Wednesday, July 17, 2013

Au-delà de neuf cents exoplanètes

Figure 1. Just as the exoplanet census passed 900, a new study reported that HD 189733 b, the nearest transiting Hot Jupiter, is a deep blue world where molten glass rains through seething hydrogen.

On Monday, July 1, the census ticker at the venerable Extrasolar Planets Encyclopaedia (EPE) turned over a new planetary century: 900 confirmed exoplanets, including the results of all current detection methods. Since then the count has moved on, hovering at 919 planets today. All these worlds are the harvest of 25 years of searching.

What planetary flavors does this numerical abundance offer? How much variety do we find in the Milky Way? How many Earth-like planets are out there?

To answer these questions, we need to take into account the fact that different search methods find different kinds of planets. The strengths and limitations of each method introduce enormous bias into the exoplanetary census.

search methods & planet populations

Most (58%) of the known exoplanets have been discovered by the radial velocity method. It returns data on the masses and orbital periods of specific kinds of planets: gas giants like Jupiter orbiting within about 5 AU of their host stars, and low-mass planets (Super Earth to Neptune-size) orbiting within about 1 or 2 AU. Under optimal conditions, it can also detect Earth-mass planets orbiting within a few tenths of an AU. This method is sensitive to multiplanet systems of many kinds, as well as to single planets. However, it isn’t very good at finding systems like our own or planets like Earth.

The next largest share of exoplanets (34%) have been observed in transit across the face of their host stars, so that we can calculate their radii and orbital periods, but not their masses. The transit method is sensitive only to planets that happen to orbit their stars precisely along the line of sight from Earth. Its effectiveness in ground-based surveys is limited, since 97% of transiting objects detected from the ground are Hot Jupiters (i.e., gas giants with periods shorter than 10 days – a monotonous selection indeed). Fortunately, detections by space-based programs now outnumber those from the ground, and the vast majority of those detections represent planets of Neptune size or less. Space-based observations have also been remarkably successful in finding low-mass multiplanet systems similar to our radial velocity neighbors, HD 40307 and HD 69830. Until other search methods improve, they remain our only hope of detecting Earth-size planets with habitable temperatures. Meanwhile, all the planets on which we have enough data to establish that they are rocky (like Earth) and smaller than two Earth radii (2 Rea) are Hellworlds, orbiting their stars in periods of a few days or less.

Figure 2. Small Planets with Measured Masses & Radii

The major drawback of transit data is the absence of constraints on an object’s mass, except in rare cases where a planet’s orbital period is modified by the influence of an adjacent transiting planet (a phenomenon known as transit timing variations or TTV). The major strength of transit data is that, when they can be supplemented by radial velocity data, the resulting synergy provides our clearest understanding of exoplanetary environments. Figure 2 shows artist’s views of half a dozen small, well-constrained planets (“Super Earths” and “Mini Neptunes”) with masses (M) and radii (R) expressed in Earth units. The three Kepler planets are located at distances of 173 to 613 parsecs (565-2000 light years away); the other three are next-door neighbors of our Sun, located within 22 parsecs (72 light years). Similar line-ups of low-mass planets are available here and here.

Three other search methods are also in use, but they offer a modest contribution to the total census (collectively, less than 10%, even allowing for overlap between transiting and radial velocity detections). Microlensing, which depends on chance alignments of distant stars , is sensitive primarily to planets separated from their host stars by about 1.5 AU to 5 AU (Bennett 2008). It has detected 20 planets so far (2% of the total), ranging from Super Earths to gas giants; none are tightly constrained. Pulsation timing has reported 15 candidate planets that EPE tags as confirmed, although several have been challenged as unphysical. The most reliable detections in this group accompany neutron stars. Imaging, the old-fashioned way of doing astronomy, has returned a much larger yield – 34 purported planets (almost 4% of the total census) – but except for a small subset, this group is dubious. Most candidates are located too far from their proposed host stars to have formed in a circumstellar disk, and many have estimated masses consistent with brown dwarfs instead of planets.

All five methods have been in use for more than a decade, each contributing its own piece of the picture. In the past few years, a new approach to detection – reanalysis of existing radial velocity and transit data, using Bayesian inference – has also proved fruitful. In fact, the exoplanetary census just surged to 900 and beyond partly as a result of two brand-new reanalyses of radial velocity data, discussed here and here.

two kinds of planets

Given the limitations of the available search methods, only radial velocity and transit data are useful for making generalizations about the known exoplanets. It seems clear that this population has two major divisions, with the dividing line at about 8 Rea and 40 Earth masses (Mea). Below this boundary are the low-mass planets, which include two subspecies: 1) rocky planets like Earth (“terrestrial” or “telluric” planets), without any substantial amount of hydrogen in their bulk composition; and 2) planets with rocky (or perhaps rocky/icy) cores surrounded by hydrogen envelopes constituting at least a few percent, but less than half, of their bulk composition (variously “gas dwarfs,” “Mini Neptunes,” “ice giants,” or “exo-Neptunes”). It remains moot whether this group also includes a third subspecies, consisting of relatively massive objects (1 Mea or more) that contain a substantial amount of volatiles (water, ammonia, methane) along with rock, without any significant hydrogen envelope. Such planets, if they exist, would be scaled-up versions of Ganymede and Titan in our Solar System.

Figure 3. Radii of 298 Transiting Planets < 18 Earth Radii: July 2013
Above the boundary are the gas giants, defined as objects whose bulk composition is dominated by hydrogen. These planets have radii larger than 8 Rea and masses in excess of 40 Mea. Although gas giants constitute a majority (72%) of exoplanets detected by radial velocity and transit searches, they are a minority among all planets in our Galaxy (Mann et al. 2010).

Figure 3 illustrates the double-peaked distribution of exoplanet radii. The letters E, U, S, and J indicate where the radii of Earth, Uranus, Saturn, and Jupiter fall on the same scale. Recent analyses argue that the population peak between 2 and 3 Rea is real, and not an artifact of detection bias (Petigura et al. 2013). In other words, the population of low-mass planets does not continue to increase at radii smaller than 2 Rea. At best, the actual number of Earth-size planets – admittedly harder to detect than larger planets – may rival that of their siblings between 2 and 3 Rea, but it probably does not exceed that population. Some well-known examples of the latter are illustrated in Figure 2.

Current radial velocity data indicate that the median mass of all extrasolar gas giants is 1.5 Jupiter masses (Mjup). This figure has remained fairly consistent over the past five years, even as the number of exoplanets has tripled. Ninety percent of gas giants have minimum masses of 7 Mjup or less. Unlike the case in lightweight planets, we see little correspondence between mass and radius for the more massive gas giants observed in transit. Instead, temperature plays the defining role. With a single exception, the objects in Figure 3 with radii above 12 Rea are Hot Jupiters, which have very high equilibrium temperatures and usually inflated atmospheres (see Figure 1 for an example). 

The fraction of Hot Jupiters is another relatively consistent feature of the overall sample. They constitute 35% of all known gas giants and 25% of all well-constrained exoplanets of any kind. However, several studies have demonstrated that the true occurrence rate for these planets is much lower: Hot Jupiters orbit far fewer than 1% of the stars in the Milky Way.

More than half (53%) of all exoplanets with well-defined orbits have semimajor axes smaller than Mercury’s (0.49 AU), while only 23% have semimajor axes larger than that of Mars (1.52. AU). All the cooler planets are gas giants, and rather massive ones at that. Just four objects on cool orbits fall in the low-mass tail of the gas giant range, with minimum masses between 45 and 60 Mea. The median mass for all cool giants (i.e., trans-Martians) is 2 Mjup, illustrating the tendency of more massive planets to travel on wider orbits.

Among low-mass planets, hot orbits predominate. Although the range of orbital periods for this group extends from 8.5 hours to 600 days, almost 40% have periods shorter than 10 days, and only 5% have periods longer than 225 days (the period of Venus around our Sun). However, this distribution is certainly the result of detection bias; a hidden population of small planets on longer-period orbits still awaits discovery. Among low-mass planets for which masses can be estimated, the median is 10 Mea, corresponding to an object more reminiscent of Uranus than of Venus. Again, this figure might be the result of detection biases instead of a reflection of reality.

three kinds of systems

One welcome trend is the increasing share of multiplanet systems within the full census of well-constrained exoplanets. Currently, just 60% of planets occur in systems with only one planet, and within this subpopulation, 92% are gas giants (among which almost half are Hot Jupiters). Another 23% of the full census are found in systems with two planets. Like the singleton systems, these are dominated by gas giants (60% of all planets in two-planet configurations). However, only 5% of planets in these systems are Hot Jupiters. The two-planet sample provides an interesting illustration of the adage that birds of a feather flock together, since 84% of systems consist of either of two gas giants or two low-mass planets, while only 16% have one of each. Even among the odd couples, the vast majority of giant companions are less massive than 0.5 Mjup.

Finally, 17% of the total appear in systems with three or more planets; 73% of planets in these systems are low-mass, and as in the two-planet configurations, only 5% of systems contain Hot Jupiters.

This breakdown of systems by planeticity (an indispensable coinage that hasn’t yet found its way into standard dictionaries) tells us something significant about the character of the two exoplanet species. The little folk are gregarious and abundant, while the heavyweights are hostile and aloof. The former play well with others, while the latter prefer solitude, or at most the company of a few representatives of their own kind. When the two species manage to co-exist in the same system, something special sometimes happens, such as life.

The study of highly planetic systems – i.e., those with three or more planets – opens a unique window on extrasolar system architectures. In a previous post I proposed three basic architectural types, defined by their planet populations: low-mass, mixed-mass, and high-mass. I continue to find these categories useful for conceptualizing the exoplanet census.

Low-mass systems contain closely-spaced planets of similar size traveling on circular orbits near their host stars. They are often extremely compact. For example, Kepler-33 has five planets (three of them larger than Neptune) within a semimajor axis of 0.26 AU, while Kepler-11 has six planets within a semimajor axis of 0.5 AU.

Mixed-mass systems, such as the Solar System, 55 Cancri, and HD 10180, contain a broader range of planet masses, with at least one, sometimes two, and rarely three gas giants. Orbits in these systems tend to be more expansive than those of low-mass systems, while their giants are typically much less massive than the median for the full sample. None have Hot Jupiters. Overall, the mixed-mass systems look more like low-mass systems than like high-mass systems

High-mass systems containing three or more gas giants, but no low-mass planets, are quite rare. (Nevertheless, one of the best-studied systems detected by direct imaging, HR 8799, seems to host at least four giants on wide orbits). The components of these systems tend to be more massive than the median for all gas giants, and their orbits are wider and more eccentric than those in the other two architectural types. Half have Hot Jupiters.

In general, gas giants are intolerant of other gas giants, especially in close proximity. Among all extrasolar systems detected to date, only six (less than 1%) contain three or more giants. Two are mixed-mass systems; the rest constitute the high-mass sample shown in the table below. The more massive the giant, the less likely it is have neighbors, in particular diminutive neighbors.

Table 1. 38 systems with three or more planets, binned by architectural category
Early in 2012, after a few months of unusually rapid growth in the census at EPE, I made a prediction about where the numbers were headed: “At this rate, we can expect the census to reach the magic number of 1000 before the end of 2013.” Given the much slower growth over the past year, my prediction may be on shaky ground. It will be interesting to see how things pan out.


Bennett DP. (2008) Detection of extrasolar planets by gravitational microlensing. In Exoplanets, ed. John Mason. Berlin: Springer, 2008. Abstract. 
Mann AW, Gaidos E, Gaudi BS. The invisible majority? Evolution and detection of outer planetary systems without gas giants. (2010) Astrophysical Journal 719, 1454–1469. Abstract: 
Petigura EA, Geoffrey MW, Howard AW. (2013) A plateau in the planet population below twice the size of Earth. In press.



  1. Nice article. It will be interesting to see how this holds up with better / different instruments.

    In particular: as we do surveys for nearby transiting systems where we can do follow up RV work; and as we get data without various biases. For example:

    Against "cool Jupiters" for both RV and transits. If we can get a large sample across the metallicity range e.g. is there a "tail" to the distribution at low metallicities?

    Against detecting multiplanet transiting systems - we only see transits for planets with orbits that are tightly coplanar.

    Towards transiting low mass (compact) systems around smaller stars (relative transit depth, transit frequency and looser constraints on coplanarity)

    Against RV & transiting bias smaller, higher period planets

    Extremely interesting times!!!

    1. "Extremely interesting times!!!"

      I couldn't agree more. (And thanks for reading!) I'm very curious to see if some of the really compact systems like Kepler-11 have gas giants in their outer regions. My hunch is that they don't.

      Mass in a planetary system seems to be concentrated either in a few gas giants, which tend to stay on cooler orbits, or a lot of Neptune-like planets, which tend to stay on warmer orbits. But a few new discoveries could change that picture.