Accepted Manuscript Uniqueness of positive radial solutions for infinite semipositone p-Laplacian problems in exterior domains
K.D. Chu, D.D. Hai, R. Shivaji
PII: DOI: Reference:
S0022-247X(18)30975-2 https://doi.org/10.1016/j.jmaa.2018.11.037 YJMAA 22723
To appear in:
Journal of Mathematical Analysis and Applications
Received date:
26 June 2018
Please cite this article in press as: K.D. Chu et al., Uniqueness of positive radial solutions for infinite semipositone p-Laplacian problems in exterior domains, J. Math. Anal. Appl. (2018), https://doi.org/10.1016/j.jmaa.2018.11.037
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Uniqueness of positive radial solutions for infinite semipositone p-Laplacian problems in exterior domains K. D. Chu Faculty of Mathematics and Statistics Ton Duc Thang University Ho Chi Minh City, Vietnam Email:
[email protected] D. D. Hai Department of Mathematics and Statistics Mississippi State University Mississippi State, MS 39762, USA Email:
[email protected] R. Shivaji Department of Mathematics and Statisitics University of North Carolina at Greensboro Greensboro, NC 27402, USA Email:
[email protected] November 13, 2018 Abstract We prove uniqueness and asymptotic behavior of positive radial solutions to the p-Laplacian problem −Δp u = λK(|x|)f (u) in |x| > r0 , u = 0 on |x| = r0 , u(x) → 0 as |x| → ∞.
1
where Ω = {x ∈ Rn : |x| > r0 > 0}, n > p, f : (0, ∞) → R is continuous, f (u) ∼ uq at ∞ for some q ∈ [0, p − 1) with possible infinite semipositone structure at 0, and λ is a large parameter.
1. INTRODUCTION Consider the boundary value problem ⎧ ⎨ −Δp u = λK(|x|)f (u) in Ω, u = 0 on |x| = r0 , ⎩ u(x) → 0 as |x| → ∞,
(1.1)
where Δp u = div(|∇u|p−2 ∇u), 1 < p < n, Ω = {x ∈ Rn : |x| > r0 > 0}, K : [r0 , ∞) → (0, ∞), f : (0, ∞) → R are continuous, and λ is a positive parameter. n−p By letting r = |x| and t = (r/r0 ) 1−p , we see that radial solutions of (1.1) satisfy the two-point boundary value problem −(φ(u )) = λh(t)f (u), t ∈ (0, 1), (1.2) u(0) = u(1) = 0, 1−p p p p(1−n) p−1 p−2 n−p where φ(s) = |s| s and h(t) = n−p r0 t K r0 t n−p . We are motivated by the uniqueness results in [1,27] in the finite semi(u) positone, sublinear case i.e. f (0) < 0 and limu→∞ ufp−1 = 0. In [1], Castro et al. obtained uniqueness of non-negative solutions to (1.2) for λ large when p = 2, f > 0 on (0, ∞), f is concave with limu→∞ f (u) = ∞, and h is positive with h < 0 and tα h(t) bounded near 0 for some α ∈ (0, 1). Uniqueness for positive solutions to (1.2) in the general case p > 1 for λ large was established in [23] when f > 0 on (0, ∞), limu→∞ f (u) = ∞, and fu(u) is nonincreasing q for u large for some q ∈ (0, p − 1). Note that the proofs in [1,27] depend heavily on the fact that f (0) is finite and can not be applied to the infinite semipositone case i.e. limu→0+ f (u) = −∞. (see figure 1.1). In this paper, we will prove uniqueness and asymptotic behavior of positive solutions to (1.2) when λ is large. In fact, this is the first paper in the literature dealing with this challenging uniqueness analysis for infinite semipositone problems. (We 2010 Mathematics Subject Classification: Primary 34B16; Secondary 34B15. Key words and phrases: infinite semipositone, positive solutions, uniqueness
2
Figure 1.1 Figure 1: * note that when h(t) ≡ constant, the problem can be analyzed by time-map analysis, but here we do not assume h is a constant). We focus on reaction terms f that satisfy f (u) ∼ uq at ∞ for some q ∈ [0, p − 1), which allows f to be bounded and we do not require any monotonicity assumption on f as in [1,27]. In particular, our results when applied to the model case 1 λ 1 q 1+u −(φ(u )) = tη − uα + u e , t ∈ (0, 1), (1.3) u(0) = u(1) = 0, where 0 ≤ η, α, η + α < 1, 0 ≤ q < p − 1, give the uniqueness of positive solutions to (1.3) when λ is large. Furthermore, the unique solution uλ of uλ (1.3) satisfies lim = 1 uniformly for t ∈ (0, 1), where wq denotes 1 λ→∞ λ p−1−q wq
the unique positive solution of −(φ(w )) = t−η wq , t ∈ (0, 1), w(0) = w(1) = 0. It should be noted that our proof can be applied to include cases when f (u) do not behave like powers of u (see Remark 1.1 (iii)). We refer to [2,7] for uniqueness results to the sublinear semilinear problem −Δu = λf (u) in Ω, (1.4) u = 0 on ∂Ω, 3
when λ is large in the semipositone case, and to [3,6,17,18,24,26,30] to the positone case. Uniqueness results to related quasilinear Dirichlet problems were discussed in [9-11,13] for the positone case on a bounded domain, and in [15] for the semipositone and the infinite semipositone case on a ball. Uniqueness of positive solutions for quasilinear semipositone problems on a bounded domain is still an open question. For existence results for λ >> 1 in the semipositone and the infinite semipositone case, see [15,16,22,23]. Related results to (1.4) when f is singular at 0 can be found in [4,5,8,12,19,20,21,28,29]. We shall make the following assumptions: (A1) h : (0, 1] → (0, ∞) is differentiable, nonincreasing, and there exists a constant η ∈ [0, 1) such that lim sup tη h(t) < ∞. t→0+
(A2) f : (0, ∞) → R is differentiable and there exists a constant β > 0 such that (z − β)f (z) > 0 for z = β. (A3) There exist numbers L > 0 and 0 ≤ q < p − 1 such that lim
z→∞
and lim sup z→∞
f (z) =L zq
z|f (z)| < p − 1. f (z)
(A4) There exist constants l < 0 and 0 ≤ α < 1 − η such that lim z α f (z) = l and lim sup z α+1 |f (z)| < ∞.
z→0+
z→0+
By a solution of (1.2), we mean a function u ∈ C 1 [0, 1] such that φ(u ) is differentiable on (0, 1] and satisfies (1.2). Our main result is Theorem 1.1. Let (A1)-(A4) hold. Then there exists a constant λ0 > 0 such that problem (1.2) has a unique positive solution uλ for λ > λ0 . Furthermore, uλ (t) lim =1 1 λ→∞ (λL) p−1−q w (t) q 4
uniformly for t ∈ (0, 1), where wq is the unique solution of −(φ(w )) = h(t)wq in (0, 1), w(0) = w(1) = 0. Remark 1.1. (i) Note that (A4) is satisfied if limz→0+ z α+1 f (z) = l1 ∈ (0, ∞). (ii) A key ingredient of the proof of Theorem 1.1 is to obtain the estimate 1 u(t) ≥ C0 λ p−1−q min(t, 1 − t) for λ large in Lemma 2.5 below. This estimate requires the assumptions (A2),(A4), and the fact that lim inf f (z) > 0 (could z→∞
be finite). It will then be used to obtain sharp upper and lower bound estimates for positive solutions of (1.2) when f (u) ∼ uq at ∞ for some q ∈ [0, p − 1). Next, we establish the existence of a positive constant c independent of u such that cv ≤ u ≤ c−1 v for any two positive solutions of (1.2), and then show that c = 1. (iii) The steps in (ii) can be used with suitable modifications to obtain the uniqueness result in Theorem 1.1 under the assumptions (A2),(A4), and (A5) There exist constants a > 0 and q ∈ [0, p − 1) such that f (z) is nondecreasing and fz(z) is nonincreasing on [a, ∞). q This allows the case when f (u) do not behave like powers of u at ∞ e.g. f (u) = − u1α + uq ln(1 + u). 2. PRELIMINARY RESULTS z We shall denote the norm in Lp (0, 1) by ||.||p . Let F (z) = 0 f (s)ds. By (A2) and (A3), there exists a number θ > β such that F (z) < 0 for 0 < z < θ and F (z) > 0 for z > θ. We first extend [27, Lemma 2.1] to include the infinite semipositone case with h not necessarily strictly decreasing. Lemma 2.1. Let u be a positive solution of (1.2). Then u has only one interior maximum at t0 and u(t0 ) ≥ θ. Proof. Using the equation in (1.2) and (A2), we see that φ(u ) is increasing for t near 0. Hence φ(u (t)) > φ(u (0)) ≥ 0 i.e. u (t) > 0 for t near 0. Let t0 ∈ (0, 1) be the first zero of u . Note that by (A4), lim+ uα−1 (t)F (u(t)) = t→0
l(1−α)−1 , and since |u(t)| ≤ ||u ||∞ t, it follows from (A1) that tη+α−1 h(t)F (u) is bounded near t = 0. Since η + α < 1, lim+ h(t)F (u(t)) = 0. Suppose t→0
u(t0 ) < θ. Then h F (u) ≥ 0 on [0, t0 ]. Hence, by multiplying the equation in
5
(1.2) by u and integrating on [0, t0 ], we obtain
t0 1 p |u (0)| = λh(t0 )F (u(t0 )) − λ h (t)F (u)dt 0≤ 1− p 0 ≤ λh(t0 )F (u(t0 )) < 0, a contradiction. Thus u(t0 ) ≥ θ. Next, we claim that u is decreasing on (t0 , 1). Suppose to the contrary that there exist t1 , t2 ∈ (t0 , 1) with t1 < t2 such that u(t1 ) ≤ u(t2 ). Since (φ(u )) (t0 ) = −λh(t0 )f (u(t0 )) < 0, it follows that u is decreasing for t near t0 , t > t0 . Hence u has a minimum value on [t0 , t2 ] at some t∗ ∈ (t0 , t2 ). This implies λh(t∗ )f (u(t∗ )) = −(φ(u )) (t∗ ) ≤ 0 i.e. 0 < u(t∗ ) ≤ β and so F (u(t∗ )) < 0. Since F (u(t0 )) ≥ 0, there exists tˆ ∈ [t0 , t∗ ) such that F (u) < 0 on (tˆ, t∗ ] and F (u(tˆ)) = 0. Multiplying the equation in (1.2) by u and integrating on [tˆ, t∗ ] gives
t∗ 1 ˆ p ∗ ∗ |u (t)| = λh(t )F (u(t )) − λ h (t)F (u)dt < 0, 0≤ 1− p tˆ a contradiction. Hence u is increasing on (0, t0 ) and decreasing on (t0 , 1), which completes the proof. Let ρ = β+θ . By Lemma 2.1, for each μ ∈ (0, θ] there exist tμ ≤ t0 ≤ t˜μ 2 such that u(tμ ) = u(t˜μ ) = μ. Lemma 2.2. There exists a constant cρ > 0 such that any positive solution u of (1.2) satisfies ⎧ p t 1/p p−1+α ⎪ ⎪ h if t ∈ [0, tρ ], ⎪ 0 ⎨ u(t) ≥ 1 ⎪ p ⎪ cρ λ p−1+α 1 1/p p−1+α ⎪ ⎩ h if t ∈ [t˜ρ , 1]. t In particular, there exists a constant c˜ρ > 0 independent of u and λ such 1 that max{tρ , 1 − t˜ρ } ≤ c˜ρ λ− p . Proof. Note that h F (u) ≥ 0 on [0, tθ ] ∪ [t˜θ , 1]. Let s ∈ [0, tθ ]. Multiplying the equation in (1.2) by u and integrating on [0, s] gives
s 1 1 p p (u (s)) = 1 − (u (0)) − λh(s)F (u) + λ h (τ )F (u)dτ 1− p p 0 6
≥ −λh(s)F (u). Hence
u (s) 1/p
≥ (λh(s))1/p ,
(−F (u)) which implies upon integrating on [0, t], t ∈ [0, tθ ], that t u(t) dz 1/p ≥λ h1/p . 1/p (−F (z)) 0 0
(2.1)
By (A4), there exists a constant k > 0 such that −F (z) ≥ kz 1−α for z ∈ [0, ρ]. Hence (2.1) implies t p−1+α p−1 1/p p u (λk) (t) ≥ h1/p , p 0 p t 1 p−1+α for t ∈ [0, tρ ] follows, where cρ = from which u(t) ≥ cρ λ p−1+α 0 h1/p p p−1+α p−1 1/p k . In particular, since h is nonincreasing, p 1
p
ρ = u(tρ ) ≥ cρ (λh(1)) p−1+α tρp−1+α , p−1+α 1 p which implies tρ ≤ c˜ρ λ−1/p , where c˜ρ = ρ/(cρ h p−1+α (1)) . Next, let s ∈ [t˜θ , 1]. Since F (θ) = 0, it follows from multiplying the equation in (1.2) by u and integrating on [t˜θ , s] that
1 1− |u (s)|p ≥ −λh(s)F (u). p Hence
|u (s)| (−F (u))
1/p
≥ (λh(s))1/p ,
and since u ≤ 0 on [t˜θ , 1], it follows upon integrating on [t, 1], t ∈ [t˜θ , 1], that 1 u(t) dz 1/p ≥λ h1/p 1/p (−F (z)) 0 t p p−1+α 1 1 and as in the above argument, we get u(t) ≥ cρ λ p−1+α t h1/p for t ∈ [t˜ρ , 1] follows. By choosing t = t˜ρ , we conclude as in the above that 1 − t˜ρ ≤ c˜ρ λ−1/p , which completes the proof. 7
The next result establishes a comparison principle when the forcing terms are close in L1 −norm, but not necessarily ordered. Lemma 2.3. Let M > 0 and hi ∈ L1 (0, 1) with ||hi ||1 < M for i = 1, 2. Let u˜i ∈ C 1 [0, 1], i = 1, 2, satisfy −(φ(˜ u1 )) ≥ h1 , t ∈ (0, 1), −(φ(˜ u2 )) ≤ h2 , t ∈ (0, 1), , u˜1 (0) = u˜1 (1) = 0, u˜2 (0) = u˜2 (1) = 0. Then (i) If h2 ≥ 0, h2 ≡ 0 then for a given ε0 > 0, there exists a constant δ > 0 such that u˜1 ≥ (1 − ε0 )˜ u2 on [0, 1], provided that ||h1 − h2 ||1 < δ. (ii) Let tˇ < 1/2 < tˆ . Suppose u˜2 defined above satisfies −(φ(˜ u2 )) = h2 on (0, 1) and there exists a constant C > 0 such that |˜ u2 | > C on [0, tˇ] ∪ ˜ [tˆ, 1]. Then there exist constants δC , kC > 0 such that if ||h1 − h2 ||1 < δ˜C then u˜1 (t) ≥ u˜2 (t) − kC ||h1 − h2 ||1 p(t) for t ∈ [0, tˇ] ∪ [tˆ, 1], where p(t) = min(t, 1 − t). Proof. Let ui ∈ C 1 [0, 1], i = 1, 2, satisfy −(φ(ui )) = hi , t ∈ (0, 1), ui (0) = ui (1) = 0. Then u˜1 ≥ u1 and u˜2 ≤ u2 on [0, 1] by the weak comparison principle (see [25, Lemma A.2]). Let ε > 0. We first show that there exists a constant δ > 0 such that ||h1 − h2 ||1 < δ =⇒ ||u1 − u2 ||∞ < ε. (2.2) By integrating, we obtain ui (t)
−1
=φ
Ci −
t 0
hi ,
1 s where Ci are constants satisfying 0 φ−1 Ci − 0 hi ds = 0, i = 1, 2. Using the fact that φ−1 is increasing with φ−1 (0) = 0, we deduce that |Ci | ≤ ||hi ||1 and |C1 − C2 | ≤ ||h1 − h2 ||1 . 8
(2.3)
Since φ−1 is uniformly continuous on [−2M, 2M ], there exists δ0 > 0 such that x, y ∈ [−2M, 2M ], |x − y| < δ0 =⇒ |φ−1 (x) − φ−1 (y)| < ε. Suppose ||h1 − h2 ||1 < δ0 /2. Then in view of (2.3), t < δ0 , C1 − C2 − (h − h ) 1 2 0
t and since Ci − 0 hi ≤ 2M, it follows that |u1 (t) − u2 (t)| < ε
(2.4)
for all t ∈ [0, 1] i.e. (2.2) holds with δ = δ0 /2. (i) If h2 ≥ 0, h2 ≡ 0 then u2 (t) ≥ ||u2 ||∞ p(t) for t ∈ [0, 1] (see e.g. [14, Lemma 3.4]. Let ε < ε0 ||u2 ||∞ . Then there exists δ > 0 such that (2.4) holds provided that ||h1 − h2 ||1 < δ. Since ui (0) = ui (1) = 0, the Mean Value Theorem gives u1 (t) ≥ u2 (t) − εp(t) ≥ (1 − ε0 )u2 (t) for t ∈ [0, 1], and therefore u˜1 ≥ (1 − ε0 )˜ u2 on [0, 1]. (ii) Suppose u˜2 satisfies −(φ(˜ u2 )) = h2 on (0, 1) and |˜ u2 | > C > 0 on I ∪ J, where I = [0, tˇ], J = [tˆ, 1]. Then u˜2 ≡ u2 and, in view of (2.4), there exists 0 < δ˜ < 12 min(φ(C/2), δ0 ) such that |u1 | > C/2 on [0, tˇ] ∪ [tˆ, 1] ˜ which we shall assume. Hence provided that ||h1 − h2 ||1 < δ, t Ci − hi = φ(|ui (t)|) > φ(C/2) > 0 (2.5) 0
for t ∈ I ∪ J and Ci − Theorem, |u1 (t)
−
t
u2 (t)|
0
hi are of the same sign, i = 1, 2. By the Mean Value
t t −1 −1 C1 − C2 − = φ h1 − φ h2 0
2−p
=
|ξ(t)| p−1
t C1 − C2 − 0 (h1 − h2 ) p−1 9
0
,
(2.6)
t t where ξ(t) is between C1 − 0 h1 and C2 − 0 h2 . By (2.3) and (2.5), φ(C/2) ≤ |ξ(t)| ≤ 2M, from which (2.3) and (2.6) gives |u1 (t) − u2 (t)| ≤ kC ||h1 − h2 ||1 , 2−p
2−p
2 for t ∈ I, where kC = p−1 max{(2M ) p−1 , φ p−1 (C/2)}. Since u1 (0) = 0 and u2 (0) = 0, it follows that
u˜1 (t) ≥ u1 (t) ≥ u2 (t) − kC ||h1 − h2 ||1 t for t ∈ I. Similarly, u˜1 (t) ≥ u1 (t) ≥ u2 (t) − kC ||h1 − h2 ||1 (1 − t) for t ∈ J, which completes the proof.
Lemma 2.4. (i)
1
(ii)
1
0
h(t) αp dt < ∞. t 1/p p−1+α h 0
1/2
h(t) αp dt < ∞. 1 1/p p−1+α h t
˜ = t−γ h. Then η + γ < p and Proof. (i) Let γ = p − 1 + α and define h 1/p 1 ˜ ∈ L (0, 1) in view of (A1). Note that therefore h t t 1/p γ/p ˜ 1/p . h ≤t h 0
Since h is nonincreasing, h(t)
p−1 p
0
p−1
t −1 1/p ≤ t h . 0
Hence 0
1
h(t) αp dt ≤ t 1/p p−1+α h 0
1
h1/p (t)
t 0
h1/p tp−1
0
10
αp p−1− p−1+α
dt
≤
1
= 0
˜ 1/p h
1
γ˜
t h
0
t 0
1/p
t
˜ 1/p h
t
γ/p 0
˜ 1/p h
αp p−1− p−1+α
αp p−1− p−1+α
dt = Cα,p
1 0
t1−p dt
˜ 1/p h
p(p−1) p−1+α
<∞
by the choice of γ, where Cα,p is a positive constant depending on α, p. (ii) Since h is nonincreasing,
1 1/2
p−1 h(t) dt ≤ (h(1/2)) p αp p−1+α 1 1/p h t
= C˜α,p
1 1/2
h
1/p
1 1/2
αp 1− p−1+α
h1/p (t) αp dt 1 1/p p−1+α h t
< ∞,
where C˜α,p is a positive constant depending on α, p. ˜ > 0 such that for λ > λ, ˜ any Lemma 2.5. There exist constants C0 , λ positive solution u of (1.2) satisfies 1
u(t) ≥ λ p−1 C0 p(t) for t ∈ [0, 1], where p(t) = min(t, 1 − t). Proof. By (A2), (A4), and continuity of z α f (z), there exists a constant a > 0 such that a f (z) ≥ − α (2.7) z for all z > 0. Since u ≥ ρ on [tρ , t˜ρ ], where tρ , t˜ρ are defined immediately preceding Lemma 2.2, it follows that f (u) ≥ δρ > 0 on [tρ , t˜ρ ] for some 1 δρ > 0. Then assuming cρ λ p−1+α > 1, (2.7) and Lemma 2.2 give ⎧ ah(t) ⎪ ⎪ − if t ∈ [0, tρ ] αp ⎪ ⎪ t 1 p−1+α ⎪ ⎪ hp 0 ⎨ ⎪ 1 − p−1 if t ∈ (tρ , t˜ρ ) ≡ h1 , δρ h(t) − φ λ u ≥ (2.8) ⎪ ⎪ ah(t) ⎪ ⎪ − if t ∈ [t˜ρ , 1] ⎪ αp ⎪ 1 1 p−1+α ⎪ ⎩ p h t
11
Defining the right-hand side of (2.8) as h1 . Then h1 ∈ L1 (0, 1) by Lemma 2.4. Let ω ∈ C 1 [0, 1] satisfy −(φ(ω )) = δρ h(t) ≡ h2 , t ∈ (0, 1), (2.9) ω(0) = ω(1) = 0. Then ω(t) ≥ ||ω||∞ p(t) for t ∈ [0, 1] (see [14, Lemma 3.4]). Note that 1 tρ tρ h(t) h+ h +a dt ||h1 − h2 ||1 = δρ αp 1 p−1+α t 0 0 t˜ρ p h 0 +a
1 t˜ρ
h(t) αp dt → 0 as λ → ∞ 1 1 p−1+α p h t
since max{tρ , 1 − t˜ρ } → 0 as λ → ∞ by Lemma 2.2. Hence it follows by 1 applying Lemma 2.3 (i) with u˜1 = λ− p−1 u, u˜2 = w, and ε0 = 1/2 that for λ large enough, 1 λ− p−1 u ≥ ω/2 ≥ (||ω||∞ /2)p(t), which completes the proof. Lemma 2.6. Let 0 < q < p − 1. Then there exists a unique positive solution wq ∈ C 1 [0, 1] to the problem −(φ(w )) = h(t)wq , t ∈ (0, 1), (2.10) w(0) = w(1) = 0. Proof. Let ω0 , ω1 ∈ C 1 [0, 1] satisfy −(φ(ω2 )) = h(t), t ∈ (0, 1), −(φ(ω1 )) = ω1q , t ∈ (0, 1), , . ω1 (0) = ω1 (1) = 0, ω2 (0) = ω2 (1) = 0. whose existence follow from [14,Theorem 2.2 (ii)] and [14, Lemma 3.1]. Then it is easy to see that εω1 is a subsolution of (2.10) if ε is small enough, and M ω2 is a supersolution of (2.10) if M is large enough. By decreasing ε and increasing M if necessary, we can assume that εω1 ≤ M ω2 on [0, 1]. Hence (2.10) has a solution wq with wq ≥ εω1 in [0, 1]. To show uniqueness, let w1 , w2 i (t) i (t) be two positive solutions of (2.10). Since sup wp(t) < ∞ and inf wp(t) > t∈(0,1)
12
t∈(0,1)
0 for i = 1, 2, there exists a largest constant c > 0 such that w1 ≥ cw2 in [0, 1]. Suppose c < 1. Then −(φ(w1 )) = h(t)w1q ≥ cq h(t)w2q q
for t ∈ (0, 1), and the weak comparison principle gives w1 ≥ c p−1 w2 in [0, 1], a contradiction with the maximality of c. Thus c ≥ 1 i.e. w1 ≥ w2 , which completes the proof. The asymptotic behavior of the solutions to (1.2) as λ → ∞ is given in the next result. Lemma 2.7. Let ε ∈ (0, 1). Let wq be the unique solution of (2.10) given ¯ > 0 such that for λ > λ, ¯ any by Lemma 2.6. Then there exists a constant λ positive solution u of (1.2) satisfies p
p
1
1
(1 − ε) p−1−q (λL) p−1−q wq ≤ u ≤ (1 + ε) p−1−q (λL) p−1−q wq on [0, 1], where L is defined in (A3). In other words, lim
λ→∞
u(t) 1
(λL) p−1−q wq (t)
=1
uniformly in t ∈ (0, 1). Proof. Let ε ∈ (0, 1). By (A3), there exists a constant A > 0 such that (1 − ε)Lz q < f (z) < (1 + ε)Lz q ˜ where λ ˜ is defined in Lemma 2.5. Since wq (t) ≤ for z > A. Suppose λ > λ, ||wq ||∞ p(t) for t ∈ [0, 1], it follows from Lemma 2.5 that u ≥ d0 wq in [0, 1], 1 ˜ p−1 where d0 = λ C0 ||wq ||−1 ∞ . Let dλ be the largest number such that u ≥ dλ wq in [0, 1]. By (A4), there exists a constant b > 0 such that |f (z)| ≤
b for z ∈ (0, A]. zα
1
Assuming λ p−1 C0 > 1 then Lemma 2.5 gives ⎧ (1 − ε)Ldλq h(t)wqq ⎪ ⎪ ⎨ −(φ(u )) ≥ λ bh(t) ⎪ ⎪ ⎩ − α p (t) 13
if u(t) > A, if u(t) ≤ A.
This implies − φ
u
≥
1
(λ(1 − ε)Ldqλ ) p−1
⎧ h(t)wqq ⎪ ⎪ ⎨
if u(t) > A,
b0 h(t) ⎪ ⎪ ⎩ − α p (t)
if u(t) ≤ A,
≡ h1
b where b0 = (1−ε)Ld q and h1 is defined to be the right-hand side of the previous 0 equation. Since −(φ(wq )) = h(t)wqq ≡ h2 in (0, 1), 1 − p−1 −1 C0 and the measure of {t : u(t) ≤ A} ≤ the measure of t : p(t) ≤ λ → 0 as λ → ∞ by Lemma 2.5, it follows that
b0 q ||h1 − h2 ||1 = dt → 0 h(t) wq + α p (t) u≤A
as λ → ∞. Hence if λ is large enough, Lemma 2.3(i) gives u 1
(λ(1 − ε)Ldλq ) p−1 p
≥ (1 − ε)wq
on [0, 1] p
1
1
i.e. u ≥ (1−ε) p−1 (λLdqλ ) p−1 wq in [0, 1]. Consequently, dλ ≥ (1−ε) p−1 (λLdqλ ) p−1 , and therefore p 1 u ≥ (1 − ε) p−1−q (λL) p−1−q wq on [0, 1]. Next, since sup t∈(0,1)
u(t) p(t)
wq (t) t∈(0,1) p(t)
< ∞ and inf
> 0, there exists a constant d1 ≥ 0
(depending on u) such that u ≤ d1 wq in [0, 1]. Let d˜λ be the smallest number such that u ≤ d˜λ wq in [0, 1]. Then d˜λ ≥ d0 and ⎧ (1 + ε)Ld˜qλ h(t)wqq if u(t) > A, ⎪ ⎪ ⎨ −(φ(u )) ≤ λ bh(t) ⎪ ⎪ ⎩ α if u(t) ≤ A. p (t) Hence − φ
u
1
(λ(1 + ε)Ld˜λq ) p−1
≤
⎧ h(t)wqq ⎪ ⎪ ⎨
if u(t) > A,
b1 h(t) ⎪ ⎪ ⎩ α p (t)
if u(t) ≤ A,
14
˜ 1, ≡h
where b1 = Since
˜ 1 be the righ-hand side of the previous equation. Let h
b 1 q ˜ 1 − h2 ||1 ≤ ||h →0 h(t) wq + α p (t) u≤A
b . (1+ε)Ldq0
p
1 gives u ≤ (1 + ε) p−1 (λLd˜λq ) p−1 wq in p 1 [0, 1] if λ is large enough. Consequently, d˜λ ≤ (1 + ε) p−1 (λLd˜λq ) p−1 , which implies p 1 u ≤ (1 + ε) p−1−q (λL) p−1−q wq in [0, 1],
as λ → ∞, Lemma 2.3(i) with ε0 =
ε ε+1
which completes the proof. Lemma 2.8. There exist positive constants K0 , tˇ , tˆ with tˇ < 1/2 < tˆ ¯ any positive solution u of (1.2) satisfies such that for λ > λ, 1
|u (t)| ≥ K0 λ p−1−q for t ∈ [0, tˇ] ∪ [tˆ, 1]. Proof. By Lemma 2.7, there exist positive constants D1 , D2 independent of u and λ such that 1
1
D1 λ p−1−q wq (t) ≤ u(t) ≤ D2 λ p−1−q wq (t)
(2.11)
¯ Consequently, there exists a constant for t ∈ [0, 1], provided that λ > λ. 1 D3 > 0 depending on D1 , D2 and wq such that u (0) ≥ D3 λ p−1−q and u (1) ≤ 1 −D3 λ p−1−q . By (A2) and (A3), there exists a constant m > 0 such that f (z) ≤ mz q for all z > 0. This, together with (2.11), gives t φ(u (t)) = φ(u (0)) − λ h(s)f (u)ds 0
p−1
≥ D3p−1 λ p−1−q − λm ≥λ
p−1 p−1−q
D3p−1
−
mD2q
t 0
t 0
h(s)uq ds
h(s)wqq ds
15
p−1
λ p−1−q D3p−1 ≥ 2
tˇ for t < tˇ, where tˇ < 1/2 is such that 0 h(s)wqq ds < D3p−1 /(2mD2q ). Hence 1 1 u (t) ≥ K0 λ p−1−q for t ∈ [0, tˇ], where K0 = D3 /2 p−1 . Similarly,
1
φ(u (t)) = φ(u (1)) + λ
h(s)f (u)ds t
≤ ≤λ
p−1 p−1−q
p−1 −D3p−1 λ p−1−q
+
mD2q
for t > tˆ, where tˆ > 1/2 is such that 1
h(s)uq ds ≤
t
−D3p−1
1
+ λm 1
q
h(s)w ds t
1 tˆ
p−1
λ p−1−q D3p−1 ≤− 2
h(s)wqq ds < D3p−1 /(2mD2q ).
Hence u (t) ≤ −K0 λ p−1−q for t ∈ [tˆ, 1], which completes the proof. 3. PROOF OF THE MAIN RESULT In what follows, we shall assume that λ > 1 is large enough so that Lemmas 2.5 and 2.7 apply. Proof of Theorem 1.1. By (A3), there exist positive constants A0 , δ, ε0 , δ1 with δ(1 + ε0 ) < δ1 < p − 1 and A0 > β such that |f (z)| ≤
and (1 − ε0 )
z1 z2
q
δf (z) z
f (z1 ) ≤ (1 + ε0 ) ≤ f (z2 )
(3.1)
z1 z2
q (3.2)
for z, z1 , z2 > A0 . Let u, v be positive solutions of (1.2). Then Lemma 2.7 with ε small gives 1 1 1 3 (λL) p−1−q wq ≤ u, v ≤ (λL) p−1−q wq 2 2
(3.3)
in [0, 1], which implies (1/3)v ≤ u ≤ 3v on [0, 1]. Let c be the largest number such that cv ≤ u ≤ c−1 v on [0, 1] and suppose that c < 1. We claim that there exists a constant L0 > 0 such that f (u) − cp−1 f (v) ≥ L0 (1 − c) 16
(3.4)
for v > A, where A = 3A0 . For v > A, we have u > A0 . Hence (3.1) and the Mean Value Theorem gives |f (u) − f (v)| = |u − v||f (ψ)| ≤
δ(1 − c)vf (ψ) cψ
δ(1 − c)f (ψ) , c2 where ψ is between u and v. Here we have used the fact that cv ≤ ψ ≤ c−1 v. By (3.2),
q ψ f (ψ) ≤ (1 + ε0 ) ≤ (1 + ε0 )c−q , f (v) v ≤
and so |f (u) − f (v)| ≤ for v > A. Since
lim
z→1−
δ1 (1 − c) δ(1 + ε0 )(1 − c) f (v) ≤ f (v) q+2 c cq+2
1 − z p−1 δ1 − 2+q 1−z z
(3.5)
= p − 1 − δ1 ≡ 2k0 > 0,
there exists c˜ ∈ (0, 1) such that 1 − z p−1 δ1 − 2+q ≥ k0 1−z z
(3.6)
for z ≥ c˜. Combining (3.5) and (3.6), we obtain f (u) − cp−1 f (v) = f (u) − f (v) + (1 − cp−1 )f (v)
δ1 1 − cp−1 − q+2 f (v) ≥ k0 (1 − c)f (v) ≥ (1 − c) 1−c c
(3.7)
for v > A if c ≥ c˜. Suppose next that c < c˜. Since c ≥ 1/3 and q < p − 1, there exists a constant m0 > 0 such that cq − cp−1 ≥ m0 (1 − c) > m0 (1 − c˜), from which (3.2) gives
f (u) − c
p−1
f (v) =
f (u) p−1 −c f (v) ≥ (1 − ε0 )cq − cp−1 f (v) f (v) 17
≥ (c − c q
p−1
− ε0 )f (v) ≥
m0 (1 − c˜) 2
f (v) ≥ B(1 − c)
(3.8)
if ε0 is decreased further, where B = (m0 (1 − c˜)/2) inf f. From (3.7), (3.8) (A,∞)
and inf f > 0, we see that (3.4) holds. (A,∞)
For v ≤ A, we have u ≤ 3A, and it follows from (A4) that there exists a constants KA > 0 such that |f (u) − f (v)| = |u − v||f (ξ)| ≤ ≤
KA (1 − c)v c|ξ|α+1
KA (1 − c) KA (1 − c) ≤ , α+2 α c v wqα
and
KA KA ≤ α α w wq
|f (w)| ≤
(3.9)
for w ∈ {u, v}, where we have assumed that λ is large enough (independent of u, v) so that (3.3) gives w ≥ wq and c2+α wα ≥ wqα on (0, 1). Consequently, |f (u) − cp−1 f (v)| ≤ |f (u) − f (v)| + (1 − cp−1 )|f (v)| ≤
˜ A (1 − c) K , wqα
(3.10)
˜ A = 2KA max{1, p − 1}. By (3.3) and (A3), there exists a constant where K L1 > 0 depending on L such that q
f (v) ≤ mv q ≤ L1 λ p−1−q
(3.11)
on [0, 1] for some m > 0. Hence (3.4) gives
L0 (1 − c) p−1 −(φ(u )) ≥ λh(t) c f (v) + q L1 λ p−1−q for v > A, which implies ⎧ q − p−1−q ⎪ λ h(t) cp−1 + ⎨ − (φ (˜ u)) ≥ ⎪ q ⎩ − p−1−q λ h(t)f (u)
L2 (1−c) q λ p−1−q
18
) f (v) if v > A, if v < A.
≡ h1 , (3.12)
1
where u˜ = λ− p−1−q u and L2 = L0 /L1 . Define h1 to be the right-hand side of (3.12). 1 p−1 q 1 Let v˜ = λ− p−1−q cp−1 + L2 (1 − c)λ− p−1−q v. Then
− (φ(˜ v )) = λ
q − p−1−q
p−1
h(t) c
+
L2 (1 − c)
f (v) ≡ h2 .
q
λ p−1−q
(3.13)
in [0, 1]. Note that, in view of (3.11), there exists a constant M0 independent of u, v, λ, such that |hi (t)| ≤ M0 h(t) for t ∈ (0, 1), i = 1, 2. By (3.9) and (3.10), q (1 − c) L 2 − p−1−q ||h1 − h2 ||1 = λ h(t) f (u) − cp−1 f (v) − f (v) dt q λ p−1−q v
≤λ
(1 − c)L3 v
h dt, wqα
(3.14)
˜ A + L2 KA . Let tˇ , tˆ be given by Lemma 2.8. Then |˜ where L3 = K v (t)| ≥ cK0 ≥ K0 /3 ≡ C for t ∈ 0, tˇ ∪ [tˆ, 1]. Since ||h1 − h2 ||1 → 0 as λ → ∞, it follows from (3,12), (3.13) and Lemma 2.3(ii) that for λ sufficiently large, u˜(t) ≥ v˜(t) − kC ||h1 − h2 ||1 p(t) for t ∈ [0, tˇ] ∪ [tˆ, 1]. This, together with (2.11), gives 1 p−1 q 1 u(t) ≥ cp−1 + L2 (1 − c)λ− p−1−q v − λ p−1−q kC ||h1 − h2 ||1 p(t)
≥
p−1
c
q − p−1−q
+ L2 (1 − c)λ
1 p−1
˜ − kC ||h1 − h2 ||1 v,
for t ∈ [0, tˇ] ∪ [tˆ, 1], where k˜C = (kC /D1 ) sup(p/wq ). Since (0,1)
q
(1/3)p−1 ≤ cp−1 , cp−1 + L2 (1 − c)λ− p−1−q ≤ 1 + L2 ,
19
(3.15)
it follows from the Mean Value Theorem that 1 p−1 1 q q − p−1−q p−1 + L2 (1 − c)λ − cp−1 p−1 ≥ L4 (1 − c)λ− p−1−q , c
(3.16)
where L4 is a positive constant depending only on p and L2 . Since v
h L4 (1 − c)v dt v ≥ q α wq 2λ p−1−q
L4 − k˜C L3 q λ p−1−q v 0 such that ≥
u ≥ (c + γ)v
on [0, tˇ] ∪ [tˆ, 1]. (3.17) In particular, u(τ ) ≥ cv(τ )+γ0 for τ ∈ tˇ , tˆ , where γ0 = γ min(v(tˇ), v(tˆ)) > 0. By (2.11), 1 v(t) ≥ D1 λ p−1−q min wq > A [tˇ,tˆ]
for t ∈ [tˇ, tˆ], provided that λ is large enough. Hence (3.4) gives −(φ(u )) = λh(t)f (u) ≥ λcp−1 h(t)f (v) on (tˇ, tˆ). Since
−(φ(cv + γ0 ) ) = λcp−1 h(t)f (v) on (tˇ, tˆ), and u(τ ) ≥ cv(τ ) + γ0 for τ ∈ tˇ , tˆ , the weak comparison principle implies u ≥ cv + γ0
on [tˇ, tˆ].
This, together with (3.17), gives the existence of a number γ1 > 0 such that u ≥ (c + γ1 )v in [0, 1]. By symmetry, there exists a constant γ2 > 0 such that v ≤ (c + γ2 )u in [0, 1]. Hence (c + γ˜ )v ≤ u ≤ (c + γ˜ )−1 v in [0, 1], a contradiction with the definition of c. Thus c ≥ 1 i.e. u = v in [0, 1]. The asymptotic behavior of the solution as λ → ∞ is given in Lemma 2.7. This completes the proof of Theorem 1.1. Ackowledgements. The authors thank the referee for carefully reading the manuscript and providing helpful suggestions. 20
References [1] A. Castro, L. Sankar, and R. Shivaji, Uniqueness of non-negative solutions for semipositone problems on exterior domains, J. Math. Anal. Appl. 394 (2012), 432-437. [2] A. Castro, M. Hassanpour, and R. Shivaji, Uniqueness of non-negative solutions for a semipositone problem with concave nonlinearities, Comm. Partial Differential Equations 20 (1995), 1927-1936. [3] A. Castro and R. Shivaji, Uniqueness of positive solutions for a class of elliptic boundary value problems, Proc. Roy. Soc. Edinburgh A 98A (1984), 267-269. [4] M. M. Coclite and G. Palmieri, On a singular nonlinear Dirichlet problem, Comm. Partial Differential Equations 14 (1989), 1315-1327. [5] M. G. Crandall, P. H. Rabinowitz, and L. Tartar, On a Dirichlet problem with a singular nonlinearity, Comm. Partial Differential Equations 2 (1977), 193-222. [6] E. N. Dancer, On the number of positive solutions of semilinear elliptic systems, Proc. London Math. Soc. 53 (1986), 429-452. [7] E. N. Dancer and J. Shi, Uniqueness and nonexistence of positive solutions to semipositone problems, Bull. London Math. Soc. 38 (2006), no. 6, 1033–1044. [8] J. I. Diaz, J. M. Morel, and L. Oswald, An elliptic equation with singular nonlinearity, Comm. Partial Differential Equations 2 (1997), 193-222. [9] J. I. Diaz and J. E. Saa, Existence et unicite de solutions positives pur certaines equations elliptiques quasilineares, Compte Rendus Acad. Sci. Paris 305 (1987), 521-524. [10] P. Drabek and J. Hernandez, Existence and uniqueness of positive solutions for some quasilinear elliptic problems, Nonlinear Anal. 27 (1996), 229-247.
21
[11] Z. M. Guo and J. R. L. Webb, Uniqueness of positive solutions for quasilinear elliptic equations when a parameter is large, Proc. Roy. Soc. Edinburgh 124 A (1994), 189-198. [12] W. Fulks and J. S. Maybee, A singular nonlinear equation, Osaka J. Math 12 (1960), 1-19. [13] D. D. Hai, Uniqueness of positive solutions for a class of quasilinear problems, Nonlinear Anal. 69 (2008), 2720-2732. [14] D. D. Hai, On singular Sturm-Liouville boundary value problems, Proc. Roy. Soc. Edinburgh 140A (2010), 49-63. [15] D. D. Hai and R. Shivaji, Existence and uniqueness for a class of quasilinear elliptic boundary value problems, J. Differential Equations 193 (2003), 500-510. [16] D. D. Hai and R. Shivaji, An existence result on positive solutions for a class of p-Laplacian systems, Nonlinear Anal. 56 (2004), 1007-1010. [17] D. D. Hai and R. C. Smith, Uniqueness for singular semilinear elliptic boundary value problems, Glasgow Math. J. 55 (2013), 399-409. [18] D. D. Hai and R. C. Smith, On uniqueness for a class of nonlinear boundary value problems, Proc. Roy. Soc. Edinburgh A 136A (2006), 779-784. [19] J. Hernandez, J. Karatson, and P. L. Simon, Multiplicity for semilinear elliptic equations involving singular nonlinearity, Nonlinear Anal. 65 (2006), 265-283. [20] J. Hernandez and F. J. Mancebo, Singular elliptic and parabolic equations, in Handbook of Differential Equations: Stationary Partial Differential Equations, (2006), 317-400. [21] S. Karlin and L. Nirenberg, On a theorem of P. Nowosad, J. Math. Anal. Appl. 17 (1967), 61-67. [22] E. K. Lee, L. Sankar, and R. Shivaji, Positive solutions for infinite semipositone problems on exterior domains, Differential Integral Equations 24 (2011), 861-875. 22
[23] E. K. Lee, R. Shivaji, and J. Ye, Classes of infinite semipositone systems, Proc. Roy. Soc. Edinburgh A 139A (2009), 853-865. [24] S. S. Lin, On the number of positive solutions for nonlinear elliptic equations when a parameter is large, Nonlinear Anal. 16 (1991), 283297. [25] S. Sakaguchi, Concavity properties of solutions to some degenerate quasilinear elliptic Dirichlet problems, Ann. Scuola Norm. Sup. Pisa Cl. Sci (4) 14 (1987), 403-421. [26] R. Shivaji, Uniqueness results for a class of positone problems, Nonlinear Anal. 7 (1983), 223-230. [27] R. Shivaji, I. Sim, and B. Son, A uniqueness result for a semipositone p-Laplacian problem on the exterior of a ball, J. Math. Anal. Appl. 445 (2017), 459-475. [28] C. A. Stuart, Existence and approximation of solutions of nonlinear elliptic equations, Math. Z. 147 (1976), 53-63. [29] C. A. Stuart, Existence theorems for a class of nonlinear integral equations, Math. Z. 137 (1974), 49-66. [30] M. Wiegner, A uniqueness theorem for some nonlinear boundary value problems with a large parameter, Math. Ann. 270 (1985), 401-402.
23