DPD is the method behind most of my thesis work on block copolymer coatings. It is coarse-grained MD with very soft beads and a thermostat that conserves momentum. Because of this, it reaches the length and time scales where polymers self-assemble, while it still keeps hydrodynamics.
Every pair of beads closer than the cutoff rc feels three forces along the line between them.
Because the energy stays finite even when beads overlap, DPD can use a time step around 0.03 to 0.05 in reduced units, far larger than with LJ.
In DPD, a water bead usually contains Nm water molecules. With the standard density of ρ = 3 beads per rc3, this sets the real cutoff length, and matching the compressibility of water sets the repulsion between like beads.
Try: three water molecules per bead give an rc close to 6.5 Å, which is a common choice.
Groot and Warren showed that the excess repulsion between unlike beads maps onto the Flory-Huggins χ parameter. This is how chemistry enters a DPD model. The χ parameter can come from experiments, solubility parameters, or all-atom simulations, and project 13 learns it from MD.
Try: below a χ of about 2, the single beads stay mixed, and above it, they demix. For polymers, the threshold is χN, so long chains separate at a much smaller χ.
The DPD friction acts between pairs, so momentum is exchanged but never lost. A flow decays slowly through viscosity, as in a real liquid. In contrast, a Langevin thermostat drags every bead against a fixed background, so flows die almost at once, and hydrodynamics is lost.
For self-assembly and phase separation, this changes the speed and even the path of structure formation.
Try: compare the decay of the flow and the total momentum with DPD and with Langevin.
If you bond A and B beads into chains, the system cannot separate fully. Instead, it forms stripes (lamellae), cylinders, or spheres, depending on the block ratio and on χN. This is called microphase separation.
Try: change the number of A beads per chain to make the blocks unequal.
This is the system from my thesis, simplified to 2D. Amphiphilic di-block copolymers in water sit above a hydrophilic surface, such as cellulose or starch. As the water evaporates, the hydrophilic blocks attach to the substrate, and the hydrophobic blocks form a layer above them.
The simulation here is standard DPD in 2D, with the soft repulsions, friction, and random forces of Groot and Warren. Evaporation is kept simple. Now and then, the highest water bead is taken out, and the top of the box moves down, so the density stays at 3. There is no vapor phase, so no many-body DPD is needed.
In 3D and with more realistic parameters, this process decides whether the coating is continuous or full of defects, and therefore how well it blocks water.
Try: change the hydrophobic block length, and stop the evaporation to compare a wet and a dry film.