2D materials exhibit emergent properties originating from their characteristic nanostructures. In general, monolayered and few-layered nanosheets are obtained by exfoliation of the precursor layered materials. However, lateral-size distribution of the nanosheets is not easily controlled through the exfoliation because of the unpredictable down-sizing processes. The present work shows selective syntheses of transition-metal-oxide nanosheets with the monodisperse and polydisperse lateral sizes by the assistance of machine learning on small experimental data. The precursor layered composites of host transition-metal oxides and interlayer organic guests are exfoliated into the surface-modified nanosheets in organic dispersion media. A prediction model of the size distribution is constructed by sparse modeling, a method of machine learning, on the experimental data. The host-guest-medium combinations achieving the monodisperse and polydisperse lateral sizes are recommended by the prediction model. Therefore, the nanosheets with the controlled lateral-size distribution are selectively obtained in a limited number of the experiments. Moreover, self-assembly of the polydispersed nanosheets forms the homogeneous thin film exhibiting interference color. The prediction model and its construction method can be applied to the other 2D materials. Moreover, the present work implies that sparse modeling is an effective approach for small-data-driven materials science.
ASJC Scopus subject areas