[seqfan] Re: Modular Partitions

Li-yao Xia li-yao.xia at ens.fr
Sat May 3 15:35:39 CEST 2014


There is a link to necklaces when p and n are coprime which generalizes 
William Keith's result.

By giving up primality, we lose the fact that rotation-equivalence of 
profiles partitions them into equal-sized classes (i.e., the object of 
paragraph 3, "If we add 1 to each part,...").

However, having p and n coprime still allows the next paragraph to be 
generalized:

 > (...) 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. (...)

In other words, equivalence classes of profiles are in bijection with 
modular partitions.

Independently of coprimality of p and n, these equivalence classes are 
in fact (in bijection with) necklaces, as is explained next.

A profile describes the number n(t) of parts of size t in a partition, 
thus it is given by a sequence (n(0) ... n(p-1)), where n(t) is 
nonnegative and n(0) + ... + n(p-1) = n (a total of n parts).

E.g., when (p, n) = (5, 4), the partition [0 3 3 4] is associated to
(1 0 0 2 1). (one 0, no 1, no 2, two 3s, one 4)

If we consider such sequences (n(t)) up to cyclic-shifts---which is the 
same as adding a constant to all parts of a partition---they are in fact 
necklaces with p white beads and n black beads, each n(t) indicating the 
number of black beads between two successive white beads.

When p and n are not coprime, that function from partitions to their 
profiles up to rotation is not a bijection. Shifting by p/gcd(n,p) 
transforms a valid p-modular partition into another valid one, possibly 
the same.

E.g., (p, n) = (6, 3), gcd(6, 3) = 3, T(6, 3) = 10

"Profiles" on the left those in the same class are gathered, partitions 
on the right with their sums:

(3 0 0 0 0 0) | [0 0 0]  0
(0 0 3 0 0 0) | [2 2 2]  6
(0 0 0 0 3 0) | [4 4 4] 12

(1 1 0 0 0 1) | [0 1 5]  6
(0 1 1 1 0 0) | [1 2 3]  6
(0 0 0 1 1 1) | [3 4 5] 12

(1 0 0 2 0 0) | [0 3 3]  6
(0 0 1 0 0 2) | [2 5 5] 12
(0 2 0 0 1 0) | [1 1 4]  6

(1 0 1 0 1 0) | [0 2 4]  6

Impossible profiles:

(1 2 0 0 0 0)
(2 1 0 0 0 0)
(1 0 2 0 0 0)
(2 0 1 0 0 0)
(1 1 0 1 0 0)
(1 0 1 1 0 0)

Li-yao Xia

-------- Original Message --------
Subject: [seqfan] Re: Modular Partitions
From: William Keith <william.keith at gmail.com>
To: Sequence Fanatics Discussion list <seqfan at list.seqfan.eu>
Date: 05/03/2014 07:19 AM

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

_______________________________________________

Seqfan Mailing list - http://list.seqfan.eu/





More information about the SeqFan mailing list