There has been plenty of observational evidence of cluster galaxy evolution, such as the Butcher-Oemler effect and the morphology-density relation. However, it is hard to identify the origin of cluster galaxy evolution, simply because it is difficult to trace the complicated process of galaxy evolution over a time-scale of several gigayears using observations that only give information about a single epoch. Here we show that gravitational interaction/friction between galaxies is the statistically dominant physical mechanism responsible for cluster galaxy evolution, and that the well-favoured ram-pressure stripping by the cluster gas is not statistically driving cluster galaxy evolution. We have constructed the largest composite cluster with 14 548 member galaxies out of 335 clusters with sigma > 300 km s(-1) carefully selected from the Sloan Digital Sky Survey. By measuring the velocity dispersions of various subsamples of galaxies in this composite cluster, we found that bright cluster galaxies (M-z < -23) have significantly smaller velocity dispersion than faint galaxies (M-z >= -23), with much greater precision than previous results (e.g. those of Adami et A). We interpret this as direct evidence of the dynamical interaction/friction between cluster galaxies, where massive galaxies lose their velocity through energy equipartition during the dynamical interaction/friction with less massive galaxies. We also found that star-forming late-type galaxies have a larger velocity dispersion than passive late-type galaxies. This is inconsistent with the ram-pressure stripping model; since ram pressure is proportional to sigma v(2) (i.e. stronger for galaxies with high velocity), ram-pressure stripping cannot explain the observed trends of passive (evolved) galaxies having low velocity rather than high velocity. On the other hand, the result is again consistent with the dynamical galaxy-galaxy interaction/friction, where more evolved (passive) galaxies lose their velocity through dynamical interaction/friction.
