Quantitative modeling of \textit{in situ} x-ray reflectivity during organic molecule thin film growth

TL;DR

This study compares three models for interpreting in situ x-ray reflectivity during organic thin film growth, highlighting their strengths and limitations in capturing surface morphology evolution, and establishing a methodology for model improvement.

Contribution

It provides a detailed comparison of existing models against experimental data and AFM measurements, revealing key insights into their accuracy and the importance of fitting local extrema.

Findings

Good early-time agreement between models and actual surface morphology.

Model fit quality does not always correlate with morphology accuracy.

Reproducing all local extrema in data is critical for accurate modeling.

Abstract

Synchrotron-based x-ray reflectivity is increasingly employed as an \textit{in situ} probe of surface morphology during thin film growth, but complete interpretation of the results requires modeling the growth process. Many models have been developed and employed for this purpose, yet no detailed, comparative studies of their scope and accuracy exists in the literature. Using experimental data obtained from hyperthermal deposition of pentane and diindenoperylene (DIP) on SiO$_2$, we compare and contrast three such models, both with each other and with detailed characterization of the surface morphology using ex-situ atomic force microscopy (AFM). These two systems each exhibit particular phenomena of broader interest: pentacene/SiO$_2$ exhibits a rapid transition from rough to smooth growth. DIP/SiO$_2$, under the conditions employed here, exhibits growth rate acceleration due to a different sticking probability between the substrate and film. In general, \textit{independent of which model is used}, we find good agreement between the surface morphology obtained from fits to the \insitu x-ray data with the actual morphology at early times. This agreement deteriorates at later time, once the root-mean squared (rms) film roughness exceeds about 1 ML. A second observation is that, because layer coverages are under-determined by the evolution of a single point on the reflectivity curve, we find that the best fits to reflectivity data --- corresponding to the lowest values of $\chi_\nu^2$ --- do not necessarily yield the best agreement between simulated and measured surface morphologies. Instead, it appears critical that the model reproduce all local extrema in the data. In addition to showing that layer morphologies can be extracted from a minimal set of data, the methodology established here provides a basis for improving models of multilayer growth by comparison to real systems.