From: Rollin Thomas (rthomas@panisse.lbl.gov)
Date: Wed Dec 31 2003 - 13:32:22 PST
Gabriele:
I have read your 1999ac draft and I have some comments that I hope
will help you out. I will start by giving you an outline of what I
think a paper like this ought to be like. It seems to me that you
have to address three basic questions in this kind of paper.
[1] Why is this SN important? I can see two things, maybe three.
First of all, this set of spectroscopy adds to the growing set of
those events with good time coverage, not just after maximum light
but before maximum too. Second, this premaximum spectroscopy shows
pretty convincing signs of C II, and the C II ejection velocities
measured are roughly consistent with the unburned layers of W7.
There is also (perhaps) evidence of C III at lower velocities. I am
not completely convinced yet that the C III is there, and I have some
suggestions below about how I might become convinced. I also have
questions about the other spectra not fit, but I will defer them to
below too. As far as where this SN fits on the spectroscopic sequence
of SNe Ia, I think that's pretty secondary, and it should only be
referred to in passing. This SN isn't super-peculiar enough for me to
care about that. It's not a 2000cx or a 2002cx, it looks ``pretty
normal'' to me. Perhaps Peter/Lifan disagree?
[2] What previous work does this SN have relevance to? Well, it's
pretty much contingent upon that really being C III there. If you
have only C II above the W7 cutoff only, then you've got evidence in
support of W7. But if you've got C III below that velocity, then
you've got something supporting the 3D deflagration results with the
ashes and fuel mixed at all ejection velocities. So if the result
that this is C III is only circumstantial, then you can only suggest
that 3D models are pretty good. If that is the case, then you need to
address some of the diversity issues that David has brought up with
respect to the 3D models. How do we reconcile the crenellated results
from Khokhlov and Hillebrant with the required upper limit on
asphericity? What new observations are needed to help us out here?
I've suggested surveys of absolute depths of Si II features, or
something similar that constrains covering factor. Perhaps these
ideas should be developed further?
[3] Implications for the future. Well, since these C II features are
rather small (I always want to see the 7000-ish feature along with the
Si II-imposed feature, too), you should estimate the required signal
to noise for looking for C in the future. This has probably already
been done by someone, but when it comes to this question I don't
really know that I can say anything useful on the S/N side. 10? 100?
A billion? This may be common knowledge to people but it's not a
thing I carry around in my head. Also, internally for SN Factory,
what triggers need to be in place to follow these kinds of events?
And which of them should be followed spectropolarimetrically? That is
the key to getting a handle on the explosion model with respect to 3D.
In your presentation of the fits in section 3, I think you might want
to restructure. Clearly you want to focus on C, and I set a bad
example of this in 2000cx by only caring about three ions, but make
each ion a separate paragraph, and end each paragraph by saying
``evidence for such-and-such is (definitive|probable|possible|
inconclusive).'' This is the clearest way to do it. With each ion,
you can address anything weird you had to do with the ion, like put
it at some velocity where there is no evidence for other ions.
Now about those other spectra after max. Why don't you do something
with them? Isn't there at least one that might show evidence of the C
II there? Maybe day +11, or +16, or +24? What was the rationale
behind only paying attention to the early time? State it clearly I
think, in the paper. Of course, I think you ought to look at these
spectra with this kind of eye on C II and C III. You don't have to
present it in the paper, but you might want to address it.
And since you only analyze those first two spectra, (and I admit I am
getting into this game pretty late), why do you need to present the
others in this paper? To my mind, there ought to be a data paper that
discusses reduction and presents the data all together, maybe with
some preliminary line identifications, and then followup analysis
papers. The reason I suggest this is that people can reference the
first data paper later on, instead of an analysis paper, which kind of
clouds the issue. This decision might have been made at the exec
level, in which case I think we should make a case for this not being
the procedure in the future. I understand the desire to have first
crack at analysis; but look at the Lifan Wang/Dan Kasen combination on
2001el. There, a data paper was presented and Dan held off on
submission of his analysis to about the same time as the first one.
There's no need to rush either the reduction or analysis, I feel, if
good science is going to be done. Perhaps there are other
constraints?
About that figure of the ``radial composition.'' PLEASE BE CAREFUL. You
are on thin ice with a plot like this I think. I realize that someone
suggested that you do this for 1999aa, but I have to reiterate that
you must make it unequivocally clear in the paper that this is *not* a
definitive plot of the composition. The reason is because with
Sobolev optical depths you really have no good sense of densities and
temperatures without further analysis and without either doing full
NLTE or LTE calculations at least. If you look carefully, for
example, your optical depths for C II show C going all the way up to
40000. This is misleading; the optical depth and e-folding velocities
show that the optical depths goes practically to zero well before
v_max, and if you were to move v_max down to 30000 kmps the difference
between the two spectra would probably be minimal at most. So I say,
remember the results of direct analysis: (1) line identifications and
(2) constraining ejection velocities for various species. This is not
the same as deriving a composition! :)
With respect to the case for C III, I think it's hard to say from your
paper whether or not it is there in the fits. Recall that before I
left for vacation, I supported the idea of showing full fits to the
spectrum turning one ion on and off. This is good because often those
without a spectroscopic background will want to see what it does
overall. So I suggest that when you are making a case for C III, you
do one fit with solid lines, and another fit with dashed ones (the one
without the C III could be the dashed one) and then you overplot them.
This can get kind of busy, so I also advocate putting an inset plot
which retains the aspect ratio of the larger diagram, but focuses on
and enlarges the C III region of interest. The referee on my 2000cx
paper complained about this, and it's pretty important that if you are
making a claim for an ion to be there, you do everything you can to
demonstrate it both to yourself and to the reader.
Also, you might have noticed that David has begun to experiment with
plotting wavelength axes in log space, and sometimes plotting weird
combinations of flux and wavelength on the flux axis. Once you show
plots of all the spectra in straight flux, it is perfectly okay (for
the purposes of analysis) to flatten the spectrum out and rescale the
wavelength axes in this way. Since we are asking our readers to
evaluate the goodness of fit basically by eye, we need to make it easy
for them to do so. By taking the log of the wavelength axis, we make
the blue features and red features have the same intrinsic width (you
can demonstrate to yourself that this kind of operation is kind of a
plot in velocity space for a given feature). So if you have two lines
that have identical optical depth profiles (regardless of what their
parent ion is), they will have the same shape no matter what
wavelength they have at rest. By plotting, say, lambda * f_lambda you
flatten out the spectrum and bring up the red end of the spectrum.
I think David goes even further. This reduces the flux contrast, and
you know that we can't ever get the flux right, we can only hope to
get the overall shapes of line profiles correct. This makes it easier
to do plots where you fit rather shallow infrared features, like the
Ca IR triplet.
Finally, I'd like to suggest a more descriptive title for the paper.
I'm not sure the C you refer to in the title qualifies as ``high
velocity'' especially since the optical depths go below 0.01 before
the v_max you set was reached. How about ``Direct Analysis of the
Type Ia Supernova 1999ac before Maximum Light: Further Evidence for
Carbon.'' This puts your work into context with the papers on 1990N,
1998aq, 1999aa and doubtless the others which will come along as
higher S/N spectra are obtained. And most importantly, it contains
the phrase ``direct analysis.'' Some people in SCP are confused by
this, but we will change that. :)
Happy holidays,
Rollin
This archive was generated by hypermail 2.1.4 : Thu Jan 01 2004 - 02:41:04 PST