necessary and sufficient conditions for the boundedness of solutions to two-dimensional quadratic...

3
ISSN 10645624, Doklady Mathematics, 2010, Vol. 81, No. 1, pp. 31–33. © Pleiades Publishing, Ltd., 2010. Original Russian Text © G.A. Leonov, 2010, published in Doklady Akademii Nauk, 2010, Vol. 430, No. 2, pp. 157–159. 31 For nonlinear dynamical systems, necessary and sufficient conditions for the global asymptotic behav ior of their solutions can rarely be formulated. In [1, 2], a method for the asymptotic integration of tra jectories of the Liénard equation was developed. Based on this method, necessary and sufficient conditions for the global boundedness of quadratic twodimen sional systems can be derived and, additionally, exist ence criteria for limit cycles can be formulated. Consider the quadratic system (1) where a i , b 1 , c 1 , α i , and β i are real constants. Proposition 1 [2–4]. Without loss of generality, it can be assumed that c 1 = 0. This result is proved by making the linear substitu tions x = x 1 + νy 1 and y = y 1 , where ν solves a third degree equation [2–4]. Proposition 2. Let c 1 = 0 and β 1 0. Then, without loss of generality, it can be assumed that α 1 = 0. This proposition is proved by making the linear substitutions x = x 1 and y = y 1 . Proposition 3. Let c 1 = 0, α 1 = 0, a 1 0, b 1 0, β 1 0. Then, without loss of generality, it can be assumed that c 1 = α 1 = 0 and a 1 = b 1 = β 1 = 1. This proposition is proved by making the substitu tions Using these results, we assume in what follows that x · a 1 x 2 b 1 xy c 1 y 2 α 1 x β 1 y , + + + + = y · a 2 x 2 b 2 xy c 2 y 2 α 2 x β 2 y , + + + + = α 1 x 1 β 1 x β 1 b 1 x 1 , y a 1 β 1 b 1 2 y 1 , t b 1 a 1 β 1 t 1 . = = = (2) Proposition 4. The halfplane is positively invariant. This result follows from the fact that (t) = x(t) 2 = 1 if x(t) = –1. Note that, under assumptions (2), system (1) can be reduced to the Liénard system [2–4] (3) Here, the trajectories of systems (1) and (2) are related by (4) and It is easy to see that transformation (4) reduces sys tem (3) to the form Theorem 1. A necessary condition for any solution of system (1) with initial data from Γ to be bounded on (0, +) is that c 2 (0, 1). The proof sketch of this theorem is as follows. Note that c 1 α 1 0 , a 1 b 1 β 1 1 , c 2 0 , = = = = = c 2 1 , c 2 b 2 a 2 . Γ x 1 > y , R 1 { } = x · x · u , u · fx () u gx () . = = u y x 2 x 1 + + x 1 + q , q c 2 , = = fx () Ψ x () x 1 + q 2 , gx () Φ x () x 1 + 2 q x 1 + ( ) 3 , = = Ψ x () 2 c 2 b 2 1 ( ) x 2 2 b 2 β 2 + + ( ) x β 2 , = Φ x () xx 1 + ( ) 2 a 2 x α 2 + ( ) = + x 2 x 1 + ( ) b 2 x β 2 + ( ) c 2 x 4 . x · x 1 + q x 1 + ( ) x 2 xy y + + ( ) , = y · x 1 + q x 1 + ( ) a 2 x 2 b 2 xy c 2 y 2 α 2 x β 2 y + + + + ( ) . = Φ x () x 1 lim c 2 . = Necessary and Sufficient Conditions for the Boundedness of Solutions to TwoDimensional Quadratic Systems in a Positively Invariant HalfPlane Corresponding Member of the RAS G. A. Leonov Received August 18, 2009 DOI: 10.1134/S1064562410010102 Faculty of Mathematics and Mechanics, St. Petersburg State University, Universitetskii pr. 28, St. Petersburg, 198504 Russia email: [email protected] MATHEMATICS

Upload: g-a-leonov

Post on 03-Aug-2016

213 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Necessary and sufficient conditions for the boundedness of solutions to two-dimensional quadratic systems in a positively invariant half-plane

ISSN 1064�5624, Doklady Mathematics, 2010, Vol. 81, No. 1, pp. 31–33. © Pleiades Publishing, Ltd., 2010.Original Russian Text © G.A. Leonov, 2010, published in Doklady Akademii Nauk, 2010, Vol. 430, No. 2, pp. 157–159.

31

For nonlinear dynamical systems, necessary andsufficient conditions for the global asymptotic behav�ior of their solutions can rarely be formulated. In[1, 2], a method for the asymptotic integration of tra�jectories of the Liénard equation was developed. Basedon this method, necessary and sufficient conditionsfor the global boundedness of quadratic two�dimen�sional systems can be derived and, additionally, exist�ence criteria for limit cycles can be formulated.

Consider the quadratic system

(1)

where ai, b1, c1, αi, and βi are real constants.Proposition 1 [2–4]. Without loss of generality, it can

be assumed that c1 = 0.This result is proved by making the linear substitu�

tions x = x1 + νy1 and y = y1, where ν solves a third�degree equation [2–4].

Proposition 2. Let c1 = 0 and β1 ≠ 0. Then, withoutloss of generality, it can be assumed that α1 = 0.

This proposition is proved by making the linear

substitutions x = x1 and y = y1 – .

Proposition 3. Let c1 = 0, α1 = 0, a1 ≠ 0, b1 ≠ 0, β1 ≠ 0.Then, without loss of generality, it can be assumed thatc1 = α1 = 0 and a1 = b1 = β1 = 1.

This proposition is proved by making the substitu�tions

Using these results, we assume in what follows that

x· a1x2 b1xy c1y2 α1x β1y,+ + + +=

y· a2x2 b2xy c2y2 α2x β2y,+ + + +=

α1x1

β1

���������

xβ1

b1

����x1, ya1β1

b12

��������y1, tb1

a1β1

��������t1.= = =

(2)

Proposition 4. The half�plane

is positively invariant.

This result follows from the fact that (t) = x(t)2 = 1if x(t) = –1.

Note that, under assumptions (2), system (1) canbe reduced to the Liénard system [2–4]

(3)

Here, the trajectories of systems (1) and (2) are related by

(4)

and

It is easy to see that transformation (4) reduces sys�tem (3) to the form

Theorem 1. A necessary condition for any solution ofsystem (1) with initial data from Γ to be bounded on(0, +∞) is that c2 ∈ (0, 1).

The proof sketch of this theorem is as follows. Notethat

c1 α1 0, a1 b1 β1 1, c2 0,≠= = = = =

c2 1, c2 b2 a2.–≠–≠

Γ x 1–> y, R1∈{ }=

x· u, u· f x( )u– g x( ).–= =

u y x2

x 1+���������+⎝ ⎠

⎛ ⎞ x 1+ q, q c2,–= =

f x( ) Ψ x( ) x 1+ q 2–, g x( ) Φ x( ) x 1+ 2q

x 1+( )3���������������,= =

Ψ x( ) 2c2 b2– 1–( )x2 2 b2 β2+ +( )x– β2,–=

Φ x( ) x x 1+( )2 a2x α2+( )–=

+ x2 x 1+( ) b2x β2+( ) c2x4.–

x· x 1+ q

x 1+( )������������� x2 xy y+ +( ),=

y· x 1+ q

x 1+( )������������� a2x2 b2xy c2y2 α2x β2y+ + + +( ).=

Φ x( )x 1–→

lim c2.–=

Necessary and Sufficient Conditions for the Boundedness of Solutions to Two�Dimensional Quadratic Systems

in a Positively Invariant Half�PlaneCorresponding Member of the RAS G. A. Leonov

Received August 18, 2009

DOI: 10.1134/S1064562410010102

Faculty of Mathematics and Mechanics, St. Petersburg State University, Universitetskii pr. 28, St. Petersburg, 198504 Russiae�mail: [email protected]

MATHEMATICS

Page 2: Necessary and sufficient conditions for the boundedness of solutions to two-dimensional quadratic systems in a positively invariant half-plane

32

DOKLADY MATHEMATICS Vol. 81 No. 1 2010

LEONOV

Therefore, for c2 < 0, there exists a sufficiently smallε > 0 such that, for x(0) ∈ (–1, –1 + ε) and u(0) < 0, wehave

Here, T is either +∞ or a constant for which x(T) = –1.Combining this with (4) yields

Now let c2 > 1. If c2 > b2 – a2, then, for sufficientlylarge x, we have g(x) < 0. Therefore, system (3) hassolutions unbounded on (0, +∞) with u(0) > 0 and suf�ficiently large x(0). Moreover, x(t) → +∞ as t → +∞.

If c2 < b2 – a2, then Eq. (3) is written as the first�order equation

. (5)

Changing to the variable z = (x + 1)q + 1 gives

(6)

Obviously,

It follows that the point F = 0, z = 0 is a saddle for Eq. (6).Consequently, for sufficiently large initial data F(1),the solution F(z) is positive on [0, 1]. This solution isassociated with that of system (3) with the initial datax(0) = 0, u(0) = F(1), for which

as t → +∞. The proof of Theorem 1 is complete.Theorem 2. The solutions of system (1) in the half�

plane Γ are bounded on (0, +∞) if and only if c2 ∈ (0, 1),c2 < b2 – a2, and 2c2 > b2 + 1 or 2c2 ≤ b2 + 1 and 4a2(c2 –1) > (b2 – 1)2.

The proof sketch of Theorem 2 is as follows. Onceagain, consider the substitution z = (x + 1)q + 1 andEq. (6) and the substitution z = (x + 1)q – 1 and theequation

u t( ) 0, x t( ) 1– –1 ε+,( ), t∀ 0 T,[ ].∈ ∈<

y t( )t T→

lim ∞.–=

F dFdx����� f x( )F g x( )+ + 0=

FdFΨ z

1q 1+���������

1–⎝ ⎠⎛ ⎞

q 1+( )z2

q 1+���������

������������������������FdzΦ z

1q 1+���������

1–⎝ ⎠⎛ ⎞

q 1+( )z4

q 1+���������

������������������������zdz+ + 0.=

Φ z1

q 1+���������

1–⎝ ⎠⎛ ⎞

q 1+( )z4

q 1+���������

������������������������z 0→

limb2 a2– c2–

q 1+( )��������������������� 0.<=

u t( ) 0, t∀ 0, x t( ) +∞→≥>

(7)

Equations (6) and (7) are well suited for the asymp�totic integration of their solutions for large initial dataand for q ∈ (–1, 0) since

It is easy to see that, for b2 < c2 + a2 and sufficientlylarge z0, the solution F(z) of Eq. (6) with initial dataF(z0) > 0 is positive on [z0, +∞]. Therefore, system (1)in Γ has solutions that are unbounded on (0, +∞).

Let b2 > c2 + a2. Then, taking into account that, forlarge initial conditions F(1) = R, the solutions ofEqs. (6) and (7) are close to those of the equations

where z ≥ 1, we find that, for 2c2 > b2 + 1, the trajecto�ries of system (3) lie as shown in the figure [1]. Thesetrajectories are located in a similar manner if 2c2 ≤ b2 + 1and 4a2(c2 – 1) > (b2 – 1)2. Therefore, the solutions ofsystem (1) are bounded in Γ.

If 2c2 ≤ b2 + 1 and 4a2(c2 – 1) > (b2 – 1)2, then it iseasy to see that there are sufficiently large z0 and ρ0 forwhich the solution of Eq. (6) with the initial dataF(z0) = ρ0 is positive on [z0, +∞]. This means that sys�tem (1) has unbounded solutions in Γ. The proof ofTheorem 2 is complete.

Based on the qualitative behavior of solutions tosystem (3) as shown in the figure, we can obtain anexistence condition for limit cycles of system (1).

Theorem 3. Suppose that c2 ∈ (0, 1), b2 > a2 + c2,2c2 > b2 + 1; and g(x) has the only zero x = 0 on theinterval (–1, +∞), which is associated with the unstableequilibrium x = x1, u = 0 of system (3).

Then system (1) has a limit cycle in the half�plane Γ.

FdFΨ z

1q 1–���������

1–⎝ ⎠⎛ ⎞

q 1–������������������������Fdz

Φ z1

q 1–���������

1–⎝ ⎠⎛ ⎞

q 1–������������������������zdz+ + 0.=

Ψ z1

q 1+���������

1–⎝ ⎠⎛ ⎞

z2

q 1+���������

������������������������z +∞→

lim 2c2 b2– 1,–=

Φ z1

q 1+���������

1–⎝ ⎠⎛ ⎞

z4

q 1+���������

������������������������z +∞→

lim b2 a2– c2,–=

Ψ z1

q 1–���������

1–⎝ ⎠⎛ ⎞

z +∞→

lim 1 2c2,+=

Φ z1

q 1–���������

1–⎝ ⎠⎛ ⎞

z +∞→

lim c2.–=

FdF2c2 b2 1–+( )

q 1+( )��������������������������Fdz

b2 a2– c2–( )q 1+( )

�������������������������zdz+ + 0,=

FdF1 2c2+q 1–( )

�������������Fdzc2

q 1–( )�������������zdz–+ 0,=

–1

u

x

Figure.

Page 3: Necessary and sufficient conditions for the boundedness of solutions to two-dimensional quadratic systems in a positively invariant half-plane

DOKLADY MATHEMATICS Vol. 81 No. 1 2010

NECESSARY AND SUFFICIENT CONDITIONS FOR THE BOUNDEDNESS OF SOLUTIONS 33

Theorem 4. Suppose that c2 ∈ (0, 1); b2 > a2 + c2; 2c2 ≤b2 + 1; 4a2(c2 – 1) > (b2 – 1)2; and g(x) has only twozeros x = 0 and x = x1 ∈ (–∞, –1), which are associatedwith the unstable equilibria x = x1, u = 0 and x = 0, u = 0of system (3).

Then system (1) has two limit cycles, of which one liesto the left of the straight line {x = –1, y ∈ R1} and theother lies to the right of this line.

The following results are useful in the verificationof the conditions of Theorems 3 and 4 [2].

Proposition 5. Let β2 = 0. Then g(x) has only twozeros x = 0 and x = x1 ∈ (–∞, –1) if and only if α2 < λ,where λ is the minimal root of the equation

Proposition 6. Let β2 = 0. Then x = x1, u = 0 is anunstable equilibrium if and only if

where

Note that β2 = 0 and α2 < 0 are necessary condi�tions for x = y = 0 to be a weak focus of system (1).

In this case, applying local bifurcation theory [1, 2,5, 6] together with Theorem 4, we can single outclasses of systems of form (1) that have four limitcycles.

REFERENCES

1. G. A. Leonov, Dokl. Phys. 54, 238–146 (2009) [Dokl.Akad. Nauk 426, 47–50 (2009)].

2. G. A. Leonov, N. V. Kuznetsov, and E. V. Kudryashova,Vestn. Sankt�Peterburg. Gos. Univ., Ser. 1, No. 3, 25–61 (2008).

3. G. A. Leonov, Differ. Equations Dyn. Syst. 5 (3/4),289–297 (1998).

4. G. A. Leonov, Vestn. S.�Peterburg. Gos. Univ., Ser. 1,No. 4, 48–78 (2006).

5. J. Li, Int. J. Bifurcation Chaos 13 (1), 47–106 (2003).6. P. Yu and G. Chen, Nonlinear Dyn. 51, 409–427

(2008).

4c2λ3– 27c22 12a2 18b2–( )c2 b2

2–+( )λ2+

+ 2 a2b22

2b23– 9a2b2c2 6a2

2c2–+( )λ a22b2

2– 4a23c2+ = 0.

α21

p 1+( )2��������������� –a2p p 1+( )2 b2p2 p 1+( ) c2p3–+( ),<

p2 b2+

2c2 b2– 1–���������������������� .=