arXiv subjects

Nihang Fu

Publications and source records attributed to Nihang Fu.

At least 19 recordsLinked to original sources

arXiv2025

Facet: highly efficient E(3)-equivariant networks for interatomic potentials

Computational materials discovery is limited by the high cost of first-principles calculations. Machine learning (ML) potentials that predict energies from crystal structures are promising, but existing methods face computational bottlenecks. Steerable graph neural networks (GNNs) encode geometry with spherical harmonics, respecting atomic symmetries -- permutation, rotation, and translation -- for physically realistic predictions. Yet maintaining equivariance is difficult: activation functions must be modified, and each layer must handle multiple data types for different harmonic orders. We present Facet, a GNN architecture for efficient ML potentials, developed through systematic analysis of steerable GNNs. Our innovations include replacing expensive multi-layer perceptrons (MLPs) for interatomic distances with splines, which match performance while cutting computational and memory demands. We also introduce a general-purpose equivariant layer that mixes node information via spherical grid projection followed by standard MLPs -- faster than tensor products and more expressive than linear or gate layers. On the MPTrj dataset, Facet matches leading models with far fewer parameters and under 10% of their training compute. On a crystal relaxation task, it runs twice as fast as MACE models. We further show SevenNet-0's parameters can be reduced by over 25% with no accuracy loss. These techniques enable more than 10x faster training of large-scale foundation models for ML potentials, potentially reshaping computational materials discovery.

cond-mat.mtrl-sci↗

arXiv2025

Data-Driven Topological Analysis of Polymorphic Crystal Structures

Polymorphism, the ability of a compound to crystallize in multiple distinct structures, plays a vital role in determining the physical, chemical, and functional properties of materials. Accurate identification and prediction of polymorphic structures are critical for materials design, drug development, and device optimization, as unknown or overlooked polymorphs may lead to unexpected performance or stability issues. Despite its significance, predicting polymorphism directly from a chemical composition remains a challenging problem due to the complex interplay between molecular conformations, crystal packing, and symmetry constraints. In this study, we conduct a comprehensive data-driven analysis of polymorphic materials from the Materials Project database, uncovering key statistical patterns in their composition, space group distributions, and polyhedral building blocks. We discover that frequent polymorph pairs across space groups, such as (71, 225), display recurring topological motifs that persist across different compounds, highlighting topology not symmetry alone as a key factor in polymorphic recurrence. We reveal that many polymorphs exhibit consistent local polyhedral environments despite differences in their symmetry or packing. Additionally, by constructing polyhedron connectivity graphs and embedding their topology, we successfully cluster polymorphs and structurally similar materials even across different space groups, demonstrating that topological similarity serves as a powerful descriptor for polymorphic behavior. Our findings provide new insights into the structural characteristics of polymorphic materials and demonstrate the potential of data mining and machine learning for accelerating polymorph discovery and design.

cond-mat.mtrl-sci↗

arXiv2025

TCSP 2.0: Template Based Crystal Structure Prediction with Improved Oxidation State Prediction and Chemistry Heuristics

Crystal structure prediction remains a major challenge in materials science, directly impacting the discovery and development of next-generation materials. We introduce TCSP 2.0, a substantial evolution of our template-based crystal structure prediction framework that advances predictive capabilities through synergistic integration of several key techniques into its major components. Building upon TCSP 1.0's template-matching foundation, this enhanced version implements three critical innovations: (1) replacement of Pymatgen with deep learning-based BERTOS model for oxidation state prediction with superior performance, (2) implementation of sophisticated element embedding distance metrics for improved chemical similarity assessment, and (3) development of a robust majority voting mechanism for space group selection that reduces prediction uncertainty. TCSP 2.0 also expands its template base by incorporating template structures from Materials Cloud, C2DB, and GNoME databases alongside the original Materials Project repository, creating a more comprehensive structural foundation. Rigorous validation across 180 diverse test cases of the CSPBenchmark demonstrates TCSP 2.0's exceptional performance, achieving 83.89% space-group success rate and 78.33% structural similarity accuracy for top-5 predictions, substantially outperforming both its predecessor and the competing modern CSP algorithms including CSPML and EquiCSP.

cond-mat.mtrl-sci↗

arXiv2024

Physical Encoding Improves OOD Performance in Deep Learning Materials Property Prediction

Deep learning (DL) models have been widely used in materials property prediction with great success, especially for properties with large datasets. However, the out-of-distribution (OOD) performances of such models are questionable, especially when the training set is not large enough. Here we showed that using physical encoding rather than the widely used one-hot encoding can significantly improve the OOD performance by increasing the models' generalization performance, which is especially true for models trained with small datasets. Our benchmark results of both composition- and structure-based deep learning models over six datasets including formation energy, band gap, refractive index, and elastic properties predictions demonstrated the importance of physical encoding to OOD generalization for models trained on small datasets.

cond-mat.mtrl-sci↗

arXiv2024

CSPBench: a benchmark and critical evaluation of Crystal Structure Prediction

Crystal structure prediction (CSP) is now increasingly used in discovering novel materials with applications in diverse industries. However, despite decades of developments and significant progress in this area, there lacks a set of well-defined benchmark dataset, quantitative performance metrics, and studies that evaluate the status of the field. We aim to fill this gap by introducing a CSP benchmark suite with 180 test structures along with our recently implemented CSP performance metric set. We benchmark a collection of 13 state-of-the-art (SOTA) CSP algorithms including template-based CSP algorithms, conventional CSP algorithms based on DFT calculations and global search such as CALYPSO, CSP algorithms based on machine learning (ML) potentials and global search, and distance matrix based CSP algorithms. Our results demonstrate that the performance of the current CSP algorithms is far from being satisfactory. Most algorithms cannot even identify the structures with the correct space groups except for the template-based algorithms when applied to test structures with similar templates. We also find that the ML potential based CSP algorithms are now able to achieve competitive performances compared to the DFT-based algorithms. These CSP algorithms' performance is strongly determined by the quality of the neural potentials as well as the global optimization algorithms. Our benchmark suite comes with a comprehensive open-source codebase and 180 well-selected benchmark crystal structures, making it convenient to evaluate the advantages and disadvantages of CSP algorithms from future studies. All the code and benchmark data are available at https://github.com/usccolumbia/cspbenchmark

cond-mat.mtrl-sci↗

arXiv2024

Structure-based out-of-distribution (OOD) materials property prediction: a benchmark study

In real-world material research, machine learning (ML) models are usually expected to predict and discover novel exceptional materials that deviate from the known materials. It is thus a pressing question to provide an objective evaluation of ML model performances in property prediction of out-of-distribution (OOD) materials that are different from the training set distribution. Traditional performance evaluation of materials property prediction models through random splitting of the dataset frequently results in artificially high performance assessments due to the inherent redundancy of typical material datasets. Here we present a comprehensive benchmark study of structure-based graph neural networks (GNNs) for extrapolative OOD materials property prediction. We formulate five different categories of OOD ML problems for three benchmark datasets from the MatBench study. Our extensive experiments show that current state-of-the-art GNN algorithms significantly underperform for the OOD property prediction tasks on average compared to their baselines in the MatBench study, demonstrating a crucial generalization gap in realistic material prediction tasks. We further examine the latent physical spaces of these GNN models and identify the sources of CGCNN, ALIGNN, and DeeperGATGNN's significantly more robust OOD performance than those of the current best models in the MatBench study (coGN and coNGN), and provide insights to improve their performance.

cond-mat.mtrl-sci↗

arXiv2024

Physics guided dual Self-supervised learning for structure-based materials property prediction

Deep learning (DL) models have now been widely used for high-performance material property prediction for properties such as formation energy and band gap. However, training such DL models usually requires a large amount of labeled data, which is usually not available for most materials properties such as exfoliation energy and elastic properties. Self-supervised learning (SSL) methods have been proposed to address this data scarcity issue by learning inherent representations from unlabeled data in various research fields. Herein, we present DSSL, a physics-guided Dual SSL framework, for graph neural networks (GNNs) based material property prediction. This hybrid framework combines node-masking based predictive SSL with atomic coordinate perturbation based contrastive SSL strategies, allowing it to learn structural embeddings that capture both local and global information of input crystals. Especially, we propose to use predicting the macroproperty (e.g. elasticity) related microproperty such as atomic stiffness as an additional pretext task to achieve physics-guided pretraining process. We pretrain our DSSL model on the Materials Project database with unlabeled data and finetune it with ten extra datasets with different material properties. The experimental results demonstrate that teaching neural networks some physics using the SSL strategy can bring up to 26.89\% performance improvement compared to the baseline GNN models. Our source code is now freely available at https://github.com/usccolumbia/DSSL

cond-mat.mtrl-sci↗

arXiv2023

Generative Design of inorganic compounds using deep diffusion language models

Due to the vast chemical space, discovering materials with a specific function is challenging. Chemical formulas are obligated to conform to a set of exacting criteria such as charge neutrality, balanced electronegativity, synthesizability, and mechanical stability. In response to this formidable task, we introduce a deep learning-based generative model for material composition and structure design by learning and exploiting explicit and implicit chemical knowledge. Our pipeline first uses deep diffusion language models as the generator of compositions and then applies a template-based crystal structure prediction algorithm to predict their corresponding structures, which is then followed by structure relaxation using a universal graph neural network-based potential. The density functional theory (DFT) calculations of the formation energies and energy-above-the-hull analysis are used to validate new structures generated through our pipeline. Based on the DFT calculation results, six new materials, including Ti2HfO5, TaNbP, YMoN2, TaReO4, HfTiO2, and HfMnO2, with formation energy less than zero have been found. Remarkably, among these, four materials, namely Ti2$HfO5, TaNbP, YMoN2, and TaReO4, exhibit an e-above-hull energy of less than 0.3 eV. These findings have proved the effectiveness of our approach.

cond-mat.mtrl-sci↗

arXiv2023

Improving realistic material property prediction using domain adaptation based machine learning

Materials property prediction models are usually evaluated using random splitting of datasets into training and test datasets, which not only leads to over-estimated performance due to inherent redundancy, typically existent in material datasets, but also deviate away from the common practice of materials scientists: they are usually interested in predicting properties for a known subset of related out-of-distribution (OOD) materials rather than a universally distributed samples. Feeding such target material formulas/structures to the machine learning models should improve the prediction performance while most current machine learning (ML) models neglect this information. Here we propose to use domain adaptation (DA) to enhance current ML models for property prediction and evaluate their performance improvements in a set of five realistic application scenarios. Our systematic benchmark studies show that there exist DA models that can significantly improve the OOD test set prediction performance while standard ML models and most of the other DAs cannot improve or even deteriorate the performance. Our benchmark datasets and DA code can be freely accessed at https://github.com/Little-Cheryl/MatDA.

cond-mat.mtrl-sci↗

arXiv2023

MD-HIT: Machine learning for materials property prediction with dataset redundancy control

Materials datasets are usually featured by the existence of many redundant (highly similar) materials due to the tinkering material design practice over the history of materials research. For example, the materials project database has many perovskite cubic structure materials similar to SrTiO$_3$. This sample redundancy within the dataset makes the random splitting of machine learning model evaluation to fail so that the ML models tend to achieve over-estimated predictive performance which is misleading for the materials science community. This issue is well known in the field of bioinformatics for protein function prediction, in which a redundancy reduction procedure (CD-Hit) is always applied to reduce the sample redundancy by ensuring no pair of samples has a sequence similarity greater than a given threshold. This paper surveys the overestimated ML performance in the literature for both composition based and structure based material property prediction. We then propose a material dataset redundancy reduction algorithm called MD-HIT and evaluate it with several composition and structure based distance threshold sfor reducing data set sample redundancy. We show that with this control, the predicted performance tends to better reflect their true prediction capability. Our MD-hit code can be freely accessed at https://github.com/usccolumbia/MD-HIT

cond-mat.mtrl-sci↗

arXiv2023

Global mapping of structures and properties of crystal materials

Understanding material composition-structure-function relationships is of critical importance for the design and discovery of novel functional materials. While most such studies focus on individual materials, we conducted a global mapping study of all known materials deposited in the Material Project database to investigate their distributions in the space of a set of seven compositional, structural, physical, and neural latent descriptors. These two-dimensional materials maps along with their density maps allow us to illustrate the distribution of the patterns and clusters of different shapes, which indicates the propensity of these materials and the tinkering history of existing materials. We then overlap the material properties such as composition prototypes and piezoelectric properties over the background materials maps to study the relationships of how material compositions and structures affect their physical properties. We also use these maps to study the spatial distributions of properties of known inorganic materials, in particular those of local vicinities in structural space such as structural density and functional diversity. These maps provide a uniquely comprehensive overview of materials and space and thus reveal previously undescribed fundamental properties. Our methodology can be easily extended by other researchers to generate their own global material maps with different background maps and overlap properties for both distribution understanding and cluster-based new material discovery. The source code for feature generation and generated maps are available at https://github.com/usccolumbia/matglobalmapping

cond-mat.mtrl-sci↗

arXiv2022

Composition based oxidation state prediction of materials using deep learning

Oxidation states are the charges of atoms after their ionic approximation of their bonds, which have been widely used in charge-neutrality verification, crystal structure determination, and reaction estimation. Currently only heuristic rules exist for guessing the oxidation states of a given compound with many exceptions. Recent work has developed machine learning models based on heuristic structural features for predicting the oxidation states of metal ions. However, composition based oxidation state prediction still remains elusive so far, which is more important in new material discovery for which the structures are not even available. This work proposes a novel deep learning based BERT transformer language model BERTOS for predicting the oxidation states of all elements of inorganic compounds given only their chemical composition. Our model achieves 96.82\% accuracy for all-element oxidation states prediction benchmarked on the cleaned ICSD dataset and achieves 97.61\% accuracy for oxide materials. We also demonstrate how it can be used to conduct large-scale screening of hypothetical material compositions for materials discovery.

cond-mat.mtrl-sci↗

arXiv2022

Probabilistic Generative Transformer Language models for Generative Design of Molecules

Self-supervised neural language models have recently found wide applications in generative design of organic molecules and protein sequences as well as representation learning for downstream structure classification and functional prediction. However, most of the existing deep learning models for molecule design usually require a big dataset and have a black-box architecture, which makes it difficult to interpret their design logic. Here we propose Generative Molecular Transformer (GMTransformer), a probabilistic neural network model for generative design of molecules. Our model is built on the blank filling language model originally developed for text processing, which has demonstrated unique advantages in learning the "molecules grammars" with high-quality generation, interpretability, and data efficiency. Benchmarked on the MOSES datasets, our models achieve high novelty and Scaf compared to other baselines. The probabilistic generation steps have the potential in tinkering molecule design due to their capability of recommending how to modify existing molecules with explanation, guided by the learned implicit molecule chemistry. The source code and datasets can be accessed freely at https://github.com/usccolumbia/GMTransformer

cond-mat.mtrl-sci↗

arXiv2022

Designing novel protein structures using sequence generator and AlphaFold2

Protein structures and functions are determined by a contiguous arrangement of amino acid sequences. Designing novel protein sequences and structures with desired geometry and functions is a complex task with large state spaces. Here we develop a novel protein design pipeline consisting of two deep learning algorithms, ProteinSolver and AlphaFold2. ProteinSolver is a deep graph neural network that generates amino acid sequences such that the forces between interacting amino acids are favorable and compatible with the fold while AlphaFold2 is a deep learning algorithm that predicts the protein structures from protein sequences. We present forty de novo designed binding sites of the PTP1B and P53 proteins with high precision, out of which thirty proteins are novel. Using ProteinSolver and AlphaFold2 in conjunction, we can trim the exploration of the large protein conformation space, thus expanding the ability to find novel and diverse de novo protein designs.

physics.chem-ph↗

arXiv2022

Materials Transformers Language Models for Generative Materials Design: a benchmark study

Pre-trained transformer language models on large unlabeled corpus have produced state-of-the-art results in natural language processing, organic molecule design, and protein sequence generation. However, no such models have been applied to learn the composition patterns of inorganic materials. Here we train a series of seven modern transformer language models (GPT, GPT-2, GPT-Neo, GPT-J, BLMM, BART, and RoBERTa) using the expanded formulas from material deposited in the ICSD, OQMD, and Materials Projects databases. Six different datasets with/out non-charge-neutral or balanced electronegativity samples are used to benchmark the performances and uncover the generation biases of modern transformer models for the generative design of materials compositions. Our extensive experiments showed that the causal language models based materials transformers can generate chemically valid materials compositions with as high as 97.54\% to be charge neutral and 91.40\% to be electronegativity balanced, which has more than 6 times higher enrichment compared to a baseline pseudo-random sampling algorithm. These models also demonstrate high novelty and their potential in new materials discovery has been proved by their capability to recover the leave-out materials. We also find that the properties of the generated samples can be tailored by training the models with selected training sets such as high-bandgap materials. Our experiments also showed that different models each have their own preference in terms of the properties of the generated samples and their running time complexity varies a lot. We have applied our materials transformer models to discover a set of new materials as validated using DFT calculations.

cond-mat.mtrl-sci↗

arXiv2022

DeepXRD, a Deep Learning Model for Predicting of XRD spectrum from Materials Composition

One of the long-standing problems in materials science is how to predict a material's structure and then its properties given only its composition. Experimental characterization of crystal structures has been widely used for structure determination, which is however too expensive for high-throughput screening. At the same time, directly predicting crystal structures from compositions remains a challenging unsolved problem. Herein we propose a deep learning algorithm for predicting the XRD spectrum given only the composition of a material, which can then be used to infer key structural features for downstream structural analysis such as crystal system or space group classification or crystal lattice parameter determination or materials property predictions. Benchmark studies on two datasets show that our DeepXRD algorithm can achieve good performance for XRD prediction as evaluated over our test sets. It can thus be used in high-throughput screening in the huge materials composition space for new materials discovery.

cond-mat.mtrl-sci↗

arXiv2022

Physics Guided Deep Learning for Generative Design of Crystal Materials with Symmetry Constraints

Discovering new materials is a challenging task in materials science crucial to the progress of human society. Conventional approaches based on experiments and simulations are labor-intensive or costly with success heavily depending on experts' heuristic knowledge. Here, we propose a deep learning based Physics Guided Crystal Generative Model (PGCGM) for efficient crystal material design with high structural diversity and symmetry. Our model increases the generation validity by more than 700\% compared to FTCP, one of the latest structure generators and by more than 45\% compared to our previous CubicGAN model. Density Functional Theory (DFT) calculations are used to validate the generated structures with 1,869 materials out of 2,000 are successfully optimized and deposited into the Carolina Materials Database \url{www.carolinamatdb.org}, of which 39.6\% have negative formation energy and 5.3\% have energy-above-hull less than 0.25 eV/atom, indicating their thermodynamic stability and potential synthesizability.

cond-mat.mtrl-sci↗

arXiv2021

Semi-supervised teacher-student deep neural network for materials discovery

Data driven generative machine learning models have recently emerged as one of the most promising approaches for new materials discovery. While the generator models can generate millions of candidates, it is critical to train fast and accurate machine learning models to filter out stable, synthesizable materials with desired properties. However, such efforts to build supervised regression or classification screening models have been severely hindered by the lack of unstable or unsynthesizable samples, which usually are not collected and deposited in materials databases such as ICSD and Materials Project (MP). At the same time, there are a significant amount of unlabelled data available in these databases. Here we propose a semi-supervised deep neural network (TSDNN) model for high-performance formation energy and synthesizability prediction, which is achieved via its unique teacher-student dual network architecture and its effective exploitation of the large amount of unlabeled data. For formation energy based stability screening, our semi-supervised classifier achieves an absolute 10.3\% accuracy improvement compared to the baseline CGCNN regression model. For synthesizability prediction, our model significantly increases the baseline PU learning's true positive rate from 87.9\% to 97.9\% using 1/49 model parameters. To further prove the effectiveness of our models, we combined our TSDNN-energy and TSDNN-synthesizability models with our CubicGAN generator to discover novel stable cubic structures. Out of 1000 recommended candidate samples by our models, 512 of them have negative formation energies as validated by our DFT formation energy calculations. Our experimental results show that our semi-supervised deep neural networks can significantly improve the screening accuracy in large-scale generative materials design.

cond-mat.mtrl-sci↗