From: Robert A. Knop Jr. (robert.a.knop@vanderbilt.edu)
Date: Tue May 20 2003 - 04:47:35 PDT
On Tue, May 20, 2003 at 12:14:39AM -0700, Greg Aldering wrote:
> I was comparing your fits to those for the 42, sparked by Gerson's
> comparison table. You may have noticed that 9739/97S is almost 0.3 mag
> different in your table than in P99. I have traced this to differences
> in kcor1. If I plot the difference in kcor1 versus redshift there is a
> smoothly increasing difference in going from z=0.18 up to 0.66 or so.
> The difference is quite large, but 9739/97S stands out from this trend
> by about 0.2 mag. Indeed, it disagrees with SN with very similar
> redshifts. Could this be a problem?
>
>
> - Greg
>
> P.S. Beyond z ~ 0.66 there is also a big change in the differences, but
> I assume this has to do with which filters were used.
That latter part may have to do more with what days are used? I think
the same filters were used. Unless you mean that kcorr1 was U-R in the
old table, whereas it's B-I now. In that case, for z>=0.7, compare
kcor1 to the old kcor2.
As for the differences in K-corrections: the U-B intrinsic color will
make some difference in this, esp. as you go to higher redshifts. How
different are the other ones that aren't 9739? Both the high-redshift
B-R and the U-R K-corrections can be explained by a U-B difference. As
I worked out before, there's an implicit difference of (if memory
serves) 0.05 (at least) in the U-B color used in P99 and now. This
might entirely explain "most" of the differences.
I think there's also an implicit B-V color difference as well-- again,
it's hard to figure out exactly what colors are assumed in the spline
files, but if memory serves I got something closer to 0 than -0.05 or
-0.1.
What's more, last time the K-corrections were not adjusted for stretch
or color. (At max-- the epoch used in finding the K-corrections
considered stretch.) I know Saul will assert I'm wrong in this, but I'm
pretty sure that I'm not; did you ever ask Don about that? The "spline
files" we use in snminuit have a given K-correction on day 0 for a given
redshift, and don't change based on anything else.
I do adjust K-corrections based on E(B-V)_host, E(B-V)_mw (which may
have been implicitly handled previously?), and the stretch/color
relationship. With 9739, it's probably E(B-V)_host making the biggest
difference, since the E(B-V) host value is so huge and negative. (In
fact, in general, the stretch/color relationship probably makes little
difference in the K-corrections.) Additionally, at z around 0.6-0.7,
that's where it's going to make the biggest difference because of
misalignments between the filters. What this really implies to me is
that, especially where we have huge color errors, we probably ought to
be taking into account propogated K-correction uncertainties, as they
are probably not negligible around the 'bad region' of z~0.65. I
suspect they still are negligible at z=0.4-0.55 and z=0.7-0.9, even with
large color uncertainties, but with large color uncertainties the
"extrapolated" K-corrections get scarier.
Practially speaking, the only supernovae likely to be seriously affected
by this (due to being in the "bad region" of the K-corrections, having a
ebv uncertainty > 0.2, and an |ebv|>0.2) are 972 (E(B-V)=-0.56,
dEBV=0.53, z=0.76), and 9739 (EBV=-0.73, dEBV=0.45, z=0.61), and *maybe*
9748 (EBV=0.14, dEBV=0.39, z=0.59). Most of the others either have
smaller color errors, smaller EBV's (so that color dispersion is
unlikely to be "screwing up" K-corrections), or are in the redshift
ranges where the K-corrections are more solid (cf: the error band in
either Alex or Peter's K-correction papers).
-Rob
-- --Prof. Robert Knop Department of Physics & Astronomy, Vanderbilt University robert.a.knop@vanderbilt.edu
This archive was generated by hypermail 2.1.4 : Tue May 20 2003 - 04:47:38 PDT