Yale researchers have published a paper describing the use of hard data in the search for distant exoplanets.
Before the launch of the Kepler space telescope in 2009, the majority of exoplanets were discovered by the Radial Velocity method, which involves observing the light spectrum from stars to determine whether planets distort the spectrum when they pass between their stars and Earth during their orbits.
The paper proposed instead looking at only the data in an effort to find the patterns that indicated a planet. The approach — used by senior author and professor of geology & geophysics John Wettlaufer, mathematics graduate student Sahil Agarwal GRD ’19 and Fabio Del Sordo, a postdoctoral fellow in Geology & Geophysics — was borrowed from other problems, like the study of the behavior of sea ice.
The method is formally known as “multifractal temporally weighted detrended fluctuation analysis” (MF-TWDFA), and Wettlaufer credited mathematicians Benoit B. Mandelbrot and Katepalli Sreenivasan with pioneering the use of multifractals in science and math at Yale. A planet’s presence is indicated when the fractality is very regular, according to Del Sordo.
The team is able to recover the same outcomes with their method as in previous studies, but they are able to do it with worse data, Wettlaufer said of the data-driven method. According to Wettlaufer, this indicates that their method is more reliable than others.
In the paper, the team tested their method on an artificial sun and planet. They have since proceeded to rework data from previously discovered planets, and have determined that their method is more accurate. The approach improves upon previous statistical methods of analysis, since it makes no prior assumptions about the data.
According to Agarwal, unlike other methods, MF-TWDFA does not rely on removing extraneous data, or “noise,” which is instead considered an additional form of information. The key data points involved in the method are time scales relating to the passing of a planet through a sun’s orbit.
“Analyzing the raw data when available is the most robust way to study any system,” Agarwal said.
Del Sordo described a possible pitfall of the MF-TWDFA method as the chance of overworking the data so much that it suggests a planet is there when it really is not. Instead, the researchers focused on trying to shed light on previously unobserved aspects of the data, he said.
Further, there is less computation associated with the MF-TWDFA method than with prior approaches to searching for exoplanets, and it is able to quickly conduct wide surveys of data, according to Del Sordo.
Eventually, the team will take its method to projects like the Kepler mission, and will sort through the data those operations have gathered, Wettlaufer said.
The search for exoplanets has geared up in the last decade with the launch of the Kepler telescope, and in August 2016 astronomers announced the discovery of one such world around Proxima Centauri, the star nearest to the Solar System.
“I envision exoplanetary science as a field where one has the possibility to simultaneously achieve both scientific discoveries, and the human need of exploring our universe,” Del Sordo said. “Humans need to explore forbidden seas and skies.”