Determinant of an infinite matrix

The Endeavour 2024-02-26

What does the infinite determinant

D = \left| \begin{array}{lllll} 1 & a_1 & 0 & 0 & \cdots \\ b_1 & 1 & a_2 & 0 & \cdots \\ 0 & b_2 & 1 & a_3 & \\ 0 & 0 & b_3 & 1 & \ddots \\ \vdots & \vdots & & \ddots & \ddots \\ \end{array} \right|

mean and when does it converge?

The determinant D above is the limit of the determinants Dn defined by

D_n = \left| \begin{array}{lllll} 1 & a_1 & & & \\ b_1 & 1 & a_2 & & \\ & b_2 & 1 & \ddots & \\ & & \ddots & \ddots & a_{n-1} \\ & & & b_{n-1} & 1 \\ \end{array} \right|

If all the a‘s are 1 and all the b‘s are −1 then this post shows that Dn = Fn, the nth Fibonacci number. The Fibonacci numbers obviously don’t converge, so in this case the determinant of the infinite matrix does not converge.

In 1895, Helge von Koch said in a letter to Poincaré that the infinite determinant is absolutely convergent if and only if the sum

\sum_{i=1}^\infty a_ib_i

is absolutely convergent. A proof is given in [1].

The proof shows that the Dn are bounded by

\prod_{i=1}^n\left(1 + |a_ib_i| \right)

and so the infinite determinant converges if the corresponding infinite product converges. And a theorem on infinite products says

\prod_{i=1}^\infty\left(1 + |a_ib_i| \right)

converges absolute if the sum in Koch’s theorem converges. In fact,

\prod_{i=1}^\infty\left(1 + |a_ib_i| \right) \leq \exp\left(\sum_{i=1}^\infty |a_ib_i| \right )

and so we have an upper bound on the infinite determinant.

Related post: Triadiagonal systems, determinants, and cubic splines

[1] A. A. Shaw. H. von Koch’s First Lemma and Its Generalization. The American Mathematical Monthly, April 1931, Vol. 38, No. 4, pp. 188–194

The post Determinant of an infinite matrix first appeared on John D. Cook.