Why use Naive Dimensional Analysis for amplitude estimates when simpler vertex-counting rules already constrain coupling strength scaling?

This is a sophisticated technical question about methods in quantum field theory amplitude calculations. Let me unpack why naive dimensional analysis remains valuable alongside vertex-counting constraints.

The complementary roles

Vertex-counting rules (like power-counting in renormalizable theories) tell you the *allowed* scaling of coupling constants based on symmetry and dimensionality—they constrain what's possible [AG-2026.02-027]. But they often underdetermine the *actual* numerical coefficients and functional forms that appear in amplitudes.

Naive Dimensional Analysis (NDA) goes further: it uses the dimensional structure of the theory—particle masses, coupling constants, and the energy scale of the process—to estimate not just scaling but actual magnitude of amplitudes. Think of it as asking: "Given only that something must respect dimensions and locality, what's the most natural size it could be?"

Where vertex-counting stops short

Consider a loop amplitude. Vertex-counting tells you whether a diagram is renormalizable, but the amplitude's actual form depends on how internal loops integrate over momentum space. This is where NDA shines: it predicts the characteristic energy scales that will govern the loop integral's behavior.

More concretely, [AG-2025.02-1050] describes how renormalized amplitudes recursively appear as effective couplings in higher-loop diagrams. A vertex-counting rule might say "this coupling is dimensionless, so it doesn't scale with energy," but NDA can predict whether quantum corrections amplify or suppress it—crucial information for understanding when perturbation theory breaks down.

A concrete example

In effective field theories of gravity coupled to matter [AG-2026.03-796], the graviton propagator scales as $1/p^4$ (higher-derivative), and naive dimensional analysis correctly predicts that ultra-Planckian scattering will exhibit universal scaling ($d\sigma/d\Omega \propto 1/s$) even though vertex-counting alone wouldn't pin down that exponent. NDA uses the Planck scale as a dimensional anchor.

Why both?

Vertex-counting is *necessary* (it rules out inconsistent theories), but NDA is *sufficient* for phenomenology—it lets you estimate cross-sections and couplings without computing every Feynman diagram. The two methods are orthogonal: one constrains possibility space, the other navigates it efficiently.

The preprints don't directly compare these two methods, so a fuller pedagogical treatment of their historical interplay would require sources outside this archive.