The Nature Paper on the First Confirmed Earth-sized Exoplanets Kepler-20e and Kepler-20f

by Dr. Tahir Yaqoob on January 19, 2012

Towards the end of 2011 there was a frenzy of reporting in the online and offline media about the discovery by the Kepler mission of the first two Earth-sized exoplanets, Kepler-20e and Kepler-20f, orbiting a sunlike star. These exoplanets were known before from the Kepler data, but had just been promoted from “candidate” to “confirmed” status. Indeed, this is a weighty milestone but the reporting left a lot to be desired, even aside from the confusion between the terms Earth-sized and Earthlike. (There were also variations on the hyphenation or otherwise of Earth-sized and Earthlike: I am following Merriam-Webster unabridged on this.)

The news reports all contained pretty much the same phrasing for the principal claims, namely that the exoplanets are rocky (usually with no caveats mentioned for the use of words like “are” instead of “may be”), and that “theoretical calculations” show that the masses are..” For example, this LA Times article says, “Scientists have confirmed the existence of two Earth-sized, rocky planets orbiting a star called Kepler-20..”

So, let’s go beyond the noisy, oversensationalized news reports and blogs and see what the original paper claims. The Nature paper (Fressin et al. 2011, doi:10.1038/nature107080) is entitled Two Earth-sized Planets Orbiting Kepler-20 (Kepler-20 is the host star).

Most of the paper discusses the detailed methods of eliminating “false positives,” (i.e., to test the hypothesis that the discovered objects are actually planets orbiting the host star in question). The exoplanets are claimed to be detected from very tiny dips in the light from the star, as they transit across the face of the star. The modulation amounts to only about 0.01 percent of the star’s light output. After eliminating obvious things like movement of the aperture, a lengthy procedure is described in which simulations are performed in order to assess the probabilities that the detections are not due to two background stars eclipsing each other, or to a larger planet eclipsing another star. There is a certain amount of circularity in the process, in that the actual data from the Kepler mission is used to construct fake distributions of key parameters from which random drawings are made in the simulations. There are also some assumptions made along the way. In the end, the authors come up with two final probabilities and say, “We find that the hypothesis of an Earth-sized planet for Kepler-20e is 3,400 times more likely than that of a false positive, and 1,370 times for Kepler-20f.”

What are we to make of this? Are these sufficiently low probabilities for a mistaken identity? It’s easy to answer that question if you ask yourself whether you would be willing to bet your life savings based on these odds. Probably not. The authors say, “Both of these odds ratios are sufficiently large to validate these objects with very high confidence as Earth-sized planets.” I wonder if the authors would bet their life savings on these odds. In general, a result is not considered to be even worth considering as not due to chance if the probability is not higher than 99.73 percent (three sigma, in jargon), and certainly not a robust discovery if the probability is not higher than 99.99994 percent (five sigma, in jargon). For Kepler-20e the paper gives 99.97 percent and for Kepler-20f gives 99.927 percent. On the other hand, the paper mentions two exoplanets that have been independently confirmed by a different method after they were confirmed with the same methods as described in the paper. Such independent confirmation has yet to be done for Kepler-20e and Kepler-20f.

So, what about the rocky nature and masses of Kepler-20e and Kepler-20f? The caption of Table 1 in the paper says “Assuming Kepler-20e and Kepler-20f are rocky bodies composed of iron and silicates, and considering their radii, the planet masses are constrained to be..” Note the use of the word assuming. The caption goes on to give these mass ranges as 0.39-1.67 Earth masses and 0.66-3.04 Earth masses for Kepler-20e and Kepler-20f respectively. There is no basis for the assumption of a rocky composition. The only motivation is based on a comparison with similar-sized planets in our own solar system, combined with an explicit reliance on the current theory of planet formation. However, many assumptions based on our own solar system have turned out to be spectacularly wrong, and we know that the current theory of planet formation doesn’t work because it conflicts with observations. You can’t invoke the standard planet formation theory when it’s convenient and at the same time say that it doesn’t work when confronted with other data.

The Nature paper does, however, give mass upper limits that are independent of the rocky composition assumption, but they are not very constraining: they are 3.08 and 14.3 Earth masses for Kepler-20e and Kepler-20f respectively, so a gaseous composition is not ruled out. In other words, we do not know if Kepler-20e and Kepler-20f are rocky, it is purely a conjecture. News reports in the popular media effortlessly turned an assumption into a deduction. The “theoretical calculation” simply takes reasonable density values corresponding to a suite of assumed rocky composition variations, and simply multiplies by the volume to get a mass range. The masses, therefore, are also effectively assumptions, not deductions.

I will talk about exoplanet temperatures at another time, as well as other properties of Kepler-20e and Kepler-20f given in the Nature paper. The last paragraph of the paper goes into some considerable speculation about the possible compositions of these exoplanets, but certain assumptions are made which may not be true. I will also leave that to a future discussion, except for a couple of points. One is that the actual wording used to dismiss a gaseous composition is: “We infer that the two planets almost certainly do not have a hydrogen-dominated gas layer, because this would readily be lost to atmospheric escape owing to their small sizes and high equilibrium temperatures.” But we don’t know how much time the planets have already had for the atmosphere to escape! Calculations in the literature show that evaporation times can be of the order of hundreds of millions of years, even this close to the host star. Moreover, the planets would not have been small if they have been evaporating for some time. They would have been much larger to start with. So yes, they could be evaporating, but so what? The argument is completely circular. But that’s probably why the phrase “Almost certainly” was used: what does that mean in this context exactly? The phrase is given no quantification which really in itself undermines all the arguments given. The second point is that, at the end of the paper, the authors do actually say that, “ the absence of a measured mass the composition cannot be determined unambiguously.”

Next post: