Star forming galaxies in the quench and control samples
-
by
mlpeck
I tried doing this exercise in tools, but it got too frustrating. With appropriate filters it is possible to isolate starforming galaxies in the BPT diagram. You should apply S/N cuts on emission line fluxes and add the following two filters to select starforming galaxies
filter .o3hbeta < 0.61/(.n2halpha - 0.05) + 1.3
filter .n2halpha < 0.05
This is just the Kauffmann 03 empirical line separating starforming from "composite" galaxies as given in Kewley et al. (2006). You need the second filter to keep some LINERS from sneaking into the starforming group.
In my database -- which differs a little from what I got in tools for reasons I haven't explored yet -- I found 1035 starforming objects in the control sample and 1015 in the quenched sample, practically the same percentage (30%) in both samples.
More significantly, I don't see any significant differences between the quench and control samples in any properties that I've looked at so far. They have the same distributions of star formation rates, Hα luminosities, stellar masses, and colors. They occupy the same general area in a color-absolute magnitude plot. Am I missing any significant properties? Most likely. I haven't looked for morphological differences for example.
This is more or less a placeholder for now, until I find time to make some graphs. Perhaps some others will have more success using tools than I do.
I have a hypothesis about this, but it really requires more detailed knowledge of how the quench sample was selected.
Posted
-
by
mlpeck
I'm going to post a series of histograms comparing various measured properties of the starforming galaxies in the quench and control samples as I get to them. First, the star formation rate (in solar masses per year) as returned by the MPA-JHU pipeline:
Just eyeballing this it appears to me the decay rate in the quench sample is faster than in the controls. The median SFR for the quench sample is 1.53 Msun/yr., vs. 1.96 for the controls.
Posted
-
by
mlpeck
Next, Hα luminosity, in ergs/sec. (dex), corrected for internal extinction per the unofficial mlpeck pipeline:
This is for the fiber only, which has the advantage of avoiding the possibly dubious extrapolation to the Petrosian radius. Median luminosities are 40.80 (quench), 40.87 (control).
Posted
-
by
mlpeck
Specific star formation rate. This is the star formation rate divided by the stellar mass. From the official MPA-JHU pipeline.
Median is -9.87 dex (quench), -9.97 (control).
Posted
-
by
mlpeck
Log of the star formation rate (
same information as 3 panels up, different presentation):
That parenthetical comment wasn't quite right: the bin width in the first plot was 1 solar mass/year; it's 0.02 dex in the above plot, or about 5%. The leftmost 12 1/2 bins in this plot all fall into the first bin in the first plot. 36% (364/1015) of the starforming quench sample have star formation rates < 1 Msun/yr., compared to 29% (298/1035) in the control sample.
Another way to do the comparison is to look at the cumulative distribution functions of the (log) star formation rates:
Note that for any quantile between about the 5th and 90th percentile the control sample star formation rate is higher than the quench. A two-tailed Kolmogorov-Smirnov test rejects the hypothesis that these distributions are equal (p < 0.0002).
Posted
-
by
mlpeck
D4000. This is the first one where I see a rather clear difference:
Posted
-
by
mlpeck
A scatterplot this time of u-r color against R band absolute magnitude.
Posted
-
by
mlpeck
Plots of empirical cumulative distribution functions are more powerful tools than I realized. I suspected that quench sample objects were slightly shifted to the red in (u-r) color relative to the controls in the plot above. Comparing the CDF's confirms that suspicion:
A two-tail KS test says these are different at arbitrary significance levels.
By the way a plot like this shows one way to deal with outliers. The full range of colors goes at least to (u-r) = 8, but a percentage that's indistinguishable from 100 have, more realistically, (u-r) < 4.
Posted
-
by
JeanTate
Very cool, mlpeck! 😃
This may seem a really dumb question, but do you get the same results if you run the CDF backwards? For example, in the last one, if the x-axis goes from 4 to 0 (say), or if you plotted (r-u) instead?
We know that the QC catalog was selected so that the log_mass of each object in the QS one had a matching counterpart in QC (and that the match was also good, in redshift, to < 0.02), so a CDF of the SSFR (specific star formation rate) should look the same as the CDF of the SFR one (and be pretty much the same, in terms of KS significance) ... but does it?
Posted
-
by
JeanTate
in response to mlpeck's comment.
It'd be nice to see this as a CDF too.
BTW, how did you correct for internal extinction? Also, how did you convert H-alpha flux to H-alpha luminosity?
Posted
-
by
mlpeck
in response to JeanTate's comment.
This may seem a really dumb question, but do you get the same results if you run the CDF backwards? For example, in the last one, if the x-axis goes from 4 to 0 (say), or if you plotted (r-u) instead?
Except for being a mirror image of the original the results would be the same. Specifically, a Kolmogorov-Smirnov test will return the same statistics and significance levels.
BTW, how did you correct for internal extinction? Also, how did you convert H-alpha flux to H-alpha luminosity?
I just take a conventional approach of calculating a color excess from the Balmer decrement assuming an intrinsic ratio of Hα/Hβ = 2.86. For expedience I just correct using the galactic extinction law of Fitzpatrick (1999). Someone who is fussy about this would probably use something else. If a measured ratio isn't available or implies negative reddening I assume no extinction. Probably also someone being fussy would perhaps assume a typical value. I have my own cosmological distance calculator, which I've verified against NED's.
If the science team decides this is an interesting topic to explore I highly recommend that they do their own calculations, since they'll probably want to use a different extinction1 law and cosmological parameters.
1I think the term of art among extragalactic astronomers is "attenuation," but since I'm just a citizen scientist I'll allow myself sloppy usage here.
Posted
-
by
mlpeck
in response to JeanTate's comment.
so a CDF of the SSFR (specific star formation rate) should look the same as the CDF of the SFR one (and be pretty much the same, in terms of KS significance) ... but does it?
That was actually the last plot I wanted to make in this series for now. Yes, the specific SFR is slightly lower in the quench sample over most of the observed range:
Posted
-
by
JeanTate
in response to mlpeck's comment.
Thanks for the answers on extinction and calculation.
This is for the fiber only, which has the advantage of avoiding the possibly dubious extrapolation to the Petrosian radius.
Aside from (small?) variations due to differences in seeing, between observations of different objects1, isn't there another factor that has to be controlled/estimated, to make the comparisons meaningful?
3" on the sky (the fiber diameter) corresponds to very different physical distances at the galaxy; the H-alpha luminosity (as you derive it) for a kpc or so of a big, low-z galaxy cannot easily be compared with that of basically the whole galaxy of a small, high-z one, can it? And comparing QS with QC amplifies this: the two catalogs are closely matched for redshift, but not for size ... and QS objects are systematically smaller - both on the sky and physically - than the corresponding QC ones.
1 everything else is already either the same (e.g. integration time) or minimized to noise (hopefully; e.g. by calibration)
Posted
-
by
mlpeck
in response to JeanTate's comment.
3" on the sky (the fiber diameter) corresponds to very different physical distances at the galaxy; the H-alpha luminosity (as you derive it) for a kpc or so of a big, low-z galaxy cannot easily be compared with that of basically the whole galaxy of a small, high-z one, can it? And comparing QS with QC amplifies this: the two catalogs are closely matched for redshift, but not for siz
Yes, absolutely! That's why I used the official pipeline star formation rate estimates for most of this discussion. I did verify here that there's a reasonably tight relation between the estimated SFR and my calculation of Hα luminosity after extrapolating to the estimated galaxy size, and that it follows Kennicutt's scaling relation to a good approximation.
Unfortunately the comparison is only useful for objects that fall in the starforming region of the BPT diagram, which is what I'm looking at here. I have no idea at the moment how useful the SFR estimates are outside that region, or even how one might go about verifying them.
Posted