Search

CeO2 nanostructures have been utilized for various biomedical, sensor, and catalysis applications owing to their unique defect structure, enabling them to have regenerative oxidative properties. Defect engineering in CeO2 nanostructures has major importance, enabling them to be utilized for specific applications. Despite various synthesis methods, it is challenging to have precise and reversible control over defect structures. Against this backdrop, in the current work, we have explored machine learning (ML) enhanced defect engineering of CeO2 nanofilms. In our earlier work [J. Vac. Sci. Technol. A 39, 060405 (2021)], we have developed an atomic layer deposition process for CeO2 using in situ ellipsometry measurements. In the current work, data collected through in situ spectroscopic ellipsometry and ex situ XPS have been correlated using two ML algorithms (gradient boost and random forest regressor) to exert better control over the chemical properties. Defect structures are one of the desired properties in CeO2 nanomaterials, characterized by the Ce3+/Ce4+ oxidation state ratio leading to its regenerative properties. We have shown that the defect structure of the CeO2 nanofilms can be predicted using in situ ellipsometry data in real time using a trained ML algorithm using two different methods. The first method involves an indirect approach of thickness prediction using an ML algorithm (k-nearest neighbors) followed by Ce3+/Ce4+ estimation using an experimental calibration curve. The second method with a more direct approach involves Ce3+/Ce4+ prediction using real-time ellipsometry data (amplitude ratio ψ and phase difference Δ) using gradient boost and random forest regressor. [ABSTRACT FROM AUTHOR]

A machine learning approach is applied to estimate film thickness from in situ spectroscopic ellipsometry data. Using the atomic layer deposition of ZnO as a model process, the ellipsometry spectra obtained contains polarization data (Ψ, Δ) as a function of wavelength. Within this dataset, 95% is used for training the machine learning algorithm, and 5% is used for thickness prediction. Five algorithms—logistic regression, support vector machine, decision tree, random forest, and k-nearest neighbors—are tested. Out of these, the k-nearest neighbor performs the best with an average thickness prediction accuracy of 88.7% to within ±1.5 nm. The prediction accuracy is found to be a function of ZnO thickness and degrades as the thickness increases. The average prediction accuracy to within ±1.5 nm remains remarkably robust even after 90% of the (Ψ, Δ) are randomly eliminated. Finally, by considering (Ψ, Δ) in a limited spectral range (271–741 nm), prediction accuracies approaching that obtained from the analysis of full spectra (271–1688 nm) can be realized. These results highlight the ability of machine learning algorithms, specifically the k-nearest neighbor, to successfully train and predict thickness from spectroscopic ellipsometry data. [ABSTRACT FROM AUTHOR]

Process development in atomic layer deposition (ALD) is often time-consuming, requiring optimization of saturation curves and temperature windows for controlled deposition rates. Any ALD process should be self-limiting in nature, exhibiting a temperature window of nominal deposition and a linear deposition rate. Meeting these criteria usually requires several ALD experiments, followed by film characterization, which are generally time, cost, and labor-intensive. Against this backdrop, we report a methodology using in situ ellipsometry to rapidly develop the ALD process for cerium oxide using Ce(iPrCp)2(N-iPr-amd) and water. The entire optimized process was realized in ten experiments of sequential pulsing as a function of temperature, requiring less than a day. In the traditional approach, tens of experiments and ex situ characterization may be required. The approach reported here generated a contour visualization of the time-temperature-thickness parameter space delineating the optimal deposition conditions. The cerium oxide deposition rate deposited in the ALD temperature window was ∼0.15 nm/cycle; the deposited film was further characterized using x-ray photoelectron spectroscopy, x-ray diffraction, and atomic force microscopy to probe the film composition and quality further. [ABSTRACT FROM AUTHOR]