Multivariate Analysis as a Tool to Identify Concentrations from Strongly Overlapping Gas Spectra

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Chemometrics

Background:

• Spectroscopic analysis of gas mixtures often faces challenges due to overlapping absorption spectra.
• Accurate quantification of individual components in complex gas mixtures is crucial for various industrial and environmental applications.
• Broadband spectroscopy, while powerful, requires advanced methods to deconvolve signals from multiple species.

Purpose of the Study:

• To develop and validate a multivariate analysis (MVA) method for determining concentrations of single volatile organic compounds (VOCs) within gas mixtures.
• To address the challenge of overlapping absorption spectra in mid-infrared (mid-IR) photoacoustic (PA) spectroscopy.
• To assess the accuracy and reliability of the proposed method for quantitative gas sensing.

Main Methods:

• Application of multivariate analysis (MVA), specifically Partial Least Squares regression (PLS), to mid-IR photoacoustic spectroscopic data.
• Utilizing spectroscopic data in the 3250 nm to 3550 nm wavelength region for detection of VOCs.
• Calibration of the PLS model using known concentrations of different VOCs.

Main Results:

• The PLS model successfully calculated concentrations of individual VOCs from complex gas mixtures with overlapping spectra.
• The developed method achieved a high relative accuracy of 2.60% in determining VOC concentrations after calibration.
• Demonstrated the effectiveness of MVA in overcoming spectral interferences in broadband spectroscopy.

Conclusions:

• Multivariate analysis, particularly PLS regression, is a robust technique for quantitative analysis of gas mixtures using mid-IR PA spectroscopy.
• The method provides accurate and reliable determination of VOC concentrations, even with significant spectral overlap.
• This approach offers a valuable tool for advanced gas sensing applications requiring precise component quantification.