You are on page 1of 9

3572

Ind. Eng. Chem. Res. 1999, 38, 3572-3580

Estimation of Diffusion Coefficients in Acetone-Cellulose Acetate Solutions


G. D. Verros and N. A. Malamataris*
Department of Mechanical and Industrial Engineering, University of Thessaly, 38334 Volos, Greece

The mutual and the self-diffusion coefficients of the acetone-cellulose acetate system are determined by using established gravimetric measurements of the acetone evaporation rate. The process is studied as a one-dimensional numerical experiment utilizing the Galerkin finite element method. The numerical technique provides simultaneous solution of the model equations and yields by comparison with gravimetric data the diffusion coefficients of acetone in cellulose acetate for a wide range of temperatures and compositions. The estimated diffusion coefficients based on free volume theory are in satisfactory agreement with the available experimental data. It is believed that this method can be applied to other systems of interest.
Introduction Asymmetric cellulose acetate membranes are widely used in a number of industrial processes such as separations, solution concentration, water desalination, waste purification, and so forth. These membranes are manufactured by two major processes:1 the dry cast process and the wet cast phase-inversion process. In the dry cast process a solution of CA-solvent-nonsolvent is allowed to solidify by solvent evaporation. In the wet cast phase-inversion process the casting solution is partially concentrated by solvent evaporation and then is solidified by immersion in a low-temperature nonsolvent bath. Several recipes and modifications based on practical experience have appeared in the literature. Most recipes utilize acetone as a solvent. The industrial importance of acetone evaporation as a major step in the dry cast and as a precursor step in the wet cast phase-inversion process has led to extensive experimental2-6 and modeling7-12 studies. TantenkinErsolmaz and Shojaie et al. performed the most complete experimental and modeling studies to date. In particular, Tantenkin-Ersolmaz4 studied the effects of the solvent evaporation process conditions on the final membrane morphology for the binary system CAacetone. She utilized infrared thermographic imaging techniques to measure the surface temperature of the casting solution coupled with gravimetric measurements of the acetone evaporation rate. She was the first to address the cooling effect of the evaporation process. Shojaie et al. extended her experimental methods to the ternary system CA-acetone-water5,10 and developed rigorous models for the evaporation step.11,12 More specifically, the process was modeled as a coupled heatand mass-transfer problem with a moving boundary. The Landau transformation13 was used to immobilize the moving boundary. In both experimental and modeling studies it is evident that the solvent evaporation process is controlled by the diffusion of the solvent in the polymer solution. Therefore, the magnitudes of the mutual and self-diffusion coefficients of acetone are of fundamental importance for cellulose acetate membrane
* Corresponding author. Current address: Verias 96, GR57008 Ionia, Greece. E-mail: nikolaos@eng.auth.gr. Telephone: +3031781463. Fax: +3031782670.

formation. Similar conclusions have been made in the modeling of the polymeric coating drying process14-16 that exhibits analogous coupled heat- and mass-transfer phenomena. Despite its industrial importance, little is known about the CA-acetone diffusion coefficient. Traditional techniques to measure diffusion coefficients include sorption and desorption techniques, radiotracer methods, chromatography, and nuclear magnetic resonance (NMR) experiments, as reviewed by Crank and Park17 and Tyrrell and Harris.18 Park19 utilized a radiotracer technique and measured the self-diffusion coefficient of acetone at cellulose acetate at 30 C for three volume fractions of acetone between 0.15 and 0.25. Anderson and Ullman7 measured the acetone self-diffusion coefficient in cellulose acetate at concentrations ranging between 0 and 40 wt % polymer at 23 C utilizing the NMR pulsed field gradient technique. Reuvers and Smolders20 measured the mutual diffusion coefficient from sedimentation experiments at 25 C for acetone concentrations between 88 and 95 vol %. So far in the modeling studies, this limited experimental data was fitted to appropriate correlations describing the dependence of diffusion coefficients on concentration only, although the nonisothermal nature of the process indicates that the temperature dependence of the diffusion coefficients must also be taken into account. The existing experimental techniques had been developed before the successful implementation of powerful numerical tools such as the finite element method in the study of transport processes. The wide use of computer-aided analysis21 enables simpler methods for the calculation of mutual and self-diffusion coefficients by utilizing gravimetric measurements of the solvent evaporation rate, which is the subject of this work. The origin of this idea has been formulated by Ataka and Sasaki3 who propose that these measurements can be used for the estimation of diffusion coefficients. A similar method in the case of drying of coatings was recently developed by Price et al.22 for the estimation of the diffusion coefficients by measuring the drying rate. In the present work, the solvent evaporation process of the CA-acetone system is modeled as a nonlinear, coupled heat- and mass-transfer problem with a moving boundary. The model equations are simultaneously

10.1021/ie990141d CCC: $18.00 1999 American Chemical Society Published on Web 08/14/1999

Ind. Eng. Chem. Res., Vol. 38, No. 9, 1999 3573

solved by the Galerkin finite element method. The diffusion coefficient is calculated from the equation of Vrentas and Duda,23,24 which is based on free volume theory and takes into account both the temperature and concentration dependence of the coefficient. Gravimetric measurements of the acetone evaporation rate are compared with model predictions in order to estimate the parameters of the Vrentas and Duda equation. The model is validated against extensive temperature measurements at the surface of the polymer solution. A parametric analysis is carried out to estimate the main sources of error in the model parameters and determine the limiting step of the process. In the following sections, the governing equations are developed along with the appropriate boundary conditions, the model parameters are presented, and the finite element formulation is given. Finally, results are discussed and conclusions are drawn. Model Equations The solvent evaporation process is illustrated in Figure 1. The system consists of a liquid layer resting upon an impermeable solid support exposed to a gas phase of temperature T0. The liquid layer is the CAacetone solution and has an initial thickness L0.The support is a flat, horizontal glass plate with constant thickness Lsup, allowing heat exchange with the polymer solution. Prior to time zero, the polymer solution is assumed to have a constant initial acetone concentration everywhere in its mass, and the whole system has the same initial temperature T0. At time t ) 0, a liquidgas interface is suddenly created by exposing the polymer solution to the gas phase. The acetone begins to evaporate, resulting in a downward motion of the liquid-gas interface. As the acetone at the surface evaporates, the gas-liquid interface is cooled. From an engineering point of view this is a coupled heat- and mass-transfer process with a moving boundary. The dimensionless governing equations, which define the evaporation model are

Figure 1. Schematic representation of the solvent evaporation process.

specific heat capacity, and the thermal conductivity of the polymer solution, respectively. Cp0 and F0 are scaling factors having units of specific heat capacity and density, respectively. The subscript sup denotes properties and variables of the support. Initial and Boundary Conditions for the Diffusion Equation:

u1 ) u10

)0

(4)

C0 u1/ ) C5(xs - x); C5 ) -L0kGP/(D0F) 1 u1/ ) 0 )0

) s (5) (6)

Equation 5 is a mass balance at the moving interface, and eq 6 specifies zero mass flux at the glass plate. xs and x are the acetone mole fraction at the liquid layergas-phase interface and far away from the interface, respectively. P is the total pressure, F denotes the 1 pure acetone density, and kG is the mass-transfer coefficient in the gas phase. Initial and Boundary Conditions for the Energy Equations:

u1 u1 ) C0 ; C0 ) D/D0 C1

) sup ) 1

)0

(7)

0<<s

(1)

C : C1 ) FCp/F0Cp0; C2 ) k/D0F0Cp0 ) 2 0 < < s (2)

C2 / ) C6( - 1) + C7(1 - s4) C8(xs - x); C6 ) L0h/D0F0Cp0; C7 ) L0 T03/D0F0Cp0; C8 ) (H)kGPL0/(T0D0F0Cp0) ) s (8) C4 sup/ ) C2 / sup/ ) 0 )0 (9) (10)

C3

sup sup ) C4 ; C3 ) FsupCpsup/F0Cp0; C4 ) ksup/F0Cp0D0 -Lsup/L0 < < 0 (3)

) -Lsup/L0

Equations 1-3 represent the conservation of mass and energy in the polymer solution and the support, respectively assuming one-dimensional, Fickian diffusion. u1 is the acetone volume fraction; ) z/L0 is the dimensionless space coordinate; ) D0t/L02 is the dimensionless time; ) T/T0 is the dimensionless temperature; s ) L(t)/L0 is the dimensionless position of the moving boundary. T represents the temperature, t denotes time, and D0 is a scaling factor having the units of diffusion coefficient. D is the mutual diffusion coefficient. An equation of state for D is presented and discussed in the next section. F, Cp, and k represent the density, the

Equation 8 is an energy balance at the moving interface, equating the heat conduction to the polymer solution with free convection heat transfer, radiant heat transfer from the ambient air, and latent heat loss due to acetone evaporation. Equation 9 implies continuity of temperature and heat flux at the glass plate-polymer solution interface, while eq 10 implies perfect insulation of the glass support lower surface. s denotes the dimensionless temperature of the liquid-gas interface, h is the heat transfer coefficient, is the emissivity of the polymer solution, denotes the Stefan-Boltzmann constant, and H is the acetone latent heat of vaporization.

3574 Ind. Eng. Chem. Res., Vol. 38, No. 9, 1999


Table 1. Thermophysical Properties of Liquid Acetone, Cellulose Acetate, and Glass Support25-27 property value units kg/m3 J/(kg K) W/(m K) J/kg kg/m3 J/(kg K) W/(m K) kg/m3 J/(kg K) W/(m K) Figure 2. Comparison between predicted and experimental data for polymer solution density (temperature 298 K). A. Liquid Acetone density 791 (20 C) 812 (0 C) specific heat capacity 2156 (20 C) 2102 (0 C) thermal conductivity 0.16 (20 C) 0.165 (0 C) latent heat of vaporization 552 103 B. Cellulose Acetate density 1300 specific heat capacity 1464 thermal conductivity 0.251 density specific heat capacity thermal conductivity C. Glass Support 2500 750 0.79

Finally, the instantaneous dimensionless solution thickness s is obtained from the conservation of the polymer mass:

chemical potential of the solvent, respectively. The above relation can be expressed in terms of the acetone volume fraction:

(1 - u1)d ) 0

(1 - u10) d 0

D) (11)

1/RT u1 u1 1 21(F - F) + F 1 2 2 D*F1

)( )

(15)

Equation 11 is the definition of the instantaneous position of the moving boundary and is solved simultaneously with the governing eqs 1-3 along with the boundary conditions in eqs 4-10, as explained in detail in the section Finite Element Formulation. Model Parameters Thermophysical Properties of the Polymer Solution and the Glass Substrate. The density, specific heat capacity, and the thermal conductivity of the polymer solution can be calculated by a simple addition rule, assuming ideal solution:

The derivative of the volume fraction u1 with respect to the weight fraction 1 can be directly obtained from eq 13. The chemical potential of the acetone is related to its volume fraction by the Flory-Huggins theory:30

1/RT ) ln u1 + (1 - u1)(1 - v1/v2 + (1 - u1) - u1(1 - u1) d/du1) (16)


where vi is the molar volume of component i and is the polymer-solvent interaction parameter, which for the CA-acetone system was measured by Altena31at 25 C and corrected for polymer solution temperature as

P ) P11 + P2(1 - 1)

(12)

) (0.645 - 011u1)298/T

(17)

where P is the property of the solution, 1 is the weight fraction of acetone, and P1 and P2 denote the corresponding property of acetone and CA, respectively. The weight fraction 1 is related to the volume fraction u1 as follows:

1 ) u1/[(F/F)(I - u1) + u1] 2 1

From the above expressions the derivative of the acetone chemical potential with respect to the volume fraction is directly calculated. According to the free volume diffusion model developed by Vrentas and Duda,23,24 the solvent self-diffusion coefficient in a polymer solution D* is given as follows:
D* ) D0* exp exp

(13)

where P and P represent the densities of pure ac1 2 etone and CA, respectively. The thermophysical properties of pure liquid acetone, CA, and the glass support are given in Table 1. The thermophysical properties of acetone were corrected for temperature changes by interpolation between their known values at 20 and 0 C. To validate eq 12, a comparison is made in Figure 2 between predicted values for polymer solution density assuming ideal solution and experimental data.28 Diffusion Coefficient of Acetone in CA. The mutual diffusion coefficient D is related to the selfdiffusion coefficient D* and the thermodynamic properties by the following expression:29

((

(-E) RT

-(1V1* + (1 - 1)V2*) K11 K12 1 (K21 - Tg1 + T) + (1 - 1) (K22 - Tg2 + T)

( )

(18)

D)

D*F1 D1 RT F1

( )

(14)

where F1 and 1 are the mass concentration and the

Here, subscripts 1 and 2 refer to acetone and CA, respectively. D0* is a pre-exponential factor, E is the critical energy which a molecule must possess to overcome the attractive forces holding it to its neighbors, and is an overlap factor which is introduced because the same free volume is available to more than one molecule. Vi* is the specific hole-free volume of the i component required for a diffusion jump, and represents the ratio of the critical molar volume of the jumping unit of the solvent to that of the polymer. K11 and K21 are free volume parameters for the solvent while K12 and K22 are those for the polymer. Tgi is the glass transition temperature of the i component.

Ind. Eng. Chem. Res., Vol. 38, No. 9, 1999 3575

The parameters K11/, K21-Tg1, D/ and E can be 0 obtained by fitting viscosity-temperature data of the pure solvent and for acetone are equal32 to 1.86 10-6 m3/(kg K), -53.33 K, 3.6 10-8 m2/s, and 0 J/mol, respectively. The two critical volumes V/ and V/ can be 1 2 estimated using group contribution methods.32 By utilizing these methods, the acetone critical volume (V/) 1 was found to be equal to 4.43 10-4 m3/kg. The free volume parameters K12/ and K22 - Tg2 are obtained by fitting viscosity-temperature data of the pure polymer. Values of these parameters have been reported by Zielinski and Duda33 for a large number of polymers. Generally, they range for solid polymers from 2 10-7 to 8 10-7 m3/(kg K) and from -80 to -400 K, respectively. Since the values of these parameters for the cellulose acetate are not available, we assume the values 5 10-7 m3/(kg K) and -240 K, respectively. This leaves only one parameter (V/) to be determined 2 by fitting experimental data for the cellulose acetateacetone system to eq 18. Heat- and Mass-Transfer Coefficients. The heat transfer coefficient h is calculated from the following empirical correlation for heat transfer under free convention conditions to a cooled, horizontal square plate, facing upward:26

be written in terms of its activity on the polymer solution side a as

xs ) RPsat/P 1

(21)

where Psat is the pure acetone vapor pressure calcu1 lated by an Antoines equation.26 The acetone activity a is related to its chemical potential in the solution by the following equation:

a ) e1/RT

(22)

where the chemical potential of the solvent is calculated by Flory-Huggins theory. Finally, the emissivity of the CA-acetone solution was equal to 0.8 according to the thermographic measurements of Greenberg et al.6 Finite Element Formulation The computational domain is shown in Figure 3 at two different time steps. It is discretized in 15 finite elements. The unknown volume fraction u and the temperature are expanded in terms of quadratic Galerkin basis functions i as

hw ) 0.54(Gr Pr)0.25 Kf

(19)

u)

uii; ) ii i)1 i)1

(23)

The mass-transfer coefficient kG is calculated from the above correlation by invoking the heat- and masstransfer analogy and using a correction term for high fluxes, obtained from film theory:34

kG ) 0.54(Gr Sc)0.25

( )

1 - x c Df MA(1 - xs) ln w P(xs - x) 1 - xs

( )

(20)

where w is the width of the plate, c is the overall molar density, and MA denotes the molecular weight of acetone. Kf and Df represent the thermal conductivity of the gas phase and the binary diffusion coefficient of acetone vapor in air, respectively. The superscript f represent properties evaluated at the mean gas-phase temperature Tf ) (Ts + T)/2 and the mean acetone vapor mole fraction xf ) (xs + x)/2, with x ) 0 and T ) T0. This means that the properties of the gas phase far away from the polymer solution interface are identical to the properties of air. Gr, Pr, and Sc represent the Grashof, Prandtl, and Scmidt numbers of the gas phase, respectively. These numbers have their standard definitions. To evaluate them, the thermal conductivity and the specific heat capacity of the gas phase are calculated from pure substance data and the weight fraction of acetone vapor, utilizing a simple addition rule (eq 12). Data for the thermophysical properties of the acetone vapor and the air are available in standard references.25-27 The gasphase viscosity is evaluated from pure substance viscosity data using the kinetic theory of gases.35 The binary diffusion coefficient of acetone in air26 is corrected to the mean gas-phase temperature Tf utilizing the correlation of Fuller et al.36 Since ideal gas behavior and equilibrium are assumed, the acetone mole fraction at the interface xs, can

In the nodes of the support, the energy equation is evaluated, the only unknown is the temperature, and this part of the domain is fixed. In the polymer solution both the energy and mass equations are evaluated, each node has the temperature and the volume fraction as unknowns, and this part of the domain deforms according to the velocity of the moving boundary. Finally, the boundary position s is added as an unknown at the last node of the computational domain. The governing equations weighted integrally with the basis functions resulted in the following mass Ri , M energy Ri , and kinematic RN residuals: E K

Ri ) M Ri ) E Ri ) E
0 -L

0s

u1 u1 i C0 d

)]

(24) (25)

0s[(C1 - (C2 ))]i d sup sup i C4 d

sup/L0

C3

)]

(26) (27)

RN ) K

0s(1 - u1) d - 01(1 - u10) d

To account for the moving boundary, the total time derivatives for the volume fraction and the temperature were calculated, introducing convective terms in the governing equations:37

du1/d ) u1/ + (d/d)(u1/) d/d ) / + (d/d)(d/d)

(28) (29)

By substituting the above equations into the weighted

3576 Ind. Eng. Chem. Res., Vol. 38, No. 9, 1999

Figure 3. Schematic representation of the computational domain deformation.

residuals in eqs 24 and 25, we obtain the following expressions:

Figure 4. Comparison of model predictions with experimental data for three differential thicknesses (u10 ) 0.87). Table 2. Model Parameters quantity initial temp, T0 length of the glass support, Lsup width of the glass support, w CA molecular weight value 294 1.2 10-3 510-2 40 000 units K m m g/mol

Ri ) M Ri E )

du1 d u1 u1 i C0 d (30) d d

)]

0 [
s

d d i C C1 C d d d 1 2

)]

(31)

To decrease the order of differentiation and project the Neuman (natural) boundary conditions of the problem, integration by parts (divergence theorem in one dimension) is applied and the weighted residuals become

Ri ) M

0s

du1 i d u1 i i u1 + C d + d d 0 )s u1 i (32) C0 )0

)]

Ri ) E

0s C1 d i - d C1 i + (C2 ) d d
C2 i
0 -L

| |

)s

details at two different time steps, to show how the domain deforms with the moving boundary. A detailed presentation of the finite element technique that enables the simultaneous solution of the primary unknowns of the problem (acetone volume fraction and temperature) with the moving boundary can be found elsewhere.38,39 Any additional mesh refinement, time step decrease, or introduction of higher order polynomial basis functions (Hermite or Lagrangian cubic) has an improvement of less than 10-6 in the accuracy of the solution. Results and Discussion To estimate the diffusion coefficients, gravimetric data of the solvent evaporation rate from Tantenkin-Ersolmaz4 were compared with model predictions utilizing nonlinear regression analysis. The objective function requires the sum of squares of the differences between the predicted and the measured acetone evaporation rates to be minimal. It was found convenient to express the acetone evaporation rate in terms of the instantaneous solvent-polymer mass ratio wr. The objective function has the form

d + (33)

)0

Ri ) E

sup/L0

C3

sup i i sup + C4 d + )0 sup i (34) C4 )-L/Lsup

)] |

In equations 32-34 the Neuman boundary conditions at the ends of the computational domain are substituted by the corresponding equations (5-6 and 8-10). Notice, that, due to the integration by parts, the continuity of the heat fluxes at the upper surface of the glass plate (eq 9) is satisfied automatically. The residuals are evaluated numerically using threepoint Gaussian integration. The time integration follows the Euler backward scheme. The resulting system of nonlinear algebraic equations is solved by the NewtonRaphson iterative method. The time step was equal to 10-2. The computer program exhibits quadratic convergence in two or three iterations at each time step. All results shown in the next section have been calculated with the mesh of Figure 3. This scheme depicts the computational domain along with initial values of the -coordinate and other computational

J ) min

(wri,obs - wri,pred)2 i)1

(35)

Available experimental data from Tantenkin-Ersolmaz4 at three different initial thicknesses and two different initial acetone concentrations were used. The only unknown in the parameter estimation procedure was the quantity, V/ (see eq 18). The estimated value 2 for this parameter was 6.38 10-4 m3/kg. The operating conditions used in our numerical experiments are given in Table 2. The resulting fitting is shown in Figures 4 and 5. The good agreement between the model predictions and the experimental data is notable.

Ind. Eng. Chem. Res., Vol. 38, No. 9, 1999 3577

Figure 7. Effect of V/2 parameter on the instantaneous acetone-CA mass ratio (u10 ) 0.87, L0 ) 300 m).

Figure 5. Comparison of model predictions with experimental data for two differential acetone concentrations (L0 ) 150 m).

Figure 6. Composition dependence of diffusion coefficients (temperature 298 K).

After the successful evaluation of the parameter the diffusion coefficients can be calculated in the temperature range from 0 to 30 C, according to the Vrentas and Duda theory (eq 18). In Figure 6 the estimated diffusion coefficients are plotted as a function of acetone weight fraction at 25 C and compared with the available experimental data. The prediction for the selfdiffusion coefficients of acetone is in good agreement with measurements obtained by NMR7 and by the radiotracer method.19 The values of the predicted mutual diffusion coefficient coincide with those of the selfdiffusion coefficient up to a concentration of 0.15 and are in good agreement with the reported values by sedimentation experiments.20 The observed difference in the mutual diffusion coefficient appears in the range from 0.8 to 1 solvent weight fraction. Since the error in the self-diffusion coefficient is very small in the same area, we believe that this discrepancy can be attributed to experimental errors introduced by the sedimentation measurements. The error in the V/ parameter is less 2 than (3%, as shown by numerical experimentation. It should be noted that the NMR technique and the radiotracer method yield experimental data in the high and low concentration regions, respectively, while the proposed method covers a wide range of concentration. Figure 7 illustrates the effect of the estimated parameter V/ on the instantaneous acetone-CA mass 2 ratio. This ratio increases as the value of the V/ 2 parameter increases due to the decrease in the mutual

V/, 2

Figure 8. Effect of initial thickness on polymer solution surface temperature (u10 ) 0.87).

diffusion coefficient. The variation of the estimated parameter V/ induces a substantial change in the 2 instantaneous acetone-CA mass ratio, indicating the significance of the magnitude of V/ in the value of the 2 diffusion coefficient and consequently in the modeling of the process. The ability of the model to describe the acetone evaporation process from CA solutions was tested by validation against temperature measurements at the surface of the polymer solution.4 Figure 8 shows the effects of the polymer solution initial thickness on the surface temperature of the solution. As expected, when the initial solution thickness increases, the temperature of the gas-liquid interface decreases due to the higher evaporation rate. The model predictions are in good agreement with experimental data for the first 80 s, but a discrepancy was observed for longer times. Since at that time the solvent evaporation has almost ceased, this discrepancy can be attributed to either errors in the estimation of the heat-transfer coefficient and the emissivity that are known with moderate accuracy or errors in the assumption of one-dimensional heat and mass transfer or even to inaccuracies in the experimental measurements. To study the effects of coupled heat and mass transfer on the model performance, a parametric analysis was carried out for various values of the pre-exponential factor in eqs 19 and 20, which define the heat and mass

3578 Ind. Eng. Chem. Res., Vol. 38, No. 9, 1999

Figure 9. Effect of heat- and mass-transfer coefficients on the instantaneous acetone-CA mass ratio (u10 ) 0.87, L0 ) 150 m).

Figure 11. Effect of emissivity on instantaneous acetone-CA mass ratio (u10 ) 0.87, L0 ) 150 m).

Figure 10. Effect of heat- and mass-transfer coefficients on the polymer solution surface temperature (u10 ) 0.87, L0 ) 150 m).

Figure 12. Effect of polymer free volume parameters on the estimated self-diffusion coefficient.

transfer. The results of this analysis for an initial solution thickness of 150 m are shown in Figures 9 and 10. Although the change in the heat-transfer coefficient has a significant effect on the solution surface temperature (Figure 10), the effects of the change of the masstransfer coefficient on the instantaneous acetone-CA mass ratio (Figure 9) are rather small. Consequently, the magnitudes of the heat- and mass-transfer coefficients have little effect on the estimation of the diffusion coefficient. This is also strong evidence that the solvent evaporation process is mainly controlled by diffusion in the polymer film. Another model parameter that is not available for most polymer-solvent systems is the emissivity of the polymer solution. Figure 11 shows the effects of the value of the emissivity in the process for the same experimental conditions as in previous runs of Figures 9 and 10. It can be seen that the evaporation rate is not affected at all even if the value of the emissivity is changed by 100%. The value of the emissivity has no effect on the estimation of the diffusion coefficients. In Figure 12 the error introduced into the model by assuming arbitrary values of polymer-free volume parameters is examined. It is shown that substantial changes by (300% have no effect on the parameter estimation procedure. The choice of a new value for these parameters leads to a new value for V/ which 2 has no effect on the magnitude of the objective function (eq 35) and the diffusion coefficient for a wide range of acetone concentration (20-100% w/w). This is attributed to the fact that these parameters are indeterminate; one cannot get estimates for the polymer-free volume parameters independent of V/. It should be noted 2

Figure 13. Effect of on the estimated self-diffusion coefficient.

though, that extrapolation for small solvent weight fraction (below 0.2) can be made only if good estimates for polymer-free volume parameters are available. In Figure 13 the effect of possible errors in the FloryHuggins interaction parameter is examined. The first case corresponds to the estimation of the self-diffusion coefficient using the Altena correlation (eq 17), while the second case corresponds to the estimates obtained by the parameter estimation procedure utilizing the constant value 0.5 (ideal solvent) for the interaction parameter. It is shown, that the Flory-Huggins interaction parameter has moderate effects on the estimated value for V/. Additional minor errors may be intro2 duced in the calculations, due to a change of thermophysical properties with respect to temperature or due to nonideal behavior of the solution. However, the error introduced by assuming constant thermophysical prop-

Ind. Eng. Chem. Res., Vol. 38, No. 9, 1999 3579

mental data from the literature. The diffusion coefficients are in good agreement with experimental measurements. A parameter analysis reveals that the evaporation process is mainly controlled by diffusion in the polymer film. The proposed method provides the opportunity to obtain data for the diffusion coefficients in the whole range of operational interest for temperature and concentration. It is believed that this technique can be applied to other polymer-solvent systems in the membrane manufacture process as well as in the drying of coatings for building efficient models, screening new materials,and so forth. Nomenclature
Figure 14. Typical acetone concentration profiles (u10 ) 0.87, L0 ) 300 m).

Figure 15. Typical temperature profiles (u10 ) 0.87, L0 ) 300 m).

c: overall molar density, mol/m3 Cp: polymer solution specific heat capacity, J/(kg K) D: acetone-CA mutual diffusion coefficient, m2/s D*: acetone self-diffusion coefficient, m2/s D0: scaling factor, m2/s h: heat-transfer coefficient, W/(m2 K) k: polymer solution thermal conductivity, W/(m K) kG : mass-transfer coefficient of the gas phase, s/m L0: initial thickness of the polymer solution, m P: total pressure, Pa R: gas constant, Pa m3/(mol K) t: time, s T: temperature, K Ts: polymer solution surface temperature, K u1: acetone volume fraction w: width of square plate, m Greek Letters : emissivity of the polymer solution : dimensionless axial length : dimensionless temperature F: polymer solution density, kg/m3

erties at 20 C was less than 1% on the estimated value of V/. 2 Finally, in Figures 14 and 15, typical profiles of the acetone volume fraction and the polymer solution temperature are shown as a function of the evaporation time. The acetone volume fraction varies from 0.05 to 0.87 (Figure 14), and steep gradients are observed due to the great variation of the diffusion coefficient. Temperature varies from 2 to 5 C (Figure 15), and almost linear profiles are observed due to the small change in the thermal diffusivity of the polymer solution. These figures also justify the ability of the method to estimate the diffusion coefficients over a wide range of temperature and concentration. Conclusions In the present work we developed a simple and effective method for determining the mutual and selfdiffusion coefficients for the acetone-CA system which is widely used in membrane manufacture. This method is based on modeling the coupled heat and mass transfer of this process as a one-dimensional numerical experiment with a moving boundary. All nonlinearities arising from the concentration and temperature dependence of the transport properties of the polymer solution as well as the dependence of the heat- and mass-transfer coefficients on the process variables are included. The resulting complex system of the governing equations along with appropriate initial and boundary conditions is solved by the Galerkin finite element method. The unknown parameters of the diffusion coefficient correlation based on the free-volume theory are calculated by fitting the numerical results with available experi-

Literature Cited
(1) Kesting, R. E. Synthetic Polymeric Membranes: A Structural Perspective, 2nd ed.; Wiley: New York, 1985. (2) Kunst, B.; Sourirajan, S. Evaporation Rate and Equilibrium Phase Separation Data in Relation to Casting Conditions and Performance of Porous Cellulose Acetate Reverse Osmosis Membranes. J. Appl. Polym. Sci. 1970, 14, 1983. (3) Ataka, M.; Sasaki, K. Gravimetric Analysis of Membrane Casting. I. Cellulose Acetate-Acetone Binary Casting Solutions. J. Membr. Sci. 1982, 11, 11. (4) Tantenkin-Ersolmaz, S. B. The Evaporation Step in Asymmetric Membrane Formation: Modeling, Gravimetric/Inframetric, and Morphology Studies. Ph.D. Dissertation, University of Colorado, Boulder, CO, 1990. (5) Shojaie, S. S.; Krantz, W. B.; Greenberg, A. R. Dense Polymer Film and Membrane Formation via the Dry-Cast Process: Part II. Model Validation and Morphological Studies. J. Membr. Sci. 1994, 94, 281. (6) Greenberg, A. R.; Shojaie, S. S.; Krantz, W. B.; TantenkinErsolmaz, S. B. Use of Infrared Thermography for Temperature Measurement During Evaporative Casting of Thin Polymeric Films. J. Membr. Sci. 1995, 107, 249. (7) Anderson, J. E.; Ullman, R. Mathematical Analysis of Factors Influencing the Skin Thickness of Asymmetric Reverse Osmosis Membranes. J. Appl. Phys. 1973, 44, 4303. (8) Krantz, W. B.; Ray, R. J.; Sani, R. L.; Gleason, K. J. Theoretical Study of the Transport Processes Occurring During the Evaporation Step in Asymmetric Membrane Casting. J. Membr. Sci. 1986, 29, 11. (9) Tsay, C. S.; McHugh, A. J. Mass Transfer Dynamics of the Evaporation Step in Membrane Formation by Phase Inversion. J. Membr. Sci. 1991, 64, 81.

3580 Ind. Eng. Chem. Res., Vol. 38, No. 9, 1999


(10) Shojaie, S. S. Polymeric Dense Films and Membranes via the Dry-cast Phase-Inversion Process: Modeling, Casting and Morphological Studies. Ph.D. Dissertation, University of Colorado, Boulder, CO, 1992. (11) Shojaie, S. S.; Krantz, W. B.; Greenberg, A. R. Development and Validation of a Model for the Formation of Evaporatively Cast Polymeric Films. J. Mater. Process Manuf. Sci. 1992, 1, 181. (12) Shojaie, S. S.; Krantz, W. B.; Greenberg, A. R. Dense Polymer Film and Membrane Formation via the Dry-Cast Process. Part I. Model Development. J. Membr. Sci. 1994, 94, 255. (13) Crank, J. Free and Moving Boundary Problems; Clarendon Press: Oxford, 1988. (14) Vrentas, J. S.; Vrentas, C. M. Drying of Solvent-Coated Polymer Films. J. Polym. Sci., Part B: Polym. Phys. 1994, 32, 187. (15) Guerrier, B.; Bouchard, C.; Allain, C.; Benard, C. Drying Kinetics of Polymer Films. AIChE J. 1998, 44 (4), 791. (16) Cairncross, R. A.; Jeyadef, S.; Dunham, R. F.; Evans, K.; Francis, L. F.; Scriven, L. E. Modeling and Design of an Industrial Dryer with Convective and Radiant Heating. J. Appl. Polym. Sci. 1995, 58, 1279. (17) Crank, J.; Park, G. S. Methods of Measurement. In Diffusion in Polymers; Crank, J., Park, G. S., Eds.; Academic Press: New York, 1968. (18) Tyrrell, H. J. V.; Harris, K. R. Diffusion in Liquids. A Theoretical and Experimental Study; Butterworths: London, 1984. (19) Park, G. S. Radioactive Studies of Diffusion in Polymer Systems. Part 3-Sorption and Self-Diffusion in the Acetone + Cellulose Acetate System. Trans. Faraday Soc. 1961, 57, 2314. (20) Reuvers, A. J.; Smolders, C. A. Formation of Membranes by Means of Immersion Precipitation. Part II. The mechanism of formation of membranes prepared from the system cellulose acetate-acetone-water. J. Membr. Sci. 1987, 34, 67. (21) Gresho, P. M.; Sani, R. L. Incompressible Flow and the Finite Element Method; Wiley: New York, 1998. (22) Price, P. E., Jr.; Wang, S.; Romdhane, I. H. Extracting Effective Diffusion Parameters from Drying Experiments. AIChE J. 1997, 43, 1925. (23) Vrentas, J. S.; Duda, J. L. Diffusion in Polymer-Solvent Systems I. Reexamination of the Free-Volume Theory. J. Polym. Sci., Part B: Polym. Phys. 1977, 15, 403. (24) Vrentas, J. S.; Duda, J. L. Diffusion in Polymer-Solvent Systems II. A Predictive Theory for the Dependence of Diffusion Coefficients on Temperature, Concentration and Molecular Weight. J. Polym. Sci., Part B: Polym. Phys. 1977, 15, 417. (25) Dean, J. A. Langes Handbook of Chemistry, 11th ed.; McGraw-Hill: New York, 1973. (26) Perry, R. H.; Green, D. Perrys Chemical Engineers Handbook, 6th ed.; McGraw-Hill: New York, 1984. (27) Verein Deutscher Ingenieure. VDI-Warmeatlas; VDIVerlag GmbH: Dusseldorf, 1974. (28) Jeffries, R. The Thermodynamic Properties of Mixtures of Secondary Cellulose Acetate and its Solvents. Trans. Faraday Soc. 1957, 53, 1592. (29) Crank, J. The Mathematics of Diffusion; Clarendon Press: Oxford, 1975. (30) Yilmaz, L.; McHugh, A. J. Analysis of NonsolventSolvent-Polymer Phase Diagrams and their Relevance to Membrane Formation Modeling. J. Appl. Polym. Sci. 1986, 31, 997. (31) Altena, F. Phase Separation Phenomena in Cellulose Acetate Solutions in Relation to Asymmetric Membrane Formation. Ph.D. Dissertation, Twente University of Technology, Enschede, Netherlands, 1982. (32) Hong, S. Prediction of Polymer/Solvent Solution Behavior Using Free-Volume Theory. Ind. Eng. Chem. Res. 1995, 34, 2536. (33) Zielinski, J. M.; Duda, J. L. Predicting Polymer/Solvent Diffusion Coefficients Using Free Volume Theory. AIChE J. 1992, 38, 405. (34) Bird, R. B.; Stewart, W. E.; Lightfoot, E. N. Transport Phenomena; Wiley: New York, 1960. (35) Wilke, C. R. A Viscosity Equation for Gas Mixtures. J. Chem. Phys. 1950, 18, 517. (36) Fuller, E. N.; Schettler, P. D.; Giddings, J. C. A New Method for Prediction of Binary Gas-Phase Diffusion Coefficients. Ind. Eng. Chem. 1966, 58 (5), 18. (37) Murray, P.; Carey, G. F. Finite Element Analysis of Diffusion with Reaction at a Moving Boundary. J. Comput. Phys. 1988, 74, 440. (38) Kistler, S. F.; Scriven, L. E. Coating Flows. In Computational Analysis of Polymer Processing; Pearson, J. R. A., Richardson, S. M., Eds.; Applied Science Publishers: London, 1983. (39) Finlayson, B. A. Numerical Methods for Problems with Moving Fronts; Ravenna Park Publishing Inc.: Seatle, 1992.

Received for review February 24, 1999 Revised manuscript received June 15, 1999 Accepted June 21, 1999 IE990141D

You might also like