Computational Physics
See recent articles
Showing new listings for Wednesday, 12 March 2025
- [1] arXiv:2503.07619 [pdf, html, other]
-
Title: Physics- and data-driven Active Learning of neural network representations for free energy functions of materials from statistical mechanicsSubjects: Computational Physics (physics.comp-ph); Machine Learning (cs.LG)
Accurate free energy representations are crucial for understanding phase dynamics in materials. We employ a scale-bridging approach to incorporate atomistic information into our free energy model by training a neural network on DFT-informed Monte Carlo data. To optimize sampling in the high-dimensional Monte Carlo space, we present an Active Learning framework that integrates space-filling sampling, uncertainty-based sampling, and physics-informed sampling. Additionally, our approach includes methods such as hyperparameter tuning, dynamic sampling, and novelty enforcement. These strategies can be combined to reduce MSE,either globally or in targeted regions of interest,while minimizing the number of required data points. The framework introduced here is broadly applicable to Monte Carlo sampling of a range of materials systems.
- [2] arXiv:2503.08039 [pdf, html, other]
-
Title: A quantum Monte Carlo algorithm for arbitrary high-spin HamiltoniansComments: 10 pages, 2 figures, 1 table. arXiv admin note: text overlap with arXiv:2307.06503Subjects: Computational Physics (physics.comp-ph); Other Condensed Matter (cond-mat.other); Strongly Correlated Electrons (cond-mat.str-el); Quantum Physics (quant-ph)
We present a universal parameter-free quantum Monte Carlo algorithm for simulating arbitrary high-spin (spin greater than $1/2$) Hamiltonians. This approach extends a previously developed method by the authors for spin-$1/2$ Hamiltonians [Phys. Rev. Research 6, 013281 (2024)]. To demonstrate its applicability and versatility, we apply our method to the spin-$1$ and spin-$3/2$ quantum Heisenberg models on the square lattice. Additionally, we detail how the approach naturally extends to general Hamiltonians involving mixtures of particle species, including bosons and fermions. We have made our program code freely accessible on GitHub.
- [3] arXiv:2503.08183 [pdf, html, other]
-
Title: Physics-based AI methodology for Material Parameter Extraction from Optical DataM. Koumans, J.L.M. van Mechelen (Eindhoven University of Technology)Comments: Submitted for IRMMW-THz 2025 conference proceedingsSubjects: Computational Physics (physics.comp-ph); Computer Vision and Pattern Recognition (cs.CV); Optics (physics.optics)
We report on a novel methodology for extracting material parameters from spectroscopic optical data using a physics-based neural network. The proposed model integrates classical optimization frameworks with a multi-scale object detection framework, specifically exploring the effect of incorporating physics into the neural network. We validate and analyze its performance on simulated transmission spectra at terahertz and infrared frequencies. Compared to traditional model-based approaches, our method is designed to be autonomous, robust, and time-efficient, making it particularly relevant for industrial and societal applications.
- [4] arXiv:2503.08207 [pdf, html, other]
-
Title: To Use or Not to Use a Universal Force FieldComments: 21 pages, 5 figuresSubjects: Computational Physics (physics.comp-ph); Materials Science (cond-mat.mtrl-sci); Machine Learning (cs.LG)
Artificial intelligence (AI) is revolutionizing scientific research, particularly in computational materials science, by enabling more accurate and efficient simulations. Machine learning force fields (MLFFs) have emerged as powerful tools for molecular dynamics (MD) simulations, potentially offering quantum-mechanical accuracy with the efficiency of classical MD. This Perspective evaluates the viability of universal MLFFs for simulating complex materials systems from the standpoint of a potential practitioner. Using the temperature-driven ferroelectric-paraelectric phase transition of PbTiO$_3$ as a benchmark, we assess leading universal force fields, including CHGNet, MACE, M3GNet, and GPTFF, alongside specialized models like UniPero. While universal MLFFs trained on PBE-derived datasets perform well in predicting equilibrium properties, they largely fail to capture realistic finite-temperature phase transitions under constant-pressure MD, often exhibiting unphysical instabilities. These shortcomings stem from inherited biases in exchange-correlation functionals and limited generalization to anharmonic interactions governing dynamic behavior. However, fine-tuning universal models or employing system-specific MLFFs like UniPero successfully restores predictive accuracy. We advocates for hybrid approaches combining universal pretraining with targeted optimization, improved error quantification frameworks, and community-driven benchmarks to advance MLFFs as robust tools for computational materials discovery.
New submissions (showing 4 of 4 entries)
- [5] arXiv:2503.07847 (cross-list from physics.flu-dyn) [pdf, html, other]
-
Title: Python-JAX-based Fast Stokesian DynamicsSubjects: Fluid Dynamics (physics.flu-dyn); Computational Physics (physics.comp-ph)
Stokesian Dynamics (SD) is a powerful computational framework for simulating the motion of particles in a viscous Newtonian fluid under Stokes-flow conditions. Traditional SD implementations can be computationally expensive as they rely on the inversion of large mobility matrices to determine hydrodynamic interactions. Recently, however, the simulation of thermalized systems with large numbers of particles has become feasible [Fiore and Swan, J. Fluid. Mech. $\textbf{878}$, 544 (2019)]. Their ``fast Stokesian dynamics'' (FSD) method leverages a saddle-point formulation to ensure overall scaling of the algorithm that is linear in the number of particles $\mathcal{O}(N)$; performance relies on dedicated graphics-processing-unit computing. Here, we present a different route toward implementing FSD, which instead leverages the Just-in-Time (JIT) compilation capabilities of Google JAX. We refer to this implementation as JFSD and perform benchmarks on it to verify that it has the right scaling and is sufficiently fast by the standards of modern computational physics. In addition, we provide a series of physical test cases that help ensure accuracy and robustness, as the code undergoes further development. Thus, JFSD is ready to facilitate the study of hydrodynamic effects in particle suspensions across the domains of soft, active, and granular matter.
- [6] arXiv:2503.08146 (cross-list from hep-ex) [pdf, html, other]
-
Title: Fast Jet Finding in JuliaSubjects: High Energy Physics - Experiment (hep-ex); Computational Physics (physics.comp-ph)
Jet reconstruction remains a critical task in the analysis of data from HEP colliders. We describe in this paper a new, highly performant, Julia package for jet reconstruction, this http URL, which integrates into the growing ecosystem of Julia packages for HEP. With this package users can run sequential reconstruction algorithms for jets. In particular, for LHC events, the Anti-${k}_\text{T}$, Cambridge/Aachen and Inclusive-${k}_\text{T}$ algorithms can be used. For FCCee studies the use of alternative algorithms such as the Generalised ${k}_\text{T}$ for $e^+e^-$ and Durham are also supported.
The performance of the core algorithms is better than Fastjet's C++ implementation, for typical LHC and FCCee events, thanks to the Julia compiler's exploitation of single-instruction-multiple-data (SIMD), as well as ergonomic compact data layouts.
The full reconstruction history is made available, allowing inclusive and exclusive jets to be retrieved. The package also provides the means to visualise the reconstruction. Substructure algorithms have been added that allow advanced analysis techniques to be employed. The package can read event data from EDM4hep files and reconstruct jets from these directly, opening the door to FCCee and other future collider studies in Julia. - [7] arXiv:2503.08184 (cross-list from hep-ex) [pdf, html, other]
-
Title: Julia in HEPGraeme Andrew Stewart, Alexander Moreno Briceño, Philippe Gras, Benedikt Hegner, Uwe Hernandez Acosta, Tamas Gal, Jerry Ling, Pere Mato, Mikhail Mikhasenko, Oliver Schulz, Sam SkipseySubjects: High Energy Physics - Experiment (hep-ex); Computational Physics (physics.comp-ph)
Julia is a mature general-purpose programming language, with a large ecosystem of libraries and more than 12000 third-party packages, which specifically targets scientific computing. As a language, Julia is as dynamic, interactive, and accessible as Python with NumPy, but achieves run-time performance on par with C/C++. In this paper, we describe the state of adoption of Julia in HEP, where momentum has been gathering over a number of years.
HEP-oriented Julia packages can already, via this http URL, read HEP's major file formats, including TTree and RNTuple. Interfaces to some of HEP's major software packages, such as through this http URL, are available too. Jet reconstruction algorithms in Julia show excellent performance. A number of full HEP analyses have been performed in Julia.
We show how, as the support for HEP has matured, developments have benefited from Julia's core design choices, which makes reuse from and integration with other packages easy. In particular, libraries developed outside HEP for plotting, statistics, fitting, and scientific machine learning are extremely useful.
We believe that the powerful combination of flexibility and speed, the wide selection of scientific programming tools, and support for all modern programming paradigms and tools, make Julia the ideal choice for a future language in HEP. - [8] arXiv:2503.08305 (cross-list from cs.LG) [pdf, html, other]
-
Title: ELECTRA: A Symmetry-breaking Cartesian Network for Charge Density Prediction with Floating OrbitalsComments: 8 pages, 3 figures, 1 tableSubjects: Machine Learning (cs.LG); Chemical Physics (physics.chem-ph); Computational Physics (physics.comp-ph)
We present the Electronic Tensor Reconstruction Algorithm (ELECTRA) - an equivariant model for predicting electronic charge densities using "floating" orbitals. Floating orbitals are a long-standing idea in the quantum chemistry community that promises more compact and accurate representations by placing orbitals freely in space, as opposed to centering all orbitals at the position of atoms. Finding ideal placements of these orbitals requires extensive domain knowledge though, which thus far has prevented widespread adoption. We solve this in a data-driven manner by training a Cartesian tensor network to predict orbital positions along with orbital coefficients. This is made possible through a symmetry-breaking mechanism that is used to learn position displacements with lower symmetry than the input molecule while preserving the rotation equivariance of the charge density itself. Inspired by recent successes of Gaussian Splatting in representing densities in space, we are using Gaussians as our orbitals and predict their weights and covariance matrices. Our method achieves a state-of-the-art balance between computational efficiency and predictive accuracy on established benchmarks.
- [9] arXiv:2503.08396 (cross-list from nucl-th) [pdf, html, other]
-
Title: Study of 14.1 MeV Neutron Moderation in BerylliumComments: 7 pages, 7 figuresSubjects: Nuclear Theory (nucl-th); Nuclear Experiment (nucl-ex); Computational Physics (physics.comp-ph)
This study investigates the moderation of 14.1 MeV neutrons in a natural beryllium moderator arranged in a spherical geometry. The neutron interactions and moderation efficiency were analyzed using Monte Carlo simulations with the GEANT4 toolkit. Various sphere radii were tested to determine the optimal moderator thickness for neutron thermalization.
Cross submissions (showing 5 of 5 entries)
- [10] arXiv:2410.12007 (replaced) [pdf, html, other]
-
Title: Machine learning of the Ising model on a spherical Fibonacci latticeComments: 13 pages, 16 figuresSubjects: Computational Physics (physics.comp-ph); Quantum Physics (quant-ph)
We investigate the Ising model on a spherical surface, utilizing a Fibonacci lattice to approximate uniform coverage. This setup poses challenges in achieving consistent lattice distribution across the sphere for comparison with planar models. We employ Monte Carlo simulations and graph convolutional networks (GCNs) to study spin configurations across a range of temperatures and to determine phase transition temperatures. The Fibonacci lattice, despite its uniformity, contains irregular sites that influence spin behavior. In the ferromagnetic case, sites with fewer neighbors exhibit a higher tendency for spin flips at low temperatures, though this effect weakens as temperature increases, leading to a higher phase transition temperature than in the planar Ising model. In the antiferromagnetic case, lattice irregularities prevent the total energy from reaching its minimum at zero temperature, highlighting the role of curvature and connectivity in shaping interactions. Phase transition temperatures are derived through specific heat, magnetic susceptibility analysis and GCN predictions, yielding $T_c$ values for both ferromagnetic and antiferromagnetic scenarios. This work emphasizes the impact of the Fibonacci lattice's geometric properties-namely curvature and connectivity-on spin interactions in non-planar systems, with relevance to microgravity environments.
- [11] arXiv:2412.09504 (replaced) [pdf, other]
-
Title: High Precision Binding Energies from Physics Informed Machine LearningComments: Version 2 contains edits including those based on referee feedbackJournal-ref: Phys. Rev. C 111 034305 (2025)Subjects: Nuclear Theory (nucl-th); Computational Physics (physics.comp-ph)
Twelve physics-informed machine learning models have been trained to model binding energy residuals. Our approach begins with determining the difference between measured experimental binding energies and three different mass models. Then four machine learning approaches are used to train on each energy difference. The most successful ML technique, both in interpolation and extrapolation, is the least squares boosted ensemble of trees. The best model resulting from that technique utilizes eight physical features to model the difference between experimental atomic binding energy values in AME 2012 and the Duflo Zuker mass model. This resulted in a model that fit the training data with a standard deviation of 17 keV and that has a standard deviation of 92 keV when compared all of the values in the AME 2020. The extrapolation capability of each model is discussed, and the accuracy of predicting new mass measurements has also been tested.