Can such long time steps really be used in dissipative particle dynamics simulations?

Commonly used time steps in dissipative particle dynamics (DPD) simulations are too large and lead to systematic errors in the computed properties. The main source of errors is the inaccurate integration of the conservative force. This error can be reduced to some extent by constructing a smoother force without any abrupt change at the cut-off distance, but the improvement is marginal. Alternatively, we tried smooth forces that also lead to the same conclusion. It is possible to find combinations of parameters for the random and dissipative forces that make errors cancel, but the combinations will depend on the system's thermodynamic state and on the particular force model. The only safe procedure is to use small time steps, i.e. comparable with those used in MD simulations. Alternatively, an improved integration algorithm should be used for the conservative force.