The Kitchin Research Group

Machine learning has revolutionized materials discovery, but ensuring reliable predictions remains a challenge. The paper “Improved Uncertainty Estimation of Graph Neural Network Potentials Using Engineered Latent Space Distances” tackles this issue by introducing advanced uncertainty quantification (UQ) techniques for Graph Neural Networks (GNNs) used in catalyst discovery. Traditional models struggle with predicting uncertainty in relaxed energy calculations, a key factor in identifying promising catalytic materials. This work proposes a novel latent space distance approach, significantly improving calibration and reliability compared to standard ensemble methods. By leveraging distribution-free techniques and engineered latent representations, the study enables more accurate and interpretable confidence estimates, ensuring that machine learning-driven catalyst discovery is not just faster but also more trustworthy.

Musielewicz, Joseph; Lan, Janice; Uyttendaele, Matt; Kitchin, John, Improved Uncertainty Estimation of Graph Neural Network Potentials Using Engineered Latent Space Distance, J. Physical Chemistry C (2024). https://doi.org/10.1021/acs.jpcc.4c04972.

@article{musielewicz-2024-improv-uncer,
  author =       {Joseph Musielewicz and Janice Lan and Matt Uyttendaele and
                  John R. Kitchin},
  title =        {Improved Uncertainty Estimation of Graph Neural Network
                  Potentials Using Engineered Latent Space Distances},
  journal =      {The Journal of Physical Chemistry C},
  volume =       128,
  number =       49,
  pages =        {20799-20810},
  year =         2024,
  doi =          {10.1021/acs.jpcc.4c04972},
  url =          {http://dx.doi.org/10.1021/acs.jpcc.4c04972},
  DATE_ADDED =   {Sat Feb 1 08:18:45 2025},
}

Copyright (C) 2025 by John Kitchin. See the License for information about copying.

org-mode source

Org-mode version = 9.8-pre

The paper “AdsorbDiff: Adsorbate Placement via Conditional Denoising Diffusion” presents a novel machine learning framework for optimizing the placement of adsorbates on catalyst surfaces. Traditional methods rely on heuristics or brute-force searches to determine the lowest energy configuration, a key step in computational catalyst discovery. AdsorbDiff introduces a conditional denoising diffusion model that predicts optimal adsorbate sites and orientations while accounting for periodic boundary conditions. The model integrates machine learning force fields and Density Functional Theory (DFT) evaluations to ensure accurate energy assessments. Compared to prior approaches, AdsorbDiff achieves up to a 5x speedup and a 3.5x improvement in accuracy. The study also explores the impact of pretraining and model architectures, demonstrating robust generalization to unseen adsorbates and slabs. This advancement significantly accelerates computational screening for new catalysts, paving the way for more efficient materials discovery.

AdsorbDiff: Adsorbate Placement via Conditional Denoising Diffusion, Adeesh Kolluru, John R. Kitchin Proceedings of the 41st International Conference on Machine Learning, PMLR 235:25042-25057, 2024.

@InProceedings{pmlr-v235-kolluru24a,
  title =        {{A}dsorb{D}iff: Adsorbate Placement via Conditional Denoising Diffusion},
  author =       {Kolluru, Adeesh and Kitchin, John R.},
  booktitle =    {Proceedings of the 41st International Conference on Machine Learning},
  pages =        {25042--25057},
  year =         {2024},
  editor =       {Salakhutdinov, Ruslan and Kolter, Zico and Heller, Katherine and Weller, Adrian and Oliver, Nuria and Scarlett, Jonathan and Berkenkamp, Felix},
  volume =       {235},
  series =       {Proceedings of Machine Learning Research},
  month =        {21--27 Jul},
  publisher =    {PMLR},
  pdf =          {https://raw.githubusercontent.com/mlresearch/v235/main/assets/kolluru24a/kolluru24a.pdf},
  url =          {https://proceedings.mlr.press/v235/kolluru24a.html},
  abstract =     {Determining the optimal configuration of adsorbates on a slab (adslab) is pivotal in the exploration of novel catalysts across diverse applications. Traditionally, the quest for the lowest energy adslab configuration involves placing the adsorbate onto the slab followed by an optimization process. Prior methodologies have relied on heuristics, problem-specific intuitions, or brute-force approaches to guide adsorbate placement. In this work, we propose a novel framework for adsorbate placement using denoising diffusion. The model is designed to predict the optimal adsorbate site and orientation corresponding to the lowest energy configuration. Further, we have an end-to-end evaluation framework where diffusion-predicted adslab configuration is optimized with a pretrained machine learning force field and finally evaluated with Density Functional Theory (DFT). Our findings demonstrate an acceleration of up to 5x or 3.5x improvement in accuracy compared to the previous best approach. Given the novelty of this framework and application, we provide insights into the impact of pretraining, model architectures, and conduct extensive experiments to underscore the significance of this approach.}
}

Copyright (C) 2025 by John Kitchin. See the License for information about copying.

org-mode source

Org-mode version = 9.8-pre

The paper “Unifying Theory of Electronic Descriptors of Metal Surfaces upon Perturbation” by Huang et al. presents a novel framework for predicting electronic properties of metal surfaces using interpretable deep learning. Unlike traditional black-box machine learning models, this approach integrates physical insights to enhance interpretability without sacrificing accuracy. The study identifies a previously overlooked factor—orbital resonance in d-electron hopping—which influences the electronic structure of transition and noble metal surfaces alongside known effects like charge transfer, strain, and ligand interactions. By leveraging a physics-infused deep learning model, the authors provide a unified method for predicting electronic descriptors, offering a powerful tool for the rational design of catalytic materials and other functional surfaces in materials science. This work bridges the gap between theory and data-driven approaches, paving the way for more efficient and interpretable materials discovery.

Yang Huang, Shih-Han Wang, Mohith Kamanuru, Luke E. K. Achenie, John R. Kitchin, and Hongliang Xin, Unifying theory of electronic descriptors of metal surfaces upon perturbation, Phys. Rev. B 110, L121404 (2024). https://doi.org/10.1103/PhysRevB.110.L121404.

@article{huang-2024-unify-theor,
  author =       {Yang Huang and Shih-Han Wang and Mohith Kamanuru and Luke E.
                  K. Achenie and John R. Kitchin and Hongliang Xin},
  title =        {Unifying Theory of Electronic Descriptors of Metal Surfaces
                  Upon Perturbation},
  journal =      {Physical Review B},
  volume =       110,
  number =       12,
  pages =        {L121404},
  year =         2024,
  doi =          {10.1103/physrevb.110.l121404},
  url =          {http://dx.doi.org/10.1103/PhysRevB.110.L121404},
  DATE_ADDED =   {Wed Jan 29 19:39:29 2025},
}

Copyright (C) 2025 by John Kitchin. See the License for information about copying.

org-mode source

Org-mode version = 9.8-pre

This paper investigates how alkali metal cations influence the oxygen reduction reaction (ORR) on various metal surfaces. By combining experimental data and computational modeling, the authors identify the potential of zero total charge (PZTC) as a key predictor of cation effects. Metals with PZTCs positive of the ORR potential window show increased activity with larger cations (e.g., Li+ < Na+ < K+ < Cs+), enhancing reaction rates by reducing hydroxide poisoning or lowering activation barriers. Conversely, metals with PZTCs negative of the ORR potential repel cations, showing no effect. This work offers a mechanistic understanding of cation effects in electrocatalysis, paving the way for electrolyte design to optimize catalytic performance.

@article{bender-2024-poten-zero,
  author =       {Jay T. Bender and Rohan Yuri Sanspeur and Angel E. Valles and
                  Alyssa K. Uvodich and Delia J. Milliron and John R. Kitchin
                  and Joaquin Resasco},
  title =        {The Potential of Zero Total Charge Predicts Cation Effects for
                  the Oxygen Reduction Reaction},
  journal =      {ACS Energy Letters},
  volume =       9,
  number =       9,
  pages =        {4724-4733},
  year =         2024,
  doi =          {10.1021/acsenergylett.4c01897},
  url =          {http://dx.doi.org/10.1021/acsenergylett.4c01897},
}

Copyright (C) 2025 by John Kitchin. See the License for information about copying.

org-mode source

Org-mode version = 9.8-pre

The paper “Investigating the Error Imbalance of Large-Scale Machine Learning Potentials in Catalysis” by Kareem Abdelmaqsoud et al. explores the persistent error imbalance in machine learning potentials (MLPs) trained on the Open Catalyst 2020 (OC20) dataset. While MLPs have significantly accelerated catalyst discovery by approximating density functional theory (DFT) calculations, the study finds that non-metals exhibit disproportionately high prediction errors compared to intermetallics and metalloids. The authors investigate two primary sources of this imbalance: (1) DFT convergence errors and (2) surface reconstruction inconsistencies. They determine that while DFT convergence errors exist, they do not significantly impact MLP performance. However, surface reconstructions introduce inconsistencies in adsorption energy referencing, making them difficult for MLPs to model accurately. By removing these reconstructions from validation sets, the study achieves a 35% reduction in mean absolute error (MAE) and a more balanced error distribution. The paper suggests an alternative approach: training MLPs on total energies rather than adsorption energies, which eliminates referencing issues and maintains comparable accuracy . This research provides valuable insights for improving MLP robustness in catalysis and optimizing large-scale computational datasets.

@article{abdelmaqsoud-2024-inves-error,
  author =       {Kareem Abdelmaqsoud and Muhammed Shuaibi and Adeesh Kolluru
                  and Raffaele Cheula and John R. Kitchin},
  title =        {Investigating the Error Imbalance of Large-Scale Machine
                  Learning Potentials in Catalysis},
  journal =      {Catalysis Science & Technology},
  volume =       {14},
  number =       {20},
  pages =        {5899-5908},
  year =         2024,
  doi =          {10.1039/d4cy00615a},
  url =          {http://dx.doi.org/10.1039/d4cy00615a},
  DATE_ADDED =   {Sat Aug 17 09:45:44 2024},
}

Copyright (C) 2025 by John Kitchin. See the License for information about copying.

org-mode source

Org-mode version = 9.8-pre