GeTe has attracted extensive research interest for thermoelectric applications.In this paper,we first train a neuroevolution potential(NEP)based on a dataset constructed by ab initio molecular dynamics,with the Gaussi...GeTe has attracted extensive research interest for thermoelectric applications.In this paper,we first train a neuroevolution potential(NEP)based on a dataset constructed by ab initio molecular dynamics,with the Gaussian approximation potential(GAP)as a reference.The phonon density of states is then calculated by two machine learning potentials and compared with density functional theory results,with the GAP potential having higher accuracy.Next,the thermal conductivity of a GeTe crystal at 300 K is calculated by the equilibrium molecular dynamics method using both machine learning potentials,and both of them are in good agreement with the experimental results;however,the calculation speed when using the NEP potential is about 500 times faster than when using the GAP potential.Finally,the lattice thermal conductivity in the range of 300 K-600 K is calculated using the NEP potential.The lattice thermal conductivity decreases as the temperature increases due to the phonon anharmonic effect.This study provides a theoretical tool for the study of the thermal conductivity of GeTe.展开更多
A machine learning(ML)potential for Au clusters is developed through training on a dataset including several different sized clusters.This ML potential accurately covers the whole configuration space of Au clusters in...A machine learning(ML)potential for Au clusters is developed through training on a dataset including several different sized clusters.This ML potential accurately covers the whole configuration space of Au clusters in a broad size range,thus expressing a good performance in search of their global minimum energy structures.Based on our potential,the low-lying structures of 17 different sized Au clusters are identified,which shows that small sized Au clusters tend to form planar structures while large ones are more likely to be stereo,revealing the critical size for the two-dimensional(2D)to three-dimensional(3D)structural transition.Our calculations demonstrate that ML is indeed powerful in describing the interaction of Au atoms and provides a new paradigm on accelerating the search of structures.展开更多
Chemical short-range order(SRO)affects the distribution of elements throughout the solid-solution phase of metallic alloys,thereby modifying the background against which microstructural evolution occurs.Investigating ...Chemical short-range order(SRO)affects the distribution of elements throughout the solid-solution phase of metallic alloys,thereby modifying the background against which microstructural evolution occurs.Investigating such chemistry–microstructure relationships requires atomistic models that act at the appropriate length scales while capturing the intricacies of chemical bonds leading to SRO.Here,we consider various approaches for the construction of training data sets for machine learning potentials(MLPs)for CrCoNi and evaluate their performance in capturing SRO and its effects on materials quantities of relevance for mechanical properties,such as stacking-fault energy and phase stability.It is demonstrated that energy accuracy on test sets often does not correlate with accuracy in capturing material properties,which is fundamental in enabling large-scale atomistic simulations of metallic alloys with high physical fidelity.Based on this analysis,we systematically derive design principles for the rational construction of MLPs that capture SRO in the crystal and liquid phases of alloys.展开更多
We propose a strategy for generating unbiased and systematically extendable training data for machine learning interatomic potentials(MLIP)for multicomponent alloys,called Automated Small SYmmetric Structure Training ...We propose a strategy for generating unbiased and systematically extendable training data for machine learning interatomic potentials(MLIP)for multicomponent alloys,called Automated Small SYmmetric Structure Training or ASSYST.Based on exploring the full space of random crystal structures with space groups,it facilitates the construction of training sets for MLIPs in an automatic way without prior knowledge of the material in question.The advantages of this approach are that only cells consisting of few atoms(≈10)are needed for the DFT training set,and the size and completeness of the data set can be systematically controlled with very few parameters.We validate that potentials fitted this way can accurately describe a wide range of binary and ternary phases,random alloys,as well as point and extended defects,that have not been part of the training set.Finally,we estimate the binary phase diagrams with good experimental agreement.We demonstrate that the overall excellent performance is not a coincidence,but a consequence of the extensive sampling in phase space of ASSYST.Overall,this means that ASSYST will enable the largely autonomous generation of highquality DFT reference data and MLIPs.展开更多
The fundamental thermal limitation of pure copper impedes progress in high-power devices,which is becoming more critical with advances in power electronics.The Cu/diamond composite becomes a promising candidate for th...The fundamental thermal limitation of pure copper impedes progress in high-power devices,which is becoming more critical with advances in power electronics.The Cu/diamond composite becomes a promising candidate for thermal management due to its excellent theoretical thermal conductivity and customizable coefficient of thermal expansion(CTE).Actually,the thermal conductivity of Cu/diamond composite is much lower than its theoretical value,for which a key bottleneck is interfacial thermal transport at the Cu/diamond interface.However,many atomic-level microscopic mechanisms of heat transport at Cu/diamond interfaces remain poorly understood at present.Especially when different interlayer materials are involved,theoretical studies become extremely complex and challenging.In this work,a machine learning potential for comprehensive simulations of thermal transport at Cu/diamond interfaces has been successfully constructed.The effects of key factors,such as interlayer material,temperature,strain,and crystal orientation,on heat transport at Cu/diamond interfaces have been studied.Furthermore,the underlying mechanisms are thoroughly analyzed and discussed.Finally,the insightful strategies are proposed to optimize and enhance the thermal properties of Cu/diamond interfaces.These advancements can lay a foundation and pave the way for further investigations into interfacial thermal transport at Cu/diamond interfaces as well as in other structures containing interlayer materials.展开更多
Quantum nuclear effects and anharmonicity impact a wide range of functional materials and their properties.One of the most powerful techniques to model these effects is the Stochastic Self-Consistent Harmonic Approxim...Quantum nuclear effects and anharmonicity impact a wide range of functional materials and their properties.One of the most powerful techniques to model these effects is the Stochastic Self-Consistent Harmonic Approximation(SSCHA).Unfortunately,the SSCHA is extremely computationally expensive,prohibiting its routine use.We propose a protocol that pairs machine learning interatomic potentials,which can be tailored for the system at hand via active learning,with the SSCHA.Our method leverages an upscaling procedure that allows for the treatment of supercells of up to thousands of atoms with practically minimal computational effort.The protocol is applied to PdCuH_(x)(x=0−2)compounds,chosen because previous experimental studies have reported superconducting critical temperatures,Tcs,as high as 17 K at ambient pressures in an unknown hydrogenated PdCu phase.We identify a P4/mmm PdCuH_(2)structure,which is shown to be dynamically stable only upon the inclusion of quantum fluctuations,as being a key contributor to the measured superconductivity.For this system,our methodology is able to reduce the computational expense for the SSCHA calculations by~96%.The proposed protocol opens the door towards the routine inclusion of quantum nuclear motion and anharmonicity in materials discovery.展开更多
Understanding and accurately predicting hydrogen diffusion in materials is challenging due to the complex interactions between hydrogen defects and the crystal lattice.These interactions span large length and time sca...Understanding and accurately predicting hydrogen diffusion in materials is challenging due to the complex interactions between hydrogen defects and the crystal lattice.These interactions span large length and time scales,making them difficult to address with standard ab-initio techniques.This work addresses this challenge by employing accelerated machine learning(ML)molecular dynamics simulations through active learning.We conduct a comparative study of different ML-based interatomic potential schemes,including VASP,MACE,and CHGNet,utilizing various training strategies such as on-the-fly learning,pre-trained universal models,and fine-tuning.By considering different temperatures and concentration regimes,we obtain hydrogen diffusion coefficients and activation energy values which align remarkably well with experimental results,underlining the efficacy and accuracy of ML-assisted methodologies in the context of diffusive dynamics.Particularly,our procedure significantly reduces the computational effort associated with traditional transition state calculations or ad-hoc designed interatomic potentials.The results highlight the limitations of pre-trained universal solutions for defective materials and how they can be improved by fine-tuning.Specifically,fine-tuning the models on a database produced during on-the-fly training of VASP ML force-field allows the retrieval of DFT-level accuracy at a fraction of the computational cost.展开更多
Machine Learning(ML)potentials can be used for the construction of potential energy surfaces(PES)to avoid computationally expensive ab initio methods.However,many such applications still require a significant number o...Machine Learning(ML)potentials can be used for the construction of potential energy surfaces(PES)to avoid computationally expensive ab initio methods.However,many such applications still require a significant number of first-principles calculations to train the ML model,prior to use.Active learning methods can address this issue by performing these calculations and trainings“on-the-fly”(OTF),based on the ML model’s uncertainty estimation.Nevertheless,current active learning approaches suffer from problems in complex simulations were frequent retraining is required,since repeated training of a largeMLmodel increases training times substantially.This work presents a solution to this limitation by introducing the FALCON(Fast Active Learning for Computational ab initio mOlecular dyNamics)calculator.Instead of relying on a single largeMLmodel,FALCON clusters the training data into subsets of similar structures and distributes them across multiple smaller ML models.This approach significantly increases the efficiency of the OTF training,drastically reducing the computational cost of training-intensive simulations.The use of FALCON is demonstrated on various molecular dynamics(MD)simulations of bulk metals,metal clusters,cathode materials and water diffusion in a carbon nanotube.However,the FALCON calculator is highly flexible and could be easily adapted for various applications and different ML models.展开更多
LiFePO_(4) is a cathode material with good thermal stability,but low thermal conductivity is a critical problem.In this study,we employ a machine learning potential approach based on first-principles methods combined ...LiFePO_(4) is a cathode material with good thermal stability,but low thermal conductivity is a critical problem.In this study,we employ a machine learning potential approach based on first-principles methods combined with the Boltzmann transport theory to investigate the influence of Na substitution on the thermal conductivity of LiFePO_(4) and the impact of Li-ion de-embedding on the thermal conductivity of Li_(3/4)Na_(1/4)FePO_(4),with the aim of enhancing heat dissipation in Li-ion batteries.The results show a significant increase in thermal conductivity due to an increase in phonon group velocity and a decrease in phonon anharmonic scattering by Na substitution.In addition,the thermal conductivity increases significantly with decreasing Li-ion concentration due to the increase in phonon lifetime.Our work guides the improvement of the thermal conductivity of Li FePO_4,emphasizing the crucial roles of both substitution and Li-ion detachment/intercalation for the thermal management of electrochemical energy storage devices.展开更多
We present a full space inverse materials design(FSIMD)approach that fully automates the materials design for target physical properties without the need to provide the atomic composition,chemical stoichiometry,and cr...We present a full space inverse materials design(FSIMD)approach that fully automates the materials design for target physical properties without the need to provide the atomic composition,chemical stoichiometry,and crystal structure in advance.Here,we used density functional theory reference data to train a universal machine learning potential(UPot)and transfer learning to train a universal bulk modulus model(UBmod).Both UPot and UBmod were able to cover materials systems composed of any element among 42 elements.Interfaced with optimization algorithm and enhanced sampling,the FSIMD approach is applied to find the materials with the largest cohesive energy and the largest bulk modulus,respectively.NaCl-type ZrC was found to be the material with the largest cohesive energy.For bulk modulus,diamond was identified to have the largest value.The FSIMD approach is also applied to design materials with other multi-objective properties with accuracy limited principally by the amount,reliability,and diversity of the training data.The FSIMD approach provides a new way for inverse materials design with other functional properties for practical applications.展开更多
This Perspective explores the integration of machine learning potentials(MLPs)in the research of heterogeneous catalysis,focusing on their role in identifying in situ active sites and enhancing the understanding of ca...This Perspective explores the integration of machine learning potentials(MLPs)in the research of heterogeneous catalysis,focusing on their role in identifying in situ active sites and enhancing the understanding of catalytic processes.MLPs utilize extensive databases from high-throughput density functional theory(DFT)calculations to train models that predict atomic configurations,energies,and forces with near-DFT accuracy.These capabilities allow MLPs to handle significantly larger systems and extend simulation times beyond the limitations of traditional ab initio methods.Coupled with global optimization algorithms,MLPs enable systematic investigations across vast structural spaces,making substantial contributions to the modeling of catalyst surface structures under reactive conditions.The review aims to provide a broad introduction to recent advancements and practical guidance on employing MLPs and also showcases several exemplary cases of MLP-driven discoveries related to surface structure changes under reactive conditions and the nature of active sites in heterogeneous catalysis.The prevailing challenges faced by this approach are also discussed.展开更多
Nanoparticles,distinguished by their unique chemical and physical properties,have emerged as focal points within the realm of materials science.Traditional theoretical approaches for atomic simulations mainly include ...Nanoparticles,distinguished by their unique chemical and physical properties,have emerged as focal points within the realm of materials science.Traditional theoretical approaches for atomic simulations mainly include empirical force field and ab initio simulations,with the former offering efficiency but limited reliability,and the latter providing accuracy but restricted to systems of relatively small sizes.Herein,we propose a systematic strategy and automated workflow designed for collecting a diverse types of atomic local environments within a training dataset.This includes small nanoclusters,nanoparticles,as well as surface and bulk systems with periodic boundary conditions.The objective is to construct a machine learning potential tailored for pure metal nanoparticle simulations of varying sizes.Through rigorous validation,we have shown that our trained machine learning potential is capable of effectively driving molecular dynamics simulations of nanoparticles across a wide temperature range,especially within the nanoscale regime.Remarkably,this is achieved while preserving the accuracy typically associated with ab initio methods.展开更多
NJmat is a user-friendly,data-driven machine learning interface designed for materials design and analysis.The platform integrates advanced computational techniques,including natural language processing(NLP),large lan...NJmat is a user-friendly,data-driven machine learning interface designed for materials design and analysis.The platform integrates advanced computational techniques,including natural language processing(NLP),large language models(LLM),machine learning potentials(MLP),and graph neural networks(GNN),to facili-tate materials discovery.The platform has been applied in diverse materials research areas,including perovskite surface design,catalyst discovery,battery materials screening,structural alloy design,and molecular informatics.By automating feature selection,predictive modeling,and result interpretation,NJmat accelerates the development of high-performance materials across energy storage,conversion,and structural applications.Additionally,NJmat serves as an educational tool,allowing students and researchers to apply machine learning techniques in materials science with minimal coding expertise.Through automated feature extraction,genetic algorithms,and interpretable machine learning models,NJmat simplifies the workflow for materials informatics,bridging the gap between AI and experimental materials research.The latest version(available at https://figshare.com/articles/software/NJmatML/24607893(accessed on 01 January 2025))enhances its functionality by incorporating NJmatNLP,a module leveraging language models like MatBERT and those based on Word2Vec to support materials prediction tasks.By utilizing clustering and cosine similarity analysis with UMAP visualization,NJmat enables intuitive exploration of materials datasets.While NJmat primarily focuses on structure-property relationships and the discovery of novel chemistries,it can also assist in optimizing processing conditions when relevant parameters are included in the training data.By providing an accessible,integrated environment for machine learning-driven materials discovery,NJmat aligns with the objectives of the Materials Genome Initiative and promotes broader adoption of AI techniques in materials science.展开更多
Surface-supported clusters forming by aggregation of excessive adatoms could be the main defects of 2D materials after chemical vapor deposition.They will significantly impact the electronic/magnetic properties.Moreov...Surface-supported clusters forming by aggregation of excessive adatoms could be the main defects of 2D materials after chemical vapor deposition.They will significantly impact the electronic/magnetic properties.Moreover,surface supported atoms are also widely explored for high active and selecting catalysts.Severe deformation,even dipping into the surface,of these clusters can be expected because of the very active edge of clusters and strong interaction between supported clusters and surfaces.However,most models of these clusters are supposed to simply float on the top of the surface because ab initio simulations cannot afford the complex reconstructions.Here,we develop an accurate graph neural network machine learning potential(MLP)from ab initio data by active learning architecture through fine-tuning pre-trained models,and then employ the MLP into Monte Carlo to explore the structural evolutions of Mo and S clusters(1-8 atoms)on perfect and various defective MoS2 monolayers.Interestingly,Mo clusters can always sink and embed themselves into MoS2 layers.In contrast,S clusters float on perfect surfaces.On the defective surface,a few S atoms will fill the vacancy and rest S clusters float on the top.Such significant structural reconstructions should be carefully taken into account.展开更多
Zr-based amorphous alloys have attracted extensive attention because of their large glassy formation ability, wide supercooled liquid region, high elasticity, and unique mechanical strength induced by their icosahedra...Zr-based amorphous alloys have attracted extensive attention because of their large glassy formation ability, wide supercooled liquid region, high elasticity, and unique mechanical strength induced by their icosahedral local structures.To determine the microstructures of Zr–Cu clusters, the stable and metastable geometry of Zr_(n)Cu(n=2–12) clusters are screened out via the CALYPSO method using machine-learning potentials, and then the electronic structures are investigated using density functional theory. The results show that the Zr_(n)Cu(n ≥ 3) clusters possess three-dimensional geometries, Zr_(n)Cu(n≥9) possess cage-like geometries, and the Zr_(12)Cu cluster has icosahedral geometry. The binding energy per atom gradually gets enlarged with the increase in the size of the clusters, and Zr_(n)Cu(n=5,7,9,12) have relatively better stability than their neighbors. The magnetic moment of most Zr_(n)Cu clusters is just 1μB, and the main components of the highest occupied molecular orbitals(HOMOs) in the Zr_(12)Cu cluster come from the Zr-d state. There are hardly any localized two-center bonds, and there are about 20 σ-type delocalized three-center bonds.展开更多
High entropy materials(HEMs, e.g. high entropy alloys, high entropy ceramics) have gained increasing interests due to the possibility that they can provide challenge properties unattainable by traditional materials. T...High entropy materials(HEMs, e.g. high entropy alloys, high entropy ceramics) have gained increasing interests due to the possibility that they can provide challenge properties unattainable by traditional materials. Though a large number of HEMs have emerged, there is still in lack of theoretical predictions and simulations on HEMs, which is probably caused by the chemical complexity of HEMs. In this work,we demonstrate that the machine learning potentials developed in recent years can overcome the complexity of HEMs, and serve as powerful theoretical tools to simulate HEMs. A deep learning potential(DLP) for high entropy(Zr(0.2) Hf(0.2) Ti(0.2) Nb(0.2) Ta(0.2))C is fitted with the prediction error in energy and force being 9.4 me V/atom and 217 me V/?, respectively. The reliability and generality of the DLP are affirmed,since it can accurately predict lattice parameters and elastic constants of mono-phase carbides TMC(TM = Ti, Zr, Hf, Nb and Ta). Lattice constants(increase from 4.5707 ? to 4.6727 ?), thermal expansion coefficients(increase from 7.85×10-6 K^(-1) to 10.58×10-6 K^(-1)), phonon thermal conductivities(decrease from 2.02 W·m-1·K^(-1) to 0.95 W·m-1·K^(-1)), and elastic properties of high entropy(Zr(0.2) Hf(0.2) Ti(0.2) Nb(0.2) Ta(0.2))C in temperature ranging from 0°C to 2400°C are predicted by molecular dynamics simulations. The predicted room temperature properties agree well with experimental measurements, indicating the high accuracy of the DLP. With introducing of machine learning potentials, many problems that are intractable by traditional methods can be handled now. It is hopeful that deep insight into HEMs can be obtained in the future by such powerful methods.展开更多
High entropy diborides are new categories of ultra-high temperature ceramics,which are believed promising candidates for applications in hypersonic vehicles.However,knowledge on high temperature thermal and mechanical...High entropy diborides are new categories of ultra-high temperature ceramics,which are believed promising candidates for applications in hypersonic vehicles.However,knowledge on high temperature thermal and mechanical properties of high entropy diborides is still lacking unit now.In this work,variations of thermal and elastic properties of high entropy(Ti_(0.2)Zr_(0.2)Hf_(0.2)Nb_(0.2)Ta_(0.2))B_(2) with respect to temperature were predicted by molecular dynamics simulations.Firstly,a deep learning potential for Ti-Zr-Hf-Nb-Ta-B diboride system was fitted with its prediction error in energy and force respectively being 9.2 meV/atom and 208 meV/A,in comparison with first-principles calculations.Then,temperature dependent lattice constants,anisotropic thermal expansions,anisotropic phonon thermal conductivities,and elastic properties of high entropy(Ti_(0.2)Zr_(0.2)Hf_(0.2)Nb_(0.2)Ta_(0.2))B_(2) from 0℃to 2400℃were evaluated,where the predicted room temperature values agree well with experimental measurements.In addition,intrinsic lattice distortions of(Ti_(0.2)Zr_(0.2)Hf_(0.2)Nb_(0.2)Ta_(0.2))B_(2) were analyzed by displacements of atoms from their ideal positions,which are in an order of 10^(-3) A and one order of magnitude smaller than those in(Ti_(0.2)Zr_(0.2)Hf_(0.2)Nb_(0.2)Ta_(0.2))C.It indicates that lattice distortions in(Ti_(0.2)Zr_(0.2)Hf_(0.2)Nb_(0.2)Ta_(0.2))B_(2) is not so severe as expected.With the new paradigm of machine learning potential,deep insight into high entropy materials can be achieved in the future,since the chemical and structural complexly in high entropy materials can be well handled by machine learning potential.展开更多
Recursively embedded atom neural network(REANN)is a general-purpose atomistic machine learning software package for representing potential energy and other physical properties.The original REANN 1.0 architecture is a ...Recursively embedded atom neural network(REANN)is a general-purpose atomistic machine learning software package for representing potential energy and other physical properties.The original REANN 1.0 architecture is a physically inspired invariant message passing neural network,which was designed for systems with a limited number of elements.It is efficient but hardly transferable to more complex multi-element systems.In this work,we release REANN 2.0 aimed at multi-element systems and universal potentials,which integrates element embedding and equivariant representation.Compared to the first version,REANN 2.0 demonstrates enhanced ele-ment transferability and higher accuracy across various periodic systems with higher efficiency.Built upon this framework,a pre-trained REANN-MPtrj model without fine-tuning accurately predicts the lithium-ion diffusion dynamics in a benchmark solid-state electrolyte Li_(3)YCl_(6).We hope this open-source software package will facilitate the development of computationally efficient universal potentials in the future.展开更多
As a representative of wide-bandgap semiconductors,wurtzite gallium nitride(GaN)has been widely utilized in highpower devices due to its high breakdown voltage and low specific on-resistance.Accurate prediction of wur...As a representative of wide-bandgap semiconductors,wurtzite gallium nitride(GaN)has been widely utilized in highpower devices due to its high breakdown voltage and low specific on-resistance.Accurate prediction of wurtzite GaN’s thermal conductivity is a prerequisite for designing effective thermal management systems for electronic applications.Machine learning-driven molecular dynamics simulation offers a promising approach to predicting the thermal conductivity of large-scale systems without requiring predefined parameters.However,these methods often underestimate the thermal conductivity of materials with inherently high thermal conductivity due to the large predicted force error compared with first-principles calculations,posing a critical challenge for their broader application.In this study,we successfully developed a neuroevolution potential for wurtzite GaN and accurately predicted its thermal conductivity,259±6 W/(m·K)at room temperature,achieving excellent agreement with reported experimental measurements.The hyperparameters of the neuroevolution potential(NEP)were optimized based on a systematic analysis of reproduced energy and force,structural features,and computational efficiency.Furthermore,a force error correction method was implemented,effectively reducing the error caused by the additional force noise in the Langevin thermostat by extrapolating to the zero-force error limit.This study provides valuable insights and holds significant implications for advancing efficient thermal management technologies in wide-bandgap semiconductor devices.展开更多
The investigation of thermal transport is crucial to the thermal management of modern electronic devices.To obtain the thermal conductivity through solution of the Boltzmann transport equation,calculation of the anhar...The investigation of thermal transport is crucial to the thermal management of modern electronic devices.To obtain the thermal conductivity through solution of the Boltzmann transport equation,calculation of the anharmonic interatomic force constants has a high computational cost based on the current method of single-point density functional theory force calculation.The recent suggested machine learning interatomic potentials(MLIPs)method can avoid these huge computational demands.In this work,we study the thermal conductivity of two-dimensional MoS_(2)-like hexagonal boron dichalcogenides(H-B_(2)VI_(2);V I=S,Se,Te)with a combination of MLIPs and the phonon Boltzmann transport equation.The room-temperature thermal conductivity of H-B_(2)S_(2)can reach up to 336 W·m^(-1)·K^(-1),obviously larger than that of H-B_(2)Se_(2)and H-B_(2)Te_(2).This is mainly due to the difference in phonon group velocity.By substituting the different chalcogen elements in the second sublayer,H-B_(2)VIV I′have lower thermal conductivity than H-B_(2)VI_(2).The room-temperature thermal conductivity of B2STe is only 11%of that of H-B_(2)S_(2).This can be explained by comparing phonon group velocity and phonon relaxation time.The MLIP method is proved to be an efficient method for studying the thermal conductivity of materials,and H-B_(2)S_(2)-based nanodevices have excellent thermal conduction.展开更多
基金Project supported by the A*STAR Computational Resource Centre through the use of its high-performance computing facilitiesfinancial support from the China Scholarship Council (Grant No.202206120136)。
文摘GeTe has attracted extensive research interest for thermoelectric applications.In this paper,we first train a neuroevolution potential(NEP)based on a dataset constructed by ab initio molecular dynamics,with the Gaussian approximation potential(GAP)as a reference.The phonon density of states is then calculated by two machine learning potentials and compared with density functional theory results,with the GAP potential having higher accuracy.Next,the thermal conductivity of a GeTe crystal at 300 K is calculated by the equilibrium molecular dynamics method using both machine learning potentials,and both of them are in good agreement with the experimental results;however,the calculation speed when using the NEP potential is about 500 times faster than when using the GAP potential.Finally,the lattice thermal conductivity in the range of 300 K-600 K is calculated using the NEP potential.The lattice thermal conductivity decreases as the temperature increases due to the phonon anharmonic effect.This study provides a theoretical tool for the study of the thermal conductivity of GeTe.
基金Computational support was provided by Supercomputing Center in USTC and National Supercomputing Center in Tianjinthe National Key Research and Development Program of China(Grant Nos.2017YFA0204904 and 2019YFA0210004)。
文摘A machine learning(ML)potential for Au clusters is developed through training on a dataset including several different sized clusters.This ML potential accurately covers the whole configuration space of Au clusters in a broad size range,thus expressing a good performance in search of their global minimum energy structures.Based on our potential,the low-lying structures of 17 different sized Au clusters are identified,which shows that small sized Au clusters tend to form planar structures while large ones are more likely to be stereo,revealing the critical size for the two-dimensional(2D)to three-dimensional(3D)structural transition.Our calculations demonstrate that ML is indeed powerful in describing the interaction of Au atoms and provides a new paradigm on accelerating the search of structures.
基金supported by the MathWorks Ignition Fund,MathWorks Engineering Fellowship Fund,and the Portuguese Foundation for International Cooperation in Science,Technology and Higher Education in the MIT—Portugal Programsupported by the Research Support Committee Funds from the School of Engineering at the Massachusetts Institute of Technology+1 种基金This work used the Expanse supercomputer at the San Diego Supercomputer Center through allocation MAT210005 from the Advanced Cyber Infrastructure Coordination Ecosystem:Services&Support(ACCESS)program,which is supported by National Science Foundation grants#2138259,#2138286,#2138307,#2137603,and#2138296the Extreme Science and Engineering Discovery Environment(XSEDE),which was supported by National Science Foundation grant number#1548562.
文摘Chemical short-range order(SRO)affects the distribution of elements throughout the solid-solution phase of metallic alloys,thereby modifying the background against which microstructural evolution occurs.Investigating such chemistry–microstructure relationships requires atomistic models that act at the appropriate length scales while capturing the intricacies of chemical bonds leading to SRO.Here,we consider various approaches for the construction of training data sets for machine learning potentials(MLPs)for CrCoNi and evaluate their performance in capturing SRO and its effects on materials quantities of relevance for mechanical properties,such as stacking-fault energy and phase stability.It is demonstrated that energy accuracy on test sets often does not correlate with accuracy in capturing material properties,which is fundamental in enabling large-scale atomistic simulations of metallic alloys with high physical fidelity.Based on this analysis,we systematically derive design principles for the rational construction of MLPs that capture SRO in the crystal and liquid phases of alloys.
基金funding from the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) through the Collaborative Research Center 1394 (SFB 1394, No. 409476157) and Project No. 405621160MP would like to thank Prince Matthews for setting up the hcp grain boundaries, Sarath Menon for providing support for CALPHY34Bengt Hallstedt for providing plots of the Mg/Ca and Al/Ca phase diagrams from his assessments, Chad Sinclair together with SFB 1394 for funding a research stay at UBC Vancouver where part of this work was conducted, as well Mira Todorova and Ali Tehranchi for fruitful discussions and Ralf Drautz for critical reading of the manuscript.
文摘We propose a strategy for generating unbiased and systematically extendable training data for machine learning interatomic potentials(MLIP)for multicomponent alloys,called Automated Small SYmmetric Structure Training or ASSYST.Based on exploring the full space of random crystal structures with space groups,it facilitates the construction of training sets for MLIPs in an automatic way without prior knowledge of the material in question.The advantages of this approach are that only cells consisting of few atoms(≈10)are needed for the DFT training set,and the size and completeness of the data set can be systematically controlled with very few parameters.We validate that potentials fitted this way can accurately describe a wide range of binary and ternary phases,random alloys,as well as point and extended defects,that have not been part of the training set.Finally,we estimate the binary phase diagrams with good experimental agreement.We demonstrate that the overall excellent performance is not a coincidence,but a consequence of the extensive sampling in phase space of ASSYST.Overall,this means that ASSYST will enable the largely autonomous generation of highquality DFT reference data and MLIPs.
基金funded by the National Natural Science Foundation of China(Grant Nos.92473102,52202045,62004141)the Shenzhen Science and Technology Program(Grant No.JCYJ20240813175906008)the Open Fund of Hubei Key Laboratory of Electronic Manufacturing and Packaging Integration(Wuhan University)(Grant No.EMPI2025007).The numerical calculations in this paper have been done on the supercomputing system in the Supercomputing Center of Wuhan University.
文摘The fundamental thermal limitation of pure copper impedes progress in high-power devices,which is becoming more critical with advances in power electronics.The Cu/diamond composite becomes a promising candidate for thermal management due to its excellent theoretical thermal conductivity and customizable coefficient of thermal expansion(CTE).Actually,the thermal conductivity of Cu/diamond composite is much lower than its theoretical value,for which a key bottleneck is interfacial thermal transport at the Cu/diamond interface.However,many atomic-level microscopic mechanisms of heat transport at Cu/diamond interfaces remain poorly understood at present.Especially when different interlayer materials are involved,theoretical studies become extremely complex and challenging.In this work,a machine learning potential for comprehensive simulations of thermal transport at Cu/diamond interfaces has been successfully constructed.The effects of key factors,such as interlayer material,temperature,strain,and crystal orientation,on heat transport at Cu/diamond interfaces have been studied.Furthermore,the underlying mechanisms are thoroughly analyzed and discussed.Finally,the insightful strategies are proposed to optimize and enhance the thermal properties of Cu/diamond interfaces.These advancements can lay a foundation and pave the way for further investigations into interfacial thermal transport at Cu/diamond interfaces as well as in other structures containing interlayer materials.
基金Funding for this research is provided by the National Science Foundation,under award DMR-2136038Calculations were performed at the Center for Computational Research at SUNY Buffalo(http://hdl.handle.net/10477/79221).
文摘Quantum nuclear effects and anharmonicity impact a wide range of functional materials and their properties.One of the most powerful techniques to model these effects is the Stochastic Self-Consistent Harmonic Approximation(SSCHA).Unfortunately,the SSCHA is extremely computationally expensive,prohibiting its routine use.We propose a protocol that pairs machine learning interatomic potentials,which can be tailored for the system at hand via active learning,with the SSCHA.Our method leverages an upscaling procedure that allows for the treatment of supercells of up to thousands of atoms with practically minimal computational effort.The protocol is applied to PdCuH_(x)(x=0−2)compounds,chosen because previous experimental studies have reported superconducting critical temperatures,Tcs,as high as 17 K at ambient pressures in an unknown hydrogenated PdCu phase.We identify a P4/mmm PdCuH_(2)structure,which is shown to be dynamically stable only upon the inclusion of quantum fluctuations,as being a key contributor to the measured superconductivity.For this system,our methodology is able to reduce the computational expense for the SSCHA calculations by~96%.The proposed protocol opens the door towards the routine inclusion of quantum nuclear motion and anharmonicity in materials discovery.
基金the“Doctoral College Advanced Functional Materials-Hierarchical Design of Hybrid Systems DOC 85 doc.funds”funded by the Austrian Science Fund(FWF)and by the Vienna Doctoral School in Physics(VDSP),For Open Access purposes,the author has applied a CC BY public copyright license to any author accepted manuscript version arising from this submission.D.M.and S.P.were supported by the European Union Horizon 2020 research and innovation program under Grant Agreement No.857470the European Regional Development Fund under the program of the Foundation for Polish Science International Research Agenda PLUS,grant No.MAB PLUS/2018/8+2 种基金the initiative of the Ministry of Science and Higher Education’Support for the activities of Centers of Excellence established in Poland under the Horizon 2020 program’under agreement No.MEiN/2023/DIR/3795L.P.and C.F.acknowledge the National Recovery and Resilience Plan(NRRP),Mission 4 Component 2 Investment 1.3-Project NEST(Network 4 Energy Sustainable Transition)of Ministero dell’Universitáe della Ricerca(MUR),funded by the European Union-NextGenerationEUL.L.and C.F.acknowledge the NRRP,CN-HPC grant no.(CUP)J33C22001170001,SPOKE 7,of MUR,funded by the European Union-NextGenerationEU.The computational results were obtained using the Vienna Scientific Cluster(VSC)and the LEONARDO cluster.We acknowledge access to LEONARDO at CINECA,Italy,via an AURELEO(Austrian Users at LEONARDO supercomputer)project.
文摘Understanding and accurately predicting hydrogen diffusion in materials is challenging due to the complex interactions between hydrogen defects and the crystal lattice.These interactions span large length and time scales,making them difficult to address with standard ab-initio techniques.This work addresses this challenge by employing accelerated machine learning(ML)molecular dynamics simulations through active learning.We conduct a comparative study of different ML-based interatomic potential schemes,including VASP,MACE,and CHGNet,utilizing various training strategies such as on-the-fly learning,pre-trained universal models,and fine-tuning.By considering different temperatures and concentration regimes,we obtain hydrogen diffusion coefficients and activation energy values which align remarkably well with experimental results,underlining the efficacy and accuracy of ML-assisted methodologies in the context of diffusive dynamics.Particularly,our procedure significantly reduces the computational effort associated with traditional transition state calculations or ad-hoc designed interatomic potentials.The results highlight the limitations of pre-trained universal solutions for defective materials and how they can be improved by fine-tuning.Specifically,fine-tuning the models on a database produced during on-the-fly training of VASP ML force-field allows the retrieval of DFT-level accuracy at a fraction of the computational cost.
基金funding by the Central Research Development Fund of the University of Bremen.
文摘Machine Learning(ML)potentials can be used for the construction of potential energy surfaces(PES)to avoid computationally expensive ab initio methods.However,many such applications still require a significant number of first-principles calculations to train the ML model,prior to use.Active learning methods can address this issue by performing these calculations and trainings“on-the-fly”(OTF),based on the ML model’s uncertainty estimation.Nevertheless,current active learning approaches suffer from problems in complex simulations were frequent retraining is required,since repeated training of a largeMLmodel increases training times substantially.This work presents a solution to this limitation by introducing the FALCON(Fast Active Learning for Computational ab initio mOlecular dyNamics)calculator.Instead of relying on a single largeMLmodel,FALCON clusters the training data into subsets of similar structures and distributes them across multiple smaller ML models.This approach significantly increases the efficiency of the OTF training,drastically reducing the computational cost of training-intensive simulations.The use of FALCON is demonstrated on various molecular dynamics(MD)simulations of bulk metals,metal clusters,cathode materials and water diffusion in a carbon nanotube.However,the FALCON calculator is highly flexible and could be easily adapted for various applications and different ML models.
基金supported by the National Natural Science Foundation of China(Grant No.12074115)the Science and Technology Innovation Program of Hunan Province(Grant No.2023RC3176)。
文摘LiFePO_(4) is a cathode material with good thermal stability,but low thermal conductivity is a critical problem.In this study,we employ a machine learning potential approach based on first-principles methods combined with the Boltzmann transport theory to investigate the influence of Na substitution on the thermal conductivity of LiFePO_(4) and the impact of Li-ion de-embedding on the thermal conductivity of Li_(3/4)Na_(1/4)FePO_(4),with the aim of enhancing heat dissipation in Li-ion batteries.The results show a significant increase in thermal conductivity due to an increase in phonon group velocity and a decrease in phonon anharmonic scattering by Na substitution.In addition,the thermal conductivity increases significantly with decreasing Li-ion concentration due to the increase in phonon lifetime.Our work guides the improvement of the thermal conductivity of Li FePO_4,emphasizing the crucial roles of both substitution and Li-ion detachment/intercalation for the thermal management of electrochemical energy storage devices.
基金funding support by the National Key Research and Development Program of China(2020YFB1506400)the National Natural Science Foundation of China(11974257 and 12188101)+1 种基金Jiangsu Distinguished Young Talent Funding(BK20200003)Soochow Municipal Laboratory for low carbon technologies and industries.
文摘We present a full space inverse materials design(FSIMD)approach that fully automates the materials design for target physical properties without the need to provide the atomic composition,chemical stoichiometry,and crystal structure in advance.Here,we used density functional theory reference data to train a universal machine learning potential(UPot)and transfer learning to train a universal bulk modulus model(UBmod).Both UPot and UBmod were able to cover materials systems composed of any element among 42 elements.Interfaced with optimization algorithm and enhanced sampling,the FSIMD approach is applied to find the materials with the largest cohesive energy and the largest bulk modulus,respectively.NaCl-type ZrC was found to be the material with the largest cohesive energy.For bulk modulus,diamond was identified to have the largest value.The FSIMD approach is also applied to design materials with other multi-objective properties with accuracy limited principally by the amount,reliability,and diversity of the training data.The FSIMD approach provides a new way for inverse materials design with other functional properties for practical applications.
基金the NKRDPC(2021YFA1500700)and NSFC(92045303).X.C.is grateful for financial support from ShanghaiTech University.
文摘This Perspective explores the integration of machine learning potentials(MLPs)in the research of heterogeneous catalysis,focusing on their role in identifying in situ active sites and enhancing the understanding of catalytic processes.MLPs utilize extensive databases from high-throughput density functional theory(DFT)calculations to train models that predict atomic configurations,energies,and forces with near-DFT accuracy.These capabilities allow MLPs to handle significantly larger systems and extend simulation times beyond the limitations of traditional ab initio methods.Coupled with global optimization algorithms,MLPs enable systematic investigations across vast structural spaces,making substantial contributions to the modeling of catalyst surface structures under reactive conditions.The review aims to provide a broad introduction to recent advancements and practical guidance on employing MLPs and also showcases several exemplary cases of MLP-driven discoveries related to surface structure changes under reactive conditions and the nature of active sites in heterogeneous catalysis.The prevailing challenges faced by this approach are also discussed.
基金supported by the National Science Fund for Distinguished Young Scholars(22225302)the National Natural Science Foundation of China(92161113,21991151,21991150 and 22021001)+2 种基金the Fundamental Research Funds for the Central Universities(20720220008,20720220009 and 20720220010)the Laboratory of AI for Electrochemistry(AI4EC)IKKEM(RD2023100101 and RD2022070501)
文摘Nanoparticles,distinguished by their unique chemical and physical properties,have emerged as focal points within the realm of materials science.Traditional theoretical approaches for atomic simulations mainly include empirical force field and ab initio simulations,with the former offering efficiency but limited reliability,and the latter providing accuracy but restricted to systems of relatively small sizes.Herein,we propose a systematic strategy and automated workflow designed for collecting a diverse types of atomic local environments within a training dataset.This includes small nanoclusters,nanoparticles,as well as surface and bulk systems with periodic boundary conditions.The objective is to construct a machine learning potential tailored for pure metal nanoparticle simulations of varying sizes.Through rigorous validation,we have shown that our trained machine learning potential is capable of effectively driving molecular dynamics simulations of nanoparticles across a wide temperature range,especially within the nanoscale regime.Remarkably,this is achieved while preserving the accuracy typically associated with ab initio methods.
基金supported by the Jiangsu Provincial Science and Technology Project Basic Research Program(Natural Science Foundation of Jiangsu Province)(No.BK20211283).
文摘NJmat is a user-friendly,data-driven machine learning interface designed for materials design and analysis.The platform integrates advanced computational techniques,including natural language processing(NLP),large language models(LLM),machine learning potentials(MLP),and graph neural networks(GNN),to facili-tate materials discovery.The platform has been applied in diverse materials research areas,including perovskite surface design,catalyst discovery,battery materials screening,structural alloy design,and molecular informatics.By automating feature selection,predictive modeling,and result interpretation,NJmat accelerates the development of high-performance materials across energy storage,conversion,and structural applications.Additionally,NJmat serves as an educational tool,allowing students and researchers to apply machine learning techniques in materials science with minimal coding expertise.Through automated feature extraction,genetic algorithms,and interpretable machine learning models,NJmat simplifies the workflow for materials informatics,bridging the gap between AI and experimental materials research.The latest version(available at https://figshare.com/articles/software/NJmatML/24607893(accessed on 01 January 2025))enhances its functionality by incorporating NJmatNLP,a module leveraging language models like MatBERT and those based on Word2Vec to support materials prediction tasks.By utilizing clustering and cosine similarity analysis with UMAP visualization,NJmat enables intuitive exploration of materials datasets.While NJmat primarily focuses on structure-property relationships and the discovery of novel chemistries,it can also assist in optimizing processing conditions when relevant parameters are included in the training data.By providing an accessible,integrated environment for machine learning-driven materials discovery,NJmat aligns with the objectives of the Materials Genome Initiative and promotes broader adoption of AI techniques in materials science.
基金supported by the National Natural Science Foundation of China(Grant No.12374253,12074053,12004064)J.G.thanks the Foreign talents project(G2022127004L),The authors also acknowledge computer support from the Shanghai Supercomputer Center,the DUT Supercomputing Center,and the Tianhe supercomputer of Tianjin Center.
文摘Surface-supported clusters forming by aggregation of excessive adatoms could be the main defects of 2D materials after chemical vapor deposition.They will significantly impact the electronic/magnetic properties.Moreover,surface supported atoms are also widely explored for high active and selecting catalysts.Severe deformation,even dipping into the surface,of these clusters can be expected because of the very active edge of clusters and strong interaction between supported clusters and surfaces.However,most models of these clusters are supposed to simply float on the top of the surface because ab initio simulations cannot afford the complex reconstructions.Here,we develop an accurate graph neural network machine learning potential(MLP)from ab initio data by active learning architecture through fine-tuning pre-trained models,and then employ the MLP into Monte Carlo to explore the structural evolutions of Mo and S clusters(1-8 atoms)on perfect and various defective MoS2 monolayers.Interestingly,Mo clusters can always sink and embed themselves into MoS2 layers.In contrast,S clusters float on perfect surfaces.On the defective surface,a few S atoms will fill the vacancy and rest S clusters float on the top.Such significant structural reconstructions should be carefully taken into account.
基金Project supported by the National Natural Science Foundation of China (Grant Nos.11864040,11964037,and 11664038)。
文摘Zr-based amorphous alloys have attracted extensive attention because of their large glassy formation ability, wide supercooled liquid region, high elasticity, and unique mechanical strength induced by their icosahedral local structures.To determine the microstructures of Zr–Cu clusters, the stable and metastable geometry of Zr_(n)Cu(n=2–12) clusters are screened out via the CALYPSO method using machine-learning potentials, and then the electronic structures are investigated using density functional theory. The results show that the Zr_(n)Cu(n ≥ 3) clusters possess three-dimensional geometries, Zr_(n)Cu(n≥9) possess cage-like geometries, and the Zr_(12)Cu cluster has icosahedral geometry. The binding energy per atom gradually gets enlarged with the increase in the size of the clusters, and Zr_(n)Cu(n=5,7,9,12) have relatively better stability than their neighbors. The magnetic moment of most Zr_(n)Cu clusters is just 1μB, and the main components of the highest occupied molecular orbitals(HOMOs) in the Zr_(12)Cu cluster come from the Zr-d state. There are hardly any localized two-center bonds, and there are about 20 σ-type delocalized three-center bonds.
基金supported financially by the National Natural Science Foundation of China(Nos.51672064 and No.U1435206)。
文摘High entropy materials(HEMs, e.g. high entropy alloys, high entropy ceramics) have gained increasing interests due to the possibility that they can provide challenge properties unattainable by traditional materials. Though a large number of HEMs have emerged, there is still in lack of theoretical predictions and simulations on HEMs, which is probably caused by the chemical complexity of HEMs. In this work,we demonstrate that the machine learning potentials developed in recent years can overcome the complexity of HEMs, and serve as powerful theoretical tools to simulate HEMs. A deep learning potential(DLP) for high entropy(Zr(0.2) Hf(0.2) Ti(0.2) Nb(0.2) Ta(0.2))C is fitted with the prediction error in energy and force being 9.4 me V/atom and 217 me V/?, respectively. The reliability and generality of the DLP are affirmed,since it can accurately predict lattice parameters and elastic constants of mono-phase carbides TMC(TM = Ti, Zr, Hf, Nb and Ta). Lattice constants(increase from 4.5707 ? to 4.6727 ?), thermal expansion coefficients(increase from 7.85×10-6 K^(-1) to 10.58×10-6 K^(-1)), phonon thermal conductivities(decrease from 2.02 W·m-1·K^(-1) to 0.95 W·m-1·K^(-1)), and elastic properties of high entropy(Zr(0.2) Hf(0.2) Ti(0.2) Nb(0.2) Ta(0.2))C in temperature ranging from 0°C to 2400°C are predicted by molecular dynamics simulations. The predicted room temperature properties agree well with experimental measurements, indicating the high accuracy of the DLP. With introducing of machine learning potentials, many problems that are intractable by traditional methods can be handled now. It is hopeful that deep insight into HEMs can be obtained in the future by such powerful methods.
基金supported by Natural Sciences Foundation of China under Grant No.51972089 and No.51672064。
文摘High entropy diborides are new categories of ultra-high temperature ceramics,which are believed promising candidates for applications in hypersonic vehicles.However,knowledge on high temperature thermal and mechanical properties of high entropy diborides is still lacking unit now.In this work,variations of thermal and elastic properties of high entropy(Ti_(0.2)Zr_(0.2)Hf_(0.2)Nb_(0.2)Ta_(0.2))B_(2) with respect to temperature were predicted by molecular dynamics simulations.Firstly,a deep learning potential for Ti-Zr-Hf-Nb-Ta-B diboride system was fitted with its prediction error in energy and force respectively being 9.2 meV/atom and 208 meV/A,in comparison with first-principles calculations.Then,temperature dependent lattice constants,anisotropic thermal expansions,anisotropic phonon thermal conductivities,and elastic properties of high entropy(Ti_(0.2)Zr_(0.2)Hf_(0.2)Nb_(0.2)Ta_(0.2))B_(2) from 0℃to 2400℃were evaluated,where the predicted room temperature values agree well with experimental measurements.In addition,intrinsic lattice distortions of(Ti_(0.2)Zr_(0.2)Hf_(0.2)Nb_(0.2)Ta_(0.2))B_(2) were analyzed by displacements of atoms from their ideal positions,which are in an order of 10^(-3) A and one order of magnitude smaller than those in(Ti_(0.2)Zr_(0.2)Hf_(0.2)Nb_(0.2)Ta_(0.2))C.It indicates that lattice distortions in(Ti_(0.2)Zr_(0.2)Hf_(0.2)Nb_(0.2)Ta_(0.2))B_(2) is not so severe as expected.With the new paradigm of machine learning potential,deep insight into high entropy materials can be achieved in the future,since the chemical and structural complexly in high entropy materials can be well handled by machine learning potential.
基金the support by the Strategic Priority Research Program of the Chinese Academy of Sciences(XDB0450101)the National Natural Science Foundation of China(Nos.22325304,22221003 and 22033007)。
文摘Recursively embedded atom neural network(REANN)is a general-purpose atomistic machine learning software package for representing potential energy and other physical properties.The original REANN 1.0 architecture is a physically inspired invariant message passing neural network,which was designed for systems with a limited number of elements.It is efficient but hardly transferable to more complex multi-element systems.In this work,we release REANN 2.0 aimed at multi-element systems and universal potentials,which integrates element embedding and equivariant representation.Compared to the first version,REANN 2.0 demonstrates enhanced ele-ment transferability and higher accuracy across various periodic systems with higher efficiency.Built upon this framework,a pre-trained REANN-MPtrj model without fine-tuning accurately predicts the lithium-ion diffusion dynamics in a benchmark solid-state electrolyte Li_(3)YCl_(6).We hope this open-source software package will facilitate the development of computationally efficient universal potentials in the future.
基金supported by the National Natural Science Foundation of China(Grant Nos.52376063 and 52306116)the Hebei Key Laboratory of Low Carbon and High Efficiency Power Generation Technology Prevention Fund(Grant No.2022-K03)the China Postdoctoral Science Foundation(Grant No.2023MD744223).
文摘As a representative of wide-bandgap semiconductors,wurtzite gallium nitride(GaN)has been widely utilized in highpower devices due to its high breakdown voltage and low specific on-resistance.Accurate prediction of wurtzite GaN’s thermal conductivity is a prerequisite for designing effective thermal management systems for electronic applications.Machine learning-driven molecular dynamics simulation offers a promising approach to predicting the thermal conductivity of large-scale systems without requiring predefined parameters.However,these methods often underestimate the thermal conductivity of materials with inherently high thermal conductivity due to the large predicted force error compared with first-principles calculations,posing a critical challenge for their broader application.In this study,we successfully developed a neuroevolution potential for wurtzite GaN and accurately predicted its thermal conductivity,259±6 W/(m·K)at room temperature,achieving excellent agreement with reported experimental measurements.The hyperparameters of the neuroevolution potential(NEP)were optimized based on a systematic analysis of reproduced energy and force,structural features,and computational efficiency.Furthermore,a force error correction method was implemented,effectively reducing the error caused by the additional force noise in the Langevin thermostat by extrapolating to the zero-force error limit.This study provides valuable insights and holds significant implications for advancing efficient thermal management technologies in wide-bandgap semiconductor devices.
基金Scientific and Technological Research of Chongqing Municipal Education Commission(Grant No.KJZD-K202100602)the funding of Institute for Advanced Sciences of Chongqing University of Posts and Telecommunications(Grant No.E011A2022326)。
文摘The investigation of thermal transport is crucial to the thermal management of modern electronic devices.To obtain the thermal conductivity through solution of the Boltzmann transport equation,calculation of the anharmonic interatomic force constants has a high computational cost based on the current method of single-point density functional theory force calculation.The recent suggested machine learning interatomic potentials(MLIPs)method can avoid these huge computational demands.In this work,we study the thermal conductivity of two-dimensional MoS_(2)-like hexagonal boron dichalcogenides(H-B_(2)VI_(2);V I=S,Se,Te)with a combination of MLIPs and the phonon Boltzmann transport equation.The room-temperature thermal conductivity of H-B_(2)S_(2)can reach up to 336 W·m^(-1)·K^(-1),obviously larger than that of H-B_(2)Se_(2)and H-B_(2)Te_(2).This is mainly due to the difference in phonon group velocity.By substituting the different chalcogen elements in the second sublayer,H-B_(2)VIV I′have lower thermal conductivity than H-B_(2)VI_(2).The room-temperature thermal conductivity of B2STe is only 11%of that of H-B_(2)S_(2).This can be explained by comparing phonon group velocity and phonon relaxation time.The MLIP method is proved to be an efficient method for studying the thermal conductivity of materials,and H-B_(2)S_(2)-based nanodevices have excellent thermal conduction.
