Exploring the Transient Optical Sky
The LSST will open a new window on the variable sky. Recent surveys have shown the power of variability for studying gravitational lensing, searching for supernovae, determining the physical properties of gamma-ray burst sources, etc. The LSST, with its repeated, wide-area coverage to deep limiting magnitudes will enable the discovery and analysis of rare and exotic objects such as neutron star and black hole binaries; gamma-ray bursts and X-ray flashes, at least some of which apparently mark the deaths of massive stars; AGNs and blazars; and very possibly new classes of transients, such as binary mergers and stellar disruptions by black holes. It is likely that the LSST will detect numerous microlensing events in the local group and perhaps beyond. The LSST would provide alerts for concerted monitoring of these events, and open the possibility of discovering planets and obtaining spectra of lensed stars in distant galaxies as well as our own. LSST can also provide multi-wavelength monitoring over time of objects discovered by the Gamma-Ray Large Area Space Telescope (GLAST) and the Energetic X-ray Imaging Survey Telescope (EXIST). With its large aperture, the LSST is well suited to conducting a Deep Supernova Search in selected areas. LSST will also provide a powerful new capability for monitoring periodic variables, such as RR Lyrae stars, which can be used to map the Galactic halo and intergalactic space to distances exceeding 400 kpc.
Since LSST extends time-volume space a thousand times over current surveys, the most interesting science may well be the discovery of new classes of objects. Exploiting the capabilities of LSST for time domain science requires large area coverage to enhance the probability of detecting rare events; time coverage, since light curves are necessary to distinguish certain types of variables and in some cases infer their properties (e.g. determining the intrinsic luminosity of supernovae Type Ia depends on measurements of their rate of decline); accurate color information to assist with the classification of variable objects; good image quality to enable differencing of images, especially in crowded fields; and rapid data reduction and classification in order to flag interesting objects for spectroscopic and other follow up with separate facilities. Time scales ranging from 1 min (to constrain the properties of fast faint transients such as those recently discovered by the Deep Lens Survey) to 10 years (to study long-period variables and quasars) should be probed over a significant fraction of the sky. It should be possible to measure colors of fast transients, and to reach r > 24.5 magnitude in individual visits. Fast reporting of transients to the community is required in order to facilitate follow-up observations.
Targets of Opportunity
Targets of Opportunity (TOOs) for LSST are defined as observations for which fast response time is required and that therefore may (given a pre-planned observation schedule and prioritization matrix) interrupt an ongoing sequence of LSST observations. These may be externally triggered (e.g. by a prompt notification of a GRB) or derived in semi-real time by quick look processing of short exposure images (e.g. from programs, such as NEO, requiring short exposure sequences). We provide brief science motivation for both types, which we define as TOO-1 (external trigger) and TOO-2 (internal trigger) and then summarize desirable technical requirements to carry out this science.
Supernovae
LSST will be a powerful SN factory. Operating in a standard mode of repeated scans of the sky with images taken every few days and with exposures of 30 seconds, LSST will discover 250,000 Type Ia SN annually. Their mean redshift will be z ~0.45 with a maximum redshift of ~0.7. These data, when combined with priors from other experiments, can constrain the lowest eigenmode of w (i.e. the mean value) in the nearby universe to 1 percent, and given the dense sampling on the sky, can be used to search for any dependence of w on direction, which would be an indicator of new physics. Some SN will be located in the same direction as foreground galaxy clusters; a measurement of the magnification of the SN will make it possible to model the cluster mass distribution. Core-collapse SN will provide estimates of the star formation rate during the epoch when star formation was changing very rapidly. Longer exposures (10-20 minutes/band) of a small area of the sky could extend the discovery of SN to a mean redshift of 0.7 with some objects having z >1.4.
The added statistical leverage on the "pre-acceleration" era will narrow the confidence interval on both w and its derivative with redshift. Spectroscopic follow-up for so many SNe will be impossible. Exploitation of the data from the LSST will require light-curves which are well-sampled both in brightness and color as a function of time. This is essential to the search for systematic differences in supernova populations which may masquerade as cosmological effects as well as for determining photometric redshifts from the supernovae themselves; the development of techniques for determining photometric redshifts from supernova light-curves is currently being pursued by several community groups. Good image quality is required to separate SNe photometrically from their host galaxies. Observations in 5-6 photometric bands will be necessary to ensure that, for any given supernova, light-curves in four bands will be obtained (due the spread in redshift). Absolute photometric calibration to 1 percent is adequate, but the importance of K-corrections to supernova cosmology implies that the calibration of the zero points between filters remains an important issue, as is stability of the response functions, especially near the edges of bandpasses where the line emission from supernovae makes this more of a problem than for stellar spectra.
The Unimagined
"As we know, there are known knowns. There are things we know we know. We also know there are known unknowns. That is to say we know there are some things we do not know. But there are also unknown unknowns, the ones we don't know we don't know." Donald Rumsfeld
A challenge for LSST is recognizing important transients -- in real time -- in a scene full of normal variations. The data stream will simply be too large for efficient transient identification by human analysts. The broad continuum of properties for both extraneous artifacts and interesting transients make them difficult to deal with on a piecemeal basis with hard-wired code. Further, understanding of the time domain is too incomplete to predict confidently the properties of important changes. We are examining the potential of modern Machine Learning (ML) techniques for solving this problem. In particular, we are applying ML techniques for automated anomaly detection to the job of identifying transients without an a priori description. Many anomalies will be instrumentation errors. Automating their identification will allow prompt action to maintain LSST data quality. But some of the anomalies are likely to be things that go bump in the night that we have not yet thought of.
Supervised Machine Learning
It's easier to show a machine what
to find ...
-Machine Learning derives classification algorithms directly from examples of
data
... than to tell a machine how to
find it
-Requires domain expertise
-Involves software development
-Demands careful attention to statistical characterization
-Entails substantial amount of trial and error
The LSST will find normal source variations and instrumental artifacts in every image, the key to success will be real-time identification of important variations in a "forest" of normal variations. To efficiently identify the important "things that go bump in the night", the LSST will require multiple approaches to automated transient detection. For example, the integration of machine learning techniques and context information provided by virtual observatories with the real-time analysis pipeline will be essential for identifying fast transients while they are still present. An advantage of using ML techniques like anomaly detection algorithms is that they can find transients with properties we have not thought of. ML techniques also allow the system to be trained, both by mining its own data and by interacting with human analysts. This will give the system an ability to bootstrap its capabilities by ignoring artifacts "like that" or by finding more "like this" without generating new hard-wired code. ML techniques can therefore enable the efficient construction of queries for the LSST to act as an autonomous discovery engine that searches the night sky.
STIS image and light curve of GRB990123. It peaked at brighter than 9th magnitude in the optical, despite having a cosmological redshift. That makes it the most energetic event yet detected and poses a severe challenge to theorists who model these events. The technology now exists to monitor the sky in the optical for rare faint transients of all types—a million times fainter. Top figure from Bloom et al., Astrophysical Journal (Letters), v518, p1.
The detection of transient emission provides a window on diverse astrophysical objects, from variable stars to stellar explosions to the mergers of compact stellar remnants. Perhaps even more exciting is the potential for discovering new, unanticipated phenomena. A few short lived optical bursts without precursor objects have already been seen in SN surveys and by the Deep Lens Survey. LSST will obtain deeper and better sampled multi-color data on tens of thousands of such events.
Optical burst detected by difference imaging (right hand frame) in the Deep Lens Survey.