Efficient Syntheses of 2D Materials from Soft Layered Composites Guided by Yield Prediction Model: Potential of Experiment‐Oriented Materials Informatics

Abstract

Exfoliation is a common route to synthesize 2D materials, such as monolayered and fewlayered nanosheets. However, it is not easy to control the breaking-down process. The repsent work shows a new exfoliation strategy for generation of nanosheets from soft layered composites in high yield. Soft layered composites of inorganic hosts and organic guests are exfoliated into nanosheets in organic dispersion media. A simple yield-prediction model is constructed on small-scale our own experimental data by sparse modeling, a data-scientific approach, combined with chemical perspective.

Size‐Distribution Control of Exfoliated Nanosheets Assisted by Machine Learning: Small‐Data‐Driven Materials Science Using Sparse Modeling

“…d) Experimental data (see Supporting Information). [ 43,44] The red and blue descriptors were used in the prediction model (Equation (1)) with the positive and negative correlations, respectively. The yellow-labelled explanatory variables were selected for the ES-LiR analysis.…”

“…Our group has proposed a potential synthetic approach to control the exfoliation. [20,[38][39][40][41][42][43][44] The soft layered composites of metaloxide host layers and interlayer organic guests are exfoliated into the surface-modified nanosheets in dispersion media (Figure 1a). The exfoliation behavior, such as the yield and size of the nanosheets, was changed by the host-guest-medium combinations.…”

“…The exfoliation behavior, such as the yield and size of the nanosheets, was changed by the host-guest-medium combinations. [20,41,43,44] In the present work, a straightforward prediction model for the lateral-size distribution was constructed by machine learning on our own small experimental data (Figure 1b,c). The exfoliated nanosheets with monodispersed and polydispersed lateral sizes were selectively obtained by the prediction on unknown host-guest-medium combinations in a limited number of experiments (Figure 1d).…”

“…[51][52][53] Our group has focused on combination of sparse modeling and chemical perspective for utilization of small data. [41,43,44,[54][55][56] In sparse modeling, the whole of high-dimensional data are explained by a limited number of strongly correlated factors, namely descriptors, on the assumption of the sparseness. [57,58] In materials science, it means that the target materials and properties can be mainly predicted by a limited number of the descriptors, such as experimental parameters and structural factors.…”

“…Therefore, combination of sparse modeling and our perspective facilitates extraction of a limited number of interpretable descriptors for construction of a straightforward prediction model. [43,44,54] In the present work, the combination approach was applied to predict the lateral-size distribution of the exfoliated nanosheets (Figure 1). The descriptors of the lateral-size distribution were extracted by combination of machine learning and chemical perspective on the small experimental data (Figure 1c).…”