Star colours as a check on photometry

NOTE This page is very out of date. They refer to Dave Balam's CFHTLS stacks from 2003. I now use exclusively my own stacks (see the appropriate field).
The colours of stars in the the CFHTLS D3 field are compared to synthetic colours produced by multiplying stellar spectra by the griz filter responses. The results indicate that the zeropoints have been calculated correctly and that the filter response curves are correct.
Motivation: In order to measure photometric redshifts using the template method, it is absolutely essential to have accurate photometry and to have accurate filter response functions. Bad photometry, particularly systematic zeropoints offsets mean that observed SED's will be poorly measured. Inaccurate filter response functions mean that the templates will be poorly determined. In either of these cases, the result be a poor match between the observed SED's and the synthetic templates and hence bad photometric redshifts.

A useful diagnostic of these effects is to examine the colours of stars. One can compare the observed colours of stars with synthetic colours generated by multiplying stellar spectra by the filter response functions. Then one plots colour-colour diagram with the observed and synthetic colours. Stars have a relatively constrained locus in colour space. Any offsets between the observed and synthetic colours indicates a problem with one or the other, either zeropoint errors or problems with the filter response functions or the software that multiplies the spectra by the filters.

The figures at right show colour-colour plots for stars in the SDSS ugriz system. The green dots represent the results of multiplying the Pickles spectra by the SDSS filter curves obtained from here The black dots show the colours of the ugriz standard stars as given by Smith et al. 2002 as retreived from this web site

The two sets of data are in good agreement. This indicates that the SDSS filter response functions are correct and that the software is function correctly.

The figures at right show the same colour plots for the stars in the CFHTLS D3 stacks prepared by Dave Balam. The stars were identified by plot half-light radius against magnitude. The stellar locus is easily identified down to z=19 and with some difficulty to z~21. This is shown in the top left panel where the stars have been highlighted in red. Note that these cuts in half-light radius and magnitude are somewhat conservative.

The other two plots show the colours of the stars selected in this manner in black overlaid in green with colours determined by multiplying the Pickles spectra by the CFHTLS griz filters. Note that the CFHTLS version of the ugriz filter set is not identical to the SDSS version, as discussed here

Overall the colours of the CFHTLS stars seem to agree well with the predictions, at least to within 0.1 mag. The only cause for concern is the elbow at (g-r)=1.3 and (r-i)=0.5 where a slight displacement may be occuring. This indicates that the zeropoints are fairly well understood.

The figures at right show the same colour plots for the stars in the CFHTLS D4 stacks prepared by Dave Balam. Overall the star colours D4 agree significantly less well than in the D3. It would seem there is 0.3 offset in the r-i colour.
Since the test works for the SDSS data and D3 D3 data, it is probable that the software and filter curves are correct. However, the test fails for the D4 data, indicating that the zero-points are poorly understood at this point for this field.
I've had a number of requests for the colours of the Pickles stars. Here they are: Each file starts with a line listing the filters. Each subsequent line has the name a the Pickles spectrum in the first column and the magnitudes (with an arbitrary but fixed offset) in the other columns. Subtracting any two columns gives you a colour. All magnitudes/colours are in the AB system (as opposed to the Vega or ST system).

Back to Stephen Gwyn's CFHTLS pages
Send comments/suggestions/problems to