(NB: I just got back to Australia yesterday, so yes, these notes describe stuff that happened a week and a half ago. But bear with me, I’m still jet-lagged.)
Day 3 at Caltech was a light day, with only a morning session lasting a couple of hours. The focus was on what the PTF people should do after PTF was done, though similar priorities might well extend to any other contemporary survey of similar capabilities, like Skymapper or LaSilla/QUEST.
Short-timescale searches demand dedicated follow-up
Shri Kulkarni kicked off the festivities by detailing a grandiose vision of what he would do next if he were in charge. He pointed out that no major optical surveys had yet really explored transients on timescales less than a day. Short-lived events, such s gamma-ray bursts (GRBs), are usually found instead by all-sky monitoring instruments aboard space telescopes, sensitive in X-rays (Swift) or gamma rays (Fermi) and with rotten positional accuracy, locating the events within a degree or so at best. A short-timescale experiment would scan the sky as often as possible, plowing the search area repeatedly in a single night to find things that die away within hours, or even minutes.
This requires a lot of things. For one thing, a search/survey instrument is needed which has a much wider field of view. It may be possible to instrument the Samuel Oschin 48-inch Schmidt telescope, the instrument previously used for Palomar/QUEST and now used for PTF, with a detector able to view as much as 40 square degrees on the sky — about seven times as much sky as PTF now searches in a single exposure. (The corners of the focal plane would be subject to some “vignetting”, lying in the shadow of some part of the telescope tube, but this could presumably be accounted for at some level in order to make use of the extra solid angle.) Such a machine could image the same area as PTF each night perhaps once per hour. The resulting images would be searched for interesting events in real time, automatically cataloging any new arrivals.
It doesn’t stop there, though — once you find something interesting you need to follow it up. A short-timescale survey would probably be limited first by follow-up resources. Virtually every new object discovered needs to be classified and followed up promptly, to accurately measure relative rates of what may be quite rare events; this will require unprecedented amounts of telescope time just to see what’s turning up, let alone acquire high-quality science data. Time is of the essence, since by waiting even half an hour the event can be over before useful data can be taken. An array of new resources dedicated solely to looking at the objects coming from the new survey might therefore be highly desirable:
SED Machine — a 2-m telescope with a low-resolution spectrograph, taking spectra just detailed enough to classify the event.
Photometric Engine — a ~1-m telescope with a camera able to take images of a new transient in several filter bands simultaneously. This allows faster follow-up of more transient events over the course of the night. Instruments like GROND already use such technology.
A dedicated UV satellite to follow up discoveries in the ultraviolet — since information on new transients at far-ultraviolet wavelengths can be crucial to understanding their physics, and since light at these wavelengths can’t get through the Earth’s ozone layer. If only UV is needed, a few million dollars may be enough to build such a telescope. If X-ray information (such as that provided by Swift) is also desired, the cost could be much higher.
(Personally, after this list I’d kind of want a pony as well. A far-infrared, helium-cooled space pony, of course. But people like Prof. Kulkarni have ways of obtaining resources that people like me don’t have access to…)
Others in the session elaborated somewhat on this theme, but generally agreed with its premises. Blue-wavelength or ultraviolet observations are in general more important than red or infrared observations, necessary to investigate a range of interesting physical processes, and offering better contrast against a mostly red background of night sky or galaxy light. Real-time classification by computer, with attendant prompt follow-up, is vital. Imaging in more than one filter of the search engine itself could be very useful for quick classification of new transients; one easy way to image in two filters with one camera would be to split filters over the field of view. This might cause some calibration and search strategy challenges, but would be the cheapest solution in dollars immediately spent on the instrument.
Another common challenge with supernova follow-up photometry is the fact that supernovae happen in galaxies. As mentioned above, this tends to reduce the contrast and contaminates the SN observations. In order to remove the contamination, one must wait for a year or more until the supernova has faded away, to get a good picture of what the galaxy underneath the SN looks like. One then numerically subtracts the pixels of this “reference” or “template” image from the original SN observations to finally get a high-quality picture of the SN itself. This subtraction process is time-consuming and requires some care to do reliably; I’ll discuss it sometime later. Matching the properties of the search and follow-up instruments, however, allows one to subtract search images of the underlying galaxy (taken before the SN went off) from follow-up images of an interesting SN (taken only after one could have known there would be a SN there). This strategy then allows astronomers to publish the follow-up photometry as soon as it is taken, rather than having to wait a year or more until a reliable reference is found.
Interesting science topics
What interesting science questions could be served by a robotic telescope imaging the sky once an hour or so? As I mentioned above, there may well be interesting things there the existence of which nobody could have foreseen, giving the theoreticians bizarre new things to theorize about. There are existing classes of interesting events, however, which have either been predicted by theory or which have been, through sheer luck, observed in very rare cases:
CC SN shock breakout. In the very early stages of a CC SN explosion, the shock wave from the core rebound travels through the star, heating up the gas therein to extremely high temperatures. The radiation from the hot gas is at first hidden inside the star, but when the shock wave reaches the star’s surface, some of it leaks out as a burst of ultraviolet light and X-rays. The event should last no more than a day, and more probably a few hours, but we could learn a great deal about CC SN progenitors by observing the properties of such a burst. This is probably the most exciting class of short events the meeting participants want to study.
Signatures of SN Ia progenitors. Similarly, in theoretical descriptions of white dwarf explosions there is usually some material left over from the white dwarf’s binary companion — whether a “normal” star (single-degenerate) or another white dwarf (double-degenerate). When the expanding SN ejecta ram into this material, it should create a short burst of energetic radiation similar to shock breakout in a CC SN. There is therefore a great incentive to get very early light curves of SNe Ia shortly after explosion to learn more about their progenitor systems, and this will be made much easier by a short-timescale transient survey.
Very early SN Ia light curves. Even apart from the progenitor question, there is some interest in examining the light curves of SNe Ia shortly after they explode, in the hopes that it will reveal new information about the physics of the explosion. To catch any such new events requires a survey that looks back at the same part of the sky very regularly.
Tidal disruption events. On an unrelated topic, most galaxies in the universe, including our own, are now believed to have gigantic black holes at the center. Another kind of exotic, explosive event occurs when stars orbiting these black holes are sucked in and torn apart by the black hole’s gravity. The first light from such an occurrence would be seen, presumably, as — you guessed it — a short burst of X-rays and ultraviolet light. It would always happen right in the middle of the host galaxy, rather than on the outskirts as many SNe do.
Many of these events are also predicted to be quite faint. For example, a SN Ia one day after it explodes shines with only about 1% of the luminosity it will eventually reach at its brightest point. It was pointed out that finding very young events also requires the search telescope to go very deep — exposing for a long time.
Finally there was a brief reference to the next-generation supertelescope LSST, which topped the “Ground-Based” section of the US Decadal Survey of Astronomy. The consensus seemed to be that while this telescope would be a great search engine and would get millions of light curves of objects, they would be hopelessly limited in the number of spectra which could be obtained for related objects. The top future science will probably involve taking spectra of sources LSST discovers, which fortuitously more or less agrees with my own opinions on the matter.
All this being done, we went to tour the Huntington Gardens, which while worthy of comment isn’t really a science topic — though I do highly recommend it to anyone who’s in town.