Number of molecules in GEMC box

[pakamore]

Hello,

I am trying to simulate molecule adsorption in MOFs using Gibbs ensemble MC algorithm. I was wondering how do you calculate the initial number of molecules in the box?

Kind regards,
Ignas

[David Dubbeldam]

The number of molecules in Gibbs is an input parameter. Usually you want a sufficient number of molecules to reduce finite size effects (like 256 or 512) in liquid/vapor equilibrium Gibbs.
In case of adsorption, you need more molecules that the saturation loading in the framework plus then enough molecules in the vapor phase.

[sara abbasi]

Dear Dr. Dubbeldam
I prepare the attached code to calculate vapor liquid equilibrium and the density of vapor and liquid phases for propane using TraPPE-UA force field parameters. But the vapor density is not accurate. I changed the number of molecules so many times and ran the code, but I did not get the right answer. Would you please guide me? Thank you for your kindness and support.
SimulationType                MonteCarlo
NumberOfCycles                25000
NumberOfInitializationCycles  5000
PrintEvery                    1000
RestartFile                   no

Forcefield                    TraPPE-UA
CutOff                        14

Box 0
BoxLengths 30 30 30
BoxAngles 90 90 90
ExternalTemperature 344.0

Box 1
BoxLengths 30 30 30
BoxAngles 90 90 90
ExternalTemperature 344.0

GibbsVolumeChangeProbability 0.1

Component 0 MoleculeName             propane
            StartingBead             1
            MoleculeDefinition       TraPPE-UA
            TranslationProbability   0.5
            RotationProbability      0.5
            ReinsertionProbability   0.5
            GibbsSwapProbability     0.5
            CreateNumberOfMolecules  185 15

[David Dubbeldam]

What do you mean wiith "not accurate"? Compared to what? experiment or the result of the TraPPE force field?
Force TraPPE, make sure you use an unshifted potential and tailcorrections.

[sara abbasi]

Dear Dr. Dubbeldum
Thanks a lot for your response. I compared my results with experimental data. The results for liquid density are accurate at all temperatures with the error of 1%, but the results for vapor density are not accurate with the error of 20%. When I compared my results with the results of a paper ,published in 1998, I understood that they could not get the accurate results for vapor density using TraPPE-UA.
What is the reason that vapor density is not achieved? Is not it because of the length of boxes?
SimulationType                MonteCarlo

NumberOfCycles                100000
NumberOfInitializationCycles  10000

RestartFile                   no
ContinueAfterCrash            no
WriteBinaryRestartFileEvery   1000

PrintEvery                    1000
PrintPropertiesEvery          1000

ChargeMethod                  Ewald
EwaldPrecision                1e-5
Forcefield                    TraPPE-UA
CutOffVDW                     14
CutOffChargeCharge            14

Box 0
BoxLengths 30 30 30
BoxAngles 90 90 90
ExternalTemperature 200

Movies                        yes
WriteMoviesEvery              1000

ComputeEnergyHistogram        yes
ComputeNumberOfMoleculesHistogram      yes
ComputeMolecularPressure      yes

Box 1
BoxLengths 30 30 30
BoxAngles 90 90 90
ExternalTemperature 200

Movies                        yes
WriteMoviesEvery              1000

ComputeEnergyHistogram        yes
ComputeNumberOfMoleculesHistogram      yes
ComputeMolecularPressure      yes

GibbsVolumeChangeProbability 0.05

Component 0 MoleculeName             propane
            MoleculeDefinition       TraPPE-UA
            TranslationProbability   0.5
            RotationProbability      0.5
            ReinsertionProbability   0.5
            GibbsSwapProbability     0.5
            CreateNumberOfMolecules  100 100
           
 

[David Dubbeldam]

The length of the boxes is important to reduce finite-size effects.

But more fundamentally, it is quite hard with a simple set of LJ parameters to get both the liquid and the vapor branch accurate. This is btw one the main reason to fit parameters on VLE-data. But the united-atom approach has its limits. More complicated models like full-atom models have more parameters and are harder to optimize.
So... having a good liquid phase and 20% off for the vapor phase is not that bad.