Skip to main content

Boundedness of solutions for semilinear Duffing’s equation with asymmetric nonlinear term

Abstract

In this paper we study the following second-order periodic system:

x + V (x)+p(x)f(t)=0,

where V(x) has a singularity. Under some assumptions on the V(x), p(x) and f(t), by Ortega’s small twist theorem, we obtain the existence of quasi-periodic solutions and boundedness of all the solutions.

1 Introduction and main result

In the early 1960s, Littlewood [1] asked whether or not the solutions of the Duffing-type equations,

x +g(x,t)=0,
(1.1)

are bounded for all time, i.e., whether there are resonances that might cause the amplitude of the oscillations to increase without bound.

The first positive result of boundedness of solutions in the superlinear case (i.e., g ( x , t ) x as |x|) was due to Morris [2]. By means of KAM theorem, Morris proved that every solution of differential equation (1.1) is bounded if g(x,t)=2 x 3 p(t), where p(t) is piecewise continuous and periodic. This result relies on the fact that the nonlinearity 2 x 3 can guarantee the twist condition of KAM theorem. Later, several authors (see [3, 4]) improved the result of (1.1) and obtained a similar result for a large class of superlinear functions g(x,t).

When g(x,t) satisfies

0k g ( x , t ) x K+,xR,

i.e., differential equation (1.1) is semilinear, similar results also hold. But the proof is more difficult since there may be a resonant case.

Liu [5] studied the following equation:

( Φ p ( x ) ) +α Φ p ( x + ) +β Φ p ( x ) =f(x,t),

where f(x,t) is 2π-periodic in t and has limits f ± (t) as x±. Under some reasonable assumptions on f(x,t), Liu [5] proved the existence of quasi-periodic solutions and the boundedness of solutions. Later, Cheng and Xu [6] studied a more general equation

( Φ p ( x ) + Ψ ( x ) ) +α Φ p ( x + ) +β Φ p ( x )=ϕ(x,t),
(1.2)

where ϕ(x,t) is 2π-periodic in t. They defined a new function ϕ ¯ (t,x)= ϕ ( t , x ) | x | p 1 σ , where σ(0,p), ϕ ¯ (t,x) has limits ϕ ± (t) and the similar property to f(x,t) in [5]. Then the authors proved the boundedness of solutions for (1.2). We observe that ϕ(x,t) in [6] is unbounded while f(x,t) in [5] is bounded and that is the major difference between [5] and [6]. The idea in [5, 6] is to change the original problem to a Hamiltonian system and then use a twist theorem of area-preserving mapping to the Poincaré map.

Recently, Capietto et al. [7] studied the following equation:

x + V (x)=F(x,t),
(1.3)

where F(x,t)=p(t) is a π-periodic function, V(x)= 1 2 x + 2 + 1 ( 1 x 2 ) ν 1, x + =max{x,0}, x =max{x,0} and ν is a positive integer. Under the Lazer-Leach assumption that

1+ 1 2 0 π p( t 0 +θ)sinθdθ>0, t 0 R,
(1.4)

they proved the boundedness of solutions and the existence of a quasi-periodic solution by the Moser twist theorem. It was the first time that the equation of the boundedness of all solutions was treated in case of a singular potential.

Motivated by the papers [57], we observe that F(x,t)=p(t) in (1.3) is smooth and bounded, so a natural question is to find sufficient conditions on F(x,t) such that all solutions of (1.3) are bounded when F(x,t) is unbounded. The purpose of this paper is to deal with this problem.

We consider the following equation:

x + V (x)+p(x)f(t)=0,
(1.5)

where

V= 1 2 x + 2 + 1 1 x 2 1,x>1.
(1.6)

In order to state our main results, we give some notations and assumptions. Let f(t) be a π-periodic function and

lim x + p ( x ) | x | α =1,P(x)= 0 x p(s)ds,
(1.7)

where 0<α<1. We suppose that the following Lazer-Leach assumption holds:

0 π f( t 0 +θ) ( sin θ ) 1 + α dθ>0, t 0 R.
(1.8)

Our main result is the following theorem.

Theorem 1 Under assumptions (1.6)-(1.8), all the solutions of (1.5) are defined for all t(,+), and for each solution x(t), we have sup t R (|x(t)|+| x (t)|)<+.

The main idea of our proof is acquired from [8]. The proof of Theorem 1 is based on a small twist theorem due to Ortega [9]. Hypotheses (1.6)-(1.8) of our theorem are used to prove that the Poincaré mapping of (1.5) satisfies the assumptions of Ortega’s theorem.

Moreover, we have the following theorem on solutions of Mather type.

Theorem 2 Assume that f(t)C satisfies (1.8); then, there is ϵ 0 >0 such that for any ω( 1 π , 1 π + ϵ 0 ), equation (1.5) has a solution ( x ω (t), x ω (t)) of Mather type with rotation number ω. More precisely,

Case 1: ω= p q is rational. The solutions ( x ω (t+2iπ), x ω (t+2iπ)), 1iq1, are independent periodic solutions of periodic ; moreover, in this case,

lim q min t R ( | x ω ( t ) | + | x ω ( t ) | ) =+.

Case 2: ω is irrational. The solution ( x ω (t), x ω (t)) is either a usual quasi-periodic solution or a generalized one.

2 Proof of the theorem

2.1 Action-angle variables and some estimates

Observe that (1.5) is equivalent to the following Hamiltonian system:

x = H y , y = H x
(2.1)

with the Hamiltonian function

H(x,y,t)= 1 2 y 2 +V(x)+P(x)f(t).

In order to introduce action and angle variables, we first consider the auxiliary autonomous equation

x =y, y = V (x),
(2.2)

which is an integrable Hamiltonian system with the Hamiltonian function

H 1 (x,y,t)= 1 2 y 2 +V(x).

The closed curves H 1 (x,y,t)=h>0 are just the integral curves of (2.2).

Denote by T 0 (h) the time period of the integral curve Γ h of (2.2) defined by H 1 (x,y,t)=h and by I the area enclosed by the closed curve Γ h for every h>0. Let 1< α h <0< β h be such that V( α h )=V( β h )=h. It is easy to see that

I 0 (h)=2 α h β h 2 ( h V ( s ) ) ds,h>0,

and

T 0 (h)= I 0 (h)=2 α h β h 1 2 ( h V ( s ) ) ds,h>0.

By a direct computation, we get

I 0 ( h ) = 2 0 β h 2 ( h V ( s ) ) d s + 2 α h 0 2 ( h V ( s ) ) d s = π h + 2 0 α h 2 ( h V ( s ) ) d s ,

so

T 0 (h)=π+ 0 α h 1 2 ( h V ( s ) ) ds.

We then have

I 0 (h)= I (h)+ I + (h), T 0 (h)= T (h)+ T + (h),

where

I ( h ) = 2 0 α h 2 ( h V ( s ) ) d s , I + ( h ) = π h , T ( h ) = 2 0 α h 1 2 ( h V ( s ) ) d s , T + ( h ) = π .

We now give the estimates on the functions I and T .

Lemma 1 We have

h n | d n T ( h ) d h n | C h 1 2 ,

and

h n | d n I ( h ) d h n | C h 1 2 ,

where n=0,1,,6, h+. Note that here and below we always use C, C 0 or C 0 to indicate some constants.

Proof Now we estimate the first inequality. We choose V ( s ) h =η as the new variable of integration, then we have

T (h)= α h 0 1 2 ( h V ( s ) ) ds= 0 1 h V ( s ( η , h ) ) 1 2 ( 1 η ) dη.

Since V(s)= 1 1 s 2 1 and V ( s ) h =η, we have s= η h 1 + η h . By a direct computation, we have

V (s)= 2 s ( 1 s 2 ) 2 = 2 η h ( 1 + η h ) 2 1 + η h ,

then we get

T ( n ) (h)= ( 3 2 ) ! ( 3 2 n ) ! 0 1 η n 2 η ( 1 η ) ( 1 + η h ) 3 2 + n dη,n=0,1,,6.

When 0η h 1 and h is sufficient large, there exits C 0 such that 1η> C 0 , so we have

0 h 1 η n 2 η ( 1 η ) ( 1 + η h ) 3 2 + n d η C 0 h 1 η n 2 η ( 1 η ) d η C C 0 0 h 1 η n 1 2 d η C h 1 2 n .
(2.3)

Since h 2 3 η1, we have

h 1 3 <1+ h 1 3 1+ηh1+h,

then

h 2 3 1 η n 2 η ( 1 η ) ( 1 + η h ) 3 2 + n d η C h 2 3 1 η n h n 2 η ( 1 η ) h n ( 1 + η h ) n ( 1 + η h ) 3 2 d η C h 2 3 1 1 2 η ( 1 η ) h n ( 1 + η h ) 3 2 d η C h 2 3 1 1 2 η ( 1 η ) h n h 1 2 d η C h 1 2 n 0 1 1 2 η ( 1 η ) d η C h 1 2 n .
(2.4)

Observing that there is C 0 >0 such that 1 η C 0 when h 1 η h 2 3 and h+, we have

h 1 h 2 3 η n 2 η ( 1 η ) ( 1 + η h ) 3 2 + n d η C 1 h 3 2 n h 1 h 2 3 1 2 η ( 1 η ) η 3 2 d η C 1 C 0 h 3 2 n h 1 h 2 3 1 η 2 d η = C 1 C 0 h 3 2 n 1 η | h 1 h 2 3 = C 1 C 0 h 3 2 n ( h h 2 3 ) C h 1 2 n .
(2.5)

By (2.3)-(2.5) we have T ( n ) (h)C h 1 2 n , n=0,1,,6.

The proof of the second inequality is similar to that of the first one, so we only give a brief proof. We choose V ( s ) h =η as the new variable of integration, so we have

s h = η V ,s= η h 1 + η h

and

V (s)= 2 s ( 1 s 2 ) 2 = 2 η h ( 1 + η h ) 2 1 + η h .

By a direct computation, we have

I (h)=2 α h 0 2 ( h V ( s ) ) ds=h 0 1 2 ( 1 η ) η ( 1 + η h ) 3 2 dη.
(2.6)

By (2.6), we can easily get

I ( n ) ( h ) = I 1 ( n ) ( h ) + I 2 ( n ) ( h ) = n ( 3 2 ) ! ( 3 2 n + 1 ) ! 0 1 2 ( 1 η ) η η n 1 ( 1 + η h ) 3 2 + n 1 d η + ( 3 2 ) ! ( 3 2 n ) ! h 0 1 2 ( 1 η ) η η n ( 1 + η h ) 3 2 + n d η ,
(2.7)

where n=0,1,,6.

By a similar way to that in estimating T ( n ) (h), we get

I 1 ( n ) (h)C h 1 2 n , I 2 ( n ) (h)C h 1 2 n ,

which means that

I ( n ) (h)C h 1 2 n ,n=0,1,,6.

Thus Lemma 1 is proved. □

Remark 1 It follows from the definitions of T + (h), T (h) and Lemma 1 that

lim h + T (h)=0, lim h + T + (h)=π.

Thus the time period T 0 (h) is dominated by T + (h) when h is sufficiently large. From the relation between T (h) and I (h), we know I 0 (h) is dominated by I + (h) when h is sufficiently large.

Remark 2 It also follows from the definition of I(h), I (h), I + (h) and Remark 1 that

| h n d n I 0 ( h ) d h n | C 0 I 0 (h)for n1.

Remark 3 Note that h= h 0 ( I 0 ) is the inverse function of I 0 . By Remark 2, we have

| I n d n h ( I ) d I n | C 0 h(I)for n1.

We now carry out the standard reduction to the action-angle variables. For this purpose, we define the generating function S(x,I)= C 2 ( h V ( s ) ) ds, where C is the part of the closed curve Γ h connecting the point on the y-axis and point (x,y).

We define the well-know map (θ,I)(x,y) by

y= S x (x,I),θ= S I (x,I),

which is symplectic since

d x d y = d x ( S x x d x + S x I d I ) = S x I d x d I , d θ d I = ( S I x d x + S I I d I ) d I = S I x d x d I .

From the above discussion, we can easily get

θ={ π T 0 ( h ( x , y ) ) ( T ( h ( x , y ) ) 2 + arcsin x 2 ( h ( x , y ) ) ) if  x > 0 , y > 0 , π T 0 ( h ( x , y ) ) ( T ( h ( x , y ) ) 2 + π + arcsin x 2 ( h ( x , y ) ) ) if  x > 0 , y < 0 , π T 0 ( h ( x , y ) ) ( α h x 1 2 ( h ( x , y ) + 1 ( 1 s 2 ) 1 ) d s ) if  x < 0 , y > 0 , π T 0 ( h ( x , y ) ) ( T 0 ( h ( x , y ) ) α h x 1 2 ( h ( x , y ) + 1 ( 1 s 2 ) 1 ) d s ) if  x < 0 , y < 0
(2.8)

and

I(x,y)= I 0 ( h ( x , y ) ) =2 α h β h 2 ( h ( x , y ) V ( s ) ) ds.
(2.9)

In the new variables (θ,I), system (2.1) becomes

θ = H I , I = H θ ,
(2.10)

where

H(θ,I,t)=π h 0 (I)+πP ( x ( I , θ ) ) f(t).
(2.11)

In order to estimate πP(I,θ), we need the estimate on the functions x(I,θ).

Lemma 2 For I sufficient large and α h x<0, the following estimates hold:

| I n n x ( I , θ ) I n | c I for 0n6.

The lemma was first proved in [3], later Capietto et al. [7] gave a different proof; using the method of induction-hypothesis, Jiang and Fang [10] also gave another proof. So, for concision, we omit the proof.

2.2 New action and angle variables

Now we are concerned with Hamiltonian system (2.10) with the Hamiltonian function H(θ,I,t) given by (2.11). Note that

IdθHdt=(HdtIdθ).

This means that if one can solve I from (2.11) as a function of H (θ and t as parameters), then

d H d θ = I t (t,H,θ), d t d θ = I H (t,H,θ)
(2.12)

is also a Hamiltonian system with the Hamiltonian function I and now the action, angle and time variables are H, t and θ.

From (2.11) and Lemma 1, we have

H I 1as I+.

So, we assume that I can be written as

I= I 0 ( H π + R ( H , t , θ ) ) ,

where R satisfies |R|< H π . Recalling that h 0 is the inverse function of I 0 , we have

H π +R(H,t,θ)= h 0 (I),

which implies that

R(H,t,θ)=P ( x ( I , θ ) ) f(t).

As a consequence, R is implicitly defined by

R(H,t,θ)=P [ x ( I 0 ( H π + R ( H , t , θ ) ) , θ ) ] f(t).
(2.13)

Now we give the estimates of R. By a similar way to that in estimating Lemma 2.3 in [7], we have the following lemma.

Lemma 3 The function R(H,t,θ) satisfies the following estimates:

| m + l R ( H , t , θ ) H m t l | H α + 1 2 for m+l6.

Moreover, by the implicit function theorem, there exists a function R 1 = R 1 (t,H,θ) such that

R(H,t,θ)=P ( x ( H , θ ) ) f(t)+ R 1 (H,t,θ).

Since

R 1 ( H , t , θ ) = R ( H , t , θ ) P ( x ( H , θ ) ) f ( t ) = P { x [ I 0 ( H π + R ( H , t , θ ) ) , θ ] } f ( t ) P ( x ( H , θ ) ) f ( t ) = 0 1 p { x [ H + s ( π R + I ) , θ ] } x I ( H + s ( π R + I ) , θ ) ( π R + I ) f ( t ) d s .

By Lemmas 1 and 3, we have the estimates on R 1 (H,t,θ).

Lemma 4 | k + l R 1 ( H , t , θ ) k H l t |< H α 2 for k+l6.

For the estimate of I( H π +R), we need the estimate on I ( H π +R). By Lemma 1 and noticing that |R|< H π , we have the following lemma.

Lemma 5 | k + l I ( H π + R ) k H l t |< H 1 2 for k+l6.

Now the new Hamiltonian function I=I(t,H,θ) is written in the form

I = I 0 ( H π + R ) = I + ( H π + R ) + I ( H π + R ) = H + π R ( H , t , θ ) + I ( H π + R ) = H + π P ( x ( H , θ ) ) f ( t ) + π R 1 ( H , t , θ ) + I ( H π + R ) .

System (2.12) is of the form

{ d t d θ = I H = 1 + π x H ( H , θ ) p ( x ( H , θ ) ) f ( t ) + π R 1 H ( H , t , θ ) + I H ( H , t , θ ) , d H d θ = I t = π P ( x ( θ , H ) ) f ( t ) π R 1 t ( t , H , θ ) I t ( H , t , θ ) .
(2.14)

Introduce a new action variable ρ[1,2] and a parameter ϵ>0 by H= ϵ 2 ρ. Then H10<ϵ1. Under this transformation, system (2.14) is changed into the form

{ d t d θ = I H = 1 + π x H ( H , θ ) p ( x ( H , θ ) ) f ( t ) + π R 1 H ( H , t , θ ) + I H ( H , t , θ ) , d ρ d θ = I t = ϵ 2 [ π P ( x ( θ , H ) ) f ( t ) + π R 1 t ( H , t , θ ) + I t ( H , t , θ ) ] ,
(2.15)

which is also a Hamiltonian system with the new Hamiltonian function

Γ(t,ρ,θ;ϵ)=ρ+π ϵ 2 P ( x ( θ , ϵ 2 ρ ) ) f(t)+π ϵ 2 R 1 ( ϵ 2 ρ , θ , t ) + ϵ 2 I ( ϵ 2 ρ , θ , t ) .

Obviously, if ϵ1, the solution (t(θ, t 0 , ρ 0 ),ρ(θ, t 0 , ρ 0 )) of (2.15) with the initial data ( t 0 , ρ 0 )R×[1,2] is defined in the interval θ[0,2π] and ρ(θ, t 0 , ρ 0 )[ 1 2 ,3]. So, the Poincaré map of (2.15) is well defined in the domain R×[1,2].

Lemma 6 ([8] Lemma 5.1)

The Poincaré map of (2.15) has the intersection property.

The proof is similar to the corresponding one in [8].

For convenience, we introduce the notation O k (1) and o k (1). We say a function f(t,ρ,θ,ϵ) O k (1) if f is smooth in (t,ρ) and for k 1 + k 2 k,

| k 1 + k 2 t k 1 ρ k 2 f ( t , ρ , θ , ϵ ) | C,

for some constant C>0 which is independent of the arguments t, ρ, θ, ϵ.

Similarly, we say f(t,ρ,θ,ϵ) o k (1) if f is smooth in (t,ρ) and for k 1 + k 2 k,

lim ϵ 0 | k 1 + k 2 t k 1 ρ k 2 f ( t , ρ , θ , ϵ ) | =0,

uniformly in (t,ρ,θ).

2.3 Poincaré map and twist theorems

We will use Ortega’s small twist theorem to prove that the Poincaré map P has an invariant closed curve if ϵ is sufficiently small. Let us first recall the theorem in [9].

Lemma 7 (Ortega’s theorem)

Let A= S 1 ×[a,b] be a finite cylinder with universal cover A=R×[a,b]. The coordinate in is denoted by (τ,ν). Consider the map

f ¯ :AS×R.

We assume that the map has the intersection property. Suppose that f:AR×R, ( τ 0 , ν 0 )( τ 1 , ν 1 ) is a lift of f ¯ and it has the form

{ τ 1 = τ 0 + 2 N π + δ l 1 ( τ 0 , ν 0 ) + δ g ˜ 1 ( τ 0 , ν 0 ) , ν 1 = ν 0 + δ l 2 ( τ 0 , ν 0 ) + δ g ˜ 2 ( τ 0 , ν 0 ) ,
(2.16)

where N is an integer, δ(0,1) is a parameter. The functions l 1 , l 2 , g ˜ 1 and g ˜ 2 satisfy

l 1 C 6 ( A ) , l 1 ( τ 0 , ν 0 ) > 0 , l 1 ν 0 ( τ 0 , ν 0 ) > 0 , ( τ 0 , ν 0 ) A , l 2 ( , ) , g ˜ 1 ( , , ϵ ) , g ˜ 2 ( , , ϵ ) C 5 ( A ) .
(2.17)

In addition, we assume that there is a function I:AR satisfying

I C 6 (A), I ν 0 ( τ 0 , ν 0 )>0,( τ 0 , ν 0 )A
(2.18)

and

l 1 ( τ 0 , ν 0 ) I τ o ( τ 0 , ν 0 )+ l 2 ( τ 0 , ν 0 ) I ν 0 ( τ 0 , ν 0 )=0,( τ 0 , ν 0 )A.
(2.19)

Moreover, suppose that there are two numbers a ˜ and b ˜ such that a< a ˜ < b ˜ <b and

I M (a)< I m ( a ˜ ) I M ( a ˜ )< I m ( b ˜ ) I M ( b ˜ )< I m (b),
(2.20)

where

I M (r)= max ρ S 1 I( ρ o , τ o ), I m (r)= min ρ S 1 I( ρ o , τ o ).

Then there exist ϵ>0 and Δ>0 such that if δ<Δ and

g ˜ 1 ( , , ϵ ) C 5 ( A ) + g ˜ 2 ( , , ϵ ) C 5 ( A ) <ϵ,

the mapping f ¯ has an invariant curve in Γ A , the constant ϵ is independent of δ.

We make the ansatz that the solution of (2.15) with the initial condition (t(0),ρ(0))=( t 0 , ρ 0 ) is of the form

t= t 0 +θ+ ϵ 1 α Σ 1 ( t 0 , ρ 0 ,θ;ϵ),ρ= ρ 0 + ϵ 1 α Σ 2 ( t 0 , ρ 0 ,θ;ϵ).

Then the Poincaré map of (2.15) is

P: t 1 = t 0 +2π+ ϵ 1 α Σ 1 ( t 0 , ρ 0 ,2π;ϵ), ρ 1 = ρ 0 + ϵ 1 α Σ 2 ( t 0 , ρ 0 ,2π;ϵ).
(2.21)

The functions Σ 1 and Σ 2 satisfy

{ Σ 1 = π ϵ α 1 0 θ x H ( ϵ 2 ρ , θ ) p ( x ( ϵ 2 ρ , θ ) ) f ( t ) d θ Σ 1 = + π ϵ α 1 0 θ ( R 1 H ( ϵ 2 ρ , t , θ ) + I H ( ϵ 2 ρ , t , θ ) ) d θ , Σ 2 = π ϵ α + 1 0 θ P ( x ( ϵ 2 ρ , θ ) ) f ( t ) d θ Σ 2 = ϵ α + 1 0 θ ( π R 1 t ( ϵ 2 ρ , t , θ ) I t ( ϵ 2 ρ , t , θ ) ) d θ ,
(2.22)

where t= t 0 +θ+ ϵ 1 α Σ 1 , ρ= ρ 0 + ϵ 1 α Σ 2 . By Lemmas 4, 6 and 7, we know that

| Σ 1 |+| Σ 2 |Cfor θ[0,2π].
(2.23)

Hence, for ρ 0 [1,2], we may choose ϵ sufficiently small such that

ρ 0 +ϵ Σ 2 ρ 0 2 1 2 .
(2.24)

Moreover, we can prove that

Σ 1 , Σ 2 O 6 (1).
(2.25)

Similar to the way of estimating R 1 , by a direct calculation, we have the following lemma.

Lemma 8 The following estimates hold:

P ( x ( ϵ 2 ρ , θ ) ) P ( x ( ϵ 2 ρ 0 , θ ) ) ϵ α O 6 ( 1 ) , x H ( ϵ 2 ρ , θ ) p ( x ( ϵ 2 ρ , θ ) ) x H ( ϵ 2 ρ 0 , θ ) p ( x ( ϵ 2 ρ 0 , θ ) ) ϵ 2 α O 6 ( 1 ) .

Now we turn to give an asymptotic expression of the Poincaré map of (2.14), that is, we study the behavior of the functions Σ 1 and Σ 2 at θ=π as ϵ0. In order to estimate Σ 1 and Σ 2 , we need to introduce the following definition and lemma. Let

Θ + (I)=meas { θ [ 0 , π ] , x ( H 0 , θ ) > 0 } , Θ (I)= T 0 Θ + (I),

where H 0 = ϵ 2 ρ 0 .

Lemma 9

Θ + (I)=π+ϵ O 6 (1), Θ (I)=ϵ O 6 (1).

Proof This lemma was proved in [7], so we omit the details. □

For estimate Σ 1 and Σ 2 , we need the estimates of x and x H .

We recall that when x<0, we have

| x ( H 0 , θ ) | = O 6 (1), | x H ( H 0 , θ ) | = ϵ 2 O 5 (1).

When x>0, by the definition of θ, we have

arcsin x ( H 0 , θ ) 2 h = T 0 ( h ) π θ T ( h ) 2 =θ+ ϵ 2 O 5 (1),

which yields that

x ( H 0 , θ ) = 2 H 0 π sin θ + O 5 ( 1 ) , x H ( H 0 , θ ) = 1 2 H 0 π sin θ + ϵ 2 O 5 ( 1 ) .

Now we can give the estimates of Σ 1 and Σ 2 .

Lemma 10 The following estimates hold true:

Σ 1 ( t 0 , ρ 0 , 2 π ; ϵ ) = ( π 2 ρ 0 ) α 1 2 0 π ( sin θ ) 1 + α f ( t 0 + θ ) d θ + o 6 ( 1 ) , Σ 2 ( t 0 , ρ 0 , 2 π ; ϵ ) = π 1 α 2 ( 2 ρ 0 ) α + 1 2 0 π ( sin θ ) 1 + α f ( t 0 + θ ) d θ + o 6 ( 1 )

for ϵ0.

Proof Firstly we consider Σ 1 . By Lemmas 3, 4, 8 and (2.22), we have

Σ 1 ( t 0 , ρ 0 , 2 π ; ϵ ) = π ϵ α 1 0 π x H ( ϵ 2 ρ , θ ) p ( x ( ϵ 2 ρ , θ ) ) f ( t ) d θ + ϵ α 1 0 π π R 1 H ( x ( ϵ 2 ρ , θ ) , t ) + I H ( x ( ϵ 2 ρ , θ ) , t ) d θ = π ϵ α 1 0 π x H ( ϵ 2 ρ 0 , θ ) p ( x ( ϵ 2 ρ 0 , θ ) ) f ( t 0 + θ ) d θ + ϵ α O 6 ( 1 ) = π ϵ α 1 Θ + x H ( ϵ 2 ρ 0 , θ ) p ( x ( ϵ 2 ρ 0 , θ ) ) f ( t 0 + θ ) d θ + π ϵ α 1 Θ x H ( ϵ 2 ρ 0 , θ ) p ( x ( ϵ 2 ρ 0 , θ ) ) f ( t 0 + θ ) d θ + ϵ α O 6 ( 1 ) .

Since lim x + p ( x ) | x | α =1 and ϵ0 means x, we have

π ϵ α 1 Θ + x H ( ϵ 2 ρ 0 , θ ) p ( x ( ϵ 2 ρ 0 ) , θ ) f ( t 0 + θ ) d θ = π ϵ α 1 Θ + x H ( θ , ϵ 2 ρ ) | x | α f ( t 0 + θ ) d θ + ϵ α O 6 ( 1 ) .
(2.26)

By the measure of Θ , we have

π ϵ α 1 Θ x H ( ϵ 2 ρ 0 , θ ) p ( x ( ϵ 2 ρ 0 , θ ) ) f( t 0 +θ)dθ= ϵ α O 6 (1).
(2.27)

By (2.26) and (2.27), we have

Σ 1 ( t 0 , ρ 0 , 2 π ; ϵ ) = π ϵ α 1 Θ + x H ( θ , ϵ 2 ρ ) | x | α f ( t 0 + θ ) d θ + ϵ α O 6 ( 1 ) = π ϵ α 1 0 π x H ( θ , ϵ 2 ρ ) | x | α f ( t 0 + θ ) d θ + ϵ α O 6 ( 1 ) = ( π 2 ρ 0 ) 1 α 2 0 π ( sin θ ) α + 1 f ( t 0 + θ ) d θ + o 6 ( 1 ) .

Now we consider Σ 2 . By Lemmas 3, 4, 8 and (2.22), we have

Σ 2 ( t 0 , ρ 0 , 2 π ; ϵ ) = π ϵ α + 1 0 π P ( x ( θ , ϵ 2 ρ ) ) f ( t ) d θ ϵ α + 1 0 π [ π R 1 t ( x ( θ , ϵ 2 ρ ) , t ) + I t ( x ( θ , ϵ 2 ρ ) , t ) ] d θ = π ϵ α + 1 0 π P ( x ( θ , ϵ 2 ρ 0 ) ) f ( t 0 + θ ) d θ + ϵ α O 6 ( 1 ) = π ϵ α + 1 Θ + P ( x ( θ , ϵ 2 ρ 0 ) ) f ( t 0 + θ ) d θ π ϵ α + 1 Θ P ( x ( θ , ϵ 2 ρ 0 ) ) f ( t 0 + θ ) d θ + ϵ α O 6 ( 1 ) .

By (1.7) and ϵ0 means x, we have

π ϵ α + 1 Θ + P ( x ( θ , ϵ 2 ρ 0 ) ) f ( t 0 + θ ) d θ = π ϵ α + 1 α + 1 Θ + | x ( θ , ϵ 2 ρ 0 ) | α x ( θ , ϵ 2 ρ 0 ) f ( t 0 + θ ) d θ + ϵ α O 6 ( 1 ) .
(2.28)

By the measure of Θ , we have

π ϵ α + 1 Θ P ( x ( θ , ϵ 2 ρ 0 ) ) f ( t 0 +θ)dθ= ϵ α O 6 (1).
(2.29)

By (2.28) and (2.29), we have

Σ 2 = π ϵ α + 1 α + 1 Θ + | x ( θ , ϵ 2 ρ 0 ) | α x ( θ , ϵ 2 ρ 0 ) f ( t 0 + θ ) d θ + ϵ α O 6 ( 1 ) = π ϵ α + 1 α + 1 0 π | x ( θ , ϵ 2 ρ 0 ) | α x ( θ , ϵ 2 ρ 0 ) f ( t 0 + θ ) d θ + ϵ α O 6 ( 1 ) = 1 α + 1 π 1 α 2 ( 2 ρ 0 ) α + 1 2 0 π ( sin θ ) 1 + α f ( t 0 + θ ) d θ + o 6 ( 1 ) .

Thus Lemma 10 is proved. □

2.4 Proof of Theorem 1

Let

Ψ 1 ( t 0 , ρ 0 ) = ( π 2 ρ 0 ) 1 α 2 0 π ( sin θ ) 1 + α f ( t 0 + θ ) d θ , Ψ 2 ( t 0 , ρ 0 ) = 1 α + 1 π 1 α 2 ( 2 ρ 0 ) α + 1 2 0 π ( sin θ ) 1 + α f ( t 0 + θ ) d θ .

Then there are two functions ϕ 1 and ϕ 2 such that the Poincaré map of (2.15), given by (2.21), is of the form

P : t 1 = t 0 + 2 π + ϵ 1 α Ψ 1 ( t 0 , ρ 0 ) + ϵ 1 α ϕ 1 , ρ 1 = ρ 0 + ϵ 1 α Ψ 2 ( t 0 , ρ 0 ) + ϵ 1 α ϕ 2 ,
(2.30)

where ϕ 1 , ϕ 2 o 6 (1).

Since 0 π p( t 0 +θ)sinθdθ>0, t 0 R, we have

Ψ 1 >0, Ψ 1 ρ 0 0.

Let

L= ρ 0 1 + α 2 0 π ( sin θ ) 1 + α f ( t 0 + θ ) d θ .

Then

L t 0 Ψ 1 ( t 0 , ρ 0 )+ L ρ 0 Ψ 2 ( t 0 , ρ 0 )=0.

The other assumptions of Ortega’s theorem are easily verified. Hence, there is an invariant curve of P in the annulus ( t 0 , ρ 0 ) S 1 ×[1,2], which implies the boundedness of our original equation (1.5). Then Theorem 1 is proved.

2.5 Proof of Theorem 2

We apply Aubry-Mather theory. By Theorem B in [11] and the monotone twist property of the Poincaré map P guaranteed by Ψ 1 ρ 0 <0, it is straightforward to check that Theorem 2 is correct.

References

  1. Littlewood J: Unbounded solutions of y +g(y)=p(t). J. Lond. Math. Soc. 1966, 41: 133–149.

    Google Scholar 

  2. Morris GR: A case of boundedness of Littlewood’s problem on oscillatory differential equations. Bull. Aust. Math. Soc. 1976, 14: 71–93. 10.1017/S0004972700024862

    Article  Google Scholar 

  3. Levi M: Quasiperiodic motions in superquadratic time-periodic potential. Commun. Math. Phys. 1991, 144: 43–82.

    Article  Google Scholar 

  4. Dieckerhoff R, Zehnder E: Boundedness of solutions via the twist theorem. Ann. Sc. Norm. Super. Pisa, Cl. Sci. 1987, 14: 79–95.

    MathSciNet  Google Scholar 

  5. Liu B: Boundedness of solutions for equations with p -Laplacian and an asymmetric nonlinear term. J. Differ. Equ. 2004, 207: 73–92. 10.1016/j.jde.2004.06.023

    Article  Google Scholar 

  6. Cheng C, Xu J: Boundedness of solutions for a second-order differential equations. Nonlinear Anal. 2008, 7: 1993–2004.

    MathSciNet  Article  Google Scholar 

  7. Capietto A, Dambrosio W, Liu B: On the boundedness of solutions to a nonlinear singular oscillator. Z. Angew. Math. Phys. 2009, 60(6):1007–1034. 10.1007/s00033-008-8094-y

    MathSciNet  Article  Google Scholar 

  8. Liu B: Quasi-periodic solutions of forced isochronous oscillators at resonance. J. Differ. Equ. 2009, 246: 3471–3495. 10.1016/j.jde.2009.02.015

    Article  Google Scholar 

  9. Ortega R: Boundedness in a piecewise linear oscillator and a variant of the small twist theorem. Proc. Lond. Math. Soc. 1999, 79: 381–413. 10.1112/S0024611599012034

    Article  Google Scholar 

  10. Jiang S, Fang F: Lagrangian stability of a class of second-order periodic systems. Abstr. Appl. Anal. 2011., 2011: Article ID 106214

    Google Scholar 

  11. Pei ML: Aubry-Mather sets for finite-twist maps of a cylinder and semilinear Duffing equations. J. Differ. Equ. 1994, 113: 106–127. 10.1006/jdeq.1994.1116

    Article  Google Scholar 

Download references

Acknowledgements

Thanks are given to referees whose comments and suggestions were very helpful for revising our paper.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Shunjun Jiang.

Additional information

Competing interests

The authors declare that they have no competing interests.

Authors’ contributions

The article is a joint work of three authors who contributed equally to the final version of the paper. All authors read and approved the final manuscript.

Rights and permissions

Open Access This article is distributed under the terms of the Creative Commons Attribution 2.0 International License (https://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Reprints and Permissions

About this article

Cite this article

Jiang, S., Rao, F. & Shi, Y. Boundedness of solutions for semilinear Duffing’s equation with asymmetric nonlinear term. J Inequal Appl 2013, 476 (2013). https://doi.org/10.1186/1029-242X-2013-476

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/1029-242X-2013-476

Keywords

  • boundedness of solutions
  • singularity
  • small twist theorem