Oxford University Press's
Academic Insights for the Thinking World

The world’s most (in)famous exoplanet vanishes

In 2012, a team of astrophysicists led by Xavier Dumusque caused a sensation when they announced the discovery of Alpha Centauri Bb: an Earth-sized planet in the Alpha Centauri star system, the star system closest to the Sun. If verified, Alpha Centauri Bb would be the closest known exoplanet to our own Solar System, and possibly also the lowest mass planet ever discovered around a star similar to the Sun.

Media coverage of this “landmark” discovery was widespread and ebullient. Popular astronomy writer Phil Plait exclaimed: “Huge news…Holy crap! A planet for Alpha Cen[tauri]. Wow.” I certainly shared Plait’s enthusiasm. Another author referred to the planet, 4.3 light years away, as being “almost close enough to touch.” Many even spoke of the possibility of sending an unmanned probe to the newly discovered exoplanet.

Fast forward to 2015, though, and there’s a catch: Alpha Cen Bb no longer seems to exist.

Where’d it go? Our recent analysis of the original data suggests the planet was never there in the first place, and was instead an unfortunate by-product of the way the observations were made.

To understand how such problems can arise, it’s worth considering just how difficult it is to detect a planetary signal similar to the one initially claimed by Dumusque’s team.

Alpha Cen B, the “Sun” around which the claimed planet orbits, is 4.37 light years from Earth: by astronomical standards, that’s really close — yet it’s still more than 40 trillion kilometres away. With current technology, an unmanned space probe would take tens or even hundreds of thousands of years to make the journey. Driving there, if that were possible, would take longer than the known Universe has been around!

To infer the presence of a planet orbiting Alpha Cen B, Dumusque’s team used the radial velocity method to look for characteristic, periodic changes in the star’s motion (“wobbles”) caused by the gravitational tug of an orbiting planet. The size of the wobbles they ascribed to Alpha Cen Bb? About 50 centimetres per second, repeating every 3.24 days. It’s a remarkable testament to modern instrumentation and analytical techniques that we do have the capability of detecting such tiny wobbles from stars trillions of kilometres away.

As it turns out, though, an orbiting planet is hardly the only thing that can cause such apparent wobbles; other sources of variability include magnetic activity internal to the star itself, the gravitational pull of companion stars, instrumental noise, and more. The effects of these ‘nuisance’ signals can often be hundreds of times greater than the effects of hypothetical exoplanets. So Dumusque’s team had to work really hard to try to model and remove all nuisance signals, to isolate the tiny signal eventually ascribed to a planet. Indeed, this prompted much initial scepticism about their claimed planet.

Where’d it go? Our recent analysis of the original data suggests the planet was never there in the first place, and was instead an unfortunate by-product of the way the observations were made

But the modelling of the nuisance signals wasn’t the problem: our recent re-analysis of the data led us to much the same conclusions about the properties of the nuisance signals, even though we used very different techniques to Dumusque’s team. Instead, the source of the problem appears to be the times the observations were made.

It’s a rather contrived analogy, but imagine you’re trying to listen to a distant orchestra playing a tune, and for some reason, you’re only able to listen to the occasional note. If you’re lucky, you might hear a few very distinctive notes, and correctly identify the tune being played. If you’re unlucky, the same few notes might happen to match a few different tunes, and you could be led to the wrong conclusion. You might even mistake the noise of passing traffic for a real tune! On the other hand, if you could’ve listened to far more notes, and perhaps without having to contend with lots of noise, you’d likely have had no trouble identifying the tune. See Figure 1.

Figure 1: Illustration of how the sporadic observation of a signal can lead it to being mistaken for a signal with a different period. Here, the blue signal is aperiodic, while the red signal contains patterns that repeat every 3.24 days. Image by Vinesh Rajpaul
Figure 1: Illustration of how the sporadic observation of a signal can lead it to being mistaken for a signal with a different period. Here, the blue signal is aperiodic, while the red signal contains patterns that repeat every 3.24 days. Image by Vinesh Rajpaul

It was much the same with the Alpha Cen B data: Dumusque’s team could only observe Alpha Cen B on clear nights, using a ground-based telescope at the La Silla Observatory in Chile, and only when this telescope wasn’t booked for other observations.

To demonstrate that the discrete observing times were the source of the bogus signal, we used computer simulations to create a fake star, with properties very similar to Alpha Cen B — including binary companion star, Alpha Cen A — but which, by careful design, had no planets around it. Whenever we ‘observed’ such a fake star at the same time as the real observations, we reliably identified a 50 cm/s signal with 3.24 d period — identical to the planetary signal claimed by Dumusque’s team. Thus we concluded that this signal was a mere “ghost” arising from the nature of the observations themselves, rather than from any real planet.

More observations are probably necessary to rule out Alpha Cen Bb’s existence with 100% certainty, but our analysis does make it seem very, very unlikely that the planet is real.

Our work has relevance to future efforts to detect exoplanets (can we choose our observing times more carefully to avoid such bogus signals from arising? if not, which tests can we run to help identify potential problems?), and also to archival data (are any other supposed exoplanets really just artefacts?)

Yet it’s not all doom and gloom. Based on current confirmed detections of exoplanets, we estimate conservatively that there is at least one exoplanet for every star in our galaxy — that’s a minimum of hundreds of billion exoplanets in our galaxy alone, and the real number might be much higher! Don’t book your interstellar flight just yet, but it’s probably safe to say that it’d be weird if there were not any planets in the Alpha Centauri system.

PS: despite some conspiracy theories to the contrary, I was not paid by aliens from Alpha Cen Bb to cover up their planet’s existence.

Feature image credit: “Artist’s impression of the planet around Alpha Centauri B.” By ESO/L. Calçada/N. Risinger  CC BY 4.0 via European Southern Observatory.

Recent Comments

  1. Oscar den Uijl

    Isn’t this just a case of sampling below the Nyquist rate leading to aliasing?

Comments are closed.