[gmx-users] NMA problem (again and again!)

[Rossen Apostolov]

Fabrizio Mancinelli wrote:
> Hi Rossen,
> before claiming a bug I'm looking for (hopefully) a mistake by myself!
> I tried to minimize further (by using l-bfgs) the cg-minimized structure, but 
> I got the same error. Maybe the structure is not properly minimized yet?
>   
Well, mine was minimized to 10e-6 with l-bfgs (great !)
and non-bonding settings for no cut-off

pbc = no
nstlist = 0
rlist = 0
rcoulomb = 0
rvdw = 0

> Another problem could be inferred to the electrostatic treatment. Following 
> David's suggestion, I applied non-cutoff treatment (molecule's in vacuum):
>
> nstlist = 5
> ns_type = simple
> pbc = no
> rlist = 2.5
> coulombtype = shift
> rcoulomb-switch = 1
> rcoulomb = 2
> vdwtype = shift
> rvdw-switch = 0
> rvdw = 1
>
> Maybe I put wrong values for the parameters or their combination doesn't make 
> sense; I accept suggestions!
>
> Unfortunately I'm afraid it's not possible to compare with older versions, 
> because they don't support sparse matrix representation (my molecule is quite 
> big).
right, i tried with 3.2.1 and failed. my system is also huge, >5000 atoms.

as for the nm run I use cut-off

nstlist                  = 10
ns_type                  = simple
rlist                    = 1.3
coulombtype              = Shift
rcoulomb_switch          = 0.
rcoulomb                 = 1.
vdw-type                  = Shift
rvdw_switch              = 0.
rvdw                     = 1.
pbc                      = no

... and I get the relatively huge Fmax ~100 at the beginning.

Here's a snip of the gmxdump-ed coordinates from the nm.tpr:

      x[    0]={ 4.28020e+00,  2.97404e+00,  2.35656e+00}
      x[    1]={ 4.20314e+00,  2.93046e+00,  2.30664e+00}
      x[    2]={ 4.32527e+00,  2.90193e+00,  2.41130e+00}
     
........................................................................................


I constructed the matrix anyway and I'm now diagonalizing it,
lets see what comes out.


Rossen