referee report:
Report from Referee:
The manuscript "Discovery of Massive, Mostly Star-Formation Quenched
Galaxies with Extremely Strong Lyman-alpha Emission at z~3" presents a
sample of nine massive galaxies at z~2.5-3.2 with Lyman-alpha equivalent width
>100A and stellar mass log(M)>10.5
As described below, I believe that they Lya emission in five out of
the sample of nine galaxies can be easily explained by significant
star formation or AGN activity. This reduces the sample of noteworthy
candidates to four.
The result that some quiescent-looking galaxies have strong Lya
emission, but no hint of significant AGN activity, is somewhat curious
and may be worthy of publication. However there are many aspects of
this analysis that need to be greatly extended and clarified. There is
also good reason to believe that the Lya size measurements, upon which
much of the physical discussion in this letter are based, are
completely unreliable. For all of these reasons I do not think this
paper can be published in the ApJL in its current form.
Below I give some extended comments about the sample selection, and
then give more comments about the rest of the analysis.
#####################
Sample selection:
Of the nine galaxies in this sample, two (ID numbers 1 and 2) are
detected in the Chandra data. These galaxies can be classified as AGN,
since star formation cannot cause an x-ray detection at these
redshifts. These galaxies also have high star formation rates as
estimated from the SED fitting; hence it is not at all surprising that
they have Lya emission.
There is also a third galaxy (ID=9) that looks as if it may have a
power-law SED, which would strongly suggest AGN activity. The authors
claim that they used the Donley et al. (2012) criteria to select AGNs
from the IRAC colors, but since this object is not detected
significantly at 8um, perhaps the authors didn't apply any IRAC AGN
selection criteria to this object?
Of the remaining galaxies, two (ID numbers 5 and 8) have
log(SFR)=1.8-2, so it is very difficult to describe these galaxies as
"mostly quenched." In fact they are not all that far below the main
sequence of star formation at these redshifts (see e.g. Fig. 10 of
Schreiber et al 2010, and note that they use a Salpeter IMF as opposed
to Chabrier as used in this work. Also, recall that the observational
scatter that is typically quoted for the main sequence is 0.2-0.3
dex.). Finding Lya emission in some fraction of typical, massive
star-forming galaxies is not surprising and is not worthy of
publication (even if the galaxies are rather red, Lya emission can be
explained by radiative transfer effects; e.g. Finkelstein et
al. 2009).
This leaves a sample of four galaxies that are plausibly quenched,
have no evidence of significant AGN activity, and have strong Lya
emission (the sample increases to five if galaxy number 9 can be shown
not to have IRAC colors consistent with an AGN). Such a very rare
sample may be of marginal interest in itself, but it is impossible to
put them into an evolutionary scenario.
#############
Other comments:
-- Title: since the authors don't quote Lya fluxes (they only quote
equivalent widths), I don't think it is justified to say that these
galaxies have "extremely strong Lyman-alpha emission." In fact, the
authors have not shown that the Lya emission is very strong; the EW
is high, but that may only be because the rest-frame UV continuum
is very faint. Indeed, the SFR estimates from Lya are quite
low. For this reason I don't think it's fair to say that these
galaxies have "extreme" Lya emission. I would suggest that they
quote Lya fluxes, and compare them to the Lya luminosity function
(e.g. Gronwall et al. 2007), if they wish to claim that the Lya
emission is "extreme." Alternatively, a more accurate title would
refer to "large Lyman-alpha equivalent widths"
-- Also, regarding the title, at least four of the galaxies have
significant, active star formation based on the SED fits. And, as
explained above, the star formation rates are not strongly
depressed compared to the main sequence. If these galaxies were in
the local universe they would be called extreme starbursts. So it
is not accurate to describe them as "mostly quenched."
-- The authors discuss two possibilities for the cause of the Lya
emission. One is winds that are blown out by AGN or supernovae, and
the other is low-level star formation activity. The quoted
conclusion that the galaxies may represent a brief transition phase
between star-forming and quiescence is completely unsupported. If
the Lya emission is caused by low level star formation, then it
could suggest that the galaxies are not 100% quenched. It may also
suggest that the galaxies are currently undergoing a minor merger
with a gas rich dwarf, that the Lya emission is related to the
mysterious Lya blobs, or that it is due to a projection effect with
a nearby source; none of these possibilities are discussed by the
authors, and neither can these possibilities be easily placed into
an evolutionary scenario. Additionally, if the Lya emission is
caused by AGN activity, then neither does that give any believable
evidence that the galaxies are in a brief transition phase.
-- Regarding the AGN activity, I found it very odd that the authors
discuss how they removed AGN using radio/x-ray/IR indicators in
Section 2, but then in Section 3 they explain that two of the
sources are detected in the x-ray. An x-ray detection at z~3
certainly means that the galaxies are AGN; it would take an extreme
level of star formation to lead to an x-ray detection, and surely
such star-formation can be ruled out by the SEDs and the lack of a
24um detection. It would make sense to remove all AGNs from the
sample, or remove none of them. However if the AGNs are kept in the
sample then the authors would have to clearly specify which
galaxies are AGN in the table and figures.
-- Why have the authors not tried stacking the Chandra data in order
to see if there is a weak detection?
-- If the galaxies do indeed represent a transition phase between star
formation and quiescence, then there should be evidence for that in
the SED fits and the colors. Do these galaxies have younger ages,
or larger tau/age, then typical quenched galaxies at similar mass?
However, given the very well-known degeneracies and systematic
uncertainties in stellar population properties inferred from
SED-fitting, it is strange that the authors don't at least mention
this issue. The SFRs, in particular, are completely unreliable for
"quenched" galaxies. So to complement the discussion of the SED
fitting, I would also strongly suggest including a U-V vs. V-J
color-color diagram, as this is a useful visual diagnostic of
galaxy's colors, dust content, and age (e.g. Williams et
al. 2009). In this diagram there is a clear separation between
star-forming and quenched galaxies, and recently-quenched galaxies
would be found at the lower-left edge of the sequence of quenched
galaxies.
-- If there is significant Lya emission, there may also be other lines
with significant emission (Ha, OII, OIII). There is the potential
that these lines are affecting the SED fits. Do the SED fitting
results change at all if the filters that are potentially affected
by these lines are removed from the fits?
-- The size measurements in the IA bands are potentially quite very
important, as they may help to shed light on the mechanism
responsible for the Lya emission. For this reason I think that it
is important to take great care in the measurements, in order to
know whether the ground-based measurements are reliable. Instead, I
find the quoted measurements very problematic. For instance, for
galaxies 6 and 2 the ground-based i*-band size measurements
disagree *very strongly* with the F814 measurements. This leads me
to believe that ground-based size measurements are completely
unreliable for small objects, and that the quoted uncertainties are
grossly underestimated. It is only for a single galaxy (ID=1) that
there is any reason to believe that the ground-based measurements
are reliable. But note that this galaxy is both quite bright and
very extended; hence this may be a special case. If we can't trust
the i*-band size measurements, why should we trust the IA-band
measurements? The authors would have to do careful tests to show
that the measurements and the uncertainties are reliable and
unbiased. There are a number of examples in the literature of such
test; for just one particularly recent example, see Straatman et
al. (2015).
-- The authors should state the FWHM of the IA and i* images.
-- At the end of the manuscript, the authors that that future studies
will allow us to constrain the mechanisms that quench galaxies. I
find this statement completely unsupported. Even if the MAESTLOs
were transition objects, I am not sure how finding more such
objects, or obtaining IFU observations, would shed any light. If
the authors can specifically point out how studying MAESTLOs will
clearly reveal the quenching mechanism, then they should explain
it.