Arxiv:1006.0182V1 [Astro-Ph.HE] 1 Jun 2010 O Hcigadhaigtegs Sn Relativistic Using Gas

Home , Lagrange point

arXiv:1006.0182v1 [astro-ph.HE] 1 Jun 2010 o hcigadhaigtegs sn relativistic Using gas. the heating Ed- mechanisms the detailed and of the shocking esti- is that of to for to independent signal arguments comparable limit, EM luminosity analytic dington observable peak used a an (2010) mate Krolik then present, likely. gas ficient Bome l 09.I uha Mcutratcould counterpart EM LISA an a such with If signal identified (GW) be 2009). wave al. gravitational et electromagnetic a (Bloom luminous interest to distinct, particular counterpart implica- a Of (EM) of astrophysical mergers. possibility the (BH) the is hole in black interest of tions increasing an to e frcn aeshv hw that num- shown a have region, BH papers central recent the of reach cannot ber or can gas known around is disks little accretion very of However, behavior luminous the universe. most the about the of some in produce objects to known the is BHs in sive gas BHs. of merging amount the all of significant almost vicinity some near proposed, potential require of been which variety recently of galaxy large have host A signatures the determined. EM then be process, likely merging could the in binary BH Amtg aaaa 02 hn ta.2009). merger al. before et shortly Chang 2002; luminosity Natarajan in leading & increase BH, (Armitage primary rapid evolving the a the around as to gas power the accretion squeezes inflow of binary an enhancement on an pre- filled calculations dict be onto semi-analytic can other However, accretion gap 2005). the the timescale. stop until Phinney at object effectively & central (Milosavljevic time the should time decouple disk inflow inspiral decoupling the also GW gaseous This of may the the edge when BHs than bi- inner point the the the shorter at losses, of becomes disk evolution GW the the to from including due Lubow When & nary Artymowicz 1996). 1991; decreas- (Pringle 1994, least significantly at it ei- or density, ing lower altogether, much accretion of gap preventing In central ther a accre- have circumbinary evolution. will a disk inspiral that tion believed their is in it disks, times Newtonian late at particularly 1 eetavne nnmrclrltvt N)hv led have (NR) relativity numerical in advances Recent a ntefr faceindssaon igemas- single around disks accretion of form the in Gas etn sd o h oetteqeto of question the moment the for aside Setting rpittpstuigL using 2018 typeset 22, Preprint October version Draft http://lisa.nasa.gov nsaiiy edsusanme fatohsclapiain fthe of applications astrophysical gravit of to headings: counterparts number Subject electromagnetic a producing discuss for We mechanism stability. in asadbcmssihl essal,wiethe while stable, less slightly becomes and mass oteprl etna ae efidta the that find we case, gr Newtonian via evolves purely system the binary the to as problem three-body restricted H ARNEEULBIMPOINTS EQUILIBRIUM LAGRANGE THE ecluaetelcto n tblt fthe of stability and location the calculate We 1. A INTRODUCTION eateto hsc n srnm,JhsHpisUnivers Hopkins Johns Astronomy, and Physics of Department T E tl mltajv 2/16/10 v. emulateapj style X 1 lc oepyis–rltvt rvttoa ae aais nuclei galaxies: – waves gravitational – relativity – physics hole black eeto fasupermassive a of detection if hr ssuf- is there rf eso coe 2 2018 22, October version Draft two eeyD Schnittman D. Jeremy how BHs, ABSTRACT the L L 4 4 L L AND 5 and 4 on oe wyfo h eodr n gains and secondary the from away moves point qiiru on oe oad h secondary the towards moves point equilibrium vralnt scale length a over etos l yaisaecniee tteprl New- purely the at level. considered GW tonian cor- are evolution non-conservative dynamics post-Newtonian these the all of rections, leading-order exception including the with ratios, With mass binary losses. BH frame the of corotating restricted of the range the radia- in a integrate motion to we for of due into paper, equations shrinks this captured re- three-body orbit In get may binary separation that reaction. move the tion binary stars to as large or free trapped at Gas main particle regions test stable plane. a these the being of orbits, objects body out Keplerian massive third planar two three- the circular, with restricted with on system circular, other a each the orbiting i.e., We in problem, merger. orbits of body time stable significant the on at a BHs to focus the lead around gas could of that amount processes dynamical tial signals. luminous more necessarily not but longer, M asi modest a is mass cm daai vlto fteB iay hsrsl imme- result This binary. the BH throughout the orbits of stable evolution adiabatic on remain points Lagrange 20)soe httepop asls u oGWs to sudden due a loss mass to prompt leads the actually al. that et showed O’Neill (2009) 2D, in simulations magneto-hydrodynamic eig hc illa oEdntnrt luminosities scat- Eddington-rate electron require we to to Thus thick lead con- 2010). optically the (Krolik will be is which “sufficient” gas tering, of is the estimate gas that sufficient One dition that present. leading event get fact thus the can in in temperatures, gas counterparts high simula- the bright to The in that to heated fluid suggest (2010). and ideal al. all shocked an et matter Farris as including and and par- tions (2010) (2010) test al. al. now 2010; as et et fields treated al. Bode Meter have gas, et van EM also Mosta in binary and of 2009; ticles 2010) al. BH presence al. et et the (Palenzuela merging Palenzuela vacuum the including a momen- a of out in tem- simulations of angular carried NR disk orbits and Full been inner few energy the efficiently. final in much accrete gas to too tum the has as porarily merger, the lowing L efidta atce ntal ls othe to close initially particles that find We poten- understand to is work this for motivation The 7 L 2 5 g = 5 − arneeulbimpit ntecircular the in points equilibrium Lagrange NABAKHL IAYSYSTEM BINARY HOLE BLACK A IN 1 M steoaiyand opacity the is BH toa-aesignals. ational-wave vttoa aito oss Relative losses. radiation avitational / (10 t,Blioe D21218 MD Baltimore, ity, 7 M M ersls npriua sa as particular in results, se ⊙ R gas .Mr a ilgnrlyla to lead generally will gas More ). ∼ ∼ 10 ρ ρR GM stecaatrsi density characteristic the is decrease 3 ≈ κρR BH 5 /c × & 2 nlmnst fol- luminosity in 10 hntetotal the Then . ,where 1, 26 M 7 2 L 4 ,where g, κ and 0 = L . 4 5 2 Schnittman diately leads to a number of astrophysical predictions. If diffuse gas is captured from the surrounding medium, 1.5 and irradiated by an accretion disk around either BH, it may produce strong, relatively narrow emission lines 1.0 L with large velocity shifts, varying on an orbital time 4 scale. If individual stars are captured, they may eventually get tidally disrupted by one of the BHs shortly 0.5 before merger, providing an enormous source of material for accretion. Stars may also get ejected from the system L L L 0.0 3 1 2

at late times, producing ultra-hyper-velocity stars mov- y/a ing away from the galactic center at a significant fraction of the speed of light. Clusters of stars could also get captured or form in situ at large binary separation, then get -0.5 compressed adiabatically as the BH system shrinks, ulti- L mately collapsing from gravitational instabilities. Com- -1.0 5 pact objects such as neutron stars or stellar-mass BHs would not get tidally disrupted in most cases, but could still possibly be detected with LISA as perturbations to -1.5 the GW signal. -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 While this manuscript was in preparation, a similar x/a work by Seto & Muto (2010) was published, with some of the same conclusions. One major distinction is that they Figure 1. Contours of the Jacobi constant CJ0 for test particles at rest in the rotating frame, for secondary mass µ2 = 0.02. The included first-order post-Newtonian terms in their equa- stable Lagrange points L4 and L5 are located at local minima of tions of motion, while we do not. Also, they “have not CJ0, and form equilateral triangles with the two massive bodies. found notable qualitative differences” between L4 and The co-linear Lagrange points L1, L2, and L3 are located at saddle- L , while we do. Many of our astronomical predictions points of CJ0 and are unstable equilibrium points. The x− and 5 y−axes have been normalized to the binary separation a. are qualitatively similar, but have been derived indepen- dently. The outline for this paper is as follows: in Section 2 we present the three-body equations of motion, including where r1 and r2 are the distances to the primary and sec- GW losses, and analyze the stability and location of the ondary bodies, respectively. These reproduce the classi- equilibrium Lagrange points. These analytic predictions cal Newtonian system in the limit ofn ˙ → 0, i.e. when are tested in Section 3 with numerical simulations of a gravitational radiation reaction is turned off. large number of test particles. In Section 4 we explore in It is well-known from classical mechanics that there more detail some potential astronomical applications of exist five equilibrium points where a test particle can re- these results, and in Section 5 we conclude. main at rest in the corotating frame (¨x =y ¨ =z ¨ = 0) (Murray & Dermott 1999). These points are known as 2. LOCATION AND EVOLUTION OF STABLE LAGRANGE the Lagrange equilibrium points (L1 ...L5), and are plot- POINTS ted in Figure 1 for a binary with mass ratio µ2 = 0.02. We consider a three-body system wherein the larger The co-linear points L1, L2, and L3 are always unstable, bodies have masses M1 and M2 (M1 > M2; M ≡ while the triangular points L4 and L5 are linearly stable M1 + M2) and the third object is a test particle of only for µ2 . 0.0385 (Murray & Dermott 1999). In the negligible mass m0. We define dimensionless masses Newtonian problem, there is a single integral of motion, µ1 ≡ M1/M and µ2 ≡ M2/M. M1 and M2 move on the Jacobi constant: circular, Keplerian orbits around the center of mass at 2 2 2 µ1 µ2 2 2 2 the origin with semi-major axis a(t). Unless otherwise CJ ≡ n (x + y )+2 + − x˙ − y˙ − z˙ . (2) stated, we use geometrized units with G = M = c = 1. r1 r2 In the corotating frame we define coordinates (x,y,z), Setting the particle’s velocity to zero in the corotating with the primary BH located at (−µ2 a, 0, 0) and the sec- frame, the contours of constant CJ0 ≡ CJ(v = 0) in the ondary at (µ1 a, 0, 0). The angular velocity of the system x − y plane define “zero velocity curves,” as plotted in −3/2 in the lab frame is n = a . The binary is orbiting Figure 1. L4 and L5 are located at local minima of CJ0, in the +φ direction for the standard definition of spher- while L1, L2, and L3 are saddle-points (note that CJ has ical coordinates. The equations of motion for the test the form of a negative energy, so that local minima are particle may be written as actually points of maximum potential energy, yet are still stable to small perturbations because of the restoring x + aµ x − aµ 2 2 1 Coriolis forces in the rotating frame). One well-known x¨ =2ny˙ + n x +ny ˙ − µ1 3 + µ2 3 (1a) r1 r2 implication of this feature is that gas bound to either M or M can pass through L and subsequently accrete 2 µ1 µ2 1 2 1 y¨= −2nx˙ + n y − nx˙ − + y (1b) onto the other object, a process known as mass trans- r3 r3 1 2 fer in standard stellar binary evolution. Additionally, µ1 µ2 any particle with CJ > CJ0(L3) would be excluded from z¨= − 3 + 3 z (1c) r1 r2 crossing into the tadpole-shaped regions around L4 and 96 L5. n˙ = µ µ n11/3 , (1d) Murray (1994) explore the effects of generalized non- 5 1 2 Lagrange Points in BH Binaries 3 gravitational forces acting on test particles in the re- y and add to (3b) times x, giving stricted three-body problem, including a detailed study of how the position and stability of the Lagrange equi- ¯ 2 1 1 knr = µ1µ2ay 3 − 3 . (7) librium points are affected. At first glance, our problem r1 r2 would seem quite similar, with then ˙ terms in equations (1a),(1b) acting as psuedo-drag forces: as the binary or- In the limit of small µ2, a ≈ r ≈ r1 at the stable Lagrange bit shrinks and n increases, test particles would feel like points and we can write they are accelerating in the −φ direction in the coro- ¯ µ1µ2 1 tating frame, thus slowing down in the inertial frame. k ≈ 3 sin φL 1 − 3/2 , (8) na (2 − 2cos φL) In this case, Murray (1994) show that L4 and L5 both move in the +φ direction, thus compensating for the in- where φL is the location of the Lagrange point in the ertial drag and still maintainingx ¨ =y ¨ = 0. However, presence of an inertial drag force. When considering the when considering the time-varying nature of n and a in orbital evolution due to GW losses, we can use equation equations (1a)-(1d) due to GW losses, we find that this (6) for k¯. We must also reverse the sign of φL because, as drag force analogy breaks down. Numerical integration mentioned above, the GW drag force is really applied to of equations (1a)-(1d) show that in fact, the opposite oc- M2 and not to the test particle as in Murray (1994). We curs, with L4 and L5 moving in the −φ direction in the now have an expression for the location of the Lagrange rotating frame. points in the presence of GW losses: Remarkably, this result can still be understood via an inertial drag term, if one applies it not to the test par- 1 32 −5/2 sin φL 3/2 − 1 = a . (9) ticle, but to the secondary mass. In order to study the (2 − 2cos φL) 5 evolution of the Trojan asteroids under the influence of a radial migration by Jupiter, Fleming & Hamilton (2000) This relation is plotted as a dashed curve in Figure 2. In- apply an artificial drag force that acts only on Jupiter terestingly, to leading order, the location of the Lagrange in the form F = −kv in the inertial frame. Of course, points is a function only of the binary orbital separation, this causes the secondary to lose energy and angular mo- not the mass ratio, even though the inspiral timescale mentum and thus orbit faster, increasing n as in the BH is strongly dependent on µ2. Note that, when including binary case. Since there is no requirement on the rela- non-conservative forces such as GW evolution, there is a tive masses of the secondary M2 and the test particle m0 lack of symmetry between L4 and L5, with L4 moving to ensure stability of L4 and L5 (Salo & Yoder 1988), we towards the secondary mass, while L5 moves away from can in fact adopt the drag force analysis of Murray (1994) it. We also find differences in the relative stability of to include GW losses by reversing the roles of the sec- orbits around each Lagrange point, as described in the ondary BH and the test particle. In the rotating frame, next section. the inertial drag force is given by Fx = −k(x ˙ − ny) and 3. NUMERICAL TESTS Fy = −k(y ˙ + nx). For stationary particles at the equilibrium points, equations (1a),(1b) become In Figure 2 we plot the position of test particles near L4 and L5 as a function of the binary separation a for 2 x + aµ2 x − aµ1 a range of mass ratios. The initial binary separation is kny¯ = −n x + µ1 + µ2 (3a) r3 r3 a0 = 40M. For the most part, the test particles remain 1 2 very close to their initial locations throughout inspiral, ¯ 2 µ1 µ2 evolving according to equation (9). While our Newtonian knx = n y − 3 + 3 y , (3b) r1 r2 equations of motion do not formally predict an inner- most stable circular orbit (ISCO), we expect the BHs to ¯ where k = k/m0, and we have replaced the GW acceler- plunge and merge around a = 5M, at which point very ation termsny ˙ andnx ˙ with those appropriate for drag little evolution in φL has taken place. The slight offsets in forces Fx/m0 and Fy/m0. the various curves at large a are due to the fact that φ is In order to reproduce the GW evolution of equation measured with respect to the origin, not M1. The inclu- (1d), we must choose the magnitude of the drag force sion of post-Newtonian (PN) corrections to the equations appropriately. This is done by equating the energy loss of motion would likely change the shape of the curves in due to the shrinking orbit with that of the inertial drag Figure 2 slightly, as well as the detailed shape of sta- force. For circular Keplerian orbits, we have bility regions around L4 and L5 (Rosswog & Trautman 1996). It should be noted that Seto & Muto (2010) in- ˙ 2 2/3 Edrag = F · v = −kv = −kn (4) clude only 1PN terms, and do find a significant reduction in stability for certain mass ratios. Yet due to the noto- and riously uneven convergence of most PN expansions (e.g. d m0 m0 −1/3 Buonanno et al. (2009) and references therein), it is not E˙ orbit = − = − n n˙ . (5) dt 2a 3 clear whether their inclusion would add significant phys- ical insight to the problem. Equating (4) and (5), along with (1d) allows us to solve When studying inertial drag forces such as those de- ¯ for k: scribed above, Murray (1994) found that the L4 point 1 n˙ 32 was more stable than L5. Since our problem is the k¯ = = µ µ n8/3. (6) 3 n 5 1 2 mirror-image of that one, it is not surprising then that we find greater stability around L5 when including GW Following Murray (1994), we multiply equation (3a) by evolution. As shown in the upper panel of Figure 2 for 4 Schnittman

the large-scale stability of orbits around each Lagrange 70 point. The ﬁrst was a simple number count: the fraction of particles that remain bound around the equilibrium 65 point. In this context, a particle is considered bound if 60 its instantaneous Jacobi constant CJ is less than CJ0(L3) for the present orbital separation, corresponding to the ) 4 tadpole-shaped regions around L4 and L5 in Figure 1.

(L 55 φ µ =0.038 Additionally, any particle that escapes with r > 10a or 2 50 µ =0.03 approaches closer than 3µ1 from the primary or 3µ2 from µ2=0.02 the secondary is removed from the system. The second µ2=0.01 45 2 method of measuring stability was to calculate the total volume ﬁlled by bound particles around each Lagrange 40 point. A larger volume corresponds to lower stability, as 5 10 15 20 25 30 the particles begin to escape the bound libration region. binary separation The fraction of bound particles fbound is plotted in the top panels of Figure 3, normalized to the bound -50 number at binary separation of a = 35M. The particles begin in a small cloud of radius δR around each -55 Lagrange point with random velocities, isotropic with 1/2 hvi = vorb(δR/a) in the corotating frame. We se- -60 lect δR to be large enough so that the cloud of parti- )

5 cles roughly ﬁlls the phase-space volume of stable or-

(L -65 bits around L and L . In practice, this typically corre- φ 4 5 µ =0.038 2 sponds to half of the particles getting ejected on roughly -70 µ =0.03 µ2=0.02 a dynamical time after the start of the simulation (hence 2 µ =0.01 fbound is defined relative to a = 35M, not a = 40M). Af- -75 2 ter a stable cloud is formed around each Lagrange point, -80 marginally bound particles are slowly ejected from the 5 10 15 20 25 30 system (or accreted onto one of the BHs) as the binary binary separation system evolves adiabatically. From Figure 2, we might expect more particles to get ejected from L4, which ap- Figure 2. Upper: The location of test particles initially at rest pears to be less stable, at least at linear order. However, near the L4 Lagrange point in the rotating frame. As the binary as evident in Figure 3, L4 retains more bound particles system evolves to smaller separation a, the L4 point moves towards the secondary BH, located at φ = 0. For µ2 = 0.038, L4 becomes when µ2 = 0.01, L5 appears more stable for µ2 = 0.02, linearly unstable around a = 15M, and the test particle is ejected and they are nearly identical for µ2 =0.038. at a = 10M. Lower: The location of test particles initially at rest It is not entirely clear why this peculiar behavior oc- near the L5 Lagrange point. Towards the end of the BH inspiral, curs, with no apparent trend in global stability as a the L5 point moves away from the secondary. In both figures, we plot the analytic prediction of equation (9) as a dashed curve. function of mass ratio, nor why one Lagrange point or the other should be more stable for a given mass ratio. As described in Murray & Dermott (1999), test parti- µ2 = 0.038, a test particle initially very close to L4 will cles near the stable Lagrange points move along epicy- eventually get ejected from the equilibrium configuration cles with two distinct frequencies. For certain values of as the binary approaches merger. On the other hand, µ2, these libration frequencies have integer commensu- test particles librating around L5 appear to become more rabilities, which naturally leads to strong resonant in- stable, asymptotically approaching the exact equilibrium teractions (Deprit & Deprit-Bartholomé1967). For ex- position. As mentioned in Erdi et al. (2009), for binaries ample, when µ2 = 0.0243, both L4 and L5 are stable with µ2 slightly above the critical value for linear stabil- under standard linear analysis, yet the two epicyclic fre- ity (0.0385 . µ2 . 0.0401), there still exist bound orbits quencies have a ratio of 2 : 1, leading to resonant ex- that librate around L4 and L5 with moderate amplitude. citations and ejection of all test particles after a few We find that for these systems, the inclusion of radiation dynamical times (Erdi et al. 2009; Seto & Muto 2010). reaction increases the stability of L5 by damping the li- Murray (1994) shows that these characteristic frequen- bration amplitude. On the other hand, the libration am- cies change in the presence of inertial drag forces, which plitude of particles around L4 grows rapidly towards the we have shown above to be directly analogous to GW end of inspiral, as in the linear case in Figure 2. In con- evolution. Thus we suspect that the different behavior trast, Seto & Muto (2010) find no difference between L4 seen in fbound(t) for different Lagrange points is due to and L5, although they only consider a few test particles a combination of resonant interactions and asymmetric around each Lagrange point. shifts in the libration frequencies due to the binary evo- To further investigate this asymmetry in stability, we lution via GW losses. carried out orbital integrations for large numbers of test In the bottom panels of Figure 3 we plot the total particles distributed in clouds around each Lagrange volume Vbound occupied by bound particles, normalized 3 point for masses µ2 = 0.01, 0.02, and 0.038. This al- to the overall scale of the binary a . In the limit of lowed us to probe the regions relatively far from the a very small δR/a, where the test particles are well equilibrium points, where standard linear stability anal- within the linear regime, the cloud around each Lagrange ysis breaks down. We used two methods for estimating Lagrange Points in BH Binaries 5

1.0 1.0 1.0

0.8 0.8 0.8 µ =0.01 µ =0.02 µ =0.038 2 2 2 L L L 0.6 4 0.6 4 0.6 4 L L L

bound 5 5 5 f 0.4 0.4 0.4

0.2 0.2 0.2

0.0 0.0 0.0 0.025 5 10 15 20 25 30 0.02535 5 10 15 20 25 30 0.02535 5 10 15 20 25 30 35

0.020 0.020 0.020

3 0.015 0.015 0.015 /a

bound 0.010 0.010 0.010 V

0.005 0.005 0.005

0.000 0.000 0.000 5 10 15 20 25 30 35 5 10 15 20 25 30 35 5 10 15 20 25 30 35 binary separation (GM/c2) binary separation (GM/c2) binary separation (GM/c2)

Figure 3. Upper: The fraction fbound of test particles that remain bound in librating orbits around L4 (solid curves) and L5 (dashed curves) for a range of mass ratios. The particles are initially distributed in a small cloud around each Lagrange point with binary separation a = 40M. fbound is normalized to the number of bound particles when a = 35M. Lower: The volume of space occupied by bound test 3 particles around L4 (solid curves) and L5 (dashed curves), normalized by the characteristic volume of the system, a .

point will actually expand adiabatically relative to the bi- consistent trend of larger Vbound around L4 than L5 for nary separation, although shrinking on an absolute scale. all values of µ2, as suggested by Figure 2. However, this Fleming & Hamilton (2000) show this evolution to scale may be less a measure of local stability than a conse- as a9/4: quence of the fact that L4 moves toward the secondary, 9/4 Vf af while L5 moves away from it. Thus the two regions be- = , (10) gin to sample different potentials, and the gravitational Vi ai force of the secondary acts to disrupt the cloud around where Vi and Vf are the initial and final volumes of test L4 more strongly than that around L5, in turn leading particle clouds. We have confirmed this result numer- to a larger Vbound. ically with simulations starting with δR ≪ a. On the We should add as a word of caution, that many of these other hand, if δR/a is large enough such that the stable numerical stability results may be altered somewhat by region of phase space is completely filled, the normalized the inclusion of additional PN terms to the equations of volume V/a3 should be constant, since the entire prob- motion (Seto & Muto 2010). In general relativity, unlike lem is scale-invariant. Any adiabatic expansion will not the case of Newtonian gravity, test particles moving in increase the volume of bound orbits, but rather decrease the Kerr metric have three distinct epicyclic frequencies. the total number of particles in these regions, as seen in These additional frequencies may lead to more compli- the upper panels of Figure 3. cated resonant interactions and thus change the behav- 3 We actually find a gradual increase in Vbound/a ior of fbound(a) and Vbound(a), at least at a quantitative as the binary shrinks, yet still much slower than the level. Yet our intuition suggests that the qualitative be- Fleming & Hamilton (2000) result of (V/a3) ∼ a−3/4, havior shown in Figure 3 should remain unchanged. Fu- suggesting that the bound region in phase space is in fact ture work will investigate this question in greater detail. 3 filled. The increase in Vbound/a during the adiabatic in- 4. ASTROPHYSICAL APPLICATIONS & spiral phase (a 10M) is likely due to a combination of 4.1. effects. First, we expect a net increase in global stability Formation Mechanisms caused by the GW losses, analogous to the effects of in- In the previous Sections, we showed analytically and ertial drag described above. Secondly, there is also some numerically how the location and stability of the L4 and inherent error in measuring the volume of bound orbits, L5 Lagrange points evolve in the presence of radiation based on our crude metric of the test particles’ instanta- reaction. We found that test particles, once captured neous Jacobi constants. Some fraction of particles that into libration orbits around one of the stable Lagrange have moved outside the stable region may still appear points, will in general remain there throughout the adi- bound for a few dynamical times before getting ejected abatic inspiral phase, as the entire length scale of the or accreted, and thus artificially contribute to Vbound. At system steadily shrinks. However, the initial premise smaller a, where the libration time becomes comparable was not addressed: will there even be gas or stars or to the inspiral time, the system no longer evolves adi- debris present in the first place? How might matter get abatically, and a cloud of particles can get “frozen” at captured into these orbits at large binary separations of roughly constant volume, leading to the rapid increase a & 1 pc? Furthermore, are binary BHs with µ2 . 0.04 3 in Vbound/a for a . 10M. Interestingly, we do see a astrophysically likely, or even possible? 6 Schnittman

Even in a system as well-studied as the Sun-Jupiter bi- orbit, surrounded by ample debris that could easily get nary, there is still no clear consensus on the origin of the trapped into resonant orbits. Trojan asteroids, despite the fact that orbital elements Another mechanism that could explain both the low- are now known for over 4000 individual objects. Rabe mass secondary and also provide a supply of stars would (1972) proposed that they may be captured comets, while be the tidal disruption of a globular cluster by the Shoemaker et al. (1989) and Kary & Lissauer (1995) primary BH. If the globular cluster hosts an IMBH suggested a more local origin as planetesimals in the early (Miller & Hamilton 2002), it could form a hard binary solar system. In most theories, some form of dissipative with the primary, and the surrounding stars could get force is required to capture the Trojans (Yoder 1979), captured into stable orbits around L4 and L5. At this such as nebular gas drag (Murray 1994; Kary & Lissauer point, it would be impossible to carry out any quan- 1995) or planetesimal collisions (Shoemaker et al. 1989). titative analysis of the relative rates and likelihoods of Additionally, an increase in Jupiter’s mass via accretion these different scenarios, considering the remarkable lack of the surrounding disk could lead to asteroids getting of observational evidence for even a single SMBH binary captured (Marzari & Scholl 1998; Fleming & Hamilton with orbit smaller than ah. Rather, we might realisti- 2000). Analogous to our BH binary system, the inward cally hope to discover such a system with help from one migration of Jupiter may have also affected the capture of the EM signatures described below. and subsequent evolution of the Trojans (Yoder 1979; Fleming & Hamilton 2000). 4.2. Tidal Disruptions The characteristic separation at which point a hard BH Tidal disruptions of solar-type stars by SMBHs are ex- binary (i.e., the orbital potential is dominated by the pected to occur at a rate of ∼ 10−4 yr−1 per MW-type two BHs, and not by the surrounding stars) is formed galaxy (Wang & Merritt 2004), and are excellent tools is comparable to the primary BH’s radius of influence for detecting otherwise-quiescent nuclei at cosmological (Merritt 2006): distances (Lidskii & Ozernoi 1979; Rees 1988). Typical luminosities are comparable to the Eddington limit, and q Rinfl ah = , (11) the timescale for the rise and subsequent decay of the (1 + q)2 4 light curve is on the order of weeks to months (Ulmer 1999; Halpern et al. 2004), which is ideal for detection by where q = M /M and 2 1 wide-field time-domain surveys such as Pan-STARRS2 GM M σ −2 or LSST3 (Gezari et al. 2009). Tidal disruptions have Rinfl = ≈ 1.1pc − , recently been proposed as tools for detecting BH merg- σ2 107M⊙ 200kms 1 ers or recoiling merger remnants, in some cases long after (12) the merger occurs (Komossa & Merritt 2008; Chen et al. with σ being the one-dimensional velocity dispersion of 2009; Stone & Loeb 2010). the galaxy’s bulge. In the limit of small µ2 ≈ q ≪ 1, we In our scenario, the main-sequence stars are captured have into resonant motion around L4 or L5 at large separation −2 −3 µ2 M σ (a ∼ ah) and then move in closer to the primary as the ah ≈ 5 × 10 pc 7 −1 , binary separation slowly shrinks. A star of mass M∗ and 0.02 10 M⊙ 200kms (13) radius R∗ will get disrupted when the tidal forces from 4 the BH are comparable to the star’s own self-gravity, corresponding to ah ∼ 10 M in geometrized units. which occurs at a distance R from the primary: Within this region, there are typically thousands of stars td (Hopman & Alexander 2006), and in active galactic nu- 1/3 M1 clei (AGN), a great deal of gas and dust as well. It is Rtd ≈ R∗ . (14) also roughly the separation at which point the binary M∗ will merge within a Hubble time due to GW losses alone In our numerical tests of Section 3, we found that during (Peters 1964), and thus equations (1a)-(1d) are appro- inspiral, some fraction of test particles get ejected from priate. L4 or L5 and disrupted by M2, the the majority remain For AGN with massive accretion disks, it is bound until the final merger stage. For stars that re- possible for significant star formation to occur in main bound to L4 or L5, the gravitational tidal forces are the region where the disk becomes self-gravitating dominated by the primary, and thus the relevant mass in (Levin & Beloborodov 2003). For standard thin disk equation (14) is M1. models Shakura & Sunyaev (1973), this self-gravitating In Figure 4 we plot the orbital separation of a BH bi- radius actually corresponds closely to ah. Since galaxy nary as a function of time before merger (black curves). mergers are often associated with gas inflow and quasar The dashed horizontal lines correspond to the points at activity, it is quite possible that stars could form in an ac- which (top-to-bottom) an O/B star, a solar-type star, cretion disk just as a hard BH binary is forming, thereby or a white dwarf would get tidally disrupted by the pri- 6 7 trapping the stars around the stable Lagrange points. A mary BH of mass 10 M⊙ (top) or 10 M⊙ (bottom), and similar possibility, suggested by Goodman & Tan (2004), secondary mass µ2 = 0.01. The vertical dotted lines 4 is that supermassive stars with M & 10 M⊙ could denote where the binary separation reaches the ISCO grow out of main-sequence seeds embedded in the ac- and the BHs merge, and the vertical dashed lines cor- cretion disk, quickly evolving and then collapsing to respond to the point where the binary enters the LISA intermediate-mass black holes (IMBHs). This scenario is particularly attractive, as it naturally provides a mecha- 2 http://pan-starrs.ifa.hawaii.edu 3 http://www.lsst.org nism for making a binary BH with µ2 . 0.01 on a circular Lagrange Points in BH Binaries 7

14 Table 1 10 Fraction of test particles that, during inspiral, are either ejected from the system or captured by one of the BHs. R td, O/B 1013 µ2 = 0.01 µ2 = 0.02 µ2 = 0.038 →∞ R f(L4 ) 0.13 0.11 0.13 td, sun f(L5 →∞) 0.20 0.12 0.21

a (cm) f(L4 → M1) 0.16 0.33 0.19 12 10 f(L5 → M1) 0.15 0.05 0.10 f(L4 → M2) 0.14 0.27 0.18 R 5 → 2 td, WD f(L M ) 0.21 0.11 0.20 1011 3 4 0.01 0.10 1.00 10.0 100 10 10 by the BHs. days before merger 4.3. Ultra-Hyper-Velocity Stars 1014 Of the stars that do get ejected from the binary system R td, O/B during inspiral, we find a wide range in their velocities at infinity. To first order, one would expect the ejection ve- 1013 R locity to be comparable to the orbital velocity. However, td, sun particles ejected at late times when the orbital velocity is highest are also coming from the deepest potentials, so

a (cm) are slowed more as they escape from the system. In prac- 1012 tice, we find that net effect is a slight increase in the av- R erage escape velocity at late times, but somewhat flatter td, WD than the orbital velocity scaling of a−1/2. This is likely due to the fact that the most efficient three-body sling- 11 10 shot interactions require the test particle to come within 3 4 0.01 0.10 1.00 10.0 100 10 10 a small fraction of a from the secondary, and when a be- days before merger comes smaller, these close approaches more likely lead to capture or disruption than ejection. Figure 4. Upper: Tidal disruption radius for a star orbiting a 6 10 M⊙ BH, as a function of time before merger. The secondary From equation (14), we see that a solar-type star would 4 6 BH has mass 10 M⊙. Lower: Tidal disruption radius for a star get tidally disrupted by a 10 M⊙ BH at a distance of 7 7 orbiting a 10 M⊙ BH, as a function of time before merger. The sec- ∼ 50M, far outside the ISCO. For M1 = 10 M⊙, how- 5 ondary BH has mass 10 M⊙. From top to bottom, the horizontal ever, a star could survive to within 10M of the primary. lines correspond to the tidal disruption radius of a main sequence From that separation, we find that test particles are O/B star, a solar-type star, and a 0.7 M⊙ white dwarf. The vertical dotted line corresponds to the point at which the secondary ejected with mean velocities of 0.1c, roughly 50 times BH reaches the ISCO and the vertical dashed lines the point at greater than the radial velocities of hyper-velocity stars which the binary enters the LIGO band (fGW ≈ 1 mHz). (HVSs) observed in the outer regions of the Milky Way (Brown et al. 2009). Of course, any star moving that fast would be observable for a much shorter time; only 6 GW frequency band. For M1 = 10 M⊙, a solar-type a few million years after the BH merger. At the same star would get disrupted roughly a year before merger. time, this allows younger, more massive stars to survive At that point, LISA would already be able to provide an out to the edge of the galaxy, making them even easier error-box of ∼ 100 square degrees for the source location to identify against the field stars. It is conceivable that on the sky (Lang & Hughes 2008), which could be easily the BH at the center of our galaxy may have experienced monitored for luminous tidal disruption flares out to red- a minor merger with µ2 ∼ 0.01 in the past 10 Myr, an shifts z ∼ 0.2 or so (Gezari et al. 2009). In all cases with event which may be convincingly inferred by the detec- 5 tion of such an ultra-HVS at a distance of ∼ 50 − 100 M1 & 10 M⊙, white dwarfs would not get disrupted, but rather plunge directly into the BH. kpc. In addition to the stars that get disrupted at late times, 4.4. we saw in the simulations of Section 3 that there are also Compressed Stellar Systems some particles that get captured by either M1 or M2 dur- In the event that many stars are initially trapped ing the inspiral phase, thus leading to tidal disruption, around L4 or L5, the gravitational interactions between or get ejected entirely from the system. Table 1 shows the test particles may become important, an effect that the fraction of test particles that meet one of these var- we have neglected in this work. Especially for larger sys- 8 ious fates during inspiral, i.e., when the binary separa- tems with M & 10 M⊙, one could even imagine forming tion is 5M ≤ a(t) ≤ 35M. Of course, if equation (14) is globular clusters around the Lagrange points, although met, then even stars on stable orbits are disrupted. As from equation (13) it is clear they would necessarily in Figure 3, there are no clear trends with varying µ2, be quite compact. When the core of a normal globu- but we do confirm that particles around L5 are generally lar cluster grows too dense, it can undergo a collapse more likely to get ejected from the system, while particles via gravitational instability, at the same time expanding around L4 are more likely to get captured or disrupted the orbits of the outer stars, due to conservation of en- 8 Schnittman ergy. In our BH binary system, the adiabatic compres- fraction could likely get trapped into orbits around L4 sion of orbits may trigger core collapse of the trapped and L5. If cold enough, this gas could get compressed cluster, while simultaneously preventing any outer stars by the evolving binary potential and form stars or even from escaping, thereby enhancing the stellar density fur- clusters of stars, as discussed above. More likely, the ther. If the density becomes high enough, stellar col- gas will be rather hot, with thermal velocities compara- lisions will become common, and the system could act ble to the libration velocities of test particles filling the as a breeding ground for supermassive stars and IMBHs stable volume around L4 and L5. One very rough esti- (Portegies Zwart et al. 1999). mate of this thermal motion may come from inspection of equation (2) and Figure 1. Trapped particles will have 4.5. Compact Objects CJ0(L4) . CJ . CJ0(L3), so typical velocities in the ro- Many of the formation scenarios considered above nat- tating frame of urally lead to high-mass stars forming in the dense 2 −1 gaseous environments around the SMBH binary. These v ≈ CJ(L3, v = 0) − CJ(L4, v = 0) ≈ 2µ2a (15) stars could then evolve and collapse into neutron stars or stellar-mass BHs on timescales short compared to the (Murray & Dermott 1999). Combining with equation inspiral time. Unlike main sequence stars that would get (13) for the separation at the time of binary formation, tidally disrupted well before the merger, compact objects we get an initial thermal velocity of would survive unperturbed until they either get ejected σ v ≈ 600kms−1 , (16) or captured whole by one of the BHs. In that case, they therm 200kms−1 would also contribute to the GW signal produced by the 6 4 7 binary. For a system with M1 = 10 M⊙, M2 = 10 M⊙, or roughly 4 × 10 K. Somewhat remarkably, vtherm is and m0 = 10M⊙, the stellar-mass BH would contribute independent of µ2 or M1 (except indirectly through the an additional 0.1% to the waveform amplitude. At a dis- M − σ relationship). tance of 1 Gpc, this small perturbation would still be We can estimate an upper limit on the density of this observable with LISA. gas ρ0 by requiring it to be stable to gravitational col- However, if the compact object is orbiting exactly at lapse. In other words, the total mass should not exceed the L4 or L5 with no libration, it would contribute a the Jeans mass: pure sinusoid component to the waveform with identi- 3/2 1/2 cal frequency as the dominant wave, only with a phase 3 5kBT 3 ◦ 0.005 ρ0ah ≤ , (17) shift of ±120 . Thus it would be indistinguishable from Gmp 4πρ0 a normal binary on a circular orbit with slightly different masses and phase. On the other hand, if the compact ob- with kB the Boltzmann constant and mp the proton ject is librating with typical amplitude ∆φ ≈ 30◦, as we mass. The leading factor of 0.005 comes from the vol- might expect in the general case, the waveform will con- ume of the stable region, estimated from Figure 3. Tak- 2 tain a distinct signal at the libration frequency (in this ing mpvtherm =3kBT , we get example, ω ≈ 0.27ω ) which may still be detectable. lib orb 2 More likely, this extra signal would be confused with a vtherm ρ0 ≤ 35 2 , (18) spin-orbit or spin-spin precession frequency, effectively Gah degrading the overall confidence in the binary parame- ter estimation. In some cases, the stellar-mass BH might corresponding to a total mass in gas of get ejected during inspiral, but with small enough velocity that it doesn’t escape the system, but rather re- 3 4 µ2 M Mgas ≈ 0.005ρ0ah ≤ 7 × 10 M⊙ 7 , mains bound on a larger, eccentric orbit even after the 0.02 10 M⊙ two massive BHs merge, ultimately merging much later (19) as an extreme mass-ratio inspiral (EMRI) source. which is far more than enough material to fuel a bright As mentioned above, the secondary BH and the “test EM counterpart. particle” can actually have comparable masses and still Even if the density of the trapped gas is orders of mag- maintain stability of the Lagrange points, as long as nitude smaller, it should still be sufficient to produce ob- µ2 . 0.0385. Thus a stable three-body BH system could servable emission lines if irradiated by an AGN or quasar 6 4 exist with M1 ≈ 10 M⊙ and M2 ≈ M3 ≈ 10 M⊙, in powered by accretion onto one of the BHs. For a cloud of which case the orbital libration could be seen very clearly gas filling the stable region around each Lagrange point, in a LISA waveform. The probability of such a system the covering fraction would be on the order of a few per- existing is perhaps not as small as one might think, espe- cent. Since the thermal velocity is roughly a factor of cially if both IMBHs form out of a massive accretion disk, 10 smaller than the orbital velocity, the emission lines not unlike Saturn’s moons Janus and Epimetheus. In- would be somewhat narrower than a classical broad line deed, only 15 years ago, the possibility of “Hot Jupiters” originating from the same radius. The coherent shape with µ2 & 0.001, orbiting very close to their host stars and motion of the cloud would also give rise to a large was considered quite remote, and now we know of hun- velocity offset relative to the host galaxy or narrow emis- dreds of such systems. The same may very well be true sion line clouds residing at a greater distance from the of SMBHs and their surrounding gas disks. central engine. This offset in wavelength would vary periodically with the binary orbit (T at our fiducial value 4.6. Diffuse Gas orb of ah would be on the order of 10 years), giving strong In the circumbinary accretion disk model, ample gas evidence for the existence of a BH binary. Such an ef- and dust will surround the BH binary, of which some fect could be detected by large spectroscopic surveys like Lagrange Points in BH Binaries 9 the Sloan Digital Sky Survey, a technique already in fre- eccentric orbits with periods 2/3 that of Jupiter. This quent use by observers searching for binary quasars or re- ensures that they avoid close encounters with Jupiter coiling black holes (Bonning et al. 2007; Komossa et al. at apocenter, where they spend most of their time. In 2008; Boroson & Lauer 2009) the corotating frame, a snapshot of their positions will appear to fill out a triangle with vertices aligned with 5. DISCUSSION L3, L4, and L5 (Bro˘z& Vokrouhlický2008). Guéron We have analyzed the circular restricted three-body (2006) showed that this resonance is stable even in the problem including leading-order terms for gravitational extreme relativistic limit, but still with small mass ratios wave losses, in particular studying the behavior of test (µ2 ≈ 0.001). Our own preliminary results suggest that particle orbits near the stable Lagrange points L4 and this 2 : 3 resonance is linearly stable for mass ratios at L5. We find that these orbit remain stable throughout least as large as µ2 ≈ 0.1, but we leave a comprehensive most of the adiabatic inspiral phase, and their locations study of the problem, including GW losses, to a future shift relative to the Newtonian limit as the binary en- paper. ters the late phases of inspiral and merger. Analogous to In addition to all the Newtonian resonance behavior the case of inertial drag forces in the planetary problem, present in the three-body problem, the two BHs may also the non-conservative GW loss terms lead to an asymme- get locked into spin-orbital resonances with each other try between L4 and L5: test particles around L4 move during inspiral (Schnittman 2004). Coupled with the closer to the secondary BH, while those initially around non-degenerate orbital frequencies of test particles in a L5 move away from it. While there appear to be compli- Kerr background, the inclusion of spinning BHs would cating resonances that affect the relative stability of the introduce many new degrees of freedom that may affect two regions differently for different mass ratios, a general the stability of L4 and L5. While we do not think ad- trend is that the inclusion of GW losses increases the sta- ditional PN terms in the equations of motion will signif- bility of L5 while decreasing the stability of L4. Particles icantly change our results on a qualitative level, future initially around L4 are more likely than those around L5 work that addresses these questions in more detail is cer- to get captured or disrupted by M2, while particles near tainly warranted. L5 are more likely to get ejected from the system entirely. However, in both cases, the vast majority of particles get The author would like to thank Doug Hamilton, Matthew neither ejected nor captured until the binary separation Holman, Scott Hughes, Julian Krolik, David Merritt, and is on the order a . 5M, at which point the Newtonian Cole Miller for helpful discussions and comments. This equations of motion break down entirely and everything work was supported by the Chandra Postdoctoral Fel- plunges into the primary BH. lowship Program. The stability of the L4 and L5 Lagrange points throughout most of the inspiral provides a natural mech- REFERENCES anism for shepharding material from relatively large radii down to the central BH, in turn producing a luminous EM counterpart to the GW signal from a BH merger. Armitage, P. J., & Natarajan, P. 2002, ApJ 567, L9. The detailed physics behind such an EM signal are Artymowicz, P., & Lubow, S. H. 1994, ApJ 421, 651. Artymowicz, P., & Lubow, S. H. 1996, ApJ 467, L77. well beyond the scope of this paper, but early results Bloom, J. S. 2009, White paper for Astro2010 Decadal Survey, from numerical relativity simulations suggest that bright [arXiv:0902.1527]. EM flares should be quite general, provided enough Bode, T., Haas, R., Bogdanovic, T., Laguna, P., & Shoemaker, D. gas is available (van Meter et al. 2010; Bode et al. 2010; 2010, ApJ 715, 1117. Farris et al. 2010). In addition to the prompt emission Bonning, E. W., Shields, G. A., & Salviander, S. 2007, ApJ 666, L13. associated with the final merger, we have shown that Boroson, T. A., & Lauer, T. R. 2009, Nature 458, 53. hyper-velocity stars with speeds as high as 0.1c might Brown, W. R., Geller, M. J., & Kenyon, S. J. 2009, ApJ 690, 1639. be observable, flying away from the galactic center for Bro˘z, M., & Vokrouhlický, D. 2008, MNRAS 390, 715. millions of years after the merger. With sufficient col- Buonanno, A., Iyer, B. R., Ochsner, E., Pan, Y., & umn densities of diffuse gas trapped around L and L , Sathyaprakash, B. S. 2009, Phys. Rev. D 80, 084043. 4 5 Chang, P., Strubbe, L. E., Menou, K., & Quataert, E. 2009, a precursor signal could take the form of highly red/blue- [arXiv:0906.0825]. shifted emission lines, narrower than typical broad lines, Chen, X., Madau, P., Sesana, A., & Liu, F. K. 2009, ApJ 697, with the line offset varying periodically with the BH or- L149. bit. Deprit, A., & Deprit-Batholomé, A. 1967, AJ 72, 173. Erdi, B., Forgacs-Dajka, E., Nagy, I., & Rajnai, R. 2009, Celest. For the majority of the results in this paper, we require Mech. Dyn. Astr. 104, 145. that the secondary mass be relatively small: µ2 ≤ 0.0385 Farris, B. D., Liu, Y.-K., & Shapiro, S. L. 2010, Phys. Rev. D 81, to guarantee linear stability of L4 and L5. It is unclear 084008. how common BH binaries with such mass ratios might Fleming, H. J., & Hamilton, D. P. 2000, Icarus 148, 479. be, and whether they are more likely to form via galac- Gezari, S., et al. 2009, ApJ 698, 1367. Goodman, J., & Tan, J. C. 2004, ApJ 608, 108. tic mergers or in situ, via the collapse of supermassive Guéron, E. 2006, Class. Quantum Grav. 23, 673. stars, accretion from a surrounding disk, or capture from Halpern, J. P., Gezari, S., & Komossa, S. 2004, ApJ 604, 572. a tidally stripped globular cluster. However, there are Hopman, C., & Alexander, T. 2006, ApJ 645, 1152. other resonances in the restricted three-body problem Kary, D. M., & Lissauer, J. J. 1995, Icarus 177, 1. that may lead to qualitatively similar behavior: the sta- Komossa, S., & Merritt, D. 2008, ApJ 683, L21. Komossa, S., Zhou, H., & Lu, H. 2008, ApJ 678, L81. ble transport of matter from large radii down to the ISCO Krolik, J. H. 2010, ApJ 709, 774. at the time of merger. One example from the solar sys- Lang, R. N., & Hughes, S. A. 2008, ApJ 677, 1184. tem is the Hilda family of asteroids that occupy stable Levin, Y., & Beloborodov, A. M. 2003, ApJ 590, L33. 10 Schnittman

Lidskii, V. V., & Ozernoi, G. M. 1979, AZh. Pis’ma 5, 28 Pringle, J. E. 1991, MNRAS 248, 754. (English transl. in Sov. Astron. Lett. 5, 16). Rabe, E. 1972, in Motion, Evolution of Orbits, and Origin of O’Neill, S. M., Miller, M. C., Bogdanovic, T., Reynolds, C. S., & Comets. Proceedings of 45th IAU Symposium, pp. 55-60. Schnittman, J. D. 2009, ApJ 700, 859. Springer-Verlag, New York/Berlin. Marzari, F., & Scholl, H. 1998, Icarus 131, 41. Rees, M. 1988, Nature 333, 523. Merritt, D. 2006, ApJ 648, 946. Rosswog, S., & Trautman, D. 1996, Plant. Space Sci. 44, 313. Miller, M. C., & Hamilton, D. P. 2002, MNRAS 330, 232. Salo, H., & Yoder, C. F. 1988, A& A 205, 309. Milosavljevic, M., & Phinney, E. S. 2005, ApJ 622, L93. Schnittman, J. D. 2004, Phys. Rev. D, 124020. Mosta, P., Palenzuela, C., Rezzolla, L., Lehner, L., Yoshida, S., Seto, N., & Muto, T. 2010, Phys. Rev. D, in press. Pollney, D. 2010, Phys. Rev. D 81, 064017. [arXiv:1005.3114]. Murray, C. D. 1994, Icarus 112, 465. Shakura, N. I., & Sunyaev, R. A. 1973, A&A, 24, 337 Murray, C. D., & Dermott, S. F. 1999, Solar System Dynamics, Shoemaker, E. M., Shoemaker, C. S., & Wolfe, R. F. 1989, in Cambridge University Press, Cambridge. Asteroids II, pp. 487-523. Univ. of Arizona Press, Tucson. Palenzuela, C., Anderson, M., Lehner, L., Liebling, S. L., & Stone, N., & Loeb, A. 2010, [arXiv:1004.4833]. Neilsen, D. 2009, Phys. Rev. Lett. 103, 081101. Ulmer, A. 1999, ApJ 514, 180. Palenzuela, C., Lehner, L., & Yoshida, S. 2010, Phys. Rev. D 81, van Meter, J. R., Wise, J. H., Miller, M. C., Reynolds, C. S., 084007. Centrella, J., Baker, J. G., Boggs, W. D., Kelly, B. J., & Peters, P. C. 1964, Phys. Rev. 136, 1224. McWilliams, S. T. 2010, ApJ 711, L89. Portegies Zwart, S. F., Makino, J., McMillan, S. L. W., & Hut, P. Wang, J., & Merritt, D. 2004, ApJ 600, 149. 1999, A & A 348, 117. Yoder, C. F. 1979, Icarus 40, 341.