A Hilbert matrix is a matrix with elements
|
(1)
|
for , 2, ...,
, where
is a positive integer.
Thus every Hilbert matrix is a rational matrix.
Hilbert matrices are implemented in the Wolfram
Language by HilbertMatrix[n]
or, for a rectangular matrix, by HilbertMatrix[{m,
n}]. The figure above shows a plot of the
Hilbert matrix with elements colored according
to their values.
Hilbert matrices whose entries are specified as machine-precision numbers are difficult to invert using numerical techniques.
The determinants for the first few values of for
, 2, ... are given by one divided by 1, 12, 2160,
6048000, 266716800000, ... (OEIS A005249).
Writing
gives the closed form
|
(2)
| |||
|
(3)
| |||
|
(4)
|
where
is the Glaisher-Kinkelin constant and
is the Barnes
G-function. The numerical values are given in the following table.
| det( | |
| 1 | 1 |
| 2 | |
| 3 | |
| 4 | |
| 5 | |
| 6 |
The elements of the matrix inverse of the Hilbert matrix are given analytically
by
|
(5)
|
(Choi 1983, Richardson 2001), showing that the matrix inverse is an integer matrix.
A simpler form of the determinant is
|
(6)
|
where
is the Barnes G-function (Vardi 1988).
An explicit Cholesky decomposition is , where
is unit lower triangular
and
is a diagonal matrix, with
|
(7)
| |||
|
(8)
|
The formula for
applies for
, with
for
(Hitotumatu 1988). The lower
triangular factor in the usual Cholesky
decomposition
is
.