Can temper command be used twice in one script?

Dear all,

I wanted to run REMD to reproduce the results in the literature. It’s said:

Dear all,

I wanted to run REMD to reproduce the results in the literature. It's said:

**************************************************************
the total simulation time of each replica was 5 ns (5x10^6 timesteps),
which were exchanged every 2 ps. For each temperature (replica), the
physical quantities were averaged within the last 2 ns. For each sweep of 2
ps, the first ps is used for the equilibrium of the system, and only the
second ps is used for statistics.
**************************************************************

I have learnt that the related command in lammps is *temper* and this
command is similar with *run *command. According to what I have learnt
about lammps, I thought the temper command should be used twice in one
script so as to split the total simulation time into two parts and average
the physcial quantitis in the latter 2ns. I wonder if the temper command
can be used twice in one script, just as the run command.

**************************************************************
variable Ttemp world 300 310 320 330 340 ...
fix tempfix all nvt temp \{Ttemp\} {Ttemp} 0.01
temper 300000 2000 \{Ttemp\} tempfix 3847 58382 fix 3 all ave/time 100 10 2000 v\_P v\_T file \./result/{Ttemp}.txt
temper 200000 2000 ${Ttemp} tempfix 3847 58382
**************************************************************

In fact, I have tested this setting. It can run. However, I am not sure
whether it was right because I found the second temper command assigned the
same temperatures for the replicas as the first one. I thought it should be
a succession of the last assignment of the first temper command.

​i think it is a pointless exercise. ​for typical systems, data points that
are less than 1ps apart are still strongly correlated, thus they will have
a low statistical efficiency and it would be practically equivalent to
simply sample the data from the (single) step right before each swap
attempt and then attempt to swap every 1ps (or every 2ps, if you feel safer
this way). since dynamical information doesn't matter, it would be fine and
not messing with complex setups are needed.

axel.

In fact, I have tested this setting. It can run. However, I am not sure whether it was right because I found > the second temper command assigned the same temperatures for the replicas as the first one. I thought > it should be a succession of the last assignment of the first temper command.