[gmx-users] Diffusion Coefficient calculations problem

[Alexandre Suman de Araujo]

Hi Gmxers

I´m running a simulation of one íon in a box of water and I´m trying to 
calculate the diffusion coefficient from MSD calculations for this ion.

I performed 5 ns of simulation and I had the MSD x time curve nearly 
linear until 1.3 ns and after this the linearity is lost (what is 
natural since I´m performing the calculations for only one íon).

Them I ran the same simulation again but now until 10 ns and the 
linearity of the curve is lost at about 800 ps. I hoped to obtain better 
results, i.e., had linearity behavior for a time greater than that 
observed in 5 ns simulation calculation.

Moreover, for various calculations MSD always decreased after the linear 
behavior. Increased MSD trends were never observed.

It could be caused by COM motion removal at every simulation step or by 
the account of the PBC in MSD calculation?
Somebody could help me? The simulations parameters could be the reason???


PS: Here is my mdp file.

;
;       Input file
;
title                   =  Simulacao de calculo     ; a string
cpp                    =  /lib/cpp                 ; c-preprocessor
dt                      =  0.001                    ; time step (ps)
integrator        =  md                       ; Tipo de simulacao a fazer
nsteps              =  10000000                 ; number of steps
nstcomm           =  1                        ; reset c.o.m. motion
nstxout             =  100                      ; write coords
nstvout             =  100                      ; write velocities
nstlog               =  100                      ; print to logfile
nstenergy         =  100                      ; print energies
nstlist               =  10                       ; update pairlist
energygrps      =  IO1 SOL
ns_type             =  grid                     ; pairlist method
pbc                   =  xyz                      ; Usar PBC em todas as 
direcoes
rlist                  =  1.5                      ; cut-off for ns
coulombtype   =  Cut-off                  ; Usa cut-off por causa do 
calculo de e livre.
rcoulomb         =  1.5                      ; cut-off for coulomb
vdwtype          =  cut-off                  ; Tipo de VdW
rvdw                =  1.5                      ; raio de calculo de Vdw
tcoupl              =  berendsen                ; Temperature coupling
ref_t                 =  300                      ; Temperatura referencia
tc-grps             =  System                   ; Quais grupos sera 
afetados pelo tcoupl
tau_t                =  0.1                      ; Tempo que sera 
verificada a temperatura
pcoupl             =  berendsen                ; pressure bath
pcoupltype      =  isotropic                ; pressure geometry
tau_p               =  1                        ; p-coupoling time
compressibility     =  4.5e-5
ref_p                =  1.0

-- 
Alexandre Suman de Araujo
asaraujo at if.sc.usp.br
UIN: 6194055
IFSC - USP - São Carlos - Brasil