The Duffin--Schaeffer conjecture for systems of linear forms

We extend the Duffin--Schaeffer conjecture to the setting of systems of $m$ linear forms in $n$ variables. That is, we establish a criterion to determine whether, for a given rate of approximation, almost all or almost no $n$-by-$m$ systems of linear forms are approximable at that rate using integer vectors satisfying a natural coprimality condition. When $m=n=1$, this is the classical 1941 Duffin--Schaeffer conjecture, which was proved in 2020 by Koukoulopoulos and Maynard. Pollington and Vaughan proved the higher-dimensional version, where $m>1$ and $n=1$, in 1990. The general statement we prove here was conjectured in 2009 by Beresnevich, Bernik, Dodson, and Velani. For approximations with no coprimality requirement, they also conjectured a generalized version of Catlin's conjecture, and in 2010 Beresnevich and Velani proved the $m>1$ cases of that. Catlin's classical conjecture, where $m=n=1$, follows from the classical Duffin--Schaeffer conjecture. The remaining cases of the generalized version, where $m=1$ and $n>1$, follow from our main result. Finally, through the Mass Transference Principle, our main results imply their Hausdorff measure analogues, which were also conjectured by Beresnevich \emph{et al} (2009).