Extended Data Fig. 1: Reactor Flux Systematic Uncertainties on the Signal Normalisation.

The 1σ uncertainty stands for 68% frequentist probability. Both rate and shape flux uncertainties are treated via covariance matrices as predicted by the data-driven reactor flux model^31,32,36 used by Double Chooz. The Bugey4 experiment provides an independent rate constraint via its 〈σ_f〉 and therefore extra precision via the cancellation of the common spectrum terms. The uncertainty coming from the reactor–detector baselines is negligible (< 0.01%). The unknown inter-reactor correlations are assumed to be correlated for the reactor power (P_th) and the fission fractions (α_f) in the single detector (SD) case and uncorrelated for any multi detector (MD) configuration in general (combined uncertainty of P_th and α_f is 0.83%). These assumptions are made to minimise the θ₁₃ sensitivity to be conservative. In the Double Chooz case with two reactors the uncertainties on P_th and α_f are reduced by about a factor of \(\sqrt{2}\). Only the uncorrelated terms are relevant for the specific MD case in Double Chooz (ND/FD-I and ND/FD-II).