Extended Data Fig. 1: Performance comparison of 3D-SMLM, 4Pi-SMLM, and 4Pi-SIMFLUX.

a-c, Side-by-side comparison of localization precision for 3D-SMLM (a), 4Pi-SMLM (b), and 4Pi-SIMFLUX (c). d, Localization precision of 4Pi-SIMFLUX at the focal plane with different photon numbers. The black dashed line denotes 1 nm localization precision. All three simulation experiments were performed using a vectorial PSF model with identical parameters: objective NA = 1.35; medium refractive index = 1.406; emission wavelength = 600 nm; pixel size = 120 nm. For 3D localization and phase-wrapping, an astigmatism of 60 nm in amplitude (root mean square error) was introduced to the pupil function to generate PSFs with elliptical patterns. For 3D-SMLM, the photon count was 4,000 per image with a background of 12 photons per pixel. For 4Pi-SMLM, the photon count was 2,000 per phase image (total of 8,000 photons) with a background of 24 photons per pixel. For 4Pi-SIMFLUX, the 2,000 photons in each phase image were distributed among the six sub-images according to the corresponding excitation patterns (total of 8,000 photons) with a background of 24 photons per pixel. Only Poisson noise was considered. The 4Pi-SIMFLUX illumination pattern was simulated with a modulation depth of 0.9 and an interference period of 210 nm, mimicking experimental conditions. For each z position in all images, 50,000 molecules were simulated. Localization precision was calculated as the standard deviation of the estimated positions.