High-speed x-ray reflectometry for characterization of thin film growth at high deposition rates

We present a high-throughput, real-time study of PTCDI-C8thin film growth using quick x-ray reflectivity (qXRR) with 12 ms time resolution, combined with machine learning-based analysis. In situ qXRR enables monitoring of vacuum deposition at growth rates from 1 to 30Å/s, accessing a previously unexplored regime in molecular beam deposition. To efficiently analyze the resulting ∼20000reflectivity curves, we employ a convolutional neural network trained on a physics-informed multilayer model. This approach robustly extracts key structural parameters—including thickness, roughness, and crystalline versus amorphous content—from noisy data. We quantify interface roughness as a function of both film thickness and growth rate, identifying rapid roughening with a scaling exponent of 𝛽=0.62with film thickness and a secondary scaling exponent of 𝛾=0.21with growth rate. Additionally, we observe a reduction in the coherently ordered film thickness and a rise in amorphous content at higher deposition rates. These results demonstrate that combining qXRR with machine learning provides quantitative access to fast kinetic growth processes, offering a powerful tool for in situ characterization and morphological control in organic thin film fabrication.