The Disappearing Habitable Planets of GJ 581

During the past few days, word has started to spread about the latest study that has resulted in the “disappearance” of a pair of planets that some had claimed over the past few years as being potentially habitable: GJ 581d and g. These planets have not physically disappeared, of course. Instead, the subtle radial velocity variations that had originally been interpreted as being the result of as many as six planets orbiting the nearby red dwarf star, GJ 581, have instead been found to be the result of just three planets, the subtle effects of stellar magnetic activity and noise in the data. In the wake of these latest findings, there have been a chorus of nonscientists on the internet touting this as yet another example of the failure of science. To the contrary, this episode is a perfect example of how the scientific process is suppose to work.


Background History

GJ 581 (also known as Gliese 581) is a type M3V dwarf star located 20.3 light years away in the constellation of Libra. With an estimated mass of 0.31 times that of the Sun, a radius of 0.29 times and a luminosity of just 0.013 times, it is a fairly typical example of the numerous red dwarf stars that litter the solar neighborhood.  And like many nearby stars, it has been a subject of long term observations for extrasolar planets. In 2005, the European team using the HARPS (High Accuracy Radial Velocity Planet Searcher) spectrograph on ESO’s 3.6-meter telescope at the La Silla Observatory in Chile announced the discovery of a planet orbiting GJ 581 with a period of 5.4 days and a minimum mass of 17 times that of the Earth.

Two years later, the same team announced the detection of an additional pair of planets: GJ 581c in a 12.9-day orbit with a minimum mass of 5 times that of the Earth and GJ 581d in a 83-day orbit with a minimum mass of about 8 times that of the Earth. It was further claimed in the discovery paper by Udry et al. that GJ 581c orbited at the inner edge of the habitable zone of that system while GJ 581d orbited at the outer edge making each of these planet potentially habitable planet candidates. However, an analysis of this system by Selsis et al. published not long afterwards concluded that this habitability claim was overstated. GJ 581c receives about 30% more energy than Venus does and their modelling found that it is unlikely to be habitable. The situation with GJ 581d made it a better candidate with the amount of energy this planet received that was about the equivalent of an optimistic “early Mars” limit to the habitable zone.

I should also remind the reader (as I have done in many previous posts) that the Doppler velocity technique used to discover these planets only provides a measure of the minimum mass of planets since the inclination of the orbits of these planets to our line of sight is unknown. Without more information to constrain the actual mass of the planets as well as their general properties, it is more probable that planets with minimum masses this large are not even terrestrial planets never mind habitable planet. They are much more likely to be gas dwarfs, mini-Neptunes or larger nonterrestrial planets.

In 2009, the same European HARPS team announced the discovery of yet another planet in this system. GJ 581e had a minimum mass of 1.9 times that of the Earth and an orbital period of 3.2 days – far too close to its sun to be habitable. This new work by Mayor et al. allowed them to further refine the properties of the other planets especially GJ 581d which now seemed to be in a moderately eccentric orbit with a period of 67 days and a slightly lower minimum mass of about 7 times that of the Earth. Dynamical simulations that they performed of this four-planet system constrained their actual masses to be no greater than 1.6 times that of their measured minimum masses. Aside from the moderately eccentric orbit, the case for GJ 581d being a potentially habitable planet appeared to be improving somewhat (although by any objective measure, was still unlikely owing to its high minimum mass and unknown physical properties).

For habitable planet enthusiasts, the situation for this planetary system seemed to improve markedly in September 2010. Steven Vogt and his team combined the previously published HARPS measurements with 11 years of precision radial velocity measurements they had acquired using the HIRES spectrograph as part of the Lick-Carnegie Exoplanet Survey at the Keck Observatory that not only refined the properties of the four previously discovered planets, but uncovered the existence of two additional planets. GJ 581f was found in a distant orbit with a period of 433 days with a minimum mass of 7.0 times that of the Earth and GJ 581g, with a minimum mass of 3.1 times that of Earth, was discovered in a 36.6-day orbit comfortably inside the habitable zone.

An analysis by Wordsworth et al. about the potential habitability of GJ 581d published not long afterwards, however, concluded that this planet’s potential habitability was marginal at best given its slow rotation and low effective stellar flux. Only the presence of a “super greenhouse” effect involving gases in addition to water vapor and CO2 could raise the global temperatures to the freezing point of water.


Diagram showing the 6-planet system of GJ 581 as known in 2010 and an optimistic view of the habitable zone compared to our Solar System. (ESO/Henrykus)


Doubts Raised

But the situation with the new discoveries was far from settled. The HARPS team quickly questioned the existence of the two new discoveries and eventually presented more data to support their doubts. Using 121 new HARPS radial velocity measurements in addition to their previously published 119, they could find no evidence for the two planets announced by Vogt et al. in 2010 and found only the four planets they had discovered earlier in their expanded data set.

But before the preprint for these new HARPS results was available in September 2011, Philip Gregory of the University of British Columbia published his own detailed Bayesian analysis of the HIRES and HARPS data sets used by the Lick-Carnegie Exoplanet Survey. His analysis found evidence to support a five-planet system that excluded GJ 581g when looking only at the HARPS data set with a false alarm probability of 1%. By looking at the combined data sets, he could only reliably detect GJ581b and c. When he included the existence of some additional Gaussian error term of unknown origin, his analysis supported a four-planet interpretation of the results. He could not confirm the presence of the potentially habitable GJ 581g nor the more distantly orbiting GJ 581f in this combined data set.

In August 2012, Steven Vogt and his colleagues officially published their own analysis of the expanded HARPS data set. They claimed that they were able to detect GJ 581g with a minimum mass of 2.2 Earth-masses squarely inside the habitable zone with a less than a 4% probability of a false alarm. With this apparent confirmation, Abel Mendez Torres of the Planetary Habitability Laboratory issued a press release stating that GJ 581g was the most Earth-like extrasolar planet then known with an Earth Similarity Index (ESI) value of 0.92. GJ 581d, with an ESI of 0.72, was claimed to be the fifth most Earth-like as of July 20, 2012 despite the reservations expressed in the earlier work by Wordsworth et al..

But the issue was far from settled in the rest of the scientific community. Roman Baluev of the Pulkovo Observatory in St. Petersburg, Russia performed his own analysis of the HARPS and HIRES data sets. He found that the data contained an important correlated noise component which produced what he characterized as “misleading effects” in the earlier analyses. Using a different analysis technique that better accounted for the type of noise present in the data, he was able to definitively confirm the existence of GJ 581b, c and e. He found that GJ 581f and g were likely just illusions caused by the noise he identified and that the reality of GJ 581d was questioned.

American astronomer Artie Hatzes performed yet another analyses of the HARPS and HIRES data sets using a Fourier-based analysis technique that was published in 2013. He found evidence for a four-planet solution but could not confirm the presence of the potentially habitable GJ 581g or a significant signal with a periodicity of about 400 days corresponding to the more distantly orbiting GJ 581f.

After almost a decade of claims and counterclaims from the HARPS and Lick-Carnegie Exoplanet Survey teams as well as several independent analyses using a range of powerful mathematical techniques, it was clear that GJ 581b, c and most likely GJ 581e were real. Except for in the Lick-Carnegie Exoplanet Survey team’s analyses, there was no independent confirmation for the existence of the potentially habitable GJ 581g or the more distantly orbiting GJ 581f. And there were definitely some questions about the situation with GJ 581d orbiting at the extreme outer limit of the habitable zone.


New Findings

On July 3, a new analysis by a team led by Paul Robertson of Penn State was published on-line by Science. Instead of looking at just radial velocity measurements alone as had been done in the previous analyses, they examined the 239 publicly available HARPS spectra of GJ 581 from which these measurements were originally derived. While the HARPS team had looked for the signature of star spots and other obvious signs of stellar activity that have been known for decades to generate signals that can mimic those of extrasolar planets, this new analysis was looking for much more subtle variations in magnetic surface activity by examining the hydrogen-α emissions of GJ 581.


An artist’s depiction of star spots on a red dwarf star orbited by a planet. (David A. Aguilar – CfA)

Robertson’s team found that the radial velocity signature for GJ 581d seemed to be strongly correlated with the star’s magnetic activity.  In other words, when magnetic activity was high, the apparent radial velocity signature for GJ 581d was strong,  When magnetic activity was low, the signature for GJ 581d was weak. When the team corrected the radial velocity measurements for the subtle effects of the observed magnetic surface activity in GJ 581, they found that the signal for GJ 581d had decreased to just 1.5 times the noise level in the data or half the value typically required to constitute a statistically significant detection. Coupled with the fact that the new rotation period the team had derived for GJ 581 of 130±2 days was almost exactly twice that of the reported 66-day orbital period of GJ 581d, the radial velocity signal corresponding to the planet was most likely just a harmonic of the star’s rotationally modulated magnetic activity signal. The residual signal left after their correction was likely due to an imperfect correction for the magnetic activity they detected probably caused by shifts in the location of surface activity across the star’s surface over time which would slightly alter its Doppler signature.

Another casualty of this correction was GJ 581g whose radial velocity signature completely disappeared in the reprocessed data. In addition to loss of any sign of GJ 581d and g, the orbital eccentricity of GJ 581e that was present in earlier analyses also completely disappeared.  Its orbit was instead perfectly circular as would be expected for a planet in such a tight orbit around its sun. The parameters derived for the other two planets changed little from the previously published values. The properties of the three-planet solution derived by Robertson et al. are summarized in the table below. The effective stellar flux or insolation, Seff, is based on these parameters and the luminosity of GJ 581 from Bonfils et al..


Table 1: Summary of GJ 581 Planetary System Parameters (based on Robertson et al.)
Planet e b c
Mass (Earth=1) ≥1.7 ≥15.8 ≥5.5
Orbit Period (days) 3.1490 5.3686 12.914
Orbit Radius (AU) 0.02815 0.04061 0.0721
Orbit Eccentricity 0.00 0.00 0.00
Seff (Earth=1) 16 7.9 2.5


So despite the hopes of habitable planet enthusiasts, there are no potentially habitable planets currently known to be orbiting GJ 581. The radial velocity signatures of the very promising GJ 581g and the less promising GJ 581d were just artifacts of stellar magnetic activity modulated by the rotation of their sun. The original claims for the habitability of GJ 581c were also found to be overstated. Even taking into account recent work by Jun et al. about the potential habitability of synchronous rotating planets, which optimistically places the inner limit of the habitable zone for GJ 581 at an Seff of about 1.6, GJ 581c receives about 50% too much energy from its sun to be habitable. While there is still the possibility that Earth-size planets exist in the habitable zone of GJ 581, their future detection by means of precision radial velocity measurements are going to be complicated by the subtle effects of stellar magnetic activity and it may prove to be impossible.

While there are those who have pointed to this episode as further proof of the failure of science, nothing could be further from the truth. The past decade of experience with GJ 581 is a perfect example of how science is suppose to work. Science seeks to find natural explanations for observed phenomena. One of the key steps in this process is the formulation of scientific hypotheses that not only explain current observations but produce predictions of what future observations will find in order to prove or disprove the hypothesis. If a hypothesis fails these tests, it is either modified or discarded in favor of a new hypothesis that better explains the observations and the process begins again. It is this constant reassessing and self-correction that makes science such a powerful tool.


Related Reading

“GJ 832c: Habitable Super-Earth or Super Venus?”, Drew Ex Machina, June 27, 2014 [Post]

“Abundance of Earth Analogs”, Drew Ex Machina. June 25, 2014 [Post]

“Habitable Planet Reality Check: Kapteyn b”, Drew Ex Machina, June 6, 2014 [Post]

“Habitable Planet Reality Check: 55 Cancri f”, Drew Ex Machina, May 7, 2014 [Post]

“Habitable Planet Reality Check: Kepler 186f”, Drew Ex Machina, April 20, 2014 [Post]

“The Transition from Super Earth to Mini Neptune”, Drew Ex Machina, March 29, 2014 [Post]

“The Extremes of Habitability”, SETIQuest, Volume 4, Number 2, pp. 1-8, Second Quarter 1998 [Article]


General References

Roman V. Baluev, “The impact of red noise in radial velocity planet searches: only three planets orbiting GJ 581?”, Monthly Notices of the Royal Astronomical Society, Vol. 429, No. 3, pp. 2052-2068, March 2013

X. Bonfils et al., “The HARPS search for southern extra-solar planets. VI. A Neptune-mass planet around the nearby M dwarf Gl 581”, Astronomy and Astrophysics, Vol. 443, No. 3, pp. L15-L18, December I 2005

T. Forveille et al., “The HARPS search for southern extra-solar planets XXXII. Only 4 planets in the Gl~581 system”, arXiv:1109.2505, submitted September 12, 2011

Philip C. Gregory, “Bayesian re-analysis of the Gliese 581 exoplanet system”, Monthly Notices of the Royal Astronomical Society, Vol. 415, No. 3, pp. 2523-2545, August 2011

A.P. Hatzes, “An investigation into the radial velocity variability of GJ 581: On the significance of GJ 581g”, Astronomische Nachrichten, Vol. 334, No. 7, pp. 616-624, August 2013

M. Mayor, “The HARPS search for southern extra-solar planets. XVIII. An Earth-mass planet in the GJ 581 planetary system”, Astronomy and Astrophysics, Vol. 507, No. 1, pp. 487-49, November 2009

Paul Robertson, Suvrath Mahadevan, Michael Endi and Arpita Roy, “Stellar activity masquerading as planets in the habitable zone of the M dwarf Gliese 581”, Science Express, July 3, 2014 [Abstract and Paper Access]

F. Selsis et al. “Habitable planets around the star Gliese 581?”, Astronomy and Astrophysics, Vol. 476, No. 3, pp. 1373-1387, December 2007

Abel Mendez Torres, “Five Potential Habitable Exoplanets Now”, Planetary Habitability Laboratory Press Release, posted July 19, 2012, revised August 1, 2012 [Press Release]

S. Udry et al., “The HARPS search for southern extra-solar planets. XI. Super-Earths (5 and 8 M) in a 3-planet system”, Astronomy and Astrophysics, Vol. 469, No. 3, pp. L43-L47, July III 2007

Steven S. Vogt et al., “The Lick-Carnegie Exoplanet Survey: A 3.1 M Planet in the Habitable Zone of the Nearby M3V Star Gliese 581”, The Astrophysical Journal, Vol. 723, No. 1, pp. 954-965, November 2010

S.S. Vogt, R.P. Butler and N. Haghighipour, “GJ 581 update: Additional evidence for a Super-Earth in the habitable zone”, Astronomische Nachrichten, Vol.333, Issue 7, pp. 561-575, August 2012

Robin D. Wordsworth et al., “Gliese 581d is the First Discovered Terrestrial-mass Exoplanet in the Habitable Zone”, The Astrophysical Journal Letters, Vol. 733, No. 2, article id. L48, June 2011

Jun Yang, Gwanel Boue, Daniel C. Fabrycky and Dorian S. Abbot, “Strong Dependence of the Inner Edge of the Habitable Zone on Planetary Rotation Rate”, The Astrophysical Journal Letters, Vol. 787, No. 1, Article id. L2, April 25, 2014