Low-Data Deep Quantum Chemical Learning for Accurate MP2 and Coupled-Cluster Correlations.

Accurate ab initio prediction of electronic energies is very expensive for macromolecules by explicitly solving post-Hartree-Fock equations. We here exploit the physically justified local correlation feature in a compact basis of small molecules and construct an expressive low-data deep neural network (dNN) model to obtain machine-learned electron correlation energies on par with MP2 and CCSD levels of theory for more complex molecules and different datasets that are not represented in the training set. We show that our dNN-powered model is data efficient and makes highly transferable predictions across alkanes of various lengths, organic molecules with non-covalent and biomolecular interactions, as well as water clusters of different sizes and morphologies. In particular, by training 800 (H2O)8 clusters with the local correlation descriptors, accurate MP2/cc-pVTZ correlation energies up to (H2O)128 can be predicted with a small random error within chemical accuracy from exact values, while a majority of prediction deviations are attributed to an intrinsically systematic error. Our results reveal that an extremely compact local correlation feature set, which is poor for any direct post-Hartree-Fock calculations, has however a prominent advantage in reserving important electron correlation patterns for making accurate transferable predictions across distinct molecular compositions, bond types, and geometries.

Phosphine-Triggered Structural Defects in Au44 Homologues Boost Electrocatalytic CO2 Reduction.

The systematic induction of structural defects at the atomic level is crucial to metal nanocluster research because it endows cluster-based catalysts with highly reactive centers and allows for a comprehensive investigation of viable reaction pathways. Herein, by substituting neutral phosphine ligands for surface anionic thiolate ligands, we establish that one or two Au3 triangular units can be successfully introduced into the double-stranded helical kernel of Au44 (TBBT)28 , where TBBT=4-tert-butylbenzenethiolate, resulting in the formation of two atomically precise defective Au44 nanoclusters. Along with the regular face-centered-cubic (fcc) nanocluster, the first series of mixed-ligand cluster homologues is identified, with a unified formula of Au44 (PPh3 )n (TBBT)28-2n (n=0-2). The Au44 (PPh3 )(TBBT)26 nanocluster having major structural defects at the bottom of the fcc lattice demonstrates superior electrocatalytic performance in the CO2 reduction to CO. Density functional theory calculations indicate that the active site near the defects significantly lowers the free energy for the *COOH formation, the rate-determining step in the whole catalytic process.

Phosphinous Acid-Phosphinito Tetra-Icosahedral Au52 Nanoclusters for Electrocatalytic Oxygen Reduction.

While the formation of superatomic nanoclusters by the three-dimensional assembly of icosahedral units was predicted in 1987, the synthesis and structural determination of such clusters have proven to be incredibly challenging. Herein, we employ a mixed-ligand strategy to prepare phosphinous acid-phosphinito gold nanocluster Au52(HOPPh2)8(OPPh2)4(TBBT)16 with a tetra-icosahedral kernel. Unlike expected, each icosahedral Au13 unit shares one vertex gold atom with two adjacent units, resulting in a "puckered" ring shape with a nuclearity of 48 in the kernel. The phosphinous acid-phosphinito ligand set, which consists of two phosphinous acids and one phosphinito motif, has strong intramolecular hydrogen bonds; the π-π stacking interactions between the phosphorus- and sulfur-based ligands provide additional stabilization to the kernel. Highly stable Au52(HOPPh2)8(OPPh2)4(TBBT)16 serves as an effective electrocatalyst in the oxygen reduction reaction. Density functional theory calculations suggest that the phosphinous acid-phosphinito ligands provide the most active sites in the electrochemical catalysis, with O* formation being the rate-determining step.