KiDS-1000: Detection of deviations from a purely cold dark matter power spectrum with tomographic weak gravitational lensing

Model uncertainties in the non-linear structure growth limit current probes of cosmological parameters. To shed more light on the physics of non-linear scales, we reconstructed the finely binned three-dimensional power-spectrum from lensing data of the Kilo-Degree Survey (KiDS), relying solely on the background cosmology, the source redshift distributions, and the intrinsic alignment (IA) amplitude of sources (and their uncertainties). The adopted Tikhonov regularisation stabilises the deprojection, enabling a Bayesian reconstruction in separate $z$-bins. Following a detailed description of the algorithm and performance tests with mock data, we present our results for the power spectrum as relative deviations from a $\Lambda\rm CDM$ reference spectrum that includes only structure growth by cold dark matter. Averaged over the full range $z\lesssim1$, a \emph{Planck}-consistent reference then requires a significant suppression on non-linear scales, $k=0.05$--$10\,h\,\rm Mpc^{-1}$, of up to $20\%$--$30\%$ to match KiDS-1000 ($68\%$ credible interval, CI). Conversely, a reference with a lower $S_8\approx0.73$ avoids suppression and matches the KiDS-1000 spectrum within a $20\%$ tolerance. When resolved into three $z$-bins, however, and regardless of the reference, we detect structure growth only in the range $z\approx0.4$--$0.13$, but not in the range $z\approx0.7$--$0.4$. This could indicate spurious systematic errors in KiDS-1000, inaccuracies in the intrinsic alignment (IA) model, or potentially a non-standard cosmological model with delayed structure growth. In the near future, analysing data from Stage IV surveys with our algorithm promises a substantially more precise reconstruction of the power spectrum.