betaFeedback

WorksMathematics

Isolated singularities of certain quasi-linear elliptic inequalities II

Abstract

In this paper, we study the asymptotic behavior of isolated singularities for quasilinear elliptic equations of pp-Laplacian type. We focus on the pp-Laplace equation with the critical Serrin exponent n(p−1)/(n−p)n(p-1)/(n-p) and the nn-Laplace equation with critical exponential growth. We prove that any positive solution either admits a C1,αC^{1,α} extension to the origin or exhibits a precise singular asymptotic behavior. These results provide a complement to our previous study on pp-Laplacian type inequalities.

Introduction

The characterization of the behavior of isolated singularities for partial differential equations of the form

−Δpu=f(x,u,∇u)-\Delta_p u = f(x,u,\nabla u)

has attracted considerable attention over the past decades with extensive literature on this topic; see, e.g., [19, 20, 11, 2, 7, 1, 8, 5, 4, 3, 21, 17, 18] and the references therein. For the study of isolated singularities of pp-superharmonic functions and applications of pp-Laplace equations in conformal geometry, we refer the reader to [12, 13] and [14, 15] for the cases 1<p<n1 < p < n and p=np=n, respectively. A comprehensive review of this field is beyond the scope of this paper. Instead, it serves to complement our previous results and provide a full classification of isolated singularities in this context. Throughout the paper, Ω⊂Rn\Omega\subset\mathbb{R}^n denotes a domain. Unless otherwise specified, BrB_r denotes the open ball in Rn\mathbb{R}^n of radius rr centered at the origin, and Br(x)B_r(x) denotes the ball of radius rr centered at xx. In [16], the authors classified the isolated singularities of positive solutions of

0≤−Δpu≤un(p−1)n−p,x∈B1∖{0}⊂Rn,1<p<n.(1)0 \leq-\Delta_p u \leq u^{\frac{n(p-1)}{n-p}}, \quad x \in B_1 \setminus\{0\} \subset\mathbb{R}^n,\quad1 < p < n. \tag*{(1)}

and obtained the following theorem.

Theorem A. ([16 Theorem 1.2]). Let uu be a C2C^2 positive solution of (1). Then lim⁡x→0u(x)/u(∣x∣)=1\lim_{x \to0} u(x)/u(|x|) = 1. Moreover, exactly one of the following holds:

  1. uu admits a C1,αC^{1,\alpha} extension to 00 for some α>0\alpha> 0;

2.

lim⁡x→0∣x∣n−pp−1u(x)=m∈(0,∞);\lim_{x\to0}|x|^{\frac{n-p}{p-1}}u(x)=m\in(0,\infty);

3.

lim inf⁡x→0(log⁡1∣x∣)n−pp(p−1)∣x∣n−pp−1u(x)≥((n−pp)(n−pp−1)p−1)n−pp(p−1).\liminf_{x\to0}\left(\log\frac{1}{|x|}\right)^{\frac{n-p}{p(p-1)}}|x|^{\frac{n-p}{p-1}}u(x)\geq\left(\left(\frac{n-p}{p}\right)\left(\frac{n-p}{p-1}\right)^{p-1}\right)^{\frac{n-p}{p(p-1)}}.

A natural remaining question is to determine the precise asymptotic behavior for the positive singular solution of the equation

−Δpu=un(p−1)n−p,1<p<n.(2)-\Delta_p u=u^{\frac{n(p-1)}{n-p}},\quad1<p<n. \tag*{(2)}

Our first result is as follows.

Theorem 1. Let u∈C2(B1∖{0})u\in C^2(B_1\setminus\{0\}) be a positive solution of (2). Then either uu admits a C1,αC^{1,\alpha} extension to {0}\{0\} for some α>0\alpha>0 or

lim⁡x→0(log⁡1∣x∣)n−pp(p−1)∣x∣n−pp−1u(x)=((n−pp)(n−pp−1)p−1)n−pp(p−1).\lim_{x\to0}\left(\log\frac{1}{|x|}\right)^{\frac{n-p}{p(p-1)}}|x|^{\frac{n-p}{p-1}}u(x)=\left(\left(\frac{n-p}{p}\right)\left(\frac{n-p}{p-1}\right)^{p-1}\right)^{\frac{n-p}{p(p-1)}}.

This result establishes the pp-Laplace analogue of [1 Theorem A]. The corresponding result for radial solutions was previously obtained by Guedda and Veron [8]. In the classical case, the moving plane method in [5] serves as a standard approach to deriving symmetry and asymptotic results. However, extending this technique to the pp-Laplace equation (2) is technically demanding due to the lack of linear structure and appropriate comparison principles. Adopting the strategy initiated in [16], we employ the Emden-Fowler transformation to convert the problem into an analysis of a dynamical system within the framework of viscosity theory, thereby circumventing the difficulties associated with the moving plane method in this context.

Turning to the case p=np=n, a similar classification issue arises for the nn-Laplace equation, where the fundamental solution behaves logarithmically. In [14], the authors studied the following equation

−Δnu=f(x,u,∇u)-\Delta_n u=f(x,u,\nabla u)

and obtained the following result.

Theorem B ([14 Theorem 4.2 and Remark 4.1]). Let u∈C2(B2∖{0})u\in C^2(B_2\setminus\{0\}) be a nonnegative solution of

−Δnu=f(x,u,∇u)(3)-\Delta_n u=f(x,u,\nabla u) \tag*{(3)}

with

lim⁡x→0u(x)=∞,\lim_{x\to0}u(x)=\infty,

where ff satisfies

0≤f(x,u,∇u)≤C∣∇u∣βeκu(4)0 \le f(x,\mathbf{u},\nabla u) \le C|\nabla u|^{\beta}e^{\kappa u} \tag*{(4)}

for β∈(0,n)\beta\in(0,n) and κ>0\kappa> 0. Then

lim⁡x→0u(x)log⁡1∣x∣=m≥0\lim_{x\to0}\frac{u(x)}{\log\frac{1}{|x|}}=m\ge0

and

u(x)≥mlog⁡1∣x∣u(x)\ge m\log\frac{1}{|x|}

for x∈B1∖{0}x\in B_1\setminus\{0\}.

The case m=0m=0 was not explicitly discussed in [14]. Our second main result complements the previous theorem by addressing this borderline case.

Theorem 2. Let w∈C2(B1∖{0})w\in C^{2}(B_1\setminus\{0\}) be a nonnegative solution of (3) with ff satisfying (4). Then either ww admits a C1,αC^{1,\alpha} extension to {0}\{0\} for some 0<α<10<\alpha<1

or

lim⁡x→0w(x)log⁡1∣x∣=m>0.\lim_{x\to0}\frac{w(x)}{\log\frac{1}{|x|}}=m>0.

Combined with Theorem B, this result completes the description of the asymptotic behavior of solutions near the isolated singularity, covering both the removable singularity case (m=0m=0) and the singular case (m>0m>0). This fully generalizes Taliaferro’s classification [22 Theorem 2.1] to the nn-Laplace setting. The proof relies on potential theoretic arguments; specifically, we apply the pointwise potential estimates of Kilpeläinen and Malý [9] and the local regularity theory of Serrin [19] to establish the removability of the singularity.

Note added. The present manuscript was submitted for publication on May 8, 2026, before being posted on arXiv. While preparing this arXiv version, we became aware of the recent preprint by Xu and Yang [XY2026], posted on October 16, 2026, which establishes the same isolated-singularity classification for 1<p<n1<p<n in the class of nonnegative weak solutions, whereas Theorem 1 is stated for positive C2C^2 solutions. Their proof builds on the spherical-extrema approach introduced in our earlier work [16], with an additional BV argument to handle weak solutions. The present paper also treats the case p=np=n in Theorem 2, which is not covered in [XY2026].

The rest of the paper is organized as follows. In Section 2, we collect some basic definitions and preliminary results. We then prove Theorem 1 and Theorem 2 in Section 3 and Section 4 respectively.

Preliminaries

In this section, we collect several definitions and basic results that will be used throughout the paper.

Definition 1 ([10 Definition 2.5]). Let Ω\Omega be a domain in Rn\mathbb{R}^{n}. We say that u∈Wloc1,p(Ω)u \in W_{\mathrm{loc}}^{1,p}(\Omega) is a weak solution to

−Δpu=0,-\Delta_p u = 0,

if for each ϕ∈C0∞(Ω)\phi\in C_0^\infty(\Omega), there holds

∫Ω⟨∣∇u∣p−2∇u,∇ϕ⟩dx=0.\int_{\Omega} \langle|\nabla u|^{p-2}\nabla u,\nabla\phi\rangle dx = 0.

If in addition, uu is continuous, we say uu is pp-harmonic.

Definition 2 ([10 Definition 5.1]). A function u:Ω→(−∞,∞]u:\Omega\to(-\infty,\infty] is called pp-superharmonic in Ω\Omega, if

  • uu is lower semi-continuous in Ω\Omega;

  • u≢∞u\not\equiv\infty in Ω\Omega;

  • for each domain D⋐ΩD\Subset\Omega the comparison principle holds: if h∈C(D‾)h\in C(\overline{D}) is pp-harmonic in DD and h∣∂D≤u∣∂Dh|_{\partial D}\le u|_{\partial D}, then h≤uh\le u in DD.

One crucial estimate for nonnegative pp-superharmonic functions is the following theorem due to Kilpeläinen and Malý [9].

Theorem 2.1 ([9 Theorem 1.6]). Let 1<p≤n1<p\le n. Suppose uu is a nonnegative pp-superharmonic function satisfying

−Δpu=μ,x∈B3r(x0).-\Delta_p u=\mu,\qquad x\in B_{3r}(x_0).

Then there exists C1(n,p),C2(n,p)>0C_1(n,p),C_2(n,p)>0 such that

C_1(n,p)W_{1,p}^{\mu}(x_0,r)\le u(x_0)\le C_2(n,p)\left(\fint_{B_r(x_0)}u+W_{1,p}^{\mu}(x_0,2r)\right).

Proof of Theorem 1

The following lemma is a generalization of Lions [11], which is the p=2p=2 case.

Lemma 3.1. Let u≥0u\ge0 be a solution of

−Δpu=uσin B1∖{0}(5)-\Delta_pu=u^\sigma\qquad\text{in }B_1\setminus\{0\} \tag*{(5)}

with 1<p<n1<p<n. If σ≥n(p−1)n−p\sigma\ge\frac{n(p-1)}{n-p}, then uσ∈L1(B1)u^\sigma\in L^1(B_1) and (5) holds in B1B_1 in the sense of distributions.

Proof. This follows directly from Phuc and Verbitsky [18 Theorem 2.18], which implies that (5) holds in B1B_1 if and only if

Cap⁡Gp,σσ+1−p({0})=0.\operatorname{Cap}_{G_p,\frac{\sigma}{\sigma+1-p}}(\{0\})=0.

Noting that a point has zero Bessel capacity Cap⁡Gα,s\operatorname{Cap}_{G_{\alpha,s}} if and only if αs≤n\alpha s \le n, this condition is equivalent to

pσσ+1−p≤n,i.e.,σ≥n(p−1)n−p.\frac{p\sigma}{\sigma+1-p} \le n,\qquad\text{i.e.,}\quad\sigma\ge\frac{n(p-1)}{n-p}.

Here, the Bessel capacity is defined for any set EE by

Cap⁡Gα,s(E):=inf⁡{∥f∥Ls(Rn):Gαf≥χE, f≥0}.\operatorname{Cap}_{G_{\alpha,s}}(E):=\inf\left\{\lVert f\rVert_{L^s(\mathbb{R}^n)}:G_\alpha f\ge\chi_E,\ f\ge0\right\}.

Lemma 3.2. Under the assumption of Theorem 1, we have

lim⁡x→0u(x)∣x∣p−np−1=0.\lim_{x\to0}\frac{u(x)}{|x|^{\frac{p-n}{p-1}}}=0.

Proof. By [16 Theorem 1.2], there exists a constant m≥0m\ge0 such that

lim⁡x→0u(x)∣x∣p−np−1=m.\lim_{x\to0}\frac{u(x)}{|x|^{\frac{p-n}{p-1}}}=m.

Moreover, Lemma 3.1 implies that u∈Ln(p−1)n−p(B1)u\in L^{\frac{n(p-1)}{n-p}}(B_1), and therefore −Δpu∈Lloc1(B1)-\Delta_pu\in L^1_{\mathrm{loc}}(B_1). As a consequence, the Radon measure −Δpu-\Delta_pu has no mass concentration at the origin. It then follows from [12 Theorem 4] that m=0m=0.

Then we follow the framework of [16]. For r>0r>0, we let

u‾(r)=inf⁡∂Bru(x),u‾(r)=sup⁡∂Bru(x).\underline{u}(r)=\inf_{\partial B_r}u(x),\qquad\overline{u}(r)=\sup_{\partial B_r}u(x).

Definition 3. Let f:(α,β)→Rf:(\alpha,\beta)\to\mathbb{R} be a function.

  1. ff is called concave if for any r1,r2∈(α,β)r_1,r_2\in(\alpha,\beta) and λ∈(0,1)\lambda\in(0,1),

    f(λr1+(1−λ)r2)≥λf(r1)+(1−λ)f(r2);f(\lambda r_1+(1-\lambda)r_2)\ge\lambda f(r_1)+(1-\lambda)f(r_2);
  1. ff is called semi-concave if there exists a constant C>0C>0 such that the function

    r↦f(r)−Cr2r\mapsto f(r)-Cr^2

    is concave in (α,β)(\alpha,\beta);

  1. ff is called locally semi-concave if for any [α1,β1]⊂(α,β)[\alpha_1,\beta_1]\subset(\alpha,\beta), ff is semi-concave in (α1,β1)(\alpha_1,\beta_1).

We recall the following properties of semi-concave functions.

Lemma 3.3. Let f:(α,β)→Rf:(\alpha,\beta)\to\mathbb{R} be a locally semi-concave function.

  1. ff is continuous, and the one-sided derivatives fr±f_r^{\pm} exist everywhere. The derivative fr−f_r^{-} is left-continuous, while fr+f_r^{+} is right-continuous. Moreover, the strict inequality fr−(r)>fr+(r)f_r^{-}(r)>f_r^{+}(r) occurs only on an at most countable set.

  1. The distributional second derivative of ff is a Radon measure ν\nu, which admits the decomposition ν=νac+νpp+νsc\nu=\nu_{\mathrm{ac}}+\nu_{\mathrm{pp}}+\nu_{\mathrm{sc}}. Here, νac\nu_{\mathrm{ac}} denotes the absolutely continuous part with respect to the Lebesgue measure, νpp\nu_{\mathrm{pp}} is the pure point component supported on {ri}\{r_i\}, and νsc\nu_{\mathrm{sc}} is the singular continuous component. Furthermore, the singular parts are nonpositive, i.e., νpp≤0\nu_{\mathrm{pp}}\le0 and νsc≤0\nu_{\mathrm{sc}}\le0.

  1. The generalized Newton-Leibniz formula holds: for any [α1,β1]⊂(α,β)[\alpha_1,\beta_1]\subset(\alpha,\beta),

fr−(β1)−fr+(α1)=∫α1β1dν.f_r^{-}(\beta_1)-f_r^{+}(\alpha_1)=\int_{\alpha_1}^{\beta_1}d\nu.

We now apply these properties to our specific setting via the Emden-Fowler transformation. Let η=n−pp−1\eta=\frac{n-p}{p-1} and consider the change of variables t=ηlog⁡ηrt=\eta\log\frac{\eta}{r} for r∈(0,η)r\in(0,\eta). We define the function w(t)w(t) on (ηlog⁡η,∞)(\eta\log\eta,\infty) by etw(t)=u(r)e^t w(t)=u(r).

Lemma 3.4. The function w(t)w(t) is locally semi-concave on (ηlog⁡η,∞)(\eta\log\eta,\infty) with

lim⁡t→∞w(t)=0.\lim_{t\to\infty}w(t)=0.

Moreover, in the decomposition wtt=νac+νpp+νscw_{tt}=\nu_{\mathrm{ac}}+\nu_{\mathrm{pp}}+\nu_{\mathrm{sc}}, the singular parts νpp\nu_{\mathrm{pp}} and νsc\nu_{\mathrm{sc}} are nonpositive, and νpp\nu_{\mathrm{pp}} is supported on a countable set {t~i}\{\tilde t_i\}.

The reader may refer to [16 Lemma 6.3] for the proof. To proceed with the analysis, we denote z(t)=wt−(t)z(t)=w_t^{-}(t) and q=n(p−1)n−pq=\frac{n(p-1)}{n-p}, and consider the trajectory (w(t),z(t))(w(t),z(t)) in the (w,z)(w,z)-plane.

Lemma 3.5. 1. w(t)w(t) is continuous for t≥ηlog⁡ηt\ge\eta\log\eta. For t∈supp⁡(νpp)={t~i}t\in\operatorname{supp}(\nu_{\mathrm{pp}})=\{\tilde t_i\}, wt−(t~i)>wt+(t~i)w_t^{-}(\tilde t_i)>w_t^{+}(\tilde t_i). Moreover, z(t)z(t) is lower semi-continuous, right continuous, of locally bounded variation in [ηlog⁡η,∞)[\eta\log\eta,\infty) and is continuous for t≠t~it\ne\tilde t_i.

  1. If w(t)+z(t)>0w(t)+z(t)>0 for t∈[t0,t1]⊂(ηlog⁡η,∞)t\in[t_0,t_1]\subset(\eta\log\eta,\infty) then

z(t1)−z(t0)≤wt−(t1)−wt+(t0)≤∫t0t1(−z−wq(p−1)(w+z)p−2) dt.z(t_1)-z(t_0)\le w_t^{-}(t_1)-w_t^{+}(t_0)\le\int_{t_0}^{t_1}\left(-z-\frac{w^q}{(p-1)(w+z)^{p-2}}\right)\,dt.

The proof is a direct consequence of Lemma 3.3 and is analogous to that of [16 Lemma 6.4].

Lemma 3.6. There is T>0T>0 such that, z(t)<0z(t)<0 for all t∈[T,∞)t\in[T,\infty).

Proof. Choose T>ηlog⁡ηT>\eta\log\eta such that z(T)<0z(T)<0. Suppose, for contradiction, that z(t)<0z(t)<0 does not hold for all t≥Tt\ge T. Set

t1=sup⁡{s≥T:z(t)<0 for all t∈[T,s)}.t_1=\sup\{s\ge T:z(t)<0\text{ for all }t\in[T,s)\}.

By definition we have z‾(t)<0\underline{z}(t) < 0 on [T,t1)[T,t_1). Since z‾(t)\underline{z}(t) is right continuous, we must have z‾(t1)≥0\underline{z}(t_1) \ge0. On the other hand, the lower semicontinuity of z‾\underline{z} together with z‾(t)<0\underline{z}(t) < 0 for t<t1t < t_1 implies

lim⁡t→t1z‾(t)=0,\lim_{t \to t_1} \underline{z}(t) = 0,

hence z‾(t1)=0\underline{z}(t_1) = 0. Moreover, since

0≥lim inf⁡t→t1z‾(t)≥z‾(t1)=0,0 \ge\liminf_{t \to t_1} \underline{z}(t) \ge\underline{z}(t_1) = 0,

we derive

lim⁡t→t1z‾(t)=0.\lim_{t \to t_1} \underline{z}(t) = 0.

Next, observe that w(t)w(t) is monotone decreasing on [T,t1][T,t_1], so w(t)≥w(t1)w(t) \ge w(t_1) for t∈[T,t1]t \in[T,t_1]. Therefore we can choose δ>0\delta> 0 sufficiently small so that on [t1−δ,t1][t_1-\delta,t_1] the following strict inequality holds

−z‾−wq(p−1)(w+z‾)p−2<0.-\underline{z} - \frac{w^q}{(p-1)(w+\underline{z})^{p-2}} < 0.

By Item 2 of Lemma 3.5, z‾(t1)<z‾(t1−δ)\underline{z}(t_1) < \underline{z}(t_1-\delta), which is a contradiction.

For fixed 0<ε0,δ0<10 < \varepsilon_0,\delta_0 < 1, we let k=k(ε0)=1(p−1)(1+ε0)k=k(\varepsilon_0)=\frac{1}{(p-1)(1+\varepsilon_0)}, and

Θkδ0={(w,z); 0<w≤δ0, z<−kwq−p+2}.\Theta_k^{\delta_0}=\{(w,z);\,0<w\le\delta_0,\ z<-kw^{q-p+2}\}.

Lemma 3.7. Suppose that for fixed ε0,δ0∈(0,1)\varepsilon_0,\delta_0\in(0,1), there exists T>0T>0 such that

(w(t),z(t))∈Θkδ0for all t>T.(w(t),z(t))\in\Theta_k^{\delta_0}\qquad\text{for all }t>T.

Then the conclusion of Theorem 1 holds.

Proof. Firstly, we establish that if the trajectory (w(t),z(t))(w(t),z(t)) remains in the region Θkδ0\Theta_k^{\delta_0} for all t>Tt>T, then the differential inequality

ddtw(t)<−kwq−p+2\frac{d}{dt}w(t)<-kw^{q-p+2}

holds in the sense of distributions on (T,∞)(T,\infty). Indeed, by Lemma 3.5, w‾t−(t~i)>w‾t+(t~i)\underline{w}_t^-(\tilde{t}_i)>\underline{w}_t^+(\tilde{t}_i) on {t~i}=supp⁡(νpp)\{\tilde{t}_i\}=\operatorname{supp}(\nu_{pp}) and we obtain

ddtwp−q−1>−k(p−q−1)\frac{d}{dt}w^{p-q-1}>-k(p-q-1)

in the sense of distributions. For fixed t1>Tt_1>T, integrating the inequality above from t1t_1 to t>t1t>t_1 gives

wp−q−1(t)−wp−q−1(t1)>−k(p−q−1)(t−t1),w^{p-q-1}(t)-w^{p-q-1}(t_1)>-k(p-q-1)(t-t_1),

and hence, for tt large,

wp−q−1(t)≥−p−q−1(p−1)(1+ε0)t+C=p(n−p)(1+ε0)t+C,≥p(n−p)(1+2ε0)t.w^{p-q-1}(t) \ge-\frac{p-q-1}{(p-1)(1+\varepsilon_0)}t+C = \frac{p}{(n-p)(1+\varepsilon_0)}t+C, \ge\frac{p}{(n-p)(1+2\varepsilon_0)}t.

Consequently, for all sufficiently large tt,

w(t)≤(p(n−p)(1+2ε0))−n−pp(p−1)t−n−pp(p−1).w(t) \le\left(\frac{p}{(n-p)(1+2\varepsilon_0)}\right)^{-\frac{n-p}{p(p-1)}}t^{-\frac{n-p}{p(p-1)}}.

Since ε0>0\varepsilon_0>0 was arbitrary (and may be taken arbitrarily small), we deduce

lim sup⁡t→∞tn−pp(p−1)w(t)≤(pn−p)−n−pp(p−1).\limsup_{t\to\infty}t^{\frac{n-p}{p(p-1)}}w(t)\le\left(\frac{p}{n-p}\right)^{-\frac{n-p}{p(p-1)}}.

Translating this estimate back to the original variable xx yields

lim sup⁡x→0(log⁡1∣x∣)n−pp(p−1)∣x∣n−pp−1u(∣x∣)≤((n−pp)(n−pp−1)p−1)n−pp(p−1).\limsup_{x\to0}\left(\log\frac{1}{|x|}\right)^{\frac{n-p}{p(p-1)}}|x|^{\frac{n-p}{p-1}}u(|x|) \le\left(\left(\frac{n-p}{p}\right)\left(\frac{n-p}{p-1}\right)^{p-1}\right)^{\frac{n-p}{p(p-1)}}.

On the other hand, if the singularity is non-removable, Theorem A provides the matching lower bound

lim inf⁡x→0(log⁡1∣x∣)n−pp(p−1)∣x∣n−pp−1uˉ(∣x∣)≥((n−pp)(n−pp−1)p−1)n−pp(p−1).\liminf_{x\to0}\left(\log\frac{1}{|x|}\right)^{\frac{n-p}{p(p-1)}}|x|^{\frac{n-p}{p-1}}\bar{u}(|x|) \ge\left(\left(\frac{n-p}{p}\right)\left(\frac{n-p}{p-1}\right)^{p-1}\right)^{\frac{n-p}{p(p-1)}}.

and moreover uˉ(∣x∣)/u(∣x∣)→1\bar{u}(|x|)/u(|x|)\to1 as x→0x\to0. Combining these facts yields

lim⁡x→0(log⁡1∣x∣)n−pp(p−1)∣x∣n−pp−1u(x)=((n−pp)(n−pp−1)p−1)n−pp(p−1),\lim_{x\to0}\left(\log\frac{1}{|x|}\right)^{\frac{n-p}{p(p-1)}}|x|^{\frac{n-p}{p-1}}u(x) =\left(\left(\frac{n-p}{p}\right)\left(\frac{n-p}{p-1}\right)^{p-1}\right)^{\frac{n-p}{p(p-1)}},

which is precisely the conclusion of Theorem 1.

Central to our argument is the following lemma.

Lemma 3.8. There exists T>0T>0 such that if (w(t0),z(t0))∈Θkδ0(w(t_0),z(t_0))\in\Theta_k^{\delta_0} for some t0>Tt_0>T, then the trajectory remains in Θkδ0\Theta_k^{\delta_0} for all t>t0t>t_0.

Proof. The proof is analogous to that of [16 Lemma 6.6]. Yet we still provide the full details below. Since

lim⁡t→∞w(t)=0,\lim_{t\to\infty}w(t)=0,

we may choose T>0T > 0 sufficiently large such that 0<w(t)<δ00 < w(t) < \delta_0 for all t≥Tt \ge T. Assume now that for some t0≥Tt_0 \ge T,

(w(t0),z(t0))∈Θkδ0(w(t_0),z(t_0)) \in\Theta_k^{\delta_0}

For t≥Tt \ge T, we have w(t)>0w(t) > 0 and z(t)z(t) is locally bounded from below. Therefore, the only possible way in which the trajectory (w(t),z(t))(w(t),z(t)) could leave Θkδ0\Theta_k^{\delta_0} for t>t0t > t_0 is that there exists t1>t0t_1 > t_0 such that z(t1)=−kwq−p+2(t1)z(t_1) = -kw^{q-p+2}(t_1) and

(w(t),z(t))∈Θkδ0for all t∈[t0,t1),(w(t),z(t)) \in\Theta_k^{\delta_0} \qquad\text{for all } t \in[t_0,t_1),

and

(w(t),z(t))→(w(t1),z(t1))as t→t1−.(w(t),z(t)) \to(w(t_1),z(t_1)) \qquad\text{as } t \to t_1^-.

From this we infer

lim sup⁡t→t1−z(t)−z(t1)w(t)−w(t1)≤d(−kwq−p+2)dw∣w=w(t1)=−k(q−p+2)wq−p+1(t1)→0\begin{aligned} \limsup_{t\to t_1^-}\frac{z(t)-z(t_1)}{w(t)-w(t_1)} &\le\left.\frac{d(-kw^{q-p+2})}{dw}\right|_{w=w(t_1)}\\ &=-k(q-p+2)w^{q-p+1}(t_1)\\ &\to0 \end{aligned}

as t1→∞t_1 \to\infty.

On the other hand, since z(t)<0z(t) < 0 for t∈[t0,t1]t \in[t_0,t_1], we have w(t1)<w(t)w(t_1) < w(t) for t<t1t < t_1. Using Lemma 3.5, we obtain

lim inf⁡t→t1−z(t1)−z(t)w(t1)−w(t)=lim inf⁡t→t1−z(t1)−z(t)t1−tt1−tw(t1)−w(t)≥lim inf⁡t→t1−∫tt1(−z−wq(p−1)∣w+z∣p−2) dτt1−tt1−tw(t1)−w(t)=(−z(t1)−wq(t1)(p−1)∣w(t1)+z(t1)∣p−2)1z(t1)=−1+wp−2(t1)k(p−1)∣w(t1)+z(t1)∣p−2→−1+1+ε0=ε0\begin{aligned} \liminf_{t\to t_1^-}\frac{z(t_1)-z(t)}{w(t_1)-w(t)} &=\liminf_{t\to t_1^-}\frac{z(t_1)-z(t)}{t_1-t}\frac{t_1-t}{w(t_1)-w(t)}\\ &\ge\liminf_{t\to t_1^-}\frac{\int_t^{t_1}\left(-z-\frac{w^q}{(p-1)|w+z|^{p-2}}\right)\,d\tau}{t_1-t}\frac{t_1-t}{w(t_1)-w(t)}\\ &=\left(-z(t_1)-\frac{w^q(t_1)}{(p-1)|w(t_1)+z(t_1)|^{p-2}}\right)\frac{1}{z(t_1)}\\ &=-1+\frac{w^{p-2}(t_1)}{k(p-1)|w(t_1)+z(t_1)|^{p-2}}\\ &\to-1+1+\varepsilon_0=\varepsilon_0 \end{aligned}

as t1→∞t_1 \to\infty.

For t1t_1 sufficiently large, the above lower bound ε0>0\varepsilon_0 > 0 contradicts the previous upper bound tending to 0, which ends the proof of Lemma 3.8. □\square

In light of Lemma 3.8, the proof of Theorem 1 is reduced to establishing the existence of an entry time into the region Θkδ0\Theta_k^{\delta_0}.

Lemma 3.9. Let ε0\varepsilon_0, δ0\delta_0, kk and TT be as defined in Lemma 3.8. Then there exists t0≥Tt_0 \ge T such that (w(t0),z(t0))∈Θkδ0(w(t_0),z(t_0)) \in\Theta_k^{\delta_0}.

Proof. Suppose by contradiction that no such t0t_0 exists; equivalently,

z‾(t)≥−kw‾(t)q−p+2for all t≥T.\underline{z}(t) \geq-k\underline{w}(t)^{q-p+2}\quad\text{for all }t \geq T.

Since z‾<0\underline{z}<0 and w‾\underline{w} is monotonically decreasing in tt, we may regard z‾\underline{z} as a function of w‾\underline{w}. In particular

lim⁡w‾→0z‾(w‾)=0.(6)\lim_{\underline{w}\to0}\underline{z}(\underline{w})=0. \tag*{(6)}

However, on the other hand

dz‾dt≤−z‾−w‾q(p−1)(w‾+z‾)p−2\frac{d\underline{z}}{dt}\leq-\underline{z}-\frac{\underline{w}^{q}}{(p-1)(\underline{w}+\underline{z})^{p-2}}

in the sense of distributions. And if t≠t~it\ne\widetilde{t}_i,

dw‾dt=z‾<0.\frac{d\underline{w}}{dt}=\underline{z}<0.

For such t≠t~it\ne\widetilde{t}_i, since z‾≥−kw‾q−p+2\underline{z}\geq-k\underline{w}^{q-p+2}, we obtain

dz‾dw‾≥−1−w‾q(p−1)(w‾+z‾)p−2z‾≥−1−w‾p−2k(p−1)(w‾+z‾)p−2≥−1−1+ε(1−kw‾q−p+1)p−2≥ε2,\begin{aligned} \frac{d\underline{z}}{d\underline{w}}\geq-1-\frac{\underline{w}^{q}}{(p-1)(\underline{w}+\underline{z})^{p-2}\underline{z}} \\ &\geq-1-\frac{\underline{w}^{p-2}}{k(p-1)(\underline{w}+\underline{z})^{p-2}} \\ &\geq-1-\frac{1+\varepsilon}{(1-k\underline{w}^{q-p+1})^{p-2}} \\ &\geq\frac{\varepsilon}{2}, \end{aligned}

for t>0t>0 large enough. If t≠t~it\ne\widetilde{t}_i, since

lim⁡t→t~i−z‾(t)≥z‾(t~i),\lim_{t\to\widetilde{t}_i^-}\underline{z}(t)\geq\underline{z}(\widetilde{t}_i),

it follows that

dz‾dw‾∣t=t~i=miδ(t−t~i)\left.\frac{d\underline{z}}{d\underline{w}}\right|_{t=\widetilde{t}_i}=m_i\delta(t-\widetilde{t}_i)

for mi>0m_i>0, where δ\delta denotes the Dirac measure. Consequently, we have

dz‾dw‾≥ε2\frac{d\underline{z}}{d\underline{w}}\geq\frac{\varepsilon}{2}

in the sense of distributions for sufficiently large tt. This contradicts the fact that z‾(w‾)<0\underline{z}(\underline{w})<0.

Proof of Theorem 2

Recall that by [4 Proposition 1.1], we have w∈Llocs(B1)w \in L_{\mathrm{loc}}^{s}(B_1) for all s∈[1,∞)s \in[1,\infty) and ∇w∈Llocq(B1)\nabla w \in L_{\mathrm{loc}}^{q}(B_1) for all q∈[1,n)q \in[1,n). Moreover, if lim⁡x→0w(x)=∞\lim_{x\to0} w(x)=\infty, then ww satisfies

−Δnw=τδ(x)+gin D′(B1),-\Delta_n w=\tau\delta(x)+g \quad\text{in }\mathcal{D}'(B_1),

where τ≥0\tau\ge0 and g∈L1(B1)g \in L^1(B_1) is nonnegative. By Theorem B, we also have

lim⁡x→0w(x)log⁡1∣x∣=m≥0.\lim_{x\to0}\frac{w(x)}{\log\frac{1}{|x|}}=m\ge0.

It remains to analyze the case m=0m=0. We claim that m=0m=0 implies τ=0\tau=0. Indeed, if τ>0\tau>0, then by Theorem 2.1, there exists a constant C>0C>0 such that

w(x)≥CW1,nτδx+g(x,110)for x∈B1/2.w(x)\ge C W_{1,n}^{\tau\delta_{x}+g}\left(x,\frac{1}{10}\right) \quad\text{for }x\in B_{1/2}.

According to [12 Theorem 1], this lower bound implies

lim inf⁡x→0w(x)log⁡1∣x∣>0,\liminf_{x\to0}\frac{w(x)}{\log\frac{1}{|x|}}>0,

which contradicts the assumption m=0m=0. Thus, we must have τ=0\tau=0. Since m=0m=0, for any ε>0\varepsilon>0, there exists rε>0r_\varepsilon>0 such that

0≤w(x)≤εlog⁡1∣x∣for 0<∣x∣<rε.0\le w(x)\le\varepsilon\log\frac{1}{|x|}\quad\text{for }0<|x|<r_\varepsilon.

This logarithmic growth bound implies that eκw∈Llocs(B1)e^{\kappa w}\in L_{\mathrm{loc}}^{s}(B_1) for every s∈[1,∞)s\in[1,\infty). Combining this integrability with the gradient estimate in [10 Theorem 5.17], we deduce that

∣∇w∣βeκw∈Lγ(B1)for some γ>1.|\nabla w|^{\beta}e^{\kappa w}\in L^{\gamma}(B_1)\quad\text{for some }\gamma>1.

Consequently, the structure conditions (5), (6), and (8) in [19] are satisfied, and we may invoke [19 Theorem 8] to conclude that ww is Hölder continuous in B1B_1. Finally, by [6 Theorem 2] (or [23 Theorem 1]), it follows that ∇w\nabla w is also Hölder continuous in B1B_1. This completes the proof.

Acknowledgements: The first named author is supported by Natural Science Foundation of Tianjin (Grant No. 22JCJQJC00130), the Fundamental Research Funds for the Central Universities.

Shiguang Ma: School of Mathematical Sciences and LPMC, Nankai University Tianjin 300071, China e-mail: [email protected]

Shengyang Zang: School of Mathematical Sciences, Nankai University Tianjin 300071, China e-mail: [email protected]

References

  1. [1]P. Aviles, Local behavior of solutions of some elliptic equations, Comm. Math. Phys. 108 (1987) 177–192.DOI
  2. [2]H. Brezis and P.L. Lions, A note on isolated singularities for linear elliptic equation, Adv. in Math. Suppl. Stud. A. 7 (1981) 263–266.
  3. [3]M.-F. Bidaut-Véron and S. Pohozaev, Nonexistence results and estimates for some nonlinear elliptic problems, J. Anal. Math. 84 (2001) 1–49.
  4. [4]M.-F. Bidaut-Véron, Local and global behavior of solutions of quasilinear equations of Emden-Fowler type, Arch. Ration. Mech. Anal. 107 (1989) 293–324.
  5. [5]L.A. Caffarelli, B. Gidas and J. Spruck, Asymptotic symmetry and local behavior of semilinear elliptic equations with critical Sobolev growth, Comm. Pure Appl. Math. 42 (1989) 271–297.DOI
  6. [6]E. Di Benedetto, C^{1+α} local regularity of weak solutions of degenerate elliptic equations, Nonlinear Anal. 7 (1983) 827–850.
  7. [7]B. Gidas and J. Spruck, Global and local behavior of positive solutions of nonlinear elliptic equations, Comm. Pure Appl. Math. 34 (1981) 525–598.DOI
  8. [8]M. Guedda and L. Véron, Local and global properties of solutions of quasilinear elliptic equations, J. Differential Equations 76 (1988) 159–189.DOI
  9. [9]T. Kilpeläinen and J. Malý, The Wiener test and potential estimates for quasilinear elliptic equations, Acta Math. 172 (1994) 137–161.DOI
  10. [10]P. Lindqvist, Notes on the p-Laplace Equation, University of Jyväskylä Lecture Notes (2017).
  11. [11]P.L. Lions, Isolated singularities in semilinear problems, J. Differential Equations 38 (1980) 441–450.DOI
  12. [12]H. Liu, S. Ma, J. Qing and S. Zhong, On the asymptotic behavior of p-superharmonic functions at singularities, preprint (2023) arXiv:2310.11610.arxiv.org/abs/2310.11610
  13. [13]H. Liu, S. Ma, J. Qing and S. Zhong, p-Laplace equations in conformal geometry, Sci. China Math. 68 (2025) 1137–1150.arxiv.org/abs/2308.02468
  14. [14]S. Ma and J. Qing, On n-superharmonic functions and some geometric applications, Calc. Var. Partial Differential Equations 60 (2021) Paper No. 234, 42 pp.arxiv.org/abs/1810.10561
  15. [15]S. Ma and J. Qing, On Huber-type theorems in general dimensions, Adv. Math. 395 (2022), Paper No. 108145, 37 pp.DOI
  16. [16]S. Ma and S. Zang, Isolated singularities of solutions of certain quasilinear elliptic inequalities, Math. Ann. 393 (2025) 733–762.arxiv.org/abs/2505.12604
  17. [17]P. Poláčík, P. Quittner and P. Souplet, Singularity and decay estimates in superlinear problems via Liouville-type theorems, I: elliptic equations and systems, Duke Math. J. 139 (2007) 555–579.DOI
  18. [18]N.C. Phuc and I.E. Verbitsky, Quasilinear and Hessian equations of Lane-Emden type, Ann. of Math. 168 (2008) 859–914.DOI
  19. [19]J. Serrin, Local behavior of solutions of quasi-linear equations, Acta Math. 111 (1964) 247–302.DOI
  20. [20]J. Serrin, Isolated singularities of solutions of quasi-linear equations, Acta Math. 113 (1965) 219–240.DOI
  21. [21]J. Serrin and H. Zou, Cauchy-Liouville and universal boundedness theorems for quasi-linear elliptic equations and inequalities, Acta Math. 189 (2002) 79–142.DOI
  22. [22]S.D. Taliaferro, Isolated singularities of nonlinear elliptic inequalities. II. Asymptotic behavior of solutions, Indiana Univ. Math. J. 55 (2006) 1791–1812.DOI
  23. [23]P. Tolksdorf, Regularity for a more general class of quasilinear elliptic equations, J. Differential Equations 51 (1984) 126–150.DOI

Paper details

vlnc:W_nmrkh6ck
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 9

downloads · 1

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents