We consider the skew circulant and skew left circulant matrices with any continuous Lucas numbers. Firstly, we discuss the invertibility of the skew circulant matrices and present the determinant and the inverse matrices by constructing the transformation matrices. Furthermore, the invertibility of the skew left circulant matrices is also discussed. We obtain the determinants and the inverse matrices of the skew left circulant matrices by utilizing the relationship between skew left circulant matrices and skew circulant matrix, respectively. Finally, the four kinds of norms and bounds for the spread of these matrices are given, respectively.
1. Introduction
Circulant and skew-circulant matrices are appearing increasingly often in scientific and engineering applications. Briefly, scanning the recent literature, one can see their utility is appreciated in the design of digital filters [1–3], image processing [4–6], communications [7], signal processing [8], and encoding [9]. They have been put on firm basis with the work of Davis [10] and Jiang and Zhou [11].
The skew circulant matrices as preconditioners for linear multistep formulae- (LMF-) based ordinary differential equations (ODEs) codes. Hermitian and skew-Hermitian Toeplitz systems are considered in [12–15]. Lyness and Sørevik employed a skew circulant matrix to construct s-dimensional lattice rules in [16]. Spectral decompositions of skew circulant and skew left circulant matrices were discussed in [17]. Compared with cyclic convolution algorithm, the skew cyclic convolution algorithm [8] is able to perform filtering procedure in approximate half of computational cost for real signals. In [2] two new normal-form realizations are presented which utilize circulant and skew circulant matrices as their state transition matrices. The well-known second-order coupled form is a special case of the skew circulant form. Li et al. [18] gave the style spectral decomposition of skew circulant matrix firstly and then dealt with the optimal backward perturbation analysis for the linear system with skew circulant coefficient matrix. In [3], a new fast algorithm for optimal design of block digital filters (BDFs) was proposed based on skew circulant matrix.
Besides, some scholars have given various algorithms for the determinants and inverses of nonsingular circulant matrices [10, 11]. Unfortunately, the computational complexity of these algorithms is very amazing with the order of matrix increasing. However, some authors gave the explicit determinants and inverse of circulant and skew circulant involving some famous numbers. For example, Jaiswal evaluated some determinants of circulant whose elements are the generalized Fibonacci numbers [19]. Lind presented the determinants of circulant and skew circulant involving the Fibonacci numbers [20]. Dazheng [21] gave the determinant of the Fibonacci-Lucas quasicyclic matrices. Shen et al. considered circulant matrices with the Fibonacci and Lucas numbers and presented their explicit determinants and inverses by constructing the transformation matrices [22]. Gao et al. [23] gave explicit determinants and inverses of skew circulant and skew left circulant matrices with the Fibonacci and Lucas numbers. Jiang et al. [24, 25] considered the skew circulant and skew left circulant matrices with the k-Fibonacci numbers and the k-Lucas numbers and discussed the invertibility of the these matrices and presented their determinant and the inverse matrix by constructing the transformation matrices, respectively.
Recently, there are several papers on the norms of some special matrices. Solak [26] established the lower and upper bounds for the spectral norms of circulant matrices with the classical Fibonacci and Lucas numbers entries. İpek [27] investigated an improved estimation for spectral norms of these matrices. Shen and Cen [28] gave upper and lower bounds for the spectral norms of r-circulant matrices in the forms of A=Cr(F0,F1,…,Fn-1), B=Cr(L0,L1,…,Ln-1), and they also obtained some bounds for the spectral norms of Kronecker and Hadamard products of matrix A and matrix B. Akbulak and Bozkurt [29] found upper and lower bounds for the spectral norms of Toeplitz matrices such that aij≡Fi-j and bij≡Li-j. The convergence in probability and in distribution of the spectral norm of scaled Toeplitz, circulant, reverse circulant, symmetric circulant, and a class of k-circulant matrices is discussed in [30].
Beginning with Mirsky [31], several authors [32–38] have obtained bounds for the spread of a matrix.
The purpose of this paper is to obtain the explicit determinants, explicit inverses, norm, and spread of skew circulant type matrices involving any continuous Lucas numbers. And we generalize the result [23]. In passing, the norm and spread of skew circulant type matrices have not been researched. It is hoped that this paper will help in changing this. More work continuing the present paper is forthcoming.
In the following, let r be a nonnegative integer. We adopt the following two conventions 00=1, and, for any sequence {an}, ∑k=inak=0 in the case i>n.
The Lucas sequences are defined by the following recurrence relations [21–23, 27–29]:
(1)Ln+1=Ln+Ln-1,whereL0=2,L1=1,
for n≥0. The first few values of the sequences are given by the following table:
(2)n0123456789Ln213471118294776.
The {Ln} is given by the formula
(3)Ln=αn+βn,
where α and β are the roots of the characteristic equation x2-x-1=0.
Definition 1 (see [17]).
A skew circulant matrix over C with the first row (a1,a2,…,an) is meant a square matrix of the form
(4)(a1a2…an-1an-ana1a2…an-1⋮-ana1⋱⋮-a3⋮⋱⋱a2-a2-a3…-ana1)n×n,
denoted by SCirc(a1,a2,…,an).
Definition 2 (see [17]).
A skew left circulant matrix over C with the first row (a1,a2,…,an) is meant a square matrix of the form
(5)(a1a2a3⋯ana2a3⋯an-a1a3⋰⋰⋰⋮⋮an-a1⋯-an-2an-a1⋯-an-2-an-1)n×n,
denoted by SLCirc(a1,a2,…,an).
Lemma 3 (see [10, 17]).
Let A=SCirc(a1,a2,…,an) be skew circulant matrix; then
A is invertible if and only if the eigenvalues of A(6)λk=f(ωkη)≠0,(k=0,1,2,…,n-1),
where f(x)=∑j=1najxj-1, ω=exp(2πi/n), and η=exp(πi/n);
if A is invertible, then the inverse of A is a skew circulant matrix.
Lemma 4 (see [17]).
Let A=SLCirc(a1,a2,…,an) be skew left circulant matrix and let n be odd; then
(7)λj=±|∑k=1nakω(j-(1/2))(k-1)|,(j=1,2,…,n-12),λ(n+1)/2=∑k=1n|ak(-1)k-1|,
where λj,j=1,2,…,(n-1)/2, (n+1)/2 are the eigenvalues of A.
Lemma 5 (see [23]).
With the orthogonal skew left circulant matrix
(8)Θ∶=(10⋯0000⋯0-100⋯-10⋮⋮⋱⋮⋮0-1⋯00)n×n,
it holds that
(9)SCirc(a1,a2,…,an)=ΘSLCirc(a1,a2,…,an).
Lemma 6 (see [23]).
If
(10)[SCirc(a1,a2,…,an)]-1=SCirc(b1,b2,…,bn),
then
(11)[SLCirc(a1,a2,…,an)]-1=SLCirc(b1,-bn,…,-b2).
Lemma 7 (see [27, 28]).
Let {Ln} be the Lucas numbers; then
(12)(i)∑i=0n-1Li=Ln+1-1,(13)(ii)∑i=0n-1Li2=LnLn-1+2,(14)(iii)∑i=0n-1iLi=(n-1)Ln+1-Ln+2+4.
Definition 8 (see [29]).
Let A=(aij) be an n×n matrix. The maximum column sum matrix norm, the spectral norm, the Euclidean (or Frobenius) norm, and the maximum row sum matrix norm of matrix A are, respectively,
(15)∥A∥1=max1≤j≤n∑i=1n|aij|,∥A∥2=(max1≤i≤nλi(A*A))1/2,∥A∥F=(∑i,j=1n|aij|2)1/2,∥A∥∞=max1≤i≤n∑j=1n|aij|,
where A* denotes the conjugate transpose of A.
Lemma 9 (see [30]).
If A is an n×n real symmetric or normal matrix, then one has
(16)∥A∥2=max1≤i≤n|λi|,
where λi (i=1,2,…,n) are the eigenvalues of A.
Definition 10 (see [31, 32]).
Let A=(aij) be an n×n matrix with eigenvalues λi, i=1,2,…n. The spread of A is defined as
(17)s(A)=maxi,j|λi-λj|.
Beginning with Mirsky [31], several authors [32–38] have obtained bounds for the spread of a matrix.
Lemma 11.
Let A=(aij) be an n×n matrix. An upper bound for the spread due to Mirsky [31] states that
(18)s(A)⩽2∥A∥F2-2n|trA|2,
where ∥A∥F denotes the Frobenius norm of A and trA is trace of A.
Lemma 12 (see [38]).
Let A=(aij) be an n×n matrix; then
if A is real and normal, then
(19)s(A)≥1n-1|∑i≠jaij|,
and if A is Hermitian, then
(20)s(A)≥2maxi≠j|aij|.
2. Determinant and Inverse of Skew Circulant Matrix with the Lucas Numbers
In this section, let Ar,n=SCirc(Lr+1,…,Lr+n) be skew circulant matrix. Firstly, we give a determinant explicit formula for the matrix Ar,n. Afterwards, we prove that Ar,n is an invertible matrix for n≥2, and then we find the inverse of the matrix Ar,n.
In the following, let x=-((Lr+Lr+n)/(Lr+1+Lr+n+1)), t=Lr+2/Lr+1, c=Lr+1+Lr+n+1, d=Lr+Lr+n, ln=Lr+1+tLr+n+∑k=1n-2(tLr+k+1-Lr+k+2)·xn-(k+1) and ln′=∑k=1n-1Lr+k+1·xn-(k+1).
Theorem 13.
Let Ar,n=SCirc(Lr+1,…,Lr+n) be skew circulant matrix; then
(21)detAr,n=Lr+1[∑i=1n-2Lr+1+tLr+n+∑i=1n-2(tLr+i+1-Lr+i+2)xn-(i+1)]·cn-2,
where Lr+n is the (r+n)th Lucas number. Specially, when r=0, one gets the result of [23].
Proof.
Obviously, detAr,1=Lr+1 satisfies the equation. In the case n>1, let
(22)Σ=(1t111-1001-1-1⋮⋰⋰⋰01⋰⋰01-1⋰001-1-1),Ω1=(100⋯000xn-20⋯010xn-30⋯10⋮⋮⋮⋱⋮⋮0x1⋯00010⋯00)
be two n×n matrices; then we have
(23)ΣAr,nΩ1=(Lr+1ln′c13⋯c1,n-1c1n0lnc23⋯c2,n-1c2n00c00d⋱⋮⋮⋱c00dc),
where
(24)c1j=Lr+n+2-j,c2j=tLr+n+2-j-Lr+n+3-j,iiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiii(j=3,4,…,n).
So it holds that
(25)detΣdetAr,ndetΩ1=Lr+1[∑k=1n-2Lr+1+tLr+n+∑k=1n-2(tLr+k+1-Lr+k+2)xn-(i+1)∑k=1n-2]·(Lr+1+Lr+n+1)n-2.
While taking detΣ=detΩ1=(-1)(n-1)(n-2)/2, we have
(26)detAr,n=Lr+1[∑k=1n-2Lr+1+tLr+n+∑k=1n-2(tLr+k+1-Lr+k+2)xn-(i+1)∑k=1n-2]·(Lr+1+Lr+n+1)n-2.
This completes the proof.
Theorem 14.
Let Ar,n=SCirc(Lr+1,…,Lr+n) be skew circulant matrix; then Ar,n is an invertible matrix. Specially, when r=0, one gets the result of [23].
Proof.
Taking n=2 in, Theorem 13, we have detAr,2=Lr+12+Lr+22≠0. Hence Ar,2 is invertible. In the case n>2, since Lr+n=αr+n+βr+n, where α+β=1, αβ=-1, we have
(27)f(ωkη)=∑j=1nLr+j(ωkη)j-1=∑j=1n(αr+j+βr+j)(ωkη)j-1=αr+1(1+αn)1-αωkη+βr+1(1+βn)1-βωkη=Lr+1+Lr+n+1+(Lr+Lr+n)ωkη1-ωkη-ω2kη2iiiiiiiiiiiiiiiiiiiiii(k=1,2,…,n-1),
where ω=exp(2πi/n), η=exp(πi/n). If there exists ωlη(l=1,2,…,n-1) such that f(ωlη)=0, we obtain Lr+1+Lr+n+1+(Lr+Lr+n)ωlη=0, for 1-ωlη-ω2lη2≠0, and hence it follows that ωlη=-((Lr+1+Lr+n+1)/(Lr+Lr+n)) is a real number. Since
(28)ωlη=exp(2l+1)πin=cos(2l+1)πn+isin(2l+1)πn,
it yields that sin((2l+1)π/n)=0, so we have ωlη=-1 for 0<(2l+1)π/n<2π. Since x=-1 is not the root of the equation Lr+1+Lr+n+1+(Lr+Lr+n)x=0(n>2). We obtain f(ωkη)≠0, for any ωkη(k=1,2,…,n-1), while
(29)f(η)=∑j=1nLjηj-1=Lr+1+Lr+n+1+(Lr+Lr+n)η1-η-η2≠0.
It follows from Lemma 3 that the conclusion holds.
Lemma 15.
Let the matrix ℋ=[hij]i,j=1n-2 be of the form
(30)hij={Lr+1+Lr+n+1=c,i=j,Lr+Lr+n=d,i=j+1,0,otherwise.
Then the inverse ℋ-1=[hi,j′]i,j=1n-2 of the matrix ℋ is equal to
(31)hij′={(-d)i-jci-j+1,i≥j,0,i<j.
Specially, when r=0, one gets the result of [23].
Proof.
Let eij=∑k=1n-2hikhkj′. Obviously, eij=0 for i<j. In the case i=j, we obtain eii=hiihii′=(Lr+1+Lr+n+1)·(1/(Lr+1+Lr+n+1))=1. For i≥j+1, we obtain
(32)eij=∑k=1n-2hikhkj′=hi,i-1hi-1,j′+hiihij′=d·(-d)i-j-1ci-j+c·(-d)i-jci-j+1=0.
Hence, we get ℋℋ-1=In-2, where In-2 is (n-2)×(n-2) identity matrix. Similarly, we can verify that ℋ-1ℋ=In-2. Thus, the proof is completed.
Theorem 16.
Let Ar,n=SCirc(Lr+1,…,Lr+n) be skew circulant matrix; then
(33)(Ar,n)-1=1ln·SCirc(y1′,y2′,…,yn′),
where
(34)y1′=1-[(Lr+3-tLr+2)·(-d)n-3cn-2+∑i=1n-3(Lr+n+2-i-tLr+n+1-i)·(-d)i-1ci],y2′=-t-∑i=1n-2(Lr+n+1-i-tLr+n-i)·(-d)i-1ci,y3′=-(Lr+3-tLr+2)·1c,y4′=-∑i=12(Lr+1+i-tLr+i)·(-d)i-1ci,yk′=-∑i=12(Lr+1+i-tLr+i)·(-d)k-5+ick-4+iiiiiiiiiiiiiiiiiiiiiiiiiiiiii(k=5,6,…,n).
Specially, when r=0, one gets the result of [23].
Proof.
Let
(35)Ω2=(1-ln′Lr+1ω13ω14⋯ω1n01ω23ω24⋯ω2n0010⋯00001⋯0⋮⋮⋮⋮⋱⋮0000⋯1),
where
(36)ω1j=1Lr+1[ln′ln(tLr+n+2-j-Lr+n+3-j)-Lr+n+2-j]ω2j=1ln·(Lr+n+3-j-tLr+n+2-j)(j=3,4,…,n).
Then, we have
(37)ΣAr,nΩ1Ω2=(Lr+1000⋯00ln00⋯000c0⋯000dc⋯0⋮⋮⋮⋮⋱⋮0000⋯c),
so ΣAr,nΩ1Ω2=𝒟⊕ℋ, where D=diag(Lr+1,ln) is a diagonal matrix and 𝒟⊕ℋ is the direct sum of 𝒟 and ℋ. If we denote Ω=Ω1Ω2, then we obtain Ar,n-1=Ω(𝒟-1⊕ℋ-1)Σ.
Since the last row elements of the matrix Ω are (0,1,ω23,ω24,…,ω2,n-1,ω2n), then the last row elements of the matrix Ω(𝒟-1⊕ℋ-1) are (0,1/ln,T23,T24,…,T2n ), where
(38)T23=∑i=1n-2ω2,2+i·(-d)i-1ci,T2k=∑i=1n+1-kω2,k-1+i·(-d)i-1ci(k=3,4,…,n).
Hence, it follows from Lemma 15 that letting Ar,n-1=SCirc(y1,y2,…,yn), then its last row elements are (-y2,-y3,…,-yn,y1) which are given by the following equations:
(39)-y2=tln+T23iiiiiiii=tln+1ln∑i=1n-2(Lr+n+1-i-tLr+n-i)·(-d)i-1ci,-y3=T2,n=1ln(Lr+3-tLr+2)·1c,-y4=T2,n-1-T2niiiiiiii=1ln∑i=12(Lr+1+i-tLr+i)·(-d)i-1ci,-y5=T2,n-2-T2n-1-T2niiiiiiii=1ln∑i=12(Lr+1+i-tLr+i)·(-d)ici+1,-yk=T2,n-k+3-T2,n-k+4-T2,n-k+5iiiiiiii=1ln∑i=12(Lr+1+i-tLr+i)·(-d)k-5+ick-4+i,iiiiiiii⋮-yn=T23-T24-T25iiiiiiii=∑i=1n-2ω2,2+i·(-d)i-1ci-∑i=1n-3ω2,3+i·(-d)i-1ciiiiiiiiiiii-∑i=1n-4ω2,4+i·(-d)i-1ciiiiiiiii=1ln∑i=12(Lr+1+i-tLr+i)·(-d)n-5+icn-4+i,iiiiy1=1ln-T23-T24iiiiiiii=1ln-1ln[(Lr+3-tLr+2)·(-d)n-3cn-2iiiiiiiiiiiiiiiiiiiiii+∑i=1n-3(Lr+n+2-i-tLr+n+1-i)·(-d)i-1ci].
Hence, we obtain
(40)y1=1ln-1ln[∑i=1n-3(-d)i-1ci(Lr+3-tLr+2)·(-d)n-3cn-2+∑i=1n-3(Lr+n+2-i-tLr+n+1-i)(-d)i-1ci],y2=-tln-1ln∑i=1n-2(Lr+n+1-i-tLr+n-i)·(-d)i-1ci,y3=-1ln(Lr+3-tLr+2)·1c,y4=-1ln∑i=12(Lr+1+i-tLr+i)·(-d)i-1ci,y5=-1ln∑i=12(Lr+1+i-tLr+i)·(-d)ici+1,yk=-1ln∑i=12(Lr+1+i-tLr+i)·(-d)k-5+ick-4+i,⋮yn=1ln∑i=12(Lr+1+i-tLr+i)·(-d)n-5+icn-4+i,Ar,n-1=1ln·SCirc(y1′,y2′,…,yn′),
where
(41)y1′=1-[(Lr+3-tLr+2)·(-d)n-3cn-2+∑i=1n-3(Lr+n+2-i-tLr+n+1-i)·(-d)i-1ci],y2′=-t-∑i=1n-2(Lr+n+1-i-tLr+n-i)·(-d)i-1ci,y3′=-(Lr+3-tLr+2)·1c,y4′=-∑i=12(Lr+1+i-tLr+i)·(-d)i-1ci,yk′=-∑i=12(Lr+1+i-tLr+i)·(-d)k-5+ick-4+i,(k=5,6,…,n).
This completes the proof.
3. Norm and Spread of Skew Circulant Matrix with the Lucas NumbersTheorem 17.
Let Ar,n=SCirc(Lr+1,…,Lr+n) be skew circulant matrix; then three kinds of norms of Ar,n are given by
(42)∥Ar,n∥1=∥Ar,n∥∞=Lr+n+2-Lr+2,(43)∥Ar,n∥F=n(Lr+nLr+n+1-LrLr+1).
Proof.
By Definition 8 and (12), we have
(44)∥Ar,n∥1=∥Ar,n∥∞=∑i=1nLr+i=Lr+n+2-Lr+2.
By Definition 8 and (13), we have
(45)(∥Ar,n∥F)2=∑i=1n∑j=1n|aij|2=n∑i=1nLr+i2=n(∑i=0r+nLi2-∑i=0rLi2)=n(Lr+nLr+n+1-LrLr+1).
Thus
(46)∥Ar,n∥F=n(Lr+nLr+n+1-LrLr+1).
Theorem 18.
Let
(47)Ar,n′=SCirc(Lr+1,-Lr+2,…,-Lr+n-1,Lr+n)
be an odd-order alternative skew circulant matrix and let n be odd. Then
(48)∥Ar,n′∥2=∑i=1nLr+i=Lr+n+2-Lr+2.
Proof.
By Lemma 3, we have
(49)λj(Ar,n′)=∑i=1n(-1)i-1Lr+i(ωjη)i-1.
So
(50)|λj(Ar,n′)|≤∑i=1n|(-1)i-1Lr+i|·|(ωjη)i-1|=∑i=1nLr+i,
for all j=0,1,…,n-1.
Since n is odd, ∑i=1nLr+i is an eigenvalue of Ar,n′; that is,
(51)(Lr+1-Lr+2⋮Lr+n-Lr+nLr+1-Lr+n-1Lr+n-1-Lr+nLr+n-2⋮⋮⋮⋮Lr+2-Lr+3Lr+1)(1-11-1⋮1)=∑i=1nLr+i·(1-11-1⋮1).
To sum up, we have
(52)max0≤j≤n-1|λj(Ar,n′)|=∑i=1nLr+i.
Since all skew circulant matrices are normal, by Lemma 9 and (12), and (52), we have
(53)∥Ar,n′∥2=∑i=1nLr+i=Lr+n+2-Lr+2,
which completes the proof.
Theorem 19.
Let Ar,n=SCirc(Lr+1,…,Lr+n) be skew circulant matrix; then the bounds for the spread of Ar,n are
(54)s(Ar,n)⩽2n(Lr+nLr+n+1-Lr+1Lr+2),s(Ar,n)≥1n-1|2Lr+n+3-(n-2)Lr+n+2-nLr+3-2Lr+4|.
Proof.
The trace of Ar,n, trAr,n=nLr+1. By (18) and (43), we have
(55)s(Ar,n)⩽2n(Lr+nLr+n+1-Lr+1Lr+2).
Since
(56)∑i≠jaij=∑k=2n(n-(k-1))Lr+k-∑k=2n(k-1)Lr+k=(n+2)∑k=2nLr+k-2∑k=2nkLr+k=(n+2)(Lr+n+2-Lr+3)-2[∑k=2n(r+k)Lr+k-∑k=2nrLr+k],
by (12) and (14),
(57)∑i≠jaij=2Lr+n+3-(n-2)Lr+n+2-nLr+3-2Lr+4.
By (19), we have
(58)s(Ar,n)≥1n-1|2Lr+n+3-(n-2)Lr+n+2-nLr+3-2Lr+4|.
4. Determinant and Inverse of Skew Left Circulant Matrix with the Lucas Numbers
In this section, let Ar,n′′=SLCirc(Lr+1,…,Lr+n) be skew left circulant matrix. By using the obtained conclusions in Section 2, we give a determinant explicit formula for the matrix Ar,n′′. Afterwards, we prove that Ar,n′′ is an invertible matrix for any positive interger n. The inverse of the matrix Ar,n′′ is also presented.
According to Lemmas 5 and 6 and Theorems 13, 14, and 16, we can obtain the following theorems.
Theorem 20.
Let Ar,n′′=SLCirc(Lr+1,…,Lr+n) be skew left circulant matrix; then
(59)detAr,n′′=(-1)n(n-1)/2Lr+1×[Lr+1+tLr+n+∑k=1n-2(tLr+1+i-Lr+2+i)xn-1-i]·cn-2,
where Lr+n is the (r+n)th Lucas number.
Theorem 21.
Let Ar,n′′=SLCirc(Lr+1,…,Lr+n) be skew left circulant matrix; then Ar,n′′ is an invertible matrix.
Theorem 22.
Let Ar,n′′=SLCirc(Lr+1,…,Lr+n) be skew left circulant matrix; then
(60)(Ar,n′′)-1=1lnSLCirc(y1′′,y2′′,…,yn′′),
where
(61)y1′′=1-[(Lr+3-tLr+2)(-d)n-3cn-2+∑i=1n-3(Lr+n+2-i-tLr+n+1-i)·(-d)i-1ci],yk′′=-yn-k+2′=∑i=12(Lr+1+i-tLr+i)·(-d)n-k-3+icn-k-2+i,iiiiiiiiiiiiiiiiiiiii(k=2,3,…,n-2).yn-1′′=-y3′=(Lr+3-tLr+2)·1c,yn′′=-y2′=t+∑i=1n-2(Lr+n+1-i-tLr+n-i)·(-d)i-1ci.
5. Norm and Spread of Skew Left Circulant Matrix with the Lucas NumbersTheorem 23.
Let Ar,n′′=SLCirc(Lr+1,…,Lr+n) be skew left circulant matrix. Then three kinds of norms of Ar,n′′ are given by
(62)∥Ar,n′′∥1=∥Ar,n∥∞=Lr+n+2-Lr+2,∥Ar,n′′∥F=n(Lr+nLr+n+1-LrLr+1).
Proof.
Using the method in Theorem 17 similarly, the conclusion is obtained.
Theorem 24.
Let
(63)Ar,n′′′=SLCirc(Lr+1,-Lr+2,…,-Lr+n-1,Lr+n)
be an odd-order alternative skew left circulant matrix; then
(64)∥Ar,n′′′∥2=∑i=1nLr+i=Lr+n+2-Lr+2.
Proof.
According to Lemma 4,
(65)λj(Ar,n′′′)=±|∑i=1n(-1)i-1Lr+iω(j-(1/2))(k-1)|,
for j=1,2,…,(n-1)/2, and
(66)λ(n+1)/2(Ar,n′′′)=∑i=1nLr+i.
So
(67)|λj(Ar,n′′′)|≤∑i=1n|(-1)i-1Lr+i(-1)i-1|=∑i=1nLr+i,(j=1,2,…,n+12).
By (66) and (67), we have
(68)max0≤i≤(n+1)/2|λi(Ar,n′′′)|=∑i=1nLr+i.
Since all skew left circulant matrices are symmetrical, by Lemma 9 and (12) and (68), we obtain
(69)∥Ar,n′′′∥2=Lr+n+2-Lr+2.
Theorem 25.
Let Ar,n′′=SLCirc(Lr+1,…,Lr+n) be skew left circulant matrix; the bounds for the spread of Ar,n′′ are
(70)2Lr+n≤s(Ar,n′′)≤{M-2nN2,ifnisodd,M,ifniseven,
where
(71)M=2n(Lr+nLr+n+1-Lr+1Lr),N=Lr+n-1+Lr-1.
Proof.
Since Ar,n′′ is a symmetric matrix, by (20),
(72)s(Ar,n′′)≥2maxi≠j|aij|=2Lr+n.
The trace of Ar,n′′ is, if n is odd,
(73)tr(Ar,n′′)=Lr+1-Lr+2+Lr+3-⋯+Lr+n=Lr+1+Lr+1+Lr+3+⋯+Lr+n-2=2Lr+1+Lr+1+Lr+2+⋯+Lr+n-3=2Lr+1+∑i=1n-3Lr+i.
By (12), we have
(74)tr(Ar,n′′)=Lr+n-1+Lr-1=N.
Let M=2n(Lr+nLr+n+1-Lr+1Lr); then, by (18), (62), and (74), we obtain
(75)s(Ar,n′′)⩽M-2nN2.
If n is even, then
(76)tr(Ar,n′′)=Lr+1-Lr+1+Lr+3-Lr+3⋯-Lr+n-1=0.
By (18), (62), and (76), we have
(77)s(Ar,n′′)⩽M.
So the result follows.
6. Conclusion
We discuss the invertibility of the skew circulant type matrices with any continuous Lucas numbers and present the determinant and the inverse matrices by constructing the transformation matrices. The four kinds of norms and bounds for the spread of these matrices are given, respectively. In [3], a new fast algorithm for optimal design of block digital filters (BDFs) is proposed based on skew circulant matrix. The reason why we focus our attention on skew circulant is to explore the application of skew circulant in the related field in medicine image, image encryption, and real-time tracking. On the basis of existing application situation [4], we conjecture that SVD decomposition of skew circulant matrix will play an important role in CT-perfusion imaging of human brain. On the basis method of [8] and ideas of [5], we will exploit real-time tracking with kernel matrix of skew circulant structure. A novel chaotic image encryption scheme based on the time-delay Lorenz system is presented in [6] with the description of circulant matrix. We will exploit chaotic image encryption algorithm based on skew circulant operation.
Conflict of Interests
The authors declare that there is no conflict of interests regarding the publication of this paper.
Acknowledgments
This research was supported by the Natural Science Foundation of Shandong Province (Grant no. ZR2011FL017), the National Nature Science Foundation of China (Grant no. F020701), and the AMEP of Linyi University, China.
DaherA.BaghiousE. H.BurelG.Fast algorithm for optimal design of block digital filters based on circulant matrices2008156376402-s2.0-6765017072910.1109/LSP.2008.2003988LiuV. C.VaidyanathanP. P.Circulant and skew-circulant matrices as new normal-form realization of IIR digital filters198835662563510.1109/31.1800MR944131FuD. Q.JiangZ. L.CuiY. F.JhangS. T.A new fastalgorithm for optimal design of block digital filters by skew-cyclic convolution2014610.1049/iet-spr.2013.0384WittsackH.-J.WohlschlägerA. M.RitzlE. K.KleiserR.CohnenM.SeitzR. J.MödderU.CT-perfusion imaging of the human brain: advanced deconvolution analysis using circulant singular value decomposition200832167772-s2.0-3594898926510.1016/j.compmedimag.2007.09.004HenriquesJ. F.CaseiroR.MartinsP.BatistaJ.Exploiting the circulant structure of tracking-by-detection with kernels20127575Berlin, GermanySpringer702715Lecture Notes in Computer Science10.1007/978-3-642-33765-9_50HuangX.YeG.WongK.-W.Chaotic image encryption algorithm based on circulant operation201320138384067MR308157810.1155/2013/384067JingY.JafarkhaniH.Distributed differential space-time coding for wireless relay networks2008567109211002-s2.0-4904908914010.1109/TCOMM.2008.060388NarasimhaM. J.Linear convolution using skew-cyclic convolutions20071431731762-s2.0-3394757254110.1109/LSP.2006.884034GulliverT. A.HaradaM.New nonbinary self-dual codes200854141541710.1109/TIT.2007.911265MR2446767DavisP. J.1979New York, NY, USAJohn Wiley & SonsMR543191JiangZ. L.ZhouZ. X.1999Chengdu, ChinaChengdu Technology University PublishingBertacciniD.NgM. K.Skew-circulant preconditioners for systems of LMF-based ODE codes20011988Berlin, GermanySpringer93101Lecture Notes in Computer Science10.1007/3-540-45262-1_12MR1938408ZBL0979.65064ChanR. H.JinX.-Q.Circulant and skew-circulant preconditioners for skew-Hermitian type Toeplitz systems199131463264610.1007/BF01933178MR1140536ZBL0743.65026ChanR. H.NgK.-P.Toeplitz preconditioners for Hermitian Toeplitz systems199319018120810.1016/0024-3795(93)90226-EMR1230358ZBL0783.65042HuckleT.Circulant and skew circulant matrices for solving Toeplitz matrix problems199213376777710.1137/0613048MR1168078ZBL0756.65047LynessJ. N.SørevikT.Four-dimensional lattice rules generated by skew-circulant matrices2004732452792952-s2.0-094227883410.1090/S0025-5718-03-01534-5MR2034122ZBL1035.65026KarnerH.SchneidJ.UeberhuberC. W.Spectral decomposition of real circulant matrices200336730131110.1016/S0024-3795(02)00664-XMR1976927ZBL1021.15007LiJ.JiangZ. L.ShenN.ZhouJ. W.On optimal backward perturbation analysis for the linear system with skew circulant coefficient matrix20132013770738110.1155/2013/707381JaiswalD. V.On determinants involving generalized Fibonacci numbers19697319330MR0258740ZBL0191.04501LindD. A.A Fibonacci circulant197085449455MR0280459ZBL0215.06802DazhengL.Fibonacci-Lucas quasi-cyclic matrices2002403280286MR1913357ZBL1081.11011ShenS.-Q.CenJ.-M.HaoY.On the determinants and inverses of circulant matrices with Fibonacci and Lucas numbers2011217239790979710.1016/j.amc.2011.04.072MR2811251ZBL1222.15010GaoY.JiangZ. L.GongY. P.On the determinants and inverses of skew circulant and skew left circulant matrices with Fibonacci and Lucas numbers2013124472481JiangX. Y.GaoY.JiangZ. L.Determinantsand inverses of skew and skew left circulant matrices involving the k-Fibonacci numbers in communications-I2013761123137JiangX. Y.GaoY.JiangZ. L.Determinants and inverses of skew and skew left circulant matrices involving the k-Lucas numbers in communications-II2013781117SolakS.On the norms of circulant matrices with the Fibonacci and Lucas numbers2005160112513210.1016/j.amc.2003.08.126MR2100179ZBL1066.15029İpekA.On the spectral norms of circulant matrices with classical Fibonacci and Lucas numbers entries2011217126011601210.1016/j.amc.2010.12.094MR2770224ZBL1211.15028ShenS.CenJ.On the bounds for the norms of r-circulant matrices with the Fibonacci and Lucas numbers2010216102891289710.1016/j.amc.2010.03.140MR2653105AkbulakM.BozkurtD.On the norms of Toeplitz matrices involving Fibonacci and Lucas numbers20083728995MR2488956ZBL1204.15031BoseA.HazraR. S.SahaK.Spectral norm of circulant-type matrices201124247951610.1007/s10959-009-0257-zMR2795050ZBL1241.60006MirskyL.The spread of a matrix19563127130MR0081875ZBL0073.00903SharmaR.KumarR.Remark on upper bounds for the spread of a matrix2013438114359436210.1016/j.laa.2013.01.018MR3034536ZBL06248217MirskyL.Inequalities for normal and Hermitian matrices1957244591599MR009057310.1215/S0012-7094-57-02467-5ZBL0081.25101BarnesE. R.HoffmanA. J.Bounds for the spectrum of normal matrices1994201799010.1016/0024-3795(94)90106-6MR1274878ZBL0803.15016JiangE.ZhanX.Lower bounds for the spread of a Hermitian matrix199725615316310.1016/S0024-3795(95)00785-7MR1438231ZBL0886.15017BhatiaR.SharmaR.Some inequalities for positive linear maps201243661562157110.1016/j.laa.2010.09.038MR2890939ZBL1252.15024WuJ.ZhangP.LiaoW.Upper bounds for the spread of a matrix2012437112813282210.1016/j.laa.2012.07.007MR2964726ZBL1258.15004JohnsonC. R.KumarR.WolkowiczH.Lower bounds for the spread of a matrix19857116117310.1016/0024-3795(85)90244-7MR813042ZBL0578.15013