[gmx-users] Trying to explain differences in behaviour between 2fs and 5fs timesteps

[Trayder]

Hi all,
I've been running an analysis of the same system (protein/ligand/water/ions)
using multiple timesteps (1-5fs) to identify differences in behaviour
between the settings. To do this I've compared the energy outputs of the 5
simulations, specifically the averages from g_energy and binning the
energies at each step from the log files.

What I've found is that 1/2/3fs simulations behave nigh identically, as do
4/5fs. However there is a notable difference between the two groups (graph
linked below). I've additionally matched timestep dependent variables
between simulations to rule out causes such as different coupling
frequencies, the difference was unaffected. 
The difference in the Total Energy seems to arise from the Coulomb (SR) and
LJ-SR terms. While the differences are in the order of 0.1% (~170kJ/mol) I
would rest easier being able to explain the sudden change in behaviour from
3-4fs.
Anyone have any ideas?

Thanks,
-Trayder

Binned total energies:
https://docs.google.com/file/d/0By3TeVcanmk8U0EwXzhNYk5JZUk/edit?usp=sharing

2fs mdp file:
integrator      =  md                          ; simulation algorithm
tinit                    = 0
dt                       = 0.002
nsteps          =  10000000                        ; # steps
;
;       Output Control
nstxout         =  500000                        ; write coordinates to .trr
nstvout         =  500000                        ; write velocities to .trr
nstlog          =  500                         ; write energies to .log
nstenergy       =  500                          ; write energies to .edr
nstxtcout       =  8000                          ; write coordinates to .xtc
energygrps      =  Protein OLAT Water_and_Ions  ;
;
;       Neighbour Searching
nstlist         =  10                           ; update neighbour list
ns_type         =  grid                         ; neighbour list method
pbc             =  xyz                          ; periodic boundary
conditions
rlist           =  0.9                          ; cut-off for short-range
neighbour (nm)
cutoff-scheme   =  verlet
;
;       Electrostatics and VdW
coulombtype     =  PME                          ; type of coulomb
interaction
rcoulomb        =  0.9                          ; cut-off distance for
coulomb
epsilon_r       =  1                            ; dielectric constant
rvdw            =  0.9                          ; cut-off for vdw
fourierspacing  =  0.12                         ; maximum grid spacing for
FFT
pme_order       =  4                            ; interpolation order for
PME
ewald_rtol      =  1e-5                         ; relative strength of
Ewald-shifted
DispCorr        =  EnerPres                     ; long range dispersion
corrections
;
;       Temperature Coupling
Tcoupl          =  v-rescale                    ; type of temperature
coupling
tc-grps         =  Protein non-Protein                       ; coupled
groups
tau_t           =  .1 .1                        ; T-coupling time constant
(ps)
ref_t           =  310 310                      ; reference temperature (K)
;
;       Pressure Coupling
Pcoupl          =  parrinello-rahman                    ; type of pressure
coupling
Pcoupltype      =  isotropic                    ; pressure coupling geometry
tau_p           =  2.0                         ; p-coupling time constant
(ps)
compressibility =  4.5e-5                       ; compressibiity
ref_p           =  1.0                          ; reference pressure (bar)
;
;       Velocity Generation
gen_vel         =  no                           ; generate initial
velocities
;
;       Bonds
constraints     =  all-bonds                    ; which bonds to contrain
constraint_algorithm    = lincs                 ; algorithm to use
lincs_order     =  4
lincs_iter     =  2




--
View this message in context: http://gromacs.5086.x6.nabble.com/Trying-to-explain-differences-in-behaviour-between-2fs-and-5fs-timesteps-tp5010304.html
Sent from the GROMACS Users Forum mailing list archive at Nabble.com.