I only have time to analyze one MaNGA galaxy for now and since it was the first to get the correctly coded LOSVD estimates I chose the Coma Cluster S0 galaxy NGC 4949 that I discussed in the last post for SFH modeling. As I mentioned a few posts ago trying to model the LOSVD and star formation history simultaneously is far too computationally intensive for my resources, so for now I just convolve the SSP model spectral templates with the elementwise means of the convolution kernels estimated as described previously and feed those to the SFH modeling code. My intuition is that the SFH estimates should be at least slightly more variable if the kinematics are treated as variable as well, but for now there’s no real alternative to just picking a fiducial value. Of course a limited test of that hypothesis could be made by pulling multiple draws from the posterior of the LOSVD distribution. I will certainly try that in the future.
My first idea was to leave both dust and emission lines out of the SFH models. The first choice was to save CPU cycles and also based on the expectation that these passively evolving galaxies should have very little dust. The results were rather interesting: here are ribbon plots of model SFH’s for all 86 binned spectra. The pink ribbons are from the original model runs, which assumed single component gaussian velocity distributions, modified Calzetti attenuation, and included emission lines (which contribute negligible flux throughout). The blue ribbons are with non-parametric LOSVD and no attenuation model:
The differences in star formation histories turn out to be almost entirely due to neglecting dust in the second set of model runs. In this galaxy there are some areas with apparently non-negligible amounts of dust:
Ignoring attenuation forces the model to use redder, hence older (and probably more metal rich although I haven’t yet checked) stellar populations and this has, in some cases, profound effects on the model SFH.
So, I tried a second set of model runs adding back in modified Calzetti attenuation but still leaving out emission:
And now the models are nearly identical, with minor differences mostly in the youngest age bin. All other quantities that I track are similarly nearly identical in both model runs.
Here is the modeled mass growth history for a single spectrum of a region just south of the nucleus that had the largest difference in model SFH in the second set of runs, and the largest optical depth of attenuation in the first and 3rd. This is an extreme case of the shift towards older populations required by the neglect of reddening. Well over 90% of the present day stellar mass was, according to the model, formed in the first half Gyr after the big bang with almost immediate and complete quenching thereafter. While not an impossible star formation history it’s not one I often see result from these models, which strongly favor more gradual mass buildup.
So, the conclusion for now is that having an attenuation model matters a lot, the detailed stellar kinematics not so much. Of course this is a relatively simple example because it’s a rapid rotator and the model convolution kernels are fairly symmetrical throughout the galaxy. The massive ellipticals and especially cD galaxies with more complex kinematics might provide some surprises.