How does the effective dimensionality reduction at high energies compare to traditional renormalization group flow?

# Effective Dimensionality Reduction vs. Traditional RG Flow

Your question touches on a subtle but important distinction that the archive illuminates well.

Traditional RG flow typically describes how coupling constants evolve smoothly as you change the energy scale—the beta functions governing this flow are usually monotonic, and dimensions of operators change gradually. This is the workhorse of renormalization: as you run from UV to IR, interactions stay in the same spacetime dimensions.

Asymmetric RG flow to lower dimensions, by contrast, is qualitatively different. A $d$-dimensional UV theory can flow to an effective theory living in genuinely *fewer* spatial dimensions in the IR [AG-2025.04-1066]. The holographic dual picture is striking: a $(d+1)$-dimensional AdS black hole has a $(d+1)$-dimensional AdS geometry in the far regions but "pinches down" to AdS$_2 \times R^{d-1}$ near the horizon. The field theory dual shows rapid suppression of correlations along $(d-1)$ spatial directions, leaving behind an effective conformal quantum mechanics—a 0+1 dimensional theory—that still encodes the full quantum information of the original system [AG-2025.04-1066].

This matters because it's *not* just a relabeling: the IR conformal dimensions of operators crucially depend on UV details in a way that traditional RG perturbation theory would miss [AG-2025.04-1066].

Another striking comparison: in far-from-equilibrium states with power-law occupation numbers, RG flow becomes truly asymmetric—a parameter $γ > 0$ can *lower* operator dimensions, transmuting marginal interactions into relevant ones [AG-2025.04-1021]. This is the opposite of the usual perturbative picture where dimensions are fixed by the spacetime dimension itself.

The archive provides good evidence for the first scenario but doesn't directly compare effective dimensionality reduction quantitatively against traditional beta-function methods.