Nonlinear differ. equ. appl.

15 (2008), 473–494
c 2008 Birkhäuser Verlag Basel/Switzerland
1021-9722/040473-22
published online 26 November 2008 Nonlinear Differential Equations
DOI 10.1007/s00030-008-7045-9 and Applications NoDEA

Existence of Positive Radial Solutions for a

Nonvariational Polyharmonic System
Yajing Zhang and Jianghao Hao1

Abstract. In this work we first prove a new Liouville-type result for a poly-
harmonic system in RN , then prove the existence of positive radial solutions
for a polyharmonic system without variational structure in bounded domains
via topological method.
Mathematics Subject Classification (2000). Primary 35J60; Secondary 35J55.
Keywords. Polyharmonic system, positive radial solution, a priori estimate,
existence.

1. Introduction
In this paper we study the existence of positive radial solutions of the following
polyharmonic system
⎧
⎪ m k p
⎪ (−Δ)m u = uq vs , in BR ,
⎨
(−Δ) v = u v , in BR ,
(1.1)
⎪
⎪ u = (−Δ)u = · · · = (−Δ)m−1 u = 0 , on ∂BR ,
⎩
v = (−Δ)v = · · · = (−Δ)m−1 v = 0 , on ∂BR ,
where m ≥ 1, N > 2m, p, q, k, s ≥ 0 and BR is the ball centered at zero and
radius R in RN .
For m = 1, this type problem were well studied, more precisely, the existence
of positive solutions of the elliptic system

−Δu = f (u, v) , in Ω ,
(1.2)
−Δv = g(u, v) , in Ω ,

The first author was supported by NSFC (Grant No. 10701051) and China Postdoctoral Science
Foundation.
The second author was supported by NSFC (Grant No. 10701051).
1 Corresponding author.
474 Y. Zhang and J. Hao NoDEA

subject to zero Dirichlet boundary conditions has been studied by a number of

authors, see for examples [2–5, 8, 12, 13, 17], and the references therein. Variational
methods are of some use when the differential problem(equations or systems) is the
Euler-Lagrange equation of some functional. In the case of a scalar equation, this
is always the case, provided some growth conditions on nonlinearity is imposed.
For the case of a system like (1.2), it is not always the case that there is a natural
functional associated to it. If the system is not variational one has to use topological
methods.
Polyharmonic equations (systems) have received great attention recently,
Bartsch and Guo [1] investigated semilinear elliptic systems for the polyharmonic
operator having a critical growth nonlinearity. They establish conditions for exis-
tence of nontrivial solutions to the systems using variational methods. System (1.1)
need not have a variational structure, and consequently we will use topological
methods to prove the existence of positive solutions. The purpose of this paper
partly is to show that how the topological method works in the study of higher
order elliptic boundary value problems. Our main result is the following.
Theorem 1.1. Suppose that N > 2m, and p, q, k, s satisfy
(i) p, q > 0, pq> 1, 0 ≤ k ≤ 1, 0 ≤ s ≤ 1, 
N (p+1)(q+1)−ks (p+1)(q+1)+k(q+1)+s(p+1)
(ii) 2m < min pq−(1−k)(1−s) , pq−1+k(q+1)+s(p+1) .
Then (1.1) possesses a positive radial solution.
The proof of Theorem 1.1 is based on a fixed point theorem on positive cones
as well as a priori estimates for the solutions. One important feature of the proof
of Theorem 1.1 is that it applies to general nonlinearities with certain growth
restriction in u and v at infinity. Here the a priori bounds for the solutions of (1.1)
are obtained via the so-called blow-up method intruduced by Gidas and Spruck
in their fundamental paper [7]. Roughly speaking, in this method one assumes
by contradiction that no universal estimate for the solutions. Then after some
limit procedures one ends up with a limiting problem for which there are positive
solutions. A contradiction comes if we prove the nonexistence of positive solutions
of the limiting problem (Liouville-type theorems).
To perform our program we first need to study nonexistence of positive radial
solutions for the following polyharmonic system

(−Δ)m u = uk v p ,
in RN . (1.3)
(−Δ)m v = uq v s ,
Compared with the case m = 1, the main difficulty we meet is that the maxi-
mum principle can not be directly applied to u, v if one does not know enough
information about (−Δ)i u, (−Δ)i v, i = 1, . . . , m − 1. Hence, we need to prove
the general superharmonic property of the solutions of the polyharmonic sys-
tem (1.3). This is the key part of proof of the nonexistence of positive solutions.
Let (u, v) ∈ C 2m (RN ) × C 2m (RN ) be a positive solution of the system (1.3),
we say that (u, v) satisfies the general superharmonic property if (−Δ)i u ≥ 0,
Vol. 15 (2008) Existence for a Polyharmonic System 475

(−Δ)i v ≥ 0, i = 1, . . . , m − 1. We obtain a new Liouville-type theorem which

extends the earlier works by the first author and his co-workers [10, 15] to more
complicated nonlinearity including the product of two powers.
The organization of the paper is as follows. In Section 2 we prove a new
nonexistence result for polyharmonic system in RN . In Section 3 we prove the a
priori estimates of positive solutions for the poyharmonic systems (1.1) via the
blow-up method. Finally in Section 4, we use a fixed point theorem to prove
Theorem 1.1. Throughout this paper, the constant c will denote various generic
constants.

2. Liouville-type theorem
In this section, we prove a Liouville type result for polyharmonic system.
Theorem 2.1. Suppose that N > 2m, and p, q, k, s satisfy
(i) p, q > 0, pq> 1, 0 ≤ k ≤ 1, 0 ≤ s ≤ 1, 
N (p+1)(q+1)−ks (p+1)(q+1)+k(q+1)+s(p+1)
(ii) 2m < min pq−(1−k)(1−s) , pq−1+k(q+1)+s(p+1) .
Then the polyharmonic system

(−Δ)m u = uk v p ,
in RN , (2.1)
(−Δ)m v = uq v s ,
has no positive radial classical solutions.
We note that the above nonexistence result can be extended to more compli-
cated system 
(−Δ)m u = |x|α uk v p ,
in RN ,
(−Δ)m v = |x|β uq v s ,
where α, β, k, s, p and q ≥ 0. The proof is along the lines of Theorem 2.1.
The following lemma is a general version of Lemma 3.1 in [14].
Lemma 2.2. Suppose y = y(r) ≥ 0 satisfies
N −1 
y  (r) + y (r) + ϕ(r) ≤ 0 , r > r0 ≥ 0 , (2.2)
r
with ϕ non-negative and non-increasing, and y  (r0 ) ≤ 0. Then
y(r) ≥ cr2 ϕ(r) , for r ≥ 2r0 ,
where c = c(N ).
Proof. From (2.2),
N −1 
y  (r) + y (r) ≤ 0 ,
r
multiplying the above inequality by rN −1 and integrating, we get
y  (r) ≤ 0 , for r ≥ r0 .
476 Y. Zhang and J. Hao NoDEA

Multiplying (2.2) by rN −1 and integrating the inequality, we get, for r > 2r0 ,
r

−y (r) ≥ r 1−N
tN −1 ϕ(t)dt
r0
r/2
≥ r1−N tN −1 ϕ(t)dt
r0

r N −1
1
r r 0
≥ ϕ − r0
N 2 2N r

1
r r 1
≥ ϕ − N −1 r0 .
N 2 2N 2
Integrating this from r to 2r then yields
 
1 2r t 1 1
y(r) ≥ y(2r) + ϕ t − N −1 r0 dt
N 2 2N 2
 r
1 1 2 1
≥ ϕ(r) 3r − N −1 r0 r
N 2N 2
1
≥ N ϕ(r)r . 2
2 N
The proof of Lemma 2.2 is completed. 
Lemma 2.3. Suppose positive radial functions u, v satisfy

(−Δ)m u(r) ≥ v p (r) , r ≥ r0 ,
(2.3)
(−Δ)m v(r) ≥ uq (r) , r ≥ r0 ,
where p, q > 0, pq > 1 and r0 ≥ 0. Then
(−Δ)i u(r) ≥ 0 , (−Δ)i v(r) ≥ 0 , for r ≥ r0 , i = 1, . . . , m − 1 .
The proof of this lemma can be found in [10] and [16].
Remark 2.4. It is easy to see that we can obtain the same result for the following
polyharmonic inequality (−Δ)m u ≥ up , r ≥ r0 , where p > 1.
Lemma 2.5. If ψ ∈ C 2m (RN ) is positive, radial and (−Δ)k ψ ≥ 0 in RN , k =
1, . . . , m, then for every r ∈ (0, +∞) we have
rψ  (r) + (N − 2m)ψ(r) ≥ 0 . (2.4)
For its proof, see [3].
Remark 2.6. Note that (2.4) is equivalent to (rN −2m ψ(r)) ≥ 0. Thus rN −2m ψ(r)
is non-decreasing.
Proposition 2.7. If (u, v) is a positive radial solution of the polyharmonic system

(−Δ)m u = a(|x|)uk v p ,
in RN , (2.5)
(−Δ)m v = b(|x|)uq v s ,
and a, b, p, q, k, s satisfy
Vol. 15 (2008) Existence for a Polyharmonic System 477

(i) a, b ∈ C[0, +∞) are non-decreasing and a(r) > 0, b(r) > 0 for r > 0,
(ii) 0 ≤ k, s ≤ 1, p > 0, q > 0, pq > 1.
Then
(−Δ)i u ≥ 0 , (−Δ)i v ≥ 0 , i = 1, . . . , m − 1 .
Proof. Step 1. We show that (−Δ)m−1 u ≥ 0, (−Δ)m−1 v ≥ 0.
The proof is by contradiction argument. Suppose the contrary, then there
exists a number r∗ ≥ 0 such that (−Δ)m−1 u(r∗ ) < 0. Set w = (−Δ)m−1 u, z =
(−Δ)m−1 v. Then it is easy to see −Δw = (−Δ)m u ≥ 0. Multiply the inequality
by rN −1 and integrate the resulting inequality, we have dwdr ≤ 0, thus w is non-
increasing and
w(r) ≤ w(r∗ ) < 0 , for r ≥ r∗ . (2.6)
In the following, we discuss two cases.
Case 1. m is an odd number.
It follows from (2.6),
Δm−1 u ≤ w(r∗ ) .
Multiplying the above inequality by rN −1 and integrating twice, we get there exists
c1 > 0 and r1 > 0 such that
Δm−2 u(r) ≤ c1 w(r∗ )r2 , for r ≥ r1 . (2.7)
Repeat the above process step by step, we get there exists c2 > 0 and r2 > 0 such
that
u(r) ≤ c2 w(r∗ )r2(m−1) , for r ≥ r2 . (2.8)
∗
Since w(r ) < 0, we have
u(r) → −∞ , as r → +∞ .
This is a contradiction since u(r) is positive.
Case 2. m is an even number.
It follows from (2.6),
Δm−1 u ≥ −w(r∗ ) > 0 .
By the same arguments as we did in the proof of Case 1, we get
u(r) → +∞ , as r → +∞ . (2.9)
In the following, we show (−Δ) v ≥ 0, i = 1, . . . , m − 1. If there exists r3 ≥ 0
i

such that z(r3 ) = (−Δ)m−1 v(r3 ) < 0, we have

Δm−1 v(r) ≥ −z(r3 ) > 0 , for r ≥ r3 ,
since z is non-increasing and m is an even number. Thus
v(r) → +∞ , as r → +∞ . (2.10)
Since a, b are non-decreasing, by (2.5), (2.9), (2.10) there exists c0 > 0, r0 > 0
such that  m
Δ u ≥ c0 v p , r ≥ r0 ,
(2.11)
Δm v ≥ c0 uq , r ≥ r0 ,
478 Y. Zhang and J. Hao NoDEA

By Lemma 2.3, we get a contradiction. Thus

(−Δ)m−1 v ≥ 0 , for r ≥ 0 .
If there exists r4 ≥ 0 such that (−Δ)m−2 v(r4 ) < 0, we have
Δm−2 v(r) ≤ Δm−2 v(r4 ) < 0 , for r ≥ r4 , (2.12)
since m is an even number. From (2.12), we get
v(r) → −∞ , as r → +∞ ,
again a contradiction. Repeating the above process, we obtain, if there exists ri ≥ 0
such that (−Δ)i v(ri ) < 0, then when i is an odd number,
v(r) → +∞ , as r → +∞ ,
and when i is an even number,
v(r) → −∞ , as r → +∞ ,
in both cases, we get a contradiction by Lemma 2.3 and v > 0. Thus
(−Δ)i v ≥ 0 , i = 1, . . . , m − 1 . (2.13)
From (2.5) we have
Δm v = b(r)uq v s . (2.14)
Since b(r) is non-decreasing and u(r) → +∞ as r → +∞, there exist c̃ > 0 and
r̃ > 0 such that
Δm v ≥ c̃v s , for r ≥ r̃ .
Thus by Lemma 2.2,
v ≥ cr2m v s (r) , for r ≥ 2r̃ . (2.15)
We get, when 0 < s < 1,
v(r) ≥ cr2m → +∞ , as r → +∞ ,
a contradiction since v(r) is non-increasing by (2.13); when s = 1,
1 ≥ cr2m , for r ≥ 2r̃ .
again a contradiction. We complete the Step 1.
Step 2. We show that (−Δ)m−i u ≥ 0, (−Δ)m−i v ≥ 0, i = 2, . . . , m − 1.
We prove this by induction on i. By step 1. this is true for i = 1. Assume
that (−Δ)m−i+1 u ≥ 0, (−Δ)m−i+1 v ≥ 0. Set w = (−Δ)m−i u, suppose by contrary
that there exists r̄ ≥ 0 such that w(r̄) < 0, we get
w(r) ≤ w(r̄) < 0 , for r ≥ r̄ , (2.16)
since −Δw = (−Δ) m−(i−1)
u ≥ 0. Again, we discuss two cases.
Case 1. m − i is an even number.
From (2.16),
Δm−i u ≤ w(r̄) < 0 ,
this deduces that u → −∞ as r → +∞, a contradiction.
Vol. 15 (2008) Existence for a Polyharmonic System 479

Case 2. m − i is an odd number.

From (2.16),
Δm−i u ≥ −w(r̄) > 0 ,
we get u → +∞ as r → +∞. We follow the same arguments in step 1 and lead to
a contradiction. Thus we complete the proof of Proposition 2.7. 

In order to prove Theorem 2.1 we make use of the following identity.

Lemma 2.8. Let Ω be a smooth bounded domain in RN , and u, v ∈ C 2m (Ω), then

Δm u(x, ∇v)dx
Ω
m
  
∂(Δm−k u)   m−k ∂(x, ∇(Δ
k−1
v))
= x, ∇(Δk−1
v) − Δ u dσ
∂n ∂n
k=1 ∂Ω
   ∂(Δm−k−i u)
m m−k k+i−1

m−k−i ∂(Δ v)
+2 Δ k+i−1
v−Δ u dσ
∂n ∂n
k=1 i=1 ∂Ω

 
+ 2m uΔm vdx + u x, ∇(Δm v) dx . (2.17)
Ω Ω

The proof is a straightforward application of Green’s formula and can be

found in [11].

Proof of Theorem 2.1. Set ui = (−Δ)i u, vi = (−Δ)i v, i = 1, . . . , m−1. From (2.1),

we get
⎧
⎪
⎪ Δu + u1 = 0 ,
⎪
⎪
⎪
⎪ Δu i + ui+1 = 0 , i = 1, . . . , m − 2 ,
⎨
Δum−1 + uk v p = 0 ,
(2.18)
⎪
⎪ Δv + v1 = 0 ,
⎪
⎪
⎪
⎪ Δvi + vi+1 = 0 , i = 1, . . . , m − 2 ,
⎩
Δvm−1 + uq v s = 0 ,
By Lemma 2.2,

u ≥ cr2 u1 , ui ≥ cr2 ui+1 , i = 1, . . . , m − 2 , um−1 ≥ cr2 u(r)k v(r)p , (2.19)

v ≥ cr2 v1 , vi ≥ cr2 vi+1 , i = 1, . . . , m − 2 , vm−1 ≥ cr2 u(r)q v(r)s . (2.20)

From (2.19), (2.20),

⎧ −2m(1−s+p)
⎪
⎪ u(r) ≤ cr pq−(1−s)(1−k) ,
⎪
⎪
⎨ −2m(1−k+q)
v(r) ≤ cr pq−(1−s)(1−k) ,
−2m(1−s+p) (2.21)
⎪
⎪ ui (r) ≤ cr−2i+ pq−(1−s)(1−k) ,
⎪
⎪
⎩ −2i−2m(1−k+q)
vi (r) ≤ cr−2i+ pq−(1−s)(1−k) .
480 Y. Zhang and J. Hao NoDEA

By Lemma 2.5,
⎧ −2m(1−s+p)
⎪
⎪
⎪ −u (r) ≤ cr−1+ pq−(1−s)(1−k) ,
⎪
⎨ −2m(1−k+q)
−v  (r) ≤ cr−1+ pq−(1−s)(1−k) ,
−2m(1−s+p) (2.22)
⎪
⎪ −ui (r) ≤ cr−1−2i+ pq−(1−s)(1−k) ,
⎪
⎪
⎩ −2m(1−k+q)
−vi (r) ≤ cr−1−2i+ pq−(1−s)(1−k) .
Using (2.17) in Ω = BR (0), we get
R
rN uk v p v  dr
0

m
  
 
=− um−k vk−1 R − um−k (rvk−1 ) RN −1
k=1
 
m m−k
−2 (um−k−i vk+i−1 − um−k−i vk+i−1

)RN −1
k=1 i=1
R R
+ 2m rN −1 uq+1 v s dr + u(uq v s ) rN dr
0 0

m
   N −1
 
=− um−k vk−1 R − um−k vk R + (N − 2)um−k vk−1 R
k=1
 
m m−k
−2 (um−k−i vk+i−1 − um−k−i vk+i−1

)RN −1
k=1 i=1
R R
N −1 q+1 s
+ 2m r u v dr + urN d(uq v s )
0 0

m
   N −1
 
=− um−k vk−1 R − um−k vk R + (N − 2)um−k vk−1 R
k=1
 
m m−k
−2 (um−k−i vk+i−1 − um−k−i vk+i−1

)RN −1
k=1 i=1
R R
+ 2m rN −1 uq+1 v s dr + RN uq+1 v s − uq v s d(urN )
0 0

m
   N −1
 
=− um−k vk−1 R − um−k vk R + (N − 2)um−k vk−1 R
k=1
 
m m−k
−2 (um−k−i vk+i−1 − um−k−i vk+i−1

)RN −1
k=1 i=1
R R
N −1 q+1 s
+ (2m − N ) r u v dr + R u N q+1 s
v − rN uq u v s dr .
0 0
Vol. 15 (2008) Existence for a Polyharmonic System 481

By (ii),


m
  N −1
− um−k vk−1
 
R − um−k vk R + (N −2)um−k vk−1 R → 0, as R → +∞ ,
k=1
 
m m−k
−2 (um−k−i vk+i−1 − um−k−i vk+i−1

)RN −1 → 0 , as R → +∞ ,
k=1 i=1
RN uq+1 v s → 0 , as R → +∞ .

Thus
R
rN uk v p v  dr
0
R R
= (2m − N ) rN −1 uq+1 v s dr − rN uq u v s dr + o(R) , (2.23)
0 0

where o(R) means infinitely small quantity.

Similarly we have
R
rN uq v s u dr
0
R R
N −1 p+1 k
= (2m − N ) r v u dr − rN uk v p v  dr + o(R) . (2.24)
0 0

From (2.23), (2.24),

R R
N k p 
r u v v dr + rN uq v s u dr
0 0
 R  R
N N −1 q+1 s N
= m− r u v dr + m − rN −1 v p+1 uk dr + o(R) .
2 0 2 0
(2.25)

Integrating the left-hand side of (2.25) by parts on (0, R) yields

R R
rN uk v p v  dr + rN uq v s u dr
0 0
R
1 N
= RN uk v p+1 − rN −1 uk v p+1 dr
p+1 p+1 0
R R
k 1 N
− rN uk−1 u v p+1 dr + RN uq+1 v s − rN −1 uq+1 v s dr
p+1 0 q+1 q+1 0
R
s
− rN uq+1 v  v s−1 dr . (2.26)
q+1 0
482 Y. Zhang and J. Hao NoDEA

Using (ii), we get

R R
rN uk v p v  dr + rN uq v s u dr
0 0
R R
N k
=− rN −1 uk v p+1 dr − rN uk−1 u v p+1 dr
p+1 0 p+1 0
R R
N s
− rN −1 uq+1 v s dr − rN uq+1 v  v s−1 dr + o(R) . (2.27)
q+1 0 q+1 0

By Lemma 2.5,
R R
N k p 
r u v v dr + rN uq v s u dr
0 0

N R
k(N − 2m) R
≤− rN −1 uk v p+1 dr + rN −1 uk v p+1 dr
p+1 0 p+1 0
R
N R
s(N − 2m)
− rN −1 uq+1 v s dr + rN −1 uq+1 v s dr + o(R)
q+1 0 q+1 0

k(N − 2m) − N R
= rN −1 uk v p+1 dr
p+1 0

s(N − 2m) − N R
+ rN −1 uq+1 v s dr + o(R) .
q+1 0

From (2.25),
 R
N s(N − 2m) − N
m− − rN −1 uq+1 v s dr
2 q+1 0
 R
N k(N − 2) − N
+ m− − rN −1 v p+1 uk dr + o(R) ≤ 0 . (2.28)
2 p+1 0

Using (ii), we get

R R
rN −1 uq+1 v s dr < +∞ , rN −1 v p+1 uk dr < +∞ .
0 0

From (2.23), (2.24),

R R
rN −1 uq+1 v s dr = rN −1 v p+1 uk dr + o(R) .
0 0

Passing to the limit R → +∞ in the (2.28),

  +∞
s(N − 2m) − N k(N − 2m) − N
2m − N − − rN −1 v p+1 uk dr ≤ 0 ,
q+1 p+1 0

which contradicts (ii), and this concludes the proof of Theorem 2.1. 
Vol. 15 (2008) Existence for a Polyharmonic System 483

3. A priori estimates
In this section we follow the approach suggested in [3] to prove a priori estimates
for positive radial solutions of the polyharmonic system (1.1).
Lemma 3.1. Let u ∈ C 1 [r0 , +∞) ∩ C 2 (r0 , +∞) be a nonnegative solution of
 
− rN −1 u (r) = rN −1 f (r) , in [r0 , +∞) ,
where f ∈ C(r0 , +∞) and is nonnegative, r0 ≥ 0. Assume that
u(r0 ) > 0 , and u (r0 ) < 0 .
Then
u(r) > 0 and u (r) ≤ 0 , for r ≥ r0 .
This lemma is a simple version of Lemma 2.1 in [3].
Proposition 3.2. Under the assumptions of Theorem 1.1, if (u, v) denotes any posi-
tive radial solution of the polyharmonic system (1.1), then there exists a constant C
independent of u and v such that
u ∞ ≤ C , (−Δ)i u ∞ ≤ C , v ∞ ≤ C , (−Δ)i v ∞ ≤ C ,
i = 1, . . . , m − 1 , (3.1)
where · ∞ is the norm in C([0, R]).
Proof. We show that (−Δ)i u ≥ 0, (−Δ)i v ≥ 0, i = 1, . . . , m − 1. Set w =
(−Δ)m−1 u, it is easy to see −Δw = (−Δ)m u ≥ 0. Multiplying the inequality
by rN −1 and integrate the resulting inequality, we have dw dr ≤ 0, thus w is non-
increasing. By the boundary condition (−Δ)m−1 u(R) = 0, we have (−Δ)m−1 u ≥
0, for r ∈ [0, R]. Similarly, we obtain (−Δ)i u ≥ 0, (−Δ)i v ≥ 0, for r ∈ [0, R], i =
1, . . . , m − 1.
Set ui = (−Δ)i u, vi = (−Δ)i v, i = 1, . . . , m−1, we rewrite the polyharmonic
system (1.1) as
⎧
⎪
⎪ Δu + u1 = 0 , in BR ,
⎪
⎪
⎪
⎪ Δu i + ui+1 = 0 , i = 1, . . . , m − 2 , in BR ,
⎪
⎪
⎨ Δum−1 + uk v p = 0 , in BR ,
Δv + v1 = 0 , in BR , (3.2)
⎪
⎪
⎪
⎪ Δv i + v i+1 = 0 , i = 1, . . . , m − 2 , in B R ,
⎪
⎪
⎪
⎪ Δvm−1 + uq v s = 0 in BR ,
⎩
u = ui = v = vi = 0 , on ∂BR , i = 1, . . . , m − 1 .
We argue by contradiction. Thus we assume that there is a sequence
{(un , u1,n , · · · , um−1,n , vn , v1,n , · · · , vm−1,n )}
of positive solutions of system (3.2) such that
 

m−1 
m−1
lim tn + ti,n + sn + si,n = +∞ , (3.3)
n→∞
i=1 i=1
484 Y. Zhang and J. Hao NoDEA

where
tn = sup un (r) = un (0) , ti,n = sup ui,n (r) = ui,n (0) , i = 1, . . . , m − 1 ,
r∈[0,R] r∈[0,R]

sn = sup vn (r) = vn (0) , si,n = sup vi,n (r) = vi,n (0) , i = 1, . . . , m − 1 .

r∈[0,R] r∈[0,R]

Let us define
2m(p + 1 − s) 2m(q + 1 − k)
λ= , μ= . (3.4)
pq − (1 − k)(1 − s) pq − (1 − k)(1 − s)
Set
1

m−1
1 1 
m−1
1
γn = (tn ) λ + (ti,n ) λ+2i + (sn ) μ + (si,n ) μ+2i , (3.5)
i=1 i=1
then γn → +∞, as n → +∞. Put y = γn r,
1 1
wn (y) = λ un (r) , wi,n (y) = λ+2i ui,n (r) , i = 1, . . . , m − 1 , (3.6)
γn γn
1 1
zn (y) = μ vn (r) , zi,n (y) = μ+2i vi,n (r) , i = 1, . . . , m − 1 . (3.7)
γn γn
We have that

wn (y) ≤ 1 , wi,n (y) ≤ 1 , zn (y) ≤ 1 , zi,n (y) ≤ 1 , y ∈ [0, Rγn ] ,

i = 1, . . . , m − 1 , (3.8)
and
wn (Rγn ) = wi,n (Rγn ) = zn (Rγn ) = zi,n (Rγn ) = 0 , i = 1, . . . , m − 1 , (3.9)
wn (0) = 
wi,n (0) = zn (0) = 
zi,n (0) = 0, i = 1, . . . , m − 1 . (3.10)
By the definition of λ, μ,
pμ + λ(k − 1) − 2m = qλ + μ(s − 1) − 2m = 0 .
Then ⎧
⎪ Δwn + w1,n = 0 ,
⎪
⎪
⎪
⎪ Δwi,n + wi+1,n = 0 , i = 1, . . . , m − 2 ,
⎪
⎨ pμ+λ(k−1)−2m k p
Δwm−1,n + γn wn zn = 0 ,
⎪
⎪ Δz n + z 1,n = 0 ,
⎪
⎪
⎪
⎪ Δzi,n + zi+1,n = 0 ,
⎩ qλ+μ(s−1)−2m q s
Δzm−1,n + γn wn zn = 0 .
or in short
ΔWn + f (Wn ) = 0 , (3.11)
where
Wn = (wn , w1,n , . . . , wm−1,n , zn , z1,n , . . . , zm−1,n ) ,
f (Wn ) = (w1,n , . . . , wm−1,n , wnk znp , z1,n , . . . , zm−1,n , wnq zns ) .
Vol. 15 (2008) Existence for a Polyharmonic System 485

Since
un (0) ui,n (0)
wn (0) = , wi,n (0) = , i = 1, . . . , m − 1 ,
γnλ γnλ+2i
vn (0) vi,n (0)
zn (0) = , zi,n (0) = μ+2i , i = 1, . . . , m − 1 ,
γnμ γn
we have that
1 
m−1 1 1 
m−1 1
wnλ (0) + λ+2i
wi,n (0) + znμ (0) + μ+2i
zi,n (0) = 1 .
i=1 i=1

Thus {wn (0)}, {wi,n (0)}, {zn (0)}, {zi,n (0)} are bounded sequences of real number.
Passing to subsequence if necessary we can assume that

wn (0) → ŵ0 , wi,n (0) → ŵi,0 , zn (0) → ẑ0 , zi,n (0) → ẑi,0 ,
as n → ∞, i = 1, . . . , m − 1 , (3.12)
and hence

m−1 
m−1
ŵ0 + ŵi,0 + ẑ0 + ẑi,0 = 1 . (3.13)
i=1 i=1
For any R̄ ∈ (0, Rγn ), {wn }, {wi,n }, {zn }, {zi,n } are uniformly bounded sequences
in C([0, R̄]). We claim {wn }, {wi,n }, {zn }, {zi,n } are equicontinuous in C([0, R]).
Multiplying (3.11)1 by wn , (3.11)2 by wi,n  
, (3.11)3 by wm−1,n , (3.11)4 by zn ,
 
(3.11)5 by zi,n , (3.11)6 by zm−1,n , we get
1   2  N − 1
(Wn ) + (Wn )2 + g(Wn ) = 0 ,
2 y
where
g(Wn ) = (w1,n wn , w2,n w1,n
 
, . . . , wnk znp wm−1,n , z1,n zn , z2,n z1,n
 
, . . . , wnq zns zm−1,n ).
Since wn , wi,n

, zn , zi,n

≤ 0, it follows from (3.8) that
1   2 
(Wn ) + Wn ≤ 0 .
2
Integrating the above inequality from 0 to y, we have
y
1   2
Wn (y) + Wn (t)dt ≤ 0 . (3.14)
2 0
and hence

|wn (y)| ≤ c , 
|wi,n (y)| ≤ c , |zn (y)| ≤ c , 
|zi,n (y)| ≤ c ,
i = 1, . . . , m − 1 , (3.15)
uniformly in n, since {wn }, {wi,n

}, {zn }, {zi,n

} are uniformly bounded sequences in
C([0, R̄]). From (3.15), we obtain the equicontinuity of the sequences {wn }, {wi,n
},
486 Y. Zhang and J. Hao NoDEA

{zn }, {zi,n

}. Thus, by Ascoli–Arzela Theorem, and passing to subsequence if nec-
essary, we can assume that

wn → ŵ , wi,n → ŵi , zn → ẑ , zi,n → ẑi ,

as n → ∞, i = 1, . . . , m − 1 , (3.16)
in C([0, R̄]). From (3.11),

y
dξ ξ  
Wn (0) − Wn (y) = η N −1 h Wn (η) dη , (3.17)
0 ξ N −1 0

where
h(Wn ) = (w1,n , w2,n , . . . , wm−1,n , wnk znp , z1,n , z2,n , . . . , zm−1,n , wnq zns ) .
By the Dominated Convergence Theorem, letting n → ∞ in (3.17), we have
y ξ
dξ
Ŵ (0) − Ŵ (y) = N −1
η N −1 h(Ŵ )(η)dη , (3.18)
0 ξ 0

which imply that ŵ ≥ 0, ŵi ≥ 0, ẑ ≥ 0, ẑi ≥ 0 belong to C 1 ([0, R̄]) ∩ C 2 ((0, R̄]) and
satisfy
−(y N −1 Ŵ  ) = y N −1 h(Ŵ ) , (3.19)
for y ∈ [0, R̄], and
ŵ (0) = ŵi (0) = ẑ  (0) = ẑi (0) = 0 , i = 1, . . . , m − 1 . (3.20)
We claim that ŵ, ŵi , ẑ, ẑi can be extended to [0, +∞). In fact we can repeat the
above argument on an interval [0, R∗ ], R∗ > R̄, for the convergent sequences
ŵ, ŵi , ẑ, ẑi on [0, R̄]. We get functions w̃, w̃i , z̃, z̃i , solutions of (3.19) on [0, R∗ ]
that satisfy

ŵ(y) = w̃(y) , ŵi (y) = w̃i (y) , ẑ(y) = z̃(y) , ẑi (y) = z̃i (y) ,
for all y ∈ [0, R̄] , i = 1, . . . , m − 1 .
Clearly, ŵ, ŵi , ẑ, ẑi can be extended to [0, +∞) as solutions of (3.19) that sat-
isfy (3.20), and ŵ(y) ≥ 0, ŵi (y) ≥ 0, ẑ(y) ≥ 0, ẑi (y) ≥ 0 for y ∈ [0, +∞), hence
the claim follows.
We now show that
ŵ(0) > 0 , ŵi (0) > 0 , ẑ(0) > 0 , ẑi (0) > 0 , i = 1, . . . , m − 1 . (3.21)
Without loss generality, suppose that ŵ(0) = 0. Since ŵ(y) ≥ 0 and ŵ (y) ≤ 0, it
follows that ŵ ≡ 0. From (3.19), we have
ŵi ≡ 0 , i = 1, . . . , m − 1 . (3.22)
From the last equation of (3.19), we get −Δẑm−1 = 0, hence ẑm−1 is constant since
ẑm−1 is a nonnegative harmonic function. By the proof of Theorem 1.1, ẑm−1 has
Vol. 15 (2008) Existence for a Polyharmonic System 487

decay property, this implies that ẑm−1 ≡ 0. Similarly, we deduce that ẑ = ẑi ≡ 0,
i = 1, . . . , m − 2. Thus

m−1 
m−1
ŵ(0) + ŵi (0) + ẑ(0) + ẑi (0) = 0 ,
i=1 i=1

which contradicts with (3.13) and we proved (3.21).

By Lemma 3.1, we have

ŵ(y) > 0 , ŵi (y) > 0 , ẑ(y) > 0 , ẑi (y) > 0 ,
for all y ∈ [0, +∞) , i = 1, . . . , m − 1 . (3.23)
Thus we obtain a positive radial classical solution of the following system

(−Δ)m u = uk v p ,
in RN .
(−Δ)m v = uq v s ,
This is impossible in view of Theorem 2.1 and the proof is completed. 
Proposition 3.3. Consider the polyharmonic system
⎧
⎪
⎪ (−Δ)m u = (u + τ )k (v + τ σ )p , in BR ,
⎨
(−Δ)m v = (u + τ )q (v + τ σ )s , in BR ,
(3.24)
⎪
⎪ u = (−Δ)u = · · · = (−Δ)m−1 u = 0 , on ∂BR ,
⎩
v = (−Δ)v = · · · = (−Δ)m−1 v = 0 , on ∂BR ,
where τ ∈ (0, +∞) is a parameter, σ satisfies

1 < k + σp ,
(3.25)
(1 − s)σ < q .
If (3.24) possesses a positive radial solution (u, v), then there is a constant
M > 0, M independent of u and v, such that 0 ≤ τ ≤ M .
Proof. Set ui = (−Δ)i u, vi = (−Δ)i v, i = 1, . . . , m − 1, we rewrite the polyhar-
monic system (3.24) as
⎧
⎪
⎪ Δu + u1 = 0 ,
⎪
⎪
⎪
⎪ Δu i + ui+1 = 0 , i = 1, . . . , m − 2 ,
⎨
Δum−1 + (u + τ )k (v + τ σ )p = 0 ,
(3.26)
⎪ Δv + v1 = 0 ,
⎪
⎪
⎪
⎪
⎪ Δvi + vi+1 = 0 , i = 1, . . . , m − 2 ,
⎩
Δvm−1 + (u + τ )q (v + τ σ )s = 0 .
From the third equation and the last equation in (3.26), we have
−um−1 ≥ cr(u + τ )k (v + τ σ )p ≥ crτ k+σp , (3.27)

−vm−1 ≥ cr(u + τ ) (v + τ ) ≥ crτ
q σ s q+σs
. (3.28)
Integrating (3.27) and (3.28) from 0 to R, we obtain that
um−1 (0) ≥ cR2 τ k+σp , vm−1 (0) ≥ cR2 τ q+σs . (3.29)
488 Y. Zhang and J. Hao NoDEA

Integrating (3.27) from r to R,

um−1 (r) ≥ cτ k+σp (R2 − r2 ) . (3.30)
Again from (3.26),
−um−2 ≥ crum−1 ,
integrating the above inequality from 0 to R, we have
R
um−2 (0) ≥ crum−1 (r)dr
0
R
≥ crτ k+σ+p (R2 − r2 )dr
0
1 4 k+σp
=
cR τ .
4
Repeating the above process, we obtain
u(0) ≥ cτ k+σp . (3.31)
Similarly, we have
v(0) ≥ cτ q+σs . (3.32)
Suppose now that proposition is not true. Then there is a sequence {τn }, τn →
+∞, as n → ∞ such that for each τn there exist solutions un , vn of (3.24).
Denoting as in Proposition 3.2,
tn = sup un (r) = un (0) , ti,n = sup ui,n (r) = ui,n (0) , i = 1, . . . , m − 1 ,
r∈[0,R] r∈[0,R]

sn = sup vn (r) = vn (0) , si,n = sup vi,n (r) = vi,n (0) , i = 1, . . . , m − 1 .

r∈[0,R] r∈[0,R]

and
1

m−1
1 1 
m−1
1
γn = (tn ) λ + (ti,n ) λ+2i + (sn ) μ + (si,n ) μ+2i ,
i=1 i=1
then from (3.31), tn → +∞, and hence γn → +∞ as n → +∞. If now as in
Proposition 3.2 we make the change of variables y, wn (y), wi,n (y), zn (y), zi,n (y)
indicated in (3.6) and (3.7), we have
⎧
⎪
⎪ Δwn + w1,n = 0 ,
⎪
⎪ i = 1, . . . , m − 2 ,
⎪
⎪ Δw i,n + wi+1,n = 0 ,
  
⎪
⎨ Δw τn k τ σ p
m−1,n + wn + γ λ zn + γnnμ = 0 ,
n (3.33)
⎪
⎪ Δzn + z1,n = 0 ,
⎪
⎪
⎪
⎪ Δzi,n + zi+1,n = 0 ,
⎪
⎩ Δz  τn q
  τ σ s
m−1,n + wn + γ λ zn + γnnμ = 0 .
n

Let R̄ be a positive number as in Proposition 3.2 and consider the restriction of

wn , wi,n , zn , zi,n to [0, R̄], which we denote again by wn , wi,n , zn , zi,n . Repeating the
argument of Proposition 3.2, we have that wn , wi,n , zn , zi,n will meet the condition
Vol. 15 (2008) Existence for a Polyharmonic System 489

of Ascoli–Arzela Theorem if we can show from (3.33)3 that the sequence {wm−1,n }
is bounded.

Multiplying (3.33)3 by wm−1,n we obtain
 p 
1  
2  τn τnσ 
(wm−1,n ) + wn + λ zn + μ wm−1 ≤ 0,
2 γn γn
which implies that
 k  p
1   τn τσ 
(wm−1,n )2 + 1 + λ 1 + nμ wm−1 ≤ 0. (3.34)
2 γn γn
Similarly we have
 q  s
1  
2  τn τnσ 
(z ) + 1+ λ 1+ μ zm−1 ≤ 0. (3.35)
2 m−1,n γn γn
 
Thus to bound the sequences {wm−1,n } and {zm−1,n } it is necessary to estimate
σ
τn τn
λ
γn
and μ.
γn
By (3.31) and (3.32), we have
1 1
τn ≤ ct k+σp m , τn ≤ cs q+σs m , (3.36)
and thus
1
τn ct k+σp
λ
≤
γn γnλ
λ
−λ
≤ cγnk+σp
λ( k+σp
1
−1)
= cγn → 0, as n → +∞ , (3.37)
since by (3.25), 1
k+σp − 1 < 0. Similarly,
σ
τnσ cs q+σs
μ ≤
γn γnμ
σμ
−μ
≤ cγnq+σs
μ( q+σs
σ
−1)
= cγn → 0, as n → +∞ . (3.38)
 
Thus (3.37) and (3.38) imply the boundedness of {wm−1,n } and {zm−1,n }.
The rest of the argument is exactly the same as that of Proposition 3.2. Thus
we complete the proof of Proposition 3.3. 

4. Existence
With the help of Proposition 3.2 and 3.3, we are able to use degree theory argument
to prove Theorem 1.1.
The following fixed point theorem on a cone is due to deFueiredo, Lions and
Naussbaum [6], which is a modified version of a theorem of Krasnosel’skii [9].
490 Y. Zhang and J. Hao NoDEA

Theorem 4.1. Let C be a cone in a Banach space X and T : C → C a compact

mapping such that T(0) = 0. Assume that there real numbers 0 < ρ < R and
t0 > 0 such that
(i) x = tT(x) for 0 ≤ t ≤ 1 and x ∈ C, ||x|| = ρ, and
(ii) There exists a compact mapping H : BR × [0, +∞) → C (where Br = {x ∈ C :
||x|| < r}) such that
(a) H(x, 0) = T(x) for ||x|| = R,
(b) H(x, t) = x for ||x|| = R and t ≥ 0,
(c) H(x, t) = x has no solution x ∈ BR for t ≥ t0 .
Then
iC (T, Bρ ) = 1 , iC (T, BR ) = 0 , iC (T, U) = −1 ,
where U = {x ∈ C : ρ < ||x|| < R}, and iC denote the Leray–Schauder index.
Let us introduce the Banach space

X = U = (u, u1 , . . . , um−1 , v, v1 , . . . , vm−1 ) : u ∈ C([0, R]), ui ∈ C([0, R]),

v ∈ C([0, R]), vi ∈ C([0, R]), i = 1, . . . , m − 1 ,
with the norm

m−1 
m−1
|U | = u ∞ + ui ∞ + v ∞ + vi ∞ .
i=1 i=1
Let us denote by C the cone of nonnegative functions

C = U ∈ X : u(r) ≥ 0, ui (r) ≥ 0, v(r) ≥ 0, vi (r) ≥ 0,

∀r ∈ [0, R], i = 1, . . . , m − 1 .
Our system (1.1) is equivalent to U = T(U ) where
 
T(U ) = T0 (U ), . . . , T2m−1 (U ) , (4.1)
where
R ξ
dξ
T0 (U ) = η N −1 u1 (η)dη ,
r ξ N −1 0
R ξ
dξ
Ti (U ) = η N −1 ui+1 (η)dη , i = 1, . . . , m − 2 ,
r ξ N −1 0
R ξ
dξ
Tm−1 (U ) = η N −1 u(η)k v(η)p dη ,
r ξ N −1 0
R ξ
dξ
Tm (U ) = η N −1 v1 (η)dη ,
r ξ N −1 0
R ξ
dξ
Tm+i (U ) = η N −1 vi+1 (η)dη , i = 1, . . . , m − 2 ,
r ξ N −1 0
R ξ
dξ
T2m−1 (U ) = η N −1 u(η)q v(η)s dη .
r ξ N −1 0
Vol. 15 (2008) Existence for a Polyharmonic System 491

It is easy to prove that T maps C in C and T is a compact operator.

The homotopy needed in this case is given by
H(U, τ ) = T ((u + τ, u1 , . . . , um−1 , v + τ σ , v1 , . . . , vm−1 )) ,
where σ > 0 satisfying (3.25). Note that H(U, τ ) is associated with a family of
systems (3.26) parametrized by τ ≥ 0.

Proof of Theorem 1.1. We shall use Theorem 4.1. and divide the proof into two
steps.
Step 1. Verification of condotion (ii).
Condition (a) clearly holds, since H(0, U ) = T(U ), ∀U ∈ C.
Set ui = (−Δ)i u, vi = (−Δ)i v, i = 1, . . . , m−1, we rewrite the polyharmonic
system (3.26) as
R ξ
dξ
u(r) = N −1
η N −1 u1 (η)dη ,
r ξ 0
R ξ
dξ
ui (r) = N −1
η N −1 ui+1 (η)dη , i = 1, . . . , m − 2 ,
r ξ 0
R ξ
dξ  p
um−1 (r) = N −1
η N −1 (u(η) + τ )k v(η) + τ σ dη ,
r ξ 0
R ξ
dξ
v(r) = N −1
η N −1 v1 (η)dη ,
r ξ 0
R ξ
dξ
vi (r) = N −1
η N −1 vi+1 (η)dη , i = 1, . . . , m − 2 ,
r ξ 0
R ξ
dξ  s
vm−1 = N −1
η N −1 (u(η) + τ )q v(η) + τ σ dη .
r ξ 0

Thus (3.26) is equivalent to

U = H(U, τ ) . (4.2)
A slight modification of the proof Proposition 3.2 and Proposition 3.3 tell us
that solutions of (4.2) are a priori bounded in C. Hence there exists R1 > 0 such
that for any τ ∈ [0, +∞) we have that
U = H(U, τ ) ,
with |U | = R1 . So (b) is verified.
If we take τ > M , M is in Proposition 3.3, it follows from this proposition
that (c) is verified.
Step 2. Verification of condition (i).
We claim that there exists a ρ̄ > 0 such that for any 0 < ρ < ρ̄,
U = λT(U ) ,
for λ ∈ [0, 1] and any U ∈ C with |U | = ρ.
492 Y. Zhang and J. Hao NoDEA

Suppose the claim is no true. Then there exist {ρn }, ρn → 0+ and sequences
{λn }, λn ∈ [0, 1], {Un } in C with |Un | = ρn such that
Un = λn Tn , (4.3)
By (4.3), we have that un , ui,n , vn , vi,n satisfy
⎧ R ξ
⎪
⎪ un (0) = 0 ξNdξ−1 0 λn η N −1 u1,n (η)dη ,
⎪
⎪  R dξ  ξ
⎪
⎪ N −1
ui+1,n (η)dη , i = 1, . . . , m − 2 ,
⎪ ui,n (0) = 0 ξN −1 0 λn η
⎪
⎪
⎨ u  R dξ  ξ N −1 k
m−1,n (0) = 0 ξ N −1 0 λn η wn (η)znp (η)dη ,
 R dξ  ξ (4.4)
⎪
⎪ vn (0) = 0 ξN −1 0 λn η N −1 v1,n (η)dη ,
⎪
⎪  R dξ  ξ
⎪
⎪ N −1
vi+1,n (η)dη , i = 1, . . . , m − 2 ,
⎪
⎪ vi,n (0) = 0 ξN −1 0 λn η
⎪
⎩ R dξ ξ
vm−1,n (0) = 0 ξN −1 0 λn η N −1 wnq (η)zns (η)dη .
Noting that un ∞ , ui,n ∞ , vn ∞ , vn ∞ can not be zero, from (4.4), we obtain
un ∞ ≤ cλn u1,n ∞ ,
ui,n ∞ ≤ cλn ui+1,n ∞ , i = 1, . . . , m − 2 ,
um−1,n ∞ ≤ cλn un k∞ vn p∞ ,
vn ∞ ≤ cλn v1,n ∞ ,
vi,n ∞ ≤ cλn vi+1,n ∞ , i = 1, . . . , m − 2 ,
vm−1,n ∞ ≤ cλn un q∞ vn s∞ ,
thus
un 1−k
∞ ≤ cλn vn ∞ ,
m p
vn 1−s
∞ ≤ cλn un ,
m q
(4.5)
which imply that when 0 ≤ k < 1 and 0 ≤ s < 1,
pq
1 ≤ c un (1−k)(1−s) −1
pq
−1
≤ cρn(1−k)(1−s) ,
this leads us to a contradiction when n → ∞. When k = 1 or s = 1, it is obvious
to obtain a contradiction. Hence the claim is proved. Consequently (i) is verified.
Thus we complete the proof of Theorem 1.1. 

References
[1] T. Bartsch, Y. Guo, Existence and nonexistence results for critical growth polyhar-
monic elliptic systems, J. Diff. Eqs. 220 (2006), 531–543.
[2] Ph. Clément, J. Fleckinger, E. Mitidieri and F. de Thélin, Existence of positive
solution for a nonvariational quasilinear elliptic system, J. Diff. Eqs. 166 (2000),
455–477.
[3] Ph. Clément, R. Manasevich and E. Mitidieri, Positive solution for a quasilinear
system via blow-up, Comm. Part. Diff. Eqs. 18 (1993), 2071–2106.
Vol. 15 (2008) Existence for a Polyharmonic System 493

[4] D. G. de Figueiredo, Semilinear elliptic system, Nonlinear Funct. Anal. Appl. Diff.
Equ., World Sci. Publishing, River Edge, NJ,1998, 122–152.
[5] D. G. de Figueiredo, P. L. Felmer, On superquadratic elliptic systems, Trans. Amer.
Math. Soc. 343 (1994), 99–116.
[6] D. G. de Figueiredo, P. Lions, R. Naussbaum, A priori estimates and existence of
positive solutions of semilinear elliptic equations, J. Math. Pures et Appl. 61 (1982),
41–63.
[7] B. Gidas, J. Spruck, A priori bounds for positive solutions of nonlinear elliptic equa-
tions, Comm. Part. Diff. Eqs. 6 (1981), 883–901.
[8] J. Hulshof, R. Van Der Vorst, Differential systems with strongly indefinite variational
structure, J. Func. Anal. 114 (1993), 32–58.
[9] M. A. Krasnosel’skii, Fixed points of cone-compressing and cone-extending operators,
Soviet. Math. Dolk. 1 (1960), 1285–1288.
[10] J. Liu, Y. Guo, Y. Zhang, Liouville-type theorems for polyharmonic system in RN ,
J. Diff. Eqs. 225 (2006), 685–709.
[11] J. Liu, Y. Guo, Y. Zhang, Existence of positive entire solutions for polyharmonic
equations and systems, J. Partial Diff. Eqs. 19 (2006), 256–270.
[12] L. A. Peletier, R. Van Der Vorst, Existence and nonexistence of positive solutions
of nonlinear elliptic systems and the biharmonic equation, Diff. Int. Eq. 5 (1992),
747–767.
[13] M. A. S. Souto, A priori estimates and existence of positive solutions of nonlinear
cooperative elliptic systems, Diff. Int. Eq. 8 (1995), 1245–1258.
[14] X. Xu, Uniqueness theorem for the entire positive solutions of biharmonic equations
in RN , Proc. Royal. Soc. Edinburgh 130A (2000), 651–670.
[15] Y. Zhang, On the nonexistence of positive solutions of polyharmonic systems in RN ,
Nonlinear Anal. 68 (2008), 3454–3464.
[16] Y. Zhang, Liouville type theorems for polyharmonic systems, PhD thesis, Peking
University (2006).
[17] H. Zou, A priori estimates for a semilinear elliptic system without variational struc-
ture and their applications, Math. Ann. 323 (2002), 713–735.

Yajing Zhang
School of Mathematical Sciences
Shanxi University
Taiyuan 030006
P.R. China
and
Institute of Mathematics
AMSS
Chinese Academy of Sciences
Beijing 100080
P.R. China
e-mail: zhangyj@sxu.edu.cn
494 Y. Zhang and J. Hao NoDEA

Jianghao Hao
School of Mathematical Sciences
Shanxi University
Taiyuan 030006
P.R. China
e-mail: hjhao@sxu.edu.cn

Received: 20 September 2007.

Revised: 24 December 2007.
Accepted: 27 January 2008.