IRAF News Number 13

FORTRAN Program and IRAF Script Produce Fast Scientific Output

Dave Latham, SAO
(IRAF Newsletter 13, December 1994, page 19 [PDF] [PS])

For more than 15 years we have been accumulating large numbers of single-order echelle spectra of stars in order to study the frequency and orbital characteristics of binary stars in a variety of stellar populations. We now have more than 120,000 spectra in our on-line database, and in a good month we get more than 1,000 new spectra. To handle this volume of data we have developed procedures for mass-producing the data reductions and analysis.

Five years ago we converted from our custom software running on Data General Nova computers to workstations. We ported across the code for doing wavelength solutions and flat-fields, but we decided to do the velocity correlations inside of IRAF. We ended up developing our own cross-correlation application, xcsao.

Each month the new spectra are reduced using an observation of the dusk sky as the template.To adjust for the small shifts in the zero point of our velocity system, we take dozens of exposures each month, both of the dusk sky and of standard stars.

When we want the best possible velocities for a given star, to see if it is variable or to solve forits orbit, we have tools for easy extraction of all the spectra of that star from the online data base. Then we can recorrelate the spectra against an optimum template, or a grid of templates, chosen from our library of several thousand synthetic templates calculated by Jon Morse, based on the latest Kurucz stellar atmospheres.

The procedure for recorrelating the spectra of a star or group of stars is quite simple. Weprepare a file which gives the name of the star and the characteristics of the template that we want to use (effective temperature, surface gravity, metallicity, and rotational velocity), one lineper star/template combination. One option is to ask for a grid of several different rotational velocities. We have FORTRAN programs which take this input and create an IRAF script forcorrelating each spectrum. The code selects the nearest template from our grid, looks up the proper zero-point shift from another data base, figures out which telescope the exposure wasmade on, and creates an output file with the necessary IRAF commands to do the correlations, typically 3 or 4 lines per correlation. The FORTRAN program also sets several hundred IRAF parameters at the beginning of the script, so that we can be sure that all the relevant parameters have been set to our default choices.

This approach to mass-producing reductions can be quite powerful. For example, last Sunday Idecided to rereduce nearly 4000 observations for 90 spectroscopic binaries using new information about the temperatures and metallicities to choose calculated templates. The biggest part of the job was the preparation of the input file specifying the star names and parameters. It took about 5 hours for my Sun to wade through the correlations, creating an output file which I couldthen feed into my orbital-solution programs. By the end of the day I had 90 new orbits!

[rvsao] [rvsao publications]