Fig 1.
Comparisons of CG potentials (a) and RDFs (b) for bulk water.
All methods fit the target RDF sufficiently well and lead to potential of similar shape. The relative entropy method has a slightly smaller error due the fact that more parameter are available in the potential form.
Table 1.
Simulated water-methanol mixtures.
Fig 2.
Comparisons of the CG potentials and RDFs of methanol-water: Mixture at different methanol mole fractions, X = 0.062,0.5,0.938 are shown.
Arrow indicates the direction of increasing X. All methods fit the target RDF very well for all mole fractions.
Table 2.
Average computational cost per iteration step (in seconds) for simplex and relative entropy-based coarse-graining.
For the simplex methods the number in the bracket denote the time spend in energy minimization before the actual molecular dynamics part.
Fig 3.
Computation times for the construction of neighbor lists: simple vs. grid search.
Shown are the results for the Lennard-Jones fluid as a function of particles in the system. In addition to the data points for the simple and grid search algorithms, lines indicate the scaling law with 2 and 1 as the exponent, respectively. These exponents result from the cost of the simple and grid search algorithm: 𝓞(N2) and 𝓞(N), respectively. The cut-off for the neighbor search was set to 1.6σ, which roughly corresponds to the first minimum in the radial distribution function.
Fig 4.
Absolute computation time for the radial distribution function calculation as a function of threads.
The small system holds 5324, the medium 17687 and the big system 60132 particles. The dashed line shows the ideal scaling line. Also shown are the results from the script-based parallelization of multi_g_rdf.