Date: Mon, 28 Oct 1996 14:41:26 +0100 (MET)
From: "Dmitrii V. Pasechnik" 
To: group-pub-forum@maths.bath.ac.uk
Subject: Poincare series for depths of positive roots?
 
Dear Forum,
 
The depth of a positive root  a in a root system F related to a Coxeter group G
is the minimal length of g in G such that the root  g a  is negative.
(Equivalently, if the refelection r_a has length l then depth a = (l+1)/2)
 
We conjecture that the function P(t)=a_1 t + a_2 t^2 + a_3 t^3 +..., where
a_i is the number of roots of depth i,
is always rational.
 
Some (computational) evidence to support this.
 
First,
For affine Coxeter groups a_i is a periodic function (of i) with a_i determined
in a simple way from a_i's for the associated finite group.
For instance, for affine A_n all a_i=n.
(this is not (yet) a theorem, but we suspect it's not so difficult to show.)
 
 
Second,
I give a (conjectural) recurrence relation for each case, and in the 1st case I
also write down the generating function.
(The latter is just a straightforward computation given the
recurrence. One could also compute a(n) as a function of n, although
it wouldn't look too nice.)
 
Note that for the sequence to satisfy a recurrence relation is
equivalent to have a rational generating function, cf. e.g.
R.Stanley "Enumerative Combinatorics I", Wadsworth 1986, Chapter 4.
 
1) o--o    n: 1   2   3    4    5    6    7    8    9   10   11 
   |\/|   2*( 2   3   6   12   27   60  138  315  726 1668 3843....)
   |/\|
   o--o      a(n)=2a(n-1)+2a(n-2)-3a(n-3).
 
        F(x)=\sum a(n) x^n=2x((2-x-10x^2)/(1-2x-2x^2+3x^3)).
 
2) o--o       1   2   3    4    5    6    7    8    9   10   11 
   | /|       4   5   8   13   24   44   83  158  303  582 1120
   |/ |
   o--o      a(n)=2a(n-1)-a(n-5).
 
3) o--o       1   2   3    4    5    6    7    8    9   10   11 
   | /        4   4   5    6    8   11   15   21   30   43   62
   |/ 
   o--o      a(n)=2a(n-1)-a(n-2)+a(n-3)-a(n-4) (for n>5).
 
 
4)
 
o
|\
| o---o---o
|/
o
 
The depths are as follows:
[ 5, 5, 6, 8, 11, 16, 25, 38, 59, 93, 148, 235, 376, 602, 966, 1550, 2491, 
  4003, 6436, 10348, 16643, 26766, 43052, 69247 ]
 
The recurrence is as follows:
                               
a(n+1)=\sum_{i=n-11}^n v(i)*a(i), for v=[0,0,-1,-1,-2,-1,0,2,2,1,0].
 
 
 
I would appreciate receiving any comments on this.
(please copy your reply to my email dima@win.tue.nl)

Dmitrii V. Pasechnik
Department of Mathematics
Eindhoven University of Technology
PO Box 513, 5600 MB Eindhoven
The Netherlands
e-mail: dima@win.tue.nl
http://www.can.nl/~pasec