Accelerated discovery of CO2 electrocatalysts using active machine learning.

The rapid increase in global energy demand and the need to replace carbon dioxide (CO2)-emitting fossil fuels with renewable sources have driven interest in chemical storage of intermittent solar and wind energy1,2. Particularly attractive is the electrochemical reduction of CO2 to chemical feedstocks, which uses both CO2 and renewable energy3-8. Copper has been the predominant electrocatalyst for this reaction when aiming for more valuable multi-carbon products9-16, and process improvements have been particularly notable when targeting ethylene. However, the energy efficiency and productivity (current density) achieved so far still fall below the values required to produce ethylene at cost-competitive prices. Here we describe Cu-Al electrocatalysts, identified using density functional theory calculations in combination with active machine learning, that efficiently reduce CO2 to ethylene with the highest Faradaic efficiency reported so far. This Faradaic efficiency of over 80 per cent (compared to about 66 per cent for pure Cu) is achieved at a current density of 400 milliamperes per square centimetre (at 1.5 volts versus a reversible hydrogen electrode) and a cathodic-side (half-cell) ethylene power conversion efficiency of 55 ± 2 per cent at 150 milliamperes per square centimetre. We perform computational studies that suggest that the Cu-Al alloys provide multiple sites and surface orientations with near-optimal CO binding for both efficient and selective CO2 reduction17. Furthermore, in situ X-ray absorption measurements reveal that Cu and Al enable a favourable Cu coordination environment that enhances C-C dimerization. These findings illustrate the value of computation and machine learning in guiding the experimental exploration of multi-metallic systems that go beyond the limitations of conventional single-metal electrocatalysts.

Stabilizing Highly Active Ru Sites by Suppressing Lattice Oxygen Participation in Acidic Water Oxidation.

In hydrogen production, the anodic oxygen evolution reaction (OER) limits the energy conversion efficiency and also impacts stability in proton-exchange membrane water electrolyzers. Widely used Ir-based catalysts suffer from insufficient activity, while more active Ru-based catalysts tend to dissolve under OER conditions. This has been associated with the participation of lattice oxygen (lattice oxygen oxidation mechanism (LOM)), which may lead to the collapse of the crystal structure and accelerate the leaching of active Ru species, leading to low operating stability. Here we develop Sr-Ru-Ir ternary oxide electrocatalysts that achieve high OER activity and stability in acidic electrolyte. The catalysts achieve an overpotential of 190 mV at 10 mA cm-2 and the overpotential remains below 225 mV following 1,500 h of operation. X-ray absorption spectroscopy and 18O isotope-labeled online mass spectroscopy studies reveal that the participation of lattice oxygen during OER was suppressed by interactions in the Ru-O-Ir local structure, offering a picture of how stability was improved. The electronic structure of active Ru sites was modulated by Sr and Ir, optimizing the binding energetics of OER oxo-intermediates.

GEOMETRIC SCATTERING ATTENTION NETWORKS.

Geometric scattering has recently gained recognition in graph representation learning, and recent work has shown that integrating scattering features in graph convolution networks (GCNs) can alleviate the typical oversmoothing of features in node representation learning. However, scattering often relies on handcrafted design, requiring careful selection of frequency bands via a cascade of wavelet transforms, as well as an effective weight sharing scheme to combine low- and band-pass information. Here, we introduce a new attention-based architecture to produce adaptive task-driven node representations by implicitly learning node-wise weights for combining multiple scattering and GCN channels in the network. We show the resulting geometric scattering attention network (GSAN) outperforms previous networks in semi-supervised node classification, while also enabling a spectral study of extracted information by examining node-wise attention weights.

An antibonding valence band maximum enables defect-tolerant and stable GeSe photovoltaics.

In lead-halide perovskites, antibonding states at the valence band maximum (VBM)-the result of Pb 6s-I 5p coupling-enable defect-tolerant properties; however, questions surrounding stability, and a reliance on lead, remain challenges for perovskite solar cells. Here, we report that binary GeSe has a perovskite-like antibonding VBM arising from Ge 4s-Se 4p coupling; and that it exhibits similarly shallow bulk defects combined with high stability. We find that the deep defect density in bulk GeSe is ~1012 cm-3. We devise therefore a surface passivation strategy, and find that the resulting GeSe solar cells achieve a certified power conversion efficiency of 5.2%, 3.7 times higher than the best previously-reported GeSe photovoltaics. Unencapsulated devices show no efficiency loss after 12 months of storage in ambient conditions; 1100 hours under maximum power point tracking; a total ultraviolet irradiation dosage of 15 kWh m-2; and 60 thermal cycles from -40 to 85 °C.

Scattering GCN: Overcoming Oversmoothness in Graph Convolutional Networks.

Graph convolutional networks (GCNs) have shown promising results in processing graph data by extracting structure-aware features. This gave rise to extensive work in geometric deep learning, focusing on designing network architectures that ensure neuron activations conform to regularity patterns within the input graph. However, in most cases the graph structure is only accounted for by considering the similarity of activations between adjacent nodes, which limits the capabilities of such methods to discriminate between nodes in a graph. Here, we propose to augment conventional GCNs with geometric scattering transforms and residual convolutions. The former enables band-pass filtering of graph signals, thus alleviating the so-called oversmoothing often encountered in GCNs, while the latter is introduced to clear the resulting features of high-frequency noise. We establish the advantages of the presented Scattering GCN with both theoretical results establishing the complementary benefits of scattering and GCN features, as well as experimental results showing the benefits of our method compared to leading graph neural networks for semi-supervised node classification, including the recently proposed GAT network that typically alleviates oversmoothing using graph attention mechanisms.

Electro-Optic Modulation in Hybrid Metal Halide Perovskites.

Rapid and efficient conversion of electrical signals to optical signals is needed in telecommunications and data network interconnection. The linear electro-optic (EO) effect in noncentrosymmetric materials offers a pathway to such conversion. Conventional inorganic EO materials make on-chip integration challenging, while organic nonlinear molecules suffer from thermodynamic molecular disordering that decreases the EO coefficient of the material. It has been posited that hybrid metal halide perovskites could potentially combine the advantages of inorganic materials (stable crystal orientation) with those of organic materials (solution processing). Here, layered metal halide perovskites are reported and investigated for in-plane birefringence and linear electro-optic response. Phenylmethylammonium lead chloride (PMA2 PbCl4 ) crystals are grown that exhibit a noncentrosymmetric space group. Birefringence measurements and Raman spectroscopy confirm optical and structural anisotropy in the material. By applying an electric field on the crystal surface, the linear EO effect in PMA2 PbCl4 is reported and its EO coefficient is determined to be 1.40 pm V-1 . This is the first demonstration of this effect in hybrid metal halide perovskites, materials that feature both highly ordered crystalline structures and solution processability. The in-plane birefringence and electro-optic response reveal that layered perovskite crystals could be further explored for potential applications in polarizing optics and EO modulation.