Future materials-science research will involve autonomous synthesis and characterization, requiring an approach that combines machine learning, robotics, and big data. In this paper, we highlight our recent experiments in autonomous synthesis and resistance minimization of Nb-doped TiO2 thin films. Combining Bayesian optimization with robotics, these experiments illustrate how the required speed and volume of future big-data collection in materials science will be achieved and demonstrate the tremendous potential of this combined approach. We briefly discuss the outlook and significance of these results and advances.
Electric double-layer capacitors are efficient energy storage devices that have the potential to account for uneven power demand in sustainable energy systems. Earlier attempts to improve an unsatisfactory capacitance of electric double-layer capacitors have focused on meso- or nanostructuring to increase the accessible surface area and minimize the distance between the adsorbed ions and the electrode. However, the dielectric constant of the electrolyte solvent embedded between adsorbed ions and the electrode surface, which also governs the capacitance, has not been previously exploited to manipulate the capacitance. Here we show that the capacitance of electric double-layer capacitor electrodes can be enlarged when the water molecules are strongly confined into the two-dimensional slits of titanium carbide MXene nanosheets. Using electrochemical methods and theoretical modeling, we find that dipolar polarization of strongly confined water resonantly overscreens an external electric field and enhances capacitance with a characteristically negative dielectric constant of a water molecule.
5 December 2017 in The Journal of Chemical Physics
To clarify atomic diffusion in amorphous materials, which is important in novel information and energy devices, theoretical methods having both reliability and computational speed are eagerly anticipated. In the present study, we applied neural network (NN) potentials, a recently developed machine learning technique, to the study of atom diffusion in amorphous materials, using Li3PO4 as a benchmark material. The NN potential was used together with the nudged elastic band, kinetic Monte Carlo, and molecular dynamics methods to characterize Li vacancy diffusion behavior in the amorphous Li3PO4 model. By comparing these results with corresponding DFT calculations, we found that the average error of the NN potential is 0.048 eV in calculating energy barriers of diffusion paths, and 0.041 eV in diffusion activation energy. Moreover, the diffusion coefficients obtained from molecular dynamics are always consistent with those from ab initio molecular dynamics simulation, while the computation speed of the NN potential is 3–4 orders of magnitude faster than DFT. Lastly, the structure of amorphous Li3PO4 and the ion transport properties in it were studied with the NN potential using a large supercell model containing more than 1000 atoms. The formation of P2O7 units was observed, which is consistent with the experimental characterization. The Li diffusion activation energy was estimated to be 0.55 eV, which agrees well with the experimental measurements.
Prediction of reaction yields by machine learning approach is demonstrated in tungsten-catalyzed epoxidation of alkenes. The various electronic and vibrational parameters of the phosphonic acids are collected by DFT simulation, and chosen by LASSO as the essential parameters for prediction of the reaction yields. With the trained model, we can predict yields of the reaction with unverified phosphonic acids with an error of 26%. Prediction of reaction yields by machine learning is demonstrated in tungsten-catalyzed epoxidation of alkenes. The electronic and vibrational parameters of the phosphonic acids are collected by DFT simulation, and chosen by LASSO as the essential parameters for the prediction. With the trained model, we can predict yields of the reaction with unverified phosphonic acids with an error of 26%.