Forbes: In 2012, scientist Xavier Dumusque and his collaborators announced something spectacular: the closest star system to the Sun, the Alpha Centauri trinary system (consisting of the Sun-like Alpha Centauri A, its lower-mass binary companion Alpha Centauri B, and the tiny, very distant trinary member, Proxima Centauri), had a planet around one of its stars! by Ethan Siegel
'Alpha Centauri B, it appeared, had a planet orbiting very close to it, completing an orbit around the star every 3.24 days! (For comparison, Mercury takes 88 days to orbit the Sun.) The way it was measured was through what’s known as the radial velocity method, where a planet’s gravitational tug on a star causes it to appear to move towards us, then away from us, then towards us again in a periodic, well-defined fashion. This results in a phenomenon known as stellar wobble, and so by measuring the frequency and magnitude of the wobbling, we can determine the mass and orbital properties of the planet that must be there. The “wobble” meant that the star moved back-and-forth by an extra speed of just 0.0005 km/s every 3.24 days. And it was measured over a long enough baseline that other explanations — internal magnetic properties of the star, instrumental noise, or the tug of other companion stars — couldn’t be the cause. It seemed they had truly discovered a planet.
'But it wasn’t to be so! There isn’t a planet there, but the data was telling us a planet was there. The hard truth is this: we tricked ourselves because of how we measured this data. You see, in an ideal world, you’d monitor a star continuously, 24 hours a day, observing its signal constantly. In the real world, you only do it when you have access to the telescope (when it’s not being used for other purposes), at night, and when the sky is both clear and has good enough atmospheric conditions to see what you’re aiming at.
'By subtracting out the inherent variation in the star itself, the team accidentally amplified other periodic signals, one recurring one of which was mistaken for a planet. That signal turned out to be the rotation of the star itself, which has only now been accounted for properly. Interestingly enough, when all the analysis is done properly, there’s a hint of a signal for a different planet significantly farther out: with a period of about 20 days. The innermost planet of Alpha Centauri B turned out to be a false signal, and not actually there at all.'
Ghost in the time series: no planet for Alpha Cen B here