 Research
 Open Access
 Published:
Selfadjoint fourth order differential operators with eigenvalue parameter dependent and periodic boundary conditions
Boundary Value Problems volume 2017, Article number: 33 (2017)
Abstract
Fourth order eigenvalue problems with periodic and separated boundary conditions are considered. One of the separated boundary conditions depends linearly on the eigenvalue parameter λ. These problems can be represented by an operator polynomial \(L(\lambda)=\lambda ^{2}Mi\alpha\lambda KA\), where \(\alpha>0\), M and K are selfadjoint operators. Necessary and sufficient conditions are given such that A is selfadjoint.
Introduction
Higher order 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. Higher order eigenvalue problems are experiencing slow but steady developments. Some recent developments of higher order problems whose eigenvalue boundary conditions may depend on the eigenvalue parameter, including asymptotics of the eigenvalues can be found in [1–14].
The generalized Regge problem and the small transversal vibration of a homogeneous beam compressed or stretched problems have boundary conditions with partial first derivatives with respect to the time variable t. The selfadjoint sixth order problem [8], the selfadjoint higher order problems [9] and the fourth order Birkhoff regular problems [10] also have the same type of boundary conditions described above. The mathematical model of these problems leads to eigenvalue problems with the eigenvalue parameter λ occurring linearly in the boundary conditions. Such problems have an operator representation of the form
in the Hilbert space \(H=L_{2}(I)\oplus\mathbb{C}^{k}\), where I is an interval, k is the number of eigenvalue dependent boundary conditions, M, K, and A are coefficient operators.
A classification of separated eigenvalue boundary conditions of 2nth order problems for which all the coefficient operators of the operator pencil (1.1) are selfadjoint is given in [9], Theorem 4.7, while an equivalent classification for fourth order problems is given in [4]. The boundary conditions investigated in [4, 9] are all separated. Möller and Pivovarchik [15] give necessary and sufficient conditions for an operator to be selfadjoint in terms of the null and image spaces of matrices defined by any type of boundary conditions for a 2nth order differential equation.
In this paper we extend the work of [1], using the same fourth order differential equation (3.1), to a class of boundary conditions, where two boundary conditions are periodic or antiperiodic at the end points, the remaining two boundary conditions are separated, one of them depends linearly on the eigenvalue parameter λ. The genesis of this problem is the problem studied by Möller and Pivovarchik [1]. In that problem the boundary condition that has dependence on the eigenvalue parameter has a specific meaning: that the hinge connection at the right end is subjected to a viscous friction \(\alpha>0\) in the hinge. In keeping with the pattern of the boundary conditions of the operator studied in [1], we confine our boundary conditions such that the two terms of the boundary condition that depends on the eigenvalue parameter are at one end point of the interval and the order of the derivative of the eigenvalue dependent part of this boundary condition is one less than the order of the eigenvalue independent part. We have associated to the problem under consideration the operator polynomial
where \(\alpha>0\), while M and K are selfadjoint operators. We give necessary and sufficient conditions for the operator A to be selfadjoint.
We give basic definitions and properties needed to conduct the study under investigation in Section 2. In Section 3 we prove that a particular fourth order periodic eigenvalue problem is selfadjoint using two different characterizations of selfadjoint operators. These characterizations are the Möller and Pivovarchik characterization for general boundary conditions [15] and the Möller and Zinsou characterization for separated boundary conditions [4, 9]. In Section 4 we present, for the fourth order eigenvalue problems investigated in this paper, the two different characterizations of selfadjoint operators used in Section 3 as matrix equations. The Möller and Pivovarchik characterization is given by \(U_{3}(N(U_{1}))=R(U^{*})\), while the Möller and Zinsou characterization is \(W(N(U_{1}))=R(U_{1}^{*})\), where \(U_{3}\) is a \(8\times10\) matrix, \(U_{1}\) is a \(3\times8\) matrix, U is a \(5\times10\) matrix and W a \(8\times8\) matrix of rank6. Finally, in Section 5 we consider a class of periodic eigenvalue problems consisting of two periodic boundary conditions and two separated boundary conditions, one of them depends on the eigenvalue parameter. We derive necessary and sufficient conditions for which the coefficient operator A is selfadjoint and we provide the structure of the boundary conditions using singular value decomposition.
Preliminaries
A Sobolev space is defined as
where \(a>0\) and \(m\in\mathbb {N}\).
Let \(n=2k\) where \(k\in\mathbb{N}\). We consider an nth order differential expression ℓ of the form
on an interval \([0,a]\), \(a>0 \), where \(g_{m}\in W_{2}^{m}(0,a)\), \(m=0,\ldots ,k\), are real valued functions and \(\vert g_{k}(x)\vert >\varepsilon\) for some \(\varepsilon>0\) and \(x\in[0,a]\). The differential expression ℓy is well defined for \(y\in W_{2}^{n}(0,a)\) in which case \(\ell y\in L_{2}(0,a)\). The operator \(L_{0}\) defined by
is called the maximal operator associated with the differential expression ℓ on \([0,a]\).
Definition 2.1
Let \(y\in W_{2}^{n}(0,a)\). For \(j=0,\ldots,n\) the jth quasiderivative of y, denoted \(y^{[j]}\), is recursively defined by
The quasiderivatives depend on the differential expression (2.1). They are convenient for the formulation of the Lagrange identity when dealing with differential operators which have fairly general coefficients. Let
be elements of the Hilbert space \(L_{2}(0,a)\oplus\mathbb{C}\), \(y,z,w\in W_{2}^{n}(0,a)\).
A formulation of the Lagrange identity and Green’s formula is quoted below from [15], Theorem 10.2.3.
Theorem 2.2
For a differential expression ℓ and \(y,z \in W_{2}^{n}(0,a)\), the Lagrange identity
holds on \([0,a]\) almost everywhere, where
and Green’s formula
is valid, where \((\cdot,\cdot)\) is the inner product in \(L_{2}(0,a)\).
Let \(r,q\in\mathbb{N}_{0}\), \(U_{1}\) a \(r\times2n\) matrix, \(U_{2}\) a \(q\times2n\) matrix, V a \(q\times2n\) matrix. Then the operator A in the Hilbert space \(L_{2}(0,a)\oplus\mathbb{C}\) is defined by
where
For \(m\in\mathbb{N}\) define
Finally define
Before stipulating a criterion of selfadjointness, we give a proposition which states conditions under which \(Z\in D(A^{*})\), quoted from [15], Proposition 10.3.3.
Proposition 2.3
Assume that \(\operatorname {rank}\bigl({\scriptsize\begin{matrix}{}U_{1} \cr U_{2}\end{matrix}} \bigr) =r+q\). Then \(Z\in D(A^{*})\) if and only if \(Z \in W^{n}_{2}(0,a) \oplus \mathbb{C}\) and there is \(e \in\mathbb{C}\) such that
for all \(\hat{Y} \in N(U_{1})\). For \(Z\in D(A^{*})\), e is unique and
A criterion of selfadjointness as given by [15], Theorem 10.3.5, is quoted below.
Theorem 2.4
Assume that
Then A is selfadjoint if and only if
In addition to determining if A is selfadjoint, we use [15], Theorem 10.3.8, quoted below to conclude that A is bounded below.
Theorem 2.5
Assume that A is selfadjoint. Then A has a compact resolvent. Assume additionally that

(i)
\((1)^{k}g_{k}>0\),

(ii)
each component of \(U_{1}\hat{Y}\) either contains only quasiderivatives \(y^{[m]}\) with \(m< k\) or contains only quasiderivatives \(m\geq k\),

(iii)
each component of \(U_{2}\hat{Y}\) either contains only quasiderivatives \(y^{[m]}\) with \(m< k\) or contains only quasiderivatives \(m\geq k\),

(iv)
for each component of \(U_{2}\hat{Y}\) which only contains quasiderivatives \(y^{[m]}\) with \(m\geq k\), the corresponding component of VŶ only contains quasiderivatives \(y^{[m]}\) with \(m< k\).
Then A is bounded below.
Any \(m\times n\) matrix can be decomposed into a diagonal matrix of its singular values and orthogonal matrices of order m and n as stated in [16], Theorem 6.1, quoted below as
Theorem 2.6
Any \(m\times n\) real matrix Γ, with \(m\geq n\), can be factorized as
where \(\Delta\in\mathbb{R}^{m\times m}\) and \(\Theta\in\mathbb {R}^{n\times n}\) are orthogonal, and \(\Sigma\in\mathbb{R}^{n\times n}\) is diagonal,
where \(\sigma_{1}\geq\sigma_{2}\geq\cdots\geq\sigma_{n}\geq0\).
A particular problem
The boundary value problem with a fourth order differential equation
together with the following boundary conditions:
defined on the interval \([0,a]\), where \(a>0\), \(\alpha>0\), and \(g\in C^{1}[0,a]\) initiates the study. The boundary conditions (3.2) and (3.3) are periodic, while the boundary conditions (3.4) and (3.5) are separated, the boundary condition (3.5) is also dependent on the eigenvalue parameter λ. We will establish an operator approach to this problem by defining operators A, K, and M, which are coefficients of the operator polynomial in the eigenvalue parameter; see (3.6) below. Then the eigenfunctions of the operator polynomial L given by
correspond to the nontrivial solutions of (3.1)(3.5), where the operators A, K, and M are defined by
and
Proposition 3.1
The operators A, K, and M are selfadjoint, M and K are bounded, K has rank 1, \(M\geq0\), \(K\geq0\), \(M+K\gg0\), \(N(M)\cap N(A)= \{0\}\) and A is bounded below and has a compact resolvent.
Proof
The statements about K and M are obvious. If \((y,c)^{\intercal}\in N(M)\cap N(A)\) then \((y,c)^{\intercal}\in N(M)\) gives \(y=0\), and \((y,c)^{\intercal}\in D(A)\) where \(c=y'(a)\) leads to \(c=y'(a)=0\). Hence \(N(M)\cap N(A)= \{0\}\). We are going to use Theorem 2.4 to verify that A is selfadjoint. By the differential expression (2.1) with \(n=4\), \(g_{0}=0\), \(g_{1}=g \in C^{1}[0,a]\) and \(g_{2}=1\) one represents (3.1) as
The quasiderivatives associated with (3.1) are
The number of eigenvalue independent boundary conditions as given by (3.2)(3.4) corresponds with \(r=3\), leaving only one boundary condition dependent on the eigenvalue parameter meaning that \(q=1\). \(U_{1}\) is a \(3 \times8\) matrix and both \(U_{2}\) and V are \(1 \times8\) matrices given by
Then the operator A can also be defined in terms of these matrices as
similar to (2.8) and (2.7). We now specify the matrices \(J_{2}\), \(U_{3}\), and U as
and
\(J_{2}\) is a \(8\times8\) matrix, \(U_{3}\) is a \(10\times8\) matrix and U is a \(5\times10\) matrix where I in (2.12) is a \(1\times1\) matrix.
We find \(N(U_{1})\) and \(R(U^{*})\) as
and
Then we compare \(U_{3}(N(U_{1}))\) with \(R(U^{*})\), showing that they are equal and A is selfadjoint as postulated by Theorem 2.4. Lastly, A has a compact resolvent in view of Theorem 2.5. The coefficient of the highest derivative in the differential component of A is \(g_{2} = 1>0\) as required by Theorem 2.5(i). Particular values in assumptions of Theorem 2.5(ii)(iv) for (3.1)(3.5) are
The first and third component of \(U_{1}\hat{Y}\) have quasiderivatives of order zero and one. Hence their order given by m is less than \(k=2\), half the order of the differential equation and the second component has order three which is \(m=3\geq k=2\). The component of \(U_{2}\hat{Y}\) has order one which is less than k. A does not have components of \(U_{2}\hat{Y}\) with quasiderivatives that are greater than k and the condition on VŶ is irrelevant. Thus all the conditions of Theorem 2.5 are fulfilled and A is bounded below. □
An alternative criterion is used to show that (3.1)(3.5) is selfadjoint. First, define
Then \(W(N(U_{1}))\) and \(R(U_{1}^{*})\) are given by
and
A comparison of \(W(N(U_{1}))\) and \(R(U_{1}^{*})\) shows that \(W(N(U_{1}))=R(U_{1}^{*})\), which is a necessary and sufficient condition of selfadjointness, see Theorem 4.4 on p.10.
Periodic and a single eigenvalue dependent boundary condition
Consider on the interval \([0,a]\), where \(a>0\), the differential equation (3.1) with boundary conditions
where the matrices \(U_{1}\), \(U_{2}\), and V are of the following form:
Consider a particular case where \(U_{2}\) and V contain exactly one nonzero element such that the nonzero element of \(U_{2}\) is in a different column to the nonzero element of V and the nonzero elements of \(U_{1}\) are positioned such that the first column of (2.12) has linearly independent rows. The operator A in (2.8) is given by
We recall that the dimension of the domain of a linear map between two spaces is given by the sum of the dimension of the null space and the rank of this linear map. In addition, two finite dimensional spaces coincide if one space is contained in the other and their dimensions are equal. A vector space \(\mathbb{C}^{8}\) acted upon by these three matrices \(U_{1}\), \(U_{2}\), and V means that \(\operatorname {rank}U_{2}\) and rankV are given by
respectively.
Proposition 4.1
Let \(U_{2}\) and V contain exactly one nonzero element such that the nonzero element in \(U_{2}\) is in a different column to the nonzero element in V. Let
Then \(U_{2}^{*}V\) and \(V^{*}U_{2}\) are \(8\times8\) matrices of rank 1, \(U_{2}^{*}VV^{*}U_{2}\) is an \(8\times8\) matrix of rank 2 and W is an \(8\times8\) matrix of rank at least 6.
Proof
Let the nonzero element of V be at \(j=p\) and that of \(U_{2}\) be at \(j=s\), \(s\neq p\). Then
has exactly one nonzero element, \(\overline{u^{2}}_{1s} v_{1p}\), at \(j=s\), \(i=p\). The position of the only nonzero element of \(V^{*}U_{2}\) is in row p and column s, thus \(U_{2}^{*}VV^{*}U_{2}\) has rank 2. \(J_{2}\) in (4.6) is invertible with rank 8 and \(U_{2}^{*}VV^{*}U_{2}\) has rank 2. Hence, the rank of W is at least 6. □
Remark 4.2
Whenever \(Y\in D(A)\) then \(\hat{Y}\in N(U_{1})\), and for every \(u \in N(U_{1})\) there is a \(Y\in D(A)\) such that \(\hat{Y}=u\).
Corollary 4.3
If A is selfadjoint then \(\operatorname {rank}W=6 \) and \(W(N(U_{1}))=R(U_{1}^{*})\).
Proof
Proposition 2.3 states that \(Z\in D(A^{*})\) if and only if \(Z \in W^{4}_{2}(0,a) \oplus \mathbb{C}\) and there is \(e \in\mathbb{C}\) such that
for all \(\hat{Y} \in N(U_{1})\). For \(Z\in D(A^{*})\), e is unique and
We use (2.3) for \(W, Z\in D(A)=D(A^{*})\) and
together with values of e and d as implied by (2.8) and (2.7), respectively, which we substitute into (4.7) to get
where Ŷ and Ẑ are as defined in (2.9). This means that \(W\hat{Y}\perp\hat{Z}\), i.e. \(W(N(U_{1}))\subset( N(U_{1}))^{\perp}=R(U_{1}^{*})\). We use this containment of \(W(N(U_{1}))\) in \(R(U_{1}^{*})\) to compare their dimensions as
Hence \(\operatorname {rank}W \leq6\). By Proposition 4.1 \(\operatorname {rank}W=6\), and hence all the inequalities in (4.8) are equalities and \(\dim(W(N(U_{1})))= \dim( R(U_{1}^{*}))\) holds. Thus \(W(N(U_{1}))= R(U_{1}^{*})\). □
Theorem 4.4
The following statements are equivalent:

(i)
A is selfadjoint,

(ii)
\(U_{3}(N(U_{1}))=R(U^{*})\),

(iii)
\(W(N(U_{1}))=R(U_{1}^{*})\).
Proof
Suppose (i) holds. Then Corollary 4.3 implies (iii).
Suppose (iii) holds. Let \(u\in N(U_{1})\). Then there is \(v \in D(U_{1}^{*}) \) such that \(Wu = U_{1}^{*}v\) i.e.
Consider
Let \(b=Vu\) and \(c=U_{2}u\) i.e. \(0=Vu+b\) and \(0=U_{2}uc\) and substitute (4.9) below. Then
Thus \(U_{3}(N(U_{1}))\subset R(U^{*})\) and \(\dim(U_{3}(N(U_{1})))\leq \operatorname {rank}U^{*}\). The map \(U_{1}: \mathbb{C}^{8} \rightarrow \mathbb{C}^{3}\), in (4.3), has \(\dim(N(U_{1}))= \dim(\mathbb{C}^{8})  \operatorname {rank}U_{1}= 8  3=5\) as given by the rank nullity theorem. Similarly U with the first column given by (4.3)(4.5) has \(\operatorname {rank}U= 5\) thus \(\dim(N(U_{1})) =\operatorname {rank}U^{*}\). We then conclude that \(U_{3}(N(U_{1}))= R(U^{*})\) by showing that \(U_{3}\) is injective i.e. 0 is the only element in \(N(U_{3})\). Suppose \(U_{3}u=0\). Then
and \(J_{2}u=0\) implies \(u=0\) since \(J_{2}\) is invertible. Hence (ii) follows.
Suppose that (ii) holds. Then by Theorem 2.4 we have (i). □
Further examples of selfadjoint operators with periodic and a single eigenvalue dependent boundary conditions
Keeping with the pattern of the boundary conditions of the operator studied in [1], using the differential equation (3.1) and Theorem 2.4, we identify the boundary conditions of the selfadjoint operators under investigation as follows:
where \(\beta_{m}\in\{0,1,2,3\}\), \(m=1,2,\ldots,5\); the \(\beta_{m}\) are distinct for \(m=1,2,3\) i.e. \(\beta_{s}\neq \beta_{m}\) for \(s\neq m\) with \(s,m=1,2,3\). \(\beta_{1}\), \(\beta_{2}\), \(\beta _{4}\), \(\beta_{5}\) are different from each other and \(\beta_{5}=\beta_{4}1\), \(\beta_{1}<\beta_{2}\), \(\epsilon_{j}=\pm1\) for \(j=1,2,3\) and \(\delta\in \{0,1\}\). We give necessary and sufficient conditions for which the main operator A is selfadjoint.
Theorem 5.1
The quadratic operator polynomial representing the fourth order differential equation (3.1) with the boundary conditions (5.1)(5.4) is selfadjoint if and only if these boundary conditions have the following structure:
Proof
Consider the matrices \(U_{1}\), \(U_{2}\), and V of the form (4.3)(4.5). Let the nonzero elements of \(U_{2}\) and V be at \(u^{2}_{1,2}\) and \(v_{1,3}\), respectively. Using the representation of (5.4), these corresponds to \(\beta_{4}=2\) and \(\beta_{5}=1\). Let \(\epsilon_{3}=1\), \(\beta_{1}=0\), \(\epsilon_{1}=1\), \(\beta_{2}=3\), \(\epsilon_{2}=1\) and \(\beta_{3}=2\). Starting with this choice of \(U_{2}\) and V, which implies that \(\delta=0\), \(U_{1}\) given by these parameters is
Then we consider \(U_{2}\) and V where the nonzero elements are at \(u^{2}_{1,6}\) and \(v_{1,7}\), respectively, correspond to \(\beta _{4}=2\), \(\beta_{5}=1\), \(\delta=1\) and \(\epsilon_{3}=1\). A matrix \(U_{1}\) with such periodic boundary conditions is given by
The assumption of Theorem 2.4 is fulfilled since \(\operatorname {rank}\bigl({\scriptsize\begin{matrix}{}U_{1} \cr U_{2}\end{matrix}} \bigr) =4\) for both (5.11) and (5.12) together with their corresponding \(U_{2}\). For each \(U_{1}\) we compute \(U_{3}(N(U_{1}))\) and the corresponding \(R(U^{*})\). The result is that \(U_{3}(N(U_{1}))=R(U^{*})\) for each of the two cases and any of the combination of the parameters stated. Thus the operator A for each of the 12 cases is selfadjoint. A selfadjoint quadratic operator polynomial representing the fourth order differential equation (3.1) with boundary conditions that satisfy (5.1)(5.4) satisfies
If we represent boundary conditions with
such that (5.1)(5.4) is satisfied. Then, using Matlab, we prove that if \(U_{3}(N(U_{1}))=R(U^{*})\) then the values of parameters β, δ, and ϵ are as given by (5.5)(5.10). □
We find an unifying structure of boundary conditions that are periodic or antiperiodic at the end points of the interval and have an eigenvalue parameter dependence in one of them as described by Theorem 2.6. The matrix \(U_{4}\) defined below was decomposed into its singular values and orthogonal matrices in an effort to find a relationship in all the cases.
Define a matrix
All the \(U_{4}\) that result from (5.1)(5.4) and satisfy Theorem 5.1 are such that each \(U_{4}\) column has at most one nonzero element and each of its rows has at least one nonzero element.
Theorem 5.2
The selfadjoint quadratic operator polynomial representing the fourth order differential equation (3.1) with boundary conditions (5.1)(5.4) that satisfy Theorem 5.1 has
where \(\Theta=I_{5}\), \(\Sigma=\operatorname {diag}(\sqrt{2},\sqrt {2},1,1,1)\) and \(\Delta^{\intercal}\in\mathbb{R}^{8\times8}\).
Proof
Consider (5.1)(5.4) with \(\beta_{1}=0\), \(\beta _{2}=3\), \(\beta_{3}=1\), \(\beta_{4}=2\), \(\beta_{5}=1\), \(\delta=0\) and \(\epsilon_{1},\epsilon_{2}, \epsilon_{3}=1\). This choice of parameters results in \(U_{1}\), \(U_{2}\), and V given in (3.8)(3.10). We then compute singular values of \(U_{4}\) with
Then
The eigenvalues of \(U_{4}U_{4}^{\intercal}\) are \(\sigma_{1}=2\) with eigenvectors \((1\ 0\ 0\ 0\ 0 ) ^{\intercal}\), \((0\ 1\ 0\ 0\ 0 ) ^{\intercal}\) and \(\sigma_{2}=1\) with \((0\ 0\ 1\ 0\ 0 ) ^{\intercal}\), \((0\ 0\ 0\ 1\ 0 ) ^{\intercal}\), and \((0\ 0\ 0\ 0\ 1 ) ^{\intercal}\). We construct a matrix C whose columns are the eigenvectors of \(U_{4}U_{4}^{\intercal}\) and order these eigenvectors by the magnitude of their eigenvalues i.e. \(C= ( e_{1}\ e_{2}\ e_{3}\ e_{4} \ e_{5} ) \). Then we implement the GramSchmidt orthonormalization process which in this case is \(\Theta= I_{5\times5}\). We repeat the process with \(U_{4}^{\intercal}U_{4}\) to find \(\Delta ^{\intercal}\). The eigenvalues of \(U_{4}^{\intercal}U_{4}\) are 2, 1, and 0 with multiplicities of two, three and three, respectively. We list the eigenvectors of \(U_{4}^{\intercal}U_{4}\) as columns of \(D= (d_{i})_{1}^{8}\) ordered below in decreasing magnitude of their eigenvalues as
Then we project the \(d_{i}\) and normalize them as before, which gives
The operators have the same Θ and Σ with \(\Delta^{\intercal}\) being the only distinguishing matrix in the decompositions of their \(U_{4}\), where \(W=J_{2}+U_{2}^{*}VV^{*}U_{2}\), and \(U_{3}= \left({\scriptsize\begin{matrix}{}J_{2}\cr V\cr U_{2}\end{matrix}} \right) \). □
References
 1.
Möller, M, Pivovarchik, V: Spectral properties of a fourth order differential equation. J. Anal. Appl. 25, 341366 (2006)
 2.
Wang, A, Sun, J, Zettl, A: The classification of selfadjoint boundary conditions: separated, coupled, and mixed. J. Funct. Anal. 255, 15541573 (2008)
 3.
Wang, A, Sun, J, Zettl, A: Characterization of domains of selfadjoint ordinary differential operators. J. Differ. Equ. 246, 16001622 (2009)
 4.
Möller, M, Zinsou, B: Selfadjoint fourth order differential operators with eigenvalue parameter dependent boundary conditions. Quaest. Math. 34, 393406 (2011). doi:10.2989/16073606.2011.622913
 5.
Möller, M, Zinsou, B: Spectral asymptotics of selfadjoint fourth order boundary value problem with eigenvalue parameter dependent boundary conditions. Bound. Value Probl. 2012, 106 (2012). doi:10.1186/168727702012106
 6.
Möller, M, Zinsou, B: Spectral asymptotics of selfadjoint fourth order differential operators with eigenvalue parameter dependent boundary conditions. Complex Anal. Oper. Theory 6, 799818 (2012). doi:10.1007/s1178501101621
 7.
Möller, M, Zinsou, B: Asymptotics of the eigenvalues of a selfadjoint fourth order boundary value problem with four eigenvalue parameter dependent boundary conditions. J. Funct. Spaces Appl. (2013). doi:10.1155/2013/280970
 8.
Möller, M, Zinsou, B: Sixth order differential operators with eigenvalue dependent boundary conditions. Appl. Anal. Discrete Math. 7, 378389 (2013). doi:10.2298/AADM130608010M
 9.
Möller, M, Zinsou, B: Selfadjoint higher order differential operators with eigenvalue parameter dependent boundary conditions. Bound. Value Probl. 2015, 79 (2015). doi:10.1186/s1366101503415
 10.
Zinsou, B: Fourth order Birkhoff regular problems with eigenvalue parameter dependent boundary conditions. Turk. J. Math. 40, 864873 (2016). doi:10.3906/mat150861
 11.
Möller, M, Zinsou, B: Asymptotics of the eigenvalues of selfadjoint fourth order differential operators with separated eigenvalue parameter dependent boundary conditions. Rocky Mt. J. Math. (to appear)
 12.
Demarque, R, Miyagaki, O: Radial solutions of inhomogeneous fourth order elliptic equations and weighted Sobolev embeddings. Adv. Nonlinear Anal. (2014). doi:10.1515/anona20140041
 13.
Baraket, S, Radulescu, V: Combined effects of concaveconvex nonlinearities in a fourth order problem with variable exponent. Adv. Nonlinear Stud. 16(3), 409419 (2016)
 14.
Han, X, Gao, T: A priori bounds and existence of nonreal eigenvalues of fourthorder boundary value problem with definite weight function. Electron. J. Differ. Equ. 2016, 82 (2016)
 15.
Möller, M, Pivovarchik, V: Spectral Theory of Operator Pencils, HermiteBiehler Functions, and Their Applications. Oper. Theory: Advances and Applications, vol. 246. Birkhäuser, Basel (2015)
 16.
Eldén, L: Matrix Methods in Data Mining and Pattern Recognition. Society for Industrial and Applied Mathematics, Philadelphia (2007)
Acknowledgements
This research was supported by a grant from the NRF of South Africa, Grant number 80956. Various of the above calculations have been verified with MathLab.
Author information
Additional information
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
The subject of this paper is part of the Ph.D. thesis of BM. The subject has been suggested and supervised by Prof Manfred Möller, BZ is the cosupervisor. The initial version of the paper has been written by BM. The submitted version has been verified and discussed by BM, BZ, and MM.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
About this article
Received
Accepted
Published
DOI
MSC
 34B07
 34B08
 47E05
Keywords
 fourth order differential operators
 eigenvalue dependent boundary conditions
 periodic boundary conditions
 quadratic operator pencil
 selfadjoint operators