In materials science, crystal structures are the cornerstone in the structure–property paradigm. The description of crystal compounds may be ascribed to the number of different atomic chemical environments, which are related to the Wyckoff sites. Hence, a set of features related to the different atomic environments in a crystal compound can be constructed as input data for artificial neural networks (ANNs). In this article, we show the performance of a series of ANNs developed using crystal-site-based features. These ANNs were developed to classify compounds into halite, garnet, fluorite, hexagonal perovskite, ilmenite, layered perovskite, -o-tp- perovskite, perovskite, and spinel structures. Using crystal-site-based features, the ANNs were able to classify the crystal compounds with a 93.72% average precision. Furthermore, the ANNs were able to retrieve missing compounds with one of these archetypical structure types from a database. Finally, we showed that the developed ANNs were also suitable for a multitask learning paradigm, since the extracted information in the hidden layers linearly correlated with lattice parameters of the crystal structures.
Attainment of a pristine copper cobaltite (CuCo2O4) phase by a fast and easy microwave (MW) assisted method and its application in the selective oxidation of propane is presented in this paper.
In this article, we present a variational autoencoder (VAE) neural network for generating in-silico diffraction patterns of multiphase samples from their single-phase components. This work makes five main contributions. First, we provide the first quantitative quality metrics for generative XRD patterns – PSNR (38.10 dB), SSIM (0.9688), and w-MAE (0.0271) – establishing benchmarks absent for this application. Second, we introduce an attention-enhanced VAE with a novel loss function that incorporates cosine similarity from a pretrained model, effectively suppressing spurious reflections. Third, we demonstrate that linear interpolation of single-phase latent representations generates realistic binary and ternary patterns without training on mixtures. Fourth, we validate generated patterns via Rietveld refinement, revealing that pixel-wise metrics do not fully capture crystallographic fidelity. Fifth, using features from the same pretrained model, a shallow neural network estimates phase concentrations with R2 = 0.9989 on test data and 5.91% MAE on experimental samples. Importantly, this work presents an alternative methodology that operates directly on measured diffraction patterns, rather than relying on crystallographic information files (CIFs) and subsequent pattern simulation. It does not aim to replace established physics-based simulation tools (e.g., TOPAS, GSAS-II), which remain the gold standard for generating ground-truth patterns when structural models are available. Instead, our approach offers a complementary pathway for scenarios where CIFs are unavailable, incomplete, or when the goal is to learn experimental artifact distributions directly from data. This distinction positions our method not as a competitor to physics-based simulation but as an additional tool in the materials informatics toolbox.