Secular Evolution of Extrasolar Planetary Systems: an Extension of the Laplace-Lagrange Secular Theory Marco Sansottera [a] [a] Namur Center for Complex Systems (naxys) Based on a research work in collaboration with Anne-Sophie Libert [a] DDA 2013 Paraty, 06.05.2013
Extrasolar systems vs. Solar System The main difference between the extrasolar systems and the Solar System regards the shape of the orbits. In the extrasolar systems, the majority of the orbits describe true ellipses (high eccentricities) and no more almost circles like in the Solar System. The classical approach uses the circular approximation as a reference. Dealing with systems with high eccentricities we need to compute the expansion at high order to study the long-term evolution of the extrasolar planetary systems.
Aim of the work 1 Can we predict the long-term evolution of extrasolar systems?
Aim of the work 1 Can we predict the long-term evolution of extrasolar systems? 2 How can we evaluate the influence of a mean-motion resonance?
Aim of the work 1 Can we predict the long-term evolution of extrasolar systems? Numerical integrations are really accurate, but have high computational cost. One has to compute a numerical integration for each initial condition. 2 How can we evaluate the influence of a mean-motion resonance?
Aim of the work 1 Can we predict the long-term evolution of extrasolar systems? Numerical integrations are really accurate, but have high computational cost. One has to compute a numerical integration for each initial condition. Normal forms provide non-linear approximations of the dynamics in a neighborhood of an invariant object. In addition, an accurate analytic approximation is the starting point for the study of the effective long-time stability. 2 How can we evaluate the influence of a mean-motion resonance?
Aim of the work 1 Can we predict the long-term evolution of extrasolar systems? Numerical integrations are really accurate, but have high computational cost. One has to compute a numerical integration for each initial condition. Normal forms provide non-linear approximations of the dynamics in a neighborhood of an invariant object. In addition, an accurate analytic approximation is the starting point for the study of the effective long-time stability. 2 How can we evaluate the influence of a mean-motion resonance? The semi-major axis ratio gives a rough indication of the proximity to the main mean-motion resonance.
Aim of the work 1 Can we predict the long-term evolution of extrasolar systems? Numerical integrations are really accurate, but have high computational cost. One has to compute a numerical integration for each initial condition. Normal forms provide non-linear approximations of the dynamics in a neighborhood of an invariant object. In addition, an accurate analytic approximation is the starting point for the study of the effective long-time stability. 2 How can we evaluate the influence of a mean-motion resonance? The semi-major axis ratio gives a rough indication of the proximity to the main mean-motion resonance. However, the impact of the proximity to a mean-motion resonance on the secular evolution of a planetary system depends on many parameters. This is due to the non-linear character of the system.
Hamiltonian of the three-body problem F(r, r) = T (0) ( r)+u (0) (r)+t (1) ( r)+u (1) (r), where r are the heliocentric coordinates and r the conjugated momenta. T (0) ( r)= 1 2 U (0) (r)= G 2 ( 1 r j 2 + 1 ), m 0 m j j=1 2 j=1 T (1) ( r)= r 1 r 2 m 0, m 0 m j r j, U (1) (r)= G m 1m 2 r 1 r 2.
Poincaré variables in the plane Λ j = m 0m j G(m 0 +m j )a j, λ j = M j +ω j, m 0 +m j } {{ } Fast variables, O(1 year). ξ j = 2Λ j 1 1 ej 2 cos(ω j ), η j = 2Λ j 1 1 ej 2 sin(ω j ), } {{ } Secular variables, O(1000 years). where a j, e j, M j and ω j are the semi-major axis, the eccentricity, the mean anomaly and periapsis argument of the j-th planet, respectively.
Expansion of the Hamiltonian In order to compute the Taylor expansion of the Hamiltonian around the fixed value Λ, we introduce the translated fast actions, The Hamiltonian reads, where h (Kep) j 1,0 H = n L+ j 1=2 L = Λ Λ. h (Kep) j 1,0 (L)+µ j 1=0j 2=0 is a hom. pol. of degree j 1 in L and h j1,j 2 is a h j1,j 2 (L,λ,ξ,η). hom. pol. of degree j 1 in L, hom. pol. of degree j 2 in ξ,η, with coeff. that are trig. pol. in λ.
Expansion of the Hamiltonian In order to compute the Taylor expansion of the Hamiltonian around the fixed value Λ, we introduce the translated fast actions, The computed Hamiltonian reads, where h (Kep) j 1,0 H = n L+ 2 j 1=2 L = Λ Λ. h (Kep) j 1,0 (L)+µ 1 12 h j1,j 2 (L,λ,ξ,η), j 1=0j 2=0 is a hom. pol. of degree j 1 in L and h j1,j 2 is a hom. pol. of degree j 1 in L, hom. pol. of degree j 2 in ξ,η, with coeff. that are trig. pol. in λ. We will choose the lowest possible limits in order to include the fundamental features of the system.
First order averaging We consider the averaged Hamiltonian, H(L,λ,ξ,η) = H(L,λ,ξ,η) λ. Namely, we get rid of the fast motion removing from the expansion of the Hamiltonian all the terms that depend on the fast angles λ. This is the so called first order averaging. We end up with the Hamiltonian, 12 H (sec) = µ h 0,j2 (ξ,η). j 2=0
Secular dynamics Doing the averaging over the fast angles (as we are interested in the secular motions of the planets), the system pass from 4 to 2 degrees of freedom, H (sec) = H 0 (ξ,η)+h 2 (ξ,η)+h 4 (ξ,η)+..., where H 2j is a hom. pol. of degree (2j +2) in (ξ, η), for all j N. ξ = η = 0 is an elliptic equilibrium point, thus we can introduce action-angle variables via Birkhoff normal form. Having the Hamiltonian in Birkhoff normal form, we can easily solve the equations of motion and finally obtain the motion of the orbital parameters.
Sketch of the procedure (1/2) H = H 0 (ξ,η)+h 2 (ξ,η)+h 4 (ξ,η)+.... Introduce the action-angle variables, ξ j = 2Φ j cosϕ j, η j = 2Φ j sinϕ j. Compute the Birkhoff normal form up to a finite order, r, H (r) = Z 0 (Φ)+Z 2 (Φ)+...+Z r (Φ)+R r (Φ,ϕ).
Sketch of the procedure (2/2) If the remainder, R r, is small enough, we can neglect it! The equations of motion are Φ j (0) = 0, ϕ j (0) = H(r) Φ j. (Φ(0),ϕ(0)) The solutions are Φ j (t) = Φ j (0), ϕ j (t) = ϕ j (0) t +ϕ j (0).
Analytical integration ( ) Secular + NF (r) ( η(0), ξ(0) Φ (r) (0), ϕ (r) (0) ) Numerical integration Φ (r) (t)=φ (r) (0) ϕ (r) (t)= ϕ (r) (0)t+ϕ (r) (0) ( η(t), ξ(t) ) (NF (r) ) 1 ( Φ (r) (t), ϕ (r) (t) )
Back to the orbital elements Remember that in the original variables, by definition, we have ξ j = 2Λ j 1 1 ej 2 cos(ω j ), η j = 2Λ j 1 1 ej 2 sin(ω j ). It s now straightforward to compute the evolution of the eccentricities, obtained by analytical integration, ) e j = 1 (1 ξ2 j +ηj 2 2 2Λ, j and compare the results with the numerical integration.
First order approximation (HD 134987) HD 134987: the system is secular. 0.26 0.24 Eccentricities analy_hd_134987_ord1 "./ecc_0.dat" "./ecc_1.dat" "./orbitals_1.dat" u 1:4 "./orbitals_2.dat" u 1:4 0.22 0.2 0.18 0.16 0.14 0.12 0.1 0.08 0 1e+06 2e+06 3e+06 4e+06 5e+06 6e+06
But... near mean-motion resonance (HD 108874) HD 108874: the system is close to the 4:1 mean-motion resonance. First order averaged Hamiltonian failed. 0.35 0.3 Eccentricities analy_hd_108874_ord1 "./ecc_0.dat" "./ecc_1.dat" "./orbitals_1.dat" u 1:4 "./orbitals_2.dat" u 1:4 0.25 0.2 0.15 0.1 0.05 0 10000 20000 30000 40000 50000 60000 70000
Second order averaging Coming back to the original Hamiltonian, H = n L+ j 1 2 h j1,0(l)+µ j 1 0 If we consider the point L(0) = 0, we have h j1,j 2 (L,λ,ξ,η). j 2 0 L j = µ h 0,j2 (λ,ξ,η). λ j j 2 0 In order to get rid of the fast motion, instead of simply erasing the terms depending on fast angles λ, we perform a canonical transformation via Lie Series to kill the terms h 0,0 (λ,ξ,η) λ, h 0,1 (λ,ξ,η) λ, h 0,2 (λ,ξ,η) λ,....
The scheme of the preliminary perturbation reduction We perform a Kolmogorov-like step of normalization. We determine the generating function, χ, by solving the equation n KS χ λ + h 0,j2 j 2=0 λ:k F = 0. where λ:kf means that we keep only the terms depending on λ and at most of degree K F. The parameters K S and K F are chosen so as to include in the secular model the main effects due to the possible proximity to a mean-motion resonance. The transformed Hamiltonian reads H (O2) = expl µχh = j=0 1 j! Lj µχh. This is our Hamiltonian at order two in the masses.
Analytical integration ( ) Secular + NF (r) ( η(0), ξ(0) Φ (r) (0), ϕ (r) (0) ) Numerical integration Φ (r) (t)=φ (r) (0) ϕ (r) (t)= ϕ (r) (0)t+ϕ (r) (0) ( η(t), ξ(t) ) (NF (r) ) 1 ( Φ (r) (t), ϕ (r) (t) )
First order approximation (HD 11506) HD 11506: the system is close to the 7:1 MMR (weak MMR). 0.45 Eccentricities analy_hd_11506_ord1 "./ecc_0.dat" "./ecc_1.dat" "./orbitals_1.dat" u 1:4 "./orbitals_2.dat" u 1:4 0.4 0.35 0.3 0.25 0.2 0 10000 20000 30000 40000 50000 60000
Second order approximation (HD 11506) HD 11506: the system is close to the 7:1 MMR (weak MMR). 0.45 Eccentricities analy_hd_11506_ord2 "./ecc_0.dat" "./ecc_1.dat" "./orbitals_1.dat" u 1:4 "./orbitals_2.dat" u 1:4 0.4 0.35 0.3 0.25 0.2 0 10000 20000 30000 40000 50000 60000
First order approximation (HD 177830) HD 177830: the system is close to the 3:1 and 4:1 MMR. 0.35 0.3 Eccentricities analy_hd_177830_ord1 "./ecc_0.dat" "./ecc_1.dat" "./orbitals_1.dat" u 1:4 "./orbitals_2.dat" u 1:4 0.25 0.2 0.15 0.1 0.05 0 0 10000 20000 30000 40000 50000
Second order approximation (HD 177830) HD 177830: the system is close to the 3:1 and 4:1 MMR. 0.35 0.3 Eccentricities analy_hd_177830_ord2 "./ecc_0.dat" "./ecc_1.dat" "./orbitals_1.dat" u 1:4 "./orbitals_2.dat" u 1:4 0.25 0.2 0.15 0.1 0.05 0 0 10000 20000 30000 40000 50000
First order approximation (HD 108874) HD 108874: the system is close to the 4:1 MMR (strong MMR). 0.35 0.3 Eccentricities analy_hd_108874_ord1 "./ecc_0.dat" "./ecc_1.dat" "./orbitals_1.dat" u 1:4 "./orbitals_2.dat" u 1:4 0.25 0.2 0.15 0.1 0.05 0 10000 20000 30000 40000 50000 60000 70000
Second order approximation (HD 108874) HD 108874: the system is close to the 4:1 MMR (strong MMR). 0.35 0.3 Eccentricities analy_hd_108874_ord2 "./ecc_0.dat" "./ecc_1.dat" "./orbitals_1.dat" u 1:4 "./orbitals_2.dat" u 1:4 0.25 0.2 0.15 0.1 0.05 0 10000 20000 30000 40000 50000 60000 70000
Proximity to a mean-motion resonance We now want to evaluate the proximity to a mean-motion resonance. The idea is to rate the proximity by looking at the canonical change of coordinates induced by the approximation at order two in the masses. ξ j = ξ j µ ( χ = ξ j 1 µ ) χ, η j ξ j η j η j = η j µ ( χ = η j 1 µ ) χ. ξ j η j ξ j In particular, we focus on the coefficients of the terms δξ j = µ ξ j χ η j and δη j = µ η j χ ξ j.
δ-criterion Secular near a MMR MMR System a 1 /a 2 δ HD 11964 0.072 9.897 10 4 sin( λ 1 +2λ 2 ) HD 74156 0.075 9.681 10 4 cos( 4λ 1 λ 2 ) HD 134987 0.140 9.897 10 4 sin( λ 1 +2λ 2 ) HD 163607 0.149 1.376 10 3 cos( 3λ 1 λ 2 ) HD 12661 0.287 1.760 10 3 sin( λ 1 +2λ 2 ) HD 147018 0.124 2.455 10 3 sin( 2λ 1 +λ 2 ) HD 11506 0.263 2.943 10 3 cos( λ 1 7λ 2 ) HD 177830 0.420 2.551 10 3 cos( λ 1 4λ 2 ) HD 9446 0.289 2.328 10 3 sin( 2λ 1 +λ 2 ) HD 169830 0.225 2.316 10 2 cos( λ 1 9λ 2 ) υ Andromedae 0.329 1.009 10 2 cos( λ 1 5λ 2 ) Sun-Jup-Sat 0.546 2.534 10 2 cos(2λ 1 5λ 2 ) HD 108874 0.380 4.314 10 2 sin( λ 1 +4λ 2 ) HD 128311 0.622 6.421 10 1 sin( λ 1 +2λ 2 ) HD 183263 0.347 5.253 10 2 cos( λ 1 5λ 2 )
Results 1 Can we predict the long-term evolution of extrasolar systems? If the system is not too close to a mean-motion resonance, providing an approximation of the motions of the secular variables up to order two in the masses, the secular evolution is well approximated via Birkhoff normal form.
Results 1 Can we predict the long-term evolution of extrasolar systems? If the system is not too close to a mean-motion resonance, providing an approximation of the motions of the secular variables up to order two in the masses, the secular evolution is well approximated via Birkhoff normal form. 2 How can we evaluate the influence of a mean-motion resonance? The secular Hamiltonian at order two in the masses is explicitly constructed via Lie Series, so the generating function contains the information about the proximity to a mean-motion. We introduce an heuristic and quite rough criterion that we think is useful to discriminate between the different behaviors: (i) δ 2.6 10 3 : secular; (ii) 2.6 10 3 < δ 2.6 10 2 : near mean-motion resonance; (iii) δ > 2.6 10 2 : in mean-motion resonance.
Future work... Resonant normal form. Kolmogorov normal form. Add the mutual inclination. Effective stability time. Improve the determination of the orbital parameters.
Thanks for your attention! Questions? Comments?