Determination of Temperature and Composition of Phosphosilicate Glass Thin Films from Infrared Emission Spectral Data

Abstract

Multivariate calibration methods were combined with full-spectrum IR emission data to determine the temperature and phosphorus (P) content of phosphosilicate glass (PSG) thin films on silicon wafers. The 12 PSG calibration samples varied in phosphorus content across the range of 1 to 6 wt %, and sample temperatures ranged from 119 to 225°C. The multivariate partial least-squares (PLS) algorithm was applied to both single-beam and emittance spectral data over the range of 450 to 1600 cm⁻¹. The predictive abilities of the PLS models were obtained from the standard errors of prediction (SEP) calculated with cross-validation procedures. The cross-validated SEPs were 1.9°C for temperature determination and 0.13 wt % for P content. These SEPs were limited by the temperature controller for temperature determination and by the reference method for P determination. Full-spectrum multivariate calibration methods have been shown to be capable of precise and accurate temperature measurements in the presence of variations in sample emittances due to concentration differences. These results demonstrate the power of multivariate calibration for use in multiwavelength pyrometry. These techniques also demonstrate the potential to perform in situ, real-time process monitoring and precise temperature measurement during the deposition of thin dielectric films on silicon wafers.