Fix a positive integer
. If
with
and
prime and
and
are not
-power smooth, then
the Pollard
-method is unlikely to work.
For example, let
and
suppose that
. Note that
neither
nor
is
-power smooth.
With
, we have
and
, so we do not find a factor of
.
As remarked above, the problem is that
is not
-power smooth for
either
or
. However, notice that
is
-power smooth. Lenstra's ECM replaces
, which has order
, by the group of points
on an elliptic curve
over
.
It is a theorem that
for some nonnegative integer
(see e.g., [#!silverman:aec!#, §V.1] for a proof).
(Also every value of
subject to this
bound occurs, as one can see using ``complex multiplication
theory''.) For example, if
is the elliptic curve
over
then by enumerating points one sees that
is cyclic of order
. The set of numbers
for
contains
numbers that are
-power smooth
for
.Thus working with an elliptic curve gives us more flexibility.
For example,
is
-power smooth and
is
-power smooth.
William
2007-06-01