Granular flow concerns large assemblages of individual solids, andis fundamentally different from any other type of flow. Common granularflows in nature, such as debris flow, landslides and chutes, are denselypacked with solid fractions, over 50{\%} by volume. The interparticlecontact forces are found not evenly distributed throughout the granularmaterial, but are concentrated in force chains supporting the bulk ofinternal compressive stress. Therefore, contact mechanics of particles anddynamic evolution of force chains have a significant effect on the rheologicalproperties of granular flows. In this review paper, we first introduceHertzian law and Mindlin-Deresiewicz theory for normal and tangentialcontact forces between non-adhesion particles. We then discuss thenew granular flowmap proposed by Campbell (2002; 2005; 2006), based on therelative importance of Bagnold's inertial stress to the elastic stress offorce chains. The limitations of Mohr--Coulomb theory and kinetics forquasistatic and the rapid-flow regimes, respectively, are also discussed. Forintermediate flow regimes, no appropriate theory is proposed so far. Over 30years' development, discrete element methods (DEM), a package of numericaltechniques to model granular materials, may well be used to study granularflows. The potential value of DEM is the ability to obtain information that isnormally inaccessible and to perform rigorous parametric studies. Thesoft-sphere model, one of the widely used DEM models in engineeringapplications, simplifies sophisticated contact mechanics and deformationdetails into an easy-to-use contact model. However, one of tricky problemsarising from this simplification is the calibration of introducedparameters with realistic physical quantities. The hard-sphere model isfurther simplified and has very limited applications so far. We concludethat only the DEM based on rigorous contact mechanics on individualparticles could help the fundamental study on granular flows, such asinternal force chains properties related with constitutive relation, and eventuallybe applicable in natural process.