Training Dataset for Deep Learning Model in Efficient Designing Optical Switch by PCM Metasurface via Tab Transformer-based Approach

This dataset is specifically constructed to support the training of the deep learning model proposed in the paper "Efficient Designing Optical Switch by Phase Change Material Metasurface via Tab Transformer-based Approach," systematically compiling the optical response data of metasurfaces based on four typical phase change materials: GST, GSST, Sb2S3, and Sb2Se3. The dataset comprehensively covers the optical constants of the materials at various wavelengths, including the refractive indices ( n_a, n_c) and extinction coefficients ( k_a, k_c ) in both amorphous and crystalline states, as well as the transmittance (T_a, T_c) and phase (P_a, P_c) distributions of the unit cells under varying geometric parameters (width W , length L, period P, and height H). The dataset aims to provide high-quality input features and labels for the Tab Transformer-based neural network to achieve efficient forward prediction and inverse design of phase change material optical switches, thereby ensuring the model accurately captures the complex nonlinear mapping relationships between geometric structures and optical performances.