Lower-viscosity fluids are commonly believed to be able to create more complex fractures in hydraulic fracturing, however, the mechanism remains stubbornly unclear. We use a new grain-scale model with accurate coupling of hydrodynamic forces to simulate the propagation of fluid-driven fracturing. The results clarify that fracturing fluid with a lower viscosity does not always create more complex fractures. The heterogeneity in the rock exerts the principal control on systematic evolution of fracture complexity. In homogeneous rock, low viscosity fluids result in low breakdown pressure, but viscosity exerts little influence on fracture complexity. However, in heterogeneous rock, lower viscosity can lead to more complex network of fracturing. A regime map shows the dependence of fracture complexity on the degree of rock heterogeneity where low viscosity fracturing fluid more readily permeates weak defects and creates complex fracture networks.

T-type intersections are commonly observed in natural fracture networks. Their impacts on the connectivity and flow results in complex fracture networks are rarely investigated. In this work, we implement the discrete fracture network method to construct complex fracture networks, denoted as original fracture networks. By implementing the rule-based fracture growth algorithm, we generate the corresponding kinematic fracture networks with a substantial proportion of T-type intersections. The connectivity and flow results of both the single-phase and two-phase flow simulations in these two types of fracture networks are systematically investigated. The results show that kinematic fracture networks tend to connect more fractures with fewer intersections and yield better connectivity than the original ones. Most kinematic fracture networks have larger permeability in the single-phase flow simulation and higher cumulative gas production in the two-phase flow simulation than original fracture networks under the same boundary conditions. The proportions of permeability and production enhancement are 68\% and 77\%, respectively. Flow results, like the permeability and production, have strong positive correlations with the connectivity of the fracture networks, but they are nonequivalent and strongly impacted by the number of inlets and outlets.