LASP(large-scale atomistic simulation with neural network potential)software developed by our group since 2018 is a powerful platform(www.lasphub.com)for performing atomic simulation of complex materials.The software ...LASP(large-scale atomistic simulation with neural network potential)software developed by our group since 2018 is a powerful platform(www.lasphub.com)for performing atomic simulation of complex materials.The software integrates the neural network(NN)potential technique with the global potential energy surface exploration method,and thus can be utilized widely for structure prediction and reaction mechanism exploration.Here we introduce our recent update on the LASP program version 3.0,focusing on the new functionalities including the advanced neuralnetwork training based on the multi-network framework,the newly-introduced S^(7) and S^(8) power type structure descriptor(PTSD).These new functionalities are designed to further improve the accuracy of potentials and accelerate the neural network training for multipleelement systems.Taking Cu-C-H-O neural network potential and a heterogeneous catalytic model as the example,we show that these new functionalities can accelerate the training of multi-element neural network potential by using the existing single-network potential as the input.The obtained double-network potential Cu CHO is robust in simulation and the introduction of S^(7) and S^(8) PTSDs can reduce the root-mean-square errors of energy by a factor of two.展开更多
For neural network potentials(NNPs)to gain widespread use,researchers must be able to trust model outputs.However,the blackbox nature of neural networks and their inherent stochasticity are often deterrents,especially...For neural network potentials(NNPs)to gain widespread use,researchers must be able to trust model outputs.However,the blackbox nature of neural networks and their inherent stochasticity are often deterrents,especially for foundationmodels trained over broad swaths of chemical space.Uncertainty information provided at the time of prediction can help reduce aversion to NNPs.In this work,we detail two uncertainty quantification(UQ)methods.Readout ensembling,by finetuning the readout layers of an ensemble of foundation models,provides information about model uncertainty,while quantile regression,by replacing point predictions with distributional predictions,provides information about uncertainty within the underlying training data.We demonstrate our approach with the MACE-MP-0 model,applying UQ to the foundation model and a series of finetuned models.The uncertainties produced by the readout ensemble and quantile methods are demonstrated to be distinct measures by which the quality of the NNP output can be judged.展开更多
The discovery and optimization of high-energy materials(HEMs)face challenges due to the computational expense and slow iteration of traditional methods.Neural network potentials(NNPs)have emerged as an efficient alter...The discovery and optimization of high-energy materials(HEMs)face challenges due to the computational expense and slow iteration of traditional methods.Neural network potentials(NNPs)have emerged as an efficient alternative to first-principles simulations.This study presents EMFF-2025,a general NNP model for C,H,N,and O-based HEMs,leveraging transfer learning with minimal data from DFT calculations.The model achieves DFT-level accuracy,predicting the structure,mechanical properties,and decomposition characteristics of 20 HEMs.Integrating EMFF-2025 with PCA and correlation heatmaps,we map the chemical space and structural evolution of these HEMs across temperatures.Surprisingly,EMFF-2025 uncovers that most HEMs follow similar hightemperature decomposition mechanisms,challenging the conventional view of material-specific behavior.EMFF-2025 offers a versatile computational framework for accelerating HEM design and optimization.展开更多
Searching for novel carbon allotropes with excellent mechanical and interesting electronic properties is valuable,but such a large structural search remains a challenge if purely based on the traditional density funct...Searching for novel carbon allotropes with excellent mechanical and interesting electronic properties is valuable,but such a large structural search remains a challenge if purely based on the traditional density functional theory(DFT)combined with Monte-Carlo(MC)methods.Herein,the neural network potential is utilized to accelerate the sampling of the stochastic surface walking algorithm for a global structural search of ordered carbons from carbon nanotubes(CNTs)under pressure.A variety of unreported ordered carbons are obtained,among which CNTs with diameters smaller than 0.7 nm are more sensitive to pressure than bigger tubes.Most ordered carbons obtained show great thermodynamical and kinetic stability,verified by ab initio molecular dynamics simulations and phonon spectra.The ordered carbons demonstrate direct or indirect band gaps in the range of 0 to 4.4 eV,including 13 superhard(H,>40 GPa)structures and 1 ductile(Pugh's Ratio G/B<0.57)structure,in which the modulus of ordered carbons exhibits a linear correlation with the density.Our study provides a pathway to create new carbons from nanotubes and a deeper understanding of the structural evolution of CNT's under pressure.展开更多
This study developed neural network potentials(NNPs)specifically tailored for pure ethylene and ethyleneammonia blended systems for the first time.The NNPs were trained on a dataset generated from density func-tional ...This study developed neural network potentials(NNPs)specifically tailored for pure ethylene and ethyleneammonia blended systems for the first time.The NNPs were trained on a dataset generated from density func-tional theory(DFT)calculations,combining the computational accuracy of DFT with a calculation speed com-parable to reactive force field methods.The NNPs are employed in reactive molecular dynamics simulations to explore the thermal decomposition reaction mechanisms of ethylene and ammonia.The simulation results revealed that adding ammonia reduces the activation energy for ethylene decomposition,thereby accelerating ethylene consumption.Furthermore,the addition of ammonia uncovers a new reaction pathway for hydrogen radical consumption,which reduces the occurrence of H-abstraction reactions from ethylene by hydrogen rad-icals.The inhibition effect of ammonia addition on soot formation mainly acts in two aspects:on the one hand,ammonia decomposition products react with carbon-containing species,ultimately producing C_(1)-N products,thereby decreasing the carbon numbers involved in soot formation.This significantly reduces the concentrations of C_(5)-C_(9)molecules and key polycyclic aromatic hydrocarbons(PAHs)precursors like C_(2)H_(2)and C_(3)H_(3).On the other hand,ammonia promotes the ring-opening reactions of six-membered carbon rings at high-temperature conditions,thereby reducing the formation of PAHs precursors.The results show that with the addition of ammonia,six-membered carbon rings tend to convert into seven-membered carbon rings at lower temperatures,while at higher temperatures,they are more likely to transform into three-and five-membered carbon rings.These variations in the transformation of six-membered carbon rings may also affect soot formation.The insights gained from understanding these fundamental chemical reaction mechanisms can guide the development of ethylene-ammonia co-firing systems.展开更多
Gallium,an elemental metal known for its distinctive thermal and electronic characteristics,holds signif-icant importance across various industrial fields including semiconductors,biomedicine,and aerospace.When subjec...Gallium,an elemental metal known for its distinctive thermal and electronic characteristics,holds signif-icant importance across various industrial fields including semiconductors,biomedicine,and aerospace.When subjected to supercooling,gallium exhibits the ability to crystallize into multiple structures that are notably more intricate compared to those found in other metallic elements,emphasizing the complex nature of its configuration space.Despite ongoing research efforts,our comprehensive understanding of the configuration space of gallium remains incomplete.In this study,we utilize an active learning strat-egy to develop an accurate deep neural network(DNN)model with strong descriptive capabilities for gallium’s entire configuration space.By integrating this DNN model with the evolutionary crystal struc-ture prediction algorithm USPEX,we conduct an extensive exploration of gallium configurations within simulation cells containing up to 120 atoms.Through this approach,we successfully identify the experi-mentally observed phases ofα-Ga,β-Ga,γ-Ga,δ-Ga,Ga-II and Ga-III.Additionally,we predict eight ther-modynamically metastable structures,labeled as mC 20,oC 8(no.63),mC 4,oP 12,tR 18,tI 20,oC 8(no.64),and mC 12,with high potential of experimental verification.Of particular interest,we identify the true struc-ture ofβ-Ga as having orthorhombic symmetry,in contrast to the widely accepted monoclini c structure.These findings offer new insights into gallium’s configuration space,demonstrating the effectiveness of the crystal structure prediction method combined with DNN model in guiding the exploration of complex systems.展开更多
Metal ions play crucial roles in various biologi-cal functions,in-cluding maintain-ing homeostasis,regulating mus-cle contraction,and facilitating enzyme catalysis.However,accurately simulating the interaction between...Metal ions play crucial roles in various biologi-cal functions,in-cluding maintain-ing homeostasis,regulating mus-cle contraction,and facilitating enzyme catalysis.However,accurately simulating the interaction between metal ions and amino acid side chain analogs using high-level wave function theories remains challenging due to the significant computational costs involved.In this study,deep potential molecular dynamics(DeePMD)simulation was employed to investigate the solvation structure of the Mg^(2+)-Ac^(−)ion pair in aqueous solution.To address the computational bottleneck associated with expensive quan-tum mechanics(QM)methods,the Deep Kohn-Sham(DeePKS)approach was utilized,which allows us to generate highly accurate self-consistent energy functionals while significantly re-ducing computational costs.The root mean square error and mean absolute error of energies and atomic forces indicate close agreement between DeePKS predictions and QM strongly constrained and appropriately normed(SCAN)calculations.Moreover,the neural network potential(NNP)generated using the SCAN-level dataset predicted by DeePKS exhibits high-er accuracy compared to previous work,which employed a moderate BLYP functional.The potential of mean force for the Mg^(2+)-Ac−system was further examined,revealing a prefer-ence for monodentate coordination of Mg^(2+)with a~5.8 kcal/mol energy barrier between bidentate and monodentate geometries.Overall,this work provides a comprehensive,precise,and reliable methodology for investigating metal ions’properties in aqueous solutions.展开更多
Cadmium selenide(CdSe)is an inorganic semiconductor with unique optical and electronic properties that make it useful in various applications,including solar cells,light-emitting diodes,and biofluorescent tagging.In o...Cadmium selenide(CdSe)is an inorganic semiconductor with unique optical and electronic properties that make it useful in various applications,including solar cells,light-emitting diodes,and biofluorescent tagging.In order to synthesize high-quality crystals and subsequently integrate them into devices,it is crucial to understand the atomic scale crystallization mechanism of CdSe.Unfortunately,such studies are still absent in the literature.To overcome this limitation,we employed an enhanced sampling-accelerated active learning approach to construct a deep neural potential with ab initio accuracy for studying the crystallization of CdSe.Our brute-force molecular dynamics simulations revealed that a spherical-like nu-cleus formed spontaneously and stochastically,resulting in a stacking disordered structure where the competition between hexagonal wurtzite and cubic zinc blende polymorphs is temperature-dependent.We found that pure hexagonal crystal can only be obtained approximately above 1430 K,which is 35 K below its melting temperature.Furthermore,we observed that the solidification dynamics of Cd and Se atoms were distinct due to their different diffusion coefficients.The solidification process was initiated by lower mobile Se atoms forming tetrahedral frameworks,followed by Cd atoms occupying these tetra-hedral centers and settling down until the third-shell neighbor of Se atoms sited on their lattice posi-tions.Therefore,the medium-range ordering of Se atoms governs the crystallization process of CdSe.Our findings indicate that understanding the complex dynamical process is the key to comprehending the crystallization mechanism of compounds like CdSe,and can shed lights in the synthesis of high-quality crystals.展开更多
We present large-scale molecular dynamics(MD)simulations based on a neural network potential(NNP)to investigate alkaline wet etching of GaN,a process critical to nitride-based semiconductor fabrication.A Behler–Parri...We present large-scale molecular dynamics(MD)simulations based on a neural network potential(NNP)to investigate alkaline wet etching of GaN,a process critical to nitride-based semiconductor fabrication.A Behler–Parrinello-type NNP is trained on extensive DFT datasets to capture chemical reactions between GaN and KOH.Using temperature-accelerated dynamics,our NNP-MD simulations accurately reproduce experimentally observed structural modifications of GaN nanorods during etching.The etching simulations reveal surface-specific morphological evolution:pyramidal pits on the−c plane,truncated pyramids on the+c plane,and planar morphologies on non-polar m and a surfaces.We also identify key chemical reactions governing the etching mechanisms.Enhanced-sampling simulations provide free-energy profiles for Ga dissolution,which critically influences the etching rate.The−c,a,and m planes exhibit moderate activation barriers,confirming their etchability,while the+c surface shows a significantly higher barrier,indicating strong resistance.Wealso observe the formation of Ga-O-Ga bridges on etched surfaces,which may act as carrier traps.This work provides atomistic insights into the mechanisms and kinetics of GaN wet etching,offering guidance for the fabrication of nanostructures in advanced GaN-based electronic and display applications.展开更多
Investigation of the electrophysiological mechanisms that induce arrhythmias is one of the most important issues in scientific research.Since computational cardiology allows the systematic dissection of causal mechani...Investigation of the electrophysiological mechanisms that induce arrhythmias is one of the most important issues in scientific research.Since computational cardiology allows the systematic dissection of causal mechanisms of observed effects,simulations based on the ionic channel mathematical models have become one of the most widely used methods.To reduce themanual classification of different types of membrane potential patterns produced during simulations,a convolutional neural network is developed in this paper.The model includes 4convolution layers,4 pooling layers and a fully connected layer.An activation function of Re LU is used.Before machine learning,all the pattems are calibrated,cut,and normalized to a uniform format with a size of 256×256.The contour boundary of each pattern is extracted using the maximum between-class variance method.In the examination,the proposed learning algorithm shows a recognition accuracy of 97%on test data set after training.展开更多
The past decade has seen a sharp increase in machine learning(ML)applications in scientific research.This review introduces the basic constituents of ML,including databases,features,and algorithms,and highlights a few...The past decade has seen a sharp increase in machine learning(ML)applications in scientific research.This review introduces the basic constituents of ML,including databases,features,and algorithms,and highlights a few important achievements in chemistry that have been aided by ML techniques.The described databases include some of the most popular chemical databases for molecules and materials obtained from either experiments or computational calculations.Important two-dimensional(2D)and three-dimensional(3D)features representing the chemical environment of molecules and solids are briefly introduced.Decision tree and deep learning neural network algorithms are overviewed to emphasize their frameworks and typical application scenarios.Three important fields of ML in chemistry are discussed:(1)retrosynthesis,in which ML predicts the likely routes of organic synthesis;(2)atomic simulations,which utilize the ML potential to accelerate potential energy surface sampling;and(3)heterogeneous catalysis,in which ML assists in various aspects of catalytic design,ranging from synthetic condition optimization to reaction mechanism exploration.Finally,a prospect on future ML applications is provided.展开更多
Researchers have been attempting to characterize heterogeneous catalysts in situ in addition to correlating their structures with their activity and selectivity in spite of many challenges.Here,we review recent experi...Researchers have been attempting to characterize heterogeneous catalysts in situ in addition to correlating their structures with their activity and selectivity in spite of many challenges.Here,we review recent experimental and theoretical advances regarding alkyne selective hydrogenation by Pd‐based catalysts,which are an important petrochemical reaction.The catalytic selectivity for the reaction of alkynes to alkenes is influenced by the composition and structure of the catalysts.Recent progress achieved through experimental studies and atomic simulations has provided useful insights into the origins of the selectivity.The important role of the subsurface species(H and C)was revealed by monitoring the catalyst surface and the related catalytic performance.The atomic structures of the Pd catalytic centers and their relationship with selectivity were established through atomic simulations.The combined knowledge gained from experimental and theoretical studies provides a fundamental understanding of catalytic mechanisms and reveals a path toward improved catalyst design.展开更多
Al_(2)O_(3)and Ga_(2)O_(3)exhibit numerous crystal phases with distinct stabilities and materialproperties.However,the phase transitions among thosematerialsare typicallyundesirable in industrial applications,making i...Al_(2)O_(3)and Ga_(2)O_(3)exhibit numerous crystal phases with distinct stabilities and materialproperties.However,the phase transitions among thosematerialsare typicallyundesirable in industrial applications,making it imperative to elucidate the transition mechanisms between these phases.The configurational similarities between Al_(2)O_(3)and Ga_(2)O_(3)allow for the replication of phase transition pathways between these materials.In this study,we investigate the potential phase transition pathway of alumina from the 0-phase to the α-phase using stochastic surface walking global optimization based on global neural network potentials,while extending an existing Ga_(2)O_(3)phase transition path.Through this exploration,we identify a novel single-atom migration pseudomartensitic mechanism,which combines martensitic transformation with single-atom diffusion.This discovery offers valuable insights for experimental endeavors aimed at stabilizing alumina in transitional phases.展开更多
Ice,a ubiquitous substance in nature,exhibits diverse forms under varying temperature and pressure conditions.However,our understanding of ice polymorphs remains incomplete.The directional nature of hydrogen bonding a...Ice,a ubiquitous substance in nature,exhibits diverse forms under varying temperature and pressure conditions.However,our understanding of ice polymorphs remains incomplete.The directional nature of hydrogen bonding and the complexity of the networks they form pose significant challenges to computational studies of ice structures.In this work,we present an extensive exploration of ice polymorphs under pressure conditions ranging from 1 bar to 10 GPa.We employ an advanced crystal-structure-prediction scheme that integrates an evolutionary algorithm,an active-learning deep neural network potential,and molecular dynamics simulations with ab initio accuracy.Among the 131,481 predicted structures,we successfully identify all experimentally known ice phases within the target pressure range,including the particularly challenging ice IV and V.These phases feature highly intricate H-bond networks,which have hindered previous efforts to fully explore ice structures.Additionally,we identify 34 new ice polymorphs that are potential candidates for experimental discovery.Notably,we predict the existence of a new stable ice phase,ice L,within the temperature range of 253–291 K and pressure range of 0.38–0.57 GPa,exhibiting a unique topology unseen in any known crystals.Our findings highlight the potential for experimental discovery of new ice phases.Furthermore,our approach can be applied to other complex systems,particularly those with network structures.展开更多
Computational tools on top of first principle calculations have played an indispensable role in revealing the molecular details,thermodynamics,and kinetics in catalytic reactions.Here we proposed a highly efficient dy...Computational tools on top of first principle calculations have played an indispensable role in revealing the molecular details,thermodynamics,and kinetics in catalytic reactions.Here we proposed a highly efficient dynamic strategy for the calculation of thermodynamic and kinetic properties in heterogeneous catalysis on the basis of efficient potential energy surface(PES)and MD simulations.Taking CO adsorbate on Ru(0001)surface as the illustrative model system,we demonstrated the PES-based MD can efficiently generate reliable two-dimensional potential-of-mean-force(PMF)surfaces in a wide range of temperatures,and thus temperature-dependent thermodynamic properties can be obtained in a comprehensive investigation on the whole PMF surface.Moreover,MD offers an effective way to describe the surface kinetics such as adsorbate on-surface movement,which goes beyond the most popular static approach based on free energy barrier and transition state theory(TST).We further revealed that the dynamic strategy significantly improves the predictions of both thermodynamic and kinetic properties as compared to the popular ideal statistic mechanics approaches such as harmonic analysis and TST.It is expected that this accurate yet efficient dynamic strategy can be powerful in understanding mechanisms and reactivity of a catalytic surface system,and further guides the rational design of heterogeneous catalysts.展开更多
Classical molecular dynamics simulations with global neural network machine learning potential are used to study early stage oxidation and dissolution behaviors of bcc Fe surfaces contacting with stagnant oxygen disso...Classical molecular dynamics simulations with global neural network machine learning potential are used to study early stage oxidation and dissolution behaviors of bcc Fe surfaces contacting with stagnant oxygen dissolved liquid leadbismuth eutectic(LBE-O).Both static and dynamic simulation results indicate that the early stage oxidation and dissolution behaviors of bcc Fe show strong orientation dependence under the liquid LBE environments,which may explain the experimental observations of uneven interface between iron-based materials and liquid LBE.Our investigations show that it is the delicate balance between the oxide growth and metal dissolution that leads to the observed corrosion anisotropy for bcc Fe contacting with liquid LBE-O.展开更多
Graph Neural Network(GNN)potentials relying on chemical locality offer near-quantum mechanical accuracy at significantly reduced computational costs.Message-passing GNNs model interactions beyond their immediate neigh...Graph Neural Network(GNN)potentials relying on chemical locality offer near-quantum mechanical accuracy at significantly reduced computational costs.Message-passing GNNs model interactions beyond their immediate neighborhood by propagating local information between neighboring particles while remaining effectively local.However,locality precludes modeling long-range effects critical to many real-world systems,such as charge transfer,electrostatic interactions,and dispersion effects.In this work,we propose the Charge Equilibration Layer for Long-range Interactions(CELLI)to address the challenge of efficiently modeling non-local interactions.This novel architecture generalizes the classical charge equilibration(Qeq)method to a model-agnostic building block for modern equivariant GNN potentials.Therefore,CELLI extends the capability of GNNs to model longrange interactions while providing high interpretability through explicitly modeled charges.On benchmark systems,CELLI achieves state-of-the-art results for strictly local models.CELLI generalizes to diverse datasets and large structureswhile providing high computational efficiency and robust predictions.展开更多
Atomic simulations aim to understand and predict complex physical phenomena,the success of which relies largely on the accuracy of the potential energy surface description and the efficiency to capture important rare ...Atomic simulations aim to understand and predict complex physical phenomena,the success of which relies largely on the accuracy of the potential energy surface description and the efficiency to capture important rare events.LASP software(large-scale atomic simulation with a Neural Network Potential),released in 2018,incorporates the key ingredients to fulfill the ultimate goal of atomic simulations by combining advanced neural network potentials with efficient global optimization methods.This review introduces the recent development of the software along two main streams,namely,higher intelligence and more automation,to solve complex material and reaction problems.The latest version of LASP(LASP 3.7)features the global many-body function corrected neural network(G-MBNN)to improve the PES accuracy with low cost,which achieves a linear scaling efficiency for large-scale atomic simulations.The key functionalities of LASP are updated to incorporate(i)the ASOP and ML-interface methods for finding complex surface and interface structures under grand canonic conditions;(ii)the ML-TS and MMLPS methods to identify the lowest energy reaction pathway.With these powerful functionalities,LASP now serves as an intelligent data generator to create computational databases for end users.We exemplify the recent LASP database construction in zeolite and the metal−ligand properties for a new catalyst design.展开更多
基金supported by the National Key Research and Development Program of China (No.2018YFA0208600)the National Natural Science Foundation of China (No.91945301, No.22033003, No.92061112, No.22122301, and No.91745201)
文摘LASP(large-scale atomistic simulation with neural network potential)software developed by our group since 2018 is a powerful platform(www.lasphub.com)for performing atomic simulation of complex materials.The software integrates the neural network(NN)potential technique with the global potential energy surface exploration method,and thus can be utilized widely for structure prediction and reaction mechanism exploration.Here we introduce our recent update on the LASP program version 3.0,focusing on the new functionalities including the advanced neuralnetwork training based on the multi-network framework,the newly-introduced S^(7) and S^(8) power type structure descriptor(PTSD).These new functionalities are designed to further improve the accuracy of potentials and accelerate the neural network training for multipleelement systems.Taking Cu-C-H-O neural network potential and a heterogeneous catalytic model as the example,we show that these new functionalities can accelerate the training of multi-element neural network potential by using the existing single-network potential as the input.The obtained double-network potential Cu CHO is robust in simulation and the introduction of S^(7) and S^(8) PTSDs can reduce the root-mean-square errors of energy by a factor of two.
基金supported by the"Transferring exascale computational chemistry to cloud computing environment and emerging hardware technologies(TEC4)"project,which is funded by the U.S.Department of Energy,Office of Science,Office of Basic Energy Sciences,the Division of Chemical Sciences,Geosciences,and Biosciences(under FWP 82037)supported by the U.S.Department of Energy(DOE),Office of Science,Office of Basic Energy Sciences,Division of Chemical Sciences,Geosciences&Biosciences(under FWP 47319)Pacific Northwest National Laboratory(PNNL)is a multiprogram national laboratory operated for the U.S.Department of Energy(DOE)by Battelle Memorial Institute under Contract No.DE-AC05-76RL0-1830.
文摘For neural network potentials(NNPs)to gain widespread use,researchers must be able to trust model outputs.However,the blackbox nature of neural networks and their inherent stochasticity are often deterrents,especially for foundationmodels trained over broad swaths of chemical space.Uncertainty information provided at the time of prediction can help reduce aversion to NNPs.In this work,we detail two uncertainty quantification(UQ)methods.Readout ensembling,by finetuning the readout layers of an ensemble of foundation models,provides information about model uncertainty,while quantile regression,by replacing point predictions with distributional predictions,provides information about uncertainty within the underlying training data.We demonstrate our approach with the MACE-MP-0 model,applying UQ to the foundation model and a series of finetuned models.The uncertainties produced by the readout ensemble and quantile methods are demonstrated to be distinct measures by which the quality of the NNP output can be judged.
基金supported by the National Natural Science Foundation ofChina(Grant 52106130)the State Key Laboratory of Explosion Science and Safety Protection(Grants QNKT23-15).
文摘The discovery and optimization of high-energy materials(HEMs)face challenges due to the computational expense and slow iteration of traditional methods.Neural network potentials(NNPs)have emerged as an efficient alternative to first-principles simulations.This study presents EMFF-2025,a general NNP model for C,H,N,and O-based HEMs,leveraging transfer learning with minimal data from DFT calculations.The model achieves DFT-level accuracy,predicting the structure,mechanical properties,and decomposition characteristics of 20 HEMs.Integrating EMFF-2025 with PCA and correlation heatmaps,we map the chemical space and structural evolution of these HEMs across temperatures.Surprisingly,EMFF-2025 uncovers that most HEMs follow similar hightemperature decomposition mechanisms,challenging the conventional view of material-specific behavior.EMFF-2025 offers a versatile computational framework for accelerating HEM design and optimization.
基金supported by the National Natural Science Foundation of China(Nos.52273234,52325202)The Fundamental Research Funds for the Central Universities(WK2060000098,WK9990000170)the open research fund of State Key Laboratory of Precision and Intelligent Chemistry(KY2490000300).
文摘Searching for novel carbon allotropes with excellent mechanical and interesting electronic properties is valuable,but such a large structural search remains a challenge if purely based on the traditional density functional theory(DFT)combined with Monte-Carlo(MC)methods.Herein,the neural network potential is utilized to accelerate the sampling of the stochastic surface walking algorithm for a global structural search of ordered carbons from carbon nanotubes(CNTs)under pressure.A variety of unreported ordered carbons are obtained,among which CNTs with diameters smaller than 0.7 nm are more sensitive to pressure than bigger tubes.Most ordered carbons obtained show great thermodynamical and kinetic stability,verified by ab initio molecular dynamics simulations and phonon spectra.The ordered carbons demonstrate direct or indirect band gaps in the range of 0 to 4.4 eV,including 13 superhard(H,>40 GPa)structures and 1 ductile(Pugh's Ratio G/B<0.57)structure,in which the modulus of ordered carbons exhibits a linear correlation with the density.Our study provides a pathway to create new carbons from nanotubes and a deeper understanding of the structural evolution of CNT's under pressure.
基金supported by UK Physical Sciences Research Council(EPSRC)under Grant No.EP/X019551/1Supercomputing time on ARCHER is provided by the“UK Consortium on Mesoscale Engineering Sciences(UKCOMES)”under the UK EPSRC Grant No.EP/R029598/1.
文摘This study developed neural network potentials(NNPs)specifically tailored for pure ethylene and ethyleneammonia blended systems for the first time.The NNPs were trained on a dataset generated from density func-tional theory(DFT)calculations,combining the computational accuracy of DFT with a calculation speed com-parable to reactive force field methods.The NNPs are employed in reactive molecular dynamics simulations to explore the thermal decomposition reaction mechanisms of ethylene and ammonia.The simulation results revealed that adding ammonia reduces the activation energy for ethylene decomposition,thereby accelerating ethylene consumption.Furthermore,the addition of ammonia uncovers a new reaction pathway for hydrogen radical consumption,which reduces the occurrence of H-abstraction reactions from ethylene by hydrogen rad-icals.The inhibition effect of ammonia addition on soot formation mainly acts in two aspects:on the one hand,ammonia decomposition products react with carbon-containing species,ultimately producing C_(1)-N products,thereby decreasing the carbon numbers involved in soot formation.This significantly reduces the concentrations of C_(5)-C_(9)molecules and key polycyclic aromatic hydrocarbons(PAHs)precursors like C_(2)H_(2)and C_(3)H_(3).On the other hand,ammonia promotes the ring-opening reactions of six-membered carbon rings at high-temperature conditions,thereby reducing the formation of PAHs precursors.The results show that with the addition of ammonia,six-membered carbon rings tend to convert into seven-membered carbon rings at lower temperatures,while at higher temperatures,they are more likely to transform into three-and five-membered carbon rings.These variations in the transformation of six-membered carbon rings may also affect soot formation.The insights gained from understanding these fundamental chemical reaction mechanisms can guide the development of ethylene-ammonia co-firing systems.
基金financially supported by the National Natural Science Foundation of China(No.92370118)the National Science Fund for Excellent Young Scientist Fund Program(Overseas)of China,the Science and Technology Activities Fund for Overseas Researchers of Shaanxi Province,and the Research Fund of the State Key Laboratory of Solidification Processing(NPU),China(No.2024-ZD-01)。
文摘Gallium,an elemental metal known for its distinctive thermal and electronic characteristics,holds signif-icant importance across various industrial fields including semiconductors,biomedicine,and aerospace.When subjected to supercooling,gallium exhibits the ability to crystallize into multiple structures that are notably more intricate compared to those found in other metallic elements,emphasizing the complex nature of its configuration space.Despite ongoing research efforts,our comprehensive understanding of the configuration space of gallium remains incomplete.In this study,we utilize an active learning strat-egy to develop an accurate deep neural network(DNN)model with strong descriptive capabilities for gallium’s entire configuration space.By integrating this DNN model with the evolutionary crystal struc-ture prediction algorithm USPEX,we conduct an extensive exploration of gallium configurations within simulation cells containing up to 120 atoms.Through this approach,we successfully identify the experi-mentally observed phases ofα-Ga,β-Ga,γ-Ga,δ-Ga,Ga-II and Ga-III.Additionally,we predict eight ther-modynamically metastable structures,labeled as mC 20,oC 8(no.63),mC 4,oP 12,tR 18,tI 20,oC 8(no.64),and mC 12,with high potential of experimental verification.Of particular interest,we identify the true struc-ture ofβ-Ga as having orthorhombic symmetry,in contrast to the widely accepted monoclini c structure.These findings offer new insights into gallium’s configuration space,demonstrating the effectiveness of the crystal structure prediction method combined with DNN model in guiding the exploration of complex systems.
基金supported by the National Natural Sci-ence Foundation of China(No.22373065,No.62072296,No.22222303,No.22173032,No.21933010)the Nation-al Key R&D Program of China(No.2023YFF1204903)NYU-ECNU Center for Computational Chemistry at NYU Shanghai,the Opening Project of Shanghai Frontiers Science Research Center for Druggability of Cardiovascular noncoding RNA.
文摘Metal ions play crucial roles in various biologi-cal functions,in-cluding maintain-ing homeostasis,regulating mus-cle contraction,and facilitating enzyme catalysis.However,accurately simulating the interaction between metal ions and amino acid side chain analogs using high-level wave function theories remains challenging due to the significant computational costs involved.In this study,deep potential molecular dynamics(DeePMD)simulation was employed to investigate the solvation structure of the Mg^(2+)-Ac^(−)ion pair in aqueous solution.To address the computational bottleneck associated with expensive quan-tum mechanics(QM)methods,the Deep Kohn-Sham(DeePKS)approach was utilized,which allows us to generate highly accurate self-consistent energy functionals while significantly re-ducing computational costs.The root mean square error and mean absolute error of energies and atomic forces indicate close agreement between DeePKS predictions and QM strongly constrained and appropriately normed(SCAN)calculations.Moreover,the neural network potential(NNP)generated using the SCAN-level dataset predicted by DeePKS exhibits high-er accuracy compared to previous work,which employed a moderate BLYP functional.The potential of mean force for the Mg^(2+)-Ac−system was further examined,revealing a prefer-ence for monodentate coordination of Mg^(2+)with a~5.8 kcal/mol energy barrier between bidentate and monodentate geometries.Overall,this work provides a comprehensive,precise,and reliable methodology for investigating metal ions’properties in aqueous solutions.
基金supported by the National Natural Science Foundation of China(No.92370118)the National Science Fund for Excellent Young Scientist Fund Program(Overseas)of China,the Science and Technology Activities Fund for Overseas Researchers of Shaanxi Province,China,and the Research Fund of the State Key Laboratory of Solidification Proceeding(NPU)of China(No.2020-QZ-03).
文摘Cadmium selenide(CdSe)is an inorganic semiconductor with unique optical and electronic properties that make it useful in various applications,including solar cells,light-emitting diodes,and biofluorescent tagging.In order to synthesize high-quality crystals and subsequently integrate them into devices,it is crucial to understand the atomic scale crystallization mechanism of CdSe.Unfortunately,such studies are still absent in the literature.To overcome this limitation,we employed an enhanced sampling-accelerated active learning approach to construct a deep neural potential with ab initio accuracy for studying the crystallization of CdSe.Our brute-force molecular dynamics simulations revealed that a spherical-like nu-cleus formed spontaneously and stochastically,resulting in a stacking disordered structure where the competition between hexagonal wurtzite and cubic zinc blende polymorphs is temperature-dependent.We found that pure hexagonal crystal can only be obtained approximately above 1430 K,which is 35 K below its melting temperature.Furthermore,we observed that the solidification dynamics of Cd and Se atoms were distinct due to their different diffusion coefficients.The solidification process was initiated by lower mobile Se atoms forming tetrahedral frameworks,followed by Cd atoms occupying these tetra-hedral centers and settling down until the third-shell neighbor of Se atoms sited on their lattice posi-tions.Therefore,the medium-range ordering of Se atoms governs the crystallization process of CdSe.Our findings indicate that understanding the complex dynamical process is the key to comprehending the crystallization mechanism of compounds like CdSe,and can shed lights in the synthesis of high-quality crystals.
基金supported by Samsung Electronics Co., Ltd (IO201214-08143-01)the Nano & Material Technology Development Program through the National Research Foundation of Korea (NRF) funded by Ministry of Science and ICT (RS-2024-00407995). We are grateful to Jisu Jung for the technical discussion.
文摘We present large-scale molecular dynamics(MD)simulations based on a neural network potential(NNP)to investigate alkaline wet etching of GaN,a process critical to nitride-based semiconductor fabrication.A Behler–Parrinello-type NNP is trained on extensive DFT datasets to capture chemical reactions between GaN and KOH.Using temperature-accelerated dynamics,our NNP-MD simulations accurately reproduce experimentally observed structural modifications of GaN nanorods during etching.The etching simulations reveal surface-specific morphological evolution:pyramidal pits on the−c plane,truncated pyramids on the+c plane,and planar morphologies on non-polar m and a surfaces.We also identify key chemical reactions governing the etching mechanisms.Enhanced-sampling simulations provide free-energy profiles for Ga dissolution,which critically influences the etching rate.The−c,a,and m planes exhibit moderate activation barriers,confirming their etchability,while the+c surface shows a significantly higher barrier,indicating strong resistance.Wealso observe the formation of Ga-O-Ga bridges on etched surfaces,which may act as carrier traps.This work provides atomistic insights into the mechanisms and kinetics of GaN wet etching,offering guidance for the fabrication of nanostructures in advanced GaN-based electronic and display applications.
基金Natural Science Foundation of Shaanxi Province in China,grant number:2019JM-137Natural Science Foundation of China,81271661
文摘Investigation of the electrophysiological mechanisms that induce arrhythmias is one of the most important issues in scientific research.Since computational cardiology allows the systematic dissection of causal mechanisms of observed effects,simulations based on the ionic channel mathematical models have become one of the most widely used methods.To reduce themanual classification of different types of membrane potential patterns produced during simulations,a convolutional neural network is developed in this paper.The model includes 4convolution layers,4 pooling layers and a fully connected layer.An activation function of Re LU is used.Before machine learning,all the pattems are calibrated,cut,and normalized to a uniform format with a size of 256×256.The contour boundary of each pattern is extracted using the maximum between-class variance method.In the examination,the proposed learning algorithm shows a recognition accuracy of 97%on test data set after training.
基金financial support from the National Key Research and Development Program of China(2018YFA0208600)the National Natural Science Foundation of China(12188101,22033003,91945301,91745201,92145302,22122301,and 92061112)the Tencent Foundation for XPLORER PRIZE,and Fundamental Research Funds for the Central Universities(20720220011)。
文摘The past decade has seen a sharp increase in machine learning(ML)applications in scientific research.This review introduces the basic constituents of ML,including databases,features,and algorithms,and highlights a few important achievements in chemistry that have been aided by ML techniques.The described databases include some of the most popular chemical databases for molecules and materials obtained from either experiments or computational calculations.Important two-dimensional(2D)and three-dimensional(3D)features representing the chemical environment of molecules and solids are briefly introduced.Decision tree and deep learning neural network algorithms are overviewed to emphasize their frameworks and typical application scenarios.Three important fields of ML in chemistry are discussed:(1)retrosynthesis,in which ML predicts the likely routes of organic synthesis;(2)atomic simulations,which utilize the ML potential to accelerate potential energy surface sampling;and(3)heterogeneous catalysis,in which ML assists in various aspects of catalytic design,ranging from synthetic condition optimization to reaction mechanism exploration.Finally,a prospect on future ML applications is provided.
文摘Researchers have been attempting to characterize heterogeneous catalysts in situ in addition to correlating their structures with their activity and selectivity in spite of many challenges.Here,we review recent experimental and theoretical advances regarding alkyne selective hydrogenation by Pd‐based catalysts,which are an important petrochemical reaction.The catalytic selectivity for the reaction of alkynes to alkenes is influenced by the composition and structure of the catalysts.Recent progress achieved through experimental studies and atomic simulations has provided useful insights into the origins of the selectivity.The important role of the subsurface species(H and C)was revealed by monitoring the catalyst surface and the related catalytic performance.The atomic structures of the Pd catalytic centers and their relationship with selectivity were established through atomic simulations.The combined knowledge gained from experimental and theoretical studies provides a fundamental understanding of catalytic mechanisms and reveals a path toward improved catalyst design.
基金supported by the National Natural Science Foundation of China(No.12188101,No.22122301,No.22033003,No.91745201,No.91945301,No.92145302,and No.92061112)the Fundamental Research Funds for the Central Universities(20720220011)+1 种基金the National Key Research and Devel-opment Program of China(2018YF A0208600)the Tencent Foundation for XPLORER PRIZE.
文摘Al_(2)O_(3)and Ga_(2)O_(3)exhibit numerous crystal phases with distinct stabilities and materialproperties.However,the phase transitions among thosematerialsare typicallyundesirable in industrial applications,making it imperative to elucidate the transition mechanisms between these phases.The configurational similarities between Al_(2)O_(3)and Ga_(2)O_(3)allow for the replication of phase transition pathways between these materials.In this study,we investigate the potential phase transition pathway of alumina from the 0-phase to the α-phase using stochastic surface walking global optimization based on global neural network potentials,while extending an existing Ga_(2)O_(3)phase transition path.Through this exploration,we identify a novel single-atom migration pseudomartensitic mechanism,which combines martensitic transformation with single-atom diffusion.This discovery offers valuable insights for experimental endeavors aimed at stabilizing alumina in transitional phases.
基金supported by the National Natural Science Foundation of China(grant nos.12374217,92370118,and 22003050)the Research Fund of the State Key Laboratory of Solidification Processing(NPU),China(grant no.2024-ZD-01)+2 种基金the Fundamental Research Funds for the Central Universities,China.The calculations were supported by the National Supercomputing Center in Xi’an,Chinathe BohriumCloud platformat DP technologythe International Center for Materials Discovery(ICMD)cluster of NPU。
文摘Ice,a ubiquitous substance in nature,exhibits diverse forms under varying temperature and pressure conditions.However,our understanding of ice polymorphs remains incomplete.The directional nature of hydrogen bonding and the complexity of the networks they form pose significant challenges to computational studies of ice structures.In this work,we present an extensive exploration of ice polymorphs under pressure conditions ranging from 1 bar to 10 GPa.We employ an advanced crystal-structure-prediction scheme that integrates an evolutionary algorithm,an active-learning deep neural network potential,and molecular dynamics simulations with ab initio accuracy.Among the 131,481 predicted structures,we successfully identify all experimentally known ice phases within the target pressure range,including the particularly challenging ice IV and V.These phases feature highly intricate H-bond networks,which have hindered previous efforts to fully explore ice structures.Additionally,we identify 34 new ice polymorphs that are potential candidates for experimental discovery.Notably,we predict the existence of a new stable ice phase,ice L,within the temperature range of 253–291 K and pressure range of 0.38–0.57 GPa,exhibiting a unique topology unseen in any known crystals.Our findings highlight the potential for experimental discovery of new ice phases.Furthermore,our approach can be applied to other complex systems,particularly those with network structures.
基金financially supported by Fujian Science&Technology Innovation Laboratory for Optoelectronic Information of China(No.2021ZR109)the National Natural Science Foundation of China(Nos.21973094,22173104,22173105)the Opening Project of PCOSS of Xiamen University(No.201908)。
文摘Computational tools on top of first principle calculations have played an indispensable role in revealing the molecular details,thermodynamics,and kinetics in catalytic reactions.Here we proposed a highly efficient dynamic strategy for the calculation of thermodynamic and kinetic properties in heterogeneous catalysis on the basis of efficient potential energy surface(PES)and MD simulations.Taking CO adsorbate on Ru(0001)surface as the illustrative model system,we demonstrated the PES-based MD can efficiently generate reliable two-dimensional potential-of-mean-force(PMF)surfaces in a wide range of temperatures,and thus temperature-dependent thermodynamic properties can be obtained in a comprehensive investigation on the whole PMF surface.Moreover,MD offers an effective way to describe the surface kinetics such as adsorbate on-surface movement,which goes beyond the most popular static approach based on free energy barrier and transition state theory(TST).We further revealed that the dynamic strategy significantly improves the predictions of both thermodynamic and kinetic properties as compared to the popular ideal statistic mechanics approaches such as harmonic analysis and TST.It is expected that this accurate yet efficient dynamic strategy can be powerful in understanding mechanisms and reactivity of a catalytic surface system,and further guides the rational design of heterogeneous catalysts.
基金the National Natural Science Foundation of China(Grant No.U1832206).
文摘Classical molecular dynamics simulations with global neural network machine learning potential are used to study early stage oxidation and dissolution behaviors of bcc Fe surfaces contacting with stagnant oxygen dissolved liquid leadbismuth eutectic(LBE-O).Both static and dynamic simulation results indicate that the early stage oxidation and dissolution behaviors of bcc Fe show strong orientation dependence under the liquid LBE environments,which may explain the experimental observations of uneven interface between iron-based materials and liquid LBE.Our investigations show that it is the delicate balance between the oxide growth and metal dissolution that leads to the observed corrosion anisotropy for bcc Fe contacting with liquid LBE-O.
基金Funded by the European Union. Views and opinions expressed are, however, those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Council Executive AgencyNeither the European Union nor the granting authority can be held responsible for them. This work was funded by the ERC (StG SupraModel) - 101077842the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) - 534045056 and 561190767.
文摘Graph Neural Network(GNN)potentials relying on chemical locality offer near-quantum mechanical accuracy at significantly reduced computational costs.Message-passing GNNs model interactions beyond their immediate neighborhood by propagating local information between neighboring particles while remaining effectively local.However,locality precludes modeling long-range effects critical to many real-world systems,such as charge transfer,electrostatic interactions,and dispersion effects.In this work,we propose the Charge Equilibration Layer for Long-range Interactions(CELLI)to address the challenge of efficiently modeling non-local interactions.This novel architecture generalizes the classical charge equilibration(Qeq)method to a model-agnostic building block for modern equivariant GNN potentials.Therefore,CELLI extends the capability of GNNs to model longrange interactions while providing high interpretability through explicitly modeled charges.On benchmark systems,CELLI achieves state-of-the-art results for strictly local models.CELLI generalizes to diverse datasets and large structureswhile providing high computational efficiency and robust predictions.
基金supported by the National Science Foundation of China(12188101,22033003,22122301,91945301,91745201,22203101)the Fundamental Research Funds for the Central Universities(20720220011)+2 种基金the National Key Research and Development Program of China(2018YFA0208600)Science&Technology Commission of Shanghai Municipality(23ZR1476100)the Tencent Foundation for XPLORER PRIZE.
文摘Atomic simulations aim to understand and predict complex physical phenomena,the success of which relies largely on the accuracy of the potential energy surface description and the efficiency to capture important rare events.LASP software(large-scale atomic simulation with a Neural Network Potential),released in 2018,incorporates the key ingredients to fulfill the ultimate goal of atomic simulations by combining advanced neural network potentials with efficient global optimization methods.This review introduces the recent development of the software along two main streams,namely,higher intelligence and more automation,to solve complex material and reaction problems.The latest version of LASP(LASP 3.7)features the global many-body function corrected neural network(G-MBNN)to improve the PES accuracy with low cost,which achieves a linear scaling efficiency for large-scale atomic simulations.The key functionalities of LASP are updated to incorporate(i)the ASOP and ML-interface methods for finding complex surface and interface structures under grand canonic conditions;(ii)the ML-TS and MMLPS methods to identify the lowest energy reaction pathway.With these powerful functionalities,LASP now serves as an intelligent data generator to create computational databases for end users.We exemplify the recent LASP database construction in zeolite and the metal−ligand properties for a new catalyst design.
