Applied Mathematics and Computation 117 (2001) 313±320 www.elsevier.com/locate/amc
Perturbing Lyapunov functions and stability criteria for initial time dierence F.A. McRae Department of Mathematics, Catholic University of America, Washington, DC 20064, USA
Abstract Stability criteria for dierential equations where the initial time for each solution is dierent is developed using the method of perturbing Lyapunov functions. Ó 2001 Elsevier Science Inc. All rights reserved. Keywords: Initial value problems; Variable initial times; Perturbing Lyapunov functions; Stability criteria
1. Introduction In dealing with real world phenomenon it is impossible not to make errors in the starting time. Hence it is important to study the variation in initial times. This creates problems in comparing any two solutions which dier in initial starting time. An investigation of initial value problems of dierential equations where the initial time changes with each solution in addition to the change of space has been initiated in [3]. In those situations where the Lyapunov function does not satisfy all the desired conditions, it is fruitful to perturb the Lyapunov functions, rather than discard it [1,2]. Thus, the notion of perturbing Lyapunov functions enables us to discuss nonuniform properties of solutions of dierential systems under weaker assumptions. In this paper we develop stability criteria for initial value problems of differential equations with initial time dierences using the method of perturbing Lyapunov function so that the usual results on stability criteria result as a consequence. E-mail address:
[email protected] (F.A. McRae). 0096-3003/01/$ - see front matter Ó 2001 Elsevier Science Inc. All rights reserved. PII: S 0 0 9 6 - 3 0 0 3 ( 9 9 ) 0 0 2 0 6 - 4
314
F.A. McRae / Appl. Math. Comput. 117 (2001) 313±320
2. Preliminaries Consider the dierential system x0 f
t; x;
2:1
where f 2 CR Rn ; Rn . Let x
t; t0 ; x0 and x
t; s0 ; y0 be the solutions of (2.1) through
t0 ; x0 and
s0 ; y0 , respectively. Suppose that x
t x
t; t0 ; x0 is the given solution relative to which we shall study stability criteria. Let g s0 ÿ t0 . We shall ®rst de®ne the notion of stability in the present framework. De®nition 2.1. The solution x
t of (2.1) is said to be (i) equistable if given > 0 and t0 2 R , there exist d d
t0 ; > 0 and r r
t0 ; > 0 such that kx0 ÿ y0 k < d, jgj < r implies kx
t g; s0 ; y0 ÿ x
t; t0 ; x0 k < , t P t0 ; (ii) uniformly stable if d and r in (1) is independent of t; (iii) equi-asymptotically stable if (1) holds and given > 0, t0 2 R , there exist d0
t0 > 0, r0
t0 > 0 and a T T
t0 ; > 0 such that kx0 ÿ y0 k < d0 , jgj < r0 implies kx
t g; s0 ; y0 ÿ x
t; t0 ; x0 k < , t P t0 T ; (iv) uniform asymptotically stable if (2) holds and d0 , r0 and T in (3) are independent of t0 . We need the following comparison theorem [4] before we can proceed. Theorem 2.1. Assume that V 2 CR Rn ; R , V
t; u is locally Lipschitzian in u and D V
t; u; g 6 g
t; V
t; u; jgj, where u u
t x
t g; s0 ; y0 ÿ x
t; t0 ; x0 , g 2 CR3 ; R and r
t; t0 ; x0 ; jgj is the maximal solution of x0 g
t; x; jgj, w
t0 ÿ w0 P 0, existing the t P t0 . Then V
t0 ; x0 ÿ y0 6 x0 implies V
t; x
t g; s0 ; y0 ÿ x
t; t0 ; x0 6 r
t; t0 ; x0 ; jgj;
t P 0:
3. Main results Theorem 3.1. Assume that (i) V1 2 CR S
q; R , V1
t; x is locally Lipschitzian in x and V1
t; 0 0 and D V1
t; x 6 g1
t; V1
t; x; jgj;
t; x 2 R S
q;
where g1 2 CR3 ; R and S
q x 2 Rn : kxk < q; (ii) for every c > 0, there exists a V2 2 R S
q \ S c
c; R , V2
t; x is locally Lipschitzian in x, b
kxk 6 V2
t; x 6 a
kxk;
t; x 2 R S
q \ S c
c;
F.A. McRae / Appl. Math. Comput. 117 (2001) 313±320
315
where a; b 2 K f/ 2 C
0; q; R : /
u is increasing in u and /
u ! 0 as u ! 0g and D V1
t; x D V2
t; x 6 g2
t; V1
t; x V2
t; x; jgj;
t; x 2 R S
q \ S c
c; where g2 2 CR3 ; R; (iii) the scalar differential equation w01 g1
t; w1 ; jgj;
w1
t0 w10 P 0;
3:1
is equistable and the scalar differential equation w02 g2
t; w2 ; jgj;
w2
t0 w20 P 0;
3:2
is uniformly stable. Then the solution x
t x
t; t0 ; x0 of the system (2.1) is equistable. Proof. Let 0 < < q and t0 2 R be given. The uniform stability of (3.2) implies that, given b
> 0 and t0 2 R , there exist a d0 d0
> 0 and r1 r1
> 0 such that 0 6 w20 < d0 and jgj < r1 implies w2
t; t0 ; w20 ; jgj < b
;
t P t0 ;
3:3
where w2
t; t0 ; w20 ; jgj is any solution of (3.2). Because of the hypothesis on a
u, there exists d2 d2
> 0 such that a
d2 <
d0 : 2
3:4
Also, since (3.1) is equistable, given
d0 =2 > 0, t0 2 R , there exists d d
t0 ; and r2 r2
t0 ; > 0 such that 0 6 w10 < d and jgj < r2 implies w1
t; t0 ; w10 ; jgj <
d0 ; 2
t P t0 ;
3:5
where w1
t; t0 ; w10 ; jgj is any solution of (3.1). Choose w10 V1
t0 ; x0 ÿ y0 . Since V1
t; x is continuous and V1
t; 0 0, there exists d1 > 0 such that kx0 ÿ y0 k < d1
and
V1
t0 ; x0 ÿ y0 < d
3:6
hold simultaneously. Set d min
d1 ; d2 , r min
r1 ; r2 . We claim that kx0 ÿ y0 k < d and jgj < r implies that kx
t g; s0 ; y0 ÿ x
t; t0 ; x0 k < , t P t0 . If not, there exists a solution x
t; s0 ; y0 of (2.1) and t1 ; t2 > t0 such that kx
t1 g; s0 ; y0 ÿ x
t1 ; t0 ; x0 k d2 ;
316
F.A. McRae / Appl. Math. Comput. 117 (2001) 313±320
kx
t2 g; s0 ; y0 ÿ x
t2 ; t0 ; x0 k
3:7
and d2 6 kx
t g; s0 ; y0 ÿ x
t; t0 ; x0 k 6 ;
t 2 t1 ; t2 :
Set d2 c so that the existence of V2 satisfying (ii) is assured. Now, setting m
t V1
t; x
t g; s0 ; y0 ÿ x
t; t0 ; x0 V2
t; x
t g; s0 ; y0 ÿ x
t; t0 ; x0 , t 2 t1 ; t2 we get, using standard arguments, D m
t 6 g2
t; m
t; jgj;
t 2 t1 ; t2 ;
which yields, using Theorem 2.1 and (ii), V1
t2 ; x
t2 g; s0 ; y0 ÿ x
t2 ; t0 ; x0 V2
t2 ; x
t2 g; s0 ; y0 ÿ x
t2 ; t0 ; x0 6 r2
t2 ; t1 ; V1
t1 ; x
t1 g; s0 ; y0 ÿ x
t1 ; t0 ; x0 V2
t1 ; x
t1 g; s0 ; y0 ÿ x
t1 ; t0 ; x0 ; jgj; where r2
t; t1 ; w20 ; jgj is the maximal solution of (3.2) r2
t1 ; t1 ; w10 ; jgj w20 . We also have using Theorem 2.1 and (i) V1
t1 ; x
t1 g; s0 ; y0 ÿ x
t1 ; t0 ; x0 6 r1
t1 ; t0 ; V
t0 ; x0 ÿ y0 ; jgj;
3:8 such
that
3:9
where r1
t; t0 ; w10 ; jgj is the maximal solution of (3.1). By (3.5) and (3.6), we get V1
t1 ; x
t1 g; s0 ; y0 ÿ x
t1 ; t0 ; x0 < d0 =2:
3:10
Also, by (3.4), (3.7) and (ii), we have V2
t1 ; x
t1 g; s0 ; y0 ÿ x
t1 ; t0 ; x0 6 a
kx
t1 g; s0 ; y0 ÿ x
t1 ; t0 ; x0 k a
d2 < d0 =2:
3:11
Therefore, using (3.8), (3.10), (3.11) and (3.3), we get V1
t2 ; x
t2 g; s0 ; y0 ÿ x
t2 ; t0 ; x0 V2
t2 ; x
t2 g; s0 ; y0 ÿ x
t2 ; t0 ; x0 < b
:
3:12
From (3.7), (ii) and the fact that V1
t; x P 0, we get b
b
kx
t2 g; s0 ; y0 ÿ x
t2 ; t0 ; x0 k 6 V2
t2 ; x
t2 g; s0 ; y0 ÿ x
t2 ; t0 ; x0 6 V1
t2 ; x
t2 g; s0 ; y0 ÿ x
t2 ; t0 ; x0 V2
t2 ; x
t2 g; s0 ; y0 ÿ x
t2 ; t0 ; x0 :
3:13
Hence (3.12) and (3.13) lead to the contradiction b
< b
, proving the equistability of (2.1). Next, we consider stability criteria for equi-asymptotic stability of (2.1).
F.A. McRae / Appl. Math. Comput. 117 (2001) 313±320
317
Theorem 3.2. Assume that (i) V1 2 CR S
q; R , V
t; x is locally Lipschitzian in x, V1
t; 0 0 and D V1
t; x 6 ÿ c
W
t; x g1
t; V1
t; x; jgj;
t; x 6 R S
q;
where g1 2 CR3 ; R, g10
t; u; jgj is increasing in u, W 2 CR S
q; R , W
t; x is locally Lipschitzian in x, D W
t; x is bounded above or below and W
t; x P b0
kxk where b0 ; C 2 K. (ii) Assumptions (ii) and (iii) in Theorem 3.1 hold. Then, the solution x
t x
t; t0 ; x0 of (2.1) is equi-asymptotically stable. Proof. The system (2.1) is equistable by Theorem 3.1. For q; let d0 d0
q; t0 > 0, r0 r0
q; t0 > 0 so that kx0 ÿ y0 k < d0 and jgj < r0 imply kx
t g; s0 ; y0 ÿ x
t; t0 ; x0 k < q, t P t0 . We want to show that kx
t g; s0 ; y0 ÿ x
t; t0 ; x0 k ! 0
3:14
0
as t ! 1 when kx0 ÿ y0 k < d and jgj < r0 . Since W
t; x is positive de®nite, to prove (3.14) it is enough to show that limt!1 W
t; x
t g; s0 ; y0 ÿx
t; t0 ; x0 0. We claim that limt!1 W
t; x
t g; s0 ; y0 ÿ x
t; t0 ; x0 0 when kx0 ÿ y0 k < d0 and jgj < r0 . If not, then for any a > 0, there exist two divergent sequences ftn g, ftn0 g such that W
ti ; x
ti g; s0 ; y0 ÿ x
ti ; t0 ; x0 a=2;
3:15
W
ti0 ; x
ti0 g; s0 ; y0 ÿ x
ti0 ; t0 ; x0 a and a 6W 2
t; x
t g; s0 ; y0 ÿ x
t; t0 ; x0 6 a;
t 2 ti ; ti0 ;
i 1; 2; . . .
We could also have instead of (3.15), W
ti ; x
ti g; s0 ; y0 ÿ x
ti ; t0 ; x0 a; W
ti0 ; x
ti g; s0 ; y0 ÿ x
ti0 ; t0 ; x0 a=2; a=2 6 W
t; x
t g; s0 ; y0 ÿ c
t; t0 ; x0 6 a; t 2 ti ; ti0 ; i 1; 2; . . . R t0 Suppose that D W
t; x 6 M. Then tii D W
s; x
s ds 6 M
ti0 ÿ ti which implies ti0 ÿ ti P a=2M
for each i:
Set
Z L
t; x V1
t; x
t0
t
c
W
s; x
s ds:
3:16
3:17
318
F.A. McRae / Appl. Math. Comput. 117 (2001) 313±320
We get, by using standard arguments and (i), D L
t; x 6 D V1
t; x C
W
t; x
t 6 g1
t; V
t; x
6 g1
t; L
t; x
3:18
Hence, by Theorem 2.1 and (i), we obtain L
t; x
t g; s0 ; y0 ÿ x
t; t0 ; x0 6 r1
t; t0 ; V1
t0 ; x0 ; jgj;
t P t0 ;
3:19
where r1
t; t0 ; w10 ; jgj is the maximal solution of (3.1). Now, using (3.15)± (3.18), we get the contradiction 0 6 V1
t; x
t g; s0 ; y0 ÿ x
t; t0 ; x0 Z t c
W
s; x
s g; s0 ; y0 ÿ x
s; t0 ; x0 ds 6 r1
t; t0 ; V1
t0 ; x0 ÿ y0 ; jgj ÿ t0
n aX a na d0 d0 <0
ti0 ÿ ti < ÿ c 6 ÿc 2 2 2 i1 2 2M
for suciently large n. Therefore, W
t; x
t g; s0 ; y0 ÿ x
t; t0 ; x0 ! 0 as t ! 1 and since, by assumption, W
t; x P b
kxk it follows that kx
t g; s0 ; y0 ÿ x
t; t0 ; x0 k ! 0 as t ! 1 and the proof is complete. Remark 3.1. In Theorem 3.1, 1. if V1
t; x 0, g2
t; u; jgj 0 and V1
t; x is positive de®nite, the system (2.1) is equistable, 2. if V1
t; x 0, g1
t; u; jgj 0, the system (2.1) is uniformly stable. Remark 3.2. In Theorem 3.2, if V2
t; x 0, g2
t; u; jgj 0 and the V1
t; x is positive de®nite, the system (2.1) is equi-asymptotically stable. Remark 3.3. In Theorem 3.1, let V1
t; x 0, g1
t; u; jgj 0. Then, for uniform asymptotic stability of (2.1), we need to strengthen the assumption on D V2
t; x, which we state as a corollary. Corollary 3.1. In Theorem 3.1, assume V1
t; x 0 and g1
t; u; jgj 0. Suppose the condition on D V2
t; x is strengthened to D V2
t; x 6 ÿ c
W
t; x g2
t; V2
t; x; jgj;
t; x 2 R S
q \ S c
c;
where W 2 CR S
q; R , W
t; x P b
kxk, c; b 2 K and g2
t; u; jgj is nondecreasing in u. Then, the solution x
t x
t; t0 ; x0 of (2.1) is uniformly asymptotically stable.
F.A. McRae / Appl. Math. Comput. 117 (2001) 313±320
319
Proof. Let 0 < < q. By uniform stability of the solution x
t of (2.1), we have kx0 ÿ y0 k < d
; implies
jgj < r
kx
t g; s0 ; y0 ÿ x
t; t0 ; x0 k < ;
t P t0 :
Choose q and let d0 d
q, r0 r
q, then kx0 ÿ y0 k < d0 implies kx
t g; s0 ; y0 ÿ c
t; t0 ; x0 k < q, t P t0 . To prove uniform attractivity, we shall show that there exists a t in t0 ; t0 T , where T
b
q 1 c
b
d
3:20
such that W
t ; x
t < b
d
q for any solution of (2.1) with kx0 ÿ y0 k < d0 . If not, for all t 2 t0 ; t0 T , W
t; x
t > b
d
;
t 2 t0 ; t0 T :
3:21
Using (3.20) and (3.21), the fact that r2
t; t0 ; w20 < b
q the assumption on D V2
t; x and Theorem 2.1, we get the contradiction 0 6 V2
t0 T ; x
t0 T ; g; s0 ; y0 ÿ x
t0 T ; t0 ; c0 6 r2
t0 T ; t0 ; V2
t0 ; x0 ÿ y0 ; jgj Z t0 T c
W
s; x
sds < b
q ÿ c
b
dT ÿ t0 b
q 1 < 0: b
q ÿ c
b
d c
b
s Hence W
t ; x
t < b
d
. Now, since W
t; x is positive de®nite, we get b
kx
t k 6 w
t ; x
t < b
d
or kx
t k < d
. Therefore, by uniform stability of (2.1), we get kx
tk < , t P t0 T , and hence the solution of x
t of (2.1) is uniform asymptotically stable. R1 Remark 3.4. The function g2
t; u; jgj jgjk
t, with 0 k
s ds N is admissible in Remark 3.1 (2) to yield uniform stability of (2.1) where d0
b
2
and
r1
b
: 2N
Remark 3.5. The function g2
t; l; jgj jgjk
t as in Remark 3.4 is admissible in Remark 3.3 to yield uniform asymptotic stability of (2.1) where d0 d
q, r0 b
q=2N , T T
q b
q=c
b
d 1.
320
F.A. McRae / Appl. Math. Comput. 117 (2001) 313±320
References [1] V. Lakshmikantham, S. Leela, On perturbing Lyapunov functions, Math. Systems Theory 10 (1976) 85±90. [2] V. Lakshmikantham, S. Leela, A.A. Martynyuk, Stability Analysis of Nonlinear Systems, Marcel Dekker, New York, 1989. [3] V. Lakshmikantham, A.S. Vatsala, Dierential inequalities with initial time dierence and applications, J. Inequalities Appl. (to appear). [4] V. Lakshmikantham, S. Leela, J. Vasundhara Devi, Stability criteria for solutions of dierential equations relative to initial time dierence (to appear).