Ground-based Exo-planet discoveries Options

[Guest_PhilCo126_*]

Well-known exoplanet researcher Dr Michel Mayor ( discoverer of Peg 51b with Dr Didier Queloz in 1995 ) today announced the discovery of the lightest exoplanet found so far. The planet, “e”, in the famous system Gliese 581, is only about twice the mass of our Earth. The team also refined the orbit of the planet Gliese 581 d, first discovered in 2007, placing it well within the habitable zone, where liquid water oceans could exist:
http://www.eso.org/public/outreach/press-r...9/pr-15-09.html

[Julius]

How can they be so sure that the super earths detected represent one planetary body ie.could it not be that 8 earth masses could represent 2 terrestrial planets and other planetary dust the likes of asteroids in orbit round the parent star?!

[Greg Hullender]

There's a branch of mathematics called "Fourier Analysis" which, among other things, let's you take a complex signal and break it into a collection of sine waves. It's pretty cool, if you haven't seen it before.

So if there's just one planet, then the velocity plot ought to be a pretty clean simple sine wave over time. If there are multiple planets (let's say three) around the same star, then it'll be a mess, but a fourier analysis ought to result in just three sine curves and very little else.

--Greg

[tfisher]

So if there's just one planet, then the velocity plot ought to be a pretty clean simple sine wave over time. If there are multiple planets (let's say three) around the same star, then it'll be a mess, but a fourier analysis ought to result in just three sine curves and very little else.

I'm afraid this skips over a few important complexities. One is that with eccentric orbits (which are common!), the radial velocity signal of a planet is not a single sine wave, but may have significantly different shape. Another problem is that in reality we never have continuous or even a complete discrete series of radial velocity observations of a given target. Instead of seeing a continuous curve, we just have a non-uniformly spaced discrete sampling from the curve. So Fourier analysis can't really be directly applied to work out the component curves. A third problem is that there is a lot of noise in the observed radial velocities, coming from stellar activity of the host star and limitations of the observation instruments. In many cases (smaller, more distant planets or younger, more active stars) this noise is of a similar or even much greater strength to the signals we are looking for.

You can get a really good understanding of all of this playing with the amazing "systemic console" free tool for fitting candidate planetary systems to radial velocity datasets. You can get this at oklo.org. I highly recommend it!

[Greg Hullender]

I'm afraid this skips over a few important complexities.

Fair enough. I still maintain that a Fourier Transform would work fine for a system whose planets had circular orbits and ought to be a great way to eliminate the high-frequency noise in any case, but, yeah, I agree it won't be very good for elliptical orbits.

You can get a really good understanding of all of this playing with the amazing "systemic console" free tool for fitting candidate planetary systems to radial velocity datasets. You can get this at oklo.org. I highly recommend it!

This is pretty cool. Without actually downloading the program, is there a link to a document that describes the actual algorithms they're using? From what I could glean from their blog, it could be anything from Expectation Maximization to Markov Chain Monte Carlo.

Thanks!

--Greg