Enumerative combinatorics related to partition shapes

Detta är en avhandling från Stockholm : KTH

Sammanfattning: This thesis deals with enumerative combinatorics applied to three different objects related to partition shapes, namely tableaux, restricted words, and Bruhat intervals. The main scientific contributions are the following.Paper I: Let the sign of a standard Young tableau be the sign of the permutation you get by reading it row by row from left to right, like a book. A conjecture by Richard Stanley says that the sum of the signs of all SYTs with n squares is 2^[n/2]. We prove a generalisation of this conjecture using the Robinson-Schensted correspondence and a new concept called chess tableaux. The proof is built on a remarkably simple relation between the sign of a permutation pi and the signs of its RS-corresponding tableaux P and Q, namely sgn(pi) = (?1)^v sgn(P)sgn(Q), where v is the number of disjoint vertical dominoes that fit in the partition shape of P and Q. The sign-imbalance of a partition shape is defined as the sum of the signs of all standard Young tableaux of that shape. As a further application of the sign-transferring formula above, we also prove a sharpening of another conjecture by Stanley concerning weighted sums of squares of sign-imbalances.Paper II: We generalise some of the results in paper I to skew tableaux. More precisely, we examine how the sign property is transferred by the skew Robinson-Schensted correspondence invented by Sagan and Stanley. The result is a surprisingly simple generalisation of the ordinary non-skew formula above. As an application, we find vanishing weighted sums of squares of sign-imbalances, thereby generalising a variant of Stanley’s second conjecture.Paper III: The following special case of a conjecture by Loehr and Warrington was proved by Ekhad, Vatter, and Zeilberger: There are 10^n zero-sum words of length 5n in the alphabet {+3,?2} such that no consecutive subword begins with +3, ends with ?2, and sums to ?2. We give a simple bijective proof of the conjecture in its original and more general setting where 3 and 2 are replaced by any relatively prime positive integers a and b, 10^n is replaced by ((a+b) choose a)^n, and 5n is replaced by (a+b)n. To do this we reformulate the problem in terms of cylindrical lattice walks which can be interpreted as the south-east border of certain partition shapes.Paper IV: We characterise the permutations pi such that the elements in the closed lower Bruhat interval [id,pi] of the symmetric group correspond to non-capturing rook configurations on a skew Ferrers board. These intervals turn out to be exactly those whose flag manifolds are defined by inclusions, as defined by Gasharov and Reiner. The characterisation connects Poincaré polynomials (rank-generating functions) of Bruhat intervals with q-rook polynomials, and we are able to compute the Poincaré polynomial of some particularly interesting intervals in the finite Weyl groups A_n and B_n. The expressions involve q-Stirling numbers of the second kind, and for the group A_n putting q = 1 yields the poly-Bernoulli numbers defined by Kaneko.