Not long ago, while making a presentation about possible destinations for an interstellar probe, I called Gl 581d the most likely candidate for habitability yet discovered among nearby stars. I knew the planet was problematic, perhaps too far on the outer edge of the habitable zone to be a realistic candidate, although this seems to depend on a variety of factors including atmospheric modeling. But what I had really been pondering in deciding whether or not to include Gl 581d in the talk was whether its purported sister world, Gl 581g, should be brought into play.
Steven Vogt (UC-Santa Cruz) and colleagues were getting ready to distribute their new paper making a further case for a super-Earth in the habitable zone, one that seemed to be ideally placed for liquid water to exist on the surface. Bring that into the discussion?
I decided against it, because the controversy over this world continues and Centauri Dreams seems a better venue than a short public talk to get into the details. Let’s begin here, then, with Michel Mayor and the Geneva team, who had already identified four planets in the system, including Gl 581c, itself a target of speculation about whether or not it might be in the habitable zone. But Gl 581c looks to be too hot to support life, leading to the renewed interest in Gl 581d. This work was accomplished using data from the HARPS spectrograph on ESO’s 3.6m La Silla instrument.
What Vogt and team did in 2010 was to combine the earlier HARPS data with 122 additional measurements made using the HIRES spectrometer at the Keck Observatory on Mauna Kea. It was from this combined dataset that Vogt drew evidence of two new planets: Gl 581f (with an orbital period of 433 days) and Gl 581g, with a period of 36.5 days. It wasn’t long after this that Francesco Pepe (Observatoire de Genève) added another 60 HARPS measurements to the earlier ones and announced his team could not confirm the presence of either of the two worlds Vogt had found. In his new paper, Vogt’s team questions whether Gl 581f or Gl 581g would have been detectable using the 179-point HARPS data set on its own.
I’m moving through the details here quickly — the paper is available on the arXiv site and I encourage you to look at it. But new work by Thierry Forveille (Institut de Planetologie et d’Astrophysique de Grenoble) added another full observing season of HARPS data and still found no trace of Gl 581f or Gl 581g. It’s this set of expanded HARPS radial velocity data that Vogt’s new paper goes to work on, and it reaches a significantly different conclusion (F11 in the excerpt below refers to Forveille’s paper):
… we have shown in the present work that the F11 Keplerian solution is dramatically unstable over a wide range of starting conditions, and is thus untenable. F11’s conclusion of there being only four planets in the system was based on this unphysical model and can thus be discounted. Furthermore, the data points that were apparently omitted from the F11 analysis were dropped solely based on deviation from their 4-planet model, thus unfairly and specifically suppressing evidence for any additional planets in the system.
Things are, as you can see, heating up. What Vogt is talking about is that his simulations of the Forveille Keplerian models — with the Gl 581 planets in eccentric rather than circular orbits — showed that these orbits were unstable. This is important because Forveille used the Keplerian model in assessing the likelihood of the existence of Gl 581f and Gl 581g. In fact, among 4000 eccentric orbit simulations, not one survived beyond 200,000 years, with only 24 surviving for at least 20,000 years. All 4000 simulations ended with a collision between the two inner planets. By contrast, all 4000 simulations based on circular orbits turn out to be stable for at least 100,000 years.
Image: Artist’s conception of five potentially habitable exoplanets, with Earth and Mars to scale. Credit: The Habitable Exoplanets Catalog, PHL @ UPR Arecibo.
Using stable, circular orbits for its modeling, Vogt’s team sees a fifth planet (V10 below refers to Vogt’s 2010 paper):
Contrary to F11’s conclusions, we find that the full 240-point HARPS data set, when properly modeled with self-consistent stable orbits, by and of itself actually offers confirmative support for a fifth periodic signal in this system near 32-33 days, and is consistent with the possibility of having been detected as GJ 581g at its 36-day yearly alias period by V10. The residuals periodograms both of our interacting and non-interacting fits and of the F11 four-planet circular fit reveal distinct peaks near 32 days and 190 days. Both of these residuals peaks are largely simultaneously accounted for by adding a fifth planet at 32.1 days to the system.
According to Vogt, we wind up with a planet with minimum mass of 2.2 times that of Earth orbiting at 0.13 AU, “solidly in the star’s classical liquid water Habitable Zone.” That, at least, is what the data analysis produces if we assume circular orbits of the four known planets and work out the reasons for the further perturbations that Vogt’s team sees as evidence for a fifth planet. Vogt believes a 5-planet model with all circular orbits trumps a 4-planet model with eccentric planetary orbits, but adds that it may take time and further data to give a definitive answer.
What to make of all this? Gl 581 is offering us an instructive example of how tricky exoplanet analysis can be. Here we’re looking at radial velocity data showing Doppler shifts in the light of the star that result from the gravitational pull of multiple planets orbiting it. You can see how complicated a problem this is, and how multiple solutions can suggest themselves. I find myself pulling for Vogt’s team because I want a planet in the habitable zone to exist somewhere near our Sun (Gl 581 is about 20 light years from us), and the new paper is compelling. But we haven’t heard the last of this exoplanet flap, and how it will end is anyone’s guess.
The paper is Vogt et al., “GJ 581 update: Additional Evidence for a Super-Earth in the Habitable Zone,” accepted for publication in Astronomische Nachrichten (abstract). The Forveille paper is “The HARPS search for southern extra-solar planets XXXII. Only 4 planets in the Gl~581 system,” submitted to Astronomy & Astrophysics (abstract).
Deliberately deleting data as to remove evidence of a planet that another team discovered? Hmm. Reverse planet fevor. That Forveille et al. paper should never be accepted to Astronomy & Astrophysics.
This is certainly interesting news concerning the possibility that these potential planets may reside in the habitable zone. What I’m wondering about but I don’t believe was addressed was the question concerning orbital dynamics that would have permitted eccentric orbits to eventually progress to a more circular type. Perhaps such modeling might suggest that the current system is in fact undergoing a deliberate change between the eccentric and the circular type of system which might explain some of the current observations.
Moving on to other topics, I was wondering Paul if you have considered providing some coverage and information about the practical aspects of either robotic or human interstellar spaceflight. What I’m suggesting, is while we’ve had many fine discussions on propulsion systems and space observations, I’ve noticed a dearth of any mention of such topics concerning material sciences, the interstellar environment and its effects on electronics (e.g. such topics as radiation hardening, etc.), how particular future explorers will be able to recycle or not water, foodstuffs, and the like, and the myriad of all sorts of other problems that perhaps has been studied someplace and are now beginning to be looked into. A series of daily blogs on such topics here would be a nice change of pace.
Gliese 581 is really becoming a sad case study in “science by press release” isn’t it? At least Gliese 667Cc seems to be somewhat more secure as a detection, and hey who doesn’t like the idea of a habitable planet with three suns?
As for the habitability diagram, I have a few problems with the Planet Habitability Catalog. I suspect their use of “surface temperature” is skewing things: HD 85512b is receiving pretty much the same insolation from its sun as Venus does from ours, yet it seems to have got a surprisingly high score. That website is also very misleading in its comparison between exoplanets and solar system planets: they appear to be using blackbody temperatures for exoplanets but real surface temperatures for solar system planets. Obviously the latter quantity is unknown for exoplanets but mixing two rather different quantities under the generic label “surface temperature” is NOT a good thing.
william writes:
All this is fair game and I’ll keep these issues in mind. Thanks.
Se le orbite di questi pianeti sono davvero circolari, è possibile rilevarli con il metodo dei transiti?
Se così fosse, si potrebbe avere qualche informazione in più, al riguardo della composizione delle loro eventuali atmosfere…
Saluti da Antonio
Via Google Translate:
If the orbits of these planets are really circles, it is possible to detect the transit method?
If so, you could have some more information by reference to the composition of their atmospheres if any …
Greetings from Antonio
Decided to play around with the HARPS data in the Systemic console.
Using free eccentricities and a Keplerian model, after the first 4 planets (which correspond to the F11 eccentric model, including the suspiciously high eccentricity of the innermost planet) I end up with the next strongest frequency in the residuals being at 189.7698 days with a false alarm probability of 138%. This is not a very convincing probability! Furthermore fitting this period results in an implausibly high eccentricity that crosses the orbits of the other planets. There is a peak at 32 days with an even higher false alarm probability value, the fit does converge to a sensible eccentricity but it is going to be difficult to make a case for its existence based on such a weak signal.
Turning on planet-planet interactions results in the innermost planet’s eccentricity dropping to about 0.1, still suspiciously high but not quite as bad. The strongest peak in the residuals is now 191.1966 days with a FAP of 15.2%, which is better but still too high to claim a discovery. The FAP for the 32-day planet still remains above 100%.
With the all-circular fit, after the 4 planets, the next strongest peak is at 32.1075 days with a FAP of 5.15% followed by the 191.1966 day peak with FAP 7.76%. Situation is almost the same for the case with planet-planet interactions taken into account. These probabilities are certainly better than the eccentric-orbits case but still not really convincing enough to claim a discovery.
Situation for the case with fixed e=0 for the inner two or three planets is basically the same for the all-eccentric case.
From what I’m seeing, you can tune the mass of planet g all the way down to zero by adjusting the eccentricity of planet d: the data do not currently seem to allow you to distinguish between these possibilities. And even in the optimistic case of circular orbits, the false alarm probability for planet g is not convincing. For now, I’m going with the 4-planet model.
@jsampson I suspect you’re right as the Forveille et al. paper has been “in press” for a LONG time and is, I think, likely to never appear in the form ably refuted by Vogt and his team.
The worst thing about this whole mess is that there are no instruments in the pipeline adequate to provide a definitive resolution to this mess.
Nice analysis andy. As Steve Vogt put – we need to collect more data. Never a truer word said in exoplanet research.
I am a Zarmina enthusiast – I’m with Vogt et al on the merits of the hypothesis, but until more data is in, it’s just a hypothesis. Tantalisingly on the edge of discovery… or disappearing in the noise. Need to get that SNR higher!
Adam: “Need to get that SNR higher!”
That will help to lower the present strong dependence on model signatures (as andy discussed), however it may not be enough. There are periodic and aperiodic stellar brightness variations that will continue to confound analysis. Better will be statistically-independent detection methods, such as using star shades, to improve (lower) false positives and false negatives.
Andy, wouldn’t you expect the greater number of data points in the combined old data set (HARPS+HIRES) to have a dramatic impact on the false alarm probability as compared to the new extended set of the HARPS data. I know it is very hard working with two different sets, but I’m wondering if you can use these methods to give a ball-park figure for a new FAP of the combined set in a perfect analysis?
Given what an important breakthrough it was for Kepler to realize, and convince others, that the movements of the planets were not constrained to divinely perfect circles, it’s kind of surprising that now we have a case where uniform circular orbits are apparently the more physically valid solution and elliptical orbits are the solution being forced by selectively discarding data. What could make those orbits so uniform?
By itself the HIRES data only seems to be able to detect planets b and c: after that the next strongest period is at 26.2474 days with a FAP of 0.292%. This period does not correspond to any of the known planets in the system. An eccentric 2-planet fit to the HIRES data yields crossing orbits, the next period is again at 26.2474 days.
Doing an eccentric planets fit to the combined HARPS+HIRES datasets yields a lower orbital eccentricity for planet e and a higher eccentricity for d as compared to the HARPS-only fit. The next most significant period after fitting the four planets is at 49.8937 days with a FAP of 4.05%. Adding this leads to a strongly unstable system with crossing orbits. For an all-circular fit to the combined HARPS+HIRES dataset, the next strongest period is at 50.0101 days with a FAP of 0.789%.
In neither case (eccentric, circular) was there anything particularly significant at 32 days. This is a strong contrast to what I found when using the previous (shorter) HARPS dataset. Have to say I didn’t expect that, thanks for the suggestion to investigate the combined dataset!
Another thing to note here is the alias between the lunar month (29.53 days) and the year (365.25 days):
1/(1/29.53 – 1/365.25) = 32.1275 days
…which is essentially the same as the claimed period for planet g.
I bet all those worlds are heavily volcanic and have thick choking hot atmospheres as they are all quite massive, is there any word on the age of the central star? things may have cooled down over the ages.
This is all truly fascinating of course.
It surprises me a bit that Kepler-22b is among the top five potentially habitable exoplanets, because, though its mother star is the only sunlike star among them (a G5 star), the planet is at least soe 30 Me, which makes it a sub/ice giant (Neptune class) at least, if not a real gas giant.
If those planets are also included in the ‘potentially habitable exoplanets’, merely based on location in the HZ, then it should be possible to include more. I will have a brief look at it.
Ok, I did a preliminary check on the Extrasolar Planet Encyclopedia;
I converted all stellar apparent magnitudes to absolute magnitudes, those to luminosity ( * solar), and calculated the inner and outer boundaries of the HZ, generously assuming an inner edge of 0.9 AU and an outer edge of 1.5 AU for our own solar system.
I then found 48 planets (of the 777 mentioned) within the HZ, but a few may be wrong, I will need to do a closer check for that later.
Remarkably:
– Gliese 667C c is not within the HZ, it is on the outside. In fact, Gliese 667C b is near the inner edge.
– HD 85512 b is also not within the HZ, it is well on the inside of it.
– Gl 581 d is well outside the HZ, g is indeed within the HZ.
Kepler 22-b is indeed well within the HZ (from 0.63 to 1.05 AU) at 0.85 AU.
All other candidates within the HZ of their star are giant planets, a few notable ones:
– 55 Cancri f (0.14 MJ) is right within the HZ (from 0.66 to 1.1 AU) at 0.78 AU.
– Mu Arae b (1.7 MJ) is also within its HZ (from about 1.2 to 2 AU) at 1.5 AU.
– 16 Cygni b (1.7 MJ) is just within its HZ (from about 1.0 to 1.7 AU) at 1.68 AU.
– Ups Andromedae d may be within the HZ.
– Others up to 70 ly: HD 147513 b, HD 176051 b, HD 210277 b.
All (super) Jupiter mass, but if they have large moons…
Correction: 16 Cygni b should have been: 16 Cygni B b, since 16 Cygni is a binary, both components being solar type stars, quite unusual (Alpha Centauri and Zeta Reticuli being the other examples of solar binaries/multiples that I can think of).
Ronald:
Yes, I found that surprising as well. In fact, I doubt that any of the “top 5” planets listed there is even more Earth-like than Venus – density estimates from the Kepler planets suggest that a planet in the habitable zone needs to have less than 2 (!) Earth masses to be rocky, which none of those planets fulfills.
(Assuming that a planet’s density decreases both with higher mass and with higher temperature (more loss of volatiles), which seems highly plausible to me.)
@Ronald: the problem with your method is that it only takes into account the visible light. Cool stars radiate more of their energy at infrared wavelengths, hot stars more in the ultraviolet. Therefore your estimates of the habitable zone location will be too close to the star for non-solar-type stars. (Or to put it another way, habitable planets around hot or cool stars will receive less visible light than their equivalents around solar-type stars). You need to apply an appropriate bolometric correction to the magnitudes.
Using the bolometric luminosity of 0.01370 times solar from Anglada-Escudé et al. (2012) and the semimajor axis of 0.123 AU, the equivalent distance for Gl 667 Cc ends up as 1.05 AU, which is within your habitable zone limits.
@andy: you are right about that! It means that for M and later K stars the HZ is further outward.
That puts Gl 667 Cc within the HZ.
However, it puts HD 85512 b even further inside of the HZ (too hot).
I did make the bolometric correction for Gl 581 (total bolometric luminosity 0.0135 * solar) and planet d is still well outside the HZ then (at 0.22 AU, outer edge of HZ at 0.17 AU), even if the outer boundary of the HZ is assumed to be the equivalent of 1.5 AU in our solar system.