Datasets for AI Agent-Assisted High-Throughput Characterization of CVD-Grown MoS2

This dataset accompanies the manuscript “AI Agent-Assisted High-Throughput Characterization of CVD-Grown MoS₂”. It provides experimental source data and a trained segmentation model supporting large-area optical mapping, spatial layer-composition analysis, differential-reflectance spectroscopy, and electrical and optoelectronic characterization of CVD-grown MoS₂. The data are organized according to the manuscript figures: Figure 3 — Large-area optical mapping and overall layer composition. The files fig3_a_mosaic.png, fig3_b_mosaic.png, and fig3_c_mosaic.png contain stitched optical images of samples grown using 5, 25, and 35 mg of sulfur, respectively, with a fixed MoO₂ loading of 10 mg. The file fig3_model_260nm.pth contains the U-Net checkpoint used for semantic layer-number segmentation on 260 nm SiO₂/Si. Figure 4 — Spatially resolved layer composition. These data describe the local distributions of monolayer (1L), bilayer (2L), and thicker-layer (TL) MoS₂ derived from the layer maps in Figure 3. Each map is divided into a 6 × 11 grid. Within each cell, layer-area fractions are normalized to the total retained classified MoS₂ area, excluding substrate/background and the designated exclusion regions. These fractions describe local layer composition rather than absolute crystal coverage. Cells without retained classified MoS₂ have undefined compositions. Whole-map fractions provide comparison with the overall statistics in Figure 3. Figure 5 — Differential-reflectance spectroscopy. The archive fig5_data.zip contains spectroscopy source data supporting the comparison of monolayer and bilayer MoS₂, background subtraction, excitonic peak fitting, and peak-energy statistics. Figure 6 — Electrical characterization. The archive fig6_data.zip contains device data supporting transfer/output characteristics and field-effect mobility analysis. Figure 7 — Photoresponse characterization. The archive fig7_data.zip contains source data for wavelength- and illumination-power-dependent photoresponse analysis. Experimental conditions, processing methods, and interpretation are described in the accompanying manuscript and Supporting Information. Associated image-processing and spatial-analysis software is available in MosaicAgent-DD.