Abstract The propped fracture dimension and fracture conductivity are controlled by the proppant placement during the hydraulic fracturing stimulation treatment. Proppant placement has traditionally been studied by flowing proppant slurries between smooth parallel plates, which is a simplification of rough fracture walls in the subsurface. In this study, for the first time we propose a simulation workflow to study proppant transport in realistic rough-walled rock fractures to optimize the proppant placement efficiency. The simulation domain was extracted from a micro-CT image of an induced fracture in a sandstone core. The fracture has many contact points (zero aperture) and all apertures on the order of a millimeter and less. Taking advantage of image processing technique, two fracture surfaces were separated and moved to obtain a wider fracture and to mimic the coupling effect of geomechanics and proppant transport. Proppant transport in the rough-walled rock fractures was simulated by coupling computational fluid dynamics with a discrete element model (CFD-DEM) and this is the first time this was done in a detailed image-based fracture geometry for an extensive range of proppant and Newtonian fluid parameters as well as non-Newtonian (shear-thinning). Proppant placement, proppant concentration profile along the flow direction, pressure difference build-up across the fracture and total proppant volume retained within the fracture are quantified under various flow conditions. Two fracture orientations including horizontal fracture without gravity influence and vertical fracture with gravity influence are investigated in this study. Proppant placement visualization shows that proppant particles distribute quite uniformly over the horizontal fracture without considering gravity effect, while for the vertical fracture an obvious proppant settling bank is observed at the bottom of the fracture except when using very viscous or non-Newtonian fluids. For the case of horizontal fracture without gravity influence, proppant placement inside the rough fracture is mainly controlled by the fracture aperture field, and the proppant size plays a crucial role in proppant placement. Fracturing fluid viscosity has to get rather high (100cP) for the fluid drag force to have a larger effect in preventing particle accumulation within the fracture and carry the proppant deeper into the fracture. The effect of fracturing fluid viscosity is much more prominent in the vertical fracture case with gravity effect. The cross-linked gel (modeled as a shear-thinning fluid with apparent viscosity between 10cP and 100cP in this study) can achieve a quite spatially uniform proppant placement. The horizontal fracture case with a higher proppant concentration shows a quite clear proppant displacement front, a slower proppant displacement speed into the fracture and a larger pressure difference build-up across the fracture. Larger fracture aperture can induce larger retained proppant volume inside. Less proppant settling and further proppant transport into the fracture are observed for a lighter proppant. For vertical fracture case a larger fluid injection velocity carries the proppant way deeper into the fracture and also gives a more even proppant placement within the fracture, while the influence of injection velocity for the horizontal fracture case is quite complicated. This study presents the first 3D simulation that tracks detailed, dynamic proppant placement in realistic rough-walled rock fractures under various flow conditions. Results from this study provide a completion engineer with directional guidance for proppant and fracturing fluid selection to optimize final proppant placement in the complex fracture networks.