'Pi of the Sky' project studies rapidly varying astrophysical objects
A number of astrophysical processes manifest themselves on a short timescale. Understanding of these processes remains a poorly explored area of astronomy. In most cases, astronomical observations are performed with a time resolution of minutes, which limits the possibilities for investigating shorter processes. However, it is now clear that the most violent processes in the universe often manifest themselves by sudden bursts in various wavelengths. Among these, probably the most interesting are gamma ray bursts (GRBs)1 and active galactic nuclei (AGN).2
GRBs are a classic example of the major difficulties of using traditional astronomy for detecting very fast processes.1 GRBs are short pulses (0.01 – 100s) of gamma radiation coming from point sources on the sky. A big step towards understanding this process was the discovery of optical counterparts to GRBs in 1997 by the Beppo-SAX satellite. It is impossible to predict where and when the next GRB will occur, so the current strategy of optical telescopes is to wait for an alert from a satellite about a new GRB and move the telescope to the burst position determined by the satellite. The GRB Coordination Network (GCN)3 was developed for the purpose of fast distribution of GRB alerts.
The main problem with this strategy is the delay caused by the time it takes for a decision to be made and an alert propagated through the GCN. Another problem is the time needed to move the telescope to the desired position. This is caused by the fact that typical telescopes have a small field of view (FOV) and the chance of a burst happening in this field is almost zero.
Research on GRBs triggered the development of new types of astronomical devices: robotic telescopes. These are small, fast telescopes with relatively large FOVs, which can rapidly move to any position. However, only in a few cases was it possible to observe the area of a GRB within the first minute after the gamma detection. One of the fastest observations was performed 22s after the gamma detection by the robotic telescope ROTSE.4 More prompt observations of optical emission from GRBs could improve our understanding of the processes powering these events. Another challenge is observing the visual counterparts of short GRBs.
In order to overcome these limitations we propose a different approach. The idea is to observe a large fraction of the sky continuously. In case of a GRB alert, we don't need to move our telescope because its FOV is already there! For most of the time there are no external alerts, so we created our own trigger looking for optical flashes.5,6 Wide field observations also give us the possibility of correlating optical events with signals in other bands, like high-energy cosmic rays, neutrinos, or gravitational waves.
This approach will be realized by a system consisting of two sets of 16 CCD cameras. These two sets will be placed in different locations separated by a distance of ∼100km, which will allow us to use parallax to reject flashes caused by objects within 300,000km of Earth. Each set will consist of four parallactic mounts7 with four cameras placed on a single mount, as shown in Figure 1. Every camera at a specific site will have a corresponding camera in the other site, and both cameras will observe the same part of the sky simultaneously. The cameras are custom-designed fast-readout cameras, optimized for taking many short exposures. Each camera covers 20° × 20° FOV. The total FOV of the system is thus 2sr, enough to cover the FOV of the Swift Burst Alert Telescope (BAT),8 which is currently the most efficient satellite detecting GRBs. Each camera has a CCD of 2000 × 2000 pixels with each pixel covering an area of 15 × 15μm2. The cameras are equipped with CANON EF f=85mm, f/d=1.2 photo lenses. Readout and control is done through a gigabit Ethernet interface.9 The expected limiting magnitude for 10s exposures is 12 - 13m and for 20 exposures added together it is 13 – 14m.10 The apparatus is currently under construction and is expected to start collecting data in the second half of 2007.
In order to test the components of the full detector, a prototype consisting of two cameras has been built and installed in Las Campanas Observatory in Chile (see Figure 2). It began operation in June 2004. The two cameras are installed on a single parallactic mount. The prototype is fully automated and controllable via the Internet. A dedicated script language was developed to control the whole system. Every night, information about satellite pointing is retrieved from the Internet and the night's observation plan is generated in the form of a script11 that is executed during the night. Most of the time we follow the FOV of the Swift or the INTEGRAL12 satellites, and twice a night the whole sky is scanned.
With the prototype, data-analyzing algorithms could be tested and optimized. Currently we have two types of algorithms. The fast online algorithms are a flash recognition algorithm looking for optical transients on single 10s exposures, and a flash recognition algorithm acting on eight images added together (∼100s time resolution).6 The offline algorithms that act on data reduced and stored in the database are: a nova recognition algorithm, looking for new objects showing up on the night images (4min time resolution)10 or scan images (1day time resolution);13 a flare recognition algorithm, looking for an increase in the brightness of existing objects (4min time resolution);10 and a variable star analysis that identifies and classifies variable stars.14
The prototype showed that the system is able to automatically detect optical flashes of astrophysical origin. A nice example is an outburst of the flare star CN Leo, illustrated in Figure 3. Since the beginning, the online algorithm has detected about 150 short optical flashes of unknown origin, mostly visible on a single 10s exposure, but in eight cases visible on two consecutive frames. It's very unlikely that these are caused by satellites. A good example of a flash visible on more than two images is that in Figure 4. This flash has not been matched with any known astrophysical source.
Unfortunately, we didn't observe any optical signals related to known GRB events. In two cases, however, we observed the position of the GRB before, during, and after the gamma emission.15
The full system is expected to start collecting data in the second half of 2007. The main work to be done is currently at the hardware level on the computer cluster that will control the system and perform data analysis. The system must be fully automatic and remotely controllable. Our results are public and can be checked on our website http://grb.fuw.edu.pl.
Marcin Sokolowski is a PhD student at The Andrzej Soltan Institute for Nuclear Studies. He is involved in the “Pi of the Sky” project, working on software at almost all levels. He previously worked for a software company.
