Talk:Mod:Hunt Research Group/Ensembles

Ensembles

Talk through the ensembles. Main ones needed are NVE, NPT and NVT.

Transport properties use NVT or NVE for production run for structure use NPT as it has a better density

1. Run NVE minimise

2. NPT to correct cell size and density => Barendsen

-looking for shrinking box size, because earlier we set a slightly larger box size as instructed for packmol

-but it is possible it could expand, if your initial density is too high, unlikely

-100ps for water, 1000ps (1ns) for IL due to higher viscosity

-keep running until box stops changing size ... box length is in history file

-a second NPT run may result in box increasing slightly OK if only a small increase i.e. 2%

-on a restart need coordinates, velocities and forces, must copy revcon to config, and revive to revolt, history and statis file (to a whole new directory so you don't accidentally overwrite), to use all of this information you must have all these files to enforce proper restart

-if the first simulation time is set to 1ns set the max time to 2ns in the second simulation, the code will find the old run, and catenate to the old history file running from 1ns-2ns and NOT from zero

-if you just give the coordinates, dlpoly will apply a random distribution of forces, this can be potentially destabilising, avoid this if at all possible

-if the box size is changing you may need to reset values for the VdW cut-off as an interim measure you can set this a little bit lower for the initial runs, don't forget to increase them for the final equilibration and production runs

Set temp and pressure normally atmospheric pressure.

Choosing NPT ensemble (constrains both TP)

Barendsen is normal quicker to equilibrate, scales all forces to match the target temp, scaling every n ps, for water 0.1ps , we found for molten salts 0.1 seems, for ILs 0.5ps seems good. You should test over a range of values, and not the same for temperature and pressure. This is system dependent but a goal is to have smaller time interval to reach equilibration more quickly, a longer interval disrupts the system less. In ILs you may have both the same, but for water you can scale temp less often and pressure more often as pressure variations can be larger.

Nose-Hoover is more mathematically rigorous, used for production for structure, not ideal for transport properties because this ensemble is altering the velocities, use NVE or NVT for transport properties, people prefer NVT because this keeps a good temperature, with NVE your temperature can go all over the place. Experimentally the temperature is normal const. For transport properties use NVE to be really accurate/theoretically correct.

during the NPT run the pressure and temperature should converge to target 1atm and temp (typically). potenital will not necessarily freeze or melt at the experimental temp for water melt at 273K, equilibrate at 298-300K (25ºC) be careful what is the melting point of your system you should be running above the melting point other than an initial check of the parameters in a short 50-500 step low temp (well frozen) simulation is there any experimental data you can compare to! closer to the melting point the closer to crystal structure for IL run 50-60ºC above melting point, for a molten salt run 100-200ºC above melting point

during equilibration P and T should converge to target values, variation pressure ±0.1 atm, temperature variation depends on magnitude, water ±5 ideally, ±10 OK, ionic liquid ±10-20K reasonable number, molten salt at 1500K ±20 is OK. Expect smaller fluctuations for shorter tilmestep in the restriction.

look at statis file, use Vincent script from [1] box size is constant look at total energy (total extended system energy), is it stable look at pressure look at the temperature (NPT) in normal simulations of 350K looking ±10 K, a range of about 10% either side of the target is acceptable, ideally within 5% of the target temp

a sharp change in temperature or pressure is indicative of a problem, potentially a phase change

Simulated Annealing

Simulated annealing allows structures to gain enough energy to overcome energy barriers and reach different minima.

Do you need to do simulated annealing? Use NVE do this for ILs to get it going, these have high viscosity, chains can get over each other for structures with alklyl chains to allow the chains to move for highly structured coulomb systems like molten salts If starting from a crystal structure to get the system moving For water no simulated annealing DES low melting, short alkyl chains probably not required

Simulated annealing involves heating the box up, in the case of molten salts in 100 K (IL's 50 K jumps) steps to a temperature that is about 2* the melting point. The heating part of annealing simulations should run for 100-500ps until temperature stabilises at target then run for 50ps at stable temp. When increasing the temp for molten salts use the Baronsen NPT thermostat and barostat then use the Nose-Hoover thermostat and barostat for final equilibrations and the production run. At each temperature an NVE is completed followed by an NPT. All you need to copy is the REVCON file from the NVE to a new directory for the NPT simulation. After the NPT simulation again copy the REVCON to a new directory to run an NVE at a higher temperature.

Once the high temperature has been reached the simulation is cooled in 50 K steps for molten salts (25 K for IL's) to reach the target temperature. The cooling part of simulated annealing should run for longer, 500ps-1ns again looking for stabilisation in temperature. It is particularly important to use simulated annealing if the starting structure is from a crystal to allow molecules to move.

command in dlpoly to print out RDF called ref repeat process until 2*mp temp look at trajectory eyeball! are things moving if they are you don't need to reach 2*mp as the maximum look at literature see what others have done

what are common problems, you can't reach the target temperature in individual steps ... run for longer eye-ball simulation => are there voids? -this indicates that your NPT is not good, i.e. before you started the annealing -you can start again ... or do an NPT with the void ... must check the density may need to be a really long run. =>do things slow down?