Abstract
We prove the existence of weak solutions for a general class of Dirichlet anisotropic elliptic problems of the form
on a bounded open subset \(\Omega \subset {\mathbb {R}}^N\) \((N\ge 2)\), where \(f\in L^1(\Omega )\) is arbitrary. Our models are \( \mathcal Au=-\sum _{j=1}^N \partial _j (|\partial _j u|^{p_j-2}\partial _j u)\) and \(\Phi (u,\nabla u)=\left( 1+\sum _{j=1}^N {\mathfrak {a}}_j |\partial _j u|^{p_j}\right) |u|^{m-2}u\), with \(m,p_j>1\),\({\mathfrak {a}}_j\ge 0\) for \(1\le j\le N\) and \(\sum _{k=1}^N (1/p_k)>1\). The main novelty is the inclusion of a possibly singular gradient-dependent term \(\Psi (u,\nabla u)=\sum _{j=1}^N |u|^{\theta _j-2}u\, |\partial _j u|^{q_j}\), where \(\theta _j>0\) and \(0\le q_j<p_j\) for \(1\le j\le N\). Under suitable conditions, we prove the existence of solutions by distinguishing two cases: 1) for every \(1\le j\le N\), we have \(\theta _j> 1\) and 2) there exists \(1\le j\le N\) such that \(\theta _j\le 1\). In the latter situation, assuming that \(f \ge 0\) a.e. in \(\Omega \), we obtain non-negative solutions for our problem.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction and main results
Motivated by our study initiated in [16], in this paper we obtain existence of solutions for general singular anisotropic elliptic equations in a bounded, open subset \(\Omega \subset {\mathbb {R}}^N\) (\(N\ge 2\)), subject to a homogeneous Dirichlet boundary condition, \(u=0\) on \(\partial \Omega \). We impose no smoothness assumptions on the boundary of \(\Omega \). The equations under consideration feature a low summability data \(f\in L^1(\Omega )\), a lower-order term \(\Phi (x,u,\nabla u)\) satisfying a “good sign” condition, an “anisotropic natural growth” in the gradient and no upper bound restriction in |u| (see (1.13) and (1.14)). The novelty of our work here consists in the introduction of a possibly singular gradient-dependent term \(\Psi (u,\nabla u)\) (as in (1.3)) which cannot be incorporated in \(\Phi \) and cannot be treated with the arguments in [16]. The main contribution in this paper is to show that, under suitable assumptions, our problem (1.11) admits solutions u in the anisotropic Sobolev space \(W_0^{1,\overrightarrow{p}}(\Omega )\) such that \(\Phi (x,u,\nabla u)\in L^1(\Omega )\).
Let \(W_0^{1,\overrightarrow{p}}(\Omega )\) be the closure of \(C_c^\infty (\Omega )\) (the set of smooth functions with compact support in \(\Omega \)) with respect to the norm
where we assume that
Here, \(p:=N/\sum _{j=1}^N (1/p_j)\) is the harmonic mean of \(p_1,\ldots , p_N\). We write \(\partial _j u\) for the partial derivative \(\partial u/\partial x_j\). We use \(W^{-1,\overrightarrow{p}'}(\Omega )\) for the dual of \(W_0^{1,\overrightarrow{p}}(\Omega )\) and \(\langle \cdot , \cdot \rangle \) for the duality between \(W^{-1,\overrightarrow{p}'}(\Omega )\) and \(W_0^{1,\overrightarrow{p}}(\Omega )\). Since \(p<N\), the embedding \(W_0^{1,\overrightarrow{p}}(\Omega )\hookrightarrow L^s(\Omega )\) is continuous for every \(s\in [1,p^*]\) and compact for every \(s\in [1,p^*)\), where \(p^*:=Np/(N-p)\) stands for the anisotropic Sobolev exponent (see Remark A.3 in the “Appendix”).
Before introducing our general problem in Sect. 1.2 and the main results associated with it (Theorems 1.4 and 1.5), we present a model. For every \((t,\xi )\in {\mathbb {R}}\times {\mathbb {R}}^N\), we define
where \(m>1\), \( {\mathfrak {a}}_0>0\), \( {\mathfrak {a}}_j\ge 0\) for \( 1\le j\le N\), whereas
with \(\theta _j>0\) and \(0\le q_j<p_j\) for all \(1\le j\le N\).
Let \(h\in W^{-1,\overrightarrow{p}'}(\Omega )\) and \(f\in L^1(\Omega )\) be arbitrary. The model for our problem is as follows:
Regarding \(\{\theta _j\}_{1\le j\le N}\), we distinguish two cases:
Case 1: (Non-singular) For every \(1\le j\le N\), we have \(\theta _j>1\).
Case 2: (Mildly singular) We have \(\theta _j\le 1\) for some \(1\le j\le N\). In this case, we will impose some restrictions, such as \(h=0\) and \(f\ge 0\) a.e. in \(\Omega \), to obtain non-negative solutions of (1.4).
The strongly singular case when \(\theta _j\le 0\) for some \(1 \le j \le N\) requires different ideas and techniques and will be considered elsewhere.
To give the notion of solution of (1.4), for \(v\in W_0^{1,\overrightarrow{p}}(\Omega )\cap L^\infty (\Omega )\) and \(U_0\in W_0^{1,\overrightarrow{p}}(\Omega )\), we define
By a solution of (1.4) we mean a function \(U_0\in W_0^{1,\overrightarrow{p}}(\Omega ) \), which is non-negative in Case 2, such that \(\Phi _0(U_0,\nabla U_0) \in L^1(\Omega )\) and for every \(v\in W_0^{1,\overrightarrow{p}}(\Omega )\cap L^\infty (\Omega )\), we have \(I_{U_0}(v)\in {\mathbb {R}}\) and
We leverage \(\Phi _0\) to get the existence of solutions of (1.4) for every \(f\in L^1(\Omega )\). This is reflected in a (lower bound) condition on \(m>1\). To be precise, we define
When either \(N_{\overrightarrow{{\mathfrak {a}}}}\) or \(P_{\overrightarrow{{\mathfrak {a}}}}\) is non-empty, we need \(m>1\) to satisfy
We first illustrate our main results for the model problem in (1.4).
Theorem 1.1
Let (1.1)–(1.3) and (1.8) hold. Let \(h\in W^{-1,\overrightarrow{p}'}(\Omega )\) and \(f\in L^1(\Omega )\) be arbitrary. When \(f\not =0\), we assume that \(\min _{1\le j\le N} {\mathfrak {a}}_j>0\). Assume Case 1 or Case 2 and, in the latter, let \(h=0\) and \(f\ge 0\) a.e. in \(\Omega \). Then, (1.4) has a solution \(U_0\in W_0^{1,\overrightarrow{p}}(\Omega ) \). Moreover, for \(f=0\), we have that \(\Phi _0(U_0,\nabla U_0)\,U_0\) and \(\Psi (U_0,\nabla U_0)\,U_0\) belong to \(L^1(\Omega )\) and (1.6) holds for \(v=U_0\).
Remark 1.2
Let us stress that if we take \(q_j=0\) for every \(1\le j \le N\), that is \(P_{\overrightarrow{{\mathfrak {a}}}}=\emptyset \), and \(\theta _j\ge p\) for some \(1 \le j \le N\), that is \(N_{\overrightarrow{{\mathfrak {a}}}}\ne \emptyset \), then (1.8) reads as \(m > \max _{j \in N_{\overrightarrow{{\mathfrak {a}}}}}\theta _j\), which is the natural condition to expect when we look for solutions in the energy space \(W_0^{1,\overrightarrow{p}}(\Omega )\).
1.1 A brief history of the problem
To understand how our results fit within the literature, we review what is known in the isotropic case, where the model problem is the following:
Here, \(-\Delta _p u=-\text{ div }\left( |\nabla u|^{p-2}\nabla u\right) \) is the p-Laplacian operator with \(1<p<\infty \), \(\lambda \ge 0\), \(m>1\), \(q\ge 0\) and \(c(\cdot )\) is a continuous, non-negative function. We start by considering \(\lambda =0\), \(c(\cdot )\) constant and f summable enough. The case \(0 \le q < p-1\) is well-known. Indeed, the existence of a solution u in \(W_0^{1,p}(\Omega )\) follows easily from a priori estimates, which are obtained using u as a test function. This is part of the general theory of pseudo-monotone operators by J. Leray and J.-L. Lions (see, for example, [41]). When f has low summability, the main questions appear to be solved (see, for instance, [7, 10] and the references therein). The limiting case \(q=p-1\) is more difficult since the operator \(-\Delta _p u- c\,|\nabla u|^q\) is not coercive for large c. This difficulty has been first overcome by Bottaro and Marina in [14] when \(p=2\), and by various authors in the nonlinear case (see, for example, [10, 24]).
We now focus our attention on the case \(p-1<q\le p\). When \(q=p\), the existence of a bounded weak solution is proved in [12] when \(f\in L^r(\Omega )\) with \(r>N/ p\). The case \(f \in L^{N/p}(\Omega )\) is treated in [29], which shows that there exists a positive constant \(C=C(\beta , N,p)\) such that, if \(\Vert f\Vert _{L^{N/p}(\Omega )}<C\), then a solution \(u\in W_0^{1,p}(\Omega )\) of problem (1.9) exists such that \( \exp \left( \frac{\beta }{p-1}|u|\right) -1\in W_0^{1,p}(\Omega ).\) Similar results are proved in the case \(p-1<q<p\) (see [28, 38] and the references therein).
The authors of [6] consider the case \(p-1<q\le p\) and look for sharp assumptions on f in order to have a solution obtained as a limit of approximations (SOLA).
As far as we know, the more challenging case is \(q >p\): it requires a completely different approach and it appears to be largely open (see, for instance, [18] and the references therein).
The case \(\lambda =0\), \(c(u)=u^\alpha \) with \(\alpha \ge 0\) and \(p=q=2\) is considered in the paper [1]. Among other things, the authors prove that if \(\alpha >0\) and \(f\ge 0\) is sufficiently small, then there exists a positive solution in \(H_0^1(\Omega )\). In [2] (see also [8, 19, 37]) any value of \(\alpha \in {\mathbb {R}}\) and \(1<q\le 2\) is allowed. The authors prove that: if \(\alpha < -1/q\) and \(f\in L^1(\Omega )\), then there exists a distributional solution; if \(-1/q\le \alpha < 0\) and \(f\in L^r(\Omega )\) with \(r>N/2\), then there exists a solution in \(H_0^1(\Omega )\); if \(\alpha \ge 0\), then there exists a solution only if f is small enough. In [35] the presence of an absorption term, which corresponds to \(\lambda >0\) and \(m=2\), is used to prove the existence of a bounded solution in \(H^1_\mathrm{{loc}}(\Omega )\) when \(\alpha <0\), \(p=q=2\) and f is a bounded, non-negative function. In [36] the authors allow the presence of a sign-changing datum f and they discuss related questions as the existence of solutions when the datum \(f \in L^r(\Omega )\), \(r \ge N/2\), or it is less regular, or the boundedness of the solutions when \(r>N/2\). Regarding the existence of classical solutions, we refer the interested reader to [47].
Sharp a priori estimates for solutions to anisotropic problems with \(\lambda =0\) and \(c \equiv 0\) have been proved by Cianchi [21] (see also [4, 5]) by introducing a convenient notion of rearrangement satisfying an anisotropic version of the Pólya-Szegö principle. For other results on anisotropic problems we refer the interested reader to the recent papers [3, 20, 22, 25,26,27, 30,31,32, 37].
We end this section by recalling the paper [13] (see the pioneering papers [23, 39, 43], as well as [15, 17, 33, 40] for the anisotropic equivalent), where the Dirichlet homogeneous problem relative to the equation \(-\Delta u=f/u^\alpha \) is considered. The authors distinguish three cases: \(0<\alpha <1\), \(\alpha =1\) and \(\alpha >1\). The first two cases can be treated using approximation techniques and providing the existence of a unique solution in \(H_0^1(\Omega )\). The validity of a strong comparison principle is a fundamental tool in order to prove the monotonicity, and also a uniform bound far from zero, of the sequence of solutions of the approximate problems. We stress that this kind of arguments cannot be generalized to the anisotropic setting because of the lack of a strong maximum principle (see [45], as well as [34, 42] for existence results without the use of a strong maximum principle).
1.2 Our general problem
We remark that the principal part in (1.4) is the anisotropic \(\overrightarrow{p}\)-Laplacian operator \({\mathcal {A}} u=-\sum _{j=1}^N \partial _j (|\partial _j u|^{p_j-2}\partial _j u)\). It is the prototype of a coercive, bounded, continuous and pseudo-monotone operator \({\mathcal {A}}:W_0^{1,\overrightarrow{p}}(\Omega )\rightarrow W^{-1,\overrightarrow{p}'}(\Omega )\) in divergence form
In this paper, we give existence results for general singular anisotropic elliptic problems such as
where \(f\in L^1(\Omega )\) and \({\mathcal {A}}\) is as in (1.10) with \(A_j(x,t,\xi ):\Omega \times {\mathbb {R}}\times {\mathbb {R}}^N\rightarrow {\mathbb {R}}\) a Carathéodory function for each \(1\le j\le N\) (that is, measurable on \(\Omega \) for every \((t,\xi )\in {\mathbb {R}}\times {\mathbb {R}}^N\) and continuous in \(t,\xi \) for a.e. \(x\in \Omega \)). Moreover, \(\Phi (x,t,\xi ),\, \Theta (x,t,\xi ):\Omega \times {\mathbb {R}}\times {\mathbb {R}}^N\rightarrow {\mathbb {R}}\) are also Carathéodory functions. For any \(r>1\), let \(r'=r/(r-1)\) be the conjugate exponent of r.
The conditions on \({\mathcal {A}}\), \(\Phi \) and \(\Theta \) are similar to those in [16]. We assume that there exist constants \(\nu _0,\nu >0\) and non-negative functions \(\eta _j\in L^{p_j'}(\Omega )\) for \(1\le j\le N\) such that for a.e. \(x\in \Omega \), for all \(t \in {\mathbb {R}}\) and every \(\xi ,{\widehat{\xi }} \in {\mathbb {R}}^N\), we have
We note that in the growth condition in (1.12), we take the greatest exponent for |t| regarding the anisotropic Sobolev inequalities. For the pseudo-monotonicity of \({\mathcal {A}}\), see [16, Lemma 2.7].
Assume that there exist a constant \(C_\Theta >0\), a non-negative function \(c\in L^1(\Omega )\) and a continuous non-decreasing function \( \phi :{\mathbb {R}}\rightarrow {\mathbb {R}}^+\) such that for a.e. \(x\in \Omega \) and all \((t,\xi )\in {\mathbb {R}}\times {\mathbb {R}}^N\),
Compared with [16], we have the extra assumption (1.14) to deal with the new term \(\Psi \) in (1.3).
The operator \({\mathfrak {B}}\) in (1.11) belongs to the general class \({\mathfrak {BC}}\) introduced in [16]. By \({\mathfrak {BC}}\) we denote the class of bounded operators \({\mathfrak {B}}:W_0^{1,\overrightarrow{p}}(\Omega )\rightarrow W^{-1,\overrightarrow{p}'}(\Omega )\) satisfying two properties:
- \((P_1)\):
-
The operator \( {\mathcal {A}}-{\mathfrak {B}}\) from \(W_0^{1,\overrightarrow{p}}(\Omega )\) into \(W^{-1,\overrightarrow{p}'}(\Omega )\) is coercive in the sense that
$$\begin{aligned} \frac{\langle \mathcal Au-\mathfrak Bu,u\rangle }{\Vert u\Vert _{W_0^{1,\overrightarrow{p}}(\Omega )}}\rightarrow \infty \quad \text{ as } \ \Vert u\Vert _{W_0^{1,\overrightarrow{p}}(\Omega )}\rightarrow \infty . \end{aligned}$$ - \((P_2)\):
-
If \(u_\ell \rightharpoonup u\) and \(v_\ell \rightharpoonup v\) (weakly) in \(W_0^{1,\overrightarrow{p}}(\Omega )\) as \(\ell \rightarrow \infty \), then
$$\begin{aligned} \lim _{\ell \rightarrow \infty } \langle {\mathfrak {B}} u_\ell ,v_\ell \rangle = \langle {\mathfrak {B}} u, v \rangle . \end{aligned}$$
We recall from [16] that our assumption \((P_2)\) is somehow reminiscent of (iii) in the Hypothesis (II) of Theorem 1 in the celebrated paper [41] by Leray and Lions. Every operator satisfying \((P_2)\) is strongly continuous (see [16]) and thus pseudo-monotone (cf. [46, p. 586]). However, unlike \({\mathcal {A}}\), the operator \(-{\mathfrak {B}}\) is not necessarily coercive (see Example 1.2).
Let \({\mathfrak {BC}}_+\) be the class of operators in \({\mathfrak {BC}}\) satisfying the extra condition
\((P_3)\) For \(\nu _0>0\) in the coercivity condition of (1.12) and each \(k>0\), it holds
We use \(T_k\) for the truncation at height k, see (1.19).
To indicate that the operator \({\mathcal {A}}\) is associated with the class \({\mathfrak {BC}}\) and \({\mathfrak {BC}}_+\), respectively, we shall write \({\mathfrak {BC}}({\mathcal {A}})\) and \({\mathfrak {BC}}_+({\mathcal {A}})\), respectively. We recall from [16] examples of \({\mathfrak {B}}\) in \({\mathfrak {BC}}({\mathcal {A}})\).
Example
Let \(F\in L^{(p^*)'}(\Omega )\) and \(h,{{\widetilde{h}}}\in W^{-1,\overrightarrow{p}'}(\Omega )\) be arbitrary. Let \(\rho ,\alpha _k\in {\mathbb {R}}\) for \(0\le k\le 4\). For every \(u\in W_0^{1,\overrightarrow{p}}(\Omega )\), we define
-
(1)
\({\mathfrak {B}} u =h\);
-
(2)
\({\mathfrak {B}} u = F+\rho \,|u|^{\vartheta -2}u\) with \(1< \vartheta <p\) if \(\rho >0\) and \(1<\vartheta <p^*\) if \(\rho <0\);
-
(3)
\({\mathfrak {B}} u =(\alpha _0+\alpha _1 \Vert u\Vert ^{{\mathfrak {b}}_1}_{L^r(\Omega )} +\alpha _2 |\langle {{\widetilde{h}}},u\rangle |^{{\mathfrak {b}}_2} ) ( \alpha _3 h+ \alpha _4 F )\) with \( r\in [1,p^*)\); we take \( {\mathfrak {b}}_1\in (0,p/p_1')\) and \( {\mathfrak {b}}_2\in (0,p_1-1)\) if \(\alpha _3\not =0\); \( {\mathfrak {b}}_1\in (0,p-1)\) and \({\mathfrak {b}}_2\in (0,p_1/p')\) if \(\alpha _3=0\);
-
(4)
\({\mathfrak {B}} u = -\sum _{j=1}^N \partial _j\left( \beta _j+|u|^{\sigma _j-1}u\right) \), where \( \beta _j\in L^{p_j'}(\Omega )\) and \(\sigma _j\in (0,p/p_j')\) for all \(1 \le j \le N\).
In each example, \({\mathfrak {B}}\) belongs to the class \({\mathfrak {BC}} ( (1-\varepsilon ){\mathcal {A}})\) for every \(\varepsilon \in [0,1)\).
Definition 1.3
A function \(U_0\in W_0^{1,\overrightarrow{p}}(\Omega )\), which is non-negative in Case 2, is said to be a solution of (1.11) if \(\Phi (x,U_0,\nabla U_0)\in L^1(\Omega )\) and for every \(v\in W_0^{1,\overrightarrow{p}}(\Omega )\cap L^\infty (\Omega )\),
where \(I_{U_0}(v)\) and \(S_{U_0,\Theta ,f}(v)\) are given respectively by (1.5) and
To simplify the notation, we have not included \({\mathcal {A}}\) and \(\Phi \) in the symbol \(S_{U_0,\Theta ,f}(v)\). When \(f=0\), we simply write \(S_{U_0,\Theta }(v)\) instead of \(S_{U_0,\Theta ,f}(v)\).
Assuming (1.12) and (1.13), we have shown in [16, Theorem 1.3] that when \(\Psi =0\) and \(f=0\), then (1.11) has a solution \(U_0\) for every \({\mathfrak {B}}\) in the class \({\mathfrak {BC}}({\mathcal {A}})\). Moreover, \(\Phi (x,U_0,\nabla U_0) \,U_0\in L^1(\Omega )\) and (1.15) holds for \(v=U_0\). If, in addition, there exist constants \(l,\gamma >0\) such that
for all \( |t|\ge l\), a.e. \( x\in \Omega \) and all \( \xi \in {\mathbb {R}}^N\), then (1.11) with \(\Psi =0\) has at least a solution for every \(f\in L^1(\Omega )\) and \({\mathfrak {B}}\) in the class \(\mathfrak {BC}_+({\mathcal {A}})\).
In this paper, under suitable hypotheses, we prove the existence of solutions for (1.11) with \(\Psi \) in (1.3) (see Theorems 1.4 and 1.5 below). Let \(v^\pm =\max \{\pm v,0\}\) be the positive and negative parts of v. In Case 2, we look for non-negative solutions of (1.11) and assume, in addition, that
Without further mention, we henceforth understand that (1.18) holds whenever Case 2 occurs.
Our main results are stated below.
Theorem 1.4
Let (1.1), (1.3), (1.8), and (1.12)–(1.14) hold. Let \(f=0\) in (1.11). Suppose that \({\mathfrak {B}}\) belongs to the class \({\mathfrak {BC}}((1-\varepsilon ){\mathcal {A}})\) for some \(\varepsilon \in (0,1)\). Assume Case 1 or Case 2. Then, there exists a solution \(U_0\in W_0^{1,\overrightarrow{p}}(\Omega ) \cap L^m(\Omega )\) of (1.11). Moreover, both \(\Phi (x,U_0,\nabla U_0)U_0\) and \(\Psi (U_0,\nabla U_0)U_0\) belong to \(L^1(\Omega )\) and (1.16) holds for \(v=U_0\).
When \(N_{\overrightarrow{{\mathfrak {a}}}}\cup P_{\overrightarrow{{\mathfrak {a}}}}=\emptyset \), then Theorem 1.4 gives that (1.11) admits a solution for every \(m>1\).
If in the framework of Theorem 1.4, we have \(\min _{1\le j \le N} {\mathfrak {a}}_j>0\) (in relation to (1.14)), then we obtain the existence of solutions for (1.11) for every \(f\in L^1(\Omega )\) and \({\mathfrak {B}}\) in the class \({\mathfrak {BC}}_+((1-\varepsilon ){\mathcal {A}})\) for some \(\varepsilon \in (0,1)\). More precisely, we prove the following result.
Theorem 1.5
Let (1.1), (1.3), (1.8) and (1.12)–(1.14) hold and, in addition, \(\min _{1\le j \le N} {\mathfrak {a}}_j>0\). Let \(f\in L^1(\Omega )\). Suppose that \({\mathfrak {B}}\) belongs to the class \({\mathfrak {BC}}_+((1-\varepsilon ){\mathcal {A}})\) for some \(\varepsilon \in (0,1)\). Assume Case 1 or Case 2. Then, (1.11) has at least a solution \(u_0\in W_0^{1,\overrightarrow{p}}(\Omega ) \).
1.3 Notation
As usual, in the following sections, we will denote by C a positive constant, the value of which can change from line to line.
For \(k>0\), we let \(T_k:{\mathbb {R}}\rightarrow {\mathbb {R}}\) stand for the truncation at height k, that is,
Moreover, we define \(G_k:{\mathbb {R}}\rightarrow {\mathbb {R}}\) by
so that \(G_k=0\) on \([-k,k]\).
For every \(u\in W_0^{1,\overrightarrow{p}}(\Omega )\) and for a.e. \(x\in \Omega \), we define
For \(u,v,w\in W_0^{1,\overrightarrow{p}}(\Omega )\), we introduce \({\mathcal {E}}_u (v,w)\) as follows
We set \(\overrightarrow{p}=\left( p_1,p_2,\ldots ,p_N\right) \) and \(\overrightarrow{p}'=(p_1',p_2',\ldots ,p_N')\).
As usual, \(\chi _\omega \) stands for the characteristic function of a set \(\omega \subset {\mathbb {R}}^N\).
1.4 Strategy for the proof of Theorems 1.4 and 1.5
We first take \(f=0\) in (1.11) and in the framework of Theorem 1.4, we obtain a solution \(U_0\) (with additional properties that \(\Phi (U_0)\, U_0\in L^1(\Omega )\) and \(\Psi (U_0) \,U_0\in L^1(\Omega )\), allowing us to take \(v=U_0\) in (1.16)). The difficulty in our analysis arises from the interaction of the absorption term \(\Phi \) with the gradient-dependent lower order term \(\Psi \). We point out that \(\Psi \) cannot be integrated into \(\Phi \) since they have the same sign but appear in the opposite sides of (1.11). Moreover, \(\Psi (u)\) is not part of \({\mathfrak {B}} u\) either (except in very special cases such as \(q_j=0\) and \(1<\theta _j<p\) for all \(1\le j\le N\)). Hence, we cannot tackle \(\Psi (u)\) directly in the framework of our paper [16]. We overcome this obstacle by approximating \(\Psi (u)\) by bounded functions \(\Psi _n(u)\) with \(\Vert \Psi _n(u)\Vert _{L^\infty (\Omega )}\le Nn\) for every \(n\ge 1\) (see Sect. 2).
We consider a sequence of approximate problems corresponding to (1.11) with \(f=0\) and \(\Psi \) replaced by \(\{\Psi _n\}_{n\ge 1}\). Then, for each \(n\ge 1\), by applying Theorem 1.3 in [16], we obtain the existence of a solution \(U_n\in W_0^{1,\overrightarrow{p}}(\Omega ) \cap L^m(\Omega )\) for the approximate problem
Moreover, \(U_n\) is non-negative in Case 2 in view of the hypothesis (1.18) (see Lemma 2.1). We capture the properties of \(U_n\) in Proposition 2.3 to be proved in Sect. 4. We are able to get a suitable upper bound for \(\int _\Omega \Psi (U_n) \,U_n\,dx\) via Lemma 4.1. To show that \( \{U_n\}_n\) is bounded in \(W_0^{1,\overrightarrow{p}}(\Omega )\) and also in \(L^m(\Omega )\), we rely on (1.8) and the property \((P_1)\) of \({\mathfrak {B}}\) in the class \({\mathfrak {BC}} ((1-\varepsilon )\,{\mathcal {A}})\). Hence, up to a subsequence, \(\{U_n\}_{n\ge 1}\) converges weakly in both \(W_0^{1,\overrightarrow{p}}(\Omega )\) and \(L^m(\Omega )\) to a function \(U_0\in W_0^{1,\overrightarrow{p}}(\Omega )\cap L^m(\Omega )\). It turns out that \(U_0\) is a good candidate for a solution of (1.11). In addition to \(\Psi _n (U_n) \), we need to handle another gradient-dependent term, namely, \(\Phi (U_n)\). To deal with these terms, we show in Proposition 2.4 that, up to a subsequence,
To prove (1.23), it is enough to show that for a subsequence of \(\{U_n\}_n\), we have
where we define
Indeed, from (1.24) we obtain that, up to a subsequence, \(\nabla U_n^\pm \rightarrow \nabla U_0^\pm \) a.e. in \(\Omega \) and \(U_n^\pm \rightarrow U_0^\pm \) (strongly) in \( W_0^{1,\overrightarrow{p}}(\Omega )\) as \(n\rightarrow \infty \). For details, see Lemma A.4 in the “Appendix”.
Broadly speaking, the proof of (1.24) is inspired by the approach in the celebrated paper [9] dealing with Leray–Lions operators from \(W_0^{1,p}(\Omega )\) into \(W^{-1,p'}(\Omega )\). We point out that, in our case, the analysis becomes more technically involved given the anisotropic setting with the modified growth condition in (1.12) and the inclusion of \({\mathfrak {B}}\) and \(\Psi \). Based on the property \((P_2)\) of \({\mathfrak {B}}\) and a careful use of the absorption term, we show that \(\limsup _{n\rightarrow \infty } \Vert G_k(U_n)\Vert _{W_0^{1,\overrightarrow{p}}(\Omega )} \le W_k\), where \(\lim _{k\rightarrow \infty } W_k=0\), see Lemma 5.2. This is an essential tool not only in the proof of (1.24) but also in that of Lemma A.4 (see Remark A.5). The technical details in the proof of Proposition 2.4 are deferred to Sect. 6.
Then, by Propositions 2.3 and 2.4, we can apply Vitali’s Theorem to obtain that
We end the proof of Theorem 1.4 by showing that, up to a subsequence of \(U_n\), we have
for every \( v \in W_0^{1,\overrightarrow{p}}(\Omega )\cap L^\infty (\Omega )\). For details see Sect. 3.
We remark that it is possible to make the proof of Theorem 1.4 work with only the strong convergence in \(W_0^{1,\overrightarrow{p}}(\Omega )\) for the truncations \(T_k(U_n)\), namely, proving that up to a subsequence,
It is this latter strategy that we adopted in our paper [16] for \(\Psi =0\) (inspired by [11]), first to obtain the existence of solutions for \(f=0\) and then building upon it also for \(f\in L^1(\Omega )\). But unlike Theorem 1.4, the approximation argument for \(f=0\) in [16] concerned the absorption term \(\Phi \).
For Theorem 1.5 dealing with a low summability term \(f\in L^1(\Omega )\), we use a well-known approximation: we replace f in (1.11) by a sequence \(\{f_n\}_{n\ge 1}\) of \(L^\infty (\Omega )\)-functions such that \(|f_n|\le |f|\) for each \(n\ge 1\) and \(f_n\rightarrow f\) in \(L^1(\Omega )\) as \(n\rightarrow \infty \). For the approximate problem, we use Theorem 1.4 to gain a solution \(u_n\in W_0^{1,\overrightarrow{p}}(\Omega )\). The additional assumption \(\min _{1\le j \le N} {\mathfrak {a}}_j>0\) and the extra property \((P_3)\) for \({\mathfrak {B}}\) in \({\mathfrak {BC}}_+((1-\varepsilon ){\mathcal {A}})\) are needed to obtain in Proposition 7.3 that the solutions \(u_n\) are uniformly bounded in \(W_0^{1,\overrightarrow{p}}(\Omega )\) with respect to n. Since here we test the approximate problem with \(T_k(u_n)\) (and not \(u_n\), which is potentially unbounded), we can only derive that \(\{\Phi (u_n)\}_n\) (and not \(\{\Phi (u_n)\,u_n\}_{n\ge 1}\)) is uniformly bounded in \(L^1(\Omega )\) uniformly with respect to n. However, this suffices to get that \(\Phi (u_0)\in L^1(\Omega )\), where \(u_0\) is the weak limit in \(W_0^{1,\overrightarrow{p}}(\Omega )\) of (a subsequence of) \(\{u_n\}_{n\ge 1}\). In Proposition 7.4, we establish the analogue of (1.27). To this end, we use [16, Lemma A.5] (and a diagonal argument) to reduce the proof to showing that for every \(k\ge 1\) and, up to subsequence,
(For the definition of \({\mathcal {E}}_u\), see (1.21).) To prove (1.28), we adapt the approach in our paper [16] by testing the approximate problem with
for \(\lambda =\lambda (k)>0\) large enough. The new ingredient here corresponds to getting a good control of \(I_{u_n}\) for this test function (see Lemma 7.6).
Bearing in mind the strong convergence of \(T_k(u_n)\) to \(T_k(u_0)\) in \(W_0^{1,\overrightarrow{p}}(\Omega )\) as \(n\rightarrow \infty \), we can obtain the analogue of (1.25) and then pass to the limit in the approximate problem to obtain suitable counterparts of (1.26) (see Lemma 7.7 for details). Putting together the above results, we conclude that \(u_0\in W_0^{1,\overrightarrow{p}}(\Omega )\) is a solution of (1.11).
1.5 Structure of the paper
In Sect. 2 we consider the sequence of approximate problems (1.22) and we establish the existence of solutions, which are non-negative in Case 2. We state the a priori estimates and the strong convergence of such solutions in \(W_0^{1,\overrightarrow{p}}(\Omega )\), deferring their proofs to Sect. 4 and 6, respectively. Based on these properties, we complete the proof of Theorem 1.4 in Sect. 3. In Sect. 5 we include several results that are invoked in Sect. 6. Sect. 7 contains the proof of Theorem 1.5. For the reader’s convenience, in the “Appendix” we present some details which are modifications of arguments known in the literature or already contained in our recent paper [16].
2 Approximate problems
We always assume that (1.1), (1.3), (1.12) and (1.13) hold. Unless otherwise stated, we also understand that \(\Phi \) satisfies (1.14) and \({\mathfrak {B}}\) belongs to the class \({\mathfrak {BC}} ((1-\varepsilon ){\mathcal {A}})\) for some \(\varepsilon \in (0,1)\) (see Sect. 2.2 below for an exception). We first take \(f=0\) in (1.11).
2.1 Setting up the approximation
We introduce the sets
Case 1 (respectively, Case 2) in Theorem 1.4 corresponds to \(J_2=\emptyset \) (respectively, \(J_2\ne \emptyset \)).
Let \(n\ge 1\) be arbitrary. For each \(1\le j\le N\), we define
In Case 1, for each \(1\le j\le N\), we extend \(H_{j,n}(t_1,t_2)\) on \((-\infty ,0]\times {\mathbb {R}}\) with the same formula as in (2.1). In Case 2, for each \(j\in J_1\) (when \(J_1\) is not empty), we extend \(H_{j,n}(t_1,t_2)\) on \((-\infty ,0]\times {\mathbb {R}}\) so that it becomes an even function in the first variable.
In Case 1 or Case 2, we define \(\Psi _n\) from \(W_0^{1,\overrightarrow{p}}(\Omega )\) into \( L^\infty (\Omega )\) as follows
where \(\Psi _{n,J_1}(u)\) and \(\Psi _{n,J_2}(u)\) are functions from \(\Omega \) to \({\mathbb {R}}\) given by
Clearly, \(\Psi _n(u)\in L^\infty (\Omega )\) for all \(u\in W_0^{1,\overrightarrow{p}}(\Omega )\) and \( \Vert \Psi _{n}(u)\Vert _{L^\infty (\Omega )}\le N n\).
As explained in Sect. 1.4, we consider the approximate problem (1.22).
2.2 Existence of solutions for (1.22)
We point out that for the existence of solutions of (1.22), we do not need \(\Phi \) to satisfy (1.14). Moreover, the operator \({\mathfrak {B}}\) can be taken in the class \({\mathfrak {BC}} ({\mathcal {A}})\) (rather than \({\mathfrak {BC}} ((1-\varepsilon ){\mathcal {A}})\) for some \(\varepsilon \in (0,1)\)).
Lemma 2.1
Let (1.1), (1.3), (1.12) and (1.13) hold. Suppose that \({\mathfrak {B}}\) belongs to the class \({\mathfrak {BC}}({\mathcal {A}})\). Assume Case 1 or Case 2. For every \(n\ge 1\), problem (1.22) admits a solution \(U_n\), which in addition satisfies \(\Phi (U_n)\,U_n\in L^1(\Omega )\) and
Moreover, in Case 2, we have \(U_n\ge 0\) a.e. in \(\Omega \).
Proof
By applying Theorem 1.3 in [16], with \(\Theta \) there replaced by \(\Theta -\Psi _n\), we obtain that (1.22) has a solution \(U_n\) (in the sense of Definition 1.3 with \(\Psi =0\)), satisfying
for all \(v \in W_0^{1,\overrightarrow{p}}(\Omega ) \cap L^\infty (\Omega )\). Moreover, \(\Phi (U_n)\,U_n\in L^1(\Omega )\) and (2.4) holds. We now show that in Case 2, we have \(U_n\ge 0\) a.e. in \(\Omega \). Since \(U_n^-\) may not be in \(L^\infty (\Omega )\), we cannot directly use \(v=U_n^-\) in (2.5). However, for every \(k>0\), we have \(T_k(U_n^-)\in W_0^{1,\overrightarrow{p}}(\Omega ) \cap L^\infty (\Omega )\). Hence, by taking \(v=T_k (U_n^-)\) in (2.5), we obtain that
Notice that \(\Vert T_k (U_n^-)\Vert _{W_0^{1,\overrightarrow{p}}(\Omega )}\le \Vert U_n^-\Vert _{W_0^{1,\overrightarrow{p}}(\Omega )}\) for all \(k>0\). Moreover, \(\partial _j (T_k (U_n^-))\rightarrow \partial _j U_n^-\) a.e. in \(\Omega \) as \(k\rightarrow \infty \), for every \(1\le j\le N\), so that \( T_k (U_n^-)\rightharpoonup U_n^-\) (weakly) in \(W_0^{1,\overrightarrow{p}}(\Omega )\) as \(k\rightarrow \infty .\) Since \({\mathcal {A}} U_n\) and \({\mathfrak {B}} U_n\) belong to \(W^{-1,\overrightarrow{p}'}(\Omega )\), it follows that
Recalling that \(\Phi (U_n)\, U_n \in L^1(\Omega )\), \( \Vert \Psi _{n}(U_n)\Vert _{L^\infty (\Omega )}\le N n\) and (1.13) holds, from the Dominated Convergence Theorem, we can pass to the limit \(k\rightarrow \infty \) in (2.6) to find that
In view of (1.18), we see that the right-hand side of (2.7) is non-negative. Using also the coercivity condition in (1.12), we infer that the left-hand side of (2.7) is bounded above by
From the sign-conditions on \(\Phi \) and \(\Theta \) in (1.13) and (1.18), respectively, we see that all terms contained in the round brackets of (2.8) are non-negative. Hence, \(\text{ meas }\,(\{U_n<0\})=0\) and so \(U_n \ge 0\) a.e. in \(\Omega \). \(\square \)
Remark 2.2
If, in addition, \(\Phi \) satisfies (1.14), then for the solution \(U_n\) of (1.22) provided by Lemma 2.1, we have \(U_n\in L^m(\Omega )\). This follows from the property \(\Phi (U_n) \,U_n\in L^1(\Omega )\).
2.3 Strong convergence of \(U_n\)
Throughout this section, we work in the framework of Theorem 1.4. Then, Lemma 2.1 and Remark 2.2 give that for every \(n\ge 1\), the approximate problem (1.22) has a solution \(U_n\in W_0^{1,\overrightarrow{p}}(\Omega )\cap L^m(\Omega )\). In Proposition 2.3 we derive essential a priori estimates in \(W_0^{1,\overrightarrow{p}}(\Omega )\) and in \(L^m(\Omega )\) for the sequence of solutions \(\{U_n\}_{n\ge 1}\), which up to a subsequence, converges weakly to some \(U_0\) both in \(W_0^{1,\overrightarrow{p}}(\Omega )\) and in \(L^m(\Omega )\). In Proposition 2.4, we show that, up to a subsequence, \(\{U_n\}_{n\ge 1}\) converges strongly to \(U_0\) in \(W_0^{1,\overrightarrow{p}}(\Omega )\) as \(n\rightarrow \infty \), see (2.12). We aim to prove that \(U_0\) is a solution of (1.11) with \(f=0\). In Sects. 4 and 6, respectively, we prove Propositions 2.3 and 2.4, which are the crux of the proof of Theorem 1.4.
Proposition 2.3
Let (1.1), (1.3), (1.8) and (1.12)–(1.14) hold. Let \(f=0\). Suppose that \({\mathfrak {B}}\) belongs to the class \({\mathfrak {BC}}((1-\varepsilon ){\mathcal {A}})\) for some \(\varepsilon \in (0,1)\). Assume Case 1 or Case 2.
-
(a)
There exists a constant \(C>0\) such that for every \(n\ge 1\), the solution \(U_n\) given by Lemma 2.1 satisfies
$$\begin{aligned} \begin{aligned}&\Vert U_n\Vert _{W_0^{1,\overrightarrow{p}}(\Omega )} + \Vert U_n\Vert _{L^m(\Omega )} + \int _\Omega \Phi (U_n)\,U_n\, dx\\&\ + \int _\Omega \Psi (U_n)\,U_n\, dx \le C. \end{aligned} \end{aligned}$$(2.9) -
(b)
There exists \(U_0 \in W_0^{1,\overrightarrow{p}}(\Omega )\cap L^m(\Omega )\) such that, up to a subsequence, as \(n\rightarrow \infty \)
$$\begin{aligned} \begin{aligned}&U_n \rightharpoonup U_0\ \text{(weakly) } \text{ both } \text{ in } W_0^{1,\overrightarrow{p}}(\Omega )\ \text{ and } \text{ in } L^m(\Omega ), \\&U_n \rightarrow U_0 \text{ a.e. } \text{ in } \Omega . \end{aligned} \end{aligned}$$(2.10)
Proposition 2.4
In the framework of Proposition 2.3, up to a subsequence, we have
Remark 2.5
Under the same assumptions as in Proposition 2.3, by Fatou’s Lemma we immediately infer that \(\Phi (U_0)\in L^1(\Omega )\). Furthermore, using Fatou’s Lemma and (2.9)–(2.11), we find that \(\Phi (U_0)\,U_0\) and \(\Psi (U_0) \,U_0\) belong to \(L^1(\Omega )\).
3 Proof of Theorem 1.4 completed
Let m satisfy (1.8) and \(f=0\). We show that the function \(U_0\) in Proposition 2.4 is a solution of (1.11). Once this is established, we readily obtain that (1.16) holds for \(v=U_0\) in both Case 1 and Case 2 with a reasoning similar to Lemma 2.1. Indeed, by taking \(v=T_k (U_0)\) in (1.16) and letting k go to infinity, we get the claim.
We now prove (1.16). As already pointed out in Sect. 1.4, we just need to check (1.26) for every \(v \in W_0^{1,\overrightarrow{p}}(\Omega )\cap L^\infty (\Omega )\). We first establish the second identity in (1.26), that is
Proof of (3.1)
Since \(U_n\rightarrow U_0\) and \(\nabla U_n\rightarrow \nabla U_0\) a.e. in \(\Omega \) as \(n\rightarrow \infty \), we have
for \( 1\le j\le N\). Let \(v \in W_0^{1,\overrightarrow{p}}(\Omega )\cap L^\infty (\Omega )\) be arbitrary. Now, \(\Theta \) satisfies (1.13). Thus, by the Dominated Convergence Theorem, we obtain that \( \Theta (U_n)\, v\rightarrow \Theta (U_0)\, v\) in \(L^1(\Omega )\) as \(n\rightarrow \infty \). Since \(\{A_j(U_n)\}_{n\ge 1}\) is uniformly bounded in \(L^{p_j'}(\Omega )\) with respect to n, from (3.2) we get that, up to a subsequence,
for every \(1\le j\le N\). It follows that \(\lim _{n\rightarrow \infty } \langle {\mathcal {A}} U_n, v\rangle = \langle {\mathcal {A}} U_0, v\rangle \). Using that \(U_n\rightharpoonup U_0\) (weakly) in \(W_0^{1,\overrightarrow{p}}(\Omega )\) as \(n\rightarrow \infty \), the property \((P_2)\) for the operator \({\mathfrak {B}}\) yields that \(\lim _{n\rightarrow \infty } \langle {\mathfrak {B}} U_n, v\rangle = \langle {\mathfrak {B}} U_0, v\rangle \). Thus, by passing to the limit as \(n\rightarrow \infty \) in (2.5), we gain (3.1) whenever
Since \(\Phi (U_n)\rightarrow \Phi (U_0)\) a.e. in \(\Omega \) as \(n\rightarrow \infty \) and \(\Phi (U_0)\in L^1(\Omega )\) (see Remark 2.5), by Vitali’s Theorem, it is enough to show that \(\{\Phi (U_n)\}_n\) is uniformly integrable over \(\Omega \). Let \(\omega \) be any measurable subset of \(\Omega \) and \(M>0\) be arbitrary. By the growth condition of \(\Phi \) in (1.13), we have
On the other hand, using (2.9) and the sign-condition on \(\Phi \) in (1.13), we see that
where \(C>0\) is a constant independent of n and \(\omega \). Since \(c\in L^1(\Omega )\) and \(\partial _j T_M(U_n)\rightarrow \partial _j T_M(U_0)\) in \(L^{p_j}(\Omega )\) as \(n\rightarrow \infty \) for all \(1 \le j \le N\) (see (2.12)), from (3.5) and (3.6) we get the equi-integrability of \(\{\Phi (U_n)\}_n\) over \(\Omega \). By Vitali’s Theorem, we end the proof of (3.4). \(\square \)
It remains to show the first identity in (1.26), that is
for every \(v \in W_0^{1,\overrightarrow{p}}(\Omega )\cap L^\infty (\Omega )\).
Recall that \(N_{\overrightarrow{{\mathfrak {a}}}}\) and \(P_{\overrightarrow{{\mathfrak {a}}}}\) are given in (1.7). We define
It follows that
For every \(1\le j\le N\), we introduce the notation
To prove (3.7), we treat Case 1 in Sect. 3.1 and Case 2 in Sect. 3.2.
3.1 Proof of (3.7) in Case 1
Here, \(\theta _j>1\) for each \(1\le j\le N\). Since \(J_2=\emptyset \), from (2.2) and (2.3), we find that \(\Psi _n(U_n)=\Psi _{n,J_1}(U_n)=\sum _{j=1}^N H_{j,n} (U_n,\partial _j U_n)\), with \(H_{j,n}(\cdot ,\cdot )\) defined in (2.1). So, to prove (3.7), it suffices to show that (up to a subsequence)
for every \(1\le j\le N\) and all \(v \in W_0^{1,\overrightarrow{p}}(\Omega )\cap L^\infty (\Omega )\).
Let \(1\le j\le N\) be arbitrary. By Proposition 2.4, we have
We next show that there exists \(s>1\) (depending on j) such that
for a positive constant C independent of n. We distinguish the following two situations.
(a) Let \(j\in N_{\overrightarrow{{\mathfrak {a}}}}\cup N_{\overrightarrow{{\mathfrak {a}}}}^{\, c}\) (when \({\mathfrak {a}}_j q_j=0\)). We define s as follows
Let \(c_j\) be given by
By Hölder’s inequality and Proposition 2.3, we infer that
where C is a positive constant independent of n.
(b) Let \(j\in P_{\overrightarrow{{\mathfrak {a}}}}\cup P_{\overrightarrow{{\mathfrak {a}}}}^{\, c}\) (when \({\mathfrak {a}}_j q_j>0\)). Let \({\mathfrak {I}}_{m,p_j}(U_n)\) be as in (3.10). In each of the situations below, we use Hölder’s inequality and Proposition 2.3 to obtain (3.13) for suitable \(s>1\).
- \((b_1)\):
-
If \(m\ge (\theta _j-1)p_j/q_j\), then by choosing \(1<s<p_j/q_j\), we see that
$$\begin{aligned} \begin{aligned}&\Vert H_{j,n}(U_n,\partial _j U_n )\Vert _{L^{s}(\Omega )} \\&\le (\text{ meas }\,(\Omega ))^{\frac{1}{s}-\frac{q_j}{p_j}} \left( {\mathfrak {I}}_{m,p_j} (U_n) \right) ^{\frac{\theta _j-1}{m}} \Vert \partial _j U_n\Vert _{L^{p_j}(\Omega )}^{q_j-\frac{(\theta _j-1)p_j}{m}}. \end{aligned} \end{aligned}$$ - \((b_2)\):
-
If \(\theta _j-1<m<(\theta _j-1)p_j/q_j\), then for \(1<s<m/(\theta _j-1)\), we have
$$\begin{aligned} \begin{aligned}&\Vert H_{j,n}(U_n,\partial _j U_n )\Vert _{L^{s}(\Omega )}\\&\le (\text{ meas }\,(\Omega ))^{\frac{1}{s}-\frac{\theta _j-1}{m}} \left( {\mathfrak {I}}_{m,p_j} (U_n) \right) ^{\frac{q_j}{p_j}} \Vert U_n\Vert _{L^{m}(\Omega )}^{\theta _j-1-\frac{q_jm}{p_j}}. \end{aligned} \end{aligned}$$ - \((b_3)\):
-
If \(1<m\le \theta _j-1\), then we always have \(m>{\mathfrak {m}}_j\). Indeed, if \(j\in P_{\overrightarrow{{\mathfrak {a}}}}\), then the assumption (1.8) gives that \(m>\min \{\theta _j,{\mathfrak {m}}_j\}={\mathfrak {m}}_j\). If, in turn, \(j\in P_{\overrightarrow{{\mathfrak {a}}}}^{\, c}\), then \({\mathfrak {m}}_j\le 1<m\). Hence, \(m>{\mathfrak {m}}_j\) for \( j\in P_{\overrightarrow{{\mathfrak {a}}}}\cup P_{\overrightarrow{{\mathfrak {a}}}}^{\, c}\) leads to
$$\begin{aligned} \begin{aligned}&\Vert H_{j,n}(U_n,\partial _j U_n )\Vert _{L^{p'}(\Omega )} \\&\le (\text{ meas }\,(\Omega ))^{\frac{q_j(m-{\mathfrak {m}}_j)}{pp_j}} \left( {\mathfrak {I}}_{m,p_j} (U_n) \right) ^{\frac{q_j}{p_j}} \Vert U_n\Vert _{L^{p}(\Omega )}^{\theta _j-1-\frac{q_jm}{p_j}}. \end{aligned} \end{aligned}$$Thus, (3.13) holds with \(s=p'\).
This proves (3.13) for every \(1\le j\le N\). Then, using (3.12), we have, up to a subsequence,
as \( n\rightarrow \infty \), where \(s>1\) is chosen according to (a), \((b_1)\), \((b_2)\) or \((b_3)\) (for the latter, we take \(s=p'\)). Thus, (3.11) follows for every \(1\le j\le N\) and all \(v \in W_0^{1,\overrightarrow{p}}(\Omega )\cap L^\infty (\Omega )\).
3.2 Proof of (3.7) in Case 2
Let v be an arbitrary non-negative function in \(W_0^{1,\overrightarrow{p}}(\Omega )\cap L^\infty (\Omega )\). By Lemma 2.1, for each \(n\ge 1\), we have \(U_n\ge 0\) a.e. in \(\Omega \) and the same applies to \(U_0\). Hence, proving (3.7) amounts to showing that
as \(n\rightarrow \infty \), where \(\Psi _{n,J_1} (U_n)\) and \(\Psi _{n,J_2} (U_n) \) can be obtained from (2.3) replacing u by \(U_n\).
From \(U_0\in W_0^{1,\overrightarrow{p}}(\Omega )\), it follows that \(\nabla U_0=0\) a.e. in \(\{U_0=0\}\). For every \(j\in J_1\) we have \(\theta _j>1\) so that with the same argument given for Case 1 in Sect. 3.1, we find that
Hence, by (3.7), we get that there exists \(\lim _{n\rightarrow \infty } \int _\Omega \Psi _{n,J_2}(U_n)\,v\,dx\). To reach (3.15), it remains to show that
To this aim, let us notice that, for every \(\sigma >0\), we have
Fix \(\sigma >0\) such that \(\sigma \not \in {\mathcal {E}}\), where we define
We show that
Assuming that the assertions in (3.19) have been proved, we end the proof of (3.16) as follows. We have \(\chi _{\{U_0>\sigma _2\}}\le \chi _{\{U_0>\sigma _1\}}\) for \(0<\sigma _1<\sigma _2\), and the set \({\mathcal {E}}\) in (3.18) is at most countable. Moreover, from (3.17) and (3.19), we see that
Hence, by the Monotone Convergence Theorem, we deduce that
Using (3.19) and (3.20) in (3.17), we obtain (3.16). It remains to show (3.19).
(i) Let \(j\in J_2\) be arbitrary. We conclude (i) by proving that
as \(n\rightarrow \infty \). For every measurable subset \(\omega \) of \(\Omega \), we have
From Proposition 2.3, using that \(\sigma \notin {\mathcal {E}}\), we obtain that
as well as
as \( n\rightarrow \infty \). By Vitali’s Theorem, we conclude the proof of (3.21).
(ii) Let \(Z_\sigma : [0,\infty )\rightarrow [0,1]\) be the following function
For \(w\in W_0^{1,\overrightarrow{p}}(\Omega )\), we define
Observe that \(Z_\sigma (U_0) \rightarrow \chi _{\{U_0=0\}}\) a.e. in \(\Omega \) as \(\sigma \rightarrow 0\) and \(U_0 \in W_0^{1,\overrightarrow{p}}(\Omega )\) implies that \(\nabla U_0=0\) a.e. in \(\{U_0=0\}\). From (1.18), we have \(\Phi (x,0,0)=0\) and \(A_j(x,0,0)=0\) a.e. in \(\Omega \), for every \(1 \le j \le N\). It follows that \({\mathcal {L}}_{\sigma ,v} (U_0)\rightarrow 0\) as \(\sigma \rightarrow 0\). Hence, we conclude the assertion of (ii) in (3.19) by showing that
From (1.18), we have \(\langle {\mathfrak {B}}U_n,Z_\sigma \left( U_n\right) v\rangle \ge 0\) and \(\Theta (U_n)\le 0\) for every \(n\ge 1\). Thus, by taking \(v\, Z_\sigma (U_n)\ge 0\) as a test function in (2.5) and using the coercivity condition in (1.12), we see that
Since \(Z_\sigma (U_n) = 1\) in \(\{U_n\le \sigma \}\), from (3.25), we derive (3.23).
Using that \(Z_\sigma (U_{n}) \rightarrow Z_\sigma (U_0)\) a.e. in \(\Omega \) as \(n\rightarrow \infty \), by Lebesgue’s Dominated Convergence Theorem, for each \(1\le j\le N\), we find that \(Z_\sigma (U_{n})\,\partial _j v \rightarrow Z_\sigma (U_0)\,\partial _jv\) (strongly) in \(L^{p_j}(\Omega )\) as \(n\rightarrow \infty \). This, jointly with (3.3), implies that
Similar to the proof of (3.4), we have \(\Phi (U_n) \,Z_\sigma (U_n)\rightarrow \Phi (U_0)\, Z_\sigma (U_0)\) in \(L^1(\Omega )\) as \(n\rightarrow \infty \). Then, using \(w=U_n\) in (3.22) and letting \(n\rightarrow \infty \), we obtain (3.24).
The proof of (3.16), and hence of (3.15), is now complete.
This ends the proof of Theorem 1.4. \(\square \)
4 Proof of Proposition 2.3
For each \(n\ge 1\), the solution \(U_n\in W_0^{1,\overrightarrow{p}}(\Omega )\cap L^m(\Omega )\) of (1.22) given in Lemma 2.1 satisfies
(a) We prove that there exists a constant \(C>0\) such that (2.9) holds for all \(n\ge 1\).
We first show that \(\{U_n\}_{n\ge 1}\) is bounded in \( W_0^{1,\overrightarrow{p}}(\Omega )\). We have assumed that \({\mathfrak {B}}\) belongs to the class \({\mathfrak {BC}} ((1-\varepsilon )\, {\mathcal {A}})\) for some \(\varepsilon >0\). Using the coercivity condition in (1.12), (1.14) and (3.10), we find that the left-hand side of (4.1) is bounded below by
We observe that \({\mathfrak {a}}_j>0\) for every \(j\in P_{\overrightarrow{{\mathfrak {a}}}} \cup P_{\overrightarrow{{\mathfrak {a}}}}^{\,c}\). We now consider the right-hand side of (4.1). Using (1.13) and the anisotropic Sobolev inequality (A.2) in the “Appendix”, we find a positive constant C, independent of n, such that
By Young’s inequality, for each \(\delta >0\), there exists a constant \(C_\delta >0\), depending on \(\delta \), such that
By (3.10), we have
In Lemma 4.1, we obtain a suitable upper bound for \(\sum _{j=1}^N {\mathfrak {I}}_{\theta _j,q_j}(U_n)\). To this end, we distinguish the case \(m\ge \theta _j p_j/q_j\) from \(m<\theta _jp_j/q_j\) whenever \(j\in P_{\overrightarrow{{\mathfrak {a}}}}\cup P^{\,c}_{\overrightarrow{{\mathfrak {a}}}}\). We observe that \(P_{\overrightarrow{{\mathfrak {a}}}}\cup P^{\,c}_{\overrightarrow{{\mathfrak {a}}}}=\{ 1\le j\le N: \ {\mathfrak {a}}_j q_j>0\}\) is a union of three sets:
where we define
Lemma 4.1
For any \(\delta >0\), there exists a positive constant \(C_\delta \) such that, for every \(n\ge 1\),
Proof
For the inequalities in (4.8), (4.9), (4.11)–(4.13) below, we use Hölder’s inequality, then Young’s inequality (see Lemma A.1 in the “Appendix”). In what follows, we understand that \(\delta >0\) is arbitrary and \(C_\delta >0\) is a suitable constant depending on \(\delta \).
(I) We first estimate \( {\mathfrak {I}}_{\theta _j,q_j}(U_n)\) for every \(j\in N_{\overrightarrow{{\mathfrak {a}}}}\cup N_{\overrightarrow{{\mathfrak {a}}}}^{\,c}\) when we let \(c_j\) and \(\lambda _j\) be as in (3.14).
\(\bullet \) Let \(j\in N_{\overrightarrow{{\mathfrak {a}}}}\). Condition (1.8) gives that \(\lambda _j>1\) so that
\(\bullet \) Let \(j\in N_{\overrightarrow{{\mathfrak {a}}}}^{\,c}\). Using Lemma A.1 and the anisotropic Sobolev inequality (A.2) in the “Appendix”, we find a positive constant C, depending on N, \(\overrightarrow{p}\), \(q_j\), \(\theta _j\) and \(\text{ meas }\,(\Omega )\), such that
(II) We now estimate \( {\mathfrak {I}}_{\theta _j,q_j}(U_n)\) for every \(j\in P_{\overrightarrow{{\mathfrak {a}}}}\cup P^{\,c}_{\overrightarrow{{\mathfrak {a}}}}\) when we define
Recall that \(P_{\overrightarrow{{\mathfrak {a}}}}:=\{1\le j\le N:\ {\mathfrak {a}}_jq_j>0, \ {\mathfrak {m}}_j> 1\}\). Condition (1.8) implies that \(m>\min \{\theta _j,{\mathfrak {m}}_j\}\) whenever \(j\in P_{\overrightarrow{{\mathfrak {a}}}}\) and, moreover, \( \min \{\theta _j,{\mathfrak {m}}_j\}=\theta _j\) if and only if \( \theta _j\ge p\).
\(\bullet \) For every \(j\in {\widehat{P}}_{\overrightarrow{{\mathfrak {a}}},1}\), we obtain that
\(\bullet \) Let \(j\in P_{\overrightarrow{{\mathfrak {a}}},3}\). In this case, we have \(m>\theta _j\) so that
\(\bullet \) Let \(j\in {\widehat{P}}_{\overrightarrow{{\mathfrak {a}}},2}\). Then \(m>{\mathfrak {m}}_j\). By Hölder’s inequality, Lemma A.1 and the anisotropic Sobolev inequality (A.2) in the “Appendix”, we find a positive constant \(C=C(N,\overrightarrow{p},q_j,\theta _j,m, \text{ meas }\,(\Omega ))\) such that
By adding the inequalities in (4.8), (4.9), (4.11)–(4.13), we complete the proof of (4.7). \(\square \)
Proof of Proposition 2.3completed. From (1.14) and the definition of \(P_{\overrightarrow{{\mathfrak {a}}}}\) and \( P^{\,c}_{\overrightarrow{{\mathfrak {a}}}}\), we have \({\mathfrak {a}}_0>0\) and \(\min _{j\in P_{\overrightarrow{{\mathfrak {a}}}}\cup P^{\,c}_{\overrightarrow{{\mathfrak {a}}}}}{\mathfrak {a}}_j>0\). We choose \(\delta >0\) small such that
By (4.3), (4.4) and Lemma 4.1, there exists a positive constant \(C_\delta \) such that for each \(n\ge 1\), the right-hand side of (4.1) is bounded above by
For ease of reference, we introduce \({\mathfrak {S}}_n\) as follows
In view of (4.1), the quantity in (4.2) is bounded above by that in (4.15). Hence, using the inequalities in (4.14), we infer that for some small constant \(\varepsilon _1>0\), we have
for every \(n\ge 1\). Now, from the hypothesis that \({\mathfrak {B}}\) belongs to the class \({\mathfrak {BC}}((1-\varepsilon ){\mathcal {A}})\) with \(\varepsilon \in (0,1)\), we have the coercivity of the operator \((1-\varepsilon ) {\mathcal {A}}-{\mathfrak {B}}\) from \(W_0^{1,\overrightarrow{p}}(\Omega )\) into \(W^{-1,\overrightarrow{p}'}(\Omega )\). Hence, (4.16) implies that \(\{U_n\}_{n\ge 1}\) is bounded in \( W_0^{1,\overrightarrow{p}}(\Omega )\). Since \({\mathfrak {B}}:W_0^{1,\overrightarrow{p}}(\Omega )\rightarrow W^{-1,\overrightarrow{p}'}(\Omega )\) is bounded, we find a positive constant C such that \(|\langle {\mathfrak {B}} U_n,U_n\rangle |\le C\) for every \(n\ge 1\). Using also the coercivity assumption in (1.12), the inequality in (4.16) gives the boundedness of \(\{{\mathfrak {S}}_n\}_{n\ge 1}\). Using this fact into (4.4) and (4.7), we conclude from (4.4) that
where \(C>0\) is a constant independent of \(n\ge 1\). Returning to (4.1) and using (4.3), we obtain that the sequence of positive functions \(\{\Phi (U_n) \,U_n\}_{n\ge 1}\) is bounded in \(L^1(\Omega )\). The proof of (2.9) is now complete. \(\square \)
(b) From (2.9), there exists a function \(U_0 \in W_0^{1,\overrightarrow{p}}(\Omega )\cap L^m(\Omega )\) such that, up to a subsequence, (2.10) holds. This completes the proof of Proposition 2.3.
\(\square \)
Remark 4.2
From (2.10), we have \( U_n^\pm \rightharpoonup U_0^\pm \) (weakly) in \( W_0^{1,\overrightarrow{p}}(\Omega )\) as \(n\rightarrow \infty \), which yields that \( \lim _{n\rightarrow \infty } \langle {\mathcal {A}} U_0^\pm , U_n^\pm -U_0^\pm \rangle = 0\).
5 Applications of Proposition 2.3
Throughout this section, the assumptions of Proposition 2.3 hold. For each \(n\ge 1\) let \(U_n\) be the solution of (1.22) provided by Lemma 2.1.
Lemma 5.1
Let \(\omega \) be a measurable subset of \(\Omega \). Assume that \(\{V_n\}_{n\ge 1}\) is a sequence in \(W_0^{1,\overrightarrow{p}}(\Omega )\cap L^m(\Omega )\) satisfying \(|V_n|\le |U_n|\) on \(\omega \) for all \(n\ge 1\). Then, for every \(\tau \in (0,1)\) small enough and \(\beta \in (1/\tau ,m/\tau )\) fixed, there exists a positive constant C, independent of \(\omega \), such that
for all \(n\ge 1\).
Proof
Fix \(\tau \) small satisfying \(0<\tau <\min \{m-1, \min _{1\le j\le N}\{\theta _j\},1\}\). Since \(|V_n|\le |U_n|\) on \(\omega \) for all \(n\ge 1\), we have \(|U_n|^{\tau -1}\le |V_n|^{\tau -1}\) on \(\omega \) so that
for every \(1\le j\le N\). Recall from (3.8), (3.9), and (4.5) that
By Hölder’s inequality, with \(c_j\) given (3.14), for every \(j\in N_{\overrightarrow{{\mathfrak {a}}}}^{\,c}\), we have
By the definition of \({\widehat{P}}_{\overrightarrow{{\mathfrak {a}}},2}\) in (4.6), we can take \(\tau \) small such that
Using \(c_j\) given by (4.10), for every \(j\in {\widehat{P}}_{\overrightarrow{{\mathfrak {a}}},2}\), we derive that
We fix \(\beta \in (1/\tau ,m/\tau )\). From (1.8), we have \(\lambda _j>1\) for every \(j\in N_{\overrightarrow{{\mathfrak {a}}}} \), where \(\lambda _j\) is given by (3.14). We choose \(\tau >0\) small such that \((m-1)\,\tau <m/\lambda _j\) for every \(j\in N_{\overrightarrow{{\mathfrak {a}}}}\), which implies that \(\lambda _j m/(m+\tau \lambda _j)<\beta \). Hence, for every \(j\in N_{\overrightarrow{{\mathfrak {a}}}}\), by defining \(c_{j,N_{\overrightarrow{{\mathfrak {a}}}}}=\left( \text{ meas }\,(\Omega ) \right) ^{\frac{1}{\lambda _j}+\frac{\tau }{m}-\frac{1}{\beta }}\), we obtain that
We diminish \(\tau \) such that \(0<\tau <(p_j-q_j)/p_j\) for every \(j\in {\widehat{P}}_{\overrightarrow{{\mathfrak {a}}},1}\). Using that \(m\ge p_j\theta _j/q_j\) for every \( j\in {\widehat{P}}_{\overrightarrow{{\mathfrak {a}}},1}\), by Hölder’s inequality, we infer that
for every \(j\in {\widehat{P}}_{\overrightarrow{{\mathfrak {a}}},1}\), where \(c_{j,{\widehat{P}}_{\overrightarrow{{\mathfrak {a}}},1}}= \left( \text{ meas }\,(\Omega )\right) ^{\frac{p_j-q_j}{p_j}-\frac{1}{\beta }}\).
Finally, for every \(j\in P_{\overrightarrow{{\mathfrak {a}}},3}\), we have \(p\le \theta _j<m<\theta _jp_j/q_j\) in view of (1.8). We let \(\tau >0\) small such that \(\tau <(m-\theta _j)/(m-1)\) for every \(j\in P_{\overrightarrow{{\mathfrak {a}}},3}\). Then, Hölder’s inequality yields that
for every \( j\in P_{\overrightarrow{{\mathfrak {a}}},3}\), where we define \(c_{j,P_{\overrightarrow{{\mathfrak {a}}},3}}:= \left( \text{ meas }\,(\Omega )\right) ^{\frac{m-\theta _j+\tau }{m}- \frac{1}{\beta }}\).
From (5.2)–(5.7), jointly with the a priori estimates in (2.9), we derive (5.1).
\(\square \)
We remark that, as \(n\rightarrow \infty \),
and
For every \(k\ge 1\), we define
In the proof of Lemma 5.2 below, we need several properties of \(\{z_{n,k}^\pm \}_n\), which we summarise next.
5.1 Properties of \(\{z_{n,k}^\pm \}_n\)
From (2.9) and (5.10), we see that \(\{z_{n,k}^\pm \}_n\) is bounded in \(W_0^{1,\overrightarrow{p}}(\Omega )\) and also in \(L^m(\Omega )\) and, up to a subsequence,
Hence, up to a subsequence, using also Remark A.3 in the “Appendix”, as \(n\rightarrow \infty \), we have
From (5.9) and (5.12), by passing to a subsequence, we deduce that
Let \(r\in (1,m)\) be arbitrary. By Vitali’s Theorem and (5.11), up to a subsequence, we get that
Since \({\mathfrak {B}}\) satisfies the property \((P_2)\), from (5.12) we have, up to a subsequence,
By applying Lemma 5.1, we obtain Lemma 5.2 to be used in the proof of Proposition 2.4.
Lemma 5.2
There exist \(\{W_k\}_{k\ge 1}\) and \(\{Z_k\}_{k\ge 1}\) with \(\lim _{k\rightarrow \infty } W_k=\lim _{k\rightarrow \infty } Z_k=0\) such that, up to a subsequence of \(\{U_n\}\), we have for each \(k\ge 1\)
Proof
By a well-known diagonal argument, it suffices to show that for every \(k\ge 1\), there exists a subsequence of \(\{U_n\}\) such that (5.15) and (5.16) hold. Let \(k\ge 1\) be arbitrary.
We prove (5.15). Since \(G_k(U_n)=U_n-T_k(U_n)\) and \(\partial _j T_k(U_n)=\partial _j U_n \chi _{\{|U_n|\le k\}}\) for \(1\le j\le N\), by the coercivity assumption in (1.12), we have
Since \(t \,G_k(t)\ge 0\) for every \(t\in {\mathbb {R}}\), by the sign-condition in (1.13), we find that \(G_k(U_n)\,\Phi (U_n)\ge 0\) for all \(n\ge 1\). Then, by Lemma 2.1, we can test (2.5) with \(v=G_k(U_n)\) and using (1.13), we get
Since \({\mathfrak {B}}\) satisfies the property \((P_2)\), using (5.8) we infer that
By (2.2), we see that
Observe that \(0<|G_k(U_n)|\le |U_n|\) on \(\{|U_n|>k\}\). By Lemma 5.1, for small \(\tau >0\) and \(\beta \in (1/\tau ,m/\tau )\) fixed, there exists a positive constant C, independent of n and k, such that
From (5.20) and (5.21), using (5.8) it follows that
Using this fact, jointly with (5.17), (5.18) and (5.19), we arrive at
Since \(G_k(U_0) \rightharpoonup 0\) (weakly) in \(W_0^{1,\overrightarrow{p}}(\Omega )\) and in \(L^m(\Omega )\) as \(k \rightarrow \infty \), using that \(\tau \beta \in (1,m)\), we find that (up to a subsequence),
Hence, \(\lim _{k\rightarrow \infty } {\mathfrak {R}}_k=0\) and, moreover, the right-hand side of (5.22) converges to 0 as \(k\rightarrow \infty \). The proof of (5.15) is complete.
We now establish (5.16). Let \(\ell >0\) be arbitrary. We take \(v=T_\ell (z_{n,k}^+) \in W_0^{1,\overrightarrow{p}}(\Omega )\cap L^\infty (\Omega )\) as a test function in (2.5) and, proceeding as in the proof of Lemma 2.1, by letting \(\ell \rightarrow \infty \), we get that (2.5) holds for \(v=z_{n,k}^+ \in W_0^{1,\overrightarrow{p}}(\Omega )\cap L^m(\Omega )\). This, jointly with (1.13), implies that
From the definition of \(\Psi _n\) in (2.2), we have
Observe that \(z_{n,k}^+\le U_n\) on \(\{z_{n,k}^+>0\}\). Then, from Lemma 5.1, for sufficiently small \(\tau >0\) and \(\beta \in (1/\tau ,m/\tau )\) fixed, there exists a positive constant C, independent of n and k, such that
By using (5.23), (5.24), (5.25), (5.13) and (5.14), we conclude (5.16) with \(Z_k\) given by
From (5.9), (5.13) and (5.14), we have \(\lim _{k\rightarrow \infty } Z_k=0\) since \(\tau \beta \in (1,m)\). This ends the proof of Lemma 5.2. \(\square \)
For \(\lambda >0\), we define \(\varphi _\lambda :{\mathbb {R}}\rightarrow {\mathbb {R}}\) as follows
We define \(I_{0}(n,k)\) by
Lemma 5.3
Up to a subsequence of \(\{U_n\}_{n}\), we have
Proof
It suffices to show that for each \(k\ge 1\), by passing to a subsequence of \(\{U_n\}\), we have \( \limsup _{n\rightarrow \infty } I_0(n,k) \le 0\). Since \(U_n\rightharpoonup U_0\) and \( \varphi _\lambda (z_{n,k}^-)\rightharpoonup 0\) (weakly) in \(W_0^{1,\overrightarrow{p}}(\Omega )\) as \(n\rightarrow \infty \), by the property \((P_2)\) for \( {\mathfrak {B}}\), we have \( \lim _{n\rightarrow \infty } \langle {\mathfrak {B}} U_n, -\varphi _\lambda (z_{n,k}^-)\rangle =0\). Moreover, up to a subsequence, \(\varphi _\lambda (z_{n,k}^-)\rightarrow 0\) (strongly) in \(L^1(\Omega )\) as \(n\rightarrow \infty \). Thus, it remains to show that
From (2.2), we have
Let \(j \in J_1\) be arbitrary. In view of Lemma 5.1, for sufficiently small \(\tau >0\) and \(\beta \in (1/\tau ,m/\tau )\) fixed, there exists a positive constant C, independent of n and k, such that
We write \(\Omega \) as the union of \(\{|U_n|< z_{n,k}^-\}\) and \(\{|U_n|\ge z_{n,k}^-\}\). Since \(\theta _j\ge 1\), we see that \(|U_n|^{\theta _j-1}\le (z_{n,k}^-)^{\theta _j-1}\) on \(\{|U_n|< z_{n,k}^-\}\). This and (5.30) imply that
With (5.29) and (5.31) in mind, to conclude (5.28), it suffices to show that for each \(j\in J_1\), each term in the right-hand side of (5.31) converges to zero as \(n\rightarrow \infty \).
In light of (5.13) and (5.14), we see that the right-hand side of (5.30) converges to 0 as \(n\rightarrow \infty \) using here that \(\tau \beta \in (1,m)\). For every \(j\in J_1\), let \(\alpha _j\in (0,\theta _j)\) satisfy \(1<\vartheta _j<m\), where we define \(\vartheta _j=(\theta _j -\alpha _j) p_j/(p_j-q_j)\). Since \(z_{n,k}^-\le k\), by Hölder’s inequality, we have
The choice of \(\alpha _j\) yields that \( \lim _{n\rightarrow \infty } \Vert z_{n,k}^{-}\Vert _{L^{\vartheta _j}(\Omega )}= 0\). Then, for every \(j\in J_1\), the last term in the right-hand side of (5.31) converges to 0 as \(n\rightarrow \infty \). This completes the proof of (5.28). \(\square \)
6 Proof of Proposition 2.4
As explained in Sect. 1.4, we conclude (2.11) and (2.12) by showing that (1.24) holds. We observe that in Case 2, we need only prove that \( {\mathcal {E}}_{U_n}(U_n^+,U_0^+)\rightarrow 0\) in \(L^1(\Omega )\) as \(n\rightarrow \infty \) since all \(U_n\) and, hence, \(U_0\) are non-negative functions. Similarly, we can establish the other convergence claim in (1.24). We thus show the details only for \( {\mathcal {E}}_{U_n}(U_n^+,U_0^+)\) in (1.24) and leave the modifications for \( {\mathcal {E}}_{U_n}(U_n^-,U_0^-)\) to the reader noting that instead of \(z_{n,k}\) in (5.10), one needs to work with \(y_{n,k}\) defined by \( y_{n,k}:=U_n^- - T_k(U_0^-)\).
In light of the monotonicity assumption in (1.12), we have \( {\mathcal {E}}_{U_n}(U_n^+,U_0^+)\ge 0\) a.e. in \(\Omega \). Hence, to attain (1.24) for \( {\mathcal {E}}_{U_n}(U_n^+,U_0^+)\), it remains to show that
Notation. Let \(\omega \) be a measurable subset of \(\Omega \) and \(v,w,z\in W_0^{1,\overrightarrow{p}}(\Omega )\). We introduce
If either of the variables v, w and z or \(\omega \) depends on n, we drop the subscript n in \(E_{n,\omega }(v,w,z)\).
Fix \(k>0\). We define \(z_{n,k}\) as in (5.10). From (1.20), we see that \(G_k(U_0^+)\ge 0\) a.e. in \(\Omega \). Since
we infer that
We show that, up to a subsequence of \(\{U_n\}\), there exist \(\mu _j\in L^{p_j'}(\Omega )\) for \(1\le j\le N\) such that
Indeed, by the growth condition in (1.12), there exists a positive constant C, independent of n and k such that, for \(1\le j\le N\), it holds
Hence, passing to a subsequence of \(\{U_n\}\), we can find \(\mu _j\in L^{p_j'}(\Omega )\) for \(1\le j\le N\) such that
This proves the claim in (6.4).
We complete the proof of (6.1) assuming that the next two results hold.
Lemma 6.1
For every \(k\ge 1\), there exist \(R_1(k)\) and \(R_2(k)\) such that, up to a subsequence of \(\{U_n\}_{n\ge 1}\), we have
where \(\lim _{k\rightarrow \infty } R_1(k)=\lim _{k\rightarrow \infty } R_2(k)=0\).
Lemma 6.2
For every \(k\ge 1\), by passing to a subsequence of \(\{U_n\}_{n\ge 1}\), we have
For the proof of Lemmata 6.1 and 6.2, we refer to Sects. 6.1 and 6.2, respectively.
Hence, by using a diagonal argument, there exists a subsequence of \(\{U_n\}_{n\ge 1}\) such that for every \(k \ge 1\) (6.4) holds and Lemmata 6.1 and 6.2 apply.
Consequently, using also (6.3), we deduce that
for every integer \(k\ge 1\).
Hence, by using (5.9) and letting \(k\rightarrow \infty \) in (6.8), we conclude the proof of (6.1).
6.1 Proof of Lemma 6.1
Let \(k\ge 1\). By Lemma 5.2, it suffices to show that there exist a positive constant C, independent of k, and \(R_0(k)\) with \(\lim _{k \rightarrow \infty } R_0(k)= 0\) such that
and
Proof of (6.9)
We define \(L_1(n,k)\), \(L_2(n,k)\) and \(L_3(n,k)\) by
It follows that
Note that \(\chi _{\{|U_n|\ge k\}}\partial _jU_n^+=\chi _{\{U_n\ge k\}}\partial _j G_k(U_n)\) for every \(1 \le j \le N\). Hence, by Hölder’s inequality and (6.5), we obtain that
where \(C>0\) is a constant independent of n and k.
By the Dominated Convergence Theorem, we see that \(R_0(k)\rightarrow 0\) as \(k\rightarrow \infty \), where we define
Using (6.11) and (6.12), we finish the proof of (6.9) by showing that
Let \(1\le j\le N\) be arbitrary. As \(n\rightarrow \infty \), we have
Since \(E_{j,U_n}(T_k(U_0^+),U_0^+)\rightarrow E_{j,U_0}(T_k(U_0^+),U_0^+)\) a.e. in \(\Omega \) as \(n\rightarrow \infty \), using (6.5) and passing to a subsequence of \(\{U_n\}\), we find that
as \(n\rightarrow \infty \). Hence, from (6.15) and (6.16), we infer that as \(n\rightarrow \infty \)
This proves the first limit in (6.14) since
Let \(1\le j\le N\) be arbitrary. We now remark that the sequences \(\{ E_{j,U_n} (T_k(U_0^+), U_0^+) \,\chi _{\{|U_n|\le k\}} \}_{n}\), \(\{\!A_j(x,U_n,\nabla T_k(U_0^+)) \,\chi _{\{|U_n|\le k\}}\}_{n} \) are uniformly integrable in \(L^{p_j'}(\Omega )\) with respect to n. Since \(U_n \rightarrow U_0\) a.e. in \(\Omega \) as \( n\rightarrow \infty \), by Vitali’s Theorem, we obtain that as \(n\rightarrow \infty \)
Recall that \(\partial _j U_n^+\rightharpoonup \partial _j U_0^+\) and \(\partial _j z_{n,k}^+\rightharpoonup \partial _j G_k(U_0^+) \) (weakly) in \(L^{p_j}(\Omega )\) as \(n\rightarrow \infty \). Since \(\chi _{\{|U_0|\le k\}}\, \partial _j G_k(U_0^+) =0\), using (6.17) and (6.18), we conclude that as \(n\rightarrow \infty \)
Moreover, from (6.19) and the squeeze law, we get the second limit in (6.14).
Proof of (6.10)
We define \(P_1(n,k)\) and \(P_2(n,k)\) as follows
Since on the set \(\{ U_n\ge k\}\) we have \(z_{n,k}^+=U_n-T_k(U_0^+)\) and \(\partial _j U_n=\partial _j G_k(U_n)\) for \(1\le j\le N\), the definition of \(P_2(n,k)\) yields that \(P_2(n,k)=P_{2,1}(n,k)+P_{2,2}(n,k)\), where
From (6.20), we get \(\lim _{n\rightarrow \infty } P_1(n,k)=0\). When \(z_{n,k}^+>0\), then \(U_n^+>0\) so that
Hence, we arrive at
Consequently, we end the proof of (6.10) once we show that
where C is a positive constant independent of k.
As for (6.12), we find a positive constant C, independent of n and k such that
For \(1\le j\le N\), by the Dominated Convergence Theorem, we get \( \chi _{\{ U_n\ge k\}} \chi _{\{0<U_0<k\}}\partial _j U_0\rightarrow 0\) (strongly) in \( L^{p_j}(\Omega )\) as \(n\rightarrow \infty \). Since \(A_j(x,U_n,\nabla U_0)\rightharpoonup A_j(x,U_0,\nabla U_0)\) (weakly) in \(L^{p_j'}(\Omega )\) as \(n\rightarrow \infty \), we infer that \( \lim _{n\rightarrow \infty } P_{2,2}(n,k) =0\). This, together with (6.22), proves (6.21).
The proof of Lemma 6.1 is now complete. \(\square \)
Remark 6.3
The reasoning in the proof of \(\lim _{n\rightarrow \infty } L_1(n,k)=0\) cannot be extended to get \(\lim _{n\rightarrow \infty } E_{\Omega } (T_k(U_0^+),U_0^+,U_n^+-U_0^+)=0\). Indeed, in the growth condition in (1.12), we have taken the greatest exponent for |t| regarding the anisotropic Sobolev inequalities so that we don’t have the compactness of the embedding \(W_0^{1,\overrightarrow{p}}(\Omega )\hookrightarrow L^{p^*}(\Omega )\). Hence, we cannot infer that \(\{ E_{j,U_n} (T_k(U_0^+), U_0^+) \}_{n\ge 1}\) is uniformly integrable in \(L^{p_j'}(\Omega )\) with respect to n.
6.2 Proof of Lemma 6.2
We need to show that, up to a subsequence, (6.7) holds, namely,
where we define D(n, k) by
We choose \(\lambda =\lambda (k)>0\) large such that \( 4\nu _0^2 \,\lambda >\phi ^2(k)\), where \(\phi \) appears in the growth assumption on \(\Phi \) in (1.13), while \(\nu _0>0\) is given by the coercivity condition in (1.12). We define \(\varphi _\lambda \) as in (5.26). Our choice of \(\lambda \) ensures that for every \( t\in {\mathbb {R}}\)
Recall that \(I_{0}(n,k)\) is defined in (5.27). For convenience, we set
We divide the proof of (6.23) into two steps.
Step 1. Let \(\nu _0\) and \(c,\phi \) be as in (1.12) and (1.13), respectively. We have
Proof of STEP 1
On the set \(\{U_n>T_k(U_0^+)\}\), we have \(D(n,k) =0\) since \(z_{n,k}^-=0\) and, hence, \(\partial _j z_{n,k}^-=0\) for \(1\le j\le N\). In turn, on the set \(\{U_n\le T_k(U_0^+)\}\), we find that \(z_{n,k}^-=T_k(U_0^+)-U_n^+\) and, by the monotonicity condition in (1.12), it follows that \(D(n,k)\ge 0\). Hence, we have
for each \(1\le j\le N\). Then, using (6.25), we find that
From (6.24) and (6.26), we observe that
Since \( z_{n,k}^-\in W_0^{1,\overrightarrow{p}}(\Omega )\cap L^\infty (\Omega )\), we have \(\varphi _\lambda (z_{n,k}^-)\in W_0^{1,\overrightarrow{p}}(\Omega )\cap L^\infty (\Omega )\) so that \(\varphi _\lambda (z_{n,k}^-) \) can be taken as a test function in (2.5). Hence, using (1.13) and \(I_0(n,k)\) given by (5.27), we find that
In view of (6.29)–(6.31), we conclude (6.27) by showing that
To this end, we next prove that
where \(I_{3}(n,k)\) is defined by
Indeed, since \(z_{n,k}^-=0\) on \(\{U_n^+>T_k(U_0^+)\}\) and \( \Phi (U_n) \le 0\le \varphi _\lambda (z_{n,k}^-)\) on \(\{U_n\le 0\}\), we have
Next, from the growth condition on \(\Phi \) in (1.13) and the coercivity condition in (1.12), we get
Consequently, the assertion of (6.33) is proved.
Since \(\varphi (z_{n,k}^-)=0\) on \(\{U_n>T_k(U_0^+)\}\), we have
where \(I_2(n,k)\) is given in (6.26). From (6.33) and (6.35), we attain (6.32). This ends the proof of (6.27) and of Step 1. \(\square \)
Step 2. Proof of (6.23) concluded.
Proof of STEP 2
Since \(0\le c(x) \,\varphi _{\lambda } (z_{n,k}^-)\le k\,e^{\lambda k^2}\,c(x) \) a.e. in \(\Omega \) and \(c(x)\, \varphi _\lambda (z_{n,k}^-)\rightarrow 0\) a.e. in \(\Omega \) as \(n\rightarrow \infty \), by the Dominated Convergence Theorem, we have \(\lim _{n\rightarrow \infty } \int _\Omega c(x) \,\varphi _\lambda (z_{n,k}^-)\,dx=0\). In view of Step 1 and Lemma 5.3, we conclude Step 2 by showing that, up to a subsequence,
From (5.12), we have that both \(z_{n,k}^-\) and \( \varphi _\lambda (z_{n,k}^-)\) converge to 0 weakly in \(W_0^{1,\overrightarrow{p}}(\Omega )\) as \(n\rightarrow \infty \). In particular, for each \(1\le j\le N\), it holds
We recall that \(z_{n,k}^-=T_k(U_0^+)\) on \(\{U_n\le 0\}\) and \(\varphi _\lambda (z_{n,k}^-)=0\) on \(\{U_n>T_k(U_0^+)\}\).
(1) We show that \(\lim _{n\rightarrow \infty } I_{1}(n,k)=0\). If \(I_{1,1}(n,k)\) is the first term in \(I_1(n,k)\) in (6.26), then
Let \(1\le j\le N\) be arbitrary. By the Dominated Convergence Theorem, we get
as \( n\rightarrow \infty \). Hence, using that
as \(n\rightarrow \infty \), we obtain that the first term in the right-hand side of (6.38) converges to 0 as \(n\rightarrow \infty \). Moreover, on \(\{0\le U_n\le T_k(U_0^+)\}\), we have \(U_n\le k\) and the family \(\{|A_j(x,U_n,\nabla T_k(U_0^+))|^{p_j'}\}_{n\ge 1}\) is uniformly integrable. Then, based on
as \(n\rightarrow \infty \), by Vitali’s Theorem, we infer that as \(n\rightarrow \infty \)
This, jointly with (6.37), implies that the second term in the right-hand side of (6.38) converges to 0 as \(n\rightarrow \infty \). This proves that \(\lim _{n\rightarrow \infty } I_{1,1}(n,k)=0\).
We now show that the remaining term in the definition of \(I_1(n,k)\) in (6.26) converges to 0 as \(n\rightarrow \infty \), that is,
By the definition of \(E_{n,\omega }(\cdot ,\cdot ,\cdot )\) in (6.2), since \(z_{n,k}^-=T_k(U_0^+)=-z_{n,k}\) on \(\{U_n\le 0\}\), we get
As for (6.6), by passing to a subsequence of \(\{U_n\}\), we get that
converge weakly in \(L^{p_j'}(\Omega )\) as \( n\rightarrow \infty \). This and (6.39) yield
(2) We show that \(\lim _{n\rightarrow \infty } I_2(n,k)=0\). From (5.11) and \(0\le z_{n,k}^-\le k\) a.e in \(\Omega \), we have \(\varphi _\lambda (z_{n,k}^-)\rightarrow 0\) a.e. in \(\Omega \) as \(n\rightarrow \infty \) and \(0\le \varphi _\lambda (z_{n,k}^-)\le k \,e^{\lambda k^2}\) a.e. in \(\Omega \). Thus, by the Dominated Convergence Theorem, for each \(1\le j\le N\), we find that as \(n\rightarrow \infty \)
Consequently, as \( n\rightarrow \infty \), we get
For \(1\le j\le N\), we have \( z_{n,k}=-z_{n,k}^-\) on \(\{0\le U_n\le T_k(U_0^+)\}\) so that using (6.40) and the weak convergence \(\partial _j z^-_{n,k}\rightharpoonup 0\) in \(L^{p_j}(\Omega )\) as \(n\rightarrow \infty \), we arrive at
as \( n\rightarrow \infty \). It follows that
Since \(\varphi _\lambda (z_{n,k}^-)=0\) on \(\{U_n>T_k(U_0^+)\}\), from (6.41) and (6.42), we find that \(\lim _{n\rightarrow \infty } I_2(n,k)=0\), completing the proof of (6.36) and of Step 2. \(\square \)
This finishes the proof of Lemma 6.2. \(\square \)
7 Proof of Theorem 1.5
Let (1.1), (1.3), (1.8) and (1.12)–(1.14) hold and, in addition, let \(\min _{1\le j \le N} {\mathfrak {a}}_j>0\). Here, we suppose that the function f in (1.11) is not identically 0. In Case 2, we assume that \(f\ge 0\) a.e. in \(\Omega \). We approximate f by a sequence of functions \(f_n\in L^\infty (\Omega )\), taking for instance
In particular, in Case 2, we have \(f_n\ge 0\) a.e. in \(\Omega \). We remark the following properties
With this approximation, assuming that \({\mathfrak {B}}\) belongs to \({\mathfrak {BC}}((1-\varepsilon )\,{\mathcal {A}})\) for some \(\varepsilon \in (0,1)\), in either Case 1 or Case 2, we can apply Theorem 1.4 for the problem generated by (1.11) with \(f_n\) instead of f. Then such an approximate problem admits at least a solution \(u_n\), namely,
Moreover, \(\Phi (u_n)\,u_n\) and \(\Psi (u_n)\,u_n\) belong to \( L^1(\Omega )\),
and
for every \( v\in W_0^{1,\overrightarrow{p}}(\Omega )\cap L^\infty (\Omega )\). Furthermore, (7.3) holds for \(v=u_n\).
In the rest of the paper, we understand that \(u_n\) is a solution of (7.2) with the above-mentioned properties that we obtain from Theorem 1.4.
But, unlike Theorem 1.4, to prove that \(\{u_n\}_{n\ge 1}\) is uniformly bounded in \(W_0^{1,\overrightarrow{p}}(\Omega )\), we need \({\mathfrak {B}}\) to satisfy the extra condition \((P_3)\) associated with \((1-\varepsilon )\,{\mathcal {A}}\), namely, for every \(k>0\),
as \( \Vert u\Vert _{W_0^{1,\overrightarrow{p}}(\Omega )} \rightarrow \infty \). Thus, \({\mathfrak {B}}\) belongs to the class \({\mathfrak {BC}}_+((1-\varepsilon )\,{\mathcal {A}})\). This assumption is made throughout this section.
All the results in this section are derived in the framework of Theorem 1.5.
7.1 A priori estimates
In order to obtain a priori estimates for \(u_n\) solving (7.2) we need the following result, which is in the spirit of Lemma 4.1.
Lemma 7.1
Let \(k\ge 1\) be arbitrary and \(\Phi _0\) be given by (1.2). Then, for every \(\rho >0\), there exists a constant \(C_\rho >0\) such that for all \(n\ge 1\), we have
Remark 7.2
The property \(\Phi (u_n)\,u_n\in L^1(\Omega )\) and (1.14) ensure that \(\int _{\{|u_n|\ge k\}} |\Phi _0(u_n)|\,dx<\infty \) for all \(k\ge 1\) and \(n\ge 1\).
We define
If in the definition of \( {\mathfrak {I}}_{m-1,p_j} (k,u_n)\), we replace \(m-1\) and \(p_j\) by \(\theta _j-1\) and \(q_j\), respectively, then we obtain \({\mathfrak {I}}_{\theta _j-1,q_j} (k,u_n)\).
Proof of Lemma 7.1
We observe that
Using the definition of \(I_{u_n}(v)\), we see that
Let \(\delta >0\) be arbitrary. By Hölder’s inequality and Young’s inequality, there exists a constant \(C_\delta >0\) such that for every \(n\ge 1\),
Let \(j\in J_1\) be arbitrary. To estimate \({\mathfrak {I}}_{\theta _j-1,q_j} (k,u_n)\), we distinguish several cases:
Case (a) Let \(q_j=0\) and \(\theta _j\ge p+1\). Then, \(j\in N_{\overrightarrow{{\mathfrak {a}}}}\) and from (1.8), we have \(m> \theta _j\). Hence, by Young’s inequality, there exists a constant \(C_{\delta }>0\) such that for all \( n\ge 1\),
Case (b) Let \(q_j=0\) and \(\theta _j< p+1\). We set \(\gamma _j=1-(\theta _j-1)/p\) and \(C_j=\left( \textrm{meas}\,(\Omega )\right) ^{\gamma _j}\). By Hölder’s inequality, Remark A.3 and Lemma A.1 in the “Appendix”, we find constants \(C>0\) (depending on N, \(\overrightarrow{p}\), \(\theta _j\) and \(\text{ meas }\,(\Omega )\)) and \(C_\delta >0\) such that for all \(n\ge 1\),
When \(q_j>0\), we define \(\zeta _j\) as follows
Case (c) Let \(q_j>0\) and \(\zeta _j\le 0\). We set \(\gamma _j=1-q_j/p_j\) and \(C_j=\left( \text{ meas }\,(\Omega )\right) ^{\gamma _j} \). Then, by Hölder’s inequality and Lemma A.1, there exists \(C_\delta >0\) such that
Case (d) Let \(q_j>0\) and \(\zeta _j>0\). We distinguish three sub-cases:
- (\(\hbox {d}_1\)):
-
Let \({\mathfrak {m}}_j>1\) and \(\theta _j\ge p\). Then, \(j\in P_{\overrightarrow{{\mathfrak {a}}}}\) and from (1.8), we have \(m>\theta _j=\min \{{\mathfrak {m}}_j,\theta _j\}\). We set \(\gamma _j:=(m-\theta _j)/(m-1)\) and \(C_j:=\left( \text{ meas }\,(\Omega )\right) ^{\gamma _j}\). It follows that
$$\begin{aligned} \begin{aligned} {\mathfrak {I}}_{\theta _j-1,q_j} (k,u_n)&\le C_j \left( {\mathfrak {I}}_{m-1}(k,u_n)\right) ^{\frac{\zeta _j}{m-1}} \left( {\mathfrak {I}}_{m-1,p_j}(k,u_n) \right) ^{\frac{q_j}{p_j}}\\&\le \delta \,{\mathfrak {I}}_{m-1}(k,u_n) +\delta \, {\mathfrak {I}}_{m-1,p_j}(k,u_n)+C_\delta , \end{aligned} \end{aligned}$$(7.12)where \(C_\delta >0\) is a suitable constant depending on \(\delta \).
- (\(\hbox {d}_2\)):
-
Let \({\mathfrak {m}}_j>1\) and \(\theta _j<p\). Then, from (1.8), we see that \(m>{\mathfrak {m}}_j=\min \{{\mathfrak {m}}_j,\theta _j\}\).
- (\(\hbox {d}_3\)):
-
Let \({\mathfrak {m}}_j\le 1\). Here, we have \(m>1\ge {\mathfrak {m}}_j\).
We next treat sub-cases (d\(_2\)) and (d\(_3\)) together to get (7.14) below. Using that \(m>{\mathfrak {m}}_j\), we define
We let \(C_j=\left( \text{ meas }\,(\Omega )\right) ^{\gamma _j}\). By Hölder’s inequality, the anisotropic Sobolev inequality (A.2) in the “Appendix” and Lemma A.1, we find constants \(C>0\) (depending on N, \(\overrightarrow{p}\), \(\theta _j\), \(q_j\), m and \(\text{ meas }\,(\Omega )\)) and \(C_\delta >0\) such that for all \(n\ge 1\), we have
Since \(\delta >0\) is arbitrary, the conclusion of Lemma 7.1 follows from (7.6) based on the inequalities in (7.7)–(7.9), (7.11), (7.12) and (7.14). \(\square \)
We now proceed with the proof of the a priori estimates of \(u_n\).
Proposition 7.3
The following hold.
-
(a)
There exists a positive constant C such that for all \(n\ge 1\), we have
$$\begin{aligned} \Vert u_n\Vert _{W_0^{1,\overrightarrow{p}}(\Omega )}+ \int _\Omega |\Phi (u_n)|\,dx \le C. \end{aligned}$$(7.15) -
(b)
There exists \(u_0\in W_0^{1,\overrightarrow{p}}(\Omega )\) such that, up to a subsequence of \(\{u_n\}_{n\ge 1}\),
$$\begin{aligned} \begin{aligned}&u_n\rightharpoonup u_0\ \text{(weakly) } \text{ in } W_0^{1,\overrightarrow{p}}(\Omega )\ \text{ as } n\rightarrow \infty ,\\&u_n \rightarrow u_0\ \text{ a.e. } \text{ in } \Omega \ \text{ as } n\rightarrow \infty . \end{aligned} \end{aligned}$$(7.16)
Proof
(a) Fix \(k\ge 1\) large such that \(k^{m-1} (k-1) \min _{1\le j\le N} {\mathfrak {a}}_j\ge \nu _0\). We define
We have \(\partial _j T_k(u_n)=\chi _{\{|u_n|< k\}} \,\partial _j u_n\) a.e. in \( \Omega \) for \(1\le j\le N\). By the sign-condition of \(\Phi \) in (1.13), we see that
Since \(\Vert f_n\Vert _{L^1(\Omega )}\le \Vert f\Vert _{L^1(\Omega )}\), by taking \(v=T_k(u_n)\in W_0^{1,\overrightarrow{p}}(\Omega )\cap L^\infty (\Omega )\) in (7.3), we find that
where \(C_0:=k \left( \Vert f\Vert _{L^1(\Omega )}+C_\Theta \,\textrm{meas}\,(\Omega )\right) \). Lemma 7.1 gives that for every \(\rho >0\), there exists a constant \(C_\rho >0\) such that (7.5) holds for all \(n\ge 1\). Using (7.5) into (7.18), we find that
We fix \(0<\rho <\min \,\{1, \varepsilon \nu _0\}\). Hence, using (7.17), (1.14), our choice of k and the coercivity condition in (1.12), we derive that
By the choice of \(\rho \) and (7.4), we conclude the boundedness of \(\{u_n\}_{n}\) in \(W_0^{1,\overrightarrow{p}}(\Omega )\). Since \({\mathfrak {B}}\) is a bounded operator from \(W_0^{1,\overrightarrow{p}}(\Omega )\) into its dual, we have \(|\langle {\mathfrak {B}}u_n,T_k (u_n)\rangle |\le C\), where C is a positive constant independent of n. Thus, from (7.19), we readily deduce that
Using the growth condition of \(\Phi \) in (1.13), we find that
for all \( n\ge 1\). This completes the proof of (7.15).
(b) Up to a subsequence, the assertion in (7.16) follows from (7.15). \(\square \)
7.2 Strong convergence of \(T_k(u_{n})\)
Our aim in this section is to prove the following Proposition 7.4.
Proposition 7.4
Up to a subsequence of \(\{u_n\}_n\), as \(n\rightarrow \infty \), we have
for every positive integer k.
Remark 7.5
We have \(\Phi (u_0)\in L^1(\Omega )\). Indeed, using the a.e. convergences of \(\{u_n\}\) and \(\{\nabla u_n\}\) in (7.16) and (7.20), respectively, we obtain that \(|\Phi (u_n)|\rightarrow |\Phi (u_0)|\) a.e. in \(\Omega \) as \(n\rightarrow \infty \). Then, the claim follows from (7.15) and Fatou’s Lemma.
To derive (7.20), we can proceed as in the proof of Lemma 4.2 in [16]. However, the new ingredient here is Lemma 7.6, which is due to the introduction of \(\Psi \) in (1.11).
We define \(Q_j(n,k)\) and \(R_{j}(n,k)\) for \(1\le j\le N\), as well as \(V_j(n,k)\) if \(j\in J_1\) and \(W_j(n,k)\) if \(j\in J_2\) as follows
Let \(\varphi _\lambda \) be as in the proof of Lemma 6.2 (see (5.26)). For every \(n,k\ge 1\), we set
Lemma 7.6
We have
Proof
Since \(\varphi _\lambda (Z_{n,k})=Z_{n,k} \,e^{\lambda (Z_{n,k})^2}\), from (1.3) we find that
Since \(|Z_{n,k}|\le 2k\) a.e. in \(\Omega \), using (7.21), we infer that
We separate the case \(j\in J_2\) from \(j\in J_1\).
(I) Let \(j\in J_2\), which pertains to Case 2 when \(u_n\ge 0\) a.e. in \(\Omega \) and, hence, \(u_0\ge 0\) a.e. in \(\Omega \). Remark that \((k-T_k(u_0)) \chi _{\{u_n\ge k\}} \rightarrow 0\) in \(L^{(p_j/q_j)'}(\Omega )\) as \(n\rightarrow \infty \). Since \(\{ \partial _j u_n\}_{n\ge 1}\) is bounded in \(L^{p_j}(\Omega )\), by Hölder’s inequality, we infer that
as \( n\rightarrow \infty \). With a similar argument, we obtain that
Let \(0<\tau <\min _{i\in J_2} \theta _i\). Hence, \(\tau \in (0,1)\) and
on the set \(\{ T_k(u_0)\le u_n\le k\}\). Since
proceeding as above, we obtain that
(II) We now assume that \(j\in J_1\) and we show that
As in the proof of Lemma 7.1, we distinguish several situations:
Case (a). Let \(q_j=0\) and \(\theta _j\ge p+1\). In this case, \(j\in N_{\overrightarrow{{\mathfrak {a}}}}\) and hence, (1.8) yields that \(m> \theta _j\).
Case (b). Let \(q_j=0\) and \(\theta _j< p+1\).
In Cases (a) and (b) above, we define \(\gamma _j=1-(\theta _j-1)/r\), where \(r=m-1\) in Case (a) and \(r=p\) in Case (b). Then, by Hölder’s inequality and (7.15), we have
For Cases (c) and (d) below, we define \(\zeta _j\) as in (7.10).
Case (c) Let \(q_j>0\) and \(\zeta _j\le 0\). Defining \(\gamma _j=1-q_j/p_j\), similar to (7.11), we get
Case (d) Let \(q_j>0\) and \(\zeta _j>0\). We have three sub-cases, see (d\(_1\))–(d\(_3\)) in Lemma 7.1.
(d\(_1\)) Let \({\mathfrak {m}}_j>1\) and \(\theta _j\ge p\). Defining \(\gamma _j=(m-\theta _j)/(m-1)\), similar to (7.12), we see that
We treat the remaining sub-cases (d\(_2\)) and (d\(_3\)) together and define \(\gamma _j\) as in (7.13). Analogous to (7.14), we find that
By (7.15), the right-hand side of each of the inequalities in (7.26), (7.27) and (7.28) converges to 0 as \(n\rightarrow \infty \). So, in any of the Cases (a)–(d), we get (7.25) for \(j\in J_1\). With the same reasoning, we obtain that \(\lim _{n\rightarrow \infty } R_j(n,k)= 0\) for every \(j\in J_1\). Using that \((u_n-T_k(u_0)) \chi _{\{ 0<|u_n|< k\}}\rightarrow 0\) in \(L^{(p_j/q_j)'}(\Omega )\) as \(n\rightarrow \infty \), we find that \(V_j(n,k)\rightarrow 0\) as \(n\rightarrow \infty \). Thus, the right-hand side of (7.24) converges to 0 as \(n\rightarrow \infty \). The proof of (7.23) is complete. \(\square \)
Proof of Proposition 7.4
Using Lemma A.5 in [16], to obtain (7.20), it suffices to show that for every integer \(k\ge 1\), there exists a subsequence of \(\{u_n\}\) (depending on k and relabeled \(\{u_n\}\)) such that (1.28) holds. We first note that, as \(n\rightarrow \infty \),
Indeed, from (6.2) and (7.22), we have
For all \(1\le j\le N\), since \(\partial _j T_k(u_n)=\chi _{\{|u_n|<k\}}\partial _j u_n \), the Dominated Convergence Theorem yields
(strongly) in \(L^{p_j}(\Omega )\) as \( n\rightarrow \infty \).
Similar to (6.6), by passing to a subsequence of \(\{u_n\}\), for each \(1\le j\le N\), we see that \(\{ E_{j,u_n}(T_k(u_n),T_k(u_0))\}_{n}\) converges weakly in \(L^{p_j'}(\Omega )\) as \(n\rightarrow \infty \). Hence, we obtain (7.29).
Using the monotonicity assumption in (1.12), we get that
Hence, in view of (7.29), to conclude (1.28), it remains to show that, up to a subsequence,
We set
where \(v\in W_0^{1,\overrightarrow{p}}(\Omega )\). Since \(T_k(u_n)=u_n\) on \(\{|u_n|<k\}\), from (6.25) and the definition of \({\mathcal {E}}_u\) in (1.21), we infer that
The proof of (7.31) follows now by establishing that
Since \(|Z_{n,k}|\le 2k\), we find a constant \(C_k>0\) such that \( |f_\lambda (n,k)| \le C_k\) for all \(n\ge 1\). For arbitrary \(1\le j\le N\), with the same reasoning as for (6.18), we have that \(A_j(x,u_n,\nabla T_k(u_0)) \,\chi _{\{|u_n|\le k\}}\) converges to \(A_j(x,u_0,\nabla T_k(u_0)) \,\chi _{\{|u_0|\le k\}}\) (strongly) in \(L^{p_j'}(\Omega )\) as \(n\rightarrow \infty \). Hence, using that \(\partial _j Z_{n,k}\rightharpoonup 0\) (weakly) in \(L^{p_j}(\Omega )\) as \(n\rightarrow \infty \), we find that \( A_j(x,u_n,\nabla T_k(u_0)) \,\chi _{\{|u_n|\le k\}} \,\partial _j Z_{n,k}\rightarrow 0\) in \( L^1(\Omega )\) as \( n\rightarrow \infty \). Thus, by the squeeze law, we obtain the first limit in (7.33).
To prove (ii) in (7.33), we take as a test function in (7.3) the function
Compared with [16], we have the extra term \(I_{u_n}(v)\) in the right-hand side of (7.3). Then, \(I_{u_n} (\varphi _\lambda (Z_{n,k}))\) is the additional term which appears when bounding from above \({\mathcal {F}}_{n,k}(u_n)\). By following the ideas in the proof of Lemmata 3.2 and 4.2 in [16] (see Lemma A.6 in the “Appendix” for details), we arrive at
where, up to a subsequence of \(\{u_n\}\), \(\lim _{n\rightarrow \infty } S_k(n)= 0\). From Lemma 7.6 and (7.35), we conclude (ii) in (7.33). This ends the proof of Proposition 7.4. \(\square \)
7.3 Proof of Theorem 1.5 concluded
Here, we obtain that \(u_0\in W_0^{1,\overrightarrow{p}}(\Omega ) \) is a solution of (1.11) by combining Propositions 7.3 and 7.4 with Lemma 7.7 below.
Lemma 7.7
Let \(u_n\) and \(u_0\) be as in Proposition 7.3. Then, up to a subsequence, we have
for every \(v\in W_0^{1,\overrightarrow{p}}(\Omega )\cap L^\infty (\Omega )\).
Proof
We start by proving the second equality in (7.36). Let \(v\in W_0^{1,\overrightarrow{p}}(\Omega )\cap L^\infty (\Omega )\) be arbitrary. From (7.1), we have
Reasoning as in the proof of (3.1), we obtain
\(\lim _{n\rightarrow \infty } \langle {\mathcal {A}} (u_n),v\rangle =\langle {\mathcal {A}} (u_0),v\rangle \) and \(\lim _{n\rightarrow \infty } \langle {\mathfrak {B}} u_n, v\rangle = \langle {\mathfrak {B}} u_0, v\rangle \). Since \(\Phi (u_0) \in L^1(\Omega )\) (see Remark 7.5), it is enough to show that
By Vitali’s Theorem, it suffices to show the uniform integrability of \(\{\Phi (u_n)\}_{n\ge 1}\) over \(\Omega \).
Fix \(M>2\) arbitrary. Let \(\omega \) be any measurable subset of \(\Omega \). We regain (3.5) with \(u_n\) instead of \(U_n\). However, the proof of (3.6) does not translate here since from Proposition 7.3 we only have the uniform boundedness in \(L^1(\Omega )\) for \(\{\Phi (u_n)\}_{n\ge 1}\) (rather than for \(\{\Phi (u_n)\,u_n\}_{n\ge 1}\)). The case \(\Psi =0\) is treated in [16, Lemma 4.3] by adapting and extending to the anisotropic case an approach from [11]. We give the details since compared with [16] we need to deal with the new term \(\Psi \) in (1.3). In (7.3), we take
Then, using the coercivity condition in (1.12) and (1.13), we obtain the estimate
Now, up to a subsequence of \(\{u_n\}\), from (7.16), we have
Using this in (7.38), jointly with (7.1) and the property \((P_2)\) for \({\mathfrak {B}}\), we find that
Since \(T_1(G_{M-1}(u_n))=0\) on \(\{|u_n|\le M-1\}\), we have
Let \(\mu _{M-1}(v):=\text{ meas }\,\{|v|\ge M-1\}\). We next bound from above the right-hand side of (7.40).
(I) For every \(j\in J_2\), using that \(M>2\) and \(\theta _j\le 1\), we find that
(II) Let \(j\in J_1\) corresponding to \(\theta _j>1\). We are guided by the reasoning in Lemma 7.6. In relation to the upper bound for \(Q_j(n,k)\) in the proof of (7.25), we replace \((k-T_k(u_0)) \chi _{\{u_n\ge k\}}\) by \(\chi _{\{|u_n|\ge M-1\}}\). Hence, using also (7.15), we obtain a positive constant C, independent of n and M, such that
where \(\gamma _j\in (0,1)\) is defined according to (a)–(d) in the proof of (7.25).
In light of (7.41) and (7.42), we infer from (7.40) that
where \(C>0\) is a constant independent of M. As \(\mu _{M-1}(u_0)\) converges to 0 as \(M\rightarrow \infty \), by choosing \(M>2\) large, we can make \( \limsup _{n\rightarrow \infty } |I_{u_n} (T_1(G_{M-1}(u_n)) )|\) as small as desired. Using this fact in (7.39), we conclude that \( \int _\omega |\Phi (u_n)| \chi _{\{ |u_n|>M\}}\,dx \) is small uniformly in n and \(\omega \). This finishes the proof of (7.37).
We now establish the first equality in (7.36) for every \(v \in W_0^{1,\overrightarrow{p}}(\Omega )\cap L^\infty (\Omega )\). We follow the ideas in the proof of (3.7), working here with \(\Psi \), \(u_n\) and \(u_0\) instead of \(\Psi _n\), \(U_n\) and \(U_0\), respectively. Hence, for every \(j\in J_1\), the reader should replace \(H_{j,n}(U_n,\partial _j U_n)\) by \(|u_n|^{\theta _j-2} u_n |\partial _j u_n|^{q_j}\).
For every \(j\in J_1\), corresponding to (3.13), we want to show that there exists \(s_j>1\) such that
for a positive constant C independent of n. We need to adjust the argument in Sect. 3.1. The reason is that instead of \(\{\Vert u_n\Vert _{L^m(\Omega )}\}_{n\ge 1}\) and \(\{{\mathfrak {I}}_{m,p_j}(u_n)\}_{n\ge 1}\) being uniformly bounded in n, we only have that \(\{\int _{\Omega } |u_n|^{m-1}\,dx\}_{n\ge 1}\) and \(\{{\mathfrak {I}}_{m-1,p_j}(u_n)\}_{n\ge 1}\) are uniformly bounded (see Proposition 7.3 (a)). With similar ideas to those given in the proof of (7.25), based on Hölder’s inequality, we obtain (7.43) by taking \(s_j=1/(1-\gamma _j)\), where \(\gamma _j\in (0,1)\) is defined as for (7.42).
Now, we use Proposition 7.3 (b) and (7.20) to deduce that
Using (7.43), we infer that, up to a subsequence,
as \(n\rightarrow \infty \), proving that
As mentioned before, for \(w\in W_0^{1,\overrightarrow{p}}(\Omega )\cap L^\infty (\Omega )\), we have \(\nabla w=0\) a.e. in \(\{w=0\}\). Hence, the above identity holds if instead of \(\Omega \) we put \(\{|u_n|>0\}\) in the left-hand side of (7.44) and \(\{|u_0|>0\}\) in the right-hand side. This completes the proof of (7.36) in Case 1.
In Case 2, the proof of (7.36) adapts almost verbatim from Sect. 3.2 remembering to work with \(\Psi \) instead of \(\Psi _n\). This ends the proof of Lemma 7.7.
\(\square \)
Data Availability
Data sharing is not applicable to this article as no new data were created or analyzed in this study.
References
Abdellaoui, B., Dall’Aglio, A., Peral, I.: Some remarks on elliptic problems with critical growth in the gradient. J. Differ. Equ. 222, 21–62 (2006), J. Differ. Equ. 246(7), 2988–2990 (2009) (Corrigendum)
Abdellaoui, B., Giachetti, D., Peral, I., Walias, M.: Elliptic problems with nonlinear terms depending on the gradient and singular on the boundary. Nonlinear Anal. 74(4), 1355–1371 (2011)
Alberico, A., Chlebicka, I., Cianchi, A., Zatorska-Goldstein, A.: Fully anisotropic elliptic problems with minimally integrable data. Calc. Var. Partial Differ. Equ. 58(6), 186 (2019)
Alberico, A., di Blasio, G., Feo, F.: A priori estimates for solutions to anisotropic elliptic problems via symmetrization. Math. Nachr. 290(7), 986–1003 (2017)
Alberico, A., di Blasio, G., Feo, F.: A symmetrization result for a class of anisotropic elliptic problems. J. Math. Sci. 224, 607–617 (2017)
Alvino, A., Ferone, V., Mercaldo, A.: Sharp a priori estimates for a class of nonlinear elliptic equations with lower order terms. Ann. Mat. Pura Appl. (4) 194(4), 1169–1201 (2015)
Alvino, A., Mercaldo, A.: Nonlinear elliptic equations with lower order terms and symmetrization methods. Boll. Un. Mat. Ital. 1(3), 645–661 (2008)
Arcoya, D., Boccardo, L., Leonori, T., Porretta, A.: Some elliptic problems with singular natural growth lower order terms. J. Differ. Equ. 249(11), 2771–2795 (2010)
Bensoussan, A., Boccardo, L., Murat, F.: On a nonlinear partial differential equation having natural growth terms and unbounded solution. Ann. Inst. H. Poincaré Anal. Non Linéaire 5(4), 347–364 (1988)
Betta, M.F., Mercaldo, A., Murat, F., Porzio, M.M.: Existence of renormalized solutions to nonlinear elliptic equations with a lower-order term and right-hand side a measure. J. Math. Pures Appl. 81(6), 533–566 (2002), J. Math. Pures Appl. 82(1), 90–124 (2003) (Corrigendum)
Boccardo, L., Gallouët, T., Murat, F.: A unified presentation of two existence results for problems with natural growth. Progress in partial differential equations: the Metz surveys, 2 (1992), Pitman Res. Notes Math. Ser., vol. 296, 127–137, Longman Sci. Tech., Harlow, 1993
Boccardo, L., Murat, F., Puel, J.-P.: Existence of bounded solutions for nonlinear elliptic unilateral problems. Ann. Mat. Pura Appl. 4(152), 183–196 (1988)
Boccardo, L., Orsina, L.: Semilinear elliptic equations with singular nonlinearities. Calc. Var. Partial Differ. Equ. 37(3–4), 363–380 (2010)
Bottaro, G., Marina, M.E.: Problema di Dirichlet per equazioni ellittiche di tipo variazionale su insiemi non limitati. Boll. Un. Mat. Ital. 8, 46–56 (1973)
Brandolini, B., Chiacchio, F., Trombetti, C.: Symmetrization for singular semilinear elliptic equations. Ann. Mat. Pura Appl. (4) 193(2), 389–404 (2014)
Brandolini, B., Cîrstea, F.C.: Anisotropic elliptic equations with gradient-dependent lower order terms and \(L^1\) data. Math. Eng. 5(4), 1–33 (2023)
Brandolini, B., Ferone, V., Messano, B.: Existence and comparison results for a singular semilinear elliptic equation with a lower order term. Ric. Mat. 63(1, suppl.), S3–S18 (2014)
Capuzzo Dolcetta, I., Leoni, F., Porretta, A.: Hölder estimates for degenerate elliptic equations with coercive Hamiltonians. Trans. Am. Math. Soc. 362(9), 4511–4536 (2010)
Carmona, J., Leonori, T., López-Martínez, S., Martínez-Aparicio, P.J.: Quasilinear elliptic problems with singular and homogeneous lower order terms. Nonlinear Anal. 179, 105–130 (2019)
Chrif, M., El Manouni, S., Mokhtari, F.: Strongly anisotropic elliptic problems with regular and \(L^1\) data. Port. Math. 72(4), 357–391 (2015)
Cianchi, A.: Symmetrization in anisotropic elliptic problems. Commun. Partial Differ. Equ. 32(4–6), 693–717 (2007)
Cîrstea, F.C., Vétois, J.: Fundamental solutions for anisotropic elliptic equations: existence and a priori estimates. Commun. Partial Differ. Equ. 40(4), 727–765 (2015)
Crandall, M.G., Rabinowitz, P.H., Tartar, L.: On a Dirichlet problem with a singular nonlinearity. Commun. Partial Differ. Equ. 2(2), 193–222 (1977)
Del Vecchio, T., Porzio, M.M.: Existence results for a class of non-coercive Dirichlet problems. Ric. Mat. 44(2), 421–438 (1995/1996)
di Blasio, G., Feo, F., Zecca, G.: Regularity results for local solutions to some anisotropic elliptic equations. Israel J. Math. (to appear)
Di Castro, A.: Anisotropic elliptic problems with natural growth terms. Manuscr. Math. 135(3–4), 521–543 (2011)
Feo, F., Vazquez, J.L., Volzone, B.: Anisotropic \(p\)-Laplacian evolution of fast diffusion type. Adv. Nonlinear Stud. 21(3), 523–555 (2021)
Ferone, V., Messano, B.: Comparison and existence results for classes of nonlinear elliptic equations with general growth in the gradient. Adv. Nonlinear Stud. 7(1), 31–46 (2007)
Ferone, V., Murat, F.: Nonlinear elliptic equations with natural growth in the gradient and source terms in Lorentz spaces. J. Differ. Equ. 256(2), 577–608 (2014)
Fragalà, I., Gazzola, F., Kawohl, B.: Existence and nonexistence results for anisotropic quasilinear elliptic equations. Ann. Inst. H. Poincaré Anal. Non Linéaire 21(5), 715–734 (2004)
Fragalà, I., Gazzola, F., Lieberman, G.: Regularity and nonexistence results for anisotropic quasilinear elliptic equations in convex domains. Discrete Contin. Dyn. Syst. suppl., 280–286 (2005)
Gao, H., Leonetti, F., Ren, W.: Regularity for anisotropic elliptic equations with degenerate coercivity. Nonlinear Anal. 187, 493–505 (2019)
Giachetti, D., Martínez-Aparicio, P.J., Murat, F.: A semilinear elliptic equation with a mild singularity at \(u=0\): existence and homogenization. J. Math. Pures Appl. 107(1), 41–77 (2017)
Giachetti, D., Martínez-Aparicio, P.J., Murat, F.: Definition, existence, stability and uniqueness of the solution to a semilinear elliptic problem with a strong singularity at \(u=0\). Ann. Sc. Norm. Super. Pisa Cl. Sci. 18(4), 1395–1442 (2018)
Giachetti, D., Murat, F.: An elliptic problem with a lower order term having singular behaviour. Boll. Un. Mat. Ital. 2(2), 349–370 (2009)
Giachetti, D., Petitta, F., Segura de León, S.: Elliptic equations having a singular quadratic gradient term and a changing sign datum. Commun. Pure Appl. Anal. 11(5), 1875–1895 (2012)
Giachetti, D., Petitta, F., Segura de León, S.: A priori estimates for elliptic problems with a strongly singular gradient term and a general datum. Differ. Integral Equ. 26(9–10), 913–948 (2013)
Grenon, N., Murat, F., Porretta, A.: A priori estimates and existence for elliptic equations with gradient dependent terms. Ann. Sc. Norm. Super. Pisa Cl. Sci. 13(1), 137–205 (2014)
Lazer, A.C., McKenna, P.J.: On a singular nonlinear elliptic boundary-value problem. Proc. Am. Math. Soc. 111(3), 721–730 (1991)
Leggat, A.R., Miri, S.: Anisotropic problem with singular nonlinearity. Complex Var. Elliptic Equ. 61(4), 496–509 (2016)
Leray, J., Lions, J.-L.: Quelques résultats de Višik sur les problèmes elliptiques nonlinéaires par les méthodes de Minty-Browder. Bull. Soc. Math. France 93, 97–107 (1965)
Oliva, F., Petitta, F.: Finite and infinite energy solutions of singular elliptic problems: existence and uniqueness. J. Differ. Equ. 264(1), 311–340 (2018)
Stuart, C.A: Existence and approximation of solutions of nonlinear elliptic problems. Mathematics Report, Battelle Advanced Studies Center, Geneva, Switzerland, vol. 86 (1976)
Troisi, M.: Teoremi di inclusione per spazi di Sobolev non isotropi. Ric. Mat. 18, 3–24 (1969)
Vétois, J.: Strong maximum principles for anisotropic elliptic and parabolic equations. Adv. Nonlinear Stud. 12(1), 101–114 (2012)
Zeidler, E.: Nonlinear Functional Analysis and its Applications. Nonlinear Monotone Operators. Springer, New York (1990)
Zhang, Z.: Two classes of nonlinear singular Dirichlet problems with natural growth: existence and asymptotic behavior. Adv. Nonlinear Stud. 20(1), 77–93 (2020)
Acknowledgements
The authors would like to thank the anonymous referee for helping improve the presentation.
Funding
Open Access funding enabled and organized by CAUL and its Member Institutions. The first author has been supported by the Sydney Mathematical Research Institute International Visitor Program (August– September 2019), by Programma di Scambi Internazionali dell’Universitá degli Studi di Napoli Federico II and by the grant “FFR 2023 Barbara Brandolini”, Università degli Studi di Palermo. The research of the second author is supported by The Australian Research Council under the Discovery Project Scheme (DP190102948 and DP220101816).
Author information
Authors and Affiliations
Contributions
Both authors wrote and reviewed the manuscript.
Corresponding author
Ethics declarations
Conflict of interest
The authors declare that they have no conflict of interest.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix A
Appendix A
In this paper, we need the following version of Young’s inequality.
Lemma A.1
(Young’s inequality) Let \(N\ge 2\) be an integer. Assume that \(\beta _1,\ldots , \beta _N\) are positive numbers and \(1<R_k<\infty \) for each \(1\le k\le N-1\). If \(\sum _{k=1}^{N-1} (1/R_k)<1\), then for every \(\delta >0\), there exists a positive constant \(C_\delta \) (depending on \(\delta \)) such that
where we define \(R_N=\left[ 1-\sum _{k=1}^{N-1} (1/R_k)\right] ^{-1}\).
We recall the anisotropic Sobolev inequality in [44, Theorem 1.2].
Lemma A.2
Let \(N\ge 2\) be an integer. If \(1< p_j <\infty \) for every \(1\le j \le N\) and \(p<N\), then there exists a positive constant \({\mathcal {S}}={\mathcal {S}}(N,\overrightarrow{p})\), such that
where, as usual, \(p^*:=Np/(N-p)\).
Remark A.3
Let \(\Omega \) be a bounded, open subset of \({\mathbb {R}}^N\) with \(N\ge 2\). If \(1< p_j <\infty \) for every \(1\le j \le N\) and \(p<N\), then by a density argument, (A.1) extends to all \(u\in W_0^{1,\overrightarrow{p}}(\Omega ) \) so that the arithmetic–geometric mean inequality yields
for all \(u\in W_0^{1,\overrightarrow{p}}(\Omega )\). Moreover, using Hölder’s inequality, the embedding \(W_0^{1,\overrightarrow{p}}(\Omega )\hookrightarrow L^s(\Omega )\) is continuous for every \(s\in [1,p^*]\) and compact for every \(s\in [1,p^*)\).
Lemma A.4
Let the assumptions of Proposition 2.3 hold. Suppose that
Then, up to a subsequence, we have
Proof
From (A.3), we see that, up to a subsequence,
We prove (A.4). Recall from Proposition 2.3 that (2.10) holds. Let Z be a subset of \(\Omega \) with \(\text{ meas }\,(Z)=0\) such that for every \(x\in \Omega \setminus Z\), we have \(|U_0^\pm (x)|<\infty \), \(|\nabla U_0^\pm (x)|<\infty \), \(|\eta _j(x)|<\infty \) for all \(1\le j\le N\), as well as
We fix \(x\in \Omega \setminus Z\). By the monotonicity and coercivity assumptions in (1.12), we find that
By Young’s inequality, for every \(\delta >0\), there exists \(C_\delta >0\) such that
for every \(1\le j\le N\). We use the growth condition in (1.12) to bound from above the right-hand side of each inequality in (A.8). Then, there exist positive constants C and \(C'_\delta \), both independent of n (only \(C'_\delta \) depends on \(\delta \)) such that \( {\mathcal {E}}_{U_n}(U_n^\pm ,U_0^\pm )(x)\) is bounded below by
Using (A.6) and choosing \(\delta \in (0,\nu _0/C)\), we conclude that
Let \(x\in \Omega \setminus Z\) be arbitrary. To prove that \(\nabla U_n^+(x)\rightarrow \nabla U_0^+(x)\) as \(n\rightarrow \infty \), we show that any accumulation point \(\Xi \) of \(\{\nabla U_n^+(x)\}_{n}\) coincides with \(\nabla U_0^+(x)\). From (A.9), we have \(|\Xi |<\infty \). By (A.6) and the continuity of \( A_j(x,\cdot ,\cdot )\) with respect to the last two variables, we get that \({{\mathcal {E}}}_{U_n}(U_n^+,U_0^+)(x)\) converges to
as \( n\rightarrow \infty \). This, jointly with (A.6) and the monotonicity condition in (1.12), gives that \(\Xi =\nabla U_0^+(x)\). Similarly, we obtain that
The proof of (A.4) is complete since \(x\in \Omega \setminus Z\) is arbitrary and \(\text{ meas }\,(Z)=0\).
In order to prove (A.5), we use (A.4), (A.7), Lemma 5.2 and Vitali’s Theorem. We see that \(\{|\partial _j U_n^\pm - \partial _j U_0^\pm |^{p_j}\}_n\) is a sequence of non-negative integrable functions, converging to 0 a.e. on \(\Omega \) as \(n\rightarrow \infty \). So, we conclude (A.5) by showing that, up to a subsequence, \( \left\{ \sum _{j=1}^N |\partial _j U_n^\pm |^{p_j}\right\} _{n}\) is uniformly integrable over \( \Omega \). Now, up to a subsequence, we have for each \(1\le j\le N\),
as \( n\rightarrow \infty \). Indeed, \(\{A_j(x,U_n,\nabla U_n^\pm )\}_n\) is bounded in \(L^{p_j'}(\Omega )\) from the growth condition in (1.12) and the boundedness of \(\{U_n\}_{n}\) in \(W_0^{1,\overrightarrow{p}}(\Omega )\) and, hence, in \(L^{p^*}(\Omega )\). Moreover, \(\{A_j(x,U_n,\nabla U_n^\pm )\}_{n}\rightarrow A_j(x,U_0,\nabla U_0^\pm )\) a.e. in \( \Omega \) as \(n\rightarrow \infty \) using (A.4), the convergence \(U_n\rightarrow U_0\) a.e. in \(\Omega \) (from (2.10)) and the continuity of \(A_j(x,\cdot ,\cdot )\) in the last two variables. Hence, up to a subsequence, we have (A.10). Consequently, for each \(1\le j\le N\), we get
as \(n\rightarrow \infty \).
Let \(k\ge 1\) be arbitrary. For each \(1\le j\le N\), we next prove that, as \(n\rightarrow \infty \),
Fix \(1\le j\le N\) arbitrary. Note that \(\{ |A_j(x,U_n,\nabla U_0^\pm )|^{p_j'} \chi _{\{|U_n|\le k\}} \!\}_n \) is uniformly integrable over \(\Omega \) and
a.e. in \(\Omega \) as \(n\rightarrow \infty \). Thus, by Vitali’s Theorem,
as \(n\rightarrow \infty \). This proves (A.12) since \(\partial _j U_n^\pm \rightharpoonup \partial _j U_0^\pm \) (weakly) in \(L^{p_j}(\Omega )\) as \(n\rightarrow \infty \) (see Remark 4.2).
By Hölder’s inequality, we get a constant \(C>0\) (independent of k) such that for all \(n\ge 1\),
Using (A.13), jointly with Lemma 5.2, we get that, up to a subsequence,
for each \(k\ge 1\), where \(\lim _{k\rightarrow \infty } W_k=0\). Using (A.3), (A.11), (A.12) and (A.14), from (A.7), we get that \( \left\{ \sum _{j=1}^N |\partial _j U_n^\pm |^{p_j}\right\} _{n}\) is uniformly integrable over \( \Omega \). This ends the proof of (A.5). \(\square \)
Remark A.5
We need Lemma 5.2 to control the integral in the left-hand side of (A.13). Indeed, we cannot conclude the convergence \( A_j(x,U_n,\nabla U_0^\pm ) \,\partial _j U_n^\pm \rightarrow A_j(x,U_0,\nabla U_0^\pm ) \,\partial _j U_0^\pm \) in \(L^1(\Omega )\) as \(n\rightarrow \infty \) for the same reason as in Remark 6.3.
Lemma A.6
In the framework of Theorem 1.5, we have (7.35).
Proof
Recall that \(Z_{n,k}=T_k(u_n)-T_k(u_0)\). From (7.16), we have \(Z_{n,k}\rightarrow 0\) a.e. in \(\Omega \) as \(n\rightarrow \infty \) and \(Z_{n,k} \rightharpoonup 0\) (weakly) in \( W_0^{1,\overrightarrow{p}}(\Omega )\) as \(n\rightarrow \infty \). Moreover, as \(n\rightarrow \infty \), we find that
Observe that \(u_n\, Z_{n,k}\ge 0\) on the set \(\{|u_n|\ge k\}\), which gives that \( \Phi (u_n) \,\varphi _\lambda (Z_{n,k}) \chi _{\{|u_n|\ge k\}}\ge 0\). Thus, by testing (7.3) with
we obtain that
where \(\ell _{n,k}\) is defined by
The first term in \(\ell _{n,k}\) converges to 0 as \(n\rightarrow \infty \) from (7.16), (A.15) and the property \((P_2)\) of \({\mathfrak {B}}\). Since \(|\Theta (u_n)|\le C_\Theta \) and (7.1) holds, by the Dominated Convergence Theorem, we get that the second, as well as the third term in \(\ell _{n,k}\), converges to 0 as \(n\rightarrow \infty \). Hence, \(\lim _{n\rightarrow \infty } \ell _{n,k}=0\).
To simplify exposition, we now introduce some notation:
We rewrite the first term in the left-hand side of (A.16) as follows
The coercivity condition in (1.12) and the growth condition of \(\Phi \) in (1.13) imply that
In the right-hand side of (A.18) we replace \(\partial _j u_n\) by \(\partial _j Z_{n,k}+\partial _j T_k (u_0)\), then we multiply the inequality by \(| \varphi _\lambda (Z_{n,k})|\) and integrate over \(\Omega \) with respect to x. It follows that the second term in the left-hand side of (A.16) is at least
Using this fact, as well as (A.17), in (A.16), we see that \({\mathcal {F}}_{n,k}(u_n)\) (defined in (7.32)) satisfies
Since \(\lim _{n\rightarrow \infty } \ell _{n,k}=0\), by showing that
we conclude (7.35). Using (A.15) and \(c\in L^1(\Omega )\), we infer from the Dominated Convergence Theorem that
Next, up to a subsequence of \(\{u_n\}\), \(A_j(u_n) \) converges weakly in \(L^{p_j'}(\Omega )\) as \(n\rightarrow \infty \) for all \(1\le j\le N\). Hence, \( \sum _{j=1}^N A_j(u_n)\,\partial _j u_0 \) converges in \(L^1(\Omega )\) as \(n\rightarrow \infty \). Then, there exists a non-negative function \(F\in L^1(\Omega )\) (independent of n) such that, up to a subsequence of \(\{u_n\}\), \( \left| \sum _{j=1}^N A_j(u_n)\,\partial _j u_0\right| \le F \) a.e. in \( \Omega \) for all \(n\ge 1\). By the Dominated Convergence Theorem, as \(n\rightarrow \infty \),
From (A.20) and (A.21), we find that \(\lim _{n\rightarrow \infty } X_k(n)=0\). Remark that \(|\varphi _\lambda '(Z_{n,k})|\) is bounded above by a constant independent of n and \(\chi _{\{|u_0|<k\}}\, \chi _{\{|u_n|\ge k\} }\rightarrow 0\) a.e. in \(\Omega \) as \(n\rightarrow \infty \). Hence, we can use a similar argument as for \(X_k(n)\), to obtain that, up to a subsequence of \(\{u_n\}\), \(\lim _{n\rightarrow \infty } Y_k(n)=0\). From (A.19), we conclude (7.35) with \(S_k(n)=X_k(n)+Y_k(n)+\ell _{n,k}\). \(\square \)
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Brandolini, B., Cîrstea, F.C. Singular anisotropic elliptic equations with gradient-dependent lower order terms. Nonlinear Differ. Equ. Appl. 30, 58 (2023). https://doi.org/10.1007/s00030-023-00864-w
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s00030-023-00864-w