Water injection into the subsurface, inherent in improved hydrocarbon recovery and extraction of geothermal energy, often suffers from injectivity decline, even when water carries only nano-sized particles at low concentrations. This study investigates the propagation of such nano-sized particles experimentally and by modelling. Water with dispersed silica nanoparticles of about 140 nm diameter was used as a proxy to ultra-filtered water. Dispersion of the nanoparticles in brine is investigated by varying their concentration, the brine composition, salinity, pH and the presence of iron ions. The measured apparent hydrodynamic size and zeta potential indicate that nanoparticles remain dispersed with the expected size only for salinity below 3000 ppm with pH ranges 6.5 to 8.5. For higher salinity or pH outside that range or presence of iron ions, agglomeration becomes strong. Core flood experiments are conducted on high permeability Bentheimer sandstone, and the transport and retention of nanoparticles in the cores was analysed using multiple pressures measured along the core and by influent/effluent analysis. Core flood results show that stable injectivity can be reached with a good propagation of the nanoparticles through the permeable core with no external filter cake formation, provided the pH and salinity of the injected fluid are kept within the dispersion range and free of iron ions. However, injectivity decline still occurs in three characteristic stages well captured by our mechanistic model used to match the data. This study will contribute to better understanding of the transport dynamics of nanoparticles in the subsurface and to better modelling prediction and assessment of technologies where transport of nanoparticles is key.