Mahler functions are special mathematical objects defined by functional equations where the input is repeatedly raised to some integer power. For example, a Mahler function might satisfy an equation relating its value at a point x to its value at x-squared. These functions appear naturally in computer science and number theory, particularly in the study of automatic sequences, which are sequences generated by simple computational rules. The paper extends the classical theory of Mahler functions from one variable to several variables simultaneously, which is considerably more difficult and less well understood.
The authors establish several foundational results about where and how these multivariate Mahler functions are well-behaved analytically. They show that the functions are always nicely behaved near the origin and extend smoothly throughout certain natural domains, both over the ordinary complex numbers and over p-adic numbers (which are alternative number systems built around a prime p). Using this analytic control, they prove a clean dichotomy: any such Mahler function is either a simple rational function or is transcendental, meaning it cannot satisfy any polynomial equation with rational coefficients. There is no messy middle ground.
The most practically significant result is a strengthened version of a theorem called the lifting theorem, originally due to Adamczewski and Faverjon. This theorem provides a systematic way to figure out all the algebraic relationships between the numerical values that multivariate Mahler functions take at specific points, given that you already know the algebraic relationships between the functions themselves. This is valuable because determining what algebraic relations hold among special values of functions is a central and generally very hard problem in transcendence theory. The paper also proves a multivariate descent theorem, which allows certain technical arguments to be reduced from many variables down to fewer, at points where the functions behave particularly nicely.