Laser Chemical Vapor Deposition of Titanium Nitride and Process Diagnostics with Laser-Induced Fluorescence Spectroscopy.

TiN films have been deposited on Ti-6Al-4V substrates by a cw CO_2 laser chemical vapor deposition (LCVD) process with TiCl_4, H_2, and N_2. Pulsed dye laser induced fluorescence (LIF) spectroscopy is used to obtain transient gas phase Ti atomic concentration above the substrate. Multi-wavelength pyrometry is applied to measure the surface temperature during deposition. Film thickness profiles are obtained by stylus profilometry, and film compositions are analyzed by Auger Electron Spectroscopy (AES). Very high film growth rates are found in the order of 1 mum/sec. The dependencies of the film growth rate on partial pressure ratio N _2:H_2, TiCl _4 partial pressure, total chamber pressure, and laser power are studied, and empirical relationships between the growth rate and TiCl_4, H_2, and N_2 partial pressures are established. The results suggest that the deposition is mainly due to chemical reactions on the substrate surface that is initiated by laser heating. The time change of Ti atomic concentration above the center of the deposition area measured by LIF is found to behave in the same way as the film thickness when experimental conditions are varied. Therefore, the LIF signal can be used as an in situ diagnostic tool for process monitoring and control. Possible surface reaction pathways and the rate-controlling steps are suggested. The apparent activation energy is found to be (115.0 +/- 10.7) kJ/mol for a substrate center temperature of 1339 K to 1515 K, a total pressure of 600 Torr, a partial pressure ratio N_2:H_2 of 3:1, and a TiCl_4 partial pressure of 27 Torr. AES analyses indicate that all the TiN _{rm x} films are in the range x = 0.8 +/- 0.1, with one exception of x = 0.6 for the lowest total pressure (Total = 100 Torr, N_2:H_2 = 3:1, TiCl_4 = 27 Torr, Laser Power = 400 W). Films are obtained with Cl atomic concentration as low as <0.5%, and O and C atomic concentrations as low as <1%. Finally, suggestions are made for improved numerical modeling based on the new findings.