Open Access

Spectral asymptotics of self-adjoint fourth order boundary value problems with eigenvalue parameter dependent boundary conditions

Boundary Value Problems20122012:106

DOI: 10.1186/1687-2770-2012-106

Received: 8 May 2012

Accepted: 17 September 2012

Published: 4 October 2012

Abstract

A regular fourth order differential equation with λ-dependent boundary conditions is considered. For four distinct cases with exactly one λ-independent boundary condition, the asymptotic eigenvalue distribution is presented.

MSC:34L20, 34B07, 34B08, 34B09.

Keywords

fourth order boundary value problems self-adjoint boundary conditions eigenvalue distribution pure imaginary eigenvalues spectral asymptotics

1 Introduction

Sturm-Liouville problems have attracted extensive attention due to their intrinsic mathematical challenges and their applications in physics and engineering. However, apart from classical Sturm-Liouville problems, also higher order ordinary linear differential equations occur in applications, with or without the eigenvalue parameter in the boundary conditions. Such problems are realized as operator polynomials, also called operator pencils. Some recent developments of higher order differential operators whose boundary conditions depend on the eigenvalue parameter, including spectral asymptotics and basis properties, have been investigated in [14]. General characterizations of self-adjoint boundary conditions have been presented in [5, 6] for singular and (quasi-)regular problems. In all these cases, the minimal operator associated with an n th order differential equation must be symmetric, see [7, 8] for necessary and sufficient conditions. A more general discussion on the spectra of fourth order differential operators can be found in [9, 10].

The generalized Regge problem is realized by a second order differential operator which depends quadratically on the eigenvalue parameter and which has eigenvalue parameter dependent boundary conditions, see [11]. The particular feature of this problem is that the coefficient operators of this pencil are self-adjoint, and it is shown in [11] that this gives some a priori knowledge about the location of the spectrum. In [12] this approach has been extended to a fourth order differential equation describing small transversal vibrations of a homogeneous beam compressed or stretched by a force g. Separation of variables leads to a fourth order boundary problem with eigenvalue parameter dependent boundary conditions, where the differential equation
y ( 4 ) ( g y ) = λ 2 y https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equa_HTML.gif

depends quadratically on the eigenvalue parameter. This problem is represented by a quadratic operator pencil, in a suitably chosen Hilbert space, whose coefficient operators are self-adjoint. In [13] we have investigated a class of boundary conditions for which necessary and sufficient conditions were obtained such that the associated operator pencil consists of self-adjoint operators, while in [14] we have continued the work of [13] in the direction of [12] to derive eigenvalue asymptotics associated with boundary conditions which lead to self-adjoint operator representations. We have considered the particular case of boundary conditions which do not depend on the eigenvalue parameter at the left endpoint and depend on the eigenvalue parameter at the right endpoint.

In this paper, we extend the work of [14] to a class of boundary conditions where exactly one of the left endpoint boundary conditions does not depend on the eigenvalue parameter, while the remaining boundary conditions depend on the eigenvalue parameter.

We define the operator pencil in Section 2 and we discuss which boundary conditions are considered. In Section 3, the eigenvalue asymptotics for the case g = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq1_HTML.gif are derived. In Section 4, it is shown that the boundary value problems under consideration are Birkhoff regular, which implies that the eigenvalues for general g are small perturbations of the eigenvalues for g = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq1_HTML.gif. Hence, in Section 5, the first four terms of the eigenvalue asymptotics are found and are compared to those obtained in [14].

2 The quadratic operator pencil L

On the interval [ 0 , a ] https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq2_HTML.gif, we consider the boundary value problem
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ1_HTML.gif
(2.1)
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ2_HTML.gif
(2.2)
where a > 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq3_HTML.gif, g C 1 [ 0 , a ] https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq4_HTML.gif is a real valued function and (2.2) are separated boundary conditions where the B j ( λ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq5_HTML.gif are constant or depend on λ linearly. The boundary conditions (2.2) are taken at the endpoint 0 for j = 1 , 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq6_HTML.gif and at the endpoint a for j = 3 , 4 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq7_HTML.gif. Further, we assume for simplicity that either B j ( λ ) y = y [ p j ] ( a j ) + i ε j α λ y [ q j ] ( a j ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq8_HTML.gif or B j ( λ ) y = y [ p j ] ( a j ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq9_HTML.gif, where a j = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq10_HTML.gif for j = 1 , 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq6_HTML.gif, a j = a https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq11_HTML.gif for j = 3 , 4 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq7_HTML.gif, α > 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq12_HTML.gif and ε j { 1 , 1 } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq13_HTML.gif. We recall that the quasi-derivatives associated with (2.1) are given by
y [ 0 ] = y , y [ 1 ] = y , y [ 2 ] = y , y [ 3 ] = y ( 3 ) g y , y [ 4 ] = y ( 4 ) ( g y ) , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equb_HTML.gif

see [[8], p.26].

Define
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equc_HTML.gif

Assumption 2.1 The numbers p 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq14_HTML.gif, p 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq15_HTML.gif, q j https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq16_HTML.gif for j Θ 1 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq17_HTML.gif are distinct as well as the numbers p 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq18_HTML.gif, p 4 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq19_HTML.gif, q j https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq16_HTML.gif for j Θ 1 a https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq20_HTML.gif.

We denote by U the collection of the boundary conditions (2.2) and define the following operators related to U:
U 0 y = ( y [ p j ] ( a j ) ) j Θ 1 and U 1 y = ( ε j y [ q j ] ( a j ) ) j Θ 1 , y W 4 2 ( 0 , a ) . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ3_HTML.gif
(2.3)
We put k = | Θ 1 | https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq21_HTML.gif and consider the linear operators A ( U ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq22_HTML.gif, K and M in the space L 2 ( 0 , a ) C k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq23_HTML.gif with domains
D ( A ( U ) ) = { y ˜ = ( y U 1 y ) : y W 4 2 ( 0 , a ) , y [ p j ] ( a j ) = 0  for  j Θ 0 } , D ( K ) = D ( M ) = L 2 ( 0 , a ) C k , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equd_HTML.gif
given by
( A ( U ) ) y ˜ = ( y [ 4 ] U 0 y ) for  y ˜ D ( A ( U ) ) , K = ( 0 0 0 I ) and M = ( I 0 0 0 ) . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Eque_HTML.gif
It is easy to check that K 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq24_HTML.gif, M 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq25_HTML.gif, M + K = I https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq26_HTML.gif and M | D ( A ( U ) ) > 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq27_HTML.gif. We associate a quadratic operator pencil
L ( λ , α ) = λ 2 M i α λ K A ( U ) , λ C https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ4_HTML.gif
(2.4)

in the space L 2 ( 0 , a ) C k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq23_HTML.gif with the problem (2.1), (2.2).

The conditions under which the differential operator A ( U ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq22_HTML.gif is self-adjoint are given in

Theorem 2.2 ([13], Theorem 1.2)

Denote by P 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq28_HTML.gif the set of p in y [ p ] ( 0 ) = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq29_HTML.gif for the λ-independent boundary conditions and by P a https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq30_HTML.gif the corresponding set for y [ p ] ( a ) = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq31_HTML.gif. Then the differential operator A ( U ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq22_HTML.gif associated with this boundary value problem is self-adjoint if and only if p + q = 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq32_HTML.gif for all boundary conditions of the form y [ p ] ( a j ) + i α ε j λ y [ q ] ( a j ) = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq33_HTML.gif and ε j = 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq34_HTML.gif if q is even in case a j = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq35_HTML.gif or odd in case a j = a https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq36_HTML.gif, ε j = 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq37_HTML.gif otherwise, { 0 , 3 } P 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq38_HTML.gif, { 1 , 2 } P 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq39_HTML.gif, { 0 , 3 } P a https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq40_HTML.gif and { 1 , 2 } P a https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq41_HTML.gif.

Proposition 2.3 The operator pencil L ( , α ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq42_HTML.gif is a Fredholm valued operator function with index 0. The spectrum of the Fredholm operator L ( , α ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq42_HTML.gif consists of discrete eigenvalues of finite multiplicities, and all eigenvalues of L ( , α ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq42_HTML.gif, α 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq43_HTML.gif, lie in the closed upper half-plane and on the imaginary axis and are symmetric with respect to the imaginary axis.

Proof As in [[12], Section 3], we can argue that for all λ C https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq44_HTML.gif, L ( λ , α ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq45_HTML.gif is a relatively compact perturbation of L ( 0 , 0 ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq46_HTML.gif, where L ( 0 , 0 ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq46_HTML.gif is well known to be a Fredholm operator. The statement on the location of the spectrum now follows as in [[12], Lemma 3.1]. □

We now consider the particular cases that exactly one of the boundary conditions at 0 depends on λ, whereas both boundary conditions at a depend on λ. Therefore, taking Assumption 2.1 and Theorem 2.2 into account, we have the four boundary conditions
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equf_HTML.gif

where 0 p 1 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq47_HTML.gif, 0 p 2 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq48_HTML.gif, 0 q 2 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq49_HTML.gif, p 2 + q 2 = 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq50_HTML.gif, and p 1 { q 2 , p 2 } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq51_HTML.gif, while { p 3 , q 3 } = { 1 , 2 } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq52_HTML.gif and { p 4 , q 4 } = { 0 , 3 } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq53_HTML.gif. Thus, we have 8 and 4 possible sets of boundary conditions at the endpoint 0 and a, respectively. Whence there are 32 different sets of boundary conditions. Recall that the parameter λ emanates from derivatives with respect to the time variable in the original partial differential equation, and it is reasonable that the highest space derivative occurs in the term without time derivative. Thus, the most relevant boundary conditions would have q 2 < p 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq54_HTML.gif, q 3 < p 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq55_HTML.gif and q 4 < p 4 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq56_HTML.gif. This leaves us with four different cases for the boundary conditions B j ( λ ) y = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq57_HTML.gif.

These four cases are uniquely determined by the value of p 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq14_HTML.gif, so that we will consider

Case 1: p 1 = 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq58_HTML.gif; Case 2: p 1 = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq59_HTML.gif; Case 3: p 1 = 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq60_HTML.gif; Case 4: p 1 = 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq61_HTML.gif.

The corresponding boundary operators are then
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ5_HTML.gif
(2.5)
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ6_HTML.gif
(2.6)
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ7_HTML.gif
(2.7)
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ8_HTML.gif
(2.8)

3 Asymptotics of eigenvalues for g = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq1_HTML.gif

In this section, we consider the boundary value problem (2.1), (2.2) with g = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq1_HTML.gif. We count all eigenvalues with their proper multiplicities and develop a formula for the asymptotic distribution of the eigenvalues, which is used to obtain the corresponding formula for general g. Observe that for g = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq1_HTML.gif, the quasi-derivatives y [ j ] https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq62_HTML.gif coincide with the standard derivatives y ( j ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq63_HTML.gif. We take the canonical fundamental system y j ( , λ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq64_HTML.gif, j = 1 , , 4 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq65_HTML.gif, of (2.1) with y j ( m ) ( 0 ) = δ j , m + 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq66_HTML.gif if j 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq67_HTML.gif for m = 0 , , 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq68_HTML.gif. It is well known that the functions y j ( , λ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq69_HTML.gif are analytic on with respect to λ. Putting
M ( λ ) = ( B i ( λ ) y j ( , λ ) ) i , j = 1 4 , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equg_HTML.gif

the eigenvalues of the boundary value problem (2.1), (2.2) are the eigenvalues of the analytic matrix function M, where the corresponding geometric and algebraic multiplicities coincide, see [[15], Theorem 6.3.2].

Setting λ = μ 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq70_HTML.gif and
y ( x , μ ) = 1 2 μ 3 sinh ( μ x ) 1 2 μ 3 sin ( μ x ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equh_HTML.gif
it is easy to see that
y j ( x , λ ) = y ( 4 j ) ( x , μ ) , j = 1 , , 4 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equi_HTML.gif

The second row of M ( λ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq71_HTML.gif has exactly two non-zero entries (for λ 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq72_HTML.gif), and these non-zero entries are:

In Cases 1 and 2, B 2 ( λ ) y 2 ( , λ ) = i α μ 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq73_HTML.gif and B 2 ( λ ) y 3 ( , λ ) = 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq74_HTML.gif;

In Cases 3 and 4, B 2 ( λ ) y 1 ( , λ ) = i α μ 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq75_HTML.gif and B 2 ( λ ) y 4 ( , λ ) = 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq76_HTML.gif.

Since the first row of M ( λ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq71_HTML.gif has exactly one entry 1 and all other entries zero, an expansion of M ( λ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq71_HTML.gif with respect to the second row shows that det M ( λ ) = ± ϕ ( μ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq77_HTML.gif, where
ϕ ( μ ) = i α μ 2 det ( B 3 ( μ 2 ) y σ ( 1 ) ( , μ ) B 3 ( μ 2 ) y σ ( 2 ) ( , μ ) B 4 ( μ 2 ) y σ ( 1 ) ( , μ ) B 4 ( μ 2 ) y σ ( 2 ) ( , μ ) ) + det ( B 3 ( μ 2 ) y σ ( 3 ) ( , μ ) B 3 ( μ 2 ) y σ ( 4 ) ( , μ ) B 4 ( μ 2 ) y σ ( 3 ) ( , μ ) B 4 ( μ 2 ) y σ ( 4 ) ( , μ ) ) , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equj_HTML.gif
with
( σ ( 1 ) , σ ( 2 ) , σ ( 3 ) , σ ( 4 ) ) = { ( 1 , 3 , 1 , 2 ) in Case 1 , ( 3 , 4 , 2 , 4 ) in Case 2 , ( 3 , 4 , 1 , 3 ) in Case 3 , ( 2 , 4 , 1 , 2 ) in Case 4 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equk_HTML.gif
In view of (2.7), (2.8) this gives
ϕ ( μ ) = i α μ 2 [ ( y σ ( 1 ) ( a , μ ) + i α μ 2 y σ ( 1 ) ( a , μ ) ) ( y σ ( 2 ) ( 3 ) ( a , μ ) i α μ 2 y σ ( 2 ) ( a , μ ) ) ( y σ ( 2 ) ( a , μ ) + i α μ 2 y σ ( 2 ) ( a , μ ) ) ( y σ ( 1 ) ( 3 ) ( a , μ ) i α μ 2 y σ ( 1 ) ( a , μ ) ) ] + ( y σ ( 3 ) ( a , μ ) + i α μ 2 y σ ( 3 ) ( a , μ ) ) ( y σ ( 4 ) ( 3 ) ( a , μ ) i α μ 2 y σ ( 4 ) ( a , μ ) ) ( y σ ( 4 ) ( a , μ ) + i α μ 2 y σ ( 4 ) ( a , μ ) ) ( y σ ( 3 ) ( 3 ) ( a , μ ) i α μ 2 y σ ( 3 ) ( a , μ ) ) = i α μ 2 [ i α μ 2 { y σ ( 1 ) ( a , μ ) y σ ( 2 ) ( 3 ) ( a , μ ) y σ ( 2 ) ( a , μ ) y σ ( 1 ) ( 3 ) ( a , μ ) + y σ ( 2 ) ( a , μ ) y σ ( 1 ) ( a , μ ) y σ ( 1 ) ( a , μ ) y σ ( 2 ) ( a , μ ) } + α 2 μ 4 { y σ ( 1 ) ( a , μ ) y σ ( 2 ) ( a , μ ) y σ ( 2 ) ( a , μ ) y σ ( 1 ) ( a , μ ) } + y σ ( 1 ) ( a , μ ) y σ ( 2 ) ( 3 ) ( a , μ ) y σ ( 2 ) ( a , μ ) y σ ( 1 ) ( 3 ) ( a , μ ) ] + i α μ 2 [ y σ ( 3 ) ( a , μ ) y σ ( 4 ) ( 3 ) ( a , μ ) y σ ( 4 ) ( a , μ ) y σ ( 3 ) ( 3 ) ( a , μ ) + y σ ( 4 ) ( a , μ ) y σ ( 3 ) ( a , μ ) y σ ( 3 ) ( a , μ ) y σ ( 4 ) ( a , μ ) ] + α 2 μ 4 ( y σ ( 3 ) ( a , μ ) y σ ( 4 ) ( a , μ ) y σ ( 4 ) ( a , μ ) y σ ( 3 ) ( a , μ ) ) + y σ ( 3 ) ( a , μ ) y σ ( 4 ) ( 3 ) ( a , μ ) y σ ( 4 ) ( a , μ ) y σ ( 3 ) ( 3 ) ( a , μ ) . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equl_HTML.gif
Each of the summands in ϕ is a product of a power in μ and a product of two sums of a trigonometric and a hyperbolic functions. The terms with the highest μ-powers in ϕ ( μ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq78_HTML.gif are non-zero constant multiples of
ϕ 0 ( μ ) = { 2 μ 4 ( y σ ( 1 ) ( a , μ ) y σ ( 2 ) ( 3 ) ( a , μ ) y σ ( 2 ) ( a , μ ) y σ ( 1 ) ( 3 ) ( a , μ ) ) in Cases 1, 2 , 2 μ 2 ( y σ ( 3 ) ( a , μ ) y σ ( 4 ) ( 3 ) ( a , μ ) y σ ( 4 ) ( a , μ ) y σ ( 3 ) ( 3 ) ( a , μ ) ) in Cases 3, 4 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equm_HTML.gif
For the above four cases, we obtain:
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equn_HTML.gif

We next give the asymptotic distributions of the zeros of ϕ 0 ( μ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq79_HTML.gif with proper counting.

Lemma 3.1 Case 1: ϕ 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq80_HTML.gif has a zero of multiplicity 8 at 0, simple zeros at
μ ˜ k = ( k 2 ) π a , k = 3 , 4 , , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equo_HTML.gif

simple zeros at μ ˜ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq81_HTML.gif, μ ˜ k = i μ ˜ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq82_HTML.gif and i μ ˜ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq83_HTML.gif for k = 3 , 4 , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq84_HTML.gif, and no other zeros.

Case 2: ϕ 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq80_HTML.gif has a zero of multiplicity 4 at 0, exactly one simple zero μ ˜ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq85_HTML.gif in each interval ( ( k 1 2 ) π a , ( k + 1 2 ) π a ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq86_HTML.gif for positive integers k with asymptotics
μ ˜ k = ( 4 k 5 ) π 4 a + o ( 1 ) , k = 2 , 3 , , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equp_HTML.gif

simple zeros at μ ˜ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq81_HTML.gif, μ ˜ k = i μ ˜ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq82_HTML.gif and i μ ˜ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq83_HTML.gif for k = 2 , 3 , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq87_HTML.gif , and no other zeros.

Case 3: ϕ 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq80_HTML.gif has a zero of multiplicity 6 at 0, simple zeros at
μ ˜ k = ( k 1 ) π a , k = 2 , 3 , , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equq_HTML.gif

simple zeros at μ ˜ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq81_HTML.gif, μ ˜ k = i μ ˜ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq82_HTML.gif and i μ ˜ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq83_HTML.gif for k = 2 , 3 , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq88_HTML.gif, and no other zeros.

Case 4: ϕ 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq80_HTML.gif has a zero of multiplicity 6 at 0, exactly one simple zero μ ˜ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq85_HTML.gif in each interval ( ( k 1 2 ) π a , ( k + 1 2 ) π a ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq89_HTML.gif for positive integers k with asymptotics
μ ˜ k = ( 4 k 5 ) π 4 a + o ( 1 ) , k = 2 , 3 , , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equr_HTML.gif

simple zeros at μ ˜ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq81_HTML.gif, μ ˜ k = i μ ˜ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq82_HTML.gif and i μ ˜ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq83_HTML.gif for k = 2 , 3 , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq88_HTML.gif, and no other zeros.

Proof The result is obvious in Cases 1 and 3. Cases 2 and 4 only differ in the factor with the power of μ, and the multiplicity of the corresponding zero of ϕ 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq80_HTML.gif at 0 is easy to verify. The choice of the indexing for the non-zero zeros of ϕ 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq80_HTML.gif in each case will become apparent later.

It, therefore, remains to describe the behavior of the non-zero zeros of ϕ 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq80_HTML.gif in Case 2. First, we are going to find the zeros of ϕ 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq80_HTML.gif on the positive real axis. One can observe that for μ 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq90_HTML.gif, ϕ 0 ( μ ) = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq91_HTML.gif implies cosh ( μ a ) 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq92_HTML.gif and cos ( μ a ) 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq93_HTML.gif, whence the positive zeros of ϕ 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq80_HTML.gif are those μ > 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq94_HTML.gif for which tan ( μ a ) + tanh ( μ a ) = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq95_HTML.gif. Since tan ( μ a ) 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq96_HTML.gif and tanh ( μ a ) > 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq97_HTML.gif for all x R https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq98_HTML.gif where the functions are defined, the function μ tan ( μ a ) + tanh ( μ a ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq99_HTML.gif is increasing with a positive derivative on each interval ( ( k 1 2 ) π a , ( k + 1 2 ) π a ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq100_HTML.gif, k Z https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq101_HTML.gif. On each of these intervals, the function moves from −∞ to ∞, thus we have exactly one simple zero μ ˜ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq85_HTML.gif of tan ( μ a ) + tanh ( μ a ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq102_HTML.gif in each interval ( ( k 1 2 ) π a , ( k + 1 2 ) π a ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq86_HTML.gif, where k is a positive integer, and no zero in ( 0 , π 2 a ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq103_HTML.gif. Since tanh ( μ a ) 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq104_HTML.gif as μ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq105_HTML.gif, we have
μ ˜ k = ( 4 k 5 ) π 4 a + o ( 1 ) , k = 1 , 2 , . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equs_HTML.gif

The location of the zeros on the other three half-axes follows by repeated application of ϕ 0 ( i μ ) = ϕ 0 ( μ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq106_HTML.gif.

The proof will be complete if we show that all zeros of ϕ 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq80_HTML.gif lie on the real or the imaginary axis. To this end, we observe that the product-to-sum formula for trigonometric functions gives
ϕ 0 ( μ ) = μ 3 [ cosh ( μ a ) sin ( μ a ) + sinh ( μ a ) cos ( μ a ) ] = 1 2 μ 3 [ sin ( ( 1 + i ) μ a ) + sin ( ( 1 i ) μ a ) i sin ( ( 1 + i ) μ a ) + i sin ( ( 1 i ) μ a ) ] = 1 2 μ 3 [ ( 1 i ) sin ( ( 1 + i ) μ a ) + ( 1 + i ) sin ( ( 1 i ) μ a ) ] . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ9_HTML.gif
(3.1)
Putting ( 1 + i ) μ a = x + i y https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq107_HTML.gif, x , y R https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq108_HTML.gif, it follows for μ 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq109_HTML.gif that
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ10_HTML.gif
(3.2)
Since cosh 2 x + cos 2 x = 1 2 cosh ( 2 x ) + 1 2 cos ( 2 x ) + 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq110_HTML.gif has a positive derivative on ( 0 , ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq111_HTML.gif, this function is strictly increasing, and ϕ 0 ( μ ) = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq91_HTML.gif therefore, implies by (3.2) that | y | = | x | https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq112_HTML.gif and thus y = ± x https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq113_HTML.gif. Then
μ = x + i y ( 1 + i ) a = 1 ± i 1 + i x a https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equt_HTML.gif

is either real or pure imaginary. □

Proposition 3.2 For g = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq1_HTML.gif, there exists a positive integer k 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq114_HTML.gif such that the eigenvalues λ ˆ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq115_HTML.gif, counted with multiplicity, of the problem (2.1), (2.5)-(2.8), where k Z { 0 } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq116_HTML.gif in Cases 1 and  2 and k Z https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq101_HTML.gif in Cases 3 and 4, can be enumerated in such a way that the eigenvalues λ ˆ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq115_HTML.gif are pure imaginary for | k | < k 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq117_HTML.gif, and λ ˆ k = λ ˆ k ¯ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq118_HTML.gif for k k 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq119_HTML.gif. For k > 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq120_HTML.gif, we can write λ ˆ k = μ ˆ k 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq121_HTML.gif, where the μ ˆ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq122_HTML.gif have the following asymptotic representation as k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq123_HTML.gif:
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equu_HTML.gif

In particular, the number of pure imaginary eigenvalues is even in Cases 1 and 2 and odd in Cases 3 and 4.

Proof Case 4: A straightforward calculation gives
ϕ ( μ ) = 1 2 i ( 2 α + α 3 ) μ 3 [ cosh ( μ a ) sin ( μ a ) sinh ( μ a ) cos ( μ a ) ] 1 2 i α μ 5 [ sinh ( μ a ) cos ( μ a ) + cosh ( μ a ) sin ( μ a ) ] 1 2 α 2 μ 4 [ 3 cosh ( μ a ) cos ( μ a ) + 1 ] 1 2 μ 4 [ cosh ( μ a ) cos ( μ a ) 1 ] α 2 μ 2 sin ( μ a ) sinh ( μ a ) . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ11_HTML.gif
(3.3)
Up to the constant factor 1 2 i α https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq124_HTML.gif, the second term equals ϕ 0 ( μ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq125_HTML.gif. It follows that for μ outside the zeros of ϕ 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq80_HTML.gif, cos ( a ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq126_HTML.gif and cosh ( a ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq127_HTML.gif, we have
ϕ 1 ( μ ) = 2 ϕ ( μ ) i α ϕ 0 ( μ ) i α ϕ 0 ( μ ) = α 2 1 i α μ 1 cosh ( μ a ) cos ( μ a ) 1 tan ( μ a ) + tanh ( μ a ) + 3 α 2 + 1 i α μ 1 tan ( μ a ) + tanh ( μ a ) + 2 α i μ 3 tan ( μ a ) tanh ( μ a ) tan ( μ a ) + tanh ( μ a ) + ( 2 + α 2 ) μ 2 [ 1 2 tanh ( μ a ) tan ( μ a ) + tanh ( μ a ) ] . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ12_HTML.gif
(3.4)
Fix ε ( 0 , π 4 a ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq128_HTML.gif and for k = 2 , 3 , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq88_HTML.gif let R k , ε https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq129_HTML.gif be the squares determined by the vertices ( 4 k 5 ) π 4 a ± ε ± i ε https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq130_HTML.gif, k Z https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq101_HTML.gif. These squares do not intersect due to ε < π 2 a https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq131_HTML.gif. Since tan z = 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq132_HTML.gif if and only if z = j π π 4 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq133_HTML.gif and j Z https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq134_HTML.gif, it follows from the periodicity of tan that the number
C 1 ( ε ) = 2 min { | tan ( μ a ) + 1 | : μ R k , ε } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equv_HTML.gif
is positive and independent of ε. Since tanh ( μ a ) 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq135_HTML.gif uniformly in the strip { μ C : Re μ 1 , | Im μ | π 4 a } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq136_HTML.gif as | μ | https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq137_HTML.gif, there is an integer k 1 ( ε ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq138_HTML.gif such that
| tan ( μ a ) + tanh ( μ a ) | C 1 ( ε ) for all  μ R k , ε  with  k > k 1 ( ε ) . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equw_HTML.gif

By periodicity, there are numbers C 2 ( ε ) > 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq139_HTML.gif and C 3 ( ε ) > 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq140_HTML.gif such that | tan ( μ a ) | < C 2 ( ε ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq141_HTML.gif and | cos ( μ a ) | > C 3 ( ε ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq142_HTML.gif for all μ R k , ε https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq143_HTML.gif and all k. Observing | cosh ( μ a ) | | sinh ( ( Re μ ) a ) | https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq144_HTML.gif, it follows that there is k 2 ( ε ) k 1 ( ε ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq145_HTML.gif such that for all μ on the squares R k , ε https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq146_HTML.gif with k > k 2 ( ε ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq147_HTML.gif the estimate | ϕ 1 ( μ ) | < 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq148_HTML.gif holds. Further, we may assume by Lemma 3.1 that μ ˜ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq85_HTML.gif is inside R k , ε https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq146_HTML.gif for k > k 2 ( ε ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq149_HTML.gif and that no other zero of ϕ 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq80_HTML.gif has this property. Hence, it follows by Rouché’s theorem that there is exactly one (simple) zero μ ˆ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq122_HTML.gif of ϕ in each R k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq150_HTML.gif for k k 2 ( ε ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq151_HTML.gif. Replacing μ with only changes the sign of the second term in (3.3) and thus the sign of ϕ 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq152_HTML.gif. Hence, the same estimates apply to corresponding squares along the other three half-axes, and we therefore have that ϕ has zeros ± μ ˆ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq153_HTML.gif, ± μ ˆ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq154_HTML.gif for k > k 2 ( ε ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq155_HTML.gif with the same asymptotic behavior as the zeros ± μ ˜ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq156_HTML.gif, ± i μ ˜ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq157_HTML.gif of ϕ 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq80_HTML.gif as discussed in Lemma 3.1.

Next, we are going to estimate ϕ 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq152_HTML.gif on the squares S k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq158_HTML.gif, k N https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq159_HTML.gif, whose vertices are ± k π a ± i k π a https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq160_HTML.gif. For k Z https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq161_HTML.gif and γ R https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq162_HTML.gif,
tan ( ( k π a + i γ ) a ) = tan ( i γ a ) = i tanh ( γ a ) i R . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ13_HTML.gif
(3.5)
Therefore, we have for μ = k π a + i γ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq163_HTML.gif, where k Z https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq164_HTML.gif and γ R https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq165_HTML.gif, that
| tan ( μ a ) | < 1 and | tan ( μ a ) ± 1 | 1 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ14_HTML.gif
(3.6)
For μ = x + i y https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq166_HTML.gif, x , y R https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq108_HTML.gif and x 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq167_HTML.gif, we have
tanh ( μ a ) = e ( a x + i a y ) e ( a x + i a y ) e ( a x + i a y ) + e ( a x + i a y ) ± 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equx_HTML.gif
uniformly in y as x ± https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq168_HTML.gif. Hence, there exists k ˆ 0 N https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq169_HTML.gif such that for all k Z https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq170_HTML.gif, | k | k ˆ 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq171_HTML.gif and γ R https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq165_HTML.gif,
| tanh ( ( k π a + i γ ) a ) sgn ( k ) | < 1 2 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ15_HTML.gif
(3.7)
It follows from (3.6) and (3.7) for μ = k π a + i γ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq172_HTML.gif, k Z https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq101_HTML.gif, | k | k ˆ 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq171_HTML.gif and γ R https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq165_HTML.gif that
| tan ( μ a ) + tanh ( μ a ) | 1 2 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ16_HTML.gif
(3.8)
Furthermore, we are going to make use of the estimates
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ17_HTML.gif
(3.9)
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ18_HTML.gif
(3.10)

which hold for all k Z https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq101_HTML.gif with | k | k ˆ 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq171_HTML.gif and all γ R https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq165_HTML.gif. Therefore, it follows from (3.6), (3.8)-(3.10) and the corresponding estimates with μ replaced by that there is k ˆ 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq173_HTML.gif such that | ϕ 1 ( μ ) | < 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq148_HTML.gif for all μ S k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq174_HTML.gif with k > k ˆ 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq175_HTML.gif. Again from the definition of ϕ 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq152_HTML.gif in (3.4) and Rouché’s theorem, we conclude that the functions ϕ 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq80_HTML.gif and ϕ have the same number of zeros in the square S k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq158_HTML.gif, for k N https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq176_HTML.gif with k k ˆ 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq177_HTML.gif.

Since ϕ 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq80_HTML.gif has 4 k + 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq178_HTML.gif zeros inside S k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq158_HTML.gif and thus 4 k + 2 + 4 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq179_HTML.gif zeros inside S k + 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq180_HTML.gif, it follows that ϕ has no large zeros other than the zeros ± μ ˆ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq181_HTML.gif found above for | k | https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq182_HTML.gif sufficiently large, and that λ ˆ k = μ ˆ k 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq121_HTML.gif account for all eigenvalues of the problem (2.1)-(2.2) since each of these eigenvalues gives rise to two zeros of ϕ, counted with multiplicity. By Proposition 2.3, all eigenvalues with non-zero real part occur in pairs λ ˆ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq115_HTML.gif, λ ˆ k ¯ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq183_HTML.gif, which shows that we can index all such eigenvalues as λ ˆ k = λ ˆ k ¯ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq118_HTML.gif. Since there is an odd number of remaining indices, the number of pure imaginary eigenvalues must be odd.

Case 2: The function ϕ in this case is
ϕ ( μ ) = 1 2 ( 2 α 2 + 1 ) μ [ cosh ( μ a ) sin ( μ a ) sinh ( μ a ) cos ( μ a ) ] 1 2 α 2 μ 3 [ sinh ( μ a ) cos ( μ a ) + cosh ( μ a ) sin ( μ a ) ] + 1 2 i α μ 2 [ 3 cosh ( μ a ) cos ( μ a ) + 1 ] + 1 2 i a 3 μ 2 [ cosh ( μ a ) cos ( μ a ) 1 ] + i α μ 2 sin ( μ a ) sinh ( μ a ) . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equy_HTML.gif

Then all the estimates are as in Case 4, and the result in Case 2 immediately follows from that in Case 4 if we observe that each S k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq158_HTML.gif for k large enough contains two fewer zeros of ϕ than in Case 4.

Case 1: A straightforward calculation gives
ϕ ( μ ) = α 2 μ 6 sin ( μ a ) sinh ( μ a ) 1 2 ( 1 + 3 α 2 ) μ 4 cos ( μ a ) cosh ( μ a ) 1 2 i ( 2 α + α 3 ) μ 5 ( sin ( μ a ) cosh ( μ a ) + cos ( μ a ) sinh ( μ a ) ) 1 2 i α μ 3 ( sin ( μ a ) cosh ( μ a ) cos ( μ a ) sinh ( μ a ) ) + 1 2 ( 1 α 2 ) μ 4 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equz_HTML.gif
Then
ϕ 1 ( μ ) = 2 ϕ ( μ ) + α 2 ϕ 0 ( μ ) ϕ 0 ( μ ) = 1 + 3 α 2 2 μ 2 cot ( μ a ) coth ( μ a ) + ( 2 α + α 3 ) i 2 μ [ coth ( μ a ) + cot ( μ a ) ] + i α 2 μ 3 [ coth ( μ a ) cot ( μ a ) ] 1 α 2 2 μ 2 1 sin ( μ a ) sinh ( μ a ) . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equaa_HTML.gif

The result follows with reasonings and estimates as in the proof of Case 4, replacing μ by μ ± π 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq184_HTML.gif and μ ± i π 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq185_HTML.gif, respectively.

Case 3: The function ϕ in this case is
ϕ ( μ ) = i α μ 4 sin ( μ a ) sinh ( μ a ) + 1 2 i ( 3 α + α 3 ) μ 2 cos ( μ a ) cosh ( μ a ) 1 2 ( 2 α 2 + 1 ) μ 3 ( sin ( μ a ) cosh ( μ a ) + cos ( μ a ) sinh ( μ a ) ) 1 2 α 2 μ ( sin ( μ a ) cosh ( μ a ) cos ( μ a ) sinh ( μ a ) ) + 1 2 i ( α α 3 ) μ 2 , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equab_HTML.gif

and a reasoning as in Case 1 completes the proof. □

4 Birkhoff regularity

We refer to [[15], Definition 7.3.1] for the definition of the Birkhoff regularity.

Proposition 4.1 The boundary value problem (2.1), (2.5)-(2.8) is Birkhoff regular for α > 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq186_HTML.gif with respect to the eigenvalue parameter μ given by λ = μ 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq187_HTML.gif.

Proof The characteristic function of (2.1) as defined in [[15], (7.1.4)] is π ( ρ ) = ρ 4 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq188_HTML.gif, and its zeros are i k 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq189_HTML.gif, k = 1 , , 4 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq190_HTML.gif. We can choose
C ( x , μ ) = diag ( 1 , μ , μ 2 , μ 3 ) ( i ( k 1 ) ( j 1 ) ) k , j = 1 4 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equac_HTML.gif
according to [[15], Theorem 7.2.4.A]. The boundary condition (2.5)-(2.8) can be written in the form
B j ( λ ) y = B ˆ j ( μ ) ( y ( a j ) , y ( a j ) , y ( a j ) , y ( 3 ) ( a j ) ) , j = 1 , 2 , 3 , 4 , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equad_HTML.gif
where
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equae_HTML.gif
and where ε ν https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq191_HTML.gif denotes the ν th unit vector in C 4 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq192_HTML.gif. Thus the boundary matrices defined in [[15], (7.3.1)] are given by
W ( 0 ) ( μ ) = ( B ˆ 1 B ˆ 2 ( μ ) 0 0 ) C ( 0 , μ ) , W ( 1 ) ( μ ) = ( 0 0 B ˆ 3 ( μ ) B ˆ 4 ( μ ) ) C ( a , μ ) . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equaf_HTML.gif
Choosing C 2 ( μ ) = diag ( μ p 1 , μ 3 , μ 3 , μ 3 ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq193_HTML.gif, it follows that C 2 ( μ ) 1 W ( j ) ( μ ) = W 0 ( j ) + O ( μ 1 ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq194_HTML.gif, where
W 0 ( 0 ) = ( 1 i r 2 i 2 ( r 2 ) i 3 ( r 2 ) θ 1 θ 2 θ 3 θ 4 0 0 0 0 0 0 0 0 ) , W 0 ( 1 ) = ( 0 0 0 0 0 0 0 0 i α α i α α 1 i 1 i ) , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equag_HTML.gif
for Case r and θ j = i j α https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq195_HTML.gif for Cases 1 and 2, while θ j = ( i ) j 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq196_HTML.gif for Cases 3 and 4. The Birkhoff matrices are
W 0 ( 0 ) Δ j + W 0 ( 1 ) ( I Δ j ) , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ19_HTML.gif
(4.1)
where Δ j https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq197_HTML.gif, j = 1 , 2 , 3 , 4 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq198_HTML.gif are the 4 × 4 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq199_HTML.gif diagonal matrices with 2 consecutive ones and 2 consecutive zeros in the diagonal in a cyclic arrangement, see [[15], Definition 7.3.1 and Proposition 4.1.7]. It is easy to see that after a permutation of columns, the matrices (4.1) are block diagonal matrices consisting of 2 × 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq200_HTML.gif blocks taken from two consecutive columns (in the sense of cyclic arrangement) of the first two rows of W 0 ( 0 ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq201_HTML.gif and the last two rows of W 0 ( 1 ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq202_HTML.gif, respectively. Hence the determinants of the Birkhoff matrices (4.1) are
± | i ( j 1 ) ( r 2 ) i j ( r 2 ) i j α i j + 1 α | | i j + 2 α i j + 3 α ( i ) j + 1 ( i ) j + 2 | = ± i j ( r 1 ) ( 1 + i 1 r ) 2 α 2 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equah_HTML.gif
in Cases 1 and 2, i.e., r { 1 , 2 } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq203_HTML.gif, whereas
± | i ( j 1 ) ( r 2 ) i j ( r 2 ) ( i ) ( j 1 ) ( i ) j | | i j + 2 α i j + 3 α ( i ) j + 1 ( i ) j + 2 | = ± i j ( r 3 ) ( i 2 r i ) 2 α 0 , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equai_HTML.gif

in Cases 3 and 4. Thus, the problem (2.1), (2.5)-(2.8) is Birkhoff regular. □

5 Asymptotic expansions of eigenvalues

Let D, as a function of μ with λ = μ 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq70_HTML.gif, be the characteristic function of the problem (2.1), (2.5)-(2.8) with respect to the fundamental system y j https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq204_HTML.gif, j = 1 , 2 , 3 , 4 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq198_HTML.gif, with y j [ m ] ( 0 ) = δ j , m + 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq205_HTML.gif for m = 0 , 1 , 2 , 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq206_HTML.gif, where δ is the Kronecker delta. Denote by D 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq207_HTML.gif the corresponding characteristic function for g = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq1_HTML.gif. Note that the characteristic functions D 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq207_HTML.gif and ϕ 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq208_HTML.gif considered in Section 3 have the same zeros counted with multiplicity. Due to the Birkhoff regularity, g only influences lower order terms in D. Therefore, it can be inferred that outside the interior of the small squares R k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq150_HTML.gif, R k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq209_HTML.gif, i R k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq210_HTML.gif, i R k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq211_HTML.gif around the zeros of D 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq207_HTML.gif, | D ( μ ) D 0 ( μ ) | < | D 0 ( μ ) | https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq212_HTML.gif if | μ | https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq213_HTML.gif is sufficiently large. Since the fundamental system y j https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq204_HTML.gif, j = 1 , 2 , 3 , 4 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq198_HTML.gif, depends analytically on μ, also D and D 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq207_HTML.gif are analytic functions. Hence, applying Rouché’s theorem both to the large squares S k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq158_HTML.gif and to the small squares which are sufficiently far away from the origin, it follows that the eigenvalues of the boundary value problem for general g have the same asymptotic distribution as for g = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq1_HTML.gif. Whence Proposition 3.2 leads to

Proposition 5.1 For g C 1 [ 0 , a ] https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq214_HTML.gif, there exists a positive integer k 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq114_HTML.gif such that the eigenvalues  λ ˆ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq115_HTML.gif, counted with multiplicity, of the problem (2.1), (2.5)-(2.8), where k Z { 0 } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq116_HTML.gif in Cases  1 and  2 and k Z https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq101_HTML.gif in Cases 3 and 4, can be enumerated in such a way that the eigenvalues  λ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq215_HTML.gif are pure imaginary for | k | < k 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq117_HTML.gif, and λ k = λ k ¯ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq216_HTML.gif for k k 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq119_HTML.gif. For k > 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq120_HTML.gif, we can write λ k = μ k 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq217_HTML.gif, where the μ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq218_HTML.gif have the following asymptotic representation as k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq123_HTML.gif:
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equaj_HTML.gif

In particular, the number of pure imaginary eigenvalues is even in Cases 1 and 2 and odd in Cases 3 and 4.

In the remainder of the section, we are going to establish more precise asymptotic expansions of the eigenvalues. According to [[15], Theorem 8.2.1], (2.1) has an asymptotic fundamental system { η 1 , η 2 , η 3 , η 4 } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq219_HTML.gif of the form
η ν ( j ) ( x , μ ) = δ ν , j ( x , μ ) e i ν 1 μ x ; ν = 1 , , 4 ; j = 0 , , 3 ; https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ20_HTML.gif
(5.1)
where
δ ν , j ( x , μ ) = [ d j d x j ] { r = 0 4 ( μ i ν 1 ) r φ r ( x ) e i ν 1 μ x } e i ν 1 μ x + o ( μ 4 + j ) , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ21_HTML.gif
(5.2)

and [ d j d x j ] https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq220_HTML.gif means that we omit those terms of the Leibniz expansion which contain a function φ r ( k ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq221_HTML.gif with k > 4 r https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq222_HTML.gif. Since the coefficient of y ( 3 ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq223_HTML.gif in (2.1) is zero, we have φ 0 ( x ) = 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq224_HTML.gif, see [[15], (8.2.3)].

We will now determine the functions φ 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq225_HTML.gif and φ 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq226_HTML.gif. In this regard, observe that n 0 = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq227_HTML.gif and l = 4 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq228_HTML.gif in the notation of [[15], (8.1.2) and (8.1.3)], see [[15], Theorem 8.1.2]. From [[15], (8.2.45)], we know that
φ r = φ 1 , r = ε 1 T V Q [ r ] ε 1 , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ22_HTML.gif
(5.3)
where ε ν https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq191_HTML.gif is the ν th unit vector in C 4 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq229_HTML.gif, V = ( i ( j 1 ) ( k 1 ) ) j , k = 1 4 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq230_HTML.gif, and Q [ r ] https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq231_HTML.gif are 4 × 4 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq199_HTML.gif matrices given by [[15], (8.2.28), (8.2.33) and (8.2.34)], that is, Q [ 0 ] = I 4 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq232_HTML.gif,
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ23_HTML.gif
(5.4)
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ24_HTML.gif
(5.5)
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ25_HTML.gif
(5.6)
where k 2 = g https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq233_HTML.gif, k 1 = g https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq234_HTML.gif, Ω 4 = diag ( 1 , i , 1 , i ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq235_HTML.gif and ε T = ( 1 , 1 , 1 , 1 ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq236_HTML.gif. Let G ( x ) = 0 x g ( t ) d t https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq237_HTML.gif. A lengthy but straightforward calculation gives
φ 1 = 1 4 G , φ 2 = 1 32 G 2 1 8 g , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ26_HTML.gif
(5.7)
and thus
η ν ( x , μ ) = ( 1 + 1 4 i ν + 1 G ( x ) μ 1 + ( 1 ) ν 1 ( 1 32 G ( x ) 2 1 8 g ( x ) ) μ 2 ) e i ν 1 μ x + { o ( μ 2 ) } e i ν 1 μ x https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ27_HTML.gif
(5.8)

for ν = 1 , 2 , 3 , 4 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq238_HTML.gif, where { o ( ) } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq239_HTML.gif means that the estimate is uniform in x.

The characteristic function of (2.1), (2.5)-(2.8) is
D ( μ ) = det ( γ j k exp ( ε j k ) ) j , k = 1 4 , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equak_HTML.gif
where
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equal_HTML.gif
Note that
D ( μ ) = m = 1 5 ψ m ( μ ) e ω m μ a , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ28_HTML.gif
(5.9)

where ω 1 = 1 + i https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq240_HTML.gif, ω 2 = 1 + i https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq241_HTML.gif, ω 3 = 1 i https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq242_HTML.gif, ω 4 = 1 i https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq243_HTML.gif, ω 5 = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq244_HTML.gif, and each of the functions ψ 1 , , ψ 5 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq245_HTML.gif has asymptotic representations of the form c k μ k + c k 1 μ k 1 + + c k 0 μ k 0 + o ( μ k 0 ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq246_HTML.gif.

It follows from (5.9) that
D 1 ( μ ) : = D ( μ ) e ω 1 μ a = ψ 1 ( μ ) + m = 2 5 ψ m ( μ ) e ( ω m ω 1 ) μ a , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ29_HTML.gif
(5.10)
where ω 2 ω 1 = 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq247_HTML.gif, ω 3 ω 1 = 2 2 i https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq248_HTML.gif, ω 4 ω 1 = 2 i https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq249_HTML.gif, ω 5 ω 1 = 1 i https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq250_HTML.gif. If arg μ [ 3 π 8 , π 8 ] https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq251_HTML.gif, we have | e ( ω m ω 1 ) μ a | e sin π 8 | μ | a https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq252_HTML.gif for m = 2 , 3 , 5 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq253_HTML.gif and the terms ψ m ( μ ) e ( ω m ω 1 ) μ a https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq254_HTML.gif for m = 2 , 3 , 5 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq253_HTML.gif can be absorbed by ψ 1 ( μ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq255_HTML.gif as they are of the form o ( μ s ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq256_HTML.gif for any integer s. Hence, for arg μ [ 3 π 8 , π 8 ] https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq257_HTML.gif,
D 1 ( μ ) = ψ 1 ( μ ) + ψ 4 ( μ ) e ( ω 4 ω 3 ) μ a = ψ 1 ( μ ) + ψ 4 ( μ ) e 2 i μ a , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ30_HTML.gif
(5.11)
where
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ31_HTML.gif
(5.12)
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ32_HTML.gif
(5.13)
A straightforward calculation gives
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ33_HTML.gif
(5.14)
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ34_HTML.gif
(5.15)
For the other two factors in (5.12) and (5.13), we have to consider the four different cases.
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equam_HTML.gif
Therefore,
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ35_HTML.gif
(5.16)
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ36_HTML.gif
(5.17)
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equan_HTML.gif
Thus, we have
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ37_HTML.gif
(5.18)
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ38_HTML.gif
(5.19)
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equao_HTML.gif
Hence, we get
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ39_HTML.gif
(5.20)
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ40_HTML.gif
(5.21)
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equap_HTML.gif
Thus, we have
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ41_HTML.gif
(5.22)
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ42_HTML.gif
(5.23)
We already know by Proposition 5.1 that the zeros μ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq218_HTML.gif of D satisfy the asymptotic representations μ k = k π a + τ 0 + o ( 1 ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq258_HTML.gif as k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq123_HTML.gif. In order to improve on these asymptotic representations, write
μ k = k π a + τ ( k ) , τ ( k ) = m = 0 n τ m k m + o ( k n ) , k = 1 , 2 , . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ43_HTML.gif
(5.24)

Because of the symmetry of the eigenvalues, we will only need to find the asymptotic expansions as k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq123_HTML.gif. We know τ 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq259_HTML.gif from Proposition 5.1, and it is our aim to find τ 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq260_HTML.gif and  τ 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq261_HTML.gif. To this end, we will substitute (5.24) into D 1 ( μ k ) = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq262_HTML.gif and we will then compare the coefficients of k 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq263_HTML.gif, k 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq264_HTML.gif and k 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq265_HTML.gif.

Observe that
e 2 i μ k a = e 2 i τ ( k ) a = e 2 i τ 0 a exp ( 2 i a ( τ 1 k + τ 2 k 2 + o ( k 2 ) ) ) = e 2 i τ 0 a ( 1 2 i a τ 1 1 k ( 2 a 2 τ 1 2 + 2 i a τ 2 ) 1 k 2 + o ( k 2 ) ) , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ44_HTML.gif
(5.25)
while
1 μ k = a π k ( 1 + a τ ( k ) k π ) 1 = a k π a 2 τ 0 k 2 π 2 + o ( k 2 ) . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ45_HTML.gif
(5.26)
Using (5.11), D 1 ( μ k ) = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq262_HTML.gif can be written as
μ k γ ψ 1 ( μ k ) + μ k γ ψ 4 ( μ k ) e 2 i τ k a = 0 , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equ46_HTML.gif
(5.27)

where γ is the highest μ-power in ψ 1 ( μ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq255_HTML.gif and ψ 4 ( μ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq266_HTML.gif. Substituting (5.25) and (5.26) into (5.27) and comparing the coefficients of k 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq263_HTML.gif, k 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq264_HTML.gif and k 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq265_HTML.gif, we get

Theorem 5.2 For g C 1 [ 0 , a ] https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq214_HTML.gif, there exists a positive integer k 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq114_HTML.gif such that the eigenvalues λ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq267_HTML.gif, k Z https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq101_HTML.gif, counted with multiplicity, of the problem (2.1), (2.5)-(2.8), where k Z { 0 } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq268_HTML.gif in Cases  1 and 2 and k Z https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq101_HTML.gif in Cases 3 and 4, can be enumerated in such a way that the eigenvalues λ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq267_HTML.gif are pure imaginary for | k | < k 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq117_HTML.gif, and λ k = λ k ¯ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq269_HTML.gif for k k 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq119_HTML.gif, where λ k = μ k 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq270_HTML.gif and the μ k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq271_HTML.gif have the asymptotic representations
μ k = k π a + τ 0 + τ 1 k + τ 2 k 2 + o ( k 2 ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equaq_HTML.gif
and the numbers τ 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq259_HTML.gif, τ 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq260_HTML.gif, τ 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq261_HTML.gif are as follows:
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_Equar_HTML.gif

In particular, the number of pure imaginary eigenvalues is even in Cases 1 and 2 and odd in Cases 3 and 4.

Remark 5.3 In [14] we have considered the differential equation (2.1) with the same boundary conditions B 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq272_HTML.gif, B 4 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq273_HTML.gif at a as in this paper but with λ-independent boundary conditions at 0, that is, the boundary conditions B 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq274_HTML.gif also occur in [14]. Whereas in [14] the number of pure imaginary eigenvalues is odd in each case, this number is even in Cases 1 and 2 of this paper. We observe that in Cases 1 and 2, the λ-dependent part is the ‘dominating’ part of the boundary condition B 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq275_HTML.gif, in the sense that it has the highest μ-power arising as μ 2 j + k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq276_HTML.gif from λ j y [ k ] https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq277_HTML.gif, whereas in Cases 3 and 4 the λ-independent part is dominating. It may be interesting to investigate if, in general, the parity of the number of pure imaginary eigenvalues can be determined by the number of dominating λ-dependent parts in the boundary conditions.

We can observe that the functions ϕ 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq80_HTML.gif in the Cases 3 and 4 are respectively the same as in [14] since the corresponding dominating terms in the boundary conditions coincide. However, the numbers τ 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq260_HTML.gif and τ 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq261_HTML.gif differ from those of [14] in each case, which is due to the λ-term in the boundary condition B 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2012-106/MediaObjects/13661_2012_Article_209_IEq275_HTML.gif.

Declarations

Acknowledgements

This research was partially supported by a grant from the NRF of South Africa, Grant number 69659. Various of the above calculations have been verified with Sage.

Authors’ Affiliations

(1)
The John Knopfmacher Centre for Applicable Analysis and Number Theory, School of Mathematics, University of the Witwatersrand

References

  1. Kerimov NB, Aliev ZS: Basis properties of a spectral problem with a spectral parameter in the boundary condition (Russian). Mat. Sb. 2006, 197(10):65-86. translation in Sb. Math. 197, 1467-1487 (2006) 10.4213/sm1433MathSciNetView Article
  2. Kerimov NB, Aliev ZS: On the basis property of the system of eigenfunctions of a spectral problem with a spectral parameter in the boundary condition (Russian). Differ. Uravn. 2007, 43: 886-895. translation in Differ. Equ. 43, 905-915 (2007)MathSciNet
  3. Marletta M, Shkalikov A, Tretter C: Pencils of differential operators containing the eigenvalue parameter in the boundary conditions. Proc. R. Soc. Edinb., Sect. A, Math. 2003, 133: 893-917. 10.1017/S0308210500002730MathSciNetView Article
  4. Shkalikov AA: Boundary problems for ordinary differential equations with parameter in the boundary conditions (Russian). Tr. Semin. Im. I.G. Petrovskogo 1983, 9: 190-229. translation in J. Sov. Math. 33, 1311-1342 (1986)MathSciNet
  5. Wang A, Sun J, Zettl A: The classification of self-adjoint boundary conditions: separated, coupled, and mixed. J. Funct. Anal. 2008, 255: 1554-1573. 10.1016/j.jfa.2008.05.003MathSciNetView Article
  6. Wang A, Sun J, Zettl A: Characterization of domains of self-adjoint ordinary differential operators. J. Differ. Equ. 2009, 246: 1600-1622. 10.1016/j.jde.2008.11.001MathSciNetView Article
  7. Möller M, Zettl A: Symmetric differential operators and their Friedrichs extension. J. Differ. Equ. 1995, 115: 50-69. 10.1006/jdeq.1995.1003View Article
  8. Weidmann J Lect. Notes Math. 1258. In Spectral Theory of Ordinary Differential Operators. Springer, Berlin; 1987.
  9. Behncke H: Spectral analysis of fourth order differential operators. I. Math. Nachr. 2006, 279: 58-72. 10.1002/mana.200310345MathSciNetView Article
  10. Behncke H: Spectral analysis of fourth order differential operators. II. Math. Nachr. 2006, 279: 73-85. 10.1002/mana.200310346MathSciNetView Article
  11. Pivovarchik V, van der Mee C: The inverse generalized Regge problem. Inverse Probl. 2001, 17: 1831-1845. 10.1088/0266-5611/17/6/318MathSciNetView Article
  12. Möller M, Pivovarchik V: Spectral properties of a fourth order differential equation. J. Anal. Appl. 2006, 25: 341-366.
  13. Möller M, Zinsou B: Self-adjoint fourth order differential operators with eigenvalue parameter dependent boundary conditions. Quaest. Math. 2011, 34: 393-406. doi:10.2989/16073606.2011.622913 10.2989/16073606.2011.622913MathSciNetView Article
  14. Möller M, Zinsou B: Spectral asymptotics of self-adjoint fourth order differential operators with eigenvalue parameter dependent boundary conditions. Complex Anal. Oper. Theory 2012, 6: 799-818. doi:10.1007/s11785-011-0162-1 10.1007/s11785-011-0162-1MathSciNetView Article
  15. Mennicken R, Möller M North-Holland Mathematics Studies 192. In Non-self Adjoint Boundary Eigenvalue Problems. Elsevier, Amsterdam; 2003.

Copyright

© Möller and Zinsou; licensee Springer 2012

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.