Euclid: Precise inference of defocus wavefront error from image diffraction spike measurements
The success of the Euclid cosmological weak lensing measurements requires unprecedented knowledge of the point spread function (PSF) shape. Defocus wavefront errors induce variations in PSF size that directly bias inferred galaxy shapes. We present a rapid, model-independent estimator of the Euclid visible instrument defocus based on measuring subpixel shifts in diffraction spikes from bright stars, arising from the non-mirror-symmetric placement of the telescope spiders. This estimate requires no a-priori PSF model and can be directly inferred from individual survey exposures within seconds. We express defocus as the secondary mirror displacement along the optical axis, $Δz$, achieving a per-exposure precision of $σ(Δz) = 0.022\,μ\text{m}$ in standard Euclid wide-survey images, corresponding to a peak-to-valley optical-path difference of $0.75\,\text{nm}$. This sensitivity resolves thermally induced shifts, typically 0.1-0.3$\,μ\text{m}$, and enables tracking of temporal evolution and field-of-view variations when combining exposures within periods of stability. Since July 2024, the Euclid telescope has been exceptionally stable, while changes to defocus are effectively homogeneous across the field of view. By correlating defocus with PSF size, we find that the Euclid DR3 requirement of a fractional PSF size bias of $\left|ΔR_\mathrm{PSF}^2/R_\mathrm{PSF}^2\right|<10^{-3}$ corresponds to field-averaged secondary mirror displacements exceeding $\langle Δz_\mathrm{thr}\rangle = (0.0683 \pm 0.0024)\,μ\text{m}$. Our estimator provides a robust real-time monitoring tool, supplies a stringent prior for computationally intensive PSF model fitting, and is applicable not only to the Euclid telescope, but to other telescopes whose spider vanes are not mirror-symmetrically arranged.