Gliese 581 is an unassuming star: it is relativity close at 20 light years away (the 87th closest cataloged star to earth), it is only a third the mass of the sun, and it is relativity quiet in terms of stellar activity (which is beneficial for life because flares scorch planets). It is the sixth planet from Gliese 581 denoted merely as g that harbors so much potential. It is not to hot, not too cold, it is just right. It is the Goldilocks planet. Vogt et al. 2010 recently reported on the discovery of this planet which is a 3.1 Earth mass (or larger) planet orbiting in the habitable zone of the M3V type star Gliese 581. The problem is that this planet may not exist.
The MediaI did not immediately discuss Gliese 581 here at The Astronomist because I wanted to read the paper before weighing in. However the authors were compelled to issue a press release about their findings before making their peer reviewed paper available. After I finally looked at the paper I was somewhat disappointed. The whole thing was a science journalism media circus. A selection of some of my favorite excerpts:
- “Found: An Earth like Planet, at Last” Time magazine
- “The chances of life on this planet are 100 percent,” Steven Vogt
- “Could contain more gold than we could ever imagine” PR Fire
- "Are the Gliesans going to Hell?" Huffington Post
- "An Alderaan Moment: Earth-Like planet disappears" Death+Taxes
All the planets around Gliese 581 were discovered using the radial velocity technique. In any gravitationally bound system the bodies orbit their common center of mass. It is a subtle effect in a star-planet system where the central star dominates the mass. The central star will move at a characteristic speed depending on the orbits of the planets around it. The movement of the star is measured through the Doppler shift of the light emitted by the star. Modern instruments are super sensitive to even the smallest movements of stars down to as little as 1 m/s. Observations of the radial velocity of the star over a period of time (usually several years) is analyzed using Fourier analysis. The Fourier analysis identifies periodic signals in the data corresponding to the orbital period of the planet or planets.
The researchers used two data sets spanning almost two decades. Most of the data came from the researcher's own instrument HIRES, and additional data came from a Swiss group with the HARPS instrument. The HIRES data spans a larger time range, but the HARPS data is more precise. This combined data set is how the researchers identified two new planets f and g.
The ProblemsA little after this new Goldilocks planet was announced the Swiss group announced that they could find no evidence of Gliese 581 g in their data. Does this mean it doesn't exist? Well this is tricky. A planetary researcher in my department, Rory Barnes, spoke to the New York times before the Swiss group had spoke up and said that the planet looked like the 'real deal'. After the announcement was made I spoke to Barnes again and he said that he would have to hold off further judgment until more information was available.
The onus of proof in science is upon those who make extraordinary claims. Vogt et al. were only able to find this planet by combing the available data sets; they actually state in their paper that they did not detect the planet in either of the data sets independently, only in combination. The damning part of the Swiss groups statement is that they say they have much more data available at this point that Vogt et al. did no have access to during their analysis. When the Swiss team forces planet g to fit their complete data they actually get a negative fit indicating that planet g really isn't there. The thing about this paper that I am least happy with is the quoted false alarm probability. The false alarm probability appears to be 1% based on the figures in the paper (see figure 3 specifically), but in the text it is quoted as ~10-5. I don't know what is going on.
Then there is their error analysis (warning this is about to get technical feel free to skip this paragraph). Vogt et al. used the peaks in the power spectrum to identify the planets in the system then subtracted off the highest power modes corresponding to the planets they had found. The power spectrum for each planet carried with it a false alarm probability, but once the planet had been subtracted out of the power spectrum its false alarm probability was washed away (you can see this happening in figure 3). They compound their errors after the 1st, 2nd, 3rd, 4th, and 5th planets which have varying false alarm rates. The proper way to do this is a joint fit model to all planets in the system using Bayesian analysis.
The strangest thing about all this is that when this paper was first submitted to The Astrophysical Journal the Swiss group was reviewing the paper and it was rejected. This Vogt paper meta chronicles its own history and discusses why it was retracted previously over concern of systematics. Unfortunatly the quality of the paper may not have improved. The Swiss group has actually leveled one specific concern, Vogt used perfectly circular orbits to find planet g, but the evidence shows the orbits are probably slightly elliptical. In fact in 2009 Vogt used elliptical orbits, but in this new paper circular orbits have been adopted. The image above illustrates this and makes a pictorial argument as to how circular vs elliptical orbits could introduce errors.
The discovery of an Earth-like planet seems imminent. I do not know if this is it. I will hold off further judgment until more information was available.
Steven S. Vogt, R. Paul Butler, Eugenio J. Rivera, Nader Haghighipour, Gregory W. Henry, & Michael H. Williamson (2010). The Lick-Carnegie Exoplanet Survey: A 3.1 M_Earth Planet in the
Habitable Zone of the Nearby M3V Star Gliese 581 ApJ accepted : arXiv: 1009.5733v1
Also thanks to Amit and Rory for discussion and figures.