Potential of far-ultraviolet absorption spectroscopy as a highly sensitive analysis method for aqueous solutions. Part II: monitoring the quality of semiconductor wafer cleaning solutions using attenuated total reflection

• Published in: Applied spectroscopy Vol. 62; no. 9; p. 1022
• Main Authors: Higashi, Noboru, Ikehata, Akifumi, Kariyama, Naomi, Ozaki, Yukihiro
• Format: Journal Article
• Language: English
• Published: United States 01.09.2008
• ISSN: 0003-7028
• DOI: 10.1366/000370208785793236

Far-ultraviolet (FUV) spectroscopy combined with attenuated total reflection (ATR) is employed for direct measurement of the concentrations of semiconductor wafer cleaning fluids such as SC-1 (aqueous solution of NH3 and H2O2) and SC-2 (aqueous solution of HCl and H2O2). FUV spectra of these aqueous solutions in the 170-200 nm region are highly sensitive to changes in both hydrogen bonding and hydration. Although ATR measurement results in lower absorptivity compared to transmittance measurement, it is possible to increase absorption with greater evanescent wave penetration depth using a low refractive index internal reflection element (IRE). We adopt quartz as an IRE material. Since the refractive index of quartz becomes lower than that of water in the low energy side of an intense absorption band due to the n-->sigma* transition of water, the quartz IRE yields non-total reflection wavelength regions. However, near 175 nm the effective absorptivity of the tail of water's absorption band can be successfully enlarged, making the FUV-ATR technique suitable for measuring the concentrations of the components in the semiconductor wafer cleaning fluids. In the present study we prepared the same cleaning fluids as those used in actual semiconductor fabrication and measured their FUV-ATR spectra in the 150-300 nm wavelength range. It was found that even with the quartz IRE one can measure FUV-ATR spectra under total reflection conditions at 175 nm or above. We created calibration models for predicting both NH3 and H2O2 in the concentration ranges of 0-10% in SC-1 using multiple linear regression (MLR). The standard deviations of the models were 0.033% and 0.265% for NH3 and H2O2, respectively. The same procedure was repeated under the same conditions for HCl and H2O2 in SC-2, yielding corresponding values of 0.018% for HCl and 0.178% for H2O2.