Whittaker's root series is a series of ratios of determinants for computing a root of an analytic function (Whittaker 1917, pp. 103-106). It applies to polynomials as well as convergent power series.
Let
|
(1)
|
where , and suppose
is an analytic function
in the disk
. Suppose
is a root with
and every other distinct root
in this disk has absolute
value strictly greater than
. The multiplicity
of
may exceed one.
Set
for
, and also above the polynomial
degree when
is a polynomial. Define
and
|
(2)
| |||
|
(3)
|
where the matrices defining are upper Hessenberg matrices
and Toeplitz matrices. Thus
,
,
, and
.
If
for every positive integer
, then
|
(4)
|
The partial sums satisfy
|
(5)
|
The hypothesis on the root of smallest absolute value alone does not ensure that all the denominators
in the series are nonzero.
Even when some early vanish, the ratio
is defined for all sufficiently large
and tends to
.
For a short proof, write
|
(6)
|
where . Equating coefficients,
or applying Cramer's rule, gives
. The determinant
identity
|
(7)
|
for
gives a telescoping series with
and
. The unique nearest pole
of
is
. Its finite principal
part gives coefficients equal to
times a polynomial in
with polynomial degree one less than the multiplicity of
, while the remaining coefficients
are exponentially smaller. It follows that
.
For example,
gives
|
(8)
|