Issue 
A&A
Volume 533, September 2011



Article Number  A7  
Number of page(s)  8  
Section  Planets and planetary systems  
DOI  https://doi.org/10.1051/00046361/201117193  
Published online  12 August 2011 
Formation and evolution of planetary systems in presence of highly inclined stellar perturbers
^{1}
Departement Cassiopée, Université de NiceSophia Antipolis, Observatoire de la Côte d’Azur, 06304 Nice, France
email: kbatygin@gps.caltech.edu
^{2}
Division of Geological and Planetary Sciences, California Institute of Technology, Pasadena, CA 91125, USA
^{3}
Section of Astrophysics, Astronomy & Mechanics, Department of Phyiscs, Aristotle University of Thessaloniki, 54 124 Thessaloniki, Greece
Received: 4 May 2011
Accepted: 20 June 2011
Context. The presence of highly eccentric extrasolar planets in binary stellar systems suggests that the Kozai effect has played an important role in shaping their dynamical architectures. However, the formation of planets in inclined binary systems poses a considerable theoretical challenge, as orbital excitation due to the Kozai resonance implies destructive, highvelocity collisions among planetesimals.
Aims. To resolve the apparent difficulties posed by Kozai resonance, we seek to identify the primary physical processes responsible for inhibiting the action of Kozai cycles in protoplanetary disks. Subsequently, we seek to understand how newlyformed planetary systems transition to their observed, Kozaidominated dynamical states.
Methods. The main focus of this study is on understanding the important mechanisms at play. Thus, we rely primarily on analytical perturbation theory in our calculations. Where the analytical approach fails to suffice, we perform numerical Nbody experiments.
Results. We find that theoretical difficulties in planet formation arising from the presence of a distant ( AU) companion star, posed by the Kozai effect and other secular perturbations, can be overcome by a proper account of gravitational interactions within the protoplanetary disk. In particular, fast apsidal recession induced by disk selfgravity tends to erase the Kozai effect, and ensure that the disk’s unwarped, rigid structure is maintained. Subsequently, once a planetary system has formed, the Kozai effect can continue to be wiped out as a result of apsidal precession, arising from planetplanet interactions. However, if such a system undergoes a dynamical instability, its architecture may change in such a way that the Kozai effect becomes operative.
Conclusions. The results presented here suggest that planetary formation in highly inclined binary systems is not stalled by perturbations, arising from the stellar companion. Consequently, planet formation in binary stars is probably no different from that around single stars on a qualitative level. Furthermore, it is likely that systems where the Kozai effect operates, underwent a transient phase of dynamical instability in the past.
Key words: planets and satellites: formation / planets and satellites: dynamical evolution and stability / methods: analytical / methods: numerical
© ESO, 2011
1. Introduction
Among the most unexpected discoveries brought forth by a continually growing collection of extrasolar planets has been the realization that giant planets can have nearparabolic orbits. Since the seminal discovery of 16Cygni B (Cochran et al. 1997), followed by HD 80606 (Naef et al. 2001), much effort has been dedicated to understanding the dynamical origin and evolution of systems with highly eccentric planets. In particular, it has been understood that in presence of a companion star on an inclined orbital plane, the most likely pathway to production of such extreme planet eccentricities is via Kozai resonance (Eggleton & KiselevaEggleton 2001).
The Kozai resonance was first discovered in the context of orbital dynamics of highlyinclined asteroids forced by Jupiter, and has been subsequently recognized as an important process in sculpting the asteroid belt (Kozai 1962) as well as being the primary mechanism by which longperiod comets become Sungrazing (Bailey et al. 1992; Thomas & Morbidelli 1996). Physically, the Kozai resonance corresponds to extensive excursions in eccentricity and inclination of a test particle forced by a massive perturber, subject to conservation of the third Delaunay momentum (where e is the eccentricity and i is the inclination), and libration of its argument of perihelion ω around ± 90°. A necessary criterion for the resonance is a sufficiently large inclination () relative to the massive perturber’s orbital plane, during the part of the cycle where the testparticle’s orbit is circular.
By direct analogy with the SunJupiterasteroid picture, the Kozai resonance can give rise to variation in orbital eccentricity and inclination of an extrasolar planet, whose orbit, at the time of formation, is inclined with respect to a stellar companion of the planet’s host star (Wu & Murray 2003). In the systems mentioned above (16Cygni B, HD 80606) the stellar companions’ (e.g. 16Cygni A, HD 80607) proper motion has been verified to be consistent with a binary solution. Other examples of planets in binary stellar systems are now plentiful (e.g. γ Cephei (Hatzes et al. 2003), HD 196885 (Correia et al. 2008), etc.) with binary separation spanning a wide range ( AU). However, all planets whose eccentricities are expected to have been excited by the Kozai resonance with the companion star are in wide binaries.
If a Kozai cycle is characterized by a sufficiently small perihelion distance, the eccentricity of the planet may subsequently decay tidally, yielding a pathway to production of hot Jupiters, whose orbital angular momentum vector is misaligned with respect to the stellar rotation axis (Fabrycky & Tremaine 2007). The presence of such objects has been confirmed via observations of the RossiterMcLaughlin effect (McLaughlin 1924), leading to a notion that Kozai cycles with tidal friction are responsible for generating at least some misaligned systems (Winn et al. 2010; Morton & Johnson 2011).
Kozai cycles may have also played an important role in systems where a stellar companion is not currently observed. Indeed, one can envision an evolutionary history where the binary companion gets stripped away as the birth cluster disperses. In fact, such a scenario may be rather likely, as the majority of stars are born in binary systems (Duquennoy & Mayor 1991). In this case, a Kozai cycle can be suddenly interrupted, causing the planet’s eccentricity to become “frozenin”.
In face of the observationally suggested importance of Kozai cycles during early epochs of planetary systems’ dynamical evolution, the formation of planets in presence of a massive, inclined perturber poses a significant theoretical challenge (Larwood et al. 1996; Marzari et al. 2009; Thébault et al. 2010). After all, in the context of the restricted problem (where only the stars are treated as massive perturbers), one would expect the protoplanetary disk to undergo significant excursions in eccentricity and inclination due to the Kozai resonance, with different temporal phases at different radial distances, resulting in an incoherent structure. Such a disk would be characterized by highvelocity impacts among newlyformed planetesimals, strongly inhibiting formation of more massive objects (planetary embryos) (Lissauer 1993).
Damping of eccentricities due to gasdrag has been considered as an orbital stabilization process. However, excitation of mutual inclination among neighboring annuli renders this mechanism ineffective (Marzari et al. 2009). Ultimately, in the context of a restricted model, one is forced to resort to competing timescales for formation of planetesimals and dynamical excitation by the companion star. Such an analysis suggests that although possible, planetary formation in binary systems is an unlikely event.
Here, we show that the theoretical difficulties in planet formation arising from the companion star, posed by the Kozai effect and other secular perturbations, can be resolved by a proper account of the selfgravity of the protoplanetary disk (i.e. planetesimals embedded in a gaseous disk). During the preparation of this manuscript, a paper was published (Fragner et al. 2011) addressing the role of the gravity of a gas disk on the relative motion of embedded planetesimals, with hydrodynamical simulations. The work of Fragner et al. (2011) neglects the gravitational effects of the gasdisk onto itself, and therefore considers a case where pressure and viscosity keep the disk more coherent against external perturbations than would be possible with selfgravity alone (Fragner & Nelson 2010). This parameter regime is characteristic of systems where external stresses are strong enough to partially overcome the role of selfgravity, but not that of the internal forces of the fluid. Examples of such systems include binary stars with moderate separations (60 AU in the simulations of Fragner et al. 2011).
In systems of this sort, the dynamics of the planetesimals tends to be somewhat different from those of the gas. Consequently, gasdrag induces sizesorted orbital evolutions, ultimately leading to highvelocity impacts among planetesimals of different sizes. This again, limits the prospects for accretion. Conversely, in the present paper we consider binary separation on the order of AU, consistent with the cases of 16Cyg B and HD HD 80606. This allows us to show, with a simple analytic approach, that gravity is a sufficient mechanism to maintain orbital coherence and planetary growth, without any need to account for other forces acting inside the disk. In fact, we show that one of the primary effects of selfgravity is to induce a fast, rigid recession in the longitudes of perihelion and ascending node of the disk. This allows for planetary formation to take place, as if the secular perturbations arising from the stellar companion were not present. It is noteworthy that such a process is in play, for instance, in the Uranian satellite system, where the Kozai effect arising from the Sun is wiped out owing to secular interactions among the satellites and the precession arising from Uranus’ oblateness (Morbidelli 2002).
Furthermore, we show that even after the formation process is complete, and the disk has evaporated, the Kozai effect may continue to be wiped out by the orbital precession, arising from planetplanet interactions. This is again in line with the example of the outer solar system, where interactions among the giant planets erase a Kozailike excitation due to the galactic tide (Fabrycky & Tremaine 2007). However, if such a planetary system undergoes a dynamical instability, which leads to a considerable change in system architecture, it may evolve to a state where the Kozai resonance is nolonger inhibited.
The purpose of this work is to identify the important physical processes at play, rather than to perform precise numerical simulations. Consequently, we take a primarily analytical approach in addressing the problem. The plan of this paper is as follows. In Sect. 2, we compute the precession rate, arising from the selfgravity of the disk and show that it is copiously sufficient to impede the Kozai effect. In Sect. 3, we show that under secular perturbations from the companion star, the reference plane of the disk precesses rigidly, implying an unwarped structure. In Sect. 4, we show how an initially stable twoplanet system enters the Kozai resonance after a transient instability causes one of the planets to be ejected from the system. We summarize and discuss our results in Sect. 5.
Fig. 1 Example of apsidal precession, γ, in a selfgravitating disk. Here the disk is assumed to contain 50 M_{ ⊕ } between 16 and 32 AU, characteristic of a typical postformation debris disk in the Nice model of solar system formation (Tsiganis et al. 2005). The solid curve shows the precession rate predicted by Eqs. (1)–(3), as a function of semi major axis. The dots and error bars show the results of a numerical calculation, integrating 3000 equalmass particles with a softening parameter of ϵ ≈ 0.005 AU to smooth the effects of their mutual close encounters. The disk was binned into 100 annuli in a and the mean frequency of the longitude of pericenter was measured from the timeseries of ϖ of the particles in each bin (dots) as well as its variance (error bars). Note that the precession frequency of a selfgravitating disk is negative. 
2. Kozai resonance in a selfgravitating disk
We begin by considering the secular dynamics of planetesimals in an isolated, flat nearlycircular disk of total mass M_{disk} around a Sunlike (M_{ ⋆ } = 1 M_{⊙}) star. Due to a nearlynull angular momentum deficit, secular interactions within the disk will not excite the eccentricities and inclinations significantly. Rather, as already mentioned above, the primary effect of disk selfgravity is to induce a fast, retrograde apsidal precession.
Our calculation of the induced precession follows the formalism of Binney & Tremaine (1987), originally developed in the context of galactic dynamics. We work in terms of a polar coordinate system, where the radial coordinate is logarithmic (ρ = lnr) and φ denotes the polar angle. The reduced potential due to a disk surface density σ reads: (1)where is the gravitational constant. We assume σ ∝ r^{1}. Consequently, axial symmetry is implicit, and the potential is only a function of ρ. The characteristic frequencies of a planetesimal in the disk are the mean motion, n, and the radial frequency, κ: (2)where a is semimajor axis. The apsidal precession that results from selfgravity, γ, can then be written as the difference between the mean motion and radial frequencies (3)In practice, the calculation of γ is performed by breaking up the disk into cells and computing the derivatives discretely. Following Levison & Morbidelli (2007), we split the disk into 1000 logarithmic radial annuli, and take the angular cell width to be Δφ = 0.5°. We assume the disk edges to be a_{in} = 0.5 AU and a_{out} = 50 AU, although the results are not particularly sensitive to these choices. The resulting precession in the disk is roughly uniform in a, except for the edges, where this linear theory breaks down. Numerical experiments of debris disks, where selfgravity is taken into account directly, however, show that the precession rate at the edges is also roughly uniform and quantitatively close to that elsewhere in the disk (see Fig. 1). In other words, the disk’s apsidal precession is approximately rigid. Consequently, for the purposes of this work, we take the precession rate evaluated at a = 10 AU, to be the characteristic γ for the entire disk.
Fig. 2 Apsidal recession of a selfgravitating disk. The recession rate, γ is plotted as a function of disk mass. Blue points are the model results. The points are well fit by a linear functional relationship γ = −2.4 × 10^{5} (M_{disk}/M_{Jup}) rad/year. 
Generally, typical protoplanetary disks contain M_{disk} ~ 10−100 M_{Jup} at the time of formation, in gas and planetesimals. We have calculated the characteristic precession rate for a planetesimal embedded in such a disk, for total disk mass range, spanning roughly two orders of magnitude, between M_{disk} = 0.1 M_{Jup} and M_{disk} = 300 M_{Jup}. Figure 2 shows the relationship between γ and M_{disk}. Note that unlike typical planetary systems, where secular interactions among planets give rise to positive apsidal precession, γ of a selfgravitating disk is negative. Quantitatively, for the assumed disk geometry, the precession rate is well fit by the functional relationship γ = −2.4 × 10^{5} (M_{disk}/M_{Jup}) rad/year. Having computed the characteristic precession rate, we can now write down the orbitaveraged Hamiltonian of a planetesimal in the disk.
We work in terms of canonically conjugated actionangle Delaunay variables (4)where the inclination i is measured relative to an arbitrary reference plane and Ω is the longitude of the node of the disk relative to such a plane. In the analysis that follows, we shall take the binary star’s orbital plane to be the reference plane. The Hamiltonian is simply (5) describes an eccentric precessing orbit on a fixed orbital plane (i.e. the plane of the disk).
Fig. 3 Dynamical phasespace portraits for a planetesimal in protoplanetary disks of various masses, perturbed by a stellar companion at various inclinations showing Kozai resonance. The eccentricity vector is plotted in cartesian coordinates on each panel (x = ecosg, y = esing). Regions of libration of argument of perihelion are shown as red curves, while blue curves depict circulation. The top panels represent a massless disk, middle panels correspond to a M_{disk} = 1 M_{Jup} disk and the bottom panels show a M_{disk} = 10 M_{Jup} disk. Note that the Kozai resonance disappears as the disk mass is increased. 
Let us now incorporate the perturbations from the stellar companion into the Hamiltonian. Due to a considerable orbital separation between the protoplanetary disk and the perturber, the interactions between the two will be secular in nature. For our purposes here, we take the perturber to lie on a circular, inclined orbit. A circular perturber implies that, to leading order, potential eccentricity excitations in the disk would arise exclusively from the Kozai resonance. The free orbital precession induced by the companion can be approximated as (Murray & Dermott 1999) (6)where and are the perturber’s mass and semimajor axis and M is the mass of the central star. In the following, we assume that (7)and neglect ġ_{free} altogether. This condition is satisfied for wellseparated binary systems (). Note that the lack of orbital eccentricity of the perturber is a mathematical convenience that makes the calculation more straight forward, without modifying our conclusions on Kozai dynamics qualitatively^{1}.
We take the stellar companion’s orbital semimajor axis to be a = 1000 AU and take the inclination as well as disk mass to be variable parameters. This choice is motivated by the estimate of the orbital separation between HD 80606 and HD 80607 (Eggenberger et al. 2003). It is noteworthy, that the particular choice of a does not have significant consequences on the dynamics of the disk, beyond setting the timescale on which the Kozai effect operates, provided that it is large enough that condition (7) is satisfied. In this section, we shall assume that the nodal reference plane of the disk precesses rigidly, and the disk remains unwarped. In other words, we assume that no mutual inclination is excited between neighboring disk annuli. This feature is implicitly essential to our argument, and we will justify this assumption quantitatively in the next section.
In accord with the reasoning outlined above, we solely retain the Kozai term in the disturbing potential of the stellar companion. Consequently, the planetesimal’s Hamiltonian now reads (Kinoshita & Nakai 1999) (8)where is the stellar perturber’s mean motion. Notice that in the Hamiltonian above, γ could be factorized, so that the magnitude of the perturbation would be proportional to . Since γ is a linear function of the disk mass, this illustrates that it is equivalent to have a closer binary companion (larger ) to a proportionally less massive disk.
With this simplified dynamical model in place, we can now explore the effect of selfgravity on the orbital excitation of the planetesimals in the disk due to the Kozai resonance. We study three choices of perturber inclination: i = 45°, i = 65° and i = 85°. To obtain a dynamical portrait of the system, we proceed as follows. Because the variable h does not appear in Eq. (8), the action H is a constant of motion. On each H = constant surface, the Hamiltonian, , describes a onedegree of freedom system, in the variables G,g. Simultaneously, because the Hamiltonian is also a constant of motion, the dynamics is described by the level curves of the Hamiltonian. For simplicity, we show the dynamics in cartesian coordinates (x = ecosg, y = esing) in Fig. 3, where e is computed from the definition of G, for the assumed value of a (here a = 10 AU). Given that H is constant on each panel, a given eccentricity also yields the inclination. The panels are identified by the value of the inclination i_{max} that corresponds to e = 0 for the given value of H (i.e. the inclination of the star relative to the initial, circular disk). Similarly, the maximal value of e on each panel corresponds to i = 0.
It is useful to begin with a discussion of a massless disk as this configuration is often assumed in formation studies. The corresponding plots are shown as the top panel of Fig. 3. In this case, there is no added precession (γ = 0) so the Kozai resonance is present for all considered choices of inclination. The phasespace portraits show that any orbit which starts out at low eccentricity (near the origin) will follow a trajectory which will eventually lead to a highly eccentric orbit, regardless of initial phase. In particular, for i_{max} = 45°, a particle which starts out on a circular orbit will attain e_{max} ≃ 0.4. For i_{max} = 65°, e_{max} ≃ 0.85 and for i_{max} = 85°, e_{max} ≃ 1. The resulting highvelocity collisions render formation of planetary embryos ineffective. Consequently, one should not expect planets to form under the massless disk approximation.
Let us now consider a M_{disk} = 1 M_{Jup} disk. The phasespace portraits of this system are shown as the middle panels of Fig. 3. Although the quoted value corresponds to a very lowmass disk, the situation is considerably different from the massless case. For i_{max} = 45°, the Kozai resonance is no longer effective^{2}. Thus, an initially nearlycircular orbit will retain its nearzero eccentricity, allowing for planetary formation to take place. The Kozai resonance still operates in the i_{max} = 65° and i_{max} = 85° cases, but the maximum eccentricities are now lower (e_{max} ≃ 0.55 and e_{max} ≃ 0.95 respectively) compared to the massless disk scenario.
Finally, the bottom panels of Fig. 2 show the phasespace portraits of a M_{disk} = 10 M_{Jup} disk. Here, the Kozai resonance is completely wiped out, for all values of i_{max}. Particularly, circular orbits remain circular (the center of each panel is a stable equilibrium point). Strictly speaking, these calculations describe the dynamics of planetesimals embedded in the disk. However, if planetesimals remain circular, the gaseous component of the disk must do so as well because it feels the same gravitational potential. On the other hand, if the Kozai resonance forces the planetesimals to acquire a considerable eccentricity during their evolution (the cases with low disk mass in Fig. 3 or, equivalently, cases with a close stellar companion) the gasdisk may remain more circular than the planetesimals, thanks to its additional dissipative forces. This is the situation illustrated in Fragner et al. (2011), where a differential evolution of planetesimals and gas, leads to sizedependent gasdrag forces.
In conclusion, recalling that 10 M_{Jup} is a lowerbound for the mass of a typical protoplanetary disk, this analysis suggests that planetary formation can take place in well separated binary systems like 16Cyg and HD 806067 as if secular perturbations, arising from the companion star were not present.
3. Rigid precession of a selfgravitating disk
In the previous section, we showed that a selfgravitating disk is not succeptible to excitation by the Kozai resonance. However, in order for our argument to be complete, it remains to be shown that the assumptions of rigid precession of the disk’s nodal reference plane, as well as the lack of the excitation of mutual inclination within the disk, hold true. To justify our assumptions, it is sufficient to consider a nearly circular selfgravitating disk and show that it is characterized by rigid nodal precession, since we have already shown that a flat disk will remain circular under external perturbations.
Intuitively, one can expect a rigidly precessing flat disk, from adiabatic invariance. Consider an isolated selfgravitating disk where mutual inclinations (inclinations with respect to the instantaneous midplane), î, are initially small (i.e. sinî ~ î ≪ 1). Forced by selfgravity, the mutual inclinations within the disk will be modulated on a characteristic (secular) timescale related to the precession of the longitude of the node, , relative to the disk midplane. One can define the action, (9)which represents the phasespace area bounded by a secular cycle (Neishtadt 1984). If the disk is subjected to an external perturbation (such as the torquing from a stellar companion), whose characteristic timescale is much longer than the secular timescale on which selfgravity modulates the mutual inclinations (the socalled adiabatic condition), J will remain a conserved quantity (Henrard & Morbidelli 1993). This implies that the mutual inclinations within the disk will remain small. This is true for each annulus of the disk in which the adiabatic condition is fulfilled. Consequently, the disk will remain unwarped and the disk midplane will precess rigidly at constant inclination relative to the binary star’s orbital plane.
The same idea can be illustrated more quantitatively in the context of classical LaplaceLagrange secular theory. One possible way to view the inclination dynamics of disk is by modeling the disk as a series of massive selfgravitating rings, adjacent to oneanother. Note that the same technique cannot be directly applied to eccentricities, since it would predict a positive apsidal precession, whereas in reality the apsidal precession would be negative, as shown in the previous section. This is because, as soon as the eccentricity is non zero, a ring would start to intersect the adjacent rings, violating the assumption on which the LaplaceLagrange secular theory is based.
Fig. 4 Inclination structure of a massless (blue) and a selfgravitating M_{disk} = 10 M_{Jup} disks. Here the inclination is measured relative to the original plane of the disk. The inclination is shown as a function of semi major axis a at t = 1,3, and 5 Myr. Note that the massless disk is considerably warped due to the perturbations from the companion star, while the selfgravitating disk maintains a uniform inclination. in this case, the growth of inclination with time is due to the rigid precession of the disk relative to the binary star plane. The inclination returns back to zero after a precession period. 
The scaled Hamiltonian of a given annulus j, where exclusively secular terms up to second order in inclination have been retained, reads (Murray & Dermott 1999) (10)where the primed quantities are expressed with respect to an fixed inertial plane (here the initial plane of the disk). In this approach, the disk is broken up into N − 1 annuli whereas the Nth index corresponds to the stellar companion. The coefficients B_{jj} and B_{jk} take the form (11)where m_{j} denotes the mass of a given annulus j if j < N, , α = a_{j}/a_{k}, if perturbation is external and otherwise, while is the Laplace coefficient of the first kind. Similarly to Eq. (6), the diagonal terms in the B matrix correspond to the free nodal precession rates. To crudely account for the large mutual inclination of the stellar companion, we reduced its mass by a factor of sin(i) because in the context of secondorder theory, it is appropriate to only consider the projection of its mass onto the disk’s reference plane. Rewriting the above Hamiltonian in terms of cartesian coordinates (q = i^{′}cosΩ^{′},p = i^{′}sinΩ^{′}), the firstorder perturbation equations () yield an eigensystem that can be solved analytically (see Ch. 7 of Murray & Dermott 1999). In our calculations, we choose N = 101 and again consider a σ ∝ r^{1} surface density across the disk.
We take the orbital properties of the stellar companion to be the same as those discussed in the previous section and take the initial mutual inclination between the disk and the stellar companion to be i = 65°. Indeed, the evaluation of the solution for various disk masses shows that the disk precesses rigidly, if the disk mass exceeds M_{disk} ≳ 1 M_{Jup}. This threshold is in rough quantitative agreement with the numerical models of Fragner et al. (2011). Figure 4 shows the evolution of the inclination as a function of semimajor axis, of a massless (blue) disk as well as a selfgravitating (black) M_{Jup} = 10 M_{Jup} disk at various epochs. The reference plane for the measure of the inclination is the initial plane of the disk. We see that for the massless disk the inclination varies considerably with semimajor axis, which means that the disk is significantly warped, as one would expect in the context of a standard restricted 3body problem. However, the inclination of a selfgravitating disk is nearly constant in semimajor axis, depicting an unwarped, rigid structure. Note, however, that the inclination changes with time. this is because the disk is precessing with a constant inclination relative to the plane of the binary star, so that the disk’s inclination relative to its initial plane has to change periodically, over a precession period, from i_{min} = 0° to a maximum of i_{max} = 130° and back.
In conclusion, the assumption of untwisted structure, that we employed in the previous section when calculating the excitation by the Kozai resonance, is valid for massive disks perturbed by distant stellar companions, such as the ones that we consider in this paper. This conclusion does not apply only to the planetesimal disk, but also to the gas disk, for the same reasons mentioned at the end of the previous section.
4. Production of highly eccentric planets
In the two preceeding sections, we have shown that in a binary stellar system, a selfgravitating disk avoids dynamical excitation, arising from the stellar companion, even if inclined. As a result, one can expect that formation of planetary systems is generally not inhibited. Furthermore, even after the disappearance of the gas, we expect that the Kozai effect can continue to be wiped out as a result of apsidal precession, induced by planetplanet interactions. As already discussed in the introduction, this is the case for the planets of the solar system with respect to the galactic tide, or the satellites of Uranus relative to solar perturbations. Consequently, the final issue we need of address is how planets, such as HD 80606b and 16Cygni Bb, do eventually end up undergoing Kozai cycles.
The evolutionary path that a planetary system can take between the birth nebula stage and the Kozai stage is necessarily nonunique. One obvious possibility is that only a single large planet forms in the disk and as the gas evaporates, selfgravity of the disk becomes insufficient to wipe out the Kozai resonance. Such a scenario, although possible, is probably far from being universal, since the observed multiplicity in planetary systems (Mugrauer et al. 2010), as well as theoretical considerations (Armitage 2010), suggest that protoplanetary disks rarely produce only a single body.
However, an alternative picture can be envisioned: a multiple system forms and after the dispersion of the disk, still protects itself from the Kozai cycles, exerted by the companion star, through selfinduced apsidal precession. Then, following a dormant period, a dynamical instability occurs, removing all planets except one, and therefore the remaining object starts to experience the Kozai resonance. Such a scenario would be considerably more likely, since dynamical instabilities are probably common among newlyformed planetary systems (Ford & Rasio 2008; Raymond et al. 2009). In particular, over the last decade or so, it has been realized that a transient dynamical instability has played an important role in shaping the architecture of the solar system (Thommes et al. 1999; Tsiganis et al. 2005; Morbidelli et al. 2007). Moreover, planetplanet scattering has been suggested to be an important process in explaining the eccentricity distribution of extrasolar planets (Jurić & Tremaine 2008) as well as the misalignment of planetary orbits with stellar rotation axes (Morton & Johnson 2011).
Fig. 5 Orbital evolution of a twoplanet system and its transition into the Kozai resonance via an instability. The figure shows the semimajor axes, as well as perihelion and aphelion distances as functions of time. The planets initially start out in a metastable configuration which is protected from Kozai resonance by apsidal precession, arising from planetplanet interactions. Following ~12 Myr of dynamical evolution, the planets suffer a dynamical instability, during which the initially outer planet is ejected. Consequently, the remaining planet enters the Kozai resonance. 
The usefulness of analytical methods is limited when it comes to the particular study of dynamical instabilities, so one must resort to numerical methods. Below, we demonstrate a numerical proofofconcept of the scenario outlined above.
The system we considered was a pair of giant planets, both with m_{p} = 1 M_{J} around a sunlike (M_{ ⋆ } = 1 M_{⊙}) star, perturbed by a companion. The planets were initialized on near circular orbits (e_{1} = e_{2} = 0.01) with a_{1} = 5 AU and a_{2} = 7.5 AU in the same plane. The stellar companion was taken to be on a circular AU orbit, inclined by i = 80° with respect to the orbital plane of the planets. We performed the simulation using a modified version of SyMBA (Duncan et al. 1998) in which a companion star is set on a distant, fixed circular orbit. The timestep was chosen to be 0.2y and throughout the integration, the fractional energy error remained below . The system was evolved over 10^{8} years.
The orbital evolution of the system is shown in Fig. 5. As can be seen, the system appears stable for the first ~6 Myr, with no sign of Kozai oscillations. However, at ~6 Myr, the system becomes unstable because the planets orbital proximity (initial orbital separation is only ~7 Hill radii) prevents them from remaining stable on long timescales (Chambers et al. 1996). The eccentricities of the planetary orbits grow and eventually, the planets begin to experience close encounters with each other. At ~12 Myr, one of the two planets is ejected onto a hyperbolic orbit. The remaining planet, now alone and with no apsidal precession, gets captured into the Kozai resonance, and starts undergoing large, coupled oscillations in eccentricity and inclination. It is noteworthy that had the remaining planet ended up on an orbit with smaller semimajor axis, general relativistic precession could have wiped out the Kozai effect (Wu & Murray 2003; Fabrycky & Tremaine 2007). Additionally, although in our setup, the stellar companion was initialized with a high inclination, this is not a necessary condition for the scenario, since the scattering event can generate mutual inclination. The transition from nonresonant motion to that characterized by Kozai cycles is depicted in Fig. 6, where orbital parameters prior to the second planet’s ejection are shown as gray dots and the resonant motion is shown as a black curve. Note the similarity of resonant motion computed numerically, to that computed analytically, shown in Fig. 3.
5. Discussion
In this paper, we have addressed the issue of how planetesimals could preserve relative velocities that are slow enough to allow planet accretion to take place, in binary stellar systems. Particularly, we focused on highly inclined systems where Kozai resonance with the perturbing stellar companion have been thought to disrupt the protoplanetary disk and inhibit planet formation (Marzari et al. 2009). Here, we have shown, from analytical considerations, that fast apsidal precession, which results from the disk selfgravity, wipes out the Kozai resonance and ensures rigid precession of the disk’s nodal reference plane.
Fig. 6 Phasespace plot of the inner planet, corresponding to the orbital evolution, shown in Fig. 4. Prior to the instability (t < 12 Myr), the motion of the planet (shown as gray points) is nonresonant. However, after a the outer planet gets ejected, the remaining planet enters the Kozai resonance (shown as a black line). 
It is useful to consider the domain of applicability of the criteria discussed here. Namely, the tradeoff between stellar binary separation and the perturbing companion’s mass should be quantified. The region of parameter space (binary separation vs disk mass to perturber mass ratio) where selfgravity suppresses secular excitation from the binary companion is delineated in Fig. 7. The red curve shows the dividing line between diskdominated and stellar companiondominated apsidal precession (as in Sect. 2). The three purple curves illustrate the disappearance of the Kozai separatrix, for various choices of maximal inclination (as in Sect. 3). The black curve delineates the boundary between rigid precession of the disk’s midplane and a warped structure (as in Sect. 4).
As can be deduced from Fig. 7, for distant stellar companions ( AU), the required total disk mass is of order M_{disk} ~ 1–10 M_{J} (depending on the perturber’s mass), considerably less than or comparable to, the total mass of the minimum mass solar nebula. This implies that generally, protoplanetary disks in binary stars can maintain roughly circular, unwarped and untwisted structures. Consequently, we can conclude that planetary formation in wide binary systems is qualitatively no different from planetary formation around single stars.
After the formation of planets is complete and the gaseous nebula has dissipated, the Kozai effect can continue to be inhibited as a result of orbital precession induced by planetplanet interactions. However, as the numerical experiment presented here suggests, if a planetary system experiences a transient dynamical instability that leaves the planets on sufficiently wellseparated orbits, the planets can start undergoing Kozai cycles. An evolutionary sequence of this kind can explain the existence of orbital architectures characterized by highly eccentric planets, such as those of HD 80606 and 16 Cygni B (Eggleton & KiselevaEggleton 2001; Wu & Murray 2003).
The work presented here resolves, at least in part, a pressing dynamical issue of planetary formation in highly inclined binary systems. As an avenue for further studies, the analytical results presented here should be explored numerically in grater detail. Particularly, hydrodynamic simulations, such as those presented by Fragner et al. (2011) can be used to quantitatively map out the parameter space that allows for planetary systems to form successfully.
Fig. 7 Domain of applicability of the arguments presented in this paper. The red curve shows the dividing line between diskdominated and stellar companiondominated apsidal precession (as in Sect. 2). The three purple curves illustrate the disappearance of the Kozai separatrix, for various choices of maximal inclination (as in Sect. 3). The black curve delineates the boundary between rigid precession of the disk’s midplane and a warped structure (as in Sect. 4). Successful formation of planets can take place in wellseparated binary systems where disk selfgravity dominates over perturbations from the stellar companion. 
The study presented here has further consequences beyond an explanation of planet formation in wide binary systems. Particularly, the model of instabilitydriven evolution of newlyformed systems into the Kozai resonance has substantial implications for orbital misalignment with the parent star’s rotation axis. In fact, Kozai cycles with tidal friction produce a particular distribution of orbitspin axis angles (Fabrycky & Tremaine 2007). This distribution differs significantly from that produced by the planetplanet scattering scenario (Nagasawa et al. 2008). This distinction has been used to statistically infer the dominant process by which misaligned hot Jupiters form (Morton & Johnson 2011). However, the model presented here suggests that the two distributions should be intimately related, as planetplanet scattering provides the initial condition from which Kozai cycles originate. Consequently, a quantitative reexamination of the orbitspin axis misalignment angle distribution, formed by Kozai cycles with tidal friction that originate from a scattered
orbital architecture, and subsequent comparison of the results with observations of the RossiterMcLaughlin effect will likely yield new insights into dynamical evolution histories of misaligned hot Jupiters.
Interestingly, the disappearance of the Kozai separatrix is not exactly symmetric with respect to the sign of γ. If γ is negative, it immediately acts to erase the Kozai effect. However, a small positive γ (i.e. γ = 10^{5} for i_{max} = 45°) can act to enhance to Kozai effect. The effect however rapidly turns over for faster positive precession (i.e. γ > 10^{4}).
Acknowledgments
We thank the referee, Y. Wu for useful suggestions.
References
 Armitage, P. J. 2010, in Astrophysics of Planet Formation, ed. J. P. Armitage, (Cambridge, UK: Cambridge University Press), 294 [Google Scholar]
 Bailey, M. E., Chambers, J. E., & Hahn, G. 1992, A&A, 257, 315 [NASA ADS] [Google Scholar]
 Binney, J., & Tremaine, S. 1987, in Galactic dynamics (Princeton, NJ, Princeton: University Press), 747 [Google Scholar]
 Chambers, J. E., Wetherill, G. W., & Boss, A. P. 1996, Icarus, 119, 261 [NASA ADS] [CrossRef] [MathSciNet] [Google Scholar]
 Cochran, W. D., Hatzes, A. P., Butler, R. P., & Marcy, G. W. 1997, ApJ, 483, 457 [NASA ADS] [CrossRef] [Google Scholar]
 Correia, A. C. M., & Udry, S., Mayor, M., et al. 2008, A&A, 479, 271 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
 Duncan, M. J., Levison, H. F., & Lee, M. H. 1998, ApJ, 116, 2067 [CrossRef] [Google Scholar]
 Duquennoy, A., & Mayor, M. 1991, A&A, 248, 485 [NASA ADS] [Google Scholar]
 Eggenberger, A., Udry, S., & Mayor, M. 2003, Res. Extrasol. Planets, 294, 43 [NASA ADS] [Google Scholar]
 Eggleton, P. P., & KiselevaEggleton, L. 2001, ApJ, 562, 1012 [NASA ADS] [CrossRef] [Google Scholar]
 Fabrycky, D., & Tremaine, S. 2007, ApJ, 669, 1298 [NASA ADS] [CrossRef] [Google Scholar]
 Ford, E. B., & Rasio, F. A. 2008, ApJ, 686, 621 [NASA ADS] [CrossRef] [Google Scholar]
 Fragner, M. M., & Nelson, R. P. 2010, A&A, 511, A77 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
 Fragner, M. M., Nelson, R. P., & Kley, W. 2011, A&A, 528, A40 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
 Hatzes, A. P., Cochran, W. D., Endl, M., et al. 2003, ApJ, 599, 1383 [NASA ADS] [CrossRef] [Google Scholar]
 Henrard, J., & Morbidelli, A. 1993, Physica D Nonlinear Phenomena, 68, 187 [NASA ADS] [CrossRef] [Google Scholar]
 Kozai, Y. 1962, ApJ, 67, 591 [NASA ADS] [CrossRef] [MathSciNet] [Google Scholar]
 Jurić, M., & Tremaine, S. 2008, ApJ, 686, 603 [NASA ADS] [CrossRef] [Google Scholar]
 Kinoshita, H., & Nakai, H. 1999, Cel. Mech. Dyn. Astron., 75, 125 [NASA ADS] [CrossRef] [Google Scholar]
 Larwood, J. D., Nelson, R. P., Papaloizou, J. C. B., & Terquem, C. 1996, MNRAS, 282, 597 [NASA ADS] [Google Scholar]
 Levison, H. F., & Morbidelli, A. 2007, Icarus, 189, 196 [NASA ADS] [CrossRef] [Google Scholar]
 Lissauer, J. J. 1993, ARA&A, 31, 129 [NASA ADS] [CrossRef] [Google Scholar]
 Marzari, F., Thébault, P., & Scholl, H. 2009, A&A, 507, 505 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
 McLaughlin, D. B. 1924, ApJ, 60, 22 [NASA ADS] [CrossRef] [Google Scholar]
 Morbidelli, A. 2002, Mechanics: aspects of solar system dynamics, ed. A. Morbidelli (London: Taylor & Francis) [Google Scholar]
 Morbidelli, A., Tsiganis, K., Crida, A., Levison, H. F., & Gomes, R. 2007, ApJ, 134, 1790 [Google Scholar]
 Morton, T. D., & Johnson, J. A. 2011, ApJ, 729, 138 [NASA ADS] [CrossRef] [Google Scholar]
 Mugrauer, M., Neuhäuser, R., Ginski, C., & Eisenbeiss, T. 2010, Highlights of Astronomy, 15, 694 [NASA ADS] [Google Scholar]
 Murray, C. D., & Dermott, S. F. 1999, Solar system dynamics, ed. C. D. Murray [Google Scholar]
 Naef, D., Latham, D. W., Mayor, M., et al. 2001, A&A, 375, L27 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
 Nagasawa, M., Ida, S., & Bessho, T. 2008, ApJ, 678, 498 [NASA ADS] [CrossRef] [Google Scholar]
 Neishtadt, A. 1984, J. Appl. Math. Mech., 48, 133 [Google Scholar]
 Raymond, S. N., Armitage, P. J., & Gorelick, N. 2009, ApJ, 699, L88 [NASA ADS] [CrossRef] [Google Scholar]
 Thébault, P., Marzari, F., & Augereau, J.C. 2010, A&A, 524, A13 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
 Thomas, F., & Morbidelli, A. 1996, Cel. Mech. Dyn. Astron., 64, 209 [Google Scholar]
 Thommes, E. W., Duncan, M. J., & Levison, H. F. 1999, Nature, 402, 635 [NASA ADS] [CrossRef] [PubMed] [Google Scholar]
 Tsiganis, K., Gomes, R., Morbidelli, A., & Levison, H. F. 2005, Nature, 435, 459 [NASA ADS] [CrossRef] [PubMed] [Google Scholar]
 Winn, J. N., Fabrycky, D., Albrecht, S., & Johnson, J. A. 2010, ApJ, 718, L145 [NASA ADS] [CrossRef] [Google Scholar]
 Wu, Y., & Murray, N. 2003, ApJ, 589, 605 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
All Figures
Fig. 1 Example of apsidal precession, γ, in a selfgravitating disk. Here the disk is assumed to contain 50 M_{ ⊕ } between 16 and 32 AU, characteristic of a typical postformation debris disk in the Nice model of solar system formation (Tsiganis et al. 2005). The solid curve shows the precession rate predicted by Eqs. (1)–(3), as a function of semi major axis. The dots and error bars show the results of a numerical calculation, integrating 3000 equalmass particles with a softening parameter of ϵ ≈ 0.005 AU to smooth the effects of their mutual close encounters. The disk was binned into 100 annuli in a and the mean frequency of the longitude of pericenter was measured from the timeseries of ϖ of the particles in each bin (dots) as well as its variance (error bars). Note that the precession frequency of a selfgravitating disk is negative. 

In the text 
Fig. 2 Apsidal recession of a selfgravitating disk. The recession rate, γ is plotted as a function of disk mass. Blue points are the model results. The points are well fit by a linear functional relationship γ = −2.4 × 10^{5} (M_{disk}/M_{Jup}) rad/year. 

In the text 
Fig. 3 Dynamical phasespace portraits for a planetesimal in protoplanetary disks of various masses, perturbed by a stellar companion at various inclinations showing Kozai resonance. The eccentricity vector is plotted in cartesian coordinates on each panel (x = ecosg, y = esing). Regions of libration of argument of perihelion are shown as red curves, while blue curves depict circulation. The top panels represent a massless disk, middle panels correspond to a M_{disk} = 1 M_{Jup} disk and the bottom panels show a M_{disk} = 10 M_{Jup} disk. Note that the Kozai resonance disappears as the disk mass is increased. 

In the text 
Fig. 4 Inclination structure of a massless (blue) and a selfgravitating M_{disk} = 10 M_{Jup} disks. Here the inclination is measured relative to the original plane of the disk. The inclination is shown as a function of semi major axis a at t = 1,3, and 5 Myr. Note that the massless disk is considerably warped due to the perturbations from the companion star, while the selfgravitating disk maintains a uniform inclination. in this case, the growth of inclination with time is due to the rigid precession of the disk relative to the binary star plane. The inclination returns back to zero after a precession period. 

In the text 
Fig. 5 Orbital evolution of a twoplanet system and its transition into the Kozai resonance via an instability. The figure shows the semimajor axes, as well as perihelion and aphelion distances as functions of time. The planets initially start out in a metastable configuration which is protected from Kozai resonance by apsidal precession, arising from planetplanet interactions. Following ~12 Myr of dynamical evolution, the planets suffer a dynamical instability, during which the initially outer planet is ejected. Consequently, the remaining planet enters the Kozai resonance. 

In the text 
Fig. 6 Phasespace plot of the inner planet, corresponding to the orbital evolution, shown in Fig. 4. Prior to the instability (t < 12 Myr), the motion of the planet (shown as gray points) is nonresonant. However, after a the outer planet gets ejected, the remaining planet enters the Kozai resonance (shown as a black line). 

In the text 
Fig. 7 Domain of applicability of the arguments presented in this paper. The red curve shows the dividing line between diskdominated and stellar companiondominated apsidal precession (as in Sect. 2). The three purple curves illustrate the disappearance of the Kozai separatrix, for various choices of maximal inclination (as in Sect. 3). The black curve delineates the boundary between rigid precession of the disk’s midplane and a warped structure (as in Sect. 4). Successful formation of planets can take place in wellseparated binary systems where disk selfgravity dominates over perturbations from the stellar companion. 

In the text 
Current usage metrics show cumulative count of Article Views (fulltext article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 4896 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.