Data-driven strategy for size-controllable growth of tin(II) sulfide nanoplates via machine learning

Precise size control during nanomaterial growth remains challenging yet crucial for optimizing their application-specific properties. This study presents a machine learning (ML) assisted framework that enables predictive, size-controllable chemical vapor deposition (CVD) growth of SnS nanoplates. A Gaussian process regression (GPR) model is trained on experimental SnS growth data, with hyperparameters fine-tuned using the Bayesian optimization algorithm (BOA) with 10-fold cross-validation. The trained GPR model exhibits high predictive accuracy on test data, outperforming alternative ML models. Experimental validation with previously unexplored parameter sets results in relative errors below 8.3 % between model predictions and experimental measurements, confirming the model’s reliability. Further application of BOA to the trained GPR model successfully identifies the maximum achievable lateral size of SnS nanoplates and its corresponding optimal growth parameters. Sensitivity analysis identifies ‘Dist’—an indicator of growth temperature—as the most influential parameter, and a classical CVD growth model is used to explain the predicted lateral size trend across different Dist values. This data-driven strategy facilitates the prediction and optimization of CVD-grown SnS nanoplate size while significantly reducing the required experimental iterations, offering a scalable and generalizable methodology for intelligent nanomaterial design and process control. • ML model predicts SnS nanoplate size based on experimental CVD growth data. • GPR model enables accurate prediction across diverse growth conditions. • BOA identifies growth parameters for maximum lateral size. • Growth temperature was revealed as the key parameter, with predicted trends supported by classical CVD theory. • Framework reduces experimental workload and enables size-controllable synthesis.