Cool stars of spectral type M, generally known as M dwarfs or red dwarfs, comprise about 70% to 75% of all stars, both in the Solar neighborhood and in the Milky Way Galaxy as a whole (Tarter et al. 2007). They range in mass from about 0.5 MSOL to less than 0.1 MSOL, with associated reductions in heat and brightness. The dim, lightweight character of red stars poses special problems to planet searches using either the standard radial velocity method or the increasingly successful transit method. As a result, search programs have only recently begun to include large numbers of M dwarfs in their target lists.

The low luminosities of M dwarfs have important consequences for the evolution and properties of any planetary systems that may form around them. Both their ice lines and their habitable zones will be located at much smaller radii than those of stars like our Sun, suggesting that M dwarf systems will be relatively compact. Further, the miserly rate at which these stars fuse hydrogen means that their lifetimes are orders of magnitude longer than those of more massive stars. All M dwarfs formed since the coalescence of the Milky Way are still burning comfortably on the main sequence, and will continue to do so for many billions of years. In fact, an M dwarf of 0.25 MSOL has a main sequence lifetime of 1 trillion years (Adams et al. 2005), compared to about 10 billion years for a G-type star like the Sun, or less than 1 billion years for an A-type star like Vega.

planet formation around M dwarfs

Several investigators have conducted theoretical studies of the evolution of planetary systems around M dwarfs. Their most robust conclusion is that gas giants form less frequently in these environments than around Sun-like stars, while terrestrial and ice giant planets may be relatively common (Laughlin et al. 2004, Ida & Lin 2005). However, most studies have focused only on a subset of potential M dwarf planets, and their diverse findings may not be mutually consistent.

Grant Kennedy and colleagues (Kennedy et al. 2006) developed a model to explain the intermediate-period Super Earths and Neptune analogs identified by microlensing surveys (Beaulieu et al. 2006, Gould et al. 2006, Bennett 2009). They argued that the pre-main sequence evolution of M dwarfs, which involves an erratically shrinking ice line, seeds the protoplanetary nebula with rocky cores. Kennedy and colleagues focused on the low-mass nebulae surrounding stars of 0.25 MSOL, where they found that several rock/ice planets of 0.1-1 MEA (Mars- to Earth-mass) can form between 1 and 4 AU. After the primordial gases dissipate, these objects interact and collide to form Super Earths of 1-5 MEA over a period of several hundred million years (much longer than the timescale for planet formation in the Solar System). In addition to Super Earths in the outer disk, their model predicts numerous Mars-mass planets orbiting within 0.5 AU of low-mass M dwarfs (Kennedy et al. 2006).

This study did not invoke any migration scenarios, implying that all planets would remain more or less in situ. With colleague Scott Kenyon, Kennedy subsequently returned to the problem of M dwarf planets in a study focusing on short-period Super Earths (Kennedy & Kenyon 2008). The initial conditions for their analyses included protoplanetary disks that lack gas giants but feature a strong enhancement of surface density at the ice line. Kennedy & Kenyon found that stars of 0.25-0.50 MSOL (i.e., most M dwarfs currently under observation) readily produce volatile-rich planets of 1-10 MEA at the ice line, which tend to undergo Type I migration to semimajor axes of about 0.04 AU. In a given system, the first such planet to migrate shepherds rocky planetesimals ahead of it, which coalesce into a small rocky planet that either achieves a stable orbit or is ultimately accreted by the migrating object. The study concluded that virtually all Super Earth-mass planets around a 0.25 MSOL star will undergo Type I migration, with the fraction of migrating planets decreasing with increasing star mass. Another of its findings – shared with many theoretical studies – is that planet formation can proceed only if the efficiency of Type I migration is reduced by a factor of 10 from classical models.

For M dwarfs and many lightweight K dwarfs, then, Kennedy & Kenyon envisioned “systems with evaporated rocky planets inside ~0.04 AU, and steam planets somewhat outside this distance,” along with “a few stalled ocean .... and icy planets” extending through the habitable zone (Kennedy & Kenyon 2008). Their model fits the observed system architectures of GJ 581, HD 40307, and HD 69830.

Sean Raymond and colleagues conducted numerical simulations of the formation of terrestrial planets around stars of sub-Solar mass, including both M and K dwarfs (Raymond et al. 2007). Their initial conditions included protoplanetary disks scaled according to observations of nearby star-forming regions, where low-mass stars harbor lightweight nebulae. Their model further assumes that no massive planets travel from beyond the ice line through the inner system, either by Type I or by Type II migration, but their results apply equally to systems with or without giant planets on longer-period orbits. Raymond and colleagues concluded that, inside the ice line, low-mass M dwarfs of 0.2 MSOL produce rocky planets in the mass range of the Moon and Mercury, while more massive stars of 0.4 MSOL produce slightly larger planets, up to the mass of Mars. These findings are consistent with the model of Kennedy et al. 2006. The authors conceded that the architecture of the GJ 581 system could not be explained by their simulations, and suggested that the system’s Super Earths accreted beyond the ice line and migrated inward – much like the explanation presented by Kennedy & Kenyon (2008). They also noted that volatile content in the inner nebula increases with overall disk mass, such that lower-mass disks yield only dry, rocky planets, while icy or ocean planets become more likely in more massive disks (Raymond et al. 2007).

More recently, Masahiro Ogihara and Shigeru Ida have also addressed the problem of planet formation in the inner systems of M dwarfs (Ogihara & Ida 2009). They found that the results of Raymond and colleagues depend sensitively on the suppression of Type I migration, which would otherwise bring icy planetesimals inside a circumstellar radius of 0.5 AU and contribute to the formation of volatile-rich terrestrial planets. However, Ogihara & Ida based their own analyses on atypically massive protoplanetary disks, enabling them to conclude – contrary to Kennedy et al. 2006, Raymond et al. 2007, and Kennedy & Kenyon 2008 – that Super Earths resembling the ones discovered in GJ 581 and HD 40307 might form in situ.

observational evidence

To date, radial velocity searches have detected 17 planetary systems around M stars. In spectral classification, all are earlier than M5, placing them at the brighter end of the M dwarf range. All but one are located within 20 parsecs (65 light years), and more than half are within 10 parsecs (32.6 light years) – evidence that successful detections are strongly biased by distance from our Solar System. If exoplanetary systems centering on M dwarfs are as common within 30 parsecs (100 light years) as they are within 10 parsecs, then more than 100 additional M dwarf systems await discovery just in this limited volume of space. (See a table of physical and orbital characteristics of the 17 systems.)

As a group, these systems harbor lower-mass planets orbiting at smaller semimajor axes than those around Sun-like stars, supporting the premise that system architecture scales roughly with stellar mass and thus with spectral type. Although more than half of these systems contain gas giants, they tend to be less massive than the ones orbiting F and G stars. In fact, 7 out of 11 gas giants orbiting red stars have minimum masses lower than Jupiter’s, well below the median for all known gas giants (~1.7 MJUP). First place in mass is currently held by GJ 676 Ab, with m sin i = 4 MJUP.

The confirmed M dwarf systems fall into two families: [1] those with gas giant planets on orbits of a few months to several years and [2] those with Super Earth to Neptune-mass planets on orbits of a few days to several weeks. Whether a given star belongs to one family or the other seems unrelated to its mass, contrary to theoretical expectations. Metallicity may be a more reliable predictor, since some investigators now detect a trend toward higher metal content in M dwarf hosts, especially those that harbor gas giants (Haghighipour et al. 2010). The fact that the gas giant family outnumbers the lower-mass family may indicate that theories of M dwarf planet formation need to be revisited.

The only real oddball in the group is GJ 876, which is unique in at least 3 ways: it is the only M dwarf system that harbors both Super Earth and gas giant planets, the only one with 2 known gas giants, and the only one with an unmistakable signature of Type II migration (i.e., 2 planets in a mean motion resonance). In addition, it hosts one of the most massive gas giants yet detected around an M dwarf (GJ 876 b, with m sin i = 1.93 MJUP, corresponding to an actual mass of 2.64 MJUP).

Curiously, the M dwarf systems with longer-period gas giants represent closer analogs of the Solar System than do all but a few of the more numerous systems orbiting Sun-like stars. Current data indicate that about one-third of detected M dwarf systems are Solar System analogs, meaning that they harbor one or more gas giant planets outside the system ice line but none inside it. Less than 10% of the systems around F, G, and K stars fit this description, and in the majority, the Jupiter surrogate has an eccentric orbit (e > 0.3) that would most likely preclude the survival of terrestrial planets in the habitable zone.

questions of habitability

The sheer abundance of M dwarfs throughout our Galaxy ensures that a large fraction of exoplanetary systems will be centered on red suns. These systems therefore loom large in contemporary thinking about extrasolar life. Given their meager energy output, the habitable zones of M dwarfs, like their ice lines, are located much closer to the primary than those of more massive stars. Mandell and colleagues provide inner and outer boundaries of 0.10 AU and 0.19 AU for the habitable zones of larger M dwarfs, with masses of about 0.4 MSOL (e.g., GJ 436). The corresponding boundaries shrink to about 0.024-0.045 AU for the smallest members of the class, with masses of about 0.1 MSOL (Mandell et al. 2007).

Such narrow boundaries, which imply belts ranging in width from 8.4 million miles down to 2 million miles, contrast strongly with the habitable zones of Sun-like stars. For example, the projected habitable zone of the Solar System extends from 0.8 AU to 1.5 AU, defining a region 65 million miles wide (Mandell et al. 2007). Simple limitations of mass and space, therefore, may make habitable planets even less likely to form in M dwarf systems than in the systems of F, G, and K stars (Raymond et al. 2007).

Additional issues arise from the habitable zone’s proximity to the host star. A planet orbiting in this space would be subject to stellar tides that would result in rotational synchronization, otherwise known as tidal locking. In this outcome, the same hemisphere would always face the parent star. Day and night would become geographical features rather than phases in a regular cycle, since the inner hemisphere would experience nonstop irradiation while the outer hemisphere would remain eternally dark.

According to some models, tidal locking would also result in complete loss of atmosphere through the freezing of volatiles on the dark side. However, this position has been challenged by recent demonstrations that atmospheric circulation (e.g., prevailing winds) would moderate global temperatures enough to maintain a substantial envelope of gases (Tarter et al. 2007).

It remains possible in theory for a tidally locked rocky planet, orbiting in the habitable zone of an M dwarf, to maintain Earthlike conditions in its so-called Twilight Zone. This is a narrow region separating the planet’s day side from its night side. In such a case, most of the day side would be burning, arid wasteland, while the night side would resemble Antarctica in winter. The Twilight Zone could sustain open bodies of water and, perhaps, both aquatic and telluric organisms. The width of any given planet's Twilight Zone will be a function of its orbital eccentricity, such that circular orbits have narrow temperate regions while increasing eccentricity results in wider Twilight Zones (Dobrovolskis 2007). (See artist Lynette Cook’s depiction of such a world.)

Another theoretical solution to the problem of M dwarf habitability lies in so-called “Ocean Planets,” as described by Leger, Selsis, Valencia, and others (Leger et al. 2004, Valencia et al. 2007). Such a planet would have about 50% of its mass in metals and silicates, and 50% in ices and other volatiles, supporting a global ocean hundreds of kilometers deep. Ocean currents would doubtless be even more successful than winds in moderating temperatures, while the watery environment might be an effective incubator for life forms. (Readers of science fiction may be reminded of the sentient global ocean imagined in Stanislaw Lem’s novel Solaris, or the oceanic Eden in C.S. Lewis's Perelandra.)

Closer to the host star, however, such volatile-rich objects would become scalding Steam Planets, and farther away they would devolve into cloud-wrapped snowballs, smaller versions of Uranus and Neptune.

For Super Earths and other terrestrial planets in the habitable zone, one alternative to the scenario of tidal locking is spin-orbit alignment. A planet in a relatively eccentric orbit may escape synchronization and establish a rotational spin that is some multiple of its orbital period (Dobrovolskis 2007). This has actually happened to the planet Mercury, which has an eccentricity of 0.206, a rotational period of almost 59 days, an orbital period of 88 days, and a spin-orbit resonance of 3:2 (Correia & Laskar 2004).

Last update March 2010