Does TsT preserve the same number of physical degrees of freedom across both formulations?

# Does TsT Preserve the Same Number of Physical Degrees of Freedom Across Both Formulations?

Your question touches on a subtle but crucial issue in modified gravity theories. The short answer is: not always, and this is a major open problem.

The Core Puzzle

When you reformulate a gravitational theory using teleparallel geometry (TsT, or more precisely the "Teleparallel Equivalent to GR"), you're switching from a curvature-based description to a torsion-based one. In the linear case—pure General Relativity recast in teleparallel form—the two formulations are indeed equivalent and preserve the same 2 physical gravitational degrees of freedom (the two polarizations of a gravitational wave).

However, once you modify the theory non-linearly, agreement breaks down. This is where the confusion deepens.

What Hamiltonian Analysis Shows

Work using the Hamiltonian (constraint-counting) approach on $f(T)$ gravity reveals an uncomfortable fact: [AG-2024.05-450] explains that modified teleparallel theories harbor "hidden extra degrees of freedom that current studies systematically miss." The extra geometric structure—the flat parallel transport—becomes *physical* when you alter GR. The problem is that breaking the gauge symmetry that keeps this structure harmless "is not stably so," leading to strong coupling instabilities.

When researchers perform detailed Hamiltonian analyses of $f(T)$ gravity, [AG-2024.05-450] notes that "results are contradictory," with some analyses yielding far fewer degrees of freedom than others. The linear perturbation treatments often see fewer modes than a full nonperturbative Hamiltonian count predicts—a striking mismatch that suggests we don't yet fully understand what is physical and what is gauge artifact.

The Degree-of-Freedom Counting Problem

Part of the confusion stems from *how* we count degrees of freedom. Recent work [AG-2025.01-981] and [AG-2025.09-408] develops systematic methods grounded in differential geometry and commutative algebra, moving beyond Hamiltonian and Lagrangian recipes. [AG-2025.09-408] applies these techniques to $f(\mathbb{Q})$ gravity (symmetric teleparallel), where standard methods fail, and finds seven propagating degrees of freedom—already a sign that modified teleparallel theories are far richer (and more problematic) than the linear case.

For quadratic nonmetricity theories, [AG-2025.12-739] finds another troubling pattern: Hamiltonian constraint analysis and linear perturbation analysis *disagree*. In some cases, the Hamiltonian yields 8 degrees of freedom while perturbations show only 5 or 6. This strongly suggests that the two formulations do *not* preserve the same count, and that hidden constraints or pathological dynamics emerge depending on how you probe the theory.

Why This Matters

If TsT formulations don't preserve degree-of-freedom content under modification, it means you cannot naively move between curvature and torsion pictures when testing theories observationally. You may be counting gravitons incorrectly, or the theory may be hiding ghost-like instabilities that only show up in certain regimes.

The provided preprints do not definitively answer whether a standard TsT preserves degrees of freedom, but they do establish that this is an open and troubling question for modified teleparallel gravity.