Robert Noll

Processing multi-species terahertz spectra for detection of a particular species

The development of tera-hertz sources and receivers recently has given rise to a desire to do spectroscopy on molecular rotational spectra in this region. In particular the use of semiconductor photo-mixers provides practical packaging concepts and good sensitivity for building absorption spectrometers. This paper deals with some data processing concepts associated with absorption spectra from such spectrometers. Specifically, the concepts of multivariate analysis developed and applied over the last 25 years to infrared spectroscopy appear to be very useful to the analysis of tera-hertz absorption spectra. These concepts are particularly useful to address problems associated with: spectra baseline variations, species concentration estimates and spectral cross talk assessments when multiple species are present in the measurement sample. This latter element is of particular importance in tera-hertz spectroscopy when exploring multiple species with overlapping rotation spectral bands. In this paper some basics of multivariate analysis are reviewed. The baseline signatures or backgrounds of our photo-mixer absorption spectrometer are described and processed. Algorithm results are presented that remove backgrounds and other instrument spectral artifacts. Finally, a simulated example of separating overlapping spectra is presented.