GETTING THE MOST OUT OF A LOT: THE ANALYSIS OF FOURIER TRANSFORM SPECTRA

Abstract

The information content in a single Fourier Transform spectrum can be enormous. Often there are on the order of $10^{6}$ independent points each with a signal to noise ratio of $10^{3}$ or even higher, all of which is readily available in computer readable form. Obtaining the best information available from these spectra requires careful consideration of the instrumental effects and the spectroscopic modeling of the spectrum. The nonlinear least squares fitting technique has been employed by a number of groups in order to derive spectral line parameters. More recently, this technique has been refined to simultaneously include several laboratory spectra obtained under different physical conditions in a single solution. Correlations between parameters are often decreased in this manner. This technique becomes very computationally intensive without careful consideration of the best mathematical techniques. Of particular interest with this technique is the evaluation of estimated uncertainties in the derived parameters. With large quantities of data in a single solution, uncertainties due to random noise in the spectra may become considerably smaller than uncertainties due to systematic errors such as modeling problems and measurement of physical parameters. One must also be careful to monitor close correlations between parameters of the solution.