On an inverse formula of a tridiagonal matrix (Q2914865)

From MaRDI portal





scientific article; zbMATH DE number 6084688
Language Label Description Also known as
English
On an inverse formula of a tridiagonal matrix
scientific article; zbMATH DE number 6084688

    Statements

    On an inverse formula of a tridiagonal matrix (English)
    0 references
    0 references
    21 September 2012
    0 references
    tridiagonal matrix
    0 references
    inversion
    0 references
    product integral
    0 references
    Volterra equation
    0 references
    computational stability
    0 references
    limit elements
    0 references
    zero minors
    0 references
    Usmani's formula
    0 references
    In this article, the author considers the inverse \(Z_n=(z_{ij})\) of \(Y_n^{-1}\) of a general \(n\times n\) tridiagonal matrix NEWLINE\[NEWLINEY_n=(y_{ij})=\begin{pmatrix} \alpha_1 & \gamma_1 & 0 & \cdots & 0 \\ \beta_2 & \alpha_2 & \gamma_2 & \ddots & \vdots \\ 0 & \beta_3 & \ddots & \ddots & 0 \\ \vdots & \ddots & \ddots & \ddots & \gamma_{n-1} \\ 0 & \cdots & 0 & \beta_n & \alpha_n \end{pmatrix}.NEWLINE\]NEWLINE By a transformation of the well-known Usmani's formula, he obtains the followingNEWLINENEWLINE\textbf{Theorem 1}: For a tridiagonal matrix \(Y_n\), define sequences \(f_l\) and \(g_l\) by NEWLINE\[NEWLINE f_l=\frac{-\gamma_l}{f_{l-1}\beta_l+\alpha_l}, g_l=\frac{-\beta_l}{g_{l+1}\gamma_l+\alpha_l} NEWLINE\]NEWLINE with \(f_0=g_{n+1}=0\). Further define NEWLINE\[NEWLINE p_{ki}=\begin{cases} \Pi_{l=k}^{i-1}f_l, & \text{if } k<i \\ 1, & \text{if } k=i \\ \Pi_{l=i+1}^kg_l & \text{if } k>i \\ \end{cases}. NEWLINE\]NEWLINE Then the \((i,j)\)-th element of \(Z_n\) is NEWLINE\[NEWLINEz_{ij}=\frac{p_{ij}}{\beta_jf_{j-1}+\alpha_j+\gamma_jg_{j+1}}.NEWLINE\]NEWLINE The author claims that, compared to Usmani's formula, the new formulation is (i) still quite simple, (ii) more stable in numerical calculation (which must be dealt with caution when zero minors are involved, as discussed in Section 5), (iii) better for study the limit form (which is discussed in Section 4).
    0 references

    Identifiers