Professional Documents
Culture Documents
Etched-silicon-wafer micromodels of the type described by four ports (one for each hole of the micromodel) that are
Sagar and Castanier6, initially developed by Hornbrook et al.7, sealed with viton O-rings, Fig. 3.
were used. They contain a repeated pattern obtained from an
SEM (Scanning Electron Microscopy) image of a rock thin Imaging System. A Nikon Optiphot-M reflected light
section, as shown in Fig. 1 for Berea sandstone. These microscope with a photo tube allowing for video imaging was
micromodels offer 1:1 size scaling of typical sandstone pores. used to view fluid movement. Because the silicon
Note the grains appear as islands and the etched pores and micromodels are opaque, an internal light source sends light
throats as channels. The etch depth is 25 µm and grains range through the objective lens to illuminate the micromodel.
in size from 30 to 200 µm, Fig. 2a. The porosity is roughly 0.2 Generally, 50X magnification was used to track the motion of
and the permeability is approximately 500 mD. The total the fluids in the micromodel. Greater magnification, 100X and
network dimension is 5-cm square, as shown in Fig. 2b. Such 200X, used to focus on a few pores.
micromodels have been used to study multi-phase flow, oil- A video camera (TMC-7RGB, PULNiX) connected to a
foam interaction, solution gas drive, and contaminant standard video cassette recorder (VCR) was used to record
hydrogeology.1,4-6 events at a speed of 30 frames per second. Then selected
These micromodels contain some modifications as images from the VCR were transferred to a Macintosh/PC
compared to those used previously4,6: narrow etched channels computer system with the aid of imaging video/recording
adjacent to the edges of the pore network carry fluid to the software. A commercial home-video digital camera provided a
inlet/outlet ports. These channels allow fluid to enter and exit second imaging device to obtain images of the entire
the porous medium at any point along the side of the pore micromodel (macro-scale) at different stages of the
network, Fig. 2b, similar to a fracture. Holes are drilled experiments. In these pictures the pattern and extent of
directly into the silicon wafer at each corner of the imbibition in the entire micromodel was observed.
micromodel providing locations for fluid exit and entry into
the “fractures”. Pump, Filter and Fluids. An ISCO 100DM syringe pump
Micromodels were fabricated at the Stanford (Instrumentation Specialties Co.) was used to deliver water or
Nanofabrication Facility (SNF). There are two main steps in oil to the micromodel. This pump is accurate to exceptionally
their construction: fabrication and anodic bonding. Fabrication low flow rates; i.e. 0.001 cm3/min to 0.0001 cm3/min. A
starts with the design and construction of a mask. The mask is Swagelock filter with a 2-micron cartridge was used to filter
placed on a silicon wafer previously coated with a out any small particles that might clog the micromodels. The
photosensitive chemical. The pattern of the mask is impressed properties of the fluids used in the experiments are shown in
on the wafer by means of passing ultraviolet light through the Table 1.
mask. The silicon wafer is developed and later etched with
hydrofluoric acid (HF) to a depth of around 25-30 µm. In this Experimental Procedure.
way pore cavities are created. Constant water injection rates into the fracture were set at two
The silicon wafer with the etched pattern is cleaned in a values: one low and one high relative to the micromodel
sulfuric acid (H2SO4) cleaning solution (NOCHROMIX, (0.001 cc/min and 0.01 cc/min). The fracture aperture was
Godax Laboratories) air dried, and then bonded to a glass plate constant in depth and width as per the fabrication etching. This
to create a two-dimensional porous medium. The procedure system mimics length, width, and approximate fracture
for bonding is simple, yet extremely important. Firstly, the aperture of the core scale set-up used with the CT Scanner to
cleaned micromodel is placed with the pattern facing up on a image in situ saturation reported in Ref.1. We performed
pre-heated hotplate at 600° F. This hotplate is connected to experiments similar to those at the core-scale trying to achieve
one of the electrodes (anode) of a power supply. The filling-fracture and instantly-filled fracture regimes using the
micromodel is left on the plate for around 45 minutes. During low and high flow rates, respectively. For reference, Figs. 3.5,
this step the silicon is oxidized resulting in the formation of a 3.6, 3.7 and 3.9 in Ref. 1 display typical core CT experiment
solid film of silicon dioxide that is water wetting. The results.
temperature is then reduced to 500° F. After the hotplate For the experiments reported here, water was injected on
temperature has stabilized, a clean glass plate (cleaned with the left at constant rate and oil and water produced from the
soap solution, then sulfuric acid solution, and then air dried) is right-hand side. The intent was uniform advance of water
placed on the micromodel. The other electrode (cathode) is along the fracture. We performed experiments with water and
placed on the top of the plate and the voltage on the power air, water and CO2, oil and CO2, and water and oil. In the case
supply is brought to 1000Volts. After 50 minutes, bonding is of experiments with oil, CO2 was first injected into the system
achieved. The electricity is turned off and the new micromodel to displace any gas and this was followed with oil. The CO2 is
is allowed to cool to room temperature. relatively soluble in oil and the combination allowed the
micromodels to be fully saturated with oil.
Micromodel Holder.
A holder to mount the micromodels for unobstructed viewing Experimental Results.
under the microscope was specially designed, Fig. 3. This Following the aforementioned procedure, experiments were
aluminum (6061 T6) micromodel holder has conduits conducted to achieve filling-fracture and instantly-filled
connected to the injection and production systems. There are fracture regimes using low and high injection rates,
respectively.
SPE 92133 3
front was found to be faster than that of the previous As oil blobs grow in the fracture, they block the flow of water
experiment (gas 'encapsulated') and very little gas was and increase the viscous pressure gradient along the fracture.
trapped. These results support the assertion that co-current This process is intermittent; i.e., a blob grows, it is displaced,
imbibition is a more efficient mechanism compared to and after some time a new blob starts to grow in the same
counter-current imbibition.1-3 location. However no constant periodicity was found.
The behavior of this second regime is similar to that
“Instantly Filled” Fracture Regime observed during both counter current and co-current
In this regime, little water imbibes before the fracture fills imbibition experiments in the previously presented CT core-
with water. Figure 9 includes the images obtained with the scale experiments (Ref. 1). These observations support the
digital camera of the entire micromodel for different times for work presented by Gautem and Mohanty8 that emphasizes the
a water-air experiment. It shows that the water advance in the importance of fluid velocity in fractures with respect to the
direction perpendicular to the fracture is nearly one- balance of capillary and viscous forces.
dimensional. Another important feature of this type of
displacement is the production of non-wetting fluid through Discussion.
“channels”. Refer to image labeled 25 min. on Fig. 9, for The micromodel experiments shed additional light on other
instance. An analysis similar to Eqs. 1 to 4 again explains the studies of imbibition. Classic experiments on imbibition in
origin of channels. fractured media are presented by Kleppe and Morse9 who
Figure 10 shows microscopic images obtained during the evaluated the effect of injection rate and fracture flow capacity
“instantly-filled fracture" regime focusing on the matrix- on the production history and oil recovery from a single
fracture interface. Almost no water imbibes while the fracture matrix block fractured system. As opposed to the classic
fills. The water fills the fracture very fast and the air becomes instantaneous immersion (instantly-filled fracture) in water,
“encapsulated” inside the micromodel; therefore, the non- Kleppe and Morse9 flooded their cylindrical core surrounded
wetting fluid is only expelled by imbibition that is both by a small space around the core (an artificial fracture) at a
microscopically and macroscopically countercurrent. Figure constant injection rate. They reported results for injection rates
11 shows images of the water advance in the matrix in the of 3.3 cm3/min and 35.0 cm3/min, that are commonly referred
direction perpendicular to the fracture during the “instantly- to in the literature as the 'low rate' and 'high rate' experiments.
filled fracture” regime. The water front has advanced almost Dutra and Aziz10 presented a diffusion-type theory to model
the same distance on the left as it has done on the right. See recovery by capillary imbibition. The main parameter
the dashed and solid lines indicating the water advance in the investigated to model the experiments of Kleppe and Morse9
matrix. As it was observed macroscopically in Fig. 9 (and the was the hydraulic diffusivity, Dcw.
CT experiments of Ref. 1), there is a well-defined spreading An interesting observation is that even though the flow rate
front parallel to the surface exposed to the imbibing fluid. in the low rate case is almost 10 times smaller than that of the
Figure 12 shows the same water advance at 10X high rate experiment, a plot of the non-wetting fluid produced
magnification close to the “channels”. There is a large amount versus the water injected shows that the low rate case is more
of gas trapped/bypassed by water in these areas. These efficient on a pore volume injected basis.
channels are not always completely connected throughout the Following the observations in the micromodel
entire micromodel. They connect and disconnect as the water experiments, a small water injection rate leads to a
advances and displaces air counter-currently. displacement by imbibition in a cocurrent mode. Moreover,
Experiments similar to the instantly-filled fracture regime for an injection flow rate under or close to the critical rate, as
were performed with water displacing CO2 with similar it seems to be in Kleppe and Morse's9 low rate case, the
results. In this case, some of the gas dissolved into the water advance in the fracture is controlled by the interaction between
during displacement. The trapped gas eventually dissolved the matrix and the fracture. Although in absolute time the
completely and was displaced at the end of this water-CO2 experiment in the low rate case (slow filling-fracture regime)
experiment. This was done with the idea of eventually having may take longer than in the high rate case (fast filling and
a matrix saturated with oil to the irreducible water saturation. filled-fracture regime) to achieve a given recovery, it
Once the matrix was completely saturated with water, oil was represents a more efficient displacement in terms of recovery
injected to displace the water. versus the volume of water injected.
The last experiment, and the most interesting, was The values for hydraulic diffusivity used by Dutra and
conducted by injecting water into an oil-filled matrix at Aziz10 to match the experiments by Kleppe and Morse9
irreducible water saturation. Water was injected at 0.01 cc/min represent a second physical explanation to support the results
and little water imbibed while the fracture filled. Water of this paper. For the high rate case at early times, the
penetrated the matrix displacing oil by counter-current experiments are well matched by a curve corresponding to a
imbibition. value of Dcw equal to 1.398 cm2/min. As shown
Figure 13 shows that the oil is displaced through experimentally, both at core- and pore-scale, the filling-
"channels" at discrete locations along the fracture. One sees fracture regime yields more efficient displacement. The
that blobs of oil grew into the fracture. The oil mobility is hydraulic diffusivity, therefore, must be larger than that for the
lower than water mobility, so viscous forces in the fracture counter-current mode. As the high rate experiment progresses,
cannot move the oil along the fracture. Once the blobs reach the results are better matched by lesser values of hydraulic
some critical volume, however, they are displaced by the diffusivity. This indicates that once the water front in the
viscous forces in the fracture at the same speed as the water fracture has reached the end of the fracture, the recovery
SPE 92133 5
switches to a counter-current imbibition mode and lower Dcw. Relative permeability functions can be X-type, Corey-type
Now, the non-wetting fluid can only be expelled through the or some other type depending on the interference between
two-phase region, reducing oil recovery efficiency. phases1. Results presented here make clear that the matrix-
On the other hand, low rate cases seem to be independent fracture interaction also affects significantly the shape of such
of the hydraulic diffusivity. Recovery is only controlled by the functions. The interaction with the matrix is strongly related to
matrix imbibition potential, and as long as there is enough the injection flow rate and to a lesser extent to the fracture
water injected into the system to match the imbibition rate, the aperture, as discussed in the experiments at core- and pore-
recovery curve behaves independent of Dcw. Non-wetting scale. Thus, it is expected that the flow behavior in fractures
fluid, is produced downstream of the water front through the connected to matrices is much different from pure fracture
single-phase region. flow; and therefore, the curves obtained for pure fracture flow
The observations from the micromodel experiments also are not necessarily applicable to fractures interacting with
suggest that for water-oil systems, the generalized Darcy matrices.
model might not be appropriate for flow in filled fractures.
The porous medium approach does not allow the possibility of Conclusions
‘blobs’ of one phase transported by another continuous phase. This paper explored hypotheses on the mode of imbibition for
It states that one phase can only move upon establishing a filling- and instantly-filled fracture regimes. Micromodels
continuous flow path. Substituting the equation for were used to achieve detailed visualization of matrix-fracture
permeability of fractures proposed by Witherspoon et al.11: interaction, water advance in the fracture and the pattern,
extent, and mode of imbibition in the matrix at the pore scale.
k f = 84x10 6 w 2f (5) Two different modes of matrix and fracture fill-up found
previously2 were verified and analyzed at the pore-scale. The
”filling-fracture” regime represents a slow advance through
into Darcy’s law results in the following equation for water
fractures resulting from rapid fluid transfer from the fracture
relative permeability in a fracture: to the matrix, wide fracture aperture, and/or slow water
injection. Recovery scales linearly with time under this
vw
k rwf = −12 µ w (6) particular co-current imbibition mode. The "instantly-filled
2 dp fracture" regime represents rapid flow through fractures due to
wf dx
relatively low rates of fracture to matrix transfer, narrow
Eq. 6 shows clearly that different relative permeability apertures, and/or high water injection rates. Recovery scales
curves result from different pressure gradient in the fracture. with the square-root of time in this countercurrent dominated
When capillary pressure in the matrix is large relative to the imbibition mode. Experimentally, cocurrent imbibition was
viscous force in the fracture, oil is expelled into the fracture at found to be more efficient than countercurrent.
a rate independent of the viscous forces. The velocity of the For a low water injection rate and low viscosity non-
water, the oil blob growth, and the roughness of the fracture wetting fluid, virtually all of the non-wetting phase saturation
wall determine the pressure gradient at which the oil blobs are above the residual saturation is expelled from the matrix by
pushed downstream; however most numerical reservoir the time water reaches the end of the fracture in roughly
simulators are written under the assumption of a Darcy flow square matrix blocks. Although this filling-fracture regime
model for the fractures. Therefore, specific relative may take longer in absolute time than a fast
permeability curves have to be obtained for a given injection injection/production rate, it represents a more efficient
flow rate. displacement in terms of recovery versus volume of water
The ‘blob transport’ observed in the water-oil micromodel injected. Reinjection in a fractured reservoir that experienced
experiments suggests that the ‘Equivalent Homogeneous early water breakthrough caused by high injection rate is
Flow’ model for fractures as proposed by Fourar et al.12 might expensive. When the fractures surrounding blocks are filled by
be more appropriate for describing the fracture relative water, non-wetting fluid from these blocks can only be
permeability curves. The homogeneous flow model is based produced by the less effective countercurrent imbibition
on pipe flow models to treat two phases as a single mechanism.
homogeneous phase. Flow rate averaged values are defined for It was found that the mechanism for the more efficient
density, velocity and viscosity. The pressure gradient is displacement within the filling-fracture regime is imbibition in
calculated by analogy to flow between rough parallel plates. a microscopically cocurrent mode. Non-wetting fluid remains
The homogeneous flow model, however, has some relatively continuous through the matrix and fracture systems
drawbacks when modeling flow in fractures connected to as the water front advances.
matrices. As observed in the micromodel experiments, new oil Because matrix-fracture interaction is controlled by the
mass is incorporated into the fracture as a result of the water injection rate, it is apparent that the relative permeability
imbibition into the matrix. Both blob growth and transport relationships for wetting and nonwetting phases are rate
effect the pressure drop along the fracture. Oil is expelled dependent. Wetting and non-wetting fluids flow
from specific locations that are not regularly distributed along simultaneously in both fractures and the matrix. The
the fracture; therefore, the pressure distribution is not linear combination of these two flow processes results in a combined
along the fracture, and the rate fractions of oil and water are relative permeability behavior that has not been determined.
not constant in time.
6 SPE 92133
References Nomenclature
1. Rangel-Germán, E.R., 2002: “Water Infiltration in h depth of the porous media [µm]
Fractured Porous Media: In-Situ Imaging, Analytical k permeability [m2]
Model, and Numerical Study,” Stanford University kf absolute fracture permeability (m2)
PhD Dissertation. km matrix permeability (m2)
2. Rangel-Germán, E.R. and Kovscek, A.R.: “Experimental kr relative permeability
and Analytical Study of Multidimensional Imbibition L length [cm]
in Fractured Porous Media,” Journal of Petroleum p pressure [atm]
Science and Engineering (2002) 36, 45-60. r radius [µm]
3. Pooladi-Darvish, M. and Firoozabadi, A.: "Cocurrent and v velocity [m/s]
Countercurrent Imbibition in a Water-Wet Matrix V volume [cm3]
Block," Soc. Pet. Eng. J. 5: 3-11, 2000. W width [cm]
4. George, D.S.: “Visualization of Solution Gas Drive in wf fracture aperture (m)
Viscous Oil”, M.S. Thesis, Stanford University, γ surface tension [dyne/cm]
September, 1999. φ porosity [fraction]
5. Buckley, J.S.: "Multiphase Displacements in µ viscosity (Pa-s)
Micromodels", Interfacial Phenomena in Petroleum
Recovery, Morrow, N.R. (ed.), Marcel Dekker Inc.,
New York, NY (1991) 157-189. Subscripts/Superscripts
6. Sagar, N.S. and Castanier, L.M.: “Oil-Foam Interactions in c capillary
a Micromodel,” SUPRI-TR110 Report, Stanford e entry
University, November 1997. f fracture
7. Hornbrook, J.W., Castanier, L.M. and Petit, P.A.: g gas
“Observation of Foam/Oil Interactions in a New, High- o oil
Resolution Micromodel,” paper SPE 22631 in w water
proceedings of the 1991 SPE Annual Technical
Conference and Exhibition, Dallas, TX, October.
8. Guatam, P.S. and Mohanty, K.K.: “Role of Fracture Flow Acknowledgment
in Matrix-Fracture Transfer,” paper SPE 77336 in
proceedings of the 2002 SPE Annual Technical This work was prepared with the support of U.S. Department
Conference and Exhibition, San Antonio, TX, October. of Energy, under Award No. DE-FC26-00BC15311.
9. Kleppe, J. and Morse, R.A.: “Oil Production from However, any opinions, findings, conclusions, or
Fractured Reservoirs by Water Displacement,” paper recommendations expressed herein are those of the authors
SPE 5084 in proceedings of the 1974 Annual Meeting and do not necessarily reflect the views of the DOE.
of the Society of Petroleum Engineers, Houston, TX, Additional support from the Stanford University Petroleum
October. Research Institute (SUPRI-A) Industrial Affiliates and
10. Dutra, T.V. and Aziz, K.: “A New Double-Porosity Consejo Nacional de Ciencia y Tencnología (CONACyT),
Reservoir Model for Oil/Water Flow Problems,” SPE México is gratefully acknowledged.
Reservoir Engineering (November 1992), 419-425.
11. Witherspoon, P. A., Wang, J. S., Iwai, K., and Gale, J. E.:
"Validity of the Cubic Law for Fluid Flow in a
Deformable Rock Fracture," Water Resources Research
(1980) 16, No. 6, 1016-1024.
12. Fourar, M., Bories, S., Lenormand, R., and Persoff, P.:
“Two-Phase Flow in Smooth and Rough Fractures:
Measurement and Correlation by Porous-Medium and
Pipe Flow Models,” Water Resources Research (1993)
29, 3699-3708.
SPE 92133 7
5 cm
fracture
25 µm
5 cm
Inlet Outlet
Figure 2. Schematic of the micromodel: The network acts as the porous medium
and the adjacent channels act as the high conductivity fractures.
SPE 92133 9
Observation Inlet/outlet
Window ports
Inlet/outlet
to MM ports
Perforations for
screws O ring gap
Figure 3. Schematic of the micromodel holder. Top (left) and bottom (right) pieces.
Air
Matrix Matri
Fracture
Fracture
Figure 4.5. Microscope images of the micromodel during “filling-fracture" regime, close to inlet.
Water Air
Air
Water
Trapped
Air
Water
Figure 4.7. Water advance in the matrix in the direction perpendicular to the fracture.
Channels
5 hr 5 min final
Figure 8. Images of the micromodel showing air escaping by 'channels'.
Air
Water
Channel
matrix
water air
air
water
air air
water
water
water
Oil blob
growing
oil
water