The Lick HdeltaA - D4000 plane; some more results
-
by
mlpeck
I posted a couple versions of this graph before. Please see here for a discussion of why this is relevant and links to some literature. The graph has been updated with data from the latest version of the control sample.
Strong Balmer absorption indicates an intermediate age population is present, and the area roughly above the outermost contour and to the left of d4000 ≈ 1.5 is, according to Kauffmann, the region where a strong starburst has occured within the last ~0.1 Gyr or so.
According to Goto (2007) the spectroscopic signature of "E+A" galaxies is HδA ≥ 5 Å and weak emission lines. It turns out that almost half the quench sample lies above the 5 Å line (1481/3000 objects), compared to a little under 5% for the control sample. Out of those 1481, 557 have no classification from the BPT diagnostic, but that just means that one or more out of 4 lines that go into that diagnostic is weak (or masked from the spectrum for some reason).
Out of those 557 I looked for galaxies with S/N < 3 in all lines tabulated in the data table. There were 37 of them. I visually inspected all of those spectra and verified that there was no obviously detectable emission. For comparison there were just 2 true "E+A" galaxies in the control set.
GZ classifiers called 34 out of 37 "smooth", with one "star or artifact" (that one is actually an extremely compact galaxy at z=0.326). GZ classifiers saw "Tidal debris" in 2 galaxies and 1 merger (the merger classification was in error).
So it appears that somewhat over 1% of our quench sample are true "E+A" galaxies, that is they are morphologically early type, have unmistakable spectroscopic signatures of recently terminated star formation, and have no detectable ionized gas.
Posted
-
by
trouille
scientist, moderator, admin
This is really interesting and a result to think more about.
Quick question -- what do the different colors in your plot represent? Might be in the thread you refer to in your post, but thought it might be quick to ask here.
Posted
-
by
trouille
scientist, moderator, admin
Hdelta_absorption and D4000 play with each other well in terms of seeing how long it has been since a galaxy had a burst of star formation.
Notice that D4000 increases steadily with age, whereas Hdelta peaks around 400 million years (Myr), and remains visible till just past 1 billion years (1 Gyr). Note: the x-axis is Log(age).
Our post-quenched galaxies are selected to be within 1 Gyr of having star formation occurring, so it makes some sense that they generally have higher Hdelta values.
Posted
-
by
trouille
scientist, moderator, admin
Never mind my earlier question about colors. Definitely get the answer from http://quenchtalk.galaxyzoo.org/#/boards/BGS0000008/discussions/DGS00001zj
Blue - post-quench galaxies that lie in the AGN region of the BPT diagram
Red - post-quench galaxies that lie in the star forming region of the BPT diagram
Gray - Control galaxies
Posted
-
by
trouille
scientist, moderator, admin
In what you've read and thought about, what is your impression for why the post-quench 'BPT-AGN' and 'BPT-LINERS' have higher Hdelta absorption line strengths than the post-quench 'BPT-star formers'?
Also, what do you think is dominating the emission in our post-quench galaxies that lie within the 'BPT-LINER' regime? Do you think their LINER appearance is the result of weak AGN activity? Or do you think it's from the winds of older evolved stars shocking with the interstellar medium?.
The nature of LINERs is a debated question, still with no clear answer. I wondered if perhaps you've seen some common characteristics in the spectra of our post-quench LINERs that might shed light on this.
Posted
-
by
trouille
scientist, moderator, admin
Backing up for a moment, this discussion about D4000 prompted me to look to see if there's a difference between our Post-Quench and Control samples. My thought was that since D4000 has a clear increasing trend with time since star formation occurred, then likely we'd see a difference (since, for our post-quenched galaxies, it's been within 1 Gyr since star formation was quenched, but for our control galaxies, there's no preference on when they formed stars or if they're still forming stars).
Check out my dashboard: http://tools.zooniverse.org/#/dashboards/galaxy_zoo_starburst/5234d3736030053702000023.
The post-quenched galaxies D4000 values are between 1 and 1.5, which corresponds with less than 1 Gyr (1 billion years) since star formation in that plot above (the Figure 2 in 2 posts above). The control galaxies span the full range of values from 1 to 2, corresponding to a million to 10 billion years since star formation.
Mlpeck's plot at the top shows the same thing with its x-axis and the comparison between the post-quenched and the control.
It's likely others already noticed this, but I find this is useful confirmation that our post-quenched galaxies generally lie within the expected time frame since star formation.
Posted
-
by
mlpeck
Good questions that I will have to think about.
I posted in the long chat thread that I've uploaded tables of H delta indexes and errors to https://www.dropbox.com/s/lryj72fjd0mb4wr/hda.quench.csv and https://www.dropbox.com/s/5hqe7ygrr7xhau2/hda.control.csv.
Getting the right spectra and associated photometric objects for the control group was a little trickier than the quench sample, but I think I have correct matches for everything.
Posted
-
by
mlpeck
In what you've read and thought about, what is your impression for why the post-quench 'BPT-AGN' and 'BPT-LINERS' have higher Hdelta absorption line strengths than the post-quench 'BPT-star formers'?
This is easy enough I guess. If there's a significant young population the light in the blue region of the spectrum will be dominated by O-B stars, which have relatively weak Balmer line absorption.
This actually needs more follow-up: there are about the same percentage of starforming objects in the quench and control samples, and so far I don't see any systematic differences in their properties. So this raises the question again of how, in technical detail, the quench sample was selected.
Posted
-
by
mlpeck
Also, what do you think is dominating the emission in our post-quench galaxies that lie within the 'BPT-LINER' regime? Do you think their LINER appearance is the result of weak AGN activity? Or do you think it's from the winds of older evolved stars shocking with the interstellar medium?.
Schawinski et al. (2007) argued that - at least for early type galaxies - position in the BPT diagram reflects an evolutionary sequence that goes like
star forming -> transition region -> AGN -> LINER -> quiescent.
In their picture as nuclear activity increases star formation fades and subsequently the nuclear activity also fades into LINERs and ultimately passively evolving galaxies. There may be some evidence for that picture in our sample.
How would you distinguish between photoionization and shock models in a fiber based study? Spatially resolved spectroscopy has turned up evidence for extended LINER like emission that can't be due to ionization from a central AGN: see Singh et al. (http://arxiv.org/abs/1308.4271) for a very recent example. They argue that the emission regions are being ionized by post-asymptotic giant branch stars, which I guess would make them planetary nebulae. Even if this is the case it would still fit within a narrative where our quenched LINERS are aging starbursts since post-AGB stars are a late evolutionary stage of intermediate mass stars (if I remember correctly).
Posted
-
by
mlpeck
Out of those 557 I looked for galaxies with S/N < 3 in all lines
tabulated in the data table. There were 37 of them. I visually
inspected all of those spectra and verified that there was no
obviously detectable emission. For comparison there were just 2 true
"E+A" galaxies in the control set.GZ classifiers called 34 out of 37 "smooth", with one "star or
artifact" (that one is actually an extremely compact galaxy at
z=0.326). GZ classifiers saw "Tidal debris" in 2 galaxies and 1 merger
(the merger classification was in error).So it appears that somewhat over 1% of our quench sample are true
"E+A" galaxies, that is they are morphologically early type, have
unmistakable spectroscopic signatures of recently terminated star
formation, and have no detectable ionized gas.I'm going to briefly go over a few more details of this subset of the quench sample. My original filter wasn't quite right and it turned out that there were 38 total galaxies with true "E+A" spectroscopic signatures, that is strong Hδ absorption and weak emission in at least the 4 lines that go into the classic BPT diagram. 35 were classified as smooth and just 2 had tidal tails.
Almost all of these are at relatively high redshift: the median was z = 0.20 and only 3 had z < 0.11, which is about the median redshift for the entire quench sample (and the entire SDSS for that matter).
There were 557 total strong Hδ galaxies that were unclassified in the BPT diagram. Out of the remaining 519 almost all had detectable emission in at least one line -- [N II] 6584 was detected at the 3 sigma level in 490, and 397 had both [N II] and Hα detected. Here is a histogram of the ratio:
About 70% of these have log [N II]/Hα > 0, which probably indicates very weak LINER emission.
1 Here's a DR10 finder chart image for the closest one, at z = 0.025:
This is our AGS00001y1
Posted
-
by
mlpeck
I may as well post pictures of the other two "nearby" E+A galaxies in the quench sample. These are actually kind of interesting because they both have compact blue cores. I see hints of structure in the first one that are completely invisible in the image we were presented to classify. Here are AGS00001p6 (z = 0.08)
and AGS00001h0 (z = 0.09)
I want to see a tidal tail in this one, but it's probably spurious. It was classified as smooth and undisturbed.
Posted
-
by
JeanTate
in response to mlpeck's comment.
I want to see a tidal tail in this one, but it's probably spurious
The cross-hairs don't help; here it is without them (and a somewhat different scale):
Here's the DR8/9 image (a.k.a. Quench project classification image):
And the DR7 one, for comparison:
Myself, I think there's a faint tail connecting the E+A galaxy with the yellow ETG to the NW, and even a hint of dust reddening in that yellow blob, where the connecting bridge 'attaches' (more likely it is foreground).
Posted
-
by
mlpeck
I'm going to try to redo some previous analyses for one or both of the proposed quench subsets. I may use dormant threads for these or start new ones. Until someone persuades me to do otherwise I'm going to concentrate on "subset 2", which are QS objects with 0.02 < z < 0.10 and Mz < -20. Where needed I'm going to use the same selection criteria for control objects. Here is a plot of the HδA-D4000n plane:
Superficially at least this looks a lot like the plot at the top of this thread. The control sample is still strongly bimodal, while the quench sample still contains a large percentage of objects with HδA eqw > 5Å, which is the traditional indicator of a post-starburst spectrum. That percentage is however somewhat smaller than in the full sample -- 39% vs. nearly 50%.
Posted
-
by
mlpeck
There's a distinct impression of stratification by BPT class in the above graph, and I've been trying to find a way to better visualize it. Here's my most recent attempt. Here I've plotted medians for each class with 95% confidence intervals estimated by bootstrapping. Similar graphs were used in the Masters et al. paper on passive red spirals. If you compare graphs you'll see that red spirals occupy completely different regions in the HδA - D4000n plane.
Quench sample:
Controls:
Well, the stratification still isn't so clear with the quench sample, although AGN/LINER, weak emission, and no emission spectra are firmly in the post-starburst part of the plane.
The control sample on the other hand does form a distinct sequence proceeding from star forming at upper left to completely passive at lower right.
Posted
-
by
mlpeck
Masters+ found some segregation by stellar mass in the Hδ - D4000n plane, with the sequence running from upper left to lower right with increasing mass, and more significantly they found the median red spiral always to lie below and to the right of blue spirals in the same mass bin. They interpreted this to mean that red spirals are systematically older (in light weighted terms) with lower recent star formation rates than blue spirals, controlling for mass.
Here are the corresponding plots for the quench and control samples. I've binned by quintiles in stellar mass, and again the error bars are 95% confidence intervals on the medians estimated by bootstrapping:
So, the control sample does show clear segregation by mass. Except perhaps for the lowest mass bin the quench sample does not. Sorry for the lame legend on these. Here are the mass intervals by quintile for the control group:
- [9.46, 10.06)
- [10.06, 10.26)
- [10.06, 10.41)
- [10.41, 10.61)
- [10.61, 11.28]
The mass bins are the same for the quench sample except the minimum and maximum stellar mass are 9.24 and 11.41.
Why is there such strong mass segregation in the control group? This plot may give some insight: I've plotted D4000_n against stellar mass for the controls:
This shows a weak, but positive correlation. In fact if we add HδA as a predictor in a linear regression there's a very strong relation:
summary(lm(d4000_n~lgm_tot_p50+lick_hd_a,data=db.sub2c)) Call: lm(formula = d4000_n ~ lgm_tot_p50 + lick_hd_a, data = db.sub2c) Residuals: Min 1Q Median 3Q Max -0.58352 -0.08535 0.00021 0.08920 0.46799 Coefficients: Estimate Std. Error t value Pr(>|t|) (Intercept) -0.162935 0.146273 -1.114 0.266 lgm_tot_p50 0.173166 0.014046 12.329 2e-16 *** lick_hd_a -0.077433 0.001945 -39.816 2e-16 *** --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 Residual standard error: 0.1372 on 1142 degrees of freedom (4 observations deleted due to missingness) Multiple R-squared: 0.7092, Adjusted R-squared: 0.7087 F-statistic: 1393 on 2 and 1142 DF, p-value: 2.2e-16This is a highly significant relationship. What does it mean? What is the direction of causality? Is some unobserved variable the causative agent? I don't know. Well, I do know that the most massive galaxies in the local universe are giant ellipticals, and for the most part these have had no recent or ongoing star formation. Actively star forming galaxies tend to be lower mass, as is shown pretty clearly in these plots. D4000 is sensitive to light weighted age and at higher ages also metallicity.
For comparison, here are the quench sample objects overlaid on a contour plot of the controls. Note that the distribution of D4000 is strongly bimodal for the controls, while the quench sample has a fairly sharp cutoff at D4000n ≈ 1.53.
Here is a summary of the same linear regression for the quench sample. This shows a much weaker, but still significant relation among D4000, stellar mass, and Hδ
summary(lm(d4000_n~lgm_tot_p50+lick_hd_a,data=db.sub2q)) Call: lm(formula = d4000_n ~ lgm_tot_p50 + lick_hd_a, data = db.sub2q) Residuals: Min 1Q Median 3Q Max -0.25037 -0.04170 0.00471 0.04670 0.18240 Coefficients: Estimate Std. Error t value Pr(>|t|) (Intercept) 0.750948 0.058008 12.946 2e-16 *** lgm_tot_p50 0.063438 0.005598 11.333 2e-16 *** lick_hd_a -0.006658 0.001046 -6.367 2.79e-10 *** --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 Residual standard error: 0.06293 on 1146 degrees of freedom Multiple R-squared: 0.1271, Adjusted R-squared: 0.1256 F-statistic: 83.41 on 2 and 1146 DF, p-value: 2.2e-16Posted