Exploration of the Time Domain

S. G. Djorgovski

VOEvent Workshop,CACR, Caltech, April 2005

Overview• Scientific motivation and opportunity

– A poorly explored portion of the observable parameter space

– A range of exciting astrophysical phenomena– Possibility of fundamental new discoveries– Technological advances and synoptic surveys

• Pilot project using DPOSS plate overlaps– A rich phenomenology of the time-variable sky

• Palomar-Quest: status and plans• Some challenges ahead

Exploration of the Time DomainThings that go bang! in the night… New and old.

Megaflares on normal stars Optical transients

Synoptic sky surveys will require a real-time mining of massive data streams and multi-PB archives

Exploration of the Time DomainThings that move…

Sedna

QuauarM. Brown et al.

NEAT

Tunguska

A Systematic Search for Transients and Highly Variable Objects Using DPOSS Plate Overlaps

~ 1.5 O overlaps between adjacent plates ~ 40% of the total survey area

Effective Area Coverage in a “Snapshot” Survey:

If texp < tburst and baseline ∆t >> tburst , then

Effective Area = Useful Area Npasses Nfilters

For DPOSS: ~ 15,000 deg2 0.4 2 3 ~ 0.9 Sky

B. Granett, A. Mahabal, S.G. Djorgovski,and the DPOSS Team

Baselines from days to ~ 10 yrs, typical ~ 2-4 yrsTypical limiting mags r ~ 20, using 3 bandpasses (JFN gri)

Plate Flaw Transient

Variable Source

Types ofDetections

Preliminary Spectroscopic Follow-Up (34 sources)

SpectroscopicSource Identifications:

35% QSOs (1/2 radio loud)18% CVs18% M dwarfs 6% Earlier type stars23% Unidentified (likely BL Lacs?)

Examples of DPOSS Transients

Examples of DPOSS Transients

flaw

asteroid

DPOSS Pilot Project Conclusions:• Faint, variable sky has a very rich phenomenology

– Spectroscopic follow-up will be a key bottleneck for any synoptic sky surveys

• Most high-amplitude variable sources down to ~ 20 mag are QSOs (Blazars, OVVs…), CVs, and flaring late-type dwarfs, with some early-type stars

• Asteroids may be a significant contaminant in a search for transients

• We find many more transients (~ 103/Sky) than expected from current models for orphan afterglows.– Most of them are probably QSOs, CVs, flaring stars, and

distant SNe; some may well be afterglows; and some may be new types of phenomena

• The Palomar-Quest survey should provide a major new venue for the exploration of the time domain

The Palomar-Quest Digital Sky Survey• The Palomar-Quest Consortium: Caltech - Yale - JPL• 50% point&shoot (NEAT, M.Brown), 50% drift scan (PQ survey)• PQ survey: a fully digital successor to the historic Palomar sky

surveys; a Caltech-JPL-Yale-NCSA-… collaboration• Data rate ~ 1 TB/month; ~100 TB total; many scientific uses• VO connections and standards built in from the start; a testbed for

many new IT/ VO applications

The PQ Survey: A new, digital, synoptic sky survey using the Quest-2 112-CCD camera

at the Palomar 48” Samuel Oschin telescope• A collaboration between Caltech, Yale/Indiana U., and

NCSA/UIUC; plus collaborations with other groups (INAOE, LBL, JPL…); started in summer of 2003

• About 50% of the telescope time, driftscan mode• Now next-day (or month… processing real-time• NVO connections and standards built in from start• Repeated observations, time baselines minutes to years• A science and technology precursor/testbed for

the LSST and other major synoptic sky surveys in the future

PQ Survey Sky Coverage• Range -25°< < +30°, excluding the Galactic plane• Ultimately cover ~ 14,000 - 15,000 deg2

• Rate ~ 500 deg2/night in 4 bands• As of Jan’05, covered ~ 13,000 deg2 in UBRI, of which

~ 11,000 deg2 at least twice, and ~ 4,700 deg2 at least 4 times; and ~ 14,100 deg2 in rizz, of which ~ 11,600 deg2

at least twice, and ~ 4,200 deg2 at least 4 times

The PQ Survey Science• Large QSO and gravitational lens survey

– Tests of the concordance cosmology– Dark mattter distribution– AGN physics and evolution

• High-redshift (z ~ 4 - 6.5) QSO survey– Probes of reionization and early structure formation

• Exploring the time domain– Supernovae, GRBs, AGN, var. stars, optical transients– Surprizes and new phenomena?– Time baselines ranging from minutes (inter-CCD) to days,

months, years (repeat scans) to decades (cross-match to DPOSS, SDSS, etc.)

• Galactic structure and stellar astrophysics• Etc. etc. - think what it can do for you

Palomar-Quest Data FlowImage

ArchiveCatalogs Archive

Master ArchivePipeline Data

Proc.Target

Selection

CIT

Yale

NCSA

JPL

LBL SNF

SDSC

P48

CITData

BrokerP200

Follow-up

PQ Survey: Recent and Upcoming Developments

• Completely redoing the DRP at Caltech– Much (much!) better data cleaning– Image remapping to a standard pixel grid, to enable

image coaddition and subtraction• Using the HyperAtlas technology for this• Using some TeraGrid resources

– Yet to come: better object detection, classification• Next: A real-time DRP for discovery of transents

and moving objects (collaborate w. JPL, M. Brown)

Improved Cleaning

of PQ Data

Improved Depth in PQ Image Coadds

PQ SDSS

A Search for Low-z Supernovae

• Calibration of the SN Ia Hubble diagram• New standard candles from SN II• Endpoints of massive star evolution

C. Baltay, R. Ellis, A. Gal-Yam, S.R. Kulkarni, and the LBL SNF

(Using the image subtraction technique)

Optical Transients and Asteroids

(Exploratory work; A. Mahabal, with P. Kollipara, a Caltech undergrad)

CITData

Broker

JPLNEAT

Archive

VariablesArchive

MasterArchive

ImageArchive

Yale

AlertDecisionEngine

SourceClassification

Engine

KnownVariablesChecker

AsteroidSeparatorEngine

CITFast

Pipeline

CIT Next-Day Pipeline

NCSA

Other?LBL SNF

Website

P48

BroadcastAlert

Off-site Archives

NVO / Multi-Off-site Archives

Palomar-Quest Real-Time TransientsDiscovery System Data Flow

ComparisonEngine:

Vars./Trans.Detector

Some Challenges Ahead• Automated, reliable, adaptive data cleaning

• High volume data generators lots of glitches• Cutting-edge systems poor stability

• High completeness / Low contamination• Must work with “solar system people” - moving

objects are the major contaminant for extra-solar-system variables and transients (and vice versa?)

• Automated, reliable event classification and alert decisions (need Machine Learning methods)– Sparse data from the event originator; folding in

heterogeneous external data; VO connections; etc.