Limitations of X-ray reflectometry in the presence of surface contamination

TL;DR

Surface contamination significantly impacts the accuracy of X-ray reflectometry measurements, and cleaning protocols are essential regardless of data source quality, as contamination effects can dominate uncertainty estimates.

Contribution

This study quantifies how surface contamination affects XRR thickness measurements and compares the impact of synchrotron versus laboratory sources, highlighting the importance of cleaning procedures.

Findings

Contamination can dominate uncertainty in XRR measurements.

Synchrotron sources do not significantly outperform laboratory sources in contaminated conditions.

Cleaning protocols are crucial for accurate XRR analysis.

Abstract

Intentionally deposited thin films exposed to atmosphere often develop unintentionally deposited few monolayer films of surface contamination. This contamination arises from the diverse population of volatile organics and inorganics in the atmosphere. Such surface contamination can affect the uncertainties in determination of thickness, roughness and density of thin film structures by X-Ray Reflectometry (XRR). Here we study the effect of a 0.5 nm carbon surface contamination layer on thickness determination for a 20 nm titanium nitride thin film on silicon. Uncertainties calculated using Markov-Chain Monte Carlo Bayesian statistical methods from simulated data of clean and contaminated TiN thin films are compared at varying degrees of data quality to study (1) whether synchrotron sources cope better with contamination than laboratory sources and (2) whether cleaning off the surface of thin films prior to XRR measurement is necessary. We show that, surprisingly, contributions to uncertainty from surface contamination can dominate uncertainty estimates, leading to minimal advantages in using synchrotron- over laboratory-intensity data. Further, even prior knowledge of the exact nature of the surface contamination does not significantly reduce the contamination's contribution to the uncertainty in the TiN layer thickness. We conclude, then, that effective and standardized cleaning protocols are necessary to achieve high levels of accuracy in XRR measurement.