arXiv subjects

Musen Li

Publications and source records attributed to Musen Li.

10 recordsLinked to original sources

arXiv2026

DeePAW: A universal machine learning model for orbital-free ab initio calculations

Developing universal machine learning models for ab initio calculations is the frontier of materials cutting edge research in the new era of artificial intelligence. Here, we present the Deep Augment Way model (DeePAW) that is a universal machine learning (ML) model for orbital-free (OF) ab initio calculations, based on the density functional theory (DFT). DeePAW is currently the best OFDFT ML model according to the three criterions, 1) covering the largest number of elements, 2) having the widest application capability to diverse crystal structures, and 3) achieving the highest prediction accuracy without further fine-tuning. These scientific merits and innovations of DeePAW are stemmed from the novel SE(3)-equivariant double massage passing neuron networks. Besides predicting electron density distributions, DeePAW predicts formation energies of crystals as well and therefore paves an efficient avenue for multiscale materials modeling beyond conventional electronic structure calculation methods.

cond-mat.mtrl-sci↗

arXiv2026

Significance of the dispersion force for ferroelectric switching in ZnO and related materials

Wurtzite-ZnO is a wide-bandgap polar material with a ferroelectric-switching barrier that is too high to utilize, but the barrier can be reduced and switching observed in substituted materials such as Zn0.5Mg0.5O. Here, we seek to understand atomic-scale features that control concerted polarization switching in these and related systems, focusing on the planar hexagonal structures h-ZnO and Zn0.5Mg0.5O that may act as metastable intermediate phases along the switching pathway. Consensus is obtained by considering a range of pure and dispersion-corrected density-functional theory (DFT) computational approaches, as well as ab initio Hartree-Fock (HF), M{\o}ller-Plesset perturbation-theory (MP2), and random-phase approximation (RPA) calculations. The perceived stability of h-ZnO is found to be strongly influenced by the dispersion correction, with the consensus being that dispersion interactions are insufficient to stabilize h-ZnO as a metastable phase in infinite crystals. In contrast, h-Zn0.5Mg0.5O is consistently predicted to be at least metastable, with some dispersion-corrected DFT approaches predicting it to be more stable than its wurtzite form; all DFT methods overestimate its stability compared to MP2 and RPA. Dispersion forces are found to be most significant for hypothetical planar hexagonal structures constrained to the lattice vectors of the wurtzite phases. In general, our results demonstrate that an accurate treatment of dispersion forces is essential when describing polarization switching and ferroelectric behavior in wurtzite-structured materials.

cond-mat.mtrl-sci↗

arXiv2025

Optimizing Data Extraction from Materials Science Literature: A Study of Tools Using Large Language Models

Large Language Models (LLMs) are increasingly utilized for large-scale extraction and organization of unstructured data owing to their exceptional Natural Language Processing (NLP) capabilities. Empowering materials design, vast amounts of data from experiments and simulations are scattered across numerous scientific publications, but high-quality experimental databases are scarce. This study considers the effectiveness and practicality of five representative AI tools (ChemDataExtractor, BERT-PSIE, ChatExtract, LangChain, and Kimi) to extract bandgaps from 200 randomly selected Materials Science publications in two presentations (arXiv and publisher versions), comparing the results to those obtained by human processing. Although the integrity of data extraction has not met expectations, encouraging results have been achieved in terms of precision and the ability to eliminate irrelevant papers from human consideration. Our analysis highlights both the strengths and limitations of these tools, offering insights into improving future data extraction techniques for enhanced scientific discovery and innovation. In conjunction with recent research, we provide guidance on feasible improvements for future data extraction methodologies, helping to bridge the gap between unstructured scientific data and structured, actionable databases.

cs.DL↗

arXiv2025

Re-refinement of the structure of the planar hexagonal phase of ZnO nanocrystals

The planar hexagonal phase of ZnO, known as h-ZnO, g-ZnO, {\alpha}-ZnO, the Bk structure, the 5-5 phase, the {\alpha}-BN phase, etc., has P63/mmc symmetry and is implicated in ferroelectric switching mechanisms for wurtzite-ZnO. It is well-known in thin films on substrates and to be stabilized by external pressure, but critical is its possible existence in high-purity nanocrystals under ambient conditions. Indeed, a crystal structure has been reported, but this work remains controversial as first-principles calculations predict very different structural properties. Herein, the original experimental data is re-refined, through phase-shift determination and Morlet wavelet transformation, that molecular dynamics simulations associate with a P63/mmc structure with lattice parameters at room temperature of a = 3.45{\pm}0.02 {\AA} and c = 4.46{\pm}0.02 {\AA}. These values are 0.35 {\AA} and 0.80 {\AA}, respectively, larger than those previously reported and in good agreement with computational predictions. This confirms that ZnO nanocrystals can form a metastable planar hexagonal phase. It provides key information pertaining to polarization switching in ZnO, its derivatives, and general wurtzite-structured materials.

cond-mat.mtrl-sci↗

arXiv2025

Reproducibility of high-throughput density-functional-theory calculations

While standard computational protocols for density functional theory (DFT) have universal applicability, differences exist in code implementations. Specific applications require manual parameter optimization, whereas high-throughput calculations employ predefined workflows. This paper uses the bandgap as a key property to reveal the impact of computational workflow differences on the reproducibility of high-throughput calculation results. The study proposes basic requirements for ensuring reproducibility: using structures optimised using the same procedure as used to calculate properties and ensuring Brillouin zone (k-point) integration grid accuracy. This research establishes a foundation for the reproducibility of DFT calculations and reliable application of results, which is of great significance for method development and artificial intelligence model training.

cond-mat.mtrl-sci↗

arXiv2025

Reliable Density Functional Theory Predictions of Bandgaps for Materials

We consider methods for optimizing the bandgap calculation of 3D materials, considering 340 sample materials. Examined are the effects of the choice of the pseudopotential to describe core electrons, the plane-wave basis set cutoff energy, and the Brillouin zone integration. Cost-saving calculations in which the structure is optimized using reduced-quality Brillouin zone integrations and cutoff energies were found to lead to experimentally significant errors exceeding 0.1 eV in 18% of cases using the PBE functional and 21% of cases using PBE0. Such cost-savings approaches are therefore not recommended for general applications. Also, the current practice of using unoptimized grids to perform the Brillouin-zone integrations in bandgap calculations is found to be unreliable for 16% of materials using PBE and for 23% using PBE0. A k-space optimization scheme is introduced that interpolates extensive PBE results to determine a generally useful approach that when used in PBE0 calculations is found to be inadequate for only 1.6% of the materials studied.

cond-mat.mtrl-sci↗

arXiv2023

Reliable density functional and G_0 W_0 approaches to the calculation of bandgaps in 2D materials

Optimizing density-functional theory (DFT) and G0W0 calculations present coupled problems as orbitals from DFT are needed as G0W0 starting points. Applied to 341 two-dimensional (2D) materials, we demonstrate that CAM-B3LYP provides minimal changes in bandgap (e.g., mean absolute deviation of 0.23 eV) when used to start G0W0 calculations, compared to traditional functionals such as PBE, PBE0, and HSE06 (1.07 eV, 1.48 eV, and 1.51 eV, respectively). CAM-B3LYP also delivers the smallest changes in orbital representation. These and other results indicate the suitability of CAM-B3LYP as a density-functional approach for modelling 2D materials, as well as for use in optimizing G0W0 calculations. Our findings parallel well established features of applications to molecules, as well as for spectroscopic applications involving 3D materials.

cond-mat.mtrl-sci↗

arXiv2021

Accurate prediction of the properties of materials using the CAM-B3LYP Density Functional

Density functionals with asymptotic corrections to the long-range potential provide entry-level methods for calculations on molecules that can sustain charge transfer, but similar applications in Materials Science are rare. We describe an implementation of the CAM-B3LYP range-separated functional within the Vienna Ab-initio Simulation Package (VASP) framework, together with its analytical functional derivatives. Results obtained for eight representative materials: aluminum, diamond, graphene, silicon, NaCl, MgO, 2D h-BN and 3D h-BN, indicate that CAM-B3LYP predictions embody mean-absolute deviations (MAD) compared to HSE06 that are reduced by a factor of 6 for lattice parameters, 4 for quasiparticle band gaps, 3 for the lowest optical excitation energies, and 6 for exciton binding energies. Further, CAM-B3LYP appears competitive compared to ab initio G0W0 and Bethe-Salpeter equation (BSE) approaches. The CAM-B3LYP implementation in VASP was verified by comparison of optimized geometries and reaction energies for isolated molecules taken from the ACCDB database, evaluated in large periodic unit cells, to analogous results obtained using Gaussian basis sets. Using standard GW pseudopotentials and energy cutoffs for the plane-wave calculations and the aug-cc-pV5Z basis set for the atomic-basis ones, the MAD in energy for 1738 chemical reactions was 0.34 kcal mol-1, whilst for 480 unique bond lengths this was 0.0036 {\AA}; these values reduced to 0.28 kcal mol-1 (largest error 0.94 kcal mol-1) and 0.0009 {\AA} by increasing the plane-wave cuttoff energy to 850 eV.

cond-mat.mtrl-sci↗

arXiv2020

Density functionals with asymptotic-potential corrections are required for the simulation of spectroscopic properties of materials

Five effects of correction of the asymptotic potential error in density functionals are identified that significantly improve calculated properties of molecular excited states involving charge-transfer character. Newly developed materials-science computational methods are used to demonstrate how these effects manifest in materials spectroscopy. Connection is made considering chlorophyll-a as a paradigm for molecular spectroscopy, 22 iconic materials as paradigms for 3D materials spectroscopy, and the VN- defect in hexagonal boron nitride as an example of the spectroscopy of defects in 2D materials pertaining to nanophotonics. Defects can equally be thought of as being "molecular" and "materials" in nature and hence bridge the realms of molecular and materials spectroscopies. It is concluded that the density functional HSE06, currently considered as the standard for accurate calculations of materials spectroscopy, should be replaced, in most instances, by the computationally similar but asymptotically corrected CAM-B3LYP functional, with some specific functionals for materials use only providing further improvements.

cond-mat.mtrl-sci↗

arXiv2018

Faraday-cage screening reveals intrinsic aspects of the van der Waals attraction

General properties of the recently observed screening of the van der Waals (vdW) attraction between a silica substrate and silica tip by insertion of graphene are predicted using basic theory and first-principles calculations. Results are then focused on possible practical applications, as well as an understanding of the nature of vdW attraction, considering recent discoveries showing it competing against covalent and ionic bonding. The traditional view of the vdW attraction as arising from pairwise-additive London dispersion forces is considered using Grimme's "D3" method, comparing results to those from Tkatchenko's more general many-body dispersion (MBD) approach, all interpreted in terms of Dobson's general dispersion framework. Encompassing the experimental results, MBD screening of the vdW force between two silica bilayers is shown to scale up to medium separations as 1.25 de/d, where d is the bilayer separation and de its equilibrium value, depicting antiscreening approaching and inside de. Means of unifying this correlation effect with those included in modern density functionals are urgently required.

cond-mat.mtrl-sci↗