Nonlinear Analysis 180 (2019) 154–169

Contents lists available at ScienceDirect

Nonlinear Analysis
www.elsevier.com/locate/na

Moser iteration applied to elliptic equations with critical growth

on the boundary
Greta Marino a ,∗, Patrick Winkert b
a
Dipartimento di Matematica e Informatica, Università degli Studi di Catania, Viale A. Doria 6, I-95125
Catania, Italy
b
Technische Universität Berlin, Institut für Mathematik, Straße des 17. Juni 136, 10623
Berlin, Germany

article info abstract

Article history: This paper deals with boundedness results for weak solutions of the equation
Received 30 April 2018
Accepted 3 October 2018 − div A(x, u, ∇u) = B(x, u, ∇u) in Ω ,
Communicated by S. Carl (P)
A(x, u, ∇u) · ν = C(x, u) on ∂Ω ,
MSC:
35J60 where the functions A : Ω × R × RN → RN , B : Ω × R × RN → R, and
35B45 C : ∂Ω ×R → R are Carathéodory functions satisfying certain p-structure conditions
35J25 that have critical growth even on the boundary. Based on a modified version of the
Keywords: Moser iteration we are able to prove that every weak solution of (P) is bounded up
Moser iteration to the boundary. Under some additional assumptions on the functions A and C this
Boundedness of solutions leads directly to regularity for weak solutions of (P).
A priori bounds © 2018 Elsevier Ltd. All rights reserved.
Elliptic operators of divergence type
Critical growth on the boundary

1. Introduction

Let Ω ⊂ RN , N > 1, be a bounded domain with a Lipschitz boundary ∂Ω . In this paper, we study the
boundedness of weak solutions of the problem

− div A(x, u, ∇u) = B(x, u, ∇u) in Ω ,

(1.1)
A(x, u, ∇u) · ν = C(x, u) on ∂Ω ,

where ν(x) denotes the outer unit normal of Ω at x ∈ ∂Ω , and A, B and C satisfy suitable p-structure
conditions, see hypotheses (H) in Section 3.
The main goal of this paper is to present a priori bounds for weak solutions of Eq. (1.1), where we allow
critical growth on the data in the domain and on the boundary. The main idea in the proof is based on a

∗ Corresponding author.
E-mail addresses: greta.marino@dmi.unict.it (G. Marino), winkert@math.tu-berlin.de (P. Winkert).

https://doi.org/10.1016/j.na.2018.10.002
0362-546X/© 2018 Elsevier Ltd. All rights reserved.
 G. Marino and P. Winkert / Nonlinear Analysis 180 (2019) 154–169 155

modified version of Moser’s iteration which in turn is based on the books of Drábek–Kufner–Nicolosi [2] and
Struwe [17].
In some sense, (1.1) is a generalization of the classical differential equation from the Yamabe problem
N +2
− ∆u = f (x)u + h(x)u N −2 , (1.2)

where f and h are smooth functions. It is well known that there is no stable regularity theory for solutions
of Eq. (1.2), which reflects the difficulty of the Yamabe problem. Nevertheless, it was proven by Trudinger [18]
that any solution W 1,2 (Ω ) of (1.2) is in fact smooth, but the regularity estimates depend on the solution
itself. In this spirit, our main result, Theorem 3.1, can thus be seen as a generalization of Trudinger’s work.
The main novelty of our paper consists in the generality of the assumptions needed to establish the
boundedness of weak solutions to (1.1). In particular, the assumptions on the nonlinearity C are rather general
allowing critical growth on the boundary. To the best of our knowledge, such a treatment with critical growth
even on the boundary has not been published before. Another novelty is a result of independent interest
which shows that a Sobolev function, which is bounded in the domain, is also bounded on the boundary,
see Proposition 2.4.
Recently, Papageorgiou–Rădulescu [14, Proposition 2.8] studied a priori bounds for problems of the form

− div a(∇u) = f0 (x, u) in Ω ,

p−2 (1.3)
a(∇u) · ν = −β|u| u on ∂Ω ,

where 1 < p < ∞, the function a : RN → RN is continuous and strictly monotone satisfying certain
regularity and growth conditions, the Carathéodory function f0 : Ω × R → R has critical growth with
respect to the second variable and β ∈ C 1,α (∂Ω ) with α ∈ (0, 1) and β ≥ 0. Note that our setting is more
general than those in [14] since we have weaker conditions on a and f0 and our boundary term is able to have
critical growth. The proof of their result is mainly based on a treatment of Garcı́a Azorero–Peral Alonso
[4], who studied Eq. (1.3) with the p-Laplacian and homogeneous Dirichlet condition, namely
q−2 p∗ −2
−∆p u = λ|u| u + |u| u in Ω ,
u=0 on ∂Ω ,

with 1 < q < p < N , λ > 0 and the Sobolev critical exponent p∗ , see Section 2 for its definition. Both works
use a different technique than the Moser iteration applied in our paper. For the semilinear case we mention
the work of Wang [19].
An alternative approach was published by Guedda–Véron [8] who studied quasilinear problems for positive
solutions given by
∗
−∆p u = a(x)up−1 + up −1 in Ω ,
u>0 in Ω ,
u=0 on ∂Ω ,
with a ∈ L∞ (Ω ). In all these works the assumptions on the functions are stronger than ours and no critical
growth on the boundary is allowed.
Finally, we mention some works concerning boundedness and regularity results of weak solutions to
quasilinear equations of the form (1.1) that have subcritical growth, see, for example, Fan–Zhao [3], Gasiński–
Papageorgiou [6], [5, pp. 737–738], Hu–Papageorgiou [9], Lê [10], Motreanu–Motreanu–Papageorgiou [12],
Pucci–Servadei [16], Winkert [20–22], Winkert–Zacher [24,23,25] and the references therein. The methods
used in these papers are mainly based on Moser’s iteration or De Giorgi’s iteration technique and no critical
growth occurs.
156 G. Marino and P. Winkert / Nonlinear Analysis 180 (2019) 154–169

The paper is organized as follows. In Section 2 we present the main preliminaries including a multiplicative
inequality estimating the boundary integrals and a result how L∞ (Ω )-boundedness implies L∞ (∂Ω )-
boundedness. In Section 3 we state our main result and the proof is divided into several parts. First we
prove that every weak solution belongs to Lq (Ω ) for any q < ∞, then we show its belonging to Lq (∂Ω )
for any finite q. In the second part of the proof we consider the uniform boundedness and show that a
weak solution belongs to L∞ (Ω ) and L∞ (∂Ω ), respectively. Finally, as an important application, we give
general conditions on the functions A and C when a solution lies in C 1,β (Ω ) for some β ∈ (0, 1) based on
the regularity results of Lieberman [11].

2. Preliminaries
( )
Let r be a number such that 1 ≤ r < ∞. We denote by Lr (Ω ), Lr Ω ; RN and W 1,r (Ω ) the usual
Lebesgue and Sobolev spaces equipped with the norms ∥ · ∥r and ∥ · ∥1,r given by
(∫ ) r1 (∫ ) r1
r r
∥u∥r = |u| dx , ∥∇u∥r = |∇u| dx
Ω Ω
(∫ ∫ ) r1
r r
∥u∥1,r = |∇u| dx + |u| dx .
Ω Ω
∞
For r = ∞ we recall that the norm of L (Ω ) is given by

∥u∥∞ = ess sup|u|.

Ω

On the boundary ∂Ω , we use the (N − 1)-dimensional Hausdorff (surface) measure denoted by σ. Then, in
a natural way we can define the Lebesgue spaces Ls (∂Ω ) with 1 ≤ s ≤ ∞ and the norms ∥ · ∥s,∂Ω which are
given by
(∫ ) 1s
s
∥u∥s,∂Ω = |u| dσ (1 ≤ s < ∞), ∥u∥∞,∂Ω = ess sup|u|.
∂Ω ∂Ω
It is well known that there exists continuous map γ : W (Ω ) → Lp∗ (∂Ω ) known as the
⏐ a unique linear1,p
1,p

trace map such that γ(u) = u⏐∂Ω for all u ∈ W (Ω ) ∩ C(Ω ), where p∗ is the critical exponent on the
boundary given by
⎨ (N − 1)p
⎧
if p < N,
p∗ = N −p (2.1)
any q ∈ (1, ∞) if p ≥ N.
⎩

For the sake of notational simplicity, we drop the use of the trace map γ. It is understood that all restrictions
of the Sobolev functions u ∈ W 1,p (Ω ) on the boundary ∂Ω are defined in the sense of traces.
Furthermore, the Sobolev embedding theorem guarantees the existence of a linear, continuous map
∗
i : W 1,p (Ω ) → Lp (Ω ) with the critical exponent in the domain given by
⎨ Np
⎧
∗ if p < N,
p = N −p (2.2)
any q ∈ (1, ∞) if p ≥ N.
⎩

We refer to Adams [1] as a reference for the embeddings above.

The norm of RN is denoted by | · | and · stands for the inner product in RN . For s ∈ R, we set
s = max{±s, 0} and for u ∈ W 1,p (Ω ) we define u± (·) = u(·)± . It is well known that
±

u± ∈ W 1,p (Ω ), |u| = u+ + u− , u = u+ − u− .

By | · | we denote the Lebesgue measure on RN .

The following proposition will be useful in our treatment and was proven in Winkert [22, Proposition 2.1]
 G. Marino and P. Winkert / Nonlinear Analysis 180 (2019) 154–169 157

Proposition 2.1. Let Ω ⊂ RN , N > 1, be a bounded domain with Lipschitz boundary ∂Ω , let 1 < p < ∞,
and let q̂ be such that p ≤ q̂ < p∗ with the critical exponent stated in (2.1). Then, for every ε > 0, there exist
constants c̃1 > 0 and c̃2 > 0 such that

∥u∥pq̂,∂Ω ≤ ε∥u∥p1,p + c̃1 ε−c̃2 ∥u∥pp for all u ∈ W 1,p (Ω ).

The next proposition is a standard argument in the application of the Moser iteration, see for example
Drábek–Kufner–Nicolosi [2].

Proposition 2.2. Let Ω ⊂ RN , N > 1, be a bounded domain with Lipschitz boundary ∂Ω . Let u ∈ Lp (Ω )
with u ≥ 0 and 1 < p < ∞ such that
∥u∥αn ≤ C (2.3)
with a constant C > 0 and a sequence (αn ) ⊆ R+ with αn → ∞ as n → ∞. Then, u ∈ L∞ (Ω ).

Proof . Let us suppose that u ̸∈ L∞ (Ω ). Then there exist a number η > 0 and a set A of positive measure
in Ω such that u(x) ≥ C + η for x ∈ A. Then it follows
(∫ ) α1
n 1
αn
∥u∥αn ≥ u dx ≥ (C + η) |A| αn .
A

Passing to the limit inferior in the inequality above gives

lim inf ∥u∥αn ≥ C + η,

n→∞

which is a contradiction to (2.3). Hence, u ∈ L∞ (Ω ). □

Remark 2.3. It is clear that the statement in Proposition 2.2 remains true if we replace the domain Ω
by its boundary ∂Ω .

Finally, we state a result that the boundedness of a Sobolev function in W 1,p (Ω ) implies the boundedness
on the boundary.

Proposition 2.4. Let Ω ⊂ RN , N > 1, be a bounded domain with Lipschitz boundary ∂Ω and let
1 < p < ∞. If u ∈ W 1,p (Ω ) ∩ L∞ (Ω ), then u ∈ L∞ (∂Ω ).

Proof . By the Sobolev embedding we have

∥v∥p∗ ,∂Ω ≤ c∂Ω ∥v∥1,p for all v ∈ W 1,p (Ω )

with the critical exponent p∗ as in (2.1). Let κ > 1 and take v = uκ in the inequality above. Note that
v ∈ W 1,p (Ω ) since u ∈ W 1,p (Ω ) ∩ L∞ (Ω ). This gives
[(∫ 1
) κp (∫ 1 ]
) κp
1
κ κ p κ p
∥u∥κp∗ ,∂Ω ≤ c∂Ω |∇u | dx + |u | dx
Ω Ω
1 [ 1 1
1− κ 1
]1
κ κ
≤ c∂Ω κ ∥u∥∞ ∥∇u∥p + ∥u∥∞ |Ω | κp .
κ

Letting κ → ∞, by applying Proposition 2.2 and Remark 2.3, we derive

∥u∥∞,∂Ω ≤ 2∥u∥∞ . □
158 G. Marino and P. Winkert / Nonlinear Analysis 180 (2019) 154–169

3. A priori bounds via Moser iteration

In this section we state and prove our main result. First, we give the structure conditions on the functions
involved in problem (1.1).

(H) The functions A : Ω × R × RN → RN , B : Ω × R × RN → R, and C : ∂Ω × R → R are Carathéodory

functions satisfying the following structure conditions:
p−1
p−1 q1 p
(H1) |A(x, s, ξ)| ≤ a1 |ξ| + a2 |s| + a3 , for a.a. x ∈ Ω ,
p q1
(H2) A(x, s, ξ) · ξ ≥ a4 |ξ| − a5 |s| − a6 , for a.a. x ∈ Ω ,
q −1
p 1q q1 −1
(H3) |B(x, s, ξ)| ≤ b1 |ξ| 1 + b2 |s| + b3 , for a.a. x ∈ Ω ,
q2 −1
(H4) |C(x, s)| ≤ c1 |s| + c2 , for a.a. x ∈ ∂Ω ,
for all s ∈ R, for all ξ ∈ RN , with positive constants ai , bj , ck (i ∈ {1, . . . , 6}, j ∈ {1, 2, 3}, k ∈ {1, 2})
and fixed numbers p, q1 , q2 such that
1 < p < ∞, p ≤ q1 ≤ p∗ , p ≤ q2 ≤ p∗
with the critical exponents stated in (2.2) and (2.1).

A function u ∈ W 1,p (Ω ) is said to be a weak solution of Eq. (1.1) if

∫ ∫ ∫
A(x, u, ∇u) · ∇φdx = B(x, u, ∇u)φdx + C(x, u)φdσ (3.1)
Ω Ω ∂Ω
holds for all test functions φ ∈ W 1,p (Ω ).
∗
By means of the embeddings i : W 1,p (Ω ) → Lp (Ω ) and γ : W 1,p (Ω ) → Lp∗ (∂Ω ) we see that the
definition of a weak solution is well-defined and all integrals in (3.1) are finite for u, φ ∈ W 1,p (Ω ).
Now we can formulate the main result of our paper.

Theorem 3.1. Let Ω ⊂ RN , N > 1, be a bounded domain with Lipschitz boundary ∂Ω and let the
hypotheses (H) be satisfied. Then, every weak solution u ∈ W 1,p (Ω ) of problem (1.1) belongs to Lr (Ω ) for
every r < ∞. Moreover, u ∈ L∞ (Ω ), that is, ∥u∥∞ ≤ M , where M is a constant which depends on the given
data and on u.

Proof . Let u ∈ W 1,p (Ω ) be a weak solution of problem (1.1). Since u = u+ − u− we can suppose, without
any loss of generality, that u ≥ 0. Furthermore, we only prove the case when q1 = p∗ and q2 = p∗ . The other
cases were already obtained in [21, Theorem 4.1] and [22, Theorem 3.1]. Moreover, we will denote positive
constants with Mi and if the constant depends on the parameter κ we write Mi (κ) for i = 1, 2, . . ..
Let h > 0 and set uh = min{u, h}. Then we choose φ = uuκp h with κ > 0 as test function in (3.1). Note
κp κp−1
that ∇φ = ∇uuh + uκpuh ∇uh . This gives
∫ ∫
A(x, u, ∇u) · ∇uuκp
h dx + κp A(x, u, ∇u) · ∇uh uκp−1
h udx
Ω∫ Ω
∫ (3.2)
κp κp
= B(x, u, ∇u)uuh dx + C(x, u)uuh dσ.
Ω ∂Ω
Applying (H2) to the first term of the left-hand side of (3.2) yields
∫
A(x, u, ∇u) · ∇uuκph dx
Ω∫
∗
[ ]
p
≥ a4 |∇u| − a5 up − a6 uκp h dx (3.3)
Ω∫ ∫
p ∗
≥ a4 |∇u| uκph dx − (a5 + a6 ) up uκp
h dx − a6 |Ω |,
Ω Ω
 G. Marino and P. Winkert / Nonlinear Analysis 180 (2019) 154–169 159

respectively to the second term on the left-hand side

∫
κp A(x, u, ∇u) · ∇uh uκp−1
h udx
Ω∫

= κp A(x, u, ∇u) · ∇uuκp

h dx
{x∈Ω:u(x)≤h}
∫ [ ∗
] (3.4)
p
≥ κp a4 |∇u| − a5 up − a6 uκph dx
{x∈Ω:u(x)≤h}
∫ ∫
p κp ∗
≥ a4 κp |∇u| uh dx − κp(a5 + a6 ) up uκp
h dx − κpa6 |Ω |.
{x∈Ω: u(x)≤h} Ω

By means of (H3) combined with Young’s inequality with ε1 > 0, the first term on the right-hand side of
(3.2) can be estimated through
∫
B(x, u, ∇u)uuκp
h dx
Ω ( )
p∗ −1 p∗ −1 p∗ −1 p∗ −1
p∗ −1 κp 1− p∗
∫
p∗ p p∗ κp ∗ − ∗
≤ b1 ε1|∇u| uh p ε1 p udx uh
Ω ∫ (3.5)
∗
+ (b2 + b3 ) up uκp
h dx + b3 |Ω |
∫ Ω
)∫
−(p∗ −1) ∗
(
p κp
≤ ε 1 b1 |∇u| uh dx + b1 ε1 + b2 + b3 up uκp
h dx + b3 |Ω |.
Ω Ω

Finally, the boundary term can be estimated via (H4). This leads to
∫ ∫
C(x, u)uuκp c1 up∗ −1 + c2 uuκp
( )
h dσ ≤ h dσ
∂Ω ∂Ω ∫ (3.6)
≤ (c1 + c2 ) up∗ uκp
h dσ + c2 |∂Ω |.
∂Ω

a4
We now combine (3.2)–(3.6) and choose ε1 = 2b1 to obtain
( ∫ ∫ )
1 p κp p κp
a4 |∇u| uh dx + κp |∇u| uh dx
2 Ω {x∈Ω: u(x)≤h}
)∫
−(p∗ −1) ∗
(
≤ (κp + 1)(a5 + a6 ) + b1 ε1 + b2 + b3 up uκp (3.7)
h dx
∫ Ω

+ (c1 + c2 ) up∗ uκp

h dσ + ((κp + 1)a6 + b3 )|Ω | + c2 |∂Ω |.
∂Ω

Observe that
∫ ∫
1 p p
|∇u| uκp
h dx + κp |∇u| uκp
h dx
2 Ω {x∈Ω: u(x)≤h}
∫ ( )∫
1 p 1 p
= |∇u| uκp h dx + κp + |∇u| uκp
h dx
2 {x∈Ω: u(x)>h} 2 {x∈Ω: u(x)≤h}
∫ ∫
κp + 1 p κp + 1 p
≥ |∇u| uκp
h dx + |∇u| uκp
h dx
2(κ + 1)p {x∈Ω: u(x)>h} 2 {x∈Ω: u(x)≤h}
∫
κp + 1 p
≥ |∇ (uuκh ) | dx
2(κ + 1)p Ω
160 G. Marino and P. Winkert / Nonlinear Analysis 180 (2019) 154–169

invoking Bernoulli’s inequality (κ + 1)p ≥ κp + 1. From (3.7) it follows

∫
κp + 1 p
a4 |∇ (uuκh )| dx
2(κ + 1)p Ω
)∫
−(p∗ −1) ∗
(
≤ (κp + 1)(a5 + a6 ) + b1 ε1 + b2 + b3 up uκph dx (3.8)
∫ Ω

+ (c1 + c2 ) up∗ uκph dσ + ((κp + 1)a6 + b3 ) |Ω | + c2 |∂Ω |.

∂Ω

Dividing by a4 , summarizing the constants and adding on both sides of (3.8) the nonnegative term
κp+1 κ p
(κ+1)p ∥uuh ∥p gives

κp + 1
∥uuκh ∥p1,p
(κ + 1)p
∫ ∫ (3.9)
κp + 1 p∗ κp
≤ κ p
∥uuh ∥p + M1 (κp + 1) u uh dx + M2 up∗ uκp
h dσ + M3 κ.
(κ + 1)p Ω ∂Ω

Part I: u ∈ Lr (Ω ) for any finite r

∗
Let us now estimate the terms on the right-hand side involving the critical exponents. We set a := up −p
and b := up∗ −p . Moreover, let L > 0 and G > 0. Then, by using Hölder’s inequality and the Sobolev
embeddings for p∗ and p∗ , see Section 2, we get
∫
∗
up uκp
h dx
Ω∫ ∫
p κp
= au uh dx + aup uκp
h dx
{x∈Ω: a(x)≤L} {x∈Ω: a(x)>L}
∫
≤L up uκp
h dx
{x∈Ω: a(x)≤L}
(3.10)
) p∗ −p ) p∗
p∗ (∫
(∫
p∗ ∗ p
p∗ −p p∗ κp
+ a dx u uh dx
{x∈Ω: a(x)>L} Ω
(∫ ) p∗ −p
∗
p∗ p
≤ L∥uuκh ∥pp + a p∗ −p dx cpΩ ∥uuκh ∥p1,p
{x∈Ω: a(x)>L}

and
∫
up∗ uκp
h dσ
∂Ω
∫ ∫
= bup uκp
h dσ + bup uκp
h dσ
{x∈∂Ω: b(x)≤G} {x∈∂Ω: b(x)>G}
∫
≤G up uκp
h dσ
{x∈∂Ω: b(x)≤G} (3.11)
(∫ ) p∗p−p (∫ ) pp
∗
p∗ ∗
+ b p∗ −p dσ up∗ uκp ∗
h dσ
{x∈∂Ω: b(x)>G} ∂Ω
(∫ ) p∗p−p
∗
p∗
≤ G∥uuκh ∥pp,∂Ω + b p∗ −p dσ cp∂Ω ∥uuκh ∥p1,p
{x∈∂Ω: b(x)>G}
 G. Marino and P. Winkert / Nonlinear Analysis 180 (2019) 154–169 161

with the embedding constants cΩ and c∂Ω . Note that

(∫ ) p∗ −p
∗
p∗ p
H(L) := a p∗ −p dx →0 as L → ∞,
{x∈Ω: a(x)>L}
(3.12)
(∫ ) p∗p−p
∗
p∗
K(G) := b p∗ −p dσ →0 as G → ∞.
{x∈∂Ω: b(x)>G}

Combining (3.9)–(3.12) finally yields

κp + 1
∥uuκh ∥p1,p
(κ + 1)p
[ ]
κp + 1 (3.13)
≤ + M 1 (κp + 1)L ∥uuκh ∥pp + M1 (κp + 1)H(L)cpΩ ∥uuκh ∥p1,p
(κ + 1)p
+ M2 G∥uuκh ∥pp,∂Ω + M2 K(G)cp∂Ω ∥uuκh ∥p1,p + M3 κ.

Now we choose L = L(κ, u) > 0 and G = G(κ, u) > 0 such that

κp + 1 κp + 1
M1 (κp + 1)H(L)cpΩ = , M2 K(G)cp∂Ω = .
4(κ + 1)p 4(κ + 1)p
Then, (3.13) becomes
κp + 1
∥uuκh ∥p1,p
2(κ + 1)p
[ ] (3.14)
κp + 1
≤ + M1 (κp + 1)L(κ, u) ∥uuκh ∥pp + M2 G(κ, u)∥uuκh ∥pp,∂Ω + M3 κ,
(κ + 1)p
where L(κ, u) and G(κ, u) depend on κ and on the solution u.
Case I.1: u ∈ Lr (Ω ) for any finite r
We can use Proposition 2.1 to estimate the remaining boundary term in form of

∥uuκh ∥pp,∂Ω ≤ ε2 ∥uuκh ∥p1,p + c̃1 ε−c̃

2
2
∥uuκh ∥pp . (3.15)
1 κp+1
Choosing ε2 = M2 G(κ,u) 4(κ+1)p and applying (3.15) to (3.14) gives

κp + 1
∥uuκh ∥p1,p
4(κ + 1)p
[ ] (3.16)
κp + 1 −c̃2
≤ + M 1 (κp + 1)L(κ, u) + M 2 G(κ, u)c̃ ε
1 2 ∥uuκh ∥pp + M3 κ.
(κ + 1)p
Inequality (3.16) can be rewritten as

∥uuκh ∥p1,p ≤ M4 (κ, u) ∥uuκh ∥pp + 1

[ ]
(3.17)

with a constant M4 (κ, u) depending on κ and on the function u. We may apply the Sobolev embedding
theorem on the left-hand side of (3.17) which leads to
]1
∥uuκh ∥p∗ ≤ cΩ ∥uuκh ∥1,p ≤ M5 (κ, u) ∥uuκh ∥pp + 1 p . (3.18)
[

Now we can start with the typical bootstrap arguments. We choose κ1 such that (κ1 + 1)p = p∗ . Then (3.18)
becomes ]1 ]1
∥uuκh1 ∥p∗ ≤ M5 (κ1 , u) ∥uuκh1 ∥pp + 1 p ≤ M6 (κ1 , u) ∥uκ1 +1 ∥pp + 1 p
[ [
[ ∗
] p1 (3.19)
= M6 (κ1 , u) ∥u∥pp∗ + 1 < ∞,
162 G. Marino and P. Winkert / Nonlinear Analysis 180 (2019) 154–169

since uh (x) = min(u(x), h(x)) ≤ u(x) for a. a. x ∈ Ω . Now we may apply Fatou’s Lemma as h → ∞ in
(3.19). This gives
1 [ ∗
] 1
(κ +1)p
≤ M7 (κ1 , u) ∥u∥pp∗ + 1 1
κ +1
∥u∥(κ1 +1)p∗ = ∥uκ1 +1 ∥p∗1 < ∞. (3.20)

∗
Hence, u ∈ L(κ1 +1)p (Ω ). Repeating the steps from (3.18)–(3.20) for each κ, we choose a sequence such that

κ2 : (κ2 + 1)p = (κ1 + 1)p∗ ,

κ3 : (κ3 + 1)p = (κ2 + 1)p∗ ,
.. ..
. . .

This shows that

∥u∥(κ+1)p∗ ≤ M8 (κ, u) (3.21)
for any finite number κ, where M8 (κ, u) is a positive constant depending both on κ and on the solution u.
Thus, u ∈ Lr (Ω ) for any r ∈ (1, ∞). This proves Case I.1.
Case I.2: u ∈ Lr (∂Ω ) for any finite r
Let us repeat inequality (3.14) which says
κp + 1
∥uuκh ∥p1,p
2(κ + 1)p
[ ] (3.22)
κp + 1
≤ + M9 (κp + 1)L(κ, u) ∥uuκh ∥pp + M10 G(κ, u)∥uuκh ∥pp,∂Ω + M11 κ.
(κ + 1)p

Taking into account (3.21), we can write (3.22) in the form

[ ]
∥uuκh ∥p1,p ≤ M12 (κ, u) ∥uuκh ∥pp,∂Ω + 1 . (3.23)

Now we may apply the Sobolev embedding theorem for the boundary on the left-hand side of (3.23). This
gives
[ ] p1
∥uuκh ∥p∗ ,∂Ω ≤ c∂Ω ∥uuκh ∥1,p ≤ M13 (κ, u) ∥uuκh ∥pp,∂Ω + 1 . (3.24)

As before we proceed with a bootstrap argument and choose κ1 in (3.24) such that (κ1 + 1)p = p∗ . This
yields
[ ] p1 [ ] p1
∥uuκh1 ∥p∗ ,∂Ω ≤ M13 (κ1 , u) ∥uuκh1 ∥pp,∂Ω + 1 ≤ M14 (κ1 , u) ∥uκ1 +1 ∥pp,∂Ω + 1
[ ] p1 (3.25)
≤ M14 (κ1 , u) ∥u∥pp∗∗ ,∂Ω + 1 < ∞.

Applying again Fatou’s Lemma we obtain from (3.25)

1 [ ] 1
(κ +1)p
∥u∥(κ1 +1)p∗ ,∂Ω = ∥uκ1 +1 ∥p∗1,∂Ω ≤ M15 (κ1 , u) ∥u∥pp∗∗ ,∂Ω + 1 1
κ +1
< ∞. (3.26)

Therefore, u ∈ L(κ1 +1)p∗ (∂Ω ). For each κ we repeat the steps from (3.24)–(3.26) and choose a sequence such
that
κ2 : (κ2 + 1)p = (κ1 + 1)p∗ ,
κ3 : (κ3 + 1)p = (κ2 + 1)p∗ ,
.. ..
. . .
We obtain
∥u∥(κ+1)p∗ ,∂Ω ≤ M16 (κ, u) (3.27)
 G. Marino and P. Winkert / Nonlinear Analysis 180 (2019) 154–169 163

for any finite number κ, where M16 (κ, u) is a positive constant depending on κ and on the solution u. Thus,
u ∈ Lr (∂Ω ) for any r ∈ (1, ∞), and therefore u ∈ Lr (Ω ) for any finite r ∈ (1, ∞). This completes the proof
of Part I.
Part II: u ∈ L∞ (Ω )
Let us recall inequality (3.9) which says
κp + 1
∥uuκh ∥p1,p
(κ + 1)p
∫ ∫ (3.28)
κp + 1 p∗ κp
≤ ∥uu κ p
∥
h p + M 17 (κp + 1) u uh dx + M 18 up∗ uκp
h dσ + M19 κ.
(κ + 1)p Ω ∂Ω

Let us fix numbers q̃1 ∈ (p, p∗ ) and q̃2 ∈ (p, p∗ ). Then, by applying Hölder’s inequality and the results of
Part I, see (3.21) and (3.27), we derive for the several terms on the right-hand side of (3.28)

q̃1 −p
(∫ ) q̃p
1
∥uuκh ∥pp
≤ |Ω | q̃1 (uuκh )q̃1 dx ≤ M20 ∥uuκh ∥pq̃1 ,
∫ ∫ Ω
p∗ κp p∗ −p
u uh dx = u (uuκh )p dx
Ω Ω
(∫ p∗ −p
) q̃1q̃−p (∫ ) q̃p
1 1
≤ u q̃1 −p q̃1 dx (uuκh )q̃1 ≤ M21 ∥uuκh ∥pq̃1 ,
Ω Ω (3.29)
∫ ∫
up∗ uκp
h dσ = up∗ −p (uuκh )p dσ
∂Ω ∂Ω
(∫ p∗ −p
) q̃2q̃−p (∫ ) q̃p
2 2
≤ u q̃2 −p q̃2 dσ (uuκh )q̃2 dσ
∂Ω ∂Ω
≤ M22 ∥uuκh ∥pq̃2 ,∂Ω .

Note that M21 , M22 are finite because of Part I. Moreover, we see from the calculations above that
( ) ( )
M21 = M21 ∥u∥ p∗ −p q̃ and M22 = M22 ∥u∥ p∗ −p q̃ . (3.30)
q̃1 −p 1 q̃2 −p 2

Using (3.29) to (3.28) leads to

κp + 1 κp + 1
∥uuκh ∥p1,p ≤ M23 ∥uuκh ∥pq̃1 + M24 ∥uuκh ∥pq̃2 ,∂Ω + M25 κ. (3.31)
(κ + 1)p (κ + 1)p

Case II.1: u ∈ L∞ (Ω )
As before, we can estimate the boundary term via Proposition 2.1 and then use Hölder’s inequality as
seen in the first line of (3.29). This gives

∥uuκh ∥pq̃2 ,∂Ω ≤ ε3 ∥uuκh ∥p1,p + c̃1 ε−c̃

3
2
∥uuκh ∥pp
(3.32)
≤ ε3 ∥uuκh ∥p1,p + c̃1 ε−c̃
3
2
M20 ∥uuκh ∥pq̃1 .
κp+1
Now we choose ε3 = 2M24 (κ+1)p and apply (3.32) in (3.31) to obtain

κp + 1 ( )
p
∥uuκh ∥p1,p ≤ M23 (κp + 1) + c̃1 ε−c̃
3
2
M20 M24 ∥uuκh ∥pq̃1 + M25 κ. (3.33)
2(κ + 1)

Inequality (3.33) can be rewritten in the form

M27
∥uuκh ∥p1,p ≤ M26 ((κ + 1)p ) ∥uuκh ∥pq̃1 + 1 .
[ ]
(3.34)
164 G. Marino and P. Winkert / Nonlinear Analysis 180 (2019) 154–169

In order so see this, note that

2(κ + 1)p ( )
M23 (κp + 1) + c̃1 ε−c̃
3
2
M 20 M 24
κp + 1
( )c̃ )
2M24 (κ + 1)p 2
(
p 1
= 2(κ + 1) M23 + c̃1 M20 M24
κp + 1 κp + 1
M27
≤ M26 ((κ + 1)p ) .

Now we may apply the Sobolev embedding on the left-hand side of (3.34) and the fact that u ∈ Lr (Ω ) for
any finite r ∈ (1, ∞) to get
]1
∥uuκh ∥p∗ ≤ cΩ ∥uuκh ∥1,p ≤ M27 (κ + 1)M28 ∥uuκh ∥pq̃1 + 1 p
( )[
]1 (3.35)
≤ M29 (κ + 1)M28 ∥uκ+1 ∥pq̃1 + 1 p < ∞.
( )[

Applying Fatou’s Lemma in (3.35) implies that

1 1 ( ) 1 [ ] 1
∥u∥(κ+1)p∗ = ∥uκ+1 ∥pκ+1
∗
κ+1
≤ M29 (κ + 1)M28 κ+1 ∥uκ+1 ∥pq̃1 + 1 (κ+1)p . (3.36)

Observe that
)√ 1 )√ 1
(κ + 1)M28 lim (κ + 1)M28 κ+1 = 1.
( (
κ+1 ≥1 and
κ→∞
Hence, we find a constant M30 > 1 such that
1 √1
(κ + 1)M28 ≤ M30 κ+1 .
( ) κ+1
(3.37)

From (3.36) and (3.37) we derive

1 √1 ] 1
M30 κ+1 ∥uκ+1 ∥pq̃1 + 1 (κ+1)p . (3.38)
κ+1
[
∥u∥(κ+1)p∗ ≤ M29

Now we are ready to prove the uniform boundedness with respect to κ. To this end, suppose there is a
sequence κn → ∞ such that
∥uκn +1 ∥pq̃1 ≤ 1,
which is equivalent to
∥u∥(κn +1)q̃1 ≤ 1,
then Proposition 2.2 implies that ∥u∥∞ < ∞.
In the opposite case there exists a number κ0 > 0 such that

∥uκ+1 ∥pq̃1 > 1 for any κ ≥ κ0 . (3.39)

Combining (3.38) and (3.39) yields

1 √1 ] 1 1 √1
M30 κ+1 2∥uκ+1 ∥pq̃1 (κ+1)p ≤ M31 M30 κ+1 ∥u∥(κ+1)q̃1 (3.40)
κ+1
[ κ+1
∥u∥(κ+1)p∗ ≤ M29

for any κ ≥ κ0 . Applying again the bootstrap arguments we define a sequence (κn ) such that

κ1 : (κ1 + 1)q̃1 = (κ0 + 1)p∗ ,

κ2 : (κ2 + 1)q̃1 = (κ1 + 1)p∗ ,
(3.41)
κ3 : (κ3 + 1)q̃1 = (κ2 + 1)p∗ ,
.. ..
. . .
 G. Marino and P. Winkert / Nonlinear Analysis 180 (2019) 154–169 165

By induction, from (3.40) and (3.41), we obtain

1 √ 1 1 √ 1
κn +1
∥u∥(κn +1)p∗ ≤ M31 M30 κn +1 ∥u∥(κn +1)q̃1 = M31
κn +1
M30 κn +1 ∥u∥(κn−1 +1)p∗
( ∗ )n
for any n ∈ N, where the sequence (κn ) is chosen in such a way that (κn + 1) = (κ0 + 1) pq̃1 . Following
this we see that ∑n ∑n
1 √ 1
i=1 κi +1 i=1 κi +1
∥u∥(κn +1)p∗ ≤ M31 M30 ∥u∥(κ0 +1)p∗
( )i
with (κn + 1)p∗ → ∞ as n → ∞. Since κ 1+1 = κ01+1 pq̃1∗ and pq̃1∗ < 1, there is a constant M32 > 0 such
i
that

∥u∥(κn +1)p∗ ≤ M32 ∥u∥(k0 +1)p∗ < ∞,

where the finiteness of the right-hand side follows from Part I. Now we may apply Proposition 2.2 to
conclude that u ∈ L∞ (Ω ), that is, there exists M > 0, which depends on the given data and on u, such that
∥u∥∞ ≤ M .
Case II.2: u ∈ L∞ (∂Ω )
This case follows directly from Case II.1 and Proposition 2.4.
Combining Case II.1 and II.2 shows that u ∈ L∞ (Ω ). □

Remark 3.2. It is clear that hypothesis (H1) is not needed in the proof of Theorem 3.1, but it is necessary
to have a well-defined definition of a weak solution.

Remark 3.3. Since problem (1.1) involves functions that can exhibit a critical growth, one cannot expect
to find a constant M which depends in an explicit way on natural norms such as ∥u∥p∗ or ∥u∥p∗ ,∂Ω . But, if
one searches for a dependence on norms that are greater than the critical ones, then a possible dependence
is given on the norms ∥u∥ p∗ −p q̃ as well as ∥u∥ p∗ −p q̃ ,∂Ω , where q̃1 ∈ (p, p∗ ) and q̃2 ∈ (p, p∗ ), as seen in the
q̃1 −p 1 q̃2 −p 2
proof of Theorem 3.1, see (3.30).

Based on the results of Theorem 3.1, we obtain regularity results for solutions of type (1.1). For
simplification we drop the s-dependence of the operator. To this end, let ϑ ∈ C 1 (0, ∞) be a function
such that
tϑ′ (t)
≤ a2 and a3 tp−1 ≤ ϑ(t) ≤ a4 1 + tp−1
( )
0 < a1 ≤ (3.42)
ϑ(t)
for all t > 0, with some constants ai > 0, i ∈ {1, 2, 3, 4} and for 1 < p < ∞. The hypotheses on
A : Ω × RN → RN read as follows.

H(A): A(x, ξ) = A0 (x, |ξ|) ξ with A0 ∈ C(Ω × R+ ) for all ξ ∈ RN , where R+ = [0, +∞) and with
A0 (x, t) > 0 for all x ∈ Ω and for all t > 0. Moreover,
(i) A0 ∈ C 1 (Ω × (0, ∞)), t → tA0 (x, t) is strictly increasing in (0, ∞), limt→0+ tA0 (x, t) = 0 for
all x ∈ Ω and
tA′0 (x, t)
lim = c > −1 for all x ∈ Ω ;
t→0+ A0 (x, t)

(ii) |∇ξ A(x, ξ)| ≤ a5 ϑ(|ξ|) N

|ξ| for all x ∈ Ω , for all ξ ∈ R \ {0} and for some a5 > 0;
ϑ(|ξ|) 2
(iii) ∇ξ A(x, ξ)y · y ≥ |ξ| |y| for all x ∈ Ω , for all ξ ∈ RN \ {0} and for all y ∈ RN .
166 G. Marino and P. Winkert / Nonlinear Analysis 180 (2019) 154–169

Remark 3.4. We chose the special structure in H(A) to apply the nonlinear regularity theory, which is
mainly based on the results of Lieberman [11] and Pucci–Serrin [15]. Closely related to this subject is also
the work by Motreanu–Motreanu–Papageorgiou [13]. If we set
∫ t
G0 (x, t) = A0 (x, s)sds,
0

then G0 ∈ C 1 (Ω × R+ ) and the function G0 (x, ·) is increasing and strictly convex for all x ∈ Ω . We set
G(x, ξ) = G0 (x, |ξ|) for all (x, ξ) ∈ Ω × RN and obtain that G ∈ C 1 (Ω × RN ) and that the function
ξ → G(x, ξ) is convex. Moreover, we easily derive that
ξ
∇ξ G(x, ξ) = (G0 )′t (x, |ξ|) = A0 (x, |ξ|)ξ = A(x, ξ)
|ξ|

for all ξ ∈ RN \ {0} and ∇ξ G(x, 0) = 0. So, G(x, ·) is the primitive of A(x, ·). This fact, the convexity of
G(x, ·) and since G(x, 0) = 0 for all x ∈ Ω imply that

G(x, ξ) ≤ A(x, ξ) · ξ for all (x, ξ) ∈ Ω × RN . (3.43)

The next lemma summarizes the main properties of A : Ω × RN → RN . The result is an easy consequence
of (3.42) and the hypotheses H(A).

Lemma 3.5. If hypotheses H(A) are satisfied, then the following hold:
(i) A ∈ C(Ω × RN , RN ) ∩ C 1 (Ω × (RN \ {0}), RN ) and the map ξ → A(x, ξ) is continuous and strictly
monotone (hence, maximal monotone) for all x ∈ Ω ;
( )
p−1
(ii) |A(x, ξ)| ≤ a6 1 + |ξ| for all x ∈ Ω , for all ξ ∈ RN and for some a6 > 0;
a3 p
(iii) A(x, ξ) · ξ ≥ p−1 |ξ| for all x ∈ Ω and for all ξ ∈ RN .

From this lemma along with (3.43) we easily deduce the following growth estimates for the primitive
G(x, ·).

Corollary 3.6. If hypotheses H(A) hold, then

a3 p p
|ξ| ≤ G(x, ξ) ≤ a7 (1 + |ξ| )
p(p − 1)

for all x ∈ Ω , for all ξ ∈ RN and for some a7 > 0.

Let A : W 1,p (Ω ) → W 1,p (Ω )∗ be the nonlinear map defined by

∫
⟨A(u), φ⟩ = A(x, ∇u) · ∇φdx for all u, φ ∈ W 1,p (Ω ). (3.44)
Ω

The next proposition summarizes the main properties of this operator, see Gasiński–Papageorgiou [7].

Proposition 3.7. Let the hypotheses H(A) be satisfied and let A : W 1,p (Ω ) → W 1,p (Ω )∗ be the map defined
in (3.44). Then, A is bounded, continuous, monotone (hence maximal monotone) and of type (S+ ).

Let us state some operators which fit in our setting and which are of much interest.

Example 3.8. For simplicity, we drop the x-dependence of the operator A. The following maps satisfy
hypotheses H(A):
 G. Marino and P. Winkert / Nonlinear Analysis 180 (2019) 154–169 167

p−2
(i) Let A(ξ) = |ξ| ξ with 1 < p < ∞. This map corresponds to the p-Laplace differential operator
defined by ( )
p−2
∆p u = div |∇u| ∇u for all u ∈ W 1,p (Ω ).
p
The potential is G(ξ) = p1 |ξ| for all ξ ∈ RN .
p−2 q−2
(ii) The function A(ξ) = |ξ| ξ + µ|ξ| ξ with 1 < q < p < ∞ and µ > 0 compares with the (p, q)-
p q
differential operator defined by ∆p u + µ∆q u for all u ∈ W 1,p (Ω ). The potential is G(ξ) = p1 |ξ| + µq |ξ|
for all ξ ∈ RN .
( ) p−2
2
2
(iii) If A(ξ) = 1 + |ξ| ξ with 1 < p < ∞, then this map represents the generalized p-mean curvature
differential operator defined by

2 p−2
[ ]
div (1 + |∇u| ) 2 ∇u for all u ∈ W 1,p (Ω ).
[ ]
1 2 p
The potential is G(ξ) = p (1 + |ξ| ) 2 − 1 for all ξ ∈ RN .

Let us write hypotheses (H) without the structure conditions on A.

H(B, C): The functions B : Ω × R × RN → R and C : ∂Ω × R → R are Carathéodory functions satisfying

the following structure conditions:
q −1
p 1q q1 −1
|B(x, s, ξ)| ≤ b1 |ξ| 1 + b2 |s| + b3 , for a.a. x ∈ Ω ,
q2 −1
|C(x, s)| ≤ c1 |s| + c2 , for a.a. x ∈ ∂Ω ,

for all s ∈ R, for all ξ ∈ RN , with positive constants bj , ck (j ∈ {1, 2, 3}, k ∈ {1, 2}) and fixed
numbers p, q1 , q2 such that

1 < p < ∞, p ≤ q1 ≤ p∗ , p ≤ q2 ≤ p∗

with the critical exponents stated in (2.2) and (2.1). Moreover, C satisfies the condition

α α
|C(x, s) − C(y, t)| ≤ L [|x − y| + |s − t| ] , |C(x, s)| ≤ L

for all (x, s), (y, t) ∈ ∂Ω × [−M0 , M0 ] with α ∈ (0, 1] and constants M0 > 0 and L ≥ 0.

Based on the hypotheses H(A) and H(B, C), problem (1.1) becomes

− div A(x, ∇u) = B(x, u, ∇u) in Ω ,

(3.45)
A(x, ∇u) · ν = C(x, u) on ∂Ω .

Combining Theorem 3.1 and the regularity theory of Lieberman [11] leads to the following result.

Theorem 3.9. Let Ω ⊂ RN , N > 1, be a bounded domain with a C 1,α -boundary ∂Ω and let the assumptions
H(A) and H(B, C) be satisfied. Then, every weak solution u ∈ W 1,p (Ω ) of problem (3.45) belongs to C 1,β (Ω )
for some β ∈ (0, 1) such that β = β(a1 , a2 , a5 , α, N ) and

∥u∥C 1,β (Ω) ≤ C(a1 , a2 , a3 , a5 , N, ϑ(1), M, α, b1 , b2 , b3 )

where M is the constant that comes from the statement of Theorem 3.1.
168 G. Marino and P. Winkert / Nonlinear Analysis 180 (2019) 154–169

Proof . We will apply Theorem 1.7 of Lieberman [11] and the comment after this theorem concerning global
Hölder gradient estimates. First, we know from Theorem 3.1 that ∥u∥∞ ≤ M . The only thing we need to
do is to check that the conditions (1.10a)–(1.10d) in [11, p. 320] are satisfied. From conditions H(A)(iii), (ii)
we see that the assumptions (1.10a) and (1.10b) are satisfied. Moreover, from H(B, C) and (3.42) we obtain
q −1
p 1q q1 −1
|B(x, s, ξ)| ≤ b1 |ξ| 1 + b2 |s| + b3
p q1 −1
≤ b1 |ξ| + b1 + b2 M + b3
p−1 q1 −1
= b1 |ξ| |ξ| + b1 + b2 M + b3
b1
≤ ϑ(|ξ|)|ξ| + b1 + b2 M q1 −1 + b3
a3 { }
b1 q1 −1
≤ max , b1 + b2 M + b3 (ϑ(|ξ|)|ξ| + 1) .
a3

This proves condition (1.10d). Assumption (1.10c) follows from the fact that the function A is continuous
differentiable in the space variable and independent of the s-variable. Then we may apply the mean value
theorem which shows (1.10c). The desired result follows from Lieberman [11, Theorem 1.7] with the constants
β, C as in the theorem (and their dependence on the data) and the constant M from Theorem 3.1. □

Acknowledgments

The authors wish to thank the referees for their corrections and insightful remarks that helped to improve
the paper.
This work has been performed in the framework of Piano della Ricerca 2016-2018–linea di intervento 2:
“Metodi variazionali ed equazioni differenziali”. The second author thanks the University of Catania for the
kind hospitality during a research stay in March 2018.

References

[1] R.A. Adams, Sobolev Spaces, Academic Press, New York-London, 1975.
[2] P. Drábek, A. Kufner, F. Nicolosi, Quasilinear Elliptic Equations with Degenerations and Singularities, Walter de Gruyter
& Co., Berlin, 1997.
[3] X. Fan, D. Zhao, A class of De Giorgi type and Hölder continuity, Nonlinear Anal. 36 (3) (1999) 295–318.
[4] J.P. Garcı́a Azorero, I. Peral Alonso, Some results about the existence of a second positive solution in a quasilinear critical
problem, Indiana Univ. Math. J. 43 (3) (1994) 941–957.
[5] L. Gasiński, N.S. Papageorgiou, Nonlinear Analysis, Chapman & Hall/CRC, Boca Raton, FL, 2006.
[6] L. Gasiński, N.S. Papageorgiou, Anisotropic nonlinear Neumann problems, Calc. Var. Partial Differential Equations 42
(3–4) (2011) 323–354.
[7] L. Gasiński, N.S. Papageorgiou, Exercises in Analysis. Part 1: Nonlinear Analysis, Springer, Heidelberg, 2014.
[8] M. Guedda, L. Véron, Quasilinear elliptic equations involving critical Sobolev exponents, Nonlinear Anal. 13 (8) (1989)
879–902.
[9] S. Hu, N.S. Papageorgiou, Nonlinear Neumann equations driven by a nonhomogeneous differential operator, Commun.
Pure Appl. Anal. 10 (4) (2011) 1055–1078.
[10] A. Lê, Eigenvalue problems for the p-Laplacian, Nonlinear Anal. 64 (5) (2006) 1057–1099.
[11] G.M. Lieberman, The natural generalization of the natural conditions of Ladyzhenskaya and Ural´tseva for elliptic
equations, Comm. Partial Differential Equations 16 (2–3) (1991) 311–361.
[12] D. Motreanu, V.V. Motreanu, N.S. Papageorgiou, Nonlinear Neumann problems near resonance, Indiana Univ. Math. J.
58 (3) (2009) 1257–1279.
[13] D. Motreanu, V.V. Motreanu, N.S. Papageorgiou, Multiple constant sign and nodal solutions for nonlinear Neumann
eigenvalue problems, Ann. Sc. Norm. Super. Pisa Cl. Sci. (5) 10 (3) (2011) 729–755.
[14] N.S. Papageorgiou, V.D. Rădulescu, Nonlinear nonhomogeneous Robin problems with superlinear reaction term, Adv.
Nonlinear Stud. 16 (4) (2016) 737–764.
[15] P. Pucci, J. Serrin, The Maximum Principle, Birkhäuser Verlag, Basel, 2007.
[16] P. Pucci, R. Servadei, Regularity of weak solutions of homogeneous or inhomogeneous quasilinear elliptic equations,
Indiana Univ. Math. J. 57 (7) (2008) 3329–3363.
[17] M. Struwe, Variational Methods, Springer-Verlag, Berlin, 2008.
 G. Marino and P. Winkert / Nonlinear Analysis 180 (2019) 154–169 169

[18] N.S. Trudinger, Remarks concerning the conformal deformation of Riemannian structures on compact manifolds, Ann.
Scuola Norm. Sup. Pisa (3) 22 (1968) 265–274.
[19] X.J. Wang, Neumann problems of semilinear elliptic equations involving critical Sobolev exponents, J. Differential
Equations 93 (2) (1991) 283–310.
[20] P. Winkert, Constant-sign and sign-changing solutions for nonlinear elliptic equations with Neumann boundary values,
Adv. Differential Equations 15 (5–6) (2010) 561–599.
[21] P. Winkert, L∞ -estimates for nonlinear elliptic Neumann boundary value problems, NoDEA Nonlinear Differential
Equations Appl. 17 (3) (2010) 289–302.
[22] P. Winkert, On the boundedness of solutions to elliptic variational inequalities, Set-Valued Var. Anal. 22 (4) (2014)
763–781.
[23] P. Winkert, R. Zacher, Corrigendum to A priori bounds for weak solutions to elliptic equations with nonstandard growth,
Discrete Contin. Dyn. Syst. Ser. S 5 (2012) 865–878, Discrete Contin. Dyn. Syst. Ser. S, published on-line as note, 2015,
1–3.
[24] P. Winkert, R. Zacher, A priori bounds for weak solutions to elliptic equations with nonstandard growth, Discrete Contin.
Dyn. Syst. Ser. S 5 (4) (2012) 865–878.
[25] P. Winkert, R. Zacher, Global a priori bounds for weak solutions to quasilinear parabolic equations with nonstandard
growth, Nonlinear Anal. 145 (2016) 1–23.