근적외선을 이용한 세정용액 조성의 온라인 분석 및 신뢰성 있는 테라헐쯔 정량분석을 위한 최적의 고체시료 조건 연구

Abstract

A simple near-infrared (NIR) spectroscopic method for on-line compositional analysis of etchant solutions has been described in the part A of this thesis. Initially, the feasibility of NIR spectroscopy for the measurements of important components in etchant solutions directly through Teflon tubing (PFA: perfluoroalkoxy fluorocarbon) was investigated. Partial least squares (PLS) prediction accuracy was acceptable and the utility of Teflon as a reliable sample container for the NIR analysis was confirmed. Then, the robustness of Teflon-based NIR measurement was evaluated by predicting the spectra of 10 independent samples that were collected in five different Teflon tubes. Although, two Teflon tubes were visually less transparent than the other three, there was no significant variation in the standard error of predictions (SEPs) among the measurements using 5 different Teflon tubes. The accuracy of the measurement was successfully maintained although different Teflon tubes were employed for NIR spectral collection. Finally, the same NIR measurement scheme has been extended for real on-line monitoring of the components in an etchant solution and the prediction selectivity among the components has been also examined. The selectivity of PLS model is the most critical issue determining reliability of the Teflon-based NIR measurement, since the spectral features of etchant solutions were indistinct with severe overlapping. To ensure the selectivity of a calibration model, an optimal spectral range was probed with the use of moving window PLS. When an optimized model was used, the prediction of concentration selectively followed according to the spike of a given component; while, the prediction selectivity was degraded when non-optimized ranges were used for calibration. In addition, to take advantage of the more resolved Raman features of the components, the compositional analysis of etchant solutions containing hydrogen peroxide (H2O2) and sulfuric acid (H2SO4) has been attempted using Raman spectroscopy. Raman spectra of the samples were directly collected through a Teflon tube. Direct Raman measurement through a Teflon tube is advantageous because a non-overlapping Teflon band could be effectively used as an external standard to correct for variation in the Raman intensity. The concentrations of H2O2 and H2SO4 were accurately predicted and the intensity variation was effectively corrected using the Teflon peak. In the part B, for reliable quantitative analysis using terahertz (THz) spectroscopy, optimal sample preparation conditions for solid samples providing reproducible THz spectral feature have been studied. In the case time-domain THz measurement, the spectral feature of a sample could be sensitively varied by its physical conditions such as a sample thickness, density, packing and heterogeneity of a sample, since these could sensitively alter the velocity of THz radiation propagating in a sample. For the investigation, different thicknesses of polyethylene (PE) pellets containing lactose from 5 to 15 wt% were prepared. The influence of pellet thickness on the reproducibility of transmission THz spectra was examined. Finally the optimum thickness for the quantification of lactose in a PE pellet matrix was evaluated with the use of PLS.