alphaFeedback

WorksPhysics and Astronomy

Nonintegrability theorem on perturbed integrable Hamiltonian dynamical systems with two independent resonances

AI contributions
Not recorded for this paper
View details

Abstract

In this paper we establish a new theorem for nonintegrability of a perturbed, autonomous, two-degree-of-freedom Hamiltonian dynamical system with two independent resonances, based on a criterion which utilizes the well-known theorem of Poincaré and a Lie-series symplectic transformation. We also make certain assumptions and suggestions as to when the theorem holds and when it does not.

Introduction

A very important qualitative feature of Hamiltonian systems is integrability [8]. According to the Liouville–Arnold theorem ([1], p. 270–272), if an autonomous Hamiltonian system with nn degrees of freedom possesses nn functionally independent first integrals in involution, then every regular, compact, and connected common level set of these integrals is diffeomorphic to an invariant nn-dimensional torus Tn\mathbb{T}^{n}. Moreover, in a neighbourhood of such a torus, action–angle coordinates can be introduced.

Let us begin with a perturbed, autonomous, two-degree-of-freedom system given by a Hamiltonian of the general form. Given that the perturbed Hamiltonian is an analytic function in terms of the parameter ε\varepsilon, on an open interval around ε=0\varepsilon=0, it can consequently be expressed as a power series in ε\varepsilon ([2], p. 40),

H(J1,J2,w1,w2)=H0(J1,J2)+εH1(J1,J2,w1,w2)(1)H(J_{1},J_{2},w_{1},w_{2}) = H_{0}(J_{1},J_{2}) + \varepsilon H_{1}(J_{1},J_{2},w_{1},w_{2}) \tag*{(1)}

The perturbed Hamiltonian is written in curvilinear symplectic coordinates called action-angle variables in an open region of the phase space:

Ji∈S⊂Rn,wi mod⁡2π∈TnJ_{i} \in S \subset\mathbb{R}^{n}, \qquad w_{i}\ \operatorname{mod} 2\pi\in\mathbb{T}^{n}

More specifically, H1(J1,J2,w1,w2)H_{1}(J_{1},J_{2},w_{1},w_{2}) is the perturbation, which is periodic in w1,w2w_{1},w_{2} and can be expanded as a double Fourier series with period 2π2\pi. The function H0(J1,J2)H_{0}(J_{1},J_{2}) is the Hamiltonian of the unperturbed problem. It is Liouville–Arnold integrable, with finite motions; that is, in the corresponding regular open region of phase space, the unperturbed system is locally foliated by invariant nn-dimensional tori. These tori are invariant with respect to the phase flow of the unperturbed system. Furthermore, H0H_{0} satisfies the nondegeneracy condition ([2], p. 40):

det⁡(∂2H0∂Ji∂Jj)≠0,i,j=1,2\det\left(\frac{\partial^{2}H_{0}}{\partial J_{i}\partial J_{j}}\right)\ne0,\qquad i,j=1,2

The motion of a phase point is defined by the solution of the system

J˙i=0,w˙i=ωi(J),ωi(J)=∂H0∂Ji,i=1,2(2)\dot{J}_i = 0,\qquad\dot{w}_i = \omega_i(J),\qquad\omega_i(J) = \frac{\partial H_0}{\partial J_i},\qquad i = 1,2 \tag*{(2)}

That is,

Ji=Ji0,wi=ωit+wi0,i=1,2(3)J_i = J_{i0},\qquad w_i = \omega_i t + w_{i0},\qquad i = 1,2 \tag*{(3)}

Each torus is determined by the value Ji=Ji0J_i = J_{i0}, while wi mod 2πw_i \bmod2\pi are the angular coordinates .

In an abstract nn-dimensional system, a torus Tn\mathbb{T}^n with frequencies

ω=(ω1,…,ωn)\boldsymbol{\omega} = (\omega_1,\ldots,\omega_n)

is called non-resonant if k⋅ω≠0\mathbf{k}\cdot\boldsymbol{\omega} \ne0 ∀k∈Zn∖{0}\forall\mathbf{k} \in\mathbb{Z}^n \setminus\{0\},

and resonant if there exists a nonzero integer vector

k∈Zn∖{0}\mathbf{k} \in\mathbb{Z}^n \setminus\{0\}

such that

k⋅ω=0\mathbf{k}\cdot\boldsymbol{\omega} = 0

Obviously, in a two-dimensional system, a torus T2\mathbb{T}^2 is said to be resonant if

∃(k1,k2)∈Z2∖{(0,0)}:k1ω1+k2ω2=0\exists(k_1,k_2) \in\mathbb{Z}^2 \setminus\{(0,0)\}:\quad k_1\omega_1+k_2\omega_2=0

For k1≠0k_1 \ne0 and ω2≠0\omega_2 \ne0, this gives

ω1ω2=−k2k1∈Q\frac{\omega_1}{\omega_2}=-\frac{k_2}{k_1}\in\mathbb{Q}

Generally, the nondegenerate condition corresponds to the case in which the frequencies are functionally independent. In nondegenerate systems the non-resonant tori form an everywhere dense set of full measure. The resonant tori form a set of measure zero, which is also everywhere dense ([3], p. 273).

A phase trajectory on a non-resonant torus covers the torus densely. One can simply say that when the frequency ratio

ω1ω2∈Q,\frac{\omega_1}{\omega_2}\in\mathbb{Q},

the invariant tori contain periodic orbits that close after a finite time, whereas when the frequency ratio is irrational,

ω1ω2∈R∖Q,\frac{\omega_1}{\omega_2}\in\mathbb{R}\setminus\mathbb{Q},

the invariant tori have quasi-periodic orbits that cover the tori densely.

Poincaré's theorem on nonintegrability

Obviously,

H(J1,J2,w1,w2)H(J_1,J_2,w_1,w_2)

is an integral of motion. Let us suppose that HH has a second integral of motion

Φ(J1,J2,w1,w2,ε)\Phi(J_1,J_2,w_1,w_2,\varepsilon)

in an open neighborhood of J1,J2J_1,J_2, for w1,w2w_1,w_2 from 0 to 2π2\pi, and for ε\varepsilon in a neighborhood of ε=0\varepsilon=0. Then Φ\Phi can be expanded in a convergent series around ε=0\varepsilon=0:

Φ=Φ0(J1,J2)+εΦ1(J1,J2,w1,w2)+ε2Φ2+O(ε3)(4)\Phi=\Phi_0(J_1,J_2)+\varepsilon\Phi_1(J_1,J_2,w_1,w_2)+\varepsilon^2\Phi_2+\mathcal{O}(\varepsilon^3) \tag*{(4)}

If we take a non-resonant torus, note that Φ0\Phi_0 is independent of the angles. A detailed explanation can be found in ([3] p. 353).

Poincaré proves that Φ0\Phi_0 is an integral of H0H_0, independent of it. Furthermore, up to order ε\varepsilon, since Φ\Phi is an integral of HH, it holds that

[Φ,H]=0[\Phi,H]=0

That leads to

[Φ0,H0]=0,[Φ0,H1]+[Φ1,H0]=0(5)[\Phi_0,H_0]=0,\qquad[\Phi_0,H_1]+[\Phi_1,H_0]=0 \tag*{(5)}

The first relation in (5) states that Φ0\Phi_0 is an integral of H0H_0. We shall examine whether Φ0\Phi_0 is independent of H0H_0. The second relation in (5) yields the relation between the Fourier coefficients Bk1,k2B_{k_1,k_2} and Ak1,k2A_{k_1,k_2}:

Bk1,k2(k1∂H0∂J1+k2∂H0∂J2)=Ak1,k2(k1∂Φ0∂J1+k2∂Φ0∂J2)(6)B_{k_1,k_2}\left(k_1\frac{\partial H_0}{\partial J_1}+k_2\frac{\partial H_0}{\partial J_2}\right)=A_{k_1,k_2}\left(k_1\frac{\partial\Phi_0}{\partial J_1}+k_2\frac{\partial\Phi_0}{\partial J_2}\right) \tag*{(6)}

Here,

H1=∑k1,k2Ak1,k2(J1,J2)exp⁡(i(k1w1+k2w2))H_1=\sum_{k_1,k_2}A_{k_1,k_2}(J_1,J_2)\exp\left(i(k_1w_1+k_2w_2)\right)

and

Φ1=∑k1,k2Bk1,k2(J1,J2)exp⁡(i(k1w1+k2w2))\Phi_1=\sum_{k_1,k_2}B_{k_1,k_2}(J_1,J_2)\exp\left(i(k_1w_1+k_2w_2)\right)

are the Fourier series of H1H_1 and Φ1\Phi_1 in w1w_1 and w2w_2.

Relation (6) is an identity on the phase space; that is, it holds for every Ji,wiJ_i,w_i in an open region d×Tnd\times\mathbb{T}^n, at least ([2] p. 47).

According to equation (6), its left-hand side becomes zero on the dense set of resonant tori of the unperturbed system H0H_0, since it is nondegenerate:

∂H0∂J1∂H0∂J2=ω1ω2=−k2k1(7)\frac{\dfrac{\partial H_0}{\partial J_1}}{\dfrac{\partial H_0}{\partial J_2}}=\frac{\omega_1}{\omega_2}=-\frac{k_2}{k_1} \tag*{(7)}

If an infinite set of Fourier coefficients Ak1,k2≠0A_{k_1,k_2} \ne0 of H1H_1 are different from zero on these dense resonant tori, then according to equation (6),

∂Φ0∂J1/∂Φ0∂J2=−k2k1(8)\frac{\partial\Phi_0}{\partial J_1}\bigg/\frac{\partial\Phi_0}{\partial J_2}=-\frac{k_2}{k_1} \tag*{(8)}

If we combine (7) and (8), we get

D(J)=∂(Φ0,H0)∂(J1,J2)=0(9)D(J)=\frac{\partial(\Phi_0,H_0)}{\partial(J_1,J_2)}=0 \tag*{(9)}

Therefore, Φ0\Phi_0 is functionally dependent on H0H_0, which is a contradiction. We point out that (9) holds for the entire domain of the phase space. Since DD is continuous, if it is zero on a dense set, then it is zero everywhere. Indeed, take any J∗J^* in the domain. Since the resonant set of tori is dense, there exists a sequence

J(n)→J∗J^{(n)} \to J^*

such that

D(J(n))=0.D\left(J^{(n)}\right)=0.

Therefore,

D(J∗)=lim⁡n→∞D(J(n))=0,D(J^*)=\lim_{n\to\infty}D\left(J^{(n)}\right)=0,

so (9) holds throughout the entire domain. Thus, Poincaré’s theorem [7] states:

If an infinite set of Fourier coefficients Ak1,k2≠0A_{k_1,k_2} \ne0 of H1H_1 are nonzero on a dense set of resonant tori of the unperturbed, integrable, nondegenerate system, then HH does not possess a second integral of motion analytic in the real variables in an open neighborhood of J1,J2J_1,J_2, for w1,w2w_1,w_2 from 0 to 2π2\pi, and for ε\varepsilon in a neighborhood of ε=0\varepsilon=0.

Meletlidou–Ichtiaroglou nonintegrability criterion and equivalence to Poincaré's theorem on nonintegrability

A more detailed presentation is given in [4]. From the second relation in (5), we obtain

[Φ0,H1]=[H0,Φ1],[\Phi_0,H_1]=[H_0,\Phi_1],

and therefore

dΦ1dt=[Φ0,H1](10)\frac{\mathrm{d}\Phi_1}{\mathrm{d}t}=[\Phi_0,H_1] \tag*{(10)}

Consider a member of the family of resonant tori for which

ω1ω2=nm,n,m∈Z∖{0}.\frac{\omega_1}{\omega_2}=\frac{n}{m}, \qquad n,m\in\mathbb{Z}\setminus\{0\}.

By integrating (10) over one period, we obtain the following. The right-hand side of this equation gives zero upon integration, since the corresponding term is the time derivative of Φ1\Phi_1 integrated over one period TT on the resonant torus.

The period TT is determined by

T=2πΩ,Ω=ω1n=ω2mT = \frac{2\pi}{\Omega}, \qquad\Omega= \frac{\omega_1}{n} = \frac{\omega_2}{m}

The left-hand side results in

(k2∂Φ0∂J1+k1∂Φ0∂J2)∂∫0TH1 dt∂φ=0(11)\left(k_2\frac{\partial\Phi_0}{\partial J_1}+k_1\frac{\partial\Phi_0}{\partial J_2}\right)\frac{\partial\int_0^T H_1\,\mathrm{d}t}{\partial\varphi}=0 \tag*{(11)}

where

φ=k2w1+k1w2\varphi=k_2w_1+k_1w_2

We note that every orbit on a resonant torus is periodic and differs from the others only in terms of w10w_{10}, w20w_{20}. Thus, we obtain an entire family of periodic orbits on a specific resonant torus.

Along the unperturbed motion,

dφdt=k2w˙1+k1w˙2=0.\frac{\mathrm{d}\varphi}{\mathrm{d}t}=k_2\dot{w}_1+k_1\dot{w}_2=0.

Therefore, φ\varphi is constant on every periodic orbit. On a resonant torus of H0H_0 there are infinitely many periodic orbits parametrized by φ\varphi.

The mean value of H1H_1 is

⟨H1⟩(φ)=1T∫0TH1(t;φ) dt\langle H_1\rangle(\varphi)=\frac{1}{T}\int_0^T H_1(t;\varphi)\,\mathrm{d}t

It can either be the same for all periodic orbits of the torus, or it can vary as φ\varphi changes. In the second case, there exists a trajectory for which

∂∫0TH1 dt∂φ≠0\frac{\partial\int_0^T H_1\,\mathrm{d}t}{\partial\varphi}\ne0

On a resonant torus, this leads to

∂Φ0∂J1∂Φ0∂J2=k1k2\frac{\dfrac{\partial\Phi_0}{\partial J_1}}{\dfrac{\partial\Phi_0}{\partial J_2}}=\frac{k_1}{k_2}

This relation concerns the entire resonant torus. It means that Φ0\Phi_0 depends on H0H_0 on this torus. This occurs for a dense set of resonant tori if

∂∫0TH1 dt∂φ≠0\frac{\partial\int_0^T H_1\,\mathrm{d}t}{\partial\varphi}\ne0

on this dense set. But this is not possible according to Poincaré’s theorem, since Φ0\Phi_0 is independent of H0H_0. Therefore, HH is nonintegrable.

More specifically, the exact proposition of [4] is that if there exists a dense set of resonant tori of H0H_0 such that, on each of them, the average value of H1H_1 depends on the particular periodic orbit, then the perturbed Hamiltonian is nonintegrable for ε∈(−ε0,ε0)∖{0}\varepsilon\in(-\varepsilon_0,\varepsilon_0)\setminus\{0\} in an open neighborhood of ε=0\varepsilon=0.

It is proven that the Meletlidou–Ichtiaroglou criterion and Poincaré’s theorem are equivalent.

A new theorem on nonintegrability in the case of two independent resonances

We are going to establish a new theorem on the nonintegrability of a perturbed Hamiltonian system with two resonances. We begin with the Hamiltonian

H=H0(J1,J2)+ε[A1(J1,J2)cos⁡(k1w1−k2w2)+A2(J1,J2)cos⁡(l1w1−l2w2)](12)H = H_0(J_1,J_2) + \varepsilon\left[A_1(J_1,J_2)\cos(k_1w_1-k_2w_2) + A_2(J_1,J_2)\cos(l_1w_1-l_2w_2)\right] \tag*{(12)}

Here J1,J2,w1,w2J_1,J_2,w_1,w_2 are the action-angle variables, H0H_0 is the integrable, nondegenerate, unperturbed part in action-angle variables, and H1H_1 is the perturbation with two resonances, k1,k2,l1,l2∈Zk_1,k_2,l_1,l_2 \in\mathbb{Z}. and arbitrary. The vectors

k=(k1,−k2),l=(l1,−l2)\mathbf{k}=(k_1,-k_2), \qquad\mathbf{l}=(l_1,-l_2)

are independent.

We suppose, as in Poincaré’s theorem of nonintegrability, that there exists a second integral of motion independent of HH, and we are going to use Poincaré’s lemma that Φ0\Phi_0 is independent of H0H_0. To study the nonintegrability generated by the interaction of these two resonances, we apply a sequence of Lie-series symplectic transformations ([6]).

The Lie operator is defined by the Poisson bracket:

LWF=[H,F].L_WF=[H,F].

Our Lie generators Wmk,nkW_{m_k,n_k} is a sequence of function in angle variables

Wmk,nk=cos⁡(mkw1−nkw2),(13)W_{m_k,n_k}=\cos(m_kw_1-n_kw_2), \tag*{(13)}

where

mk=pkk1+qkl1,nk=jkk2+skl2,m_k=p_kk_1+q_kl_1, \qquad n_k=j_kk_2+s_kl_2,

with

pk,qk,jk,sk,k1,k2,l1,l2∈Z−{0},k∈N−{0}i=1,2,p_k,q_k,j_k,s_k,k_1,k_2,l_1,l_2 \in\mathbb{Z}-\{0\}, \quad k\in\mathbb{N}-\{0\} \qquad i=1,2,

Then the transformed Hamiltonian is

H′=exp⁡(εLWmk,nk)H=∑n=0∞εnn!LWmk,nknH==H0+εH1+ε[Wmk,nk,H0]+ε2[Wmk,nk,H1]+ε22[Wmk,nk,[Wmk,nk,H0]]+O(ε3).(14)\begin{aligned} H'=\exp\left(\varepsilon L_{W_{m_k,n_k}}\right)H=\sum_{n=0}^{\infty}\frac{\varepsilon^n}{n!}L_{W_{m_k,n_k}}^nH = \\ &=H_0+\varepsilon H_1+\varepsilon[W_{m_k,n_k},H_0]+\varepsilon^2[W_{m_k,n_k},H_1]+\frac{\varepsilon^2}{2}[W_{m_k,n_k},[W_{m_k,n_k},H_0]]+\mathcal{O}(\varepsilon^3). \tag*{(14)} \end{aligned}

According to Cauchy’s theorem for Lie series, the series for H′H' converges, since it does not have any small divisors[5].

At each step we apply Wmk,nkW_{m_k,n_k} to the initial Hamiltonian HH. At each step we shall prove that eq. (16) is non-zero at the resonant torus with ω1ω2=k2+nkk1+mk\frac{\omega_1}{\omega_2}=\frac{k_2+n_k}{k_1+m_k}. After applying this infinite sequence of Lie generators, we shall prove that (16) is non-zero at the resonant tori ω1ω2=k2+nkk1+mk\frac{\omega_1}{\omega_2}=\frac{k_2+n_k}{k_1+m_k} where pk,qk,jk,skp_k,q_k,j_k,s_k take infinite values and ω1ω2\frac{\omega_1}{\omega_2} at these resonant tori form a dense set in action space.

We also apply the Poisson bracket with Wmk,nkW_{m_k,n_k} to equation (5):

[Wmk,nk,[Φ0,H1]]+[Wmk,nk,[Φ1,H0]]=0[W_{m_k,n_k},[\Phi_0,H_1]]+[W_{m_k,n_k},[\Phi_1,H_0]]=0

Using again Jacobi’s identity for

[Wmk,nk,[Φ0,H1]]=[Φ0,[H1,Wmk,nk]]+[[Wmk,nk,Φ0],H1][W_{m_k,n_k},[\Phi_0,H_1]]=[\Phi_0,[H_1,W_{m_k,n_k}]]+[[W_{m_k,n_k},\Phi_0],H_1]

and

[Wmk,nk,[Φ1,H0]]=[Φ1,[Wmk,nk,H0]]+[[Wmk,nk,Φ1],H0],[W_{m_k,n_k},[\Phi_1,H_0]]=[\Phi_1,[W_{m_k,n_k},H_0]]+[[W_{m_k,n_k},\Phi_1],H_0],

we obtain

[Φ0,[Wmk,nk,H1]]+[[Wmk,nk,Φ0],H1]+[Φ1,[Wmk,nk,H0]]+[[Wmk,nk,Φ1],H0]=0(15)[\Phi_0,[W_{m_k,n_k},H_1]]+[[W_{m_k,n_k},\Phi_0],H_1]+[\Phi_1,[W_{m_k,n_k},H_0]]+[[W_{m_k,n_k},\Phi_1],H_0]=0 \tag*{(15)}

According to the criterion by E. Meletlidou and S. Ichtiaroglou, we integrate (16) on the periodic orbits of the unperturbed invariant resonant tori of H0H_0, where the resonance condition holds:

ω1k2+nk=ω2k1+mk\frac{\omega_1}{k_2+n_k}=\frac{\omega_2}{k_1+m_k}

And also

ω1l2+nk=ω2l1+mk\frac{\omega_1}{l_2+n_k}=\frac{\omega_2}{l_1+m_k}

The results of the analysis using (α)(\alpha) also hold for (β)(\beta). Let us consider the resonance condition (α)(\alpha). The period TT is

T=2πΩ,Ω=ω1k2(1+jk)+skl2=ω2k1(1+pk)+qkl1T=\frac{2\pi}{\Omega},\qquad\Omega=\frac{\omega_1}{k_2(1+j_k)+s_kl_2}=\frac{\omega_2}{k_1(1+p_k)+q_kl_1}

We integrate over one period:

∫0T([Wmk,nk,[Φ0,H1]]+[Wmk,nk,[Φ1,H0]]) dt\int_0^T\left([W_{m_k,n_k},[\Phi_0,H_1]]+[W_{m_k,n_k},[\Phi_1,H_0]]\right)\,\mathrm{d}t

Thus,

∫0T([Φ0,[Wmk,nk,H1]]+[[Wmk,nk,Φ0],H1]+[Φ1,[Wmk,nk,H0]]+[[Wmk,nk,Φ1],H0]) dt=0(16)\int_0^T\left([\Phi_0,[W_{m_k,n_k},H_1]]+[[W_{m_k,n_k},\Phi_0],H_1]+[\Phi_1,[W_{m_k,n_k},H_0]]+[[W_{m_k,n_k},\Phi_1],H_0]\right)\,\mathrm{d}t=0 \tag*{(16)}

Consider

w1=ω1t+w10,w2=ω2t+w20,w_1=\omega_1t+w_{10},\qquad w_2=\omega_2t+w_{20},

with frequencies of the form

ω1=[k2(1+jk)+skl2]Ω,\omega_1=[k_2(1+j_k)+s_kl_2]\Omega,
ω2=[k1(1+pk)+qkl1]Ω\omega_2=[k_1(1+p_k)+q_kl_1]\Omega

For the first term of eq. (16) which is

∫0T[Φ0,[W,H1]] dt\int_{0}^{T} [\Phi_0,[W,H_1]]\,dt

and selecting the term cos⁡((k1+m1)ω1−(k2+m2)ω2)\cos((k_1+m_1)\omega_1-(k_2+m_2)\omega_2) as a general term, we obtain

12(∂A1∂J2nk−∂A1∂J1mk)(∂Φ0∂J2(k2+nk)−∂Φ0∂J1(k1+mk))⋅⋅∫0Tcos⁡((k1+mk)ω1−(k2+nk)ω2) dt.\begin{aligned} \frac{1}{2}\left(\frac{\partial A_1}{\partial J_2}n_k-\frac{\partial A_1}{\partial J_1}m_k\right)\left(\frac{\partial\Phi_0}{\partial J_2}(k_2+n_k)-\frac{\partial\Phi_0}{\partial J_1}(k_1+m_k)\right)\cdot\\ &\cdot\int_{0}^{T}\cos((k_1+m_k)\omega_1-(k_2+n_k)\omega_2)\,dt. \end{aligned}

Provided that Φ0\Phi_0 is independent of H0H_0, the term

(k2+nk)∂Φ0∂J2−(k1+mk)∂Φ0∂J1(k_2+n_k)\frac{\partial\Phi_0}{\partial J_2}-(k_1+m_k)\frac{\partial\Phi_0}{\partial J_1}

is different from zero

Also the other coefficient is nonzero if

nkmk≠∂A1∂J1∂A1∂J2\frac{n_k}{m_k}\ne\frac{\frac{\partial A_1}{\partial J_1}}{\frac{\partial A_1}{\partial J_2}}

as well as

The integral is nonzero since, on the above resonance (α)(\alpha),

∫0Tcos⁡[(k1+mk)ω1−(k2+nk)ω2] dt=∫0Tcos⁡ϕk dt=Tcos⁡ϕk.(17)\int_{0}^{T}\cos[(k_1+m_k)\omega_1-(k_2+n_k)\omega_2]\,dt=\int_{0}^{T}\cos\phi_k\,dt=T\cos\phi_k. \tag*{(17)}

where ϕk=(k1+mk)ω10−(k2+nk)ω20\phi_k=(k_1+m_k)\omega_{10}-(k_2+n_k)\omega_{20} since the resonance does not vanish the initial conditions ωi0\omega_{i0}.

Looking at the last term of (17), it can be proven that it is zero, since it is the time derivative of [Wmk,nk,Φ1][W_{m_k,n_k},\Phi_1]:

∫0T[[Wmk,nk,Φ1],H0] dt=∫0T(∂[Wmk,nk,Φ1]∂ω1∂H0∂J1+∂[Wmk,nk,Φ1]∂ω2∂H0∂J2) dt==∫0T[∂[Wmk,nk,Φ1]∂ω1ω˙1+∂[Wmk,nk,Φ1]∂ω2ω˙2] dt.\begin{aligned} \int_{0}^{T}[[W_{m_k,n_k},\Phi_1],H_0]\,dt=\int_{0}^{T}\left(\frac{\partial[W_{m_k,n_k},\Phi_1]}{\partial\omega_1}\frac{\partial H_0}{\partial J_1}+\frac{\partial[W_{m_k,n_k},\Phi_1]}{\partial\omega_2}\frac{\partial H_0}{\partial J_2}\right)\,dt= \\ &=\int_{0}^{T}\left[\frac{\partial[W_{m_k,n_k},\Phi_1]}{\partial\omega_1}\dot{\omega}_1+\frac{\partial[W_{m_k,n_k},\Phi_1]}{\partial\omega_2}\dot{\omega}_2\right]\,dt. \end{aligned}

That is,

∫0Td[Wmk,nk,Φ1]dt dt=0\int_{0}^{T}\frac{d[W_{m_k,n_k},\Phi_1]}{dt}\,dt=0

The other terms are

∫0T([[Wmk,nk,Φ0],H1]+[Φ1,[Wmk,nk,H0]]) dt(18)\int_{0}^{T}\left([[W_{m_k,n_k},\Phi_0],H_1]+[\Phi_1,[W_{m_k,n_k},H_0]]\right)\,dt \tag*{(18)}

The second term is

∫0T[[Wmk,nk,Φ0],H1] dt\int_{0}^{T}[[W_{m_k,n_k},\Phi_0],H_1]\,dt

and has the coefficient

12[∂A1∂J1mk(∂Φ0∂J2nk−∂Φ0∂J1mk)+∂A1∂J2nk(∂Φ0∂J1mk−∂Φ0∂J2nk)+A1(−∂2Φ0∂J1∂J2k1k2(pk+jk)+k1l2sk+k2l1qk+k1∂2Φ0∂J12mk+k2∂2Φ0∂J22nk)]\frac{1}{2}\left[\frac{\partial A_1}{\partial J_1}m_k\left(\frac{\partial\Phi_0}{\partial J_2}n_k-\frac{\partial\Phi_0}{\partial J_1}m_k\right)+\frac{\partial A_1}{\partial J_2}n_k\left(\frac{\partial\Phi_0}{\partial J_1}m_k-\frac{\partial\Phi_0}{\partial J_2}n_k\right)+A_1\left(-\frac{\partial^2\Phi_0}{\partial J_1\partial J_2}k_1k_2(p_k+j_k)+k_1l_2s_k+k_2l_1q_k+k_1\frac{\partial^2\Phi_0}{\partial J_1^2}m_k+k_2\frac{\partial^2\Phi_0}{\partial J_2^2}n_k\right)\right]

It can be proven that this it is nonzero when H0H_0 does not have a linear integral in J1,J2J_1,J_2, namely Φ0\Phi_0, that is continued in the perturbed system. Also, the integral (18) and the last term in (19) at the resonance is nonzero.

Also, the terms do not balance.

We shall prove in Appendix A. that the ratios

ω1ω2=k2+nkk1+mk\frac{\omega_1}{\omega_2}=\frac{k_2+n_k}{k_1+m_k}

and

ω1ω2=l2+nkl1+mk\frac{\omega_1}{\omega_2}=\frac{l_2+n_k}{l_1+m_k}

form a dense set in the relevant interval. Therefore, the resonant tori form a dense set in the phase space. After proving that the resonant tori are dense in the action space, the Hamiltonian given by (12) is nonintegrable

For HH to possess a second independent first integral analytic in ε\varepsilon, equation (17) must vanish. Hence, the vanishing of (17) is a necessary condition for the existence of such an integral. Under the assumptions stated below, equation (17) is non-zero on a dense set of resonant tori, which yields a contradiction.

Finally,according to the criterion, the Hamiltonian system given by (12) is nonintegrable, except in the cases described above.

We formulate the following theorem:

Theorem: Consider the analytic autonomous Hamiltonian system

H(J,w;ε)=H0(J1,J2)+ε[A1(J1,J2)cos⁡(k1w1−k2w2)+A2(J1,J2)cos⁡(l1w1−l2w2)],H(J,w;\varepsilon)=H_0(J_1,J_2)+\varepsilon\left[A_1(J_1,J_2)\cos(k_1w_1-k_2w_2)+A_2(J_1,J_2)\cos(l_1w_1-l_2w_2)\right],

where H0H_0 is Liouville-integrable and non-degenerate, and

k=(k1,−k2),l=(l1,−l2)\mathbf{k}=(k_1,-k_2),\qquad\mathbf{l}=(l_1,-l_2)

are linearly independent integer vectors.

Let

Wmk,nk=cos⁡(mkw1−nkw2),W_{m_k,n_k}=\cos(m_kw_1-n_kw_2),

with

mk=pkk1+qkl1,nk=jkk2+skl2,m_k=p_kk_1+q_kl_1,\qquad n_k=j_kk_2+s_kl_2,

where

pk,qk,jk,sk∈Z−{0},k∈N−{0}.p_k,q_k,j_k,s_k\in\mathbb{Z}-\{0\},\qquad k\in\mathbb{N}-\{0\}.

Assume that the corresponding resonance ratios

ω1ω2=k2+nkk1+mk(19)\frac{\omega_1}{\omega_2}=\frac{k_2+n_k}{k_1+m_k} \tag*{(19)}

or, analogously,

ω1ω2=l2+nkl1+mk\frac{\omega_1}{\omega_2}=\frac{l_2+n_k}{l_1+m_k}

generate a dense set of resonant tori in the considered action domain. Assume further that, on this dense resonant set, the second-order equation (16) does not vanish. In particular, either for the relevant resonant contribution,

nk∂A1∂J2−mk∂A1∂J1≠0,n_k\frac{\partial A_1}{\partial J_2}-m_k\frac{\partial A_1}{\partial J_1}\ne0,

with analogous non-vanishing condition for A2A_2, or H0H_0 does not possess a linear integral with respect to J1,J2J_1,J_2 or that the remaining averaged Poisson-bracket contributions do not cancel each other we proved that for sufficiently small

ε≠0,\varepsilon\ne0,

the perturbed Hamiltonian HH admits no second first integral which is functionally independent of HH and analytic in ε\varepsilon in the considered open domain that is continued in the perturbed system. Consequently HH is nonintegrable.

Remark: The theorem is inconclusive in cases:

  1. Where all second-order resonant coefficient vanishes, as well as the coefficients,

nk∂Ai∂J2−mk∂Ai∂J1=0,i=1,2,k∈N−{0}n_k\frac{\partial A_i}{\partial J_2}-m_k\frac{\partial A_i}{\partial J_1}=0,\quad i=1,2,k\in\mathbb{N}-\{0\}
  1. The two Fourier vectors are not independent, i.e.

k∥l;\mathbf{k}\parallel\mathbf{l};

then it is known that the system is integrable.

Failure of any of these hypotheses does not imply integrability; it only means that the present criterion is inconclusive.

Conclusions

In this work we have investigated a two-degree-of-freedom nearly integrable Hamiltonian system whose perturbation contains two linearly independent Fourier modes. By combining a Lie-series symplectic transformation with the non-integrability approach of Poincaré and the criterion of Meletlidou–Ichtiaroglou, we obtain a second-order resonant obstruction to the existence of an additional analytic first integral.

The Lie transformation generates combination harmonics at order O(ε2)\mathcal{O}(\varepsilon^2). When one of these harmonics becomes resonant with an unperturbed invariant torus, its average over the corresponding periodic orbit does not, in general, vanish and depends on the phase labelling the periodic orbit on that torus.

If this non-vanishing, orbit-dependent averaged contribution occurs on a dense set of resonant tori, and if the remaining second-order averaged terms do not cancel it, the necessary condition for the existence of a second first integral analytic in ε\varepsilon is violated. Consequently, for sufficiently small ε≠0\varepsilon\ne0, no second first integral functionally independent of HH and analytic in the stated class can exist in the considered domain.

Thus, the obstruction is not produced merely by the presence of two independent Fourier modes, but by their second-order interaction together with the density and non-vanishing conditions stated in the theorem.

The Lie-series formulation is particularly convenient for this analysis because it preserves the canonical structure and provides an explicit order-by-order description of the harmonics generated by the interaction of the perturbing terms.

Dense set of Resonant Tori

We prove that the resonance ratios

k2+nkk1+mk\frac{k_{2}+n_{k}}{k_{1}+m_{k}}

form a dense set in the relevant positive frequency-ratio interval.

Recall that

mk=pkk1+qkl1,nk=jkk2+skl2,m_{k}=p_{k}k_{1}+q_{k}l_{1}, \qquad n_{k}=j_{k}k_{2}+s_{k}l_{2},

where

pk,qk,jk,sk∈Z∖{0}.p_{k},q_{k},j_{k},s_{k}\in\mathbb{Z}\setminus\{0\}.

Let

d1=gcd⁡(k1,l1),d2=gcd⁡(k2,l2).d_{1}=\gcd(k_{1},l_{1}), \qquad d_{2}=\gcd(k_{2},l_{2}).

By Bézout’s identity,

{pk1+ql1:p,q∈Z}=d1Z,\{pk_{1}+ql_{1}:p,q\in\mathbb{Z}\}=d_{1}\mathbb{Z},

and similarly,

{jk2+sl2:j,s∈Z}=d2Z.\{jk_{2}+sl_{2}:j,s\in\mathbb{Z}\}=d_{2}\mathbb{Z}.

Therefore mkm_{k} and nkn_{k} may attain arbitrary integer multiples of d1d_{1} and d2d_{2}, respectively.

Consider an arbitrary open interval

(α,β)⊂R+,0<α<β.(\alpha,\beta)\subset\mathbb{R}^{+}, \qquad0<\alpha<\beta.

Since mkm_{k} may take arbitrarily large values, we can choose mkm_{k} such that

Dk:=k1+mk>0D_{k}:=k_{1}+m_{k}>0

and

d2Dk<β−α.(20)\frac{d_{2}}{D_{k}}<\beta-\alpha. \tag*{(20)}

Now choose nkn_{k} such that

Nk:=k2+nkN_{k}:=k_{2}+n_{k}

which is a large enough multiple of d2d_{2} satisfying

Nk>αDk⇒NkDk>αN_{k}>\alpha D_{k}\Rightarrow\frac{N_{k}}{D_{k}}>\alpha

Therefore the set of integers

B={nk:k∈N, NkDk>α}B=\left\{n_{k}:k\in\mathbb{N},\ \frac{N_{k}}{D_{k}}>\alpha\right\}

is nonempty. In the above set there is a minimum integer nkn_{k} such that Sk=min⁡nk, nk∈BS_{k}=\min n_{k},\ n_{k}\in B for which it holds that Sk∈BS_k \in B.

We will show that for SkS_k it holds that

SkDk<β\frac{S_k}{D_k} < \beta

Let us suppose the opposite

SkDk≥β\frac{S_k}{D_k} \ge\beta

then,

SkDk−d2Dk≥β−d2Dk>β−(β−α)=α\frac{S_k}{D_k} - \frac{d_2}{D_k} \ge\beta- \frac{d_2}{D_k} > \beta- (\beta- \alpha) = \alpha

which is a contradiction, so

SkDk<β\frac{S_k}{D_k} < \beta

since SkS_k is the minimum of the set BB, and DkD_k may attain arbitrary large values,

α<SkDk<β\alpha< \frac{S_k}{D_k} < \beta

The interval (α,β)(\alpha, \beta) is arbitrary, the set

{k2+nkk1+mk}\left\{ \frac{k_2+n_k}{k_1+m_k} \right\}

is dense in R+\mathbb{R}^{+}.

The same argument applies to the second family

{l2+nkl1+mk},\left\{ \frac{l_2+n_k}{l_1+m_k} \right\},

and therefore it is also dense in R+\mathbb{R}^{+}.

Finally, since the unperturbed Hamiltonian H0H_0 is non-degenerate, the map

J↦ω(J)=∇H0(J)J \mapsto\omega(J) = \nabla H_0(J)

is locally invertible. Hence the the above rational frequency ratios imply that the corresponding resonant invariant tori form a dense set in the considered regular action domain.

References

  1. [4]Meletlidou, E. and Ichtiaroglou, S. (1994). “A criterion for non-integrability based on Poincaré’s theorem.” Physica D: Nonlinear Phenomena, 71(3), 261–268. ISSN 0167-2789. https://doi.org/10.1016/0167-2789(94)90148-1
  2. [5]Giorgilli, A. (2002). “Notes on Exponential Stability of Hamiltonian Systems.” Centro di Ricerca Matematica Ennio De Giorgi.
  3. [7]H.Poincaré, 1899, “Les Méthodes Nouvelles de la Mécanique Céleste”, Vol III Gauthier- Vil-lars, Paris, (“New Methods of Celestial Mechanics”, Vol.III, Ed. D.Goroff).
  4. [8]Gaetano Vilasi, 2001, “Hamiltonian Dynamics”, World Scientific.

Paper details

vlnc:W_2yc36cf6
v1 · latestHistory (1)

published · Fresh — added to the Hub within the last 30 days

added ·

Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

CC BY 4.0

Original source

  • Physics and Astronomy

Not recorded for this paper · View statement →

Looking for related papers…
Contents