Dear Colleague,
Three bugs have been discovered in the Dissipative Particle Dynamics (DPD) code
of DL_MESO 2.6 and two of its post-processing utilities.
Switching off pairwise thermostatting across shearing boundaries (see INFOMAIL
24) has been shown to break spatial homogeneity and thus Galilean invariance. To
ensure this does not happen while still preventing overly high dissipative
forces close to shearing boudnaries, the proposed solution is to modify the
relative velocities of particle pairs to take the boundary velocities into
account. (Many thanks are due to Xiaocheng Shang at the University of Edinburgh
for bringing this to our attention and outlining the solution.)
For systems with charges, contributions to one of the non-symmetric pressure
tensor components from the reciprocal space part of the Ewald sum are not
correctly accumulated. While this means the xz- and zx-components of the tensor
are incorrectly recorded in the CORREL file, it has no effect on the accuracy of
simulations as it does not affect calculations of reciprocal space contributions
to forces or energy. (Many thanks are due to Jony Castagna at STFC Daresbury
Laboratory for bringing this bug to our attention.)
When using the post-processing utilities for calculating radial distribution
functions, rdf.f90 and rdfmol.f90, the search algorithm for finding particle
pairs and the distances between them may encounter the same particle pairs
multiple times if the maximum distance is more than a third of a system
dimension. (Many thanks are due to Silvia Chiacchiera at STFC Daresbury
Laboratory for identifying this bug.)
While the bug calculating pressure tensor contributions in reciprocal space can
be solved by replacing a single line of code, the other two bugs require more
extensive changes to both the main DPD code and two of the post-processing
utilities. Because the implementations of the Ewald sum and thermostatting
across Lees-Edwards boundaries have changed, the Polyelectrolyte and ShearFlow
test cases also need to be re-run and the output files replaced.
We therefore strongly recommend that all registered users re-download DL_MESO
using the instruction they received on registration to obtain the corrected
version of the software. The decryption password has not changed and thus re-
registration is not required. The changes to pairwise thermostatting across
shearing boundaries are also reflected in a revised version of the user manual,
which can be found both in the DL_MESO distribution and via the DL_MESO website
(www.ccp5.ac.uk/DL_MESO).
Michael Seaton