Currently, reactor temperature is mostly controlled by a thermocouple or a back-side pyrometer. Therefore, the true wafer temperature on the wafer top side might change during the run and go unnoticed by the temperature control loop.
Figure 1 gives a typical example of these effects. It shows reflectance and temperature measurements during a typical GaN growth on sapphire. At around 10000s, the reactor pressure is reduced, leading to a drop in wafer temperature (red curve). This remains unnoticed by the back-side light pipe providing the control temperature for the recipe (black curve).
The drop in temperature is accompanied by a change in curvature (blue curve, lower graph). While the process temperature is constant during the run, the measured wafer temperature is changing significantly due to changed process conditions (in this case, reduced pressure).
Furthermore, run-to-run wafer temperature variations can occur due to different pre-coating conditions of quartz parts in the reactor, or different satellite rotation speeds, which would remain unnoticed without true temperature measurement, and could be fatal for the final LED devices.
Read the complete article at Solid State Technology magazine's web site.