From: Andy Howell (DAHowell@lbl.gov)
Date: Thu May 01 2003 - 18:24:35 PDT
Greg Aldering wrote:
>Hi Andy,
>
>I see, basically your program threw away everything below 4000A in the
>restframe, i.e, the Ca feature that is present in the observed spectrum,
>to obtain your Ic fit to 98104. I understand the constraints imposed by
>the limited spectral coverage of the template spectra, but throwing away a
>key portion of the spectrum is obviously not the approach to take in this
>case.
>
>
Greg,
No it did not throw away everything below 4000A for every fit, it only
did so fit the
fit I chose to show, which I picked semi-randomly. That is the point --
nothing fits
that break!
Again, I put no faith in that or any of the fits to that SN. The data
are just too bad
to say anything conclusive. I have a slider which lets you control how
much overlap
you want to require. I can set it to 100% to require full wavelength
coverage, but
that cuts down the number of possible matches at high-z.
>Also, I don't know whether bluening was important in getting a Ic fit for
>this spectrum, but I think it is important to impose some limits on the
>amount of bluening allowed. The only mechanisms for bluening are flux
>calibration errors or atmospheric dispersion combined with guiding at a
>blue wavelength. I doubt that the flux calibration errors are very large,
>and since the Keck guider response peaks in R-band, atmospheric dispersion
>should not be bluening the spectra much at all. (A check I did for a pivot
>star that was in the slit when we observed Albinoni, using LRIS just as
>for many of these SNe, showed excellent agreement with a model spectrum
>for the pivot star.)
>
>
No, there is another major reason for allowing the fit to go blue -- in
most cases the
template spectra have not been corrected for reddening, because the host
galaxy
reddening is unknown. This can make a big difference in Ib/c's
especially. SN 1994I,
for example is thought to have Av~2 if I remember correctly. But I have
a control
for that too in my fits so that if you want to limit that, I can do that
too.
>I understand that it is hard to code for all the different types of
>observational situations which may arise, but to remove from the Hubble
>diagram a SN based on such hard-won and expensive data, without first
>imposing every realistic constraint on the spectral fits, just doesn't
>make sense to me.
>
>
Actually, all of this is already coded for, I just chose the parameters
for my fit
that from experience cover the broadest range of possible situations
that might arise.
And you see from 97226, which is at higher redshift than the other two,
if there is
an appreciable SN signal there, it comes shining through with the
parameters I have
chosen.
I would say it another way: Why do we spend all of these hours on
spectroscopy if
we are going to ignore it if we don't like what it is telling us in the
end?
Again, I am not saying these are not Ia's -- it looks like they are from
the colors and light
curves. The spectra are simply inconclusive. We can't say (as the
draft does now),
"all eleven in the set have strong [spectroscopic] confirmation as type
Ia..."
But I do agree that we should perform the best analysis we can on the
data we have.
I think I have done that, but I am willing to do it whichever way you want.
>Do you have some fits that are better in this regard
>
You tell me what you want, and I will run it. I can require full
wavelength coverage,
no blueening, limit the host galaxy contribution, etc.
Cheers,
-Andy
>Cheers,
>
>Greg
>
>
>On Wed, 30 Apr 2003, Andy Howell wrote:
>
>
>
>>Greg,
>>This mainly a psychological effect of the rebinning -- I rebinned the
>>data to 20A.
>>I put up on the web page the same plot rebinned to 3A -- you can see it
>>looks like the
>>untouched data.
>>
>>Also, the spectrum I show is cropped to only show where the template
>>spectrum and
>>the observation overlap. I put up another plot showing a different fit
>>with more of the data.
>>
>>My program has the freedom to do other things to the data as well. In
>>my plots I have alread
>>subtracted host galaxy. Plus, my program either reddens or bluens the
>>spectrum to find the best fit.
>>
>>So the spectrum I show may not correspond to reality -- I let it mangle
>>the observed
>>spectrum to find the best fit possible. Often it will find junk, but
>>the point is that
>>if there is a Ia signal there it should find it. Here I just chose one
>>example to show,
>>but I don't believe it -- I don't believe any of the fits.
>>
>>-Andy
>>
>>Greg Aldering wrote:
>>
>>
>>
>>>Hi Andy,
>>>
>>>I am looking at the spectral fits to 98104 on your website. I don't see a
>>>relation between your spectrum and Isobel's. Am I even looking at the
>>>right thing? (Yours is labeled "Observed: 98104cc_comb.Ic.asc" and
>>>Isobel's is labeled "98104 (Keck) : 98104cc_comb.asc". Yours is a
>>>power law and Isobel's have plenty of wriggles.
>>>
>>>
>>>
>>>
>>>
>>>>Due to the E's 4000A break, you subtrack off more light blueward of 4000A
>>>>than redward, and this can mimic Ca from the SN.
>>>>
>>>>
>>>>
>>>>
>>>No. The effect of subtracting an elliptical is to decrease the strength
>>>of Ca - unless you let the elliptical go negative! But I do agree that
>>>9878 isn't convincing.
>>>
>>>- Greg
>>>
>>>
>>>
>>>
>>>
>>
>>
>
>
>
This archive was generated by hypermail 2.1.4 : Thu May 01 2003 - 18:24:39 PDT