Saturday, July 5, 2014

Phantom Planets of Gliese 581

Figure 1. Cartoon published in 2007 on the announcement of the alleged exoplanet Gliese 581 d, which has now been declared an illusion induced by correlated noise in the stellar spectra
Exoplanetary astronomy is a brand-new field of inquiry, originating only in 1992 with the detection of two barren planets orbiting a distant neutron star (Wolszczan & Frail 1992). As an infant science, it is especially prone to controversy and revision.

No extrasolar hullabaloo has generated so much wordage or persisted so long as the debate on the nature of the planetary system around Gliese 581 (GJ 581), a red dwarf star located just 20 light years away. For the past seven years, this system has been repeatedly proposed as the home of one or more habitable Super Earths, although the precise number and characteristics of these marvelous worlds have remained in flux.

Just yesterday came a study that claims to settle, once and for all, the question of habitable planets around GJ 581. According to Paul Robertson and colleagues, the candidate planet most favored to support life – GJ 581 d – simply does not exist (Robertson et al. 2014). This news means that our Sun’s back yard lacks any plausible contenders for habitability, with the possible exception of GJ 667C.

Whether the study by Robinson’s team will actually close the book on this controversy remains to be seen. Members of the original discovery team (Stephane Udry, Xavier Bonfils, Xavier Delfosse, Thierry Forveille, Michel Mayor) have generally been quite protective of their analyses of the radial velocity data on GJ 581, and have publicly repudiated previous attempts to revise their picture of the system.

Here is an abbreviated rundown of the controversy, which has been mentioned more than once in this blog. Relevant publications appear in the reference list at the end.

origin of the controversy

In 2005, Bonfils and colleagues announced that radial velocity data collected by the HARPS spectrograph indicated the existence of a Neptune-mass exoplanet in a tight orbit around the nearby star GJ 581. The star has a spectral type of M3, a mass 30% Solar (0.3 Msol), and a luminosity only 1.3% Solar. Following standard naming protocols, the new planet was designated GJ 581 b. It is at least 16 times as massive as Earth (16 Mea), yet its orbital period is only about 5 days. Thus it merits the nickname of Hot Neptune, as it is similar in mass to the real Neptune (17.2 Mea) but even hotter than Mercury and Venus. At the time, this discovery was notable because GJ 581 was only the third M dwarf to be identified as a planet host. Nevertheless, astrobiologists took no interest, since a planet with the characteristics of GJ 581 b has no hope of nurturing life.

In 2007, things became much more interesting when the same HARPS team, now led by Udry, announced two new Super Earths orbiting in or near the proposed habitable zone of GJ 581. Planet c had a minimum mass of about 5 Mea, a semimajor axis of only 0.07 astronomical units (AU), and an orbital period of 13 days. Planet d, which was detected with less confidence, had a minimum mass of 7.7 Mea, a semimajor axis of 0.25 AU, and a period of 83 days. The discovery team described the orbital environment of GJ 581 c as “the warm edge of the habitable zone” and characterized the planet itself as “the closest Earth twin to date.” The more massive outer candidate, GJ 581 d, was positioned “close to the cold edge of the habitable zone.” The team called both objects “promising targets” for future studies. One member of the team, Xavier Forveille, delighted the press (CNN, Time Magazine, Agence France-Presse) with an especially mediagenic remark about GJ 581 c: “On the treasure map of the universe, one would be tempted to mark this planet with an X.”

In those pre-Kepler days, many researchers were inclined to believe that a planet 5 to 10 times the mass of Earth would just be a bigger version of Earth, rather than a smaller version of Uranus. I never bought that argument myself, so I was surprised by the flood of articles on the habitability of planet c that followed Dr. Forveille’s enthusiastic speculations. However, the exoplanet community recovered from this irrational exuberance pretty quickly. Before the year was out, at least two studies conceded that even for a dim red dwarf, a semimajor axis of 0.07 AU (one-sixth of Mercury’s) is just too hot for water and life (Selsis et al. 2007, Von Bloh et al. 2007). GJ 581 c might be a Super Venus, but it can't be a big sister to Earth.

Two years later, nevertheless, the HARPS team successfully refreshed its picture of GJ 581 with new data and analyses. Now led by Michel Mayor, the discoverer of the very first Hot Jupiter, they announced yet another planet, GJ 581 e, orbiting interior to planet b with a period of only three days. Its minimum mass – a quantity known as m sin (i) in the astronomical literature – was only 1.9 Mea, at that time the lowest value ever reported for an extrasolar planet. While any object with those characteristics must be a partially molten Hellworld, the identification of planet e led to a revision in the parameters of GJ 581 d and a welcome constraint on the orbital inclinations of the whole system. The orbital period, semimajor axis, and m sin (i) of planet d were all revised downward to 67 days, 0.22 AU, and 7 Mea, respectively. These reductions placed GJ 581 d more securely in the system habitable zone. Meanwhile, a stability analysis of the full ensemble of planets found that the maximum inclination of their orbits (presumed co-planar) to the line of sight was only 30%, enabling upper as well as lower limits on planet masses.

With this publication, planet d definitively replaced planet c as the most likely “habitable Super Earth.” Today, if you do a Google Image search for “Gliese 581 d,” you can still find an abundance of imaginary views of this putative Earth-like planet; here’s a pretty one by DarinK on deviantART.

In 2010, a completely different team led by Steven Vogt used radial velocity data from the HIRES program to confirm two new planets around GJ 581. One member of the pair, GJ 581 g, was described as a Super Earth of about 3 Mea with a period of 37 days, placing it squarely in the system habitable zone (Vogt et al. 2010). The other candidate, GJ 581 f, seemed to be a more straightforward Mini Neptune of about 7 Mea, orbiting at an astrocentric distance of 0.76 AU, implying chilly temperatures at its cloud tops. The smaller and warmer of these two worlds received instant media acclaim as the latest truly habitable exoplanet, inspiring still prettier pictures.

Figure 2. Gliese 581, Then and Now
In the upper panel, planets proposed by the HARPS team are labeled in red; planets proposed by the HIRES team are labeled in blue; planet b is assumed to have a hydrogen/helium envelope; planets e, c, g, and d are rocky; and planet f is rocky with an ice veneer. In the lower panel, planet e is assumed to be rocky, and planets b and c have hydrogen/helium envelopes. Colors are arbitrary.

Alas, even before the official publication of the discovery paper, astronomers with the HARPS survey sharply challenged the reality of GJ 581 g. As Francesco Pepe argued, “The signal amplitude of this potential fifth planet is very low and basically at the level of the measurement noise” (Mullen 2010). He considered it likely that the purported “signal” was “just produced ‘by chance’ out of the noise.”

The consensus of the astronomical community, as expressed over the next few years (Gregory 2011, Tuomi 2011, Forveille et al. 2011, Tuomi & Jenkins 2012), coalesced around the four-planet model favored by the HARPS team. This model was limited to the four objects designated by red letters in the upper panel of Figure 2.

Yet the rumblings continued. Roman Baluev, a Russian astronomer with no association with either HARPS or HIRES, published a new analysis of both datasets (Baluev 2013). His results not only refuted the existence of GJ 581 f and g but challenged the reality of GJ 581 d. Baluev argued that both planet-hunting teams, while recognizing the need to filter random white noise from their analyses, failed to realize that their data were also contaminated by “red noise” – a non-random signal whose origin was indeterminate. Once Baluev removed the red noise from his model, planet g disappeared and planet d became dubious; yet the signals for planets b, c, and e became stronger.

Figure 3. Red noise (red) and white noise (gray) in simulated radial velocity data; based on Figure 1 of Baluev 2013.
Baluev proposed two potential sources for this vexing red noise: either instrumental effects or some feature of the host star itself. Since he was working with two different datasets collected by two different instruments, he could easily rule out the first option. That left GJ 581 itself as the source of the variation, perhaps through magnetic activity resulting in starspots. Baluev also noted a bit ominously that “red RV noise might be a rather common phenomenon,” possibly relegating many other widely accepted detections to the freak tent of “unconfirmed, controversial.”

Despite so many naysayers, the ghost of GJ 581 g has found no rest. It appeared as recently as 2012 in a list of habitable planets published by the University of Puerto Rico Arecibo, and it was still discussed as a potentially real object in a study published only last month (Joiner et al. 2014).

And now, with the work of Robertson and colleagues, GJ 581 d has also been consigned to the extrasolar dustbin. Their approach resembles that of Baluev, insofar as they filtered correlated (non-random) noise from the radial velocity data on the host star. However, they analyzed only the HARPS dataset, since it covered a longer time span. Their conclusions are even more robust than those of Baluev: they confirm planets b, c, and e while completely ruling out planet d as well as planet g.

Robertson’s study foregrounds a key weakness of all previous analyses: until this work, no one had ever managed to define the rotation period of the host star. This was a crucial limitation, since one standard approach to avoiding spurious exoplanet detections is to test whether the proposed orbital period is related to the stellar rotation period. For the first time, Robertson et al. determined the rotation of GJ 581, which turns out to be a leisurely 130 days. As they argue, the signal identified as the phantom planet d was simply a harmonic of this activity, corresponding to approximately half of the rotation period.

where does all this leave us?

I have never been attached to the notion that either GJ 581 c or GJ 581 d might be habitable, because even the low end of their proposed mass ranges always exceeded 3 Mea (not to mention the hellish temperatures likely for the warmer of the two). Therefore, losing planet d as a possibly potentially habitable Super Earth brings no tears.

But I do find it troubling that a star system so nearby, to which so much attention has been devoted for so many years, was so profoundly misunderstood. If GJ 581 d isn’t real, then what about all the other controversial objects in the Sun’s back yard? Does Epsilon Eridani really harbor a cool gas giant? Does Alpha Centauri B really host a Hot Super Earth? Do HD 40307 and Tau Ceti really have extensive low-mass planetary systems? Or do all our exoplanetary dreams amount to a handful of dust?

Things probably aren’t that bad. Still, the GJ 581 affair once again reminds us that most data on extrasolar planets are provisional and subject to enormous changes at any moment. And so, on the assumption of innocence rather than guilt, here is a revision of a chart I published just last week, summarizing all systems discovered by radial velocity searches with three or more low-mass planets:

Figure 4. All twelve radial velocity systems with at least three low-mass planets
Semimajor axes are measured in astronomical units (AU), where 1 AU is the Earth-Sun separation; planet masses are indicated in Earth units (ME); star masses appear in Solar units at right.


Anglada-Escudé G. (2010) Aliases of the first eccentric harmonic: Is GJ 581g a genuine planet candidate? (First version, unpublished; a second unpublished draft with the collaboration of Rebekah Dawson is available at
Baluev R. (2013) The impact of red noise in radial velocity planet searches: Only three planets orbiting GJ 581? Monthly Notices of the Royal Astronomical Society 429, 2052-2068. Abstract:
Bonfils X, Forveille T, Delfosse X, et al. (2005) The HARPS search for southern extra-solar planets VI. A Neptune-mass planet around the nearby M dwarf Gl 581. Astronomy & Astrophysics, 443: L15-L18. Abstract.
Cameron AC. (2012) Astrophysical false positives. Nature 6 December 2012 (492:48-50).
Forveille T, Bonfils X, Delfosse X, et al. (2011) The HARPS search for southern extra-solar planets XXII. Only 4 planets in the Gl 581 system. Abstract:
Gregory PC. (2011) Bayesian re-analysis of the Gliese 581 exoplanetary system. Monthly Notices of the Royal Astronomical Society 415, 2523-2545. Abstract:
Joiner DA, Sul C, Dragomir D, Kane SR, Kress ME. (2014) A consistent orbital stability analysis for the GJ 581 system. Astrophysical Journal 788, 160.
Lestrade J-F, Matthews BC, Sibthorpe B, Kennedy GM, Wyatt MC, Bryden G, Greaves JS, Thilliez E, Moro-Martın A, and 11 others. (2012) A DEBRIS disk around the planet hosting M-star GJ 581 spatially resolved with Herschel. Astronomy & Astrophysics 548, A86.
Mayor M, Bonfils X, Forveille T, et al. (2009) The HARPS search for southern extra-solar planets XVIII. An Earth-mass planet in the GJ 581 planetary system. Astronomy & Astrophysics, 507: 487–494. Abstract.
Mullen L. (2010) Doubt Cast on Existence of Habitable Alien World. Astrobiology Magazine, 12 October 2010. Linked article.
Robertson P, Mahadevan S, Endl M, Roy A. (2014) Stellar activity masquerading as planets in the habitable zone of the M dwarf Gliese 581. Science, in press.
Selsis F, Kasting JF, Levrard B, et al. (2007) Habitable planets around the star Gliese 581? Astronomy & Astrophysics, 476: 1373–1387.
Tuomi M. (2011) Bayesian re-analysis of the radial velocities of Gliese 581. Evidence in favour of only four planetary companions. Astronomy & Astrophysics 528, L5. Abstract:
Tuomi M, Jenkins JS. (2012) Counting the number of planets around GJ 581. False positive rate of Bayesian signal detection methods. Abstract:  
Udry S, Bonfils X, Delfosse X, et al. (2007) The HARPS search for southern extrasolar planets XI. Super-Earths (5 & 8 Mea) in a 3-planet system. Astronomy & Astrophysics, 469: L43–L47. Abstract.
Vogt SS, Butler RP, Rivera EJ, Haghighipour N, Henry GW, Williamson MH. (2010) The Lick-Carnegie exoplanet survey: A 3.1 Mea planet in the habitable zone of the nearby M3V star Gliese 581. Astrophysical Journal 723, 954-965.
Von Bloh W, Bounama C, Cuntz M, Franck S. (2007) The habitability of super-Earths in Gliese 581. Astronomy & Astrophysics, 476: 1365-1371.
Von Braun K, Boyajian TS, Kane SR, van Belle GT, Ciardi DR, Lopez-Morales M, et al. (2011) Astrophysical parameters and habitable zone of the exoplanet hosting star GJ 581. Astrophysical Journal Letters, 729: L26. Abstract.
Von Paris P, Gebauer S, Godolt M, Grenfell JL, Hedelt P, Kitzmann D, et al. (2010) The extrasolar planet GL 581 d: A potentially habitable planet? Astronomy & Astrophysics, 522: A23. Abstract.
Wolszczan A, Frail DA. (1992) A planetary system around the millisecond pulsar PSR 1257+12. Nature 355, 145-147.
Wordsworth RD, Forget F, Selsis F, Madeleine J-B, Millour E, Eymet V. (2010) Is Gliese 581 d habitable? Some constraints from radiative-convective climate modeling. Astronomy & Astrophysics, 522: A22. Abstract.
Wordsworth RD, Forget F, Selsis F, Millour E, Charnay B, Madeleine J-B. (2011) Gliese 581 d is the first discovered terrestrial-mass exoplanet in the habitable zone. Astrophysical Journal Letters, 733: L48. Abstract.