How we found a planet orbiting Proxima Centauri

As exciting as it is to find a full-blown extrasolar planet, it is sometimes equally important to know what kinds of planets might have been seen orbiting a set of stars, but weren’t. That was the goal of my research group at the University of Hertfordshire in early 2013. We were examining radial velocity data for a sample of nearby red dwarfs, most of which had no known planets, to estimate the probabilities of detecting planets of varying sizes and orbits around similar stars.

One of the red dwarfs in our dataset was the nearest star to the Sun, Proxima Centauri. Analyzing the combined data from two instruments, the Ultraviolet and Visual Echelle Spectrograph (UVES) at the Very Large Telescope and the High Accuracy Radial Velocity Planet Searcher (HARPS) at the La Silla Observatory, I started calculating the detection threshold of planets as a function of their orbital period. That’s when I spotted something interesting in the data: an unexpected maximum, a tiny bimodal bump with modes at periods of roughly 300 and 2000 days .

I flagged Proxima Centauri as noteworthy, but then I quickly moved on to other targets. I had to analyze data from GJ 163, which I was about to report as a host to up to four planets .

The plot thickened once I turned my attention back to Proxima Centauri a few months later. My plan had been to investigate whether a Keplerian periodicity could explain the variations in the noisy data from the two spectrographs. If the maximum I spotted earlier had less than 0.1% probability of being caused by random noise, then it would be a signal worth subjecting to further scrutiny.

However, the variations that I spotted were not independently present in measurements from the two different instruments. The UVES showed them, but HARPS did not. Such a discrepancy indicated that the variations did not come from Proxima Centauri; rather, something was probably wrong with the UVES data between 2000 and 2009. A member of my group at the time, Guillem Anglada-Escudé , diagnosed the problem: The UVES data suffered from an inaccurate barycentric correction. The team that compiled the measurements had not properly accounted for Earth’s movement around the Sun, an effect that has to be addressed with a precision of less than 1 mm/s.

After making the correction, I saw immediately how together the HARPS and UVES datasets showed unmistakable evidence for a statistically significant 11.2-day periodicity. Photometric brightness measurements revealed beyond any doubt that Proxima Centauri rotated with a period of 83 days, which ruled out the star’s own rotation as a source of the radial velocity signal. In the summer of 2013 I submitted a paper along with Anglada-Escudé and colleagues suggesting that there is a planet orbiting Proxima Centauri. But we could not convince the referee that the result was sufficiently robust.

One problem was that the signal’s statistical significance wasn’t sufficient to raise the result beyond a reasonable doubt. We also didn’t know enough about Proxima Centauri’s activeness. Stellar magnetic cycles and starspots can cause periodic Doppler shifts, which in turn create periodically changing radial velocities that mimic Keplerian signals of planets. This is especially true of starspots, which periodically block redshifted and blueshifted light from the two sides of the rotating star’s surface. However, such periodicities are not constant in time due to differential rotation—the fact that spots on different latitudes rotate around the star with slightly different periods.

The Pale Red Dot campaign , led by Anglada-Escudé, was initiated to take new measurements of Proxima Centauri and resolve the source of the 11.2-day periodicity. Radial velocity data taken in tandem with brightness measurements would distinguish between magnetic activity and the gravitational pull of a planet. The detection of a signal in 2016 would also demonstrate that the periodicity is stable and does not change in the lifetime of an observer. All we needed were observations covering a series of orbits. We obtained observing time on the oversubscribed HARPS spectrograph this past spring.

As is so often the case in science, targets of observations do not behave as expected. Proxima Centauri showed considerable evidence of flares and magnetic eruptions on the surface, compromising our observations. Fortunately, we could identify flaring events in the simultaneously taken brightness data, which enabled us to simply abandon the handful of radial velocities corrupted by flares. After only 30 nights of observations—half our allotted time—we were convinced that the signal’s significance was increasing every night and thus was time-invariant, as a planet’s signature should be (see accompanying graph). The signal was not caused by Proxima Centauri’s activity, and it was certainly not a statistical fluke. We finally put the champagne on ice.

Now we know that there is strong evidence for the planet Proxima b . But what do the observations really tell us? Because the orbital inclination of the planet is unknown, we know only that its minimum mass—the product of mass and the sine of the inclination angle—is roughly 30% more than Earth’s mass. It’s possible that Proxima b is considerably more massive. Although planetary statistics from NASA’s Kepler spacecraft show that short-period massive planets are rare around red dwarfs, all we really know is that Proxima b is unlikely to be more massive than Neptune.

We also know that the orbital period implies an orbital distance, or semi-major axis, of only 0.05 AU. That means Proxima b experiences strong tidal forces, which circularize its orbit and lock its rotation to its orbital period. Even if the planet resembles Earth in mass, it would be a wholly different world, with one side in constant daylight and the other in eternal darkness.

Yet even if not much is known about Proxima b, it is remarkable how much information about a star 4.2 light-years away can be squeezed out of a tiny probability maximum. This particular one is indicative of variations in the velocity of about 1 m/s, a human’s average walking speed. Proxima Centauri may host additional planets, too: The initial bump I spotted in 2013 didn’t totally disappear after the barycentric correction, so we suspect that there is another planet with an orbital period of 200–400 days.

Now it’s time to imagine what information could be obtained by seeing light that comes from Proxima b’s surface. That is an adventure within our reach during the next decade of more powerful telescopes.

Mikko Tuomi is an astronomer at the University of Hertfordshire in the UK. His initial detection of Proxima b gave rise to the Pale Red Dot campaign and the Nature paper announcing the planet’s presence in August 2016.