We wanted to share a plot showing some of the very early science results coming from theSkyNet POGS. There has been debate in the astronomy research community about the trustworthiness of estimated stellar mass and star formation rate values derived from spectral energy distribution (SED) fitting of entire galaxies at once. Traditionally, the total luminosity (at each wavelength) coming from the sum of all galactic sub-components is modeled without regard for the spatially-varying contribution of stellar populations having different characteristic age (old vs. young). This variation is clearly significant in certain galaxy types - for instance, it is one reason the spiral arms of a star-forming disk galaxy appear so prominently.
By lumping all the light from sub-components together before interpretation, potential biases are introduced into results from integrated (total) SED fitting. We have the opportunity to do better than this traditional approach. With your CPU-power, we can very efficiently fit for the diverse stellar populations seen on small (pixel-size) scales and only later sum our parameter maps (examples given in the earlier post linked below) across entire galaxies. In this way, we can use the extra information provided by the multi-wavelength morphology of each galaxy to get a more reliable answer for questions such as how much stellar mass resides in a galaxy, what is the rate of star formation, and how are mass/SFR related in galaxies of varied type.
Our POGS analysis for a small sample of objects run late last year shows that significant bias can be removed by working in the pixel-SED regime. In particular, using the traditional approach the stellar mass of a galaxy appears to be preferentially underestimated, by a factor which increases for decreasing galaxy mass. Also, a significant SFR bias is also demonstrated when comparing integrated-SED versus pixel-SED results. The plot we wanted to share (below) shows these results by drawing a vector for each galaxy in the mass-SFR parameter plane, connecting the estimated parameters evaluated via the two competing methods. The filled circles show the 'improved' location for each galaxy. This is not simply a trivial result. It likely has wide-reaching implications. As one example, scatter in the relation between stellar mass and SFR seen in this plot is reduced (the so-called galaxy main sequence becomes tighter), allowing us to track galaxy evolution with more fidelity. Adoption of the pixel-SED method will also influence interpretation of galaxy rotation curves and even the star formation law.
We are about to start using the much larger number of galaxies processed by now to attempt establishing parametric corrections for application to distant (progressively unresolved) galaxy populations. This would not be possible without your assistance -- thanks!