RADIAL SYMMETRY OF POSITIVE SOLUTIONS FOR SEMILINEAR ELLIPTIC EQUATIONS IN jRn
YUKI NAITO
ABSTRACT. Symmetry properties of positive solutions for semilin- ear elliptic problems in ]Rn are considered. We give a symmetry result for the problem in the general case, and then derive various results for certain classes of semilinear elliptic equations. We em- ploy the moving plane method based on the maximum principle on unbounded domains to obtain the result on symmetry.
1. Introduction
In this paper, we study the radial symmetry of positive solutions for semilinear elliptic equations in jRn. We consider the problem of the form
{
~u + f(lxl, u) = ° in jRn,
(1.1)
u(x) -+ ° as Ixl -+ 00,
where n ::::: 3. We establish a symmetry result for the problem (1.1) in the general case, and then derive various results for certain classes of semilinear elliptic equations. Typical equations we shall be interested in are as follows:
flu - u + uP = 0, p > 1;
A 1 P _ 0
uu + (1 + Ixl)e u - 0, to::::: 0, p > 1;
flu + (1 + /3 Ixl)/L uP - (1 + 1 Ixl)v u q = ° ' /3 > ° , ,." 11. 1/ > 2 " - ' P q > 1
Some of those results are already known, but our aim is to treat those results from a unified point of view.
Received August 18, 1999.
2000 Mathematics Subject Classification: 35J60.
Key words and phrases: radial symmetry, semilinear elliptic equation, moving
plane method.
752 YUki Naito
Our arguments are based on the moving plane method. The method was first developed by Serrin [13] in PDE theory, and later extended and generalized by Gidas, Ni, and Nirenberg [2, 3]. In this paper, we employ the moving plane method based on the maximum principle on unbounded domains to obtain the result on symmetry.
In Section 2, we state the main theorem, and give some corollaries of the theorem. In Section 3, we prove the theorem by using the method . of moving planes.
2. Statement of the results
In (1.1), we assume that f(r, u) is continuous and Cl in u ~ 0, and that f(r, u) is nonincreasing in r > 0 for each fixed u ~ o. Our main result is the following:
THEOREM 1. Let u E C 2 (]Rn) be a positive solution of (1.1). Define U and <I> as
(2.1)
U(r) = sup{u(x) : Ixl ~ r} and <I>(r) = sup{fu(r, s) : 0:::; s:::; U(r)}, respectively. Assume tbat tbere exists a positive function W on Ixl ~ Ho for some Ho > 0 satisfying
(2.2) (2.3)
~W + <I>(lxl)W :::; 0 in Ixl > R o and lim U(lxl) = O.
Ixl-+oo W(x)
Tben u must be radially symmetric about some point Xo E Rn and UT < 0 for r = Ix - xol > o.
We give some corollaries of the theorem. Some of the results presented here are already known, but our aim is to treat these results from a unified point of view.
COROLLARY 1. Assume tbat
fu(r,u):::; 0 forr ~ ro, 0:::; u:::; uo,
witb some constants ro ~ 0 and Uo > O. Let u be a positive solution of (1.1). Tben u must be radia1ly symmetric about some point Xo E ]Rn
and UT < 0 for r = Ix - xol > o.
REMARK. This result has been obtained by [9]. Related results have been obtained by [3, 5].
Proof. We see that the function U defined by (2.1) satisfies U(r) -+ 0 as r -+ 00. Take Ro 2: ro so large that U(r) ::; Uo for r 2: Ro. Define W as W(x) == 1 on Ixl 2: Ro. Then W satisfies (2.3). Since <P(r) = max{fu(r, u) : 0 ::; u ::; U(r)} ::; 0 for r 2: Ro, we have (2.2). Therefore, we can apply Theorem 1 to conclude the assertion.
For simplicity we consider the equation of the form (2.4) D.u + </>(Ixl)f(u) = 0
In equation (2.4), we assume that </> E e[O, (0) satisfies
</>(r) 2: 0 for r 2: 0 and </>(r) is nonincreasing in r> 0, and that f E G 1 [0, (0) with f(u) > 0 for u > o. 0
COROLLARY 2. Assume, furthermore, that
</>(r) = O(r f ) If'(u)1 = O(u p - 1 )
as r -+ 00 for some f E [-2,0]" and as u -+ 0 for some p > ~~;.
Let u be a positive solution of (2.4) in !Rn satisfying
as Ixl -+ 00 if - 2 < f ::; 0
aslxl-+oo iff=-2.
Then u must be radia1ly symmetric about some point Xo E Rn and U r < 0 for r = Ix - xol > O.
REMARK. Related results have been obtained by [3, 5, 7,8]. Li[6] and Zou[14, 15] have obtained the symmetry results for positive solutions which have slow decay at infinity.
Proof. We see that the function U defined by (2.1) satisfies
as r -+ 00 if - 2 < f ::; 0,
as r -+ 00 if f = -2.
754 Yilki Naito
(i) The case where -2 < f ::; O. Let W(x) = Ixl-~=~ for Ixl > o. Then W satisfies (2.3) and
.6.W+ m (n-2-m)W=0 Ixl>O,
Ixl 2 '
where m = (f + 2)j(n - 2). We note that n - 2 - m > 0 from p >
(n + f)j(n - 2). We find that
<P(r) = max{et>(r)J'(u) : 0 ::; u ::; U(r)} = O(lxj-2) as r -+ 00.
Then we have (2.2) for sufficiently large Ro > o. Therefore, we can apply Theorem 1 to conclude the assertion.
(ii) The case where f = -2. Let W(x) = (log Ixl)-P-l for Ixl 1 > l.
Then W satisfies (2.3) and
.6.W + (; =~ - (p -l~lOg Ix') IX12l~g Ixl W = 0, Ixl > l.
We find that
<P(r) = 0 (lxl-2(log Ixlt 1 ) as r -+ 00.
Then we have (2.2) for sufficiently large Ro > o. Therefore, we can apply
Theorem 1 to conclude the assertion. 0
COROLLARY 3. In equation (2.4), assume furthermore that et> t= 0
satisfies
(2.5) 100 ret>(r)dr < 00.
Let u be a positive solution of (2.4) in lR n satisfyingu(x) -+ cas lxl -+ 00
for some constant c 2: O. Then u must be radia1ly symmetric about the origin and UT < 0 for r = Ixl > O.
REMARK. Related results have been obtained by [7, 8, 11]. If et>
is locally Holder continuous, then every bounded positive solution u satisfies u(x) -+ c as Ixl -+ 00 for some constant c 2: O. (See [11].)
Proof.
(2.6)
Let v(x) = u(x) - c. Then v satisfies
{
.6.v + et>(lxl)h(v) = 0 in lR n ,
v(x) -+ 0 as Ixl -+ 00,
where h(v) = g(v + c). Since -~v = cjJh :::: 0, we have v > 0 in lR. n by the maximum principle. We apply Theorem 1 to the problem (Z.6). We define U and <I> as
U(r) = sup{v(x) : Ixl :::: r} and <I>(r) = sup{cjJ(r)h'(s) : 0:::; s:::; U(r)}, respectively. Since <I>(r) :::; McjJ(r) for some constant M > 0 and (2.5) holds, we have
100 r<I>(r)dr < 00.
Then there exists a positive function W on Ixl :::: Ra for some Ra > 0 satisfying
~W + <I>(lxl)W = 0 and liminf W(x) > o.
Ixl->oo
(See, e.g., [11, Lemma Bol].) Then W satisfies (2.2) and (2.3). Therefore, Theorem 1 can be applied to conclude the assertion. 0
Finally, we consider the semilinear elliptic equations of the form
(2 7) A f3 p 1 q - 0
0 lJ1)no