Hi Christine,
Nice to hear from you again. Yes, those were some crazy days back in 2010 when we announced GJ581g. I’m glad you decided to go ahead with your Sci-Fi novel. Thanks very much for sending me an advance copy. I will peruse it as soon as I can get some free bandwidth.
You needn’t be saddened about reports that others could not confirm my findings. Let me give you a bit of history and an update that you won’t find in the public media.
We worked on this result for many, many months, playing Devil’s advocate with our data and errors, to convince ourselves that we were on solid ground for claiming the detection of GJ 581g. We then submitted it to the Astrophysical Journal, the world’s premier journal of record for work in my field. It passed two independent peer reviews, both with flying colors. Indeed, our editor was/is also one of the world’s pre-eminent experts in statistical analysis, and literally “wrote the book” on statistical techniques in science. Even he agreed that our detection was rigorously significant using all the usual metrics for false alarms, etc. He even advised us that we were under-selling the confidence level of the detection. It was only after passing this rigorous peer-review process, and the paper was accepted, that we then organized the NSF press release announcement.
Several weeks after our announcement, the Swiss team from Geneva (the only other group then, and now, capable of providing confirmation of such a difficult detection) gave a talk at a conference in Trieste, Italy. At that conference, they verbally stated that they had an additional bunch of 60 new points of data, data that was also superior in quality to ours, and that they saw no evidence of GJ581g in their expanded data set. They did not provide that data in their talk, nor make it available to the public.
We were a bit disappointed in this, but could not respond in any way as they did not divulge any of their data. But we were skeptical, and so we simulated up a data set (actually we simulated 1000 data sets) using Monte Carlo techniques, that had the exact same cadence characteristics (they Swiss HARPS team doesn’t get to observe every day of the month, only in lunar bright time, etc) and signal/noise characteristics as their typical HARPS data. In these mock data sets, we inserted real signals at the periods of all the 4 known planets (b,c,d, and e) plus the ones we had claimed (f, and g). In doing so, we quickly convinced ourselves that, even if the Swiss had 60 more high quality data points, they would not have been able to either refute or confirm GJ581g. So we knew that their claim (that they would have been able to see GJ581g with their expanded, superior data set) was untrue. But since we did not have access to their data, we did not feel we could write this up and get it published in the peer-reviewed literature.
So we sat back, and waited, while we were roundly pilloried in the media about our GJ 518g claim being a false alarm. It was rather frustrating, and we were accused of being scoundrels, of grand-standing in order to get grants renewed, etc. etc. I was accused of making a “mistake” by assuming circular rather than eccentric orbits. All of this was complete and utter nonsense, but that was how it played out. And we just could not join the public discussion to defend our conclusions and data.
So we kept acquiring further precision velocity data on this system, and waited for the Swiss to eventually publish their data set. It took about a year of waiting for the Swiss to submit a preprint to the European journal Astronomy & Astrophysics, their equivalent of our Astrophysical Journal. That preprint by Forveille et al 2011 is attached fyi. Once we had their actual data set, I set about combining it with our data to do an extremely careful analysis of the joint data set and see what was what.
In doing this detailed analysis, we learned several things. First, the Swiss’s 4-planet Keplerian model was dynamically unstable, and was therefore manifestly unphysical. Secondly, we could not reproduce their values of Chi-squared, and RMS from using their 4-planet Keplerian model as fit to their data set. Digging in further, we realized that this was because they had omitted 5 of their data points from their analysis. These were specifically the 5 data points which had the worst fit to their model. They did this omission without disclosure. Indeed, they even stated in their paper that they had omitted no data points in their analysis.
Once we realized this, we re-did their analysis, including all of their data points, and omitting none. We also included the effects of planet-planet gravitational interaction, something the Swiss had not done, but that is critical for properly modeling this system. That kind of fit is called a dynamically “self-consistent” fit, and is generally not done as it requires a much more sophisticated level of numerical modeling. Anyway, our self-consistent fit, using their expanded data set, combined with our full data set, revealed that, after accounting for the 4 “known” planets b,c,d, and e, a statistically significant signal for planet g still emerged from the data set, with a period of 32-days, and consistent with a planet of minimum mass 2.2 times the mass of the Earth, and smack dab in the liquid water habitable zone. Furthermore, our 5-planet model was statistically superior to that of the Swiss, and was completely dynamically stable.
I wrote this all up and submitted it to the Astrophysical Journal to finally join the debate. Again, it passed several peer-reviews at ApJ, and was soon accepted for publication. Unfortunately, the Swiss’s Forveille et al paper had yet to be accepted by A&A, and to appear in print. So the ApJ editor put our accepted paper on-hold until the Forveille paper appeared in print. So we waited another 6 months.
Eventually, we gave up waiting for the Forveille et al paper to be accepted and to appear in print. And the ApJ editors just would not relent and let us publish until that happened. So we pulled the paper from ApJ and submitted it to Astronomische Nachrichten, a European journal in Potsdam. It passed peer-review and was published in only a few weeks. I attach a copy of that also fyi.
The Forveille et all paper was never accepted nor published. No doubt the A&A editors rejected it and they never re-submitted. They know they were wrong, and we called them out on it publicly. There is little they can say to change that. At the same time, the Swiss have had a habit of making their mistakes just disappear conveniently from the web. So, just in case they did this, I included all of their actual data in our A.N. paper, for all to see, for all time. Anyone can repeat my analysis, with all of the available data, and will come to the same conclusions I did.
So, even though the press has long since moved on and forgotten about GJ581g, the scientific peer-reviewed record shows that our result still stands, our analysis is the last and best word on the reality of GJ581g, and the available data and most sophisticated analyses all indicate that planet g exists, albeit with a slightly smaller period (32d instead of 36d) and slightly smaller mass (2.2 Mearth). What will it take to firmly confirm this planet? That’s a tough one. We already have many hundreds of data points. We’d need more than twice that data set to even begin to make further improvements (as I discussed at the end of our rebuttal paper). Also, this is a particularly complex system in that the 67d orbit could be either circular or eccentric. Slight amounts of eccentricity in the 67d orbit could be used to mask the signal from the 32d. So the reality or not of planet g is coupled to whether or not the 67d planet’s orbit has a small amount of eccentricity. This is a very tough thing to assess, even though we now know from KEPLER results of thousands of exoplanet orbits, that it is most likely that all of these GJ581 orbits are very nearly, if not perfectly circular.
So GJ581g still lives, at least according to the present set of available data, and the best and most sophisticated analyses of those data. But we have moved on to many other systems that are much simpler and easier to confirm. These are systems such as Kapteyn b, GJ 667Cc, GJ667Ce, and GJ667Cf. FYI, a list of these nearest potentially habitable Earth-sized planets can be found here:
