ISSN 0012-2661, Differential Equations, 2012, Vol. 48, No. 5, pp. 630–646.


c Pleiades Publishing, Ltd., 2012.
Original Russian Text 
c V.M. Evtukhov, E.S. Vladova, 2012, published in Differentsial’nye Uravneniya, 2012, Vol. 48, No. 5, pp. 622–639.

ORDINARY DIFFERENTIAL EQUATIONS

Asymptotic Representations of Solutions

of Essentially Nonlinear Cyclic Systems
of Ordinary Differential Equations
V. M. Evtukhov and E. S. Vladova
Odessa National University, Odessa, Ukraine
Received May 20, 2010

Abstract—We obtain asymptotic representations for a class of solutions of cyclic systems of

ordinary differential equations with properly varying singularities.
DOI: 10.1134/S0012266112050023

1. STATEMENT OF THE PROBLEM AND AUXILIARY NOTATION

We consider the system of differential equations1
yi = αi pi (t)ϕi+1 (yi+1 ) (i = 1, . . . , n), (1.1)
where αi ∈ {−1, 1} (i = 1, . . . , n), the pi : [a, ω[ → ]0, +∞[ (i = 1, . . . , n) are continuous functions,
−∞ < a < ω ≤ +∞ (if ω = +∞, then a > 0), the ϕi : Δ(Yi0 ) → ]0, n+∞[ (i = 1, . . . , n) are
normalized functions properly varying as yi → Yi0 of orders σi such that i=1 σi = 1, where Δ(Yi0 )
is some one-sided neighborhood of the point Yi0 and Yi0 is equal to either zero or ±∞.
If ϕi (yi ) = |yi |σi (i = 1, . . . , n), then such a system of equations is called a system of Emden–
Fowler type. Asymptotic representations of its nonoscillating solutions as t ↑ ω were obtained
in [1–5] for n = 2.
In the present paper, we drop the assumption that ϕi (yi ) (i = 1, . . . , n) are power-law functions
and assume that they are close to power-law functions in neighborhoods of the points Yi0 in the
sense of the definition of normalized properly varying functions [6, pp. 9–11, 31].
A solution (yi )ni=1 of system (1.1) defined on an interval [t0 , ω[ ⊂ [a, ω[ is called a
Pω (Λ1 , . . . , Λn−1 )-solution if it satisfies the conditions

yi (t) ∈ Δ(Yi0 ) for t ∈ [t0 , ω[, lim yi (t) = Yi0 ,

t↑ω

yi (t)yi+1 (t) (1.2)
lim  = Λi (i = 1, . . . , n − 1).
t↑ω yi (t)yi+1 (t)

The aim of the present paper is to derive necessary and sufficient conditions for the existence of
Pω (Λ1 , . . . , Λn−1 )-solutions of system (1.1) and asymptotic formulas for such solutions as t ↑ ω for
the case in which the Λi (i = 1, . . . , n − 1) are nonzero real constants.
Let us introduce notation needed in forthcoming considerations.
First, we set μi = 1 if either Yi0 = +∞ or Yi0 = 0 and Δ(Yi0 ) is a right neighborhood of the point
0 and μi = −1 if either Yi0 = −∞ or Yi0 = 0 and Δ(Yi0 ) is a left neighborhood of the point 0; note
that the numbers μi (i = 1, . . . , n) determine the signs of the components of a Pω (Λ1 , . . . , Λn−1 )-
solution in some left neighborhood of ω.
1
Here and in the following, all functions and parameters with subscript n + 1 are assumed to coincide with those
with subscript 1.

630
 ASYMPTOTIC REPRESENTATIONS OF SOLUTIONS OF ESSENTIALLY NONLINEAR . . . 631

Next, we note that the definition of a Pω (Λ1 , . . . , Λn−1 )-solution does not provide a straightfor-
ward relationship between the first and nth components of the solution, which occur in the nth
equation of the system. But if Λi ∈ R\{0} (i = 1, . . . , n − 1), then, by virtue of (1.2), we have
  

yn (t)y1 (t) yn (t)yn−1 (t) yn−1 (t)yn−2 (t) y2 (t)y1 (t) 1
Λn = lim  = lim  · · · = .
t↑ω yn (t)y1 (t) 
t↑ω yn (t)yn−1 (t) yn−1 (t)yn−2 (t) y2 (t)y1 (t) Λ1 · · · Λn−1
n n
Hence it follows that i=1 Λi = 1; therefore, by the condition i=1 σi = 1, the expression 1−Λi σi+1
is nonzero for at least one i ∈ {1, . . . , n}. Let

I = {i ∈ {1, . . . , n} : 1 − Λi σi+1 = 0}, Ī = {1, . . . , n}\I,

and let l be the minimal element of the set I.

Taking into account the choice of l, we introduce auxiliary functions Ii (i = 1, . . . , n) and nonzero
constants βi (i = 1, . . . , n) by the relations
⎧ ⎧
⎪ t ⎪
⎪ 1 − Λi σi+1 for i ∈ I
⎪
⎪ ∈ ⎪
⎪ β
⎨ p i (τ ) dτ for i I ⎨ l
for i ∈ {l + 1, . . . , n}\I
Ii (t) = A ti βi = Λl · · · Λi−1
⎪
⎪ ⎪
⎪
⎪
⎩ Il (τ )pi (τ ) dτ for i ∈ Ī, ⎪
⎪ βl
⎩ for i ∈ {1, . . . , l − 1}\I,
Ai Λl · · · Λn Λ1 · · · Λi−1

where each of the integration limits Ai ∈ {ω, a} is chosen so as to ensure that the corresponding
integral Ii tends either to zero or to infinity as t ↑ ω.
In addition, we introduce the numbers

1 if Ai = a
A∗i = (i = 1, . . . , n), (1.3)
−1 if Ai = ω,
which permit one to determine the signs of the functions Ii (i = 1, . . . , n) on the interval ]a, ω[.
Note also that, by virtue of the properties of normalized properly varying functions [6, pp. 9–11,
31], we have
zϕi (z)
lim = σi , ϕi (z) = |z|σi θi (z) (i = 1, . . . , n), (1.4)
z→Yi
0
ϕi (z)
z∈Δ(Yi )

where θi (z) is a function slowly varying as z → Yi0 .

2. MAIN RESULTS

Theorem 2.1. Let Λi ∈ R\{0} (i = 1, . . . , n − 1) and l = min I. Then, for the existence of
Pω (Λ1 , . . . , Λn−1 )-solutions of system (1.1), it is necessary and, if the algebraic equation
 i−1  i−1
n  n 
Λj + ν − σi Λj = 0 (2.1)
i=1 j=1 i=1 j=1

has no roots with zero real part, sufficient that the relation

Ii (t)Ii+1 (t) βi+1
lim 
= Λi (2.2)
t↑ω Ii (t)Ii+1 (t) βi
and the sign conditions

A∗i βi > 0 if Yi0 = ±∞, A∗i βi < 0 if Yi0 = 0, (2.3)

sgn[αi A∗i βi ] = μi (2.4)

DIFFERENTIAL EQUATIONS Vol. 48 No. 5 2012

632 EVTUKHOV, VLADOVA

hold for each i ∈ {1, . . . , n}. Moreover , each solution of this kind admits the asymptotic represen-
tations
yi (t)
= αi βi Ii (t)[1 + o(1)] for i ∈ I, (2.5)
ϕi+1 (yi+1 (t))
yi (t) Ii (t)
= αi βi [1 + o(1)] for i ∈ Ī (2.6l )
ϕi+1 (yi+1 (t)) Il (t)
as t ↑ ω; moreover , there even exists a k-parameter family of such solutions provided that the
algebraic equation (2.1) has k roots (with regard of multiplicities) whose real parts have sign opposite
to the sign of the number A∗l βl .
n
Remark 2.1. Note that Eq. (2.1) necessarily has no roots with zero real part if i=1 |σi | < 1.
Proof of Theorem 2.1. Without loss of generality, we assume that l = 1. The general case
can be reduced to this one by rearranging all functions, variables, and constants as follows (we show
the change of subscripts) :
l → 1, ..., n → n − l + 1, 1 → n − l + 2, ..., l − 1 → n.
If l = 1, then the formulas defining the nonzero constants βi (i = 1, . . . , n) and the functions Ii
(i = 1, . . . , n) acquire the form
⎧
⎧ ⎪
⎪ t
⎨ 1 − Λ σ
i i+1 for i ∈ I ⎪
⎨ pi (τ ) dτ for i ∈ I
Ai
βi = β1 Ii (t) = t
⎩ for i ∈ Ī, ⎪
⎪
Λ1 · · · Λi−1 ⎪
⎩ I1 (τ )pi (τ ) dτ for i ∈ Ī.
Ai

By taking into account definition (1.3) and the choice of the integration limits Ai (i = 1, . . . , n),
we note that the relation  ∗
Ai for i ∈ I
sgn Ii (t) = (2.7)
A∗i A∗1 for i ∈ Ī
holds for t ∈ ]a, ω[.
Necessity. Let yi : [t0 , ω[ → Δ(Yi0 ) (i = 1, . . . , n) be an arbitrary Pω (Λ1 , . . . , Λn−1 )-solution of
system (1.1). Then, by virtue of (1.1), we have
yi (t)
= αi pi (t) (i = 1, . . . , n) for t ∈ [t0 , ω[. (2.8)
ϕi+1 (yi+1 (t))
By integrating each of these relations for i ∈ I from Bi to t, where Bi = ω if Ai = ω and Bi = t0 if
Ai = a, we obtain the relation
t
yi (τ ) dτ
= αi Ii (t)[1 + o(1)] as t ↑ ω. (2.9)
ϕi+1 (yi+1 (τ ))
Bi

yi (t)
Let us compare the expression with the integral occurring on the left-hand side in
ϕi+1 (yi+1 (t))
relation (2.9). By the l’Hospital rule in the Stolz form, we have
 t
yi (t) yi (τ ) dτ
lim
t↑ω ϕi+1 (yi+1 (t)) ϕi+1 (yi+1 (τ ))
Bi
  
yi (t) yi (t)ϕi+1 (yi+1 (t))yi+1
(t) ϕi+1 (yi+1 (t))
= lim −
t↑ω ϕi+1 (yi+1 (t)) ϕ2i+1 (yi+1 (t)) yi (t)
yi+1 (t)ϕi+1 (yi+1 (t)) 
yi (t)yi+1 (t)
= 1 − lim lim  = 1 − Λi σi+1 = βi = 0 for i ∈ I.
t↑ω ϕi+1 (yi+1 (t)) t↑ω yi (t)yi+1 (t)

DIFFERENTIAL EQUATIONS Vol. 48 No. 5 2012

 ASYMPTOTIC REPRESENTATIONS OF SOLUTIONS OF ESSENTIALLY NONLINEAR . . . 633

By virtue of this limit relation, from (2.9), we obtain the asymptotic representations (2.5). In ad-
dition, they, together with (2.8), imply that

yi (t) I  (t)

= i [1 + o(1)] as t ↑ ω. (2.10)
yi (t) βi Ii (t)

Let us multiply each relation in (2.8), where i ∈ Ī, by I1 (t) and integrate the resulting relation
from Bi to t, where Bi is chosen just as above. As a result, we obtain the relation

t
yi (τ )I1 (τ ) dτ
= αi Ii (t)[1 + o(1)] (i ∈ Ī) as t ↑ ω. (2.11)
ϕi+1 (yi+1 (τ ))
Bi

By virtue of the l’Hospital rule in the Stolz form, we have

 t
yi (t)I1 (t) yi (τ )I1 (τ ) dτ
lim
t↑ω ϕi+1 (yi+1 (t)) ϕi+1 (yi+1 (τ ))
Ai

yi (t)I1 (t) yi+1 (t)ϕi+1 (yi+1 (t)) 

yi (t)yi+1 (t) yi (t)y1 (t)
= 1 + lim  − lim lim  = 1 − Λi σi+1 + β1 lim 
t↑ω yi (t)I1 (t) t↑ω ϕi+1 (yi+1 (t)) t↑ω yi (t)yi+1 (t) t↑ω yi (t)y1 (t)
  

yi (t)yi−1 (t) y2 (t)y1 (t) β1
= β1 lim  ···  = = βi = 0 for i ∈ Ī.
t↑ω yi (t)yi−1 (t) y2 (t)y1 (t) Λ1 · · · Λi−1

This, together with (2.11), implies that the asymptotic relations (2.61 ) hold; then, by virtue of (2.8),
the asymptotic representations (2.10) take place for i ∈ Ī as well.
Since relations (2.10) hold for i = 1, . . . , n, and the solution in question satisfies the last limit
relation in the definition of a Pω (Λ1 , . . . , Λn−1 )-solution, it follows that condition (2.2) is satisfied
for each i ∈ {1, . . . , n}. In addition, it follows from (2.10) that

|yi (t)| = |Ii (t)|1/βi +o(1) (i = 1, . . . , n) as t ↑ ω,

which, together with the condition limt↑ω yi (t) = Yi0 , the definition of a Pω (Λ1 , . . . , Λn−1 )-solution,
and the definition of the number A∗i , implies the sign conditions (2.3).
The validity of the sign condition (2.4) readily follows from the representations (2.5) and (2.61 )
if we take into account the signs of the functions yi and Ii (i = 1, . . . , n) on the interval [t0 , ω[.
Sufficiency. Suppose that Λi ∈ R\{0}, 1 − Λ1 σ2 = 0 (l = 1), and, in addition to condi-
tions (2.2)–(2.4), the algebraic equation (2.1) has no roots with zero real part. Let us show that,
in this case, system (1.1) has at least one Pω (Λ1 , . . . , Λn−1 )-solution that admits the asymptotic
representations (2.5) and t ↑ ω as (2.61 ), and consider the problem on evaluating the number of
such solutions.
First, considering the system of relations of the form
yi
= Qi (t)[1 + vi ] (i = 1, . . . , n), (2.12)
ϕi+1 (yi+1 )

where 
αi βi Ii (t) for i ∈ I
Qi (t) =
αi βi Ii (t)/I1 (t) for i ∈ I,
we show that it uniquely determines continuously differentiable implicit functions yi = Yi (t, v̄)
(i = 1, . . . , n) of the form

Yi (t, v̄) = μi |I1 (t)|(1/β1 )Li +zi (t,v̄) (i = 1, . . . , n) (2.13)

DIFFERENTIAL EQUATIONS Vol. 48 No. 5 2012

634 EVTUKHOV, VLADOVA

defined on the set D0 = [t0 , ω[ × V0 , t0 ∈ [a, ω[, V0 = {v̄ ≡ (v1 , . . . , vn ) : |vi | ≤ 1/2 (i = 1, . . . , n)};
i−1
here L1 = 1, Li = j=1 Λj (i = 2, . . . , n), and the functions zi (i = 1, . . . , n) satisfy the conditions

1
|zi (t, v̄)| ≤ |Li | (i = 1, . . . , n) for (t, v̄) ∈ D0
2|β1 |
and limt↑ω zi (t, v̄) = 0 uniformly with respect to v̄ ∈ V0 .
To this end, by setting

yi = μi |I1 (t)|(1/β1 )Li +zi (i = 1, . . . , n), (2.14)

in (2.12) and by taking into account (1.4), we obtain a system of relations of the form

|I1 (t)|(1/β1 )Li +zi −(σi+1 /β1 )Li+1 −σi+1 zi+1

= |Qi (t)| θi+1 (μi+1 |I1 (t)|(1/β1 )Li+1 +zi+1 )(1 + vi ) (i = 1, . . . , n). (2.15)

By condition (2.2), we have

βi βi
ln |Ii (t)| ∼ Λi−1 ln |Ii−1 (t)| ∼ · · · ∼ Li ln |I1 (t)| (i = 2, . . . , n) as t ↑ ω,
βi−1 β1
and therefore,
|Ii (t)| = |I1 (t)|(βi /β1 )Li +ui (t) (i = 1, . . . , n), (2.16)
where the ui : ]a, ω[ → R (i = 2, . . . , n) are continuous functions tending to zero as t ↑ ω. This,
together with condition (2.3), implies that limt↑ω μi |I1 (t)|(1/β1 )Li = Yi0 (i = 1, . . . , n). Consequently,
the system of relations (2.15) is defined on the set Ω0 = [t1 , ω[ × Z0 × V0 , where t1 is some number
in the interval [a, ω[ and Z0 = {z̄ ≡ (z1 , . . . , zn ) : |zi | ≤ (1/(2|β1 |))|Li |, i = 1, . . . , n}.
By using relations (2.16), from (2.15), we obtain

ln[|βi | θi+1 (μi+1 |I1 (t)|(1/β1 )Li+1 +zi+1 )(1 + vi )]

zi − σi+1 zi+1 = ui (t) + (i = 1, . . . , n).
ln |I1 (t)|
By solving this system partly for z1 , . . . , zn (as a linear inhomogeneous system), we obtain
zi = ai (t) + bi (t, v̄) + Zi (t, z̄) (i = 1, . . . , n), (2.17)
where2
 −1    

n 
n 
k 
n 
i−1 
k
ai (t) = 1− σj uk (t) σj + σj uk (t) σj ,
j=1 k=i j=i+1 j=i+1 k=1 j=1
 −1  

n
−1

n 
k
bi (t, v̄) = 1− σj ln |I1 (t)| ln(|βk |(1 + vk )) σj
j=1 k=i j=i+1
 

n 
i−1 
k
+ σj ln(|βk |(1 + vk )) σj ,
j=i+1 k=1 j=1
 −1  

n
−1

n 
k
Zi (t, z̄) = 1− σj ln |I1 (t)| ln θk+1 (μk+1 |I1 (t)| (1/β1 )Lk+1 +zk+1
) σj
j=1 k=i j=i+1
 

n 
i−1 
k
+ σj ln θk+1 (μk+1 |I1 (t)|(1/β1 )Lk+1 +zk+1 ) σj (i = 1, . . . , n).
j=i+1 k=1 j=1

2
i i
Here and in the following, we assume that j=i+1 = 1 and j=i+1 = 0.

DIFFERENTIAL EQUATIONS Vol. 48 No. 5 2012

 ASYMPTOTIC REPRESENTATIONS OF SOLUTIONS OF ESSENTIALLY NONLINEAR . . . 635

Here
lim bi (t, v̄) = 0 (i = 1, . . . , n) uniformly with respect to v̄ ∈ V0 (2.18)
t↑ω

and, by virtue of the properties of slowly varying functions (see [6, pp. 9–15, 23–25]), we have the
relations
lim Zi (t, z̄) = 0 (i = 1, . . . , n) uniformly with respect to z̄ ∈ Z0 . (2.19)
t↑ω

Since the functions ui (t) (i = 1, . . . , n) tend to zero as t ↑ ω, it follows that

lim ai (t) = 0 (i = 1, . . . , n). (2.20)
t↑ω

In addition, we have

∂Zi (t, z1 , . . . , zn ) μm |I1 (t)|(1/β1 )Lm +zm θm (μm |I1 (t)|(1/β1 )Lm +zm )
= im (i, m = 1, . . . , n),
∂zm θm (μm |I1 (t)|(1/β1 )Lm +zm )
where
⎧ −1 
⎪ 
⎨ 1 − n σj n
for 1 ≤ i ≤ n, 1 ≤ m ≤ i
j=1 j=i+1 σj
im =   −1 n m−1
⎪
⎩ 1 − nj=1 σj for 1 ≤ i ≤ n − 1, i + 1 ≤ m ≤ n.
j=i+1 σj j=1 σj

This, together with the conditions limz→Yi0 (zθi (z)/θi (z)) = 0 (i = 1, . . . , n), which are satisfied
by the slowly varying functions θi , (i = 1, . . . , n), implies that limt↑ω (∂Zi (t, z1 , . . . , zn )/∂zm ) = 0
(i, m = 1, . . . , n) uniformly with respect to z̄ ∈ Z0 .
By virtue of the above-represented limit relations, there exists a number t0 ∈ [t1 , ω[ such that
the inequalities
β0 1
|ai (t) + bi (t, v̄) + Zi (t, z̄)| ≤ (i = 1, . . . , n), β0 = min{|L1 |, . . . , |Ln |}, (2.21)
n 2|β1 |
and the Lipschitz conditions
1  1
n
|Zi (t, z̄ 1 ) − Zi (t, z̄ 2 )| ≤ |z − zk2 | (i = 1, . . . , n) (2.22)
n + 1 k=1 k

hold on the set [t0 , ω[ × Z0 × V0 for t ∈ [t0 , ω[ and for arbitrary z̄ 1 , z̄ 2 ∈ Z0 .

Having chosen the number t0 this way, by B we denote the Banach space of continuous vector
functions z = (zi )ni=1 : Ω → Rn bounded on the set Ω = [t0 , ω[ × V0 and equipped with the norm
 n 

z = sup |zi (t, v̄)| : (t, v̄) ∈ Ω .
i=1

In this space, we single out the subspace B0 of functions in B for which z ≤ β0 , and, arbitrarily
choosing a number ν ∈ (0, 1), on B0 we consider the operator Φ = (Φi )ni=1 defined by the relations
Φi (z)(t, v̄) = zi (t, v̄)−ν[zi (t, v̄)−ai (t)−bi (t, v̄)−Zi (t, z1 (t, v̄), . . . , zn (t, v̄))] (i = 1, . . . , n). (2.23)
By virtue of conditions (2.21), the inequalities
|Φi (z)(t, v̄)| ≤ (1 − ν)|zi (t, v̄)| + νβ0 /n (i = 1, . . . , n) for (t, v̄) ∈ Ω
hold for each z ∈ B0 . Therefore, we have

n 
n
|Φi (z)(t, v̄)| ≤ (1 − ν) |zi (t, v̄)| + νβ0 ≤ (1 − ν)z + νβ0 ≤ (1 − ν)β0 + νβ0 = β0
i=1 i=1

on the set Ω. Hence it follows that Φ(z) ≤ β0 ; i.e., Φ(B0 ) ⊂ B0 .

DIFFERENTIAL EQUATIONS Vol. 48 No. 5 2012

636 EVTUKHOV, VLADOVA

Now let z, z̃ ∈ B0 . Then, by virtue of condition (2.22), for (t, v̄) ∈ Ω, we obtain the inequalities
|Φi (z)(t, v̄) − Φi (z̃)(t, v̄)|
≤ (1 − ν)|zi (t, v̄) − z̃i (t, v̄)| + ν|Zi (t, z1 (t, v̄), . . . , zn (t, v̄)) − Zi (t, z̃1 (t, v̄), . . . , z̃n (t, v̄))|
ν 
n
≤ (1 − ν)|zi (t, v̄) − z̃i (t, v̄)| + |zk (t, v̄) − z̃k (t, v̄)| (i = 1, . . . , n).
n + 1 k=1

Therefore, the estimate


n  
n  
ν ν
|Φk (z)(t, v̄) − Φk (z̃)(t, v̄)| ≤ 1− |zk (t, v̄) − z̃k (t, v̄)| ≤ 1− z − z̃
k=1
n+1 k=1
n+1

holds on the set Ω, which implies that

 
ν
Φ(z) − Φ(z̃) ≤ 1− z − z̃.
n+1
We have thereby shown that the operator Φ maps the space B0 into itself and is a contraction
operator on it. Then, by the contraction mapping principle, there exists a unique vector function
z ∈ B0 such that z = Φ(z). By virtue of relations (2.23), this continuous vector function defined on
the set Ω is the unique solution of system (2.17) satisfying the condition z ≤ β0 . Relation (2.17),
together with this condition and relations (2.18)–(2.20), implies that the components of this solution
tend to zero uniformly with respect to v̄ ∈ V0 as t ↑ ω. The continuous differentiability of this
solution on the set Ω follows directly from the well-known local theorem on the existence of implicit
functions defined by a system of relations. By the change of variables (2.14), the resulting vector
function z = (zi )ni=1 corresponds to a vector function (Yi )ni=1 with components of the form (2.13),
which is a solution of system (2.12).
Now, by applying the transformation
yi (t) = Yi (t, v1 (x), . . . , vn (x)) (i = 1, . . . , n), x = A∗1 ln |I1 (t)|, (2.24)
to the system of differential equations (1.1) and by using the fact that the vector function
(Yi (t, v1 (x), . . . , vn (x)))ni=1 , t ∈ [t0 , ω[, (v1 (x), . . . , vn (x)) ∈ V0 , is a solution of the system of equa-
tions
yi (t)
= Qi (t)[1 + vi (x)] (i = 1, . . . , n), (2.25)
ϕi+1 (yi+1 (t))
we obtain the system of differential equations of the form
 
 A∗1 1 + vi
vi = hi (x) − hi+1 (x)ξi+1 (x, v1 , . . . , vn ) − gi (x)(1 + vi ) (i = 1, . . . , n), (2.26)
β1 1 + vi+1
where
β1 Ii (t)I1 (t)
hi (x) = hi (x(t)) = (i = 1, . . . , n),
βi Ii (t)I1 (t)
⎧
⎪
⎪ Ii (t)I1 (t)
⎪ β
⎨ Ii (t)I  (t)
1 for i ∈ I
1
gi (x(t)) =    (i = 1, . . . , n),
⎪
⎪ Ii (t)I1 (t)
⎪
⎩ β1 −1 for i ∈ I, ¯
Ii (t)I1 (t)
Yi (t, v1 , . . . , vn )ϕi (Yi (t, v1 , . . . , vn ))
ξi (x, v1 , . . . , vn ) = ξ(x(t), v1 , . . . , vn ) = (i = 1, . . . , n).
ϕi (Yi (t, v1 , . . . , vn ))
Here, by virtue of conditions (2.2),
lim hi (x) = lim hi (x(t)) = Li , lim gi (x) = lim gi (x(t)) = (1 − Λi σi+1 )Li (2.27)
x→+∞ t↑ω x→+∞ t↑ω

DIFFERENTIAL EQUATIONS Vol. 48 No. 5 2012

 ASYMPTOTIC REPRESENTATIONS OF SOLUTIONS OF ESSENTIALLY NONLINEAR . . . 637

for i = 1, . . . , n. Since, by virtue of (2.3) and (2.13), the relations limt↑ω Yi (t, v1 , . . . , vn ) = Yi0
(i = 1, . . . , n) hold uniformly with respect to (v1 , . . . , vn ) ∈ V0 and the first condition in (1.4) is
satisfied, it follows that, in addition, the representation ξi (x, v1 , . . . , vn ) = σi + Ri1 (x, v1 , . . . , vn )
(i = 1, . . . , n) holds, where

Ri (x, v1 , . . . , vn ) → 0 uniformly with respect to (v1 , . . . , vn ) ∈ V0 as x → +∞. (2.28)

By taking into account these relations and the representations

(1 + vi )/(1 + vi+1 ) = 1 + vi − vi+1 + Ri2 (v1 , . . . , vn ) (i = 1, . . . , n),

where the Ri2 (i = 1, . . . , n) are functions such that

∂Ri2 (v1 , . . . , vn )
lim =0 (i, k = 1, . . . , n), (2.29)
|v1 |+···+|vn |→0 ∂vk

we rewrite the system of differential equations (2.26) in the form

A∗1
vi = [fi (x) + pii (x)vi + pii+1 (x)vi+1 + Vi1 (x, v1 , . . . , vn ) + Vi2 (x, v1 , . . . , vn )], i = 1, . . . , n,
β1
(2.30)
where

fi (x) = hi (x) − σi+1 hi+1 (x) − gi (x), pii (x) = −σi+1 hi+1 (x) − gi (x), pii+1 (x) = σi+1 hi+1 (x),
1 + vi
Vi1 (x, v1 , . . . , vn ) = −hi+1 (x)Ri1 (x, v1 , . . . , vn ) ,
1 + vi+1
Vi2 (x, v1 , . . . , vn ) = −hi+1 (x)ξi (x, v1 , . . . , vn )Ri2 (x, v1 , . . . , vn ), i = 1, . . . , n.

By virtue of conditions (2.24)–(2.26), in this system, we have the relations

lim fi (x) = 0 (i = 1, . . . , n),

x→+∞

lim Vi1 (x, v1 , . . . , vn ) = 0 (i = 1, . . . , n) uniformly with respect to (v1 , . . . , vn ) ∈ V0 ,

x→+∞

Vi2 (x, v1 , . . . , vn )
lim = 0 (i = 1, . . . , n) uniformly with respect to t ∈ [t0 , ω[,
|v1 |+···+|vn |→0 |v1 | + · · · + |vn |

and the matrix P (x) = (pij (x))ni,j=1 of coefficients multiplying vk (k = 1, . . . , n) occurring in the
brackets on the right-hand sides satisfies the condition
⎛ ⎞
−L1 σ2 L2 0 ... 0 0
⎜ ⎟
⎜ 0 −L2 σ3 L3 . . . 0 0 ⎟
⎜ ⎟
P0 = lim P (x) = ⎜
⎜ ··· ··· ··· ··· ··· ··· ⎟⎟.
x→+∞
⎜ ⎟
⎝ 0 0 0 . . . −Ln−1 σn Ln ⎠
σ1 L1 0 0 ... 0 −Ln

In addition, note that the characteristic equation det[P0 − νEn ] = 0, where En is the n × n
identity matrix, can be represented in the form (2.1) and hence has no roots with zero real part.
Consequently, system (2.30) satisfies all assumptions of Theorem 2.2 in [7]. This theorem implies
that system (2.30) has at least one solution {vi }ni=1 : [x1 , +∞[ → Rn (x1 ≥ x0 = A∗1 ln |I1 (t0 )|) that
tends to zero as x → +∞; moreover, such solutions form a k-parameter family provided that the
characteristic equation (2.1) has k roots (with regard of multiplicities) whose real parts have sign
opposite to that of A∗1 β1 . By the change of variables (2.24) and system (2.25), which holds for

DIFFERENTIAL EQUATIONS Vol. 48 No. 5 2012

638 EVTUKHOV, VLADOVA

the functions Yi (t, v1 (x(t)), . . . , vn (x(t))) (i = 1, . . . , n), all solutions of this kind of system (2.30)
corresponds to solutions (y1 , . . . , yn ) of system (1.1) admitting the asymptotic representations

yi (t)
= Qi (t)[1 + o(1)] (i = 1, . . . , n) as t ↑ ω.
ϕi+1 (yi+1 (t))

It remains to show that each of the above-mentioned solutions of system (1.1) is a

Pω (Λ1 , . . . , Λn−1 )-solution.
Since each of them corresponds to a solution (v1 (x), . . . , vn (x)) of system (2.30) tending to
zero as x → +∞, it follows from the above-proved properties of the functions Yi (t, v1 , . . . , vn )
(i = 1, . . . , n) that the first two conditions in (1.2) are necessarily satisfied. In addition, by virtue
of relations (2.25) and (2.2), for these solutions of system (1.1), we have
 
yi+1 (t)yi (t) βi Ii+1 (t)Ii (t)
lim 
= lim = Λi .
t↑ω yi+1 (t)yi (t) βi+1 t↑ω Ii+1 (t)Ii (t)

Therefore, the third condition (1.2) in the definition of a Pω (Y10 , . . . , Yn0 , λ1 , . . . , λn )-solution is
satisfied as well. The proof of the theorem is complete.
Now let us write out conditions under which the asymptotic representations (2.5) and (2.6l ) can
be written out in closed form.
Definition 2.1. We say that a function ϕi (i ∈ {1, . . . , n}) satisfies Condition S if the asymp-
totic relation
θi (zl(z)) = θi (z)[1 + o(1)] as z → Yi0 (z ∈ Δ(Yi0 )) (2.31)
holds for any continuously differentiable function l : Δ(Yi0 ) → ]0, +∞[ such that

z l (z)
lim0 = 0.
z→Yi l(z)
z∈Δ(Yi0 )

Condition S is satisfied for functions ϕi (i ∈ {1, . . . , n} for which the function θi has a finite
limit as z → Yi0 as well as for functions of the form

ϕi (z) = |z|σ | ln z|γ1 , ϕi (z) = |z|σ | ln z|γ1 | ln | ln z||γ2 ,

where γ1 , γ2 = 0, and for many other functions.

Remark 2.2. If a function ϕi (i ∈ {1, . . . , n}) satisfies Condition S and yi : [t0 , ω[ → Δ(Yi0 ) is
a continuously differentiable function such that

yi (t) ξ  (t)

lim yi (t) = Yi0 , = [r + o(1)] as t ↑ ω,
t↑ω yi (t) ξ(t)

where r is a nonzero real constant and ξ is a real function continuously differentiable in some left
neighborhood of ω and such that ξ  (t) = 0, then

θi (yi (t)) = θi (μi |ξ(t)|r )[1 + o(1)] as t ↑ ω,

because in this case, we have

yi (t) = z(t)l(z(t)), z(t) = μ0 |ξ(t)|r ,

and  
z l (z) z(t) l (z(t)) z(t)(yi (t)/z(t)) ξ(t)yi (t)
lim = lim = lim = lim − 1 = 0.
z→Y0 l(z) t↑ω l(z(t)) t↑ω (yi (t)/z(t))z  (t) t↑ω rξ  (t)yi (t)
z∈ΔY0

DIFFERENTIAL EQUATIONS Vol. 48 No. 5 2012

 ASYMPTOTIC REPRESENTATIONS OF SOLUTIONS OF ESSENTIALLY NONLINEAR . . . 639

Theorem 2.2. Let Λi ∈ R\{0} (i = 1, . . . , n − 1), l = min I, and let all the functions ϕi
(i = 1, . . . , n) satisfy Condition S. Then each Pω (Λ1 , . . . , Λn−1 )-solution (if any) of system (1.1)
admits the asymptotic representations

n
yi (t) = μi |Qk (t)θk+1 (μk+1 |Ik+1 (t)|1/βk+1 )|ik [1 + o(1)] (i = 1, . . . , n) (2.32)
k=1

as t ↑ ω, where
⎧
⎨ αk βk Ik (t) for k ∈ I
Qk (t) = Ik (t)
⎩ αk βk for k ∈ I,
Il (t)
⎧  
⎪ n k n (2.33)
⎪
⎨ j=i+1 σj j=1 σj 1 − j=1 σj for k = 1, . . . , i − 1
ik =   
⎪
⎪ k n
⎩ j=i+1 σj 1 − j=1 σj for k = i, . . . , n.

Proof. In the proof of Theorem 2.1, we have shown that, for the existence of Pω (Λ1 , . . . , Λn−1 )-
solutions of the system of differential equations (1.1), it is necessary that conditions (2.2)–(2.4)
be satisfied, and any solution of that kind admits the asymptotic representations (2.5) and (2.6l )
as t ↑ ω. In addition, the asymptotic relation (2.10) has been obtained for such solutions. By virtue
of this relation and Remark 2.2, we have θi (yi (t)) = θi (μi |Ii (t)|1/βi )[1 + o(1)] (i = 1, . . . , n) as t ↑ ω.
Therefore, the asymptotic representations (2.5) and (2.6l ) can be rewritten in the form
yi (t)
= Qi (t)θi+1 (μi+1 |Ii+1 (t)|1/βi+1 )[1 + o(1)] (i = 1, . . . , n) as t ↑ ω.
|yi+1 (t)|σi+1
By solving this system of algebraic equations for y1 , . . . , yn , we obtain the asymptotic representa-
tions (2.32). The proof of the theorem is complete.

3. APPLICATION OF THE MAIN RESULTS

First, as an example illustrating our results, consider the system of differential equations of the
form
yi = αi pi (t)|yi+1 |σi+1 | ln |yi+1 ||γi+1 (i = 1, . . . , n), (3.1)
where αi , σi , and pi (i = 1, . . . , n) are the same as in system (1.1) and γi ∈ R (i = 1, . . . , n). Here
ϕi (yi ) = |yi |σi | ln |yi ||γi (i = 1, . . . , n). They are properly varying functions of orders σi (i = 1, . . . , n)
both as yi → 0 and as yi → ±∞. Note that in this case each of the Yi0 (i ∈ {1, . . . , n}) can be either
zero, or +∞, or −∞. Moreover, for each of the above-mentioned values of Yi0 (i ∈ {1, . . . , n}), the
function ϕi : Δ(Yi0 ) → ]0, +∞[ satisfies condition S.
Therefore, Theorems 2.1 and 2.2 readily imply the following assertion.

Corollary 3.1. Let Λi ∈ R\{0} (i = 1, . . . , n − 1) and l = min I. Then for the existence of
Pω (Λ1 , . . . , Λn−1 )-solutions of system (3.1), it is necessary and, if the algebraic equation (2.1) has
no roots with zero real part, sufficient that conditions (2.2)–(2.4) be satisfied for each i ∈ {1, . . . , n};
moreover , each solution of this kind admits the asymptotic representations
n ! ! !γk+1 !ik
! ! 1 ! !
!Qk (t) ! ! !
yi (t) = μi ! ! βk+1 ln |Ik+1 (t)|! ! (i = 1, . . . , n) (3.2)
k=1

as t ↑ ω, where Qk and ik are given by formulas (2.33). Moreover , under the above-mentioned
conditions, there exists a k-parameter family of such solutions provided that the algebraic equa-
tion (2.1) has k roots (with regard of multiplicities) whose real parts have signs opposite to the sign
of A∗l βl .

DIFFERENTIAL EQUATIONS Vol. 48 No. 5 2012

640 EVTUKHOV, VLADOVA

Remark 3.1. The asymptotic representations (3.2) are new even in the case of a system of
Emden–Fowler type, i.e., for γ1 = · · · = γn = 0.
Next, consider the equation
u = α0 p(t)ϕ1 (u)ϕ2 (u ); (3.3)
here α0 ∈ {−1, 1}, p : [a, ω[ → ]0, +∞[ is a continuous function, and the ϕi : Δ(Ui0 ) → ]0, +∞[
(i = 1, 2) are continuously differentiable functions properly varying as z → Ui0 of orders σi such
that σ2 = 1 and σ1 + σ2 = 1, where Δ(Ui0 ) is some one-sided neighborhood of the point Ui0 = 0 or
Ui0 = ±∞.
A solution u of Eq. (3.3) is called a Pω (λ0 )-solution, where −∞ ≤ λ0 ≤ +∞, if it is defined on
some interval [t0 , ω[ ⊂ [a, ω[ and satisfies the condition

(u (t))2
u(i−1) (t) ∈ Δ(Ui0 ) for t ∈ [t0 , ω[, lim u(i−1) (t) = Ui0 (i = 1, 2), lim = λ0 .
t↑ω t↑ω u(t)u (t)

Let μ0i = 1 if either Ui0 = +∞ or Ui0 = 0 and Δ(Ui0 ) is a right neighborhood of the point 0 and
μ0i= −1 if either Ui0 = −∞ or Ui0 = 0 and Δ(Ui0 ) is a left neighborhood of the point 0 (i = 1, 2);
these numbers determine the signs of a Pω (λ0 )-solution and its derivative in some left neighborhood
of ω. In addition, we note that the conditions

μ01 μ02 < 0 if U10 = 0, μ01 μ02 > 0 if U10 = ±∞, (3.4)
α0 μ02 < 0 if U20 = 0, α0 μ02 > 0 if U20 = ±∞ (3.5)

are satisfied. They are necessary for the existence of Pω (λ0 )-solutions.
In addition, we introduce the function
z
ds
ψ(z) = ,
ϕ2 (s)
B

U0
where B = U20 if the integral b 2 ϕ−1 2 (s) ds is convergent and B = b if it is divergent, b being an
arbitrary number in the interval Δ(U20 ).
Since ψ  (z) > 0 for z ∈ Δ(U20 ), it follows that ψ : Δ(U20 ) → Δ(Y20 ) is an increasing function,
where Y20 = limz→U20 ψ(z); therefore, Y20 is either zero or ±∞, and Δ(Y20 ) is a one-sided neighbor-
hood of the point Y20 . In addition, by using the properties of properly varying functions and the
l’Hospital rule, we obtain the limit

zψ  (z) z (z/ϕ2 (z))

lim0 = lim0 = lim0 = 1 − σ2 . (3.6)
z→U2 ψ(z) z→U2 ψ(z)ϕ2 (z) z→U2 ψ  (z)

Consequently, ψ is a properly varying function of order 1 − σ2 as z → U20 .

Now, by using the transformation

u = y1 , ψ(u ) = y2 , (3.7)

we reduce Eq. (3.3) to the system of differential equations

y1 = ψ −1 (y2 ), y2 = α0 p(t)ϕ1 (y1 ). (3.8)

Here ψ −1 : Δ(Y20 ) → Δ(U20 ) is a properly varying function of order 1/(1 − σ2 ) as y2 → Y20 , because

y2 (ψ −1 (y2 )) y2 ϕ2 (ψ −1 (y2 )) ψ(z)ϕ2 (z) 1

lim 0 = lim = lim0 = ,
y2 →Y2 −1
ψ (y2 ) y2 →Y20 −1
ψ (y2 ) z→U2 z 1 − σ2
DIFFERENTIAL EQUATIONS Vol. 48 No. 5 2012
 ASYMPTOTIC REPRESENTATIONS OF SOLUTIONS OF ESSENTIALLY NONLINEAR . . . 641

and ϕ1 : Δ(Y10 ) → ]0, +∞[, where Y10 = U10 , Δ(Y10 ) = Δ(U10 ), is a properly varying function of
order σ1 as y1 → Y10 ; moreover, by virtue of the condition σ1 +σ2 = 1, the product of orders of these
functions is not unity. Therefore, system (3.8) is a system of the type (1.1) (for n = 2), where

α1 = sgn ψ −1 (y2 ) = μ02 , p1 (t) ≡ 1, α2 = α0 , p2 (t) = p(t). (3.9)

In addition, one can readily note that for λ0 ∈ R\{0} a solution u of Eq. (3.3) is a Pω (λ0 )-solution
if and only if the solution (y1 , y2 ) of system (3.8) corresponding to it by virtue of the changes of
variables (3.5) is a Pω ((1 − σ2 )/λ0 )-solution. To clarify the problem on the existence, asymptotic
behavior, and number of such solutions of system (3.3) with the use of Theorem 2.1, one should
compute the numbers μi , A∗i , and βi (i = 1, 2) and the functions Ii (i = 1, 2) in this case with
regard of properties of the function ψ and relation (3.9).
First, by taking into account the form and properties of the function ψ, we note that μ2 =
μ02 sgn(1 − σ2 ) and μ1 = μ01 . Next, we single out two cases depending on whether the number
1 − 1/(1 − σ2 )Λ1 = 1 − 1/λ0 is zero or not.
If λ0 ∈ R\{0, 1}, i.e., it is nonzero, then we have
⎧
⎪ 1 − σ2 − λ0 σ1
λ0 − 1 ⎨ for λ0 σ1 = 1 − σ2
β1 = 1 −
, β2 = 1 − σ − σ σ 2
λ0 ⎪
⎩ 1 2
for λ0 σ1 = 1 − σ2 ,
σ1 (1 − σ2 )
⎧
⎪
⎪ t
t ⎪
⎨ p(τ ) dτ for λ0 σ1 = (1 − σ2 )
I1 (t) = dτ = πω (t), I2 (t) = A t2
⎪
⎪
A1 ⎪ πω (τ )p(τ ) dτ
⎩ for λ0 σ1 = (1 − σ2 ),
A2
⎧ 
⎪
⎪ t
⎪
⎪ for λ0 σ1 = (1 − σ2 )
⎨ sgn p(τ ) dτ
A∗1 = sgn πω (t), ∗
A2 =  A2

⎪
⎪ t
⎪
⎪ |πω (τ )|p(τ ) dτ for λ0 σ1 = (1 − σ2 ),
⎩ sgn
A2

where the integration limits A1 , A2 ∈ {ω, a} are chosen so as to ensure that the corresponding
integrals tend either to zero or to ±∞ as t ↑ ω,

πω (t) = t − a for ω = +∞
t − ω for ω < +∞.
If λ0 = 1, i.e., 1 − (1/(1 − σ2 ))Λ1 = 0, then we have

t t
1 − σ1 − σ2
β1 = 1 − σ1 − σ2 , β2 = , I1 (t) = I2 (τ ) dτ, I2 (t) = p(τ ) dτ,
1 − σ2
A1 A2
 t
A∗1 = sgn |I2 (τ )| dτ , A∗2 = sgn I2 (t) for t ∈ ]a, ω[,
A1

where the integration limits A1 , A2 ∈ {ω, a} are chosen just as above.

For λ0 ∈ R\{0}, the algebraic equation (2.1) for system (3.8) and hence for Eq. (3.3) acquires
the form ν 2 + (λ0 + σ2 − 1)λ−1 −1
0 ν + (1 − σ1 − σ2 )λ0 = 0. This equation has no roots with zero real
part if and only if one of the following two conditions is satisfied:

either λ0 = σ2 − 1 or λ0 = σ2 − 1 and (1 − σ1 − σ2 )(1 − σ2 ) > 0. (3.10)

DIFFERENTIAL EQUATIONS Vol. 48 No. 5 2012

642 EVTUKHOV, VLADOVA

Finally, by using relations (3.6) and (3.7), we note that

y1 (t) u(t) y2 (t) ψ(u (t)) u (t)

=  , = ∼ as t ↑ ω.
−1
ψ (y2 (t)) u (t) ϕ1 (y1 (t)) ϕ1 (u(t)) (1 − σ2 )ϕ1 (u(t))ϕ2 (u (t))

By taking into account the above-performed considerations and Theorem 2.1, we obtain the
following two assertions for Eq. (3.3).

Corollary 3.2. Let λ0 ∈ R\{0, 1}. Then, for the existence of Pω (λ0 )-solutions of Eq. (3.3), it is
necessary and, if one of conditions (3.10) is satisfied, sufficient that, in addition to conditions (3.4)
and (3.5), the following requirements hold :
 t 
1 − σ2 − λ0 σ1
lim πω (t)p(t) p(τ ) dτ = if λ0 σ1 = 1 − σ2 ,
t↑ω λ0 − 1
A2
 t 
2
lim πω (t)p(t) πω (τ )p(τ ) dτ = 1 if λ0 σ1 = 1 − σ2 ,
t↑ω
A2

μ01 μ02 λ0 (λ0 − 1)πω (t) > 0 for t ∈ ]a, ω[, α0 λ0 μ01 > 0;

moreover , each solution of this kind admits the asymptotic representation

πω (t)u (t) λ0 u (t)

= + o(1), = α(λ0 − 1)πω (t)p(t)[1 + o(1)] (3.11)
u(t) λ0 − 1 ϕ1 (u(t))ϕ2 (u (t))

as t ↑ ω. Moreover , under the above-mentioned conditions, there exists a one-parameter family

of such solutions if λ0 (1 − σ1 − σ2 ) > 0 and a two-parameter family if λ0 (1 − σ1 − σ2 ) < 0 and
λ0 (λ0 + σ2 − 1) < 0.

Corollary 3.3. For the existence of Pω (1)-solutions of Eq. (3.3), it is necessary and, if either
σ2 = 2 or σ2 = 2 and σ1 > −1, sufficient that, in addition to conditions (3.4) and (3.5),
 t  τ  t −2 
lim p(t) p(s) ds dτ p(s) ds = 1,
t↑ω
A1 A2 A2
t
μ01 μ02 (1 − σ1 − σ2 ) p(τ ) dτ > 0 for t ∈ ]a, ω[, α0 μ01 > 0.
A2

In addition, each solution of this kind admits the asymptotic representations

t  t  τ −1
u (t)
= p(τ ) dτ (1 − σ1 − σ2 ) p(s) ds dτ [1 + o(1)], (3.12)
u(t)
A2 A1 A2
t
u (t)
= α0 (1 − σ1 − σ2 ) p(τ ) dτ [1 + o(1)] (3.13)
ϕ1 (u(t))ϕ2 (u (t))
A2

as t ↑ ω; moreover , such solutions form a one-parameter family if 1 − σ1 − σ2 > 0 and a two-

parameter family if 1 − σ1 − σ2 < 0 and σ2 < 0.

DIFFERENTIAL EQUATIONS Vol. 48 No. 5 2012

 ASYMPTOTIC REPRESENTATIONS OF SOLUTIONS OF ESSENTIALLY NONLINEAR . . . 643

Remark 3.2. If the functions ϕ1 and ϕ2 satisfy condition S, then, by using Remark 2.2, from
the representations (3.11)–(3.13), one can readily obtain closed-form asymptotic representations
for u and u .
Remark 3.3. The above-proved representations (3.11)–(3.13) differ from the representations of
Pω (λ0 )-solutions (for λ0 ∈ R\{0}) obtained in [8–11]. In addition, in the present paper, the result
on Pω (1)-solutions has been obtained under less restrictive assumptions.
In conclusion, consider the nth-order differential equation
u(n) = α0 p(t)ϕ(u), (3.14)
where α0 ∈ {−1, 1}, p : [a, ω[ → ]0, +∞[ is a continuous function and ϕ : Δ(U00 ) → ]0, +∞[ is
a continuously differentiable function properly varying of order σ = 1 as z → U00 , where Δ(U00 ) is a
one-sided neighborhood of the point U00 and either U00 = 0 or U00 = ±∞.
A solution u of Eq. (3.14) is called a Pω (λ0 )-solution, where −∞ ≤ λ0 ≤ +∞, if it is defined on
some interval [t0 , ω[ ⊂ [a, ω[ and satisfies the conditions
u : [t0 , ω[ → Δ(U00 ), lim u(t) = U00 , (3.15)
t↑ω

either 0 [u(n−1) (t)]2
lim u(k) (t) = Uk0 = (k = 1, . . . , n − 1), lim = λ0 . (3.16)
t↑ω
or ±∞ t↑ω u(n) (t)u(n−2) (t)

By virtue of the form of Eq. (3.14), each of its Pω (λ0 )-solutions and their derivatives of order
≤ n have constant sign on some interval [t1 , ω[ ⊂ [t0 , ω[.
Let μ00 = 1 if either U00 = +∞ or U00 = 0 and Δ(U00 ) is a right neighborhood of the point 0;
μ0 = −1 if either U00 = −∞ or U00 = 0 and Δ(U00 ) is a left neighborhood of the point 0; μ01 = 1
0

if Δ(U00 ) is a left neighborhood of the point U00 ; and μ01 = −1 if Δ(U00 ) is a right neighborhood of
the point U00 ; these numbers specify the signs of a Pω (λ0 )-solution and its first derivative. In this
connection, note that the following conditions should be satisfied:
μ00 μ01 < 0 if U00 = 0, μ00 μ01 > 0 if U00 = ±∞, (3.17)
which is necessary for the existence of Pω (λ0 )-solutions.
Theorems 2.1 and 2.2 permit one to study the problem on the existence and asymptotic behavior
of Pω (λ0 )-solutions of Eq. (3.14) for the case in which
" 1 2 n−2
#
λ0 ∈ R 0, , , . . . , . (3.18)
2 3 n−1
Indeed, by using the changes of variables
u(i−1) = yi (i = 1, . . . , n), (3.19)
one can reduce Eq. (3.14) to the system of differential equations
yi = yi+1 (i = 1, . . . , n − 1), yn = α0 p(t)ϕ(y1 ). (3.20)
This is a system of the form (1.1), where
αi = μi+1 = sgn yi+1 (i = 1, . . . , n − 1), αn = α0 ,
pi (t) ≡ 1 (i = 1, . . . , n − 1), pn (t) = p(t),
ϕi (yi ) = |yi | (i = 2, . . . , n), ϕ1 (y1 ) = ϕ(y1 ), μ1 = μ00 .
Here the ϕi (i = 2, . . . , n) are properly varying functions of order 1 both as yi → 0 and as yi → ±∞;
i.e., σi = 1 (i = 2, . . . , n) and σ1 = σ. In addition, by Lemma 10.1 in [12], if condition (3.18) is
satisfied, then each Pω (λ0 )-solution of Eq. (3.14) satisfies the limit relations
[u(i) (t)]2 (n − i)λ0 − (n − i − 1)
lim = = 0 (i = 1, . . . , n − 1). (3.21)
t↑ω u(i−1) (t)u (i+1) (t) (n − i − 1)λ0 − (n − i − 2)
DIFFERENTIAL EQUATIONS Vol. 48 No. 5 2012
644 EVTUKHOV, VLADOVA

By virtue of conditions (3.19), we have


yi (t)yi+1 (t) u(i−1) (t)u(i+1) (t)
= (i = 1, . . . , n − 1).
yi (t)yi+1 (t) [u(i) (t)]2

Therefore, a solution u of Eq. (3.14) is a Pω (λ0 )-solution if and only if the solution (y1 , . . . , yn ) of
system (3.20) corresponding to it by the changes of variables (3.19) is a Pω (Λ1 , . . . , Λn−1 )-solution,
where
(n − i − 1)λ0 − (n − i − 2)
Λi = (i = 1, . . . , n − 1).
(n − i)λ0 − (n − i − 1)
1
In this case, we have Λn = = (n − 1)λ0 − (n − 2).
Λ1 . . . Λn−1
By virtue of the above-mentioned form of Λi (i = 1, . . . , n), we have

1 − σi+1 Λi = 1 − Λi = (λ0 − 1)/[(n − i)λ0 − (n − i − 1)] (i = 1, . . . , n − 1),

1 − σ1 Λn = 1 − σΛn = 1 − σ[(n − 1)λ0 − (n − 2)].

Consequently, one should consider two cases in which λ0 = 1 and λ0 = 1.

In the first case, the constants βi and A∗i (i = 1, . . . , n) and the functions Ii (i = 1, . . . , n)
occurring in Section 1 are defined as follows:

βi = 1 − Λi = (λ0 − 1)/[(n − i)λ0 − (n − i − 1)] (i = 1, . . . , n − 1),


1 − σΛn = 1 − σ[(n − 1)λ0 − (n − 2)] for 1 − σΛn = 0
βn =
(1 − Λ1 )/(Λ1 · · · Λn−1 ) = λ0 − 1 for 1 − σΛn = 0,
⎧
⎪
⎪ t
t ⎪
⎨ p(τ ) dτ for 1 − σΛn = 0
Ii (t) = dτ = πω (t), i = 1, . . . , n − 1, In (t) = A nt
⎪
⎪
Ai ⎪
⎩ πω (τ )p(τ ) dτ for 1 − σΛn = 0,
An
⎧ 
⎪
⎪ t
⎪
⎪ sgn p(τ ) dτ for 1 − σΛn = 0
⎨ An
A∗i = sgn πω (t), i = 1, . . . , n − 1, A∗n = 
⎪
⎪ t
⎪
⎪ |πω (τ )|p(τ ) dτ for 1 − σΛn = 0,
⎩ sgn
An

where

t for ω = +∞
πω (t) =
ω−t for ω < +∞,
and each of the integration limits Ai ∈ {ω, a} (i = 1, . . . , n) is chosen so as to ensure that the
corresponding integral Ii tends either to zero or to ∞ as t ↑ ω.
If λ0 = 1, then we have

t t
βi = 1 − σ, i = 1, . . . , n, In (t) = p(τ ) dτ, Ii (t) = In (τ ) dτ, i = 1, . . . , n − 1,
An A
 t
A∗n = sgn In (t), A∗i = sgn |In (τ )| dτ , i = 1, . . . , n − 1,
A

where each of the integration limits A, An ∈ {ω, a} is chosen just as above.

DIFFERENTIAL EQUATIONS Vol. 48 No. 5 2012

 ASYMPTOTIC REPRESENTATIONS OF SOLUTIONS OF ESSENTIALLY NONLINEAR . . . 645

Under condition (3.18), we rewrite the algebraic equation (2.1) for system (3.20) in the form
 n   n  
(n − i)λ0 − (n − i − 1) (n − i)λ0 − (n − i − 1)
+ν −σ = 0,
i=1
(n − 1)λ0 − (n − 2) i=1
(n − 1)λ0 − (n − 2)
or

n−1 
n−1
(1 + ) ((n − i)λ0 − (n − i − 1) + ) = σ ((n − i)λ0 − (n − i − 1)), (3.22)
i=1 i=1

where  = ν[(n − 1)λ0 − (n + 2)]. For λ0 = 1, it acquires the form

(1 + ν)n = σ. (3.23)
Finally, note that, by virtue of the changes of variables (3.19),
yi (t) u(i−1) (t) yn (t) u(n−1) (t)
= (i) , i = 1, . . . , n − 1, = ,
yi+1 (t) u (t) ϕ1 (y1 (t)) ϕ(u(t))
and hence
yn (t)  yi (t)
n−1
u(t)
= .
ϕ(u(t)) ϕ1 (y1 (t)) i=1 yi+1 (t)

In addition, it follows from (2.21) that

u(i) (t) (n − i)λ0 − (n − i − 1) u (t)
= [1 + o(1)] (i = 2, . . . , n) as t ↑ ω.
u(i−1) (t) (n − 1)λ0 − (n − 2) u(t)
By virtue of the above argument and Theorem 2.1, the following two assertions are true for the
differential equation (3.14).
Corollary 3.4. Let condition (3.18) be satisfied. Then, for the existence of Pω (λ0 )-solutions of
the differential equation (3.14), it is necessary and, if the algebraic equation (3.22) has no roots
with zero real part, sufficient that, in addition to (3.17), the relations
  t −1 
1 − σ((n − 1)λ0 − (n − 2))
lim πω (t)p(t) p(τ ) dτ = if 1 − σ((n − 1)λ0 − (n − 2)) = 0,
t↑ω λ0 − 1
An
  t −1 
lim πω2 (t)p(t) πω (τ )p(τ ) dτ =1 if 1 − σ((n − 1)λ0 − (n − 2)) = 0
t↑ω
An

hold and the inequalities


n−1
α0 μ00 [(λ0 − 1)πω (t)] n
[(n − i)λ0 − (n − i − 1)] > 0,
i=1

μ00 μ01 [(n − 1)λ0 − (n − 2)](λ0 − 1)πω (t) > 0

be satisfied for t ∈ ]a, ω[; moreover , each solution of that kind admits the asymptotic representations
u(t) α0 [(λ0 − 1)πω (t)]n p(t)
= n−1 [1 + o(1)], (3.24)
i=1 [(n − i)λ0 − (n − i − 1)]
ϕ(u(t))
u(i) (t) (n − i)λ0 − (n − i − 1)
= [1 + o(1)], i = 1, . . . , n − 1,
u(i−1) (t) (λ0 − 1)πω (t)
as t ↑ ω. Moreover , under these conditions, there exists a k-parameter family of such solutions
provided that the algebraic equation (3.22) has k roots (with regard of multiplicities) whose real
parts have sign coinciding with the sign of the function (1 − λ0 )πω (t) for t ∈ ]a, ω[.

DIFFERENTIAL EQUATIONS Vol. 48 No. 5 2012

646 EVTUKHOV, VLADOVA

Corollary 3.5. For the existence of Pω (1)-solutions of Eq. (3.14), it is necessary and, if
Eq. (3.23) has no roots with zero real part, sufficient that, in addition to (3.17), the condition
 t  τ  t −2 
lim p(t) p(s) ds dτ p(s) ds =1
t↑ω
A An An

be satisfied and the inequalities

t  t n
μ00 μ01 (1 − σ) p(s) ds > 0, α0 μ00 (1 − σ) p(s) ds >0
An An

hold for t ∈ ]a, ω[; moreover , each solution of that kind admits the asymptotic representations
 t n
u(t)
= α0 (1 − σ) p(s) ds p1−n (t)[1 + o(1)], (3.25)
ϕ(u(t))
An
 t −1
u(i) (t)
= p(t) (1 − σ) p(τ ) dτ [1 + o(1)], i = 1, . . . , n − 1,
u(i−1) (t)
An

as t ↑ ω. In addition, under these conditions, there exists a k-parameter family of such solutions
provided that the algebraic equation (3.23) has k roots (with regard of multiplicities) whose real
parts have sign opposite to the sign of μ00 μ01 .
Remark 3.4. If the function ϕ satisfies Condition S, then, by using Remark 2.2, from the
representations (3.24) and (3.25), one can obtain closed-form asymptotic representations for u.

REFERENCES
1. Mirzov, D.D., Asymptotic Properties of Solutions of a System of Emden–Fowler Type, Differ. Uravn.,
1985, vol. 21, no. 9, pp. 1498–1504.
2. Mirzov, D.D., Some Asymptotic Properties of the Solutions of a System of Emden–Fowler Type, Differ.
Uravn., 1987, vol. 23, no. 9, pp. 1519–1532.
3. Mirzov, D.D., Asimptoticheskie svoistva reshenii sistem nelineinykh neavtonomnykh obyknovennykh dif-
ferentsial’nykh uravnenii (Asymptotic Properties of Solutions of Systems of Nonlinear Nonautonomous
Ordinary Differential Equations), Maikop, 1993.
4. Evtukhov, V.M., Asymptotic Representations of Regular Solutions of a Two-Dimensional System of
Differential Equations, Dopov. Nats. Akad. Nauk Ukr., 2002, no. 4, pp. 11–17.
5. Evtukhov, V.M., Asymptotic Representations of Regular Solutions of a Semilinear Two-Dimensional
System of Differential Equations, Dopov. Nats. Akad. Nauk Ukr., 2002, no. 5, pp. 11–17.
6. Seneta, E., Pravil’no menyayushchiesya funktsii (Properly Varying Functions), Moscow, 1985.
7. Evtukhov, V.M. and Samoilenko, A.M., Existence Conditions for Solutions Vanishing at Singular Point
for Real Nonautonomous Systems of Quasilinear Differential Equations, Ukrain. Mat. Zh., 2010, vol. 62,
no. 1, pp. 52–80.
8. Belozerova, M.A., Asymptotic Properties of a Class of Solutions of Essentially Nonlinear Second-Order
Differential Equations, Mat. Stud., 2008, vol. 29, no. 1, pp. 52–62.
9. Bilozerova, M.O., Nauk. Visn. Cherniv. Univ., Chernivtsi, 2008, vol. 374, pp. 34–43.
10. Belozerova, M.A., Asymptotic Representations of Solutions of Second-Order Nonautonomous Differential
Equations with Nonlinearities Close to Power Type, Nelin. Koleb., 2009, vol. 12, no. 1, pp. 3–15.
11. Evtukhov, V.M. and Belozerova, M.A., Asymptotic Representations of Solutions of Essentially Nonlinear
Nonautonomous Differential Equations of the Second Order, Ukrain. Mat. Zh., 2008, vol. 60, no. 3,
pp. 310–331.
12. Evtukhov, V.M., Asymptotic Representations of Solutions of Nonautonomous Ordinary Differential
Equations, Doctoral (Phys.–Math.) Dissertation, Kiev, 1998.

DIFFERENTIAL EQUATIONS Vol. 48 No. 5 2012