-
A Recipe for Charge Density Prediction
Authors:
Xiang Fu,
Andrew Rosen,
Kyle Bystrom,
Rui Wang,
Albert Musaelian,
Boris Kozinsky,
Tess Smidt,
Tommi Jaakkola
Abstract:
In density functional theory, charge density is the core attribute of atomic systems from which all chemical properties can be derived. Machine learning methods are promising in significantly accelerating charge density prediction, yet existing approaches either lack accuracy or scalability. We propose a recipe that can achieve both. In particular, we identify three key ingredients: (1) representi…
▽ More
In density functional theory, charge density is the core attribute of atomic systems from which all chemical properties can be derived. Machine learning methods are promising in significantly accelerating charge density prediction, yet existing approaches either lack accuracy or scalability. We propose a recipe that can achieve both. In particular, we identify three key ingredients: (1) representing the charge density with atomic and virtual orbitals (spherical fields centered at atom/virtual coordinates); (2) using expressive and learnable orbital basis sets (basis function for the spherical fields); and (3) using high-capacity equivariant neural network architecture. Our method achieves state-of-the-art accuracy while being more than an order of magnitude faster than existing methods. Furthermore, our method enables flexible efficiency-accuracy trade-offs by adjusting the model/basis sizes.
△ Less
Submitted 29 May, 2024;
originally announced May 2024.
-
Addressing the Band Gap Problem with a Machine-Learned Exchange Functional
Authors:
Kyle Bystrom,
Stefano Falletta,
Boris Kozinsky
Abstract:
The systematic underestimation of band gaps is one of the most fundamental challenges in semilocal density functional theory (DFT). In addition to hindering the application of DFT to predicting electronic properties, the band gap problem is intimately related to self-interaction and delocalization errors, which make the study of charge transfer mechanisms with DFT difficult. In this work, we prese…
▽ More
The systematic underestimation of band gaps is one of the most fundamental challenges in semilocal density functional theory (DFT). In addition to hindering the application of DFT to predicting electronic properties, the band gap problem is intimately related to self-interaction and delocalization errors, which make the study of charge transfer mechanisms with DFT difficult. In this work, we present two key innovations to address the band gap problem. First, we design an approach for machine learning density functionals based on Gaussian processes to explicitly fit single-particle energy levels. Second, we introduce novel nonlocal features of the density matrix that are expressive enough to fit these single-particle levels. Combining these developments, we train a machine-learned functional for the exact exchange energy that predicts molecular energy gaps and reaction energies of a wide range of molecules in excellent agreement with reference hybrid DFT calculations. In addition, while being trained solely on molecular data, our model predicts reasonable formation energies of polarons in solids, showcasing its transferability and robustness. Our approach generalizes straightforwardly to full exchange-correlation functionals, thus paving the way to the design of novel state-of-the-art functionals for the prediction of electronic properties of molecules and materials.
△ Less
Submitted 10 April, 2024; v1 submitted 25 March, 2024;
originally announced March 2024.
-
Transferability and Accuracy of Ionic Liquid Simulations with Equivariant Machine Learning Interatomic Potentials
Authors:
Zachary A. H. Goodwin,
Malia B. Wenny,
Julia H. Yang,
Andrea Cepellotti,
Jingxuan Ding,
Kyle Bystrom,
Blake R. Duschatko,
Anders Johansson,
Lixin Sun,
Simon Batzner,
Albert Musaelian,
Jarad A. Mason,
Boris Kozinsky,
Nicola Molinari
Abstract:
Ionic liquids (ILs) are an exciting class of electrolytes finding applications in many areas from energy storage to solvents, where they have been touted as ``designer solvents'' as they can be mixed to precisely tailor the physiochemical properties. As using machine learning interatomic potentials (MLIPs) to simulate ILs is still relatively unexplored, several questions need to be answered to see…
▽ More
Ionic liquids (ILs) are an exciting class of electrolytes finding applications in many areas from energy storage to solvents, where they have been touted as ``designer solvents'' as they can be mixed to precisely tailor the physiochemical properties. As using machine learning interatomic potentials (MLIPs) to simulate ILs is still relatively unexplored, several questions need to be answered to see if MLIPs can be transformative for ILs. Since ILs are often not pure, but are either mixed together or contain additives, we first demonstrate that a MLIP can be trained to be compositionally transferable, i.e., the MLIP can be applied to mixtures of ions not directly trained on, whilst only being trained on a few mixtures of the same ions. We also investigate the accuracy of MLIPs for a novel IL, which we experimentally synthesize and characterize. Our MLIP trained on $\sim$200 DFT frames is in reasonable agreement with our experiments and DFT.
△ Less
Submitted 15 July, 2024; v1 submitted 4 March, 2024;
originally announced March 2024.
-
Nonlocal Machine-Learned Exchange Functional for Molecules and Solids
Authors:
Kyle Bystrom,
Boris Kozinsky
Abstract:
The design of better exchange-correlation functionals for Density Functional Theory (DFT) is a central challenge of modern electronic structure theory. However, current developments are limited by the mathematical form of the functional, with efficient semilocal functionals being inaccurate for many technologically important systems and the more accurate hybrid functionals being too expensive for…
▽ More
The design of better exchange-correlation functionals for Density Functional Theory (DFT) is a central challenge of modern electronic structure theory. However, current developments are limited by the mathematical form of the functional, with efficient semilocal functionals being inaccurate for many technologically important systems and the more accurate hybrid functionals being too expensive for large solid-state systems due to the use of the exact exchange operator. In this work, we use machine learning combined with exact physical constraints to design an exchange functional that is both orbital-dependent and nonlocal, but which can be evaluated at roughly the cost of semilocal functionals and is significantly faster than hybrid DFT in plane-wave codes. By training functionals with several different feature sets, we elucidate the roles of orbital-dependent and nonlocal features in learning the exchange energy and determine that both types of features provide vital and independently important information to the model. Having trained our new exchange functional with an expressive, nonlocal feature set, we substitute it into existing hybrid functionals to achieve hybrid-DFT accuracy on thermochemical benchmark sets and improve the accuracy of band gap predictions over semilocal DFT. To demonstrate the scalability of our approach as well as the practical benefits of improved band gap prediction, we compute charged defect transition levels in silicon using large supercells. Due to its transferability and computational efficiency for both molecular and extended systems, our model overcomes the cost-accuracy trade-off between semilocal and hybrid DFT, and our general approach provides a feasible path toward a universal exchange-correlation functional with post-hybrid DFT accuracy and semilocal DFT cost.
△ Less
Submitted 15 August, 2024; v1 submitted 1 March, 2023;
originally announced March 2023.
-
Complexity of Many-Body Interactions in Transition Metals via Machine-Learned Force Fields from the TM23 Data Set
Authors:
Cameron J. Owen,
Steven B. Torrisi,
Yu Xie,
Simon Batzner,
Kyle Bystrom,
Jennifer Coulter,
Albert Musaelian,
Lixin Sun,
Boris Kozinsky
Abstract:
This work examines challenges associated with the accuracy of machine-learned force fields (MLFFs) for bulk solid and liquid phases of d-block elements. In exhaustive detail, we contrast the performance of force, energy, and stress predictions across the transition metals for two leading MLFF models: a kernel-based atomic cluster expansion method implemented using sparse Gaussian processes (FLARE)…
▽ More
This work examines challenges associated with the accuracy of machine-learned force fields (MLFFs) for bulk solid and liquid phases of d-block elements. In exhaustive detail, we contrast the performance of force, energy, and stress predictions across the transition metals for two leading MLFF models: a kernel-based atomic cluster expansion method implemented using sparse Gaussian processes (FLARE), and an equivariant message-passing neural network (NequIP). Early transition metals present higher relative errors and are more difficult to learn relative to late platinum- and coinage-group elements, and this trend persists across model architectures. Trends in complexity of interatomic interactions for different metals are revealed via comparison of the performance of representations with different many-body order and angular resolution. Using arguments based on perturbation theory on the occupied and unoccupied d states near the Fermi level, we determine that the large, sharp d density of states both above and below the Fermi level in early transition metals leads to a more complex, harder-to-learn potential energy surface for these metals. Increasing the fictitious electronic temperature (smearing) modifies the angular sensitivity of forces and makes the early transition metal forces easier to learn. This work illustrates challenges in capturing intricate properties of metallic bonding with current leading MLFFs and provides a reference data set for transition metals, aimed at benchmarking the accuracy and improving the development of emerging machine-learned approximations.
△ Less
Submitted 26 September, 2023; v1 submitted 25 February, 2023;
originally announced February 2023.
-
CIDER: An Expressive, Nonlocal Feature Set for Machine Learning Density Functionals with Exact Constraints
Authors:
Kyle Bystrom,
Boris Kozinsky
Abstract:
Machine learning (ML) has recently gained attention as a means to develop more accurate exchange-correlation (XC) functionals for density functional theory, but functionals developed thus far need to be improved on several metrics, including accuracy, numerical stability, and transferability across chemical space. In this work, we introduce a set of nonlocal features of the density called the CIDE…
▽ More
Machine learning (ML) has recently gained attention as a means to develop more accurate exchange-correlation (XC) functionals for density functional theory, but functionals developed thus far need to be improved on several metrics, including accuracy, numerical stability, and transferability across chemical space. In this work, we introduce a set of nonlocal features of the density called the CIDER formalism, which we use to train a Gaussian process model for the exchange energy that obeys the critical uniform scaling rule for exchange. The resulting CIDER exchange functional is significantly more accurate than any semi-local functional tested here, and it has good transferability across main-group molecules. This work therefore serves as an initial step toward more accurate exchange functionals, and it also introduces useful techniques for developing robust, physics-informed XC models via ML.
△ Less
Submitted 24 January, 2022; v1 submitted 6 September, 2021;
originally announced September 2021.
-
Pawpyseed: Perturbation-extrapolation band shifting corrections for point defect calculations
Authors:
Kyle Bystrom,
Danny Broberg,
Shyam Dwaraknath,
Kristin A. Persson,
Mark Asta
Abstract:
Significant progress has been made recently in the automation and standardization of ab initio point defect calculations. However, the task of developing, implementing, and benchmarking charge corrections for density functional theory (DFT) point defect calculations is still an open challenge. Here we present a high-performance Python package called pawpyseed, which can read PAW DFT wave functions…
▽ More
Significant progress has been made recently in the automation and standardization of ab initio point defect calculations. However, the task of developing, implementing, and benchmarking charge corrections for density functional theory (DFT) point defect calculations is still an open challenge. Here we present a high-performance Python package called pawpyseed, which can read PAW DFT wave functions and calculate the overlap between wavefunctions from different structures. Using pawpyseed, we implement a new band shifting correction derived from first order perturbation theory. We benchmark this method by calculating the transition levels of several point defects in silicon and comparing to experimental and hybrid functional results. The new band shifting method can shift single-particle energies to improve transition level predictions and can be automated and parallelized using pawpyseed, suggesting it could be a useful method for high-throughput point defect calculations.
△ Less
Submitted 25 April, 2019;
originally announced April 2019.
