ACTA MATHEMATICA VIETNAMICA 251

Volume 23, Number 3, 2000, pp. 251-260

PERMANENCE AND POSITIVE BOUNDED SOLUTIONS

OF KOLMOGOROV COMPETING SPECIES SYSTEM

TRINH TUAN ANH, TRAN VAN NHUNG AND LE HONG LAN

Abstract. We study the Kolmogorov equation for n-species. Under certain

conditions, it is shown that the equation is permanent and there exists a
solution defined on whole R whose components are bounded above and below
by positive constants.

1. Introduction

Consider the n-species Kolmogorov-type nonautononous differential system

(1.1) ẋi = xi fi (t, x1 , ..., xn ), i = 1, ..., n,
where fi : R × R+ n → R is uniformly continuous on R × Rn (Rn := {x ∈ Rn :
+ +
xi ≥ 0, i = 1, ..., n}).
A special case of (1.1) is the Lotka-Volterra-type system
h n
X i
(1.2) ẋi = xi bi (t) − aij (t)xj , i = 1, ..., n,
j=1

where the functions bi , aij : R → R are bounded and continuous.

A fundamental ecological question associated with the study of multispecies
population interactions is the long term coexistence of the involved populations.
Such questions arise also in many other situations (see [6]). Mathematically, this
is equivalent to the so-called permanence of the populations. We recall that the
system (1.1) is permanent if there exist positive constants m, M (m ≤ M ) such
that any noncontinuable solution x(.) of (1.1) with x(t0 ) ∈ int(R+ n )-the interior
n
of R+ , is defined on [t0 , +∞) and the following condition is satisfied
(1.3) m ≤ lim inf xi (t) ≤ lim sup xi (t) ≤ M, i = 1, ..., n.
t→+∞ t→+∞

Zanolin [12] and Zhao [13] studied the permanence and the existence of positive
periodic solutions of the systems (1.1) and (1.2) in the periodic case. Some results
on the existence of periodic and almost periodic solution and the stability of the
system (1.2) in the periodic and almost periodic cases were given [2, 3, 4, 9].
Received January 25, 1999; in revised form July 6, 2000.
1991 Mathematics Subject Classification. 13D45.
Key words and phrases. differential system.
This work is supported by the National Basic Research Program.
252 TRINH TUAN ANH, TRAN VAN NHUNG AND LE HONG LAN

In general, if system (1.1) is not periodic, then it may have positive bounded
solutions (defined on whole R). For system (1.2) in the case n = 2, some sufficient
conditions for the existence of a positive bounded solution were given in [1, 10].
In this paper we present a result on the permanence and the existence of
positive and bounded solutions of the system (1.1) and (1.2).

2. Main results

We introduce the following hypotheses:

(H1 ) There exists a positive number α such that Dxi fi (t, x1 , . . . , xn ) ≤ −α,
n , i = 1, . . . , n, where D
(t, x) ∈ R × R+ xi is any Dini derivative in xi ,

(H2 ) inf fi (t, 0) > 0 and fi (t, x) is bounded on the sets of the form R × S, where
t∈R
n (i = 1, . . . , n).
S is any compact subset of R+

(H3 ) fi (t, x1 , . . . , xn ) is decreasing in xj (i, j = 1, . . . , n).

From now on, let R+ := [0, +∞) and θ := (0, . . . , 0) ∈ Rn . For x, z ∈ Rn we set
x ≤ z if xi ≤ zi , i = 1, . . . , n. We denote by C + the set of continuous functions
from R into R which are bounded above and below by positive constants.
The following lemmas were proved in [11].
Lemma 1. Suppose that g : R × R+ → R is continuous such that
(G1 ) g(., 0) ∈ C + ,

(G2 ) There exists a positive α such that Dy g(t, y) ≤ −α for all (t, y) ∈ R × R+ ,

(G3 ) There exists ε > 0 such that g(t, x) is uniformly continuous on R × [0, ε].
Then the problem
(2.1) ẏ = yg(t, y), y(.) ∈ C + ,
has a unique solution y 0 (.). Moreover, we have lim |y 0 (t) − y(t)| = 0 for any
t→+∞
solution y(.) of the equation in (2.1) with y(t0 ) > 0.
Lemma 2. Suppose that g : R × R+ → R is almost periodic in t uniformly for
y ∈ R+ such that
Zω
1
(G∗1 ) lim g(t, 0)dt > 0,
ω→∞ ω
0

(G∗2 ) There exists a positive α such that Dy g(t, y) ≤ −α for all (t, y) ∈ R × R+ .
Then problem (2.1) has a unique solution y 0 (.). Moreover, y 0 (.) is almost periodic
and we have lim |y 0 (t) − y(t)| = 0 for any solution y(.) of the equation in (2.1)
t→+∞
with y(t0 ) > 0.
 PERMANENCE AND POSITIVE BOUNDED SOLUTIONS 253

By Lemma 1, for each i = 1, . . . , n the problem

(2.1i ) ẋi = xi fi (t, 0, ..., 0, xi , 0, ..., 0), xi (.) ∈ C + ,
has a unique solution, say Xi0 (.), which is bounded above and below by positive
constants. From now on we set X 0 (.) = (X10 (.), ..., Xn0 (.)).
Our main results are the following:
Theorem 1. Assume (H1 ), (H2 ), (H3 ). If
(H4 ) inf fi (t, X10 (t), ..., Xi−1
0 0
(t), 0, Xi+1 (t), ..., Xn0 (t)) > 0, i = 1, ..., n,
t∈R

then system (1.1) is permanent and it has at least one solution x∗ (.) = (x∗1 (.), ..., x∗n (.))
defined on whole R, whose components are bounded above and below by pos-
itive constants.
Theorem 2. Suppose that fi (t, x) (i = 1, ..., n) is almost periodic in t uniformly
n and satisfies (H ), (H ) and the following conditions for i = 1, . . . , n,
for x ∈ R+ 1 3
Zω
1
(H2∗ ) lim fi (t, 0) > 0,
ω→∞ ω
0

Zω
1
(H4∗ ) lim fi (t, X10 (t), ..., Xi−1
0 0
(t), 0, Xi+1 (t), ..., Xn0 (t))dt > 0.
ω→∞ ω
0

Then system (1.1) is permanent and it has at least one solution x∗ (.) = (x∗1 (.), ..., x∗n (.))
defined on whole R, whose components are bounded above and below by positive
constants.

Applying Theorem 2 to system (1.2) we get the following

Corollary 1. Suppose that bi (t), aij (t) (i, j = 1, ..., n) are almost periodic
functions such that
Zω
1
(L1 ) inf aii (t) > 0, lim bi (t)dt > 0, aij (t) ≥ 0, (i, j = 1, ..., n, t ∈ R).
t∈R ω→∞ ω
0

If
Zω h n
1 X i
(L2 ) lim bi (t) − aij (t)Xj0 (t) dt > 0, i = 1, ..., n,
ω→∞ ω
0 j=1,j6=i

where Xj0 (.) is the unique almost periodic solution of the following problem
ẋj = xj [bj (t) − ajj (t)xj ], xj (.) ∈ C + ,
then system (1.2) is permanent and it has at least one solution defined on whole
R, whose components are bounded above and below by positive constants.
254 TRINH TUAN ANH, TRAN VAN NHUNG AND LE HONG LAN

It is clear that
sup Xj0 (t) ≤ Kj := sup bj (t)/ajj (t).
t∈R t∈R
Therefore (L2 ) holds if
Zω n
0 1 h X i
(L2 ) lim bi (t) − Kj aij (t) dt > 0, i = 1, ..., n.
ω→∞ ω
0 j=1,j6=i
Thus we have the following corollary.
Corollary 2. Suppose that the functions bi (t), aij (t) (i, j = 1, ..., n) are almost
periodic and (L1 ), (L02 ) hold. Then the assertion of Corollary 1 is valid.

Let M be the space of continuous functions from R into Rn equipped with the
topology of uniform convergence on compact subsets of R. It is well-known that
M is a Frechet space. Let
M1 := {p ∈ M : θ ≤ p(t) ≤ X 0 (t), t ∈ R}.

By Lemma 1 and the hypotheses (H1 ), (H2 ), (H3 ), for each p ∈ M1 the
following system of n uncouple differential equations
(2.2) żi = zi fi (t, p1 (t), ..., pi−1 (t), zi , pi+1 (t), ..., pn (t)), i = 1, ...n
has a unique solution z(p)(.) ∈ M whose components are bounded above and
below by positive constants. Hence, we can introduce the operator
T : M1 → M
p 7→ T (p) = z(p)(.).
Clearly, p(.) is a solution in M1 of (1.1) if and only if it is a fixed point of T . We
shall apply an extension of Schauder’s fixed point theorem, namely the Tychonov
fixed point theorem, for proving the existence of a fixed point for the operator T .
Theorem 3 (Tychonov). Let X be locally convex and Hausdorff, C ⊂ X closed
convex, F : C → C continuous and F (C) precompact. Then F has a fixed point.

Moreover, Ascoli’s theorem is also used in our proof.

Theorem 4 (Ascoli). Let X be a topological space which is locally compact Haus-
dorff, Y a metric space, and C(X , Y) the space of continuous functions from X
into Y. Consider C(X , Y) with the topology of uniform convergence on compact
subsets of X . A subset F of C(X , Y) has compact closure if and only if it is
equicontinuous and the subset Fx = {h(x)|h ∈ F} of Y has compact closure for
each x ∈ X .

Note that in this paper we do not assume the uniqueness of solutions to the
Cauchy problem for system (1.1). The proof of Theorem 1 is based on the fol-
lowing lemmas.
Lemma 3. If p1 , p2 ∈ M1 , p1 (t) ≤ p2 (t) (t ∈ R), then T (p1 )(t) ≥ T (p2 )(t).
 PERMANENCE AND POSITIVE BOUNDED SOLUTIONS 255

Proof. Suppose the assertion of the lemma is false, i.e., there exist i ∈ {1, ..., n}
and t0 ∈ R such that (T (p1 ))i (t0 ) < (T (p2 ))i (t0 ). Put z 1 = T (p1 ), z 2 = T (p2 )
and A(t) = ln zi1 (t) − ln zi2 (t). Note that A(t0 ) < 0 and z k (t) (k = 1, 2) satisfies
the system (2.2) where p is replaced by pk , i.e.,
(2.2k ) żik = zik fi (t, p1 (t), ..., pi−1 (t), zik , pi+1 (t), ..., pn (t)), i = 1, ..., n.
By (H3 ), we have Ȧ(t0 ) > 0.
Claim. Ȧ(t) > 0 for all t ∈ (−∞, t0 ].
Suppose the claim were false, i.e., there exists t1 < t0 such that Ȧ(t) > 0 for all
t ∈ (t1 , t0 ] and Ȧ(t1 ) = 0. This implies that A(t) is strictly increasing on [t1 , t0 ].
Thus zi1 (t1 ) < zi2 (t1 ) and, consequently, Ȧ(t1 ) > 0. This is a contradiction.
It follows from the claim that zi1 (t) < zi2 (t) for all t ∈ (−∞, t0 ]. By (2.2k ),
(H1 ), (H2 ) we have
Ȧ(t) = fi (t, p11 (t), ..., p1i−1 , zi1 (t), p1i+1 (t), ..., p1n (t))
− fi (t, p21 (t), ..., p2i−1 , zi2 (t), p2i+1 (t), ..., p2n (t))
≥ α[zi2 (t) − zi1 (t)] > 0, t ∈ (−∞, t0 ].
Since zi1 (.), zi2 (.) ∈ C + , there exists β > 0 such that
Zt0
zi1 (t0 )zi2 (ω)
0< Ȧ(t)dt = ln <β for all ω ≤ t0 .
zi2 (t0 )zi1 (ω)
ω
Thus
Zt0
0< α[zi2 (t) − zi1 (t)]dt < +∞.
−∞
Since żi1 (t), żi2 (t) are bounded, we have
lim [zi2 (t) − zi1 (t)] = 0
t→−∞

and, consequently, lim A(t) = 0. This implies that

t→−∞

Zt0
Ȧ(t)dt = A(t0 ) < 0,
−∞
which contradicts
Zt0 Zt0
Ȧ(t)dt ≥ α [zi2 (t) − zi1 (t)]dt > 0.
−∞ −∞
The lemma is proved.

By Lemma 3, we have T (X 0 )(t) ≤ T (θ)(t) = X 0 (t), t ∈ R. Let us set

M2 = {P ∈ M : T (X 0 )(t) ≤ p(t) ≤ X 0 (t), t ∈ R},
256 TRINH TUAN ANH, TRAN VAN NHUNG AND LE HONG LAN

δ= inf T (X 0 )i (t),
t∈R, 1≤i≤n

∆= sup Xi0 (t).

t∈R, 1≤i≤n

It is clear that 0 < δ ≤ ∆ < +∞. It follows from (H2 ) that

0 < L := sup |fi (t, x)| < +∞.

(t,x)∈R×[δ,∆], 1≤i≤n

Let us set

M3 = {p ∈ M2 : |pi (t) − pi (t0 )| ≤ L|t − t0 |, i = 1, ..., n, t ∈ R}.

It is easily seen that M3 is a closed convex subset of M. By Theorem 4, M3 is

compact. Moreover, Lemma 3 implies that T (M3 ) ⊂ M3 .

Lemma 4. The operator T is continuous on M3 (in the topology of uniformly

convergence on compact subsets of R).

Proof. Let {pk }∞ k

k=1 ∈ M3 such that p → p̄ as k → ∞. Clearly, p̄ ∈ M3 . We
shall prove that T (pk ) → T (p̄) as k → ∞.
∞
Since T (pk ) k=1 is precompact, it suffices to show that if a subsequence


{T (pks )} converges to p̃ then p̃ = T (p̄). To this end, let us consider the sys-
tems

(2.2ks ) żi = zi fi (t, pk1s (t), ..., pki−1

s
(t), zi , pki+1
s
(t), ..., pkns (t)), i = 1, ..., n

and

(2.3) żi = zi fi (t, p̄1 (t), ..., p̄i−1 (t), zi , p̄i+1 (t), ..., p̄n (t)), i = 1, ..., n.

Clearly, the right hand side of (2.2ks ) converges to the right hand side of (2.3)
uniformly on any compact subset of R × R+ n . Therefore, from [5, Theorem 2.4, p.

4] it follows that p̃(.) is a solution of (2.2ks ). Since (2.2ks ) has a unique solution
in C + (by Lemma 1), T (p̄) = p̃.

Proof of Theorem 1. (i) The existence. By Lemma 4 and Tychonov’s fixed point
theorem, there exists x∗ ∈ M3 such that T (x∗ ) = x∗ . Thus x∗ (t) is a solution
of system (1.1) whose components are bounded above and below by positive
constants.
(ii) The permanence. First of all we prove the following claim.
Claim. If x(.) is a solution of (1.1) and xi (t0 ) > 0 for some i ∈ {1, ..., n}, then
xi (t) > 0 for all t ≥ t0 where the solution is defined.
Indeed, if xi (t1 ) = 0 for a minimal value t1 > t0 , we have that w1 (.) = xi (.)
and w2 (.) = 0 both are solutions of the scalar differential equation

ẇ = wfi (t, x1 (t), ..., xi−1 (t), w, xi+1 (t), ..., xn (t))
 PERMANENCE AND POSITIVE BOUNDED SOLUTIONS 257

satisfying the same condition at t1 . Hence, for t ∈ [t0 , t1 ], we have

d 
1 2
 (w (t) − w (t)) = |w1 (t)fi (t, x1 (t), ..., xn (t))| ≤ γ|w1 (t)|
 
dt
= γ|w1 (t) − w2 (t)|,
for a suitable constant γ > 0. Since w1 (t1 )−w2 (t1 ) = 0, the Gronwall lemma gives
w1 (t)−w2 (t) = 0 for all t ∈ [t0 , t1 ], which contradicts the fact that w1 (t)−w2 (t) >
0 for t ∈ [t0 , t1 ). The claim is proved.
Let x(.) be any noncontinuable solution of (1.1) with xi (t0 ) > 0, (i = 1, ..., n).
By the above claim and hypothesis (H3 ), x(.) is defined in [t0 , +∞) and xi (t) > 0
for all t ≥ t0 , i = 1, ..., n. For each i = 1, ..., n, let ui (.) be the (right) noncon-
tinuable maximum solution of the scalar equation (2.2i ) with ui (t0 ) = xi (t0 ). By
(H3 ), the relation
ẋi (t) = xi (t)fi (t, x1 (t), ..., xn (t)) ≤ xi (t)fi (t, 0, ..., 0, xi (t), 0, ..., 0),
holds for each i and t ≥ t0 . By [8, Lemma 2.6, p.318], we have
(2.4) xi (t) ≤ ui (t) (t ≥ t0 , i = 1, ..., n).
Since fi (t, x) (i = 1, ..., n) is uniformly continuous, the hypothesis (H4 ) implies
that there exists  > 0 such that, for i = 1, ..., n, it holds
(2.5) 0
inf fi (t, X10 (t) + ε, ..., Xi−1 0
(t) + ε, 0, Xi+1 (t) + ε, ..., Xn0 (t) + ε) > 0.
t∈R

By Lemma 1, lim |ui (t) − Xi0 (t)| = 0, (i = 1, ..., n). Hence (2.4) implies that
t→+∞
there exists t1 ≥ t0 such that
(2.6) xi (t) ≤ Xi0 (t) + ε, t ≥ t1 , i = 1, ..., n.
By Lemma 1, for each i = 1, ..., n, the scalar differential equation
(2.6i ) żi = zi fi (t, X10 (t) + ε, ...., Xi−1
0
(t) + ε, zi , Xi0 (t) + ε, ..., Xn0 (t) + ε)
has a unique solution Zi0 (t) which is bounded above and below by positive con-
stants.
Let zi (.) (i = 1, ..., n) be the (right) noncontinuable minimum solution of the
scalar equation (2.6i ) with zi (t1 ) = xi (t1 ). We have, for t ≥ t1 ,
ẋi (t) ≥ xi (t)fi (t, Xi0 (t) + ε, ..., Xi−1
0 0
(t) + ε, xi (t), Xi+1 (t) + ε, ..., Xn0 (t) + ε).
Consequently, by [8, Lemma 2.7, p.319] we have
zi (t) ≤ xi (t), t ≥ t1 , i = 1, .., n.
Once again, by Lemma 1, lim |zi (t) − Zi0 (t)| = 0, (i = 1, ..., n).
t→+∞
We set
n o n o
M = ε + max sup Xi0 (t) , m = min inf Zi0 (t) .
1≤i≤n t∈R 1≤i≤n t∈R

Then (1.3) holds, i.e., system (1.1) is permanent. The proof of Theorem 1 is now
complete.
258 TRINH TUAN ANH, TRAN VAN NHUNG AND LE HONG LAN

Proof of Theorem 2. For i = 1, ..., n, let us set εi = γi /4 and

Zω
1
γi = lim fi (t, X10 (t), ..., Xi−1
0 0
(t), 0, Xi+1 (t), ..., Xn0 (t))dt > 0.
ω→∞ ω
0

Since 0 (t), 0, X 0 (t), ..., X 0 (t))

fi (t, X10 (t), ..., Xi−1 i+1 (i = 1, ..., n) is almost periodic,
n
the approximation theorem [7, p. 17] implies that there exists a trigonometric
polynomial ∆i (t) such that
sup |fi (t, X10 (t), ..., Xi−1
0 0
(t), 0, Xi+1 (t), ..., Xn0 (t)) − ∆i (t)| < εi .
t∈R

Clearly, we have that

Zω
3γi ¯ i := lim 5γi
γi − εi = <∆ ∆i (t)dt ≤ εi − γi =
4 ω→∞ 4
0

Zt
¯ it −
and exp{∆ ∆i (s)ds} is almost periodic.
0
By a change of the variable
n Zt o
ui = xi exp ∆¯ i t − ∆i (s)ds (i = 1, ..., n)
0

the system (1.1) becomes

(2.7) u̇i = ui Fi (t, u), i = 1, ..., n,
where
 Zt
¯ 1 t + ∆1 (s)ds , ...,

Fi (t, u) = fi t, u1 exp − ∆
0
Zt

¯ nt + ¯ i − ∆i (t).

un exp − ∆ ∆n (s)ds +∆
0

Hence, it suffices to show that system (2.7) is permanent and it has at least one
solution defined on whole R, whose components are bounded above and below
by positive constants.
It is clear that
¯ i − ∆i (t)
Fi (t, θ) = fi (t, θ) + ∆
≥ fi (t, X10 (t), ..., Xi−1
0 0
(t), 0, Xi+1 (t), ..., Xn0 (t))
+∆ ¯ i − ∆i (t) ≥ γi > 0.
2
 PERMANENCE AND POSITIVE BOUNDED SOLUTIONS 259

Zt
Ui0 (t) Xi0 (t) exp ¯ i−

Therefore (by Lemma 1) := ∆ ∆(s)ds is a unique solution
0
in C + of the following equation

u̇i = ui Fi (t, 0, ..., 0, ui , 0, ..., 0) (i = 1, ..., n).

Moreover,

Fi (t, U10 (t), ..., Ui−1

0 0
(t), 0, Ui+1 (t), ..., Un0 (t))
= fi (t, X10 (t), ..., Xi−1
0 0
(t), 0, Xi+1 ¯ i − ∆i (t)
(t), ..., Xn0 (t)) + ∆
γi
≥ > 0.
2
Thus system (2.7) satisfies all the conditions of Theorem 1. The proof of Theo-
rem 2 is now complete.

References
[1] S. Ahmad, On almost periodic solutions of competing species problems, Proc. Amer. Math.
Soc. 102 (1988), 855-861.
[2] C. Alvarez and A. C. Lazer, An application of topological degree to the periodic competing
species problem, J. Austral. Math. Soc. Ser. B 28 (1986), 202-219.
[3] K. Gopalsamy, Global asymptotic stability in a periodic Lotka-Volterra system, J. Austral.
Math. Soc. Ser. B 27 (1985), 66-72.
[4] K. Gopalsamy, Global asymptotic stability in an almost periodic Lotka- Volterra system, J.
Austral. Math. Soc. Ser. B 27 (1986), 346-360.
[5] P. Hartman, Ordinary Differential Equations, Birkhauser, Boston, 1982.
[6] V. Hutson and K. Schmitt, Permanence and the dynamics of biological systems, Math.
Biosci. 111 (1992), 1-71.
[7] B. M. Levitan and V. V. Zhikov, Almost Periodic Functions and Differential Equations,
Cambridge University Press, Cambridge, 1982.
[8] N. Rouch, P. Habets and M. Laloy, Stability Theory by Liapunovs Direct Methods, Springer-
Verlag, New York, 1977.
[9] A. Tineo and Alvarez, A different consideration about the globally asymptotically stable
solution of the periodic n-competing species problem, J. Math. Anal. Appl. 159 (1991),
44-50.
[10] A. Tineo, Asymptotic behaviour of posotive solutions of the nonautonomous Lotka-Volterra
competition equations, Differential and Integral Equations 2 (1993), 449-457.
[11] Trinh T. Anh, Stability of Differential Equations in some Mathematical Models of Popula-
tion Dynamics, Ph. D. Thesis, Hanoi National University, 1998 (in Vietnamese).
[12] F. Zanolin, Permanence and positive periodic solutions for Kolmogorov competing species
systems, Results in Math. 21 (1992), 224-250.
[13] X. Q. Zhao, The qualitative analysis of n-species Lotka-Volterra periodic competition sys-
tems, Math. Comput. Modelling 15 (1991), 3-8.

Department of Mathematics,
Hanoi University of Pedagogy,
Cau giay, Hanoi
E-mail address: ttanh@thevinh.ncst.ac.vn
260 TRINH TUAN ANH, TRAN VAN NHUNG AND LE HONG LAN

Ministry of Education and Training,

49 Dai Co Viet Str., Hanoi
E-mail address: tvnhung@thevinh.ncst.ac.vn

Department of Mathematics,
Hanoi University of Transportation and Telecommunication
E-mail address: lhlan@thevinh.ncst.ac.vn