Sample selection – finding disk galaxies

This post is partially based on a series I wrote on the previous version of Galaxy Zoo Talk. I hope it will be shorter.

The analysis in the previous posts is applicable to galaxies whose stars and/or ionized gas are confined to a thin disk with no net vertical bulk motion. In that case velocities can be treated as two dimensional and, for moderately inclined disks, the full velocity vectors recovered from the observed radial velocities. The thin disk approximation is an idealization of course, but it’s reasonable at least for late type spiral galaxies that lack significant bulges.

I ultimately tried three different ideas to acquire a statistically meaningful sample of disk galaxies without actually visually examining all ∼2700 of them (this is certainly feasible, but somewhat error prone and tedious). The first idea was to use the drpall catalog to select a sample of probable disk galaxies. The catalog contains a single morphology proxy in nsa_sersic_n derived from single component sersic index fits to r-band SDSS imaging. A forgotten paper suggests disk galaxies have n ≤ 2.5, so I adopted that as an upper limit. There are several estimates of the minor to major axis size ratio, which for a disk galaxy is a proxy for the cosine of the inclination. The one recommended by the MaNGA science team is nsa_elpetro_ba. I chose a conservative range of inclinations of between 30 and 60^o, or \(0.5 \le \cos i \le 0.866\).

The full MaNGA sample contains several subsamples: the main ones are a primary sample intended to be observed out to ≈1.5 effective radii and a secondary sample observed to ≈2.5R_eff. I initially thought the latter would be most suitable for analysis. The subsample an observation belongs to is identified by three sets of bitmasks, with mngtarg1 used for galaxies in the primary and secondary samples. So, the R code to select this sample is:

attach(drpcat)
maybedisks2 <- which((mngtarg1==4 | mngtarg1==32 | (mngtarg1 >= 2048 & mngtarg1 < 4096)) & nsa_sersic_n < 2.5 & nsa_elpetro_ba>=0.5 & nsa_elpetro_ba<=0.866)
detach(drpcat)

This returns a vector of 254 row indexes in drpcat (in the DR14 release). The following short R function takes the indexes and writes out a text file of MaNGA file names suitable for use with wget

writefn <- function(indexes, ftype="rss", cat=drpcat, outfile="fgets.txt") {
  ftype <- toupper(ftype)
  plates <- cat$plate[indexes]
  plateifus <- cat$plateifu[indexes]
  fnames <- paste(plates, "/stack/manga-", plateifus,
                  "-LOG", ftype, ".fits.gz", sep="")
  cat(fnames, file=outfile, sep="\n")
  fnames
}

I soon found out that this sample has problems. The main issue is that although the IFU covers almost the full visible extent of these galaxies the exposures are no deeper than for the primary sample. Even with optimistic S/N thresholds there is little usable data at large radii.

My next idea was to try the very pure sample of spirals of Masters et al. (2010) selected from the original Galaxy Zoo project. A position cross-reference turned up 69 observations of 68 unique galaxies in the first two MaNGA releases. This is a bit small for the purpose I had in mind, and the sample size was further reduced by another 10-12 galaxies that were too nearly face-on to analyze.

In search of a larger sample I again turned to Galaxy Zoo data, this time the main spectroscopic sample of GZ2 (Willett et al. 2013) which is conveniently included in the SDSS databases. A complication here is that although the vast majority of MaNGA galaxies have single fiber spectra from SDSS their id's aren't tabulated in drpall. Although I could probably have done a simple position based cross match of GZ2 galaxies with MaNGA targets I decided instead to look for SDSS specObjID's near to MaNGA target positions, and then join to matching GZ2 targets. This required a command I hadn't used before and a table valued function that was also new to me:

select into gz2disks
  m.mangaid,
  m.plateifu,
  m.objra,
  m.objdec,
  m.ifura,
  m.ifudec,
  m.mngtarg1,
  m.mngtarg3,
  m.nsa_z,
  m.nsa_zdist,
  m.nsa_elpetro_phi,
  m.nsa_elpetro_ba,
  m.nsa_elpetro_th50_r,
  m.nsa_sersic_n,
  z.specObjID
from mangaDrpAll m
cross apply dbo.fGetNearbySpecObjEq(m.objra, m.objdec, 0.05) as n
join zoo2MainSpecz z on z.specobjid=n.specObjID
where
  m.mngtarg2=0 and
  z.t01_smooth_or_features_a02_features_or_disk_weighted_fraction > 0.5 and
  z.t02_edgeon_a05_no_weighted_fraction > 0.5 and
  z.t06_odd_a15_no_weighted_fraction > 0.5 and
  m.nsa_elpetro_ba >= 0.5 and
  m.nsa_elpetro_ba <= 0.866 and
  m.mngtarg1 > 0

The line that starts with cross apply builds a table of position matches of MaNGA objects to spectroscopic primaries, then in the next line joins data from the GZ2 spectroscopic sample. The last lines set some thresholds for weighted vote fractions for a few attributes: yes votes for "features or disk", no votes for "edge on", and no votes for "something odd". I chose, more or less arbitrarily, to set all thresholds to 0.5. The final lines set inclination proxy limits as discussed above. There were also questions about bulge size, spiral structure, and bars in the GZ2 decision tree that I did not consider but certainly might if I return to this subject.

This query run in the DR14 "context" produced 359 hits out of about 2700 galaxy targets in the first two MaNGA public releases, which is certainly an incomplete census of disk galaxies. On the other hand all 359 appeared to me actually to be disk galaxies, so the purity of the sample was quite high. Next time I'll do something with the sample.