A reactive neural network framework for water-loaded acidic zeolites
Open accessThe DOI matches its Crossref record; 109 of the 112 references it cites were found in a registry or an index; 23 of the 64 sources its claims cite were checked for support.
Abstract
Under operating conditions, the dynamics of water and ions confined within protonic aluminosilicate zeolite micropores are responsible for many of their properties, including hydrothermal stability, acidity and catalytic activity. However, due to high computational cost, operando studies of acidic zeolites are currently rare and limited to specific cases and simplified models. In this work, we have developed a reactive neural network potential (NNP) attempting to cover the entire class of acidic zeolites, including the full range of experimentally relevant water concentrations and Si/Al ratios. This NNP has the potential to dramatically improve sampling, retaining the (meta)GGA DFT level accuracy, with the capacity for discovery of new chemistry, such as collective defect formation mechanisms at the zeolite surface. Furthermore, we exemplify how the NNP can be used as a basis for further extensions/improvements which include data-efficient adoption of higher-level (hybrid) references via Δ-learning and the acceleration of rare event sampling via automatic construction of collective variables. These developments represent a significant step towards accurate simulations of realistic catalysts under operando conditions.
Overview
CoChat Summary
## TL;DR
The paper develops a reactive neural network potential (NNP) for water-loaded Brønsted acidic zeolites (BAS) that aims to cover broad chemical and configurational space, including varied water loadings and Si/Al ratios. It reports that the model retains roughly meta-GGA DFT-level accuracy across diverse tests and can also support Δ-learning and machine-learned collective variables for rare-event sampling.
## Key findings
- The final SCAN+D3(BJ) training database contained 248,439 structures; an ensemble of six SchNet NNPs trained on it achieved average test RMSEs of 5.3 meV atom−1 for energies and 186 meV Å−1 for forces.
- Across 2700 test configurations, the NNP gave an average reaction-energy error of about 6 kJ mol−1 for FAU reaction-path tests, and the authors say this is within DFT accuracy.
- For the in-domain Δ-learning example (proton jump in CHA), 150 training structures were sufficient to reach test-set RMSEs of 1.3 meV atom−1 and 69 meV Å−1, with the ΔNNP deviating by less than 3 kJ/mol from the ωB97X-D3(BJ) reference.
- In out-of-domain tests, performance remained good but worsened modestly: proton jump in FAU showed about 5 kJ mol−1 relative-energy error, and Al–O(H) bond dissociation in FAU reached up to 8 kJ mol−1 error.
- The model captured multiple reactive behaviors not in the training set, including a collective hydrolysis mechanism at an external MFI surface and a high-temperature GIS liquid-state run with thousands of bond-breaking events.
## Methods
- The authors built a broad training set from 500 short (10 ps) AIMD trajectories plus lattice deformations, bulk-water simulations, water clusters, ice/alumina polymorphs, and prior CHA trajectories; the resulting reference dataset used SCAN+D3(BJ) single-point labels.
- Generalization was tested with MD and NEB on unseen systems, including FAU, MFI, GIS, bulk-water-solvated fragments (Si(OH)4 and Al(OH)3), and high-temperature simulations at 1600 K and 3000 K.
- They trained an ensemble of six SchNet-based NNPs and also built two extensions: Δ-learning to ωB97X-D3(BJ) and machine-learned collective variables via a variational autoencoder on NNP representations.
## Limitations
- The paper states that finite-size effects hinder the interpretation of confined-water dynamics in FAU because a small primitive cell was used primarily for benchmarking.
- The authors note that the very high-temperature conditions used in some tests (1600 K and 3000 K) are not applicable to standard applications of water-zeolite systems; this is an evident limitation of the stress tests, not the target use case.
- For the biased-dynamics free-energy calculations, the authors explicitly say their restraints may bias the reported free-energy profiles and that the goal was to demonstrate capability rather than maximal accuracy.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures



Verification
Claims the paper makes
17 Valid · 2 Not valid · 15 Not checked · 66 Own results and interpretations
Some cited sources support this claim and some don’t — read each source’s verdict below.
In the paper
“As a result, the proper mechanistic understanding of BAS-water-zeolite interactions is still lacking, limited to either static calculations at ultra-high vacuum conditions13,14 or exploratory (ab initio) dynamical simulations of narrow scope15–18.”
§ A reactive neural network framework for water-loaded acidic zeolites > Introduction
What the sources say
- Not checked — no full text of the sourcecites [13] Silaghi, M.-C. et al. Regioselectivity of Al–O bond hydrolysis during zeolites dealumination unified by brønsted–evans–polanyi relationship. ACS Catal. 5, 11–15 (2015). source
- Not checked — no full text of the sourcecites [14] Silaghi, M.-C., Chizallet, C., Sauer, J. & Raybaud, P. Dealumination mechanisms of zeolites and extra-framework aluminum confinement. J. Catal. 339, 242–255 (2016). source
- Partly supportedcites [15] Heard, C. J. et al. Fast room temperature lability of aluminosilicate zeolites. Nat. Commun. 10, 1–7 (2019). sourceExperimental — accuracy not yet measured
The Results (Computational studies) describe biased ab initio molecular dynamics of the initial hydrolysis of water-loaded chabazite, so this paper is an example of a dynamical study with a specific scope. It does not state that mechanistic understanding of Brønsted acid site–water–zeolite interactions is generally lacking or that prior work is limited to the two approaches named in the citing sentence. (LLM, conf=0.84)
- Partly supportedcites [16] Nielsen, M. et al. Collective action of water molecules in zeolite dealumination. Catal. Sci. Technol. 9, 3721–3725 (2019). sourceExperimental — accuracy not yet measured
The Introduction of Nielsen et al. (2019) supports the 'static calculations' half of the claim: 'so far all DFT studies of zeolite dealumination have employed a static approach, in which structures are located as minima in the potential energy landscape. Free energies were subsequently derived from nuclear vibrations in the harmonic approximation, which has obvious and severe limitations, in particular at the high temperatures employed in dealumination', and 'the static DFT approach in practice does not allow adequate sampling of all relevant states when taking the collective action of multiple water molecules into account'. However, our copy never describes those static studies as being at 'ultra-high vacuum conditions' (the limitation it names is the harmonic free-energy treatment and the gas-phase reference state for adsorbed water), and it does not characterise prior work as 'exploratory (ab initio) dynamical simulations of narrow scope' — it states the opposite, that all previous dealumination DFT studies were static, with its own DFT-MD/umbrella-sampling work presented as the first in situ dynamical investigation at realistic steaming conditions (Results and discussion; Conclusions). The claim's framing is also broader (BAS–water–zeolite interactions in general) than the dealumination-specific statements in the source. (LLM, conf=0.72)
- Disputedcites [17] Grifoni, E. et al. Confinement effects and acid strength in zeolites. Nat. Commun. 12, 2630 (2021). sourceExperimental — accuracy not yet measured
The Introduction (page unavailable) describes full ab initio metadynamics across four zeolite frameworks and multiple water loadings, yielding thermodynamic and mechanistic information at atomistic detail—not a dynamical study of narrow scope. The Results and discussion report general findings about BAS solvation, contrary to the claim that mechanistic understanding remains limited to such studies. (LLM, conf=0.85)
Checked by a reasoning model — may be wrong.
- Not checked — no full text of the sourcecites [18] Jin, M. et al. The role of water loading and germanium content in germanosilicate hydrolysis. J. Phys. Chem. C 125, 23744–23757 (2021). 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.
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“A standard tool for accelerating the reactive sampling in zeolite-water systems, and thus reaching experimentally relevant timescale or realistic levels of model complexity is the class of reactive analytical potentials, for example, ReaxFF19. However, due to their fixed functional form, these potentials have limited transferability to systems with different chemical composition20.”
§ A reactive neural network framework for water-loaded acidic zeolites > Introduction
What the sources say
- Partly supportedcites [19] Senftle, T. P. et al. The ReaxFF reactive force-field: development, applications and future directions. npj Comput. Mater. 2, 1–14 (2016). sourceExperimental — accuracy not yet measured
The section on ReaxFF parameterisations and development branches (page unavailable in our copy) says that parameter sets cannot be combined arbitrarily and that transferring them between branches requires extensive refitting. It does not attribute this limitation to a fixed functional form; it discusses the branches within a common functional form and points instead to ReaxFF’s lack of atom typing as a complication. (LLM, conf=0.86)
- Supportedcites [20] Vandermause, J., Xie, Y., Lim, J. S., Owen, C. J. & Kozinsky, B. Active learning of reactive Bayesian force fields applied to heterogeneous catalysis dynamics of H/Pt. Nat. Commun. 13, 5183 (2022). source
The Introduction says ReaxFF models require expert fine-tuning for each system and can deviate substantially from ab initio results because of their limited parametric functional form. That supports the citing sentence’s point that the potentials do not readily transfer between chemically different systems. (LLM, conf=0.86)
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.
- Partly supported
The sources with a verdict support this claim; the rest weren’t checked yet, have no full text we could read, or their verdicts aren’t rated yet.
In the paper
“This relationship governs many features of BAS zeolites including (i) proton solvation, and thus acidity7,8”
§ A reactive neural network framework for water-loaded acidic zeolites > Introduction
What the sources say
- Not checked — no full text of the sourcecites [7] Pfriem, N. et al. Role of the ionic environment in enhancing the activity of reacting molecules in zeolite pores. Science 372, 952–957 (2021). source
- Supportedcites [8] Bates, J. S., Bukowski, B. C., Greeley, J. & Gounder, R. Structure and solvation of confined water and water–ethanol clusters within microporous Brønsted acids and their effects on ethanol dehydration catalysis. Chem. Sci. 11, 7102–7122 (2020). source
The Introduction (page not available in our copy) says water stabilizes solvated protons and that solvent polarity changes protonation equilibrium constants in zeolites. It also reports that solvent properties can change the strength of Brønsted acid sites, supporting the link between water interactions, proton solvation, and acidity. (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.
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“(iii) water mobility and clustering in zeolite pores10”
§ A reactive neural network framework for water-loaded acidic zeolites > Introduction
What the sources say
- Partly supportedcites [10] Fasano, M. et al. Interplay between hydrophilicity and surface barriers on water transport in zeolite membranes. Nat. Commun. 7, 12762 (2016). sourceExperimental — accuracy not yet measured
The Results section (‘Intracrystalline diffusion in zeolites’; page not specified) reports that water–zeolite interactions affect water mobility in MFI pores at low hydration, while water–water interactions predominate at higher hydration. It examines water–water hydrogen bonds but does not establish that Brønsted acid sites govern water clustering. (LLM, conf=0.86)
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.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Two paradigms dominate the MLP field currently: (i) training of universal MLP that cover large parts of the chemical space with dozens of elements, but with limited coverage of the configuration space, e.g., not considering all relevant chemical reactions with the associated transition states, e.g., OC22, CHGNet, and others25–29”
§ A reactive neural network framework for water-loaded acidic zeolites > Introduction
What the sources say
- Partly supportedcites [25] Takamoto, S. et al. Towards universal neural network potential for material discovery applicable to arbitrary combination of 45 elements. Nat. Commun. 13, 2991 (2022). sourceExperimental — accuracy not yet measured
The Abstract and Methods describe a potential spanning 45 elements and a training dataset with no transition-state structures, supporting the claim about broad chemical coverage and a gap in training coverage. But the Results report that the model predicts transition-state energies and evaluates chemical-reaction pathways, so the paper does not support an unqualified claim that it does not consider reactions or their transition states. The indexed copy provides section labels but no page numbers. (LLM, conf=0.85)
- Not checked yetcites [26] Tran, R. et al. The open catalyst 2022 (OC22) dataset and challenges for oxide electrocatalysts. ACS Catal. 13, 3066–3084 (2023). source
Checked before — its earlier verdict is held back until a new check runs.
- Not checked yetcites [27] Deng, B. et al. CHGNet as a pretrained universal neural network potential for charge-informed atomistic modelling. Nat. Mach. Intell. 5, 1031–1041 (2023). source
Checked before — its earlier verdict is held back until a new check runs.
- Not checked yetcites [28] Zhang, D. et al. DPA-2: Towards a universal large atomic model for molecular and material simulation. Preprint at https://doi.org/10.48550/arXiv.2312.15492 (2023). source
Checked before — its earlier verdict is held back until a new check runs.
- Not checked yetcites [29] Batatia, I. et al. A foundation model for atomistic materials chemistry. Preprint at https://doi.org/10.48550/arXiv.2401.00096 (2024). source
Checked before — its earlier verdict is held back until a new check runs.
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
49 more claims · 4 with their cited sources checked
Create a free account to see all 100 claims and to verify the 23 references still open.
- Probably not a substantial claim. 14
- It may state two separate claims as one. 38
Citations
Cites
Generalized Gradient Approximation Made Simple
Perdew, J. P., Burke, K., & Ernzerhof, M. (1996) · Physical Review Letters · doi:10.1103/physrevlett.77.3865 · Cited by 215,824 · cited 2×
“All AIMD simulations and structure optimizations were performed at the computationally less demanding PBE+D3(BJ)[29] level employing the dispersion correction of Grimme et al.”
MethodEfficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set
Kresse, G., & Furthmüller, J. (1996) · Physical Review B · doi:10.1103/physrevb.54.11169 · Cited by 123,652
Projector augmented-wave method
Blöchl, P. E. (1994) · Physical Review B · doi:10.1103/physrevb.50.17953 · Cited by 92,732
From ultrasoft pseudopotentials to the projector augmented-wave method
Kresse, G., & Joubert, D. (1999) · Physical Review B · doi:10.1103/physrevb.59.1758 · Cited by 85,307 · cited 1×
“4) along with standard versions of the Projector Augmented-Wave (PAW) potentials.[79, 80] Calculations at constant volume were performed with an energy cutoff of 400 eV.”
BackgroundKresse, G., & Furthmüller, J. (1996) · Computational Materials Science · doi:10.1016/0927-0256(96)00008-0 · Cited by 75,638
Grimme, S., Antony, J., Ehrlich, S., & Krieg, H. (2010) · The Journal of Chemical Physics · doi:10.1063/1.3382344 · Cited by 59,686
Ab initio molecular dynamics for liquid metals
Kresse, G., & Hafner, J. (1993) · Physical Review B · doi:10.1103/physrevb.47.558 · Cited by 45,727
Effect of the damping function in dispersion corrected density functional theory
Grimme, S., Ehrlich, S., & Goerigk, L. (2011) · Journal of Computational Chemistry · doi:10.1002/jcc.21759 · Cited by 28,221 · cited 2×
“The bulk of the data-base is derived from 500 short (10 ps) ab initio molecular dynamics (AIMD) trajectories (at PBE +D3(BJ) level) 35 – 37 using a set of BAS zeolite models.”
MethodCanonical dynamics: Equilibrium phase-space distributions
Hoover, W. G. (1985) · Physical Review A · doi:10.1103/physreva.31.1695 · Cited by 23,465 · cited 2×
“The AIMD equilibration used the canonical (NVT) ensemble along with a 1 fs time step, the Nosé–Hoover thermostat[70, 71] and with hydrogen being replaced by tritium.”
MethodKresse, G., & Hafner, J. (1994) · Physical Review B · doi:10.1103/physrevb.49.14251 · Cited by 23,019
Research works printed but not linked: 3
- Schütt, K., Unke, O. & Gastegger, M. Equivariant message passing for the prediction of tensorial properties and molecular spectra. In Proceedings of the 38th International Conference on Machine Learning 9377–9388 (PMLR, 2021).
- van der Maaten, L. & Hinton, G. Visualizing Data using t-SNE. J. Mach. Learn Res. 9, 2579–2605 (2008).
- Kingma, D. P. & Ba, J. Adam: a method for stochastic optimization. In 3rd International Conference on Learning Representations, ICLR 2015, San Diego, CA, USA, May 7-9, 2015, Conference Track Proceedings (eds Bengio, Y. & LeCun, Y.) (ICLR, 2015).
Cited by
Application of machine learning interatomic potentials in heterogeneous catalysis
Olajide, G., Baral, K., Ezendu, S., Soyemi, A., & Szilvási, T. (2025) · Journal of Catalysis · doi:10.1016/j.jcat.2025.116202 · Cited by 34
Bocus, M., Vandenhaute, S., & Speybroeck, V. V. (2024) · Angewandte Chemie International Edition · doi:10.1002/anie.202413637 · Cited by 28
Wang, X., Wang, Q., Wang, C., Chu, Y., Hu, M., Deng, F., Yu, J., & Xu, J. (2025) · Journal of the American Chemical Society · doi:10.1021/jacs.5c01756 · Cited by 24
Learning a reactive potential for silica-water through uncertainty attribution
Roy, S., Dürholt, J. P., Asche, T. S., Zipoli, F., & Gómez‐Bombarelli, R. (2024) · Nature Communications · doi:10.1038/s41467-024-50407-9 · Cited by 28
“Several NNIP have been constructed for pure water 18 – 20 , amorphous and liquid silica 21,22 and water-fi lled aluminosilicate zeolites 23 .”
Zeolite–Water Chemistry: Characterization Methods to Unveil Zeolite Structure
Almeida, J. N., Song, L., Askarli, S., Chung, S., & Ruiz-Martínez, J. (2025) · Chemistry - Methods · doi:10.1002/cmtd.202400076 · Cited by 29
Metadata
- Authors
- Erlebach, A., Šípka, M., Saha, I., Nachtigall, P., Heard, C. J., & Grajciar, L.
- Affiliations
- Charles University (CZ)
- Corresponding author
- Erlebach, Andreas
- Grajciar, Lukáš
- ORCID
- Venue
- Nature Communications
- Published
- 2024-05-17
- Volume
- 15
- Issue
- 1
- Pages
- 4215-4215
- ISSN
- 2041-1723 (ISSN-L)
- PMC
- PMC11101627
- OpenAlex
- W4396999021
- Cited by
- 47
- References
- 112
- Citations by year
- 2026: 13 · 2025: 20 · 2024: 13
- FWCI
- 4.47
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- cc-by
- creativecommons.org/licenses/by/4.0 (tdm)
- creativecommons.org/licenses/by/4.0 (vor)
- Locations
- Nature Communications (publishedVersion, cc-by)
- PubMed Central (publishedVersion, cc-by-nc-sa)
- PubMed (publishedVersion)
- arXiv (Cornell University) (submittedVersion, cc-by-nc-sa)
- RePEc: Research Papers in Economics (submittedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
- arXiv (Cornell University)
- Funders
- Univerzita Karlova v Praze (PRIMUS/20/SCI/004, UNCE/SCI/014)
- Ministerstvo Školství, Mládeže a Tělovýchovy (ID:90254, CZ.02.1.01/0.0/0.0/15 003/0000417)
- Grantová Agentura České Republiky (23-07616S)
- Topics
- Physical Sciences › Materials Science › Materials Chemistry › Machine Learning in Materials Science
- Physical Sciences › Chemistry › Inorganic Chemistry › Zeolite Catalysis and Synthesis
- Physical Sciences › Physics and Astronomy › Atomic and Molecular Physics, and Optics › Advanced Chemical Physics Studies
- Keywords
- acid zeolites, zeolite acidity, hydrothermal stability, Si/Al ratio, delta learning, rare event sampling, collective variables, machine learning interatomic potentials
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
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: A reactive neural network framework for water-loaded acidic zeolites. https://app.cochat.ai/papers/10.1038/s41467-024-48609-2
- The paper
- Erlebach, A., Šípka, M., Saha, I., Nachtigall, P., Heard, C. J., & Grajciar, L. (2024). A reactive neural network framework for water-loaded acidic zeolites. Nature Communications. https://doi.org/10.1038/s41467-024-48609-2
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.