J. C. B. Papaloizou - R. P. Nelson
Astronomy Unit, Queen Mary, University of London, Mile End Rd, London E1 4NS, UK
Received 20 September 2004 / Accepted 14 December 2004
Abstract
We present evolutionary models of gas giant planets forming in
protoplanetary disks. We first consider protoplanet models
that consist of solid cores surrounded by hydrostatically supported
gaseous envelopes that are in contact with the boundaries of
their Hill spheres, and
accrete gas from the surrounding disk. We neglect planetesimal accretion, and
suppose that the luminosity arises from gas accretion alone.
This generally occurs on a long time scale which may be
comparable to the protostellar disk lifetime.
We classify
these models as being of type A, and follow their quasi static evolution
until the point of rapid gas accretion is reached.
We consider a second class of protoplanet models that have not hitherto
been considered. These models have a free surface, their energy supply
is determined by gravitational contraction, and mass accretion
from the protostellar disk
that is assumed to pass through a circumplanetary disk.
An evolutionary sequence is obtained by specifying the accretion rate
that the protostellar disk is able to supply.
We refer
to these models as being of type B. An important
result is that these protoplanet models
contract quickly to a radius
cm
and are able to accrete
gas from the disk at any
reasonable rate that may be supplied without any consequent expansion
(e.g. a Jupiter mass in
few
years, or more slowly
if so constrained by the disk model).
We speculate that the early stages of gas giant
planet formation proceed along evolutionary paths described by models
of type A, but at the onset of rapid gas accretion
the protoplanet contracts interior to its Hill sphere,
making a transition to an evolutionary path
described by models of type B, receiving gas through a circumplanetary
disk that forms within its Hill sphere, which is in turn
fed by the surrounding protostellar disk.
We consider planet models with solid core masses of 5 and
,
and consider evolutionary sequences
assuming different amounts of dust opacity
in the gaseous envelope.
The initial protoplanet mass doubling time scale is very approximately
inversely proportional
to the magnitude of this opacity.
Protoplanets with
cores, and standard dust opacity
require
years to grow to a Jupiter mass,
longer than reasonable disk life-times.
A model with 1% of standard dust opacity requires
years.
Rapid gas accretion in both these cases ensues once the planet
mass exceeds
with substantial time
spent in that mass range.
Protoplanets with
cores grow to a Jupiter mass
in
years if standard dust opacity is assumed,
and in
years if 1% of standard dust opacity is adopted.
In these cases, the planet
spends substantial time with mass between 30-40
before
making the transition to rapid gas accretion. We emphasize that these growth times
apply to the gas accretion phase and not to the prior core formation phase.
According to the usual theory of protoplanet migration, although there is some dependence on disk parameters, migration in standard model disks is most effective in the mass range where the transition from type A to type B occurs. This is also the transitional regime between type I and type II migration. If a mechanism prevents the type I migration of low mass protoplanets, they could then undergo a rapid inward migration at around the transitional mass regime. Such protoplanets would end up in the inner regions of the disk undergoing type II migration and further accretion potentially becoming sub Jovian close orbiting planets. Noting that more dusty and higher mass cores spend more time at a larger transitional mass that in general favours more rapid migration, such planets are more likely to become close orbiters.
We find that the luminosity of the forming protoplanets
during the later stages of gas accretion is dominated by the circumplanetary
disk and protoplanet-disk boundary layer.
For final accretion times for one Jupiter mass
in the range 105-6 y, the luminosities are in the range
and the characteristic
temperatures are in the range 1000-2000 K.
However, the luminosity may reach
for shorter time periods at the faster rates of accretion
that could be delivered by the protoplanetary disk.
Key words: accretion, accretion disks - solar system: formation - stars: planetary systems
The process is presumed to begin
with the accumulation of the solid cores by the
accretion of planetesimals typically exceeding
a kilometer in radius which have been formed
through the collisional growth and sedimentation of dust grains in
the protoplanetary disk (see Lissauer 1993, and references therein).
Once the solid core becomes massive enough
a significant gaseous atmosphere forms.
The mass required depends to some extent on physical conditions
in the disk, the rate of planetesimal
accretion and the dust opacity but is typically several earth
masses
(e.g. Mizuno 1980; Stevenson 1982; Bodenheimer & Pollack 1986).
This is consistent with models of
Jupiter which indicate that it has a solid core typically of
this magnitude (Podolak et al. 1993). We note, however, that more recent models
suggest that Jupiter's core may be less massive than previously thought
(Saumon & Guillot 2004). Models of Saturn still indicate a core mass of
10
.
During the early build up of the core the luminosity is due to the liberation of gravitational energy by accreting planetesimals. However, once the mass of the gaseous envelope starts to become significant the gravitational settling of the gas becomes important and at some cross over point becomes dominant (Pollack et al. 1996). At this point models assuming strict thermal equilibrium break down. This is manifest through the fact that for fixed luminosity due to planetesimal accretion, there is a maximum or critical core mass for which a strict thermal equilibrium model can be constructed (see e.g. Papaloizou & Terquem 1999). Beyond this point the evolution is no longer in thermal equilibrium and if the protoplanet remains in contact with adequate disk material, gas accretion may ensue.
The purpose of this paper is to examine
the protoplanet models subsequent to the attainment
of the critical core mass in the context
of the protoplanetary disk environment
and disk planet interactions. We assume that the core becomes isolated
from further planetesimal accretion so that settling of accreted gas
is the only energy source. The rationale for this assumption is discussed
in Sect. 4.2. We consider two types of model.
The first type, which we denote as type A, is fully embedded
in the protostellar disk
and hence has an effective radius equal to that of the Roche lobe
or Hill sphere. This is the correct radius to use rather than the
Bondi radius which is never significantly smaller for any of the models
we study. At some mass, these models tend to
enter a rapid accretion phase. This occurs when the planet mass
,
denoting a Jupiter mass, and
is similar to that for which either significant perturbation
to the protoplanetary disk through local mass
accretion or disk-planet interaction begins (e.g. Nelson et al. 2000).
These processes eventually lead to gap formation.
Accordingly we consider models of a second type, type B,
which are no longer enveloped at the Roche lobe but are assumed to
have a free surface and accrete from a circumplanetary disk
at a rate determined by the external throughput from the protostellar disk.
We find that these can be constructed for a wide range of accretion rates
indicating that during the later stages a forming protoplanet
can comfortably
accrete at any rate reasonably supplied by the protostellar disk.
We supplement these models of protoplanet evolution with hydrodynamical simulations of the interaction between low mass protoplanets and protostellar disks. The purpose of these models is to establish plausible accretion times scales for the freely accreting protoplanet models of type B.
This paper is organised as follows. We present the basic equations for the protoplanet models in Sect. 2, and discuss the appropriate boundary conditions in Sect. 3. In Sect. 4 we describe how evolutionary sequences are constructed for protoplanet models of type A and B, accounting for gas accretion from the protostellar disk. We discuss the numerical procedure adopted for the hydrodynamic simulations of disk-planet interactions in Sect. 5. The results of our calculations are presented in Sect. 6, and their implications are discussed in Sect. 7. Finally we draw our conclusions in Sect. 8.
The interior state variables at any point in a
model are functions only of the distance to the centre,
r, also characterized as the spherical polar radius.
We assume the models are in
hydrostatic equilibrium and neglect rotation. The
equation of hydrostatic equilibrium is
For the calculations presented here, we adopt the
equation of state for a hydrogen and helium mixture given by Chabrier
et al. (1992). The mass fractions of hydrogen and helium are taken to
be 0.7 and 0.28, respectively. The luminosity
transported
by radiation satisfies
The radiative and adiabatic temperature gradients
and
are defined through
![]() |
(5) |
The total luminosity is Lr. During the phase of solid core growth it is expected that this is produced by the gravitational energy of accreting planetesimals (e.g., Mizuno 1980; Bodenheimer & Pollack 1986). However, for the later phases considered here, the source of energy is primarily settling and accretion of gas (see Sect. 4.2).
When
,
the gas is
convectively stable and the energy is transported
entirely by radiation. On the other hand
when
the medium is
convectively unstable and some of the energy is transported by
convection. We write the total luminosity
passing through a sphere of radius r
as
,
where
is the
luminosity associated with convection. Adopting the mixing length theory
(Cox & Giuli 1968) we have
![]() |
(7) |
![]() |
(8) |
Thus the boundary conditions
are that at
,
and the temperature is given by
Here we approximate the additional optical depth above the
protoplanet atmosphere, through which radiation passes, by (Papaloizou & Terquem 1999)
![]() |
(10) |
For most models we adopt disk parameters appropriate to 5 AU
from the disk model of Papaloizou & Terquem (1999) with Shakura & Sunyaev (1973)
and steady state accretion rate of
Accordingly
K and
dyn cm-2.
For these models, in contrast to those of type A,
the effect of the exterior disk material on the
surface boundaries is small. Thus for the boundary condition on Twe again adopt
Eq. (9) but with
.
We note for these models
is
in general significantly larger than
For the boundary condition on P we adopt
In order to have a complete system for which the evolution can be calculated Eqs. (1)-(6) need to be supplemented by an equation governing internal energy production and the internal luminosity, normally the first law of thermodynamics. Here we simplify matters by using the fact that most of the internal energy of the models is contained within a deep convection zone. The thermal time scale associated with relaxation of the exterior layers is expected to be much shorter than the thermal relaxation time scale associated with the model as a whole. Under these conditions, if the model evolution time scale is on the global Kelvin-Helmholtz time scale or longer, it is a reasonable assumption that Lr is constant in the outer layers. Because of efficient convection the inner convection zone is unaffected by the distribution of Lr. Accordingly we make the assumption that Lr = L is constant in the outer layers. This is expected to hold during the longest lasting evolutionary phases for all masses and and at all times for the larger masses which tend to have only a very thin surface radiative shell, but we bear in mind that it may fail when the evolution time becomes very short. With the above assumption we obtain a complete system for which the evolution may be calculated.
When such a model increases its mass slightly so that
the change of energy content is
If the energy change balances losses by radiation in time dt,
then conservation of energy requires that
![]() |
(14) |
In arriving there all available gravitational binding energy of
per unit mass,
being the surface radius,
has been liberated and so
an amount of energy
must be subtracted
from dE in order to obtain the energy available to replace radiation losses.
Therefore if the changes occur over an interval dt, we must have
Thus total energy conservation
for models of type B enables the calculation of evolutionary tracks
through.
Thus Eqs. (13) and (15) constitute the basic equations governing the evolution of models of type A and type B respectively.
Note that we neglect any input from planetesimal accretion during and after the phase when the core becomes critical. The primary reason for doing this is that we are interested in examining the fastest time scales possible for giant planet formation via the core instability scenario. The inclusion of planetesimal accretion and the associated accretion luminosity will have the effect of lengthening the time scale of formation, provided that the core mass itself does not increase significantly. However, there are also uncertainties about how large the planetesimal accretion rate ought to be.
Previous work on the formation of gas giant planets via the core
instability model assumed that core formation can proceed through runaway growth
in which a protoplanetary core can grow by accreting essentially all
planetesimals in its feeding zone (e.g. Pollack et al. 1996). This resulted
in a core formation time of few
years. Simulations by
Ida & Makino (1993) indicate, however, that runaway growth slows down prior
to the completion of core formation, and proceeds through a more
orderly mode of planetesimal accretion known as oligarchic growth.
This arises because neighbouring planetary embryos stir up the random motions
of the planetesimal swarm, reducing the effectiveness of gravitational focusing.
N-body simulations of protoplanetary core formation indicate that obtaining
cores of the necessary mass is not an easy task to achieve during the
oligarchic growth phase, in part due to planetary cores of a few
repelling the surrounding planetesimals and opening gaps in the
planetesimal disk, and in part due to the excitation of planetesimal
eccentricities and inclinations by the "oligarchs'' (e.g. Thommes et al. 2003).
After core formation, and during the longest phase of evolution involving gas
settling onto the core,
the calculations of Pollack et al. (1996) result in planetesimal accretion
rates that are a factor of times smaller than
those of the gas accretion
rate, and this planetesimal accretion results in significant accretion
luminosity.
This planetesimal accretion arises because the feeding zone expands as
the planet mass increases due to gas accretion, and depends on the strict
assumption
that planetesimals are not allowed to enter or leave the feeding zone.
Thus the possibility of gap formation in the planetesimal disk, as found
by Thommes et al. (2003), is not accounted for in these models, although
one may reasonably expect its effect to be increasingly important
as the planet mass increases.
The generation of significant luminosity from planetesimal accretion depends on where it is assumed the energy is deposited within the protoplanet. Large (100 km) planetesimals are able to penetrate deep into the planetary interior, and so provide a significant source of energy by virtue of descending deep into the gravitational potential well. Smaller planetesimals or fragments are more likely to dissolve higher up in the planet atmosphere, and so will contribute less accretion luminosity. A possible resolution of the long time scales of formation for planetary cores reported by Thommes et al. (2003) is that collisions between planetesimals result in fragmentation when their random motions are excited by the forming planetary embryos (e.g. Rafikov 2004). This possible generation of smaller planetesimals results in increased efficiency of gas drag by the nebula in damping random motions, thus speeding up planetesimal accretion by planetary embryos. This potential modification of the size distribution will also have an impact on the accretion luminosity generated by accreted planetesimals.
In the light of these uncertainties in the radial distribution and size distribution of planetesimals, and its effect on planetesimal accretion rates during the gas settling, and rapid gas accretion phase of giant planet formation, we believe it is justified to treat the planetesimal accretion rate and its associated luminosity generation as a free parameter of the problem. A similar approach has been argued for by Ikoma et al. (2000). As we are interested in examining the shortest possible time scales for giant planet formation, we neglect the effects of planetesimal accretion in this study.
With the above assumption we have a complete system of Eqs. (1)-(6), (13) and (15) for which the evolution may be calculated.
The number of grid cells used was
.
The inner boundary of the computational domain was placed at
R=0.4 and the outer boundary at R=3. Reflecting boundary conditions
were used at both radial boundaries, and linear viscosity was used between
and
to reduce reflection of waves excited by
the planet. The gravitational potential of the planet was softened using a
softening parameter
b=0.5 H(Rp) - i.e. half of the local disk
semi-thickness.
Simulations were initiated by placing a low mass planet (either 15 or 30 Earth
masses) at a radius Rp=1 in the disk. The planet was assumed to accrete gas
that entered its Hill sphere. This was achieved by removing gas
from any cells that lay within half of the planet Hill sphere. The e-folding
time of this gas removal was
,
where
.
Thus this corresponds
to the extreme case when the planet accretes material within
the Hill sphere on the dynamical time scale. The gas that was removed from the Hill sphere
was added to the planet at each time step, such that the planet mass
is a function of time. Similar models are described in Nelson et al. (2000).
For a fixed accretion rate onto a core
at a given radius, there is a critical
core mass
above which no solution can be found
in hydrostatic and thermal equilibrium that joins on to
the protoplanetary disk model assumed at the Roche lobe.
In this paper we consider cores with
and
Our evolutionary calculations commence close to the state when the cores
are critical, that is no further gas
can be added in strict hydrostatic equilibrium. At this stage the evolution
is slowest and we compute a type A model sequence by use of Eq. (13). These models are in contact with the Roche lobe.
We also construct type B model sequences. These satisfy the free surface boundary conditions given by Eqs. (9) and (11). Because the surface is free, these form a two parameter sequence in that evolutionary tracks for a given mass can be started for a range of radii (or equivalently the luminosity may be used as a parameter). This the same situation as in standard pre-main sequence contraction where a stellar model of a given mass can be started at different points on an evolutionary track corresponding to different radii.
We have considered models using the Bell & Lin (1994) opacities hereafter referred to as standard. These have a very large contribution from dust grains for T < 1600 K and because there is clearly some uncertainty about the disposition of the dust particularly under circumstances where the protoplanet is assumed isolated from further planetesimal accretion, we have explored the effect of reducing this contribution to the opacity by factors of up to 100for both models of type A and B. We have done this, by making the reduction for the opacity as a whole, for T <1600 K only, and with a reduction factor that is constant for T <1600 K and which then decreases linearly to unity at T = 1700 K. In practice we find that the results are essentially independent of whether such a linear join is made or not. The uncertainty in the magnitude of the surface opacity as well as its important role in controlling the evolutionary time scale of an embedded protoplanet has been pointed out by Ikoma Nakazawa & Emori (2000).
We begin by describing some typical models of type A.
In Fig. 1, state variables are plotted for a protoplanet model
with
which has
a total mass
As expected the deep interior
of this model is convective with little energy transported by radiation.
However, there are two convective regions which occur for
K and T > 2100 K.
Approximately ninety eight percent of the mass
is in the inner convective zone. This means that most of the thermal inertia
is contained within the deep convection zone rendering the assumption
of little spatial variation of the luminosity in the upper layers a reasonable
approximation. The existence of two separate convective regions
is in contrast to what we find for models of type B that approach
In those cases we find a single interior convection zone for
T > 1000-2000 K,
with negligible mass in the outer radiative region.
![]() |
Figure 1:
State variables are plotted for a protoplanet model
of total mass
![]() ![]() ![]() |
Open with DEXTER |
![]() |
Figure 2:
As in Fig. 1, but
for a protoplanet model
of total mass
![]() ![]() ![]() |
Open with DEXTER |
![]() |
Figure 3:
This figure illustrates the evolution of
protoplanet models which maintain contact with the
protoplanetary disk and fill their Roche lobes while
they accrete from it. They have fixed
solid core masses of
![]() ![]() ![]() ![]() |
Open with DEXTER |
![]() |
Figure 4:
As in Fig. 3 but for two models with standard
opacity and solid cores of
![]() |
Open with DEXTER |
![]() |
Figure 5: This figure shows the mass accretion onto planetary cores obtained from the hydrodynamic simulations described in the text. The left panel shows accumulated mass onto each planet, and the right hand panel shows the mass accretion rate in units of Jupiter masses per year. We note that the values obtained for this quantity span the range of values used for the accretion rates onto the detached planet models described in Sect. 6.3. |
Open with DEXTER |
![]() |
Figure 6:
This figure illustrates the evolution of
protoplanet models which accrete from the
protoplanetary disk at an assumed
rate of one Jupiter mass in
![]() ![]() ![]() |
Open with DEXTER |
![]() |
Figure 7: As in Fig. 6 but for an assumed accretion rate from the disk that is ten times slower. As a consequence of that it takes about ten times longer to attain one Jupiter mass in these cases. |
Open with DEXTER |
![]() |
Figure 8:
As in Fig. 6 but for models with a
![]() |
Open with DEXTER |
![]() |
Figure 9:
As in Fig. 6 but for models with a
![]() |
Open with DEXTER |
![]() |
Figure 10:
As in Fig. 6 but for models with a
![]() |
Open with DEXTER |
![]() |
Figure 11:
As in Fig. 6 but for models with a
![]() |
Open with DEXTER |
![]() |
Figure 12:
As in Fig. 6 but for models with a
![]() |
Open with DEXTER |
![]() |
Figure 13:
As in Fig. 6 but for models with a
![]() |
Open with DEXTER |
In Fig. 2,
we illustrate the behaviour of the state variables
for a protoplanet model with the smaller
core mass.
The total mass is
This has similar
properties to the previous case regardless of the
fact that the core mass is three times smaller.
Convective heat
transport occurs when
K and when T > 2100 K.
The inner eighty percent of the mass
is convective.
In Fig. 3 we illustrate the evolution of
models of type A for
Cases with standard opacities and with opacity reductions of
three, ten and one hundred made globally and for T < 1600 K
are shown.
In all cases as the models gain in mass from the protoplanetary disk
their luminosity increases and their evolutionary time
measured through their accretion time
decreases. In the standard opacity case the accretion
time is very long, exceeding 108 y. However, this time
reduces by the opacity reduction factor independently
of where this is applied even though the situation
might have appeared to have been complicated by the existence of two
convective regions.
Thus times
y are attained
for reduction factors of one hundred.
The evolutionary time scale of the models begins
to decrease rapidly once
,
becoming less that 105 y even for the standard opacity case.
This phenomenon, which is sensitive to relatively minor model
details, can be traced to the fact that
becomes small
or that less and less binding energy is liberated
as the mass increases. This is likely to indicate the
onset of a rapid collapse and possible detachment from the Roche lobe.
As the effect of disk planet interactions and local gas depletion
are likely to become important, we have not tried to follow
such rapid evolution with the simplistic models adopted here.
Rather we have considered evolutionary sequences
of type B which are likely to be the outcome.
Although the position where such a sequence is joined
cannot of course
be determined without considering the above rapid phase of evolution.
Figure 3 also shows that
the effective temperature needed to
supply the luminosity of the model, is always small when compared to
the surrounding protostellar disk temperature.
This indicates negligible thermal perturbation of the protostellar disk.
In Fig. 4
we illustrate the evolution of
models of type A for
but for two models with standard
opacity
For these models the longest evolutionary times
are in the
y range.
The attainment of short evolutionary times likely
leading to detachment from the Roche lobe occurs
for
in this case.
The two models illustrated differ in surface boundary conditions.
The model illustrated with dotted curves
is embedded
in a disk
with the same temperature
but with a density ten times larger than usual. Except
during the beginning of the rapid evolution phase
the models show very similar behaviour.
In Fig. 4 we also plot
evolutionary tracks for which the opacity
was reduced by factors of ten and one hundred
in the surface layers for which T > 1600 K
with a linear transition to standard opacities occurring
for
K.
For these sequences the transition mass is unaffected but
the evolutionary time scales are factors of three and thirty
faster respectively. This means that the formation time scale
is reduced to
y in the latter case.
The accretion times obtained in Fig. 5 range from a
few thousand years to a few tens of thousands of years, and show that
the actual accretion time
obtained is sensitive to the disk viscosity assumed. In the planet models of
type B presented below, we consider accretion times of between
to
years.
These cover the accretion times obtained for a protoplanet on short time
scales in the extreme case when it is
immersed ab initio into an unperturbed disk as in the above simulations.
They also allow for the situation where there is gas depletion
such that the protoplanet can only accrete for longer time scales at an assumed mass flow
rate through the protostellar disk of
![]() |
Figure 14: As in Fig. 13 but for a sequence of models with the opacity reduction applied only for T<1600 K. |
Open with DEXTER |
Figure 7 illustrates evolutionary tracks
for a fixed assumed accretion rate
that is ten times slower.
As a consequence of this the evolution times are ten times
longer. The convergence of the evolutionary tracks is greater in this case
with all radii being close to
cm after
y.
The indication is that values of
K
for these models are sustained for
y.
However, the luminosity expected from circumplanetary
disk accretion at the later phases
erg s-1(calculated adopting a radius of
cm
for the protoplanet) is only exceeded at early times
by the most luminous model
which then becomes fainter at later times. But note that for this
sequence of models and others presented later, models accreting from the
disk can exist which have small luminosities
the circumplanetary
disk luminosity and also that due to the protoplanet disk boundary layer, equal to
(see e.g. Lynden-Bell & Pringle 1974).
In Fig. 8 we show tracks for an accretion rate
of
in
y for models with
The behaviour is similar to that in the higher core mass case.
In Fig. 9 we show models with
with the same accretion rate
which have opacities globally reduced by a factor of three
and in Fig. 10
the reduction is by a factor of ten. In all of these cases
there is a tendency of the tracks to converge especially
the radii of different models to a value of about
cm,
with the lower opacity models being
slightly smaller. In all cases the protoplanet luminosities
is exceeded at late stages by the circumplanetary disk luminosity.
In Fig. 11 we illustrate models
with
and standard opacities accreting
from the disk at a rate that is ten times slower while in
Fig. 12
the opacity is globally reduced by a factor of three at that accretion rate.
In these cases the evolution is prolonged by a factor of 10.These models are again similar to the previous ones and indicate that
a model starting from one Saturn mass and radius
cm
could sustain effective temperatures
K for times approaching 106 y.
Finally in Figs. 13 and 14
we explore models with
subjected to a very high
accretion rate from the disk at a rate of
in
y
with an opacity reduction by a factor of ten applied globally
in the former case and applied only for T<1600 K in the latter.
Paradoxically (see the discussion in Sect. 7) these
models may appear somewhat cooler
and less intrinsically luminous than those calculated for lower accretion rates.
However, convergence of model radii towards
cm
again occurs.
We plot the growth times for three different evolutionary models
in Fig. 15. The dashed line represents the model with standard
opacity, the dashed-dotted line the model with one tenth the standard
opacity, and the dashed-dot-dot-dotted line the model with opacity
reduced by a factor of 100. During the earliest phases of evolution
the growth times of these models are
,
and
years respectively.
The two models with largest opacity are thus unable to form giant planets
within the disk life-time. Such systems will result in rock and ice cores
forming that are unable to accrete significant gas envelopes.
The lowest opacity model has a sufficiently
low growth time that it will be able to form a giant planet before
disk dispersal. However, Fig. 15 shows that during the early
stages of evolution, while the planet mass is below
,
the growth time is significantly longer than the type I migration time scale,
implying that the protoplanet will migrate into the central star before
forming a gas giant. This is a problem for all reasonable core
instability models of gas giant formation, since there exists a
bottle neck for gas accretion while the planet mass is relatively small,
but massive enough to undergo quite rapid migration. If the core instability
model is correct, then we are inevitably led
to the conclusion that some process must operate to prevent type
I migration in a standard quiescent disk
for at least some protoplanets in order that gas giant
planets can form.
Although many issues
remain outstanding,
a number of processes may operate to prevent type I migration.
Being the result of a linear disk response it depends on the
temperature and density structure of the disk and special
features such as rapid spatial variation of opacity may slow
or stop migration (e.g. Menou & Goodman 2004).
Recent simulations by Nelson & Papaloizou (2004) and Nelson (2004)
show that low mass planets migrating in magnetised, turbulent accretion disks
undergo stochastic migration rather than monotonic inward
migration. This leads to a distribution of migration rates
for embedded planets, with some undergoing rapid inward migration,
and others perhaps migrating outward or not at all. A well defined direction
of migration is likely to occur when the planet mass is large enough to
dominate over turbulent fluctuations, with simulations indicating that
this is likely to arise for planet masses exceeding
.
The occurrence of global disk structures such as eccentric m=1 modes
are also capable of disrupting both type I (Papaloizou 2002) and
type II (Nelson 2003) migration, and if established within a disk
are likely to be long lived entities. Finally, low mass planets on
eccentric orbits may undergo type I torque reversal (Papaloizou & Larwood
2000). For an isolated planet the eccentricity is quickly damped,
but a system of mutually interacting planetary cores may be able to
maintain eccentric orbits and hence reduce or even prevent type I
migration.
In light of these (and perhaps other) mechanisms for overcoming
type I migration, which may operate in tandem rather than in isolation,
we make the assumption that for masses below
,
type I migration is essentially ineffective for at least some protoplanets
below that mass range, such that a population of giant planet can form.
It seems likely that for planet masses larger than this, where the
disk-planet interaction starts to become non linear, the ability of
the planet to impose itself on the disk will lead to inward migration being
re-established. When it does so Fig. 15 indicates
it will be at near the maximum rate. The transition from type A to type B
models is near to where
Note that m0 is likely to
depend on location
and parameters in the disk making it uncertain whether the protoplanet
undergoes some rapid inward migration.
Returning to our discussion of the low opacity planet model in
Fig. 15, we can see that once the planet mass reaches
and moves to a type B track,
the growth time of the planet decreases dramatically
down to a value that is determined by the rate at which the protostellar
disk can supply mass to the planet. It is at this stage that we suppose
that the protoplanet undergoes a transition from being an extended
structure in contact with its Roche lobe, and accreting slowly from the
disk, to a more compact protoplanet with a free surface that accretes
rapidly from the protostellar disk via a circumplanetary accretion disk.
The type B models presented in Figs. 13 and 14 suggest
that these compact models can accrete rapidly from the disk, and we
specify a growth time of
years for this stage of growth
in Fig. 15, corresponding to the more rapid growth rates
presented in Fig. 7. Such a rate ensures that the planet can grow to
become a Jovian mass gas giant on a time scale shorter than any likely
migration time.
As shown in Fig. 4, the growth time for the standard opacity model
presented in Fig. 16 remains larger than the
corresponding migration time for planet masses up to 34
,
at which stage rapid gas accretion can ensue.
Because such planet models spend longer time at these higher masses
they may be
more susceptible to undergoing
a period of rapid migration either prior to or during the
early stages of rapid gas accretion than are the models with lower core masses.
This may be related to an indication that extrasolar planets
in systems with high metallicity tend to be found at shorter periods
commented on by Santos et al. (2003). However, because of the small
numbers involved, the statistical significance of such a trend is not yet
established.
The result of a rapid inward migration is that the planet will move into the inner regions of the disk where: (i) the local reservoir of disk material is reduced relative to larger radii; (ii) the disk aspect ratio H/R decreases making gap formation and a transition to type II migration easier (Papaloizou & Terquem 1999). The result is likely to be a tendency for larger and more dusty cores to produce a distribution of planets with a greater bias toward low mass, short period objects.
The reason for the behaviour of type B models
where they fail to expand even at high accretion rates can be
related to some simple properties of barotropic stellar models
that would apply in the completely degenerate limit.
For these P is a specified function of
and is related to the internal
energy per unit mass, U, through
Although the protoplanet models are not of this type, they
are similar enough to make the discussion relevant.
The total energy is given by
Eq. (12). For polytropes of index n and
,
it is well known that
(Chandrasekhar 1939)
![]() |
(16) |
The expected circumplanetary
disk or disk/protoplanet boundary layer luminosity
for a Jovian mass with radius
cm,
and final accretion times
in the range 105-6 y, lies in the range
and the characteristic
temperatures are expected to be in the range 1000-2000 K.
![]() |
Figure 15:
This diagram provides a schematic representation of
the formation and migration time scales of planet models as a function
of protoplanet mass for core masses of
![]() ![]() |
Open with DEXTER |
![]() |
Figure 16:
This diagram is similar to Fig. 15, except that
it applies to planet models with
![]() |
Open with DEXTER |
We present a second class of planet models (type B) where the planet has a free surface, and accretes gas from a circumplanetary disk that is fed by the surrounding protostellar disk at a specified rate. We find that these models can accrete gas at any reasonable rate that may be supplied by the protostellar disk without expansion.
We suggest that the earliest stages of giant planet formation
are described by models of type A. For all such models,
the standard type I migration time is shorter than the accretion time
prior to rapid gas accretion. We suggest that type I migration is
inoperative for at least some protoplanets with masses below those
for which disk-planet interactions becomes non-linear (i.e.
), beyond which planets are more
likely to undergo rapid inward migration. In such a scenario
planets with low mass cores and low opacity envelopes will have a
greater tendency to remain at larger radii up to the point of
rapid gas accretion. Those with more massive cores will tend to
undergo more rapid inward migration prior to or during rapid gas
accretion.
At the point of rapid gas accretion, we suppose that planets contract within their Hill sphere, and are described by type B models. The planets may now accrete at any rate supplied by the protostellar disk, and can undergo rapid growth on a time scale shorter than the migration time. If planets with low mass cores tend to exist at larger radii during this stage, they may make a rapid transition to Jovian mass objects, forming gaps and entering a phase of slower type II migration. If planets with larger mass cores have a tendency to undergo more rapid inward migration, they may exist at smaller radii during the rapid gas accretion phase. The disk is thinner here - such that it is easier to form gaps, and the local reservoir of gas is smaller. Such objects are more likely to end up with sub-Jovian masses.
We note that this general picture is likely to be blurred by variations in disk parameters and life-times. But we also note that the current extrasolar planet data shows a mass-period correlation in line with the simple ideas presented here (Zucker & Mazeh 2002). Furthermore there is a hint of a correlation between host star metallicity and period such that higher metallicity stars appear to host shorter period planets (e.g. Santos et al. 2003). Such a correlation, while not statistically significant in the data at present, may turn out to be as more data is accumulated and is accordingly a topic for scientific consideration (e.g. Sozzetti 2004). We comment that a correlation of this type might be expected if planetary core mass and envelope opacity scale with the metallicity of the protoplanetary environment.
Acknowledgements
The hydrodynamic simulations performed here were carried out using the QMUL HPC facility funded by the SRIF initiative.