«Received ; accepted ms-press UCO/Lick Observatory, University of California, Santa Cruz, CA 95064 Department of Terrestrial Magnetism, Carnegie ...»
The Lick-Carnegie Exoplanet Survey: A 3.1 M⊕ Planet in the
Habitable Zone of the Nearby M3V Star Gliese 581
Steven S. Vogt1, R. Paul Butler2, E. J. Rivera1, N. Haghighipour3, Gregory W. Henry4,
and Michael H. Williamson4
Received ; accepted
UCO/Lick Observatory, University of California, Santa Cruz, CA 95064
Department of Terrestrial Magnetism, Carnegie Institution of Washington, 5241 Broad
Branch Road, NW, Washington, DC 20015-1305
Institute for Astronomy and NASA Astrobiology Institute, University of Hawaii-Manoa, Honolulu, HI 96822 Tennessee State University, Center of Excellence in Information Systems, 3500 John A.
Merritt Blvd., Box 9501, Nashville, TN. 37209-1561 –2– ABSTRACT We present 11 years of HIRES precision radial velocities (RV) of the nearby M3V star Gliese 581, combining our data set of 122 precision RVs with an existing published 4.3-year set of 119 HARPS precision RVs. The velocity set now indicates 6 companions in Keplerian motion around this star. Diﬀerential photometry indicates a likely stellar rotation period of ∼ 94 days and reveals no signiﬁcant periodic variability at any of the Keplerian periods, supporting planetary orbital motion as the cause of all the radial velocity variations. The combined data set strongly conﬁrms the 5.37-day, 12.9-day, 3.15-day, and 67-day planets previously announced by Bonﬁls et al. (2005), Udry et al. (2007), and Mayor et al. (2009). The observations also indicate a 5th planet in the system, GJ 581f, a minimum-mass 7.0 M⊕ planet orbiting in a 0.758 AU orbit of period 433 days and a 6th planet, GJ 581g, a minimum-mass 3.1 M⊕ planet orbiting at
0.146 AU with a period of 36.6 days. The estimated equilibrium temperature of GJ 581g is 228 K, placing it squarely in the middle of the habitable zone of the star and oﬀering a very compelling case for a potentially habitable planet around a very nearby star. That a system harboring a potentially habitable planet has been found this nearby, and this soon in the relatively early history of precision RV surveys, indicates that η⊕, the fraction of stars with potentially habitable planets, is likely to be substantial. This detection, coupled with statistics of the incompleteness of present-day precision RV surveys for volume-limited samples of stars in the immediate solar neighborhood suggests that η⊕ could well be on the order of a few tens of percent. If the local stellar neighborhood is a representative sample of the galaxy as a whole, our Milky Way could be teeming with potentially habitable planets.
–3– Subject headings: stars: individual: GJ 581 HIP 74995 –
There are now nearly 500 known extrasolar planets, and discovery work continues apace on many fronts: by radial velocities (RV), gravitational microlensing, transit surveys, coronography, nulling interferometry, and astrometry. By far the most productive discovery technique to date has been through the use of precision RVs to sense the barycentric reﬂex velocity of the host star induced by unseen orbiting planets. In recent years, the world’s leading RV groups have improved precision down to the ∼1 ms−1 level, and even below, extending detection levels into the range of planets with masses less than 10 M⊕, commonly referred to as “Super-Earths”. This level of precision is now bringing within reach one of the holy grails of exoplanet research, the detection of ∼Earth-size planets orbiting in the habitable zones (HZ) of stars. Nearby K and M dwarfs oﬀer the best possibility of such detections, as their HZ’s are closer in, with HZ orbital periods in the range of weeks to months rather than years. These low mass stars also undergo larger reﬂex velocities for a given planet mass. To this end, we have had a target list of ∼400 nearby quiet K and M dwarfs under precision RV survey with HIRES at Keck for the past decade.
One of these targets, the nearby M3V star GJ 581 (HIP 74995), has received considerable attention in recent years following the announcement by Bonﬁls et al. (2005), hereafter Bonﬁls05, of a 5.37-day hot-Neptune (GJ 581b, or simply planet-b) around this star. More recently, the Geneva group (Udry et al. 2007), hereafter Udry07, announced the detection of two additional planets (c and -d) in this system, one close to the inner edge of the HZ of this star and the other close to the outer edge. Planet-c was reported to have a period of 12.931 days and m sin i = 5.06 M⊕ whereas planet-d was reported to have a period of 83.4 days and m sin i = 8.3 M⊕.
our solar system and potentially in the regime of rocky planets or Super-Earths) and its location near the inner edge of the HZ of this star. An assumed Bond albedo of 0.5 yielded a simple estimate of ∼320 K for the equilibrium temperature of the planet, suggesting the possibility that it was a habitable Super-Earth. However, a more detailed analysis by Selsis et al. (2007), that included the greenhouse eﬀect and the spectral energy distribution of GJ 581, concluded that planet-c’s surface temperature is much higher than the equilibrium temperature calculated by Udry07 and that it is unlikely to host liquid water on its surface.
Selsis et al. (2007) concluded that both planets c and d are demonstrably outside the conservative HZ of this star, but that given a large atmosphere, planet-d could harbor surface liquid water. Chylek & Perez (2007) reached a similar conclusion that neither planets c nor d is in the HZ, but that planet-d could achieve habitability provided a greenhouse eﬀect of 100 K developed. Moreover, if these planets are tidally spin-synchronized, planet-c could conceivably have atmospheric circulation patterns that might support conditions of habitability. von Bloh et al. (2007) also concluded that planet-c is too close to the star for habitability. They argue, however, that if planet-d has a thick atmosphere and is tidally locked, it may lie just within the outer edge of the HZ. Both von Bloh et al. (2007) and Selsis et al. (2007) conclude that planet-d would be an interesting target for the planned TPF/Darwin missions.
Beust et al. (2008) studied the dynamical stability and evolution of the GJ 581 system using the orbital elements of Udry07, which they integrated forward for 108 years.
They observed bounded chaos (see e.g. Laskar (1997)), with small-amplitude eccentricity variations and stable semi-major axes. Their conclusions were unaﬀected by the presence of any as-yet-undetected outer planets. On dynamical stability grounds, they were able to exclude inclinations i ≤ 10◦ (where i = 0◦ is face-on).
Last year, Mayor et al. (2009), hereafter Mayor09, published a velocity update wherein –6– they revised their previous claim of an 8 M⊕ planet orbiting with an 83-day period, to a
7.1 M⊕ planet orbiting at 67-days, citing confusion with aliasing for the former incorrect period. Mayor09 also reported another planet in the system at 3.148 days with a minimum mass of 1.9 M⊕. They also presented a dynamical stability analysis of the system. In particular, the addition of the 3.15d planet, GJ 581e, greatly strengthened the inclination limit for the system. The planet was quickly ejected for system inclinations less than 40◦.
This dynamical stability constraint implies an upper limit of 1.6 to the 1/ sin i correction factor for any planet’s minimum mass (assuming coplanar orbits). Most recently, Dawson and Fabrycky (2010) published a detailed study of the eﬀects of aliasing on the GJ 581 data set of Mayor09. They concluded that the 67-day period of GJ 581c remains ambiguous, and favored a period of 1.0125 days that produced aliases at both 67 days and 83 days.
The Gliese 581 system exerts an outsize fascination when compared to many of the other exoplanetary systems that have been discovered to date. The interest stems from the fact that two of its planets lie tantalizingly close to the expected threshold for stable, habitable environments, one near the cool edge, and one near the hot edge. We have had GJ 581 under survey at Keck Observatory for over a decade now. In this paper, we bring 11 years of HIRES precision RV data to bear on this nearby exoplanet system. Our new data set of 122 velocities, when combined with the previously published 119 HARPS velocities, eﬀectively doubles the amount of RVs available for this star, and almost triples the time base of those velocities from 4.3 years to 11 years. We analyze the combined precision RV data set and discuss the remarkable planetary system that they reveal.
yielding a typical S/N ratio per pixel of 140. Doppler shifts are measured by placing an Iodine absorption cell just ahead of the spectrometer slit in the converging f/15 beam from the telescope. This gaseous absorption cell superimposes a rich forest of Iodine lines on the stellar spectrum, providing a wavelength calibration and proxy for the point spread function (PSF) of the spectrometer. The Iodine cell is sealed and temperature-controlled to 50 ± 0.1 C such that the column density of Iodine remains constant (Butler et al. 1996).
For the Keck planet search program, we operate the HIRES spectrometer at a spectral resolving power R ≈ 70,000 and wavelength range of 3700 – 8000 ˚, though only the region A 5000 – 6200 ˚ (with Iodine lines) is used in the present Doppler analysis. Doppler shifts A from the spectra are determined with the spectral synthesis technique described by Butler et al. (1996). The Iodine region is divided into ∼700 chunks of 2 ˚ each. Each chunk A produces an independent measure of the wavelength, PSF, and Doppler shift. The ﬁnal measured velocity is the weighted mean of the velocities of the individual chunks.
In August 2004, we upgraded the focal plane of HIRES to a 3-chip CCD mosaic of ﬂatter and more modern MIT-Lincoln Labs CCD’s. No zero point shift in our RV pipeline was incurred from the detector upgrade. Rather, the new CCD mosaic eliminated a host of photometric problems with the previous Tek2048 CCD (non-ﬂat focal plane, non-linearity of CTE, charge diﬀusion in the silicon substrate, overly-large pixels, and others). The deleterious eﬀects of all these shortcomings can be readily seen as larger uncertainties on the pre-August 2004 velocities.
In early 2009, we submitted a paper containing our RVs up to that date for GJ 581 that disputed the 83-day planet claim of Mayor09. One of the referees (from the HARPS team) kindly raised the concern (based partly on our larger value for apparent stellar jitter) that we may have some residual systematics that could be aﬀecting the reliability of some of our conclusions. In the precision RV ﬁeld there are no suitable standards by which –8– teams can evaluate their performance and noise levels; so, it is rare but also extremely useful for teams to be able to check each other using overlapping target stars, like GJ 581, for inter-comparison. So, we took the HARPS team’s concerns to heart and withdrew our paper to gather another season of data, to do a detailed reanalysis of our uncertainty estimates, and to scrutinize our 15-year 1500-star data base for evidence of undiscovered systematic errors.
Soon after we withdrew our 2009 paper, Mayor09 published a revised model wherein they altered their 83-day planet period to 66.8 days (citing confusion by yearly aliases) and also announced an additional planet in the system near 3.15 days. For our part, as a result of our previous year’s introspection, we discovered that the process by which we derive our stellar template spectra was introducing a small component of additional uncertainty that added about 17% to our mean internal uncertainties. This additional noise source stems from the deconvolution process involved in deriving stellar template spectra. This process works quite well for G and K stars, but it is prone to extra noise when applied to heavily line-blanketed M dwarf spectra. We have included this in our present reported uncertainties for GJ 581, and are working on improvements to the template deconvolution process. Furthermore, our existing template for this star, taken many years ago, was not up to the task of modeling RV variation amplitudes down in the few ms−1 regime. So, over the past year, we obtained a much higher quality template for GJ 581.
The HIRES velocities of GJ 581 are presented in Table 1, corrected to the solar system barycenter. Table 1 lists the JD of observation center, the RV, and the internal uncertainty.
The reported uncertainties reﬂect only one term in the overall error budget, and result from a host of systematic errors from characterizing and determining the PSF, detector imperfections, optical aberrations, eﬀects of under-sampling the Iodine lines, etc. Two additional major sources of error are photon statistics and stellar jitter. The former is –9– already included in our Table 1 uncertainties. The latter varies widely from star to star, and can be mitigated to some degree by selecting magnetically-inactive older stars and by time-averaging over the star’s unresolved low-degree surface p-modes. The best measure of overall precision for any given star is simply to monitor an ensemble of planet-free stars of similar spectral type, chromospheric activity, and apparent magnitude, observed at similar cadence and over a similar time base. Figures 2, 3, and 4 of Butler et al. (2008) show 12 M dwarfs with B-V, V magnitude, and chromospheric activity similar to GJ 581. In any such ensemble, it is diﬃcult to know how much of the root-mean-square (RMS) of the RVs is due to as-yet-undiscovered planets and to stellar jitter. However, these stars do establish that our decade-long precision is better than 3 ms−1 for M dwarfs brighter than V=11, including contributions from stellar jitter, photon statistics, undiscovered planets, and systematic errors.
3. Properties of GJ 581