I am preparing a series of water-propanol binary mixtures for simulations. I have started by keeping a box length of 4.5 nm (cubic) and put 500 propanol and added 60 water molecules (using gmx solvate -maxsol 60)
I have chosen 500 propanol molecules as that is the maximum number of propanol that I can insert using gmx insert-molecules tool. I need to prepare solutions for water-mol-fractions (Xw) varying 0.1 to 0.9.
As maximum number of propanol I can add to the box is 500 (for 4.5 nm cubic), 500 propanol molecules should be considered for the Xw = 0.1 system, that alcohol having the higher mole-fraction. In that case for Xw = 0.1, we get a total number of molecules of around 560 and number of water required for Xw=0.1 will be ~60 water molecules.
Thus I kept the total number of molecules constant at 560 throughout the Xw variations and prepared the systems in 4.5 nm cubic boxes.
Now as I increase Xw, number of propanols will decrease and water will increase, but as propanol is larger is size than water, for higher Xw the simulation box is remaining vacant at multiple locations
E.g. for Xw = 0.9, N(propanol) = 60, N(water) = 500; Now I run energy minimization followed by 5 ns NVT equilibration (unrestrained) followed by 5 ns of NPT equilibration for the system. What I see is that the box sizeis drastically shrinking from 4.5 nm to approximately 2.94 nm. With increaseing Xw, box size is decreasing.
I tried with larger box sizes of 5 nm, 5.5 nm, similar things are happening. Is there any other way to keep the box size fluctuations from happening?
Can we run simulations with different total number of molecules but say keep either number of water or propanol molecules fixed throughout the solutions? If we are to keep the total number of molecules constant, how can we prepare this binary mixture across varying water mole-fractions so that the box-size will not shrink down drastically? Any help is much appreciated, thank you.
I don’t understand why you don’t want the box size to fluctuate, but there are two options for you to achieve that:
Do not run NPT simulations, but stick to NVT. The results may not be comparable to experimental data due to the low pressure you will have in your simulation system.
Start with a realistic box size, seemingly ~3nm in each dimension. But that may make it difficult to fit enough molecules using gmx insert-molecules.
Thank you for your reply. If the box sizes fluctuate to a large degree, would not that affect the comparison of results across the binary mixtures with different water mole-fractions? Or is it okay to have simulations of different systems with fixed total number of molecules but having different equilibrium box dimensions? Can we compare their results, e.g. H-bond propensities, rdf etc.?
If you would fix the box dimensions the density would be different from one system to another (based on the size of the original box and the number of molecules you put in it). I think you’d have more problems with arbitrary densities than arbitrary simulation box sizes (as long as the boxes are large enough considering the non-bonded interactions).
Densities are supposed to be different for different compositions of the binary mixture right? So is it okay if we achieve desired experimental densities even with large fluctuations in box dimensions?
Yes, I don’t see how large box size fluctuations are a problem, especially if you are starting from a system generated with gmx insert-molecules. If starting from a pre-equilibrated state using gmx solvate there might be concerns if the box size fluctuates a lot, but in your case I would say that the box size adaptions are a feature rather than a problem.