Startseite Periodic or homoclinic orbit bifurcated from a heteroclinic loop for high-dimensional systems
Artikel Open Access

Periodic or homoclinic orbit bifurcated from a heteroclinic loop for high-dimensional systems

  • Bin Long EMAIL logo und Yiying Yang
Veröffentlicht/Copyright: 10. April 2025

Abstract

Consider an autonomous ordinary differential equation in R n , which has a heteroclinic loop. Assume that the heteroclinic loop consists of two degenerate heteroclinic orbits γ 1 , γ 2 and two saddle points with different Morse indices. The degenerate heteroclinic orbit in the sense that variational equation along the heteroclinic orbit γ i has d i ( d i > 1 , i = 1 , 2) linearly independent bounded solutions. By the different Morse indices and d i , the heteroclinic loop is a heterodimensional loop, at the same time, it has high codimension in this situation. Applying Lin’s method to the heteroclinic loop, we derived the bifurcation function. The zeros of this function correspond to the conditions under which periodic or homoclinic orbits can bifurcate from the high-codimension heteroclinic loop in the perturbed system.

MSC 2010: 34C23; 34C25; 34C37

1 Introduction

Homoclinic and heteroclinic bifurcations play an important role in dynamical systems. More and more mathematicians have devoted themselves to study the bifurcation problems of homoclinic or heteroclinic orbits. An overview of homoclinic and heteroclinic bifurcation is given in [1]. A heteroclinic loop consists of two distinct hyperbolic saddle points and two heteroclinic orbits connecting them. When there are two saddle points with different Morse indices, this heteroclinic loop is called a heterodimensional loop. Otherwise, the heteroclinic loop is called a equidimensional loop [2]. Numerical and explicit examples of a heteroclinic loop are given in [3,4].

There is rich and complex recurrent dynamics near homoclinic or heteroclinic orbits. Hence, a central task is to find all orbits that stay near the homoclinic orbits or heteroclinic loop for all times. There are two different approaches to treat those problems. The first approach is to use Poincare or first-return maps. The existence of these special orbits is equivalent to the existence of the fixed points of Poincare or first-return maps. These methods are called geometric approaches. The second is the analytical approach. The core is using Lyapunov-Schmidt reduction. The heart of Lyapunov-Schmidt method is the Fredholm property. Chow et al. studied the persistence of the homoclinic orbit of the Duffing equation by the Fredholm property [5]. Following this work, many people have helped to develop the analytical approach to homoclinic or heteroclinic bifurcation problems. In 1990, Lin investigated the existence of periodic or aperiodic solutions near the heteroclinic chains for systems of ordinary differential equations and delay equations by analytical approach. This method was generalized by Fiedler, Vanderbauwhede, Sandstede, and many others as Lin’s method [6]. The idea of Lin’s method is to construct a sequence of piecewise continuous solutions near the original heteroclinic chain, and the bifurcation function can be obtained from these solutions. If the bifurcation function has zeros, then there exist periodic or aperiodic solutions near the heteroclinic chain. Lin’s method can also be used in discrete dynamical systems, singularly perturbed systems, and numerical computation, cf. [79].

For the periodic or aperiodic solutions bifurcated from homoclinic orbit by analytical approach, refer [1012]. Chow et al. [13] considered the equidimensional heteroclinic loop, which is the equilibria that form the heteroclinic loop that has the same dimension of the unstable manifold. Meanwhile, the authors considered non-degenerate heteroclinic orbit. They used geometric approach to seek homoclinic or periodic orbit bifurcated from a heteroclinic loop. Rademacher [14], studied the homoclinic orbit bifurcated from a heteroclinic loop with one equilibrium and one periodic orbit. They assumed that the unperturbed heteroclinic orbits are one- or two-dimensional. By exponential trichotomy and Lin’ method, they found 1-homoclinic orbits near the heteroclinic loop. Jin et al. [15] considered an equidimensional loop for high-dimensional systems. They used local coordinate systems in a neighborhood of a heteroclinic loop to construct the Poincare maps and the bifurcation equations and then obtained the coexistence and coexistence regions of the 1-homoclinic loop, 1-periodic orbit, 2-homoclinic loop, and 2-periodic orbit near the heteroclinic loop. Zhu and Sun [16] considered the same subject that is homoclinic and periodic orbits bifurcated from the heteroclinic cycle connecting saddle-foci and saddle. Bykov cycle is a special heteroclinic cycle between two hyperbolic equilibria of saddle types p1 and p2, where one of the connections is transverse and isolated. Labouriau and Rodrigues [17] considered a differential equation in a three-dimensional manifold having a heteroclinic cycle that consists of two saddle-foci of different Morse indices whose one-dimensional invariant manifolds coincide and whose two-dimensional invariant manifolds intersect transversely. So, the heteroclinic cycle is defined by the presence of the Bykov cycle. They showed the existence of mixed dynamics in the neighborhood of the Bykov cycle. In the recent work by Knlbloch [18], this subject was extended to higher dimensions using Lin’s method. Long and Xu [19] investigated the persistence of a heterodimensional loop under periodic perturbation. Under some conditions, the perturbed system can have a heteroclinic loop near the unperturbed heterodimensional loop. For more research results regarding the recurrent dynamic near heteroclinic loop, refer [2024].

Based on the above background, we apply Lin’s method to investigate periodic or homoclinic orbits near the heterodimensional loop under periodic perturbation for a high-dimensional system. We consider the following autonomous differential equation:

(1.1) x ˙ ( t ) = f ( x ( t ) )

and its periodic perturbed equation is as follows:

(1.2) x ˙ ( t ) = f ( x ( t ) ) + j = 1 2 μ j g j ( x ( t ) , μ , t ) ,

where x R n , μ = ( μ 1 , μ 2 ) R 2 and we give the following assumptions:

  1. f C 3 .

  2. The unperturbed equation (1.1) has two distinct hyperbolic equilibria P + and P . Namely, f ( P ± ) = 0 and the eigenvalues of D f ( P ± ) lie off the imaginary axis, where D denotes the derivative operator.

  3. The unperturbed equation (1.1) has two heteroclinic solutions γ 1 ( t ) and γ 2 ( t ) , which are asymptotic to equilibria P + and P , respectively. That is, γ i ˙ ( t ) = f ( γ i ( t ) ) , i = 1 , 2 and

    lim t + γ 1 ( t ) = P + , lim t γ 1 ( t ) = P , lim t + γ 2 ( t ) = P , lim t γ 2 ( t ) = P + .

  4. g j C 3 , g j ( P ± , μ , t ) = 0 , g j ( x , 0 , t ) = 0 , and g j ( x , μ , t + 2 ) = g j ( x , μ , t ) .

  5. dim ( W s ( P + ) ) = d + , dim ( W s ( P ) ) = d , where W s ( P + ) and W s ( P ) are the stable manifolds of the equilibria P + and P , respectively.

  6. dim ( T γ 1 ( 0 ) W s ( P + ) T γ 1 ( 0 ) W u ( P ) ) = d 1

    and

    dim ( T γ 2 ( 0 ) W s ( P ) T γ 2 ( 0 ) W u ( P + ) ) = d 2 ,

    where T γ i ( 0 ) W s u ( P ± ) is the tangent spaces of the corresponding invariant manifolds at γ i ( 0 ) and d i > 1 , i = 1 , 2.

From ( H 3 ) , we know that the unperturbed equation (1.1) has a heteroclinic loop Γ . Refer Figure 1, where

Γ = { P } { γ 1 ( t ) : t R } { P + } { γ 2 ( t ) : t R } .

Figure 1 
               Heteroclinic loop 
                     
                        
                        
                           Γ
                        
                        \Gamma 
                     
                  .
Figure 1

Heteroclinic loop Γ .

From ( H 5 ) , the hyperbolic equilibria P + and P can have different saddle point indices or Morse indices when d + d [14,18]. Long and Yang [25] defined the splitting index of heteroclinic orbit in the unperturbed heteroclinic chain. Analogously, we define splitting index of heteroclinic orbit γ i as S ( γ i ) , which are expressed as follows:

S ( γ 1 ) = d + d = s , S ( γ 2 ) = d d + = s .

The variational equation of (1.1) along the heteroclinic orbit γ i is:

(1.3) u ˙ ( t ) = D f ( γ i ( t ) ) u ( t ) .

Since γ i ( t ) is the heteroclinic solution of equation (1.1), γ i ( t ) is situated at the intersection of the stable manifold and the unstable manifold. From ( H 6 ) , we know equation (1.3) has d i ( d i > 1 ) linearly independent bounded solutions, i = 1 , 2. And the dimension of the intersection of the correspondence stable manifold and unstable manifold is d i . If the intersection is non-transversal, the bifurcation phenomenon will occur in the original heteroclinic loop under periodic perturbation. By the definition of the splitting index of heteroclinic orbit γ i and a simple calculation, we have the codimension of γ i as

codim ( γ i ) = n dim ( T γ i ( 0 ) W s u ( P + ) + T γ i ( 0 ) W u s ( P ) ) = n dim ( T γ i ( 0 ) W s u ( P + ) ) dim ( T γ i ( 0 ) W u s ( P ) ) + dim ( T γ i ( 0 ) W s u ( P + ) ) ( T γ i ( 0 ) W u s ( P ) ) = d i + ( 1 ) i s ,

for i = 1 , 2. When d i + ( 1 ) i s > 0 , the intersection of the stable and unstable manifolds is non-transversal, i = 1 , 2. The heteroclinic loop has high codimension in this situation. Therefore, under small perturbation, the heteroclinic loop is broken. So, in this study, we are mainly concerned with the heterodimensional loop bifurcation of high codimension under periodic perturbation. We will apply Lin’s method to construct the periodic solution and homoclinic solution near the unperturbed heteroclinic loop.

The structure of this study is organized as follows. In Section 2, we first study the variational equation of (1.1) along the degenerate heteroclinic orbit γ i and establish two-side exponential dichotomies. We introduce some notations and present the main result. In Section 3, we provide the proof of the main result. First, we study the existence of periodic solution near the heteroclinic loop Γ in Section 3.1. In Section 3.2, we construct the homoclinic solution near the heteroclinic loop Γ . Hence, under some conditions, periodic or homoclinic solution can bifurcated from heteroclinic loop Γ under periodic perturbation.

2 Preliminaries and main result

2.1 Lin’s method

We give a brief description of the idea of Lin’s method in this section, refer [6] for details. Lin’s method is an analytical tool to deal with heteroclinic loop bifurcation.

We assume that for μ = 0 , a system in R n has saddle points P , P + and the heteroclinic orbits are asymptotic to P + and P , respectively. The system is as follows:

(2.1) x ( t ) = f ( x ( t ) , μ ) .

By the assumptions, we know that the system has a heteroclinic loop consisting of saddles P , P + and heteroclinic solutions γ 1 ( t ) , γ 2 ( t ) .

We apply Lin’s method to construct Lin orbits x 1 , x 2 that are composed of piecewise continuous orbits x 1,1 , x 1,2 , x 2,1 , x 2,2 near the heteroclinic loop, which is characterized by a gap on t = 0 in a distinguished direction. When gaps ζ i disappear, Lin orbits can glue together. So, we can obtain the existence of homoclinic or periodic orbits equivalent to the existence of the zeros of the corresponding bifurcation function. That is,

x i , 1 = f ( x i , 1 , μ ) , t [ T , 0 ] , x i , 2 = f ( x i , 2 , μ ) , t [ 0 , T ] , x i , 1 ( 0 ) , x i , 2 ( 0 ) Σ P ± P ± , x 1,2 ( T ) = x 2,1 ( T ) , x 1,1 ( T ) = x 2,2 ( T ) , x i 1 ( t ) γ i ( t ) < ε , t [ T , 0 ] , x i 2 ( t ) γ i ( t ) < ε , t [ 0 , T ] , x i 1 ( 0 ) x i 2 ( 0 ) R ψ i ( 0 ) .

There are constants 0 < ε 1 , T 0 1 , such that the boundary-value problem has a unique solution { x i } i = 1,2 with T > T 0 and x i < ε , and the solution is smooth. we can obtain the bifurcation function as follows:

ζ i ψ i ( 0 ) , x i 1 ( 0 ) x i 2 ( 0 ) .

That is, the existence of perodic orbits is equivalent to the existence of ζ i 0 .

2.2 Exponential dichotomy

Since P + and P are hyperbolic equilibria of equation (1.1), at the same time

lim t + D f ( γ 1 ( t ) ) = D f ( P + ) , lim t D f ( γ 1 ( t ) ) = D f ( P ) , lim t + D f ( γ 2 ( t ) ) = D f ( P ) , lim t D f ( γ 2 ( t ) ) = D f ( P + ) .

By the exponential dichotomy roughness theorem, we know (1.3) has two-side exponential dichotomies on R and R + . Most of our analysis depends on the exponential dichotomy of the linear variational equation around the heteroclinic orbit γ i , so these basic properties are given in detail, respectively. For a detailed analysis, we refer Coppel [26] and Palmer [27]. We start with the following lemma (refer [28] for similar results in a different setting).

Lemma 2.1

Assume that (H5) and (H6) hold. There exists a fundamental matrix solution U 1 for the variational equation

(2.2) u ˙ ( t ) = D f ( γ 1 ( t ) ) u ( t ) ,

where constants M > 0 , K 0 > 0 , n × n nonsingular matrix C, projections P s s 1 , P u s + , P s u , P u u 1 with P s s 1 + P u s + + P s u + P u u 1 = I , I be the n × n unit matrix, and the following hold:

(2.3) ( i ) U 1 ( t ) ( P s s 1 + P u s + ) U 1 1 ( s ) K 0 e 2 M ( s t ) , 0 < s t , ( i i ) U 1 ( t ) ( P s u + P u u 1 ) U 1 1 ( s ) K 0 e 2 M ( t s ) , 0 < t s , ( i i i ) U 1 ( t ) ( P s s 1 + P s u ) U 1 1 ( s ) K 0 e 2 M ( t s ) , t s < 0 , ( i v ) U 1 ( t ) ( P u s + + P u u 1 ) U 1 1 ( s ) K 0 e 2 M ( s t ) , s t < 0 , ( v ) lim t + U 1 ( t ) ( P s s 1 + P u s + ) U 1 1 ( t ) = C I s + 0 0 0 C 1 , ( v i ) lim t + U 1 ( t ) ( P s u + P u u 1 ) U 1 1 ( t ) = C 0 0 0 I u 1 C 1 , ( v i i ) lim t U 1 ( t ) ( P s s 1 + P s u ) U 1 1 ( t ) = C 0 0 0 I u C 1 , ( v i i i ) lim t U 1 ( t ) ( P u s + + P u u 1 ) U 1 1 ( t ) = C I s 1 0 0 0 C 1 ,

where I s + , I u 1 , I u , I s 1 are the d + × d + , ( n d s ) × ( n d s ) , ( n d ) × ( n d ) , ( d + s ) × ( d + s ) unit matrixes, respectively.

Moreover, rank ( P s s 1 ) = d 1 , rank ( P u u 1 ) = d 1 s .

For the variational equation

(2.4) u ˙ ( t ) = D f ( γ 2 ( t ) ) u ( t ) ,

we have analogous two-side exponential dichotomies on R and R + . Without loss of generality, we take these constants to be the same as above. Then, we have the following Lemma.

Lemma 2.2

Assume that (H5) and (H6) hold. There exists a fundamental matrix solution U 2 for (2.4), constants M > 0 , K 0 > 0 , n × n nonsingular matrix C, projections P s s 2 , P u s , P u u 2 , P s u such that P s s 2 + P u s + P u u 2 + P s u + = I , I be the n × n unit matrix, and the following hold:

(2.5) ( i ) U 2 ( t ) ( P s s 2 + P u s ) U 2 1 ( s ) K 0 e 2 M ( s t ) , 0 < s t , ( i i ) U 2 ( t ) ( P s u + + P u u 2 ) U 2 1 ( s ) K 0 e 2 M ( t s ) , 0 < t s , ( i i i ) U 2 ( t ) ( P s s 2 + P s u + ) U 2 1 ( s ) K 0 e 2 M ( t s ) , t s < 0 , ( i v ) U 2 ( t ) ( P u s + P u u 2 ) U 2 1 ( s ) K 0 e 2 M ( s t ) , s t < 0 , ( v ) lim t U 2 ( t ) ( P s s 2 + P u s ) U 2 1 ( t ) = C I s 0 0 0 C 1 , ( v i ) lim t U 2 ( t ) ( P s u + + P u u 2 ) U 2 1 ( t ) = C 0 0 0 I u 2 C 1 , ( v i i ) lim t U 2 ( t ) ( P s s 2 + P s u + ) U 2 1 ( t ) = C 0 0 0 I u + C 1 , ( v i i i ) lim t U 2 ( t ) ( P u s + P u u 2 ) U 2 1 ( t ) = C I s 2 0 0 0 C 1 ,

where I s , I u 2 , I u + , I s 2 are the d × d , ( n d + + s ) × ( n d + + s ) , ( n d + ) × ( n d + ) , ( d + s ) × ( d + s ) unit matrixes, respectively.

Moreover, rank ( P s s 2 ) = d 2 , rank ( P u u 2 ) = d 2 + s .

2.3 Statement of the main result

Before presenting the main result, we introduce some notations. Let u k i denote the k th column of U i . Renumbering if necessary, we can assume that

P u u i = I d i + ( 1 ) i s 0 0 0 0 d i 0 0 0 0 , P s s i = 0 d i + ( 1 ) i s 0 0 0 I d i 0 0 0 0 ,

where I d i + ( 1 ) i s and 0 d i + ( 1 ) i s are ( d i + ( 1 ) i s ) × ( d i + ( 1 ) i s ) identity and zero matrixes, respectively, for i = 1 , 2. Then, we have

P u u i U i = ( u 1 i , u 2 i , , u d i + ( 1 ) i s i , 0 , , 0 ) , P s s i U i = ( 0 , , 0 , u d i + ( 1 ) i s + 1 , u d i + ( 1 ) i s + 2 , , u d i + ( 1 ) i s + d i , 0 , , 0 ) ,

for i = 1 , 2.

Let U i 1 denote the inverse of U i . Then, we have U i 1 U i = I . Differentiating with respect to t , we obtain U ˙ i 1 U i + U i 1 U ˙ i = 0 and hence, U ˙ i 1 = U i 1 U ˙ i U i 1 = U i 1 D f ( γ i ) . Then, ( U ˙ i 1 ) T = D f ( γ i ) T ( U i 1 ) T , where T denotes the transpose of a matrix. Hence, ( U i 1 ) T is a fundamental matrix solution of the adjoint equation (2.4). Let ( u j i ) denote the j th row of U i 1 . ( u j i ) can be obtained from the fundamental matrix solution of the adjoint equation. Clearly, ( u p i ) , u q i = δ p q , the Kronecker delta. By the definition of P u u i , we have

P u u i U i 1 = ( ( u 1 i ) T , ( u 2 i ) T , , ( u d i + ( 1 ) i s i ) T , 0 , , 0 ) T ,

for i = 1 , 2.

Let

a j , k i ( α i ) = + ( u j i ) ( s ) , g k ( γ i ( s ) , 0 , s + α i ) d s b j , p q i = + ( u j i ) ( s ) , D 11 f ( γ i ( s ) ) u p i ( s ) u q i ( s ) d s ,

i = 1 , 2, j = 1 , , d i + ( 1 ) i s , p , q = d i + ( 1 ) i s + 1 , , d i + ( 1 ) i s + d i 1 . Using those notations, we let

M j i ( β , μ , α i ) = k = 1 2 a j , k i ( α i ) μ k + 1 2 p = 1 d i 1 q = 1 d i 1 b j , p q i β p i β q i ,

for i = 1 , 2,  j = 1 , , d i + ( 1 ) i s , p , q = d i + ( 1 ) i s + 1 , , d i + ( 1 ) i s + d i 1 . And define M i ( β , μ , α i ) : R d 1 + d 2 2 × R 2 × R R d i + ( 1 ) i s by

M i ( β , μ , α i ) = ( M 1 i ( β , μ , α i ) , , M d i + ( 1 ) i s i ( β , μ , α i ) ) .

Further, we let M : R d 1 + d 2 2 × R 2 × R 2 R d 1 s × R d 2 + s be given by

(2.6) M ( β , μ , α ) = ( M 1 ( β , μ , α 1 ) , M 2 ( β , μ , α 2 ) ) ,

where α = ( α 1 , α 2 ) .

Our main result can be stated as follows.

Theorem 1

Assume that (H1)–(H5) hold. Let M ( β , μ , α ) be as in (2.6). If there are some points ( β 0 , μ 0 , α 0 ) R d 1 + d 2 2 × R 2 × R 2 , such that

M ( β 0 , μ 0 , α 0 ) = 0

and

D ( β , μ ) M ( β 0 , μ 0 , α 0 )

is a nonsingular ( d 1 + d 2 ) × ( d 1 + d 2 ) matrix, then there exists constants T 0 > 0 , r 0 > 0 and a differentiable function ψ ˆ * : B 2 ( α 0 , r 0 ) × ( T 0 , + ) B ¯ 1 ( r 0 ) , where B 2 ( α 0 , r 0 ) R 2 , B ¯ 1 ( r 0 ) R d 1 + d 2 , such that the perturbed equation (1.2) has periodic or homoclinic solution γ ( t ) near the unperturbed heteroclinic loop Γ when μ = μ 0 + ψ ˆ * .

The proof establishing the existence of a periodic solution bifurcating from the heteroclinic loop Γ is detailed in Section 3.1. Figure 2 illustrates the associated bifurcation phenomenon. The proof of the homoclinic solution bifurcated from heteroclinic loop Γ is given in Section 3.2. And, the Figure 3 illustrates the associated bifurcation phenomenon.

Figure 2 
                  The periodic solution bifurcated from heteroclinic loop 
                        
                           
                           
                              Γ
                           
                           \Gamma 
                        
                     .
Figure 2

The periodic solution bifurcated from heteroclinic loop Γ .

Figure 3 
                  The homoclinic solution bifurcated from heteroclinic loop 
                        
                           
                           
                              Γ
                           
                           \Gamma 
                        
                     . (a) Homoclinic solution 
                        
                           
                           
                              γ
                              
                                 (
                                 
                                    t
                                 
                                 )
                              
                           
                           \gamma \left(t)
                        
                      asymptotic to hyperbolic equilibrium 
                        
                           
                           
                              
                                 
                                    P
                                 
                                 
                                    −
                                 
                              
                           
                           {P}_{-}
                        
                      near the unperturbed heteroclinic loop 
                        
                           
                           
                              Γ
                           
                           \Gamma 
                        
                      and (b) homoclinic solution 
                        
                           
                           
                              γ
                              
                                 (
                                 
                                    t
                                 
                                 )
                              
                           
                           \gamma \left(t)
                        
                      asymptotic to hyperbolic equilibrium 
                        
                           
                           
                              
                                 
                                    P
                                 
                                 
                                    +
                                 
                              
                           
                           {P}_{+}
                        
                      near the unperturbed heteroclinic loop 
                        
                           
                           
                              Γ
                           
                           \Gamma 
                        
                     .
Figure 3

The homoclinic solution bifurcated from heteroclinic loop Γ . (a) Homoclinic solution γ ( t ) asymptotic to hyperbolic equilibrium P near the unperturbed heteroclinic loop Γ and (b) homoclinic solution γ ( t ) asymptotic to hyperbolic equilibrium P + near the unperturbed heteroclinic loop Γ .

3 Proof of Theorem 1

For the proof of the conclusion of Theorem 1, we apply Lin’s method for constructing Lin orbits x i , i = 1 , 2 that are composed of piecewise continuous orbits x i , 1 , x i , 2 , i = 1 , 2 near the heteroclinic loop, which is characterized by a gap ζ i on Σ P ± P ± in a distinguished direction Z . The orbit x 1 = { x 1,1 , x 2,2 } starts in Σ P P + , follows Σ P P + until it reaches a neighborhood of P + follows then Σ P + P , and x 2 = { x 1,2 , x 2,1 } starts in Σ P + P , stays further close to Σ P + P until it reaches a neighborhood of P , follows then Σ P P + again, and terminates finally in Σ P P + . Because the periodic and homoclinic solutions are constructed in different ways, we divide the proof of Theorem 1 into two parts, which prove that the periodic solution can be bifurcated from heteroclinic loop Γ in Section 3.1 and the homoclinic solution bifurcated from heteroclinic loop Γ in Section 3.2, respectively. First, we seek the periodic solution near the unperturbed heteroclinic loop Γ .

3.1 Periodic solution bifurcated from heteroclinic loop Γ

In this section, our objective is to find periodic solution near the unperturbed heteroclinic loop Γ for equation (1.2). The Lin orbits have been glued together at t = T and t = T , with a gap at t = 0 . When the gap disappears, they can be glued together to form a periodic solution near the heteroclinic loop Γ .

We define functions b 1 : R d 1 + d 2 2 × ( 0 , + ) R n by

b 1 ( β , t ) = ( γ 1 ( t ) γ 2 ( t ) ) ( γ 1 ( t ) γ 2 ( t ) ) i = 1 d 1 1 β i 1 u d 1 s + i 1 ( t ) i = 1 d 2 1 β i 2 u d 2 + s + i 2 ( t ) i = 1 d 1 1 β i 1 u d 1 s + i 1 ( t ) i = 1 d 2 1 β i 2 u d 2 + s + i 2 ( t ) ,

and b 2 : R d 1 + d 2 2 × ( 0 , + ) R n by

b 2 ( β , t ) = ( γ 1 ( t ) γ 2 ( t ) ) + ( γ 1 ( t ) γ 2 ( t ) ) i = 1 d 1 1 β i 1 u d 1 s + i 1 ( t ) i = 1 d 2 1 β i 2 u d 2 + s + i 2 ( t ) + i = 1 d 1 1 β i 1 u d 1 s + i 1 ( t ) i = 1 d 2 1 β i 2 u d 2 + s + i 2 ( t ) ,

where β = ( β 1 1 , , β d 1 1 1 , β 1 2 , , β d 2 1 2 ) . For i = 1 , 2, note that

b i ( β , t ) = O ( e M t ) ,

uniformly with respect to β from any bounded subset of R d 1 + d 2 2 .

For i = 1 or i = 2 , we suppose x i ( t ) is a solution of equation (1.2). Fix T N and take the change in variable

(3.1) x 1 ( t + α 1 ) = γ 1 ( t ) + y 1 ( t ) + i = 1 d 1 1 β i 1 u d 1 s + i 1 ( t ) + 1 2 b 1 ( β , T ) ,

(3.2) x 2 ( t + α 2 ) = γ 2 ( t ) + y 2 ( t ) + i = 1 d 2 1 β i 2 u d 2 + s + i 2 ( t ) + 1 2 T b 2 ( β , T ) t .

Then, y i ( t ) satisfies the following equation:

(3.3) y i ˙ = D f ( γ i ) y i + g ˜ i ( y i , β , μ , α i ) ,

with

(3.4) g ˜ 1 ( y 1 , β , μ , α 1 , T ) ( t ) = f ( γ 1 ( t ) + y 1 ( t ) + i = 1 d 1 1 β i 1 u d 1 s + i 1 ( t ) + 1 2 b 1 ( β , T ) ) f ( γ 1 ( t ) ) D f ( γ 1 ( t ) ) y 1 ( t ) i = 1 d 1 1 β i 1 D f ( γ 1 ( t ) ) u d 1 s + i 1 ( t ) + j = 1 2 μ j g j γ 1 ( t ) + y 1 ( t ) + i = 1 d 1 1 β i 1 u d 1 s + i 1 ( t ) + 1 2 b 1 ( β , T ) , μ , t + α 1 .

and

(3.5) g ˜ 2 ( y 2 , β , μ , α 2 , T ) ( t ) = f γ 2 ( t ) + y 2 ( t ) + i = 1 d 2 1 β i 2 u d 2 + s + i 2 ( t ) + 1 2 T b 2 ( β , T ) t f ( γ 2 ( t ) ) D f ( γ 2 ( t ) ) y 2 ( t ) i = 1 d 2 1 β i 2 D f ( γ 2 ( t ) ) u d 2 + s + i 2 ( t ) 1 2 T b 2 ( β , T ) + j = 1 2 μ j g j γ 2 ( t ) + y 2 ( t ) + i = 1 d 2 1 β i 2 u d 2 + s + i 2 ( t ) + 1 2 T b 2 ( β , T ) t , μ , t + α 2 .

By the definition of the function b i ( β , t ) , if y 1 ( T ) = y 2 ( T ) and y 1 ( T ) = y 2 ( T ) , then x 1 ( T + α 1 ) = x 2 ( T + α 2 ) and x 1 ( T + α 1 ) = x 2 ( T + α 2 ) . Hence, under the above change in variable, we will seek solution y i ( t ) of equation (3.3) satisfying y 1 ( T ) = y 2 ( T ) and y 1 ( T ) = y 2 ( T ) .

Let D i g or D i j h denote the derivatives of a multivariate function g with respect to its ith or the i, jth variables. Based on the formula of g ˜ i ( y i , β , μ , α i , T ) , the following lemma is derived through a straightforward computation.

Lemma 3.1

For any i = 1 or i = 2 , the function g ˜ i ( y i , β , μ , α i , T ) satisfies the following properties:

( i ) lim T g ˜ i ( 0,0,0 , α i , T ) = 0 ; lim T D 1 g ˜ i ( 0,0,0 , α i , T ) = 0 ; ( i i ) lim T D 11 g ˜ i ( 0,0,0 , α i , T ) = D 11 f ( γ i ) ; ( i i i ) lim T g ˜ i μ j ( 0,0,0 , α i , T ) ( t ) = g j ( γ i , 0 , t + α i ) ; ( i v ) lim T g ˜ i β j i ( 0,0,0 , α i , T ) = 0 ; lim T 2 g ˜ i β j i β k i ( 0,0,0 , α i , T ) = D 11 f ( γ i ) u j i u k i .

Let C 1 ( [ a , b ] , R n ) be the Banach space of C 1 functions, which take [ a , b ] into R n , and we equipped norm z = max t [ a , b ] z ( t ) , for z C 1 ( [ a , b ] , R n ) . Assume

X T = C 1 ( [ T , 0 ] , R n ) , Y T = C 1 ( [ 0 , T ] , R n ) .

For any η i R n and ξ i R n , i = 1, 2, 3, 4, consider the functions z 1 , z 3 X T and z 2 , z 4 Y T by

z 1 ( t ) = U 1 ( t ) P s u ξ 1 + U 1 ( t ) 0 t ( P s s 1 + P s u ) U 1 1 ( s ) g ˜ 1 ( z 1 , β , μ , α 1 , T ) ( s ) d s + U 1 ( t ) ( P u u 1 + P u s + ) U 1 1 ( T ) η 1 + U 1 ( t ) T t ( P u u 1 + P u s + ) U 1 1 ( s ) g ˜ 1 ( z 1 , β , μ , α 1 , T ) ( s ) d s , z 2 ( t ) = U 1 ( t ) P u s + ξ 2 + U 1 ( t ) 0 t ( P s s 1 + P u s + ) U 1 1 ( s ) g ˜ 1 ( z 2 , β , μ , α 1 , T ) ( s ) d s + U 1 ( t ) ( P u u 1 + P s u ) U 1 1 ( T ) η 2 + U 1 ( t ) T t ( P u u 1 + P s u ) U 1 1 ( s ) g ˜ 1 ( z 2 , β , μ , α 1 , T ) ( s ) d s , z 3 ( t ) = U 2 ( t ) P s u + ξ 3 + U 2 ( t ) 0 t ( P s s 2 + P s u + ) U 2 1 ( s ) g ˜ 2 ( z 3 , β , μ , α 2 , T ) ( s ) d s + U 2 ( t ) ( P u u 2 + P u s ) U 2 1 ( T ) η 3 + U 2 ( t ) T t ( P u u 2 + P u s ) U 2 1 ( s ) g ˜ 2 ( z 3 , β , μ , α 2 , T ) ( s ) d s , z 4 ( t ) = U 2 ( t ) P u s ξ 4 + U 2 ( t ) 0 t ( P s s 2 + P u s ) U 2 1 ( s ) g ˜ 2 ( z 4 , β , μ , α 2 , T ) ( s ) d s + U 2 ( t ) ( P u u 2 + P s u + ) U 2 1 ( T ) η 4 + U 2 ( t ) T t ( P u u 2 + P s u + ) U 2 1 ( s ) g ˜ 2 ( z 4 , β , μ , α 2 , T ) ( s ) d s .

Lemma 3.2

Given η i and ξ i , the functions z 1 , z 2 , z 3 , z 4 are solutions of equation (3.3) for i = 1, 2, 3, 4.

Proof

For given η i and ξ i , by the definition of z i ( t ) , if z i ( t ) is a continuous solution of the above integral equation, then taking derivatives with respect to t , a simple computation shows that z i ( t ) is a solution of equation (3.3) near the heteroclinic loop Γ , i = 1 , 2 , 3 , 4 .

Next we will prove that the above integral equation has a continuous solution. Using the equation z i , we define the operator F i by the right hand side of equation z i . Next we will show that the operator F i has a fixed point in the space X T or Y T for i = 1, 2, 3, 4. We only prove the situation of the operator F 1 in the space X T . Other proofs are similar.

Let the constants K 0 and M be given in (2.3). By Lemma 3.1 and smoothness of f and g j , there are constants δ , δ 1 , δ 2 , and T 0 such that

g ˜ 1 ( z 1 , β , μ , α 1 , T ) δ M 2 K 0 g ˜ 1 ( z 1 , β , μ , α 1 , T ) g ˜ 1 ( z ˜ 1 , β , μ , α 1 , T ) M 2 K 0 z 1 z ˜ 1

for t [ T , 0 ] , ( z 1 , z ˜ 1 ) B ¯ ( 0 , δ ) , and ( β , μ , α 1 , T ) B ¯ 1 ( 0 , δ 1 ) × B ¯ 2 ( 0 , δ 2 ) × R × ( T 0 , + ) , where B ¯ ( 0 , δ ) , B ¯ 1 ( 0 , δ 1 ) , and B ¯ 2 ( 0 , δ 2 ) are closed subsets with radius δ > 0 , δ 1 > 0 , and δ 2 > 0 centered at the origin of X T , R d 1 + d 2 2 and R 2 , repetitively.

With this choice of δ and for any η 1 R n and ξ 1 R n with η 1 δ 4 K 0 and ξ 1 δ 4 K 0 , define a Banach space

X T ( ξ 1 , η 1 ) = { z 1 ( t ) z 1 ( t ) B ¯ ( 0 , δ ) , P s u z 1 ( 0 ) = P s u ξ 1 , and ( P u u 1 + P u s + ) z 1 ( T ) = U 1 ( T ) ( P u u 1 + P u s + ) U 1 1 ( T ) η 1 }

For any z 1 ( t ) X T ( ξ 1 , η 1 ) , define

F 1 ( z 1 ) ( t ) = U 1 ( t ) P s u ξ 1 + U 1 ( t ) 0 t ( P s s 1 + P s u ) U 1 1 ( s ) g ˜ 1 ( z 1 , β , μ , α 1 , T ) ( s ) d s + U 1 ( t ) ( P u u 1 + P u s + ) U 1 1 ( T ) η 1 + U 1 ( t ) T t ( P u u 1 + P u s + ) U 1 1 ( s ) g ˜ 1 ( z 1 , β , μ , α 1 , T ) ( s ) d s ,

for t [ T , 0 ] . It is easy to see that P s u F 1 ( z 1 ) ( 0 ) = P s u ξ 1 and ( P u u 1 + P u s + ) F 1 ( z 1 ) ( T ) = U 1 ( T ) ( P u u 1 + P u s + ) U 1 1 ( T ) η 1 . By the exponential dichotomy of the variational equation, we obtain

(3.6) F 1 ( z 1 ) ( t ) U 1 ( t ) P s u ξ 1 + U 1 ( t ) ( P u u 1 + P u s + ) U 1 1 ( T ) η 1 + 0 t U 1 ( t ) ( P s s 1 + P s u ) U 1 1 ( s ) g ˜ 1 ( z 1 , β , μ , α 1 , T ) ( s ) d s + T t U 1 ( t ) ( P u u 1 + P u s + ) U 1 1 ( s ) g ˜ 1 ( z 1 , β , μ , α 1 , T ) ( s ) d s , K 0 e 2 M t ξ 1 + 0 t K 0 e 2 M ( t s ) g ˜ 1 d s + K 0 e 2 M ( T + t ) η 1 + T t K 0 e 2 M ( s t ) g ˜ 1 d s K 0 ξ 1 + K 0 2 M g ˜ 1 + K 0 η 1 + K 0 2 M g ˜ 1 δ 4 + δ 4 + δ 4 + δ 4 = δ ,

for t [ T , 0 ] . Thus, F 1 ( z 1 ) ( t ) δ , so F 1 : X T ( ξ 1 , η 1 ) X T ( ξ 1 , η 1 ) .

For any ( z 1 , z ˜ 1 ) X T ( ξ 1 , η 1 ) , we have

F 1 ( z 1 ) ( t ) F 1 ( z ˜ 1 ) ( t ) = 0 t U 1 ( t ) ( P s s 1 + P s u ) U 1 1 ( s ) g ˜ 1 ( z 1 , β , μ , α 1 , T ) ( s ) d s 0 t U 1 ( t ) ( P s s 1 + P s u ) U 1 1 ( s ) g ˜ 1 ( z ˜ 1 , β , μ , α 1 , T ) ( s ) d s + T t U 1 ( t ) ( P u u 1 + P u s + ) U 1 1 ( s ) g ˜ 1 ( z 1 , β , μ , α 1 , T ) ( s ) d s T t U 1 ( t ) ( P u u 1 + P u s + ) U 1 1 ( s ) g ˜ 1 ( z ˜ 1 , β , μ , α 1 , T ) ( s ) d s 0 t U 1 ( t ) ( P s s 1 + P s u ) U 1 1 ( s ) g ˜ 1 ( z 1 , β , μ , α 1 , T ) g ˜ 1 ( z ˜ 1 , β , μ , α 1 , T ) d s + T t U 1 ( t ) ( P u u 1 + P u s + ) U 1 1 ( s ) g ˜ 1 ( z 1 , β , μ , α 1 , T ) g ˜ 1 ( z ˜ 1 , β , μ , α 1 , T ) d s k 0 M g ˜ 1 ( z 1 , β , μ , α 1 , T ) g ˜ 1 ( z ˜ 1 , β , μ , α 1 , T ) 1 2 z 1 z ˜ 1 ,

for t [ T , 0 ] . Thus, F 1 is a contraction on X T ( ξ 1 , η 1 ) and there is a unique fixed point z 1 * ( β , μ , α 1 , T ) X T ( ξ 1 , η 1 ) , for ( β , μ , α 1 , T ) B ¯ 1 ( 0 , δ 1 ) × B ¯ 2 ( 0 , δ 2 ) × R × ( T 0 , + ) . Therefore, we can prove that there exists a solution z 1 * of equation (3.3) on [ T , 0 ] . Similarly, we can prove that there exists a solution z 2 * of equation (3.3) on [ 0 , T ] , and a solution z 3 * of equation (3.3) on [ T , 0 ] , and a solution z 4 * of equation (3.3) on [ 0 , T ] .

Thus, we conclude that z 1 , z 2 , z 3 , z 4 are said to be four piecewise continuous solutions of equation (3.3) near the heteroclinic Γ (Figure 4).

The proof is complete.□

Figure 4 
                  
                     
                        
                           
                           
                              
                                 
                                    z
                                 
                                 
                                    1
                                 
                              
                              
                                 (
                                 
                                    t
                                 
                                 )
                              
                           
                           {z}_{1}\left(t)
                        
                     ,…,
                        
                           
                           
                              
                                 
                                    z
                                 
                                 
                                    4
                                 
                              
                              
                                 (
                                 
                                    t
                                 
                                 )
                              
                           
                           {z}_{4}\left(t)
                        
                      near the heteroclinic loop 
                        
                           
                           
                              Γ
                           
                           \Gamma 
                        
                     .
Figure 4

z 1 ( t ) ,…, z 4 ( t ) near the heteroclinic loop Γ .

If we can seek some η i and ξ i such that z 1 ( 0 ) = z 2 ( 0 ) , z 3 ( 0 ) = z 4 ( 0 ) , then z 1 ( t ) and z 2 ( t ) stick together at t = 0 and z 3 ( t ) and z 4 ( t ) stick together at t = 0 . So, z 1 ( t ) and z 2 ( t ) can form a solution y 1 ( t ) of equation (3.3) in [ T , T ] near the heteroclinic orbit γ 1 , and z 3 ( t ) and z 4 ( t ) form a solution y 2 ( t ) of equation (3.3) in [ T , T ] near the heteroclinic orbit γ 2 . Moreover, if z 4 ( T ) = z 1 ( T ) and z 2 ( T ) = z 3 ( T ) , then y 1 ( t ) and y 2 ( t ) satisfy y 1 ( T ) = y 2 ( T ) and y 1 ( T ) = y 2 ( T ) . Hence, equation (3.3) has a periodic solution y ( t ) with 4 T period consisting of z 1 ( t ) , z 2 ( t ) , z 3 ( t ) , and z 4 ( t ) .

y ( β , μ , α , T ) ( t ) = y 1 ( β , μ , α 1 , T ) ( t ) = z 1 ( β , μ , α 1 , T ) ( t ) , t [ T , 0 ] , z 2 ( β , μ , α 1 , T ) ( t ) , t [ 0 , T ] , y 2 ( β , μ , α 2 , T ) ( t ) = z 3 ( β , μ , α 2 , T ) ( t ) , t [ T , 0 ] , z 4 ( β , μ , α 2 , T ) ( t ) , t [ 0 , T ] .

By the definition of b i ( β , t ) and the change in variables (3.1) and (3.2), we have x 1 ( T + α 1 ) = x 2 ( T + α 2 ) , x 1 ( T + α 1 ) = x 2 ( T + α 2 ) . Hence, equation (1.2) has a periodic solution with 4 T period consisting of x 1 ( t ) and x 2 ( t ) near the heteroclinic loop.

Next we will seek some η i and ξ i such that z 1 ( 0 ) = z 2 ( 0 ) , z 3 ( 0 ) = z 4 ( 0 ) , z 2 ( T ) = z 3 ( T ) , and z 4 ( T ) = z 1 ( T ) hold. We can decompose z 1 ( 0 ) = z 2 ( 0 ) into the following three equations:

(3.7) P s u ξ 1 = P s u U 1 1 ( T ) η 2 + T 0 P s u U 1 1 ( s ) g ˜ 1 d s ,

(3.8) P u s + ξ 2 = P u s + U 1 1 ( T ) η 1 + T 0 P u s + U 1 1 ( s ) g ˜ 1 d s ,

(3.9) T T P u u 1 U 1 1 ( s ) g ˜ 1 d s + P u u 1 U 1 1 ( T ) η 1 P u u 1 U 1 1 ( T ) η 2 = 0 .

From z 3 ( 0 ) = z 4 ( 0 ) , we can obtain something similar

(3.10) P s u + ξ 3 = P s u + U 2 1 ( T ) η 4 + T 0 P s u + U 2 1 ( s ) g ˜ 2 d s ,

(3.11) P u s ξ 4 = P u s U 2 1 ( T ) η 3 + T 0 P u s U 2 1 ( s ) g ˜ 2 d s ,

(3.12) T T P u u 2 U 2 1 ( s ) g ˜ 2 d s + P u u 2 U 2 1 ( T ) η 3 P u u 2 U 2 1 ( T ) η 4 = 0 .

From z 2 ( T ) = z 3 ( T ) and z 4 ( T ) = z 1 ( T ) , we can obtain

(3.13) U 1 ( T ) ( P u u 1 + P s u ) U 1 1 ( T ) η 2 U 2 ( T ) ( P u u 2 + P u s ) U 2 1 ( T ) η 3 = U 2 ( T ) 0 T ( P s s 2 + P s u + ) U 2 1 ( s ) g ˜ 2 d s U 1 ( T ) 0 T ( P s s 1 + P u s + ) U 1 1 ( s ) g ˜ 1 d s + U 2 ( T ) P s u + ξ 3 U 1 ( T ) P u s + ξ 2

and

(3.14) U 2 ( T ) ( P u u 2 + P s u + ) U 2 1 ( T ) η 4 U 1 ( T ) ( P u u 1 + P u s + ) U 1 1 ( T ) η 1 = U 1 ( T ) 0 T ( P s s 1 + P s u ) U 1 1 ( s ) g ˜ 1 d s U 2 ( T ) 0 T ( P s s 2 + P u s ) U 2 1 ( s ) g ˜ 2 d s + U 1 ( T ) P s u ξ 1 U 2 ( T ) P u s ξ 4 .

In (3.13), taking T approach infinity, we have

(3.15) C 0 0 0 I u 1 C 1 η 2 C I s 2 0 0 0 C 1 η 3 = U 2 ( ) 0 ( P s s 2 + P s u + ) U 2 1 ( s ) g ˜ 2 d s U 1 ( ) 0 ( P s s 1 + P u s + ) U 1 1 ( s ) g ˜ 1 d s ,

where (2.3) and (2.5) are used to ensure the existence of the limit, I u 1 and I s 2 are ( n d s ) × ( n d s ) , and ( d + s ) × ( d + s ) unit matrixes, respectively. Assume

C 1 η 2 = 0 u 2 , C 1 η 3 = u 1 0 , u = u 1 u 2 .

From (3.15), we can solve

u = C 1 ( U 2 ( ) 0 ( P s s 2 + P s u + ) U 2 1 ( s ) g ˜ 2 d s U 1 ( ) 0 ( P s s 1 + P u s + ) U 1 1 ( s ) g ˜ 1 d s ) .

Then, there exists T 0 > 0 so that we can solve η 2 and η 3 in (3.13) whenever T > T 0 . We denote that the solutions are η 2 * = η 2 * ( β , μ , α 1 , T ) and η 3 * = η 3 * ( β , μ , α 2 , T ) . Applying the same method we can solve η 1 and η 4 in (3.14) whenever T > T 0 . We denote that the solutions are η 1 * = η 1 * ( β , μ , α 1 , T ) , η 4 * = η 4 * ( β , μ , α 2 , T ) . And η i * satisfies lim T η i * ( 0,0 , α j , T ) = 0 for α j R and i = 1 , , 4 , j = 1 , 2.

Substituting η 1 * , η 2 * , η 3 * , η 4 * for η 1 , η 2 , η 3 , η 4 in (3.7), (3.8), (3.10), and (3.11), we can obtain ξ 1 , ξ 2 , ξ 3 , ξ 4 . In addition, if η 1 * , η 2 * , η 3 * , η 4 * satisfy (3.9) and (3.12), then z 1 ( 0 ) = z 2 ( 0 ) , z 2 ( T ) = z 3 ( T ) , z 3 ( 0 ) = z 4 ( 0 ) , z 4 ( T ) = z 1 ( T ) , that is, (3.3) has a periodic solution with 4 T period consisting of z 1 ( t ) , z 2 ( t ) , z 3 ( t ) , and z 4 ( t ) . Hence, we have the following result for equation (1.2).

Lemma 3.3

Assume U 1 , U 2 , P u u 1 , P u u 2 be as in (2.3) and (2.5). There exists T 0 > 0 and if

(3.16) T T P u u 1 U 1 1 ( s ) g ˜ 1 d s + P u u 1 U 1 1 ( T ) η 1 * P u u 1 U 1 1 ( T ) η 2 * = 0 ,

(3.17) T T P u u 2 U 2 1 ( s ) g ˜ 2 d s + P u u 2 U 2 1 ( T ) η 3 * P u u 2 U 2 1 ( T ) η 4 * = 0 ,

for T > T 0 , then equation (1.2) has a periodic solution with 4 T period near the unperturbed heteroclinic loop Γ , where g ˜ 1 and g ˜ 2 be as in (3.4) and (3.5).

Next we will give a sufficient condition for the existence of zeros of (3.16) and (3.17). Based on g ˜ 1 , g ˜ 2 , and η i * , the left hand side of (3.16) and (3.17) depend on ( β , μ , α i , T ) . To simplify, we define function H ˜ i ( β , μ , α i , T ) : R d 1 + d 2 2 × R 2 × R R n by

H ˜ 1 ( β , μ , α 1 , T ) = T T P u u 1 U 1 1 ( s ) g ˜ 1 d s + P u u 1 U 1 1 ( T ) η 1 * P u u 1 U 1 1 ( T ) η 2 * , H ˜ 2 ( β , μ , α 1 , T ) = T T P u u 2 U 2 1 ( s ) g ˜ 2 d s + P u u 2 U 2 1 ( T ) η 3 * P u u 2 U 2 1 ( T ) η 4 * .

By the properties of g ˜ i and η j * , i = 1 , 2 , j = 1 , 2 , 3 , 4 , we have

H ˜ i ( 0,0 , α i , T ) = 0 and lim T P u u i U i 1 ( ± T ) η j * = 0 ,

uniformly with respect to ( β , μ , α i ) B ¯ 1 ( 0 , δ 1 ) × B ¯ 2 ( 0 , δ 2 ) × R , for i = 1 , 2 , j = 1 , 2 , 3 , 4 , B ¯ 1 ( 0 , δ 1 ) , and B ¯ 2 ( 0 , δ 2 ) are closed subsets with radius δ 1 > 0 and δ 2 > 0 centered at the origin of R d 1 + d 2 2 and R 2 , repetitively. For i = 1 , 2, assume

(3.18) M ˜ i ( β , μ , α i ) = P u u i U i 1 ( s ) g ˜ i ( y i , β , μ , α i , ) d s ,

and by

P u u i U i 1 = ( ( u 1 i ) T , ( u 2 i ) T , , ( u d i + ( 1 ) i s i ) T , 0 , , 0 ) T ,

hence

(3.19) M ˜ i ( β , μ , α i ) = ( M 1 i ( β , μ , α i ) , , M d i + ( 1 ) i s i ( β , μ , α i ) , 0 , , 0 ) ,

where

(3.20) M j i ( β , μ , α i ) = ( u j i ) ( s ) , g ˜ i ( y i , β , μ , α i , ) ( s ) d s ,

where j = 1 , 2 , , d i + ( 1 ) i s . By Lemma 3.1, we have

M j i ( β , μ , α i ) = k = 1 2 a j , k i ( α i ) μ k + 1 2 p = 1 d i 1 q = 1 d i 1 b j , p q i β p i β q i ,

where

a j , k i ( α i ) = + ( u j i ) ( s ) , g k ( γ i ( s ) , 0 , s + α i ) d s b j , p q i = + ( u j i ) ( s ) , D 11 f ( γ i ( s ) ) u p i ( s ) u q i ( s ) d s ,

i = 1 , 2 , j = 1 , , d i + ( 1 ) i s , p , q = d i + ( 1 ) i s + 1 , , d i + ( 1 ) i s + d i 1 . Hence,

H ˜ i ( β , μ , α i , T ) = M ˜ i ( β , μ , α i ) + hot ,

for i = 1 , 2. Define M i ( β , μ , α i ) : R d 1 + d 2 2 × R 2 × R R d i + ( 1 ) i s by

M i ( β , μ , α i ) = ( M 1 i ( β , μ , α i ) , , M d i + ( 1 ) i s i ( β , μ , α i ) ) .

Let

H i ( β , μ , α i , T ) = M i ( β , μ , α i ) + hot .

It follows from the definition of the projection P u u i that

H ˜ i ( β , μ , α i , T ) = ( H i ( β , μ , α i , T ) , 0 , , 0 ) .

Hence, H ˜ i ( β , μ , α i , T ) = 0 is equivalent to H i ( β , μ , α i , T ) = 0 . Define a function H : R d 1 + d 2 2 × R 2 × R 2 × R R d 1 + d 2 by

H ( β , μ , α , T ) = ( H 1 ( β , μ , α 1 , T ) , H 2 ( β , μ , α 2 , T ) ) ,

where α = ( α 1 , α 2 ) . So if there exists some ( β , μ , α , T ) such that H ( β , μ , α , T ) = 0 , then (3.16) and (3.17) are valid. By Lemma 3.6, equation (1.2) then has a periodic solution with 4 T period. Define a function M : R d 1 + d 2 2 × R 2 × R 2 R d 1 + d 2 by

M ( β , μ , α ) = ( M 1 ( β , μ , α 1 ) , M 2 ( β , μ , α 2 ) ) ,

where α = ( α 1 , α 2 ) . So

(3.21) H ( β , μ , α , T ) = M ( β , μ , α ) + hot .

Lemma 3.4

If there exists ( β 0 , μ 0 , α 0 ) R d 1 + d 2 2 × R 2 × R 2 such that M ( β 0 , μ 0 , α 0 ) = 0 and D ( β , μ ) M ( β 0 , μ 0 , α 0 ) is a nonsingular ( d 1 + d 2 ) × ( d 1 + d 2 ) matrix, then there exist constants T 0 > 0 , r 0 > 0 , and a differentiable function ψ * = ( ψ ˜ * , ψ ˆ * ) : B 2 ( α 0 , r 0 ) × ( T 0 , + ) B ¯ 1 ( r 0 ) , where B ¯ 1 ( r 0 ) R d 1 + d 2 and B 2 ( α 0 , r 0 ) R 2 , such that lim T ( ψ ˜ * ( α , T ) , ψ ˆ * ( α , T ) ) = 0 and H ( β 0 + ψ ˜ * ( α , T ) , μ 0 + ψ ˆ * ( α , T ) , α , T ) = 0 for ( α , T ) B 2 ( α 0 , r 0 ) × ( T 0 , + ) .

Proof

Let ω 0 = ( β 0 , μ 0 ) , ω = ( β , μ ) ω 0 , and A D ( β , μ ) M ( β 0 , μ 0 , α 0 ) 1 . Define

(3.22) ( ω , α ) = ω A M ( ω 0 + ω , α ) , ( ω , α , T ) = ω A H ( ω 0 + ω , α , T ) .

By the definition of , we know that the fixed points ψ * of ( , α , T ) correspond to H ( ω 0 + ψ * , α , T ) = 0 . Through direct calculations, we obtain that

(3.23) ( 0 , α 0 ) = 0 , D ω ( 0 , α 0 ) = 0 .

From (3.21) and (3.22), it is easy to check that

(3.24) lim T ( 0 , α 0 , T ) ( 0 , α 0 ) = lim T A ( H ( ω 0 , α 0 , T ) M ( ω 0 , α 0 ) ) = 0 , lim T D ω ( 0 , α 0 , T ) D ω ( 0 , α 0 ) = lim T A ( D ω H ( ω 0 , α 0 , T ) D ω M ( ω 0 , α 0 ) ) = 0 .

Let B 1 ( r ) R d 1 + d 2 be an open ball with radius r > 0 centered at the origin and B 2 ( α 0 , r ) R 2 be an open ball with radius r > 0 centered at α 0 . From (3.23), we obtain that there exists r 0 > 0 such that

(3.25) D ω ( ω , α ) 1 4 for ( ω , α ) B 1 ( r 0 ) × B 2 ( α 0 , r 0 ) .

Note that ( 0 , α 0 ) = 0 . We can obtain from (3.24) that there exists a constant T 0 > 0 such that

(3.26) ( 0 , α 0 , T ) = ( 0 , α 0 ) ( 0 , α 0 , T ) r 0 2 , D ω ( ω , α , T ) D ω ( ω , α ) 1 4 ,

for ( ω , α , T ) B 1 ( r 0 ) × B 2 ( α 0 , r 0 ) × ( T 0 , ) . Hence, we can obtain that

(3.27) D ω ( ω , α , T ) = D ω ( ω , α ) + D ω ( ω , α , T ) D ω ( ω , α ) D ω ( ω , α ) + D ω ( ω , α , T ) D ω ( ω , α ) 1 2 ,

for ( ω , α , T ) B 1 ( r 0 ) × B 2 ( α 0 , r 0 ) × ( T 0 , ) .

For any ( ω , α , T ) B 1 ( r 0 ) × B 2 ( α 0 , r 0 ) × ( T 0 , ) , define a map ψ 1 : [ 0 , 1 ] R d 1 + d 2 by ψ 1 ( s ) = ( s ω , α , T ) . We obtain

( ω , α , T ) = ψ 1 ( 1 ) = ψ 1 ( 0 ) + 0 1 ψ 1 ˙ ( s ) d s ( 0 , α , T ) + 0 1 D ω ( ω , α , T ) ω d s r 0 2 + 0 1 1 2 r 0 d s = r 0 ,

which implies that ( , α , T ) maps B ¯ 1 ( r 0 ) into itself.

For ω 1 , ω 2 B ¯ 1 ( r 0 ) , ( α , T ) B 2 ( α 0 , r 0 ) × ( T 0 , ) , define a map ψ 2 : [ 0 , 1 ] R d 1 + d 2 by ψ 2 ( s ) = ( s ω 1 + ( 1 s ) ω 2 , α , T ) . Clearly, ψ 2 C 2 . Then, there exists s 0 ( 0 , 1 ) such that

( ω 1 , α , T ) ( ω 2 , α , T ) = ψ 2 ( 1 ) ψ 2 ( 0 ) = 0 1 ψ 2 ˙ ( s 0 ) d s D ω ( ω , α , T ) ω 1 ω 2 1 2 ω 1 ω 2 .

Hence, for any ( α , T ) B 2 ( α 0 , r 0 ) × ( T 0 , ) , ( , α , T ) is a uniformly contraction map in B ¯ 1 ( r 0 ) . By the contraction mapping principle, there exists a C 2 function ψ * ( α , T ) : B 2 ( α 0 , r 0 ) × ( T 0 , ) B ¯ 1 ( r 0 ) such that lim T ψ * ( α , T ) = 0 and

( ψ * ( α , T ) , α , T ) = ψ * ( α , T ) .

By the definition of , this equality is equivalent to

H ( ω 0 + ψ * ( α , T ) , α , T ) = H ( β 0 + ψ ˜ * ( α , T ) , μ 0 + ψ ˆ * ( α , T ) , α , T ) = 0 ,

where ψ * = ( ψ ˜ * , ψ ˆ * ) R d 1 + d 2 1 × R 2 . The proof is complete.□

By transformations (3.1) and (3.2), we know that for ( α , T ) B 2 ( α 0 , r 0 ) × ( T 0 , ) , equation (1.2) has a periodic solution γ ( t ) with period 4T near the heteroclinic loop Γ as follows:

γ ( t ) = x 1 ( t ) , t [ α 01 T , T + α 01 ] , x 2 ( t ) , t [ α 02 T , T + α 02 ] ,

where x 1 ( t ) and x 2 ( t ) are

x 1 ( t ) = z 1 ( ω 0 + ψ * , α 01 , T ) + 1 2 b ( β 0 + ψ ˜ * , T ) + i = 1 d 1 1 ( β 0 , i 1 + ψ ˜ 1 , i * ) u d 1 s + i 1 ( t α 01 ) + γ 1 ( t α 01 ) , t [ α 01 T , α 01 ] z 2 ( ω 0 + ψ * , α 01 , T ) + 1 2 b ( β 0 + ψ ˜ * , T ) + i = 1 d 1 1 ( β 0 , i 1 + ψ ˜ 1 , i * ) u d 1 s + i 1 ( t α 01 ) + γ 1 ( t α 01 ) , t [ α 01 , T + α 01 ] ,

x 2 ( t ) = z 3 ( ω 0 + ψ * , α 02 , T ) + 1 2 T b ( β 0 + ψ ˜ * , T ) ( t α 02 ) + i = 1 d 2 1 ( β 0 , i 2 + ψ ˜ 2 , i * ) u d 2 + s + i 2 ( t α 02 ) + γ 2 ( t α 02 ) , t [ α 02 T , α 02 ] z 4 ( ω 0 + ψ * , α 02 , T ) + 1 2 T b ( β 0 + ψ ˜ * , T ) ( t α 02 ) + i = 1 d 2 1 ( β 0 , i 2 + ψ ˜ 2 , i * ) u d 2 + s + i 2 ( t α 02 ) + γ 2 ( t α 02 ) , t [ α 02 , T + α 02 ] ,

where ψ ˜ * = ( ψ ˜ 1 * , ψ ˜ 2 * ) R d 1 1 × R d 2 1 , α 0 = ( α 01 , α 02 ) .

Next we seek homoclinic solution near the unperturbed heteroclinic loop Γ .

3.2 Homoclinic solution bifurcated from heteroclinic loop Γ

In this section, we consider the homoclinic solution bifurcated from heteroclinic loop Γ . The homoclinic solution can be asymptotic to hyperbolic equilibrium P or asymptotic to hyperbolic equilibrium P + near the unperturbed heteroclinic loop Γ . No matter which equilibrium, the construct method of the homoclinic solutions is similar. Hence, in these situations, one of those classes is chosen as the proof. The construction of the homoclinic solutions is just a modification of the constructed above periodic solutions. For the completeness of the study, we will give it briefly. Next we will find a homoclinic solution asymptotic to P of equation (1.2). On the other hand, the method can be referred in [25].

Define a function b : R d 1 + d 2 2 × ( 0 , + ) R n by

b ( β , t ) = γ 2 ( t ) γ 1 ( t ) + i = 1 d 2 1 β i 2 u d 2 + s + i 2 ( t ) i = 1 d 1 1 β i 1 u d 1 s + i 1 ( t ) ,

where β = ( β 1 1 , , β d 1 1 1 , β 1 2 , , β d 2 1 2 ) . Note that

b ( β , t ) = O ( e M t ) ,

uniformly with respect to β from any bounded subset of R d 1 + d 2 2 . For i = 1 or i = 2 , we suppose x i ( t ) is a solution of equation (1.2). Fix T N and take the change in variable

(3.28) x 1 ( t + α 1 ) = γ 1 ( t ) + y 1 ( t ) + i = 1 d 1 1 β i 1 u d 1 s + i 1 ( t ) + 1 2 b ( β , T ) ,

(3.29) x 2 ( t + α 2 ) = γ 2 ( t ) + y 2 ( t ) + i = 1 d 2 1 β i 2 u d 2 + s + i 2 ( t ) 1 2 b ( β , T ) .

Then, y i ( t ) satisfies the following equation:

(3.30) y i ˙ = D f ( γ i ) y i + g ˜ i ( y i , β , μ , α i ) ,

with

(3.31) g ˜ 1 ( y 1 , β , μ , α 1 , T ) ( t ) = f ( γ 1 ( t ) + y 1 ( t ) + i = 1 d 1 1 β i 1 u d 1 s + i 1 ( t ) + 1 2 b ( β , T ) ) f ( γ 1 ( t ) ) D f ( γ 1 ( t ) ) y 1 ( t ) i = 1 d 1 1 β i 1 D f ( γ 1 ( t ) ) u d 1 s + i 1 ( t ) + j = 1 2 μ j g j ( γ 1 ( t ) + y 1 ( t ) + i = 1 d 1 1 β i 1 u d 1 s + i 1 ( t ) + 1 2 b ( β , T ) , μ , t + α 1 )

and

(3.32) g ˜ 2 ( y 2 , β , μ , α 2 , T ) ( t ) = f ( γ 2 ( t ) + y 2 ( t ) + i = 1 d 2 1 β i 2 u d 2 + s + i 2 ( t ) 1 2 b ( β , T ) ) f ( γ 2 ( t ) ) D f ( γ 2 ( t ) ) y 2 ( t ) i = 1 d 2 1 β i 2 D f ( γ 2 ( t ) ) u d 2 + s + i 2 ( t ) + j = 1 2 μ j g j ( γ 2 ( t ) + y 2 ( t ) + i = 1 d 2 1 β i 2 u d 2 + s + i 2 ( t ) 1 2 b ( β , T ) , μ , t + α 2 ) .

With regard to g ˜ i ( y i , β , μ , α i , T ) , a simple computation yields the same result as Lemma 3.1, so it is omitted.

By the definition of the function b i ( β , t ) , if y 1 ( T ) = y 2 ( T ) , then x 1 ( T + α 1 ) = x 2 ( T + α 2 ) . From (3.28) and (3.29), if lim t y 1 ( t ) = 0 and lim t + y 2 ( t ) = 0 , then lim t x 1 ( t + α 1 ) = lim t + x 2 ( t + α 2 ) = P . Therefore, to seek such solutions, define the following Banach spaces:

Z + = { z C 1 ( R + , R n ) : sup t R + z ( t ) e M t < } , Z = { z C 1 ( R , R n ) : sup t R z ( t ) e M t < } ,

with the norm z ± = sup t R ± z ( t ) e M t , for z ± Z ± . And M be the same as in (2.3) and (2.5).

For any η 2 , η 3 R n and ξ i R n , i = 1, 2, 3, 4, consider the functions z 1 Z , z 2 Y T , z 3 X T , z 4 Z + by

z 1 ( t ) = U 1 ( t ) P s u ξ 1 + U 1 ( t ) 0 t ( P s s 1 + P s u ) U 1 1 ( s ) g ˜ 1 ( z 1 , β , μ , α 1 , T ) ( s ) d s + U 1 ( t ) t ( P u u 1 + P u s + ) U 1 1 ( s ) g ˜ 1 ( z 1 , β , μ , α 1 , T ) ( s ) d s , z 2 ( t ) = U 1 ( t ) P u s + ξ 2 + U 1 ( t ) 0 t ( P s s 1 + P u s + ) U 1 1 ( s ) g ˜ 1 ( z 2 , β , μ , α 1 , T ) ( s ) d s + U 1 ( t ) ( P u u 1 + P s u ) U 1 1 ( T ) η 2 + U 1 ( t ) T t ( P u u 1 + P s u ) U 1 1 ( s ) g ˜ 1 ( z 2 , β , μ , α 1 , T ) ( s ) d s , z 3 ( t ) = U 2 ( t ) P s u + ξ 3 + U 2 ( t ) 0 t ( P s s 2 + P s u + ) U 2 1 ( s ) g ˜ 2 ( z 3 , β , μ , α 2 , T ) ( s ) d s + U 2 ( t ) ( P u u 2 + P u s ) U 2 1 ( T ) η 3 + U 2 ( t ) T t ( P u u 2 + P u s ) U 2 1 ( s ) g ˜ 2 ( z 3 , β , μ , α 2 , T ) ( s ) d s , z 4 ( t ) = U 2 ( t ) P u s ξ 4 + U 2 ( t ) 0 t ( P s s 2 + P u s ) U 2 1 ( s ) g ˜ 2 ( z 4 , β , μ , α 2 , T ) ( s ) d s + U 2 ( t ) t ( P u u 2 + P s u + ) U 2 1 ( s ) g ˜ 2 ( z 4 , β , μ , α 2 , T ) ( s ) d s .

Lemma 3.5

Given η 2 , η 3 , and ξ i , the functions z 1 , z 2 , z 3 , z 4 are solutions of equation (3.30) for i = 1 , 2, 3, 4.

We only need to show that the definitions of z 1 and z 4 are reasonable. And the others have the same proof as Lemma 3.2, so it is omitted. It is needed to verify the infinite integral is convergence in the definition of z 1 and z 4 . By (2.3) and (2.5), we know

U 1 ( t ) ( P u s + + P u u 1 ) U 1 1 ( s ) K 0 e 2 M ( s t ) , s t < 0 , U 2 ( t ) ( P s u + + P u u 2 ) U 2 1 ( s ) K 0 e 2 M ( t s ) , 0 < t s ,

and g ˜ i is bounded in suitable regions. So the infinite integral

U 1 ( t ) t ( P u u 1 + P u s + ) U 1 1 ( s ) g ˜ 1 ( z 1 , β , μ , α 1 , T ) ( s ) d s , U 2 ( t ) t ( P u u 2 + P s u + ) U 2 1 ( s ) g ˜ 2 ( z 4 , β , μ , α 2 , T ) ( s ) d s ,

are convergent. Hence the definition of z 1 , z 4 are reasonable.

Here η 2 , η 3 , and ξ i are arbitrary. If we can seek some η 2 , η 3 , and ξ i such that z 1 ( 0 ) = z 2 ( 0 ) , z 3 ( 0 ) = z 4 ( 0 ) , then z 1 ( t ) and z 2 ( t ) stick together at t = 0 and form a solution y 1 ( t ) of equation (3.30) in ( , T ] near the heteroclinic orbit γ 1 , z 3 ( t ) and z 4 ( t ) stick together and form a solution y 2 ( t ) of equation (3.30) in [ T , ) near the heteroclinic orbit γ 2 . By the definition of b i ( β , t ) , (3.28), and (3.29), we have x 1 ( T + α 1 ) = x 2 ( T + α 2 ) . Hence, equation (1.2) has a homoclinic solution asymptotic to P consisting of x 1 ( t ) and x 2 ( t ) near the unperturbed heteroclinic loop Γ . From the above description, we know that the homoclinic solution consists of z 1 ( t ) , z 2 ( t ) , z 3 ( t ) , and z 4 ( t ) near the heteroclinic loop Γ (Figure 5). The construction of homoclinic solution asymptotic to P + near the unperturbed heteroclinic loop Γ is similar, so it is omitted. We can decompose z 1 ( 0 ) = z 2 ( 0 ) into the following three equations:

(3.33) P s u ξ 1 = P s u U 1 1 ( T ) η 2 0 T P s u U 1 1 ( s ) g ˜ 1 d s ,

(3.34) P u s + ξ 2 = 0 P u s + U 1 1 ( s ) g ˜ 1 d s ,

(3.35) T P u u 1 U 1 1 ( s ) g ˜ 1 d s = P u u 1 U 1 1 ( T ) η 2 .

From z 3 ( 0 ) = z 4 ( 0 ) , we can obtain something similar

(3.36) P s u + ξ 3 = + 0 P s u + U 2 1 ( s ) g ˜ 2 d s ,

(3.37) P u s ξ 4 = P u s U 2 1 ( T ) η 3 + T 0 P u s U 2 1 ( s ) g ˜ 2 d s ,

(3.38) P u u 2 U 2 1 ( T ) η 3 + T + P u u 2 U 2 1 ( s ) g ˜ 2 d s = 0 .

From z 2 ( T ) = z 3 ( T ) , we can obtain

(3.39) U 1 ( T ) ( P u u 1 + P s u ) U 1 1 ( T ) η 2 U 2 ( T ) ( P u u 2 + P u s ) U 2 1 ( T ) η 3 = U 2 ( T ) 0 T ( P s s 2 + P s u + ) U 2 1 ( s ) g ˜ 2 d s U 1 ( T ) 0 T ( P s s 1 + P u s + ) U 1 1 ( s ) g ˜ 1 d s + U 2 ( T ) P s u + ξ 3 U 1 ( T ) P u s + ξ 2 ,

Figure 5 
                  
                     
                        
                           
                           
                              
                                 
                                    Z
                                 
                                 
                                    1
                                 
                              
                              
                                 (
                                 
                                    t
                                 
                                 )
                              
                           
                           {Z}_{1}\left(t)
                        
                     ,…,
                        
                           
                           
                              
                                 
                                    Z
                                 
                                 
                                    4
                                 
                              
                              
                                 (
                                 
                                    t
                                 
                                 )
                              
                           
                           {Z}_{4}\left(t)
                        
                      near the the heteroclinic loop 
                        
                           
                           
                              Γ
                           
                           \Gamma 
                        
                     .
Figure 5

Z 1 ( t ) ,…, Z 4 ( t ) near the the heteroclinic loop Γ .

In (3.39), taking the T approach infinity, we have

(3.40) C 0 0 0 I u 1 C 1 η 2 C I s 2 0 0 0 C 1 η 3 = U 2 ( ) 0 ( P s s 2 + P s u + ) U 2 1 ( s ) g ˜ 2 d s U 1 ( ) 0 ( P s s 1 + P u s + ) U 1 1 ( s ) g ˜ 1 d s ,

where (2.3) and (2.5) are used to make sure the existence of the limit, and I u 1 and I s 2 are ( n d s ) × ( n d s ) , ( d + s ) × ( d + s ) unit matrixes, respectively. Assume

C 1 η 2 = 0 k 2 , C 1 η 3 = k 1 0 , k = k 1 k 2 .

From (3.40), we can solve

k = C 1 ( U 2 ( ) 0 ( P s s 2 + P s u + ) U 2 1 ( s ) g ˜ 2 d s U 1 ( ) 0 ( P s s 1 + P u s + ) U 1 1 ( s ) g ˜ 1 d s ) .

Then, there exists T 0 > 0 so that we can solve η 2 and η 3 in (3.40) whenever T > T 0 . We denote that the solutions are η 2 * = η 2 * ( β , μ , α 1 , T ) and η 3 * = η 3 * ( β , μ , α 2 , T ) . And η j * satisfies lim T η j * ( 0 , 0 , α i , T ) = 0 for α i R and i = 1 , 2 , j = 2 , 3 .

Substituting η 1 * , η 2 * for η 1 , η 2 in (3.33), (3.34), (3.36), and (3.37), we can obtain ξ 1 , ξ 2 , ξ 3 , ξ 4 . Furthermore if η 1 * , η 2 * satisfy (3.35) and (3.38), then z 1 ( 0 ) = z 2 ( 0 ) , z 2 ( T ) = z 3 ( T ) , z 3 ( 0 ) = z 4 ( 0 ) , that is, (3.30) has a homoclinic solution consisting of z 1 ( t ) , z 2 ( t ) , z 3 ( t ) , and z 4 ( t ) . Hence, we have the following result for equation (1.2).

Lemma 3.6

Assume U 1 , U 2 , P u u 1 , P u u 2 be as in (2.3) and (2.5). There exists T 0 and if

(3.41) T P u u 1 U 1 1 ( s ) g ˜ 1 d s P u u 1 U 1 1 ( T ) η 1 * = 0 ,

(3.42) T + P u u 2 U 2 1 ( s ) g ˜ 2 d s P u u 2 U 2 1 ( T ) η 2 * = 0 ,

for T > T 0 , then equation (1.2) has a homoclinic solution near the unperturbed heteroclinic loop Γ , where g ˜ 1 and g ˜ 2 are as in (3.31) and (3.32).

Next, in a similar way, we give a sufficient condition for the existence of zeros of (3.41) and (3.42) and obtain the same bifurcation function as (3.21). Hence, the zeros of the bifurcation function correspond to the existence of a homoclinic solution for the perturbed equation. So, under the conditions of Lemma 3.4, we know that for ( α , T ) B 2 ( α 0 , r 0 ) × ( T 0 , ) , equation (1.2) has a homoclinic solution γ ( t ) near the heteroclinic loop Γ as follows:

γ ( t ) = x 1 ( t ) = z 1 ( ω 0 + ψ * , α 01 , T ) + i = 1 d 1 1 ( β 0 , i 1 + ψ ˜ 1 , i * ) u d 1 s + i 1 ( t α 01 ) + γ 1 ( t α 01 ) + 1 2 b 1 ( β 0 + ψ ˜ * , T ) , t ( , α 01 ] z 2 ( ω 0 + ψ * , α 01 , T ) + i = 1 d 1 1 ( β 0 , i 1 + ψ ˜ 1 , i * ) u d 1 s + i 1 ( t α 01 ) + γ 1 ( t α 01 ) + 1 2 b 1 ( β 0 + ψ ˜ * , T ) , t [ α 01 , T + α 01 ] x 2 ( t ) = z 3 ( ω 0 + ψ * , α 02 , T ) + i = 1 d 2 1 ( β 0 , i 2 + ψ ˜ 2 , i * ) u d 2 + s + i 2 ( t α 02 ) + γ 2 ( t α 02 ) 1 2 b 2 ( β 0 + ψ ˜ * , T ) , t [ α 02 T , α 02 ] z 4 ( ω 0 + ψ * , α 02 , T ) + i = 1 d 2 1 ( β 0 , i 2 + ψ ˜ 2 , i * ) u d 2 + s + i 2 ( t α 02 ) + γ 2 ( t α 02 ) 1 2 b 2 ( β 0 + ψ ˜ * , T ) , t [ α 02 , )

where ψ ˜ * = ( ψ ˜ 1 * , ψ ˜ 2 * ) R d 1 1 × R d 2 1 , α 0 = ( α 01 , α 02 ) .

In summary, we have demonstrated the perturbed equation (1.2) has periodic or homoclinic solution γ ( t ) near the unperturbed heteroclinic loop Γ .

Acknowledgement

We are grateful to the anonymous referees for the constructive comments on the manuscript.

  1. Funding information: This work was supported by the National Natural Science Foundation of China (Grant No. 11801343).

  2. Author contributions: Both authors equally contributed to this work. Both authors read and approved the final manuscript.

  3. Conflict of interest: The authors state no conflict of interest.

  4. Data availability statement: All relevant data analyzed during this study are included in this published article.

References

[1] A. J. Homburg and B. Sandstede, Homoclinic and heteroclinic bifurcations in vector fields, Chapter 8 in Handbook of Dynamical Systems, vol. 3, Elsevier, Amsterdam, 2010. 10.1016/S1874-575X(10)00316-4Suche in Google Scholar

[2] C. Bonatti, L. J. Daz, and M. Viana, Dynamics Beyond Uniform Hyperbolicity: A Global Geometric and Probabilistic Perspective, Springer-Verlag, Berlin, 2004. Suche in Google Scholar

[3] K. H. Alfsen, and J. Frøyland, Systematics of the Lorenz model at σ=10, Phys. Scr. 31 (1985), 15–20, DOI: https://doi.org/10.1088/0031-8949/31/1/003. 10.1088/0031-8949/31/1/003Suche in Google Scholar

[4] W. Zhang, B. Krauskopf, and V. Kirk, How to find a codimension-one heteroclinic cycle between two periodic orbits, Discrete Contin. Dyn. Syst. 32 (2012), 2825–2851, DOI: https://doi.org/10.3934/dcds.2012.32.2825. 10.3934/dcds.2012.32.2825Suche in Google Scholar

[5] S. N. Chow, J. K. Hale, and J. Mallet-Parret, An example of bifurcation to homoclinic orbits, J. Differential Equations 37 (1980), 551–573, DOI: https://doi.org/10.1016/0022-0396(80)90104-7. 10.1016/0022-0396(80)90104-7Suche in Google Scholar

[6] X. B. Lin, Lin’s method, Scholarpedia 3 (2008), no. 9, 6972, DOI: https://dx.doi.org/10.4249/scholarpedia.6972.10.4249/scholarpedia.6972Suche in Google Scholar

[7] J. Knobloch, Lin’s method for discrete dynamical systems, J. Difference Equ. Appl. 6 (2000), no. 5, 577–623, DOI: https://doi.org/10.1080/10236190008808247. 10.1080/10236190008808247Suche in Google Scholar

[8] H. P. Krishnan, Uniqueness of rapidly oscillating periodic solutions to a singularly perturbed differential-delay equation, Electron. J. Differential Equations 2000 (2000), no. 56, 1–18. Suche in Google Scholar

[9] B. E. Oldeman, R. A. Champneys, and B. Krauskopf, Homoclinic branch switching: A numerical implementation of Lin’s method, Internat. J. Bifur. Chaos Appl. Sci. Engrg. 13 (2003), no. 10, 2977–2999, DOI: https://doi.org/10.1142/S0218127403008326. 10.1142/S0218127403008326Suche in Google Scholar

[10] C. Zhu, From homoclinics to quasi-periodic solutions for ordinary differential equations, Proc. Roy. Soc. Edinburgh Sect. A 145 (2015), no. 5, 1091–1114, DOI: https://doi.org/10.1017/S0308210515000189. 10.1017/S0308210515000189Suche in Google Scholar

[11] C. Zhu and B. Long, The periodic solution bifurcated from homoclinic orbit for a coupled ordinary differential equations, Math. Methods Appl. Sci. 40 (2017), no. 8, 2834–2846, DOI: https://doi.org/10.1002/mma.4200. 10.1002/mma.4200Suche in Google Scholar

[12] C. Zhu and W. Zhang, Multiple chaos arising from single-parametric perturbation of a degenerate homoclinic orbit, J. Differential Equations 268 (2020), no. 10, 5672–5703, DOI: https://doi.org/10.1016/j.jde.2019.11.024. 10.1016/j.jde.2019.11.024Suche in Google Scholar

[13] S. N. Chow, B. Deng, and D. Terman, The bifurcation of homoclinic and periodic orbits from two heteroclinic orbits, SIAM J. Math. Anal. 21 (1990), no. 1, 179–204, DOI: https://doi.org/10.1137/0521010. 10.1137/0521010Suche in Google Scholar

[14] J. D. M. Rademacher, Homoclinic orbits near heteroclinic cycles with one equilibrium and one periodic orbit, J. Differential Equations 218 (2005), no. 2, 390–443, DOI: https://doi.org/10.1016/j.jde.2005.03.016. 10.1016/j.jde.2005.03.016Suche in Google Scholar

[15] Y. Jin, D. Zhang, N. Wang, and D. Zhu, Bifurcations of twisted of fine heteroclinic Loop for high-dimensional systems, J. Appl. Anal. Comput. 13 (2023), no. 5, 2906–2921, DOI: https://doi.org/10.11948/20230052. 10.11948/20230052Suche in Google Scholar

[16] D. M. Zhu and Y. Sun, Homoclinic and periodic orbits arising near the heteroclinic cycle connecting saddle-focus and saddle under reversible condition, Acta Math. Sin. (Engl. Ser.) 23 (2007), no. 8, 1495–1504, DOI: https://doi.org/10.1007/s10114-005-0746-7. 10.1007/s10114-005-0746-7Suche in Google Scholar

[17] I. S. Labouriau and A. A. P. Rodrigues, Dense heteroclinic tangencies near a Bykov cycle, J. Differential Equations 259 (2015), no. 11, 5875–5902, DOI: https://doi.org/10.1016/j.jde.2015.07.017. 10.1016/j.jde.2015.07.017Suche in Google Scholar

[18] J. Knobloch, J. S. W. Lamb, and K. N. Webster, Using Lin’s method to solve Bykov’s problems, J. Differential Equations 257 (2014), no. 8, 2984–3047, DOI: https://doi.org/10.1016/j.jde.2014.06.006. 10.1016/j.jde.2014.06.006Suche in Google Scholar

[19] B. Long and S. S. Xu, Persistence of the heteroclinic loop under periodic perturbation, Electron. Res. Arch. 31 (2023), no. 2, 1089–1105, DOI: https://doi.org/10.3934/era.2023054. 10.3934/era.2023054Suche in Google Scholar

[20] F. Chen, A. Oksasoglu, and Q. Wang, Heteroclinic tangles in time-periodic equations, J. Differential Equations 254 (2013), no. 3, 1137–1171, DOI: https://doi.org/10.1016/j.jde.2012.10.010. 10.1016/j.jde.2012.10.010Suche in Google Scholar

[21] I. S. Labouriau and A. A. P. Rodrigues, Bifurcations from an attracting heteroclinic cycle under periodic forcing, J. Differential Equations 269 (2020), no. 5, 4137–4174, DOI: https://doi.org/10.1016/j.jde.2020.03.024. 10.1016/j.jde.2020.03.024Suche in Google Scholar

[22] I. S. Labouriau and A. A. P. Rodrigues, Periodic forcing of a heteroclinic network, J. Differential Equations 35 (2023), 2951–2969, DOI: https://doi.org/10.1007/s10884-021-10054-w. 10.1007/s10884-021-10054-wSuche in Google Scholar

[23] X. Sun, Exact bound on the number of limit cycles arising from a periodic annulus bounded by a symmetric heteroclinic loop, J. Appl. Anal. Comput. 10 (2020), no. 1, 378–390, DOI: https://doi.org/10.11948/20190294. 10.11948/20190294Suche in Google Scholar

[24] Y. Xiong and G. Hu, Heteroclinic loop bifurcations by perturbing a class of Z2-equivariant quadratic switching Hamiltonian systems with nilpotent singular points, J. Math. Anal. Appl. 532 (2024), no. 2, 127977, DOI: https://doi.org/10.1016/j.jmaa.2023.127977. 10.1016/j.jmaa.2023.127977Suche in Google Scholar

[25] B. Long and Y. Yang, Applying Lin’s method to constructing heteroclinic orbits near the heteroclinic chain, Math. Methods Appl. Sci. 47 (2024), no. 12, 10235–10255, DOI: https://doi.org/10.1002/mma.10118. 10.1002/mma.10118Suche in Google Scholar

[26] W. Coppel, Dichotomies and Stability Theory, Springer-Verlag, Berlin, 1978. 10.1007/BFb0067780Suche in Google Scholar

[27] K. J. Palmer, Exponential dichotomies and transversal homoclinic points, J. Differential Equations 55 (1984), no. 2, 225–256, DOI: https://doi.org/10.1016/0022-0396(84)90082-2. 10.1016/0022-0396(84)90082-2Suche in Google Scholar

[28] M. Feckan and J. Gruendler, Bifurcation from homoclinic to periodic solutions in singular ordinary differential equations, J. Math. Anal. Appl. 246 (2000), no. 1, 245–264, DOI: https://doi.org/10.1006/jmaa.2000.6791. 10.1006/jmaa.2000.6791Suche in Google Scholar

Received: 2024-06-19
Revised: 2024-12-19
Accepted: 2025-01-25
Published Online: 2025-04-10

© 2025 the author(s), published by De Gruyter

This work is licensed under the Creative Commons Attribution 4.0 International License.

Artikel in diesem Heft

  1. Special Issue on Contemporary Developments in Graphs Defined on Algebraic Structures
  2. Forbidden subgraphs of TI-power graphs of finite groups
  3. Finite group with some c#-normal and S-quasinormally embedded subgroups
  4. Classifying cubic symmetric graphs of order 88p and 88p 2
  5. Simplicial complexes defined on groups
  6. Two-sided zero-divisor graphs of orientation-preserving and order-decreasing transformation semigroups
  7. Further results on permanents of Laplacian matrices of trees
  8. Special Issue on Convex Analysis and Applications - Part II
  9. A generalized fixed-point theorem for set-valued mappings in b-metric spaces
  10. Research Articles
  11. Dynamics of particulate emissions in the presence of autonomous vehicles
  12. The regularity of solutions to the Lp Gauss image problem
  13. Exploring homotopy with hyperspherical tracking to find complex roots with application to electrical circuits
  14. The ill-posedness of the (non-)periodic traveling wave solution for the deformed continuous Heisenberg spin equation
  15. Some results on value distribution concerning Hayman's alternative
  16. 𝕮-inverse of graphs and mixed graphs
  17. A note on the global existence and boundedness of an N-dimensional parabolic-elliptic predator-prey system with indirect pursuit-evasion interaction
  18. On a question of permutation groups acting on the power set
  19. Chebyshev polynomials of the first kind and the univariate Lommel function: Integral representations
  20. Blow-up of solutions for Euler-Bernoulli equation with nonlinear time delay
  21. Spectrum boundary domination of semiregularities in Banach algebras
  22. Statistical inference and data analysis of the record-based transmuted Burr X model
  23. A modified predictor–corrector scheme with graded mesh for numerical solutions of nonlinear Ψ-caputo fractional-order systems
  24. Dynamical properties of two-diffusion SIR epidemic model with Markovian switching
  25. Classes of modules closed under projective covers
  26. On the dimension of the algebraic sum of subspaces
  27. Periodic or homoclinic orbit bifurcated from a heteroclinic loop for high-dimensional systems
  28. On tangent bundles of Walker four-manifolds
  29. Regularity of weak solutions to the 3D stationary tropical climate model
  30. A new result for entire functions and their shifts with two shared values
  31. Freely quasiconformal and locally weakly quasisymmetric mappings in metric spaces
  32. On the spectral radius and energy of the degree distance matrix of a connected graph
  33. Solving the quartic by conics
  34. A topology related to implication and upsets on a bounded BCK-algebra
  35. On a subclass of multivalent functions defined by generalized multiplier transformation
  36. Local minimizers for the NLS equation with localized nonlinearity on noncompact metric graphs
  37. Approximate multi-Cauchy mappings on certain groupoids
  38. Multiple solutions for a class of fourth-order elliptic equations with critical growth
  39. A note on weighted measure-theoretic pressure
  40. Majorization-type inequalities for (m, M, ψ)-convex functions with applications
  41. Recurrence for probabilistic extension of Dowling polynomials
  42. Unraveling chaos: A topological analysis of simplicial homology groups and their foldings
  43. Global existence and blow-up of solutions to pseudo-parabolic equation for Baouendi-Grushin operator
  44. A characterization of the translational hull of a weakly type B semigroup with E-properties
  45. Some new bounds on resolvent energy of a graph
  46. Carmichael numbers composed of Piatetski-Shapiro primes in Beatty sequences
  47. The number of rational points of some classes of algebraic varieties over finite fields
  48. Singular direction of meromorphic functions with finite logarithmic order
  49. Pullback attractors for a class of second-order delay evolution equations with dispersive and dissipative terms on unbounded domain
  50. Eigenfunctions on an infinite Schrödinger network
  51. Boundedness of fractional sublinear operators on weighted grand Herz-Morrey spaces with variable exponents
  52. On SI2-convergence in T0-spaces
  53. Bubbles clustered inside for almost-critical problems
  54. Classification and irreducibility of a class of integer polynomials
  55. Existence and multiplicity of positive solutions for multiparameter periodic systems
  56. Averaging method in optimal control problems for integro-differential equations
  57. On superstability of derivations in Banach algebras
  58. Investigating the modified UO-iteration process in Banach spaces by a digraph
  59. The evaluation of a definite integral by the method of brackets illustrating its flexibility
  60. Existence of positive periodic solutions for evolution equations with delay in ordered Banach spaces
  61. Tilings, sub-tilings, and spectral sets on p-adic space
  62. The higher mapping cone axiom
  63. Continuity and essential norm of operators defined by infinite tridiagonal matrices in weighted Orlicz and l spaces
  64. A family of commuting contraction semigroups on l 1 ( N ) and l ( N )
  65. Pullback attractor of the 2D non-autonomous magneto-micropolar fluid equations
  66. Maximal function and generalized fractional integral operators on the weighted Orlicz-Lorentz-Morrey spaces
  67. On a nonlinear boundary value problems with impulse action
  68. Normalized ground-states for the Sobolev critical Kirchhoff equation with at least mass critical growth
  69. Decompositions of the extended Selberg class functions
  70. Subharmonic functions and associated measures in ℝn
  71. Some new Fejér type inequalities for (h, g; α - m)-convex functions
  72. The robust isolated calmness of spectral norm regularized convex matrix optimization problems
  73. Multiple positive solutions to a p-Kirchhoff equation with logarithmic terms and concave terms
  74. Joint approximation of analytic functions by the shifts of Hurwitz zeta-functions in short intervals
  75. Green's graphs of a semigroup
  76. Some new Hermite-Hadamard type inequalities for product of strongly h-convex functions on ellipsoids and balls
  77. Infinitely many solutions for a class of Kirchhoff-type equations
  78. On an uncertainty principle for small index subgroups of finite fields
Heruntergeladen am 5.9.2025 von https://www.degruyterbrill.com/document/doi/10.1515/math-2025-0134/html?lang=de
Button zum nach oben scrollen