Finding a habitable world around any one of the three Alpha Centauri stars would be huge. If the closest of all stellar systems offered a blue and green target with an atmosphere showing biosignatures, interest in finding a way to get there would be intense. Draw in the general public and there is a good chance that funding levels for exoplanet research as well as the myriad issues involving deep space technologies would increase. Alpha Centauri planets are a big deal.

The problem is, we have yet to confirm one. Proxima Centauri continues to be under scrutiny, but the best we can do at this point is rule out certain configurations. It appears unlikely, as per the work of Michael Endl (UT-Austin) and Martin Kürster (Max-Planck-Institut für Astronomie), that any planet of Neptune mass or above exists within 1 AU of the star. Moreover, no ‘super-Earths’ have been detected in orbits with a period of less than 100 days. This doesn’t rule out planets around Proxima, but if they are there, so far we don’t see them.


Image: The Alpha Centauri stellar system, consisting of the red dwarf Proxima Centauri and the two bright stars forming a close binary, Centauri A and B. Credit: NASA.

Centauri B, the K-class star in close proximity to G-class Centauri A, was much in the news a few years back with the announcement of Centauri Bb, a candidate world announced by Swiss planet hunters. This is radial velocity work based on data gathered by the HARPS (High Accuracy Radial Velocity Planetary Searcher) spectrometer on the 3.6-meter telescope at the European Southern Observatory in La Silla, Chile. The signal that Xavier Dumusque and team drew out of the data was 0.5 meters per second, a fine catch if confirmed.

What we thought we had in Centauri Bb was a mass just a little over the Earth’s and an orbit of a scant 3.24 days. As the blistering first planet detected around one of the Centauri stars, it would be a significant find even if it’s a long way from the temperate, life-sustaining world we’d like to find further out. The putative Centauri Bb supported the idea that there might be other planets there, and we’ve known since the work of Paul Wiegert and Matt Holman back in the 1990s that sustainable habitable zone orbits are possible around both the primary Alpha Centauri stars.

But Centauri Bb has remained controversial since Artie Hatzes (Thuringian State Observatory, Germany), using different data processing strategies, looked at the same data and found a signal he considered too noisy, indicating that what might be a planet might also be stellar activity on Centauri B itself. Debra Fischer’s team at Cerro Tololo Inter-American Observatory has also been studying Centauri Bb using the CHIRON spectrometer but has not been able to confirm it. And while a transit search using the Hubble Space Telescope did find a promising lightcurve (about which more in a moment), it couldn’t confirm Centauri Bb.


Image: Of the three stars of Alpha Centauri, the dimmest, Proxima Centauri, is actually the nearest star to the Earth. The two bright stars, Alpha Centauri A and B form a close binary system; they are separated by only 23 times the Earth – Sun distance. This is slightly greater than the distance between Uranus and the Sun. The Alpha Centauri system is not visible from much of the northern hemisphere. The image above shows this star system and other objects near it in the sky. Credit/copyright: Akira Fujii / David Malin Images.

Now we have a new paper from Vinesh Rajpaul (University of Oxford) and colleagues that makes Centauri Bb look more unlikely than ever. Rajpaul praises the thorough work of Xavier Dumusque and the team at the Geneva Observatory, but notes that their attempts to filter stellar activity out of their data evidently boosted other periodic signals that had nothing to do with a planet. The signal grows out of the time sampling, or ‘window function,’ of the data.

What is left behind is what the paper calls ‘the ‘ghost’ of a signal’ that was present all along. The paper argues that when a signal is sampled at discrete times (and the Dumusque team had to use the La Silla instrument only when it was not otherwise booked), periodicities can be imposed on the signal. Rajpaul was able to simulate a star with no planets, generating synthetic data out of which the exact same 3.24-day planetary signal emerged. The problem is particularly acute when working with planetary ‘signals’ as weak as these. From the paper:

D12’s data set [i.e., the data gathered by Dumusque and team] was particularly pathological because the window function happened to contain periodicities that coincided with the stellar rotation period of α Cen B, and its first harmonic; when these signals were filtered out, the significance of the 3.24 d signal was preferentially boosted.

All this is going to be quite useful if it helps us refine our techniques for identifying small planets. Rajpaul proposes that his team will carry out a new study of the spurious but coherent signals that can emerge from noisy datasets that should help us learn how to mitigate the problem:

We alluded to a number of other tests we believe worth carrying out when considering the reliability of planet detections from noisy, discretely-sampled signals. These include using the same model used to detect the planet instead to fit synthetic, planet-free data (with realistic covariance properties, and time sampling identical to the real data), and checking whether the ‘planet’ is still detected; comparing the strength of the planetary signal with similar Keplerian signals injected into the original observations; performing Bayesian model comparisons between planet and no-planet models; and checking how robust the planetary signal is to datapoints being removed from the observations.

Xavier Dumusque praises the Rajpaul team in this story in National Geographic, saying “This is really good work… We are not 100 percent sure, but probably the planet is not there.” We’re going to get a lot out of this investigation even though we lose Centauri Bb.

But back to that HST transit study run by Brice-Olivier Demory (University of Cambridge). I mentioned that it could detect no transit of Centauri Bb, which certainly fits with what we’ve just seen, but there was an interesting lightcurve suggesting a different possible planet, this one in an orbit that might range from 12 to 20 days. If this planet exists, radial velocity confirmation would be even more challenging than for Centauri Bb. Its signal, as Andrew LePage notes in The Discovery of Alpha Centauri Bb: Three Years Later, would be only half that of Centauri Bb.

LePage’s work at Drew ex Machina is definitive, and he has devoted a good deal of attention to Alpha Centauri. Here he explains why that second ‘planet’ is going to be so hard to spot:

Unfortunately with such a poorly constrained orbit, three weeks of nearly continuous photometric monitoring of α Centauri B will be required to confirm this hypothesis. HST is too busy to accommodate a dedicated search of this length and no other space telescope currently available is capable of making the needed observations. In addition, since the radial velocity signature for this planet would be expected to be maybe half that of α Centauri Bb, this method has little likelihood of providing independent confirmation of this sighting any time soon. Once again, we will have to wait for a few more years for new telescopes to become available such as NASA’s TESS (Transiting Exoplanet Survey Satellite) mission or ESA’s CHEOPS (Characterizing Exoplanets Satellite) which are both scheduled for launches in 2017 and may be capable of making the required observations of such a bright target.

Alpha Centauri is frustrating in many ways because you would expect the closest stellar system to have revealed more of its secrets by now. One of the problems, though, and a huge one, is that the angular separation (as viewed from Earth) of the primary Centauri stars has been decreasing as they move through their orbits. It won’t be until December of this year that they’ll reach minimum separation as seen from Earth. We’ll need to give Alpha Centauri a little time, in other words, before we can hope to get data on other possible worlds around Centauri B.


Image (click to enlarge): Apparent and true orbits of Alpha Centauri. The A component is held stationary and the relative orbital motion of the B component is shown. The apparent orbit (thin ellipse) is the shape of the orbit as seen by an observer on Earth. The true orbit is the shape of the orbit viewed perpendicular to the plane of the orbital motion. According to the radial velocity vs. time [10] the radial separation of A and B along the line of sight had reached a maximum in 2007 with B being behind A. The orbit is divided here into 80 points, each step refers to a timestep of approx. 0.99888 years or 364.84 days. Credit: Wikimedia Commons.

The Rajpaul paper is Rajpaul, Aigrain & Roberts, “Ghost in the time series: no planet for Alpha Cen B,” accepted for publication at Monthly Notices of the Royal Astronomical Society (preprint). The Hatzes paper is “Radial Velocity Detection of Earth-Mass Planets in the Presence of Activity Noise: The Case of α Centauri Bb”, The Astrophysical Journal, Vol. 770, No. 2, (2013) (preprint).