You are on page 1of 80

Journal of

Porous Media
Volume 6, Number 1, 2003

Contents
Nonlinear Convection in Porous Media: A Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . N. Rudraiah, P. G. Siddheshwar, and T. Masuoka Mass Transfer Jump Condition at the Boundary between a Porous Medium and a Homogeneous Fluid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . J. J. Valencia-Lopez,G. Espinosa-Paredes, and J. A. Ochoa-Tapia Effects of Gross Heterogeneity and Anisotropy in Forced Convection in a Porous Medium: Layered Medium Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D. A. Nield and A. V. Kuznetsov The Effect of the Local Inertial Term on the Free-Convection Fluid Flow in Vertical Channels Partially Filled with Porous Media . . . . . . . . . . . . . . . . . . . . . A. F. Khadrawi and M. A. Al-Nimr Numerical Study of Boiling in an Inclined Porous Layer . . . . . . . . . . . . . . . . . . . . . . . . Mustapha Najjari and Sassi Ben Nasrallah 1

33

51

59 71

Journal of Porous Media 6(1), 132 (2003)

1RQOLQHDU &RQYHFWLRQ LQ 3RURXV 0HGLD $ 5HYLHZ


N. Rudraiah,1 P. G. Siddheshwar,2 and T. Masuoka2
1

National Research Institute for Applied Mathematics (NRIAM), 492/G, 7th Cross 7th Block (West), Jayanagar, Bangalore, 560 082,India, and UGC-CAS in Fluid Mechanics, Department of Mathematics, Central College, Bangalore University, Bangalore, 560 001, India and 2 Department of Mechanical Engineering, Kyushu University, Kyushu, Japan ABSTRACT The review article deals mainly with nonlinear convection (NLC) in porous media and discusses analytical and numerical techniques to handle them. A section on experimental heat transfer in porous media relates to the state of the art on the subject. Five techniques for studying NLC are presented in a logical order of traversal. The first technique, due to Lyapunov, is useful to obtain energy bounds and in conjunction with a variation technique can be used to investigate for possible subcritical instabilities. The power integral and the spectral methods concern steady finite-amplitude convection. The former method is useful in obtaining information on physically preferred cell patterns, whereas the latter can handle cross-interaction of different modes in addition to estimating heat transfer. The Fourier decomposition for unsteady large-amplitude convection is capable of predicting chaos and quantifying heat transfer. The finite-difference method, or any numerical method, when guided by the results of the existing analytical methods and experiment, can be used effectively to handle a more general problem with realistic boundary conditions. The results of the experimental and theoretical study are meant to mutually ratify the respective findings. The present scenario on heat transfer in porous media is such that not all observed aspects can be covered in a theoretical study and also not all results predicted by the theory are experimentally realizable. It thus calls for concerted effort from various quarters. It is on this ground that the review puts together many aspects of NLC in porous media, taking essential excerpts from previous works, with an unavoidable lean on the authors own works. 1

Received May 30, 2001; Accepted February 4, 2002 Copyright 2003 Begell House, Inc.

Rudraiah et al.

Nonlinear Convection in Porous MediaA Review INTRODUCTION Ever since the pioneering work of Darcy (1856) resulting in the Darcy equation, considerable interest has been evinced in the study of different aspects of flow through porous media (see Muskat, 1937; Scheidegger, 1960; Bear, 1972; Rudraiah et al., 1979; Greenkorn, 1983; and Vafai, 2000), flow past a porous medium (see Beavers and Joseph, 1967; Taylor, 1971; Richardson, 1971; Saffman, 1971; Rudraiah, 1985; and Prasad 1991), and heat and mass transfer in porous media (see Rudraiah, 1984, 1989; Kaviany, 1991; Nield and Bejan, 1999; and Vafai, 2000), because of its natural occurrence and of its broad range of scientific, engineering, and technological activities, particularly in petroleum engineering, geographic activities, geohydrology, transpiration cooling, cooled thermal protection of systems, nuclear and thermal waste disposal in geologic media, materials processing, biomechanics, and so on. Our main interest in this article is to review some of the literature on techniques used to study nonlinear convection (NLC) in porous media. Earlier works/reviews (see Wooding and White, 1984; Rudraiah, 1984, 1989, 2000; Rudraiah and Friedrich, 1984; Kaviany, 1991; Nield and Bejan, 1999; Masuoka, 1999; and Vafai 2000) concentrated on explaining the complicated flow paths and phenomena in porous media depending on certain simplifications based on volume-averaging techniques involving spatial and temporal scales specifying the phenomena under consideration. Some of the reviews have also covered flows in porous media that exhibit time-dependent, chaotic, and turbulent flow behaviors in spite of highly dissipative effects of viscosity, thermal conductivity, and resistance offered by the solid particles to the flow. In view of the difference of opinion in the use of suitable nonlinear equations as well as on merits and demerits of using different techniques to solve these nonlinear equations governing convection in porous media, it is our opinion that an article addressing these aspects is warranted, even though other articles have appeared on the same topic but on different aspects; e.g., the most recent book on porous media, edited by Vafai (2000), has some focused articles on thermal and double-diffusive convection and allied topics. To the knowledge of the authors, there is no circumspect article devoted exclusively to NLC and hence this review article on the topic. Before we embark on a survey of the literature and the methods, we include a short discussion on the non-Darcy porous media equations and refer the reader to Vafai (2000)

3 for more details. The most general non-Darcy momentum equation, known as the DarcyLapwoodForchheimer Brinkman (DLFB) equation, is 1 1 q q + (q )q = p + g k t 2 cbq q + ef 2 q k

(1)

The second term on the left-hand side (LHS) of the above equation is the Lapwood term. The third, fourth, and fifth terms on the right-hand side (RHS) are the Darcy, Brinkman, and Forchheimer terms, respectively. The last term, signifying quadratic form drag for laminar flows, also goes by the name of Ergun and Dupuit, and this has the distinct advantage of also taking care of the superlinear drag in the case of one-dimensional flow, unlike the Lapwood term, which fails to take care of this (see Vafai and Tien, 1981; and Georgiadis and Catton, 1988). The constant quadratic drag coefficient is independent of the viscosity and density of the fluid but depends on the geometry of the flow. In the case of turbulent flow the theory and experiments of Masuoka and Takatsu (1996), Takatsu and Masuoka (1998), and Masuoka (2000) bring out very important results on the quadratic drag. They classify the observed vortices into two types: 1. Interstitial vortices of the size of the interstitial gap 2. Pseudo vortices of the size of the particle diameter These authors opine that the interstitial vortices on mixing give rise to the Forchheimer drag and the pseudovortices give rise to thermal dispersion. It thus seems that the Forchheimer term is a drag due to mixing of interstitial vortices at high Reynolds number. Using a Reynolds number Re based on the hydraulic diameter dh and the mean (Darcian) velocity VD, their range of experimentation can be taken to be Re > 40. They examined the experimental Lyapunov exponents for the time series of velocity fluctuations and observed that it becomes positive for Re > ~60, indicating weak chaos. The Lyapunov exponent rapidly increases for Re > ~(300500), indicating the end of transition or in other words the initiation of turbulence. The advantage of the DLFB Eq. (1) is that as k , with chosen to be unity, one may recover the results of viscous flow, i.e., flow in the absence of porous media. Another advantage of the DLFB equation is that it can be used to study heat transport by turbulent convection in

4 porous media (see Rudraiah et al., 1985; and Masuoka, 1999) when the flow is not small. The books by Nield and Bejan (1999) and Vafai (2000) include good discussions of this equation. In the absence of the Brinkman term the differential equation becomes one order less, leading to an underspecified system (see Beck, 1972). In addition to making the equation properly specified, the Brinkman term takes care of the boundary effects. Katto and Masuoka (1967), Masuoka et al. (1981), and others have pointed out that for low velocities and when the microscopic dimension k becomes small compared to the macroscopic reference length h of the system, say k/h2 << 103, the long-distance Brinkman effect is generally weak. From the above observation it is obvious that the Brinkman term serves as a bridge between the viscous and Darcy regimes. One important observation can be made when we neglect the acceleration, Brinkman, and Forchheimer terms in Eq. (1), viz., = k [p g] q (2)

Rudraiah et al. tion of clear fluids. This was set right by the work of Katto and Masuoka (1967), who also showed that the classical result Ra/2 > 42 is valid for values of 2 > 103. Together with Eqs. (1) and (4), the continuity equation and the equation of state are required to study thermal convection: q = 0 = o[1 (T To)] (5) (6)

In the study of a composite medium (see Rudraiah and Wilfred, 1982; Masuoka et al., 1988; Somerton and Catton, 1982; Nishimura et al., 1986; and Masuoka et al., 1992, 1994), one has to use the usual NavierStokes equation for the clear fluid flow and the DLFB equation for flow in a porous layer with suitable boundary conditions between the fluid and porous layers. In that case a most general unified equation that accommodates different forms of momentum and energy equations in both the fluid layer and the fluid-saturated porous layer are

On comparing this equation with the HeleShaw cell model (see Masuoka et al., 1987), viz.,
2 = h [p g ] q 12

q 1 1 Xp 1 + 1 t + Xp 2 1 + 1 (q )q
q = p + g Xp( 1) + 1 2 cbq Xp q k k Xp(M 1) + 1 T + ()T q t

(3)

(7)

we can infer the analogy between the two, with an equivalent permeability k = h2 12. Since the analogy is for densely packed porous media characterized by Darcy velocity, it follows that the theoretical and experimental results with respect to low-porosity media can be extracted from the HeleShaw model. The most accepted, in-vogue single-phase heat transport equation for two-phase fluid-saturated porous media is M T q + ()T = ef2T t (4)

= Xp(K 1) + 1 2T
Here, 1 for a porous medium Xp = , 0 for a fluid

(8)

This equation was, in fact, the outcome of a series of theoretical investigations and experiments performed by Katto and Masuoka (1967) in attempts to bridge the gap between the theoretical and experimental results observed by a number of investigators, including Horton and Rogers (1945), Morrison et al. (1949), Rogers and Morrison (1950), Rogers et al. (1951), and a host of others. All these and many other works had sought to correlate the experimental and theoretical results using the heat transport equa-

ef ,

K =

ef

Note that different models governing the flow in a porous medium can be obtained from Eq. (7) by suitably assigning the values of , Cb, and , e.g., the DarcyLapwood (DL) equation can be obtained by choosing Xp = 1, Cb 0, and 0. If Cb 0 and Xp = 1 we get the

Nonlinear Convection in Porous MediaA Review DarcyLapwoodBrinkman (DLB) equation. We also note that if Xp =1 we get from Eq. (8) the energy equation used by Katto and Masuoka (1967), viz., Eq. (4). In the case of rapid transient flows the contribution to the heat capacity does not come from the entire body of the solid matrix but only from the surface. Thus for these types of flows Eq. (8) needs to be modified. In recent years there has been considerable interest in the study of different types of convection because of their interest in many practical problems discussed earlier, particularly in providing effective thermal insulation in the utilization of thermal energy, in some type of nuclear reactors, in packed beds, in high-temperature heat exchangers, in thermal energy storage, and so on (for example, see the reviews by Cheng, 1978; Combarnous, 1978; Rudraiah, 1984, 1989; Masuoka, 1989; Tien and Vafai, 1990; and Masuoka, 1999). The first type, called type 1, is concerned with natural convection in a horizontal singlecomponent (heat or mass) or two-component (heat and mass, called double-diffusive convection) in a Newtonian or non-Newtonian fluid-saturated porous medium with and without the external constraints of rotation and/or magnetic field. The convection due to buoyancy force is either in a horizontal porous layer heated from below and cooled from above called RayleighBenardDarcy (RBD) convection, analogous to the RayleighBenard (RB) convection in a layer of fluid in the absence of porous media, or in a vertical porous layer called OberbeckDarcy convection, analogous to Oberbeck convection in a fluid in the absence of porous media (see Vafai, 2000). The above convections may also be generated by surface tension force. We note that a granular porous medium whose particle size is sufficiently small, such as fine sand, will entrain a liquid which wets the particles of the medium. The liquid retention and the liquid flow in such media is dominantly affected by the surface tension characteristics of the entrained liquids (see Moore et al., 1974; Rudraiah and Prasad, 1998). Therefore many have studied either surface tension-driven convection in porous media, called MarangoniDarcy (MD) convection, analogous to Marangoni convection in clear liquid in the absence of porous media, or combined buoyancy- and surface tension-driven convection in either a single-component or a two-component liquid-saturated porous medium with or without the external constraints of rotation and/or magnetic field. The effect of Coriolis force on linear and nonlinear convection in a single-component fluid-saturated porous medium (see Rudraiah and Srimani, 1976; Friedrich and

5 Rudraiah, 1981; and more recently Vadasz, 1998) and in a two-component fluid-saturated porous medium (see Rudraiah et al., 1986; Guo and Kaloni, 1995; Kaloni and Guo, 1996; and references therein), has been investigated with the motivation of understanding geothermal energy processes. The external constraint of magnetic field on linear and nonlinear magnetoconvection in fluid-saturated porous media has also been investigated (see Rudraiah, 1984, and references therein) because of their application in geothermal energy and also to study bifurcation and chaotic phenomena. Another interesting study of a buoyant plume through a permeable porous layer located above a line heat source was carried out both experimentally and numerically by Masuoka et al. (1995). The study shows the expansion and contraction of the plume at the lower and upper interfaces of the horizontally held permeable layer and also serves to check the validity of the BeaversJoseph (1967) slip condition for the problem. In a double-diffusive convective system where the diffusivity of heat is greater than that of salt, D, two types of instabilities called "fingering" and "diffusive" instabilities are of particular interest. In a fingering instability the gradient of heat is stabilizing (i.e., cooling from below) and that of salt is destabilizing (i.e., salting from above), while the net density gradient is stable. However, in diffusive instability the gradient of heat is destabilizing and that of salt is stabilizing, while the net density gradient is stable. A physical mechanism in fingering instability is that when a small parcel of fluid is displaced downward, it finds itself in a colder and fresh environment. Since heat diffuses much faster than salt, the parcel of fluid will lose its heat quickly to the surrounding fluid while remaining saltier. The parcel of fluid then finds itself denser than its surroundings and so continues to fall. This instability is called fingering because the fluid tends to form tall, slender vertical convective cells. In diffusive instability, when a small parcel of fluid is displaced downward it gains heat from its surroundings while gaining only a little salt. It is then less warm than its surrounding and so returns to its original position. Due to the phase lag between the displacement and temperature fields it is warmer and less dense than the surrounding fluid and tends to continue rising, overshooting its position. The process may be repeated, leading to an overstable oscillatory growth of the initial disturbance, called diffusive instability. In addition to single- and double-diffusive convection in porous media, there are works on cross-diffusion (Soret

6 and Dufour effects) of substances (such as mass and heat or species, chemicals. etc.) in porous media of type 1 (see Rudraiah and Prabhamani, 1974, 1980; Rudraiah and Siddheshwar, 1998; and references therein) because of their applications in chemical engineering and geophysics, particularly in saline geothermal fields where hot brines remain beneath less saline, cooler groundwater. Here the diffusivity of heat is usually more than the diffusivity of the salt and hence a displaced parcel of fluid loses any excess heat more rapidly than any excess solute. The resulting buoyancy force may tend to increase the displacement of the particle from its original position, causing instability either of finger or of diffusive type. A theory behind it (see De Groot and Mazur, 1962; Schechter et al., 1972) is based on considering two substances, say mass denoted by concentration C and heat denoted by temperature T, that are activating or inhibiting each other according to some law of reaction and diffusion in a spatial domain. Here diffusion of one of the substances influences the other and vice versa. For example, at a time t and position let r, r the concentration be C(r, ) and temperature be T(t, ). r A substance, say C, flows from places where its density is high toward places where its density is low. At the same time T has an attracting or repelling effect on C, so that C flows toward high (or low)-temperature places of T. Accordingly, the diffusion flux, say JC, of C is a linear combination of the gradients of C and T and is given by JC = D11C D12T M where the concentration diffusion coefficient D11 > 0 and the cross-diffusion coefficient D12 0 (or 0) according as T attracts (or repels) C. Similarly the diffusion flux, say JT, of T is JT = D21C D22T where the thermal diffusion coefficient D22 > 0 and the cross-diffusion coefficient D21 0 (or 0) according as C attracts (or repels) T. Here the diffusivities D11 and D22 and the cross-diffusivities D12 and D21 may be constants or functions of C or T depending on the situation. The works on porous media mentioned above are concerned with constant diffusivities and cross-diffusivities. Further, these cross-diffusivities play a significant role depending on whether we are dealing with liquids or gases. In the case of liquid-saturated porous media the mass of the species C arises due to thermal cross-diffusion known as Soret effect (see Rudraiah and

Rudraiah et al. Prabhamani, 1974, 1980), where the Dufour effect is negligible because it is of order less than 103 oC. Similarly, in gases Soret effect is negligible and crossdiffusion through Dufour effect is significant (see De Groot and Mazur, 1962). The influence of both the molecular and cross-diffusion effects on double-diffusive convection in a porous medium has been investigated by Rudraiah and Siddheshwar (1998). In the works of Rudraiah and Prabhamani (1974, 1980) and of Rudraiah and Siddheshwar (1998), the effects of reaction due to C and T have been neglected. In the dynamics of fission reactors with temperature feedback (see Leung, 1997), it is essential to consider reaction terms such as R(C, T) and S(C, T), where R and S are the effect of interaction between C and T on the creation and dissemination of these substances. The basic momentum equation will be Eq. (1) or one of its limiting forms, and in addition we need energy and species equations to study the effect of cross-diffusion on double-diffusive convection in porous media. The energy and diffusion equations with reactiondiffusion terms can be obtained from standard procedure (see De Groot and Mazur, 1962; and Schechter et al., 1972) in the form C + (q )C = S(C, T) + D112C + D122T t T R(C T) + (q )T = + D212C + D222T (cp)f t (9)

(10)

Rudraiah and Prabhamani (1974, 1980) have used this equation with R = S = D21= 0 to study the Soret effect on linear and nonlinear double-diffusive convection in a porous medium. We note that when R = S = D12 = D21 = 0 (i.e., in the absence of reaction and cross-diffusion terms), these equations reduce to the usual equations to study double-diffusive convection in porous media discussed earlier. Recently, Rudraiah and Siddheshwar (1998) have made a linear and nonlinear analyses of double-diffusive convection with cross-diffusion in porous media. More recently, Rudraiah, Siddheshwar, and Masuoka (2002) have made a linear and nonlinear stability analyses of double-diffusive convection with/without cross-diffusion in anisotropic porous media. Brand and Steinberg (1983a, 1983b) and Steinberg and Brand (1983, 1984) were the first to investigate the effects of homogeneous first-order chemical reaction on double-

Nonlinear Convection in Porous MediaA Review diffusive convection in porous media. Rudraiah and Siddheshwar (2002) performed linear and nonlinear analyses of double-diffusive convection with fast chemical reaction in anisotropic porous media against the background of the results of isotropic porous media. The works cited above are concerned only with type-I convection, namely, a horizontal porous layer heated from below and cooled from above. Type-II is concerned with convection in a vertical porous layer either heated from one side and cooled from the other side where the temperature difference is maintained normal to gravity, or heated from below and cooled from above as in RBD convection. The natural convection in a vertical porous layer generated by maintaining the temperature difference normal to gravity (i.e., differentially heated side walls) is usually called Oberbeck convection (see Rudraiah and Nagaraj, 1977; Rudraiah et al., 1982, 1983; Lauriat and Prasad, 1987; and references therein). Heat transfer by Oberbeck convection in a vertical porous layer with differentially heated side walls has also received considerable interest because of its relevance to geothermal applications and thermal insulation (see Masuoka, 1968; Masuoka et al., 1981, 1992, 1994; Chan et al., 1970; Weber, 1975). The works of Rudraiah and Nagaraj (1977) and Rudraiah et al. (1988) are concerned with analytical studies, whereas the works of Jannot et al. (1973) and Masuoka et al. (1983) are concerned with numerical study with the motive of understanding the effects of variation of viscosity, thermal conductivity, and density with temperature on Oberbeck convection in a porous layer. The numerical results reveal that the flow patterns were considerably distorted in a horizontal direction in contrast to the symmetrical patterns predicted by analytical study when viscosity and thermal conductivity were constant. However, the overall heat transfer characteristics were not affected so much, provided that representative properties were evaluated at a mean temperature. Masuoka et al. (1981) made a significant contribution to this problem. They extended the boundary-layer analysis to account for the core vertical temperature gradient (CVTG) and the apparent wall-film thermal resistance (WFTR), caused by a local increase in porosity near the wall. Not-so-tall and tall slender cells were considered and the heat transfer characteristic was presented. In the absence of WFTR, used by most investigators, the heat transfer is overpredicted and hence the mismatch between theory and experiment. It is important to note that the nature of CVTG near the cold boundary changes drastically with aspect ratio compared to the hot wall. It also depends

7 on the combined effect of buoyancy, porous permeability, and aspect ratio. The type-III convection is concerned with forced and mixed convection in a porous layer. Forced convection is concerned with basic flow driven by an external force such as pressure gradient, through flow (injection or suction) at the boundaries, shear produced by motion of the boundaries, magnetic and electric fields, etc., and temperature field rises due to thermal convection, diffusion, and viscous dissipation. In this forced convection the momentum equation is decoupled from the energy equation. First solving the momentum equation, we can then solve the energy equation for temperature and here heat transfer is mainly by conduction. Mixed convection is concerned with combined forced and natural convection. Although natural convection (i.e., RBD convection) in porous media has received considerable attention in recent years because of its natural occurrence and relevance in many practical problems discussed earlier, much work has not been done, to our knowledge, either on forced convection, which may occur in certain types of insulation layers, or on mixed convection in porous media. In mixed convection some internal forced flow might superimpose on RBD convection in some cases, e.g., when difference in thermal expansion between liner walls and pressure boundaries is not negligible and when perforation is needed for liner walls to be adapted to a pressure transient in a sudden charging or discharging process of an internal gas. Uncertainties still remain as to the effect of the internal forced flow on thermal convection in porous insulation, particularly for inclined geometries (see Rudraiah and Wilfred, 1982; and Masuoka et al., 1987). Rudraiah et al. (1979) have studied forced convection in a layer bounded by a porous medium. Masuoka et al. (1987) have studied the stability of mixed convection in a sloping HeleShaw slot, a rectangular channel of large aspect ratio heated from below, both theoretically and experimentally. They have investigated linear stability using the Galerkin method and found that thermal convection superimposed on the basic forced flow can occur in the singular form of transverse rolls with axes oriented across the slope and also found oscillations in fluid temperatures. In particular, they found good agreement between theoretical and experimental results on temperature oscillations. Recently, Calmidi and Mahajan (2000) have studied forced convection in high-porosity metal foams, both numerically and experimentally, and excellent agreement was achieved.

8 So far we have given a brief outline of different types of convection in porous media as a background for the study of nonlinear convection in porous media using different techniques. The main object of this review is to study the merits and demerits of the following techniques: (1) Lyapnuov technique, (2) power integral technique, (3) spectral method, (4) Fourier decomposition and numerical techniques, and (5) experimental investigation. To achieve this objective, in the remaining sections of this article, we explain how the techniques listed above can be applied to study NLC in porous media. LYAPUNOV TECHNIQUE The problem of analysis and synthesis of nonlinear systems is still of interest even though there are numerous books and articles dedicated to this subject. The Lyapunov (1892) method and its various modifications have been and still remain one of the most powerful means to solve this problem. In the Lyapunov method, to study universal stability of flow of the quiescent state and for synthesis of nonlinear system control, we start with a trial function which is an appropriate positive definite function, say V(.), and continue with a verification of negative (semi-) definiteness of its total time derivative. This method, called classical Lyapunov method, is based essentially only on the sufficiency of the corresponding condition because the selection of the function V(.) is not governed by both necessary and sufficient conditions. Therefore, the stability condition obtained from this analysis will be too restrictive on the system under control. Further, it does not provide an algorithm for a Lyapunov function construction to guarantee an exact determination of the universal stability domain. In spite of this, it has been widely and effectively applied to study the stability of various complex quantitative dynamical problems. A main advantage of this technique is that it does not require knowledge of the solutions and therefore it has great power in application to many dynamical systems arising from many disciplines. To improve the stability criterion obtained from the Lyapunov method one usually applies the variation principle. Because of the fact that the method of calculus of variations and the method of Lyapunov functions are both extremely useful and effective techniques in the investigation of nonlinear stability problems, it is natural to combine these two approaches in order to gain benefits of these two important methods. These methods are used effectively in the study of stability of flow with and without porous media (see

Rudraiah et al. Carmi and Lalas, 1970; Shir and Joseph, 1968; Joseph, 1976; Rudraiah and Prabhamani, 1973, 1974; Rudraiah and Vortmeyer, 1978; Guo and Kaloni, 1995; Qin and Chadam, 1996). Most works on convection of single- and doublecomponent fluids in porous media are concerned with studying linear theory using normal mode analysis and nonlinear theory using the energy method, that is, the Lyapunov method. In this section, we explain the use of Lyapunov technique to study nonlinear convection in porous media. The Lyapunov method is also called the energy method because of the following physical consideration. It is intuitively clear that, if the total energy of a physical system has a local minimum at a certain point, then that system is stable at that point. This idea was generalized by Lyapunov (1907) into a single but powerful method for studying stability problems in a broader context. A formal discussion of this method is given by Pritchard (1968) and by Rudraiah and Vortmeyer (1978). A brief outline of this method is as follows. Let = (x, y, z) be the position vector defined over fixed r and finite domain with a boundary and t be the time. Our aim is to study the stability of the basic state of an electrically conducting fluid in a porous medium in the presence of an applied magnetic field H0, given by

( r, 0) = 0 qb

d2Tb H = H0 and = 0 dz2

For this, we suppose that a metric space X(a, a) be given such that the metric between any two states a0 and a at time t denote by B(a0, a). Let the basic state a0 and the perturbed state a, both belong to . The basic state is stable in the sense of Lyapunov if for each real number > 0 there exists a real number () such that for every a , B(a0, a) < at the initial time t0 implies that B(a0, a) < for all t > t0. The basic state is asymptotically stable if it is stable and in addition B 0 as t . The basic state is unstable if it is not stable. Among the unstable systems, we do not consider those that have finite escape time, i.e., B for finite t. A theorem discussed by Pritchard (1968) establishes asymptotic stability in the Lyapunov sense. The crucial point in the Lyapunov method is to obtain a suitable Lyapunov functional because there is no unique way of obtaining it. To obtain a suitable Lyapunov functional we consider a small disturbance of the form

Nonlinear Convection in Porous MediaA Review

9 based on the energy method (see Shir and Joseph, 1968). The method is as follows. First we subtract the basic flow equations from the altered flow equations and we multiply the resulting difference equations secularly by q , H, and T, respectively, where for simplicity primes are omitted, and integrating over the volume V and making the resulting equations dimensionless, we get

^ u = u, v = v, w = w, P = P0 + P, H = H0k + H
and 1 z T = T0 T + T (x, y, z, t) h 2 (11)

Here (u, v, w) are the components of velocity, T is the temperature difference, suffix zero denotes the basic equilibrium state, primed quantities denote perturbed quantities which are assumed to be small for linear analysis and arbitrary for nonlinear analysis compared to the basic state, and ^ denotes the unit vector in the z direction. k The basic equations are 1 h 1 1 q (H )H + 2 (q )q = P 0 0 t ^ [1 (T T0)] gk q k u v w + + = 0 y x z H + (q )H = (H )q + m2H t = 0[1 t(T T0)] T + )T = 2T (q t (12)

dK ^q = R2E + q2 + RaT(k) q q d

q 2RmH H Pm dL = d

J + 2RmA4

(18)

HEH I2RPm + 2H : H

+ 2Rmq H H Rm2A4

(19) (20)

Pr (13)

d = d

I3RaTq T1 + 2T : TJ 2INuT 2J

(14) (15) (16)

where IJ denotes the volume integral over V(). IJ denotes the surface integral over , kinetic energy K = _ modulus =

I 1 q2J, magnetic energy L = I 1 h2J, temperature 2 2 I 1 T 2J, A4 = IH H + H H , = qJ q 2


t , Tl = T T, T is made dimensionless using k

(Th2 gk) V0, magnetic field is made dimension


less using Ho P1 2 A, Ho is the reference magnetic m field. Nu = h3 is the Nusselt number, h is the characteristic length, and 3 is a piecewise continuous function such that T + 3T = 0 on n Since the surface integral tends to stabilize the flow as Nu grows larger, we neglect it. Adding Eqs. (18)(20), we get dI = I2H :H + T :T + q 2 d

The boundary conditions are W = = hz = D = DJz = 0 at z = 0 and z = h Universal Stability In this section, we discuss the stability for arbitrary perturbations, called universal stability, using the Lyapunov method. We use Eq. (11) with the observation that primed quantities are not small and that the basic and perturbed (i.e., altered) quantities satisfy the same basic equations (12)(16) satisfying the boundary conditions given by Eq. (17). The main concern here is to know whether the altered flow will approach the basic flow as t . For this we use the Lyapunov technique, which is (17)

k + RaT(q ^) + 3RaT(T1) + R2E q q q


R2PmH E H + 22RmG H q

(21)

10 where I = K + PmL + Pr is the total energy and G is the antisymmetric part of H . Using Schwarzs inequality, we get J 2K Iq E q (22) IH E H J 2L J 2n n (K + L) Iq G H 0 KL 0 where n0 = max G . We can find a > 0 (see Shir and Joseph, 1968) such that IH : H J T IT : J 2L 2 K q IT1TJ 2 K + Ra2 98 1 n0 (0.01) Rm N
2

Rudraiah et al.

(27)

and RaL, obtainable from linear theory, is Ra2 990 1 + L

2 1 + (0.01)2 M1

(28)

Thus Ra2 < Ra2 and hence subcritical instabilities are N L possible. We note that the stability criterion Eq. (26) can be further improved by using the techniques of calculus of variations (see Shir and Joseph, 1968). POWER INTEGRAL TECHNIQUE The concept of power integral was first introduced by Howard (1964) as one of the constraints in finding the maximum upper bound, using the variation method, for the heat transport by RayleighBenard convection in a clear fluid. Howard (1964) obtained two power integrals as the integral consequences of Boussinesq equations using horizontal averages and average over the layer. The first power integral was obtained from the momentum equation and the second was obtained from the energy equation using the above averages, and Howard gave simple physical interpretation for these two integrals. Later, Busse and Joseph (1972) and Gupta and Joseph (1973) extended Howards (1964) variation method and the bifurcation property of Busses (1969) multi- solution of Howards (1964) variation problem to find bounds on the strongly nonlinear heat transport across a porous layer. Palm et al. (1972) examined the dependence of Nusselt number in two-dimensional convection neglecting the inertia term in the momentum equation and using the expansion method of Kuo (1967). The above theoretical models have limited applications in that they predicted heat transfer accurately only for a restricted range of Rayleigh number Ra (<400). Most of the available literature on porous media focuses mainly on the onset of convection and the amount of heat transport, but is silent about the prediction of the preferred cell pattern, the nature of the amplitude, and the effect of the Prandtl number on the solution. Rudraiah and Srimani (1980) investigated these aspects using an iterative technique which combines the best features of Howards (1964) power integral technique, Galerkin technique, and Stuarts (1974) shape assumption (see Malkus and Veronis, 1958). In this method the linear stability theory is used to predict a condition for the onset of convection and in so doing, a shape (i.e., a mathematical formula for convection velocity) is also predicted. This shape, derived from the

(23)

Equation (21), using Eqs. (22) and (23), takes the form

_ ~ ~ dI 2N Ra(1 + 2) 2R 2Rm I d
where _ = max (1, 2) and 1 1 N = minimum of 1, Pm , Pr

(24)

~ R = max (R, Ra Pm)

~ Rm = n0Rm

Integrating Eq. (24) with equality sign, we get _ ~ ~ I = I0 exp 2N Ra(1 + 2) 2R 2Rm (25) where I is the value of I when T = 0.

For stability we must have

_ ~ ~ 2R + Rm + 1 + 2 Ra

(26)

This establishes a stability region near the origin of 12 12 2 1 2 R , m , (1 + )Ra sp ace, wh ich is a _ sphere of radius 1 2 and center at the origin for a fixed Pm. To compare with the linear theory we take = 3.742 (for channel flow, see Carmi and Lalas, 1970), R = 0 for basic stationary flow, and 2 = 0.01. In that case the critical Rayleigh number for nonlinear theory, RaN is

Nonlinear Convection in Porous MediaA Review linear analysis, persists into the nonlinear convection regime, which is Stuarts (1974) shape assumption. This method is capable of predicting: (1) the nature of the amplitude at a given value of Ra, (2) the energetics of the fluid, and (3) the physically preferred cell pattern, i.e., the one that transports maximum heat when the viscosity and permeability of the porous media are constants. In this method, the perturbations on the basic flow are assumed to be finite so that higher-order terms in the stability equation have to be retained. It relies on a system of differential equations for each of the field variables. These are accomplished by introducing parametric expansions of the motion and then equating the coefficients of each order of the parameter. From the zero-order system of differential equations, corresponding to the linear theory, the form of one of the field variables is determined. Then one uses Stuarts (1974) shape assumption, which implies that the plan form that exists at the onset of convection, i.e., at the critical Rayleigh number Rac, persists even after the onset of convection, i.e., R > Rc (nonlinear theory) to determine the remaining variables. The investigation of higher-order approximations based on the solution of the linear stability problem is called local nonlinear stability analysis. This analysis is useful to study in detail both two- and three-dimensional convection. Rudraiah and Srimani (1980) have used this method to study two- and three-dimensional nonlinear convection in porous media using the DarcyLapwood (DL) equation. They have shown that the wave number is independent of the porous parameter and replaced it by unity. This implies that permeability does not influence the cell pattern at onset. However, if we consider the DLB equation (i.e., boundary effects), then we can show that permeability influences the cell pattern at onset. Therefore, in this section, following Rudraiah and Srimani (1980), we apply the power integral method combined with local nonlinear stability analysis to study RayleighBenardDarcy (RBD) convection in porous media using the DLB equation. First, we find power integrals by applying an averaging process in which the nonlinear terms ( )q and q )T are divided into terms that are finite when averaged (q over the horizontal plane, and that therefore modify the average field, and into terms of zero average. Applying the averaging process on the energy equation (4), we get ___ ___ + WT = H = m + (W T)m (29) where

11 ___ ___ ___ 1 = 1 + WT + G ( WT) WT) ( m t m m _ ~ _ T ~ ~ ~ Here = , T = T T, T being the actual temperaz ture, H is the constant heat flux between the horizontal surfaces due to both conduction () and convection ___ (WT) and G is the___ of which vanishes when the part time variations of WT vanish. The overbar denotes the horizontal average and the subscript m denotes the layer average. These equations are made dimensionless using the transformations (x, y, z) = (x, y, z) = q h q h2 t = t gh3 T (30)

T =

In the remainder of this section all unprimed quantities are nondimensional unless otherwise stated. To study the local nonlinear stability, we assume that (u, v, w) = (u0, v0 w0) + 2(u1, v1, w1) + 3(u2, v2, w2) + ... Ra = Ra0 + Ra1 + 2Ra2 + ... where is a constant perturbation parameter and Rai (i = 0, 1, ...) are integrals of Wi and Ti. Linear stability is concerned with the solution of the first-order equation (i.e., of order ) only, where the amplitude varies with time. In other words, for the first-order solutions to be complete, must be proportional to the amplitude of the disturbance, and this amplitude must be infinitesimal. Eliminating the pressure in the DLB equation and expressing u and v in terms of w, using Eq. (31) and equating the like powers of , we get the required equations for wi (I = 0, 1, 2, ...) and Ti (i = 0, 1, 2, ...). These equations can easily be obtained and for want of space they are omitted here; the interested reader can look into the work of Rudraiah and Srimani (1980). We note that the equation for w0 is the Lapwood (1948) equation with linear constant coefficient operator. We assume that it is solved for w0 (normalized). The equations for wi (i = 1, 2, ...) are nonhomogeneous partial differential equations which will pose a problem because of the pres(31)

12 ence of resonance terms. These can be taken into account, following Malkus and Veronis (1958), as explained in the work of Rudraiah and Srimani (1980). The Rai are evaluated in such a way that they eliminate resonant inhomogeneous terms. Also, from Eq. (31) it is evident that all wi (i > 0) must be orthogonal to w0, but they need not be mutually orthogonal. Using this procedure the solutions for these inhomogeneous two- and three-dimensional equations can be obtained to determine the physically preferred cell pattern. As we are interested in local nonlinear stability analysis, the results of linear stability analysis are important. The interested reader can refer to the linear theory in Rudraiah and Srimani (1980). We note that the Rayleigh number from the linear theory for rolls is Ra0 = 2 2 + 1 2 2 ( + 1) + 2 22

Rudraiah et al.

Ra2 = =

Ra0
1(2 + 1)

(33)

(2 + 1)2 2 2 ( + 1) + 2 2

(32)

We note that since Ra2 is always positive, subcritical solutions are not possible. In other words, the steady solution with finite amplitude is stable and the bifurcation of conduction into convection is supercritical rather than twosided. The Nusselt number Nu, for the second-order approximation, is ___ 2(Ra Ra0) (WT)m (34) Nu = 1 + = 1 + Ra Ra The derivative of Nu with respect to Ra at Ra = Ra0 is given by dNu 1 2 = = dRa Ra0 22 Ra=Ra0

Rolls Here, we study nonlinear stability using the local nonlinear analysis which is pivoted on the linear theory discussed by Rudraiah and Srimani (1980). From the first-order solutions, we develop, using the method of iteration, the higher-order solutions that give Ra1 = 0. Further, vanishing of all the zero-average advection terms from the first-order solutions leads to w1 = u1 = T1 = 0. Then from the third-order equation, we can get

(35)

Thus, the slope of the NuRa relation at Rac is the same as that given by Masuoka (1968, 1972), Busse and Joseph (1972), and Buretta and Berman (1976). Similarly, we can work out Ra4 approximation, Ra6 approximation, and so on. The ratio ___ RaNu H(Ra)Ra Ra + (WT)w ____ _ (36) = = Ra0Nu0 H(Ra0)Ra0 Ra0 + (W0T0)m

Figure 1. Heat transport versus Rayleigh number for different approximations for the case Pr = 8 and 2 = 103.

Nonlinear Convection in Porous MediaA Review where Nu0 is the value of Nu at Ra = Ra0 (i.e., conduction alone). This ratio [Eq. (36)] is plotted in Fig. 1 against Ra Ra0 for Pr = 8 and 2 = 103 for various approximations Ra2, Ra4, Ra6, and Ra8 (ZANL means Zero-Average NonLinear). From this figure it is clear that heat transport decreases as the order of approximation increases. It is of interest to compare the results of the present analysis with those of Malkus and Veronis (1958), where they observed that the Ra6 approximation for heat transport is very close to the Ra2 approximation. In the present work it is not so, and the curves of Ra2 and Ra6 approximations diverge and the Ra6 (ZANL) curve is very close to the Ra2 curve in the range Ra0 Ra 2.5Ra0. One can anticipate that the Ra6 curve, like the Ra4 curve, will diverge to the right at some R greater than 2.5Ra and the Ra10 curve, like the Ra6 curve, may be more nearly parallel to that for Ra2. From this figure it is also clear that neglect of zero-average nonlinear (ZANL) terms increases the amplitude of the predicted heat transport. Rectangles, Squares, and Hexagons The power integral technique used in the previous section is now extended to three-dimensional motion with the object of determining the preferred cell pattern. Since the analysis is similar, only the results are presented and discussed. The linear solution analogous to that obtainable for rolls in the linear theory is ln1 2 u0 = 2 2 sin cos (my) cos (n1z) x mn1 v0 = 2 2 2 cos (lx) sin (my) cos (n1z) w0 = 22 cos (lx) cos (my) sin (n1z) 2 T0 = 2 N0 cos (lx) cos (my) sin (n1z)

13 infinite set of values. The ratio l/m cannot be chosen arbitrarily and has to be determined by the physics of the system. The question as to which of the infinite number of values of l/m is chosen by the fluid can be answered by considering the relative stability criterion as done by Rudraiah and Srimani (1980). By using the value of Ra2 the ___ heat transport (WT)m 2(Ra Ra0) is computed for different values of Pr, , and l/m. We find that heat transport is maximum at l/m = 0 (i.e., limiting rectangles) for Pr = 0.1. As Pr increases to 0.45 and above, the occurrence of maximum convective heat transport crosses from l/m = 0 to l/m = 1. By fixing Pr = 0.1 and varying 2 from 102 to 103, the occurrence of maximum convective heat transport changes from a limiting rectangle to a square. In particular, we note that for values of 2 (104) the convective heat transport is almost independent of Pr, a property exhibited by rolls. We also find that the limiting rectangle will convert maximum heat only when 2 102 and Pr = 0.45. In all other cases, a square cell will convert more heat. If we plot Nu versus Ra/2, we see that Nu increases with increase in l/m. The results for limiting rectangles can be obtained easily from the general rectangle and are compared with those for rolls. We find that the value of heat transport for the limiting rectangle differs markedly from that for rolls. In general, therefore, we conclude that rolls are preferred cells rather than squares and limiting rectangles, because rolls transport more heat. However, if we consider only the three-dimensional motion, square cells are preferred to limiting rectangles. We note that these conclusions are true only for uniform viscosity and for isotropic porous media. The study of hexagonal forms using the iterative procedure discussed earlier follows a similar procedure. We find that the finite-amplitude heat transport for the hexagonal plan form is smaller than that due to squares and greater than that due to limiting rectangles. Hence we conclude that square cells are the physically preferred cell patterns. In this section it is appropriate to draw attention to the experimental works of Shattuck et al. (1995, 1997) and Howle et al. (1997). They adopted the shadowgraph and magnetic resonance imaging (MRI) techniques to come to the conclusion that the findings of the theory for homogeneous media match partially with the experiments. The reason given by them is that the structure of the medium plays a vital role in deciding on the plan forms. In the case of most experiments the medium is disordered, whereas in the theory it is generally ordered. The ultimate conclusion seems to be that much needs to be done from both the

(37)

and Ra0 is given by Eq. (32). Here 2 = l2 + m2 and N0 = (2 + n2)2 2. These solutions are such that the 1 porous parameter influences the temperature distribution only through N0 and not the velocity distribution directly. To determine the first finite-amplitude results, Ra2 is computed in a way similar to that for rolls and the influence of Pr is found even in the second-order approximation itself. This Ra2, as a function of Pr, , l, and m is a triply

14 theoretical and experimental sides to relate them to each other better. SPECTRAL METHOD The power integral technique discussed in the power integral technique section studies finite-amplitude convection in a porous layer and is pivoted on the linear theory, which is a local nonlinear analysis. It is an iterative technique which combines the best features of the Galerkin method and Stuarts shape assumption, as in Malkus and Veronis (1958). This power integral technique, however, is mathematically cumbersome, and the built-in orthogonal process to take care of the resonance and secular terms considers only the even modes. In the process, some interesting results, obtained by considering cross-interactions of modes, are missed. Rudraiah and Rao (1982) have overcome this deficiency using the method of spectral analysis, following Kuo and Platzman (1965), where the flow is governed by the Darcy equation. When we use the Darcy equation to study convection in porous media, we face the difficulty of more boundary conditions than the order of the stability equation and hence an overspecified system. This is because from physical arguments there will be six boundary conditions based on no-slip condition for the fourth-order differential equation. From a physical point of view, the no-slip condition is as valid as the other two conditions, and there appears a priori no reason to reject it. A suitable statistical approach is needed to resolve this problem. If a porous medium is sparsely packed, however, the Brinkman extended DarcyLapwood equation is valid and this leads to a sixth-order differential equation with six boundary conditions for the study of convection in porous media and hence leads to a well-posed problem. However, care must be taken in using the no-slip conditions in the case of a porous medium (see Masuoka et al., 1981). Such difficulties, however, will not arise when a porous medium is bounded by free-free boundaries where slip is allowed. In this section we briefly explain the salient features of spectral analysis considering nonlinear convection in porous media using the DLB equation. We consider the same physical model, namely, RayleighBenard convection in a sparsely packed porous layer using the Boussinesq approximation. We use the total nondimensional temperature T given by T = T1 Raz + (38) and the stream function such that u = , w = z x

Rudraiah et al.

(39)

Eliminating the pressure in the DLB equation and making the resulting equation dimensionless using Eq. (30) with h replaced by h/, assuming the marginal state is valid, we get the steady-state vorticity equation (2 2) 2 + 1 1 B = Pr x (40)

where B = (, 2) (x, z) is the vorticity advection Jacobian and 1 = is the modified permeability parameter. The nondimensional steady-state energy equation obtainable from Eq. (4) is 2 + Ra = H x (41)

where H = (, ) (x, z) is the thermal advection Jacobian. We assume that the porous layer is bounded by stress-free isothermal boundaries and hence the boundary conditions are = 2 = = 0 (42)

Spectral Representation Let = ^ l,n sin (lax) sin (nz)


l=0 n=1

(43)


l=0 n=1

^ l,n cos (lax) sin (nz)

(44)

where and satisfy Eq. (42) and symmetry conditions, l and n are integers, a is the wave number of the ^ first mode (i.e., l = n = 1), and the coefficients l,n and ^ are functions of R . The representations (43) and
l,n a

(44) transform the basic equations into the spectral domain of the spectra of the linear case. Equations (43) and (44) can be written as

Nonlinear Convection in Porous MediaA Review =

15 where ,+ is the Kronecker delta. We note that L,, vanishes unless the selection rule = + (53)

S = Exp i (lax + nz)

(45)

C = Exp i(lax + nz)

(46)

holds. This means l = l + l and n = n + n

where

represents summation over all integral lattice To find the solutions of Eqs. (48) and (49), we write _ (x, z) = (z) + (x, z) (54) where
2 a _ a (z) = (x, z) dz 1 0

points in the ln-plane and is a vector with components (l, n). The orthogonal of S and C may be expressed in the form
S S ds
s

C C ds s

= ,

(47)

where S and C are the complex conjugates of S and C, S is the surface 0 x 2 a, z , and ds is the elementary area divided by the total area 42 a of the region. Equations (40) and (41), using Eqs. (45) and (46), become 2(2 + 2) la = 1 2 laRa = aH where 2 = l2a2 + n2 B =
1

is the horizontal average over a full wavelength. Then Eqs. (48) and (49) can be written as _ _ _ H (55) = 2 for l = 0 n for l 0 a 2(2 + 2) = al B 1 Pr 2 alRa = aH

a B P

(48) (49)

(56)

These expressions are used to find the modal Rayleigh number and to study the interaction of different modes. (50) Modal Rayleigh Number Eliminating in Eq. (56), we get (Ra R) = where R = 2 1 H + 2 B l al Pr (57)

(l1n2 l2n1) 2 1 2

H =
1

(l1n2 l2n1) 1 2

(51)

the pairs 1 and 2 satisfy the selection rule = 1 + 2, i.e., l = l1 + l2 and n = n1 + n2. In the spectral method the nonlinear contributions are expressed in terms of coupling coefficients L,, given by L,, = (C, C) 1 C (x, z) d (52)

4 (2 + 2) 1
a2l2

(58)

= (l, n l, n) ,+

is called the modal Rayleigh number. We note that as 1 0, R tends to that given by Kuo and Platzman (1967) for a clear fluid. R( tends to that given by Rudraiah and Rao (1982) for Darcy flow in the limit 1 . In the linear theory B = H = 0 and hence a nontrivial solution of Eq. (57) requires Ra = R.

16 The critical modal Rayleigh number, Rc and the corresponding critical wave number ac are given by Rrc = 1 2 3n 2 + 1 a 1 32n2 (59) (60)

Rudraiah et al. of interesting possibilities. The reader is referred to Rudraiah and Rao (1983) for an interesting discussion of such results. We also note that this method needs to be pursued further. Solution of the Spectral Equations The effect of nonlinear advection B and H on the onset of convection is delineated here using the solution of Eq. (57). For this we define Ra (R11 + )c (61)

a 3n2 + 2 + 1 1 a2 = c where a1 = (2 + n2) (2 + 9n2) 1 1 and a2 = (2 + n2) (2 9n2) 1 1 1 (2 + n2) + 2 a 1 2 4l

which is a deviation from the critical Rayleigh number (R11)c and measures the amplitude of the disturbance. We expand in the form = ,r r + ,r+1 r+1 + ... where r is the order of magnitude of an element being the lowest power of in the expansion. For example, the first-order element (11 can be expanded as 11 = 111 + 113 3 + ... We note that the exclusion of odd parity elements implies that the series expansion of a spectral element is in terms either of odd powers of only or even powers only, according as r, the order of the spectral element, is either odd or even. In the coefficients of type lnp are all constants, where lnp is the coefficient of p in the expansion of (i.e., ln). Expressing , H, and B in Eq. (57) as powers in , using Ra R = (Ra R11) 2 and equating the coefficients of r, we get (R R11) ,r = ,r2 + 2 1 H,r + 2 Br l al Pr (62)

These are true for any mode consistent with the selection rule explained in the spectral representation section of this article. We can easily obtain the minimal Rc and ac for the fundamental mode (1, 1) by replacing l and n in Eqs. (59) and (60) by unity. We note that when 2 0, we get the minimal Rc = 6.75 and 1 minimal a2 = 0.5 by putting l = 1 and n = 1 in Eqs. c (59) and (60), which are the known values for viscous flow in the absence of porous media given by Kuo and Platzman (1967). We note that their Rc differs from our Rc by a factor 4 because of our choice of length scale. Further, for large 1, say 2 = 105, we get from 1 Eqs. (59) and (60) with l = n = 1, the value Rc = 42 and ac = 1, which are the known values given by 1 Lapwood (1948). From these limiting values we may conclude that a transition zone from the Brinkman model to the Darcy model exists for values of 2 in the range 102 < 2 < 103. For values of 2 up to 103 we should use the usual Rayleigh number R11 whereas for values of 2 > 103, the Lapwood Rayleigh number R11/2 is more suitable. This aspect, emerging from the 1 spectral method, corroborates with the results of Rudraiah and Masuoka (1982). The variation of R with a2, for different modes , can be computed and plotted on a graph in order to understand the interaction of different modes for different 2. Such a study reveals that the fundamental mode (1, 1) may not be a self-excited mode in some regions, thus ruling out the existence of a steady solution. In certain regions there might be a double-mode steady solution corresponding to two self-excited modes. This method opens up a plethora

Some of the important coefficients of the flow spectrum are: 111 = 1 (1 + a2) 2 a2(83 + 22a2 + 3a4) 162 (a2 + 1)3 A

113 =

Nonlinear Convection in Porous MediaA Review a2 82 (a2 + 1)3A 2a3 [P1 (a2 + 1) + a2 + 5] r 3(a2 + 1)4 (5a2 + 5 + 2) A 1

17 tween the Nusselt number Nu and the Rayleigh number Ra. The Nu is the ratio of the actual heat transport rate (i.e., combined conduction and convection) to the rate at which heat would be transported by conduction alone. Figure 2 provides a comparison of heat transport. Thus _ T _ z z=0 1 T = Nu = Ra z T z=0 h = 1

133 =

224 = 244 =

a3a2 + 5 + P1 (a2 + 4) r

A(a2+1)1023+765a2+189a4+15a6+32(a4+10a2+21) 1
A = 91 + 102 + 2 (15 + 2) a2 + 3a4 1 1 The solutions obtained by the spectral method are useful to know the effect of each coefficient on different modes and the effect of Prandtl number on the coefficients. We see that in all these coefficients, A, involving 2, appear in 1 the denominator and hence the increase in 2 is to decrease the spectral elements. Therefore, the effect of is to dampen the convection. It is also clear that the effect of Pr appears only in the fourth-order spectral coefficients. In other words, 111, 113, 133 are independent of Pr whereas 224 and 244 decrease with an increase in Pr because Pr appears in the denominator. These observations are useful in studying the amount of heat transport, as explained in the next section. Heat Transport The heat transport, which is the realm of nonlinear theory, is usually expressed as a functional relation be-

_ 1 n Ra

(63)

where = (0, n) and n covers both positive and negative integers. _ To determine Nu, we expand in Eq. (63) in powers of the parameter , as we did for , in the form _ _ _ _ 02 = 022 2 + 024 4 + 0266 + ... _ _ _ 04 = 044 4 + 046 6 + ... _ where onp is a constant, being the coefficient of p in _ _ the power series expansion on. Since contain even powers of , following Rudraiah and Rao (1983) we use

=
that is,

(Ra R11) 2 = R11 R11

(64)

Figure 2. Comparison of heat transport curve with experimental results.

18 Ra = (1 + ) R11 We expand Nu as a power series in and write Nu = N0 + N2 + N42 + ... where _ N0 = 1 N2 = 4022 _ _ N4 = N2 4R11 (024 + 2044) (66) (65) d (Nu) = 2 d(RL RLc) and

Rudraiah et al.

d2 (Nu) < 0 at R = RLc d(RL RLc)2

(67)

Equation (65) is obtained from Eq. (61) by replacing Ra using Eq. (63) and expanding (1 + )1 as a power series in . This implies that Eq. (65) is valid only for < 1. _ The spectrum is determined using Eqs. (55) and (56), and some of the spectral coefficients are _ 1 022 = 2 _ a2 024 = 16A

_ a2(a4 + 10a2 + 41) 044 = 16(a2 + 1)2 A Substituting these in Eq. (66), we get N0 = 1 N4 = N2 = 2

Thus the NuRL curve must be a concave-downward curve [see also Eq. (35)]. This is in keeping with experimental observations for the range RLc < RL < 2RLc ~ 3RLc (see Masuoka, 1968, 1972; Combarnous, 1970). It may thus be inferred that for the linear relationship between Nu and RL to be true, RL must take values in excess of 2RLc ~ 3RLc. Masuoka (1972) also experimentally ratified the above theoretical findings, using glass balls and water. The results of Rudraiah and Rao (1983) indicate a concave-upward curve for the NuRL plot. This is possibly due to their taking insufficient number of terms in the series expansion. This experience of knowing the overall nature of heat transfer goes a long way in qualitatively validating the results of any study on heat transfer, be it analytical or numerical, or experimental or of simulation. FOURIER DECOMPOSITION AND NUMERICAL TECHNIQUES In the previous sections we have discussed the nonlinear convection in porous media using Lyapunov, power integral, and spectral techniques. These methods cover limited aspects of the convective flow phenomena in porous media. They are, however, unable to cover flows in porous media that exhibit time-dependent, chaotic, and turbulent behaviors. A Fourier decomposition together with numerical techniques is best suited to take care of such flows which are essentially nonlinear and time-dependent (see Friedrich and Rudraiah, 1993). In this section we explain the salient features of this method by using it to study nonlinear double-diffusive convection in a sparsely packed porous medium. Here we limit our attention to some basic aspects of the solidfluid mixture making up the porous medium, in a rotating frame, and examine the chaotic behavior of time-dependent convection. We also focus our attention on determining the flow patterns and the associated characteristic scales and how these are reflected by the flow and heat transfer characteristics. The time-dependent convective motions in horizontal porous layers in the absence of rotation have been studied by many authors (see Howard, 1964; Straus, 1974; Kvernk-

3a6 + a4 135a2 645 + 32 (a2 + 1)2 1 4A

Then Nu, in terms of different orders, is Nu(0) = N0 = 1

Nu(2) = N0 + N2
(68)

Nu(4) = N0 + N2 + N42 ...

The variation of Nu with respect to the Rayleigh number is shown in Fig. 2. The results are compared with the experimental results of Combarnous and Lefur (1969), and good agreement is found for 2 = 104. It is appropriate to mention in this section the important analytical and experimental results of Masuoka (1968, 1972), who used the Platzman (1965) method. For values of RL near RLc(42), the relation between the Nusselt number Nu and RL was found analytically by him to be RLc Nu = 1 + 2 1 RL which means

Nonlinear Convection in Porous MediaA Review old, 1979; Rudraiah and Friedrich, 1984; Kimura et al., 1986, 1987; Steen and Aidun, 1988; Graham et al., 1992; Graham and Steen, 1994; Kimura, 1998; Masuoka, 1999; Rudraiah, 2000) using Darcys law and related to the instability of the thermal boundary layers. However, it is of interest to know to what extent the effect of Pr and Coriolis force can contribute to chaotic convection in porous media at high Ra, where the inertia force and zonal velocity can act so as to maintain the flow direction by even opposing the buoyant forces and bringing about the instability of flow directions, from which chaos arises. Using unsteady Darcy law, Masuoka (1999) has studied this aspect in the absence of Coriolis force and Friedrich and Rudraiah (1981) have studied chaos in the presence of Coriolis force. We now move on to explain the analysis involving Fourier decomposition. Mathematical Formulation We consider a sparsely packed Boussinesq fluid saturated porous layer of infinite horizontal extent confined between parallel stress-free boundaries at z = 0 and z = h at which the temperatures are T0 and T0 T, respectively. The layer rotates uniformly about the z-axis with a uniform angular velocity . We denote salinity by C and it is held at C0 and C0 C at the lower and upper boundaries, respectively. We write the total temperature and salinity as z (T, C) = (T0, C0) (T, C) h + (, ) (x, y, z, t) (69)

19 where s is the solute analogue of . For simplicity of analysis, we deal with two-dimensional motion so that all the physical quantities are independent of y and we introduce stream function such that u = z

w =

(72)

Eliminating the pressure from the momentum Eq. (12) and using the scales h, h2/, h, T, and C for velocity, time, distance, temperature, and concentration, respectively, the basic equations may be written in terms of stream function and vorticity , in dimensionless form as V 1 (2 2) T1 2 a Pr z t

= J(, ) +

1 Pr x x

(73) (74)

1 2 2 1 2 Pr t ( ) V = J(, V) Ta z 2 = PrJ(, ) PrRa x t Rs 2 = PrJ(, ) Pr x t

(75) (76)

where = 2 and V is the zonal velocity (also called thermal wind component) induced by rotation. The boundary conditions are = 2 z
2

where T, C > 0. In the Boussinesq approximation the density is taken to be (70) = 0 1 t (T T0) + s (C C0) where 0 is the density at T = T0 and C = C0, t (> 0) is the thermal expansion coefficient and s is the solute analog of t. In this case, the required momentum equation is the same as Eq. (12) but with (h v) ^ q (H ) H replaced by Coriolis force 2k and also 2 adding g to its right-hand side. In addition to the energy equation given by Eq. (16), we have to consider the diffusion equation C + (q )C = s2C t (71)

= = =

V = 0 at z = 0, 1 z

(77)

From the linear stability analysis we can find the critical Rayleigh number, Rac, to be
2 12 Ta 2 2 1 + 1 + 2
2

Rac =

(78)

which represents a sufficient condition for the onset of convection, and the critical horizontal wave number is

20
2 14

Rudraiah et al. bpq = 2 (p2 2 + q2)bpq cpapq c (1 0.5po) 2 c (np mq)apm,qnbmn 4 m=0 n=1
M N

Ta c = 1 + 2

(79)

p1 q1

For a large large-amplitude analysis, we now consider the Fourier decomposition in the form =

m=1 n=1 M N

amn sin (mcx) sin (nz)

(80)

(np
m=p n=q+1

mq)amn bmp,nq

V =


m=1 n=0 M N

+ cmn sin (mcx) cos (nz) (81)

m=p+1 n=q+1 M N


M N

[p(q n) mq] amp,n bm,nq

+ (82) +

[p(q n) mq] am,nq bmp,n

m=0 n=1

bmn cos (mcx) sin (nz)

m=p n=q+1

which satisfy the stress-free and perfectly conducting boundary conditions: V 2 = 0, =0 = 0 = 0 on z = 0 1 z z2 and the following symmetry conditions: 2 = 0 = 0 V = 0 2 x x (84) (83)

m=p+1 n=q+1 p1


M N N

(np mq) amp,nq bmn

[p(n q) + mq] apm,n bm,nq

m=0 n=q+1 p1

(mq np) apm,nq bmn

m=0 n=q+1

on x = 0, n ... (n = 1, 2, ...)

We truncate the sums, for convenience of computations, and calculate systems with a maximum number of modes K = M + N not larger than 12. The set of ordinary differential equations for the amplitudes amn, bmn, cmn of the harmonic components is derived and we obtain . apq = apq + Ra Ta 2 c p bpq 2 q cpq 2 2 2 2 Pr (p c + q )


m=0 M

q1

[q(m p) + np] am,nq bmp,n

n=1 q1

(mq np) amp,qn bmn (86)

m=p+1 n=1

(0 p M, 1 q N) ... Ta . cpq = cpq + 2 q apq Pr 2 2c (1.05oq) 4


p1 q1

p1 q1 + (mq np) (p m)2 2 c 2 2 2 4(p c + q ) m=1 n=1

2 2 c

(qm pn) apm,qn cmn + ... m=1 n=0

(1 p M, 1 q N)

(85)

(0 p M, 1 q N)

(87)

Nonlinear Convection in Porous MediaA Review The nonlinear terms in Eqs. (85) and (87) have not been written completely because their contribution to the momentum flux is negligible for large Pr 2. Moreover these terms are the same as those in the case of pure viscous flow in the absence of porous media discussed by Veronis (1968). The dots on apq, bpq, and Cpq on the left-hand side of Eqs. (85)(87) represent the time derivative. In the next section, following the work of Friedrich and Rudraiah (1981), we explain the numerical procedure. Numerical Procedure Based on Different Integration Codes Because of the excessive length of Eqs. (85)(87) for K = M + N > 3, special codes have to be used to generate the right-hand sides of Eqs. (85)(87) in a form most suitable for rapid integration. The analytical solution of Eqs. (85) (87) for M = N = 2 can be obtained following the procedure of Rudraiah et al. (1986). This provides a set of initial values for five amplitudes (the remaining amplitudes are zero) for the Fourier coefficients, which is used in solving systems M > 2, N > 2. Solutions can be obtained using RungeKutta and predictorcorrector combinations of Adam (for details, see Friedrich and Rudraiah, 1981) with proper specification of relative and absolute error tolerances. A steady state of the solution can be defined through the criterion Nu(t + t) Nu(t) < 106 The total vertical heat flux is given by . Nu = 1 nbn
n=1 N

21

(88)

To know how well different K represent reality, a maximum number of modes needs to be varied. To predict chaotic solutions there are two approaches. One is through specifying a suitable number of modes. Solutions are obtained for maximum number of modes, and the results are depicted in Fig. 3. We see that for K = 6 and 8 the solutions exhibit chaotic behavior (i.e., random behavior), whereas they behaved well for K = 10 and for Ra Rac , Ta = 0 , 2 = 104 . The second approach is to choose Rayleigh numbers larger than 6Rac. Figure 4 illustrates this effect for a system of order 120 (K = 12) at Ra = 10Rac. Note that Lorenz (1963) was the first to point out that the solution of a nonlinear third-order system of ordinary differential equations is, for a certain range of parameters, unpredictable over a large time interval, regardless of the initial conditions. From Figs. 3 and 4 we conclude that if the parameter Ra (= 6Rac) is fixed, then for lower modes of K (= 6 or 8), chaos appears; and if we increase the number of modes K (= 10 and 12), chaos disappears. Similarly, if we fix the number of modes K (= 12) and increase the parameter Ra

Figure 3. Certain modes produce chaotic solution.

22

Rudraiah et al.

Figure 4. Chaotic solutions as a result of too high Rayleigh numbers.

Figure 5. Nu versus various Rayleigh (Ta 2)2 numbers.

Table
Ra 2 (Ta )
2 2

100 3.76

200 11.26

300 19.36

400 27.79

500 36.42

(= 10Rac), then chaos appears. We also note that further extensive studies are necessary to find the stable and unstable critical points of such large systems. As pointed out by Rudraiah et al. (1981), a finite Prandtl number, say of order 1 like as for water (here we take Pr = 6.8 for water), the values of porous parameter 2 in the

range 0.2 105 2 102, do not affect the heat flux and flow quantities, which enables one to discuss the effects of rotation and heating in term of modified Taylor number (Ta 2)2 and Rayleigh numbers Ra 2. The Nu is computed for various values of (Ta 2)2 and Ra 2 and the results are depicted in Fig. 5. From this figure it is clear that

Nonlinear Convection in Porous MediaA Review

23

Figure 6. Nu versus Ra 2 for different rotation rates.

Figure 7. Influence of Taylor number on horizontally averaged temperature.

Nu decreases with increasing (Ta 2)2 if Ra 2 is kept constant. Thus, rotation is able to dampen convection considerably. The corresponding Taylor numbers at which pure conduction occurs, follow from Eq. (77). The constraining effect of rotation is compared with the available experimental results in Fig. 6, for different rotation rates ranging from 0 to 100. We found that our calculations for (Ta 2)2 = 0 do not deviate much from numerical results of Combarnous (1970) and available experimental data as (Ta 2)2 increases at constant Ra 2. The horizontally averaged temperature approaches more and more the heat conduction profile as (Ta 2)2 increases at constant Ra 2 (see Fig. 7). The dotted line in

Fig. 7 indicates starting instability. Thus while Nu varies uniformly the temperature field shows oscillations and proves to be a more sensitive indicator of proximity to chaos. Finally, we note that the computer execution time to compute Nu increases with increasing Taylor number, but it seems to be independent of variation in the Rayleigh number. From this study we conclude that further extensive studies are necessary to find the stable and unstable critical points of large systems. Having a fondness for "analytical solutions," we note that a steady-state solution does exist for M = 2 and N = 2 in the system of equations (85) and (87) (see Rudraiah et

24 al., 1986). It is also possible to make a qualitative study of such truncated Lorenz systems to obtaining information on limit cycles, spirals, and other possible solutions of the autonomous system (see Rudraiah, 1981; Masuoka, 1999). Masuoka (1999) observed that such truncated models cannot induce small-scale convection, and also for large-scale scale convection to form it requires a long waiting time after the sweep of the heating and cooling surface of entrained plumes. Also, they observed that the actual heat transfer rate and the actual scale of flow will be significantly affected by truncation. In view of this, they also resorted to a numerical solution to take the action of the smaller scales into consideration and to confirm chaos to a certain degree for the low-Pr range, the behavior of which is different from those of Pr . Finite-Difference and Other Methods Masuoka (1999) solved the unsteady Darcy equation and the energy equation (4) by the finite-difference method, using a square grid system with 101 meshes in the vertical direction. The alternate direct implicit (ADI) method together with the upwind scheme was adopted with a fine time step of t = 105. Masuoka (1999) has depicted his numerical results graphically. We see the qualitative tendency of the effect of Pr on the time variations of the stream function at the midpoint of the cell for 2 = 104 and Ra = 3000. It is also seen that the apparently different oscillatory behavior can be discerned for Pr < 10 and that the dominant frequency remains almost constant for Pr > ~100, although the flow becomes chaotic at Pr = 100, 1000, and 4000, periodic at Pr = 10, chaotic at Pr = 0.7, and steady at Pr = 0.01, where the nature of chaos at low Prandtl numbers differs from that at high Prandtl numbers due to the inertia effect. Masuoka (1999) has observed that the heat transfer characteristics do not change substantially for Pr > ~ 100 but it decreases at Pr < ~ 10 due to the change in the dominant frequency as well as flow patterns, while the heat transfer characteristics at Pr = 0.01 become relatively high due to the transition of convection to the stable mode with higher wave number. Masuoka (1999) also found that the flow pattern evolution can include chaos with intermittency for convection in porous media at Ra ~ 3000, as observed in the previous sections, and in a fluid in the absence of porous media. He has also observed that the dominant frequency normalized by the modified Ra remains almost constant for the moderately high Pr range as f = constant Ra

Rudraiah et al. (89)

where the value of the constant changes with the wave number or the aspect ratio of the convective cell. Masuoka (1999) has depicted the time variations of the stream function and its power spectra for the low-Pr range graphically. He has observed local flow reversals against the action of gravity which are not observed in the limit of Pr . This suggests that the flow in the low-Pr regime can be affected by the unsteady inertia term and that flow retardation in the flow-reversal process can yield a relative decrease in the dominant frequency or its deviation from Eq. (89). Masuoka (1999) has also observed that when the flow paths of the large-scale circulation are time-dependent and oscillatory, the inertia term, which causes the time-delay in the velocity field, will bring about the interaction across the boundary between the large-scale and secondary circulations of smaller scales. Then it seems that the interference between the neighboring circulations due to the chaotic motion induces a relative decrease of the dominant frequency or its deviation from Eq. (89), particularly at high Ra. Masuoka (1999) has also observed that a porous layer of infinite horizontal extent yields the interference of convection between the neighboring cells, or the long-distance coherence. From the above discussion we conclude that these effects still need further study. Also, though the Lorenz model cannot allow small-scale motion or small-scale mixing, when the basic effect of the characteristic length of k k chaos hc becomes of the same order of as , i.e., hc = , i.e., c = 1, the chaotic behavior should be connected to the limitation of the macroscopic momentum equation where the microscopic acceleration or deceleration becomes significant. At the present juncture we also note that there are other numerical methods like such as the finite-element method (see Rathish Kumar et al., 1998) which may be used to take care of surface imperfections in the case of porous media. A good survey of this is given in by Vafai (2000). EXPERIMENTAL INVESTIGATIONS In the earlier sections, we have presented theoretical aspects of heat transfer in porous media, which is the realm of NLC. In the abstract we remarked that not all observed aspects can be covered in a theoretical study and also not all results predicted by the theory are experimentally real-

Nonlinear Convection in Porous MediaA Review izable. With the object of connecting theory and experiment, to whatever extent possible, we first put in place some aspects on porous media. Phenomena in porous media seem to be complicated for two reasons: 1. Complicated flow passages at the microscopic level, i.e., at the level of the pores and 2. Complicated flow in each microscopic passage. Contrary to the general perception of mixing being prevalent in porous media, flow visualization experiments suggest that the laminar flow streaks in porous media are steady, without any tearing or merging or mixing (see Takatsu and Masuoka, 1998). Except for the meandering of the flow, the features seem to be essentially those seen in clear fluids for laminar flow. However, in the case of high Re, mixing and turbulence can be observed. Before embarking on a review of the techniques to study heat transfer in porous media, we first appreciate the practical difficulties in performing experiments in porous media with theory as the background. In theoretical studies the porous media are generally ordered and the models for fluids saturating them are essentially "equivalent continuums" or "representative continuums" of the actual porous media, which are obtained by averaging techniques. The averaging has to be done keeping in mind the nature of the phenomena to be analyzed. An interesting analysis of the invariance of the flow solution with the order of time and volume averaging is given by Pedras and de Lemos (2001). In addition, a proper choice of characteristic scale is crucial, for a wrong choice may lead to failure in the description of the essential features. In the case of laminar flows there are less fewer problems involved in the choice of scales. However, the wall-film thermal resistance can affect the deterioration of the heat transfer characteristics, the degree of which is closely connected to the relative thickness of the thermal boundary layer (Masuoka, 1968). The Brinkman length scale is relevant to the concept of the slip-flow approximation, where the coarse grid system of numerical calculation may cause the underestimation of the heat transfer characteristics (see Masuoka et al., 1981). The large-scale inhomogeneity of permeability or thermal conductivity also may bring about the a change of convective patterns with a jump of wave number from small-scale local convection to large-scale convection, which changes the mode of heat transfer characteristics (see Masuoka et al., 1981). For a turbulent flow, further, the choice has to

25 be done with discretion, e.g., if the turbulent scale is smaller than the particle diameter, then Eq. (1) does not give a proper description of turbulence. In addition, the heat transfer characteristics by chaotic behavior in the transition region from laminar to turbulent flow are not well understood. With this background we now discuss quite briefly some recent experimental techniques, i.e., magnetic resonance imaging (MRI) techniques and shadowgraph techniques. Using the MRI technique it is possible to provide accurate imaging of the flow or patterns selected (see Shattuck et al., 1995, 1997; Ogawa et al., 2000). It is also possible to construct media with fluidsold interfaces (even if both do not share the same refractive index) which are perpendicular or parallel to some direction of light travel. Such a novel shadowgraph experiment has been reported by Howle et al. (1997) and seems to be the first non-intrusive determination of the horizontal flow patterns for porous media convection. As in Shattuck et al. (1997), Howle et al. (1997) reached a conclusion that the pattern selection criterion in the case of clear fluids does not fully apply for porous media, in contrast to what is reported in the power integral technique section of this article. The structure of a porous medium affecting the flow pattern in turn affects the heat transport. In the case of homogeneous porous media, also the classical NuR slope of 2 was not realized by their experiments. The experiments also did not predict the concave-downward NuR curve for the RaL range of 42 to 82. In a typical porous medium constructed from spherical particles, the light which enters the medium is rapidly diffused, thus making visualization by optical techniques ineffective. A common drawback of both the above methods is the difficulty involved in handling the dp << h situation. This may be the reason for mismatch between most theoretical and experimental results on heat transfer in porous media. The methods require careful experimentation to take care of near-wall inhomogeneities. Also, it seems that unsteady and turbulent flows are difficult to be handled by the MRI technique. In what follows we pay particular attention to characteristic scales in relation to boundary-layer flow, wall-film thermal resistance, turbulence, and equivalent heat capacities, with the emphasis placed on the choice of characteristic scales. This may probably bridge the gap between the experimental and the theoretical results on heat transfer in porous media, especially that in RayleighBenard convection in a porous medium.

26 The Boundary Layer Concept for Porous MediaThe Effect of Finite Dimension of Heated Length The boundary-layer thickness relative to the pore diameter or the characteristic system dimension becomes the crucial factor in describing the heat transfer characteristic. The effect of the finite dimension of various configurations is related to the development of the boundary layer along the heated length. First, the applicability of the boundarylayer concept for convection in porous media becomes the problem. The boundary-layer concept generally assumes the boundary-layer thickness to be thin enough. However, this imposes an intrinsic difficulty from the viewpoint of the continuum concept of a porous medium, for the boundary-layer thickness cannot become smaller than the characteristic dimension of the microscopic porous structure from the viewpoint of the homogeneity of a porous medium. Thus, the boundary-layer thickness is to be thin enough compared to the characteristic dimension of the heated length and to be thick enough compared to the characteristic dimension of the porous structure. The applicability of the boundary-layer concept to natural-convection heat transfer from a vertical flat plate in a porous medium was examined by Masuoka (1968), and the existence of the boundary-layer regime for porous media was confirmed. For natural convection along a flat plate in a porous medium, the velocity boundary layer and the thermal boundary layer are considered to have the same thickness because the action of the viscous shearing force is obstructed by the presence of the solid matrix so that its effect to induce flow at far distance is impeded, although, for a clear-fluid system, the flow is induced outside the thermal boundary layer due to the action of the viscous shear. Here the viscous force acts only as a local damping force proportional to velocity. Thus, the velocity induced in the thermal boundary layer is expressed as (Masuoka, 1968) u = k g0 (T T) (90)

Rudraiah et al. Thus, the introduction of the similarity variable = (y x yields a solution of the form (g(Tw T)kx) (Masuoka, 1968):
O

6.4 R k a O O 2
12

(92)

For a porous medium consisting of glass beads of diameter dp = 3 mm and of water, at the midheight (x = O/2) of the plate of length O (= 100 mm), the existence of a boundary-layer regime is possible for the range of the boundary-layer thickness of the order of ~40 mm. This confirms experimentally the boundary-layer regime for Rax (k O2) > ~150. However, a too-high RaO (k O2) is incompatible with the assumption of the macroscopic homogeneity or macroscopic continuity. The boundary-layer thickness at the mid-height is required to be, at least, thicker than the mean diameter of the packed beads, which is expressed as = dp 4.58
O

Ra k2 O
O

dp

The value of dp is 4.6 at RaO (k O) = 1000 for the above experimental configuration of O dp = 33. Thus, the boundary-layer regime was experimentally confirmed to exist for RaO (k O2) 1501000 on a vertical layer in a porous medium. Thus, based on the concept of the boundary layer, the heat transfer characteristics of the vertical plate are given by
12 hm O k = 0.887 Ra O 2 K O

(93)

which characterizes the time scale. Then, essentially, the sweep time determines the thickness of the boundary layer as
O

t
O

R k aO O 2

1 2

(91)

and this is fundamentally valid for RaO (k O2) 150 1000 (Masuoka, 1968). This result is reiterated by the numerical work of Chan et al. (1970) and Cheng and Minkowycz (1977). The boundary-layer concept can be applied not only to the heat transfer characteristics of a vertical porous layer but also to horizontal concentric cylinders, concentric spheres, and so on. In these, the developments of the boundary layers control heat transfer from both the heating and cooling surfaces. The distance, s, between the heated and cooled surfaces does not affect the convective heat

Nonlinear Convection in Porous MediaA Review transfer if the boundary layers do not interfere with each other. In this case, the characteristic heated or cooled length, O, should be incorporated into the heat transfer correlations for the configurations discussed above. The heat transfer in a vertical porous medium can be defined in terms of the height O of the vertical layer and the conduction thickness, i.e., the gap width s, as the characteristic length, as (see Masuoka, 1968):
12 hm s R k = C1 as sO K

27 porous medium (seen in Caltagirone, 1976) completely disappears on the introduction of the characteristic dimension O. This innovative idea guided by the physics reflects the developing boundary layer. Further, if we rewrite Eq. (96) in the form
1 2 hms k = 0.312 Ras K sD

(97)

(94)

where C1 = 0.313. Equation (94) essentially means that the heat transfer characteristics are fundamentally independent of the gap width s, reflecting the independent developments of the boundary layers along the both surfaces. Weber (1975) incorporated the effect of the vertical temperature stratification in the core region with C1 = 3 3. This was experimentally confirmed by Masuoka et al. (1981), who also extended Webers analysis to include the effect of wallfilm thermal resistance (to be discussed later in the section). We now discuss the work of Caltagirone (1976), who studied numerically the three-dimensional structure of oscillatory natural convection in an annular porous layer between concentric cylinders and described the heat transfer correlation in terms of a modified Rayleigh number based on the gap width, which is dependent on the ratio of the outer and inner cylinder diameters. The exact solution for the heat transfer characteristics for the boundary layer, which develops along the cylindrical surface of diameter D, in Caltagirones problem is given by (see Masuoka and Katsuhara, 1974): hmD 4 = K 5 k R aD 2 D
1 2

where D (= r0 + ri) is the mean diameter, then the nondependence of the heat transfer characteristic on the radius ratio becomes more explicit. This fundamental natural-convection heat transfer characteristics for an annular porous layer has also been experimentally confirmed by Masuoka et al., (1980). Following the above approach for the region between concentric spheres the heat transfer characteristics can be obtained in the form (see Masuoka et al. 1980):
1 2 hms k = 0.363 Ras K sD

(98)

Similarly, we may apply the boundary-layer concept to the problem of natural convection due to a constant heat-flux from the vertical plate in a porous medium. We note here that the relative thickness of the boundary layer compared to the gap width s is a problem in this case. This concept can also be extended by default to RayleighBenard problem in porous enclosures. So far we have considered three problems on heat transfer to highlight the importance of (1) the boundarylayer concept and (2) the need for proper choice of scales. In the next sub-section, we introduce another concept called the "wallfilm thermal resistance," which is important when the boundary-layer thickness is greater than the characteristic dimension of a small scale. Wall Film Thermal Resistance Darcys law allows velocity slip at the solid boundary for flow in porous media, where the thermal boundary layer develops as in slug flow with a uniform velocity profile for the case of Pr 0 (Cheng, 1978; Kaviany, 1991). In this case, the rather restrictive packing of solid materials near the bounding wall produces a variation of the porosity near the wall, different from that in regions away from the wall. Vafai and Tien (1981) incorporated and studied the nonDarcy effects of Brinkman shear and Forchheimer inertia in the momentum equation for a forced-flow heat transfer.

(95)

If we now apply the aforementioned boundary-layer concept to Caltagirones problem, as done by Masuoka et al. (1980) in terms of the mean length of the inner and outer peripheral lengths, viz., O = (r0 + ri) 2 as the characteristic length, then we obtain:
12 hms k = 0.391 Ras s O K

(96)

From Eq. (96) we see that the dependence of the heat transfer correlation on the radius ratio for the annular

28 The Brinkman shear allowed the formation of a velocity boundary layer with thickness of the order of the characteristic dimension of the pore structure. Vafai (1984) examined the effects of decrease in the viscous shear, due to the variation of the porosity in the vicinity of the boundary wall, and an increase in the flow resistance, due to the inertia, on the heat transfer characteristics. Tien and Hunt (1987) discussed the effects of thermal dispersion in addition to non-Darcy and channeling effects near the boundary wall. Nakayama et al. (1988) expressed the scale of the boundary-layer thickness in terms of both the Darcy resistance and the inertia resistance and analyzed the heat transfer. Vafai (1985) further made experimental examination of the effect of variation of porosity on the heat transfer characteristics. Georgiadis and Catton (1988), using an extended ForchheimerBrinkman model, examined the effects of the porous Prandtl number and the geometric parameters reflecting the inertia term on heat transfer at high Rayleigh number. Hunt and Tien (1988) made an analysis of the effects of change in thermal conductivity and inertia term due to the variation in porosity near the boundary. When the thermal conductivity of the solid matrix differs from that of the fluid phase, the sharp increase of the porosity in the vicinity of the bounding wall will locally decrease the effective thermal conductivity. This sharp decrease in the effective thermal conductivity in the vicinity of a wall can be modeled through the introduction of the concept of WFTR (see Kunii and Smith, 1961; Yagi et al., 1961). We now introduce the concept. In the case of natural convection in porous media, the difficulty related to the concept of macroscopic continuum is largely reduced because the boundary layer becomes thicker than the scale of pore structures. This, however, does not mean that we can neglect the WFTR effect in the experimental verification of the theoretical characteristics of heat transfer. Let Tw Twa be the apparent temperature drop due to the excess boundary thermal resistance, where Twa is the apparent wall temperature extrapolated from the temperature profile in the interior region of a porous medium. The wallfilm heat-transfer coefficient hw, which is the reciprocal of the WFTR, is defined with the wall heat flux Q as Q = hw(Tw Twa) (99)

Rudraiah et al. can be obtained as (Masuoka, 1968; Masuoka et al., 1980):


12 hm O (hmO ) k = 0.887 RaO 2 1 (hwO ) O 32

(100)

Masuoka (1968) experimentally examined the WFTR and found that the WFTR effect was substantial for natural convection in a porous medium consisting of steel beads and water (hw O ) = 10.6, while it was not naturally severe for a porous medium consisting of glass beads and water (hw O ) . Thus, we may infer that the difference in thermal conductivities between the constituent solid and fluid phases and the restrictive packing near the wall warrant the incorporation of the WFTR. It is to be mentioned here that in chaotic motions or turbulent flows in porous media which are dominated by small-scale convective motions, it is important to know how the WFTR is affected by chaos and turbulence. This is still an open problem. The majority of the reported works on the RayleighBenard convection in porous media also do not address the WFTR effect and hence there is immense scope for work in this direction. In the case of rapid unsteady thermal flows there is also a need for fundamental studies because the small-scale response in the solid phase contributes to the heat transfer characteristics and this calls for further modeling exercise. ACKNOWLEDGMENT The work was started when one of us (NR) was visiting TM at Kyusha University under an INSA JSPS Senior fellowship exchange program. This support is gratefully acknowledged. One of the authors (PGS) is grateful to the Indian National Science Academy and the Japanese Society for Promotion of Science for an Inter-Academy Exchange Fellowship to visit Japan for 3 months. The work of NR is also supported by DST under a research project No. SP/12-PC-03/2003 and he (NR) gratefully acknowledges it. REFERENCES
Bear, J., Dynamics of Fluids in Porous Media, American Elsevier, New York, 1972.

Using an approximation that the apparent wall temperature is maintained at a uniform temperature, the mean heat transfer coefficient hm for a vertical wall in a porous medium affected by the excess thermal resistance

Nonlinear Convection in Porous MediaA Review


Beavers, G. S. and Joseph, D. D., Boundary condition at a naturally permeable wall, J. Fluid Mech., vol. 30, pp. 197 207, 1967. Beck, J. L., Convection in a box of porous material saturated with liquid, Phys. Fluids, vol. 15, p. 1377, 1972. Brand, H. R. and Steinberg, V., Convective instabilities in binary mixtures in a porous medium, Physica A, vol. 119, p. 327, 1983a. Brand, H. R. and Steinberg, V., Nonlinear effects in the convective instability of a binary mixture in a porous medium near threshold, Phys. Lett., vol. 93A, no. 7, p. 333, 1983b. Buretta, R. J. and Berman, A. S., Convective heat transfer in a liquid saturated porous layer, J. Appl. Mech., vol. 43, p. 249, 1976. Busse, F. H., On Howards upper bound for heat transport in turbulent convection, J. Fluid. Mech., vol. 37, p. 457, 1969. Busse, F. H. and Joseph, D. D., Bounds for heat transport in a porous layer, J. Fluid Mech., vol. 54, pp. 521543, 1972. Calmidi, V. V. and Mahajan, R. L., Forced convection in high porosity metal foams, Trans. ASME, J. Heat Transfer, vol. 122, pp. 557565, 2000. Caltagirone, J. P., Thermoconvective instabilities in a porous medium bounded by two concentric horizontal cylinders, J. Fluid Mech., vol. 76, pp. 337362, 1976. Carmi, S. and Lalas, D. P., Universal stability of hydromagnetic flows, J. Fluid. Mech., vol. 43, no. 4, pp. 711719, 1970. Chan, B. K. C., Ivey, C. M., and Barry, J. M., Natural convection in enclosed porous media with rectangular boundaries, Trans. ASME, J. Heat Transfer, vol. 92, pp. 2127, 1970. Cheng, P., Heat transfer in geothermal systems, Adv. Heat Transfer, vol. 14, pp. 1105, 1978. Cheng, P. and Minkowycz, J. M., Free convection about a vertical flat plate embedded in a porous medium with application to heat transfer from dike, J. Geoohys. Res., vol. 82, no. 14, pp. 20402044, 1977. Combarnous, M., Etude numerique de la convection naturelle dans une couche porous horizontale, C. R. Acad. Sci., Paris, vol. 271B, pp. 357360, 1970. Combarnous, M., Natural convection in porous media and geothermal system, Proc. 6th Int. Heat Transfer Conf., vol. 6, pp. 4559, 1978. Combarnous, M. and Lefur, B., Transfert de chaleur par convection naturelle dans une couche poreuse horizontale, Compt. Rend., vol. 269, pp. 10091020, 1969. Darcy, H. P. G., Les fontaines publiques de la ville de Dijon, V. Dalmont, Paris, 1856. De Groot, S. R. and Mazur, P., Non-Equilibrium Thermodynamics, North Holland, Amsterdam, 1962. Friedrich, R. and Rudraiah, N., Numerical study of large amplitude convection in a rotating fluid saturated porous layer, Proc. 4th GAMM Conf. on Numerical Methods in Fluid Mechanics, H. Viand (ed.), pp. 117126, Fried Viewieg and Sons, Berlin, 1981.

29
Friedrich, R. and Rudraiah, N., Order and disorder in nonlinear dynamical systems, Encyclopedia of Fluid Mechanics (Supplement I), Applied Mathematics in Fluid Dynamics, chap. 20, pp. 441507, 1993, Gulf Publishing Co., Houston. Georgiadis, J. G. and Catton, I., Stochastic modeling of unidirectional fluid transport in uniform and random packed beds, Phys. Fluids, vol. 30, pp. 10171022, 1988. Graham, M. D., Muller, U., and Steen, P. H., Time periodic thermal convection in HeleShaw slits: The diagonal oscillation, Phys. Fluids A, vol. 4, pp. 23822393, 1992. Graham, M. D. and Steen, P. H., Plume formation and resonant bifurcations in porous media convection, J. Fluid Mech., vol. 272, pp. 6789, 1994. Greenkorn, R. A., Flow Phenomenon in Porous Media, Marcel Dekker, New York, 1983. Guo, J. and Kaloni, P. N., Double diffusive convection in a porous medium, nonlinear stability and the Brinkman effect, Stud. Appl. Math., vol. 94, no. 3, pp. 341358, 1995. Gupta, V. P. and Joseph, D. D., Bounds for heat transport in a porous layer, J. Fluid Mech., vol. 57, pp. 491514, 1973. Horton, C. W. and Rogers, F. T., Convective currents in a porous medium, J. Appl. Phys., vol. 16, pp. 367370, 1945. Howard, L. N., Convection of high Rayleigh number, Proc. 11th Cong. Appl. Mech., H. Gortler (ed.), pp. 11091115, 1964. Howle, L. E., Behringer, R. P., and Georgiadis, J. G., Convection and flow in porous media. Part 2. Visualization by shadowgraph, J. Fluid Mech., vol. 332, p. 247, 1997. Hunt, M. L. and Tien, C. L., Effects of thermal dispersion on forced convection in a porous media, Int. J. Heat Mass Transfer, vol. 31, pp. 301309, 1988. Jannot, M., Naudin, P., and Vinnay, S., Convection mixte en milieu poreux, Int. J. Heat Mass Transfer, vol. 16, no. 2, pp. 395410, 1973. Joseph, D. D., Stability of fluid motions I and II, Tracts in Nat. Phil., vol. 28, Springer-Verlag, Berlin, 1976. Kaloni, P. N. and Guo, J., Study of nonlinear double diffusive convection in a porous medium based upon the Brinkman Forchheimer model, J. Math. Anal. Appln., vol. 204, no. (1), p. 138, 1996. Katto, Y. and Masuoka, T., Criteria for the onset of convective flow in a fluid in a porous medium, Int. J. Heat Mass Transfer, vol. 10, pp. 297309, 1967. Kaviany, M., Principles of Heat Transfer in Porous Media, Springer-Verlag, Berlin, 1991. Kimura, S., Onset of oscillatory convection in a porous medium, in Transport Phenomena in Porous Media, D. B. Inghan and I . Pop (eds.), Pergamon Press, Oxford, U.K., 1998. Kimura, S., Sehbent, G., and Straus, J. M., Route to chaos in porous medium thermal convection, J. Fluid Mech., vol. 166, pp. 305324, 1986. Kimura, S., Sehbent, G., and Straus, J. M., Instabilities of steady, periodic and quasi-periodic modes of convection in porous

30
media, Trans. ASME, J. Heat Transfer, vol. 109, pp. 350355, 1987. Kunii, D. and Smith, J. M., Heat transfer characteristics of porous rocks: 2. Thermal conductivities of unconsolidated particles with flowing fluids, AIChE J., vol. 7, no. 1, pp. 2933, 1961. Kuo, H. L., Solution of the nonlinear equation of cellular convection and heat transport, J. Fluid. Mech., vol. 10, p. 611, 1961. Kuo, H. L. and Platzman, G. W., A normal mode nonlinear solution of the Rayleigh convection problem, Bet: Phys. Frei Atmos. ed. 33, p. 137, 1967. Kvernkold, O., On the stability of nonlinear convection in Hele Shaw cell, Int. J. Heat Mass Transfer, vol. 22, pp. 394400, 1979. Lapwood, E. R., Convection of a fluid in a porous medium, Proc. Camb. Phil. Soc., vol. 44, p. 508, 1948. Lauriat, G. and Prasad, V., Natural convection in a vertical porous cavity: A numerical study for Brinkman extended Darcy formulation, Trans. ASME, J. Heat Transfer, vol. 109, no. 3, pp. 688696, 1987. Leung, A. W., Reaction-diffusion systems with temperature feedback: Bifurcations and stability, Nonlinear Analy., Meth. and Appln., vol. 30, no. 6, pp. 33793390, 1997. Lorenz, E. N., Deterministic non-periodic flow, J. Atmos. Sci., vol. 20, pp. 130141, 1963. Lyapunov, A. M., The general problem of the stability of motion, French translation: Problem general de la stabilite der mouvement. Aun. Fac. Toulouse, vol. 9, pp. 203474, (1907); English translation: Int. J. Control, vol. 55, pp. 531773, Taylor and Francis, London, 1992. Malkus, W. V. R. and Veronis., G., Finite amplitude cellular convection, J. Fluid Mech., vol. 4, p. 225, 1958. Masuoka, T., Natural convection heat transfer from a vertical wall to a porous medium, Trans. Jpn. Soc. Mech. Eng., vol. 34, no. 259, pp. 491500, 1968. Masuoka T., Natural convection and heat transfer in a porous medium, Ph.D. thesis, Univ. of Tokyo, Tokyo, Japan, 1968. Masuoka, T., Heat transfer by free convection in a porous layer heated from below, Heat Transfer Japanese Research, vol. 1, no. 1, pp. 3945, 1972. Masuoka, T., Heat transfer in porous media, J. Heat Transfer Soc. of Japan, vol. 28, no. 108, pp. 147161, 1989. Masuoka, T., Some aspects of fluid flow and heat transfer in porous media, Proc. 5th ASME/JSME Joint Thermal Eng. Conf., AJTE 99, p. 6394, 1999. Masuoka T., Chaotic behaviors and transition to turbulence in porous media, Proc. Int. Conf. Heat Transfer and Transport Phenomena in Microscale, Banff, Canada, pp. 299303, 2000. Masuoka, T., Ishizaka, K., and Katsuhara, T., Heat transfer by natural convection in porous media between two concentric spheres, Natural Convection in Enclosures, ASME HTD-vol. 8, ASME, pp. 115120, 1980.

Rudraiah et al.
Masuoka, T. and Katsuhara, T., An experimental study of natural convention in an annular insulating layer, Bull. Kyushu Inst. of Technol., vol. 29, pp. 3742, 1974. Masuoka, T., Kawamura, H., and Okamoto, Y., On velocity slip problems of natural convection in porous media, Trans. Jpn. Soc. Mech. Eng., ser. B., vol. 47, no. 423, pp. 21572160, 1981. Masuoka T, Nakashima, H., and Tsurutu, T., Stability of mixed convection in HeleShaw slot, ASME-JSME Thermal Eng. Joint Conf., Honolulu, HI, pp. 3540, 1987. Masuoka, T., Sakamoto, N., and Katsuhara, T., Natural convection in porous media between two concentric cylinders, Trans. Jpn. Soc. Mech. Eng., vol. 46, no. 405, pp. 919926, 1980. Masuoka, T., Shibata, K., Nakamura, H., Tanaka, T., and Tsuruta, T., Natural convection in porous insulation layer with peripheral gaps, Heat Transfer on the Second International Symposium, Tsinghua University, Beijing, China, vol. 1, August 911, 1988. Masuoka, T., Shimomura, H., and Okamoto, Y., Effects of variable physical properties on free convection in a porous insulating layer, Bull. Jpn. Soc. Mech. Eng., vol. 26, no. 212, pp. 247253, 1983. Masuoka, T. and Takatsu, Y., Turbulence model flow through porous media, Int. J. Heat Mass Transfer, vol. 39, no. 13, pp. 28032804, 1996. Masuoka, T., Takatsu, Y., Kawamoto, S., Koshino, H., and Tsuruta, T., Buoyant plume through a permeable porous layer located above a line heat source in an infinite fluid space, Jpn. Soc. Mech. Eng. Int. J., ser. B, vol. 38, no. 1, pp. 7985, 1995. Masuoka, T., Takatsu, Y., and Tsuruta, T., Buoyancy-driven channnelling flow in a vertical porous layer, Proc. 2nd JSME KSME Thermal Engineering Conf., October 19-21, pp. 321, 32133.218, 1992. Masuoka, T., Takatsu, Y., Tsuruta, T., and Nakamura, H., Buoyancy-driven channeling flow for vertical porous layer, Jpn. Soc. Mech. Eng. Int. J., ser. B, vol. 37, no. 4, pp. 4752, 1994. Masuoka, T., Yokote, Y., and Katsuhara, T., Heat transfer by natural convection in a vertical porous layer, Bull. Jpn. Soc. Mech. Eng., vol. 24, no. 192, pp. 9951001, 1981. Moore, N. R., Deaver, F. K., and Wolf, H., Thermally induced transport phenomena in partially saturated porous media, 5th Int. Heat Transfer Conf., Tokyo, Japan, CT 3.6, vol. 5, pp. 103107, 1974. Morrison, H. L., Rogers, F. T., Jr., and Horton, C. W., Convective currents in porous media, II: Observations of conditions at the onset of convection, J. Appl. Phys., vol. 20, pp. 10271029, 1949. Muskat, M., The Flow of Homogeneous Fluids through Porous Media, McGraw-Hill, Book Company Inc., New York and London, 1937. Nakayama, A., Koyama, H., and Kuwahara, F., An analysis on forced convection in a channel filled with a BrinkmanDarcy

Nonlinear Convection in Porous MediaA Review


porous medium: Exact and approximate solutions, Warme und Stoffubert., vol. 23, pp. 291295, 1988. Nield, D. A. and Bejan, A., Convection in Porous Media, Springer-Verlag, New York, 1999. Nishimura, T., Takumi, T., Shiraishe, M., Kawamura, Y., and Ozoe, H., Numerical analysis of natural convection in a rectangular enclosure horizontally divided into fluid and porous regions, Int. J. Heat Mass Transfer, vol. 29, no. 6, pp. 889 898, 1986. Ogawa, K., Yokouchi, Y., and Hirai, S., Correlation between interstitial flow and pore structure in packed bed (1st report, Axial velocity measurement using MRI and visualization of axial channel flow), Trans. Jpn. Soc. Mech. Eng., ser. B, vol. 66, no. 642, pp. 445452, 2000. Palm, E., Weber, J. E., and Kvernkold, O., On steady convection in a porous medium, J. Fluid. Mech., vol. 54, p. 153, 1972. Pedras, M. H. J. and de Lemos, M. J. S., Macroscopic turbulence modeling for incompressible flow through undeformable porous media, Int. J. Heat Mass Transfer, vol. 44, p. 1081, 2001. Platzman, G. W., The spectral dynamics of laminar convection, J. Fluid Mech., vol. 23, no. 3, p. 481, 1965. Prasad, V., Convective flow interaction and heat transfer between fluid and porous layer, in Convective Heat and Mass Transfer in Porous Media, S. Kakac et al. (eds.), pp. 563615, 1991, Kluwer Academic Publishers, London. Pritchard, A. J., A study of two of the classical problems of hydrodynamic stability by the Lyapunov method, J. Inst. Math. Appl., vol. 4, p. 78, 1968. Qin, Y. and Chadam, J., Nonlinear convective stability in a porous medium with temperature dependent viscosity and inertia drag, Stud. Appl. Math., vol. 96, pp. 273288, 1996. Rathish Kumar, B. V., Murthy, P. V. S. N., and Singh, P., Free convection heat transfer from an isothermal wavy surface in a porous enclosure, Int. J. Numer. Meth. Fluids, vol. 28, pp. 633661, 1998. Richardson, S. J., A model for the boundary condition of a porous material, Part II, J. Fluid Mech., vol. 49, pp. 327336, 1971. Rogers, F. T., Jr. and Morrison, H. L., Convection currents in porous media. Extended theory of the critical gradient, J. Appl. Phys., vol. 21, pp. 11771180, 1950. Rogers, F. T., Jr., Schilberg, L. E., and Morrison, H. L., Convection currents in porous media, IV. Remarks on the theory, J. Appl. Phys., vol. 22, pp. 14761479, 1951. Rudraiah, N., Turbulent convection in porous media using spectral method, Proc. Second Asian Cong. of Fluid Mech., Seicha Press, Beijing, China, pp. 10151020, 1983. Rudraiah, N., Linear and nonlinear magnetoconvection in a porous medium, Proc. Indiand. Acad. Sci. (Math. Sci.), vol. 93, nos. 2 and 3, pp. 117135, 1984. Rudraiah, N., Coupled parallel flow in a channel and a bounding porous medium of finite thickness, J. Fluid Eng., vol. 107, pp. 322329, 1985.

31
Rudraiah, N., Turbulent convection in porous media with nonDarcy effects, Proc. ASME Int. Symp. on Convection in Porous Media Non-Darcy Effects, ASME HTD-96, vol. 1, pp. 747754, 1988. Rudraiah, N., Effects of inertia and variation of viscosity on convective instability in porous media with through flow, 4th Asian Cong. Fluid Mech., Hong Kong, vol. I, pp. H33H36, 1989. Rudraiah, N., Nonlinear flow through in porous media, in Nonlinear Phenomena, S.K. Malik et al. (eds.), INSA Publications, pp. 601681, 2000, INSA Publications, New Delhi, India. Rudraiah, N., Chandrasekhara, B. C., Veerabhadraiah, R., and Nagaraj, S. T., Some Flow Problems in Porous Media, Bangalore University, PGSAM Bull. 2, 1979. Rudraiah, N. and Friedrich, R., Periodic and chaotic behaviour of highly truncated spectral method for cellular convection in fluids and porous media, Proc. Int. Cong. on Theoretical and Applied Mechanics, Lyngby, Denmark, August, 1984. Rudraiah, N., and Masuoka, T., Asymptotic analysis of natural convection through horizontal porous layer, Int. J. Eng. Sci., vol. 20, pp. 827839, 1982. Rudraiah, N., Masuoka, T., and Malashetty, M. S., Heat transfer by natural convection in a vertical porous layer, Int. J. Heat Mass Transfer, vol. 10, pp. 5976, 1983. Rudraiah, N. and Nagaraj, S. T., Natural convection through vertical porous stratum, Int. J. Eng. Sci., vol. 15, pp. 589600, 1977. Rudraiah, N. and Prabhamani, R. P., Stability of hydromagnetic thermoconvective flow through a porous medium, J. Appl. Mech., (ASME) ser. E, vol. 10, pp. 879884, 1973. Rudraiah, N. and Prabhamani, R. P, Thermal Convection of a conducting fluid in a porous medium, Israel J. Technol., vol. 12, pp. 8998, 1974. Rudraiah, N. and Prabhamani, P. R., Linear convective stability and thermal diffusion of a horizontal quiescent layer of twocomponent fluid in a porous medium, Int. J. Eng. Sci., vol. 18, pp. 10551059, 1980. Rudraiah, N. and Prasad, V., Effects of Brinkman boundary layer on the onset of Marangoni convection in a fluid-saturated porous layer, Acta Mechanica, vol. 127, pp. 234246, 1998. Rudraiah, N. and Rao, S. B., Nonlinear cellular convection and heat transfer in the porous medium, Appl. Sci. Res., vol. 39, pp. 2743, 1982. Rudraiah, N. and Rao, S. B., Study of nonlinear convection in a sparsely packed porous medium using spectral analysis, Appl. Sci. Res., vol. 40, pp. 223245, 1983. Rudraiah, N., Shivakumara, I. S., and Friedrich, R., The effect of rotation on linear and nonlinear double diffusive convection in a sparsely packed porous medium, Int. J. Heat Mass Transfer, vol. 29, no. 9, pp. 13011317, 1986. Rudraiah, N. and Siddheshwar, P. G., A weak nonlinear stability analysis of double diffusive convection with cross-diffusion

32
in a porous medium, Heat and Mass Transfer, Springer-Verlag, vol. 33, pp. 287293, 1998. Rudraiah, N. and Siddheshwar, P. G., Double-diffusive convection with fast chemical reaction in an anisotropic porous medium, J. Porous Media (Submitted), 2003. Rudraiah, N., Siddheshwar, P. G., and Masuoka, T., Turbulent convection in a high porosity medium with convective acceleration, J. Math. Phys. Sci., vol. 19, no. 2, pp. 93117, 1985. Rudraiah, N., Siddheshwar, P. G., and Masuoka, T., Turbulent convection in a densely-packed fluid-saturated porous medium, 3 ACFM, Tokyo, Japan, pp. 402405, 1986. Rudraiah, N., Siddheshwar, P. G., and Masuoka, T., Linear and nonlinear analysis of double-diffusive convection with/without cross-diffusion in a fluid-saturated and isotropic porous medium, Int. J. Heat Mass Transfer (Submitted),2003. Rudraiah, N. and Srimani, P. K., Thermal convection of a rotating fluid through a porous medium, Vignana Bharathi, Bangalore Univ. J., vol. 2, p. 11, 1976. Rudraiah, N. and Srimani, P. K., Finite-amplitude cellular convection in a fluid-saturated porous layer, Proc. Roy. Soc. Lond., vol. A 373, p. 199, 1980. Rudraiah, N., Srimani, P. K., and Friedrich, R., Finite-amplitude convection in two-component fluid-saturated porous layer, Int. J. Heat Mass Transfer, vol. 25, no. 5, pp. 715722, 1982. Rudraiah, N., Venkatachalappa, M., and Malashetty, M. S., Oberbeck convection through vertical porous stratum, Proc. Ind. Acad. Sci. (Math. Sci.), vol. 91, pp. 1737, 1982. Rudraiah, N. and Vortmeyer, D., Stability of finite amplitude and overstable convection of a conducting fluid through fixed .. porous bed, Warme und Stoffubertr., vol. 11, pp. 241254, 1978. Rudraiah, N. and Wilfred, V., Natural convection past inclined porous layer, J. Appl. Mech. (ASME) vol. 49, pp. 266272, 1982. Saffman, P. G., On the boundary condition at the surface of a porous material, MIT,. Stud. Appl. Math., vol. 510, pp. 93 101, 1971. Schechter, R. S., Prigogine, I., and Hamm, J. R., Thermal diffusion and convective stability, Phys. Fluids, vol. 15, no. 3, p. 379, 1972. Scheidegger, A. E., The Physics of Flow through Porous Media, rev. ed., Macmillan, New York, (1960). Shattuck, M. D., Behringer, R. P., Johnson, G. A., and Georgiadis, J. G., Onset and stability of convection in porous media: Visualization by magnetic resonance imaging, Phys. Rev. Lett., vol. 75, no. 10, p. 1934, 1995. Shattuck, M. D., Behringer, R. P., Johnson, G. A., and Georgiadis, .G., Convection and flow in porous media. Part 1. Visualization by magnetic resonance imaging, J. Fluid Mech., vol. 332, pp. 215245, 1997. Shir, C. C. and Joseph, D. D., Convective instability in a temperature and concentration field, Arch. Rational Mech. Anal., vols. 3031, pp. 3880, 1968.

Rudraiah et al.
Somerton, C. W. and Cotton, I., On the thermal stability of superimposed porous and fluid layer, Trans. ASME, J. Heat Transfer, vol. 104, pp. 160165, 1982. Steen, P. H. and Aidun, C. K., Time-periodic convection in porous media, J. Fluid Mech., vol. 196, pp. 263290, 1988. Steinberg, V. and Brand, H. R., Convective instabilities of binary mixtures with fast chemical reaction in a porous medium, J. Chem. Phys., vol. 78, no. 5, p. 2655, 1983. Steinberg, V. and Brand, H. R., Amplitude equations for the onset of convection in a reactive mixture in a porous medium, J. Chem. Phys., vol. 80, no. 1, p. 431, 1984. Straus, J. M., Large amplitude convection in porous media, J. Fluid Mech., vol. 64, p. 51, 1974. Stuart, J. T., On the cellular patterns in thermal convection, J. Fluid Mech., vol. 33, p. 481, 1974. Takatsu, Y. and Masuoka, T., Turbulent phenomena in porous media, J. Porous Media, vol. 3, p. 243, 1998. Takatsu, Y., Masuoka, T., Hashimoto, Y., and Yokota, K., Natural convection along a vertical porous surface, Proc. 3rd KSME-JSME Thermal Eng. Conf., Oct. 20-23, Kyongju, Korea, pp. II-457II-462, 1996. Taylor, G. I., A model for boundary condition of a porous material, J. Fluid Mech., vol. 49, pp. 319326, 1971. Tien, C. L. and Hunt, M. L., Boundary-layer flow and heat transfer in porous beds, Chem. Eng. Proces., vol. 21, pp. 5363, 1987. Tien, C. L. and Vafai, K., Convective and radioactive heat transfer in porous media, Adv. Appl. Mech., vol. 27, pp. 225281, 1990. Vadasz, P., Coriolis effect on gravity driven convection in a rotating porous layer heated from below, J. Fluid Mech., vol. 376, pp. 351375, 1998. Vafai, K., Handbook of Porous Media, Marcel Dekker, New York, 2000. Vafai, K., Convective flow and heat transfer in variable porosity media, J. Fluid. Mech., vol. 147, pp. 233259, 1984. Vafai, K., Alkire, R. L., and Tien, C. L., An experimental investigation of heat transfer in variable porosity media, Trans. ASME, J. Heat Transfer, vol. 105, pp. 642647, 1985. Vafai, K. and Tien, C. L., Boundary and inertial effects on flow and heat transfer porous media, Int. J. Heat Mass Transfer, vol. 24, pp. 195203, 1981. Veronis, G., Large amplitude Benard convection in a rotating fluid, J. Fluid Mech., vol. 31, pp. 113139, 1968. Weber, J. E., The boundary-layer regime for convection in a vertical porous layer, Int. J. Heat Mass Transfer, vol. 18, pp. 569573, 1975. Wooding, R. A. and White, I., Convective flows in porous media, Proc. of Seminar Organized by DSIR and CSIRO, Institute of Physical Sciences, Wairakei, New Zeland, 34 March, 1984. Yagi, S., Kunii, D., and Wakao, N., Radically effective thermal conductivities in packed beds, Int. Dev. Heat Transfer, vol. 13, pp. 742750, 1961.

Journal of Porous Media 6(1), 3349 (2003)

0DVV 7UDQVIHU -XPS &RQGLWLRQ DW WKH %RXQGDU\ EHWZHHQ D 3RURXV 0HGLXP DQG D +RPRJHQHRXV )OXLG
J. J. Valencia-Lopez,1,2 G. Espinosa-Paredes,1 and J. A. Ochoa-Tapia1*
Divisin de Ciencias Bsicas e Ingeniera, Universidad Autnoma, Metropolitana-Iztapalapa, Apartado Postal 55-534, 09340 Mxico D.F., Mxico and 2 Instituto Mexicano del Petrleo, Eje Central Lazaro Crdenas 152, Mxico D.F. 07730 Mxico * E-mail: jaot@xanum.uam.mx ABSTRACT The mass transfer condition that applies at the boundary between a porous medium (the region) and a homogeneous fluid (the region) is developed as a jump condition based on the nonlocal form of the volume-averaged mass transport equation that is valid within the boundary region. Outside the boundary region this nonlocal form reduces to the classic volume-averaged transport equation in the region and to the point equation in the region, respectively. The jump condition takes the form of a surface transport equation that contains terms representing the excess surface accumulation, convection, diffusion, adsorption, and a nonequilibrium source term, in addition to a term representing the exchange with the surrounding regions. When the excess surface terms are represented using dimensionless adjustable parameters, the jump condition can be expressed as follows: n (D sCAt DsCAt) = avKeq where is the only adjustable parameter. sCAt
1

33 Received July 10, 2001; Accepted January 7, 2002 Copyright 2003 Begell House, Inc.

34

Valencia-Lpez et al.

Mass Transfer Jump Condition at the Boundary INTRODUCTION The process of mass transport at the boundary between a porous medium and a homogeneous fluid occurs in a wide variety of technological applications and has therefore been the object of a great deal of study. As examples, we note adsorption columns (Aris, 1959), and the phenomena of convective dispersion that result as a consequence of the interface mass transfer in a cylindrical cavity (Lenhoff and Lightfoot, 1984), which can include first-order reactions (Lenhoff and Lightfoot, 1986). The mass transfer boundary condition is also important in the modeling of chromatographic separation. Moreover, different separation schemes are still being evaluated (Lee et al., 1990; Huang et al., 1988). An extension of chromatographic separations can be found in bed reactors packed with radial flow. These are used in a great variety of industrial applications such as methanol synthesis, catalyst regeneration, desulfurization in the steam phase (Balakotaiah and Luss, 1981), among others. In this situation the mass boundary condition may also be crucial. The boundary conditions developed by Ochoa-Tapia and Whitaker (1995a, 1995b, 1997) were utilized by Kuznetsov (1996, 1997, 1998), where solutions were obtained for different channels partly filled with a porous material. These boundary conditions have recently been applied to obtain an exact boundary-layer solution for velocity and temperature distributions by Kuznetsov (1999). Alazmi and Vafai (2001) analyzed different types of interfacial conditions between a porous medium and a fluid layer. The main objective of their study was to assess the differences between different models and to examine the effect of using them on heat transfer and fluid flow in the interface region. These authors showed that, in general, the differences have a more pronounced effect on the velocity field and a substantially smaller effect on the temperature field and even a smaller effect on the Nusselt number distribution. In a previous study of systems with a boundary between a porous medium and a homogeneous fluid, flux jump conditions were developed based on the nonlocal form of the volume-averaged transport equations for momentum transfer (Ochoa-Tapia and Whitaker, 1995a, 1995b, 1998a) and for heat transfer (Ochoa-Tapia and Whitaker, 1997, 1998b). The momentum transfer condition that applies at the boundary between a porous medium and a homogeneous

35 fluid was developed by Ochoa-Tapia and Whitaker (1995a), and their theoretical results were compared with experimental data (Ochoa-Tapia and Whitaker, 1995b). These authors constructed the jump condition to join Darcys law with the Brinkman correction to Stokess equation, and this leads to a volume-averaged velocity field that is continuous. Ochoa-Tapia and Whitaker (1998b) developed the jump condition when local thermal equilibrium is imposed and found that the nonlocal form simplifies to the classic oneequation model for thermal energy transport. This jump condition contains terms representing the accumulation, conduction, and convection of excess surface thermal energy, in addition to an excess nonequilibrium thermal source that results from the failure of local thermal equilibrium in the boundary region. These studies are the starting point to obtain the mass transfer condition that applies at the boundary between a porous medium and a homogeneous fluid, which is the purpose of the present work. Point Equations The system under consideration is illustrated in Fig. 1, where we have identified the porous medium as the region and the homogeneous fluid as the region. The region consists of the phase. Also shown in Fig. 1 is a sample of the region that is made up of the solid phase ( phase) and the fluid phase ( phase). The phase is flowing in both the and the regions. The governing point equation, boundary, and initial conditions that describe the mass transfer of species A in this system are given by CA t

+ (CAv) = (DCA) in the phase


CAs at the interface t at Ae

(1) (2) (3) (4)

B.C.1 nDCA = B.C.2 CA = F(r, t) I.C.

CA = G(r) t = 0

where n is the unit normal vector directed from the phase toward the phase, A represents the area of entrances and exits of the phase associated with the region. The boundary conditions at the fluidsolid interface given by Eq. (2) is a special form of the general

36

Valencia-Lpez et al.

Figure 1. Flow in a system composed of a porous medium ( region) and a homogeneous fluid ( region). The fluid phase is indicated by phase, and phase indicates the impermeable solid phase.

jump condition (Whitaker, 1992). In this work, we have restricted the analysis to dilute solutions of species A, and this means that the mole fraction of species A should be less than 0.1. The solution of Eqs. (1) and (2) requires additional information to relate the surface concentration, CAs, to the bulk concentration, CA. The simplest interfacial flux model is given by Flux model nDCA = k1CA k1CAs (5)

in which Keq, the equilibrium coefficient is defined by Keq = k1 k1 (8)

at the interface

and this can only be valid when the surface concentration is small enough so that it has no influence on the forward rate of adsorption represented by k1CA. When the characteristic process time, t*, is constrained by (Whitaker, 1986b, 1999) k1t >> 1 (6)

The physical statement of the process under consideration is relatively simple; however, the analysis becomes difficult because of the complexity associated with the position of the fluidsolid interface in the region. The boundary-value problem given by Eqs. (1) and (4) is similar in form to previous diffusion problems that have been studied using the method of volume averaging (Carbonell and Whitaker, 1984; Whitaker, 1986a, 1986b, 1999). In addition to these equations for the mass transport, we will also draw upon results associated with the continuity equation, v = 0 and the NavierStokes equations, v + vv = p + g + 2v t in the phase (10) The analysis of Eqs. (9) and (10) is given by OchoaTapia and Whitaker (1995a, 1995b) and Whitaker (1996); (9)

then the condition of local adsorption equilibrium is valid. Under these circumstances, Eqs. (2) and (5) can be combined to yield CA B.C.1 nDCA = Keq t at the interface (7)

Mass Transfer Jump Condition at the Boundary thus, a framework for the fluid mechanics problem is available to support the analysis of the mass transfer processes. In order to describe the process of mass transfer illustrated in Fig. 1, we need the volume-averaged form of Eq. (1) in the region . The development of this mathematical model for this system is relatively straightforward when classic length-scale constraints are satisfied (Zanotti and Carbonell, 1984; Carbonell and Whitaker, 1984); however, difficulties arise in the neighborhood of the boundary, where there are rapid changes in the porosity and the length-scale constraints fail. At the boundary between the and the regions we are also confronted with the mismatch of scales, i.e., the region is characterized by a volume-averaged mass transfer, while the region is characterized by a point mass transfer. When the volume-averaged equations in the homogeneous region are equivalent to the point equations we can use standard techniques to develop the jump condition (Whitaker, 1992). We address both of these problems by deriving a generalized mass transport equation that is devoid of length-scale constraints and that is valid in both regions, and in particular, in the boundary region. Given a single transport equation that is valid in both regions, we can follow the methods given by Slattery (1990) and Whitaker (1992) to develop a jump condition for the flux at the boundary between the two regions. Volume Averaging

37

We define the superficial volume average of some function, , associated with the phase according to
s

tx =

1 (x + y)dVy V V(x)

(11)

where V(x) is the volume of the phase contained within the averaging volume illustrated in Fig. 2. In this figure we have indicated that x represents the position vector locating the centroid of the averaging volume, while y represents the position vector locating points in the phase relative to the centroid. In Eq. (11) we used dVy to indicate that the integration is carried out with respect to the components of y, and the nomenclature in Eq. (11) clearly indicates that volumeaveraged quantities are associated with the centroid. In order to simplify the notation, we will avoid the precise nomenclature used in Eq. (11) and represent the superficial average of as
s

t =

1 dV V V

(12)

while the intrinsic average is expressed in the form


s

t =

1 dV V V

(13)

Figure 2. Position vectors x, y, and r associated with the averaging volume.

38 Both of these averages will be used in the theoretical development of this article, and they are related by
s

Valencia-Lpez et al. Fig. 3. Following the nomenclature given by Eq. (12), we express the superficial average of Eq. (1) as

t = st

(14)

CA t

+ s(vCA)t = s(DCA)t

(17)

The porosity is defined explicitly as V(x) (x) = V (15)

The volume V is independent of time. This allows us to interchange integration and differentiation to express the first term on the left-hand side in this result as follows:

and we note that in the boundary undergoes significant changes over a distance equal to the radius of the averaging volume, r0. In addition to the definitions given by Eqs. (12) and (15), we will make use of the averaging theorem (Howes and Whitaker, 1985) that can be expressed as
s

sC t CA J = tA t

(18)

t = st +

1 n dA V A

(16)

Use of the relationship between the superficial and intrinsic averages given by Eq. (14), and because is independent of time, leads to the final form for the local volume average of the accumulation term: sCAt t = sCAt t (19)

Generalized Mass Transfer Equation In this study we need a volume-averaged mass transfer equation for species A that is valid everywhere in this system and, in the boundary (Fig. 1). We begin this process by locating an averaging volume at every point in space, and three of these averaging volumes are shown in In order to interchange integration and differentiation in the second term on left-hand side of Eq. (17), we make use of the volume-averaging theorem given by Eq. (16) to express the superficial average of the convective flux in the following form:

Figure 3. Averaging volume located in the homogeneous fluid bulk, the interregion, and the porous medium bulk.

Mass Transfer Jump Condition at the Boundary 1 n (v C ) dA V AA A (20)

39 gions. To avoid imposing length-scale constrains, we define the excess dispersive flux, which is given by Ochoa-Tapia and Whitaker (1997, 1998a, 1998b):
s

(CAv)t = svCAt +

Since the interface is impermeable, this result reduces to


s

(CAv)t = sCAvt

(21)

~ v vCAtex = svCAt svtsCAt s~CAt (26)

with the idea that and we move on to the diffusive term on the right-hand side of Eq. (17). Use of the averaging theorem leads directly to 1 s(DCA)t = sDCAt + n D C dA V A A (22) and the substitution of Eqs. (19), (21), and (22) into (17) provides the following form of the superficial-averaged mass transport equation:
s

vCAtex = 0

in the homogeneous regions (27)

The excess dispersive flux defined by Eq. (26) will not be equal to zero in the boundary, where the length-scale constraints developed by Carbonell and Whitaker (1984) are not valid. Then, the convective term in Eq. (23) can be expressed as
  vCA = ( v CA ) + vCA +

vCA ex

convection

t
C A + v C A = D C A + convection

dispersion nonlocal dispersion

accumulation

D C dA
V n
diffusion
1
A

(28) Here, we have identified svCAtex as a nonlocal term, since it involves, indirectly, values of sCAt that are not associated with the centroid of the averaging volume illustrated in Fig. 2. Substituting Eq. (28) into Eq. (23) leads to a form that contains the traditional convective and dispersive transport terms in addiction to the excess dispersion:

interfacial flux

(23) It is important to note at this point that we have not imposed any length-scale constraints on the volume-average transport equation. The absence of any length-scale constraint means that Eq. (23) is also valid in the boundary between the and regions, illustrated in Fig. 3. The traditional representation of the convective flux is given by (Carbonell and Whitaker, 1983)

CA t

  + ( v CA ) + vCA +

vCA ex

vCAt = svtsCAt (24) ~ v in which ~ and CA represent the spatial deviations around averaged values of the local variables, and are defined by the following decompositions (Gray, 1975):
s

~ + s~CAt v

accumulation

convection 1 V

dispersive

nonlocal dispersion

= D CA +

diffusion

n D CA dA

(29)

interfacial flux

~ CA = sCAt + CA
v = svt + ~ v

(25a) (25b)

Turning our attention to the diffusion transport term and making use of the averaging theorem, we obtain:
s

Use of Eq. (24) would require the imposition of lengthscale constraints (Carbonell and Whitaker 1984; Quintard and Whitaker 1994), and we need to avoid this if we are to develop a transport equation that is valid within the boundary region between the and re-

1 DCAt = DsCAt + nCA dA V A

(30)

In writing Eq. (30) we have ignored variations of the diffusion coefficient D within the averaging volume. Now, we employ Eq. (14) to express this result in

40 terms of the intrinsic average concentration according to


s

Valencia-Lpez et al. values of sCAt that are evaluated at points within the averaging volume not located at the centroid. The substitution of Eq. (35) into Eq. (29) leads to the following result:

DCAt = D sCAt + sCAt 1 A n CAdA V

(31)

sCAt ~ + svtsCAt + s~CAt v t

In order to obtain a volume-average transport equation that contains only average quantities and spatial deviations, the interfacial area integral in this equation is developed with the same idea given by Ochoa-Tapia and Whitaker (1998b). From the volume-averaging theorem given by Eq. (16) we can obtain the result

~ 1 + svCAtex = DsCAt + nCAdA V A + 1 n (sCAt sCAtx) dA V A 1 n D C dA V A A

1 n dA = V A
which can be used to write

(36)

(32)

1 n sC t dA = sCAtx V A A x = sCAt

Use of the boundary condition given by Eq. (7) in the interfacial flux term in Eq. (36), and the fact that is independent of time, allow us to interchange integration and differentiation to obtain the following expression:

(33)

CA 1 1 A n DCA dA = V A Keq t dA V

It can be observed that the convention used here is that averaged quantities located outside an integral are always evaluated at the centroid unless there is some specific indication to the contrary. Use of this result together with Eq. (31) leads to the following expression for the diffusive transport:

= Keq

1 CA dA t V A

(37)

This equation can be rewritten as sCAt 1 n D C dA = avKeq V A A t

(38)

1 sDCAt = D sCAt + n (CA V A


(34) sCAtx)dA Use of the decomposition given by Eq. (25a) into Eq. (34) yields the following representation for the diffusion transport term: ~ 1 sDCAt = D sCAt + n CAdA V A + 1 A n (sCAt sCAtx)dA V

where av represents the interfacial area per unit volume of the porous medium, which is given by

av(x) =

A(x) V

(39)

and the area-averaged concentration is defined by

sCAt =

1 C dA A A A

(40)

(35)

Substituting Eq. (38) into Eq. (36) yields a general form of mass transport equation, which is valid in and regions, given by:
avKeq sCAt 1 + + (svtsCAt) t

The second integral on the right-hand side of this equation is identified as a nonlocal term, since it involves

Mass Transfer Jump Condition at the Boundary ~ + s~CAt + svCAtex v = (DsCAt)

41 (45)

~ 1 = D sCAt + n C dA V A A
+

where represents the nonequilibrium concentration source and is defined as follows: = avKeq (sCAt sCAt) t (46)

1 n (sC t sCAtx) dA V A A (41)

av Keq (sCAt sCAt) t This equation can be rewritten as follows:


avKeq sCAt 1 + + (svtsCAt) t

This source term is negligible in both homogeneous regions: the bulk of the porous medium and the bulk of the clear fluid region. The equations that are valid in the homogeneous region and homogeneous region are given in the following sections. Region It is important to note that, in the homogeneous region, the area-averaged concentration, sCAt is essentially equal to the volume-average concentration, sCAt. Thus, the nonequilibrium concentration source, , is essentially null in the region. The starting point to justify this result is the area-average concentration given by Eq. (40). The decomposition, given by Eq. (25a), allows expressing sCAt as
s

= (Deff + DD) sCAt av Keq (sCAt sCAt) t (42)

In this equation we have introduced the following definitions: ~ 1 DeffsCAt = D sCAt + n C dA V A A 1 (43) + n (sCAt sCAt x) dA V A ~ DDsCAt = (s~CAt + svCAtex) (44) v These representations have been adopted because, in the bulk of the porous medium ( region), Deff and DD reduce to the effective diffusivity tensor and to the dispersion tensor, respectively. This idea is better emphasized if one observes that the last integral term in Eq. (43) represents the nonlocal contribution to the diffusive transport, and it is negligible when the tensor Deff reduces to the classical representation for the effective diffusivity tensor (Whitaker, 1999). According to the previous line of thought, we define D* as the total dispersion tensor, and it is used to replace Deff + DD in Eq. (42). Then, the generalized mass transport equation is given by
avKeq sCAt 1 + + (svt sCAt) t

CAt =

~ 1 A sCAtx+y + CAx+y dA (47) A

Whitaker (1999) showed, that for most practical problems of diffusion and reaction in porous media, the spatial deviation concentration is small compared to the volumeaveraged concentration, i.e.,

~ CA << sCAt

(48)

This inequality is valid whenever the following lengthscale constrains are satisfied:
O

<< LC << r0

(49a) (49b)

Here O is the characteristic length associated with the phase within the region, r0 is the characteristic length associated with the radius of an averaging volume; where according with Whitaker (1999) r0 << LC, and LC is the characteristic length associated with changes in CA, i.e.,

42 sCAt sCAt = O LC Then, Eq. (47) simplifies to


s

Valencia-Lpez et al. On the basis of this estimate, and the restriction given by Eq. (54), we can express Eq. (52) as
s

(50)

CAt =

1 sCAt x+y dA A A

(51)

r2 o CAt = sCAt + O sCAt + ... (58) LC1LC

This indicates that when the following two length-scale constraints, r0 << LC r2 << LC1LC 0 are valid, Eq. (58) can be expressed as (59a) (59b)

At this point, we can use Taylor series expansion of


s

CAt, around the centroid of the averaging volume, to express Eq. (51) as
s

CAt = sCAt x + syt sCAt x


1 sy y t : sCAtx + 2 (52)

CAt = sCAt

(60)

No precise estimates of syt are available; however, one can assume that it is small compared to the characteristic length of the averaging volume, i.e.,
s

This result and the fact that in the homogeneous region the nonlocal term is null allow us to simplify Eq. (45) to obtain
avKeq sCAt 1 + + (svt sCAt ) t

yt << r0

(53)

On the basis of Eq. (50) this result leads to


s

r0 yt sCAt << CA <<< sCAt L C

(54)

= (D sCAt )

(61)

where it is expected that r0 << LC will be satisfied in locations away from the interregion. Now, we move on to the term involving sCAt. An order-of-magnitude estimate of the second derivative of the average concentration is given by sCAt sCAt = O LC1LC

which is a mass transport equation that is valid only in the homogeneous region. In Eq. (61) the subscript is used to identify variables and parameters in the phase associated with the homogeneous region. The total dispersion tensor, D , in the homogeneous re gion is given by D sCAt = DsCAt ~ ~ 1 v n C dA s~CAt + V A A

(55)

(62)

where LC1 represents a characteristic length associated with the sCAt gradients [(sCAt)]. We can draw on the work of Whitaker (1999) to obtain the order-ofmagnitude estimate
s

yyt = Or2 o

(56)

and this result, along with our estimate of sCAt given by Eq. (55), leads to
s

The result given by Eq. (60) implies that, when the length-scale constrains given by Eq. (59) are fulfilled, is negligible. Most probably this will not occur in the neighborhood of the boundary , therefore 0. Region In the homogeneous region the volume fraction of the phase () is equal to 1. This implies that av = 0 and that the nonlocal term is null. Therefore, Eq. (45) reduces

r2 o yyt : sCAt = O sCAt LC1LC

(57)

Mass Transfer Jump Condition at the Boundary sCAt t

43 In the process of deriving Eq. (63) from Eq. (45), we have made use of the following relation:
s

+ (svt sCAt ) = (DsCAt ) (63)

This is the mass trahsport equation valid in the homogeneous region. In Eq. (63) the subscript is used to identify variables and parameters in the phase associated with the homogeneous region. It is important to note that the mass transport equation in the homogeneous region has exactly the same form as the original point equation given by Eq. (1). This is based on the approximation that the local volume-averaged values in the homogeneous region are equal to the corresponding point values, i.e.,
s

vCAt = vCA
(69)

in the homogeneous region

in order to conclude that there is no dispersion in the homogeneous region. For a spherical averaging volume 2 syyt = O(r0) and this means that the constraint given by Eq. (67) takes the following form: r2 << o Thus we can express Eq. (66) as
s

(70)

t x = stx = x (64)

in the homogeneous region

tx = x + Or2 0 L1L

(71)

The justification of this result is reported by OchoaTapia and Whitaker (1995a). They showed that average and point values in the homogeneous region are related by a Taylor series expansion about the centroid, which allows expressing Eq. (11) as
s

where L and L1 are characteristic lengths associated with changes in () and changes in the gradients [()], respectively. With Eq. (71), we can conclude that the condition indicated by Eq. (64) is valid whenever the following length-scale constraint is satisfied: r2 0 << 1 L1L in the homogeneous region

tx = stx + syt x + + 1 sy y y t e x + 6

1 sy y t : x 2 (65)

(72)

where y is the position vector relative to the centroid V. Therefore, syt = 0 at points located in the homogeneous region. For a spherical averaging volume syyyt is zero. Thus, for most transport processes in the homogeneous fluid region, Eq. (65) reduces to
s

tx = x +

1 sy y t : x 2

(66)

When this length-scale constraint is satisfied, the analysis of the jump condition is greatly simplified (Whitaker, 1992). Before moving on to the derivation of the jump condition, it is important to be aware that length-scale constraints associated with are applied only in the homogeneous

This means that Eq. (64) is valid whenever


s

yyt : x << x

(67)

Under such conditions the velocity and concentration are given by


s

region, and that no length-scale constraints have been imposed on the generalized mass transport equation [Eq. (45)] in the boundary region. Equation (45) is the starting point to develop the jump condition at the boundary, and this is the objective of the following section.
JUMP CONDITION The coefficients and the nonequilibrium source in the generalized transport equation given by Eq. (45) undergo extremely rapid variations in the boundary region. To avoid the difficulty associated with these rapid variations, we will

CAt = sCAt = CA

(68a) (68b)

vt = svt = v

Here it is understood that the average and the point quantities are evaluated at the same position.

44 apply the transport equations that are valid in the homogeneous parts of the and regions to the entire space occupied by the and regions. For example, this means that the computed values of sCAt and sCAt in the boundary region will not be equal to the value of sCAt that would be determined by Eq. (45). We have indicated this situation in Fig. 4, where concentration profiles are illustrated for two cases, nD*n > D and nD*n < D. It is important to understand that the profiles for sCAt and sCAt are not extrapolations, but are solutions to the transport equations that are valid in the homogeneous and regions and applied everywhere, while sCAt is continuous in the boundary region, as illustrated in Fig. 4. It should be clear that the fluid volume fraction is a continuos function of position, which takes the value in the homogeneous porous medium region and in the homogeneous fluid region. Therefore, the superficial concentration sCAt is also continuous everywhere in the two-phase system, including the interregion neighborhood. In this analysis we will make use of the idea that the generalized mass transport must be satisfied within the large-scale averaging volume V shown in Fig. 5. The volume of the and regions contained in V are designated by V and V, respectively, thus we have V(t) + V(t) = V (73)

Valencia-Lpez et al. bounding surfaces located in the and regions according to A + A = A (74)

The procedure to develop a jump condition associate with the boundary is given by Ochoa-Tapia and Whitaker (1995a). One first integrates the equations that are valid in the homogeneous and regions over V and V. Next, one integrates the generalized mass transport equations over V and the former integrals are subtracted from the latter to eventually obtain a jump condition. The use of this jump condition in conjunction with the equations that are valid in the homogeneous and regions, but applied everywhere in those regions, provides a solution for which the generalized mass transport equation is satisfied on the average. Ochoa-Tapia and Whitaker (1995a) have carried out this procedure for the continuity equations. They present the result
s

vtn = svtn
(75)

at the boundary

Now, to develop the jump condition for the mass transport process, we first integrate Eqs. (61) and (63) over volumes V and V, respectively, and then add them to obtain:

sCAt avKeq sCAt V 1 + t dV + V t dA

The area of the surface, A, that defines the large-scale averaging volume can also be represented in terms of the

Figure 4. Concentration profiles in the boundary region: (a) n D n > D and (b) n D n > D.

Mass Transfer Jump Condition at the Boundary + n(svtsCAt)dA n(svtsCAt)dA CAt)dA +

45

+ =

A n(D

A n(D

CAt)dA (77)

A dV

V dV

When we subtract Eq. (76) from Eq. (77), we obtain the lengthy result that is given below:
Figure 5. Large-scale averaging volume V and unit vectors.
avKeq sCAt 1 + V t
)

+ + + = +

A n(

vt sCAt )dA

A n(

vt sCAt )dA

avKeq sCAt 1 + dV t

n(svt sCAt ) svt sCAt )dA


+

+ + +

sCAt avKeq sCAt 1 + dV t t


s

n(D sCAt )dA n(D sCAt )

n(DsCAt )dA

A n(

vt sCAt svt sCAt )dA

DsCAt )dA

(76)

A n(

vt sCAt svt sCAt )dA

The various unit vectors used in this result are identified in Fig. 5. To obtain Eq. (76) we have used the divergence theorem to change several of the volume integrals to area integrals. In this result, we have also used A = A to represent the area of dividing surface contained within the volume V, and the convention associated with the unit normal vectors at the dividing surface is that n = n. The next step requires that we integrate Eq. (45) over the volume identified by V, thus the following result is obtained:

A
= +

n(svt sCAt svt sCAt )dA

A n(D

CAt D sCAt )dA

A n(D

CAt DsCAt )dA

n(D sCAt ) DsCAt )dA

V 1

avKeq sCAt dV t

A dV

V dV

(78)

avKeq sCAt + 1 + dV t V

The type of terms containing integrals in the surfaces A and A is the characteristic that first led Gibbs (1928) to define excess functions at phase interfaces; an explanation that is applicable to the particular problem under investigation is given elsewhere (Ochoa-Tapia and Whi-

46 taker, 1995a). The excess surface accumulation associated with Eq. (78) is defined next. Excess Surface Accumulation

Valencia-Lpez et al.

n (D* C A D C A )dA

(81)

A
=

sCAts t

dA

Here we use s to represent the surface gradient operator and it is related to the nabla operator by s = (I nn) (Ochoa-Tapia et al., 1993). Nonequilibrium Excess Surface

avKeq sCAt 1 + V t

avKeq sCAt 1 + dV t

s dA =

V dV + V dV

(82)

sCAt avKeq sCAt 1 + dV t t V

(79)

in which is defined by Eq. (46). This definition, in accord with previous works (Ochoa-Tapia and Whitaker, 1998b), suggests that the nonequilibrium surface excess can be represented as s = (83)

where s represents the excess surface porosity, and sCAts is the surface concentration. Excess Surface Convective Transport

Cns s
=

vt sCAt d s s

where represents the thickness associated with the boundary region and can be estimated as the root of the norm of the permeability tensor in Darcys law. Substituting Eqs. (84), (85)(87) into Eq. (78) leads to the following result:

* n (D* C A D C A )dA

n ( v C A v C A )dA

n ( v C A v C A ) dA

(80)

* n ( D C A D C A )dA

In this equation, as can be observed in Fig. 5, C represents a closed curve lying on the dividing surface, A, while ns represents the unit vector that is tangent to the boundary and normal to the curve C, and represents the arc length along this curve. The surface velocity is represented by ssvt (Ochoa-Tapia and s Whitaker, 1995a). Excess Surface Diffusive Transport

sCAt s t

dA +

C ns s

vt sCAt d s s s dA (84)

ns s D s sCAt d s s
C

Using the surface divergence theorem (Ochoa-Tapia et al., 1993) to write all of the terms under the same area integral, and then requiring that the integratal be zero, we obtain a surface transport equation, or a jump condition:

C ns s D s s

CAt d s

s
* A

n (D C A D C ) dA
*



t excess surface transport

C A s

+ s s v s C A s sD* s C A s s

Mass Transfer Jump Condition at the Boundary


n ( v C A v C A )

47 Then, if the excess surface transport and the nonequilibrium terms are neglected, the boundary condition takes the form n(D sCAt DsCAt ) = 0 (90)



convective transport
s

* = n ( D C A D C A )



nonequilibrium
diffusive transport surface excess

(85)

Although this is a simplified expression, we should remember that it has to be used together with the velocity and concentration conditions given by Eqs. (75) and (86). General Form of the Boundary Condition Here we follow the ideas proposed by Ochoa-Tapia and Whitaker (1995b) for their study on momentum transport. On the basis of the definitions of the excess surface terms given by Eqs. (79)(82), the following representations are proposed. Excess Surface Accumulation: s sCAt s t = a sCAt t (91)

We must remember that at this point we have not imposed any condition on the average concentration at the dividing surface of the porous and fluid regions. Therefore, based on the continuity of the fields sCAt and sCAt shown in Fig. 4, we choose to impose the continuity of the intrinsic average concentrations at the boundary:
s

CAt = sCAt
(86)

at the boundary

This also implies that the surface concentration must satisfy


s

Excess Surface Diffusive Transport: s(s DssCAt) = d n( D sCAt s s

CAt = sCAt = sCAt s


(87)

at the boundary

D sCAt) Nonequilibrium Excess Surface: s = s av Keq sCAt t

(92)

Equation (87) together with the continuity of the superficial velocity lead to the result n(svt sCAt at the boundary With this result, Eq. (85) is reduced to
svtsCAt)

= 0 (88)

(93)



t excess surface transport

C A s

+ s s v s C A s sD* s C A s s

Note that, due to Eq. (88), it was not necessary to introduce an equation for the excess surface convective transport. In Eqs. (91)(93), a, d, and s are dimensionless parameters of order one. Using the Eqs. (91)(93) in Eq. (89) yields
a

* = n ( D C A D C A )



nonequilibrium
s

excess surface accumulation



excess surface transport
t

C A

* + d n ( D C A D C A )

diffusive transport

surface excess

* = n ( D C A D C A ) s av K eq

(89)


t
diffusive transport
nonequilibrium surface excess

C A

(94)

48 which can be written in terms of only one adjustable parameter as n(D sCAt DsCAt ) = av Keq sCAt t (95)

Valencia-Lpez et al. ous fluid that contains excess surface mass transport and nonequilibrium surface excess terms. The introduction of representations for the excess terms, as functions of bulk average concentrations, allows expressing the flux jump condition in terms of only one adjustable coefficient. This coefficient indicates the significance of the accumulation rate at the interregion between the porous medium and the homogeneous fluid. Experimental studies are needed to determine the adjustable parameter.In this development, we have also imposed the condition that the superficial average velocity and the intrinsic average concentration are continuous at the boundary. ACKNOWLEDGMENT The authors wish to acknowledge the financial support provided by the Consejo Nacional de Ciencia y Tecnologia (CONACyT) and the Instituto Mexicano del Petroleo (IMP). The authors also appreciate the comments and suggestions from Prof. Stephen Whitaker of the University of California at Davis. REFERENCES
Alazmi, B. and Vafai, K., Analysis of fluid flow and heat transfer interfacial conditions between a porous medium and a fluid layer, Int. J. Heat Mass Transfer, vol. 44, pp. 17351749, 2001. Aris, R., The longitudinal diffusion coefficient in flow through a tube with stagnant pockets, Chem. Eng. Sci., vol. 11, pp. 194198, 1959. Balakotaiah, V. and Luss, D., Effect of flow direction on conversion in isothermal radial flow fixed-bed reactors, AIChE J., vol. 27, pp. 442450, 1981. Carberry, J. J., Chemical and Catalytic Reaction Engineering, McGraw-Hill, New York, 1976. Carbonell, R. G. and Whitaker, S., Dispersion in pulsed system: II Theoretical developments for passive dispersion in porous medium, Chem. Eng. Sci., vol. 38, pp. 17951802, 1983. Carbonell, R. G. and Whitaker, S., Heat and mass transport in porous media, in Fundamentals of Transport Phenomena in Porous Media, J. Bear and M. Y. Corapcioglu (eds.), Martinus-Nijhoff, Dordrecht, The Netherlands, pp. 123198, 1984. Fogler, S., Elements of Chemical Reaction Engineering, PrenticeHall, Englewood Clifts, NJ, 1992. Gibbs, J. W., The Collected Works of J. Willard Gibbs, vol. 1, Yale University Press, New Haven, CT, 1928. Gray, W. G., A derivation of the equations for multi-phase transport, Chem. Eng. Sci., vol. 30, pp. 229233, 1975.

In this equation the parameter involves the dimensionless parameters a, d, and s. It is clear that for steady-state conditions the boundary condition reduces to the one under the assumption of excess surface neglegible effects given by Eq. (90). At this point we must recognize that if the excess surface accumulation term in Eq. (95) is important, we are faced with the problem of designing experiments that will allow for the direct measurement of that quantity. This situation occurs routinely when a heterogeneous reaction rate (Carberry, 1976) or adsorption isotherm (Fogler, 1992) needs measuring. To measure these heterogeneous reaction rates and adsorption isotherms, one attempts to carry out experiments where these effects dominate the observed phenomena; however, this is not easily done for the unknown terms that appear in Eq. (95). Numerical experiments, such as those described in previous works (Larson and Higdon, 1987a, 1987b; Prat, 1992; Sahraoui and Kaviany, 1992, 1993, 1994), provide a means of examining the details of the boundary region. From a practical point of view, it seems plausible that the excess surface accumulation term in Eq. (95) can be neglected. This means that the flux jump condition takes the form indicated in Eq. (90). In further contributions, we will develop a jump condition for the boundary between a porous medium and a homogeneous fluid, where the importance of the surface curvature should be enhanced. CONCLUSION In this study we have derived a jump condition for the boundary between a porous medium and a homogeneous fluid with adsorption effects that ensures that the generalized transport equation will be satisfied, on the average. The development is based on a generalized, nonlocal form of the volume-average mass transport equation that is valid everywhere in a system composed of a porous medium and a homogeneous fluid. This generalized mass transport equation was used to construct a jump condition at the boundary between the porous medium and the homogene-

Mass Transfer Jump Condition at the Boundary


Howes, F. A. and Whitaker, S., The spatial averaging theorem revisited, Chem. Eng. Sci., vol. 40, pp. 13871392, 1985. Huang, H. S., Roy, S., Hou, K. C., and Tsao, G. T., Scaling-up of affinity chomatography by radial-flow cartridges, Biotechnol. Prog., vol. 4, pp. 159165, 1988. Kuznetsov, A. V., Analytical investigation of the fluid flow in the interface region between a porous medium and a clear fluid in channels partially filled with a porous medium, Appl. Sci. Res., vol. 56, pp. 5367, 1996. Kuznetsov, A. V., Influence of the stress jump boundary condition at the porous-medium/clear-fluid interface on a flow at a porous wall, Int. Commun. Heat Mass Transfer, vol. 24, pp. 401410, 1997. Kuznetsov, A. V., Analytical study of fluid flow and heat transfer during forced convection in a composite channel partly filled with a Brinkman-Forchheimer porous medium, Flow, Turbulence and Combustion, vol. 60, pp. 173192, 1998. Kuznetsov, A. V., Fluid mechanics and heat transfer in the interface region between a porous medium and a fluid layer: A boundary-layer solution, J. Porous Media, vol. 2, pp. 309 321, 1999. Larson, R. E. and Higdon, J. J. L., Microscopic flow near the surface of two-dimensional porous media. Part 1. Axial flow, J. Fluid Mech., vol. 166, pp. 449472, 1987a. Larson, R. E. and Higdon, J. J. L., Microscopic flow near the surface of two-dimensional porous media. Part 2. Transverse flow, J. Fluid Mech., vol. 178, pp. 119136, 1987b. Lee, W.-C., Tsai, G.-J., and Tsao, G. T., Radial-flow affinity chromatography for trypsin purification, ACS Symp. Ser., vol. 427, pp. 104117, 1990. Lenhoff, A. M. and Lightfoot, E. N., The effects of axial diffusion and permeability barriers on the transient response of tissue cylinders. II Solution in time domain, J. Theor. Biol., vol. 106, pp. 207238, 1984. Lenhoff, A. M. and Lightfoot, E. N. Convective dispersion and interphase mass transfer, Chem. Eng. Sci., vol. 41, pp. 2795 2810, 1986. Ochoa-Tapia, J. A., Del Rio, P. J. A., and Whitaker, S., Bulk and surface diffusion in porous media: An application of the surface averaging theorem, Chem. Eng. Sci., vol 48, pp. 20612082, 1993. Ochoa-Tapia, J. A. and Whitaker, S., Momentum transfer at the boundary between a porous medium and a homogeneous FluidI. Theoretical development, Int. J. Heat Mass Transfer, vol. 38, pp. 26352646, 1995a.

49
Ochoa-Tapia, J. A. and Whitaker, S., Momentum transfer at the boundary between a porous medium and a homogeneous fluidII. Comparison with experiment, Int. J. Heat Mass Transfer, vol. 38, pp. 26472655, 1995b. Ochoa-Tapia, J. A. and Whitaker, S., Heat transfer at the boundary between a porous medium and a homogeneous fluid, Int. J. Heat Mass Transfer, vol. 40, pp. 26912707, 1997. Ochoa-Tapia, J. A. and Whitaker, S., Momentum jump condition at the boundary between a porous medium a homogeneous fluid: Inertial effects, J. Porous Media, vol. 1, pp. 201217, 1998a. Ochoa-Tapia, J. A. and Whitaker, S., Heat transfer at the boundary between a porous medium and a homogeneous fluid: The one-equation model, J. Porous Media, vol. 1, pp. 3146, 1998b. Prat, M., Some refinements concerning the boundary conditions at the macroscopic level, Transport Porous Media, vol. 7, pp. 147161, 1992. Quintard, M. and Whitaker, S., Transport in ordered and disordered porous media I: The cellular average and the use of weighting functions, Transport Porous Media, vol. 14, pp. 163177, 1994. Sahraoui, M. and Kaviany, M., Slip and no slip temperature boundary conditions at the interface of porous, plain media: Convection, Int. J. Heat Mass Transfer, vol. 37, pp. 1029 1044, 1994. Slattery, J. C., Interfacial Transport Phenomena, Springer-Verlag, New York, 1990. Whitaker, S., Transport processes with heterogeneous reaction, in Concepts and Design of Chemical Reactors, S. Whitaker and A. E. Cassano (eds.), Gordon & Breach, New York, pp. 194, 1986a. Whitaker, S., Transient diffusion, absorption and reaction in porous catalysts: The reaction controlled, quasi-steady catalytic surface, Chem. Eng. Sci., vol. 41, pp. 30153022, 1986b. Whitaker, S., The species mass jump condition at a singular surface, Chem. Eng. Sci., vol. 47, pp. 16771685, 1992. Whitaker, S., The Forchheimer equation: A theoretical development, Transport Porous Media, vol. 25, pp. 2761, 1996. Whitaker, S., The Method of Volume Averaging, Kluwer Academic, Dordrecht, The Netherlands, 1999. Zanotti, F. and Carbonell, R. G., Development of transport equations for multiphase systems I: General development for twophase systems, Chem. Eng. Sci., vol. 39, pp. 263278, 1984.

Journal of Porous Media 6(1), 5157 (2003)

(IIHFWV RI *URVV +HWHURJHQHLW\ DQG $QLVRWURS\ LQ )RUFHG &RQYHFWLRQ LQ D 3RURXV 0HGLXP /D\HUHG 0HGLXP $QDO\VLV
D. A. Nield1 and A. V. Kuznetsov2*
1

Department of Engineering Science, University of Auckland, Private Bag 92019, Auckland, New Zealand 2 Department of Mechanical and Aerospace Engineering, North Carolina State University, Campus Box 7910, Raleigh, NC 27695-7910, USA E-mail: avkuznet@eos.ncsu.edu ABSTRACT The effects of gross heterogeneity and anisotropy, associated with horizontal fissures in a porous medium filling a parallel-plate channel, on forced convection are studied. An approximate analysis, based on a piecewise-constant (layered) distribution of permeability across the channel, is used to model the experiments performed by Paek et al. (1999) on foam material with drilled-out tubes. The analysis leads to estimates of the Nusselt number, Nu. Drilling out the tubes leads to a reduction in the value of Nu but an even greater reduction in the friction factor, f, so that the net result is an increase in the value of Nu/f, i.e., for a given driving pressure gradient the heat transfer is increased.

51 Received October 11, 2001; Accepted February 26, 2002 Copyright 2003 Begell House, Inc.

52

Nield and Kuznetsov

NOMENCLATURE
d f G H k _ k K _ _ K Nu q S T* Tm Tw T u u*
^

tube diameter friction factor applied pressure gradient channel height thermal conductivity mean value of k permeability mean value of K Nusselt number heat flux fissure spacing temperature bulk mean temperature wall temperature = (T* Tw)/(Tm Tw) _ = u GH2 velocity

u U W x x y y

= u U mean velocity channel width longitudinal coordinate = x H transverse coordinate = y H

Greek symbols void fraction of a channel interface parameter defined before Eq. (4) viscosity density dimensionless parameters defined by Eq. (7) , porosity of a porous medium

INTRODUCTION Because it possesses the properties of high surface areato-volume ratio together with relatively high permeability, metallic foam material has recently become widely employed in industrial applications that require compact and highly efficient thermal systems. A typical aluminum foam material can be modeled as a porous medium of large porosity. A number of recent articles have been concerned with forced convection in such media, and in the context of enhanced heat transfer a very useful survey has been made by Lage and Narasimhan (2000). A particularly interesting article is that by Paek et al. (1999). They report the results of experiments in aluminum foam metal material that has been made anisotropic by the drilling of holes in the material. The motivation for this fabrication is that, compared with a fluid that is clear of solid material, the foam material has an increased resistance to flow because of friction, and usually this is undesirable. The required pressure drop can be reduced if holes in the form of axial tubes are drilled out. Clearly, this introduces gross anisotropy into the porous medium. A look at the book by Nield and Bejan (1999) reveals that a

large number of articles have dealt with natural convection in an anisotropic porous medium. However, it appears that the topic of forced convection in an anisotropic porous medium has received little study, and this is why the Paek et al. (1999) article is highly significant. The situation dealt with in this article has another feature of interest. For one type of material investigated, the drilled holes were distributed across the channel in a relatively uniform manner, but in another type they were concentrated at mid-channel. Thus, in the latter case, the material was heterogeneous as well as anisotropic. The study by Paek et al. (1999) was entirely experimental. The situations studied by those authors involved complicated geometries, and so one should not expect to model these situations quantitatively by simple analysis. Nevertheless, some simple analysis can provide some qualitative information which can be expected to have wide applicability, and that is the purpose of the present article. ANALYSIS: DARCY MODEL First, we note the relationship between flow through an array of parallel straight capillary tubes of diameter d and

Effects of Gross Heterogeneity and Anisotropy in Forced Convection flow through an equivalent porous medium. If the void fraction is , then the equivalent permeability is [see, for example, Eq. (2.11) of Kaviany (1995)]

53

our model, and suppose that sandwiched between two layers of thickness (1 )H/2 we have a layer of thickness H of effective permeability

K eff =

d 2 32

(1)

K eff =

d 2 + 1 K 32

(4)

We now extend this result to the case where a fraction of the cross-sectional area of the channel is occupied by fissures in the form of capillary tubes, and the remainder is occupied by a porous medium of permeability K. The equivalent permeability of the composite structure is found, using the same argument as in Kaviany (1995), to be

It is clear that we have to take > d/H. Ideally, for an accurate model, we would like to have 1 > >> d/H, but this condition cannot be met in the case of the experiments of Paek et al. (1999). However, we confirm below that the Nusselt number depends only weakly on and so the inaccuracy is not too important. Let km be the thermal conductivity of the porous foam material, and kf be that of the fluid. Then in the middle layer the thermal conductivity can be taken to be

K eff =

d 2 + (1 ) K 32

(2)

keff =

It is assumed that the flow through the tubes and that through the porous medium are in parallel and are driven by the same pressure gradient. Finally, we apply this result to the medium denoted as Type 1 by Paek et al. (1999). In the cross section of Fig. 1, the medium has height H and width W. The fissures have diameter d and are spaced along the midline of the channel with spacing S between the centers of adjacent tubes. It follows that the number of tubes is W/S, and the cross-sectional area of the array of tubes is (W/S)(d2/4), and thus the void fraction is = d2 4SH (3)

k + 1 k f m

(5)

and hence the equivalent permeability is given by Eq. (2) with given by Eq. (3). So far we have not taken account of the fact that the medium is not homogeneous and that the tubes are concentrated at mid-channel. For more accuracy, we further refine

This equation is obtained on the assumption that the heat transfer through the tubes and that through the porous medium in the middle layer are in parallel. In the problem considered by Paek et al. (1999), the upper boundary was maintained at constant temperature and the lower boundary was adiabatic. This circumstance, together with the fact that we are using the Darcy model and hence we do not have to satisfy a no-slip boundary condition, allows us to use symmetry and model the situation by a layer between walls at y = H and y = H, with constant temperature at both walls. We recognize that we have a special case of the situation treated in Section 2.3.1 of Kuznetsov and Nield (2001), namely, a triple layer with isothermal boundaries, and a permeability and thermal conductivity distribution given by K = K1 and k = k1 for 0 < |y| < H K = K2 and k = k2 for H < |y| < H (6a) (6b)

Figure 1. Definition sketch. Channel cross section for the anisotropic foam metallic material of Type 1 investigated by Paek et al. (1999).

54 K = K3 and k = k3 for H < |y| < H In the present case we have (6c)

Nield and Kuznetsov longitudinal conduction assumed to be negligible and with the assumption of local thermal equilibrium)

1 = 2
and

1+ = 2

(7)

d 2T1 2 = 1 T1 dy 2 d 2T2 2 = 2 T2 2 dy

for 0 < y < for < y < for < y < 1

(13a)

(13b)

K1 = K, K2 = Keff, K3 = K, k1 = km, k2 = keff, k3 = km The mean values are given by (8)

d 2T3 2 = 3 T3 dy 2
where

(13c)

K = K1 + ( ) K 2 + (1 ) K 3
k = k1 + ( )k2 + (1 )k3

(9a) (9b)

 Nu Ki i =  2ki

1/ 2

for i = 1, 2, 3

(14)

We write
 Ki =

Ki K

 and ki =

ki k

for i = 1, 2, 3

(10)

The mean velocity U and the bulk mean temperature Tm are defined by

[The reader should note that once equivalent permeabilities and thermal conductivities are introduced, one has a standard problem for each of the layers. The situation is closely analogous to that treated by Kuznetsov and Nield (2001).] The Nusselt number Nu is defined as

U=

1 u * dy * H
0

Tm =

1 u * T * dy * HU
0

Nu =
(11)

2 Hq (Tw Tm ) k

(15)

The solutions of Eqs. (13a)(13c) satisfying the boundary conditions

Here the asterisks denote dimensional quantities. Dimensionless variables are defined by

u=

u* U

T * Tw T= Tm Tw

dT (0) = 0 dy
are

T (1) = 0

(16)

For the Darcy model, the velocity distribution is given by

T1 = A1 cos 1 y T2 = A2 cos 2 y + B2 sin 2 y


T3 = A3 sin 3 (1 y)

(17a) (17b) (17c)

u1 u2

 =K  =K

for 0 < y < for < y < for < y < 1

(12a) (12b) (12c)

 u3 =K 3

The continuity of temperature and heat flux at the interfaces y = and y = then implies the matching conditions

The temperature distribution is given by the solution of the equations (derived from the usual energy equation with

A1 cos 1 = A2 cos 2 + B2 sin 2

(18a)

Effects of Gross Heterogeneity and Anisotropy in Forced Convection

55 (22)

   k11 A1 sin 1 = k2 2 A2 sin 2 k2 2 B2 cos 2 (18b)

d 2 K = 1 32 K Kh

A2 cos 2 + B2 sin 2 = A3 sin 3 (1 )

(18c)

   k2 2 A2 sin 2 k2 2 B2 cos 2 = k33 A3 cos 3 (1 )

The drilling of the holes causes the resistance to fluid flow to be reduced in this ratio, so that the mean velocity is increased by the reciprocal of this ratio. Paek et al. (1999) work in terms of a friction factor f defined by f = H P / L 2 U (23)

(18d) The condition that Eqs. (18a)(18d) have a nontrivial solution is that
cos 1 k sin  1 det 1 1 0 0 cos 2  k22 sin 2 cos 2  k22 sin 2 sin 2
 k22 cos 2

sin 2
 k22 cos 2

=0 sin 3 (1 )  k33 cos 3 (1) 0 0

where P is the pressure drop over the length L and U is the uniform inlet velocity, so the drilling of holes results in this factor being decreased by the square of the factor given in Eq. (22). RESULTS AND DISCUSSION All figures are computed utilizing K = 7108 m2, km = 6.2 W/mK, and kf = 0.026 W/mK, in order to model the situation in Paek et al. (1999). In Fig. 2 we have plotted values of Nu versus for parameters corresponding to the Type 1 material of Paek et al. (1999). As one might expect, Nu is only weakly dependent on , and so the choice of (provided that it is not less than d/H) is not critical. Accordingly, in Fig. 3 we have varied the hole spacing ratio S/H, for various values of hole diameter ratios d/H with = d/H. Again as one would expect, for large spacing the holes have negligible effect and Nu is close to 2/2 = 4.93, and the smaller the spacing and the larger the holes, the greater is the reduction in the value of Nu as defined in Eq. (15). For comparison with experimental results presented in Fig. 12 of Paek et al. (1999), we have in our Fig. 4 plotted 2Nu/2 versus d/H for the appropriate value of S/H. Our predictions agree well with the experimental results for small values of d/H (of the order 0.1 or smaller), but our predictions markedly overestimate the values for larger values of d/H. We believe that the discrepancy can be explained by the fact that our analysis is based on the assumption of thermally developed flow, whereas in the experiments the flow was still developing. (A quick calculation, based on the experimental Reynolds number value (1,600) and a Prandtl number estimated as 0.7, shows the estimated development length-to-channel width ratio was of the order of 10 times the experimental channel lengthto-channel width ratio.)

(19) In view of Eq. (14), this equation may be regarded as an eigenvalue equation for Nu. As soon as the value of Nu has been found, a compatibility condition gives A3 = Nu 23 (20)

and then (18a)(18c) give A1, A2, and B2, to complete the solution. In general, Eq. (19) must be solved for Nu numerically. In the problem considered by Paek et al. (1999), involving aluminum foam with drilled-out tubes, the prime quantity of interest is the increase in heat transfer rate with the applied pressure gradient kept constant, or the increase in the ratio of heat transfer rate to applied pressure gradient. When Darcys law holds, the pressure gradient is proportional to the reciprocal of the permeability, and in the case of a layered medium it is the harmonic mean of the permeability Kh that is pertinent. Here

1 1 = + + K h K1 K2 K3

(21)

Inserting the parameter values given in Eqs. (7) and (8), and using the fact that and d2/32K are each small compared with unity, we find that approximately

56

Nield and Kuznetsov

Figure 2. Plot of Nusselt number versus interface parameter for the material of Type 1 used by Paek et al. (1999).

Figure 3. Plot of Nusselt number for various hole spacing ratios and hole diameter ratios.

Even though Nu is reduced by the drilling, the ratio Nu/f can be increased by the drilling, as illustrated by Fig. 13 of Paek et al. (1999). The value of K in their experiments was 7108 m2 (S. Y. Kim, private communication). For a medium of Type I with d/H = 0.2, our theory predicts that drilling will produce an increase in 1/f of about 10%, which is more than enough to offset the reduction in Nu.

CONCLUSION We have shown that a simple layered model is able to provide qualitative information applicable to the complicated heterogeneous and anisotropic porous medium used in forced-convection experiments reported by Paek et al. (1999). The methodology employed in this note can be employed to treat the case of more general horizontal fissures in a porous medium, provided that it is reasonable to

Effects of Gross Heterogeneity and Anisotropy in Forced Convection

57

Figure 4. Plot of Nusselt number for various hole diameter ratios, for comparison with the experimental values obtained by Paek et al. (1999).

assume that both the fluid flow and the heat transfer in the fissures and the porous medium are in parallel. ACKNOWLEDGMENTS A. V. Kuznetsov would like to acknowledge support provided by the North Carolina Supercomputing Center (NCSC) under an Advanced Computing Resources Grant. We thank Dr. S. Y. Kim for supplying us with additional data for the experiments reported in Paek et al. (1999). REFERENCES
Kaviany, M., Principles of Heat Transfer in Porous Media, 2nd ed., Springer-Verlag, New York, 1995.

Kuznetsov, A. V. and Nield, D. A., Effects of heterogeneity in forced convection in a porous medium: Triple layer or conjugate problem, Numer. Heat Transfer A, vol. 40, pp. 363385, 2001. Lage, J. L. and Narasimhan, A., Porous media enhanced forced convection fundamentals and applications, in Handbook of Porous Media (K. Vafai, ed.), pp. 357394, Marcel Dekker, New York, 2000. Nield, D. A. and Bejan, A., Convection in Porous Media, 2nd ed., Springer-Verlag, New York, 1999. Paek, J. W., Kim, S. Y., Kang, B. H., and Hyun, J. M., Forced convective heat transfer from anisotropic aluminum foam in a channel flow, Proceedings of the 33rd National Heat Transfer Conference, August 1517, 1999, Albuquerque, NM, Paper NHTC99-158, pp. 18.

Journal of Porous Media 6(1), 5970 (2003)

7KH (IIHFW RI WKH /RFDO ,QHUWLDO 7HUP RQ WKH )UHH&RQYHFWLRQ )OXLG )ORZ LQ 9HUWLFDO &KDQQHOV 3DUWLDOO\ )LOOHG ZLWK 3RURXV 0HGLD
A. F. Khadrawi1 and M. A. Al-Nimr2*
2

Mechanical Engineering, Al-Balqa Applied University Al-Salt, Jordan and Mechanical Engineering, Jordan University of Science and Technology Irbid, Jordan E-mail: malnimr@just.edu.jo ABSTRACT The transient hydrodynamics and thermal behaviors of the free-convection fluid flow in open-ended vertical parallel-plate channels partially filled with porous material are investigated. The role of the local macroscopic inertial term in the porous domain momentum equation is studied. The ranges of different dimensionless parameters within which the local inertial term is of significant effect are presented.

59 Received October 26, 2001; Accepted March 21, 2002 Copyright 2003 Begell House, Inc.

60

Khadrawi and Al-Nimr

NOMENCLATURE
Ca Da g k k1 k2 K KR L1 L2 Pr r t T T Tw u U acceleration coefficient tensor Darcy number (= K2/L2) 2 acceleration of gravity thermal conductivity thermal conductivity of clear domain thermal conductivity of porous domain permeability of the porous medium thermal conductivity ratio (= k2/k1) porous domain width channel width Prandtl number dimensionless ratio (= L2/L1) time temperature ambient temperature wall temperature axial velocity dimensionless axial velocity

Greek symbols thermal diffusivity thermal diffusivity of clear domain 1 thermal diffusivity of porous domain 2 thermal diffusivity ratio (= 2/1) R volumetric coefficient of thermal expansion porosity of the medium dimensionless temperature [= (T T)/(Tw T)] dynamic viscosity dynamic viscosity of clear domain 1 dynamic viscosity of porous domain 2 dynamic viscosity ratio (= 2/1) R kinematics viscosity dimensionless time (= t1 L2) 2 Subscripts R ratio, property 2/property 1 w 1 2 wall condition ambient condition refers to clear domain refers to porous domain

= u1 [L2g1(Tw T)] 2

y Y z

transverse coordinate dimensionless transverse coordinate (= y/L2) axial coordinate

INTRODUCTION The use of porous substrates to improve convection heat transfer in channels has many practical geophysical, environmental, and technological applications. Examples include electronic cooling, chemical and nuclear reactors, heat transfer from hair-covered skin, porous flat-plate collectors, grain and food storage and drying, solidification of concentrated alloys, packed-bed thermal storage, and fibrous and granular insulation where the insulation occupies only part of the space separating the heated and cooled walls. Examples of geophysical applications are found in geothermal reservoirs, in which the meteoric water, percolating down through the permeable formation, is heated by the intruding magma. Also, many geothermal areas that consist of volcanic debris confined by the walls on nonfragmented ignimbrite can be modeled as a fluid layer bounded by vertical porous walls subject to heat flux. An example of an environmental application is the thermal circulation in lakes, in which the shallow coastal water is

influenced significantly by the interaction between the overlying layer of water and the water-saturated substrate. In the literature, extensive research has been performed in recent years to examine the effects of different geometric, fluid, and solid matrix parameters on the natural convection in domains partially filled with porous materials. Many of these applications investigate the thermal stability of a composite fluid and porous layers for different configurations (Somerton and Catton, 1982; Poulikakos, 1987; Chen and Chen, 1988; Taslim and Narusawa, 1989; Prasad and Tian, 1990; Kuznetsov, 1998, 2000). Other studies deal with natural convection in cavities or enclosures (Nishimura et al., 1986; Tong and Subramanian, 1986; Tong et al., 1986; Beckermann et al., 1987; Vafai and Sarkar, 1987; Chang and Chang, 1996) partially filled with porous materials. Most of the above-reported works are either numerical (Somerton and Catton, 1982; Nishimura et al., 1986; Beckermann et al., 1987; Poulikakos, 1987; Vafai and Sarkar, 1987; Chen and Chen, 1988; Taslim and Narusawa, 1989) or experimental (Tong and Subramanian,

The Effect of the Local Inertial Term on the Free-Convection Fluid Flow 1986; Prasad and Tian, 1990), and they consider closed cavities and enclosures. Few of these studies investigate the free-convection transient thermal behavior in open-ended vertical channels partially filled with porous domain (AlNimr and Darabseh, 1997; Al-Nimr and Haddad, 1999; Al-Nimr, 1995; Al-Nimr and Massoud, 1998). The present work considers the transient free-convection fluid flow problem in open-ended vertical channels partially filled with porous media. The unsteadiness in the channel thermal behavior is due to a sudden change in the channel wall temperature. The main goal of the present study is to investigate the role of the macroscopic local inertial term in the porous domain momentum equation and its effect on the hydrodynamics and thermal behavior of channels partly filled with porous material. In the literature about fluid flow in domains totally filled with porous material, it has been realized that the local macroscopic inertial term is usually small compared to the microscopic Darcy drag term, and hence can be neglected (Nield, 1991). In most practical situations, the velocity responds to an imposed temperature change within a second or less. The local inertial term may be important if an oscillatory temperature change is imposed at the channel boundary or if the porous domain is of large void fraction. However, it is obvious that the local inertial term may retain its importance in applications involving very thin porous substrates or at large Darcy numbers. Abu-Hijleh and Al-Nimr (2001) have investigated the importance of the local inertial term in forced-convection fluid flow problems in channels partially filled with porous material. A quantitative mapping of the operating and geometric parameters within which the local inertial term may be significant in free-convection problems in domains partially filled with porous materials is not available in the literature yet. Also, the effect of these geometric and operating conditions on the importance of the local inertial term in free-convection problems needs to be investigated. In this study, the DarcyBrinkman model is adopted to describe the fluid flow hydrodynamic behavior. The tangential velocities and stresses are assumed to be matched at the clear fluid/porous domain interface. The inclusion of the Brinkman term is justified when the porous domain is light, i.e., > 0.6 (Lundgren, 1972). The continuity of the tangential velocities and shear stresses at the interface is widely used in the literature. It is also believed that this approach gives good predictions, especially in light porous domains, when > 0.6 (Kakac et al., 1991). The following section demonstrates the effect of the macroscopic local

61

inertial term on unsteady fully developed flow in openended vertical channel partly filled with porous material. The unsteadiness in the flow is due to sudden change in the temperature of the channel wall. MATHEMATICAL FORMULATION Consider an unsteady, laminar, fully developed freeconvection flow inside an open-ended vertical parallelplate channel partly filled with porous material. The fluid is assumed to be Newtonian with uniform properties, and the porous medium is isotropic and homogeneous. It is assumed that the unsteadiness in the channel thermal and hydrodynamic behavior is due to a sudden change in the temperature of the channel wall. Also, it is assumed that both viscous dissipation and internal heat generation are absent. Referring to Fig. 1, under the above-mentioned assumptions and using the dimensionless parameters given in the Nomenclature, the equations of motion and energy in both the clear and porous domains are given as, respectively, subject to the following initial and boundary conditions,

U 1 2U 1 = + 1 Y 2

(1) (2)

1 1 2 1 = Pr Y 2
Ca U 2 2U 2 1 U2 +2 = R 2 Da Y

(3)

2 2 2 = R Pr Y 2
U1(, 1) = 0.0

(4)

U1 (0, Y ) = U 2 (0, Y ) = 0.0 1 (0, Y ) = 2 (0, Y ) = 0.0

U 2 (, 0 ) = 2 (, 0 ) = 0.0 Y Y
1(, 1) = 1.0
(5)

62

Khadrawi and Al-Nimr

Figure 1. Schematic diagram of the problem under consideration.

U1(, t) = U2(, r)

1(, r) = 2(, r)

U1 U (, r ) = R 2 (, r ) Y Y 1 ( , r ) = K R 2 ( , r ) Y Y
where subscripts 1 and 2 refer to the clear and porous domains, respectively. In Eq. (3), Ca is an acceleration coefficient tensor that depends on the geometry of the porous medium (Nield, 1991). The value of this coefficient is far from being settled, but it may assumed to be or 1 in light domains having void fraction close to 1. The other parameters appearing in Eqs. (2)(5) are defined as 2 R = 1 k2 KR = k1 K Da = 2 L2 1 Pr = 1 L2 r= L1

Equations (1)(5) are solved using Laplace transformation technique. Now, with the notation that L[U(, Y)] = W(S, Y) and L[(, Y)] = V(S, Y), Laplace transformation of Eqs. (1)(4) yields

d 2W1 SW1 = V1 dY 2 d 2V1 Pr SV1 = 0.0 dY 2


d 2W2 V AW2 = 2 2 dY R

(6)

(7)

(8)

d 2V2 Pr .S .V2 = 0.0 dY 2 R


These equations assume the following solutions:
W1 ( S , Y ) = C1e
SY

(9)

+ C2e

SY

It is worth mentioning here that the DarcyBrinkman model is adopted to describe the fluid flow hydrodynamic behavior. The term describing the quadratic drag is neglected, which is justified in natural-convection problems especially at small Rayleigh numbers.

( Pr1) S

C3

eM1Y

( Pr1) S

C4

e M1Y

(10)

V1 ( S , Y ) = C3e M1Y + C4 e M1Y

(11)

The Effect of the Local Inertial Term on the Free-Convection Fluid Flow
W2(S, Y) = C5eA Y + c6eA Y

63

C7 eM2Y R[(Pr S R) A] (12)

C8 eM2Y R[(Pr S R) A]

V2 ( S , Y ) = C7 e M 2Y + C8 e M 2Y
where Pr M1 = S M2 =

(13)

where k = 1 for the clear domain and k = 2 for the porous domain. In Eqs. (15) and (16), Re represents the real part of the summation and i = . The quantity 1 = 4.7 gives the most satisfactory results (Tzou, 1997), where is a floating parameter used to get faster convergence. Equations (15) and (16) yield the exact temperature and velocity distributions in both domains. RESULTS AND DISCUSSION

S Pr
R

Ca 1 A = S + (R Da) R

Also, the Laplace transformation of Eq. (5), yields

W2 V ( S , 0 ) = 2 ( S , 0 ) = 0.0 Y Y W1 ( S ,1) = 0.0


W1(S, r) = W2(S, r) V1(S, r) = V2(S, r)

V1 ( S ,1) =

1 S
(14)

W1 W ( S , r ) = R 2 ( S , r ) = 0.0 Y Y V1 V ( S , r ) = K R 2 ( S , r ) = 0.0 Y Y
The constants C1C8 are found after inserting Eqs. (10)(13) into the boundary conditions (14). Equations (10)(13) are inverted using a computer program based on Riemann sum approximation as

Figures 24 show the effect of Darcy number on the velocity at different locations within the channel and at different values of the acceleration coefficient tensor Ca. Figure 2 shows the effect of Da and Ca on the velocity within the porous domain. For Da < 104, the frictional drag resistance against free convection is very large due to the small permeability of the porous domain, and as a result, the velocity is very small. As Da increases, the fluid velocity increases and the effect of the local inertial term becomes significant. As one approaches the clear domain part of the channel, the effect of Ca becomes more significant, as is clear from Figs. 3 and 4. Figures 24 show that small Da numbers, less than 104, have insignificant effect on the velocity in all locations of the channel. The effect of Da number on the channel hydrodynamic behavior becomes more significant as Da increases. Figure 5 shows the effect of Ca on the velocity spatial distribution. Again, it is clear that Ca has more significant effect on the clear domain hydrodynamic behavior than on that of the porous domain. Also, and due to its higher microscopic frictional drag resistance, the porous domain has much lower velocity than that of the clear domain. The effect of R on the clear domain velocity at different Ca values is shown in Fig. 6. It is clear that R has insignificant effect on the hydrodynamic behavior when the local inertial term is neglected. Also, the local inertial term has insignificant effect on the channel hydrodynamics behavior for R > 1. The effect of R on the channel hydrodynamic behavior is insignificant in channels having R > 1. Figures 7 and 8 show the effect of Ca on the velocity and temperature spatial distribution at different values of the thermal conductivity ratio KR. Although KR has significant effect on both velocity and temperature distribution, changing KR does not cause any change in the effect of Ca on both velocity and temperature distributions. From Fig.

k ( ,Y )

e 1 in n Wk Y , + ( 1) Wk (Y , ) + Re 2 n=1

(15)
n

Uk ( , Y )

e 1 V (Y, ) + Re 2 k

V Y, + (1)
in
k n=1

(16)

64

Khadrawi and Al-Nimr

Figure 2. Effect of Da on the velocity within porous domain at different Ca.

Figure 3. Effect of Da on interfacial velocity at different Ca.

The Effect of the Local Inertial Term on the Free-Convection Fluid Flow

65

Figure 4. Effect of Da on the velocity within clear domain at different Ca.

Figure 5. Effect of Ca on the spatial velocity distribution.

66

Khadrawi and Al-Nimr

Figure 6. Effect of R on the velocity within the clear domain at different Ca.

Figure 7. Effect of Ca on the velocity spatial distribution at different KR.

The Effect of the Local Inertial Term on the Free-Convection Fluid Flow

67

Figure 8. Effect of Ca on the temperature spatial distribution at different KR.

Figure 9. Effect of Ca on the temperature spatial distribution at different R.

68

Khadrawi and Al-Nimr

Figure 10. Effect of Ca on the velocity spatial distribution at different R.

Figure 11. Effect of Ca on the velocity spatial distribution at different Pr.

The Effect of the Local Inertial Term on the Free-Convection Fluid Flow

69

Figure 12. Effect of Ca on the velocity distribution within the clear domain.

7 it is clear that the velocity increases as KR decreases and Ca increases. The conductivity ratio KR decreases as the clear domain conductivity k1 increases, and in this case more heat is carried from the heated wall to the channel. This causes an increase in the buoyancy-driving forces, which enhances the free convection. Bringing more heat from the boundary to the clear and porous domains causes an increase in their temperatures as shown in Fig. 8. Figures 9 and 10 show the effect of Ca on the temperature and velocity spatial distributions at different thermal diffusivity ratios R. Although R has significant effect on the channel hydrodynamic and thermal behavior, changing R does not cause any change in the effect of Ca on both the velocity and temperature distributions. Figure 11 shows the effect of Ca on the velocity spatial distribution at different values of Pr. It is clear from this figure that the effect of Pr number on the channel hydrodynamic behavior is insignificant at all values of Ca. Figure 12 shows the effect of the local inertial term on the hydrodynamic transient behavior within the clear domain. As is clear from this figure, the local inertial term has significant effect on the hydrodynamic transient behavior over the entire time domain.

CONCLUSION The importance of the macroscopic inertial term in transient free-convection problems in vertical channels partly filled with porous materials has been investigated. The effect of Da, KR, R, R, and Pr on the role of the macroscopic inertial term has been studied. It has been found that the local inertial term has insignificant effect on the channel hydrodynamic and thermal behavior for Da < 104, especially in the porous part of the channel. Also, the effect of the macroscopic inertial term is insignificant at large values of R, where R > 1. In general, R has insignificant effect on the channel behavior when R > 1. Changing KR, R, and Pr may change the channel behavior, but the effect of Ca on the channel behavior remains insignificant over the entire range of KR, R, and Pr. REFERENCES
Abu-Hijleh, B. A. and Al-Nimr, M. A., The effect of the local inertial term on the fluid flow in channels partially filled with porous material, Int. J. Heat Mass Transfer, vol. 4, pp. 1565 1572, 2001.

70
Al-Nimr, M. A., Fully developed free convection in open-ended vertical concentric porous annuli, Int. J. Heat Mass Transfer, vol. 38, pp. 112, 1995. Al-Nimr, M. A. and Darabseh, T., Analytical solution to transient laminar fully developed free convection in open-ended vertical channel embedded in porous media, Appl. Mech. Eng., vol. 2, no. 1, pp. 932, 1997. Al-Nimr, M. A. and Haddad, O. M., Fully developed free convection in open-ended vertical channels partially filled with porous material, J. Porous Media, vol. 2, no. 2, pp. 179189, 1999. Al-Nimr, M. A. and Masoud, S., Unsteady free convection flow over a vertical flat plate immersed in a porous medium, Fluid Dynam. Res., vol. 23, pp. 153160, 1998. Beckermann, C., Ramadhyani, S., and Viskanta, R., Natural convection flow and heat transfer between a fluid layer and a porous layer inside a rectangular enclosure, J. Heat Transfer, vol. 109, pp. 363370, 1987. Chang, W. and Chang, W., Mixed convection in a vertical parallel-plate channel partially filled with porous media of high permeability, Int. J. Heat Mass Transfer, vol. 39, pp. 1331 1342, 1996. Chen, F. and Chen, C. F., Onset of finger convection in horizontal porous layer underlying a fluid layer, J. Heat Transfer, vol. 110, pp. 403409, 1988. Kakac, S., Kilkis, B., Kulacki, F., and Arinc, F., Convective Heat and Mass Transfer in Porous Media, Kluwer, Dordrecht, The Netherlands, 1991. Kuznetsov, V., Analytical study of fluid flow and heat transfer during forced convection in a composite channel partly filled with a BrinkmanForchheimer porous medium, Flow, Turbulence and Combustion, vol. 60, pp. pp. 173192, 1998. Kuznetsov, V., Analytical studies of forced convection in partly porous configurations, in Handbook of Porous Media, K. Vafai (ed.), pp. 269312, Marcel Dekker, New York, 2000.

Khadrawi and Al-Nimr


Lundgren, T. S., Slow flow through stationary random beds and suspension of spheres, J. Fluid Mech., vol. 51, pp. 273299, 1972. Nield, D. A., The limitations of the BrinkmanForchheimer equation in modeling flow in a saturated porous medium and at an interface, Int. J. Heat Fluid Flow, vol. 12, no. 3, pp. 269272, 1991. Nishimaura, T., Takumi, T., Shiraishi, M., Kawamura, Y., and Ozoe, H., Numerical analysis of natural convection in a rectangular enclosure horizontally divided into fluid and porous region, Int. J. Heat Mass Transfer, vol. 29, pp. 889898, 1986. Poulikakos, D., Thermal instability in a horizontal fluid layer superposed on a heat generating porous bed, Numer. Heat Transfer, vol. 12, pp. 8399, 1987. Prasad, V. and Tian, Q., An experimental study of thermal convection in fluid-superposed porous layers heated from below, Proc. 9th Int. Heat Transfer Conf., Tel Aviv, Israel, 1990. Somerton, C. W. and Catton, I., On the thermal instability of superposed porous and fluid layers, J. Heat Transfer, vol. 104, pp. 160165, 1982. Taslim, M. E. and Narusawa, U., Thermal stability of horizontally superposed porous and fluid layers, J. Heat Transfer, vol. 111, pp. 357362, 1989. Tong, T. W. and Subramanian, E., Natural convection in rectangular enclosures partially filled with a porous medium, Int. J. Heat Fluid Flow, vol. 7, pp. 310, 1986. Tong, T., Faruque, M. A., Orangi, S., and Sathe, S. B., Experimental results for natural convection in vertical enclosures partly filled with a porous media, ASME HTD-Vol. 56, pp. 8593, 1986. Tzou, D. Y., Macro to Microscale Heat Transfer, The Lagging Behavior, Taylor & Francis, Philadelphia, 1997. Vafai, K. and Sarkar, S., Heat and mass transfer in partial enclosures, J. Thermophys., vol. 1, pp. 253259, 1987.

Journal of Porous Media 6(1), 7181 (2003)

1XPHULFDO 6WXG\ RI %RLOLQJ LQ DQ ,QFOLQHG 3RURXV /D\HU


Mustapha Najjari and Sassi Ben Nasrallah*
Laboratoire dEtudes des Systmes Thermiques et Energetiques, Ecole Nationale dIngenieurs de Monastir, Monastir 5019 Tunisia E-mail: sassi.bennasrallah@enim.rnu.tn ABSTRACT A numerical study, employing an enthalpic method, of boiling in porous layer with a discrete heating and crossed by a through flow is presented. The effects of porous medium permeability, inlet flow velocity and inclination angle are presented. It was found that boiling depends strongly on inclination angle. Profile of generated volume vapor and flow structure are obtained for 0o 180o. In the case of opposing vertical flow, mixed convection regime can be detected when evaporated volume reaches its maximum value. For weak inlet velocity and high permeability, a critical value of arises corresponding to a maximum value of evaporated volume.

71 Received November 28, 2001; Accepted April 1, 2002 Copyright 2002 Begell House, Inc.

72

Najjari and Ben Nasrallah

NOMENCLATURE
c D ex ev g h hfg H j J(s) k kr keff l L p pc qw s t T u x, y specific heat capacity (J/kg K1) capillary diffusion coefficient (m2/s) unit vector for the x coordinate evaporated volume fraction gravitational acceleration (m/s2) enthalpy (J/kg) latent heat of liquid-vapor phase change (J/kg) volumetric enthalpy (J/m3) diffusive mass flux (kg/m2s) capillary pressure function absolute permeability (m2) relative permeability effective thermal conductivity (W/mK) porous layer width (m) porous layer length (m) pressure (Pa) capillary pressure (Pa) heat flux density (W/m2) liquid saturation time (s) temperature (K) Darcian velocity vector (m/s) longitudinal and transversal coordinates Greek symbols thermal expansion coefficient (K1) convection correction factor diffusion coefficient (kg/ms) porosity of the porous medium dynamic viscosity (kg/ms) kinematic viscosity (m2/s) density (kg/m3) surface tension, N/m Subscripts eff in irr k l s sat v 0

effective inlet irreducible kinetic liquid phase solid phase saturated state vapor phase initial

Superscripts 0 reference

INTRODUCTION Analysis of boiling in porous media is motivated by applications related to drying process, to high quality insulation of bildings, to post-accident boiling of fluids in nuclear reactor debris, to heat transfer from buried nuclear wastes in geologic repositories, etc. Numerical and experimental studies dealing with this problem, in rectangular or cylindrical geometries, have been published during the last decades. Udell (1985) obtained theoretical and experimental results for boiling in a rectangular cavity filled with porous medium and heated on one of the wall sides. The experimental apparatus allows variable orientation with respect to gravity. Torrance et al. (1976, 1990a, 1990b) examined the case of a saturated porous bed heated from below and cooled from above. Two-dimensional numerical calculations were reported by Ribando and Torrance (1976). Critical values of heat flux for which free convection appears

were calculated for different boundary conditions and permeability values. This study was extended to the case of boiling (Ramesh and Torrance, 1990a, 1990b), where the authors presented linear stability diagrams for boiling in a square cavity with free convection. Wang and co-workers (Wang and Beckermann, 1993; Wang et al., 1994a, 1994b; Wang, 1997) proposed an enthalpic method (two-phase mixture model) to remove difficulties in numerical analysis of boiling in porous media related to the presence of moving and irregular interfaces between the single- and two-phase zones. If the gaseous phase is composed of the vapor of the liquid and another gas, the enthalpic method is not convenient. Such a problem was resolved by Topin and co-workers (Rahli et al., 1996; Topin et al., 1999) when studying drying with superheated steam. Boiling in a horizontal porous layer heated on the bottom and cooled by a forced liquid flow was studied by Wang (1997) and by Peterson and Chang (1997). The

Numerical Study of Boiling in Inclined Porous Layer imposed heat flux chosen by Wang (1997) is high enough that a dry zone occurs. Good agreement was found between numerical results predicted by mathematical formulation based on a two-phase mixture model and experimental observations. Peterson and Chang (1997) found that with a high-conductivity porous-channel, heat exchange is enhanced. Comparatively few works deal with boiling in porous media with mixed convection. Numerical investigations using the enthalpic method to examine boiling effects on descending mixed-convection flow through a porous layer with a finite wall heat source have been conducted by the present authors (Najjari and Ben Nasrallah, 2002) for a wide range of intrinsic porous medium permeabilities and inlet liquid velocities. As a continuing effort toward a complete understanding of boiling in porous media with mixed-convection effects, we consider in this note the influence of porous layer inclination. The mathematical formulation is based on the twophase mixture model and takes into account, simultaneously, buoyancy, forced-convection, and capillary effects. The set of equations is solved numerically by the finite-volume method. Results concerning the effect of inclination on flow and heat and mass transfer are presented and analyzed. MATHEMATICAL FORMULATION The geometric configuration and the coordinate system are illustrated in Fig. 1. A porous layer with a discrete heating source on one side is initially saturated by liquid injected with inlet velocity uin and at inlet temperature T0. The layer walls are impervious and inclined to the gravity direction at an angle . The density of heat flux qw is constant and high enough that the temperature of the liquid can reach the boiling temperature Tsat and thus a two-phase zone (nonsaturated zone) can occur. After vaporization starts in the heated zone, the porous medium consists of two regions: (1) a saturated liquid region where T < Tsat, and (2) a nonsaturated zone (T = Tsat), where the void space in the porous medium is filled by a mixture of liquid and vapor phases. Saturated and nonsaturated zones are separated by a vaporization front. In this study, basic assumptions employed, are * Local thermal equilibrium: Tl = Tv = Ts = T. * Individual phase velocity is given by Darcys law. * Heat transfer by radiation is neglected.

73 * The Boussinesq approximations are available. * The liquid-gas-solid zone is isothermal and is at 100oC. With these assumptions, the macroscopic governing equations, derived from the two-phase mixture model (Wang and Beckermann, 1993; Wang et al., 1994a; Wang, 1997; Najjari and Ben Nasrallah, 2002) and valid in all regions, are Continuity Equation for the Two-Phase Mixture where and are the density and velocity of the mixu

+ div u = 0 t

(1)

ture, given by l, v, and are the densities and the velocities of ul, uv
u = l u l + v u v
= l s + v (1 s)

(2) (3)

liquid and vapor phases, s is the liquid saturation denoting the ratio of liquid volume and the void-space volume in a representative elementary volume. The evaporated volume fraction of liquid is then obtained as:

ev =

1 L .l

0< y < l 0< x < L

(1 s) dx dy

Momentum Equation for the Two-Phase Mixture


k u = (grad p k g )

(4)

where

k = l [1 l (T T0 )]l + v [1 v (T Tsat )] v = [l s + v (1 s)] krl / l + krv / v

74

Najjari and Ben Nasrallah

x=0 y=0 impervious adiabatic qw

u=uin T=To x1 x2

y=l g impervious isotherm T=T0 y x

u= uin

impervious adiabatic x= L

Figure 1. Schematic diagram of physical problem.

l =

l l

l =

krl l (5)

where H = (h 2hvsat) and h = lshl + v(1 s)hv Here hl and hv are the enthalpies of the liquid and vapor phases; they are related to temperature by the relations hs = csT + h0 s hl = clT + h0 l hv = cvT + [(cl cv)Tsat + hfg] + h0 l (8) (9) (10)

grad p = l grad pl + (1 l) grad pv

Here l and v are the thermal coefficients of expansion, l and v are the viscosities of the liquid and vapor, is the viscosity of the mixture, krl and krv are the relative permeabilities of the liquid and vapor, pl and pv are the values of pressure in the liquid and vapor, and p is the mixture pressure. By replacing with its expression in (1), we obtain the u pressure equation:
k 2 k k + p grad p.grad + div k g = 0 t

(6)

where Tsat is the temperature of phase change, hfg is the latent heat of vaporization, and hvsat = (hv)T=Tsat. h0 and h0 are reference enthalpies. cs, cl, and cv are the s l specific heats of the solid, liquid, and vapor phases. The coefficients , h, h/, and f(s) are given by

Energy Equation A unified form of the energy conservation equations for solid, liquid, and vapor phases is given by the volumetric enthalpy H equation: H + div (h u H) t (7)

= + s cs (1 )

dT dH

h =

[ sl + v (1 s )][hvsat (1 + l ) hlsat l ] (2 hvsat hlsat ) sl + v (1 s)hvsat

h khfg = div grad H + div f(s) g v

l h fg h dT D + keff = dH l h fg + (l v ) hvsat

Numerical Study of Boiling in Inclined Porous Layer

75
H + 2 l hvsat l cl T = Tsat H + v hvsat Tsat + c v v

D=

krl krv k J ' ( s) l ( v / l ) krl + krv

H l (2 hvsat hlsat ) l (2 hvsat hlsat ) < H v hvsat v hvsat < H

f (s) =

krl krv / l krl / l + krv / v

= l v
Capillary pressure pc(s) is represented by the Leverett function (Wang, 1997)

(11)

pc = k

1/ 2

J ( s )

J(s) = 1.417(l s) 2.120(l s)2 + 1.263(1 s)3 The intrinsic permeability of the porous medium, k, can be evaluated as a function of porosity and particle diameter d by the KozeneyCarman equation (Monicard, 1975):

1 H + v hvsat s = l h fg + (l v ) hvsat 0

H l (2 hvsat hlsat ) l (2 hvsat hlsat ) < H v hvsat v hvsat < H

(12)

Then the expression of dT/dH is different in each zone: * In the liquid zone dT 1 = (H l (2hvsat hlsat)), dH lcl * In the vapor zone 1 dT (v hvsat < H), = dH v cv In the two-phase zone (l(2hvsat hlsat) < H v hvsat) <

k=

d2 3 180 (1 )2

The relative permeabilities of the liquid and vapor phases are chosen as a linear function of saturation: krl = s krv = (l s) These expressions are simplified forms of relative permeabilities. In fact, the liquid phase (gaseous phase) will be discontinuous for values of saturation under (over) the irreducible saturation of the liquid phase (gaseous phase). In these conditions, the relative permeabilities become null:

dT = 0 dH

The velocities of individual phases can be also calculated as follows:


l u l = l u + j
and

(13)

krl = 0 krv = 0

s < slirr s > svirr v u v = (1 l ) u j


(14)

Numerical resolution of Eqs. (4), (6), and (7) gives the values of volumetric enthalpy H, the pressure, and the velocity for the mixture. The temperature and liquid saturation can be deduced from the volumetric enthalpy H by (11) and (12):

where j is a mass diffusion flux:

j = l D grad s + f ( s )

k g v

(15)

Initial and boundary conditions are expressed as follows:

76 At the inlet (x = 0), the porous medium is saturated with liquid at constant and uniform velocity uin, and kept at a constant temperature T0: H = H0 = l(clT0 2hvsat)

Najjari and Ben Nasrallah p = lg sin y

H = H0 = l(clT0 2hvsat);

At t = 0, the porous medium is at constant temperature T0 and the liquid velocity is uniform: u ex H = H0 = l(clT0 2hvsat) = uin RESULTS AND DISCUSSION Thermophysical properties of the porous medium (0.1 m 1 m) are listed in Table 1. Initial temperature T0 and inlet temperature for the liquid are equal to 293 K. The density heat flux qw is constant. The heat source length is 0.1 m. The set of equations (4), (6), and (7) is numerically solved by the finite-volume method (Najjari and Ben Nasrallah, 2002). To validate our code, comparisons with existing results have showed good agreement (Najjari and Ben Nasrallah, 2002). As shown in Figs. 2 and 3, characteristics of flow and heat and mass transfer are affected by inclination angle of porous layer. In vertical positions ( = 0o and = 180o), unicellular flow is observed (Fig. 2). For = 0o, in the zone adjacent to the heated surface, flow induced by buoyancy forces opposes forced flow. The cells are located near the heat source. However, for = 180o, cells are deflected toward the cold wall. In this case, ascending forced flow assists buoyancy-induced flow near the heated wall. If increases from 0o, the buoyancy force in the longitudinal direction (x direction) decreases and recirculating flow is reduced. However, the buoyancy force in the transverse direction (y direction) increases and reaches its maximum value at = 90o. If exceeds a certain value, we can observe in Fig. 2 bicellular flow where two recirculating cells are located at the edges of heated surface. The sizes of recirculating cells are smaller than those in unicellular flow.

p = l uin + l g cos k x
= u e u m x At the outlet (x = L), the porous medium is saturated with liquid, in thermally developed conditions and at constant and uniform velocity uin (in all cases studied, the volume expansion with time due to vapor generation in the transient regime is negligible compared to the inlet mass flow rate):

H =0 x

p = l uin + l g cos k x

u = uin . e x

The face (y = 0) is impervious and adiabatic, except for the heated segment between x1 and x2 :

p = l g sin y
k h fg h H g sin = qw + f ( s) y v

for x1 x x2

k h fg h H g sin = 0 + f (s ) y v

for x < x1 and x > x2 The face (y = l) is impervious, saturated with liquid, and kept at constant temperature T0:

Table 1 Thermophysical property data for a watersteamglass bead system (Wang et al., 1994a)
, N/m 0.059 keff, W/mK 0.85 cl, J/kgK 4.178103 cs, J/kgK 8.79102 cv, J/kgK 1.55103 v, K1 3103 l, K1 5.23104 hfg J/kg 2.257106 l, kg/m3 957.9 s, kg/m3 2.645103 v, kg/m3 l, kg/ms v, kg/ms

0.35

0.598 4.473104 1.2105

Numerical Study of Boiling in Inclined Porous Layer

77

=0

=30

=60

=90

=120

=150

=180

Figure 2. Streamlines for the liquid phase, at various inclination angles, with k = 71011 m2, uin = 4.24106 m/s and qw = 3000 W/m2.

The effects of inclination angle on temperature field are illustrated in Fig. 3. The contours are closely related to the flow structure represented by streamlines. Then, for horizontal layers, the heated zone becomes nearly symmetric and trapped near the heat source. For inclined and vertical layers, the heated zone is more extended and is stretched

toward the inlet for opposing flow and toward the outlet for assisting flow. When approaches 90o, the flow due to the pair of counter recirculating cells is aided by forced flow at the center of the heated segment and then the boiling interface is moved considerably toward the cooled wall (Fig. 3, = 90o).

78

Najjari and Ben Nasrallah

=0

=30

=60

=90

=120

=150

=180

Figure 3. Temperature fields for various inclination angles with k = 71011 m2, uin = 4.24106 m/s and qw = 3000 W/m2.

When the porous layer is inclined from = 0o, the buoyancy force magnitude becomes smaller, recirculating flow is reduced, and evacuated heat from the heat source is diminished. Consequently, for high permeability or low inlet velocity, the volume of vapor is first increased with (Fig. 4). We can notice in Fig. 4 the existence of a critical inclination angle m depending on the values of inlet ve-

locity and permeability which correspond to a maximum value of the vapor volume. By increasing from this value the volume of vapor will decrease. In Fig. 5 we can see that inlet velocity vectors near the heated wall advance more and more toward the two-phase zone when is increased. If is greater than m (m = 26o), forced flow contacts the two-phase zone directly (Fig. 5) and exerts a cooling effect

Numerical Study of Boiling in Inclined Porous Layer

79

Figure 4. Volume vapor fraction for various inclination angles at different values of inlet velocity uin and absolute permeability k; qw = 3000 W/m2.

Figure 5. Liquid velocity vectors for various inclination angles with k = 71011 m2, uin = 4.24 106 m/s and qw = 3000 W/m2.

80

Najjari and Ben Nasrallah

Longitudinal 4E-5 component of liquid at point A


2E-5

0E+0

k= 1 e-11 m, uin= 4.24 e-6 m.s-1 k= 3 e-11 m, uin= 4.24 e-6 m.s-1 k=10 e-11 m, uin= 4.24 e-6 m.s-1 k=7 e-11 m, uin= 10.60 e-6 m.s-1 k=7 e-11 m, uin= 42.40 e-6 m.s-1

-2E-5 0 40

()

80

120

160

Figure 6. Longitudinal component of liquid velocity at the first attachment point (point A) of vaporization interface; for various inclination angles and different values of inlet.

on the heated zone. If continues to increase, evacuated heat from the two-phase zone by forced flow is enhanced and then evaporated volume is reduced. This inversion in the evaporated volume curve can be best demonstrated by studying the variation with of the longitudinal velocity component at the first attachment point between the vaporization interface and the heated wall (Fig. 6, point A). This component is negative for small values of , increases with , and becomes positive for m. At intermediate values of uin, similar behavior is demonstrated by the curve of evaporated volume as compared to the case of uin = 2.12106 m/s because forced flow still opposes recirculating flow at small values of (Fig. 6). However, the peak becomes much smaller and is displaced toward low values of . For small values of k or large values of uin, the longitudinal component of liquid velocity at the attachment point A is positive for all values of . This means that buoyancy effects are not important. Forced flow overrides recirculating flow and contacts the heat source directly. Thus evaporated volume decays with inclination (Fig. 4), its maximum value is reached for = 0o, and no peak is observed. The influence of inclination angle disappears gradually for high

values of , meaning that the two components of buoyancy forces are negligible. REFERENCES
Lai, F.-C., Prasad, V., and Kulacki, F. A., Aiding and opposing mixed convection in a vertical porous layer with a finite wall heat source, Int. J. Heat Mass Transfer, vol. 31, no. 5, pp. 10491061, 1988. Monicard, R., Caracteristiques des roches reservoirs Analyse des carottes, Societe des Editions Techniques, Paris, 1975. Najjari, M. and Ben Nasrallah, S., Etude de lebullition en convection mixte dans une couche poreuse verticale, Int. J. Thermal Sci., vol. 41, no. 9, pp. 913925, 2002. Peterson, G. P. and Chang, C. S., Heat transfer analysis and evaluation for two-phase flow in porous-channel heat sinks, Numer. Heat Transfer A, vol. 31, pp. 113130, 1997. Rahli, O., Topin, F., Tadrist, L., and Pantaloni, J., Analysis of heat transfer with liquid-vapor phase change in a forced-flow fluid moving through porous media, Int. J. Heat Mass Transfer, vol. 39, no. 18, pp. 39593975, 1996. Ramesh, P. S. and Torrance, K. E., Stability of boiling in porous media, Int. J. Heat Mass Transfer,vol. 33, pp. 18951908, 1990a. Ramesh, P. S. and Torrance, K. E., Numerical algorithm for problems involving boiling and natural convection in porous materials, Numer. Heat Transfer B, vol. 17, pp. 124, 1990b.

Numerical Study of Boiling in Inclined Porous Layer


Ribando, R. J. and Torrance, K. E., Natural convection in a porous medium: Effects of confinement, variable permeability, and thermal boundary conditions, ASME J. Heat Transfer, vol. 98, pp. 4248, 1976. Topin, F., Rahli, O., and Tadrist, L., Experimental and numerical analysis of drying of particles in superheated steam, J. Porous Media, vol. 2, no. 3, pp. 205229, 1999. Udell, K. S., Heat transfer in porous media considering phase change and capillarity-the heat pipe effect, Int. J. Heat Mass Transfer, vol. 28, no. 2, pp. 485495, 1985. Wang, C. Y. and Beckermann, C., A two-phase mixture model of liquid-gas flow and heat transfer in capillary porous mediaI: Formulation, Int. J. Heat Mass Transfer, vol. 36, no. 11, pp. 27472758, 1993.

81
Wang, C. Y., Beckermann, C., and Fan, C., Numerical study of boiling and natural convection in capillary porous media using the two-phase mixture model, Numer. Heat Transfer A, vol. 26, pp. 375398, 1994a. Wang, C. Y., Beckermann, C., and Fan, C., Transient natural convection and boiling in porous layer heated from below, 10th Int. Heat Transfer Conf., Brighton, UK, 1994b. Wang, C. Y., A fixed-grid numerical algorithm for two-phase flow and heat transfer in porous media, Numer. Heat Transfer B, vol. 31, pp. 85105, 1997.

You might also like