This paper explains how simple geometric features of certain string compactifications produce discrete flavor symmetries. The authors study symmetric heterotic orbifolds with non‑Abelian point groups. Their goal is to find the moduli‑independent discrete symmetries that act on matter fields and that could constrain particle masses and mixings.

To do this they develop a set of geometric and algebraic tools. One ingredient is the Abelianization of the orbifold space group. Abelianization captures the phase factors that appear in string coupling selection rules and so gives a set of discrete charges for matter states. A second ingredient is the set of geometric affine transformations that permute equivalent localized states. The authors show these must normalize the space group and also preserve the metric and the antisymmetric tensor (the B‑field). A third ingredient are outer automorphisms of the space group that act like some target‑space modular transformations but leave the compactification shape unchanged. Taking the multiplicative closure of these elements builds the traditional, non‑Abelian flavor group.

The paper works through the complications that appear when point groups are non‑Abelian. In that case twisted sectors are labeled by conjugacy classes rather than single elements, and rotations combined with fractional translations (so‑called roto‑translations) affect the charges. The authors present an algorithm and apply their methods to all 331 non‑Abelian six‑dimensional affine geometries that can give N=1 supersymmetric heterotic compactifications in four dimensions. Appendices and catalogs list the resulting symmetries and charges for these geometries.

Why this matters: discrete flavor symmetries can restrict the allowed Yukawa couplings in particle physics and so help explain patterns in fermion masses and mixing. Finding such symmetries from the compactification geometry ties those model‑building ingredients to string theory. The methods here also let one derive these moduli‑independent symmetries without a separate conformal‑field‑theory computation.