Step 6 - Initial Comparison
Now that we have our structure factor calculations set up let’s run for a few steps and see what they look like.
Ctrl-5
You can go to the
tab of any of the
NeutronSQ modules and see the comparison of the F(Q) - we re-plot all seven together for convenience below (this is output from the
EPSR module):
Calculated total structure factors (dashed lines) compared to the experimental data (solid lines) in the equilibrated simulation
Overall the calculate F(Q) agree rather well with the experimental curves, with the exception that the peak at Q = 1.5 Å-1 is too intense and well-defined, particularly for the H3D8-D and H3D8-D samples. This peak is related to ion-ion neighbour interactions - we can lightly abuse an equation related to Bragg scattering, $r = 2 \pi / Q$, to calculate the real-space correlation distance as 4.19 Å - and so suggests that the forcefield, even with the scaled charges, is giving a liquid that is overstructured from too-strong interactions between the cation and anion.
We’ll let the
EPSR module sort that out for us!
Layer ⇨ Create ⇨ Refinement ⇨ Standard EPSR
Because of the stronger intermolecular interactions manifesting in a system of ions we will need to push the refinement a bit harder than usual. You can run the refinement with the default value of EReq of 3, but you will see very little improvement in the structure factors. Standard practice is to then sequentially increase this value in small steps (e.g. increments of 3 at a time) until no further improvement to the fits (as measured by the total R-factor) can be achieved. Simply putting in a huge number that is “definitely enough” is not a good idea - experimental data is imperfect to some degree, and at some point the generated potentials are highly likely to force the system into a worse state than it was before.
For the present system a value of about 17 is optimal, and we’ll put this in straight away:
Go to the
Refine (EPSR) layer tab
Select the
In the module change the EReq value to
17.0
Click the tab
Running the simulation for another 1000 iterations should give us a good refined structure:
Ctrl-F
Set the number of iterations to
1000and hit