1: \documentclass[useAMS,usenatbib,usegraphicx]{mn2e}

2: \usepackage{ifpdf}

3: \usepackage{pdfscreen}

4: %-------------------------------------------------------------------------------

5: % preamble

6: %-------------------------------------------------------------------------------

7: \usepackage{pslatex}

8: \usepackage{amsmath,amsfonts}

9: \newcommand{\Hfull}{\mathcal{H}}

10: \newcommand{\Hsec}{\mathcal{H}_{\idm{sec}}}

11: \newcommand{\Hrel}{\mathcal{H}_{\idm{GR}}}

12: \newcommand{\Hflat}{\mathcal{H}_{\idm{QM}}}

13: \newcommand{\Hbulge}{\mathcal{H}_{\idm{TB}}}

14: \newcommand{\HflatS}{\mathcal{H}_{\idm{spin}}}

15: \newcommand{\HflatT}{\mathcal{H}_{\idm{TB}}}

16: \newcommand{\HsecJ}{\mathcal{H}^J_{\idm{sec}}}

17: \newcommand{\Hpert}{\mathcal{H}_{\idm{pert}}}

18: \newcommand{\Hmut}{\mathcal{H}_{\idm{NG}}}

19: \newcommand{\HpertJ}{\mathcal{H}^J_{\idm{pert}}}

20: \newcommand{\HKepl}{\mathcal{H}_{\idm{kepl}}}

21: \newcommand{\Hint}{\mathcal{H}_{\idm{int}}}

22: \newcommand{\Rsec}{\mathcal{R}_{\idm{sec}}}

23: \newcommand{\Xint}{\mathcal{X}}

24: \newcommand{\R}{\mathcal{R}}

25: \newcommand{\K}{\mathcal{K}}

26: \newcommand{\RP}{${\cal S}$}

27: \newcommand{\RPm}{${\cal S}_{\pi}$}

28: \newcommand{\RPp}{${\cal S}_0$}

29: \newcommand{\TSR}{TSR}

30: \newcommand{\UE}{UE}

31: \def\vec#1{{\mathbf{#1}}}

32: \def\idm#1{{\mbox{\scriptsize #1}}}

33: %\def\corr#1{{\em #1}}

34: \def\corr#1{{ #1}}

35: %

36: \voffset=-0.4in

37: %-------------------------------------------------------------------------------

38: \title[Generalized secular planetary problem]

39: {Secular dynamics of coplanar, non-resonant planetary system

40: under the general relativity and quadrupole moment perturbations}

41: \author[C. Migaszewski and K. Go\'zdziewski]{

42: Cezary Migaszewski$^{1}$\thanks{E-mail: c.migaszewski@astri.uni.torun.pl} and 

43: Krzysztof Go\'zdziewski$^{1}$\footnotemark[1]\thanks{E-mail: k.gozdziewski@astri.uni.torun.pl}\\

44: $^{1}$Toru\'n Centre for Astronomy, Nicolaus Copernicus University, 

45: Gagarin Str. 11, 87-100 Toru\'n, Poland}

46: \begin{document}

47: %-------------------------------------------------------------------------------

48: 

49: \date{Accepted 2008 September 20.  Received 2008 September 19; in original form

50: 2008 April 14}

51: \pagerange{\pageref{firstpage}--\pageref{lastpage}} \pubyear{2008}

52: 

53: \maketitle

54: 

55: \label{firstpage}

56: 

57: %-------------------------------------------------------------------------------

58: \begin{abstract}

59: We construct a secular theory of a coplanar system of $N$-planets not involved

60: in strong mean motion resonances, and which are far from collision zones.

61: Besides the point-to-point Newtonian mutual interactions, we  consider the

62: general relativity corrections to the gravitational potential of the star and

63: the innermost planet, and also a modification of this potential by the

64: quadrupole moment and tidal distortion of the star. We focus on hierarchical

65: planetary systems.  After averaging the model Hamiltonian with a simple

66: algorithm making use of very basic properties of the Keplerian motion, we obtain

67: analytical secular theory of the high order in the semi-major axes ratio. A

68: great precision of the analytic approximation is demonstrated with the numerical

69: integrations of the equations of motion.  A survey regarding model parameters

70: (the masses, semi-major axes, spin rate of the star) reveals a rich and

71: non-trivial dynamics of the secular system. Our study is focused on its

72: equilibria. Such solutions predicted by the classic secular theory,  which

73: correspond to aligned (mode~I) or anti-aligned (mode~II) apsides, may be

74: strongly affected by the gravitational corrections. The so called true secular

75: resonance, which is a new feature of the classic two-planet problem discovered

76: by Michtchenko \& Malhotra (2004), may appear in other, different regions of the

77: phase space of the generalized model. 

78: \corr{

79: We found bifurcations of mode~II which emerge new, yet unknown in the

80: literature, secularly unstable equilibria and a complex structure of the phase

81: space. These equilibria may imply secularly unstable orbital configurations

82: even for initially moderate eccentricities.

83: } 

84: The point mass gravity corrections can affect the long term-stability in the

85: secular time scale,  which may directly depend on the age of the host star

86: through its spin rate. We also analyze the secular dynamics of the

87: $\upsilon$~Andromede system in the realm of the generalized model. Also in this

88: case of the three-planet system, new secular equilibria may appear.

89: \end{abstract}

90: %-------------------------------------------------------------------------------

91: 

92: \begin{keywords}

93: celestial mechanics -- secular dynamics -- relativistic effects -- 

94:  quadrupole moment -- analytical methods -- stationary

95: solutions --  extrasolar planetary systems

96: \end{keywords}

97: 

98: %-------------------------------------------------------------------------------

99: \section{Introduction}

100: %-------------------------------------------------------------------------------

101: The currently known sample of extrasolar planets\footnote{See

102: www://exoplanet.eu, http://exoplanets.org, http://exoplanets.eu} comprises of

103: many multiple-planet configurations. Their architectures are diverse and,

104: usually, very different from the Solar system configuration. Some of them

105: consist of so called hot-Jupiters or hot-Neptunes, with  semi-major axes $\sim

106: 0.05$~au and the orbital periods of a few days.  The multiple-planet systems may

107: also contain companions in relatively distant orbits. Such systems are 

108: non-resonant (or hierarchical). 

109: 

110: The discovery of multi-planet systems and their unexpected orbital diversity

111: initiated interest in their secular evolution. Generally, the analysis of

112: secular interactions, including tidal effects, general relativity (GR), and

113: quadrupole moment (QM) corrections to the Newtonian gravitation (NG) of point

114: masses, follows the theory of compact multiple stellar systems

115: \citep{Mardling2002,Nagasawa2005}.  The secular evolution  of planetary systems

116: has been already intensively studied in this context by many authors. For

117: instance, \cite{Adams2006a} consider the GR corrections in the sample of

118: detected extrasolar systems. They have shown that the GR interactions can be particularly 

119: important for the long-term dynamics of  short-period planets. Also

120: \cite{Adams2006b,Mardling2007} study the evolution of such objects taking into

121: account the tidal circularization of their orbits, and secular excitation of the

122: eccentricity of the innermost planet by more distant companions. These secular

123: effects imply constraints on the orbital elements of the innermost orbits,

124: including yet undetected Earth-like planets, which are expected to be found in

125: such systems. The  quadrupole moment  and relativistic effects may detectably

126: affect the light-curves of stars hosting transiting planets in a few years

127: time-scale  \citep{MiraldaEscude2002,Agol2005,Winn2005}.  In  general, the

128: interplay of apparently subtle perturbations with the  point mass Newtonian

129: gravity depends on many physical and orbital parameters, and it may lead to

130: non-trivial and interesting dynamical phenomena. \corr{Still, 

131: the problem is of particular

132: importance for studies of the long-term stability of the Solar system

133: \citep[see very recent works by][]{Benitez2008,Laskar2008}.}

134: 

135: Because a fully general study of such effects would be a very difficult task, we

136: introduce some simplification to the planetary model. We skip tidal effects

137: caused by dissipative interactions of extended star and planetary bodies, in

138: particular  the tidal friction (TF) damping the eccentricity and the tidal

139: distortion (TD) that modifies planetary figures,  and could influence their

140: gravitational interaction with the parent star. The tidal effects are typically

141: much smaller than the leading Newtonian, relativistic and quadrupole moment

142: contributions \citep{Mardling2007}. Also their time-scale is usually much longer

143: than of the most  significant conservative effects.    In this work, we focus on

144: the secular dynamics of planetary systems over an intermediate time scale

145: relevant to the  conservative perturbations.  Our goal is to obtain a

146: qualitative picture of the secular system that may be useful as the first order

147: approximation to more general model of the long-term planetary dynamics. 

148: 

149: Moreover, we show in this paper that even if we skip the non-conservative  tidal

150: effects then the GR and/or QM {\em corrections} to the 

151: point mass Newtonian gravity can be

152: much more significant for the secular dynamics than the mutual NG interactions

153: alone. These effects may change qualitatively  the view of the dynamics of

154: planetary systems  as predicted by the Laplace-Lagrange theory

155: \citep{Murray2000} or its recent versions

156: \citep[e.g.,][]{Lee2003,Michtchenko2004,Michtchenko2006,Libert2005,Veras2007,Migaszewski2008a}. 

157: 

158: The model without dissipative effects may be investigated  with the help of

159: conservative Hamiltonian theory. Assuming that planetary orbits are well

160: separated and the system is far from mean motion resonances and collision zones,

161: we can apply the averaging proposition \citep{Arnold1993} to derive the long

162: term evolution of their orbital elements. This approach can be classified among

163: secular theories having a long history in the context of the Solar system

164: \citep{Brouwer1961,Murray2000}.  

165: 

166: The present work is a step towards a generalization of the secular model of

167: coplanar 2-planet system by \cite{Michtchenko2004} and its analytical version

168: applied to $N$-planet system in \citep{Migaszewski2008a}.  To explore the phase

169: space globally without restrictions on eccentricities, we simplify the equations

170: of motion through the averaging of perturbations to the Keplerian motion, with

171: the help of semi-numerical method proposed by \cite{Michtchenko2004}. To obtain

172: analytical results, we also use  a very simple averaging algorithm that can be

173: applied to perturbations dependent on the mutual distance of interacting bodies

174: \citep{Migaszewski2008a}. These works demonstrate that the secular evolution can

175: be precisely described in wide ranges of the orbital parameters, including

176: eccentricities up to 0.8--0.9, for well separated (hierarchical)  orbits with

177: small ratio of the semi-major axes, $\alpha\sim 0.1$.  Here, the secular NG

178: theory is very helpful to derive 

179: the more general model including the GR and QM  effects. Basically, without the

180: averaging, the only possibility of investigating the long-term secular dynamics

181: relies on numerical solutions of the equations of motion \citep{Mardling2002}.

182: However, due to extremely different time scales which are related as days (the

183: orbital periods of inner planets) to $10^3$--$10^6$~years of the secular orbital

184: evolution, the numerical integrations are of very limited use when we want to

185: investigate large volumes of initial conditions rather than isolated orbits. In

186: such a case,  the analytical theory can be very helpful to get a deep insight

187: into the secular dynamics. Moreover, when necessary, particular solutions can be studied

188: in detail with the help of the direct numerical integrations.

189: 

190: The plan of this paper is the following. In Sect.~2 we formulate the generalized

191: model of a coplanar, $N$-planet system, accounting for the relativistic and

192: quadrupole moment corrections to the NG-perturbed motion of the innermost

193: planet. To make the paper self-consistent, we average out the perturbations to

194: the Keplerian motion  with the help of the averaging algorithm described in

195: \citep{Migaszewski2008a}. In Sect.~3 we apply the analytical and numerical tools

196: to study the influence of the GR and QM corrections on the secular evolution. 

197: In particular, we recall the concept of the so called representative  plane of

198: initial conditions \citep{Michtchenko2004}. We focus on the  search for

199: stationary solutions in the averaged and reduced systems (periodic orbits in the

200: full systems), and we investigate their stability and bifurcations with the help

201: of phase diagrams.  In this section we also derive interesting conclusions on

202: the stability of the unaveraged systems. In Sect.~4, we study the phase space of

203: two-planet systems in wide ranges of parameters governing their orbital

204: configurations (masses, semi-major axes ratios, eccentricities) and physical

205: parameters (flattening of the star). To illustrate the application of the

206: secular theory to multi-planet configurations, we consider the three-planet

207: $\upsilon$~Andr system, and we investigate the QM (stellar rotation) influence

208: on its secular orbital evolution.

209: %

210: %-------------------------------------------------------------------------------

211: \section{Generalized model of $N$-planet system}

212: %-------------------------------------------------------------------------------

213: %

214: The dynamics of the planetary system can be modeled by the Hamiltonian function

215: written with respect to canonical Poincar\'e  variables \citep[see,

216: e.g.,][]{Poincare1897,Laskar1995}, and    expressed by a sum of two terms,

217: $

218: \Hfull = \HKepl + \Hpert,

219: $

220: where

221: \begin{equation}

222: \HKepl = \sum_{i=1}^{N} {\bigg( \frac{\mathbf{p}_i^2}{2 \beta_i} 

223: - \frac{\mu_i \beta_i}{r_i} \bigg)}

224: \label{HKepl}

225: \end{equation}

226: stands for  integrable part comprising of the direct sum of the relative,

227: Keplerian motions of $N$ planets and the host star. Here, the dominant point

228: mass of the star is ${m_0}$, and $m_i \ll m_0$, $i=1,\ldots,N$ are the point

229: masses of the $N$-planets.  For each planet--star pair we define the mass

230: parameter ${\mu_i=k^2~(m_0+m_i)}$ where $k$ is the Gauss gravitational constant,

231: and ${\beta_i=(1/m_i+1/m_0)^{-1}}$ are the so called reduced masses. We consider

232: the perturbing Hamiltonian  $\Hpert$ as a sum of three terms, 

233: \begin{equation}

234: \Hpert = \Hmut + \Hrel + \Hflat,

235: \label{Hpert}

236: \end{equation}

237: where $\Hmut$ is for the mutual point-mass interactions  between planets,

238: $\Hrel$ is for the general (post-Newtonian) relativity corrections to the

239: Newtonian gravity, and $\Hflat$ takes into account the dynamical flattening and

240: tidal distortion of the parent star. 

241: 

242: We consider the secular effects of $\Hmut$, which can be expressed as follows:

243: \begin{equation}

244: \Hmut = \sum_{i=1}^{N-1} \sum_{j>i}^{N} {\bigg(

245:  - \underbrace{\frac{k^2 m_i m_j}{\Delta_{i,j}}}_{\textrm{\small direct part}} +

246: \underbrace{\frac{\mathbf{p}_i \cdot \mathbf{p}_j}{m_0}}_{\textrm{\small indirect

247: part}}\bigg)},

248: \label{Hmut}

249: \end{equation}

250: where  ${\mathbf{r}_i}$ are for the position vectors of planets relative to the

251: star,  ${\mathbf{p}_i}$ are for their conjugate momenta relative to the {\em

252: barycenter} of the whole $(N+1)$-body system, 

253: ${\Delta_{i,j}=\|\mathbf{r}_i-\mathbf{r}_j\|}$ denote the relative distance

254: between planets $i$  and $j$.

255: 

256: For  hierarchical planetary  systems (weakly interacting binaries), the secular

257: time scale of Newtonian point-mass interactions may be as short as 

258: $10^{3}$~years, up to Myrs. In general, these interactions cause slow

259: circulation of the apsidal lines. The GR correction to the NG potential of the

260: star--planet system also leads to the circulation of pericenters.

261: Moreover, the time scale of this effect may be comparable to that one forced by

262: mutual Newtonian interactions between planets. In  such a situation, one should

263: necessarily include the relativistic corrections to the model of motion.

264: %

265: %-------------------------------------------------------------------------------

266: \subsection{General relativity corrections}

267: %-------------------------------------------------------------------------------

268: %

269: The GR correction will be applied only to the innermost planet, so we skip the

270: direct GR perturbations on the motion of the more distant companions, as well as

271: the mutual GR interactions caused by planetary masses. Usually, the GR 

272: corrections to the Newtonian potential are expressed in terms of the PPN

273: formalism \cite[e.g.,][]{Kidder1995}. Alternatively, we found a very clear paper

274: of \cite{Richardson1988} {conveniently providing explicit Hamiltonian of the two

275: body problem with the GR term}. Following these authors, $\Hrel' \equiv

276: \Hrel/\beta$ (i.e., $\Hrel$ rescaled by the  reduced mass) may be written as

277: follows:

278: \begin{equation}

279: \Hrel' = \gamma_1 \vec{P}^4 + \gamma_2 \frac{\vec{P}^2}{r} + \gamma_3

280: \frac{\left(\vec{r} \cdot \vec{P}\right)^2}{r^3} + \gamma_4 \frac{1}{r^2},

281: \label{Hrel}

282: \end{equation}

283: where $\gamma_1,\gamma_2,\gamma_3,\gamma_4$ are coefficients defined 

284: through

285: \[

286: \gamma_1 = - \frac{\left(1 - 3 \nu\right)}{8 c^2}, \quad

287:    \gamma_2 = - \frac{\mu \left(3 + \nu\right)}{2 c^2},\quad

288: \gamma_3 = \frac{\mu^2}{2 c^2}, \quad

289:   \gamma_4 = - \frac{\mu \nu}{2 c^2},

290: \]

291: and  $c$ is the velocity of light in a vacuum, $\mu = k^2 (m_0 + m_1)$,  $\nu

292: \equiv m_0 m_1 / (m_0 + m_1)^2$, $r$ is the astrocentric distance, and

293: $\vec{P}$ is  the astrocentric momentum of the

294: innermost planet (normalized through the reduced mass): 

295: \begin{equation} \vec{P} = \vec{v} + \frac{1}{c^2}

296: \left[ 4\gamma_1 (\vec{v} \cdot \vec{v}) \vec{v} + 

297: \frac{2\gamma_2}{r} \vec{v} + \frac{2\gamma_4}{r^3} (\vec{r} \cdot \vec{v})

298: \vec{r} \right],

299: \end{equation} 

300: where $ \vec{v} \equiv \dot{\vec{r}}$ stands for the  astrocentric velocity of

301: the innermost planet (the relativistic corrections from  other  planetary bodies

302: in the system are skipped). Hence, in the relativistic Hamiltonian, we put 

303: $\vec{P} = \vec{v}$ with the accuracy of $\vec{O}({c^{-2}})$ and then the

304: Hamiltonian is conserved up to the order of $O({c^{-4}})$.

305: %

306: %-------------------------------------------------------------------------------

307: \subsection{Rotational and tidal distortions of the star}

308: %-------------------------------------------------------------------------------

309: %

310: Fast rotating stars are significantly flattened and that in turn may lead to

311: important deviations of the NG potential. In the absence of a close planet

312: and  non-radial pulsations,  the star has the rotational symmetry, regarding its

313: mass density and shape. The gravitational potential of such a body can  be

314: expanded in harmonic series expressed in terms of  the Stokes coefficients. The

315: well known property of this expansion is that the gravitational potential of

316: rotationally symmetric bodies retains only terms related to the so called {\em

317: zonal harmonics}:

318: \begin{equation}

319: \HflatS = \frac{\mu \beta}{r} \sum_{l=2}^{\infty} 

320: {J_{l} \left(\frac{R_0}{r}\right)^{l}} P_l(\sin\phi),

321: \label{Hflat}

322: \end{equation}

323: where $R_0$ stands for the characteristic radius of a sphere encompassing the

324: body, and it can be fixed as the equatorial radius of the star, $P_l(\sin\phi)$

325: is the Legendre polynomial of the $l$-th order, $\phi$ is the astrocentric

326: latitude and $J_l$ are non-dimensional Stokes coefficients, $l>1$. It can be

327: shown that for rotationally symmetric objects, $J_l=0$ for $l$ odd.  Basically,

328: $J_l$ (and the leading term with $J_2$, in particular) can be determined

329: numerically  with the help of the theory of stellar interiors combined with a

330: model of rotation and helio-seismic data \citep{Godier1999,Pijpers1998}.  To

331: calculate these coefficients, one must know the mass density as well as the

332: hydrostatic figure of the star.  There are attempts to estimate $J_2$ for stars

333: hosting planets.  For instance, \cite{Iorio2006} determined quadrupole moment of

334: HD~209458~\citep{Charbonneau2000} as $J_2 \sim 3.5 \times 10^{-5}$, 

335: \corr{however with large error of $\sim 10^{-3}$, making the result not very

336: credible}.  In general, the estimates of $J_{l}$ are very  uncertain even for  the

337: well explored Sun. For this slowly rotating dwarf (with rotational period of

338: $\sim 30$~days), $J_2 \sim 10^{-7}$ with an upper bound of $10^{-6}$. Moreover,

339: in the literature we found approximations of current  $J_2$ of the Sun in the

340: range up to $10^{-5}$. 

341: 

342: Due to indefiniteness of higher order zonal harmonics, we consider the 

343: dynamical effects of the leading $J_2$ term only.  For a coplanar system, in

344: which planets move around the star in its equatorial plane,  the first non-zero

345: term in harmonic series, Eq.~\ref{Hflat}, has the form of:

346: \begin{equation}

347: \label{eq:hrf}

348: \HflatS \approx -\frac{1}{2} \mu \beta J_2 R_0^2 \frac{1}{r^3}.

349: \end{equation}

350: The above formulae may be conveniently expressed 

351: \citep{Mardling2002} in terms of

352: the stellar spin frequency:

353: \begin{equation}

354: \HflatS \approx -\frac{1}{6} \beta R_0^5 \left(1 + m_1/m_0\right) k_L \Omega^2

355: \frac{1}{r^3},

356: \end{equation}

357: where $k_L$ is the tidal Love number \citep{Murray2000} 

358: which can be related to $J_2$ through:

359: \begin{equation}

360: k_L = 3 \frac{J_2}{q}, \quad q = \frac{\Omega^2 R_0^3}{k^2 m_0},

361: \end{equation}

362: $q$ is the ratio of centrifugal acceleration and gravitation at the

363: surface of the star, and $\Omega$ is the spin rate.

364: Hence, we have:

365: \begin{equation}

366: J_2 = \frac{1}{3} k_L \frac{R_0^3 \Omega^2}{k^2 m_0}.

367: \end{equation}

368: In this work we adopt the standard value of $k_L=0.02$ \citep{Nagasawa2005}.

369: Obviously, the dynamical flattening is more significant for fast rotating stars.

370: Before the Zero Age Main Sequence (ZAMS) stage, the rotational periods may be as

371: low as a few days, down to 1~day for young ($\sim 100$~Myr) Sun-like stars. The

372: rotational periods of $\sim 40$~days are typical for 8~Gyrs old, evolved

373: objects.  According to the scaling rule of  $J_2 \sim \Omega^2$, the zonal

374: harmonics of young objects may be as large as $10^{-4}$.   Because the

375: rotational period may change by two orders of magnitude during the life-time of

376: the star, also the quadrupole moment may change by a few orders of magnitude. 

377: Hence, flattening caused by fast rotation not only can affect significantly the

378: orbital acceleration which can compete with the GR and  mutual NG contributions

379: but it may also introduce a dependence of the system dynamics on the age of the

380: parent star \citep[][see also Sect.~5 and Sect.~6]{Nagasawa2005}.

381: 

382: We also consider a contribution of the tidal bulge (TB) caused by  the presence

383: of point-mass innermost planet, assuming that only the extended star is

384: distorted due to the tidal interactions. Then the Hamiltonian is corrected by the

385: following term \citep{Mardling2002}:

386: \begin{equation}

387: \label{eq:htb}

388: \HflatT = -\beta R_0^5 k^2 m_1 \left(1 + m_1/m_0\right) k_L \frac{1}{r^6}. 

389: \end{equation}

390: Both gravitational corrections are then $\Hflat = \HflatS + \HflatT$.

391: 

392: We account for direct dynamical effects related to $\Hflat$ only in the motion

393: of the innermost planet although these perturbations can be also  quite easily

394: included for the remaining star-planet pairs.  In the realm of the secular

395: theory of non-resonant, hierarchical systems, we assume that other companions

396: are much more distant from the star. For the semi-major axes ratio $\sim 0.1$,

397: the GR+QM perturbations acting on such outer companions are by orders of

398: magnitude smaller than for the innermost planet. However, the dynamics of the

399: innermost body influences  indirectly the secular dynamics of the whole system.

400: This will be demonstrated in this work in the cases of two- and three-planet

401: configurations. We also underline that with the simplified  model of

402: interactions, we can study the dynamics of quite compact star--planets 

403: configurations  because they can be parameterized, in general, by 

404: individual planetary

405: masses, semi-major axes, and physical parameters of the star.

406: %

407: %-------------------------------------------------------------------------------

408: \subsection{The secular model of non-resonant planetary system}

409: %-------------------------------------------------------------------------------

410: %

411: To apply the canonical perturbation theory, we first transform $\Hfull$ to the

412: following form:

413: \begin{equation}

414:  \Hfull(\vec{I},\vec{\phi}) = \HKepl(\vec{I}) + \Hpert(\vec{I},\vec{\phi}),

415: \label{poincare} 

416: \end{equation}

417: where $(\vec{I},\vec{\phi})$ stand for the action-angle variables, and

418: $\Hpert(\vec{I},\vec{\phi}) \sim \epsilon \HKepl(\vec{I})$,  where $\epsilon \ll

419: 1$ is a small parameter. In this work, we apply the well known 

420: approach to analyze the equations of motion induced by Hamiltonian,

421: Eq.~\ref{poincare}, that relies on the averaging proposition \citep[see,

422: e.g.,][]{Arnold1993}.  By averaging the perturbations with respect to the fast

423: angles ({\em the mean longitudes} or {\em the mean anomalies}) over their

424: periods,  we obtain the secular Hamiltonian which does not depend on these fast

425: angles. Simultaneously,  the  conjugate momenta to the fast angles become

426: integrals of the secular problem.  In the planetary system with a dominant

427: stellar mass, the orbits (apsidal lines) also slowly rotate due to mutual

428: interactions, hence the longitudes of periastron and the longitudes of node

429: become slow angles. Assuming that no strong  mean motion resonances are present,

430: and the system is far enough from collisions, the averaging makes it possible to

431: reduce the number of the degrees of freedom, and to obtain qualitative

432: information on the long-term changes of the slowly varying orbital elements

433: (i.e., on the slow angles and their conjugate momenta).

434: 

435: The transformation of the Hamiltonian to the required form (Eq.~\ref{poincare})

436: may be accomplished   by expressing this Hamiltonian with respect to the

437: (modified) Delaunay canonical elements. These variables can be related to the

438: Keplerian canonical elements \citep{Murray2000}. Actually, we use the following

439: set of canonical action-angle variables which can be obtained after an

440: appropriate canonical transformation of the Delaunay elements:

441: \begin{eqnarray}

442: \label{dvars}

443: {l_i \equiv \mathcal{M}_i}, & \quad {L_i=\beta_i~\sqrt{\mu_i~a_i}},\nonumber\\

444: {g_i \equiv \varpi_i}, & \quad  {G_i=L_i~\sqrt{1-e_i^2}},\\

445: {h_i \equiv \Omega_i}, & \quad{H_i=G_i~(\cos~I_i}-1),\nonumber

446: \end{eqnarray}

447: where $\mathcal{M}_i$ are the mean anomalies,  $a_i$ stand for canonical

448: semi-major axes,  $e_i$ are the eccentricities,   $I_i$ denote inclinations,

449: $\varpi_i$ are the longitudes of pericenter, and $\Omega_i$ denote the

450: longitudes of ascending node. The choice of $\varpi_i$ instead of $\omega_i$ is

451: important for further applications, because $\Omega_i$ are undefined (and

452: irrelevant) for the dynamics of the coplanar system. We note that the

453: geometrical, canonical elements ($a_i$, $e_i$, $I_i$, $\varpi_i$, $\Omega_i$)

454: may be derived through the formal transformation between classic (astro-centric)

455: Keplerian elements and the relative  Cartesian coordinates

456: \citep[e.g.,][]{FerrazMello2006a,Morbidelli2003}, with appropriate rescaling of the astrocentric

457: velocities. In the settings adopted here, the Cartesian  coordinates are

458: understood as Poincar\'e coordinates, i.e., {\em astrocentric} positions of

459: planets, and canonical momenta taken relative to the {\em barycenter} of the

460: system. 

461: 

462: The $N$-planet Hamiltonian expressed in terms of the modified  Delaunay variables

463: (\ref{dvars}) has the form of:

464: \[

465: \Hfull = -\sum_{i=1}^N \frac{\mu_i^2 \beta^3_i}{2 L_i^2} +

466: \Hpert\underbrace{(L_i,l_i,G_i,g_i,H_i,h_i)}_{i=1,\ldots,N}.

467: \]

468: In this Hamiltonian, $l_i$ play the role of the fast angles. In the absence of

469: strong mean motion resonances, these angles can be eliminated by the following

470: averaging formulae:

471: \begin{eqnarray}

472: \label{secular}

473: && \Hsec=\frac{1}{(2 \pi)^N}\underbrace{\int_0^{2 \pi} \ldots 

474: \int_0^{2\pi}}_{i=1,\ldots,N}{\Hmut \, d\mathcal{M}_1 \ldots d\mathcal{M}_N} +\\

475: && \quad + \frac{1}{2\pi} \int_0^{2\pi}{\Hrel \,d\mathcal{M}_1} 

476: + \frac{1}{2\pi} \int_0^{2\pi}{\Hflat \,d\mathcal{M}_1}\nonumber.

477: \end{eqnarray}

478: Hence, we can rewrite the secular Hamiltonian in the symbolic form of:

479: \begin{equation}

480: \Hsec = \left<\Hmut\right> + \left<\Hrel\right> + \left<\Hflat\right>.

481: \label{Hsec}

482: \end{equation}

483: Now, we try to average out each component of this Hamiltonian expressed  with

484: respect to the fast angles (the

485: mean anomalies).

486: %

487: %-------------------------------------------------------------------------------

488: \subsection{Averaging Newtonian point-to-point interactions}

489: %-------------------------------------------------------------------------------

490: %

491: A simple averaging of  the Hamiltonian of the classic planetary

492: $N+1$-body model is described in our previous paper \citep{Migaszewski2008a}. 

493: The algorithm makes use on the very basic properties of the Keplerian motion and

494: it relies on appropriate change of integration variables in Eq.~\ref{secular}.

495: It may be also applied to  perturbations expressed through powers of the

496: relative distance.

497: 

498: To recall the main result, the secular Hamiltonian  of the $N$-planet system can be

499: described as a sum of two-body Hamiltonians evaluated over all pairs of planets:

500: \begin{equation}

501: \left<\Hmut\right> = \sum_{i=1}^{N-1}  \sum_{j>i}^{N}{\left<\Hmut^{(i,j)}\right>}.

502: \label{avHmut}

503: \end{equation}

504: The direct part of the disturbing Hamiltonian reads as follows:

505: \begin{eqnarray}

506: \label{expanssion}

507: & & \left<\Hmut^{(i,j)}\right> = -\frac{k^2 m_i m_j}{a_j} \times  \nonumber \\

508: && \quad \times \left[1 + \sqrt{1-e_j^2} \sum_{l=2}^{\infty}

509: {\left(\frac{\alpha_{i,j}}{1-e_j^2}\right)^l

510: \mathcal{R}^{(i,j)}_l(e_i,e_j,\Delta\varpi_{i,j})}\right].

511: \end{eqnarray}

512: The explicit formulae for functions $\R^{(i,j)}_l(e_i,e_j,\Delta\varpi_{i,j})$ are

513: given in \citep{Migaszewski2008a}. It is well known that the indirect part

514: averages out to a constant and it does not contribute to the secular dynamics

515: \citep{Brouwer1961}. 

516: %

517: %-------------------------------------------------------------------------------

518: \subsection{Averaging the PPN relativistic potential}

519: %-------------------------------------------------------------------------------

520: %

521: Making use of the formulae for the relativistic PPN Hamiltonian by

522: \cite{Richardson1988}, we write down the mean relativistic potential as follows:

523: \begin{equation}

524: \left<\Hrel'\right> = \gamma_1 \left<\vec{v}^4\right> + \gamma_2

525: \left<\frac{\vec{v}^2}{r}\right> + \gamma_3 \left<\frac{\left(\vec{r} \cdot

526: \vec{v}\right)^2}{r^3}\right> + \gamma_4 \left<\frac{1}{r^2}\right>,

527: \label{Hrelsec}

528: \end{equation}

529: with the accuracy of $O(c^{-2})$.

530: We should average out each component of this Hamiltonian over the mean

531: anomalies. It appears to be possible with  the algorithm in

532: \citep{Migaszewski2008a}. To average out the whole PPN Hamiltonian, we must

533: calculate a  few integrals  written in the general form of:

534: \begin{equation}

535: \left<\Xint\right> \equiv \frac{1}{2\pi} \int_0^{2\pi} {\Xint \,

536: d\mathcal{M}}

537: \equiv \frac{1}{2\pi} \int_0^{2\pi} {\left[\Xint \, {\cal J}\right] \, df},

538: \end{equation}

539: where ${\cal J}$ is a scaling function defined with: 

540: \begin{equation}

541: {\cal J} \equiv \frac{d\mathcal{M}}{df} = \frac{\left(1 - e^2\right)^{3/2}}{\left(1 + e

542: \cos{f}\right)^2},

543: \end{equation}

544: and $f$ is the true anomaly of the innermost planet (note that we skip index

545: "1" of that planet). Components  of the

546: mean Hamiltonian can be written explicitly as follows:

547: \begin{eqnarray}

548: \label{v4}

549: &&\left<\vec{v}^4\right> = \frac{n^4 a^4}{\sqrt{1 - e^2}} \left[1 + \frac{3}{2}

550: e^2 + e^4 \mathcal{F}_1\right],\\

551: \label{v2r}

552: &&\left<\frac{\vec{v}^2}{r}\right> = \frac{n^2 a}{\sqrt{1 - e^2}} \left[1 + e^2

553: \mathcal{F}_2\right],\\

554: \label{rvr3}

555: &&\left<\frac{\left(\vec{r} \cdot \vec{v}\right)^2}{r^3}\right> = \frac{n^2 a}{\sqrt{1 - e^2}} e^2

556: \mathcal{F}_2,\\

557: \label{r2}

558: &&\left<\frac{1}{r^2}\right> = \frac{1}{a^2 \sqrt{1 - e^2}}.

559: \end{eqnarray}

560: Functions $\mathcal{F}_1$ and $\mathcal{F}_2$ are given through:

561: \begin{eqnarray}

562: && \mathcal{F}_1 = \frac{1}{2\pi} \int_0^{2\pi} {\frac{\sin^4{f}}{\left(1 + e

563: \cos{f}\right)^2}~df},\\

564: && \mathcal{F}_2 = \frac{1}{2\pi} \int_0^{2\pi} {\frac{\sin^2{f}}{1 + e \cos{f}}~df}.

565: \end{eqnarray}

566: These integrals can be calculated in the closed form:

567: \begin{equation}

568: \mathcal{F}_1 = \frac{3 \left(2 - e^2 - 2\sqrt{1-e^2}\right)}{2 e^4}, \quad

569: \mathcal{F}_2 = \frac{1-\sqrt{1-e^2}}{e^2}.

570: \end{equation}

571: Finally, the secular relativistic potential is the following:

572: \begin{equation}

573: \left<\Hrel'\right> = - \frac{3 \mu^2}{c^2 a^2 \sqrt{1 - e^2}} +

574: \frac{\mu^2 \left(15 - \nu\right)}{8 a^2 c^2}.

575: \label{eq28}

576: \end{equation} 

577: {

578: Because all secular Hamiltonian corrections we account for in this work (see

579: below) do not depend on the  mean anomalies, $L_{1,2}$ are constants of motion. 

580: It also means  that the second term

581: in Eq.~\ref{eq28} reduces to

582: the constant and does not contribute to the long-term dynamics.

583: }

584: Writing down the secular Hamiltonian with respect to  the action-angle

585: variables, we have:

586: \begin{equation}

587: \left<\Hrel'\right> = -\frac{3 \mu^4 \beta^4}{c^2 L^3 G} + \mbox{const}.

588: \label{avHrel}

589: \end{equation}

590: Here, the action variables $(L,G)$ are defined with usual formulae known

591: in the Keplerian motion theory. We recall that in the secular PPN Hamiltonian,

592: only the star-innermost planet interactions are considered.  The  perturbation

593: has been also scaled by the reduced mass. To  properly add the GR star-inner

594: planet  interactions to the Hamiltonian of $N$-planets, we should account for

595: the mass factor. Because the canonical Delaunay elements of the $N$-planet system are

596: defined   through Eq.~\ref{dvars}, hence, to obtain the proper scaling, we should  multiply

597: the two-body Hamiltonian, Eq.~\ref{avHrel}, through $\beta \equiv 1/(1/m_0 +

598: 1/m_1)$.  Finally, we obtain the following secular Hamiltonian related to the GR

599: correction: 

600: \begin{equation} 

601: \left<\Hrel\right> = -\frac{3 \mu^4 \beta^5}{c^2 L^3 G} + \mbox{const}. 

602: \label{avbHrel}

603: \end{equation} 

604: In this Hamiltonian all slow angles are cyclic, 

605: hence $L$ and $G$ would be integrals of motion in the absence of 

606: {mutual planetary interactions}. However, the element $G$ is

607: no longer constant when we consider the NG contributions. 

608: The evolution of canonical angle $\varpi$, reads as follows: 

609: \begin{equation}

610: \dot{\varpi}_{\idm{GR}} = \frac{\partial \left<\Hrel\right>}{\partial G} =

611: \frac{3 \mu^4 \beta^5}{c^2 L^3 G^2} = \frac{3 \mu^{3/2}}{c^2 a^{5/2} \left(1 -

612: e^2\right)}. 

613: \label{orel}

614: \end{equation} 

615: This is the well known formulae describing the  {\em

616: relativistic advance of pericenter}. We derive it here to keep the paper

617: self-consistent.  In the above formulae,  we skip  the symbol

618: $\left<\ldots\right>$ of the mean, which basically should encompass the

619: canonical angle $\varpi$  and the conjugate actions ($L, G$) as well as the

620: symbols of orbital elements ($a,e$).  However, we should keep in mind that after

621: calculating the average over the fast angles, these actions/elements have the

622: sense of the {\em mean} actions/elements.

623: %

624: %-------------------------------------------------------------------------------

625: \subsection{The effect of quadrupole moment of the star}

626: %-------------------------------------------------------------------------------

627: %

628: To average out the quadrupole moment of the star, we have to calculate integrals

629: from the astrocentric distance of the planet taken in the power of ``-3'' for the spin

630: distortion, and the power of ``-6'' for the tidal distortion. To calculate these averages,

631: we express the astrocentric distance $r$ of the inner planet with respect to the

632: true anomaly and  we replace the integration variables $d\mathcal{M} =

633: \mathcal{J} df$. For $l>1$, we  find

634: \begin{equation}

635: \left<\frac{1}{r^l}\right> = \frac{(1 - e^2)^{1/2}}{a^l 

636: \left(1 - e^2\right)^{l-1}}

637: \sum_{s=0}^{l-2} {{{l-2}\choose{s}} e^s \left<\cos^s{f}\right>},

638: \end{equation}

639: where, for $s$ even, the average of $\left<\cos^s{f}\right>$  over the mean

640: anomaly reads as follows:

641: \begin{equation}

642: \left<\cos^s{f}\right> = \frac{2^s \pi }{\Gamma \left(\frac{1}{2}-\frac{s}{2}\right)^2

643: \Gamma (s+1)},

644: \end{equation}

645: while for $s$ odd, the average over the mean anomaly

646: $\left<\cos^s{f}\right>=0$.  The average of the leading term in

647: $\left<\HflatS\right>$ has the form of:

648: \begin{equation}

649: \left<\HflatS\right> \approx -\frac{\beta  k_L

650:     \left(1 + m_1/m_0\right) R_0^5 \Omega^2}{6 a^3

651:     \left(1-e^2\right)^{3/2}}.

652: \label{avHflat}

653: \end{equation}

654: Having this secular Hamiltonian, we can calculate the apsidal 

655: frequency forced by the spin-induced QM of the star:

656: \begin{equation}

657: \dot{\varpi}_{\idm{spin}} = 

658: \frac{\partial\,\left<\HflatS\right>}{\partial\,G} \approx 

659: \frac{\beta^7 k_L (1 + m_1/m_0) \mu ^3 R_0^5 \Omega^2}{2 G^4 L^3}.

660: \label{oqm}

661: \end{equation}

662: 

663: Similarly, we calculate the leading term in $\left<\HflatT\right>$:

664: \begin{equation}

665: \left<\HflatT\right> \approx -\frac{\beta k_L R_0^5 \left(\frac{3}{8} e^4+3 e^2+1\right) k^2

666:     m_1 \left(1+m_1/m_0\right)}{a^6 \left(1-e^2\right)^{9/2}},

667: \label{avHflatT}

668: \end{equation}

669: and the apsidal frequency due to this correction:

670: \begin{equation}

671: \dot{\varpi}_{\idm{TB}} = 

672: \frac{\partial\,\left<\HflatT\right>}{\partial\,G} \approx 

673: \frac{15 \beta^{13} k_L R_0^5 \left(G^4-14 L^2 G^2+21 L^4\right)

674:     m_1 \mu^7}{8 G^{10} L^7 m_0}.

675: \label{oqmT}

676: \end{equation}

677: The tidal bulge induced by a close companion may be 

678: very important in some  configurations.  In fact, the TB effect may be 

679: even dominant

680: over the spin-induced quadrupole moment perturbations. For instance, for $\upsilon$~Andr~b it is

681: $\sim 20$~times as big \citep[see][for details]{Mardling2002,Nagasawa2005}.

682: %

683: %-------------------------------------------------------------------------------

684: \section{Tools to analyze the secular dynamics}

685: %-------------------------------------------------------------------------------

686: %

687: To overview the dynamical influence of the GR and QM  corrections on

688: the point mass NG dynamics, we compare the {\em secular} evolution of the

689: eccentricity of the innermost planet in the sample of detected multiple extrasolar

690: systems. In this experiment, we basically follow \cite{Adams2006a}, however,

691: solving the {\em averaged} equations of motion by means of the numerical

692: integrator. Moreover, we carry out this test to illustrate the different and rich

693: behaviors of  multi-planet systems in the detected sample; a systematic

694: analysis is described in subsequent sections. The results are shown 

695: in Fig.~\ref{fig:fig1}. Each panel in this figure is labeled with the  parent star name.

696: The orbital elements of the selected systems, and the parameters of their parent

697: stars come  from the Jean Schneider  Encyclopedia of Extrasolar

698: Planets\footnote{http://exoplanet.eu}. In this experiment, we fixed 

699: rotational period of the star, $T_{\idm{rot}}=30$~days, 

700: $k_L=0.02$ (hence $J_2 \sim 10^{-7}$--$10^{-6}$).   In some cases, the differences between predictions of the

701: generalized theory and by the classical model are significant, in some other

702: cases both theories give practically the same  outcome.  Looking at the elements

703: of the examined systems, we may conclude that the corrections become very

704: important  for systems with the innermost planet  close to the star, and with

705: other bodies relatively distant.  

706: 

707: In fact, it is already well known \cite[e.g.,][]{Adams2006a} that the additional

708: effects are important when the strength and characteristic time scale of

709: perturbations stemming from the GR and QM become comparable with the secular

710: time-scale of mutual NG interactions.  Then the GR and QM  regarded as {\em

711: perturbing} effects may induce significant  contribution to the secular

712: evolution and the interplay of these effects with the NG interactions may lead

713: to very complex and rich dynamics. As we will see below, this condition is

714: particularly well satisfied  for strictly hierarchical  systems, e.g.,

715: HD~217107, HD~38529 \citep{Fischer2001},  HD~190360~\citep{Vogt2005},

716: HD~11964~\citep{Butler2006},  $\upsilon$ Andromedae~\citep{Butler1999}. The

717: secular time scales can be also comparable when the planetary masses are

718: relatively small because the NG-induced apsidal  frequency decreases

719: proportionally to the products of these masses while, in the first

720: approximation, the GR and spin-induced apsidal motion of the innermost orbit

721: does not depend directly on the planetary mass (see Eqs.~\ref{orel},~\ref{oqm}).

722: 

723: Still, the  results illustrated in Fig.~\ref{fig:fig1}  have limited

724: significance for the study of the global dynamics. Drawing one-dimensional

725: time--orbital  element plots, we can analyze the dynamical evolution only for a

726: few isolated initial conditions. This can be a serious drawback, if we recall

727: that the initial conditions of the discovered systems are still known with large

728: uncertainties. Some  critically important parameters governing the secular

729: evolution, like the masses, nodal lines and inclinations of orbits are poorly

730: constrained by the observations or undetermined at all.  There is also a

731: technical problem: the  direct integrations over the secular time-scale are CPU

732: intensive. Yet looking at  isolated configurations, we obtain only a local view of

733: the dynamics. Instead, following the methodology of Poincar\'e, we should try to

734: understand globally the perturbing effects and the resulting dynamics. We should

735: investigate the whole families of solutions rather than a few isolated

736: phase-space trajectories. In such a case, the application  of secular analytical

737: theories become critically important.

738: \ifpdf

739: \begin{figure*}

740: \centering

741: \includegraphics[width=17.6cm]{fig1.pdf}

742: \caption{

743: The long-term, secular  evolution of eccentricities of the innermost planets of

744: known non-resonant extrasolar systems. Red curves are for the point-mass

745: planets, interacting mutually through Newtonian forces. The blue curves are for 

746: generalized model of motion, including additional effects  (general relativity

747: and the quadrupole moment of the star). For all  tested planetary systems, we

748: fix $T_{\idm{rot}} = 30$~days, $k_L=0.02$. The names of  parent stars are

749: labeling each panel together with  a number of planets written in brackets.

750: }

751: \label{fig:fig1}

752: \end{figure*}

753: \else

754: \begin{figure*}

755: \centering

756: \includegraphics[width=17.6cm]{fig1.eps}

757: \caption{

758: The long-term, secular  evolution of eccentricities of the innermost planets of

759: known non-resonant extrasolar systems. Red curves are for the point-mass

760: planets, interacting mutually through Newtonian forces. The blue curves are for 

761: generalized model of motion, including additional effects  (general relativity

762: and the quadrupole moment of the star). For all  tested planetary systems, we

763: fix $T_{\idm{rot}} = 30$~days, $k_L=0.02$. The names of  parent stars are

764: labeling each panel together with  a number of planets written in brackets.

765: }

766: \label{fig:fig1}

767: \end{figure*}

768: \fi

769: %

770: %-------------------------------------------------------------------------------

771: \subsection{Representative plane of initial conditions and equilibria}

772: %-------------------------------------------------------------------------------

773: %

774: The simplest class of solutions which can be studied most effectively, are the

775: equilibria (or stationary solutions).  In the multi-parameter dynamical systems,

776: their positions (coordinates), stability, and bifurcations provide  information

777: on the general structure of the phase space. Hence, we focus on the

778: stationary solutions emerging in the secular model of a two-planet system with

779: the GR and QM corrections. 

780: 

781: After the averaging of $\Hfull$ over the mean anomalies (i.e., over

782: the orbital periods), the secular Hamiltonian ($\Hsec$) does not depend on $l_i

783: \equiv \mathcal{M}_i$ anymore. Therefore, as we mentioned already,

784: the conjugate actions $L_i$ become

785: constants of motion (hence, the mean semi-major axes are also constant).

786: Moreover, in the coplanar problem the longitudes of nodes are undefined and

787: irrelevant for the dynamics. The secular energy of two-planet system does not

788: depend on individual longitudes of pericenters $\varpi_i$, but only on the their

789: difference  $\Delta\varpi = \varpi_1-\varpi_2$ \citep{Brouwer1961}. These basic

790: facts can be expressed  with the help of appropriate canonical transformation

791: \citep{Michtchenko2004}:

792: \begin{eqnarray}

793: \label{eqtot1}

794: \Delta\varpi \equiv \varpi_1 - \varpi_2, \quad &&G_1,\\

795: \label{eqtot2}

796: \varpi_2, \quad &&C = G_1 + G_2.

797: \end{eqnarray}

798: Because $\varpi_2$ is the cyclic angle, the conjugate momentum equal to the

799: total angular momentum $C$ is conserved. For a fixed 

800: angular momentum as a constant parameter, the  phase-space  of reduced system

801: become two-dimensional,  and  the system is integrable.  We can now choose

802: $\Delta\varpi$ as the canonical angle and $G_1$ as the

803: conjugate momentum. Alternatively, the role of the momentum may be attributed to

804: $e_1$. Obviously,  $G_2$ (or $e_2$) becomes a dependent

805: variable (through the $C$ integral).  

806: 

807: To study the dynamics of the reduced secular system in a global manner, we

808: follow a concept of the so called  {\em representative plane} of initial

809: conditions introduced by \cite{Michtchenko2004}. The representative plane

810: (\RP{}) comprises of points in the phase space which  lie on a specific plane

811: crossing all phase trajectories of the system. In the cited paper, it has been

812: shown that a good choice of the \RP{}-plane flows from the condition of

813: vanishing derivatives of the secular Hamiltonian over $\Delta{\varpi}$. It is

814: equivalent to the symmetry of secular interactions with respect to fixed apsidal

815: line of a selected orbit and it means that time derivatives of the conjugate

816: momenta (which can be expressed by eccentricities) must vanish.  Indeed, in

817: accord with the law of conservation of the  total angular momentum,  the

818: eccentricities must reach at the same time maximal and minimal values along the phase

819: trajectories, hence $\dot{e_i}=0$, $i=1,2$.  Regarding the classic problem with

820: Newtonian point-to-point interactions, this condition is satisfied when

821: $\Delta{\varpi}=0$ or $\Delta{\varpi}=\pi$ \citep{Michtchenko2004} [see also

822: \citep{Migaszewski2008a}]. If we consider the generalized problem of two-planet

823: system then the concept of the representative plane is still valid. The {\em

824: small} perturbations which we introduce do not change the dimension of the phase

825: space and the motion remains on perturbed Keplerian orbits, only the form of the

826: secular Hamiltonian is modified. Moreover, we may expect that qualitative

827: properties of the secular system may be changed because we introduce 

828: a few new

829: parameters describing physical setup of the  studied system and interactions

830: governing its dynamical evolution. To encompass both the specific values of

831: angle $\Delta{\varpi}=0,\pi$, the representative plane can be defined by the

832: following set of points:

833: \[

834: {\cal S} = \{ e_1 \cos{\Delta{\varpi}} \times e_2;  e_1, e_2 \in [0,1) \}, 

835: \]

836: where $\Delta\varpi=\pi$,  $e_1\cos{\Delta{\varpi}}<0$ for the left-hand

837: half-plane (called the \RPm-plane, from hereafter), and $\Delta\varpi=0$,

838: $e_1\cos{\Delta{\varpi}}>0$ are for the right-hand half-plane (the \RPp-plane).

839: Hence, the sign of $e_1 \cos\Delta\varpi$ on the $x$-axis tells us on the value of angle

840: $\Delta\varpi$.

841: 

842: \subsection{A general view of the representative plane}

843: To illustrate the influence of additional perturbations on the secular dynamics

844: of the classic model, we compute a set of the \RP{}-planes for wide ranges of

845: orbital and physical parameters of the studied two-planet  configurations.

846: Before we discuss these results, at first  we explain how the \RP-{plane}

847: illustrates  the dynamical structure of the phase space.  Conveniently,

848: Figure~\ref{fig:fig2}, which is derived for {\em specific} orbital parameters

849: (given in the caption), reveals  {\em all} relevant features in a condensed form

850: and its description can be regarded as a guide useful for the further analysis.

851: 

852: We start from the left-hand panel of Fig.~\ref{fig:fig2}. Smooth half-ellipse

853: like curves marked with thin lines 

854: are for the levels of the secular energy of the

855: generalized problem. The straight lines  are for the collision line of orbits

856: defined through  $a_1 (1\pm e_1) = e_2 (1-e_2)$.

857: %

858: \ifpdf

859: \begin{figure*}

860: \centerline{

861: \hbox{\includegraphics[width=8.8cm]{fig2a.pdf}

862:       \hspace*{0cm}   

863:       \includegraphics[width=8.8cm]{fig2b.pdf}

864: }

865: }

866: \caption{

867: The representative plane $(e_1 \cos{\Delta{\varpi}}, e_2)$ for the two-planet

868: system. The elements are $m_1=1.359~\mbox{m}_{\idm{J}}$,

869: $m_2=0.453~\mbox{m}_{\idm{J}}$, $a_1=0.1$~au, $a_2=1.0$~au. Stellar mass is

870: 1~$M_{\sun}$, rotational period is $T_{\idm{rot}}=30$~days, stellar radius is

871: 1~$R_{\sun}$, $k_L=0.02$. The thick curves are for the stationary solutions: red

872: curves are for stable equilibria in the classic model, blue and violet curves

873: are for stable and unstable equilibria of the generalized model, respectively.  

874: {\em The left panel}:  Color contours are for the ratio of the  apsidal

875: frequency induced by the general relativity and quadrupole moment  to the 

876: apsidal frequency caused by mutual interactions between planets, $\kappa$.

877: \corr{The scale of color code is limited by $\kappa=1$, meaning that the

878: corrections become more important than the point mass Newtonian gravity. In

879: general, the yellow colour may refer to $\kappa>1$.} Energy levels of the

880: generalized model are marked with thin lines.  {\em The right panel}: A few

881: particular energy levels of the generalized model are labeled with

882: \textbf{a}---\textbf{f}, accordingly.  See also phase diagrams for these energy

883: levels  which are shown in Fig.~\ref{fig:fig3} and Fig.~\ref{fig:fig4}. Curves

884: of equilibria are labeled: mode~I are for aligned apsides ($\Delta\varpi=0$),

885: mode~II are for anti-aligned apsides ($\Delta\varpi=\pi$),  \UE{} means unstable

886: equilibria.  The thin half-ellipse like curves are for energy levels computed

887: for the classic model (red) and generalized model (green). More details can be

888: found in the text. 

889: }

890: \label{fig:fig2}

891: \end{figure*}

892: \else 

893: \begin{figure*}

894: \centerline{

895: \hbox{\includegraphics[width=8.8cm]{fig2a.eps}

896:       \hspace*{0cm}   

897:       \includegraphics[width=8.8cm]{fig2b.eps}

898: }

899: }

900: \caption{

901: The representative plane $(e_1 \cos{\Delta{\varpi}}, e_2)$ for the two-planet

902: system. The elements are $m_1=1.359~\mbox{m}_{\idm{J}}$,

903: $m_2=0.453~\mbox{m}_{\idm{J}}$, $a_1=0.1$~au, $a_2=1.0$~au. Stellar mass is

904: 1~$M_{\sun}$, rotational period is $T_{\idm{rot}}=30$~days, stellar radius is

905: 1~$R_{\sun}$, $k_L=0.02$. The thick curves are for the stationary solutions: red

906: curves are for stable equilibria in the classic model, blue and violet curves

907: are for stable and unstable equilibria of the generalized model, respectively.  

908: {\em The left panel}:  Color contours are for the ratio of the  apsidal

909: frequency induced by the general relativity and quadrupole moment  to the 

910: apsidal frequency caused by mutual interactions between planets, $\kappa$.

911: \corr{The scale of color code is limited by $\kappa=1$, meaning that the

912: corrections become more important than the point mass Newtonian gravity. In

913: general, the yellow colour may refer to $\kappa>1$.} Energy levels of the

914: generalized model are marked with thin lines.  {\em The right panel}: A few

915: particular energy levels of the generalized model are labeled with

916: \textbf{a}---\textbf{f}, accordingly.  See also phase diagrams for these energy

917: levels  which are shown in Fig.~\ref{fig:fig3} and Fig.~\ref{fig:fig4}. Curves

918: of equilibria are labeled: mode~I are for aligned apsides ($\Delta\varpi=0$),

919: mode~II are for anti-aligned apsides ($\Delta\varpi=\pi$),  \UE{} means unstable

920: equilibria.  The thin half-ellipse like curves are for energy levels computed

921: for the classic model (red) and generalized model (green). More details can be

922: found in the text. 

923: }

924: \label{fig:fig2}

925: \end{figure*} 

926: \fi

927: %

928: The relative magnitude of the corrections to the secular Hamiltonian

929: is represented by contour levels of the following coefficient:

930: \begin{equation}

931: \label{eq:kappa}

932: \kappa(e_1 \cos \Delta\varpi,e_2)

933: = \left|\frac{\partial\left<\Hrel\right>/\partial{G_1} +

934: \partial\left<\Hflat\right>/\partial{G_1}}

935: {\partial\left<\Hmut\right>/\partial{G_1}}\right|,

936: \end{equation}

937: i.e., the ratio of the apsidal frequency of the inner pericenter induced by

938: $\left<\Hrel + \Hflat\right>$ relative to the  ``natural''  apsidal frequency in

939: the point mass Newtonian model. Regions of the \RP{}-plane  where $\kappa>0.1$, are

940: color-coded, according to the levels of constant $\kappa$. Yellow color (light

941: gray in the printed paper) encodes $\kappa \ge 1$, meaning that

942:  the apsidal frequency of

943: the innermost pericenter forced by  {\em perturbing} GR+QM corrections is

944: larger than the relative pericenter frequency $\dot{\Delta\varpi}$ 

945: caused by the secular NG interactions between  planets. 

946: 

947: The thick curves defined through:

948: \begin{equation}

949: \frac{\partial{\Hsec}}{\partial{G_1}} = 0,

950: \end{equation}

951: can be attributed to the stationary solutions of the reduced, two-dimensional system

952: with $(G_1,\Delta\varpi)$-variables, because  coordinates of all points of 

953: these curves must also satisfy

954: \begin{equation}

955: \frac{\partial{\Hsec}}{\partial{\Delta{\varpi}}} = 0,

956: \end{equation}

957: according to the definition of the \RP{}-plane.  Such solutions are periodic

958: orbits of the full secular system (Eqs.~\ref{eqtot1}--\ref{eqtot2}). We  examine

959: the Lyapunov stability of these equilibria with the help of the Lyapunov

960: theorem, adopting  $\Hsec$ as the Lyapunov function in the cases when they 

961: correspond to its maximum (or a minimum) in the reduced two-dimensional phase

962: space.  \corr{Because the investigated dynamical system has one degree of

963: freedom, the stable or unstable equilibria can be easily identified as extrema

964: or saddles of the secular Hamiltonian in the representative plane,

965: respectively}. 

966: 

967: The positions of equilibria [or stationary modes \citep{Michtchenko2004}] help

968: us to distinguish between different types (families) of orbits characterized by

969: librations of $\Delta\varpi$ around a particular value (libration center).  The

970: red, thick curves drawn in both panels of Fig.~\ref{fig:fig2} are for stationary

971: modes in the classic model \citep{Michtchenko2004}, while the equilibria in the

972: generalized model are drawn with thick, blue curves. In the \RPp{} half-plane,

973: these stationary solutions are classified by \cite{Michtchenko2004} as mode~I

974: equilibria, and they are characterized  by librations of $\Delta\varpi$

975: around~$0$ in the neighboring trajectories.  In the \RPm{} half-plane, we can

976: find also  Lyapunov stable mode~II solutions related to librations of

977: $\Delta\varpi$ around~$\pi$ in close trajectories.  Some parts of the equilibria

978: curves are marked with   violet color (for the generalized model). These points

979: denote unstable equilibria (\UE{}) accompanied by {\em the true secular

980: resonance} (the \TSR{} from hereafter). Such unstable equilibria in the

981: \RPp{}-plane  are discovered  by \cite{Michtchenko2004} in the secular  classic

982: coplanar model of 2-planets. Obviously, mode~I and mode~II solutions known as

983: generic features of the classic model, exist also in the generalized problem.

984: However, their  positions in the phase space may be heavily affected by

985: apparently subtle GR and QM perturbations.  

986: 

987: Note that along the red thick curves representing equilibria in the classic

988: model, $\kappa$ is undefined. 

989: 

990: Now, let us examine the right-hand panel of Fig.~\ref{fig:fig2}. In this plot,

991: besides levels of the secular energy of the classic model (red, thin curves), we

992: also plot  such levels for the generalized problem (green, thin curves). We can

993: observe a significant discrepancy between the shapes of contour levels of both

994: Hamiltonians. 

995: 

996: In this plot we mark also a few specific levels of $\Hsec$: 

997: $\mathcal{E}_{\idm{\bf a}} = -2.247$,

998: $\mathcal{E}_{\idm{\bf b}} = -2.24955$,

999: $\mathcal{E}_{\idm{\bf c}} = -2.25$,

1000: $\mathcal{E}_{\idm{\bf d}} = -2.2525$,

1001: $\mathcal{E}_{\idm{\bf e}} = -2.256$,

1002: $\mathcal{E}_{\idm{\bf f}} = -2.26$,

1003: respectively; here, we skipped all constant terms (the Keplerian part, indirect

1004: term, constant relativistic term) in the full secular Hamiltonian, and  the

1005: energy values are given in terms of $10^{-5}~M_{\sun}\mbox{au}^2\mbox{yr}^{-2}$ 

1006: when 1~$M_{\sun}$, 1~au, and 1~sideral year are taken as units of mass,

1007: distance, and time, respectively (then the Gauss constant $k=2\pi$).  Because

1008: the motion of the secular system  must be confined to fixed energy level, points

1009: at which a particular energy level crosses the curve representing stationary

1010: solutions, tell us on distinct equilibria and their bifurcations. For instance,

1011: following energy level labeled with {\bf a} in the right-hand panel of Fig.~2,

1012: we see that it  crosses the equilibria curves at two points. The first one,

1013: found in the \RPp{} half-plane, corresponds to mode~I solution.  The second

1014: cross-point is found in the \RPm{} half-plane, and it marks the mode~II

1015: equilibrium.

1016: 

1017: To estimate the precision of the analytical theory, the energy levels  in

1018: Fig.~\ref{fig:fig2} are calculated with exact semi-analytical averaging

1019: \citep{Michtchenko2004} that makes use of precise adaptive Gauss-Legendre

1020: quadratures \citep{Migaszewski2008b}.  For a comparison, the stationary modes

1021: are computed with both methods: we recall that thick curves represent equilibria

1022: calculated with the semi-numerical averaging while the thin curves (over-plotted

1023: on them) are for  stationary solutions calculated with the help of the

1024: analytical theory outlined in Sect.~2 (the right-hand panel of

1025: Fig.~\ref{fig:fig2}). The results of these methods are in excellent accord. The

1026: most significant  discrepancies between the results appear in the \RPp{}

1027: half-plane (for $\Delta\varpi=0$), in the regime of large eccentricity $e_2$. In

1028: such a case, the secular series representing $\Hmut$ diverge over the {\em

1029: anti-collision} line (marked with red, straight line). This problem is discussed

1030: in detail in \citep{Migaszewski2008a}.  Obviously, over this anti-collision

1031: line, $r_2>r_1$ at some parts of orbits, breaking the underlying assumption that

1032: we require to expand the inverse of the mutual distance in convergent series.

1033: Moreover, we demonstrate that the analytical theory reproduces the dynamics of 

1034: hierarchical configurations up to very large $e_2$. Another, direct test of the

1035: precision of the analytic theory is given in Sect.~3.4.

1036: %

1037: %-------------------------------------------------------------------------------

1038: \subsection{Phase diagrams and the non-linear secular resonance}

1039: %-------------------------------------------------------------------------------

1040: %

1041: To investigate more closely the secular dynamics related to the new mode~II and

1042: \UE{} solutions, and to understand the structure of the phase space in more

1043: details, we compute a number of {\em phase plots} (or {\em phase

1044: diagrams}). The secular energy is kept constant and we draw the phase

1045: trajectories in the ($e_1 \cos{\Delta{\varpi}}, e_1 \sin{\Delta{\varpi}}$)--, as

1046: well as  ($e_2 \cos{\Delta{\varpi}}, e_2 \sin{\Delta{\varpi}}$)--planes,

1047: choosing the initial conditions along the fixed energy level.  The phase plots

1048: are computed with the help of the analytic theory.  For a reference, we recall

1049: the right-hand panel of Fig.~\ref{fig:fig2} which illustrates the  \RP{}-plane

1050: of $(e_1 \cos{\Delta{\varpi}}, e_2)$.   We recall that the \UE{} of the

1051: generalized model are marked with violet curves. 

1052: 

1053: 

1054: The phase diagrams are illustrated in Fig.~\ref{fig:fig3}. The top row is for

1055: the energy level of $\mathcal{E}_{\idm{a}} = -2.247\times10^{-5}$  which is

1056: labeled with {\bf a} in the right-hand panel of Fig.~\ref{fig:fig2}.  This

1057: energy level crosses the equilibria curves in two points which  can be

1058: recognized in the phase diagrams as libration centers of $\Delta\varpi=0$

1059: (labeled with mode~I), and of $\Delta\varpi=\pi$  (labeled with mode~II),

1060: respectively. Overall, the phase diagrams look like qualitatively the same as in

1061: the classic model. Both libration modes are separated from the circulation zone

1062: of $\Delta\varpi$ by {\em false} separatrices

1063: \citep{Pauwels1983,Michtchenko2004}, i.e., the transition between each mode and

1064: the circulation  of $\Delta\varpi$ {\em does not} involve solutions with

1065: infinite period. \corr{ This is illustrated in smaller, bottom plots

1066: accompanying each phase diagram, which show the secular, fundamental frequency

1067: $g$ of the mean system for initial conditions lying on the $x$ axis of the

1068: respective phase diagram.   This frequency has been determined through 

1069: solutions of the secular equations of motion. Clearly, when the true separatrix

1070: crosses the $e_{1,2}\cos\Delta\varpi$-axis, $g$ decreases to 0. Crossings of

1071: false separatrices do not lead to any discontinuity of smooth plots of $g$. }

1072: %

1073: \ifpdf

1074: \begin{figure*}

1075: \centerline{

1076: \vbox{

1077: \hbox{\includegraphics[width=68mm]{fig3a.pdf}

1078: \hspace*{6mm}

1079:       \includegraphics[width=68mm]{fig3b.pdf}}    

1080: \hbox{\includegraphics[width=68mm]{fig3c.pdf}

1081: \hspace*{6mm}

1082:       \includegraphics[width=68mm]{fig3d.pdf}}

1083: }

1084: }      

1085: \caption{

1086: Phase diagrams computed for two-planet secular system described with the same

1087: parameters as used for the construction of Fig. \ref{fig:fig2}. \corr{ Each

1088: phase diagram is accompanied by a smaller  plot of the 

1089: fundamental frequency $g$ of

1090: the secular solutions  calculated for initial conditions lying on  the $y \equiv

1091: e_{1,2}\cos\Delta\varpi$-axis. } Panels in the top row are for the secular

1092: energy $E_{\bf a} = -2.247\times10^{-5}$, panels in the bottom row are for

1093: $E_{\bf c} = -2.25\times10^{-5}$. These levels are labeled in Fig.

1094: \ref{fig:fig2} with ({\bf a}) and ({\bf c}), respectively.  The left-hand panels

1095: are for the ($e_1 \cos{\Delta{\varpi}}, e_1 \sin{\Delta{\varpi}}$)-plane, the

1096: right-hand panels are  for ($e_2 \cos{\Delta{\varpi}}, e_2

1097: \sin{\Delta{\varpi}}$)-plane. Shaded regions mark different  libration zones

1098: around mode~I ($\Delta\varpi=0$) and mode~II ($\Delta\varpi=\pi$), respectively.

1099: \UE{} marks  unstable equilibria  accompanied by libration zones around

1100: $\Delta\varpi=\pi$ and centered at the \TSR{} solutions.  \corr{The true

1101: separatrices are indicated by $g \rightarrow 0$}. See the text for more details.

1102: }

1103: \label{fig:fig3}

1104: \end{figure*}

1105: \else

1106: \begin{figure*}

1107: \centerline{

1108: \vbox{

1109: \hbox{\includegraphics[width=68mm]{fig3a.eps}

1110: \hspace*{6mm}

1111:       \includegraphics[width=68mm]{fig3b.eps}}    

1112: \hbox{\includegraphics[width=68mm]{fig3c.eps}

1113: \hspace*{6mm}

1114:       \includegraphics[width=68mm]{fig3d.eps}}

1115: }

1116: }      

1117: \caption{

1118: Phase diagrams computed for two-planet secular system described with the same

1119: parameters as used for the construction of Fig. \ref{fig:fig2}. \corr{ Each

1120: phase diagram is accompanied by a smaller  plot of the 

1121: fundamental frequency $g$ of

1122: the secular solutions  calculated for initial conditions lying on  the $y \equiv

1123: e_{1,2}\cos\Delta\varpi$-axis. } Panels in the top row are for the secular

1124: energy $E_{\bf a} = -2.247\times10^{-5}$, panels in the bottom row are for

1125: $E_{\bf c} = -2.25\times10^{-5}$. These levels are labeled in Fig.

1126: \ref{fig:fig2} with ({\bf a}) and ({\bf c}), respectively.  The left-hand panels

1127: are for the ($e_1 \cos{\Delta{\varpi}}, e_1 \sin{\Delta{\varpi}}$)-plane, the

1128: right-hand panels are  for ($e_2 \cos{\Delta{\varpi}}, e_2

1129: \sin{\Delta{\varpi}}$)-plane. Shaded regions mark different  libration zones

1130: around mode~I ($\Delta\varpi=0$) and mode~II ($\Delta\varpi=\pi$), respectively.

1131: \UE{} marks  unstable equilibria  accompanied by libration zones around

1132: $\Delta\varpi=\pi$ and centered at the \TSR{} solutions.  \corr{The true

1133: separatrices are indicated by $g \rightarrow 0$}. See the text for more details.

1134: }

1135: \label{fig:fig3}

1136: \end{figure*}

1137: \fi

1138: %

1139: The phase diagrams are much more complicated for energy levels labeled with {\bf

1140: b}--{\bf f}, which cross the equilibria curves at more than two points. We

1141: analyze in detail the energy level ({\bf c})  which intersects the curve of

1142: stationary solutions in {\em four} points. These points can be recognized in the

1143: phase diagrams shown in the bottom panels of Fig.~\ref{fig:fig3}. We can

1144: identify them easily in the $x$-axis of these diagrams  (because,  such points

1145: of the \RP{}-plane have $y \equiv  e_{1,2}\sin \Delta\varpi=0$). Starting at the

1146: \RPp{}-plane and following the energy level counter-clockwise, we have a stable

1147: mode~I equilibrium surrounded by large zone of librations of $\Delta\varpi$

1148: around~$0$ which corresponds to single crossing point of the energy level and

1149: the equilibria curve in the \RPp{}-plane. In the \RPm{}-plane, we have {\em

1150: three} such points, two of them are Lyapunov stable, and one point in the middle

1151: is an unstable equilibrium. This part of the phase space, as seen in

1152: Fig.~\ref{fig:fig3}, encompass figure-eight shaded area, involving two islands

1153: of librations (\TSR{}s, or elliptic points) around $\Delta\varpi=\pi$,  and the

1154: hyperbolic \UE{} point lying in the middle between them. Both libration centers

1155: are characterized by $\Delta\varpi=\pi$. Because they are related to stable

1156: equilibria separated by hyperbolic structure of the \UE{}, two parts of  the

1157: phase curve surrounding the islands of the \TSR{}s and that meet in the \UE{},

1158: must form a real separatrix. The whole structure may be  still surrounded by a

1159: zone of librations of $\Delta\varpi$ around $\pi$. It is shaded in light-gray.

1160: Let us note, that this mode~II libration area is confined to the \RPm{}-plane

1161: (hence, in this particular case, angle $\Delta\varpi$  does not pass through

1162: $0$).

1163: 

1164: A sequence of ($e_2 \cos{\Delta{\varpi}}, e_2 \sin{\Delta{\varpi}}$)-diagrams

1165: for the secular energy levels {\bf a}--{\bf f} are shown in Fig.~\ref{fig:fig4}.

1166: Looking at these levels plotted in the \RP{}-plane, we can now follow a

1167: development of dynamical structures related to the different modes of motion. In

1168: particular, phase diagrams Fig.~\ref{fig:fig4}b and Fig.~\ref{fig:fig4}d reveal

1169: bifurcations of mode~II which emerge the \UE{} and  \TSR{} solutions. Clearly,

1170: the bifurcations may  be identified with points at which the given energy level

1171: is tangent to the equilibria curves in the \RP{}-plane. 

1172: 

1173: The phase diagrams assure us that the secular dynamics of the generalized model

1174: can be much more complex and rich due to the GR and QM corrections to the NG

1175: Hamiltonian than the secular dynamics in the classic model.

1176: %

1177: %-------------------------------------------------------------------------------

1178: \subsection{Numerical test of the analytic secular theory}

1179: %-------------------------------------------------------------------------------

1180: %

1181: Finally, we illustrate limitations of  the secular theory and we compare its

1182: results with the outcome of the direct numerical integrations. This comparison

1183: is also directly related to the dynamical stability of the planetary system.

1184: 

1185: First, we constructed the numerical model of the generalized system

1186: independently on the analytical model. We wrote the equations of motion with

1187: respect to the Jacobi reference frame, using formulation of \cite{Mardling2002} 

1188: who call it {\em the direct code}. In this code, the GR acceleration is modeled 

1189: with the PPN formulae given in \citep{Kidder1995}. In that way we have a

1190: possibility to check the analytic theory in completely independent way, which

1191: also prevents copying logical errors which could be done during the averaging.

1192: After some experiments, we also found that the choice of the reference frame

1193: (e.g., related to  Jacobi, Poincar\'e, or classic-astrocentric coordinates) is

1194: in fact irrelevant for the results of this test. We also do not

1195: account for the difference between the {\em osculating} and the {\em mean}

1196: elements. 

1197: 

1198: Using the direct code, we integrated numerically a few phase diagrams 

1199: for  nominally {\em

1200: the same initial conditions and parameters} used to draw plots in

1201: Fig.~\ref{fig:fig4}a--f with the help of the analytic, secular model. We computed osculating elements related

1202: to the Jacobi reference  frame over a few secular cycles. The numerically

1203: derived phase curves are drawn with black, filled circles in  respective panels

1204: of Fig.~\ref{fig:fig4}. The solutions obtained with the analytic theory are

1205: over-plotted on these numerical solutions with thiner, green curves. Subsequent

1206: five panels  of Fig.~\ref{fig:fig4}a--e reveal that the agreement of both sets

1207: of solutions is excellent. The analytic theory reproduces qualitative features

1208: seen in the phase plots, and their structure with great accuracy. We find that

1209: both solutions coincide even in the regime of large eccentricities.

1210: Remarkably, the direct code integrations last over CPU time which is by a few

1211: orders of magnitude longer than the calculations carried out with the help of

1212: the analytical theory.

1213: 

1214: However, in the last panel of Fig.~\ref{fig:fig4}f we can observe significant

1215: deviations of the analytic solutions from the numerical theory, particularly in

1216: the outer parts of the phase diagram.  In fact, in this case $e_2$ is so large that

1217: the assumptions of the secular theory are broken. After examining the

1218: \RP{}-plane (Fig.~\ref{fig:fig2}), we can see that the energy level

1219: corresponding to the last panel passes close to the collision line. To 

1220: illustrate the real border of the dynamical stability, we examined the dynamical

1221: character of solutions in the \RP{}-plane with the help of the Spectral Number

1222: technique \citep{Michtchenko2001}. This simple FFT-based algorithm makes it

1223: possible to distinguish between chaotic and regular solutions. The dynamical

1224: maps shown in Fig.~\ref{fig:fig5} are constructed by counting the number of

1225: frequencies in the FFT-spectrum of the time series, $\{ \sigma(t) = a(t) \exp

1226: \mbox{i}\lambda_i(t) \}$, where $a_i(t)$ and $\lambda_i(t)$ are temporal {\em

1227: canonical} semi-major axes and mean longitude of each planet. The number of

1228: peaks (the Spectral Number, SN from hereafter)  in the spectrum over some noise

1229: level tells us on the character of orbit. Orbits with large SN (grater than

1230: 1000) are very chaotic, while the SN $\sim 1$ means small number of frequencies

1231: and a regular, quasi-periodic phase trajectory. Each point in the dynamical maps

1232: represents a phase trajectory  that was integrated over $\sim 10^4~P_{\idm{2}}$.

1233: Although such time span is relevant for the short-term dynamics only, 

1234: calculations took a very long CPU time (a few days on 24 AMD-CPU cores). The

1235: results are shown in two panels of Fig.~\ref{fig:fig5}. The left panel is for

1236: the classic model (only NG interactions are included), while the right panel is

1237: constructed for the generalized model.

1238: 

1239: Let us analyze the left-hand panel of Fig.~\ref{fig:fig5}

1240:  for the classic model. Solutions, which appear

1241: strongly chaotic  are marked with colors (darker point means larger SN and more

1242: chaotic system).  Clearly, the border of stable motions is irregular and is

1243: shifted towards small $e_2$ by $0.1$--$0.2$ with respect to the formal,

1244: geometrical collision line bordering the triangular region (it is drawn in both

1245: panels of Fig.~\ref{fig:fig5}). The thick red curves mark the equilibria of mode

1246: I and mode~II, respectively. Shaded regions are for the initial conditions in

1247: the \RP{}-plane corresponding to orbital configurations with librating

1248: $\Delta\varpi$.  In the right-hand panel of Fig.~\ref{fig:fig5}, we show the

1249: equilibria curves of  the generalized model and  libration zones of

1250: $\Delta\varpi$ associated with these equilibria. We mark again the SN signature

1251: of  the short-term dynamics. We note that the border of stability is

1252: quite different from that ones of the classic model (compare with the

1253: left-hand panel of Fig.~\ref{fig:fig5}).  This is a very clear example showing

1254: that apparently subtle GR+QM effects may affect the {\em short-term} stability

1255: of the system in a significant way. 

1256: 

1257: The dynamical map for the generalized model 

1258: (Fig.~\ref{fig:fig5}, the right hand panel) helps us to identify the source of

1259: unstable behavior seen in Fig.~\ref{fig:fig4}f,  revealing that some initial

1260: conditions lead to erratic and irregular behavior. In the dynamical map, we mark

1261: two levels of $\Hsec$ corresponding to phase diagrams drawn in

1262: Fig.~\ref{fig:fig4}e,f, respectively. The energy level ({\bf e}) lies entirely

1263: in the regular region, in spite that $e_2$ may reach values as large as 0.8. The

1264: neighboring level ({\bf f}) may touch the unstable zone, and that is why the

1265: orbital evolution at this energy level  may become very unstable.  Because the

1266: motion must be confined to the fixed energy level, due to the secular evolution,

1267: some initial conditions may be transported to the chaotic zone  (by excitation

1268: of the eccentricity) during  roughly a half of the secular period. Then the

1269: short-term, strong chaos can destabilize the system immediately.  Following the

1270: fixed energy curve, we can also identify three islands of stable motions. The

1271: first one lies in the right half-plane of the phase diagram and is associated

1272: with librations of $\Delta\varpi$ around $0$, see the neighborhood of the

1273: corresponding  cross point in the \RPp-plane, Fig.~\ref{fig:fig2}.  In the

1274: \RPm-plane, we can find corresponding unstable equilibrium and two stable

1275: solutions with associated libration islands shown in Fig.~\ref{fig:fig4}f (see

1276: the left-hand half-plane of the phase diagram) around $(0,-0.1)$ and

1277: $(0,-0.78)$, respectively. 

1278: %

1279: \ifpdf

1280: \begin{figure*}

1281: \centerline{

1282: \vbox{

1283: \hbox{\includegraphics[width=58mm]{fig4a.png}

1284:       \includegraphics[width=58mm]{fig4b.png}

1285:       \includegraphics[width=58mm]{fig4c.png}

1286:       }

1287: \hbox{\includegraphics[width=58mm]{fig4d.png}

1288:       \includegraphics[width=58mm]{fig4e.png}

1289:       \includegraphics[width=58mm]{fig4f.png}

1290:       }

1291: }

1292: }      

1293: \caption{

1294: Phase diagrams in the $(e_2\cos{\Delta{\varpi}}, e_2

1295: \sin{\Delta{\varpi}})$-plane drawn  to illustrate a comparison of the secular

1296: evolution of the mean system (green, thin curves) with the numerical solutions

1297: of full,  unaveraged system (larger, filled black circles). The orbital

1298: parameters are the same as in Fig. \ref{fig:fig2}. Subsequent panels labeled

1299: with \textbf{a}-\textbf{f}  correspond to  relevant energy levels marked and

1300: labeled in Fig.~\ref{fig:fig2}, accordingly.

1301: }

1302: \label{fig:fig4}

1303: \end{figure*}

1304: \else

1305: \begin{figure*}

1306: \centerline{

1307: \vbox{

1308: \hbox{\includegraphics[width=58mm]{fig4a.eps}

1309:       \includegraphics[width=58mm]{fig4b.eps}

1310:       \includegraphics[width=58mm]{fig4c.eps}

1311:       }

1312: \hbox{\includegraphics[width=58mm]{fig4d.eps}

1313:       \includegraphics[width=58mm]{fig4e.eps}

1314:       \includegraphics[width=58mm]{fig4f.eps}

1315:       }

1316: }

1317: }      

1318: \caption{

1319: Phase diagrams in the $(e_2\cos{\Delta{\varpi}}, e_2

1320: \sin{\Delta{\varpi}})$-plane drawn  to illustrate a comparison of the secular

1321: evolution of the mean system (green, thin curves) with the numerical solutions

1322: of full,  unaveraged system (larger, filled black circles). The orbital

1323: parameters are the same as in Fig. \ref{fig:fig2}. Subsequent panels labeled

1324: with \textbf{a}-\textbf{f}  correspond to  relevant energy levels marked and

1325: labeled in Fig.~\ref{fig:fig2}, accordingly.

1326: }

1327: \label{fig:fig4}

1328: \end{figure*}

1329: \fi

1330: %

1331: \ifpdf

1332: \begin{figure*}

1333: \centerline{

1334: \hbox{\includegraphics[width=8.7cm]{fig5a.pdf}

1335:    \hspace*{0.4cm}   

1336:    \includegraphics[width=8.7cm]{fig5b.pdf}

1337: }

1338: }

1339: \caption{

1340: The \RP{}-planes for the classic model (the left-hand panel) and for the

1341: generalized model (the right-hand panel) for the two-planet system analyzed in

1342: Fig.~\ref{fig:fig2}. The thick, red curves mark stationary modes in  the classic

1343: model, the thick blue curves are for the equilibria in the generalized model.

1344: Shaded areas indicate zones of $\Delta\varpi$ librations. The thick red lines

1345: are for the collision line of orbits. Colors code $\log$~SN of the {\em outer

1346: orbit}, characterizing  solutions derived numerically with the help of the

1347: direct code. Black points are for strongly chaotic solutions with the

1348: $\log\mbox{SN} \sim 3$, white color  is for regular solutions with

1349: $\log\mbox{SN}\sim 0$, intermediate values are marked with the color scale above

1350: the panels, accordingly (see the text for more details).

1351: }

1352: \label{fig:fig5}

1353: \end{figure*}

1354: \else

1355: \begin{figure*}

1356: \centerline{

1357: \hbox{\includegraphics[width=8.7cm]{fig5a.eps}

1358:    \hspace*{0.4cm}   

1359:    \includegraphics[width=8.7cm]{fig5b.eps}

1360: }

1361: }

1362: \caption{

1363: The \RP{}-planes for the classic model (the left-hand panel) and for the

1364: generalized model (the right-hand panel) for the two-planet system analyzed in

1365: Fig.~\ref{fig:fig2}. The thick, red curves mark stationary modes in  the classic

1366: model, the thick blue curves are for the equilibria in the generalized model.

1367: Shaded areas indicate zones of $\Delta\varpi$ librations. The thick red lines

1368: are for the collision line of orbits. Colors code $\log$~SN of the {\em outer

1369: orbit}, characterizing  solutions derived numerically with the help of the

1370: direct code. Black points are for strongly chaotic solutions with the

1371: $\log\mbox{SN} \sim 3$, white color  is for regular solutions with

1372: $\log\mbox{SN}\sim 0$, intermediate values are marked with the color scale above

1373: the panels, accordingly (see the text for more details).

1374: }

1375: \label{fig:fig5}

1376: \end{figure*} 

1377: \fi

1378: %

1379: \corr{

1380: Finally, to illustrate the development of the secular instability, we solved the

1381: equations of motion of the full system, starting very close to the UE lying on

1382: the energy level between levels {\bf e} and {\bf f}, as marked in

1383: Fig.~\ref{fig:fig2}, and Fig.~\ref{fig:fig4}e,f. The solution is illustrated in

1384: the phase diagram in the left-hand panel of Fig.~\ref{fig:fig6}. For a reference, the

1385: numerical solution is over-plotted on the analytically derived separatrices of

1386: the UE. The corresponding time evolution of the orbital elements are illustrated

1387: in the right-hand panels of Fig.~\ref{fig:fig6}. Clearly, during quite a long

1388: time the full system stays close to the UE, but after $\sim 10^5$~yrs it follows

1389: a trajectory  close to the inner separatrix, and finally begins to move close to

1390: the outer separatrix, approaching large $e_2$. During this evolution, we observe

1391: not only very irregular behavior of $a_2$ and both  eccentricities, but also

1392: $\Delta\varpi$ changing from large amplitude  librations around $0$ to

1393: circulations. Although the configuration seems bounded during many secular

1394: periods, such behavior may be classified as strongly chaotic.

1395: }

1396: 

1397: %

1398: \ifpdf

1399: \begin{figure*}

1400: \centerline{

1401: \vbox{

1402: \hbox{

1403: \includegraphics[height=61mm]{fig6a.png}\hskip5mm

1404: \includegraphics[height=61mm]{fig6b.png}

1405: }

1406: }

1407: }

1408: \caption{

1409: The phase diagram in the $(e_2\cos{\Delta{\varpi}}, e_2

1410: \sin{\Delta{\varpi}})$-plane (the left-hand panel) drawn to illustrate a development

1411: of the secularly unstable behavior of the full system.  The numerical solution

1412: of the full system (thin, black curve) is over-plotted on the analytical,

1413: secular solution (gray, thicker curve). The right-hand panels illustrate the

1414: relative changes of semi-major axes (top panel), eccentricities (middle panel)

1415: [grey curves are for outer orbit, black curves are for the inner orbit], and the

1416: apsidal angle $\Delta\varpi$ (bottom panel).  The initial orbital parameters are

1417: the same as in Fig. \ref{fig:fig2}.  The energy level of this solution lies

1418: between levels \textbf{e}-\textbf{f} marked in Fig.~\ref{fig:fig2}, accordingly,

1419: compare it also with phase diagrams Fig.~\ref{fig:fig4}e,f.

1420: }

1421: \label{fig:fig6}

1422: \end{figure*}

1423: \else

1424: \begin{figure*}

1425: \centerline{

1426: \vbox{

1427: \hbox{

1428: \includegraphics[height=61mm]{fig6a.eps}\hskip5mm

1429: \includegraphics[height=61mm]{fig6b.eps}

1430: }

1431: }

1432: }

1433: \caption{

1434: The phase diagram in the $(e_2\cos{\Delta{\varpi}}, e_2

1435: \sin{\Delta{\varpi}})$-plane (the left-hand panel) drawn to illustrate a development

1436: of the secularly unstable behavior of the full system.  The numerical solution

1437: of the full system (thin, black curve) is over-plotted on the analytical,

1438: secular solution (gray, thicker curve). The right-hand panels illustrate the

1439: relative changes of semi-major axes (top panel), eccentricities (middle panel)

1440: [grey curves are for outer orbit, black curves are for the inner orbit], and the

1441: apsidal angle $\Delta\varpi$ (bottom panel).  The initial orbital parameters are

1442: the same as in Fig. \ref{fig:fig2}.  The energy level of this solution lies

1443: between levels \textbf{e}-\textbf{f} marked in Fig.~\ref{fig:fig2}, accordingly,

1444: compare it also with phase diagrams Fig.~\ref{fig:fig4}e,f.

1445: }

1446: \label{fig:fig6}

1447: \end{figure*}

1448: \fi

1449: %

1450: %-------------------------------------------------------------------------------

1451: \section{Parametric survey of two-planet systems}

1452: %-------------------------------------------------------------------------------

1453: %

1454: The characterization of the phase space with the help of the representative

1455: plane  can be  very useful to  conduct a survey of the basic features of the

1456: secular dynamics. In particular, we want to understand how it depends on the 

1457: physical and orbital parameters governing the magnitude of the GR and QM

1458: interactions. In subsequent diagrams of the \RP{}-plane, we will always mark the

1459: collision and anti-collision lines. In this way, we can determine the border of

1460: validity of the analytic approach. Yet to derive the stationary modes possibly

1461: exactly, in the whole permitted range of eccentricity,  we compute their

1462: locations with the help of the semi-analytical averaging algorithm.

1463: %

1464: %-------------------------------------------------------------------------------

1465: \subsection{Dependence of the secular dynamics on the masses}

1466: \label{mass_dependence}

1467: %-------------------------------------------------------------------------------

1468: %

1469: The results of the survey of the secular dynamics of two-planet systems,

1470: including GR and QM interactions, for varied planetary masses, are illustrated

1471: in   Fig.~\ref{fig:fig7}. 

1472: %

1473: \ifpdf

1474: \begin{figure*}

1475: \centerline{

1476: \vbox{

1477: \hbox{\includegraphics[width=5.9cm]{fig7a.png}

1478:       \includegraphics[width=5.9cm]{fig7b.png}

1479:       \includegraphics[width=5.9cm]{fig7c.png}}

1480: \hbox{\includegraphics[width=5.9cm]{fig7d.png}

1481:       \includegraphics[width=5.9cm]{fig7e.png}

1482:       \includegraphics[width=5.9cm]{fig7f.png}}

1483: }

1484: }      

1485: \caption{

1486: The representative energy planes $(e_1 \cos{\Delta\varpi}, e_2)$ for

1487: $\Delta\varpi=0$ (the right half-planes,\RPp{}) and $\Delta\varpi=\pi$ (the left

1488: half-planes, \RPm{}). 

1489: Color contours are for the ratio of the  apsidal frequency induced by

1490: the general relativity and quadrupole moment  to the  apsidal frequency caused by mutual

1491: interactions between planets. 

1492: Thick curves mark positions of stationary solutions. Red

1493: lines are for stable equilibria in the classic model. Blue and violet 

1494: curves mark

1495: the positions of stable and unstable equilibria of the generalized model. The

1496: thin lines are for the secular energy levels of the generalized model. The thick, skew lines are for the

1497: collision lines defined with $a_1 (1 \pm e_1) = a_2 (1 - e_2)$. Parameters of

1498: these systems are as follows: $m_0 = 1 \mbox{M}_{\sun}$, $a_1=0.1 \mbox{au}$,

1499: $a_2=1 \mbox{au}$, $R_0=1 \mbox{R}_{\sun}$,  $T_{\idm{rot}}=30$~days,

1500: $k_L=0.02$.  Each panel was calculated for varied planetary masses, under the

1501: condition of constant mass ratio $m_1/m_2 = 3$. The mass of the inner planet is

1502: written in the top left corner of each panel.

1503: }

1504: \label{fig:fig7}

1505: \end{figure*}

1506: \else

1507: \begin{figure*}

1508: \centerline{

1509: \vbox{

1510: \hbox{\includegraphics[width=5.9cm]{fig7a.eps}

1511:       \includegraphics[width=5.9cm]{fig7b.eps}

1512:       \includegraphics[width=5.9cm]{fig7c.eps}}

1513: \hbox{\includegraphics[width=5.9cm]{fig7d.eps}

1514:       \includegraphics[width=5.9cm]{fig7e.eps}

1515:       \includegraphics[width=5.9cm]{fig7f.eps}}

1516: }

1517: }      

1518: \caption{

1519: The representative energy planes $(e_1 \cos{\Delta\varpi}, e_2)$ for

1520: $\Delta\varpi=0$ (the right half-planes,\RPp{}) and $\Delta\varpi=\pi$ (the left

1521: half-planes, \RPm{}). 

1522: Color contours are for the ratio of the  apsidal frequency induced by

1523: the general relativity and quadrupole moment  to the  apsidal frequency caused by mutual

1524: interactions between planets. 

1525: Thick curves mark positions of stationary solutions. Red

1526: lines are for stable equilibria in the classic model. Blue and violet 

1527: curves mark

1528: the positions of stable and unstable equilibria of the generalized model. The

1529: thin lines are for the secular energy levels of the generalized model. The thick, skew lines are for the

1530: collision lines defined with $a_1 (1 \pm e_1) = a_2 (1 - e_2)$. Parameters of

1531: these systems are as follows: $m_0 = 1 \mbox{M}_{\sun}$, $a_1=0.1 \mbox{au}$,

1532: $a_2=1 \mbox{au}$, $R_0=1 \mbox{R}_{\sun}$,  $T_{\idm{rot}}=30$~days,

1533: $k_L=0.02$.  Each panel was calculated for varied planetary masses, under the

1534: condition of constant mass ratio $m_1/m_2 = 3$. The mass of the inner planet is

1535: written in the top left corner of each panel.

1536: }

1537: \label{fig:fig7}

1538: \end{figure*}

1539: \fi

1540: %

1541: We fix the system parameters as follows: the mass of the  parent star is $m_0 =

1542: 1~\mbox{M}_{\sun}$,  the equatorial radius of the star $R_0 = 

1543: 1~\mbox{R}_{\sun}$, and  $T_{\idm{rot}}=30$~days, $k_L=0.02$ (then $J_2 \sim

1544: 10^{-7}$),  the semi-major axes of the planets are $a_1 = 0.1~\mbox{au}$,

1545: and $a_2 = 1.0~\mbox{au}$, respectively. Hence, we consider a typical

1546: hierarchical configuration with the orbital periods ratio $\sim 30$. In this

1547: experiment, the planetary masses are varied, but their ratio is kept constant,

1548: $m_1/m_2=3$. For a reference, the mass of the inner planet is labeled in the

1549: top-left corner of each respective panel in Fig.~\ref{fig:fig7}. Basically, in

1550: this test we can also analyze the effect of unknown inclination of the co-planar

1551: system on the long-term dynamics and stability. However, as we show in a recent

1552: work regarding the 14~Herculis planetary system \citep{Gozdziewski2008},

1553: planetary masses  derived from observations do not necessarily always scale

1554: according to the law of the mass-factor $1/\sin~i$. If the minimal masses are

1555: large then the mutual interactions in low-inclination configurations can

1556: strongly modify the RV signal and even {\em the mass hierarchy} may be reversed 

1557: in the orbital fits.

1558: 

1559: Figure~\ref{fig:fig7} reveals that curves representing stationary modes,  which

1560: are known in the classic problem, are usually significantly shifted and/or

1561: distorted.  Also new features of the \RP{}-plane appear and  it can be seen in

1562: the top-left panel of Fig.~\ref{fig:fig7}. In general,  the distortions of

1563: equilibria curves are the more stronger when the masses are smaller. It is quite

1564: straightforward to explain this effect. When the planetary masses decrease, also

1565: their mutual interactions (scaled by $m_1 m_2$) are decreasing.  Yet the

1566: pericenter frequency induced by $\left<\Hmut\right>$ is scaled by the mass

1567: product. Simultaneously, the GR and the spin-induced apsidal frequencies do not depend on the

1568: planetary masses directly (see Eq.~\ref{orel} and  Eq.~\ref{oqm}, respectively),

1569: and they can be regarded as approximately constant in the given mass range.

1570: Therefore  $\kappa$ increases with decreasing $m_1$ and $m_2$. 

1571: \corr{Then, also the assumptions of the secular theory are better

1572: fulfilled.}

1573: 

1574: Some parts of the stationary curves in the \RPm{} half-plane (for

1575: $\Delta{\varpi}=\pi$) comprise of unstable equilibria (they are marked with

1576: violet color). As we mentioned already, to the best of our knowledge, such

1577: solutions are yet unknown in the literature. Similarly to the non-classic

1578: equilibria discovered by \cite{Michtchenko2004}, these solutions are accompanied

1579: by the \TSR{} solutions and correspond to saddles of the secular

1580: Hamiltonian.

1581: The behavior of neighboring solutions tells us that they are Lyapunov unstable.

1582: This has been analyzed  in Sect.~3.

1583: 

1584: 

1585: Actually, the sequence of panels in Fig.~\ref{fig:fig7} illustrates a

1586: characteristic development of curves representing the  equilibria, including the

1587: \UE{} solutions. When the masses are relatively large (see the top-left panel of

1588: Fig.~\ref{fig:fig7}), the equilibria  curves are distorted and the unstable

1589: equilibria appear at the very edge of the \RPm{}-plane, in the range of moderate

1590: and large values of eccentricity. On contrary, in the classic model, the

1591: \UE{} solutions

1592: can appear (in fact, they were found) only for $\Delta\varpi=0$

1593: \citep{Michtchenko2004}.

1594: %

1595: %%In the analyzed range of masses, it does not exist in the generalized model. 

1596: %

1597: Seemingly, the new \UE{} branch located in the \RPm{}-plane is

1598: specific only for this model. When the masses decrease then $\kappa$ grows (so

1599: the GR+QM effects become comparable in magnitude to the NG interactions). This

1600: leads to further distortion of mode~II curves and to expanding the \UE{} part

1601: towards moderate $e_1$. At some point (between $m_1\sim 1.8~\mbox{m}_{\idm{J}}$

1602: and $m_1\sim 1.2~\mbox{m}_{\idm{J}}$) both  stationary curves meet in a

1603: bifurcation point. Here, we can explain the particular choice of parameters used

1604: to construct  Fig.~\ref{fig:fig2}. When the masses become smaller, the 

1605: equilibria curves separate along $e_2$. We note that already for  $m_1\sim

1606: 1.2~\mbox{m}_{\idm{J}}$,  the \RP{}-plane is dominated by the GR+QM corrections.

1607: We recall that in the classic case, the qualitative features of the \RP{}-plane

1608: do not depend on the masses individually \citep{Michtchenko2004}, only on their

1609: ratio in the approximation of small values (see also the sequence of plots in

1610: Fig.~\ref{fig:fig7}). This conclusion is not true

1611: anymore in the

1612: realm of the generalized model.

1613: %

1614: %-----------------------------------------------------------------------------

1615: \subsection{Dependence of the secular dynamics on semi-major axes}

1616: %-----------------------------------------------------------------------------

1617: %

1618: In the next experiment, we investigate the dependence of the  secular dynamics

1619: of the generalized model on individual semi-major axes;

1620: note that the dynamics of the classic model depend only

1621: on their ratio, $\alpha$. The results are illustrated in

1622: Fig.~\ref{fig:fig8}. We proceed in the same manner as to draw 

1623: Figs.~\ref{fig:fig2} and \ref{fig:fig7}. We seek for stationary solutions, and we

1624: overplot the found equilibria on color-coded  contour levels of coefficient

1625: $\kappa$. The primary parameters of the tested configurations are the

1626: following:  $m_0 = 1~\mbox{M}_{\sun}$, $R = 1~\mbox{R}_{\sun}$,

1627: $T_{\idm{rot}}=30$~days, $k_L=0.02$, $m_1 = 0.4~\mbox{m}_{\idm{J}}$,  $m_2 =

1628: 0.2~\mbox{m}_{\idm{J}}$.  The ratio of semi-major axes is kept constant, $\alpha

1629: \equiv a_1/a_2 = 0.1$, while the individual $a_1, a_2$ are varied.   For a

1630: reference, the nominal value of $a_1$ is labeled in the top-left corner in each

1631: respective panel.  

1632: %

1633: \ifpdf

1634: \begin{figure*}

1635: \centerline{

1636: \vbox{

1637: \hbox{\includegraphics[width=5.9cm]{fig8a.png}

1638:       \includegraphics[width=5.9cm]{fig8b.png}

1639:       \includegraphics[width=5.9cm]{fig8c.png}}

1640: \hbox{\includegraphics[width=5.9cm]{fig8d.png}

1641:       \includegraphics[width=5.9cm]{fig8e.png}

1642:       \includegraphics[width=5.9cm]{fig8f.png}}

1643:       }

1644:       }

1645: \caption{

1646: The survey of the secular dynamics of two-planet system, when the semi-major

1647: axes are varied. Panels are constructed in the same way,  as in Fig.

1648: \ref{fig:fig7}.  

1649: Color contours are for the ratio of the  apsidal frequency induced by

1650: the general relativity and quadrupole moment  to the  apsidal frequency caused by mutual

1651: interactions between planets. 

1652: The thick skew

1653: line is for collisions line defined with $a_1 (1 \pm e_1) = a_2 (1 - e_2)$. The

1654: masses are $m_0 = 1 \mbox{M}_{\sun}$, $m_1=0.4 \mbox{m}_{\idm{J}}$, $m_2=0.2

1655: \mbox{m}_{\idm{J}}$, respectively, the characteristic radius of the star  $R_0=1

1656: \mbox{R}_{\sun}$, $T_{\idm{rot}}=30$~days, $k_L=0.02$. Each panel is for

1657: different semi-major axes, fulfilling the condition of constant $\alpha \equiv

1658: a_1/a_2 = 0.1$.  The semi-major axis of the inner planet labels each respective

1659: panel. Compared to  Fig. \ref{fig:fig7}, an additional  unstable equilibrium for

1660: the classic model appears and it is marked with thick green lines.

1661: }

1662: \label{fig:fig8}

1663: \end{figure*}

1664: \else

1665: \begin{figure*}

1666: \centerline{

1667: \vbox{

1668: \hbox{\includegraphics[width=5.9cm]{fig8a.eps}

1669:       \includegraphics[width=5.9cm]{fig8b.eps}

1670:       \includegraphics[width=5.9cm]{fig8c.eps}}

1671: \hbox{\includegraphics[width=5.9cm]{fig8d.eps}

1672:       \includegraphics[width=5.9cm]{fig8e.eps}

1673:       \includegraphics[width=5.9cm]{fig8f.eps}}

1674:       }

1675:       }

1676: \caption{

1677: The survey of the secular dynamics of two-planet system, when the semi-major

1678: axes are varied. Panels are constructed in the same way,  as in Fig.

1679: \ref{fig:fig7}.  

1680: Color contours are for the ratio of the  apsidal frequency induced by

1681: the general relativity and quadrupole moment  to the  apsidal frequency caused by mutual

1682: interactions between planets. 

1683: The thick skew

1684: line is for collisions line defined with $a_1 (1 \pm e_1) = a_2 (1 - e_2)$. The

1685: masses are $m_0 = 1 \mbox{M}_{\sun}$, $m_1=0.4 \mbox{m}_{\idm{J}}$, $m_2=0.2

1686: \mbox{m}_{\idm{J}}$, respectively, the characteristic radius of the star  $R_0=1

1687: \mbox{R}_{\sun}$, $T_{\idm{rot}}=30$~days, $k_L=0.02$. Each panel is for

1688: different semi-major axes, fulfilling the condition of constant $\alpha \equiv

1689: a_1/a_2 = 0.1$.  The semi-major axis of the inner planet labels each respective

1690: panel. Compared to  Fig. \ref{fig:fig7}, an additional  unstable equilibrium for

1691: the classic model appears and it is marked with thick green lines.

1692: }

1693: \label{fig:fig8}

1694: \end{figure*}

1695: \fi

1696: %

1697: For decreasing $a_1$, the derivatives of $\Hmut, \Hrel, \Hflat$ over $G_1$ 

1698: increase, hence the magnitude of the respective correction to 

1699: the apsidal  frequency grows.

1700: Moreover, the GR and QM induced apsidal frequency increase faster than the rate

1701: of the pericenter advance forced by the NG interactions. In the region of

1702: \RP{}-plane

1703: painted in yellow (light gray), the GR and QM perturbations dominate over the NG

1704: interactions.  This region expands quickly with decreasing semi-major axis of

1705: the inner planet, $a_1$. Already for $a_1 \sim 0.4$~au (which is similar to the

1706: semi-major axis of Mercury in the Solar system), the apparent {\em corrections}

1707: to the classic model, may contribute much more to the pericenter frequency  of

1708: the innermost planet than the point mass Newtonian interactions.

1709: 

1710: The top-left panel in Fig. \ref{fig:fig8} is for  $a_1 = 2$~au and $a_2 =

1711: 20$~au, respectively. For these parameters, a shift of the curve of stationary

1712: solutions, when compared to the ones in the classic model, is already

1713: significant. In the next  panel ($a_1=1$~au, $a_2=10$~au) the distortion of

1714: curves representing stationary modes is even stronger. Moreover, the \UE{} mode

1715: present in the classic model in the \RPp{}-plane, cannot be found  in that

1716: half-plane anymore. Simultaneously, new solutions appear at the very edge of

1717: the  \RPm-plane, in the range of large $e_1$. For $a_1=0.5$~au, this branch of

1718: stationary modes is even more extended. Starting with this value of  $a_1$, the

1719: structure of the \RP{}-plane with respect to the generalized  model is very

1720: different from that ones in the classic case. For smaller $a_1$, the curves of 

1721: stationary modes are still more distorted.  Clearly, these distortions cannot be

1722: regarded as small. This result is  quite unexpected, recalling that the

1723: semi-major axes and the planetary masses by no means  are ``extreme''. In spite

1724: of these ``typical'' parameters, the secular theories of the classic and

1725: generalized models lead to qualitatively different view of the phase space. We

1726: stress again that the notion of the GR and QM effects as corrections (or small

1727: perturbations) to the  secular Hamiltonian should be understood in quite a new

1728: light.

1729: %

1730: %-------------------------------------------------------------------------------

1731: \subsection{Dependence of the secular dynamics on the stellar spin}

1732: %-------------------------------------------------------------------------------

1733: %

1734: In the last parametric survey,  we study the dependence of the secular dynamics

1735: in the realm of the generalized model on the stellar spin (or, effectively, on

1736: the second zonal harmonic $J_2$). Figure \ref{fig:fig9} illustrates the

1737: \RP{}-plane  computed for  following  parameters: $m_0 = 1~\mbox{M}_{\sun}$, \

1738: $m_1=1.25~\mbox{m}_{\idm{J}}$, $m_2=0.25~\mbox{m}_{\idm{J}}$, $a_1=0.1$~au,

1739: $a_2=1.0$~au,  $R_0=1~\mbox{R}_{\sun}$, $k_L=0.02$. The top-left panel in

1740: Fig.~\ref{fig:fig9} is for the GR correction only and subsequent plots are for

1741: decreasing rotation period (generalized model)  $T_{\idm{rot}}$ of the star (its particular values 

1742: label the respective plots). This sequence corresponds to increasing $J_2$.

1743: %

1744: \ifpdf

1745: \begin{figure*}

1746: \centerline{

1747: \vbox{

1748: \hbox{\includegraphics[width=5.9cm]{fig9a.png}

1749:       \includegraphics[width=5.9cm]{fig9b.png}

1750:       \includegraphics[width=5.9cm]{fig9c.png}

1751:       }

1752: \hbox{\includegraphics[width=5.9cm]{fig9d.png}

1753:       \includegraphics[width=5.9cm]{fig9e.png}

1754:       \includegraphics[width=5.9cm]{fig9f.png}

1755:       }

1756: }}      

1757: \caption{

1758: The representative energy planes $(e_1 \cos{\Delta\varpi}, e_2)$ for 

1759: $\Delta\varpi=0$ (the right half-planes, \RPp{}) and $\Delta\varpi=\pi$  (the

1760: left half-planes, \RPm{}). Colors are for the ratio of the  apsidal frequency

1761: induced by the general relativity and quadrupole moment  to the  apsidal

1762: frequency caused by mutual interactions between planets.  The thick lines mark

1763: the stationary solutions. The thick red curves are for stable equilibria in the

1764: classic model. The blue and violet curves mark  positions of stable and unstable

1765: equilibria in the generalized model, respectively.  Thin lines are for the energy levels of a

1766: generalized model. The thick skew line is for collisions line defined with $a_1

1767: (1 \pm e_1) = a_2 (1 - e_2)$. Parameters of the system are the following: $m_0 =

1768: 1.0~\mbox{M}_{\sun}$,  $m_1=1.25~\mbox{m}_{\idm{J}}$,

1769: $m_2=0.25~\mbox{m}_{\idm{J}}$,  $a_1=0.1 \mbox{au}$, $a_2=1.0~\mbox{au}$, 

1770: $R_0=1~\mbox{R}_{\sun}$, $k_L=0.02$.  Subsequent panels are for  different

1771: rotational periods of the star, $T_{\idm{rot}}$ which is labeled in each

1772: respective plot. 

1773: \corr{For a reference,

1774: the top left panel is for the GR correction only.}

1775: }

1776: \label{fig:fig9}

1777: \end{figure*}

1778: \else

1779: \begin{figure*}

1780: \centerline{

1781: \vbox{

1782: \hbox{\includegraphics[width=5.9cm]{fig9a.eps}

1783:       \includegraphics[width=5.9cm]{fig9b.eps}

1784:       \includegraphics[width=5.9cm]{fig9c.eps}

1785:       }

1786: \hbox{\includegraphics[width=5.9cm]{fig9d.eps}

1787:       \includegraphics[width=5.9cm]{fig9e.eps}

1788:       \includegraphics[width=5.9cm]{fig9f.eps}

1789:       }

1790: }}      

1791: \caption{

1792: The representative energy planes $(e_1 \cos{\Delta\varpi}, e_2)$ for 

1793: $\Delta\varpi=0$ (the right half-planes, \RPp{}) and $\Delta\varpi=\pi$  (the

1794: left half-planes, \RPm{}). Colors are for the ratio of the  apsidal frequency

1795: induced by the general relativity and quadrupole moment  to the  apsidal

1796: frequency caused by mutual interactions between planets.  The thick lines mark

1797: the stationary solutions. The thick red curves are for stable equilibria in the

1798: classic model. The blue and violet curves mark  positions of stable and unstable

1799: equilibria in the generalized model, respectively.  Thin lines are for the energy levels of a

1800: generalized model. The thick skew line is for collisions line defined with $a_1

1801: (1 \pm e_1) = a_2 (1 - e_2)$. Parameters of the system are the following: $m_0 =

1802: 1.0~\mbox{M}_{\sun}$,  $m_1=1.25~\mbox{m}_{\idm{J}}$,

1803: $m_2=0.25~\mbox{m}_{\idm{J}}$,  $a_1=0.1 \mbox{au}$, $a_2=1.0~\mbox{au}$, 

1804: $R_0=1~\mbox{R}_{\sun}$, $k_L=0.02$.  Subsequent panels are for  different

1805: rotational periods of the star, $T_{\idm{rot}}$ which is labeled in each

1806: respective plot. 

1807: \corr{For a reference,

1808: the top left panel is for the GR correction only.}

1809: }

1810: \label{fig:fig9}

1811: \end{figure*}

1812: \fi

1813: %

1814: The top-left panel of Fig.~\ref{fig:fig9} is for a spherical,

1815: non-rotating star ($J_2=0$, no tidal bulge),  the bottom-right panel is for 

1816: very fast rotation characteristic for a young object (then

1817: $J_2=1\times10^{-4}$). Curiously,  changes of the spin encompassing that range

1818: are well enough to emerge new families and  stationary solutions which we

1819: described above. They are represented, as before, by thick violet curves drawn in the

1820: \RPm{}-plane.  Actually,  the sequence of plots simulates  variations of

1821: flattening during the lifetime of the star. Hence, after skipping 

1822: dissipative tidal perturbations, we may conclude that the structure of the phase space of

1823: the secular system (and, in general, also its dynamical stability) may depend

1824: not only on the observed or measured  orbital parameters but also on the age and

1825: the spin rate of the host star. 

1826: 

1827: The structure of the \RP{}-plane and stationary modes which are illustrated

1828: in Fig.~\ref{fig:fig9} remind us closely Fig.~\ref{fig:fig7} and

1829: Fig.~\ref{fig:fig8}. In fact, as we already mentioned,  the dependence of the dynamics on the model

1830: parameters may be described uniformly through  $\kappa$, see Eq.~\ref{eq:kappa}.

1831: Indeed, the increase of the stellar spin leads to increase of the  pericenter

1832: frequency and the nominator of Eq.~\ref{eq:kappa} (note that other terms are

1833: constant). The same behavior of $\kappa$ is caused by decreasing

1834: masses (see the sequence of diagrams in Fig.~\ref{fig:fig7} and also

1835: discussion in Sect.~4.1). In that

1836: case, two terms of the nominator of $\kappa$ do not change, but the denominator

1837: decreases. Finally, if the semi-major axes decrease in constant ratio, $\kappa$

1838: also grows because the GR and QM-induced  correction to the apsidal frequency

1839: grows {\em faster} than the NG-forced apsidal frequency. In that sense, all 

1840: point-mass gravity corrections governed by the described above parameter 

1841: changes, modify similarly the structure of the \RP-plane. 

1842: 

1843: Considering our simplified secular model, the existence of the branch

1844: of stationary solutions in

1845: the regime of large $e_1$ and small $e_2$ may seem questionable in {\em the real}

1846: planetary configurations. In that regime, the tidal perturbations

1847: may become very significant for the secular evolution.  However,

1848: the position of the bifurcation point 

1849: in the \RPm{}-plane, where the two branches

1850: of equilibria curves meet, and which marks the extent of the branch, is

1851: shifted towards moderate $e_1 \sim 0.6$--$0.7$ when the  mass

1852: ratio $m_1/m_2$  grows.

1853: This effect can be observed in the geometric evolution of

1854: these branches in Figs.~\ref{fig:fig8}, \ref{fig:fig7} and \ref{fig:fig9}

1855: which form a sequence of $m_1/m_2=2,3,5$, respectively.

1856: Hence, planetary configurations in that regime may really be found  

1857: in the Nature.

1858: 

1859: %

1860: %-------------------------------------------------------------------------------

1861: \section{The secular dynamics of the $\upsilon$~And system}

1862: %-------------------------------------------------------------------------------

1863: %

1864: Finally, we consider the generalized coplanar problem of three planets.  This

1865: model can be still described by the Hamiltonian written  in the general form of

1866: Eqs.~(\ref{HKepl}--\ref{Hrel}), and (\ref{eq:hrf}--\ref{eq:htb}) with $N=3$. The

1867: averaged perturbing Hamiltonian has the form of Eqs.~(\ref{Hsec}),

1868: (\ref{avHmut}), (\ref{expanssion}), (\ref{avbHrel}), (\ref{avHflat}) and

1869: (\ref{avHflatT}). To study the properties of the secular system, we introduce  

1870: the following set of angle-action variables related to the Poincar\'e canonical

1871: elements

1872: \citep{Migaszewski2008a}:

1873: \begin{eqnarray}

1874: \label{3pl}

1875: && \sigma_1 \equiv \varpi_3 - \varpi_1,  \quad G'_1, \nonumber\\

1876: && \sigma_2 \equiv \varpi_3 - \varpi_2 ,  \quad G'_2, \\

1877: && \sigma_3 \equiv -\varpi_3,  \quad \mbox{AMD} = G'_1 + G'_2 + G'_3, \nonumber

1878: \end{eqnarray}

1879: where $G'_i \equiv L_i-G_i$ (see Eq.~\ref{dvars}). Because $\sigma_3$ is the

1880: cyclic angle, the Angular Momentum Deficit (AMD) is conserved, and the reduced

1881: system has two degrees of freedom. These variables make it possible to construct

1882: the representative energy plane in a similar manner as  for the two-planet

1883: system. Moreover, the choice of such a plane is not unique. One of possible

1884: definitions can be  derived through a direct analogy to the two-planet system

1885: case. The {\em symmetric} representative plane can be defined as the set of

1886: phase-space  points fulfilling the following relation:

1887: \[

1888: \frac{\partial\,\Hsec}{\partial\,\sigma_i} = 0, \quad i=1,2,3,

1889: \]

1890: with the simultaneous conditions that $\sigma_i=0,\pi$. For details, see

1891: \citep{Migaszewski2008a}.

1892: 

1893: A sequence of symmetric  representative planes shown in Fig.~\ref{fig:fig10}

1894: illustrates the qualitative properties of the generalized secular model of the

1895: $\upsilon$ Andromedae planetary system \citep{Butler1999}. This system comprises

1896: of three planets with masses and semi-major axes  derived through the radial

1897: velocity observations:

1898: $m_0=1.27~\mbox{M}_{\sun}$,

1899: $m_1=0.69~\mbox{m}_{\idm{J}}$, 

1900: $m_2=1.98~\mbox{m}_{\idm{J}}$, 

1901: $m_3=3.95~\mbox{m}_{\idm{J}}$, 

1902: $a_1=0.059~\mbox{au}$, 

1903: $a_2=0.83~\mbox{au}$, 

1904: $a_3=2.51~\mbox{au}$.

1905: The constant of the AMD integral (effectively, the integral

1906: of the total angular momentum) was obtained for the following eccentricities of the

1907: nominal system:  $e_1=0.029$, $e_2=0.254$, $e_3=0.242$. 

1908: %For a reference, the top-left panel in Fig.~\ref{fig:fig10} is computed for the classic model.

1909: 

1910: We consider the representative plane for varying age of the parent star,

1911: starting with approximately 30~Myr before entering the ZAMS.  Subsequent plots

1912: are labeled by the lifetime $\tau$ relative to this moment taken  as the

1913: zero-point of the time scale, rotation period $T_{\idm{rot}}$ and stellar radius

1914: ($R_{0}$) expressed in Sun's radius. For a reference, the  classic model and

1915: with the GR correction  are illustrated in the first two top-left panels of

1916: Fig.~\ref{fig:fig10}. These panels are derived for  non-rotating, non-distorted

1917: spherical star. When the star is spinning, its second zonal harmonic may be as

1918: large $J_2 \sim  10^{-3}$.  The current equatorial  radius 

1919: of $\upsilon$~Andr is approximately 

1920: $R_0=1.26~\mbox{R}_{\sun}$. We note that the stellar radius and the spin period  of

1921: the star at $\tau=-30$~Myr are taken from \citep{Nagasawa2005}, and were

1922: linearly interpolated over $\tau \in [-30,0]$~Myr.

1923: 

1924: The results are again quite surprising.  After adding the GR corrections to the

1925: Hamiltonian of the classic model, the  overall view of the phase space changes

1926: significantly.  The saddle of the secular Hamiltonian which is present  in the

1927: classic model now vanishes.  At its place, a new {\em maximum} of the secular

1928: Hamiltonian appears. Moreover, at the bottom half-plane of the representative

1929: plane, close to the border of the permitted region of motion,  two new saddle

1930: points appear. 

1931: 

1932: The next diagrams illustrate changes of the structure of the \RP{}-plane and a

1933: development  of equilibria in a sequence simulating time evolution of the

1934: stellar spin.  At the beginning, before the star enters the ZAMS, the

1935: characteristic plane reveals a sharp maximum and a saddle in the very edge of

1936: the region of permitted motions.  The thin curve surrounding the maximum marks

1937: the energy level of the nominal system. When the rotation period increases

1938: up to $\sim 8$~days, the secondary extremum (the minimum) emerges in the place of the

1939: saddle and it persists shortly before the ZAMS stage and for longer rotational

1940: periods.

1941: 

1942: Curiously, the only feature seen in the energy diagrams, which survives the spin

1943: variations during the whole lifetime of the star, and persist in  the

1944: generalized model  (with the GR and QM corrections) is the stable equilibrium

1945: point related to the maximum of the secular Hamiltonian, which  is found in the

1946: range of  small $e_1$ and moderate $e_2$. Curiously, the nominal system appears

1947: in the energy level drawn with grey (green), thick line surrounding this

1948: maximum of $\Hsec$. 

1949: \corr{We also may notice that close to

1950: this equilibrium of the generalized model, a

1951: saddle of the classic model appears which is {\em

1952: linearly stable}. 

1953: }

1954: %

1955: \ifpdf

1956: \begin{figure*}

1957: \centerline{

1958: \vbox{

1959: \hbox{

1960: \includegraphics[width=44mm]{fig10a.pdf}

1961: \includegraphics[width=44mm]{fig10b.pdf}

1962: \includegraphics[width=44mm]{fig10c.pdf}

1963: \includegraphics[width=44mm]{fig10d.pdf}

1964: }

1965: \hbox{

1966: \includegraphics[width=44mm]{fig10e.pdf}

1967: \includegraphics[width=44mm]{fig10f.pdf}

1968: \includegraphics[width=44mm]{fig10g.pdf}

1969: \includegraphics[width=44mm]{fig10h.pdf}

1970: }

1971: }}

1972: \caption{

1973: The secular energy levels on the \textit{symmetric representative plane} for

1974: three-planet system $\upsilon$ Andromede. The map coordinates are $x \equiv e_1

1975: \cos{\sigma_1}$, $y \equiv e_2 \cos{\sigma_2}$, where $ \sigma_1, \sigma_2$ are

1976: $0$ (positive values of $x$ or $y$) or $\pi$ (negative values of $x$ or $y$).

1977: Grey-colored region means forbidden motions with 

1978: $e_3<0$. Black curves are for the energy levels, green, thick

1979: levels are for the energy of the nominal $\upsilon$~Andr  system. Each panel is for a different

1980: setup of the planetary system model.  

1981: From the {\em left-top panel}: the first panel is for

1982: the classic NG model, the next panel is for the NG+GR model. 

1983: Next panels are for generalized model with the QM corrections parameterized

1984: by the spin rate of the parent star and its lifetime $\tau$ before

1985: the ZAMS stage. 

1986: }

1987: \label{fig:fig10}

1988: \end{figure*}

1989: \else

1990: \begin{figure*}

1991: \centerline{

1992: \vbox{

1993: \hbox{

1994: \includegraphics[width=44mm]{fig10a.eps}

1995: \includegraphics[width=44mm]{fig10b.eps}

1996: \includegraphics[width=44mm]{fig10c.eps}

1997: \includegraphics[width=44mm]{fig10d.eps}

1998: }

1999: \hbox{

2000: \includegraphics[width=44mm]{fig10e.eps}

2001: \includegraphics[width=44mm]{fig10f.eps}

2002: \includegraphics[width=44mm]{fig10g.eps}

2003: \includegraphics[width=44mm]{fig10h.eps}

2004: }

2005: }}

2006: \caption{

2007: The secular energy levels on the \textit{symmetric representative plane} for

2008: three-planet system $\upsilon$ Andromede. The map coordinates are $x \equiv e_1

2009: \cos{\sigma_1}$, $y \equiv e_2 \cos{\sigma_2}$, where $ \sigma_1, \sigma_2$ are

2010: $0$ (positive values of $x$ or $y$) or $\pi$ (negative values of $x$ or $y$).

2011: Grey-colored region means forbidden motions with 

2012: $e_3<0$. Black curves are for the energy levels, green, thick

2013: levels are for the energy of the nominal $\upsilon$~Andr  system. Each panel is for a different

2014: setup of the planetary system model.  

2015: From the {\em left-top panel}: the first panel is for

2016: the classic NG model, the next panel is for the NG+GR model. 

2017: Next panels are for generalized model with the QM corrections parameterized

2018: by the spin rate of the parent star and its lifetime $\tau$ before

2019: the ZAMS stage. 

2020: }

2021: \label{fig:fig10}

2022: \end{figure*}

2023: \fi

2024: %

2025: %-------------------------------------------------------------------------------

2026: \section{Conclusions}

2027: %-------------------------------------------------------------------------------

2028: %

2029: In this work, we consider a generalized secular theory of coplanar, $N$-planet

2030: system. Extending the model analyzed in the recent works devoted 

2031: to the secular planetary dynamics  with mutual Newtonian

2032: point-to-point interactions

2033: \citep[e.g.,][]{Michtchenko2004,Libert2005,Gallardo2005,Migaszewski2008a},  we

2034: consider the influence of the general relativity and quadrupole moment of the

2035: parent star on the secular dynamics of the innermost planet and stability of 

2036: the whole planetary system. In general, these corrections to the classic model

2037: still do not cover all physics governing the dynamics of such systems. In some

2038: cases (for instance, of the short-period hot-Jupiters), the tidal, dissipative

2039: torques acting between the inner planet and the star may be significant for the

2040: orbital evolution. However,  our main  goal is rather to extend the classic

2041: model with the perturbations that are conservative and may be well modeled in

2042: the realm of the Hamiltonian mechanics than to build a complete, general

2043: secular theory. Still, this approach is useful to a wide class of systems, when

2044: the tidal interactions may be regarded as secondary effects, or are acting 

2045: during much longer characteristic time-scale than the GR and QM perturbations.

2046: In reward, for paying the price of less general model, we may investigate the

2047: secular dynamics in a global manner. 

2048: 

2049: Our analytic model follows assumptions required by the averaging theorem.

2050: Technically, the averaging has been done with  the help of a very simple method.

2051: This algorithm relies on  appropriate change of integration variables. It does

2052: not incorporate any classic Fourier expansion of the perturbing function.  We

2053: obtain a very precise analytic model of the coplanar, $N$-planet system in

2054: terms of the semi-major axes ratio. It can be regarded as a generalization of

2055: the recent analytic secular theories  of the classic model investigated in

2056: many recent papers \citep[e.g.,][]{Ford2000,Lee2003,Michtchenko2004,Libert2005}. On the other

2057: hand,  our work also covers, to some extent, the global dynamics  of the

2058: generalized model studied in \cite{Mardling2002,Nagasawa2005} with the help of

2059: the Gauss/Lagrange planetary equations of motion. We stress, however, that our

2060: investigations are devoted to  more narrow class of systems (regarding the

2061: conservative perturbations).

2062: 

2063: A general conclusion which can be derived on the basis of the generalized theory

2064: is quite unexpected. Even in a case when the orbital parameters cannot be

2065: regarded as {\em extreme}, the corrections to the classic Hamiltonian stemming

2066: from the general relativity and the quadrupole moment of the star, may affect

2067: the secular dynamics dramatically. Not only  the structure of the phase space of

2068: the secular model changes, and new branches  of stationary solutions appear.

2069: These solutions may bifurcate within small relative ranges of the parameters

2070: (for instance, when the spin of the parent star is changing). We show that there

2071: is no simple and general recipe to predict the behavior of the secular system,

2072: when the perturbations are ``switched on''. The secular dynamics of  the

2073: generalized model becomes extremely complex and rich. For some combinations of

2074: the system parameters, the notion of the GR and QM  effects as {\em

2075: corrections}  to the point-mass  NG interactions does not seem proper anymore.  In some

2076: cases, these effects may be {\em more important} for the secular dynamics than

2077: the mutual, point mass Newtonian interactions between the planets. 

2078: 

2079: We also show that these effects may be significant for the dynamical stability

2080: of planetary systems both in the short-term and 

2081: in the secular time scales. For instance,

2082: the QM generated perturbations may directly depend on the star age and its

2083: physical parameters ($k_L$, $R_0$). In turn, these effects may strongly

2084: influence the structure of the phase space and can imply short-term, strong

2085: chaotic orbital evolution during a few secular periods.

2086: 

2087: The direct tests of the analytic theory are very encouraging.  The results

2088: justify its great accuracy. The precision of the analytic calculations is  very

2089: important for studying the global dynamics of hierarchical systems. The

2090: alternative numerical approach would require huge CPU time because the

2091: hierarchical planetary systems evolve during very different time scales. Then

2092: the CPU requirements of the direct numerical integrations are  by orders of

2093: magnitude larger than those ones needed by the analytic formulae. 

2094: %

2095: %-------------------------------------------------------------------------------

2096: \section*{Acknowledgments}

2097: %-------------------------------------------------------------------------------

2098: %

2099: We are very grateful to Rosemary Mardling for careful reading of the manuscript,

2100: constructive and informative review, many suggestions and invaluable remarks

2101: that greatly improved the work. We would like to thank Tatiana Michchenko for a

2102: discussion and comments on the manuscript. This work is supported by the Polish

2103: Ministry of Science and Education, Grant No. 1P03D-021-29. C.M. is also

2104: supported by Nicolaus Copernicus University Grant No.~408A.

2105: 

2106: \bibliographystyle{mn2e}

2107: \bibliography{ms}

2108: \label{lastpage}

2109: \end{document}

2110: 

2111: 

2112: