Wednesday August 23, 2006

Defining Planet

If you've been keeping an eye on any of the wire services (or Language Log), you may have noticed that the members of the International Astronomical Union meeting in Prague have been wrangling, Vatican-style, over the definition of the word planet.  Judging by the periodic puffs of contradictorily-colored smoke the meeting is emitting, it sounds like they may be getting bogged down in the details.  (That's where the Devil is, you know.)  Well, defining words involves language, and language is what linguists study, so that means this is a linguistic problem.  Let's roll up our sleeves and see what all the trouble is.

As you probably don't recall, I'm on record supporting a definition similar to the initial IAU proposal from last week.  I wrote, "any object that's (a) large enough to pull itself into a spherical shape and (b) orbiting a star (or some kind of star-remnant) is good enough for me."  Last week's draft reportedly read:

A planet is a celestial body that (a) has sufficient mass for its self-gravity to overcome rigid body forces so that it assumes a hydrostatic equilibrium (nearly round) shape, and (b) is in orbit around a star, and is neither a star nor a satellite of a planet.

This is pretty close to my formulation (surely a coincidence!), but it includes the additional restriction that the celestial body is question must not be "...a satellite of a planet."

This first proposal was apparently voted down, and another definition proposed, whose first bullet-point read:

(1) A planet is a celestial body that (a) is by far the largest object in its local population[1], (b) has sufficient mass for its self-gravity to overcome rigid body forces so that it assumes a hydrostatic equilibrium (nearly round) shape [2], (c) does not produce energy by any nuclear fusion mechanism [3].

The bracketed numbers are footnotes, the most important of which is number [1]: "The local population is the collection of objects that cross or close approach the orbit of the body in consideration."

An article in today's New York Times describes a further proposal that would classify the celestial bodies in question into three categories, labeling the controversial ones "dwarf planets".  Still another idea would be to define a planet as a body large enough to hold an atmosphere (you can read the two papers mentioned here and here).

Having trouble keeping track?  Me too.  There are a lot of proposals, some of which partially overlap.  Let's tease apart the various criteria that are being proposed as tests for planethood and consider the edge cases they're designed to rule out.

Not a Star

Everyone seems to agree with this criterion.  Nobody wants to start calling stars planets, but it turns out there's a gray area—or rather, a brown area—between stars and planets, namely the brown dwarfs. Depending on how you squint at them, they're either super-duper-Jupiters or small failed stars.  It's not crystal clear to me how the various proposed definitions would deal with brown dwarfs, but everybody seems to agree it's the small bodies we're having trouble with, so let's not worry about the large ones.  (Much.)


This criterion is uncontroversial, too, with a little bit of qualification.  The problem is there's a lot of small, irregularly-shaped bits and pieces floating around the solar system (see belt, asteroid) and nobody wants to call most of them planets.  The nine traditional planets are a tremendously varied bunch of objects, but they are, at least, all more or less spherical.  Shape alone isn't quite sufficient, though, leading to the qualification.  Some fraction of asteroids and comets are presumably, by random chance, roughly spherical, so it's additionally required that a body "assumes a hydrostatic equilibrium"—that is, it's so big that its material can't withstand the pull of its own gravity and it collapses into a sphere.

This criterion can be hard to apply in practice.  It's not just a function of the size of a body, because different materials (and at different temperatures) have more resistance to mechanical strain, so just knowing a body's mass isn't sufficient.  On top of that, just how much deviation from spherical would be allowed might be hard to nail down.  Saturn's moon Iapetus is big enough that it is roughly spherical, but it has a distinctly irregular shape, with some flattened areas and a large equatorial ridge.  It's not in the running for planethood because it's a moon (see below), but if it were orbiting out by itself would we accept it as sufficiently round to be a planet?  Small or distant bodies present another problem: what if they're too far away to measure their shape accurately?  The possible tenth planet 2003 UB313 (whose name I'm tired of typing and am going to refer to as "Xena" until they give it a real name, come what may) is far enough away that its shape cannot, as yet, be directly observed.  The same was true of even the relatively nearby asteroid Ceres until recent observations showed that it is quite round.

Not Orbiting a Planet

This is where things start to get interesting.  There is general agreement that a moon is a different sort of a thing than a planet, but making the distinction is tricky.  It would be nice if all planets looked one way and all moons looked another, but that just isn't so, even in the tiny sample of celestial bodies we have in our solar system.  Moons are smaller, you say?  Ganymede, a moon of Jupiter, and Titan, a moon of Saturn, are both bigger than Mercury, which is incontrovertibly a planet.  (Hands Off Mercury!)  Moons are airless?  Titan has an atmosphere; Mercury doesn't.  The distinction between moons and planets doesn't have to do with composition, temperature, or any other physical property, it depends only on location, location, location: moons aren't planets, moons orbit planets.

Even that seemingly simple distinction is tricky.  Suppose Earth had a twin of equal or nearly equal size instead of our Moon.  We'd want to call that a double planet, right?  But wait, both of them are orbiting a planet, so they have to be moons according to this criterion.  But if they're both moons, then neither is orbiting a planet, my head hurts.  Such double planets aren't just hypothetical, either.  Pluto has a moon, Charon, that's so large relative to Pluto that they both orbit each other around an empty point in space, their center of mass or barycenter.

Orbiting a Star

Now things are getting even fuzzier.  The reason for requiring that a planet orbit a star is, I think, primarily to give us a way to distinguish planets from moons, but the criterion that a planet must orbit an object undergoing core fusion is tricky.  What about bodies orbiting dead supernova remnants like pulsars?  Some have been found; do we call them planets?  Suppose you had an object that everyone would agree was a planet if it were orbiting a star, except that it wasn't orbiting a star.  Suppose it's big, seven Jupiter masses—something that big must be a planet, right, even out in interstellar space?  In fact, it's not a hypothetical question.  There's a pair of objects about 400 light years away called Oph 162225-240515 (a.k.a. Oph1622) that are 14 and seven Jupiter masses respectively and don't orbit any star.

There's a term for such objects that are planet-sized but which some people might not consider actual, honest-to-God planets: planemos.  The term was first proposed by Gibor Basri in a 2003 paper about the definition of planet that you can read here.  Basri actually proposes a nice, simple system for naming celestial objects that's worth repeating here:

FUSOR - an object that achieves core fusion during its lifetime.
PLANEMO - a round non-fusor.
PLANET - a planemo orbiting a fusor.

Basri's system is straightforward, but I haven't seen it mentioned among the candidates in the current debate.  Oh, well, too bad—back to the criteria.


This criterion was mentioned in that article I mentioned above, although it's not clear whether it's been proposed to the IAU.  It's an interesting idea.  Its proponent, Bojan Pečnik, argues that planets should be distinguished by some physical property, and that hydrostatic equilibrium (roundness) isn't workable because of the difficulty of measuring it for distant objects like extrasolar planets.  He suggests that any object large enough to hold an atmosphere should be considered a planet.  This sounds immediately suspicious, since Mercury doesn't have an atmosphere (Hands Off Mercury!), but the article says it's just the ability to hold an atmosphere that matters, so Mercury would just squeak in.

Locally Largest

The most complex, and I suspect the most controversial, of the criteria proposed has to do not with the properties of object in question, but with other nearby objects.  This is the requirement, quoted above, that a body be "by far the largest object in its local population", where local population is defined as: "the collection of objects that cross or close approach the orbit of the body in consideration."  This criterion is specifically designed to distinguish between the Big Eight (Mercury through Neptune modulo the Asteroids) but to exclude Pluto and other outer solar system bodies.

The current understanding of the way planets form is that they coalesce out of a disk of material surrounding a young star.  At first, there are many, many small bodies called planetesimals in all sorts of intersecting orbits around the star.  Over time, they collide and form larger planetesimals, and as they get larger, the process tends to run away, since larger bodies have more gravity and so attract more material.  Eventually, the theory goes, you end up with a solar system like ours, in which most large bodies have swept out a "lane" all to themselves—Venus is always outside of the orbit of Mercury, Earth is always outside the orbit of Venus, and so forth.  The idea, then, is that since this is how planets form, any body that does not represent the end product of this process can't be a planet.

This criterion excludes the asteroids of the main belt, and in particular Ceres, the largest of them, because they occupy a set of similar orbits between Mars and Jupiter, and none of them is overwhelmingly larger than all the rest.  It excludes Pluto (and Charon) as well, since Pluto's orbit crosses that of the much larger Neptune.  (I'm not sure whether this criterion is also intended to exlude Pluto and Charon because they're orbiting each other.)  The other outer solar system bodies in the Kuiper Belt and Oort Cloud are similar excluded because none of them (at least, none yet discovered) are overwhelmingly larger than all the other bodies in similar orbits.  Since the material in those regions hasn't coalesced into a single body, goes the argument, there's nothing there we can call a planet.

Applying the Criteria

How, then, do each of these criteria divide up the celestial bodies in question?  Let's find out.  The table below contains a list of potential planetary bodies and the six criteria mentioned above, suitably abbreviated: !* = not a star, Round = roughly spherical, !Moon = not orbiting a planet, Around* = orbiting a star, Atmo = has an atmosphere, and MOID = locally largest (Master of Its Domain).  Bodies that meet a criterion are marked with an 'X', and questionable cases are marked with a '?'.

Celestial Body!*Round!MoonAround*AtmoMOID
Earth X X X X X X
Moon X X
Mercury X X X X ? X
Pluto X X ? ? ?
Charon X X ?? ?
Ceres X X X X
Vesta X ? X X
Oph1622 X X ? X X

Fortunately, Earth appears to be a planet by every criterion.  (Whew!)  Mercury is a little less certain; it's big enough to hold only a tiny trace of an atmosphere against the solar wind, but otherwise it's solid.  The Moon is right out.  All it has going for it is that it's round and not a star.  Pluto and Charon really are border cases.  Since they form a double planet, whether they meet the "not orbiting a planet" criterion depends on how you define that special case (although if only one of them is a planet, it's definitely the larger Pluto, hence the double question mark for Charon).  Pluto does have an atmosphere even though it's not large enough to retain one, because when it gets close to the sun some of its icy surface boils away and eventually escapes, like a huge comet.  Ironically, Ceres and the second-largest asteroid Vesta (which I haven't seen mentioned in any of the media accounts) both meet more criteria clearly than Pluto and Charon.  [Update: Some further reading, including an updated Wikipedia entry, implies that Vesta isn't in hydrostatic equilibrium, but rather assumed a round shape at some time in the past when it was partly molten, making it an interesting edge case.  Accordingly, I've changed Vesta's value in the Round column to a ? and updated the places where I mention it below.]  The oddballs Oph1622 (that pair of planemos out in interstellar space) also meet most of the criteria.

All of the proposed definitions of planet mentioned above can be described as a logical combination of these criteria.  The first proposal is (Round and Around* and !Moon and !*), which includes Earth, Mercury, Ceres, and possibly Vesta, Pluto, and Charon.  The second proposal is (MOID and Round and !*) and includes only Earth and Mercury.  Other proposals can be similarly expressed.

A Definition

So far, I've described the problem and talked about a bunch of the edge cases.  Now I'm going to argue for and against some of the criteria, but before I do, I want to warn you in the strongest possible terms: I Am Not An Astronomer, just an interested outsider and space-geek.

Over this last week I've been reading objections to various proposals.  First, it was objected that treating Pluto and Charon as a double planet was wrong, and it was pointed out that in a few billion years the Earth and the Moon will be far enough apart that their barycenter will be above the surface of the Earth.  Should we then call the Moon a separate planet?  Well, in a word, yes.  If the physical circumstances of the solar system change, I'd be willing to change terminology.  If the Earth weren't here, wouldn't the Moon qualify as a planet?  (Let alone what might happen if we started storing nuclear waste on the Moon and there was a huge explosion...)

The criterion of "largest in its local neighborhood", devised to include the Big Eight but exclude the asteroids and Pluto/Charon, seems overly complicated to me.  (I'm willing to accept the simple exception that moons aren't planets because being a moon somehow overrides being a planet.)  For example, it's true that Neptune is much larger than Pluto and that Pluto crosses its orbit (Pluto, like all the bodies known as Plutinos, is in a 2:3 orbital resonance with Neptune, so this arrangement is stable).  But suppose—I'm going to get hypothetical here, but remember that the definition of planet has to work for all celestial bodies everywhere, and it's a big universe—suppose Neptune were only Mercury-sized.  There's no reason two planets can't be in an orbital resonance.  Two of the three exoplanets around Gliese 876 are about 0.6 and 2 Jupiter masses in size and they're in a 1:2 resonance, but surely they're still planets.

Worse, there are a bunch more odd-but-possible orbital arrangements that would, under this criterion, disqualify bodies that should be considered planets.  Suppose there were a Mercury-sized body orbiting in one of Jupiter's Trojan points.  Their orbits cross and Jupiter is overwhelmingly larger, but does it seem reasonable to declare Trojan Mercury to be, what, a moon?  Before you answer, consider a slightly different circumstance: suppose there was another body the same size as the Earth in one of the two stable Earth-Sun Lagrangian points; would that disqualify both from planetary status?  What about two planet-sized bodies orbiting a star in two interacting orbits like Janus and Epimetheus (moons of Saturn) or like Earth and 3753 Cruithne?  The definition of planet has to be ready to accomodate oddities like these as extrasolar planets are discovered, or else exclude them on a more principled basis than "well, we don't have any planets like that in our solar system."

What's more, the criteria designed to exclude Pluto in particular (MOID and Atmo) seem to be too carefully tailored to the particular circumstances of our solar system rather than to a general consideration of what a body is.  If the giants planets were all no more than Earth-sized, so that Pluto wasn't so relatively small, would we be having this argument?  What if the planets decreased in size the further you got from the Sun?  It seems to me that any place with a resonably flat horizon and enough gravity to make me stick to it (or sink into it, in the case of the gas giants) is enough like a place that it deserves to be called a planet.

So, if I were a Professional Astronomer instead of just some goofball, here is the definition I'd propose:

A planet is a celestial body (a) that is not a star or stellar remnant, (b) that is not orbiting another planet, and (c) that is large enough for self-gravity to pull it into a roughly spherical shape.

This is different from the definition I suggested a while ago in that post about Xena. I've added the criterion that excludes moons (a case I'd frankly forgotten about) and removed the one about orbiting a star—if  a body is wandering around in the cold, cold of interstellar space, taking away its status as a planet seems like adding insult to injury.  This definition is very close to the initial proposal at the IAU meeting that was voted down last week.  It would certainly include Ceres, and Xena, and possibly Vesta depending on whether we consider it round or not, and Pluto and Charon if we can define the "not a moon" criterion non-circularly, which involves being able to decide what an object is "orbiting".  To address that difficulty, here's an algorithm for deciding what a candidate body CB "orbits" for the purposes of the definition:

  1. Make a list of all other bodies such that CB passes through 360 degrees of sky as seen from that body.  This is the list of bodies that CB revolves around or that revolve around CB.
  2. Consider the bodies starting with the closest.  For each body, is the center of rotation of that body and CB below the surface of the body?  If so, CB "orbits" the body.  If not, discard the body and continue.
  3. If there are no bodies left, CB doesn't "orbit" anything.

If this works as intended, moons would be considered to be orbiting their primaries, Pluto and Charon would be considered to be orbiting the Sun (which is why "orbit" is in quotes above, since by any reasonable definition they're orbiting each other), and wandering interstellar planemos would be orbiting nothing (or possible the huge black hole at the center of the galaxy).  I have to say, this is the criterion that I'm least wedded to.  Excluding Pluto and any other large outer solar system objects entirely strikes me as arbitrary, but the difference between double planet and planet-and-moon really is a fine hair to split, and I could go either way.

So, assuming the above procedure for deciding what's a moon and what's not, I would include the following as planets: Mercury, Venus, Earth, Mars, (possibly) Vesta, Ceres, Jupiter, Saturn, Uranus, Neptune, Pluto and Charon, Xena, and any other spherical outer solar system bodies not orbiting another.  The more we find, the more there are.  That's not a problem—that's discovery.

Welcome to the solar system.  I can't wait to visit.

[Yipes, I'd better post this before the professionals vote tomorrow and make it all moot.]

I am The Tensor, and I approve this post.
10:54 PM in News , Science | Submit: | Links:


TrackBack URL for this entry:

Listed below are links to weblogs that reference Defining Planet:


Interesting that you should quote the saying "the devil is in the details" in a post concerning the status of a planet named after the god of the Underworld.

Posted by: Pete Bleackley at Aug 24, 2006 1:13:55 AM

Yay! Thanks for agreeing with me that finding lots of new planets is NOT a bad thing ... If defining planet in such a way that new planets might be discovered, then the solar system gets bigger and more complicated and more exciting. Over at the Language Log Geoff Pullum just lumped those who want Pluto to be a planet in with those who want all the "rules" they learned to be inviolable: "For people who want to make sure the material they were taught in elementary school and high school stays unchanged forever, the path is clear: stay away from all intellectual activity, avoid contact with anyone who is intellectually curious, live a dull and unexamined life." I object - I think it's perfectly possible to want Pluto to be a planet without your motivation being "that's what I learned in school!" After all, if we want Ceres and Vesta to be planets, too, that's not an "unchanged" solar system...

I think we'll be revising our definition a lot if we try to tailor it to our solar system to meet any agenda. If we want to say there are several types of planets, why not? There are types of stars, after all. We could have "inner system small rocky planets" and "outer system small icy planets" as well as gas giants and rocks with atmosphere, right? Why not?

Posted by: The Ridger at Aug 24, 2006 2:24:23 AM

If four moons of Jupiter was good enough for Galileo, we shouldn't be futzing about with these new-fangled teleoscopes discovering things we weren't meant to discover, and confusing the wee bairns by crossing out bits in their textbooks.

Posted by: nw at Aug 24, 2006 6:14:42 AM

First, there are no equations that depend on the definition of the word "planet". The word in isolation has no physical meaning. The definition is no sort of science other than linguistics. That means that this whole argument is the worst sort of prescriptivism.

In my dialect (or perhaps "current idiolect"), "planet" includes Pluto. I believe that this is a very common (perhaps even majority) opinion among native-speakers of English, which should put the entire argument to rest.

"If defining planet in such a way that new planets might be discovered, then the solar system gets bigger and more complicated and more exciting."

No. The definition of a word changes the solar system in no way. The only importance attaching to the definition is what ends up in lists of planets that school children must memorize.

FWIW, Pullum seems awfully wedded to idiosyncratic definitions of words and the ridicule of those who disagree. (See also his insistence that "hyperbole" means only what he says it means when "about a million" people disagree.)

Posted by: Doug Sundseth at Aug 24, 2006 11:22:03 AM

And among these other indefeasible facts of English usage that make Pluto a planet so there, are the fact that Quaoar, Sedna, and UB313 weren't known when I was a boy, so can't ever be included for consideration in anything. Some arbitrary state of scientific discovery between 1930 and 1990 is more important than the possibility that understanding of the world can change and our language with it. _Your_ attitude, Doug Sundseth, is that of the hardened prescriptivists who learnt something in school from people who knew little more, and will never abandon it.

Posted by: x at Aug 25, 2006 12:27:44 AM

I wish the IAU had read this post before voting. There are several definitions by which I would have no trouble accepting Pluto's demotion, but "has cleared the neighbourhood around its orbit" is way too vague. Maybe it's a better criterion than "If I learned it was a planet in school, it's a planet," but it's not much better.

Now we can move on to discuss similarly important topics like "Is Antarctica a continent? What about Greenland? Europe?"

Posted by: Big Ben at Aug 25, 2006 1:07:47 AM

My reaction to "has cleared the neighbourhood around its orbit" is, "You mean, like Neptune did?" Very poorly worded.

It's interesting that this is being spun as a declaration that Pluto isn't a planet. After all, they did declare it, along with Ceres and Xena (but not Vesta!), to be dwarf planets. I'd call that a refinement rather than an outright rejection -- in other words, just the sort of hair-splitting that satisfies no one.

The good news is, it looks like they didn't pass the awful "classical planet" amendment. The "classical" planets are surely the ones visible to the naked eye that have been known since antiquity -- Mercury, Venus, Mars, Jupiter, and Saturn, and maybe the Sun and Moon if you're feeling particularly old-school.

Posted by: The Tensor at Aug 25, 2006 4:19:03 AM

"_Your_ attitude, Doug Sundseth, is that of the hardened prescriptivists who learnt something in school from people who knew little more, and will never abandon it."

Nope. I don't require that your dialect or current idiolect includes "Pluto" in the definition of "planet". This is an attempt to change usage by "learned" decree in a case where nothing significant depends on the definition used. (This differs from the definition of "second" as a measure of time, for instance, where many things depend critically on the definition.)

The definition of "prescriptivism" that I am familiar with would include the disallowing as illegitimate a word-definition in common usage. Perhaps the definition is different in your current idiolect; please feel free to enlighten me.

Posted by: Doug Sundseth at Aug 25, 2006 8:07:12 AM

It's interesting how hard it is even to define the term "orbits". I think your definition is a pretty good one, even if it means the Pluto and Charon both orbit the sun.
If you haven't thought about it much, I think the most obvious initial suggestion is that a body orbits whatever other body attracts it most strongly. But that has all sorts of problems. By that definition the Moon is not a satellite of the Earth, since the gravitational pull of the Sun on the Moon is 2.2 times stronger than Earth's. Pluto orbits Charon AND Charon orbits Pluto.
I wrote some orbit simulation software recently, so getting this definition right was a real practical problem for me.

Posted by: Mark Naught at Aug 26, 2006 6:28:33 AM

I heard an Australian astronomer recently claim that nailing down a good definition of 'planet' will (paraphrasing) lead to a better understanding of them, which will in turn lead to the discovery of previously unknown planets/planemos/satelites/stars/celestial bodies.

Now honestly, since when did arbitrary definition affect the referent in such a way as to make them more easily detectable?! The same goes for identifying and labelling grammatical phenomena. To quote my favourite linguist: "You can call it whatever you want. Call it 'Fred' if you like, you won't change what it actually does," (refering to, I don't know, post-lexical argument coalition or, coindexation). Anyway, my point is that you can define it any way you like, but what is the point?

Besides, all these criteria of different definitions seem to be non-discrete variables. Orbit, for example, is a function of relative size and distance of other bodies, as per the Newton's (sufficiently good) law of universal gravitation. A planet, such as Earth, may orbit the sun and still orbit its moon, in fact it may technically orbit everything in the universe, but most of those gravitational interactions aren't very salient. How then can a continuous variable be used in constructing a discrete definition (I assume 'planethood' is supposed to be discrete)?

I am happy with some sort of continuum of celestial bodies, starting with, well, I suppose, stars, followd by brown dwarfs, other fusors, planemos, planets, twin planets, satelites, bloody big boulders, meteors/comets, etc.

Posted by: Aidhoss at Aug 26, 2006 8:39:25 PM