A strategic roadmap for an atomistic machine-learning ecosystem
Abstract
Data-driven machine learning (ML) techniques have become an essential tool in many domains of science. Their application to atomistic simulations of matter is particularly widespread and impactful. This success is due largely to the existence of a well-developed and established physics-based modeling framework, ranging from first-principles electronic-structure calculations to molecular dynamics and statistical sampling, into which ML was integrated naturally to reshape long-standing trade-offs between accuracy, efficiency, and scale. Nevertheless, this integration raises both conceptual and practical challenges, from choosing between data-centric and physics-based modeling approaches to adapting established software stacks to modern hardware accelerators and ML libraries. As the field evolves rapidly, fueled in part by widespread enthusiasm but also by tangible impact, it seems appropriate to take a moment to consider the current state of the art and open challenges, and reflect on what can be done to better coordinate efforts across the community. With this goal in mind, several members of this community met in Lausanne in January 2026 at CECAM to discuss algorithms, models, software and hardware infrastructure, and the most promising scientific applications that have become possible thanks to the use of artificial intelligence in atomic-scale simulations. This strategic roadmap paper summarizes the outcomes of these discussions, suggesting some long-term goals, and some concrete actions, to establish a healthy, sustainable and impactful atomistic ML ecosystem.
Related
- A Lightweight Universal Machine-Learning Interatomic Potential via Knowledge Distillation for Scalable Atomistic Simulations
- Overcoming the limitations of body-ordered potentials for atomistic machine learning
- From sparse quantum-computing data to atomistic simulation with universal machine-learning interatomic potentials
- Enabling Domain-Specific Atomistic Models: A Machine Learning Potential for the Solid Acid Family
- RACE: A Modular and Interpretable Radial–Angular Correlation Descriptor for Atomistic Machine Learning
BibTeX
@misc{behler2026strategic,
title = {{A strategic roadmap for an atomistic machine-learning ecosystem}},
author = {Jörg Behler and Michele Ceriotti and Cecilia Clementi and Gábor Csányi and Alin-Marin Elena and Aditi Krishnapriyan and Joseph W. Abbott and Fabio Affinito and Albert P. Bartók and Ilyes Batatia and Filippo Bigi and Florian N. Brünig and Yannick Calvino Alonso and Giuseppe Carleo and Aurélie Champagne and Stefan Chmiela and Marc L. Descoteaux and Ralf Drautz and Alexandra Farcas and Meng Gao and Rohit Goswami and Michael F. Herbst and Christian Holm and James R. Kermode and Alexander L. M. Knoll and Tobias Kreiman and Marcel F. Langer and Hoang-Thien Luu and Yury Lysogorskiy and Mihai-Cosmin Marinica and Rocco Meli and Klaus-Robert Müller and Frank Noé and Mohamadhosein Nosratjoo and Simon Olsson and Christoph Ortner and Aldo S. Pasos-Trejo and Anyang Peng and Eric Qu and Andrea Rizzi and Mariana Rossi and Bassem Sboui and Gregor N. C. Simm and Alexandre Tkatchenko and Jacopo Venturin and O. Anatole von Lilienfeld and William C. Witt and Brandon M. Wood and Tigany Zarrouk and Fabian Zills},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2609.39090},
eprint = {2609.39090},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.39090},
}
RIS
TY - UNPB
TI - A strategic roadmap for an atomistic machine-learning ecosystem
AU - Jörg Behler
AU - Michele Ceriotti
AU - Cecilia Clementi
AU - Gábor Csányi
AU - Alin-Marin Elena
AU - Aditi Krishnapriyan
AU - Joseph W. Abbott
AU - Fabio Affinito
AU - Albert P. Bartók
AU - Ilyes Batatia
AU - Filippo Bigi
AU - Florian N. Brünig
AU - Yannick Calvino Alonso
AU - Giuseppe Carleo
AU - Aurélie Champagne
AU - Stefan Chmiela
AU - Marc L. Descoteaux
AU - Ralf Drautz
AU - Alexandra Farcas
AU - Meng Gao
AU - Rohit Goswami
AU - Michael F. Herbst
AU - Christian Holm
AU - James R. Kermode
AU - Alexander L. M. Knoll
AU - Tobias Kreiman
AU - Marcel F. Langer
AU - Hoang-Thien Luu
AU - Yury Lysogorskiy
AU - Mihai-Cosmin Marinica
AU - Rocco Meli
AU - Klaus-Robert Müller
AU - Frank Noé
AU - Mohamadhosein Nosratjoo
AU - Simon Olsson
AU - Christoph Ortner
AU - Aldo S. Pasos-Trejo
AU - Anyang Peng
AU - Eric Qu
AU - Andrea Rizzi
AU - Mariana Rossi
AU - Bassem Sboui
AU - Gregor N. C. Simm
AU - Alexandre Tkatchenko
AU - Jacopo Venturin
AU - O. Anatole von Lilienfeld
AU - William C. Witt
AU - Brandon M. Wood
AU - Tigany Zarrouk
AU - Fabian Zills
PY - 2026
DA - 2026/10/13
PB - arXiv
DO - 10.48550/arxiv.2609.39090
UR - https://arxiv.org/abs/2609.39090
AB - Data-driven machine learning (ML) techniques have become an essential tool in many domains of science. Their application to atomistic simulations of matter is particularly widespread and impactful. This success is due largely to the existence of a well-developed and established physics-based modeling framework, ranging from first-principles electronic-structure calculations to molecular dynamics and statistical sampling, into which ML was integrated naturally to reshape long-standing trade-offs between accuracy, efficiency, and scale. Nevertheless, this integration raises both conceptual and practical challenges, from choosing between data-centric and physics-based modeling approaches to adapting established software stacks to modern hardware accelerators and ML libraries. As the field evolves rapidly, fueled in part by widespread enthusiasm but also by tangible impact, it seems appropriate to take a moment to consider the current state of the art and open challenges, and reflect on what can be done to better coordinate efforts across the community. With this goal in mind, several members of this community met in Lausanne in January 2026 at CECAM to discuss algorithms, models, software and hardware infrastructure, and the most promising scientific applications that have become possible thanks to the use of artificial intelligence in atomic-scale simulations. This strategic roadmap paper summarizes the outcomes of these discussions, suggesting some long-term goals, and some concrete actions, to establish a healthy, sustainable and impactful atomistic ML ecosystem.
ER -