Re: SMD and Temperature control

From: Marcos Sotomayor (
Date: Mon Aug 10 2009 - 20:35:24 CDT

Dear S

You can easily adapt the ft.tcl script given in the tutorial and
compute the work done on the system during the sample simulation (by
integrating F times d). The total work done on the system is about
1922.938 kcal/mol, and NAMD reports a fairly similar increase in energy
of about 1966.36 kcal/mol (the difference arising likely due to numerical
errors in the integration and the use of a multiple time stepping
algorithm). The increase in temperature is thus caused by the large
forces applied in a very short simulation time and is not related at all
to the system equilibration or any problem with either the simulation or

The whole NAMD tutorial was designed to teach how to do MD simulations and
run sample simulations quickly in a not-so powerful computer. Therefore,
and as explained many times in the tutorial, the sample SMD simulations in
the tutorial are done in "unrealistic" conditions (vacuum) and at
pulling speeds that are *very* large. If you pull more slowly, the
applied forces will be smaller, the work done on the system will be
smaller, and the temperature increase (due to the work done on the
system) will be smaller. If you have a big enough system, pull slow
enough, and do not use multiple time stepping, the temperature increase
will be negligible.

The previous conversation about temperature increase was related to
artifacts arising from the use of a multiple time stepping algorithm. In
the tutorial case, the temperature increase is mainly related to the
conditions used.

Needless to say, as pointed out in the tutorial, that if you want to do a
proper SMD simulation you need to have an explicit solvent, an
equilibrated system, and pull at speeds that are relevant to whatever
system/process you are studying.

Last but not least, there is no space here for luck. Its better to
understand what is going on your simulations!


On Fri, 7 Aug 2009, snoze pa wrote:

> Dear Namd users,
> Thank you very much for your support. I have a question related to
> increase of temperature in SMD simulation. I was reading the past
> conversation between Sterling, Marcos and Gianluca and observed that
> the system temperature actually increases a lot in SMD simulation
> compared to what Marcos statement of 20K.
> For example, in SMD tutorial provided in the NAMD website(ubiqutin
> 3-1-pullcv, ubq_ww_pcv.log) , the initial TEMPAVG was 297.5483 which
> increases up to TEMPAVG equal to 637.5142 at the end of 20000 step.
> This is a huge increase in temperature. Do you think in the example
> the system was not properly equilibrated?
> Somehow I was luck that my system temperature increases only by 10 and
> 15 in two SMD simulation.
> Thank you
> S

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:53:09 CST