Vietnam Journal

of
MATHEMATICS
Formerly Ta.p chı́ Toán ho.c (Journal of Mathematics)

Volume 38, Number 3 June 2009

Editor-in-Chief
Nguyen Khoa Son
Institute of Mathematics, Hanoi

Deputy Editor-in-Chief
Pham Ky Anh Nguyen Tu Cuong
Hanoi University of Science, Hanoi Institute of Mathematics, Hanoi

Hoang Xuan Phu

Institute of Mathematics, Hanoi

Editorial Board
Hans George Bock, Universität Heidelberg
Markus Brodmann, Universität Zürich
Ngo Bao Chau, Universite Paris-Sud, Orsay
Nguyen Dinh Cong, Institute of Mathematics, Hanoi
Dinh Dung, Institute of Information Technology, VNU, Hanoi
Phung Ho Hai, Institute of Mathematics, Hanoi
Nguyen Huu Du, Vietnam National University, Hanoi
Karl-Heinz Hoffmann, Caesar, Bonn
Nguyen Huu Viet Hung, Vietnam National University, Hanoi
Masami Ito, Kyoto Sangyo University, Kyoto
Ha Huy Khoai, Institute of Mathematics, Hanoi
Pham The Long, Le Qui Don University, Hanoi
Charles A. Micchell, State University of New York, New York
Tran Van Nhung, Mistry of Education and Traing, Hanoi
Shum Kar Ping, The Chinese University of Hong Kong
J. P. Ramis, Université Paul Sabatier, Toulouse
Ngo Dac Tan, Institute of Mathematics, Hanoi
Vladimir N. Temlyakov, University of South Carolina, USA
Do Duc Thai, Vietnam National University, Hanoi
Nguyen Quoc Thang, Institute of Mathematics, Hanoi
Dao Trong Thi, Vietnam National University, Hanoi
Dang Duc Trong, University of Ho Chi Minh City
Tran Duc Van, Institute of Mathematics, Hanoi
Nguyen Thanh Van, Université Paul Sabatier, Toulouse
Ha Huy Vui, Institute of Mathematics, Hanoi
Robert Wishbauer, Heinrich - Heine - Universität, Düsseldorf

Vietnam Academy of
Scien and Technology
&
Vietnam Mathematical Society
 Vietnam Journal
of
MATHEMATICS
Volume 38, Number 3 June 2009

Vietnam Academy of
Scien and Technology
&
Vietnam Mathematical Society
 Vietnam Journal
of
MATHEMATICS
AIMS AND SCOPE
The Vietnam Journal of Mathematics seeks to publish research papers, review articles and
short communications in all areas of mathematics with the aim of disseminating original
research material to both the regional and international scientific community.

SUBSCRIPTION INFORMATION
Suggested list price: US$ 300,00 plus carriage charges.
Orders should be addressed to:

Department of Distribution Marketing, Room i.2.2,

Pullishing House of Science and Technology
A11 Building, VAST, 18 Hoang Quoc Viet, Hanoi, Vietnam
Tel: +84-4-2149040 Fax: +84-4-7910147
Email: phongphathanh@gmail.com

Change of address: Allow 6 weeks for all changes to become effective. All communications
should include both and new addresses (with postal codes) and should be accompanied by
a mailing label from a recent issue.

COPYRIGHT INFORMATION
Submission of a manuscript implies:

• that the work described has not been published before (except in the form of an abstract
or as part of a published lecture, review, or thesis);
• that it is not under consideration for publication elsewhere;
• that its publication has been approved by all co-authors, if any, as well as by the re-
sponsible authorities at the institute whre the work has been carried out;
• that, if and when the manuscript is accepted for publication, the authors agree to
automatic transfer of the copyright to the editorial office;
• that the manuscript will not be published elsewhere in any language without the consent
of the copyright holders; and
• that written permission of the copyright holder is obtained by the authors for material
used from other copyrighted sources.

All articles published in this journal are protected by copyright, which covers the exclusive
rights to reproduce and distribute the article (e.g., as offprints), as well as all translation
rights. No material published in this journal may be reproduced photographically or stored
on microfilm, in electronic databases, video disks, etc., without first obtaining written
permission from the editorial board.

c
2009 Vietnam Academy of Science and Technology (VAST)
Vietnam Journal of Mathematics 38:3(2009) 125-136
  


       
© VAST 2009

T1 Theorems for Inhomogeneous Besov

and Triebel-Lizorkin Spaces over
Space of Homogeneous Type ?

Yanchang Han

Department of Mathematics
South China Normal University, Guangzhou, 510631, P.R. China

Received November 17, 2006

Abstract. The author establishes T 1 theorems for inhomogeneous Besov and Triebel-
Lizorkin spaces by discrete Calderón type reproducing formula and the Plancherel-
Pôlya characterization for inhomogeneous Besov and Triebel-Lizorkin spaces. These
results are new even for Rd .

2000 Mathematics Subject Classification: 42B25, 42B35, 46E35.

Key words: T 1 theorem, inhomogeneous Besov and Triebel-Lizorkin spaces, spaces of
homogeneous type.

1. Introduction

In the past years, there has been significant progress on the problem of proving
the boundedness of generalized Calderón-Zygmund operators on various function
spaces. A remarkable result is the famous T 1 theorem of David and Journé in
[3]. T 1 theorem has been extended for Besov and Triebel-Lizorkin spaces. For a
broader view of this active area of research, see e.g. [5, 10, 12–14, 16, 17] and
references therein.
The main purpose of this paper is to establish T 1 theorems for the inhomo-
d
geneous spaces Bpα,q (X) when d+α < p ≤ ∞, 0 < q ≤ ∞, 0 < α <  and for
d d
Fpα,q (X) when d+α < p < ∞, d+α < q ≤ ∞, 0 < α < , and for Bpα,q (X)
?
Project supported in part by the NNSF
126 Yanchang Han

d
when d+α+ < p ≤ ∞, 0 < q ≤ ∞, − < α < 0 and for Fpα,q (X) when
d
d+α+ < p < ∞, d+dα+  < q ≤ ∞, − < α < 0 for some  > 0 by discrete
Calderón type reproducing formula and Plancherel-Pôlya characterization for
the inhomogeneous Besov and Triebel-Lizorkin spaces. Roughly speaking, T is
bounded on Bpα,q (X) and Fpα,q (X) for the range of α, p, and q indicated above,
respectively, if its kernel satisfies only half smoothness and moment conditions.
An application of these results is given in [4].
To state main results of this paper, we begin by recalling the definitions
necessary for inhomogeneous Besov and Triebel-Lizorkin spaces on spaces of
homogeneous type and some basic facts about the Calderón-Zygmund operator
theory. A quasi-metric ρ on a set X is a function ρ : X × X → [0, ∞) satisfying:

(i) ρ(x, y) = 0 if and only if x = y;

(ii) ρ(x, y) = ρ(y, x) for all x, y ∈ X;
(iii) There exists a constant

A ∈ [1, ∞) such that for all x, y, z ∈ X,

ρ(x, y) ≤ A[ρ(x, z) + ρ(z, y)].

Any quasi-metric defines a topology, for which the balls B(x, r) = {y ∈ X :

ρ(y, x) < r} for all x ∈ X and all r > 0 form a basis.
The following spaces of homogeneous type are variants of those introduced
by Coifman and Weiss in [2].

Definition 1 (13). Let d > 0 and 0 < θ ≤ 1. A space of homogeneous type

(X, ρ, µ)d,θ is a set X together with a quasi-metric ρ and a nonnegative Borel
measure µ on X with suppµ = X, and there exists a constant 0 < C < ∞ such
that for all 0 < r < diam X and all x, x0 , y ∈ X,

µ(B(x, r)) ∼ rd , (1.1)

|ρ(x, y) − ρ(x0 , y)| ≤ Cρ(x, x0 )θ [ρ(x, y) + ρ(x0 , y)]1−θ . (1.2)

In [14], Macı́as and Segovia have proved that one can replace a quasi-metric
ρ of spaces of homogeneous type in the sense of Coifman and Weiss by another
quasi-metric % which yields the same topology on X as ρ such that (X, %, µ) is
the space defined by Definition 1.1 with d = 1.
Suppose that T is a continuous linear mapping from C0η (X) to (C0η (X))0 ,
associated to a kernel K(x, y) in the following sense that
Z Z
hT f, gi = g(x)K(x, y)f (y)dµ(x)dµ(y)

for all test functions f and g in C0η with disjoint supports.

Assume that K(x, y) satisfies the pointwise conditions:
T1 Theorems for Inhomogeneous Besov and Triebel-Lizorkin Spaces... 127

|K(x, y)| ≤ Cρ(x, y)−d for ρ(x, y) 6= 0, (1.3)

|K(x, y)| ≤ Cρ(x, y)−d−σ for ρ(x, y) ≥ 1, (1.4)
ρ(x, y)
|K(x, y) − K(x0 , y)| ≤ Cρ(x, x0 ) ρ(x, y)−d− for ρ(x, x0 ) ≤ , (1.5)
(2A)
ρ(x, y)
|K(x, y) − K(x, y 0 )| ≤ Cρ(y, y 0 ) ρ(x, y)−d− for ρ(y, y 0 ) ≤ , (1.6)
(2A)

where  ∈ (0, θ), σ > 0.

The conditions (1.3)-(1.6) are natural when one considers the boundedness
of Calderón-Zygmund operators on inhomogeneous function spaces, which were
pointed out by Meyer in [16].
Assume also that T satisfies the Weak Boundedness Property, denote this by
T ∈ WBP,
|hT f, gi| ≤ Crd+2η kf kC0η (X) kgkC0η (X)
for all f and g in C0η (X) with diameters of supports not greater than r.
To state the definition of the inhomogeneous Besov and Triebel-Lizorkin
spaces, we need the following definitions. Let Z+ = N ∪ {0}.

Definition 2. [9] A sequence {Sk }k∈Z+ of operators is said to be an approxima-

tion to the identity if Sk (x, y), the kernel of Sk , are functions from X × X into
C such that for all k ∈ Z+ and all x, x0 , y and y 0 in X, and some 0 <  ≤ θ and
C > 0,
2−k
|Sk (x, y)| ≤ C −k ; (1.7)
(2 + ρ(x, y))d+
 
ρ(x, x0 ) 2−k
|Sk (x, y) − Sk (x0 , y)| ≤ C −k
(1.8)
2 + ρ(x, y) −k
(2 + ρ(x, y))d+
1
for ρ(x, x0 ) ≤ 2A (2
−k
+ ρ(x, y));
 
0 ρ(y, y 0 ) 2−k
|Sk (x, y) − Sk (x, y )| ≤ C (1.9)
2−k + ρ(x, y) (2−k + ρ(x, y))d+
1
for ρ(y, y 0 ) ≤ 2A (2
−k
+ ρ(x, y));
 
[Sk (x, y) − Sk (x, y 0 )] − [Sk (x0 , y) − Sk (x0 , y 0 )]

 0  0
ρ(x, x0 ) ρ(y, y 0 ) 2−kσ
≤C −k −k
2 + ρ(x, y) 2 + ρ(x, y) (2−k + ρ(x, y))d+σ
1
for 0 < 0 < , σ =  − 0 > 0, ρ(x, x0 ) ≤ 2A (2−k + ρ(x, y)) and ρ(y, y 0 ) ≤
1 −k
2A (2 + ρ(x, y)); Z
Sk (x, y)dµ(x) = 1 (1.10)

for all k ∈ Z+ ;
 128 Yanchang Han
Z
Sk (x, y)dµ(y) = 1 (1.11)

for all k ∈ Z+ .

Definition 3. [12] Fix two exponents 0 < β ≤ θ and γ > 0. A function f defined
on X is said to be a test function of type (β, γ) centered at x0 ∈ X with width
d > 0 if f satisfies the following conditions:
dγ
|f (x)| ≤ C ; (1.12)
(d + ρ(x, x0 ))d+γ
 β
0 ρ(x, x0 ) dγ
|f (x) − f (x )| ≤ C (1.13)
d + ρ(x, x0 ) (d + ρ(x, x0 ))d+γ
1
for ρ(x, x0 ) ≤ 2A (d + ρ(x, x0 )).

If f is a test function of type (β, γ) centered at x0 with width d > 0, we write

f ∈ M(x0 , d, β, γ), and the norm of f in M(x0 , d, β, γ) is defined by

kf kM(x0,d,β,γ) = inf{C ≥ 0 : (1.12) and (1.13) hold}.

We denote by M(β, γ) the class of all f ∈ M(x0 , 1, β, γ). It is easy to see

that M(x1 , d, β, γ) = M(β, γ) with equivalent norms for all x1 ∈ X and d > 0.
Furthermore, it is also easy to check that M(β, γ) is a Banach space with respect
to the norm in M(β, γ). We denote by (M(β, γ))0 the dual space of M(β, γ)
consisting of all linear functionals L from M(β, γ) to C with the property that
there exists a constant C such that for all f ∈ M(β, γ),

|L(f )| ≤ Ckf kM(β,γ).

We denote by hh, f i the natural pairing of elements h ∈ (M(β, γ))0 and f ∈

M(β, γ). Since M(x1 , d, β, γ) = M(β, γ) with the equivalent norms for all x1 ∈
X and d > 0, thus, for all h ∈ (M(β, γ))0 , hh, f i is well defined for all f ∈
f γ) be the
M(x0 , d, β, γ) with x0 ∈ X and d > 0. In what follows, we let M(β,
completion of M(θ, θ) in M(β, γ) when 0 < β, γ < θ.
We also need the following construction of Christ in [1], which provides an
analogue of the grid of Euclidean dyadic cubes on spaces of homogeneous type.

Lemma 4. Let X be a space of homogeneous type. Then there exist a collection

{Qkα ⊂ X : k ∈ Z+ , α ∈ Ik } of open subsets, where Ik is some (possible finite)
index set, and constants δ ∈ (0, 1) and C1 , C2 > 0 such that

(i) µ(X\ ∪α Qkα ) = 0 for each fixed k and Qkα ∩ Qkβ = ∅ if α 6= β;

(ii) for any α, β, k, l with l ≥ k, either Qlβ ⊂ Qkα or Qlβ ∩ Qkα = ∅;
(iii) for each (k, α) and each l < k there is a unique β such that Qkα ⊂ Qlβ ;
(iv) diam(Qkα ) ≤ C1 δ k ;
 T1 Theorems for Inhomogeneous Besov and Triebel-Lizorkin Spaces... 129

(v) each Qkα contains some ball B(zαk , C2 δ k ), where zαk ∈ X.

In fact, we can think of Qkα as being a dyadic cube with diameter roughly
δ k and centered at zαk . In what follows, we always suppose δ = 1/2. See [12] for
how to remove this restriction. Also, in the following, for k ∈ Z+ , τ ∈ Ik , we will
k+M
denote by Qk,ντ , ν = 1, . . . , N (k, τ, M ), the set of all cubes Qτ 0 ⊂ Qkτ , where
M is a fixed large positive integer.
Now, we can introduce the inhomogeneous Besov spaces Bpα,q (X) and Triebel-
Lizorkin spaces Fpα,q (X) via approximations to the identity.

Definition 5. Suppose that −θ < α < θ, and β and γ satisfying

max(0, −α + max(0, d(1/p − 1))) < β < θ, 0 < γ < θ. (1.14)

Suppose {Sk }k∈Z+ is an approximation to identity and let D0 = S0 , and Dk =

0,ν
Sk − Sk−1 for k ∈ N. Let M be a fixed large positive
 integer,
 Qτ be as above.
d d
Inhomogeneous Besov space Bpα,q (X) for max d+θ , d+θ+α < p ≤ ∞, 0 < q ≤
f γ))0 such that
∞ is the collection of all f ∈ (M(β,
  p1
X N (0,τ,M)
X 
kf kBpα,q (X) = µ(Q0,ν
τ )[mQ0,ν (|D0 (f )|)]p
 τ 
τ ∈I0 ν=1
( ∞
) q1
X  q
kα
+ 2 kDk (f )kLp (X) < ∞.
k=1
 
d d
Inhomogeneous Triebel-Lizorkin space Fpα,q (X) for max d+θ , d+θ+α <p<∞
 
d
and max d+θ d
, d+θ+α f γ))0 such that
< q ≤ ∞ is the collection of f ∈ (M(β,

  p1
X N (0,τ,M)
X 
kf kFpα,q (X) = µ(Q0,ν
τ )[mQ0,ν (|D0 (f )|)]p
 τ 
τ ∈I0 ν=1
( ) q1
X
∞  kα q

+ 2 |Dk (f )| < ∞,

k=1
Lp (X)

where mQ0,ν
τ
(D0 (f )) are averages of D0 (f ) over Q0,ν
τ .

T 1 theorems for inhomogeneous Besov and Triebel-Lizorkin spaces were

proved in [10]. Roughly speaking, T is bounded on Bpα,q , 1 ≤ p, q ≤ ∞ and
0 < α < , and on Fpα,q , 1 < p, q < ∞ and 0 < α < , if T has the weak
boundedness property, T 1 = 0 and the conditions (1.3)–(1.5) hold in [10]. In
this paper, we will prove the following results.
130 Yanchang Han

Theorem A. Let 0 <  ≤ θ, 0 < α < . Suppose that T (1) = 0, T ∈ WBP, and
K(x, y), the kernel of T, satisfies (1.3) − (1.5) with σ > max(0, d( 1p − 1)). Then
d
T is bounded on Bpα,q (X), for d+α < p ≤ ∞, 0 < q ≤ ∞, and on Fpα,q (X), for
d d
d+α < p < ∞, d+α < q ≤ ∞.

Theorem B. Let 0 <  ≤ θ, − < α < 0. Suppose that T ∗ (1) = 0, T ∈ WBP, and
K(x, y), the kernel of T, satisfies (1.3), (1.4) and (1.6) with σ > max(0, d( p1 −1)).
d
Then T is bounded on Bpα,q (X), for d+α+ < p ≤ ∞, 0 < q ≤ ∞, and on
α,q d d
Fp (X), for d+α+ < p < ∞, d+α+ < q ≤ ∞.

Theorems A and B are to give a uniform treatment in [10]. To be precise,

to deal with the case where 0 < α < , p, q > 1, the main tools used were
the continuous Calderón reproducing formula. The proof of the case where
− < α < 0, and p, q > 1 then follows from the duality argument. How-
ever, the continuous Calderón reproducing formula and duality argument do
not work for the cases where either p or q, or both p and q are less than or
equal to 1. The key point of the present paper is to use discrete Calderón repro-
ducing formula and Plancherel-Pôlya characterization of the Besov and Triebel-
Lizorkin spaces developed in [6, 11]. T 1 theorems for inhomogeneous Triebel-
d d
Lizorkin space Fpα,q (X) with − < α < , max{ d+ , d+α+ } < p < ∞ and
d d
max{ d+ , d+α+ } < q ≤ ∞ in [17] are also stated, if T has the weak bound-
edness property, T (1) = 0, T ∗ (1) = 0 and the conditions (1.3)–(1.6) hold.
Furthermore, by use of the real interpolation theorems the author obtained
the T 1 theorem for inhomogeneous Besov space Bpα,q (X) with − < α < ,
d d
max{ d+ , d+α+ } < p ≤ ∞ and 0 < q ≤ ∞ under the same conditions. The
d d
range of index p and q has the change from d+ to d+α with 0 < α < , the
main reason is that the smoothness and moment conditions of Theorems A and
B decrease a half compared with the corresponding results of [17].

2. Proofs of Theorems A and B

The basic tool to show main results is the discrete Calderón reproducing formulae
in [6]. It can be stated as follows.

Lemma 1. Suppose that {Sk }k∈Z+ is an approximation to the identity as in

Definition 1.2. Set Dk = Sk − Sk−1 for k ∈ N and D0 = S0 . Then there exist
functions De 0,ν , τ ∈ I0 and ν ∈ {1, . . . , N (0, τ, M )} and {D e k (x, y)}k∈N such
Qτ
that for any fixed yτk,ν ∈ Qk,ν
τ , k ∈ N, τ ∈ Ik and ν ∈ {1, . . . , N (k, τ, M )} and all
f ∈ (M(β, γ))0 with 0 < β, γ < θ,
T1 Theorems for Inhomogeneous Besov and Triebel-Lizorkin Spaces... 131

N (0,τ,M)
X X
f (x) = µ(Q0,ν e 0,ν (x)
τ )mQ0,ν
τ
(D0 (f ))D Qτ
τ ∈I0 ν=1
N (k,τ,M)
X X X
+ µ(Qk,ν e k,ν k,ν
τ )Dk (x, yτ )Dk (f )(yτ ), (2.1)
k∈Z+ τ ∈Ik ν=1

where diam (Qk,ν τ ) ∼ 2

k+M
for k ∈ Z+ , τ ∈ Ik , ν ∈ {1, . . . , N (k, τ, M )} and
a fixed large M ∈ N, the series converges in the norm of Lp (X), 1 < p < ∞,
and M(β 0 , γ 0 ) for f ∈ M(β, γ) with β 0 < β and γ 0 < γ, and (M(β 0 , γ 0 ))0 for
f ∈ (M(β, γ))0 with θ > β 0 > β and θ > γ 0 > γ. Moreover, D e k (x, y), k ∈ N,
satisfies for any given  ∈ (0, θ), all x, y ∈ X the following conditions:
0

e k (x, y)| ≤ C 2−k

|D ; (2.2)
−k
(2 + ρ(x, y))d+0
 0 0

e k (x, y) − D
e k (x0 , y)| ≤ C ρ(x, x0 ) 2−k
|D (2.3)
2−k + ρ(x, y) (2−k + ρ(x, y))d+0
1
for ρ(x, x0 ) ≤ 2A (2
−k
+ ρ(x, y));
Z Z
e k (x, y)dµ(y) =
D e k (x, y)dµ(x) = 0
D
X X

for all k ∈ Z+ .
e 0,ν (x) for τ ∈ I0 and ν ∈ {1, . . . , N (0, τ, M )} satisfies
D Qτ
Z
e 0,ν (x)dµ(x) = 1,
D Qτ
X

e 0,ν (x)| ≤ C
|D Qτ (2.4)
(1 + ρ(x, y))d+
for all x ∈ X and y ∈ Q0,ν
τ and
 
e 0,ν (z)| ≤ C
e 0,ν (x) − D ρ(x, z) 1
|D Qτ Qτ (2.5)
1 + ρ(x, y) (1 + ρ(x, y))d+
1
for all x, z ∈ X and y ∈ Q0,ν
τ satisfying ρ(x, z) ≤ 2A (1 + ρ(x, y)); the constant
C in (2.2) − (2.5) is independent of M .

To prove Theorem A and Theorem B, we need the following lemmas. Their

proofs are similar to that of Lemma 4.1 in [10].

Lemma 2. With notation as in Lemma 2.1 and Theorem A, then

0
(i) for k ∈ Z+ , τ 0 ∈ I0 and ν 0 ∈ {1, . . . , N (0, τ 0 , M )}, yτ0,ν
0 is any fixed point of
0
Q0,ν
τ 0 , x ∈ X,
 132 Yanchang Han

e 0,ν 0 (x)| ≤ C(1 + k)2−k 1

|Dk T D (2.6)
(1 + ρ(x, yτ0,ν
0
Q τ0 0 ))d+σ0

where σ 0 = σ when k = 0 and σ 0 =  when k ∈ N,

(ii) for k ∈ Z+ , k 0 ∈ N, x, y ∈ X,
  2−(k∧k )
0 0

e k0 (x, y)| ≤ C[1+|k−k 0 |] 2(k0 −k)0 ∧ 1

|Dk T D . (2.7)
(2−(k∧k0 ) + ρ(x, y))d+0
Lemma 3. With notation as in Lemma 2.1 and Theorem B, then
0
(i) for k ∈ Z+ , τ 0 ∈ I0 and ν 0 ∈ {1, . . . , N (0, τ 0 , M )}, yτ0,ν
0 is any fixed point of
0
Q0,ν
τ 0 , x ∈ X,
e 0,ν 0 (x)| ≤ C 1
|Dk T D (2.8)
(1 + ρ(x, yτ0,ν
0
Qτ 0
0 ))d+σ0
where σ 0 = σ when k = 0 and σ 0 =  when k ∈ N,
(ii) for k ∈ Z+ , k 0 ∈ N, x, y ∈ X,
  2−(k∧k )
0 0

e k0 (x, y)| ≤ C[1+|k−k 0 |] 2(k−k0 )0 ∧ 1

|Dk T D . (2.9)
(2−(k∧k0 ) + ρ(x, y))d+0
f γ),
Proof of Theorem A. By Lemma 2.1 and Theorem 1.5 in [6], for f ∈ M(β,
we write

nX N (0,τ,M) h N (0,τ 0 ,M)

X X X 0
kT (f )kBpα,q (X) ≤ mQ0,ν
τ
µ(Q0,ν
τ0 )
τ ∈I0 ν=1 τ 0 ∈I0 ν 0 =1

 o 1
e 0,ν 0 (·)|m 0,ν 0 (|D0 (f )|) p p
|D0 T D Q Q
τ0 τ0

nX N (0,τ,M) ∞ N (k0 ,τ 0 ,M)

X  X X X 0 0
+ mQ0,ν
τ
µ(Qkτ 0 ,ν )
τ ∈I0 ν=1 k0 =1 τ 0 ∈Ik0 ν 0 =1

 o 1
e k0 (·, y k0 ,ν )||Dk0 (f )(y k0 ,ν )| p p
0 0 0 0
|D0 T D τ τ

N (0,τ 0 ,M)
nX
∞ X N (l,τ,M)
X  X X 0
+ inf µ(Q0,ν
τ0 )
l=1 τ ∈Il ν=1 z∈Ql,ν
τ τ 0 ∈I0 ν 0 =1

−α 1
p
pq o 1q
× µ(Ql,ν
τ ) l
e
d + p |D T D
0,ν 0 (z)|m 0,ν 0 (|D 0 (f )|)
Q Q τ0 τ0

nX N (k0 ,τ 0 ,M)
∞ X N (l,τ,M)
X  ∞
X X X 0 0
+ inf µ(Qkτ 0 ,ν )
l=1 τ ∈Il ν=1 z∈Ql,ν
τ k0 =1 τ 0 ∈Ik0 ν 0 =1


q o 1
−α 1
e k0 (z, y k0 ,ν )||Dk0 (f )(y k0 ,ν )| p p q
0 0 0 0
× µ(Ql,ν d + p |D T D
τ ) l τ τ
.
= A1 + A2 + A3 + A4 .
T1 Theorems for Inhomogeneous Besov and Triebel-Lizorkin Spaces... 133

The estimate of A4 is similar to Theorem 1 in [5]. It remains to deduce the

estimates of A1 , A2 and A3 .
From (2.6), the Hölder inequality for p > 1 and (a + b)p ≤ ap + bp for p ≤ 1,
we deduce

nX N (0,τ,M) N (0,τ 0 ,M) h ip∧1

X X X 1
A1 ≤ C 0
d+σ 0
τ ∈I0 ν=1 τ 0 ∈I0 ν 0 =1 1 + ρ(yτ0,ν , yτ0,ν
0 )
o p1
[mQ0,ν 0 (|D0 (f )|)]p
τ0

n X N (0,τ 0 ,M) o p1
X
≤C [mQ0,ν 0 (|D0 (f )|)]p
τ0
τ 0 ∈I0 ν 0 =1

≤ C kf kBpα,q (X)
0 0
0,ν
where yτ0,ν is any point of Q0,ν
τ , yτ 0 is any point of Q0,ν
τ0 .
By (2.7), it follows that

nX
∞ N (k0 ,τ 0 ,M) N (0,τ,M) h
X X X X 0 0
A2 ≤ C 2−k d 2−k α [1 + k 0 ]
k0 =1 τ 0 ∈Ik0 ν 0 =1 τ ∈I0 ν=1

1 ip∧1 h 0 0 α 0 0 p o p1
×
d+ 0 µ(Qkτ 0 ,ν )− d |Dk0 (f )(yτk0 ,ν )|
1 + ρ(yτ0,ν , yτk0 ,ν )
0 0

nX ∞ 
 −k0 d −k0 α p∧1 k0 d
pq ∧1
≤ C 2 2 (1 + k 0 ) 2
k0 =1

 X N (k0 ,τ 0 ,M) h ip  pq o 1q
X 0 0 α 1
× µ(Qkτ 0 ,ν )− d + p sup |Dk0 (f )(z)|
0 0
τ 0 ∈Ik0 ν 0 =1 z∈Qk
τ0
,ν

≤ Ckf kBpα,q (X) ,

where these inequalities follow from the fact that

∞ N (k0 ,τ 0 ,M)
X X X 0 0 1
2−k d 2−k α [1 + k 0 ] 0 0
d+0
≤ C,
k0 =1 τ 0 ∈Ik0 ν 0 =1 1+ ρ(yτ0,ν , yτk0 ,ν )
∞
X  0 0 p∧1 k0 d X  0 0 p∧1  pq ∧1
2−k d 2−k α (1 + k 0 ) 2 + →0 2−k d 2−k α (1 + k 0 ) ≤C
k
k0 =1

and the last inequality follows from the Plancherel-Pôlya characterization of the
Besov spaces [6].
134 Yanchang Han

By (2.6), it follows that

N (0,τ 0 ,M)
nX
∞  X N (l,τ,M)
X X X 0  p
A3 ≤ C 2 −dl
µ(Q0,ν
τ 0 ) mQ0,ν (|D0 (f )|)
0
τ0
l=1 τ ∈Il ν=1 τ 0 ∈I0 ν 0 =1
h 1 ip∧1  pq o 1q
× 2lα (1 + l)2−l   d+
1 + ρ(yτl,ν , yτ0,ν
0
0 )
nX  X N (0,τ 0 ,M)
∞
  q
−l (p∧1) p
X 0  p  pq o 1q
≤ C lα
2 (1 + l)2 µ(Q0,ν
τ 0 ) m Q0,ν 0 (|D0 (f )|)
τ0
l=1 τ 0 ∈I0 ν 0 =1

≤ Ckf kBpα,q (X) .

f γ), we have
Similarly, for f ∈ M(β,

kT (f )kFpα,q (X)
nX N (0,τ,M) h  X N (0,τ 0 ,M)
X X 0
≤ µ(Q0,ν
τ ) mQ0,ν
τ
µ(Q0,ν
τ0 )
τ ∈I0 ν=1 τ 0 ∈I0 ν 0 =1
ip o p1
e 0,ν 0 (·)|m 0,ν 0 (|D0 (f )|)
|D0 T D Q τ0
Q τ0

nX N (0,τ,M) h X
∞ N (k0 ,τ 0 ,M)
X X X 0 0
+ mQ0,ν
τ
µ(Qkτ 0 ,ν )
τ ∈I0 ν=1 k0 =1 τ 0 ∈Ik0 ν 0 =1
0 0 0 0
ip o p1
e k0 (·, y k0 ,ν )||Dk0 (f )(y k0 ,ν )|
|D0 T D τ τ

n X
∞ X N (l,τ,M) N (0,τ 0 ,M)
X h X X α
+ inf µ(Ql,ν
τ )
−d
l,ν
l=1 τ ∈Il ν=1 z∈Qτ τ 0 ∈I0 ν 0 =1
iq o q1
e 0,ν 0 (z)|m 0,ν 0 (|D0 (f )|)χ l,ν
|Dl T D Q Q Qτ
τ0 τ0 Lp (X)

n X
∞ N (l,τ,M) h ∞ N (k0 ,τ 0 ,M)
X X X X X 0 0
+ inf | µ(Qkτ 0 ,ν )
l=1 τ ∈Il ν=1 z∈Ql,ν
τ k0 =1 τ 0 ∈Ik0 ν 0 =1
iq o 1q
−α
0 0
e k0 (·, y k0 ,ν )(z)Dk0 (f )(y k0 ,ν )|χ l,ν
0 0
× µ(Ql,ν
τ )
d D TD
l τ τ Qτ
Lp (X)
.
= B1 + B2 + B3 + B4 ,
0 0 0 0
where yτk0 ,ν are any point in Qkτ 0 ,ν .
The estimates of B1 and B4 are similar to A1 above and Theorem 2 in [5],
respectively. It remains to deduce the estimates of B2 and B3 .
From (2.7), the Hölder inequality for q > 1 and (a + b)q ≤ aq + bq for q ≤ 1,
Lemma A.2 in [8], the Fefferman-Stein vector-valued inequality in [7], it follows
T1 Theorems for Inhomogeneous Besov and Triebel-Lizorkin Spaces... 135

that
nh X
∞
0 0 k0 d
B2 ≤ C 2−k d 2−k α [1 + k 0 ]2 r

k0 =1
h  X N (k0 ,τ 0 ,M) r i 1r iq o q1
X 0 0 α 0 0
× M µ(Qkτ 0 ,ν )− d |Dk0 (f )(yτk0 ,ν )|χQk0 ,ν 0
τ0 Lp (X)
τ 0 ∈Ik0 ν 0 =1
n X
∞ h i
k0 d q∧1
0 0
≤ C 2−k d 2−k α [1 + k 0 ]2 r
k0 =1

h  X N (k0 ,τ 0 ,M) r i qr o q1
X 0 0 α 0 0
× M µ(Qkτ 0 ,ν )− d |Dk0 (f )(yτk0 ,ν )|χQk0 ,ν 0
τ0 Lp (X)
τ 0 ∈Ik0 ν 0 =1

n X
∞ N (k0 ,τ 0 ,M) h iq o q1
X X 0 0 α 0 0
≤ C µ(Qkτ 0 ,ν )− d |Dk0 (f )(yτk0 ,ν )|χQk0 ,ν 0
τ0 Lp (X)
k0 =1 τ 0 ∈Ik0 ν 0 =1

≤ Ckf kFpα,q (X) ,

d
where d+α < r < min(p, q, 1).
From (2.6), the Hölder inequality for p > 1 and (a + b)p ≤ ap + bp for p ≤ 1,
the Lemma A.2 in [7], it follows that

nZ X
∞ X N (l,τ,M)
X h
B3 ≤ C χQl,ν
τ
(x) 2lα (1 + l)2−l
l=1 τ ∈Il ν=1
N (0,τ 0 ,M) iq  pq o p1
X X 1
×  d+ m Q0,ν
0 (|D0 (f )|) dµ(x)
1 + ρ(x, yτ0,ν
0 τ0
τ 0 ∈I0 ν 0 =1 0 )
nZ X
∞ h  X N (0,τ 0 ,M)
X
lαq q −lq
≤ C 2 (1 + l) 2 M mQ0,ν 0
τ0
l=1 τ 0 ∈I0 ν 0 =1
r i rq  pq o p1
(|D0 (f )|)χQ0,ν 0 (x) dµ(x)
τ0

nZ h  X N (0,τ 0 ,M)
X r i pr o p1
≤ C M mQ0,ν 0 (|D0 (f )|)χQ0,ν 0 (x) dµ(x)
τ0 τ0
τ 0 ∈I0 ν 0 =1

≤ Ckf kFpα,q (X) ,

p
where we used the L r (X) boundedness of Hardy-Littlewood maximal functions.
This proves Theorem A. Proof of Theorem B. The main difference of proof
between Theorem B and Theorem A is that we should replace Lemma 2.2 by
Lemma 2.3. We leave the details to the reader.
136 Yanchang Han

References

1. M. Christ, A T (b) theorem with remarks on analytic capacity and the Cauchy
integral, Colloq Math. 60/61 (1990), 601-628.
2. R. R. Coifman and G. Weiss, Analyse Harmonique Noncommutative Sur Certains
Espaces Homogenes, Lecture Notes in Math, 242, Springer-Verlag, Berlin and New
York, 1971.
3. G. David and J. L. Journé, A boundedness criterion for generalized Calderón-
Zygmund operators, Ann. of Math. 120 (1984), 371-397.
4. Y. C. Han, New characterizations of inhomogeneous Besov and Triebel-Lizorkin
spaces over spaces of homogeneous type, Submitted.
5. D. G. Deng and Y. S. Han, T 1 Theorem for Besov and Triebel-Lizorkin spaces,
Science in China Ser.A Math. 34 (2004), 657-664.
6. D. G. Deng, Y. S. Han and D. C. Yang, Inhomogeneous Plancherel-Pôlya type
inequality on spaces of homogeneous type and their applications, Communications
in Contemporary Mathematics, 6 (2004), 221-243.
7. C. Feferman and E. M. Stein, Some maximal inequalities, Amer. J. Math., 93
(1971), 107-115.
8. M. Frazier and B. Jawerth, A discrete transform and decompositions of distribu-
tion spaces, J. Funct. Anal. 93 (1990), 34-170.
9. Y. S. Han, Plancherel-Pôlya type inequality on spaces of homogeneous type and
its applications, Proc. Amer. Math. Soc., 126 (1998), 3315-3327.
10. Y. S. Han, S. Z. Lu and D. C. Yang, Inhomogeneous Besov and Triebel-Lizorkin
space on spaces of homogeneous type, Approx. Theory Appl. 15 (1999), 37-65.
11. Y. S. Han, S. Z. Lu and D. C. Yang, Inhomogeneous discrete Calderón reproducing
formulas for spaces of homogeneous type, J. Fourier Anal. Appl., 7 (2001), 571-
600.
12. Y. S. Han and E. T. Sawyer, Littlewood-Paley theory on spaces of homogeneous
type and classical function spaces, Mem. Amer. Math. Soc., 110 (1994), 1-126.
13. Y. S. Han and D. C. Yang, New characterizations and applications of inhomo-
geneous Besov and Triebel-Lizorkin space on spaces of homogeneous type and
fractals, Dissertationes Math. (Rozprawy Mat.) 403 (2002), 1-102.
14. P. G. Lemarie, Continuite sur les espaces de Besov des operateurs definis par des
integrales, Ann. Inst. Fouier(Grenoble), 4 (1985), 175-187.
15. R. A. Macias and C. Segovia, Lipschitz function on spaces of homogeneous type,
Adv. Math. 33 (1979), 257-270.
16. Y. Meyer, Wavelets and Operators, (Cambridge Studies in Advanced Mathematics:
37), Cambridge University Press, 1992.
17. D. C. Yang, T 1 Theorem on Besov and Triebel-Lizorkin spaces on spaces of homo-
geneous type and their applications. Zeitschrift für Analysis ihre Anwendungen,
22 (2003), 53-72.
Vietnam Journal of Mathematics 38:3(2009) 137-146    
!
"# $ % & "# $ ' ()
© VAST 2009

On Oscillation, Convergence and

Boundedness of Solutions of Some Nonlinear
Difference Equations with Multiple Delay ?

Dinh Cong Huong and Phan Thanh Nam

Department of Mathematics, Quy Nhon University

170 An Duong Vuong, Quy Nhon, Binh Dinh, Vietnam

Received February 12, 2007

Revised March 8, 2008

Abstract. The oscillation, convergence and boundedness of the solutions of some non-
linear difference equations with multiple delay of the form

xn+1 = G(xn , xn−m1 , . . . , xn−mr ), n = 0, 1, . . .

are investigated, where mi ∈ N0 , ∀i = 1, . . . , r and G is a function mapping Rm+1 to

R.

2000 Mathematics Subject Classification: 39A12.

Key words: Nonlinear difference equation, multiple delay, oscillation, convergence,
boundedness, equilibrium.

1. Introduction

It is well-known that the difference equation

xn+1 = G(xn , xn−m1 , . . . , xn−mr )

includes difference equations

?
This paper was partly supported by CTTDQG 08.09
138 Dinh Cong Huong and Phan Thanh Nam
r
X
xn+1 = λn xn + αi (n)F (xn−mi ), (1)
i=1

and
xn+1 = G(xn , xn−1 , . . . , xn−m ). (2)
In [2, 5], the oscillation of solutions of the difference equation (1) was discussed.
In addition, the convergence and the boundedness of solutions of (1) were also
investigated in [5]. The authors in [1, 4] studied the global attractivity, the local
stability and the existence of positive periodic solutions of some equations, which
are special cases of (2).
Equation (2) includes many discrete versions of the most celebrated de-
lay differential equation for single species. For example, the Mackey-Glass
hematopoiesis model, the Lasota-Wazewska red blood model, and the Nichol-
son’s blowflies model.
Studying the population dynamics has attracted much attention from both
mathematicians and mathematical biologists recently. Many authors have inves-
tigated the extinction, persistence, global stability and the existence of positive
periodic solutions for several population models; see for example [3, 6–9] and the
references therein.
Motivated by the work above, in the present paper, we aim to study the
oscillation as well as nonoscillation of (1) and investigating the convergence,
boundedness of (2).

2. The Oscillation

Consider the difference equation (1), for n ∈ N, n > a for some a ∈ N, where
r, mi > 1, 1 6 i 6 r are fixed positive integers ;{λn }n , {αi (n)}n are sequences of
numbers and the function F is defined on R. Recall that, the solution {xn }n>a
of (1) is called oscillatory if for any n1 > a there exists n2 > n1 such that
xn2 xn2 +1 6 0. The difference equation (1) is said to be oscillatory if all its
solutions are oscillatory. The solution {xn }n>a of (1) is called nonoscillatory if it
is eventually positive or negative, i.e. there exists a n1 > a such that xn xn+1 > 0
for all n > n1 .

Theorem 1. Assume that λn = 1, ∀n ∈ N; αi (n) > 0, n ∈ N, 1 6 i 6 r;

−F (x)
xF (x) < 0, ∀x 6= 0; sup = M > 0. Then, (1) has a nonoscillatory
x6=0 x
solution if the following holds
r
X ∗
1 mm∗
sup αi (n) 6
n
i=1
M (m∗ + 1)m∗ +1

where m∗ = min{m1 , m2 , . . . , mr }, m∗ = max{m1 , m2 , . . . , mr }.

On Oscillation, Convergence and Boundedness of Solutions... 139
xn
Proof. Setting vn = and dividing (1) by xn , we obtain
xn+1

1 hXr m
F (xn−mi ) Yi i
=1+ αi (n) vn−` , n ∈ N,
vn i=1
xn−mi
`=1

or
r
X h F (x iY
mi
n−mi )
vn−1 = 1 − αi (n) − vn−` , n ∈ N. (3)
i=1
xn−mi
`=1

We shall prove that the equation (3) has a positive solution. Indeed, we define
m∗ + 1
v−1 = v−2 = · · · = v−m∗ = .
m∗
We have
n r
X h F (x iY
mi o−1
−mi )
v0 = 1 − αi (0) − v−` .
i=1
x−mi
`=1

Since
r
X h F (x iY
mi Xr mi
Y
−mi )
06 αi (0) − v−` 6 M αi (0) v−`
i=1
x−mi i=1
`=1 `=1
r
X  m + 1 mi
∗
=M αi (0)
i=1
m ∗
r
X  m + 1 m∗
∗
6M αi (0)
i=1
m ∗

1 m∗
m∗  m + 1  m∗
∗
6M·
M (m∗ + 1)m +1 ∗
m∗
1
= < 1,
m∗ + 1
Pr h F (x i
−mi ) Qmi
we obtain 1 > 1 − i=1 αi (0) − `=1 v−` > 0 and therefore v0 > 1.
x−mi
m∗ + 1
We can check that v0 6 . So, by (3) we have
m∗
n Xr h F (x iY
mi o−1 m + 1
1−mi ) ∗
v1 = 1 − αi (1) − v1−` 6
i=1
x1−mi m ∗
`=1

m∗ + 1
and now by induction 1 < vn 6 for all n = 2, 3, . . . so that {vn } is a
m∗
positive solution of (3). Next, we define
 m + 1 m∗ −i xn−1
∗
xi−m∗ = , 0 6 i 6 m∗ , xn = , n = 1, 2, . . . ,
m∗ vn−1
140 Dinh Cong Huong and Phan Thanh Nam

it follows that {xn } is a nonoscillatory solution of (1).

Example 2. Consider the difference equation

mm
xn+1 = xn + (−xn−m ). (4)
(m + 1)m+1

It is clear that this equation is a particular case of (1), where λn = 1, αi (n) =

1 mm
, ∀n ∈ N, 1 6 i 6 r, mi = m, 1 6 i 6 r and F (x) ≡ −x. It is
r (m + 1)m+1
easy to check that the assumptions of the Theorem 1 are satisfied. If we put
 mm i−m
xi−m = β, 1 6 i 6 m, β 6= 0, the solution of (4) is
(m + 1)m+1
 mm n
xn = β, n ∈ N,
(m + 1)m+1

which is nonoscillatory.

Theorem 3. Assume that λn = 1, ∀n ∈ N; αi (n) > 0, n ∈ N, 1 6 i 6 r;

xF (x) < 0, −F (x) > x, ∀x 6= 0. Then, (1) is oscillatory if the following inequal-
ity holds
n−1
X ∗
1 m∗ m
lim inf ∗ αm∗ (`) >
n→∞ m (m∗ + 1)m∗ +1
`=n−m
∗

∗
where m = max{m1 , m2 , . . . , mr }.

Proof. The proof of the Theorem 2 can be obtained similarly as the proof of
Theorem 3, in [2], so we omit it here.

Theorem 4. Assume that λn = λ > 1, ∀n ∈ N; αi (n) > 0, n ∈ N, 1 6 i 6 r;

xF (x) < 0, ∀x 6= 0 and there exists i0 ∈ {1, 2, . . . , r} such that
X 1
αi (`) = ∞. (5)
λ` 0
`∈N

Suppose further that, if |u| > c then |F (u)| > c1 where c and c1 are positive
constants. Then, every solution {xn }n of (1) is either oscillatory or
xn
lim = 0.
n→∞ λn

Proof. Let {xn }n be a nonoscillatory solution of (1). Suppose that {xn }n is

an eventually positive solution. Then there is n1 ∈ N such that xn > 0 and
xn−mi > 0 for all n > n1 and i = 1, 2, . . . , r. Since
xn+1 xn 1
− n = n+1 (xn+1 − λxn ) 6 0, ∀n > n1 ,
λn+1 λ λ
On Oscillation, Convergence and Boundedness of Solutions... 141
xn
we have { n }n>n1 is nonincreasing for all n > n1 . Therefore, there exists
λ
xn xn
lim . Putting β = lim n , we shall show β = 0. Suppose β > 0, then
n→∞ λn n→∞ λ
there exists n2 > n1 such that

xn > βλn , ∀n > n2 .

Putting n3 = n2 + mi0 , where i0 ∈ {1, 2, . . . r}, we get

xn−mi0 > βλn−mi0 > β, ∀n > n3

and by hypotheses, there exists a positive constant β1 such that

|F (xn−mi0 )| = −F (xn−mi0 ) > β1 , ∀n > n3 .

This implies
1 1
αi0 (n)F (xn−mi0 ) 6 −β1 αi0 (n).
λn+1 λn+1
On the other hand, from (1) we have
xn+1 xn 1
− n 6 n+1 αi0 (n)F (xn−mi0 ).
λn+1 λ λ
Hence
xn+1 xn 1 xn β1 1
6 n − β1 n+1 αi0 (n) = n − αi (n), ∀n > n3
λn+1 λ λ λ λ λn 0
or
n−1
xn xn3 β1 X 1
6 − αi (`), ∀n > n3 .
λn λn3 λ λ` 0
`=n3

But, in view of (5) this leads to a contradiction to our assumption that xn > 0
eventually. The case xn < 0 eventually can be considered similarly.

Theorem 5. If the given hypothesis on the parameter λ in Theorem 3 is replaced

by 0 < λ < 1, then every solution {xn }n of (1) is either oscillatory or
xn
lim = 0.
n→∞ n

Proof. Let {xn }n be a nonoscillatory solution of (1). Suppose that {xn }n is an

eventually positive solution. We have
xn+1 xn
6 n, ∀n ≥ n1 .
λn+1 λ
This yields
xn n+1
xn+1 6 λ = λxn < xn because λ ∈ (0, 1)
λn
and
142 Dinh Cong Huong and Phan Thanh Nam
xn+1 xn xn
< < , ∀n > n1 .
n+1 n+1 n
xn
Hence, { xnn }n is nonincreasing for all n > n1 and therefore lim = β exists.
n→∞ n
We can prove β = 0 similarly as in the proof of Theorem 3. Thus Theorem 4 is
proved.

Theorem 6. Assume that λn = λ > 1, ∀n ∈ N; αi (n) > 0, n ∈ N, 1 6 i 6 r;

xF (x) < 0, ∀x 6= 0 and there exist i0 ∈ {1, 2, . . . , r} and L > 0 such that
1
|F (u)| > L, ∀u ∈ R and Lαi0 (n) mi0 +1 > 1, ∀n ∈ N.
λ
Then (1) is oscillatory.
xn
Proof. Let {xn }n be as in Theorem 3 so that { n }n>n1 is nonincreasing for all
λ
n > n1 . Thus, for all n > n1 we have

xn−mi0 > xn λ−mi0

and
r
X
xn+1 = λxn + αi (n)F (xn−mi )
i=1
6 λxn + αi0 (n)F (xn−mi0 ),
6 λxn − αi0 (n)Lxn−mi0 ,
6 λxn − αi0 (n)Lxn λ−mi0 ,
= λxn [1 − αi0 (n)Lλ−mi0 −1 ],
1
= λxn [1 − Lαi0 (n) mi +1 ] 6 0.
λ 0
This contradicts our assumption.

3. Convergence and Boundedness

Consider the difference equation (2), where n ∈ N, x−m , x−m+1 , . . . , x0 are pos-
itive initial values and the function

G(z0 , z1 , . . . , zm ) : R+ × . . . × R+ → R+ .

We give conditions under which every solution of this equation is convergent

or bounded. First of all we have

Lemma 1. If λ0 + λ1 + λ2 + · · · + λm < 1 then there exists a number s > 1 such

that
On Oscillation, Convergence and Boundedness of Solutions... 143

λ0 s + λ1 s2 + λ2 s3 + · · · + λm sm+1 < 1.

Lemma 2. Let {βn }n be a sequence which satisfies the following relations:

β0 = β−1 = · · · = β−m = 1,

βn+1 = λ0 βn + λ1 βn−1 + · · · + λm βn−m .

If P := λ0 + λ1 + λ2 + · · · + λm > 1 where λi ≥ 0, then βn > 1, ∀n ∈ N0 and βn
is monotone increasing for n ∈ N0 .
Pm Pm
Theorem 3. Assume that G(z0 , z1 , . . . , zm ) 6 i=0 λi zi and i=0 λi < 1.
Then every solution of (2) converges to zero.
Pm
Proof. Since G(z0 , z1 , . . . , zm ) 6 i=0 λi zi , for a positive number a > 1 we get

axn+1 = aG(xn ,...,xn−m ) 6 aλ0 xn aλ1 xn−1 . . . aλm xn−m .

Put yn = axn . Clearly

yn+1 6 [yn ]λ0 . [yn−1 ]λ1 . . . [yn−m ]λm

and yn ≥ 1. Hence, η = max{y−m , y−m+1 , . . . , y0 } ≥ 1. Using Lemma 1, we can

prove the following estimations by induction
−n
yn+1 6 η s , n ∈ N0 , (6)

where s was given in Lemma 1. For n = 0, we have

−0
y1 6 [y0 ]λ0 , [y−1 ]λ1 . . . [y−m ]λm 6 η λ0 +λ1 +...+λm < η 1 = η s .

Assume that (6) holds for the steps 1, 2, . . . , n, we estimate the solution at step
n + 1 as follows

yn+1 6 [yn ]λ0 · [yn−1 ]λ1 . . . [yn−m ]λm

−(n−1) −(n−2) −(n−m+1)
6 ηs .λ0
.η s .λ1
. . . ηs .λm

−n
.(λ0 s+λ1 s2 +λ2 s3 +...+λm sm+1 )
= ηs
−n
6 ηs .

This implies lim yn 6 η 0 = 1. Since yn ≥ 1 for all n, we have lim yn = 1,

n→∞ n→∞
hence lim xn = 0. The proof is complete.
n→∞

Assume that equation (2) has a unique positive equilibrium x. We have a

sufficient condition for convergence to x.
144 Dinh Cong Huong and Phan Thanh Nam

Corollary 4. If G(z0 , z1 , . . . , zm ) satisfies P

Lipschitz condition in every variable
m
zi with Lipschitz factors Li which satisfy i=0 Li < 1, then every solution of
(2) is convergent to the positive equilibrium x.

Proof. We have

|xn+1 − x| = |G(xn , xn−1 , . . . , xn−m ) − G(x, x, . . . , x)|

6 |G(xn , xn−1 , . . . , xn−m ) − G(x, xn−1 , . . . , xn−m )|
+ |G(x, xn−1 , . . . , xn−m ) − G(x, x, xn−2 , . . . , xn−m )|
···
+ |G(x, x, . . . , x, xn−m ) − G(x, x, . . . , x)|
6 L0 |xn − x| + L1 |xn−1 − x| + . . . + Lm |xn−m − x|.

Putting yn = |xn − x|, we have

yn+1 6 L0 yn + L1 yn−1 + . . . + Lm yn−m .

By Theorem 6, the proof is complete.

Remark 5. In the case of

m
X
G(xn , xn−1 , . . . , xn−m ) = λn xn + αi (n)F (xn−i ),
i=1
Pm
where αi (n) ≥ 0, i=1 αi (n) = 1, ∀n ∈ N and F : [0, ∞) → [0, ∞) is a continu-
ous function, applying Theorem 6 to equation (2), we obtain some convergence
results presented in [5, 6].

Under converse conditions, the following theorem gives a sufficient condition

for the non-convergence to zero of the solutions of (2).
Pm Pm
Theorem 6. Assume that G(z0 , z1 , . . . , zm ) ≥ i=0 λi zi and i=0 λi > 1.
Then, every solution {xn } of (2) satisfies

lim inf xn > 0.

n→∞

Proof. As in the proof of Theorem 6, we put yn = axn . Then we have

yn+1 ≥ [yn ]λ0 . . . [yn−1 ]λ1 . . . [yn−m ]λm

and θ = min{y0 , y−1 , . . . , y−m } > 1. We prove yn ≥ θβn by induction.

Clearly, y1 ≥ [y0 ]λ0 ˙[y−1 ]λ1 . . . [y−m ]λm ≥ θλ0 +λ1 +λ2 +...+λm = θβ1 .
Assuming that yn ≥ θβn for the steps 1, 2, . . . , n, we have
On Oscillation, Convergence and Boundedness of Solutions... 145

yn+1 ≥ [yn ]λ0 .[yn−1 ]λ1 . . . [yn−m ]λm

≥ θλ0 βn .θλ1 βn−1 . . . θλm βn−m
= θλ0 βn +λ1 βn−1 +...+λm βn−m
= θβn+1 .

By Lemma 2, we get yn+1 > θβn+1 > θβ1 = θP , ∀n ∈ N0 . This yields xn+1 >
P. loga θ > 0. Hence, lim inf xn > P. loga θ > 0.
n→∞

Definition 7. A solution {xn }n of (2) is called persistent if

0 < lim inf xn 6 lim sup xn < ∞.

n→∞ n→∞

The following theorem gives a sufficient condition for the persistence of (2).

Theorem 8. Assume that

G(x0 , x1 , . . . , xm ) = H(x0 , x1 , . . . , xm ; x0 , x1 , . . . , xm )

where
H(x0 , x1 , . . . , xm ; y0 , y1 , . . . , ym ) : [0, ∞)2(m+1) → [0, ∞)
is a continuous function, increasing in xi and decreasing in yi and

H(x0 , x1 , . . . , xm ; y0 , y1 , . . . , ym ) > 0

if xi , yi > 0. Suppose further that

H(x0 , x1 , . . . , xm ; y0 , y1 , . . . , ym ) 1
lim sup < ,
xi ,yi →∞ x0 + x 1 + . . . + x m m +1

H(x0 , x1 , . . . , xm ; y0 , y1 , . . . , ym ) 1
lim inf > .
xi ,yi →0+ x0 + x1 + . . . + xm m+1
Then every solution {xn }∞
n=−m of (2) is persistent.

Proof. The proof of this theorem can be obtained similarly as the proof of The-
orem 2 in [6].

4. Conclusion

New results for oscillation or nonoscillation of the difference equation (1) and
the extensive results for convergence and boundedness of a class of general dif-
ference equations (2) are given in this paper. Note that, some results in [5, 6]
are particular cases of Theorem 6 and Theorem 8.
146 Dinh Cong Huong and Phan Thanh Nam

Acknowledgements The authors would like to thank the referees for useful com-
ments, which improve the presentation of this paper.

References

1. Karakostas-Stevic G. L. Karakostas and S. Stevic, On the recursive sequence

x
xn+1 = α + f (xn ,···n−k,xn−k+1 )
, Demonstratio. Math. 3 (2005), 595-610.
2. G. Ladas, Ch. G. Philos and Y. G. Sficas, Sharp conditions for the oscillation of
delay differerence equations, J. App. Math. Simulation 2 (1989), 101-111.
3. V. Tkachenko and S. Trofimchuk, Global stability in difference equations satisfying
the generalized Yorke condition, J. Math. Anal. Appl. 303 (2005), 173-187.
4. H. EL-Metwally, E. A. Grove and G. Ladas, A Global Convergence Result with
Applications to Periodic Solutions, J. Math. Anal. Appl. 245 (2000), 161-170.
5. Dinh Cong Huong, On the asymptotic behaviour of solutions of a nonlinear differ-
ence equation with bounded multiple delay, Vietnam J. Math. 34 (2006), 163-170.
6. Dang Vu Giang and Dinh Cong Huong, Extinction, Persistence and Global sta-
bility in models of population growth, J. Math. Anal. Appl. 308 (2005), 195-207.
7. Dang Vu Giang and Dinh Cong Huong, Nontrivial periodicity in discrete delay
models of population growth, J. Math. Anal. Appl. 305 (2005), 291-295.
8. S. H. Saker and S. Agarwal, Oscillation and global attractivity in a periodic Nichol-
son’s blowfilies model, Math. Comput. Mod. 35 (2002), 719-731.
9. S. Saker, Oscillation and global attractivity in hematopoiesis model with delay
time, Appl. Math. Comput. 136 (2003), 241-250.
Vietnam Journal of Mathematics 38:3(2009) 147-156 * +,-./0 1 234./5
26
78 9 : ; 78 9 < =>
© VAST 2009

?
On Harada Rings and Serial Artinian Rings

Thanakarn Soonthornkrachang1 , Phan Dan2

Nguyen Van Sanh3 , and Kar Ping Shum4

1,3
Department of Mathematics, Mahidol University
Bangkok 10400, Thailand
2
Department of Mathematics
University of Transport of Ho Chi Minh City, Vietnam
4
Department of Mathematics
The University of Hong Kong, Hong Kong, China (SAR)

Received November 18, 2007

Abstract. A ring R is called a right Harada ring if it is right Artinian and every
non-small right R-module contains a non-zero injective submodule. The first result in
our paper is the following: Let R be a right perfect ring. Then R is a right Harada
ring if and only if every cyclic module is a direct sum of an injective module and a
small module; if and only if every local module is either injective or small. We also
prove that a ring R is QF if and only if every cyclic module is a direct sum of a
projective injective module and a small module; if and only if every local module is
either projective injective or small. Finally, a right QF-3 right perfect ring R is serial
Artinian if and only if every right ideal is a direct sum of a projective module and a
singular uniserial module.

2000 Mathematics Subject Classification: 16D50, 16D70, 16D80.

Key words: Harada ring, Artinian Ring, Small module, Co-small module.
?
Partially supported by UGC grant (HK) #2160187.
148 Authors Suppressed Due to Excessive Length

1. Introduction and Preliminaries

Throughout this paper, all rings are associative rings with identity and all right
R-modules are unitary. For a right R - module M, we denote by E(M ), J(M ) and
Z(M ) the injective hull, radical and the singular submodule of M , respectively.
Especially, J(R) is the Jacobson radical of the ring R. A right R-module M is
called uniserial if the lattice of its submodules is linear. Call M a serial module
if it is a direct sum of uniserial modules. A ring R is right serial if RR is serial
as a right R-module. A ring R is serial if it is right and left serial. Call a ring R
serial Artinian if it is serial and two-sided Artinian.
Call M a small module if M is small in E(M ) otherwise, we call it a non-
small module. Dually, a right R-module M is called a co-small module if for
any epimorphism f from P to M with P is projective, ker(f ) is essential in P.
A non-cosmall module is defined as a not co-small module. A right R-module
is a local module if it has the greatest proper submodule. A ring R is called a
Quasi-Frobenius ring (briefly QF-ring) if it is right self-injective right Artinian.
Call a ring R a right QF-3 ring if E(RR ) is projective (see [14, 21]).
Manabu Harada [9, 10, 11] has studied some generalizations of QF-rings by
introducing the following two conditions:
(*) Every non-small right R-module contains a non-zero injective submodule;
(**) Every non-cosmall right R-module contains a non-zero projective direct
summand.
It should be noted that right perfect rings with (*) and semiperfect rings with
(**) are characterized in terms of ideals in [11, 12, 13]. Oshiro [16] gave some
characterizations of these kinds of rings and introduced the definitions of right
Harada and right co-Harada rings as follows.
A ring R is called a right Harada ring if it is right Artinian and (*) holds. A
ring R is a right co-Harada ring if it satisfies (**) and ACC on right annihilators.
In this paper, we give some characterizations of right Harada rings and serial
Artinian rings. In Sec. 1, we recall some well-known results which will be used
in this paper. We characterize the classes of right Harada rings and QF-rings by
right perfect rings and cyclic or local modules. Section 3 is concerned with serial
Artinian rings. For convenience, we list some well-known results here to use in
this paper.

Theorem 1 (Theorem A). ([16, Theorem 2.11]) For a ring R, the following
conditions are equivalent:

1. (1) R is a right Harada ring;

2. (2) R is right perfect and satisfies the condition that the family of all projec-
tive modules is closed under taking small covers, i.e., for any exact sequence
ϕ
P →→ E → 0, with E is injective and P is projective, ker(ϕ) is small in P ;
3. (3) Every right R-module is a direct sum of an injective module and a small
module;
On Harada Rings and Serial Artinian Rings 149

4. (4) Every injective module is a lifting module.

Theorem 2 (Theorem B). ([11, Theorem 3.6]) Let R be a semiperfect ring

and {ei }ni=1 ∪ {fj }m
j=1 a complete set of orthogonal primitive idempotents of R,
where each ei R is non-small, i = 1, . . . n, and fj R is small, j = 1, . . . m. Then
(∗∗) holds if and only if:

1. (a) n ≥ 1 and ei R is injective, i = 1, . . . , n;

2. (b) For each fj , there exists ei such that fj R can be embedded in ei R;
3. (c) For each fj , there exists an integer nj such that ei J t is projective for
0 ≤ t ≤ ni and ei J ni +1 is a singular module, where J = J(R);
4. Further in this case, it is shown that every submodule ei B in ei R is either
contained in ei J ni +1 or equal to some ei J i .

Theorem 3 (Theorem C). ([16]) Let R be a right Artinian ring. Then R is a

serial ring if and only if for any primitive idempotent f of R, the injective hull
E(f R) is a uniserial module.

2. Right Harada Rings

In this section, we will prove that the classes of right Harada rings and QF-rings
are both characterized by perfect rings and cyclic (or local) modules. The proof
of the following lemma is routine and is therefore omitted.
P
Lemma 1. If {Xi , i = 1, . . . , n} is a family of small modules, then X = Xi
i
is also small.

Proposition 2. Let R be a right perfect ring. Then the following conditions are
equivalent:

1. (1) The condition (∗) holds;

2. (2) Every non-small cyclic module contains a non-zero injective module;
3. (3) Every local module is either injective or small.

Proof. (1) ⇒ (2) is clear. We now prove (2) ⇒ (3). Let M be a local module.
Since R is right perfect, it follows from [5, Proposition 18.23] that there exists
a primitive idempotent g of R such that M ∼ = gR/G, G ⊂ gR. Hence M is an
indecomposable cyclic module. Therefore M either injective or small by (2).
(3) ⇒ (1) Let M be a non-small module and E = E(M ). Since R is right
perfect, it follows that M 6⊂ EJ (since EJ is small in E). Take any m ∈ M \ EJ.
Then mR is a non-small module. Let {ei |i = 1, 2, . . . , n} be an orthogonal system
Pn
of primitive idempotents of R. Then mR = mei R. By Lemma 1, there exists
i=1
an idempotent ei such that mei R is non-small. Since mei R 6= 0 and mei R ∼
=
150 Authors Suppressed Due to Excessive Length

ei R/H, H ⊂ ei R, the module mei R is local. It follows from (3) that mei R is
injective. Thus M contains a non-zero mei R, hence (∗) holds.

Theorem 3. Let R be a right perfect ring. Then the following conditions are
equivalent:
1. (1) R is a right Harada ring;
2. (2) Every cyclic module is a direct sum of an injective module and a small
module;
3. (3) Every local module is either injective or small.

Proof. (1) ⇒ (3). This part follows from Theorem A.

(2) ⇒ (3) By the same argument as that of (2) ⇒ (3) in the proof of Propo-
sition 2.
(3) ⇒ (1) By Proposition 2, R satisfies (∗). Hence by [12, Theorem 5], R is a
right Artinian ring, and therefore it is a right Harada ring.

Theorem 4. Let R be a right perfect ring. Then the following conditions are
equivalent:

1. (1) R is a QF-ring;
2. (2) Every cyclic module is a direct sum of a projective injective module and a
small module;
3. (3) Every local module is either projective injective or small.

Proof. (1) ⇒ (2) Since R is QF, it is a right Harada ring by [16, Theorem 4.3].
For a right R-module M, we have M = I ⊕ S, with I is injective and S is small
by Theorem A. It is clear that I is projective, proving (2).
(2) ⇒ (3) It follows from Proposition 2.
(3) ⇒ (1) It suffices to show that every injective module is projective. Since
R is a right perfect ring satisfying (3), in view of Theorem 3, R is a right Harada
L
ring. Let Q be an injective module. Then Q has a decomposition Q = Qi ,
I
where each Qi is a non-zero indecomposable module. We will show that each Qi
is projective for each i ∈ I. Since R is right Artinian, Qi contains a maximal
submodule Q0i (see [1, Theorem 28.4]). Let X be a proper submodule of Qi such
that X 6⊂ Q0i . Then Q0i +X = Qi = E(Q0i ). Hence Q0i is a non-small module. Since
R is a right Harada ring, it implies that Q0i contains a proper direct summand,
a contradiction. Thus Q0i is the greatest proper submodule of Qi , i.e., Qi is a
local module. Hence, by (3), Qi is projective. It follows that Q is a projective
module, proving that R is QF.

3. Characterizations of Serial Artinian Rings

First we recall a remark due to M. Harada in [11] as folllows:

On Harada Rings and Serial Artinian Rings 151

Remark 1. Let R be a right perfect ring. Then R has a decomposition of the

form
M n m
M
R= ei R ⊕ fj R
i=1 m=1

where {ei , i = 1, . . . , n} ∪ {fj , j = 1, . . . , m} is the set of mutually orthogonal

idempotents with each ei R non-small and fj R being a small module, and always
we have n ≥ 1.

Lemma 2. Let R be a right perfect ring and M a uniserial right R-module. Then
every submodule of M is cyclic and hence M is a Noetherian module.

Proof. Let N be a non-zero submodule of M. Since R is right perfect, it follows

from [1, Theorem 28.4] that N contains a maximal submodule N 0 . Take any
x ∈ N \ N 0 . Then xR 6⊂ N 0 . Since M is uniserial, we must have N 0 ⊂ xR. Hence
N = xR, proving that M is a Noetherian module.

Lemma 3. Let R be a right perfect ring. If E(gR) is uniserial for any primitive
idempotent g of R, then R is serial Artinian.

Proof. By Theorem C, it is enough to show that R is right Artinian. Since R is

right perfect, we can write
Mn
R= gi R
i=1

where {gi , i = 1, . . . n} is a system of orthogonal primitive idempotents. For

each i, E(gi R) is uniserial, therefore by Lemma 6, it is Noetherian and hence R
is right Noetherian. Combining with the assumption that R is right perfect, it
follows that R is right Artinian by [1, Corollary 15.23] and this completes our
proof.

Theorem 4. Let R be a right perfect ring. Then the following conditions are
equivalent:

1. (1) R is a serial Artinian ring;

2. (2) Every cyclic module is a direct sum of an injective module and a uniform
small module;
3. (3) Every local module is either injective or uniform small.

Proof. (1) ⇒ (2) Let R be a right perfect serial ring. Then by [16, Theorem 4.5],
R is a right Harada ring. For a cyclic module M, we have M = I ⊕ S, where LI is
injective and S is small (see Theorem A). Since R is serial Artinian, S = Sj
j
with Sj uniserial. It follows from [16, Lemma 1.1] that Sj is small, proving (2).
(2) ⇒ (3) by Proposition 2.
(3) ⇒ (1) Supppose that R has a decomposition as in Remark 5. Then by (3),
ei R is injective for each ei ∈ {ei , i = 1, . . . , n}. Hence by [11, Theorem 1.3], the
152 Authors Suppressed Due to Excessive Length

ring R is right QF-3. It follows that

LE(fk R) is projective for each fk ∈ {fj , j =
1, . . . , m}. Therefore E(fk R) = eij R, eij R ∼ e
= i R, I ⊂ {1, . . . .n}. Since
i∈I,j∈J
fk R is uniform by (3), it implies E(fk R) ∼ = ei R for some idempotent ei . Next,
we will show that ei R is uniserial for all ei , i = 1. . . . , n.
Let U, V be non-zero submodules of eR, where e ∈ {ei , 1 ≤ i ≤ n}. Put
I = U ∩ V. Assume that I 6= U and I 6= V. Then the module B = eR/I is not
uniform, and hence B is not injective because it is indecomposable. Thus the
local module B is neither uniform, nor injective, this contradicts the assumption
(3). Therefore, either I = U or I = V, proving that eR is a uniserial module. In
view of Lemma 7, it follows that R is a serial Artinian ring.
Lemma 5. ([11, 19]) The following statements hold for non-cosmall modules:
1. (1) A right R-module M is non-cosmall if it does not coincide with its singular
submodule;
2. (2) If an R-module M contains a non-zero projective submodule, then it is
non-cosmall.
Lemma 6. Let R be a right QF-3 semiperfect ring. If every uniform principal
right ideal of R is either projective or singular as a right R-module, then R
satisfies two conditions (a) and (b) of Theorem B.
Proof. Since R is semiperfect, it has a decomposition of the form

R = e1 R ⊕ e2 R ⊕ · · · ⊕ en R,

where {ei 1 ≤ i ≤ n} is the set of mutually orthogonal primitive idempotents.

Since R is right QF-3 (i.e., E(RR ) is projective), it follows that there exists at
least one ei such that ei R is injective. Without loss of generality we may assume
that ei R is injective for 1 ≤ i ≤ k and ej R is not injective with k + 1 ≤ j ≤ n.
Take e = ej , k + 1 ≤ j ≤ n. Since E(RR ) is projective, so is E(eR). Hence

t(i)
k M
M
E(eR) = eit R
i=1 t=1

where eit R ∼
= ei R for 1 ≤ t ≤ t(i). Put
t(i)
k M
M
Q= eit R
i=1 t=1

k t(i)
L L
and let α : E(eR) → Q be an isomorphism and πit : eit R −→ eit R the
i=1 t=1
projections. Let F = α(eR) and Fit = πit (F ) ⊂ eit R = ∼ ei R. Then Fit is cyclic.
It is easy to see that Fit is isomorphic to a principal right ideal of R. Moreover,
Fit is uniform. Hence by hypothesis, Fit is either projective or singular for all
pairs (i, t).
On Harada Rings and Serial Artinian Rings 153

SupposeL
that Fit are singular for all pairs (i, t). It follows that F is singular,
since F ⊂ Fit , which is a contradiction to the fact that F ∼ = eR. Therefore
i
there exists a pair (i0 , t0 ) such that Fi0 t0 is non-zero projective. Let p = πi0 t0 |F ,
p
induced by the projection, and consider the exact sequence F →→ Fi0 t0 → 0
with Fi0 t0 projective. Then F = ker p⊕F 0 , for some F 0 ∼ = Fi0 t0 . Since ∼
L F (= eR) is
indecomposable and F 0 6= 0, it implies that ker p = 0, hence F ∩ eit R = 0,
i6=i0 ,t6=t0
because F ⊂ eQ ( i.e., F is essential in Q). Therefore Q = ei0 t0 R. Thus eR ∼ =
F ⊂ ei0 t0 R = ei0 R. Hence R satisfies both conditions (a) and (b) of Theorem B
and our Lemma has been proved.

Lemma 7. ([11, Theorem 3.6]) Let R be a semiperfect ring and e a primitive

idempotent of R such that eR is injective. Suppose that every submodule of eR
is either projective or singular. Then there exists an integer n such that eJ t is
projective for 0 ≤ t ≤ n and eJ n+1 is singular, where J is the Jacobson radical
of R.

Lemma 8. Let R be a right QF-3 right perfect ring and e ∈ R a primitive

idempotent of R such that eR is injective. If every 2-generated right submodule of
eR is either projective or singular, then every submodule of eR is either projective
or singular.

Proof. It is clear that eR is uniform. Let Z(eR) be the singular submodule of

eR. We first prove that eR/Z(eR) is uniserial. Suppose on the contrary that
there are submodules U and V of eR such that Z(eR) ⊂ U ∩ V and U 6⊂
V, V 6⊂ U. Take u ∈ U \ V and v ∈ V \ U. Consider the module X = uR + vR.
Since X is a 2-generated right submodule of eR, by assumption, it is projective
and singular. But both uR and vR are not singular, X must be not singular.
Hence X is projective. Let X1 and X2 be maximal submodules of uR and vR
respectively. Then X1 + vR and X2 + uR are distinct maximal submodules of
X. This contradicts the fact that X is a projective indecomposable module on
a right perfect ring. It would imply that eR/Z(eR) is a uniserial module. We
now show that eR/Z(eR) has finite length. Let X1 be the largest submodule of
eR. Since R is right perfect, it follows from [1, Theorem 28.4] that X1 (here,
X1 = rad(eR)) contains a maximal submodule X2 . Continueing this process, we
get a strictly descending chain

eR ⊃ X1 ⊃ X2 ⊃ · · · ⊃ Xn ⊃ · · · ⊃ Z(eR).

We will prove that this chain is stationary. It is clear that each Xi is a local
module, hence each Xi is an epimorphism image of f R for some primitive idem-
potent f of R. Since R is right QF-3, it follows that f R is uniform, and hence
Xi is projective or singular. But Xi ⊃ Z(eR), and Xi 6= Z(eR), it follows that
each Xi is projective. We claim that for i 6= j, Xi 6∼
= Xj .
Suppose on the contrary that there is an isomorphism ϕ : Xi → Xj . Since
eR is injective, ϕ can be extended to ϕ̄ : eR → eR and ϕ̄ is also an isomor-
phism. From this we obtain eR/Xi ∼ = eR/Xj , and this contradicts the fact that
154 Authors Suppressed Due to Excessive Length

length(eR/Xi ) 6=length (eR/XCj ). Since the representative set of R is finite, it

implies that the chain X1 ⊃ X2 ⊃ . . . must be stationary. Therefore the condi-
tion (c) of Theorem B is satisfied, proving our Lemma.

Theorem 9. Let R be a semiperfect ring. Then the following conditions are

equivalent:
1. (1) (∗∗) holds;
2. (2) R is right QF-3 and every right ideal is a direct sum of a projective module
and a singular module;
3. (3) R is right QF-3 and every uniform right ideal is either projective or sin-
gular.

Proof. (1) ⇒ (2). In view of Theorem B, we see that E(RR ) is projective, i.e., R
is right QF-3. Moreover, R has finite right Goldie dimension. Let B be a right
ideal of R. If B is non-cosmall, then by (**), we have B = B1 ⊕ B10 with B1 is
non-zero and projective.
Again, if B10 is non-cosmall, then B10 = B2 ⊕ B20 , with B2 being non-zero
and projective. Since RR has finite Goldie dimension, we get that Bi is finite
dimensional and therefore after a finite number of steps, we get B = B1 ⊕
· · · ⊕ Bk ⊕ Bk0 , where B1 , . . . , Bk are projective and Bk0 is cosmall, i.e., singular
(Lemma 9). Hence (2) holds.
(2) ⇒ (3) is obvious.
(3) ⇒ (1) by Lemmas 10 and 11.

Theorem 10. Let R be a right perfect ring. The following conditions are equiv-
alent:
1. (1) R satisfies (∗∗);
2. (2) R is right QF-3 and every 2-generated right ideal is a direct sum of a
projective module and a singular module;
3. (3) R is right QF-3 and every uniform 2-generated right ideal is either pro-
jective or singular.

Proof. The proof of (1) ⇒ (2) is similar to that of Theorem 13. The implication
(2) ⇒ (3) is obvious. From Lemmas 10 and 12, it follows that R satisfies three
conditions (a), (b) and (c) of Theorem B. Therefore, R satisfies (**), proving
(3) ⇒ (1).

Theorem 11. The following conditions are equivalent for a right QF-3 semiper-
fect ring R.
1. (1) R is a serial Artinian ring;
L
2. (2) Every right ideal B of R has a decomposition of the form B = B0 Bi ,
1≤i≤n
with B0 projective and each Bi (1 ≤ i ≤ n) being a singular uniserial module
of finite length;
On Harada Rings and Serial Artinian Rings 155

3. (3) Every right ideal of R is either a projective module or a singular uniserial

module of finite length.

Proof. (1) ⇒ (2). Let R be a serial Artinian ring. Then R is a right co-Harada
ring by [16, Theorem 4.5]. Hence for any right ideal B of R, B = B0 ⊕ B 0 , L
where
B0 is projective and B 0 is singular by [16, Theorem 3.18]. Then B 0 = Bi ,
i∈I
where each Bi is uniserial with finite length, proving (2).
(2) ⇒ (3) is obvious.
(3) ⇒ (1). Since R is right QF-3 and semiperfect, it follows from Theorem 13
that R satisfies the condition (**). By applying Theorem B, R has a decompo-
sition n n
M M
R= ei R ⊕ fj R,
i=1 j=1

where ei R is injective and fj R is small. Moreover for each j we have E(fj R) ∼ =

ei R for some i. By Theorem C, in order to prove that R is serial, it suffices to
show that R is right Artinian and each ei R is uniserial, i ∈ {1, 2, . . . , n}. Take
any e ∈ {ei , i = 1, . . . n} and consider the module eR with its singular submodule
Z = Z(e). By Theorem B and (3), Z is a uniserial module with finite length.
Therefore, eR is an Artinian module. On the other hand, again by Theorem
B, we have either eB ⊂ Z or Z ⊂ eB. Since Z and eR/Z are both uniserial,
it follows that eR is a uniserial module. Moreover, eR/Z is of finite length. It
is now clear that R is right Artinian and E(eR) is uniserial for any primitive
idempotent e of R. The proof is now complete.

Theorem 12. Let R be a right QF-3 right perfect ring. The following conditions
are equivalent:

1. (1) R is serial Artinian;

2. (2) Every right ideal of R is a direct sum of a projective module and a singular
uniserial module;
3. (3) Every uniform 2-generated right ideal is either projective or singular unis-
erial.

Proof. The proof of (1) ⇒ (3) is similar to that of Theorem 15 and (2) ⇒ (3) is
obvious. We now prove that (3) ⇒ (1). Clearly, R satisfies (**) by Theorem 14.
By Remark 5, we can write R in the form
n
M m
M
R= ei R ⊕ fj R
i=1 j=1

with properties in the Remark 5.

By Theorem B, we can see that each ei R is injective and for each j, we have
E(fj R) ∼ = ei R for some i = 1, . . . , n. Using Lemma 7, we now show that each
ei R is a uniserial module, i = 1, . . . , n. Put Z = Z(ei R). Since R satisfies (**),
156 Authors Suppressed Due to Excessive Length

in the same way as in Theorem 15, we can see that ei R/Z is a uniserial module
with finite length, and for every right ideal B of R, we have either Z ⊂ ei B or
ei B ⊂ Z. Therfore, it suffices to show that Z is also a uniserial module. We can
suppose that Z 6= 0.
Let U, V be non-zero submodules of Z. If U 6⊂ V and V 6⊂ U, we can take
any u ∈ U \ V and v ∈ V \ U, and consider the module C = uR + vR. Then C
is a uniform 2-generated right ideal of R and C is singular. However C is not
uniserial, since uR 6⊂ vR and vR 6⊂ uR, and this contradicts the hypothesis (3).
Hence, either U ⊂ V or V ⊂ U, proving that Z(eR) is uniserial. The proof is
now complete.

References

1. F. W. Anderson and K. R. Fuller, Rings and Categories of Modules, Graduate

Texts in Math. No.13, Springer-Verlag, New York, Heidelberg, Berlin, 1974.
2. H. Bass, Finitistic dimension and a homological generalization of semiprimary
rings, Trans. Amer. Math. Soc. 95 (1960), 466-488.
3. R. Colby and E. A. Rutters, Jr, Semiperfect QF-3 and p.p rings, Osaka J. Math.
5 (1968), 99-102.
4. R. Colby and E. A. Rutters, Jr, Generalizations of QF-3 algebras, Trans. Amer.
Math. Soc. 153 (1971), 37-386.
5. C.Faith, Algebra II. Ring Theory, Springer-Verlag, 1976.
6. C. Faith and E.A. Walker, Direct sum presentations of injective modules, J.Algebra
5 (1967), 203-221.
7. A. W. Fuller, On direct representations of quasi-injectives and quasi-projectives,
Arch. Math. 20 (1969), 495-502 (Correction 21 (1970), 478).
8. A. W. Goldie, Torsion-free modules and rings, J. Algebra 1 (1964), 268-287.
9. M. Harada, QF-3 and semiprimary p.p rings I, Osaka J.Math. 2 (1965), 357-368.
10. M. Harada, A note on hollow modules, Rev. Un. Matx. Argentina, 28 (1978),
186-194.
11. M. Harada, Non-small modules and non-cosmall modules, Proc. of the 1978 Antw.
Conf., F. van Ostaen, Ed., Marcel Dekker, New York, pp. 669-689.
12. M. Harada, On one-sided QF-2 rings I, Osaka J. Math. 17 (1980), 421-431.
13. M. Harada, On one-sided QF-2 rings II, Osaka J. Math. 17 (1980), 433-438.
14. J. P. Jans, Projective injective modules, Pacific J. Math. 9(1959), 1103-1108.
15. D. Jonah, Rings with minimum condition for principal right ideals have the max-
imum condition for principal left ideals, Math. Z. 113 (1965) 106-112.
16. K. Oshiro, Lifting modules, extending modules and their applications to QF-rings,
Hokkaido Math. J. 13 (1984), 310-338.
17. D. V.Huynh and P. Dan, Some characterizations of right co-H-rings, Math. J.
Okayama Univ. 34 (1992), 165-174.
18. P. Dan, Right perfect rings with the extending property on finitely generated free
modules, Osaka J. Math. 26 (1989), 226-273.
19. M. Rayar, Small and Cosmall Modules, Ph. D. Dissertation, Indiana Univ., 1971.
20. F. L. Sandomierski, Non-singular rings, Proc. Amer. Math. Soc. 19 (1968), 225-
230.
21. R. M. Thrall, Some generalizations of quasi-Frobenius Algebras, Trans. Amer.
Math. Soc. 64 (1984), 173-183.
22. C. I. Vinsonhaler, Orders in QF-3 rings, J. Algebra 14 (1970), 83-90.
Vietnam Journal
of
MATHEMATICS
INSTRUCTIONS FOR AUTHORS
Papers submitted for publication must be written in En-
glish.njh Manuscripts must be in their final form and
typed on one side of the paper only in double-line spac-
ing with wide margins. The author should keep a copy
of the manuscript. Manuscripts can be sent by post (in
triplicate) or by email to one of the following editors:
Nguyen Khoa Son, Editor-in-Chief, Vietnam Academy
of Science and Technology, 18 Hoang Quoc Viet Road,
Cau Giay Str., Hanoi, Vietnam nkson@vast.ac.vn
Pham Ky Anh, Deputy Editor-in-Chief, Hanoi Univer-
sity of Science, 334 Nguyen Trai, Thanh Xuan, Hanoi,
Vietnam anhpk@vnu.edu.vn
Nguyen Thi Hoai Linh, Editorial Secretary, Vietnam
Journal of Mathematics Editorial Office, 70 Tran Hung
Dao Str., Hanoi, Vietnam vjm@isi.ac.vn
All papers will be refereed. If necessary, the author(s)
would be asked to revise the manuscript according to
the referee’s report. The Editorial Board reserves the
right to make reasonable modifications in wording. Pa-
pers should be written in an elegant and clear style.
Prospective authors not proficient in the language are
urged to seek proper assistance in this matter before
submitting papers. If the language used is poor, the
manuscript will be returned to the author.
IMPORTANT: To facilitate communication between
authors, editors and the publisher, papers must be sub-
mitted with full address(es), and telefax numbers and
e-mail address(es) of the corresponding author.
Authors are strongly advised to submit electronically
prepared manuscripts (AMSTeX or LaTeX).
Each paper should be preceded by an abstract in En-
glish (not exceeding 100 words), the primary and sec-
ondary classification numbers according to the 2000
Mathematics Subject Classification, and a short list of
keywords describing the subject of the paper.
On the first page of the manuscript, a short running ti-
tle, not exceeding 60 typewrite strokes including spaces,
should be provided.
Authors are requested to choose mathematical sym-
bols as simply as possible, and place superscripts and
subscripts accurately so that they will not be misun-
derstood. Distinction should be made between easily
confused characters.
Illustrations and diagrams should be submitted on sep-
arate sheets and not included in the text. They should
be of good quality and drawn in India ink using clear
uniform lines. Computer-drawn figures are acceptable
provided they are of comparable quality. Lines and
curves must be smooth. Letters 2mm high are rec-
ommended. Illustrations and diagrams may be reduced
where necessary. The author should indicate in the mar-
gin of the manuscript where illustrations and diagrams
are to be inserted.
Indicators such as ”Definition”, ”Theorem”,
”Lemma”, ”Corollary”, ”Proposition”, etc. are nor-
mally in boldface followed by the formulation in italics,
the end of which should be clearly indicated. The words
”Proof ”, ”Remark”, ”Note”, etc. are in italics with
the formulation in ordinary roman (upright) typeface.
Formulae should be typewritten whenever possible.
Special marking should be explained in a ”Note to the
printer”. Copies produced by matrix printers are not ac-
ceptable unless clearly legible. Letters in formulae are
printed in italics and number in roman unless marked
otherwise. Please indicate clearly the position of expo-
nents and indices. Also, certain symbols can be mis-
taken for other characters and therefore, the intended
expression or meaning should be made clear.
Footnotes, other than those referring to the title of the
paper, should be avoided. If absolutely necessary, they
should be numbered consecutively and placed at the
foot of the page on which they occur (not at the end
of the article).
The list of references at the end of the paper always
be in alphabetical order, numbered consecutively and
include the names and initials of all authors, or edi-
tors, year of publication, and name and location of the
publisher (see examples below). Names of journals and
book series should be abbreviated in accordance with
Zentralblatt fr Mathematik. Whenever possible, please
replace all references to papers accepted for publica-
tion, preprints or technical report by exact name of the
journal as well as the volume, first and last page num-
bers and year, if the article has already been published
or accepted for publication.
Citations in the text should be by numbers in square
brackets, for example [1], or Freed and Melrose [1], re-
ferring to an alphabetically ordered and numbered list.
If a work with more than three author’s or editor’s name
followed by ”et al.” need be given, for example, Komor
et al. [1].
In the list of references, the following examples should
be observed:
Journal article:
1. D. S. Freed and R. B. Melrose, A mode k index
theorem, Invent. Math. 107 (1992), 283-299.
Complete book:
2. J. H. Conway and N. J. Sloane, Sphere Packings,
Lattices and Groups (Grundlehren Math. Wiss. Bd.
290), Springer-Verlag, 1988.
Single contribution in a book:
3. J. Sjöstrand, Microlocal analysis for the periodic
magnetic Schrdinger equations and related questions,
in: Microlocal Analysis and Applications, Montecatini
Terme 1989, J. M. Bony et al. (eds.), Lecture Notes
in Mathematics, Vol. 1495, Springer- Verlag, 1991, pp.
237-332.
Normally, only printer’s errors should be corrected
in the proofs. A charge may be made for extensive
changes, not due to typesetting errors, introduced at
the proofing stage.
Thirty (30) offprints of each paper will be supplied free
of charge. In case of a joint paper, these offprints will
be sent to the corresponding author.
For further information, please contact:
Editorial Office at the above address
 Volume 38, Number 3 June 2009

T1 Theorems for Inhomogeneous Besov and Triebel-Lizorkin Spaces over

Space of Homogeneous Type 125
Yanchang Han

On Oscillation, Convergence and Boundedness of Solutions of Some

Nonlinear Difference Equations with Multiple Delay 137
Dinh Cong Huong and Phan Thanh Nam

On Harada Rings and Serial Artinian Rings 147

Thanakarn Soonthornkrachang, Phan Dan, Nguyen Van Sanh, and
Kar Ping Shum