Further Evidence for Near-Tsirelson Bell-CHSH Violations in Quantum Field Theory via Haar Wavelets
David Dudal, Ken Vandermeersch
TL;DR
The work reframes near-Tsirelson Bell-CHSH violations in the vacuum of a free massless 1+1D spinor QFT as a spectral problem for a family of symmetric matrices built from Haar-wavelet integrals. Step 1 demonstrates an exact wavelet-based construction that yields explicit test functions achieving correlations close to the bound, with the key link being the maximal eigenvalue of $A(N,K)$ approaching $\pi$ in the large-resolution limit for $K=1$, giving a numerical value $\approx 3.1105202$ for the asymptotic maximum. Step 2 then shows how to replace Haar wavelets by smooth bumpified versions via Planck-taper windows, preserving the necessary inner-product relations to arbitrary precision and producing $C^{\infty}$ test functions compatible with Algebraic QFT. The general case $K\ge1$ is discussed through block Toeplitz and matrix-valued Fourier-series analysis, with numerical evidence suggesting the maximal eigenvalue tends to $\pi$ as the translate count grows, supporting the conjecture that near-maximal Bell-CHSH violations exist in this QFT setting. Together, these results offer a more efficient route to certify near-Tsirelson violations in QFT and illuminate how wavelet-based constructions can yield explicit, smooth test functions in the vacuum sector.
Abstract
This paper investigates a recent construction using bumpified Haar wavelets to demonstrate explicit violations of the Bell-Clauser-Horne-Shimony-Holt inequality within the vacuum state in quantum field theory. The construction was tested for massless spinor fields in $(1+1)$-dimensional Minkowski spacetime and is claimed to achieve violations arbitrarily close to an upper bound known as Tsirelson's bound. We show that this claim can be reduced to a mathematical conjecture involving the maximal eigenvalue of a sequence of symmetric matrices composed of integrals of Haar wavelet products. More precisely, the asymptotic eigenvalue of this sequence should approach $π$. We present a formal argument using a subclass of wavelets, allowing us to reach $3.11052$. Although a complete proof remains elusive, we present further compelling numerical evidence to support it.
