So today I got started on another chore: adding flux upper limits into the light curves being produced by our supernova search. Usually the way we find things is to look for point sources in our processed images and see what we find. However, sometimes it can be equally useful to know what we didn’t find.
If all we have is a little dot on the image staring us in the eye and we’re asked what it is, all we can really do is shrug. Knowing what that little dot was doing over the course of the last week or two at least, and ideally the whole survey history, is much more powerful. If something blinks on and off every other day, it is probably not a supernova. Same if it was there two years ago at the same location. If it’s been there for the last month, it either isn’t a supernova or it’s an old, stale supernova unlikely to be of interest. However, if something has a smooth, quickly rising light curve and it was never found at those coordinates in your data before three days ago, it’s much more likely to be a young supernova and worth spending some spectroscopy time on.
The bookkeeping associated with this for five years’ worth of survey and hundreds of repeat visits with millions of detections can be considerable, so I talked with Fang about how best to go about it. I started work today on what I hope is the least troublesome option.
There is also some math to figuring out at what level you would have expected to see something at a given point in the image if there were really anything there, which I dealt with. It’s never enough to just say “it’s not there!” — you need some estimate of the maximum flux a source could have without being detected. That’s where the real information comes in, and how you can tell whether your non-detection is a useful constraint.