Browsing by Author "Dr. F. Yates Sorrell, Chair"

Now showing 1 - 1 of 1

Real-Time Control of Polysilicon Deposition in Single-Wafer Rapid Thermal Chemical Vapor Deposition Furnaces
(1998-10-26) Smith, Robert Merrill; Dr. F. Yates Sorrell, Chair; Dr. Dennis E. Garoutte, Member; Dr. Kevin M. Lyons, Member; Dr. Richard R. Johnson, Member
This thesis describes the development of a real-time control system for depositing polysilicon films on silicon wafers by means of rapid thermal chemical vapor deposition. Results are presented which characterize the ability of the control system to deposit films of an average desired thickness and predict the film's spatial thickness distribution. A rapid thermal chemical vapor deposition system was used to individually process wafers. During processing, a mass spectrometer monitored the chemical species present in the exhaust gases to determine the total volume of material deposited. Simultaneously, optical probes resolved the spatial temperature distribution of the wafer. The mass spectrometry and optical temperature data were combined with an Arrhenius equation to model the deposition process. Validation of the model was exsitu. After processing, film thickness measurements were made on each wafer and compared to the computer model's predictions.Experimental results identified hydrogen, a by-product of the deposition reaction, as the metric for determining the total volume of polysilicon deposited. Process recipe control (today's standard control technique) produced films varying over a range of 280 Å when repeatedly employed to deposit film's of 900 Å. Application of the real-time control system produced films varying a maximum of 74 Å when attempting to deposit films of average thickness ranging from 800 to 1200 Å. Modeling results predicted the thickness of the deposited film to within 20 Å at the center of the wafer. Predictions at the wafers edge were off by a maximum of 160 Å. From the experience gained during this project, the following two recommendations are made to guide future efforts. First, the mass spectrometer's reaction time to an event occurring in the furnace was found to be one second. Employing an optical sensor could improve control by reducing the time lag of the system. Second, designing the furnace with the necessary optical access so that the sensors can be located outside the vacuum system would greatly facilitate the accuracy and reliability of the system. This would eliminate exposure of the sensors to the high temperatures and corrosive gases present inside the furnace which can adversely affect their performance.