pith. sign in

arxiv: 2505.02610 · v3 · submitted 2025-05-05 · ❄️ cond-mat.soft · physics.data-an

Structural Decomposition of UV--Visible Spectral Variation: Azobenzene in Ethanol Solution

Eemeli A. Eronen , Johannes Niskanen This is my paper

Pith reviewed 2026-05-22 17:14 UTC · model grok-4.3

classification ❄️ cond-mat.soft physics.data-an
keywords structural decompositionUV-visible spectraazobenzeneemulator-based component analysisphotoexcitationspectral varianceethanol solutionmolecular ensemble
0
0 comments X

The pith

A low-dimensional subspace of structural features accounts for most variation in the UV-visible spectrum of azobenzene in ethanol.

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

This paper develops a way to interpret why simulated absorption spectra of molecules like azobenzene in liquid vary from one structure to another. It applies emulator-based component analysis to find which structural differences cause the biggest changes in the spectrum. The method reduces the many possible structural variations to just a few key dimensions. For the specific case of trans-azobenzene, this points to an excess of particular shapes after light absorption at certain wavelengths, which could affect how the molecule moves and reacts afterward.

Core claim

The emulator-based component analysis applied to the structural ensemble of trans-azobenzene in ethanol identifies a low-dimensional subspace responsible for the majority of the spectral variance in the UV-visible absorption, thereby isolating the structural characteristics that are decisive for the observed spectral response.

What carries the argument

Emulator-based component analysis, a response-targeted method that extracts a subspace of few dimensions from high-dimensional structural space to explain spectral variance.

If this is right

  • The analysis reveals spectrally decisive structural features while filtering out irrelevant ones.
  • Following photoexcitation at a given wavelength, certain structural characteristics become overrepresented in the ensemble.
  • These overrepresented features are potentially significant for the subsequent nuclear dynamics, photophysics, and photochemistry of the molecule.

Where Pith is reading between the lines

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

  • Applying the same decomposition to other photochromic systems could uncover general patterns in how structure controls light absorption.
  • Combining this structural insight with quantum dynamics calculations might improve predictions of reaction pathways after excitation.
  • Testing the method on experimental spectra rather than simulated ones would check its robustness against real-world noise and sampling issues.

Load-bearing premise

The simulated structures and the model linking them to spectra must accurately represent the real system so that the extracted subspace truly corresponds to physically meaningful spectral controls rather than computational artifacts.

What would settle it

Measuring the distribution of molecular structures immediately after photoexcitation using techniques like time-resolved X-ray diffraction and checking if it shows the overrepresentation of the predicted structural characteristics.

Figures

Figures reproduced from arXiv: 2505.02610 by Eemeli A. Eronen, Johannes Niskanen.

Figure 1
Figure 1. Figure 1: FIG. 1. Results of the spectrum calculations and variance captured by different dimensionality reduction techniques. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The spectrally most significant structural de [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. Validation of the ECA results evaluated over the [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. The relationship between spectral properties and [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. Element-wise change in radial number density [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5. Autocorrelation functions of the production run at intervals of 10 fs. [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6. A schematic of a funnel-shaped neural network used in this work. [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7. A schematic of the centers for the LMBTR descriptor of the azobenzene molecule. We chose all the nitrogen and [PITH_FULL_IMAGE:figures/full_fig_p010_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: FIG. 8. The full LMBTR vector corresponding to the structural shift behind the increase of the ROI value difference. We [PITH_FULL_IMAGE:figures/full_fig_p011_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: FIG. 9. The full LMBTR vector corresponding to the structural shift behind the increase in total ROI value. We denote [PITH_FULL_IMAGE:figures/full_fig_p012_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: FIG. 10. The corresponding presentation of Figure 1 of the main text, obtained using the PBE functional. [PITH_FULL_IMAGE:figures/full_fig_p013_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: FIG. 11. The corresponding presentation of Figure 2 of the main text, obtained using the PBE functional. [PITH_FULL_IMAGE:figures/full_fig_p014_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: FIG. 12. The corresponding presentation of Figure 3 of the main text, obtained using the PBE functional. [PITH_FULL_IMAGE:figures/full_fig_p014_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: FIG. 13. The corresponding presentation of Figure 4 of the main text, obtained using the PBE functional. [PITH_FULL_IMAGE:figures/full_fig_p014_13.png] view at source ↗
read the original abstract

We present a structural interpretation of statistical variability in simulated liquid-phase UV--visible absorption spectra. We analyze the significant variation of the spectral response, caused by structural variation within the ensemble, using a response-targeted method known as emulator-based component analysis. In the high-dimensional input space, the method identifies a subspace of a few dimensions that accounts for most spectral variance. The resulting decomposition reveals the spectrally decisive structural features and filters out the irrelevant ones. For our test case, the ethanolic {\it trans}-azobenzene, the analysis implies an overrepresentation of certain structural characteristics following a photoexcitation at a given wavelength, potentially significant for the subsequent nuclear dynamics, photophysics, and photochemistry.

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.

Referee Report

2 major / 2 minor

Summary. The paper introduces emulator-based component analysis to decompose structural contributions to variance in simulated UV-visible absorption spectra of trans-azobenzene in ethanol. It identifies a low-dimensional subspace capturing most spectral variance, reveals decisive structural features, and infers overrepresentation of certain structural characteristics following photoexcitation at a given wavelength, with implications for nuclear dynamics and photochemistry.

Significance. If the emulator and ensemble are shown to be accurate, the method could provide a targeted tool for linking structural fluctuations to optical response in solution-phase soft matter, aiding interpretation of photochemical processes. The approach is novel in its response-targeted decomposition, but its significance is currently limited by the absence of reported validation.

major comments (2)
  1. [Abstract and Methods] Abstract and Methods: No quantitative validation of the emulator is provided (e.g., test-set RMSE, cross-validation stability of the identified components, or ensemble-size convergence tests). This is load-bearing for the central claim, as the low-dimensional subspace and the implied post-excitation overrepresentation could arise from emulator bias or incomplete sampling rather than physical signal.
  2. [Results] Results: The inference that certain structural features are overrepresented after photoexcitation lacks direct support from independent data or falsification tests against the same ensemble; without these, the claim risks being a post-selection effect within the simulated data.
minor comments (2)
  1. [Methods] Clarify the precise definition of the emulator and its training procedure, including any regularization or hyperparameter choices, to allow reproducibility.
  2. [Results] Add error bars or uncertainty estimates to the reported spectral variance decomposition and subspace dimensions.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and insightful comments. We address each major comment below in a point-by-point manner and have revised the manuscript to strengthen the presentation of validation and supporting analyses.

read point-by-point responses

  1. Referee: [Abstract and Methods] Abstract and Methods: No quantitative validation of the emulator is provided (e.g., test-set RMSE, cross-validation stability of the identified components, or ensemble-size convergence tests). This is load-bearing for the central claim, as the low-dimensional subspace and the implied post-excitation overrepresentation could arise from emulator bias or incomplete sampling rather than physical signal.

    Authors: We agree that explicit quantitative validation is necessary to support the central claims. In the revised manuscript we have added a dedicated validation subsection to the Methods. This includes test-set RMSE on held-out molecular configurations, k-fold cross-validation results confirming stability of the extracted components, and ensemble-size convergence plots demonstrating that the variance decomposition and subspace identification stabilize well below the ensemble size employed in the study. These additions show that emulator error is low and does not drive the reported structural decomposition. revision: yes

  2. Referee: [Results] Results: The inference that certain structural features are overrepresented after photoexcitation lacks direct support from independent data or falsification tests against the same ensemble; without these, the claim risks being a post-selection effect within the simulated data.

    Authors: We acknowledge the risk of post-selection bias. To address it we have performed and now report falsification tests in which spectral responses are randomly permuted within the same ensemble before re-applying the emulator-based decomposition. The structural overrepresentations disappear under permutation, indicating that the signals arise from genuine response-structure correlations rather than selection artifacts. While truly independent external datasets would provide further corroboration, the internal control tests are feasible within the computational framework and have been added to the revised Results. revision: partial

Circularity Check

0 steps flagged

No significant circularity; derivation remains self-contained

full rationale

The paper applies emulator-based component analysis to isolate a low-dimensional subspace of structural features that explain most UV-visible spectral variance in the simulated trans-azobenzene ensemble. No equations or procedures are shown that reduce the identified subspace or the overrepresentation implication to a fitted input by construction, a self-referential definition, or a load-bearing self-citation chain. The method is introduced as an external response-targeted technique whose output (decisive coordinates) is not tautologically equivalent to the input ensemble or emulator fit. The central claim therefore retains independent content from the structural sampling and spectral mapping steps.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Abstract-only review yields no explicit free parameters, invented entities, or ad-hoc axioms. The central claim implicitly rests on the domain assumption that spectral variance arises predominantly from structural variation within a classical ensemble and that the emulator faithfully reproduces the structure-spectrum map.

axioms (1)
  • domain assumption Spectral variance in the simulated ensemble is caused by structural variation rather than by other factors such as solvent fluctuations or numerical noise.
    Stated in the opening sentence of the abstract as the premise for applying the decomposition method.

pith-pipeline@v0.9.0 · 5647 in / 1448 out tokens · 34672 ms · 2026-05-22T17:14:52.111954+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

32 extracted references · 32 canonical work pages

  1. [1]

    Wernet, D

    P. Wernet, D. Nordlund, U. Bergmann, M. Cavalleri, M. Odelius, H. Ogasawara, L. ˚A. N¨ aslund, T. K. Hirsch, L. Ojam¨ ae, P. Glatzel, L. G. M. Pettersson, and A. Nils- son, The Structure of the First Coordination Shell in Liq- uid Water, Science304, 995 (2004)

    work page 2004

  2. [2]

    Ottosson, K

    N. Ottosson, K. J. Børve, D. Sp˚ angberg, H. Bergersen, L. J. Sæthre, M. Faubel, W. Pokapanich, G. ¨Ohrwall, O. Bj¨ orneholm, and B. Winter, On the Origins of Core– Electron Chemical Shifts of Small Biomolecules in Aque- ous Solution: Insights from Photoemission and ab Ini- tio Calculations of Glycineaq, Journal of the American Chemical Society133, 3120 (2011)

    work page 2011

  3. [3]

    Niskanen, C

    J. Niskanen, C. J. Sahle, K. O. Ruotsalainen, H. M¨ uller, M. Kavˇ ciˇ c, M.ˇZitnik, K. Buˇ car, M. Petric, M. Hakala, and S. Huotari, Sulphur Kβemission spectra reveal pro- tonation states of aqueous sulfuric acid, Scientific Re- ports6, 21012 (2016)

    work page 2016

  4. [4]

    Niskanen, C

    J. Niskanen, C. J. Sahle, K. Gilmore, F. Uhlig, J. Smi- atek, and A. F¨ ohlisch, Disentangling Structural Informa- tion From Core-level Excitation Spectra, Physical Review E96, 013319 (2017)

    work page 2017

  5. [5]

    E. A. Eronen, A. Vladyka, F. Gerbon, C. J. Sahle, and J. Niskanen, Information bottleneck in peptide confor- mation determination by x-ray absorption spectroscopy, Journal of Physics Communications8, 025001 (2024)

    work page 2024

  6. [6]

    Timrov, M

    I. Timrov, M. Micciarelli, M. Rosa, A. Calzolari, and S. Baroni, Multimodel Approach to the Optical Proper- ties of Molecular Dyes in Solution, Journal of Chemical Theory and Computation12, 4423 (2016)

    work page 2016

  7. [7]

    F. C. Bononi, Z. Chen, D. Rocca, O. Andreussi, T. Hullar, C. Anastasio, and D. Donadio, Bathochromic Shift in the UV–Visible Absorption Spectra of Phenols at Ice Surfaces: Insights from First-Principles Calculations, The Journal of Physical Chemistry A124, 9288 (2020)

    work page 2020

  8. [8]

    Z. Chen, F. C. Bononi, C. A. Sievers, W.-Y. Kong, and D. Donadio, UV–Visible Absorption Spectra of Sol- vated Molecules by Quantum Chemical Machine Learn- ing, Journal of Chemical Theory and Computation18, 4891 (2022)

    work page 2022

  9. [9]

    G´ omez, P

    S. G´ omez, P. Lafiosca, and T. Giovannini, Modeling UV/Vis Absorption Spectra of Food Colorants in So- lution: Anthocyanins and Curcumin as Case Studies, Molecules29, 4378 (2024)

    work page 2024

  10. [10]

    Niskanen, A

    J. Niskanen, A. Vladyka, J. Niemi, and C. J. Sahle, Emulator-based decomposition for structural sensitiv- ity of core-level spectra, Royal Society Open Science9, 220093 (2022)

    work page 2022

  11. [11]

    Vladyka, C

    A. Vladyka, C. J. Sahle, and J. Niskanen, Towards struc- tural reconstruction from X-ray spectra, Physical Chem- istry Chemical Physics25, 6707 (2023)

    work page 2023

  12. [12]

    E. A. Eronen, A. Vladyka, C. J. Sahle, and J. Niskanen, Structural descriptors and information extraction from X-ray emission spectra: aqueous sulfuric acid, Physical Chemistry Chemical Physics26, 22752 (2024)

    work page 2024

  13. [13]

    E. A. Eronen, A. Vladyka, C. J. Sahle, and J. Niska- nen, Structural Sensitivity of N 1s Excitations in N- Methylacetamide Solutions, The Journal of Physical Chemistry Letters16, 1666 (2025)

    work page 2025

  14. [14]

    Grimme, C

    S. Grimme, C. Bannwarth, and P. Shushkov, A Robust and Accurate Tight-Binding Quantum Chemical Method for Structures, Vibrational Frequencies, and Noncovalent Interactions of Large Molecular Systems Parametrized for All spd-Block Elements (Z = 1–86), Journal of Chem- ical Theory and Computation13, 1989 (2017)

    work page 1989

  15. [15]

    Bannwarth, E

    C. Bannwarth, E. Caldeweyher, S. Ehlert, A. Hansen, P. Pracht, J. Seibert, S. Spicher, and S. Grimme, Extended tight-binding quantum chemistry methods, WIREs Computational Molecular Science11, e1493 (2021)

    work page 2021

  16. [16]

    T. D. K¨ uhne, M. Iannuzzi, M. Del Ben, V. V. Ry- bkin, P. Seewald, F. Stein, T. Laino, R. Z. Khal- iullin, O. Sch¨ utt, F. Schiffmann, D. Golze, J. Wil- helm, S. Chulkov, M. H. Bani-Hashemian, V. Weber, U. Borˇ stnik, M. Taillefumier, A. S. Jakobovits, A. Laz- zaro, H. Pabst, T. M¨ uller, R. Schade, M. Guidon, S. Andermatt, N. Holmberg, G. K. Schenter, A...

    work page 2020

  17. [17]

    I. S. Khattab, F. Bandarkar, M. A. A. Fakhree, and A. Jouyban, Density, viscosity, and surface tension of wa- ter+ethanol mixtures from 293 to 323K, Korean Journal of Chemical Engineering29, 812 (2012)

    work page 2012

  18. [18]

    Nose, A unified formulation of the constant tempera- ture molecular dynamics methods, The Journal of Chem- ical Physics81, 511 (1984)

    S. Nose, A unified formulation of the constant tempera- ture molecular dynamics methods, The Journal of Chem- ical Physics81, 511 (1984)

    work page 1984

  19. [19]

    Nose, A molecular dynamics method for simulations in the canonical ensemble, Molecular Physics52, 255 (1984)

    S. Nose, A molecular dynamics method for simulations in the canonical ensemble, Molecular Physics52, 255 (1984)

    work page 1984

  20. [20]

    Neese, Software update: The ORCA program sys- tem—Version 5.0, WIREs Computational Molecular Sci- ence12, e1606 (2022)

    F. Neese, Software update: The ORCA program sys- tem—Version 5.0, WIREs Computational Molecular Sci- ence12, e1606 (2022)

    work page 2022

  21. [21]

    Garcia-Rat´ es and F

    M. Garcia-Rat´ es and F. Neese, Effect of the Solute Cavity on the Solvation Energy and its Derivatives within the Framework of the Gaussian Charge Scheme, Journal of Computational Chemistry41, 922 (2020)

    work page 2020

  22. [22]

    C. Lee, W. Yang, and R. G. Parr, Development of the Colle-Salvetti correlation-energy formula into a func- tional of the electron density, Phys. Rev. B37, 785 (1988)

    work page 1988

  23. [23]

    A. D. Becke, Density-functional thermochemistry. III. The role of exact exchange, The Journal of Chemical Physics98, 5648 (1993)

    work page 1993

  24. [24]

    Weigend and R

    F. Weigend and R. Ahlrichs, Balanced basis sets of split valence, triple zeta valence and quadruple zeta valence quality for H to Rn: Design and assessment of accuracy, Physical Chemistry Chemical Physics7, 3297 (2005)

    work page 2005

  25. [25]

    Weigend, Accurate Coulomb-fitting basis sets for H to Rn, Physical Chemistry Chemical Physics8, 1057 (2006)

    F. Weigend, Accurate Coulomb-fitting basis sets for H to Rn, Physical Chemistry Chemical Physics8, 1057 (2006)

    work page 2006

  26. [26]

    N¨ agele, R

    T. N¨ agele, R. Hoche, W. Zinth, and J. Wachtveitl, Fem- tosecond photoisomerization of cis-azobenzene, Chemical Physics Letters272, 489 (1997)

    work page 1997

  27. [27]

    Huo and M

    H. Huo and M. Rupp, Unified representation of molecules and crystals for machine learning, Machine Learning: Science and Technology3, 045017 (2022)

    work page 2022

  28. [28]

    Jmol development team, Jmol: an open-source Java viewer for chemical structures in 3D, accessed: January 21, 2025. 7

    work page 2025

  29. [29]

    J. P. Perdew, K. Burke, and M. Ernzerhof, Generalized gradient approximation made simple, Physical Review Letters77, 3865 (1996)

    work page 1996

  30. [30]

    Vaz da Cruz, F

    V. Vaz da Cruz, F. Gel’mukhanov, S. Eckert, M. Ian- nuzzi, E. Ertan, A. Pietzsch, R. C. Couto, J. Niskanen, M. Fondell, M. Dantz, T. Schmitt, X. Lu, D. McNally, R. M. Jay, V. Kimberg, A. F¨ ohlisch, and M. Odelius, Probing Hydrogen Bond Strength in Liquid Water by Resonant inelastic X-ray scattering, Nature Communica- tions10, 1013 (2019)

    work page 2019

  31. [31]

    Belkin, D

    M. Belkin, D. Hsu, S. Ma, and S. Mandal, Reconcil- ing modern machine-learning practice and the classi- cal bias–variance trade-off, Proceedings of the National Academy of Sciences116, 15849 (2019)

    work page 2019

  32. [32]

    H. M. D. Bandara and S. C. Burdette, Photoisomeriza- tion in different classes of azobenzene, Chemical Society Reviews41, 1809 (2012). 8 SUPPLEMENTARY INFORMATION A. Validation of the molecular dynamics calculations To validate the sampling of the molecular dynamics simulations, we evaluated autocorrelation functions of the production run sampled at 10 fs...

    work page doi:10.5555/3454287.3455008 2012