![Arxiv:1206.4316V2 [Astro-Ph.SR] 6 Dec 2013 and (SMA) a Ofiuain:A Inr Iay(Ihmasses (With Binary “Inner” an Configurations: Cal Evolution](http://data.docslib.org/img/7edc1845b9c50514c03183e57b3341e7-1.webp)
arXiv:1206.4316v2 [astro-ph.SR] 6 Dec 2013 and (SMA) a ofiuain:a inr iay(ihmasses (with binary “inner” an configurations: cal evolution. com- then triple through two can produced with We likely binaries are triples. stellar objects least pact close at many are of that mag. conclude sample 42 10 a that than noted brighter surveyed and have binaries, (2006) contact Rucinski & Pribulla httefato ftilsadhge utpe among multiples higher found (10 and also period triples He with of binaries companion. fraction additional the one that least at have hwdta 0 fbnr tla ytm ihperiod with systems stellar binary sub- of (1997) < actually Tokovinin 40% is confi- observed. that systems reasonably that showed triple than be of greater faint number can stantially finding the least we against that at dent companions, effects are selection distant see the and we al. Given stars et Probably, bright Eggleton 2007). orbit. 1997; (Tokovinin the wider systems much of binary (double) a 50% or- on is than body binary more third (inner) a the by which 2007). bited in al. triples, et must hierarchical systems Eggleton be these arguments, 1997; stability Tokovinin dynamical From (e.g., Nature in a,M 91,Uie States. United 59718, MT man, Evanston USA University, 60208, Northwestern (CIERA), Astrophysics Cambri 02138 St.; USA Garden MA, 60 Astrophysics, for Center Smithsonian [email protected] ogtr tblt ftil ytmrqie hierarchi- requires system triple of stability Long-term rpittpstuigL using 2018 typeset 21, Preprint August version Draft rpeselrssesaeblee ob eycommon very be to believed are systems stellar Triple 0dy,i hc h rmr sadaf(0 dwarf a is primary the which in days, 10 † 3 2 1 m eateto hsc,MnaaSaeUiest,Boze- University, State Montana Physics, Harvard- of in Department Research and Computation, Exploration Interdisciplinary and for Center Theory for Institute isenFellow Einstein 2 nanal elra ri ihsm-ao axis semi-major with orbit Keplerian nearly a in ) a iecl ftefre smc ogrta h iecl ftesecu the of timescale the than longer much is and eccentricity former perturbations. the the increase of further timescale Kozai-Lid can eccentric the corrections the where eccentric binary, Newtonian to resonant-like inner comparable a comparable-mass to a is rise with timescale give exc interactions pa Relativity can three-body other General ones, the In Relat the body three General where one. the the Newtonian systems with secular where combined the corrections space than Newtonian phase e shorter of Relativity magnitude General parts of that the i.e., prod with lore, can consistent standard mechanism The the Kozai-Lidov explore eccentricities. the We secula the where new triple. well- a systems, the the find of eccentric to we evolution and addition long-term orbits, the outer In affect and can orbits. inner two that the the of precession of GR axis post-Newtonian the semi-major first-order the the of expand ratio We Relativity. General of 1 n n“ue”bnr nwhich in binary “outer” an and , esuytesclr irrhcltrebd rbe ofirst-o to problem three-body hierarchical secular, the study We 1. A INTRODUCTION T E EOATPS-ETNA CETIIYEXCITATION ECCENTRICITY POST-NEWTONIAN RESONANT mdrNaoz Smadar tl mltajv 5/2/11 v. emulateapj style X − 0 a)is day) 100 NHEACIA HE-OYSYSTEMS THREE-BODY HIERARCHICAL IN ± %o 5 fthem of 151 of 5% 1,2, ∼ † 0.Moreover, 10%. ec Kocsis Bence , m . 5 3 − rf eso uut2,2018 21, August version Draft risthe orbits 1 . 5 M ABSTRACT dge, IL , m ⊙ ), 1 1 baa Loeb Abraham , etro aso h ne iay ihSMA with binary, inner the of mass of center h oa-io ehns ly nipratrl in role that showed important (2002) an al. multiples plays et mechanism or Blaes Kozai-Lidov 2012). binaries Loeb the al. & SMBH et Kulkarni Dotti 2007; re- of Loeb 2012; inevitably & formation Hoffman should lo- 1996; ex- the (Valtonen mergers the one in galaxy of (SMBHs), sult major abundance BHs that high supermassive pects a host galaxies that cal evidence strong the binary. be dwarf may BH-white clusters a in globular transfer in mass binaries al. triple-induced X-ray mech- et formation BH Ivanova important for Furthermore, most anism the that showed 2002). (2010) clusters 2003; Hamilton (Wen stellar BHs & binary dense Miller short-period of al. of centers formation et of the the growth Naoz and at the both (BHs) 2008; in holes role black been important al. has an et play mechanism Takeda to Kozai-Lidov suggested the 2007; Tremaine addition, com- In & Fabrycky al. 2012). stellar et 1997; Wu distant al. et additional 2007; Murray Holman plan- & an (e.g., extrasolar Prodan with panion and 1998; 2013; systems 2012) 2009; Thompson etary al. al. & et et Fabrycky Sharpee Kiseleva Naoz & 2012; Perets 2010; 1979; 2007; Thompson Harrington Tremaine Shaham (e.g., & & Fabrycky stars Mazeh triple important of an 1969; evolution as the proposed in been in element has dy- common process two-body This is from that absent namics. but period) systems orbital triple hierarchical the timescales on to interaction compared coherent long (i.e., effect secular a of so-called the system, nism incli- the and of eccentricity nation produce this the can In in perturber oscillations orbit. inclined large-amplitude binary highly inner a regime, the stability to not approaches does perturber close the make that is condition stability Another ie h irrhclglx omto aaim and paradigm, formation galaxy hierarchical the Given example important an is mechanism Kozai-Lidov The Kzi16;Ldv1962). Lidov 1962; (Kozai t xiain utemr,frtriples for Furthermore, excitation. ity aitna olaigodri the in leading-order to Hamiltonian dri otNwoinexpansion post-Newtonian a in rder 1 edt ria isee hnthe when even flips orbital to lead Nicol , aaee pc o ihyinclined highly for space parameter t cetiiis npriua,for particular, In eccentricities. ite vt ieclsaesvrlorders several are timescales ivity et upesecnrct,i only is eccentricity, suppress ffects c ag-mltd siltosin oscillations large-amplitude uce t fpaesae oee,post- however, space, phase of rts otNwoinitrcinterm interaction post-Newtonian r vmcaimi upesd post- suppressed, is mechanism ov nw em htcrepn to correspond that terms known eua etna timescales, Newtonian secular a oa-io quadrupole Kozai-Lidov lar sYunes as ´ 3 oa-io mecha- Kozai-Lidov a 2 ≫ a 1 . 2 Naoz et al. the evolution of SMBH triples, where high eccentricity in- established that the eccentricity and inclination are con- duced by the outer perturber can lead to a more efficient stant in the 1PN two-body problem (Damour & Deruelle merger rate, due to gravitational wave (GW) emission 1985), it is not true for hierarchical triples. In addition (see also Seto 2012). Also, recently Antonini & Perets to the standard GR precession of the inner and outer (2012) showed that secular three-body effects play an im- orbits, the 1PN corrections lead to a new secular in- portant role in the evolution of binary compact objects teraction between the inner and outer binaries that af- near SMBH. fects their long-term evolution. We find that the stan- GWs emitted during Kozai-Lidov–induced, highly dard lore, i.e., that GR effects suppress eccentricity, is eccentric orbits of compact binaries might be de- only true when the GR timescales are several orders tectable using LIGO5 and VIRGO6 (e.g., Wen 2003 of magnitude shorter than the secular Newtonian ones. but see Mandel et al. 2008 and O’Leary et al. 2006), When the GR timescales are comparable to the secular pulsar timing arrays (e.g., Finn & Lommen 2010; Newtonian ones, we show that three-body interactions Amaro-Seoane et al. 2010; Kocsis et al. 2012), and fu- generally give rise to a resonant-like eccentricity excita- ture space-based GW observatories, such as eLISA/NGO tion (see also Ford et al. 2000b). We will be using the (Amaro-Seoane et al. 2012a,b)7. In fact, GWs associated term “resonance” here to describe the rapid excitations with eccentric orbits are stronger and have a very differ- of the inner orbit’s eccentricity, which occurs when the ent spectrum relative to their circular counterparts for 1PN timescales are comparable to the secular Newto- sources at the same distance and with the same mass and nian timescales. We demonstrate that even for systems spin. This may allow for the GW detection of eccentric with comparable inner binary masses, where the Kozai- inspirals with higher masses, larger SMAs or farther away Lidov mechanism is suppressed, and even when the GR from Earth relative to their quasi-circular counterparts timescales are much longer than the secular Newtonian (Arun et al. 2007a,b; Yunes et al. 2009; O’Leary et al. ones, 1PN corrections continue to excite the eccentricity. 2009; Kocsis & Levin 2011). Using GW information This paper is organized as follows. We begin with a emitted by the close binary, it might be possible to con- definition of the parameters used to describe a hierar- strain the parameters of the third body, such as its mass chical triple system based on Newtonian and 1PN three- or distance, provided the GW signal-to-noise ratio is suf- body Hamiltonians ( 2). We then show that three-body ficiently high (Yunes et al. 2011; Galaviz & Br¨ugmann evolution is modified§ by 1PN effects ( 3). We discuss the 2011). different time-scales corresponding to§ the 1PN effects, The Kozai-Lidov mechanism is therefore tremendously and identify the region in phase space where important important and there is still much to be understood. Re- deviations might arise due to these terms ( 4). We then cently, Naoz et al. (2011, 2013) showed that an eccen- show that 1PN terms can, in many cases, excite§ the ec- tric outer orbit (and even a circular one with compa- centricity of the inner orbit instead of suppressing it ( 5). rable mass inner binary) can behave significantly dif- We conclude with a discussion in 6. § ferently than previously assumed, the so-called “ec- § centric Kozai-Lidov mechanism”. Specifically, they 2. HAMILTONIAN PERTURBATION THEORY FOR showed that the inner orbits can flip from prograde HIERARCHICAL TRIPLE SYSTEMS to retrograde and back, and can also reach extremely high eccentricities close to unity, and the system be- A triple system consists of a binary (with masses m1 haves chaotically (Lithwick & Naoz 2011). Most previ- and m2) and a third body (with mass m3) in orbit ous secular three body dynamics studies that incorpo- about the center of mass of the former. It is conve- rated GR effects did so through a pseudopotential, con- nient to describe the orbits using Jacobi coordinates structed mainly to model accretion disks and 1st post- (Murray & Dermott 2000). Let r be the relative position Newtonian (1PN) shifts in the innermost stable circu- vector from m1 to m2 and R3 be the position vector of lar orbit (Nowak & Wagoner 1991; Artemova et al. 1996; m3 relative to the center of mass of the inner binary (see Miller & Hamilton 2002). It has been shown that the for more details Naoz et al. 2013), as shown in Fig. 1. 1PN precession of the inner body may play an impor- In the PN approximation, corrections to Newtonian n tant role in secular evolution (e.g. Ford et al. 2000b; mechanics arise in powers of (v/c) , where v is the or- bital velocity and c is the speed of light, with n 2 an Miller & Hamilton 2002; Blaes et al. 2002; Mardling ≥ 2007; Fabrycky & Tremaine 2007; Zhang et al. 2013). integer. Here we concentrate on the 1PN order correc- tions to Newtonian motion, which are (v2/c2) relatively Here we expand our investigation to include both the ec- O centric Kozai-Lidov mechanism and the three body 1PN smaller than the Newtonian terms. The Hamiltonian can effects. We show here ( 3 and Appendix A) that al- then be divided into a Newtonian part ( N) and a 1PN § part ( ): H though this pseudo–potential does capture some 1PN ef- H1PN fects, such as the precession rate, the full 1PN three-body Hamiltonian introduces other corrections that cannot be tot,1PN = N + 1PN , (1) modeled with this potential. H H H In this paper, we study the consistent inclusion of 1PN where the Newtonian part is simply terms in the secular dynamical evolution of hierarchical triple systems. We restrict attention to the 1PN approx- 3 2 3 2 imation of the three-body Hamiltonian. While it is well 1 pi 1 k mimj N = , (2) H 2 mi − 2 rij 5 http://www.ligo.caltech.edu/ Xı=1 i,jX=6 i 6 http://www.ego-gw.it/ 7 http://elisa-ngo.org/ and the 1PN part is (e.g. Sch¨afer 1987; Moore 1993; Resonant 1PN Eccentricity Excitation in Hierarchical Triples 3
orbits. Let the SMAs of the inner and outer orbits be a1 and a2, respectively. Then, the coupling term in the Hamiltonian can be written as a power series in the ra- tio of the SMAs α = a1/a2 (e.g., Harrington 1968). In a hierarchical system, by definition, this parameter α is small. The Newtonian part of the Hamiltonian, expanded in powers of α, is (e.g., Harrington 1968), 2 2 k m1m2 k m3(m1 + m2) N = (4) H − 2a1 − 2a2 ∞ k2 r j a j+1 αj M 2 P (cos Φ) , −a j a R j 2 Xj=2 1 3
where Pj are Legendre polynomials, Φ is the angle be- tween r1 and r2 (see Fig. 1) and j−1 j−1 m1 ( m2) Mj = m1m2m3 − − j . (5) Fig. 1.— Coordinate system used to describe the hierarchical (m1 + m2) triple system (not to scale). Here ’c.m.’ denotes the center of mass of the inner binary, containing objects of masses m1 and m2. The Note that most secular studies follow the convention of separation vector r points from m1 to m2; R3 points from ’c.m.’ to Harrington (1969) and choose the Hamiltonian to be the m3. The angle between the vectors r and R3 is Φ. The distances r r r negative of the total energy, so that > 0 for bound from the bodies to a field point are labeled by 1, 2 and 3. systems. Here we did not follow thisH convention. The equations of motion in Naoz et al. (2013) did use this Lousto & Nakano 2008): convention, and thus, a reader that wishes to combine the two sets of equations need to introduce a minus sign 3 2 2 2 2 to one of the sets. 1 pi k mimj pi 1PN = 2 mi 2 2 6 2 We adopt canonical variables, known as Delaunay’s el- H −8c m − 4c rij m Xi=1 i i,iX=6 j i ements, which provide a particularly convenient dynam- (p p ) (n p )(n p ) ical description of hierarchical three-body systems (e.g. 7 i · j ij · i ij · j Valtonen & Karttunen 2006). The coordinates are cho- − mimj − mimj sen to be the mean anomalies, l1 and l2, the arguments k4 m m m ′ of periastron, g1 and g2, and the longitudes of ascend- + i j k . (3) 2 ′ ing nodes, h1 and h2, where subscripts 1, 2 denote the 2c ′ rij rik i,j=6 Xi,k =6 i inner and outer orbits, respectively. Their Newtonian 2 conjugate momenta are In these equations, k is the gravitational constant, rij m1m2 (rij ) is the relative position vector (magnitude) from L = k2(m + m )a , (6) 1 m + m 1 2 1 mass mi to mj , pi (pi) is the momentum vector (mag- 1 2 p nitude) of mass mi in an arbitrary plane (we shall later m (m + m ) L = 3 1 2 k2(m + m + m )a , transform to center-of-mass coordinates). In the 1PN 2 m + m + m 1 2 3 2 Hamiltonian, i, j and k′ run from 1 to 3 (the three 1 2 3 p ′ masses), where k is an index while k2 is the gravita- 2 2 G1 = L1 1 e1 , G2 = L2 1 e2 , (7) tional constant, and nij = rij /rij . q − q − Many gravitational triple systems are in a hierarchical and configuration: two objects orbit each other in a relatively H1 = G1 cos i1 , H2 = G2 cos i2 , (8) tight inner binary while the third object is on a much wider orbit. If the third object is sufficiently distant, an respectively, where e1 (e2) is the inner (outer) orbital ec- analytic, perturbative approach can be used to calculate centricity. Note that G1 and G2 are also the magnitudes the evolution of the system over long timescales (rela- of the angular momentum vectors (G1 and G2), and H1 tive to the orbital period). In the usual secular approx- and H2 are the z-components of these vectors. The fol- imation (e.g., Marchal 1990), the three orbiting objects lowing geometric relations between the momenta follow torque each other and exchange angular momentum, but from the law of cosines: 2 2 2 not energy. Therefore, on timescales much longer than Gtot G1 G2 their orbital periods, the eccentricity and orientation can cos itot = − − , (9) 2G1G2 change, but not the SMA. 2 2 2 Given this, the orbital motion of a triple system can Gtot + G1 G2 H1 = − , (10) be divided into two separate Keplerian orbits: the rel- 2Gtot ative orbit of bodies 1 and 2, and the orbit of body 3 2 2 2 Gtot + G2 G1 around the center of mass of the system. The Hamilto- H2 = − , (11) 2G nian for the system can then be decomposed accordingly tot into two Keplerian Hamiltonians plus a coupling term where Gtot = G1 + G2 is the (conserved) total angu- that describes the (weak) interaction between the two lar momentum vector, and the angle between G1 and 4 Naoz et al.
G2 defines the mutual inclination itot = i1 + i2. From the SMAs, and are constant in the secular approxima- Eqs. (10) and (11) we find that the inclinations i1 and i2 tion without dissipative effects. Furthermore, we have are determined by the orbital angular momenta via defined 2 2 2 Gtot + G1 G2 m1 m2 a1 e2 cos i1 = − , (12) ǫM = − 2 , (21) 2GtotG1 m1 + m2 a2 1 e2 2 2 2 4 7 − 7 4 Gtot + G2 G1 k (m1 + m2) m3 L1 cos i2 = − . (13) C2 = 3 3 3 3 , (22) 2GtotG2 16 (m1 + m2 + m3) (m1m2) L2G2 5 In addition to these geometrical relations we also have A =4+3e2 B sin i2 , (23) that 1 − 2 tot H + H = G =const (14) B =2+5e2 7e2 cos(2g ) , (24) 1 2 tot 1 − 1 1 since we are here neglecting dissipative effects such as and GW radiation-reaction, and thus, the Hamiltonian is conserved. Given this parameterization, the Hamilto- cos φ = cos g1 cos g2 cos itot sin g1 sin g2 . (25) nian or canonical equations describe the orbital motion − − via Note that the octupole coefficient in Ford et al. (2000b), is simply C3 = C2(ǫM /e2)15/4. Also, following our def- dL ∂ dl ∂ j = H , j = H , (15) initions (see Figure 1) m1 and m2 refer to the com- dt − ∂lj dt ∂Lj ponent masses of the inner orbit, while e2 refers to the eccentricity of the outer orbit. In the test-particle dGj ∂ dgj ∂ = H , = H , (16) limit (i.e., m1 m2) ǫM [Eq. (21)] reduces to the oc- dt −∂gj dt ∂Gj tupole coefficient≫ introduced in Lithwick & Naoz (2011) dH ∂ dh ∂ and Katz et al. (2011), j = H , j = H , (17) dt −∂h dt ∂H j j a e ǫ = 1 2 . (26) where j =1, 2. 2 a2 1 e2 The secular Hamiltonian (both the Newtonian and the − 1PN parts) is given by taking Eq. (1) expanded in pow- In these Hamiltonians (and in the following 1PN parts), ers of α and averaging over the rapidly varying l1 and we have eliminated the nodes (i.e., h1 and h2) by using l2. The averaging technique we use is known as the the conservation of total angular momentum, which leads Von Zeipel transformation (for more details, see Brouwer to h1 h2 = π. As shown in Naoz et al. (2013) this can 1959; Naoz et al. 2013), also see Appendix B, a canoni- be done− only as long as one does not conclude that the cal transformation that eliminates the rapidly-oscillating conjugate momenta are constant (e.g., Dirac 1950). The parts of . We apply this transformation twice, lead- full equations of motion up to the Newtonian octupole ing to aH Hamiltonian that is the double average of the order are presented in Naoz et al. (2013). original Hamiltonian over both orbital periods. We thus The averaged 1PN Hamiltonian can be separated into refer to the resulting quantity as the “double-averaged different terms. First, let us use the fact that for Ke- Hamiltonian.” plerian orbits the momentum can be related to the ra- The double-averaged Newtonian Hamiltonian [Eq. (4)], dius and SMA; for the inner orbit, we can write pin = 3 2 up to octupole order (i.e., up to (α ) beyond the leading µin k (m1 + m2)(2/r 1/a1), where µin is the reduced −O1 8 − order term proportional to a1 , see Naoz et al. 2013) massp of the inner orbit and a similar relation can be writ- can be written as: ten to the outer orbit. Second, we substitute this relation into the three body 1PN Hamiltonian, i.e., Eq. (3). After ¯ = ¯N + ¯N (18) HN Hquad Hoct transforming to the center of mass frame, the 1PN cor- rections is expanded in powers of α up to relative (α3). where This produces a similar expansion to Eq. (4) for theO 1PN N 2 2 ¯ = C2 2+3e 3cos itot 1 Hamiltonian, but due to its length we have chosen not Hquad − { 1 − to present it here. To investigate the long-term dynam- 2 2 +15e1 sin itot cos(2g1) , (19) ics of the three-body system, we eliminate all terms with } N 15 2 short-periods in the Hamiltonian, which depend on the ¯ = ǫM e1C2 A cos φ + 10cos itot sin itot Hoct 4 { rapidly changing l1 and l2, using a double Von Zeipel transformation (Brouwer 1959), see for more details Ap- (1 e2) sin g sin g . (20) × − 1 1 2} pendix B. In doing so, we must first calculate the angle Note that Equation (19) has a minus sign compare to between the vectors pin pout and pout r and pin R3, · · · Naoz et al. (2013) that used the sign convention for where pin (pout) is the momentum of the inner (outer) which the Hamiltonian is positive. Here, we did not in- orbit, as defined in the invariable plane. clude the terms which correspond to the Keplerian or- The leading-order term in an α 1 expansion is pro- −2 ≪ bital energy of the three objects which depend on only portional to a1 in the double-averaged 1PN Hamilto- nian. Keeping all terms up to (α3) beyond leading 8 Note that Krymolowski & Mazeh (1999) showed that the von gives O Zeipel transformation results in higher orders terms proportional to 7/2 3 1PN 1PN 1PN 1PN 1PN α , however here we consider only O(α ) level of perturbations. ¯ ¯ −2 ¯ ¯ −2 ¯ = + a1a2 + + int (27) H Ha1 H Ha2 H Resonant 1PN Eccentricity Excitation in Hierarchical Triples 5 where IN POST-NEWTONIAN THEORY k4µ 15m 2 + 29m m + 15m 2 The secular evolution of a three-body hierarchical ¯1PN in 1 1 2 2 −2 = (28) system to Newtonian, octupole order was studied in Ha1 8a 2c2 1 Naoz et al. (2013). As mentioned in 1, they showed that 3k4m m (m + m ) § 1 2 1 2 , the commonly assumed conservation of the z-component 2 2 2 − a1 c √1 e1 of the angular momenta of the inner and outer orbits 4 − (H1 and H2) is only correct in the test-particle approxi- ¯1PN k m1m2m3(2(m1 + m2)+3m3) a1a2 = 2 , (29) mation to quadrupole order. Newtonian octupole terms H 4a1a2c (m1 + m2 + m3) further modulate the eccentricity and inclination oscil- k4µ 15(m + m )2 + 29(m + m )m + 15m 2 lations. Specifically, for an eccentric and inclined outer ¯1PN out 1 2 1 2 3 3 −2 = perturber, these terms can lead to extremely high eccen- Ha2 8a 2c2(m + m + m ) 2 1 2 3 tricities and flip the inner orbit from prograde to ret- 4 3k (m1 + m2)m3(m1 + m2 + m3) rograde. This type of behavior also appears in the test- (30) 2 2 2 − a2 c √1 e2 particle limit for an eccentric orbit (e.g. Lithwick & Naoz 2 − 2011; Katz et al. 2011; Naoz et al. 2013). ¯1PN k int = G1G2[8(m1 + m2) Figure 2 presents the secular evolution of a three-body H 3 2 2 3/2 4a2c (1 e2 ) (m1 + m2) hierarchical system to Newtonian, octupole order (red − 2 lines). We chose a system with inner binary masses a1k m1m2m3 2 +6m3]cos itot + (fme1 3fme1 cos itot m1 = 1 M⊙ and m2 = 0.001 M⊙, and an outer bi- 8(m + m ) − 4 1 2 nary companion with mass m3 = 10 M⊙. For this g 4 g 2 2 2 2 system, we set a1/R1 = 10 and a2/R3 = 202, where, +9e1 m1 + m1m2 + m2)cos(2g1) sin itot , (31) g 2 2 g 2 2 R1 = k (m1 +m2)/c and R3 = k m3/c are the gravita- tional radii of the inner and the outer orbits, respectively. ◦ and where We also set initially e2 = 0.6, e1 = 0.01, g1 = g2 = 0 ◦ m1m2 and itot = 85 . The Newtonian quadrupole terms in- µin = , (32) m1 + m2 duce the “standard” eccentricity–inclination oscillations, while octupole terms modulate it. As can be seen in m3(m1 + m2) µout = , (33) the figure, the modulation does not have a precise pe- m1 + m2 + m3 riodicity and, in fact, the octupole terms introduce the 2 2 2 2 fme1 = (2 5e1)(m1 + m2) 3(2 e1)m1m2 (34) chaotic aspects to the evolution (Lithwick & Naoz 2011). − − − When 1PN corrections become significant however, the Here ¯1PN includes all terms of (α5/2) and (α3) be- Hint O O evolutionary orbital tracks can be significantly different ¯1PN 1/2 1/2 yond −2 , since G1 a1 and G2 a2 according to (already at quadrupole order). Ha1 ∝ ∝ Different 1PN terms have different effects on the evo- Eqs. (6) and (7). lutionary orbital tracks, where the perturbations to the Not all of the different 1PN Hamiltonian terms affect 1PN equations of motion follow from Eqs. (15–17). ¯ −2 ¯1PN a the dynamical evolution of the triple. The a1a2 term H 1 only depends on the masses and the SMAs,H i.e., it does gives rise to the standard GR precession of the argument 1PN not depend on the canonical coordinates, and thus, it of periapsis of the inner orbit, while ¯ −2 is responsi- Ha2 does not affect the canonical equations, although it does ble for the precession of the argument of periapsis of the change the total energy of the system. On the other 1PN 1PN outer orbit, hand, ¯ −2 and ¯ −2 do contribute to the dynamical a a H 1 H 2 dg 3k3(m + m )3/2 evolution, as they clearly depend on e1 and e2. A pos- 1 = 1 2 , (35) sible, (intuitive) physical explanation for this is the fol- dt −2 5/2 2 2 1PN(a1 ) a1 c (1 e1) lowing. In the a2 limit, one would expect only two − → ∞ dg 3k3(m + m + m )3/2 physical effects: precession of the inner orbit and preces- 2 = 1 2 3 . (36) −2 5/2 2 sion of the outer orbit about the inner binary. These two dt 1PN(a ) a c2(1 e ) 1PN 1PN 2 2 2 physical effects arise because of ¯ −2 and ¯ −2 , and − Ha1 Ha2 ¯1PN ¯1PN ¯1PN These contributions can be recovered independently from thus, a1a2 , (which satisfies a1a2 −2 for large H H ≫ Ha2 the individual two-body 1PN Hamiltonians of the inner a2/a1) cannot contribute to the motion. and outer binary (see Appendix A Eq. A1), or from an ¯1PN effective potential, or directly from the 1PN metric (e.g., The quantity int is an “interaction term,” in that it represents theH coupling between the outer and the in- Misner et al. 1973, chapter 25 p. 668–670). Other than 1PN 1PN ner orbits. Notice that this term would not be present this precession, ¯ −2 and ¯ −2 , do not directly affect 2 Ha1 Ha2 if we had truncated the α 1 expansion at (α ). No- 9 ¯1PN ≪ O the other orbital elements. The a1a2 term just modi- tice also that the interaction term does not depend on fies the total energy and does notH modify the long–term the argument of periapsis of the outer orbit, g2, just dynamical evolution at all, as long as dissipative effects like the quadrupole Newtonian Hamiltonian. Therefore, are neglected. at quadrupole order, e.g. for a circular outer perturber, In the standard lore, if the GR precession rate of the in- the absence of g2 in the Hamiltonian implies that the ner orbit is faster than the quadrupole secular Newtonian outer orbital angular momentum, G2, is conserved (the so-called “happy coincidence” of Lidov & Ziglin 1976). 9 Note, however, that the precession indirectly affects the evolu- ¯N 3. TRIPLE BODY EVOLUTION tion of the other orbital elements through H , as shown below. 6 Naoz et al.
Fig. 2.— An example of the evolution of a triple to Newto- Fig. 3.— Two examples of the time evolution of the system dom- nian octupole order neglecting PN contributions (red lines) and inated by 1PN effects, the right and left side panels differ only including them up to the 1PN interactions terms in the double- in the initial relative inclination and the outer orbital separation. averaged Hamiltonian (blue lines). The system has an inner binary In the top panels we consider the three-body orbit evolution due of m1 =1M⊙, m2 = 0.001 M⊙, and a the third object mass is: to Newtonian dynamics and the lowest–level 1PN correction, i.e., 4 g 4 g 1PN m3 = 10 M⊙. We set a1/R1 = 10 and a2/R3 = 202. We set HN + H −2 . In the middle panels we add the next level of ap- initially e = 0.6, e = 0.01 g = g = 0◦ and i = 85◦. We a1 2 1 1 2 tot 1PN proximation +H −2 , and in the bottom panels we consider the consider, from top to bottom, the inclination of the inner orbit, a i , the eccentricity of the inner orbit in terms of 1 − e , and the 2 1 1 H1PN z-component of the angular momentum normalized to the total approximation up to the highest level discussed here, i.e., + int . The inner binary contains an object of mass 1 M⊙ and an object angular momentum. Note that we do not plot G2 since in con- → → of mass 1 MJ (can be considered as a test particle), while the outer figuration where m2 0, G2 Gtot, (Lithwick & Naoz 2011). 6 In this case, the quadrupole Newtonian terms induce eccentricity– object is a BH with mass of 10 M⊙. The inner orbital separa- g 5 inclination oscillations, modulated by the octupole terms, while the tion is a1/R1 = 5.06 × 10 , corresponding to 0.005 AU. The initial 1PN effect suppresses them. eccentricities are e1 = 0.001 and e2 = 0.7. The arguments of pericenter of the inner orbit initially set to 240◦ and outer orbit initially is set to zero. In the left column, we consider an initial ◦ relative inclination of itot = 95 , and an outer orbital separation timescales, the GR effect is presumed to suppress the ec- g 3 of a2/R3 = 5.2 × 10 , corresponding to 51.4 AU. In the right col- ◦ centricity growth (for an m2 test particle to quadrupole umn, we consider an initial relative inclination of itot = 65 , and g 3 Newtonian order, see Fabrycky & Tremaine 2007). In an outer orbital separation of a2/R3 = 4.8 × 10 , corresponding to Fig. 2 (blue lines) we show an example where this is in- 47.35 AU. Observe that as one includes more 1PN effects, qualita- deed the case, even when including all 1PN terms (see tively different behavior emerges. below). In this example, eccentricity (and orbital flips) are suppressed by the 1PN corrections (the variations shown by the blue lines are shorter than red lines). the calculation in the two columns, the initial eccentrici- The usual precession term [Eq. (35)] is not sufficient to ties are e1 =0.001 and e2 =0.7, and the initial argument model the system, as one must also account for the pre- of pericenter of the inner and outer orbits is set to 240◦ cession of the outer orbit [Eq. (36)] and the other effects and zero, respectively. This system configuration is such −2 introduced by the 1PN interaction terms (Appendix C). that the 1PN(a1 ) timescales for circular orbits ( 59 yr, The inclusion of these terms leads to qualitatively differ- for the left column example, see 4 for more details)∼ are ent behavior because they directly drive the evolution of shorter than the Newtonian quadrupole§ ones ( 145 yr inner and outer orbital eccentricity and inclination, while for the left column example). Nevertheless, the∼ secu- Eq. (35) [Eq. (36)] drives the evolution of only the argu- lar eccentricity oscillations are still present. The bottom ¯1PN ment of periapsis of the inner (outer) orbit. Figure 3 panel shows that the interaction term int introduces shows the evolution of the eccentricity when different a significant new periodic modulation toH the eccentricity terms in the Hamiltonian are included. We considered a evolution. We discuss in more details the conditions in system with parameters m1 =1 M⊙, m2 =1 MJ (essen- parameter space that lead to this behavior in 4 and 5 6 g 5 § tially a test particle), m3 = 10 M⊙, a1/R1 =5.06 10 (note that this system represents the resonance peak of corresponding to 0.005 AU. In the left column we× con- the 95◦ and 65◦ cases of Fig. 5 below). ◦ sider initial relative inclination itot = 95 and a separa- The usual precession term (mostly for the inner) in g 3 tion of the outer orbit a2/R3 =5.2 10 corresponding the presence of three body secular evolution was com- to 51.4 AU. In the right column, we× consider initial rel- pared to direct N-body calculation in the literature be- ◦ ative inclination itot = 65 and a separation of the outer fore (e.g., Ford et al. 2000b; Zhang et al. 2013). To re- orbit a /Rg =4.8 103 corresponding to 47.35 AU. For solve the effects of the interaction term one needs to be in 2 3 × Resonant 1PN Eccentricity Excitation in Hierarchical Triples 7 the the regime where the GR precession time scales are (Lithwick & Naoz 2011). We will see that this is suf- much shorter than the quadrupole precession time scales ficient to understand the regions of phase space where (see 4). The examples considered in Figure 3 represent 1PN terms become important. 3 § 108 of the inner orbital period. Numerical integra- The 1PN timescale can be estimated as in Eq. (37), tions∼ × algorithms that conserve the energy over such long which gives timescales in the 3-body post Newtonian regime are not 5/2 2 2 easy to implement or develop, and thus, they are left to 1PN a1 c (1 e1) −2 ta 2π 3 − 3/2 , (39) future work. 1 ∼ 3k (m1 + m2) One might worry that a 1PN treatment might not be 5/2 2 2 1PN a2 c (1 e2) sufficient to model certain regions of phase space, as we −2 ta 2π 3 − 3/2 , (40) have neglected 2PN and higher PN order terms in the 2 ∼ 3k (m1 + m2 + m3) evolution. Such terms become important when the PN 3 2 2 3/2 3/2 1PN 16 a2c (1 e2) (m1 + m2) perturbation parameter, v/c with v any system velocity, tint − (41). 2 g g 2 3 2 2 i.e. the pericenter velocities (v /c) = (R +R )/[a (1 ∼ 9 √a1e1 1 e1k (m1 + m1m2 + m2)m3 p1 1 2 1 − − e1)] and similarly for the outer orbit, are not sufficiently p small. In most of our examples we ensured that our All of these timescales depend on the secularly varying, systems are well within the PN regime, however, for very inner orbital eccentricity, which implies that we need eccentric systems, 2PN and higher PN order terms may to explore different eccentricity values in phase space. be important. In fact, one might naively expect the 1PN Equations (38–41) show that the Newtonian and 1PN corrections accounted for here to be negligible if v/c 1. timescales have a simple dependence on the inner and This is not so, because although the 1PN terms are small≪ outer orbital eccentricity, on the mass ratio qm = m3/m1, at any point in time, their effects may accumulate and and on the SMAs. become significant over long timescales in the three-body If any of the above timescales is much smaller than problem. all others, then the evolution of the triple is dominated by the corresponding term in the Hamiltonian. Next, 4. TIMESCALES we examine the three-body parameters where any two In order to explore the regions of phase space where timescales are equal, which defines the region where the the 1PN effects may be expected to become significant, corresponding two terms are equally important. The cor- we compare the various timescales associated with the responding regions are shown in Figure 4 in the m2 0 limit. → individual terms in the Hamiltonian. 1PN N Equating t −2 to t (Eqs. 37 and 39) gives a relation The timescale associated with the Newtonian a1 quad quadrupole term can be estimated from the canonical between the SMAs which normalized to the gravitational equations. More precisely, tN 2πG /C , where C radius of the inner and outer binaries, Rg and Rg, as quad ∼ 1 2 2 1 3 is given in Eq. (22): defined above, can be written as:
3 2 3/2 1/3 4/3 2 1/3 N 2πa2(1 e2) √m1 + m2 a2 1 a1 1 (1 e1) tquad − 3/2 . (37) g g − . ∼ a m k R −2 ∼ 3 R 2/3 1 e2 1 3 3 quad=1PN(a1 ) 1 qm 2 − The timescales associated with the Newtonian oc- p (42) tupole terms are more difficult to estimate due to their This relation is shown by the blue hatched area bounded by solid blue lines in Figure 4 for 0 e 0.999. A chaotic effect on the orbits. For example, in Fig. 2 ≤ 1 ≤ the first modulation period is shorter than the second resonant-like 1PN excitation of eccentricity is possible in this region, as we will show in the next section. For (see, Naoz et al. 2013, for more examples). However, as 1PN N much larger a2 or smaller a1, t −2 tquad, and thus demonstrated in Lithwick & Naoz (2011), although the a1 ≪ system is chaotic when the octupole terms are included, the Kozai-Lidov eccentricity excitations are suppressed there are two general features in the evolution: one as- by the 1PN effects. 1PN N sociated with an octupole timescale, defined below, and Next, equating t −2 to t gives a quad a shorter one (see for example their Fig. 7 of surfaces of 2 section, which shows the two different evolutionary be- 3 a2 a1 qm 1 haviors). We define a timescale for the regular part of g 3 g 3 2 . (43) R −2 ∼ R (1 + qm) 1 e the Newtonian octupole evolution through the rough es- 3 quad=1PN(a2 ) 1 − 2 timate tN (4/15)tN /ǫ for a given inner and outer oct quad M This is shown by a blue short–long dashed line in Fig. 4 eccentricity,∼ on the top and bottom panels for e2 = 0.5 and 0.7, re- g 4 a4(1 e2)5/2 1 e2(m + m )3/2 spectively. For a2/R3 much larger than this value, the tN 2π 2 − 2 − 1 1 2 . (38) Kozai-Lidov oscillations are suppressed and the 1PN ef- oct ∼ 15 5/2 p a1 e2k m1 m2 m3 fects dominate. | − | 1PN N Note that when the inner binary is very eccentric, these Let us next compare tint and toct by setting them timescales can change by orders of magnitude. At oc- equal to each other: tupole order, the eccentricity of the outer orbit can also a 32 a 2 1 e e oscillate, although usually these oscillations are small in 2 1 1 2 . Rg ∼ 135π Rg q (1 e2)(1 e2) magnitude. For the remainder of this section, we employ 3 oct=1PN(int) 1 m 2 1 − − a test-particle approximation for one of the components (44) of the inner binary, i.e. m 0, for which e = const. shown by long-dashed brown lines in Fig. 4. The 1PN 2 → 2 8 Naoz et al.
Fig. 4.— Regions of binary parameters where the various secular Newtonian and 1PN effects are expected to become significant. We g show the SMAs where the timescales, corresponding to individual terms in the Hamiltonian, are equal to each other for different a1/R1 g 4 6 and a2/R3 . We consider four mass ratios between the outer object and the inner binary qm = m3/m1 = 0.01, 1, 10 and 10 (see labels in each panel) and two different choices for outer orbital eccentricity (e2 = 0.5, top panels and e2 = 0.7 lower panels), setting m2 → 0 in N 1PN N 1PN N 1PN all panels. We consider tquad = t −2 (Eq. (42), solid blue lines), tquad = t −2 (Eq. (43), thick long-short dashed blue lines), toct = tint a1 a2 N 1PN (Eq. (44), long dashed brown lines) and toct = t −2 (Eq. (45), short dashed purple lines). For the latter, we show e1 = 0 (top purple a2 line) in all panels and in the top left panel we also show the eccentric case e1 = 0.999 (bottom purple line). The gray and blue shaded areas cover the range 0 ≤ e1 ≤ 0.999 between the brown shaded lines and the blue solid lines respectively In the top left panel and bottom right, we also show a green band for which the timescale to shrink the inner orbit by a factor two, due to GW emission, is equal to the quadrupole timescale, which covers the range 0 ≤ e1 ≤ 0.999. This line is generated by specifying m1 = 100 M⊙ and m2 = m3 =1M⊙ (top left panel) and m2 = m1 =1M⊙ (bottom right panel). We also show two stability criteria, ǫ = 0.1, dotted magenta lines and the Mardling & Aarseth (2001) criterion [Eq. (46), in the top right panel]. The green dot in the right top panel represents (up to a factor 3) the location in this phase space of the system considered in Fig. 7, while the black dot in the bottom right panel represent the location of the example considered in Fig. 3. The rectangle in the top left panel shows the parameter space considered in Figure 6. The black rectangles in the top and bottom right panels roughly represent the region in parameter space where planets in stellar binaries (top right) and stars in the galactic nuclei (bottom right) live. effects are equally important as the Newtonian ones in purple lines in Fig. 4): the gray shaded area in Fig. 4 for 0 e1 0.999. This is ≤ ≤ 1/3 5/3 the regime in which the 1PN interaction term introduces a2 4 a1 qualitatively different behavior in the orbital motion (i.e., g g (45) R −2 ∼ 45 R modulation) as shown in Fig. 3. Outside the gray region 3 oct=1PN(a2 ) 1 in Fig. 4, the interaction term is negligible. Note that 1 e2/3 comparing t1PN with tN leads to a vertical line in the 2 . int quad × 1/3 5/3 (1 e2)(1 e2)1/3 phase diagram of Fig. 4. This is because both timescales qm (1 + qm) − 2 − 1 have the same dependence on the outer SMA ( a3), ∼ 2 This relation also provides a range of parameters for dif- resulting in a1/R1 < 10 (not shown). In 5, we explore § ferent values of e1 and e2, but to avoid cluttering, we plot the parameter space that also leads to excitations in the this timescale only for a circular inner binary (e = 0). eccentricity (as shown in Fig. 3). 1 1PN The boundary shifts to larger a2 in the eccentric case. In Comparing t −2 to the octupole timescale gives usually a2 top-left panel of Fig. 4 (the qm =0.01 case), we show the a longer timescale than the quadrupole (short–dashed range of Eq. (45) for e2 (0, 0.999), since, in this case, this ratio is smaller than∈ Eq. (44). Resonant 1PN Eccentricity Excitation in Hierarchical Triples 9
Figure 4 also shows the regime of validity of the hierar- (2000b), studying the triple system PSR B1620 26, chical triple approximation, where we choose ǫ > 0.1 as showed that the Newtonian octupole and the leading− or- a rule of thumb for stability (dotted magenta line). For der 1PN corrections [i.e., Eq. (35)] can lead to the exci- the qm = 1 case, this rule of thumb seems to agree with tation of the eccentricity of the inner orbit. We repeat the Mardling & Aarseth (2001) stability criterion, which and extend the investigations of that study for a broader defines a stable three-body system as one that obeys range of three-body systems and examine (i) whether the 2/5 new 1PN terms derived in 2 give rise to different behav- a2 2/5 (1 + e2) 0.3itot ior; and (ii) whether Newtonian§ octupole terms can sig- > 2.8(1 + qm) 6/5 1 ◦ , (46) a1 (1 e2) − 180 nificantly change the evolution of the three-body system − in the presence of the 1PN terms by producing flips of the where in the top right panel we considered this crite- ◦ inner orbit and eccentricities close to unity (Naoz et al. rion with prograde itot = 0 . Note that retrograde mo- 2011, 2013). We begin by exploring systems in which tions are usually more stable (e.g. Innanen 1979, 1980; m3 m1 (qm 1), and investigate the opposite limit Morais & Giuppone 2012). in the≫ next subsection.≫ Another consistency requirement for the 1PN Kozai- Let us systematically examine the parameters where Lidov effects to operate is that gravitational radiation the competition between the secular Newtonian Hamil- does not reaction change the SMA significantly over this tonian and 1PN corrections lead to the excitation of the timescale. We define tGW1, the GW in-spiral timescale inner orbital eccentricity as opposed to a suppression. of the inner binary for the SMA to change by factor of We do this by preforming a large number of simulations two using Peters (1964) (see also Arun et al. 2009). Note starting from a nearly zero eccentricity for the inner bi- that tGW1 when m2 0, but the GW inspiral may nary and varying the following dimensionless parameter: become very→ important ∞ in the→ comparable-mass limit and 1PN shut off the Kozai-Lidov effect. This is the case above t −2 g 4 a 1 (a /R ) 1 the green bands in the top left and bottom right panels, 1 1 1 = N = g 3 2 2 3/2 . (47) which show the range of SMAs where tN = t for R tquad 3 (a2/R3) qm(1 e2) quad GW1 e1=0 − m = m = 1 M (top left panel) and m = m = 1 M 2 3 ⊙ 2 1 ⊙ where we have substituted Eqs. (37) and (39) with e = (bottom right panel) for 0 e 0.999. Note that there 1 ≤ 1 ≤ 0. The quantity in Eq. (47) compares the timescales is a region below the green band where the 1PN approx- of the leading order 1PN and secular Newtonian effects. imation is insufficient, and where 2PN and higher PN We find that a resonant–like eccentricity excitation can order corrections need to be included; we leave this to take place around = 1 for zero initial eccentricity as future work. A similar regime, were the 1PN level of ap- shown below. Note,R however, that another important in- proximation is insufficient, was considered by Blaes et al. gredient to this resonant-like behavior is the Newtonian (2002) for which the evolution was mostly affected by octupole term, which is most obvious in simulations with GW emission and resonant eccentricity growth was not a low initial mutual inclination, for which the quadrupole observed. approximation is subdominant (see also 5.2 below and We conclude that 1PN effects may be important for Ford et al. 2000b). In other words, neglecting§ the contri- a vast range of parameters as shown in Fig. 4. Note bution of the octupole–level of approximation, one would however, that physical timescales are not shown there; miss entirely the resonant behavior. this figure is independent of an overall dimensional scale Figure 5 shows the maximum eccentricity achieved (e.g. total mass or the scale of the SMAs). The physical during the course of the evolution of the system studied, timescale may be smaller or larger than the Hubble time after 1000 quadrupole cycles10 as a function of . We depending on the actual system parameters considered. also examined the effects of various 1PN terms byR repeat- ¯ ¯1PN 5. EXCITATION OF THE INNER ORBITAL ECCENTRICITY ing the calculations using N + −2 (blue triangles), H Ha1 1PN As stated above the 1PN effects may suppress the then adding ¯ −2 (green crosses), and finally including a Newtonian Kozai-Lidov oscillations if the correspond- H 2 ¯1PN (red circles). The fiducial example chosen (black ing 1PN timescale is much smaller than the Newtonian Hint lines in all panels) has m1 = 1 M⊙, m2 = 0.001 M⊙, quadrupole one. However, we identify two regimes where 6 6 and m3 = 10 M⊙ (i.e. qm = 10 ) with initial condi- the combined secular Newtonian and 1PN effects pro- ◦ ◦ tions e1 = 0.001, e2 = 0.7, itot = 65 , g1 = 240 and duce interesting different behavior in three-body sys- ◦ tems: (i) if the quadrupole order terms are comparable g2 = 0 . To generate Fig. 5, we used different runs and to or somewhat smaller than the 1PN ones, or (ii) for changed both a1 and a2. For a choice of mass ratio and initial mutual inclination the parameter collapses the comparable-mass inner binaries where the 1PN effects R are subdominant relative to the Newtonian quadrupole different runs with different SMAs into one curve. Note terms but they are comparable to the Newtonian oc- that changing the initial value of the argument of peri- tupole timescale. We discuss these two regimes in detail apsis does not change the location of the resonance, but below. it does change its width, as shown in the inset of Fig. 5, where we compare the fiducial example with an initial 5.1. Eccentricity Peak in the Restricted Three Body 10 Each point in this figure corresponds to a separate, high- Problem for a Massive Perturber resolution three-body evolution with 1PN effects, each of which takes approximately 3 days to complete per computer core. We also The standard lore says that GR effects (or 1PN ef- preformed convergence tests using longer integration times at high fects in our case) suppress the eccentricity growth of resolutions, and found that for this type of systems at least 1000 the inner orbit in a three body system (e.g., Blaes et al. quadrupole cycles are needed for convergence, over the parameter 2002; Fabrycky & Tremaine 2007). However, Ford et al. range considered in the Figure. 10 Naoz et al.
6 Fig. 5.— The maximum eccentricity in a triple system in the test-particle approximation for different mass ratios (qm = 10 , left hand 4 panels, and qm = 10 , right hand panels) as a function of the relative timescales of the 1PN and secular Newtonian quadrupole effects −2 −2 (R, see Eq. 47). We consider the 1PN evolution including terms only up to O(a1 ), O(a2 ) and the interaction term (blue triangles, green crosses, and red squares, respectively). We show three different initial outer orbital eccentricities: e2 = 0.7 (top panels) e2 = 0.6 and 0.5, bottom right and left hand respectively. In all these examples, we set m1 =1M⊙, m2 = 0.001 M⊙, m3 = qmm1 and we vary ◦ both the inner and outer SMAs to match the different values of R, see Eq. (47). The system is initialized with e1 = 0.001, g2 = 0 and ◦ ◦ ◦ ◦ g1 = 240 (inset shows the results for initializing g1 = 0 and 240 ). In all panels we set the initial mutual inclination to 65 . In the top left panel, we also set the initial inclination to 95◦ (left set of lines) and 50◦ (right set of lines) . We compare the fiducial example 6 ◦ 4 (qm = 10 ,itot = 65 ,e2 = 0.7), black lines in each panel, to systems with a different mass ratio (qm = 10 ), right hand panels), and different eccentricities, see labels top right panel. Observe the emergence of resonant-like eccentricity excitations.
◦ g1 = 240 to an initial g1 = 0. The different panels many inspiraling systems some time during their evolu- show that the amplitude and width of the resonance are tion. Whether this occurs or not depends on the masses different for different qm and e2, as they depend on the and SMAs; a detailed analysis is left to future work. Newtonian octupole terms. The top left panel shows that 5.2. The Case of PSR B1620 26 different initial inclinations change both the location of − the resonance and its width (as a function of ), i.e., Although the above discussion of the resonant behav- low initial inclination leads to broad resonances.R The ior assumed a test particle (m2 0), general mass-ratio time evolution of the systems with parameters on the triples also exhibit a similar effect∼ (as was first consid- ◦ ◦ 1PN peak of the resonance of the itot = 95 and 65 cases are ered by Ford et al. 2000b), provided t −2 is shorter then shown in Fig. 3. a1 tN [see Eq. (39)]. To find in the general case, we To the left of the resonance peak, the maximum ec- quad R centricity converges asymptotically to the initial inner can simply set eccentricity for 1 (i.e. in this case the eccentric- qm m3/(m1 + m2) (48) R ≪ −3 ity is not excited). This was chosen to be e1 = 10 in ≡ Fig. 5. Decreasing the initial eccentricity (not shown) in Eq. (47). Ford et al. (2000b) observed a resonant- changes the figure at 1, but it does not change the like eccentricity increase while studying the triple sys- R ≪ tem PSR B1620 26, which is located near the core of amplitude and location of the resonance and the behav- − ior at larger . To the right of the resonance ( & 1), the globular cluster M4. They showed that this resonant the maximumR eccentricity starts to increase whenR the ec- behavior may explain the unusually large eccentricity of centric Kozai-Lidov mechanism begins to dominate. For the inner binary, which contains a millisecond radio pul- q 1 and 1, we find asymptotically e 1 (see sar of m1 = 1.4M⊙ and a companion of m2 = 0.3M⊙ m ≫ R ≫ 1 → (McKenna & Lyne 1988). 5.2 for small qm). § A binary that slowly shrinks due to GW emission, For completeness, we repeat and extend the calcula- slowly changes and may sweep across the resonant re- tion of Ford et al. (2000b) by fixing the inner orbit’s gions of eccentricityR excitation shown in Fig. 5. Thus, SMA, changing the outer orbit’s SMA, and including all resonant 1PN eccentricity increase may take place in 1PN terms. We choose two different values for the in- ner binary to explore the sensitivity to these parameters. Resonant 1PN Eccentricity Excitation in Hierarchical Triples 11
First, following Ford et al. (2000a), we consider a1/R1 = 7 5.6 10 (i.e. a1 =0.77 AU). Additionally, we consider × 8 a1/R1 = 3.6 10 (i.e. a1 = 5 AU). We adopt param- eters for the× outer perturber from Ford et al. (2000a): m3 = 0.01 M⊙ and e2 = 0.45. We initialize the system ◦ −4 with g1 = g2 = 0 , e1 = 10 , and we also choose two ◦ ◦ different initial inclinations, itot = 65 and itot = 0 . A ◦ mutual inclination of itot = 0 highlights that the per- turbations of the outer orbit affect the inner orbit due to the Newtonian octupole term, even far from the nomi- 1PN nal Kozai-Lidov regime. In this configuration, the ¯ −2 Ha1 term is the most significant 1PN effect, as shown in Fig- ure 4 (see the rectangle in the top left panel). The left-panel of Fig. 6 shows the inner orbit’s maxi- mum eccentricity as a function of the outer orbit’s SMA (or equivalently the value for a1 = 0.77 AU). This figure confirms the resonant–likeR increase in eccentricity found in Ford et al. (2000b) (their Figure 14). This fig- ure also shows that the resonant-like eccentricity increase is present for a large range of a2 values, even for sys- tems where the excitation of the eccentricity due to the Newtonian octupole term is somewhat suppressed, due to comparable masses for the inner orbit (e.g., Naoz et al. Fig. 6.— The maximum eccentricity as a function of the outer 2013). Although changes in the outer orbit’s eccentricity orbit’s SMA (left panel) in a triple for the case of PSR B1620−26. (see the right panels of Fig. 6) do not change the loca- The inner binary is a millisecond pulsar of mass 1.4 M⊙ with a tion of the resonant peaks, their amplitude does change companion of m2 = 0.3M⊙, and the outer body has mass m3 = 0.01M⊙. The inner orbit has a1 = 5 AU and 0.77 AU, in two for large inclinations. The a1 = 5 AU case has irregular different sets of simulations (see labels). The initial eccentricities −4 behavior and results in a higher inner and outer orbital are e1 = 10 and e2 = 0.45 and the initial relative inclination 1PN ◦ ◦ eccentricity. Note that −2 term is the dominant one i = 65 and i = 0 . The argument of pericenter of the inner a tot tot H 1 and outer orbits are initially set to zero. The top axis show the here and the other 1PN terms are negligible in this con- value of R for the case of a1 = 0.77 AU, see equation (47). Note ◦ figuration, as can be see from the black rectangle in the that the itot = 0 resonance happens when R∼ 1. The thin black top left panel in Figure 4. lines show the results of integrating the system over a single Kozai- Lidov cycle. The left thin black line is for the case of a1 = 0.77 AU The right panels of Fig. 6 show the time evolution ◦ ◦ and itot = 65 , while the right one is for a1 = 5 AU and itot = 0 . of the inner and outer orbital eccentricity for the a1 = Although we include all 1PN terms described in previous sections, ◦ 11 0.77 AU case with itot = 0 and a2/R3 = 4.9 10 the curves corresponding to the lower order ones exactly overlap the ◦ × (a2 = 48 AU, bottom panel) and itot = 65 and a2/R3 = curves including the interaction term (i.e., the leading 1PN term 2.2 1011 (a = 22 AU, top panel). Unlike the systems is the most dominant in the evolution of the system). In the right × 2 panels, we show the time evolution of the eccentricity of the inner considered in Fig. 5, the outer orbit’s eccentricity oscil- (blue line) and outer binary (green line). Here we set a1 = 0.77 AU ◦ ◦ lates slightly (see right hand panels in Figure 6). Fur- itot = 0 , a2 = 48 AU in the bottom and itot = 65 , a2 = 22 AU in thermore, the eccentricity peak is even larger than the the top panel, respectively. In this case, the resonant eccentricity excitation due to the 1PN terms reaches higher values then the one eccentricity reached for & 1, which shows that the R achieved by the eccentric Kozai-Lidov mechanism for small (large) 1PN terms further increase the inner orbit’s eccentricity a2 (R). above the excitation induced by the Newtonian eccen- tric Kozai-Lidov mechanism. This was not the case in ◦ the itot = 65 case in the left panel. This also explains Figure 5 primarily because qm 1 there. Although we integrated the≫ system for up to 1000 the somewhat larger e1,max values we found compared to Kozai-Lidov cycles, the eccentric Kozai-Lidov process did Ford et al. (2000b) in the regime where the 1PN effects not seem to induce chaotic behavior in this configura- are subdominant and the eccentricity of the inner orbit increases due to the eccentric Kozai-Lidov mechanism. tion (since m3 portance of the Newtonian octupole term. Without 1PN effects, comparable mass triples result in ǫM 0 [Eq. (21)], which suppresses the eccentricity excitations→ and the flipping of the inner orbits. This can be seen in Fig. 8 (black and gray lines) for two examples, where we consider ∆m = m m = 2 M⊙, (black lines) | 1 − 2| and ∆m = 0.2 M⊙ (gray lines). For both of these, m3 = 3m1 and the two systems were initialized with ◦ e2 = 0.7, e1 = 0.001 and itot = 94 . However, we find that although 1PN effects are small during a single orbit, they can be significant on much longer timescales, and can lead to significant eccentricity growth and orbital- 11 flips . The top panel shows that for ǫM 0.002 a flip is triggered due to the 1PN terms. In other≥ words, for ǫM 0.002 the colored curves deviate (including 1PN effects)≥ from the black and grey ones (Newtonian effects only). For larger ǫM , where the eccentric Kozai-Lidov mechanism is triggered, the maximum eccentricity can be very close to unity, and thus due to the chaotic nature of the system, the maximum eccentricity shown should be considered as a lower limit. For ǫM 0.005 we could not reach convergence after 1000 quadrupole≫ cycles, since the eccentricity is very close to unity (see Teyssandier et Fig. 7.— Eccentricity excitation and orbital flips for the New- al; in preparation). tonian octupole and the 1PN approximations (up to the interac- In the above examples, a requirement for eccentricity tion term, see text for details) as a function of time. We com- 1PN excitation is that the 1PN timescale, t −2 , be shorter pare the test-particle case (m2 → 0 left panels) to a comparable a1 mass case (m2 = 8M⊙ right panels) in the inner binary with than or comparable to the octupole timescale tN , [see m1 = 10 M⊙, always with an outer object of mass m3 = 30 M⊙. oct g 8 N 1PN N The separation of the inner orbit is a1/R = 1 × 10 (correspond- Eq. (38)], i.e., tquad . t −2 toct. A possible explana- 1 a1 ∼ ing to 10 AU and orbital period P1 ∼ 10 yr), and the outer orbit’s g 9 tion for the excitation of the eccentricity in these cases separation is a2/R3 = 1.69 × 10 (corresponding to 502 AU and 3 is the following. Neglecting 1PN effects, comparable- P2 = 2.8 × 10 yr). The initial eccentricities are e1 = 0.001 and ◦ masses in the inner binary suppress the Newtonian e2 = 0.7 and the initial relative inclination itot = 94 . The argu- ments of pericenter of the inner and outer orbits are initially set to octupole effects and the outer potential is effectively 240◦ and zero respectively. For these examples R ≫ 1. We show quadrupolar. However, GR precession of the inner orbit with red lines evolutions without 1PN corrections (curves including breaks this symmetry. As long as GR precession occurs only the lower order 1PN terms simply overlap this). We also show the minimum eccentricity corresponding to the detectable LIGO on a timescale comparable to (or slightly smaller than) frequency range (horizontal lines in the bottom panels). The 1PN the octupole one, the eccentric Kozai-Lidov mechanism corrections help to further increase the eccentricity and lead to is then triggered. orbital flips for the inner binary for comparable masses. Let us now discuss the implications of these finding for direct GW detections using Earth-based instruments like LIGO and VIRGO. The characteristic frequency of the evolution can also be triggered without mass loss, but GW signal is fp = vp/rp, where vp and rp are the orbital accounting for 1PN effects, as shown in Fig. 7. For this velocity and radius at pericenter (Peters 1964). Thus, g 8 figure, we set m = 10 M⊙,m = 30 M⊙, a /R =1 10 1/2 −3/2 −1 1 3 1 1 fp =2π(1 + e) (1 e) P , where P is the orbital (corresponding to 10 AU), and a /Rg = 1.7 109 ×(cor- − 2 3 × period. We assume that the GW signal is in the de- responding to 502 AU). The initial eccentricities were tectable frequency band if fp > 5 Hz. For a comparable- e1 = 0.001 and e2 = 0.7 and the initial relative incli- mass inner binary, the Newtonian eccentric Kozai-Lidov ◦ nation itot = 94 . The argument of pericenter of the mechanism is suppressed12 and the eccentricity remains ◦ inner and outer orbits was initially set to 240 and zero, smaller than in the test particle case. The GW frequency respectively. The two panels in this figure differ in the emitted by a circular binary with an orbital period larger choice of m2, i.e. in the left panel m2 =0.001 and in the than a second is too small for a LIGO detection. How- right panel m2 = 8M⊙. While a test-particle evolution ever, the 1PN eccentricity excitations discussed in this is relatively insensitive to the 1PN terms in this case, paper lead to a much larger fp and might lead to GWs comparable mass systems present qualitatively different in the LIGO band. In particular, Figs. 7 and 8 show −5 −6 behavior. In particular, while the Newtonian eccentric examples where 1 e1 can be as small as 10 to 10 Kozai-Lidov effect is suppressed for comparable masses, for a comparable-mass− inner binary. These sources enter 1PN effects can trigger it. A possible reason for the qual- the LIGO GW frequency band if their orbital period is itative difference is that changing m2 from zero to 8 M⊙ less than 1 to 60 years. However, note that this estimate resulted in a configuration for which the leading order 1PN timescale is closer to (but still slightly longer than) 11 The choice of argument of periapsis does not change the out- the octupole time scale. In the test particle case, the oc- ◦ come; the first example was initialized with g1 = 240 while the ◦ ◦ tupole timescale is two orders of magnitude shorter than second had g1 = 0 , both had g2 = 0 . the shortest 1PN timescale. 12 Although suppressed, note that the eccentricity can still reach The excitation of the eccentricity depends on the im- 0.999 in this case. Resonant 1PN Eccentricity Excitation in Hierarchical Triples 13 Fig. 8.— Excitation of the eccentricity and inclination due to 1PN effects for Kozai-Lidov timescales much shorter than the lowest 1PN timescales (as parametrized through ǫM , see Eq. 21). We examine the maximum eccentricity reached, and plot 1 − e1,max (bottom panel), and the maximum and minimum inclination reached in the runs (top panel). We show two examples. The first system is initialized with ◦ ◦ m1 = 10 M⊙, m2 =8M⊙ m3 = 30 M⊙ and e2 = 0.7. We initialize the system with e1 = 0.001, g1 = 240 , g2 = 0 and itot = 94 . We vary both the inner and outer SMAs to match the different values of ǫM depicted in the figure. We consider the Newtonian approximation −2 −2 (filled black squares) and the three 1PN level of approximations, O(a1 ), O(a2 ) and the interaction term (filled blue triangles, green cross and filled red squares, respectively). In the second system, we set m1 =1M⊙, m2 = 1.2 M⊙ m3 =3M⊙ and e2 = 0.7. We initialize ◦ ◦ the system with e1 = 0.001, g1 = 0 , g2 = 0 and itot = 94 . We vary both the inner and outer SMAs to match the different values of ǫM depicted in the figure. For this example, we consider the Newtonian approximation (empty gray squares) and up to the interaction level of the 1PN approximation (empty magenta circles). We also show the detectable LIGO frequency limit for the first example, where we set a2 = 501 AU and varied a1 between 5 AU to 15 AU (solid brown line). is oversimplified because it neglects the backreaction of nian, focusing on secular and hierarchical three body sys- GW emission on the evolution. The latter strongly re- tems. We expanded the 1PN Hamiltonian in the ratio of duces the SMA of the binaries during close approaches, SMAs (α) to third order beyond leading, i.e. the leading- and may lead to an eccentric inspiral and merger within order terms in the 1PN Hamiltonian perturbation scale −2 a Kozai-Lidov period. The GW inspiral may deliver the here as a1 and we carried out an expansion up to rela- binaries to the LIGO frequency band even if the signal is tive (α3). We also averaged over the orbital timescale outside the LIGO band during the Kozai-Lidov oscilla- of theO inner and outer binary to investigate the long- tions. If the event rate of these sources is sufficiently term secular evolution of the system ( 2). We examined large within the LIGO detection range, these sources the effects of the different 1PN terms§ in this expansion: 1PN could constitute a distinct population for LIGO (Wen 1PN precession of the inner orbit due to ¯ −2 (Eq. 28); 2003; Antonini & Perets 2012). Ha1 1PN 1PN precession of the outer orbit due to ¯ −2 (Eq. 30); Ha2 6. DISCUSSION and a new 1PN interaction term between the two orbits, ¯1PN (Eq. 31), which introduces a new inclination and The Kozai-Lidov mechanism (Kozai 1962; Lidov 1962, int eccentricityH dependent modulation (e.g., Fig. 3). see below), has been shown to play an important role for We compared the different timescales associated with highly inclined hierarchical triples, from planetary sys- the secular Newtonian and different 1PN terms (see tems to stellar size and/or massive compact objects (e.g., Fig. 4). If the timescales associated with the 1PN effects Naoz et al. 2013, and references therein). For an eccen- are much shorter than the timescales associated with the tric outer perturber, the eccentricity of the inner orbit eccentric Kozai-Lidov mechanism, i.e. the secular New- can reach values extremely close to unity, and the incli- tonian timescales, the growth of the eccentricity in the nation can flip from prograde to retrograde (Naoz et al. inner orbit tends to be suppressed. We confirm that the 2011, 2013). The quadrupole Kozai-Lidov oscillations excitation of the eccentricity is indeed suppressed for sys- between the eccentricity and inclination still persist at tems where the Kozai-Lidov timescale is many orders of octupole order, but they are further modulated on long magnitude longer than the 1PN timescales. However, if timescales. the timescales of the 1PN effects are comparable to the We have here studied how the Kozai-Lidov mechanism secular Newtonian ones (see Fig. 4), we found two inter- is affected by 1PN corrections to the three-body Hamilto- 14 Naoz et al. esting regimes that present qualitatively different behav- 1PN effects, is suppressed when m1 m2, since the ior. outer orbit’s potential is effectively quadrupolar.→ As we The first regime is where the 1PN timescales are showed in this paper, 1PN effects can break symmetry comparable but slightly shorter than the Newtonian and excite eccentricity, triggering the eccentric Kozai- Kozai-Lidov timescale. Ford et al. (2000b), studying the Lidov mechanism. As long as 1PN precession occurs on PSR B1620 26 triple system, noted that the inner ec- a comparable timescale (or lower) than the Newtonian − N 1PN N centricity may be greatly increased around some critical octupole precession, i.e, t . t −2 toct, the eccen- 1PN quad a1 value of the outer SMA, due to the ¯ −2 term and the ∼ Ha1 tric Kozai-Lidov mechanism will be triggered. octupole term. We extended this calculation by including Eccentricity excitations are particularly interesting all averaged 1PN terms up to (α3) and the Newtonian in the context of possible GW detections (Wen 2003; octupole term (Naoz et al. 2013),O as well as exploring a Brown & Zimmerman 2010; Armitage & Natarajan wide region of phase space. We confirmed Ford et al. 2005; Sesana 2010). If such excitations were not present, (2000b) result and found a resonant-like behavior, where the frequency of the GWs emitted by the inner binary the inner orbital eccentricity is greatly increased com- would be typically too low for detection with LIGO pared to the Newtonian case. This behavior exists also (see however O’Leary et al. 2009; Kocsis & Levin 2011, when including all averaged 1PN terms and for a wide for eccentric binaries which form in the LIGO band). range of mass ratios and orbital parameters. We param- However, if eccentricity is secularly excited through a eterized the location of the resonant peak in terms of the three-body interaction, the frequency of the GWs is also SMAs by defining a parameter, in Eq. (47), as the ratio increased during pericenter passage, thus bringing the of the leading-order 1PN and secularR Newtonian terms. signals into the detector’s sensitivity band. Such large This parameter depends on the ratio of the mass of the eccentricities would then lead to GW-driven inspiral and outer perturber to the total mass of the inner binary. the eventual merger of binaries. Whether such eccentric The presence of the octupole term is important for the signals can be detected or not will depend on how close resonant 1PN eccentricity excitation, which is most ap- such sources are to Earth. But if detections are made parent in the examples with a small mutual inclination. with sufficiently high signal-to-noise ratio, then GWs For systems where either the inner or the outer binary could be used to measure the eccentricity of the inner shrinks, for example due to GW radiation-reaction, the binary, and thus, distinguish between different source triple may pass through this three-body 1PN resonance. populations. The amplitude and location of the resonance changes due to 1PN terms as a function of . We found that lower ACKNOWLEDGMENTS R mutual inclinations in the prograde regime cause a wider We thank Alessandra Buonanno, Fred Rasio and peak (in terms of ), while a less massive outer body R Gongjie Li for useful discussions, and we also thank Cole tends to produce wider and higher amplitude peaks. A Miller for carefully reading the first draft of the paper detailed investigation of the properties of the resonance is and sending useful comments. We thank our anonymous beyond the scope of this paper, but could be the subject referee for useful remarks. We thank Yoram Lithwick of future investigations. for the use of his allocation time on the computer clus- It is important to note that the outer orbit precession ter Quest. This research was supported in part through and the interaction term affect the overall time evolution the computational resources and staff contributions pro- (see Figure 3). Since these terms are a result of the ex- vided by Information Technology at Northwestern Uni- pansion of the three body 1PN Hamiltonian in α, it is not versity as part of its shared cluster program, Quest. SN surprising that the different terms affect the location of acknowledge partial supported by NASA through a Ein- the resonant like behavior (e.g., Figure 5). It is interest- stein Post- doctoral Fellowship awarded by the Chandra ing however, that they produce a qualitatively different X-ray Center, which is operated by the Smithsonian As- time evolution of the system (e.g., bottom panels of Fig- trophysical Observatory for NASA under contract PF2- ure 3). This suggests that a system evolved under GR 130096. This work was supported in part by NSF grant effects in the presence of a third body has richness to PHY-1114374 and AST-0907890, as well as NASA grants it that should be examined in more detail. This is the NNX08AL43G and NNA09DB30A and NNX11AI49G. subject of future investigation in the framework of direct BK acknowledges support from NASA through Einstein 3-body integration. Postdoctoral Fellowship Award Number PF9-00063 is- The second regime that exhibits qualitatively different sued by the Chandra X-ray Observatory Center, which behavior from that obtained with a quadrupole Newto- is operated by the Smithsonian Astrophysical Observa- nian Kozai-Lidov treatment is when the quadrupolar sec- tory for and on behalf of the National Aeronautics Space ular Newtonian timescales are shorter than the 1PN ones Administration under contract NAS8-03060. NY also and when the inner binary has comparable mass compo- thanks the Institute for Theory and Computation at the nents. The eccentric Kozai-Lidov mechanism, neglecting Harvard Smithsonian Center for Astrophysics for their hospitality. APPENDIX A. TWO BODY SYSTEMS - EFFECTIVE ONE BODY GR pericenter precession has been studied in great detail and used to test Einstein’s theory in the Solar System, for example with observations of the perihelion precession of Mercury (e.g., Shapiro et al. 1972). The simplest method to derive such precession is to consider test-particle motion in an effective potential, assuming that GR introduces Resonant 1PN Eccentricity Excitation in Hierarchical Triples 15 small corrections to Newtonian dynamics and small eccentricities (e.g., Misner et al. 1973, chapter 25 p. 668–670)13. The Hamiltonian (e.g. Artemova et al. 1996; Miller & Hamilton 2002) is simply derived by integrating over the pre- cession rate. Although this Hamiltonian leads to the correct ISCO location, if one uses the full expansion given in Artemova et al. (1996, Eq. 4), it is not equal to the 1PN Hamiltonian. The purely orbital (non-spinning) 3PN Hamiltonian was derived in Jaranowski & Sch¨afer (1998, 2001) (in the center of mass frame, and after subtraction of the total rest-mass term). Here, we focus only on expansions to 1PN order for a two body system (with masses m1 and m2 and momenta p1 and p2, respectively). The Hamiltonian is then (e.g., Buonanno et al. 2006): 3 3 4 2 2 2 2 2 2 4 2 2body (m1 + m2)p k (3m1 +7m1m2 +3m2)p k (p r) k (m1 + m2) µ 1PN = 2 3 3 2 2· 3 + 2 2 , H − 8c m1m2 − 2c m1m2r − 2c r 2c r where r is the radius vector between the two bodies, with magnitude r, the linear momentum of the effective one body problem is simply p = p1 = p2 and µ = m1m2/(m1 + m2). Eliminating the short-period terms in the Hamiltonian, using the Von Zeipel transformation− (for more details, see Brouwer 1959) for an orbit with SMA a and eccentricity e, the double average Hamiltonian is given by 4 2 2 4 k µin 15m1 + 29m1m2 + 15m2 3k m m (m + m ) ¯2body = 1 2 1 2 , (A1) H1PN 8a2c2 − a2c2√1 e2 − which is the same as Eq. (28). B. THE VON ZEIPEL TRANSFORMATION The technique, known as the Von Zeipel transformation (for more details, see Brouwer 1959) is being used in order to eliminate the short-period terms in the Hamiltonian that depend of l1 and l2. The technique had been used to derive the double average hierarchical three body Hamiltonian (e.g. Kozai 1962; Harrington 1968, 1969; Krymolowski & Mazeh 1999; Naoz et al. 2013). Here the Hamiltonian we consider is simply = + (see 2). Following Htot,1PN HN H1PN § Naoz et al. (2013) Appendix A, we replace by our tot,1PN. The equivalent of Equation (A7) at Naoz et al. (2013) is simply: H H = K + K + N + , (B1) Htot,1PN H1 H2 H2 H1PN K K where 1 and 2 are the Kepler Hamiltonians that describe the inner and outer Newtonian orbits in the triple system, N H H 2 describes the Newtonian quadrupole interaction between the orbits (for the octupole interaction one can simply a HN 3 3 ), and 1PN describes the 1PN correction up to (α ). In this technique, we use a canonical transformation that Hcan eliminateH the l and l terms from the N + O parts (which depends on l and l ), where the momenta are 1 2 H2 H1PN 1 2 pi Li, Gi,Hi , and the coordinates are qi li,gi,hi . Replacing 2 from Naoz et al. (2013) Appendix A, with tot∈= { N + } , we find (after following their∈ { derivation)} the equivalentH of their equation (A22) H2 H2 H1PN 2π tot,∗ ∗ ∗ 1 ∗ ∗ tot ∗ ∗ 2 (qi ,pi )= 2 dl1dl2 2 (qi ,pi ) , (B2) H 4π Z0 H where the new momenta and coordinates have a superscript asterix. Since Hamiltonian is an additive quantity, and integral is an additive operation the overall new Hamiltonian after the canonical transformation is simply: 2π 2π tot,∗ 1 ∗ ∗ N ∗ ∗ 1PN 2 = 2 dl1dl2 2 + dl1dl2 . (B3) H 4π Z0 H Z0 H Therefore we can simply use the double averaged Newtonian Hamiltonian derived in Naoz et al. (2013) and separately derive the double averaged 1PN Hamiltonian. C. EQUATION OF MOTIONS FOR THE 1PN INTERACTION Using the canonical relations [eqs. (15)], we find the equations of motion for the interaction part of the 1PN Hamil- tonian: dg k4m m m a 1 = 1 2 3 1 (m + m + m )[(1 e2)(5m2 3m m +5m2) dt 16a3c2(1 e2)3/2(m + m )2(m + m + m )G 1 2 3 − 1 1 − 1 2 2 1PN(int) 2 2 1 2 1 2 3 1 − 2 2 2 1 9fm1m2 ((1 e1)cos2g1 +2cos itot sin g1)] + [ 8fLL + fi] (C1) − − G2 − where L˜1,2 = L1,2/µin,out, 2 2 fm1m2 = m1 + m1m2 + m2 , (C2) 13 Note that the same precession rate can be also derived directly 1112). from the 1PN metric (e.g., Misner et al. 1973, chapter 40 p. 1100- 16 Naoz et al. fLL = L˜1L˜2(m1 + m2)(4(m1 + m2)+3m3) , (C3) 2 2 2 2 2 f 1 = (2 5e )m + 3( 2+ e )m m + (2 5e )m , (C4) e − 1 1 − 1 1 2 − 1 2 and also 2 fi =3a1(m1 + m2 + m3)cos itot(fe1 +3e1fm1m2 cos2g1) (C5) dg k4m m m 1 2 1 2 3 f f = 3 2 2 3/2 2 [8 LL i] (C6) dt −16a c (1 e ) (m + m ) (m + m + m )G1 − 2 − 2 1 2 1 2 3 1 2 [2 cos itot( 8fLL + fi) 16(m1 + m2) cos itotL˜1L˜2[16(m1 + m2)L˜1L˜2(7(m1 + m2)+6m3)cos itot − 2G2 − − 3 2 2 + a (m + m + m )( f 1 [1 + 3 cos 2i ]+18e f 1 2 cos2g sin i )] 2 1 1 2 3 − e tot 1 m m 1 tot de 9a e 1 e2k4m m (m2 + m m + m2)m sin2 i sin(2g ) 1 = 1 1 − 1 1 2 1 1 2 2 3 tot 1 . (C7) dt p 16a3c2(1 e2)3/2L (m + m )2 1PN(int) 2 2 1 1 2 − The change of the inner orbital angular momentum is simply dG 9a e2k4m m (m2 + m m + m2)m sin2 i sin(2g ) 1 = 1 1 1 2 1 1 2 2 3 tot 1 , (C8) dt 16a3c2(1 e2)3/2(m + m )2 1PN(int) 2 2 1 2 − while for the outer orbit it is simply zero. Thus, dH sin i dG 1 = 2 1 , (C9) dt sin i dt 1PN(int) tot i.e., dH sin i 9a e2k4m m (m2 + m m + m2)m sin2 i sin(2g ) 1 = 2 1 1 1 2 1 1 2 2 3 tot 1 , (C10) dt sin i 16a3c2(1 e2)3/2(m + m )2 1PN(int) tot 2 2 1 2 − The inclinations evolve according to (cos˙ i )= H˙ /G G˙ /G cos i (e.g. Naoz et al. 2013), and thus, 1 1 1 − 1 1 1 9a e2k4m m (m2 + m m + m2)m sin2 i sin(2g ) (cos˙ i ) = 1 1 1 2 1 1 2 2 3 tot 1 (C11) 1 16a3c2(1 e2)3/2(m + m )2 1PN(int) 2 2 1 2 − 1 sin i2 cos i1 , × G1 sin itot − and since (cos˙ i )= H˙ /G G˙ /G cos i and H˙ = H˙ (e.g. Naoz et al. 2013) we find 2 2 2 − 2 2 2 2 − 1 2 4 2 2 2 ˙ sin i2 9a1e1k m1m2(m1 + m1m2 + m2)m3 sin itot sin(2g1) (cos i2)= 3 2 2 3/2 2 , (C12) G2 sin itot 16a c (1 e ) (m + m ) 2 − 2 1 2 REFERENCES Amaro-Seoane, P. et al. 2012a, ArXiv e-prints, 1202.0839 Brown, D. A., & Zimmerman, P. J. 2010, Phys. Rev. D, 81, ——. 2012b, ArXiv e-prints, 1201.3621 024007, 0909.0066 Amaro-Seoane, P., Sesana, A., Hoffman, L., Benacquista, M., Buonanno, A., Chen, Y., & Damour, T. 2006, Phys. Rev. D, 74, Eichhorn, C., Makino, J., & Spurzem, R. 2010, MNRAS, 402, 104005, arXiv:gr-qc/0508067 2308, 0910.1587 Damour, T., & Deruelle, N. 1985, Journal des Astronomes Antonini, F., & Perets, H. 2012, ArXiv e-prints, 1203.2938 Francais, 25, 21 Armitage, P. J., & Natarajan, P. 2005, ApJ, 634, 921, Dirac, P. A. M. 1950, Can. J. Math., 2, 937 arXiv:astro-ph/0508493 Dotti, M., Sesana, A., & Decarli, R. 2012, Advances in Artemova, I. V., Bjoernsson, G., & Novikov, I. D. 1996, ApJ, 461, Astronomy, 2012, 1111.0664 565 Eggleton, P. P., Kisseleva-Eggleton, L., & Dearborn, X. 2007, in Arun, K., Iyer, B. R., Sathyaprakash, B., & Sinha, S. 2007a, IAU Symposium, Vol. 240, IAU Symposium, ed. Phys.Rev., D75, 124002, 0704.1086 W. I. Hartkopf, E. F. Guinan, & P. Harmanec, 347–355 Arun, K., Iyer, B. R., Sathyaprakash, B., Sinha, S., & Broeck, C. Fabrycky, D., & Tremaine, S. 2007, ApJ, 669, 1298, 0705.4285 V. D. 2007b, Phys.Rev., D76, 104016, 0707.3920 Finn, L. S., & Lommen, A. N. 2010, ApJ, 718, 1400, 1004.3499 Arun, K. G., Blanchet, L., Iyer, B. R., & Sinha, S. 2009, Ford, E. B., Joshi, K. J., Rasio, F. A., & Zbarsky, B. 2000a, ApJ, Phys. Rev. D, 80, 124018, 0908.3854 528, 336, arXiv:astro-ph/9905347 Blaes, O., Lee, M. H., & Socrates, A. 2002, ApJ, 578, 775, Ford, E. B., Kozinsky, B., & Rasio, F. A. 2000b, ApJ, 535, 385 arXiv:astro-ph/0203370 Galaviz, P., & Br¨ugmann, B. 2011, Phys. Rev. D, 83, 084013, Brouwer, D. 1959, AJ, 64, 378 1012.4423 Resonant 1PN Eccentricity Excitation in Hierarchical Triples 17 Harrington, R. S. 1968, AJ, 73, 190 Naoz, S., Farr, W. M., Lithwick, Y., Rasio, F. A., & Teyssandier, ——. 1969, Celestial Mechanics, 1, 200 J. 2011, Nature, 473, 187, 1011.2501 Hoffman, L., & Loeb, A. 2007, MNRAS, 377, 957, ——. 2013, MNRAS, 431, 2155 arXiv:astro-ph/0612517 Naoz, S., Farr, W. M., & Rasio, F. A. 2012, ApJ, 754, L36, Holman, M., Touma, J., & Tremaine, S. 1997, Nature, 386, 254 1206.3529 Innanen, K. A. 1979, AJ, 84, 960 Nowak, M. A., & Wagoner, R. V. 1991, ApJ, 378, 656 ——. 1980, AJ, 85, 81 O’Leary, R. M., Kocsis, B., & Loeb, A. 2009, MNRAS, 395, 2127, Ivanova, N., Chaichenets, S., Fregeau, J., Heinke, C. O., 0807.2638 Lombardi, J. C., & Woods, T. E. 2010, ApJ, 717, 948, O’Leary, R. M., Rasio, F. A., Fregeau, J. M., Ivanova, N., & 1001.1767 O’Shaughnessy, R. 2006, ApJ, 637, 937, Jaranowski, P., & Sch¨afer, G. 1998, Phys. Rev. D, 57, 7274, arXiv:astro-ph/0508224 arXiv:gr-qc/9712075 Perets, H. B., & Fabrycky, D. C. 2009, ApJ, 697, 1048, 0901.4328 ——. 2001, Phys. Rev. D, 63, 029902 Peters, P. C. 1964, Physical Review, 136, 1224 Katz, B., Dong, S., & Malhotra, R. 2011, ArXiv e-prints, Pribulla, T., & Rucinski, S. M. 2006, AJ, 131, 2986, 1106.3340 arXiv:astro-ph/0601610 Kiseleva, L. G., Eggleton, P. P., & Mikkola, S. 1998, MNRAS, Prodan, S., & Murray, N. 2012, ApJ, 747, 4, 1110.6655 300, 292 Sch¨afer, G. 1987, Physics Letters A, 123, 336 Kocsis, B., & Levin, J. 2011, ArXiv e-prints, 1109.4170 Sesana, A. 2010, ApJ, 719, 851, 1006.0730 Kocsis, B., Ray, A., & Portegies Zwart, S. 2012, ApJ, 752, 67, Seto, N. 2012, Phys. Rev. D, 85, 064037, 1202.4761 1110.6172 Shapiro, I. I., Pettengill, G. H., Ash, M. E., Ingalls, R. P., Kozai, Y. 1962, AJ, 67, 591 Campbell, D. B., & Dyce, R. B. 1972, Physical Review Letters, Krymolowski, Y., & Mazeh, T. 1999, MNRAS, 304, 720 28, 1594 Kulkarni, G., & Loeb, A. 2012, MNRAS, 2670, 1107.0517 Sharpee, B. J., & Thompson, T. A. 2012, ArXiv e-prints, Lidov, M. L. 1962, planss, 9, 719 1204.1053 Lidov, M. L., & Ziglin, S. L. 1976, Celestial Mechanics, 13, 471 Takeda, G., Kita, R., & Rasio, F. A. 2008, ApJ, 683, 1063, Lithwick, Y., & Naoz, S. 2011, ArXiv e-prints, 1106.3329 0802.4088 Lousto, C. O., & Nakano, H. 2008, Classical and Quantum Thompson, T. A. 2010, ArXiv e-prints, 1011.4322 Gravity, 25, 195019, 0710.5542 Tokovinin, A. A. 1997, Astronomy Letters, 23, 727 Mandel, I., Brown, D. A., Gair, J. R., & Miller, M. C. 2008, ApJ, Valtonen, M., & Karttunen, H. 2006, The Three-Body Problem, 681, 1431, 0705.0285 ed. Valtonen, M. & Karttunen, H. Marchal, C. 1990, The three-body problem, ed. Marchal, C. Valtonen, M. J. 1996, MNRAS, 278, 186 Mardling, R. A. 2007, MNRAS, 382, 1768, 0706.0224 Wen, L. 2003, ApJ, 598, 419, arXiv:astro-ph/0211492 Mardling, R. A., & Aarseth, S. J. 2001, MNRAS, 321, 398 Wu, Y., Murray, N. W., & Ramsahai, J. M. 2007, ApJ, 670, 820, Mazeh, T., & Shaham, J. 1979, AA, 77, 145 0706.0732 McKenna, J., & Lyne, A. G. 1988, Nature, 336, 226 Yunes, N., Arun, K., Berti, E., & Will, C. M. 2009, Phys.Rev., Miller, M. C., & Hamilton, D. P. 2002, ApJ, 576, 894, D80, 084001, 0906.0313 arXiv:astro-ph/0202298 Yunes, N., Miller, M. C., & Thornburg, J. 2011, Phys. Rev. D, 83, Misner, C. W., Thorne, K. S., & Wheeler, J. A. 1973, Gravitation, 044030, 1010.1721 ed. Misner, C. W., Thorne, K. S., & Wheeler, J. A. Zhang, K., Hamilton, D. P., & Matsumura, S. 2013, ArXiv Moore, C. 1993, Physical Review Letters, 70, 3675 e-prints, 1302.1620 Morais, M. H. M., & Giuppone, C. A. 2012, ArXiv e-prints, arXiv:1204.4718 Murray, C. D., & Dermott, S. F. 2000, Solar System Dynamics, ed. Murray, C. D. & Dermott, S. F.