• Energy Research

We present an update of the DScribe package, a Python library for atomistic descriptors. The update extends DScribe’s descriptor selection with the Valle–Oganov materials fingerprint and provides descriptor derivatives to enable more advanced machine learning tasks, such as force prediction and structure optimization. For all descriptors, numeric derivatives are now available in DScribe. For the many-body tensor representation (MBTR) and the Smooth Overlap of Atomic Positions (SOAP), we have also implemented analytic derivatives. We demonstrate the effectiveness of the descriptor derivatives for machine learning models of Cu clusters and perovskite alloys.

Perovskites are promising materials candidates for optoelectronics, but their commercialization is hindered by toxicity and materials instability. While compositional engineering can mitigate these problems by tuning perovskite properties, the enormous complexity of the perovskite materials space aggravates the search for an optimal optoelectronic material. We conducted compositional space exploration through Monte Carlo (MC) convex hull sampling, which we made tractable with machine learning (ML). The ML model learns from density functional theory calculations of perovskite atomic structures, and can be used for quick predictions of energies, atomic forces, and stresses. We employed it in structural relaxations combined with MC sampling to gain access to low-energy structures and compute the convex hull for CsPb(Br1−xClx)3. The trained ML model achieves an energy prediction accuracy of 0.1 meV per atom. The resulting convex hull exhibits two stable mixing concentrations at 1/6 and 1/3 Cl contents. Our data-driven approach offers a pathway towards studies of more complex perovskites and other alloy materials with quantum mechanical accuracy. Peer reviewed

Finding low-energy structures of ligand-protected clusters is challenging due to the enormous conformational space and the high computational cost of accurate quantum chemical methods for determining the structures and energies of conformers. Here, we adopted and utilized a kernel rigid regression based machine learning method to accelerate the search for low-energy structures of ligand-protected clusters. We chose the Au25(Cys)18 (Cys: cysteine) cluster as a model system to test and demonstrate our method. We found that the low-energy structures of the cluster are characterized by a specific hydrogen bond type in the cysteine. The different configurations of the ligand layer influence the structural and electronic properties of clusters.