Gaps and approximations in the space of growth functions

TL;DR

This paper demonstrates that not all increasing, submultiplicative functions are realizable as growth functions of finitely generated algebras, revealing gaps in the possible growth behaviors and settling longstanding questions.

Contribution

It proves the existence of arbitrarily rapid increasing submultiplicative functions that cannot be realized as algebra growth functions, resolving open problems in algebraic growth characterization.

Findings

Existence of non-realizable rapid growth functions

Identification of 'holes' in the space of growth functions

Negative solution to a conjecture on graded algebra growth

Abstract

An important problem in combinatorial noncommutative algebra is to characterize the growth functions of finitely generated algebras (equivalently, semigroups, or hereditary languages). The growth function of every finitely generated, infinite-dimensional algebra is increasing and submultiplicative. The question of to what extent these natural necessary conditions are also sufficient -- and in particular, whether they are sufficient at least for sufficiently rapid functions -- was posed and studied by various authors and has attracted a flurry of research. While every increasing and submultiplicative function is realizable as a growth function up to a linear error term, we show that there exist arbitrarily rapid increasing submultiplicative functions which are not equivalent to the growth of any algebra, thus resolving the aforementioned problem and settling a question posed by Zelmanov (and repeated by Alahmadi-Alsulami-Jain-Zelmanov). These can be interpreted as `holes' in the space of growth functions, accumulating to exponential functions in the order topology. We show that there exist monomial algebras and hereditary languages whose growth functions encode the existence of non-prolongable words, and algebras whose growth functions encode the existence of nilpotent ideals (in the graded case). This negatively solves another conjecture of Alahmadi-Alsulami-Jain-Zelmanov in the graded case.