Jul 2026· Proceedings of the 3rd Foundations of Process/Product Analytics and Machine Learning (FOPAM 2026)· pp. 6-7· 0 citations· 8 references
TL;DR
The Genarris code is developed, and it is shown that MLIPs can completely replace both early-stage screening with classical force fields and final ranking with DFT, paving the way to high-throughput CSP.
Abstract
Molecular crystals are bound by dispersion (van der Waals) interactions, whose weak nature gives rise to polymorphism, the ability of a compound to crystallize in different structures. Crystal structure profoundly influences the physical and chemical properties, and hence the functionality of molecular solids in applications including pharmaceuticals, electronic devices, and energetic materials. Therefore, the ability to predict the structure and properties of molecular crystals is of paramount importance. To this end, we combine first principles simulations with machine learning. Molecular crystal structure prediction (CSP) is challenging because it requires searching a high-dimensional configuration space with high accuracy. CSP workflow have two main components, structure generation and stability ranking. For structure generation, we develop the Genarris code [1], which generates random structures in all compatible space groups with physical constraints on intermolecular distances. Machine learned interatomic potentials (MLIPs) are trained on large data sets of first principles simulations [2], typically density functional theory (DFT) to achieve DFT-level accuracy at the computational cost of classical force fields. We have interfaced Genarris with several types of MLIPs for geometry optimization and stability ranking [1,3,4]. We have shown that MLIPs can completely replace both early-stage screening with classical force fields and final ranking with DFT, paving the way to high-throughput CSP [4]. One of the optoelectronic applications of molecular crystals is singlet fission (SF), the conversion of one photogenerated singlet exciton into two triplet excitons. SF has the potential to increase the efficiency of solar cells by harvesting two charge carriers from one high-energy photon, whose excess energy would otherwise be lost to heat. The realization of SF-based solar cells is hindered by the dearth of suitable materials. The excited-state properties of molecular crystals can be calculated using many-body perturbation theory (MBPT) within in the GW approximation and the Bethe-Salpeter equation (BSE) [5]. The computational cost of GW+BSE is prohibitive for large-scale exploration of the chemical space, and also for generating large amounts of training data. This calls for ML approaches that work well with small data.
Due to its characteristics of high electron mobility, moderate band gap, excellent radiation resistance, and high-frequency performance, gallium arsenide (GaAs) is widely used as an important semiconductor material in microelectronics and optoelectronics. Here, we systematically investigated the crystallization mechanism of GaAs by machine learning molecular dynamics simulation. By combining unbiased molecular dynamics simulations with deep neural network potentials, the melting temperature of GaAs crystals was predicted, revealing the crystallization mechanism and phase competition rules. The results show that under undercooling conditions, the cubic zinc blende (ZB) phase has a slower crystallization rate than hexagonal wurtzite (WZ), and its excellent thermodynamic stability makes it the dominant phase in the crystallization process. These factors ultimately determine that a GaAs melt forms a mixed crystal mainly composed of the ZB phase. Ga and As atoms cooperatively complete the lattice arrangement at the solid-liquid interface, and the short-range order is the core driving force for rapid crystal growth. Thus, the microscopic mechanism of the solid-liquid phase transition of GaAs has been probed by applying atomic-scale simulation methods, providing theoretical support for the design, preparation, and performance control of semiconductor materials.
Hongbin Zhang, Zijun Meng, Haichao Li et al.· Inorganic Chemistry· 0 citations
Melting point (MP) is an important thermophysical property for the chemical process industry, yet accurate prediction of MP for organic compounds in the absence of experimental data remains challenging due to the complex interplay between molecular packing, intermolecular interactions, and electronic structure. Traditional group contribution and quantitative structure-property relationship models, which rely primarily on static molecular descriptors, often fail to capture these critical condensed-phase effects. In this study, we present a hybrid machine learning framework that integrates cheminformatics descriptors with quantum chemical features and dynamic condensed-phase descriptors derived from molecular dynamics (MD) simulations. Using a curated subset of the DIPPR 801 database, multiple machine learning architectures, including light gradient boosting machine (LightGBM) and graph convolutional networks, were evaluated with feature sets of increasing physical fidelity. The best-performing model, based on LightGBM trained on Dragon descriptors augmented with MD and quantum chemical features, achieves a mean absolute error of 22.5 K, outperforming descriptor-only models and structure-based deep learning baselines. Shapley additive explanations interpretability analysis reveals that melting behavior is governed primarily by molecular topology, surface-area-weighted electronic descriptors, and condensed-phase interaction properties. In contrast, many isolated functional group and single molecule electronic descriptors contribute negligibly once these effects are accounted for. These results demonstrate that incorporating physics-informed, multi-scale descriptors enables more accurate and physically interpretable MP predictions.
Frank T. Mtetwa, N. Giles, W. Wilding et al.· Journal of Chemical Physics· 0 citations
Hydrates are common solid forms that can significantly affect a compound’s stability, physicochemical properties, and commercial viability. Despite their importance in pharmaceutical development, hydrate structures, their relationship to material properties, and their propensity for formation remain poorly understood. In 2003, Gillon et al. introduced a simple yet powerful framework for the structural classification of hydrates based on the hydrogen-bonding environment of crystalline water molecules. This framework categorizes water molecules according to the number of hydrogen-bond donor (D) and acceptor (A) interactions in which they participate, giving rise to eight distinct environments. In the original study, the DDA environment, where water donates two hydrogen bonds and accepts one, was reported as the most common. The statistical distribution of these environments in the Cambridge Structural Database (CSD) has since provided a useful benchmark for the qualitative assessment of hydrate structures. Here, we revisit Gillon’s water environment classification using a substantially expanded data set comprising 13,881 hydrate entries from the CSD and provide updated statistics on hydrate structures. Our analysis confirms that the DDA environment remains the most prevalent, followed by DDAA and DD, while the remaining environments occur less frequently. Extending beyond the original work, we quantify the energetics associated with each environment and demonstrate that the four-hydrogen-bond DDAA environment is energetically the most favorable, followed by environments involving three, two, and one hydrogen bonds, respectively. We further show that water-mediated intermolecular interactions contribute up to 40% of the total lattice interaction energy in the hydrate structures, highlighting the surprisingly large influence of this small solvent molecule on crystal stability. Despite its superior energetic stability, the DDAA environment is not the most frequently observed experimentally. This apparent discrepancy arises because the hydrogen-bonding capability and topology of the main component in the hydrate constrains the maximum hydrogen-bonding environment that water can achieve. Overall, this work provides a comprehensive analysis of hydrate structures and energetics across the CSD. The resulting insights offer a valuable framework for the qualitative assessment of newly discovered hydrate forms and for evaluating whether their structures conform to established crystallographic trends.
Henry A. Holleb, Fragkoulis Theodosiou, Pablo Martinez-Bulit et al.· Crystal Growth & Design· 0 citations
Abstract The single crystal structure of Sodium Suphanilate Dihydrate (SSDH) features infinitely connected cationic metal-organic frameworks (MOFs) of Sodium and Oxygen atoms blended with strong and moderate hydrogen bonding interactions. The crystal packing shows alternate hydrophilic and hydrophobic regions along the c-axis of the unit cell at z = 0, ½ and 1 and at z = ¼ and ¾, respectively. The powder XRD pattern confirms the crystalline phase of the grown crystal. The molecular structure of SSDH was quantum chemically optimized by DFT method. The various functional group species of the compound was investigated by Fourier Transformed Infra-Red and Raman spectroscopic techniques and compared with quantum chemically computed counterparts. The calculated Frontier Molecular Orbital energies reveal the charge transfer characteristic of SSDH along with the chemical hardness, electro-negativity and electrophilicity index which are key to make optoelectronic devices. The atomic charge distribution of the molecular assembly was analyzed by Mulliken Charge analysis method. The UV–Visible absorbance spectra exhibit a characteristic peak of absorbance at λ = 248 nm. The nonlinear optical parameters of the crystal such as absorption coefficient, susceptibilities and nonlinear refractive index were determined by the Z-scan technique.
J. J. Belciya, R. Anitha, M. Roshan et al.· Zeitschrift fur physikalisch...· 0 citations
Machine learning interatomic potentials (MLIPs) have become emerging tools in molecular modeling and computational chemistry. By learning high-dimensional potential energy surfaces from quantum chemical data, MLIPs enable accurate and efficient predictions of structural, thermodynamic, and dynamical properties. However, such models have limitations in predictions of electronic properties and the effects of static electron correlation due to their lack of electronic structure information. This work presents OrbGNN, an electronic structure graph architecture analogous to molecular graph and MLIP frameworks, where pair-orbital interactions constitute the graph representation, while orbital entanglement encodes the connectivity between them. By embedding information derived from orbital correlation metrics directly into the graph topology, OrbGNN provides a compact representation of a molecule s orbital landscape and electron correlation patterns. Analysis of the behavior of the feature space in an orbital graph are shown to demonstrate model robustness. The model is evaluated for the dissociation of nitrogen and for a larger dataset of diatomic molecules. Finally, the OrbGNN model is applied to a set of octahedral iron(II) complexes to predict spin-state energy gaps.
Brody Quebedeaux, Shahzad Akram, Markus Reiher et al.· 0 citations
The structure and dynamical behavior of water confined at or within nanostructures is a topic central to many fields, from biology to emerging electronics such as carbon nanostructures. Nanoporous graphene (NPG) containing periodic nanoscale pores with specific topologies has emerged as a promising material in carbon-based nanoelectronics; however, its interaction with ambient water remains poorly understood. Here, we combine density functional theory (DFT), ab initio molecular dynamics (AIMD), and interpretable machine learning (ML) to reveal how water controls quantum transport in NPGs. Depending on the local hydration structure, the bandgap varies by more than a factor of two across NPG and nitrogen-doped hybrid (h-NPG) systems. To uncover the underlying mechanism, we develop Smooth Overlap of Atomic Positions (SOAP)-based black-box and physics-informed grey-box ML models. The Gaussian process regression model achieves near-DFT accuracy while enabling physical interpretation. Analysis identifies water dipole orientation, water-substrate distance, water center-of-geometry, and ribbon-resolved dipole moments as the dominant factors controlling bandgap modulation across NPG and h-NPG systems.
Sneha Mittal, Alan E. Anaya Morales, V. Rosendal et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
