The use of machine learning in computational molecular design has great potential to accelerate the discovery of innovative materials. However, its practical benefits still remain unproven in real-world applications, particularly in polymer science. We demonstrate the successful discovery of new polymers with high thermal conductivity, inspired by machine-learning-assisted polymer chemistry. This discovery was made by the interplay between machine intelligence trained on a substantially limited amount of polymeric properties data, expertise from laboratory synthesis and advanced technologies for thermophysical property measurements. Using a molecular design algorithm trained to recognize quantitative structure-property relationships with respect to thermal conductivity and other targeted polymeric properties, we identified thousands of promising hypothetical polymers. From these candidates, three were selected for monomer synthesis and polymerization because of their synthetic accessibility and their potential for ease of processing in further applications. The synthesized polymers reached thermal conductivities of 0.18-0.41 W/mK, which are comparable to those of state-of-the-art polymers in non-composite thermoplastics .

A layered oxychloride BiNbOCl is a visible-light responsive catalyst for water splitting, with its remarkable stability ascribed to the highly dispersive O-2p orbitals in the valence band, the origin of which, however, remains unclear. Here, we systematically investigate four series of layered bismuth oxyhalides, BiOX (X = Cl, Br, I), BiNbOX (X = Cl, Br), BiGdOX (X = Cl, Br), and SrBiOX (X = Cl, Br, I), and found that Madelung site potentials of anions capture essential features of the valence band structures of these materials. The oxide anion in fluorite-like blocks (e.g., [BiO] slab in BiNbOCl) is responsible for the upward shift of the valence band, and the degree of electrostatic destabilization changes depending on building layers and their stacking sequence. This study suggests that the Madelung analysis enables a prediction and design of the valence band structures of bismuth and other layered oxyhalides and is applicable even to a compound where DFT calculation is difficult to perform.

Mixed anion compounds such as oxynitrides and oxychalcogenides are recognized as potential candidates of visible-light-driven photocatalysts since, as compared with oxygen 2p orbitals, p orbitals of less electronegative anion (e.g., N, S) can form a valence band that has more negative potential. In this regard, oxyfluorides appear unsuitable because of the higher electronegativity of fluorine. Here we show an exceptional case, an anion-ordered pyrochlore oxyfluoride PbTiOF that has a small band gap (ca. 2.4 eV). With suitable modification of PbTiOF by promoters such as platinum nanoparticles and a binuclear ruthenium(II) complex, PbTiOF worked as a stable photocatalyst for visible-light-driven H evolution and CO reduction. Density functional theory calculations have revealed that the unprecedented visible-light-response of PbTiOF arises from strong interaction between Pb-6s and O-2p orbitals, which is enabled by a short Pb-O bond in the pyrochlore lattice due to the fluorine substitution.

The aim of computational molecular design is the identification of promising hypothetical molecules with a predefined set of desired properties. We address the issue of accelerating the material discovery with state-of-the-art machine learning techniques. The method involves two different types of prediction; the forward and backward predictions. The objective of the forward prediction is to create a set of machine learning models on various properties of a given molecule. Inverting the trained forward models through Bayes’ law, we derive a posterior distribution for the backward prediction, which is conditioned by a desired property requirement. Exploring high-probability regions of the posterior with a sequential Monte Carlo technique, molecules that exhibit the desired properties can computationally be created. One major difficulty in the computational creation of molecules is the exclusion of the occurrence of chemically unfavorable structures. To circumvent this issue, we derive a chemical language model that acquires commonly occurring patterns of chemical fragments through natural language processing of ASCII strings of existing compounds, which follow the SMILES chemical language notation. In the backward prediction, the trained language model is used to refine chemical strings such that the properties of the resulting structures fall within the desired property region while chemically unfavorable structures are successfully removed. The present method is demonstrated through the design of small organic molecules with the property requirements on HOMO-LUMO gap and internal energy. The R package iqspr is available at the CRAN repository.Electronic supplementary materialThe online version of this article (doi:10.1007/s10822-016-0008-z) contains supplementary material, which is available to authorized users.

We have applied the diffusion Monte Carlo method, for the first time, to an organic molecular crystal (para-diiodobenzene) in order to determine the relative stability of its two well-known polymorphs. The DMC result predicts that the R phase is more stable than the β phase at zero temperature, in agreement with experiment. In comparison, we evaluated four commonly used local, semilocal, and hybrid density functionals, including an empirical correction to include the effects of dispersion. We conclude that while density functional theory may provide the most practical method for estimating the effects of electron correlation, conventional functionals which do not accurately describe noncovalent interactions may not be considered reliable for determining highly accurate energies in such systems.

We report fixed-node diffusion Monte Carlo (DMC) calculations of stacking interaction energy between two adenine(A)-thymine(T) base pairs in B-DNA (AA:TT), for which reference data are available, obtained from a complete basis set estimate of CCSD(T) (coupled-cluster with singles, doubles, and perturbative triples). We consider four sets of nodal surfaces obtained from self-consistent field calculations and examine how the different nodal surfaces affect the DMC potential energy curves of the AA:TT molecule and the resulting stacking energies. We find that the DMC potential energy curves using the different nodes look similar to each other as a whole. We also benchmark the performance of various quantum chemistry methods, including Hartree-Fock (HF) theory, second-order Møller-Plesset perturbation theory (MP2), and density functional theory (DFT). The DMC and recently developed DFT results of the stacking energy reasonably agree with the reference, while the HF, MP2, and conventional DFT methods give unsatisfactory results.

No abstract

Herein, we performed ab initio screening to identify the best doping of LiNiO2 to achieve improved cycle performance in lithium ion batteries. The interlayer interaction that dominates the c-axis contraction and overall performance was captured well by density functional theory using van der Waals exchange-correlation functionals. The screening indicated that Nb-doping is promising for improving cycle performance. To extract qualitative reasonings, we performed data analysis in a materials informatics manner to obtain a reasonable regression to reproduce the obtained results. LASSO analysis implied that the charge density between the layers in the discharged state is the dominant factor influencing cycle performance.