![MathJax Logo](/templates/jsp/_style2/_tandf/pb2/images/math-jax.gif)
ABSTRACT
The set of eigenvalues of a square matrix P is denoted by , and the set of eigenvalues of the submatrix obtained from P by deleting its first i rows and columns of P is denoted by
. It is known that a symmetric pentadiagonal matrix may be constructed from
and
. The pairs
and
must interlace; the construction is not unique. In this paper we solve the inverse eigenvalue problem for H-Symmetric pentadiagonal matrices, not necessarily symmetric. Using
, not necessarily interlacing, and
may contain complex eigenvalues, we construct the solution by modified Lanczos algorithm.
1. Introduction
Let be a diagonal matrix such that
, where I is identity matrix. In
we define the inner product
A square real matrix P is said to be H-Symmetric if
The matrix
is called
adjoint of P and it is denoted by
. Let
be real numbers such that
, for
. It can be easily proved that any pentadiagonal real matrix P of the form
(1)
(1)
is a H-Symmetric matrix, where
. It is proved that [Citation1–3], if P is H-Symmetric and
is real and distinct, then the eigenvectors are H-Orthonormal, i.e. there are eigenvectors
of P such that
If H = I then P is pentadiagonal symmetric matrix. Pentadiagonal symmetric matrices arise in discrete vibrating beams, see [Citation4,Citation5,Citation8,Citation9] for more details. Inverse eigenvalue problems for symmetric pentadiagonal matrices are studied by many authors, for example see [Citation4–7,Citation9,Citation14]. In most cases the reconstruction procedure requires three interlacing real spectra. Using a finite difference method for discretizing non-smooth beams may lead to a nonsymmetric stiffness matrix, see [Citation12]. H-Symmetric pentadiagonal matrices of the form (Equation1
(1)
(1) ) appear in non-hermitian quantum mechanics [Citation11,Citation15]. Other inverse eigenvalue problems for H-symmetric matrices are studied in [Citation13,Citation16–18]. The objective of this paper is to study the inverse eigenvalue problem for H-Symmetric pentadiagonal matrices. Indeed using three given spectra that may or may not have interlacing property, we construct H-Symmetric pentadiagonal matrices of the form (Equation1
(1)
(1) ) such that
and
are the prescribed spectra. We use the modified form of Lanczos algorithm to construct the solution and we show that the solution is not unique. The solution obtained by this algorithm produces eigenvectors that for large size matrices may not be H-Orthonormal. To resolve this case we use a modified Gram–Schmidt orthogonalization procedure to make eigenvectors to be H-Orthonormal.
2. Construction of the solution
In this section, we state the main inverse eigenvalue problem and construct the solution. We consider conditions on the given data such that this problem has solution.
Inverse Problem.
Given three spectra construct a H-Symmetric pentadiagonal matrix P of the form (Equation1
(1)
(1) ) such that
We give three theorems that lead to the solution of this inverse problem. During the proof we give the conditions on the data for solvability of the inverse problem.
2.1. Real eigendata
In this subsection, we consider the case that the three given spectra are real. We denote the column vector consisting of ith components of the eigenvectors with . Indeed
where
are the eigenvectors. In the following lemma, we prove that the vectors
are also H-Orthonormal.
Lemma 2.1.
Let be H-Orthonormal eigenvectors of the matrix P then, the vectors
are also H-Orthonormal.
Proof.
Suppose and
, then we have
. The H-Orthonormality of the eigenvectors
implies that
. This implies that HU is invertible (HU is a square matrix), therefore
. Thus
. This means that the vectors
are H-Orthonormal which completes the proof.
In the following theorem, we construct H-Symmetric pentadiagonal matrix and H-Orthonormal eigenvectors by using eigenvalues and the first and second components of the eigenvectors. The proposed algorithm is called the modified Lanczos algorithm.
Theorem 2.2.
Let P be a H-Symmetric matrix with Let
and
are the vectors of first and second components of the eigenvectors of P, respectively, such that
Then the entries of P and other components of the eigenvectors can be constructed as follows:
where
.
Proof.
Suppose that is a matrix such that its columns are eigenvectors of P and
then,
. On transposing we find
, where
Thus the eigenvalue equation is as follows
(2)
(2)
Equating the first column on both sides of Equation (Equation2
(2)
(2) ) implies that
(3)
(3)
Multiplying both sides of (Equation3
(3)
(3) ) by
and summing up from j = 1 to j = n and using H-Orthonormal property of the vectors
and
we find
Similarly, multiplying both sides of (Equation3
(3)
(3) ) by
and summing up from j = 1 to j = n we find
which implies that
Using Equation (Equation3
(3)
(3) ) implies that
Taking to the power 2 and multiplying the last equation by
and summing up implies that
which concludes
with
.Again using Equation (Equation3
(3)
(3) ) we find
Now consider the second column of Equation (Equation2
(2)
(2) ) which is
(4)
(4)
Multiplying both sides by
and taking sigma, we find
Multiplying (Equation4
(4)
(4) ) by
and taking sigma we find
We can rewrite (Equation4
(4)
(4) ) as follows
Squaring both sides of the last equation and multiplying by
and summing up implies that
with
, and
In general for
equating ith column in Equation (Equation2
(2)
(2) ) we obtain
(5)
(5)
where
,
,
,
,
and
are computed previously and we suppose that
. Multiplying both sides of (Equation5
(5)
(5) ) by
and taking sum over j we find
Again multiplying both sides of (Equation5
(5)
(5) ) by
and taking sum over j we find
Equation (Equation5
(5)
(5) ) can be written as follows:
Squaring both sides of the last equation, multiplying by
, summing up over j and using H-Orthonormality we can find
and
. Thus all entries of the matrix P and the corresponding H-Orthonormal eigenvectors are constructed. Note that for solvability of the inverse problem by this Theorem the data
,
and
must be chosen such that
given by theorem to be nonnegative, otherwise the problem has no solution.
Now we are ready to construct the solution of the inverse problems by three given spectra. First, we compute the first entries of the eigenvectors of P by two given spectra.
Theorem 2.3.
Let P be H-Symmetric and
are real and distinct such that there exists a permutation of
like
such that
Then, the first entries of the eigenvectors of P i.e.
is computed as follows:
(6)
(6)
Proof.
We modify the proof in [Citation10] for H-symmetric case. Let P be H-Symmetric. By definition of the eigenvalues, we have
That is to say that
is the set of stationary points of the function
with condition
for
and
Thus
are stationary points of the function
where ν is a lagrange multiplier. Thus the equation
, implies
that is
Setting
, we find
Multiplying both sides by
implies that
Therefore
Thus
Equalizing the first components implies that
Imposing the condition
, implies that
Since
are the roots of the last equation, thus
Multiplying both sides of the last equation by
then taking limit
we find the required result.
Example 2.4.
Consider the matrix P as follows
The matrix P is H-Symmetric with
and we have
Conditions of Theorem 2.2 hold and using (Equation6
(6)
(6) ) we obtain, the first components of the eigenvectors of P as follows
Let be a matrix obtained from P by deleting the first row and column. The following theorem finds the relation between the eigenvalues of matrices P and
.
Theorem 2.5.
Let and
be
Symmetric matrix. Suppose
is the matrix with columns consisting the eigenvectors of
. Then the eigenvalues of P are the roots of the following equation:
(7)
(7)
Proof.
Let be an arbitrary eigenvector of P, then this vector can be written as follows:
(8)
(8)
where
and
. The equation
can be written as follows:
(9)
(9)
where
and
Multiplying both sides by
and simple calculation shows that
(10)
(10)
Combining the last equations together we find
According to Theorem 2.3 and Equation (Equation6
(6)
(6) ), if the matrices P and
have distinct eigenvalues then the first entry of the eigenvectors is not zero, i.e.
, thus
Remark 2.6.
Suppose that P and have a common eigenvalue
, then Equation (Equation6
(6)
(6) ) implies that the first entry of the eigenvector corresponding to
is zero. Substituting
and
in the system (Equation10
(10)
(10) ), we obtain
Note that
is different from zero otherwise the eigenvector X in (Equation8
(8)
(8) ) corresponding to the eigenvalue
will be zero which is not true, thus
. Therefore, if P and
have the set of common eigenvalues
then other eigenvalues of P are the roots of the following equation:
Note that since are the eigenvalues of Equation (Equation7
(7)
(7) ) thus
Multiplying both sides by
then replacing
, we find
(11)
(11)
For existence of the solution the right-hand side of the last equations must be nonnegative. Thus, there must be a permutation
of
such that the following fractions to be negative:
Now given three spectra, we find the second entries of the eigenvectors in the following theorem.
Theorem 2.7.
Let P be H-Symmetric and be
-Symmetric such that
. Then
where
is computed using (Equation11
(11)
(11) ).
Proof.
According to Theorem 2.3, we have
Now let
where
and
is eigenvector of P corresponding to the
. Thus we have
therefore
and
(12)
(12)
Since
is
-Symmetric and its eigenvalues are distinct, thus
(13)
(13)
Multiplying both sides of Equation (Equation12
(12)
(12) ) by
implies that
Thus
Therefore
can be computed as follows:
Equalizing the first components of both sides in the last equation and simple calculation completes the proof.
Now given the spectrum and having the first and second components of eigenvectors, i.e.
and
we can construct the matrix P using modified Lanczos algorithm (Theorem 2.2).
Remark 2.8.
Lanczos algorithm constructs pentadiagonal matrix P and H-Orthonormal eigenvectors. For large scale matrices the constructed eigenvectors might not be H-Orthonormal. To overcome this difficulty we use the modified Gram–Schmidt orthogonalization. This algorithm transforms the vectors into H-Orthonormal vectors as follows:
Put
, for
do the steps
and
,
Define
Set
Remark 2.9.
Due to the fact that the components of and
have signs +, −, moreover
have signs +, −, therefore the solution matrix is not unique. Therefore, in order to show the efficiency of the modified algorithm in the numerical examples, we compare the prescribed eigenvalues with the eigenvalues of the constructed matrix.
2.2. Complex eigenvalues in spectrum
In this section, we consider an inverse eigenvalues problem for pentadiagonal H-Symmetric matrix P using three spectra consisting of complex eigenvalues. That is we want to construct pentadiagonal matrix P such that may have some complex numbers. Indeed, we construct a pentadiagonal matrix P by prescribed eigendata
such that
are real and distinct, but
has some complex eigenvalues. Since P has complex eigenvalues, thus the corresponding eigenvectors do not make a H-Orthonormal matrix. Therefore, we can not use the Lanczos algorithm to construct matrix P. Using (Equation11
(11)
(11) ) we compute
, where
(14)
(14)
On the other hand, the eigenvectors of
are
-Orthonormal, thus simple calculation shows that
(15)
(15)
From (Equation14
(14)
(14) ), we find
(16)
(16)
Multiplying both sides of (Equation16
(16)
(16) ) by
and taking sum in i and considering the condition
, we obtain
(17)
(17)
therefore, we find
Now using the first and the second components of
, i.e.
and the spectrum
we may use the Lanczos algorithm to construct
. The entries
and
are found by (Equation15
(15)
(15) ) and (Equation17
(17)
(17) ). Using trace formula we find
as follows:
(18)
(18)
Therefore, P is constructed completely.
3. Numerical examples
In this section, to display the efficiency of the proposed method, some numerical experiments are considered. The numerical computations are performed using MATLAB R2015a on an Intel(R) Core(TM) i5 system. In order to test and evaluate the obtained numerical results, first we consider P as a known matrix and compute the eigendata ,
,
using eig function in Matlab. Then by using these eigendata and proposed method we reconstruct the matrix and call it
. Finally, we compare the spectral data of P and
.
Example 3.1.
Consider the H-symmetric pentadiagonal matrix of the form (Equation1(1)
(1) ) with entries
In Tables and , for different values of n, we compared the spectra of computed matrix
with the initial given matrix P.
Table 1. Numerical results for Example 3.1 without modified Gram–Schmidt method.
Table 2. Numerical results for Example 3.1 with applying modified Gram–Schmidt method.
We consider the following numerical examples where the given matrix P has some complex eigenvalues.
Example 3.2.
Consider the H-symmetric pentadiagonal matrix of the form (Equation1(1)
(1) ) with entries
In Tables and , for different values of n, we compared the spectra of computed matrix
with the initial given matrix P.
Table 3. Numerical results for Example 3.2 without Modified Gram–Schmidt method.
Table 4. Numerical results for Example 3.2 with applying Modified Gram–Schmidt method.
Example 3.3.
Consider the H-symmetric pentadiagonal matrix of the form (Equation1(1)
(1) ) with entries
In Tables and , for different values of n, we compared the spectra of computed matrix
with the initial given matrix P.
Table 5. Numerical results for Example 3.3 without Modified Gram–Schmidt method.
Table 6. Numerical results for Example 3.3 with applying Modified Gram–Schmidt method.
4. Conclusions
In this paper, given three spectra , not necessarily interlacing, we construct H-Symmetric pentadiagonal matrix P that admits the spectrum, i.e.
The construction procedure shows that the solution is not unique. The significance of this paper is to consider the cases where
may contain complex eigenvalues. Moreover, in general the eigenvalues do not need to satisfy interlacing properties.
Acknowledgments
The authors would like to thank three anonymous reviewers for their grateful comments on earlier version of the paper.
Disclosure statement
No potential conflict of interest was reported by the author(s).
References
- Bebiano N, da Providencia J. Inverse spectral problems for structured pseudo-symmetric matrices. Linear Algebra Appl. 2013;438:4062–4074.
- Bebiano N, Fonseca CM, da Providencia J. An inverse eigenvalue problem for periodic Jacobi matrices in Minkowski spaces. Linear Algebra Appl. 2011;435:2033–2045.
- Bebiano N, da Providencia J. Inverse eigenvalue problems for pseudo-Jacobi matrices: existence and uniqueness. Inverse Probl. 2011;27:025005. 12 pp.
- Caddemi S, Calia I. The influence of the axial force on the vibration of the Euler-Bernoulli beam with an arbitrary number of cracks. Arch Appl Mech. 2012;82:827–839.
- Caddemi S, Calia I. Exact closed-form solution for the vibration modes of the Euler-Bernoulli beam with multiple open cracks. J Sound Vib. 2009;327:473–489.
- Chu MT, Diele F, Ragni S. On the inverse problem of constructing symmetric pentadiagonal Toeplitz matrices from their three largest eigenvalues. Inverse Probl. 2005;21:1879–1894.
- Li J, Dong L, Li G. A class of inverse eigenvalue problems for real symmetric banded matrices with odd bandwidth. Linear Algebra Appl. 2018;541:131–162.
- Ghanbari K, Mirzaei H. Inverse eigenvalue problem for pentadiagonal matrices. Inverse Probl Sci Eng. 2014;22(4):530–542.
- Gladwell GML. Inverse problems in vibration. New York: Kluwer Academic; 2004.
- Golub GH. Some modified matrix eigenvalue problem. SIAM Rev. 1973;15(2):318–334.
- Han RPS, Zu JW. Pseudo non-selfadjoint and non-selfadjoint systems in atructural dynamics. J Sound Vibrat. 1995;184(4):725–742.
- Metrovic M. Finite difference method for non-smooth beam bending. Int'l Conf. Scientific Computing; CSC'17.
- Mirzaei H. Inverse eigenvalue problem for pseudo-symmetric Jacobi matrices with two spectra. Linear Multilinear Algebra. 2018;66(4):759–768.
- Moghaddam MR, Mirzaei H, Ghanbari K. On the generalized inverse eigenvalue problem of constructing symmetric pentadiagonal matrices from three mixed eigendata. Linear Multilinear Algebra. 2015;63(6):1154–1166.
- Shubo MA. Generation of Gevrey class semigroup by non-selfadjoint Euler-Bernoulli beam model. Math Methods Appl Sci. 2006;29:2181–2199.
- Xu WR, Bebiano N, Chen GL. An inverse eigenvalue problem for pseudo-Jacobi matrices. Appl Math Comput. 2019;346:423–435.
- Xu WR, Bebiano N, Chen GL. On the construction of real non-selfadjoint tridiagonal matrices with prescribed three spectra. Electr Trans Numer Anal. 2019;51:363–386.
- Xu WR, Bebiano N, Chen GL. Generalized inverse spectral problem for pseudo-jacobi matrices with mixed eigendata. Inverse Probl Sci Eng. 2019;27(6):773–789.