Periodic Reporting for period 2 - FIAMMA (Fully Integrating Atomistic Modeling with Machine Learning)
Reporting period: 2023-07-01 to 2024-12-31
FIAMMA aims to leverage the synergy between physics-based modeling of matter and state-ot-the-art machine-learning techniques to achieve predictive accuracy when modeling complex materials under realistic conditions. By intimately combining deductive and inductive paradigms of modeling it strives to get the best of both worlds - the transferability and interpretability of bottom-up simulations, and the speed and accuracy of data-driven methods. Expanding the reach and predictive power of computational materials discovery will accelerate the search for technological solutions to several pressing societal challenges, from green chemistry to sustainable energy production and storage, exploiting the ability of simulations to perform high-throughput searches of potential materials, feeding the R&D pipeline with promising candidates for applications.
The primary objective of the project, which has a clear methodological focus, is to create an integrated framework that combines machine learning (ML) with quantum mechanical (QM) and statistical mechanical (SM) methods. This framework aims to enhance the predictive power and applicability of atomic-scale simulations. By achieving this integration, the project seeks to:
* Improve the mathematical frameworks used to describe atomic structures, addressing fundamental shortcomings that hinder predictive accuracy, in particular developing long-range models that incorporate non-local physics, going beyond current models that are either limited to short-range interactions, or to the inclusion of traditional electrostatic descriptions. (Work package ML1)
* Integrate ML at the core of quantum mechanical calculations, by predicting intermediate ingredients of an electronic-structure calculation. (Work package ML2)
* Advance uncertainty quantification (UQ) techniques to enhance the reliability of predictions, and to assess the impact of ML approximations on the end result of complicated statistical-mechanical simulations. (Work package ML3)
In order to make these methodological advances widely available, one of the milestones of the project involves implementing them within a modular, open-source software stack that can be interfaced seamlessly with established modeling tools. (Work package SW)
Even if the core objectives are methodological, FIAMMA also includes applications to the materials science of real materials, chosen to serve as stringent tests of the accuracy and efficiency of the methodology, while demonstrating the impact on relevant technological challenges. (Work package APP)
The primary objective of the project, which has a clear methodological focus, is to create an integrated framework that combines machine learning (ML) with quantum mechanical (QM) and statistical mechanical (SM) methods. This framework aims to enhance the predictive power and applicability of atomic-scale simulations. By achieving this integration, the project seeks to:
* Improve the mathematical frameworks used to describe atomic structures, addressing fundamental shortcomings that hinder predictive accuracy, in particular developing long-range models that incorporate non-local physics, going beyond current models that are either limited to short-range interactions, or to the inclusion of traditional electrostatic descriptions. (Work package ML1)
* Integrate ML at the core of quantum mechanical calculations, by predicting intermediate ingredients of an electronic-structure calculation. (Work package ML2)
* Advance uncertainty quantification (UQ) techniques to enhance the reliability of predictions, and to assess the impact of ML approximations on the end result of complicated statistical-mechanical simulations. (Work package ML3)
In order to make these methodological advances widely available, one of the milestones of the project involves implementing them within a modular, open-source software stack that can be interfaced seamlessly with established modeling tools. (Work package SW)
Even if the core objectives are methodological, FIAMMA also includes applications to the materials science of real materials, chosen to serve as stringent tests of the accuracy and efficiency of the methodology, while demonstrating the impact on relevant technological challenges. (Work package APP)
The project has already achieved significant milestones, including addressing fundamental mathematical issues in atomic structure descriptions, developing robust long-range models, and achieving substantial progress in integrating ML with QM and SM methods (ML1). We already started releasing some of the planned software tools, making these advancements accessible to the broader scientific community (SW).
Simulations of transport in solid-state electrolyte materials and of the thermodynamics and dielectric behavior of prototypical ferroelectrics provide early proof of the relevance of this line of research to develop technological solutions that are of great relevance for contemporary societal issues (APP).
One result worth highlighting is the derivation of a method to achieve a complete local description of an atomic structure based on finite-correlation-order descriptors.
Soon after submitting the proposal for this project, we realized that one of the basic assumptions we made, the widely shared belief that the *local* descriptors used in the field were a solved problem, did not hold. We found a few arrangements of atoms that could not be distinguished by those descriptors - meaning that the ML models built on them were not able, even in principle, to provide a universal approximation of the quantum mechanical structure-property relations. Fully understanding -- and resolving -- this problem was one of the early achievements in FIAMMA, which allowed us to stay on track with the original plans.
With this potential roadblock removed, we could make excellent progress on the three methodological work packages: We've developed models that accurately describe different types long-range interactions (ML1), and integrated machine learning Hamiltonians into quantum mechanical calculations (ML2). Our work the integration of ML models with sophisticated statistical mechanical calculations has also yielded promising early results (ML3), which allowed us to secure additional funding to go beyond what initially envisaged to tackle the issue of assessing the level of trust that can be put in ML-driven predictions in a reliable and inexpensive way.
On the software-development front, we have completely overhauled our ML software infrastructure, making it more modular and efficient. Most of the core elements are still in heavy-development, but already openly accessible. We have already released some production libraries, e.g. Sphericart for efficient spherical harmonics computation and Scikit-Matter for domain-agnostic machine learning methods. We have also finalized a new release of the i-PI advanced simulation code, focusing on ML model integration (SW, ML3).
In terms of applications (APP), we've slightly deviated from our original plans in terms of the target systems, but have already made impactful discoveries.
Our focus shifted to ferroelectrics and lithium thiophosphate electrolytes, materials that allowed us to explore long-range physics and electronic properties while establishing valuable collaborations with experimentalists and industry partners. Although we intend to return to our initial focus on silicate frameworks and molecular materials, our current applications have effectively demonstrated the practical implications of our methodological advancements.
Simulations of transport in solid-state electrolyte materials and of the thermodynamics and dielectric behavior of prototypical ferroelectrics provide early proof of the relevance of this line of research to develop technological solutions that are of great relevance for contemporary societal issues (APP).
One result worth highlighting is the derivation of a method to achieve a complete local description of an atomic structure based on finite-correlation-order descriptors.
Soon after submitting the proposal for this project, we realized that one of the basic assumptions we made, the widely shared belief that the *local* descriptors used in the field were a solved problem, did not hold. We found a few arrangements of atoms that could not be distinguished by those descriptors - meaning that the ML models built on them were not able, even in principle, to provide a universal approximation of the quantum mechanical structure-property relations. Fully understanding -- and resolving -- this problem was one of the early achievements in FIAMMA, which allowed us to stay on track with the original plans.
With this potential roadblock removed, we could make excellent progress on the three methodological work packages: We've developed models that accurately describe different types long-range interactions (ML1), and integrated machine learning Hamiltonians into quantum mechanical calculations (ML2). Our work the integration of ML models with sophisticated statistical mechanical calculations has also yielded promising early results (ML3), which allowed us to secure additional funding to go beyond what initially envisaged to tackle the issue of assessing the level of trust that can be put in ML-driven predictions in a reliable and inexpensive way.
On the software-development front, we have completely overhauled our ML software infrastructure, making it more modular and efficient. Most of the core elements are still in heavy-development, but already openly accessible. We have already released some production libraries, e.g. Sphericart for efficient spherical harmonics computation and Scikit-Matter for domain-agnostic machine learning methods. We have also finalized a new release of the i-PI advanced simulation code, focusing on ML model integration (SW, ML3).
In terms of applications (APP), we've slightly deviated from our original plans in terms of the target systems, but have already made impactful discoveries.
Our focus shifted to ferroelectrics and lithium thiophosphate electrolytes, materials that allowed us to explore long-range physics and electronic properties while establishing valuable collaborations with experimentalists and industry partners. Although we intend to return to our initial focus on silicate frameworks and molecular materials, our current applications have effectively demonstrated the practical implications of our methodological advancements.
Since the beginning of the project, FIAMMA-funded research has resolved a major issue with the most established form of atom-centered descriptors, provided a rigorous framework to represent bond-centered properties (that are needed to built ML models of the electronic structure), developed uncertainty-quantification schemes that apply to complicated statistical-mechanical frameworks, proposed general and flexible descriptors of long-range interactions and developed proof-of-principle software implementations that are modular, efficient and openly accessible.
These methodological and software advances are already supporting applicative breakthroughs in the group, and are increasingly used by others. We are well on track to deliver the fully-integrated conceptual and computational framework that is the core aim of FIAMMA, and expect to be able to demonstrate with compelling examples the enhancement in predictive power of atomic-scale simulations, and its impact on the solution of concrete technological problems.
These methodological and software advances are already supporting applicative breakthroughs in the group, and are increasingly used by others. We are well on track to deliver the fully-integrated conceptual and computational framework that is the core aim of FIAMMA, and expect to be able to demonstrate with compelling examples the enhancement in predictive power of atomic-scale simulations, and its impact on the solution of concrete technological problems.