Granular mixtures such as sand and powders tend to segregate or unmix by particle property. The details are important for many natural and industrial applications. While kinetic theory provides a mechanistic framework for modeling segregation in energetically agitated granular mixtures, there is no analogous framework for dense sheared granular mixtures. A number of segregation mechanisms have been identified as important, including those associated with pressure gradients, gravity, gradients in shear rates, and gradients in granular temperature --- the kinetic energy of velocity fluctuations. All likely contribute to segregation in densely sheared systems, though there is no constitutive relationship for mixtures, and the details are difficult to determine. Further, in typical experimental systems designed to study segregation in dense granular flow (such as chutes and rotated drums), gravity, velocity gradients, and porosity gradients coexist in the direction of segregation.
The research in this thesis uses physical and computational experiments to elucidate particular segregation mechanisms in dense granular flow and develop a theoretical model incorporating these segregation mechanisms. Experiments are conducted in a relatively new geometric configuration called split-bottom cell which can isolate shear rate and porosity gradients from gravity. Distinct Element Method (DEM) simulations of experiments in this geometry and in a chute flow provide details inaccessible experimentally including particle concentrations and velocities at every point in space and interparticle forces. These simulations reveal unique dynamics associated with shear-induced segregation in dense systems. Based on the results, a theoretical framework is developed to model segregation associated with shear gradients in dense sheared granular mixtures.