The following images came from I did with the GEMS survey (Rix et al. 2004, ApJS, 152, 163). Some of the original data were also from the GOODS Legacy program using the Advanced Camera for Surveys (ACS) on-board the Hubble Space Telescope (HST). These images were automatically fitted using a master wrapping script that extracted basic galaxy information from a large database of objects, and fitted them in batch-mode using GALFIT without any human interaction. In addition to fitting the galaxies, the wrapping script figured out which neighboring galaxies needed to be simultaneously fitted and deblended by GALFIT.

One of the challenges in automating GALFIT for batch fitting is to figure out how to translate global galaxy parameters such as total magnitude, some kind of radius, etc., into initial parameters for a wide variety of galaxy types. Being a gradient algorithm, a common concern is that GALFIT might be easily confused, therefore require initial parameters to be finely ``tuned'' for each galaxy type, especially when trying to do bulge-to-disk decomposition (B/D). Interestingly, this has not been the case for the examples shown below. Although in theory a gradient-based algorithm is not the most robust algorithm to be using for automation, in practice it is at least as robust as other algorithms which use Metropolis or annealing methods, but much faster. When doing single component Sérsic fits, GALFIT was very robust and insensitive to input parameters. Even when doing B/D decomposition, simple ways of converting global galaxy photometry into initial parameters, using a single, generic prescription for all galaxies worked to better than 90% for a test sample of real galaxies. It mattered little what that prescription was, or what the initial parameters were as long as they were sensible (or often when not -- the initial guesses were typically off by a factors of 5-30 to start with). The other 5-10% or so were problematic because the use of 2 component model was not a good assumption. A common example was when a galaxy had a strong bar, or multiple halo components. Another fairly common example included galaxies which had large central concentrations (Sérsic index n>4) because of a compact nucleus, or because the bulge may be a compound bulge.

The examples below show some of the fits done in batch mode by using a GALFIT wrapper script, for doing bulge-to-disk decomposition. Only one galaxy in each image -- the central one (called the ``primary'') -- is fitted with a Sérsic + exponential disk component. Neighboring galaxies are fitted using a single Sérsic component to reduce their contamination on the central galaxy, not to get their parameters exactly right. Every galaxy eventually gets to be fitted as the ``primary,'' when the objects are cycled through sequentially.

Left = original data, Middle = model fit, Right = residual image.

Go back to GALFIT Home Page.
Go back to Chien Peng's Home Page.

Chien Peng ()