Dear Colleague,

A number of bugs have been discovered in the Dissipative Particle Dynamics (DPD) 
code of DL_MESO 2.5 and its post-processing utilities.

In systems created from scratch involving multiple species of unbonded particles 
and frozen particle walls, frozen particles are mistakenly included in the 
algorithm for assigning species for unbonded non-frozen particles, causing an 
infinite loop. (Many thanks are due to Mary McDonald from The University of New 
South Wales, Sydney for bringing our attention to this bug.)

To fix this bug, changes to start_module.f90 are required. Please change line 
829 to

  gbltotnbd (idnode+1) = nubeads - nfbeads

replace line 836 with

  IF (lfrzn (ispe)==0) THEN

and insert

  END IF

immediately after line 840. The same IF statement as above should be inserted 
immediately before line 846 and END IF should be included immediately after line 
859, i.e. lines 846 to 859 should only be executed for non-frozen species. Line 
865 should be changed to

  DO i = 1, (nubeads - nfbeads)

and line 883 should be modified to the following:

  k = (nubeads - nfbeads) * mtrnd (idnode) + nfbeads + 1

Minor errors in the ewald_frozen_slater routine of ewald_module.f90 - used to 
calculate corrections for pairs of charged frozen particles - were also 
discovered. To correct these errors, please insert the following immediately 
after line 572:

  nfcboffset = 0

change line 575 to

  IF (lfrzn (ltp (i))>0 .AND. ABS(chge (ltp (i)))>1.0e-16_dp) THEN

and change line 583 to

  IF (idnode>0) nfcboffset = SUM (nfcbead(1:idnode))

The endianness checks in the traject.f90, trajectselected.f90, local.f90, 
exportconfig.f90 and exportimage.f90 utilities have been shown to be unreliable. 
In particular, it has been demonstrated that these checks are less likely to 
work when HISTORY* files are particularly large, i.e. larger than approximately 
350MB. (Many thanks are due to Ming-Tsung Lee from Rutgers University, New 
Jersey for bringing our attention to this problem.)

To fix the problem, we recommend that the endianness checks in each of these 
codes are commented out or removed: these can be found in lines 123-140 of 
exportconfig.f90, lines 152-169 of exportimage.f90, lines 64-81 of local.f90, 
lines 59-76 of traject.f90 and lines 58-75 of trajectselected.f90. All 
subsequent instances of opening export* or HISTORY* files should also be 
modified; for instance, lines 82 to 88 of traject.f90:

  IF (bigend) THEN
    OPEN (ntraj, file = 'HISTORY0000', access = 'sequential', form = 'unformatted', &
                &status = 'unknown', convert = 'big_endian')
  ELSE
    OPEN (ntraj, file = 'HISTORY0000', access = 'sequential', form = 'unformatted', &
                &status = 'unknown', convert = 'little_endian')
  END IF

should be replaced with

  OPEN (ntraj, file = 'HISTORY0000', access = 'sequential', form = 'unformatted', &
              &status = 'unknown')

noting that both the IF statement detecting the value of big end and the convert 
parameter in the OPEN statement are omitted. If any of the utilities are to be 
run on different computers to those used for DPD calculations, it may be 
necessary to compile them with a compiler flag to switch the endianness: please 
refer to the documentation for your Fortran90 compiler to determine what form 
this should take.

We strongly recommend that all registered users of DL_MESO version 2.5 should 
either make the code changes listed above or re-download DL_MESO using the 
instructions they received on registration to obtain the corrected version of 
the software. Re-registration is not required and the same password is used for 
decryption.

Michael Seaton
2014-05-06