Boundary Value Problems
Volume 2009 (2009), Article ID 853607, 23 pages
doi:10.1155/2009/853607
Research Article

Stagnation Zones for 𝒜 -Harmonic Functions on Canonical Domains

1Department of Mathematics, Volgograd State University, Universitetskii prospect 100, Volgograd 400062, Russia
2Department of Mathematics and Systems Analysis, Aalto University, P.O. Box 11100, FI-00076 Aalto, Finland
3Department of Mathematics, University of Turku, 20014 Turku, Finland

Received 1 July 2009; Revised 31 October 2009; Accepted 15 November 2009

Academic Editor: Sandro Salsa

Copyright © 2009 Vladimir M. Miklyukov et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

We study stagnation zones of 𝒜 -harmonic functions on canonical domains in the Euclidean 𝑛 -dimensional space. Phragmén-Lindelöf type theorems are proved.

1. Introduction

In this article we investigate solutions of the 𝒜 -Laplace equation on canonical domains in the 𝑛 -dimensional Euclidean space.

Suppose that 𝐷 is a domain in 𝑛 , and let 𝑓 𝐷 be a function. For 𝑠 > 0 , a subset Δ 𝐷 is called 𝑠 -zone (stagnation zone with the deviation 𝑠 ) of 𝑓 if there exists a constant 𝐶 such that the difference between 𝐶 and the function 𝑓 is smaller than 𝑠 on Δ . We may, for example, consider difference in the sense of the sup norm

𝑓 ( 𝑥 ) 𝐶 𝐶 ( Δ ) = s u p 𝑥 Δ | | | | 𝑓 ( 𝑥 ) 𝐶 < 𝑠 , ( 1 . 1 ) the 𝐿 𝑝 -norm

𝑓 ( 𝑥 ) 𝐶 𝐿 𝑝 ( Δ ) = Δ | | | | 𝑓 ( 𝑥 ) 𝐶 𝑝 𝑑 𝑛 1 / 𝑝 < 𝑠 , ( 1 . 2 ) or the Sobolev norm

𝑓 ( 𝑥 ) 𝐶 𝑊 1 𝑝 ( Δ ) = Δ | | | | 𝑓 ( 𝑥 ) 𝑝 𝑑 𝑛 1 / 𝑝 < 𝑠 , ( 1 . 3 ) where 𝑑 is the 𝑑 -dimensional Hausdorff measure in 𝑛 .

For discussion about the history of the question, recent results and applications the reader is referred to [1, 2].

Some estimates of stagnation zone sizes for solutions of the 𝒜 -Laplace equation on locally Lipschitz surfaces and behavior of solutions in stagnation zones were given in [3]. In this paper we consider solutions of the 𝒜 -Laplace equation in subdomains of 𝑛 of a special form, canonical domains. In two-dimensional case, such domains are sectors and strips. In higher dimensions, they are conical and cylindrical regions. The special form of domains allows us to obtain more precise results.

Below we study stagnation zones of generalized solutions of the 𝒜 -Laplace equation

d i v 𝒜 ( 𝑥 , 𝑓 ) = 0 , ( 1 . 4 ) (see [4]) with boundary conditions of types (see Definitions 1.1 and 1.2 below)

𝒜 ( 𝑥 , 𝑓 ) , 𝐧 𝑓 = 0 , 𝑥 𝜕 𝐷 𝐺 , 𝒜 ( 𝑥 , 𝑓 ) , 𝐧 = 0 , 𝑥 𝜕 𝐷 𝐺 ( 1 . 5 ) on canonical domains in the Euclidean 𝑛 -dimensional space, where 𝐺 is a closed subset of 𝜕 𝐷 . We will prove Phragmén-Lindelöf type theorems for solutions of the 𝒜 -Laplace equation with such boundary conditions.

1.1. Canonical Domains

Let 𝑛 2 . Fix an integer 𝑘 , 1 𝑘 𝑛 , and set

𝑑 𝑘 ( 𝑥 ) = 𝑘 𝑖 = 1 𝑥 2 𝑖 1 / 2 . ( 1 . 6 ) We call the set

𝐵 𝑘 ( 𝑡 ) = 𝑥 𝑛 𝑑 𝑘 ( 𝑥 ) < 𝑡 ( 1 . 7 ) a 𝑘 -ball and

Σ 𝑘 ( 𝑡 ) = 𝑥 𝑛 𝑑 𝑘 ( 𝑥 ) = 𝑡 ( 1 . 8 ) a 𝑘 -sphere in 𝑛 . In particular, the symbol Σ 𝑘 ( 0 ) denotes the 𝑘 -sphere with the radius 0 , that is, the set

Σ 𝑘 𝑥 ( 0 ) = 𝑥 = 1 , , 𝑥 𝑘 , , 𝑥 𝑛 𝑥 1 = = 𝑥 𝑘 . = 0 ( 1 . 9 ) For every 0 < 𝑘 < 𝑛 , we set

𝑝 𝑘 ( 𝑥 ) = 𝑛 𝑗 = 𝑘 + 1 𝑥 2 𝑗 1 / 2 , Σ 𝑘 ( 𝑡 ) = 𝑥 𝑛 𝑝 𝑘 ( 𝑥 ) = 𝑡 , 𝑡 0 . ( 1 . 1 0 )

Let 0 < 𝛼 < 𝛽 < be fixed, and let (see Figure 1)

fig1
Figure 1: 𝐷 1 1 , 2 (a) and 𝐷 2 1 , 2 in 3 .

𝐷 𝑘 𝛼 , 𝛽 = 𝑥 𝑛 𝛼 < 𝑝 𝑘 . ( 𝑥 ) < 𝛽 ( 1 . 1 1 ) For 𝑘 = 𝑛 1 , we also assume that 𝑥 𝑛 > 0 . Then for 𝑘 = 𝑛 1 , the 𝐷 𝑛 1 𝛼 , 𝛽 is the a layer between two parallel hyperplanes, and for 1 𝑘 < 𝑛 1 the boundary of the domain 𝐷 𝑘 𝛼 , 𝛽 consists of two coaxial cylindrical surfaces. The intersections Σ 𝑘 ( 𝑡 ) 𝐷 𝑘 𝛼 , 𝛽 are precompact for all 𝑡 > 0 . Thus, the functions 𝑑 𝑘 ( 𝑥 ) are exhaustion functions for 𝐷 𝑘 𝛼 , 𝛽 .

1.2. Structure Conditions

Let 𝐷 be a subdomain of 𝑛 and let

𝒜 ( 𝑥 , 𝜉 ) 𝐷 × 𝑛 𝑛 ( 1 . 1 2 ) be a vector function such that for a.e. 𝑥 𝐷 the function

𝒜 ( 𝑥 , 𝜉 ) 𝑛 𝑛 ( 1 . 1 3 ) is defined and is continuous with respect to 𝜉 . We assume that the function

𝑥 𝒜 ( 𝑥 , 𝜉 ) ( 1 . 1 4 ) is measurable in the Lebesgue sense for all 𝜉 𝑛 and

𝒜 | | 𝜆 | | ( 𝑥 , 𝜆 𝜉 ) = 𝜆 𝑝 2 𝒜 ( 𝑥 , 𝜉 ) , 𝜆 { 0 } , 𝑝 1 . ( 1 . 1 5 ) Suppose that for a.e. 𝑥 𝐷 and for all 𝜉 𝑛 the following properties hold:

𝜈 1 | | 𝜉 | | 𝑝 | | | | 𝜉 , 𝒜 ( 𝑥 , 𝜉 ) , 𝒜 ( 𝑥 , 𝜉 ) 𝜈 2 | | 𝜉 | | 𝑝 1 , ( 1 . 1 6 ) with 𝑝 1 and some constants 𝜈 1 , 𝜈 2 > 0 . We consider the equation

d i v 𝒜 ( 𝑥 , 𝑓 ) = 0 . ( 1 . 1 7 ) An important special case of (1.17) is the Laplace equation

Δ 𝑓 = 𝑛 𝑖 = 1 𝜕 2 𝑓 𝜕 𝑥 2 𝑖 = 0 . ( 1 . 1 8 ) As in [4, Chapter 6], we call continuous weak solutions of (1.17) 𝒜 -harmonic functions. However we should note that our definition of generalized solutions is slightly different from the definition given in [4, page 56].

1.3. Frequencies

Fix 𝑡 0 and 𝑝 1 . Let 𝑂 be an open subset of Σ 𝑘 ( 𝑡 ) (with respect to the relative topology of Σ 𝑘 ( 𝑡 ) ), and let 𝒫 be a nonempty closed subset of 𝜕 𝑂 . We set

𝜆 𝑝 , 𝒫 ( 𝑂 ) = i n f 𝑢 𝑂 | | | | 𝑢 𝑝 𝑑 𝑛 1 𝑂 𝑢 𝑝 𝑑 𝑛 1 , ( 1 . 1 9 ) where 𝑢 L i p l o c ( 𝑂 ) 𝐶 0 ( 𝑂 ) with 𝑢 | 𝒫 = 0 . If 𝒫 = 𝜕 𝑂 , then we call 𝜆 𝑝 ( 𝑂 ) 𝜆 𝑝 , 𝒫 ( 𝑂 ) the first frequency of the order 𝑝 1 of the set 𝑂 . If 𝒫 𝜕 𝑂 , then the quantity 𝜆 𝑝 , 𝒫 ( 𝑂 ) is the third frequency.

The second frequency is the following quantity:

𝜇 𝑝 ( 𝑂 ) = s u p 𝐶 i n f 𝑢 𝑂 | | | | 𝑢 𝑝 𝑑 𝑛 1 𝑂 ( 𝑢 𝐶 ) 𝑝 𝑑 𝑛 1 , ( 1 . 2 0 ) where the supremum is taken over all constants 𝐶 and 𝑢 L i p l o c ( 𝑂 ) 𝐶 0 ( 𝑂 ) . See also Pólya and Szegö [5] as well as Lax [6].

1.4. Generalized Boundary Conditions

Suppose that 𝐷 is a proper subdomain of 𝑛 . Let 𝜑 𝐷 be a locally Lipschitz function. We denote by 𝐷 𝑏 ( 𝜑 ) the set of all points 𝑥 𝐷 at which 𝜑 does not have the differential. Let 𝑈 𝐷 be a subset and let 𝜕 𝑈 = 𝜕 𝑈 𝜕 𝐷 be its boundary with respect to 𝐷 . If 𝜕 𝑈 is ( 𝑛 1 , 𝑛 1 ) -rectifiable, then it has locally finite perimeter in the sense of De Giorgi, and therefore a unit normal vector 𝐧 exists 𝑛 1 -almost everywhere on 𝜕 𝑈 [7, Sections 3.2.14, 3.2.15].

Let 𝐷 𝑛 be a domain and let 𝐺 𝜕 𝐷 be a subset of the boundary of 𝐷 . Define the concept of a generalized solution of (1.17) with zero boundary conditions on 𝜕 𝐷 𝐺 . A subset 𝑈 𝐷 is called admissible, if 𝑈 𝐺 = and 𝑈 have a ( 𝑛 1 , 𝑛 1 ) -rectifiable boundary with respect to 𝐷 .

Suppose that 𝐷 is unbounded. Let 𝐺 𝜕 𝐷 be a set closed in 𝑛 { } . We denote by ( 𝐺 , 𝐷 ) the collection of all subdomains 𝑈 𝐷 with 𝜕 𝑈 ( 𝐷 ( 𝜕 𝐷 𝐺 ) ) and ( 𝑛 1 , 𝑛 1 ) -rectifiable boundaries 𝜕 𝑈 = 𝜕 𝑈 𝜕 𝐷 .

Definition 1.1. We say that a locally Lipschitz function 𝑓 𝐷 is a generalized solution of (1.17) with the boundary condition 𝒜 ( 𝑥 , 𝑓 ) , 𝐧 = 0 , 𝑥 𝜕 𝐷 𝐺 , ( 1 . 2 1 ) if for every subdomain 𝑈 ( 𝐺 , 𝐷 ) , 𝑛 1 𝜕 𝑈 𝐷 𝑏 ( 𝑓 ) = 0 , ( 1 . 2 2 ) and for every locally Lipschitz function 𝜑 𝑈 𝐺 the following property holds: 𝜕 𝑈 𝜑 𝒜 ( 𝑥 , 𝑓 ) , 𝐧 𝑑 𝑛 1 = 𝑈 𝒜 ( 𝑥 , 𝑓 ) , 𝜑 𝑑 𝑛 . ( 1 . 2 3 ) Here 𝐧 is the unit normal vector of 𝜕 𝑈 and 𝑑 𝑛 is the volume element on 𝑛 .

Definition 1.2. We say that a locally Lipschitz function 𝑓 𝐷 is a generalized solution of (1.17) with the boundary condition 𝑓 𝒜 ( 𝑥 , 𝑓 ) , 𝐧 = 0 , 𝑥 𝜕 𝐷 𝐺 , ( 1 . 2 4 ) if for every subdomain 𝑈 ( 𝐺 , 𝐷 ) with (1.22) and for every locally Lipschitz function 𝜑 𝑈 𝐺 the following property holds: 𝜕 𝑈 𝜑 𝑓 𝒜 ( 𝑥 , 𝑓 ) , 𝐧 𝑑 𝑛 1 = 𝑈 𝒜 ( 𝑥 , 𝑓 ) , ( 𝜑 𝑓 ) 𝑑 𝑛 . ( 1 . 2 5 )

In the case of a smooth boundary 𝜕 𝐷 , and 𝑓 𝐶 2 ( 𝐷 ) , the relation (1.23) implies (1.17) with (1.21) everywhere on 𝜕 𝐷 𝐺 . This requirement (1.25) implies (1.17) with (1.24) on 𝜕 𝐷 𝐺 . See [8, Section 9.2.1].

The surface integrals exist by (1.22). Indeed, this assumption guarantees that 𝑓 ( 𝑥 ) exists 𝑛 1 a.e. on 𝜕 𝑈 . The assumption that 𝑈 ( 𝐺 , 𝐷 ) implies existence of a normal vector 𝐧 for 𝑛 1 a.e. points on 𝜕 𝑈 [7, Chapter 2, Section 3.2]. Thus, the scalar product 𝒜 ( 𝑥 , 𝑓 ) , 𝐧 is defined and is finite a.e. on 𝜕 𝑈 .

2. Saint-Venant’s Principle

In this section, we will prove the Saint-Venant principle for solutions of the 𝒜 -Laplace equation. The Saint-Venant principle states that strains in a body produced by application of a force onto a small part of its surface are of negligible magnitude at distances that are large compared to the diameter of the part where the force is applied. This well known result in elasticity theory is often stated and used in a loose form. For mathematical investigation of the results of this type, see, for example, [9].

In this paper the inequalities of the form (2.5), (2.4) are called the Saint-Venant principle (see also [9, 10]). Here we consider only the case of canonical domains. We plan to consider the general case in another article.

Let 0 < 𝑘 < 𝑛 . Fix a domain 𝐷 0 in 𝑘 with compact and smooth boundary, and write

𝐷 = 𝐷 0 × 𝑛 𝑘 = 𝑥 𝑛 𝑥 1 , , 𝑥 𝑘 𝐷 0 . ( 2 . 1 ) We write 𝒫 = { 𝑥 𝜕 𝐷 𝑝 𝑘 ( 𝑥 ) = 𝛼 } , 𝒬 = { 𝑥 𝜕 𝐷 𝑝 𝑘 ( 𝑥 ) = 𝛽 } , and 𝐺 = 𝒫 𝒬 . Let 𝑡 , 𝜏 ( 𝛼 , 𝛽 ) , 𝑡 < 𝜏 and

Δ 𝑘 ( 𝑡 , 𝜏 ) = 𝑥 𝐷 𝑡 < 𝑝 𝑘 ( . 𝑥 ) < 𝜏 ( 2 . 2 ) For 𝑠 0 , we set

𝜎 𝑘 ( 𝑠 ) = 𝑥 Δ 𝑘 ( 0 , ) 𝑝 𝑘 ( . 𝑥 ) = 𝑠 ( 2 . 3 )

Theorem 2.1. Let 𝛼 < 𝜏 < 𝜏 < 𝛽 , and let 0 < 𝑘 < 𝑛 . If 𝑓 𝐷 is a generalized solution of (1.17) with the generalized boundary condition (1.21) on 𝜕 𝐷 𝐺 , then the inequality 𝐼 1 𝑡 , 𝜏 + 𝐶 1 ( 𝑡 ) 𝜈 1 𝐼 1 𝑡 , 𝜏 + 𝐶 1 ( 𝑡 ) 𝜈 1 𝜈 e x p 1 𝜈 2 𝜏 𝜏 𝜇 𝑝 𝜎 𝑘 ( 𝜏 ) 𝑑 𝜏 ( 2 . 4 ) holds for all 𝑡 ( 𝛼 , 𝜏 ] .
If 𝑓 𝐷 is a generalized solution of (1.17) with the generalized boundary condition (1.24), then 𝐼 1 𝑡 , 𝜏 + 𝐶 2 ( 𝑡 ) 𝜈 1 𝐼 1 𝑡 , 𝜏 + 𝐶 2 ( 𝑡 ) 𝜈 1 𝜈 e x p 1 𝜈 2 𝜏 𝜏 𝜆 1 / 𝑝 𝑝 , 𝑍 𝑓 ( 𝜏 ) 𝜎 𝑘 ( 𝜏 ) 𝑑 𝜏 ( 2 . 5 ) holds for all 𝑡 ( 𝛼 , 𝜏 ] . Here 𝐼 1 ( 𝑡 , 𝜏 ) = Δ 𝑘 ( 𝑡 , 𝜏 ) | | | | 𝑓 𝑝 𝑑 𝑛 , ( 2 . 6 ) 𝑍 𝑓 ( 𝜏 ) = 𝑥 Σ 𝑘 ( 𝜏 ) 𝜕 𝐷 l i m 𝑦 𝑥 . 𝑓 ( 𝑦 ) = 0 ( 2 . 7 )

Proof. Case A. At first we consider the case in which 𝑓 is a generalized solution of (1.17) with the generalized boundary condition (1.24) on 𝜕 𝐷 𝐺 . It is easy to see that a.e. on 𝐷 𝑘 𝛼 , 𝛽 , | | 𝑝 𝑘 | | ( 𝑥 ) = 1 . ( 2 . 8 ) The domain Δ 𝑘 ( 𝑡 , 𝜏 ) belongs to ( 𝐺 , 𝐷 ) . Let 𝜑 𝑈 𝐺 be a locally Lipschitz function. By (1.25) we have 𝜕 Δ 𝑘 ( 𝑡 , 𝜏 ) 𝜑 𝑓 𝒜 ( 𝑥 , 𝑓 ) , 𝐧 𝑑 𝑛 1 = Δ 𝑘 ( 𝑡 , 𝜏 ) 𝒜 ( 𝑥 , 𝑓 ) , ( 𝜑 𝑓 ) 𝑑 𝑛 . ( 2 . 9 ) But 𝜕 Δ 𝑘 ( 𝑡 , 𝜏 ) = 𝜎 𝑘 ( 𝑡 ) 𝜎 𝑘 ( 𝜏 ) . ( 2 . 1 0 ) For 𝜑 1 , we have by (1.16) and (1.25) 𝜈 1 𝐼 1 ( 𝑡 , 𝜏 ) Δ 𝑘 ( 𝑡 , 𝜏 ) 𝒜 ( 𝑥 , 𝑓 ) , 𝑓 𝑑 𝑛 = 𝜎 𝑘 ( 𝜏 ) 𝑓 𝒜 ( 𝑥 , 𝑓 ) , 𝑝 𝑘 ( 𝑥 ) 𝑑 𝑛 1 𝜎 𝑘 ( 𝑡 ) 𝑓 𝒜 ( 𝑥 , 𝑓 ) , 𝑝 𝑘 ( 𝑥 ) 𝑑 𝑛 1 , ( 2 . 1 1 ) since 𝐧 = 𝑝 𝑘 ( 𝑥 ) for 𝑥 𝜎 𝑘 ( 𝜏 ) and 𝐧 = 𝑝 𝑘 ( 𝑥 ) for 𝑥 𝜎 𝑘 ( 𝑡 ) . We obtain 𝜈 1 𝐼 1 ( 𝑡 , 𝜏 ) + 𝐶 2 ( 𝑡 ) 𝜎 𝑘 ( 𝜏 ) 𝑓 𝒜 ( 𝑥 , 𝑓 ) , 𝑝 𝑘 ( 𝑥 ) 𝑑 𝑛 1 , ( 2 . 1 2 ) where 𝐶 2 ( 𝑡 ) = 𝜎 𝑘 ( 𝑡 ) 𝑓 𝒜 ( 𝑥 , 𝑓 ) , 𝑝 𝑘 ( 𝑥 ) 𝑑 𝑛 1 . ( 2 . 1 3 ) Note that we may also choose 𝐶 2 ( 𝜏 ) = 𝜎 𝑘 ( 𝜏 ) 𝑓 𝒜 ( 𝑥 , 𝑓 ) , 𝑝 𝑘 ( 𝑥 ) 𝑑 𝑛 1 ( 2 . 1 4 ) to obtain an inequality similar to (2.12).
Next we will estimate the right side of (2.12). By (1.16) and the Hölder inequality, | | | | 𝜎 𝑘 ( 𝜏 ) 𝑓 𝒜 ( 𝑥 , 𝑓 ) , 𝑝 𝑘 ( 𝑥 ) 𝑑 𝑛 1 | | | | 𝜎 𝑘 ( 𝜏 ) | | 𝑓 | | | | 𝒜 | | ( 𝑥 , 𝑓 ) 𝑑 𝑛 1 𝜈 2 𝜎 𝑘 ( 𝜏 ) | | 𝑓 | | | | | | 𝑓 𝑝 1 𝑑 𝑛 1 𝜈 2 𝜎 𝑘 ( 𝜏 ) | | 𝑓 | | 𝑝 𝑑 𝑛 1 1 / 𝑝 𝜎 𝑘 ( 𝜏 ) | | | | 𝑓 𝑝 𝑑 𝑛 1 ( 𝑝 1 ) / 𝑝 . ( 2 . 1 5 ) By using (1.19), we may write 𝜎 𝑘 ( 𝜏 ) | | 𝑓 | | 𝑝 𝑑 𝑛 1 𝜆 1 𝑝 , 𝑍 𝑓 ( 𝜏 ) 𝜎 𝑘 ( 𝜏 ) 𝜎 𝑘 ( 𝜏 ) | | | | 𝑓 𝑝 𝑑 𝑛 1 , ( 2 . 1 6 ) | | | | 𝜎 𝑘 ( 𝜏 ) 𝑓 𝒜 ( 𝑥 , 𝑓 ) , 𝑝 𝑘 ( 𝑥 ) 𝑑 𝑛 1 | | | | 𝜈 2 𝜆 1 / 𝑝 𝑝 , 𝑍 𝑓 ( 𝜏 ) 𝜎 𝑘 ( 𝜏 ) 𝜎 𝑘 ( 𝜏 ) | | | | 𝑓 𝑝 𝑑 𝑛 1 . ( 2 . 1 7 )
By (2.12) and the Fubini theorem, 𝜈 1 𝐼 1 ( 𝑡 , 𝜏 ) + 𝐶 2 ( 𝑡 ) 𝜈 2 𝜆 1 / 𝑝 𝑝 , 𝑍 𝑓 ( 𝜏 ) 𝜎 𝑘 ( 𝜏 ) 𝑑 𝐼 1 𝜈 𝑑 𝜏 ( 𝑡 , 𝜏 ) , 1 𝜈 2 𝜆 1 / 𝑝 𝑝 , 𝑍 𝑓 ( 𝜏 ) 𝜎 𝑘 ( 𝜏 ) 𝑑 𝐼 1 / 𝑑 𝜏 ( 𝑡 , 𝜏 ) 𝐼 1 ( 𝑡 , 𝜏 ) + 𝐶 2 ( 𝑡 ) / 𝜈 1 . ( 2 . 1 8 ) By integrating this differential inequality, we have 𝜈 e x p 1 𝜈 2 𝜏 𝜏 𝜆 1 / 𝑝 𝑝 , 𝑍 𝑓 ( 𝜏 ) 𝜎 𝑘 𝐼 ( 𝜏 ) 𝑑 𝜏 1 𝑡 , 𝜏 + 𝐶 2 ( 𝑡 ) / 𝜈 1 𝐼 1 ( 𝑡 , 𝜏 ) + 𝐶 2 ( 𝑡 ) / 𝜈 1 ( 2 . 1 9 ) for arbitrary 𝜏 , 𝜏 ( 𝛼 , 𝛽 ) with 𝜏 < 𝜏 . We have shown that 𝐼 1 𝑡 , 𝜏 + 𝐶 2 ( 𝑡 ) 𝜈 1 𝐼 1 𝑡 , 𝜏 + 𝐶 2 ( 𝑡 ) 𝜈 1 𝜈 e x p 1 𝜈 2 𝜏 𝜏 𝜆 1 / 𝑝 𝑝 , 𝑍 𝑓 ( 𝜏 ) 𝜎 𝑘 𝑝 . ( 𝜏 ) 𝑑 𝜏 ( 2 . 2 0 )
Case B. Now we assume that 𝑓 is a generalized solution of (1.17) with the boundary condition (1.21) on 𝜕 𝐷 𝐺 . Fix 𝑡 < 𝜏 . By choosing 𝜑 1 in (1.23), we see that 𝜎 𝑘 ( 𝑡 ) 𝜎 𝑘 ( 𝜏 ) 𝒜 ( 𝑥 , 𝑓 ) , 𝐧 𝑑 𝑛 1 = 0 . ( 2 . 2 1 ) For an arbitrary constant 𝐶 , we get from this and (1.23) 𝜎 𝑘 ( 𝑡 ) 𝜎 𝑘 ( 𝜏 ) ( 𝑓 𝐶 ) 𝒜 ( 𝑥 , 𝑓 ) , 𝐧 𝑑 𝑛 1 = Δ 𝑘 ( 𝑡 , 𝜏 ) 𝒜 ( 𝑥 , 𝑓 ) , 𝑓 𝑑 𝑛 . ( 2 . 2 2 ) Thus Δ 𝑘 ( 𝑡 , 𝜏 ) 𝒜 ( 𝑥 , 𝑓 ) , 𝑓 𝑑 𝑛 𝐶 1 ( 𝑡 ) + 𝜎 𝑘 ( 𝜏 ) | | | | | | 𝒜 𝑓 𝐶 𝑥 , 𝑝 𝑘 ( | | 𝑥 ) 𝑑 𝑛 1 , ( 2 . 2 3 ) where 𝐶 1 ( 𝑡 ) = 𝜎 𝑘 ( 𝑡 ) | | | | | | 𝒜 𝑓 𝐶 𝑥 , 𝑝 𝑘 ( | | 𝑥 ) 𝑑 𝑛 1 ( 2 . 2 4 ) or 𝜈 1 𝐼 1 ( 𝑡 , 𝜏 ) + 𝐶 1 ( 𝑡 ) 𝜈 2 𝜎 𝑘 ( 𝜏 ) | | | | | | | | 𝑓 𝐶 𝑓 𝑝 1 𝑑 𝑛 1 . ( 2 . 2 5 ) As above, we obtain 𝜎 𝑘 ( 𝜏 ) | | | | | | | | 𝑓 𝐶 𝑓 𝑝 1 𝑑 𝑛 1 𝜎 𝑘 ( 𝜏 ) | | | | 𝑓 𝐶 𝑝 𝑑 𝑛 1 1 / 𝑝 𝜎 𝑘 ( 𝜏 ) | | | | 𝑓 𝑝 𝑑 𝑛 1 ( 𝑝 1 ) / 𝑝 . ( 2 . 2 6 ) By using (1.20), we get 𝜎 𝑘 ( 𝜏 ) | | 𝑓 𝐶 3 | | 𝑝 𝑑 𝑛 1 1 / 𝑝 𝜇 𝑝 1 / 𝑝 𝜎 𝑘 ( 𝜏 ) 𝜎 𝑘 ( 𝜏 ) | | | | 𝑓 𝑝 𝑑 𝑛 1 1 / 𝑝 , ( 2 . 2 7 ) where 𝐶 3 = 𝐶 3 ( 𝑓 ) is the constant from (1.20). Then by (2.26) and (2.27), 𝜎 𝑘 ( 𝜏 ) | | 𝑓 𝐶 3 | | | | | | 𝑓 𝑝 1 𝑑 𝑛 1 𝜇 𝑝 1 𝜎 𝑘 ( 𝜏 ) 𝜎 𝑘 ( 𝜏 ) | | | | 𝑓 𝑝 𝑑 𝑛 1 , ( 2 . 2 8 ) and by (2.25) we have 𝜈 1 𝐼 1 ( 𝑡 , 𝜏 ) + 𝐶 1 ( 𝑡 ) 𝜈 2 𝜇 𝑝 1 𝜎 𝑘 ( 𝜏 ) 𝜎 𝑘 ( 𝜏 ) | | | | 𝑓 𝑝 𝑑 𝑛 1 ( 2 . 2 9 ) or 𝜈 1 𝐼 1 ( 𝑡 , 𝜏 ) + 𝐶 1 ( 𝑡 ) 𝜈 2 𝜇 𝑝 1 𝜎 𝑘 ( 𝜏 ) 𝑑 𝐼 1 𝑑 𝑡 ( 𝑡 , 𝜏 ) . ( 2 . 3 0 ) By integrating this inequality, we have shown that 𝐼 1 𝑡 , 𝜏 + 𝐶 1 ( 𝑡 ) 𝜈 1 𝐼 1 𝑡 , 𝜏 + 𝐶 1 ( 𝑡 ) 𝜈 1 𝜈 e x p 1 𝜈 2 𝜏 𝜏 𝜇 𝑝 𝜎 𝑘 . ( 𝜏 ) 𝑑 𝜏 ( 2 . 3 1 )

3. Stagnation Zones

Next we apply the Saint-Venant principle to obtain information about stagnation zones of generalized solutions of (1.17). We first consider zones with respect to the Sobolev norm. Other results of this type follow immediately from well-known imbedding theorems.

3.1. Stagnation Zones with Respect to the 𝑊 1 𝑝 -Norm

We rewrite (2.4) and (2.5) in another form. Let 0 < 𝑘 < 𝑛 and let 0 < 𝛼 < 𝛽 . Fix a domain 𝐷 0 in 𝑘 with compact and smooth boundary, and write

𝐷 = 𝐷 0 × 𝑛 𝑘 = 𝑥 𝑛 𝑥 1 , , 𝑥 𝑘 𝐷 0 . ( 3 . 1 ) We write

𝑝 𝑘 ( 𝑥 ) = 𝑝 𝑘 ( 𝑥 ) 𝛼 + 𝛽 2 . ( 3 . 2 ) For 𝑥 𝐷 𝑘 𝛼 , 𝛽 and

𝛽 = 𝛽 𝛼 2 , ( 3 . 3 ) we have

𝛽 < 𝑝 𝑘 ( 𝑥 ) < 𝛽 , ( 3 . 4 ) and we denote

𝐷 𝛽 , 𝑘 = 𝑥 𝐑 𝑛 𝛽 < 𝑝 𝑘 ( 𝑥 ) < 𝛽 . ( 3 . 5 ) Let 𝛽 < 𝜏 𝜏 < 𝛽 . We write

Δ , 𝑘 𝜏 , 𝜏 = 𝑥 𝐷 𝜏 < 𝑝 𝑘 ( 𝑥 ) < 𝜏 , ( 3 . 6 ) 𝐼 2 𝜏 , 𝜏 = Δ , 𝑘 ( 𝜏 , 𝜏 ) | | | | 𝑓 𝑝 𝑑 𝑛 . ( 3 . 7 )

Let 0 < 𝜏 < 𝜏 < 𝛽 . By (2.5) we have, for 𝑡 ( 𝜏 , 𝜏 ) ,

𝐼 2 𝑡 , 𝜏 + 𝐶 4 ( 𝑡 ) 𝜈 1 𝐼 2 𝑡 , 𝜏 + 𝐶 4 ( 𝑡 ) 𝜈 1 𝜈 e x p 1 𝜈 2 𝜏 𝜏 𝜆 1 / 𝑝 𝑝 , 𝑍 𝑓 ( 𝜏 ) 𝜎 , 𝑘 , ( 𝜏 ) 𝑑 𝜏 ( 3 . 8 ) where

𝑍 𝑓 ( 𝜏 ) = 𝑥 𝜕 𝐷 𝑝 𝑘 ( 𝑥 ) = 𝜏 l i m 𝑦 𝑥 . 𝑓 ( 𝑦 ) = 0 ( 3 . 9 )

By choosing the estimate as in (2.14), we also have

𝐼 2 𝜏 + 𝐶 , 𝑡 4 ( 𝑡 ) 𝜈 1 𝐼 2 𝜏 + 𝐶 , 𝑡 4 ( 𝑡 ) 𝜈 1 𝜈 e x p 1 𝜈 2 𝜏 𝜏 𝜆 1 / 𝑝 𝑝 , 𝑍 𝑓 ( 𝜏 ) 𝜎 , 𝑘 , ( 𝜏 ) 𝑑 𝜏 ( 3 . 1 0 ) where

𝜎 , 𝑘 ( 𝑠 ) = 𝑥 Δ , 𝑘 ( , ) 𝑝 𝑘 ( . 𝑥 ) = 𝑠 ( 3 . 1 1 ) By adding these inequalities and noting that 𝐶 4 𝐶 ( 𝑡 ) + 4 ( 𝑡 ) = 0 , we obtain

𝐼 2 𝜏 , 𝑡 + 𝐼 2 𝑡 , 𝜏 𝐼 2 𝜏 , 𝑡 + 𝐼 2 𝑡 , 𝜏 𝜈 × m a x e x p 1 𝜈 2 𝜏 𝜏 𝜆 1 / 𝑝 𝑝 , 𝑍 𝑓 ( 𝜏 ) 𝜎 , 𝑘 𝜈 ( 𝜏 ) 𝑑 𝜏 , e x p 1 𝜈 2 𝜏 𝜏 𝜆 1 / 𝑝 𝑝 , 𝑍 𝑓 ( 𝜏 ) 𝜎 , 𝑘 . ( 𝜏 ) 𝑑 𝜏 ( 3 . 1 2 ) Thus we have the estimate

𝐼 2 𝜏 , 𝜏 𝐼 2 𝜏 , 𝜏 𝜈 × m a x e x p 1 𝜈 2 𝜏 𝜏 𝜆 1 / 𝑝 𝑝 , 𝑍 𝑓 ( 𝜏 ) 𝜎 , 𝑘 𝜈 ( 𝜏 ) 𝑑 𝜏 , e x p 1 𝜈 2 𝜏 𝜏 𝜆 1 / 𝑝 𝑝 , 𝑍 𝑓 ( 𝜏 ) 𝜎 , 𝑘 . ( 𝜏 ) 𝑑 𝜏 ( 3 . 1 3 ) Similarly, from (2.4) we obtain

𝐼 2 𝜏 , 𝜏 𝐼 2 𝜏 , 𝜏 𝜈 × m a x e x p 1 𝜈 2 𝜏 𝜏 𝜇 𝑝 𝜎 , 𝑘 𝜈 ( 𝜏 ) 𝑑 𝜏 , e x p 1 𝜈 2 𝜏 𝜏 𝜇 𝑝 𝜎 , 𝑘 . ( 𝜏 ) 𝑑 𝜏 ( 3 . 1 4 )

From this we obtain the following theorem on stagnation 𝑊 1 𝑝 -zones.

Theorem 3.1. Let 0 < 𝑘 < 𝑛 , 𝛽 > 𝛼 > 0 , and let 𝛽 < 𝜏 𝜏 < 𝛽 where 𝛽 is as in (3.3). If 𝑓 is a solution of (1.17) on 𝐷 with the generalized boundary condition (1.21) on 𝜕 𝐷 𝐺 , where 𝐺 = { 𝑥 𝜕 𝐷 𝑝 𝑘 ( 𝑥 ) = ± 𝛽 } and