With arguments
suppressed,
For
,
, and
see (33.4.1).
where
The continued fraction (33.8.1) converges for all finite values of
, and (33.8.2) converges for all
.
If we denote
and
, then
The ambiguous sign in (33.8.4) has to agree with that of the final denominator in (33.8.1) when the continued fraction has converged to the required precision. For proofs and further information see Barnett et al. (1974) and Barnett (1996).