OF
SELÇUK UNIVERSITY
ON COMPUTING OF POSITIVE INTEGER POWERS OF SOME SPECIAL TYPES OF
MATRICES Ahmet ÖTELEŞ Ph.D THESIS Department of Mathematics June-2015 KONYA All Rights Reserved
iv ABSTRACT Ph.D THESIS
ON COMPUTING OF POSITIVE INTEGER POWERS OF SOME SPECIAL TYPES OF MATRICES
Ahmet ÖTELEŞ
THE GRADUATE SCHOOL OF NATURAL AND APPLIED SCIENCE OF SELÇUK UNIVERSITY
THE DEGREE OF DOCTOR OF PHILOSOPHY IN MATHEMATICS
Advisor: Assoc. Prof. Dr. Mehmet AKBULAK 2015, 66 Pages
Jury
Assoc. Prof. Dr. Mehmet AKBULAK Prof. Dr. Durmuş BOZKURT Prof. Dr. Ahmet Sinan ÇEVİK
Prof. Dr. Süleyman SOLAK Assoc. Prof. Dr. Ramazan TÜRKMEN
In this dissertation, we obtain formulas for the entries of positive integer powers of some certain tridiagonal and circulant matrices, in terms of the Chebyshev polynomials. Then we present some examples concerned with these formulas. Finally, we give some Maple 13 procedures that calculate these formulas.
Keywords: Chebyshev polynomials, Circulant matrices, Eigenvalues, Eigenvectors, Jordan's
v ÖZET
DOKTORA TEZİ
BAZI ÖZEL TİPTEKİ MATRİSLERİN POZİTİF TAMSAYI KUVVETLERİNİN HESAPLANMASI
Ahmet ÖTELEŞ
Selçuk Üniversitesi Fen Bilimleri Enstitüsü Matematik Anabilim Dalı
Danışman: Doç. Dr. Mehmet AKBULAK 2015, 66 Sayfa
Jüri
Doç. Dr. Mehmet AKBULAK Prof. Dr. Durmuş BOZKURT Prof. Dr. Ahmet Sinan ÇEVİK
Prof. Dr. Süleyman SOLAK Doç. Dr. Ramazan TÜRKMEN
Bu tezde, bazı tridiagonal ve circulant matrislerin pozitif tamsayı kuvvetlerinin elemanlarını Chebyshev polinomlarına bağlı olarak veren formüller elde ettik. Daha sonra bu formüllerle ilgili bazı örnekler sunduk. Son olarak bu formülleri hesaplayan Maple 13 prosedürleri verdik.
Anahtar Kelimeler: Chebyshev polinomları, Circulant matrisler, Jordan formu, Özdeğer,
vi
ACKNOWLEDGMENTS
I would like to thank my previous advisor Prof. Dr. Durmuş Bozkurt, has supported and guided me wonderfully during my academic life. I am very glad to meet with such an excellent person.
I would like to express my gratitude to Prof. Dr. Ahmet Sinan Çevik for his valuable suggestions.
I would like to thank the other members of my committee: Prof. Dr. Süleyman Solak and Prof. Dr. Ramazan Türkmen for their contributions during the preparation of this dissertation.
I am very thankful for the financial support of the Council of Higher Education of Turkey (YÖK).
I would like to thank the Chair of the Department of Mathematics, Prof. Dr. Tin-Yau Tam and Assoc. Prof. Dr. Erkan Nane for their great support during my studies at the Auburn University.
I would like to acknowledge my family for their love, support, attention and encouragement during my entire life.
Assoc. Prof. Dr. Mehmet Akbulak, my advisor, has supported, encouraged and guided me excellently throughout my studies. I am very grateful to him for teaching me the clear ways of doing research, providing me significant documentation for my research, listening me for hours, accompanying me during my studies. I am so pleased to meet such a wonderful person and great mathematician in my life. This dissertation has become valuable with his contributions.
Ahmet ÖTELEŞ KONYA-2015
vii
TABLE OF CONTENTS
ABSTRACT ... iv
ÖZET ... v
ACKNOWLEDGMENTS ... vi
TABLE OF CONTENTS ... vii
LIST OF SYMBOLS ... ix
1. INTRODUCTION ... 1
2. PRELIMINARIES ... 12
2.1. Eigenvalues, Eigenvectors and Similarity ... 12
2.2. Chebyshev Polynomials ... 13
2.2.1. The first-kind polynomial T ... 13n 2.2.2. The second-kind polynomial U ... 15n 2.3. Homogeneous Linear Systems ... 16
2.4. Jordan Blocks and Jordan Form ... 16
3. ON COMPUTING OF POSITIVE INTEGER POWERS FOR SOME CERTAIN TYPES OF COMPLEX TRIDIAGONAL MATRICES ... 18
3.1. Positive Integer Powers of One Type of Complex Tridiagonal Matrices ... 18
3.1.1. Eigenvalues and eigenvectors ofXn ... 19
3.1.2. Derivation of the formula for the entries of X ... 22nr 3.1.3. Numerical Considerations ... 24
3.2. Positive Integer Powers of One Type of Complex Tridiagonal Matrices with Eigenvalues on Imaginary Axis ... 27
3.2.1. Eigenvalues and eigenvectors of Y ... 27n 3.2.2. Derivation of the formula for the entries of Ynr ... 31
3.2.3. Numerical Considerations ... 33
3.3. Positive Integer Powers of A Certain Type of Complex Tridiagonal Matrix Family ... 35
3.3.1. Eigenvalues and eigenvectors of B ... 36n 3.3.2. Derivation of the formula for the entries of Bnr ... 38
viii
4. ON COMPUTING OF POSITIVE INTEGER POWERS FOR CERTAIN
REAL CIRCULANT MATRIX ... 46
4.1. General Expression for the Entries of A ... 46nr 4.2. Numerical Considerations ... 51
5. CONCLUSION ... 56
REFERENCES ... 57
APPENDICES ... 61
ix
LIST OF SYMBOLS
ij
A a : matrix Awith entries aij
T
A : transpose of matrix A
A : conjugate of matrix A
*
A : conjugate transpose of matrix A, i.e., * T
A A
1
A : inverse of matrix A
0 1 1
circ c c, , ,cn : circulant matrix with c c0, ,1 ,cn1 on the first row
det A : determinant of matrix A
1, 2, , n
diag
: diagonal matrix with
1, 2, ,
n on the diagonalI or E : identity matrix
n
I or E n : the n n identity matrix
n
M : n n matrices with complex number entries
z : conjugate of the complex number z
: the set of natural numbers
: the set of complex numbers
n
1
INTRODUCTION
A certain type of transformation of a set of numbers can be represented as the
multiplication of a vector by a square matrix. Repetition of the operation is
equivalent to multiplying the originial vector by a power of the matrix. Solving some
di¤erence equations, di¤erential and delay di¤erential equations and
boundary value problems, we need to compute the arbitrary integer powers of a square matrix. Properties of powers of matrices are thus of considerable importance (Cobb, 1958; Rimas, 1977; Rimas and Leonaite, 2006).
One can …nd the r th power (r 2 N) of an n n matrix A using the well-known expression
Ar = P JrP 1; (1.1)
where J is the Jordan’s form of A and P is the transforming matrix. Matrices J and P can be found, provided eigenvalues and eigenvectors of A are known (Horn and Johnson, 1985).
An n n tridiagonal matrix is one having the form
Tn := 2 6 6 6 6 6 6 6 6 6 6 6 6 4 a1 b1 0 0 c1 a2 b2 . .. ... 0 c2 a3 . .. . .. ... .. . . .. ... ... bn 2 0 .. . . .. cn 2 an 1 bn 1 0 0 cn 1 an 3 7 7 7 7 7 7 7 7 7 7 7 7 5 :
That is, Tn = [ti;j]1 i;j n is tridiagonal if ti;j = 0 for ji jj > 1 (Horn and
Johnson, 1985). These types of matrices often arise in many areas of mathematics
and engineering such as boundary value problems, parallel computing,
telecommunication system analysis, image processing and so on. In these areas, the computation of powers of tridiagonal matrices are necessary (Cobb, 1958; Rimas, 1977; Rimas and Leonaite, 2006). Therefore, there are a lot of studies dealing with the powers of tridiagonal matrices by using equation (1.1). Now we speak to those.
n n tridiagonal matrix B of the form B = 2 6 6 6 6 6 6 6 6 6 6 6 6 4 0 1 0 0 1 0 1 . .. ... 0 1 0 . .. ... ... .. . . .. ... ... 1 0 .. . . .. 1 0 1 0 0 1 0 3 7 7 7 7 7 7 7 7 7 7 7 7 5 ; (1.2)
of order n = 2p (p 2 N) ; in terms of Chebyshev polynomials.
The author …nds the rth power (r 2 N) of this matrix in (1.2) applying the
expression Br = T JrT 1 given by (1.1), where J is the Jordan’s form of B, T is
the transforming matrix. Matrices J and T can be found provided eigenvalues and eigenvectors of B are known. The eigenvalues of B are de…ned by the characteristic equation
jB Ej = 0: (1.3)
He …rstly de…nes n( ) as the following
n( ) = 1 0 0 1 1 . .. ... 0 1 . .. ... ... .. . . .. ... ... 1 0 .. . . .. 1 1 0 0 1 ; (1.4)
here 2 R. Taking (1.3) and (1.4) into account he writes
jB Ej = n( ) : (1.5)
From (1.4) follows:
n = n 1 n 2 2 = 2 1, 1 = , 0 = 1 ; (1.6)
here n = n( ). Solving di¤erence equation (1.6) he obtains
n( ) = Un
All the roots of the polynomial Un(x) are included in the interval [ 1; 1] and can
be found using the relation (Fox and Parke, 1968) xnk = cos
k
n + 1, k = 1; 2; 3; : : : ; n. (1.8)
Taking (1.5), (1.7) and (1.8) into account he …nds the roots of the
characteristic equation (1.3) (eigenvalues of B):
k = 2 cos
k
n + 1, k = 1; 2; 3; : : : ; n. Let n be an even number (n = 2p, p 2 N).
Since all the eigenvalues k (k = 1; 2; : : : ; n, n = 2p, p 2 N) are simple
eigenvalues (lk = 1; lk denotes the multiplicity of the eigenvalue k), for each
eigenvalue corresponds single Jordan cell J1( k) in the matrix J . Taking this into
account and applying the relation
k = n k+1 k = 1; 2; : : : ;
n
2 ,
he writes down the Jordan’s form of B:
J = diag n n 1 n 2 n2+1 n2+1 n 2 n 1 n .
Using the equality J = T 1BT he …nds the matrices T , T 1 and derives the
expression of the rth power (r 2 N) of B:
Br = T JrT 1 = 1 2n + 2Q (r) = 1 2n + 2(qij(r)) here qij(r) = h 1 + ( 1)r+i+ji n=2 X k=1 4 2n k+1 rn k+1Ui 1 n k+1 2 Uj 1 n k+1 2 ;
k (k = 1; n) are the eigenvalues of B and Uk(x) be the k th degree Chebyshev
polynomial of the second kind
Uk(x) =
sin ((k + 1) arccos x)
sin arccos x ; 1 x 1
(Rimas, 2005e).
There are a lot of studies on computing powers of tridiagonal matrices by using the method mentioned above.
Rimas (2005f) derives the general expression of the rth power (r 2 N) of the
n n matrix B given by (1.2) for n = 2p + 1 (p 2 N).
Rimas (2005g, 2005h) derives the general expression of the rth power
(r2 N, r > 2) for one type of tridiagonal matrices of the form
2 6 6 6 6 6 6 6 6 6 6 6 6 4 0 0 0 0 1 0 1 . .. ... 0 1 0 . .. ... ... .. . . .. ... ... 1 0 .. . . .. 1 0 1 0 0 2 0 3 7 7 7 7 7 7 7 7 7 7 7 7 5
of order n = 2p (p 2 N, p > 2) and n = 2p + 1 (p 2 N) with zeros in the …rst row. Rimas (2006c) derives the general expression of the rth power (r 2 N) for one type of skew-symmetric tridiagonal matrices of the form
2 6 6 6 6 6 6 6 6 6 6 6 6 4 0 1 0 0 1 0 1 . .. ... 0 1 0 . .. ... ... .. . . .. ... ... 1 0 .. . . .. 1 0 1 0 0 1 0 3 7 7 7 7 7 7 7 7 7 7 7 7 5 (1.9) of order n = 2p + 1, (p 2 N).
Rimas (2007c) derives the general expression of the rth power (r 2 N) for one type of tridiagonal matrices of the form
2 6 6 6 6 6 6 6 6 6 6 6 6 4 1 1 0 0 1 0 1 . .. ... 0 1 0 . .. ... ... .. . . .. ... ... 1 0 .. . . .. 1 0 1 0 0 1 0 3 7 7 7 7 7 7 7 7 7 7 7 7 5 of arbitrary order.
(r2 N; r > 2) for one type of tridiagonal matrices of the form 2 6 6 6 6 6 6 6 6 6 6 6 6 4 0 0 0 0 1 0 1 . .. ... 0 1 0 . .. ... ... .. . . .. ... ... 1 0 .. . . .. 1 0 1 0 0 0 0 3 7 7 7 7 7 7 7 7 7 7 7 7 5 of order n = 2p + 1, (p 2 N, p > 2).
Rimas (2007d) derives the general expression of the rth power (r 2 N) for one type of tridiagonal matrices of the form
2 6 6 6 6 6 6 6 6 6 6 6 6 4 1 1 0 0 1 0 1 . .. ... 0 1 0 . .. ... ... .. . . .. ... ... 1 0 .. . . .. 1 0 1 0 0 1 1 3 7 7 7 7 7 7 7 7 7 7 7 7 5 of order n, (n 2 N, n > 2).
Rimas (2007e) derives the general expression of the rth power (r 2 N) for one type of tridiagonal matrices of the form
2 6 6 6 6 6 6 6 6 6 6 6 6 4 1 1 0 0 1 0 1 . .. ... 0 1 0 . .. ... ... .. . . .. ... ... 1 0 .. . . .. 1 0 1 0 0 1 1 3 7 7 7 7 7 7 7 7 7 7 7 7 5 of order n (n 2 N).
(r2 N, r > 2) for one type of tridiagonal matrices of the form 2 6 6 6 6 6 6 6 6 6 6 6 6 4 0 0 0 0 1 0 1 . .. ... 0 1 0 . .. ... ... .. . . .. ... ... 1 0 .. . . .. 1 0 1 0 0 0 0 3 7 7 7 7 7 7 7 7 7 7 7 7 5
of order n = 2p + 1 (p 2 N) with zeros in the …rst and nth rows.
Kiyak et al (2010) derive the general expression of the rth power (r 2 N) for one type of tridiagonal matrix of the form
2 6 6 6 6 6 6 6 6 6 6 6 6 4 0 1 0 0 1 0 1 . .. ... 0 1 0 1 . .. ... .. . . .. 1 . .. . .. 0 .. . . .. ... . .. ( 1)n 1 0 0 ( 1)n 1 0 3 7 7 7 7 7 7 7 7 7 7 7 7 5 in terms of the Chebyshev polynomials of the second kind.
Rimas (2012) gives the eigenvalue decomposition for odd order tridiagonal 2-Toeplitz matrix of the form
2 6 6 6 6 6 6 6 6 6 6 6 6 4 a1 b1 0 0 c1 a2 b2 . .. ... 0 c2 a1 b1 . .. ... .. . . .. c1 a2 b2 0 .. . . .. c2 a1 . .. 0 0 . .. ... 3 7 7 7 7 7 7 7 7 7 7 7 7 5
and derive the explicit expression for integer powers of this matrix.
power (r 2 N) of the n n tridiagonal matrix Un of the form Un:= 2 6 6 6 6 6 6 6 6 6 6 6 6 4 0 2 0 0 1 0 1 . .. ... 0 1 0 . .. ... ... .. . . .. ... ... 1 0 .. . . .. 1 0 1 0 0 2 0 3 7 7 7 7 7 7 7 7 7 7 7 7 5 ; (1.10) for all n 2 N:
Rimas (2007a, 2007b) also obtain a general expression for the entries of the
r th power (r 2 N) of the n n tridiagonal matrix Vn of the form
Vn := 2 6 6 6 6 6 6 6 6 6 6 6 6 4 0 2 0 0 1 0 1 . .. ... 0 1 0 . .. ... ... .. . . .. ... ... 1 0 .. . . .. 1 0 1 0 0 2 0 3 7 7 7 7 7 7 7 7 7 7 7 7 5 ; (1.11) for all n 2 N:
Gutiérrez studies powers of tridiagonal matrices di¤erently from the method introduced by Rimas (2005e). For example; he consider the n n complex tridiagonal matrix A of the form
A = 2 6 6 6 6 6 6 6 6 6 6 6 6 4 a0 a 1 0 0 a1 a0 a 1 . .. ... 0 a1 a0 . .. ... ... .. . . .. ... ... ... 0 .. . . .. ... ... a 1 0 0 a1 a0 3 7 7 7 7 7 7 7 7 7 7 7 7 5 (1.12)
where a1a 1 6= 0: He …rstly give the following theorem for eigenvalue decomposition
of the matrix A in (1.12) (Gutiérrez, 2008c).
Theorem 1 Let a1; a0; a 1 2 C; a1a 1 6= 0 and n 2 N: Then W DW 1 is an
matrix W are Wjk = r 2 n + 1 r a 1 a 1 j sin jk n + 1; 1 j; k n;
and D = diag ( 1; 2; : : : n) with j
j = 0+ 2a 1 r a 1 a 1 cos j n + 1; 1 j n (Gutiérrez, 2008c).
After that, based on this decomposition the author presents the following result that provides a general expression for the entries of the qth power (q 2 N) of
A; in terms of Chebyshev polynomials of the second kind.
Theorem 2 Consider a1; a0; a 1 2 C; a1a 1 6= 0 and n 2 N: Let A be the
n n complex tridiagonal matrix in (1.12), = q a1
a 1 and h = 2 cos h n+1 for every 1 h n: Then [Aq]jk = j k 2n + 2 2 6 42 1 + ( 1)n+1 aq0Uj 1(0) Uk 1(0) + bn 2c X h=1 4 2n h+1 Uj 1 n h+1 2 Uk 1 n h+1 2 h (a0+ a 1 n h+1) q + ( 1)j+k (a0 a 1 n h+1) q ] 3 5
for all q 2 N and 1 j; k n; where bxc denotes the largest integer less than or
equal to x (Gutiérrez, 2008c).
Since the matrix A in (1.12) is a matrix family of the matrices given by (1.2) and (1.9), Gutiérrez deduce the expression given by Rimas for the entries of the powers of these matrices by using Theorem 2.
By the similar way mentioned above, Gutiérrez (2008d) derives the entries of
powers of an n n Hermitian tridiagonal matrix of the form
2 6 6 6 6 6 6 6 6 6 6 6 6 4 a0 a1 0 0 a1 a0 a1 . .. ... 0 a1 a0 . .. ... ... .. . . .. ... ... ... 0 .. . . .. ... ... a1 0 0 a1 a0 3 7 7 7 7 7 7 7 7 7 7 7 7 5
where a0 2 R and a1 6= 0.
An n n circulant matrix Cn := circn(c0; c1; : : : ; cn 1) is a square matrix
having the form
Cn:= 2 6 6 6 6 6 6 6 6 6 6 6 6 4 c0 c1 c2 : : : cn 2 cn 1 cn 1 c0 c1 . .. cn 2 cn 2 cn 1 c0 . .. . .. ... .. . . .. ... ... . .. c2 c2 . .. ... . .. c1 c1 c2 : : : cn 2 cn 1 c0 3 7 7 7 7 7 7 7 7 7 7 7 7 5 ;
where each row is a cyclic shift of the row above it. We may also write a circulant matrix in the form
Cn = cij = c(j i) mod n:
Circulant matrices arise; for example, in applications involving the discrete Fourier transform (DFT) and the study of cyclic codes for error correction (Gray, 2006).
Circulant matrices have a wide range of applications in mathematics and engineering such as graph theory, signal processing, coding theory and image processing, etc. Numerical solutions of certain types of elliptic and parabolic partial di¤erential equations with periodic boundary conditions often involve linear systems associated with circulant matrices (Tsitsas et al, 2007; Zhao, 2009).
The well-known eigenvalue decomposition of an n n circulant matrix (Davis, 1994) is that
circn(c0; c1; : : : ; cn 1) = FnDnFn; (1.13)
where * denotes conjugate transpose (i.e Fn = FTn), Fn is the n n Fourier matrix:
[Fn]u;v = 1 p ne 2 (u 1)(v 1) n i; 1 u; v n;
and Dn= diag ( 1; 2; : : : ; n) with
k = n X r=1 cr 1e 2 (k 1)(r 1) n i; 1 k n:
It can be given the entries of the r th power (r 2 N) of an n n circulant matrix by using (1.13) as the following:
[Cnr]i;j = [(FnDnFn) r ]i;j = [FnDrnFn]i;j = n X k=1 [Fn]i;k[DrnFn]k;j = n X k=1 [Fn]i;k r k[Fn]j;k = 1 n n X k=1 r ke 2 (i 1)(k 1) n ie 2 (j 1)(k 1) n i;
according to above the last equation, we get [Cnr]i;j = 1 n n X k=1 r ke 2 (k 1)(j i) n i: (1.14)
We can easily calculate the entries of the r th power (r 2 N) of an
n n circulant matrix by using (1.14). But in recent years, computing the integer
powers of certain types of circulant matrices depending on Chebyshev polynomials has been a very popular problem. For instance, Rimas obtain a general expression for the entries of the r th power (r 2 N) of the n n real symmetric circulant matrix of the form circn(0; 1; 0; : : : ; 0; 1) := 2 6 6 6 6 6 6 6 6 6 6 6 6 4 0 1 0 0 1 1 0 1 . .. 0 0 1 0 . .. ... ... .. . . .. ... ... 1 0 0 . .. 1 0 1 1 0 0 1 0 3 7 7 7 7 7 7 7 7 7 7 7 7 5 ;
in terms of the Chebyshev polynomials (see Rimas (2005c, 2005d) for the odd case and Rimas (2006a, 2006b) for the even case).
Gutiérrez (2008b) derive a single formula by generalizing the results of Rimas
(2005c, 2005d) for the entries of the positive integer powers of the n n complex
symmetric circulant matrix of odd and even order given as circn b0; b1; : : : ; bn 1 2 ; b n 1 2 ; : : : ; b1 T if n is odd, circn b0; b1; : : : ; bn2 1; bn2; bn2 1; : : : ; b1 T if n is even, (1.15)
in terms of the Chebyshev polynomials of the …rst kind and the author also derives two separate formulas for the entries of the positive integer powers of the
n n complex skew-symmetric circulant matrix of odd and even order given as circn 0; b1; : : : ; bn 1 2 ; b n 1 2 ; : : : ; b1 T if n is odd, circn 0; b1; : : : ; bn2 1; 0; bn2 1; : : : ; b1 T if n is even,
in terms of the Chebyshev polynomials of the …rst and second kinds (Gutiérrez 2008a).
Köken and Bozkurt (2011) consider the n n non-symmetric complex
circulant matrix of the form
circn(0; a; 0; : : : ; 0; b) := 2 6 6 6 6 6 6 6 6 6 6 6 6 4 0 a 0 0 b b 0 a . .. 0 0 b 0 . .. ... ... .. . . .. ... ... a 0 0 . .. b 0 a a 0 0 b 0 3 7 7 7 7 7 7 7 7 7 7 7 7 5 ; (1.16)
di¤erently from complex symmetric circulant matrix given by (1.15). They derive a general expression for the entries of the r th power (r 2 N) of the matrix of odd order in (1.16), in terms of the Chebyshev polynomials.
The dissertation is organized as follows.
In Chapter 2, we present some basic notations, de…nitions and fundamental properties that we will need for the main results during this dissertation.
In Chapter 3, we present formulas for the entries of positive integer powers of some certain types of tridiagonal matrices, in terms of Chebyshev polynomials of the …rst kind.. Then we give some numerical examples.
In Chapter 4, we present a single formula for the entries of the r th power
(r2 N) of a certain type of circulant matrix of odd and even order, in terms of the
2
PRELIMINARIES
In this chapter, we present preliminary de…nitions and the necessary theorems which will be referred throughout the presentation of the main results of this dissertation.
2.1
Eigenvalues, Eigenvectors and Similarity
De…nition 3 If A 2 Mn and x 2 Cn; we consider the equation
Ax = x; x6= 0;
where is a scalar. If a scalar and a nonzero vector x happen to satisfy this
equation, then is called an eigenvalue of A and x is called an eigenvector of
A associated with (Horn and Johnson,1985).
De…nition 4 The set of all 2 C that are eigenvalues of A 2 Mn is called the
spectrum of A and is denoted by (A) (Horn and Johnson,1985).
Example 5 Consider the matrix
A = 2 4 7 2 4 1 3 5 2 M2:
Then we have 3 2 (A) with 2
4 1
2 3
5 as an associated eigenvector since
A 2 4 1 2 3 5 = 2 4 3 6 3 5 = 3 2 4 1 2 3 5 :
De…nition 6 The characteristic polynomial of A 2 Mn, denoted by pA(t), is the
polynomial de…ned by
pA(t) = det (tI A) ;
where I denotes the identity matrix (Horn and Johnson,1985).
De…nition 7 A matrix B 2 Mn is said to be similar to a matrix A 2 Mn if there
exists a nonsingular matrix S 2 Mn such that
The transformation A ! S 1AS is called a similarity transformation by the
similarty matrix S: The relation "B is similar to A" sometimes abbreviated as
B A (Horn and Johnson,1985).
Theorem 8 Let A; B 2 Mn:If B is similar to A; then the characteristic polynomial
of B is the same as that of A (Horn and Johnson,1985).
Corollary 9 If A; B 2 Mn and if A and B are similar, then they have the same
eigenvalues (Horn and Johnson,1985).
De…nition 10 If the matrix A 2 Mn is similar to a diagonal matrix, then A is said
to be diagonalizable (Horn and Johnson,1985).
Theorem 11 Let A 2 Mn: Then A is diagonalizable if and only if there is set
of n linearly independent vectors, each of which is an eigenvector of A (Horn and Johnson,1985).
Theorem 12 If A 2 Mn has n distinct eigenvalues, then A is diagonalizable (Horn
and Johnson,1985).
Theorem 13 If A 2 Mn is diagonalizable, then for k 2 N
Ak = S kS 1
(Horn and Johnson,1985).
2.2
Chebyshev Polynomials
2.2.1 The …rst-kind polynomial Tn
De…nition 14 The Chebyshev polynomial Tn(x) of the …rst kind is a polynomial in
x of degree n, de…ned by the relation
Tn(x) = cos n when x = cos : (2.1)
If the range of the variable x is the interval [ 1; 1] ; then the range of the
corresponding variable can be taken as [0; ] : These ranges are traversed in
opposite directions, since x = 1 corresponds to = and x = 1 corresponds
It is well known (as a consequence of de Moivre’s Theorem) that cos n is a polynomial of degree n in cos , and indeed we are familiar with the elementary formula
cos 0 = 1; cos 1 = cos ; cos 2 = 2 cos2 1;
cos 3 = 4 cos3 3 cos ; cos 4 = 8 cos4 8 cos2 + 1; : : : :
We may immediately deduce from (2.1), that the …rst few Chebyshev polynomials are
T0(x) = 1; T1(x) = x; T2(x) = 2x2 1;
T3(x) = 4x3 3x; T4(x) = 8x4 8x2+ 1:
In practice it is neither convenient nor e¢ cient to work out each Tn(x)from
…rst principles. Rather by combining the trigonometric identity
cos n + cos (n 2) = 2 cos cos (n 1) ;
with De…nition 14, we obtain the fundamental recurrence relation Tn(x) = 2xTn 1(x) Tn 2(x) ; n = 2; 3; : : : ;
which together with the initial conditions
T0(x) = 1; T1(x) = x;
recursively generates all the polynomials fTn(x)g very e¢ ciently.
Tn(x) can also be satis…ed the determinant equation as the following
Tn(x) = x 1 0 0 1 2x 1 . .. ... 0 1 2x . .. ... ... .. . . .. ... ... 1 0 .. . . .. 1 2x 1 0 0 1 2x
2.2.2 The second-kind polynomial Un
De…nition 15 The Chebyshev polynomial Un(x) of the second kind is a polynomial
of degree n in x de…ned by
Un(x) = sin (n + 1) = sin when x = cos : (2.2)
The ranges of x and are the same as for Tn(x).
sin 1 = sin ; sin 2 = 2 sin cos ; sin 3 = sin 4 cos2 1 ;
sin 4 = sin 8 cos3 4 cos ; : : : ;
so that we see that the ratio of sine functions (2.2) is indeed a polynomial in
cos ; and we may immediately deduce that
U0(x) = 1; U1(x) = 2x; U3(x) = 4x2 1;
U4(x) = 8x3 4x; : : : :
By combining the trigonometric identity
sin (n + 1) + sin (n 1) = 2 cos sin n ;
with De…nition 15, we …nd that Un(x) satis…es the recurrence relation
Un(x) = 2xUn 1(x) Un 2(x) ; n = 2; 3; : : : ;
which together with the initial conditions
U0(x) = 1; U1(x) = 2x;
provides an e¢ cient procedure for generating the polynomials. A similar trigonometric identity
sin (n + 1) sin (n 1) = 2 sin cos ;
leads us to a relationship
Un(x) Un 2(x) = 2Tn(x) ; n = 2; 3; : : : ;
Un(x) can also be satis…ed the determinant equation as the following Un(x) = 2x 1 0 0 1 2x 1 . .. ... 0 1 2x . .. ... ... .. . . .. ... ... 1 0 .. . . .. 1 2x 1 0 0 1 2x
(Mason and Handscomb, 2003).
2.3
Homogeneous Linear Systems
De…nition 16 A system of linear equations is said to be homogeneous if the
constant terms are all zero; that is, the system has the form a11x1+ a12x2+ : : : + a1nxn= 0
a21x1+ a22x2+ : : : + a2nxn= 0
.. .
am1x1+ am2x2+ : : : + amnxn= 0:
Every homogeneous system of linear equations is consistent, since all such systems
have x1 = 0; x2 = 0; : : : ; xn = 0 as a solution. This solution is called the trivial
solution; if there are other solutions, they are called nontrivial solutions.
Because a homogeneous linear system always has the trivial solution, there are only two possibilities for its solutions:
i. The system has only the trivial solution,
ii. The system has in…nitely many solutions in addition to the trivial solution (Anton and Rosses, 2005).
2.4
Jordan Blocks and Jordan Form
A Jordan Block of size m and value is a matrix Jm( )having the value repeated
For example: J2(4) = 2 4 4 1 0 4 3 5 ; J4( 2) = 2 6 6 6 6 6 6 4 2 1 0 0 0 2 1 0 0 0 2 1 0 0 0 2 3 7 7 7 7 7 7 5 :
A Jordan Form is a block diagonal matrix consisting of several Jordan blocks.
For example, we can form a Jordan Form from two copies of J2(4) and one copy of
J4( 2) as follows. J = 2 6 6 6 4 J2(4) 0 0 0 J2(4) 0 0 0 J4( 2) 3 7 7 7 5= 2 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 4 4 1 0 0 0 0 0 0 0 4 0 0 0 0 0 0 0 0 4 1 0 0 0 0 0 0 0 4 0 0 0 0 0 0 0 0 2 1 0 0 0 0 0 0 0 2 1 0 0 0 0 0 0 0 2 1 0 0 0 0 0 0 0 2 3 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 5 (Klavins, 2008).
3
ON COMPUTING OF POSITIVE INTEGER POWERS
FOR
SOME
CERTAIN
TYPES
OF
COMPLEX
TRIDIAGONAL MATRICES
There are three sections in this chapter. In the …rst section, we consider one type of complex tridiagonal matrix. Then, in the second section we consider another type of complex tridiagonal matrix with eigenvalues on imaginary axis. In the last section, we consider a matrix family that is the general type of the matrices given in Section 1 and Section 2. We study on powers of these matrices. Hence, we …rstly give theoretical background required for …nding the eigenvalues and eigenvectors of these matrices. Secondly we obtain the eigenvalues and eigenvectors of these matrices. Based on the eigenvalues and eigenvectors we introduce the formulas that provides a general expression for the entries of the r th power (r 2 N) of these matrices, depending on the Chebyshev polynomials of the …rst kind. Finally we present some numerical examples related to the formulas.
3.1
Positive Integer Powers of One Type of Complex
Tridiagonal Matrices
In this section, we study the entries of positive integer powers of an n n complex
tridiagonal matrix having the form
Xn:= 2 6 6 6 6 6 6 6 6 6 6 6 6 4 a 2b 0 0 b a b . .. ... 0 b a . .. ... ... .. . . .. ... ... b 0 .. . . .. b a b 0 0 2b a 3 7 7 7 7 7 7 7 7 7 7 7 7 5 ; (3.1)
where b 6= 0: This matrix is a more general type of Un given by (1.10). For the
solution of this problem we shall apply the similarity transformation
Xnr = NnJnrNn1; where Jn is the Jordan’s form of Xn and Nn is the
transforming matrix. Since Matrices Jn and Nn can be found by providing
3.1.1 Eigenvalues and eigenvectors of Xn
We …rstly give some preliminaries required for …nding the eigenvalues and eigenvectors of Xn:
Lemma 17 Let fHn; n = 1; 2; : : :g be sequence of tridiagonal matrices of the form
Hn= 2 6 6 6 6 6 6 6 6 6 6 6 6 4 h1;1 h1;2 0 0 h2;1 h2;2 h2;3 . .. ... 0 h3;2 h3;3 . .. . .. ... .. . . .. ... ... . .. 0 .. . . .. ... . .. hn 1;n 0 0 hn;n 1 hn;n 3 7 7 7 7 7 7 7 7 7 7 7 7 5 :
Then the succesive determinants of Hn are given by the recursive formula:
jH1j = h1;1;
jH2j = h1;1h2;2 h1;2h2;1;
jHnj = hn;njH(n 1)j hn 1;nhn;n 1jH(n 2)j
(Cahill et al, 2003).
Let Hy
n; n = 1; 2; : : : be sequence of tridiagonal matrices of the form
Hny = 2 6 6 6 6 6 6 6 6 6 6 6 6 4 h1;1 h1;2 0 0 h2;1 h2;2 h2;3 . .. ... 0 h3;2 h3;3 . .. . .. ... .. . . .. . .. . .. . .. 0 .. . . .. . .. . .. hn 1;n 0 0 hn;n 1 hn;n 3 7 7 7 7 7 7 7 7 7 7 7 7 5 :
Since the determinant of matrices Hn and Hny have the same recursive formula, it
can be written that
jHnj = Hny (3.2)
Let Un be the n n tridiagonal matrix given by (1.10). By using (3.2), we
can write the characteristic polynomial of Un as the following:
jtIn Unj = t 2 0 0 1 t 1 . .. ... 0 1 t . .. ... ... .. . . .. ... ... 1 0 .. . . .. 1 t 1 0 0 2 t : (3.3)
The eigenvalues of Un are
tk = 2 cos
(k 1)
n 1 ; for k = 1; 2; : : : ; n; (3.4)
where tk denotes k th eigenvalue of Un (Rimas, 2005a):
Lemma 18 Let Qn be the following n n tridiagonal matrix
Qn= 2 6 6 6 6 6 6 6 6 6 6 6 6 4 a 2 0 0 1 a 1 . .. ... 0 1 a . .. ... ... .. . . .. ... ... 1 0 .. . . .. 1 a 1 0 0 2 a 3 7 7 7 7 7 7 7 7 7 7 7 7 5 ; (3.5)
where a 2 C: Then the eigenvalues of Qn are
k = a + 2 cos
(k 1)
n 1 ; for k = 1; 2; : : : ; n (3.6)
(Ötele¸s and Akbulak, 2013).
Proof. The eigenvalues of Qn are the roots of its characteristic polynomial. From
(3.2) we can write the characteristic polynomial of Qn as the following:
j In Qnj = a 2 0 0 1 a 1 . .. ... 0 1 a . .. . .. ... .. . . .. . .. . .. 1 0 .. . . .. 1 a 1 0 0 2 a :
Substituting t = aand taking (3.3) and (3.4) into account, we …nd the eigenvalues
of Qn as (3.6). So, the lemma is proved.
Theorem 19 Consider a; b 2 C and b 6= 0. Let Xn be the n n complex tridiagonal
matrix in (3.1). Then the eigenvalues of Xn are
k = a + 2b cos
(k 1)
n 1 ; for k = 1; 2; : : : ; n (3.7)
and the eigenvectors of Xn with associated with eigenvalues are given by
&j = 2 6 6 6 6 6 6 6 4 T0 j a 2b T1 j a 2b .. . Tn 1 j a 2b 3 7 7 7 7 7 7 7 5 ; for j = 1; 2; : : : ; n; (3.8)
where Tk(x) is the kth degree Chebyshev polynomial of the …rst kind such that
Tk(x) = cos k (arccos x) ; 1 x 1 (3.9)
(Ötele¸s and Akbulak, 2013).
Proof. In order to prove the theorem we need a relation between the matrices Xn
and Qn given by (3.5). Dividing all entries of Xn by nonzero b; we get a new matrix
Mn with Mn= 2 6 6 6 6 6 6 6 6 6 6 6 6 4 a=b 2 0 0 1 a=b 1 . .. ... 0 1 a=b . .. ... ... .. . . .. ... ... 1 0 .. . . .. 1 a=b 1 0 0 2 a=b 3 7 7 7 7 7 7 7 7 7 7 7 7 5 :
Taking (3.5) and (3.6) into account we …nd the eigenvalues of Mn to be
a
b + 2 cos
(k 1)
n 1 ; for k = 1; 2; : : : ; n:
Since the eigenvalues of Xn are just b times the eigenvalues of Mn; we get the
Let us …nd the eigenvectors corresponding to each eigenvalue of Xn: Each
eigenvector of Xnis the solution of the following homogeneous linear equation system
( jI Xn) x = 0; (3.10)
where j is the jth eigenvalue of Xn (1 j n) : We clearly write the expression
(3.10) as follows: ( j a) x1 2bx2 = 0 bx1+ ( j a) x2 bx3 = 0 bx2+ ( j a) x3 bx4 = 0 .. . bxn 2+ ( j a) xn 1 bxn= 0 2bxn 1+ ( j a) xn= 0 : (3.11)
Dividing all terms of the each equations in (3.11) by b 6= 0, choosing x1 = 1arbitrarily
and solving the set of systems (3.11) according to x1; we get (3.8). So the theorem
is proved.
Let us begin to get the formula for Xr
n.
3.1.2 Derivation of the formula for the entries of Xnr
Theorem 20 Consider a; b 2 C; b 6= 0. Let Xn be an n n complex tridiagonal
matrix in (3.1). Then (i,j)th entry of Xr
n is [Xnr]i;j = j 2n 2 n X k=1 k r kTi 1 k a 2b Tj 1 k a 2b ; (3.12) for i; j = 1; 2; :::; n: Here j = 8 < : 1; if j = 1; n; 2; if 1 < j < n: and k = 8 < : 1; if k = 1; n; 2; if 1 < k < n; (3.13)
k and Tk(x) are de…ned by (3.7) and (3.9), respectively (Ötele¸s and Akbulak, 2013).
Proof. We shall consider the relation Xr
n = NnJnrNn1. In order to get the formula
for the entries of Xr
n; we …rstly …nd the matrices Jn and Nn:
Since all the eigenvalues k (k = 1; 2; : : : ; n) are simple, each eigenvalue k
corresponds single Jordan cell Ji( k)in Jn:Taking this into account we write down
the Jordan’s form of Xn
Let us …nd the transforming matrix Nn and its inverse Nn1:
From (3.8) we get the transforming matrix Nn as:
Nn= 2 6 6 6 6 6 6 6 4 T0 12ba T0 22ba : : : T0 n2ba T1 12ba T1 22ba : : : T1 n2ba .. . ... . .. ... Tn 1 12ba Tn 1 22ba : : : Tn 1 n2ba 3 7 7 7 7 7 7 7 5 : (3.15)
Denoting the j th column of the inverse matrix N 1
n by j (Nn1 = ( 1; 2; : : : ; n)) ;
from (Rimas, 2005b) we get
j = j 2 6 6 6 6 6 6 6 4 f1Tj 1 12ba f2Tj 1 22ba .. . fnTj 1 n2ba 3 7 7 7 7 7 7 7 5 for j = 1; 2; : : : ; n; (3.16)
where fk = 2n 2k and j; k are de…ned by (3.13). Taking into account (3.16), we
write down N 1 n as Nn1 = 1 2n 2 2 6 6 6 6 6 6 6 6 6 4 T0 12ba 2T1 12ba 2Tn 2 12ba Tn 1 12ba 2T0 22ba 4T1 22ba 4Tn 2 22ba 2Tn 1 22ba .. . ... . .. ... ... 2T0 n2b1 a 4T1 n2b1 a 4Tn 2 n2b1 a 2Tn 1 n2b1 a T0 n2ba 2T1 n2ba 2Tn 2 n2ba Tn 1 n2ba 3 7 7 7 7 7 7 7 7 7 5 : (3.17)
Denoting i th row of Nn by Li (i = 1; 2; ::; n) ;from (3.15) we have
LTi = 2 6 6 6 6 6 6 6 4 Ti 1 12ba Ti 1 22ba .. . Ti 1 n2ba 3 7 7 7 7 7 7 7 5 ;
where T means operation of transposition. Using the equalities Xnr = NnJnrNn1 (see
(3.14), (3.15), (3.16) and (3.17)) and Jr n = diag ( r 1; r 2; : : : ; r
entry of Xr n by [Xnr]i;j; we get [Xnr]i;j = LiJn jr = Li 2 6 6 6 6 6 6 6 4 jf1 r1Tj 1 12ba jf2 r2Tj 1 22ba .. . jfn rnTj 1 n2ba 3 7 7 7 7 7 7 7 5 ; Consequently, [Xnr]i;j = j 2n 2 n X k=1 k r kTi 1 k a 2b Tj 1 k a 2b ; for i; j = 1; 2; :::; n. It is desirable. 3.1.3 Numerical Considerations
In this section, we give two examples related to the formula of Xnr. One of them is
4 4real tridiagonal matrix and the other is 3 3 complex tridiagonal matrix. We
calculate 3th and 4th powers of these matrices, respectively. These examples can be veri…ed by using Maple 13 procedure given in Appendix 1.
Example 21 Let X4 be a 4 4 real tridiagonal matrix given in (3.1) as in the
following: X4 = 2 6 6 6 6 6 6 4 2 6 0 0 3 2 3 0 0 3 2 3 0 0 6 2 3 7 7 7 7 7 7 5
for a = 2 and b = 3: 3th power of X4 is computed as in the following.
From (3.7) eigenvalues of X4 can be written for k = 1; 2; 3; 4 as:
k= 2 + 6 cos
(k 1)
3 ;
matrix N4; whose columns consist of eigenvectors of X4; and its inverse as: N4 = 2 6 6 6 6 6 6 6 6 4 T0 16 2 T0 26 2 T0 36 2 T0 46 2 T1 16 2 T1 26 2 T1 36 2 T1 46 2 T2 16 2 T2 26 2 T2 36 2 T2 46 2 T3 16 2 T3 26 2 T3 36 2 T3 46 2 3 7 7 7 7 7 7 7 7 5 = 2 6 6 6 6 6 6 4 1 1 1 1 1 12 12 1 1 1 2 1 2 1 1 1 1 1 3 7 7 7 7 7 7 5 and N4 1 = 1 6 2 6 6 6 6 6 6 6 6 4 T0 16 2 2T1 162 2T2 162 T3 16 2 2T0 262 4T1 262 4T2 262 2T3 262 2T0 362 4T1 362 4T2 362 2T3 362 T0 46 2 2T1 462 2T2 462 T3 46 2 3 7 7 7 7 7 7 7 7 5 = 2 6 6 6 6 6 6 6 6 4 1 6 1 3 1 3 1 6 1 3 1 3 1 3 1 3 1 3 1 3 1 3 1 3 1 6 1 3 1 3 1 6 3 7 7 7 7 7 7 7 7 5 : Then we get X43 = N4J43N4 1 = 2 6 6 6 6 6 6 4 116 234 108 54 117 170 171 54 54 171 170 117 54 108 234 116 3 7 7 7 7 7 7 5 :
(i; j)th entry of the X43 can be veri…ed by formula given in (3.12).
Example 22 Let X3 be a 3 3 complex tridiagonal matrix given in (3.1) as in the
following: X3 = 2 6 6 6 4 1 i 10 + 4i 0 5 + 2i 1 i 5 + 2i 0 10 + 4i 1 i 3 7 7 7 5:
for a = 1 i and b = 5 + 2i: 4th power of X3 is computed as in the following.
From (3.7) eigenvalues of X3 can be written for k = 1; 2; 3 as:
k = (1 i) + (10 + 4i) cos
(k 1)
2 ;
namely, 1 = 11 + 3i; 2 = 1 iand 3 = 9 5i. We also write the transforming
matrix N3; whose columns consist of eigenvectors of the X3; and its inverse as:
N3 = 2 6 6 6 6 6 4
T0 110+4i(1 i) T0 210+4i(1 i) T0 310+4i(1 i)
T1 110+4i(1 i) T1 210+4i(1 i) T1 310+4i(1 i)
T2 1 (1 i) 10+4i T2 2 (1 i) 10+4i T2 3 (1 i) 10+4i 3 7 7 7 7 7 5 = 2 6 6 6 4 1 1 1 1 0 1 1 1 1 3 7 7 7 5 and N3 1 = 1 4 2 6 6 6 6 6 4 T0 1 (1 i) 10+4i 2T1 1 (1 i) 10+4i T2 1 (1 i) 10+4i
2T0 210+4i(1 i) 4T1 210+4i(1 i) 2T2 210+4i(1 i)
T0 310+4i(1 i) 2T1 310+4i(1 i) T2 310+4i(1 i)
3 7 7 7 7 7 5 = 2 6 6 6 6 6 4 1 4 1 2 1 4 1 2 0 1 2 1 4 1 2 1 4 3 7 7 7 7 7 5 : Then we get X34 = N3J34N 1 3 = 2 6 6 6 4
804 + 6216i 6576 + 2352i 808 + 6216i
3288 + 1176i 1612 + 12432i 3288 + 1176i
808 + 6216i 6576 + 2352i 804 + 6216i
3 7 7 7 5: (i; j)th entry of the X4
3.2
Positive Integer Powers of One Type of Complex
Tridiagonal Matrices with Eigenvalues on Imaginary Axis
In this section, we study the entries of positive integer powers of an n n complex
tridiagonal matrix having the form
Yn := 2 6 6 6 6 6 6 6 6 6 6 6 6 4 a 2b 0 0 b a b . .. ... 0 b a . .. ... ... .. . . .. ... ... b 0 .. . . .. b a b 0 0 2b a 3 7 7 7 7 7 7 7 7 7 7 7 7 5 ; (3.18)
where b 6= 0: This matrix is a more general type of Vn given by (1.11). For the
solution of this problem we shall apply the similarity transformation
Yr
n = NnJnrNn1; where Jn is the Jordan’s form of Yn and Nn is the
transforming matrix. Since Matrices Jn and Nn can be found by providing
eigenvalues and eigenvectors of Yn, we …nd eigenvalues and eigenvectors of Yn.
3.2.1 Eigenvalues and Eigenvectors of Yn
We …rstly give some preliminaries required for …nding the eigenvalues and eigenvectors of Yn:
Let+Hn and Hn be two n n tridiagonal matrices as the following:
+Hn= 2 6 6 6 6 6 6 6 6 6 6 6 6 4 h1;1 h1;2 0 0 h2;1 h2;2 h2;3 . .. ... 0 h3;2 h3;3 . .. . .. ... .. . . .. . .. ... . .. 0 .. . . .. ... . .. hn 1;n 0 0 hn;n 1 hn;n 3 7 7 7 7 7 7 7 7 7 7 7 7 5 ;
and Hn= 2 6 6 6 6 6 6 6 6 6 6 6 6 4 h1;1 h1;2 0 0 h2;1 h2;2 h2;3 . .. ... 0 h3;2 h3;3 . .. . .. ... .. . . .. . .. . .. . .. 0 .. . . .. . .. . .. hn 1;n 0 0 hn;n 1 hn;n 3 7 7 7 7 7 7 7 7 7 7 7 7 5 :
From Lemma 17 since the matrices+Hn and Hn have the same recursive formula,
it can be written that
j+Hnj = j Hnj (3.19)
(Ötele¸s and Akbulak, 2014b).
Let Vn be the n n tridiagonal matrix given by (1.11). By using (3.19), we
can write the characteristic polynomial of Vn as the following:
jtIn Vnj = t 2 0 0 1 t 1 . .. ... 0 1 t . .. ... ... .. . . .. ... ... 1 0 .. . . .. 1 t 1 0 0 2 t : (3.20)
The eigenvalues of Vn are
tk= 2i cos
(k 1)
n 1 ; for k = 1; 2; : : : ; n (3.21)
where tk denotes k th eigenvalue of Vn and i :=
p
1 (Rimas, 2007a).
Lemma 23 Consider a 2 C and i = p 1: Let Rn be the following n n complex
tridiagonal matrix Rn= 2 6 6 6 6 6 6 6 6 6 6 6 6 4 a 2 0 0 1 a 1 . .. ... 0 1 a . .. ... ... .. . . .. ... ... 1 0 .. . . .. 1 a 1 0 0 2 a 3 7 7 7 7 7 7 7 7 7 7 7 7 5 ; (3.22)
where a 2 C: Then the eigenvalues of Rn are
k = a 2i cos
(k 1)
n 1 ; for k = 1; 2; : : : ; n (3.23)
(Ötele¸s and Akbulak, 2014b).
Proof. Since the eigenvalues of Rn are the roots of its characteristic polynomial,
we can write the characteristic polynomial of Rn by using (3.19) as the following:
j I Rnj = a 2 0 0 1 a 1 . .. ... 0 1 a . .. . .. ... .. . . .. . .. . .. 1 0 .. . . .. 1 a 1 0 0 2 a :
Substituting t = a and taking (3.20) and (3.21) into account, we …nd the
eigenvalues of Rn as (3.23). So the theorem is proved.
Theorem 24 Consider a; b 2 C ; b 6= 0; and i =p 1. Let Yn be the n n complex
tridiagonal matrix given by (3.18) and j = ( j a) =b. Then the eigenvalues of Yn
are
k = a 2ib cos
(k 1)
n 1 ; for k = 1; 2; : : : ; n; (3.24)
and the eigenvectors matrix of Yn with associated with eigenvalues are given by,
&j = 2 6 6 6 6 6 6 6 4 e (0) T0 i2j e (1) T1 i j 2 .. . e (n 1) Tn 1 i j 2 3 7 7 7 7 7 7 7 5 ; j = 1; 2; : : : ; n; (3.25) where e (k) = e ikn for k = 0; 1; 2; 3; : : : ; (3.26)
and Tk(x) is the kth degree Chebyshev polynomial of the …rst kind such that
Tk(x) = cos k (arccos x) ; 1 x 1 (3.27)
Proof. In order to prove the theorem, we need a relation between the matrices Yn
and Rn given by (3.22): Dividing all entries of Yn by nonzero b; we get a new matrix
+Mn with +Mn = 2 6 6 6 6 6 6 6 6 6 6 6 6 4 a=b 2 0 0 1 a=b 1 . .. ... 0 1 a=b . .. ... ... .. . . .. ... ... 1 0 .. . . .. 1 a=b 1 0 0 2 a=b 3 7 7 7 7 7 7 7 7 7 7 7 7 5 :
Taking (3.22) and (3.23) into account, we …nd the eigenvalues of+Mn to be
a
b 2i cos
(k 1)
n 1 ; for k = 1; 2; : : : ; n:
Since the eigenvalues of Ynare just b times the eigenvalues of+Mn; we get eigenvalues
of Yn as in (3.24)
k= a 2ib cos
(k 1)
n 1 ; for k = 1; 2; : : : ; n;
where k denotes kth eigenvalue of Yn:
Let us …nd the eigenvectors corresponding to each eigenvalue of Yn: Each
eigenvector of Ynis the solution of the following homogeneous linear equation system
( jI Yn) x = 0; (3.28)
where j is the jth eigenvalue of Yn (1 j n) : We clearly write the expression
(3.28) as follows: ( j a) x1 2bx2 = 0 bx1+ ( j a) x2 bx3 = 0 bx2+ ( j a) x3 bx4 = 0 .. . bxn 2+ ( j a) xn 1 bxn = 0 2bxn 1+ ( j a) xn= 0 : (3.29)
Dividing all terms of the each equations in (3.29) by b 6= 0, substituting
j = ( j a) =b, choosing x1 = 1 arbitrarily and solving the set of systems (3.29)
according to x1, we get (3.25). So the theorem is proved.
3.2.2 Derivation of the formula for the entries of Yr n
Theorem 25 Consider a; b 2 C; b 6= 0; and i =p 1. Let Yn be the n n complex
tridiagonal matrix given by (3.18). Then (i,j)th entry of Yr
n is
[Ynr]i;j = je(i 1)e(j 1)
2n 2 n X k=1 k r kTi 1 i( k a) 2b Tj 1 i( k a) 2b ; (3.30) for i; j = 1; 2; :::; n: Here j = 8 < : 1; if j = 1; n; 2; if 1 < j < n: and k = 8 < : 1; if k = 1; n; 2; if 1 < k < n: (1)
k; e (k)and Tk(x) are de…ned by (3.24), (3.26) and (3.27), respectively (Ötele¸s and
Akbulak, 2014b).
Proof. We shall consider the relation Ynr = NnJnrNn1. In order to get the formula
for the entries of Yr
n; we …rstly …nd the matrices Jn and Nn:
Since all the eigenvalues k (k = 1; 2; : : : ; n) are simple, each eigenvalue k
corresponds single Jordan cell Ji( k)in Jn:Taking this into account we write down
the Jordan’s form of Yn
Jn= diag ( 1; 2; : : : ; n) : (3.31)
Let us …nd the transforming matrix Nn and its inverse Nn1:
From (3.25) we get the transforming matrix Nn as:
Nn= 2 6 6 6 6 6 6 6 4 e (0) T0 i21 e (0) T0 i22 : : : e (0) T0 i2n e (1) T1 i21 e (1) T1 i22 : : : e (1) T1 i2n .. . ... . .. ... e (n 1) Tn 1 i21 e (n 1) Tn 1 i22 : : : e (n 1) Tn 1 i2n 3 7 7 7 7 7 7 7 5 : (3.32)
Denoting the j th column of the inverse matrix N 1
n by j (Nn1 = ( 1; 2; : : : ; n)) ;
from (Rimas, 2007b) we get
j = j 2 6 6 6 6 6 6 6 4 f1e (j 1) Tj 1 i21 f2e (j 1) Tj 1 i22 .. . fne (j 1) Tj 1 i2n 3 7 7 7 7 7 7 7 5 ; j = 1; 2; : : : ; n; (3.33)
where e (j 1) is the complex conjugate of the complex number e (j 1),
fk = 2n 2k and j; k are de…ned by (1): Taking (3.33) into account, we write
down N 1 n as Nn1 = 2 6 6 6 6 6 6 6 6 4 1f1e (0) T0 i21 2f1e (1) T1 i21 nf1e (n 1) Tn 1 i21 1f2e (0) T0 i22 2f2e (1) T1 i22 nf2e (n 1) Tn 1 i22 .. . ... . .. ... 1fne (0) T0 i2n 2fne (1) T1 i2n nfne (n 1) Tn 1 i2n 3 7 7 7 7 7 7 7 7 5 : (3.34)
Denoting i th row of Nn by Li (i = 1; 2; ::; n) ;from (3.32) we have
LTi = 2 6 6 6 6 6 6 6 4 e(i 1)Ti 1 i21 e(i 1)Ti 1 i22 .. . e(i 1)Ti 1 i2n 3 7 7 7 7 7 7 7 5 ;
where T means operation of transposition. Using the equalities Ynr = NnJnrNn1 (see
(3.31), (3.32), (3.33) and (3.34)) and Jr n = diag ( r 1; r 2; : : : ; r
n) and denoting (i,j)th
entry of Ynr by [Ynr]i;j; we get
[Ynr]i;j = LiJn jr = Li 2 6 6 6 6 6 6 6 4 jf1e (j 1) r1Tj 1 i21 jf2e (j 1) r2Tj 1 i22 .. . jfne (j 1) rnTj 1 i2n 3 7 7 7 7 7 7 7 5 : Consequently,
[Ynr]i;j = je(i 1)e(j 1) n X k=1 fk( k)rTi 1 i k 2 Tj 1 i k 2 ; for i; j = 1; 2; :::; n.
Substituting k= ( k a)=b and fk= 2n 2k ; we get
[Ynr]i;j = je(i 1)e(j 1)
2n 2 n X k=1 k r kTi 1 i( k a) 2b Tj 1 i( k a) 2b ; for i; j = 1; 2; :::; n. It is desirable.
3.2.3 Numerical Considerations
In this section, we give two examples related to the formula of Yr
n. One of them is
4 4real tridiagonal matrix and the other is 3 3 complex tridiagonal matrix. We
calculate 3th and 2th powers of these matrices, respectively. These examples can be veri…ed by using Maple 13 procedure given in Appendix 2.
Example 26 Let Y4 be a 4 4 real tridiagonal matrix given in (3.18) as in the
following: Y4 = 2 6 6 6 6 6 6 4 2 6 0 0 3 2 3 0 0 3 2 3 0 0 6 2 3 7 7 7 7 7 7 5 :
for a = 2 and b = 3: 3th power of Y4 is computed as in the following.
From (3.24) eigenvalues of Y4 can be written for k = 1; 2; 3; 4 as:
k = 2 6i cos
(k 1)
3 ;
namely, 1 = 2 6i; 2 = 2 3i; 3 = 2 + 3i and 4 = 2 + 6i. We also write the
transforming matrix N4;whose columns consist of eigenvectors of Y4;and its inverse
as: N4 = 2 6 6 6 6 6 6 6 6 6 4 e (0) T0 i( 16 2) e (0) T0 i( 26 2) e (0) T0 i( 36 2) e (0) T0 i( 46 2) e (1) T1 i( 16 2) e (1) T1 i( 26 2) e (1) T1 i( 36 2) e (1) T1 i( 46 2) e (2) T2 i( 16 2) e (2) T2 i( 26 2) e (2) T2 i( 36 2) e (2) T2 i( 46 2) e (3) T3 i( 1 2) 6 e (3) T3 i( 2 2) 6 e (3) T3 i( 3 2) 6 e (3) T3 i( 4 2) 6 3 7 7 7 7 7 7 7 7 7 5 = 2 6 6 6 6 6 6 6 4 1 1 1 1 i 2i 2i i 1 12 12 1 i i i i 3 7 7 7 7 7 7 7 5
and N4 1 = 2 6 6 6 6 6 6 6 4 1f1e (0) T0 i( 1 2) 6 2f1e (1) T1 i( 1 2) 6 3f1e (2) T2 i( 1 2) 6 1f2e (0) T0 i( 26 2) 2f2e (1) T1 i( 26 2) 3f2e (2) T2 i( 26 2) 1f3e (0) T0 i( 36 2) 2f3e (1) T1 i( 36 2) 3f3e (2) T2 i( 36 2) 1f4e (0) T0 i( 46 2) 2f4e (1) T1 i( 46 2) 3f4e (2) T2 i( 46 2) 4f1e (3) T3 i( 16 2) 4f2e (3) T3 i( 2 2) 6 4f3e (3) T3 i( 36 2) 4f4e (3) T3 i( 46 2) 3 7 7 7 7 7 7 7 7 7 5 = 1 6 2 6 6 6 6 6 6 4 1 2i 2 i 2 2i 2 2i 2 2i 2 2i 1 2i 2 i 3 7 7 7 7 7 7 5 ; Then we get Y43 = N1J13N 1 1 = 2 6 6 6 6 6 6 4 100 90 108 54 45 154 99 54 54 99 154 45 54 108 90 100 3 7 7 7 7 7 7 5 :
(i; j)th entry of the Y3
4 can be veri…ed by formula given in (3.30).
Example 27 Let Y3 be a 3 3 complex tridiagonal matrix given in (3.18) as in the
following: Y3 = 2 6 6 6 4 1 i 10 + 4i 0 5 2i 1 i 5 + 2i 0 10 4i 1 i 3 7 7 7 5:
for a = 1 i and b = 5 + 2i: 2th power of Y3 is computed as in the following.
From (3.24) eigenvalues of Y3 can be written for k = 1; 2; 3 as:
k= (1 i) + (4 10i) cos
(k 1)
namely, 1 = 5 11i; 2 = 1 iand 3 = 3 + 9i. We also write the transforming
matrix N3; whose columns consist of eigenvectors of the Y3; and its inverse as:
N3 = 2 6 6 6 4 e (0) T0 i( 16 2) e (0) T0 i( 26 2) e (0) T0 i( 36 2) e (1) T1 i( 1 2) 6 e (1) T1 i( 2 2) 6 e (1) T1 i( 3 2) 6 e (2) T2 i( 16 2) e (2) T2 i( 26 2) e (2) T2 i( 36 2) 3 7 7 7 5 = 2 6 6 6 4 1 1 1 i 0 i 1 1 1 3 7 7 7 5; and N3 1 = 2 6 6 6 4 1f1e (0) T0 i( 16 2) 2f1e (1) T1 i( 16 2) 3f1e (2) T2 i( 16 2) 1f2e (0) T0 i( 26 2) 2f2e (1) T1 i( 26 2) 3f2e (2) T2 i( 26 2) 1f3e (0) T0 i( 36 2) 2f3e (1) T1 i( 36 2) 3f3e (2) T2 i( 36 2) 3 7 7 7 5 = 1 4 2 6 6 6 4 1 2i 1 2 0 2 1 2i 1 3 7 7 7 5: Then we get Y32 = N3J32N 1 3 = 2 6 6 6 4
42 42i 28 12i 42 + 40i
14 + 6i 84 82i 14 6i
42 + 40i 28 + 12i 42 42i
3 7 7 7 5: (i; j)th entry of the Y32, can be veri…ed by formula given in (3.30).
3.3
Positive Integer Powers of A Certain Type of Complex
Tridiagonal Matrix Family
In this section, we study the entries of positive integer powers of an n n complex
tridiagonal matrix having the form
Bn= 2 6 6 6 6 6 6 6 6 6 6 6 6 4 a 2b 0 0 c a b . .. ... 0 c a . .. ... ... .. . . .. ... ... b 0 .. . . .. c a b 0 0 2c a 3 7 7 7 7 7 7 7 7 7 7 7 7 5 ; (3.35)
where bc 6= 0: This matrix is a matrix family of the matrices Un; Vn; Xnand Yngiven
by (1.10), (1.11), (3.1) and (3.18) respectively. For the solution of this problem we
shall apply the similarity transformation Br
n= NnJnrNn1; where Jn is the Jordan’s
form of Bn and Nn is the transforming matrix. Since Matrices Jn and Nn can be
found by providing eigenvalues and eigenvectors of Bn, we shall eigenvalues and
eigenvectors of Bn.
3.3.1 Eigenvalues and Eigenvectors of Bn
Theorem 28 Consider a; b; c 2 C; bc 6= 0. Let Bn be the n n complex tridiagonal
matrix given by (3.35). Then the eigenvalues of Bn are
k= a + 2
p
bc cos(k 1)
n 1 ; for k = 1; 2; : : : ; n; (3.36)
and the eigenvectors of Bn associated with eigenvalues are given by
&j = 2 6 6 6 6 6 6 6 4 pc b 0 T0 2jpbca pc b 1 T1 j a 2pbc .. . pc b n 1 Tn 1 j a 2pbc 3 7 7 7 7 7 7 7 5 ; for j = 1; 2; : : : ; n: (3.37)
where Tk(x) is the kth degree Chebyshev polynomial of the …rst kind such that
Tk(x) = cos k (arccos x) ; 1 x 1 (3.38)
(Ötele¸s and Akbulak, 2014a).
Proof. In order to prove the theorem, we need a relation between the matrices Bn
and Qn given by (3.5). Now let us consider the n n tridiagonal matrix fMn as the
following: f Mn= 2 6 6 6 6 6 6 6 6 6 6 6 6 4 a=pbc 2 0 0 1 a=pbc 1 . .. ... 0 1 a=pbc . .. . .. ... .. . . .. . .. . .. 1 0 .. . . .. 1 a=pbc 1 0 0 2 a=pbc 3 7 7 7 7 7 7 7 7 7 7 7 7 5 ;
where bc 6= 0: Taking (3.5) and (3.6) into account, we …nd the eigenvalues of fMnto be a p bc+ 2 cos (k 1) n 1 ; for k = 1; 2; : : : ; n:
Dividing all entries of Bn by
p
bc; we also get a new n n matrix My
n as the following: Mny = 2 6 6 6 6 6 6 6 6 6 6 6 6 4 a=pbc 2b=pbc 0 0 c=pbc a=pbc b=pbc . .. ... 0 c=pbc a=pbc . .. . .. ... .. . . .. . .. . .. b=pbc 0 .. . . .. c=pbc a=pbc b=pbc 0 0 2c=pbc a=pbc 3 7 7 7 7 7 7 7 7 7 7 7 7 5 :
Since the multiplication of subdiagonal entries of the matrices fMn and Mny are
the same, the characteristic polynomials of fMn and Mny have the same recursive
formula from Lemma 17. Therefore, the eigenvalues of these matrices are the same.
Moreover, since the eigenvalues of Bn are just
p
bctimes the eigenvalues of My
n; we
get eigenvalues of Bn as in (3.36).
Let us …nd the eigenvectors corresponding to each eigenvalue of Bn: Each
eigenvector of Bnis the solution of the following homogeneous linear equation system
( jIn Bn) x = 0; (3.39)
where j is the j th eigenvalue of Bn (1 j n) : We clearly write the expression
(3.39) as follows: ( j a) x1 2bx2 = 0 cx1+ ( j a) x2 bx3 = 0 cx2+ ( j a) x3 bx4 = 0 .. . cxn 2+ ( j a) xn 1 bxn= 0 2cxn 1+ ( j a) xn= 0 : (3.40)
Dividing all terms of the each equations in (3.40) bypbc, choosing x1 = 1arbitrarily
and solving the set of systems (3.40) according to x1;we get (3.37). So the theorem
is proved.
3.3.2 Derivation of the formula for the entries of Br n
Theorem 29 Consider a; b; c 2 C; bc 6= 0. Let Bn be an n n complex tridiagonal
matrix given by (3.35). Then (i,j)th entry of Br
n is [Bnr]i;j = j 2n 2 n X k=1 k r k r c b i j Ti 1 k a 2pbc Tj 1 k a 2pbc (3.41) for i; j = 1; 2; :::; n: Here j = 8 < : 1; if j = 1; n; 2; if 1 < j < n: and k = 8 < : 1; if k = 1; n; 2; if 1 < k < n; (3.42)
k and Tk(x) are de…ned by (3.36) and (3.38).
Proof. We shall consider the relation Br
n = NnJnrNn1. In order to get the general
expression for the entries of Bnr; we …rstly …nd the matrices Jn and Nn:
Since all the eigenvalues k (k = 1; 2; : : : ; n) are simple, each eigenvalue k
corresponds single Jordan cell Ji( k)in Jn:Taking this into account we write down
the Jordan’s form of Bn
Jn= diag ( 1; 2; : : : ; n) : (3.43)
Let us …nd the transforming matrix Nn and its inverse Nn1:
From (3.37) we get the transforming matrix Nn as:
Nn= 2 6 6 6 6 6 6 6 4 pc b 0 T0 21pbca pc b 0 T0 22pbca : : : pc b 0 T0 2npbca pc b 1 T1 21pbca pc b 1 T1 22pbca : : : pc b 1 T1 2npbca .. . ... . .. ... pc b n 1 Tn 1 21pbca pc b n 1 Tn 1 22pbca : : : pc b n 1 Tn 1 2npbca 3 7 7 7 7 7 7 7 5 : (3.44)
Denoting the j th column of the inverse matrix N 1
n by j (Nn1 = ( 1; 2; : : : ; n)) ;
from (Rimas, 2005b) we get
j = j 2 6 6 6 6 6 6 6 6 4 f1 q b c j 1 Tj 1 21pbca f2 q b c j 1 Tj 1 22pbca .. . fn q b c j 1 Tj 1 2npbca 3 7 7 7 7 7 7 7 7 5 for j = 1; 2; : : : ; n; (3.45)
where fk = 2n 2k ; and j; k are de…ned by (3.42). Taking (3.45) into account, we write down N 1 n as Nn1 = 1 2n 2 2 6 6 6 6 6 6 6 6 6 6 6 6 6 4 q b c 0 T0 21pbca q b c 1 2T1 21pbca q b c 0 2T0 22pbca q b c 1 4T1 22pbca .. . ... . .. q b c 0 2T0 n2p1bca q b c 1 4T1 n2p1bca q b c 0 T0 2npbca q b c 1 2T1 2npbca q b c n 2 2Tn 2 21pbca q b c n 1 Tn 1 21pbca q b c n 2 4Tn 2 22pbca q b c n 1 2Tn 1 22pbca .. . ... q b c n 2 4Tn 2 n2p1bca q b c n 1 2Tn 1 n2p1bca q b c n 2 2Tn 2 2npbca q b c n 1 Tn 1 2npbca 3 7 7 7 7 7 7 7 7 7 7 7 7 7 5 : (3.46)
Denoting i th row of Nn in (3.44) by Li (i = 1; 2; ::; n) ;we have
LTi = 2 6 6 6 6 6 6 6 4 pc b i 1 Ti 1 21pbca pc b i 1 Ti 1 22pbca .. . pc b i 1 Ti 1 2npbca 3 7 7 7 7 7 7 7 5 ;
where T means operation of transposition. Using the equalities Br
n= NnJnrNn1 (see (3.43), (3.44), (3.45) and (3.46)) and Jr n = diag ( r 1; r 2; : : : ; r
n) and denoting (i,j)th
entry of Br n by [Bnr]i;j; we get [Bnr]i;j = LiJn jr = Li 2 6 6 6 6 6 6 6 6 4 q b c j 1 jf1 r1Tj 1 21pbca q b c j 1 jf2 r2Tj 1 22pbca .. . q b c j 1 jfn rnTj 1 2npbca 3 7 7 7 7 7 7 7 7 5 : Consequently, [Brn]i;j = j 2n 2 n X k=1 k( k) r r c b i j Ti 1 k a 2pbc Tj 1 k a 2pbc ;
for i; j = 1; 2; :::; n. It is desirable.
Now, if we write c = b in Bn given by (3.35), we get Xn given by (3.1). From
Theorem 29 we can directly obtain the expression given in (3.12) for the entries of the powers of Xn.
Corollary 30 Consider a; b; c 2 C; bc 6= 0: Let Bn be an n n complex tridiagonal
matrix given by (3.35). If c = b; then (i,j)th entry of Br
n is [Bnr]i;j = j 2n 2 n X k=1 k r kTi 1 k a 2b Tj 1 k a 2b ;
for i; j = 1; 2; :::; n: Here k; Tk(x) and j; k are de…ned by (3.36), (3.38) and
(3.42), respectively.
If we write c = b in Bn given by (3.35), we get Yn given by (3.18). From
Theorem 29 we can also obtain the expression given in (3.30) for the entries of the powers of Yn:
Corollary 31 Consider a; b; c 2 C; bc 6= 0 and i = p 1. Let Bn be the
n n complex tridiagonal matrix given by (3.35), If c = b; then (i,j)th entry
of Bnr is
[Brn]i;j = je(i 1)e(j 1)
2n 2 n X k=1 k r kTi 1 i( k a) 2b Tj 1 i( k a) 2b ;
for i; j = 1; 2; :::; n: Here e(k); k; Tk(x) and j; k are de…ned by (3.26), (3.36),
(3.38) and (3.42), respectively.
Proof. From Theorem 29 we have
[Bnr]i;j = j 2n 2 n X k=1 k r k(i) i j Ti 1 i( k a) 2b Tj 1 i( k a) 2b ;
for c = b: Since Tk( x) = ( 1)kTk(x) (Mason and Handscomb, 2003), We get
[Bnr]i;j = j( 1) i+j 2 (i)i j 2n 2 n X k=1 k r kTi 1 i( k a) 2b Tj 1 i( k a) 2b :
To conclude the proof we need to show that
Since ( 1)i+j 2 = ( 1)i+j = ( 1)j i = (i)2j 2i; we get ( 1)i+j 2(i)i j = (i)j i:
Since e(i 1)e(j 1) = ei(j i)2and (i)l
= eil2 (Mason and Handscomb, 2003),
( 1)i+j 2(i)i j = (i)j i = ei(j i)2
= e(i 1)e(j 1):
It is desirable.
3.3.3 Numerical Considerations
In this section, we give three examples related to the formula of Bnr. First one is
4 4real tridiagonal matrix, the other is 3 3complex tridiagonal matrix and the
last one is 6 6 real tridiagonal matrix with eigenvalues on imaginary axis. We
calculate 3th, 4th and 5th powers of these matrices, respectively. These examples can be veri…ed by using Maple 13 procedure given in Appendix 3.
Example 32 Let B4 be a 4 4 real tridiagonal matrix (for a = 3; b = 1 and c = 4,
given in (3.35)) as in the following:
B4 = 2 6 6 6 6 6 6 4 3 2 0 0 4 3 1 0 0 4 3 1 0 0 8 3 3 7 7 7 7 7 7 5 :
3th power of B4 is computed as in the following.
From (3.36) the eigenvalues of B4 can be written for k = 1; 2; 3; 4 as:
k = 3 + 4 cos
(k 1)
3 ;
matrix N4; whose columns consist of eigenvectors of B4; and its inverse as: N4 = 2 6 6 6 6 6 6 6 6 4 T0 14 3 T0 243 T0 34 3 T0 44 3 2T1 14 3 2T1 24 3 2T1 34 3 2T1 443 (22) T 2 14 3 (22) T2 24 3 (22) T2 343 (22) T2 44 3 (23) T 3 14 3 (23) T3 24 3 (23) T3 343 (23) T3 44 3 3 7 7 7 7 7 7 7 7 5 = 2 6 6 6 6 6 6 4 1 1 1 1 2 1 1 2 4 2 2 4 8 8 8 8 3 7 7 7 7 7 7 5 ; and N4 1 = 1 6 2 6 6 6 6 6 6 6 6 4 T0 14 3 12 2T1 143 212 2T2 14 3 213 T3 14 3 2T0 24 3 12 4T1 243 212 4T2 24 3 213 2T3 24 3 2T0 34 3 12 4T1 343 212 4T2 34 3 213 2T3 34 3 T0 44 3 12 2T1 443 212 2T2 44 3 213 T3 44 3 3 7 7 7 7 7 7 7 7 5 = 2 6 6 6 6 6 6 6 6 6 4 1 6 1 6 1 12 1 48 1 3 1 6 1 12 1 24 1 3 1 3 1 12 1 24 1 6 1 3 1 12 1 48 3 7 7 7 7 7 7 7 7 7 5 : Then we get B43 = N4J43N 1 4 = 2 6 6 6 6 6 6 4 99 78 18 2 156 135 47 9 144 188 135 39 128 288 312 99 3 7 7 7 7 7 7 5 :
(i,j)th entry of the B43, can be veri…ed by formula given in (3.41).
b = 3 + 4i and c = 3 4i given in (3.35)) as in the following: B3 = 2 6 6 6 4 1 i 6 + 8i 0 3 4i 1 i 3 + 4i 0 6 8i 1 i 3 7 7 7 5:
4th power of B3 is computed as in the following.
From (3.36) the eigenvalues of B3 can be written for k = 1; 2; 3 as:
k= (1 i) + 10 cos
(k 1)
2 ;
namely, 1 = 11 i; 2 = 1 i and 3 = 9 i. We also write the transforming
matrix N3; whose columns consist of eigenvectors of the B3; and its inverse as:
N3 = 2 6 6 6 6 6 4 T0 1 10(1 i) T0 2 10(1 i) T0 3 10(1 i) 3 5 4 5i T1 1 (1 i) 10 3 5 4 5i T1 2 (1 i) 10 3 5 4 5i T1 3 (1 i) 10 3 5 4 5i 2 T2 1 (1 i) 10 3 5 4 5i 2 T2 2 (1 i) 10 3 5 4 5i 2 T2 3 (1 i) 10 3 7 7 7 7 7 5 = 2 6 6 6 6 4 1 1 1 3 5 4 5i 0 3 5 + 4 5i 7 25 24 25i 7 25+ 24 25i 7 25 24 25i 3 7 7 7 7 5 ; and N31 = 2 6 6 6 6 6 4 T0 1 10(1 i) 2 35 +45i T1 1 10(1 i) 35 + 45i 2 T2 1 10(1 i) 2T0 2 (1 i) 10 4 3 5 + 4 5i T1 2 (1 i) 10 2 3 5 + 4 5i 2 T2 2 (1 i) 10 T0 3 10(1 i) 2 35 +45i T1 3 10(1 i) 35 + 45i 2 T2 3 10(1 i) 3 7 7 7 7 7 5 = 1 4 2 6 6 6 6 6 4 1 65 +85i 257 + 2425i 2 0 1425 4825i 1 65 85i 257 + 2425i 3 7 7 7 7 7 5 : Then we get B43 = N3J34N3 1 = 2 6 6 6 4
4996 600i 5616 + 688i 824 + 4968i
456 2792i 9996 1200i 2808 + 344i
1976 4632i 912 5584i 4996 600i
3 7 7 7 5: (i,j)th entry of the B34, can be veri…ed by formula given in (3.41).
Example 34 Let B6 be a 6 6 real tridiagonal matrix (for a = 5; b = 3 and
c = b = 3, given in (3.35)) as in the following:
B6 = 2 6 6 6 6 6 6 6 6 6 6 6 6 4 5 6 0 0 0 0 3 5 3 0 0 0 0 3 5 3 0 0 0 0 3 5 3 0 0 0 0 3 5 3 0 0 0 0 6 5 3 7 7 7 7 7 7 7 7 7 7 7 7 5 :
5th power of B6 is computed as in the following.
From (3.36) the eigenvalues of B6 can be written for k = 1; 6 as:
k = 5 + 6i cos (k 1) 3 ; namely, 1 = 5 + 6i; 2 = 5 + 4:8541i; 3 = 5 + 1:8541i; 4 = 5 1:8541i; 5 = 5 4:8541i; 6 = 5 6i:
We also write the transforming matrix N6;whose columns consist of eigenvectors of
B6; and its inverse as:
N6 = 2 6 6 6 6 6 6 6 6 6 6 6 6 4 1 1 1 1 1 1
i 0:8090i 0:3090i 0:3090i 0:8090i i
1 0:3090 0:8090 0:8090 0:3090 1
i 0:3090i 0:8090i 0:8090i 0:3090i i
1 0:8090 0:3090 0:3090 0:8090 1 i i i i i i 3 7 7 7 7 7 7 7 7 7 7 7 7 5 ;
and N61 = 2 6 6 6 6 6 6 6 6 6 6 6 6 4
0:1 0:2i 0:2 0:2i 0:2 0:1i
0:2 0:3236i 0:1236 0:1236i 0:3236 0:2i
0:2 0:1236i 0:3236 0:3236i 0:1236 0:2i
0:2 0:1236i 0:3236 0:3236i 0:1236 0:2i
0:2 0:3236i 0:1236 0:1236i 0:3236 0:2i
0:1 0:2i 0:2 0:2i 0:2 0:1 3 7 7 7 7 7 7 7 7 7 7 7 7 5 : Then we get B65 = N6J65N 1 6 = 2 6 6 6 6 6 6 6 6 6 6 6 6 4 7225 16890 6300 11070 4050 486 8445 10375 13980 1125 5049 2025 3150 13980 5200 7959 1125 5535 5535 1125 7959 5200 13980 3150 2025 5049 1125 13980 10375 8445 486 4050 11070 6300 16890 7225 3 7 7 7 7 7 7 7 7 7 7 7 7 5 :
(i,j)th entry of the B5
4
ON COMPUTING OF POSITIVE INTEGER POWERS
FOR CERTAIN REAL CIRCULANT MATRIX
In this chapter, we present a single formula for the entries of the r th power (r 2 N) of the n n real circulant matrix An := circn(a0; a1; 0; : : : ; 0; a 1) of odd and even
order, in terms of the Chebyshev polynomials. Namely,
An := 2 6 6 6 6 6 6 6 6 6 6 6 6 4 a0 a1 0 : : : 0 a 1 a 1 a0 a1 . .. 0 0 a 1 a0 . .. ... ... .. . . .. ... ... ... 0 0 . .. ... ... a1 a1 0 : : : 0 a 1 a0 3 7 7 7 7 7 7 7 7 7 7 7 7 5 : (4.1)
As we will see, that formula obtained for the powers of the matrix in (4.1) is an extension of the one obtained for the powers of the matrix in (1.16). The chapter …nishes some numerical examples concerned with the formula.
4.1
General Expression for the Entries of
A
rnThe well-known eigenvalue decomposition of an n ncirculant matrix (Davis, 1994)
is that
circn(c0; c1; : : : ; cn 1) = FnDnFn; (4.2)
where * denotes conjugate transpose (i.e Fn = FTn), Fn is the n n Fourier matrix:
[Fn]u;v = 1 p ne 2 (u 1)(v 1) n i; 1 u; v n;
and Dn= diag ( 1; 2; : : : ; n) with
k = n X r=1 cr 1e 2 (k 1)(r 1) n i; 1 k n: (4.3)
It is also known that Fn is unitary (Davis, 1994).
Let Uk(x) be the k th degree Chebyshev polynomial of the second kind:
Uk(x) =
sin ((k + 1) arccos x)
and Tk(x) is the k th degree Chebyshev polynomial of the …rst kind, with
k 2 N [ f0g :
Tk(x) = cos (k arccos x) ; 1 x 1 (4.5)
(Mason and Handscomb, 2003).
Based on the eigenvalue decomposition (4.2), we can prove Theorem 35 that
provides a general expression for the entries of Ar
n.
Theorem 35 Let An= circn(a0; a1; 0; : : : ; 0; a 1) be an n n real circulant matrix
and k = cos
2 (k 1)
n ; 1 k n; 3 n 2 N: Then (u,v)th entry of A
r
n is given by:
[Arn]u;v = 1
n(S1+ S2) ;
for all r 2 N and 1 u; v n; where S1 and S2 are respectively that
S1 = bn 2c+1 X k=1 a0+ (a 1+ a1) T1( k) + (a 1 a1) sign n 2 + 1 k q 1 2 kilimj !kU0( j) r Tjv uj( k) + sign (v u) sign n 2 + 1 k q 1 2 kilim j!kUjv uj 1( j) ; and S2 = bn+1 2 c X k=2 a0+ (a 1+ a1) T1( k) (a 1 a1) sign n 2 + 1 k q 1 2 kilimj !kU0( j) r Tjv uj( k) sign (v u) sign n 2 + 1 k q 1 2 kiUjv uj 1( k) :
Here bxc is the largest integer less than or equal to x:
Proof. By using (4.2), [Arn]u;v = [(FnDnFn) r ]u;v = [FnDrnFn]u;v = n X k=1 [Fn]u;k[DnrFn]k;v = n X k=1 [Fn]u;k r k[Fn]v;k = 1 n n X k=1 r ke 2 (u 1)(k 1) n ie 2 (v 1)(k 1) n i;
according to above the last equation, we get [Arn]u;v = 1 n n X k=1 r ke 2 (k 1)(v u) n i: (4.6)
From (4.3) and e 2 (k 1)(rn 1) = e
2 (k 1)(n+2 r 1)
n i for all 2 r n (see (Gutiérrez,
2008a), we can write k as
k = a0+ a1e 2 (k 1) n i+ a 1e 2 (k 1) n i = a0+ (a 1+ a1) cos 2 (k 1) n + (a 1 a1) i sin 2 (k 1) n :
Observe that from (4.4) and (4.5), we have
Tm cos 2 (k 1) n = cos 2 (k 1) m n ; and Um 1 cos 2 (k 1) n = sin2 (k 1)mn sin2 (k 1)n : (4.7)
In (4.7), we have the indeterminate form 0=0 for k = 1 and k = n2 + 1: So we can
write k = a0+(a 1+ a1) T1 cos 2 (k 1) n +(a 1 a1) sin 2 (k 1) n ijlim!kU0 cos 2 (j 1) n :
Since k = cos2 (k 1)n and
sin2 (k 1) n = 8 > < > : p 1 2 k if n 2 + 1 k > 0; p 1 2 k if n 2 + 1 k < 0; (4.8) k is obtained as k= a0+ (a 1+ a1) T1( k) + (a 1 a1) sign n 2 + 1 k q 1 2 kilimj !kU0( j) : (4.9)
From (4.9) we can write k = n+2 k for 2 k n: Namely,
Dn = diag 1; 2; : : : ; n+1 2 ; n+1 2 ; : : : ; 2 if n is odd, Dn = diag 1; 2; : : : ; n2; n2+1; n2; : : : ; 2 if n is even. By using (4.6), [Ar
n]u;v can be expressed as
[Arn]u;v = 8 > > > > > > > < > > > > > > > : 1 n 2 4 r 1+ n+1 2 X k=2 r ke 2 (k 1)(v u) n i+ n X k=n+12 +1 r ke 2 (k 1)(v u) n i 3 5 if n is odd, 1 n 2 4 r 1+ n 2+1 X k=2 r ke 2 (k 1)(v u) n i+ n X k=n 2+2 r ke 2 (k 1)(v u) n i 3 5 if n is even,