How does observability of extra geometric structure determine whether TsT is a true duality or reparametrization?

This is a sophisticated question about the distinction between true dualities and mere reparametrizations, using TsT (a string theory transformation) as the test case. Let me unpack what the archive says about this criterion.

The Observability Principle

The key insight comes from work on geometric reformulations of gravity. When we have two formulations of a theory—say, general relativity (GR) versus teleparallel gravity—the question "is this a duality or a reparametrization?" hinges on whether extra structure introduced in one formulation is *observable* (physically real) or merely *auxiliary* (pure gauge) [AG-2024.11-413, AG-2024.05-450].

In GR, gravity emerges from spacetime *curvature* alone. Teleparallel gravity reformulates the same equations using *torsion* instead, but on a flat spacetime. The catch: teleparallel gravity posits strictly more geometric structure—a flat parallel transport—than GR does [AG-2024.06-418]. However, in models equivalent to GR, this extra structure is "absolutely abstract and unpredictable" (pure gauge), meaning it has no observable consequences [AG-2024.11-413].

The distinction becomes sharp when you *modify* the theory. In modified teleparallel gravity, the extra geometric structure becomes *physical*—it influences the equations of motion—which is why those theories genuinely predict new physics [AG-2024.05-450]. A true duality preserves physics; a reparametrization that masks unobservable scaffolding does not.

Applied to TsT

For TsT (T-duality followed by a shift and T-duality again), the criterion would be analogous: does TsT introduce extra geometric structure whose effects are observable in principle, or does it merely re-express the same physical content?

The archive discusses T-duality in several contexts. One approach uses *categorical equivalence* and *functor maps* between geometric objects [AG-2024.02-714]. Another extends T-duality to non-Abelian settings (transgressive fibrations) and shows that T-duals preserve *twisted cohomology* and *Clifford–Courant algebroids*—invariants that capture observable structure [AG-2025.03-1203]. If TsT preserves such topological and algebraic invariants, it acts as a true duality; if it only shuffles coordinates or unobservable degrees of freedom, it's a reparametrization.

---

Caveat: The archive does not directly address TsT transformations in string theory or discuss criteria for when TsT counts as a duality versus a reparametrization, so the reasoning above extrapolates from the teleparallel gravity literature.