The Unitary Architecture of Renormalization

TL;DR

This work develops an on-shell bootstrap for renormalization by enforcing unitarity on massless 4D scalar theories, showing that all-loop recursion relations among scattering amplitudes mirror the renormalization group identities. By comparing unitarity-derived recursions with those from the Callan–Symanzik equation, the authors demonstrate a deep equivalence between on-shell unitarity constraints and RG flow, up to subleading logarithmic order, once the appropriate initial data (β and γ) are matched. The large-N O(N) warm-up clarifies the essential structure, and the λφ^4 analysis extends it to full multi-channel dynamics, including Steinmann-like constraints and phase-space-enhanced cuts. The results yield explicit closed-form expressions for leading and subleading amplitude coefficients, connect renormalization to on-shell data, and suggest a path to deriving RG data purely from unitarity, with potential applications to more realistic theories and EFTs.

Abstract

We set up a bootstrap problem for renormalization. Working in the massless four-dimensional O$(N)$ model and the $λφ^4$ theory, we prove that unitarity leads to all-loop recursion relations between coefficients of scattering amplitudes with different multiplicities. These turn out to be equivalent to the identities imposed by renormalization of the coupling and the wavefunction through subleading logarithmic order, except with different initial conditions. Matching the initial conditions thus fixes the beta function and wavefunction anomalous dimension to these orders. We explain how to connect this new on-shell renormalization picture with the standard renormalized perturbation theory, highlighting a rich interplay between finiteness, dimensional regularization, and unitarity cuts.

Figures (6)

• Figure 1: The unitary architecture of renormalization. Each row represents a given order in perturbation theory, while each column indicates the logarithmic order. The directions of the arrows denote proven (orange) and conjectured (gray) relationships between amplitude coefficients $m_{L,k}$. Note that $m_{0,0}=1$ and all other $m_{L,0} = 0$ in the on-shell scheme. Unitarity and renormalization give similar recursion relations, except for the initial conditions. Left: Unitarity fixes each column solely in terms of the $m_{L,0}$ as initial conditions, see \ref{['fig:results']} for the concrete realization. Right: Renormalization fixes each column in terms of $\beta_L$, $\gamma_L$, and $m_{L,0}$ as initial conditions, see \ref{['fig:resultsRG2']} and \ref{['fig:resultsRG']} for the concrete realization.
• Figure 2: Left and right angles. This is the 2-cut for a 4-particle amplitude. The kinematic invariants for the left and right amplitude can be expressed in terms of two angles, $\theta_L$ and $\theta_R$, which have the geometric interpretation of being the angle between $\vb{p}_2$ and $\vb*{{\ell}}_1$ and $\vb{p}_3$ and $\vb*{{\ell}}_1$ respectively.
• Figure 3: Summary of the unitarity constraints. The first values of $m_{L,k}$ and $n_{L,k}$ through subleading order are given in terms of $m_{1,0}$, which is the only initial condition (scheme dependence). The arrow pointing for the 4-particle coefficients to the 2-particle coefficients indicates that, at subleading order, the 2-particle coefficients are constrained in terms of 4-particle coefficients by the unitarity equation. The analogous recursions coming from renormalization are illustrated in \ref{['fig:resultsRG2']} and \ref{['fig:resultsRG']}.
• Figure 4: Summary of the renormalization constraints on $n_{L,k}$. The first few coefficients are displayed in terms of their dependence on the initial conditions. This table is to be compared with \ref{['fig:results']} (right).
• Figure 5: Summary of the renormalization constraints on $m_{L,k}$. The first few coefficients are displayed in terms of their dependence on the initial conditions. This table is to be compared with \ref{['fig:results']} (left).