Вы находитесь на странице: 1из 9

ISSN 0012-2661, Differential Equations, 2011, Vol. 47, No. 1, pp. 19. c Pleiades Publishing, Ltd., 2011.

Original Russian Text c I.M. Burkin, D.V. Soboleva, 2011, published in Differentsialnye Uravneniya, 2011, Vol. 47, No. 1, pp. 310.

ORDINARY DIFFERENTIAL EQUATIONS

On a Smale Problem
I. M. Burkin and D. V. Soboleva
Tula State University, Tula, Russia
Received December 12, 2008

Abstract We derive conditions under which a linear coupling between two globally stable nonlinear nth-order systems results in a system of order 2n whose almost every solution asymptotically approaches an orbitally stable cycle. These results permit one to solve a problem posed by Smale and pertaining to the theory of chemical kinetics of biological cells. DOI: 10.1134/S0012266111010010

1. STATEMENT OF THE PROBLEM A mathematical model in cell biology was suggested in [1, pp. 274283]. This model describes two identical cells interacting by diusion through a membrane. The system of dierential equations describing the chemical kinetics of substances (ferments) contained in each cell is of fourth order. This means that the cells contain four ferments reacting with one another. The dynamics of this system has the property that each solution tends to a unique stationary point as time tends to innity. In other words, each cell is dead on its own (the concentration of its ferments achieves an equilibrium state); however, if the cells begin to interact by diusion through a membrane, then the cell system begins to pulse (returns to life). From the mathematical viewpoint, this means that almost every solution of the new dynamical system tends to some periodic solution. Mathematically, this situation can be described as follows. The state space of each cell is the space P = {x Rn : x = (x1 , x2 , . . . , xn ), xi 0}, where n is the number of ferments. Moreover, a smooth mapping R : P Rn and an n n matrix D are given such that the following conditions are satised. 1. The nth-order system of dierential equations dx = R(x) dt describing an isolated cell has a unique equilibrium, which is globally asymptotically stable. 2. The 2nth-order system of dierential equations (the Turing system [2]) dx(1) = R(x(1) ) + D(x(2) x(1) ), dt dx(2) = R(x(2) ) + D(x(1) x(2) ) dt (2) (1)

describing the interaction of two adjacent cells is a global oscillator; i.e., it has a nontrivial stable periodic solution to which almost all remaining solutions of the system tend. An example of system (1), (2) with the desired properties was constructed in [1, p. 278] for n = 4, and it was shown there that this example can be used to derive more complicated examples for an arbitrary number of cells (more than one) or substances (more than three). Smale [1] also posed the problem on the axiomatization of properties of the pair (R, D) which ensure that the solutions of systems (1) and (2) have the desired properties and pointed out that the aim is to reduce the number of substances in the problem to 2 or even 3 (n 3) and construct a model with three cells and two or more chemicals. The problem of axiomatizing the properties of the pair (R, D ) was partly solved in [3, 4]. However, the requirement of global oscillation of the system was replaced in [3, 4] by the less restrictive condition of self-oscillation of this system.
1

BURKIN, SOBOLEVA

In the present paper, we solve the problem on the axiomatization of the properties of the pair (R, D ) ensuring the desired properties of solutions of systems (1) and (2). In particular, our results permit one to solve the problem of reducing the number of chemicals to 3. Before proceeding to the exposition of the results, we note that, by [1, p. 279], it suces to construct a mapping R of Rn into Rn rather than a mapping R of P into Rn ; however, one should require that the cycle of system (2) has a suciently small amplitude. In addition, in accordance with the meaning of the problem, the matrix D should be similar to a positive diagonal matrix. 2. STATEMENT OF MAIN ASSERTIONS Following [1, p. 278], we take system (1) in the form dx = Rx + q, dt = ( ), = r x, (3)

where R is an n n matrix, q and r are n-vectors, and ( ) is a continuously dierentiable function. The symbol stands for the transposition and, in the complex case below, for Hermitian conjugation. Then system (2) has the form dx(1) = Rx(1) + q1 + D(x(2) x(1) ), dt where i = (i ), dx(2) = Rx(2) + q2 + D(x(1) x(2) ), dt i = 1, 2. (4)

i = r x(i) ,

Throughout the following, we assume that the pairs (R, q ) and (R 2D, q ) are completely controllable; i.e., the ranks of the n n matrices q, Rq, . . . , Rn1 q and q, (R 2D)q, . . . , (R 2D)n1 q are equal to n. It is well known [5, p. 48] that the complete controllability of the pairs (R, q ) and (R 2D, q ) takes place, for example, if the rational fractional functions W1 (p) = r (R pIn )1 q = m1 (p)[n1 (p)]1 and W2 (p) = r (R 2D pIn )1 q = m2 (p)[n2 (p)]1 of the complex argument p are nonsingular, i.e., the degrees of polynomials occurring in the denominators of these fractions are equal to n and fractions are irreducible. Here In is the n n identity matrix. System (4) can be represented in the form dx = Ax + b ( ), dt = c x. (5)

Here = col(1 , 2 ), i = (i ), i = r x(i) , i = 1, 2, and the 2n 2n matrix A and the 2n 2 matrices b and c have the form A= RD D D RD , b= q 0 0 q , c= r 0 0 r .

Note also that, by using the special form of the matrices A and b and well-known criteria of complete controllability [5, Th. 1.2.1], one can readily show that the complete controllability of the pairs (R, q ) and (R 2D, q ) implies the complete controllability of the pair (A, b). Set W (p) = c (A pI2n )1 b. One can readily show that W (p) = u(p) v (p) v (p) u(p) , u(p) = W1 (p) + W2 (p) , 2 v (p) = W1 (p) W2 (p) . 2 (6)

Theorem 1. Let the following conditions be satised. 1. (0) = 0, and the inequality 0 < ( ) < 1 holds for some 1 > 0 and for all (, ).
DIFFERENTIAL EQUATIONS Vol. 47 No. 1 2011

(7)

ON A SMALE PROBLEM

2. There exists a number > 0 such that all eigenvalues of the matrix R lie in the half-strip Re p < and
1 1 + Re W1 (i ) > 0 1 1 + Re W2 (i ) > 0

if if

0, 0.

(8) (9)

3. All eigenvalues of the matrix R 2D have negative real parts (R 2D is a Hurwitz matrix ), and there exist positive numbers 2 < 1 and 0 such that
1 2 + Re W1 (i ) > 0, 1 2 + Re W2 (i ) > 0 0 < ( ) 2 2

if 0, if | | 0 .

(10) (11)

4. The matrix R 2D + (0)qr has exactly two eigenvalues with positive real parts and no eigenvalue in the strip Re p 0. 1 1 and W2 (0) hold. 5. The relations W1 (0) 1 1 Then the following assertions are valid. 10 . The equilibrium x = 0 of system (3) is globally stable. 20 . System (4) has the unique equilibrium x(1) = 0, x(2) = 0 [system (5) has the unique equilibrium x = 0]. 30 . There exists a closed bounded set R2n not containing the point x = 0 and such that , for an arbitrary x0 , the trajectory of the solution x(t, x0 ) converges to a closed trajectory of that system as t +. The set contains at least one closed trajectory that is orbitally stable. If ( ) is an analytic function , then the set contains at most countably many closed trajectories , at least one of which is asymptotically orbitally stable. 40 . There exists a set R2n of zero measure such that x(t, x0 ) 0 as t + for x0 / there exists a t(x0 ) such that x(t, x0 ) for t > t(x0 ). and for x0 Remark 1. If relations (9)(11) hold, then the function ( ) can always be chosen so as to ensure that the inclusion {x : |x| h} is true for an arbitrary number h > 0. In other words, the amplitude of cycles of system (5) can be made arbitrarily small. Remark 2. Theorem 1 does not provide a complete solution of the Smale problem on the construction of a global oscillator, since it does not provide the existence of a unique stable cycle in the domain to which all remaining trajectories in that domain converge. Moreover, as the following theorem shows, if the assumptions of Theorem 1 are satised, then the nonlinearity can always be chosen so as to ensure that system (4) has any prescribed number of asymptotically orbitally stable cycles. Theorem 2. Suppose that there exist numbers 1 , 2 , and > 0 satisfying the condition 0 < 2 < 1 and such that relations (8)(10) hold , and suppose that assumption 5 in Theorem 1 is qr is a Hurwitz matrix and the matrix satised. For some (2 , 1 ), suppose that R + In + R 2D + qr has exactly two eigenvalues in the half-plane Re p > 0 and no eigenvalues in the strip Re p 0. Then there exists a dierentiable function ( ) such that all assertions of Theorem 1 are true and in addition , system (4) has an arbitrary prescribed number of orbitally stable cycles lying in the ball {x : |x| h} for any given h > 0. 3. PROOF OF THEOREM 1 It follows from assumption 1 of the theorem that 0 < ( ) < 1 2 for (, ). This, together with properties of the spectrum of the matrix R, implies that the assumptions of Theorem 1.1.1 in [5] and Remark 2 to it are satised, and consequently, system (3) has a unique equilibrium, which is globally asymptotically stable. Lemma 1. Under assumptions 1 and 5 of the above theorem , system (4) has the unique equilibrium x(1) = 0, x(2) = 0.
DIFFERENTIAL EQUATIONS Vol. 47 No. 1 2011

BURKIN, SOBOLEVA

Proof. It follows from the condition (0) = 0 that x(1) = 0, x(2) = 0 is an equilibrium of system (4). Let us show that this system has no other equilibrium. Suppose the contrary: there 2 2 + 2 = 0) such that Ax + b = 0, where exist numbers 1 and 2 (1 = r x(1) , 2 = r x(2) , 1 (1) (2) x = col(x , x ) and = col((1 ), (2 )). It follows from the denition of the function W (p) that this relation can be represented in the form 1 + u(0)(1 ) + v (0)(2 ) = 0, 2 + v (0)(1 ) + u(0)(2 ) = 0. (12)

Assumptions 1 and 5 of the theorem imply that if = 0, then 0 < ( )/ < 1 and the line + u(0) = 0 does not meet the sector {(, ) : 0 < / < 1 } on the plane (, ). If 1 = 0 and 2 = 0, then it follows from the second equation in system (12) that 2 + u(0)(2 ) = 0. This relation can be true only for 2 = 0. Likewise, it follows from the assumption that 2 = 0 and the rst equation in system (12) that 1 = 0. Let 1 = 0, 2 = 0 be a solution of system (12). Then (1 , 2 ) = 1 + u(0) (1 ) 2 + u(0) v 2 (0) = 0. (2 )

1 1 1 On the other hand, 1 (1 ) > 1 , 2 (2 ) > 1 , and u(0) + 1 0. Therefore, 1 2 2 1 1 (1 , 2 ) > (u(0) + 1 ) v (0) = (u(0) v (0) + 1 )(u(0) + v (0) + 1 ) 1 1 = (W2 (0) + 1 )(W1 (0) + 1 ) 0.

The resulting contradiction implies the assertion of the lemma. Lemma 2. If R is a Hurwitz matrix and assumption 3 of the theorem is satised , then there exists a positive denite 2n 2n matrix Q and a number > 0 such that the set G = {x : x Qx 0} / G, there exists a (x0 ) > 0 is positively invariant for the trajectories of system (5) and for each x0 such that x(t, x0 ) G for t > (x0 ). Proof. Let us introduce a quadratic form of the variables x and as follows: F (x, ) = 1 (2 r x(1) 1 ) + 2 (2 r x(2) 2 ) = 2 (x c ) . ) | |2 we denote the extension of the form F (x, ) to complex values of ) = 2 Re ( ( xc By F x, the argument with preservation of the Hermitian property. We try to choose a matrix Q = Q such that the inequality (13) 2x Q(Ax + b ) + F (x, ) < (|x|2 + | |2 ) with some > 0 holds for arbitrary x R2n and R2 . Since R and R 2D are Hurwitz matrices, it follows that A is a Hurwitz matrix as well. The pair (A, b) is completely controllable. Therefore, for the existence of a matrix Q satisfying inequality (13), it is sucient [5, Th. 1.2.7] ] < 0 holds for all [0, ). By (6), this inequality [(iI2n A)1 b, that the inequality F acquires the form [0, ). (14) 2 Re W (i ) + I2 > 0, By taking into account the form (6) of the matrix W (p), we nd that relation (14) is a consequence of inequality (10). By setting = 0 in (13), we nd that the matrix Q satises the inequality AQ + Q A < I2n , which, together with Lemma 1.2.4 in [5], implies the inequality Q > 0. Let (0, 2 ). Set ( ) = 1 ( ) + 2 ( ) in (4), where 1 ( ) = ( ) if | | < 0 if | | 0 , 2 ( ) = ( ) 1 ( ).
Vol. 47 No. 1 2011

DIFFERENTIAL EQUATIONS

ON A SMALE PROBLEM

Note that 2 ( ) is a bounded function for (, ). Set = 1 + 2 , where 1 = col(1 (1 ), 1 (2 )), 2 = col(2 (1 ), 2 (2 )).

Let Z (x) = x Qx , where is a number to be specied below. By taking into account inequality (13), for the derivative of the function Z (x) according to system (5), we obtain (x) = 2x Q[Ax + b(1 + 2 )] < |x|2 F (x, 1 ) + 2x Qb2 . Z It follows from the denition of the function F (x, ), the form of the function 1 , and relation (11) that F (x, 1 ) 0 on solutions of system (5). Since 2 is a bounded function, we nd that, for an arbitrary number > 0, there exists a number > 0 such that 2x Qb2 < |x|2 for |x|2 > . (x) 1 |x|2 for |x| > , where 1 = > 0. Let > 0 be the maximum eigenvalue Therefore, Z of the matrix Q. Set > . Then the inequality |x|2 > 1 > holds for Z (x) 0. Therefore, (x) < 1 Z if Z (x) 0. (15)

Hence it follows that, for any solution x(t, x0 ) with x0 {x : Z (x) 0}, there exists a (x0 ) > 0 such that the inequality Z [x(t, x0 )] < 0 holds for t > (x0 ). Indeed, by integrating relation (15) from 0 to t, we obtain Z [x(t, x0 )] Z (x0 ) 1 t. Therefore, the inequality Z [x(t, x0 )] 0 can be true only for t < Z (x0 )(1 )1 . This completes the proof of Lemma 2. Lemma 3. Under assumptions 1, 2, and 4 of the theorem , there exists a nonsingular 2n 2n matrix H = H with exactly two negative eigenvalues and a number > 0 such that two arbitrary solutions x1 (t) and x2 (t) of system (5) satisfy the relation [x1 (t) x2 (t)] + 2V [x1 (t) x2 (t)] < |x1 (t) x2 (t)|2 , V (x) is the derivative of the function V (x) according to system (5). where V (x) = x Hx and V Proof. The complete controllability of the pair (R, q ) implies that of the pair (R + In , q ). It follows from condition (8) and Theorem 1.2.7 in [5] that there exists a matrix M = M and a number 1 > 0 such that the inequality 2x M [(R + In )x + q ] + (1 r x ) < 1 (|x|2 + 2 ) (17) (16)

is true for all x Rn and R1 . In (17), we set = r x, where [0, 1 ]. It follows from (17) that (18) x [M (R + In + qr ) + (R + In + qr ) M ]x < 1 |x|2 . If = 0, then, by Lemma 1.2.4 in [5], from (18) and from the conditions imposed on the spectrum of the matrix R, we obtain M > 0. Therefore, by the same lemma, all matrices R + In + qr with [0, 1 ] have the same number of eigenvalues in the half-plane Re p > 0, i.e., are Hurwitz matrices. We introduce the quadratic form F1 (x, ) = 1 (1 r x(1) 1 ) + 2 (1 r x(2) 2 ) = 1 (x c ) . Note that the complete controllability of the pair (A, b) implies that of the pair (A + I2n , b). Now, proceeding by analogy with the proof of Lemma 2, we nd that conditions (8) and (9) ensure the existence of a 2n 2n matrix H = H such that the inequality 2x H [(A + I2n )x + b ) + F1 (x, ) < (|x|2 + | |2 ) (19)

holds for all x R2n and R2 and for some > 0. In (19), we set 1 = 2 = (0)qr . Then it follows from (19) that x [H (A + I2n + (0)bc ) + (A + I2n + (0)bc ) H ]x < |x|2 .
DIFFERENTIAL EQUATIONS Vol. 47 No. 1 2011

BURKIN, SOBOLEVA

Since det(A + I2n + (0)bc ) = det(R + In + (0)qr ) det(R 2D + In + (0)qr ) and, as was proved above, R + In + (0)qr is a Hurwitz matrix, it follows from assumption 4 of the theorem that the matrix A + I2n + (0)bc has exactly two eigenvalues in the half-plane Re p > 0. Therefore, H is a nonsingular matrix and has exactly two negative eigenvalues. Let x1 (t) and x2 (t) be two solutions of system (5). Set z (t) = x1 (t) x2 (t). Then z (t) is a solution of the system dz/dt = Az + bc f (t)z , where the matrix f (t) has the form f (t) = diag{ [1 (t)r x1 (t) + (1 1 (t))r x2 (t)]In , [2 (t)r x1 (t) + (1 2 (t))r x2 (t)]In }, 0 2 (t) 1. 0 1 (t) 1,
(1) (1) (1) (1)

By setting x = z (t) and = c f (t)z (t) in (19) and by taking into account the structure of the quadratic form F1 (x, ) and assumption 1 of the theorem, we nd that relation (16) holds. The proof of Lemma 3 is complete. Consider the set = {x : x Hx 0}. This set is positively invariant for trajectories of system (5), and its boundary is intersected inwards by all trajectories of that system that are adjacent to it. The last assertion implies that the condition x(t0 )Hx(t0 ) = 0 (x(t0 ) = 0) results in the inequality x (t)Hx(t) < 0 for t > t0 . Indeed, by setting x2 (t) 0 in (16), for an arbitrary solution x(t) of system (5), we have [x(t)] + 2V [x(t)] < |x(t)|2 . V This, together with Theorem 2.9 in [6], implies the above assertions. Let us clarify the local behavior of trajectories of system (5) in a neighborhood of the point x = 0 in the invariant set . To this end, note that it follows from the argument used in the proof of Lemma 2 that the matrix A + (0)bc has exactly two eigenvalues in the half-plane Re p > 0 and no eigenvalues in the strip Re p 0. We rewrite system (5) in the form dx = (A + (0)bc )x + ( (0)bc x). dt Obviously, |g(x)| = | (0)bc x| = o(|x|). Let x = T y be a nonsingular transformation reducing system (16) to the form y 1 = B1 y1 + g1 (y ), y 2 = B2 y2 + g2 (y ), y = col(y1 , y2 ),

where B1 and B2 are anti-Hurwitz matrices of sizes (2n 2) (2n 2) and 2 2, respectively; is a positively denite matrix, g1 (y ) = o(|y |), and g2 (y ) = o(|y |). Set F y = moreover, B2 + B2 col(0, y2 ) and N = T HT . Then, as was shown in [6, p. 133], y N y < 0 for F y = 0. Since the relation |y1 | |y2 | with some > 0 is satised for an arbitrary y T = {y : y N y 0}, we have d (|y2 |2 ) = y2 (B2 + B2 )y2 + 2y2 g2 (y ) k|y2 |2 + 2y2 g2 (y ) > 0 dt and is some for y T , |y2 | < , where k is the minimum eigenvalue of the matrix B2 + B2 suciently small number. In other words, we have proved the following assertion. Lemma 4. There exists a number > 0 such that the trajectories of all solutions x(t, x0 ) of system (5) such that x0 {x : |F T 1 x| < } leave that set for some t > 0 and the trajectories issuing from the exterior of that set in the positively invariant set never get into it. In addition , S = {x : |F T 1 x| = , x Hx 0} is a contact-free surface for trajectories of system (5). It follows from the properties of the spectrum of the matrix A + (0)bc that system (5) has solutions such that |x(t, x0 )| 0 as t +. The well-known Hartman theorem [8, Chap. IX,
DIFFERENTIAL EQUATIONS Vol. 47 No. 1 2011

ON A SMALE PROBLEM

Sec. 7] implies that, in a neighborhood of the point x = 0, the set of trajectories of such solutions is a (2n 2)-dimensional manifold. By virtue of the assumption on the continuous dierentiability of the function ( ), the solutions x(t, x0 ) of system (5) continuously depend on the initial data for the values of t with arbitrary sign. Therefore, the set of initial conditions of solutions of system (5) with the property |x(t, x0 )| 0 as t + is a (2n 2)-dimensional manifold in R2n . / and V (x0 ) > 0. Then there exists It follows from Lemmas 3 and 4 that R2n \. Let x0 a t1 > 0 such that V [x(t, x0 )] > 0 for t (0, t1 ). Let us show that t1 < +. If we suppose the contrary, then from (11) with x1 (t) = x(t, x0 ) and x2 (t) 0, we obtain 0 |x(t, x0 )|2 dt V (x0 ). By Lemma 2, the solution x(t, x0 ) is bounded. It follows from (5) that the derivative of this solution is bounded as well. Therefore, by Theorem 2.12 in [5], limt+ |x(t, x0 )| = 0. The contradiction obtained implies that t1 < +. Let = G {x : |F T 1 x| }, where the sets G and and the number have been dened above. It follows from Lemmas 14 that this set is bounded and does not contain the unique equilibrium x = 0 of system (5), and its boundary is intersected inwards by all trajectories of that system that are adjacent to it. Then, as was shown in [9], it follows from (11) that any trajectory of system (5) in converges to a closed trajectory of that system as t +, and contains at least one closed trajectory that is orbitally stable. If ( ) is an analytic function, then in there are at most nitely many closed trajectories of system (5); in addition, at least one of them is asymptotically orbitally stable. It follows from the above considerations that, for the trajectories of all solutions x(t, x0 ) of system (5) except for those satisfying the inclusion x0 , there exists a t(x0 ) such that the inclusion x(t, x0 ) is true for t > t(x0 ). The proof of Theorem 1 is complete. Note that, by replacing the function ( ) in (5) by the function l1 (l ) with suciently large l, one can ensure the inclusion {x : |x| h} with arbitrary h > 0. We do not present the proof of Theorem 2, since it is cumbersome. Note only that this theorem can be proved by the scheme that was used in the proof of Theorems 3 and 5 in [7]. 4. EXAMPLES Example 1. Consider the system of kinetics of two cells with three chemicals, that is, system (4) with 0 1 0 0 1 1 0 0 R= 0 q = 0 , r = 0, D = 0 2 0 , 0 1 , 1 0 0 3

where = 1 + 2 + 3 , = 1 2 + 1 3 + 2 3 , = 1 2 3 , and 0 < 3 < 2 < 1 . Then W1 (p) = (p3 + p2 + p + )1 . By expanding W1 (p) into partial fractions, we obtain the inequality Re W1 (i ) = 2 1 2 2 (1 3 )(1 2 )( + 1 ) (1 2 )(2 3 )( 2 + 2 2) 2 3 + (2 3 )(1 3 )( 2 + 2 (1 2 )(2 3 )( 2 + 2 3) 2) 1 . 2 (1 2 )(2 3 )

1 Therefore, 0 + Re W1 (i ) > 0 for 0 < 2 (1 2 )(2 3 ). In other words, for an arbitrary 0 < 2 (1 2 )(2 3 ), the sector [0, 0 ] is a sector of absolute stability of system (3). By proceeding in a similar way, we obtain the inequality 1 0 + Re W1 (i ) > 0

for 0 < (2 )(1 2 )(1 3 ). Therefore, if is small, then 1 satisfying relation (8) can be chosen to be suciently close to 0 .
DIFFERENTIAL EQUATIONS Vol. 47 No. 1 2011

BURKIN, SOBOLEVA

Consider the matrix D with 1 = 2 = 3 = 0. Then W2 (p) = (p3 + p2 + p + + 2 )1 . In addition, det(R 2D + qr pI ) = p3 + p2 + p + + 2 + . The matrix R 2D + qr is a Hurwitz matrix for < 2 . Therefore, one can choose a number > 0 such that 0 < < 2 2 (1 2 )(2 3 ). Obviously, there always exists a positive number 2 < 1 for which relation (10) holds. root , > 1 , and Take a such that the polynomial p3 + p2 + p + + 2 + has one real a pair of complex conjugate roots m ni, where m is small and n is close to . Then 1 + Re W2 (i ) 1 |W2 (i )| 1 ( )[(m )2 + n2 ]1 > 0, m < , for < ( )[(m )2 + n2 ]. Therefore, one can nd and 1 < min(2 )(1 2 )(1 3 ), ( )[(m )2 + n2 ] = (2 )(1 2 )(1 3 )

such that relations (8) and (9) are satised. If 1 > 0, 2 > 0, and 3 > 0, then the matrix D is similar to a positive diagonal matrix, and W2 (p, 1 , 2 , 3 ) = [ p3 + (1 , 2 , 3 )p2 + (1 , 2 , 3 )p + 2 + (1 , 2 , 3 )]1 W2 (p) as i 0 (i = 1, 2, 3). Therefore, for a system with the above-mentioned values of R, q , and r , one can always choose 1 , 2 , and D with suciently small 1 , 2 , and 3 such that relation (10) and assumption 2 of Theorem 1 are satised. In this case, assumption 5 is obviously satised. If ( ) is a dierentiable function satisfying conditions (7) and (11) and the relations 2 < (0) < 1 and (0) = 0, then assumption 4 of Theorem 1 is satised, and therefore, all of its assertions hold. At the same time, as Theorem 2 claims, under the above-mentioned conditions, one can choose a nonlinearity ( ) so as to ensure that the considered system has an arbitrary prescribed number of orbitally asymptotically stable cycles. Example 2. Consider system (4) with R= 0 0 1 0 0 1 , q = 0 , 1 0 r = 1 , 0 1 D= 0 0.1 0 0 2 0 .

15 23 9 Then

450 0 1

W1 (p) = (1 p)(p3 + 9p2 + 23p + 15)1 , W2 (p) = (0.8 p)(p3 + 15.2p2 + 72.0p + 950.4)1 .

Assumption 2 of Theorem 1 is true for = 0.3 and 1 = 13. Obviously, assumption 5 of that theorem is satised as well. Relation (10) holds for 2 = 8.5. The matrix D is similar to the positive diagonal matrix diag(0.1, 2, 1). Let a dierentiable function ( ) satisfy assumption 1 of Theorem 1, relation (11), and the inequality 9 < (0) < 13. Then assumption 4 of the theorem is true as well. Therefore, by the assertion of the theorem, all trajectories of system (3) tend to the equilibrium point x = 0 as t +, and almost every trajectory of system (5) tends to some periodic trajectory of that system as t +. REFERENCES
1. Smale, S., Mathematical Model of the Interaction of Two Cells with the Use of Turing Equations, in Bifurkatsiya rozhdeniya tsikla i ee prilozheniya (Hopf Bifurcation and Its Applications), Moscow, 1980. 2. Turing, A.M., The Chemical Basis of Morphogenesis, Philos. Trans. R. Soc., 1925, pp. 3772. 3. Tomberg, E.A. and Yakubovich, V.A., Conditions for Self-Induced Oscillations in Nonlinear Systems, Sib. Mat. Zh., 1989, vol. 30, no. 4, pp. 180194.
DIFFERENTIAL EQUATIONS Vol. 47 No. 1 2011

ON A SMALE PROBLEM

4. Tomberg, E.A. and Yakubovich, V.A., On a Problem of Smale, Sib. Mat. Zh., 2000, vol. 41, no. 4, pp. 926928. 5. Gelig, A.Kh., Leonov, G.A., and Yakubovich, V.A., Ustoichivost nelineinykh sistem s needinstvennym sostoyaniem ravnovesiya (Stability of Nonlinear Systems with a Nonunique Equilibrium State), Moscow: Nauka, 1978. 6. Leonov, G.A., Burkin, I.M., and Shepelyavyi, A.I., Chastotnye metody v teorii kolebanii. Ch.1. Mnogomernye analogi uravneniya Van-der-Polya i dinamicheskie sistemy s tsilindricheskim fazovym prostranstvom (Frequency Methods in the Theory of Oscillations. Part 1. Multidimensional Analogs of the van der Pol Equation and Dynamical Systems with Cylindrical Phase Space), St. Petersburg: SanktPeterburg. Gos. Univ., 1992. 7. Burkin, I.M., The Buer Phenomenon in Multidimensional Dynamical Systems, Dier. Uravn., 2002, vol. 38, no. 5, pp. 585595. 8. Hartman, Ph., Ordinary Dierential Equations , New York, 1969. Translated under the title Obyknovennye dierentsialnye uravneniya , Moscow, 1970. 9. Smith, R.A., Orbital Stability for Ordinary Dierential Equations, J. Dierential Equations , 1987, vol. 69, no. 2, pp. 265287.

DIFFERENTIAL EQUATIONS

Vol. 47

No. 1

2011

Вам также может понравиться