Coupling a Crystal Graph Multilayer Descriptor to Active Learning for Rapid Discovery of 2D Ferromagnetic Semiconductors/Half-Metals/Metals

From bulk effective mass to 2D carrier mobility accurate prediction via adversarial transfer learning

Data scarcity is one of the critical bottlenecks to utilizing machine learning in material discovery. Transfer learning can use existing big data to assist property prediction on small data sets, but the premise is that there must be a strong correlation between large and small data sets. To extend its applicability in scenarios with different properties and materials, here we develop a hybrid framework combining adversarial transfer learning and expert knowledge, which enables the direct prediction of carrier mobility of two-dimensional (2D) materials using the knowledge learned from bulk effective mass. Specifically, adversarial training ensures that only common knowledge between bulk and 2D materials is extracted while expert knowledge is incorporated to further improve the prediction accuracy and generalizability. Successfully, 2D carrier mobilities are predicted with the accuracy over 90% from only crystal structure, and 21 2D semiconductors with carrier mobilities far exceeding silicon and suitable bandgap are successfully screened out. This work enables transfer learning in simultaneous cross-property and cross-material scenarios, providing an effective tool to predict intricate material properties with limited data.

Strain-Induced Ferromagnetism in Monolayer T″-Phase VTe2: Unveiling Magnetic States and Anisotropy for Spintronics Advancement

Two-dimensional (2D) ferromagnets have attracted significant interest for their potential in spintronic device miniaturization, especially since the discovery of ferromagnetic ordering in monolayer materials such as CrI3 and Fe3GeTe2 in 2017. This study presents a detailed investigation into the effects of the Hubbard U parameter, biaxial strain, and structural distortions on the magnetic characteristics of T″-phase VTe2. We demonstrate that setting the Hubbard U to 0 eV provides an accurate representation of the observed structural, magnetic, and electronic features for both bulk and monolayer T″-phase VTe2. The application of strain reveals two distinct ferromagnetic states in the monolayer T″-phase VTe2, each characterized by minor structural differences, but notably different magnetic moments. The T″-1 state, with reduced magnetic moments, emerges under compressive strain, while the T″-2 state, featuring increased magnetic moments, develops under tensile strain. Our analysis also compares the magnetic anisotropy between the T and T″ phases of VTe2, highlighting that the periodic lattice distortion in the T″-phase induces an in-plane anisotropy, which makes it a material with an easy-axis of magnetization. Monte Carlo simulations corroborate our findings, indicating a high Curie temperature of approximately 191 K for the T″-phase VTe2. Our research not only sheds light on the critical aspects of the VTe2 system but also suggests new pathways for enhancing low-dimensional magnetism, contributing to the advancement of spintronics and straintronics.

When Machine Learning Meets 2D Materials: A Review

The availability of an ever-expanding portfolio of 2D materials with rich internal degrees of freedom (spin, excitonic, valley, sublattice, and layer pseudospin) together with the unique ability to tailor heterostructures made layer by layer in a precisely chosen stacking sequence and relative crystallographic alignments, offers an unprecedented platform for realizing materials by design. However, the breadth of multi-dimensional parameter space and massive data sets involved is emblematic of complex, resource-intensive experimentation, which not only challenges the current state of the art but also renders exhaustive sampling untenable. To this end, machine learning, a very powerful data-driven approach and subset of artificial intelligence, is a potential game-changer, enabling a cheaper - yet more efficient - alternative to traditional computational strategies. It is also a new paradigm for autonomous experimentation for accelerated discovery and machine-assisted design of functional 2D materials and heterostructures. Here, the study reviews the recent progress and challenges of such endeavors, and highlight various emerging opportunities in this frontier research area.

MatGPT: A Vane of Materials Informatics from Past, Present, to Future

Combining materials science, artificial intelligence (AI), physical chemistry, and other disciplines, materials informatics is continuously accelerating the vigorous development of new materials. The emergence of "GPT (Generative Pre-trained Transformer) AI" shows that the scientific research field has entered the era of intelligent civilization with "data" as the basic factor and "algorithm + computing power" as the core productivity. The continuous innovation of AI will impact the cognitive laws and scientific methods, and reconstruct the knowledge and wisdom system. This leads to think more about materials informatics. Here, a comprehensive discussion of AI models and materials infrastructures is provided, and the advances in the discovery and design of new materials are reviewed. With the rise of new research paradigms triggered by "AI for Science", the vane of materials informatics: "MatGPT", is proposed and the technical path planning from the aspects of data, descriptors, generative models, pretraining models, directed design models, collaborative training, experimental robots, as well as the efforts and preparations needed to develop a new generation of materials informatics, is carried out. Finally, the challenges and constraints faced by materials informatics are discussed, in order to achieve a more digital, intelligent, and automated construction of materials informatics with the joint efforts of more interdisciplinary scientists.

Universal machine learning aided synthesis approach of two-dimensional perovskites in a typical laboratory

The past decade has witnessed the significant efforts in novel material discovery in the use of data-driven techniques, in particular, machine learning (ML). However, since it needs to consider the precursors, experimental conditions, and availability of reactants, material synthesis is generally much more complex than property and structure prediction, and very few computational predictions are experimentally realized. To solve these challenges, a universal framework that integrates high-throughput experiments, a priori knowledge of chemistry, and ML techniques such as subgroup discovery and support vector machine is proposed to guide the experimental synthesis of materials, which is capable of disclosing structure-property relationship hidden in high-throughput experiments and rapidly screening out materials with high synthesis feasibility from vast chemical space. Through application of our approach to challenging and consequential synthesis problem of 2D silver/bismuth organic-inorganic hybrid perovskites, we have increased the success rate of the synthesis feasibility by a factor of four relative to traditional approaches. This study provides a practical route for solving multidimensional chemical acceleration problems with small dataset from typical laboratory with limited experimental resources available.

Material symmetry recognition and property prediction accomplished by crystal capsule representation

Learning the global crystal symmetry and interpreting the equivariant information is crucial for accurately predicting material properties, yet remains to be fully accomplished by existing algorithms based on convolution networks. To overcome this challenge, here we develop a machine learning (ML) model, named symmetry-enhanced equivariance network (SEN), to build material representation with joint structure-chemical patterns, to encode important clusters embedded in the crystal structure, and to learn pattern equivariance in different scales via capsule transformers. Quantitative analyses of the intermediate matrices demonstrate that the intrinsic crystal symmetries and interactions between clusters have been exactly perceived by the SEN model and critically affect the prediction performances by reducing effective feature space. The mean absolute errors (MAEs) of 0.181 eV and 0.0161 eV/atom are obtained for predicting bandgap and formation energy in the MatBench dataset. The general and interpretable SEN model reveals the potential to design ML models by implicitly encoding feature relationship based on physical mechanisms.

Machine-Learning-Enabled Framework in Engineering Plastics Discovery: A Case Study of Designing Polyimides with Desired Glass-Transition Temperature

Great and continuous efforts have been made to discover high-performance engineering plastics with specific properties to replace traditional engineering materials in many fields. The utilization of machine learning (ML) has brought more opportunities for the discovery of high-performing engineering plastics. However, hindered by either the relatively small database or a lack of accurate structure descriptors with clear physical and chemical meanings relating to polymer properties, the current ML studies show some flaws in the accuracy and efficiency in polymer development. Herein, we collected a dataset of 878 polyimides (PI), one of the best engineering plastics, with experimentally measured glass-transition temperature (Tg) values, and developed a rapid and accurate ML approach to design PI candidates with the desired Tg value. After the conversion from PI structures into "mechanically identifiable" SMILES (Simplified molecular input line entry system) language, the eight most critical descriptors were ultimately obtained by multiple analysis methods. The physiochemical meaning of the key descriptors was further analyzed carefully to translate the implicit "machine language" to chemical knowledge. The artificial neural network (ANN)-based model gave the most accurate results with a root-mean-square error of ∼11 K among the studied ML methods. More importantly, three potential PI candidates with desired Tg (DPIs) were designed according to the chemical insight of the key descriptors, which were then verified by experiments. The experimental and predicted Tg values of DPIs have an acceptable average deviation of ca. 3.66%. This accuracy has reached the level of the traditional molecular simulation, but the time consumption and hold-up computing resource are tremendously reduced. Furthermore, the current ML approach could offer a scalable and adaptable framework in future engineer plastics innovation.

Adaptive Design of Alloys for CO2 Activation and Methanation via Reinforcement Learning Monte Carlo Tree Search Algorithm

Data-driven machine learning (ML) has earned remarkable achievements in accelerating materials design, while it heavily relies on high-quality data acquisition. In this work, we develop an adaptive design framework for searching for optimal materials starting from zero data and with as few DFT calculations as possible. This framework integrates automatic density functional theory (DFT) calculations with an improved Monte Carlo tree search via reinforcement learning algorithm (MCTS-PG). As a successful example, we apply it to rapidly identify the desired alloy catalysts for CO₂ activation and methanation within 200 MCTS-PG steps. To this end, seven alloy surfaces with high theoretical activity and selectivity for CO₂ methanation are screened out and further validated by comprehensive free energy calculations. Our adaptive design framework enables the fast computational exploration of materials with desired properties via minimal DFT calculations.

Descriptor engineering in machine learning regression of electronic structure properties for 2D materials

We build new material descriptors to predict the band gap and the work function of 2D materials by tree-based machine-learning models. The descriptor's construction is based on vectorizing property matrices and on empirical property function, leading to mixing features that require low-resource computations. Combined with database-based features, the mixing features significantly improve the training and prediction of the models. We find R[Formula: see text] greater than 0.9 and mean absolute errors (MAE) smaller than 0.23 eV both for the training and prediction. The highest R[Formula: see text] of 0.95, 0.98 and the smallest MAE of 0.16 eV and 0.10 eV were obtained by using extreme gradient boosting for the bandgap and work-function predictions, respectively. These metrics were greatly improved as compared to those of database features-based predictions. We also find that the hybrid features slightly reduce the overfitting despite a small scale of the dataset. The relevance of the descriptor-based method was assessed by predicting and comparing the electronic properties of several 2D materials belonging to new classes (oxides, nitrides, carbides) with those of conventional computations. Our work provides a guideline to efficiently engineer descriptors by using vectorized property matrices and hybrid features for predicting 2D materials properties via ensemble models.

Large tunneling magnetoresistance in van der Waals magnetic tunnel junctions based on FeCl2 films with interlayer antiferromagnetic couplings

Antiferromagnets (AFMs) are some of the most promising candidates for next-generation magnetic memory technology owing to their advantages over conventional ferromagnets (FMs), such as zero stray field and THz-range magnetic resonance frequency. Motivated by the recent synthesis of FeCl₂ films with interlayer AFM and intralayer FM couplings, we investigated the magnetic properties of few-layer FeCl₂ and the spin-dependent transmissions of graphite/bilayer FeCl₂/graphite and Au/n-layer FeCl₂/Au magnetic tunnel junctions (MTJs) using first-principles calculations combined with the nonequilibrium Green's function. The interlayer AFM coupling of FeCl₂ is certified to be stable and independent of the stacking orders and relative displacement between layers. Furthermore, based on the Au electrode with better conductive performance than the graphite electrode and monolayer 1T-FeCl₂ with complete spin polarization, high Curie temperature and large magnetic anisotropic energy, a high tunnel magnetoresistance (TMR) ratio of 2.7 × 10³% is achieved in Au/bilayer FeCl₂/Au MTJs at zero bias and it increases with different layers of FeCl₂ (n = 2-10). These excellent spin transport properties of Au/n-layer FeCl₂/Au MTJs based on two-dimensional (2D) AFM barriers with out-of-plane magnetization directions suggest their great potential for application in high-reliability, high-speed and high-density spintronic devices.

Atomic scale insights into the epitaxial growth mechanism of 2D Cr3Te4 on mica

Two-dimensional (2D) magnetic materials are of wide research interest owing to their promising applications in spintronic devices. Among them, chromium chalcogenide compounds are some of the limited available systems that present both high stability in air and high Curie temperatures. Epitaxial growth techniques based on chemical vapour deposition (CVD) have been demonstrated to be a robust method for growing 2D non-layered chromium chalcogenides. However, the growth mechanism is not well-understood. Here, we demonstrate the epitaxial growth of Cr₃Te₄ nanoplates with high quality on mica. Atomic-resolution scanning transmission electron microscopy (STEM) imaging reveals that the epitaxial growth is based on nanosized chromium oxide seed particles at the interface of Cr₃Te₄ and mica. The chromium oxide nanoparticle exhibits a coherent interface with both mica and Cr₃Te₄ with a lattice mismatch within 3%, suggesting that, as a buffer layer, chromium oxide can release the interfacial strain, and induce the growth of Cr₃Te₄ although there is a distinct oxygen-content difference between mica and Cr₃Te₄. This work provides an experimental understanding behind the epitaxial growth of 2D magnetic materials at the atomic scale and facilitates the improvement of their growth procedures for devices with high crystalline quality.

Clear-Box Machine Learning for Virtual Screening of 2D Nanozymes to Target Tumor Hydrogen Peroxide

Targeting tumor hydrogen peroxide (H₂ O₂ ) with catalytic materials has provided a novel chemotherapy strategy against solid tumors. Because numerous materials have been fabricated so far, there is an urgent need for an efficient in silico method, which can automatically screen out appropriate candidates from materials libraries for further therapeutic evaluation. In this work, adsorption-energy-based descriptors and criteria are developed for the catalase-like activities of materials surfaces. The result enables a comprehensive prediction of H₂ O₂ -targeted catalytic activities of materials by density functional theory (DFT) calculations. To expedite the prediction, machine learning models, which efficiently calculate the adsorption energies for 2D materials without DFT, are further developed. The finally obtained method takes advantage of both interpretability of physics model and high efficiency of machine learning. It provides an efficient approach for in silico screening of 2D materials toward tumor catalytic therapy, and it will greatly promote the development of catalytic nanomaterials for medical applications.

Graph neural networks for materials science and chemistry

Machine learning plays an increasingly important role in many areas of chemistry and materials science, being used to predict materials properties, accelerate simulations, design new structures, and predict synthesis routes of new materials. Graph neural networks (GNNs) are one of the fastest growing classes of machine learning models. They are of particular relevance for chemistry and materials science, as they directly work on a graph or structural representation of molecules and materials and therefore have full access to all relevant information required to characterize materials. In this Review, we provide an overview of the basic principles of GNNs, widely used datasets, and state-of-the-art architectures, followed by a discussion of a wide range of recent applications of GNNs in chemistry and materials science, and concluding with a road-map for the further development and application of GNNs.

Tunneling transport of 2D anisotropic XC (X = P, As, Sb, Bi) with a direct band gap and high mobility: a DFT coupled with NEGF study

Direct bandgap and significant anisotropic properties are crucial and beneficial for nanoelectronic applications. In this work, through first-principles calculations, we investigate novel two-dimensional (2D) α-XC (X = P, As, Sb, Bi) materials, which possess a direct bandgap of 0.73 to 1.40 eV with remarkable anisotropic electronic properties. Intriguingly, 2D α-XC presents the highest electron mobility near 8 × 10³ cm² V^-1 s^-1 along the Γ-X direction. Moreover, the transfer characteristics of the 2D α-XC TFETs are thoroughly assessed through NEGF methods. AsC TFETs demonstrate an on-state current larger than 2.2 × 10³ μA μm^-1, which can satisfy the International Technology Roadmap for Semiconductors (ITRS) for high-performance requirements. In particular, the minimum value of subthreshold swing of devices is as low as 15 mV dec^-1, indicating excellent device switching characteristics. The relevant calculation results show that 2D α-XC monolayers could be a promising candidate in next-generation high-performance device applications.

Computational prediction of new magnetic materials

The discovery of new magnetic materials is a big challenge in the field of modern materials science. We report the development of a new extension of the evolutionary algorithm USPEX, enabling the search for half-metals (materials that are metallic only in one spin channel) and hard magnetic materials. First, we enabled the simultaneous optimization of stoichiometries, crystal structures, and magnetic structures of stable phases. Second, we developed a new fitness function for half-metallic materials that can be used for predicting half-metals through an evolutionary algorithm. We used this extended technique to predict new, potentially hard magnets and rediscover known half-metals. In total, we report five promising hard magnets with high energy product (|BH|_MAX), anisotropy field (H_a), and magnetic hardness (κ) and a few half-metal phases in the Cr-O system. A comparison of our predictions with experimental results, including the synthesis of a newly predicted antiferromagnetic material (WMnB₂), shows the robustness of our technique.

Transition-Metal Interlink Neural Network: Machine Learning of 2D Metal-Organic Frameworks with High Magnetic Anisotropy

Two-dimensional (2D) metal-organic framework (MOF) materials with large perpendicular magnetic anisotropy energy (MAE) are important candidates for high-density magnetic storage. The MAE-targeted high-throughput screening of 2D MOFs is currently limited by the time-consuming electronic structure calculations. In this study, a machine learning model, namely, transition-metal interlink neural network (TMINN) based on a database with 1440 2D MOF materials is developed to quickly and accurately predict MAE. The well-trained TMINN model for MAE successfully captures the general correlation between the geometrical configurations and the MAEs. We explore the MAEs of 2583 other 2D MOFs using our trained TMINN model. From these two databases, we obtain 11 unreported 2D ferromagnetic MOFs with MAEs over 35 meV/atom, which are further demonstrated by the high-level density functional theory calculations. Such results show good performance of the extrapolation predictions of TMINN. We also propose some simple design rules to acquire 2D MOFs with large MAEs by building a Pearson correlation coefficient map between various geometrical descriptors and MAE. Our developed TMINN model provides a powerful tool for high-throughput screening and intentional design of 2D magnetic MOFs with large MAE.

A Universal Descriptor for Complicated Interfacial Effects on Electrochemical Reduction Reactions

Supported catalysts have exhibited excellent performance in various reactions. However, the rational design of supported catalysts with high activity and certain selectivity remains a great challenge because of the complicated interfacial effects. Using recently emerged two-dimensional materials supported dual-atom catalysts (DACs@2D) as a prototype, we propose a simple and universal descriptor based on inherent atomic properties (electronegativity, electron type, and number), which can well evaluate the complicated interfacial effects on the electrochemical reduction reactions (i.e., CO₂, O₂, and N₂ reduction reactions). Based on this descriptor, activity and selectivity trends in CO₂ reduction reaction are successfully elucidated, in good agreement with available experimental data. Moreover, several potential catalysts with superior activity and selectivity for target products are predicted, such as CuCr/g-C₃N₄ for CH₄ and CuSn/N-BN for HCOOH. More importantly, this descriptor can also be extended to evaluate the activity of DACs@2D for O₂ and N₂ reduction reactions, with very small errors between the prediction and reported experimental/computational results. This work provides feasible principles for the rational design of advanced electrocatalysts and the construction of universal descriptors based on inherent atomic properties.

Coexistence of Semiconducting Ferromagnetics and Piezoelectrics down 2D Limit from Non van der Waals Antiferromagnetic LiNbO3-Type FeTiO3

Stable two-dimensional (2D) ferromagnetic semiconductors (FMSs) with multifunctional properties have attracted extensive attention in device applications. Non van der Waals (vdW) transition-metal oxides with excellent environmental stability, if ferromagnetic (FM), may open up an unconventional and promising avenue for this subject, but they are usually antiferromagnetic or ferrimagnetic. Herein, we predict an FMS, monolayer Fe₂Ti₂O₉, which can be obtained from LiNbO₃-type FeTiO₃ antiferromagnetic bulk, has a moderate band gap of 0.87 eV, large perpendicular magnetization (6 μ_B/fu) and a Curie temperature up to 110 K. The intriguing magnetic properties are derived from the double exchange and negative charge transfer between O_p orbitals and Fe_d orbitals. In addition, a large in-plane piezoelectric (PE) coefficient d₁₁ of 5.0 pm/V is observed. This work offers a competitive candidate for multifunctional spintronics and may stimulate further experimental exploration of 2D non-vdW magnets.

Graph-based discovery and analysis of atomic-scale one-dimensional materials

Recent decades have witnessed an exponential growth in the discovery of low-dimensional materials (LDMs), benefiting from our unprecedented capabilities in characterizing their structure and chemistry with the aid of advanced computational techniques. Recently, the success of two-dimensional compounds has encouraged extensive research into one-dimensional (1D) atomic chains. Here, we present a methodology for topological classification of structural blocks in bulk crystals based on graph theory, leading to the identification of exfoliable 1D atomic chains and their categorization into a variety of chemical families. A subtle interplay is revealed between the prototypical 1D structural motifs and their chemical space. Leveraging the structure graphs, we elucidate the self-passivation mechanism of 1D compounds imparted by lone electron pairs, and reveal the dependence of the electronic band gap on the cationic percolation network formed by connections between structure units. This graph-theory-based formalism could serve as a source of stimuli for the future design of LDMs.

VSi2P4/FeCl2 van der Waals heterostructure: A two-dimensional reconfigurable magnetic diode

The reconfigurable magnetic tunnel diode has recently been proposed as a promising approach to decrease the base collector leakage currents. However, conventional bulk interfaces usually suffer from strong Fermi level...

Air-Stable 2D Cr5 Te8 Nanosheets with Thickness-Tunable Ferromagnetism

2D magnetic materials have aroused widespread research interest owing to their promising application in spintronic devices. However, exploring new kinds of 2D magnetic materials with better stability and realizing their batch synthesis remain challenging. Herein, the synthesis of air-stable 2D Cr₅ Te₈ ultrathin crystals with tunable thickness via tube-in-tube chemical vapor deposition (CVD) growth technology is reported. The importance of tube-in-tube CVD growth, which can significantly suppress the equilibrium shift to the decomposition direction and facilitate that to the synthesis reaction direction, for the synthesis of high-quality Cr₅ Te₈ with accurate composition, is highlighted. By precisely adjusting the growth temperature, the thickness of Cr₅ Te₈ nanosheets is tuned from ≈1.2 nm to tens of nanometers, with the morphology changing from triangles to hexagons. Furthermore, magneto-optical Kerr effect measurements reveal that the Cr₅ Te₈ nanosheet is ferromagnetic with strong out-of-plane spin polarization. The Curie temperature exhibits a monotonic increase from 100 to 160 K as the Cr₅ Te₈ thickness increases from 10 to 30 nm and no apparent variation in surface roughness or magnetic properties after months of exposure to air. This study provides a robust method for the controllable synthesis of high-quality 2D ferromagnetic materials, which will facilitate research progress in spintronics.

Bayesian Machine Learning for Efficient Minimization of Defects in ALD Passivation Layers

Atomic layer deposition (ALD) is an enabling technology for encapsulating sensitive materials owing to its high-quality, conformal coating capability. Finding the optimum deposition parameters is vital to achieving defect-free layers; however, the high dimensionality of the parameter space makes a systematic study on the improvement of the protective properties of ALD films challenging. Machine-learning (ML) methods are gaining credibility in materials science applications by efficiently addressing these challenges and outperforming conventional techniques. Accordingly, this study reports the ML-based minimization of defects in an ALD-Al₂O₃ passivation layer for the corrosion protection of metallic copper using Bayesian optimization (BO). In all experiments, BO consistently minimizes the layer defect density by finding the optimum deposition parameters in less than three trials. Electrochemical tests show that the optimized layers have virtually zero film porosity and achieve five orders of magnitude reduction in corrosion current as compared to control samples. Optimized parameters of surface pretreatment using Ar/H₂ plasma, the deposition temperature above 200 °C, and 60 ms pulse time quadruple the corrosion resistance. The significant optimization of ALD layers presented in this study demonstrates the effectiveness of BO and its potential outreach to a broader audience, focusing on different materials and processes in materials science applications.

Computational Discovery of New 2D Materials Using Deep Learning Generative Models

Two-dimensional (2D) materials have emerged as promising functional materials with many applications such as semiconductors and photovoltaics because of their unique optoelectronic properties. Although several thousand 2D materials have been screened in existing materials databases, discovering new 2D materials remains challenging. Herein, we propose a deep learning generative model for composition generation combined with a random forest-based 2D materials classifier to discover new hypothetical 2D materials. Furthermore, a template-based element substitution structure prediction approach is developed to predict the crystal structures of a subset of the newly predicted hypothetical formulas, which allows us to confirm their structure stability using DFT calculations. So far, we have discovered 267 489 new potential 2D materials compositions, where 1485 probability scores are more then 0.95. Among them, we have predicted 101 crystal structures and confirmed 92 2D/layered materials by DFT formation energy calculation. Our results show that generative machine learning models provide an effective way to explore the vast chemical design space for new 2D materials discovery.

Size Doesn't Matter: Predicting Physico- or Biochemical Properties Based on Dozens of Molecules

The use of machine learning in chemistry has become a common practice. At the same time, despite the success of modern machine learning methods, the lack of data limits their use. Using a transfer learning methodology can help solve this problem. This methodology assumes that a model built on a sufficient amount of data captures general features of the chemical compound structure on which it was trained and that the further reuse of these features on a data set with a lack of data will greatly improve the quality of the new model. In this paper, we develop this approach for small organic molecules, implementing transfer learning with graph convolutional neural networks. The paper shows a significant improvement in the performance of the models for target properties with a lack of data. The effects of the data set composition on the model's quality and the applicability domain of the resulting models are also considered.

High-efficient ab initio Bayesian active learning method and applications in prediction of two-dimensional functional materials

Beyond the conventional trial-and-error method, machine learning offers a great opportunity to accelerate the discovery of functional materials, but still often suffers from difficulties such as limited materials data and the unbalanced distribution of target properties. Here, we propose the ab initio Bayesian active learning method that combines active learning and high-throughput ab initio calculations to accelerate the prediction of desired functional materials with ultrahigh efficiency and accuracy. We apply it as an instance to a large family (3119) of two-dimensional hexagonal binary compounds with unbalanced materials properties, and accurately screen out the materials with maximal electric polarization and proper photovoltaic band gaps, respectively, whereas the computational costs are significantly reduced by only calculating a few tenths of the possible candidates in comparison with a random search. This approach shows the enormous advantages for the cases with unbalanced distribution of target properties. It can be readily applied to seek a broad range of advanced materials.

Physics-Based Feature Makes Machine Learning Cognizing Crystal Properties Simple

Machine learning (ML) accelerates the rational design and discovery of materials, where the feature plays a critical role in the ML model training. We propose a low-cost electron probability waves (EPW) descriptor based on electronic structures, which is extracted from high-symmetry points in the Brillouin zone. In the task of distinguishing ferromagnetic or antiferromagnetic material, it achieves an accuracy (ACC) at 0.92 and an area under the receiver operating characteristic curve (AUC) at 0.83 by 10-fold cross-validation. Furthermore, EPW excels at classifying metal/semiconductors and judging the direct/indirect bandgap of semiconductors. The distribution of electron clouds is an essential criterion for the origin of ferromagnetism, and EPW acts as an emulation of the electronic structure, which is the key to the achievements. Our EPW-based ML model obtains ACC and AUC equivalent to crystal graph features-based deep learning models for tasks with physical recognitions in electronic states.

Accelerated design of promising mixed lead-free double halide organic-inorganic perovskites for photovoltaics using machine learning

Mixed double halide organic-inorganic perovskites (MDHOIPs) exhibit both good stability and high power conversion efficiency and have been regarded as attractive photovoltaic materials. Nevertheless, due to the complexity of structures, large-scale screening of thousands of possible candidates remains a great challenge. In this work, advanced machine learning (ML) techniques and first-principles calculations were combined to achieve a rapid screening of MDHOIPs for solar cells. Successfully, 204 stable lead-free MDHOIPs with optimal bandgaps were selected out of 11 370 candidates. The accuracy of ML models for perovskite structure formability and bandgap is over 94% and 97%, respectively. Moreover, representative MDHOIP candidates, MA₂GeSnI₄Br₂ and MA₂InBiI₂Br₄, stand out with suitable direct bandgaps, light carrier effective masses, small exciton binding energies, strong visible light absorption, and good stability against decomposition.

Full quantum search for high Tc two-dimensional van der Waals ferromagnetic semiconductors

Atomic thin two-dimensional (2D) ferromagnetic (FM) semiconductors with high Curie temperatures (T_c) are essential for future spintronic applications. However, reliable theoretical searching for 2D FM semiconductors is still hard due to the complexity of strong quantum fluctuations in 2D systems. We have proposed a full quantum search (FulQuanS) method to tackle the difficulty, and finally identified five 2D semiconductors of CrX₃ (X = I, Br, Cl), CuCl₃ and FeCl₂ with FM order at finite temperature from the pool of 3721 potential 2D structures. Via the method of renormalized spin wave theory (SW) and quantum Monte Carlo simulations (QMC), we located the T_c for CrX₃ (X = I, Br, Cl), CuCl₃ and FeCl₂ at 48 K, 31 K, 18 K, 74 K and 931 K respectively, which excellently agree with experiments for CrX₃ and reveal the superior performances of the new predicted structures. Furthermore, our QMC results demonstrated that the systems with low-spin numbers and/or low anisotropies have much higher T_c than the estimations of classical models e.g., Monte Carlo simulations based on classical Heisenberg models. Our findings suggest excellent candidates for future room-temperature spintronics, and shed light on the quantum effects inherent in 2D magnetism.

Conversation from antiferromagnetic MnBr2 to ferromagnetic Mn3Br8 monolayer with large MAE

A pressing need in low energy spintronics is two-dimensional (2D) ferromagnets with Curie temperature above the liquid-nitrogen temperature (77 K), and sizeable magnetic anisotropy. We studied Mn3Br8 monolayer which is obtained via inducing Mn vacancy at 1/4 population in MnBr2 monolayer. Such defective configuration is designed to change the coordination structure of the Mn-d5 and achieve ferromagnetism with sizeable magnetic anisotropy energy (MAE). Our calculations show that Mn3Br8 monolayer is a ferromagnetic (FM) half-metal with Curie temperature of 130 K, large MAE of - 2.33 meV per formula unit, and atomic magnetic moment of 13/3μB for the Mn atom. Additionally, Mn3Br8 monolayer maintains to be FM under small biaxial strain, whose Curie temperature under 5% compressive strain is 160 K. Additionally, both biaxial strain and carrier doping make the MAE increases, which mainly contributed by the magneto-crystalline anisotropy energy (MCE). Our designed defective structure of MnBr2 monolayer provides a simple but effective way to achieve ferromagnetism with large MAE in 2D materials.

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.

Large family of two-dimensional ferroelectric metals discovered via machine learning

Ferroelectricity and metallicity are usually believed not to coexist because conducting electrons would screen out static internal electric fields. In 1965, Anderson and Blount proposed the concept of "ferroelectric metal", however, it is only until recently that very rare ferroelectric metals were reported. Here, by combining high-throughput ab initio calculations and data-driven machine learning method with new electronic orbital based descriptors, we systematically investigated a large family (2964) of two-dimensional (2D) bimetal phosphates, and discovered 60 stable ferroelectrics with out-of-plane polarization, including 16 ferroelectric metals and 44 ferroelectric semiconductors that contain seven multiferroics. The ferroelectricity origins from spontaneous symmetry breaking induced by the opposite displacements of bimetal atoms, and the full-d-orbital coinage metal elements cause larger displacements and polarization than other elements. For 2D ferroelectric metals, the odd electrons per unit cell without spin polarization may lead to a half-filled energy band around Fermi level and is responsible for the metallicity. It is revealed that the conducting electrons mainly move on a single-side surface of the 2D layer, while both the ionic and electric contributions to polarization come from the other side and are vertical to the above layer, thereby causing the coexistence of metallicity and ferroelectricity. Van der Waals heterostructures based on ferroelectric metals may enable the change of Schottky barrier height or the Schottky-Ohmic contact type and induce a dramatic change of their vertical transport properties. Our work greatly expands the family of 2D ferroelectric metals and will spur further exploration of 2D ferroelectric metals.

Voting Data-Driven Regression Learning for Accelerating Discovery of Advanced Functional Materials and Applications to Two-Dimensional Ferroelectric Materials

Regression machine learning is widely applied to predict various materials. However, insufficient materials data usually leads to poor performance. Here, we develop a new voting data-driven method that could generally improve the performance of the regression learning model for accurately predicting properties of materials. We apply it to investigate a large family (2135) of two-dimensional hexagonal binary compounds focusing on ferroelectric properties and find that the performance of the model for electric polarization is indeed greatly improved, where 38 stable ferroelectrics with out-of-plane polarization including 31 metals and 7 semiconductors are screened out. By unsupervised learning, actionable information such as how the number and orbital radius of valence electrons, ionic polarizability, and electronegativity of constituent atoms affect polarization was extracted. Our voting data-driven method not only reduces the size of materials data for constructing a reliable learning model but also enables one to make precise predictions for targeted functional materials.

Two-dimensional magnetic materials: structures, properties and external controls

Since the discovery of intrinsic ferromagnetism in atomically thin Cr2Gr2Te6 and CrI3 in 2017, research on two-dimensional (2D) magnetic materials has become a highlighted topic. Based on 2D magnetic materials and their heterostructures, exotic physical phenomena at the atomically thin limit have been discovered, such as the quantum anomalous Hall effect, magneto-electric multiferroics, and magnon valleytronics. Furthermore, magnetism in these ultrathin magnets can be effectively controlled by external perturbations, such as electric field, strain, doping, chemical functionalization, and stacking engineering. These attributes make 2D magnets ideal platforms for fundamental research and promising candidates for various spintronic applications. This review aims at providing an overview of the structures, properties, and external controls of 2D magnets, as well as the challenges and potential opportunities in this field.

Data-driven studies of magnetic two-dimensional materials

We use a data-driven approach to study the magnetic and thermodynamic properties of van der Waals (vdW) layered materials. We investigate monolayers of the form \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\hbox {A}_2\hbox {B}_2\hbox {X}_6$$\end{document}A2B2X6, based on the known material \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\hbox {Cr}_2\hbox {Ge}_2\hbox {Te}_6$$\end{document}Cr2Ge2Te6, using density functional theory (DFT) calculations and machine learning methods to determine their magnetic properties, such as magnetic order and magnetic moment. We also examine formation energies and use them as a proxy for chemical stability. We show that machine learning tools, combined with DFT calculations, can provide a computationally efficient means to predict properties of such two-dimensional (2D) magnetic materials. Our data analytics approach provides insights into the microscopic origins of magnetic ordering in these systems. For instance, we find that the X site strongly affects the magnetic coupling between neighboring A sites, which drives the magnetic ordering. Our approach opens new ways for rapid discovery of chemically stable vdW materials that exhibit magnetic behavior.