Patterns in Riordan arrays
This repository collects our thesis about Riordan arrays, which is our work to graduate at the University of Florence, defended on October 10, 2015.
Candidate: Massimo Nocentini (massimo.nocentini@unifi.it)
Supervisor: prof. Donatella Merlini (donatella.merlini@unifi.it)
Files
Abstract
This work aims to study a subset of objects belonging to the field of analytic combinatorics, in particular generating functions, formal power series and infinite lower triangular matrices. Many interesting books by Flajolet and Sedgewick [1], Knuth [2] and Graham et al. [3] exist on those topics, where skillful methods to handle sequences of counting numbers, combinatorial sums and classes of combinatorial discrete objects, such as graphs, words, lattice paths and trees, are presented. The concept of Riordan array is the core of the present work. We are interested to show new characterizations to spot some properties of their structure. One example is the $h$-characterization $\mathcal{R}_{h(t)}$ of a Riordan array $\mathcal{R}$ and a second one is the generalization of the $A$-sequence $\lbrace a_{n}\rbrace_{n\in\mathbb{N}}$ and $A$-matrix $\lbrace a_{ij}\rbrace_{i,j\in\mathbb{N}}$ concepts.
Although Riordan group theory has been studied intensively in the recent past, we would like to give an introduction with our words, rethink about the original and introductory papers of this theory, in particular those by Shapiro, who introduces the Riordan group and builds a combinatorial triangle counting non-intersecting paths; by Rogers, who introduces the concept of renewal arrays and finds the important concept of their $A$-sequences; by Eplett, who provides an identity involving determinants and Catalan numbers; and, finally, by Sprugnoli, who uses Riordan arrays in order to find generating functions of combinatorial sums in a constructive way, not just proving that a sum equals a given value (usually denoted by a closed formula).
The other topic of this work is the description and formalization of Riordan arrays under the light of modular arithmetic. We have shown some congruences about Pascal array $\mathcal{P}$ and its inverse $\mathcal{P}^{-1}$. We have also proved a formal characterization for the Catalan array $\mathcal{C}$. These results were presented in a talk contributed at a recent conference held in Lecco [4]. All major researchers involved in Riordan group theory were present and some of them threw some important ideas about possible enhancement of our results.
Finally, we have implemented a subset of Riordan group theory using the Python programming language, on top of Sage mathematical framework. Our implementation is written in pure object-oriented style and allows us to do raw matrix expansion, computing inverse arrays, applying modular transformation, using a set of partition functions, and building LaTeX,code for representing such modular arrays as coloured triangles.
[1] Flajolet and Sedgewick, Analytic Combinatorics, Cambridge University Press, 2009
[2] Knuth, The Art of Computer Programming, vol. 1-3, Addison-Wesley, 1973
[3] Graham, Knuth and Patashnik, Concrete Mathematics, Addison-Wesley, 1994
[4] Second International Symposium on Riordan Arrays and Related Topics, RART2015
RART2015 and paper submission
The content about modular transformations, applied to Pascal and Catalan arrays in particular, was shown in a talk at RART2015; moreover, a paper that collects theorems about the Catalan triangle has been submitted.
Also, we provide an implementation to do matrix expansion of Riordan matrices and computing inverses, symbolically; finally, the code base includes the inductive procedure given in the paper to build the modular Catalan triangle, choosing modulo 2, avoiding the brute force approach; for details and other fractal objects, have a look in this notebook.