Sample Selection: Post-quenched galaxy and control galaxy
-
by
trouille
scientist, moderator, admin
Check out this discussion thread for an overview on the process we followed for selecting our post-quenched galaxy and control galaxy sample. http://quenchtalk.galaxyzoo.org/#/boards/BGS000000a/discussions/DGS00001xk
Post updated on December 13, 2013 to bring information buried in this thread up to the forefront:
Response about sample selection based on emails with Yanmei Chen, now a faculty in astronomy in China (who recently had a baby girl). She was Christy Tremonti’s postdoc at UW-Madison when she did the sample selection.
http://en.wikipedia.org/wiki/Principal_component_analysis helps provide an overview of this method.
In brief, the method looks for features (emission and absorption lines, size of the 4000 Angstrom break, etc.) in the galaxy spectra that indicate that they're a member of this special sub-group of galaxies.
The features the program looks for are based on those described in http://postquench.blogspot.com/2013/06/wong-et-al-article-galaxy-zoo-building.html. But it takes this to the next level, by creating a whole range of model spectra with different star formation history, time since quenching, rate of quenching, stellar population models, metallicity, etc. and correlates the SDSS galaxies against all the spectral components of these model spectra.
The power of the principal component analysis is that it uses many spectral features to determine if a galaxy is or isn't a post-quenched galaxy.
In detail:
First we create a set of galaxy model spectra. Details about the model spectra are provided at the end of this post.
We apply the principal component analysis (PCA) method to our set of galaxy model spectra. This process allows us to create eigenspectra.
In our PCA analysis, we use the restframe 3750-4150Ang wavelength range. At wavelengths around 4000Angstrom, galaxy spectra vary in both spectral shape (4000Ang break) and strength of the hydrogen Balmer absorption lines. Both of these are very sensitive to the recent star-formation history.
Figure 1 shows the mean spectrum of the model library and the first seven eigenspectra (PC1-PC7). PC1 corresponds to the continuum shape. PC2 shows the Balmer absorption lines, PC3 contains Ca II (H+K).
Link to Figures 1 and 2: https://vault.it.northwestern.edu/let412/GZQuench/Talk_Images/fig1and2.pdf
We then project the SDSS DR7 spectra through the eigenspectra. Through this projection, we get the strength of the eigenspectra for each spectrum (i.e., for each galaxy). We call the strengths of each eigenspectra PC1, PC2 ..., which stand for Principal Component 1, Principal Component 2…
Figure 2 is an example of the projection of the continuum fit. Black is a DR7 spectrum in the range of 3750-4200Angstrom, green is our best fit using our Eigenspectra.
Our sample selection is based on PC1 and PC2.
Figure 3: Figure 3 shows how we select the post-quenched galaxies using PC1 and PC2. Our post-quenched galaxies are the red dots in the figure. The overplotted lines show different models for how quickly the star formation is quenched in different models (tau = 100, 300, and 500 Myr, from top to bottom). The left bottom point of each of these model lines are where galaxies lie whose star formation has just started. As a galaxy evolves over time, it moves to the right along those model lines. The red part of the lines are where the models’ ages are between tau + 0.2Gyr to 2 Gyr.
The PSB selection criteria (sel = where(pc2 gt a*pc1+b and pc1 lt -1.8), with a=-0.146667 and b=0.33332) is built according to the \tau = 500Myr model and Yanmei’s comparison with the selected post-quenched galaxy spectra with S/N > 15.
More details are as follows:
(1) model library: Our input data set for the creation of the PCA eigenspectra is a set of model spectra generated using the Bruzual & Charlot (2003) BC03 stellar population synthesis code. The model library is similar to that used in Kauffmann et al. (2003) and Salim et al. (2005), although with a more restricted parameter range:
(i) The time tform when the galaxy begins to form its stars is distributed uniformly between 0 and 5.7 Gyr after the big bang (the age of the universe is assumed to be 13.7 Gyr)
(ii) The model galaxies have exponentially declining star formation histories
SFR~exp(−t/tao) with tao distributed uniformally between 1~2Gyr(iii) Top-hat bursts of star formation are superimposed on these continuous star formation models. Two parameters describe the bursts: fburst, the fraction of the total stellar mass formed in bursts, is distributed logarithmically between 0.0 and 0.1; tburst, the duration of the burst, is distributed uniformly between 0.03 and 0.3 Gyr.
During the burst, stars form at a constant rate. Bursts occur with equal probability at all times after tform and the probability is set so that 50 per cent of the galaxies in
the library have experienced a burst over the past 2 Gyr. Finally, the fraction of galaxies with ongoing starbursts is reduced to 25 percent of those initially created.(iv) The metallicity is distributed linearly in the range 0.5~2Zsolar. No metallicity
evolution is included.As you can see, there's no selection for/against AGN activity within our post-quenched galaxy sample. That was something we were excited about -- that this sample selection doesn't exclude AGN. That way we can actually consider AGN activity as a potential mechanism for shutting off star formation in these types of galaxies.
Other selection methods for post-quenched galaxies, particularly those based on [OII] equivalent width (EW), excluded AGN. People used EW[OII] LT some value because it was a way to identify sources that didn't have current star formation ([OII] is a good star formation indicator). However, AGN activity can also boost your [OII] emission, so you can have a galaxy with little current star formation but with AGN activity that has a high EW[OII] that would be excluded from previous post-quenched galaxy selection.
https://vault.it.northwestern.edu/let412/GZQuench/yan_2009_psbEnviro_DEEP2.pdf provides a nice description of why we don't base our post-quenched galaxy sample selection on emission lines like [OII].
Post updated on December 20, 2013 to bring additional information from Yanmei up to the forefront:
In 2012, Yanmei (one of the Quench science team members) published an article using PCA analysis to estimate stellar masses, ages, and star formation histories of SDSS galaxies. To share the results and method with the public, she created a schema website within the SDSS server that includes those eigenspectra values used in the article.
There is a great question in a QuenchTalk post about whether we can use that SDSS website and those eigenvalues to recreate the post-quenched galaxy selection that Yanmei provided for us.
Yanmei wanted to respond to this question directly.
She explained: In principle, you can use Calpha_0 (correlates D4000) & Calpha_3 (correlates to Balmer absorption) from that website to select post-quenched galaxies as she did for our project. However, she recommends against this route because the eigenspectra on that website were built for calculating galaxy parameters like stellar mass, velocity dispersion, etc. They weren't optimized for selecting post-quenched galaxies.
The main reason they aren't optimized for post-quench galaxy selection is that those eigenspectra in that website are based on 3700-5500 angstroms in wavelength. The information above 4200 angstroms does not help in identifying post-quenched galaxies. This causes the eigenspectra related to Balmer absorption (a key feature in post-quenched galaxy selection) to drop down in importance in the principal component analysis listing (from the number 2 spot to the number 4 spot). This is not what we want because it degrades the post-quenched galaxy selection robustness.
Instead, she advises we use the eigenspectra she created specifically to select post-quenched galaxies, which was based on using 3700-4200 angstroms in wavelength. She has also provided us the relevant files to recreate the sample selection ourselves.
This text file lists the pc1, pc2, pc3, pc4, pc5, pc6, and pc7 for the 927,552 objects in the SDSS-DR7 catalog released by MPA/JHU.. You'll want to download the gal_info file at this website to get all the information (RA, Dec, z, ect) for the 927,552 SDSS-DR7 objects. The order in that gal_info file matches the order in the psb_dr7.txt file.
Stellar masses for the 927,552 galaxies can be gotten here.
From those principal component values, she chose a very inclusive sample of almost 14,000 potential post-quenched galaxies. The information for these galaxies is here.
These ~14,000 objects were chosen based on the criteria listed earlier in this post (i.e., sel = where(PC2 GT 'a' times PC1 + 'b' and PC1 LT -1.8, with a=-0.146667 and b=0.33332). Figure 3 above visually shows the PC1 vs PC2 selection and the text that follows the figure explains a bit more in detail.
From this sample of ~14,000 objects, we then set a Signal-to-Noise cut of S/N_median GE 10. (Note: GE = greater than or equal to). This cut the sample down to ~3060 objects. We then got rid of duplicates to end up with our final sample of 3002 post-quenched galaxies, which we placed on Tools.
With this information, you can recreate the sample selection as well as play with different S/N limits, compare other samples of galaxies with the PC1, PC2, etc., and any other tests to see the impact on our results.
Posted
-
by
zutopian
In the topic "Spectrum, which doesn't look like post-starburst! QS galaxy", which was started by me, mlpeck wrote following.:
30% of the quench sample objects are, spectroscopically, starforming. I examined the statistics of the starforming subset in some detail here. The percentage of starforming objects is virtually the same in the quench and control sample.
Star forming galaxies in the quench and control samples started by mlpeck:
http://quenchtalk.galaxyzoo.org/#/boards/BGS0000008/discussions/DGS0000204
Posted
-
by
zutopian
Topics started by Jean:
-
What is needed to get a clean QS and a clean QC?
http://quenchtalk.galaxyzoo.org/#/boards/BGS0000008/discussions/DGS00001zr -
Outliers - summary
http://quenchtalk.galaxyzoo.org/#/boards/BGS0000008/discussions/DGS00001yg
Topic started by mlpeck:
- Towards a clean control sample
http://quenchtalk.galaxyzoo.org/#/boards/BGS0000008/discussions/DGS00001yo
Posted
-
-
by
zutopian
News: mzevin wrote in another topic today.:
I'm working to get the specifics of the selection criteria - I should have that info fairly soon. I will keep you posted.
Posted
-
by
zutopian
New topic by Jean:
Redshift and size cuts: Proposal and discussion
Posted
-
by
zutopian
I copied following post, which mlpeck did in another topic today.:
Wong et al.'s selection criteria effectively excluded AGNs, since they required weak Hα emission (SN<4).
I found one object in common between the Wong et al. sample of 80 and the quench sample: AGS000022s. That was also the only object in QS that met their stated selection criteria. I found no objects in the control sample.
Here's the DR10 explore page. Notice in the spectrum that Hα was masked but there are other strong emission lines, so it's questionable whether it really belonged in their sample.The GZQ sample contains some AGN galaxies. Why weren't they excluded? Wong et al. required weak Hα emission, but I guess, that GZQ didn't. If so, why not? I cite the from Wong et al paper.:
The [OII] forbidden lines were not used in the selection criteria because the H alpha emission line is a more accurate tracer of current star formation in the Local Universe.
http://arxiv.org/abs/1111.1785I guess, that the following comment by mlpeck might be somehow related.:
30% of the quench sample objects are, spectroscopically, starforming. I examined the statistics of the starforming subset in some detail here. The percentage of starforming objects is virtually the same in the quench and control sample.
Is this the case, because GZQ maybe didn't require weak Halpha emission?
I wonder, why there is just one "object in common between the Wong et al. sample of 80 and the quench sample"? I wonder, what the difference between the selection criteria of both samples is? I cite from the Wong et al paper.:
To minimise the Malmquist bias, and create a volume and magnitude-limited (proxy for stellar-mass limited) sample of galaxies, we select all the galaxies within 0.02 < z< 0.05 with Mz,Petro< −19.5 magnitudes.
http://arxiv.org/abs/1111.1785What is the Malmquist bias? In a wikipedia article there is given.: "Anytime a magnitude-limited sample is used, one of the methods described above should be used to correct for the Malmquist bias."
The GZQ sample contains many galaxies with redshift greater than 0.05. What is the redshift selection range of the GZQ sample and the selection criteria for MzPetro?
UPDATE: There is not just one object, but 2 ones in common between the Wong et al. sample of 80 and the quench sample.
Posted
-
by
zutopian
Referring to my previous post.:
I copy another comment by mlpeck.:
30% of the quench sample objects are classified as starforming by their position in the classic BPT diagram. See here and here. This is almost identical to the percentage of starforming objects in the control sample.
This gets directly to the sample selection. Does this mean that the selection criteria resulted in a significant fraction of false positives? Or are the quenched starformers actually quenched relative to _ (fill in the blank here)?
I see some evidence that the quench sample starformers are in fact edging their way into the "green valley" -- scroll down through the second thread and also see here.
http://quenchtalk.galaxyzoo.org/#/boards/BGS0000001/discussions/DGS000021d
Posted
-
by
mzevin1
scientist, moderator
From my understanding of the selection criteria (which, I should add, I did not have a direct role in) this project's criteria was a good deal different than previous studies. This is also the reason why we call the galaxies in question 'post-quenched galaxies' instead of the commonly used terminology 'post-starburst galaxies'. With previous studies, galaxies that were still starforming or had AGN signatures were often excluded by the selection criteria, even if they had signs of a period of intense star formation that was 'quenched' in their pasts.
Posted
-
by
zutopian
in response to mzevin1's comment.
There is following related discussion topic.:Post-Quenched" = "E+A" = "K+A"?
http://quenchtalk.galaxyzoo.org/#/boards/BGS0000001/discussions/DGS0000006
There is said.:
The spectra for the galaxies in our sample give the indication that they had star formation within the past few hundred million years, but they're not currently forming new stars.
In the Wong et al paper is said following.:
The [OII] forbidden lines were not used in the selection criteria because the H alpha emission line is a more accurate tracer of current star formation in the Local Universe. http://arxiv.org/abs/1111.1785
mlpeck says "30% of the quench sample objects are, spectroscopically, starforming".
Do those 30% QS galaxies currently form new stars or not?
Posted
-
by
mlpeck
in response to mzevin1's comment.
this project's criteria was a good deal different than previous studies. This is also the reason why we call the galaxies in question 'post-quenched galaxies' instead of the commonly used terminology 'post-starburst galaxies'.
Going back to the top of this thread and following a few links we eventually get a mention of Yan-Mei Chen and principal components, which with a bit of digging on arxiv/ADS leads to Chen et al. 2012.
Starting from simple stellar population models from Bruzual & Charlot or Maraston Chen et al. generated tens of thousands of star formation histories, each of which produces a synthetic spectrum. They took pieces of those spectra with rest frame wavelengths between 3700 and 5500 Å, masked off all emission lines which means all Balmer lines in that wavelength range plus lines of [O II], [O III] and [Ne III], and calculated 7 principal components from those segments of model spectra. The claim is that those 7 PCs adequately span the multidimensional (~1500) space of model spectra, and furthermore that they adequately span the space of real galaxy spectra.
I don't see a mention of this application in the paper, but presumably one could select interesting sets of objects from the positions of their spectra in that 7 dimensional space. For example some fraction of the models have recently quenched star formation, and those have to occupy some region, possibly a compact set, in PC space.
I'm guessing that principal component scores (and most likely no other criteria) were used somehow to select the quench sample. I haven't figured out exactly how they were used, but I have deduced a few properties of the PC scores:
-
They're highly predictive of important indices in that part of the spectrum, specifically Hδ equivalent width and D4000. This is discussed in their paper.
-
They can effectively separate red sequence spectra from everything else. There are no "red and dead" objects in the quench sample.
-
They have little or no ability to predict position in a BPT diagram. This is not surprising given that emission lines were masked out of the input data.
Posted
-
-
by
JeanTate
There's another aspect to the selection process which I think is worth spending some time on; namely, the objects.
It's likely that the key inputs to the selection process are the DR7 spectra, specifically the SpecObjs. Each is the output of the SDSS DR7 spectroscopic pipeline, the 'upstream source' for each being the light that fell on the 3" diameter fibers that fed the spectrographs.
A common assumption, starting with the spectroscopic pipeline, is that the light which fell on each fiber comes from only one kind of astronomical object, a galaxy, star, supernova, or (at least once) an asteroid or comet. In reality - as we know in the Quench project - many objects are composites: star+galaxy, galaxy+galaxy, QSO+star, supernova+galaxy. So, surely a first step in this project should be to identify and exclude all such composites (objects mistakenly classified as 'galaxy' by the pipeline - but which are stars - should also be identified and excluded).
A separate definitional issue is 'galaxy'. For many QS objects, the covering fraction is > 20% (as measured by luminosity), and there is support in the literature for assuming such spectra are sufficiently representative of the whole galaxy for analyses involving extinction, star-formation, and metallicity (thank you jtmendel! 😃). But we surely want just one spectrum per galaxy, don't we? If so, we need to exclude one of the QS-QS 'duplicate objects'.
Perhaps the most interesting aspect of 'what is a galaxy?' for our purposes is whether a merger is one galaxy or two (or more) 😮 Among the ~700k galaxies classified by zooites in GZ1, the fraction classed as 'merger' is tiny, barely a percent or two; however, in the QS catalog, the fraction is almost 10% (counting just consensus votes for 'merging' and 'both'). In many of these, whether a covering fraction of 20% is sufficient is surely unknown ... after all, 'merger' is not a Hubble type, and Kewley et al. (2005) did not include any mergers in their analyses. And then there's the question of what a 'nuclear spectrum' might be for many mergers, given that there's often no obvious nucleus, or more than one nucleus, or that the fiber collected light from a part of the merger far from any nucleus.
Of course I do not know, but I doubt many - if any - of these considerations went into the QS catalog selection process.
Posted
-
by
JeanTate
in response to zutopian's comment.
There are quite a few questions in this post of yours, zutopian, which do not seem to have been answered yet.
Are you - or any other reader - still interested in answers?
Posted
-
by
JeanTate
in response to JeanTate's comment.
I have just re-read the Wong et al. (2012) paper. I could find no mention of any of the issues I raised in my earlier post (two upstream).
There are images of 12 (of 80) PSGs in their paper. Several seem to have covering fractions below 20%, but none appear to have the fiber centered on anything other than the obvious nucleus. While some of these 12 appear somewhat disturbed, none are as disturbed as many of the obvious mergers in the QS catalog. I'll look at the other 68 later; if all have centered-on-the-nucleus fibers AND none appear as disturbed as AGS000048a1 or AGS00000wq (say), then my questions (re merger/definition of galaxy) will be moot (composite and star questions might still apply, of course).
1 UPDATE: AGS000048a is, as zutopian noticed, a QC object, not a QS one. So the challenge of deriving parameters-from-SDSS spectra is not limited to QS objects. There are, of course, plenty of other QS examples; e.g. AGS00001rm
Posted
-
by
JeanTate
in response to JeanTate's comment.
There are some which seem to have covering factions < 20%, and certainly some are disturbed and/or asymmetric. However, none are at all like AGS000048a or AGS00000wq. Perhaps the closest is 587741722823754043; the fiber does not seem to coincide with the nucleus (the only obvious case I found), and it's clearly a merger (there are certainly other mergers too).
For overlaps - star/galaxy or galaxy/galaxy - 587733603187556413 is perhaps the most extreme, though it's not clear how much of the foreground star is in the galaxy's spectrum (there is certainly some):
Posted
-
by
zutopian
in response to zutopian's comment.
And below there is a related question by Jean. However, she did the post in the GZ forum. The post contains information about the selection criteria for the Wong et al sample .So I copy it to this topic.:
What is a "therms" level?
As in "As we aim to study the properties of all galaxies that have ceased star formation recently, we define a PSG to be a galaxy with a recently truncated star formation history (i.e. where the observed Hα emission line is weaker than four times therms level), while still exhibiting strong Balmer absorption lines from recently formed young stars (where the Hδ equivalent width is wider than 3 Å)."
The source is Wong et al. (2012), "Galaxy Zoo: building the low-mass end of the red sequence with local post-starburst galaxies".
http://www.galaxyzooforum.org/index.php?topic=281421.msg643743#msg643743
BTW, in a post by mlpeck, he said that Wong et al. "required weak Hα emission (SN<4)."
EDIT:
As I asked before, I wonder, if the 30% starforming GZQ sample galaxies are currently starforming?Posted
-
by
zutopian
in response to mzevin1's comment.
Here is a paper about "post-starburst with AGN" galaxies.:
Post-starburst--AGN Connection: Spatially Resolved Spectroscopy of Hdelta-Strong AGNs
840 galaxies with both a post-starburst signature (strong Balmer absorption lines) and an AGN (based on the emission line ratio).
Tomotsugu Goto (JAXA)
(Submitted on 8 May 2006)
http://arxiv.org/abs/astro-ph/0605202In the paper there are presented three of the 840 galaxies, which the author had selected. One of those is also in the GZQ sample.: AGS000009g
The other two aren't, if I checked right.And here is a paper, where there is given a sample of "12,105 quenched galaxies in SDSS".:
Towards a physical picture of star-formation quenching: the photometric properties of recently-quenched galaxies in the Sloan Digital Sky Survey
J. Trevor Mendel (MPE), Luc Simard (HIA), Sara L. Ellison (UVic), David R. Patton (Trent)
(Submitted on 26 Nov 2012)
http://arxiv.org/abs/1211.6115Posted
-
by
zutopian
Referring to my previous post.:
In the Mendel et al. paper there is given "consider only those galaxies with z=<0.2, ensuring that Halpha stays blue-ward of significant night-sky emission at lambda > 8000Å" and a "lower redshift limit of z = 0.01".
As far as I know, the GZQ sample contains however also galaxies, which have redshifts greater than 0.2. Nonetheless, the GZQ sample contains just 3000 galaxies, but the Mendel et al sample contains 12.105 galaxies.
Besides the GZQ sample however contains just 1 galaxy from the Wong et al sample and not all AGN galaxies from the Goto sample.Posted
-
by
zutopian
in response to JeanTate's comment.
Yes, I am still interested and continued to post. 😃
Posted
-
by
jules
moderator
I am also interested in the answers. I have reached a bit of a halt until I know we have a clean dataset now.
Posted
-
by
JeanTate
in response to zutopian's comment.
Towards a physical picture of star-formation quenching: the photometric properties of recently-quenched galaxies in the Sloan Digital Sky Survey ... J. Trevor Mendel (MPE), ...
A MODERATOR, SCIENTIST who recently posted a very helpful comment here is "jtmendel". Google tells me that there is a J. Trevor Mendel who is a post-doctoral research fellow at the University of Victoria in Canada. His 'research' webpage says he's interested in "Tracing galaxy growth with the post-starburst population", among other topics.
Posted
-
by
JeanTate
in response to zutopian's comment.
Indeed. I worked out that it's a typo; it should be "the rms", and so "x times the rms level" is a synonym for S/N (signal to noise) ratio. I've written up my (embarrassing) discovery in the GZ forum thread. I also discovered a new class of object, and a new parameter, 'sturbed'; one PSG in Table 1 of that paper is described as being "Bisturbed", as opposed, I guess, to Monosturbed or Tristurbed :p
Posted
-
by
JeanTate
in response to zutopian's comment.
Thanks zutopian, and jules, for saying you're still interested in answers to the questions you included in earlier posts in this thread, zutopian. Some are certainly far beyond what I could even imagine attempting to answer, but I'll have a go at answers to the others.
The GZQ sample contains some AGN galaxies. Why weren't they excluded? Wong et al. required weak Hα emission, but I guess, that GZQ didn't. If so, why not?
I assume by "QZQ sample" you mean the Quench sample catalog, the one containing the post-quenched galaxies selected by a method which includes application of the PCA (principal component analysis) developed by Chen (the control - QC - sample will surely include some AGN, because it's a random1 selection from SDSS galaxies with redshifts, and those galaxies certainly include ones with AGNs). Even within QS, and even accepting something like the Wong et al. selection criteria, some "AGNs" may be selected. As is so often the case, however, it depends on definitions ... in a common version of the BPT diagram, there are just three regions, "star-forming", "AGN", and "composite" (I see that mlpeck once called this "transition", IIRC); in others, "AGN" is divided into "AGN" and "LINER". A LINER is a low ionization nuclear emission line region (galaxy); here is a good summary, from NED. A galaxy may have weak emission lines, and the ratios of their fluxes may put the nuclear emission line region in the "AGN" or "LINER" part of the BPT diagram. A higher resolution spectrum with greater S/N may show that some such "AGNs" are, in fact, not. In any case, there's no particular reason to exclude all AGNs from a selection of "post-quenched" galaxies; it is not yet known how activity associated with a nuclear SMBH is related to the quenching of star-formation in a galaxy, so excluding AGNs would pre-suppose an answer to a very interesting research question!
Is this the case, because GZQ maybe didn't require weak Halpha emission?
I don't know what the selection criteria were, for choosing objects to belong to the QS catalog; however, it seems pretty obvious that there was no explicit requirement - one way or the other - concerning the strength of H-alpha emission (or absorption). You'll have read mlpeck's excellent post, in which he tries to reverse engineer the selection criteria; he may well be spot on! 😃 However, he may be quite off the mark; unless and until we hear from the Quench Science Team, we cannot know for sure.
What is the Malmquist bias?
There are good resources on the internet which explain it; one is NED: The Classical Malmquist Bias. In simple terms, it's why you'd be utterly wrong to conclude what the most common kinds of star are, by using just the ~6,000 which are visible to the unaided eye: intrinsically bright stars can be easily seen, even if they're a very long way away; intrinsically faint stars may all be too faint to be seen at all, even if they are very close (this is, in fact, reality; the most common kinds of star are the "M dwarfs", the closest of which is just a few light-years from us, yet is far too faint to see without a telescope). "Volume and magnitude-limited sample" is a common approach used in extra-galactic astronomy to minimize Malquist bias (eliminating it entirely is ... extraordinarily difficult). If you're interested, I'll explain how this works, in more detail.
What is the redshift selection range of the GZQ sample
There's no explicit selection criterion, as far as I know; there's an indirect one that comes from the fact that, in DR7, very few (MGS2) galaxies with redshifts > ~0.35 have spectra.
and the selection criteria for MzPetro?
There is none, as far as I can tell.
The design of the Quench project has a very strong 'unbiased' aspect at its heart: every QS galaxy has a matched control galaxy, with 'the same' stellar mass and 'almost the same' redshift. This design feature was not, initially, implemented very well - there'd have been no QS-QC duplicates if it had been, to take just one example - but has subsequently been fixed. With 'matched pairs' design, the sorts of selection criteria Wong et al. used are unnecessary.
1 yes, it's not entirely random; to be selected a QC object must have a log_mass (per MPA-JHU) "the same" as that of a QS one, and a redshift that differs by < 0.02
2 Main Galaxy Sample (I think); up to DR7, there were basically two different spectroscopic projects, one that aimed to get as many spectra of galaxies whose r-band mag is < 17.7 (IIRC), and the other the get spectra of LRGs (luminous red galaxies)
Posted
-
by
JeanTate
in response to JeanTate's comment.
It goes without saying that the above is just my understanding, and it likely contains mistakes. I hope a SCIENTIST will be along soon to correct any mistakes I made.
Posted
-
by
mlpeck
in response to JeanTate's comment.
You'll have read mlpeck's excellent post, in which he tries to reverse engineer the selection criteria; he may well be spot on! 😃 However, he may be quite off the mark; unless and until we hear from the Quench Science Team, we cannot know for sure.
Indeed. That was more or less an "am I getting warm?" query that I hoped someone on the science team would decide to answer.
As of DR10 principal component coefficients from the Chen models are available through CASJobs or the explore pages. I was able to verify the statement that they are highly predictive of Hδ EW and D4000.
Posted
-
by
JeanTate
in response to zutopian's comment.
I think this is last currently open question, in this thread (at least, of those which I think I can answer):
Do those 30% QS galaxies currently form new stars or not?
The ongoing formation of stars in a galaxy is not like the spin of an electron, either one state (up, or on) or one other (down, or off); rather, it's more a question of 'what is the current star-formation rate?' It is usually expressed as the mass of new stars formed per year, expressed in 'sols' (this does NOT mean, say, 10 sols/year -> ten stars of mass ~the same as that of our Sun are formed each year).
So a 'red and dead' galaxy would have a star-formation rate (SFR) of "less than X", rather than "is not currently forming stars" (though that may be true too).
There's another parameter, the 'specific SFR', SSFR, which is the SFR per unit of galaxy stellar mass. This is more useful when comparing galaxies of different masses; after all, an SFR of 100 sols/year (say) is a very different thing in a galaxy of only ~10^7 sols (a dwarf galaxy) than it is in one of ~10^12 sols (a really big spiral, say).
So, from an analysis of the spectra of the QS objects, it may be possible to conclude that ~30% have SFRs of > Y (some threshold), or an SSFR of > Z (whatever these numbers might be).
Posted
-
by
zutopian
in response to mzevin1's comment.
In the Wong et al. paper there is following statement.:
It should be noted that we do not find many of the E+A galaxies found by Goto (2005, 2007) (hereafter known as the “G05” sample) from the SDSS DR5 catalogue because we detect strong Halpha emission in these G05 objects using the SDSS DR7 catalogue. Hence, these E+A galaxies appear to have current, on-going star formation and is inconsistent with our definition of a PSG. Of the overlap galaxies between our full galaxy sample and that of the G05 sample, we find that our method of using the Balmer absorption line strength finds every galaxy found via its absorption line equivalent width.
So actually Wong et al criteria was different than in previous studies.
I wonder, if there are G05 sample galaxies, which are also in the GZQ sample?
EDIT:
I didn't crossmatch the whole G05 sample with the GZQ sample, but I found a GZQ galaxy in the G05 sample. Some other G05 galaxies, which I checked, are not in the GZQ sample.
The G05 sample contains also galaxies with z greater than the redshift selection range of the Wong et al sample.: The mentioned GZQ/G05 galaxy has z=0.1. I should crossmatch the low redshifts galaxies in the redshift range of Wong et al.. to find out, if there are G05 sample galaxies, which were rejected by Wong et al, but might be in the GZQ sample. (I hope, that my post makes sense.)PS: I mentioned another paper by Goto in this topic before, but it is a different paper than the ones in above statement.
Posted
-
by
zutopian
in response to zutopian's comment.
I cross-matched some of the low redshifts galaxies in the redshift range by Wong et al.. to find out, if there are G05 sample galaxies, which were rejected by Wong et al, but are in the GZQ sample.:
I found following one.: AGS00001nt with z=0.038 .: Since 0.02 < z< 0.05, it is within the redshift range of Wong et al., but it isn't in the Wong et al. sample. So it is a G05 galaxy, which was rejected by Wong et al. Not all G05 galaxies were rejected by Wong et al., but I have to check, which galaxies are in both samples: Wong et al and G05.
EDIT: Here is one of the G05 sample galaxies (0.02 < z <0.05), which wasn't rejected by Wong et al.: http://cas.sdss.org/dr7/en/tools/explore/obj.asp?id=587726032236183669
I don't need to check, if it is also in the GZ-Q Sample. I know, that it isn't. As we know, there is however just one Wong et al sample galaxy in the GZQ Sample and that is a different one.
Finally, the Wong et al sample contains 80 galaxies. I wonder, how many galaxies within the redshift range by Wong et al. are in the G05 sample and how many of these are also in the Wong et al. sample? Wong et al. selected 0.02 < z< 0.05 with Mz,Petro< −19.5 magnitudes. I don't know, if Goto used a different MZPetro selection criteria than Wong et al..PS: According to the cited stament in my previous post, Goto used for his catalog SDSS DR5, hence G05 sample, but Wong et al. used DR7.
I think, that maybe some sample content differences might be due to different SDSS versions.: e.g. Maybe there was more spectra available in DR7 than in DR5.UPDATE: There is not just one object, but two ones in common between the Wong et al. sample of 80 and the quench sample, but it isn't 587726032236183669.
Posted
-
by
mlpeck
I wonder, if there are G05 sample galaxies, which are also in the GZQ sample? EDIT: I didn't crossmatch the whole G05 sample with the GZQ sample, but I found a GZQ galaxy in the G05 sample.
Out of 564 in the Goto (2007) catalog I count 290 that are in the quench sample, based on position matches. There's also one in the control sample.
Posted
-
by
zutopian
in response to mlpeck's comment.
I wonder, which G05 galaxy is in the control sample?
Posted
-
by
zutopian
I copied following post, which mlpeck did today, from another topic.:
I missed one in the quench sample -- AGS00000y2 is also in the Wong catalog.
And in the control sample AGS0000495 is in Wong's catalog.
Both of these have rather strong emission lines in their spectra.So there is not just one object, but two ones in common between the Wong et al. sample of 80 and the quench sample.:
AGS00000y2 and AGS000022s
Besides the control sample contains however following Wong et al sample galaxy.: AGS0000495Posted
-
by
zutopian
in response to mlpeck's comment.
I guess, that following post, which mlpeck did in this topic, is meant.:
Going back to the top of this thread and following a few links we eventually get a mention of Yan-Mei Chen and principal components, which with a bit of digging on arxiv/ADS leads to Chen et al. 2012. (...)
In another topic Jean did following comment about the spectrum of a QS galaxy, which is one of the 30% QS Starforming galaxies.:
It's in the "SFR" (star-forming region) part of the BPT diagram, and all the H Balmer lines are emission (at least to H-epsilon1). However, the model spectrum differs considerably from the 'observed' one, especially in that the model is waaay lower (in peak flux) than the observed emission lines; it is even ~zero for some quite prominent (observed) emission lines (e.g. the [OIII] line red-ward of H-gamma).
1 strangely, this is at a significantly different redshift than the other H-Balmer lines
http://quenchtalk.galaxyzoo.org/#/boards/BGS000000e/discussions/DGS000021e?page=1&comment_id=5257ba4a74983225cd00014eI guess, that it means, that the spectrum however doesn't fit?
I wonder, if Jean's comment also applies to the spectra of the other 30% QS starforming galaxies? If so, I guess, that this might be insightful.Posted
-
by
JeanTate
in response to JeanTate's comment.
Several seem to have covering fractions below 20%, ...
From an analysis I've just done, "several" would be a considerable understatement.
DR7 does not have cModelMag fields, so I used the ModelMag ones to obtain an estimate of the total galaxy flux/luminosity. By comparing these with the fiberMag values, I found that 39 (of 80) have covering fractions < 20% in at least one band, and 31 in all five bands1. This was rather a surprise.
Here's an example, DR7 ObjId 587722982822379684, a nice face-on two-arm barred spiral with a redshift of 0.034:
I calculate the covering fractions as 4.3%, 5.2%, 6.5%, 7.2%, and 7.4% in the u, g, r, i, and z bands, respectively.
The spectrum has a nice big H-alpha/[NII] complex, strong [SII] , and obvious [OII] emission too, but little else (emission-wise). It's also got a continuum which slopes downward to the blue end, perhaps indicating lots of dust.
Some consistency puzzles: SDSS gives this galaxy an r-band magnitude (ModelMag) of 15.13, yet it is listed as 13.29 in the Wong et al. paper; similarly, the u-band mag is 17.7, giving a u-r color of 2.57, but it's just 1.22 in the paper. fracDeV_r is the same, 0.477, which is strange as this is obviously a late-type spiral!
1 Preliminary results; I have not yet checked these
Posted
-
by
zutopian
Referring to the above post by Jean about 587722982822379684, which is one of the 80 Wong et al sample galaxies.:
DR8 spectrum chart says AGN, but Wong et al had stated in their paper, that they didn't detect AGN in their sample.:
"However, apart from two PSGs which exhibit spectral properties of LINERs, we do not observe any spectral signatures of AGN within our PSG sample."
http://arxiv.org/abs/1111.1785
I wonder, which ones are the 2 PSGs with spectral properties of LINERS?
I found out, that 587739810494218509 and 587739809961804053 are the Liners according to Table 1 in the Wong et al paper.Besides this galaxy has another GZ paper reference.: "Galaxy Zoo: Passive Red Spirals" by Master et al.:
We provide an online table listing our full sample of red spirals along with the normal/blue spirals used for comparison.
http://arxiv.org/abs/0910.4113
This galaxy has this Reference in ADS, but however not in NED. (It has the GZ post-staburst paper as a Ref in ADS and also in NED.)
I checked the tables.: According to Master et al. it is a Passive Red Spiral.
I think, that it seems like a contradiction, that this galaxy is a sample galaxy in both GZ papers.: In one paper it is a Post-Starburst Galaxy and in the other paper it is a Passive Red Spiral. Besides there is following statement in the Master et al. paper.:While the presence of spiral arms suggests that major star formation cannot have ceased long ago, we show that these are not recent post-starbursts, so star formation must have ceased gradually.
http://arxiv.org/abs/0910.4113
Actually, I think, that it looks like a blue spiral. In the paper on page 7 there are shown some images of blue and red spirals.
As mentioned, I checked the tables and I found out, that this one is listed in the "Passive Red Spirals" Table aka Table A1.PS: K. Masters is also one of the authors of the GZ Post-starburst paper.
Posted
-
by
zutopian
Referring to my previous post:
Now I found a QS spiral galaxy, which Masters et al. listed in the table Blue spirals=Control sample for Red Passive Spirals.:
AGS000013a z=0.083
I agree, that it is blue. I think, that it is an inconsistency, that it was selected as QS galaxy and on the other hand it is a galaxy in the control sample of Masters et al..
Posted
-
by
zutopian
Concerning negative and bad masses, there was a discussion in the topic "Office Hours" 2 months ago.:
It turns out these 5 control galaxies with bad masses correspond to 5 post-quenched galaxies with bad masses. In our results, we'll filter out those bad post-quenched galaxies from our sample.
Are these still in the QS and QC samples? If so, which are these?
I'm really glad and excited you are talking about the 112 post-quenched galaxies with negative mass values and that you're interested in their implications for the data analysis.
While Yanmei (my collaborator who did the sample selection) was able to use the spectra for these galaxies to carry out her principal component analysis method (to identify them as post-quenched galaxies), it's not too surprising that for some galaxies she was not able to get an accurate mass estimate through her automated pipeline. It may be that these sources have something odd about their spectra. Or her pipeline came across something it didn't know how to handle and rejected these sources. (...)Which are these 112 QS galaxies? And did they found out the cause of the problem?
Posted
-
by
mlpeck
Which are these 112 QS galaxies? And did they found out the cause of the problem?
I'm not sure what the cause was, but a simple solution was to take the mass values from the DR10 database which has values for all but a few objects: scroll down to the end of this thread.
Posted
-
by
JeanTate
in response to JeanTate's comment.
From the Wong et al. (2012) paper:
Depending on the size of the galaxy, the 3 arcsec fibres (from which the SDSS spectra are obtained) may only correspond to the central 1.2–2.9 kpc of galaxies within our redshift range. In our sample, the observed spectral properties may not be representative for approximately 5 per cent of our sample where the outer galaxy regions are much greater than 3 arcsec and appear to be bluer than the central region.
So my random choice - one of 31 (or 39) - picked one of the four (=80 x 5%). Perhaps (more later).
Also:
We derive the u − r colours of our sample using the modelMag values (from SDSS DR7; Abazajian et al. 2009) that are determined
from the best fit of each galaxy profile to the linear combination of the exponential and the de Vaucouleurs profiles. In addition,
these magnitude measurements are corrected for dust attenuation using the models of Calzetti et al. (2000).Again, by chance my random choice may have been of the galaxy with the greatest correction for dust attenuation. Doesn't explain why the r-band mags are so different though. And it also seems, by chance, I picked the galaxy with the greatest difference.
An implication for the Quench project: to compare our results with those of Wong et al., we need to 'correct [the magnitude measurements] for dust attenuation using the models of Calzetti et al. (2000)'.
Time to check out Calzetti et al. (2000) ...
Posted
-
by
JeanTate
in response to zutopian's comment.
Very interesting findings, zutopian; well done!
To summarize: one galaxy (DR7 587722982822379684) is classed as both a PSG (in Wong et al. 2012) and as a passive red spiral (in Masters et al. 2010); one galaxy ( AGS000013a, DR7 588017625628278943) is classed as both a control blue spiral with active star-formation in Masters et al. 2010) and as a post-quenched galaxy in the Quench project.
Posted
-
by
JeanTate
in response to JeanTate's comment.
So my random choice - one of 31 (or 39) - picked one of the four (=80 x 5%). Perhaps (more later).
Um, no. These are the 31 PSGs with fiber covering fractions < 20% in all five bands, from smallest (5.5% max) to largest (19.8% max). The scale is 0.3" per pixel and the DR10 images are 120x120 pixels in size (so 36" x 36"); The fiber aperture thus just a circle 10 pixels in diameter; the last "image" is 64x64 pix (the smallest image possible) and 3" x 3" (different scale, obviously; can you guess which nucleus it is?).
Myself, I think there are obvious signs of on-going star-formation, well away from the nuclear region sampled by the spectroscopic fiber, in > four of these; and very different colors in parts of the galaxies away from the central few kpc, also in > four of these.
Posted
-
by
ivywong
scientist
Very nice work you all.
Apologies for not keeping up with all this as well as I could have. The galaxy identified above is indeed 1 of 4 galaxies that I initially threw out for the exact reasons that you all discovered. The other 3 are: 587733080268931236, 587739405703577638, 587739719754514434
In the end, we kept them in the sample because: 1) we wanted to keep our selection method consistent throughout. That is, we were selecting them simply from their post-starburst spectroscopic properties and 2) they didn't change the results of our finding since these were the only 4 galaxies where the central colours (or star formation history) appear very different to the colours beyond the fibre size.
Our sample was also specifically selected so that these galaxies are too nearby so that a larger fraction of each galaxy is observed by each fibre. Can't remember the fraction but most of the galaxies look like the galaxies shown in Figure 1 so we do not have very many galaxies where the centres are of a different colour/SFR to that in the outer regions (beyond the fibre). In general, the Kewley+05 paper is correct. It's just that because our sample consists of mostly dwarf/smaller galaxies, the central star formation histories sampled by the SDSS fibre are consistent with the general star formation history of the entire galaxies. (To first order, the colours are the same inside and outside the fibre).
Posted
-
by
JeanTate
in response to ivywong's comment.
Thanks very much ivywong! 😃
The other 3 are: 587733080268931236, 587739405703577638, 587739719754514434
Here are the DR10 images of those other three, respectively (same scale and size as above):
Posted
-
by
ivywong
scientist
Another slight subtlety is that these PSGs can still have a little bit of residual star formation and still have a youngish population as observed in the optical.
Our method of selecting galaxies without H-alpha only selects out galaxies that are no longer forming the most massive O-type stars. H-alpha emission is an indirect star formation indicator because a Hydrogen cloud has to be ionised by a hot young star to temperatures >~10,000 Kelvins. Now, these massive O-stars only live for 30 million years (blink of an eye in astronomical terms) and so even if there are no longer anymore O-stars around, the smaller young stars which have typical lifetimes from 100Myr- 1 Gyr will still shine quite brightly and the galaxy will still appear blue. This is because for every episode or burst of star formation, there are many more smaller stars formed compared to the number of super massive O-stars being formed.
Does this help?
Posted
-
by
JeanTate
in response to ivywong's comment.
Yes, it does; thanks.
I have to keep reminding myself that all these inferences apply, strictly speaking, to only the part of the galaxy sampled by the fiber; there could be very little recent star-formation in the bulge, for example, yet substantial amounts - with lots of HII regions - elsewhere. And what Kewley et al. (2005) showed is that, for a range of normal galaxies across essentially all Hubble types, one cannot infer SFRs for the whole galaxy if the fiber covering fraction is < ~20%. Also, though they didn't study this, if the fiber is not centered on the nucleus (or it does not include it at all) - as is the case for quite a few of the QS objects (and one PSG too) - inferred SFRs are unreliable.
And then there's LINERs. 😉
Earlier I wrote: "Time to check out Calzetti et al. (2000) ...", referring to the fact that you "corrected [the SDSS modelMags] for dust attenuation using the models of Calzetti et al. (2000)". I've now read that paper, but have no idea how you made those corrections! 😦
Would you mind at least sketching the steps involved? Specifically, what galaxy-specific inputs did you use? Just the five modelMags?
Posted
-
by
ivywong
scientist
Using the BPT diagram, I think we may have on order of 5-10% of LINERs but another way to study the star formation rate is looking of UV emission. These will come from both the O and B-type (less massive and not able to ionise the Hydrogen clouds) stars that are formed. But I suspect that this might be beyond the main aim of this paper so I would not try to confuse things further.
Also, if you look at figure 4 (from Wong+12), the scatter of red dots do include some bluer galaxies too so it's actually the fractional plot with the orange cloud which show that most PSGs (not all) are low-mass living in the green valley and they typically have early-type morphologies.
Posted
-
by
ivywong
scientist
For the dust correction, I corrected the ugriz magnitudes from the Balmer decrement using the dust models from Calzetti+00. One also has to know the E(B-V) extinction factor. I think it can be found from the published Galaxy Zoo catalogues. I'm not sure if this is true so it might be worth checking.
Basically the Calzetti model boils down to:
if (l ge 0.63 and l le 2.2) then k(i) = 2.659*(-1.857+1.040/l)+4.05
if (l lt 0.63) then k(i) = 2.659*(-2.156+1.509/l-0.198/l^2+0.011/l^3)+4.05
if (l gt 2.2) then k(i) = 0.0
where l is the wavelength in microns and k is the correctiondust_correction_factor = 10^(-0.4ebvk)
corrected_magnitude = observed_mag + 2.5 log(dust_correction_factor)Note that astronomers always use log base 10 unless otherwise specified. The dust correction factor will correct for both the dust in our Galaxy as we look out as well as the dust inherent in each of our observed galaxies (Calzetti's models).
Posted
-
by
zutopian
in response to zutopian's comment.
Copy from the topic "Spectrum, which doesn't look like post-starburst! QS galaxy".:
My question:
I guess, that it means, that the spectrum however doesn't fit?Reply by Jean:
I don't know, and I wish I did. I do not yet understand what - exactly - the model spectrum (red line) is (well, I've read the Redshifts, Classifications and Velocity Dispersions page, but I don't understand how the H-Balmer line fitting is done, nor how the flux estimates are derived).My question:
I wonder, if your comment also applies to the spectra of the other 30% QS starforming galaxies?Reply by Jean:
I'm sure it does to at least some. In my experience, a single object, selected at random from a set defined in some fairly consistent way, is rarely unique.Posted
-
by
zutopian
in response to JeanTate's comment.
Thanks! I continued my work.:
I checked all 80 Post-Starburst galaxies from the Wong et al sample, in order to find out, if there are further spirals, which have also the Reference Masters et al., but I didn't spot any further cases. So the spiral, which you had posted for another reason, is the only one.
Of the 80 galaxies in the Wong et al sample 15 were classed as spiral.
BTW, please be informed, that the Masters et al. paper had been published in 2010 and the Wong et al. paper was published in 2012. However, the "GZ:Passive Red Spirals" paper isn't listed in the References of the Wong et al GZ paper.
I am not sure, what the following statement in the Masters et al paper means.:While the presence of spiral arms suggests that major star formation cannot have ceased long ago, we show that these are not recent post-starbursts, so star formation must have ceased gradually.
http://arxiv.org/abs/0910.4113
And in the paper there are further similiar statements.
Concerning GZQ:
I guess, that the QS sample might contain further blue control spirals or maybe also red passive spirals from the Masters et al paper.Posted
-
by
mlpeck
I took a look at the Wong sample of 80 and the Goto (2007) sample of 564 yesterday.
In the Wong sample there was one object with a starburst spectrum although according to the MPA pipeline its star formation rate is only ~0.1 Msun/yr. Here is the link to its DR10 explore page.
There were also 6 LINERs and 5 in the composite/transitional region of the BPT diagram. There were ~3 that appear to be well in the red sequence. I'm using the MPA-JHU pipeline estimates for emission line fluxes and line indexes.
I may write about the Goto sample later and somewhere else.
One thing I discovered is that there are some massive errors in the DR8+ spectroscopic redshift estimates. 2 of the 80 in the Wong sample were affected and 29 of the Goto sample. The spectroscopic pipeline seems vulnerable to huge errors in objects with strong Balmer and H+K absorption. I may write more about this later as well.
Thank you Ivywong for joining in.
Posted
-
by
zutopian
I wonder, what MPA says concerning AGS0000021 : The one, whose spectrum doesn't fit.
Posted
-
by
zutopian
in response to mlpeck's comment.
Thanks for your reply. I downloaded the table.
You found the missing masses for the 112 galaxies with negative masses in DR10.
Were the mass values for the other QS galaxies not also taken from the DR10 database?
As far as I understood, they were taken from another database, in which mass values for the 112 galaxies were missing/negative. If so, do both databases agree concerning the stellar masses of the QS galaxies, whose masses weren't missing?Posted
-
by
mlpeck
in response to zutopian's comment.
I wonder, what MPA says concerning AGS0000021 : The one, whose spectrum doesn't fit.
Doesn't fit what? It's a starforming galaxy.
As far as I understood, they were taken from another database, in which mass values for the 112 galaxies were missing/negative. If so, do both databases agree concerning the stellar masses of the QS galaxies, whose masses weren't missing?
As far as I know they were from the same MPA pipeline applied to DR7 data. The remaing non-missing values are the same -- not identically for the most part but the differences are trivial (the standard deviation of the difference is 4E-5).
Posted
-
by
ivywong
scientist
Most of the confusion between overlapping samples can be solved once we look at how and where these samples were selected. The red spirals were picked out because of their morphology and colour. The Wong sample relies on the spectral properties from the central regions of a galaxy. It is conceivable that the the central regions can show a post-starburst signature while in fact, its spiral arms are fading gradually because both selection methods are not mutually exclusive. Also, the Masters sample come from DR6 and use the MPA-Garching's measured properties, while the Wong sample extracts the spectral properties from Gandalf (Sarzi+06). Different methods of spectral fitting may result in the observed differences in samples. The differences between the Goto and the Wong sample are also there because Goto has a slightly different selection criteria.
@mlpeck With regards to the one galaxy with a starburst spectrum, are you sure it's not a post-starburst instead of a starburst spectrum? Reason being that 0.1 Msun/year is close enough to zero once you factor in the uncertainties and errors in estimating this number. Dust corrections are notoriously uncertain so the measured values will also reflect this between different databases.
Posted
-
by
mlpeck
With regards to the one galaxy with a starburst spectrum, are you sure it's not a post-starburst instead of a starburst spectrum? Reason being that 0.1 Msun/year is close enough to zero once you factor in the uncertainties and errors in estimating this number.
What do you think? Visually, to me the spectrum looks pretty strongly starforming but, if I calculated it correctly the Hα luminosity is ~1040.3 erg/s without attenuation correction or 1040.6 with correction based on the Balmer decrement. Using Kennicutt's scaling relation that implies a SFR of ~0.15-0.3 Msun/yr, which is not so vigorously starforming. This is for the fiber only.
I'm using the MPA-JHU emission line flux estimates here and my own cosmological distance calculator with cosmological parameters that may differ a little from what the GZ science team is using these days.
Posted
-
by
ivywong
scientist
Having a young/blue spectrum just says that your stellar population is young and has been star-forming up to maybe 1Gyr ago. If the H-alpha emission is non-existent then it is by definition not currently star-forming. Perhaps you can try looking at another star formation indicator such as the far-ultraviolet. I did not want to suggest this earlier because it is beyond the scope of this project but as a sanity check, far-ultraviolet emission is an indicator of young O/B type stars which are up to 100 Myr ago.
So I took the liberty of search an online archive of past UV observations using the Galex Space Telescope and found that they may be a little bit of star formation but not as much as you'd think. Check out the following link (for reference the blue is Far-UV & yellow is near-UV--> yellow shows young stars up to 1Gyr and blue are the 100Myr stars so fairly current/recent):
UV image of this galaxy---source is centered
Posted
-
by
zutopian
Comment by mlpeck:
One thing I discovered is that there are some massive errors in the DR8+ spectroscopic redshift estimates. 2 of the 80 in the Wong sample were affected and 29 of the Goto sample. The spectroscopic pipeline seems vulnerable to huge errors in objects with strong Balmer and H+K absorption. I may write more about this later as well.
New topic by mlpeck:
Major redshift measurement errors in the SDSS spectroscopic pipeline
http://quenchtalk.galaxyzoo.org/#/boards/BGS000000b/discussions/DGS000021u?page=1&comment_id=526024447498327189000007Posted
-
by
zutopian
Here is a statement, which I found in another paper, in which Goto was involved as author.:
In this paper, we use publicly available true E+A galaxies (without Halpha nor [OII] emission) selected from the Sloan Digital Sky Survey,(...),hereafter DR5) by Goto (2007b)
E+A and Companion Galaxies - I : A Catalogue and Statistics
Chisato Yamauchi (ISAS/Jaxa), Masafumi Yagi (NAOJ), Tomotsugu Goto (Ifa/Uh)
(Submitted on 4 Sep 2008)
http://arxiv.org/abs/0809.0890I think, that there is Halpha and OII in some cases in DR8.: e.g.: pair ID 16:
http://skyserver.sdss3.org/dr8/en/tools/explore/obj.asp?id=1237658917995610179
They took another spectrum at KPNO.: KPNO Observation #8Posted
-
by
zutopian
in response to zutopian's comment.
Referring to my previous post:
One of those GPairs were observed by using the Subaru telescope. That GPair is neither in the Wong et al sample nor in the QS sample.: http://skyserver.sdss3.org/dr8/en/tools/explore/obj.asp?id=1237665357237518394
z=0.034. The redshift in DR8 is different and wrong.Spatially-Resolved Medium Resolution Spectroscopy of an Interacting E+A (post-starburst) System with the Subaru telescope
Tomotsugu Goto (Ifa/UH), Masafumi Yagi (NAOJ), Chisato Yamauchi (ISAS/JAXA)
(Submitted on 4 Sep 2008)
http://arxiv.org/abs/0809.0891Posted
-
by
zutopian
Below there is a further paper, in which Goto is listed as an author. The paper is from 2011/2012. They used DR7.:
The current star formation rate of K+A galaxies
We derive the stacked 1.4 GHz flux from FIRST (Faint Images of the Radio Sky at Twenty Centimeters) survey for 811 K+A galaxies selected from the SDSS DR7.
Danielle Nielsen (Wisconsin), Roberto De Propris (CTIO), Susan E. Ridgway (CTIO), Tomotsugu Goto (IfA, Hawaii)
(Submitted on 3 Jun 2011 (v1), last revised 22 Oct 2012 (this version, v2))
http://arxiv.org/abs/1106.0757Citations to the Article:
2013MNRAS.431.2034M by Melnick, J.; De Propris, R.
The spectral energy distributions of K+A galaxies from the UV to the mid-IR: stellar populations, star formation and hot dustThey used the sample from the other paper. De Propris, R. is listed as an author in both papers.
Posted
-
by
zutopian
The Fundamental Planes of E+A galaxies and GALEX UV-excess early-type galaxies: Revealing their intimate connection
Strong Balmer absorption lines and the lack of Ha and [OII] emission lines signify that E+As are post-starburst systems. (...)
In particular, we examine the FP scaling relations of the largest sample of ~1,000 E+As selected from the SDSS and ~20,000 morphologically-selected SDSS ETGs with GALEX UV data.(...)Yumi Choi (1), Tomotsugu Goto (2,3), Suk-Jin Yoon (1) ((1) Yonsei University, (2) ISAS/JAXA, (3) University of Hawaii)
(Submitted on 7 Apr 2009)
http://arxiv.org/abs/0904.1209
Posted
-
by
ivywong
scientist
Yes, there are errors and uncertainties involved with measuring redshifts but the exact definitions of K+A or E+A can differ. The following paper might be of interest too:
http://adsabs.harvard.edu/abs/2000ApJ...529..157P
Posted
-
by
trouille
scientist, moderator, admin
Response about sample selection based on emails with Yanmei Chen, now a faculty in astronomy in China (who recently had a baby girl). She was Christy Tremonti’s postdoc at UW-Madison when she did the sample selection.
http://en.wikipedia.org/wiki/Principal_component_analysis helps provide an overview of this method.
In brief, the method looks for features (emission and absorption lines, size of the 4000 Angstrom break, etc.) in the galaxy spectra that indicate that they're a member of this special sub-group of galaxies.
The features the program looks for are based on those described in http://postquench.blogspot.com/2013/06/wong-et-al-article-galaxy-zoo-building.html. But it takes this to the next level, by creating a whole range of model spectra with different star formation history, time since quenching, rate of quenching, stellar population models, metallicity, etc. and correlates the SDSS galaxies against all the spectral components of these model spectra.
The power of the principal component analysis is that it uses many spectral features to determine if a galaxy is or isn't a post-quenched galaxy.
In detail:
First we create a set of galaxy model spectra. Details about the model spectra are provided at the end of this post.
We apply the principal component analysis (PCA) method to our set of galaxy model spectra. This process allows us to create eigenspectra.
In our PCA analysis, we use the restframe 3750-4150Ang wavelength range. At wavelengths around 4000Angstrom, galaxy spectra vary in both spectral shape (4000Ang break) and strength of the hydrogen Balmer absorption lines. Both of these are very sensitive to the recent star-formation history.
Figure 1 shows the mean spectrum of the model library and the first seven eigenspectra (PC1-PC7). PC1 corresponds to the continuum shape. PC2 shows the Balmer absorption lines, PC3 contains Ca II (H+K).
https://vault.it.northwestern.edu/let412/GZQuench/Talk_Images/fig1and2.pdfWe then project the SDSS DR7 spectra through the eigenspectra. Through this projection, we get the strength of the eigenspectra for each spectrum (i.e., for each galaxy). We call the strengths of each eigenspectra PC1, PC2 ..., which stand for Principal Component 1, Principal Component 2…
Figure 2 is an example of the projection of the continuum fit. Black is a DR7 spectrum in the range of 3750-4200Angstrom, green is our best fit using our Eigenspectra.
https://vault.it.northwestern.edu/let412/GZQuench/Talk_Images/fig1and2.pdfOur sample selection is based on PC1 and PC2.
Figure 3 shows how we select the post-quenched galaxies using PC1 and PC2. Our post-quenched galaxies are the red dots in the figure. The overplotted lines show different models for how quickly the star formation is quenched in different models (tau = 100, 300, and 500 Myr, from top to bottom). The left bottom point of each of these model lines are where galaxies lie whose star formation has just started. As a galaxy evolves over time, it moves to the right along those model lines. The red part of the lines are where the models’ ages are between tau + 0.2Gyr to 2 Gyr.
The PSB selection criteria (sel = where(pc2 gt a*pc1+b and pc1 lt -1.8), with a=-0.146667 and b=0.33332) is built according to the \tau = 500Myr model and Yanmei’s comparison with the selected post-quenched galaxy spectra with S/N > 15.
https://vault.it.northwestern.edu/let412/GZQuench/Talk_Images/psb_selection.pngMore details are as follows:
(1) model library: Our input data set for the creation of the PCA eigenspectra is a set of model spectra generated using the Bruzual & Charlot (2003) BC03 stellar population synthesis code. The model library is similar to that used in Kauffmann et al. (2003) and Salim et al. (2005), although with a more restricted parameter range:
(i) The time tform when the galaxy begins to form its stars is distributed uniformly between 0 and 5.7 Gyr after the big bang (the age of the universe is assumed to be 13.7 Gyr)
(ii) The model galaxies have exponentially declining star formation histories
SFR~exp(−t/tao) with tao distributed uniformally between 1~2Gyr(iii) Top-hat bursts of star formation are superimposed on these continuous star formation models. Two parameters describe the bursts: fburst, the fraction of the total stellar mass formed in bursts, is distributed logarithmically between 0.0 and 0.1; tburst, the duration of the burst, is distributed uniformly between 0.03 and 0.3 Gyr.
During the burst, stars form at a constant rate. Bursts occur with equal probability at all times after tform and the probability is set so that 50 per cent of the galaxies in
the library have experienced a burst over the past 2 Gyr. Finally, the fraction of galaxies with ongoing starbursts is reduced to 25 percent of those initially created.(iv) The metallicity is distributed linearly in the range 0.5~2Zsolar. No metallicity
evolution is included.Posted
-
by
trouille
scientist, moderator, admin
As you can see, there's no selection for/against AGN activity within our post-quenched galaxy sample. That was something we were excited about -- that this sample selection doesn't exclude AGN. That way we can actually consider AGN activity as a potential mechanism for shutting off star formation in these types of galaxies.
Other selection methods for post-quenched galaxies, particularly those based on [OII] equivalent width (EW), excluded AGN. People used EW[OII] < some value because it was a way to identify sources that didn't have current star formation ([OII] is a good star formation indicator). However, AGN activity can also boost your [OII] emission, so you can have a galaxy with little current star formation but with AGN activity that has a high EW[OII] that would be excluded from previous post-quenched galaxy selection.
Posted
-
by
trouille
scientist, moderator, admin
https://vault.it.northwestern.edu/let412/GZQuench/yan_2009_psbEnviro_DEEP2.pdf provides a nice description of why we don't base our post-quenched galaxy sample selection on emission lines like [OII].
Posted
-
by
mlpeck
in response to trouille's comment.
This is great, thank you. Unfortunately these links go to vault.it.northwestern.edu, which wants a user name and password for access. Can you copy them to an open access location or give us a password?
The details of the principal components calculations appear to be slightly different than is described in Chen et al. (2012) -- for example a different restframe wavelength range is used. Is this correct?
Posted
-
by
trouille
scientist, moderator, admin
Very sorry. Didn't realize those links were password protected! Please try again and let me know if I've correctly changed the permissions.
Posted
-
by
trouille
scientist, moderator, admin
Yes, the different rest-frame wavelength is correct. Are there other differences you noticed between Chen et al. (2012) and the description she provides above for our sample selection?
Posted
-
by
zutopian
Quenching of Star Formation
Here we will review a new robust and iterative Principal Component Analysis (PCA) algorithm, which solves several common issues with classic PCA. Application to the 4000AA break region of galaxies in the VIMOS VLT Deep Survey (VVDS) and Sloan Digital Sky Survey (SDSS) gives new high signal-to-noise ratio spectral indices easily interpretable in terms of recent star formation history.
Vivienne Wild, Tamas Budavari, Jeremy Blaizot, C. Jakob Walcher, Peter H. Johansson, Gerard Lemson, Gabriella de Lucia, Stephane Charlot
(Submitted on 7 Nov 2008)
http://arxiv.org/abs/0811.1174Posted
-
by
zutopian
in response to zutopian's comment.
I informed about the paper also in the topic "scientific articles" and mlpeck did following comment.:
The technical paper describing the technique used by Wild et al. is Budavari et al. (2009).
Their (and others) methods for filling in missing values were criticized by Tsalmantza and Hogg (2012), who present an alternative algorithm based on iterative matrix factorization. Essentially the same algorithm was used earlier by Blanton and Roweis (2007) to create templates for spectral fitting.
I've implemented versions of Tsalmantza and Hogg's algorithm for PCA and the Blanton & Roweis algorithm for template generation.
Posted
-
by
zutopian
in response to zutopian's comment.
Budavari is a co-author of the Wild et al. paper and Wild is a co-author of the Budavari et al paper..
In the below paper by Yan-Mei Chen et al., the Budavari et al 2009 paper is mentioned as follows.:
Our method makes use of Principle Component Analysis (PCA), a standard multivariate analysis technique (see Budavari
et al 2009., for a recent discussion).Evolution of the Most Massive Galaxies to z=0.6: I. A New Method for Physical Parameter Estimation
Yan-Mei Chen, Guinevere Kauffmann, Christy A. Tremonti, Simon White, Timothy M. Heckman, Katarina Kovac, Kevin Bundy, John Chisholm, Claudia Maraston, Donald P. Schneider, Adam S. Bolton, Benjamin A. Weaver, Jon Brinkmann
http://arxiv.org/abs/1108.4719Posted
-
by
mlpeck
in response to trouille's comment.
The links work now, thanks.
Are there other differences you noticed between Chen et al. (2012) and the description she provides above for our sample selection?
They used at least a slightly different range of star formation histories and a broader range of metallicities. There's no mention above of whether they included dust in their models or varied stellar velocity dispersions.
The PC coefficient plot looks like a tilted and slightly distorted plot of Hδ vs. D4000, which helps explain the rather sharp cutoff in D4000 values.
Posted
-
by
zutopian
The features the program looks for are based on those described in http://postquench.blogspot.com/2013/06/wong-et-al-article-galaxy-zoo-building.html. But it takes this to the next level, by creating a whole range of model spectra with different star formation history, time since quenching, rate of quenching, stellar population models, metallicity, etc. and correlates the SDSS galaxies against all the spectral components of these model spectra.
The power of the principal component analysis is that it uses many spectral features to determine if a galaxy is or isn't a post-quenched galaxy.It is still unclear to me, why the QS sample contains just 2 galaxies of the 80 galaxies from the Wong et al sample?
Does this mean, that the Wong et al sample is incorrect? Are 78 actually non-PSGs? If so, the conclusions in that paper are also incorrect.
Do I miss something?Posted
-
by
zutopian
in response to trouille's comment.
That way we can actually consider AGN activity as a potential mechanism for shutting off star formation in these types of galaxies.
In the Wong et al paper there is given following statement.:
Therefore, following the results of Yan et al. (2009) and Wild et al. (2009), we hypothesise that the evolution of our local PSG sample is likely to be different to that of PSGs at higher redshifts.
Consistent with the results of Kaviraj et al. (2007), it is unlikely that AGN feedback will be a dominant quenching process for star formation in local PSGs.http://arxiv.org/abs/1111.1785
Refs.:
Yan et al. 2009 : The DEEP2 Galaxy Redshift Survey: environments of post-starburst galaxies at z ~ 0.1 and ~0.8
Wild et al. 2009 : Post-starburst galaxies: more than just an interesting curiosity
Kaviraj et al. 2007: The UV properties of E+A galaxies: constraints on feedback-driven quenching of star formationPS: Kaviraj is one of the co-authors of the Wong et al paper.
Posted
-
by
zutopian
in response to JeanTate's comment.
I reviewed this topic. I reply to a post by Jean, which she did on 12 Oct.:
There are some which seem to have covering factions <. 20%, and certainly some are disturbed and/or asymmetric. However, none are at all like AGS000048a or AGS00000wq. Perhaps the closest is 587741722823754043; the fiber does not seem to coincide with the nucleus (the only obvious case I found), and it's clearly a merger (there are certainly other mergers too).
587741722823754043 is from the Wong et al sample.:
It isn't a merger, but an overlap.: Both galaxies have different redshifts.: z=0.031 and z=0.013. It is listed in the GZ overlap catalog.
Posted
-
by
zutopian
Below galaxy is missing in the QS sample, though it is one of the two SDSS images, which are shown in a GZ blog post and the spectrum looks like Post-starburst.:
http://cas.sdss.org/dr7/en/tools/explore/obj.asp?id=588848899898867738GZ blog post:
http://blog.galaxyzoo.org/2013/07/10/galaxy-zoo-quench-experience-the-full-scientific-process/PS: The other SDSS image is in the QS sample.
Posted
-
by
ivywong
scientist
in response to zutopian's comment.
What you are finding are the main differences between the Wong+ sample selection (a very simple approach of no Halpha emission & some H-delta absorption) and a full PCA analysis of an entire spectrum which includes these lines. The reason I did not name my sample as local analogues of E+As (or K+As) is because there is no selection for an underlying old stellar population in my selection. This is can only be found if I used the entire spectrum in my selection such as the PCA approach. The point of my selection is to find all the local galaxies that have recently and quite sharply truncated its star formation. So it's a selection based on its current star formation history. So, the result of this paper (that AGN is not the cause for shutting off the star formation in local galaxies) is valid provided one understands that not all post-quenched galaxies are E+As but all E+As are post-quenched. If the QS sample are indeed selected to be E+As, then it will only find E+As and none of the other post-quenched galaxies.
Posted
-
by
mlpeck
If the QS sample are indeed selected to be E+As, then it will only find E+As and none of the other post-quenched galaxies.
I can't seem to find the blog or talk post at the moment, but I think the quench sample selection criteria were supposed to be broader than is implied by "E+A", "K+A", or "post-starburst." That would seem to imply that it includes those as subsets, but that's not the case. At least for the few studies where I've been able to retrieve catalogs of objects there's a non-null intersection of sets of objects, but for example only a little over half of Goto's 2007 catalog is included in the Quench sample.
It would be difficult if not impossible to reproduce Trouille et al.'s sample selection, so it's hard to say why any particular sample merely overlaps with the one we have at hand.
Posted
-
by
JeanTate
in response to mlpeck's comment.
It would be difficult if not impossible to reproduce Trouille et al.'s sample selection
This was certainly true when Stage 2 began, but since then there's been a fair bit of discussion on the selection criteria, and SCIENTIST trouille ("Laura") has provided quite a bit of background (if memory serves, which these days I must confess it doesn't always, not well anyway), right?
For clarity, I assume you're referring to the selection of the QS objects; the selection rules for the QC ones are quite unambiguous, aren't they? Sure, since there's a random element involved, it is not possible to reproduce the QC selection exactly (i.e. produce a catalog of 3002 objects that exactly matches the 3002 objects actually in the current version of the QC catalog), but otherwise zero uncertainty, right?
Posted
-
by
mlpeck
in response to JeanTate's comment.
This was certainly true when Stage 2 began, but since then there's
been a fair bit of discussion on the selection criteria, and SCIENTIST
trouille ("Laura") has provided quite a bit of background (if memory
serves, which these days I must confess it doesn't always, not well
anyway), right?For clarity, I assume you're referring to the selection of the QS
objectsWell, she explained in satisfactory detail how the quench sample was selected, however the sample selection was based on unpublished models with a slightly different range of star formation histories than the published version. Also, and maybe more importantly a different wavelength range was used in calculating the principal components.
That means that the PCs will encode different information than the published versions and the coefficients of the fits to the PCs will differ. As of DR10 those are now tabulated in the CAS table stellarMassPCAWiscBC03 but it isn't possible to reconstruct the quench sample with a query on that table because the coefficients are different.
If it were possible I would probably have seen it since I had downloaded almost every column of that table for both the quench and control samples some time ago.
Posted
-
by
trouille
scientist, moderator, admin
Excellent question and resource. I had not seen that this was available. I've emailed Yanmei Chen (our collaborator who created the Quench sample selection) to see how/if we can use the publicly available information to recreate her sample selection. If we cannot use that website, I'll push on having her make available the equivalent for us to be able to recreate her Quench sample selection.
In the meanwhile, definitely check out http://quenchtalk.galaxyzoo.org/#/boards/BGS0000008/discussions/DGS0000223 and focus our efforts there for now.
Posted
-
by
trouille
scientist, moderator, admin
Also, below are the different Talk threads related to Sample selection (the first being the most detailed):
Go the first post in this discussion thread. I revised the post so that all the details of the selection criteria are upfront. And yes, by 'this discussion thread', I mean page 1 of this discussion thread that we're writing in.
http://quenchtalk.galaxyzoo.org/#/boards/BGS0000001/discussions/DGS00001xyhttp://quenchtalk.galaxyzoo.org/#/boards/BGS000000a/discussions/DGS00001xk
http://quenchtalk.galaxyzoo.org/#/boards/BGS0000001/discussions/DGS0000006
Posted
-
by
trouille
scientist, moderator, admin
in response to trouille's comment.
Hi all,
Please go to the first post in this discussion thread to read Yanmei's response about why the SDSS with eigenspectra values is not best to use.
Instead she has provided all the data needed (including the eigenspectra values that she created specifically with post-quenched galaxy selection in mind) to not only recreate her sample selection, but also explore if any changes would make for a better selection and how other sample selections compare with this one.
Definitely go check out that updated post (you'll see the line "Post updated on December 20, 2013 to bring additional information from Yanmei up to the forefront" about 2/3 of the way down the first post).
Posted
-
by
mlpeck
in response to trouille's comment.
Laura posted a very important update a few weeks ago on the quench sample selection process that I had completely overlooked until now, probably because she edited a 5 month old post in order to place it at the very top of this thread.
Unfortunately I think editing an old post doesn't put it back on the recent discussions list, which is how I keep track of recent activity.
Anyway, it's the very first post in this thread.
Posted
