Edirisuriya Mudalige Dilanga Siriwardane

Crystal structure prediction (CSP) of inorganic materials is a fundamental challenge in computational materials science, yet the field has historically lacked well‐defined benchmarks and comprehensive evaluations. We address this gap by introducing CSP180, a standardized benchmark of 180 inorganic materials, and evaluate 13 state‐of‐the‐art CSP algorithms spanning template‐based, machine learning (ML) potential‐based, and deep learning‐based approaches, none relying on density functional theory (DFT). As a baseline, we also assess the leading DFT‐based algorithms, CALYPSO and USPEX, on a 23‐structure subset with simpler compositions. Template‐based methods, such as TCSP and CSPML, achieve the highest space group match rates (58.9% and 47.2%) when a suitable template exists, while de novo algorithms struggle to identify correct space groups, with the best performers reaching only around 17%. On the 23‐structure subset, ML potential‐based algorithms prove highly competitive with DFT‐based methods at identifying lower‐energy structures, and this performance depends strongly on both potential quality and search algorithm efficiency. This assessment provides a standardized framework, quantitative metrics, and open‐source tools to guide future CSP development and accelerate materials discovery. Code and benchmark data are available at https://github.com/usccolumbia/cspbenchmark .

Layered materials, with their distinctive electronic and surface structures, have garnered significant attention in catalysis and energy conversion research. MXenes, a family of two-dimensional transition metal carbides, nitrides, and carbonitrides, exhibit remarkable electrical conductivity and surface chemistry. MXene restacking is commonly observed property during the exfoliation process. Druffel et al. synthesized 3D Y2CF2 MXene through a high-temperature solid-state reaction. We performed density functional theory (DFT) computation to study the structural and electronic property changes when Y2XT2 MXenes formed bilayer and 3D structures. We considered X as C and N, while surface termination T as O, F, and Cl. Except monolayer Y2NO2, the other two Nitride monolayers show metallic character. Our calculations show that Y2NO2 monolayer become a metal when stacking occurs to form bilayers and 3D structures. On the contrary, all the carbide monolayers exhibit non-zero band gaps showing they are semiconductors. Their semiconductor character is preserved even after stacking occurs. Our HSE06 hybrid functional calculations reveal that monolayer Y2CO2, Y2CF2, and Y2CCl2 have band gaps of 2.27, 1.82, and 1.77 eV, respectively. Upon formation of bilayers, the band gaps shift to 2.55, 1.94, and 1.72 eV, while 3D materials exhibit band gaps of 1.98, 2.00, and 1.65 eV, respectively. Moreover, we also studied the electronic properties of bulk MXenes under the external pressure range of 1-10 GPa. Under external pressure, our calculations shows that the band gap change of Y2CO2 and Y2CF2 bulk MXenes is negligible. Conversely, increasing external pressure leads to a reduction in the band gap of bulk Y2CCl2.

Ultralow compressible materials, which have high bulk modulus (K), are invaluable in extreme conditions due to their ability to undergo significant compression without structural failure. As large number of borides can be found with high K, this study develop a computational framework to scan the vast chemical space to identify the ultralow compressible borides. Transformer-based networks are helpful to generate new chemical compositions due to their self-attention mechanism, scalability, and ability to capture long-range dependencies. First, we developed a transformer-based network to generate new binary and ternary boride compositions based on the known boride compositions. Next, we trained a hybrid model based on AdaBoost and Gradient Boosting algorithms with mean absolute error (MAE) of 14.1 GPa to scan the high K borides. The CALYPSO code was used to find the possible structures for those materials. After predicting K for broad chemical domain, we found that Re-B and W-B systems are promising ultralow compressible materials. We then performed density functional theory (DFT) calculations to investigate the stability of high K materials. Our computations suggest that Re3B2, Re2B3, W5VB4, and Re5CrB4 materials exhibit K > 300 GPa with negative formation energy and energy-above-hull less than 40 meV. Those materials are mechanically and dynamically stable based on the elastic constant calculations and the phonon dispersion.

A DFT study explored the properties of MnCrNO₂ MXene, revealing a linear relationship of its band gap and magnetic anisotropy energy as a function of biaxial strain. Adsorption of alkali and alkaline earth metals significantly modulates the band gap.

In this study, we explored the phase stability of quaternary Mo1−xMxAlB (M = Cr, Fe, Mn, Nb, Sc, Ta, Ti, V, and W) solid solutions by employing a cluster expansion method to generate structures with different concentrations of M atoms. Using the first-principles calculations based on density functional theory, we predicted that these compounds exhibit a preference for either fully random structures or phase-segregated (M-rich regions) phases against the competing phases. To evaluate the Gibbs free energy of Mo1−xMxAlB alloys, we investigated the impact of various entropy contributions, including configurational, electronic, and vibrational entropy. Our study revealed that configurational entropy plays an important role in stabilizing the random phases observed in Mo1−xMxAlB compounds, highlighting its importance in understanding the thermodynamic behavior of these alloys. However, the vibrational and electronic entropy changes with respect to competing phases can stabilize or destabilize Mo1−xMxAlB depending on their sign. Our results indicate that, while W is soluble across the entire range of mixing ratios, Sc and Ti are completely insoluble in any ratio. On the other hand, Cr, Ta, Nb, and V can be successfully incorporated into the MoAlB lattice at varying fractions at elevated temperatures. The size, valence electron concentration, and electronegativity differences between Mo and M can be utilized as descriptors to identify stable Mo1−xMxAlB compounds. We extensively examined the structural, dynamical stability, thermal conductivity, and mechanical properties of Mo1−xMxAlB compounds. We analyze their dependence on the choice of the M element incorporated in the alloy. Our findings can guide the design and engineering of these materials to tailor their properties to specific applications based on the choice of the M element.

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. 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, Ti2HfO5, 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.

Discovering novel magnetic materials is essential for advancing the spintronic technology with significant applications in data communication, data storage, quantum computing, etc. While density functional theory (DFT) has been widely used for designing materials, its high computational demand for estimating the magnetic ground states of even a single material limits its ability to explore the vast chemical design space to find the right materials for spintronic applications. In this work, we developed a computational framework combining generative adversarial networks (GANs), machine learning (ML) classifiers, and DFT for de novo magnetic material discovery. We used the CubicGAN generative crystal structure design model for creating new ternary cubic structures. Machine learning classifiers were developed with around 90% accuracy to screen candidate ternary magnetic materials, which were then subjected to DFT based stability validation. Our calculations discovered and confirmed that Na₆TcO₆, K₆TcO₆, and BaCuF₆ are stable ferromagnetic compounds, while Rb₆IrO₆ is a stable antiferromagnetic material. All these materials have zero energy above hull. Moreover, Na₆TcO₆ and BaCuF₆ are found to be half metals that are highly favorable for spintronic applications. Due to the structural differences, the A₆MO₆ materials have a higher thermal capacity (C_v) compared to BaCuF₆. At 300 K temperature, the C_v of the A₆MO₆ materials is around 1100 J K⁻¹ mol⁻¹ and that of BaCuF₆ is about 176 J K⁻¹ mol⁻¹. This work demonstrates the promising potential of deep generative design for discovering novel functional materials.

Crystal structure prediction is one of the major unsolved problems in materials science. Traditionally, this problem is formulated as a global optimization problem for which global search algorithms are combined with first-principles free energy calculations to predict the ground-state crystal structure of a given material composition. These ab initio algorithms are currently too slow for predicting complex material structures. Inspired by the AlphaFold algorithm for protein structure prediction, herein, we propose AlphaCrystal, a crystal structure prediction algorithm that combines a deep residual neural network model for predicting the atomic contact map of a target material followed by three-dimensional (3D) structure reconstruction using genetic algorithms. Extensive experiments on 20 benchmark structures showed that our AlphaCrystal algorithm can predict structures close to the ground truth structures, which can significantly speed up the crystal structure prediction and handle relatively large systems.