Ismaila Dabo

Minimum-energy paths and intrinsic reaction coordinates are central tools for analyzing chemical reactions. As one-dimensional objects, however, individual paths do not provide a continuous representation of the surrounding potential energy surface (PES), including the arrangement of adjacent valleys, ridges, and higher-order saddle regions. We introduce the nudged elastic membrane method, which constructs a two-dimensional reduced PES in a selected region of configuration space using energies and forces, without requiring Hessian evaluations. The method is demonstrated for a three-dimensional analytical model and for triplet formaldehyde. In both cases, the membrane recovers known reaction-path and stationary-point regions. For formaldehyde, it also reveals a previously unreported second-order saddle point on the fitted PES. Structures extracted from the membrane provide starting points for subsequent stationary-point refinement.

Hydrogen embrittlement, a critical concern for the mechanical response of engineering materials, can arise due to an influx of hydrogen atoms at interstitial sites and at grain boundaries. The acoustic nonlinearity parameter (ANP) is used in nondestructive evaluation as a sensitive parameter for the early detection of material degradation. From a measurement perspective, the ANP can be determined from the distortion of elastic waves. From a modeling perspective, the ANP is computed from second and third-order elastic constants. This study investigates the influence of solute hydrogen atoms on the ANP in aluminum using results of density functional theory calculations as input to continuum-scale computations of elastic constants. Based on the sensitivity of the ANP to hydrogen solute atoms, the findings suggest that an additive decomposition of the ANP is not applicable. Additionally, approaches based upon the stress or strain caused by local heterogeneity (such as solute atoms), without including the heterogeneity itself may be misleading with regard to the ANP. Moreover, the general expectation that atomistic and microscale defects increase ANP may not be universally valid because we observed a decrease in ANP due to interstitial hydrogen atoms and grain boundaries. This work provides novel insights into the application of nonlinear acoustics for detecting atomistic-scale defects and lays the groundwork for a more accurate connection between acoustic measurements and hydrogen-related degradation in materials.

Ferroelectrics are of practical interest for non‐volatile data storage due to their reorientable, crystallographically defined polarization. Yet efforts to integrate conventional ferroelectrics into ultrathin memories have been frustrated by film‐thickness limitations, which impede polarization reversal under low applied voltage. Wurtzite materials, including magnesium‐substituted zinc oxide (Zn,Mg)O, have been shown to exhibit scalable ferroelectricity as thin films. In this work, the origins of ferroelectricity in (Zn,Mg)O are explained, showing that large strain fluctuations emerge locally in (Zn,Mg)O and can reduce local barriers to ferroelectric switching by more than 40%. Concurrent experimental and computational evidence of these effects are provided by demonstrating polarization switching in ZnO/(Zn,Mg)O/ZnO heterostructures featuring built‐in interfacial strain gradients. These results open up an avenue to develop scalable ferroelectrics by controlling strain fluctuations atomistically.

Recent progress towards universal machine-learned interatomic potentials holds considerable promise for materials discovery. Yet the accuracy of these potentials for predicting phase stability may still be limited. In contrast, cluster expansions provide accurate phase stability predictions but are computationally demanding to parameterize from first principles, especially for structures of low dimension or with a large number of components, such as interfaces or multimetal catalysts. We overcome this trade-off via transfer learning. Using Bayesian inference, we incorporate prior statistical knowledge from machine-learned and physics-based potentials, enabling us to sample the most informative configurations and to efficiently fit first-principles cluster expansions. This algorithm is tested on Pt:Ni, showing robust convergence of the mixing energies as a function of sample size with reduced statistical fluctuations.