arxiv: 2604.19964 · v1 · submitted 2026-04-21 · ❄️ cond-mat.stat-mech · physics.chem-ph

Recognition: unknown

Nudged Elastic Membranes for Constructing Reduced Two-Dimensional Potential Energy Surfaces

Uday Sankar Manoj , Nicole Drew , Ismaila Dabo , Lukas Muechler

Authors on Pith no claims yet

Pith reviewed 2026-05-10 00:53 UTC · model grok-4.3

classification ❄️ cond-mat.stat-mech physics.chem-ph

keywords nudged elastic membranepotential energy surfacereaction pathssaddle pointsformaldehydemultidimensional PESpath optimizationcomputational chemistry

0 comments

The pith

The nudged elastic membrane constructs two-dimensional reduced potential energy surfaces from energies and forces alone.

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

The paper introduces the nudged elastic membrane method to build reduced two-dimensional potential energy surfaces for chemical systems. It relies solely on energies and forces to nudge an elastic membrane into alignment with relevant PES regions, avoiding Hessian computations. Tests on a model system and triplet formaldehyde show it captures both linear reaction paths and two-dimensional features including a new second-order saddle point. This matters because it provides a direct way to explore the full shape of energy landscapes rather than isolated paths.

Core claim

We introduce the nudged elastic membrane method, a framework for constructing two-dimensional reduced potential-energy surfaces in chemically relevant regions of a PES using only energies and forces without requiring more costly Hessian information. The method is demonstrated for a three-dimensional prototype model and for the triplet formaldehyde molecular system. In both cases, the resulting membrane captures one-dimensional reaction-path features as well as genuinely two-dimensional structures such as a yet unreported second-order saddle point in the PES of triplet formaldehyde. The method further provides direct access to structures that can serve as starting points for subsequent refin

What carries the argument

The nudged elastic membrane: a discretized elastic sheet whose nodes are displaced by a force combining the PES gradient with elastic tension to conform to the surface while preserving two-dimensional connectivity.

If this is right

It offers a practical route to exploring multidimensional PES topography beyond the single-path picture.
The membrane identifies structures that serve as starting points for subsequent refinement of critical points.
It captures both one-dimensional minimum-energy paths and higher-dimensional features such as second-order saddles.
The approach extends to other molecular systems where only energies and forces are available from electronic structure calculations.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

If the membrane reliably covers low-energy regions, it could be chained with existing path optimizers to automatically locate branching reaction channels.
The method's force-only requirement suggests direct integration with on-the-fly ab initio molecular dynamics for exploring surfaces in larger molecules.
It may reduce reliance on hand-selected collective variables when generating 2D cuts through high-dimensional PES.

Load-bearing premise

That nudging an elastic membrane with only energies and forces will consistently converge to the chemically relevant two-dimensional regions of the PES without missing key features or requiring Hessian data or manual adjustments.

What would settle it

Running the nudged elastic membrane on the triplet formaldehyde PES, then computing the full numerical Hessian at the identified critical point and checking whether it has exactly two negative eigenvalues.

Figures

Figures reproduced from arXiv: 2604.19964 by Ismaila Dabo, Lukas Muechler, Nicole Drew, Uday Sankar Manoj.

**Figure 2.** Figure 2: FIG. 2. (a) Optimized [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** Figure 3: FIG. 3. Initial [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

**Figure 4.** Figure 4: FIG. 4. (a) Contour plot of the energy surface obtained from [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗

read the original abstract

Path optimization methods have been widely used and highly successful for the analysis of chemical reactions. Yet, they can fail to capture intrinsically multidimensional features of potential energy surfaces (PES). We introduce the nudged elastic membrane method, a framework for constructing two-dimensional reduced potential-energy surfaces in chemically relevant regions of a PES using only energies and forces without requiring more costly Hessian information. The method is demonstrated for a three-dimensional prototype model and for the triplet formaldehyde molecular system. In both cases, the resulting membrane captures one-dimensional reaction-path features as well as genuinely two-dimensional structures such as a yet unreported reported second-order saddle point in the PES of triplet formaldehyde. The method further provides direct access to structures that can serve as starting points for subsequent refinement. Our results show that the method offers a practical route to exploring multidimensional PES topography beyond the single-path picture.

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

The nudged elastic membrane is a straightforward 2D extension of NEB that avoids Hessians, but the unreported second-order saddle in formaldehyde rests on verification steps that need clearer numbers.

read the letter

The paper's core contribution is a nudged elastic membrane that builds a reduced 2D surface from energies and forces alone. It treats the surface like a flexible sheet pulled by the local gradients, then relaxes it in the chemically interesting region. This is a direct lift from the nudged elastic band idea but spread across two dimensions instead of one path. On the three-dimensional prototype it recovers the expected features without trouble. On triplet formaldehyde it surfaces a stationary point the authors call a second-order saddle that had not been noted before, plus nearby structures that could seed further searches. That is the practical payoff: a way to scan for multidimensional character without paying for second derivatives up front. The implementation looks clean and the method stays parameter-light, which is a plus for reproducibility. The main soft spot is the formaldehyde saddle claim. A true index-2 point requires two negative Hessian eigenvalues at a stationary point. The membrane itself is built only from first derivatives, so confirming the curvature must involve a separate calculation—finite differences or a small Hessian evaluation at the candidate point. The abstract and the stress-test note both leave that step implicit, with no grid density, convergence checks, or error bars shown. If that post-processing is stable and the eigenvalues really are negative, the result stands; if it is marginal or omitted, the “genuinely two-dimensional structure” is only a geometric feature of the membrane rather than a verified saddle of the PES. The prototype demonstration is too low-dimensional to test this issue. The work is aimed at computational chemists who already run force evaluations and want a lightweight way to map branching or flat regions before committing to full Hessian or dynamics runs. It is not a replacement for existing path tools but a useful complement when single paths miss the picture. The idea is coherent and the authors engage the right literature on NEB variants. I would bring it to a reading group to see the actual implementation and the numerical checks on the saddle. It is worth sending to peer review so referees can ask for the missing verification data and test the method on a few more systems.

Referee Report

2 major / 2 minor

Summary. The manuscript introduces the nudged elastic membrane (NEM) method for constructing reduced two-dimensional potential energy surfaces (PES) in chemically relevant regions using only energies and forces, without Hessian information. It demonstrates the approach on a three-dimensional prototype model and the triplet formaldehyde system, claiming that the resulting membrane reproduces one-dimensional reaction-path features while also capturing genuinely two-dimensional structures, including an unreported second-order saddle point in the formaldehyde PES. The method is further positioned as supplying starting structures for subsequent refinement.

Significance. If the claims are substantiated, the NEM framework would provide a computationally attractive route to mapping multidimensional PES topography in molecular systems where traditional path-optimization methods are insufficient. A notable strength is the explicit avoidance of Hessian evaluations while still targeting higher-order stationary points, which could enable broader exploration of reaction mechanisms. The dual demonstration on a controlled prototype and a realistic chemical system, together with the provision of refinement-ready structures, adds practical value if quantitative validation is supplied.

major comments (2)

[Abstract and triplet formaldehyde demonstration] Abstract and triplet formaldehyde results: The headline claim that the membrane locates a 'yet unreported second-order saddle point' in the triplet formaldehyde PES is load-bearing for the assertion of 'genuinely two-dimensional structures.' Because the NEM is constructed exclusively from energies and forces, the manuscript must specify the post-processing procedure (finite-difference Hessian, grid parameters, convergence criteria, and numerical stability checks) used to confirm that the identified point has exactly two negative eigenvalues on the underlying PES rather than merely appearing as a geometric extremum of the membrane.
[Prototype model section] Prototype model demonstration: The abstract states that the membrane 'captures' one- and two-dimensional features of the three-dimensional prototype, yet no quantitative error metrics, comparison to analytic reference surfaces, or reproduction accuracy for known stationary points are reported. Without such benchmarks, it is unclear whether the membrane faithfully represents the PES topography or merely provides a plausible interpolation.

minor comments (2)

[Abstract] The abstract contains a duplicated word: 'yet unreported reported second-order saddle point' should read 'yet unreported second-order saddle point.'
[Method description] Notation for the membrane tension, spring constants, and the mapping from the two-dimensional parameter space to Cartesian coordinates should be introduced with an explicit equation or schematic to improve reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful and constructive review of our manuscript. We address each major comment below and will revise the manuscript accordingly to strengthen the presentation and validation of the nudged elastic membrane method.

read point-by-point responses

Referee: [Abstract and triplet formaldehyde demonstration] Abstract and triplet formaldehyde results: The headline claim that the membrane locates a 'yet unreported second-order saddle point' in the triplet formaldehyde PES is load-bearing for the assertion of 'genuinely two-dimensional structures.' Because the NEM is constructed exclusively from energies and forces, the manuscript must specify the post-processing procedure (finite-difference Hessian, grid parameters, convergence criteria, and numerical stability checks) used to confirm that the identified point has exactly two negative eigenvalues on the underlying PES rather than merely appearing as a geometric extremum of the membrane.

Authors: We agree that the verification procedure must be documented explicitly to substantiate the identification of a second-order saddle point. In the revised manuscript we will add a new subsection detailing the post-processing: the Hessian is obtained via central finite differences of the forces with a step size of 0.01 bohr; the eigenvalues are computed on a 3x3x3 grid centered at the candidate point with convergence checked to 10^{-6} hartree/bohr^2; numerical stability is confirmed by repeating the calculation at step sizes of 0.005 and 0.02 bohr and verifying that exactly two negative eigenvalues persist. The two negative eigenvalues and their eigenvectors will be reported, confirming that the point is a genuine second-order saddle on the underlying PES. revision: yes
Referee: [Prototype model section] Prototype model demonstration: The abstract states that the membrane 'captures' one- and two-dimensional features of the three-dimensional prototype, yet no quantitative error metrics, comparison to analytic reference surfaces, or reproduction accuracy for known stationary points are reported. Without such benchmarks, it is unclear whether the membrane faithfully represents the PES topography or merely provides a plausible interpolation.

Authors: The referee is correct that quantitative benchmarks are required. In the revised manuscript we will insert a new paragraph and accompanying table in the prototype-model section that reports (i) the root-mean-square deviation between membrane-interpolated energies and the analytic PES evaluated on an independent 20x20x20 test grid, (ii) the Euclidean distance and energy difference between each stationary point located on the membrane and the corresponding analytic location, and (iii) the maximum absolute error along the minimum-energy path. These metrics will demonstrate that the membrane reproduces the known topography to within chemical accuracy. revision: yes

Circularity Check

0 steps flagged

No circularity: method derives from energies/forces without self-referential fitting or definition

full rationale

The paper defines the nudged elastic membrane construction explicitly from input energies and forces on a grid or along paths, then uses the resulting surface to locate candidate structures. No equation or step equates an output feature (such as the reported index-2 point) back to the membrane parameters by construction; the second-order saddle claim is presented as a post-construction discovery on the original PES, not as a tautological consequence of the membrane definition itself. The derivation chain therefore remains independent of the specific results it produces.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The central claim rests on the domain assumption that elastic membrane dynamics driven by energies and forces suffice to capture multidimensional PES features without second-derivative information.

axioms (1)

domain assumption Potential energy surfaces can be explored in chemically relevant regions using only energies and first derivatives via elastic membrane dynamics.
Invoked to justify constructing the membrane without Hessian calculations.

invented entities (1)

Nudged elastic membrane no independent evidence
purpose: To construct reduced two-dimensional potential energy surfaces
New construct introduced to map multidimensional features from energies and forces.

pith-pipeline@v0.9.0 · 5453 in / 1205 out tokens · 40714 ms · 2026-05-10T00:53:43.525990+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

43 extracted references

[1]

Grid and optimization The NEM grid is constructed in cartesian coordinates, with distances in angstroms (Å) and all forces are eval- uated numerically in eV/Å. The anchor points for the initial grid are chosen such that it encompasses three known local minima that are important for the formalde- hydedecompositionreactiononthelowesttripletsurface. The firs...
[2]

C. J. Cerjan and W. H. Miller, J. Chem. Phys75, 2800 (1981)

1981
[3]

Quapp and D

W. Quapp and D. Heidrich, Theoret. Chim. Acta66, 245 7 TABLE II. Geometry and energy comparisons of the predicted TSs, obtained from the optimized17×17grid, and the actual TSs from literature. TS1 geometry was reconstructed from the work by Yamaguchiet al.[37]. TS2 and TS3 geometries were obtained from the work by Hicksonet al.[36] Geometry Predicted Geom...

1984
[4]

Fukui, The Journal of Physical Chemistry74, 4161 (1970)

K. Fukui, The Journal of Physical Chemistry74, 4161 (1970)

1970
[5]

Jónsson, G

H. Jónsson, G. Mills, and K. W. Jacobsen, inClassical and quantum dynamics in condensed phase simulations (World Scientific, 1998) pp. 385–404

1998
[6]

Henkelman and H

G. Henkelman and H. Jónsson, J. Chem. Phys111, 7010 (1999)

1999
[7]

Peters, A

B. Peters, A. Heyden, A. T. Bell, and A. Chakraborty, J. Chem. Phys120, 7877 (2004)

2004
[8]

H. B. Schlegel, J. Comput. Chem.24, 1514 (2003)

2003
[9]

A. L. Dewyer, A. J. Argüelles, and P. M. Zimmerman, WIREs Comput. Mol. Sci.8, e1354 (2018)

2018
[10]

A. G. Suits, Annu. Rev. Phys. Chem.71, 77 (2020)

2020
[11]

S. R. Hare and D. J. Tantillo, Pure Appl. Chem.89, 679 (2017)

2017
[12]

Vidossich and M

P. Vidossich and M. De Vivo, Chem Catalysis1, 69 (2021)

2021
[13]

G. S. Ezra and S. Wiggins, J. Phys. A: Math. Theor.42, 205101 (2009)

2009
[14]

Rehbein and B

J. Rehbein and B. K. Carpenter, Phys. Chem. Chem. Phys.13, 20906 (2011)

2011
[15]

Y. J. Hong and D. J. Tantillo, Nat. Chem.6, 104 (2014)

2014
[16]

Pradhan and U

R. Pradhan and U. Lourderaj, Phys. Chem. Chem. Phys. 21, 12837 (2019)

2019
[17]

D. H. Ess, S. E. Wheeler, R. G. Iafe, L. Xu, N. Celebi- Olçüm, and K. N. Houk, Angew. Chem. Int. Ed Engl. 47, 7592 (2008)

2008
[18]

Heidrich and W

D. Heidrich and W. Quapp, Theor. Chim. Acta70, 89 (1986)

1986
[19]

Yadav, R

K. Yadav, R. Pradhan, and U. Lourderaj, Faraday Dis- cuss.238, 183 (2022)

2022
[20]

Y. Yang, Y. Gao, J. Eaton, Y. Meng, H. Zhu, and H.-X. Zhou, J. Chem. Phys.151, 070902 (2019)

2019
[21]

Mehdi, Z

S. Mehdi, Z. Smith, L. Herron, Z. Zou, and P. Tiwary, Annu. Rev. Phys. Chem.75, 347 (2024)

2024
[22]

Awasthi and N

S. Awasthi and N. N. Nair, WIREs Comput. Mol. Sci.9, e1398 (2019)

2019
[23]

Piccini, M.-S

G. Piccini, M.-S. Lee, S. F. Yuk, D. Zhang, G. Collinge, L. Kollias, M.-T. Nguyen, V.-A. Glezakou, and 1 R. Rousseau, Catal. Sci. Technol.12, 12 (2022)

2022
[24]

Chuang, D

H.-H. Chuang, D. J. Tantillo, and C.-P. Hsu, J. Chem. Theory Comput.16, 4050 (2020)

2020
[25]

Henkelman and H

G. Henkelman and H. Jónsson, J. Chem. Phys.113, 9978 (2000)

2000
[26]

S. J. Leon, Å. Björck, and W. Gander, Numerical Linear Algebra with Applications20, 492 (2013)

2013
[27]

Bitzek, P

E. Bitzek, P. Koskinen, F. Gähler, M. Moseler, and P. Gumbsch, Phys. Rev. Lett.97, 170201 (2006)

2006
[28]

Guénolé, W

J. Guénolé, W. G. Nöhring, A. Vaid, F. Houllé, Z. Xie, A. Prakash, and E. Bitzek, Comput. Mater. Sci.175, 109584 (2020)

2020
[29]

Ribaldone and S

C. Ribaldone and S. Casassa, AIP Adv.12(2022)

2022
[30]

H. C. Herbol, J. Stevenson, and P. Clancy, J. Chem. The- ory Comput.13, 3250 (2017)

2017
[31]

Sheppard, R

D. Sheppard, R. Terrell, and G. Henkelman, J. Chem. Phys128(2008)

2008
[32]

Shuang, P

F. Shuang, P. Xiao, R. Shi, F. Ke, and Y. Bai, Comput. Mater. Sci.156, 135 (2019)

2019
[33]

E. L. Kolsbjerg, M. N. Groves, and B. Hammer, The Journal of chemical physics145(2016)

2016
[34]

C. Liu, C. T. Kelley, and E. Jakubikova, J. Phys. Chem. A.123, 4543 (2019)

2019
[35]

Y. Shu, Z. Varga, D. Zhang, and D. G. Truhlar, J. Phys. Chem. A.127, 9635 (2023)

2023
[36]

J. Li, C. Xie, and H. Guo, Phys. Chem. Chem. Phys.19, 23280 (2017)

2017
[37]

K. M. Hickson, J.-C. Loison, D. Nuñez-Reyes, and R. Méreau, J. Phys. Chem. Lett.7, 3641 (2016)

2016
[38]

Yamaguchi, S

Y. Yamaguchi, S. S. Wesolowski, T. J. Van Huis, and H.F.Schaefer,TheJournalofchemicalphysics108,5281 (1998)

1998
[39]

Makri, C

S. Makri, C. Ortner, and J. R. Kermode, The Journal of Chemical Physics150(2019)

2019
[40]

S. A. Trygubenko and D. J. Wales, The Journal of chem- ical physics120, 2082 (2004)

2082
[41]

P. M. Zimmerman, J. Chem. Phys138(2013)

2013
[42]

H. B. Schlegel, Wiley Interdisciplinary Reviews: Compu- tational Molecular Science1, 790 (2011)

2011
[43]

Henkelman, B

G. Henkelman, B. P. Uberuaga, and H. Jónsson, The Journal of chemical physics113, 9901 (2000). 2 I. SUPPLEMENT AR Y INFORMA TION In this SI, we provide additional information about the two models studied. A. 3D Model Fig. S1 displays the chosen initial grid and energy surface of the 3D model discussed in the main text, and displays the convergence of the ...

2000