Advanced two-objective optimization of thickness and large-area homogeneity of ZnO ultrathin films deposited by atomic layer deposition

Semiconductor thin films and coatings have become one of the most relevant research fields due to their significant applications in priority energy-related technologies such as solar cells, photocatalysts, and smart windows. Since all these fields are conceived as tools to fight against the effects of climate change, a real impact requires the successful deposition of semiconductor films on large-area substrates such as windows, panels, pipes, and containers, to give rise to photoactive components suitable for buildings, industries, cars, and parks. However, scalability remains one of the major issues in almost all methodologies known for the deposition of semiconductor films, irrespective of the phase approach used, i.e., either from vapor- or liquid-phase. Here, a mathematical metamodel was applied to simulate the atomic layer deposition (ALD) of zinc oxide (ZnO) ultrathin films (a versatile photoactive material in energy-related research) and optimized their thickness and homogeneity over the whole area of 8 in.-diameter Si wafers. Knowing all ALD parameters that define the quality and properties of the deposited films, we delimitated a set of four metamodel-inputs (zinc precursor dose, purge, and the inner and outer carrier gas flows) based on literature review, expertise, costs, and reactor design aspects specific to the deposition of ZnO. The average thickness and homogeneity of the films were established as the two outputs of the metamodel, which were the object of optimization. Using advanced iterative procedures, we carried out three rounds of experiments that lead us to a set of ALD parameters to deposit a ZnO ultrathin film with an average thickness of 11.38 nm that leads to a deposition rate of 1.9 Å/cycle, which represents 90% of the highest reported value for ZnO by ALD (2.1 Å/cycle). The homogeneity over the whole 8 in.-diameter wafer reached 2.61 nm, which represents the smoothest distribution of thickness values in the entire deposited area. Given the origin of the limits constraining this optimization procedure, our results hold promise in supporting the transition from the laboratory-level synthesis of thin-film-based optoelectronic devices to their large-scale production. This could ultimately help to circumvent the difficulties faced in scaling the ALD technology and enable alternative deposition methodologies such as thermal ALD, otherwise inaccessible to the production chain.