alfa: automated line fitting algorithm

installmanualchange logsource code on github

Animation showing ALFA fitting a spectrum

Cutting edge astronomical instruments can now deliver data in unbelievable quantities. MUSE on the VLT, for example, can give you some 90,000 spectra simultaneously, each one covering 4800-9300Å at a resolution of λ/Δλ=3000. Extracting all the available information from such a large quantity of data in a timely fashion is now the challenge. Fast and reliable algorithms for the automated measurement of spectra are required.

For nebular spectra, which typically consist of narrow emission lines superimposed on a weak continuum, a typical analysis requires first the fitting of Gaussian profiles to observed emission lines, then the identification of the ion responsible for each of the lines thus measured, and finally the derivation of physical parameters such as temperature, density and ionic abundances from the list of measured line fluxes. Automated fitting routines still leave the user with the task of identifying the lines which have been detected.

alfa does the first two of these three steps with no need for any user input at all. It works in the opposite direction to the traditional analysis - from a list of lines likely to be present in a nebular spectrum, alfa constructs a synthetic spectrum to match the observations. It optimises the parameters of all the Gaussian line profiles by means of a genetic algorithm. In this approach, the line identification is known from the beginning. The processing time depends on the number of wavelength bins in the spectrum and the number of lines assumed to be present; the current version of the code takes a few seconds to fit a spectrum containing 2000 data points in a wavelength range covering 285 emission lines. Running on four processors, alfa can fit 2 million lines in about 20,000 spectra in around four hours.

Output from alfa can be fed directly into the nebular empirical analysis tool, which will turn a list of line fluxes into measurements of temperatures, densities and abundances in a small fraction of a second. The two codes in combination thus carry out in a few minutes work that would formerly have taken days, weeks or months.

A paper describing alfa was published in January 2016: Wesson R., 2016, MNRAS, 456, 3774.