From: Robert A. Knop Jr. (robert.a.knop@vanderbilt.edu)
Date: Fri May 02 2003 - 15:49:06 PDT
On Fri, May 02, 2003 at 09:07:29PM +0200, Reynald Pain wrote:
> It may just be a wrong impression but most of the ground based data seem
> to be above the lightcurve template. It looks like there is some
> systematic shift in HST vs groundbased zeropoints ?
There may well be. For most of them, it's not very serious; the only
one that looks particularly egregious is the R band lightcurve SN1998ay.
That one has an uncertainty of 0.05 on the ground based zeropoint, which
is an error 100% correlated between all of the groundbased data points
(but not, I believe, included in the plotted error bars-- only in the
covariance matrix used in the fit; I could be wrong about that).
None of the I band lightcurves appear to have an appreciable systematic
offset. For the other R-band lightcurves, beyond SN1998ay, the only two
I might question are 1997ek and 1997ez. (1997eq has a couple of
large-error-bar points which are high, but the rest of the ground points
are much closer.) Those aren't as off as 1998ay. Interestingly, they
are also the two at highest redshift; problems with the uberspectrum in
the U-band region could lead to this kind of systematic offset in the R
band at those redshifts. I'm not sure I really want to get into this at
this point; there's nothing we can do about it without having much
better spectrophotmetry out to the blue edge of the U-band on
supernovae. I suspect what *might* happen is that we'd clean up that,
get slightly different R-I colors for those supernovae, and then decide
that we really got better fits with a U-B=-0.5 than U-B=-0.4 as I used,
and have the same cosmology. (The peak magnitudes were estimated from
the I band lightcurves in this case.) I'm a little hesitant to say too
much about this in the paper, because it's difficult to state without
opening a pandoras box. On the other hand, not addressing it may be a
bigger pandora's box.
Probably all I would do is:
(1) observe that there appear to be systematic offsets in R-band
lightcurves between the ground and HST based data at the highest
redshifts; in the case of 1998ay there is a fairly large
correlated error between all the ground based data points.
(2) these offsets could result from imperfections in the template
spectrum in the rest-frame U-band, where spectrophtometry is
weakest.
(3) this will not affect our primary cosmological fits, because at the
highest redshifts m_B is estimated from m_I, where there is no
systematic offset, and where K-corrections are more dependent on
the much better BV range of the template spectrum;
(4) The consistency of E(B-V) at higher redshifts argues that what we
are doing between the K-corrections of HST filters and the assumed
intrinsic U-B colors used in color excess estimates is yielding a
correct and robust analysis.
(5) we will consider in the systematic error section the effect of
a simple change to the U-band region of the uberspectrum,
specifically smoothly making the U-B color bluer by 0.1
magnitudes, and finally
(6) as more and better spectrophotmetry becomes available, we will be
able to refine K-corrections so that in the future the U-band
region will be as reliable as the B and V band regions today.
The trick will be to do this in such a way that we don't completely
undermine the E(B-V) analysis for the highest redshift supernovae.
Probably only the points 1 shoudl be noted in the section discussing the
lightcurve fits, with a pointer forward to the "colors and extinction"
section for the rest of this discussion.
-Rob
-- --Prof. Robert Knop Department of Physics & Astronomy, Vanderbilt University robert.a.knop@vanderbilt.edu
This archive was generated by hypermail 2.1.4 : Fri May 02 2003 - 15:49:11 PDT