From: Robert A. Knop Jr. (robert.a.knop@vanderbilt.edu)
Date: Sat May 03 2003 - 10:21:06 PDT
On Fri, May 02, 2003 at 05:32:20PM -0700, Greg Aldering wrote:
> z SN % error Source
> ---------------------------------------
> 0.53 97eq +10% WIYN
> 0.86 97ek +10% BTC & WIYN
> 0.78 97ez +15% BTC
> 0.44 98aw 0% BTC & WIYN
> 0.64 98be +10% BTC
> 0.64 98ay +15% BTC
> 0.50 98ax 0% BTC
> 0.36 98as 0% BTC
> 0.74 98bi +10% BTC
> 0.54 00fr 0% Keck & NTT
* I've checked the ground-based zeropoints; I believe they are good.
Most calibrations for *all* of these lightcurves are coming from BTC,
and what tends to have the most weight generally is the final
reference run. This isn't always the case, and in (I believe) all
cases there are at least two nights with confirming photometric
calibration.
* I've checked the ground-based relative photometry (ratios), and
believe they are good.
* I've checked the aperture corrections on the ground-based data, and
believe they are good.
* I've checked the HST CTE corrections (using the new Dolphin
coefficients), and they didn't change more than <<1%.
* I've played with using energy instead of photon units for color
corrections (that made things worse, which is probably a good thing)
* I've played with using a different U-band filter description I found
on the CTIO website (that made very little difference)
* I've played with different ways of mangling the uberspectrum to get
the U-B colors that I want in it (that made very little difference).
(Note that we just have photometry in various integrated filters, and
you can slope the spectrum various different ways in the (say) U-band
and still get the same broadband colors... but you can over or
underweight one side versus another.) I suspect that the real answer
lies in here somewhere, but it's nothing I can do with the data I've
got available. We need really good real spectrophotmetry in the U
band of many Ia's, I think.
There you have a summary of the last ~18 hours of my life... (there was
four hours of sleep in there somewhere too).
One thing to note is the redshifts (added to your list above). It's all
the *highest* redshift supernoave that are affected (except for 97eq,
which has a relatively large ground-based zeropoint error of 0.05-- that
will be correlated between all ground based points). This may suggest
it's got something to do with the U-band region of the template
spectrum.
I will put a note about this into the text, stating that this should be
revisited as better U-band spectrophotometry becomes available, but that
in the mean time (a) our primary cosmological analysis will be hardly
affected, since we're getting B band from I band for z>0.7, and (b) we
will investigate the systematic effect of bluening the U-B of our
template spectrum on K-correction and U-B analysis. (That doesn't bring
the ground and HST data together, though, so there is probably more
going on.)
The Bessel R-band filter extends more to the blue, and has its peak to
the blue, compared to the F675W filter. There are probably some issues
with color corrections. If anything, I suspect that the K-corrections
are likely to be more robust from the F675W filter; at z=0.86, it's
almost a perfect match to the U-band filter.
What is the answer? I don't know. But I don't think I can make those
differences go away doing anything reasonable at the moment. Also, for
the reasons stated, I don't think this is seriously biasing our
cosmology or even our E(B-V) analysis. It's mostly just ugly.
-Rob
-- --Prof. Robert Knop Department of Physics & Astronomy, Vanderbilt University robert.a.knop@vanderbilt.edu
This archive was generated by hypermail 2.1.4 : Sat May 03 2003 - 10:21:33 PDT