NOTES ON ASYMPTOTIC BEHAVIOR OF RADIAL SOLUTIONS FOR

SOME WEIGHTED ELLIPTIC EQUATIONS ON THE ANNULUS

FUTOSHI TAKAHASHI
arXiv:2405.18854v1 [math.AP] 29 May 2024

Abstract. In this paper, we study the asymptotic behavior of radial solutions for several
weighted elliptic equations with power type or exponential type nonlinearities on an annulus.

1. Introduction.
Let A = {x ∈ RN | 0 < a < |x| < b} be an annulus in RN , N ≥ 2. In this paper, we study
the following elliptic problem with the inverse-square weight in R2 :
up (x)

−∆u = , x ∈ A ⊂ R2 ,
|x|2


(1)

u(x) > 0, x ∈ A,
u(x) = 0, x ∈ ∂A,


where p > 1. In higher dimensions N ≥ 3, we also consider

up (x)

−∆u = (N − 2)2 2(N −1) ,
 x ∈ A ⊂ RN ,
 |x|
(2)

 u(x) > 0, x ∈ A,
u(x) = 0, x ∈ ∂A.


1
Since the embedding of the radial Sobolev space H0,rad (A) ֒→ Lp+1 (A) is compact for any
p > 1, we can easily obtain a radially symmetric solution to (1) or (2), by considering a
suitable minimization problem. Later it will be shown that the positive radial solution is
unique for both problems and it is denoted by up . In this paper, we study the asymptotic
behavior of the radial solution up to (1) or (2) as p → ∞ and p ց 1. In the following, we
denote
Z 1
ds
Lp = √ , (p > 0). (3)
0 1 − sp+1
Theorem 1. For any p > 1, the problem (1) admits a unique radially symmetric solution
up (x) = up (|x|). Furthermore, it holds that
( √
2 log |x| − log a, for a ≤ |x| ≤ ab
up (|x|) → √ (4)
log b − log a log b − log |x|, for ab ≤ |x| ≤ b

Date: May 30, 2024.

2020 Mathematics Subject Classification. Primary 34C23; Secondary 37G99.
Key words and phrases. Weighted elliptic equations, radial solutions, asymptotic behavior, Green’s function.
1
2 FUTOSHI TAKAHASHI

in C 0 (A) as p → ∞, and
 2 
 p−1  1    
2 p + 1 p−1 p−1
2
π(log b − log |x|)
up (|x|) = Lp sin + op (1) (5)
log b − log a 2 log b − log a
in C 0 (A) as p ց 1, where op (1) → 0 in C 0 (A) as p ց 1.
Theorem 2. For any p > 1, the problem (2) admits a unique radially symmetric solution
up (x) = up (|x|). Furthermore, it holds that
1
  2−N 2−N  2−N
2−N 2−N a +b
a − |x| , for a ≤ |x| ≤

2 
2
up (|x|) → 2−N 1 (6)
a − b2−N 
 2−N 2−N  2−N
|x|2−N − b2−N , a +b
for 2
≤ |x| ≤ b

in C 0 (A) as p → ∞, and
2   1
π(|x|2−N − b2−N )
  p−1    
2 p + 1 p−1 p−1
2
up (|x|) = Lp sin + op (1) (7)
a2−N − b2−N 2 a2−N − b2−N
in C 0 (A) as p ց 1, where op (1) → 0 in C 0 (A) as p ց 1.
Note that in both Theorem 1 and Theorem 2, L∞ -blow up of up does not occur when
p → ∞. On the other hand, kup kL∞ (A) could diverge as p ց 1 if the annulus is “thin” in the
sense that log b − log a < 2 when N = 2, or a2−N − b2−N < 2 when N ≥ 3.
Also there are local convergence results as follows.
Theorem 3. Let up (x) = up (r), r = |x|, be the unique radial solution to (1) for p > 1. Define
εp > 0 and ũp by


 pε2p kup kp−1
L∞ (A) ≡ 1,
 n  log ab
 o
u˜p (t) = kup kLp∞ (A) up e−(εp t− 2 ) − kup kL∞ (A) , (8)

 t ∈ I = (− log b−log a , log b−log a ).

p εp εp

Then εp → 0 and √
4e 2t 1
u˜p (t) → U(t) = log √ in Cloc (R)
(1 + e 2t )2
as p → ∞.
Theorem 4. Let up (x) = up (r), r = |x|, be the unique radial solution to (2) for p > 1. Define
εp > 0 and ũp by


 pε2p kup kp−1
L∞ (A) ≡ 1,

    1  
p a2−N +b2−N 2−N
u˜p (t) = kup kL∞ (A) up εp t − 2
− kup kL∞ (A) , (9)


 t ∈ I = (− a2−N −b2−N , a2−N −b2−N ).

p εp εp

Then εp → 0 and √
4e 2t 1
u˜p (t) → U(t) = log √ in Cloc (R)
(1 + e 2t )2
as p → ∞.
 ASYMPTOTIC BEHAVIOR OF RADIAL SOLUTIONS 3

Note that in Theorem 3 and Theorem 4, U(t) is the unique solution of the 1-D Liouville
equation
(
−U ′′ (t) = eU (t) , t ∈ R,
(10)
U(0) = U ′ (0) = 0
√
with R eU (t) dt = 2 2.
R
In [3], M. Grossi considered the asymptotic behavior as p → ∞ of radial solutions up of the
problem

p
−∆u = u ,
 in A ⊂ RN ,
u > 0, in A,

u = 0, on ∂A
on an annulus A = {a < |x| < b} in RN , N ≥ 2. He proved that the same asymptotic formulae
(4), (6) hold true for up according to the case N = 2 or N ≥ 3. Note that the limit functions
in (4), (6) are the constant multiple of Ga,b (|x|, r0 ), where Ga,b (r, s) is the Green’s function of
d2 N −1 d
the operator − dr 2 − r dr
on r ∈ (a, b) defined by
(
sN −1 (b2−N − s2−N )(r 2−N − a2−N ), for a ≤ r ≤ s,
Ga,b (r, s) = (N ≥ 3),
(N − 2)(a2−N − b2−N ) (s2−N − b2−N )(b2−N − r 2−N ), for s < r ≤ b,
(
s (log r − log a)(log b − log s), for a ≤ r ≤ s,
Ga,b (r, s) = (N = 2), (11)
log b − log a (log s − log a)(log b − log r), for s < r ≤ b,

and r0 ∈ (a, b) is defined by

   1
a2−N +b2−N 2−N
, (N ≥ 3),

r0 = √ 2 (12)
 ab, (N = 2).

Furthermore he proved the local convergence result that

√
4e 2t 1
u˜p (t) → U(t) = log √ in Cloc (R),
(1 + e 2t )2
where εp > 0 and u˜p are defined as


 pε2p kup kp−1
L∞ (A) ≡ 1,

p
 u˜ (t) = 
up (εp t + rp ) − kup kL∞ (A) ,

p kup kL∞ (A)


 up (rp ) = kup kL∞ (A) , rp ∈ (a, b),
 t ∈ ( a−rp , b−rp ).

εp εp

Theorems 1-4 are strongly related with results in [3] and the proofs are much simpler.
The key ingredients are the explicit norm computations and the asymptotic behaviors of the
unique solution Wp of 1-D Emden equation (13) as p → ∞ or p ց 1. Based on these, we
prove theorems via the well-known transformations.
4 FUTOSHI TAKAHASHI

2. Asymptotic behaviors of Wp .
For any p > 1 and ā, b̄ ∈ R with ā < b̄, let Wp denote the unique function satisfying

′′ p
−Wp (s) = Wp (s), s ∈ I = (ā, b̄),

Wp (s) > 0, s ∈ I, (13)

W (ā) = W (b̄) = 0.
p p

By [1], Wp is symmetric with respect to s = s0 = ā+2 b̄ and kWp kL∞ (ā,b̄) = Wp ( ā+2 b̄ ). Also Wp is
strictly monotone increasing on (ā, s0 ) and strictly monotone decreasing on (s0 , b̄).

The following explicit norm computations are done by T. Shibata [5]. For the readers
convenience, we give the proof here.

Proposition 1. ([5]) For a given p > 1, let Wp be the unique solution of the problem (13).
Then
2
! p−1 r
p+1 2
ξp = kWp kL∞ (ā,b̄) = Lp , (14)
2 b̄ − ā
r  
q 2 2q−p+1 q+1 1
kWp kLq (ā,b̄) = ξp 2
B , , (q > 0) (15)
p+1 p+1 2
R1
where Lp is defined in (3) and B(x, y) = 0
tx−1 (1 − t)y−1 dt denotes the Beta function.

Especially by (15) with q = p, we see

 p+1 
 p−1 1
 p−1
2 p+1 1 2π
kWp kpLp (ā,b̄) = Lp B(1, ) → (16)
b̄ − ā 2 2 b̄ − ā
as p → ∞.

Proof. We use a time-map method to compute kWp kLq (ā,b̄) .

Multiply the equation by Wp′ (s), we have
 ′
1 1
− (Wp′ (s))2 − Wpp+1 (s) = 0,
2 p+1

thus
1 1 1 p+1
− (Wp′ (s))2 − Wpp+1(s) = constant = 0 − ξ .
2 p+1 p+1 p

Therefore, we have
r
2
Wp′ (s) = (ξpp+1 − Wpp+1 (s)). (17)
p+1
 ASYMPTOTIC BEHAVIOR OF RADIAL SOLUTIONS 5

By (17) and the change of variables t = Wp (s), we compute

Z ā+b̄ Z ā+b̄
b̄ − ā 2 2 Wp′ (s)ds
= 1ds = q
2 ā ā 2
(ξ p+1 − W p+1 (s))
p+1 p p
r ξp
p+1
Z
dt
= q
2 0 ξpp+1 − tp+1
r 1
p + 1 1−p
Z
dτ
= ξp 2 √ .
2 0 1 − τ p+1
From this, we obtain
r 2
! p−1
2 p+1
ξp = kWp kL∞ (ā,b̄) = Lp .
b̄ − ā 2
Next, by the formula ξp and the change of variables t = Wp (s) again, we see
Z ā+b̄
2
q
kWp kLq (ā,b̄) = 2 |Wp (s)|q ds
ā
ā+b̄
Wp′ (s)ds
Z
2
=2 Wpq (s) q
ā 2
(ξ p+1
p+1 p
− Wpp+1 (s))
ξp
tq dt
p Z
= 2(p + 1) q
ξpp+1 − tp+1
0

Z 1
p 2q−p+1 τ q dt
= 2(p + 1)ξp 2
√
0 1 − τ p+1
r  
2 2q−p+1 q+1 1
= ξp 2
B , ,
p+1 p+1 2
which proves the formula of kWp kLq (ā,b̄) . 
The following two propositions were essentially proven by M. Grossi [3]; see Theorem 1.1,
and Proposition 3.2 in [3]. Though Theorem 1.1 (1.3) in [3], N ≥ 3 is assumed but the proof
there works for the case N = 1. Also when N = 1, ā > 0 is not needed. We give proofs here
for the sake of the convenience of the readers.
Proposition 2. Let Wp be the unique solution of the problem (13). Then
(
(s − ā) ā ≤ s ≤ ā+2 b̄
   
4 ā + b̄ 2
Wp (s) → G s, = ā+b̄ in C 0 [ā, b̄]
b̄ − ā 2 b̄ − ā (b̄ − s) 2
≤ s ≤ b̄
d 2
as p → ∞ holds true. Here G(s, t) is the Green’s function of the operator − ds 2 with the

Dirichlet boundary condition on [ā, b̄]:

(
1 (s − ā)(b̄ − t), ā ≤ s ≤ t ≤ b̄,
G(s, t) = (18)
b̄ − ā (b̄ − s)(t − ā), ā ≤ t ≤ s ≤ b̄.
6 FUTOSHI TAKAHASHI

Proof. Put I = (ā, b̄). For ā < r < s0 = ā+2 b̄ , we have from (13) that
Z s0 Z s0 Z b̄
′′
−Wp (s)ds = p
Wp (s)ds ≤ Wpp (s)ds = kWp kpLp (I) .
r r ā

Thus we have Wp′ (r)

≤ kWp kpLp (I) , (ā < ∀r < s0 ). Similarly, we see −Wp′ (r) ≤ kWp kpLp (I) ,
(s0 < ∀r < b̄), so we obtain
|Wp′ (s)| ≤ kWp kpLp (I) = O(1) (ā < ∀s < b̄) (19)
as p → ∞ by (16). Then by Ascoli-Arzela Theorem, there exists a function W : I → R such
that
Wp → W uniformly on I¯ = [ā, b̄]
as p → ∞ (up to a subsequence). Since kWp kL∞ (I) = ξp → 1 as p → ∞ by Proposition 1, we
assure that kW kL∞ (I) = W (s0 ) = 1.
Now, we claim that for any s ∈ [ā, b̄], s 6= s0 = ā+2 b̄ , there exists p0 > 1 such that
Wp (s) < 1 (p > p0 ) (20)
holds true. We prove the claim for s < s0 first. Indeed, assume the contrary that there exists
s̄ < s0 and pn → ∞ such that Wpn (s̄) ≥ 1 for any n ∈ N. Since Wpn is strictly monotone
increasing on (s̄, s0 ), we obtain Wpn (t) > Wpn (s̄) ≥ 1 for any t ∈ (s̄, s0 ). Then there exists
ε0 > 0 such that Wpn (t) > 1 + ε0 for any t ∈ (s̄, s0 ), which implies that
Z s0 Z s0
′ ′′
Wpn (s̄) = −Wpn (s)ds = Wppnn (s)ds ≥ (1 + ε0 )pn (s0 − s̄) → +∞
s̄ s̄

as pn → ∞. This contradicts to (19). The proof when s > s0 is similar. Therefore, we obtain
(20).
Note that by the equation of Wp and (20), the convergence Wp → W actually holds in
1
C ([ā, b̄] \ {s0 }). Thus letting p → ∞ with (20) in (13), we obtain that the limit function
satisfies that 
′′
−W (s) = 0, s ∈ I = (ā, b̄) \ {s0 },

W (ā) = W (b̄) = 0,

W (s ) = 1.
0

From this, we have the conclusion that

(  
2 (s − ā), (ā ≤ s < s0 ) 4
W (s) = = G(s, s0 ),
b̄ − ā (b̄ − s), (s0 < s ≤ b̄) b̄ − ā
d 2
where G(s, t) is the Green’s function of the operator − ds 2 with the Dirichlet boundary condi-

tion on [ā, b̄] defined in (18). 

Proposition 3. Let Wp be the unique solution of the problem (13). Define εp > 0 and W̃p be
such that 
2 p−1
 pεp kWp kL∞ (ā,b̄) ≡ 1,


W̃p (t) = kWp k p∞ {Wp (εp t + s0 ) − Wp (s0 )} , (21)
 L (ā,b̄)
 t ∈ Ip = ( ā−s0 , b̄−s0 ).

εp εp
 ASYMPTOTIC BEHAVIOR OF RADIAL SOLUTIONS 7

Then εp → 0 and
√
4e 2t 1
W̃p (t) → U(t) = log √ in Cloc (R)
(1 + e 2t )2
as p → ∞.
Proof. By Proposition 1 (14) and Lp > 1 for any p > 1, we observe
r !2  2  
p−1 p−1 2 p+1 2 p+1
kWp kL∞ (ā,b̄) = ξp = Lp > .
b̄ − ā 2 b̄ − ā 2
1
1
Thus ξp ≥ Cp p−1 and εp = √ ≤ C ′ ( p1 ) for some C, C ′ > 0. Then εp → 0 and “Ip → R”
pξpp−1
as p → ∞. Direct computation shows that W̃p solves the equation
  p
′′ W̃p (t)


 −W̃p (t) = 1 + p ,

ā−s0 b̄−s0
t ∈ Ip = ( εp , εp ) =: (āp , b̄p ),

(22)


 W̃p (āp ) = W (b̄p ) = 0,

W̃p (0) = 0, W̃p′ (0) = 0.


Also by (19) and the definition of εp , we see

p
|W̃p′ (t)| = εp |Wp′ (εp t + s0 )|
kWp kL∞ (ā,b̄)
1
p2
= p+1 |Wp′ (s)|
s=εp t+s0
kWp k 2
L∞ (ā,b̄)
1
p2 1
≤C 1 p+1
= Cp− p−1 = O(1)
(p p−1 ) 2

for any t ∈ Ip as p → ∞. Thus by Ascoli-Arzela Theorem, we assure that there exists a

function U : R → R such that
W̃p → U uniformly on compact sets on R.

By using the equation of W̃p , we obtain the uniform boundedness of W̃p′′ in p so this convergence
1
holds actually on Cloc (R). Letting p → ∞ in (22), we see that the limit function satisfies
(
−U ′′ (t) = eU (t) , t ∈ R,
U(0) = U ′ (0) = 0.

By the uniqueness of the initial value problem of the second order ODE, we conclude that
√
4e 2t
U(t) = log √ .
(1 + e 2t )2

8 FUTOSHI TAKAHASHI

Next proposition concerns the behavior of Wp as p ց 1 and is proven in [4]. In [4], the
asymptotic behavior as p ց 1 of solutions of

p N
−∆u = u in Ω ⊂ R ,

u > 0 in Ω ⊂ RN ,

u = 0 on ∂Ω,

is considered, where Ω is a bounded convex domain. Here we give a simple proof of the fact
in one-dimensional case.
Proposition 4. Let Wp be the unique solution of the problem (13). Then
   
π(s − ā)
Wp (s) = kWp kL∞ (ā,b̄) sin + op (1)
b̄ − ā
 2 
 p−1  1    
2 p + 1 p−1 p−1 2
π(s − ā)
= Lp sin + op (1)
b̄ − ā 2 b̄ − ā
in C 1 ([ā, b̄]) as p ց 1, where op (1) → 0 in C 1 ([ā, b̄]) as p ց 1.
Wp
Proof. Let wp = = Wξpp . Then kwp kL∞ (ā,b̄) = 1 and wp solves
kWp kL∞ (ā,b̄)

′′ p−1 p
−wp (s) = ξp wp (s), s ∈ I = (ā, b̄),

wp (s) > 0, s ∈ I, (23)

w (ā) = w (b̄) = 0.
p p

By Proposition 1 (14), we see

 2    2
2 p+1 π
ξpp−1 = L2p →
b̄ − ā 2 b̄ − ā
R 1 ds π ′′
as p ց 1. Note that L1 = 0 √1−s 2 = 2 . Thus wp is uniformly bounded on [ā, b̄] and the

Ascoli-Arzela Theorem implies that there exists a function w : [ā, b̄] → R such that wp → w
as p ց 1 in C 1 ([ā, b̄]). Taking a limit p ց 1 in (23), we see w satisfies
  2
′′ π
−w (s) = b̄−ā w(s), s ∈ I = (ā, b̄),



 w(s) > 0, s ∈ I,

w(ā) = w(b̄) = 0.

d 2
Thus w is the first eigenfunction of − ds 2 with the Dirichlet boundary condition on I = (ā, b̄),

which leads to  
π(s − ā)
w(s) = C sin , in C 1 ([ā, b̄])
b̄ − ā
for some C > 0. Since kwkL∞ (I) = C = limpց1 kwp kL∞ (I) = 1, we conclude
 
Wp (s) π(s − ā)
lim wp = lim = sin
p→1 p→1 ξp b̄ − ā
in C 1 ([ā, b̄]), which ends the proof. 
 ASYMPTOTIC BEHAVIOR OF RADIAL SOLUTIONS 9

3. Proofs of Theorems.
Theorem 1-4 in §1 will be proven by combining well-known transformations and the asymp-
totic behaviors of Wp as p → ∞, or p ց 1 in §2. First we prove Theorem 1.
Proof. For a radially symmetric solution up (|x|) = up (r) to (1), define a new function vp as
up (r) = vp (s), s = − log r, s ∈ (− log b, − log a). (24)
Then a direct computation shows that vp solves

′′ p
−vp (s) = vp (s), s ∈ (− log b, − log a),

vp (s) > 0, s ∈ (− log b, − log a),

v (− log b) = v (− log a) = 0.
p p

By the uniqueness, vp ≡ Wp , where Wp is the unique solution of (13) with ā = − log b,

b̄ = − log a. Thus the radial solution up to (1) is unique for any p > 1 and is represented as
up (r) = Wp (s), s = − log r, s ∈ (− log b, − log a). (25)
Applying Proposition 2 to Wp and going back to up with s = − log |x|, we obtain (4). On the
other hand, applying Proposition 4 to Wp and going back to up with s = − log |x| and (14),
we obtain (5). 
Proof of Theorem 2 is similar:
Proof. For a radially symmetric solution up (|x|) = up (r) to (2), define a new function vp as
up (r) = vp (s), s = r 2−N , s ∈ (b2−N , a2−N ). (26)
Then a direct computation and the uniqueness of Wp assures that
up (r) = Wp (s), s = r 2−N , s ∈ (b2−N , a2−N ), (27)
where Wp is the unique solution of (13) with ā = b2−N , b̄ = a2−N . Applying Proposition 2 to
Wp and going back to up with s = |x|2−N , we obtain (6). Applying Proposition 4 to Wp with
(14), we obtain (7). 
Proofs of Theorem 3 and Theorem 4 is as follows:
Proof. By (25) and Proposition 3 with ā = − log b, b̄ = − log a, we obtain the conclusion of
Theorem 3. Note that kup kL∞ (A) = kWp kL∞ (ā,b̄) and s = εp t + s0 = − log r, so (21) can be
rewritten as (8).
For the proof of Theorem 4, just we use (27) and Proposition 3 with ā = b2−N , b̄ = a2−N .
In this case (21) is the same as (9). 

4. Other application.
Let N ≥ 1 and consider the following elliptic equation on an annulus:
(N−1)(p−1)

CN
−∆u − |x|2 u = |x|
 2 up , x ∈ A = {a < |x| < b} ⊂ RN ,
u > 0, x ∈ A, (28)

u = 0 x ∈ ∂A,


(N −1)(N −3)
where CN = 4
.
10 FUTOSHI TAKAHASHI

Theorem 5. For any p > 1, the positive radially symmetric solution up (|x|) = up (r) of (28)
exists and unique. Moreover, the following asymptotic behaviors hold:
(i) As p → ∞, it holds
(
(|x| − a), (a ≤ |x| ≤ a+b
   
4 N−1 a+b 2 N−1
2
)
up (|x|) → |x| 2 G |x|, = |x| 2 a+b
b−a 2 b−a (b − |x|), ( 2 ≤ |x| ≤ b)
in C 0 (A).
(ii) Put εp > 0 and ũp such that
 N−1

 pε2p k| · |− 2 up kL∞ (A) ≡n1, o
N−1 N−1

ũp (t) = − N−1 p (εp t + s0 )− 2 up (εp t + s0 ) − k| · |− 2 up kL∞ (A) ,
 k|·| 2 up kL∞ (A)
 t ∈ (− b−a , b−a ).

εp εp

Then εp → 0 as p → ∞ and
√
4e 2t 1
ũp (t) → U(t) = log √ in Cloc (R)
(1 + e 2t )2
as p → ∞.
(iii) As p ց 1, it holds
   
N−1
− N−1 π(|x| − a)
up (|x|) = |x| 2 k| · | 2 up kL∞ (A) sin + op (1)
b−a

in C 1 (A), where op (1) → 0 in C 1 (A) as p ց 1.

Proof. Let up (|x|) = up (r) be any positive radial solution to (28) and put a new function vp
by
N−1
vp (r) = r − 2 up (r).
Then a direct computation shows that

′′ p
−vp (r) = vp (r), r ∈ (a, b),

vp (r) > 0, r ∈ (a, b),

v (a) = v (b) = 0.
p p

Thus we have vp ≡ Wp , where Wp is the unique solution to (13) with ā = a, b̄ = b. Therefore

N−1
positive radial solution up to (28) is unique and can be written as up (r) = r 2 Wp (r). Then
we obtain the results by applying Proposition 2-4. 
5. Exponential nonlinearity case.
In [2], the authors considered the Liouville equation
(
−∆u = λeu , in A,
(29)
u = 0, on ∂A,
where λ > 0 and A = {x ∈ RN | 0 < a < |x| < b} is an annulus in RN , N ≥ 2. It is well-known
that the problem (29) adimits no solution if λ is sufficiently large, but there exist at least two
radially symmetric solutions uλ and Uλ for λ > 0 sufficiently small. uλ is called the minimal
 ASYMPTOTIC BEHAVIOR OF RADIAL SOLUTIONS 11

solution to (29), which is strictly stable and bounded, and the other solution Uλ is unstable and
can be obtained, for example, by the Mountain Pass Theorem applied in the radial Sobolev
1
space H0,rad (A). The authors in [2] studied the asymptotic behavior of non-minimal (blowing-
up) radially symmetric solutions Uλ to (29) as λ → +0, and obtained the following R results: Let
Uλ (r) = Uλ (|x|) be any radially symmetric solution to (29) such that γλ = λ A eUλ dx → +∞
as λ → +0. Define δλ > 0 by λδλ2 ekUλ kL∞ (a,b) ≡ 1 and Ũλ (t) = Uλ (δλ t + rλ ) − Uλ (rλ ),
t ∈ ( a−r
δλ
λ b−rλ
, δλ ), where rλ ∈ (a, b) be such that Uλ (rλ ) = kUλ kL∞ (a,b) . Then
√
4e 2t 1
Ũλ (t) → U(t) = log √ in Cloc (R),
(1 + e 2t )2
and √
δλ Uλ (|x|) → 2 2Ga,b (r, r0 )
d 2 N −1 d
in C 0 (A) as λ → +0, where Ga,b (r, s) is the Green’s function of − dr 2− r dr
with the Dirichlet
boundary conditions defined in (11) and r0 is in (12).

For 1-D case, let ā, b̄ ∈ R with ā < b̄ and consider

(
−w ′′ (s) = λew , s ∈ I = (ā, b̄),
(30)
w(ā) = w(b̄) = 0.
By the maximum principle, any solution to (30) is positive on I, and by [1], it is symmetric
with respect to s = s0 = ā+2 b̄ , which is the unique maximum point of w in I. It is a classical fact
that (see for example [6]), there exists λ∗ > 0 such that (30) adimits no solution for λ > λ∗ ,
only one solution for λ = λ∗ , just two solutions wλ and Wλ for 0 < λ < λ∗ . Here w λ is the
minimal solution, and Wλ is the unique unstable solution which satisfies kWλ kL∞ (I) → +∞
as λ → +0.
The following is the one-dimensional counter part of the results in [2], and the proof of it
is much simpler.
Proposition 5. Let λ > 0 sufficiently small and let Wλ be the unique unstable solution of the
problem (30). Then we have
Z b̄
γλ = λ eWλ (s) ds → +∞ (λ → +0).
ā

Define δλ > 0 and W̃λ such that


2 kWλ kL∞ (ā,b̄)
 λδλ e
 ≡ 1,
ā+b̄
W̃λ (t) = Wλ (δλ t + s0 ) − Wλ (s0 ), s0 = 2
, (31)
 t ∈ I = ( ā−s0 , b̄−s0 ).

λ δλ δλ

Then δλ → 0 and √
4e 2t 1
W̃λ (t) → U(t) = log √ in Cloc (R) (32)
(1 + e )2t 2

as λ → +0. Furthermore,
(
√ √ (s − ā), (ā ≤ s ≤ ā+b̄ ),
δλ Wλ (s) → 2 2G (s, s0 ) = 2 2 in C 0 [ā, b̄] (33)
(b̄ − s), ( ā+2 b̄ ≤ s ≤ b̄),
12 FUTOSHI TAKAHASHI

d 2
holds as λ → +0. Here G(s, t) is the Green’s function of the operator − ds 2 with the Dirichlet

boundary condition on [ā, b̄] defined by (18).

Proof. Integrating the equation −Wλ′′ (s) = λeWλ (s) on [s0 , s] when s > s0 and on [s, s0 ] when
s < s0 , we have as in the derivation of (19) that
Z b̄
′
|Wλ (s)| ≤ γλ = λ eWλ (s) ds, ∀s ∈ I. (34)
ā
Therefore, if γλ = O(1) as λ → +0, Ascoli-Arzela Theorem implies that there exists W0 such
that Wλ → W0 uniformly on I as λ → +0 (up to a subsequence). By Green’s representation
formula, we see Z
Wλ (s) = λ G(s, t)eWλ (t) dt
I
for s ∈ I. Letting λ → +0, we obtain W0 (s) = 0 for any s ∈ I. Also it is known that the
minimal solution uλ to (30) satisfies limλ→0 uλ = 0 uniformly on I. This fact contradicts to
the local uniqueness of the minimal solution uλ , since the minimal solution could be obtained
by the implicit function theorem, which results in the local uniqueness of uλ . Thus we have
R b̄
proved that γλ → ∞ as λ → +0. Since γλ = λ ā eWλ (s) ds ≤ λ(b̄ − ā)ekWλ kL∞ (I) , we have
1
δλ2 = → 0, (λ → 0).
λekWλ kL∞ (I)
Next, integrating the equation −Wλ′′ (s) = λeWλ (s) on [ā, b̄] and noting that Wλ′ (ā) = −Wλ′ (b̄),
we see γλ = Wλ′ (ā)−Wλ′ (b̄) = 2Wλ′ (ā). Also, multiplying Wλ′ (s) to the equation and integrating
on [ā, s0 ], we have
Z s0  ′ Z s0
1 ′ 2

′
− (Wλ (s)) ds = λ eWλ (s) ds,
ā 2 ā
so
1
− 0 − (Wλ′ (a))2 = λ eWλ (s0 ) − 1 .


2
2
Multiplying δλ to the above and noting the definition of δλ , we get
 2
1 1
γλ δλ = 1 − λδλ2 → 1 (λ → +0).
2 2
Therefore we obtain that √
lim γλ δλ = 2 2.
λ→+0

Note that W̃λ satisfies the equation


′′
−W̃λ (t) = e
 W̃λ (t)
, t ∈ Iλ = ( ā−s
δλ
0 b̄−s0
, δλ ),
ā−s0 b̄−s0
W̃λ ( δλ ) = W̃λ ( δλ ) = 0, (35)

W̃λ (0) = W̃λ′ (0) = 0.


Now, by (31) and (34), we see

|W̃λ′ (t)| = δλ |Wλ′ (δλ t + s0 )| ≤ δλ γλ = O(1)
as λ → +0. Also Iλ = I−s δλ
0
“ → R” as λ → +0. Therefore, Ascoli-Arzela Theorem implies
that there exists U0 : R → R such that W̃λ → U0 uniformly on compact sets on R. By
 ASYMPTOTIC BEHAVIOR OF RADIAL SOLUTIONS 13

1
using (35), this convergence
√
holds on Cloc (R) and the limit function U0 solves (10). Thus
4e √2t
U0 (t) ≡ U(t) = log (1+e 2t )2 and (32) is proven.
Next, we prove (33). By Green’s representation formula, we have
Z b̄ Z b̄−s0
δλ
Wλ (τ )
Wλ (s) = λ G(s, τ )e dτ = λ G(s, δλ t + s0 )eWλ (δλ t+s0 ) δλ dt
ā−s0
ā δλ
Z
= λδλ ekWλ kL∞ (I) G(s, δλ t + s0 )eW̃λ (t) dt
Iλ
1
Z
= G(s, δλ t + s0 )eW̃λ (t) dt.
δλ Iλ
Thus we have Z
δλ Wλ (s) = G(s, δλ t + s0 )eW̃λ (t) dt. (36)
Iλ
On the other hand, we can show the estimate
W̃λ (t) ≤ C1 U(t) + C2 (37)
for some constants C1 , C2 > 0. The proof of this fact is completely the same as Proposition
4.1 in [2]. Then by (32) and eW̃λ ≤ CeCU (t) ∈ L1 (R) by (37), we can use the Lebesgue’s
dominated convergence theorem in (36) to obtain
Z √
lim δλ Wλ (s) = G(s, s0 ) eU (t) dt = 2 2G(s, s0 ).
λ→+0 R

By (34) and δλ γλ = O(1), this convergence holds uniformly on [ā, b̄]. Thus we have proved
Proposition 5. 
As in the derivation of Theorem 1-4, we obtain the following assertion from Proposition 5
and the transformation (24).
Theorem 6. There exists λ∗ > 0 such that the problem
eu(x)

−∆u = λ 2 , x ∈ A = {a < |x| < b} ⊂ R2 ,

|x| (38)
u(x) = 0, x ∈ ∂A,


admits no radially symmetric solution for λ > λ∗ , one radially symmetric solution for λ = λ∗ ,
and just two radially symmetric solutions uλ and Uλ for 0 < λ < λ∗ . Also it holds that
Z u Z Uλ
e λ e
λ 2
dx → 0, λ 2
dx → +∞,
A |x| A |x|

and uλ → 0 uniformly on A as λ → +0.

Define δλ > 0 and Ũλ by

2 kUλ kL∞ (A)
 λδλ e  ≡ 1, log ab 


Ũλ (t) = Uλ e−(δλ t− 2 ) − kUλ kL∞ (A) , (39)

 t ∈ I˜ = (− log b−log a , log b−log a ).

λ 2δλ 2δλ
14 FUTOSHI TAKAHASHI

then δλ → 0 as λ → +0 and
√
4e 2t 1
Ũλ (t) → U(t) = log √ in Cloc (R)
(1 + e 2t )2
as λ → +0. Also √ √
δλ Uλ (|x|) → 2 2G(− log |x|, − log ab)
d2
as λ → +0 uniformly in C 0 (A), where G(s, t) is the Green’s function of the operator − ds 2

with the Dirichlet boundary condition on [ā, b̄] defined in (18) with ā = − log b, ā = − log a.
Proof. For a radially symmetric solution uλ (|x|) = uλ (r) to (38), we use a transformation (24)
to define vλ , i.e.,
uλ(r) = vλ (s), s = − log r, s ∈ (− log b, − log a).
Then a direct computation shows that vλ solves
(
−vλ′′ (s) = λevλ (s) , s ∈ (− log b, − log a),
vλ (− log b) = vλ (− log a) = 0.
By [6], there exists λ∗ > 0 such that vλ is either wλ or Wλ if 0 < λ < λ∗ . Also by the
transformation above, we see
Z − log a Z uλ (x)
vλ (s) λ e
λ e ds = dx.
− log b 2π A |x|2
Applying Proposition 5 with ā = − log b, b̄ = − log a, we obtain the result. 
Similar argument using the transformation (26) and Proposition 5 leads to the following.
The simple proof of it is omitted.
Theorem 7. There exists λ∗ > 0 such that the problem
eu(x)

−∆u = λ(N − 2)2 2(N −1) , x ∈ A = {a < |x| < b} ⊂ RN , (N ≥ 3)

|x| (40)
u(x) = 0, x ∈ ∂A,


admits no radially symmetric solution for λ > λ∗ , one radially symmetric solution for λ = λ∗ ,
and just two radially symmetric solutions uλ and Uλ for 0 < λ < λ∗ . Also it holds that
euλ eUλ
Z Z
λ 2(N −1)
dx → 0, λ 2(N −1)
dx → +∞,
A |x| A |x|

and uλ → 0 uniformly on A as λ → +0.

Define δλ > 0 and Ũλ by

2 kUλ kL∞ (A)
 λδλ e  ≡ 1, 2−N 2−N 1 


Ũλ (t) = Uλ (δλ t + a +b
2
) 2−N − kUλ kL∞ (A) , (41)
 t ∈ I˜ = (− a2−N −b2−N , a2−N −b2−N ).


λ 2δλ 2δλ

then δλ → 0 as λ → +0 and
√
4e 2t 1
Ũλ (t) → U(t) = log √ in Cloc (R)
(1 + e 2t )2
 ASYMPTOTIC BEHAVIOR OF RADIAL SOLUTIONS 15

as λ → +0. Also
√ 2−N
+ b2−N
 
2−N a
δλ Uλ (|x|) → 2 2G |x| ,
2
d 2
as λ → +0 uniformly in C 0 (A), where G(s, t) is the Green’s function of the operator − ds 2

with the Dirichlet boundary condition on [ā, b̄] defiend in (18) with ā = b2−N , b̄ = a2−N .

Data Availability. Data sharing is not applicable to this article as no datasets were generated
or analysed during the current study.

Conflicts of Interest Statement. The author has no conflicts of interest to disclose.

Acknowledgement. The author was supported by JSPS Grant-in-Aid for Scientific Re-
search (B), No. 23H01084, and was partly supported by Osaka Central University Advanced
Mathematical Institute (MEXT Joint Usage/Research Center on Mathematics and Theoreti-
cal Physics).
References
[1] B. Gidas, W. M. Ni, and L. Nirenberg: Symmetry and related properties via the maximum principle,
Comm. Math. Phys., 68 (1979), 209–243.
[2] F. Gladiali, and M. Grossi: Singular limit of radial solutions in an annulus, Asymptotic Anal., 55 (2007),
73–83.
[3] M. Grossi: Asymptotic behavior of the Kazdan-Warner solution in the annulus, J. Differential Equations,
223 (2006), 96–111.
[4] C. S. Lin: Uniqueness of least energy solutions to a semilinear elliptic equation in R2 , Manuscripta Math.,
84 (1994), 13–19.
[5] T. Shibata: Global and asymptotic behaviors of bifurcation curves of one-dimensional nonlocal elliptic
equations, J. Math. Anal. Appl., 516 (2022),126525.
[6] D. D. Joseph, and T. S. Lundgren: Quasilinear Dirichlet problems driven by positive sources, Arch.
Rational. Mech. Anal., 49 (1973), 241–269.

Department of Mathematics, Osaka Metropolitan University, 3-3-138, Sumiyoshi-ku, Sugimoto-

cho, Osaka, Japan,
Email address: [email protected]