COMPUTATIONAL PROBLEMS IN ABSTRACT ALGEBRA.
COMPUTATIONAL PROBLEMS IN ABSTRACT ALGEBRA.
COMPUTATIONAL PROBLEMS IN ABSTRACT ALGEBRA.
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
108<br />
C. Brott and J. Neubiiser<br />
elements of G in q, are computed and printed. The programme stops as<br />
soon as r irreducible representations have been found. This often happens<br />
long before all triplets N, U, V have been investigated.<br />
4.7. In debugging the programme we have used the special version (2.1 d)<br />
of the programme @ to generate a matrix group from the images of the<br />
generators of G under an irreducible representation T. This must be isomorphic<br />
to G/ker T.<br />
5. The numerical part of the programme.<br />
5.1. The part of the programme described so far finds only the monomial<br />
irreducible representations and characters. For low orders there are very<br />
few groups having non-monomial irreducible representations. Up to order<br />
96 there are one of order 24, four of order 48, one of order 60 and two of<br />
order 72. No method is included yet in our programme for the determination<br />
of non-monomial irreducible representations, but at least an attempt is<br />
made to complete the character table by a more numerical part of the programme.<br />
For its description we define:<br />
Then from [2], p. 235,<br />
we have<br />
I<br />
WsJ)W(i) = C CijkW,Q, lei,j,s==r,<br />
k=l<br />
kg1 (Ciik -&&y))w~) = 0 ,<br />
(5.1.1)<br />
(5.1.2)<br />
(5.1.3)<br />
i.e. for each s the r values WV), 1 4 I 4 r, belonging to the sth character satisfy<br />
the ra equations<br />
k$l(Cijk-SikXj)Xk = 0, 1 =S i, j e r. (5.1.4)<br />
To solve this system we choose a fixed j = jo and consider only the r equations<br />
k~l(c&&-djJ&jo)x~ = 0, 1 4 i < r. (5.1.5)<br />
Then for each s, 1~ s 4 r, the vector (IV?), . . . , WY)) is an eigenvector of the<br />
matrix (cildr) belonging to the eigenvalue wji’. If for some j,-, this matrix has<br />
r different eigenvalues, the eigenvectors are essentially uniquely determined<br />
and hence must coincide up to a factor with the vectors (WY), . . . , w$@),<br />
1 4 s 4 r. This factor is calculated from<br />
h<br />
From (21, 31.18,<br />
Group characters and representations 109<br />
(5.1.6)<br />
we obtain the dimensions dl, . . . , d, and hence the values #).<br />
In our programme we use the characters already obtained by the process<br />
of induction to reduce the task of finding the eigenvalues of the (cijk) as<br />
roots of the characteristic polynomials.<br />
5.2. For the numerical programme we first compute the structure constants<br />
cijke Let X1, . . . , x, be representatives of the r classes of elements conjugate<br />
in G. Then for all k = l( 1)r and all b E Ci the number of solutions of<br />
bx = xk in Cj is counted.<br />
For each matrix (cu.& 2 ==j--r, its characteristic polynomial is computed<br />
by the Hessenberg procedure [12]. Zeros of this polynomial belonging to<br />
known characters are split off and the zeros of the remaining polynomial are<br />
computed by the Bairstow procedure. For a simple root of this polynomial<br />
a character is found as an eigenvector.<br />
The numerical method described above does not work if there is a nonmonomial<br />
character x(“) such that for each j there exists s’ + s with<br />
$I = $‘I. This case has not yet been covered in our programme, but we<br />
intend to replace the numerical part by a method proposed by John D.<br />
Dixon+ which makes use of the fact that the vectors w@), l< s < r, are essentially<br />
UniqUdy determined as common eigenvectors of all matrices (cijk),<br />
1