Galaxy Zoo Starburst Talk

Quench Sample - BPT diagram

  • JeanTate by JeanTate

    My first attempt to create a BPT diagram, using Tools, was unsuccessful (Automated creation of BPT diagrams, in GZ Tools?). I will try again later.

    In another thread, I asked about a recent paper The Cosmic BPT Diagram: Confronting Theory with Observations - relevant to us? (Kewley et al. 2013). Louise93 kindly that the standard z=0 line should be fine. That made me realize that I didn't know what the standard line(s) is (are)! I found a PDF by Carl Ferkinhof, Distinguishing Starburst from AGN: Application of the BPT diagram.

    From the Quench Sample, removing the outliers (so far) and objects with zero or negative Halpha, Hbeta, Oiii, or Nii flux, leaves 2779 (of 3002). Here's a plot1

    enter image description here

    The orange line is Kewley et al. (2013) for z=0.1 (the mean redshift of the 2779 is 0.108); the green is from Kewley et al. (2001); the brown Kaufmann et al. (2003).

    To the left of the orange/green line, the objects are dominated by star-formation.

    To the right of brown line, the objects are dominated by AGNs.

    In between? "Composite"! 😃

    Of course, more work needed (e.g. identifying and removing other outliers), but it's interesting that the QS contains a nice mix of pure starforming, composite, and AGN-dominated objects ...

    1 The axes are logs (I need to re-label them)

    Posted

  • jtmendel by jtmendel scientist, moderator

    Very nice!

    I'm a bit surprised at the abundance of star-forming/composite values in the quenched sample: my presumption would be that the selection would be more biased against such objects -- hence their "quenched" label -- but I don't know the details of it enough to say.

    Have you put any cuts on the signal-to-noise in the different emission lines when you made this figure? This can also have some effect on which regions of the digram are populated.

    Just a quick note that I think the references for BPT demarcations might have gotten swapped (green looks like Kauffmann et al. 2003, and brown looks like Kewley et al. 2001).

    Posted

  • JeanTate by JeanTate in response to jtmendel's comment.

    Thanks!

    No, no cuts (other than (NOT outlier) AND (X_flux > 0, where X = Halpha, Hbeta, Oiii, Nii)). In the next iteration, I'll add S/N cuts.

    Re the lines: you may well be right; I was focused on getting a formula that I could use, and did not check the source(s) for accuracy!

    Posted

  • mlpeck by mlpeck

    In all of the SDSS related literature I've read the threshold for detection of an emission line is S/N > 3, so until we have flux error estimates making BPT diagrams is useful for practice only, I'm afraid.

    BTW a fairly recent GZ paper introduced a criterion for dividing AGN from "LINERS". I'll try to find the reference sometime soon.

    Posted

  • JeanTate by JeanTate in response to mlpeck's comment.

    Thanks.

    Practice is good! 😃 I'll bet most of the "BPT outliers" that I've been posting this afternoon will turn out to have poor S/Ns ...

    Posted

  • JeanTate by JeanTate

    Here's a BPT diagram I created using Tools; hope you can see it:

    http://tools.zooniverse.org/#/dashboards/galaxy_zoo_starburst/520647fe0aab2a3a980001fd

    Posted

  • wassock by wassock moderator in response to JeanTate's comment.

    Hi Jean - So if we could take out all the non-star forming galaxies we'll be left with a data set which closely tracks the orange/green curves?

    BPT? Probably a term that could be added to the Glossary?

    Posted

  • JeanTate by JeanTate in response to wassock's comment.

    Yes, we do need an explanation (I'll write one "later today", unless someone else beats me to it).

    My plot is just a first go, so there's not much you can read into it, for all sorts of reasons. However, once a decent one in made (see below), galaxies to the left of the green/orange line(s) - shouldn't matter which - will (very likely, p>0.99?) be dominated by star-formation. Between that line and the brown (or whatever color you call it), both AGN-like and star-formation contribute to the emission lines, hence "composite". To the right of that line, the emission lines are caused by an AGN (usual caveats).

    "Once a decent plot is made ...":

    • if what the spectrum is of is merely a clump in a galaxy that's much
      larger than 3", you can't really say much about the galaxy (see this
      thread for more: Is the aperture that fed the SDSS spectrograph
      fibers 3" in diameter, or 3" in
      radius?      )
    • if any of the four em line fluxes (Halpha, Hbeta, Oiii, Nii) are wrong or highly uncertain (one reason why we need the corresponding _Err values!), the classification (star-forming, composite, AGN) will be equally wrong or highly uncertain1
    • even if the line fluxes look good, the galaxy's size is ~<3" (or it's not a clump), and it's not a star, if the spectrum is of an overlap or a 'violent merger', the classification could be quite wrong
    • I've noticed that spectra with nice strong Balmer series abs lines (the "A" in "E+A") often have very noticeable mismatches between the red 'Best Fit' line and the black 'data' line, in and around these lines; there may be a systematic effect here that we (or rather, I) haven't considered ...

    1 we may be able to do some analyses with these, but it won't be easy, or pretty!

    Posted

  • JeanTate by JeanTate in response to jtmendel's comment.

    Update: my 'sample size' is still quite small, but I have found several galaxies - with decent fiber covering fractions, apparently quite high S/N ratios, etc - that have nice 'A-type' Balmer absorption lines ... yet Hbeta_flux (and even Halpha_flux) is recorded - in the QS catalog - as an emission line.

    For example, AGS00000v1 (587732701254779013): in the catalog, Halpha_flux is 14.1, and Hbeta_flux is 2.2 (and Petro_R50 is a mere 1.0"), yet both are clearly abs lines (spectrum, interactive):

    enter image description here

    Posted

  • mlpeck by mlpeck in response to JeanTate's comment.

    JeanTate:

    This is a good example of why the lack of flux error data is not at all helpful. If you look in the galSpecLine table for this object it lists the hbeta flux as 2.2 with a flux error of 2.1, in other words no emission is detected. It lists the halpha flux as 14.1 with flux error of 4.26, which makes it a very marginal (3.3 sigma) detection.

    The SDSS pipeline on the other hand, which is what is shown in the interactive spectrum page, lists H_beta as -9 (no emission) and H_alpha = +31.2.

    The web page doesn't show errors for the flux estimates, but it's 14.6 for H_alpha (in other words, no detection), and 8.5 for H_beta which also means not significantly different from 0.

    Basically this object shouldn't even show up in a BPT diagram. At most there might be some very weak LINER-like emission and even that depends on which data processing pipeline you believe.

    Posted

  • JeanTate by JeanTate in response to wassock's comment.

    "Later today" ended up being "in three days' time", but at last it's done: What is a "BPT diagram"? Why do astronomers like to use them?

    Comments - especially those pointing out glaring mistakes - very welcome! 😃

    Posted