An application of matricial Fibonacci identities to the computation of spectral norms (Q457949)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | An application of matricial Fibonacci identities to the computation of spectral norms |
scientific article |
Statements
An application of matricial Fibonacci identities to the computation of spectral norms (English)
0 references
30 September 2014
0 references
The authors present the exact value for the spectral norms of Toeplitz matrices constructed from Fibonacci and Lucas numbers. Let \(A=(a)_{i j}\) be a matrix with rank \(k\) and adjoint \(A^*\), and let the singular values of \(A\) be defined by the non-zero eigenvalues of \(|A|=\left (A^* A\right )^{1/2} \), labelled \(s_1,s_2,\dots,s_k\) with \(s_1\geq s_2\geq \ldots \geq s_k>0\). Then, the Schatten norms of the matrix \(A\) are a family of unitary invariant norms, denoted by \(\| A \|_p\) \((1\leq p \leq \infty)\), with \[ \| A \|_p = \left (\sum_{i=1}^k s_i^p\right )^{1/p},\qquad 1 \leq p <\infty,\qquad \text{and}\qquad \| A \|_\infty =s_1. \] When \(p=2\), the Schatten norm equates to the Frobenius norm, and when \(p=\infty\), the Schatten norm returns the maximal singular value of \(A\), denoted by \(\|A\|\), and referred to as the spectral norm. The authors consider the Frobenius and spectral norms of the family of Fibonacci and Lucas matrices, denoted by \(F\) and \(L\), and defined such that \[ F=\left (f_{i-j}\right )_{i,\, j=0}^{n-1},\qquad\text{and}\qquad L=\left (l_{i-j}\right )_{i,\, j=0}^{n-1}, \] where \(f_j\) and \(l_j\) are respectively the \(j\)th Fibonacci and Lucas sequence terms. It follows from the definitions that the matrices \(F\) and \(L\) are Toeplitz matrices. The main results of this paper express the spectral norm of the matrices \(F\) and \(L\) in terms of Fibonacci and Lucas numbers, and state for \(n\) even that \[ \|F\|=\frac{1}{\sqrt{2}}\|F\|_2=\sqrt{\sum_{i=0}^{n-1}f_i\,f_{i+1}}=f_n,\qquad \|L\|=\sqrt{5}\|F\|=\sqrt{5}f_n, \] and for \(n=2t+1\) odd that \[ \|F\|=\frac{1}{\sqrt{2}}\|F\|_2=\sqrt{\sum_{i=0}^{n-1}f_i\,f_{i+1}}=\sqrt{f_n^2-1}, \] with \(\|L\|=l_t l_{t+1}\) for even \(t\), and \(\|L\|= 5f_t f_{t+1}\) for odd \(t\). A corollary for the Lucas matrices extends these results to all Schatten-\(p\) norms with \(0\leq p<\infty\), and \(n=2t+1\), such that \[ \|L\|_p=\left (l_t^p l_{t+1}^p+5^p f_t^p f_{t+1}^p\right )^{1/p}. \] As a by-product of these results, it is demonstrated that the Fibonacci and Lucas matrices also obey analogous identities to those for the Fibonacci and Lucas numbers. In particular, it is shown that \[ \left (F^*\, F\right )^2=f_n\left (F^*\,F\right ),\qquad \text{and}\qquad L^*\, L=5F^*\, F, \] and the algebra constructed from such Toeplitz matrices is discussed. Concluding results concern the spectral norms of circulant matrices constructed from the Fibonacci and Lucas sequences.
0 references
matrix norms
0 references
Fibonacci matrices
0 references
Toeplitz matrices
0 references
circulant matrices
0 references
Fibonacci numbers
0 references
Lucas numbers
0 references
unitary invariant norm
0 references
Frobenius norm
0 references
singular value
0 references
spectral norm: Lucas matrices
0 references
Schatten norm
0 references