We demonstrate the effectiveness of a hybrid Monte Carlo (MC)/molecular dynamics (MD) method in analysis of grain-boundary segregation, a long-time phenomenon from the atomistic-simulation point of view. The Σ9 coincidence grain boundary of Si crystal is adopted as a testbed for the method. We find an efficient sampling of hopping diffusion of a vacancy in the system can be achieved, while the internal energy evaluated from the MD runs at the same temperature as the MC simulation includes the structural relaxation of the system at the temperature.