betaFeedback

WorksMathematics

A sufficient condition for the Liouville property of solutions to the Euler-Lagrange equations of variational integrals with nonstandard growth in terms of the ellipticity modulus near the origin

Abstract

We consider entire solutions uu: Rn→R\mathbb{R}^{n}\rightarrow\mathbb{R} of the Euler-Lagrange equation div⁡(∇f(∇u))=0\operatorname{div}(\nabla f(\nabla u))=0 with convex density ff: Rn→R\mathbb{R}^{n}\rightarrow\mathbb{R} of nonstandard growth and nonuniform ellipticity. It turns out that the Liouville property can be characterized through the behavior of D2f(p)D^{2}f(p) for points p∈Rnp\in\mathbb{R}^{n} close to the origin: if, for example, D2f(p)D^{2}f(p) behaves as ∣p∣ρ−21|p|^{ρ-2}\mathbf{1}, 1=unit matrix\mathbf{1}=\text{unit matrix}, with exponent ρ≥2ρ\ge2 and for ∣p∣≤1|p|\le1 then the validity of n≤ρ+2n\leρ+2 together with the boundedness of uu implies its constancy.

Introduction

In this note we discuss the Liouville property for entire solutions u∈C2(Rn)u \in C^2(\mathbb{R}^n) of the nonlinear elliptic equation

div⁡(∇f(∇u))=0(1)\operatorname{div}(\nabla f(\nabla u))=0 \tag*{(1)}

arising as the Euler-Lagrange equation of the variational integral ∫f(∇u) dx\int f(\nabla u)\,dx with convex density f:Rn→Rf:\mathbb{R}^n \to\mathbb{R} of class C2C^2. Here we mean by Liouville property the fact that any bounded entire solution u:Rn→Ru:\mathbb{R}^n \to\mathbb{R} of (1.1) must be a constant function.

Surprisingly, it turns out that a positive answer just depends on the behavior of D2f(p)D^2f(p) for pp close to the origin, for example, if D2f(p)D^2f(p) is of order ∣p∣ρ−2|p|^{\rho-2} as p→0p \to0 with exponent ρ≥2\rho\ge2, then we have the Liouville property provided that n≤ρ+2n \le\rho+2, and this is true under rather general hypotheses concerning the growth of ff at infinity.

Before going into details we like to remark that the study of equations with nonstandard growth and nonuniform ellipticity is a very active field, which has attracted the research of many scientists over the last decades focussing on the questions of existence and regularity of solutions. For a detailed overview including historical comments the reader is referred to [1]. However, we could not find a discussion of the Liouville property in this general setting, but it should be mentioned that there are contributions concerning quasilinear or uniformly elliptic equations, which can be found for instance in [2], [3], [4], [5] and in the references quoted therein. We also advise the reader to consult the textbooks [6], [7].

Let us now state our assumptions concerning the density f∈C2(Rn)f \in C^2(\mathbb{R}^n) occurring in equation (1.1): we have

D2f(p)(q,q)>0,p,q∈Rn, q≠0,∇f(0)=0,(2)D^2f(p)(q,q)>0,\quad p,q\in\mathbb{R}^n,\ q\ne0,\quad\nabla f(0)=0, \tag*{(2)}

implying the strict convexity of ff. We further require (c,c1,…c,c_1,\ldots denoting positive constants)

∣∇f(p)∣≤c∣p∣m−1for p∈Rn, ∣p∣≥1(3)|\nabla f(p)|\le c|p|^{m-1}\quad\text{for }p\in\mathbb{R}^n,\ |p|\ge1 \tag*{(3)}

with an arbitrary exponent m>1m>1, and addition we assume

p⋅∇f(p)≥c1∣p∣−μ∣p∣∣∇f(p)∣,p∈Rn, ∣p∣≥1(4)p\cdot\nabla f(p)\ge c_1|p|^{-\mu}|p||\nabla f(p)|,\quad p\in\mathbb{R}^n,\ |p|\ge1 \tag*{(4)}

for some number μ∈[0,1)\mu\in[0,1) for which we also have

∣D2f(p)∣ln⁡2(1+∣p∣)≤c2∣p∣−μ∣∇f(p)∣,p∈Rn, ∣p∣≥1.(5)|D^2f(p)|\ln^2(1+|p|)\le c_2|p|^{-\mu}|\nabla f(p)|,\quad p\in\mathbb{R}^n,\ |p|\ge1. \tag*{(5)}

The conditions (1.3)–(1.5) are conditions at infinity, since the bound ∣p∣≥1|p|\ge1 for p∈Rnp\in\mathbb{R}^n can be replaced by ∣p∣≥M|p|\ge M for some arbitrary number MM. Inequality (1.3) excludes exponential growth, but the exponent mm may be as large as we want. Condition (1.4) states that the radial part of ∇f(p)\nabla f(p), i.e. the quantity ∇f(p)⋅p∣p∣\nabla f(p)\cdot\frac{p}{|p|}, is not allowed to degenerate at infinity in the sense that we have ∣p∣−μ∣∇f(p)∣|p|^{-\mu}|\nabla f(p)| as a lower bound for some exponent μ∈[0,1)\mu\in[0,1). Finally, the growth of ∣D2f(p)∣|D^2f(p)| as p→∞p\to\infty is limited in terms of ∣∇f(p)∣|\nabla f(p)| equipped with the weight function ∣p∣−μ|p|^{-\mu}, which is made precise in (1.5). We refer to § 2 for a list of prominent examples.

The conditions (1.3)–(1.5) are completed by

∃ρ≥2:c3∣p∣ρ−2∣q∣2≤D2f(p)(q,q)≤c4∣p∣ρ−2∣q∣2,p,q∈Rn, ∣p∣≤1(6)\exists\rho\ge2:\quad c_3|p|^{\rho-2}|q|^2\le D^2f(p)(q,q)\le c_4|p|^{\rho-2}|q|^2,\quad p,q\in\mathbb{R}^n,\ |p|\le1 \tag*{(6)}

and we can state

Theorem 1.1. Let (1.2)–(1.6) hold and let u∈C2(Rn)u\in C^2(\mathbb{R}^n) denote an entire solution of equation (1.1) for which

lim⁡x→∞∣u(x)∣∣x∣β<∞(7)\lim_{x\to\infty}\frac{|u(x)|}{|x|^\beta}<\infty \tag*{(7)}

with

β:=β(ρ):=1ρ(ρ+2−n).(8)\beta:=\beta(\rho):=\frac{1}{\rho}(\rho+2-n). \tag*{(8)}

Then uu is an affine function, i.e. ∂i∂ju=0\partial_i\partial_j u=0 on Rn\mathbb{R}^n for $i,j=1,\ldots,n.

If n=2n=2, then β=1\beta=1 and in conclusion:

Corollary 1.2. Under the assumptions of Theorem 1.1 consider the case n=2n=2 and let u∈C2(R2)u\in C^2(\mathbb{R}^2) denote an entire solution of (1.1) such that

lim⁡x→∞∣u(x)∣∣x∣<∞.\lim_{x\to\infty}\frac{|u(x)|}{|x|}<\infty.

Then uu is an affine function.

Note that the number β=β(ρ)\beta=\beta(\rho) from (8) is ≥0\ge0 if and only if n≤ρ+2n\le\rho+2. This implies the Liouville property for dimensions n≤n(ρ)n\le n(\rho) with upper bound n(ρ):=2+ρn(\rho):=2+\rho determined just through the behavior of D2f(p)D^2f(p) for pp close to the origin as expressed in (6).

Corollary 1.3 (Liouville property). Let (2)–(6) hold and assume that n≤ρ+2n\le\rho+2. Then the constant functions are the only bounded entire solutions of equation (1).

In fact, if u∈L∞(Rn)u\in L^\infty(\mathbb{R}^n) and n≤ρ+2n\le\rho+2, then (7) clearly holds, hence uu is an affine function and thereby constant.

Remark 1.4. It is possible to replace (6) by the anisotropic condition

c5∣p∣t−2∣q∣2≤D2f(p)(q,q)≤c6∣p∣ρ−2∣q∣2,p,q∈Rn, ∣p∣≤1,c_5|p|^{t-2}|q|^2\le D^2f(p)(q,q)\le c_6|p|^{\rho-2}|q|^2,\qquad p,q\in\mathbb{R}^n,\ |p|\le1,

with exponents 2≤ρ≤t2\le\rho\le t. Then Theorem 1.1 continues to hold, provided we define in (8)

β:=β(t,ρ):=1T(2+T−n),T:=tt−ρ+1\beta:=\beta(t,\rho):=\frac{1}{T}(2+T-n),\qquad T:=\frac{t}{t-\rho+1}

as the appropriate value of β\beta, and if we require ρ+2≥t\rho+2\ge t. Some comments will be given in § 4.

Remark 1.5. One might wonder about the occurrence of the quantity ln⁡2(1+∣p∣)\ln^2(1+|p|) on the left-hand side of inequality (5). Of course we can replace (5) by the stronger condition (keeping (4)) that

∣D2f(p)∣≤c2∣p∣−(μ+ε)∣∇f(p)∣,p∈Rn, ∣p∣≥1|D^2f(p)|\le c_2|p|^{-(\mu+\varepsilon)}|\nabla f(p)|,\qquad p\in\mathbb{R}^n,\ |p|\ge1

holds with exponent μ\mu from (4) and for some arbitrarily small number ε>0\varepsilon>0. However, we prefer to work with a less restrictive assumption, and at the same time it turns out that the expression ∣D2f(∇u)∣ln⁡2(1+∣∇u∣)|D^2f(\nabla u)|\ln^2(1+|\nabla u|) occurs in a limit version of a Caccioppoli type inequality, which will be presented in Lemma 3.1 being valid for solutions uu of equation (1) under very general hypotheses concerning the density ff.

Our paper is organized as follows: in § 2 we collect some examples to which Theorem 1.1 and the corollaries apply. § 3 is devoted to an inequality of Caccioppoli type, which is crucial for the proof of Theorem 1.1 presented in § 4. The Caccioppoli inequality Lemma 3.1 will explain the occurrence of the factor ln⁡2(1+∣p∣)\ln^2(1+|p|) on the left-hand side of (5) and during the proof of Lemma 3.1 the comments from Remark 1.5 will become more transparent.

Examples

We present some examples of densities f:Rn→Rf:\mathbb{R}^n\to\mathbb{R} to which Theorem 1.1 can be applied.

Example 2.1 (anisotropic power growth). Let f∈C2(Rn)f \in C^2(\mathbb{R}^n) satisfy (2) and (6). In addition we require: there exist constants c1c_1, c2c_2 and exponents 1<τ≤r1 < \tau\le r such that

c1∣p∣τ−2∣q∣2≤D2f(p)(q,q)≤c2∣p∣r−2∣q∣2(9)c_1|p|^{\tau-2}|q|^2 \le D^2f(p)(q,q) \le c_2|p|^{r-2}|q|^2 \tag*{(9)}

holds for p,q∈Rnp,q \in\mathbb{R}^n, ∣p∣≥1|p| \ge1. Then we claim: if

r<τ+12,(10)r < \tau+ \frac{1}{2}, \tag*{(10)}

then we have (3), (4), (5) for some μ∈[0,1)\mu\in[0,1) and with arbitrary m≥rm \ge r.

The validity of (3) for m≥rm \ge r is immediate by the second inequality in (9). For (4) and (5) we choose p∈Rnp \in\mathbb{R}^n, ∣p∣≥1|p| \ge1, and observe that by (9)

∣∇f(p)∣∣p∣≥∇f(p)⋅p=∫01D2f(tp)(p,p) dt≥c1∫1/21tτ−2 dt∣p∣τ\begin{aligned} |\nabla f(p)||p| \ge\nabla f(p) \cdot p = \int_0^1 D^2f(tp)(p,p)\,dt \\ &\ge c_1\int_{1/2}^1 t^{\tau-2}\,dt|p|^\tau \end{aligned}

for ∣p∣≥2|p| \ge2, whereas in the case 1≤∣p∣≤21 \le|p| \le2 we use (6) to get

{∇f(p)⋅p≥c∣p∣τ,∣∇f(p)∣≥c∣p∣τ−1,p∈Rn, ∣p∣≥1,(11)\begin{cases} \nabla f(p) \cdot p \ge c|p|^\tau,\quad|\nabla f(p)| \ge c|p|^{\tau-1},\\ p \in\mathbb{R}^n,\ |p| \ge1, \tag*{(11)} \end{cases}

with a positive constant cc. The second part of (9) implies with the help of (11)

∣D2f(p)∣ln⁡2(1+∣p∣)≤cln⁡2(1+∣p∣)∣p∣r−2=cln⁡2(1+∣p∣)∣p∣τ−1∣p∣r−τ−1≤cln⁡2(1+∣p∣)∣p∣r−τ−1∣∇f(p)∣≤c∣p∣−μ1∣∇f(p)∣\begin{aligned} |D^2f(p)|\ln^2(1+|p|) \le c\ln^2(1+|p|)|p|^{r-2} \\ &= c\ln^2(1+|p|)|p|^{\tau-1}|p|^{r-\tau-1} \\ &\le c\ln^2(1+|p|)|p|^{r-\tau-1}|\nabla f(p)| \\ &\le c|p|^{-\mu_1}|\nabla f(p)| \end{aligned}

for any μ1∈[0,τ+1−r)\mu_1 \in[0,\tau+1-r).

Moreover, again for ∣p∣≥1|p| \ge1, quoting (6) and (9)

∣∇f(p)∣=∣∫01ddt∇f(tp) dt∣≤∫01∣D2f(tp)∣ dt∣p∣=∣p∣(∫01/∣p∣∣D2f(tp)∣ dt+∫1/∣p∣1∣D2f(tp)∣ dt)≤c∣p∣(1∣p∣+∣p∣r−2)\begin{aligned} |\nabla f(p)| = \left|\int_0^1 \frac{d}{dt}\nabla f(tp)\,dt\right| \le\int_0^1 |D^2f(tp)|\,dt|p| \\ &= |p|\left(\int_0^{1/|p|}|D^2f(tp)|\,dt+\int_{1/|p|}^1|D^2f(tp)|\,dt\right) \\ &\le c|p|\left(\frac{1}{|p|}+|p|^{r-2}\right) \end{aligned}

so that ∣∇f(p)∣≤c∣p∣r−1|\nabla f(p)| \le c|p|^{r-1} and in conclusion (see (11))

p⋅∇f(p)≥c∣p∣τ=c∣p∣r−1∣p∣τ−r+1≥c∣p∣τ−r∣∇f(p)∣∣p∣≥c∣p∣−μ2∣∇f(p)∣∣p∣\begin{aligned} p\cdot\nabla f(p) \ge c|p|^\tau= c|p|^{r-1}|p|^{\tau-r+1} \\ &\ge c|p|^{\tau-r}|\nabla f(p)||p| \ge c|p|^{-\mu_2}|\nabla f(p)||p| \end{aligned}

provided that μ2≥r−τ\mu_2 \ge r-\tau.

In order to find a common value μ\mu for μ1\mu_1 and μ2\mu_2 we need to know that r−τ<τ+1−rr-\tau< \tau+1-r, which is just a consequence of (10) completing the discussion of (4) and (5).

Example 2.2 (densities ff with linear growth). Again we consider f∈C2(Rn)f \in C^2(\mathbb{R}^n) with (2) and (6). We further assume that

a(∣p∣−1)≤f(p)≤A(∣p∣+1)p∈Rn(12)a(|p|-1) \le f(p) \le A(|p|+1) \qquad p \in\mathbb{R}^n \tag*{(12)}

holds with constants a,A>0a,A>0. Moreover, we require

D2f(p)(q,q)≤Θ(∣p∣)∣q∣2p,q∈Rn, ∣p∣≥1,(13)D^2f(p)(q,q) \le\Theta(|p|)|q|^2 \qquad p,q \in\mathbb{R}^n,\ |p| \ge1, \tag*{(13)}

for a function Θ:[0,∞)→[0,∞)\Theta:[0,\infty) \to[0,\infty) such that

lim sup⁡t→∞Θ(t)ln⁡2(t)<∞.(14)\limsup_{t\to\infty} \Theta(t)\ln^2(t) < \infty. \tag*{(14)}

Then we are in the situation of Theorem 1.1, i.e. the conditions (3)–(6) hold: for (3) we just observe that from the convexity of ff (see (2)) together with (12) we immediately get the boundedness of ∇f\nabla f.

Let

ϑ(r)=inf⁡∣p∣=r∣q∣=1D2f(p)(q,q),ϑ:[0,∞)→[0,∞).\vartheta(r)=\inf_{\substack{|p|=r\\|q|=1}}D^2f(p)(q,q), \qquad\vartheta:[0,\infty)\to[0,\infty).

Again from (2) we deduce that ϑ(r)>0\vartheta(r)>0 for r>0r>0. Observing this, it follows for ∣p∣≥1|p|\ge1:

∣p∣∣∇f(p)∣≥p⋅∇f(p)=∫01D2f(tp)(p,p) dt|p||\nabla f(p)| \ge p\cdot\nabla f(p)=\int_0^1D^2f(tp)(p,p)\,\mathrm{d}t
≥∣p∣2∫01ϑ(t∣p∣) dt=∣p∣∫0∣p∣ϑ(s) ds\ge|p|^2\int_0^1\vartheta(t|p|)\,\mathrm{d}t=|p|\int_0^{|p|}\vartheta(s)\,\mathrm{d}s
≥∣p∣∫01ϑ(s) ds=:α∣p∣\ge|p|\int_0^1\vartheta(s)\,\mathrm{d}s=:\alpha|p|

for some positive constant α\alpha, and the boundedness of ∇f\nabla f clearly implies (4) with the choice μ=0\mu=0. Condition (5) (with μ=0\mu=0) is a consequence of (13) and (14).

Example 2.3 (densities with radial structure). We let f(p):=g(∣p∣)f(p):=g(|p|), p∈Rnp\in\mathbb{R}^n for a function g∈C2([0,∞))g\in C^2([0,\infty)) such that

{g′(0)=0,g′′(t)>0for t>0,c1tρ−2≤g′′(t)≤c2tρ−2,0≤t≤1,(15)\left\{ \begin{aligned} g'(0)&=0,\\ g''(t)&>0 \quad\text{for }t>0 \end{aligned} \right. ,\qquad c_1t^{\rho-2}\le g''(t)\le c_2t^{\rho-2},\qquad0\le t\le1, \tag*{(15)}
{g′(t)≤c3tm−1,t≥1,g′′(t)ln⁡2(1+t)≤c4g′(t)for t≥1(16)\left\{ \begin{aligned} g'(t)&\le c_3t^{m-1},\quad t\ge1,\\ g''(t)\ln^2(1+t)&\le c_4g'(t)\quad\text{for }t\ge1 \end{aligned} \right. \tag*{(16)}

with constants c1,…,c4>0c_1,\ldots,c_4 > 0 and exponents m>1m > 1, ρ≥2\rho\ge2. We claim that ff satisfies the hypotheses of Theorem 1.1: noting that

∇f(p)=g′(∣p∣)p∣p∣,∣∇f(p)∣=g′(∣p∣),\nabla f(p)=g'(|p|)\frac{p}{|p|},\qquad|\nabla f(p)|=g'(|p|),
D2f(p)(q,q)=g′(∣p∣)∣p∣[∣q∣2−(p∣p∣⋅q)2]+g′′(∣p∣)(p∣p∣⋅q)2D^2f(p)(q,q)=\frac{g'(|p|)}{|p|}\left[|q|^2-\left(\frac{p}{|p|}\cdot q\right)^2\right]+g''(|p|)\left(\frac{p}{|p|}\cdot q\right)^2

the validity of (1.2) and (1.6) is an immediate consequence of (2.7), whereas (1.3) follows from (2.8).

Let p∈Rnp\in\mathbb{R}^n, ∣p∣≥1|p|\ge1. Then we get (again quoting (2.8)):

∣D2f(p)∣ln⁡2(1+∣p∣)≤max⁡{g′(∣p∣)∣p∣,g′′(∣p∣)}ln⁡2(1+∣p∣)≤cmax⁡{g′(∣p∣)∣p∣ln⁡2(1+∣p∣),g′(∣p∣)}≤cg′(∣p∣)=c∣∇f(p)∣,\begin{aligned} |D^2f(p)|\ln^2(1+|p|)\le\max\left\{\frac{g'(|p|)}{|p|},g''(|p|)\right\}\ln^2(1+|p|) \\ &\le c\max\left\{\frac{g'(|p|)}{|p|}\ln^2(1+|p|),g'(|p|)\right\} \\ &\le cg'(|p|)=c|\nabla f(p)|, \end{aligned}

hence (1.5) holds for the choice μ=0\mu=0. From p⋅∇f(p)=∣p∣∣∇f(p)∣p\cdot\nabla f(p)=|p||\nabla f(p)| we obtain (1.4).

A Caccioppoli type inequality

A preliminary version of the following lemma can be found in Proposition 6.1 of [8]. Here we extend Proposition 6.1 in some sense to the limit case α=−1/2\alpha=-1/2.

Lemma 3.1. Let f∈C2(Rn)f\in C^2(\mathbb{R}^n) denote a convex function, which means that D2f(p)D^2f(p) is just positive semidefinite for any p∈Rnp\in\mathbb{R}^n.

We define λ:[1,∞)→[0,∞)\lambda:[1,\infty)\to[0,\infty), Λ:[1,∞)→[0,∞)\Lambda:[1,\infty)\to[0,\infty) by

λ(t):={t−1/2,t≥2,12(t−1),1≤t≤2,Λ(t):={t−1/2(t−1)ln⁡2(t),t≥2,12ln⁡2(t),1≤t≤2.(17)\lambda(t):= \begin{cases} t^{-1/2}, & t\ge2,\\ \frac{1}{\sqrt{2}}(t-1), & 1\le t\le2, \end{cases} \qquad \Lambda(t):= \begin{cases} t^{-1/2}(t-1)\ln^2(t), & t\ge2,\\ \frac{1}{\sqrt{2}}\ln^2(t), & 1\le t\le2. \end{cases} \tag*{(17)}

Let Ω⊂Rn\Omega\subset\mathbb{R}^n denote a domain and consider a solution u∈C2(Ω)u\in C^2(\Omega) of equation (1.1) now on Ω\Omega.

Then it holds for any η∈C01(Ω)\eta\in C^1_0(\Omega) and for all indices i∈{1,…,n}i\in\{1,\ldots,n\} (Γi:=1+∣∂iu∣2\Gamma_i:=1+|\partial_i u|^2):

∫ΩD2f(∇u)(∇∂iu,∇∂iu)η2λ(Γi) dx≤c(∫spt⁡(∇η)D2f(∇u)(∇∂iu,∇∂iu)η2λ(Γi) dx)1/2(∫spt⁡(∇η)D2f(∇u)(∇η,∇η)Λ(Γi) dx)1/2,(18)\int_{\Omega}D^2f(\nabla u)(\nabla\partial_i u,\nabla\partial_i u)\eta^2\lambda(\Gamma_i)\,dx \le c\left(\int_{\operatorname{spt}(\nabla\eta)}D^2f(\nabla u)(\nabla\partial_i u,\nabla\partial_i u)\eta^2\lambda(\Gamma_i)\,dx\right)^{1/2} \left(\int_{\operatorname{spt}(\nabla\eta)}D^2f(\nabla u)(\nabla\eta,\nabla\eta)\Lambda(\Gamma_i)\,dx\right)^{1/2}, \tag*{(18)}

cc denoting a constant independent of η\eta, Ω\Omega, uu, and in conclusion

∫ΩD2f(∇u)(∇∂iu,∇∂iu)η2λ(Γi) dx≤c∫ΩD2f(∇u)(∇η,∇η)Λ(Γi) dx.(19)\int_{\Omega} D^{2}f(\nabla u)(\nabla\partial_i u,\nabla\partial_i u)\eta^{2}\lambda(\Gamma_i)\,dx \le c\int_{\Omega}D^{2}f(\nabla u)(\nabla\eta,\nabla\eta)\Lambda(\Gamma_i)\,dx. \tag*{(19)}

Moreover, on the right-hand side of (19) we may replace Λ(Γi)\Lambda(\Gamma_i) by Λ(Γ)\Lambda(\Gamma), where Γ:=1+∣∇u∣2\Gamma:= 1+|\nabla u|^{2}.

Proof. The last claim follows from the monotonicity of Λ\Lambda. Now we fix η∈C01(Ω)\eta\in C_{0}^{1}(\Omega) and i∈{1,…,n}i\in\{1,\ldots,n\}. Let

Φ:[1,∞)→[0,∞),Φ(t):=t−1/2ln⁡t,\Phi:[1,\infty)\to[0,\infty),\qquad\Phi(t):=t^{-1/2}\ln t,

and observe that Ψ:=η2∂iuΦ(Γi)\Psi:=\eta^{2}\partial_i u\Phi(\Gamma_i) is admissible in the equation

∫ΩD2f(∇u)(∇∂iu,∇Ψ) dx=0.(20)\int_{\Omega}D^{2}f(\nabla u)(\nabla\partial_i u,\nabla\Psi)\,dx=0. \tag*{(20)}

Here equation (20) is a direct consequence of the weak form of (1.1). From (3.4) we obtain

∫ΩD2f(∇u)(∇∂iu,∇∂iu)η2[Φ(Γi)+2∣∂iu∣2Φ′(Γi)] dx=−2∫spt⁡(∇η)D2f(∇u)(∇∂iu,∇η)∂iuηΦ(Γi) dx\int_{\Omega}D^{2}f(\nabla u)(\nabla\partial_i u,\nabla\partial_i u)\eta^{2}\left[\Phi(\Gamma_i)+2|\partial_i u|^{2}\Phi'(\Gamma_i)\right]\,dx = -2\int_{\operatorname{spt}(\nabla\eta)}D^{2}f(\nabla u)(\nabla\partial_i u,\nabla\eta)\partial_i u\eta\Phi(\Gamma_i)\,dx

or equivalently

∫ΩD2f(∇u)(∇∂iu,∇∂iu)η2[Φ(Γi)+2(Γi−1)Φ′(Γi)] dx=−2∫spt⁡(∇η)D2f(∇u)(∇∂iu,∇η)∂iuηΦ(Γi) dx.(21)\int_{\Omega}D^{2}f(\nabla u)(\nabla\partial_i u,\nabla\partial_i u)\eta^{2}\left[\Phi(\Gamma_i)+2(\Gamma_i-1)\Phi'(\Gamma_i)\right]\,dx = -2\int_{\operatorname{spt}(\nabla\eta)}D^{2}f(\nabla u)(\nabla\partial_i u,\nabla\eta)\partial_i u\eta\Phi(\Gamma_i)\,dx. \tag*{(21)}

Letting t:=Γit:=\Gamma_i (≥1\ge1) for the moment, we get for the expression […][\ldots] in (21), using the definition of Φ\Phi,

[…]=2(t−1/2−t−3/2)+t−3/2ln⁡t.[\ldots]=2(t^{-1/2}-t^{-3/2})+t^{-3/2}\ln t.

This implies

Φ(t)+2(t−1)Φ′(t)≥{t−1/2,t≥2,12(t−1),1≤t≤2(22)\Phi(t)+2(t-1)\Phi'(t)\ge \begin{cases} t^{-1/2}, & t\ge2,\\ \frac{1}{\sqrt{2}}(t-1), & 1\le t\le2 \end{cases} \tag*{(22)}

as a lower bound for the term on the left-hand side of (3.5). On the right-hand side we decompose

−2∫spt⁡(∇η)⋯=−2∫spt⁡(∇η)∩[Γi<2]⋯−2∫spt⁡(∇η)∩[Γi≥2]⋯=:(I)+(II)-2\int_{\operatorname{spt}(\nabla\eta)}\cdots = -2\int_{\operatorname{spt}(\nabla\eta)\cap[\Gamma_i<2]}\cdots -2\int_{\operatorname{spt}(\nabla\eta)\cap[\Gamma_i\ge2]}\cdots =:(I)+(II)

discussing the terms (I) and (II) separately: it holds

∣integrand of (II)∣=2∣∂iu∣∣η∣Φ(Γi)∣D2f(∇u)(∇∂iu,∇η)∣=2(Γi−1)1/2Γi−1/2ln⁡(Γi)∣η∣∣D2f(∇u)(∇∂iu,∇η)∣=2∣D2f(∇u)(∣η∣Γi−1/4∇∂iu,Γi−1/4(Γi−1)1/2ln⁡(Γi)∇η)∣≤2(D2f(∇u)(∇∂iu,∇∂iu)Γi−1/2η2)1/2\begin{aligned} |\text{integrand of }(II)| = 2|\partial_i u||\eta|\Phi(\Gamma_i)|D^{2}f(\nabla u)(\nabla\partial_i u,\nabla\eta)| \\ &= 2(\Gamma_i-1)^{1/2}\Gamma_i^{-1/2}\ln(\Gamma_i)|\eta||D^{2}f(\nabla u)(\nabla\partial_i u,\nabla\eta)| \\ &= 2\left|D^{2}f(\nabla u)\left(|\eta|\Gamma_i^{-1/4}\nabla\partial_i u,\Gamma_i^{-1/4}(\Gamma_i-1)^{1/2}\ln(\Gamma_i)\nabla\eta\right)\right| \\ &\le 2\left(D^{2}f(\nabla u)(\nabla\partial_i u,\nabla\partial_i u)\Gamma_i^{-1/2}\eta^{2}\right)^{1/2} \end{aligned}
×(D2f(∇u)(∇η,∇η)Γi−1/2(Γi−1)ln⁡2(Γi))1/2\qquad\times\left(D^{2}f(\nabla u)(\nabla\eta,\nabla\eta)\Gamma_i^{-1/2}(\Gamma_i-1)\ln^{2}(\Gamma_i)\right)^{1/2}

where in the last step the Cauchy-Schwarz inequality has been used for the nonnegative symmetric bilinear form D2f(∇u)D^2f(\nabla u). Thus, after using Hölder’s inequality, we find (recall (3.1))

∣(II)∣≤(∫spt⁡(∇η)∩[Γi≥2]D2f(∇u)(∇∂iu,∇∂iu)Γi−1/2η2 dx)1/2×(∫spt⁡(∇η)∩[Γi≥2]D2f(∇u)(∇η,∇η)Γi−1/2(Γi−1)ln⁡2(Γi) dx)1/2≤(r.h.s. of (3.2))(23)\begin{aligned} |(II)| &\le\left(\int_{\operatorname{spt}(\nabla\eta)\cap[\Gamma_i\ge2]} D^2f(\nabla u)(\nabla\partial_i u,\nabla\partial_i u)\Gamma_i^{-1/2}\eta^2\,dx\right)^{1/2}\\ &\quad\times\left(\int_{\operatorname{spt}(\nabla\eta)\cap[\Gamma_i\ge2]}D^2f(\nabla u)(\nabla\eta,\nabla\eta)\Gamma_i^{-1/2}(\Gamma_i-1)\ln^2(\Gamma_i)\,dx\right)^{1/2}\\ &\le\text{(r.h.s. of (3.2))} \tag*{(23)} \end{aligned}

In a similar way we proceed with (I): we have

∣integrand of (I)∣=2∣η∂iuΦ(Γi)∣∣D2f(∇u)(∇∂iu,∇η)∣≤c∣η(Γi−1)1/2ln⁡(Γi)∣∣D2f(∇u)(∇∂iu,∇η)∣≤c(D2f(∇u)(∇∂iu,∇∂iu)η2(Γi−1))1/2×(D2f(∇u)(∇η,∇η)ln⁡2(Γi))1/2\begin{aligned} |\text{integrand of }(I)|&=2|\eta\partial_i u\Phi(\Gamma_i)||D^2f(\nabla u)(\nabla\partial_i u,\nabla\eta)|\\ &\le c|\eta(\Gamma_i-1)^{1/2}\ln(\Gamma_i)||D^2f(\nabla u)(\nabla\partial_i u,\nabla\eta)|\\ &\le c\left(D^2f(\nabla u)(\nabla\partial_i u,\nabla\partial_i u)\eta^2(\Gamma_i-1)\right)^{1/2}\\ &\quad\times\left(D^2f(\nabla u)(\nabla\eta,\nabla\eta)\ln^2(\Gamma_i)\right)^{1/2} \end{aligned}

and using Hölder’s inequality again (recall (3.1))

∣(I)∣≤c(∫spt⁡(∇η)∩[Γi<2]D2f(∇u)(∇∂iu,∇∂iu)η2(Γi−1) dx)1/2×(∫spt⁡(∇η)∩[Γi<2]D2f(∇u)(∇η,∇η)ln⁡2(Γi) dx)1/2≤(r.h.s. of (3.2))(24)\begin{aligned} |(I)|&\le c\left(\int_{\operatorname{spt}(\nabla\eta)\cap[\Gamma_i<2]}D^2f(\nabla u)(\nabla\partial_i u,\nabla\partial_i u)\eta^2(\Gamma_i-1)\,dx\right)^{1/2}\\ &\quad\times\left(\int_{\operatorname{spt}(\nabla\eta)\cap[\Gamma_i<2]}D^2f(\nabla u)(\nabla\eta,\nabla\eta)\ln^2(\Gamma_i)\,dx\right)^{1/2}\\ &\le\text{(r.h.s. of (3.2))} \tag*{(24)} \end{aligned}

We return to (3.5) noting that, on account of (3.6) it holds (recall (3.1))

(l.h.s. of (3.5))≥c∫ΩD2f(∇u)(∇∂iu,∇∂iu)η2λ(Γi) dx.(25)\text{(l.h.s. of (3.5))}\ge c\int_{\Omega}D^2f(\nabla u)(\nabla\partial_i u,\nabla\partial_i u)\eta^2\lambda(\Gamma_i)\,dx. \tag*{(25)}

Thus, combining (3.5) with (3.7)–(3.9), inequality (3.2) follows, which completes the proof of Lemma 3.1.

Proof of Theorem 1.1

We want to prove Theorem 1.1 in the setting of Remark 1.4, which means that we assume the conditions (1.2)–(1.5) for the density ff but replace (1.6) by (p,q∈Rn, ∣p∣≤1)(p,q\in\mathbb{R}^n,\ |p|\le1)

c∣p∣t−2∣q∣2≤D2f(p)(q,q)≤C∣p∣ρ−2∣q∣2(26)c|p|^{t-2}|q|^2\le D^2f(p)(q,q)\le C|p|^{\rho-2}|q|^2 \tag*{(26)}

for exponents 2≤ρ≤t2\le\rho\le t, t≤ρ+2t\le\rho+2 and constants c,C>0c,C>0.

Our arguments proceed as follows: let i∈{1,…,n}i \in\{1,\ldots,n\} and consider η∈C01(Rn)\eta\in C^1_0(\mathbb{R}^n) such that spt⁡(η)⊂B2R‾\operatorname{spt}(\eta) \subset\overline{B_{2R}}, 0≤η≤10 \leq\eta\leq1, η=1\eta= 1 on BRB_R and ∣∇η∣≤c/R|\nabla\eta| \leq c/R, where Br:={x∈Rn:∣x∣<r}B_r := \{x \in\mathbb{R}^n : |x| < r\} for radii r>0r > 0. We will show that we can find a constant c∗∈(0,∞)c^* \in(0,\infty) for which

R−2∫B2R−BR∣D2f(∇u)∣Λ(Γ) dx≤c∗,0<R<∞,(27)R^{-2}\int_{B_{2R}-B_R}|D^2f(\nabla u)|\Lambda(\Gamma)\,dx \leq c^*, \qquad0<R<\infty, \tag*{(27)}

holds with Λ\Lambda defined in (17) and Γ:=1+∣∇u∣2\Gamma:= 1+|\nabla u|^2.

Accepting (27) for the moment, we quote (19) (recalling Λ(Γi)≤Λ(Γ)\Lambda(\Gamma_i) \leq\Lambda(\Gamma), Γi:=1+∣∂iu∣2\Gamma_i := 1+|\partial_i u|^2) and get

∫RnD2f(∇u)(∇∂iu,∇∂iu)λ(Γi) dx<∞.\int_{\mathbb{R}^n}D^2f(\nabla u)(\nabla\partial_i u,\nabla\partial_i u)\lambda(\Gamma_i)\,dx<\infty.

Finally, we use inequality (18) with

∫BRD2f(∇u)(∇∂iu,∇∂iu)λ(Γi) dx=:Θi(R)(28)\int_{B_R}D^2f(\nabla u)(\nabla\partial_i u,\nabla\partial_i u)\lambda(\Gamma_i)\,dx=:\Theta_i(R) \tag*{(28)}

as a lower bound for the left-hand side of (18), whereas the right-hand side is bounded from above by the quantity (c=c= a constant independent of RR)

c(∫B2R−BRD2f(∇u)(∇∂iu,∇∂iu)λ(Γi) dx)1/2⋅(R−2∫B2R−BR∣D2f(∇u)∣Λ(Γ) dx)1/2c\left(\int_{B_{2R}-B_R}D^2f(\nabla u)(\nabla\partial_i u,\nabla\partial_i u)\lambda(\Gamma_i)\,dx\right)^{1/2}\cdot\left(R^{-2}\int_{B_{2R}-B_R}|D^2f(\nabla u)|\Lambda(\Gamma)\,dx\right)^{1/2}

which goes to zero as R→∞R\to\infty on account of (27) and (28).

Thus we have lim⁡R→∞Θi(R)=0\lim_{R\to\infty}\Theta_i(R)=0 and in conclusion

D2f(∇u)(∇∂iu,∇∂iu)λ(Γi)=0on Rn(29)D^2f(\nabla u)(\nabla\partial_i u,\nabla\partial_i u)\lambda(\Gamma_i)=0 \quad\text{on }\mathbb{R}^n \tag*{(29)}

for i∈{1,…,n}i \in\{1,\ldots,n\}. The definition of λ\lambda (see (17)) shows that λ(Γi(x))=0\lambda(\Gamma_i(x))=0 if and only if ∂iu(x)=0\partial_i u(x)=0, and by Lemma 7.7, p. 152, in [6] it holds ∇∂iu=0\nabla\partial_i u=0 almost everywhere on [∂iu=0][\partial_i u=0]. However, on the set [∂iu≠0][\partial_i u\neq0], (29) implies ∇∂iu=0\nabla\partial_i u=0, hence ∇∂iu=0\nabla\partial_i u=0 almost everywhere and by continuity everywhere on Rn\mathbb{R}^n, being true for i=1,…,ni=1,\ldots,n, completing the proof of Theorem 1.1.

Let us discuss our claim (27). On the set [∣∇u∣≥1][|\nabla u|\geq1] it holds by the definition of Λ\Lambda given in (17) and on account of (5)

∣D2f(∇u)∣Λ(Γ)=∣D2f(∇u)∣Γ−1/2(Γ−1)ln⁡2(Γ)≤∣D2f(∇u)∣∣∇u∣ln⁡2(Γ)≤c∣∇u∣1−μ∣∇f(∇u)∣.\begin{aligned} |D^2f(\nabla u)|\Lambda(\Gamma) &=|D^2f(\nabla u)|\Gamma^{-1/2}(\Gamma-1)\ln^2(\Gamma)\\ &\leq|D^2f(\nabla u)||\nabla u|\ln^2(\Gamma)\\ &\leq c|\nabla u|^{1-\mu}|\nabla f(\nabla u)|. \end{aligned}

Note that in (5) we can evidently replace ln⁡(1+∣p∣)\ln(1+|p|) by ln⁡(1+∣p∣2)\ln(1+|p|^2) for p∈Rnp\in\mathbb{R}^n such that ∣p∣≥1|p|\geq1. Using inequality (4), we get

∣D2f(∇u)∣Λ(Γ)≤c∇u⋅∇f(∇u)on [∣∇u∣≥1].(30)|D^2f(\nabla u)|\Lambda(\Gamma)\leq c\nabla u\cdot\nabla f(\nabla u)\quad\text{on }[|\nabla u|\geq1]. \tag*{(30)}

On the set [∣∇u∣<1][\lvert\nabla u\rvert< 1] we observe using (17) and (26)

∣D2f(∇u)∣Λ(Γ)≤c∣D2f(∇u)∣ln⁡2(Γ)≤c∣D2f(∇u)∣∣∇u∣4≤c∣∇u∣ρ+2\lvert D^{2}f(\nabla u)\rvert_{\Lambda(\Gamma)} \le c\lvert D^{2}f(\nabla u)\rvert\ln^{2}(\Gamma) \le c\lvert D^{2}f(\nabla u)\rvert\lvert\nabla u\rvert^{4} \le c\lvert\nabla u\rvert^{\rho+2}

since ln⁡(1+α)≤α\ln(1+\alpha)\le\alpha for α≥0\alpha\ge0. At the same time it holds (again for [∣∇u∣<1][\lvert\nabla u\rvert< 1])

∇u⋅∇f(∇u)=∫01D2f(r∇u)(∇u,∇u) dr≥c∫01rt−2 dr∣∇u∣t=c∣∇u∣t,\nabla u\cdot\nabla f(\nabla u)=\int_{0}^{1}D^{2}f(r\nabla u)(\nabla u,\nabla u)\,dr\ge c\int_{0}^{1}r^{t-2}\,dr\lvert\nabla u\rvert^{t}=c\lvert\nabla u\rvert^{t},

and from our assumption t≤ρ+2t\le\rho+2 we find

∣D2f(∇u)∣Λ(Γ)≤c∣∇u∣ton [∣∇u∣<1].(31)\lvert D^{2}f(\nabla u)\rvert_{\Lambda(\Gamma)}\le c\lvert\nabla u\rvert^{t}\quad\text{on }[\lvert\nabla u\rvert<1]. \tag*{(31)}

Let M:Rn→[0,∞)\mathcal{M}:\mathbb{R}^{n}\to[0,\infty),

M(p):={∣∇f(p)∣∣p∣1−μ,∣p∣≥1,∣p∣t,∣p∣<1.(32)\mathcal{M}(p):= \begin{cases} \lvert\nabla f(p)\rvert\lvert p\rvert^{1-\mu}, & \lvert p\rvert\ge1,\\ \lvert p\rvert^{t}, & \lvert p\rvert<1. \end{cases} \tag*{(32)}

Then our calculations leading to (30) and (31) yield

∣D2f(∇u)∣Λ(Γ)≤cM(∇u)≤c∇u⋅∇f(∇u),(33)\lvert D^{2}f(\nabla u)\rvert_{\Lambda(\Gamma)}\le c\mathcal{M}(\nabla u)\le c\nabla u\cdot\nabla f(\nabla u), \tag*{(33)}

and (27) will follow as soon as we can show

R−2∫B2R−BRM(∇u) dx≤c∗(34)R^{-2}\int_{B_{2R}-B_{R}}\mathcal{M}(\nabla u)\,dx\le c^{*} \tag*{(34)}

for a finite constant c∗c^{*} independent of RR.

For proving (34) we select η∈C01(B52R−BR2)\eta\in C_{0}^{1}(B_{\frac{5}{2}R}-B_{\frac{R}{2}}), 0≤η≤10\le\eta\le1 such that η=1\eta=1 on B2R−BRB_{2R}-B_{R}, ∣∇η∣≤c/R\lvert\nabla\eta\rvert\le c/R and let l∈Nl\in\mathbb{N}. Abbreviating TR:=B52R−BR2T_{R}:=B_{\frac{5}{2}R}-B_{\frac{R}{2}} we quote (33) and use equation (1) to get

∫TRη2lM(∇u) dx≤c∫TR∇u⋅∇f(∇u)η2l dx=−2lc∫TR∇f(∇u)⋅∇η η2l−1u dx=−2lc{∫TR∩[∣∇u∣≥1]∇f(∇u)⋅∇η η2l−1u dx+∫TR∩[∣∇u∣<1]⋯ dx}=−2lc((I)+(II)).(35)\begin{aligned} \int_{T_{R}}\eta^{2l}\mathcal{M}(\nabla u)\,dx\le c\int_{T_{R}}\nabla u\cdot\nabla f(\nabla u)\eta^{2l}\,dx \\ &= -2lc\int_{T_{R}}\nabla f(\nabla u)\cdot\nabla\eta\,\eta^{2l-1}u\,dx \\ &= -2lc\left\{\int_{T_{R}\cap[\lvert\nabla u\rvert\ge1]}\nabla f(\nabla u)\cdot\nabla\eta\,\eta^{2l-1}u\,dx+\int_{T_{R}\cap[\lvert\nabla u\rvert<1]}\cdots\,dx\right\} \\ &= -2lc\bigl((I)+(II)\bigr). \tag*{(35)} \end{aligned}

For ϵ,δ>0\epsilon,\delta>0 we have (see (3))

∣integrand of (I)∣≤∣∇f(∇u)∣η2l−1∣∇η∣∣u∣=∣∇f(∇u)∣1−ϵ∣∇f(∇u)∣ϵη2l−1∣∇η∣∣u∣≤c∣∇f(∇u)∣1−ϵ∣∇u∣ϵ(m−1)η2l−1∣∇η∣∣u∣≤δ∣∇f(∇u)∣∣∇u∣(m−1)ϵ1−ϵη2l−11−ϵ+c(δ)∣∇η∣1/ϵ∣u∣1/ϵ,\begin{aligned} \lvert\text{integrand of }(I)\rvert\le\lvert\nabla f(\nabla u)\rvert\eta^{2l-1}\lvert\nabla\eta\rvert\lvert u\rvert\\ &=\lvert\nabla f(\nabla u)\rvert^{1-\epsilon}\lvert\nabla f(\nabla u)\rvert^{\epsilon}\eta^{2l-1}\lvert\nabla\eta\rvert\lvert u\rvert\\ &\le c\lvert\nabla f(\nabla u)\rvert^{1-\epsilon}\lvert\nabla u\rvert^{\epsilon(m-1)}\eta^{2l-1}\lvert\nabla\eta\rvert\lvert u\rvert\\ &\le\delta\lvert\nabla f(\nabla u)\rvert\lvert\nabla u\rvert^{\frac{(m-1)\epsilon}{1-\epsilon}}\eta^{\frac{2l-1}{1-\epsilon}}+c(\delta)\lvert\nabla\eta\rvert^{1/\epsilon}\lvert u\rvert^{1/\epsilon}, \end{aligned}

and if we fix ϵ>0\epsilon>0 such that ϵ1−ϵ(m−1)≤1−μ\frac{\epsilon}{1-\epsilon}(m-1)\leq1-\mu (recall 0≤μ<10\leq\mu<1), we obtain after increasing ll (recall (32)):

∣integrand of (I)∣≤δ∣∇f(∇u)∣∣∇u∣1−μη2l+c(δ)∣∇η∣1/ϵ∣u∣1/ϵ≤δM(∇u)η2l+c(δ)∣∇η∣1/ϵ∣u∣1/ϵ.\lvert\text{integrand of }(I)\rvert\leq\delta\lvert\nabla f(\nabla u)\rvert\lvert\nabla u\rvert^{1-\mu}\eta^{2l}+c(\delta)\lvert\nabla\eta\rvert^{1/\epsilon}\lvert u\rvert^{1/\epsilon} \leq\delta\mathcal{M}(\nabla u)\eta^{2l}+c(\delta)\lvert\nabla\eta\rvert^{1/\epsilon}\lvert u\rvert^{1/\epsilon}.

Returning to (35) and choosing δ\delta sufficiently small, it is shown

∫TRη2lM(∇u) dx≤c{∣(II)∣+∫TR∣∇η∣1/ϵ∣u∣1/ϵ dx}.\int_{T_R}\eta^{2l}\mathcal{M}(\nabla u)\,dx\leq c\left\{\lvert(II)\rvert+\int_{T_R}\lvert\nabla\eta\rvert^{1/\epsilon}\lvert u\rvert^{1/\epsilon}\,dx\right\}.

The remaining term (II)(II) is handled as follows (see (26) and (32)):

∣integrand of (II)∣≤∣∇f(∇u)∣η2l−1∣u∣∣∇η∣≤c∣∇u∣ρ−1η2l−1∣∇η∣∣u∣≤δ∣∇u∣tη(2l−1)tρ−1+c(δ)∣u∣T∣∇η∣T≤δM(∇u)η(2l−1)tρ−1+c(δ)∣u∣T∣∇η∣T.(36)\begin{aligned} \lvert\text{integrand of }(II)\rvert\leq\lvert\nabla f(\nabla u)\rvert\eta^{2l-1}\lvert u\rvert\lvert\nabla\eta\rvert\\ &\leq c\lvert\nabla u\rvert^{\rho-1}\eta^{2l-1}\lvert\nabla\eta\rvert\lvert u\rvert\\ &\leq\delta\lvert\nabla u\rvert^{t}\eta^{(2l-1)\frac{t}{\rho-1}}+c(\delta)\lvert u\rvert^{T}\lvert\nabla\eta\rvert^{T} \\ &\leq\delta\mathcal{M}(\nabla u)\eta^{(2l-1)\frac{t}{\rho-1}}+c(\delta)\lvert u\rvert^{T}\lvert\nabla\eta\rvert^{T}. \tag*{(36)} \end{aligned}

with exponent T:=t/(t−ρ+1)T:=t/(t-\rho+1) as in Remark 1.4.

Finally, for ll sufficiently large, we can arrange that (2l−1)t/(ρ−1)≥2l(2l-1)t/(\rho-1)\geq2l, and after appropriate choice of δ\delta we deduce from (36)

∫TRη2lM(∇u) dx≤c∫TR(∣u∣T∣∇η∣T+∣u∣1/ϵ∣∇η∣1/ϵ) dx.(37)\int_{T_R}\eta^{2l}\mathcal{M}(\nabla u)\,dx\leq c\int_{T_R}\left(\lvert u\rvert^{T}\lvert\nabla\eta\rvert^{T}+\lvert u\rvert^{1/\epsilon}\lvert\nabla\eta\rvert^{1/\epsilon}\right)\,dx. \tag*{(37)}

We recall that ϵ>0\epsilon>0 just has to satisfy ϵ1−ϵ(m−1)≤1−μ\frac{\epsilon}{1-\epsilon}(m-1)\leq1-\mu. Now we require in addition that ϵ≤1/T\epsilon\leq1/T holds and use (1.7) on the right-hand side of (37) with the choice of β=β(t,ρ)\beta=\beta(t,\rho) as defined in Remark 1.4. This gives (recall the choice of η\eta) for R≫1R\gg1:

r.h.s. of (4.12)≤cRn[RβT−T+Rβϵ−1ϵ]≤cRn+βT−T=cR2,\mathrm{r.h.s.\ of\ }(4.12)\leq cR^n\left[R^{\beta T-T}+R^{\frac{\beta}{\epsilon}-\frac{1}{\epsilon}}\right]\leq cR^{n+\beta T-T}=cR^2,

and (4.9) is a consequence of (37). This completes the proof of Theorem 1.1 and of Remark 1.4, respectively.

Revisiting Example 2.1

We finish by remarking that the arguments presented during the proof of Theorem 1.1 are rather general, which means that they can easily be adjusted to particular cases, for instance it is possible to weaken condition (2.2) required in Example 2.1. We can state the following extension:

Theorem 4.1. Let the density f∈C2(Rn)f\in C^2(\mathbb{R}^n) satisfy (1.2), (1.6) and (2.1), where in (2.1) we choose exponents 1<τ≤r1<\tau\leq r such that

r<τ+1(38)r<\tau+1 \tag*{(38)}

holds.

Then, if u∈C2(Rn)u \in C^2(\mathbb{R}^n) is an entire solution of (1.1) with the property (1.7), β\beta defined according to (1.8), it follows that uu is an affine function. Moreover, Corollary 1.2 and Corollary 1.3 extend to this situation.

Proof. We present the necessary adjustments. As before, it suffices to prove inequality (27). On the set [∣∇u∣≥1][|\nabla u| \ge1] we have by (9) (recall also (17))

∣D2f(∇u)∣Λ(Γ)≤∣D2f(∇u)∣∣∇u∣ln⁡2(Γ)≤c∣∇u∣r−2∣∇u∣ln⁡2(Γ)=c∣∇u∣r−1ln⁡2(Γ)≤c∣∇u∣τ(39)\begin{aligned} |D^2 f(\nabla u)|\Lambda(\Gamma) \le|D^2 f(\nabla u)||\nabla u|\ln^2(\Gamma) \\ &\le c|\nabla u|^{r-2}|\nabla u|\ln^2(\Gamma)=c|\nabla u|^{r-1}\ln^2(\Gamma) \\ &\le c|\nabla u|^\tau \tag*{(39)} \end{aligned}

since

∇u⋅∇f(∇u)=∫01D2f(t∇u)(∇u,∇u) dt≥c∣∇u∣τ,\nabla u\cdot\nabla f(\nabla u)=\int_0^1 D^2f(t\nabla u)(\nabla u,\nabla u)\,dt\ge c|\nabla u|^\tau,

whereas (31) (now with t=ρt=\rho, see (6)) continues to hold. We replace (32) by

M~(p):={∣p∣τ,∣p∣≥1,∣p∣ρ,∣p∣<1(40)\widetilde{M}(p):= \begin{cases} |p|^\tau, & |p|\ge1,\\ |p|^\rho, & |p|<1 \end{cases} \tag*{(40)}

and deduce from (39) and (40)

∣D2f(∇u)∣Λ(Γ)≤cM~(∇u)≤c∇u⋅∇f(∇u).(41)|D^2 f(\nabla u)|\Lambda(\Gamma)\le c\widetilde{M}(\nabla u)\le c\nabla u\cdot\nabla f(\nabla u). \tag*{(41)}

From (41) in turn it follows that our claim will be a consequence of (compare (34))

R−2∫B2R−BRM~(∇u) dx≤c∗(42)R^{-2}\int_{B_{2R}-B_R}\widetilde{M}(\nabla u)\,dx\le c^* \tag*{(42)}

again with a finite constant c∗c^* independent of RR. For proving (42) we pass to inequality (35) discussing the quantity (II) exactly as before (recalling that now, for notational simplicity, we just assume that ρ=t\rho=t). In addition we have for the integrand in (I) (compare (9) and (38))

∣∇f(∇u)⋅∇η η2l−1u∣≤∣∇f(∇u)∣η2l−1∣∇η∣∣u∣≤c∣∇u∣r−1η2l−1∣u∣∣∇η∣≤ϵ∣∇u∣τητ−r+1τ−r+1(2l−1)+c(ϵ)∣u∣τ−r+1τ−r+1∣∇η∣τ−r+1τ−r+1\begin{aligned} |\nabla f(\nabla u)\cdot\nabla\eta\,\eta^{2l-1}u|\le|\nabla f(\nabla u)|\eta^{2l-1}|\nabla\eta||u| \\ &\le c|\nabla u|^{r-1}\eta^{2l-1}|u||\nabla\eta| \\ &\le\epsilon|\nabla u|^\tau\eta^{\frac{\tau-r+1}{\tau-r+1}(2l-1)}+c(\epsilon)|u|^{\frac{\tau-r+1}{\tau-r+1}}|\nabla\eta|^{\frac{\tau-r+1}{\tau-r+1}} \end{aligned}

After increasing ll and decreasing ϵ\epsilon we get in place of (37)

∫TRη2lM~(∇u) dx≤c∫TR(∣u∣a∣∇η∣a+∣u∣b∣∇η∣b) dx,a:=ρ,b:=ττ−r+1.(43)\int_{T_R}\eta^{2l}\widetilde{M}(\nabla u)\,dx\le c\int_{T_R}\left(|u|^a|\nabla\eta|^a+|u|^b|\nabla\eta|^b\right)\,dx,\qquad a:=\rho,\quad b:=\frac{\tau}{\tau-r+1}. \tag*{(43)}

Assume now that we have (1.7) with β(≤1)\beta(\le1) from (1.8). Observe that in (2.1) we can increase rr such that rr is very close to τ+1\tau+1. Assuming this from the beginning, it follows that aβ−a≥bβ−ba\beta-a\ge b\beta-b and (4.18) gives

∫TRη2lM~(∇u) dx≤cRn+ρ(β−1)=cR2,\int_{T_R}\eta^{2l}\widetilde{M}(\nabla u)\,dx\le cR^{n+\rho(\beta-1)}=cR^2,

hence we have (4.17) and the proof is complete.

Michael Bildhauer Saarland University Department of Mathematics P.O. Box 15 11 50 66041 Saarbrücken, Germany [email protected]

Martin Fuchs Saarland University Department of Mathematics P.O. Box 15 11 50 66041 Saarbrücken, Germany [email protected]

References

  1. [1]G. Mingione, V. Rădulescu: Recent developments in problems with nonstandard growth and nonuniform ellipticity. J. Math. Anal. Appl. 501 (2021), no. 1, Paper No. 125197, 41 pp.DOI
  2. [2]L. D’Ambrosio: Liouville theorems for anisotropic quasilinear inequalities. Nonlinear Anal. 70 (2009), no. 8, 2855–2869.DOI
  3. [3]A. Farina: Liouville-type theorems for elliptic problems. Handbook of differential equations: stationary partial differential equations. Vol. IV, 61–116. Elsevier/North-Holland, Amsterdam, 2007.DOI
  4. [4]T. Adamowicz, P. Górka: The Liouville theorems for elliptic equations with nonstandard growth. Commun. Pure Appl. Anal. 14 (2015), no. 6, 2377–2392.arxiv.org/abs/1408.6359
  5. [5]S. Dudek: The Liouville-type theorem for problems with nonstandard growth derived by Caccioppoli-type estimate. Monatsh. Math. 192 (2020), no. 1, 75–91.DOI
  6. [6]D. Gilbarg, N. S. Trudinger: Elliptic partial differential equations of second order. Reprint of the 1998 edition. Classics Math. Springer-Verlag, Berlin, 2001. xiv+517 pp.
  7. [7]O. A. Ladyzhenskaya, N. N. Ural’tseva: Linear and quasilinear elliptic equations. Academic Press, New York-London, 1968. xviii+495 pp.
  8. [8]M. Bildhauer, M. Fuchs: On the global regularity for minimizers of variational integrals: splitting-type problems in 2D and extensions to the general anisotropic setting. J. Elliptic Parabol. Equ. 8 (2022), no. 2, 853–884.

Paper details

vlnc:W_3tcnmwcn
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

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