The Causal Bootstrap: Bounding Smeared Spectral Functions from Non-Perturbative Euclidean Data

TL;DR

The paper introduces the causal bootstrap, a method for bounding smeared spectral observables from finite Euclidean data, using convex optimization and semidefinite programming to provide rigorous bounds.

Contribution

It develops a novel framework that combines convex optimization, duality, and numerical techniques to bound spectral functions from non-perturbative Euclidean data.

Findings

Provides rigorous bounds on spectral observables using the causal bootstrap.

Reduces dual problems to finite-dimensional semidefinite programs for various kernels.

Demonstrates the method with numerical examples.

Abstract

This work introduces the causal bootstrap, a framework for bounding smeared spectral observables from finite non-perturbative Euclidean data. The method optimizes over the convex set of positive spectral densities compatible with the data to compute rigorous upper and lower bounds on the target observable. Statistical uncertainties, including correlations, are incorporated through compatibility regions using the covariance matrix. The bounds are equivalent, via Lagrange duality, to certified bounds on the target smearing kernel. For polynomial, rational, and piecewise rational kernels, the resulting dual problems can be reduced to finite-dimensional semidefinite programs using techniques familiar, e.g., in the numerical conformal bootstrap. The present formulation clarifies the relation to moment problems, Nevanlinna--Pick interpolation, and linear kernel-reconstruction methods. Numerical examples demonstrate the method.