#### Abstract

This work extends the high-order Melnikov method established by FJ Chen and QD Wang to heteroclinic orbits, and it is used to prove, under a certain class of perturbations, the heteroclinic orbit in a planar vector field that remains unbroken. Perturbations which have this property together form the heteroclinic persisting space. The Van der Pol system is analysed as an application.

#### 1. Introduction

Since the pioneering work of Poincaré on celestial mechanics [1, 2], homoclinic tangles created by transversal intersection of the stable and unstable manifolds have been playing a prominent role in the progress of chaos theory. In 1967, Smale observed that horseshoe chaos occurs in all homoclinic tangles [3]. For the past few years, the studies in [4–7] presented that periodic sinks and H nonlike attractors are also residing in homoclinic and heteroclinic tangles. The H nonlike attractors provide observable chaos in the sense of Sinai–Ruelle–Bowen (SRB) measure [8, 9].

In general, the standard Melnikov function is enough to verify whether the homoclinic tangle or heteroclinic tangle occurs [10, 11]. However, there exist perturbations which make the standard Melnikov function identically equal to zero. In such a case, no simple zero exists. As a result, we need to calculate high-order Melnikov functions for further investigation. There have been many previous works to refer high-order Melnikov method (see [12–19]). We focus on the method presented in [13] and extends the high-order Melnikov functions for homoclinic orbits to heteroclinic orbits.

Naturally, another question arises: does there exist a perturbation such that all the high-order Melnikov functions identically equal to zero? This means the homoclinic or heteroclinic solution remains unbroken. Through the high-order Melnikov method, we prove that the answer of the above question is positive. The union of such perturbations will be defined as heteroclinic persisting space.

This paper is built up as follows: In Section 2, we outline the main results. Section 3 is devoted to focus on the derivation of high-order Melnikov functions for heteroclinic orbits. Then, in Section 4, we prove the heteroclinic orbit remains unbroken under a certain class of autonomous perturbations. Finally, Van der Pol system is considered in Section 5.

#### 2. Statement of Results

We start with a planar system:

Assume that (1) has two hyperbolic saddle fixed points and with a heteroclinic solutionsuch that

The functions and are real analytic near .

Adding time-periodic perturbations to (1) yieldswhere is a small parameter. The periodic functions are real analytic with respect to , and they are terms of second order and higher at and . Let be the unstable manifold of and be the stable manifold of . In general, for . We let be the splitting distance, which can be expressed in the form ofwhere is equivalent to the standard Melnikov function in [10, 11] and , are the high-order Melnikov functions. Moreover, we write , , , , , and for easy of notations. We also introducewhere

In what follows, we give the formulas of and .

Theorem 1. *In (5), we have the following:*(i)*Integral formula for :*(ii)*Integral formula for :*

On the other hand, we consider the question whether there exist any perturbations such that the heteroclinic solution remains unbroken, which implies the splitting distance vanishes identically. By using the high-order Melnikov functions above, such class of perturbations are obtained. Before giving our main theorem, we first introduce some definitions.

*Definition 1. *Let be a smooth manifold and and be vector fields on it. We assume there exists a heteroclinic orbit of and a heteroclinic orbit of , where is a small parameter. We call a heteroclinic persisting perturbation of ifin the sense of Hausdorff limit.

*Definition 2. *All heteroclinic persisting perturbations of are referred to as the heteroclinic persisting space of , which we denote by *.*

It is not difficult to find a heteroclinic persisting perturbation. Let in (4) be satisfyingThen, we have

It follows that is also a heteroclinic solution of (4). By Definition 1, we claim is a heteroclinic persisting perturbation.

However, this heteroclinic persisting perturbation is trivial, since the heteroclinic orbit is just the previous one. We would like to quest for more general ones. For this purpose, we focus on an autonomous perturbed system:

Then, splitting distance (5) can be rewritten aswhere is the -th order Melnikov integral, The following fact is trivial.

The fact is that is a heteroclinic persisting perturbation if and only if for all , where is sufficiently small.

Note that the geometric shape of heteroclinic orbits in planar vector fields are often symmetric which contributes to the computations of high-order Melnikov integrals. We obtain the following theorem.

Theorem 2. *For the given vector field defined by (1), we assume is odd and is even in and is even and is odd in . Then, we have*

The example considered in Section 5 will show that the parity assumption made in Theorem 2 is without loss of generality.

#### 3. High-Order Melnikov Method

This section is intended to derive the high-order Melnikov functions for equation (4). The method here is based on the homoclinic case introduced in [13], but there still exist some nonnegligible differences between them.

##### 3.1. Canonical Equation around the Heteroclinic Orbit

We introduce new phase variables by letting

Differentiating both sides of (16) with respect to gives

By Cramer’s Rule and (4), we havewhere

We suppose to be a stable solution or an unstable solution of system (18) such that and . Let and ; then, is well-defined for all or , and it is the solution of the systemsatisfying . To differ from the other solutions, we call them the primary stable and unstable solutions.

##### 3.2. Integral of Primary Solutions

We letas one more change of variables for system (20). Then, we have

Expanding , , , and at , we obtain

It then follows that

We solve (24) to obtain the primary stable solutionand the primary unstable solutionin which and are the initial values.

Before proceeding further, we need to consider the asymptotic behaviors of and as which are crucial to get the integral equations for primary solutions. Observe thatand due to the local linearization theory, there exists a small number such thatwhere and are terms of second order and higher at and and are terms of second order and higher at . We assume that are the eigenvalues of and are the eigenvalues of . There exist two invertible matrices and such that

In what follows, we introduce two coordinate transformations:

Unperturbed system (1) is transformed to

In the case , our problem is about the unstable solution in a small neighborhood of . Let be sufficiently large. For , we can find a near-identity coordinate transformation:where and are homogeneous polynomials of degree in , such that the unperturbed system near is linearized to

Assume that the power series in (33) are convergent on . It is not difficult to verify thatwhere be such that and . We solve (35) to obtain

On the other hand, let . The problem is then about the stable solution in a small neighborhood of . We similarly havewhere and are homogeneous polynomials of degree .

Lemma 1. *For the above heteroclinic solution , we have*

*Proof. *We recall thatand letThen,This is to implyAs , we infer from (36) thatAccordingly, we haveRecall (29), and we haveSubstituting (45) into (44) givesAs , we infer from (37) thatAfter similar derivation, we haveThis completes the proof.

By using Lemma 2, one can easily verify that

Observe that and should be bounded as , and the initial values in (25) and (26) are claimed to be

Substituting (50) and (51) into (25) and (26), we have the integral equations of the primary stable solutionand the primary unstable solution

##### 3.3. Derivation of High-Order Melnikov Functions

The properties of here are coincident with the ones in [13]. We have the following power series:

In what follows, we substitute (54) into (52) and (53) to determine and inductively. They arefor . Here, are nonnegative integrals, and , , and . and are coefficient functions of , which have the form of

Here, is a constant depending on the specific term. In the next section, these notations contribute to the proof of Theorem 2.

Recall the definition of , and we have

Then,and for the -th order Melnikov function iswhere and are the values by letting in (56). Particularly, we have

Up to now, the formula of can be obtained.

#### 4. Heteroclinic Persisting Perturbations

This section serves as the proof of Theorem 2. For autonomous perturbations, all the functions above having the variable are now dropped off. Before starting our proof, we need some lemmas.

Lemma 2. *Under the conditions of Theorem 2, we have the following:*(1)*If is odd, then and are even in , while and are odd in *(2)*If is even, then and are odd in , while and are even in .*

*Proof. *We will show the parity of , and the other three can be verified similarly. Recall thatSince is an odd function while is even, then and are even and and are odd. It follows thatDenote . We haveMoreover, the parity in will not change by taking a derivative with respect to . Inductively, we obtainSince differs from by a constant factor, the proof is now completed by letting .

At the end of this section, let us prove Theorem 2 with the help of the preceding preparations.

*Proof of Theorem 2. *To start with, we prove, for all ,By using mathematical induction, we first consider the case . Recall thatwhere is an odd function in and and are even functions in . Moreover, we have