• Views 291
• Citations 0
• ePub 17
• PDF 146
`Abstract and Applied AnalysisVolume 2014 (2014), Article ID 260287, 19 pageshttp://dx.doi.org/10.1155/2014/260287`
Research Article

## Semi-Fredholm Solvability in the Framework of Singular Solutions for the (3+1)-D Protter-Morawetz Problem

1Department of Mathematics and Informatics, University of Sofia, 1164 Sofia, Bulgaria
2Department of Mathematics, City University of New York, College of Staten Island, Staten Island, NY 10314, USA

Received 28 February 2014; Accepted 11 May 2014; Published 14 October 2014

Academic Editor: Donal O’Regan

Copyright © 2014 Nedyu Popivanov 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

For the four-dimensional nonhomogeneous wave equation boundary value problems that are multidimensional analogues of Darboux problems in the plane are studied. It is known that for smooth right-hand side functions the unique generalized solution may have a strong power-type singularity at only one point. This singularity is isolated at the vertex of the boundary light characteristic cone and does not propagate along the bicharacteristics. The present paper describes asymptotic expansions of the generalized solutions in negative powers of the distance to . Some necessary and sufficient conditions for existence of bounded solutions are proven and additionally a priori estimates for the singular solutions are obtained.

#### 1. Introduction

In the present paper, boundary value problems for the wave equation in with points , are studied in the domain bounded by the two characteristic cones and the ball , centered at the origin , . The following BVPs were proposed by Protter [1].

Problem P2. Find a solution of the wave equation (1) in which satisfies the boundary conditions and its adjoint problem.

Problem P. Find a solution of the wave equation (1) in which satisfies the adjoint boundary conditions

Protter [1] formulated in 1952 some versions of and in (i.e., in (2+1)-D case) as a multidimensional analogue of the planar Darboux problems with boundary data prescribed on one characteristic and on the noncharacteristic segment. Initially the expectation was that such BVPs are classical solvable for very smooth right-hand side functions. However, soon it became clear that contrary to this traditional belief, unlike the plane Darboux problem, Protter’s problems are not well posed. The reason is that the homogeneous adjoint Problem has an infinite number of nontrivial classical solutions (Tong [2], Popivanov and Schneider [3], and Khe [4]). It is known from [5] that for each there exists a right-hand side function of the wave equation, for which the uniquely determined generalized solution of Problem has a strong power-type singularity like at the origin .

In the present paper we examine the exact behavior of the singular solutions of Problem . In the case when the right-hand side function is harmonic polynomial, the Problem is Fredholm and we find the asymptotic expansion at of the unique generalized solution. On the other hand, in the general case when , the problem is not Fredholm because it has an infinite dimensional cokernel. We show that there are an infinite number of necessary conditions for the existence of bounded solutions. We discuss the semi-Fredholm solvability of Problem and for we prove that the necessary conditions for the existence of bounded solutions are also sufficient.

In a historical perspective, Protter studied Problems and in connection with BVPs for mixed type equations that model transonic flow phenomena. In fact, in [1], he also proposes a multidimensional analogue to the two-dimensional Guderley-Morawetz problem for the Gellerstedt equation of hyperbolic-elliptic type. The Guderley-Morawetz problem describes flows around airfoils and is well studied. The existence of weak solutions and the uniqueness of the strong ones were first established by Morawetz [6] by reducing the problem to a first-order system. Lax and Phillips [7] established that these weak solutions are strong. A survey for the classical 2D mixed-type BVPs and their transonic background can be found in [8]. The domain of Protter’s analogue could be constructed by rotation in of a symmetric planar domain for Guderley-Morawetz problem around the axis of symmetry. As a result the set forms the hyperbolic part of the domain. Although it was expected that the multidimensional mixed-type problems would be similar to the two-dimensional BVPs, for the Protter hyperbolic-elliptic problems a general understanding of the situation is still not at hand. Even the question of well posedness is surprisingly subtle and not completely resolved. One has uniqueness results for quasiregular solutions, a class of solutions introduced by Protter, but there are real obstructions to existence in this class. The Protter problems in the hyperbolic part of the domain illustrate some of the difficulties and differences between the planar BVPs and the multidimensional analogues.

In order to construct the solutions of the homogenous Problem we need the spherical functions in . Traditionally, are defined on the unit sphere (see [9]). For convenience in the discussions that follows, we keep the same notation for the radial extension of the spherical function to ; that is, for . For the definition and properties of the spherical functions see Section 3. For define the functions where the coefficients are with and . Then the functions are classical solutions of the homogeneous adjoint Protter problem.

Lemma 1 (see [10]). The functions are classical solutions from of the homogeneous Problem for .

A necessary condition for the existence of classical solution for the Problem is the orthogonality with respect to the inner product, of the right-hand side function to all functions from Lemma 1. To avoid an infinite number of necessary conditions in the framework of classical solvability, we introduce generalized solutions for the Problem (see the similar definition for the (2+1)-D case in [5]).

Definition 2 (see [10]). A function is called a generalized solution of the Problem in , if the following conditions are satisfied:(1), ;(2)the identity holds for all such that and in a neighborhood of .

This definition allows the generalized solution to have singularity at the origin and there is a uniqueness result (see Theorem 18). Without any additional conditions imposed on the right-hand side function , it is known (see [3, 10]) that the generalized solution may have power type singularity. Alternatively, we will prove the following necessary conditions for the existence of bounded solutions.

Theorem 3. Suppose that there is a bounded generalized solution of the Protter Problem with right-hand side function . Then for all , , .

The proof of Theorem 3 is given in Section 4, but before that we will describe the exact influence of the conditions (10) on the behavior of the generalized solution.

First, we consider the case when the right-hand side function of the wave equation (1) has the representation with . In particular, notice that in the case when the function is a harmonic polynomial in of order , whose coefficients are functions of (see the properties of in Section 3). For convenience further by “harmonic polynomial of order ” we will mean a function from that has the more general form (11). The coefficients are and must have some special properties at (see, e.g., Lemma 17).

According to the results from [10] we know that the generalized solution of Problem may have a power type singularity at the origin , . In the present paper we study more accurately the exact behavior of the solution of Problem at . It is governed by the parameters where ; and . We find the asymptotic formula for the generalized solution of Problem .

Theorem 4. Suppose that the right-hand side function has the form (11). Then the unique generalized solution of Problem belongs to and has the following asymptotic expansion at the singular point : where(i)the function and satisfies the a priori estimate with constant independent of and ;(ii)the functions , , satisfy the equalities with functions bounded and independent of ;(iii)if at least one of the constants in (16) is different from zero, then for the corresponding function there exists a direction with for , such that

After the case of the harmonic polynomials, here we deal with the more general situation when the right-hand side function is smooth, but it cannot be expanded simply as a sum (11). Now, Lemma 1 shows that the Problem is not Fredholm solvable.

Remark 5. Consider the operator where is the unique classical solution to Protter Problem for the right-hand side function . According to Lemma 1 we have . This means that is not Fredholm operator for example in . On the other hand, the uniqueness result Theorem 18 shows that and could be a semi-Fredholm operator. A semi-Fredholm operator is a bounded operator that has a finite dimensional kernel or cokernel and closed range (see, e.g., [11]). Accordingly we need to find the range of .

The next result suggests that is a semi-Fredholm operator.

Theorem 6. Let the function belong to . Then the necessary and sufficient conditions for existence of bounded generalized solution of the Protter Problem are for all , , and .
Moreover, this generalized solution and satisfies the a priori estimates where the constant is independent of the function .

Obviously, the set of all functions from that satisfy the orthogonality conditions (19) is closed. Therefore, Theorem 6 shows that the operator defined in Remark 5 with a domain has a closed range in , and we get the following result.

Corollary 7. The operator is a semi-Fredholm operator from to .

We have briefly announced some of the results from this section in [12] with the assumption .

The main results in this work are discussed in Section 2 and the proofs are in Sections 37. In more detail the paper is organized as follows: estimates for the spherical functions involved in the representation of the solution are proven in Section 3. In Section 4 the necessary conditions for bounded solution Theorem 3 are proved. In Section 5 we consider some two-dimensional boundary value problems connected to Problem , the Problems and . Exact formulas for the solution of the Problem are presented in Lemma 20. In Section 6 the proofs of the main Theorems 4 and 6 are given based on the results from the previous sections and an asymptotic expansion formula for the generalized solution of the Problem (Theorem 22). The long and technical proof of Theorem 22 is postponed to Section 7.

#### 2. Historical Remarks on the Main Results

Let us point out several related recent works on Protter problems. Necessary and sufficient conditions for the existence of solutions with fixed order of singularity were obtained in [10]. Similarly, for the -analogues of Protter problems, some results are presented in [13, 14]. For the problem with Dirichlet type boundary condition on , a formula for the asymptotic expansion of the singular solution can be found in [15], and the semi-Fredholm solvability is discussed in [16] for . A comparison of various recent results for Protter problems is made in [13].

Various authors adopted a variety of approaches to Protter problems over the last sixty years, for example, Wiener-Hopf method, special Legendre functions, a priori estimates, nonlocal regularization, and so forth (see [5] and references therein; see also [4, 10, 14, 1719]). Alternatively, different multidimensional analogues of the classical Darboux problem for the wave equation are considered in [2022], while for some related semilinear equations and systems see [23]. The existence of bounded or unbounded solutions for the wave equation in and , as well as for the Euler-Poisson-Darboux equation, has been studied in [4, 10, 17, 19, 2426].

Regarding the Protter problems with lower order terms see [27] and references therein. Problems with more general boundary condition on are studied in [26, 27]. Some possible regularization methods involving integrodifferential or nonlocal terms can be found in [18].

For the Protter problems for equations of mixed hyperbolic-elliptic type proposed in [1], Aziz and Schneider [28] proved an uniqueness result in the linear case (see also [21]). Concerning nonexistence principle for nontrivial solution of semilinear mixed-type equations in multidimensional case, we refer to [29].

In 1960 Garabedian [30] proved the uniqueness of a classical solution of Proter problem. However, generally, Problem is not classically solvable and a necessary condition for the existence of a classical solution is the orthogonality of the right-hand side function to all solutions of the corresponding homogeneous adjoint Problem . Here, in Lemma 1, the solutions were constructed with the help of the functions defined by (6). The alternate representation in terms of the Gauss hypergeometric function can be found in Khe [4]. In [3] there are some solutions for the three-dimensional analogue of the homogeneous Problem .

Let us look back at Theorem 3 and the necessary orthogonality conditions (10) for the existence of bounded solutions of Problem . Naturally, these conditions include the functions from Lemma 1. However, notice that there are also some others.

Remark 8. It is interesting that conditions (10) include the case of even . Notice that the functions are not classical solutions of the homogenous adjoint Problem . Actually, they satisfy the homogenous wave equation in and vanish on , but is not zero on . In addition, the functions have a singularity at the origin like ; however, this singularity is integrable in the domain .

Instead of imposing an infinite number of orthogonality conditions on , Popivanov and Schneider [3, 5] introduced the concept of generalized solution that allows the solution to have singularity on the inner cone . Here, Theorem 4 describes the effect of the parameters on the behavior of the generalized solution of Problem . The constants are defined by (13) and are obviously related to the orthogonality conditions (10). When the right-hand function is a harmonic polynomial (11), the asymptotic expansion in Theorem 4 shows that the generalized solution could be bounded only if all involved are zero.

Corollary 9. Suppose that the right-hand side function has the form (11) and satisfies the orthogonality conditions for all ; and . Then the unique generalized solution of Problem belongs to , is bounded, and satisfies the a priori estimate

On the other hand, without any orthogonality conditions on , the following result is obtained.

Corollary 10. The generalized solution of Problem with a right-hand side function in the form (11) satisfies the a priori estimate

The influence of the orthogonality conditions (22) on the exact behavior of the generalized solution is clarified by case (iii) of Theorem 4. It shows that, for fixed indexes , the corresponding condition (22) “controls” one power-type singularity.

In Corollaries 9 and 10 the emphasis is on the extreme cases: when all orthogonality conditions (22) are fulfilled or, alternatively, when none of them are satisfied. In both cases the exact behavior of the solution is given. The estimate (24), presented here, is analogous to known estimates for Protter problems in [5] and in [24]. It is interesting that singularities of the generalized solutions are isolated at the origin and do not propagate in the direction of the bicharacteristics on the characteristic cone . Traditionally, it is assumed that the wave equation, with sufficiently smooth right-hand side, cannot have a solution with an isolated singular point as in Hörmander [31, Chapter 24.5]. The case here is different since the point of singularity lies on the noncharacteristic part of the boundary , as well as on the characteristic part .

Remark 11. The Problem in with harmonic polynomial on the right-hand side is also studied in [10]. However, the explicit asymptotic expansion here has no analogue in [10], where only the behavior of the singularities is given. Additionally, if the orthogonality conditions (22) are fulfilled, Corollary 9 states that the generalized solution is in fact bounded, while the estimates in [10, Theorem 1.1] still allow the solution to have some logarithmic singularities.

Remark 12. Let us compare Protter problems in (as treated in [13, 14]) and here (see also [10, 15]). In both cases the study of these BVPs is based on the properties of the special Legendre functions. Instead of Legendre polynomials here, in the three-dimensional case, the Legendre functions with noninteger indexes are used (for their properties see [32]). One can easily modify both these techniques to obtain similar results for the -dimensional problems, for even (analogous to case) or for odd (the present case ). Some related results for Protter problems in are presented in [17, 24].

In the general case when the right-hand side function is smooth enough, Theorem 6 implies that the necessary conditions (10) for existence of bounded solutions from Theorem 3 are also sufficient. Further, this means that there are no other nontrivial classical solutions of the homogenous adjoint Problem except those listed in Lemma 1.

Remark 13. We point out the differences between Theorem 6 and the results from [16] for Protter Problem with Dirichlet type boundary condition on . First, notice that in the case when right-hand side function is harmonic polynomial of order the solution of Problem may have worse singularity (like ; see Theorem 4) than the solution of Problem (like ; see [15]). For the general case we are able to reduce the assumptions on ; in Theorem 6 we assume , while in [16, Theorem 1.1] smoother is required. In order to achieve this, we rely on the more accurate estimates for the special functions proven in Section 3.

#### 3. Estimates for the Special Functions

For the proof of the main results we will need some properties of the spherical functions in . They are naturally expressed on the unit sphere in spherical polar coordinates. Let us introduce polar coordinates in : where , , . Then the spherical functions, expressed in terms of and as in the traditional definition on (see [9]), are given by and . Here are the Legendre polynomials defined by the Rodrigues’ formula while are the associated Legendre polynomials that can be defined as The functions satisfy the differential equation and form a complete orthonormal system in (see [9]).

Using Cartesian coordinates as in Section 1, one can define the spherical functions as for , or by where are constants. In the present paper, we keep the same notation for the radial extension of the spherical function to ; that is, for . According to the properties of , the function is a homogenous harmonic polynomial of order in the variables , , .

We will need some estimates for and the special functions involved in the representations of the solutions of the Protter problems. Let us start with the Legendre polynomials .

Lemma 14. The following estimates hold for :

Proof. The estimates (31) are proved, for example, in [10]. Here we will show that (32) holds. Using Bonnet’s recursion formula we get and Thus From here and the equalities and we get the estimate (32) by induction.

In order to study the first derivatives of the generalized solution of the Protter problem, we will need to estimate also the first derivatives with respect to of , the radial extension of the spherical function to : Using the so-called addition theorem for Legendre polynomials we get the following result.

Lemma 15. For the functions satisfy the equalities

Proof. From the definition (26) of it follows that According to the addition theorem (see [33]) With and one derives the equality This means that Using (39) again, we get the required property of the derivatives of . Directly from (36) for the squares of the derivatives with respect to we have and from the definition (26) of we find
Put , and differentiation of (39) with respect to and gives With we derive
Analogously, after differentiating (39) twice with respect to , Then substituting and we have Since we conclude that

As a direct consequence, we find the next estimates for the radial extension of the spherical functions to .

Corollary 16. For the functions satisfy the inequalities

Proof. Obviously The estimates (50) and (51) follow directly from the Cauchy-Schwarz inequality and Lemma 15: Finally (52) follows from the equality (42):

Generally, if the function is smooth enough, it can be represented as a harmonic series. In order to estimate the coefficients of the series we will use the next result. It is based on the fact that the spherical functions are eigenfunctions for the Laplace operator on the sphere .

Lemma 17. Let .(1)(See [16]). Let . Then it has the representation and for (2)Suppose ; then for

Proof. The estimate in case follows directly from the fact that the spherical functions are eigenfunctions for the spherical Laplacean : That is, (29) shows that . In fact is bounded by the Fourier coefficient of : For detail proof of (1) see [16, 34].
Next we will prove case with . Let ; then Integrating by parts gives where Therefore using (52) we have To prove case it remains to substitute for in this estimate.

#### 4. Necessary Conditions for Bounded Solution

Here we will prove of the necessity of the orthogonality conditions (10) for the existence of a bounded solution.

Proof of Theorem 3. Let be a bounded generalized solution of Problem . Let us fix a function such that for , and for .
For fixed indexes , , , consider the functions for . Obviously, vanishes on a neighborhood of and on . Therefore according to Definition 2 we have and thus We want to prove that for all , , . Notice that and then the integral is convergent. The function and thus Therefore it is sufficient to prove that when , the right-hand side of the equality (67) tends to zero for , , .
Using the fact that the functions are solutions of the homogenous wave equation in , straightforward computations show that where , while and stand for and , respectively. Expressed with spherical polar coordinates in , For simplicity, we consider only one of the terms from the definition of the function (see (6)). Let us denote defined in , and consider the subsets and .
Since the functions , , , and are bounded, the function in , and is zero in , it is sufficient to prove that the integrals tend to zero as , for , , and .
We get from the estimate in and thus For let us compute first because could be negative only when . Therefore, we have and thus .
Finally, to evaluate , we use the estimate and find that shows that .

#### 5. Previous Results

In this section we quote some results from [10] that will be essentially the starting point for the proofs of Theorems 4 and 6. We start with the following uniqueness result.

Theorem 18 (see [10]). The Problem has at most one generalized solution.

In [10] the right-hand side function of the wave equation (1) is fixed as a harmonic polynomial (11). Then the following existence result for the generalized solution is valid.

Theorem 19 (see [10]). Suppose that the right-hand side has the form (11) where . Then, the unique generalized solution of the Problem in exists and has the form

In fact, the function from (78) is the solution of a two-dimensional boundary value problem that involves only the corresponding coefficient from (11). In order to formulate this BVP, it is natural to introduce polar coordinates in : ; and are such that , , and In the special case when has the form according to Theorem 19, we may look for a solution of the same form Then we can reduce the (3+1)-D Protter problem to some BVPs in . From the properties of the spherical functions it follows that the function is a solution of the equation with right-hand side , in the domain , bounded by Thus, we arrive at the next two-dimensional problems.

Problem P21. Find a solution of (82) in the domain which satisfies the boundary conditions

Finally, we substitute and get the following problem.

Problem P22. Find a solution of the equation in the domain with boundary conditions

In [10] the solution of Problem is constructed with the help of the Riemann’s function for equation (86) found by Copson [35]. The problem is reduced to an integral equation of Volterra type. Then this integral equation is solved using some formulas from the book by Samko et al. [36] and the properties of the Mellin transform. According to formulas , , and in the proof of Theorem 3.1 from [10] we can write down the following result.

Lemma 20. The solution of Problem is given by where

Finally, we will need the relation between the functions , with , defined in (6) and the Legendre polynomials .

Lemma 21 (see [10]). Define the functions Then the equality holds for with some nonzero constants .

#### 6. Proofs of the Main Results

In this section we will prove Theorems 4 and 6. The proofs are based on an asymptotic expansion formula for the solution of the two-dimensional Problem . In order to formulate it, let us concentrate first on Problem with right-hand side functions of the form with fixed and . We will use the results stated in Section 5. The unique generalized solution of Problem also has the form The function is the solution of Problem with the function as a right-hand side in (82). We are interested in the exact behavior of at . One expects it to depend on the constants that correspond to defined in (13) for the Problem . For simplicity, denote further

Theorem 22. Let and . Then the generalized solution of Problem belongs to and has the following asymptotic expansion at : where (1)the functions are independent of , bounded, and satisfy ;(2)the function satisfies the following estimates: and for If also and , then and if additionally and , then where in all inequalities the constant is independent of and .

The proof of this result is quite long and technical and we leave it for Section 7. Here we will use Theorem 22 to prove Theorems 4 and 6.

Proof of Theorem 4. Assume that the right-hand side function is a harmonic polynomial (11). Then the unique generalized solution also is a harmonic polynomial (78), according to Theorem 19. Furthermore, the functions are solutions of Problem with right-hand side that can be represented as When we have and obviously The definition of functions from Lemma 1 and (103) gives the identity