Exam 3: Soil Physics due: 12/12 A recent article (attached) analyzes Darcy’s law from angles not covered in lecture. Summarize the main point(s) and discuss the practical and theoretical implications of the material presented. Please type your answers in the following format: double space; with 1" margins on top, bottom, left and right; and12 point type font. 1 Environ. Sci. Technol. 2004, 38, 5895-5901 Examination of Darcy’s Law for Flow in Porous Media with Variable Porosity† W. G. GRAY* AND C. T. MILLER Department of Environmental Sciences and Engineering, University of North Carolina, Chapel Hill, North Carolina 27599-7431 Henry Darcy’s experimental studies in 1856 of saturated water flow through a homogeneous porous medium contained in a vertical column have provided the basis for the quantitative description of fluid flow in a wide variety of both natural and engineered porous medium environmental systems. Extrapolation of Darcy’s original observations and conclusions has led to several commonly applied equations used to model flow in porous media. This work examines this original experimental study, summarizes the appropriate mathematical expressions that ensue directly from the data, and indicates expressions in common use that are suggested, but not actually supported, by the data. The paradoxes that exist in the common approaches for the case of a porous medium with a spatially variable porosity are illustrated. A modified form of Darcy’s law, and also of the Hubbert potential, is derived based upon fundamental notions of averaging. The modified form of Darcy’s law derived here reduces to the conventional form for a homogeneous porous medium. Darcy’s Experiments The study of flow in porous media is important in a wide range of applications including groundwater flow, exploitation of petroleum reservoirs, filter design, manufacture of composites, capillary circulation, chemical reactors, etc. Efforts to quantitatively describe flow and transport processes in porous media date from the experimental studies of Henry Darcy in 1856 (1, 2). Darcy’s studies involved packing a vertical column with sand as depicted in Figure 1. Four different packings were used, each of different height, with the sand used in each packing differing primarily by the degree of washing employed. Manometers tapped reservoirs at each end of the column, and water was allowed to flow through the packing. The experiments involved steady, single-fluidphase (hereafter designated as single-phase) flow with water assumed to fill completely the pore space. Despite difficulties in obtaining a constant water source, the experiments demonstrated that the volumetric flow rate through the porous medium (Q) is proportional to the total head loss across the sand column (h2 - h1) and the cross-sectional flow area (A) and inversely proportional to the packed height of the column (L): h2 - h1 L Q ) KA † (1) This paper is part of the Walter J. Weber Jr. tribute issue. * Corresponding author phone: (919)966-3013; fax: (919)966-7911; e-mail: graywg@unc.edu. 10.1021/es049728w CCC: $27.50 Published on Web 10/09/2004 2004 American Chemical Society In this algebraic expression, K is referred to as the hydraulic conductivity, and it is a function of both the porous medium and the fluid properties. Darcy found that K was essentially constant for a particular packing that he employed. It is important to note that Darcy’s experiments actually provide no information about any properties within the packed column. All data was collected at locations external to the packed column, and Darcy’s algebraic expression provides effective information for the column as a whole. In eq 1, the hydraulic conductivity is characteristic of the column as a whole and provides no indication of the degree of homogeneity of the packing in the column. Neither the volumetric flow rate (Q) nor the volumetric flow rate per area (Q/A) provides any indication of the speed of the water flowing within the pores. The area (A) is a property of the column that was packed and does not necessarily relate to any effective cross-sectional area of flow. The length parameter (L) is the distance between the sampling points of the manometers and does not indicate the travel distance for fluid moving through the porous medium, which is influenced by the tortuous path created by the medium. Finally, even the head values (h1 and h2) are obtained from measurements taken in reservoirs outside of the column. The fact that Darcy’s law is expressed in terms of variables that may not describe the actual porous medium flow can be further demonstrated by building an experimental apparatus intrinsically different from that in Figure 1. Consider the sketch in Figure 2 where the visible, packed column of Darcy is replaced by an unusually shaped shell that houses some arbitrary packed, saturated flow system. For example, this system could be a spiral tube that connects the reservoirs; it could be a straight column with a small cross-section; it could be a bent tube or a tube of varying cross-section that follows some tortuous path between the reservoirs. In fact, any geometric shape is allowable that can be housed within, but need not completely fill, the shell depicted. The volumetric flow rate through the system could be measured as could the heads in the two reservoirs. Then, Darcy’s equation for saturated flow in this ill-defined region may be expressed as: Q ) Keff(h2 - h1) (2) Thus the single parameter (Keff) accounts for the effects of the permeability of the medium and the geometry of the flow area. Still, no information is obtained about the actual fluid velocity within the study system. Indeed, a revised definition of Keff can be used to express eq 2 as: q ) K′eff(h2 - h1) (3) by dividing by some cross-sectional area of the flow. However, the parameter q with dimensions of length/time provides no additional insight beyond that of eq 2 into actual flow velocities. The measured quantities in Darcy’s experiment are the volumetric flow rate and the difference in the heads in the reservoirs that bound the study region. These will be found to be proportional for conditions of slow flow. The coefficient of proportionality is constant over a range of specified head differences for a fixed system. In reporting his experimental results, Darcy was careful not to overstate their utility and implications. However, the need for scientists to model flow and transport in porous medium systems has led to general acceptance of equations with similarities to Darcy’s eq 1 with the state of the art reported in references such as refs 3-7. Despite their utility, VOL. 38, NO. 22, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 5895 The objectives of this work are (i) to examine commonly used expressions based upon Darcy’s law, (ii) to illustrate paradoxes inherent in some of these expressions for simple heterogeneous systems, and (iii) to provide modified forms of common expressions that apply to both homogeneous and heterogeneous systems. Considerations for Hydraulic Conductivity Because of the geometry of the original apparatus and the form of the equation put forth by Darcy, it is tempting to write a differential form in the limit of a very short column as: ( ) h 2 - h1 dh Q ) -K ) q ) - lim K Lf0 A L dL FIGURE 1. Darcy’s experimental apparatus. FIGURE 2. Alternative Darcy-like experimental apparatus (1). these differential equations cannot be justified based on Darcy’s experimental data. Indeed, if examined carefully, even the simplest differential form of Darcy’s equation for single-phase flow in a porous medium is highly restricted. Subsequent extensions to multiphase flow are even more problematic (e.g., ref 8). This situation has arisen because systematic procedures for derivation of equations for flow in porous media, such as averaging theory, began to develop approximately 100 years after Darcy’s experiments. The need to model porous media systems in the intervening years produced heuristic equations with variables and parameters that are not precisely related to measurements. However, our current theoretical understanding makes it possible to relate experimental measurements to equation variables, thus facilitating transfer of data between scales. 5896 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 22, 2004 (4) where a minus sign has been introduced because flow is in the direction of decreasing head (i.e., negative head gradient). Based on Darcy’s experimental data, however, there is no way to determine if the hydraulic conductivity, which is a characteristic of the full column length, retains its same value if one looks at a smaller segment of that column. Although Darcy carefully packed columns of different lengths, he had no way of examining flow behavior in smaller segments of a single packed column. His manometers provided information only from the reservoirs at the ends of the column, not from within the pore space. Based in Figure 2, which depicts an arbitrary apparatus, it is reasonable to project that some proportionality relation exists between the head drop and the flow rate for slow saturated flow. However, the proportionality parameter is expected to vary depending on the characteristics of the flow region under consideration. Thus, as has been widely recognized in experimental and mathematical analyses of porous medium systems, the hydraulic conductivity is generally a function of the length scale of the system being studied. For a column of constant cross-section with very uniform packing, K may not vary with the length of the column until L becomes small, on the order of a few pore diameters. However, below this length, the heterogeneities in the packing and inherent in the sand will cause the value of hydraulic conductivity to depend strongly on the orientation and shape of grains. It will vary so significantly from point to point and as a function of the length scale that modeling of averaged system behavior at this scale provides little information. The concept of a homogeneous differential length of porous medium is inherently contradictory. For the more general geometry of Figure 2, similar conceptual considerations apply. In addition to our inability to handle an arbitrarily small element of a medium, observe also that if eq 4 were to be applied to a general system, the value of K would depend on the location within the apparatus of the medium element being considered. Only for the case of a perfectly homogeneous packing would one expect K to be a constant for all equally sized elements of the porous medium. Efforts to pack columns for porous medium flow experiments have demonstrated the difficulty in obtaining a truly uniform packing. Although Darcy did use care in packing his columns, the chance of the medium truly being homogeneous is small. Thus the values of hydraulic conductivity he obtained are characteristic of the full columns that he examined, and reduction to the differential form of eq 4 with the value of K being independent of position cannot be justified mathematically. Length Scales The scale issues mentioned in the preceding discussion have been considered by porous medium scientists for decades (e.g., ref 3). The use of common length-scale notions can aid this discussion. This discussion is concerned with three different length scales: the microscale, the macroscale, and the megascale. Since designation of these scales takes on different meanings in different disciplines, we will expand on the meanings intended in this study. First, the microscalesor smallest length scale discusseds is a length scale over which the details of the pore morphology and topology are considered. At this scale the fundamental equations, such as the Navier-Stokes equations or Boltzmann’s equation, may be applied and solved (e.g., ref 9). In general, the level of detail needed to consider porous medium flow at the microscale is not available for non-ideal systems. So while theoretical work is increasingly relying upon descriptions at such scales, the field of porous medium science is often concerned with length scales considerably above that needed to completely define the geometry of the pore structure. The length scale at which Darcy’s law is typically written and applied in a differential form is termed the macroscale. The macroscale is a scale well above the microscale or pore scale, say on the order of 10 or more mean particle diameters. The porous medium field relies upon conservation equations developed, closed, and applied at this scale in terms of quantities that are averaged over this length scale. The minimum length scale for this approach to have meaning is often referred to as a representative elementary volume (REV) scale (3). Darcy’s experiments relied upon measures of flow and head loss external to the system and indicative of some averaged properties of the entire system. We will term this length scale a megascale, which is the largest of the three length scales discussed. The point made above through reference to Darcy’s work and the modified apparatus suggested by Figure 2 is that at the megascale only averaged quantities characteristic of the entire system are well defined, and great care must be taken to extend knowledge based upon observations of such systems to other systems. Subsequent efforts to extend Darcy’s work and produce a differential form of Darcy’s law that is applicable within a domain become paradoxical unless a point is interpreted as a macroscale average about an REV. Clearly the length scale of an REV is much larger than the microscale or pore scale, and the details of flow at the microscale are not resolved using Darcy’s law. In terms of the notion of an REV, a differential form of Darcy’s law is an acceptable notion, since taking a limit as the length scale tends toward zero merely implies evaluating the difference in averages over an REV as the distance between the centroids of equal-sized macroscale neighborhoods tends toward zero. However, care must be taken in deriving and defining the averaged quantities over an REV for use in any macroscale conservation equation. Differential macroscale conservation equations and their associated closure relations are used to describe transport phenomena in porous medium systems. For scientific and mathematical consistency, precise definition and measurability of quantities that appear in those equations is required. We show below that attention to achieving this consistency has been lacking historically even for the simple case of steady single-phase flow. We also derive a modified Darcy’s law expression that is physically and mathematically consistent. Considerations for Hydraulic Head The differential equation purportedly describing single-phase flow in a homogeneous column of porous medium is typically stated as: dh dL Q ) -KA (5) with K considered constant. Although this equation integrates to the Darcy eq 1 at steady state, this differential form contains an important complexity related to scale. The equation proposed by Darcy is in terms of quantities at the ends of the column (megascale) where no solid phase is present, whereas eq 5 is expressed in terms of quantities that must be defined within the porous medium (macroscale) (i.e., in regions where both fluid and solid are present). At the ends of the column, only fluid is present. However, at some points within the column, solid grains occupy the space such that a differential form for the fluid phase at that microscopic point would not be relevant. Thus, if one is to use a differential extension of Darcy’s equation at the macroscale, a definition of h within the porous medium at the macroscale that is precisely defined, measurable, and mathematically consistent across all scales is needed. In natural groundwater flow systems, measurements of hydraulic head are made in wells rather than in situ. Thus, as with Darcy’s experiments, the measurement is made outside the porous medium system such that the actual average value of h within the porous medium is not directly measured. In fact, if all measurements are made externally to the porous medium, the quantity h that pertains to the macroscale average within the porous medium and which appears in the differential equation need never be precisely related to any measured physical quantity. Additionally, the volumetric flow rate (Q) is not actually measured within a porous medium. It is inferred by the rate at which water is pumped from or injected into the medium. For the case of the column flow as described by eq 5, all of the fluid flow forms the average flux. However in natural systems, the flow within the pores is actually three-dimensional as it finds a path through the system. Despite the difficulties mentioned, the differential form of Darcy’s law has proven to be quite useful for modeling single-phase flow in a porous medium. For single-phase, three-dimensional flow through an isotropic medium, eq 5 is typically expressed as: k j - Fg) q ) - (3p µ (6) where q is the volumetric flow rate per unit area vector, p j is the pressure, F is the fluid density, g is the gravity vector, µ is the dynamic viscosity of the fluid, and k is the intrinsic permeability, a property of the solid medium with: K) kFg µ (7) where g is the magnitude of the gravity vector. The intrinsic permeability must be determined for each porous medium system and depends on the distribution of the sizes, shapes, and orientations of the pores that provide pathways for flow within the solid system (i.e., the morphology and topology of the pore space). In eq 6, the hydraulic head has been replaced by the pressure and gravity terms with: Fg3h ) 3p j - Fg (8) In fact, the use of head rather than pressure in eq 6 is only employed when the density is either assumed to be a constant or dependent only on pressure. As has been shown by Hubbert (10), eq 8 may be rearranged to an integral expression: ∫ h(x) h0 dh′ ) ∫ p(x) p0 dp′ + gF(p′) ∫ Φ(x) Φ0 dΦ′ g (9) where x is a position vector, Φ is the gravitational potential VOL. 38, NO. 22, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 5897 FIGURE 3. Homogeneous two-dimensional experimental apparatus. with g ) -3Φ. As written, the integration in eq 9 can be performed to obtain the value of h(x). However, if F is a function of any variables in addition to density, such as temperature or composition, the integral cannot be evaluated. In that instance, the more general form of Darcy’s law as given in eq 6 is employed; and evolution or conservation equations for the additional variables are also needed to model the system. Despite the fact that eq 6 is the equation of choice in most single-phase flow applications, it is interesting to note that the pressure appearing in the equation, a pressure existing within the porous medium, has not been well defined. It is some average characteristic of the medium associated with the macroscale point under consideration, but the precise definition has not been explicitly considered. The fact that Darcy’s experiments considered only the regions outside the porous medium for measurement means that those experiments do not provide any indication of the meaning of this pressure. However, if one is to derive point equations from a theoretical perspective, it is necessary to have quantities that appear in the equations be measurable. This problem will be demonstrated by an example. Considerations for Pressure The experimental apparatus depicted in Figure 3 is a horizontal rectangular column packed with a homogeneous isotropic sand. At equilibrium, the water levels in the two reservoirs at the ends of the column will be equal. Equation 1 is satisfied by this situation with Q ) 0 and h1 ) h2. Additionally, for this apparatus pressure gages are located, as indicated, at positions aligned with the vertical center of the column. The pressure readings of the two gages will be identical for conditions of no flow. Furthermore, at conditions of equilibrium, the average pressure calculated for the fluid within any volume that occupies the entire vertical cross section of the column will also be equal to the reading at the gages. This last assertion assumes that the length scale of the cross section is sufficient to serve as an REV. Under more general conditions in which flow may occur, the form of the Darcy equation for this experimental setup given by eq 6 in the x direction is: qx ) - j k dp µ dx (10) since the gravity vector in the x direction is zero. In this expression, p j is the average pressure in the fluid obtained by averaging the microscale pressure over the volume of fluid contained within a cross-sectional slice that constitutes a macroscale REV such that: p j) 1 Vf ∫ p dV Ωf (11) where Vf is the volume of fluid, and Ωf represents the averaging region. Note that p j can vary with x in that this volume-averaged pressure can be calculated at any x position 5898 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 22, 2004 FIGURE 4. Heterogeneous two-dimensional experimental apparatus. along the column. In fact for the case of a homogeneous medium and assuming a constant density fluid: p j (x) x B ) h1 + (h2 - h1) Fg L 2 [ ] (12) Equation 10 is satisfied at positions within the porous medium for this volume-averaged pressure. Furthermore, differentiation of eq 12 with respect to x and elimination of dp j /dx between eq 10 and 12 yields: qx ) - ( ) Fgk h2 - h1 µ L (13) which is consistent with Darcy’s law. Additionally, the gage pressures are known as functions of h according to: h1 ) p1 B Fg 2 (14a) h2 ) p2 B Fg 2 (14b) and: Substitution of these expressions into eq 13 yields a flow equation: qx ) - ( ) k p2 - p1 µ L (15) In this form, the difference between the gage pressures is the driving force for the flow through the column. This equation, which accurately describes the flow, has a character similar to Darcy’s expression in that the driving force and the flow per area describe the column as a whole at steady state but provides no information of the situation within the column. Now consider a second experiment using the column depicted in Figure 4. This column differs from that of the previous experiment by the presence of a homogeneous layer of constant thickness that may have porosity and permeability different from those in the previous experiment. By examination of the flow in the column under these conditions, some interesting information may be developed. Our discussion will be concerned with a model formulation that is macroscopic in the direction along the axis of the column but megascopic in the directions orthogonal to the axis. Thus, the model will contain gradients in properties along the flow direction but will be integrated over the flow cross-sectional area. From this perspective, issues of anisotropy that would require consideration using a fully macroscale formulation do not arise. The equilibrium configuration for this system will be the same as with the homogeneous packing with h1 ) h2, p1 ) p2, and Q ) 0. For purposes of illustration, we will work with pressure rather than head. Analysis of experiments performed on this packed column for slow flows will confirm the governing equation: qx ) - ( ) k* p2 - p1 µ L (16) The medium outside the layer is assumed to be the same medium that completely fills the domain in Figure 3. Therefore, k* will be different from k of the original system when the permeability of the added layer is different from that of the surrounding medium. For this setup, the value of k* will not be a function of position along the direction of flow as each cross-section contains the same fractional area of each of the two media. Thus, the reasoning that led to eq 10 for the single medium case would seem to apply here as well with the differential equation taking the form: qx ) - j k* dp µ dx (17) There is no need to employ a tensor form of k*, as would be appropriate for an anisotropic system, because of the scales used in formulation of this problem. Define ∆ ) 2 - 1 where 1 is the porosity outside the stripe and 2 is the porosity within the stripe. Also, let B be the height of the flow domain and b be the height of the stripe. Then at any position x along the column with p j defined by eq 11 for the constant density fluid case and assuming hydrostatic conditions, p j can be calculated as: [ ] b b 2x 11∆ p j B B L x B ) h1 + (h2 - h1) - 1 Fg L 2 b 1 + ∆ B (18) [ ] ( )( ( ) ) This equation is the same as eq 12 with the addition of a correction term that accounts for a different average pressure due to the porosity of the stripe being different from that of the surrounding medium. The reason for this correction term is that the nonconstant porosity assigns different weightings to the pressure variation in the vertical direction when calculating the average pressure over the fluid volume. Differentiation of eq 18 with respect to x and substitution into eq 17 yields: { Fgk* qx ) µ [ ]} b b 1h 2 - h1 B B B∆ L L b 1 + ∆ B ( ) ( ) (19) This equation differs from Darcy’s law as given by eq 13 by the presence of the last term that accounts for the difference in porosities of the two solid materials. We can observe that if the layer has only a different permeability from the original material but the same porosity, the form of Darcy’s law is preserved as ∆ ) 0. However, a material with a different porosity causes the equation to take a different form. In fact, we know that if h1 ) h2, there will be no flow in the system. However, eq 19 erroneously indicates that there will be a flow due to a nonconstant porosity. Note that this result is obtained even though the porosity of the medium obtained by integration over its cross section is not a function of position along the direction of flow. It is important to examine the factors that contributed to the nonphysical result of eq 19. The most obvious reaction is to state that the average pressure as provided by eq 18 is incorrect. Indeed, if the fact that the fluid occupies only a portion of the porous medium space is neglected and the average pressure is calculated by integration without regard to a porosity distribution, the correction term in the definition of the average porosity will disappear and eq 19 will reduce to the expected Darcian form. However, there are at least two problems with this approach. First, it is important to work with averaged quantities that are physically based. It is useful to be able to average physical measurements over regions of porous media to arrive at new quantities at a larger scale. Averaging while “pretending” that the solid phase is not present precludes such a systematic approach. Second, in the present example, not considering the presence of the solid phase in doing the averaging leads to the expected result because of the hydrostatic distribution in the vertical. For a general fully three-dimensional flow problem, such an assumption would not be appropriate. Thus, the mathematical formulation of the average pressure, as proposed in eq 11, is reasonable on physical and computational grounds. The use of this average pressure in a differential form of Darcy’s law is where the problem arises. Recall, that Darcy’s experiment and reported results involved measurements of quantities made outside the porous medium, at the megascale. These measurements do not capture information about the pressure distribution within the system. Recognition of the differences between measurements and mathematical forms leads to the resolution of the present paradox. Properly Formulated Equation When flow is in the horizontal direction, as in the examples considered here, it certainly seems reasonable to discard the gravitational term. If eq 6 is the equation that governs flow in a porous medium, then the vector component of this expression in the horizontal direction will not have the gravitational term. This leads to the conundrum found above for an inhomogeneous medium. In addition, for any porous medium, if we average Fg over the fluid volume, for the case when F is constant, the average will be equal to the point value, since g is also constant, regardless of the distribution of porosity. Thus, direct incorporation of gravity into the flow equation does not solve the problem. Consider the implications of using the gravitational potential, 3Φ, in the formulation instead of the gravitational vector. At the microscale, g ) -3Φ; but after transformation to the macroscale, it is important to make use of some averaged potential. The microscale gravitational potential is: Φ ) Φ0 + gz (20) where Φ0 is a reference constant gravitational potential, and z is the spatial vertical coordinate. Since the only spatial dependence of this microscale function is on the vertical coordinate, the derivative of this function in the net direction of flow in the example problem is zero: dΦ )0 dx (21) Following a similar procedure as that used to derive eq 18, we obtain the volume average of Φ over the fluid at a position along the x direction as: [ ] b b 2x 11∆ gB B B L Φ h ) Φ0 + 12 b 1 + ∆ B ( )( ) (22) When the porosity is not constant, a correction term appears that accounts for the unequal weighting of the gravitational potential through the vertical dimension. Furthermore, the VOL. 38, NO. 22, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 5899 gradient of this average potential has a nonzero component in the direction of flow with: [ ] b b 1 - ∆ B gB B dΦ h ) b dx L 1 + ∆ B ( ) ( (23) ) Although the microscale gravitational potential does not vary in the direction of flow, the macroscale potential will vary if the porosity is not constant. The correspondence between eq 23 and the correction term in eq 19 is directly observable. Now consider conditions in the reservoirs at the ends of the flow region. In a reservoir, the hydraulic head does not vary with depth. Furthermore, the hydraulic head may be expressed in terms of the pressure and the gravitational potential. In the reservoir: p ) Fg[h - (z - zb)] (24) Φ ) g (z - zb) (25) within the porous medium is speculative. Megascale data collected at the exterior of the system is being used to project macroscale behavior within the system. The generally employed differential extension of Darcy’s law given by eq 6 is limited in utility unless the averaged pressure can be defined. In cases where the porous medium is homogeneous in porosity, p j in eq 6 is simply the microscale pressure averaged over the fluid volume within the averaging region. In cases where the volume fraction of the fluid within the averaging region is not constant, inconsistencies may arise if this volume-averaged pressure is used in eq 6. When multiphase flow is considered, the distribution of volume fractions within an averaging volume could cause complications in presenting and applying a consistent flow equation. The calculations presented to this point have involved the case of a constant microscale density. If the density is not constant within the averaging volume, then the average gravitational potential needed is a mass weighted quantity with: where zb is a vertical reference location at the bottom of the domain. The head is then obtained as: h) 1 (p + FΦ) Fg (26) This expression for head is for the reservoirs. The appropriate difference form in terms of pressure that corresponds to Darcy’s equation, which was written in terms of head, is then: ( ) k* (p + FΦ)2 - (p + FΦ)1 qx ) µ L (27) as opposed to eq 15. Thus, the differential form written in terms of average quantities within the porous medium system will be: j + FΦ h) k* d(p qx ) µ dx ( ) Fgk* h2 - h1 µ L (31) and: 1 Vf Fj ) Vf With these definitions, the extension to the differential form of Darcy’s law is: k q ) - (3p j + Fj3Φ h) µ (32) Hubbert (10) derived the relation describing flow in a porous medium as: Fk 3H µ (33) where the Hubbert potential (H) is defined as: (29) Extensions The derivation here is intended to indicate the care that must be exercised in employing the differential form of Darcy’s law to model flow in porous media. There are several important observations that can be made. The experimental work of Darcy involving a porous medium made use only of data collected outside the column or at the megascale. Darcy’s law is an algebraic expression that provides a correlation between flow through the system and the measured heads at the ends of the system under study. The actual measured results of Darcy’s experiment are expressed by eq 2. The extension of this form to eq 1 followed by differentiation to obtain an expression at a point 9 ∫ F dV Vf (28) This equation applies at each x location within the porous medium and is consistent with the equation obtained for the system as a whole based on measurements outside the porous medium. Thus, the importance of properly incorporating the averaged gravitational potential into the governing equations is demonstrated. 5900 (30) q)- The presence of the averaged potential is a correction to eq 17. Now substitution of the expressions for p j and Φ h from eqs 18 and 22 simplifies eq 28 to: qx ) - ∫ FΦ dV ∫ F dV Vf Φ h ) ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 22, 2004 H) ∫ p(x) p0 dp′ + gz F(p′) (34) When F depends on pressure only, this expression may be differentiated to obtain: 3H ) 3p -g F (35) and a flow equation similar to eq 6 is recovered. However, if the pressure appearing in this equation is to be an average explicitly defined in terms of microscale quantities, the definition of the Hubbert potential should be: H h ) ∫ p j (x) p0 dp′ +Φ h Fj(p′) (36) This expression makes use of an average pressure as well as an average gravitational potential, both clearly defined in terms of microscale counterparts. The form of eq 36 is equivalent to replacing the microscale elevation coordinate (z) in eq 34 with the average value of this coordinate within the fluid in the averaging volume. When the pore space within the averaging region is distributed uniformly and the density is constant, the gravitational portions of eqs 34 and 36 are identical, and the pressure is that obtained from volume averaging of the microscale pressure. In many cases of flow in porous media when the variation in volume fraction is small, the corrections presented here may also be small. Nevertheless, they are important for ensuring that the fundamental flow equations being employed are consistent. If one wishes to invoke a simplifying assumption at some point in the analysis to make use of the traditional equations, this approximation may be made. In many instances, the approximation will be a good one. However, beginning an analysis with unrecognized assumptions built in can lead to faulty and inconsistent models. For porous medium problems in which change of scale is an important component of the modeling process, rigorously defining variables to allow for a consistent change of scales is an important consideration. One final disclaimer that may be important is to state that this analysis has not considered the case where temperature and/or concentration gradients would drive the flow. Although, in the general form provided as eq 32, Fj may depend on temperature or concentration, variations in these quantities are not deemed significant enough to provide potentials for flow. Further examination of this issue awaits a more detailed analysis than a visitation of the experimental results of Darcy and the use of his correlation as a starting point for a more general physical description. Acknowledgments The authors are pleased to submit this manuscript in recognition of the distinguished research career of Walter J. Weber, Jr. C.T.M. acknowledges the guidance of Professor Weber early in his career and his encouragement to do fundamental work; hopefully, this current collaborative work is some evidence that this message got through. This work was supported in part by the National Science Foundation (NSF) Grant DMS-0327896. Partial support of this work was also provided by NSF through DMS-0112069 to the Statistical and Applied Mathematical Sciences Institute in Research Triangle Park, where the initial ideas for this work were formulated. C.T.M. was also supported in part by Grant P42 ES05948 from The National Institute of Environmental Health Sciences. Literature Cited (1) Darcy, H. Les Fontaines Publiques de la Ville de Dijon; Dalmont: Paris, 1856. (2) Darcy, H. Determination of the laws of flow of water through sand. In Physical Hydrology; Freeze, R. A., Back, W., Eds.; Hutchinson Ross: Stroudsburg, PA, 1983. (3) Bear, J. Dynamics of Fluids in Porous Media; Elsevier: New York, 1972. (4) Bear, J. Hydraulics of Groundwater; McGraw-Hill: New York, 1979. (5) de Marsily, G. Quantitative Hydrogeology: Groundwater Hydrology for Engineers; Academic Press: San Diego, CA, 1986. (6) Domenico, P. A.; Schwartz, F. W. Physical and Chemical Hydrogeology; John Wiley and Sons: New York, 1998. (7) Freeze, R. A.; Cherry, J. A. Groundwater; Prentice-Hall: Englewood Cliffs, NJ, 1979. (8) Gray, W. G.; Hassanizadeh, S. M. Paradoxes and realities in unsaturated flow theory. Water Resour. Res. 1991, 27 (8), 18471854. (9) Pan, C.; Hilpert, M.; Miller, C. T. Pore-scale modeling of saturated permeabilities in random sphere packings. Phy. Rev. E 2001, 64 (6), 066702. (10) Hubbert, M. K. The Theory of Ground-Water Motion and Related Papers; Hafner Publishing Company: New York, 1969. Received for review February 20, 2004. Revised manuscript received July 27, 2004. Accepted September 1, 2004. ES049728W VOL. 38, NO. 22, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 5901