Hyperspectral Fingerprints for Atomic Layer Mapping of Two-Dimensional Materials with Single-Layer Accuracy

Abstract

Hyperspectral imaging microscopy with multivariate analysis is an efficient technique for rapid and accurate atomic layer mapping of two-dimensional materials, of which the thickness distributes randomly over a large area. In this work, the identification accuracy of a dual-illumination hyperspectral imaging microscope was systematically studied for mapping mono-, bi-, tri-, and few-layer MoS2 flakes fabricated by the chemical vapor deposition method. Hyperspectral fingerprints of MoS2 flakes were extracted and implemented to identify distinct flakes of new samples for cross-validation and generalizability analysis with single-layer accuracy. Mechanically exfoliated WSe2 flakes on the SiO2/Si substrate with small sizes were identified by multiline laser illumination. To reduce the computational consumption when processing hyperspectral data sets with high dimensions, the influence of the number of hyperspectral channels on the identification performance was investigated using hyperspectral fingerprints of mono- and bilayer flakes with high spectral similarity.