| Real-time monitoring and control of epitaxial growth has been recognized as an enabling technology for volume manufacturing of epitaxial-based compound semiconductor optoelectronic and electronic devices (lasers, microwave transistors etc.). Conventional approach which is based on pre-growth calibration does not allow real-time compensation for the drift of process parameters that occurs during the growth process (growth time varies from I to 10 hours). By monitoring the deposited material directly and continuously, the new approach in this thesis based on model-reference system and least square phase not only ensure higher growth reproducibility, but also eliminate calibration runs completely. To date, in situ non-invasive normal incidence laser reflectance, pyrometric interferometry and diffused reflection spectroscopy are increasing being used for in situ growth calibrations for both molecular beam epitaxy and metal-organic vapor phase deposition growths.; The physical original of non-invasive normal incidence laser reflectance, pyrometric interferometry, diffused reflectance spectroscopy has been studied. The interpretation of sample temperature from the measured diffused reflectance spectroscopy spectrum is influenced by interference effect in the heterostructure. The compensation algorithm with the aid of curvature has been investigated. And our experiments have discovered that temperature drift during MBE growth of heterostructures is caused by change of sample emissivity as well as radiative heating from the Knudsen cells. The identification of these mechanisms will lead to a more precise modeling of sensor signal and thus improving the performance of the closed-loop control system.; Using ‘signal phase’ and a model-reference system, we have achieved continuous growth rate estimation with thickness resolution of 30 nm on both Pyrometric Interferometry (PI) and Laser Reflectance (LR). This is nearly five-fold improvement over the previous technique. This new method also achieved improved accuracy (thickness error < 2%) over conventional methods (error ∼4–7%). In addition, the new method shows excellent dynamic tracking of growth rate when the growth rate is intentionally changed by as much as 18%.; Finally, the automation of MBE system and measurement synchronization during wafer rotation are also designed and investigated in this thesis. |
