Astropulse. A Search for Microsecond Transient Radio Signals Using Distributed Computing
- Author(s): Von Korff, Joshua Solomon
- Advisor(s): Werthimer, Dan
- Boggs, Steven E
- et al.
I performed a transient, microsecond timescale radio sky survey, called “Astropulse,” using the Arecibo telescope in Puerto Rico. Astropulse searches for brief (0.4 μs to 204.8 &mus;s), wideband (relative to its 2.5 MHz bandwidth) radio pulses centered at 1,420 MHz, a range that includes the hyperfine hydrogen line.
Astropulse is a commensal survey, obtaining its data by sharing telescope time with other surveys, such as PALFA. I scanned the sky visible to Arecibo, between declinations of -1.33 and 38.03 degrees, with varying dwell times depending on the requirements of our partner surveys. I analyzed 1,540 hours of data in each of 7 beams of the ALFA receiver, with 2 polarizations per beam, for a total of 21,600 hours of data. The data were 1-bit complex sampled at the Nyquist limit of 0.4 μs per sample.
Examination of timescales less than 12.8 μs would have been impossible if not for my use of coherent dedispersion, a technique that has frequently been used for targeted observations, but has never before been associated with a radio sky survey. I performed nonlinear coherent dedispersion, reversing the broadening effects on signals caused by their passage through the interstellar medium (ISM). Coherent dedispersion requires intensive computations, and needs far more processing power than the more usual incoherent dedispersion. This processing power was provided by BOINC, the Berkeley Open Infrastructure for Network Computing. BOINC is a distributed computing system, which allowed me to utilize hundreds of thousands of volunteers’ computers to perform the necessary calculations for coherent dedispersion.
Each volunteer’s computer requires about a week to process a single 8 MB “workunit,” corresponding to 13 s of data from a single beam and polarization. In all, Astropulse analyzed over 48 TB of data.
I did not aim to detect any particular astrophysical source, intending rather to perform a survey of the transient radio sky.
Astrophysical events that might produce brief radio pulses include giant pulses from pulsars, RRATs, or exploding primordial black holes.
In discussing the results of the Astropulse project, I have taken our sensitivity to primordial black holes with a certain size and spatial distribution to indicate our overall sensitivity relative to other surveys.
Radio frequency interference (RFI) and noise contaminated the data; these were mitigated by a number of techniques including multi-polarization correlation, DM repetition detection, and frequency profiling.
I also made use of a number of programs that specifically blank RFI from the FAA and aerostat radars near Arecibo.
Ultimately, Astropulse's sensitivity turned out to be similar to that of other very recent surveys, demonstrating that with enough computing power, a radio sky survey can make use of coherent dedispersion. We were unable to prove decisively that any of the signals came from astrophysical sources, but we did notice a surplus of pulses coming from inside the Galactic disk, as opposed to the halo.
In addition to Astropulse, I programmed a “distributed thinking” project called Stardust@home.
The two projects are not related by their science content, but they are closely connected by their use of distributed processing methods.
The Stardust spacecraft returned pristine interstellar dust samples, and Stardust@home recruited volunteers to locate these dust particles in microscopic-scale images of aerogel.
“Distributed thinking” means that volunteers examine our data with their own eyes, judging whether they see the dust particles.
In contrast, Astropulse volunteers utilize their computers’ processing power. Both methods create opportunities for public outreach, encouraging non-scientists to participate in scientific research. By signing up for Astropulse or Stardust@home, anyone can learn about astronomy and make a contribution to the field.