Open Access

A generalization on the solvability of integral geometry problems along plane curves

Boundary Value Problems20132013:202

DOI: 10.1186/1687-2770-2013-202

Received: 2 August 2013

Accepted: 19 August 2013

Published: 8 September 2013

Abstract

This paper is concerned with a general condition for the solvability of integral geometry problems along the plane curves of given curvatures. As two important results, the solvabilities of integral geometry problems along the family of circles with fixed radius and along the family of circles of varying radius centered on a fixed circle are given. By using some extension of the class of unknown functions, the proofs are based on the solvabilities of equivalent inverse problems for transport-like equation.

MSC:35R30, 53C65, 65N30.

Keywords

integral geometry problem inverse problem Galerkin method transport-like equation

1 Introduction

The problems of integral geometry are to determine a function, given (weighted) integrals of this function over a family of manifolds, and there has been significant progress in the classical Radon problem when manifolds are hyperplanes and the weight function is unity, there are interesting results in the plane case when a family of curves is regular or in the case of a family of straight lines with arbitrary regular attenuation [[1], Chapter 7]. It is assumed that the basis of the integral geometry problems is the Radon transform [2]. The Radon transform R integrates a function f on R n https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq1_HTML.gif over hyperplanes. Let H ( s , Φ ) = { x R n : x Φ = s } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq2_HTML.gif be the hyperplane perpendicular to Φ S n 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq3_HTML.gif (unit sphere) with signed distance s R 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq4_HTML.gif from the origin, and the Radon transform ( R f ) ( s , Φ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq5_HTML.gif is defined as the integral of f over H ( s , Φ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq6_HTML.gif, i.e.,
( R f ) ( s , Φ ) = H ( s , Φ ) f ( x ) d x https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equa_HTML.gif

(see [[3], Chapter 2]).

The problems of integral geometry have important applications in imaging and provide the mathematical background of tomography, where the main goal is to recover the interior structure of a nontransparent object using external measurements. The object under investigation is exposed to radiation at different angles, and the radiation parameters are measured at the points of observation. The basic problem in computerized tomography is the reconstruction of a function from its line or plane integrals, and there are many applications related with computerized tomography: medical imaging, geophysics, diagnostic radiology, astronomy, seismology, radar and many other fields (see, e.g., [4]).

From the applied point of view, the importance of integral geometry problem over a family of straight lines in the plane is indicated in [5], where the problem models X-rays, and applicable to the problems of radiology and radiotherapy. Because of their many practical applications, a considerable attention has been devoted to other family of curves in the plane as well as straight lines. Invertibility of the Radon transforms on some families of curves in the plane is given with explicit inversion formulas via circular harmonic decomposition in [6] and for the explicit inversion formulas of the attenuated Radon transform, see, e.g., [7, 8]. Note that the circle is the simplest non-trivial curve in the plane next to the straight line, and the representation of a function by its circular Radon transform also arises in applications. In [9], invertibility of the Radon transforms over all translations of a circle of fixed radius and circles of varying radius centered on a fixed circle is considered, where the proofs require microlocal analysis of the Radon transforms and a microlocal Holmgren theorem. In [10], some existence and uniqueness results on recovering a function from its circular Radon transform with partial data are presented and the relations to applications in medical imaging are described. There are several other ways related to the selection of a family of curves, such as circles of varying radius centered on a straight line or a fixed curve, circles passing through a fixed point, along paths that are not on the zero sets of harmonic polynomials, circular arcs having a chord of fixed length rotating around its middle point etc., which are meaningful in applications on thermo-acoustic and photoacoustic tomography, synthetic aperture radar, Compton scattering tomography, ultrasound tomography etc. (see, e.g., [1012] and the references therein).

In fact, since the seminal work of Radon [2], the various integral geometry problems with numerous applications have been considered in several important aspects which are not mentioned here, but for a comprehensive list, see, e.g., [1316] and the references therein. Furthermore, the problems of integral geometry and inverse problems for transport equations are interrelated and the latter are also of great importance in theory and applications; see, e.g., [1720] and for the derivation and applications of transport equations, see, e.g., [2123].

In this paper, a general condition for the solvability of integral geometry problems along plane curves of given curvatures is presented and its relation with some previous results is indicated. Moreover, as two important results, the solvabilities of integral geometry problems along the family of circles with fixed radius and along the family of circles of varying radius centered on a fixed circle are given. Since the curvature of a circle is defined to be the reciprocal of the radius of the circle, in the former case the curvature is a constant, while in the latter one the curvature depends on the point and the direction. To investigate the solvability of the integral geometry problem (IGP) given in Section 2.1, which is overdetermined since the underlying operator of the IGP is compact and its inverse operator is unbounded (see Section 2.2), it is reduced to an equivalent overdetermined inverse problem for a transport-like equation, and then, with the use of a similar method which was proposed in [24] (see also [[13], Chapter 1]), on using some extension of the class of unknown functions, this inverse problem is replaced by a determined one. Thus, the solvability of IGP is proved via the solvability of an inverse problem for a transport-like equation. The above mentioned method on the solvability was also previously utilized in [2527] for IGP along some family of plane curves of given curvatures and straight lines, in [28] for IGP along geodesics and in [29] for IGP along the family of curves whose curvatures are given by the Christoffel symbols. Here, the presented general condition for the solvability covers those of [2528], in the manner indicated by Remark 1 in Section 3.1. Moreover, these previous solvability conditions do not hold for IGP along the above given two families of circles, and this is the main importance and motivation of this study.

In Section 2, IGP and its reduction to the equivalent inverse problem for a transport-like equation, the method to overcome the difficulty on investigating the solvability arising from overdeterminacy of these problems are presented and some definitions and notations which will be used throughout the paper are introduced. Section 3 is devoted to the statements of main results, and finally in Section 4, proofs of the results are given.

The investigation of approximate solutions of the concerned integral geometry problems is beyond the scope of this paper, but similar procedures as in [26] can be carried out by using the Galerkin method or the finite difference method.

2 Statement of the problem and overdeterminacy

2.1 Statement of the problem

Let D be a bounded domain in R 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq7_HTML.gif. It is assumed that in D, a family of regular curves is given by curvature K ( x , φ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq8_HTML.gif which is the curvature of the curve passing from the point x = ( x 1 , x 2 ) D https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq9_HTML.gif in the direction ν = ( cos φ , sin φ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq10_HTML.gif, and there exists a unique sufficiently smooth curve of this family which is passing from any point x D https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq11_HTML.gif in the arbitrary direction ν, with the endpoints on the boundary of D. Suppose that the lengths of these curves in D are bounded above by the same constant. Let us denote the family of these curves by { Γ } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq12_HTML.gif. IGP is stated below.

IGP Determine a function λ ( x ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq13_HTML.gif in the domain D from the integrals of λ along the curves of a given family of curves { Γ } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq12_HTML.gif.

Suppose that λ ( x ) C ( R 2 ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq14_HTML.gif vanishes outside D, and let us introduce an auxiliary function
u ( x , φ ) = γ ( x , φ ) λ d σ , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equ1_HTML.gif
(1)

where γ ( x , φ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq15_HTML.gif is a part of the curve that belongs to { Γ } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq16_HTML.gif, with one end of it being the point x and the other one on ∂D, and is the arc length element along γ ( x , φ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq15_HTML.gif.

Investigating the uniqueness of a solution of a problem of integral geometry by reducing it to the equivalent inverse problem for a differential equation was first carried out in [30]. Similar reduction is demonstrated for IGP formulated below.

Differentiating (1) in the direction ν at x, we obtain the following transport-like equation:
L u u x 1 cos φ + u x 2 sin φ + K ( x , φ ) u φ = λ ( x ) . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equ2_HTML.gif
(2)
From (1), u is 2π-periodic with respect to φ, and since the integrals of λ along the curves of { Γ } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq17_HTML.gif are known, u is known on D × ( 0 , 2 π ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq18_HTML.gif, i.e.,
u | D × ( 0 , 2 π ) = u 0 ( x , φ ) , u ( x , φ ) = u ( x , φ + 2 π ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equ3_HTML.gif
(3)

(see [30] and [[13], p.11]). So, we have the following inverse problem.

Problem 1 Determine a pair of functions ( u , λ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq19_HTML.gif from the transport-like equation (2) provided that the function K is known and u satisfies conditions (3).

2.2 Overdeterminacy

Generally, in the theory of integral geometry, reconstruction of a function of n variables from a function of m > n https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq20_HTML.gif variables is said to be an overdetermined problem of integral geometry (see, e.g., [[14], Chapter 5]), and in the theory of inverse problems, overdetermination usually means that the number m of independent variables in the data exceeds the number n of independent variables in the unknown target function (see, e.g., [[31], Section 1.3]). However, since the data u 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq21_HTML.gif given on the two-dimensional surface D × ( 0 , 2 π ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq18_HTML.gif and λ ( x ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq13_HTML.gif is a function of two variables, these are not the cases for IGP or Problem 1. Here, since the operator given in (1) (for γ ( x , φ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq15_HTML.gif, where ( x , φ ) D × ( 0 , 2 π ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq22_HTML.gif) is compact, its inverse is unbounded, and therefore it is not possible to prove a general existence result. So, IGP and Problem 1 are called overdetermined in this sense. Hence, because of the overdeterminacy, the initial data for these problems should not be arbitrary and satisfy some ‘solvability conditions’ (see [[13], p.4] and [[15], p.18, Theorem 1.4]) which are difficult to establish. It should be noted that the set of functions u 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq21_HTML.gif for which IGP is solvable is not everywhere dense in any of the spaces L 2 ( D × ( 0 , 2 π ) ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq23_HTML.gif, C m ( D × ( 0 , 2 π ) ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq24_HTML.gif and H m ( D × ( 0 , 2 π ) ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq25_HTML.gif. Moreover, the data in problems of integral geometry are of quasianalytic character, i.e., their values specified in a domain of the Lebesgue measure can be as small as desired, determine their values in an essentially larger domain (see [[32], Chapter 6, Section 17] and [[33], Chapter 6, Section 1]). In particular, this implies that it is impossible to avoid overdeterminacy of the problem by specifying the data on a part of the boundary rather than on the whole boundary. Even if it were possible to find the solvability conditions for the mentioned overdetermined problems, since the real data usually have some errors in practice, and thus fall out of the data class for which the existence of a solution is established, it appears that these conditions would not always be satisfactory in applications. Therefore, to prove the existence results, such special conditions on the data u 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq21_HTML.gif have to be posed.

Let us propose the procedure for establishing the solvability of IGP. Assume that the unknown function λ in IGP depends not only upon the space variables x, but also upon the direction φ in some special manner, i.e., consider λ ( x , φ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq26_HTML.gif, where this dependence upon φ is impossible to be arbitrary, for in the opposite case the problem would be underdetermined and the examples on the nonuniqueness of a solution can be easily constructed. Herein the special dependence of λ ( x , φ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq26_HTML.gif upon the direction means that λ ( x , φ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq26_HTML.gif satisfies a certain differential equation ( L ˆ λ = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq27_HTML.gif, where the expression of L ˆ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq28_HTML.gif is given in Section 2.3) with the following properties:
  1. (1)

    The IGP or Problem 1 with the function λ ( x , φ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq26_HTML.gif becomes a determined one.

     
  2. (2)

    The sufficiently smooth functions λ depending only on x satisfy this equation.

     

Suppose that a differential equation for λ ( x , φ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq26_HTML.gif satisfying properties (1) and (2) has been found and that a priori the function u 0 e https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq29_HTML.gif, which represents the exact data of IGP related to a function λ depending only on x, is known. Then, utilizing u 0 e https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq29_HTML.gif, a solution λ ˜ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq30_HTML.gif to IGP can be constructed. By uniqueness of a solution, λ ˜ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq30_HTML.gif and λ ( x ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq13_HTML.gif coincide. At the same time, knowing the approximate data u 0 a https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq31_HTML.gif with u 0 e u 0 a H 3 ( D × ( 0 , 2 π ) ) ε https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq32_HTML.gif, an approximate solution λ a ( x , φ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq33_HTML.gif can be constructed such that λ λ a L 2 ( Ω ) ε C https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq34_HTML.gif. Recall that if λ depends only on x and u 0 a https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq31_HTML.gif does not satisfy the ‘solvability conditions’, the solution λ a https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq35_HTML.gif depending only x does not exist. Here the data are specified on D × ( 0 , 2 π ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq18_HTML.gif and C > 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq36_HTML.gif is independent of u 0 e https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq29_HTML.gif and u 0 a https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq31_HTML.gif. In other words, a regularizing procedure is constructed for the IGP.

In general, the equation with the properties (1) and (2) for the same problem is not uniquely defined. Hence, the class of unknown functions λ extends so that IGP for this class becomes a determined problem and all sufficiently smooth functions in x belong to it. On using some extension of the class of functions λ, the overdetermined Problem 1 is replaced by a determined one (Problem 2 in Section 3.1).

The above method of solvability of the IGP or Problem 1 leads to the Dirichlet-type problem with conditions (3) for the third-order equation of the form A u L ˆ L u = F https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq37_HTML.gif. In investigating and proving the solvability of IGP over any regular family { Γ } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq12_HTML.gif of curves with curvature K, since the quadratic form J ( u ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq38_HTML.gif in (14) is required to be positive definite, the construction of L ˆ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq28_HTML.gif is important and to be able to this, in Theorem 1 condition (6) is given.

2.3 Definitions and notations

In this section, some notations are given based on [13]. Let D R 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq39_HTML.gif with the boundary D C 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq40_HTML.gif, Ω = { ( x , φ ) : x D , φ ( 0 , 2 π ) } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq41_HTML.gif and Ω ¯ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq42_HTML.gif be the closure of Ω. By ( u , v ) L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq43_HTML.gif we denote a scalar product of functions u and v in L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq44_HTML.gif and by C 0 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq45_HTML.gif the set of all functions defined in Ω which have continuous partial derivatives of order up to all k < https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq46_HTML.gif, whose supports are compact subsets of Ω. For a differential expression A, by A https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq47_HTML.gif we denote the conjugate of A in the sense of Lagrange. For x = ( x 1 , x 2 ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq48_HTML.gif,
x i u = u x i = u x i ( i = 1 , 2 ) , φ u = u φ = u φ , | x u | 2 = u x 1 2 + u x 2 2 and | x , φ u | 2 = | x u | 2 + u φ 2 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equb_HTML.gif

Let C π 3 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq49_HTML.gif denote the set of real-valued functions u C 3 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq50_HTML.gif that are 2π-periodic with respect to φ in the domain Ω, i.e., x 1 α 1 x 2 α 2 φ α 3 u ( x , 0 ) = x 1 α 1 x 2 α 2 φ α 3 u ( x , 2 π ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq51_HTML.gif, where α i https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq52_HTML.gif are nonnegative integers such that 0 α 1 + α 2 + α 3 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq53_HTML.gif.

The proof of Theorem 1 involves energy-like estimates and the Galerkin method (see, e.g., [[34], Chapter 5, Section 2.3], [[35], Chapter 7]), and therefore some class of functions are introduced below. In C π 3 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq54_HTML.gif, let us introduce the scalar product
( u , v ) 1 , 2 = Ω ( u v + i = 1 2 ( u x i v x i + u x i φ v x i φ ) + u φ v φ + u φ φ v φ φ ) d Ω , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equc_HTML.gif

where d Ω = d x 1 d x 2 d φ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq55_HTML.gif and set u 1 , 2 = [ ( u , u ) 1 , 2 ] 1 / 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq56_HTML.gif. Let H 1 , 2 π ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq57_HTML.gif and H m π ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq58_HTML.gif be the completions of C π 3 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq49_HTML.gif with respect to the norms 1 , 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq59_HTML.gif and H m ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq60_HTML.gif ( m = 1 , 2 , 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq61_HTML.gif), respectively (for the space H m https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq62_HTML.gif, see, e.g., [34, 36]).

Let C π 0 3 = { w : w | D × ( 0 , 2 π ) = 0 , w C π 3 ( Ω ) } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq63_HTML.gif, and H ˚ 1 , 2 π ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq64_HTML.gif and H ˚ m π ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq65_HTML.gif be the completions of C π 0 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq66_HTML.gif with respect to the norms 1 , 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq59_HTML.gif and H m ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq60_HTML.gif ( m = 1 , 2 , 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq61_HTML.gif). Let us take a set { w 1 , w 2 , } C π 0 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq67_HTML.gif which is complete and orthonormal in L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq44_HTML.gif, then we may assume that the linear span of this set is everywhere dense in H ˚ 1 , 2 π ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq64_HTML.gif. Since H ˚ 1 , 2 π ( Ω ) H ˚ 1 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq68_HTML.gif is separable, there exists a countable set { φ i } i = 1 C π 0 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq69_HTML.gif which is everywhere dense in this space and this set up can be extended to a set which is everywhere dense in L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq44_HTML.gif. Orthonormalizing the latter in L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq70_HTML.gif, we obtain { w 1 , w 2 , } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq71_HTML.gif. We denote by P n https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq72_HTML.gif the orthogonal projector of L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq73_HTML.gif onto M n https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq74_HTML.gif which is the linear span of { w 1 , w 2 , , w n } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq75_HTML.gif.

Let
L ˆ u = l ( φ u ) , l = ( sin φ ) x 1 ( cos φ ) x 2 + g φ + g φ , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equd_HTML.gif
where g ( x , φ ) C π 1 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq76_HTML.gif and it can be easily verified that
( l ) = ( sin φ ) x 1 + ( cos φ ) x 2 g φ . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Eque_HTML.gif

The existence of the function g in the expression of l https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq77_HTML.gif leads to a generalization on conditions for the solvability of integral geometry problems. In Theorem 1, it is shown that if there exists a function g satisfying condition (6) which depends on the curvature K and the domain D, the solvability holds.

Let A u L ˆ L u https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq78_HTML.gif and Γ ( A ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq79_HTML.gif be the set of all functions u L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq80_HTML.gif such that for any u Γ ( A ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq81_HTML.gif there exists y L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq82_HTML.gif such that A u = y https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq83_HTML.gif in the generalized functions sense, i.e., ( u , A η ) L 2 ( Ω ) = ( y , η ) L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq84_HTML.gif holds for every η C 0 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq85_HTML.gif. Take a subset Γ ( A ) Γ ( A ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq86_HTML.gif such that for any u Γ ( A ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq87_HTML.gif there exists a sequence { u k } C π 0 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq88_HTML.gif such that u k u https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq89_HTML.gif weakly in L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq44_HTML.gif and ( A u k , u k ) L 2 ( Ω ) ( A u , u ) L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq90_HTML.gif as k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq91_HTML.gif. If we denote the closure of C π 0 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq66_HTML.gif with respect to the norm u Γ ( A ) = u L 2 ( Ω ) + A u L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq92_HTML.gif by Γ ( A ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq93_HTML.gif, then we have Γ ( A ) Γ ( A ) Γ ( A ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq94_HTML.gif and it can be shown that the inclusions H ˚ 3 π ( Ω ) Γ ( A ) H ˚ 1 , 2 π ( Ω ) Γ ( A ) L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq95_HTML.gif hold.

3 Statements of results

3.1 Solvability of IGP along plane curves

Since Problem 1 is overdetermined, as indicated in Section 2.2, we consider the following determined problem.

Problem 2 Determine a pair of functions ( u , λ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq19_HTML.gif defined in Ω that satisfies
L u = λ ( x , φ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equ4_HTML.gif
(4)

provided that L ˆ λ = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq27_HTML.gif, u is 2π-periodic with respect to φ, u | D × ( 0 , 2 π ) = u 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq96_HTML.gif and K are known.

In (4), it is assumed that the unknown function λ depends also on φ and the condition L ˆ λ = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq27_HTML.gif holds in the generalized functions sense.

If u 0 C 3 ( D × ( 0 , 2 π ) ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq97_HTML.gif and D C 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq40_HTML.gif, then there exists a function G C π 3 ( Ω ¯ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq98_HTML.gif (see [[34], p.130, Theorem 2]) such that G | D × ( 0 , 2 π ) = u 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq99_HTML.gif, and we can consider the new unknown function u ¯ = u G https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq100_HTML.gif. Hence from (4) we obtain the equation L u ¯ = λ + F https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq101_HTML.gif , where F = L G https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq102_HTML.gif and u ¯ | D × ( 0 , 2 π ) = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq103_HTML.gif. Let us denote u ¯ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq104_HTML.gif again by u for simplicity, then Problem 2 can be reduced to Problem 3 given below (see [[13], p.20]) and the solvability of the former follows from that of the latter and does not depend on the choice of G.

Problem 3 Determine a pair of functions ( u , λ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq19_HTML.gif defined in Ω that satisfies
L u = λ ( x , φ ) + F , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equ5_HTML.gif
(5)

provided that L ˆ λ = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq27_HTML.gif, u is 2π-periodic with respect to φ, u | D × ( 0 , 2 π ) = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq105_HTML.gif, K and F are known.

The existence, uniqueness and the stability of the solution of Problem 3 are given by the following theorem.

Theorem 1 If F H 2 π ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq106_HTML.gif and a function g ( x , φ ) C π 1 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq107_HTML.gif exists such that
g x 1 cos φ + g x 2 sin φ + g φ K K x 1 sin φ + K x 2 cos φ K φ g > K 2 + g 2 , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equ6_HTML.gif
(6)
for all ( x , φ ) D ¯ × ( 0 , 2 π ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq108_HTML.gif, then Problem 3 has a unique solution ( u , λ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq19_HTML.gif such that u Γ ( A ) H ˚ 1 π ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq109_HTML.gif, λ L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq110_HTML.gif and
u H ˚ 1 π ( Ω ) + λ L 2 ( Ω ) C ( F L 2 ( Ω ) + F φ L 2 ( Ω ) ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equ7_HTML.gif
(7)

holds, where C > 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq36_HTML.gif depends on K and the Lebesgue measure of D.

Remark 1 In fact, without being aware of (6), the function g, with the appropriate choices of it, was used previously in [25, 26, 28]. The convenience of (6) with those of previous solvability results is indicated below.
  1. (i)

    In [26], when the curvature K is sufficiently smooth and 2π-periodic, the condition for the solvability is ( sin φ ) K x 1 + ( cos φ ) K x 2 > K 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq111_HTML.giffor all ( x , φ ) D ¯ × ( 0 , 2 π ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq108_HTML.gif, where this condition holds for g = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq112_HTML.gif in (6).

     
  2. (ii)

    In [25], when K ( x , φ ) = f 2 ( x ) cos φ f 1 ( x ) sin φ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq113_HTML.gif, where f 1 ( x ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq114_HTML.gif and f 2 ( x ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq115_HTML.gif are sufficiently smooth functions, the condition for the solvability is f 1 x 1 + f 2 x 2 > 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq116_HTML.gif for all x D ¯ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq117_HTML.gif, which holds for g ( x , φ ) = f 1 ( x ) cos φ + f 2 ( x ) sin φ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq118_HTML.gif in (6).

     
  3. (iii)

    In [28], when K ( x , φ ) = f 1 ( x , φ ) cos φ + f 2 ( x , φ ) sin φ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq119_HTML.gif, where f 1 ( x , φ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq120_HTML.gif and f 2 ( x , φ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq121_HTML.gif are sufficiently smooth 2π-periodic functions, the condition for the solvability is f 1 x 2 f 2 x 1 + f 2 f 1 φ f 1 f 2 φ > 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq122_HTML.gif for all ( x , φ ) D ¯ × ( 0 , 2 π ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq123_HTML.gif, and this condition holds for g ( x , φ ) = f 1 ( x , φ ) sin φ f 2 ( x , φ ) cos φ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq124_HTML.gif in (6).

     

Note that in none of the above cases, the solvability conditions hold for the IGP along the family of circles of varying radius centered on a fixed circle and the family of curves of constant curvatures, i.e., the family of circles with fixed radius where the curvature is a nonzero constant or the family of straight lines where the curvature is zero. The former cases are investigated in Section 3.2 below and the latter case is considered in [27], where the term K u φ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq125_HTML.gif in the expression of Lu will not be present since K = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq126_HTML.gif and the proof of solvability is given for g = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq112_HTML.gif. In fact, the strict inequality in (6) can be written as a non-strict inequality, with the equality only for K = g = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq127_HTML.gif.

3.2 Solvability of IGP along the family of circles

Since Theorem 1 was given for IGP along a regular family of plane curves for the general case, in this section the given results on the solvabilities of IGP along the family of circles depend on finding an appropriate function g satisfying (6) for the given curvature K and the domain D under the assumptions of Theorem 1.

3.2.1 Solvability of IGP along the family of circles with fixed radius

Let D R 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq39_HTML.gif and M = sup ( x 1 , x 2 ) D ( x 1 2 + x 2 2 ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq128_HTML.gif. Let us take the family of curves in IGP as the family of circular arcs (the segments of circles inside D with the endpoints on ∂D) with fixed radius r, passing from the point ( x 1 , x 2 ) D https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq129_HTML.gif in the direction ν = ( cos φ , sin φ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq130_HTML.gif and denote this family of circles by { Γ r } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq131_HTML.gif. Since the curvature of a circle is defined to be the reciprocal of the radius of this circle, the curvature of the elements of the family { Γ r } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq132_HTML.gif becomes K = 1 r https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq133_HTML.gif.

The solvability of IGP along the curves of { Γ r } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq131_HTML.gif follows from the following lemma.

Lemma 1 Let us define g on D × ( 0 , 2 π ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq134_HTML.gif as
g ( x 1 , x 2 , φ ) = 1 3 M ( x 1 cos φ + x 2 sin φ ) . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equ8_HTML.gif
(8)

If r > 3 M https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq135_HTML.gif, then (6) holds for K = 1 r https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq133_HTML.gif.

Remark 2 The above choice of g is not unique, and since the curvature K = 1 r https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq136_HTML.gif is constant, and hence K x 1 = K x 2 = K φ = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq137_HTML.gif, condition (6) reduces to
g x 1 cos φ + g x 2 sin φ + g φ 1 r > 1 r 2 + g 2 , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equ9_HTML.gif
(9)

for all ( x , φ ) D ¯ × ( 0 , 2 π ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq108_HTML.gif, which depends on r and the domain D.

3.2.2 Solvability of IGP along the family of circles of varying radius centered on a fixed circle

Let D R 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq39_HTML.gif and M = sup ( x 1 , x 2 ) D ( x 1 2 + x 2 2 ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq128_HTML.gif. Let D R https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq138_HTML.gif be a fixed disk of radius R > 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq139_HTML.gif and, without any loss of generality, centered at the origin such that R > M https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq140_HTML.gif, and take the family of curves in IGP as the family of circular arcs (the segments of circles inside D with the endpoints on ∂D) of varying radius r ( x 1 , x 2 , φ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq141_HTML.gif centered on D R https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq142_HTML.gif, passing from the point ( x 1 , x 2 ) D https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq129_HTML.gif in the direction ν = ( cos φ , sin φ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq143_HTML.gif and denote this family of circular arcs by { Γ R } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq144_HTML.gif. It can be shown that the radius of the circles of the family { Γ R } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq144_HTML.gif is defined by the function
r ( x 1 , x 2 , φ ) = x 1 sin φ x 2 cos φ + ( R 2 ( x 1 cos φ + x 2 sin φ ) 2 ) 1 / 2 , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equ10_HTML.gif
(10)
and hence the curvature of the elements of the family { Γ R } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq144_HTML.gif is defined by
K ( x 1 , x 2 , φ ) = ( x 1 sin φ x 2 cos φ + ( R 2 ( x 1 cos φ + x 2 sin φ ) 2 ) 1 / 2 ) 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equ11_HTML.gif
(11)

on D × ( 0 , 2 π ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq134_HTML.gif. Note that since R > M https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq140_HTML.gif, we have r ( x 1 , x 2 , φ ) > 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq145_HTML.gif and K ( x 1 , x 2 , φ ) > 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq146_HTML.gif on D × ( 0 , 2 π ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq147_HTML.gif.

The solvability of IGP along the curves of { Γ R } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq144_HTML.gif follows from Lemma 2 given below.

Lemma 2 If the function g is defined on D × ( 0 , 2 π ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq134_HTML.gif as
g ( x 1 , x 2 , φ ) = 1 2 M ( x 1 cos φ + x 2 sin φ ) , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equ12_HTML.gif
(12)

then (6) holds for K defined in (11).

4 Proof of results

Proof of Theorem 1 First, we will prove the uniqueness of the solution ( u , λ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq19_HTML.gif of Problem 3 under the assumptions of the theorem. To this end, it is sufficient to show that the corresponding homogeneous problem has only a trivial solution. Then, by taking into account that L ˆ λ = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq148_HTML.gif and F = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq149_HTML.gif, from (5) we obtain L ˆ L u = A u = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq150_HTML.gif.

Since u Γ ( A ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq87_HTML.gif, there exists a sequence { u k } C π 0 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq151_HTML.gif such that u k u https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq89_HTML.gif weakly in L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq44_HTML.gif and ( A u k , u k ) L 2 ( Ω ) ( A u , u ) L 2 ( Ω ) = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq152_HTML.gif as k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq91_HTML.gif. Now we want to decompose the product ( A u k ) u k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq153_HTML.gif into the sum of a positive definite quadratic form and a divergence form. For this purpose, we have the following identities:
( A u k ) u k = ( L ˆ L u k ) u k = ( l ( φ L u k ) ) u k = φ L u k ( l ) u k + x 1 ( u k ( φ L u k ) sin φ ) x 2 ( u k ( φ L u k ) cos φ ) + φ ( u k ( φ L u k ) g ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equ13_HTML.gif
(13)
and
2 φ L u k ( l ) u k = 2 ( u k x 1 φ cos φ u k x 1 sin φ + u k x 2 φ sin φ + u k x 2 cos φ + K φ u k φ + K u k φ φ ) ( u k x 1 sin φ + u k x 2 cos φ g u k φ ) = u k x 1 2 + u k x 2 2 + 2 K u k φ ( u k x 1 cos φ + u k x 2 sin φ ) + 2 g u k φ ( u k x 1 sin φ u k x 2 cos φ ) + ( g x 1 cos φ + g x 2 sin φ + g φ K K x 1 sin φ + K x 2 cos φ g K φ ) u k φ 2 + x 1 ( u k x 2 u k φ + K u k φ 2 sin φ g u k φ 2 cos φ ) x 2 ( u k x 1 u k φ + K u k φ 2 cos φ + g u k φ 2 sin φ ) + φ ( u k x 1 2 sin φ cos φ + u k x 2 2 sin φ cos φ + u k x 1 u k x 2 cos 2 φ 2 K u k x 1 u k φ sin φ + 2 K u k x 2 u k φ cos φ g K u k φ 2 ) . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equf_HTML.gif
If (6) holds, then the quadratic form J ( u k ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq154_HTML.gif in u k x 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq155_HTML.gif, u k x 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq156_HTML.gif, u k φ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq157_HTML.gif is positive definite, where
J ( u k ) = u k x 1 2 + u k x 2 2 + 2 K u k φ ( u k x 1 cos φ + u k x 2 sin φ ) + 2 g u k φ ( u k x 1 sin φ u k x 2 cos φ ) + ( g x 1 cos φ + g x 2 sin φ + g φ K K x 1 sin φ + K x 2 cos φ g K φ ) u k φ 2 = u k x 1 2 + u k x 2 2 + 2 u k φ u k x 1 ( K cos φ + g sin φ ) + 2 u k φ u k x 2 ( K sin φ g cos φ ) + ( g x 1 cos φ + g x 2 sin φ + g φ K K x 1 sin φ + K x 2 cos φ g K φ ) u k φ 2 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equ14_HTML.gif
(14)
Indeed, we can estimate the terms 2 u k φ u k x 1 ( K cos φ + g sin φ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq158_HTML.gif and 2 u k φ u k x 2 ( K sin φ g cos φ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq159_HTML.gif as follows:
2 u k φ u k x 1 ( K cos φ + g sin φ ) ε u k x 1 2 ε 1 ( K cos φ + g sin φ ) 2 u k φ 2 , 2 u k φ u k x 2 ( K sin φ g cos φ ) ε u k x 2 2 ε 1 ( K sin φ g cos φ ) 2 u k φ 2 , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equg_HTML.gif
where 0 < ε < 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq160_HTML.gif, and we obtain
J ( u k ) u k x 1 2 + u k x 2 2 ε u k x 1 2 ε u k x 2 2 ε 1 ( K cos φ + g sin φ ) 2 u k φ 2 ε 1 ( K sin φ g cos φ ) 2 u k φ 2 + ( g x 1 cos φ + g x 2 sin φ + g φ K K x 1 sin φ + K x 2 cos φ g K φ ) u k φ 2 = ( 1 ε ) ( u k x 1 2 + u k x 2 2 ) + ( g x 1 cos φ + g x 2 sin φ + g φ K K x 1 sin φ + K x 2 cos φ g K φ ε 1 ( K 2 + g 2 ) ) u k φ 2 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equh_HTML.gif
Moreover, whenever (6) holds, for sufficiently close value of ε to 1, there exists an α R https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq161_HTML.gif such that g x 1 cos φ + g x 2 sin φ + g φ K K x 1 sin φ + K x 2 cos φ K φ g ε 1 ( K 2 + g 2 ) α > 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq162_HTML.gif in Ω, and we obtain
J ( u k ) ( 1 ε ) | x u k | 2 + α u k φ 2 β | x , φ u k | 2 , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equi_HTML.gif

where β = min { ( 1 ε ) , α } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq163_HTML.gif.

Since the domain D is bounded, u k = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq164_HTML.gif on D × ( 0 , 2 π ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq18_HTML.gif and J ( u k ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq154_HTML.gif is positive definite, we have
u k L 2 ( Ω ) 2 C 0 Ω | x u k | 2 d Ω C Ω J ( u k ) d Ω , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equj_HTML.gif

where C = C 0 β 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq165_HTML.gif and C 0 > 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq166_HTML.gif depends on the Lebesgue measure of D and does not depend on k.

Thus, since u k C π 0 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq167_HTML.gif, K and g are 2π-periodic with respect to φ, after integrating (13) over Ω, the divergent terms disappear and we obtain
2 ( A u k , u k ) L 2 ( Ω ) = Ω J ( u k ) d Ω . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equ15_HTML.gif
(15)
Since u Γ ( A ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq87_HTML.gif, from (15) we have
u L 2 ( Ω ) 2 lim ̲ k u k L 2 ( Ω ) 2 C lim k Ω J ( u k ) d Ω = 2 C lim k ( A u k , u k ) L 2 ( Ω ) = 0 , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equ16_HTML.gif
(16)

which implies that u = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq168_HTML.gif, and since F = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq149_HTML.gif, from (5) we get λ = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq169_HTML.gif. So, the uniqueness part of the proof is completed.

Now we will prove that there exists a solution ( u , λ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq19_HTML.gif of Problem 3 in ( Γ ( A ) H ˚ 1 π ( Ω ) ) × L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq170_HTML.gif by means of the following auxiliary problem.

Determine u defined in Ω that satisfies
A u = F , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equ17_HTML.gif
(17)
u | D × ( 0 , 2 π ) = 0 , u ( x , 0 ) = u ( x , 2 π ) , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equ18_HTML.gif
(18)

where F = L ˆ F https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq171_HTML.gif.

The solution u of problem (17)-(18) will be approximated by
u N = i = 1 N α N i w i ( x , φ ) ; α N = ( α N 1 , α N 2 , , α N N ) R N , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equk_HTML.gif
construction of which is based on finding the vector α N https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq172_HTML.gif from the system of linear algebraic equations
( A u N , w j ) L 2 ( Ω ) = ( F , w j ) L 2 ( Ω ) , j = 1 , 2 , , N , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equ19_HTML.gif
(19)

where the system of functions { w j } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq173_HTML.gif is taken as indicated in Section 2.3 and u N = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq174_HTML.gif on D × ( 0 , 2 π ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq18_HTML.gif.

We must show that the solution of system (19) exists and is unique for any F H 2 π ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq106_HTML.gif. To demonstrate this, let us assume that the homogeneous version of (19), i.e., the system
( A u N , w j ) L 2 ( Ω ) = 0 , j = 1 , 2 , , N , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equl_HTML.gif
has a nonzero solution α ¯ N = ( α ¯ N 1 , α ¯ N 2 , , α ¯ N N ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq175_HTML.gif. Substituting α ¯ N https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq176_HTML.gif for α N https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq172_HTML.gif, multiplying the j th equation of the above homogenous system by 2 α ¯ N j https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq177_HTML.gif and summing with respect to j from 1 to N, we obtain
2 ( A u ¯ N , u ¯ N ) L 2 ( Ω ) = 0 , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equ20_HTML.gif
(20)
where u ¯ N = i = 1 N α ¯ N i w i https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq178_HTML.gif. So, from (15) and (20) we obtain
2 ( A u ¯ N , u ¯ N ) L 2 ( Ω ) = Ω J ( u ¯ N ) d Ω = 0 , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equm_HTML.gif

and since the quadratic form J ( u ¯ N ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq179_HTML.gif defined in (14) is positive definite and u ¯ N = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq180_HTML.gif on D × ( 0 , 2 π ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq181_HTML.gif, we have u ¯ N = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq180_HTML.gif in Ω. But { w i } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq182_HTML.gif is linearly independent and this implies that α ¯ N i = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq183_HTML.gif, i = 1 , 2 , , N https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq184_HTML.gif, which contradicts with the assumption α ¯ N 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq185_HTML.gif. So, it is shown that system (19) has a unique solution α N https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq186_HTML.gif for any F H 2 π ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq106_HTML.gif.

Now we estimate the solution u N https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq187_HTML.gif of system (19) in terms of F. For this purpose, we multiply both sides of the j th equation of (19) by 2 α N j https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq188_HTML.gif and sum the obtained equations with respect to j from 1 to N to obtain
2 ( A u N , u N ) L 2 ( Ω ) = 2 ( F , u N ) L 2 ( Ω ) = 2 ( L ˆ F , u N ) L 2 ( Ω ) . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equ21_HTML.gif
(21)
Since u N C π 0 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq189_HTML.gif, applying integration by parts, the right-hand side of (21) can be estimated as
2 | ( L ˆ F , u N ) L 2 ( Ω ) | γ Ω F φ 2 d Ω + γ 1 Ω ( ( l ) u N ) 2 d Ω , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equn_HTML.gif
for γ > 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq190_HTML.gif, and from (15) and (21) we have
Ω J ( u N ) d Ω γ Ω F φ 2 d Ω + γ 1 Ω ( ( l ) u N ) 2 d Ω . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equ22_HTML.gif
(22)
It can be verified that for sufficiently large γ > 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq190_HTML.gif, from (22) we obtain
u N H ˚ 1 π ( Ω ) C F φ L 2 ( Ω ) , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equ23_HTML.gif
(23)
where the constant C is independent of N. This implies that { u N } N = 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq191_HTML.gif is bounded in L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq44_HTML.gif and H ˚ 1 π ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq192_HTML.gif, and since L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq44_HTML.gif and H ˚ 1 π ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq192_HTML.gif are Hilbert spaces, it is weakly compact in L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq44_HTML.gif and H ˚ 1 π ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq192_HTML.gif. Therefore, there exists a subsequence, which we again denote by { u N } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq193_HTML.gif, such that u N u https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq194_HTML.gif weakly in L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq44_HTML.gif and H ˚ 1 π ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq195_HTML.gif as N https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq196_HTML.gif and
u H ˚ 1 π ( Ω ) C F φ L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equo_HTML.gif
holds. Since u N | D × ( 0 , 2 π ) = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq197_HTML.gif and u N u https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq194_HTML.gif weakly in H ˚ 1 π ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq192_HTML.gif, we have u | D × ( 0 , 2 π ) = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq198_HTML.gif. From (23) we have also that { u N x 1 } N = 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq199_HTML.gif, { u N x 2 } N = 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq200_HTML.gif and { u N φ } N = 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq201_HTML.gif are bounded and there exists a subsequence of { u N } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq193_HTML.gif, which is again denoted by { u N } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq193_HTML.gif, such that u N x 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq202_HTML.gif, u N x 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq203_HTML.gif and u N φ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq204_HTML.gif converge weakly in L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq44_HTML.gif to u x 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq205_HTML.gif, u x 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq206_HTML.gif and u φ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq207_HTML.gif, respectively. Taking into account that u N , w j C π 0 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq208_HTML.gif, F H 2 π ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq106_HTML.gif and applying integration by parts in (19), for N j https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq209_HTML.gif, we obtain
( L u N F , ( L ˆ ) w j ) L 2 ( Ω ) = 0 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equp_HTML.gif
Since the linear span of { w j } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq210_HTML.gif is dense on the space H ˚ 1 , 2 π ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq211_HTML.gif, passing to the limit as N https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq212_HTML.gif, we get
( L u F , ( L ˆ ) η ) L 2 ( Ω ) = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equ24_HTML.gif
(24)
for every η H ˚ 1 , 2 π ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq213_HTML.gif. If we set λ = L u F https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq214_HTML.gif, since C 0 ( Ω ) H ˚ 1 , 2 π ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq215_HTML.gif, from (24) we have L ˆ λ = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq27_HTML.gif in the generalized functions sense. Moreover,
λ L 2 ( Ω ) C u H ˚ 1 π ( Ω ) + F L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equq_HTML.gif

holds and, by using u H ˚ 1 π ( Ω ) C F φ L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq216_HTML.gif, it can be seen that (7) holds. In the above expressions, by C we denote generic constants which depend only on the given functions and Lebesgue measure of the domain D.

Now it remains to show that u Γ ( A ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq87_HTML.gif. Since u L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq80_HTML.gif and F H 2 π ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq106_HTML.gif, for F = L ˆ F L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq217_HTML.gif, from (24) we have
( u , A η ) L 2 ( Ω ) = ( u , L ( L ˆ ) η ) L 2 ( Ω ) = ( L u , ( L ˆ ) η ) L 2 ( Ω ) = ( F , ( L ˆ ) η ) L 2 ( Ω ) = ( F , η ) L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equr_HTML.gif

for any η C 0 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq218_HTML.gif, which implies that F = A u https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq219_HTML.gif in the generalized functions sense, i.e., u Γ ( A ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq220_HTML.gif.

Moreover, from (19) we have P N A u N = P N F https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq221_HTML.gif, where P N https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq222_HTML.gif is the orthogonal projector. Since the system { w 1 , w 2 , } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq223_HTML.gif is orthogonal and complete in L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq44_HTML.gif, P N F F https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq224_HTML.gif strongly, i.e., P N A u N F = A u https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq225_HTML.gif strongly in L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq44_HTML.gif as N https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq196_HTML.gif. Then, since u N u https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq194_HTML.gif weakly in L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq44_HTML.gif as N https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq196_HTML.gif, we have ( P N A u N , u N ) L 2 ( Ω ) ( A u , u ) L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq226_HTML.gif as N https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq196_HTML.gif. On the other hand, since the projection operator P N https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq222_HTML.gif is self-adjoint in L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq227_HTML.gif,
( P N A u N , u N ) L 2 ( Ω ) = ( A u N , P N u N ) L 2 ( Ω ) = ( A u N , P N u N ) L 2 ( Ω ) = ( A u N , u N ) L 2 ( Ω ) , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equs_HTML.gif

thus ( A u N , u N ) L 2 ( Ω ) ( A u , u ) L 2 ( Ω ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq228_HTML.gif as N https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq196_HTML.gif, and so u Γ ( A ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq229_HTML.gif. The proof of Theorem 1 is complete. □

Proof of Lemma 1 Since K = 1 r https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq133_HTML.gif is constant, and hence
K x 1 = K x 2 = K φ = 0 , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equt_HTML.gif
to prove that condition (6) holds, we only need to show that g satisfies
g x 1 cos φ + g x 2 sin φ + g φ 1 r > 1 r 2 + g 2 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equu_HTML.gif
By taking into account that M = sup ( x 1 , x 2 ) D ( x 1 2 + x 2 2 ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq230_HTML.gif and r > 3 M https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq135_HTML.gif, for the function
g ( x 1 , x 2 , φ ) = 1 3 M ( x 1 cos φ + x 2 sin φ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equv_HTML.gif
given in (8) defined on D × ( 0 , 2 π ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq134_HTML.gif, we obtain
g x 1 cos φ + g x 2 sin φ + g φ 1 r 1 r 2 g 2 = 1 3 M ( 1 + 1 r ( x 1 sin φ + x 2 cos φ ) ) 1 r 2 1 9 M 2 ( x 1 cos φ + x 2 sin φ ) 2 1 3 M ( 1 1 r M ) 1 r 2 1 9 M > 1 3 M ( 1 1 3 ) 2 9 M = 0 , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equw_HTML.gif

and the proof is complete. □

Proof of Lemma 2 For
K ( x 1 , x 2 , φ ) = ( x 1 sin φ x 2 cos φ + ( R 2 ( x 1 cos φ + x 2 sin φ ) 2 ) 1 / 2 ) 1 = ( R 2 ( x 1 cos φ + x 2 sin φ ) 2 ) 1 / 2 x 1 sin φ + x 2 cos φ R 2 x 1 2 x 2 2 , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equx_HTML.gif
we have
K x 1 = 1 ( R 2 x 1 2 x 2 2 ) 2 ( ( ( x 1 cos φ + x 2 sin φ ) cos φ ( R 2 ( x 1 cos φ + x 2 sin φ ) 2 ) 1 / 2 sin φ ) ( R 2 x 1 2 x 2 2 ) + 2 x 1 ( ( R 2 ( x 1 cos φ + x 2 sin φ ) 2 ) 1 / 2 x 1 sin φ + x 2 cos φ ) ) , K x 2 = 1 ( R 2 x 1 2 x 2 2 ) 2 ( ( ( x 1 cos φ + x 2 sin φ ) sin φ ( R 2 ( x 1 cos φ + x 2 sin φ ) 2 ) 1 / 2 + cos φ ) ( R 2 x 1 2 x 2 2 ) + 2 x 2 ( ( R 2 ( x 1 cos φ + x 2 sin φ ) 2 ) 1 / 2 x 1 sin φ + x 2 cos φ ) ) , K φ = ( x 1 cos φ + x 2 sin φ ) ( R 2 x 1 2 x 2 2 ) ( ( x 1 sin φ + x 2 cos φ ) ( R 2 ( x 1 cos φ + x 2 sin φ ) 2 ) 1 / 2 + 1 ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equy_HTML.gif
and
K x 1 sin φ + K x 2 cos φ = 1 ( R 2 x 1 2 x 2 2 ) 2 ( R 2 ( x 1 cos φ + x 2 sin φ ) 2 + ( x 1 sin φ x 2 cos φ ) 2 2 ( x 1 sin φ x 2 cos φ ) ( R 2 ( x 1 cos φ + x 2 sin φ ) 2 ) 1 / 2 ) = K 2 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equz_HTML.gif
Since M = sup ( x 1 , x 2 ) D ( x 1 2 + x 2 2 ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq128_HTML.gif and R > M https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq140_HTML.gif, for
g ( x 1 , x 2 , φ ) = 1 2 M ( x 1 cos φ + x 2 sin φ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equaa_HTML.gif
given in (12) defined on D × ( 0 , 2 π ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq134_HTML.gif, we have
K φ g = ( x 1 cos φ + x 2 sin φ ) ( R 2 x 1 2 x 2 2 ) ( ( x 1 sin φ + x 2 cos φ ) ( R 2 ( x 1 cos φ + x 2 sin φ ) 2 ) 1 / 2 + 1 ) × 1 2 M ( x 1 cos φ + x 2 sin φ ) = ( x 1 cos φ + x 2 sin φ ) 2 ( x 1 sin φ + x 2 cos φ + ( R 2 ( x 1 cos φ + x 2 sin φ ) 2 ) 1 / 2 ) 2 M ( R 2 x 1 2 x 2 2 ) ( R 2 ( x 1 cos φ + x 2 sin φ ) 2 ) 1 / 2 0 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equab_HTML.gif
Hence, to prove that condition (6) holds, we only need to show that
g x 1 cos φ + g x 2 sin φ + g φ K > g 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equac_HTML.gif
holds. If we take into account again that M = sup ( x 1 , x 2 ) D ( x 1 2 + x 2 2 ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq231_HTML.gif and R > M https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_IEq140_HTML.gif, then we obtain
g x 1 cos φ + g x 2 sin φ + g φ K g 2 = 1 2 M ( 1 + x 1 sin φ + x 2 cos φ x 1 sin φ x 2 cos φ + ( R 2 ( x 1 cos φ + x 2 sin φ ) 2 ) 1 / 2 ) 1 4 M 2 ( x 1 cos φ + x 2 sin φ ) 2 1 2 M ( R 2 ( x 1 cos φ + x 2 sin φ ) 2 ) 1 / 2 x 1 sin φ x 2 cos φ + ( R 2 ( x 1 cos φ + x 2 sin φ ) 2 ) 1 / 2 1 4 M > 0 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-202/MediaObjects/13661_2013_Article_455_Equad_HTML.gif

The proof is complete. □

Declarations

Authors’ Affiliations

(1)
Department of Mathematics, Bulent Ecevit University

References

  1. Isakov V: Inverse Problems for Partial Differential Equations. Springer, New York; 2006.MATH
  2. Radon J: Über die Bestimmung von Funktionen durch ihre Integralwerte längs gewisser Mannigfaltigkeiten. Ber. Verh. Sächs. Akad. Wiss. Leipz., Math.-Nat.wiss. Kl. 1917, 69: 262-277.MATH
  3. Natterer F, Wübbeling F Monographs on Mathematical Modeling and Computation 5. In Mathematical Methods in Image Reconstruction. SIAM, Philadelphia; 2001.View Article
  4. Natterer F: The Mathematics of Computerized Tomography. Vieweg+Teubner, Wiesbaden; 1986.MATH
  5. Cormack AM: Representation of a function by its line integrals, with some radiological applications. J. Appl. Phys. 1963, 34: 2722-2727. 10.1063/1.1729798View ArticleMATH
  6. Cormack AM: The Radon transform on a family of curves in the plane. Proc. Am. Math. Soc. 1981, 83: 325-330. 10.1090/S0002-9939-1981-0624923-1MathSciNetView ArticleMATH
  7. Natterer F: Inversion of attenuated Radon transform. Inverse Probl. 2001, 17: 113-119. 10.1088/0266-5611/17/1/309MathSciNetView ArticleMATH
  8. Novikov RG: An inversion formula for the attenuated X-ray transformation. Ark. Mat. 2002, 40(1):145-167. 10.1007/BF02384507MathSciNetView ArticleMATH
  9. Quinto ET: Radon transforms on curves in the plane. Lectures in Applied Mathematics 30. In Tomography, Impedance Imaging and Integral Geometry. Am. Math. Soc., Providence; 1994:231-244.
  10. Ambartsoumian G, Gouia-Zarrad R, Lewis M: Inversion of the circular Radon transform on an annulus. Inverse Probl. 2010., 26: Article ID 105015
  11. Nguyen MK, Truong TT: Inversion of a new circular-arc Radon transform for Compton scattering tomography. Inverse Probl. 2010., 26: Article ID 065005
  12. Rigaud G, Nguyen MK, Louis AK: Novel numerical inversions of two circular-arc Radon transforms in Compton scattering tomography. Inverse Probl. Sci. Eng. 2012, 20(6):809-839. 10.1080/17415977.2011.653008MathSciNetView Article
  13. Amirov AK: Integral Geometry and Inverse Problems for Kinetic Equations. VSP, Utrecht; 2001.View ArticleMATH
  14. Gelfand IM, Gindikin SG, Graev MI: Selected Topics in Integral Geometry. Am. Math. Soc., Providence; 2003.MATH
  15. Romanov VG: Integral Geometry and Inverse Problems for Hyperbolic Equations. Springer, Berlin; 1974.View ArticleMATH
  16. Sharafutdinov VA: Integral Geometry of Tensor Fields. VSP, Utrecht; 1994.View ArticleMATH
  17. Arridge SR: Optical tomography in medical imaging. Inverse Probl. 1999, 15: R41-R93. 10.1088/0266-5611/15/2/022MathSciNetView ArticleMATH
  18. Bal G: Inverse transport theory and applications. Inverse Probl. 2009., 25: Article ID 053001
  19. Stefanov P: Inverse problems in transport theory. MSRI Publications 47. In Inside Out: Inverse Problems and Applications. Edited by: Uhlmann G. Cambridge University Press, Cambridge; 2003.
  20. Tamasan A: An inverse boundary value problem in two-dimensional transport. Inverse Probl. 2002, 18: 209-219. 10.1088/0266-5611/18/1/314MathSciNetView ArticleMATH
  21. Anikonov DS, Kovtanyuk AE, Prokhorov IV: Transport Equation and Tomography. VSP, Utrecht; 2002.MATH
  22. Case KM, Zweifel PF: Linear Transport Theory. Addison-Wesley, Reading; 1967.MATH
  23. Klibanov MV, Yamamoto M: Exact controllability for the time dependent transport equation. SIAM J. Control Optim. 2007, 46(6):2071-2095. 10.1137/060652804MathSciNetView ArticleMATH
  24. Amirov AK: Existence and uniqueness theorems for the solution of an inverse problem for the transport equation. Sib. Math. J. 1986, 27: 785-800.MathSciNetView ArticleMATH
  25. Amirov A, Yildiz M, Ustaoglu Z: Solvability of a problem of integral geometry via an inverse problem for a transport-like equation and a numerical method. Inverse Probl. 2009., 25: Article ID 095002
  26. Ustaoglu Z, Heydarov B, Amirov S: On the solvability and approximate solution of a two dimensional coefficient inverse problem for a transport-like equation. Inverse Probl. 2010., 26: Article ID 115019
  27. Amirov A, Ustaoglu Z, Heydarov B: Solvability of a two dimensional coefficient inverse problem for transport equation and a numerical method. Transp. Theory Stat. Phys. 2011, 40(1):1-22. 10.1080/00411450.2010.529980MathSciNetView ArticleMATH
  28. Golgeleyen I: An integral geometry problem along geodesics and a computational approach. An. Univ. “Ovidius” Constanţa, Ser. Mat. 2010, 18(2):91-112.MathSciNetMATH
  29. Golgeleyen I: An inverse problem for a generalized transport equation in polar coordinates and numerical applications. Inverse Probl. 2013., 29: Article ID 095006
  30. Lavrent’ev MM, Anikonov YE: A certain class of problems in integral geometry. Sov. Math. Dokl. 1967, 8: 1240-1241.MATHMathSciNet
  31. Klibanov MV, Timonov A: Carleman Estimates for Coefficient Inverse Problems and Numerical Applications. VSP, Utrecht; 2004.View ArticleMATH
  32. Courant R, Hilbert D 2. In Methods of Mathematical Physics. Partial Differential Equations. Interscience, New York; 1962.
  33. Lavrent’ev MM, Romanov VG, Shishatskii SP: Ill-Posed Problems of Mathematical Physics and Analysis. Am. Math. Soc., Providence; 1986.
  34. Mikhailov VP: Partial Differential Equations. Mir, Moscow; 1978.
  35. Evans LC: Partial Differential Equations. Am. Math. Soc., Providence; 1998.MATH
  36. Lions JL, Magenes E: Nonhomogeneous Boundary Value Problems and Applications. Springer, London; 1972.View ArticleMATH

Copyright

© Ustaoglu; licensee Springer 2013

This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://​creativecommons.​org/​licenses/​by/​2.​0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.