MADField: Multi-fidelity Amortized Density Field for Adsorption in Nanoporous Materials

Honghui Kim; Seongsu Kim; Sungsoo Ahn; Yoonho Kim

arxiv: 2606.21284 · v2 · pith:6EPGX7CZnew · submitted 2026-06-19 · ⚛️ physics.comp-ph

MADField: Multi-fidelity Amortized Density Field for Adsorption in Nanoporous Materials

Yoonho Kim , Seongsu Kim , Sungsoo Ahn , Honghui Kim This is my paper

Pith reviewed 2026-07-02 22:14 UTC · model grok-4.3

classification ⚛️ physics.comp-ph

keywords adsorptionnanoporous materialsdensity functional theorygrand canonical Monte Carlomulti-fidelity learningdensity field estimationgas storage screening

0 comments

The pith

MADField predicts equilibrium adsorbate density fields from multi-fidelity cDFT and GCMC data to estimate gas uptake in nanoporous materials.

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

The paper presents MADField as a method that treats adsorption equilibrium as a density-field prediction task rather than direct uptake calculation. It trains an amortized model on density fields computed by classical density functional theory across many structures and refines it with higher-fidelity grand canonical Monte Carlo density labels on a smaller set. The predicted fields are integrated to obtain uptake values and can also initialize or accelerate the original cDFT solvers. On screening tasks this produces substantially more accurate rankings than either simulation method used alone. The approach is demonstrated on a database of hundreds of thousands of candidate structures for methane working capacity.

Core claim

MADField reframes adsorption prediction as equilibrium density-field estimation. It learns from two complementary fidelities by combining broad and scalable cDFT density supervision with higher-fidelity GCMC density labels, then recovers gas uptake by integrating the predicted density field. This yields uptake accuracy improvements of 6.0x over the strongest cDFT baseline and 15.4x over GCMC, accelerates cDFT solvers by a factor of two while recovering 42 percent of previously failing cases, and on the 270k-structure ARC-MOF database delivers 56x higher average precision than the strongest baseline at five orders of magnitude lower inference cost than GCMC.

What carries the argument

Multi-fidelity amortized neural network that takes material structure as input and outputs a predicted adsorbate density field trained jointly on cDFT and GCMC data.

If this is right

Uptake predictions improve by a factor of 6.0 over cDFT baselines and 15.4 over GCMC baselines.
Predicted density fields reduce cDFT solver iterations by a factor of two and recover 42 percent of cases that fail under standard settings.
On the 270k-structure database the method achieves 56 times higher average precision than the strongest baseline.
Inference runs five orders of magnitude faster than GCMC while still allowing selection of the top 1.7 percent of candidates to recover 95 percent of high-capacity targets.

Where Pith is reading between the lines

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

The same density-field representation could support prediction of adsorption at multiple temperatures or pressures without separate models for each condition.
Density-field outputs may highlight local structural motifs that drive high uptake, offering a route to inverse design beyond ranking.
Because the model produces a full spatial field rather than a scalar uptake, it could be coupled to transport or reaction models that depend on local concentration.

Load-bearing premise

The multi-fidelity trained model generalizes accurately to unseen structures in large databases such as ARC-MOF without significant bias introduced by the lower-fidelity cDFT supervision.

What would settle it

Running full GCMC simulations on the top 1.7 percent of ARC-MOF candidates ranked by MADField and measuring whether the fraction of true high-capacity structures recovered matches the reported 95 percent.

read the original abstract

High-throughput computational screening of nanoporous materials for gas storage and separation requires fast and accurate characterization of adsorption equilibrium. Particle-based grand canonical Monte Carlo (GCMC) and density-based classical density functional theory (cDFT) provide simulation-based estimates of gas uptake and adsorbate density fields, but their speed-accuracy tradeoff remains insufficient for large-scale screening. In this work, we address this gap with Multi-fidelity Amortized Density Field for Adsorption in Nanoporous Materials (MADField), which reframes adsorption prediction as equilibrium density-field estimation. MADField learns from two complementary fidelities, combining broad and scalable cDFT density supervision with higher-fidelity GCMC density labels, and recovers gas uptake by integrating the predicted density field. MADField improves uptake accuracy over the strongest baselines by 6.0x for cDFT and 15.4x for GCMC, and its predicted fields accelerate cDFT solvers with 2.0x fewer iteration steps while recovering 42 percent of cases that fail under standard settings. Finally, we evaluate MADField for conventional CH4 working capacity screening on the 270k-structure ARC-MOF database. Within this space of extremely rare high-capacity targets, 167 in total, the model achieves 56x higher average precision than the strongest baseline and accelerates inference by five orders of magnitude compared to GCMC. By prioritizing the MADField rankings, selecting the top 1.7 percent of candidates recovers 95 percent of all targets, while the top 6 percent ensures 100 percent recall.

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

MADField introduces a multi-fidelity amortized density field model that reports large gains in adsorption screening speed and precision on ARC-MOF, but the abstract leaves the generalization and bias checks thin.

read the letter

Hi,

The main point is that MADField treats adsorption uptake as the integral of a learned density field, trained on lots of cDFT data plus some GCMC labels. It claims 6x better accuracy than cDFT baselines, 15.4x over GCMC, plus 2x faster cDFT convergence and 56x higher average precision on the 270k ARC-MOF CH4 screen, recovering 95% of the 167 high-capacity targets from the top 1.7% of ranked structures.

What stands out is the practical framing: amortizing the density field across fidelities and then using it both for direct uptake and to warm-start cDFT. The screening numbers on a real large database are concrete and relevant to gas storage work.

The soft spot is the one flagged in the stress test. The abstract gives no information on whether the training structures were filtered for overlap or similarity with ARC-MOF, nor does it show direct GCMC validation of the integrated uptakes on the high-capacity subset. If cDFT systematic errors propagate through the integral or if the model saw similar structures during training, the reported gains could be inflated. That is a load-bearing assumption for the headline result, and it is not addressed in what is visible.

This is for people running high-throughput MOF screens who need something faster than GCMC but more accurate than plain cDFT. A reader working on ML surrogates for classical simulations would find the multi-fidelity setup worth looking at.

It deserves a serious referee because the claims are specific, the application is well-motivated, and the gaps are fixable with added validation details rather than fundamental problems with the approach.

I'd send it to review.

Referee Report

3 major / 1 minor

Summary. The paper introduces MADField, a multi-fidelity amortized model for predicting adsorbate density fields in nanoporous materials. It trains on broad cDFT density data plus sparse GCMC labels, integrates the predicted fields to recover uptake, and reports accuracy gains over baselines (6.0x vs cDFT, 15.4x vs GCMC), acceleration of cDFT solvers (2x fewer steps, 42% recovery of failures), and strong screening performance on the 270k ARC-MOF database (56x average precision, 95% target recovery by ranking top 1.7% of candidates).

Significance. If the generalization and integration claims hold, MADField would provide a practical bridge between scalable low-fidelity and accurate high-fidelity simulations, enabling reliable identification of rare high-capacity adsorbents in databases far larger than those accessible to direct GCMC.

major comments (3)

[Methods and Results] Methods and Results sections: the manuscript supplies no details on model architecture, training data construction (including how cDFT and GCMC structures were chosen), data splits, structural similarity filtering, or validation protocols, preventing assessment of whether the stated quantitative gains are supported by the underlying data and methods.
[Screening experiment on ARC-MOF] Screening experiment on ARC-MOF (abstract and §5): the headline claims of 56x average precision and 95% recovery of the 167 high-capacity targets at top 1.7% require explicit confirmation that the 270k evaluation structures are disjoint from the training set and that integrated uptakes from the predicted density fields match direct GCMC on a held-out high-capacity subset; without this, bias from cDFT supervision or distributional overlap cannot be ruled out.
[§4] §4 (performance claims): the reported 6.0x and 15.4x uptake accuracy improvements, 2.0x iteration reduction, and 42% failure recovery lack accompanying error metrics, statistical significance tests, or baseline implementation details, making the load-bearing accuracy claims difficult to evaluate.

minor comments (1)

[Methods] Clarify the precise form of the multi-fidelity loss and the numerical integration step from predicted density field to uptake.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the detailed and constructive review. We address each major comment below and will revise the manuscript to provide the requested details and clarifications.

read point-by-point responses

Referee: [Methods and Results] Methods and Results sections: the manuscript supplies no details on model architecture, training data construction (including how cDFT and GCMC structures were chosen), data splits, structural similarity filtering, or validation protocols, preventing assessment of whether the stated quantitative gains are supported by the underlying data and methods.

Authors: We agree that the current manuscript lacks sufficient methodological detail for full reproducibility and evaluation. In the revised version we will expand the Methods section to specify the model architecture, the procedure for selecting and constructing the cDFT and GCMC training structures, the data splitting strategy, any structural similarity filtering applied, and the complete validation protocols. These additions will directly support assessment of the reported gains. revision: yes
Referee: [Screening experiment on ARC-MOF] Screening experiment on ARC-MOF (abstract and §5): the headline claims of 56x average precision and 95% recovery of the 167 high-capacity targets at top 1.7% require explicit confirmation that the 270k evaluation structures are disjoint from the training set and that integrated uptakes from the predicted density fields match direct GCMC on a held-out high-capacity subset; without this, bias from cDFT supervision or distributional overlap cannot be ruled out.

Authors: We acknowledge the necessity of demonstrating independence between training and evaluation sets. The revised manuscript will add explicit statements confirming that the 270k ARC-MOF structures are disjoint from the training data. We will also include a direct comparison of integrated uptakes from the predicted density fields against GCMC on a held-out high-capacity subset to substantiate the screening metrics. revision: yes
Referee: [§4] §4 (performance claims): the reported 6.0x and 15.4x uptake accuracy improvements, 2.0x iteration reduction, and 42% failure recovery lack accompanying error metrics, statistical significance tests, or baseline implementation details, making the load-bearing accuracy claims difficult to evaluate.

Authors: We agree that additional quantitative support is warranted. The revised §4 will incorporate error metrics (including standard deviations), statistical significance tests where appropriate, and expanded descriptions of the baseline implementations to allow rigorous evaluation of the accuracy, iteration, and failure-recovery claims. revision: yes

Circularity Check

0 steps flagged

No significant circularity; standard multi-fidelity ML surrogate trained on external simulation outputs

full rationale

The paper frames MADField as a neural network trained on cDFT density supervision plus sparse GCMC labels to predict density fields, with uptake recovered by integration. This is a conventional supervised learning pipeline using independent simulation codes as data sources and evaluated on the external ARC-MOF database. No equations reduce a claimed prediction to a fitted parameter by construction, no uniqueness theorems are imported from self-citations, and no ansatz or renaming of known results is presented as a derivation. The reported gains are empirical performance numbers on held-out structures rather than tautological identities.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract provides insufficient detail to identify specific free parameters, axioms, or invented entities; review is limited to the provided summary.

pith-pipeline@v0.9.1-grok · 5815 in / 1271 out tokens · 32387 ms · 2026-07-02T22:14:06.125135+00:00 · methodology

Review history (2 revisions) →

discussion (0)

Reference graph

Works this paper leans on

48 extracted references · 8 canonical work pages · 4 internal anchors

[1]

Large-scale screening of hypothetical metal–organic frameworks.Nature chemistry, 4(2):83–89, 2012

Christopher E Wilmer, Michael Leaf, Chang Yeon Lee, Omar K Farha, Brad G Hauser, Joseph T Hupp, and Randall Q Snurr. Large-scale screening of hypothetical metal–organic frameworks.Nature chemistry, 4(2):83–89, 2012

2012
[2]

Evaluating metal–organic frameworks for natural gas storage

Jarad A Mason, Mike Veenstra, and Jeffrey R Long. Evaluating metal–organic frameworks for natural gas storage. Chemical Science, 5(1):32–51, 2014

2014
[3]

In silico screening of carbon-capture materials.Nature materials, 11(7):633–641, 2012

Li-Chiang Lin, Adam H Berger, Richard L Martin, Jihan Kim, Joseph A Swisher, Kuldeep Jariwala, Chris H Rycroft, Abhoyjit S Bhown, Michael W Deem, Maciej Haranczyk, et al. In silico screening of carbon-capture materials.Nature materials, 11(7):633–641, 2012

2012
[4]

A holistic platform for accelerating sorbent-based carbon capture.Nature, 632(8023):89–94, 2024

Charithea Charalambous, Elias Moubarak, Johannes Schilling, Eva Sanchez Fernandez, Jin-Yu Wang, Laura Herraiz, Fergus Mcilwaine, Shing Bo Peh, Matthew Garvin, Kevin Maik Jablonka, et al. A holistic platform for accelerating sorbent-based carbon capture.Nature, 632(8023):89–94, 2024

2024
[5]

Grand canonical ensemble monte carlo for a lennard-jones fluid.Molecular Physics, 29(1):307–311, 1975

DJ Adams. Grand canonical ensemble monte carlo for a lennard-jones fluid.Molecular Physics, 29(1):307–311, 1975

1975
[6]

Raspa: molecular simulation software for adsorption and diffusion in flexible nanoporous materials.Molecular Simulation, 42(2):81–101, 2016

David Dubbeldam, Sofía Calero, Donald E Ellis, and Randall Q Snurr. Raspa: molecular simulation software for adsorption and diffusion in flexible nanoporous materials.Molecular Simulation, 42(2):81–101, 2016

2016
[7]

Raspa3: A monte carlo code for computing adsorption and diffusion in nanoporous materials and thermodynamics properties of fluids.The Journal of Chemical Physics, 161(11), 2024

YA Ran, S Sharma, SRG Balestra, Z Li, S Calero, TJH Vlugt, RQ Snurr, and D Dubbeldam. Raspa3: A monte carlo code for computing adsorption and diffusion in nanoporous materials and thermodynamics properties of fluids.The Journal of Chemical Physics, 161(11), 2024

2024
[8]

The nature of the liquid-vapour interface and other topics in the statistical mechanics of non-uniform, classical fluids.Advances in physics, 28(2):143–200, 1979

Robert Evans. The nature of the liquid-vapour interface and other topics in the statistical mechanics of non-uniform, classical fluids.Advances in physics, 28(2):143–200, 1979. 11

1979
[9]

Fundamental measure theory for hard-sphere mixtures: a review.Journal of Physics: Condensed Matter, 22(6):063102, 2010

Roland Roth. Fundamental measure theory for hard-sphere mixtures: a review.Journal of Physics: Condensed Matter, 22(6):063102, 2010

2010
[10]

Classical density functional theory as a fast and accurate method for adsorption property prediction of porous materials.AIChE Journal, 71(6):e18779, 2025

Vincent Dufour-Décieux, Philipp Rehner, Johannes Schilling, Elias Moubarak, Joachim Gross, and André Bardow. Classical density functional theory as a fast and accurate method for adsorption property prediction of porous materials.AIChE Journal, 71(6):e18779, 2025

2025
[11]

Efficient prediction of multicomponent adsorption isotherms and enthalpies of adsorption in mofs using classical density functional theory.The Journal of Physical Chemistry B, 2026

Nadine Thiele, Tiong Wei Teh, Benjamin Bursik, Marcel Granderath, Gernot Bauer, Vincent Dufour-Décieux, Philipp Rehner, Rolf Stierle, André Bardow, Niels Hansen, et al. Efficient prediction of multicomponent adsorption isotherms and enthalpies of adsorption in mofs using classical density functional theory.The Journal of Physical Chemistry B, 2026

2026
[12]

Gas adsorption meets deep learning: voxelizing the potential energy surface of metal-organic frameworks.Scientific Reports, 14(1):2242, 2024

Antonios P Sarikas, Konstantinos Gkagkas, and George E Froudakis. Gas adsorption meets deep learning: voxelizing the potential energy surface of metal-organic frameworks.Scientific Reports, 14(1):2242, 2024

2024
[13]

Interpretable graph transformer network for predicting adsorption isotherms of metal–organic frameworks.Journal of Chemical Information and Modeling, 62(22): 5446–5456, 2022

Pin Chen, Rui Jiao, Jinyu Liu, Yang Liu, and Yutong Lu. Interpretable graph transformer network for predicting adsorption isotherms of metal–organic frameworks.Journal of Chemical Information and Modeling, 62(22): 5446–5456, 2022

2022
[14]

Understanding and predicting the spatially resolved adsorption properties of nanoporous materials.Journal of Chemical Theory and Computation, 20(12):5259–5275, 2024

Yangzesheng Sun and J Ilja Siepmann. Understanding and predicting the spatially resolved adsorption properties of nanoporous materials.Journal of Chemical Theory and Computation, 20(12):5259–5275, 2024

2024
[15]

Rapid prediction of single-site adsorbate probability distributions in metal–organic frameworks using graph neural networks.Digital Discovery, 2026

Jake Burner, Olivier Marchand, Rosa Cicciarella, Marco Gibaldi, and Tom K Woo. Rapid prediction of single-site adsorbate probability distributions in metal–organic frameworks using graph neural networks.Digital Discovery, 2026

2026
[16]

High-throughput computational screening of metal–organic frameworks

Yamil J Colón and Randall Q Snurr. High-throughput computational screening of metal–organic frameworks. Chemical Society Reviews, 43(16):5735–5749, 2014

2014
[17]

The materials genome in action: identifying the performance limits for methane storage.Energy & Environmental Science, 8(4):1190–1199, 2015

Cory M Simon, Jihan Kim, Diego A Gomez-Gualdron, Jeffrey S Camp, Yongchul G Chung, Richard L Martin, Rocio Mercado, Michael W Deem, Dan Gunter, Maciej Haranczyk, et al. The materials genome in action: identifying the performance limits for methane storage.Energy & Environmental Science, 8(4):1190–1199, 2015

2015
[18]

Elmar Sauer and Joachim Gross. Classical density functional theory for liquid–fluid interfaces and confined systems: A functional for the perturbed-chain polar statistical associating fluid theory equation of state.Industrial & Engineering Chemistry Research, 56(14):4119–4135, 2017

2017
[19]

Computational screening of trillions of metal–organic frameworks for high-performance methane storage.ACS Applied Materials & Interfaces, 13(20):23647–23654, 2021

Sangwon Lee, Baekjun Kim, Hyun Cho, Hooseung Lee, Sarah Yunmi Lee, Eun Seon Cho, and Jihan Kim. Computational screening of trillions of metal–organic frameworks for high-performance methane storage.ACS Applied Materials & Interfaces, 13(20):23647–23654, 2021

2021
[20]

Kaihang Shi, Zhao Li, Dylan M Anstine, Dai Tang, Coray M Colina, David S Sholl, J Ilja Siepmann, and Randall Q Snurr. Two-dimensional energy histograms as features for machine learning to predict adsorption in diverse nanoporous materials.Journal of chemical theory and computation, 19(14):4568–4583, 2023

2023
[21]

Unified physio-thermodynamic descriptors via learned co2 adsorption properties in metal-organic frameworks.npj Computational Materials, 11(1):225, 2025

Emily Lin, Yang Zhong, Gang Chen, and Sili Deng. Unified physio-thermodynamic descriptors via learned co2 adsorption properties in metal-organic frameworks.npj Computational Materials, 11(1):225, 2025

2025
[22]

Direct prediction of gas adsorption via spatial atom interaction learning.Nature Communications, 14 (1):7043, 2023

Jiyu Cui, Fang Wu, Wen Zhang, Lifeng Yang, Jianbo Hu, Yin Fang, Peng Ye, Qiang Zhang, Xian Suo, Yiming Mo, et al. Direct prediction of gas adsorption via spatial atom interaction learning.Nature Communications, 14 (1):7043, 2023

2023
[23]

A multi-modal pre-training transformer for universal transfer learning in metal–organic frameworks.Nature Machine Intelligence, 5(3):309–318, 2023

Yeonghun Kang, Hyunsoo Park, Berend Smit, and Jihan Kim. A multi-modal pre-training transformer for universal transfer learning in metal–organic frameworks.Nature Machine Intelligence, 5(3):309–318, 2023

2023
[24]

A comprehensive transformer-based approach for high-accuracy gas adsorption predictions in metal-organic frameworks.Nature Communications, 15(1):1904, 2024

Jingqi Wang, Jiapeng Liu, Hongshuai Wang, Musen Zhou, Guolin Ke, Linfeng Zhang, Jianzhong Wu, Zhifeng Gao, and Diannan Lu. A comprehensive transformer-based approach for high-accuracy gas adsorption predictions in metal-organic frameworks.Nature Communications, 15(1):1904, 2024

1904
[25]

Deepdft: Neural message passing network for accurate charge density prediction

Peter Bjørn Jørgensen and Arghya Bhowmik. Deepdft: Neural message passing network for accurate charge density prediction. InMachine Learning for Molecules Workshop@ NeurIPS 2020, 2020

2020
[26]

Equivariant graph neural networks for fast electron density estimation of molecules, liquids, and solids.npj Computational Materials, 8(1):183, 2022

Peter Bjørn Jørgensen and Arghya Bhowmik. Equivariant graph neural networks for fast electron density estimation of molecules, liquids, and solids.npj Computational Materials, 8(1):183, 2022

2022
[27]

Swin transformer: Hierarchical vision transformer using shifted windows

Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo. Swin transformer: Hierarchical vision transformer using shifted windows. InProceedings of the IEEE/CVF international conference on computer vision, pages 10012–10022, 2021. 12

2021
[28]

Machine learning the quantum-chemical properties of metal–organic frameworks for accelerated materials discovery.Matter, 4(5):1578–1597, 2021

Andrew S Rosen, Shaelyn M Iyer, Debmalya Ray, Zhenpeng Yao, Alan Aspuru-Guzik, Laura Gagliardi, Justin M Notestein, and Randall Q Snurr. Machine learning the quantum-chemical properties of metal–organic frameworks for accelerated materials discovery.Matter, 4(5):1578–1597, 2021

2021
[29]

Synthetic data enable experiments in atomistic machine learning.Digital Discovery, 2(3):651–662, 2023

John LA Gardner, Zoé Faure Beaulieu, and Volker L Deringer. Synthetic data enable experiments in atomistic machine learning.Digital Discovery, 2(3):651–662, 2023

2023
[30]

Boyd, Stephen Maley, Marco Gibaldi, Scott Simrod, and Tom K

Jake Burner, Jun Luo, Andrew White, Adam Mirmiran, Ohmin Kwon, Peter G. Boyd, Stephen Maley, Marco Gibaldi, Scott Simrod, and Tom K. Woo. ab initio repeat charge mof database (arc-mof), July 2022. URL https://doi.org/10.5281/zenodo.6908728

work page doi:10.5281/zenodo.6908728 2022
[31]

Metal-organic framework ch4@1bar adsorbate probability distributions, August 2025

Jake Burner. Metal-organic framework ch4@1bar adsorbate probability distributions, August 2025. URL https://doi.org/10.5281/zenodo.16800893

work page doi:10.5281/zenodo.16800893 2025
[32]

Metal-organic framework ch4@65bar adsorbate probability distributions, August 2025

Jake Burner. Metal-organic framework ch4@65bar adsorbate probability distributions, August 2025. URL https://doi.org/10.5281/zenodo.16801034

work page doi:10.5281/zenodo.16801034 2025
[33]

Metal-organic framework xe@1bar adsorbate probability distributions, August 2025

Jake Burner. Metal-organic framework xe@1bar adsorbate probability distributions, August 2025. URLhttps: //doi.org/10.5281/zenodo.16801181

work page doi:10.5281/zenodo.16801181 2025
[34]

A database of porous rigid amorphous materials.Chemistry of Materials, 32(18):8020–8033, 2020

Raghuram Thyagarajan and David S Sholl. A database of porous rigid amorphous materials.Chemistry of Materials, 32(18):8020–8033, 2020

2020
[35]

Transferable potentials for phase equilibria

Marcus G Martin and J Ilja Siepmann. Transferable potentials for phase equilibria. 1. united-atom description of n-alkanes.The Journal of Physical Chemistry B, 102(14):2569–2577, 1998

1998
[36]

Perturbed-chain saft: An equation of state based on a perturbation theory for chain molecules.Industrial & engineering chemistry research, 40(4):1244–1260, 2001

Joachim Gross and Gabriele Sadowski. Perturbed-chain saft: An equation of state based on a perturbation theory for chain molecules.Industrial & engineering chemistry research, 40(4):1244–1260, 2001

2001
[37]

Rolf Stierle, Gernot Bauer, Nadine Thiele, Benjamin Bursik, Philipp Rehner, and Joachim Gross. Classical density functional theory in three dimensions with gpu-accelerated automatic differentiation: Computational performance analysis using the example of adsorption in covalent-organic frameworks.Chemical Engineering Science, 298: 120380, 2024

2024
[38]

Uff, a full periodic table force field for molecular mechanics and molecular dynamics simulations.Journal of the American chemical society, 114(25):10024–10035, 1992

Anthony K Rappé, Carla J Casewit, Kent S Colwell, William A Goddard III, and W Mason Skiff. Uff, a full periodic table force field for molecular mechanics and molecular dynamics simulations.Journal of the American chemical society, 114(25):10024–10035, 1992

1992
[39]

Decoupled Weight Decay Regularization

Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization.arXiv preprint arXiv:1711.05101, 2017

work page internal anchor Pith review Pith/arXiv arXiv 2017
[40]

Lora: Low-rank adaptation of large language models.Iclr, 1(2):3, 2022

Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Liang Wang, Weizhu Chen, et al. Lora: Low-rank adaptation of large language models.Iclr, 1(2):3, 2022

2022
[41]

An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale

Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, et al. An image is worth 16x16 words: Transformers for image recognition at scale.arXiv preprint arXiv:2010.11929, 2020

work page internal anchor Pith review Pith/arXiv arXiv 2010
[42]

FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning

Tri Dao. Flashattention-2: Faster attention with better parallelism and work partitioning.arXiv preprint arXiv:2307.08691, 2023

work page internal anchor Pith review Pith/arXiv arXiv 2023
[43]

GLU Variants Improve Transformer

Noam Shazeer. Glu variants improve transformer.arXiv preprint arXiv:2002.05202, 2020

work page internal anchor Pith review Pith/arXiv arXiv 2002
[44]

Scalable diffusion models with transformers

William Peebles and Saining Xie. Scalable diffusion models with transformers. InProceedings of the IEEE/CVF international conference on computer vision, pages 4195–4205, 2023

2023
[45]

Pcp-saft parameters of pure substances using large experimental databases.Industrial & Engineering Chemistry Research, 62(37):15300–15310, 2023

Timm Esper, Gernot Bauer, Philipp Rehner, and Joachim Gross. Pcp-saft parameters of pure substances using large experimental databases.Industrial & Engineering Chemistry Research, 62(37):15300–15310, 2023

2023
[46]

Pytorch: An imperative style, high-performance deep learning library.Advances in neural information processing systems, 32, 2019

Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, et al. Pytorch: An imperative style, high-performance deep learning library.Advances in neural information processing systems, 32, 2019

2019
[47]

Feos: an open-source framework for equations of state and classical density functional theory.Industrial & Engineering Chemistry Research, 62(12):5347–5357, 2023

Philipp Rehner, Gernot Bauer, and Joachim Gross. Feos: an open-source framework for equations of state and classical density functional theory.Industrial & Engineering Chemistry Research, 62(12):5347–5357, 2023

2023
[48]

The atomic simulation environment—a python library for working with atoms.Journal of Physics: Condensed Matter, 29(27):273002, 2017

Ask Hjorth Larsen, Jens Jørgen Mortensen, Jakob Blomqvist, Ivano E Castelli, Rune Christensen, Marcin Dułak, Jesper Friis, Michael N Groves, Bjørk Hammer, Cory Hargus, et al. The atomic simulation environment—a python library for working with atoms.Journal of Physics: Condensed Matter, 29(27):273002, 2017. 13 A Additional Background on Adsorption Simulati...

2017

[1] [1]

Large-scale screening of hypothetical metal–organic frameworks.Nature chemistry, 4(2):83–89, 2012

Christopher E Wilmer, Michael Leaf, Chang Yeon Lee, Omar K Farha, Brad G Hauser, Joseph T Hupp, and Randall Q Snurr. Large-scale screening of hypothetical metal–organic frameworks.Nature chemistry, 4(2):83–89, 2012

2012

[2] [2]

Evaluating metal–organic frameworks for natural gas storage

Jarad A Mason, Mike Veenstra, and Jeffrey R Long. Evaluating metal–organic frameworks for natural gas storage. Chemical Science, 5(1):32–51, 2014

2014

[3] [3]

In silico screening of carbon-capture materials.Nature materials, 11(7):633–641, 2012

Li-Chiang Lin, Adam H Berger, Richard L Martin, Jihan Kim, Joseph A Swisher, Kuldeep Jariwala, Chris H Rycroft, Abhoyjit S Bhown, Michael W Deem, Maciej Haranczyk, et al. In silico screening of carbon-capture materials.Nature materials, 11(7):633–641, 2012

2012

[4] [4]

A holistic platform for accelerating sorbent-based carbon capture.Nature, 632(8023):89–94, 2024

Charithea Charalambous, Elias Moubarak, Johannes Schilling, Eva Sanchez Fernandez, Jin-Yu Wang, Laura Herraiz, Fergus Mcilwaine, Shing Bo Peh, Matthew Garvin, Kevin Maik Jablonka, et al. A holistic platform for accelerating sorbent-based carbon capture.Nature, 632(8023):89–94, 2024

2024

[5] [5]

Grand canonical ensemble monte carlo for a lennard-jones fluid.Molecular Physics, 29(1):307–311, 1975

DJ Adams. Grand canonical ensemble monte carlo for a lennard-jones fluid.Molecular Physics, 29(1):307–311, 1975

1975

[6] [6]

Raspa: molecular simulation software for adsorption and diffusion in flexible nanoporous materials.Molecular Simulation, 42(2):81–101, 2016

David Dubbeldam, Sofía Calero, Donald E Ellis, and Randall Q Snurr. Raspa: molecular simulation software for adsorption and diffusion in flexible nanoporous materials.Molecular Simulation, 42(2):81–101, 2016

2016

[7] [7]

Raspa3: A monte carlo code for computing adsorption and diffusion in nanoporous materials and thermodynamics properties of fluids.The Journal of Chemical Physics, 161(11), 2024

YA Ran, S Sharma, SRG Balestra, Z Li, S Calero, TJH Vlugt, RQ Snurr, and D Dubbeldam. Raspa3: A monte carlo code for computing adsorption and diffusion in nanoporous materials and thermodynamics properties of fluids.The Journal of Chemical Physics, 161(11), 2024

2024

[8] [8]

The nature of the liquid-vapour interface and other topics in the statistical mechanics of non-uniform, classical fluids.Advances in physics, 28(2):143–200, 1979

Robert Evans. The nature of the liquid-vapour interface and other topics in the statistical mechanics of non-uniform, classical fluids.Advances in physics, 28(2):143–200, 1979. 11

1979

[9] [9]

Fundamental measure theory for hard-sphere mixtures: a review.Journal of Physics: Condensed Matter, 22(6):063102, 2010

Roland Roth. Fundamental measure theory for hard-sphere mixtures: a review.Journal of Physics: Condensed Matter, 22(6):063102, 2010

2010

[10] [10]

Classical density functional theory as a fast and accurate method for adsorption property prediction of porous materials.AIChE Journal, 71(6):e18779, 2025

Vincent Dufour-Décieux, Philipp Rehner, Johannes Schilling, Elias Moubarak, Joachim Gross, and André Bardow. Classical density functional theory as a fast and accurate method for adsorption property prediction of porous materials.AIChE Journal, 71(6):e18779, 2025

2025

[11] [11]

Efficient prediction of multicomponent adsorption isotherms and enthalpies of adsorption in mofs using classical density functional theory.The Journal of Physical Chemistry B, 2026

Nadine Thiele, Tiong Wei Teh, Benjamin Bursik, Marcel Granderath, Gernot Bauer, Vincent Dufour-Décieux, Philipp Rehner, Rolf Stierle, André Bardow, Niels Hansen, et al. Efficient prediction of multicomponent adsorption isotherms and enthalpies of adsorption in mofs using classical density functional theory.The Journal of Physical Chemistry B, 2026

2026

[12] [12]

Gas adsorption meets deep learning: voxelizing the potential energy surface of metal-organic frameworks.Scientific Reports, 14(1):2242, 2024

Antonios P Sarikas, Konstantinos Gkagkas, and George E Froudakis. Gas adsorption meets deep learning: voxelizing the potential energy surface of metal-organic frameworks.Scientific Reports, 14(1):2242, 2024

2024

[13] [13]

Interpretable graph transformer network for predicting adsorption isotherms of metal–organic frameworks.Journal of Chemical Information and Modeling, 62(22): 5446–5456, 2022

Pin Chen, Rui Jiao, Jinyu Liu, Yang Liu, and Yutong Lu. Interpretable graph transformer network for predicting adsorption isotherms of metal–organic frameworks.Journal of Chemical Information and Modeling, 62(22): 5446–5456, 2022

2022

[14] [14]

Understanding and predicting the spatially resolved adsorption properties of nanoporous materials.Journal of Chemical Theory and Computation, 20(12):5259–5275, 2024

Yangzesheng Sun and J Ilja Siepmann. Understanding and predicting the spatially resolved adsorption properties of nanoporous materials.Journal of Chemical Theory and Computation, 20(12):5259–5275, 2024

2024

[15] [15]

Rapid prediction of single-site adsorbate probability distributions in metal–organic frameworks using graph neural networks.Digital Discovery, 2026

Jake Burner, Olivier Marchand, Rosa Cicciarella, Marco Gibaldi, and Tom K Woo. Rapid prediction of single-site adsorbate probability distributions in metal–organic frameworks using graph neural networks.Digital Discovery, 2026

2026

[16] [16]

High-throughput computational screening of metal–organic frameworks

Yamil J Colón and Randall Q Snurr. High-throughput computational screening of metal–organic frameworks. Chemical Society Reviews, 43(16):5735–5749, 2014

2014

[17] [17]

The materials genome in action: identifying the performance limits for methane storage.Energy & Environmental Science, 8(4):1190–1199, 2015

Cory M Simon, Jihan Kim, Diego A Gomez-Gualdron, Jeffrey S Camp, Yongchul G Chung, Richard L Martin, Rocio Mercado, Michael W Deem, Dan Gunter, Maciej Haranczyk, et al. The materials genome in action: identifying the performance limits for methane storage.Energy & Environmental Science, 8(4):1190–1199, 2015

2015

[18] [18]

Elmar Sauer and Joachim Gross. Classical density functional theory for liquid–fluid interfaces and confined systems: A functional for the perturbed-chain polar statistical associating fluid theory equation of state.Industrial & Engineering Chemistry Research, 56(14):4119–4135, 2017

2017

[19] [19]

Computational screening of trillions of metal–organic frameworks for high-performance methane storage.ACS Applied Materials & Interfaces, 13(20):23647–23654, 2021

Sangwon Lee, Baekjun Kim, Hyun Cho, Hooseung Lee, Sarah Yunmi Lee, Eun Seon Cho, and Jihan Kim. Computational screening of trillions of metal–organic frameworks for high-performance methane storage.ACS Applied Materials & Interfaces, 13(20):23647–23654, 2021

2021

[20] [20]

Kaihang Shi, Zhao Li, Dylan M Anstine, Dai Tang, Coray M Colina, David S Sholl, J Ilja Siepmann, and Randall Q Snurr. Two-dimensional energy histograms as features for machine learning to predict adsorption in diverse nanoporous materials.Journal of chemical theory and computation, 19(14):4568–4583, 2023

2023

[21] [21]

Unified physio-thermodynamic descriptors via learned co2 adsorption properties in metal-organic frameworks.npj Computational Materials, 11(1):225, 2025

Emily Lin, Yang Zhong, Gang Chen, and Sili Deng. Unified physio-thermodynamic descriptors via learned co2 adsorption properties in metal-organic frameworks.npj Computational Materials, 11(1):225, 2025

2025

[22] [22]

Direct prediction of gas adsorption via spatial atom interaction learning.Nature Communications, 14 (1):7043, 2023

Jiyu Cui, Fang Wu, Wen Zhang, Lifeng Yang, Jianbo Hu, Yin Fang, Peng Ye, Qiang Zhang, Xian Suo, Yiming Mo, et al. Direct prediction of gas adsorption via spatial atom interaction learning.Nature Communications, 14 (1):7043, 2023

2023

[23] [23]

A multi-modal pre-training transformer for universal transfer learning in metal–organic frameworks.Nature Machine Intelligence, 5(3):309–318, 2023

Yeonghun Kang, Hyunsoo Park, Berend Smit, and Jihan Kim. A multi-modal pre-training transformer for universal transfer learning in metal–organic frameworks.Nature Machine Intelligence, 5(3):309–318, 2023

2023

[24] [24]

A comprehensive transformer-based approach for high-accuracy gas adsorption predictions in metal-organic frameworks.Nature Communications, 15(1):1904, 2024

Jingqi Wang, Jiapeng Liu, Hongshuai Wang, Musen Zhou, Guolin Ke, Linfeng Zhang, Jianzhong Wu, Zhifeng Gao, and Diannan Lu. A comprehensive transformer-based approach for high-accuracy gas adsorption predictions in metal-organic frameworks.Nature Communications, 15(1):1904, 2024

1904

[25] [25]

Deepdft: Neural message passing network for accurate charge density prediction

Peter Bjørn Jørgensen and Arghya Bhowmik. Deepdft: Neural message passing network for accurate charge density prediction. InMachine Learning for Molecules Workshop@ NeurIPS 2020, 2020

2020

[26] [26]

Equivariant graph neural networks for fast electron density estimation of molecules, liquids, and solids.npj Computational Materials, 8(1):183, 2022

Peter Bjørn Jørgensen and Arghya Bhowmik. Equivariant graph neural networks for fast electron density estimation of molecules, liquids, and solids.npj Computational Materials, 8(1):183, 2022

2022

[27] [27]

Swin transformer: Hierarchical vision transformer using shifted windows

Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo. Swin transformer: Hierarchical vision transformer using shifted windows. InProceedings of the IEEE/CVF international conference on computer vision, pages 10012–10022, 2021. 12

2021

[28] [28]

Machine learning the quantum-chemical properties of metal–organic frameworks for accelerated materials discovery.Matter, 4(5):1578–1597, 2021

Andrew S Rosen, Shaelyn M Iyer, Debmalya Ray, Zhenpeng Yao, Alan Aspuru-Guzik, Laura Gagliardi, Justin M Notestein, and Randall Q Snurr. Machine learning the quantum-chemical properties of metal–organic frameworks for accelerated materials discovery.Matter, 4(5):1578–1597, 2021

2021

[29] [29]

Synthetic data enable experiments in atomistic machine learning.Digital Discovery, 2(3):651–662, 2023

John LA Gardner, Zoé Faure Beaulieu, and Volker L Deringer. Synthetic data enable experiments in atomistic machine learning.Digital Discovery, 2(3):651–662, 2023

2023

[30] [30]

Boyd, Stephen Maley, Marco Gibaldi, Scott Simrod, and Tom K

Jake Burner, Jun Luo, Andrew White, Adam Mirmiran, Ohmin Kwon, Peter G. Boyd, Stephen Maley, Marco Gibaldi, Scott Simrod, and Tom K. Woo. ab initio repeat charge mof database (arc-mof), July 2022. URL https://doi.org/10.5281/zenodo.6908728

work page doi:10.5281/zenodo.6908728 2022

[31] [31]

Metal-organic framework ch4@1bar adsorbate probability distributions, August 2025

Jake Burner. Metal-organic framework ch4@1bar adsorbate probability distributions, August 2025. URL https://doi.org/10.5281/zenodo.16800893

work page doi:10.5281/zenodo.16800893 2025

[32] [32]

Metal-organic framework ch4@65bar adsorbate probability distributions, August 2025

Jake Burner. Metal-organic framework ch4@65bar adsorbate probability distributions, August 2025. URL https://doi.org/10.5281/zenodo.16801034

work page doi:10.5281/zenodo.16801034 2025

[33] [33]

Metal-organic framework xe@1bar adsorbate probability distributions, August 2025

Jake Burner. Metal-organic framework xe@1bar adsorbate probability distributions, August 2025. URLhttps: //doi.org/10.5281/zenodo.16801181

work page doi:10.5281/zenodo.16801181 2025

[34] [34]

A database of porous rigid amorphous materials.Chemistry of Materials, 32(18):8020–8033, 2020

Raghuram Thyagarajan and David S Sholl. A database of porous rigid amorphous materials.Chemistry of Materials, 32(18):8020–8033, 2020

2020

[35] [35]

Transferable potentials for phase equilibria

Marcus G Martin and J Ilja Siepmann. Transferable potentials for phase equilibria. 1. united-atom description of n-alkanes.The Journal of Physical Chemistry B, 102(14):2569–2577, 1998

1998

[36] [36]

Perturbed-chain saft: An equation of state based on a perturbation theory for chain molecules.Industrial & engineering chemistry research, 40(4):1244–1260, 2001

Joachim Gross and Gabriele Sadowski. Perturbed-chain saft: An equation of state based on a perturbation theory for chain molecules.Industrial & engineering chemistry research, 40(4):1244–1260, 2001

2001

[37] [37]

Rolf Stierle, Gernot Bauer, Nadine Thiele, Benjamin Bursik, Philipp Rehner, and Joachim Gross. Classical density functional theory in three dimensions with gpu-accelerated automatic differentiation: Computational performance analysis using the example of adsorption in covalent-organic frameworks.Chemical Engineering Science, 298: 120380, 2024

2024

[38] [38]

Uff, a full periodic table force field for molecular mechanics and molecular dynamics simulations.Journal of the American chemical society, 114(25):10024–10035, 1992

Anthony K Rappé, Carla J Casewit, Kent S Colwell, William A Goddard III, and W Mason Skiff. Uff, a full periodic table force field for molecular mechanics and molecular dynamics simulations.Journal of the American chemical society, 114(25):10024–10035, 1992

1992

[39] [39]

Decoupled Weight Decay Regularization

Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization.arXiv preprint arXiv:1711.05101, 2017

work page internal anchor Pith review Pith/arXiv arXiv 2017

[40] [40]

Lora: Low-rank adaptation of large language models.Iclr, 1(2):3, 2022

Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Liang Wang, Weizhu Chen, et al. Lora: Low-rank adaptation of large language models.Iclr, 1(2):3, 2022

2022

[41] [41]

An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale

Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, et al. An image is worth 16x16 words: Transformers for image recognition at scale.arXiv preprint arXiv:2010.11929, 2020

work page internal anchor Pith review Pith/arXiv arXiv 2010

[42] [42]

FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning

Tri Dao. Flashattention-2: Faster attention with better parallelism and work partitioning.arXiv preprint arXiv:2307.08691, 2023

work page internal anchor Pith review Pith/arXiv arXiv 2023

[43] [43]

GLU Variants Improve Transformer

Noam Shazeer. Glu variants improve transformer.arXiv preprint arXiv:2002.05202, 2020

work page internal anchor Pith review Pith/arXiv arXiv 2002

[44] [44]

Scalable diffusion models with transformers

William Peebles and Saining Xie. Scalable diffusion models with transformers. InProceedings of the IEEE/CVF international conference on computer vision, pages 4195–4205, 2023

2023

[45] [45]

Pcp-saft parameters of pure substances using large experimental databases.Industrial & Engineering Chemistry Research, 62(37):15300–15310, 2023

Timm Esper, Gernot Bauer, Philipp Rehner, and Joachim Gross. Pcp-saft parameters of pure substances using large experimental databases.Industrial & Engineering Chemistry Research, 62(37):15300–15310, 2023

2023

[46] [46]

Pytorch: An imperative style, high-performance deep learning library.Advances in neural information processing systems, 32, 2019

Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, et al. Pytorch: An imperative style, high-performance deep learning library.Advances in neural information processing systems, 32, 2019

2019

[47] [47]

Feos: an open-source framework for equations of state and classical density functional theory.Industrial & Engineering Chemistry Research, 62(12):5347–5357, 2023

Philipp Rehner, Gernot Bauer, and Joachim Gross. Feos: an open-source framework for equations of state and classical density functional theory.Industrial & Engineering Chemistry Research, 62(12):5347–5357, 2023

2023

[48] [48]

The atomic simulation environment—a python library for working with atoms.Journal of Physics: Condensed Matter, 29(27):273002, 2017

Ask Hjorth Larsen, Jens Jørgen Mortensen, Jakob Blomqvist, Ivano E Castelli, Rune Christensen, Marcin Dułak, Jesper Friis, Michael N Groves, Bjørk Hammer, Cory Hargus, et al. The atomic simulation environment—a python library for working with atoms.Journal of Physics: Condensed Matter, 29(27):273002, 2017. 13 A Additional Background on Adsorption Simulati...

2017