Bikash Kanungo, Jeffrey Hatch, Paul M. Zimmerman, Vikram Gavini
Finding accurate exchange-correlation (XC) functionals remains the defining challenge in density functional theory (DFT). Despite 40 years of active development, the desired chemical accuracy is still elusive with existing functionals. We present a data-driven pathway to learn the XC functionals by utilizing the exact density, XC energy, and XC potential. While the exact densities are obtained from accurate configuration interaction (CI), the exact XC energies and XC potentials are obtained via inverse DFT calculations on the CI densities. We demonstrate how simple neural network (NN) based local density approximation (LDA) and generalized gradient approximation (GGA), trained on just five atoms and two molecules, provide remarkable improvement in total energies, densities, atomization energies, and barrier heights for hundreds of molecules outside the training set. Particularly, the NN-based GGA functional attains similar accuracy as the higher rung SCAN meta-GGA, highlighting the promise of using the XC potential in modeling XC functionals. We expect this approach to pave the way for systematic learning of increasingly accurate and sophisticated XC functionals.
{"title":"Learning local and semi-local density functionals from exact exchange-correlation potentials and energies","authors":"Bikash Kanungo, Jeffrey Hatch, Paul M. Zimmerman, Vikram Gavini","doi":"arxiv-2409.06498","DOIUrl":"https://doi.org/arxiv-2409.06498","url":null,"abstract":"Finding accurate exchange-correlation (XC) functionals remains the defining\u0000challenge in density functional theory (DFT). Despite 40 years of active\u0000development, the desired chemical accuracy is still elusive with existing\u0000functionals. We present a data-driven pathway to learn the XC functionals by\u0000utilizing the exact density, XC energy, and XC potential. While the exact\u0000densities are obtained from accurate configuration interaction (CI), the exact\u0000XC energies and XC potentials are obtained via inverse DFT calculations on the\u0000CI densities. We demonstrate how simple neural network (NN) based local density\u0000approximation (LDA) and generalized gradient approximation (GGA), trained on\u0000just five atoms and two molecules, provide remarkable improvement in total\u0000energies, densities, atomization energies, and barrier heights for hundreds of\u0000molecules outside the training set. Particularly, the NN-based GGA functional\u0000attains similar accuracy as the higher rung SCAN meta-GGA, highlighting the\u0000promise of using the XC potential in modeling XC functionals. We expect this\u0000approach to pave the way for systematic learning of increasingly accurate and\u0000sophisticated XC functionals.","PeriodicalId":501369,"journal":{"name":"arXiv - PHYS - Computational Physics","volume":"6 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142204042","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The effect of Diabatisation is reported in the excited argon isomers using the Diatomic-In-Molecules (DIM) method. In previous work using DIM, the lowest energy isomers of Ar$_N^*$ were shown as Ar$_3^*-$Ar$_{N-3}$, however, using the Hole-Particle-Psedopotential (HPP) method, it was shown that the excitation is localised over dimer not trimer; Ar$_2^*-$Ar$_{N-2}$. In this work we improve the DIM calculations by including previously ignored strongly avoided crossing between 3p4s and 3p4p $^{1,3}Sigma$ states.
利用二原子分子内（DIM）方法报告了激发氩异构体的二原子化效应。在以前使用 DIM 方法进行的研究中，Ar$_N^*$ 的最低能量异构体被显示为 Ar$_3^*-$Ar$_{N-3}$，然而，使用孔-粒子-位移电位（HPP）方法，结果表明激发被定位在二聚体而不是三聚体上；Ar$_2^*-$Ar$_{N-2}$。在这项工作中，我们将以前忽略的 3p4s 和 3p4p $^{1,3}Sigma$ 态之间的强回避交叉包括在内，从而改进了 DIM 计算。
{"title":"An Update to Isomers of Rydberg Excitations in Argon Clusters","authors":"Mukul Dhiman, Benoit Gervais","doi":"arxiv-2409.06484","DOIUrl":"https://doi.org/arxiv-2409.06484","url":null,"abstract":"The effect of Diabatisation is reported in the excited argon isomers using\u0000the Diatomic-In-Molecules (DIM) method. In previous work using DIM, the lowest\u0000energy isomers of Ar$_N^*$ were shown as Ar$_3^*-$Ar$_{N-3}$, however, using\u0000the Hole-Particle-Psedopotential (HPP) method, it was shown that the excitation\u0000is localised over dimer not trimer; Ar$_2^*-$Ar$_{N-2}$. In this work we\u0000improve the DIM calculations by including previously ignored strongly avoided\u0000crossing between 3p4s and 3p4p $^{1,3}Sigma$ states.","PeriodicalId":501369,"journal":{"name":"arXiv - PHYS - Computational Physics","volume":"50 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142204044","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The use of Green's function in quantum many-body theory often leads to nonlinear eigenvalue problems, as Green's function needs to be defined in energy domain. The $GW$ approximation method is one of the typical examples. In this article, we introduce a method based on the FEAST eigenvalue algorithm for accurately solving the nonlinear eigenvalue $G_0W_0$ quasiparticle equation, eliminating the need for the Kohn-Sham wavefunction approximation. Based on the contour integral method for nonlinear eigenvalue problem, the energy (eigenvalue) domain is extended to complex plane. Hypercomplex number is introduced to the contour deformation calculation of $GW$ self-energy to carry imaginary parts of both Green's functions and FEAST quadrature nodes. Calculation results for various molecules are presented and compared with a more conventional graphical solution approximation method. It is confirmed that the Highest Occupied Molecular Orbital (HOMO) from the Kohn-Sham equation is very close to that of $GW$, while the Least Unoccupied Molecular Orbital (LUMO) shows noticeable differences.
{"title":"FEAST nonlinear eigenvalue algorithm for $GW$ quasiparticle equations","authors":"Dongming Li, Eric Polizzi","doi":"arxiv-2409.06119","DOIUrl":"https://doi.org/arxiv-2409.06119","url":null,"abstract":"The use of Green's function in quantum many-body theory often leads to\u0000nonlinear eigenvalue problems, as Green's function needs to be defined in\u0000energy domain. The $GW$ approximation method is one of the typical examples. In\u0000this article, we introduce a method based on the FEAST eigenvalue algorithm for\u0000accurately solving the nonlinear eigenvalue $G_0W_0$ quasiparticle equation,\u0000eliminating the need for the Kohn-Sham wavefunction approximation. Based on the\u0000contour integral method for nonlinear eigenvalue problem, the energy\u0000(eigenvalue) domain is extended to complex plane. Hypercomplex number is\u0000introduced to the contour deformation calculation of $GW$ self-energy to carry\u0000imaginary parts of both Green's functions and FEAST quadrature nodes.\u0000Calculation results for various molecules are presented and compared with a\u0000more conventional graphical solution approximation method. It is confirmed that\u0000the Highest Occupied Molecular Orbital (HOMO) from the Kohn-Sham equation is\u0000very close to that of $GW$, while the Least Unoccupied Molecular Orbital (LUMO)\u0000shows noticeable differences.","PeriodicalId":501369,"journal":{"name":"arXiv - PHYS - Computational Physics","volume":"68 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142204039","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Ariadni Boziki, Frédéric Ngono Mebenga, Philippe Fernandes, Alexandre Tkatchenko
Vibrational spectroscopy is an indispensable analytical tool that provides structural fingerprints for molecules, solids, and interfaces thereof. This study introduces THeSeuSS (THz Spectra Simulations Software) - an automated computational platform that efficiently simulates IR and Raman spectra for both periodic and non-periodic systems. Utilizing DFT and DFTB, THeSeuSS offers robust capabilities for detailed vibrational spectra simulations. Our extensive evaluations and benchmarks demonstrate that THeSeuSS accurately reproduces both previously calculated and experimental spectra, enabling precise comparisons and interpretations of vibrational features across various test cases, including H2O and glycine molecules in the gas phase, as well as solid ammonia and solid ibuprofen. Designed with a user-friendly interface and seamless integration with existing computational chemistry tools, THeSeuSS enhances the accessibility and applicability of advanced spectroscopic simulations, supporting research and development in chemical, pharmaceutical, and material sciences. Future updates aim to expand its methodological diversity by incorporating machine learning techniques to analyze larger and more complex systems, solidifying THeSeuSS's role as an essential tool in the computational chemist's arsenal.
{"title":"A Journey with THeSeuSS: Automated Python Tool for Modeling IR and Raman Vibrational Spectra of Molecules and Solids","authors":"Ariadni Boziki, Frédéric Ngono Mebenga, Philippe Fernandes, Alexandre Tkatchenko","doi":"arxiv-2409.06597","DOIUrl":"https://doi.org/arxiv-2409.06597","url":null,"abstract":"Vibrational spectroscopy is an indispensable analytical tool that provides\u0000structural fingerprints for molecules, solids, and interfaces thereof. This\u0000study introduces THeSeuSS (THz Spectra Simulations Software) - an automated\u0000computational platform that efficiently simulates IR and Raman spectra for both\u0000periodic and non-periodic systems. Utilizing DFT and DFTB, THeSeuSS offers\u0000robust capabilities for detailed vibrational spectra simulations. Our extensive\u0000evaluations and benchmarks demonstrate that THeSeuSS accurately reproduces both\u0000previously calculated and experimental spectra, enabling precise comparisons\u0000and interpretations of vibrational features across various test cases,\u0000including H2O and glycine molecules in the gas phase, as well as solid ammonia\u0000and solid ibuprofen. Designed with a user-friendly interface and seamless\u0000integration with existing computational chemistry tools, THeSeuSS enhances the\u0000accessibility and applicability of advanced spectroscopic simulations,\u0000supporting research and development in chemical, pharmaceutical, and material\u0000sciences. Future updates aim to expand its methodological diversity by\u0000incorporating machine learning techniques to analyze larger and more complex\u0000systems, solidifying THeSeuSS's role as an essential tool in the computational\u0000chemist's arsenal.","PeriodicalId":501369,"journal":{"name":"arXiv - PHYS - Computational Physics","volume":"47 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142204043","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Jorge I. Hernandez-Martinez, Gerardo Rodriguez-Hernandez, Andres Mendez-Vazquez
We propose a data-driven approach using a Restricted Boltzmann Machine (RBM) to solve the Schr"odinger equation in configuration space. Traditional Configuration Interaction (CI) methods, while powerful, are computationally expensive due to the large number of determinants required. Our approach leverages RBMs to efficiently identify and sample the most significant determinants, accelerating convergence and reducing computational cost. This method achieves up to 99.99% of the correlation energy even by four orders of magnitude less determinants compared to full CI calculations and up to two orders of magnitude less than previous state of the art works. Additionally, our study demonstrate that the RBM can learn the underlying quantum properties, providing more detail insights than other methods . This innovative data-driven approach offers a promising tool for quantum chemistry, enhancing both efficiency and understanding of complex systems.
我们提出了一种数据驱动的方法，利用受限玻尔兹曼机（RBM）来求解构型空间中的薛定谔方程。传统的配置交互（CI）方法虽然功能强大，但由于需要大量的行列式，因此计算成本很高。我们的方法利用 RBM 高效地识别和采样最重要的行列式，加快了收敛速度并降低了计算成本。与完整的 CI 计算相比，这种方法即使减少了四个数量级的行列式，也能实现高达 99.99% 的相关能量，比以前的技术水平低两个数量级。此外，我们的研究表明，RBM 可以学习潜在的量子特性，提供比其他方法更详细的见解。这种创新的数据驱动方法为量子化学提供了一种前景广阔的工具，既提高了效率，又加深了对复杂系统的理解。
{"title":"Configuration Interaction Guided Sampling with Interpretable Restricted Boltzmann Machine","authors":"Jorge I. Hernandez-Martinez, Gerardo Rodriguez-Hernandez, Andres Mendez-Vazquez","doi":"arxiv-2409.06146","DOIUrl":"https://doi.org/arxiv-2409.06146","url":null,"abstract":"We propose a data-driven approach using a Restricted Boltzmann Machine (RBM)\u0000to solve the Schr\"odinger equation in configuration space. Traditional\u0000Configuration Interaction (CI) methods, while powerful, are computationally\u0000expensive due to the large number of determinants required. Our approach\u0000leverages RBMs to efficiently identify and sample the most significant\u0000determinants, accelerating convergence and reducing computational cost. This\u0000method achieves up to 99.99% of the correlation energy even by four orders of\u0000magnitude less determinants compared to full CI calculations and up to two\u0000orders of magnitude less than previous state of the art works. Additionally,\u0000our study demonstrate that the RBM can learn the underlying quantum properties,\u0000providing more detail insights than other methods . This innovative data-driven\u0000approach offers a promising tool for quantum chemistry, enhancing both\u0000efficiency and understanding of complex systems.","PeriodicalId":501369,"journal":{"name":"arXiv - PHYS - Computational Physics","volume":"29 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142204053","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Compared to ground state electronic structure optimizations, accurate simulations of molecular real-time electron dynamics are usually much more difficult to perform. To simulate electron dynamics, the time-dependent density matrix renormalization group (TDDMRG) has been shown to offer an attractive compromise between accuracy and cost. However, many simulation parameters significantly affect the quality and efficiency of a TDDMRG simulation. So far, it is unclear whether common wisdom from ground state DMRG carries over to the TDDMRG, and a guideline on how to choose these parameters is missing. Here, in order to establish such a guideline, we investigate the convergence behavior of the main TDDMRG simulation parameters, such as time integrator, the choice of orbitals, and the choice of MPS representation for complex-valued non-singlet states. In addition, we propose a method to select orbitals that are tailored to optimize the dynamics. Lastly, we showcase the TDDMRG by applying it to charge migration ionization dynamics in furfural, where we reveal a rapid conversion from an ionized state with a $sigma$ character to one with a $pi$ character within less than a femtosecond.
{"title":"Simulating real-time molecular electron dynamics efficiently using the time-dependent density matrix renormalization group","authors":"Imam S. Wahyutama, Henrik R. Larsson","doi":"arxiv-2409.05959","DOIUrl":"https://doi.org/arxiv-2409.05959","url":null,"abstract":"Compared to ground state electronic structure optimizations, accurate\u0000simulations of molecular real-time electron dynamics are usually much more\u0000difficult to perform. To simulate electron dynamics, the time-dependent density\u0000matrix renormalization group (TDDMRG) has been shown to offer an attractive\u0000compromise between accuracy and cost. However, many simulation parameters\u0000significantly affect the quality and efficiency of a TDDMRG simulation. So far,\u0000it is unclear whether common wisdom from ground state DMRG carries over to the\u0000TDDMRG, and a guideline on how to choose these parameters is missing. Here, in\u0000order to establish such a guideline, we investigate the convergence behavior of\u0000the main TDDMRG simulation parameters, such as time integrator, the choice of\u0000orbitals, and the choice of MPS representation for complex-valued non-singlet\u0000states. In addition, we propose a method to select orbitals that are tailored\u0000to optimize the dynamics. Lastly, we showcase the TDDMRG by applying it to\u0000charge migration ionization dynamics in furfural, where we reveal a rapid\u0000conversion from an ionized state with a $sigma$ character to one with a $pi$\u0000character within less than a femtosecond.","PeriodicalId":501369,"journal":{"name":"arXiv - PHYS - Computational Physics","volume":"9 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142204055","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Manuel Ballester, Jaromir Kaspar, Francesc Massanes, Srutarshi Banerjee, Alexander Hans Vija, Aggelos K. Katsaggelos
CdZnTe-based detectors are highly valued because of their high spectral resolution, which is an essential feature for nuclear medical imaging. However, this resolution is compromised when there are substantial defects in the CdZnTe crystals. In this study, we present a learning-based approach to determine the spatially dependent bulk properties and defects in semiconductor detectors. This characterization allows us to mitigate and compensate for the undesired effects caused by crystal impurities. We tested our model with computer-generated noise-free input data, where it showed excellent accuracy, achieving an average RMSE of 0.43% between the predicted and the ground truth crystal properties. In addition, a sensitivity analysis was performed to determine the effect of noisy data on the accuracy of the model.
{"title":"Characterization of Crystal Properties and Defects in CdZnTe Radiation Detectors","authors":"Manuel Ballester, Jaromir Kaspar, Francesc Massanes, Srutarshi Banerjee, Alexander Hans Vija, Aggelos K. Katsaggelos","doi":"arxiv-2409.06738","DOIUrl":"https://doi.org/arxiv-2409.06738","url":null,"abstract":"CdZnTe-based detectors are highly valued because of their high spectral\u0000resolution, which is an essential feature for nuclear medical imaging. However,\u0000this resolution is compromised when there are substantial defects in the CdZnTe\u0000crystals. In this study, we present a learning-based approach to determine the\u0000spatially dependent bulk properties and defects in semiconductor detectors.\u0000This characterization allows us to mitigate and compensate for the undesired\u0000effects caused by crystal impurities. We tested our model with\u0000computer-generated noise-free input data, where it showed excellent accuracy,\u0000achieving an average RMSE of 0.43% between the predicted and the ground truth\u0000crystal properties. In addition, a sensitivity analysis was performed to\u0000determine the effect of noisy data on the accuracy of the model.","PeriodicalId":501369,"journal":{"name":"arXiv - PHYS - Computational Physics","volume":"9 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142204040","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Cluster expansions are commonly employed as surrogate models to link the electronic structure of an alloy to its finite-temperature properties. Using cluster expansions to model materials with several alloying elements is challenging due to a rapid increase in the number of fitting parameters and training set size. We introduce the embedded cluster expansion (eCE) formalism that enables the parameterization of accurate on-lattice surrogate models for alloys containing several chemical species. The eCE model simultaneously learns a low dimensional embedding of site basis functions along with the weights of an energy model. A prototypical senary alloy comprised of elements in groups 5 and 6 of the periodic table is used to demonstrate that eCE models can accurately reproduce ordering energetics of complex alloys without a significant increase in model complexity. Further, eCE models can leverage similarities between chemical elements to efficiently extrapolate into compositional spaces that are not explicitly included in the training dataset. The eCE formalism presented in this study unlocks the possibility of employing cluster expansion models to study multicomponent alloys containing several alloying elements.
{"title":"Constructing multicomponent cluster expansions with machine-learning and chemical embedding","authors":"Yann L. Müller, Anirudh Raju Natarajan","doi":"arxiv-2409.06071","DOIUrl":"https://doi.org/arxiv-2409.06071","url":null,"abstract":"Cluster expansions are commonly employed as surrogate models to link the\u0000electronic structure of an alloy to its finite-temperature properties. Using\u0000cluster expansions to model materials with several alloying elements is\u0000challenging due to a rapid increase in the number of fitting parameters and\u0000training set size. We introduce the embedded cluster expansion (eCE) formalism\u0000that enables the parameterization of accurate on-lattice surrogate models for\u0000alloys containing several chemical species. The eCE model simultaneously learns\u0000a low dimensional embedding of site basis functions along with the weights of\u0000an energy model. A prototypical senary alloy comprised of elements in groups 5\u0000and 6 of the periodic table is used to demonstrate that eCE models can\u0000accurately reproduce ordering energetics of complex alloys without a\u0000significant increase in model complexity. Further, eCE models can leverage\u0000similarities between chemical elements to efficiently extrapolate into\u0000compositional spaces that are not explicitly included in the training dataset.\u0000The eCE formalism presented in this study unlocks the possibility of employing\u0000cluster expansion models to study multicomponent alloys containing several\u0000alloying elements.","PeriodicalId":501369,"journal":{"name":"arXiv - PHYS - Computational Physics","volume":"61 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142204054","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Understanding and predicting interface diffusion phenomena in materials is crucial for various industrial applications, including semiconductor manufacturing, battery technology, and catalysis. In this study, we propose a novel approach utilizing Graph Neural Networks (GNNs) to investigate and model material interface diffusion. We begin by collecting experimental and simulated data on diffusion coefficients, concentration gradients, and other relevant parameters from diverse material systems. The data are preprocessed, and key features influencing interface diffusion are extracted. Subsequently, we construct a GNN model tailored to the diffusion problem, with a graph representation capturing the atomic structure of materials. The model architecture includes multiple graph convolutional layers for feature aggregation and update, as well as optional graph attention layers to capture complex relationships between atoms. We train and validate the GNN model using the preprocessed data, achieving accurate predictions of diffusion coefficients, diffusion rates, concentration profiles, and potential diffusion pathways. Our approach offers insights into the underlying mechanisms of interface diffusion and provides a valuable tool for optimizing material design and engineering. Additionally, our method offers possible strategies to solve the longstanding problems related to materials interface diffusion.
{"title":"Investigating Material Interface Diffusion Phenomena through Graph Neural Networks in Applied Materials","authors":"Zirui Zhao, Haifeng-Li","doi":"arxiv-2409.05306","DOIUrl":"https://doi.org/arxiv-2409.05306","url":null,"abstract":"Understanding and predicting interface diffusion phenomena in materials is\u0000crucial for various industrial applications, including semiconductor\u0000manufacturing, battery technology, and catalysis. In this study, we propose a\u0000novel approach utilizing Graph Neural Networks (GNNs) to investigate and model\u0000material interface diffusion. We begin by collecting experimental and simulated\u0000data on diffusion coefficients, concentration gradients, and other relevant\u0000parameters from diverse material systems. The data are preprocessed, and key\u0000features influencing interface diffusion are extracted. Subsequently, we\u0000construct a GNN model tailored to the diffusion problem, with a graph\u0000representation capturing the atomic structure of materials. The model\u0000architecture includes multiple graph convolutional layers for feature\u0000aggregation and update, as well as optional graph attention layers to capture\u0000complex relationships between atoms. We train and validate the GNN model using\u0000the preprocessed data, achieving accurate predictions of diffusion\u0000coefficients, diffusion rates, concentration profiles, and potential diffusion\u0000pathways. Our approach offers insights into the underlying mechanisms of\u0000interface diffusion and provides a valuable tool for optimizing material design\u0000and engineering. Additionally, our method offers possible strategies to solve\u0000the longstanding problems related to materials interface diffusion.","PeriodicalId":501369,"journal":{"name":"arXiv - PHYS - Computational Physics","volume":"29 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142204059","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Jonathan Maes, Diego De Gusem, Ian Lateur, Jonathan Leliaert, Aleksandr Kurenkov, Bartel Van Waeyenberge
We present Hotspice, a Monte Carlo simulation software designed to capture the dynamics and equilibrium states of Artificial Spin Ice (ASI) systems with both in-plane (IP) and out-of-plane (OOP) geometries. An Ising-like model is used where each nanomagnet is represented as a macrospin, with switching events driven by thermal fluctuations, magnetostatic interactions, and external fields. To improve simulation accuracy, we explore the impact of several corrections to this model, concerning for example the calculation of the dipole interaction in IP and OOP ASI, as well as the impact of allowing asymmetric rather than symmetric energy barriers between stable states. We validate these enhancements by comparing simulation results with experimental data for pinwheel and kagome ASI lattices, demonstrating how these corrections enable a more accurate simulation of the behavior of these systems. We finish with a demonstration of 'clocking' in pinwheel and OOP square ASI as an example of reservoir computing.
我们介绍的 Hotspice 是一款蒙特卡罗模拟软件，旨在捕捉平面内（IP）和平面外（OOP）几何形状的人造自旋冰（ASI）系统的动力学和平衡态。我们使用了一个类似伊辛的模型，其中每个纳米磁体都被表示为一个大自旋，开关事件由热波动、磁静相互作用和外场驱动。为了提高模拟精度，我们探讨了对该模型进行若干修正的影响，例如 IP 和 OOP ASI 中偶极子相互作用的计算，以及允许稳定状态之间的非对称能垒而非对称能垒的影响。我们将模拟结果与 PINWELL 和 KAGOME ASI 晶格的实验数据进行了比较，从而验证了这些改进，证明了这些修正如何能够更精确地模拟这些系统的行为。最后，我们以储层计算为例，演示了pinwheel 和 OOP square ASI 中的 "时钟"。
{"title":"The design, verification, and applications of Hotspice: a Monte Carlo simulator for artificial spin ice","authors":"Jonathan Maes, Diego De Gusem, Ian Lateur, Jonathan Leliaert, Aleksandr Kurenkov, Bartel Van Waeyenberge","doi":"arxiv-2409.05580","DOIUrl":"https://doi.org/arxiv-2409.05580","url":null,"abstract":"We present Hotspice, a Monte Carlo simulation software designed to capture\u0000the dynamics and equilibrium states of Artificial Spin Ice (ASI) systems with\u0000both in-plane (IP) and out-of-plane (OOP) geometries. An Ising-like model is\u0000used where each nanomagnet is represented as a macrospin, with switching events\u0000driven by thermal fluctuations, magnetostatic interactions, and external\u0000fields. To improve simulation accuracy, we explore the impact of several\u0000corrections to this model, concerning for example the calculation of the dipole\u0000interaction in IP and OOP ASI, as well as the impact of allowing asymmetric\u0000rather than symmetric energy barriers between stable states. We validate these\u0000enhancements by comparing simulation results with experimental data for\u0000pinwheel and kagome ASI lattices, demonstrating how these corrections enable a\u0000more accurate simulation of the behavior of these systems. We finish with a\u0000demonstration of 'clocking' in pinwheel and OOP square ASI as an example of\u0000reservoir computing.","PeriodicalId":501369,"journal":{"name":"arXiv - PHYS - Computational Physics","volume":"23 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142204056","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：