Theory of Mobilization Pressure Gradient of Flowing Foams in Porous Media I1. Effect of Compressibility W I L L I A M R. R O S S E N l Chevron Oil Field Research Company, P.O. Box 446, La Habra, California 90633-0446 September 29, 1988; accepted July 31, 1989 This report extends the theory of the minimum pressure gradient (Vp)m~"for flowingfoams in porous media to include the effects of gas compressibility.The compressibilityof the gas phase raises (Vp) rain because, in compressiblefoams, lamellaetend to lodge in pore throats, where capillaryresistanceto flow is greatest. The effect of compressibilitydepends on bubble size, gas compressibility,pore-throat geometry, compressibilityof adjacent trapped-gas bubbles, capillary pressure Pc, and bubble-train length, i.e., the number of consecutive bubbles between slugs of liquid. With reasonable parameter estimates, theory predicts that medium-textured steam foams (bubbles 400 um in diameter in bulk) have (Vp) rainof 1.2 MPa/m (55 psi/ft). For nearly incompressibleCO2 foams, the compressibilityeffect depends strongly on capillarypressure and (Vp)minis reduced by lowergas-liquid surfacetension. Theory predicts (Vp)rain = 125 and 20-40 kPa/m (6 and 1-2 psi/ft) for medium-textured CO2 foams at low and high capillary pressure, respectively. Finer textured CO2 foams have a still higher (xup)min. Clearly, limited foam coalescence is essential for deep foam penetration away from the wellbore in an oil reservoir. Modified for the effect of compressibility,this theory of (~Tp)min fits the beadpack data of Falls et al. (SPE Reservoir Eng. 4, 155, 1989). ©1990AcademicPress,lnc. INTRODUCTION where ( A p ) t oavgt 1S - the average capillary Ap per Foams are a promising means to improve reservoir sweep and oil recovery in gas-injection enhanced oil recovery ( E O R ) projects (1). However, the existence of a substantial m i n i m u m pressure gradient to create and sustain a flowing foam in porous media could limit the successful application of foams to lamella in the train a n d nL is the n u m b e r of lamellae per u n i t length of the train i n the direction of flow. By c o n v e n t i o n , Ap is positive if the forward b u b b l e is at a lower pressure. Analysis of the passage of a single lamella t h r o u g h a single pore gives the time-average Ap EOR (2). At low pressure gradients, foams flow as " b u b b l e trains" along s i n u o u s paths t h r o u g h regions of trapped b u b b l e s (3, 4). I n a comp a n i o n paper (2) we derive a theory for the m i n i m u m pressure gradient (Vp) rain required to keep a train of incompressible b u b b l e s flowing. T h e f u n d a m e n t a l e q u a t i o n of this theory is (VP) rain = ( An'~avg~ /J]tot 'tC, (Ap) avg = f0 ' Apd-r, [21 where Ap is the pressure difference between bubbles and r is the fraction of the pore transit time. Analysis also gives ¢, the standard deviation of the distribution Ap(r). The curvature of the lamella governs Ap, 43, zXp --- Pl - P2 = - - , [1] [3] rl where 3' is the liquid/gas surface tension and rl is the radius of the spherical lamella at a Present address: Department of Petroleum Engineering, The University of Texas at Austin, Austin, TX 78712. 17 0021-9797/90 $3.00 Journal of Colloid and Interface Science, Vol. 136, No. 1, April 1990 Copyright © 1990 by Academic Press, Inc. All rights of reproduction in any form reserved. 18 WILLIAM R. ROSSEN given position. By convention, we define q and of linear compressibility [4], though an apAp as positive when the lamella bulges forward proximation, greatly simplifies the analysis to and negative when it bulges backward. follow. Plugging the equation of state ( c f [ 4 ] ) (Ap)ta'g depends both on (Ap)avg and on for each bubble into Eq. [3] gives (2). Large values of (Ap) avg means lamellae ~_ = 2 7 k ( rl ]-l + v_! require, on average, large positive pressure difrtVp \ rt ] Vp ferences between bubbles in order to flow. Large values of a mean that the random fluctuations in pressure between bubbles cause large excursions in pressure along the train and frequent bubble separation and snap-off in the with interior of the train. Both bubble separation K ~ ~'~T"el"Ov° [6] and snap-off increase (Ap) ~vg and (~Tp)min. Up In this paper we extend our theory to account for bubble compressibility. In addition, Pc~ 23/ [7] we consider several effects that magnify the rt effective compressibility of the foam: first, the contraction and expansion of trapped-gas where vl is the volume behind the lamella in bubbles and liquid films to accommodate the the pore of interest; vp is volume of a pore; compression and decompression of moving and p~ is the capillary entry pressure of the bubbles, and, second, the autocorrelation of pore. For a given pore geometry, both (rffrt) lamella movement in trains of multiple bub- and (v~/Vp) are functions of dimensionless lables. We show good agreement between theory mella position ( x / l ) , so it is convenient to and the data of Falls et al. (3) when these ef- solve Eq. [5] for r as a function o f x / l (2). fects are taken into account. A preliminary Portions of the curve z ( x / l ) where T decreases with increasing x are replaced by jumps in poreport of this work was given elsewhere (5). sition (6). Equation [ 6 ] reveals that the effective comTHEORY AND RESULTS pressibility is proportional to gas compressiEFFECT OF GAS COMPRESSIBILITY bility, bubble volume, and the capillary entry pressure of the given pore. High gas compresTheory for One Lamella sibility and large bubble size both contribute to large values of K; for a given pore shape, We employ the same conceptual model as that in Ref. (2), that is, a single lamella tra- so does a large capillary entry pressure, versing a single pore, impelled by the steady whether due to large surface tension 3' or small flow produced by moving pistons upstream throat radius rt. If the pressure either upstream or downand downstream. Here, however, we allow the gas in both forward and rearward bubbles to stream of the bubble is constant, then Eq. [ 5 ] be compressible. We assume that the pressure still obtains, with K twice the value given by Eq. [ 6 ]. If constant-pressure conditions apply in each bubble is linear in volume both upstream and downstream, then either Ov - - =- - k ~- - k xo vbo ~-- constant, [4] the lamella is stranded at a position where Eq. Op [ 3 ] is satisfied or there is no position at static equilibrium at any time. where k is the linear coefficient of compressibility and kT and V0b are, respectively, the conventional coefficient of compressibility ( - 0 In V/Op)T and bubble volume at the average pressure of the foam. The assumption Journal of Colloid and Interface Science, Vol. 136, No. 1, April 1990 Results for One Lamella Figure 1 shows the progress of the lamella through a conical pore with dimensionless 19 FLOWING FOAMS IN POROUS MEDIA, II a) K =0 / 1.0 r b) / i 1.0 I ./,v 0.5 K = 0.2 /V T 0.5 I 0.5 1II 0 1.0 01.5 x/~ c) 1.0 x~ K = 0.6 d) K = 2.0 1.0 7- 0.5 [ "._.,;' 0 0 I / T 0.5b ~, / ¢" / L 0.5 1.0 x/~ ii / OL o 0,5 x/,( // '\ 1.0 FIG. 1. Effect of compressibility factor K on lamella transit through a conical pore with Pb Pt = 0.1. pore-body radius P b ~--- rb/l = 0.5 and dimensionless throat radius pt = rt/l = 0.1 as a function of K. Horizontal segments correspond to j u m p s in position. For K = 0 (Fig. l a ) , there is one j u m p as the lamella reaches the corner at the midpoint of the pore and no change in bubble volume at the j u m p (2). For K = 0.2, the rearward bubble compresses and the forward bubble decompresses for a time, until the rearward bubble has enough pressure to force its way through the pore throat. Then it jumps to a position at which it is decompressed enough to be at equilibrium at the lower capillary ~xp. There is the j u m p again at the pore body, this time including some expansion of the rearward bubble as a result of the j u m p and change in sign of ~xp. Finally, there is a third j u m p as increasing capillary force pulls the lamella into the converging pore throat. For higher K(Figs. lc and ld), the j u m p s become greater until they merge into one jump: the lamella hops from the position of maxim u m ~xp in the pore throat to the next pore = 0.5, throat, near the position of m a x i m u m Ap. A similar progression is obtained for a sinusoidal pore with Pb = 0 . 5 , Pt = 0 . 1 , although larger values of K are required to induce the multiple jumps. Figure 2 shows for the conical pore how the frequency distribution of Ap (2) changes with K. As K increases, the lamella spends a large fraction of its time in the diverging pore throat until, for sufficiently large K, the lamella is always in the throat with ~xp near its m a x i m u m value, (Lxp) ma~. For uniform, conical pores, (Ap)max/p e = 2 s i n [ t a n - l ( 2 ( P b -- Pt))]. Figure 3 shows how (Lxp) avg and ¢ depend on K for conical and sinusoidal pores with pt = 0.5, pt = 0.1. For conical pores, (Lxp) avg approaches the incompressible limit for K less than 0.1 and approaches (~xp) maxfor Kgreater than 10. As K increases beyond 1.0, ¢ approaches 0 because all lamellae are at the pore throat with Lxp near (~xp) maX. For sinusoidal pores, although there is a small j u m p in position for K as low as 0.12, (Ap) ,vg is insignifJournal of ColloM and Interface Science, Vol. 136, No. 1, April 1990 20 WILLIAM R. ROSSEN b) K = 0 . 2 a) K = 0 Diverging ~ COrn:B r g ~ _ ~ e Body Diverging Pore Body Converging Pore Body Diverging Converging / ~t Por~Throat I - 1.0 0 1.0 (Ap) max . ~p/pc. ~ _ (~p) . . c) K = 0.6 (Ap)max A p/pce Diverging Pore Body Converging Pore Throat Diverging Pore Throat Diverging Pore Throat \ / . . . . . . . . . . . 1-'1.o _ (~p) . 0 d) K = 2.0 / _ 1,0 (AP) max - 1.0 I o . ~p/pc e . 1.0 (Ap) max . , l- .o (~p)max o v , /,4 1.o ,x zXp/pce FIG. 2. Frequency distribution of Ap as a function of K for a conical pore with Pb = 0.5, Pt = 0.1. icant for K less than 3. (Ap) avg approaches its maximum value as K increases beyond 100. These results show that for long or highly compressible bubbles, only the geometry of the pore throat, which determines (Ap) max, matters to (Vp) rain (Eq. [1]). In contrast, for incompressible bubbles in smooth symmetric pores, only the geometry of the pore body, where the jump in position occurs, matters to (Vp) mi" (2). Multiple Pore Shapes If a train of highly compressible bubbles passes through a variety of pore shapes, the pores with the highest (Ap) max determine (Ap) "~g for the train: after compressing to overcome (Ap) max at the tightest throats, lamellae pass completely through pore throats with lower (Ap) m"x. Figure 4 illustrates this phenomenon for a train passing through conical pores with the same length l and body Journal of Colloid and Interface Science, Vol. 136,No. 1, April 1990 radius Pb = 0.5 in all pores but with porethroat radius pt alternating between pt = 0.1 and 0.25. Also pictured is (Ap) avg for uniform conical pores with pt = 0.1 and with ot = 0.25. For comparison, we have normalized all the curves by using pe for the tighter pore throat, pt = 0.1. For the sequence of alternating pore types, in the incompressible limit K --~ 0, (Ap) "vgis simply the volume average of(Ap) avg for the two pore types. As K increases, however, the lamella spends more of its time in the tighter pore throats; as K ~ oe, (Ap) avg reflects only the presence of the tighter throats. Thus, (Ap) avgfor a train of highly compressible bubbles reflects only the tightest pore throats in the train. The train follows the path of least resistance, however. Therefore, out of all the pores in the porous medium, the train samples the percolation-threshold (7-10) fraction (about 15%) of pores with the largest pore throats (2). As K approaches infinity, (Ap) a v g reflects FLOWING FOAMS IN POROUS i 21 M E D I A , II i j fs. 1.0 c°n7 (Ap) avg pce 0.5 " I i 0,1 1. ~ ] 10. 100. 10. 100. K \ pce C o n i al 0.1 1. K FIG. 3. Effect of compressibility factor K o n ( A p ) avg a n d ~r for conical a n d sinusoidal pores with/0 b - 0.5, pt = 0.1. the tightest throats in this fraction; i.e., it reflects the throats at the percolation threshold itself--the throat size measured at the "threshold pressure" in conventional mercury porosimetry (9, 10). EFFECT OF BUBBLES ALONGSIDE TRAIN Theory Trapped-gas bubbles alongside the flowing bubble train expand and contract in response to fluctuations in pressure of individual bubbles in the train. This increases the effective compressibility of the bubbles in the train. We augment our model to allow for movement of lamellae separating trapped-gas bubbles from the train as follows. For each bubble i in the train, we assume there are nsi trapped bubbles surrounding it numbered j = 1 through n~i. We assume initially that the surrounding bubbles are isolated from each other and fixed in volume except for the m o v e m e n t of the lamellae they share with the bubble train. The m o v e m e n t of side lamella ij produces a change Avsi; in the volume of the side bubble and an opposite change in the volume of bubble i. By convention we call 2xvs o > 0 if the volume of bubble i increases and that of side bubble ij decreases. For each lamella alongside the train we choose a reference position such that 2xv~u = 0 at r = 0. The pressure in side bubble ij is P w , and we define ps°j =- p ( AVsij = 0 ) . From the bubble equations of state (cf. Eq. [4]) and the capillary Ap conditions (cf. Eq. [ 3 ] ) we obtain 47 - 2(vor- vl)/k rl + 47 '~PS O' - - rso (-~ Vsu + ~ vs2j)/k [81 - Pi - Ps°j - Avsij/k, o, i = 1,2; j= 1. . . . . nsi. [9] Journal of Colloidand InterfaceScience, Vol. 136, No. 1, April 1990 22 WILLIAM R. ROSSEN 1.0 / pe Pt = 0"1/ (Ap)avg /Alternating Pore Types 0.5 ~ 01 0 ~ I Pt = 0.25 / 0.1 I 1. K 10. FIG. 4. Effect of compressibility factor K on (Ap)aVg for conical pores with pb p~ = 0.25, or an alternating sequence of the two pore types. Here, ks o ~- (OAvsij/Op~ij), a function of the initial volume and pressure of side bubble ij (cf. Eq. [4]). For the side bubbles we define rsij as the radius of the spherical lamella separating bubble ij from the train, with r~ij > 0 if the lamella bulges into the side pore. We assume here for simplicity that the simple conical and sinusoidal pore shapes adequately describe both the pores of the bubble train and the side-pore throats. One could solve Eqs. [8] and [91 simultaneously for positions of all lamellae as functions of time given their initial positions and knowledge of how r~ varies with Vl and how the r~ij vary with AvsO. Instead, we offer here an approximate solution. 100. = 0.5 and either pt = 0.1, keff keffdpl = d(vpr - 1)1) [11] with 1 keff/ k - ~ 1 + ~c ]" [121 Differentiating Eq. [ 9 ] for i = 1 with respect to P l g i v e s dAl)slj dpl [ dApslj] -1 - kslj 1 + ksl j dAvslj J " [13] The capillary entry pressure of the side throat is 2X Pc~sU~ [141 rts lj A p p r o x i m a t e Solution From the definition of k =- - ( O v / O p ) and the equation of state for the rearward bubble 1, kdpl + d ~ AVsij dpl dpl = d(vpr - Vl). with rtslj the side throat radius. Combining Eq. [ 13 ] with Eq. [ 12 ], noting k ~ U 0b k 0T and kslj ~- l ) 0b s k T0 , Gfr/k ~- 1 V°s [ d-(APslHP~slj)] -I + ~ V--~D 1 + Ks d(AVslj/Vp ) j [151 [101 with The effect of movement of side lamellae is to increase the effective compressibility to a value Journal of Colloid and Interface Science, Vol. 136, No. 1, April 1990 V°s Ks =- peslj - - g°T. Vp [161 FLOWING FOAMS IN POROUS Here we have assumed that k ° has roughly the same value in the side bubbles and in the bubble train and that all side bubbles have roughly the same volume, v°s. Equation [15] indicates that the m o v e m e n t of side lamellae increases the effective compressibility of bubbles in the train. If the large term in brackets in Eq. [ 15 ] is 1, then bubble compressibility in the train increases as though each flowing bubble included the trapped bubbles around it. The term in brackets thus indicates the capillary resistance to m o v e m e n t of the side lamellae. If this term is near 1, then the first row of side lamellae does little to d a m p the expansion and contraction of bubbles in the train. In a real porous m e d i u m this effect would then propagate through m a n y rows of trapped bubbles and greatly increase effective compressibility. This capillary resistance is proportional to (Eq. [ 16 ] ) the capillary entry pressure of the side pores, the volume of side bubble relative to a pore in the bubble train, gas compressibility, and (Eq. [ 15 ] ) the dimensionless change in a capillary Ap per unit change in bubble volume. From the definition of compressibility factor K, we have Keff/K = 1 V°s [ d(Aps*J/P~slJ)]-' [171 + ~ V~b 1 +Ks d(AVslffVv) j " If we l e t m s ~- ~ 1)bs//) 0 0b be the total volume of side bubbles relative to the volume of the flowing bubble, and replace the elements in the summation by a representative average value, Eq. [ 17 ] becomes Keff/K = 1 + m s 1 d(Aps/P~s)]- 1 +Ks~j__ . [18] Two Examples Beadpack. The effect of the side bubbles depends on the geometry of the pore throats alongside the train. We consider two examples. The first is a pack of beads of radius R at the m a x i m u m packing density. There are two pore 23 M E D I A , II types in such a pack: for one there are eight pore throats, l = 1.73R, Pb --~ 0.24, and Pt - 0.086; for the other there are four pore throats, l - R, Pb ~ 0.23, and Pt ~ 0.155 ( 11, 12). We assume that there are both pore types in the train; the train must alternate pore types if the beads are in a hexagonal-close-pack arrangement. Using a conical-pore model for pore bodies and taking the average for the two pore types, we estimate Vp = 0.59R 3. All throats in such a pack have the same diameter (rts = rt = 0.155R), and, thus, when gas invades the interior of the pack, it fills virtually all the pores. We assume there are, on average, four gas-filled side bubbles per pore in the train, and, if each side bubble occupies one "average" pore, v°s/Vp = 1 and ms = 4. Equation [ 18 ] is, for this case, [ Kerf/K~- 1 + 4 1 + 0 . 1 5 5 ~ d(Aps/pes)] -1 ×k ° ~ ~ j . [19] For instance, in the 3 - m m beadpack experiments of Falls et al. (3) at 1 atm, R = 0.3 cm, kT -~ ( 14.7 psi) -1 = 9.86 × 10 -7 ( d y n / c m 2) 1, 3' = 30 d y n / c m , and Kecf/ K = 1 +4 d(Apslpes)] 1 1+0.0013 ~ j . [20] To estimate the derivative in Eq. [ 20 ], we approximate the pore throats by sinusoidal pores with Pb = 0.177, Pt = 0.0896; this approximates the curvature in a toroidal pore throat with the same geometry as pore throats in the beadpack. The result is shown in Fig. 5a. In the center of the pore throat (Aps = 0), the derivative is large but as the side lamella is displaced from the throat, the derivative falls, passing through zero for large displacements. If Ks is sufficiently large (side bubble size very large or gas compressibility very low), as the derivative falls below zero the denominator in Eq. [ 18 ] also m a y pass through zero for some lamellae: these side lamellae j u m p Journal of Colloid and Interface Science, Vol. 136, No. 1, April 1990 24 W I L L I A M R. R O S S E N a) i 15 i Pore Throat 10 d Aps/pes d AVs/V p 0 Pore Body ~ _/2 I -1 -3 J 0 Pore Body I i 1 2 APs/P~ b) i i i i i Pore Throat 5000 dZ~p~/p e d AVs/Vp I ,I 0 -1000 -3 II I 't1 Pore Body I : ZXPs/pg FIG. 5. F u n c t i o n used to estimate effect of trapped-gas bubbles alongside train. (a) Bead pack. ( b ) Rock. from their throats and the effective compressibility is, briefly, infinite. We employ a generous estimate of the derivative in Eq. [20] for all side lamellae, 15 (cf. Fig. 5a). Inserting this value into Eq. [20] gives Keer/K = 1 + 4 [1.02] -1. [21] Equation [21] indicates that the effective compressibility is quintupled due to the expansion and contraction of the first row of trapped-gas bubbles alongside the train. It is a conservative estimate because it excludes the possibility of discontinuous jumps by side lamellae. More important, the value near 1 within the brackets in Eq. [ 21 ] means that the first row of side lamellae damps little of the expansion and contraction of bubbles in the train: in other words, what Falls et al. called the "breathing" effect (3) propagates through many rows of lamellae around the train, effectively engaging the entire beadpack in the pulsating pressure fluctuations of the train itself, as Falls et al. observed. If we allowed for Journal of Colloid and Interface Science, Vol. 136, No. 1, April 1990 many rows of bubbles in our model, Keffwould increase by orders of magnitude. In general, pore geometry, bed size, and gas compressibility together determine the magnitude of this effect in beadpacks. Rounded pore throats facilitate breathing (Fig. 5a). On the other hand, "breathing" would be less pronounced in a pack of smaller beads, while it would be magnified by lower gas compressibility (Eq. [ 16 ] ). Rock. Our second example is oilfield rock. We approximate both pore shape in the train and shape of the side throats by angular, conical geometry. Rocks can have a wide distribution of pore-throat sizes (13). Since the bubble train selects the largest throats for its path, most throats alongside the train are significantly tighter than those in the path of the train. Therefore, we describe pore geometry in the train by a conical pore with Pb = 0.5, Ot = 0.1, and pore-throat geometry alongside the train by a conical pore with Pb = 0.5, 0t = 0.05. We assume initially, as before, that FLOWING FOAMS IN POROUS each side bubble occupies one pore (VObs/Vp 1) and 3' = 30 d y n / c m . We take kT ~ (50 psi) -I = (350 kPa) -1 = 2.9 × 10 -7 ( d y n / cm2) -1 as in a steamflood away from the nearwellbore region (14). We assume that there are four gas-filled pores beside each pore in the train; this generous estimate corresponds to a Bethe-tree pore network of coordination number 7 ( 15, 16) with one side pore occupied by water and two of the throats, of course, taken by the path of the train. We assume a pore length l of 150 #m. Equation [18] becomes d(Aps/P~s)] -1 geff/g ~ 1 -]-4 1 + 0.023 ~ j beyond the first row of bubbles, and Ke~ would be larger than estimated here. A reasonable estimate is that "breathing" increases Kerr by a factor of about 10 for CO2 foams. Side lamellae make a smaller contribution to Kerr in rock than in beadpacks primarily because the derivative in Eq. [18] is small in narrow, angular pore throats. The wider distribution of pore-throat sizes in rock than in beadpacks means that P~s is relatively small as well (Eq. [ 16 ] ). "Breathing" is expected to be less significant in rock than in beadpacks unless the pore throats in a given rock are large, rounded, and of uniform size. EFFECT OF LIQUID FILMS AND LIQUIDFILLED PORE THROATS Figure 5b shows how (d(2xpJpec~)/ d(2XVs/Vv)) varies with 2xps for the given side-pore-throat geometry. If 2xps exceeds +2.7pce, then the derivative discontinuously changes sign: if Ks is large enough, the denominator term in Eq. [22] may fall below zero, and the lamella may jump from the throat. For 2xps between +_2.7p~, a reasonable value for the derivative is 5500. Equation [22] is then [23] The effective compressibility increases by only a few percent. Moreover, because the first row of side lamellae damps almost all of the "breathing" effect at the first row of bubbles, the effect does not propagate further into the surrounding rows of trapped bubbles as it does in beadpacks. According to Eqs. [ 15 ] and [ 16 ], if k ° and 3' are lower, the effect of side lamellae is somewhat greater. For instance, using the same parameters as above except k ° = 6.7 × 10-4 psi-i = 9.7 × 10 -9 ( d y n / c m 2 ) -~ and 3' = 5 d y n / cm (reasonable values for CO2 floods ( 17, 18 )) gives Liquid Films Liquid films filling corners and crevices in pores can swell and shrink to accommodate fluctuations in pressure in the bubble train. An important practical consideration is whether these films have time to respond to the rapid changes in pressure as bubbles cascade from pore to pore. Flow along films presumably is slower than the bulging in and out of trapped lamellae analyzed above. Here, nevertheless, for simplicity, we assume that these films maintain equilibrium with the surrounding liquid at constant (though spatially nonuniform) pressure. An alternative asymptotic assumption would treat the films as isolated from surrounding liquid and fixed in volume. In that case, of course, they would have no effect on compressibility. By arguments similar to those in the previous section (cf. Eq. [ 12 ]), one can obtain Kerr~K= 1 - [24] Again, the relatively small value in brackets means the "breathing" effects extends outward (1)d(2xvf/vp) ~f d(pi/p~) [25] with Kf Kaf/K= 1 + 4 [ 1 + 0 . 7 ] -1 = 3 . 4 . 25 II . [221 Keff/K~ 1 + 411 + 130] I = 1.03. MEDIA, e 0 =- Pckx, [26 ] where Avf is the change in the volume occupied by films in a single pore. The swelling and draining of liquid films can give a finite Journal of Colloid and Interface Science, Vol. 136, No. 1, April 1990 26 W I L L I A M R. ROSSEN ( d y n / c m 2 ) -l, then Kf -~ 0.012. For low capillary pressures, Keff increases by about a factor of 5; at high capillary pressures, by only about 9%. For virtually incompressible CO2 foams, the effect of liquid films depends on bubble size as well as on capillary pressure (Eq. [27]). For instance, assuming a medium foam texture, with bubbles about 400 #m in diameter in bulk (2) and v ° / v v = 30, and low capillary pressure Pci = 0.5p~, gives Ke~ = 1.5. According to Fig. 3, such a foam would behave as though highly compressible, with (Ap) avg about half of (~xp)max. Finer foam texture and higher capillary pressure would reduce Kef~, however. Real pores in rocks may have larger crevices and more substantial liquid films than assumed in this model, especially if the surfaces are deeply roughened. For example, other simple geometric pore shapes, such as cubes and tetrahedra, have about double the edge length per unit volume of the conical pore, and this difference can increase the effect of liquid films on effective compressibility. Moreover, as capillary pressure increases, smaller scale roughness (20) retains liquid on surfaces that are effectively smooth at lower capillary pressure. This mitigates the reduction compressibility effect even if k ° -- 0 for the gas phase. Rearranging Eq. [ 25 ] gives K~ff = K - pt V ° d(Avf/Vp) Vp d(pi/p~) " [27] In deriving the relation between Avf and Pi, we consider a conical pore and, more specifically, the liquid film filling the edge encircling the pore body. We assume that this film is not in contact with a lamella and that the axial cross section of the film surface is circular. This second assumption, though an approximation, greatly simplifies the calculation of ~vf. One can determine vf/v v as a complicated algebraic function (19) of P b , P t , and Of ~ rf/l. The capillary condition on rf is _3'_ Pci ~- P i - - Pliq ot rf e [28] 2pf pc" Figure 6 plots the derivative in Eq. [25 ] for a conical pore with 0b = 0 . 5 , P t = 0.1 as a function of dimensionless capillary pressure (Pci/P~). We expect Pci/P~ to take values between 0.5 and 2 (2). At the low end of this range, the derivative is about 0.05; at the high end, about 0.001. If we take the same parameter values as above, i.e., 3/ = 30 d y n / c m , l = 150 urn, and kT = (50 psi) -1 --- 2.9 X l 0 -7 1.0 i i I ~ i 0.1 0.01 d (A Vf/Vp) d (pi/pce) 0.001 0.0001 0.00001 0 5 Pci/Pce 10 I~G. 6. Function used to evaluate effect of liquid films on effective compressibility for a conical pore with Pb = 0.5, Pt = 0.1. Journalof Colloid and InterfaceScience, Vol. 136, No. 1, April 1990 FLOWING FOAMS IN POROUS MEDIA, II in the effect of liquid films predicted above at high capillary pressures. Thus, the effect of liquid films on effective compressibility of foams in rock may be somewhat larger than estimated here. Liquid-Filled Throats A similar analysis covers the intrusion of gas into liquid-filled throats. By arguments similar to those above, one can obtain Keff/K = 1 + (~f) d(Al)t/I)P) ml d ( p i / p ~ ) ' [29] where m~ is the average number of liquid-occupied throats adjacent to each pore in the bubble train, Art is the change in bubble volume due to movement into or out of a liquidfilled throat, Kf is defined by Eq. [ 26 ], and rt, Vp, and p~ apply to the pores and throats in the path of the bubble train. For virtually incompressible foams v° d(Avt/vp) Vp d ( p i / p e) " Keff= K + p~ - - ml [30] Three examples. We consider three cases. In each case we use conical pore geometry for both the pore shape in the train and the liquidfilled pore throats alongside the train. The derivative in Eq. [ 29 ] can be determined for this geometry as a function of pe given values for 0.1 ~0.01 d(vttVp) Wide, ~ Uniform \ 27 parameters, Pt, Pb, and Pts, the throat radius of liquid-filled throats. We assume that initially there are one or two liquid-filled pore throats adjacent to each bubble in the bubble train ( m j -- 1 . 5 ) . In the first case, a wide distribution of porethroat sizes means that liquid-occupied pore throats are significantly narrower, on average, than those occupied by gas: P b = 0 . 5 , P t = 0 . 1 , Pts = 0.05. We believe this model is appropriate for reservoir rock. Figure 7 shows the derivative in Eq. [ 29 ] as a function of capillary pressure in bubble i (curve labeled "narrow, varied throats"). This factor is large in two instances: if capillary pressure is low, liquid fills the throat completely but small changes in pressure cause significant liquid drainage. As capillary pressure approaches 2p~, gas invades the liquid-filled throat. For Pci > 2P e, gas breaks through the throat and fills the liquid-filled pore on the other side. The effective compressibility is, briefly, infinite. If snap-off does not create a lamella in the newly invaded throat during invasion (11, 21), one will be placed there by lamella division ( 11 ) when the next lamella in the train passes by. In either case, the liquid-filled side pore is replaced by a trapped bubble. If Pci falls below about 0.5pc~, on the other hand, bubble separation or snap-off will occur in the bubble train (2) and disrupt the train's movement: therefore we do not expect Pci to fall below 0.5p~. If we [ ] Throats ~' I Narrow, /Uniform ~/ "'"-' ~~ThroatSNarrow, VariedJ 0.001 0.00010 I (Pci/Pe FIG. 7. Function used to evaluateeffectof liquid-filledthroats on effectivecompressibility. Journal of Colloid and Interface Science, Vol. 136, No. 1, April 1990 28 WILLIAM take Kf = 0.012 as above, then swelling and drainage of liquid-filled pore throats can increase Keg significantly only if bubble capillary pressure is much higher or lower than 1. For Pci = 0.5p¢e, Keg is increased by a factor of about 3. For virtually incompressible foams, Keg again depends on bubble size V°Vp(Eq. [ 30] ). For v°/vp = 30, Ken is as large as 1.3 at Pci = 0.5 pc~. Keg is smaller for finer textured foams and at higher capillary pressures. K~f~is still larger for other throat geometries shown in Fig. 7. The curve labeled "narrow, uniform throats" corresponds to conical pores with Pb = 0.5, Pt = Pts = 0.1. The final case in Fig. 7 has uniform, relatively wide pore throats, a model we believe is more appropriate for a beadpack, although for simplicity here we use a conical model with Pb = 0.5, Pt = Ors = 0.25. In this case (Fig. 7), if v°/Vp = 30, Keg ~> 5 regardless of the values ofpci, k°t, and K. Wider side throats experience significant liquid drainage upon small changes in capillary pressures. With the random fluctuations in pressure expected in a bubble train (2), it is likely that all of the larger pore throats alongside the bubble train are rapidly occupied by gas. This suggests that within a short time ml approaches zero and Keg approaches K. If pore-throat sizes are widely distributed, then it is the narrowest of throats that remain occupied by liquid and their effect on capillary Kefrwould be moderate at best. This analysis, and that for "breathing" of trapped-gas bubbles, is complicated when, at low capillary pressures, the lamellae separating trapped-gas bubbles from the train swell and isolate the trapped-gas bubbles from the train. This is expected to occur at or below about one-half the capillary entry pressure of the side throats (2). Thus, at low capillary pressures, "breathing" of side lamellae stops: bubbles in the train push against the swollen liquid lenses occupying the side throats instead. At these low capillary pressures, Eq. [ 29 ] then applies with m~ representing the throats filled with liqJournal of Colloid and Interface Science, Vol. 136, No. 1, April 1990 R. R O S S E N uid, whether the pores across the throats are occupied by liquid or by gas. Summary In summary, the drainage and swelling of liquid films in the crevices and comers of pores and of the liquid in liquid-filled pore throats can help relieve the compression and decompression of gas bubbles in a train and raise the effective compressibility. At low capillary pressures, the effect ofliquidfilms can increase Keg by orders of magnitude, especially for otherwise incompressible foams. For porous media of uniform throat size, we expect the liquid-occupied pores alongside the train to fill with gas quickly. Then the effect of side lamellae depends on capillary pressure. If capillary pressure is relatively high, these lamellae move back and forth in response to bubbletrain pressure and the process of "breathing" covered by Eq. [ 17 ]. If the capillary pressure is low, these lamellae swell into lenses that swell and drain with changes in capillary pressure as described by Eq. [29]. This effect can raise Kerr by orders of magnitude. For porous media of varied pore sizes, the effect of the narrow throats that remained filled with liquid even at high capillary pressure would increase Kerr more moderately. BACKWARDS LAMELLA MOVEMENT AS noted, when pressure builds exceptionally high in rock (Fig. 5b) or builds more moderately in a beadpack (Fig. 5a), a lamella may jump from the side pore throat into the body of the side pore. In that event, the affected bubble decompresses discontinuously; the lamella behind it may pull forward and the lamella ahead may pull back. Conceivably, the lamella ahead could pull back to a position it had previously jumped over, e.g., into one of the dotted "inaccessible" branches in Fig. 1. This is the only means we are aware of by which a lamella in an isolated bubble train could take a position it had previously jumped over. We believe this happens infrequently. 29 FLOWING FOAMS IN POROUS MEDIA, II Side lamellae jump only if the side bubbles are highly compressible, i.e., if Ks is large in Eq. [26]. In addition, side lamellae jump only if the bubble in the train is compressed to a relatively high pressure. This high pressure is most likely if there is a large, positive Ap for the forward lamella, i.e., a position in the pore throat. A sudden decompression behind this lamella would merely cause it to retreat slightly and spend more time in the throat, not to jump back into the preceding pore body. Individual lamellae do retreat when, at higher pressure gradients, the bubbles in multiple bubble trains converge on the same pore body (22, 23 ). We do not expect this to be a c o m m o n occurrence at the threshold of flow at low Vp, where isolated bubble trains move separately through the porous medium ( 3, 4). A U T O C O R R E L A T I O N OF L A M E L L A POSITIONS In a real bubble train, the lamellae act as the "pistons" for each other. In a compressible foam, when one lamella pauses in a pore throat, it stops pushing and pulling, respectively, the lamellae ahead of and behind it. As a result, these lamellae advance more slowly, and they in turn displace the lamella between them from its pore throat more slowly. Thus, lamellae in real bubble trains spend more time in pore throats than they do in the model of relentless pistons, behaving, in effect, as if they were more compressible. When jumps do occur, all the lamellae tend to jump together. Model for Multiple Lamellae Our model for a train of many bubbles envisions steadily advancing pistons separated by (n + 1 ) bubbles and n lamellae. While relentless pistons on each end of the train drive its overall movement, in the interior of the train lamella movement is governed by the movement of adjacent lamellae. Our algorithm selects random and uncorrelated bubble volumes at a fixed reference pressure and then tracks bubble progress through a chain of uniform pores or repeating sequences of pores. In the results below we assume uniform conical pores with Pb ----"0.5, Pt = 0.1. For solving for lamella positions, we selected a slow, but, for this application, sure algorithm: successive substitution. That is, after each advance of the pistons, we allowed one lamella at a time to relax with all others fixed. Over many iterations, the entire train relaxed to equilibrium, and then we advanced the pistons again. Other algorithms, more efficient for most applications, cannot preclude overshoot of lamella position in an iteration followed by backwards "correction" into a position the lamella would not have taken in forward flow. An additional problem with derivative-based algorithms such as Newton-Raphson (24) is that the functions required to form the Jacobian possess sharp corners and discontinuities (see, e.g., Figs. la and 8 below). Even taking small time steps cannot preclude large overshoots because there are times when lamellae 1 Lamella AP/Pe +'t--o _lqO ~----~1~ r 10 hamellae NO. 1 Ap/pee ~/JGi' ~ 2 ! 3 ! 4 0-- t 5 o- t 6 o 7 8 o~ o ~ 10 0 ~_t ! ~]..__.._._ I ~''-'~-- +10 Total _,°r ~ _ ~ ~/z FIG. 8. Capillary Ap over time for trains of 1 and 10 lamellae. K = 0. Journal of Colloid and Interface Science, Vol. 136,No. 1, April 1990 30 W I L L I A M R. R O S S E N must make large jumps across these corners and discontinuities. 1 Lamella +1 zxP/Pc 0 4 -1 u 10 Lamellae Results Figure 8 contrasts how Ap varies with time ~- for trains of 1 and 10 lamellae between incompressible bubbles ( K = 0). The top curve, for one lamella, is the same as in Fig. 3 of Ref. (2), except that here we have not assumed that the lamella passes through the pore throat at ~- = 0. The dot and arrow to the right of the Ap curve in Fig. 8 indicate the time-average #i and standard deviation ag of Ap for each lamella during transit through the pore. The rapid rise in Ap corresponds to the rapid movement of the lamella through the pore throat, from ( - - ( A p ) max) to (Ap) max. The smaller, discontinuous drop corresponds to the lamella j u m p at the pore body. In the train of 10 lamellae, the fluctuations in ~p are uncorrelated because the bubble volumes and hence lamella positions are uncorrelated, as we assumed in Ref. (2). The time-average total Ap is 10 times the time-average Ap for the 10 individual lamellae, and the standard deviation over time of total Ap across the train, o-, is 1/~ times the standard deviation over time of ~xp for the individual lamellae. Thus, the model of relentless pistons, and the random-walk model based theoreon (2), is reasonable for a long train of incompressible bubbles, if the movement at the ends of the train is steady. Figure 9 illustrates how £xp varies with time for trains of 1 and 10 lamellae with K = 1. The steady rise in Ap for one lamella corresponds to slow lamella movement through the pore throat (cf. Figs. lc and 2c for K = 0.6). The discontinuous drop in Ap corresponds to the jump from the pore throat to the diverging pore body, and the nearly steady Ap thereafter (actually, a slight decline) corresponds to steady lamella movement to the corner at the pore body. Another discontinuous jump from the pore body to the pore throat begins the steady buildup of ~p to its maximum again. Journal of Colloid and Interface Science, Vol. !36, No. 1, April 1990 r °i NO. 1 /XP/P e 0 -1~ 2 0 4 0 5 0 6 0 7 0 8 0 9 0 ~ ~z/'t' I-! I--I t o +10 Total ~ 0 3 lO t"~l +1 • 0 -10 FIG. 9. Capillary Ap over time for trains of 1 and 10 lamellae. K = 1. The most striking feature of the Ap curves for a train of 10 lamellae is the apparent lack of action except near the pistons. Lamellae numbers 4 to 9 are at virtually the same Ap at all times. Even near the pistons there are only three events where Ap jumps discontinuously. The time-average Ap for the train of 10 lamellae is 0.79, not 10 times but 18 times that for the single-lamella case; and the variance over time of total Ap is nearly zero for the train of 10 bubbles, while it is substantial for the single-lamella case (Fig. 8 ). Figure 10 makes plain that autocorrelated lamella jumps cause the higher Ap for the 10bubble train in Fig. 9. This figure illustrates the position of each lamella within its pore as well as the relentless movement of the pistons. Although plotted as though the lamellae are in adjacent pores, bubbles in highly compressible foams are, in general, many pores long (Eq. [ 8 ] ). Nine out of ten lamellae in this train jump at the same time; eight out of ten lamellae reside exclusively in pore throats; and FLOWING FOAMS Piston _ _ _ _ _ - - - - - - - - - - - ~ 10 Lamellae lmella NO. 10 9 8 7 6 5 4 3 aF 7 1j p~stOe_----------~---- 'L--" T 1 T 1 =1 .= FIG. 10. L a m e l l a p o s i t i o n s o v e r t i m e for trains o f 1 a n d 10 lamellae. K = 1. seven out of ten reside always at virtually the same 2xp within the throats (Fig. 9). The autocorrelation of lamella jumps in long trains allows lamellae to spend more time in pore throats, behaving in effect as though K were higher. IN POROUS MEDIA, II 31 Figure 11 shows how (2xp) avg varies with K for trains of one lamella and 10, 100, and 300 lamellae (cf. Fig. 3 ). Because the set of initial bubble volumes is a random variable and affects (2xp) avg, (2xp) avgitself varies from case to case for trains of many lamellae: in fact, by as much as 0.12p~ for medium-length trains (3 to 30 lamellae) with intermediate values of K (0.03 to 0.3). In Fig. 11, we plot the expected value of (Ap) avg averaged over enough realizations that its uncertainty is less than 0.02p e. Figure 11 shows that at low, finite compressibility, longer trains behave as though bubbles were more compressible. (All train lengths have the same (2xp) avgfor K = 0, however.) For given K, as train length n increases, (Ap) avg approaches a limit below (Ap) m a x i.e., infinitely long trains do not behave as though they were infinitely incompressible. As K increases, behavior approaches the asymptotic, long-train (2xp) avgat smaller values of n. The behavior of infinite-length trains is academic, however, because in real bubble trains bubble separation and snap-off acts periodically to truncate train length, especially at low capillary pressures (2). When bubble separation or snap-off occurs, a liquid-filled throat , I (Ap)ma x _-- I . . . . 1.0 (ZXp)avg P~ 0.5 0 0.01 i I i 0.1 1. 10. K FIG. 11. Effect o f b u b b l e - t r a i n length n a n d c o m p r e s s i b i l i t y f a c t o r K o n (2xp) avg for c o n i c a l p o r e s w i t h Pb = 0.5, 0t = 0.1. Journal of Colloid and Interface Science, Vol. 136, No. 1, April 1990 32 W I L L I A M R. R O S S E N temporarily isolates the movement of the lamellae upstream from those downstream, severing the autocorrelation of movement of these lamellae. (Bubble separation and snap-off do increase (Ap)avg by other means (2).) We expect train length in real foams to be from a few to a few tens of bubbles (2). Figure 11 shows that for some values of K these train lengths could more than double (Ap) avg. DISCUSSION Compressibility significantly increases the minimum pressure gradient for foam flow, ( ~ p ) rain if either gas compressibility or bubble size is large (Eq. [6]). As compressibility increases, the random-walk contribution (2) to (Vp) min decreases because a, the standard deviation of the Ap between bubbles, decreases (Fig. 3). In addition, trains of compressible bubbles have yet-smaller fluctuations in pressure because lamella movement is correlated (Fig. 9). However, for the model pore shape examined (conical pore, Pb = 0.5, Ot = 0.1 ), the increase in the average ~p between bubbles (Ap) ~vg more than compensates for the drop in ~. As compressibility factor K increases from zero to infinity (Fig. 3), (~p)avg increases by a factor of 10, to a value five times that for incompressible bubbles including a reasonable estimate of the random-walk effect (2). Estimates for Real Foams One can estimate K for real foams from surface tension, bubble and pore sizes, and pressure. For instance, for steam foams, we take 3/ = 30 d y n / c m , rt = 15 # m , p = 345 kPa (50 psi) (12), P b : 0 . 5 , P t = 0 . 1 , /)O/l)p - - 30, and nL 2.5 cm -1. This bubble size corresponds to bubbles about or 400 #m in diameter in bulk, roughly the size we see exiting Berea core in nitrogen-foam-propagation experiments in our laboratory. If steam were an ideal gas, its compressibility at this pressure would be 2.9 X 10 -7 ( d y n / c m 2 ) -1. However, the compressibility of pure steam in equilibrium with water Journal of Colloid and Interface Science, Vol. 136, No. 1, April 1990 is infinite; a reasonable estimate with noncondensible gas present (Appendix 1 ) would be 20 times the ideal gas figure, 5.8 × 10 -6 ( d y n / cm 2)-1. This gives a value o f K = 7, near the high-compressibility limit in Fig. 3. Autocorrelation oflamella jumps, swelling, and drainage of liquid films and "breathing" of trapped bubbles would further increase (•p)aVg. Therefore, we estimate (~xp)"Vg/p~ ~ 1.25 and (~Tp)rain ~ 1.2 M P a / m (55 psi/ft) (Eq. [1]). Finer-textured foams would have higher (~7p) min. Clearly, steam foams cannot flow from the near-wellbore region unless coalescence coarsens foam texture. For CO2 foams, we use again rt = 15 t~m Pb = 0.5, Pt = 0.1, vO/vo ~-- 30, and n~ = 2.5 cm -~. Surface tension is lower, 3' = 5 d y n / c m (18), however, and at typical reservoir conditions of 66°C and 13.8 MPa (2000 psi), k ° is much smaller, at 9.7 × 10 -9 ( d y n / cm2) -1 ((17) and Appendix 1 ). As a result, K is roughly 0.0023, at the incompressible limit on Fig. 3. For nearly incompressible foams like CO2 foams, however, compressibility-like effects are important, and these effects depend on capillary pressure and bubble size as well as on gas compressibility. For low capillary pressures and v°/vp = 30, Eqs. [27] and [30] indicate that the combined effects of swelling and draining of liquid raises Keerto the range of 1 to 3; a reasonable estimate of(Ap) avg(Fig. 3) is then 0.6p~, and (Vp) min ~ 100 k P a / m (4.4 psi/ft). Smaller bubble sizes, e.g., v°/Vp = 5, reduce Kerr toward the incompressible limit, but with the finer foam texture (n~ = 15 cm -1) even the incompressible-limit value (XTp)avg = 0.126p~ gives ( A p ) min - - 126 k P a / m (5.6 psi/ft) (Eq. [1]). Bubble separation at low capillary pressure would further increase (XTp)min. At high capillary pressures, the effect of liquid-filled throats is unimportant, because all but the tightest throats have been occupied by gas. Equation [27] and Fig. 6 indicate, however, that even at fairly high capillary pressures (Pci = 2p~), liquid films give Kerr = 0.03 for 33 FLOWING FOAMS IN POROUS MEDIA, II medium-textured foams (v°/Vp = 30); for finer textured foams ( vOb/vp = 5 ) Keff is lower (0.05), and at extremely high capillary pressures, the effect is insignificant. As shown above, "breathing" could raise Keg for CO2 foams by a factor of about 10, to 0.02 from the value of 0.002 based on intrinsic compressibility. All of these effects are too small an increase to raise (Ap)"vg by themselves (Fig. 3), and moreover, they are not additive, because the higher effective compressibility due to films reduces the "breathing" effect (Eqs. [ 16 ] and [ 18 ]). However, in this range of values of K around 0.01, autocorrelation of lamella jumps in long trains, expected at high capillary pressures, could raise (Ap) avg by a factor of 1 to 2 or more above the incompressible limit (Fig. 11). For v°/vp = 30, (Ap) avg- 0.126 to 0.25p e gives (Vp) min ~ 2142 k P a / m (0.9 to 1.8 psi/ft); for v ° / % = 5, (XTp)rain - 125-250 k P a / m (5.6 to 11 psi/ft) due to finer texture (Eq. [ 1]). Our predictions for both steam and CO2 foams reflect the pore geometry assumed, i.e., a conical pore with Pb = 0 . 5 , Pt = 0 . 1 . A n o t h e r reasonable model is a conical pore with Pb = 0.5, Pt = 0.25. If one holds r~ constant between the two models, however, (At))avg is only about 30% lower for Pt = 0.25. Our value of rt, 15 t~m, agrees with threshold pressures measured on a variety of Berea sandstones with air permeabilities of 0.3 to 0.5 #m 2 (300- 1.0 500 md) (25). Unconsolidated sands presumably would have lower ( • p ) m i n due to larger rt, and tighter rocks, higher (Vo)rain. Fit of Model to Data of Falls et al. For the atmospheric-pressure beadpack experiments of Falls et al. (3), K ranged from 0.055 to 0.43 (Eq. [ 6 ] ). "Breathing," however, both predicted by theory and observed in these experiments, would raise the effective K b y orders of magnitude to the infinite-compressibility limit and raise (Ap) avg to (Ap) max. For highly compressible foams, only the pore throat matters to (Ap) avg. Here we approximate the pore throat of a beadpack as a torus with the same ratio of the radius of the solid matrix rg to the radius o f the pore throat rt: rt/rg = 0.155. For such a torus, ( A p ) m a x / p e c = 0.54. We replot the data from Table 2 of Fails et al. in Fig. 12 with this new estimate of theory, accounting for compressibility and "breathing" of bubbles alongside the train. The new estimate of theory falls near the middle of the widely scattered data. Thus, our model fits these beadpack data well when compressibility and "breathing" are taken into account. CONCLUSIONS 1. The compressibility of the gas phase in foams can dramatically increase the minimum ~o o o o o (Ap) avg o o __Q_ . . . . a o 0.5 o o~oa J o 0~) Theory IYield Pressure Dropl | o Sustain Flow / aStart Flow ~ 110 1'5 20 Bubble Length (cm) FIG. 12. Fit of model for (Ap) a~g for compressible foams to data of Falls et al. (Ref. (3)). Journal of Colloid and Interface Science, Vo]. 136,No. l, April 1990 34 WILLIAM R. ROSSEN pressure gradient required to maintain flow of these foams through porous media, ( ~ p ) m i n . When gas compressibility is high, foam lamellae spend most of their time in pore throats, where capillary resistance to flow is greatest, and j u m p over positions where capillary forces promote foam flow. 2. Steam foams behave as highly compressible foams: low pressure and vapor-liquid equilibrium give the gas phase high compressibility. At low capillary pressures, CO2 foams also behave as though highly compressible due to the swelling and drainage of liquid films, despite the low intrinsic compressibility of CO2. At high capillary pressures CO2 foams behave as though slightly to moderately compressible. Reasonable parameter estimates give values of(Vp) rain of 1.2 M P a / m , 125 k P a / m , and 20-40 k P a / m (55, 6, and 1-2 psi/if) for medium-textured stream, low-capillary-pressure CO2, and high-capillary-pressure CO2 foams, respectively, in rock. Finer textured CO2 foams have a more substantial (Ap) min. Clearly, limited coalescence is essential for deep foam penetration into the formation. 3. The effect of compressibility depends on both bubble size and gas compressibility. Large, moderately compressible bubbles, occupying many pores, behave like smaller, highly compressible bubbles. 4. The contribution of bubble separation t o ( ~ p ) m i n decreases for highly compressible foams because the magnitude of the pressure fluctuations along the train of bubbles decreases. However, the net effect of compressibility is to raise (Vp)min significantly. 5. ( V p ) rain for highly compressible foams depends only on pore-throat geometry. In porous media of widely varied throat sizes, (~p) minfor highly compressible foams depends only on throats represented in the "threshold pressure" (10) of the porous medium as a whole. 6. Other phenomena increase the effective compressibility of foams: (a) Compression and decompression of trapped-gas bubbles alongside the train can Journal of Colloid and Interface Science, Vol. 136,No. 1, April 1990 raise the effective compressibility of a foam by orders of magnitude in beadpacks. The effect on flow of foam through rock can be significant in some cases as well. (b) At low capillary pressures, the effect of swelling and drainage of liquid films and liquid-filled throats can establish a substantial effective compressibility even for foams of an incompressible gas. At high capillary pressures, these effects are negligible, however. (c) In long trains of compressible bubbles, lamellae tend to stop in pore throats simultaneously, stay there longer, and jump together. This can increase (Ap) avgsubstantially for long trains. 7. The theory of (Vp) rain of compressible foams fits the beadpack data of Falls et al. (3) well within the scatter of the data. APPENDIX 1: COMPRESSIBILITY OF STEAM AND CO2 FOAMS The compressibility of gases in equilibrium with liquid is complicated by the exchange of mass with the liquid. For instance, the isothermal compressibility of pure steam in equilibrium with pure water is infinite: changes in system volume merely cause condensation of steam or evaporation of water, with no change in pressure (26). For steam foams compressibility is limited by the concentration of noncondensible gas (27); for CO2 foams it is limited by the solubility of CO2 in water. For steam foams we make assumptions similar to those in Ref. (27). Specifically, we assume equilibrium between an ideal gas mixture of an insoluble, noncondensible gas and water vapor and an aqueous phase whose properties are constant. With these assumptions, one can easily show that the compressibility of steam foam increases proportionately to the inverse of the mole fraction of the noncondensible gas. As injected, the vapor phase of steam foams may contain from a few hundredths of a percent to a few percent of noncondensible gas (27). Therefore, the mole fraction of noncondensible gas in the steam- FLOWING FOAMS IN POROUS MEDIA, II swept zone is expected to be low, e.g., only a few percent, and the increase in compressibility great. For CO2 foams, the effect of dissolution of CO2 depends on the mass of water in which it can dissolve, i.e., on the rate at which CO2 can diffuse away from the bubble. Here we simplify the rate issue by assuming, generously, that CO2 in a bubble can equilibrate with five times the bubble volume of water during the period of interest. By interpolating from data in Ref. (28), we determine that at 66°C and 13.8 MPa (2000 psi), approximately 7 × 10 -6 additional g of CO2 dissolve in 1 g water due to a pressure increase of 6.9 KPa ( 1 psi). Therefore, the contribution of dissolution of C O 2 to kT is k s O. = l= ms z rnl = n z /7 L = nsi = Pl, P2 = kT ~ 5(7 × 10-6)/Pco2 - 7.5 × 10 -5 psi -l, [A1] where pco2 is the density of CO2, 0.45 g / c m 3. This result is an order of magnitude less than the contribution due to the compressibility of the CO2 itself (17). APPENDIX 2: NOMENCLATURE K ~ Ke¢ = Kfz Ks= k z ko= compressibility factor in foam flow (Eq. [6]). effective value of K, increased by "breathing" of side bubbles (Eq. [ 18 ]) and by drainage of liquid films (Eq. [25]) and of liquid-filled throats (Eq. [ 29 ] ). compressibility factor for effect of liquid films (Eq. [ 26 ] ). compressibility factor for effect of "breathing" of side bubbles (Eq. [161). linear coefficient of compressibility (Eq. [41). conventional coefficient of compressibility of gas phase, ( - 0 In v~ Op)T, at the average pressure of the foam. p~= Pci = e z Pcs0 p~ = pi = Psij = Ps = ap= ( A p ) max = (ap)av~ = An/avg ~/~ t tot = 35 linear coefficient of compressibility of trapped bubbles alongside train. length of a pore. total volume of side bubbles relative to volume of bubbles in train (Eqs. [181). average number of liquid-filled throats adjacent to each pore in the bubble train. number of lamellae in a bubble train. foam texture: number of lamellae per unit length in the direction of flow. number of lamellae in throats alongside bubble i in the train. for a pair of bubbles, the pressure of the rearward and forward bubbles, respectively. capillary entry pressure of pore throats in the bubble train (Eq. [71). capillary pressure of bubble i in a train. capillary entry pressure of throat ij alongside bubble train. representative average value of pecsij (Eq. [18]). the pressure of bubble i in the bubble train. pressure in trapped bubble ij alongside bubble train. representative average value of Ps0 (Eq. [181). difference in pressure between two adjacent bubbles in a train; positive if the forward bubble is at lower pressure (Eq. [ 3 ] ). maximum value of £xp--occurs in pore throat. average capillary Ap per lamella, not accounting for lower bound in capillary pressure (Eq. [ 2 ] ). average capillary Ap per lamella, accounting for lower bound in capillary pressure ) Eq. [ 1] ). Journal of Colloid and Interface Science, Vol. 136, No. 1, April 1990 36 WILLIAM R. ROSSEN (~7p) min = minimum R z rb = rf = r1 rt = r s ij = rts = rts0. z I) 1 = v°= VOs = /)p = AVf = Avsij = mrs A/)t = X z -y= #= pressure gradient to maintain foam flow. bead radius in a beadpack. pore-body radius. radius of curvature of liquid film in pore. radius of curvature of lamella; positive if convex toward rearward bubble. pore-throat radius. radius of the side lamella ij. average throat radius of liquidfilled throats alongside the bubble train. radius of throat ij alongside bubble train. volume occupied by bubble 1 in the pore of interest. average bubble volume in bubble train. average volume of a trapped bubble alongside the bubble train. volume of a pore. the change in volume of water films in one pore. change in volume of bubble i due to movement of lamella ij alongside train. representative average value of Avsij (Eq. [181). change in bubble volume due to gas intrusion into a liquid-filled throat. axial position of lamella attachment to the pore throat. gas/liquid surface tension. the mean of Ap(r); equivalent to ( A p ) avg p b = dimensionless pore-body radius, rb/l. p f = dimensionless radius of curvature of liquid film in pore, rf/I. p t ~- dimensionless pore-throat radius, rt/l. Pts = dimensionless throat radius of liq- uid-filled throats alongside bubble train, rts/l. Journal of Colloidand InterfaceScience, Vol. 136,No. 1, April1990 the a = the standard deviation of Ap(z) for the entire train. ~-= dimensionless time: fraction of pore transit time. ACKNOWLEDGMENTS We thank P. A. Gauglitz and F. Friedmann for fruitful discussions. L. A. Young helped accelerate the execution of the algorithm used to prepare Fig. 11. REFERENCES 1. Bailey, R. E., and Curtis, L. B., "Enhanced Oil Recovery." National Petroleum Council, Washington, DC, 1984. 2. Rossen, William, W. R., J. Colloid lnterface Sci. 136, 1 (1990). 3. Fails, A. H., Musters, J. J., and Ratulowski, J., SPE Reservoir Eng. 4, 155 (1989). 4. Prieditis, J., and Flumerfelt, R. W., in "SurfactantBased Mobility Control: Progress in Miscible-Flood Enhanced Oil Recovery" (D. H. Smith, Ed.), ACS Symposium Series No. 373, p. 295. American Chemical Society, Washington, DC, 1988. 5. Rossen, W. R., "Theories of Foam Mobilization Pressure Gradient," SPE 17358, presented at the SPE/DOE Symposium on EOR, Tulsa, April 1720, 1988. 6. Rossen, William, W. R., J. Colloid Interface Sci. 136, 38 (1990). 7. Stauffer, D., "Introduction to Percolation Theory." Taylor and Francis, London, 1986. 8. Larson, R. G., Davis, H. T., and Scriven, L. E., Chem. Eng. Sci. 36, 75 (1981). 9. Larson, R. G., Scriven, L. E., and Davis, H. T., Chem. Eng. Sci. 36, 57 (1981). 10. Larson, R. G., and Morrow, N. R., Powder Technol. 30, 123 (1981). 11. Ransohoff, T. C., and Radke, C. J., SPE Reservoir Eng. 3, 573 (1988). 12. Ransohoff, T. C., "Foam Generation in Constricted Noncircular Capillaries and Bead Packs," pp. 202. MS thesis, U. Calif. Berkeley, 1986. 13. Katz, A. J., and Thompson, A. H., Phys. Rev. B 34, 8179 (1986). 14. Hong, K. C., SPE Reservoir Eng. 2, 67 (1987). 15. Heiba, A. A., Sahimi, M., Scriven, L. E., and Davis, H. T., "Percolation Theory of Two-Phase Relative Permeability," SPE 11015, presented at the Annual Meeting of the Soc. Pet. Eng., New Orleans, Sept. 26-29, 1982. 16. Heiba, A. A., Davis, H. T., and Scriven, L, E., "Statistical Network Theory of Three-Phase Relative Permeabilities," SPE 12690, presented at the SPE/ DOE Symposium on EOR, Tulsa, 1984. FLOWING FOAMS IN POROUS MEDIA, II 17. Canjar, L. N., and Manning, F. S., "Thermodynamic Properties and Reduced Correlations for Gases," p. 131. GulfPubl., Houston, 1967. 18. Wellington, S. L., and Vinegar, H. J., J. Pet. Technol. 39, 885 (1987). 19. See, e.g., Selby, S. M., "Standard Mathematical Tables," 18th ed., p. 17. Chemical Rubber, Cleveland, 1970. 20. Thompson, A. H., Katz, A. J., and Krohn, C. E., Adv. Phys. 36, 625 (1987). 21. Roof, J. G., Soc. Pet. Eng. J. 10, 85 (1970). 22. Huh, D. G., Cochrane, T. D., and Kovarik, F. S., SPE Reservoir Eng. 4, 872 (1989). 23. Shirley, A. I., in "Surfactant-Based Mobility Control: Progress in Miscible-Flood Enhanced Oil Recov- 24. 25. 26. 27. 28. 37 ery" (D. H. Smith, Ed.), ACS Symposium Series No. 373, p. 234. American Chemical Society, Washington, DC, 1988. Carnahan, B., Luther, H. A., and Wilkes, J. O., "Applied Numerical Methods." Wiley, New York, 1969. Wunderlich, R. W., personal communication, 1984. Steam tables, e.g., in Weast, R. C., Ed., "Handbook of Chemistry and Physics," 53rd ed., p. E-12. Chemical Rubber, Cleveland, 1972. Falls, A. H., Lawson, J. B., and Hirasaki, G. J., J. Pet. Technol. 40, 95 (1988). Stalkup, F. I., "Miscible Displacement," p. 142. Soc. Pet. Eng., New York, 1983. Journal of Colloid and Interface Science, Vol. 136, No. 1, April 1990
© Copyright 2026 Paperzz