_m$ and $R_n$, $P_\gamma$ orders the product along a path, and each factor is an overlap matrix between nearest-neighbor grid points. The derivation inserts electronic identities along the path and drops the complementary projection $\hat{Q}$ at each intermediate geometry, which becomes exact in a complete electronic basis. The paper shows that in the two-state test model the approximate matrix is globally phase-consistent, reproduces the random $\pm 1$ phase structure, and yields conical-intersection dynamics in almost exact agreement with the exact overlap matrix, including the geometric-phase node in the nuclear wave packet.

Load-bearing premise

The approximation assumes that electronic states not included in the small active set make negligible contributions at every intermediate geometry along the path; if a higher-lying state participates, the missing piece has no error bound.

Editorial extensions

If this is right

Electronic structure calculations are needed only for nearest-neighbor geometry pairs, reducing the overlap-matrix cost from $O(n^{2d})$ to $O(d n^d)$.
The approximate overlap matrix remains globally phase-consistent, so the geometric phase accumulated around a loop is carried by the short-range links without gauge fixing.
Conical-intersection population dynamics, proton position, and the geometric-phase node in the wave packet are reproduced almost exactly despite visible long-range differences in the overlap matrix.
The path dependence of the approximation is immaterial for the dynamics in the tested model, so any shortest path between two geometries can be used.
The recursive construction extends the approximation to higher-dimensional grids while keeping the computational gain per added dimension.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Beyond the paper, the same link-product construction should be tested with larger active spaces and more than two electronic states; the error will likely grow as higher-lying states acquire physical weight at intermediate geometries.
The success of the approximation suggests that conical-intersection dynamics is insensitive to errors in long-range overlap elements, a statement about dynamical averaging that could be probed directly by comparing exact and approximate long-range blocks.
Because the approximation is path-dependent at the level of matrix elements but path-independent in the observed dynamics, it may be possible to average over multiple paths to estimate the error without computing the exact global overlap matrix.
The approach may also be combined with on-the-fly electronic structure, since only nearest-neighbor overlaps are needed; a sparse, link-based global overlap could be assembled without storing all pairs.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The linked product approximation is a genuinely useful idea. The paper replaces the prohibitive O(n^{2d}) cost of building the global electronic overlap matrix in the local diabatic representation with O(d n^d) nearest-neighbor overlap calculations, then reconstructs long-range overlaps by path-ordered products of links. That is a real bottleneck, and this is a simple, parameter-free way to attack it. The derivation via insertion of electronic resolutions of identity and neglect of complementary projectors is transparent, and the recursive dimension-by-dimension implementation is a nice practical touch. The paper also earns credit for showing the approximation is exact in the complete-basis limit and for demonstrating that it captures the geometric phase node in the Shin-Metiu wave packet.

The numerical evidence, however, is thinner than the title suggests. The validation is one 2D model with two electronic states, and the authors report average differences of ~0.03 and maximum differences ~0.31 between approximate and exact overlap matrices. The claim that the dynamics is "almost exact" rests on visual overlap of population and position expectation values, not on quantitative error metrics. The path dependence of the approximation is acknowledged but not controlled; the fact that the two paths give identical dynamics is reassuring but is a single observation. The stress-test concern about neglected complementary-projector terms is legitimate: for larger active spaces or longer paths the dropped terms could matter, and the paper offers no error bound. That does not fatally undermine the paper, but it does mean transferability is unproven.

None of this is disqualifying for a first methods paper. The approximation is new, the computational savings are real, and the single model is at least a nontrivial test with a conical intersection. What the paper needs before broad claims are made: code and data release, quantitative dynamics errors instead of "almost exact," and tests on a three-state or larger system with a longer path. The authors are also candid about the approximation's limitations, which I appreciate.

This deserves a serious referee. It is the kind of paper where a careful referee can push for the extra validation without rejecting the core idea. I would cite it if I worked on nonadiabatic dynamics, and I would bring it to a reading group to discuss the path-ordering construction, but I would not take the transferability on faith.

Referee Report

2 major / 5 minor

Summary. The paper proposes a linked product approximation for the global many-electron overlap matrix used in the discrete variable local diabatic representation (LDR). For non-nearest-neighbor nuclear configurations, the overlap matrix is approximated by a path-ordered product of nearest-neighbor overlap matrices along a connecting path [Eq. (10)]. This reduces the electronic-structure cost for constructing the overlap matrix from O(n^{2d}) to O(d n^d). The approximation is derived by inserting electronic resolutions of identity along the path and neglecting all terms containing complementary projectors [Eq. (11)]. The approach is validated on a two-dimensional Shin-Metiu model with two electronic states, showing that although the approximate overlap matrix differs from the exact one (especially for long-range elements), the resulting conical-intersection population dynamics, proton position, and geometric-phase node are in nearly exact agreement with the reference calculations. The path dependence of the approximation is quantified (average difference ~0.03, maximum ~0.31) and is reported to be immaterial for the dynamics in this model.

Significance. If the approximation is transferable, it directly addresses the main computational bottleneck of the LDR method and could enable exact nonadiabatic dynamics for systems with more nuclear degrees of freedom. The derivation is transparent, the approximation is parameter-free, it is exact in the complete-basis limit, and it preserves the geometric-phase structure encoded in the nearest-neighbor links. The numerical demonstration on the Shin-Metiu model is a useful proof of concept. However, the central claim is currently supported by only one two-state, two-dimensional model, and no quantitative control is provided for the neglected complementary-projector terms. The significance is therefore conditional on additional analysis or benchmarks establishing that the error remains small for larger active spaces, longer paths, and more complex electronic structures.

major comments (2)

[Section II.B, Eqs. (10)-(11)] The approximation neglects all terms containing the complementary projectors Q(R_k) in Eq. (11) with no estimate or bound for the dropped contributions. At each link the error involves quantities such as ||Q(R_k)|phi_alpha(R_{k-1})>||, and because the approximation is multiplicative over the path, errors can accumulate with path length. The Shin-Metiu demonstration uses only two active states well separated from higher states in the sampled region, so it does not establish that the truncation is safe when a third state approaches the crossing or when the wave packet accesses regions with small gaps to higher states. Please add a quantitative error analysis, for example by computing the norm of Q-projected states along representative paths or by benchmarking with a larger active space.
[Section III, Figs. 4-6] The validation is restricted to a single two-dimensional model with two electronic states. The path dependence shown in Fig. 5 (average difference ~0.03, maximum ~0.31) confirms that different shortest paths give different approximate overlap matrices, yet the manuscript only claims that this path dependence is immaterial for the particular dynamics studied. Longer paths, higher-dimensional grids, and more than two electronic states may amplify the uncontrolled error of Eq. (10). To support the general cost-scaling claim, the authors should test at least one additional case with more electronic states or more grid points along the path, and report the path dependence of the observables there.

minor comments (5)

[Section II.B, Eq. (11)] The expression after 'inserting the electronic identity' contains an undefined index M and is not displayed as a well-formed product; please rewrite Eq. (11) to unambiguously show the ordered product of link matrices and the placement of projectors.
[Section II.C, Eq. (15)] The recursive relation would benefit from an explicit statement of matrix dimensions and index ranges, since A(d) is a matrix over nuclear grid indices and electronic state indices.
[Section III, Fig. 5] Please state which matrix norm or elementwise statistic is used for the average (~0.03) and maximum (~0.31) differences.
[Section III, text after Eq. (20)] Change 'What'more' to 'Moreover'.
[Section IV] The statement that the approximate overlap matrix 'perfectly matches' short-range values is expected because nearest-neighbor overlaps are exact inputs; consider quantifying the short-range error as a function of grid distance instead.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the linked product approximation is an uncontrolled but non-circular truncation of an identity insertion, validated against independently computed exact overlaps.

full rationale

The paper's central claim, Eq. (10), is an ansatz: the global overlap is approximated by a path-ordered product of nearest-neighbor overlaps. The supporting derivation, Eq. (11), inserts the electronic identity P_k + Q_k at each intermediate geometry and then drops all terms containing complementary projectors Q_k. This is a controlled truncation of an exact expansion in the finite active space, not a renaming of inputs or a fitted result. No parameters are fitted; the nearest-neighbor links are exact electronic-structure inputs, and the approximate long-range overlaps are compared against reference overlap matrices computed independently in the same simulation. The paper explicitly acknowledges the approximation's path dependence (Fig. 5) and its deterioration for long-range geometries, which further confirms that the result is an honest numerical approximation rather than a tautology. The authors cite their own prior LDR framework papers [10,13,14], but these citations establish the propagation framework and the role of the overlap matrix; the linked product approximation itself does not reduce to those citations. The main limitation is the absence of an error bound for the dropped Q terms, but that is a correctness/robustness concern, not a circularity. Overall, the derivation chain is self-contained apart from routine framework citations, and no step equates the conclusion to the inputs by construction.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

No free parameters are fitted to the target dynamics, and no invented entities appear. The approach relies on the domain assumption that the included electronic states form a sufficient active space (Q neglected) and that short L1 paths capture the geometric phase. Both are reasonable but untested beyond one model.

assumptions (4)

domain assumption The active electronic subspace used in the overlap matrices is closed enough that the complementary projection operators Q_n in Eq. (11) can be neglected.
This is the defining approximation of the method; without it the product of links is not exact. The paper validates it only for two excited states of the Shin-Metiu model.
domain assumption A nuclear DVR grid with 63 points per dimension in [-3, 3] Bohr resolves the geometry dependence of the electronic overlap sufficiently well.
The accuracy of the linked product approximation depends on nearest-neighbor overlaps being representative; no convergence study with grid density is reported.
ad hoc to paper The L1 shortest path and the two selected path orderings (Path A and Path B) are representative, and the dynamics are path-independent.
Only two paths are compared, in one model; path independence of the dynamics is asserted from this single example.
standard math Standard DVR, electronic structure, and Strang splitting results are used as background.
These are established tools cited by the paper and not derived here.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Path-ordered linked product approximation to the global electronic overlap matrix." pith.science (2026). https://pith.science/paper/HMTOURPR

@misc{pith2026250105003,
  author       = {Pith},
  title        = {Pith review of: Path-ordered linked product approximation to the global electronic overlap matrix},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HMTOURPR}},
  note         = {Machine review of arXiv:2501.05003}
}

read the original abstract

The global many-electron wave function overlap matrix accounts for all effects beyond the Born-Oppenheimer approximation in the discrete variable local diabatic representation, a numerically exact framework for modeling nonadiabatic conical intersection wave packet dynamics. Nevertheless, calculating the electronic overlap matrix from electronic structure is computationally expensive. Here, we introduce an approximation for constructing the electronic overlap matrix between any two long-range geometries by the product of nearest-neighbor overlap matrices along a path connecting these two geometries. This approximation significantly reduces the computational effort by only requiring electronic structure calculations for the nearest-neighbor overlap matrices. The accuracy of this approximation is demonstrated through an exact simulation of a proton-coupled electron transfer model. Our results show that although the approximate overlap matrix can exhibit noticeable differences from the exact ones, the conical intersection dynamics is in almost exact agreement with those from the exact overlap matrix.

Figures

Figures reproduced from arXiv: 2501.05003 by the authors.

**Figure 3.** FIG. 3. The schematic and the adiabatic potential en [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

**Figure 2.** FIG. 2. Schematic of calculating the global electronic overlap [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

Figures from the paper (4 more)

**Figure 5.** Figure 5: FIG. 5. The approximate electronic overlap matrices of the [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]

**Figure 4.** Figure 4: FIG. 4. The exact (left) and approximate (middle) overlap [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]

**Figure 6.** Figure 6: FIG. 6. (a) The electronic population on two excited states [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]

**Figure 7.** Figure 7: FIG. 7. The nuclear wave packet distribution of the first (top, [PITH_FULL_IMAGE:figures/full_fig_p006_7.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

Topological Quantum Molecular Dynamics
physics.chem-ph 2025-05 conditional novelty 5.0 of 10

Molecular quantum dynamics can be written so that all effects beyond the Born-Oppenheimer approximation are carried by the overlap between electronic states at different nuclear geometries.

Reference graph

Works this paper leans on

22 extracted references · 22 canonical work pages · cited by 1 Pith paper

[1]

Domcke, D

W. Domcke, D. R. Yarkony, and H. K¨ oppel, Coni- cal Intersections: Theory, Computation and Experiment (World Scientific, 2011)

work page 2011
[2]

Larson, E

J. Larson, E. Sj¨ oqvist, and P. ¨Ohberg, Conical Intersec- tions in Physics: An Introduction to Synthetic Gauge Theories, Lecture Notes in Physics, Vol. 965 (Springer International Publishing, Cham, 2020)

work page 2020
[3]

H. J. W¨ orner, J. B. Bertrand, B. Fabre, J. Higuet, H. Ruf, A. Dubrouil, S. Patchkovskii, M. Spanner, Y. Mairesse, V. Blanchet, E. M´ evel, E. Constant, P. B. Corkum, and D. M. Villeneuve, Conical Intersection Dynamics in NO2 Probed by Homodyne High-Harmonic Spectroscopy, Sci- ence 334, 208 (2011)

work page 2011
[4]

Kowalewski and S

M. Kowalewski and S. Mukamel, Manipulating molecules with quantum light, Proc. Natl. Acad. Sci. 114, 3278 (2017)

work page 2017
[5]

Gu and S

B. Gu and S. Mukamel, Cooperative Conical Intersec- tion Dynamics of Two Pyrazine Molecules in an Optical Cavity, J. Phys. Chem. Lett. 11, 5555 (2020)

work page 2020
[6]

Gu and S

B. Gu and S. Mukamel, Manipulating nonadiabatic con- ical intersection dynamics by optical cavities, Chem. Sci. 11, 1290 (2020)

work page 2020
[7]

L.-P. Yang, F. Khosravi, and Z. Jacob, Quantum spin operator of the photon, arXiv:2004.03771 [cond- mat, physics:hep-th, physics:physics, physics:quant- ph] (2020), arXiv:2004.03771 [cond-mat, physics:hep-th, physics:physics, physics:quant-ph]

work page Pith review arXiv 2020
[8]

C. Xie, J. Ma, X. Zhu, D. R. Yarkony, D. Xie, and H. Guo, Nonadiabatic Tunneling in Photodissociation of Phenol, J. Am. Chem. Soc. 138, 7828 (2016)

work page 2016

Show all 22 references

[9]

S. Han, C. Xie, X. Hu, D. R. Yarkony, H. Guo, and D. Xie, Quantum Dynamics of Photodissociation: Re- cent Advances and Challenges, J. Phys. Chem. Lett. 14, 10517 (2023)

2023
[10]

Zhu and B

X. Zhu and B. Gu, Making Peace with Random Phases: Ab Initio Conical Intersection Quantum Dynamics in Random Gauges, J. Phys. Chem. Lett. 15, 8487 (2024)

2024
[11]

Aleotti, D

F. Aleotti, D. Aranda, M. Yaghoubi Jouybari, M. Gar- avelli, A. Nenov, and F. Santoro, Parameterization of a linear vibronic coupling model with multiconfigura- tional electronic structure methods to study the quantum dynamics of photoexcited pyrene, J. Chem. Phys. 154, 104106 (2021)

2021
[12]

Y. Shu, Z. Varga, S. Kanchanakungwankul, L. Zhang, and D. G. Truhlar, Diabatic States of Molecules, J. Phys. Chem. A 126, 992 (2022)

2022
[13]

Gu, A Discrete-Variable Local Diabatic Representa- tion of Conical Intersection Dynamics, J

B. Gu, A Discrete-Variable Local Diabatic Representa- tion of Conical Intersection Dynamics, J. Chem. Theory Comput. 19, 6557 (2023)

2023
[14]

Gu, Nonadiabatic Conical Intersection Dynamics in the Local Diabatic Representation with Strang Splitting and Fourier Basis, J

B. Gu, Nonadiabatic Conical Intersection Dynamics in the Local Diabatic Representation with Strang Splitting and Fourier Basis, J. Chem. Theory Comput. 20, 2711 (2024)

2024
[15]

Shin and H

S. Shin and H. Metiu, Nonadiabatic effects on the charge transfer rate constant: A numerical study of a simple model system, The Journal of Chemical Physics 102, 9285 (1995)

1995
[16]

Shin and S.-I

S. Shin and S.-I. Cho, Quantum dynamics of model proton-coupled electron transfer reactions, Chem. Phys. 259, 27 (2000)

2000
[17]

J. C. Light and T. Carrington Jr., Discrete-Variable Rep- resentations and their Utilization, in Advances in Chemi- cal Physics(John Wiley & Sons, Ltd, 2000) pp. 263–310

2000
[18]

R. G. Littlejohn, M. Cargo, T. Carrington, K. A. Mitchell, and B. Poirier, A general framework for dis- crete variable representation basis sets, J. Chem. Phys. 116, 8691 (2002)

2002
[19]

Y. Gu, H. Yong, B. Gu, and S. Mukamel, Chemical bond reorganization in intramolecular proton transfer revealed by ultrafast X-ray photoelectron spectroscopy, Proceedings of the National Academy of Sciences 121, e2321343121 (2024)

2024
[20]

Lu and J

Y. Lu and J. Gao, Multistate Density Functional Theory of Excited States, J. Phys. Chem. Lett. 13, 7762 (2022)

2022
[21]

Y. Yang, L. Shen, D. Zhang, and W. Yang, Conical Intersections from Particle–Particle Random Phase and Tamm–Dancoff Approximations, J. Phys. Chem. Lett. 7, 2407 (2016)

2016
[22]

S. K. Min, A. Abedi, K. S. Kim, and E. K. U. Gross, Is the Molecular Berry Phase an Artifact of the Born- Oppenheimer Approximation?, Phys. Rev. Lett. 113, 263004 (2014)

2014

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.

[1] [1]

Domcke, D

W. Domcke, D. R. Yarkony, and H. K¨ oppel, Coni- cal Intersections: Theory, Computation and Experiment (World Scientific, 2011)

work page 2011

[2] [2]

Larson, E

J. Larson, E. Sj¨ oqvist, and P. ¨Ohberg, Conical Intersec- tions in Physics: An Introduction to Synthetic Gauge Theories, Lecture Notes in Physics, Vol. 965 (Springer International Publishing, Cham, 2020)

work page 2020

[3] [3]

H. J. W¨ orner, J. B. Bertrand, B. Fabre, J. Higuet, H. Ruf, A. Dubrouil, S. Patchkovskii, M. Spanner, Y. Mairesse, V. Blanchet, E. M´ evel, E. Constant, P. B. Corkum, and D. M. Villeneuve, Conical Intersection Dynamics in NO2 Probed by Homodyne High-Harmonic Spectroscopy, Sci- ence 334, 208 (2011)

work page 2011

[4] [4]

Kowalewski and S

M. Kowalewski and S. Mukamel, Manipulating molecules with quantum light, Proc. Natl. Acad. Sci. 114, 3278 (2017)

work page 2017

[5] [5]

Gu and S

B. Gu and S. Mukamel, Cooperative Conical Intersec- tion Dynamics of Two Pyrazine Molecules in an Optical Cavity, J. Phys. Chem. Lett. 11, 5555 (2020)

work page 2020

[6] [6]

Gu and S

B. Gu and S. Mukamel, Manipulating nonadiabatic con- ical intersection dynamics by optical cavities, Chem. Sci. 11, 1290 (2020)

work page 2020

[7] [7]

L.-P. Yang, F. Khosravi, and Z. Jacob, Quantum spin operator of the photon, arXiv:2004.03771 [cond- mat, physics:hep-th, physics:physics, physics:quant- ph] (2020), arXiv:2004.03771 [cond-mat, physics:hep-th, physics:physics, physics:quant-ph]

work page Pith review arXiv 2020

[8] [8]

C. Xie, J. Ma, X. Zhu, D. R. Yarkony, D. Xie, and H. Guo, Nonadiabatic Tunneling in Photodissociation of Phenol, J. Am. Chem. Soc. 138, 7828 (2016)

work page 2016