Code:gmxliquid

In this tutorial, I'm assuming you already have the gromacs commands (grompp, mdrun, etc.) loaded. I'm also assuming you have gromacs 5, which made significant changes to the way command-line tools are used.

You'll also need the starting files [[Media:liq_files.tgz|here]]. Or, on circe, just run

cp -R /shares/mri_workshop/Liquids. cd Liquids

= Create an input system =

First, we do some size calculations in the python prompt.

>>> # Compute number of EtOH molecules >>> # 8 nm^3 = 8e-21 cc >>> # 0.4 % By vol. >>> # rho_EtOH = 0.78924 g / cc >>> # FW = 46 g/mol >>> >>> 8e-21*0.4*0.78924/46.0*6.022e23 33

Next, we need to generate an 8 nm$$^3$$ box with 33 EtOH molecules

gmx insert-molecules -box 2 2 2 -ci etoh.pdb -o box.pdb -nmol 33 nano topol.top # check for line "ETH 33" (no changes required)

If the topology file and forcefields are 'there', we can create a test simulation using just these molecules. We'll do this just to check that grompp works

gmx grompp -c box.pdb -f equ.mdp -o equ.tpr

If this completes without errors, we know that we could simulate the ETH molecules we have now. That would give a high-density gas, rather than the liquid we want. Next, the complete water+EtOH mixture can be made by filling the voids with water. We'll use TIP4P here.

gmx solvate -cs tip4p -cp box.pdb -o sys.pdb

Since this last command added waters, you have to update the topology file, adding "SOL 130" (or your number of waters)

nano topol.top

= run minimization =

Molecular structures can be touchy, since contacting atoms causes large forces and 'blows up' a system. The risk of this can be reduced by minimizing the energy of initial starting systems.

First, we have to create a file describing how gromacs should do the minimization.

gmx grompp -p topol.top -f min.mdp -c sys.pdb -o min

A lot of backup files (starting with '#') accumulate, and we remove them like so:

rm -f \#* mdout.mdp

Now we're ready to run minimization. Since the system is small, minimization is cheap and fast - so we do it on the head node. For larger systems, this command would be put in a job-script.

gmx mdrun -deffnm min

The latest gromacs has a bug of some sort or can't work with OPLS-AA for EtOH, since I get a segfault here. Hopefully this will be fixed soon. Otherwise, you can manually delete the EtOH from sys.pdb and topol.top and then re-run the grompp and mdrun steps.

Now we're ready to take the minimized structure and run dynamics. This first round is called equilibration, since we intend to let the system settle into a thermal equilibrium state.

gmx grompp -c min.gro -f equ.mdp -o equ.tpr

To run this one, we'll use the cluster by writing down the command in a script file. That script gets sent to the cluster. No changes to equ.sh should be required.

nano equ.sh sbatch equ.sh squeue -u `whoami`

As the job is running you can read through the log file.

less equ.log tail -f equ.log
 * 1) enter Ctl-C to stop tailing

Note that the energy is output at every time-step. Why is the total energy changing with time? What about the total volume?

= run dynamics =

Running dynamics uses the same procedure as before, but now we need to worry about how much and which output to produce.

nano run.mdp # no changes required gmx grompp -c equ.gro -f run.mdp -o run.tpr

Check and start the job (no changes required).

nano run.sh sbatch run.sh squeue -u `whoami`

Read through run.log

less run.log tail -f run.log
 * 1) enter Ctl-C to stop tailing

Notice that the total energy remains relatively constant now, but the individual energies are still very noisy. What does this say about the shape of the molecules?

= analyze dynamic data =

Let's first check what happened during equilibration. Since the volume was allowed to change, we should be able to plot it vs. time:

gmx energy -f equ.edr -o equ.en.xvg
 * 1) Select: Potential Kinetic-En. Total-Energy Volume


 * At what time does the potential energy stabilize?
 * What is the density?
 * answer: (33*46 + 18*133)/Avogadro / 7.0333834916722191e-21 = 0.9236 g/mL
 * These two quantities inform on the heat and volume change of mixing.
 * Pure H2O (TIP4P): rho = 1.001, H ~ 11.6 kcal/mol, D = 3.9e-5 cm^2/s, eps = 52
 * Pure H2O (expt): rho = 0.9971, H = 10.5 kcal/mol, D = 2.3e-5 cm^2/s, eps = 78.4
 * Gas-phase TIP4P energy = 0 (no self interactions in the model)
 * Gas-phase OPLSAA-EtOH energy = 25 kJ/mol ($$\pm$$1)
 * (J. Chem. Phys. 123, 234505, 2005)
 * Pure EtOH: rho = 0.7873, H = 10.0 kcal/mol, D = 1.0e-5 cm^2/s, eps(expt) = 23
 * (J. Phys. Chem. B 1997, 101, 78-86, J. Phys. Chem. B, 2014, 118 (34), 10156-66)

Gnuplot can make this plot with

plot 'equ.en.xvg' u 1:2 w l, 'equ.en.xvg' u 1:5 axis x1y2 w l

The radial distribution function is a classical measure of the structuring of liquid water. The one we calculate here shows the density of molecules within spherical shells around the central molecule.

gmx rdf -rdf mol_com -ng 2 -s run -f run.trr -o run.w-rdf.xvg gmx rdf -rdf mol_com -ng 1 -s run -f run.trr -o run.e-rdf.xvg
 * 1) select Water as "reference group", select "Water", then "ETH"
 * 2) re-run for ETH-ETH rdf

gnuplot can plot these together with:

plot 'run.w-rdf.xvg' u 1:2 w l, 'run.w-rdf.xvg' u 1:3 w l, 'run.e-rdf.xvg' u 1:2 w l

Notice how waters stack much more closely together than EtOH? How is this related to their liquid densities?

The dielectric of a liquid measures its ability to be polarized by an applied field. High dielectric materials make good capacitors.

gmx dipoles -f run.trr -s run.tpr -c dipcorr.xvg -corr total -P 1

Dipole moment (Debye) - Average =   2.2234  Std. Dev. =  0.1095  Error =   0.0001

The following averages for the complete trajectory have been calculated:

Total < M_x > = 7.79882 Debye Total < M_y > = -1.11977 Debye Total < M_z > = 9.49511 Debye

Total < M_x^2 > = 882.516 Debye^2 Total < M_y^2 > = 1239.78 Debye^2 Total < M_z^2 > = 925.457 Debye^2

Total < |M|^2 > = 3047.75 Debye^2 Total |< M >|^2 = 152.233 Debye^2

< |M|^2 > - |< M >|^2 = 2895.52 Debye^2

Finite system Kirkwood g factor G_k = 3.52836 Infinite system Kirkwood g factor g_k = 2.37948

Epsilon = 43.1739

Water and EtOH are constantly undergoing random collisions in solution. The net effect of this is diffusion of the two molecules. Einstein showed that this can be tracked by watching the mean squared displacement of the molecules over time. Since the distance traveled over a random walk with diffusion constant $$D$$ has a Gaussian distribution with variance $$2D\Delta t$$, where $$\Delta t$$ is the elapsed time, the mean squared displacement vs time should be a straight line with slope $$2D$$. Of course, over short times, this picture is only approximate, but it usually holds over longer times.

gmx msd -f run.trr -ngroup 2 -o run.msd.xvg

Fitting from 100 to 900 ps

D[    Water] 1.4336 (+/- 0.0850) 1e-5 cm^2/s D[      ETH] 0.7312 (+/- 0.1279) 1e-5 cm^2/s

How do these compare with the literature values (above) for the pure liquids?

More detail on velocity correlation:

gmx velacc -mol -s run -f run.trr -o run.w-acf.xvg # Choose group 3 (Water) gmx velacc -mol -s run -f run.trr -o run.e-acf.xvg # Choose group 2 (ETH)

you can compare these to some of the above references.