Search for: All records

In the present paper, we introduce a new neural network-based tool for the prediction of formation energies of atomic structures based on elemental and structural features of Voronoi-tessellated materials. We provide a concise overview of the connection between the machine learning and the true material–property relationship, how to improve the generalization accuracy by reducing overfitting, how new data can be incorporated into the model to tune it to a specific material system, and preliminary results on using models to preform local structure relaxations. The present work resulted in three final models optimized for (1) highest test accuracy on the Open Quantum Materials Database (OQMD), (2) performance in the discovery of new materials, and (3) performance at a low computational cost. On a test set of 21,800 compounds randomly selected from OQMD, they achieve a mean absolute error (MAE) of 28, 40, and 42 meV/atom, respectively. The second model provides better predictions in a test case of interest not present in the OQMD, while the third reduces the computational cost by a factor of 8. We collect our results in a new open-source tool called SIPFENN (Structure-Informed Prediction of Formation Energy using Neural Networks). SIPFENN not only improves the accuracy beyond existing models but also ships in a ready-to-use form with pre-trained neural networks and a GUI interface. By virtue of this, it can be included in DFT calculations routines at nearly no cost. 

Forming metallurgical phases has a critical impact on the performance of dissimilar materials joints. Here, we shed light on the forming mechanism of equilibrium and non-equilibrium intermetallic compounds (IMCs) in dissimilar aluminum/steel joints with respect to processing history (e.g., the pressure and temperature profiles) and chemical composition, where the knowledge of free energy and atomic diffusion in the Al–Fe system was taken from first-principles phonon calculations and data available in the literature. We found that the metastable and ductile (judged by the presently predicted elastic constants) Al₆Fe is a pressure (P) favored IMC observed in processes involving high pressures. The MoSi₂-type Al₂Fe is brittle and a strongP-favored IMC observed at high pressures. The stable, brittle η-Al₅Fe₂is the most observed IMC (followed by θ-Al₁₃Fe₄) in almost all processes, such as fusion/solid-state welding and additive manufacturing (AM), since η-Al₅Fe₂is temperature-favored, possessing high thermodynamic driving force of formation and the fastest atomic diffusivity among all Al–Fe IMCs. Notably, the ductile AlFe₃, the less ductile AlFe, and most of the other IMCs can be formed during AM, making AM a superior process to achieve desired IMCs in dissimilar materials. In addition, the unknown configurations of Al₂Fe and Al₅Fe₂were also examined by machine learning based datamining together with first-principles verifications and structure predictions. All the IMCs that are notP-favored can be identified using the conventional equilibrium phase diagram and the Scheil-Gulliver non-equilibrium simulations.