Open ramess101 opened 5 years ago
@msoroush
Yeah, that is what is shown in this figure:
This is the my vapor density data for TraPPE. This difference is not visible when you plot the density.
This data for one replica must have better quality.Can you use this data for TAMie to compare with their data?
@ramess101
If you are looking for Keasler uncertainties in VLE data, here is the link for it.
@msoroush
Where did you get the "GOMC" values for the figure you plotted above? Was that from recent simulations? I think if you convert that to a percent deviation plot you are going to see a similar discrepancy because the green points appear to be slightly shifted (on a log plot).
@msoroush
Thanks for directing me to Keasler uncertainties. They weren't reported in the original manuscript, but I didn't think to look on the website.
@ramess101 I ran it before, when I was trying to find the problem with torsion.
@msoroush
OK, but do you agree that there is also a small shift in vapor density? If you have the data, I could throw it on a deviation plot to see if it really does deviate slightly
@ramess101 I totally agreeing with you. I can see the deviation, but when you calculate the error, you can see it more clearly. Lets see how TAMie results looks like. One other thing I found out, in their paper, they mentioned that they used tabulated parameter for TAMie and used reservoir method for molecule insertion/deletion.
@msoroush
Cool, just wanted to make sure we were on the same page. Here are the TAMie results with the updated values you uploaded:
@msoroush
On Towhee and GROMACS I have always obtained consistent results for tabulated or native LJ potentials, but I guess it could impact your results if the spacing is not refined.
@msoroush
Here I have replotted the Keasler comparison with the uncertainties:
@msoroush
Here I have replotted the Keasler comparison with the uncertainties:
Our data are between Keasler and Yiann, data :D
@msoroush
It is curious that our rhol and DeltaHv are much lower for TraPPE, while our rhol is higher and DeltaHv lower for TAMie
@msoroush
So do we anticipate having results by Monday for TAMie with 10 A cut-off and maybe 14 A?
@ramess101 Hopefully they both will be finished by tomorrow. As soon as I get the results. I will send it to you.
@msoroush
Alright sounds good. Thanks again!
@msoroush @jpotoff
I think we need to anticipate that our TAMie results are going to have slight (but irreconcilable) deviations from Weidler et al. How do we want to address this in the manuscript/review? The deviations in DeltaHv are particularly concerning since the reviewer claims that our approach will suffer from finite-size effects (although our comparison for system sizes didn't seem to suggest that was true).
@ramess101 we tried their method to calculate the dHv. Results where changed slightly for molecules with intrabonded energy, which would result in change in the error value. Maybe we can calculate dHV with their method, and plot them together to see if the dHv for cyclohexane changes or not.
@msoroush
OK, that might be the cause of the discrepancy. We assume that intrabonded is the same in liquid and vapor phase right? This certainly could contribute around 1% to dHv. Do you have those dHv values readily available?
The TAMie method for calculating heat of vaporization appears to be more resistant to finite size effects in the near critical region. It doesn't seem to offer better performance with respect to finite size effects at lower temperatures. We did see a small shift in the calculated error going from our method to the TAMie way of doing things. It might be due to the bonded interactions, but we don't have conclusive evidence, yet. @msoroush Could you post Niloo's figures for methane and butane to illustrate the difference? The only way to really tell would be to compare to GEMC simulations where we calculate dHv using bonded and non-bonded interactions. Unfortunately, the uncertainty in those data is a lot larger than the GCMC results...
@jpotoff @msoroush
OK, so the difference at low T is not due to finite size effects. But it could be caused by how dHv is being computed, namely, their approach makes no assumption about intramolecular energies cancelling between the two phases.
Yeah, the fact that there might be some discrepancy between GEMC and GCMC is not totally unexpected. I don't think we need to run GEMC simulations for this study, although it could be worth looking into.
For the purpose of this review, we just need to show that the deviation in TAMie is: 1) Not due to finite-size effects 2) Caused by different analysis methods, where our approach assumes intramolecular energies are the same in vapor and liquid. Right?
@ramess101 @jpotoff
Butane:
Methane:
@msoroush
So the left panels use our method and the right panels is TAMie method? Discrepancy in methods appear larger for butane, supporting the intramolecular theory
On Friday, March 8, 2019, Mohammad Soroush Barhaghi < notifications@github.com> wrote:
Butane: [image: butane-averages] https://user-images.githubusercontent.com/13317444/54060637-7b0b1080-41cb-11e9-8614-97a3e9404bf5.png
Methane: [image: methane-averages] https://user-images.githubusercontent.com/13317444/54060669-96761b80-41cb-11e9-96aa-106418e94a91.png
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/ramess101/JCED_FOMMS_Manuscript/issues/26#issuecomment-471105175, or mute the thread https://github.com/notifications/unsubscribe-auth/AWUvhBUFatDs7AKMy4YU_i9tx9E3Jh07ks5vUurXgaJpZM4bSc9v .
@ramess101
Here are the results for rcut10 and 14 for TAMie rcut10.zip rcut14.zip
@msoroush
The rcut14 looked marginally better, and I think we should include the rcut14 results. The rcut14.zip folder was missing the Average_Results_m1.dat file. Could you re-zip rcut14 with this file and re-upload it?
@ramess101 Here are the average results rcut14.zip
@msoroush @jpotoff
OK, here are the comparisons for our TAMie simulations:
rcut = 10 A
rcut = 14 A
@msoroush @jpotoff
rcut = 14 A is slightly better, but still a clear discrepancy.
Is this unexpected? Of course our DeltaHv analysis is different, but would you expect the disagreement for rholiq and Psat? Could the difference in Psat be related to the different approaches for computing absolute pressures (recall Issue #29)? But this shouldn't affect rholiq, right? Because equating the pressures in the two phases should still be the same.
Instead, could the difference in rholiq be related to how the reweighting scheme only includes the non-bonded energies? In other words, would the phase equilibria change some if the histograms used the total energy (with intramolecular)?
@msoroush @jpotoff
In case you didn't know, the response to the reviewer comments is due today. I think this is really the only issue that we still need to resolve. Please let me know if you have any thoughts for how we can adequately explain the discrepancy between our TAMie results.
@msoroush
Would you be able to compute DeltaHv for our TAMie simulations (14 A) using the TAMie methodology? Including those values might be helpful.
The agreement for rho_liquid and dHv, is reasonably decent; at least we have overlap of the error bars. It's the vapor density and vapor pressure that are way off. Given that the vapor densities are different, the difference in the vapor pressure is unlikely to be due to the different techniques used to get the reference state. Without seeing their simulation and histogram analysis code, it is impossible to explain why the results are different with any certainty.
Looking at Gross's data, if those were our data, I would concerned about the large error bars, especially near the critical point. That seems very odd.
@jpotoff
Good point about the vapor densities. I will plot up the Zvap really quick to see what those look like.
@jpotoff
However, the rholiq deviations are actually not within the error bars at most temperatures (compare "TAMie, This Work" with "TAMie, Weidler et al." DeltaHv are closer though
However, the rholiq deviations are actually not within the error bars at most temperatures (compare "TAMie, This Work" with "TAMie, Weidler et al." DeltaHv are closer though
Right. I was incorrectly comparing our results and TAMie instead of comparing TAMie (us) with TAMie (published). We may just have to throw our hands up on this one since the TAMie calculations are done with a non-open source simulation engine.
@jpotoff
Yeah, currently I just try to pass this off as relatively small deviations for rhol and DeltaHv (less than 0.5%)
@msoroush @jpotoff
Here is a comparison for the Zvap values:
I think this shows that the deviation in Psat is due to the deviations in rhovap because we get the same Z, right?
I think my brain just exploded. Pvap and rho_vap are different between our results and Gross, but we get cancellation of errors to give an exact match for Z=PV_gas/RT? I'm having a hard time reconciling these results.
@jpotoff
My mindset was that this demonstrates that the ideal gas is not to blame for the discrepancy in Psat. Because we still get Z converging to 1, the pressure calculations are consistent with the densities, right?
@ramess101 What if we plot error in Z? I think we will see some deviation.
@msoroush
Yeah, but we can tell that the error bars do overlap. Of course, the Z error bars might be too large since I assumed independence between rhovap and Psat...let me fix that first
I can't do anything about how the error bars are reported for Weidler. I could do it rigorously for the 20 replicates, but it might not be worth my time.
@msoroush @jpotoff
That resolves the comment about reference states.
Still having a hard time resolving that rho_vap and Pvap don't agree with us. I guess we have to accept that we aren't going to be able to answer that question conclusively. Then again, we weren't asked to do that by the reviewers. If we stick with just Z and put the plot in the supporting info, or in the response to the reviewers, that should be sufficient.
@jpotoff
Agreed.
@msoroush @jpotoff
This is actually in response to comment #28 Any suggestions or concerns for this portion of the response:
This looks good. We've gone above and beyond addressing the reviewer comments.
@jpotoff
Thanks, agreed!
On Monday, March 11, 2019, Jeffrey Potoff notifications@github.com wrote:
This looks good. We've gone above and beyond addressing the reviewer comments.
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/ramess101/JCED_FOMMS_Manuscript/issues/26#issuecomment-471752488, or mute the thread https://github.com/notifications/unsubscribe-auth/AWUvhJWjtUUWauT2kyNpnHBseiPJM1ZKks5vVtFbgaJpZM4bSc9v .
@msoroush
Do you have the state points (chemical potentials) that you used to simulate TAMie (just for 3.5 nm box and 1.4 nm cutoff simulations)? Just needed for the SI
On Monday, March 11, 2019, Richard Messerly ramess101@gmail.com wrote:
@jpotoff
Thanks, agreed!
On Monday, March 11, 2019, Jeffrey Potoff notifications@github.com wrote:
This looks good. We've gone above and beyond addressing the reviewer comments.
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/ramess101/JCED_FOMMS_Manuscript/issues/26#issuecomment-471752488, or mute the thread https://github.com/notifications/unsubscribe-auth/AWUvhJWjtUUWauT2kyNpnHBseiPJM1ZKks5vVtFbgaJpZM4bSc9v .
@ramess101 Here are the state point for TAMie
RUNS
===============================================
- # T(K) Chem. Pot CBMC_LJ_1st CBMC_LJ_nth
-----------------------------------------------------------
R 1 450 -4358 8 4
R 2 510 -4358 8 4
R 3 550 -4358 8 4
R 4 510 -4165 8 4
R 5 470 -4036 8 4
R 6 430 -3923 16 10
R 7 400 -3852 16 10
R 8 370 -3793 16 10
R 9 350 -3762 16 10
@msoroush
Perfect
On Monday, March 11, 2019, Mohammad Soroush Barhaghi < notifications@github.com> wrote:
@ramess101 https://github.com/ramess101 Here are the state point for TAMie
RUNS
T(K) Chem. Pot CBMC_LJ_1st CBMC_LJ_nth
R 1 450 -4358 8 4 R 2 510 -4358 8 4 R 3 550 -4358 8 4 R 4 510 -4165 8 4 R 5 470 -4036 8 4 R 6 430 -3923 16 10 R 7 400 -3852 16 10 R 8 370 -3793 16 10 R 9 350 -3762 16 10
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/ramess101/JCED_FOMMS_Manuscript/issues/26#issuecomment-471770911, or mute the thread https://github.com/notifications/unsubscribe-auth/AWUvhB_JYiIv87GPTcgUfmbO_9bFIIYTks5vVt1agaJpZM4bSc9v .
@mrshirts @jpotoff @msoroush
The reviewer never clarifies what paper has a Mie force field for cyclohexane. @jpotoff @msoroush Any ideas which paper the reviewer is referring to? Who are "these authors" that are included in our reference list? I can't find one for TAMie (the only other Mie family we cite). I know there are parameters obtained with SAFT-gamma that use fractional lambda values and some single-cite cyclohexane Mie models. But I don't think we have any data for those models.