This seems to be a common problem. So I am just pasting the answer that I
gave sometime back as how I could get over it. Hope it helps. But I
suspect that the problem with num_ks I faced was slightly different, but
nothing worng in tryng though.
Best wishes,
-Sanjeev
___________________________________________________________________________
num_ks error, as I vaguely remember, comes with problem of
'vacuum' that exists between the boxes due to periodic boundary condition.
You must be using leap to add water. It gives this problem and you cannot
visualize it looking at the structure (which looks perfectly fine)!
The problem of num_ks is this: sander expects and inisits on
partioning atoms for different processors as equally as possible, at
least nearly equal. It fails if density in some regions is very low. So
the recipie would be to run on a single processor for a small amount of
time, so that, hopefully in that short time, the vacuum is 'filled'. (It
should be alarming if your density is below 0.8.) Then you can run on
multiple processors. Hope this works.
___________________________________________________________________________
hi folks.....set up a simulation just as i always do, nothing fancy.....when i try to minimize it,
i get the following message:
EWALD BOMB in subroutine ew_recip/grid_weights
nmine exceeds num_ks
System must be very inhomogeneous. Cannot continue.
Terminated
logout
the first lines in min.out are
NSTEP ENERGY RMS GMAX NAME NUMBER
1 nan nan 3.3290E+10 CZ 11209
BOND = 819.3590 ANGLE = 5097.4239 DIHED = 3670.0325
VDWAALS = ************* EEL = nan HBOND = 0.0000
1-4 VDW = 5915.4236 1-4 EEL = 30333.7495 CONSTRAINT = 0.0000
fyi, the crystal structure was minimized before i added counterions and solvated; this minimization
(in vacuo) worked fine. any idea what gives? tjm
--
Todd J. Minehardt
Department of Chemistry
Princeton University
Princeton, NJ 08544-1009
Phone: 609.258.1932
Fax: 609.258.6746
tjm_at_princeton.edu
Received on Fri Jul 13 2001 - 11:30:38 PDT
