nLab Faddeev-LeVerrier algorithm

Contents

Overview

The Faddeev–Le Verrier algorithm (or Frame–Faddeev algorithm) is an nn-step procedure to provide the inverse of an n×nn\times n square matrix AA if it exists. Along the way it computes the coefficients of the characteristic polynomial of AA using matrix multiplication, calculating the trace and subtraction of matrices.

Statement

Let AA be a n×nn\times n square matrix over a field and II the unit matrix of the same size. Define

A 1=A σ 1=trA 1 B 1=A 1σ 1I A 2=AB 1 σ 2=12trA 2 B 1=A 2σ 2I A k=AB k1 σ k=1ktrA k B k=A kσ kI A n=AB n1 σ k=1ntrA n B n=A nσ nI. \begin{matrix} A_1 = A & \sigma_1 = tr A_1 & B_1 = A_1 - \sigma_1 I \\ A_2 = A B_1 & \sigma_2 = \frac{1}{2}tr A_2 & B_1 = A_2 - \sigma_2 I \\ \vdots &\vdots &\vdots \\ A_k = A B_{k-1}& \sigma_k = \frac{1}{k}tr A_k & B_k = A_k -\sigma_k I\\ \vdots &\vdots &\vdots\\ A_n = A B_{n-1}& \sigma_k = \frac{1}{n}tr A_n & B_n = A_n -\sigma_n I \mathrlap{\,.} \end{matrix}

Then B n=0B_n = 0 and the characteristic polynomial κ A(λ)=det(λIA)\kappa_A(\lambda) = det(\lambda I - A) of AA is:

κ A(λ)=λ nσ 1λ n1σ nI. \kappa_A(\lambda) = \lambda^n - \sigma_1 \lambda^{n-1} - \ldots - \sigma_n I.

The matrix AA is regular (has an inverse) iff σ n0\sigma_\n\neq 0 and its inverse A 1A^{-1} is

A 1=B n1σ n. A^{-1} = \frac{B_{n-1}}{\sigma_n}.

References

Named after:

  • Urbain Le Verrier: Sur les variations séculaires des éléments des orbites pour les sept planètes principales, J. de Math. 5 1 (1840) 230 [gallica:pdf]

  • D. K. Faddeev, I. S. Sominskii (translated from the Russian by J. L. Brenner): Problems in Higher Algebra, W. H. Freeman and Company (1965)

Last revised on April 15, 2026 at 08:24:27. See the history of this page for a list of all contributions to it.