Rtvoleom brapneeving Sob Feservoiv ladle Roper ties Leston 2 - Els tvaglePaovs Meda Septamber 16, 1941flow Flow ta Fovevs Media O Te alelar |Lamlnas Flow in Piges ~ Flow we Bros Media by Drovetion | by hawinay, vsotheroh daw of tloids i pre, we eat |use the Prtevlle relation | us od? bp vo 2 aK where | av: average velocity, em sec”! de pipe olaneter, em dpe presare change, dyne ent (noises clo‘dhyeem # labuslhet pn) Axe bfferentiol Length, em — (th 2 them) Ae Hold vcority, 9 ent! 2" Upobe af 100 ep) A chttaram of Laminar Hew i thes reveals vee at wall deer a K— a —____ > | The velocity prble doring Lemivay tow 15 given as | ye Mp ref l= fle @ | eee OL (ET Theretore ¥,lv'sr)e0 (at the wall), The maiounn teloerte Nagy eccore at v'z0 (the center of the tobe) and this velocity is ' Umaee 4p rt @ Gace| \ O 146 dela the average velocity, a, fs or variable of interest. The average velocity across. the tube ts detined as errr We f a, ride! de Zn LE ride! de © LUT ET Ieee Z = tyr 4uar vr en ph | Ee eee Reta eas Zé Grats er 7 r | Be eee ae Arde Bee | gan ee | eae i LE z Jde | rmx a = o ri? hb do 4ruae 4 | Ve ap ris bp ad? @ oAhK Sen Ax which verifies &g, 1. The purpose of this exercise u's to doctrate the restrictions on hamtnar flow, In porticolar, be wlocity at the wall 's devo.i OTs dish | ae alisiay | mnoltiphypray 4g, 4 by the evess-sectlonal avea of the [fee tr?, we obtalv | ge vhs 2 hp mr? f 8 wax | = om hp rt () | 6 aan by | 4g vo Ac = 4 bp Ay &) | & aly wax by ded ree 4, where by ts a Condbetance" term, when the ube ce ® large, be us ae and the How apne ly of the jtvbe cs Large. Conversely, when by is small, the capacity V4 the tube ¢o How floide 1's redveed!. | der that we fave measured 4 and &) and that 1, bt, and lA are brown, Then we may vse fg 6 to compete by, the vorduotance of the pipe. kearrangiuey ba, 6 and solvlog. ee 4 gives | ye fade a | Ade oo tn the case of many meacvrements, we may wish to slightly remnauge Gb tito a peeing tunetion, This giver F= ke bp (s) Ae wax ypemn +h where yet jx 4) meh, boo i | | | pth ts of the form i | Ae he |4 erample plot of dada veiny %.8 as a model would Orme a)ela look Like | Z 45 0 ip ts the tepe ok the stale hie a | ud ” i = ’ | Ay of | of” awinay Flaw | Now-Lasunar ! a Flow j 0. 2 a ' aby le still have pot consideved pores medi, but by shdvetion, we ean write. the tlourag A ve ae, bp tb valid tor Lamirar tow in pipes | Ay atk THEN. Ve #2 k bp should be valid tor Lamar How — (4) i | | Ax abn Un por media. (rock) here kis detned as the perma lityy of the veeck, Ty He units the components of Gq,4 are expresied as a2 volumetric tlowmte, cm “see a i Aye cross-sectional area, em* | Ve average velocity, em sec”! | dpe pressure difference, atm me viscosity | ¢ ate lentil " stane, on| Ore theh \4ud the pormesby bby, £, ds un clareys, | — eni'set™' hep lem \ 1 cm? latm or as the stotemant reack on page To of the ABW text.200 Propuction PRACTICE ‘THE PERMEABILITY OF POROUS MEDIA TO LIQUIDS AND GASES} L. J. Kunxpnorre * aBsTRACT ‘The standard procedare for determining the permes- bility of porous media according to API Code No. 27 Girst edition, October 1985) is based on the fanda- mental assumption thet, 25 Tong as the rate of flow is Proportional to the pressure gradient, the permeability constant of a porous medium is « property of the medium, and is independent of the fluld used in its determination. Although this is true for most liquide, the permes- L Introduction 1 has teoane como prt in the lindsey to dstarmpe to pemosty of coe nner lk thy any the equine anally employed tor Bis de termination fe arrange to optte wth salt of She'samne tor meat aaper= pressor "he prac ie and on the fondant! eaempton that ar lng te Davey nw syns ay elon 2 the rat of dow Ie portion! foe presare radio the presi emctant of pores mates E's port f the med, sd ie ndopenten fe fli'wed tn sdtarnnation” Therefore, the merle cbtained by laboratory measures it ai are taken {2 be'tptoio tte haogenons ow of eter ol gua andegrooné reer {Po andar soap fat the permet ot 4 porous maam i inept of the Rd an Sctemisaen lari ty Mast wi few SHencrements of highly perms sandones Sa ie THowere, Muslat* giver stale of relis of mea- suranonta on the peeblty water and st of 6 omer oe nds, care’ ott ty Hane, Lavi nd Berne? showing large dept betwee Perma tr aod water~ioet aloe found fer Water tang lover than Yor ait Many con of ch reendes bemecn the pemedtiger oir tnd thee tolerant eter Hg to wate ceted Soringivengnins cated ot inthe lnoraes of = Fh ernest Pa ES ais leet tol sr cert het bility constant as determined with gates is dependent ‘upon the nature of the gas, and is approximately Unear function of the reciprocal mean pressure. ‘This effect can be explained by taking into account the phenomena of slip, which are related closely to the mean free paths of the gas molecules. The apparent permeability extrapolated to infinite pressure gives a permeability constant which ie a characteristic of the Porous mediam only. the Bataafsche Petroleum Maatschapp{j, Amsterdam, ‘The Netherlands, and of the Shell Development Comi- pany, Emeryville, Calif. In general it was found ‘that, ‘with highly permeable media, the differences be- ‘tween liguid and air permeabilities were emall, whereas ‘these differences were considerable for media of low permeability. ‘These discrepancies made it destrable to investigate the validity of the assumption that the permeability of @ porous medium is independent of the nature of the flaid with which the determination is carried out, ‘The investigation has shown that the permeability to a gas is a function of the mean free path of the gas molecules, and thus depends on factors which induence ‘the mean free path, such as the pressure, temperature, and the nature of the gas. ‘Therefore, when the mean, free paths are small, eg, at high pressures, the permeability to gas should be expected to approach that for liquids. ‘The experimental data which support these ‘conclusions are discussed hereinafter, IL. Permeability of Porous Media to Different Liquids 1 is, of coures, obvious that if a liguid reacts with ‘come constituent of the core material, eg, if waler ‘causes clay-containing core matarial ‘to ewell, then thee = hd qui se réduit A a=b tee quand /1, = 0, ou lorsque la pression sous le flere est égale A la pression atmosphérique. “Il est facile de déterminer Ia Joi de décroissance de la hauteur d'eau fr sur le filtre; en effet, soit dh Ia quantité dont cette hauteur s'abaisse’ pen- ant un temps di, sa vitesse dabaissement sera dh — SP ; mais Péquation ci-dessus donne encore pour cette vitesse Vexpression wtate mn aura done ccrooee suenaveusen 2.10357 Fic, 2—Grarus ComPiten From Darcy's Tanutan Data on His EXPERIMENTS OF Ocr. 29 ro Nov. 2, 1855, AND oF Fez, 17-18, 1856, SHowina Linear RE- LATION BETWEEN FLOW RATE AND DIFFERENCES 1™ Hercits oF Equivalent WATER MANOMETERS, ah_k = Ks eo aon Git ah oe Kae are a raeonc-£s th grit tetas 3 “Sila valeur correspond au temps 1, et hun temps quelconque 1, il viendra rate atte ~Aumad a “Si on remplace maintenant h + ¢ eth, + e pat » il viendra k Ig =Ig.- U8) @ et les deux équations (1) et (2) donnent, soit la loi abaissement de Ja hauteur sur le filtte, soit la loi de variation des volumes débités i’ partir du temps 1. ‘Si ket ¢ étaient inconnus, on voit qu'll faudrait ‘deux expériences préliminaires pour faire disparattre de a seconde le rapport inconnu Translating Darcy's statements into the notation which will subsequently be used in the present paper, what Darcy found and stated was that, when water flows vertically downward through a sand, the volume fof water O passing through the system in unit time is siven by ; oh 2 KAM orty eae H, «1 andthe vole emg it en it ie by hn hh eee : or by - KOA . . ) where K is factor of proportionality, 4 the area of cross section and I the thickness of the sand, and h, and hz the heights above a standard reference elevation of ‘water in equivalent water manometers terminated above and below the sand, respectively. ‘Writing Eq. 2 in differential form gives q=-K (dh/dl) . @ Soon after the publication of Darcy's account of these experiments, the relationship expressed by Eqs. 1 t0 3 became known, appropriately, as Darcy’s law. It has subsequently come to be universally acknowl. ‘edged that Darcy's law plays the same role in the theory fof the conduction of fluids through porous solids as ‘Ohm's law in the conduction of electricity, or of Four ier’s law in the conduction of heat. On the other hand, Darey's own statement of the law was in an empirical form which conveys no insight into the physics of the phenomenon. Consequently, during the succeeding cen- tury many separate attempts were made (o give the statement of the law a more general and physically sat- isfactory form, with the result that there appeared in the technical literature a great variety of expressions, ‘many mutually contradictory, but all credited directly or indirectly to Henry Darey. It has accordingly become recently the fashion, when some of these expressions have been found to be physically untenable, to attribute the error to Darcy himself. In fact one recent author, in discussing a sup- posed statement of Darcy's law which is valid for hori~ zontal flow only, has gone so far as to explain thatFic. 3 APPARATUS FOR VERIFYING Dancy’s Law FoR FLow in Various DIRECTIONS. Darcy was led to the commission of this error by re stricting his experiments to flow in a horizontal direc- tioa On this centennial occasion of Darcy's original pub- lication, it would appear to be fitting, therefore, i stead of merely paying our respects to Darcy in the form of an empty homage, that we first establish un- equivocally what Darcy himself did and said with respect to the relationship which bears his name; second, try {o ascertain the generality and physical content of the relationship and to give it a proper physical expression; and third, atiempt to see how this fits into a general field theory of the flow of fluids through porous solids in three-dimensional space. The first of these objectives has already been ac- complished; the second and third will now be given our attention, THE PHYSICAL CONTENT OF DARCY'S LAW As we have seen heretofore, what Darcy determined was that, when water flows vertically downward through «a sand, the relation of the volume of water crossing unit ‘area normal to the flow direction in unit time, to the thickness of the sand, and to the difference in heights of equivalent water manometers terminated above and below tne sand, is given by the following equation: I= ge KS where K is “a coefficient depending upon the per- rmeability of the sand.” Questions immediately arise regarding the generality of this result. Would it still be true if the water flowed ‘upward through the sand? or horizontally? What changes ‘would be effected in the relationship if some different liquid characterized by a different density and viscosity were used? In what manner does K depend upon the permeability of the sand, or upon its measurable sta~ tistical parameters such as coarseness and shape? And finally, what physical expression can be found which properly embodies all of these variables? The answer to most of these questions can be de- termined empirically by an extension of Darcy's original experiment. If, for example, we construct an apparatus such as that shown in Fig. 3, consisting of a movable cylinder with a rigid sand pack into which two man- org @ meters, at an axial distance apart, are connected by flexible rubber tubing, we can determine the validity of Darey’s law with respect to the ditection of flow. With the apparatus vertial and the low dowaward at a tolal rate, the manometer diffrence fh ~ J will have some fixed value a. Now, Keeping Q constant and inverting the column so thatthe flow will be vet= tically upward, it will be found that Af alo remains constant. Nest, sting the column horizoatal, af stil remains constant. In this manner we easly establish that Darcy's Taw is invariant with respect to the dire: tion of the flow in the earth's gravity fel, and that for a given Qh the flow rate Q zemains constant whether the flow be in the direction of gravity Or apposed to it or in any other direction in three-dimensional space. This leads immediately to a generalization for flow in threedimensional space. Atetch point in such space there must exist a particular value of a scalat quantity hy defined as the height above standard clevation datum of the water column in a manometer termiaated atthe given point. The ensemble of such values then aives rise to a scalar field in the quantity A with ts at Cendant family. of surfaces, = constant. Tn sucha sealar field water will flow in the direction perpen- dicular to the surfaces, f= constant, and ata Tate siven by a= ~Keradh -.@ Continuing our empirical experimentation, we fad that wen we change ether ofthe Nuid properties, den- sity or viscosity, or the geometrical properties of the sand, Eq, 4 stil remains valid but the value of K changes. In particular, by varying one factor ata time, we find Kap . reins} + & winrep isthe density and xis the visosity of the Aid, Likewise, if we use a number of geometrically similar sands which differ only in grain sie, we find that Ked, 6 where d is a length such a the mean grain diameter, Which characterizes the size scale of the pore sucture ofthe sand Introducing the results of Eqs. 5 and 6 into Eq. 4 then gives a= Kd (o/p)( ~ grad h), -.@ in which K’ is a new factor of proportionality contain: ing all other vaviables not hitherto explicitly evaluated This remains, however, an empirical equation devoid of ed” which, when the vector quantities are resolved into their components in the flow direction s, becomes kaNG = pot 8. — ep/0s’ Co It is found in ths manner that for randomly packed, uniform spheres of diameter d, the valve of the shape: factor Nis approximately 6% 10" Then, for a pack of uniform spheres of any size, the permeability will be sppronimately Fe (6x 104 & 1d, for diferent packs, is allowed to vary from about 10" to 10° em, corresponding tothe approximate range of grain sizes from fine sills to coarse sands, the pet taeabity will vary fom about 10°" to 10" en, which is alo approximately the range of the permeabiies of the corresponding clastic sediments. In view ofthe fact that magnitudes of permeabilities of rocks are remote from that of the square of sty nit of length in common use, there fs some advantage in having a. precical ‘unit’ such that most measured ‘lus fall Within the range 1~ 10,000 practical unit, Te such a practical unit isto fi into a consistent sys tem of measoremeat without awkward conversion fae- tors, then it must also be a submulple of the funda mesial unit ofthe form 1 practical unit = 10" fundamental units In the cgs system with the fundamental unit the (centimeter)’, the optimum value of the exponent ‘would be about 12, or 1 practical unit = 10 em’ Regrettably the unit of permeability used almost universally in the petroleum industry, for which the name “darcy” has been pre-empted, was defined origin- ally in terms both of an incomplete statement of Darey's pes as) and an inconsistent system of measurement. The per- meability & is defined to be 1 darcy when q = 1 (cm'/ cem')/sec, x = 1 op, and dp/as = 1 atmosphere/em. In the complete Darcy’s law of Eq. 18, the factor pee, except when the motion is horizontal, is of com- parable magnitude to @p/@s, and so cannot be ignored. Consequently, since it ig not practical to measure pe. in atmospheres/em, it follows that permeabilities, ex- pressed in dareys, cannot be used in a proper statement of Darcy's law without the insertion of a numerical fac~ tor to convert pge from cgs units into atmospheres/em. The only alternative is to convert permeabilities ex- pressed in darcys into the cgs unit, the cm’. For this conversion 1 darcy = 0.987 x 10* em’, which is within 1.3 per cent of the submultiple, 10* _——— 1) we Fic, 5—MetHon or Drrininc Pour VALUES oF MAcro- scopic QUANTITIES ILLUSTRATED WITH THE POROSITY j. Few permeability measurements are accurate to with in 1.3 per cent and, when several specimens from the same formation are measured, the scatter is much greater than this. Consequently, for all ordinary com- Putations, the approximate conversions: 1 darey == 10* em’, 1 md 10” cm’, are more accurate than the permeability data available; ‘though if the data warrant it, the more precise conver: sion can of course be used DERIVATION OF DARCY'S LAW FROM NAVIER-STOKES EQUATION Having thus achieved the desired generalization and 4 proper physical statement of Darcy's law by an ex- tension of the empirical method which Darcy himself employed, let us now see if the same result can be de- rived directly from the fundamental equation of Navier and Stokes for the motion of a viscous fluid Macroscopic anp Microscopic ScaLes In order to do this we must first distinguish between the two size scales, the macroscopic and the micro- scopic, on which the phenomena considered are 10 be viewed. The macroscopic scale, which is the one we have been using thus far, is a scale that is large as compared ‘with the grain or pore size of the porous solid. On this seale the flow of a fluid through a porous solid is seen as a continuous phenomenon in space. However, when we are dealing with macroscopic quantities which have particular values at each point in space, but which may vary with position, it is necessary for us to define more clearly what is meant by the value of a macro- scopic quantity at a given point, This can be illustrated with the concept of porosity Suppose that we are interested in the porosity at @ Particular point, About this point we take a finite volume element AV, which is large as compared with the grain or pore size of the rock. Within this volume element the average porosity is defined to be A av" where AV; is the pore volume within AV. We then allow AV to contract about the point P and note the value of f as AV diminishes. If we plot 7 as a function of AV (Fig. 5), it will approach smoothly a limiting value as AV diminishes until 4V approaches the grain (20) —_——‘or pore size of the solid. At this stage 7 will begin to vary erratically and will ultimately attain the value of either 1 or 0, depending upon whether P falls within the void or the solid space. However, if we extrapolate the smooth part of the curve of # vs AV to its limit as AV tends (0 zero, we shall obtain an unambiguous value of f at the point P. We thus define the value of the porosity f at the point P to be extrap lim AV, avs0 av" where “extrap lim” signifies the extrapolated limit as ob- tained in the manner just described. By an analogous operation the point value of any ‘other macroscopic quantity may be obtained, so that hereafter, when such quantities are being considered, their values will be understood to be defined in the foregoing manner, and we may state more simply: 1 4) = av — lim 5, 7 Ur OP) = sy B45), 2 or Lim ow) qi Beao, | where the quantity of interest is a function of a volume, an area, or a length, respectively. ‘The microscopic scale, on the contrary, is a scale commensurate with the grain or pore size of the solid, but still large as compared with molecular dimensions cer of the motional irregularities due to Brownian or molecular movements. Mieroscoric Equations oF Motion Let us next consider the steady, macroscopically reeti- linear flow of an incompressible fluid through @ porous solid which is macroscopically homogeneous and iso- tropic with respect to porosity and permeability. We shall then have the fluid flowing with a constant macro- scopic flow rate q under a constant impelling force per unit of mass E, and in virtue of the isotropy of the system, we shall have q=oE,. 2... 23) where a is an unknown scalar whose value we shall seck to determine. ‘Then, choosing x-, y-, and z-axes, aniatia tka, ieUe ee. } a and of, Ga, be (25) a= 08. where i, j, and k are unit vectors parallel to the x, y=, and z-axes, respectively, and the subscripts signify the corresponding scalar components of the vectors. Further, there will be no loss of generality, and our analysis will be somewhat simplified, if we choose the “x-axis in the macroscopic direction of flow. Then (26) Next, consider the microscopic flow through a macro- scopic volume element AV of sides Ax, Ay, and Az. ‘The void space in such an element will be seen to be an intricately branching, three-dimensional network of flow ‘channels, each of continuously varying cross section. fluid particle passing through such a system will foliow ‘2 continuously curving tortuous path. Moreover, the speed of the particle will alternately increase and de- Crease as the cross section of the channel through which it flows becomes larger or smaller. Such a particle will accordingly be seen to be in a continuous state of accel- eration with the acceleration vector free to assume any possible direction in space, Consider now the forces which act upon a small vol- lume element dV of this fluid. By Newton’s second law of motion dma =XdF , ey where dt is the mass of the fluid, athe acceleration, and SAF is the sum of all the forces acting upon the Mid contained within dV, There are many ways in which these forces may be resolved, but for present purposes it will be convenient to resolve them into a driving or impelling force dF, and a resistive force arising from the viscous resistance of the fluid element to deformation, dF... Eq, 27 then becomes dma ak, + dF, a (28) By the principle of D’Alembert we may also introduce a force dE, = — dm a, which is the inertial reaction of the mass dm to the acceleration a, and with this substitu tion Eq, 28 becomes dB, + dB, + dk, = 0 9) Of these forces, dF, which is imposed from without ‘and does not depend primarily upon the motion of the fivid, may be regarded as the independent variable. The forces dF, and dF, both owe their existences to the fluid motion, and their effect is to impede that motion. ‘The relation of the separate terms of Eq. 29 to the externally applied forces and the fluid motion are given by the equation of Navier and Stokes, which in vector Form may be written as follows: pg ~ (1/p) grad p] dV = p (Dy/Dr) av ~ niv'y + (1/3) vow) dP, in which vis the microscopic velocity, and p= — (1/3) (oy * 05 + a4) is the microscopic pressure fat a point. The a's are normal components of microscopic stress. ‘The expression Dv/D1 is the total derivative with re- spect to time of the velocity v, and is equal to the accel- eration a. This can be expanded into Go) ‘SimtLaR FLOW SysTEMs.in which the term @v/at signifying the rate of change of the velocity at a particular point is 2er0 for steady motion. The expression V-y, in the last term of Eq. 30, is the divergence of the velocity; and for an incompressible Bud this also is zero. Since the flow being considered is the steady motion of an incompressible fluid, Eq. 30 simplifies to lg ~ (I/p) grad p] dV = p (w-V¥) dV = nO) av ee BD ‘TRANSFORMATION FROM MICROSCOPIC TO Macroscopic Equations oF Morion Eq, 31 expresses the relation of the Muid velocity and its derivatives in a small microscopic volume element t0 the applied force dF, acting upon that element. If we could integrate the three terms of this equation with re- spect to the volume, over the macroscopic volume AV, and then convert the results into equivalent macroscopic variables, our problem would be solved, In fact the inte- gration of the first two terms presents no difficulty. The total driving force on the fluid content of the volume AV. as obtained by integrating the microscopie forces dF is: Ry = f dks = f (oe ~ grad p) dV av = (on ~Brad p) fav, 2) ‘where gf0d7 is the volumetric average of the microscopic rad p over the fluid volume fV- From the macroscopic equations, the driving foree Fy is given by: F.= p/AV E = pig ~ (1/5) grad pl fa¥- (33) ‘Then by combining Eqs. 32 and 33, f (oe = grad p) dV = (pg ~ Brady) faV fav = pfa¥ tg (i/o) grad p}, (34) from which itis seen that the mi to the volumetric average, grad p. Tategrating the inertial term: F=f dE =p frovdy YaV 0 f (uiufox + v du/ay TAY + wau/te) de dy de +5 pS (uavfax + vovjoy FAV. + wav/ez) de dy de + kp f (uow/ax + vow/ey {AV + waw/az) dx dy dz.| Since there is no net gain in velocity with macroscopic distanes, each separate integral ofthe expanded form of Eq, 35 is equal to zero, and we obtain forthe volume ele- ment AV, roscopic grad p is equal id p, of the microscopic | I + G5) | R= fdk=0 Ls GH ‘Then, in virtue of Eq. 36, SAB, = pf VAY =~ Ry 6 tv Ordinarily the evaluation of Svvay 1a would require a detailed consideration of the geometry of the void space through which the flow occurs and of the flow field within that space. This difficulty can be ‘circumvented, however, and the integral evaluated except for @ dimensionless factor of proportionality, provided the flow field is kinematically similar for different rates of flow. Carrems oF SmaaLaniry Consider two flow systems consisting of two geometri- cally similar porous solids through which two different fluids are flowing. The criterion of geometrical similarity is that if f,and f, are any corresponding lengths of the two systems, then for every pair of such lengths Bk = I = const G8) ‘The eriterion of kinematic similarity is that if v, and vs are the velocities at corresponding points in the two sys- tems, the two velocities must have the same direction and their magnitudes the ratio w/o, = ve = const Then, since the forees dF, and dF, acting upon a Hid clement are each determined by the velocities and the fluid density, or viscosity, and dF is determined by dF, and dF, ifthe fluid motions of the to systems are kine= matically similar, all corresponding forces will have the same directions ‘and their magnitudes the same ratio (Fig. 6) Thus (aF) (aE): Ey. GE GE Since only two of the three forces are independent, we red to consider the ratios of only the first two, and, by reciprocation, indicating that for each system the ratio of the inertial 10 the viscous force must be the same. (dF): G9) (40) From Eq. 31 CFD _ pe (WOW), a GF), ~ vw). a, oe as (WW)e dV = (de (a2y In Eq. 41 v-9v expands into the sum of a series of terms, each of the form u du/2x, which is a velocity squared divided by a length. In Eq. 42 V'v expands into a series of terms, each of the form 2'/2x", which is a veloc ity divided by the square of a length. Then, since the ratios of all corresponding velocities and of all corre- sponding lengths of the two systems are constant, we may choose any suitable velocity and any convenient length, We accordingly choose for the velocity the macro- scopic flow rate q whose dimensions are (L? L* T", or [LT", For the characteristic length we choose d, which ust be some convenient statistical length parameter of the microscopic geometry of the system. With these sub- stitutions the first two terms of Eq. 39 become (43)‘The dimensionless quantity (gd)/(u/p) is the Reyn- ‘olds number R of the system, which, as seri from its deri- vation, is a measure of the ratio of the inertial to the vis- cous forces of the system. Our criterion for kinematic similarity between the two systems thus reduces to the requirement that ae = (45) ‘Now let us specialize the two systems by making ds = dhs p= puts = By which is equivalent to requiring the same fluid to flow through the same porous solid at velocities, . and 4. Howover, when these values are substituted into Eq. 44, wwe obtain Ga indicating that, in general, when the same fiuid flows through a given porous solid at two different rates, the resulting flow fields cannot be kinematically similar. However, since dF, is proportional to q’ and dF to 4, then as q is decreased dF, diminishes much more rapidly than dF, Consequently there must be some limiting value of ¢ = 9°, orof R = Re, at and below which the inertial force dF, is so much less than the viscous force dF, that the effect of the former is negligible as compared with the latter. For flow in this domain we may then write a8, =~ ak, and the force ratios become (46) GP. whereby kinematical similarity is maintained forall rates of flow g < q*. InrecRaTion oF Viscous Forces With this result established let us now return to the integration of dF, over the volume element (AV. Sar, vudv + is var sav fav fav +k sf vwdV) « fav Ta virtue of the fact that, by our choice of axes, ay and 4, are both zer0 and there is no net flow in the )~ oF 2- direction, the lat two integrals to the vight re both zero, and Eq, 47 simplifies to ° dk, =in f Vuav, 5 co) fav fav From our earlier discussion, so Jong 2s the flow remains kinematically similar for different rates the quantity Ve, Which is a velocity divided by the square of a length, is related to the macroscopic parameters by Vu= alae), (a9) where a is a dimensionless constant of proportionality for the element dV but has a different value for each different element of volume. ‘Substituting Eq. 49 into Eq. 48 then gives Sdk. =~ ipta/d) faav fav tov mo Fp fav. (s0) where 1/N is the average value of « over /AV. Substituting this result into Eq. 37, we obtain lg ~ (1/e) sad pl fav = ~ f dk, = EE fav, a= Ne (/u) fg ~ (1/e) grad pl, . » (51) which is the derived Darcy's Iaw in the same form as Eq, 16 deduced earlier from empirical data. Discussion oF Darcy's Law ‘The direct derivation of Darcy’s law from fundamen- tal mechanies affords a further insight into the physics of the phenomena involved over what was obtainable from the earlier method of empirical experimentation. It bas ong been known empirically, for example, that Darcy’s law fails at sufficiently high rates of flow, or at a Reyn~ olds number, based on the mean grain diameter as the characteristic length, of the order of R= 1+ ‘At the same time one of the most common statements made about Darcy's law has been that it is a special case of Poiseuile’s law; and most efforts at its derivation have been based upon various models of capillary tubes for of pipes. It also has been known since the classical studies Of Osborne Reynolds’ in 1883 that Poiseuille’s law fails when the Row makes the transition from laminar { turbulent motion, so the conclusion most often reached as to the cause of the failure of Darey’s law has been that the motion has become turbulent From what we have seen, this represents a serious mis interpretation and lack of understanding of Darcy's law In the Darcy flow each particle moves along a contint- ously curvilinear path at a continuously varying speed, and hence with a continuously varying acceleration; in the Poiseuille flow each particle moves along a rectilinear path at constant velocity and zero acceleration. There fore, instead of Darcy's law being a special case of Poi seuille’s Jaw, the converse is true; Poiseuille’s law is in fact a very Special case of Darcy’s law, Another special cease of Darcy's law is the rectilinear flow between paral Tel plates. Consequently deductions concerning the Darcy-type flow made from the simpler Poiseuille flow are likely to be seriously misleading. The deduction that the Reynolds ‘number at which Darcy’ law fails is also the one at which turbulence begins is 2 case in point. We have seen that the ccause of the failure of Darey’s law is the distortion that results in the flowlines when the velocity is great enough that the inertial force becomes significant. This occurs at aa very slow creeping rate of flow which, for water, has the approximate value of gt =H = (1X 10%em' see") (1/d) , oa corresponding to R* = 1, when d is the mean grain diameter. ‘Thus when d = 10° cm Darcy's law fails at a low rate a= Lem/see. Since this represents the threshold at which the effects of inertial forces first become perceptible, and since tur- bulence is the result of inertial forces becoming predom- inant with respect to resistive forces, it would be inferred that the incidence of turbulence in the Darey flow would ‘occur at very much higher velocities or at very much higher Reynolds numbers than those for which linearity between the flow rate and the driving force ceases. That this ig in fact the case has been verified by visual obser- vations of the flow of water containing a dilute suspension of colloidal bentonite through a transparent cell contain ing cylindrical obstacles, This system, when observed in PETROLEUM TRANSACTIONS, AIME ESpolarized light, exhibits flow birefringence which is sta- tionary for steady laminar flow but highly oscillatory when the motion is turbulent. Observations of only mod- erate precision indicate that the incidence of turbulence occurs at a Reynolds number of the order of 600 or 700, or at a flow velocity of the order of several hundred times that at which Darcy’s law fails. EXAMINATION OF THE SuaPE-FactoR N In both procedures used thus far, the factor N’ has ‘emerged simply as a dimensionless factor of proportional- ity whose magnitude is a function of the statistical geo- ‘metrical shape of the void space through which the flow ‘occurs. For systems which are either identically or sta- tistically similar geometrically, N has the same value. Beyond this we have little idea of the manner in which N js related to the shape or of what its numerical magni: tude should be, except as may be determined by experi ment. Let us now see if the value of N, at least to within an order of magnitude, can be determined theoretically. Since we have already seen in Eq, 50 that 1/ = 3; where a is the factor of proportionality between the macroscopic quantity g/d? and the microscopic quantity ‘x,t fllows that in order fo determine the mognitade Of N we must fist determine that of the average value of ‘Vis, For this purpose, withthe uid incompressible, the rmacroscapic flow parallel to the a-anis, andthe inertial forces negligible, Only the "component of the Navier Stokes Ba 31, du, ou Oe oa. tpfos= — 0 (25 + 38 + SE), 52) needs to be considered. When this s integrated with re Speet tothe volume over the uid space 7.1 becomes Ga aprB av =| ST av Gee TTR ta i, ve + +S Fy ve dea | + (53) Of the three integrals to the right, the first, whieh rep- resents the expansion of the fluid in the x-direction, is zero; and from symmetry, the last two, both. being the integrals of derivatives with respect to axes at right angles to the flow, are equal to each other. Tn consequence, Ea, 53 is reduced to the simpler form: Cie ~ OPR) FAV = — nf (u/ay’) av — 2u FUP) f¥, in which Fu/3y" is the average value of°u/2)" throughout the macroscopic volume ‘element. Solving this for Fu/ay" then gives Gulay) = ~ 1/2») (pee PpPeA). - (54) Our probiem now reduces to one of attempting to determine the average value of 2'u/@y" for the system in terms of the kinematics of the flow itself. If we extend any line through the system parallel to the y-axis, this line will pass alternately through solid and fluid spaces. At each point on the line in the fluid space, there will be a particu- lar value of the x-component u of the velocity, which will be zero at each fluid-solid contact, but elsewhere will have finite, and usually positive, values giving some kind of a velocity profile across each fluid gap. If this profile for each gap could be determined, then we could also com: pute @'u/éy* at each point on the line and thereby deter- mine 7u7oy" for the line, which, in a homogeneous and Vor. 207, 1956 isotropic system, would also be the average value for a volume. To attempt to do this in detail would be a statistical undertaking beyond the scope of the present paper. AS a first approximation, however, we may simplify the prab- Tem by assuming: 1, That all the gaps are equal and of width 2%, where J is the average half-width of the actual gaps. 2. That through each gap the velocity profile sat- isfies the differential equation Pufay' = Fufey (55) 3. That the total discharge through the averaged gaps is the same as that through the actual aps. The velocity profile for the averaged gaps can then be obtained by integrating Eq. 55 with respect to y. Taking a local origin of coordinates at the middle of the gap, ‘and integrating Eq. 55 twice with respect to y, We obtain us Cy + Ay +B, (56) in which A and B are constants of integration, Then, supplying the boundary conditions, u = 0 when y A=0;B=CK"/2 Substituting these into Eq. 56, we obtain w= (C2) B= y) on ‘as the equation of the parabolic profile of the aver- aged velocity across the averaged gap. ‘The mean value, u, of w across this gap is given by y re T= aA way |e ao Then, replacing C by (1/2n) (pee Eqs. 55'and 54, we obtain eat Ua eral (59) ‘This can be converted into terms of the macroscopic velocity, q,, by noting that for a macroscopic length of line I normal to the flow direction ¥) dy = 8/3. (58) apfex) from ql = Wy, = WA) =aF, 2s + (60) where f is the porosity With this substitution Eq, 59 becomes ca [ge — (1/p) @p/ex} , . (61) or, more generally, BLL. tea) arade , . (62) ® which is a-statement of Darcy's law that is valid to the extent of the validity of the averaging approxima- tion used ia its derivation,it will be noted that the geometrical factor (4/6) ¥ represents the permeability, so that (4/6) P= NK, "6, . + (63) where Nr is the shape factor corresponding to X as the characteristic length of the system. Derersanation of X ‘The mean half gap-width X along a linear traverse ccan be determined in either of two Ways. The most ob- vious way is by direct observation by means of microme- ter measurements along rectilinear traverses across a plane section of the porous solid. Of greater theoretical interest, however, is an indirect method due to Corrsin* Instead of a line, let a rec tangular prism of cross-sectional area 8, where 8 can bbe made arbitrarily small, be passed through a porous solid which is macroscopically homogeneous and iso- tropic. This prism will pass alternately through solid segments and void segments. Let n be the number of each which is traversed per unit length. Then the num. ber of intersections with the solid surface per unit length will be 2n,.and if @ is the average area of the solid surface cut out by the prism at each intersection, the total area per unit length, dB, will be dp = 2a (64) If a parallel family of such prisms is made to fill all space, the number per unit area perpendicular to the axis of the prisms will be 1/8", and the solid surface in- tersected per unit volume will be B=2n (as) (6s) In addition, the otal length of the void spaces per unit length of line will be anX=f, n= f/QR) . . + + (66) Substituting the value of m from Eq. 66 into 65 and solving for X then gives fa Be Since f and f can be measured, the value of X could be determined if 2/8° were known. ‘The latter can be determined in the following man- xr: A homogeneous and isotropic distribution of the internal surface S, inside a macroscopic space, implies that if all equal elements dS of the surface were placed without rotation at the center of a reference sphere, their normals would intersect the sphere with a uni- form surface density. Then, with the normals fixed in ditection, if the surface elements were all moved equal radial distances outward, at some fixed radius they would coalesce to form the surface of a sphere. If the prisms of cross-sectional area 8, parallel to a given line, were then passed through this sphere, the aver- age value of a/8* would be x (on) and, when the summ: isphere, area of hemisphere _ 2s" a7 = 0/8 = ex of diametral plane —eP * (69) Introducing this result into Eq. 67 then gives X= 2/8. oe ~ (10) ‘A method for measuring has been descibed by rca ae Penal but far present purposes the date aad packed mom spheres fur whieh 6 can 9 Staph aa slice For sucha yc, ath n Mbnciee por wai volume, eee ee ~ a oar a = 6=p =f, 3a -/ am where d is the sphere diameter ‘Then, introducing this result into Eq, 70, we obtain for packs of uniform spheres .& (72 and =~he 13 vI=T ™) Comparison: wins ExerninestTat. DATA Tn order to compare the value of Ns, correspond ing to the characteristic length X, with Na, corresponds ing to the sphere diameter d, we must fist establish the relation between Nz and N,. This can be done by noting that, by definition, MAN =k, where k i the permeability ofthe system, Consequently Nr= Na (/®) (74) “Then, introducing the vale of / from Eq. 73 into Eq, 74, gives for a pack of uniform spheres 9p" F The value of Ny for wellrounded quart. sands, screened to nearly uniform sizes, has been determined by the authors former research, assistant, Terry. Con- ner. Using packs of different uniform sands with mean fain diameters ranging. from 1.37 X 10" to 7415 X TO" cm, Conner made seven independent. determina- tions of Ny. The average value oblained was N= 60x10", cs vith individual values falling within the range between 53X10" and 6.7 X 10% Conner did not determine the porosity, but the aver age value of the porosity of randomly packed, uniform spherical glass beads found by Brooks and Purcell’ was 0.37. Inserting this value of f, and Conners value of ‘Ne, into Eq. 75 then gives forthe equivalent experi- ‘mental value of NB Ne=262 N= 1STX IO? aT) Comparing this experimental result with the approxi- mate theoretical result of Eq. 63, it will be seen that [Nz (theoretical) ‘Ni (observed) Nr Nes - 5) (78) PETROLEUM TRANSACTIONS, IME STSince our object at the outset was merely to gain some insight into the nature of the shape-factor N ‘eceurring in Darcy's law, no particular concern is to be felt over the discrepancy in Eq. 78 between the ob- served and the theoretical values. All that this really indicates is that the system of averaging required should be better than the oversimplified one actually used. For ‘a more complete analysis, account needs to be taken of the frequency distribution of the half gap-width A, and also of the functional relation between @ and A, The fact that our approximate analysis yields a result in er~ ror by only a factor of 4 makes it appear promising that if account is taken of the variability of A and of WW as a function of A, much better approximations may be obtainable. Darcy's Law For COMPRESSIBLE FLUIDS Our analysis thus far has been restricted to the flow of incompressible fluids for which the divergence term, = (1/3) u VV ¥, could be eliminated from the Navier ‘Stokes equation, For the flow of a compressible fluid, this term must be retained, and with the flow parallel to the axis, 4%u udu, (te , tw ae f Sart at Hs * sl rosa f(GBe Be av 2p ((2/3) Fufor) + Fufoy av. (80) Here, 7/6 represents the gradient of the divergence of the velocity in the x-direction, or the rate of the fluid expansion, Should this term be of the same order of magnitude as 7u7oy", and if the flow of a gas through the porous system is otherwise similar to that of a liquid, then the viscous resistance to a gas should be greater than that for a liquid of the same viscosity. To compare the two terms Fu7av and Fujayr, it will be noted that each is of the form: velocity/(length)* We have already seen that Bufay = - 3upX = ~ al indicating that the magnitude of this term is determined by the fact that large variations of u in the y-direction take place within the width of a single pare, Compar- able variations of u in the x-direction, however, due to the expansion of the fluid, occur only in fairly large macroscopic distances. Consequently, we may write Fufik = va/ex = alt, where 1 is a macroscopic distance. The ratio of the two terms is accordingly Fae Tuy ‘Then, since > > 7X, it follows that the additional frictional drag caused by the divergence term is negli- sible, and this term may be deleted from the equation. We conclude, therefore, from this approximate analy- sis, that Darey’s Jaw in its differential form is the same for a gas as for a liquid, provided that the flow «en behavior of a gas in small pore spaces, other than ex- pansion, is similar to that of a liquid. It has been conclusively shown, however, by L. J Klinkenberg’ that the two flows are not similar, and that, im general, ky, the permeability to gas based on the ‘assumed validity of Darcy's law for gases, is not equal to ky, the permeability to Tiquids; and, in fact, is not even a constant, In the case of the flow of 2 liquid through small pores, the microscopic velocity v becomes zero at the fluid-solid boundary; for gas flow, on the contrary, there exists slong the boundary a zone of slippage of thickness 8, which is proportional to the length of the mean-free path of the molecules. Consequently the ‘gas velocity does not become zero at the boundaries, and the frictional resistance to the flow of gas is less than that for a liquid of the same viscosity and macro- scopic velocity Since 8 is proportional to the mean-free path, it is ‘also. approximately proportional to 1/p. Consequently when the gas permeability, ke, of @ given porous solid is determined with the same gas at a number of differ- ent mean pressures, the resulting values of ky, when plotted as a function of 1/p, give a curve wnich is ap- proximately linear with I/p. Moreover, different gases, having different mean-free paths, give curves of differ- tent slopes, The limiting value of ke as 1/p—> 0, or as becomes very large, is also equal to k;, the permea. bility obtained by means of a liquid (Fig. 7). In view of this fact it is clear that, in general, the flow of gases through porous solids is not in accord- ance with Darcy's law. However, from Klinkenberg's data, at pressures greater than about 20 atmospheres (2. 10° dynes/emr, or 300 psi), the value of ky differs from k, by less than 1 per cent. Therefore, since most cil and. gas reservoir pressures are much higher than this, it can be assumed that gases do obey Darcy's law under most reservoir conditions. FIELD EQUATIONS OF THE FLOW OF FLUIDS THROUGH POROUS SOLIDS ‘The establishment of Darcy's law provides a basis ‘upon which we may now consider the field equations that must be satisfied by the flow of fluids through porous solids in general, three-dimensional space. This 6. 7—VARIATION AS A FUNCTION OF 1/p OF THE APPAR~ ENT PERMEABILITY OF A GIVEN SoLtp 5 DETERMINED By THE DIFFERENT Gasts. THE VALUE OF 2.75 MD, as (1/p) Tens To 0, DIFFERS BUT SLIGHTLY FROM THAT oF 2.55 mp Optainen Usine « Ligup (arrer L. J KLINKENBERG, API Drill. and Prod. Prac., 1941) eeproblem is complicated, however, by the fact that the fivids considered may be either of constant or of var- iable density; that one or several different fluids, either intermixed or segregated into separate macroscopic spaces, may be present simultaneously; and that all de- agrees of saturation of the space considered are possible. ‘The problem of dealing with such cases becomes tractable when we recognize that the behavior of each different fluid can be treated separately. Thus, for 2 specified fluid, there will exist at each point in space capable of being occupied by that fiuid a macroscopic force intensity vector F, defined as the force per unit ‘mass that would act upon a macroscopic element of the fluid if placed at that point, In addition, if the fluid does occupy the space, its macroscopic flow rate at the given point will be indicated by the velocity vec- tor q, the volume of the fitid crossing unit area nor- ‘mal to the flow direction in unit time. We shall thus have for each fluid two superposed fields, a field of force and a field of flow, each inde- pendently determinable. The equations describing the properties of each of these fields, and their mutual in- terrelations, comprise the field equations of the system, these in turn, in conjunction with the boundary condi- tions, determine the nature of the flow. ‘The FieLp oF Force We have seen already that the force per unit mass is given by E=g-(1/p) grad p , . as) and the force per unit volume by H= B= pg grad p an Either of these force vectors could he used, and either is determinable from the other, but before choosing fone in preference to the other, let us first consider the propertics of their respective fields, of which the most important for present purposes is whether or not the field has a potential. To simplify our analysis we will make the approximations that = const , (92) “and for chemically homogeneous liquids under the range of temperatures and pressures normally encoun- tered in the earth to drillable depths, pacomt. . ao (83) For gases, on the other hand, we shall have an equa tion of state p=, 7) (84) where T is the absolute temperature. The vector F has a particular value at each point in space and the ensemble of such values comprises its vector field. The criterion of whether this field has a potential, that is to say, of whether E=~ grado, where @ is a scalar field, is whether the field Fis icto- tational, which can be determined from its curl. From Eq. 15, curl E = curl [g ~ (1/p) grad pl =v Xg~ 9X [(/o) Vel. As is well known, the gravity field is irotational even without the assumption that g = const,” so. that VX E=0. Also —¥ XIU/p) vp] = — VU1/p) x VP = Of) 9X vp ‘V(/e) X Ve, ‘Consequently curl B= ~ V(1/p) X Vp = Vp X VM/p) ce : (85) so that curl E=0 when Vp X V(1/p) =0 . (86) ‘Therefore, in onder for the field to be irrotational, and hence derivable from a scalar potential, itis necessary either that Vp = 0, corresponding to’ constant pres- sure, or V(1/p) 0, corresponding to constant den- sity, or else thatthe vectors Vp and V(L/p) be collinear, corresponding to a coincidence of the surfaces of equal density and equal pressure ‘The second of these three cases is satisfied by a liquid of constant density, and the third by a gas whase density is a function of the pressure only, such as oc- curs under either isothermal or adiabatic conditions For the general ease, however, of & gas for which p = i(p. T), and the surfaces of equal temperature do not coincide with those of equal pressure, then the sur faces p = const will also not coincide with the surfaces p= const and we shall have two intersecting families of surfaces, (1/p) = const, aid p = const, for which Eq. 85 applies ‘This is the condition corresponding to thermal con- section, and the fluid will have a convective circula- tion in the direction that will tend to bring the surfaces of equal density into coincidence with those of equal pressure, with the less dense fluid uppermost Hence, subject to the condition that either » = const, orp = fr), culE=OandE=—gnde. (s7) ‘The value of at any arbitrary point P in space (Fig. 8) is then obtained by FP BP) = O(P,)— f E. ds P, P = 9 — Fea Blas ?, a] z 2 mrt SF edet 5 ® o PP Ray = ar) tere fF, Pp? . (88) where the integral from P, to P is taken along any path 5. Then by setting &(P,) = 0 when z= 0 and ps = 1 atmosphere, we obtain ea wernt SP... (89) where p is now the gauge presstre, or the absolute pres- sure less 1 atmosphere. If the fuid is incompressible and chemically homogen- ‘cous, this reduces to the simpler form B(P) = az + p/p + (90) For this case, if a manometer is tapped into the sys- tem at the point P, the height i above the level z = 0, to which the liquid will rise, will be Wattle 8 ee ey ON)Fic, 8 Tae PoreNtiat & 4S A LINE INTEGRAL OF THE Fie oF Force from which it follows that gh = ez + p/p=% ee) in agreement with our earlier definition of & in Eq. 14 If the fluid is incompressible and chemically inho- mogeneous, as in the case of water of variable salinity, the density p will not be a function of pressure only, and, in general, surfaces of constant density will not be parallel to surfaces of constant pressure. For such 2 system curl E70, and no potential exists, We thus see that, with the exception of cases of ther- mal convection, and of inhomogeneous liquids of var- iable density, the fields E for both Fiquids and gases are irrotational and are derivable from a potential 4. Since E is a force per unit of mass, then @ is an energy per unit of mass, and represents the work required to transport the given fluid by a frictionless process along a prescribed (p, T)-path from a standard position and state (0 that of the point considered. Surfaces = const fare accordingly equipotential surfaces, or surfaces of constant energy of position, and the fluid will tend to flow from higher to lower potentials or energy levels. ‘The field of force per unit volume, H,, can be dis- posed of summarily. Since H= pg — grad p, then curl H = VX (pg) — 9 X Up Ve Xe pv XE— VX VP But, since VX Vp and YX are each zero, then curl H = Vp Xg. (93) ‘This is zero only when p is constant or when the sur: faces of constant density are horizontal. The last con- dition never occurs except when the fluid is at rest or ‘when the motion is vertical. Hence, for motion in any direction other than vertical, the field of the vector H does not have a potential except when the density of the fluid is constant. For the special case of constant density, H=-gndt, . os (4) T= p@=pertp.. - - - . - - (95) is the energy per unit volume of the fluid at any given point. ° In view of the fact that the field Eas a potential for both liquids and gases under the conditions specified above, Whereas, in general, the field H has a potential where only for the special case of liquids of constant density, then there is no advantage in using the Tatter in pref erence to the former, and henceforth it shall be dropped from further consideration, ‘The generality of the field of force, as herein de- fined, merits attention. The force vector E for any gi fluid not only has values in space occupied by that fluid, ‘but also in any space capable of being occupied by the fluid, At a point in air, for example, the force Ey for water would be Ey =g~ (pe) arad p, and since, in ait, grad p is py. g, then b= Pat “The field F for a given fluid thus extends throughout all space of continuous permeability. When several fluids are to be considered, then at each point in space there will be a different value of E, for each separate fluid, given by: E,=g-(1/p) grad , ] E,=g — (1/p.) grad p , | on E.=g~ (1/p.) grad p . | The vectors F for the separate fluids of different den- sities will differ among themselves, both in magnitude and direction, but will ll fallin the same vertical plane, that defined by g and — grad p (Fig. 9) Similarly the potentials of diferent fluids at the same point will be: at f (dp/p) » It (96) ®, = 2+ Sldp/o) ) = 92+ f (dp/pn) ) where pn pps ate the variable densities of the sep- arate fluids. Since the equipotential surfaces for the separate fluids must be normal to the respective vectors B, then it fol- Tows that the equipotential surfaces of different fluids passing through a given point will not be parallel to fone another, although they will all intersect along a line normal to the (g, — grad p)-plane, Tue Fito or Frow We have already defined the flow vector q for a space which is entirely filled with a single fluid. For a space which is incompletely filled with a single fluid, ‘or is occupied by two or more intermixed fluids, then there will be two or more superposed flow fields, not jn general in the same direetion, and a separate value of 4 for each separate fluid. ‘The principal condition which must be satisfied by the field of flow independently of the field of force is, that it must be in accord with the principle of the con- servation of mass. Thus, if a closed surface, 5, fixed with respect to the porous solid, is inscribed within the field of flow, then the total net outward mass flux of any given fluid in unit time will be equal to the diminu- tion of the mass of that fluid enclosed by S, assuming ‘that processes which create or consume the given fluid are forbidden, “This condition is expresed by SS paedS = ~Amjn sy .. . . . (99) s S—where q. the outwarddirected normal component of ‘dand mthe mass enclosed (Fig. 10) In the ease of «space completely saturated with the siven fia, by dividing the integral (eq. 99) by the folume and then leting V tend to zor, we Obtain im 1 1 en yoo ig m= 7a (100) Which is the rate of loss of mass per unit macroscopic volume at a given point. Then, since ‘am/et = f¥ Gps) , Eq, 100 becomes div pg = ~ F(@e/2t) , aot) which is the so-called “equation of continuity” of the flow. If the motion is steady, then dp/2t is zero, and the equation simplifies to 0 (102) When this is expanded it becomes div pq = V pq = Ve.q+ pV.q=9, and when p is constant over space, corresponding to div pa = div 0 the flow of a homogeneous liquid, Vp = 0, and diva =0 7 + (403) _ grad p Pe Fic. 9—Force Vectors E at tit Same Pont Corre: SPONDING TO FLUIDS OF DIFFERENT DENSITIES. RELATION BETWEEN FIELD OF FORCE anp FIELD oF FLow In_a space completely saturated by 2 single fluid, the field of flow and the field of force are linked to- gether by Darcy's law q=or, Se es C08) which, in those eases for which curl E = 0, becomes a= ~ cand e. ++ 105) If the solid is isotropic with respect to permeability, the conductivity o is a scalar and the flowlines and the lines of force will coincide; if the solid is anisotropic, « will be a tensor and E and q will then differ somewhat in direction except when parallel to the principal axes of the tensor. Limiting our discussion to isotropic systems, by (ak- ing the curl of Eq. 105, we obtain cul q = — UX eV =~ Ve X TH ad X VO Then, since the last term to the right is zero, this be curl q=— Vex UD, (106) which is zero only when @ is constant throughout the field of flow. Therefore, in general, cul g #0 (07) and this circumstance precludes the derivation of the flow field from an assumed velocity potential, for, with the exception of the flow of a fluid of constant density and viscosity in a space of constant permeability, n0 such function exists The flow of a given Mid through a porous solid incompletely saturated with that Auld is equivalent to flow through a solid of reduced permeability, because the space available to the flow diminishes as the sa tration decreases. For saturations greater tan some caiied minim valve, the low obeys Darey’s law subject to the petmeabiiy having this reduced, Or 20- called, “relative-permeabilty” value, ‘Thus, with two interspersed but immiscible Mids in the seme macro- ee = on Ey = on fg — (1/p.) grad p] y tack mente yep ard fe 208) wihere on aad on are the relative condvetiities of the two fois, Tt will be noted that, except for vertical mation, F, and are not parallel. Consequently the two force filds and flow fields will be, im general, transverse to one another: Pioneer work on relative permeability as a function Of saturation was done on the single fluid, water, by L. A. Richards’ in 1931. Subsequently studies of the simultaneous flow of two or more fuids were initiated by Wyckoff and Botset,” and by Hassler, Rice, and Leeman’ in 1936. Since that time many other such stud- jes for the systems water-oil-gas have been published One flaw which has been common to most of these ‘multifluid experiments has been that the experimental arrangements and their interpretation were usually based upon the premise that the flowlines of the various components are all parallel and im the ditectionFic. 10—Mass Dischaxcr Turoucn « Fixe CLosei SURFACE $ AS THE BASIS FoR THE EQuaTion ‘oF Continutry. — grad p. Since this is far from truc, there is some ques- tion of the degree of reliability of the results of such experiments, Even so, the existing evidence indicates that for saturations greater than some critical minimum the flow obeys Darcy's Jaw in the form given in Eqs. 108. For saturations less than this limit, there should still be a general drift of discontinuous fluid elements in the direction of the field E, but with, as yet, no well- defined relationship between q and E. “The migration of petroleum and natural gas, through an otherwise water-saturated underground environmen from an initial state of high dispersion to final positions of concentration and entrapment, constitutes an exam ple of the latter kind, PROPERTIES OF THE COMBINED FiELD oF FLow Although the present paper does not permit of their elaboration, practically everything that is now known concerning the flow of fluids in a porous-solid, three- dimensional space is deducible from the foregoing field equations. The equations for the flow of an incom- pressible fluid are of the same form as those for the steady conduction of electricity. Consequently the well- known solutions to the electrical equations can be ap- plied directly to analogous situations in fluid flow. The tunsteady flow of a compressible fuid, when the force field is irrotational, is closely analogous, although some- \what more complex, to the unsteady conduction of heat, and the solutions of the heat equations can be adapted with some modification (© the analogous fluid-flow problems. ‘The fluid phenomena which are unique and have no counterpart in other more familiar field theory are those involving multiple fluids. If we consider the flow of two immiscible fluids of unequal densities, such as water and oil, interspersed in the same macroscopic space, we have seen from Eq. 97 that, in general, the force fields E, and E, of the two fluids will lie in the same vertical plane, but will have divergent directions, that for the less dense fluid being upward with respect to that for the more dense, As a consequence the two fluids will drift in these respective directions,, and, in response to suitable impermeable barriers, will tend to become completely segregated. Once this segregation is achieved, for any further steady motion of either of the fuids, the interface will appear as an impermenble berrie. Across this in- terface, neglecting minor pressure differences due to capillarty, the pressure in the two fluids must be the same. Then, in case neither of the fds is in motion, the interface will have t be horizontal, with the less dense fluid uppecmost, since this isthe only inctination fof the surface for which the presses on opposite sides can be the same ‘When either or both of the fluids isin motion, how: ever, in # nonvertical direction, the vectors (1/m) aed p and (1/3) g12d p will be inclined from the vertical, and the corresponding equipressure surfaces will be ine clined from the horizontal by the same amounts, At the interface every equipressure surface in one system must match that of the same value in the other. [At the same time, the flowlines in each system must be parallel to the interface so that the (B,, E,)-plane ‘must be tangent to the interface. Consequently the in- terfacial surface and the flow patters,in the wo sys- tems must mutslly adjust themselves until these eondi- tions are simultaneously satisfied before a steady state of flow becomes possible Of particular interest is the special case of this gen- ‘eral situation wherein one of the two fluids is in motion land the other is completely static. In this case the equipressure surfaces in the static fluid are horizontal and surfaces having equal difference of pressure are ‘equally spaced; whereas, in the moving fluid the equipres- sure surfaces are inclined downward in the direction of the horizontal component of the flow. Their lines of inter- section at the interface must accordingly be horizontal. Also, since the vector, grad p, lies in the same vertical plane as the vector E of the flowing fluid, it follows that the horizontal component of the fiow direction ‘must be parallel to the direction of steepest slope of the interface. If we consider a vertical plane parallel to the flow direction and perpendicular to the interface, then in this plane the surfaces of constant pressure will appear 3s lines of constant pressure, inclined in the flowing fluid and refracting into the horizontal across the interface in the static Mluid. Let the flowing fluid be denoted by the subseript 1 and the static fluid by the subscript 2. Then along the interface in the direction of the flow (pies), = Gp/as (109) In the static fluid (epfes), = - pgsin®, . (10) and consequently depends only upon the density p: and the angle of slope 6 of the interface, where @ is posi- tive when the slope is upward in the flow direction. Tn the flowing fluid, since E= ~ (1/p) (@p/2s)s» then (@p/Ps). = pas ~ pE ’ ur =—pesing + p.Jenaa).f OM) ——Equating 110 and 111 then gives pag sind =pgsin® — p [grad |, which, when solved for sin 0, becomes 1m sing = 2. made]... (112) 2 lerad ©] (ii2) = IE the less dense uid is flowing (Fig. 11a), ps > px the term to the right will be positive, and the interface will tit upward; if the more dense fuid is lowing (Fig. 11b), p, > ps the term to the right will be negative, and the interface will tilt downward Eq. 112, derived earlier by Hubbert," is the funda mental equation pertaining to fresh-water ~ salt-water relations along shore lines, and to oil-water and oil-gas interfacial relations such as water or gas coning during oil and gas exploitation, It is also the basic equation governing the underground positions of oil and gas en- teapment, If the water is static, accumulations of pe- troleum or of natural gas will occur beneath downwardly concave impermeable barriers, with the oil-water or gas-water interface horizontal. If the water is in mo- tion, as often is the case, the oil-water or gas-Wwater in- terface will be inclined, and oil or gas may be trapped beneath structures completely unclosed in the hydro: static sense. RESUME In paying our respects to M, Henry Darcy on this centennial occasion, we stated at the outset that it should be our endeavor: 1. To show unequivocally what Darcy himself did and stated with respect to the relationship which now bears his name, and to give his results a more general, Dut still equivalent, physical formulation. 2. To derive Darcy's law directly from the funda- mental Navier-Stokes equation of motion of viscous fiuids, 3. To develop, in at least their primitive forms, the principal field equations of the flow of fluids through porous solids. These objectives have now been accomplished, and the result is that, despite a number of (roublesome complexities such as those arising from thermal con- vection and from waters of variable salinity, the field theory of the flow of underground fluids is capable of ‘being brought into the same kind of a comprehensive unification as that already achieved for the more fa- miliar phenomena of electrical and thermal conduction, In closing it is pertinent to reiterate that Darcy's ‘empirical formulation: qa Ku = hy, which, as we have seen is valid for flow of a homogen- cous liquid in any direction, is physically equivalent to the expressions: = ~ (NE) (p/p) gerad hy a= + (NE) (9/1 = (A/p) grad 7] From these it follows that fluids can and do flow in any er Fig. 11—Inrerraces BeTweeN FLowin axp Staric Fiuws oF Ungquat Denstries. (a) Urwarb Tit IN THE Flow Dinecrion WuEN Less Dense Fiuio Is Frowinc. (6) Downwaxo Tit Wuen More Dense Fu Flows. direction whatever with respect to that of the pressure gradient, So far as we have seen no error of any kind with re- spect to the flow of fluids through porous solids can be attributed to Henry Darcy; and the errors which have been alleged by various authors during recent years appear on closer inspection to have been those com- ‘mitted by the authors themselves, ACKNOWLEDGMENTS Since this paper is the result of some 20 years of intermittent reflection upon the phenomena encompassed by Darcy's law, the author is indebted to many people whose oral or writlen discussions of various aspects of the problems involved have contributed to his own understanding of them. He is particularly indebted, how- ever, to his current and recent colleagues who have been of direct assistance in the present study. These in- clude: Jerry Conner who made an extensive series of experiments verifying the author's earlier theoretical deductions; David G. Willis who, in addition to reading critically the manuscript and assisting in the design of the illustrations, has been of invaluable assistance in the clarification of many difficult points; R. L. Chuoke and A. S. Ginzbarg who gave important mathematical assist- ance; and R. H, Nanz who made an extensive series of ‘measurements of the gap-widths along linear traverses through randomly packed uniform spheres. Any errors, however, are the author's own.REFERENCES 1. Darey, Henry: Les Fontaines Publiques de la Ville de Dijon, Victor Dalmont, Paris (1856). 2. Wyckoff, R. D., Botset, H. G., Muskat, M., and Reed, D. W.: “The Measurement of the Permeatil ity of Porous Media for Homogeneous Fluids, Rey. Sci. Instruments (1933), 4, 394, 3. A.P.L Code No. 27, “Standard Procedure for De termining Permeability of Porous Media” (Tenta- tive), Ist Edition (Oct, 1935), American Petro- eum Institut. 4, Fancher, G. H., Lewis, J. A., and Bames, K. B. “Some Physical Characteristics of Oil Sands, Pennsylvania State College Mineral Industries Ex- periment Station Bulletin 12 (1933), 65. 5. Reynolds, Osborne: “An Experimental Investiga- tion of the Circumstances Which Determine Whether the Motion of Water Shall Be Direct or Sinuous and of the Law of Resistance in Parallel Channels,” Philos. Trans. Royal Soc. (1883), 174, 935; or Papers on Mechanical and Physical Subjects, University Press, Cambridge (4901), Vol. Ht, 51. London 13. Corrsin, Stanley: “A Measure of the Area of @ ‘Homogeneous Random Surface in Space,” Quart. Appl. Math, (1954-1955), 12, 404, Brooks, C. §., and Purcell, W. R.; “Surface Area Measurements on Sedimentary Rocks, AIME (1952), 195, 289. Klinkenberg, L. J.: “The Permeability of Porous Media to Liquids and Gases,” API Drill. and Prod. Prac. 1941 (1942), 200, Richards, L. A.: “Capillary Conduction of Liquids ‘Through Porous Mediums,” Physics (1931), 1, 318. : Wyckoff, R. D., and Botset, H. G.: “The Flow of Gas-Liquid Mixtures. Through Unconsolidated Sands,” Phystes (1936), 7, 325. Trans. |. Hassler, Gerald L., Rice, Raymond R., and Lee- man, Erwin H. sions on the Recovery of il from Sandstones by Gas Drive,” Trans. AIME, (1936), 118, 116. Hubbert, M. King: “The Theory of Ground-Water Motion,” Jeur. Geol. (1940), 48, 785, Hubbert, M. King: “Entrapment of Petroleum Under Hydrodynamic Conditions,” Buil. AAPG (1953), 37, 1954. see LS EEESEEESEnsees a!Material from: Muskat, M.: The Flow of Homogencous Fluids Through Porous Media, YHRDC, Boston (1982).2.2. The Range of Validity of Darcy’s Law.—In order to under- stand more clearly the general nature of the “Jaw of flow” with which to represent the data of flow experiments such as those of Darcy, it is well to consider first the implications of the theory of dimensions. Applying the well-known rules of this theory, it is readily found that the pressure drop Ap over a column of sand of lengths As, carrying a fluid of density y and viscosity # with an average velocity », should be related to these variables by a relation of the form Ap = const. = 1 7p = const. - Tl" ——* a (a where the unknown functions / and ¢ are to be determined by the experiments, and d is a length characterizing cither the size of the pore openings or the size of the sand grains. Now it is clear physically that the function ¢ should be simply the first power of its argument, and indeed this appears to be the only clement of this analysis upon which there has been universal agreement among the numerous investigators of Darcy’s law. This observa- tion reduces Eq. (1) to the form Ap _ #2 (doy SB = const. Er a), (2) where now the left-hand side represents the pressure gradient in the linear system. 1 Due to the constancy of yw over the length of a linear system in a condi- a Oy tion of steady-state flow, it is clear that the pressure gradient Zz in the case of liquid flow will be uniform along such a system regardless of the nature of the function F, and hence character of the flow, and similarly for iP in case the fluid is a gas.‘The argument dvy/u = R, of the function F, is a well-known term in the ordinary hydrodynamics and in engineering hydraulics in the discussion of the fiow of fluids through sand-free pipes.? With d representing the diameter of the pipe, the term doy/p is known as the “Reynolds number,” and, as indicated by Eq. (2)— which must also apply to the flow through sand-free vessels as it is nothing more than a dimensional equation—determines the character of the flow. Thus, in particular, for low velocities, fluid densities, or pipe diameters, the function F is found to simply equal its argument, so that (3) ‘As this result is the same as that given by the classical hydro- dynamics for viscous fluids, where it is known as Poiseuille’s law, the constant coefficient having the value 32, and v represent- ing the average velocity over the seetion of the tube,? Eq. (3) is usually denoted as representing the pressure gradients in linear “viscous flow.” 21 has also more recently been applied with a generalized interpretation to problems in aerodynamics. 2 One may note here again the fundamental difference between the flow through « sand-free tube and what will be denoted here as the “ viscous” flow—governed by Darcy’s law—through a tube filled with a porous medium. For, while in the former the velocity traverse is essentially parabolic across the section (exaetly so in the special case of a circular tube), falling from = maximum at the center to zero at the walls, the macroscopic velocity in a linear porous medium is uniform over the cross section. Thus, whereas for Poiseuille flow the total flux is proportional to the square of the cross- sectional area, that in a linear porous medium is proportional only to the first nawer of the area, The reason for this difference evidently lies in the iscous tremendous surface exposed in a porous medium and its uniform dissemina- tion throughout the medium, thus giving in @ rough sense the equivalent of an enormous number of parallel capillaries, the fluid in each having the same average velocity, although within each there are undoubtedly micro- scopic velocity traverses similar to those in sand-free capillaries.For larger values of d, v, y, or 1/u, or more specifically as the Reynolds number dvy/y increases so as to exceed a critical value— of the order of 2,000—the nature of the flow in sand-free pipes suddenly changes from a smooth streamline character to one permeated by an irregular and fluctuating distribution of eddies The motion is then termed as “turbulent,” the transition in the direction of cither increasing or decreasing velocity (the most convenient parameter for continuous variation) being qitite sharp, although the cycle of increasing and decreasing velocities across the transition region usually shows some hysteresis. This type of flow is characterized dynamically by the fact that the function F is now proportional to the square! of its argument—in particular the velocity—so that Eq. (2) takes the form 2 Ay

It should be noted that Eqs. (5) and (6) are, in principle, consistent with the dimensional requirements of Eq. (2), provided the constants a and b are adjusted to absorb the factor 4?/yd* and the powers of dy/x left over on separating out v from the Reynolds number. In fact, as far as dimensional theory is concerned, F can be represented by any functional form including even such as have infinite power-series expansions. On the other hand, from a physical point of view it must be remembered that @ choice of F carries with it an implication as to the variation of 2 with the other vari- ables d,y, and ». ‘Thus, in particular, in the case of Eq. (6), values of n exceeding 2 would, by Eq. 2), imply that S2 would deerease with increasing viscosity, which clearly is physically unreasonable.‘69 spressune ORO? TC SUENGTH OF CORE 7 Truvo oensiTY vivewery ‘Teste oF eLow ‘CROSS-SECTIGN AREA = kaso WiscosiTe SYMBOLS a rc waren} "mrSe™ aR OL, WcLav0 cas.us euneay oF MINES or 10 0s 0 000 aad v Fic. 8.—Friction-factor chart for the flow of fluids through sands. (4fter Fancher, Lewis, and Barnes, Bull. Pa. State Expt. Sta.)Figure 8, published by Fancher, Lewis, and Barnes, gives 2 representation of the flow data for a number of consolidated and unconsolidated sands in the form of a “friction-factor chart.” Here the quantity ¢ = dAp/2Lye, which is dimensionally equivalent to the “friction factor” (¢f. footnote on page 58) used extensively in the study of the fluid flow in pipes, is taken for the ordinates. The abscissas are the Reynolds numbers dvy/y, the d here and in ¢ being the “average grain diameter”? defined by the equation 3 - r Oe mi zn.” wy) where d, is the arithmetic mean of the openings in any two con- secutive sieves of the Tyler or U. S. Standard series, and n, is the number of grains of the diameter d,, as found by a sieve analysis. Regardless of the physical significance of this definition of d, the important implication of Fig. 8 is that up to a Reynolds number (as defined above) of the order of 1, the data strietly obey the relation log ¢ = a — log R, (9) from which it follows at once that AP _ ZB = const. » (10) in accordance with Darcy’s law. ? Physically, of course, the term d should represent the average pore rather than grain diameter. However, as the former can be directly meas- ured only by microscopic examination of a cross section of the porous medium itself, all attempts to define or use a value of d to enter into the Reynolds number have referred to the averages of the actual grain diameters.While this law admittedly loses its strict validity and the flow becomes partially or completely turbulent! as the Reynolds number or velocity is increased, there appears to be no unique Ap eaae for large values of the latter. About all that appears to be definitely established is that the empirical data can be expressed in all cases by an equation of the form of Eq. (5), with n having a value in the neighborhood of 2. modification which is assumed by the relation between. and v 1 It may be noted that Lindquist and Nemenyi (¢f. infra, and also Forel heimer, loc. cit., p. 66) attribute the deviation from the condition of viscous flow, as the velocities are increased, to the rise in importance of the inertial forces as compared to the viscous forces, rather than to a real turbulence. This view, however, seems open to question, since there is no net kinetic energy gained in a linear channel, so that the pressure drop is consumed only in supplying the friction losses. Furthermore, as pointed out by Fancher, Lewis, and Barnes (loc. cit.), the turbulent character of the flow can be readily. demonstrated by injecting a stream of fluorescein solution into the main current of liquid flowing through a column of shot. At low velocities the dye passes around the shot in streamers with but little diffusion, whereas as a certain velocity is exceeded the streamers begin to bounce from one shot to another in a chaotic manner and become completely diffused. Finally, as Nemenyi has already observed, the fact that the Reynolds numbers at which the deviations appear are so much lower than those at which turbulence develops in sand-free vessels is in itself not surprising when account is taken of the fact that the actual velocities in the pores of porous medium are higher by a factor of 4-8 than the macroscopic velocities *, and that the flow actually takes place in channels of sharply varying cross section, a condition which, as is well known, ix conducive to the early establishment of turbulence. It, therefore, is not unlikely that the devia tions from Darcy’s law appearing at the low Reynolds numbers of the order of 1-10 (cf. Figs. 8, and 9) are manifestations of a real turbulence dissemi- nated through the flow system.149 900°T 88 HOHE) &1 04 9¥ LOyUS JO AyfeoDsIA !AyfSo>eIA do T JO ping v OF JoJor eyHUN ano} eA 1 E01 x FORT 1-01 X soge's | 0rx sone'@ |-01 x 426r6 | Lor x B6ez's sores'0 - Goa , : 1 = QUOD ayaa, 230°6 1 01x orer9 — |eorx re01's | s-01 x onset | 01 x 96g6'9 e1s0L Seon ne OLX F9T'T set t 01 X £009" SHEL) OE X SHOT cor X sey [= CRASHES D ses A 0.02) OH T 18) 45-008 wl X ctes't 201 X 0086'T sere 1 oor] OFX vEFeT OLX £908T - Gederee yar) eis 008 01 X 610 rer zope | 01 X 9x09 | ror x s80r'6 | LOT X e062'6 - Giewaree OX ¥9eNT 01 X 6986'T 01 X 9282°6 Was] Or X 9060"T 1 wrx zerot |= i (uo/ ane) 9 “999 start | sor X 18161 HOF X toes — | OFX eveE'L |e-0r x yoL0T| Lor X zoo8's 1 - oe foxep ‘oxy 40 |ov/ orm Deas we |cwoy one wo « 9°20 w/e ae | yur) ee] wo/ Go oa4p) oe] (OL wE) arvoe oun (O.00) FHIAA 1] O03) OF FT (0.08) FHT aT ut cor -o10p ° SUN] AUTaVEWusg UOL aTAV], NOISHBANOD—'T BITaVy,Material from: Bird R.B., Stewart, W.E., and Lightfoot, E.N.: Transport Phenomena, Wiley, Inc., New York (1960).§2.3 FLOW THROUGH A CIRCULAR TUBE The flow of fluids in circular tubes is encountered frequently in physics, chemistry, biology, and engineering. The laminar flow of fluids in circular tubes may be analyzed by means of the momentum balance described in §2.1. The only new feature introduced here is the use of cylindrical coordinates, which are the natural coordinates to describe positions in a circular pipe. We consider then the steady laminar flow of a fluid of constant density p ina “very long” tube of length L and radius R; we specify that the tube be “very long” because we want to assume that there are no “end effects”; that is, we ignore the fact that at the tube entrance and exit the flow will not necessarily be parallel everywhere to the tube surface. ‘Momentum in by flow Pressure py SS SST | Momentury flow in |: and out by viscous transfer 4 th Shell of thickness “f= Ar over which ‘momentum balance is made Momentum out by flow Pressure p, Fig. 23-1. Cylindrical shell of fluid over which momentum balance is made to get the velocity profile and the Hagen-Poiseuille formula for the volume rate of flow.We select as our system a cylindrical shell of thickness Ar and length L (see Fig. 2.3-1), and we begin by listing the various contributions to the momentum balance in the z-direction: rate of momentum in across cylindrical surface at r Qarlr Je (23-1) rate of momentum out across cylindrical surface at r+ Ar Qarlr, Jet ar (23-2) tate of momentum in across annular surface at z = 0 Qnrdr 0,\(prJlena 23-3) rate of momentum out across annular surface at z = L Qnrdr 1 \(podlenr (23-4) gravity force acting on cylindrical (QarArL)pg (23-5) shell pressure force acting on annular (QarAr)py (2.3-6) surface at z = 0 pressure force acting on annular —(Qnrdr)p, (23-7) surface at z= L Note once again that we take “in” and “out” to be in the positive direction of the axes. We now add up the contributions to the momentum balance: QarLr,,)|, — Qarkt Near + ar Arp! luo = Qnr Arpt), + 2nr ArLpg + 2nr Ar(py — p,) =O (2.3-8)Because the fluid is assumed to be incompressible, v, is the same at 2 = Zz and L, hence the third and fourth terms cancel one another, We now divide Eq. 2.3-8 by 2nL Ar and take the limit as Ar goes to zero; this gives ((Codlesas = (r7,2)], lim Aro (252 + va) r (23-9) ‘The expression on the left side is the definition of the first derivative. Hence Eq. 2.3-9 may be written as (2.3-10) (23-11) The constant C, must be zero if the momentum flux is not to be infinite at r= 0. Hence the momentum flux distribution is (2 =P; T= | 2L (23-12) This distribution is shown in Fig, 2.3-2. Parabolic velocity distribution v,(r) | T= 07; 1 Linear momentum flux distribution T(r) Fig. 2.3-2. Momentum flux and velocity distributions in flow in evlindrical tubes.Newton’s law of viscosity for this situation is dv, ar (2.3-13) * The quantity # represents the combined effect of static pressure and gravitational force, To allow for other flow orientations, # may be defined more generally as ¥ =p + pgh, where A is the distance upward (that is, in the direction opposed to gravity) from any chosen reference plane. . Fluid initially al rest Lower plate set in motion 1, — 1 ' Velocity buildup Small ¢ in uns aco a in unsteady a ~ — —— Final velocity Large «distribution in steady flow = v a Fig. 1-1. Buildup to steady laminar velocity profile for fluid contained between two plates. subscripts to indicate differentiation of velocity components. Then, in terms of these symbols, Eq. 1.1-1 is rewritten ast (11-2) ay ‘This states that the shear force per unit area is proportional to the negative of the local velocity gradient; this is known as Newton's law of viscosity, and fluids that behave in this fashion are termed Newtonian fluids. All gases and most simple liquids are described by Eq. 1.1-2; fluids that do not obey this simple law (primarily pastes, slurries, and high polymers) are discussed in §1.2.Substitution of this relation into Eq. 2.3-12 then gives the following differen- tial equation for the velocity: do, (2 - #1) de, _ _ (Po= Fa) 23-14 ar Lt |” O14) Integration of this gives A= #1) —_ L) 246, 23-15 - ( au ° aa Because of the boundary condition that v, be zero at r = R, the constant C, has the value (P, — #;)R*/4uL. Hence the velocity distribution is eet] This result tells us that the velocity distribution for laminar, incompressible flow in a tube is parabolic. (See Fig. 2.3-2.) Once the velocity profile has been established, various derived quantities are easily calculated (2.3-16) (i) The maxintum velocity v, mae Occurs at r = 0 and has the value = 2 eee (2.3-17) au. (ii) The average velocity {v,) is calculated by summing up all the velocities over a cross section and then dividing by the cross-sectional area: re a pee rd (ay One 2) =e i ff r dr dO ae o Jo The details of the integration are left to the reader. Note that (v,) = We.max- (ili) The volume rate of flow Q is the product of area and average velocity; thus (2.3-18) (Po — Pi)R* Bul This rather famous result is called the Hagen-Poiseuille? law in honor of the two scientists** credited with its formulation. It gives the relationship Q= (23-19) * Pronounce Poiseuille as “Pwah-z0'-yuh,” in which o is roughly the same as the “oo” in the American pronunciation of “book”. *G, Hagen, Ann, Phys. Chem, 46, 423-442 (18389). “J.L, Poiseuille, Compte Rendus, 11, 961 and 1041 (1840); 12, 112 (1841).between the volume rate of flow and the forces causing the flow—the forces associated with the pressure drop and the gravitational acceleration. (iv) The z-component of the force of the fluid on the wetted surface of the pipe, F,, is just the momentum flux integrated over the wetted area: dv, Grrt)(—p22)| = aR, 91) oR = a R*(py — py) + 7 R*Lpg (23-20) This result is not surprising —all it says is that the net force acting downstream on the cylinder of fluid by virtue of the pressure difference and gravitational acceleration is just counterbalanced by the viscous force F,, which tends to resist the fluid motion. The results of this section are valid only for values of Reynolds numbers less than about 2100, for which the flow is laminar. For this system, it is customary to define the Reynolds number by Re = D(v,)p/u, where D = 2Ris the tube diameter. We may now summarize all the assumptions that are implied in the development of the Hagen-Poiseuille law: a. The flow is laminar—Re less than about 2100. b. The density p is constant (“incompressible fiow”). c. The flow is independent of time (“steady state”)—the corresponding unsteady-state problem is discussed in §4.1. d. The fiuid is Newtonian—that is, 7,, = —p(dv,]dr). e. End effects are neglected—actually an “entrance length” (beyond the tube entrance) on the order of L, = 0.035 Re is required for build-up to the parabolic profiles; if the section of pipe of interest includes the entrance region, a correction must be applied.® The fractional correction introduced in either AF or Q never exceeds L,/L if L > L,. J. The fluid behaves as a continuum—this assumption is valid except for very dilute gases or very narrow capillary tubes, in which the molecular mean free path is comparable to the tube diameter (“slip flow” regime) or much greater than the tube diameter (“Knudsen flow” or “free molecule flow” regime).* g. There is no slip at the wall—this is an excellent assumption for pure fluids under the conditions assumed in (f). 54, H. Perry, Chemical Engineers Handbook, McGraw-Hill, New York (1950), Third Edition, pp. 388-389, W. M. Kays and A. L. London, Compact Heat Exchangers, McGraw- Hill, New York (1958), p. 49. *M. Knudsen, The Kinetic Theory of Gases, Methuen, London (1934). E. H. Kennard, Kinetic Theory of Gases, McGraw-Hill, New York (1938). G. N. Patterson, Molecular Flow of Gases, Wiley, New York (1956).§8.1 FOURIER'S LAW OF HEAT CONDUCTION Consider a slab of solid material of area A between two large parallel plates a distance Y apart. We imagine that initially (for time r <0) the solid material is at a temperature Ty throughout. At time 1 = 0 the lower plate is suddenly brought to a slightly higher temperature 7; and maintained at that temperature. As time proceeds, the temperature profile in the slab changes, 1, Solid initially at Y 2 temperature Tp —1_| Lower plate £=0 suddenly raised to temperature 7; Fig. 8.1-1. Build up to steady-state temperature profile for a solid slab between two plates; see Fig. I.I-1 for analogous situation for momentum transport. and ultimately a linear steady-state temperature distribution is attained. (See Fig. 8.1-1.) When this steady-state condition has been reached, a constant rate of heat flow through the slab is required to maintain the temperature difference AT = T, — Ty. Itis found then that for sufficiently small values of AT the following relation holds: o_, aT AY That is, the heat flow per unit area is proportional to the temperature decrease in the distance Y; the constant of proportionality & is the thermal conductivity of the slab. (8.1-1)Equation 8.1-1 is also valid if a liquid or gas is placed between the plates, provided that suitable precautions are taken to eliminate convection and radiation. Equation 8.1-1 therefore describes the heat-conduction process in solids, liquids, and gases. Convection and radiation are treated in later chapters. In the analytic treatments that follow it will be more useful to work with Eq. 8.1-1 in differential form; that is, we shall use the limiting form of this equation as the slab thickness Y approaches zero. The local heat flow per unit area (heat flux) in the positive y-direction is designated g,. In this notation Eq. 8.1-1 becomes! wake (8.1-2) dy This equation is the one-dimensional form of Fourier’s law of heat conduction, valid when T = T(y). It states that the heat flux by conduction is proportional to the temperature gradient, or, to put it somewhat pictorially, heat “slides downhill on the temperature versus distance graph.” In an isotropic? medium in which the temperature varies in all three directions we can write an equation like Eq. 8.1-2 for each of the coordinate directions ar c, 7 8.1-3,4,5 en qe ( ) These three relations are the components of the single vector equation q=—-KVT (8.1-6) which is the three-dimensional form of Fourier’s law. It states that the heat flux vector q is proportional to the temperature gradient® V7 and is oppositely directed. Thus in an isotropic medium heat flows by conduction in the direction of steepest temperature descent. In a moving fluid q represents the flux of thermal energy relative to the local fluid velocity The reader will have noticed by this time that there is a striking similarity between Eq. 8.1-2 for one-dimensional energy flux and Eq. 1.1-2 for one- dimensional momentum flux. In both cases the flux is proportional to the + We shall have use for three kinds of heat quantities in the following chapters: the heat flow Q (Btu ht"), the heat flux q (Btu hr ft), and the heat source strength S (Btu bet), * By isotropic we mean that the coefficient & in Eqs. 8.1-3, 4, and 5 has the same value in all three directions. The assumption of isotropy is satisfactory for fluids and for most homogeneous solids; the principal nonisotropic materials are single noncubic crystals and fibrous or laminated solids. (See wood in Table 8.1-4.) # The vector quantity V7'is read “gradient of T” or “del 7,” and in some books it is written as grad T. For a discussion of the gradient of a scalar field, see G. B. Thomas, Analytic Geometry and Calculus, Addison-Wesley (1953), Second Edition pp. 497-500. See Table 10.2-1 for the components of Eq. 8.1-6 in curvilinear coordinates.negative of the gradient of a macroscopic variable, and the coefficients of proportionality are physical properties dependent on the substance and on the local temperature and pressure. For more complicated situations in which the temperature and velocity vary in all three directions, however, we notice that Eqs. 8.1-3, 4, and 5 for the energy flux are simpler than Eqs. 3.2-I1 through 16 for the momentum flux. This difference in form arises because energy is a scalar quantity but momentum is a vector quantity. As a result, the energy flux is a vector with three components, whereas the momentum flux is a tensor with nine components. We can thus anticipate that problems of momentum transport and energy transport will not be mathematically analogous except in certain geometrically simple situations.§16.2 FICK’S LAW OF DIFFUSION In Eq. 1.1-2 the viscosity 1 is defined as the proportionality factor between momentum flux and velocity gradient (Newton's law of viscosity). In Eq. 8.1-6 the thermal conductivity k is defined as the proportionality factor between heat flux and temperature gradient (Fourier’s law of heat conduction). TABLE 16.2-1 EQUIVALENT FORMS OF FICK’S FIRST LAW OF BINARY DIFFUSION Flux Gradient Form of Fick's First Law m4 Vox Mg — 04g +p) = —p%yyVog (A) i Vr4 Ng — 24(Nq + Ny) = ~c% 44 V2q () ja Vag 5a = ~0%4y Very «© Jat Ve Iagt = CP 4, V0, (D) 2 ja Vag da = (5) aeadtatan Vig (2) Jyh Vo, 244 Vo, (F) ; . A au an vg — vp) Vg (og — vy) = ——“2 va, ©) rary Now we define the mass diffusivity 2 ,,, analogous fashio: Py in a binary system in an Jyh = cP yy Ne, (16.2-1) This is Fick's first law of diffusion written in terms of the molar diffusion flux J4*. This equation states that species A diffuses (moves relative to the mixture) in the direction of decreasing mole fraction of A, just as heat flows by conduction in the direction of decreasing temperature. A number of other mathematically equivalent statements of Fick’s first Jaw have appeared in the literature, and some of them are summarized in Table 16.2-1 for reference only. The diffusivity %4y is identical in all these equations. Of special importance in the following chapters is the form of Fick's first law in terms of N,, the molar flux relative to stationary co- ordinates: Ny =24(Nq + Np) — ayV2, (16.2-2) + Temperature gradients, pressure gradients, and external forces also contribute to the diffusion flux, although their effects are usually minor. More complete expressions for the diffusion flux are given in §18.4. A. Fick, Ann. der Physik, 94, 59-86 (1855)This equation shows that the diffusion flux N,, relative to stationary co- ordinates is the resultant of two vector quantities: the vector x4(Ny + Ny), which is the molar flux of A resulting from the bulk motion of the fluid, and the vector J* = —cP4,Vx,, which is the molar flux of A resulting from the diffusion superimposed on the bulk flow. Thus in Fig. 16.1-16 the bulk flow and diffusion terms in Eq. 16.2-2 are in the same direction for species A (because A is diffusing with the current) and are opposed for species B (because B is diffusing against the current). The units of the mass diffusivity D4 are cm? sec! or ft2 hr, Note that the kinematic viscosity » and the thermal diffusivity « also have the same units. ‘The way in which these three quantities are analogous can be seen from the following equations for the fluxes of mass, momentum, and energy in ‘one-dimensional systems: 4 4 (4) — (Fick's law! for constant p) (16.23) y of (pv.) (Newton's law for constant p) (16.24) ty y= a4 (66,1) (Fourier’s law for constant pG,) (16.2-5) ly ‘These equations state, respectively, that (a) mass transport occurs because of a gradient in mass concentration, (6) momentum transport occurs because of a gradient in momentum concentration, and (c) energy transport occurs because of a gradient in energy concentration. These analogies do not apply in two- and three-dimensional problems, however, because t is a tensor quantity, with nine components, whereas j, and q are vectors with three components.