In Thrust 3, we are focusing on the stability of the catalyst and how to prevent corrosion. We want to understand how the catalyst material changes during the reaction and the formation of chemical bonds. We will explore strategies like applying protective layers and controlling the timing of surface activation. By studying the behavior of the catalyst over multiple reaction cycles, we hope to uncover the connection between surface reconstruction and corrosion.

Thrust 2 is all about figuring out how the speed of the reaction is influenced by the surface properties and the environment of the catalyst. We will manipulate the surroundings of the catalyst and use different methods to study the chemical reactions happening on the surface. Our goal is to understand how electronic and chemical states can be adjusted to speed up the reaction and improve catalysis.

Atomic Layer Deposition (ALD) is sequential and conformal process that provides precise thickness control due to its self-limiting nature even in high aspect ratio structures. The exponential growth in the use of ALD is due the availability of hundreds of precursor chemistries for diverse technological applications beyond just semiconductors and microelectronics. However, one of recurring challenges is the problem of results disparity, and errors that underline the causes of disparity, primarily may be due to lack of consistency in process optimization and reproducibility. In this research, a Design of Experiment (DoE) approach is adopted for parametric optimization for synthesis of Al2O3 thin films using ALD. The objective of this work is to adopt DoE for reproducible ALD synthesis processes within statistical reliability. Briefly, the precursors used for Al2O3 ALD (Ultratech Veeco Thermal ALD) were Trimethylaluminum (TMA) and H2O with Ozone. The effect of four parameters, namely pulse duration of TMA and water, number of cycles, and temperature were evaluated for the response parameter of Al2O3 thickness on Si with 300 nm of SiO2. E-Chip DoE software was chosen for data modeling. E-Chip considers multivariate analysis to provide the parametric relationship for Analysis of Variances (ANOVA) and predictive surface response analysis, thus eliminating the need for extensive process optimization experiments that are required for full factorial design. Initially, screening experiments were conducted to determine the levels of input parameters. The pulse duration for TMA and H2O were 0.01-0.02 sec, the number of cycles were varied between 50 and 100, the temperature range was 100-200 0C. For each sample, five thickness and roughness readings were measured on an ex-situ ellipsometer and Atomic Force Microscope (AFM). The stoichiometry of Al2O3 thin films were evaluated using XPS. The data analysis reveals that the influential parameters are number of cycles, pulse duration, and temperature in the order of their significance. The direct relationship between number of cycles and Al2O3 thickness infers that flux available is effectively used for layer growth. The consistency of layer thickness is related not just to the conformality but also to the adsorption of reactive species effectively. Uneven pulse duration also contributes to the linear growth of layer thickness, which supports the effectiveness of reactive species saturation and purging of un-reactive elements. The temperature variation was also an influential parameter as it affected the vapor pressure for TMA and H2O. This mechanistic model based on DoE is a suitable tool to predict the ALD’s growth per cycle, which can further help in understanding of any chemisorption with substrate treatment or variation in film roughness characteristics. Study of other DoE parameters such as type of substrate, substrate functionalization, reactor size in relation to the size of substrate, and purge duration are ongoing. Figure 1 describes the ANOVA chart estimates and surface response graphs, which identifies the bounded parametric region along with the ability to predict the influence of change in parameters based on developed mechanistic model. Figure 1