vestrand 2nd Zwicky Workshop

Thinking Telescopes,

RAPTOR, and GRB Follow-up

by

Tom Vestrand

Los Alamos National Laboratory

vestrand 2nd Zwicky Workshop

Machine LearningGENIE,

ML Classifiers,Anomaly Detection

Context KnowledgeRecord of

Sky variability (Virtual Observatories),

Massive Distributed Disk Array

Robotic Hardware

Wide-Field Sky Monitoring

Rapid Response Telescopes,

Real Time Pipeline

Thinking TelescopesAn Engine for Discovery

in the Time Domain

Goal is to Integrate Three Components

vestrand 2nd Zwicky Workshop

Thinking Telescope Objectives

• Find fast transients, distinguish foreground from celestial sources

• Monitoring of persistent sources for important changes in real time

• Machine learning merged with real time context information

• Anomaly detection, automated classification• “find more like this”• Learn to optimize telescope network

response, respond in real time• System Adaptability; Querying the Sky

vestrand 2nd Zwicky Workshop

vestrand 2nd Zwicky Workshop

RAPTOR design approach

• Must be a full system capable of both finding and following-up transients

• Modular, scalable, techniques• COTS (commercial off-the-shelf)

components• Distributed aperture approach• Must be robust, not requiring optimal

imaging and tuning

vestrand 2nd Zwicky Workshop

Raptor: Sky Monitoring with Both Eyes Open

• Wide-field imaging system monitors ~1300 square-deg with resolution ~35 arcsec and limiting magnitude of R~13th in 60 seconds. ( like the rod cells of the retina )

• Each array has a “fovea” telescope with limiting magnitude of R~16.5 (60 sec), resolution of ~7 arcsec and Gunn g (or r) filter. Provides color, better resolution, and faster cadence light curves (cone cells of fovea)

• Rapidly slewing mount places the “fovea” anywhere in the field in <3 seconds. (rapid eye movement).

• Two identical arrays are separated by ~38 km. (stereoscopic vision)

vestrand 2nd Zwicky Workshop

Memory and Context

http://skydot.lanl.gov

vestrand 2nd Zwicky Workshop

Machine Learning

• Automated identification of artifacts and transients in direct and difference images.

• Automated classification of celestial objects based on temporal and spectral properties.

• Real time recognition of important deviations from normal behavior for persistent sources.

vestrand 2nd Zwicky Workshop

vestrand 2nd Zwicky Workshop

vestrand 2nd Zwicky Workshop

vestrand 2nd Zwicky Workshop

vestrand 2nd Zwicky Workshop

Taxonomy of GRB Optical Emission in three classes

• Prompt Optical Emission varying simultaneously with prompt gamma-rays.

• Early Afterglow Emission that may start during prompt gamma-rays, but persists after gamma-rays fade.

• Late Afterglow Emission that can last for hours to days.

vestrand 2nd Zwicky Workshop

In the Standard Theoretical Framework it makes sense to attribute the components to

• Prompt optical emission is generated by internal shocks in ejecta---driven by engine.

• Early afterglow is a reverse shock driven into ejecta by interaction with surroundings.

• Late afterglow is generated by forward external shocks driven into surrounding medium.

vestrand 2nd Zwicky Workshop

vestrand 2nd Zwicky Workshop

vestrand 2nd Zwicky Workshop

vestrand 2nd Zwicky Workshop

Conclusions: What one needs to search for fast optical transients

• Ability to filter false positives robustly

• Real time follow-up

• Rank follow-up priorities

• Configure the response to optimize scientific yield, networking of telescopes

• All this must be autonomously without a human in the loop.

vestrand 2nd Zwicky Workshop