From: Serena Nobili (serena@lpnhep.in2p3.fr)
Date: Tue Jul 20 2004 - 02:29:11 PDT
Dear Chris,
thanks for your comments. I agree, it is a pity not to be able to meet
during the SCP meeting this year. I hope we will have soon other
opportunities for sharing a beer.
The spectral template I used for the paper are available from the web page:
http://www.physto.se/~snova/internal/papers/iband/#changes
Remember, however, that this is not optimal for k-corrections
involving rest frame U-band.
I have implemented part of your comments, and I will update the
version in the web page as soon as possible. I answer below to some of
your questions, and to those comments or changes we would need to discuss
further.
On Sat, 16 Jul 2004, Chris Lidman wrote:
>
> Section 2.2
> -----------
>
> In the second last paragraph
>
> "The uncertainties for this supernova may be underestimated."
>
> I find this to be too vague. It might be true, but do you have any
> evidence other than the poor fit that this is true. Otherwise, it
> would be better to leave this sentence out. It might be that the
> template is a poor description for the light curve of this
> supernova. Indeed, the latter is more likely to be true because the
> fit to the early and late parts of the light curve (figure 1.)
> clearly show a systematic trend. Compare the rising part of 94D with
> the rising part in 94ae, 95bd, 92bo, 92bc, etc. Compare the tail 94D
> with the tail of 94M.
>
I wrote this because also Rob found the uncertainties for this SN to be
under-estimated. Thus, I have increased the uncertainties to the
floor used by Rob in his paper. However, I agree with you that most
probably the main problem is in the template. As I think the text says.
> Section 3.
> ---------
>
> For my own benefit, geometrically means
>
> (sigma_intrinsic)^2 = (sigma_total)^2 - (sigma_measured),
>
> correct?
>
(sigma_intrinsic)^2 = (sigma_total)^2 - (sigma_measured)^2
where sigma_measured is the average r.m.s. measured on the simulated
data sets, and sigma_total is the r.m.s. measured on the real data.
> Section 4.1.
> -----------
>
> I think we can improve the discussion on the cross filter k-corrections.
> There has been quite a bit of work on trying to relate the IR and
> optical photometric systems and we can say that the we know the
> relationship
> to 2%. I attach an e-mail which I had sent out on May 16th. The
> most important part of the e-mail is that we need to add an extra term
> to the IR to optical k-correction which accounts for the small
> difference
> in IR and optical systems. The text should be changed to reflect this.
>
I will extend this part by including the "new" formula for
computing the k-correction asap.
> Section 4.2
> -----------
>
> How different are the J to I k corrections for 99ff between the Serena
> mark I and Serena mark II templates?
>
See the attachment: kcorr_99ff.ps
> For 99ff and 99Q, I'd like us to come up with a more diplomatic
> way of stating that the other group are wrong.
>
> For 99ff, simply drop the sentences starting with "We found differences
> ..." and ending with "...communication." You can also drop the section
> saying that the I-band magnitudes have been adjusted, since I presume
> that
> this will now be absorbed in the k-correction as described below.
>
> For 99Q, say that "The mean difference was found to be xx magnitudes
> rather
> than quoting the maximum magnitude difference."
>
I am definitely pro diplomacy. However, by not saying that they made a
mistake, simply referring to our way of computing k-corrections, could
rise questions about the reliability of the k-correction calculations in
general. The "careful" reader, could conclude that we simply get different
results because we used two different methods or spectral template for
computing K-correction (not careful readers will not even notice the
difference). While this might be true (e.g. for 99Q), this is not the
problem for 99ff, for which they used the wrong filter. I don't quite
know how to make this clear, while being diplomatic. I am open to
suggestions.
> Fit to the high z SNe.
> ---------------------
>
> The following is not in the paper and it is something that we do not
> have to add, but it is something that should be done at some stage.
>
> The high redshift SNe are poorly sampled in time. We have explored the
> systematics of how photometric errors might affect the
> fits for both the high and low z SNe, but we have explored the
> systematics caused by poor sampling.
>
As you know, I am convinced that the Monte Carlo I am performing now, does
test for systematics caused by poor sampling. However, I will try to check
for this after the summer.
>
> Section 6
> ---------
>
> For the R_V = 4.5, you mention that the ellipses would only move by 0.03
> and
> 0,04 magnitudes in E(B-V) and E(B-I). Can you please check this.
>
As you see in figure 13, the two dust curves, corresponding to Rv=4.5 and
Rv=9.5, are very close to each other. The step between the two curves does
not depend on the time, since the dust density does not vary. Therefore,
the same offset is in the ellipse graph. I think it is correct. For more
details about the models, you can directly ask Ariel, who generated the
models for me using SNOC.
> In the next paragraph you mention that R_V=4.5 is disfavored at the 90%
> confidence level and that something, presumably R_V=9.5 is disfavored
> at the 97% level.
>
I am sorry if it is not clear. Both percentages refer to Rv=4.5, in
the host galaxy extinction corrected case and not corrected case.
> Figure 13
> ---------
>
> Could you also double check figure 13. It surprises me that there
> is so little difference between the R_V=4.5 and R_V=9.5 models.
> I would have expected the R_V models (dotted line) to be much closer
> to the solid line. Have you used the formulas in Cardelli, Clayton and
> Mathis ApJ 345, 245? The formulas are very involved (see table 3 and
> equations 2a to 3b). I imagine that the calculation for SNe at z~0.5
> must be very complex because one has to take into account the redshift.
> Dust at low redshift will have a relatively low effect when compared
> to the same dust at high redshift because the spectra are already
> redshifted.
>
I am pretty sure of the dust models. Please, refer to Goobar et al. 2003
for more details, or again, ask directly to Ariel. In brief, the value of
Rv corresponds roughly to the minimum size of the grains, i.e. the
minumum size of grains between 0.08 and 0.12 micron (as proposed by
Aguirre) corresponds to Rv between 5.5 and 9.5. The larger the minimum
size, the lower the dependence on wavelengths of the dust properties, the
closer the corresponding colour curve will be to the no-dust model (that
is why in my plots the curve for Rv=9.5 is closer to the no-dust than the
Rv=4.5). I believe the parametrization by Cardelli was also used in the
calculations. I cannot help you more, because I know what I learned by
reading the paper and discussing with Ariel about it.
> Figure 14.
> ----------
>
> I think that you should avoid using E(B-V) as the vertical axis as the
> reader might take it that dust is the cause for the spread in
> SNe colours. I think \Delta(B-V) = (B-V)_SN - (B-V)_local may be
> clearer.
>
You mean both axes, not only the vertical one, right? Isn't the difference
between the B-V measured on the SN and the average value measured on the
local sample, exactly the definition of E(B-V)? I know it can be
confusing, and I got confused many times myself. However, every time I
try to change the labels of figure 14, I end up by
going back to the labels you see there now. In fact if you think about it,
that is exactly E(B-V). Now, if we found that that is not zero, then we
have a problem, e.g. presence of dust in the IGM.
> Conclusion
> ----------
>
> In the last sentence, remove the sentence
>
what sentence?
> SN 99ff and SN 99Q
> ------------------
>
> In the revised version, 99Q is even more of an outlier? 99ff is less
> problematic, but there are only two data points in the light
> curve. Since these SNe were not observed with by the SCP, do we want
> to keep them in the paper, or do we keep just 99ff.
>
> I am not advocating that we remove them, but I think it is something
> we should openly discuss. From might point of view, here are the
> negatives
>
> The negatives of 99Q and 99ff
>
> - We are relying on the photometry of the other group.
>
> - We find significant differences in the k-correction
>
This is true only for 99Q (Brian has admitted they reported the wrong
value in the paper). Moreover, Riess uses one value for k-correcting all
epochs for 99Q. This is really difficult to understand. Any spectral
template you use, would give you a k-correction varying with time.
> - We've already excluded one their SNe because we it was probably
> heavily reddened.
>
It was surely reddened by the Milky Way. This was observed near
the galactic plane.
> - For 99ff, there are only two points in the light curve.
>
I know this isn't great.
> - I have not see a spectrum of 99Q. It is faint and the second maximum
> is not convincing. If it were not for the B-band stretch, one could
> argue that it is an sub-luminous SNe, both from the I-band maximum and
> the
> lack of a second maximum.
>
I am afraid the I-band lightcurve of 99Q is not compatible with a
sub-luminous SN. The second peak is not prominent as in Beethoven,
however, the lightcurve shape will not agree with a sub-luminous SN. That
is, if we trust the time of B-max given by the optical lightcurve fit (but
it would be difficult to fit the I-band with a sub-luminous one in any
case). The chisq to the template of 91bg or 97cn is about 21 or 27 worse
than in the case of best fit (see table 13). To conclude, the I-band shape
of 99Q does not support the hypothesis of it being sub-luminous.
> The positives about Beethoven
>
> - The Beethoven observations are more precise as we know the Js filter
> curve very precisely (I measured it myself).
>
> - We measured all aspects of this SNe.
>
> Overall, I do not think we should put these two SNe in the paper
> unless we are comfortable with the work that has been done by the
> other group, and, quite frankly, I do not feel at ease with their
> work. I do not think that we would weaken the paper if these 2 SNe
> were excluded; however, the last couple of sections in the paper
> would be considerably shorter.
>
I have to agree on the fact that we don't control the work done by the
other team, although Vitalyi has been able to check the optical lc for
99Q. However, we should keep in mind that this is mainly a feasibility
work, at least for what concerns the high-z SNe, both in the Hubble
diagram and in the test for IG dust. We don't base any strong conclusions
on the high-z SNe. I think excluding these 2 SNe from the paper will
diminish the significance of the work by a lot. I believe the paper is
interesting, apart from the analysis on the low-z sample, also because it
shows that it is feasible to measure high-z SNe in the J-band and use the
measurements for doing cosmology, although there are possible systematic
effects involved that we need to understand. Given the uncertainties
involved, both statistic and systematic, none of the high-z SNe is really
outlier.
Cheers
Serena
----------------------------------------------------------------------
www.physto.se/~serena
Tel +33 1 44 27 75 82
Give free food at:
http://www.porloschicos.com/
http://www.thehungersite.com/
This archive was generated by hypermail 2.1.4 : Tue Jul 20 2004 - 02:29:24 PDT