What is a Planet?

Gibor Basri, Astronomy Dept., UC Berkeley

 

Abstract


A spate of new discoveries of objects in the mass range 0.5-15 Jupiter masses (jupiters), along with the Pluto controversy, has caused increasing debate over the meaning of the word "planet" (as differentiated from "stellar or substellar" objects). A variety of uses for this word have occurred, each with a somewhat different intent. No formal definition has been officially adopted. I examine the problem in the light of current astrophysical observations and theory. The debate can be framed in terms of three distinct arenas that drive it: characteristics, circumstance, and cosmogony. By "characteristics" I mean the physical properties of an object - primarily its means of support against gravity and its source of luminosity. By "circumstance" I mean the environment in which it is found, most importantly whether it is in orbit, what primary object and other companions are present, and their orbital characteristics. By "cosmogony" I mean the origin or mode of formation of the object. The current lack of consensus derives from differing weights being applied to these three arenas in forming the definition of "planet". Each arena contains dividing lines between "planets" and "stars" which are partly matters of opinion, and end up sorting objects differently. The main purpose of this paper is to bring order to the discussion, so that we may eventually find a broadly accepted consensus definition. I cannot resist, however, offering a possible solution that steers a middle path through the various quandaries that come up in attempting a formal definition.

"When I use a word", Humpty-Dumpty said in a rather scornful tone, "it means just what I choose it to mean neither more nor less".
"The question is", said Alice, "whether you can make words mean so many different things".
"The question is", said Humpty-Dumpty, "which is to be master that's all."
Lewis Carroll in Alice Through the Looking Glass (1872)
 

I. Introduction

Until 1991, the only planets we knew were the 9 objects given that appellation, which comprise our Solar System. These all lie in a disk around our Sun, with small rocky objects near it, and larger gaseous objects further out (Pluto is an anomaly). Since then, however, we have discovered many substellar objects in the solar neighborhood, ranging in mass from as perhaps as low as Saturn up to the bottom of the main sequence of stars. Some of these are in orbit about other stars and appear to have masses similar to Jupiter, while others are different in various ways, and some are clearly brown dwarfs. We have begun to determine the numbers of objects found with various characteristics and circumstances. During the same period, our understanding of the star formation process has moved rapidly forward. We now understand through observations and theory that circumstellar disks are a natural and common early part of this process. At the same time, the configuration of the extrasolar planetary systems has challenged pre-discovery theories of planet formation, making us realize that Nature (as always) is more inventive than suggested by a single example (our Solar System). A good recent summary of relevant information can be found in Protostars and Protoplanets IV (Univ. of Arizona Press, 2000).
The original meaning of the English word "planet" is rooted in Greek and means "wanderer" (it would be interesting to include other cultures and languages). This refers to the apparent positions of the objects in the sky from Earth, and so is not much use as an astrophysical definition. After examining a number of definitions from dictionaries and encyclopedias, a fair summary of the "cultural" definition of "planet" seems to be something like "an object resembling the planets in the Solar System, which is orbiting a star". The problem with this definition is that it is not very specific about how closely an object must resemble one of our planets, or in what ways, and it does not take into account the recent discoveries bearing on this question.
There have been a number of announcements of new objects that stretch the use of the word "planet". There are the many announcements of "exoplanets" in orbit about solar-type stars. These, especially considering that their published masses are lower limits, extend up to the mass needed for the onset of deuterium fusion. Whether that is the appropriate upper mass limit for planets depends on the definition of "planets" as well as "brown dwarfs". The big surprise in these systems is the presence of objects with masses up to several jupiters in the "inner" system, sometimes just above the star. Furthermore, they are mostly in reasonably eccentric orbits, counter to the expectation engendered by the Solar System. There have also been several recent announcements of "free-floating planets" (called that because they are inferred to be below the deuterium-burning limit) in young stellar clusters. Papers on brown dwarfs have also sometimes referred to objects below this limit as "planets", without regard to any other properties.
These pronouncements have been accompanied by objections to the terminology employed. It has been retorted that since the free-floating objects are not orbiting stars, they do not conform to the cultural meaning of "planet". It is further suggested they should be given some sort of "stellar" name under the presumption that they formed by themselves in a "star-like" fashion. There has been a suggestion that most of the exoplanets deserve similar treatment, because their eccentric orbits are presumed to suggest that they also didn't form the way Jupiter did. Recently it was proposed that most exoplanets are really brown dwarfs, because the lower limits from Doppler searches consistently and substantially underestimate their true mass. If this were actually the case (which their mass-limit distribution makes highly unlikely), it would be a sound reason for not calling them "exoplanets".
In this paper, I examine each of the three arenas (characteristics, circumstances, and cosmogony) separately. For each I summarize the state of our current theories and observations. I try to separate what is "known" from what is conjectured, or just a matter of opinion. I try to show how each might be used to construct the definition of "planet", and on what basis one would separate planets from stars in each arena. I begin with a table that organizes all the pieces one might use, and invite the reader to see how your predilections would put them together in defining "planet". I then provide more detailed reasoning about why the table looks as it does, and justifying my proposed definition.

Table I. The Defining Arenas for Planets

I. Characteristics
     A. Source of Internal Pressure (inferred from well-accepted models)
          1. Coulomb forces (free electron, crystalline, or liquid); ideal gas law
          2. Free electron degeneracy
          3. Thermal pressure, supplied by fusion, due to gravity
     B. Source of Luminosity (inferred from well-accepted models)
          1. Internal Heat Capacity (and radioactive decay)
          2. Gravitational Contraction
          3. Fusion of Deuterium
          4. Fusion of Hydrogen
II. Circumstance
     A. Object is in Orbit (observed)
          1. In circular orbit around a main sequence star
          2. In any orbit around an object capable of fusion (fusor)
          3. In any orbit around a more massive object
     B. Object is in a system (observed or observable)
          1. Object is in unique, non-crossing orbit, dynamically cleared
          2. Other similar objects are in similar orbits
    C. Object is not in orbit (observed)
         1. Object was never in orbit (difficult to ascertain)
         2. Object used to be in orbit, but was ejected (difficult to ascertain)
III. Cosmogony
    A. Formed in a disk around a fusor (depends on developing models)
         1. Built up by agglomeration of planetesimals (difficult to ascertain)
         2. Gas added by accretion onto rock/ice planetary core (difficult to ascertain)
         3. Formed directly by gravitational instability in disk (difficult to ascertain)
    B. Formed in a disk as a result of  gravitational perturbations
         1. Perturbations were due to other companions (ascertainable?)
         2. Perturbations were due to passing external bodies (difficult to ascertain)
    C. Formed at the center of an isolated disk (observable)
 

Proposed Definitions:

"A planet is a spherical non-fusor which is born in orbit around a fusor."

"A fusor is an object capable of core fusion at some time during its life."
 

II. Characteristics


One of the most basic characteristics of astrophysical objects is their mass. It is mass which determines their structure and evolution, and often serves as the basis for classification. Does mass make sense as a basis for defining planets? It certainly plays a role in one of the few points on which there seems to be complete consensus: if an object has core conditions capable of generating nuclear fusion (which depends on its mass), it is not a planet. The proposition that we should therefore define planets as spherical objects incapable of fusion, however, has been made but enjoys no such consensus. The role of mass in the characteristics of objects is important here in two contexts: determining the dominant pressure-support mechanism in the interior, and setting the sources of luminosity (all opaque objects emit thermal radiation). These two ways do not, unfortunately, boil down to the same thing, so a mass-based definition has to combine them somehow. It also probably makes sense to continue to tell young children that planets shine by reflected light while stars make their own.
The structural divisions are fairly straightforward and sensible. Most stars (especially after their pre-main sequence phase) derive their internal support from thermal pressure. The heat may be generated by fusion, or by gravitational collapse. Of course, without support the object will just collapse further and heat up, so one might view core fusion as a strategy for keeping the core relatively cool (I am indebted to John Faulkner for pointing this out to me). As one moves to the lowest mass stars, the core densities become high enough that free electron degeneracy provides increasing pressure. It is the eventual dominance of this source of pressure that relieves the stars from "cooling" themselves by fusion, and causes the substellar realm to begin (at about 75 jupiters, depending on metallicity). Degeneracy support extends below this limit down to about 2 jupiters. At this point the internal density and pressure has decreased again enough (due to the low mass of the object), that ordinary Coulomb forces can begin to support the object. An obvious difference between "degenerate" and "ordinary" objects is that adding mass to the former make them smaller, while adding mass to the latter make them larger. All of the Solar System planets fall into the "ordinary" category.
The luminosity divisions are similar, but not identical. One problem here is that objects change their luminosity sources as they evolve, so it is harder to put an object into one bin. Almost everything is first luminous due to gravitational release of energy (by accretion and contraction). If fusion can occur, it will first occur for deuterium, then hydrogen. Main sequence stars (especially low mass ones) can burn hydrogen for a very long time. Hydrogen burning can occur in objects down to 60 jupiters (well into the substellar regime). Deuterium is much less abundant, and can only burn for a short time in any object. This occurs in objects down to 13 jupiters, early in their lives. All objects in the substellar regime derive most of their time-integrated luminosity from gravitational contraction. When an object is finally neither fusing nor contracting, its emission comes only from trapped heat inside. This happens fairly soon for rock/ice planets. Jupiter and Saturn are still contracting slightly. Although legal definitions would be hard to formulate, astronomers have been  comfortable with the following categories. A "star" is an object that burns hydrogen for a "long" time, during which its luminosity is practically stable and derives exclusively from that (i.e. it achieves the main sequence). If hydrogen fusion is never the sole source of luminosity and the object continually fades, it is a (substellar) "brown dwarf". It is still a brown dwarf if it doesn't burn hydrogen at all, but only deuterium. Most would agree that the lower mass limit of brown dwarfs occurs at the deuterium limit.
At this point, cosmogony is often thrown into the mix. The deuterium boundary was originally thought to be close to the lower limit for isolated star formation as well, but that is no longer clear. Some have recently proposed that degenerate non-fusors are still brown dwarfs (or maybe "grey dwarfs" to differentiate them) if they form the same way as fusing brown dwarfs. Others would like to call such objects sub-substellar (with various names proposed), and still others would like to call them planets (without regard to cosmogony or circumstance). It has been suggested that since deuterium fusion is not really a significant long-term source of luminosity, distinctions based on it are forced. Astrophysically that is reasonable, although it is often the case that objects very near the boundary of classes are harder to differentiate than those well away from the boundary. But the importance of the fusion/non-fusion boundary is another example of a strong "cultural" influence (that stars have fusion and planets don't). Because of that, it may make sense to coin new terms, say "fusor" and "non-fusor", to make this distinction explicit (without additional baggage). Such terms would take much of the heat out of several of the current disagreements.
We can see that defining objects purely on the basis of mass has its problems. It would be nice if the degeneracy boundary coincided with fusion boundaries at both ends, but it does not. There is a mismatch at high masses between the end of stable hydrogen burning (the main sequence), and the actual hydrogen-burning limit (the latter term is often confused with the former condition). The problematic objects for our purposes are degenerate non-fusors. As regards the definition of "planet", of course, the other question is whether one is willing to use a purely mass-based definition, or must consider cosmogony and circumstance as well. Finally, to do something about the low mass limit for planets, the best choice seems to be a requirement that they have sufficient mass for gravity to force a spherical shape.
 

III. Circumstance


I pointed out in the Introduction that the cultural definition of planet generally includes the necessity of being "in orbit about a star". We can therefore agree that if "circumstance" is to be included in the formal definition, this is a minimal requirement. Our Solar System suggests that planets should furthermore be found in circular orbits, and this was thought to be a natural consequence of the fact that planets form in disks. Indeed, as the first extrasolar planets were found, their eccentric orbits caused some to doubt that they are really "planets". As more systems were found, however, they inspired theoretical work that makes it likely that planets can indeed inhabit eccentric orbits (even planets which formed just as Jupiter did). We realized that if one has several massive planets, it is quite easy for them to disturb each other's orbits. We further realized that disks can move planets from their birth orbits (in the most extreme cases, dumping them into the star itself). Interactions between the planet and disk can also damp or excite eccentricity, depending on the particulars. Thus, even if planets built from planetesimals really must form in circular orbits, that is not how we may find them. Furthermore, planet formation may be influenced by mergers with smaller planets, or may occur instead by direct formation through disk instabilities (see section on cosmogony); either of which might generate eccentric orbits. It now seems unwise to use a circumstance based on orbital eccentricity in a definition of "planet".
The next question is what sort of object a planet must be orbiting. This seems relatively uncontroversial; it is generally agreed that the object being orbited should be a fusor. My experience is that few raise objections if it is a brown dwarf (so long as that is defined so only fusors are brown dwarfs). The only uncertainty comes if the "central" object is a degenerate non-fusor. Both objects may be like that, or at a minimum the center of mass may be well outside either of them. Such a configuration may be hard (especially if the secondary is too massive) to differentiate from a "binary planet", so it seems safer to demand that the primary be a fusor.
The object may not be the only one in orbit about the central fusor. Indeed, it is more convincing to have several objects in a system in "planetary" orbits (meaning they at least vaguely resemble those in the Solar System), since then the system more closely resembles ours. A single object in an eccentric orbit is uncomfortably similar to a binary stellar system (though this may just reflect our Solar System bias). Nonetheless, if all the objects in the Solar System were removed, leaving only the Sun and Jupiter, it is unlikely that anyone would wish to revoke Jupiter's planetary status.
On the other hand, having too many other similar objects in similar orbits has already been fatal to an object's planetary status. Ceres was originally designated a planet (and found in the position anticipated by Bode's Law, between Mars and Jupiter). It lost its status when several objects somewhat smaller than it, and in similar orbits, were discovered. Precisely the same situation has now arisen with Pluto. If we were historically consistent, there would be no question about the demotion of Pluto to "minor planet". In Pluto's case, another strike against it is that it crosses inside Neptune's orbit. If "major" planets do that, they will inevitably perturb each other into new orbits. Stable planetary systems must have all the major planets in non-crossing orbits (by the very definition of "stable"). It is a matter of taste whether one wants to include this requirement in a definition of "planet", but it seems to make sense to call small objects in crossing orbits "minor planets" by default.
Finally, we must deal with an object which is not in orbit about a more massive object at all. Given the cultural understanding of "planet" it would seem at first that, in a discussion of circumstance, this case cannot be included in the definition of "planet". We must, however, consider the instance where an object that is born as an acceptable planet is later ejected from the system. Turning the reasoning in the previous paragraph around, if Jupiter were suddenly lost from the Solar System, would we then consider it no longer a planet? That is again a matter of opinion, but I doubt that many would consider ejection grounds to revoke its status. It should be pointed out that smaller planets are even more likely to be thrown out (witness the vast number of Oort cloud comets). An ejected Earth would certainly still be considered a planet (what else could it be?). An attractive option is to append an adjective to reflect the new situation, such as "ejected" or "free-floating", to planet. The problem here is in determining whether a free-floating object has a history like this, or was formed without a primary. If the latter is true, a clear majority prefers not to call the object a planet. Those who would like to do so don't want circumstance considered in the definition at all, and insist on a definition based purely on characteristics.
 

IV. Cosmogony

One of the most fundamental properties of our Solar System, the one which best evokes its origin, is the fact that it is arrayed in a disk around the Sun, with all the major angular momentum vectors in (almost) the same direction. This allowed Laplace to state his "nebular hypothesis", the modern form of which is now universally accepted. This scenario is strongly supported by the ample evidence of commonly occurring disks around newly forming stars. These disks have been shown to often contain sufficient material to produce a system like ours (at least in principle). The theory of star formation, furthermore, demands that such disks should form early in the construction of the star; it is a fundamental property of the molecular cloud cores which produce stars that they have far more angular momentum than we see ends up in stars.
Thus, it is often stated that a fundamental property of planets is that they "formed in a disk". This is offered in contra-distinction to "stars" which form by "direct collapse of a molecular cloud core". I would first like to argue that this is an aspect of cosmogony that is not useful as commonly stated (I will replace it with a useful statement in the same spirit afterwards). The fact is that stars form in disks as inevitably as planets. One may immediately object that even if that is so, stars form at the centers of disks but planets form in the disks around stars. This viewpoint is misleading as well. The majority of stars actually form in multiple systems. In that case, there is typically a circumstellar disk centered on each star. When the separation of the stars is on solar system dimensions, there is often a circumbinary disk as well. From the point of view of the circumbinary disk, neither star is at its center (the center of mass is).
This may still seem like quibbling, until one goes through the exercise of shrinking one of the stars (increasing the mass ratio of the binary). Now there is a circumstellar disk around the primary, whose size is determined by the binary separation. There is one around the secondary, which decreases in size and mass along with the mass of the secondary. And there is a circumbinary disk that increases in importance relative to the primary's disk as the binary separation decreases. Eventually, the secondary and its disk become small enough compared to the primary and circumbinary disk that we begin to speak of a purely circumstellar disk around the primary, with a "gap" in it at the location of the secondary, and a small secondary disk. There is a smooth changeover from one point of view to the other, with no qualitative difference between them. In fact, Jupiter seems to have had a disk around it (as evidenced by the compositional gradient of the Galilean moons with distance from Jupiter), and so is equivalent to a very small secondary in a 5 AU binary system, at least from the point of view of whether it "formed in a disk".  With low mass fusor primaries and high mass non-fusor secondaries, the configuration looks even more like a stellar binary.
We can save the conceptual importance of disks to planet formation, while evading the above criticism, if we look a little deeper. The principle behind using "formation in a disk" as a criterion for planets should not concentrate on the disk itself, but the way in which the disk makes planets. A much more powerful and justifiable distinction to be drawn between the formation of planets and stars could be a requirement that planets form through mechanisms which require the prior formation of planetesimals. These might be defined as macroscopic  bodies too small to be forced into spheres by their own gravity, which are composed of solids, liquids, and/or ices. Current theory supposes that one will not produce such bodies except as a consequence of settling and agglomeration of dust and frozen gas in a disk (though even that is not indisputable).
For all we know, such bodies may form at the early core of an incipient star as well, but they are not required for its subsequent formation. They certainly seem to be required for the construction of the terrestrial planets, and also for any spherical icy bodies further out (which can include planets, moons, and other cases). The current favored theory for the formation of the gas giant planets in our Solar System posits as an initial step the construction of a rock/ice core of 10-15 earth masses, which would be built by planetesimal/planet mergers. The predominant gas component of Jupiter and Saturn can then be accreted directly from the gas disk by this core. Uranus and Neptune appear to be cases where the core is present but the gas accretion process was terminated before it went as far, perhaps by the disappearance of the gas disk while they formed (more slowly). It was thought that this mode of origin must result in circular orbits, but we now realize that they need not stay that way (see the section on "circumstance"). It was also thought that this mode of formation can only operate beyond the "ice boundary" where icy planetesimals are numerous; it is unclear in light of the exoplanets whether that idea is still valid (though it might be if their current orbits are greatly modified from birth).
Originally it was thought that this mode of formation could not produce objects greater than about 2 jupiters, because of the large tidal gap they open in the disk. If this is the case, then there is a natural congruence between the lower mass limit for degenerate objects and a formation modal boundary. Lately, however, it has become clear that accretion can occur across even the gaping tidal gaps in stellar binary systems. It is currently unclear how large an object the runaway gas accretion mode can produce.
It has now been proposed that there may be a third mode of planet formation. Due to gravitational instabilities in a sufficiently dense and cold disk, one might force disk material (both gas and dust, and any nearby planetesimals) to collapse directly into an object whose mass is a few jupiters (the mass limits for this process are not known). This might occur by itself in the right kind of mixed disk, or it might be promoted by the settling of dust to the midplane within a gas disk. The result could be what most people would be comfortable calling "gas giants", as well as more massive objects (including most of the current "exoplanets"), up to what people would be comfortable calling "brown dwarfs". It is unclear whether this is also the mode of formation for (some) stellar binaries; that has to do with whether the "fragmentation" of the original material occurs before or after it is forced into a disk (which we don't really know). It is possible that disk instabilities work for both fusors and non-fusors. There may or may not be a good boundary based on cosmogenesis near in mass to the lower fusion boundary.
If disk instabilities are the mode of formation for the inner massive exoplanets, it might explain why they have orbits that more closely resemble stellar binary systems. The question is whether the objects this mode produces should be called "planets". That is a matter of opinion, but it is interesting to note that the rock/ice core we expect to find in Jupiter seems to be absent, down to a current upper limit of 5 earth masses. Those who would like to exclude a "direct" mode of formation for "planets" may end up having to argue that Jupiter is not really a planet (or, more likely, having to change their definition).
The justification for excluding directly collapsed objects is also more problematic now that a system in which there are several massive objects arrayed in "planetary" orbits around a star has been found (HD 168443). We know of no stellar multiple systems that look like it. Stellar triples are all hierarchical, with a close inner pair and a distant companion. At first glance one might also claim this one is too (the outer orbit is 10 times further than the inner orbit). But in stellar cases, the outer object is always the least massive. Furthermore, since the massive fusor companion is at 3 AU, it is hard to imagine that the inner (probably non-fusor) companion's core formed from the agglomeration of icy planetesimals, or obtained most of its (quite substantial) mass by late gas accretion in its very restricted inner chunk of the disk. With such massive objects, it is unlikely that disk accretion could shrink their orbits a great deal, and also difficult to arrange a transfer of angular momentum to other objects to get them both where they are.
At this point, it probably only makes sense to insist that directly collapsed objects are not planets as a matter of taste: they are not planets because one does not wish them to be. However, there is no general agreement that the circumstantial or theoretical case for directly collapsing planets is fully made, nor that the direct mechanism can produce objects as light as Jupiter. This will clearly be a fertile area for theorists and observers  for the next decade or two. Perhaps the most telling point is that almost nobody would care which mode it turns out Jupiter formed with; we would still insist it is a planet!
Finally there is the question of the "free-floating planets". There are two possibilities for their formation. One is that they formed in planetary systems around stars (by whatever mechanisms are allowed to planets), and were subsequently ejected from the system by interaction with other massive bodies in the system (these could be planets or stars). Here the cosmogony question conflicts with the "circumstance" criterion; does a "properly generated" planet cease being a planet because of its circumstances?
The other possibility is that these objects formed in isolation, or at least were not originally bound to a star. If they form at the center of their own disk, and there is no circumbinary disk encompassing them and a fusor, most would agree that their cosmogony was "star-like". A majority takes this as sufficient reason to exclude such objects from the class of planets. Even here, however, the situation is somewhat muddled. Suppose (as already happens in numerical simulations) a binary star system is forming. The interaction between the stars, their circumstellar disks, and the circumbinary disk, produces various instabilities, shocks, and pile-ups in or between the disks. One of these collapses to form a non-fusor surrounded by its own little disk. It is not in a stable orbit, and is eventually ejected from the stellar system. Is this a planet? Does it matter whether it was first in one of the circumstellar disks, in the circumbinary disk, or had no real identifiable locus of "origin"? This problem disappears if "direct" objects are not allowed to be called planets, but otherwise it is rather confusing. Perhaps we should call such objects "confusors" (just kidding).
The problem is, as it was with runaway gas accretion onto a core, that it is very difficult to determine observationally what the cosmogony of these free-floating objects was. Whether they form in isolation or are ejected from a system leaves few lasting or easily observed traces. It is possible in principle that their current motion might provide such information; a young object seen in a cluster with much greater than the cluster escape velocity would be the best case. There are no current examples of objects that are known to have been ejected from planetary systems. We also have only glimmerings of the mass function for non-fusors that form "like stars". It is unclear what their minimum mass might be, what the true number of such objects is, or how it compares with the number of ejected objects.
We thus come to the primary difficulty with using cosmogony at all as a basis for defining planets. The truth is that we don't really know very much about the details of planet formation, and have rather few observational diagnostics for it. The situation is not entirely in hand even in the Solar System, and certainly not for the increasing zoo of extrasolar planetary systems and other substellar objects. Without an orbiting spacecraft, it is extremely difficult to ascertain the presence or absence of a rock/ice core in a massive gas giant (even with a spacecraft its tough). This problem gets harder the greater the mass of gas that overlies the core. We also no longer feel able to argue that the semimajor axis or eccentricity of an orbit is a firm piece of evidence for the cosmogony of the object; it is clear that orbits can be modified. Part of the problem has been that without firm theories or confirming observations, the discussion on cosmogony comes down on the wrong side of the border between science and religion (and has some of the heat inherent in that as well). I feel it is currently unproductive to use it as even a partial basis for a definition of planets, yet am finally compelled to include a bit of it for "cultural" reasons.
 

V. A Proposed Definition for "Planet"


"That's a great deal to make one word mean", said Alice.
"When I make a word do a lot of work like that", said Humpty-Dumpty, "I always pay it extra."
Lewis Carroll in Alice Through the Looking Glass (1872)

Given all the above considerations, it seems difficult to compose a definition that would satisfy the many conflicting constraints (cultural and astrophysical), deal with the three arenas, and encompass our new knowledge. But here is an attempt that may qualify (without even having to pay the word exorbitant wages):
"A planet is an object that is spherical due to its own gravity, that is never capable of core fusion, and which is formed in orbit around an object in which core fusion occurs at some time". Or more succinctly: "A planet is a spherical non-fusor born in orbit around a fusor".
This captures the two clearest cultural imperatives (that a planet be in orbit, and not be a fusor). It avoids the difficulties associated with our ignorance (both observationally and theoretically) of cosmogenetic issues, and allows planets to form as they will. There is increasing evidence from the current mass distribution of "exoplanets" that a different formation mode operates below the deuterium-burning limit (or something close to it) than for more massive substellar objects, at least in proximity to solar-type stars. This definition does not conflict with that evidence. It is not so specific that it excludes any of the current objects that most would agree are planets. The definition does not include non-fusors that form "like stars" by themselves (but it is fair to say that such objects have "planetary masses"). I have modified my own previous stance on the definition (which was purely mass-based) as a concession to thinking about the overall problem in the cultural context. To apply the definition requires observations of the mass and environment of an object (or inferring these from a combination of observations and well-accepted modeling).
As an example let us take the example of HD 168443 (a solar-type star). At 3 AU there is an object which is indisputably well above the fusion boundary (with a lower mass limit of 17 jupiters). At 0.3 AU there is an object above 7 jupiters. This could also be above the fusion limit if the orbital inclination is low enough, but there is evidence from Hipparcos which makes that relatively unlikely. There is certainly no need to speculate on what to call the outer massive object it is very clearly a brown dwarf. The identity of the inner object is also easy (subject to determination of the true orbital inclination): it nicely fits my definition of "planet". It should be pointed out that there may well be other "stellar" close binaries with inner giant planets; the Doppler searchers have quite purposefully avoided examining this question by avoiding close stellar binaries as targets. At any rate, the identity of objects in this system are only puzzling to those who insist on cosmogenetic preconceptions as defining characteristics.
To the basic definition one is encouraged to add adjectives that make it clearer what one is talking about.  For our Solar System, one can refer to "historical" planets (which include Pluto).  As regards characteristics, one can speak of "ordinary" or "degenerate" planets (perhaps the latter could be called "superplanets" in deference to their mass).  As regards circumstance, one can refer to planets as "minor" or "major" depending on whether they are in unique, non-crossing orbits or not.  This may also depend on how many similar objects there are in the system.  This paper does not address the question of exactly how to make this distinction.  For cosmogony, one can apply adjectives for each mode of formation (once it is theoretically well-accepted and can be justified observationally).  These might include "agglomerated", "core-accretion", "direct", or different and/or additional other terms.
My  proposed definition has the intended "flaw" that it includes ejected planets; those subject to the following sad fate (with apologies to Gilbert and Sullivan's 1885 Mikado)

In an inner stable orbit, round a warm and yellow sun,
with an outer fellow planet that's a huge and gaseous one.
Which will migrate ever closer, wielding gravity's sharp kicks,
whose result is cold careening to the interstellar "sticks".

To confirm such a history may be difficult in most cases. The class of "non-fusors" is therefore, however,  not congruent with "planets", because of the possibility that non-fusors may be also be formed as isolated free-floaters.  Such objects should in principle be given a "stellar" sort of name (such as the recently proposed "grey dwarf") to distinguish them from ejected planets.  As a practical matter, it will be difficult to attach the proper label to a given object (although "free-floating non-fusor" is indisputable if the mass is known, since a fusor companion would always be found if a non-fusor has been detected).  I think it worth living with this difficulty, because it is likely that many ejected analogs of our local terrestrial and gas giant planets are floating out between the stars. They deserve to be called planets despite their misfortune. Even accepting  this, the difficulty of confirmation once again reinforces the problem with including any cosmogony in the definition. An easy way out would be to drop "born" from the definition (and abandon ejected planets as worthy of the name).
One must eventually deal with the question of planets versus moons, and other binary or multiple configurations.  For two non-fusors in orbit about each other, it has been suggested that one differentiate between a "binary companion" and a "satellite" by demanding that the center of mass for the objects be outside both of them for the term "binary" to be appropriate.  I further suggest that this be independent of the masses involved.  Thus, a bound pair of objects with 9 and 11 jupiter masses would be a "binary planet", so long as the pair were born in orbit around a fusor (also independent of its mass).  If they were by themselves, they would be "binary grey dwarfs" (or whatever term applies to isolated non-fusors).  A substantial disadvantage with this proposed definition is that it is very "circumstantial" - to the extent that our own Moon would become a planet if it recedes sufficiently far from us. For the purpose of this paper, it is probably better to duck this issue altogether. This problem makes explicit the difficulty of the role which circumstance can play if used in the identity of objects.  It is this which leads some to suggest that a purely "characteristics-based" definition makes more sense.  It seems, however, that most are not ready for that conceptual change. A circumstantial component is required if we are to retain the cultural imperative of being "in orbit around a fusor" for planets.
Thus, in the end, I think it is possible to fashion a formal definition of "planet" that will accomplish the astrophysical point of having such a word. It can also fit in well with our cultural preconceptions (these are unavoidable given that we live on a planet, in a planetary system), but at some cost. This definition should be able to survive subsequent discoveries and understandings about the characteristics, circumstances, origin, and evolution of planets, and the many undiscovered configurations that Nature has probably produced, which will delight new generations of astronomers. Perhaps, as the general public becomes familiar with the new astrophysical context in which "planets" are discussed, the cultural imperatives surrounding the word will eventually change.

I welcome discussion on this paper; please send email to basri@astro.berkeley.edu

Home Page