From: Alex Conley (aconley@panisse.lbl.gov)
Date: Fri Jan 02 2004 - 13:24:06 PST
Hello all,
I hope you had a pleasant holiday season.
We've discovered that the filter transmission functions that we have
used in the past for the Bessell filters are slightly incorrect.
Basically, the passbands for UBVRI given in Bessell (1990), Table 2, are
not what we thought they were, and hence our K-corrections have been
slightly incorrect in some of our previous work (including Knop et al.
2003 and possibly the 42 SNe paper). This is NOT a large effect, so
nobody should get too worked up about it, but it's worth pointing out.
The problem is that the numbers that Bessell gives in his table are
actually not R (the dimensionless response, or transmission, or whatever
you feel like calling it), but actually lambda * R (suitably
renormalized).
Note that this is absolutely not the energy vs. counts issue again. It
takes a similar form mathematically, but there is no physics in this one,
just a question of a paper that isn't as clear as it should have been.
If you actually read Bessell (1990) you will find no explanation of what
he means by passband, and so it is very natural to (incorrectly) assume
that the numbers in Table 2 are just the dimensionless response. In
subsequent papers is he much more clear about this issue, and always
tabulates lambda*R.
This discovery arose when Lifan pointed out a footnote in Jha's thesis
that points this out, and references some papers by Suntzeff that have
also discussed this issue. I ended up contacting Nick Suntzeff about
this, and after a few email exchanges I contacted Mike Bessell directly.
He confirms that the numbers in Table 2 of Bessel 1990 are actually
lambda*R, and not R. Therefore, the real filter is slightly bluer than we
have been using. Fortunately, as I said before, the size of the effect is
not large, generally less than 0.02 magnitudes. If you would like to see
the size of the effect, take a look at
panisse.lbl.gov/~aconley/kdiff.eps
This is the old K correction minus the new K correction (the new one using
the right form for the Bessell passbands) for a variety of redshifts.
The blue lines are rest frame B band, the red lines rest frame V band. The
sign of the effect is such that SNe are actually slightly dimmer and
redder at higher redshifts than we believed (recall that K-corrections are
subtracted from observed magnitudes). At z=0.5, a SNe is about 0.015 mag
dimmer and 0.012 mag redder with the correct filter. With extinction
correction, these counteract each other to some extent. This makes the
Omega_Lambda case a little bit weaker (with extinction correction), but by
an amount much smaller than we care about -- so far.
So -- how do you figure out if this affects you? Well, take a look at
whatever dimensionless filter passband you have been using. If, for the
I band it looks something like
lambda R
700 0.0
710 0.024
720 0.232
730 0.555
740 0.785
750 0.910
760 0.965
770 0.985
780 0.990
790 0.995
800 1.000
etc.
you are probably affected. If this is so, you have to decide if the size
of the problem is big enough for you to redo whatever you have been using
them for.
Note that this lambda*R thing does not affect all passband information.
For example, as far as I can tell (I would appreciate somebody
cross-checking this) the HST filters that synphot provides are simply R,
and not lambda*R. I think that the Sloan filters as specified by Fukugita
et al. are also R. Some other papers (the Y band of Hillenbrand, the Js,
H, Ks of Persson) don't define their terms clearly enough that I can be
sure either way.
This issue has given me a new appreciation of why the HST folks decided to
define their own passband formalism instead of trying to work with the
very messy and creakily defined standard filter stuff.
Incidentally, the full reference to Bessel 1990:
Bessel, M.S. 1990 Pub. A.S.P. 102, 1191.
Alex
This archive was generated by hypermail 2.1.4 : Fri Jan 02 2004 - 13:24:34 PST