Greg, HST/NICMOS proposal status. It's due this Friday the fifth, Greg believes 8PM or 5PM, Gerson wants to know which standard time.... Ahhh, the conversations of compusive last-minuters
Greg says that talking to Andy Fructer, apparently NICMOS doesn't perform even as well as the Web page claims its post-launch performance is. With IR arrays, normally with multiple readouts you beat down the read noise. Andy Fructer, however, says that this readnoise is highly correlated, so that even by multiple reads you don't get the read noise better than to 20 e- or something. (They don't know why the multiple reads are correlated.) So, whereas before we were in the photon limited regime and could use NICMOS 1 which had things like the F140 filter, now we're in the read noise dominated regime, even with the SN at max at z=0.5. This forces us to NICMOS 2. This means that we can't use F140, since that's only on NICMOS 1... so we're stuck with F110. Also, the quality of the photometry will be somewhat worse than we had been predicting.
Additionally, Andy strongly recommended dithering, (cosmetic problems, persistent comsic ray problems, spots of bad QE, want to avoid all these things) but as you break it into more exposures, you get the read noise in more times. Non-negligible chance that your object would get hit if you don't dither. Even though we're looking at a very small spot on the array, Andy didn't think that they odds were low enough of getting nuked to make it worth risking not dithering.
CR rejection mode, where you keep reading it out throughout the exposure. This gives you the ability to see CRs as they hit, and reject them. Howver, the cosmetic problems and persistent CRs can't be dealt with this way.
Right now, Greg has made modifications based on discussions with Andy and Saul. People should have a look at this... soon, since it's due on Friday.
Other question is whether or not they are dividing into small/medium/large proposals. We're asking for 42 orbits (out of 1500 available). We don't know what size this would be or if it matters. Even yet another question is if the TAC will be told a limit as to how long NICMOS will be around. The worry is that the TAC will see us asking for time after the putative death date of NICMOS, and we will get flat out buzzed.
Rob should start looking at the Keck proposal and figuring out what needs to be changed and so forth.
Also, we should talk to Tom Broadhurst again about looking for high-redshift lensed supernovae in bright time with LRIS at Keck. The idea is to take images in I-band (or somethin glike that) over and over again once a month at clusters that Broadhurst is interested in anyway. We'll look to see if any background SNe, a mag or two brighter due to the lensing. Lunctime rate calculations suggest that we ought to get a few, especially if the rates go up at SN~1 as things like Pilar's calculations suggest. The SN would be useful for rate calculations, perhaps, and also for mapping the cluster. Since they want that data anyway, we ought to look into this. (Who will be in charge of this?)
Alex is running templates again; the computers crashed over the weekend, so he lost some time. He wants to finish what he's doing right now before he leaves... Saul is pushing for the single color version.
Reason for single-color: one he's doing now has been optimized to model the relationship between R and I to reduce the dispersion of the lightcurves. However, you might be able to better (and also have the advantage of comparing apples to apples) if you do a single color template for a single color fit. Alex is going to pass these templates to Greg before he leaves.
Color due to intrinsic color variations of different stretch SNe and color due to extinction. There was a problem with Adam's form that there was crosstalk between the two colors. To what extent is ours clean? Alex claims that it's not clean because he assumes a certain peak V for a stretch=1 supernova, but no assumption about peak B. When he puts in the nearby SNe to the fit, he has to specify which don't have extinction. He does this by using earlier type galaxies (E/S0) and higher galactic latitude. Right now, the calibration is fixed assuming no extinction. (To check this, you should plot a color histogram, to see a tail due to extincted SNe.) Earlier verison, Alex thinks he was getting the templates too blue. However, that shouldn't be the case any more.
Saul reports that Ariel said that the relationship between width and brightness wasn't as strong in V as in B. Saul showed a plot (a "M&M plot") which he and Alex did, showing that using a stretch corretion gives a nice lineup. First step was aligning the vertical scale at maximum. Next step is to realize that they don't all have the same maximum. Narrow ones are fainter in V, broader ones are brighter V. Lines plotted are Alex's tepmlates with stretch correctio put in, and they seem to align with observed SNe. Using a linear relationship between stretch and height, all the corrected SNe lightcurves line up well with each other again. (Saul had the plots for B as well, and it worked there too.)
Saul suggests that perhaps the difference was that Alex considered extinction and Ariel didn't. Alex replies that if extinction was the effect, it should have screwed up B worse than V. (The report is that Ariel did see the effect in B, not in V.)
As long as the extinctions are well understood, this plot looks to be convincing about the existence of a stretch/brightness relationship in V.
Theoretically, Sebastien knows how to fit Omega and Lambda countours even when Alex isn't around, so we should still be able to do that after Alex is gone. So far he's done it fitting script-M at the same time. Alex should make sure that Sebastien has seen how Alex runs that program before Alex vanishes. Also old results should be compared with new results.
Alex leaves a week from Friday. Sebastien is going to France for something like a few weeks to a month. Craig is also leaving; he'll be working for Remote Sensing Systems in Santa Rosa, a small company that takes raw satellite data and turns it into climate variables over the oceans. The guy who runs the company has the only succesful data algorithms. One of Craig's first jobs will be to figure what to do about El Nino. Most of it is funded by NASA's "Mission to Planet Earth" project.
Visitation schedule: the Richards will be in town sometime around October 25th. Sebastien will bring news to Reynald, and see if he can be in town then also.
Portugese students are coming around October 15th or something like that. Saul may have their E-mail addresses, and Jeanne Miller may have them. Peter wants to impose a tarif of 2 cases of port, plus one case of port for each week it takes them to learn IDL.
Peter, K-correction: K-correction depends on extinction as well as SN stretch. (We've been ignoring both of these before now, because Alex's first paper didn't seen any evidence for it in the few spectra we had.) Peter drew on the whiteboard some sample K-correction-over-spectral-feature plots, and the effects of when it gets reddened. If emission-type features which are relatively strongly affected by reddening (in other wods, features on the blue side of the filter) move into the observed band (R) (relative to the rest band (B)), then the K-correction is significantly different than it would have been without extinction.
Some worry that the extinction correction might be double-counted. Care must be taken. I.e., don't want to fold the extinction correction into the K-correction, and then also explicitly fold in an extinction correction.
Petter plotted Kbr vs. B-V due to strectch, and Kbr vs. B-V due to extinction, and the two fell along almost the same line. The major difference due to stretch variations, it indicated, was due to the color differences. In short, you only have to worry about the stretch B-V and the extinction B-V simultaneously. All you need is the B-V of the SN light once it emerges from the outer reaches of its host galaxy; you don't need separate stretch and extinction colors. (This we can measure by knowning the extinction due to our own galaxy and measuring good I and R colors.) (If you don't screw up and fold the correction in multiple times.)
Peter notes that the K correciton where the filters don't match up well can vary with time having shapes that aren't purely linear. If you give a different stretch supernova, does it just shift the K correction up and down, or does it change the shape of K vs. t? Peter is checking this right now. If this is true, then he can apply a single K-correction. He'll know the stretch, so he'll know the color the thing _should_ be when it comes out of the galaxy. (Then it got complicated and I slightly lost the flow of it.)
Question: when you apply this K-correction, have you already applied the stretch/magnitude correction and the extinction correction? I think not, right now, because the K-correction is from "B affected by extinction" to "measured R affected by extinction," not from "B unaffected by extinction" to "measured R affected by extinction." However, this needs to be carefully pondered.
Peter says: generic K-correction, this tells you what the stretch is (from the fit?). You need this K-correction, since it's a function of time, From the stretch, you know what the color's supposed to be. What do you do when.... Allright, everybody think about this so that enough of us can eventually be happy with everything so that they can convince the rest of us that we all know what we're doing.
Colors to use: Peter says use B-V at Bmax or at Vmax. Bmax-Vmax is meaningless, Peter tells us. Unfortunately, right now we're reporting Bmax-Vmax.... although Don says that the Leib templates have Bmax and Vmax at the same day but Alex disagrees with this. Alex says this is due to a bug in the table, but Don just says that's what Leibengut told us.
Since there's a bug, there's no problem.
--A. Kim
This needs to be worried about, but Alex says that his rewrite takes care of this, but currently gives Bmax-Vmax. He hems and haws about getting B-V at Bmax. (Stretch changes Bmax-Vmax.) Program is keyed on doing Bmax and Vmax. Saul says this is fine, but then we need a lookup table on the templates. This cannot be forgotten.
Peter is speaking again. Alex gave him tables of SNe at various stretches and their color curves. Peter is taking SN spectra and reddening or bluening them to match Alex's. He'll run these through to see what happens with the time variation of the K correction for SNe of different stretch (or different B-V). He's reddening them rather than using the stretch of the SNe because of his findings that extinction and stretch affect the K-correction the same way.
Wrench in the works department?: Saul says that B-V at Bmax is not a great predictor of stretch, for the blueist supernovae. Peter says, however, that the prediction may have been good if there hadn't been extinction. Peter says, yes, it does flatten out, but it's still monotonic. Slope is higher in red range, but there is still a slope in the blue range.
Results coming.
Robert, host galaxy color stuff. He knows how to due it, he knows how to work aperture. He needs to talk to Rob about finding good reference images. He will talk to Greg about choosing good aperture sizes. Greg says that they will try using a metric aperture and an isophotal aperture to see if it makes any difference.
Gerson notes that in searchscan, sometimes the distance to the galaxy is wrong, because in about 20% of the times the searchscan program finds the distance to a different galaxy. Rob notes that the lightcurve software does some host galaxy finding, which isn't great but perhaps better than what searchscan does. This should be evaluated and also reported.
I haven't done anything.
--D. Groom
(Is this an alibi or an admission?)
Craig is going to talk about the work with the spectra and can we clean up our sample in the future. Original question: can we set cuts on the discovery parameters such as %inc in order to enrich the sample with SNe with good spectra.
Defined code of spectra. 0 is the best one, 4 is unusable, 5 is QSO. Code was based on telegrams and what was in the database and what we said in correspondence about how good the spectra were.
QSO's all cluster at 20-30% increase, 23-24 magnitude. (These are parameters from the scanning program.) Most of the rest of the bad ones are at less than 100% increase.
Next was get an Isobel ranking, which was more pessimistic than the Craig code. Still is trend to worse objects having lower %Inc, but no obvious place to set a hard cut. Can also plot in magnitude and %INC at the same time, most of bad ones are still dimmer than 23rd mag, less than 100 %Inc, but the good ones are quite blurred with the bad oens.
The next thing to try is the new perent increase defined by the lightcurve photometry software, with (theoretically) year-old references. Plotting searchscan %Inc vs. lightcurve %Inc (still to discovery image), the lightcurve %Incs are often higher. Plotting quality vs. New %Inc, all of the good ones are at greater than 90% %Inc. However, there are bad ones there too. Also tried using new SN mag as well because Rob is slow and there were only 31 zeropoints. Some of the new (i.e. lightcurve software) values are slightly brighter than the old (i.e. searchscan). No good spot to set a cut when you do a plot with this.
Craig goes back to original %Inc. Plots %Inc vs. "0" for bad spectrum and "1" for good spectrum. Craig fits a cut function to this which is a step function, where the assumption is that below the cut everything is bad and above the cut eveyrthing is good. He'll form a chi^2(x) and minimize that. So, chi^2 is the total number of "mistakes" you make. Result, for original %Inc, get that you should set the cut at about 22%. This corresponds to a place where you eliminate a bunch of bad spectra, but you haven't touched the good ones yet. Another way of doing this is ratio of good/bad above the cut. Between 10 and 20 percent there is a jump in this; after that it's only slowly changing.
Using the lightcurve software %Inc, the cut comes out at about 60%, and looks more convincing to this humble obsever. Suggests you would set the cut at somewhere between 50% and 70%. That is the spot where you've thrown out a bunch of bad ones but haven't gotten rid of very many good ones.
Other statistics: doesn't look like spectrum quality is a function of z. There are other questions about galaxy surface brightnesses, and who would %Inc change for the same galaxy/SN transported by z. From our data, there is not a real strong correlation between %Inc and z.
Craig will write this up so that we can all have it in front of us, and not forget about this.
Carl has no good news. (Mike Lesser had a head injury and is in the hospital.) Chip thinning has as a result stopped for the time being. Saul will talk to the British to find out what they want to do.
Alex got a hole-in-one from 183 yards yesterday. Peter estimates the probability is about 1,000,000 in 1. The sad thing was that he didn't see it, he turned his head.