Peter summarizes what he does with the K-corrections. He has a time series of flux-calibrated, composite supernovae spectra everywhere from -10 to 40 days, everywhere from 2000 to 10000 angstroms. He flux calibrates by tilting the spectra to match the UBVRI photometry that Adam Reiss gets from his MLCS stuff.
He interpolates, and extrapolates a little on the end, to fill in every day. The gap depends on the wavelength range. In the IR, it's sometimes 5-6 days, same in the UV. In the optical, it's less (presumably). The extrpolation to earlier times.
Peter says that if you have a handful of spectra all within +-5 days, and you force them to have the same color, you get the same K-correction.
For stretch, Peter asserts that the thing driving the difference is that different stretch supernovae have different colors. Ariel questions whether the color/stretch correlation is that convincing. Peter says it's not great, but it's also not horrendous.
He has supernova spectra, wavelength, and time. You have to give his program the color (B-V) somehow -- because of stretch, E(B-V), whatever -- and it spits back a K-correction that should be pretty good.
How are these K-corrections used in the program to fit the lightcurves? Right now, you stretch the K-correction along with the template lightcurve. Peter asserts that this only introduces 0.01-0.02 magnitudes at even the worst redshifts. This stretched K-correction is an approximation to K-correction as a function of stretch.
Note that the uncertainty on the K-corrections does not currently go into the snminuit fit. However, the assertion is that the uncertainty on the K-corrections are a lot smaller than the uncertainties on the individual photometry points.
Issue: should K-corrections be done in counts or in Energy?
Isobel asks if the K-corrections were done using an average CCD response? Peter says yes... he's used the Harris R and I filters, which (he asserts) has been mixed in with a CCD response. Peter does say something about mixing it all together at once when we do the HST, which makes Rob think that he should pay attention to what is going on when fits to HST+ground are done.
Now we're talking about extinction and what R_B means, but for various reasons that discussion isn't transcripped.
Rob talked about photometry, and so didn't type while he was talking.
Gerson talks. He's been "for some time" in the process of writing a paper. He's done this work on 40 SNe (hasn't been extended to 42). He threw out 3 at z>0.7 and 2 at z<0.2, because there R band doesn't map to B band. This leaves him with 35 SNe. He threw out 9733, because it's pretty severely an outlyer on the Omega/Lambda fit... doesn't seem to belong to the regular type Ia. It's too dim for it's z value by an order of magnitude.
Gerson plots our 34 supernovae, separated into z=0.3-0.45 and z=0.45-0.7, and plots them together with the Hamuy supernovae. Note that the Hamuy data doesn't have anything in the early times, whereas we have a fair amount of data on the early times.
The subject of this paper Gerson is working on is 3-fold. "Type Ia supernova lightcurves: cosmological time dilation, a single-parameter description, and the turn-on behavior."
Gerson takes all of the photometric lightcurves. He normalizes them to unit flux at the maximum of the fit. Plotting everything on top of each other is a mess. He then averages the three different subsets. This gives a clear indiciation of time dilation -- the higher redshift supernovae have broader lightcurves.
Next, Gerson applies K-corrections to everything, and then divides by (1+z), which takes you to the rest system in time. Then, all of the points coallesce within their errors. This demonstrates the time dilation. However, it's still not quite right.
Finally, Gerson takes out the measured stretch of each individual lightcurve. He divides by the stretch. Some object that this is a circular arguement.
Gerson says we now have a curve that goes all the way down to zero on the rising side. We've got data there, whereas Hamuy didn't. He calculates the explosion time to be -18.8 days +- 0.2 (error is still uncertain) times the stretch. The key is that this one parameter seems to be enough to describe the shape of the curve.
There is discussion as to whether 9733 should be omitted. Peter says that that supernova looks like it's just reddened.