Prediction and control of thermo-fluid-chemical deposition of large-scale thin films for solar cells

Abstract

A numerical method to predict and control physical and chemical phenomena during a selenization process for large-scale CIGS thin films was developed. The selenization process is a deposition process of the CIGS layer. The CIGS, which is a chemical compound of Cu, In, Ga, and Se, has a high light absorption coefficient, while has an electro-optically stable structure. The CIGS solar cells are expected to be the next-generation solar cells with the above-mentioned advantages. During the selenization process, the CIGS layer should be uniformly deposited on a substrate to get a high-conversion efficiency solar cell. However, resulting from the turbulent convective heat transfer that occurs as a reactor size grows; thus, the uniformity and efficiency of the cells decrease. Resulting from the toxicity of working gas (H$_2$Se), to figure out the physical and chemical phenomena experimentally is difficult. Therefore, the present study investigates the phenomena during the selenization process by using a numerical method.

A fully implicit second-order large-eddy simulation (LES) method is used to simulate physical phenomena in a fluid domain. The LES method predicts the turbulent convective heat transfer and transport of energy and chemical species. A second-order finite element method (FEM) is used to predict the thermal conduction and radiation of substrates and a thermal reactor. For chemical phenomena, a chemical reaction model was developed to calculate the chemical reactions in gas and on the substrate surface. Also, a film growth model determines a growth rate of CIGS thin films. Then, an efficiency prediction model calculates a conversion efficiency based on a thin-film thickness and substrate's chemical composition. Each solver exchanges values of variables. The LES method and FEM communicate with each other to calculate the temperature at the interface between the domains. In the fluid solver, scalar transport equations are also solved to predict the transport of energy and gaseous species. Temperature and mass fractions in the control volumes are sent to the chemical reaction model. The model calculates each species' molar production rate and sends this information to the species transport equations. The reaction model also predicts the surface adsorption of reactant gas and the generation of volatile compounds during the surface reaction. As a selenization continues, the composition of chemical compounds on the substrates is changed; thus, the film growth model calculates the film thickness by using the molar production rates of surface species. Then, after the selenization process is finished, the thin-film thickness and chemical composition on the substrates is transferred to the efficiency prediction model to predict the cell performance.

The developed numerical method was validated. First, a numerical study for natural convection was conducted. When a vertical wall's temperature is higher than the atmosphere, natural convection develops along the vertical wall. Time-averaged velocity, temperature distribution, and local Nusselt numbers along the wall-normal direction were compared with experimental results and were in good agreement with the experiments. Next, thermal conduction and radiation were verified. For a side-vented open cavity, which has two vertical walls with a horizontal wall, the wall surface's emissivity and aspect ratio of the bottom to the left walls were changed. The temperature distribution along the right wall fitted well with experimental results. Next, physical phenomena during a selenization process were validated. Temperature distribution on three substrates was in good agreement with an experimental result. Then, a numerical study for a chemical vapor deposition process of a gallium arsenide (GaAs) thin-film growth was conducted for validating the developed chemical reaction model. The present result was compared with experimental and numerical results regarding the growth rate of the GaAs film. The present result was in good agreement with the previous results. Lastly, the efficiency prediction model was validated. The present results were compared with experimental and numerical results regarding a short-circuit current density, open-circuit voltage, fill factor, and conversion efficiency. The efficiency prediction model showed good agreement with the experimental and numerical studies.

In the present study, a numerical method, which integrated an LES-FEM-chemistry model for predicting a selenization process, was developed. Results of the present numerical study showed that stain patterns and non-uniform chemical composition on substrates were closely related to turbulent thermal fluid motions in a reactor. Therefore, the reactor and process design which generates uniform flow on the substrate and minimizes the effect of the turbulent convective heat transfer on the substrate's side are necessary. Also, the developed efficiency prediction model showed meaningful results. Though buffer and window layers are assumed to be uniformly applied, the results agreed well with experimental results. It indicates that the selenization process has the most significant influence on determining the solar cell performance. Lastly, the developed numerical method can apply to other processes in which bulk fluid motion and chemical reaction occur simultaneously.