On Wed, 2015-04-15 at 18:02 +0530, Swithin Hanosh wrote:
> Dear Jason M. Swails,
>
> Thank you for the reply. I followed your instructions, but now I am
> encountering errors : "RESTARTED DUE TO LINMIN FAILURE" and "Program
> received signal SIGSEGV".
>
> Program received signal SIGSEGV: Segmentation fault - invalid memory
> reference.
>
> Backtrace for this error:
> #0 0xB7473163
> #1 0xB7473800
> #2 0xB77D13FF
> #3 0x811115E in nb_adjust_
> #4 0x8112D1A in ewald_force_
> #5 0x827E934 in force_
> #6 0x80C277A in runmin_
> #7 0x80B1077 in sander_
> #8 0x80AA74F in MAIN__ at multisander.F90:?
>
> The output file is attached with the message. Please suggest how to solve
Something seems to be very wrong with your minimization. If you look at
the output file, you see that the energy goes down, as expected, for the
first 150 steps. On step 200, the electrostatic energy overflows its
field and the energy plummets by 2 orders of magnitude. This is *not*
expected. The energy continues to plummet until it hovers about 3 orders
of magnitude smaller than it started.
When it hits the LINMIN failure (i.e., the minimizer got "stuck"), it
resets configuration and you see a better energy at step 550. But at
step 600, the energy starts to plummet again, the maximum gradient
shoots up by 7 orders of magnitude, and you ultimately get a segfault.
I *suspect* what's happening is that some atoms are collapsing on top of
one another during the minimization. This is usually caused in one of
two situations:
1. Some atoms have van der Waals parameters that are 0 when they
shouldn't. (Did you create an frcmod file for part of your system? Did
you check that none of the lines had "ATTN:" at the end after filling in
0s for all of the parameters?)
2. A hydrogen attached to an oxygen (which has 0 vdW radius) collapses
into another atom it forms a torsion with since the oxygen it's bound to
is excluded from interacting with that atom.
Go back and check the files you used to build the prmtop in the first
place and make sure all parameters have been defined. You can also set
the ntwx variable in your input file to get a minimization "trajectory"
so you can see where the collapse is occurring. This should help narrow
down what is happening.
HTH,
Jason
--
Jason M. Swails
BioMaPS,
Rutgers University
Postdoctoral Researcher
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Apr 15 2015 - 06:30:03 PDT
