From: Mattia Felice Palermo (mattiafelice.palerm2_at_unibo.it)
Date: Fri Oct 23 2015 - 08:57:37 CDT
Dear Zhe,
thank you very much for your accurate answer.
> Could you specify how you calculate the dielectric constant?
It has been computed with a inhouse program using the classical
expression for the dielectric constant from the average dipole moment
<M^2> as in your JCTC paper.
> The dielectric constant of TIP3P-EW water is 89 (D. J. Price, C. L.
> Brooks, J. Chem. Phys. 2004, 121, 10096), while you are getting 96.8
> with PME and 83.3 with MSM. You can find the way we calculated the
> dielectric constant on page 773 of our JCTC paper.
We suspect that the higher value found in our simulation with PME might
be related to the size of the sample (N=1000 in the TIP3P-EW original
paper against N=11000 in our simulations). However the value is not far
from the literature value of 89 (TIP3P-EW, model B) and from the
value of 104 reported in your JCTC paper. We also checked other
properties e.g. the oxygen-oxygen radial distribution function and
the average module of the molecular dipole moment, and they are in
agreement with the expected results for the model.
> It is very surprising to see a dielectric constant of water to be
> 258. How is the water structure? With such a large dielectric
> constant, the water structure should be very unphysical.
The first surprising fact is that the dielectric constant of bulk
water differs if the MD simulations is carried out with PME and MSM
methods. The dielectric constant of water with value 258 was obtained
for a thin film of water supported on silicon dioxide and using the
MSM method, while when using the PME method we obtain a dielectric constant
of 91.5. Visual inspection of the samples through VMD did not show any
unphysical structure of the water. We also computed the oxygen-oxygen
radial distribution g(r)_O-O for the films, which you can find attached, and
it does not show any significant difference between the two simulations.
> Also, could you give us a little bit more details about your VDW
> cutoff scheme? As reported in TIP3P-EW’s original paper, a model
> that incorporates a long-range correction for truncated VDW will
> give a dielectric constant of 76 instead of 89. Are you using the
> same cutoff scheme for both PME and MSM?
Yes we used the same cutoff scheme for both simulations. More in detail,
these are the settings for the MSM simulations:
cutoff 12.0
switching on
switchdist 10.0
pairlistdist 14.0
timestep 2.0
nonbondedFreq 1
fullElectFrequency 2
stepspercycle 10
langevin on
langevinDamping 1
langevinTemp 298.15
LangevinPiston on
LangevinPistonTarget 1.01325
LangevinPistonPeriod 250
LangevinPistonDecay 100
LangevinPistonTemp 298.15
cellBasisVector1 63.340 0.0 0.0
cellBasisVector2 0.0 63.340 0.0
cellBasisVector3 0.0 0.0 63.340
rigidBonds all
MSM yes
The same setting were used for the PME simulations, except of course for
the calculation of electrostatics:
PME yes
PMEGridSpacing 1.5
Thank you very much for your help!
P.s.: I changed the subject of this email to have all future
contributions sorted in the same thread. Sorry for the confusion!
--
Mattia Felice Palermo - Ph.D.
Università di Bologna
Dipartimento di Chimica Industriale "Toso Montanari"
> ________________________________________
> From: Zhe Wu [zephyrbless_at_gmail.com]
> Sent: mercoledì 21 ottobre 2015 21.16
> To: Mattia Felice Palermo
> Cc: namd-l_at_ks.uiuc.edu; David Hardy
> Subject: Re: namd-l:
> simulations
>
> Hi Mattia,
>
> Could you specify how you calculate the dielectric constant? The
> dielectric constant of TIP3P-EW water is 89 (D. J. Price, C. L. Brooks,
> J. Chem. Phys. 2004, 121, 10096), while you are getting 96.8 with PME
> and 83.3 with MSM. You can find the way we calculated the dielectric
> constant on page 773 of our JCTC paper.
>
> It is very surprising to see a dielectric constant of water to be 258.
> How is the water structure? With such a large dielectric constant, the
> water structure should be very unphysical.
>
> Also, could you give us a little bit more details about your VDW cutoff
> scheme? As reported in TIP3P-EW’s original paper, a model that
> incorporates a long-range correction for truncated VDW will give a
> dielectric constant of 76 instead of 89. Are you using the same cutoff
> scheme for both PME and MSM?
>
> Hopefully, the above helps.
>
> Best,
> Zhe
>
>
>
> Zhe Wu
> Klaus Schulten's Group,
> Postdoc Fellow in NSF's Center for the Physics of Living Cells,
> University of Illinois Urbana-Champaign
> Beckman Institute, RM 3125,
> 405 N. Mathews Ave. Urbana, IL 61820
> Office: 217-224-3160
> http://www.ks.uiuc.edu/~zhewu/
>> ________________________________________
>> From: owner-namd-l_at_ks.uiuc.edu [owner-namd-l_at_ks.uiuc.edu] per conto di
>> Mattia Felice Palermo [mattiafelice.palerm2_at_unibo.it]
>> Sent: mercoledì 21 ottobre 2015 11.07
>> To: namd-l_at_ks.uiuc.edu
>> Subject: namd-l:
>>
>> Dear developers of NAMD,
>>
>> I am running several simulations of TIP3P-EW water using the new MSM
>> method to evaluate the electrostatic interactions. I then computed the
>> dielectric constant of water but it does not compare well with that
>> obtained using the classic PME method. For the MSM simulation I have
>> only specified the keyword:
>>
>> msm yes
>>
>> and left all the default options. The resulting dielectric constants of
>> bulk water (~11k molecule) are 83.3 with MSM and 96.8 with the PME
>> method. Even if the MSM value is closer to the experimental value, the
>> one obtained with PME is in line with the expected value for this
>> specific model of water. Furthermore, I would have expected a similar
>> value for both methods, consistently with the benchmark results
>> published in JCTC 11, 766 (2015).
>>
>> Then I considered a thin film of TIP3P-EW water on top of an amorphous
>> silica surface (2D periodic). In this case I have specified the
>> following options for the MSM simulation:
>>
>> cellBasisVector1 57.108000 0.0 0.0
>> cellBasisVector2 0.0 57.108000 0.0
>> msm yes
>> MSMGridSpacing 2.8554
>> MSMzmin 8.0
>> MSMzmax 208.0
>>
>> The grid spacing ensure that a even number of points divides the basis
>> vectors, while the z_min and z_max values are big enough to contain the
>> sample. An harmonic potential wall was also included to contain escaping
>> water molecules. The computed dielectric constant of water is now 258,
>> while in the simulation with PME, the dielectric constant is 91.5. The
>> potential wall was maintained in the PME simulation for better comparison.
>>
>> Do you have any idea what may be wrong in my simulations using the MSM
>> method?
>>
>> Thank you very much for the help!
>>
>> --
>> Mattia Felice Palermo - Ph.D.
>> Università di Bologna
>> Dipartimento di Chimica Industriale "Toso Montanari"
This archive was generated by hypermail 2.1.6 : Thu Dec 31 2015 - 23:22:10 CST