Slashdot Mirror

← Back to Stories (view on slashdot.org)

Alpha Centauri Turns Out Not To Have a Planet After All. At Least, Not Yet (forbes.com)

Posted by Soulskill on from the either-that-or-The-First-Order-has-some-explaining-to-do dept.

StartsWithABang writes: In 2012, astronomers announced that the nearest star system to us, the Alpha Centauri system, possessed at least one exoplanet around it. A periodic signal that recurred just every 3.24 days was consistent with an Earth-sized exoplanet orbiting and gravitationally tugging on the second largest member of the star system: Alpha Centauri B. That planet, named Alpha Centauri Bb, turns out not to actually be there. A reanalysis of the data shows that a combination of stellar properties and the times at which the observations were made conspired to produce this spurious signal: a signal that goes away if the data is handled correctly. Accounting for everything correctly reveals something else of interest, a periodic 20-day signal, which may turn out — with better observations — to be Alpha Centauri's first exoplanet after all.

2 of 91 comments (clear)

  1. Re:Known unknowns by bondsbw · · Score: 3, Interesting

    And how many exoplanets discovered to date are the result of handling data incorrectly?

    And how has that information influenced scientific research?

    --
    All my liberal friends think I'm a conservative, all my conservative friends think I'm a liberal.
  2. Re:Known unknowns by arcctgx · · Score: 4, Interesting

    It's noisy data. In the plots in TFA, you'll see that the residuals are expressed in meters per second. Meters! It's at the limit of detection even for our best spectrographs.

    It's very hard to work with noisy data. If you work on bad data the results get extremely dependent on methods of analysis. How do you prepare the data? Do you reject outlying measurements before you even get to analysis? If so, how? Why reject *this* point, but leave *that* one? Are you doing any filtering of the data (and how)? Any windowing? Smoothing? There's a lot of tricks you can use to make bad data appear acceptable. But in the end, it's garbage in, garbage out. That other signal can very well be an artifact. Or could be real, but not a planet. Or indeed a planet. We have no way of knowing without getting more observations of better quality (which is difficult and costs a lot of $$$).

    On the other hand, if the data is good, then any data analysis method will give you consistent results (provided that the method is used correctly).