Discovering melting temperature prediction models of inorganic solids by combining supervised and unsupervised learning.

The melting temperature is important for materials design because of its relationship with thermal stability, synthesis, and processing conditions. Current empirical and computational melting point estimation techniques are limited in scope, computational feasibility, or interpretability. We report the development of a machine learning methodology for predicting melting temperatures of binary ionic solid materials. We evaluated different machine-learning models trained on a dataset of the melting points of 476 non-metallic crystalline binary compounds using materials embeddings constructed from elemental properties and density-functional theory calculations as model inputs. A direct supervised-learning approach yields a mean absolute error of around 180 K but suffers from low interpretability. We find that the fidelity of predictions can further be improved by introducing an additional unsupervised-learning step that first classifies the materials before the melting-point regression. Not only does this two-step model exhibit improved accuracy, but the approach also provides a level of interpretability with insights into feature importance and different types of melting that depend on the specific atomic bonding inside a material. Motivated by this finding, we used a symbolic learning approach to find interpretable physical models for the melting temperature, which recovered the best-performing features from both prior models and provided additional interpretability.

Origin of enhanced water oxidation activity in an iridium single atom anchored on NiFe oxyhydroxide catalyst.

The efficiency of the synthesis of renewable fuels and feedstocks from electrical sources is limited, at present, by the sluggish water oxidation reaction. Single-atom catalysts (SACs) with a controllable coordination environment and exceptional atom utilization efficiency open new paradigms toward designing high-performance water oxidation catalysts. Here, using operando X-ray absorption spectroscopy measurements with calculations of spectra and electrochemical activity, we demonstrate that the origin of water oxidation activity of IrNiFe SACs is the presence of highly oxidized Ir single atom (Ir5.3+) in the NiFe oxyhydroxide under operating conditions. We show that the optimal water oxidation catalyst could be achieved by systematically increasing the oxidation state and modulating the coordination environment of the Ir active sites anchored atop the NiFe oxyhydroxide layers. Based on the proposed mechanism, we have successfully anchored Ir single-atom sites on NiFe oxyhydroxides (Ir0.1/Ni9Fe SAC) via a unique in situ cryogenic-photochemical reduction method that delivers an overpotential of 183 mV at 10 mA ⋅ cm-2 and retains its performance following 100 h of operation in 1 M KOH electrolyte, outperforming the reported catalysts and the commercial IrO2 catalysts. These findings open the avenue toward an atomic-level understanding of the oxygen evolution of catalytic centers under in operando conditions.

Machine learning with bond information for local structure optimizations in surface science.

Local optimization of adsorption systems inherently involves different scales: within the substrate, within the molecule, and between the molecule and the substrate. In this work, we show how the explicit modeling of different characteristics of the bonds in these systems improves the performance of machine learning methods for optimization. We introduce an anisotropic kernel in the Gaussian process regression framework that guides the search for the local minimum, and we show its overall good performance across different types of atomic systems. The method shows a speed-up of up to a factor of two compared with the fastest standard optimization methods on adsorption systems. Additionally, we show that a limited memory approach is not only beneficial in terms of overall computational resources but can also result in a further reduction of energy and force calculations.

Elucidation of temperature-programmed desorption of high-coverage hydrogen on Pt(211), Pt(221), Pt(533) and Pt(553) based on density functional theory calculations.

In this work we compute high-coverage hydrogen adsorption energies and geometries on the stepped platinum surfaces Pt(211) and Pt(533) which contain a (100)-step type and the Pt(221) and Pt(553) surface with a (111) step edge. We discuss these results in relation to ultra-high-vacuum temperature programmed desorption (TPD) data to elucidate the origin of the desorption features. Our results indicated that on surfaces with a (100)-step type, two distinct ranges of adsorption energy for the step and terrace are observed, which mirrors the TPD spectra for which we find a clear separation of the desorption peaks. For the (111) step type, the TPD spectra show much less separation of the step and terrace features, which we assign to the low individual adsorption energies for H atoms on this step edge. From our results we obtain a much clearer understanding of the surface-hydrogen bonding at high coverages and the origin of the different TPD features present for the two step types studied.

Low-Scaling Algorithm for Nudged Elastic Band Calculations Using a Surrogate Machine Learning Model.

We present the incorporation of a surrogate Gaussian process regression (GPR) atomistic model to greatly accelerate the rate of convergence of classical nudged elastic band (NEB) calculations. In our surrogate model approach, the cost of converging the elastic band no longer scales with the number of moving images on the path. This provides a far more efficient and robust transition state search. In contrast to a conventional NEB calculation, the algorithm presented here eliminates any need for manipulating the number of images to obtain a converged result. This is achieved by inventing a new convergence criteria that exploits the probabilistic nature of the GPR to use uncertainty estimates of all images in combination with the force in the saddle point in the target model potential. Our method is an order of magnitude faster in terms of function evaluations than the conventional NEB method with no accuracy loss for the converged energy barrier values.

Strong Substrate Mediation of Attractive Lateral Interactions of CO on Cu(110).

The mechanism of chemical reactions between adsorbed species is defined by the combined effects of the adsorbate-substrate potential landscape and lateral interactions. Such lateral interactions are therefore integral to catalytic processes, but their study is often complicated by "substrate mediation", the regulation of a two-body potential between adsorbed particles by the surface itself. Substrate mediation can influence the sign and magnitude of lateral interactions. There are notable exceptions of ordered structures forming at low coverage, indicative of short-range attractive forces where repulsive forces are expected to dominate, suggesting a strong substrate-mediated contribution. To explore further the origins of such interactions, we have investigated the adsorption of CO on Cu(110) using a combination of low-temperature microscopy and first-principles calculations. Our studies reveal that lateral adsorbate interactions, which are constrained by the metal surface, regulate the bonding between the adsorbate and substrate. Anisotropic CO-CO coupling is seen to arise from a perfect balance between the intermolecular accumulation of charge that acts as a glue (chemical coupling) at sufficiently large distances to avoid repulsive effects (dipole-dipole coupling and Pauli's repulsion between electron clouds).

Prediction of Stable and Active (Oxy-Hydro) Oxide Nanoislands on Noble-Metal Supports for Electrochemical Oxygen Reduction Reaction.

Developing cost-effective oxygen electrocatalysts with high activity and stability is key to their commercialization. However, economical earth-abundant catalysts based on first-row transition-metal oxides suffer from low electrochemical stability, which is difficult to improve without compromising their activity. Here, using density functional theory calculations, we demonstrate that noble-metal supports lead to bifunctional enhancement of both the stability and the oxygen reduction reaction (ORR) activity of metal (oxy-hydro) oxide nanoislands. We observe a significant stabilization of supported nanoislands beyond the intrinsic stability limits of bulk phases, which originates from a favorable lattice mismatch and reductive charge transfer from oxophilic supports. We discover that interfacial active sites (located between the nanoisland and the support) reinforce the binding strength of reaction intermediates, hence boosting ORR activity. Considering that both stability and activity lead to discovery of CoOOH|Pt, NiOOH|Ag, and FeO2|Ag as viable systems for alkaline ORR, we then use a multivariant linear regression method to identify elementary descriptors for efficient screening of promising cost-effective nanoisland|support catalysts.

On-Demand Final State Control of a Surface-Bound Bistable Single Molecule Switch.

Modern electronic devices perform their defined action because of the complete reliability of their individual active components (transistors, switches, diodes, and so forth). For instance, to encode basic computer units (bits) an electrical switch can be used. The reliability of the switch ensures that the desired outcome (the component's final state, 0 or 1) can be selected with certainty. No practical data storage device would otherwise exist. This reliability criterion will necessarily need to hold true for future molecular electronics to have the opportunity to emerge as a viable miniaturization alternative to our current silicon-based technology. Molecular electronics target the use of single-molecules to perform the actions of individual electronic components. On-demand final state control over a bistable unimolecular component has therefore been one of the main challenges in the past decade (1-5) but has yet to be achieved. In this Letter, we demonstrate how control of the final state of a surface-supported bistable single molecule switch can be realized. On the basis of the observations and deductions presented here, we further suggest an alternative strategy to achieve final state control in unimolecular bistable switches.