Welcome To Support Community

Materials Studio

Advanced Search
Ask Search:
AleNuce95AleNuce95 

Sorption: Bug?

Hi all,

I have to optimize a force field for guest-guest interaction for the probe of Argon at 87 Kelvin and for doing that, I have to compare the simulated isotherm with the real gas density found on the NIST site, mapping the same gas phase and condensing the gas at the right pressure point.

At the beginning, I took a 1000 A^3 void box to simulate my isotherm in this way:
Quality ultra-fine, no charge, electrostatic method: Ewald, van der Waals methos: atom based, cut off 20 A .
I thought I had come to convergence by finding two parameters of epsilon and sigma that would resume the NIST trend and condense the gas at 95 KPa pressure.

To certify what I found I tried with two bigger boxes (8000 and 27000 A^3) with the same setup used before. With my great surprise my optimized forcefield didn't condense the argon to 95 KPa, but around 300 KPa in both case.

How is this happening explained?? The average loading in the box of 1000 A^3 is much less than 1 atom (0,050).
Could this be the cause?

Thanks,

Ale
Best Answer chosen by AleNuce95
reinierreinier
Hi Ale,

I think what you observe is hysteresis due to nucleation. If you reverse the pressure interval in Sorption you may find the opposite, i.e.  evaporation happening at a pressure lower than the experimental vapor pressure. The simulation is dominated by interfacial effects.

To study phase transitions people tend to eliminate the interface through the Gibbs ensemble [1], but this is not available in Sorption.

Another way you could go is through calculation of the solvation free energy in Forcite. From this you can estimate the vapor pressure.  In fact, I did this for argon recently, see [2].

Best,
Reinier

[1] A. Z. Panagiotopoulos, N. Quirke , M. Stapleton, D. J. Tildesley, Phase equilibria by simulation in the Gibbs ensemble, Molecular Physics, 63(4) 527545 (1988).
[2] Reinier L. C. Akkermans, Solvation Free Energy of Regular and Azeotropic Molecular Mixtures, J.Phys.Chem.B 121, 1675−1683 (2017).

All Answers

reinierreinier
Hi Ale,

I think what you observe is hysteresis due to nucleation. If you reverse the pressure interval in Sorption you may find the opposite, i.e.  evaporation happening at a pressure lower than the experimental vapor pressure. The simulation is dominated by interfacial effects.

To study phase transitions people tend to eliminate the interface through the Gibbs ensemble [1], but this is not available in Sorption.

Another way you could go is through calculation of the solvation free energy in Forcite. From this you can estimate the vapor pressure.  In fact, I did this for argon recently, see [2].

Best,
Reinier

[1] A. Z. Panagiotopoulos, N. Quirke , M. Stapleton, D. J. Tildesley, Phase equilibria by simulation in the Gibbs ensemble, Molecular Physics, 63(4) 527545 (1988).
[2] Reinier L. C. Akkermans, Solvation Free Energy of Regular and Azeotropic Molecular Mixtures, J.Phys.Chem.B 121, 1675−1683 (2017).
This was selected as the best answer
AleNuce95AleNuce95
Thanks Dott. Reiner for the answer.

So ,in few words, "Adsorption isotherms" in the Sorption Package isn't reliable to study the point of gas's condensation throught simulated isotherms.

Right?
reinierreinier
Yes I suspect you will suffer hysteresis, as condensation takes place through nucleation. Instead of vacuum you could take a zeolite as host to calibrate the force field.
 
AleNuce95AleNuce95
Thanks again for the help, Dott. Reiner.

I have read the articles and I would like to ask a thing. I don't understand why the ensemble Gibbs should be a problem during the transition gas->liquid in the vacuum and not in a matrix, like a zeolite?

I ask this because looking in the "literature" (surfing on the internet) I have found some presentation like this http://www.acmm.nl/molsim/frenkel_smit/Presentations/Gibbs.pdf where at the end it's reccomended the Gibbs ensemble for determining the gas-liquid phase.

However, I think, I have confirmed the hysteresis problem as you can see in the attachment. With the same FF adsorption and desorption are very different....

Thanks for the patience and time.

Ale
reinierreinier
Not sure I follow you; I agree with you that the Gibbs ensemble is best for simulating gas-liquid equilibria. But unfortunately this technique is no longer available in Materials Studio.

However if you aim to calibrate a forcefield for sorption in zeolites, it is common to optimize the Lennard-Jones parameters against experimental Henry constant and heat of absorption. As an example see Fig 3 in [1]. This is something you could do in Materials Studio.
User-added image

Best,
Reinier

[1] B. Smit, Simulating the Adsorption Isotherms of Methane, Ethane, and Propane in the Zeolite Silicalite, J. Phys. Chem. 99, 5597-5603 (1995).
 
moondogmoondog
Hi,
You not specify the  types of moves and the production and equilibration steps.
In all simulation you always report some parameters and same cutoff for the sum of non electrostatic contributes. 
Are you sure of the convergence of all parameters of GCMG calculation?
A wrong measurement of density can bring erroneus results.

M.
AleNuce95AleNuce95
Thank you for the answering.

Dott. Reiner:
Sorry I didn't understand. I believe that the problem was the presence of the Gibbs ensemble in the Materials Studio' s simulated isotherms. It's the contrary. Sorry

Moondog:
1) "You not specify the  types of moves and the production and equilibration steps."
For "types of moves" do you you refer to this?
User-added image
Should I also modify these parameters?
2) "Are you sure of the convergence of all parameters of GCMC calculation?
A wrong measurement of density can bring erroneus results." 
When I started my job I found these parameters of LJ which reproduce the "experimental" values coming from NIST in a empty box of 10^3 A^3 starting form a pcff standard with a ultrafine quality.
User-added image
Ok I didn't consider the comparison between heat of adsorption and Henry coefficient, but is this the only reason why, changing the volume of the box, the final results are so different?
I have tried to change the equilibration and the productional steps (medium,fine,ultra-fine), the cut off (18,27,34), and the electrostatic method (ewald and atom based), but the results are always the same:
A) 20^3 cube: condensation at about 300 KPa
B) 10^3 cube: condensation at about 100 KPa
C) 5^3 cube: condensation at about 50 KPa

Thanks for any help,

Ale