AI detects 72 fast radio bursts from a distant and unknown source

In a few milliseconds, fast radio gusts (FRB) are among the newest and most surprising astronomical phenomena. They come from extreme distances and appear without any known rhyme or reason, which makes them almost impossible to study in detail.

But fortunately, artificial intelligence helps researchers discover this mysterious event. Using machine learning technology, a team of astronomers was able to study 72 new radio frequency explosions from FRB 121102 – the only known FRB for continuously firing signals. Their conclusions, accepted for publication in the Astrophysical Journal, gives an overview of the periodicity of radio waves and specific frequencies, but unfortunately leaves us with more questions than answers.

Unknown FRB origin

FRBs are bursts of radio waves that last only a few milliseconds, coming from far away from the Milky Way. They appeared for the first time in 2007 when astronomy professor Duncan Lorimer and his student David Narkevic detected a 5 millisecond flash of radio waves surpassing the Small Magellanic Cloud, but they could not know where he was coming from neither he.

Researchers have since detected much more FRB, but their origins remain a mystery to this day. There is reason to believe that they come from energy supernovae or that they are ejected during mergers of neutron stars or black holes. Some even like to speculate that they are artificial signals from distant and intelligent civilizations.

The problem with obtaining concrete answers, however, is that the vast majority of FRBs are of a singular nature and emit only one explosion. But for some unknown reason, FRB 121102 continues to fire radio waves from a galaxy 3 billion light years from Earth.

As part of a program called Breakthrough Listen, which scours the skies for signs of extraterrestrial intelligence, a team of researchers from the SETI Research Center at the University of California at Berkeley decided to look into FRB 121102. A five-hour observation session with the Green Bank telescope in Virginia, they detected 21 FRB shots from the far galaxy, all over a period of one hour. The data indicated that radio waves were going from a confusing period of extremely intense activity to periods of nothing at all.

Other FRB surveys

Knowing that human observations are not always perfect, UC Berkeley Ph.D. Student Gerry Zhang has created a machine learning algorithm to apply to the dataset, hoping to collect the FRB that researchers could have missed. The algorithm revealed an additional 72 FRBs emitting from the galaxy – the largest number of FRBs detected in a single period of observation.

From the new data, Zhang and his colleagues were able to identify previously unknown details of FRB 121102. They found that the 93 FRBs in total had no specific pattern in their arrival time, and that the bandwidth of the airwaves radio amplifications. What this means for the FRB, no one really knows it. But with further observations, Zhang and his team hope to continue tracking his arrival times and bandwidth frequencies, to finally identify the source of the FRB.

The new discoveries still leave the researchers to break their heads, but with a little time, artificial intelligence will, hopefully, detect more features of the FRB and solve this astronomical mystery once and for all.