Galaxy Zoo Starburst Talk

Objects - Galaxies? - indistinguishable from point sources

  • JeanTate by JeanTate

    Background

    In Bamford et al. (2009), which describes how the classification bias was identified and quantified, for the original Galaxy Zoo project, there is a size cut-off: Petro_R50 ~1" ("size greater than twice the angular resolution"). In Willett et al. (2013), which describes something similar for the GZ2 project, one of the selection criteria is a size one, "a size limit of petroR90_r>3 arcsec (petroR90_r is the radius containing 90% of the r-band Petrosian aperture flux)."

    In DR9 there are five parameters (one per band) which provide estimates of the probability that an object has a radial intensity profile consistent with that of the PSF (point source function) at that position in the field, probPSF_x (where x is u, g, r, i, or z).

    The JPG images we got to classify in the Quench project are produced from the data in just three bands, g, r, and i.

    The downloadable Quench project data, the data you can analyze using Tools, includes a field Petro_R50, which is the radius of the object; more exactly, the radius containing 50% of the r-band Petrosian aperture flux. A fair approximation is that an object with a Petro_R50 of ~0.7" or smaller would likely be indistinguishable from a star (a point source). Why? Because the mean r-band seeing during the SDSS runs which produced the data we ended up using was ~1.4", which is a FWHM (full width, half-maximum) measure.

    What I would like to do, in this thread, is examine the QS and QC objects which are "small"; specifically, to:

    • see how they were classified (what 'the consensus morphology' of each is)
    • compare their probPSF_x (x=g, r, i) values
    • discuss how likely it is that they are not galaxies

    First, the data.

    Posted

  • JeanTate by JeanTate

    QS data

    There are 177 QS objects with Petro_R50 ("R50") values < 1.0". In bins of 0.1":

    • < 0.6": 2
    • > 0.6" to < 0.7": 8
    • > 0.7" to < 0.8": 23
    • > 0.8" to < 0.9": 57
    • > 0.9" to < 1.0": 87

    Only three are classified as "Star or artifact" ( AGS000005o, R50=0.84", AGS00001uo, R50=0.85", and AGS00002ak, R50=0.95"); 14 as "Features or disk", and 160 as "Smooth".

    The probPSF_x (x=g, r, i) and Petro_R90 ("R90") for the ten smallest - all with R50 < 0.7" - are as follows (R50, R90, probPSF_g, _r, _i):

    enter image description here AGS00000mt (0.58, 0.98, 1, 0, 0)

    enter image description here AGS00001z3 (0.58, 1.3, 1, 1, 1)

    enter image description here AGS00001lz (0.62, 1.5, 0, 0, 0)

    enter image description here AGS000021r (0.63, 0.91, 1, 1, 1)

    enter image description here AGS00001eg (0.66, 2.15, 0, 0, 0)

    enter image description here AGS00001fy (0.66, 2.0, 0, 0, 0)

    enter image description here AGS00000wq (0.68, 6.9, 0, 0, 0)

    enter image description here AGS000028i (0.68, 1.7, 0, 0, 0)

    enter image description here AGS00001bh (0.70, 2.2, 0, 0, 0)

    enter image description here AGS00001hn (0.70, 1.9, 0, 0, 0)

    And the three 'Star/artifact' objects:

    enter image description here AGS000005o (0.84, 1.7, 0, 0, 0)

    enter image description here AGS00001uo (0.85, 2.0, 1, 1, 0)

    enter image description here AGS00002ak (0.95, 2.3, 1, 0, 0)

    Posted

  • JeanTate by JeanTate

    QC data

    There are 37 QC objects with Petro_R50 ("R50") values < 1.0". In bins of 0.1":

    • < 0.6": 1
    • > 0.6" to < 0.7": 4
    • > 0.7" to < 0.8": 5
    • > 0.8" to < 0.9": 9
    • > 0.9" to < 1.0": 18

    Only three are classified as "Star or artifact" ( AGS00002x2, R50=0.62", AGS00003fj, R50=0.82", and AGS00003ag, R50=0.89"); six as "Features or disk", and 28 as "Smooth".

    The probPSF_x (x=g, r, i) and Petro_R90 ("R90") for the ten smallest - all with R50 < 0.8" - are as follows (R50, R90, probPSF_g, _r, _i):

    enter image description here AGS00002ds (0.56, 1.2, 1, 1, 1)

    enter image description here AGS00002x2 (0.62, 1.6, 0, 0, 0) - "Star/artifact"

    enter image description here AGS00002n6 (0.65, 1.2, 1, 0, 0)

    enter image description here AGS0000462 (0.65, 1.5, 0, 1, 1)

    enter image description here AGS00003jv (0.67, 1.4, 1, 1, 1)

    enter image description here AGS00004fz (0.71, 1.4, 0, 0, 0)

    enter image description here AGS00004ig (0.73, 40.4, 0, 0, 0)

    enter image description here AGS00004ac (0.74, 7.2, 0, 0, 0)

    enter image description here AGS0000445 (0.76, 30.0, 0, 0, 0)

    enter image description here AGS00002ud (0.77, 6.3, 0, 0, 0)

    And the two other 'Star/artifact' objects:

    enter image description here AGS00003fj (0.82, 4.6, 0, 0, 0)

    enter image description here AGS00003ag (0.89, 1.9, 0, 0, 0)

    Posted

  • JeanTate by JeanTate

    A quick check on the other objects with R50 <~1", to note those which are obviously - often wildly - wrong.

    QS

    enter image description here AGS00000ji (0.82, 5.35, 0, 0, 0)

    enter image description here AGS00001se (0.85, 15.2, 0, 0, 0) - fg star?1

    enter image description here AGS00000y2 (0.97, 2.77, 0, 0, 0)

    QC

    enter image description here AGS0000352 (0.84, 22.9, 0, 0, 0)

    enter image description here AGS00004bm (0.85, 4.9, 0, 0, 0)

    enter image description here AGS000048a (0.92, 10.7, 0, 0, 0)

    enter image description here AGS0000470 (0.92, 6.92, 0, 0, 0)

    enter image description here AGS00003hq (0.96, 2.7, 0, 0, 0)

    enter image description here AGS000031r (1.0, 11.5, 0, 0, 0)

    1 No. I wrote this up as a post in the GZ forum, here. The DR10 image is much clearer than the DR8 one:

    enter image description here

    Posted

  • JeanTate by JeanTate

    Some seem, well, odd.

    Consider this triple, all QS, with almost identical values of R50:

    enter image description here AGS00001ng (0.981")

    enter image description here AGS000018b (0.982")

    enter image description here AGS00001ho (0.984")

    And this pair from QC:

    enter image description here AGS00003x7 (0.943")

    enter image description here AGS000038g (0.943")

    Posted

  • JeanTate by JeanTate

    Test of the visualization technique described here, using the 100 smallest QC objects (as measured by Petro_R50). The images are DR10, and the scale is 0.4:

    AGS00002dsAGS00002x2AGS00002n6AGS0000462AGS00003jvAGS00004fzAGS00004igAGS00004acAGS0000445AGS00002ud
    AGS00002qzAGS00003fjAGS0000352AGS00003oqAGS00004bmAGS00004m3AGS000044oAGS00003agAGS000037jAGS00004gz
    AGS000048aAGS00002hqAGS0000470AGS00003w4AGS00003hqAGS00003x7AGS000038gAGS00002q1AGS00002v3AGS00004d5
    AGS00004c3AGS000030bAGS00002sqAGS00003b5AGS0000475AGS000047oAGS00002r4AGS00002drAGS00002v5AGS000031r
    AGS00002v2AGS00002y7AGS0000304AGS00004jdAGS00002y4AGS000034bAGS000044cAGS00004h8AGS00002n0AGS00004ho
    AGS00003ctAGS00003k2AGS00002f6AGS00003x0AGS00002z3AGS00004lnAGS00003w2AGS000034qAGS00002r0AGS00002nr
    AGS00002qgAGS00004l3AGS00002jbAGS00003ozAGS00002fhAGS00003c0AGS0000319AGS00003k3AGS00002ogAGS0000388
    AGS00004diAGS000038bAGS00003d1AGS00002t3AGS000045mAGS00003l0AGS00002kfAGS00003uaAGS00003rkAGS000030x
    AGS00004bgAGS00002d7AGS00003esAGS00003e3AGS00002q4AGS00004dnAGS00003izAGS000041oAGS00003pnAGS000046p
    AGS00003m5AGS00002hzAGS00002zeAGS00003wrAGS000035yAGS000043oAGS000032qAGS00002saAGS00002nhAGS00004kc

    By eye, I'd say the numbers of catastrophic failures, in each row of ten, are (from top, smallest, to bottom):

    3, 2, 1, 1, 2, 0, 1, 1, 0, 1.

    Here are the same 100, with a scale of 0.2:










    UPDATE: I added the AGS IDs as the 'alt text' in the image URLs of the first set of 100. This should make it much easier to identify any particular image/object, so making it easier to do follow-ups on apparent outliers/anomalies.

    Posted

  • JeanTate by JeanTate

    A quick cross-check: my detailed look at the ten smallest QC objects, brief eyeball of the other 27 (<~1"), and the 100 60x60 pix images:

    • AGS00002n6 and AGS00004ac (both in the first 10, both co/m1) not among the 3/10 in the first row of 'the 100'
    • the other three outliers in this first ten were identified in 'the 100'
    • AGS0000352, AGS00004bm, AGS000048a, and AGS000031r are the 2/10, 1/10, and 1/10 in rows 2, 3, and 4 (respectively); all are noted in both the 'brief eyeball' and in 'the 100'
    • AGS0000470 and AGS00003hq are among 'the 27', and noted in the brief eyeball, but not in the 100.

    The five others I noted in the 100 are (AGS ID, R50, R90):

    enter image description here AGS00002y7 (1.02, 36.6)

    enter image description here AGS000034b (1.03, 5.1)

    enter image description here AGS00002qg (1.07, 3.0)

    enter image description here AGS000045m (1.1, 8.6)

    enter image description here AGS00002nh (1.14, 3.46)

    One other, which I was of two minds on:

    enter image description here AGS00003k3 (1.09, 4.6)

    1 complex overlap/merger

    Posted

  • JeanTate by JeanTate

    The smallest 100 QS objects, per DR10; scale=0.3:

    AGS00000mtAGS00001z3AGS00001lzAGS000021rAGS00001egAGS00001fyAGS00000wqAGS000028iAGS00001bhAGS00001hn
    AGS0000038AGS00000toAGS00001koAGS000002eAGS00000vyAGS00001chAGS00000j6AGS00001w6AGS00000jrAGS00000ds
    AGS00001lmAGS000017vAGS00000nzAGS00001fiAGS00001sfAGS00001d7AGS00000w5AGS000012qAGS00000wnAGS00000vz
    AGS0000152AGS000026mAGS00000k6AGS00001g2AGS00001thAGS00002apAGS00001ltAGS000027oAGS00000pjAGS00000ji
    AGS000025kAGS000014wAGS000005dAGS00000u0AGS00000w7AGS000022xAGS00001xkAGS000013tAGS00001vpAGS000005o
    AGS00001jhAGS00001seAGS00001nxAGS00001vsAGS00000o0AGS00001uoAGS00000neAGS00000upAGS00000grAGS0000290
    AGS000022cAGS00000iaAGS00000y3AGS00000iyAGS00001bdAGS0000253AGS00000rtAGS00001kxAGS000026aAGS00000fd
    AGS000022kAGS000020yAGS00000vxAGS00001x9AGS000018iAGS00001elAGS00001eqAGS00001wpAGS000025xAGS00001ov
    AGS00001p1AGS00001dhAGS000019jAGS00000nfAGS000022fAGS00000p4AGS000019bAGS00001c7AGS00000o9AGS00001m8
    AGS00001l9AGS00001xbAGS00001c4AGS00000n6AGS000024lAGS00001jdAGS000005yAGS00001jyAGS000022yAGS000021q

    Posted

  • JeanTate by JeanTate

    Big galaxies with small cores

    Consider the QC object AGS00003hq:

    enter image description here

    It has an R50 value of 0.96 and an R90 one of 2.7. Somewhat simplified, this means that, in the r-band, it is 2.81 times bigger at the 9th decile than at the 5th ... or that the ratio of R50 to R90 is 0.356. This last number is called the 'inverse concentration ratio' (in the r-band), and it is a parameter used to determine if a galaxy is more likely to be a spiral (disk) or an elliptical (no disk)1.

    0.385 is, as the value of the r-band inverse concentration ratio, a good first choice, to distinguish disk galaxies from ellipticals. So AGS00003hq is, very likely, an elliptical. But how low can values of this ratio go? For a truly on-its-own galaxy - and not a merger or overlap, or a clump or region of a larger galaxy - what is a reasonable minimum for this parameter? If we have such a minimum, it would be possible to separate apparently anomalous objects (i.e. on-their-own galaxies which have R50 values that seem too low) from (very likely) anomalous ones.

    I haven't checked the relevant literature (again), but in my Huds research, I have not found any galaxy2 with an inverse concentration ratio less than 0.21. So a working value, for the minimum, might be 0.20.

    How do the galaxies2 I've noted as having anomalous values of R50, in this thread, fare, when measured using the inverse concentration ratio ruler?

    Well, here's one that remains anomalous AGS00003fj (R50 is 0.82, R90 4.6; the ratio is 0.18):

    enter image description here

    And another, AGS000031r (1.0, 11.5; 0.09):

    enter image description here

    Which confirms what you probably would have guessed anyway ... being close to a really bright star can mess with the photometric pipeline's head conclusions.

    But what about this one AGS000045m (1.1, 8.6; 0.13)?

    enter image description here

    Is there some sort of overlap (the odd feature crossing the disk near its top tip), or is that an artifact? Is the proximity to the bright star enough to mess up the estimate of R90? Do highly elongated Eos - especially if they have a tiny (or non-existent) bulge - have really small inverse concentration ratios?

    1 I used this parameter to help me decide if my Huds (huge distant spiral) candidates were, in fact, spirals. See Discovery of Four Giant, Distant Spirals

    2 on-it-own, not merging, not an overlap, and where R50 and R90 refer to the whole galaxy (not a clump, or region, within a galaxy)

    Posted