Analysis of spectra using correlation functions.

We present a novel method for the quantitative analysis of spectra based on the properties of the cross correlation between a real spectrum and either a numerical synthesis or a laboratory simulation. We propose a new goodness-of-fit criterion called the heteromorphic coefficient H that has the property of being zero when a fit is achieved and varying smoothly through zero as the iteration proceeds, providing a powerful tool for automatic or near-automatic analysis. We also show that H can be rendered substantially noise-immune, permitting the analysis of very weak spectra well below the apparent noise level and, as a by-product, providing Doppler shift and radial velocity information with excellent precision. The technique is in regular use in the Atmospheric Trace Molecule Spectroscopy (ATMOS) project and operates in an interactive, real time computing environment with turn-around times of a few seconds or less.