You are on page 1of 5

International Communications in Heat and Mass Transfer 81 (2017) 51–55

Contents lists available at ScienceDirect

International Communications in Heat and Mass Transfer

journal homepage: www.elsevier.com/locate/ichmt

Fractal description of microstructures and properties of dynamically


evolving porous media
A. Verma, R. Pitchumani ⁎
Advanced Materials and Technologies Laboratory, Department of Mechanical Engineering, Virginia Tech, Blacksburg, Virginia 24061-0238, United States

a r t i c l e i n f o a b s t r a c t

Available online xxxx Transport through porous media is encountered in several engineering and biological applications. The porous
media can be subjected to changes in structure owing to deposition, erosion, swelling or shrinkage which, in
Keywords: turn, affects the transport properties of the media. A dynamic fractal model (DFM) is developed to describe the
Fractal dimensions evolution in pore structure undergoing deposition using fractal dimensions and to predict the changes in the ef-
Dynamic fractal model fective diffusivity in terms of the dynamic fractal dimensions. Evolving microstructures undergoing deposition
Porous media
are analyzed at various saturation levels to determine the effective diffusivity using the dynamic fractal model.
Saturation
Diffusivity
The effective diffusivity values of the evolving porous media are compared against existing data in the literature.
Image analysis © 2016 Elsevier Ltd. All rights reserved.

1. Introduction analysis of reconstructed pore structure to characterize the transport


properties. Chen et al. [6] used X-ray computed microtomography
The structure of porous media commonly undergoes changes over (XCMT) to construct three-dimensional (3D) geometry of the pore struc-
time as part of several physical processes such as flow in fuel cell layers, ture before and after colloid deposition, followed by Lattice Boltzmann
transport in batteries, fouling of membranes during filtration, percola- (LB) simulations to evaluate change in local permeability and tortuosity.
tion of minerals through rocks, contaminant transport and carbon- It was found that the change in permeability followed a power law var-
dioxide storage as seen in oil and gas industry, flow in biological tissue, iation with respect to porosity and the results significantly differed from
and chemical vapor infiltration. Describing transport properties such as the predictions using Kozeny-Carman relationship. Similar studies in-
permeability, diffusivity and conductivity in these dynamically evolving volving XCMT and LB modeling were performed by Okabe et al. [7].
porous media is critical to fundamental understanding, design and opti- The most common practice is to use the Bruggeman equation [15]
mization of such systems and forms the focus of the proposed study. that relates the effective transport properties to the porosity of the me-
Several experimental studies in the past have focused on measure- dium and a constant term denoting the tortuosity. A limitation of this
ment of transport properties in porous media by quantifying pressure description is that any two geometries (pore structure) with same po-
drop, species concentration and flow. Recent literature has seen in- rosity would exhibit similar behavior irrespective of their morphology,
crease in numerical work involving reconstruction of porous media which is inaccurate. In the absence of a comprehensive and accurate
combined with pore scale modeling to predict transport properties theoretical model, researchers use experimental data to obtain a corre-
[1–7]. Owing to their disordered nature, pore structures can be well de- lation between tortuosity and porosity for different pore structures. The
scribed by fractal dimensions that, in turn, are used to predict transport resulting models are, therefore, correlatory and not predictive. In the
properties such as permeability, conductivity and diffusivity [8–14]. Al- case of dynamic changes in pore structures, the number of experiments
though the properties of porous media have been explored extensively, increases exponentially for a comprehensive description and, in some
the study of changes in pore structure and, in turn, the corresponding cases, the measurements require sophisticated tools and extensive ef-
properties owing to deposition, erosion swelling or shrinking remains fort to effectively determine the changes in pore structure. Moreover,
a relatively unexplored area. the use of oversimplified models that are inaccurate, or the reliance on
Previous theoretical and experimental studies have mainly focused empirical expressions that involve tuning factors or post-facto correla-
on determining the properties for pore structures at the beginning of de- tions of experiments do not offer a truly predictive approach to describ-
position or erosion, followed by use of analytical models or experimental ing evolving porous media and their properties.
correlations to describe changes in transport properties with deposition. In this communication, we address the challenge of predicting the
The few studies relating to permeability reduction have used pore scale evolution of pore structure as a function of deposition, saturation or ero-
sion using a fractal model. A model to predict the changes in fractal di-
⁎ Corresponding author. mensions for a pore structure undergoing deposition is developed and
E-mail address: pitchu@vt.edu (R. Pitchumani). compared with the predictions obtained from image analysis. The

http://dx.doi.org/10.1016/j.icheatmasstransfer.2016.11.008
0735-1933/© 2016 Elsevier Ltd. All rights reserved.
52 A. Verma, R. Pitchumani / International Communications in Heat and Mass Transfer 81 (2017) 51–55

variation in effective diffusivity is predicted with the dynamic fractal


model and compared with analytical solutions from other studies.

2. Fractal model

Fig. 1 shows the microstructural image of a disordered porous medi-


um [12], where the black regions represent the pores and the white
areas denote the solid matrix formed by the agglomeration of copper
particles. Owing to the randomness and presence of multiple length
scales, common to a wide range of porous media encountered in practi-
cal applications, the porous medium in Fig. 1 can be represented by frac-
tal dimensions [1–10,14]. In a fractal representation, the pore
architecture is usually quantified in terms of fractal dimensions dT and
dN, corresponding to the tortuosity and area dimensions, respectively.
The porous medium in this description can be envisioned to be a distri-
bution of tortuous capillaries of various sizes, with the effective length,
Lc, of a capillary of size λ given by [8–13]:

Lc ðλÞ ¼ λ1−dT Ld0T ð1Þ

where L0 is a representative length and dT is the tortuosity fractal di-


mension, such that a value of dT = 1 corresponds to a straight capillary
and dT = 2 corresponds to a highly tortuous medium. For a complete de-
scription of the pore structure, it is essential to also quantify the number
of capillary pathways corresponding to every pore size, λ. The popula-
tion of the intersecting pores in a cross section exhibits the general
trend that N(L ≥ λ), which denotes the total number of pores of size ex-
ceeding or equal to a value λ, increases as the pore size decreases.
Hence, the cumulative pore distribution in a cross section can be repre-
sented by [8–13]: The theory of fractal dimensions and representation
Fig. 2. Box-counting analysis for (a) area flow dimension, dN and (b) tortuosity dimension,
of porous media using these dimensions is well established and is not dT.
discussed in detail here. Readers are referred to Refs. [8–14] for a
more detailed explanation.
 
λmax dN dN (= 1.76 for Fig. 1), which is representative of the cumulative pore
NðL ≥λÞ¼ ð2Þ
λ distribution as given by Eq. (2). The value calculated by Yu and Cheng
[12] for the above geometry was 1.79 considering a bi-dispersed porous
The area and tortuosity dimensions can be obtained from a box- media, which is very close to the value obtained here, while treating the
counting analysis [9,10] of the area and the perimeter of the pores, re- media as mono-dispersed. Similarly, by using the box-counting method
spectively, in a microstructural image of the porous structure. The re- on the perimeter of the pores [9,10], the tortuosity dimension, dT, can be
sults of box-counting analysis of the microstructure in Fig. 1 are obtained from the slope of a linear fit through data on a log-log plot to
presented in Fig. 2, where Fig. 2(a) describes the log-log variation of be dT (=1.18 for Fig. 1), as seen in Fig. 2(b).
the cumulative pore distribution as a function of pore size. The negative
of slope of a linear fit through the data gives the area fractal dimension, 3. Deposition in porous media

Based on the fractal description discussed above, the changes in pore


structure as a function of saturation or deposition is modeled. Deposi-
tion (or saturation) is simulated in porous media by numerically dilat-
ing [16] the solid boundaries in the pore structure image to mimic a
physical deposition in pores. Fig. 3 shows the pore structure images
with different saturation values, where saturation is defined
as, s = 1 − εnew/ε0, with ε0 and εnew as the porosity of the medium be-
fore and after deposition, respectively. The porosity and saturation
values for Fig. 3(a), (b), (c) are s = 0, 0.116, 0.208 and ε = 0.575,
0.508, 0.456, respectively. The methodology to quantify changes in frac-
tal dimensions with changes in saturation levels is discussed next.

3.1. Evaluating changes in dN

According to Yu and Cheng [12], the area fractal dimension, dN, can
be expressed as:

lnε
dN ¼ k− ð3Þ
λ
ln min
Fig. 1. Microstructure image of a disordered porous media [12]. λmax
A. Verma, R. Pitchumani / International Communications in Heat and Mass Transfer 81 (2017) 51–55 53

s=0 s = 0.116 s = 0.208

Fig. 3. Pore structure images with different levels of saturation.

where k represents the dimensionality namely, 2 for two dimensions where α = λmin/λmax. For very small α,
and 3 for three dimensions, and λmin and λmax are the minimum and
maximum size of the pores, respectively. For the present case,  dT
1 dN
Leff ¼ λmax ð10Þ
lnε β ð1−dT −dN Þ
dN ¼ 2− : ð4Þ
λ
ln min
λmax
Similarly, the effective length at any level of saturation, s, can be
0 0 given by:
It follows that for change in porosity from ε to ε(s) = ε (1− s):

2−dN ln ε 0 ln ε 0  0 0
1 dT dN
0 ¼ ¼ ð5Þ Leff ðsÞ ¼ λmax 
2−dN lnεðsÞ ln ε 0 ð1−sÞ 0 0  ð11Þ
βðsÞ 1−dT −dN

where dN′ is the area fractal dimension of the pore structure at satura-
tion, s, such that dN′b dN for increase in saturation, s. Pitchumani and
Ramakrishnan [9,10] used area fraction to establish relationship be-
tween the pore geometry and fractal dimensions. Since the area fraction
is a representative of volume fraction it was derived that:

 
dN ¼ 2 1−β2 ; β ¼ λmax =L0 ð6Þ

3.2. Evaluating changes in dT

For small values of deposition, it can be envisioned that there is no


change in the effective length of the pores, and it follows from Eq. (1)
that

!d0T −1  
Lc ðλÞ L0 L0 dT −1
¼ pffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ ð7Þ
L0 λ0 ð1−sÞ λ0

pffiffiffiffiffiffiffiffiffiffiffiffiffi
where dT′ and λ0 ð1−sÞ are, respectively, the area dimension and pore
size at saturation, s. Accordingly, dT′ can be written as:

 
L0
ln
0 λ0
dT ¼ 1 þ ðdT −1Þ ! ð8Þ
L0
ln pffiffiffiffiffiffiffiffiffiffiffiffiffi
λ0 ð1−sÞ

The cumulative length of the pores, Leff, is given by:

 dT  
eff λ 1 1−α 1−dT −dN
L ¼ ∫ λmax Lc ðλÞdN ðλÞ ¼ λmax dN ð9Þ Fig. 4. Variation in (a) area dimension, dN and (b) tortuosity dimension, dT, as a function of
min
β ð1−dT −dN Þ saturation.
54 A. Verma, R. Pitchumani / International Communications in Heat and Mass Transfer 81 (2017) 51–55

where Deff is the effective diffusion coefficient, A0 is the representative


area and m(λ) is the molar mass transfer rate through a single capillary
of size λ [8–13]. The effective diffusion coefficient is then given as a
function of fractal dimensions dN and dT as:

 
1−α 1þdT −dN
Deff ¼ g q D βdT þ1 dN ð14Þ
ð1 þ dT −dN Þ

where gq = π/4.
The change in effective diffusivity as a function of saturation (change
in porosity) is presented in Fig. 5, where the fractal model is compared
to available models in the literature. A commonly used relationship to
describe changes in Deff as a function of porosity is Deff = ε1.5D, referred
to as Bruggeman equation [15]. Other models describe change in diffu-
Fig. 5. Effective diffusivity as a function of porosity.
sivity as, Deff = D/NM, where NM is the MacMullin number, a function of
porosity [17,18]. The shaded area in Fig. 5 represents region between
For small values of saturation, it may be assumed that Leff = Leff(s), the upper and lower bounds formed by the values of Deff as calculated
and from Eqs. (10) and (11) it can be deduced that: using the MacMullin numbers given in Table 1. The solid line in Fig. 5
denotes the effective diffusivity as calculated using the Bruggeman
 dT −d0T 
1 pffiffiffiffiffiffiffiffiffiffiffiffiffi1−d0T ð1−dT −dN Þ d0N equation. The dashed line represents the results obtained using
¼ ð1−sÞ  0 0  ð12Þ Eq. (14), where the fractal dimensions are function of porosity (equiva-
β 1−dT −dN dN
lently, saturation) as given by Eqs. (5) and (12). The chain-dashed line
in Fig. 5 denotes the results calculated from Eq. (14) using the fractal di-
Thus, from Eq. (12) it can be inferred that dT′(β, dN, s) b dT for s N 0. mensions obtained from image analysis of pore structures with simulat-
Fig. 4(a) presents the variation in dN as calculated from the box- ed deposition as seen in Fig. 3. It is noted from Fig. 5 that the trends
counting method and that obtained from Eq. (5). The images with dif- described by dynamic fractal model [Eq. (14)] and Deff = ε1.5D are sim-
ferent levels of saturation were created, similar to that described for ilar, and show a decrease in Deff with decrease in porosity, which is
Fig. 3, for extracting fractal dimensions using image analysis. It can be equivalent to increase in saturation. It can also be seen that the predic-
seen that dN decreases with increase in saturation, with Eq. (5) tions by the fractal model, using the values obtained through image
predicting a similar trend to that described by image analysis. It is analysis (chain-dashed line) and those given by the dynamic fractal
noted that the predictions from Eq. (5) show a larger deviation from model (dashed line) lie within the bounds in the literature (shaded
the results obtained by box-counting method (image analysis) with in- area), confirming the validity of a fractal description of the pore struc-
crease in saturation. This can be attributed to the fact that a theoretical ture, and that the dynamic fractal model can be effectively used to pre-
Sierpinski triangle model [12] was used to derive the relationship in dict transport properties for evolving porous media.
Eq. (4) and, in turn, Eq. (5) whereas, in reality, the porous structure dif- The present model, based on uniform saturation works well for cases
fers from the idealized Sierpinski construct, as evident from the fractal of low saturation, where there are no significant changes to the flow
dimension obtained through image analysis. In the absence of a better path. A more comprehensive description may be developed by, relating
comprehensive theoretical model to accurately describe the changes the deposition, pore size distribution and pore networking, such that dT
in dN′ as seen in Fig. 4(a), Eq. (5) may be considered to best represent has different variations for range of saturation values. A pore network
the evolution of area fractal dimension for very small values of model for random three-dimensional geometry, coupled with image
saturation. analysis can provide further insights and can be studied as a part of fu-
Fig. 4(b) shows the variation in tortuosity fractal dimension dT as a ture work. Also, the effects of multiphase flow can be characterized
function of saturation where the results from image analysis (box- using the above formulation.
counting method) are represented by solid line. It can be seen that dT re-
duces with the increase in saturation, and that the results from the dy- 4. Conclusions
namic model (Eq. (12)) closely predict the changes as observed from
image analysis. The reduction in dT may be understood by considering A dynamic fractal model was presented for describing dynamically
that the pores with smaller diameter become less relevant to describing evolving porous media as function of saturation. The model closely re-
the pore structure with increase in saturation, resulting in increase in sembles the values obtained through image analysis for small values
average pore size and correspondingly smaller values of dT. It is to be of saturation. The dynamic fractal model was used to predict the chang-
noted here that above relationship is derived for smaller values of satu- es in diffusion coefficient as a function of saturation. The model provides
ration and might not hold true for larger saturations where there is sig- an effective approach to predicting the properties of porous media with
nificant restructuring of transport path, and a distribution of dT over evolving microstructures encountered in many fields.
different ranges of saturations.

Table 1
3.3. Evaluating changes in effective diffusivity
MacMullin number as a function of porosity for various geometries, arrangement and size
Refs. [17,18].
Following an analysis of the transport through the fractal capillaries
Label Expression
representing the porous medium, as in Refs. [8–13], the molar mass
transfer rate, M, through the porous medium by diffusion is given by: I NM = [(5 − ε)(3 + ε)]/[8(1 + ε)ε]
II NM = [(3 − ε)(43 + 0.409(1 − ε)7/3)
− 1.315(1 − ε)10/3]/[2ε(43 + 0.409(1 − ε)7/3) − 1.315(1 − ε)10/3]
ΔC λ NM = [2 − ε − 0.3058(1 − ε)4 − 1.334(1-ε)8]/[ε
M ¼ A0 Deff ¼ ∫ λmax qðλÞdNðλÞ III
L0 min
  − 0.3058(1 − ε)4 − 1.334(1 − ε)8]
dT 1−α 1þdT −dN IV NM = [2 − ε − 0.3058(1 − ε)4–1.334(1 − ε)8]/[ε − 0.3058(1 − ε)4
¼ g q D ΔC β λmax dN ð13Þ − 1.334(1 − ε)8]
ð1 þ dT −dN Þ
A. Verma, R. Pitchumani / International Communications in Heat and Mass Transfer 81 (2017) 51–55 55

References [9] R. Pitchumani, B. Ramakrishnan, A fractal geometry model for evaluating permeabil-
ities of porous preforms used in liquid composite molding, Int. J. Heat Mass Transf.
[1] C. Manwart, U. Aaltosalmi, A. Koponen, R. Hilfer, J. Timonen, Lattice-Boltzmann and 42 (12) (1999) 2219–2232.
finite-difference simulations for the permeability for three-dimensional porous [10] B. Ramakrishnan, R. Pitchumani, Fractal permeation characteristics of preforms used
media, Phys. Rev. E 66 (1) (2002) 016702. in liquid composite molding, Polym. Compos. 21 (2) (2000) 281–296.
[2] L. Chen, L. Zhang, Q. Kang, H.S. Viswanathan, J. Yao, W. Tao, Nanoscale simulation of [11] B. Yu, J. Li, Some fractal characters of porous media, Fractals 9 (03) (2001) 365–372.
shale transport properties using the lattice Boltzmann method: permeability and [12] B. Yu, P. Cheng, A fractal permeability model for bi-dispersed porous media, Int. J.
diffusivity, Sci. Rep. 5 (2015) 8089. Heat Mass Transf. 45 (14) (2002) 2983–2993.
[3] T. Gebäck, M. Marucci, C. Boissier, J. Arnehed, A. Heintz, Investigation of the effect of [13] B. Yu, W. Liu, Fractal analysis of permeabilities for porous media, AICHE J. 50 (1)
the tortuous pore structure on water diffusion through a polymer film using lattice (2004) 46–57.
Boltzmann simulations, J. Phys. Chem. B 119 (16) (2015) 5220–5227. [14] B.B. Mandelbrot, The Fractal Geometry of Nature, W.H. Freeman and Company,
[4] H. Koku, R.S. Maier, K.J. Czymmek, M.R. Schure, A.M. Lenhoff, Modeling of flow in a 1982.
polymeric chromatographic monolith, J. Chromatogr. A 1218 (22) (2011) [15] D.A.G. Bruggeman, Calculation of various physics constants in heterogeneous sub-
3466–3475. stances I Dielectricity constants and conductivity of mixed bodies from isotropic
[5] L. Chen, G. Wu, E.F. Holby, P. Zelenay, W.Q. Tao, Q. Kang, Lattice Boltzmann pore- substances, Ann. Phys. 24 (7) (1935) 636–664.
scale investigation of coupled physical-electrochemical processes in C/Pt and non- [16] Image Processing Toolbox™, The Math Works Inc. http://www.mathworks.com/
precious metal cathode catalyst layers in proton exchange membrane fuel cells, help/images (viewed July 2016).
Electrochim. Acta 158 (2015) 175–186. [17] M.J. Martínez, S. Shimpalee, J.W. Van Zee, Measurement of MacMullin numbers for
[6] C. Chen, A.I. Packman, J.F. Gaillard, Pore-scale analysis of permeability reduction PEMFC gas-diffusion media, J. Electrochem. Soc. 156 (1) (2009) B80–B85.
resulting from colloid deposition, Geophys. Res. Lett. 35 (7) (2008), L07404. [18] Y. Wang, S.C. Cho, K.S. Chen, PEM Fuel Cells: Thermal and Water Management Fun-
[7] H. Okabe, M.J. Blunt, Prediction of permeability for porous media reconstructed damentals, Momentum Press, 2013.
using multiple-point statistics, Phys. Rev. E 70 (6) (2004) 066135.
[8] D.A. Weitz, M. Oliveria, Fractal structures formed by kinetic aggregation of aqueous
gold colloids, Phys. Rev. Lett. 52 (16) (1984) 1433.

You might also like