powers of Antisymmetric Signed binary Matrices
Marc LeBrun
mlb at fxpt.com
Tue Aug 26 01:45:30 CEST 2003
> wouter meeussen
> it's becoming:
> "invertable (-1,0,1) nXn matrices having (Tij = -Tji ;i<j) such that all
T^k (k= 1..12) are also
> (-1,0,1) matrices"
> and comment:
> "the matrix powers T^k reach identity I for k a divisor of 12. All are
both invertible (non-zero
> determinant) (-1,0,1)-matrices and "pseudo-anti-symmetric" (that is
Tij=-Tji except for the main
> diagonal, or, equivalently, the sum of an anti-symmetric and a diagonal
matrix)."
"invertible"
^
If you say "all T^k for k=1..12 have Property X" then it includes both the
cases k=1 and k>1 and so you don't have to append that repetitious "are
also X" part. (Also the T^0=I case apparently also satisfies X<;-).
Moreover, I agree, it's beginning to seem, from the proliferation of
attributes and restrictions, that we ought to analyze what's going on a bit
more before settling on a description and comment.
It's usually telling when we count something and all the instances that
meet the criteria also meet some apparently stronger restrictions.
It's not clear exactly what combination of independent attributes
constitutes Property X here, which of the nontrivial powers (k=-1, >+1)
share all or some of these attributes comprising X, and which attributes
are simply inevitable consequences of others.
For example, it seems significant that all T are "roots of unity". Isn't
being invertible a consequence of this? If T^p=I then T^(-1)=T^(p-1). And
so mustn't they also all have determinant +1 or -1?
We might be counting some special subset embedded in a larger
structure. In particular, apparently they can all be expressed as powers
of twelfth roots. However any given one of these 12th roots itself might
not qualify as fully having Property X--are there some that do (and hence
are also included in the T count) and some that don't?
It might be a good idea to look at the entire graph of the explicit Ts,
their underlying 12th roots and their powers for some small n and see if
they form some recognizable structure.
More information about the SeqFan
mailing list