Machine Learning for Chemistry:Basics and Applications

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.

Selectivity control in alkyne semihydrogenation:Recent experimental and theoretical progress

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.

Theoretical Aspects on Doped-Zirconia for Solid Oxide Fuel Cells:from Structure to Conductivity

Solid oxide fuel cells(SOFCs)are regarded to be a key clean energy system to convert chemical energy(e.g.H_(2) and O_(2))into electrical energy with high efficiency,low carbon footprint,and fuel flexibility.The electrolyte,typically doped zirconia,is the"state of the heart"of the fuel cell technologies,determining the performance and the operating temperature of the overall cells.Yttria stabilized zirconia(YSZ)have been widely used in SOFC due to its excellent oxide ion conductivity at high temperature.The composition and temperature dependence of the conductivity has been hotly studied in experiment and,more recently,by theoretical simulations.The characterization of the atomic structure for the mixed oxide system with different compositions is the key for elucidating the conductivity behavior,which,however,is of great challenge to both experiment and theory.This review presents recent theoretical progress on the structure and conductivity of YSZ electrolyte.We compare different theoretical methods and their results,outlining the merits and deficiencies of the methods.We highlight the recent results achieved by using stochastic surface walking global optimization with global neural network potential(SSW-NN)method,which appear to agree with available experimental data.The advent of machine-learning atomic simulation provides an affordable,efficient and accurate way to understand the complex material phenomena as encountered in solid electrolyte.The future research directions for design better electrolytes are also discussed.

Recent Implementations in LASP 3.0:Global Neural Network Potential with Multiple Elements and Better Long-Range Description

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.

Directly Determining the Interface Structure and Band Offset of a Large-Lattice-Mismatched CdS/CdTe Heterostructure

The CdS/CdTe heterojunction plays an important role in determining the energy conversion efficiency of CdTe solar cells.However,the interface structure remains unknown,due to the large lattice mismatch between CdS and CdTe,posing great challenges to achieving an understanding of its interfacial effects.By combining a neuralnetwork-based machine-learning method and the stochastic surface walking-based global optimization method,we first train a neural network potential for CdSTe systems with demonstrated robustness and reliability.Based on the above potential,we then use simulated annealing to obtain the optimal structure of the CdS/CdTe interface.We find that the most stable structure has the features of both bulks and disorders.Using the obtained structure,we directly calculate the band offset between CdS and CdTe by aligning the core levels in the heterostructure with those in the bulks,using one-shot first-principles calculations.Our calculated band offset is 0.55 eV,in comparison with 0.70 eV,obtained using other indirect methods.The obtained interface structure should prove useful for further study of the properties of CdTe/CdS heterostructures.Our work also presents an example which is applicable to other complex interfaces.

Confined Mn^(2+) enables effective aerobic oxidation catalysis

Effective and mild activation of O_(2) is essential but challenging for aerobic oxidation. In heterogeneous catalysis, high-valence manganese oxide(e.g., +4) is known to be active for the oxidation, whereas divalent MnO is ineffective due to its limited capacity to supply surface oxygen and its thermodynamically unstable structure when binding O_(2) in reaction conditions. Inspired by natural enzymes that rely on divalent Mn^(2+), we discovered that confining Mn^(2+) onto the Mn_(2)O_(3) surface through a dedicated calcination process creates highly active catalysts for the aerobic oxidation of 5-hydroxymethylfurfural, benzyl alcohol, and CO.The Mn_(2)O_(3)-confined Mn^(2+) is undercoordinated and efficiently mediates O_(2) activation, resulting in 2–3 orders of magnitude higher activity than Mn_(2)O_(3) alone. Through low-temperature infrared spectroscopy, we distinguished low-content Mn^(2+) sites at Mn_(2)O_(3) surface, which are difficult to be differentiated by X-ray photoelectron spectroscopy. The combination of in-situ energydispersive X-ray absorption spectroscopy and X-ray diffraction further provides insights into the formation of the newly identified active Mn^(2+) sites. By optimizing the calcination step, we were able to increase the catalytic activity threefold further.The finding offers promising frontiers for exploring active oxidation catalysts by utilizing the confinement of Mn^(2+)and oftenignored calcination skills.

Machine-learning atomic simulation for heterogeneous catalysis

Heterogeneous catalysis is at the heart of chemistry.New theoretical methods based on machine learning(ML)techniques that emerged in recent years provide a new avenue to disclose the structures and reaction in complex catalytic systems.Here we review briefly the history of atomic simulations in catalysis and then focus on the recent trend shifting toward ML potential calculations.The advanced methods developed by our group are outlined to illustrate how complex structures and reaction networks can be resolved using the ML potential in combination with efficient global optimization methods.The future of atomic simulation in catalysis is outlooked.

Generalized Mechanism for the Solid Phase Transition of M_(2)O_(3)(M=Al,Ga)Featuring Single Cation Migration and Martensitic Lattice Transformation

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.

Structure and Dynamics of Energy Materials from Machine Learning Simulations:A Topical Review

Energy materials featuring the capability to store and release chemical energy reversibly involve generally complex geometrical structures with multiple elements.It has been a great challenge to establish the quantitative relationship between the structure of materials and their dynamic physicochemical properties.In recent years,machine learning(ML)technique has demonstrated its great power in accelerating the research on energy materials.This topical review introduces the key ingredients and typical applications of ML to energy materials.We mainly focus on the ML based atomic simulation via ML potentials in different architectures/implementations,including high dimensional neural networks(HDNN),Gaussian approximation potential(GAP),moment tensor potentials(MTP)and stochastic surface walking global optimization with global neural network potential(SSW-NN)method.Three cases studies,namely,Si,LiC and LiTiO systems,are presented to demonstrate the ability of ML simulation in assessing the thermodynamics and kinetics of complex material systems.We highlight that the SSW-NN method provides an automated solution for global potential energy surface data collection,ML potential construction and ML simulation,which boosts the current ability for large-scale atomic simulation and thus holds the great promise for fast property evaluation and material discovery.