Linear-growth harmonic functions for nonsymmetric random walks on groups of polynomial growth
Let $G$ be a finitely generated group of polynomial growth and let $μ$ be an adapted, Abelian-centered probability measure with a finite exponential moment (not necessarily symmetric or finitely supported). We prove that every $μ$-harmonic function of at most linear growth is globally Lipschitz. In particular, if $G$ is nilpotent, these functions are exactly the affine characters: $$ \mathrm{HF}_1(G,μ)=\operatorname{LHF}(G,μ)=P^1(G). $$ The analytic machinery required for this is a convolution-gradient estimate that is uniform over families of Abelian-centered, finitely supported probability measures satisfying fixed ellipticity and exponential-moment bounds. A quantitative induction-restriction theorem transfers $\operatorname{LHF}$ across finite-index subgroups. We also determine the normed structure of these spaces. In particular, on a nilpotent group, the Lipschitz seminorm of an affine character is exactly the dual stable norm of the word metric. Under the same nonsymmetric hypotheses, linear-growth harmonic functions modulo constants identify canonically with the virtual first cohomology. This identification is an isometry for the asymptotic Lipschitz seminorm. For the ordinary Lipschitz seminorm it is a contraction with bounded inverse; the inverse bound seems to depend on $μ$, as shown by an infinite dihedral group example.