ASYMPTOTIC ESTIMATE FOR DIFFERENTIAL

tm
Mathemati al Publi ations
DOI: 10.2478/v10127-011-0009-1
Tatra Mt. Math. Publ. 48 (2011), 91–99

EQUATION WITH POWER COEFFICIENTS AND

POWER DELAYS
Petr Kundrát

ABSTRACT. The paper analyzes the asymptotic bounds of solutions of differ-

ential equation with power coefficients, power delays and a forcing term in the
Pm αj
form ẏ(t) = j=0 aj t y(tλj ) + f (t), where a0 is a negative real, λ0 = 1 and
0 < λi < 1, i = 1, . . . , m. Some additional assumptions on power coefficients and
a forcing term f (t) are considered to obtain an asymptotic estimate for solutions
of the studied differential equation. The application of the result is illustrated by
several examples.

1. Introduction

The paper deals with the asymptotic estimate of solutions of the differential
equation with several power delays and a forcing term in the form
m
X
ẏ(t) = aj tαj y(tλj ) + f (t), t ∈ I := [t0 , ∞), t0 ≥ 1, (1)
j=0

where
a0 < 0, λ0 = 1, 0 < λ1 ≤ λ2 ≤ · · · ≤ λm < 1, ai , αi ∈ R, i = 1, . . . , m and f
is a continuous function on I.
The asymptotic behaviour of solutions of delay differential equations is still
widely investigated as evidenced by many papers dealing with this topic, see, e.g.,

c 2011 Mathematical Institute, Slovak Academy of Sciences.

2010 M a t h e m a t i c s S u b j e c t C l a s s i f i c a t i o n: 34K25, 39B22.
K e y w o r d s: asymptotic estimate, delay differential equation.
The work was supported by the grant # GA201/08/0469 “Oscillatory and asymptotic prop-
erties of solutions of differential equations” of Czech Science Foundation, the research plan
MSM 0021630518 “Simulation modelling of mechatronic systems” of the Ministry of Edu-
cation, Youth and Sports of the Czech Republic and the grant FSI-S-10-14 “Modern meth-
ods of mathematical modeling of engineering problems” of Grant agency of Brno University
of Technology.

91
 PETR KUNDRÁT

I s e r l e s [5], [6], L i u [7] or M a k a y and T e r j é k i [8]. To recall some pa-

pers closely related to the studied problem we mention Č e r m á k [2], where
the equation with power coefficients and a proportional delays have been stud-
ied. Paper [1] gives asymptotic bounds of solutions of differential equation with
a more general form of coefficients and delays and with a forcing term
m
X

ẏ(t) = aj (t)y τj (t) + f (t), t ∈ [t0 , ∞),
j=0

where a0 (t) < 0 and ai (t), i = 1, . . . , m are continuous functions, τ0 (t) = t,

τi (t) < t, i = 1, . . . , m and some additional assumptions have been considered.
The following analysis of the asymptotics of the equation (1) takes advantage
of the approach utilized in [1].
The main goal of the paper is to formulate the asymptotic estimate of (1)
and discuss the obtained result on several examples. The structure of the paper
is the following: Section 2 introduces some auxiliary statements. In Section 3 is
derived the asymptotic estimate of delay differential equation (1) and Section 4
illustrates the obtained result on several examples and gives some final remarks.

2. Preliminaries

In this section we introduce some fundamentals connected with the studied

problem. First, we recall the notion of solution of (1). Denote

t−1 := tλ0 1 and I−1 := [t−1 , ∞).

By a solution of (1) we understand a real-valued function y ∈ C(I−1 ) ∩ C 1 (I)

such that y satisfies (1) on I.
Analogously, as in papers [1] or [2], we utilize a system of auxiliary functional
equations, which can be in our special case of power delays rewritten as

ψ(tλi ) = ψ(t) − log λ−1

i , t ∈ [1, ∞), i = 1, . . . , m. (2)

One can see that the common solution of the system (2) is

ψ(t) = log log t. (3)

The existence of the common solution of (2) enables us to embed the system of
delayed arguments {tλ1 , . . . , tλm } into an iteration group [ψ]. For more informa-
tion about this topic see, e.g., N e u m a n [9].

92
 ASYMPTOTIC ESTIMATE FOR DIFFERENTIAL EQUATION

3. Asymptotic estimate of the equation (1)

The aim of this section is to formulate and prove the asymptotic estimate of
solutions of the delay differential equation (1).
α −η
Theorem 1. Let y be a solution of (1), where −a0 t 0 ≥ K(log t), 0 <
P m αi α0
|a
i=1 i t | ≤ −M a0 t for all t ∈ I and suitable real constants K > 0, M > 0,
η < 1. Let f (t) ∈ C(I) be such that f (t) = O(logν t) as t → ∞ for a suitable
real ν. Then
 
γ log M log M
y(t) = O(log t) as t → ∞, γ > max η + ν, ,..., . (4)
log λ−1
1 log λ−1
m

P r o o f. First we transform the equation (1). Utilizing the substitution (arising

from the common solution of the auxiliary system (2))

s = ψ(t) = log log(t), z(s) = exp −γ log log(t) y(t) (5)
we obtain
h
α i
z ′ (s) = a0 h(s) 0 h′ (s) − γ z(s)
m
X
αi
exp −γ log λ−1 h′ (s)z µi (s)


+ ai h(s) i
i=1

+ f h(s) exp{−γs}h′ (s),

where ”′ ” means the differentiation with respect to s, h(s) = ψ −1 (s) = exp exp(s)
and µi (s) = s − log λ−1
i on ψ(I), i = 1, . . . , m. We reorganize the equation into
the form
   
 h(s) 
d   Z 
α
exp γs − a0 u 0 du z(s)

ds 
 

s0
 
h(s)
m 
 Z 

X
αi  −1 ′ α0


= ai h(s) exp −γ log λi h (s) exp γs − a0 u du z µi (s)
 
i=1 
s0

 
 h(s)
Z 
 
+ exp γs − a0 u du f h(s) exp{−γs}h′ (s),
α0


(6)

 

s0

where s0 ∈ ψ(I) is such that

α0
γ > a0 h(s) h′ (s) for all s ≥ s0 .

93
 PETR KUNDRÁT

Denote
sk := s0 + kλ−1
m , Jk := [sk−1 , sk ], k = 1, 2, . . .

Let s∗ ∈ Jk+1 . The integration of (6) over [sk , s∗ ] gives

  s∗
 h(s)
Z 
 
α0
exp γs − a0 u du z(s)

 

s0
sk
 
∗ h(s)
m Zs 
 Z 

X
αi
exp −γ log λ−1 ′ α0


= ai h(s) i h (s) exp γs − a0 u du z µi (s) ds
 
i=1 s 
s0

k
 
Zs ∗
 h(s)
Z 
 
a0 uα0 du f h(s) exp{−γs}h′ (s) ds .


+ exp γs −

 

sk s0

We express the value of z in the instant s∗ ∈ Jk+1 as

z(s∗ )
 ∗


 Z )
h(s 

= exp γ(sk − s∗ ) + a0 uα0 du z(sk )

 

h(sk )
 ∗ 
 Z )
 h(s 

+ exp a0 uα0 du − γs∗

 

s0
 
s∗ h(s)
m Z 
 Z 

X
αi
exp −γ log λ−1 h′ (s) exp a0 uα0 du


× ai h(s) i γs − z µi (s) ds
 
i=1 s 
s0

k
 ∗ 
 Z )
 h(s 

α0 ∗
+ exp a0 u du − γs

 

s0
 
Zs
∗
 h(s)
Z 
 
× exp γs − a0 u du f h(s) exp{−γs}h′ (s) ds.
α0



 

sk s0

94
 ASYMPTOTIC ESTIMATE FOR DIFFERENTIAL EQUATION

Denote

s ∈ ∪kp=1 Jp ,

Mk := max |z(s)|, k = 1, 2, . . .

Then the value of z(s∗ ) can be estimated as

|z(s∗ )|
 ∗


 Z )
h(s 

≤ Mk exp γ(sk − s∗ ) + a0 uα0 du

 

h(sk )
 ∗ 
 h(s ) 
 Z 
+ Mk exp a0 uα0 du − γs∗

 

s0
 
∗
Zs X h(s)
m 
 Z 


αi
exp −γ log λ−1 ′ α0

× ai h(s) i h (s) exp γs − a0 u du ds
 
sk i=1
 
s0
 ∗ 
 h(s ) 
 Z 
+ exp a0 uα0 du − γs∗

 

s0
 
Zs
∗
 h(s)
Z 
 
× exp γs − a0 uα0 du f h(s) exp{−γs}h′ (s) ds.


(7)

 

sk s0

Further, considering the relations

m
X
αi
exp −γ log λ−1

ai h(s) i
i=1

α0 
≤ M exp −γ log λ−1

i −a0 h(s)

α
≤ −a0 h(s) 0

and


f h(s) exp{−γs} ≤ K1 exp (ϑ − γ)s , K1 > 0

95
 PETR KUNDRÁT

the estimate (7) can be rewritten as

|z(s∗ )|
 ∗


 Z )
h(s 

≤ Mk exp γ(sk − s∗ ) + a0 uα0 du

 

h(sk )
 ∗ 
 Z )
 h(s 

α0 ∗
+ Mk exp a0 u du − γs

 

s0
 
∗
Zs h  h(s)
Z 
i

α0 ′  
α0
× −a0 h(s) h (s) exp γs − a0 u du ds

 

sk s0
 ∗ 
 Z )
 h(s 

α0 ∗
+ K1 exp a0 u du − γs

 

s0
 
Zs
∗
 h(s)
Z 
 
a0 uα0 du h′ (s) exp (ϑ − γ)s ds.

× exp γs −

 

sk s0

Hence

|z(s∗ )|
 ∗


 Z )
h(s 

∗ α0
≤ Mk exp γ(sk − s ) + a0 u du

 

h(sk )
 ∗ 
 n o  Z )
 h(s 

α0 ∗
+ Mk + K2 exp (η + ϑ − γ)sk exp a0 u du − γs

 

s0
 
∗
Zs h  h(s)
Z 

α0 i ′  
α0
× −a0 h(s) h (s) exp γs − a0 u du ds,

 

sk s0


where K2 = K1 K.

96
 ASYMPTOTIC ESTIMATE FOR DIFFERENTIAL EQUATION

Since a0 is negative, the integral on the right-hand side can be estimated as

 
∗
Zs h  h(s)
Z 
i

α0 ′  
α0
−a0 h(s) h (s) exp γs − a0 u du ds

 

sk s0
  s∗
 h(s)
Z 
 
α0
≤ exp γs − a0 u du (1 + K3 e−ωsk ), K3 > 0, ω = 1 − η > 0.

 

s0
sk

Then we obtain

|z(s∗ )|
 ∗

 h(s
Z ) 
 
∗ α0
≤ Mk exp γ(sk − s ) + a0 u du

 

h(sk )
 ∗ 
 h(s
Z ) 
   
α0 ∗

+ Mk + K2 exp (η + ϑ − γ)sk exp a0 u du − γs

 

s0
  s∗
 h(s)
Z 
 
α0


× exp γs − a0 u du 1 + K3 exp{−ωsk }

 

s0
sk
   

≤ Mk 1 + M exp −ωsk + K2 exp (η + ϑ − γ)sk 1 + K3 exp{−ωsk }

≤ Mk∗ 1 + N exp{−κsk } ,

where

Mk∗ = max (Mk , K2 ) and κ = min (ω, γ − η − ϑ) > 0 and N > 0

is a constant large enough. Since s∗ ∈ Jk+1 was arbitrary,

k
Y
∗
≤ Mk∗ 1 + N exp{−κsk } ≤ M1∗



Mk+1 1 + N exp{−κsj } .
j=1

The boundedness of the sequence (Mk∗ ) as k → ∞ gives the asymptotic estimate

(4) with respect to the substitution (5). 

97
 PETR KUNDRÁT

4. Examples and final remarks

First we introduce such a case of differential equation (1) that there are con-
sidered constant coefficients and one delayed term in the equation. Furthermore,
the forcing term is omitted in this case:

Example 1. Consider the equation

ẏ(t) = a0 y(t) + a1 y(tλ1 ), t ∈ [1, ∞), (8)

where a0 < 0, a1 are real constants, 0 < λ1 < 1. Then Theorem 1 gives the
asymptotic estimate
|a1 |
γ
log −a
y(t) = O(log t), as t → ∞, γ= 0
.
log λ−1
1

The above asymptotic estimate coincides with the result of H e a r d [4], which
can be reformulated for equation (8) as follows.

Theorem 2 (Heard, 1975). Let a0 < 0, a1 6= 0 be real constants, 0 < λ1 < 1.

Then for every solution of (8) there exists a continuous periodic function g with
period log λ−1
1 such that

y(t) = (log t)ξ g(log log t) + O logξr −1 t

as t → ∞,

where ξ is a root of a0 + a1 λξ1 = 0 and ξr = ℜ(ξ).

The next example illustrates the application of the Theorem 1 to the equa-
tion (1) with nonconstant coefficients at delayed terms.

Example 2. Consider the equation

1h
i
ẏ(t) = a0 y(t) + a1 y tλ1 + a2 y tλ2 + f (t),


(9)
t
where a0 < 0, a1 , a2 are real constants, 0 < λ1 < λ2 < 1 and f (t) = O(logν t) as
t → ∞ for a suitable real ν. Then in accordance with Theorem 1 we obtain the
asymptotic estimate of solution of (9) in the form

y(t) = O(logγ t), as t → ∞,

where
log |a1−a
|+|a2 |
log |a1−a
|+|a2 |
( )
γ = max ν, 0
, 0
.
log λ−1
1 log λ −1
2

98
 ASYMPTOTIC ESTIMATE FOR DIFFERENTIAL EQUATION

REFERENCES

[1] ČERMÁK, J.—KUNDRÁT, P.: Linear differential equations with unbounded delays and
a forcing term, Abstr. Appl. Anal. 2004 (2004), 337–345.
[2] ČERMÁK, J.: On the differential equation with power coefficients and proportional de-
lays, Tatra Mt. Math. Publ. 38 (2007), 57–69.
[3] DERFEL, G.—VOGL, F.: On the asymptotics of solutions of a class of linear functional-
-differential equations, European J. Appl. Math. 7 (1996), 511–518.
[4] HEARD, M. L.: A change of variables for functional differential equations, J. Differential
Equations 18 (1975), 1–10.
[5] ISERLES, A.: On the generalized pantograph functional–differential equation, European
J. Appl. Math. 4 (1993), 1–38.
[6] ISERLES, A.: Exact and discretized stability of the pantograph equation, Appl. Numer.
Math. 24 (1997), 295–308.
[7] LIU, Y.: Asymptotic behaviour of functional–differential equations with proportional time
delays, European J. Appl. Math. 7 (1996), 11–30.
[8] MAKAY, G.—TERJÉKI, J.: On the asymptotic behaviour of the pantograph equations,
Electron. J. Qual. Theory Differ. Equ. 2 (1998), 1–12.
[9] NEUMAN, F.: Simultaneous solutions of a system of Abel equations and differential
equations with several deviations, Czechoslovak Math. J. 32(107) (1982), 488–494.

Received November 15, 2010 Institute of Mathematics FME BUT

Technická 2
CZ–616-69 Brno
CZECH REPUBLIC
E-mail: kundrat@[Link]

99