<b>Closing the Loop in Epitaxy with Machine Learning: Joint Optimization of Growth and Geometry in On-Chip Lasers</b>

<b>Datasets for “Closing the Loop in Epitaxy with Machine Learning: Joint Optimization of Growth and Geometry in On-Chip Lasers”</b>These datasets support multi-objective Bayesian optimization (MOBO) for variance reduction in bottom-up grown InP/InAsP multi-quantum well (MQW) microring lasers. The optimization considers field-level lasing performance, including median lasing threshold, median lasing wavelength, and lasing threshold variance. The datasets also include optical microscopy images and their corresponding binarized microring images, which were used to train a variational autoencoder (VAE) for morphology–performance correlation analysis. Together, the datasets link epitaxial growth conditions, device geometry, morphology, lasing threshold, lasing wavelength, yield, and device-to-device variability.The training data used for the multi-objective Bayesian optimization framework for variance reduction are available separately at:<br>https://doi.org/10.48420/273305281. Batch 3 Dataset with Optical and Binarized Images<b>File:</b> <code>Batch_3_with_images.pkl</code>This is the main Batch 3 dataset. It contains device-level information for the MQW microring lasers, including growth parameters, geometry parameters, lasing performance metrics, and image data for individual devices.<b>Contents include:</b><b>Growth parameters:</b>Number of quantum wellsQuantum well growth temperatureAs/P ratio estimated from precursor flow rates in the vapor phaseV/III ratio during InP barrier growthInP capping layer growth duration<b>Geometry parameters:</b>MQW microring diameterPitch, defined as the center-to-center spacing between neighboring MQW microrings<b>Performance metrics:</b>Lasing thresholdLasing wavelength<b>Image data:</b>Optical microscopy images of the MQW microringsCorresponding binarized microring images used for VAE analysisThe dataset is indexed by <code>sample_ID</code>, <code>field_ID</code>, and <code>ring_ID</code>, where each row corresponds to an individual MQW microring device.2. Field-Level Statistics Dataset<b>Files:</b><br><code>Field_statistics_Batch_3A.pkl</code><br><code>Field_statistics_Batch_3B.pkl</code>These datasets contain the field-level statistics used for Batch 3 performance analysis and variance-aware MOBO evaluation.<b>Contents include:</b>Median lasing thresholdLasing threshold varianceMedian lasing wavelengthYield3. Batch 3 Power-Dependent PL Spectroscopy Dataset<b>File:</b> <code>Batch_3_pdep.h5</code>This dataset contains the power-dependent photoluminescence (PL) spectroscopy measurements for the Batch 3 MQW microring lasers.It includes power-dependent spectra, excitation fluence/power values, device identifiers, and processed light-in–light-out curve data. These data are used to reproduce power-dependent PL spectra, extract lasing wavelengths, construct L–L curves, and determine lasing thresholds.4. Reference / Repeated-Measurement Dataset<b>File:</b> <code>ref_data.h5</code>This dataset contains reference measurements from repeated experimental runs of Batch 1. It includes ring and sample identifiers, along with lasing threshold and wavelength values measured across different runs.These data are used to assess measurement reproducibility and apply run-to-run corrections to the extracted lasing performance metrics.5. Batch 3 Measurement Settings<b>File:</b> <code>Measurement_Settings_Batch_3.h5</code>This file contains the measurement metadata associated with the Batch 3 power-dependent PL spectroscopy experiments.This file should be used together with <code>Batch_3_pdep.h5</code> to correctly interpret the spectroscopy data.