Integrating Explainability into Graph Neural Network Models for the Prediction of X-ray Absorption Spectra

The use of sophisticated machine learning (ML) models, such as graph neural networks (GNNs), to predict complex molecular properties or all kinds of spectra has grown rapidly. However, ensuring the interpretability of these models' predictions remains a challenge. For example, a rigorous understanding of the predicted X-ray absorption spectrum (XAS) generated by such ML models requires an in-depth investigation of the respective black-box ML model used. Here, this is done for different GNNs based on a comprehensive, custom-generated XAS data set for small organic molecules. We show that a thorough analysis of the different ML models with respect to the local and global environments considered in each ML model is essential for the selection of an appropriate ML model that allows a robust XAS prediction. Moreover, we employ feature attribution to determine the respective contributions of various atoms in the molecules to the peaks observed in the XAS spectrum. By comparing this peak assignment to the core and virtual orbitals from the quantum chemical calculations underlying our data set, we demonstrate that it is possible to relate the atomic contributions via these orbitals to the XAS spectrum.

Oxidation-State Dynamics and Emerging Patterns in Magnetite

Magnetite is an important mineral with many interesting applications related to its magnetic, electrical, and thermal properties. Typically studied by electronic structure calculations, these methods are unable to capture the complex ion dynamics at relevant temperatures, time, and length scales. We present a hybrid Monte Carlo/molecular dynamics (MC/MD) method based on iron oxidation-state swapping for accurate atomistic modeling of bulk magnetite, magnetite surfaces, and nanoparticles that captures the complex ionic dynamics. By comparing the oxidation-state patterns with those obtained from density functional theory, we confirmed the accuracy of our approach. Lattice distortions leading to the stabilization of excess charges and a critical surface thickness at which the oxidation states transition from ordered to disordered were observed. This simple yet efficient approach paves the way for elucidating aspects of oxidation-state ordering of inverse spinel structures in general and battery materials in particular.

Reconstructing the infrared spectrum of a peptide from representative conformers of the full canonical ensemble

Leucine enkephalin (LeuEnk), a biologically active endogenous opioid pentapeptide, has been under intense investigation because it is small enough to allow efficient use of sophisticated computational methods and large enough to provide insights into low-lying minima of its conformational space. Here, we reproduce and interpret experimental infrared (IR) spectra of this model peptide in gas phase using a combination of replica-exchange molecular dynamics simulations, machine learning, and ab initio calculations. In particular, we evaluate the possibility of averaging representative structural contributions to obtain an accurate computed spectrum that accounts for the corresponding canonical ensemble of the real experimental situation. Representative conformers are identified by partitioning the conformational phase space into subensembles of similar conformers. The IR contribution of each representative conformer is calculated from ab initio and weighted according to the population of each cluster. Convergence of the averaged IR signal is rationalized by merging contributions in a hierarchical clustering and the comparison to IR multiple photon dissociation experiments. The improvements achieved by decomposing clusters containing similar conformations into even smaller subensembles is strong evidence that a thorough assessment of the conformational landscape and the associated hydrogen bonding is a prerequisite for deciphering important fingerprints in experimental spectroscopic data.

Atomistic Insight into the Hydration States of Layered Double Hydroxides

Effective protective coatings are an essential component of lightweight engineering materials in a large variety of applications as they ensure structural integrity of the base material throughout its whole service life. Layered double hydroxides (LDHs) loaded with corrosion inhibitors depict a promising approach to realize an active corrosion protection for aluminum and magnesium. In this work, we employed a combination of density functional theory and molecular dynamics simulations to gain a deeper understanding of the influence of intercalated water content on the structure, the stability, and the anion-exchange capacity of four different LDH systems containing either nitrate, carbonate, or oxalate as potential corrosion inhibiting agents or chloride as a corrosion initiator. To quantify the structural change, we studied the atom density distribution, radial distribution function, and orientation of the intercalated anions. Additionally, we determined the stability of the LDH systems by calculating their respective hydration energies, hydrogen-bonded network connected to the intercalated water molecules, as well as the self-diffusion coefficients of the intercalated anions to provide an estimate for the probability of their release after intercalation. The obtained computational results suggest that the hydration state of LDHs has a significant effect on their key properties like interlayer spacing and self-diffusion coefficients of the intercalated anions. Furthermore, we conclude from our simulation results that a high self-diffusion coefficient which is linked to the mobility of the intercalated anions is vital for its release via an anion-exchange mechanism and to subsequently mitigate corrosion reactions. Furthermore, the presented theoretical study provides a robust force field for the computer-assisted design of further LDH-based active anticorrosion coatings.

Elucidating Curvature-Capacitance Relationships in Carbon-Based Supercapacitors

Nanoscale surface curvatures, either convex or concave, strongly influence the charging behavior of supercapacitors. Rationalizing individual influences of electrode atoms to the capacitance is possible by interpreting distinct elements of the charge-charge covariance matrix derived from individual charge variations of the electrode atoms. An ionic liquid solvated in acetonitrile and confined between two electrodes, each consisting of three undulated graphene layers, serves as a demonstrator to illustrate pronounced and nontrivial features of the capacitance with respect to the electrode curvature. In addition, the applied voltage determines whether a convex or concave surface contributes to increased capacitance. While at lower voltages capacitance variations are in general correlated with ion number density variations in the double layer formed in the concave region of the electrode, for certain electrode designs a surprisingly strong contribution of the convex part to the differential capacitance is found both at higher and lower voltages.

Osmotic Transport at the Aqueous Graphene and hBN Interfaces: Scaling Laws from a Unified, First-Principles Description

Osmotic transport in nanoconfined aqueous electrolytes provides alternative venues for water desalination and "blue energy" harvesting. The osmotic response of nanofluidic systems is controlled by the interfacial structure of water and electrolyte solutions in the so-called electrical double layer (EDL), but a molecular-level picture of the EDL is to a large extent still lacking. Particularly, the role of the electronic structure has not been considered in the description of electrolyte/surface interactions. Here, we report enhanced sampling simulations based on ab initio molecular dynamics, aiming at unravelling the free energy of prototypical ions adsorbed at the aqueous graphene and hBN interfaces, and its consequences on nanofluidic osmotic transport. Specifically, we predicted the zeta potential, the diffusio-osmotic mobility, and the diffusio-osmotic conductivity for a wide range of salt concentrations from the ab initio water and ion spatial distributions through an analytical framework based on Stokes equation and a modified Poisson-Boltzmann equation. We observed concentration-dependent scaling laws, together with dramatic differences in osmotic transport between the two interfaces, including diffusio-osmotic flow and current reversal on hBN but not on graphene. We could rationalize the results for the three osmotic responses with a simple model based on characteristic length scales for ion and water adsorption at the surface, which are quite different on graphene and on hBN. Our work provides fundamental insights into the structure and osmotic transport of aqueous electrolytes on 2D materials and explores alternative pathways for efficient water desalination and osmotic energy conversion.

Constant potential simulations on a mesh

Molecular dynamics simulations in a constant potential ensemble are an increasingly important tool to investigate charging mechanisms in next-generation energy storage devices. We present a highly efficient approach to compute electrostatic interactions in simulations employing a constant potential method (CPM) by introducing a particle-particle particle-mesh solver specifically designed for treating long-range interactions in a CPM. Moreover, we present evidence that a dipole correction term-commonly used in simulations with a slab-like geometry-must be used with caution if it is also to be used within a CPM. It is demonstrated that artifacts arising from the usage of the dipole correction term can be circumvented by enforcing a charge neutrality condition in the evaluation of the electrode charges at a given external potential.

A Molecular Simulation Approach to Bond Reorganization in Epoxy Resins: From Curing to Deformation and Fracture

We model bond formation and dissociation processes in thermosetting polymer networks from molecular dynamics simulations. For this, a coarsened molecular mechanics model is derived from quantum calculations to provide effective interaction potentials that enable million-atoms scale simulations. The importance of bond (re)organization is demonstrated for (i) simulating epoxy resin formation-for which our approach leads to realistic network models which can now account for degrees of curing up to 98%. Moreover, (ii) we elucidate the competition of bond dissociation and bond reformation during plastic deformation and fracture. On this basis, we rationalize the molecular mechanisms that account for the irreversible nature of damaging epoxy polymers by mechanical load.

A first-principles analysis of the charge transfer in magnesium corrosion

Magnesium is the lightest structural engineering material and bears high potential to manufacture automotive components, medical implants and energy storage systems. However, the practical use of untreated magnesium alloys is restricted as they are prone to corrosion. An essential prerequisite for the control or prevention of the degradation process is a deeper understanding of the underlying corrosion mechanisms. Prior investigations of the formation of gaseous hydrogen during the corrosion of magnesium indicated that the predominant mechanism for this process follows the Volmer-Heyrovský rather than the previously assumed Volmer-Tafel pathway. However, the energetic and electronic states of both reaction paths as well as the charge state of dissolved magnesium have not been fully unraveled yet. In this study, density functional theory calculations were employed to determine these parameters for the Volmer, Tafel and Heyrovský steps to gain a comprehensive understanding of the major corrosion mechanisms responsible for the degradation of magnesium.