From: Alex Curutiu (alexcurutiu_at_gmail.com)
Date: Sun Aug 15 2010 - 05:49:52 CDT
Hi all,
I am having problems calculating specific heat, basically on each run (using
same configuration file and same positions/velocities ) I am getting a
different result. Finally I end up with testing from namd-tutorial files, to
be more precise with spec-heat example. I will present now the results I am
getting when I run namd ( 2.7b2 binary for linux and 2.7b3 for win32 ) for
ubq-nvt.conf from 2-5-spec-heat folder. The only change in the configuration
was the following line :"set inputname
./1-2-sphere/example-output/ubq_ws_eq" instead of "set inputname
./1-2-sphere/ubq_ws_eq ". Basically to point to the already computed
positions/velocities from tutorial files. A second change I made in the
configuration file is related to the number of steps, first I try with
100000, then I continue until 400000. I will present now the results I am
getting (I also try to increase the number of steps, but with no results):
linux
ID
Nr steps
( <E^2> - <E>^2)
average.tcl script
Cv (J/(KgK))
1
100000
1309.54
3315.11
2
100000
2099.18
5314.07
3
200000
2686.75
6801.49
4
300000
3305.07
8366.78
5
300000
2416.38
6117.06
6
400000
3398.27
8602.71
7
400000
4058.48
10274.01
windows
ID
Nr steps
( <E^2> - <E>^2)
average.tcl script
Cv (J/(KgK))
1
100000
2402.185
6081.12
2
100000
4265.493
10798.076
Would you please advise? Is there an instability in namd code? Should I run
more steps? What actually is the specific heat of Ubiquitin since I am
getting results from around 3000J/KgK until 10000 J/KgK. To duplicate the
problem is straightforward, as input I am using only data from tutorial of
namd,nothing else (configuration file from 2-5-spec-heat folder with the
change for original position/velocites to the one already computed (
basically to the onle present in the tutorial files) ). As executable I am
using as well form namd web site. I run the simulation on a single core /
single processor. Is there any chance to compute specific heat with namd or
Thank you very much.
-- Best Regards, Alex Curutiu
This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:54:25 CST