CHIME Rings in Epic Sky Survey

A Fast Radio Burst (FRB) is a mysterious flash of light, detected primarily in radio wavelengths, and it can come from any direction in the sky. The first FRB was recognized in 2007 by accident while a team of radio astronomers analyzed archived data from 2001, collected by the 64-meter radio telescope at the Parkes Observatory in Australia. What they found was a broadband (consisting of many frequencies) burst of energy less than five milliseconds long, coming from a point in the sky a few degrees away from the Small Magellanic Cloud (a galaxy 200,000 light years from our home galaxy, the Milky Way).

Since then, less than 30 FRB signals have been found, making it difficult to identify their sources. In 2014, the first real-time FRB observation occurred at the Parkes radio telescope, and the additional data this event provided allowed Emily Petroff’s team to measure additional properties of the burst that prior events had been unable to capture. For example, the degree of polarization in this burst told researchers that large magnetic fields must be near the source of the emission, ruling out supernovae a possible cause.

Researchers including Victoria Kaspi estimate that between 100 and 10,000 FRBs arrive at Earth per day, so why have we only discovered 30 in the past 10 years? Since we only have a handful of large radio telescopes on the planet, and an FRB can only be observed if a telescope happens to be pointed in the right direction, seeing an event requires a lot of luck.

It would be infeasible to build thousands more radio telescopes, but that’s not our only option! A team of 50 Canadian scientists from the University of British Columbia, the University of Toronto, McGill University, and the National Research Council of Canada (NRC) had a better idea: a telescope that can see more of the sky at once.

Completed earlier this month, the Canadian Hydrogen Intensity Mapping Experiment (CHIME) is a radio telescope with an innovative shape. The physical telescope consists of a series of four 100-meter long “half-pipe” shaped reflectors that focus a large strip of incoming radiation onto a linear antenna hanging above each half-pipe. With a total area of about five hockey rinks (1.7 American football fields), CHIME can view a rectangular area of sky 120° in the N-S direction, and 1.3°-2.5° across, depending on the frequency being captured. This lets us observe almost half a percent of the entire sky! (Which doesn’t sound like much, but if you’re looking for something that happens 100 to 10,000 times a day, that means you can hope to detect between 3 and 300 events every week.)

But the magic is in processing what the telescope receives. Capturing this much of the sky results in a problem that most astronomers would love to have — too much data! Approximately 13 terabits are captured per second, which is about the same as the combined data used by every mobile device in the world. Since it would be impossible to store the raw data stream, CHIME analyzes and compresses what it sees in real time.

To build this capability on budget, CHIME’s team had to find a source of readily available processing hardware. Fortunately, the graphics processing unit (GPU) on video cards used for gaming suited their computational needs. Gaming hardware is perfect for space applications, as much of the computation is similar: both involve calculating the position of objects in moving reference frames. For example, determining where on a screen to display the hockey puck in a video game depends on where the player is skating and which direction the player’s head is looking. Similarly, determining where in space a radio signal originated involves calculating the position of an object relative to the Earth, which is rotating about its own axis, as well as orbiting the Sun. Since both tasks involve calculating matrix expressions, building a cluster of GPUs is a good fit for space processing.

To give CHIME the ability to detect FRBs, a specialized instrument was developed to search the entire viewable sky in real time. The instrument will scan 130 billion bits per second of data to detect arriving bursts as they happen, and direct the telescope’s main data channels to take a close look at the indicated portion of the sky.

As this new capability increases the number of bursts observed each year, researchers will be able to further refine models to learn more about these events, and provide clues as to whether they’re generated by magnetar flares, black holes collapsing, or some other exotic structure!