• 検索結果がありません。

Then we apply the averaging theory for first-order discontinuous differential systems to show that for anynandmthere are non-smooth Lienard polynomial equations having at least max{n, m}limit cycles

N/A
N/A
Protected

Academic year: 2022

シェア "Then we apply the averaging theory for first-order discontinuous differential systems to show that for anynandmthere are non-smooth Lienard polynomial equations having at least max{n, m}limit cycles"

Copied!
8
0
0

読み込み中.... (全文を見る)

全文

(1)

Electronic Journal of Differential Equations, Vol. 2013 (2013), No. 195, pp. 1–8.

ISSN: 1072-6691. URL: http://ejde.math.txstate.edu or http://ejde.math.unt.edu ftp ejde.math.txstate.edu

LIMIT CYCLES FOR DISCONTINUOUS GENERALIZED LIENARD POLYNOMIAL DIFFERENTIAL EQUATIONS

JAUME LLIBRE, ANA CRISTINA MEREU

Abstract. We divideR2 into sectorsS1, . . . , Sl, withl >1 even, and define a discontinuous differential system such that in each sector, we have a smooth generalized Lienard polynomial differential equation ¨x+fi(x) ˙x+gi(x) = 0, i= 1,2 alternatively, wherefiandgiare polynomials of degreen1 andm respectively. Then we apply the averaging theory for first-order discontinuous differential systems to show that for anynandmthere are non-smooth Lienard polynomial equations having at least max{n, m}limit cycles. Note that this number is independent of the number of sectors.

Roughly speaking this result shows that the non-smooth classical (m= 1) Lienard polynomial differential systems can have at least the double number of limit cycles than the smooth ones, and that the non-smooth generalized Lienard polynomial differential systems can have at least one more limit cycle than the smooth ones.

1. Introduction

A large number of problems from mechanics and electrical engineering, theory of automatic control, economy, impact systems among others cannot be described with smooth dynamical systems. This fact has motivated many researchers to the study of qualitative aspects of the phase space of non-smooth dynamical systems.

One of the main problems in the qualitative theory of real planar continuous and discontinuous differential systems is the determination of their limit cycles. The non-existence, existence, uniqueness and other properties of limit cycles have been studied extensively by mathematicians and physicists, and more recently also by chemists, biologists, economists, etc (see for instance the books [2, 5, 24]). This problem restricted to continuous planar polynomial differential equations is the well known 16th Hilbert’s problem [10]. Since this Hilbert’s problem turned out a strongly difficult one Smale [23] particularized it to Lienard polynomial differential equations in his list of problems for the present century.

The classical Lienard polynomial differential equations

¨

x+f(x) ˙x+g(x) = 0, (1.1)

where f(x) and g(x) =xgoes back to [11]. The dot denotes differentiation with respect to the time t. This second-order differential equation (1.1) can be written

2000Mathematics Subject Classification. 34C29, 34C25, 47H11.

Key words and phrases. Limit cycles; non-smooth Li´enard systems; averaging theory.

c

2013 Texas State University - San Marcos.

Submitted May 7, 2013. Published September 3, 2013.

1

(2)

as the following first-order differential system inR2

˙

x=y−F(x),

˙

y=−g(x), (1.2)

whereF(x) =Rx 0 f(s)ds.

Many results on the number of limit cycles has been obtained for the continuous generalized polynomial differential equations (1.1) beingf(x) andg(x) polynomials in the variable x of degrees n−1 and m respectively. The continuous classical Lienard polynomial differential equations (1.1) were studied in 1977 by Lins, de Melo and Pugh [12] who stated the conjecture:

Iff(x) has degree n−1>0 andg(x) =x, then (1.1) has at most [(n−1)/2] limit cycles.

Here [z] denotes the integer part function ofz∈R. They also proved the conjecture forn= 2,3. Forn= 4 this conjecture has been proved in 2012 (see [13]). Forn≥7 Dumortier, Panazzolo and Roussarie proved that this conjecture is not true in [7], they show that these differential equations can have [(n−1)/2] + 1 limit cycles.

Recently De Maesschalck and Dumortier proved in [22] that the classical Lienard equation of degreen≥6 can have [(n−1)/2] + 2 limit cycles. The conjecture for n= 5 is still open.

Results on the number of limit cycles for continuous generalized Lienard polyno- mial differential equations can be found in [14] where the authors show that there are differential equations (1.1) having at least [(n+m−2)/2] limit cycles. See also [1, 3, 6, 8, 18, 19, 20, 21, 25].

The objective of this work is to star the study of the number of limit cycles for a kind of discontinuous generalized Lienard polynomial differential systems.

Here we shall play with many straight lines of discontinuities through the origin of coordinates and with two different continuous generalized Lienard polynomial differential systems located alternatively in the sectors defined by these straight lines.

A similar work but with only one classical Lienard polynomial differential system and only one straight line of discontinuity was studied in [16] obtaining [n/2] limit cycles, instead of the [(n−1)/2] of the continuous classical Lienard polynomial differential equation obtained in [12].

Now we shall define the discontinuous generalized Lienard polynomial differential system that we will study. We consider the functionh:R2→Rdefined by

h(x, y) =

l 2−1

Y

k=0

y−tan α+2kπ l

x , wherel >1 even. The set

h−1(0) =∪k=02l−1{(x, y) :y = tan α+2kπ l

x},

divides R2 into l sectors, S1, S2, . . . , Sl, i.e. h−1(0) is the product ofl/2 straight lines passing through the origin of coordinates dividing the plane in sectors of angle 2π/l.

In this work we study the maximum number of limit cycles given by the averaging theory of first order, which can bifurcate from the periodic orbits of the linear

(3)

center ˙x=y, ˙y=−x, perturbed inside the following class of discontinuous Lienard polynomial differential systems

X˙ =Z(x, y) =

(Y1(x, y) ifh(x, y)>0,

Y2(x, y) ifh(x, y)<0, (1.3) where

Y1(x, y) =

y−εF1(x)

−x−εg1(x)

, Y2(x, y) =

y−εF2(x)

−x−εg2(x)

, (1.4)

where εis a small parameter, andFi(x) and gi(x), for i= 1,2 are polynomials in the variablex, and degreesnandmrespectively. System (1.3) can be written using the sign function as

X˙ =Z(x, y) =G1(x, y) + sgn(h(x, y))G2(x, y), (1.5) whereG1(x, y) =12(Y1(x, y) +Y2(x, y)) andG2(x, y) = 12(Y1(x, y)−Y2(x, y)).

Our main result reads as follows.

Theorem 1.1. Assume that for i = 1,2 the polynomials Fi(x) and gi(x) have degreen≥1andm≥1respectively, and thatl >1is even. Then for|ε|sufficiently small there are discontinuous Lienard polynomial differential systems (1.3)having at least max{n, m} limit cycles bifurcating from the periodic orbits of the linear centerx˙ =y,y˙=−x.

Taking into account Theorem 1.1 and roughly speaking, we can say that the non- smooth classical Lienard polynomial differential systems can have at least max{n,1}

limit cycles; i.e. the double number of limit cycles than the smooth ones which at least have [(n−1)/2] + 2 for n ≥ 6. Comparing the mentioned result from [14], that smooth generalized Lienard polynomial differential systems have at least [(n+m−1)/2] limit cycles with Theorem 1.1, we can say that the non-smooth generalized Lienard polynomial differential systems can have at least one more limit cycle than the smooth ones. Of course all these comparisons are done with the present known results.

2. Averaging theory for first-order discontinuous differential systems

The first-order averaging theory developed for discontinuous differential systems in [15] is presented in this section. It is summarized as follows.

Theorem 2.1. We consider the discontinuous differential system

x0(t) =εF(t, x) +ε2R(t, x, ε), (2.1) with

F(t, x) =F1(t, x) + sgn(h(t, x))F2(t, x), R(t, x, ε) =R1(t, x, ε) + sgn(h(t, x))R2(t, x, ε),

whereF1, F2:R×D→Rn,R1, R2:R×D×(−ε0, ε0)→Rn andh:R×D→R are continuous functions, T–periodic in the variablet andD is an open subset of Rn. We also suppose that his a C1 function having0 as a regular value. Denote by M=h−1(0), byΣ ={0} ×D *M, byΣ0 = Σ\M 6=∅, and its elements by z≡(0, z)∈ M./

(4)

Define the averaged functionf :D→Rn as f(x) =

Z T

0

F(t, x)dt.

We assume the following three conditions.

(i) F1, F2, R1, R2 andhare locally L-Lipschitz with respect to x;

(ii) for a ∈ Σ0 with f(a) = 0, there exist a neighborhood V of a such that f(z)6= 0for all z∈V\{a}anddB(f, V, a)6= 0, (i.e. the Brouwer degree of f atais not zero).

(iii) If ∂h/∂t(t0, z0) = 0for some(t0, z0)∈ M, then h∇xh, F1i2− h∇xh, F2i2

(t0, z0)>0.

Then, for |ε| > 0 sufficiently small, there exists a T-periodic solution x(·, ε) of system (2.1)such that x(t, ε)→aasε→0.

Remark 2.2. We note that if the functionf(z) isC1and the Jacobian of f ata is not zero, then dB(f, V, a)6= 0. For more details on the Brouwer degree see [4]

and [17].

3. Proof of Theorem 1.1

The discontinuous Lienard differential systems (1.3) in polar coordinates (r, θ) become

˙

r=−ε(cosθ Fi(rcosθ) + sinθ gi(rcosθ)), θ˙=−1 + ε

r(sinθ Fi(rcosθ)−cosθ gi(rcosθ)),

with i = 1 if sgn(h(rcosθ, rsinθ)) > 0 and i = 2 if sgn (h(rcosθ, rsinθ)) <

0. Taking the angle θ as new independent variable the discontinuous differential systems become

˙

r=ε(cosθ Fi(rcosθ) + sinθ gi(rcosθ)) +O(ε2). (3.1) This discontinuous differential system is studied under the assumptions of Theorem 2.1, taking

t=θ, T = 2π, x=r, M=h−1(0) =∪k=02l−1{(θ, r) :θ=α+2kπ

l , r >0}.

So according to Theorem 2.1 we must study the zeros of the averaged function f(r) =

l

X

k=1

hZ α+(2k−1)πl

α+2(k−1)πl

(cosθ F1(rcosθ) + sinθ g1(rcosθ))dθ

+

Z α+2kπl

α+(2k−1)πl

(cosθ F2(rcosθ) + sinθ g2(rcosθ))dθi .

(3.2)

Denoting F1(x) =

n

X

i=0

aixi, F2(x) =

n

X

i=0

bixi, g1(x) =

m

X

i=0

cixi, g2(x) =

m

X

i=0

dixi we have

f(r) =

l

X

k=1

hZ α+(2k−1)πl

α+2(k−1)πl

Xn

i=0

airicosi+1θ+

m

X

i=0

ciricosiθsinθ dθ

(5)

+

Z α+2kπl

α+(2k−1)πl

Xn

i=0

biricosi+1θ+

m

X

i=0

diricosiθsinθ dθi

.

To calculate the exact expression off(r) we use [9, formulae 2.513 3 and 2.513 4]:

Z

cos2mθ dθ= 1 22m

2m m

θ+ 1 22m−1

m−1

X

j=0

2m j

sin(2m−2j)θ 2m−2j , Z

cos2m+1θ dθ= 1 22m

m

X

j=0

2m+ 1 j

sin(2m−2j+ 1)θ 2m−2j+ 1 . Thus we have

f(r) =f1(r) +f2(r) +f3(r) +f4(r), where

f1(r) =

l

X

k=1

Z α+(2k−1)πl

α+2(k−1)πl

Xn

i=0

airicosi+1θ dθ

=

l

X

k=1

h Xn

i=1, iodd

airih 1 2i+1

i+ 1 (i+ 1)/2

π l + 1

2i

i−1 2

X

j=0

i+ 1 j

ϕi,j,k]

+

n

X

i=0, ieven

airi 2i

i 2

X

j=0

i+ 1 j

ϕi,j,k

i

=

l

X

k=1

[

n

X

i=1, iodd

airi 2i+1

i+ 1 (i+ 1)/2

π l +

n

X

i=0

airi 2i

[2i]

X

j=0

i+ 1 j

ϕi,j,k], with

ϕi,j,k= sin

(i−2j+ 1)

α+(2k−1)πl

−sin

(i−2j+ 1)

α+2(k−1)πl

i−2j+ 1 6= 0;

f2(r) =

l

X

k=1

Z α+(2k−1)πl

α+2(k−1)πl

Xm

i=0

ciricosiθsinθ dθ=−

l

X

k=1 m

X

i=0

ciri i+ 1φi,k, with

φi,k= cosi+1

α+(2k−1)π l

−cosi+1

α+2(k−1)π l

6= 0;

f3(r) =

l

X

k=1

Z α+(2k−1)πl

α+2(k−1)πl

Xn

i=0

biricosi+1θ dθ

=

l

X

k=1

h Xn

i=1, iodd

biri 2i+1

i+ 1 (i+ 1)/2

π l +

n

X

i=0

biri 2i

[2i]

X

j=0

i+ 1 j

ψi,j,k

i ,

with ψi,j,k=

sin (i−2j+ 1) α+2kπl

−sin

(i−2j+ 1) α+(2k−1)πl

i−2j+ 1 6= 0;

(6)

f4(r) =

l

X

k=1

Z α+2kπl

α+(2k−1)πl

Xm

i=0

diricosiθsinθ dθ=−

l

X

k=1 m

X

i=0

diri i+ 1ζi,k, with

ζi,k= cosi+1 α+2kπ l

−cosi+1 α+(2k−1)π l

6= 0.

Thus f(r) =

l

X

k=1

h Xn

i=1, iodd

ri 2i+1

i+ 1 (i+ 1)/2

π

l(ai+bi) +

n

X

i=0

ri 2i

[2i]

X

j=0

i+ 1 j

(aiϕi,j,k+biψi,j,k)−

m

X

i=0

ri

i+ 1(ciφi,k+diζi,k)i . The functionf(r) is a polynomial in the variablerof degree max{n, m}therefore f(r) has at most max{n, m}positive roots. Ifris a simple zero off(r); i.e. f(r) = 0 and drdf

r=r6= 0, then the Brouwer degreedB(f, V, r)6= 0 being V a convenient open neighborhood ofr(see Remark 2.2). We can choose the coefficientsai,bi,ci

edi in such a way thatf(r) has exactly max{n, m} simple positive roots. Hence Theorem 1.1 is proved.

4. Examples

In this section we illustrate Theorem 1.1 by studying the existence of 2π-periodic solutions for two non-smooth Lienard polynomial differential systems.

Example 4.1. We consider l = 2 and α= 0. Thus the function h: R2 →R is defined byh(x, y) =y andh−1(0) ={(x, y)∈R2:y= 0}. System (1.3) becomes

X˙ =Z(x, y) =

(Y1(x, y) ify >0,

Y2(x, y) ify <0, (4.1) where

F1= 1 +x+x2+ 1 9π−1

x3, F2= 1 + 11 12π−1

x+x2+x3, g1= 7

8+x+5

8x2, g2= 1 +x+x2. Thus we have

Y1(x, y) =

y−ε 1 +x+x2+ 1 −1 x3

−x−ε 78 +x+58x2

, Y2(x, y) =

y−ε 1 + 12π11 −1

x+x2+x3

−x−ε(1 +x+x2)

. The averaging function (3.2) is

f(r) = Z π

0

(cosθ F1(rcosθ) + sinθ g1(rcosθ))dθ +

Z

π

(cosθ F2(rcosθ) + sinθ g2(rcosθ))dθ

=−6 + 11r−6r2+r3.

(7)

The zeros of f(r) are r = 1, r = 2 and r = 3, and they are simple. Hence, by Theorem 1.1 it follows that forε6= 0 sufficiently small the discontinuous differential system (4.1) has three periodic solutions.

Example 4.2. We considerl = 4 andα=π/4. Thus the function h:R2→Ris defined byh(x, y) = (y−x)(y+x) andh−1(0) ={(x, y) :y=x}∪{(x, y) :y=−x}.

System (1.3) becomes

X˙ =Z(x, y) =

(Y1(x, y) if (y−x)(y+x)>0,

Y2(x, y) if (y−x)(y+x)<0, (4.2) whereF1=x2,F2= 12√

2π+72

2

5 x2,g1= 1+x+x2+x3andg2=−88πx−32π3 x3. Thus we have

Y1(x, y) =

y−εx2

−x−ε(1 +x+x2+x3)

, Y2(x, y) =

y−ε 12√ 2π+72

2 5 x2

−x−ε −88πx−32π3 x3

. The averaging function (3.2) is

f(r) = Z 3π/4

π/4

(cosθ F1(rcosθ) + sinθ g1(rcosθ))dθ +

Z 5π/4

3π/4

(cosθ F2(rcosθ) + sinθ g2(rcosθ))dθ +

Z 7π/4

5π/4

(cosθ F1(rcosθ) + sinθ g1(rcosθ))dθ +

Z 9π/4

7π/4

(cosθ F2(rcosθ) + sinθ g2(rcosθ))dθ

=−6 + 11r−6r2+r3.

The zeros of f(r) are r = 1, r = 2 and r = 3, and they are simple. Hence, by Theorem 1.1 it follows that forε6= 0 sufficiently small the discontinuous differential system (4.2) has three periodic solutions.

Acknowledgments. The first author is partially supported by a MICINN/FEDER grant MTM 2008–03437, an AGAUR grant number 2009SGR-0410, an ICREA Academia, and FP7-PEOPLE-2012-IRSES-316338 and 318999. The second author is partially supported by a FAPESP-BRAZIL grant 2012/20884-8. Both authors are also supported by the joint project CAPES–MECD grant PHB-2009-0025-PC.

References

[1] S. Badi, A. Makhlouf;Limit cycles of the generalized Li´enard differential equation via aver- aging theory, Electron. J. Differential Equations 2012, No. 68, 11 pp.

[2] N. N. Bogoliubov;On some statistical methods in mathematical physics, Izv. vo Akad. Nauk Ukr. SSR, Kiev, 1945.

[3] I. Boussaada, A. R. Chouikha;Existence of periodic solution for perturbed generalized Li´enard equations, Electron. J. Differential Equations2006, No. 140, 10 pp.

[4] F. Browder; Fixed point theory and nonlinear problems, Bull. Amer. Math. Soc.9(1983), 1–39.

[5] S. N. Chow, C. Li, D. Wang;Normal Forms and Bifurcation of Planar Vector Fields, Cam- bridge Univ. Press., 1994.

[6] C. J. Christopher, S. Lynch; Small-amplitude limit cycle bifurcations for Li´enard systems with quadratic or cubic dumping or restoring forces, Nonlinearity12(1999), 1099–1112.

(8)

[7] F. Dumortier, D. Panazzolo, R. Roussarie;More limit cycles than expected in Lienard sys- tems, Proc. Amer. Math. Soc.135(2007), 1895–1904.

[8] A. Gasull, J. Torregrosa;Small-amplitude limit cycles in Li´enard systems via multiplicity, J.

Differential Equations159(1998), 1015–1039.

[9] I. S. Gradshteyn, I. M. Ryshik;Table of Integrals, Series, and Products, Edited by A. Jeffrey.

Academic Press, New York, 5th edition, 1994.

[10] D. Hilbert;Mathematische Probleme (lecture), Second Internat. Congress Math. Paris, 1900, Nach. Ges. Wiss. Gottingen Math.–Phys. Kl., 1900, 253–297.

[11] A. Li´enard; Etude des oscillations entrenues, Revue G´´ enerale de l’ ´Electricit´e 23(1928), 946–954.

[12] A. Lins, W. de Melo, C.C. Pugh; On Li´enard’s Equation, Lecture Notes in Math 597, Springer, Berlin, (1977), pp. 335–357.

[13] C. Li, J. Llibre;Uniqueness of limit cycle for Li´enard equations of degree four, J. Differential Equations252(2012), 3142–3162.

[14] J. Llibre, A. C. Mereu, M. A. Teixeira;Limit cycles of the generalized polynomial Li´enard differential equations, Math. Proc. Cambridge Philos. Soc.148(2010) 363–383.

[15] J. Llibre, D. D. Novaes, M. A. Teixeira;Averaging methods for studying the periodic orbits of discontinuous differential systems, http://arxiv.org/pdf/1205.4211. pdf

[16] J. Llibre, M. A. Teixeira;Periodic solutions of discontinuous second order differential sys- tems, preprint, 2012

[17] N. G. Lloyd;Degree Theory, Cambridge University Press, 1978.

[18] S. Lynch; Limit cycles if generalized Li´enard equations, Applied Math. Letters 8 (1995), 15–17.

[19] S. Lynch;Generalized quadratic Li´enard equations, Applied Math. Letters11(1998), 7–10.

[20] S. Lynch;Generalized cubic Li´enard equations, Applied Math. Letters12(1999), 1–6.

[21] S. Lynch, C. J. Christopher;Limit cycles in highly non-linear differential equations, J. Sound Vib.224(1999), 505–517

[22] P. De Maesschalck, F. Dumortier; Classical Li´enard equation of degree n 6 can have [n−12 ] + 2limit cycles, J. Differential Equations250(2011), 2162–2176.

[23] S. Smale; Mathematical Problems for the Next Century, Mathematical Intelligencer 20 (1998), 7–15.

[24] Ye Yanqian;Theory of Limit Cycles, Translations of Math. Monographs66(Providence, RI Amer. Math. Soc.), 1986.

[25] P. Yu, M. Han;Limit cycles in generalized Li´enard systems, Chaos, Solitons and Fractals30 (2006), 1048–1068.

Jaume Llibre

Departament de Matem`atiques, Universitat Aut`onoma de Barcelona, 08193 Bellaterra, Barcelona, Catalonia, Spain

E-mail address:[email protected]

Ana Cristina Mereu

Department of Physics, Chemistry and Mathematics, UFSCar, 18052-780, Sorocaba, SP, Brazil

E-mail address:[email protected]

参照

関連したドキュメント

By using the averaging theory of the first and second orders, we show that under any small cubic homogeneous perturbation, at most two limit cycles bifurcate from the period annulus

By using the first order averaging method and some mathematical technique on estimating the number of the zeros, we show that under a class of piecewise smooth quartic

Also we show the existence of mero- morphic solutions with finite order for differential-difference equations related to the Fermat type functional equation.. This article

g(t,Y) is small in some sense, it is reasonable to expect that the behavior of the solutions of the perturbed equation will be similar to that for the solutions of the

One of the motivations of this work is that parameter dependent systems with ex- ponential nonlinearities have been recently shown to be very involved in relativistic

The purpose of this work is to advance and simplify our understanding of some of the basic theory of linear dynamic equations and dynamic inequalities on time scales, with

Yoshida; A Picone-type identity and Sturmian comparison and oscillation theorems for a class of half-linear partial differential equations of second order, Nonlinear Anal.. Maˇ

We then apply the generali- zed Delzant construction simultaneously to each rhombohedron and we show that there is one symplectic quasifold, M b , associated to each of the