Showing papers by "Fritz Haber Institute of the Max Planck Society published in 2019"

Rational catalyst and electrolyte design for CO2 electroreduction towards multicarbon products

Abstract: The CO2 electroreduction reaction (CO2RR) to fuels and feedstocks is an attractive route to close the anthropogenic carbon cycle and store renewable energy. The generation of more reduced chemicals, especially multicarbon oxygenate and hydrocarbon products (C2+) with higher energy densities, is highly desirable for industrial applications. However, selective conversion of CO2 to C2+ suffers from a high overpotential, a low reaction rate and low selectivity, and the process is extremely sensitive to the catalyst structure and electrolyte. Here we discuss strategies to achieve high C2+ selectivity through rational design of the catalyst and electrolyte. Current state-of-the-art catalysts, including Cu and Cu–bimetallic catalysts, as well as some alternative materials, are considered. The importance of taking into consideration the dynamic evolution of the catalyst structure and composition are highlighted, focusing on findings extracted from in situ and operando characterizations. Additional theoretical insight into the reaction mechanisms underlying the improved C2+ selectivity of specific catalyst geometries and compositions in synergy with a well-chosen electrolyte are also provided. The electrochemical reduction of carbon dioxide to fuels and feedstocks has received increased attention over the past few years. In this Review, Roldan Cuenya and co-workers discuss strategies to achieve high selectivity towards multicarbon products via rational catalyst and electrolyte design.

New tolerance factor to predict the stability of perovskite oxides and halides

Abstract: Predicting the stability of the perovskite structure remains a long-standing challenge for the discovery of new functional materials for many applications including photovoltaics and electrocatalysts. We developed an accurate, physically interpretable, and one-dimensional tolerance factor, τ, that correctly predicts 92% of compounds as perovskite or nonperovskite for an experimental dataset of 576 ABX 3 materials ( X = O 2− , F − , Cl − , Br − , I − ) using a novel data analytics approach based on SISSO (sure independence screening and sparsifying operator). τ is shown to generalize outside the training set for 1034 experimentally realized single and double perovskites (91% accuracy) and is applied to identify 23,314 new double perovskites ( A 2 BB′X 6 ) ranked by their probability of being stable as perovskite. This work guides experimentalists and theorists toward which perovskites are most likely to be successfully synthesized and demonstrates an approach to descriptor identification that can be extended to arbitrary applications beyond perovskite stability predictions.

The 2019 materials by design roadmap

Abstract: Advances in renewable and sustainable energy technologies critically depend on our ability to design and realize materials with optimal properties. Materials discovery and design efforts ideally involve close coupling between materials prediction, synthesis and characterization. The increased use of computational tools, the generation of materials databases, and advances in experimental methods have substantially accelerated these activities. It is therefore an opportune time to consider future prospects for materials by design approaches. The purpose of this Roadmap is to present an overview of the current state of computational materials prediction, synthesis and characterization approaches, materials design needs for various technologies, and future challenges and opportunities that must be addressed. The various perspectives cover topics on computational techniques, validation, materials databases, materials informatics, high-throughput combinatorial methods, advanced characterization approaches, and materials design issues in thermoelectrics, photovoltaics, solid state lighting, catalysts, batteries, metal alloys, complex oxides and transparent conducting materials. It is our hope that this Roadmap will guide researchers and funding agencies in identifying new prospects for materials design.

Ni Single Atom Catalysts for CO2 Activation.

Abstract: We report on the activation of CO2 on Ni single-atom catalysts. These catalysts were synthesized using a solid solution approach by controlled substitution of 1–10 atom % of Mg2+ by Ni2+ inside the MgO structure. The Ni atoms are preferentially located on the surface of the MgO and, as predicted by hybrid-functional calculations, favor low-coordinated sites. The isolated Ni atoms are active for CO2 conversion through the reverse water–gas shift (rWGS) but are unable to conduct its further hydrogenation to CH4 (or MeOH), for which Ni clusters are needed. The CO formation rates correlate linearly with the concentration of Ni on the surface evidenced by XPS and microcalorimetry. The calculations show that the substitution of Mg atoms by Ni atoms on the surface of the oxide structure reduces the strength of the CO2 binding at low-coordinated sites and also promotes H2 dissociation. Astonishingly, the single-atom catalysts stayed stable over 100 h on stream, after which no clusters or particle formation could ...

Helical cobalt borophosphates to master durable overall water-splitting

Abstract: The sustainable electrochemical production of hydrogen from water is a key element in the transition toward carbon-neutral energy systems. Application at a global scale requires the discovery of precious metal-free electrocatalysts that unify high energetic efficiency, long-term stability and economic viability. Here we report the striking properties of noble metal-free, alkali-metal cobalt borophosphates acting as robust and efficient materials for bi-functional electrocatalytic water-splitting to give hydrogen and oxygen. Alkali-metal cobalt borophosphates are porous crystalline inorganic materials with chiral DNA-like helical structures bearing two chemically distinct types of water molecules, coordinated and strands of hydrate water associated via hydrogen bonds, located in the channels, which are predestined to possess superior catalytic performance. Depending on the applied electrode potential, they can be reversibly switched between catalysis of the hydrogen and oxygen evolution reactions, both at low overpotentials. This bifunctionality provides access to technologically simple overall water-splitting systems with energetic efficiencies exceeding the 75% level (above 90% based on a higher heating value) and uncompromised long-term stability, now verified with a two-and-a-half month period.

Selective CO 2 Electroreduction to Ethylene and Multicarbon Alcohols via Electrolyte-Driven Nanostructuring.

Abstract: Production of multicarbon products (C2+ ) from CO2 electroreduction reaction (CO2 RR) is highly desirable for storing renewable energy and reducing carbon emission. The electrochemical synthesis of CO2 RR catalysts that are highly selective for C2+ products via electrolyte-driven nanostructuring is presented. Nanostructured Cu catalysts synthesized in the presence of specific anions selectively convert CO2 into ethylene and multicarbon alcohols in aqueous 0.1 m KHCO3 solution, with the iodine-modified catalyst displaying the highest Faradaic efficiency of 80 % and a partial geometric current density of ca. 31.2 mA cm-2 for C2+ products at -0.9 V vs. RHE. Operando X-ray absorption spectroscopy and quasi in situ X-ray photoelectron spectroscopy measurements revealed that the high C2+ selectivity of these nanostructured Cu catalysts can be attributed to the highly roughened surface morphology induced by the synthesis, presence of subsurface oxygen and Cu+ species, and the adsorbed halides.

Beyond Scaling Relations for the Description of Catalytic Materials

Abstract: Computational screening for new and improved catalyst materials relies on accurate and low-cost predictions of key parameters such as adsorption energies. Here, we use recently developed compressed...

Operando Insight into the Correlation between the Structure and Composition of CuZn Nanoparticles and Their Selectivity for the Electrochemical CO2 Reduction.

Abstract: Bimetallic CuZn catalysts have been recently proposed as alternatives in order to achieve selectivity control during the electrochemical reduction of CO2 (CO2RR). However, fundamental understanding...

The role of the copper oxidation state in the electrocatalytic reduction of CO2 into valuable hydrocarbons

Abstract: Redox-active copper catalysts with accurately prepared oxidation states (Cu0, Cu+, and Cu2+) and high selectivity to C2 hydrocarbon formation, from electrocatalytic cathodic reduction of CO2, were fabricated and characterized. The electrochemically prepared copper-redox electro-cathodes yield higher activity for the production of hydrocarbons at lower oxidation state. By combining advanced X-ray spectroscopy and in situ microreactors, it was possible to unambiguously reveal the variation in the complex electronic structure that the catalysts undergo at different stages (i.e., during fabrication and electrocatalytic reactions). It was found that the surface, subsurface, and bulk properties of the electrochemically prepared catalysts are dominated by the formation of copper carbonates on the surface of cupric-like oxides, which prompts catalyst deactivation by restraining effective charge transport. Furthermore, the formation of reduced or partially reduced copper catalysts yields the key dissociative proto...

Plasma-Modified Dendritic Cu Catalyst for CO2 Electroreduction.

Abstract: Efficient and active catalysts with high selectivity for hydrocarbons and other valuable chemicals during stable operation are crucial. We have taken advantage of low-pressure oxygen plasmas to modify dendritic Cu catalysts and were able to achieve enhanced selectivity toward C2 and C3 products. Utilizing operando spectroscopic methods such as X-ray absorption fine-structure spectroscopy (XAFS) and quasi in situ X-ray photoelectron spectroscopy (XPS), we observed that the initial presence of oxides in these catalysts before the reaction plays an inferior role in determining their catalytic performance as compared to the overall catalyst morphology. This is assigned to the poor stability of the Cu x O species in these materials under the conditions of electrocatalytic conversion of CO2 (CO2RR). Our findings shed light into the strong structure/chemical state-selectivity correlation in CO2RR and open venues for the rational design of more effective catalysts through plasma pretreatments.

Machine-learned multi-system surrogate models for materials prediction

Abstract: Surrogate machine-learning models are transforming computational materials science by predicting properties of materials with the accuracy of ab initio methods at a fraction of the computational cost. We demonstrate surrogate models that simultaneously interpolate energies of different materials on a dataset of 10 binary alloys (AgCu, AlFe, AlMg, AlNi, AlTi, CoNi, CuFe, CuNi, FeV, NbNi) with 10 different species and all possible fcc, bcc and hcp structures up to 8 atoms in the unit cell, 15 950 structures in total. We find that the deviation of prediction errors when increasing the number of simultaneously modeled alloys is less than 1 meV/atom. Several state-of-the-art materials representations and learning algorithms were found to qualitatively agree on the prediction errors of formation enthalpy with relative errors of <2.5% for all systems.

Oxidation of Reduced Ceria by Incorporation of Hydrogen

Abstract: The interaction of hydrogen with reduced ceria (CeO2-x ) powders and CeO2-x (111) thin films was studied using several characterization techniques including TEM, XRD, LEED, XPS, RPES, EELS, ESR, and TDS. The results clearly indicate that both in reduced ceria powders as well as in reduced single crystal ceria films hydrogen may form hydroxyls at the surface and hydride species below the surface. The formation of hydrides is clearly linked to the presence of oxygen vacancies and is accompanied by the transfer of an electron from a Ce3+ species to hydrogen, which results in the formation of Ce4+ , and thus in oxidation of ceria.

Time- and angle-resolved photoemission spectroscopy of solids in the extreme ultraviolet at 500 kHz repetition rate

Abstract: Time- and angle-resolved photoemission spectroscopy (trARPES) employing a 500 kHz extreme-ultraviolet light source operating at 21.7 eV probe photon energy is reported. Based on a high-power ytterbium laser, optical parametric chirped pulse amplification, and ultraviolet-driven high-harmonic generation, the light source produces an isolated high-harmonic with 110 meV bandwidth and a flux of more than 1011 photons/s on the sample. Combined with a state-of-the-art ARPES chamber, this table-top experiment allows high-repetition rate pump-probe experiments of electron dynamics in occupied and normally unoccupied (excited) states in the entire Brillouin zone and with a temporal system response function below 40 fs.

Simultaneous learning of several materials properties from incomplete databases with multi-task SISSO

Abstract: The identification of descriptors of materials properties and functions that capture the underlying physical mechanisms is a critical goal in data-driven materials science. Only such descriptors will enable a trustful and efficient scanning of materials spaces and possibly the discovery of new materials. Recently, the sure-independence screening and sparsifying operator (SISSO) has been introduced and was successfully applied to a number of materials-science problems. SISSO is a compressed-sensing based methodology yielding predictive models that are expressed in form of analytical formulas, built from simple physical properties. These formulas are systematically selected from an immense number (billions or more) of candidates. In this work, we describe a powerful extension of the methodology to a 'multi-task learning' approach, which identifies a single descriptor capturing multiple target materials properties at the same time. This approach is specifically suited for a heterogeneous materials database with scarce or partial data, e.g., in which not all properties are reported for all materials in the training set. As showcase examples, we address the construction of materials-properties maps for the relative stability of octet-binary compounds, considering several crystal phases simultaneously, and the metal/insulator classification of binary materials distributed over many crystal-prototypes.

Molecular force fields with gradient-domain machine learning: Construction and application to dynamics of small molecules with coupled cluster forces.

Abstract: We present the construction of molecular force fields for small molecules (less than 25 atoms) using the recently developed symmetrized gradient-domain machine learning (sGDML) approach [Chmiela et al., Nat. Commun. 9, 3887 (2018) and Chmiela et al., Sci. Adv. 3, e1603015 (2017)]. This approach is able to accurately reconstruct complex high-dimensional potential-energy surfaces from just a few 100s of molecular conformations extracted from ab initio molecular dynamics trajectories. The data efficiency of the sGDML approach implies that atomic forces for these conformations can be computed with high-level wavefunction-based approaches, such as the “gold standard” coupled-cluster theory with single, double and perturbative triple excitations [CCSD(T)]. We demonstrate that the flexible nature of the sGDML model recovers local and non-local electronic interactions (e.g., H-bonding, proton transfer, lone pairs, changes in hybridization states, steric repulsion, and n → π* interactions) without imposing any restriction on the nature of interatomic potentials. The analysis of sGDML molecular dynamics trajectories yields new qualitative insights into dynamics and spectroscopy of small molecules close to spectroscopic accuracy.

Shape-Controlled Nanoparticles as Anodic Catalysts in Low-Temperature Fuel Cells.

Abstract: The great dependence of the electrocatalytic activity of most electrochemical reactions on the catalytic surface area and specific surface structure is widely accepted. Building on the extensive knowledge already available on single-crystal surfaces, this Perspective discusses the recent progress made in low-temperature fuel cells through the use of the most active shape-controlled noble metal-based nanoparticles. In particular, we will focus on discussing structure-composition-reactivity correlations in methanol, ethanol, and formic acid oxidation reactions and will offer a general vision of future needs.

The Oxidation of Platinum under Wet Conditions Observed by Electrochemical X-ray Photoelectron Spectroscopy.

Abstract: During the electrochemical reduction of oxygen, platinum catalysts are often (partially) oxidized. While these platinum oxides are thought to play a crucial role in fuel cell degradation, their nature remains unclear. Here, we studied the electrochemical oxidation of Pt nanoparticles using in situ XPS. When the particles were sandwiched between a graphene sheet and a proton exchange membrane that is wetted from the back, a confined electrolyte layer was formed, allowing us to probe the electrocatalyst under wet conditions. We show that the surface oxide formed at the onset of Pt oxidation has a mixed Pt delta+/Pt2+/Pt4+ composition. The formation of this surface oxide is suppressed when a Br-containing membrane is chosen due to adsorption of Br on Pt. Time-resolved measurements show that oxidation is fast for nanoparticles: even bulk PtO2 center dot nH(2)O growth occurs on the subminute time scale. The fast formation of Pt4+ species in both surface and bulk oxide form suggests that Pt4+-oxides are likely formed (or reduced) even in the transient processes that dominate Pt electrode degradation.

Enhanced Stability and CO/Formate Selectivity of Plasma-Treated SnOx/AgOx Catalysts during CO2 Electroreduction

Abstract: CO2 electroreduction into useful chemicals and fuels is a promising technology that might be used to minimize the impact that the increasing industrial CO2 emissions are having on the environment. Although plasma-oxidized silver surfaces were found to display a considerably decreased overpotential for the production of CO, the hydrogen evolution reaction (HER), a competing reaction against CO2 reduction, was found to increase over time. More stable and C1-product-selective SnOx/AgOx catalysts were obtained by electrodepositing Sn on O2-plasma-pretreated Ag surfaces. In particular, a strong suppression of HER (below 5% Faradaic efficiency (FE) at −0.8 V vs the reversible hydrogen electrode, RHE) during 20 h was observed. Ex situ scanning electron microscopy (SEM) combined with energy-dispersive X-ray spectroscopy (EDS), quasi in situ X-ray photoelectron spectroscopy (XPS), and operando X-ray absorption near-edge structure spectroscopy (XANES) measurements showed that our synthesis led to a highly roughened...

Strong Metal-Support Interactions between Copper and Iron Oxide during the High-Temperature Water-Gas Shift Reaction.

Abstract: The commercial high-temperature water-gas shift (HT-WGS) catalyst consists of CuO-Cr2 O3 -Fe2 O3 , where Cu functions as a chemical promoter to increase the catalytic activity, but its promotion mechanism is poorly understood. In this work, a series of iron-based model catalysts were investigated with in situ or pseudo in situ characterization, steady-state WGS reaction, and density function theory (DFT) calculations. For the first time, a strong metal-support interaction (SMSI) between Cu and FeOx was directly observed. During the WGS reaction, a thin FeOx overlayer migrates onto the metallic Cu particles, creating a hybrid surface structure with Cu-FeOx interfaces. The synergistic interaction between Cu and FeOx not only stabilizes the Cu clusters, but also provides new catalytic active sites that facilitate CO adsorption, H2 O dissociation, and WGS reaction. These new fundamental insights can potentially guide the rational design of improved iron-based HT-WGS catalysts.

Iron Carbides: Control Synthesis and Catalytic Applications in COx Hydrogenation and Electrochemical HER

Abstract: Catalytic transformation of COx (x = 1, 2) with renewable H2 into valuable fuels and chemicals provides practical processes to mitigate the worldwide energy crisis. Fe-based catalytic materials are widely used for those reactions due to their abundance and low cost. Novel iron carbides are particularly promising catalytic materials among the reported ferrous catalysts. Recently, a series of strategies has been developed for the preparation of iron carbide nanoparticles and their nanocomposites. Control synthesis of FeCx -based nanomaterials and their catalytic applications in COx hydrogenation and electrochemical hydrogen evolution reaction (HER) are reviewed. The discussion is focused on the unique catalytic activities of iron carbides in COx hydrogenation and HER and the correlation between structure and catalytic performance. Future synthesis and potential catalytic applications of iron carbides are also summarized.

The ultrafast Einstein-de Haas effect.

Abstract: The Einstein-de Haas effect was originally observed in a landmark experiment1 demonstrating that the angular momentum associated with aligned electron spins in a ferromagnet can be converted to mechanical angular momentum by reversing the direction of magnetization using an external magnetic field. A related problem concerns the timescale of this angular momentum transfer. Experiments have established that intense photoexcitation in several metallic ferromagnets leads to a drop in magnetization on a timescale shorter than 100 femtoseconds—a phenomenon called ultrafast demagnetization2–4. Although the microscopic mechanism for this process has been hotly debated, the key question of where the angular momentum goes on these femtosecond timescales remains unanswered. Here we use femtosecond time-resolved X-ray diffraction to show that most of the angular momentum lost from the spin system upon laser-induced demagnetization of ferromagnetic iron is transferred to the lattice on sub-picosecond timescales, launching a transverse strain wave that propagates from the surface into the bulk. By fitting a simple model of the X-ray data to simulations and optical data, we estimate that the angular momentum transfer occurs on a timescale of 200 femtoseconds and corresponds to 80 per cent of the angular momentum that is lost from the spin system. Our results show that interaction with the lattice has an essential role in the process of ultrafast demagnetization in this system. Femtosecond time-resolved X-ray diffraction reveals that in the ultrafast demagnitization of ferromagnetic iron, about 80% of the angular momentum lost from the spins is transferred to the lattice on a sub-picosecond timescale.

Step-flow growth in homoepitaxy of β-Ga2O3 (100)—The influence of the miscut direction and faceting

Abstract: We present a systematic study on the influence of the miscut orientation on structural and electronic properties in the homoepitaxial growth on off-oriented β-Ga2O3 (100) substrates by metalorganic chemical vapour phase epitaxy. Layers grown on (100) substrates with 6° miscut toward the [001¯] direction show high electron mobilities of about 90 cm2 V−1 s−1 at electron concentrations in the range of 1–2 × 1018 cm−3, while layers grown under identical conditions but with 6° miscut toward the [001] direction exhibit low electron mobilities of around 10 cm2 V−1 s−1. By using high-resolution scanning transmission electron microscopy and atomic force microscopy, we find significant differences in the surface morphologies of the substrates after annealing and of the layers in dependence on their miscut direction. While substrates with miscuts toward [001¯] exhibit monolayer steps terminated by (2¯01) facets, mainly bilayer steps are found for miscuts toward [001]. Epitaxial growth on both substrates occurs in step-flow mode. However, while layers on substrates with a miscut toward [001¯] are free of structural defects, those on substrates with a miscut toward [001] are completely twinned with respect to the substrate and show stacking mismatch boundaries. This twinning is promoted at step edges by transformation of the (001)-B facets into (2¯01) facets. Density functional theory calculations of stoichiometric low index surfaces show that the (2¯01) facet has the lowest surface energy following the (100) surface. We conclude that facet transformation at the step edges is driven by surface energy minimization for the two kinds of crystallographically inequivalent miscut orientations in the monoclinic lattice of β-Ga2O3.

Elucidating the Nuclear Quantum Dynamics of Intramolecular Double Hydrogen Transfer in Porphycene.

Abstract: We address the double hydrogen transfer (DHT) dynamics of the porphycene molecule, a complex paradigmatic system in which the making and breaking of H-bonds in a highly anharmonic potential energy surface require a quantum mechanical treatment not only of the electrons but also of the nuclei. We combine density functional theory calculations, employing hybrid functionals and van der Waals corrections, with recently proposed and optimized path-integral ring-polymer methods for the approximation of quantum vibrational spectra and reaction rates. Our full-dimensional ring-polymer instanton simulations show that below 100 K the concerted DHT tunneling pathway dominates but between 100 and 300 K there is a competition between concerted and stepwise pathways when nuclear quantum effects are included. We obtain ground-state reaction rates of 2.19 × 1011 s-1 at 150 K and 0.63 × 1011 s-1 at 100 K, in good agreement with experiment. We also reproduce the puzzling N-H stretching band of porphycene with very good accuracy from thermostated ring-polymer molecular dynamics simulations. The position and line shape of this peak, centered at around 2600 cm-1 and spanning 750 cm-1, stem from a combination of very strong H-bonds, the coupling to low-frequency modes, and the access to cis-like isomeric conformations, which cannot be appropriately captured with classical-nuclei dynamics. These results verify the appropriateness of our general theoretical approach and provide a framework for a deeper physical understanding of hydrogen transfer dynamics in complex systems.

Dinitrogen Fixation and Reduction by Ta3N3H0,1- Cluster Anions at Room Temperature: Hydrogen-Assisted Enhancement of Reactivity

Abstract: Dinitrogen activation and reduction is one of the most challenging and important subjects in chemistry. Herein, we report the N2 binding and reduction at the well-defined Ta3N3H– and Ta3N3– gas-pha...

Partial Oxidation of Methane to Syngas Over Nickel-Based Catalysts: Influence of Support Type, Addition of Rhodium, and Preparation Method.

Abstract: There is great economic incentive in developing efficient catalysts to produce hydrogen or syngas by catalytic partial oxidation of methane (CPOM) since this is a much less energy-intensive reaction than the highly endothermic methane steam reforming reaction, which is the prominent reaction in industry. Herein, we report the catalytic behavior of nickel-based catalysts supported on different oxide substrates (Al2O3, CeO2, La2O3, MgO, and ZrO2) synthesized via wet impregnation and solid-state reaction. Furthermore, the impact of Rh doping was investigated. The catalysts have been characterized by X-ray diffraction, N2 adsorptiondesorption at -196°C, temperature-programmed reduction, X-ray photoelectron spectroscopy, O2-pulse chemisorption, transmission electron microscopy, and Raman spectroscopy. Supported Ni catalysts were found to be active for CPOM but can suffer from fast deactivation caused by the formation of carbon deposits as well as via the sintering of Ni nanoparticles (NPs). It has been found that the presence of Rh favors nickel reduction, which leads to an increase in the methane conversion and yield. For both synthesis methods, the catalysts supported on alumina and ceria show the best performance. This could be explained by the higher surface area of the Ni NPs on the alumina surface and presence of oxygen vacancies in the CeO2 lattice, which favor the proportion of oxygen adsorbed on defect sites. The catalysts supported on MgO suffer quick deactivation due to formation of a NiO/MgO solid solution, which is not reducible under the reaction conditions. The low level of carbon formation over the catalysts supported on La2O3 is ascribed to the very high dispersion of the nickel NPs and to the formation of lanthanum oxycarbonate, through which carbon deposits are gasified. The catalytic behavior for catalysts with ZrO2 as support depends on the synthesis method; however, in both cases, the catalysts undergo deactivation by carbon deposits.

Chemical diversity in molecular orbital energy predictions with kernel ridge regression.

Abstract: Instant machine learning predictions of molecular properties are desirable for materials design, but the predictive power of the methodology is mainly tested on well-known benchmark datasets. Here, we investigate the performance of machine learning with kernel ridge regression (KRR) for the prediction of molecular orbital energies on three large datasets: the standard QM9 small organic molecules set, amino acid and dipeptide conformers, and organic crystal-forming molecules extracted from the Cambridge Structural Database. We focus on the prediction of highest occupied molecular orbital (HOMO) energies, computed at the density-functional level of theory. Two different representations that encode the molecular structure are compared: the Coulomb matrix (CM) and the many-body tensor representation (MBTR). We find that KRR performance depends significantly on the chemistry of the underlying dataset and that the MBTR is superior to the CM, predicting HOMO energies with a mean absolute error as low as 0.09 eV. To demonstrate the power of our machine learning method, we apply our model to structures of 10k previously unseen molecules. We gain instant energy predictions that allow us to identify interesting molecules for future applications.

Molecular Force Fields with Gradient-Domain Machine Learning: Construction and Application to Dynamics of Small Molecules with Coupled Cluster Forces

Abstract: We present the construction of molecular force fields for small molecules (less than 25 atoms) using the recently developed symmetrized gradient-domain machine learning (sGDML) approach [Chmiela et al., Nat. Commun. 9, 3887 (2018); Sci. Adv. 3, e1603015 (2017)]. This approach is able to accurately reconstruct complex high-dimensional potential-energy surfaces from just a few 100s of molecular conformations extracted from ab initio molecular dynamics trajectories. The data efficiency of the sGDML approach implies that atomic forces for these conformations can be computed with high-level wavefunction-based approaches, such as the "gold standard" CCSD(T) method. We demonstrate that the flexible nature of the sGDML model recovers local and non-local electronic interactions (e.g. H-bonding, proton transfer, lone pairs, changes in hybridization states, steric repulsion and $n\to\pi^*$ interactions) without imposing any restriction on the nature of interatomic potentials. The analysis of sGDML molecular dynamics trajectories yields new qualitative insights into dynamics and spectroscopy of small molecules close to spectroscopic accuracy.