Accurate, affordable, and generalizable machine learning simulations of transition metal x-ray absorption spectra using the XANESNET deep neural network

Emerging Atomistic Modeling Methods for Heterogeneous Electrocatalysis

Heterogeneous electrocatalysis lies at the center of various technologies that could help enable a sustainable future. However, its complexity makes it challenging to accurately and efficiently model at an atomic level. Here, we review emerging atomistic methods to simulate the electrocatalytic interface with special attention devoted to the components/effects that have been challenging to model, such as solvation, electrolyte ions, electrode potential, reaction kinetics, and pH. Additionally, we review relevant computational spectroscopy methods. Then, we showcase several examples of applying these methods to understand and design catalysts relevant to green hydrogen. We also offer experimental views on how to bridge the gap between theory and experiments. Finally, we provide some perspectives on opportunities to advance the field.

Improving sensitivity of XANES structural fit to the bridged metal-metal coordination

Hard X-ray absorption spectroscopy is a valuable in situ probe for non-destructive diagnostics of metal sites. The low-energy interval of a spectrum (XANES) contains information about the metal oxidation state, ligand type, symmetry and distances in the first coordination shell but shows almost no dependency on the bridged metal-metal bond length. The higher-energy interval (EXAFS), on the contrary, is more sensitive to the coordination numbers and can decouple the contribution from distances in different coordination shells. Supervised machine-learning methods can combine information from different intervals of a spectrum; however, computational approaches for the near-edge region of the spectrum and higher energies are different. This work aims to keep all benefits of XANES and extend its sensitivity towards the interatomic distances in the first and second coordination shells. Using a binuclear bridged copper complex as a case study and cross-validation analysis as a quantitative tool it is shown that the first 170 eV above the edge are already sufficient to balance the contributions of Cu-O/N scattering and Cu-Cu scattering. As a more general outcome this work highlights the trivial but often overlooked importance of using `longer' energy intervals of XANES for structural refinement and machine-learning predictions. The first 200 eV above the absorption edge still do not require parametrization of Debye-Waller damping and can be calculated within full multiple scattering or finite difference approximations with only moderately increased computational costs.

Accurate, Uncertainty-Aware Classification of Molecular Chemical Motifs from Multimodal X-ray Absorption Spectroscopy

Accurate classification of molecular chemical motifs from experimental measurement is an important problem in molecular physics, chemistry, and biology. In this work, we present neural network ensemble classifiers for predicting the presence (or lack thereof) of 41 different chemical motifs on small molecules from simulated C, N, and O K-edge X-ray absorption near-edge structure (XANES) spectra. Our classifiers not only achieve class-balanced accuracies of more than 0.95 but also accurately quantify uncertainty. We also show that including multiple XANES modalities improves predictions notably on average, demonstrating a "multimodal advantage" over any single modality. In addition to structure refinement, our approach can be generalized to broad applications with molecular design pipelines.

Stability Trends in disubstituted Cobaltocenium Based on the Analysis of the Machine Learning Models

Cobaltocenium derivatives have shown great potential as components of anion exchange membranes in fuel cells because they exhibit excellent thermal and alkaline stability under operating conditions while allowing for high anion mobility. The properties of the cobaltocenium-anion complexes can be chemically tuned through the substituent groups on the cyclopentadienyl (Cp) rings of the cation CoCp2+. However, the synthesis and characterization of the full range of possible derivatives are very challenging and time-consuming, and while the computational tools can greatly expedite this process, full screening of the electronic structure at a high level of theory is still computationally intensive. Therefore, in this work, we consider the machine learning (ML) modeling as a tool of predicting stability of disubstituted [CoCp2]OH complexes measured by their bond-dissociation energy (BDE). The relevant process here is the dissociation of the cobaltocenium-hydroxide complex into fragments [CoCpY']OH and CpY, where Y and Y' each represent one out of 42 substituent groups of experimental interest. In agreement with the previous ML study of 120 mono- and selected disubstituted species [Wetthasinghe et al. J. Chem. Phys. A (2022) 126], our analysis of the data set expanded to all possible disubstituted cobaltoceniums, points to the highest occupied and lowest unoccupied molecular orbitals, along with the Hirshfeld charge on the singly substituted benzene, to be the key features predicting the BDE of the unseen complexes. Based on the examination of the outliers, the acidity of substituents ((CO)NH2 in our case) is found to be of special significance for the cobaltocenium stability and for the model development. Moreover, we demonstrate that upon the data set refinement, the conventional ML models are capable of predicting the BDE close to 1 kcal/mol based on the properties of just the fragments, thereby greatly reducing the total number of species and of the computational time of each calculation. Such fragment-based "combinatorial" approach to the BDE modeling is noteworthy, since the geometry optimization of complexes in solution is conceptually challenging and computationally demanding, even when leveraging high-performance computing resources.

Machine learning for analysis of experimental scattering and spectroscopy data in materials chemistry

The rapid growth of materials chemistry data, driven by advancements in large-scale radiation facilities as well as laboratory instruments, has outpaced conventional data analysis and modelling methods, which can require enormous manual effort. To address this bottleneck, we investigate the application of supervised and unsupervised machine learning (ML) techniques for scattering and spectroscopy data analysis in materials chemistry research. Our perspective focuses on ML applications in powder diffraction (PD), pair distribution function (PDF), small-angle scattering (SAS), inelastic neutron scattering (INS), and X-ray absorption spectroscopy (XAS) data, but the lessons that we learn are generally applicable across materials chemistry. We review the ability of ML to identify physical and structural models and extract information efficiently and accurately from experimental data. Furthermore, we discuss the challenges associated with supervised ML and highlight how unsupervised ML can mitigate these limitations, thus enhancing experimental materials chemistry data analysis. Our perspective emphasises the transformative potential of ML in materials chemistry characterisation and identifies promising directions for future applications. The perspective aims to guide newcomers to ML-based experimental data analysis.

A Δ-learning strategy for interpretation of spectroscopic observables

Accurate computations of experimental observables are essential for interpreting the high information content held within x-ray spectra. However, for complicated systems this can be difficult, a challenge compounded when dynamics becomes important owing to the large number of calculations required to capture the time-evolving observable. While machine learning architectures have been shown to represent a promising approach for rapidly predicting spectral lineshapes, achieving simultaneously accurate and sufficiently comprehensive training data is challenging. Herein, we introduce Δ-learning for x-ray spectroscopy. Instead of directly learning the structure-spectrum relationship, the Δ-model learns the structure dependent difference between a higher and lower level of theory. Consequently, once developed these models can be used to translate spectral shapes obtained from lower levels of theory to mimic those corresponding to higher levels of theory. Ultimately, this achieves accurate simulations with a much reduced computational burden as only the lower level of theory is computed, while the model can instantaneously transform this to a spectrum equivalent to a higher level of theory. Our present model, demonstrated herein, learns the difference between TDDFT(BLYP) and TDDFT(B3LYP) spectra. Its effectiveness is illustrated using simulations of Rh L3-edge spectra tracking the C-H activation of octane by a cyclopentadienyl rhodium carbonyl complex.

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.

Deep Reinforcement Learning Environment Approach Based on Nanocatalyst XAS Diagnostics Graphic Formalization

The most in-demand instrumental methods for new functional nanomaterial diagnostics employ synchrotron radiation, which is used to determine a material's electronic and local atomic structure. The high time and resource costs of researching at international synchrotron radiation centers and the problems involved in developing an optimal strategy and in planning the control of the experiments are acute. One possible approach to solving these problems involves the use of deep reinforcement learning agents. However, this approach requires the creation of a special environment that provides a reliable level of response to the agent's actions. As the physical experimental environment of nanocatalyst diagnostics is potentially a complex multiscale system, there are no unified comprehensive representations that formalize the structure and states as a single digital model. This study proposes an approach based on the decomposition of the experimental system into the original physically plausible nodes, with subsequent merging and optimization as a metagraphic representation with which to model the complex multiscale physicochemical environments. The advantage of this approach is the possibility to directly use the numerical model to predict the system states and to optimize the experimental conditions and parameters. Additionally, the obtained model can form the basic planning principles and allow for the optimization of the search for the optimal strategy with which to control the experiment when it is used as a training environment to provide different abstraction levels of system state reactions.

An on-the-fly deep neural network for simulating time-resolved spectroscopy: predicting the ultrafast ring opening dynamics of 1,2-dithiane

Revolutionary developments in ultrafast light source technology are enabling experimental spectroscopists to probe the structural dynamics of molecules and materials on the femtosecond timescale. The capacity to investigate ultrafast processes afforded by these resources accordingly inspires theoreticians to carry out high-level simulations which facilitate the interpretation of the underlying dynamics probed during these ultrafast experiments. In this Article, we implement a deep neural network (DNN) to convert excited-state molecular dynamics simulations into time-resolved spectroscopic signals. Our DNN is trained on-the-fly from first-principles theoretical data obtained from a set of time-evolving molecular dynamics. The train-test process iterates for each time-step of the dynamics data until the network can predict spectra with sufficient accuracy to replace the computationally intensive quantum chemistry calculations required to produce them, at which point it simulates the time-resolved spectra for longer timescales. The potential of this approach is demonstrated by probing dynamics of the ring opening of 1,2-dithiane using sulphur K-edge X-ray absorption spectroscopy. The benefits of this strategy will be more markedly apparent for simulations of larger systems which will exhibit a more notable computational burden, making this approach applicable to the study of a diverse range of complex chemical dynamics.

Beyond structural insight: a deep neural network for the prediction of Pt L2/3-edge X-ray absorption spectra

X-ray absorption spectroscopy at the L_2/3 edge can be used to obtain detailed information about the local electronic and geometric structure of transition metal complexes. By virtue of the dipole selection rules, the transition metal L_2/3 edge usually exhibits two distinct spectral regions: (i) the "white line", which is dominated by bound electronic transitions from metal-centred 2p orbitals into unoccupied orbitals with d character; the intensity and shape of this band consequently reflects the d density of states (d-DOS), which is strongly modulated by mixing with ligand orbitals involved in chemical bonding, and (ii) the post-edge, where oscillations encode the local geometric structure around the X-ray absorption site. In this Article, we extend our recently-developed XANESNET deep neural network (DNN) beyond the K-edge to predict X-ray absorption spectra at the Pt L_2/3 edge. We demonstrate that XANESNET is able to predict Pt L_2/3 -edge X-ray absorption spectra, including both the parts containing electronic and geometric structural information. The performance of our DNN in practical situations is demonstrated by application to two Pt complexes, and by simulating the transient spectrum of a photoexcited dimeric Pt complex. Our discussion includes an analysis of the feature importance in our DNN which demonstrates the role of key features and assists with interpreting the performance of the network.