High-throughput First-principles Calculations Research Articles

Unpredicted but It Exists: Trigonal Sc2Ru with a Significant Metal-Metal Charge Transfer.

The Sc2Ru compound, obtained by high-temperature synthesis, was found to crystallize in a new trigonal hP45 structure type [space group P3̅m1; a = 9.3583(9) Å and c = 11.285(1) Å]: Ru@Sc8 cubes, Ru@Sc12 icosahedra, and uncommon Ru@Sc10 sphenocoronae are the building blocks of a unique motif tiling the whole crystal space. According to density functional theory studies, Sc2Ru is a metallic compound characterized by multicenter interactions: a significant charge transfer occurs from Sc to Ru, indicating an unexpectedly strong ionic character of the interactions between the two transition metals. Energy calculations support our experimental results in terms of stability of this compound, contributing to the recurrent discussion on the limits of the high-throughput first-principles calculations for metallic materials design.

JAMIP: an artificial-intelligence aided data-driven infrastructure for computational materials informatics

Materials informatics has emerged as a promisingly new paradigm for accelerating materials discovery and design. It exploits the intelligent power of machine learning methods in massive materials data from experiments or simulations to seek new materials, functionality, and principles, etc. Developing specialized facilities to generate, collect, manage, learn, and mine large-scale materials data is crucial to materials informatics. We herein developed an artificial-intelligence-aided data-driven infrastructure named Jilin Artificial-intelligence aided Materials-design Integrated Package (JAMIP), which is an open-source Python framework to meet the research requirements of computational materials informatics. It is integrated by materials production factory, high-throughput first-principles calculations engine, automatic tasks submission and monitoring progress, data extraction, management and storage system, and artificial intelligence machine learning based data mining functions. We have integrated specific features such as an inorganic crystal structure prototype database to facilitate high-throughput calculations and essential modules associated with machine learning studies of functional materials. We demonstrated how our developed code is useful in exploring materials informatics of optoelectronic semiconductors by taking halide perovskites as typical case. By obeying the principles of automation, extensibility, reliability, and intelligence, the JAMIP code is a promisingly powerful tool contributing to the fast-growing field of computational materials informatics.

Data driven discovery of conjugated polyelectrolytes for optoelectronic and photocatalytic applications

Conjugated polyelectrolytes (CPEs), comprised of conjugated backbones and pendant ionic functionalities, are versatile organic materials with diverse applications. However, the myriad of possible molecular structures of CPEs render traditional, trial-and-error materials discovery strategy impractical. Here, we tackle this problem using a data-centric approach by incorporating machine learning with high-throughput first-principles calculations. We systematically examine how key materials properties depend on individual structural components of CPEs and from which the structure–property relationships are established. By means of machine learning, we uncover structural features crucial to the CPE properties, and these features are then used as descriptors in the machine learning to predict the properties of unknown CPEs. Lastly, we discover promising CPEs as hole transport materials in halide perovskite-based optoelectronic devices and as photocatalysts for water splitting. Our work could accelerate the discovery of CPEs for optoelectronic and photocatalytic applications.

The atlas of ferroicity in two-dimensional MGeX3 family: Room-temperature ferromagnetic half metals and unexpected ferroelectricity and ferroelasticity

Two-dimensional (2D) ferromagnetic and ferroelectric materials attract unprecedented attention due to the spontaneous-symmetry-breaking induced novel properties and multifarious potential applications. Here we systematically investigate a large family (148) of 2D MGeX3 (M = metal elements, X = O/S/Se/Te) by means of the high-throughput first-principles calculations, and focus on their possible ferroic properties including ferromagnetism, ferroelectricity, and ferroelasticity. We discover eight stable 2D ferromagnets including five semiconductors and three half-metals, 21 2D antiferromagnets, and 11 stable 2D ferroelectric semiconductors including two multiferroic materials. Particularly, MnGeSe3 and MnGeTe3 are predicted to be room-temperature 2D ferromagnetic half metals with Tc of 490 and 308 K, respectively. It is probably for the first time that ferroelectricity is uncovered in 2D MGeX3 family, which derives from the spontaneous symmetry breaking induced by unexpected displacements of Ge-Ge atomic pairs, and we also reveal that the electric polarizations are in proportion to the ratio of electronegativity of X and M atoms, and IVB group metal elements are highly favored for 2D ferroelectricity. Magnetic tunnel junction and water-splitting photocatalyst based on 2D ferroic MGeX3 are proposed as examples of wide potential applications. The atlas of ferroicity in 2D MGeX3 materials will spur great interest in experimental studies and would lead to diverse applications.

A generally reliable model for composition-dependent lattice constants of substitutional solid solutions

Solid solutioning has long been employed to improve the performance of enegineering materials, the degree of improvement generally correlates closely with the resultant lattice parameters. It is therefore of great importance for materials design to describe accurately the composition-dependent lattice constants of the solid solutions (SSs). However, existing models could hardly reproduce the usually non-linear relationship between the compositions and the lattice constants. Herein, we present a new model within the framework of virtual crystal approximation by taking into account both the size factor and the electronic effect. The model takes inputs as simple as the fundamental property parameters of the elementary substances and N referential SSs for an N-component system, and can then predict the lattice constant of SS with any composition within the system. Systematical validation using datasets obtained from high-throughput first-principles calculations and available experiments confirmed the high reliability and general applicability of our model for various substitutional SSs. Applications and limitations of the model as well as outlooks were also discussed. It is expected that this model will deepen the understanding of the relationship between the composition and the properties of materials.

All-boron planar ferromagnetic structures: from clusters to monolayers.

Ferromagnetism in all-boron planar clusters is revealed based on high-throughput first-principles calculations. Magnetic boron clusters induced from p electrons have been confirmed with large spins, e.g., S = 3 in a B34 cluster, which can be assembled to construct all-boron ferromagnetic monolayers. Notably, the ferromagnetic semiconductors of boron monolayers can be designed with the hybridization of a nonmagnetic B36 cluster in experimental synthesis. The ferromagnetism-paramagnetism transition and semiconductor-metal transition in these boron nanostructures will occur around 500 K according to ab initio molecular dynamics simulation, indicating the potential applications in nano-devices at room temperature. The coexisting ferromagnetic and semiconducting properties in boron monolayers are attributed to the unique multicenter bonds together with the modulation of structural symmetry, which might be worth experimental attempts in the future.

High-throughput computation on relationship between composition and magnetic phase transition temperature of LaFe<sub>11.5</sub>Si<sub>1.5</sub>-based magnetic refrigeration materials

La(Fe,<i> </i>Si)<sub>13</sub>-based alloys have attracted more and more attention, for they exhibit giant magnetocaloric effects. In order to broaden their magnetic refrigeration temperatureranges, achieving a series of La(Fe,<i> </i>Si)<sub>13</sub>-based alloys with different magnetic phase transition temperatures is of great significance. Unlike the traditional research method, in this paper, a high-throughput first-principles computation is performed to estimate the magnetic phase transition temperature of the LaFe<sub>11.5</sub>Si<sub>1.5</sub>-based alloy by employing AMS-BAND software and the mean field theory. We investigate the effects of doping Mn, Co, Ni, Al atoms and Fe-vacancies on the magnetic phase transition temperature of LaFe<sub>11.5</sub>Si<sub>1.5</sub>-based alloy, and give the phase diagrams between the composition and magnetic phase transition temperature. The calculated results demonstrate that the magnetic phase transition temperature of the LaFe<sub>11.5</sub>Si<sub>1.5</sub>-based alloy increases with the increase of Co and Ni content. However, it shows an opposite result when Mn atom is doped. As for the LaFe<sub>11.5</sub>Si<sub>1.5</sub>-based alloy with the Fe-vacancies, the research results indicate that the absence of Fe atoms will reduce the magnetic phase transition temperature. Furthermore, when Mn, Co, Ni and Al atoms are doped in the alloys with Fe-vacancies, the variation tendency of the magnetic phase transition temperature with the change of the doping content is similar to that without the Fe-vacancies. Some estimated results are compared with the experimental or reported results, showing that they are in good agreement with each other. The PDOS and the magnetic moments of Fe atoms in the Mn, Co, Ni, Al-doped LaFe<sub>11.5</sub>Si<sub>1.5</sub>-based alloys are calculated, in which only the doping of Mn atoms can increase the magnetic moments of Fe atoms. Using the method of high-throughput first-principles calculation can effectively reduce the research cost and improve the working efficiency. In addition, it can provide technical support for the experimental selection of magnetocaloric materials with appropriate magnetic phase transition temperatures.

Theoretical prediction of solution in Sc<sub><i>x</i></sub>Y<sub>1–<i>x</i></sub> Fe<sub>2</sub> and order-disorder transitions in V<sub>2<i>x</i></sub> Fe<sub>2(1–<i>x</i>)</sub>Zr

Alloying is an important way to increase the diversity of material structure and properties. In this paper, we start from Ising model considering nearest neighbor interaction, in which a ferromagnetic system corresponds to a low temperature phase separation and high temperature solid solution of binary alloy, while antiferromagnetic system corresponds to a low temperature ordered solid solution and a high temperature disorder. The high-throughput first-principles calculation based on the structure recognition is realized by the program SAGAR (structures of alloy generation and recognition) developed by our research group. By considering the contribution of structural degeneracy to the partition function, theoretical prediction of alloy materials can be carried out at finite temperature. Taking hydrogen storage alloy (Sc<sub><i>x</i></sub>Y<sub>1–<i>x</i></sub> Fe<sub>2</sub> and V<sub>2<i>x</i></sub> Fe<sub>2(1–<i>x</i>)</sub>Zr) for example, the formation energy of ground state (at zero temperature) can be obtained by the first-principles calculations. It is found that the formation energy of Sc<sub><i>x</i></sub>Y<sub>1–<i>x</i></sub> Fe<sub>2</sub> is greater than zero, thereby inducing the phase separation at low temperature. The free energy will decrease with the temperature and concentration increasing, where the critical temperature of solid solution of alloy is determined according to the zero point of free energy. The formation energies of V<sub>2<i>x</i></sub> Fe<sub>2(1–<i>x</i>)</sub>Zr are all lower than zero, and the ordered phase occurs at low temperature. The order-disorder transition temperature of V<sub>0.5</sub>Fe<sub>1.5</sub>Zr and V<sub>1.5</sub>Fe<sub>0.5</sub>Zr are both about 100 K, while the transition temperature of VFeZr is nearly 50 K. The calculation process will effectively improve the high throughput screening efficiency of alloy, and also provide relevant theoretical reference for experimental research.

Ammonia Synthesis Using Single-Atom Catalysts Based on Two-Dimensional Organometallic Metal Phthalocyanine Monolayers under Ambient Conditions.

We have identified three novel metal phthalocyanine (MPc, M = Mo, Re, and Tc) single-atom catalyst candidates with excellent predicted performance for the production of ammonia from electrocatalytic nitrogen reduction reaction (NRR) through a combination of high-throughput screening and first-principles calculations on a series of 3d, 4d, and 5d transition metals anchored onto extended Pc monolayer catalysts. Analysis of the energy band structures and projected density of states of N2-MPc revealed significant orbital hybridization and charge transfer between the adsorbed N2 and catalyst MPc, which accounts for the high catalytic activity. Among 30 MPc catalysts, MoPc and TcPc monolayers were found to be the most promising new NRR catalysts, as they exhibit excellent stability, low onset potential, and high selectivity. A comprehensive reaction pathway search found that the maximum free energy changes for the MoPc and TcPc monolayers are 0.33 and 0.54 eV, respectively. As a distinctive nature of this work, the hybrid reaction pathway was considered extensively and searched systematically. The onset potential of the hybrid pathway is found to be smaller than or comparable to that of the commonly known pure pathway. Thus, the hybrid path is highly competitive with low onset potential and high activity. The hybrid pathway is expected to have an important impact on future research on the mechanism of NRR, and it will open up a new way to explore the mechanism of the NRR reaction. We hope that our work will provide impetus to the creation of new catalysts for reduction of N2 to NH3. This work provides new insights into the rational design of NRR catalysts and explores novel reaction pathways under ambient or mild conditions.

Non-hydrostatic pressure-dependent structural and transport properties of BiCuSeO and BiCuSO single crystals

High-pressure experiments usually expect a hydrostatic condition, in which the physical properties of materials can be easily understood by theoretical simulations. Unfortunately, non-hydrostatic effect is inevitable in experiments due to the solidification of the pressure transmitting media under high pressure. Resultantly, non-hydrostaticity affects the accuracy of the experimental data and sometimes even leads to false phenomena. Since the non-hydrostatic effect is extrinsic, it is quite hard to analyze quantitatively. Here, we have conducted high pressure experiments on the layered BiCuXO (X = S and Se) single crystals and quantitatively analyzed their pronounced non-hydrostatic effect by high throughput first-principles calculations and experimental Raman spectra. Our experiments find that the BiCuXO single crystals sustain the tetragonal structure up to 55 GPa (maximum pressure in our experiment). However, their pressure-dependent Raman shift and electric resistance show anomalous behaviors. Through optimization of thousands of crystal structures in the high throughput first-principles calculations, we have obtained the evolution of the lattice constants under external pressures, which clearly substantiates the non-hydrostatical pressure exerted in BiCuXO crystals. Our work indicates that the high throughput first-principles calculations could be a handy method to investigate the non-hydrostatic effect on the structural and electronic properties of materials in high pressure experiments.

Rapid screening of alloy elements to improve the elastic properties of dilute Pt-based alloys: High-throughput first-principles calculations and modeling

Platinum (Pt)-based dilute solid solutions are an important category of high-temperature alloys and bond coatings. In this study, the effects of 33 alloying elements on the mechanical and electronic properties of dilute Pt-based solid solutions are systematically investigated under atom relaxation and full relaxation using first-principles calculations based on density functional theory. The negative mixing enthalpy of Pt-dilute solid solutions means that the solubility of the solute elements in the Pt-based dilute alloys is energetically favorable at 0 K. Niobium, rhenium, and scandium are promising candidate elements for increasing the hardness and ductility of dilute Pt-based solid solutions. In addition, the electronic basis for the mechanical properties of Pt-dilute solid solutions is investigated in terms of the electronic density and mean bond population. The results demonstrate that the Pt–X bond lengths are shorter than the Pt–Pt bond length, resulting in greater hardness. Moreover, the model for the composition dependent elastic properties is built based on the CALPHAD approach, which will be used to the Pt-based multiphase alloys in the future. As certain alloying elements improve the hardness and ductility of Pt, this research expands our knowledge of the mechanism of dilute Pt-based solid solutions and provides a basis for next-generation superalloys or bond coatings at higher temperatures.

Atomic Wires of Transition Metal Chalcogenides: A Family of 1D Materials for Flexible Electronics and Spintronics.

As analogues of two-dimensional (2D) layered materials, searching for one-dimensional (1D) van der Waals wired materials as 1D Lego blocks for integration and device applications has been pursued. Motivated by the recently synthesized atomic wires of molybdenum chalcogenide, here we explored the structures and stability of 66 atomic wires of 3d, 4d, and 5d transition metal chalcogenides in the M6X6 stoichiometry (M = transition metal, X = chalcogen). After high-throughput first-principles calculations, 53 unprecedented and experimentally feasible M6X6 wires have been identified. Diverse functionalities are found in these 1D materials, including semiconductors, metals, and ferromagnets with high Young’s modulus and large fracture strain. Notably, six kinds of M6X6 wires are robust ferromagnets with Curie temperatures up to 700 K, which can be further elevated under axial strains. Moreover, these M6X6 atomic wires possess high stability and resistance to oxidation, humidity, and aggregation; both merits are desirable for device applications. This large family of 1D materials with definite structures and rich properties allows atomically precise integration for flexible electronics and spintronics.

Self-consistent assessment of Li+ ion cathodes: Theory vs. experiments

A self-consistent DFT framework without fitting parameters has been carried out to make assessments for the well-recognized cathode materials for lithium ion batteries. It offers a dependable framework for theoretical assessment of cathode materials. Transition metal oxide cathodes such as layered LiCoO 2 , spinel LiMn 2 O 4 and olivine LiFePO 4 have been commercialized for several decades and widely used in the rechargeable Li-ion batteries (LIBs). While great theoretical efforts have been made using the density functional theory (DFT) method, leading to insightful understanding covering materials stability and functional properties, the lack of consistency in choices of functionals and/or convergence criteria makes it somewhat difficult to compare results. It is therefore highly useful to assess these established systems towards self-consistency, thus offering a reliable working basis for theoretical formulation of novel cathodes. Here in this work, we have carried out systematic DFT calculations on the basis of recently established framework covering both thermodynamic stability, functional properties and associated mechanisms. Efforts have been made in self-consistent selection of exchange–correlation (XC) functionals in terms of dependable accuracy with affordable computational cost, which is essential for high-throughput first-principles calculations. The outcome of the current work on three established cathode systems is in very good agreement with experimental data, and the methodology is to provide a solid basis for designing novel cathode materials without using costing non-local exchange–correlation functionals for structure-energy calculations.

Integrating data mining and machine learning to discover high-strength ductile titanium alloys

Based on the growing power of computational capabilities and algorithmic developments, with the help of data-driven and high-throughput calculations, a new paradigm accelerating materials discovery, design and optimization is emerging. Titanium (Ti) alloys have been chosen herein to highlight an integrated computational materials engineering case study with the aim of improving their strength and ductility. The electronic properties of elemental building blocks were derived from high-throughput first-principles calculations and presented in the form of the Mendeleev periodic table, including their electron work function (Ф), Fermi energy (EF), bonding charge density (Δρ), and lattice distortion energy. The atomic and electronic insights of the composition–structure–property relationships were revealed by a data mining approach, addressing the key features/principles for the design strategies of advanced alloys. Guided by defect engineering, the deformation fault energy and dislocation width were treated as the dominating criteria in improving the ductility. The proposed yield strength model was utilized quantitatively to present the contributions of solid-solution strengthening and grain refinement hardening. Machine learning was used collaboratively with fundamental knowledge and feed back into a new training model, shown to be superior to the empirical molybdenum equivalence method. The results draw a conclusion that the integration of data mining and machine learning will not only generate plausible explanations and address new hypotheses, but also enable the design of strong and ductile Ti alloys in a more efficient and cost-effective way.

High-throughput screening of two-dimensional materials and prediction of photocatalytic performances

<p indent=0mm>In recent years, a variety of excellent two-dimensional (2D) materials have been successfully prepared by mechanical isolation, liquid-phase exfoliation and molecular beam epitaxial growth, which greatly promotes the exploration of the special properties of 2D materials. Traditional “trail and error methods” are facing challenges such as low efficiency, high cost and lack of guidance, so it is extremely difficult to explore novel 2D materials. In contrast, first-principles calculations can not only compute and simulate a large number of new 2D materials to predict their structure and physical properties, but also design their possible synthesis paths and methods by referring to the predicted properties. With the continuous improvement of supercomputers and the establishment of a large number of material databases, potential 2D structures can be screened out by using different algorithms, and targeted design of 2D materials can be achieved by combining first-principles calculations. The first-principles calculations can simulate the structure of 2D materials, predict their stability, and rationally design the synthesis route according to the simulation results. At present, many 2D material structures have been predicted by theoretical calculations and many of them have been synthesized by experiments. For example, the graphene analogue silicene has attracted much attention. The results of first-principles calculations show that itexhibitsa buckled honeycomb structure. Based on the calculation of phonon spectrum and molecular dynamics simulation, the structure is stable below<sc>1000 K.</sc> The calculation results show that Si can be well adsorbed on the Ag(110) surface and silicene was successfully obtained on the Ag(111) surface several years later. Compared with the traditional “trial and error method”, although the first-principles calculations can accelerate the discovery of 2D materials, it still does not get rid of the constraints of “intuition” and causes a waste of resources. First-principles calculations heavily rely on scientific intutition, which would lead to the waste of calcualtion resources. The emergence of high-throughput computing can reduce the waste of computing resources. It provides an effective method for the discovery and screening of novel 2D materials. Through various algorithms, novel 2D materials are excavated from theconstructed databases, such as inorganic crystal structure database, crystal open database, Materials Project database, etc. High-throughput screening combined with first-principles calculations can realize the targeted design of materials. For example, through high-throughput calculations, potential 2D materials are screened from materials databases. After that, the stability, band gap, carrier mobility, light absorption and the position of band edge for the screened 2D materials can be predicted by first-principles calculations. Therefore, the design of 2D photocatalytic materials with guidance and the evaluation of photocatalytic performance greatly promote the development of 2D photocatalytic materials. Based on high-throughput screening and first-principles calculations, predicting the structural and physical properties of 2D materials exhibits advantages that traditional experiments can not achieve. These advantages have greatly promoted the discovery and research of 2D materials. We introduced the methods of theoretical calculation and summarized the research results of structure design and performance prediction of 2D materials. Finally, based on these research advances, the future development of this field is prospected.

Machine learning models for predicting the dielectric constants of oxides based on high-throughput first-principles calculations

Prediction models of both the electronic and ionic contributions to the static dielectric constants have been constructed using data from density functional perturbation theory calculations of approximately 1200 metal oxides via supervised machine learning. We developed two types of random forest regression models for oxides with the ground-state crystal structures: one model requires only compositional information and the other model also uses structural information. Although the training data included various atomic frameworks, the prediction models performed well even when only compositional information was used as feature descriptors. In prediction of the electronic contributions to the dielectric constants, the accuracies of the regression models with and without structural information were comparable, while the structural descriptors more clearly improved the prediction accuracy for the ionic contributions. We also analyzed the feature importance for prediction of the dielectric constants. The mean atomic mass and mass density were determined to be significant features in prediction of the electronic contributions without and with structural information, respectively. The standard deviation of the principal quantum number and mean neighbor distance variation were found to be important for the respective prediction models of the ionic contributions. The correlations between the dielectric constants and these features are discussed, along with the underlying physical mechanisms.

Balance between strength and ductility of dilute Fe2B by high-throughput first-principles calculations

In boride ceramics, Fe2B typically functions as the primary wear-resistant phase owing to its high hardness, but the elastic and ductile-brittle properties of Fe2B greatly influence its wear resistance and industrial application. First-principles calculations using density functional theory (DFT) were utilized to systematically investigate the effects of Al, Si, and transition elements (including 3d, 4d, and 5d transition elements) on the mechanical properties and electronic structure of Fe31MB16 with a dilute solution model. The bulk modulus (B), shear modulus (G), and Young's modulus (E) were calculated through the stress-strain method. The Vickers hardness (HV) and fracture toughness (KIC) were calculated by semi-empirical models. The electronic structures and chemical bonding were analyzed using the charge density difference, electron localization function, and Mulliken bond population. Overall, Fe31YB16 had the largest moduli (B, G, and E) among all the Fe31MB16 compounds that were investigated, while Fe31CrB16 had the highest hardness. Compared with pure Fe2B, the ductility of Fe31MB16 improved to varying degrees according to the B/G and Poisson's ratio (σ) criteria, respectively. However, adding alloying elements in this concentration did not change the intrinsic brittleness of Fe2B. Moreover, the trends in the calculated KIC of Fe31MB16 (M = Cr, Mn, Mo, and W respectively) were consistent with those of experiments. The electronic structure and Mulliken population analysis showed that the differences in the mechanical properties of the Fe31MB16 compounds were primarily determined by the M-B and B–B bonds because the alloying element M replaced Fe. These results provide guidance for improving the ductility of Fe2B and its industrial application.

ALKEMIE: An intelligent computational platform for accelerating materials discovery and design

Developing new materials with target properties via the traditional trial-and-error ways is cost-inefficient, and sometimes ends up with fruitlessness, therefore, simulation-driven materials design plays an important role in the past decades. Nevertheless, the advent of the era of data-driven material science requires an intelligent computational platform to accelerate the discovery and design for advanced materials. Here, we present an open-source computational platform named as ALKEMIE, acronyms for Artificial Learning and Knowledge Enhanced Materials Informatics Engineering, which enables easy access of data-driven techniques to broad communities. ALKEMIE is incorporated with three key components for the computational design of materials for the forthcoming data-driven sciences: data generation via high-throughput calculations, data management and data mining via machine learning models. Briefly speaking, the high-throughput calculations in ALKEMIE are implemented through the integration of automatic frameworks of model constructions, calculation performances and data analysis. And the used high-level application programming interface for the database makes the data mining through machine learning more applicable in material science. In particular, ALKEMIE is integrated with a module for the generation of machine-learned interatomic potential for large-scale molecular dynamic simulations where the dataset is obtained from high-throughput first-principles calculations. More importantly, ALKEMIE has an elaborately designed user-friendly graphical user-interface which makes the workflow and dataflow more maneuverable and transparent, facilitating its easy-to-use for scientists with broad backgrounds. Finally, the main characters of ALKEMIE are demonstrated using three computational examples.

Explainable machine learning for materials discovery: predicting the potentially formable Nd-Fe-B crystal structures and extracting the structure-stability relationship.

New Nd-Fe-B crystal structures can be formed via the elemental substitution of LA-T-X host structures, including lanthanides (LA), transition metals (T) and light elements, X = B, C, N and O. The 5967 samples of ternary LA-T-X materials that are collected are then used as the host structures. For each host crystal structure, a substituted crystal structure is created by substituting all lanthanide sites with Nd, all transition metal sites with Fe and all light-element sites with B. High-throughput first-principles calculations are applied to evaluate the phase stability of the newly created crystal structures, and 20 of them are found to be potentially formable. A data-driven approach based on supervised and unsupervised learning techniques is applied to estimate the stability and analyze the structure-stability relationship of the newly created Nd-Fe-B crystal structures. For predicting the stability for the newly created Nd-Fe-B structures, three supervised learning models: kernel ridge regression, logistic classification and decision tree model, are learned from the LA-T-X host crystal structures; the models achieved maximum accuracy and recall scores of 70.4 and 68.7%, respectively. On the other hand, our proposed unsupervised learning model based on the integration of descriptor-relevance analysis and a Gaussian mixture model achieved an accuracy and recall score of 72.9 and 82.1%, respectively, which are significantly better than those of the supervised models. While capturing and interpreting the structure-stability relationship of the Nd-Fe-B crystal structures, the unsupervised learning model indicates that the average atomic coordination number and coordination number of the Fe sites are the most important factors in determining the phase stability of the new substituted Nd-Fe-B crystal structures.

Two-Dimensional Dirac Semimetals without Inversion Symmetry.

Realizing stable two-dimensional (2D) Dirac points against spin-orbit coupling (SOC) has attracted much attention because it provides a platform to study the unique transport properties. In previous work, Young and Kane [Phys. Rev. Lett. 115, 126803 (2015)PRLTAO0031-900710.1103/PhysRevLett.115.126803 proposed stable 2D Dirac points with SOC, in which the Berry curvature and edge states vanish due to the coexistence of inversion and time-reversal symmetries. Herein, using the tight-binding model and k·p effective Hamiltonian, we present that 2D Dirac points can survive in the presence of SOC without inversion symmetry. Such 2D Dirac semimetals possess nonzero Berry curvature near the crossing nodes, and two edge states are terminated at one pair of Dirac points. In addition, according to symmetry arguments and high-throughput first-principles calculations, we identify a family of ideal 2D Dirac semimetals, which has nonzero Berry curvature in the vicinity of Dirac points and visible edge states, thus facilitating the experimental observations. Our work shows that 2D Dirac points can emerge without inversion symmetry, which not only enriches the classification of 2D topological semimetals but also provides a promising avenue to observe exotic transport phenomena beyond graphene, e.g., nonlinear Hall effect.