Stray PME grid charges on increasing number of replicas and T-range with T-REMD

From: Francesco Pietra (chiendarret_at_gmail.com)
Date: Tue Sep 10 2013 - 05:24:58 CDT

Hello:
With debug (64 nodes, 1024 processors) T-remd (parallel tempering), on
increasing the number of replicas, from 16 to 32, and the corresponding
temperature range from 314-320 to 314-338 (expecting a good exchange for T
gaps of 0.8K as it occurred at 314-320) I am noticing stray PME grid
charges.

In a first attempt this occurred at step 1590 (out of expected ca 2000), on
a second attempt at step 190 with, of course, all 32 replicas.

Never had a problem with same input data during extensive NPT MD, or 16
replicas, same ts=2fs.

I should say that PMEGridSizeZ differs much from what is advised on the
manual. See below please what was used for both successful MD, 16 replicas,
and 32 replicas:

cellBasisVector1 60.88 0. 0.
cellBasisVector2 0. 92.45 0.
cellBasisVector3 0. 0. 102.99

cellOrigin -12.409886360168457 36.25075912475586 -53.04564666748047
PMEGridSpacing 1.0
PMEGridSizeX 65
PMEGridSizeY 96
PMEGridSizeZ 105

The system is a 34aa in TIP3 water, total number of atoms 55,118.

Thanks for advice.

francesco pietra

This archive was generated by hypermail 2.1.6 : Wed Dec 31 2014 - 23:21:38 CST