[seqfan] Re: Modular Partitions

William Keith william.keith at gmail.com
Sat May 3 07:19:54 CEST 2014


Chris' conjectures are accurate.

Suppose we are in the case n !==0 mod p.  Diagram one of Jen's partitions
in Z/pZ into n parts with a Ferrers diagram, with parts of size 0 noted.
It will fit in a p-1 by n box.  Choose any profile, to create a Ferrers
diagram; there are (n+p-1)-choose-n possible profiles.

If we add 1 to each part, rotating any new parts of size p to the end as
parts of size 0, we can do so p times and get back our original profile.
If p is prime and n !==0 mod p, then none of this group of p modpartitions
can possibly be the same: Suppose you had k parts of size 0, t additions
ago, then now you have k parts of size t.  If this is the same partition as
the one t additions ago, then you must have had k parts of size t, so now
you have k parts of size 2t, etc, until you had k parts of each size to
begin with, i.e. p divides n.

So we can group the p resulting profiles.  But adding 1 to each of n parts
adds n to the total mod p, so the residue of the profiles mod p are all
distinct, and hence exactly one of them is 0 mod p.  Thus there are (1/p)
(n-1+p choose n) partitions of 0 mod p with n parts when p is prime, p not
dividing n.

When n = kp, then exactly the one partition listed above will be separate
from the groups; take (1/p)[(n+p-1 choose n) - 1] + 1.    QED.

I think this diagram technique, and some sort of rotation counting, should
also be able to prove symmetry.  I don't know anything about Molien
series.  (Yet.)  If something from that subject proves symmetry I'd be
interested in seeing it.  But I would also like to see a map defined in
terms of the modpartitions themselves that gives symmetry in n and k.

William Keith



More information about the SeqFan mailing list