The challenge of controlling the temperature uniformity of wafer processing
The deposition reactor design engineer knows very well that the gradient heat distribution inside the reactor will not be conducive to the precise control of the uniformity of temperature distribution required for wafer processing. Generally, the simplified approach is to optimize certain components within the system to achieve the purpose of improving the overall performance of the system. Simulation and analysis of the system through the optimization of individual components and the computer fluid dynamics model is a common method to achieve precise control of the wafer temperature uniformity. That is to say, experienced analysts can make the control of certain components such as heating plates more precise through computer modeling software such as Ansys (Finite Element Analysis) and CFD (Computer Fluid Dynamics Analysis) based on assumptions.
However, even if the accuracy of these individual components is well controlled, it is actually difficult to achieve the same accuracy during wafer processing, which makes the system unable to achieve the expected performance, and the process yield and repeatability are unsatisfactory. In order to narrow the gap between individual components and system performance, it is necessary to replace standard assumptions with actual conditions and information of the reactor when performing model analysis. Among them, the key information includes various factors that can affect the reactor system, including heaters, wafers, reactor side walls, support systems, pedestals, reactor inlet and outlet ports, detection instruments and methods, and changes in the external physical and thermal environment of the reactor , Fluid type and flow rate, etc. This information allows analysts to more accurately assess the impact of various factors, including surface finish, edge loss, free convection or radiation (depending on whether the reactor is a vacuum system). In the finally designed heating and control platform, the components can cooperate with each other to achieve the best overall performance, rather than the best performance of a single component.