a–d, Reflectance contrast spectra differentiated with respect to photon energy as a function of displacement field D at a filling factor of ν = −2/3 (a) and ν = −1 (b–d). The spectra were acquired under B = 50 mT in two circular polarizations (σ+: top panels, σ−: bottom panels) using a weak, linearly polarized broadband white-light excitation. Horizontal lines indicate the displacement fields at which the optical switching efficiency is investigated below. e–g, Degree of spin polarization of the AP resonance before (grey circles) and after (coloured circles) illuminating the sample with a narrow-bandwidth (<1 meV FWHM) pump beam as a function of the pump centre wavelength. The wavelength dependence is investigated at the filling factors and displacements fields indicated in the respective reflectance contrast plots above. The pump powers are 90 nW (e), 80 nW (f), 65 nW (g) and 50 nW (h), respectively. In each graph, the spins are first oriented using B = 50 mT, which is then turned off. The spin polarization before and after optical pumping is examined by measuring the helicity-resolved reflection of a linearly polarized white-light probe beam. We pump with σ− (blue circles) and σ+ (red circles) circularly as well as with linearly polarized light (pink circles in g). The spin polarization is inferred by fitting double Lorentzian spectral profiles to the optical resonances in the σ+-polarized and σ−-polarized reflection contrast spectra. The fit parameter corresponding to the amplitude of the stronger Lorentzian is constrained from below and above to reduce fitting uncertainty and an amplitude of zero is assigned to a spectrum when the χ2 of the fit exceeds a global tolerance, that is, when no Lorentzian with a finite amplitude can sensibly be fitted.