Microuid Nanouid (2008) 4:257260 DOI 10.

1007/s10404-007-0182-3

SHORT COMMUNICATION

Determination of the effective diffusion coefcient in porous media including Knudsen effects
Deqiang Mu Zhong-Sheng Liu Cheng Huang Ned Djilali

Received: 8 May 2006 / Accepted: 8 May 2007 / Published online: 16 June 2007 Springer-Verlag 2007

Abstract A 3-D bond pore network model is presented and used to evaluate the effect of pore size and connectivity on the effective diffusion coefcient in random porous media. The control equations of the system are set up and the simulation method is discussed. The simulation results show that when the average pore size dm < 1 lm, the effective diffusion coefcient is strongly dependent on pore size. The analysis shows that Knudsen and bulk diffusion effects can be decoupled, and for any given diffusion conditions, the effect coefcient accounting for Knudsen diffusion can be obtained. Thus, the effect of pore size can been readily accounted for by correcting the bulk effective diffusivity with the Knudsen effect coefcient. The simulations also show that the percolation threshold in random porous materials decreases with increased pore network connectivity. Keywords Effective diffusion coefcient Pore network Knudsen diffusion Percolation threshold Porous media

1 Introduction Gas transport through a porous material is of importance in many applications, including for example design and analysis of gas electrodes. In particular it is often necessary to calculate the ux of gas diffusing through the material, and this depends on the internal microstructure of the pores. Because of the solid portions through which, in general, there is no diffusion, the gas diffusion rates through porous materials can be signicantly lower than through air, depending on porosity, pore sizes and pore connections. Usually the effective diffusion coefcient Deff is correlated to the corresponding coefcient through air Db and the porosity e. Such relations can be derived for simplied limiting case assumptions. For example, under the assumption that no diffusion takes place through the solid portions of the porous material and for ordered inorganic whiskers, the cross section area available for diffusion would be the average area of the pore space, that is, eA, where A is the total area of the cross-section. This yields Deff/Db e. Similar relationships have been proposed for porous materials such as textiles (Mezedur et al. 2002; Nam and Kaviany 2003; Tomadakis and Sotirchos 1993) and porous solid particle materials (Maxwell 1954; Rayleigh 1892; Weissberg 1963; Bruggeman 1935) and they are often adequate for calculating gas diffusion. However, such formulae are inadequate for calculating the dependence of Deff/Db on the internal microstructure including porosity, pore shape, pore size distribution and pore connectivity. Furthermore, the complexity increases considerably when Knudsen effects have to be considered. The objective of the present work is to evaluate the effect of pore size distribution and connectivity (coordination number) on Deff/Db including Knudsen effects by using a 3-D pore network model. This work was motivated by our

D. Mu Z.-S. Liu (&) C. Huang NRC Institute For Fuel Cell Innovation, 4250 Wesbrook Mall, Vancouver, BC, Canada V6T 1W5 e-mail: simon.liu@nrc.gc.ca Present Address: D. Mu Changchun University of Technology, Changchun, China N. Djilali Institute for Integrated Energy Systems, University of Victoria, PO Box 3055 STN CSC, Victoria, BC, Canada V8W 3P6

123

258

Microuid Nanouid (2008) 4:257260

recent efforts to evaluate different concepts of fuel cell porous catalyst layers with very different microstructures, random or ordered, wide range of pore sizes from 5 to 100 nm, corresponding to a Knudsen number > 1, and dominance of Knudsen diffusion (Kast and Hohenthanner 2000) The model presented here allows a direct resolution of the porous microstructures and hence accurate quantication of Deff/Db as a function of the porous media geometry.

where Di,j is the effective diffusion coefcient of the pore connecting site i and site j, qi and qj are gas concentrations at site i and site j, respectively. Equations (1) and (2) can be recast into a matrix form as AX B 3

2 Description of the modeling method We assume that the microstructure of porous materials can be approximately represented by the 3-D bond pore network model shown in Fig. 1. In the model, a pore is represented by a bond segment, which connects two nodes (sites). It also assumed that the shape of a bond segment is circular with its diameter denoted as dp. The pore size dp is specied by its probability distribution function. The porosity is achieved by randomly deleting a portion of bond segments. The length of a bond segment, the distance between two nodes (sites), is assumed to be longer than the maximum pore diameter dmax. The number of the bond segments that are connected to the same site is dened as the coordination number z, representing the connectivity of the pores. When different gas concentrations, q1 and q2, are applied, respectively to the two opposite surfaces, gas diffusion takes place through the pores (bond segments) only. Here we consider steady state diffusion. The mass balance at site i requires
z X j1

where X is a vector containing the unknown gas densities at different sites; A is a matrix with its diffusion coefcients as its elements, and B is a vector dependent on the gas concentration values at inlet and outlet faces, and the diffusion coefcients of the pore segments connected to these boundaries. Here matrices A and B are constant, so Eq. (3) is easily solved. To estimate the effect of pore size on the effective diffusion coefcient and the Knudsen effect, we use the Dusty-gas model (DGM) (Mason et al. 1967; Mason and Malinauskas 1983) and Bosanquet formula (Pollard and Present 1948), where the diffusion coefcient Dp is given by 1 1 1 Di;j Db Dk 4

where Di,j is the diffusion coefcient in a single cylindrical pore, Db is the bulk diffusion coefcient, and Dk is the Knudsen diffusion coefcient (Kast and Hohenthanner 2000; Cunningham and Williams 1980) given by r T ; Dk % 48:5 dp M

mi;j 0

where mi,j is mass transport rate in the pore connecting sites i and j. According to Ficks law, the mass transport rate for a single-phase pore is mi;j Di;j qi qj 2

in which T is temperature and M is molecular weight. Using Eqs. (1)(5), one can calculate the gas ux F, and then the effective diffusion coefcient Deff can be obtained using F Deff q2 q1 6

From Eq. (6) one can get the effective diffusivity Deff/Db.

3 Simulation results and discussions Along the lines of the model described above, we build the 3-D bond pore network geometry, as shown in Fig. 1, to represent the internal microstructure of a random porous solid. For simplicity, the sites in Fig. 1 are uniformly distributed in the cube, and all of the sites except those on the boundary surface have the same coordination number. After the network is established, a fraction of the total bond segments, e, is randomly chosen from the total bond segments and is designated as pores, and the rest as solid. On the front and back surfaces of the cube, gas concentrations

Fig. 1 Pore network

123

Microuid Nanouid (2008) 4:257260

0.4

259
1.0

Effective diffusivity Deff / Db

Knudesn factor K (dm )

0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 1 10

Porosity= 0.4 porosity= 0.6

0.3

0.2

Porosity=0.4, No Kn. Porosity=0.4, Kn. Porosity=0.6, No Kn. Porosity=0.6, Kn.

0.1

0 0 200 400 600 800 1000

100

1000

10000

100000

Mean pore size d m / nm

Mean pore size d m / nm

(a) Z= 6
0.6

(a) Z= 6
1.0

Effective diffusivity Deff / Db

Knudsen factor K (d m)

0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 1 10

Porosity= 0.4 Porosity= 0.6

0.5 0.4 0.3 0.2 0.1 0 0 200 400 600 800 1000

Porosity=0.4, No Kn. Porosity=0.4, Kn. Porosity=0.6, No Kn. Porosity=0.6, Kn.

100

1000

10000

100000

Mean pore size d m /nm

Mean pore size d m / nm

(b) Z=12
Fig. 3 Effect of mean pore size on Knudsen factor

(b) Z=12
Fig. 2 Effect of mean pore size on effective diffusivity
1.0

Deff /Db

0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0.0 0.1 0.2

q1 and q2, are applied, respectively, and on the other four surfaces zero ux (q(x,y,z)/n = 0) is specied. The pore diameter dp distribution is assigned using the Gauss probability density function. Since the Knudsen effect is dependent on temperature and molecular weight, a temperature 343 K (70C) and oxygen molecular weight M = 32 kg kmol1 and a bulk diffusion coefcient Db = 2.9 105 m2/s are prescribed used in the calculations. Calculations show that the boundary effect can be neglected when the number of total sites is as large as 20 20 20. 3.1 Effect of pore diameter distribution In Fig. 2a, b, we show the dependence of Deff/Db upon the mean pore diameter. In these calculations, we assume the standard deviation of pore diameter is one tenth of the mean. The results in Fig. 2a, b correspond to coordination numbers of 6 and 12, respectively. Consistent with DGM predictions for a single pore, Deff/Db increases with the mean pore diameter when the Knudsen effect is considered, whereas Deff/Db is independent of the mean pore diameter s when the Knudsen effect is neglected. Deff/Db with the Knudsen effect is smaller than Deff/Db without the Knudsen

Effective diffusivity

z =6 z =8 z = 12

0.3

porosity

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Fig. 4 Effect of coordination number on effective diffusivity

effect counted, but the two values converge as the mean pore diameter increases. In order to quantify the Knudsen effect contribution, we express Deff/Db as Deff =Db Kdm Fe 7

where K(dm) represents the effect of pore diameter on diffusion, which we refer to as a Knudsen factor, and F(e) represents the dependence of Deff/Db on porosity e only. Based on the data from Fig. 2a, b, we extracted K(dm), as

123

260 Fig. 5 Distribution of gas concentration in the pore network

Microuid Nanouid (2008) 4:257260

shown in Fig. 3a, b. Interestingly, K(dm) is almost independent of porosity e and coordination number z. This supports the decomposition Deff/Db into the two terms, the Knudsen term K(dm) and porosity term F(e), Deff/Db = K(dm) F(e). 3.2 Effect of pore connectivity The coordination number z represents pore connectivity. Figure 4 shows the effect of the coordination number z on the effective diffusivity Deff/Db. Here we specify the mean pore size dm = 400 nm, its the standard deviation r = 10 nm. The Knudsen effect is not considered in these calculations. It can be seen that the coordination number z affects signicantly the effective diffusivity, in particular, near the region of percolation threshold. The larger the coordination number z, the higher Deff/Db. It can also be seen that within the low porosity region, the relationship between Deff/Db and the porosity becomes increasingly non-linear. This non-linearity is partly due to the rapid increase of the number of dead pores when porosity is low, as shown in Fig. 5.

For dm < 1 lm, it is essential to account for the effect of Knudsen diffusion on effective diffusion coefcient. A percolation threshold exists in random porous materials and decreases with increasing connectivity of the pore network. For low porosity, the relationship between the effective diffusion coefcient and porosity is strongly nonlinear.

Acknowledgments The authors thank Prof. M. Eikerling, Department of Chemistry, Simon Fraser University, British Columbia, Canada, for his helpful discussions and comments.

References
Bruggeman DAG (1935) Berechnung verschiedener physicalischer konstaten von heterogenen substanzen. Ann Phys 24:636664 Cunningham RE, Williams RJJ (1980) Diffusion in gases and porous media. Plenum, New York Kast W, Hohenthanner CR (2000) Mass transfer within the gas-phase of porous media. Int J Heat Mass Transfer 43:807823 Mason EA, Malinauskas AP (1983) Gases transport in porous media: the dusty-gas model. Elsevier, Amsterdam Mason EA, Malinauskas AP, Evans RB (1967) III. Flow and diffusion of gases in porous media. J Chem Phys 46:31993216 Maxwell JC (1954) A treatise on electricity and magnetism, vol 1, 3rd edn. Dover, New York Mezedur MM, Kaviany M, Moore W (2002) Effect of pore structure, randomness and size on effective mass diffusivity. AIChE J 48:1524 Nam JH, Kaviany M (2003) Effective diffusivity and water-saturation distribution in single- and two-layer PEMFC diffusion medium. Int. J Heat Mass Transfer 46:45954611 Pollard WG, Present RD (1948) On gaseous self-diffusion in long capillary tubes. Phys Rev 73:752774 Rayleigh (1892) On the inuence of obstacles arranged in rectangular order upon the properties of a medium. Phil Mag 34:481503 Tomadakis MM, Sotirchos SV (1993) Ordinary and transition regime diffusion in random ber structure. AIChE J 39:397412 Weissberg HL (1963) Effective diffusion coefcients in porous media. J Appl Phys 34:297303

4 Conclusions In this paper, a 3-D bond pore network model and simulations are used to evaluate the effect of pore size and connectivity on the effective diffusion coefcient of random porous media. The simulations show that: For a given diffusion conditions, the effective coefcient accounting for Knudsen diffusion can be obtained using the present approach. This allows a direct and accurate correction of the effective diffusion formula accounting for the effect of pore size.

123