Vous êtes sur la page 1sur 3

Analytical method of finding polynomial roots

by using the eigenvectors, eigenvalues apparatus.


Dr. Alexander Ivanov Dr. Humbet Aliyev Bakhyt Bakiyev
Computer Engineering Department Economics Faculty Department of Computer Engineering
Suleyman Demirel University Suleyman Demirel University and Telecommunication
Almaty, Kazakhstan Almaty, Kazakhstan International IT University
alexander.ivanov@sdu.edu.kz humbet.aliyev@sdu.edu.kz Almaty, Kazakhstan
b.bakiev@iitu.kz

AbstractAlgorithms of finding polynomials roots have been


a) x 0 and
proposed and they are continuously suggesting in different works
[5] which are devoted for solving as theoretical so application b) , such that x = x (1)
problems because of theirs complexity and sufficient perfection
(accuracy) in series of certain situations. A method of finding In the given case is called eigen or characteristic value
polynomials roots which is reduced to obtaining analytical of the matrix which satisfies to the given eigenvector;
representation of polynomial roots, and for this purpose the backwards, they say that vector x satisfies the eigenvalue
methods of eigenvectors of associated to this polynomial a
companion (Frobenius or similar to it) matrix is proposed. The order of finding eigenvalues is reduced to the
following operations:
Keywords eigenvalues; eigenvectors; Frobenius matrix;
polynomial roots;

I. INTRODUCTION
For polynomial a companion
(Frobenius) matrix is a matrix [1,2,3]:

The solution of the last equation represents the main


For all that the roots of the given polynomial coincide with complexity and executed as a rule with help of numerical
eigenvalues of matrix that is satisfies to the characteristic methods (approximately).
equation In the given work an order of solving eigenvalue and
eigenvectors problem is changed by the opposite, and it allows
us to find analytically necessary components (exactly).
On an example of third order polynomial we will illustrate
how its roots may be find by use eigenvectos of the matrix
similar to the a companion (Frobenius) one, which is chosen in
some suitable basis (use of the suggested method directly for
the a companion (Frobenius) matrix in the given basis is not
reasonable because it will be reduced to the initial
Method of finding eigenvalues and corresponding (characteristic) polynomial; it should be noted that
eigenvectors is described in details in [1,2,3,4]. Lets give eigenvectors depend upon chosen basis and for some of bases
some more important for us definitions from those sources: eigenvectors may be not determined). In some suitable basis
well build such eigenvectors the relations between which will
Vector is an eigenvector of matrix , if be possible to use for finding roots of the polynomial (the
eigenvalues of the matrix) by rather simple methods without
solving the characteristic equation).

II. EXAMPLE
Lets see an example. It is necessary to find roots of is the eigenvector of the matrix .
polynomial:
Simultaneously,

Companion matrix for the given polynomial has the form:

is the eigenvector of matrix ;

Meanings of those vectors may be necessary in future (its not


obligatory requirement) for composing linear combinations
For application the suggested method of finding roots of
which will allow to build second (or even first) order
the initial polynomial (or finding meanings of eigenvalues of
equations with only one unknown variable (as unknown
characteristic polynomial of the companion matrix) lets
variables will be considered components of eigenvector).
transfer this matrix to the similar one with help of similarity
transformation:
From (1) it follows that:
.
From the vectors collinearity it follows:
This is equal to:
(2).

In (2), they are angles of slopes of the considered vectors do


not change after the matrix action on them, since these vectors
are eigenvectors of this matrix.
Well get a matrix: For compactness of the description the indexes of vectors
coordinates will not be used and more simple type of notation
will be used, namely:

The properties of eigenvectors of this matrix well use for


finding roots of the initial polynomial (eigenvalues of
complain matrix).[1,2].
We have:
Such representation is possible, due to the property of any
eigenvector to have at least one non zero coordinate, and all
these coordinates are determined with a precision to a
constant, in a case when such representation doesnt
correspond the available meanings it may be changed by the
Lets denote: other one:

In our case we have:

It is an eigenvector of the matrix A, where:

is the initial value of the eigenvector. .

From here we get:


Also, it is possible to find it by using the much simpler way if
we take into account the well-known relation:

III. ONE ILLUSTRATIVE PARTICULAR CASE


The given example doesnt exhaust the all possible
variants of using the suggested method for finding roots of
From (2), and also taking into account meanings of vector
polynomials. In particular, it should be noted, this method in
coordinates in (3) and (4), we get:
some cases allows us to obtain analytically roots of
characteristic polynomial when eigenvectors of the considered
matrix are not determined. For example, for orthogonal matrix
eigenvectors are not determined. But the n - th degree of this
or matrix (the degree coincides with the dimension of the matrix)
is a diagonal matrix, and if we denote:

From this it follows that

It is the unknown eigenvector. The properties of this vector


By using ratio are used in the suggested method for finding one of the roots;
here j th coordinate of the vector which is chosen to be equal
, to constant (as a rule they take it as equal to 1) and the
we get: following equality is also used:

By analogy it is possible to find:

Here: j is a vectors coordinate, i and 0 degrees of the


So we have find eigenvectors: matrix, s- number of the root of polynomial.
From (5) it is easy to find one of the roots by evident:

Which is corresponds to .
And

References

Which is corresponds to . [1] Voevodin VV, Kuznetsov YuA. Matritsy i vychisleniya [Matrices
and computation]. Moscow, Nauka Publ., 1984.
[2] P. Lancaster and M. Tismenetsky, The Theory of Matrices, Second
By substitution in (1) ( x = x) instead of x and those vectors Edition, Academic Press, Orlando, 1985.
, we get: [3] Roger A. Horn and Charles R. Johnson, Matrix analysis, Cambridge
University Press, Cambridge, 1990.
[4] F. R. Gantmacher, Applications of the Theory of Matrices, Dover
Publications, 2005.
[5] Ruskakova Ya. A., Candidate of science dissertation topic:
By using some other relations it is not difficult to obtain: Primeneniye konstruktivnyh metodov issledovaniya ustoychivosti
system bolshogo poryadka v vychislitelnoy praktike, 01.01.09 VAK,
Russia, 2003.

Vous aimerez peut-être aussi