Given an arbitrary finite group, the computation of its automorphism group is a very difficult task. Pioneer work in this area was carried out by Felsch & Neubüser (1970), whose algorithm used the output of their subgroup lattice program. A technique developed by Neubüser in the early 1970s sought to compute the automorphism group viewed as a permutation group acting on unions of certain conjugacy classes of the group. A similar method was implemented by Hulpke (1997) in the GAP 4 library. Recently, Cannon & Holt (1999) presented a new algorithm which uses a ``hybrid group'' approach.
More efficient approaches are available to determine the automorphism group for groups satisfying certain properties. Following the work of Shoda (1928), Hulpke in 1997 implemented a practical method for finite abelian groups in the GAP 4 library. Wursthorn (1993) adapted modular group algebra techniques to compute the automorphism groups of p-groups; the GAP 3 share package Sisyphos includes an implementation. Smith (1994) introduced an algorithm for finite solvable groups which is available in the AutAg share package of GAP 3.
Moreover, the p-group generation method of Newman (1977) and O'Brien
(1990) can be modified to compute the automorphism group of a finite
p-group as outlined in O'Brien (1995). This algorithm is implemented
in the ANU pq C program.
Here we introduce a new function to compute the automorphism group of
a finite p-group. The underlying algorithm is a refinement of the
methods described in O'Brien (1995). In particular, this implementation
is more efficient in both time and space requirements and hence has a
wider range of applications than the ANU pq method. Our package is
written in GAP code and it makes use of a number of methods from the
GAP library such as the MeatAxe for matrix groups and
permutation group functions.
The GAP 4 package ANUPQ, which is an interface to most of
the functionality of the ANU pq C program, uses the AutPGrp package
to compute automorphism groups of p-groups.
We have compared our method to the others available in GAP. Our package usually out-performs all but the method designed for finite abelian groups. We note that our method uses the small groups library in certain cases and hence our algorithm is more effective if the small groups library is installed.
AutPGrp manual