systemic

Philippe Thebault sent me a link to an article on the Alpha Centauri planet search published earlier this month in the Frankfurter Allgemeine Zeitung. The text is in German, but the Google translator does a passable job of getting the gist across.

I got my first inkling of the Geneva Planet Search’s Alpha Centauri campaign through Lee Billings’ article in Seed Magazine. (See this post). In the Frankfurter Allgemeine article, Francesco Pepe gives further details — Alpha Cen B is one out of ten stars that are receiving special scrutiny for terrestrial planets at HARPS. They are getting one observation every two weeks, meaning that the star is being hit roughly one out of every two of their planet search nights:

“Allerdings müssen wir uns Harps mit anderen Gruppen teilen”, sagt er. Zudem ist Alpha Centauri B nur einer von zehn Sternen, die sie auf erdähnliche Planeten absuchen wollen. “Aber alle zwei Wochen schauen wir damit auf Alpha Centauri, und das Gerät ist sehr effizient.”

This quote implies that my speculations regarding the Geneva team’s data collection rate on Alpha Cen B were somewhat overheated. Instead of getting 100 ultra-high-precision HARPS velocities per year, it looks like a more realistic estimate of their current rate is 25 velocities per year. Since signal-to-noise increases as the root of the number of observations, this means that the minimum mass threshold for Alpha Cen Bb at any given time is approximately doubled relative to my estimates at the beginning of the Summer. Instead of arriving at 2.5 Earth masses in the habitable zone a bit more than a year from now, they’ll be at roughly 5 Earth masses.

Now nobody likes backseat drivers. As the saying goes, “theorists know the way, but they can’t drive”, and theorists have had a particularly dismal record in predicting nearly everything exoplanetary.

But nevertheless, I’m urging a factor-of-four increase to that data rate on Alpha Cen B. I would advocate two fully p-mode averaged velocities per night, 50 nights per year. I know that because Alpha Cen B is so bright, the duty cycle isn’t great. I know that there are a whole panoply of other interesting systems calling for time. It is indeed a gamble, but from the big-picture point of view, there’s a hugely nonlinear payoff in finding a potentially habitable planet around Alpha Centauri in comparison to any other star.

During the next few months, it’s inevitable that one of the numerous Super-Earths that have been turning up in the radial velocity surveys will be announced to be observable in transit (see, e.g. this post). When that occurs, we’ll effectively have had our last first look at a truly new category of planet — the logarithmic mass interval between Earth an Uranus is currently by far the largest among the 70-odd planets that have accurately determined radii. My own guess is that the emerging population of super-Earths will be better described as a population of sub-Neptunes. That is, the transit depths will indicate compositions that are largely water.

So if 5-Earth mass planets turn out to be primarily water-based rather than rock-based, it’s (in my mind) an argument in favor of cranking up the data rate on Alpha Cen B. There were no structurally substantial quantities of water in the Alpha Cen planet-forming environment. If we’re seeing sub-Neptunes rather than super-Earths in the HD 40307, Gliese 581, et al. systems, then the odds are heightened that any planets orbiting Alpha Cen B are less than 2 Earth masses. There’s no payoff in tuning your Alpha Cen B strategy for sub-Neptunes. Finding truly terrestrial-mass planets will require paying full freight.

In the early nineteenth century, the detection of stellar parallax was a problem fully equivalent in both scientific excitement and prestige to the modern-day detection of the first potentially habitable extrasolar planet. I think it’s worth noting that the prize of discovery of the first stellar parallax went not to the eminently capable (but overly cautious and slow-moving) observer who accumulated data on the best star in the sky, but rather to an observer who focused on a rather obscure star in the constellation Cygnus.

Here’s a link to the article, “Thomas Henderson and Alpha Centauri” by Brian Warner of the University of Cape Town.

During my visit to the Paris Observatory earlier this summer, Alain Lecavelier showed me the work that he and David Sing and their collaborators have been doing to get a better handle on the atmospheric conditions on HD 209458b. Using the STIS spectrograph on HST, they’ve obtained both medium-resolution and low-resolution visible-wavelength absorption spectra of starlight shining through the atmosphere of the planet as it transits the parent star.

HST is sensitive enough to allow startlingly detailed portraits of “sunsets” that took place back in the mid-1850s. Here’s a reworking of Figure 1 from Sing et al. (2008):

Illustrator-editable .pdf of above with title and source.

Sing et al. manage to do a good job of matching the features in the spectrum. The big absorption spike in the orange is due to the presence of atomic sodium. Their atmospheric models also include Raleigh scattering by hydrogen molecules, a temperature inversion in the atmosphere, condensation of sodium sulfate on the planet’s night side, and the presence of titanium and vanadium oxide in the atmosphere. (Titanium oxide can be invoked to play a big role in modulating the visual appearance of hot Jupiters for much the same reason that it’s used as an opacifier in ordinary paint.)

With a detailed atmospheric model in hand, it’s possible to calculate both the color of the sky and the color of HD 209458b at various sight lines through the air column. David and Alain did exactly that, and have made an animation from the perspective of an observer in an asbestos-coated balloon drifting nightward across the terminator. The effect is reminiscent of a Turrell skyspace:

Here’s a link to their French-language press release. According to the inimitable google translator, “star at bedtime absorption is cyan”

In reviewing grant proposals and observing proposals that seek to study extrasolar planets, one notices that two cliches turn up with alarm-clock regularity. Number one is Rosetta Stone, as in this or that planetary system is a Rosetta Stone that will enable astronomers to obtain a better understanding of the formation and evolution of planetary systems. Number two is ideal laboratory, as in this or that system is an ideal laboratory for studying the processes that guide the formation and evolution of planetary systems.

A terse unsolicited e-mail from Gaspar Bakos always means that a big discovery is in the offing, and today was no exception:

Hello Greg,

You may like this.
http://xxx.lanl.gov/abs/0907.3525

Best wishes
Gaspar

Indeed! HAT-P-13b and c constitute a really exciting discovery. For a number of reasons, this system is a Rosetta Stone among extrasolar planets, and in large part, this is because the system is an ideal laboratory for studying processes such as tidal dissipation and orbital evolution.

HAT-P-13 harbors the first transiting planet that has a well-characterized companion planet. In this case, the outer companion has a P=428 day orbit, an Msin(i) of 15 Jupiter masses, and an eccentricity, e=0.7. In the following diagram, the orbits and the star are shown to scale; the small filled circles that delineate the outer orbit show the position of the outer planet at 4.28 day intervals.

Illustrator-editable PDF of the above

Of obvious interest is the question of whether planet c can be observed in transit. The a-priori probability is seemingly enhanced by the transit of the inner planet. (Give that one to the good Reverend Bayes). The next opporunity rolls around in April 2010, with the opportunity to observe secondary transit following a bit more than two months later.

It’ll be quite something if planet “c” does transit. A sense of the wide open spaces in the system can be obtained by plotting the star and the two planets to scale with their respective separations at the moment of inferior conjunction. Given the width restriction of the blog post format, one needs to present this plot vertically:

There’s a lot more to say about the HAT-P-13 system — so much in fact, that Peter Bodenheimer, Konstantin Batygin and I are furiously writing an ApJ letter. Should have it out the door in a day or so, with a roundup to follow here on oklo.org immediately thereafter…

oklo.org is heading into its fifth year, and we’ve just hit something of a milestone: this is the 300th post. A great deal has been learned about extrasolar planets in the past half-decade, and I’ve found that participative reporting has been a great way to keep up with, and even, sometimes, to influence the course of events.

We’ve also just hit another big milestone with the release of the Systemic Console Paper. Our manuscript has just been accepted for publication in the peer-reviewed journal Publications of the Astronomical Society of the Pacific, and the article is now available on astro-ph.

In coming posts, we’ll be highlighting the many new features of the console, and we’ll be updating the now badly-out-of-date tutorials. If you are interested, then by all means download the latest “cutting edge” version, which is available on Stefano’s website.

And finally, if you use the console, and find it useful, please consider citing Meschiari et al. 2009 in your publications.

Earth occulting the Sun, seen from Apollo 12 (source).

The year 1995 fades into increasingly ancient history, but I vividly remember the excitement surrounding Mayor and Queloz’s Nature article describing the discovery of 51 Peg b. Back in the day, the idea of a Jovian planet roasting in a 4.2-day orbit was outlandish to the edge of credibility.

In the five years following the Mayor-Queloz paper, four additional Doppler-wobble planets with periods less than a week (Ups And b, Tau Boo b, HD 187123b, and HD 75289b) were announced. Each one orbited close enough to its parent star to have a significant a-priori probability of transiting, and by mid-1999, the summed expectation for the number of transiting planets grew to N=0.68. Each new planet-bearing star was monitored for transits, and each star came up flat. Non-planet explanations for the radial velocity variations gained credence. The “planets” were due to stellar oscillations. The “planets” were actually mostly brown dwarfs or low-mass stars on orbits lying almost in the plane of the sky.

The discovery of HD 209458b, the first transiting extrasolar planets was therefore a huge deal. Instantly, the hot Jupiters gained true planetary status. There’s a huge leap from a mass-times-a-sine-of-an-inclination to density, temperature, composition, weather. 209458 was the moment when the study of alien solar systems kicked into high gear.

At the moment, we’re within a year of getting news of the first Earth-mass planet orbiting a solar-type star. It’s effectively a coin flip whether the announcement will come from Kepler or from the radial velocity surveys. In either case, the first Earth will likely be too hot for habitability, but within a few years we’ll be seeing genuinely habitable, multi-million dollar worlds. Kepler, for one, will deliver them in bulk.

Enter the TESS mission.

Here’s the scoop: The TESS satellite consists of six wide-field cameras placed on a satellite in low-Earth orbit. If it’s selected, then during its two-year mission, it will monitor the 2.5 million brightest stars with a per-point accuracy of 0.1 millimagnitude (one part in ten thousand) for the brightest, most interesting stars. It will find all of the transiting Jovian and Neptune-mass planets with orbital periods of less than 36 days, and it can make fully characterized detections of transiting planets with periods up to 72 days. Where transits are concerned, brighter stars are better stars. TESS will locate all the bright star transits for Neptune-mass planets and up, and equally important, it will find the best examples of large transiting terrestrial planets that exist.

TESS also provides an eminently workable path to the actual characterization of a potentially habitable planet. Included in the 2.5 million brightest stars are a substantial number of M dwarfs. Detailed Monte-Carlo simulations indicate that there’s a 98% probability that TESS will locate a potentially habitable transiting terrestrial planet orbiting a red dwarf lying closer than 50 parsecs. When this planet is found, JWST (which will launch near the end of TESS’s two year mission) can take its spectrum and obtain resolved measurements of molecular absorption in the atmosphere.

If TESS is selected for flight, we’re literally just five years away from probing the atmospheres of transiting planets in the habitable zone.

An interesting discovery announcement came across the wire on Friday. In an article to be published in the Astrophysical Journal, Steven Prado and Stuart Shaklan of JPL write up their detection of a ~6 Jupiter mass companion orbiting the nearby ultra-low-mass red dwarf VB 10. Their discovery was made astrometrically, using a modern CCD camera attached to the venerable Palomar 200-inch telescope. JPL put out a press release to go along with the article.

VB 10 contains about 78 Jupiter masses, just barely lifting it above the minimum mass required to qualify as a bona-fide hydrogen-burning main sequence star. It’s got roughly ten times the mass and ten times the density of its companion. In the center-of-mass frame, the system configuration looks like this, where I’m taking a guess at the unknown eccentricity:

I wouldn’t call VB 10b a planet in the usual sense. With a mass of order one-tenth that of its parent star, it’s almost certainly straggling in at the very bottom of the stellar initial mass function. It’s a low-mass brown dwarf impinging into the “planet desert” from above. Gravitational instabilities tend to crop up if a protostellar disk exceeds 10% the mass of its central star, so the VB 10 system probably formed via the fragmentation process that leads to binary stars rather than via the core accretion mechanism that seems to be responsible for the majority of Jovian planets. Presumably, a similar fragmentation-based process had a hand in the formation of 2M1207, in which a ~4 Jupiter-mass secondary orbits a ~25 Jupiter-mass primary:

Photo credit: ESO (VLT/NACO)

At a distance of only 19 light years, VB 10 is (relatively speaking) just right next door. In tandem with its wide binary companion Wolf 1055, it currently ranks as the 68th-nearest known stellar system. That one need not travel far afield to find VB 10b means that objects like VB 10b are probably common in orbit around the most dimunitive red dwarfs.

As instrumentation improves, it’ll eventually become possible to survey the satellite systems of objects like VB 10b. In our solar system, Jupiter, Saturn and Uranus all have roughly 2×10-4 of their primary mass locked up in satellites. I’m guessing that this rule of thumb will continue to hold when exomoons start getting detected, but I bet that it won’t hold true for objects that formed via fragmentation.

The VB 10 system is built to last. The primary will enjoy a main-sequence lifetime of close to ten trillion years, during which time the Milky Way-M31 merger remnant will become increasingly isolated from all the other mass that makes up the currently observable universe. Tidal evolution will gradually tighten up the orbit of VB 10b, meaning that the binary will quite possibly survive and harden further during quadrillions of years of encounters with passing degenerates. Barring other catastrophes, gravitational radiation will eventually bring VB 10 and VB 10b together into merger. That shot of good pure H will revive the dead helium remnant of VB 10, causing it to shine for a further hundred billion years or so.

It’s a struggle to stay afloat in the non-stop flow of results. As a case in point, the Mayor et al. discovery preprint for HD 40307 b, c, and d has already been up on astro-ph for several weeks, and I only just a chance to read it carefully. The paper spells out the details of the announcement made at the Nantes conference last month, and ends with some bromides that seem to telegraph that the photometric transit search for planets b, c, and d is not yet definitive:

One of the most exciting possibilities offered by this large emerging population of low-mass planets with short orbital periods is the related high probability to have transiting super Earths among the candidates. If detected and targeted for complementary observations, these transiting super-Earths would bring a tremendous contribution to the study of the expected diversity of the structure of low-mass planets.

No controversy in that paragraph. It’ll be undeniably dope when the super-Earths start materializing in transit. Given that population of hot sub Neptunes in our Galaxy is apparently more than five times larger than the human population, it’s also likely that a significant number of these planets transit bright stars, and that’s good news for JWST.

In the interim, it’s not hard to see why the jury is still out on transits for HD 40307 b,c, and d. With its period of 4.61 days, the ~4 Earth-mass HD 40307b has a healthy a-priori transit probability of ~7%. Its expected transit depth, however, is a meager 0.05%. So far, the shallowest known transit for an extrasolar planet is that of HD 149026 b, which, at 0.3%, is fully six times deeper.

A ground-based detection of transits by HD 40307b would be quite a coup indeed. Is it feasible?The parent star HD 40307 is a K dwarf that’s quite similar in both spectral type and apparent magnitude to HD 189733. We can thus draw on the HD 189733 transits to get a ball park idea of the quality of the photometric data that one might expect from HD 40307. The best published ground-based light curve for HD 189733 that I could find comes from Bakos et al. (2008), who used the FLWO 1.2m telescope in Arizona to get the time series that I’ve reproduced just below. The skimpy expected depth of a central transit by HD 40307b is shown for comparison. The situation looks daunting.

The out-of-transit data in this light curve has a reported RMS scatter of 2.6 mmag for photometric points taken every 17 seconds (binned data is shown in the figure). Naive statistics thus imply that a 0.5 mmag central transit by HD 40307b could be detected by the FLWO 1.2m with at least several sigma confidence. Life, however, is more than root N. Systematic errors are probably large enough to scotch a discovery on a single night of observation, but nevertheless, by repeatedly observing, either with multiple nights or with multiple telescopes, a detection seems within reach. And it’s worth in excess of USD 5M. (At the moment, it seems there’s little need for European or Asian observers to hedge their currency risk.)

In the event that photometric campaigns aren’t up to the task, it’s in the realm of possibility that a transit by HD 40307b could be extracted via a spectroscopic detection of the Rossiter-McLaughlin effect. Assuming a 1 km/sec rotational velocity for the star, the expected half-amplitude of the Rossiter distortion is similar to the error bars on the published radial velocities. In the following figure, I’ve dished up a simulated Rossitered data set from HARPS, superimposed (with an offset for clarity) on a blown-up version of the radial velocity plot in the paper. During a single occultation, the radial velocities can produce a ~0.85 sigma detection.

In this case, the economics are a bit steeper, but still viable. At the current dollar-euro exchange rate, I’d estimate that USD 15K is a fair price for a HARPS night. (Forgive all this yak yak about currency — as an American traveling in Europe at the moment, I’m rather shocked to be seeing $6.24 0.7l bottles of water at the airport newstand!). One would need 4 hours, or half a night to observe the transit and get adequate baseline. To be at least four-sigma sure, you’d want to rack up ~20 full transits (which would take quite a while). Factoring in the expectation value of 0.07 arising from the transit probability, this works out to a USD ~2M detection.

This morning, I awoke to an inbox full of indications that there was indeed plenty of drama in the club.

From one of our correspondents:

He threw out dozens of new systems, very graphically, on slips of paper, like playing cards, floating down on a pile on the screen. Very dramatic. But no HD numbers on those slips!

He predicts 1 to 1.5 Earth sensitivity by around 2010 (extrapolating a trend).

He has been monitoring about 400 FGK slow rotators since 2004, with HARPS.

Can do 0.5 m/s today, 0.1 m/s in near future.

Noise sources are astroseismology, which settles to about 0.1 m/s after 15 minutes of integration, and a worse one, star spots, which settle to about 0.5 m/s after 15 minutes but do not drop lower, even though theory says level should drop to about 0.1 m/s.

He says he has 40 new candidates in the 30-50 day period range, and mass less than 30 Earths.

Nevertheless, after the drama, he did report 3 Neptune-type systems, all focused on the Super Earth theme of the meeting.

The centerpiece was definitely HD 40307, a deep southern K2.5 V star only 40 light years distant, with a metallicity roughly half that of the Sun. It has three detected planets, with Msin(i)’s of 4.2, 6.9, and 9.2 Earth masses, and corresponding periods of 4.31, 9.62, and 20.45 days. It’s fascinating that these planets are close to, but aren’t actually in a 4:2:1 resonance. This is really a remarkable detection.

With 40 candidates in the pocket, the Geneva team does, however, seem to be keeping some of their powder dry, perhaps in anticipation of a low-mass transit. Here’s a link to the ESO press release, which has triggered 93 news articles and counting.

In the press release image, HD 40307d is definitely all that and a bag of chips. Puffy white clouds, azure seas, continents, soft off-stage lighting…

On December 29th 1959 at the annual meeting of the American Physical Society at Caltech, Richard Feynman gave a remarkable talk entitled “There’s plenty of room at the bottom” in which he foresaw the impact that nanotechnology could have on materials science. At the beginning of the lecture he remarked (in a vernacular that dates him to the Eisenhower era):

I imagine experimental physicists must often look with envy at men like Kamerlingh Onnes, who discovered a field like low temperature, which seems to be bottomless and in which one can go down and down. Such a man is then a leader and has some temporary monopoly in a scientific adventure.

Over the past several years, the oklo.org party line has been that the radial velocity method for exoplanet detection is similarly equipped with the potential to go down and down in planet mass, and to continue with at least a respectable share of the lead in the ongoing scientific adventure.

That said, the Doppler returns so far this year have been underwhelming. If we look at the latest planet-mass vs year of discovery diagram on exoplanet.eu (no pulsar planets, no microlenses), the detection rate seems to be holding up, but the crop of announced low-mass planets is nonexistent. Of the 22 new planets so far in ’08 that have been detected via radial velocity, 16 were initially detected by the transit surveys.

What’s up with that?

We’re seeing core accretion in action. The baseline prediction of the core accretion theory for giant planet formation is that once a planet reaches a crossover threshold, where the mass of gas and solids is equal, then rapid gas accretion ensues, and the planet grows very rapidly to Jovian size or even larger. When the galactic planetary census is complete, one thus expects a relative dearth of planets with masses in the range between ~20 and ~100 Earth masses. In the freewheelingly unrefereed forum of a blog post, I can go ahead and dispense with an analysis that takes all the thorny completion issues and selection biases into account and state unequivocally that:

(Courtesy as usual of the exoplanet.eu statistics plot generators)

Planets that do make the grade and blow up to truly Jovian size are the beneficiaries of protostellar disks that had solid surface densities that were well above the average. At a given disk mass, a disk with a higher metallicity has a higher surface density of solids, which is the reason for the planet-metallicity correlation. Disks with higher oxygen and silicon fractions relative to iron will also have high solid surface densities, which is the reason for the planet-silicon correlation. And M stars have trouble putting their Jovian cores together fast enough to get the gas while it’s still there, which is the source of the planet-stellar mass correlation.

As one pushes below Neptune-mass, these correlations should all get much weaker, and the fraction of producing stars should go way up. It’s hard, at the ~10% success rate level for a protostellar disk, to make a Jupiter, and it should be straightforward, at (I’ll guess) the 50% success level for a protostellar disk to make a Neptune.

The gap between Neptune and Saturn is the source of the current RV planet drought. At given velocity precision (in the absence of stellar jitter), it takes ~25x more velocities to detect a Neptune than to detect a Saturn. To make progress, it’s necessary to stop down the number of stars in the survey and focus on as many old, quiet K-type stars as possible. We’re talking HD 69830.

The indications at Harvard were that the Geneva group has been doing just that. In a few hours, Michel Mayor is scheduled to give the lead-off talk at the Nantes meeting on extrasolar super Earths. I’ll post a rundown of what he has to say just as soon as the Oklo foreign correspondents file their reports…

On Friday, I flew back from the Boston IAU meeting, still buzzing with excitement. On Saturday, I woke up with what might best be described as a transit-induced hangover (an entirely distinct condition from transit fever). I’d basically allowed all my professorial responsibilities to slide for a week. On my desk is a mountain of work, a preliminary exam to assemble, and a horrifying backlog of e-mail.

Ahh, but like an exotic sports car bought on credit, it was worth it. The meeting was amazing, certainly the most exciting conference that I’ve ever attended. Big ups to the organizers! Planetary transits are no longer the big deal of the future. They’re the big deal of the right here right now. Spitzer, Epoxi, MOST, HST and CoRoT are firing on all cylinders. The ground-based surveys are delivering bizarre worlds by the dozen. And we’re clearly in the midst of very rapid improvement of our understanding of the atmospheres and interiors of the planets that are being discovered.

From a long-term perspective, the conference’s biggest news was probably provided by the Geneva group, in the form of Christophe Lovis’ presentation on Tuesday afternoon. In his 15-minute talk to a packed auditorium, Lovis covered a lot of ground. I scrambled to take notes. My reconstructed summary (hopefully without major errors) runs like this:

The HARPS planet survey of solar-type stars contains ~400 non-active, slowly rotating FGK dwarfs. Observations with the 3.6-meter telescope have been ongoing since 2004, and over time, their emphasis has been progressively narrowed to focus on stars that harbor low-amplitude radial velocity variations with RMS residuals in the 0.5-2.0 m/s range. The current observing strategy is to obtain a nightly multiple-shot composite velocity of an in-play candidate during block campaigns that run for 7-10 nights.

During the first few minutes, Lovis reviewed the current status of the published results. The Mu Arae planets (including the hot Neptune on the 9.6-day orbit, see here and here) are all present and accounted for. The HD 69830 triple-Neptune data set (see here, here and here) now contains twice as many velocities, with virtually no changes to the masses and orbits of the three known planets. Long-term scatter in the HD 69830 data set is at the ~90 cm/sec level, indicating either the effect of residual stellar jitter, or perhaps the presence of additional as-yet uncharacterized bodies.

He then announced that there are currently forty-five additional candidate planets with Msin(i)<30 Earth masses, P<50 days and acceptable orbital solutions. And that’s not counting candidates orbiting red dwarfs.

He then began to highlight specific systems. To say that planets were flying thick and fast is an understatement. Here’s the verbatim text that I managed to type out while simultaneously attempting to focus on the talk:

Rumor has it that some of these systems will be officially unveiled at the upcoming Nantes meeting on Super Earths. Odds-on, with 45 candidates in play, we’ll soon be hearing about a transiting planet with a mass of order ten times Earth’s. I won’t be at the Nantes meeting, but the stands will be harboring agents of the Oklo Corporation.

The talk finished with an overview of the statistics of the warm Neptune population. Most strikingly, a full 80% of the candidates appear to belong to multiple planet systems, but cases of low-order mean motion resonance seem to be rare [as predicted --Ed.] . There is a concentration of these planets near the 10-day orbital period, and the mass function is growing toward lower masses. Significant eccentricities seem to be the rule. And finally, I think it was mentioned that the planet-metallicity correlation is weaker for the warm Neptunes than for the population of higher-mass planets.

Seems like core accretion is standing the test of time.

Note on the images: Gaspar Bakos (of HAT fame) had the cool idea of machining metal models for the planets of known radius which are correct in terms of relative size, and which have the actual density of their namesakes. HAT-P-6, for example, is constructed from a hollow aluminum shell, and with a density of ~0.6 gm/cc it would float like a boat. HAT-P-2b, on the other hand, which packs 8.6 Jupiter masses into less than a Jovian radius, has the density of lead and (not coincidently) is made out of lead. It’s startling to pick it up. CoRoT-Exo-3b, which was announced at the meeting, has a mass of twenty Jovian masses, and a radius just less than Jupiter. I guess that one will have to be made from Osmium.

Earth, at ~5.5 gm/cc, on the other hand, can be readily manufactured from a variety of different alloys.