Special Paper A Reappraisal of the Habitability of Planets ... · emission and scattered light...

of 36/36
30 ASTROBIOLOGY Volume 7, Number 1, 2007 © Mary Ann Liebert, Inc. DOI: 10.1089/ast.2006.0124 Special Paper A Reappraisal of the Habitability of Planets Around M Dwarf Stars JILL C. TARTER, 1 PETER R. BACKUS, 1 ROCCO L. MANCINELLI, 1 JONATHAN M. AURNOU, 2 DANA E. BACKMAN, 1 GIBOR S. BASRI, 3 ALAN P. BOSS, 4 ANDREW CLARKE, 5 DRAKE DEMING, 6 LAURANCE R. DOYLE, 1 ERIC D. FEIGELSON, 7 FRIEDMANN FREUND, 1 DAVID H. GRINSPOON, 8 ROBERT M. HABERLE, 9 STEVEN A. HAUCK, II, 10 MARTIN J. HEATH, 11 TODD J. HENRY, 12 JEFFERY L. HOLLINGSWORTH, 9 MANOJ M. JOSHI, 13 STEVEN KILSTON, 14 MICHAEL C. LIU, 15 ERIC MEIKLE, 16 I. NEILL REID, 17 LYNN J. ROTHSCHILD, 9 JOHN SCALO, 18 ANTIGONA SEGURA, 19 CAROL M. TANG, 20 JAMES M. TIEDJE, 21 MARGARET C. TURNBULL, 4 LUCIANNE M. WALKOWICZ, 22 ARTHUR L. WEBER, 1 and RICHARD E. YOUNG 9 ABSTRACT Stable, hydrogen-burning, M dwarf stars make up about 75% of all stars in the Galaxy. They are extremely long-lived, and because they are much smaller in mass than the Sun (between 0.5 and 0.08 M Sun ), their temperature and stellar luminosity are low and peaked in the red. We have re-examined what is known at present about the potential for a terrestrial planet forming within, or migrating into, the classic liquid–surface–water habitable zone close to an 1 SETI Institute, Mountain View, California. 2 University of California Los Angeles, Los Angeles, California. 3 University of California Berkeley, Berkeley, California. 4 Carnegie Institution of Washington, Washington, D.C. 5 British Antarctic Survey, Cambridge, United Kingdom. 6 NASA Goddard Space Flight Center, Greenbelt, Maryland. 7 Pennsylvania State University, University Park, Pennsylvania. 8 Southwest Research Institute; and 14 Ball Aerospace & Technologies Corporation, Boulder, Colorado. 9 NASA Ames Research Center, Moffett Field, California. 10 Case Western Reserve University, Cleveland, Ohio. 11 Ecospheres Project, Greenwich, United Kingdom. 12 Georgia State University, Atlanta, Georgia. 13 U.K. Meteorological Office, Reading, United Kingdom. 15 University of Hawaii, Honululu, Hawaii. 16 National Center for Science Education, Oakland, California. 17 Space Telescope Science Institute, Baltimore, Maryland. 18 University of Texas at Austin, Austin, Texas. 19 California Institute of Technology, Pasadena, California. 20 California Academy of Sciences, San Francisco, California. 21 Michigan State University, East Lansing, Michigan. 22 University of Washington, Seattle, Washington.
  • date post

    26-Dec-2019
  • Category

    Documents

  • view

    0
  • download

    0

Embed Size (px)

Transcript of Special Paper A Reappraisal of the Habitability of Planets ... · emission and scattered light...

  • 30

    ASTROBIOLOGYVolume 7, Number 1, 2007© Mary Ann Liebert, Inc.DOI: 10.1089/ast.2006.0124

    Special Paper

    A Reappraisal of the Habitability of Planets Around M Dwarf Stars

    JILL C. TARTER,1 PETER R. BACKUS,1 ROCCO L. MANCINELLI,1JONATHAN M. AURNOU,2 DANA E. BACKMAN,1 GIBOR S. BASRI,3

    ALAN P. BOSS,4 ANDREW CLARKE,5 DRAKE DEMING,6 LAURANCE R. DOYLE,1ERIC D. FEIGELSON,7 FRIEDMANN FREUND,1 DAVID H. GRINSPOON,8ROBERT M. HABERLE,9 STEVEN A. HAUCK, II,10 MARTIN J. HEATH,11

    TODD J. HENRY,12 JEFFERY L. HOLLINGSWORTH,9 MANOJ M. JOSHI,13STEVEN KILSTON,14 MICHAEL C. LIU,15 ERIC MEIKLE,16 I. NEILL REID,17

    LYNN J. ROTHSCHILD,9 JOHN SCALO,18 ANTIGONA SEGURA,19 CAROL M. TANG,20JAMES M. TIEDJE,21 MARGARET C. TURNBULL,4 LUCIANNE M. WALKOWICZ,22

    ARTHUR L. WEBER,1 and RICHARD E. YOUNG9

    ABSTRACT

    Stable, hydrogen-burning, M dwarf stars make up about 75% of all stars in the Galaxy. Theyare extremely long-lived, and because they are much smaller in mass than the Sun (between0.5 and 0.08 MSun), their temperature and stellar luminosity are low and peaked in the red.We have re-examined what is known at present about the potential for a terrestrial planetforming within, or migrating into, the classic liquid–surface–water habitable zone close to an

    1SETI Institute, Mountain View, California.2University of California Los Angeles, Los Angeles, California.3University of California Berkeley, Berkeley, California.4Carnegie Institution of Washington, Washington, D.C.5British Antarctic Survey, Cambridge, United Kingdom.6NASA Goddard Space Flight Center, Greenbelt, Maryland.7Pennsylvania State University, University Park, Pennsylvania.8Southwest Research Institute; and 14Ball Aerospace & Technologies Corporation, Boulder, Colorado.9NASA Ames Research Center, Moffett Field, California.10Case Western Reserve University, Cleveland, Ohio.11Ecospheres Project, Greenwich, United Kingdom.12Georgia State University, Atlanta, Georgia.13U.K. Meteorological Office, Reading, United Kingdom.15University of Hawaii, Honululu, Hawaii.16National Center for Science Education, Oakland, California.17Space Telescope Science Institute, Baltimore, Maryland.18University of Texas at Austin, Austin, Texas.19California Institute of Technology, Pasadena, California.20California Academy of Sciences, San Francisco, California.21Michigan State University, East Lansing, Michigan.22University of Washington, Seattle, Washington.

  • HABITABLE PLANETS AROUND M DWARFS 31

    1. INTRODUCTION

    STARTING WITH THE WORK OF HUANG (1959, 1960)and continuing with Dole’s Habitable Planetsfor Man (1964), astronomers have considered starswhose masses greatly exceed or fall below themass of the Sun to be inhospitable to biology, par-ticularly the complex, intelligent variety. Thisconclusion stemmed from simple arguments.Massive stars live their lives too quickly—mea-sured in millions of years—and exhaust their nu-clear fuel too rapidly (self-destructing in spectac-ular fashion) to permit the multi-billion yearevolutionary time scales that, on Earth, were re-quired to convert stardust into beings capable ofcontemplating the stars. While low mass starshave much longer lifetimes, their luminosity is sofeeble that any planet would need to be nestledvery close to the star to permit the possibility ofhaving a surface temperature conducive to liquidwater, which seems to be the sine qua non of lifeas we know it. In such small orbits [0.1–0.35 as-tronomical units (AU) for an M0 star, and closerstill for smaller stars], any planet would becometidally locked to the star. That is, it would keepthe same side continuously facing the star (ananalog is the Earth’s moon). Early, and perhapsoverly simplistic, arguments suggested that sucha planet’s atmosphere would boil off on the star-facing side and freeze out on the dark side—notthe most clement conditions for life. As a result,for decades, exobiologists and search for ex-traterrestrial intelligence (SETI) researchers ex-

    cluded M dwarf stars from their consideration.However, in 1994, an international conference re-considered the question of circumstellar habitablezones (HZs) and concluded that late K and earlyM dwarf stars provide the best opportunity forthe existence of environments that are continu-ously habitable over a 4.6-Gyr period, as was required by evolutionary time scales on Earth(Doyle, 1996). Based on a simple energy-balancemodel, it was argued that atmospheric heat trans-port could prevent freeze-out on the dark side(Haberle et al., 1996). More sophisticated three-dimensional climate modeling by Joshi et al.(1997) suggested that a surface pressure of as lit-tle as 0.1 bar could prevent atmospheric collapseon the dark side, while a surface pressure of 1–2bars could allow liquid water on most parts ofthe surface. However, the magnetic activity andconsequent flaring of M dwarfs were consideredproblematic because of the large flux of ultravio-let (UV)-B that would be delivered to the surfaceof the planet. Later studies (Segura et al., 2005) in-dicated that abiotic production of atmosphericozone could mitigate any flare activity. However,because planets in the HZs of dwarf M starswould likely be so very unearthly, they have notattracted a great deal of attention until now.

    The discovery of the first extrasolar planet(Mayor and Queloz, 1995) and the realization thattype I and II migration plus N-body dynamicscan dramatically rearrange the geometry of plan-etary systems subsequent to their formation (nu-merous authors in Deming and Seager, 2003) and,

    M dwarf star. Observations of protoplanetary disks suggest that planet-building materials arecommon around M dwarfs, but N-body simulations differ in their estimations of the likeli-hood of potentially habitable, wet planets that reside within their habitable zones, which areonly about one-fifth to 1/50th of the width of that for a G star. Particularly in light of theclaimed detection of the planets with masses as small as 5.5 and 7.5 MEarth orbiting M stars,there seems no reason to exclude the possibility of terrestrial planets. Tidally locked syn-chronous rotation within the narrow habitable zone does not necessarily lead to atmosphericcollapse, and active stellar flaring may not be as much of an evolutionarily disadvantageousfactor as has previously been supposed. We conclude that M dwarf stars may indeed be vi-able hosts for planets on which the origin and evolution of life can occur. A number of plan-etary processes such as cessation of geother mal activity or thermal and nonthermal atmos-pheric loss processes may limit the duration of planetary habitability to periods far shorterthan the extreme lifetime of the M dwarf star. Nevertheless, it makes sense to include M dwarfstars in programs that seek to find habitable worlds and evidence of life. This paper presentsthe summary conclusions of an interdisciplinary workshop (http://mstars.seti.org) sponsoredby the NASA Astrobiology Institute and convened at the SETI Institute. Key Words: Plan-ets—Habitability—M dwarfs—Stars. Astrobiology 7, 30–65.

  • thus, deliver and remove potentially habitableplanets from an HZ expanded astronomers’ at-tention beyond planetary systems that were ex-act analogs of our Solar System. This broadeningof astronomical perspective combined with thediscovery of hydrothermal habitats that are ap-parently independent of the Sun (Corliss et al.,1979), an increasing appreciation for the incredi-ble tenacity of extremophiles on Earth (Roth-schild and Mancinelli, 2001), and the realizationthat M dwarfs make up perhaps 75% or more ofthe stars (excluding brown dwarfs) and half ofthe total stellar mass in the Galaxy (Henry, 2004)have, once again, raised the issue of the suitabil-ity of M dwarfs as hosts for habitable worlds. Sev-eral authors (for example, Basri et al., 2004) haveeven challenged their colleagues with conferenceposters arguing that M dwarfs may be the favoredlocations for the origin and evolution of life.

    We have every reason to believe that M starsare favorable hosts for planets. Protoplanetary ac-cretion disks are as common around low massstars as they are around solar-type stars, witharound half of all objects at ages of a few Myrpossessing such disks (e.g., Haisch et al., 2001; Liuet al., 2004; Sicilia-Aguilar et al., 2005; Lada et al.,2006). Older debris disks around M stars are thusfar rare, but their paucity may be due to the lim-ited sensitivity of previous surveys. One of thevery nearest young M stars—the 12 Myr old M0.5star AU Mic, at a distance of only 10 parsecs(pc)—possesses a prominent cold, dusty disk,which is easily detected by thermal continuumemission and scattered light (Kalas et al., 2004; Liuet al., 2004). The AU Mic disk likely represents alate stage of planetary accretion in which plane-tary-mass objects have stirred up remnant plan-etesimals to collide and fragment in a Kuiper Belt-like structure. Thus, the potential for planetformation around M stars appears to be robust.

    Ground-based radial velocity searches havebeen looking for extrasolar planets around sev-eral hundred M dwarfs and have found severalplanets to date (http://exoplanets.org, http://obswww.unige.ch/!udry/planet/planet.html). Thelowest mass exoplanets orbiting main sequencestars, with claimed masses of 5.5 and !7.5 MEarth,orbit M dwarfs (Rivera et al., 2005; Beaulieu et al.,2006). Regardless of whether small planets arepreferentially formed around M dwarf stars, thecurrent detection technologies preferentially de-tect them there. For a given planet mass and ra-dius, both the radial velocity reflexes and the pho-

    tometric transit depths are larger for lower massstellar hosts. However, no rocky planets have yetbeen discovered within an M dwarf HZ, and N-body simulations (Ida and Lin, 2005; Mont-gomery and Laughlin, 2006; Raymond et al., 2006)provide conflicting answers regarding the likeli-hood of such planets. Ground-based observationsusing radial velocity techniques as well as tran-sit photometry may soon resolve the issue.

    Plans for increasingly sophisticated space mis-sions to search for terrestrial planets around nearbystars and biosignatures in their atmospheres (e.g.,Darwin at http://sci.esa.int/science-e/www/area/index.cfm?fareaid"28, CoRoT at http://sci.esa.int/science-e/www/area/index.cfm?fareaid"39, andthe Terrestrial Planet Finder Coronagraph andInferometer at http://planetquest.jpl.nasa.gov/TPF/tpf_index.html) demand high-fidelity findinglists of nearby stars that are most likely to harborhabitable planets. If these should turn out to be M dwarf stars, then constraints imposed by theplanet/star contrast ratio will be eased, but thechallenges of spatially resolving a planet in the HZwill be significantly enhanced (Turnbull, personalcommunication). The Kepler mission (Borucki et al.,2003) is scheduled to launch in 2008 and shouldprovide, within a few years thereafter, the demo-graphics of terrestrial-size planets within the HZsof stars as a function of spectral type, from F to atleast M5 dwarfs.

    The Allen Telescope Array is currently underconstruction in Northern California (http://astron.berkeley.edu/ral/ata/), and efficient use of itswide field of view for commensal search for SETIobservations requires a large list of target starsaround which habitable planets and their lifeforms might exist. In creating a list of “Habstars”to be used to guide SETI observations with theAllen Telescope Array, Turnbull and Tarter(2003a,b) did not explicitly exclude M dwarfs, ex-cept for those that express high levels of chro-mospheric activity. Even so, the magnitude limitof the parent Hipparcos sample meant that pro-portionately few M dwarfs ended up in the finaltarget list (!600 M stars of !19,000 total Hab-stars). A SETI target list of Habstars incorporatesan additional selection criterion related to thestellar age, requiring at least 3 Gyr for evolutionto have potentially led to technological civiliza-tions. If M dwarfs are viable Habstars, then ad-ditional efforts will be required to inventory thelocal neighborhood and overcome the biases ofmagnitude limited sampling.

    TARTER ET AL.32

  • HABITABLE PLANETS AROUND M DWARFS 33

    Although we may soon know whether appro-priate planets lie within the HZs of M dwarf stars,we are a long way from knowing whether theyare likely to be inhabited. This paper presents thesummary conclusions of an interdisciplinaryworkshop (http://mstars.seti.org) sponsored bythe NASA Astrobiology Institute and convenedat the SETI Institute to consider what is reason-ably known about the topic and what theoreticaland observational projects might be undertakenin the near term to constrain the possibilities. Sec-tion 2 assesses what is known about M dwarfstars, the formation of planets around M dwarfs,the probable atmospheric characteristics of plan-ets orbiting within the HZ of such stars, and thebasic requirements for the origin and evolutionof biology on any such planets. Section 3 makesa differential comparison between a G dwarf andM dwarf stellar host, enumerates the most favor-able and unfavorable environmental conditionsfor biology (in a relative way), and examinesthose processes that might terminate the habit-ability of a favored planet long before the host Mdwarf star ceases fusing hydrogen in its core. Inconclusion (Section 4), we suggest a frameworkfor research to be conducted in the next 2 yearsthat should better delimit the constraints on thespectral type and ages of M dwarf stars that maybe appropriate targets for missions in search ofevidence of life beyond Earth, including that oftechnological civilizations. A second workshop in2008 will provide the opportunity for re-exami-nation and further refinements to our ap-proaches.

    2. REQUIREMENTS FOR HABITABILITY

    2.1. Nomenclature

    To deal with large numbers and phenomenol-ogy they can measure, but perhaps not under-stand, astronomers have a long history of usingclassification schemes that are not always intu-itively understood by scientists in other fields orthe general public. At the risk of offending somewell-schooled readers, this section presents aminiguide to those classification schemes that areused in this paper to discuss the habitability ofplanets orbiting M dwarf stars.

    “M” is a stellar classification based on the char-acteristic features found in relatively low-resolu-tion spectra, initially obtained at optical wave-

    lengths nearly a century ago. It was later under-stood that the observable spectral features pro-vide information about stellar surface tempera-tures and mass, but by then the classificationalphabet was well established, requiring the or-dering OBAFGKML from the hottest most mas-sive star to the coolest objects, some of which maynot be massive enough to be true stars and sta-bly fuse H to He at their cores. These alphabeti-cal classifications also have fractional divisionsfrom 0 to 9, running from hottest to coolest withinthe class. Colors are also used as an abbreviationfor temperatures; as one might expect, blue starsare hot, and red stars are cool. Also, astronomersuse the terms “earlier” and “later” to refer to thespectral sequence from hotter to cooler. “Dwarf”(as opposed to supergiant, giant, and subgiant) isa luminosity classification and tells us that a starresides on the “main sequence” (a locus of pointsin a plot of stellar luminosity vs. temperature)where it will spend the majority of its lifetime ina stable configuration with nuclear fusion as itspower source. These same named luminosityclasses are also numbered for abbreviation; I, II,III, IV, and V runs from supergiant to dwarf. Ourown Sun is spectral type G2 and luminosity classV (or dwarf). Less massive dwarf stars are cool,such as M dwarfs. The length of time a particu-lar star remains on the main sequence and therate at which the end phases of its evolution causeit to expand, contract, change its surface gravity,and, therefore, appear within other luminosityand spectral classifications are determined by itsmass. Massive stars consume their nuclear fuelrapidly, while M dwarfs remain stable for a verylong time. In the history of the Milky Way Galaxy,no M dwarf star has yet had sufficient time toevolve away from the main sequence and end itslife.

    Since the human eye is a logarithmic sensor,astronomers have historically used a logarithmicor magnitude scale for recording and relatingstellar brightness. On this scale, a difference of 5in magnitude implies a factor of 100 difference inthe brightness, and the sense of the comparisonis such that a 6th magnitude star is 100 timesfainter than a 1st magnitude star. The zero pointof this scale is set by observations of A0 V stars,which are assigned a magnitude of 0. To distin-guish between the brightness of a star as it ap-pears on the sky, which depends on its distance,and the intrinsic luminosity of the star, there aretwo magnitude scales: apparent magnitudes,

  • which are designated as lowercase m, and ab-solute magnitudes, which are designated as up-percase M and defined as the apparent magni-tude that a star would have if it were at a distanceof 10 pc (!33 light years) from the observer. Ob-servations were historically made with the nakedeye, then eyes with telescopic assistance, thenrecorded on photographic plates, and nowrecorded with instrumentation spanning a muchlarger range of frequencies than can be perceivedby the human eye. Each of these sensors has dif-ferent sensitivities over different spectral ranges.Significant energy has been expended and con-tinues to be invested in calibrating different mea-surement results against one another. One ap-proach is to attach a letter designation tomagnitude measurements, where the letter indi-cates the spectral band and response of the re-ceiver, e.g., MV denotes absolute magnitude asmeasured in the visual or V band. The difficultycomes in practice when not all “V” bands are thesame, but that level of detail is beyond the scopeof this paper. Common spectroscopic bands are:U (UV), B (blue), V (visual), R (red), and I [in-frared (IR)], plus J, H, and K (other near-IR). Thedifference between the brightness of an object intwo spectral bands, e.g., MB-MV, is an astronom-ical color and designated (B-V). In contrast tothese wavelength-specific measurements, as-tronomers use the term bolometric to connote aproperty, such as luminosity, that includes con-tributions from all wavelengths.

    Finally, because of the ungainliness of astro-nomical quantities when expressed in cgs or otherscientific systems, astronomers tend to use rela-tive measures, e.g., the mass of an object ex-pressed in terms of the mass of the Sun (MSun "1.99 # 1033 g), the distance light travels in 1 year(l year " 9.46 # 1017 cm), or the pc, which is thedistance of an object whose measured angularparallax is 1 arc sec (1 pc " 3.08 # 1018 cm).

    Most of the stellar members of the solar neigh-borhood and, presumably, the Universe are red

    dwarfs (low mass, cool, main sequence stars).They make up at least 75% of all stars (excludingbrown dwarfs) and continue to be found at dis-tances less than 5 pc (Henry et al., 1997; Jao et al.,2005; Henry et al., 2006). These objects span a hugerange in brightness (9 $ MV $ 20), have massesbetween 0.5–0.6 and 0.08 MSun (Delfosse et al.,2000; Henry et al., 2004), and have been assignedspectral type M. Because of their overwhelmingnumbers, they contribute more to the total stellarmass budget of the Galaxy than any other spec-tral-type star, even at their relatively low masses.Their intrinsic faintness makes them elusive, ev-idenced by the fact that not a single one can beseen with the naked eye.

    Table 1 outlines the basic characteristics forstars that represent the broad range of M dwarfcharacteristics, with our Sun as a benchmark. Ab-solute brightnesses at optical (MV, 0.55 !m) andIR (MK, 2.2 !m) wavelengths are provided, aswell as a broad baseline color (V-K). Guidelinesfor effective surface temperatures and the massesand luminosities relative to our Sun are alsolisted. It is worth noting that, for the later M types,the mass inferred from color temperature can de-pend on age (younger stars are hotter, then cooloff).

    Given that a magnitude of 1 corresponds to afactor of !2.5 in brightness, one notes that an M0V star produces only 1.9% of the light in the vi-sual band as the Sun, but 16% of the light in theIR K band. Corresponding values for an M9 Vstar are 0.00014% at V and 0.13% at K. Thus, thebalance of radiation relative to our Sun is verydifferent for M dwarfs, which produce relativelymore IR radiation than visible.

    The range of different stellar environments islarger for dwarf stars defined as spectral type Mthan for any other stellar type. The full gamut ofstellar spectral types runs OBAFGKML. To matchthe range in visual brightness of the M dwarfs(magnitudes 9.0–19.4), more than four full spec-tral types of stars are required: A0.0 V stars have

    TARTER ET AL.34

    TABLE 1. BASIC CHARACTERISTICS OF STARS WITH M DWARF CHARACTERISTICS

    Spectral type MV MK V-K Temperature Mass Luminosity

    Sun (G2V) 4.8 3.3 1.5 5,800 1.00 100.%M0 V 9.0 5.3 3.8 3,900 0.50 6.%M3 V 11.7 6.8 4.9 3,600 0.29 3.%M6 V 16.6 9.4 7.2 3,000 0.10 0.5%M9 V 19.4 10.5 8.9 2,400 0.08 0.02%

  • HABITABLE PLANETS AROUND M DWARFS 35

    MV !0, whereas K7.0 V stars have MV !9. Whenconsidering stellar mass, which is arguably thesingle most important characteristic of a star inthat mass determines nearly every other charac-teristic, the factor of 6 in mass from M0.0 V toM9.0 V is again only matched by spanning typesAFGK.

    In summary, M dwarfs are low mass stars thatspan a substantial mass range (!0.1 to !0.5 so-lar masses) and exhibit a wide range of funda-mental properties. During an extended youth(ages less than !1 Gyr), they are significantlymore magnetically active at UV and X-ray wave-lengths than solar-type stars and exhibit power-ful flares whose radiation could inhibit the emer-gence of life. They dominate the local stellarpopulation by number and contribute approxi-mately half of the stellar mass in the Galaxy.

    2.2. The stellar properties of M stars

    As described above, M stars have relativelylow stellar temperatures, which makes their lightred. They also have smaller radii (the size scalesapproximately with the mass), so that a 0.1 solarmass object is roughly 0.1 solar radii as well. Thisaccounts for their low luminosities, though onecould also say that these low luminosities arecaused by the low central pressure and tempera-ture required to support the lower mass againstgravity, which leads to much smaller nuclear fu-sion rates. On the other hand, the surface pres-sures are actually higher than for solar-type starsdespite the low mass (because gravity is an in-verse square force). Because of the low outer tem-peratures and high pressures, the atmosphere cansupport molecules as well as atoms, and the op-tical and IR spectra are dominated by molecularbands. In the optical, these arise from heavy metaloxides like TiO and VO, not because those mole-cules are abundant but because they happen tohave high optical opacities. In the IR, the spec-trum is dominated by steam and carbon monox-ide spectral features.

    The high outer opacities and generally lowertemperatures throughout mean that the interioropacities are also higher, and the stars find it eas-ier to move luminosity outward by convectionrather than radiation throughout their bulk. Thisconvective mixing has the effect of making thewhole star available as nuclear fuel. Given thatthe Sun only burns 0.1 of its mass (because of itsradiative core), coupled with the fact that M stars

    have much lower luminosities, this makes theirhydrogen burning lifetimes (the main sequence)much longer. These can range from 50 Gyr for themost massive M stars to several trillion years forthe least massive ones. The mixing also meansthat the core composition has an extremely slowgain in mean molecular weight and a conse-quently very slow change in emitted flux. This isin contrast to stars like the Sun, in which hydro-gen is converted to helium in an isolated (non-convective) core, and the relatively faster increasein mean molecular weight results in a steadybrightening of solar-type stars even while on themain sequence. As a result, the HZ (discussednext) for an M star stays in place for much longeras well, and the continuously HZ is a much largerpercentage of it. Thus, if a planet exists in the HZof a typical M star, it may remain in that zone for100 Gyr. In essence, therefore, M stars last so longthat the length of habitability becomes more of aplanetary than a stellar issue. An important ex-ception to this could be the effect of magnetic ac-tivity on the star; we defer discussion of this un-til subsection 3.1.

    2.3. The HZ

    The HZ is a concept that is used extensively inthe context of searching for planets that might besuitable abodes for life. There are a number of dif-ferent definitions of the HZ (e.g., Dole, 1964;Heath et al., 1999), but here we define it as thatregion around a star in which a planet with anatmosphere can sustain liquid water on the sur-face. This is because liquid water is the basic re-quirement for life on Earth, so its presence wouldappear to be the most important criterion for in-cluding a given location in any search for life, ifindeed any selection of locations is desirable inthe first place. An additional reason for the choiceof this definition is that extrasolar planets onwhich liquid water and life are present at the sur-face should be observable spectroscopically in asearch for evidence of life (Leger et al., 1993; An-gel and Woolf, 1996; Tinetti et al., 2005), whereassubsurface biospheres may not be detectable.

    The nature of the HZ around F, G, and M starshas been dealt with elsewhere (Kasting et al., 1993;Heath et al., 1999), and the reader is referred tothose papers for detailed overviews of the sub-ject. We only summarize some of their resultshere. The outer and inner edges of the HZ are de-fined as those values of the stellar insolation (SI)

  • at which liquid water ceases to be stable at a plan-etary surface. In terms of actual size, the HZaround an M dwarf will be smaller in size thanthe HZ around a G star. The edges of the HZ arecloser to the star because of the lower luminosityof M dwarfs and the inverse square law decreaseof radiation with distance. The wide range of Mdwarf masses means that the width of the HZ willvary by an order of magnitude from 0.2 AU to0.02 AU. The seriousness of this constraint onhabitability, given the likelihood of a planet form-ing at a given distance from its star, is discussedin subsection 3.2.

    For the outer edge, Kasting et al. (1993) con-sidered that, on geological time scales, theamount of CO2 in the Earth’s atmosphere is con-trolled by processes such as weathering. The de-pendence of weathering rate on temperature ensures that, if a planet’s surface freezes, weath-ering ceases, and CO2 builds up in the atmo-sphere and thus warms the planet up again. Thisnegative feedback acts to extend the width of theHZ outward in terms of distance from the parentstar and lower values of SI. The limit to the neg-ative feedback, i.e., the outer edge of the HZ, oc-curs at that value of SI at which CO2 condensesin the atmosphere, rendering the weatheringfeedback irrelevant.

    The inner edge of the HZ can be defined as thatpoint at which a planet’s stratosphere (which on Earth lies between 15 and 40 km in altitude)changes in composition from being orders ofmagnitude drier than the surface (as it is on Earth)to being almost as wet. The rate of hydrogen pro-duction from water photolysis in the middleatmosphere increases accordingly, as does escapeof hydrogen to space. Over time, the planet’s sur-face dries (Kasting and Pollack, 1983). A numberof caveats on this mechanism, such as climatefeedbacks, are present and are reviewed by Kast-ing et al. (1993).

    An M dwarf presents specific issues for thehabitability of planets that lie in the HZ. Fordecades, the main problem was thought to betidal locking; a planet in the HZ of an M dwarfwould lie so close to its parent star that it wasvery likely to become tidally locked. In otherwords, the planet always presented the same faceto its star, much as Earth’s moon does to the Earth(Dole, 1964). It was assumed that a consequenceof locking was that atmospheric volatiles wouldfreeze out on the dark side on the planet and re-sult in an atmosphere where the surface pressure

    is controlled by a balance between the latent heatof condensation and radiative cooling on the darkside. Such an atmosphere would have a surfacepressure orders of magnitude below that of pre-sent-day Earth and far below the triple point ofwater—the minimum pressure at which watercan stably exist as a liquid. It would therefore beconsidered not habitable.

    The above view was first challenged byHaberle et al. (1996), who used an energy balancemodel that parameterized atmospheric heattransport to show that an atmosphere containingonly 100 mbars of CO2 transported enough heatto the dark side to prevent freezing of volatiles.The result of Haberle et al. (1996) was given sig-nificantly more weight by Joshi et al. (1997), whoexplored the sensitivity of surface temperaturegradient to factors such as surface pressure andatmospheric IR optical depth using a three-di-mensional global circulation model that actuallysimulated the climate and weather of the atmo-sphere. They, like Haberle et al. (1996), found thatapproximately 100 mbars of CO2 was enough to prevent atmospheric collapse. Joshi (2003)showed that inclusion of the hydrological cycle(including advection of water vapor) on a tidallylocked Earth did not destabilize the climate, andachieved qualitatively similar results to Joshi etal. (1997).

    Note that the above studies are “worst-case”scenarios. For instance, not all close-in planets be-come tidally locked; another possibility is thatthey have a spin/orbit resonance like Mercury,which rotates three times every two orbits aroundthe Sun. If a planet is not tidally locked or the ef-fect of heat transport by oceans is included, thetemperature difference between day and nightwill be significantly lower than the model pre-dictions. It is perhaps counterintuitive to thinkthat atmospheric motions can dramaticallychange the surface environment of a planet. Nev-ertheless, the models do show that while tidallocking might alter the expected edge of the HZ(see below), it can no longer be considered a se-rious barrier to habitability.

    For a given SI, the substellar point (SP), the lo-cation directly beneath the star, is warmer than ifthe stellar radiation was distributed evenly in lon-gitude (Joshi, 2003). Thus, one might expect theinner edge of the HZ (as defined by a wet strato-sphere) to happen at a lower value of SI than isthe case for a G star. On the other hand, climatemodel studies show that the planetary albedo of

    TARTER ET AL.36

  • HABITABLE PLANETS AROUND M DWARFS 37

    an ocean-covered synchronously rotating Earth is0.35, which is 20% higher than if tidal locking didnot happen (Joshi, 2003). A higher albedo wouldmove the inner edge of the HZ nearer its parentstar. The rise is due to the presence of ice andclouds, and works in the opposite direction to therelatively low planetary albedo expected of aplanet orbiting an M dwarf (Kasting et al., 1993).

    The distribution over wavelength of the UV ra-diation from M dwarfs could lead to a differentphotochemistry in their planets, moving the outeredge of the HZ away from the star, because gasessuch as CH4 or N2O might exist in much higherconcentrations than on Earth (Segura et al., 2005).The CO2 condensation limit would no longer ap-ply, as even without any CO2, significant green-house gases would exist in the atmosphere. Thiseffect would make the HZ wider.

    Kasting et al. (1993) used the concept of a con-tinuous HZ to take account of changes in SI overbillions of years. Since M dwarfs are very longlived, the continuous HZ might exist for an orderof magnitude longer in time around an M dwarfthan around a G dwarf. The time that life wouldhave to take hold on a planet orbiting an M dwarfwould, therefore, be much longer, and life might,therefore, be more likely on such a planet. Suchconsiderations are, of course, based on thepremise that a planet orbiting an M dwarf couldkeep a substantial atmosphere despite the vari-ous loss processes that would inevitably occur.This is discussed in subsection 3.3 below.

    While the amplitude of long-term variability ofSI on an M dwarf is small, short-term variabilitymight present a problem. For instance, stellar ra-diation can oscillate by several percent on timescales of days because of rotationally modulatedstarspots (Rodono, 1986). However, Joshi et al.(1997) showed that an atmosphere that has a sur-face pressure of 1 bar would not freeze out evenif subjected to a starspot that reduced SI by 40%and lasted a month. In addition, a tidally lockedplanet would be prevented from undergoinglarge swings in obliquity that might destabilizeclimate. A tidally locked planet in the HZ of anM dwarf would have an orbital period of days toweeks (Heath et al., 1999), and so the climatic con-sequences of an eccentric orbit would be heavilydamped by a planet’s atmosphere and oceans. Wefurther discuss the effects of stellar variabilityand, indeed, orbital variations in subsection 4.2.6.The effects of flares, which might present an ob-stacle to habitability (Kasting et al., 1993; Heath et

    al., 1999), are also considered later in subsection3.4. The potential effects of coronal mass ejectionsare discussed in papers by Khodachenko et al.(2007) and Lammer et al. (2007) in this issue.

    To summarize, the limits of the HZ around anM star might be expected to be defined by simi-lar processes to those that define the limitsaround a G star. However, the atmospheres ofplanets in the HZ of M stars would certainly bevery unfamiliar to us in terms of their circulation,radiation, and chemistry. However, if planets cankeep their atmospheres and water inventories,none of these differences presents a large obsta-cle to their potential habitability. Therefore, fromthe point of view of atmospheric and climate sci-ence, planets in the HZ of M dwarfs should havealmost as high a chance of being habitable asplanets in the HZ of G stars.

    2.4. Planets within the HZ

    The formation of planets around M stars re-quires that the protostar be surrounded by a pro-tostellar disk with sufficient material. Primordial(optically thick) circumstellar disks are readilydetected by their strong thermal IR emission, farabove that produced by stellar photospheres.Both ground-based and space-based IR surveyshave identified many disks around low massstars and brown dwarfs in the nearest young star-forming regions (e.g., Muench et al., 2001;Jayawardhana et al., 2003; Liu et al., 2004; Luh-man et al., 2005), with about half of the low masspopulation (spectral type of M0–M9) havingdisks. In at least some cases, the masses of theseprimordial disks appear to be quite substantial,about a few percent of the stellar host mass (e.g.,Andrews and Williams, 2005). Thus, it appearsthat M stars are as likely a venue for planet for-mation as more massive stars.

    Less theoretical work has been done on planetformation around low mass stars because of theemphasis on explaining the origin of our ownplanetary system. Our terrestrial planets are uni-versally thought to have formed through the col-lisional accumulation of successively larger solidbodies—starting with submicron-sized dustgrains, through kilometer-sized planetesimals,Moon-sized planetary embryos, and culminatingafter about 30 Myr in the formation of the Earthand the other terrestrial planets (Wetherill, 1990).Wetherill (1996) extended his Monte Carlo mod-els of collisional accumulation to include the case

  • of lower mass stars. He found that Earth-likeplanets were just as likely to form from the colli-sional accumulation of solids around M dwarfswith half the mass of the Sun as they were to formaround solar-mass stars. More recent modelingby Ida and Lin (2005) investigated the effects ofvarying stellar (and presumably protoplanetarydisk) metallicities and mass. The results are con-sistent with the detection statistics on exoplanetsand predict a peak in formation of Neptune-massplanets in short orbits around M stars. Simula-tions by Montgomery and Laughlin (2006) sug-gested that terrestrial mass planets form consis-tently in the HZ of late-type M stars, whereasRaymond et al. (2006) found that it may only bethe early-type M stars that form terrestrial plan-ets within the HZ having sufficient mass to retainan atmosphere and a favorable water content.

    Because the collisional accumulation process isinherently stochastic, it is not possible to makedefinitive predictions regarding the outcome of agiven set of initial conditions for the protoplane-tary disk, but basic trends can be discerned, ofwhich perhaps the most important is the effect ofJupiter-mass planets on terrestrial planet forma-tion (Wetherill, 1996). If a Jupiter-mass planetforms prior to the final phases of terrestrial planetformation, as seems to be necessary to explain thegaseous bulk composition of a Jupiter-like planet,then the location of such a massive gravitationalperturber can have a controlling influence on theformation of Earth-mass planets in the HZ of so-lar-mass stars; a Jupiter closer than about 4 AUto its star could prevent the growth of Earth-massplanets at 1 AU. In the early stages of planetaryevolution, it may well be necessary for life to havea long-period Jupiter-sized planet to scatter wa-ter-bearing planetary embryos toward innerplanets as was the likely case for the early Earth(Morbidelli et al., 2000). In later stages, a Jupiter-like planet could intercept and reduce thecometary flux onto the inner planets that mightotherwise frustrate the evolution of advanced lifethough periodic catastrophic impacts (Wetherill,1994). N-body simulations with many more par-ticles by Raymond et al. (2007) suggested that thedelivery of water to terrestrial planets near theHZ is statistically more robust than previouslythought, particularly if the water-bearing em-bryos form slowly. A long-period Jupiter mightthen be a prerequisite for a habitable Earth-likeplanet.

    Interstellar dust is the main contributor to the

    formation of accretionary disks and, thence, to theformation of Earth-like planets. The interstellardust consists largely of silicate minerals, mostlyolivine and pyroxenes (Draine and Lee, 1984;Adamson et al., 1990; Greenberg and Li, 1996), orof amorphous condensates in the multicompo-nent system MgO-FeO-SiO2 (Rietmeijer et al.,1999; Rietmeijer et al., 2002a,b). Judging from therelative intensities of the silicate and “organic”bands in the IR spectra, a surprisingly large frac-tion, about 10%, of the mineral dust grains in thediffuse interstellar medium across our galaxy andeven in neighboring galaxies consists of organics(Sandford et al., 1991; Pendleton et al., 1994). Mostof these organics are expected to survive intactthe accretionary processes leading to planetesi-mals and comet-sized bodies. A smaller but stillsignificant fraction may survive collisions be-tween planetesimals and even cometary impactson larger planets such as the Earth (Chyba, 1989).In this way, organics that formed in the outflowof distant dying stars (Dey et al., 1997; Matsuuraet al., 2004) and persisted for hundreds of millionsof years in the interstellar medium may have con-tributed to the pool of complex organic moleculeson the early Earth from which life arose.

    The organics associated with the interstellardust have the spectroscopic signature of fully sat-urated (aliphatic) hydrocarbons, in which Catoms are linked to other C atoms by single bondsand each C atom is bonded to either two or threeH atoms. They are essentially straight orbranched Cn chains, where n can be a large num-ber, such as in carboxylic (fatty) acids or parafinicsections linking more sturdy polyaromatic hy-drocarbons (Sandford et al., 1991; Pendleton et al.,1994; Pendleton and Allamandola, 2002). Howthe delicate saturated hydrocarbons survive theintense UV radiation field that permeates the dif-fuse interstellar medium and the incessant bom-bardment by high-energy particles has long beena question of great interest and discussion.

    Various models have been proposed, of whichthe “core–mantle” concept (Greenberg, 1968) hasreceived the widest attention (Sandford et al.,1991; Allamandola et al., 1999). It is based on theassumption that dust grains that form in the out-flow of stars acquire a layer of ice composed notonly of H2O but also other gaseous componentssuch as CO, NH3, etc. Through UV photolysis inthe ice matrix, organic molecules would formthat, upon sublimation of the ice, will remain onthe grain surface as a thin veneer (Greenberg et

    TARTER ET AL.38

  • HABITABLE PLANETS AROUND M DWARFS 39

    al., 1995). Such a thin veneer of organic matter onthe surface of grains, however, is not expected tosurvive for long the intense radiation field andconstant sputtering in the diffuse interstellarmedium. Another model (Mathis et al., 1977) pos-tulates separate grain population of silicate min-erals and hydrogenated amorphous carbon to ac-count for the observed IR spectral features andthe dielectric properties of the dusty interstellarspace (Draine and Lee, 1984; Draine, 2003; Zubkoet al., 2004). It is in disagreement with the obser-vational fact that the silicate-to-organics ratiothroughout our galaxy and even other galaxies issurprisingly constant, at about 10:1, in spite ofvery different conditions of dust formation in dif-ferent astrophysical environments.

    Recently, a dust grain model was presented,based on the laboratory observation that C-Hbonds can form inside the mineral matrix (Fre-und and Freund, 2006). The model treats the or-ganics associated with the dust as Cn–Hm entitiesembedded into the mineral matrix of the dustgrains, the combination forming a single phase,solid solution. This model appears to be mostconsistent with the astronomical observations(Sandford et al., 1991; Pendleton et al., 1994). It ex-plains both the complexity and hardiness of theorganics in interstellar space and their surviv-ability through time and space.

    While a general consensus exists regarding thebasic mechanism for terrestrial planet formation,the situation is quite different for the formationof giant planets. The conventional wisdom is thatgas giant planets form by the process of core ac-cretion, where a roughly 10 Earth-mass solid coreforms first by collisional accumulation, and thenaccretes disk gas to form a massive gaseous en-velope (Pollack et al., 1996). However, even whenthe surface density of solids is high enough forrunaway accretion (Lissauer, 1987) to assemble aseed mass for the solid core in about 0.5 Myr, thetime scale for subsequent growth and accretionof the gaseous envelope can be sufficiently long(typically a few to 10 Myr) to raise the danger thatthe disk gas will have dissipated well before thesolid cores can accrete enough gas to grow toJupiter-mass. Instead, “failed cores” similar toUranus and Neptune might result. An alternativemechanism has been proposed and investigated,where gas giant planets form rapidly through agravitational instability of the gaseous portion ofthe disk (Boss, 1997, 2001). Disk instability can oc-cur within thousands of years, well before the

    disk dissipates, and if this mechanism occurs, itwould presumably outrace the core accretionmechanism.

    Boss (1995) studied the thermodynamics ofprotoplanetary disks around stars with massesfrom 0.1 to 1.0 solar mass and found that the lo-cation of the ice condensation point only movedinward by a few AU at most when the stellar masswas decreased to that of late M dwarfs. In the coreaccretion model of giant planet formation, thisimplies that gas giant planets should be able toform equally well around M dwarf stars and, per-haps, at somewhat smaller orbital distances.However, the longer orbital periods at a givendistance from a lower mass star mean that coreaccretion will generally be too slow to produceJupiter-mass planets in orbit around M dwarfsbefore the disk gas disappears (Laughlin et al.,2004). In the competing disk instability model forgas giant planet formation, however, calculationsfor M dwarf protostars (Boss, 2006a) show thatJupiter-mass clumps can form in less than 1,000years in marginally gravitationally unstable diskswith masses of 0.02–0.07 solar masses orbitingaround stars with masses of 0.5 and 0.1 solarmasses. These ongoing models suggest that, ifdisk instability occurs, there is no reason why Mdwarf stars might not rapidly form gas giant pro-toplanets. Hence, it may well be that mini-solarsystems, similar to our own except reduced inspatial scale, are able to form in orbit about Mdwarf stars. Zhou et al. (2005) have recently pro-posed an observational test to discriminate be-tween these two models of planet formation. Nu-merous short-period terrestrial planets can beexpected to be found in systems with jovian massplanets as the result of sequential accretion, butnot as the result of gravitational instability. How-ever, models of terrestrial planet formation byKortenkamp et al. (2001) have shown that rapidformation of gas giant planets by disk instabilitycan actually facilitate the growth of inner rockyplanets, contrary to the assumptions made byZhou et al. (2005). Ground-based spectroscopicsurveys have begun to detect planets in orbitaround M dwarfs, with masses ranging fromabout 7.5 Earth-masses to Jupiter-mass and above(Rivera et al., 2005). Boss (2006b) has shown thatdisk instability in the outer regions coupled withcollisional accumulation in the inner regions isable to explain the super-Earths and gas giantsfound in orbit around M dwarfs by both mi-crolensing and spectroscopic planet searches.

  • 2.5. Atmospheres of habitable planets

    Since the circulations of terrestrial planetaryatmospheres are primarily driven by the spatialdistribution of short-wave radiation, one can ex-pect that the atmospheres of planets orbiting Mdwarfs will be significantly different in characterto the terrestrial paradigm, in terms of both cir-culation and composition. Again, for a fuller de-scription of the atmospheric circulation on tidallylocked planets, the reader is referred to Joshi etal. (1997), who used a simple two-stream gray ra-diation scheme to model IR emission and as-sumed that all solar radiation hit the surface, andJoshi (2003), who explicitly modeled the effects ofthe hydrological cycle and had higher horizontaland vertical resolution. The effect of a water cy-cle was to have more uplift on the dayside anddownwelling on the nightside than in the drymodel, which resulted in different flow at lowlevels across the terminator. The strengths of theeast–west, or zonal, jet streams in the upper lay-ers of each model were also different as a resultof the different configurations.

    It appears to be quite commonly thought that,even if atmospheric collapse is prevented ontidally locked planets, the large thermal gradientsand associated high winds across the terminatorinhibit habitability, especially forest habitability.However, atmospheric heat transport tends to re-duce thermal gradients, both by reducing theiramplitude and by smearing them out to largerhorizontal scales. Figure 8 of Joshi et al. (1997)shows the so-called radiative-convective equilib-rium temperature field (obtained by running theatmospheric model without the effect of large-scale horizontal motions) and the actual time-av-eraged temperature field. The former case haslarge gradients on the terminator, which are com-pletely smeared out in the latter case. Thermalgradients and any potential high wind speeds as-sociated with them should, therefore, not be con-sidered a barrier to habitability.

    The terminator is thought to present the mostlikely location for life due to the relatively smalleffect of flares at that location (potentially dam-aging radiation travels through a longer atmos-pheric pathlength suffering greater degradation,and the absolute flux is diminished by the slantangle). However, the terminator would also re-ceive far less sunlight than the SP, the locationwhere the parent star is directly overhead, and soforest habitability might be impeded here (Heath

    et al., 1999). The presence of thick clouds associ-ated with convection would have an effect on thisconclusion, by changing the amount of direct anddiffuse solar radiation at the SP. Plants at this lo-cation, for instance, might be less susceptible toflares.

    The remotely observed IR properties of atidally locked planet will be greatly affected bythe presence of an atmosphere with an active hy-drological cycle. On a dry world or a planetwhose atmosphere has collapsed, IR emissionmostly comes from the surface, so that the darkside will appear far colder than the dayside whenviewed at these frequencies. However, with a hy-drological cycle comes the presence of water va-por and clouds—the SP will have clouds presentfor most of the time—and so IR emission here willcome mostly from the cold cloud tops. Indeed,the dayside of such a planet might actually ap-pear as cold in the IR as the darkside, in the sameway as the Western Pacific region, which has thehottest sea surface temperature, appears verycold when seen in the IR. Such observationsmight actually provide evidence for a planet withan active hydrological cycle. Significant effortshave been expended on predicting and inter-preting the spectra to be observed by the Terres-trial Planet Finder Coronagraph and TerrestrialPlanet Finder Inferometer missions, which seekto directly image terrestrial planets in the HZs ofnearby stars. The moisture content and averageplanetary albedo are critical to these models, asare considerations of phase and smearing over or-bital cycles (Segura et al., 2003). A great deal moreresearch needs to be done on the topic.

    Photolysis processes on a planetary atmo-sphere are controlled by the UV radiation of itsparent star. Because of the peculiar flux distribu-tion over wavelength of the active M dwarfs andthe low radiation in the UV region of the spec-trum from quiescent M stars, the atmosphericchemistry of trace species may be greatly affected.Simulations of planets that lie in the HZ of an Mstar and have the same atmospheric compositionand input of biogenic compounds as present-dayEarth showed that the atmospheric lifetimes ofcompounds like methane, nitrous oxide, andmethyl chloride are larger than on Earth (Seguraet al., 2005). As a result, the abundance of thesecompounds could be potentially three orders ofmagnitude more than their terrestrial abundance(Table 2). Planets around quiescent M stars (re-ferred as “model” in Table 2) show a “methane

    TARTER ET AL.40

  • HABITABLE PLANETS AROUND M DWARFS 41

    runway,” i.e., methane builds up in the atmo-sphere because there is not enough UV radiationto destroy this compound in the stratosphere, andthe main sink of methane, OH, is highly reducedin the troposphere (see discussion in Segura et al.,2005). To avoid this problem, a methane mixingratio of 500 ppm is set for the simulated planetsaround the quiescent M stars. The methane fluxneeded to maintain this mixing ratio is 21% of thepresent methane production on Earth for a planetaround the coolest star considered in Segura et al.(2005). Since methane is a greenhouse gas, this result points to the possibility of extending theouter limit of the HZ for planets that producemethane either biologically or not.

    For life detection purposes, these biogenicgases may be more detectable on planets aroundquiescent or active M dwarfs than on Earth (Figs.5–8 in Segura et al., 2005). Finally, a planet withthe present concentration of O2 in the HZ of anactive M dwarf may be able to develop an ozonelayer as large as the terrestrial one, which willprotect the planetary surface from UV radiation(Tables 2 and 3 in Segura et al., 2005).

    Another aspect of synchronously rotatingplanet habitability that needs attention in futurework is how the carbonate–silicate cycle wouldcontrol the partial pressure of the atmosphericgreenhouse gas CO2 (pCO2) and, thereby, climateon time scales of %0.5 Myr. Changes in out-gassing rates from the interior of the planet andin continental weathering regimes due to changesin the distribution of continents across climaticzones would be expected to be associated withadjustments in pCO2. As mentioned earlier, Kast-ing et al. (1993) used the classic work by Walkeret al. (1981), which demonstrated how the car-

    bonate–silicate cycle could act as a natural ther-mostat to dampen climatic change and establishthe limits of the HZ. The greenhouse gas CO2 en-ters the atmosphere through outgassing from theEarth’s interior and is removed from the atmo-sphere as it dissolves in rainwater to form dilutecarbonic acid. In the event that insolation in-creased, the effect on weathering processeswould be to adjust pCO2 downward. This wouldreduce the atmospheric greenhouse effect and socompensate for increased insolation. Reduced in-solation would result in an increased pCO2 so thatan accentuated greenhouse effect would com-pensate and prevent a geologically active planetfrom freezing over.

    In the early climatic models for synchronouslyrotating planets, rather high pCO2 values of up to1,000 mbars and 1,500 mbars were assumed, butit is now clear that extreme values are not needed.The model of Joshi et al. (1997) actually employeda gray radiation scheme whose optical depth "could be specified independently of the surfacepressure. The only aspect of their model atmo-sphere that was actually “CO2” was the gas con-stant. A value of " of 1.0 was chosen for their control scenario because this happened to ap-proximately match the gray optical depth of thepresent terrestrial atmosphere with !350 ppmCO2 (" !0.9), as well as an atmosphere having1,000 mbars pure CO2 (" !1). It is, therefore, notnecessary to invoke high pCO2 to prevent freeze-out on synchronously rotating planets—a con-clusion that was repeated by Joshi (2003), whoused present-day atmospheric composition.

    These results meant that it was not necessaryto limit discussions of habitable synchronouslyrotating planets to planets on which pCO2 was

    TABLE 2. MIXING RATIOS OF BIOGENIC GASES CALCULATED BY A ONE-DIMENSIONAL PHOTOCHEMICALMODEL FOR EARTH-LIKE PLANETS AROUND DIFFERENT STARS (SEGURA ET AL., 2005)

    Parent star Methane Nitrous oxide Methyl chloride

    Sun 1.6 # 10&6 3.0 # 10&7 5.0 # 10&10AD Leo (M4.5V)a 4.6 # 10&4 1.3 # 10&6 1.4 # 10&6GJ 643 (M3.5V)a 3.8 # 10&4 1.1 # 10&6 8.8 # 10&7Modelb

    M1 (Teff " 3,650 K) 5.0 # 10&4 3.5 # 10&7 1.2 # 10&7M3 (Teff " 3,400 K) 5.0 # 10&4 1.0 # 10&4 2.0 # 10&7M5 (Teff " 3,100 K) 5.0 # 10&4 1.3 # 10&3 5.4 # 10&7

    aFor these planets the surface flux of these compounds was consider to be the same as present Earth: 9.54 # 1014g of CH4/year, 1.32 # 1013 g of N2O/year, and 7.29 # 1012 g of CH3Cl/year.

    bOn these planets the boundary conditions were: fixed mixing ratio for CH4 (5.0 # 10&4), fixed deposition veloci-ties for H2 (2.4 # 10&4 cm/s) and CO2 (1.2 # 10&4 cm/s), and fixed surface fluxes for N2O and CH3Cl.

  • constrained to extreme values, nor was it neces-sary to invoke and explain special circumstancesin which this was possible with Earth-level inso-lation. Once again, then, the discussion of foresthabitability was found to follow on naturallyfrom the climatic models without unduly con-trived conditions.

    There has been no investigation of the possiblelong-term evolution of climate in relation to thecarbonate–silicate cycle on synchronously rotat-ing planets; values of pCO2 for different conti-nental distributions have not been estimated. Thisis a major area for future work, the results ofwhich will be of fundamental importance to tak-ing discussion of synchronously rotating planethabitability forward.

    The geophysical regime of a synchronously ro-tating planet will play an important part in de-termining whether it is habitable. On a planetwith plate tectonics, the drift of plates throughthe colder zones near the terminator would ex-tinguish Earth-type arboreal forms, while driftonto the dark side would spell extinction for awide range of organisms. Alternatively, one-plateplanets such as Mars and Venus with histories oflong-term geologic activity (e.g., Basilevsky et al.,1997; Hauck et al., 1998; Head et al., 2001) mayalso have regions of persistent habitability. Inparticular, volcanically active localities can act asa source of juvenile or recycled volatiles and gen-erate surface heat fluxes commensurate withplate tectonics settings.

    It was pointed out in Heath et al. (1999) that,even if a layer of ice were to form on the darkside of a synchronously rotating planet, Earth-like levels of geothermal heat from the interior ofa geologically active planet should ensure thatoceans would not freeze to their floors. Bada etal. (1994) had advanced this mechanism todemonstrate that liquid water environmentswould have been possible on an early Earth thatreceived reduced insolation from the early Sun,even in the absence of a massive atmosphericgreenhouse effect. What it means for a synchro-nously rotating planet is that a supply of snow tothe top of a thick layer of sea ice should be bal-anced by melting at its base, and so, if deep oceanbasins communicate between the dark and lithemispheres, a vigorous hydrological cycle ispossible. This mechanism demonstrates one wayin which the hydrological cycle and geological ac-tivity are linked, and future research should in-vestigate how the volume of liquid water beneath

    the ice responds as a planet’s internal heat fluxdiminishes with time and the thickness of theoceanic crust declines.

    However, it should be noted that the connec-tion between climate and geological activity is notone-directional. Rather, geologic activity, espe-cially through volcanism, may be necessary toprovide appropriate greenhouse gasses to theatmosphere. In the case of some planets, such asVenus, there may be a closed feedback loopwhere increasing or decreasing atmospheric tem-peratures temper the amount of geologic activity(e.g., Phillips et al., 2001). Water also strongly af-fects the melting and deformational behavior ofthe silicate minerals that make up the crust andmantle of terrestrial planets (e.g., Mackwell et al.,1998), and hence affects the type and vigor of ge-ological activity such as has been suggested forVenus (Kaula, 1995). However, water alone likelydoes not explain even the diversity of tectonicregimes within our own Solar System. For exam-ple, like Venus, Mars is a one-plate planet; how-ever, it clearly has had water throughout its his-tory. Geodynamical regimes of geologicallyactive Earth-sized planets subject to substantialtidal torque remain to be investigated.

    2.6. Requirements for life

    The envelope of physical conditions for life onEarth is in many ways remarkably wide. Envi-ronmental pH can range from $1 to 12, pressurefrom $1 to %500 bars, salinity from zero (fresh-water) to saturation in saline lakes, and temper-ature from !380K (%100°C) to below 233K(&40°C), the absolute limit for undercooled wa-ter at a pressure of 1 bar. The discovery of so-called extremophiles that have expanded theboundaries of habitable real estate has dramati-cally changed perceptions about the possibility oflife beyond Earth; a modern review of currentlyunderstood limits on habitation can be found inRothschild and Mancinelli (2001). Life as weknow it does, however, depend on the presenceof liquid water, and the existence of the liquidstate ultimately sets the boundary conditions forlife on Earth. Not only does water provide thesolvent for biochemistry, but there are very fewphysiological processes in which water does notplay a part as reactant or product.

    Life on Earth depends on solution chemistry inwater, a solvent that possesses unique physicaland chemical characteristics, a set of chemicals

    TARTER ET AL.42

  • HABITABLE PLANETS AROUND M DWARFS 43

    primarily consisting of a few simple and commonatoms (C, H, N, O, P, and S), and relatively smalldifferences of free energy (Ball, 2004; Benner etal., 2004). Within organisms, energy tends to betransduced using electrical gradients acrossmembranes or by the oxidation of organic mole-cules to generate energy-rich molecules (typicallycontaining phosphate) and organic electron paircarriers (often involving sulfur) that are used todrive the synthesis of essential biomolecules.Such energy transduction processes frequentlyinvolve the use of transition metals, which makestheir presence in trace amounts necessary.

    It is often assumed that life elsewhere, if it ex-ists, will be based on liquid water and carbon.Liquid water has a complex and intimate in-volvement in all life processes on Earth, and car-bon’s four valence electrons endow it with extra-ordinary versatility to create multiple chemicalbonds. Further, water and carbon are common inthe universe.

    Alternative physiologies may exist, however,that could widen the range of physical conditionsover which life could exist. Extrapolation fromthe one example of life known to us, however,would suggest that the two solvents most likelyto act as a basis for life are water and ammonia.In the absence of evidence, however, it seemspragmatic to confine our search to life based uponwater and carbon (Ball, 2004; Benner et al., 2004).

    Given the basic requirements outlined above,we are therefore looking for a rocky planet, withsubsurface or surface liquid water (the latter re-quiring an atmosphere), that is geologically ac-tive to provide a continual recycling of elementsand capable of supporting liquid water for a suf-ficient period of time for life to arise and evolve.The question of life on planets around M dwarfstars thus centers on whether these planets couldexist within the continuously HZ of their hoststar, and how these conditions might be alteredby the effects of tidal locking on the climate andthe effects of an M dwarf’s radiation on life.

    Since we are also interested in the ways inwhich M dwarf planets might contribute to theevolution of technological civilizations (of inter-est to SETI searches), it is worthwhile to describethe requirements of ecosystems with multicellu-lar organisms, such as plants, that have beentermed “forest habitability” (Heath, 1996). OnEarth, forests dominate biomass at thecrust–atmosphere interface. It is estimated that,in the absence of human interference, forests

    would cover around 40% of our planet’s land areaand account for 60% of net primary productivityin land environments. This massive photosyn-thetically produced biomass implies that photo-synthesis is a highly favorable solution to theproblem of harvesting and exploiting natural en-ergy sources, and not merely an accident of his-tory. Further support for this is the fact that al-ternative forms of autotrophy exist among theprokaryotes, and yet none has dominated any butsmall, specialized niches. Forests also play an im-portant role in biogeochemical cycles, the hydro-logical cycle, and the modification of climate (no-tably through control of transpiration and bychanging ground albedo). In addition, they haveprovided an important environment in which ourown primate ancestors developed key adapta-tions concerned with hand–eye–brain coopera-tion that was important in the evolution of tool-making.

    The temperature tolerance of higher plants onEarth is narrower than the &40°C to %100°Crange of microorganisms. For example, the scle-rophyll trees and shrubs include forms that suf-fer cold injury at &5°C to &2°C and heat injury at50–60°C. The cold and hot limits of CO2 uptakeare &5 to 0°C and 45–50°C, respectively (Larcher,1995). The temperature ranges at which CO2 up-take is 50% that at the temperature optimum are15–20°C to 40–45°C. Also, in the Earth’s seasonalregime, full-sized trees are found where meandaily temperatures exceed 10°C for a minimumof 1 month a year. On synchronously rotating Mdwarf planets, therefore, where there are no sea-sons, we might take the 10°C isotherm as the coldlimit for trees. Although another planet wouldhave its own unique biosphere, these Earth-bi-ased considerations suggest that the temperaturerange of M star “forest habitability” would prob-ably be near 10–50°C. Furthermore, since theJoshi et al. (1997) model predicts typical windspeeds near the surface to be 5–10 m/s (even atthe terminator), no special biomechanics wouldbe required for tree growth.

    In summary, if a rocky planet orbiting withinthe HZ of an M dwarf star possesses liquid wa-ter and the chemical constituents necessary forlife, and if these conditions persist for a sufficienttime to allow life to originate (a time scale that isnot yet constrained) and evolve (again this is rel-atively unconstrained, but the bias is certainlytoward billions of years), then there is no reasonto conclude that the planet is not habitable.

  • 3. A DIFFERENTIAL ANALYSISBETWEEN G DWARFS AND M DWARFS

    3.1. Magnetic activity and stellar evolution

    M dwarf variability was first detected and re-ported by Ejnar Hertzsprung in 1924 and Adrianvan Maanen in 1939 and 1945. However, the re-alization that these changes in brightness weredue to short time scale flares only came in 1947,when Edwin Carpenter noted a 3 magnitudebrightening of UV Ceti (M5.5e, 2.7 pc) in 12 min.By 1970, 50 flare stars were known, and it wasrecognized that they were nearly all mid- to late-type M stars lying within 15 pc of the Sun(Gurzadyan, 1970). By the 1990s, the number haddoubled. A much greater number of flare starswas found in young open clusters; the Pleiadesalone has over 500 observed with another 500 es-timated to be present (Ambartsumian et al., 1970).A considerable phenomenology of M star flareshas been garnered in the optical and UV bands(Houdebine, 2003, and references therein;Walkowicz et al., 2004; West et al., 2004). The Uband flux is correlated with Balmer line emissionfor M dwarf stars whose levels of activity span 5orders of magnitude. Chromospheric surfacefluxes in hydrogen Balmer lines can reach 108erg/s/cm2, as seen in the young system AU Mic.Flare intensity and duration are correlated from1- to 100-min decay phases. The chromosphericcooling budget is often dominated by UV con-tinuum, rather than emission line. Stars earlierthan M7 maintain a persistent quiescent chro-mosphere and transition region between flares asindicated by H# activity, with later-type stars re-maining active longer (Hawley and Johns-Krull,2003; Silvestri et al., 2005). Significant fractions ofthe stellar photosphere can be covered with long-lived, cool, strongly magnetized starspots.

    The activity of M stars follows a rotation–ac-tivity relation, as seen in hotter solar-type stars,but only when the rotational velocities are below4 (10) km/s for mid- (late-) M dwarfs (Mohantyand Basri, 2003). For the more rapidly rotatingstars, the dynamos or surfaces of most M starsappear to be fully saturated. The fact that M starsare convective, conductive, and spin means thatthey can generate magnetic fields by dynamo ac-tion. When they are fully convective (those M3and later), the cyclical solar-type magnetic dy-namos (which operate at the radiative–convectiveinterface) cannot work, and the magnetic fields

    must come from partially or fully turbulent dy-namos. This likely means that the typical fieldstructures tend to be smaller and the sensitivityto rotation less, which may account for the gen-erally longer spindown time for M stars (perhapsthrough the field geometry). There is no evidenttransition in activity properties around M3–M4when the radiative core disappears and the inte-rior becomes fully convective. This is not under-stood and suggests that a saturated convectivedynamo either dominates the field generation oralters the interior structure of the cooler stars(Mullan and MacDonald, 2001).

    We do know that M stars, when they areyoung, can generate fields stronger than solar be-cause of the higher gas pressure, and that theycan cover nearly the whole star (Saar and Linsky,1985; Johns-Krull and Valenti, 1996; Valenti andJohns-Krull, 2001). These lead to flares that some-times reach stellar dimensions (Osten et al., 2005;Güdel et al., 2004) and create bursts of luminos-ity that are a greater fraction of the total stellarluminosity than the strongest solar flares. Therange of flare intensities is very wide (up to 105or more), with young flare stars at the maximumvalues. Flare luminosities decrease with stellarmass and maintain roughly a maximum ratio of10&4 with the bolometric luminosity (Pettersen,1991). Of course, similar energies are also emit-ted at shorter wavelengths, including X-rays. Anumber of recent X-ray studies with simultane-ous multi-wavelength coverage provide a close-up view of individual flares. On Proxima Cen-tauri, low-level X-ray emission formerly labeled“quiescent” is now clearly shown to be the su-perposition of many small flares, and modelinga powerful X-ray flare implies a loop size com-parable to the stellar radius (Güdel et al., 2004).A continuous study of EV Lac (M4.5e, 5.1 pc) over2 days revealed a powerful radio flare accompa-nied by an impulsive U-band event. It also re-vealed numerous optical white light and X-rayflares (Osten et al., 2005). For EV Lac, and in otherM stars, flare emissions in the different wave-bands are largely decoupled from each other(Smith et al., 2005).

    The activity in electromagnetic radiation is un-doubtedly accompanied by high-energy particlefluxes as well. Since the X-rays in flare stars areharder (have relatively more high energy pho-tons) than those of solar flares, it is reasonable toassume that the particles may be more energeticas well. They also sometimes expel substantial

    TARTER ET AL.44

  • HABITABLE PLANETS AROUND M DWARFS 45

    quantities of mass at high speed (mostly in theform of energetic protons). Such ejections may bea substantial contributor to the early mass lossrates, which could be as high as 104 greater thanthe current solar rate (Mullan et al., 1992). Thesemass ejections can also affect the atmosphere ofa planet in the HZ (Khodachenko et al., 2007;Lammer et al., 2007). We also know, however, thatby the age of several Gyr (for example, ProximaCentauri), and likely well before that, the massloss rate has fallen below the solar rate (Wood etal., 2001). Thus, it is unlikely that an M star losesa significant amount of its mass in either coronalmass ejections or a steady stellar wind. Perhapsthe best evidence for this is that we do observespindown in most M stars. The spindown is afeedback mechanism that has the seeds of its owndestruction; the slower the star spins, the less ac-tivity and mass loss it has.

    The M flare stars identified prior to 1970 con-stitute roughly 1/10th of the dM population in thesolar neighborhood (Gurzadyan, 1970). Compar-ison of flare star populations in clusters of dif-ferent ages shows a decay of activity over severalhundred million years, and several groups foundthat the lowest mass M stars exhibit a longer flarestar phase than higher mass M stars (Mirzoyan,1990; Stauffer et al., 1991; Hawley et al., 2000).However, a kinematic study of 93 UV Ceti starswithin 25 pc of the Sun show a fairly high veloc-ity dispersion $ " 30 km/s, which indicates amean age of around 3 Gyr and no dependence onmass (Poveda et al., 1996). A small fraction ofthese nearby flare stars are either identifiablemembers of young disk groups or members of anold thick disk population.

    Two recent studies give new insight into thelong-term decay of M star activity. Silvestri et al.(2005) examined a sample of 139 older M starswhose ages can be estimated from their whitedwarf companions. Using V-I color as a chro-mospheric activity indicator, they found thatmany dM and dMe stars with ages 1–10 Gyr aremore magnetically active than predicted by thedecline of chromospheric activity seen in youngerclusters with ages $1 Gyr. Feigelson et al. (2004)obtained a small sample of 11 X-ray selected starsfrom a very deep Chandra X-ray Observatoryfield at high galactic latitude. Seven of these aredM stars at distances 50–500 pc and represent thehigh-activity tail of older stars in the upper disk.Most of these stars exhibit X-ray flares with peakluminosities around 1027–1028 erg/s, similar to

    young flare stars. Comparison with evolutionmodels suggests a steep decay law Lx ! t&2 inflare activity. This is consistent with X-ray stud-ies of dM stars in younger cluster populationsand the solar neighborhood, which indicate aflare decay law of Lx ! t&1 for ages $1 Gyr andsteepening at later ages (Preibisch and Feigelson,2005).

    The fraction of M stars with H# emission de-noting strong magnetic activity was known to below among early-M stars, but increases to nearly100% around M5.5–7 stars, and declines rapidlyamong early-L stars (Joy and Abt, 1974; Gizis etal., 2000). The largest samples studied to date in-clude 499 M dwarfs from the volume-completePalomar/Michigan State University survey ofnearby stars and nearly 8,000 M- and early-L-typestars from a flux-limited sample obtained fromthe Sloan Digital Sky Survey (Gizis et al., 2002;West et al., 2004). The Palomar/Michigan StateUniversity sample suggests that the magnetic ac-tivity of M0–M3 stars is bimodal, with most ex-hibiting no H# emission, some strong H# emis-sion, but with few emisions at intermediatelevels. The Sloan sample shows the fraction ofdMe stars peaks at 75% of M8 stars with %50%between M5 and L0; the somewhat lower fractionof dM3 stars is probably due to the higher frac-tion of older stars in this high galactic latitudesurvey. Using a more quantitative measure, how-ever, the average LH#/Lbol " 2 # 10&4 from M0to M5 and declines to 5 # 10&5 for M7–L0. ThePalomar/Michigan State University sampleshows an average around LH#/Lbol " 1 # 10&4from M0 to M3, followed by an increased rangeof activity (with a small decline in averageLH#/Lbol) from M3 to M5.

    3.2. Truncation of planetary habitability

    Although the stable lifetimes of M dwarf starscan be counted in the hundreds of gigayears, hab-itable planets that form initially may not remainhabitable for the lifetime of the star.

    3.2.1. Mass loss and the evolution of planetary po-sition relative to the HZ. Mass loss from young so-lar-type stars was first suggested by Whitmireand Doyle (1995) as an explanation for the faintSun paradox (Sagan and Mullen, 1972). Using the&2/3 spectral index excess indicative of a stellarwind at radio wavelengths, Mullan et al. (1989,1992) placed limits on the stellar wind from a K2

  • dwarf and two M dwarf stars, the latter indica-tive of mass loss at the rate of up to 10&10 solarmasses per year. Following this procedure, Doyleet al. (1996) put an upper limit on mass loss fromtwo young solar-type stars, the smaller upperlimit being about 9 # 10&11 solar masses per year,which also indicates what the effect of such massloss would have been on the early Solar System(the circumstellar HZ migrating inward while theplanets migrated outward). The migration of thecircumstellar HZ due to mass loss and tidal fric-tion (from the hydrological cycle on a synchro-nously rotating planet) and the possible loss of atidally locked planet’s atmosphere due to a re-duced magnetic field have been pointed out inDoyle (2006) and Lammer et al. (2007).

    We calculate what level of stellar mass losswould significantly affect a planet’s habitability.As the star loses mass, its luminosity decreases,and the HZ will contract. But, stellar mass losswill also cause the planet’s orbit to spiral outwardby conservation of angular momentum. On sucha planet, the combined effects of stellar mass losson the position of the HZ and the planet’s orbitalposition would exacerbate life’s ability to survive.A planet initially within the HZ could find itselfoutside after sufficient mass loss by the parentstar.

    The calculation proceeds as follows: lettingm " stellar mass, mp " planet mass, L " stellarluminosity, r " planet orbit radius (orbit as-sumed to be always circular for slow mass loss),then in order to keep constant insolation, r2 ' L(applies to center or boundaries of HZ).

    Main sequence luminosity evolution (theoreti-cally understood to be driven by stellar structureand nuclear fusion rates) depends on stellar mass,with L ' mq. For high-mass stars, q !4.5, but forM dwarfs the luminosity relationship has flat-tened to q !2.5. Therefore, since r ' Ll/2, constantinsolation requires that the radius decrease withmass as r ' m1.25. But, to conserve orbital angu-lar momentum of the planet as the parent starloses mass, mpvr must remain constant, and,therefore, the radius must increase in time asr1/ro " mo/m1.

    A factor of 2 characterizes the inner and outerradii of the HZ for the Solar System (about0.75–1.5 AU). So for stellar mass change to causea planet starting out at the inner edge of the HZto drift to the outer edge of the (shrinking) HZrequires that r1/ro " 2, and, therefore, m1/m0 "2&1/(1(1.25) " 0.73. Therefore, the star can lose at

    most about 25% of its mass before a planet onceresident at the inner edge of the HZ moves outbeyond the outer edge.

    At the Sun’s mass loss rate of 2 # 10&14 MSunper year, a main sequence M star would take sev-eral hundred billion years to lose 25% of its mass.Wood et al. (2005) found that the sustained massloss rate from M dwarfs is far below solar, soeven an initial burst of mass loss cannot produceany substantial effects on orbital change. Migra-tion out of the HZ due to sustained mass loss ofthe M dwarf star is unlikely to be a contributingfactor to early truncation of habitability; a ter-restrial planet that starts in the HZ will remainthere.

    3.2.2. Planetary interiors. For planets of both Mand G dwarfs, truncation of planetary habitabil-ity is likely to occur with the loss of large-scaleendogenic geologic activity. Loss of heat via con-vective heat transfer in a terrestrial planet’s sili-cate mantle is the primary driver of endogenic ge-ologic activity over solar system time scales.Mantle convection, in turn, can prove importantfor chemical recycling, mantle degassing of at-mospheric components, and sustaining sufficientinternal heat flow to power a magnetic dynamoin the planet’s metallic core. The lifetime of man-tle convective motions will tend to increase with(1) the thickness of the mantle (which will tendto increase with planetary radius) and (2) the ra-dioactive element content that provides heat dur-ing the decay process. The time scale over whichradioactive decay is relevant is roughly 3–10 Gyrtime scales for Earth-like planets (e.g., Sleep,2000). However, the dynamics of mantle flowmay be rather different on planets considerablymore massive than Earth because of the effects ofextreme pressures and temperatures on mantlematerials (e.g., Van den Berg et al., 2005). In ad-dition to planetary radius, the other primary con-trol on the depth of the mantle is the relativeamount of metallic iron in the planet. For exam-ple, the Earth’s metallic core makes up !55% ofthe planet’s radius compared with !75% inferredfor Mercury (e.g., Lodders and Fegley, 1998).Therefore, bulk composition and internal struc-ture of 1–10 Earth-mass planets (e.g., Valencia etal., 2006) will also affect the truncation of terres-trial planet habitability. Assuming typical en-richment in radioactive elements, planetary geo-logical activity on planets orbiting M dwarfs islikely to follow a similarly broad spectrum as ob-

    TARTER ET AL.46

  • HABITABLE PLANETS AROUND M DWARFS 47

    served in our Solar System. In addition, tidal dis-sipation may be an important additional mantleheat source on bodies in eccentric orbits, as oc-curs on the galilean satellites (Showman and Mal-hotra, 1999; Moore, 2003).

    Internally generated planetary magnetic fieldsare just as likely to occur on planets orbiting Mdwarfs as on planets around G dwarfs. Planetarymagnetic fields are generated on terrestrial plan-ets by fluid motions within the liquid portion ofthe planet’s metallic core (Olson and Aurnou,1999; Stevenson, 2003). Such fields can create a magnetosphere that surrounds the planet,which buffers it from the surrounding spaceenvironment and limits the atmospheric loss rate from cosmic ray sputtering (Johnson, 1994;Grießmeier et al., 2005). The critical ingredientsfor planetary dynamo action are a sufficient vol-ume of electrically conducting fluid, an energysource to drive motions in the fluid, and net or-ganization of the flow field (Roberts and Glatz-maier, 2000; Aurnou, 2004). Typically, Coriolisforces produced by planetary rotation act to or-ganize the planetary core flow. Planetary rotationperiods within the M dwarf habitability zone are estimated to be between 10 and 100 days.These rotation rates, though slower than that ofEarth, will still produce strong Coriolis forces in!100–1,000-km-deep core fluid layers. Thus, interms of dynamo theory (Stevenson, 2003), tidallylocked planets orbiting M dwarfs are viable can-didates for core dynamos.

    Recent numerical studies of planetary dynamoaction propose the following scaling law for plan-etary magnetic dipole moment (Christensen andAubert, 2006; Olson and Christensen, 2006):

    M ! (QBD)1/3R3C (1)

    where M is the magnetic dipole moment, QB isthe buoyancy flux, D is the thickness of the dy-namo generation region, and RC is the radius ofthe planet’s core. This study has been carried outusing only dipole-dominant dynamo solutions.Multipolar numerical dynamo solutions also ex-ist, although no clear scaling behavior has beenidentified for these cases (Christensen andAubert, 2006). In Eq. 1, note that, quite surpris-ingly, the scaling predicts that the planetary di-pole moment does not depend on the rotation rateof the planet. This does not mean that rotationrate is irrelevant in the dynamo process. It means,instead, that in the regime where the core con-

    vection is dominated by the effects of planetaryrotation, the typical strength of the dynamo fieldis controlled by the amount of available buoyancypower (Christensen and Aubert, 2006). Thus, interms of planetary dynamo action, a fast or slowrotator is correctly defined by the ratio of buoy-ancy forces versus Coriolis forces (Aurnou et al.,2006), not by the ratio of a planet’s rotation rateversus that of the Earth. A planet that is a fast ro-tator must have a buoyancy flux large enough togenerate core convection, but not so large that thebuoyancy forces swamp out the Coriolis forcesthat lead to the production of a well-organizedplanetary-scale magnetic field.

    3.2.3. Atmospheric escape. The ability of planetsorbiting M stars to retain their atmosphere willdepend critically on the nature of atmospheric es-cape processes that such planets experience.There are a variety of ways by which an atmo-sphere can escape into space. Perhaps the mostwell known are the thermal escape mechanismsof Jeans escape and hydrodynamic escape (for de-tails, see Walker, 1977). Jeans escape occurs abovethe exobase where molecular collisions are infre-quent and some of the upward traveling mole-cules have high enough velocities to escape intospace. This type of escape will occur in any plan-etary atmosphere with the escape flux dependenton planet mass. For the terrestrial planets, onlythe lightest elements (hydrogen and helium) canescape by this process, and the total loss over thelifetime of the solar system does not significantlydeplete their atmospheres. This type of escape isalso likely to be inconsequential for M star plan-ets.

    Hydrodynamic escape, however, may be moreof a factor. This kind of escape occurs when ther-mospheric temperatures are high enough suchthat the thermal energy of molecular motions iscomparable to the kinetic energy at escape ve-locity. In this situation, the escaping molecules,generally hydrogen, expand into the vacuum ofspace with such intensity that they can drag theheavier elements with them (e.g., C, O, N, etc.).Though still hypothetical, hydrodynamic escapehas been invoked to explain the fractionation ofthe xenon isotopes in the atmospheres of Marsand Earth (Hunten et al., 1987; Bogard et al., 2001).It is believed to have occurred very early in So-lar System history, when the solar extreme UVflux was much higher than it is today because itis the absorption of solar extreme UV that pow-

  • ers hydrodynamic escape (Chassefière, 1996).Since M stars are very active in the extreme UVfor much longer periods of time than G stars (Al-lard et al., 1997), it raises the possibility that hy-drodynamic escape could significantly depletethe atmospheres of planets that orbit them (Scaloet al., 2007).

    Nonthermal escape mechanisms include dis-sociative recombination, ion escape, and sputter-ing (Chassefière and Leblanc, 2004; Lammer et al.,2007). Unlike thermal escape, which mainly af-fects the lightest elements, nonthermal escape canremove heavier elements and molecules. Disso-ciative recombination occurs when UV photonsionize molecules, which then recombine withelectrons and produce energetic neutrals that canescape. This process is most efficient for smallplanets with weak gravity fields, and is believedto drive the escape of O, C, and N at Mars (McEl-roy, 1972; Hunten, 1993; Fox and Hac, 1997).However, the only flux of escaping material mea-sured at Mars is due to ion escape, wherein theelectric field associated with the solar wind di-rectly accelerates ions to the escape velocity(Lundin et al., 1989). This process differs fromsputtering in that the latter occurs when the ionsreimpact the atmosphere with sufficient energyto eject all molecules at the exobase (mainly C, O,CO, N, N2, and CO2). Both ion escape and sput-tering are facilitated by the lack of a magneticfield (which allows direct interaction of the solarwind with the atmosphere) and are sensitive tothe UV flux. Thus, M star planets that lack astrong enough magnetic field are susceptible toloss by each of these nonthermal escape mecha-nisms.

    This is far from a simple problem, and thoughsome effort has been expended to understand theevolution of the atmospheres of close-in hot-Jupiters (Grießmeier et al., 2004), very little re-search on atmospheric escape from M star plan-ets appears in the published literature. Clearly,this is a major issue and needs to be investigated.

    3.3. Origin versus sustainability of biology

    Heath et al. (1999) explored the opportunitiesfor known types of higher plants in terms of threemain factors: the distribution of climatic zonesacross the surfaces of synchronously rotatingplanets, the amount of photosynthetically activeradiation (PAR) arriving in M star insolation, andstellar variability.

    PAR, from about 4,000 Å to 7,000 Å, is ratherof special biological interest because it is neitherso energetic that it damages cells, nor so weakthat it cannot power water-splitting photosyn-thesis. As regards a perfect black body, the wave-length of peak energy output is inversely pro-portional to its temperature (Wien’s displacementlaw). However, stars are more complex than thisbecause light emitted from any level of the pho-tosphere (a layer hundreds of kilometers deep)can be absorbed and re-emitted by the higher lay-ers of the star in a manner determined by stellarcomposition. The proportion of IR in M star in-solation is substantially greater than that of ourSun, and as effective photospheric temperaturesdrop through the range for spectral class M, theamount of PAR falls. Molecular bands of TiO2 areparticularly important in reducing emitted stellarradiation in the PAR range, and red dwarf starsachieve energy balance by back-warming of theircontinua at other wavelengths, notably around10,000 Å. Stars that form out of material with lowcontents of “metals” (astronomical parlance forelements heavier than hydrogen and helium)should put out the most PAR but will havesmaller amounts of material with which to formterrestrial planets. These would be stars of the an-cient galactic halo and the older stars of the disk,while stars with the highest metallicities, those ofthe galactic bar/bulge and the intermediate andyoung disk populations, will have more materialto form terrestrial planets but suffer the greatestreduction in PAR output.

    Rough estimates were made of the amount ofenergy in and adjacent to the PAR region by ref-erence to published (Allen, 1973)