The DOI matches its Crossref record; 56 of the 56 references it cites were found in a registry or an index; 3 of the 65 sources its claims cite were checked for support.
Efficient neural network models of chemical kinetics using a latent asinh rate transformation
Open accessPublished version of a preprint
Abstract
The proposed latent transformation approach allows building lightweight neural networks that accelerate reactor simulations significantly.
Overview
CoChat Summary
TL;DR
The paper proposes a latent inverse-hyperbolic-sine (asinh) transformation inside neural networks to model steady-state chemical kinetics when rates or source terms change sign. Across two detailed surface mechanisms, the approach improves accuracy and yields reactor-simulation speed-ups of up to 100,000×.
Key findings
- For preferential oxidation of CO, direct neural-network modeling of sign-changing CO source terms gave relative prediction errors around 100% or higher, conventional asinh preprocessing reduced errors, and the latent asinh version produced application-ready predictions with relative errors below 1%.
- In a plug-flow reactor simulation for CO preferential oxidation, the neural-network surrogate was reported to be about 50,000× faster than exact steady-state kinetics on the tested hardware, and up to 100,000× faster using a consumer-grade GPU.
- For ammonia oxidation in the Ostwald process, direct modeling again gave errors near 100%; conventional asinh preprocessing reduced errors to about 15%, while the latent asinh model achieved errors near 1%.
- The paper reports small models: e.g. CO source terms were modeled with either 34 nodes in 5 hidden layers for CO and 94 nodes in 1 hidden layer for O₂, and the ammonia-oxidation models used 63 nodes in a single hidden layer (≈500 parameters) for each key species.
Methods
- Two detailed surface-kinetics test cases were used: preferential oxidation of CO on Pt and ammonia oxidation in the Ostwald process; both were simulated under reactor-relevant operating ranges defined in Tables 1 and 2.
- The dataset comprised 35,000 input-output pairs for each test case, split into 25,000 training, 5,000 validation, and 5,000 test points; neural networks were fully connected, tanh-activated, and trained in PyTorch with full-batch LBFGS.
- The main comparison was between three strategies: direct modeling, conventional preprocessing with asinh transformation, and the proposed latent asinh transformation, evaluated with mean absolute relative error (MARE) and validated in plug-flow reactor simulations and selectivity calculations.
Limitations
- The paper states that its previous, reaction-path-analysis-based method was more accurate, but it requires mechanism-specific insight; the new latent asinh method is less accurate than that prior method when such insight is available.
- For the ammonia-oxidation mechanism, mass transfer was not considered in the validation at Ostwald-process conditions; this is stated in the text.
- Evident from the text: the study is limited to two detailed mechanistic case studies and steady-state source-term prediction, so broader claims about other mechanisms or transient kinetics are not directly tested here.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures













Verification
Claims the paper makes
3 Valid · 1 Not valid · 28 Not checked · 80 Own results and interpretations
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“The computationally most demanding part of those simulations is the solution of the chemical kinetics that often takes 70 % to 90 % of the computation time for both gas-phase 5,6 and surface-reactive 1,7 systems.”
§ Efficient Neural Network Models of Chemical Kinetics Using a Latent asinh Rate Transformation † > 1 Introduction
What the sources say
- Partly supportedcites [5] 5. T. S. Brown, H. Antil, R. Löhner, F. Togashi and D. Verma, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinfor- matics), Springer International Publishing, 2021, vol. 12761 LNCS, pp. 23-39. sourceExperimental — accuracy not yet measured
The paper's introduction (p. 1) says chemical reactions substantially increase the computing requirements of reacting-flow simulations, supporting the claim that kinetics is computationally demanding. But its opening summary (p. 1) gives a figure of over 99% for combustion, not 70%–90%, and the passages found do not address surface-reactive systems. (LLM, conf=0.86)
- Not matchedcites [6] 6. V. F. Nikitin, I. M. Karandashev, M. Y. Malsagov and E. V. Mikhalchenko, Acta Astronautica, 2022, 194, 376-382.
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Not matchedcites [1] 1. D. Micale, C. Ferroni, R. Uglietti, M. Bracconi and M. Maestri, Chemie Ingenieur Technik, 2022, 94, 634-651.
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Not checked — no full text of the sourcecites [7] 7. M. Bracconi, M. Maestri and A. Cuoci, AIChE Journal, 2017, 63, 95-104. source
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“This can be accounted to the fact that it makes the target function more linear over a wide range of reaction conditions 17,29 as demonstrated in equations S1 -S3 in the ESI † .”
§ Efficient Neural Network Models of Chemical Kinetics Using a Latent asinh Rate Transformation † > 1.2 Data Transformation
What the sources say
- Supportedcites [17] 17. S. Matera, M. Maestri, A. Cuoci and K. Reuter, ACS Catalysis, 2014, 4, 4081-4092. source
Section 2.4 (First-Principles Kinetic Monte Carlo Simulations) states that it is 'numerically convenient to rather interpolate the logarithm of the reaction rates as a function of the logarithms of the partial pressures and the inverse temperature', and that 'this particular choice of functional dependence leads to an almost linear behavior in the regions of (T,{p k})-space corresponding to "stable" phases', which the same section identifies as 'large portions of (T, {p k})-space' where the raw TOF-map is otherwise largely exponential. This matches the claim that logarithmic transformation linearizes the target function over a wide range of reaction conditions; the only nuance is that the source confines the near-linearity to stable-phase regions and notes strongly nonlinear behaviour at narrow phase-transition boundaries. (LLM, conf=0.82)
- Not matchedcites [29] 29. F. A. Döppel and M. Votsmeier, Chemical Engineering Science, 2022, 262, 117964.
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
boostedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“In accordance with our findings, it is reported that the parameter z of the inverse hyperbolic sine transformation (equation 20) can substantially affect regression results 39 .”
§ Efficient Neural Network Models of Chemical Kinetics Using a Latent asinh Rate Transformation † > 4 Conclusions and Future Work
What the sources say
- Supportedcites [39] 39. G. B. D. Aihounton and A. Henningsen, The Econometrics Jour- nal, 2021, 24, 334-351. source
The paper states exactly this: Section 2 (IHS transformation) says a regression with IHS-transformed variables "is not invariant to the units of measurement of these variables so that the arbitrary choice of the units of measurement affects the regression results," and the Introduction notes that "the results of the empirical analysis of Pence (2006) are substantially affected by the scale factor." The Conclusion (Section 5) reiterates that "the unit of measurement of IHS-transformed variables can substantially affect the regression results," which the paper confirms with its own Monte Carlo and empirical application, so the point is this paper's own finding and not merely credited to another work. The only wording slack is that our copy denotes this parameter as the scale factor θ (eq. 6) rather than z, but the attributed content — that this parameter of the IHS transformation can substantially affect regression results — is stated. (LLM, conf=0.88)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“In consequence, all species except NH 3 , H 2 O and N 2 show source terms changing sign in the range of reaction conditions considered.”
§ Efficient Neural Network Models of Chemical Kinetics Using a Latent asinh Rate Transformation † > 2.2.1 The Reaction Mechanism.
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on 2 assumptions not shown here.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Implementing the surrogate models into reactor simulations shows accurate results with a speed-up of 100 000.”
- Also stated in: § 1.3 Scope of this Work · plain
“The preferential oxidation of CO in the presence of H 2 is simulated in a plug-flow reactor showing a speed-up of 100 000 when using neural networks instead of solving the full mechanism.”
§ Efficient Neural Network Models of Chemical Kinetics Using a Latent asinh Rate Transformation † > Abstract
plainstates a causeSingle case- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption not shown here.
- Also stated in: § 1.3 Scope of this Work · plain
44 more claims
Create a free account to see all 112 claims and to verify the 10 references still open.
- It may state two separate claims as one. 38
- Probably not a substantial claim. 36
Citations
Cites
Advances in neural information processing systems 7
(1996) · Computers & Mathematics with Applications · doi:10.1016/s0898-1221(96)90463-0 · Cited by 14,281
Shampine, L. F., & Reichelt, M. W. (1997) · SIAM Journal on Scientific Computing · doi:10.1137/s1064827594276424 · Cited by 3,599
Elasticities and the Inverse Hyperbolic Sine Transformation
Bellemare, M. F., & Wichman, C. J. (2020) · Oxford Bulletin of Economics and Statistics · doi:10.1111/obes.12325 · Cited by 1,159
Computationally efficient implementation of combustion chemistry using in situ adaptive tabulation
Pope, S. B. (1997) · Combustion Theory and Modelling · doi:10.1080/713665229 · Cited by 1,124
An Alternative Family of Transformations
John, J. A., & Draper, N. R. (1980) · Applied Statistics · doi:10.2307/2986305 · Cited by 357
Matera, S., Schneider, W. F., Heyden, A., & Savara, A. (2019) · ACS Catalysis · doi:10.1021/acscatal.9b01234 · Cited by 222
Blasco, J. A., Fueyo, N., Dopazo, C., & Ballester, J. (1998) · Combustion and Flame · doi:10.1016/s0010-2180(97)00211-3 · Cited by 169
Autonomous Discovery of Unknown Reaction Pathways from Data by Chemical Reaction Neural Network
Ji, W., & Deng, S. (2021) · The Journal of Physical Chemistry A · doi:10.1021/acs.jpca.0c09316 · Cited by 161
Mhadeshwar, A. B., & Vlachos, D. G. (2004) · The Journal of Physical Chemistry B · doi:10.1021/jp048698g · Cited by 159
Dual layer automotive ammonia oxidation catalysts: Experiments and computer simulation
Scheuer, A., Hauptmann, W., Drochner, A., Gieshoff, J., Vogel, H., & Votsmeier, M. (2012) · Applied Catalysis B: Environmental · doi:10.1016/j.apcatb.2011.10.032 · Cited by 143
Self-citationTonse, S., Moriarty, N. W., Brown, N. J., & Frenklach, M. (1999) · Israel Journal of Chemistry · doi:10.1002/ijch.199900010 · Cited by 130
Wan, K., Barnaud, C., Vervisch, L., & Domingo, P. (2020) · Combustion and Flame · doi:10.1016/j.combustflame.2020.06.008 · Cited by 119
Zhang, T., Yi, Y., Xu, Y., Chen, Z. X., Zhang, Y., Weinan, E., & Xu, Z. J. (2022) · Combustion and Flame · doi:10.1016/j.combustflame.2022.112319 · Cited by 118
A bi-symmetric log transformation for wide-range data
Webber, J. (2013) · Measurement Science and Technology · doi:10.1088/0957-0233/24/2/027001 · Cited by 111
Structure- and Temperature-Dependence of Pt-Catalyzed Ammonia Oxidation Rates and Selectivities
Ma, H., & Schneider, W. F. (2019) · ACS Catalysis · doi:10.1021/acscatal.8b04251 · Cited by 95
Matera, S., Maestri, M., Cuoci, A., & Reuter, K. (2014) · ACS Catalysis · doi:10.1021/cs501154e · Cited by 92
Whittaker, J., Whitehead, C., & Somers, M. (2005) · Journal of the Royal Statistical Society Series C (Applied Statistics) · doi:10.1111/j.1467-9876.2005.00520.x · Cited by 88
Mullahy, J., & Norton, E. (2022) · doi:10.3386/w30735 · Cited by 87
Quo vadis multiscale modeling in reaction engineering? – A perspective
Wehinger, G. D., Ambrosetti, M., Cheula, R., Ding, Z.-B., Isoz, M., Kreitz, B., Kuhlmann, K., Kutscherauer, M., Niyogi, K., Poissonnier, J., et al. (2022) · Chemical Engineering Research and Design · doi:10.1016/j.cherd.2022.05.030 · Cited by 64
Partopour, B., & Dixon, A. G. (2016) · Computers & Chemical Engineering · doi:10.1016/j.compchemeng.2016.02.015 · Cited by 62
Kinetics-informed neural networks
Gusmão, G. S., Retnanto, A. P., da Cunha, S. C., & Medford, A. J. (2022) · Catalysis Today · doi:10.1016/j.cattod.2022.04.002 · Cited by 60
Efficient implementation of detailed surface chemistry into reactor models using mapped rate data
Votsmeier, M. (2009) · Chemical Engineering Science · doi:10.1016/j.ces.2008.12.006 · Cited by 58
Self-citationRandom Forests for mapping and analysis of microkinetics models
Partopour, B., Paffenroth, R. C., & Dixon, A. G. (2018) · Computers & Chemical Engineering · doi:10.1016/j.compchemeng.2018.04.019 · Cited by 57
Modeling the simultaneous oxidation of CO and H2 on Pt – Promoting effect of H2 on the CO-light-off
Hauptmann, W., Votsmeier, M., Vogel, H., & Vlachos, D. G. (2011) · Applied Catalysis A: General · doi:10.1016/j.apcata.2011.02.031 · Cited by 55
Self-citationvan Keken, P. E., Yuen, D. A., & Petzold, L. R. (1995) · Geophysical & Astrophysical Fluid Dynamics · doi:10.1080/03091929508229763 · Cited by 55
Votsmeier, M., Scheuer, A., Drochner, A., Vogel, H., & Gieshoff, J. (2010) · Catalysis Today · doi:10.1016/j.cattod.2010.01.018 · Cited by 52
Self-citationComputational Fluid Dynamics Simulation of Catalytic Reactors
Deutschmann, O. (2008) · Wiley eBooks · doi:10.1002/9783527610044.hetcat0097 · Cited by 50
Bracconi, M., & Maestri, M. (2020) · Chemical Engineering Journal · doi:10.1016/j.cej.2020.125469 · Cited by 48
Electrons to Reactors Multiscale Modeling: Catalytic CO Oxidation over RuO2
Sutton, J. E., Lorenzi, J. M., Krogel, J. T., Xiong, Q., Pannala, S., Matera, S., & Savara, A. (2018) · ACS Catalysis · doi:10.1021/acscatal.8b00713 · Cited by 46
A Neural Network-Inspired Matrix Formulation of Chemical Kinetics for Acceleration on GPUs
Barwey, S., & Raman, V. (2021) · Energies · doi:10.3390/en14092710 · Cited by 46
Bracconi, M. (2022) · Chemical Engineering and Processing - Process Intensification · doi:10.1016/j.cep.2022.109148 · Cited by 42
Computational Fluid Dynamics of Reacting Flows at Surfaces: Methodologies and Applications
Micale, D., Ferroni, C., Uglietti, R., Bracconi, M., & Maestri, M. (2022) · Chemie Ingenieur Technik · doi:10.1002/cite.202100196 · Cited by 41
Bracconi, M., Maestri, M., & Cuoci, A. (2017) · AIChE Journal · doi:10.1002/aic.15441 · Cited by 41
Efficient simulation of an ammonia oxidation reactor using a solution mapping approach
Scheuer, A., Hirsch, O., Hayes, R. E., Vogel, H., & Votsmeier, M. (2011) · Catalysis Today · doi:10.1016/j.cattod.2011.03.036 · Cited by 40
Self-citationAn improved approach towards more robust deep learning models for chemical kinetics
Xu, H., Jia, M., Chang, Y., & Li, Y. (2021) · Combustion and Flame · doi:10.1016/j.combustflame.2021.111934 · Cited by 38
Döppel, F. A., & Votsmeier, M. (2022) · Chemical Engineering Science · doi:10.1016/j.ces.2022.117964 · Cited by 38
Self-citationResolved‐particle fixed bed CFD with microkinetics for ethylene oxidation
Partopour, B., & Dixon, A. G. (2017) · AIChE Journal · doi:10.1002/aic.15422 · Cited by 37
Hierarchical multi-scale model reduction in the simulation of catalytic converters
Nien, T., Mmbaga, J. P., Hayes, R. E., & Votsmeier, M. (2013) · Chemical Engineering Science · doi:10.1016/j.ces.2013.01.059 · Cited by 34
Self-citationGeneralized reactor neural ODE for dynamic reaction process modeling with physical interpretability
Yin, J., Li, J., Karimi, I. A., & Wang, X. (2022) · Chemical Engineering Journal · doi:10.1016/j.cej.2022.139487 · Cited by 33
Uglietti, R., Bracconi, M., & Maestri, M. (2018) · Reaction Chemistry & Engineering · doi:10.1039/c8re00050f · Cited by 33
(2020) · doi:10.2514/6.2020-0181 · Cited by 33
Accelerating particle-resolved CFD simulations of catalytic fixed-bed reactors with DUO
Daymo, E. A., Hettel, M., Deutschmann, O., & Wehinger, G. D. (2022) · Chemical Engineering Science · doi:10.1016/j.ces.2021.117408 · Cited by 29
Units of measurement and the inverse hyperbolic sine transformation
Aihounton, G. B. D., & Henningsen, A. (2021) · The Econometrics Journal · doi:10.1093/ectj/utaa032 · Cited by 28
Efficient interpolation of precomputed kinetic data employing reduced multivariate Hermite Splines
Klingenberger, M., Hirsch, O., & Votsmeier, M. (2016) · Computers & Chemical Engineering · doi:10.1016/j.compchemeng.2016.12.005 · Cited by 27
Self-citationApproach to combustion calculation using neural network
Nikitin, V. F., Karandashev, I. M., Malsagov, M. Y., & Mikhalchenko, E. V. (2022) · Acta Astronautica · doi:10.1016/j.actaastro.2021.10.034 · Cited by 27
Blasi, J. M., & Kee, R. J. (2016) · Computers & Chemical Engineering · doi:10.1016/j.compchemeng.2015.08.020 · Cited by 26
Uglietti, R., Bracconi, M., & Maestri, M. (2020) · Reaction Chemistry & Engineering · doi:10.1039/c9re00440h · Cited by 22
Mazumder, S. (2005) · Computers & Chemical Engineering · doi:10.1016/j.compchemeng.2005.08.008 · Cited by 20
Rebughini, S., Cuoci, A., Dixon, A. G., & Maestri, M. (2017) · Computers & Chemical Engineering · doi:10.1016/j.compchemeng.2016.11.033 · Cited by 19
Lorenzi, J. M., Stecher, T., Reuter, K., & Matera, S. (2017) · The Journal of Chemical Physics · doi:10.1063/1.4997286 · Cited by 15
Acceleration of Chemical Kinetics Computation with the Learned Intelligent Tabulation (LIT) Method
Haghshenas, M., Mitra, P., Santo, N. D., & Schmidt, D. P. (2021) · Energies · doi:10.3390/en14237851 · Cited by 14
(2021) · doi:10.1007/978-3-030-90539-2_2 · Cited by 11
KiNet: A Deep Neural Network Representation of Chemical Kinetics
Ji, W., & Deng, S. (2021) · arXiv · doi:10.48550/arxiv.2108.00455 · Cited by 10
Insights into reactors through CFD simulations
Wehinger, G. D. (2022) · Chemie Ingenieur Technik · doi:10.1002/cite.202255272 · Cited by 2
Research works printed but not linked: 3
- Chase, 1998, J. Phys. Chem. Ref. Data, 14, 927
- Klumpers, 2023, SSRN Journal, 1
- Chen, 2018, Adv. Neural. Inf. Process. Syst., 31, 1
Cited by
Kircher, T., Döppel, F. A., & Votsmeier, M. (2024) · Chemical Engineering Journal · doi:10.1016/j.cej.2024.149863 · Cited by 51
Self-citationRobust mechanism discovery with atom conserving chemical reaction neural networks
Döppel, F. A., & Votsmeier, M. (2024) · Proceedings of the Combustion Institute · doi:10.1016/j.proci.2024.105507 · Cited by 37
Self-citationBiermann, F., Uglietti, R., Döppel, F. A., Kircher, T., Bracconi, M., Maestri, M., & Votsmeier, M. (2025) · Chemical Engineering Journal · doi:10.1016/j.cej.2025.163598 · Cited by 25
Self-citationRuth, M., Gensch, T., & Schreiner, P. R. (2024) · Angewandte Chemie International Edition · doi:10.1002/anie.202410308 · Cited by 22
Machine Learning Model for CFD Simulations of Fluidized Bed Reactors
Kumar, R., Mohnin Gopinath, M., Kumar, B. K., & Goyal, H. (2025) · Industrial & Engineering Chemistry Research · doi:10.1021/acs.iecr.4c02885 · Cited by 15
Metadata
- Authors
- Döppel, F. A., & Votsmeier, M.
- Affiliations
- Technische Universität Darmstadt (DE)
- Corresponding author
- Votsmeier, Martin
- Venue
- Reaction Chemistry & Engineering
- Publisher
- Royal Society of Chemistry (RSC)
- Published
- 2023
- Volume
- 8
- Issue
- 10
- Pages
- 2620-2631
- ISSN
- 2058-9883 (ISSN-L)
- OpenAlex
- W4383960100
- Cited by
- 24
- References
- 55
- Citations by year
- 2026: 6 · 2025: 8 · 2024: 9 · 2023: 1
- FWCI
- 2.38
- Type
- journal article
- Language
- en
- Open access
- hybrid
- doi.org
- Locations
- Reaction Chemistry & Engineering (publishedVersion, cc-by)
- ChemRxiv (acceptedVersion, cc-by-nc-nd)
- Funders
- Bundesministerium für Bildung und Forschung (05M20RDA)
- Topics
- Physical Sciences › Physics and Astronomy › Statistical and Nonlinear Physics › Model Reduction and Neural Networks
- Physical Sciences › Computer Science › Artificial Intelligence › Neural Networks and Applications
- Physical Sciences › Engineering › Control and Systems Engineering › Control Systems and Identification
- Keywords
- reaction kinetics, neural network surrogate model, lightweight neural networks
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol and qwen3.8-omni-flash).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.