Search

Due to their compositional complexity, refractory multi-principal element alloys (RMPEAs) exhibit a diverse range of material properties, making them highly suitable for several applications. Importantly, single-phase BCC RMPEAs are known for their superior strength. Nevertheless, identifying the stability of single-phase BCC is challenging because of the vast compositional space of RMPEAs. In this study, we develop a deep learning framework that predicts the fraction of RMPEA phases with high accuracy at several temperatures. Using the accurate predictive framework, we navigate the compositional space of Ti, Fe, Al, V, Ni, Nb and Zr elements to search for potentially stable single-phase BCC RMPEAs and provide design insights that can guide the synthesis of new RMPEAs in experiments. Furthermore, we develop a mathematical equation that enables the rapid and accurate identification of single-phase BCC RMPEAs with high accuracy. The machine learning (ML) model is finally evaluated against a high-fidelity experimental data from literature, confirming that our model accurately predicts the constituent phases of the experimental samples while provides insights that are laborious to extract from experiments.

The tunability of the mechanical properties of refractory multi-principal-element alloys (RMPEAs) make them attractive for numerous high-temperature applications. It is well-established that the phase stability of RMPEAs control their mechanical properties. In this study, we develop a deep learning framework that is trained on a CALPHAD-derived database that is predictive of RMPEAs phases with high accuracy up to eight phases within the elemental space of Ti, Fe, Al, V, Ni, Nb, and Zr with an accuracy of approximately 90%. We further investigate the causes for the low out of domain performance of the deep learning models in predicting phases of RMPEA with new elemental sets and propose a strategy to mitigate this performance shortfall.

A central challenge of the clean energy transition is the development of catalysts for low-emissions technologies. Recent advances in Machine Learning for quantum chemistry drastically accelerate the computation of catalytic activity descriptors such as adsorption energies. Here we introduce AdsorbRL, a Deep Reinforcement Learning agent aiming to identify potential catalysts given a multi-objective binding energy target, trained using offline learning on the Open Catalyst 2020 and Materials Project data sets. We experiment with Deep Q-Network agents to traverse the space of all ~160,000 possible unary, binary and ternary compounds of 55 chemical elements, with very sparse rewards based on adsorption energy known for only between 2,000 and 3,000 catalysts per adsorbate. To constrain the actions space, we introduce Random Edge Traversal and train a single-objective DQN agent on the known states subgraph, which we find strengthens target binding energy by an average of 4.1 eV. We extend this approach to multi-objective, goal-conditioned learning, and train a DQN agent to identify materials with the highest (respectively lowest) adsorption energies for multiple simultaneous target adsorbates. We experiment with Objective Sub-Sampling, a novel training scheme aimed at encouraging exploration in the multi-objective setup, and demonstrate simultaneous adsorption energy improvement across all target adsorbates, by an average of 0.8 eV. Overall, our results suggest strong potential for Deep Reinforcement Learning applied to the inverse catalysts design problem., Comment: 37th Conference on Neural Information Processing Systems (NeurIPS 2023), AI for Accelerated Materials Design Workshop

Statistics of grain sizes and orientations in metals correlate to the material's mechanical properties. Reproducing representative volume elements for further analysis of deformation and failure in metals, like 316L stainless steel, is particularly important due to their wide use in manufacturing goods today. Two approaches, initially created for video games, were considered for the procedural generation of representative grain microstructures. The first is the Wave Function Collapse (WFC) algorithm, and the second is constraint propagation and probabilistic inference through Markov Junior, a free and open-source software. This study aimed to investigate these two algorithms' effectiveness in using reference electron backscatter diffraction (EBSD) maps and recreating a statistically similar one that could be used in further research. It utilized two stainless steel EBSD maps as references to test both algorithms. First, the WFC algorithm was too constricting and, thus, incapable of producing images that resembled EBSDs. The second, MarkovJunior, was much more effective in creating a Voronoi tessellation that could be used to create an EBSD map in Python. When comparing the results between the reference and the generated EBSD, we discovered that the orientation and volume fractions were extremely similar. With the study, it was concluded that MarkovJunior is an effective machine learning tool that can reproduce representative grain microstructures., Comment: 6 pages, 18 figures

Plasticity in hexagonal close packed (HCP) metals and alloys such as Titanium (Ti) and Zirconium (Zr) is carried out by the motion of $\langle a \rangle$ dislocations. Above room temperature, in situ transmission electron microscopy straining experiments have shown that basal slip of $\langle a \rangle$ screw dislocations in Ti is activated and its preponderance increases with temperature. To discern the mechanism of basal slip of these dislocations in pure Ti, molecular dynamics (MD) simulations and nudged elastic band (NEB) calculations were performed for a screw dislocation gliding on the basal plane. From MD simulations, basal and pyramidal slip were shown to be intertwined. The screw dislocation glides by kink pair nucleation and propagation mechanism on one of the two equiprobable pyramidal planes adjacent to the maximum resolved shear stress basal plane, before switching to the other pyramidal plane, thus leading to an average basal slip at low stress and/or high temperature. On the other hand, NEB calculations revealed that at low stresses kinks do not nucleate in the pyramidal plane but rather directly in the basal plane fully compatible with the motion of a screw dislocation confined to the basal plane, in agreement with experimental observations in Ti and Zr. The discrepancy between these two mechanisms is shown to be related to the thermal activation of the cross-slip from the pyramidal to the prismatic core configuration of the $\langle a \rangle$ screw dislocation.

Slip transmission across $\alpha-\beta$ interfaces is of great significance in understanding the strength of $\alpha-\beta$ Ti alloys for aerospace applications. Molecular statics (MS) and molecular dynamics (MD) simulations were conducted to investigate the mechanisms of slip transmission of $\langle a \rangle$ prismatic screw dislocations across the $\alpha-\beta$ interface side face. In these simulations, the $\alpha$ phase consisted of pure HCP Ti whereas the $\beta$ phase was modeled as Ti$_{60}$Nb$_{40}$ BCC random alloy using a Ti-Nb interatomic potential. Firstly, the misfit dislocations structure on the $\alpha-\beta$ interface side face has been characterized from MS simulations. This predicted dislocation structure is in good agreement with experimental observations in Ti-alloys. Secondly, MD simulations of the interaction of $[a_1]$, $[a_2]$, and $[a_3]$ prismatic screw dislocations with the interface side face in an $\alpha-\beta$ bi-crystal were performed at different temperatures. A distinct barrier of the interface side face to different types of dislocation transmission was found. The origin of this anisotropy in the slip transmission is due to the relative misalignment of the slip systems between the $\alpha$ and the $\beta$ phases from the Burgers orientation relationship. Finally, the mechanisms of slip transmission of each dislocation type were analyzed in a detailed atomistic description and compared to experimental observations.

Mesoscale simulations of discrete defects in metals provide an ideal framework to investigate the micro-scale mechanisms governing the plastic deformation under high thermal and mechanical loading conditions. To bridge size and time-scale while limiting computational effort, typically the concept of representative volume elements (RVEs) is employed. This approach considers the microstructure evolution in a volume that is representative of the overall material behavior. However, in settings with complex thermal and mechanical loading histories careful consideration of the impact of modeling constraints in terms of time scale and simulation domain on predicted results is required. We address the representation of heterogeneous dislocation structure formation in simulation volumes using the example of residual stress formation during cool-down of laser powder-bed fusion (LPBF) of AISI 316L stainless steel. This is achieved by a series of large-scale three-dimensional discrete dislocation dynamics (DDD) simulations assisted by thermo-mechanical finite element modeling of the LPBF process. Our results show that insufficient size of periodic simulation domains can result in dislocation patterns that reflect the boundaries of the primary cell. More pronounced dislocation interaction observed for larger domains highlight the significance of simulation domain constraints for predicting mechanical properties. We formulate criteria that characterize representative volume elements by capturing the conformity of the dislocation structure to the bulk material. This work provides a basis for future investigations of heterogeneous microstructure formation in mesoscale simulations of bulk material behavior.

The mechanical properties of a material are intimately related to its microstructure. This is particularly important for predicting mechanical behavior of polycrystalline metals, where microstructural variations dictate the expected material strength. Until now, the lack of microstructural variability in available datasets precluded the development of robust physics-based theoretical models that account for randomness of microstructures. To address this, we have developed a probabilistic machine learning framework to predict the flow stress as a function of variations in the microstructural features. In this framework, we first generated an extensive database of flow stress for a set of over a million randomly sampled microstructural features, and then applied a combination of mixture models and neural networks on the generated database to quantify the flow stress distribution and the relative importance of microstructural features. The results show excellent agreement with experiments and demonstrate that across a wide range of grain size, the conventional Hall-Petch relationship is statistically valid for correlating the strength to the average grain size and its comparative importance versus other microstructural features. This work demonstrates the power of the machine-learning based probabilistic approach for predicting polycrystalline strength, directly accounting for microstructural variations, resulting in a tool to guide the design of polycrystalline metallic materials with superior strength, and a method for overcoming sparse data limitations., Comment: The dataset generated and the demo codes used in this paper are available on Zenodo (https://doi.org/10.5281/zenodo.7762663)

We reconstruct a 3D model of the surface of a material undergoing fatigue testing and experiencing cracking. Specifically we reconstruct the surface depth (out of plane intrusions and extrusions) and lateral (in-plane) motion from multiple views of the sample at the end of the experiment, combined with a reverse optical flow propagation backwards in time that utilizes interim single view images. These measurements can be mapped to a material strain tensor which helps to understand material life and predict failure. This approach offers an alternative to the commonly used Digital Image Correlation (DIC) technique which relies on tracking a speckle pattern applied to the material surface. DIC only produces in-plane (2D) measurements whereas our approach is 3D and non-invasive (requires no pattern being applied to the material)., Comment: 8 pages, 12 figures

There is a massive underserved market for small business lending in the US with the Federal Reserve estimating over \$650B in unmet annual financing needs. Assessing the credit risk of a small business is key to making good decisions whether to lend and at what terms. Large corporations have a well-established credit assessment ecosystem, but small businesses suffer from limited publicly available data and few (if any) credit analysts who cover them closely. We explore the applicability of (DL-based) large corporate credit risk models to small business credit rating., Comment: 5 pages

While friction stems from the fundamental interactions between atoms at a contact interface, its best descriptions at the macroscopic scale remain phenomenological. The so called "rate-and-state" models, which specify the friction response in terms of the relative sliding velocity and the "age" of the contact interface, fail to uncover the nano-scale mechanisms governing the macro-scale response, while models of friction at the atomic scale often overlook how roughness can alter the friction behavior. Here we bridge this gap between nano and macro descriptions of friction by correlating the physical origin of macroscopic friction to the existence, due to nanometric roughness, of contact junctions between adsorbed monolayers. Their dynamics, as we show, emerges from molecular motion. Through coupled experimental and atomic simulations, we highlight that transient friction overshoots its steady-state value after the system is allowed to rest, with the friction force decaying to a steady-state value over a distance of a few nanometers, much smaller than the junction size, even with a root-mean-square roughness of 0.6 nm. We demonstrate how this transient decay is intrinsically related to the evolution of the number of cross-surface attractive physical links between adsorbed molecules on rough surfaces. We also show that roughness is a sufficient condition for the appearance of frictional aging. In systems that show structural aging, this paints contact junctions as a key component in the observation of the transient friction overshoot, and shows how infrajunction molecular motion can control the macroscopic response.

Additive manufacturing (AM) of metallic components promises many advantages over conventional manufacturing processes through high design flexibility across multiple length scales and precision coupled with an astonishing combination of mechanical properties. Characterizing the relationship between microstructure and mechanical properties remains one of the major challenges for this novel technology. A natural precursor is identifying the influence of the processing path on the developing microstructure. We combine experimental studies of single-track laser powder bed fusion (LPBF) scans of AISI 316L stainless steel, finite element analyses, and large-scale three-dimensional discrete dislocation dynamics simulations to provide a unique understanding of the underlying mechanisms leading to the formation of heterogeneous defect structures in additively manufactured metals. Our results show that the interruption of dislocation slip at solidification cell walls is responsible for the formation of cellular dislocation structures, highlighting the significance of solute segregation for plastic deformation of additively manufactured components. This work provides a mechanistic perspective on heterogeneous microstructure formation and opens the potential for a reliable prediction of the resulting mechanical properties of additively manufactured parts., Comment: Accepted manuscript