An integer function of finite sets of points on spheres
Paul C. Leopardi
leopardi at bigpond.net.au
Sun Apr 10 13:53:39 CEST 2005
Hi all,
Here is an integer function of a set of points on a sphere, which can be used
to generate integer sequences from sequences of point sets on spheres.
In brief, in Matlab:
function R = rsg(x)
%rsg The number of distinct sets of inner products of x
R = rank(sort(x'*x));
In detail:
Take a set of N points on the sphere S^d in R^(d+1), expressed in Cartesian
coordinates as the (d+1) by N matrix x.
Now, using (eg.) Matlab, calculate
R = rank(sort(x'*x));
ie.,
M = x'*x;
Y = sort(M);
R = rank(Y);
The expression M = x'*x gives the usual Gram matrix of inner products of the
points of x. This is symmetric by construction.
The sort function Y = sort(M) (at least in Matlab) sorts the columns of the
matrix M, so that each column Y(:,k) of Y contains the inner products of x
with respect to x(k) sorted in increasing order.
The rank function R = rank(Y) then determines the number of different columns
of Y, which is the number of different sets of distinct inner products of x.
Since R is a rank of a non-zero N*N matrix, it must be an integer between 1
and N inclusive. The ratio R/N could be used to measure a kind of geometric
irregularity. If R == 1, ie. R/N == 1/N, then the set of points would be very
regular in this sense and if R == N, ie. R/N == 1, the set of points would be
very irregular.
Example:
When used on Hardin and Sloane's Library of 3-d designs
http://www.research.att.com/~njas/sphdesigns/dim3
for spherical designs with N=(n+1)^2 points, the function rsg yields the
following sequences N, R:
n: 1, 2, 3, 4, 5, 6, 7, ...
N: 4, 9, 16, 25, 36, 49, 64, ...
R: 1, 2, 2, 3, 3, 48, 13, ...
Notice the large contrast between R(5)/N(5) == 1/12 and R(6)/N(6) == 48/49.
I have not yet calculated the rsg of all of the point sets in the library of
3-d designs, because it takes too long to download each file individually. Is
there somewhere I can download a compressed (eg. Gzipped) file containing all
of these point sets?
Best regards
References:
"Uniqueness of the 120-point spherical 11-design in four dimensions", Archiv
der Mathematik, P. Boyvalenkov and D. Danev, Volume 77, Number 4,
October 2001, 360 - 368.
http://www.springerlink.com/link.asp?id=vkuwt9wda3wj0jnv
"McLaren's Improved Snub Cube and Other New Spherical Designs in Three
Dimensions", R. H. Hardin and N. J. A. Sloane, Discrete and Computational
Geometry, 15 (1996), pp. 429-441.
http://www.research.att.com/~njas/sphdesigns/
Matlab documentation on x'*x, sort and rank
http://www.mathworks.com/access/helpdesk/help/techdoc/ref/arithmeticoperators.html
http://www.mathworks.com/access/helpdesk/help/techdoc/ref/sort.html
http://www.mathworks.com/access/helpdesk/help/techdoc/ref/rank.html
More information about the SeqFan
mailing list