See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/355043308

Iterative Determinant Method for Solving Eigenvalue Problems

Article · September 2014

CITATIONS READS

0 177

3 authors, including:

Owus. Ibearugbulem Uchenna Joseph Maduh

Federal University of Technology Owerri Yeungnam University
171 PUBLICATIONS 627 CITATIONS 9 PUBLICATIONS 684 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Owus. Ibearugbulem on 04 October 2021.

The user has requested enhancement of the downloaded file.

 ISSN (e): 2250 – 3005 || Vol, 04 || Issue, 9 || September – 2014 ||
International Journal of Computational Engineering Research (IJCER)

Iterative Determinant Method for Solving Eigenvalue Problems

Owus M. Ibearugbulem1, Osasona, E. S. and Maduh, U. J.2
1,2
Civil Engineering Department, Federal University of Technology, Owerri, Nigeria

ABSTRACT:
This paper presents iterative determinant method for solving eigenvalue problems. A matlab program that
operates iterative to evaluate the determinant of the problem was written. In the program, a trial
eigenvalue is used in the program to compute the determinant. A small value (iterator) is added to trial
eigenvalue to obtain a new trial eigenvalue, which is used to compute new determinant. The trial
eigenvalue in the sequence that made the determinants to move from negative values to positive value or
move from positive values to negative value becomes required eigenvalue. Some engineering problems
were used to validate the eigensolver. Some of the data from the eigensolver and the corresponding exact
data are 2.468 and 2.4678; 9.876 and 9.875; 60.001 and 60.00; 12.37 and 12.3695. Close look at

Key word: eigensolver, iterative, determinant, eigenvalue, matlab program, iterator

I. INTRODUCTION
According to Ibearugbulem et al (2013), the stiffness matrix [k], the geometric matrix [kg] and the
mass (inertia) matrix [ki] are formulated using the assumed shape function. This shape function is usually
assumed to approximate the deformed shape of the continuum. If the shape function assumed is the exact one, it
means the solution will converge to the exact solution. The inertia matrix and the geometric matrix formulated
using the assumed shape functions are called consistent mass matrix and consistent geometric matrix
respectively (Paz, 1980 and Geradin, 1980). In the work of Ibearugbulem et al (2013), the dynamic equation in
structural dynamic as:
[k] - [ki] = 0 (1)
these data
For static reveals
stability that the
problem, thepresent eigensolver
eigenvalue equationhas high degree
is given as: of accuracy in predicting the eigenvalues.
– Nc[kg] =the
[k] However, 0 accuracy depends on the size(2)of iterator. The smaller the iterator, the higher the accuracy
WhereandKslower
is thethe
material
speed ofstiffness matrix,
computation andkivise
is the matrix of inertia, kg is the geometric matrix  is natural
versa.
frequency parameter and Nc is the critical buckling load parameter. Of note here is that both k g and ki are
consistent matrices
Keywords: (that is iterative,
eigensolver, they are determinant,
non-diagonaleigenvalue,
matrices). matlab
The difficulty
program,posed by this type of eigenvalue
iterator
problem led many researchers to transform the consistent matrices to diagonal matrices as:
[k] - λ A[I] = 0 (3)
[k] - Nc A[I] = 0 (4)

Here, A[I] is the transformed diagonal inertia or geometric matrix. In dynamics, the diagonal inertia
matrix is often called lumped mass matrix. Many methods are adopted so far by researchers for solutions of
eigenvalue problems of equation (3) and (4). According to Ibearugbulem et al, (2013), the methods include
Jacobi method, Polynomial method, Iterative method and Householder’s method were used by (Greenstadt,
1960; Ortega, 1967; and James, Smith and Wolford, 1977. Others are Power method, Inverse iterative
(Wilkinson, 1965), Lanczos method (lanczos, 1950), Arnoldi method (Arnoldi, 1951; Demmel, 1997; Bai et al,
2000;Chatelin, 1993; and Trefethenand Bau, 1997), Davidson method, Jacobi-Davidson method (Hochstenback
and Notay, 2004; and Sleijpen and Vander Vorst, 1996), Minimum residual method, generalized minimum
residual method were used by Barrett et al, (1994), Multilevel preconditioned iterative eigensolvers (Arbenz and
Geus, 2005), Block inverse-free preconditioned Krylov subspace method (Quillen and Ye, 2010), Inner-outer
iterative method (Freitag, 2007), and adaptive inverse iteration method (Chen, Xu and Zou, 2010), Matrix
iterative-inversion (Ibearugbulem et al, 2013). Sadly, of all these methods, only polynomial and iterative-
inversion methods can handle the problems of equations (1) and (2). Other methods can only be used for
equation (3) and (4). However, polynomial method also becomes very difficult to use when the size of the
matrix exceeds 3x3. In the same way, iterative-inversion method is also difficult when the matrix size is large
and the speed of computation is very slow.
The essence of this paper is to present a method that can be used in solving eigenvalue problems of equations
(1), (2), (3), and (4) for any size of matrix with high speed and good accuracy

www.ijceronline.com Open Access Journal Page 28

 Iterative Determinant Method…

II. ITERATIVE DETERMINANT

In this method, a trial eigenvalue 0 or Nc0 of default value of zero shall be substituted into the
eigenvalue equation to obtain eigenvalue matrix as:
[k]- [ki] = [kk]0 (5)
The determinant of the eigenvalue matrix, [kk]0 shall then be computed and the value kept. The value
of this initial determinant, dt0 may be positive or negative. Now the value of the default eigenvalue shall be
increased by adding iterator (say 1, 0.1, 0.01, 0.001 etc. as desired) to it to obtain new eigenvalue, 1. This shall
be substituted into the eigenvalue equation to obtain [kk] 1. The determinant, dt1 of [kk]1 shall be computed. If
dt0 is negative and dt1 is also negative, then the actual eigenvalue has not been obtained and the iterator has to
be added to 1 to obtain 2. This process has to be continued until we reach a stage where n-1 > 0 and n < 0 or
n-1 < 0 and n > 0. Note here that the bigger the iterator, the faster the computation and less the accuracy, the
smaller the iterator, the slower the computation and more the accuracy. To ease the use of this method, a general
matlab program was written. The user is at will to modify the iterator, f and sub iterators ff (i) in the program,
where is 1, 2, 3 … n and n is the size of the square matrix involved. For instance, if the order of the expected
eigenvalue is 100, iterator and sub iterators of 1 or 0.1 can be used. In this case the expected accuracy shall be of
the order of iterator, 1 or 0.1. In the same way, if the order of expected eigenvalue is 1, iterator of 0.001 can be
used. The choice of iterator is guided by the speed and accuracy of the computation. In all, for high accuracy,
iterators of 0.001 and below are advisable but the speed may be very slow. Thus, the level of accuracy and speed
shall guide you in using your choice iterator in this program. The user is also at liberty to choose the range of
eigenvalues to compute for by adjusting “for r = 1:3” to say “for r = 1:5” if the size of the matrix is 5x5. The
stiffness matrix “prs” and the geometric matrix or inertia matrix “pri” in the program are to be entered by the
user. The iterating count ceiling, “k” in the program can also be raised from 6260 to any higher value to suit the
user. Number of sub iterators “ff(i) as we have them in the program is 7. The user is at liberty also to introduce
more sub iterators say ff(8), ff(9) etc. as matches the size of the matrix. Some engineering problems were used
to ascertain the adequacy of the method and the program.

III. NUMERICAL EIGENVALUE PROBLEMS

The program will be used to test the following problems.
2 0 1 1 0 0
1. −1 4 −1 − 𝜆 0 1 0 = 0 (Stroud, 1982)
−1 2 0 0 0 1
0.1 0.1 0.1 1 0 0
2. 0.1 0.2 0.2 − 𝜆 0 1 0 = 0 (James, Smith and Wolford, 1977)
0.1 0.2 0.3 0 0 1
204.8 −102.4 25.6 4.8762 −2.438 0.0762
3. −102.4 63.2 −18.8 −𝜆 −2.438 2.419048 −0.1381 = 0
25.6 −18.8 7.2 0.0762 −0.1381 0.0762
7.2 −25.6 −1.2 0.07619 −0.076 0.02381
4. −25.6 204.8 25.6 −𝜆 −0.07619 4.8762 0.07619 = 0
−1.2 25.6 7.2 0.02381 0.0762 0.07619

204.8 −102.4 25.6 0.406349 0.0635 −0.00635

5. −102.4 63.2 −18.8 −𝜆 0.0635 0.2063 −0.01587 = 0
25.6 −18.8 7.2 −0.00635 −0.016 0.001587

126.4 18.8 18.8 -102.4 0 -102.4 2.419048 0.138095 0.138095 -2.4381 0

18.8 14.4 -1.2 -25.6 -25.6 0 0.138095 0.15238 0.02381 -0.07619 -0.07619
18.8 -1.2 14.4 0 25.6 -25.6 0.138095 0.02381 0.15238 0 0.07619
6. - =0
-102.4 -25.6 0 204.8 0 0 -2.4381 -0.07619 0 4.87619 0
0 -25.6 25.6 0 204.8 0 0 -0.07619 0.07619 0 4.87619
-102.4 0 -25.6 0 0 204.8 -2.4381 0 -0.07619 0 0

www.ijceronline.com Open Access Journal Page 29

 Iterative Determinant Method…

IV. RESULT AND DISCUSSION

Table 1 shows result data from eigenvalue problems herein. Eigenvalues obtained from the program were
compared with the exact eigenvalues. The exact eigenvalues were obtained by trial and error means using
Microsoft excel worksheet. Any eigenvalue that made the determinant of the eigenvalue matrix [kk] to become
zero is the exact eigenvalue. The data from the program compared very well with the values from Microsoft
excel worksheet with high degree of accuracy. As said earlier, it was observed that with small iterators say 0.01,
we obtained more accurate eigenvalues with slow computing speed. We also confirmed that with big iterators
say 1.0 we obtained less accurate eigenvalues with fast computing speed.

Table 1: Result data from the Eigenvalue Problems

Eigenvalues
From iterative Determinant method Exact Eigenvalues
Problems 1st 2nd 3rd 4th 5th 1st 2nd 3rd 4th 5th
1 1 2 3 1 2 3
2 0.0308 0.0644 0.5049 0.0308 0.0644 0.5049
3 2.468 23.392 109.143 2.4678 23.3912 109.1422
4 9.876 60.001 170.128 9.875124 60 170.1276
5 12.37 494.266 12.3695 494.2658
6 15.1 27 42.1 83.1 133.8 15.075832 26.95929142 42 83.07804 133.72546
248

V. APPENDIX A (MATLAB PROGRAM)

nn = input('enter size of matrix');nn = nn*1;
p=0;m=1;k=6260;j=1;f=0.1;ff(1)=0.1;ff(2)=0.01;ff(3)=0.001;ff(4)=0.001;ff(5)=0.001;ff(6)=0.001;ff(7)=0.001;
for r = 1 : 3
while p < k
prs = [43.2 0 0 ;0 6.400093 0 ;0 0 6.4];
pri = [0.54824 0 0 ;0 0.01268 0 ;0 0 0.01268];
a = prs - p*pri;
for x = 1:nn
for y = 1:nn
c(x,y)= a(x,y) ;
end
end
d = det(c);
if (m < 1.1); t1 = d;end
m = m + 1;
if (j >nn); break; end
if ((d >= 0)&& (t1 <= 0));py(j) = p;j = j + 1;t1 = d;end
if ((d <= 0)&& (t1 >= 0));py(j) = p;j = j + 1;t1 = d; end
p = p + f;
end
p = p-f; f =ff(r);m=1;
end
REFERENCES
[1] Arnoldi, W. E. (1951). The principle of minimized iteration in the solution of the matrix eigenvalue problem, Quarterly of Applied
Mathematics, vol. 9, pp. 17– 29.
[2] Arbenz, P. andGeus, R. (2005). Multilevel preconditioned iterative eigensolvers for maxwell eigenvalue problems. Applied
numerical mathematics.Vol. 54 , issue 2 .pp 107 – 121
[3] Bai, Z. Demmel, J., Dongarra, J. Ruhe, A., and van der Vorst, H. (2000). Templates for the Solution of Algebraic Eigenvalue
Problems - A Practical Guide, SIAM, Philadelphia, PA,
[4] Barrett, R., Berry, M., Chan, T. F., Demmel, J., Donato, J., Dongarra, J., Eijkhout, V., Pozo, R., Romine, C.and van der Vorst, H. A.
(1994). Templates for the Solution of Linear Systems: Building Blocks for Iterative Methods, 2nd Edition, SIAM, Philadelphia, PA,
[5] Chatelin, F. (1993).Eigenvalues of matrices,JohnWiley&SonsLtd,Chichester,West Sussex, England, Originally published in two
separate volumes by Masson, Paris: Valeurspropres de matrices (1988) and Exercises de valeurspropres de matrices (1989).
[6] Demmel, J. W. (1997). Applied Numerical Linear Algebra, SIAM, Philadelphia, PA A.

www.ijceronline.com Open Access Journal Page 30

 Iterative Determinant Method…

[7] Freitag, M. (2007).Inner-outer Iterative Methods for Eigenvalue Problems - Convergence and Preconditioning.PhD Thesis
submitted to University of Bath
[8] Fullard, K. (1980). The Modal Method for Transient Response and its Application to Seismic Analysis. In J. Donea (Ed.), Advan ced
Structural Dynamics (pp.43-70)
[9] Geradin, M. (1980).Variational Methods of Structural Dynamics and their Finite Element Implementation. In J. Donea (Ed.),
Advanced Structural Dynamics (pp.1-42)
[10] (Greenstadt, J. (1960). Mathematical Methods for Digital Computers. Vol. 1, ed. A. Ralston and H. S. Wilf (New York: John Wiley
& sons). Pp. 84-91.
[11] Hochstenbach, M. E. andNotay, Y. (2004). The Jacobi-Davidson method, GAMM Mitt.,
[12] Ibearugbulem, O. M., Ettu, L. O., Ezeh, J.C., and Anya, U.C. (2013).Application of Matrix Iterative – Inversion in Solving
Eigenvalue Problems in Structural Engineering. International Journal of Computational Engineering Research, vol.3. pp: 17-22
[13] James, M. L., Smith, G. M. and Wolford, J. C. (1977). Applied Numerical Methods for Digital Computation: with FORTRAN and
CSMP. Harper and Row Publishers, New York.
[14] Key, S. W. (1980). Transient Response by Time Integration: Review of Implicit and Explicit Operators. In J. Donea (Ed.),
Advanced Structural Dynamics.
[15] Key, S. W. and Krieg, R. D. (1972) Transient Shell Response by numerical Time Integration.2nd US-JAPAN Seminar on Matrix
Methods in Structural Analysis. Berkeley, California.
[16] Lanczos, C. (1950). An iterative method for the solution of the eigenvalue problem of linear differential and integral operators,
Journal of Research of the National Bureau of Standards, vol. 45 , pp. 255–282.
[17] [Ortega, J. (1967). Mathematical Methods for Digital Computers. Vol. 2, ed. A. Ralston and H. S. Wilf (New York: John Wiley &
sons). Pp. 94-115.

www.ijceronline.com Open Access Journal Page 31

View publication stats