Download as pdf or txt
Download as pdf or txt
You are on page 1of 60

Page 1 of 60 AIChE Journal

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
Predicting NRTL Binary Interaction Parameters from

Molecular Simulations

Ashwin Ravichandran, Rajesh Khare*, Chau-Chyun Chen*

Department of Chemical Engineering, Texas Tech University, Box 43121

Lubbock, TX 79409

Email: rajesh.khare@ttu.edu; chauchyun.chen@ttu.edu

This is the author manuscript accepted for publication and has undergone full peer review but has not been
through the copyediting, typesetting, pagination and proofreading process, which may lead to differences
between this version and the Version of record. Please cite this article as doi:10.1002/aic.16117.

This article is protected by copyright. All rights reserved.


AIChE Journal Page 2 of 60

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
Abstract

A predictive approach for calculating the binary interaction parameters ( ) of the nonrandom

two liquid (NRTL) local composition model is developed, combining molecular simulations with

the two-fluid theory. The binary interaction parameters are determined for the following three

sets of model binary mixtures: water + methanol, methanol + methyl acrylate, and water +

methyl acrylate. For each binary mixture, the interaction parameters are expressed in terms of

molecular size and strength of interactions, which are in turn, calculated from molecular

simulations. We show that the binary interaction parameters determined from simulations are in

qualitative agreement with those estimated from regressing experimental data. The major factors

that determine the binary interaction parameters are outlined based on simple thermodynamic

arguments for each mixture.

2
AIChE Journal
This article is protected by copyright. All rights reserved.
Page 3 of 60 AIChE Journal

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
Introduction

Designing energy efficient and sustainable industrial processes requires reliable

prediction of thermodynamic and phase equilibrium properties. Molecular thermodynamic

models developed over the past several decades have been successful in quantifying vapor-

liquid, liquid-liquid, and vapor-liquid-liquid equilibria of organic mixtures1-3 and of electrolyte

systems4,5. In some cases, these models have even been applied to large molecules such as

polymers6 and biomolecules7. Most of these semi-empirical models predict the thermodynamic

properties based on the local composition of the mixture. Notable among these models are the

Wilson model8, the nonrandom two liquid (NRTL)9 model, and their variants1,10.

In a binary mixture of species  and , the local composition models relate the energy

required for the formation of two types of local domains ( centered and  centered domains that

are surrounded by both  and  molecules) to the excess Gibbs energy of the system9. In general,

the expression for local composition for an  centered domain can be stated as:

  
=
 
 
(1)

Here  and  are the local mole fractions of type  molecule around type  molecule and type 

molecule around type  molecule respectively,  and  are the bulk mole fractions of type  and

 molecules,  is the non-randomness factor, and  is the binary interaction parameter which is

the input to these local composition models. The binary interaction parameter quantifies the

nature of molecular interaction between different chemical species in a mixture. High positive

values of τ indicate unfavorable interaction between the solute and the solvent molecules (e.g.

3
AIChE Journal
This article is protected by copyright. All rights reserved.
AIChE Journal Page 4 of 60

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
interaction between hydrophilic and hydrophobic groups) while negative numbers indicate

favorable interactions (e.g. interaction between hydrophilic and polar groups)1. The non-

randomness factor is usually set to 1.0 or 0.2-0.5 for the Wilson model and the NRTL model,

respectively8,9. An expression for the local composition around a  centered domain can be

written in a similar way, resulting in the interaction parameter  . It must be noted that in the

local composition models, the binary interaction parameters are not symmetric, i.e.  ≠  9.

Although the expressions for the local composition in Wilson and NRTL models are

written in a semi-heuristic fashion, a rigorous interpretation of the local composition approach of

Eq. 1, is provided by the two-liquid theory11-13. Kemény and Rasmussen11, used two-liquid

theory to formulate the partition function of a binary mixture as a product of the partition

functions of two hypothetical fluid domains. The resulting partition function was used to derive

a Wilson-like expression relating the energetic interaction between the two liquids (), bulk

concentration, and the local composition. Brandani and Prauznitz14 used similar arguments to

find the probability of forming hypothetical fluid domains by solving a combinatorial problem,

thereby arriving at an expression analogous to the Wilson model. The validity of the local

composition approach has also been confirmed independently by many experimental studies,

most of which used spectroscopic techniques to measure the local compositions15-17, that were

then compared with the local fractions predicted by the NRTL model. In one such study, Phillips

and Brennecke17 measured the local composition of various mixtures of organic solvents. They

concluded that, for binaries without any specific chemical interactions (e.g. hydrogen bonding

interactions), the local compositions predicted by the NRTL model agree closely with

experiments. Owing to their simplicity in application and rigorous treatment of the underlying

4
AIChE Journal
This article is protected by copyright. All rights reserved.
Page 5 of 60 AIChE Journal

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
physical phenomena, local composition models are widely used in the design of industrial

processes18.

Accurate values of the binary interaction parameters are crucial for the success of these

thermodynamic models. It is customary to obtain the values of the binary interaction parameters

by fitting the model to the experimental data, which include among others, vapor pressure,

osmotic coefficients19, phase equilibrium20, solubility, and excess enthalpy data. However, the

regression based approach is not always feasible as it requires many reliable experimental data

sets to calculate the binary interaction parameters. Also, careful analysis of data uncertainty and

selection of the optimization approach are essential, as such regression problems can often result

in different values of the binary interaction parameters due to the existence of multiple local

minima21 in the parameter estimation problem.

Thus, it is desirable to predict the binary interaction parameters from first principles, as

such a technique can circumvent the problems associated with the regression based approach and

the subsequent need for availability of many reliable experimental data sets. To our knowledge,

there have been very few attempts in the literature for calculating the interaction parameters of

the activity coefficient models from a fundamental approach. Most of these approaches focus on

the parameters of the UNIQUAC model22-24, where the model parameters are determined directly

using the interaction energies of either molecular pairs22 or small clusters of molecules23.

Neiman et al.24 noted that a similar methodology can be applied for calculating the binary

interaction parameters of the NRTL model using the bulk intermolecular energy difference

between the liquid and the gas phase. They concluded that the interaction parameters obtained

using the bulk intermolecular energies cannot be used with the NRTL model to predict the phase

diagrams of binary mixtures. In another recent work by Seyf and Haghtalab,25,26 the local

5
AIChE Journal
This article is protected by copyright. All rights reserved.
AIChE Journal Page 6 of 60

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
coordination number obtained by integrating the radial distribution function was used to estimate

the binary interaction parameters. However, that approach is not predictive as it requires an

adjustable parameter, the cut-off length for integrating the radial distribution function, this cut-

off length was calculated from the knowledge of the experimental activity coefficient data.

In this work, we combine two-fluid theory27 with molecular simulations to determine the

binary interaction parameters of the NRTL model. The approach we present here is purely

predictive and considers the local structural and energetic information for calculating τ. The

binary interaction parameters are calculated from the knowledge of molecular size and

interaction strength, both of which are obtained from molecular simulations. The following three

sets of binaries were chosen as model systems to validate the method proposed here: water +

methanol (fully miscible binary), methanol + methyl acrylate (micro-clusters forming binary),

water + methyl acrylate (partially miscible binary). The reason for choosing these mixtures are

two-fold; one, experimentally determined interaction parameters are available for each of these

binaries20 that can be used for validating our approach, and two, the proposed binaries exhibit a

wide range of thermodynamic characteristics (hydrophilicity/hydrophobicity, hydrogen bonding

capabilities, etc.), which serve as a stringent test for the validity of the proposed approach.

Rest of the article is organized as follows: Section II presents a brief description of the

two-fluid theory, followed by the details of the molecular dynamics (MD) simulations. The

calculated interaction parameters and their molecular significance are discussed in Section III,

followed by conclusions.

Theoretical Background and Simulation Details

Two-Fluid Theory

6
AIChE Journal
This article is protected by copyright. All rights reserved.
Page 7 of 60 AIChE Journal

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
Two-fluid theory of Brandani and Prausnitz14,27,28 relates the binary interaction

parameters of a Wilson-like model with the molecular size, the effective interaction strength

between different species of the mixture, and the size of the first neighbor shell. In this work, we

extend the two-fluid theory to NRTL model and apply molecular simulations to determine the

binary interaction parameters. A brief account of the two-fluid approach used in this work is

presented herein; the reader is referred to a series of articles published by Brandani and

Prausnitz14,27,28 for a detailed discussion of the approach.

To begin, consider a binary mixture of species  and . Further, consider a domain in the

binary with molecule of type  at the center that is locally coordinated by z molecules out of

which  are of type  and ( − ) are of type 11,27. The probability of arranging  molecules of

type  and ( − ) molecules of type , (randomly choosing them from an infinite pool of  and 

molecules) in a domain around a central molecule of type  is:

 =     (2)

where  is the combinatorial factor,  and  are the probabilities of choosing an  type or a 

type molecule, respectively. It must be noted that,  denotes the probability of forming an 

centered domain without taking into account the interaction energies between the molecules of

species  and . To account for the preferential interactions between the components, the

distribution in Eq. 2 is weighted by the local potential function  , which could be written as27:

−  + ( − )  !


 =
(3)
"# () $

7
AIChE Journal
This article is protected by copyright. All rights reserved.
AIChE Journal Page 8 of 60

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Here, %# () & is the radius of the coordination shell with molecule of type  at center,  is the

Author Manuscript
molecular diameter and  is the interaction strength between the components of the binary

mixture. Note that the interaction strength,  is symmetric, i.e.  =  at a given concentration.

The interaction potential represented by the above form assumes that the local interactions are

only governed by the molecules within the first neighbor shell of the center molecule. This is the

central assumption of the local composition approaches.

The distribution function for a non-random arrangement (including specific interactions)

of a local domain is given by weighting  by the Boltzmann factor:

^   
()*
   

 =
(4)
+

^
where  denotes the domain distribution function of an interacting system, and +  is the

normalization factor that is analogous to a local partition function, with , = 1/#/.

+ = 0    


()*


(5)

12

Substituting the expression for potential from Eq. 3 into the Eq. 4 and rearranging we get:

^ 1
 =   ( ′) ( ′)  
+
(6)

where

8
AIChE Journal
This article is protected by copyright. All rights reserved.
Page 9 of 60 AIChE Journal

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
, 

Author Manuscript  ′ =     
5 6 ()  7
(7)
"# $

, 
 ′ =   
5 6 7
(8)
"# () $

By definition,  and  are the bulk mole fractions of components  and , respectively. Hence,

by construction of Eq. 7 and Eq. 8,  ′ and  ′ can be interpreted as the local compositions of

molecules of types  and  around a center molecule of type , respectively. Taking ratio of Eq. 7

and Eq. 8, one obtains:

< <
9: ; :  ;  =
  8 ( ?
(9)
<
=
5
%> () &
 

Equation 9 is thus equivalent to the NRTL equation (Eq. 1).

Comparing the mathematical forms of Eq. 9 and Eq. 1, the binary interaction parameter

of the NRTL model can be written as:

,   −  


 =
(10)
 "# () $

Here,  is the non-randomness factor. We note that a similar expression was derived by

Brandani and Prausnitz14 for calculating the interaction parameter of the Wilson model. An

equivalent expression for the interaction parameter with the molecule of type  as the center of

the domain can be derived in a similar fashion:

9
AIChE Journal
This article is protected by copyright. All rights reserved.
AIChE Journal Page 10 of 60

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
,   −  

Author Manuscript  =


(11)
 "# () $

Equations 10 and 11 are the basis for calculating the NRTL binary interaction parameters

from the two-fluid theory. Note that in all of the double indexed variables, consistent with the

notations of NRTL model, the second index denotes the molecule at the center of the domain. In

what follows, %# () & is denoted as # () for the sake of simplicity in text.

Calculation Methodology

The size of the molecule (), interaction strength (), and the neighbor cage radius (#)

which are required for calculating the binary interaction parameters are obtained from the short-

range structure of the mixture as determined from molecular simulations. The interaction

potential defined by the two-fluid theory (Eq. 3) is of square well type, consistent with the local

composition approach. Hence, the simulation results for short-range structure and interactions

(potential of mean force) from MD simulations should be mapped onto square well type

potential, in order to determine the binary interaction parameters. In this work, we have

followed two different approaches for evaluating , , and # from liquid structure as follows:

Approach I
The molecular diameter () of each species is calculated from the pure component radial

distribution function (RDF). For this purpose, the position of the first peak in the RDF was taken

as the molecular diameter of the pure component29 ( or  ). Using these values,  was

; @;
calculated as the average of pure component radii ( = ). An alternate way to obtain 
A

would be to define it as the location of the first peak in the mixture RDF between components 

10
AIChE Journal
This article is protected by copyright. All rights reserved.
Page 11 of 60 AIChE Journal

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
and . The arithmetic mean approach chosen here is consistent with the Lorentz-Berthelot

Author Manuscript
mixing rules30 that are used in our simulations for evaluating interactions between species  and

. In fact, the force field parameters were optimized using such rules. Furthermore, for mixtures

where the two components are not fully miscible, first peak in the RDF is not well defined and

hence evaluation of  from the - RDF becomes difficult.

The radius of the neighbor cage (# () ) was calculated from the simulation of binary

mixtures of components  and . Simulations were performed at different concentrations of the

binary mixture and the location of the first minimum in the RDF involving the molecules of

interest was taken as the cage radius. Since RDF can be defined between different groups in a

molecular system, # () was defined as the largest of the neighbor shell locations of all such group

RDFs across all concentrations (i.e. largest # () value in the system over the entire concentration

range). Such a definition for the calculation of  includes, the length scale up to which the short-

range structure of the liquid extends.

Finally, the effective molecular interaction strength () was defined as the depth of the

first minimum in the potential of mean force (PMF) required for separating the given molecular

pair. Since the net interactions between molecules determine the local structure12 in a liquid

phase,  was calculated from the PMF rather than using the bare pair-wise interactions. While

calculating the interaction strength, it is necessary to ensure that the solute and solvent are

homogeneously mixed. Hence, the PMF between species  and  was calculated in the dilute

limit of the center molecule. In the calculation of  , both  and  were calculated from the

binary mixture that was very dilute in component . Similarly,  and  , that are required to

calculate  , were obtained in the limit of dilute . Defining the interaction strength in the dilute

11
AIChE Journal
This article is protected by copyright. All rights reserved.
AIChE Journal Page 12 of 60

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
limit of the solute also ensures that  and  have no contribution from the solute-solute

Author Manuscript
interactions.

Approach II

The expression for  from the two-fluid theory (Eq. 10) can be rearranged as:

  () 


  
CD ()  E %# & − F ()  G %# () & H
"# $ "# $
(12)
,
 = 6 7
 "# () $

Here, # () and # () denote the size of the neighbor shell with molecule of type  at the center

that is surrounded by molecules of type  and type , respectively. # () and # () were evaluated

from the first minimum in the RDFs for  and  pairs respectively and # () was then obtained as

# () = maxL# () , # () N, consistent with Approach I.

The two quantities in the square brackets in the numerator are the square well potentials

as used by Brandani and Prausnitz for the  and  pairs27. These were evaluated by following

the suggestion by Hirschfelder et al.31 for connecting the square well potential with the usual

Lennard-Jones (LJ) type interactions. Specifically, it was suggested that the strength of the

square well interaction should be 0.56 times the barrier height of the LJ type interaction.

Replacing the terms in the square brackets of Eq. 12 using this approach, we get:

12
AIChE Journal
This article is protected by copyright. All rights reserved.
Page 13 of 60 AIChE Journal

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
()  

Author Manuscript , O0.56 %#


 = 6 7
& − 0.56 %# () & T (13)
 "# () $

The values of  and  required to calculate  from the above expression were obtained using

the same procedure as described in Approach I, i.e. depth of the first minima in the PMF.

Molecular Simulations

The components of interest in this work are water, methanol (MeOH), and methyl

acrylate (MA). Pure component simulations of each of these species were performed to

determine the molecular diameter (). Further, to obtain the size of the neighbor cage (# () ),

three sets of binary mixtures, (water + methanol, water + methyl acrylate, and methanol + methyl

acrylate) were simulated at different concentrations of the mixtures (0.002-0.998 mole fraction of

one of the components). All simulations were carried out using the TraPPE-UA potentials for

acrylate and alcohol32,33 while the water molecules were represented by the SPC/E model34. The

TraPPE-UA force field is known to predict the phase equilibrium properties of organic liquids

with good accuracy and hence it is an appropriate choice for the present calculations.

Simulations were performed using the LAMMPS package35. One small change made in the

force field was that, while TraPPE-UA fixes the bond length values of all pairs of atoms at their

equilibrium values, we used strong harmonic constraints to represent the bond stretching

interactions, with force constants for the corresponding bonds taken from the General AMBER

Force Field (GAFF)36. TraPPE-UA force field was tested for its consistency by comparing the

13
AIChE Journal
This article is protected by copyright. All rights reserved.
AIChE Journal Page 14 of 60

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
densities of the pure components and the binaries of water + MeOH and MeOH + MA with

experimental data. Overall there was a good agreement, with the simulated densities

representing the experimental data within 2% error. Most of the MD simulation systems

reported in this study consisted of approximately 9000 united atoms. However, simulations at

very low concentrations (0.002 and 0.998 mole fractions) consisted of smaller systems

containing 3000 to 6000 united atoms. Since the properties of interest here are pertaining to the

local structure of the molecular system, the chosen system sizes were considered to be sufficient

for these calculations. Simulations were performed at 303.15 K and 1 bar in the constant number

of particles, pressure, and temperature (constant NPT) ensemble using Nosé-Hoover thermostat

and barostat37. All binary simulations were carried out for a duration of 20 ns with a time step of

1 fs. The first 5 ns of the trajectory were discarded as initial equilibration period and the rest of

the trajectory was used for subsequent calculations.

The PMF between two species, which is required for obtaining the strength of the

effective interaction (), was calculated from umbrella sampling technique38. As mentioned

above, these umbrella sampling simulations were performed on dilute mixtures. The binary

mixtures that were equilibrated for 20 ns using MD simulations were used as the starting

configurations for the umbrella sampling simulations. The reaction coordinate (U) for these

calculations was chosen as the center of mass distance between the molecular pair of interest.

Harmonic biasing potential, with a spring constant of 3012.48 kJ/mol nm2 (7.2 kcal/mol Å2) 39,

was applied to restrain the molecular pair at the desired value of the reaction coordinate. The

molecules were moved towards each other starting from a distance of 1.2 nm. The size of each

umbrella sampling window was chosen as 0.03 nm, resulting in a total of 34 windows. Umbrella

sampling simulations were performed using the COLVARS module40 in LAMMPS. Under the

14
AIChE Journal
This article is protected by copyright. All rights reserved.
Page 15 of 60 AIChE Journal

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
biasing potential, each window was sampled for 3 ns out of which the first 2 ns were discarded

as the initial equilibration period. The last 1 ns of the trajectory at each reaction coordinate value

was used to calculate the potential of mean force using the Weighted Histogram Analysis

Method (WHAM)41, as implemented in the WHAM code from Grossfield Laboratory42. Finally,

the entropic term 2#/ln(Y) was added back43,44 and the PMF was shifted to zero at a separation

distance of 1.2 nm, to be consistent with its definition.

Data Regression

The binary interaction parameters obtained from molecular simulations were compared

with those obtained by regressing experimental data. For this purpose,  values for the three

binaries studied here were determined using data regression. ASPEN Properties software,

version 8.8 was used for calculating the interaction parameters45. Vapor-liquid equilibrium,

liquid-liquid equilibrium, and azeotrope data were primarily considered for data regression. The

standard error on the regressed interaction parameters was estimated using error propagation

technique46. Further details of the regression are provided in the Supplementary Material.

Results and Discussion

Molecular Diameters

As described in the previous section, the molecular sizes were obtained from the

corresponding pure component RDFs. The effective molecular sizes () were taken from the

location of the first peak in the pure component RDF as shown in Fig. 1. Accordingly, the size

of a water molecule calculated here to be 0.267 nm (based on the water oxygen RDF) agrees well

with the value reported in a previous study47. Size of a methanol molecule was found to be 0.424

15
AIChE Journal
This article is protected by copyright. All rights reserved.
AIChE Journal Page 16 of 60

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
nm in a similar fashion using the location of the first peak in the RDF between methyl groups.

The value of the effective molecular diameter of a methanol molecule was reported to be 0.42

nm by Perera et al.29 Thus, the value obtained in the current study matches with that reported in

the literature. Similarly, the size of a methyl acrylate molecule was determined to be 0.595

nm from the location of the carbonyl carbon peak.

Water (1) + Methanol (2) Binary System: Size effects determine the interaction parameter

for the completely miscible mixture

The first binary mixture studied in this work is that between water and methanol. RDFs

were determined for water + methanol mixtures over a wide range of concentrations. We

obtained # () as the size of the first neighbor shell with molecule  as center from those RDFs.

Correspondingly, # (Z[\]^) was calculated as 0.534 nm and # (_]`a) as 0.597 nm. Table 1.a

summarizes the sizes of the neighbor shells with water and methanol as center. Note that the

first hydration shell with methanol as the center molecule is slightly larger than the shell with

water as the center molecule due to the difference in their molecular sizes. Also, the value of

# () does not change significantly with the concentration of the mixture. The  and  values

in Eqs. 10, 11, 13 were obtained from the PMF curves for the water + methanol system (see Fig.

2).

The binary interaction parameters for water + methanol mixture obtained from

simulations are listed in Table 1.b, along with the  values determined from regressing

experimental data. The value of the non-randomness factor was set to 0.3 for all the calculations

(both MD and regression  calculations)9. However in practice, the non-randomness factor is

usually set between 0.2-0.5 depending on the chemical mixture9. The value of  was

16
AIChE Journal
This article is protected by copyright. All rights reserved.
Page 17 of 60 AIChE Journal

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
consistently chosen as 0.3 in this work to compare the binary interaction parameters predicted

using our method with those in the literature20 (calculated using data regression). Furthermore,

we focus on the role played by molecular sizes and interaction strength in determining the binary

interaction parameters and hence the non-randomness factor was conveniently fixed at 0.3 for all

mixtures. The uncertainty in  was estimated by propagating the error46 in the interaction

strength, . As the uncertainties in the molecular and hydration shell sizes are generally small,

they were not considered in the error estimation calculation.

The binary interaction parameters estimated from molecular simulation in combination

with the two-fluid theory (using both of the proposed approaches) show reasonable agreement

with the regressed parameters, within the limits of statistical uncertainties. It can be noted that

the interaction parameters found by regressing experimental data from two different sources (this

work and literature20), differ significantly in their magnitude. This is common occurrence in

parameter estimation using data regression, where the final converged solution depends on

various factors such as the data set used in the regression, uncertainty in the data, and the

optimization algorithm which can converge to different local minima based on the initial

conditions21,48. Table 1.c shows the infinite dilution activity coefficient values of water +

methanol binary calculated using the NRTL equations with the binary interaction parameters

obtained from both MD simulations and by regressing experimental data. The uncertainties in

those values were calculated by propagating the error in the binary interaction parameters. The

values of ln (b c ) calculated using τ obtained from Approach I are in reasonable agreement with

those predicted using τ from experimental data regression, within statistical uncertainties.

However, ln (b c ) obtained from Approach II shows considerable deviation from the other

values. We attribute this observation to overprediction of Ad by Approach II. We also note that

17
AIChE Journal
This article is protected by copyright. All rights reserved.
AIChE Journal Page 18 of 60

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
the values of infinite dilution activity coefficient are sensitive to the binary interaction parameter

values.

The binary interaction parameter has contributions from both local energetic and size

effects as evident from Eq. 10 and Eq. 11. It can be seen from the potential of mean force (Fig.

2) that e = |Ad − dd | < #/ (similarly, e = |dA − AA | < #/) at low (high) water fraction,

implying that there is effectively no energetic penalty for mixing water and methanol at these

concentrations. Consequently in the case of water + methanol binary, the two-fluid theory (Eq.

10 and 11) suggests that, molecular size is the dominating effect in determining the local

composition (or the preferential solvation of components) of the mixture and hence the NRTL

binary interaction parameters. The results of Chatterjee and Bagchi16 suggest a similar

conclusion that, even in the absence of specific interactions, the difference in molecular sizes can

cause preferential solvation between components.

The P-x-y diagrams and the Gibbs free energy of mixing for water + methanol system

calculated using different binary interaction parameters at 303.15 K are compared in Fig. 3. The

phase equilibrium diagrams are calculated using ASPEN Properties software, version 9.045. The

P-x-y diagram predicted using τ values obtained from MD simulations (both Approach I and

Approach II) shows reasonable agreement with that predicted using the regressed interaction

parameters. However, in this case, predictions using τ obtained from Approach I and the

regressed τ are in closer agreement. Similarly, the Gibbs free energy of mixing calculated using

τ from Approach I and regression are in good agreement (the maximum difference between these

two predictions is 0.29 kJ/mol). As discussed earlier, one of the interaction parameters

calculated using Approach II (Ad ) has a high negative value (indicating attractive interactions

between the components) and consequently, ∆jkl is overpredicted in this case.

18
AIChE Journal
This article is protected by copyright. All rights reserved.
Page 19 of 60 AIChE Journal

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
Methanol (1) + Methyl Acrylate (2) Binary System: Combined size and energetic effects

determine the strength of the interaction parameter

The potential of mean force curves used in the calculation of  for this system are shown

in Fig. 4. Note that for this mixture methanol is designated as component 1 and methyl acrylate

as component 2. The sizes of the neighbor shells for this binary mixture are reported in Table

2.a. The binary interaction parameters and the infinite dilution activity coefficients for methanol

+ methyl acrylate mixture are shown in Table 2.b and Table 2.c, respectively. Values of these

parameters as determined using both the approaches (Approach I and Approach II) show

reasonable agreement with those calculated from regression, within the limits of statistical

uncertainties. From the magnitude of the interaction parameters (determined from both

simulation and regression), it is clear that, methyl acrylate is not preferentially solvated by either

methanol or methyl acrylate (since dA is close to zero). On the other hand, methanol is

preferentially solvated only by other methanol molecules, as can be seen from the magnitude of

Ad .

Unlike the previous case of water + methanol binary, the thermodynamics of mixing in

this case is dominated by both energetic and size effects. At low concentrations of methanol, it

was found that methanol molecules have a higher tendency to form micro-clusters. Fig. 5 shows

the cluster size distribution of methanol at 25% mole fraction along with the corresponding

RDFs. Molecular clusters, in this work were determined using distance criterion. If two

molecules are within the specified cut-off distance of each other, they are considered to belong to

the same cluster. The cut-off distance criterion for determining methanol clusters was taken to

be 0.35 nm based on the first minimum in the oxygen-oxygen RDF. As seen from Fig. 5, though

approximately 32% of the total methanol molecules present in the system are fairly dispersed in

19
AIChE Journal
This article is protected by copyright. All rights reserved.
AIChE Journal Page 20 of 60

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
methyl acrylate, micro-clusters of larger sizes are also observed. Clustering of methanol can also

be confirmed from the strength of the first peak of methanol-methanol RDF (Fig. 5b). This

clustering behavior of methanol is attributed to specific hydrogen bonded interactions between

methanol molecules that are not prominent in methyl acrylate. Consequently, methanol-

methanol interaction is much stronger than methanol-methyl acrylate interaction (as seen in Fig.

4a). This unfavorable interaction between methanol and methyl acrylate leads to a large positive

value of Ad . On the contrary, the binary becomes completely miscible at high methanol

concentrations, as solvation of methyl acrylate by methanol can take place without the breakage

of the hydrogen bonded network. It can be seen from Fig. 4b that there is effectively no

energetic penalty for mixing (since e < #/) in an acrylate centered domain. Hence, according

to two-fluid theory expression, the size effects determine the local composition and thereby the

binary interaction parameter (dA ).

In order to validate the local composition as predicted by the NRTL model, we examine

the short-range structure obtained from MD simulations. The local mole fractions were

calculated from the coordination numbers which in turn were determined by integrating the RDF.

It can be seen from Fig. 6 that in general, NRTL model predicts depletion of methyl acrylate in

the vicinity of the methanol molecules (consequently enhancement of methanol around methanol

as shown in Fig. S1 of Supplementary Material), consistent with MD simulations. However, at

concentrations where there is self-association of methanol (low methanol fraction), NRTL model

over-predicts the local mole fraction of methyl acrylate around methanol. Due to clustering of

methanol molecules (as shown in Fig. 5) at low methanol fractions, there is a depletion of

acrylate molecules to distances that are beyond the second neighbor shell, i.e ∼ 0.98 nm (based

on RDF between methanol oxygen and methyl acrylate CH3, shown in Fig. 5). However, the

20
AIChE Journal
This article is protected by copyright. All rights reserved.
Page 21 of 60 AIChE Journal

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
local fractions predicted by NRTL are close to the bulk concentrations. The NRTL predictions

are improved as the molecules disperse well in the solution at low acrylate concentrations. Since

methanol molecules are capable of forming hydrogen bonded networks, we suggest that the

structural correlation between these molecules extends further beyond the first neighbor shell and

a simple local domain based structural representation can only capture such behavior

qualitatively. Our observation is consistent with that by Phillips and Brennecke17 who concluded

that in mixtures with specific chemical interactions, local composition values predicted by the

NRTL model do not agree with experimental measurements. Despite this observation, it is

remarkable that the NRTL model can predict the phase behavior of such binary mixtures with

good accuracy.

Fig. 7 shows the predicted phase behavior of methanol + methyl acrylate binary at 303.15

K. The P-x-y diagrams and ∆jkl calculated using different binary interaction parameters are in

reasonable agreement with each other. All the interaction parameters predict the presence of an

azeotrope in the range of 0.6-0.66 mole fraction of methanol.

Water (1) + Methyl Acrylate (2) Binary System: Repulsive interactions dominate the

thermodynamics for this partially miscible binary

Unlike the other two binary systems studied here, MD simulations suggest that water and

methyl acrylate are not miscible over most of the concentration range. In the wide range of mole

fractions studied with simulations (from 0.2% to 99.8% water mole fraction), solubility of

methyl acrylate in water (or vice versa) was observed only in the limit of infinite dilution. Water

and methyl acrylate were observed to phase separate at all other concentrations. The PMF

curves at infinite dilution of components were calculated (see Fig. 8) in order to quantify the

strength of various interactions in the binary mixture. As expected, interaction between

21
AIChE Journal
This article is protected by copyright. All rights reserved.
AIChE Journal Page 22 of 60

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
molecules of the same species is much stronger than the interaction between water and methyl

acrylate in both the cases (i.e. water centered domain and MA centered domain), explaining why

the binary is thermodynamically incompatible. It must also be noted that water-water interaction

has a deeper potential minimum than water-MA and MA-MA interactions due to its ability to

form hydrogen bonds.

The interaction strength values ( and  ) calculated from the PMF shown in Fig. 8 are

used to determine . Since water and acrylate phase separate at most of the concentrations

studied in the simulation, the radius of the solvation shell (# () ) was obtained from the RDF only

at concentrations where the mixture is homogeneous (i.e. at infinite dilution). The radii of the

solvation shell thus obtained are shown in Table 3.a. The values of the binary interaction

parameters obtained from these data are listed in Table 3.b. It can be seen from Fig. 8b that at

higher water concentration, water-methyl acrylate interaction does not have a potential

minimum, which is indicative of unfavorable interaction between the components. As a

consequence, the magnitude of the binary interaction parameter (dA ) calculated from molecular

simulation (using both Approach I and Approach II) is large and the value is positive, consistent

with that estimated by regressing the experimental data.

On the other hand, the binary parameter Ad calculated from simulation using Approach I

shows significant deviation from that obtained by regressing the experimental data. However,

Ad calculated from simulation using Approach II is in line with that calculated by Zuo et al.,20

but differs from the values obtained by regressing experimental data in this work. The infinite

dilution activity coefficients, P-x-y diagrams, and the Gibbs free energy of mixing of water +

methyl acrylate systems are shown in Table 3.c, Fig. 9a, and Fig. 9b respectively. The values of

ln (b c ) for methyl acrylate are reasonably consistent between the different methods used (MD

22
AIChE Journal
This article is protected by copyright. All rights reserved.
Page 23 of 60 AIChE Journal

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
simulation and data regression) while the infinite dilution activity coefficient of water is

significantly different. This is due to the inconsistency in the predicted value of Ad for this

mixture. Similar discrepancy is also reflected in the calculated Gibbs free energy of mixing.

The free energy of mixing predicted using MD Approach I does not exhibit two-liquid region.

On the contrary,  from MD Approach II and the  determined by regressing experimental data

(both this work and the work of Zuo et al.20) predict phase separation of the mixture. Similar

conclusion about the phase behavior can also be made from the P-x-y diagram shown in Fig. 9a.

Conclusions
An approach that combines two-fluid theory and molecular simulations was formulated

to predict the binary interaction parameters of the NRTL model. The approach was validated

using three binary systems: water + methanol, methanol + methyl acrylate, and water + methyl

acrylate. The binary interaction parameters were expressed as a function of molecular diameters,

radius of the neighbor-shell, and the effective interaction strength between the molecular species.

Molecular diameters and the neighbor shell size were determined from the RDFs obtained from

MD simulation of pure components and binary mixtures, respectively. Interaction strength

between the molecules was obtained from the potential of mean force which was calculated from

umbrella sampling simulations. The concept of local composition assumes that all of the

molecules in the neighbor shell of a center molecule experience the same potential, irrespective

of their position in the cage. This idea is incorporated into the two-fluid theory by considering

an interaction potential of the square well form. However, to map the net potential experienced

by a molecule (the barrier height in the PMF) onto a square well type potential, different

23
AIChE Journal
This article is protected by copyright. All rights reserved.
AIChE Journal Page 24 of 60

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
methodologies have been suggested in the literature. We have used the mapping approaches

similar to that suggested by Brandani and Prausnitz27 (Approach I), and that by Hirschfelder et

al.,31 (Approach II) to calculate the binary interaction parameters. Both of the approaches

resulted in the binary interaction parameters of similar magnitude for majority of the systems

except one, indicating the robustness of our method. The technique proposed in this work is

purely predictive and requires no adjustable parameters. Such a technique, apart from providing

molecular insight into the thermodynamics of the mixtures, can also be used as a reliable method

to determine the binary interaction parameters for mixtures where experimental data are scarce

or are not available.

Overall, the binary interaction parameters predicted using molecular simulations showed

good agreement with those calculated using experimental data. Among the systems studied, the

water-methyl acrylate pair exhibits the largest disparity in terms of size and interactions and the

predicted binary interaction parameters exhibit stronger sensitivities to the approaches in

mapping molecular simulation results to two-fluid theory. The observation suggests that our

technique might need to be refined for such systems that exhibit large differences in molecular

characteristics of the two components.

The other important aspect to consider is the temperature dependence of the binary

interaction parameters which is not investigated in this work. Based on the framework proposed

here, it is rather straight forward to extend the methodology to high temperatures, given the

availability of a reliable force field over the temperature range of interest. We defer these

investigations to future work. The approach can also be extended to other complex mixtures

such as electrolyte systems.

24
AIChE Journal
This article is protected by copyright. All rights reserved.
Page 25 of 60 AIChE Journal

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
Acknowledgments
This work is supported by J.F Maddox Foundation and also by National Science Foundation

under the grant NSF CMMI-1335082. Computational resources provided by Chemistry

Computational Cluster (CCC) at Texas Tech University are acknowledged.

Literature Cited

1. Chen C-C, Song Y. Solubility modeling with a nonrandom two-liquid segment activity

coefficient model. Industrial & Engineering Chemistry Research. 2004;43:8354-8362.

2. Kato R, Krummen M, Gmehling J. Measurement and correlation of vapor--liquid

equilibria and excess enthalpies of binary systems containing ionic liquids and

hydrocarbons. Fluid Phase Equilibria. 2004;224:47-54.

3. Döker M, Gmehling J. Measurement and prediction of vapor--liquid equilibria of ternary

systems containing ionic liquids. Fluid Phase Equilibria. 2005;227:255-266.

4. Song Y, Chen C-C. Symmetric electrolyte nonrandom two-liquid activity coefficient

model. Industrial & Engineering Chemistry Research. 2009;48:7788-7797.

5. Zhang Y, Chen C-C. Thermodynamic modeling for CO2 absorption in aqueous MDEA

solution with electrolyte NRTL model. Industrial & Engineering Chemistry Research.

2011;50:163-175.

6. Chen C-C. A segment-based local composition model for the Gibbs energy of polymer

solutions. Fluid Phase Equilibria. 1993;83:301-312.

25
AIChE Journal
This article is protected by copyright. All rights reserved.
AIChE Journal Page 26 of 60

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
7. Wu Y-T, Lin D-Q, Zhu Z-Q. Thermodynamics of aqueous two-phase systems - the effect

of polymer molecular weight on liquid-liquid equilibrium phase diagrams by the

modified NRTL model. Fluid Phase Equilibria. 1998;147:25-43.

8. Wilson GM. Vapor-Liquid Equilibrium. XI. A New Expression for the Excess Free

Energy of Mixing. Journal of the American Chemical Society. 1964;86:127-130.

9. Renon H, Prausnitz JM. Local compositions in thermodynamic excess functions for

liquid mixtures. AIChE Journal. 1968;14:135-144.

10. Chen C-C, Song Y. Generalized electrolyte-NRTL model for mixed-solvent electrolyte

systems. AIChE Journal. 2004;50:1928-1941.

11. Kemény S, Rasmussen P. A derivation of local composition expressions from partition

functions. Fluid Phase Equilibria. 1981;7:197-203.

12. Prausnitz JM, Lichtenthaler RN, de Azevedo EG. Molecular thermodynamics of fluid-

phase equilibria (3rd edition). New Jersey: Prentice Hall, 1999.

13. Scott RL. Corresponding States Treatment of Nonelectrolyte Solutions. J Chem Phys.

1956;25:193-205.

14. Brandani V, Prausnitz JM. Two-fluid theory and thermodynamic properties of liquid

mixtures: Application to simple mixtures of nonelectrolytes. Proceedings of the National

Academy of Sciences. 1982;79:5729-5733.

15. Chatterjee P, Bagchi S. Study of preferential solvation in mixed binary solvents by

ultraviolet--visible spectroscopy. Journal of the Chemical Society, Faraday Transactions.

1990;86:1785-1789.

26
AIChE Journal
This article is protected by copyright. All rights reserved.
Page 27 of 60 AIChE Journal

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
6. Chatterjee P, Bagchi S. Preferential solvation of a dipolar solute in mixed binary solvent:

a study of UV-visible spectroscopy. The Journal of Physical Chemistry. 1991;95:3311-

3314.

7. Phillips DJ, Brennecke JF. Spectroscopic Measurement of Local Compositions in Binary

Liquid Solvents and Comparison to the NRTL Equation. Industrial & Engineering

Chemistry Research. 1993;32:943-951.

8. Chen C-C, Mathias PM. Applied thermodynamics for process modeling. AIChE Journal.

2002;48:194-200.

9. Chen C-C. Representation of solid-liquid equilibrium of aqueous electrolyte systems with

the electrolyte NRTL model. Fluid Phase Equilibria. 1986;27:457-474.

0. Zuo C, Li Y, Li C, Cao S, Yao H, Zhang S. Thermodynamics and separation process for

quaternary acrylic systems. AIChE Journal. 2016;62:228-240.

1. Vasquez VR, Whiting WB. Regression of binary interaction parameters for

thermodynamic models using an inside-variance estimation method (IVEM). Fluid Phase

Equilibria. 2000;170:235-253.

2. Jónsd'ottir SÓ, Rasmussen K, Fredenslund A. UNIQUAC parameters determined by

molecular mechanics. Fluid Phase Equilibria. 1994;100:121-138.

3. Sum AK, Sandler SI. Use of ab initio methods to make phase equilibria predictions using

activity coefficient models. Fluid Phase Equilibria. 1999;158:375-380.

4. Neiman M, Cheng H, Parekh V, Peterson B, Klier K. A critical assessment on two

predictive models of binary vapor-liquid equilibrium. Physical Chemistry Chemical

Physics. 2004;006:3474-3483.

27
AIChE Journal
This article is protected by copyright. All rights reserved.
AIChE Journal Page 28 of 60

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
5. Haghtalab A, Seyf JY. A new insight to validation of local composition models in binary

mixtures using molecular dynamic simulation. AIChE Journal. 2016;62:275-286.

6. Seyf JY, Haghtalab A. A junction between molecular dynamics simulation and local

composition models for computation of solid-liquid equilibrium-A pharmaceutical

solubility application. Fluid Phase Equilibria. 2017;437:83-95.

7. Brandani V, Prausnitz JM. Two-fluid theory and thermodynamic properties of liquid

mixtures: General theory. Proceedings of the National Academy of Sciences.

1982;79:4506-4509.

8. Brandani V, Prausnitz JM. Two-fluid theory and thermodynamic properties of liquid

mixtures. Application to hard-sphere mixtures. Proceedings of the National Academy of

Sciences. 1982;79:5103-5106.

9. Perera A, Sokolić F, Zoranić L. Microstructure of neat alcohols. Physical Review E.

2007;75:60502-60502.

0. Allen MP, Tildesley DJ. Computer simulation of liquids. Oxford: Clarendon Press, 1987.

1. Hirschfelder J, Bird RB, Curtiss CF. Molecular theory of gases and liquids. New York:

Wiley, 1954.

2. Maerzke KA, Schultz NE, Ross RB, Siepmann JI. TraPPE-UA force field for acrylates

and Monte Carlo simulations for their mixtures with Alkanes and alcohols. Journal of

Physical Chemistry B. 2009;113:6415-6425.

3. Chen B, Potoff JJ, Siepmann JI. Monte Carlo calculations for alcohols and their mixtures

with alkanes. Transferable potentials for phase equilibria. 5. United-atom description of

primary, secondary, and tertiary alcohols. The Journal of Physical Chemistry B.

2001;105:3093-3104.

28
AIChE Journal
This article is protected by copyright. All rights reserved.
Page 29 of 60 AIChE Journal

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
4. Berendsen HJC, Grigera JR, Straatsma TP. The missing term in effective pair potentials.

Journal of Physical Chemistry. 1987;91:6269-6271.

5. Plimpton S. Fast parallel algorithms for short-range molecular dynamics. Journal of

Computational Physics. 1995;117:1-19.

6. Wang J, Wolf RM, Caldwell JW, Kollman PA, Case DA. Development and testing of a

general amber force field. Journal of Computational Chemistry. 2004;25:1157-1174.

7. Shinoda W, Shiga M, Mikami M. Rapid estimation of elastic constants by molecular

dynamics simulation under constant stress. Physical Review B. 2004;69:134103-134103.

8. Torrie GM, Valleau JP. Nonphysical sampling distributions in Monte Carlo free-energy

estimation: Umbrella sampling. Journal of Computational Physics. 1977;23:187-199.

9. Park S, Schulten K. Calculating potentials of mean force from steered molecular

dynamics simulations. Journal of Chemical Physics. 2004;120:5946-5961.

0. Fiorin G, Klein ML, Hénin J. Using collective variables to drive molecular dynamics

simulations. Molecular Physics. 2013;111:3345-3362.

1. Kumar S, Rosenberg JM, Bouzida D, Swendsen RH, Kollman PA. Multidimensional

free-energy calculations using the weighted histogram analysis method. Journal of

Computational Chemistry. 1995;16:1339-1350.

2. Grossfield A. WHAM: an implementation of the weighted histogram analysis method.

http://membrane.urmc.rochester.edu/content/wham/. Version 2.0.9.

3. Trzesniak D, Kunz A-PE, van Gunsteren WF. A comparison of methods to compute the

potential of mean force. ChemPhysChem. 2007;8:162-169.

4. de Oliveira TE, Netz PA, Mukherji D, Kremer K. Why does high pressure destroy co-

non-solvency of PNIPAm in aqueous methanol? Soft matter. 2015;11:8599-8604.

29
AIChE Journal
This article is protected by copyright. All rights reserved.
AIChE Journal Page 30 of 60

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
5. Aspen Properties Burlington, MA: Aspen Technology.

6. Bevington PR, Robinson DK, Blair JM, Mallinckrodt AJ, McKay S. Data reduction and

error analysis for the physical sciences. Computers in Physics. 1993;7:415-416.

7. Israelachvili JN, Pashley RM. Molecular layering of water at surfaces and origin of

repulsive hydration forces. Nature. 1983;306:249-250.

8. Ahmad SA, Tanwar RS, Gupta RK, Khanna A. Interaction parameters for multi-

component aromatic extraction with sulfolane. Fluid Phase Equilibria. 2004;220:189-198.

30
AIChE Journal
This article is protected by copyright. All rights reserved.
15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

This article is protected by copyright. All rights reserved.


AIChE Journal

AIChE Journal
31
Author Manuscript
Page 31 of 60
Page 32 of 60

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

This article is protected by copyright. All rights reserved.


AIChE Journal

AIChE Journal
1
Author Manuscript
Page 33 of 60 AIChE Journal

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
List of Figure Captions

Figure 1. Pure component RDFs for water, methanol, and methyl acrylate showing effective

molecular sizes.

Figure 2. Potential of mean force of water + methanol system at two different concentrations

and at 303.15 K. (a) PMF around a center water molecule that is used in the calculation of τ ,

and (b) PMF around a center methanol molecule that is used in the calculation of τ . The

symbols denote the calculated PMF values and the lines represent the PMF smoothed with cubic

splines. The depths of the potential wells are marked in their corresponding colors.

Figure 3. Comparison of water + methanol binary phase behavior at 303.15 K using NRTL

equations with  obtained from data regression (performed in this work) and molecular

simulations. (a) P-x-y diagram, and (b) Gibbs free energy of mixing.

Figure 4. Potential of mean force of methanol + methyl acrylate system at two different

concentrations and at 303.15 K. (a) PMF around a center methanol molecule that is used in the

calculation of τ , and (b) PMF around a center methyl acrylate molecule that is used in the

calculation of τ . The symbols denote the calculated PMF values and the lines represent the

PMF smoothed with cubic splines. The depths of the potential wells are marked in their

corresponding colors.

Figure 5. (a) Cluster size distribution of methanol at 0.25 mole fraction of methanol, and (b)

RDF of different components at 0.25 mole fraction of methanol in methanol + methyl acrylate

binary.

2
AIChE Journal
This article is protected by copyright. All rights reserved.
AIChE Journal Page 34 of 60

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
Figure 6. Local mole fraction of methyl acrylate around methanol at different concentrations of

methanol. The horizontal lines are NRTL predictions using regressed parameters20.

Figure 7. Comparison of methanol + methyl acrylate binary phase behavior at 303.15 K using

NRTL equations with  obtained from data regression (performed in this work) and molecular

simulations. (a) P-x-y diagram, and (b) Gibbs free energy of mixing.

Figure 8. Potential of mean force of water + methyl acrylate system at two different

concentrations and at 303.15 K. (a) PMF around a center water molecule that is used in the

calculation of τ , and (b) PMF around a center methyl acrylate molecule that is used in the

calculation of τ . The symbols denote the calculated PMF values and the lines represent the

PMF smoothed with cubic splines. The depths of the potential wells are marked in their

corresponding colors.

Figure 9. Comparison of water + methyl acrylate binary phase behavior at 303.15 K using

NRTL equations with  obtained from data regression (performed in this work) and molecular

simulations. (a) P-x-y diagram, and (b) Gibbs free energy of mixing.

3
AIChE Journal
This article is protected by copyright. All rights reserved.
15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

This article is protected by copyright. All rights reserved.


AIChE Journal

AIChE Journal
4
Author Manuscript

Figure 1
Page 35 of 60
Page 36 of 60

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

This article is protected by copyright. All rights reserved.


AIChE Journal

AIChE Journal
(a)

5
Author Manuscript
15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

This article is protected by copyright. All rights reserved.


AIChE Journal

AIChE Journal
(b)

6
Author Manuscript

Figure 2.
Page 37 of 60
Page 38 of 60

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

This article is protected by copyright. All rights reserved.


AIChE Journal

AIChE Journal
(a)

7
Author Manuscript
15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

This article is protected by copyright. All rights reserved.


AIChE Journal

AIChE Journal
(b)

8
Author Manuscript

Figure 3.
Page 39 of 60
Page 40 of 60

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

This article is protected by copyright. All rights reserved.


AIChE Journal

AIChE Journal
(a)

9
Author Manuscript
15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

This article is protected by copyright. All rights reserved.


AIChE Journal

AIChE Journal
(b)

10
Author Manuscript

Figure 4.
Page 41 of 60
Page 42 of 60

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

This article is protected by copyright. All rights reserved.


AIChE Journal

AIChE Journal
(a)

11
Author Manuscript
15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

This article is protected by copyright. All rights reserved.


AIChE Journal

AIChE Journal
(b)

12
Author Manuscript

Figure 5.
Page 43 of 60
Page 44 of 60

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

This article is protected by copyright. All rights reserved.


AIChE Journal

AIChE Journal
13
Author Manuscript

Figure 6.
15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

This article is protected by copyright. All rights reserved.


AIChE Journal

AIChE Journal
(a)

14
Author Manuscript
Page 45 of 60
Page 46 of 60

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

This article is protected by copyright. All rights reserved.


AIChE Journal

AIChE Journal
(b)

15
Author Manuscript

Figure 7.
15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

This article is protected by copyright. All rights reserved.


AIChE Journal

AIChE Journal
(a)

16
Author Manuscript
Page 47 of 60
Page 48 of 60

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

This article is protected by copyright. All rights reserved.


AIChE Journal

AIChE Journal
(b)

17
Author Manuscript

Figure 8.
15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

This article is protected by copyright. All rights reserved.


AIChE Journal

AIChE Journal
(a)

18
Author Manuscript
Page 49 of 60
Page 50 of 60

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

This article is protected by copyright. All rights reserved.


AIChE Journal

AIChE Journal
(b)

19
Author Manuscript

Figure 9.
15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
()
(nm)

0.534

0.597

This article is protected by copyright. All rights reserved.


Table 1a. Neighbor Shell Size for Water + Methanol Binary Mixture

( )
(nm)

0.534

0.534
AIChE Journal

AIChE Journal
20
()
(nm)

0.377

0.597
Center molecule

Methanol (2)
Water (1)
Author Manuscript
Tables
Page 51 of 60
AIChE Journal Page 52 of 60

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
Table 1b. Comparison of Binary Interaction Parameters for Water + Methanol System obtained

from Simulations with those Obtained from Experimental Data at 303.15 K: Non-randomness

Factor  was set to 0.3 in the Calculations

Center  Simulation  Simulation  Regression  Regression20

molecule
(Approach I) (Approach II) (this work)

Water (1) -0.395±0.187 ( ) -1.287±0.369 ( ) -0.358±0.233 ( ) 0.105 ( )

Methanol (2) 0.446±0.233 ( ) 0.512±0.417 ( ) 0.976±0.368 ( ) 0.441 ( )

21
AIChE Journal
This article is protected by copyright. All rights reserved.
Page 53 of 60 AIChE Journal

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
Table 1c. Infinite Dilution Activity Coefficient of Water + Methanol System obtained using

NRTL Equations with τ Calculated from Simulations and from Experimental Data at 303.15 K

Center ln (  ) Simulation ln (  ) ln (  ) ln (  )

molecule Regression Regression20


(Approach I) Simulation

(this work)
(Approach II)

Water (1) -0.005 ± 0.257 -0.848 ± 0.477 0.37 ± 0.303 0.491

Methanol (2) 0.001 ± 0.331 -1.381 ± 0.86 0.577 ± 0.467 0.543

22
AIChE Journal
This article is protected by copyright. All rights reserved.
Page 54 of 60

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
()
(nm)

0.855

0.856
Table 2a. Neighbor Shell Size for Methanol + Methyl Acrylate Binary Mixture

This article is protected by copyright. All rights reserved.


( )
(nm)

0.855

0.855
AIChE Journal

AIChE Journal
23
()
(nm)

0.611

0.856
Methyl acrylate (2)
Center molecule

Methanol (1)
Author Manuscript
Page 55 of 60 AIChE Journal

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
Table 2b. Comparison of Binary Interaction Parameters for Methanol + Methyl Acrylate System

obtained from Simulations with those obtained from Experimental Data at 303.15 K: Non-

randomness Factor  was Set to 0.3

Center  Simulation  Simulation  Regression  Regression20

molecule
(Approach I) (Approach II) (this work)

Methanol (1) 1.229±0.114 ( ) 1.547±0.289 ( ) 0.838±1.141 ( ) 0.801 ( )

Methyl 0.392±0.175 ( ) 0.298±0.365 ( ) 0.456±1.257 ( ) 0.592 ( )

acrylate (2)

24
AIChE Journal
This article is protected by copyright. All rights reserved.
AIChE Journal Page 56 of 60

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
Table 2c. Infinite Dilution Activity Coefficient of Methanol + Methyl Acrylate System obtained

using NRTL Equations with τ Calculated from Simulations and from Experimental Data at

303.15 K

Center ln (  ) Simulation ln (  ) ln (  ) ln (  )

molecule Regression Regression20


(Approach I) Simulation

(this work)
(Approach II)

Methanol (1) 1.578 ± 0.178 1.82 ± 0.419 1.236 ± 1.482 1.297

Methyl 1.242 ± 0.182 1.271 ± 0.378 1.108 ± 1.422 1.222

acrylate (2)

25
AIChE Journal
This article is protected by copyright. All rights reserved.
15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
()
(nm)

0.660

0.795

This article is protected by copyright. All rights reserved.


Table 3a. Neighbor Shell Size for Water + Methyl Acrylate Binary Mixture

( )
(nm)

0.660

0.660
AIChE Journal

AIChE Journal
26
()
(nm)

0.525

0.795
Methyl acrylate (2)
Center molecule

Water (1)
Author Manuscript
Page 57 of 60
AIChE Journal Page 58 of 60

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
Table 3b. Comparison of Binary Interaction Parameters for Water + Methyl Acrylate System

obtained from Simulations with those obtained from Experimental Data at 303.15 K: Non-

randomness Factor  was Set to 0.3

Center  Simulation  Simulation  Regression  Regression20

molecule
(Approach I) (Approach II) (this work)

Water (1) 0.311±0.147 ( ) 3.826±0.306 ( ) 1.084±0.014 ( ) 2.294 ( )

Methyl 2.078±0.189 ( ) 2.741±0.279 ( ) 2.687±0.056 ( ) 2.410 ( )

acrylate (2)

27
AIChE Journal
This article is protected by copyright. All rights reserved.
Page 59 of 60 AIChE Journal

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Author Manuscript
Table 3c. Infinite Dilution Activity Coefficient of Water + Methyl Acrylate System obtained

using NRTL Equations with τ Calculated from Simulations and from Experimental Data at

303.15 K

Center ln (  ) Simulation ln (  ) ln (  ) ln (  )

molecule Regression Regression20


(Approach I) Simulation

(this work)
(Approach II)

Water (1) 1.425 ± 0.152 5.03 ± 0.307 2.284 ± 0.015 3.464

Methyl 2.361 ± 0.225 3.955 ± 0.279 3.47 ± 0.056 3.563

acrylate (2)

28
AIChE Journal
This article is protected by copyright. All rights reserved.
Page 60 of 60

15475905, 2018, 7, Downloaded from https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.16117 by Libya Hinari NPL, Wiley Online Library on [25/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

This article is protected by copyright. All rights reserved.


AIChE Journal

AIChE Journal
29
Author Manuscript

You might also like