SPE 92196 Experimental Study of Effective Diffusion in Porous Media

R. Islas-Jurez, Instituto Mexicano del Petrleo; F. Samanego-V., UNAM; and E. Luna, C. Prez-Rosales, and J. Cruz, Instituto Mexicano del Petrleo
Copyright 2004, Society of Petroleum Engineers Inc. This paper was prepared for presentation at the 2004 SPE International Petroleum Conference in Mexico held in Puebla, Mexico, 89 November 2004. This paper was selected for presentation by an SPE Program Committee following review of information contained in a proposal submitted by the author(s). Contents of the paper, as presented, have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material, as presented, does not necessarily reflect any position of the Society of Petroleum Engineers, its officers, or members. Papers presented at SPE meetings are subject to publication review by Editorial Committees of the Society of Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper for commercial purposes without the written consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to a proposal of not more than 300 words; illustrations may not be copied. The proposal must contain conspicuous acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O. Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.

Abstract In this work the effective molecular diffusion in homogeneous porous media was studied. An experimental setup was constructed to measure the effective molecular diffusion coefficient in packed unconsolidated sands, which allowed to take samples of fluids at different positions and times. The fluid samples were analyzed by gas chromatography to determine fluid concentration; these data were analyzed using the Ficks second law to calculate the molecular diffusion coefficient. These results were compared with those obtained using a different methodology based in pressure changes developed by Luna et al. A good agreement between both methodologies was obtained. Introduction Most of the Mexican oil is produced from mature reservoirs located at the offshore Campeche zone; because of this, the implementation of enhanced oil recovery (EOR) technologies is becoming rapidly necessary and important. Due to the fractured carbonated rock structure of these reservoirs, one of the most recommended technique is the injection of gases, such as nitrogen or carbon dioxide. This was shown in the giant Cantarell reservoir, where a nitrogen injection process has a better performance than expected, or in the Artesa field where carbon dioxide has been successfully injected. From this point of view, the laboratory study of the behavior of these gases is very important before the implementation of any EOR process. When a gas is injected in a reservoir, its behavior is governed by convection, dispersion and diffusion. Dispersion and convection work together especially in the fracture network, whereas diffusion is important in the matrix where fluid velocities are small; in fact, mass transfer between the matrix and its surrounding fractures is governed by gravitational segregation at short times, and molecular diffusion at long times.

Diffusion within porous media has been studied since the early years of the 20th century, especially in homogeneous media with intergranular porosity. Some works1,2 focused on finding the relationship between the molecular diffusion, tortuosity, formation resistivity factor and effective molecular diffusion coefficient (MDC); others, were interested in measuring the MDC using indirect methods.4,5,6,7 For example, one common methodology is to establish a gas flow at one end of a liquid saturated porous block, and measure the concentration changes in the resulting current flow. The problem here is that it is very difficult to measure the concentration variations in the current, because changes are generally very small. The main aim of the present paper is the molecular diffusion process of a gas within a liquid saturated porous media. In particular, sand of well defined size (0.032 0.051 cm) and porosity (0.44) was used. 3 This work presents an experimental study to find the effective molecular diffusion coefficient in porous media and its relationship to free molecular diffusion. The measured concentration variations through the porous medium are coupled with the second Ficks law for diffusion, to obtain the effective MDC. Model Consider a PVT cell with a cylindrical geometry of internal radio r and height H, where constant temperature and volume are established. A cylindrical porous matrix of porosity , was placed at the bottom of the cylinder (Fig. 1). The porous matrix height is hp < H and radio equal to that of the cylindrical cell. Initially a liquid (hexane) is saturating completely the porous medium and a gas (nitrogen) is occupying the free space at the top of the matrix, both at the same temperature and pressure. The height of the gas column is hg, and that of the liquid hl . Immediately two transport processes begin: the liquid molecules spread into the gas zone and those of gas go into the liquid. Both processes are governed by mechanisms of free diffusion and diffusion in porous media (effective diffusion), that are present until the thermodynamic balance is reached. Since the diffusive process is faster in the gaseous phase than in the liquid phase, thermodynamic stability in the former is reached first. The binary coefficient of diffusion in the liquid zone is Dl,eff, while in the gaseous zone is Dg. In the theoretical model of this study, it is considered that the diffusion coefficient does not depend on the species concentration. The process of nitrogen diffusion in hexane is governed by the diffusion equation or second Ficks law , expressed as:

www.petroman.ir

SPE 92196

2r

hg

xg

Gas

H
xl hl
Porous matrix + liquid

hp

Constant Temp. and Vol.

2). The cell was designed to be partially filled with sand, which constitutes an unconsolidated porous media. A sieve was used to have a grain size ranging between the meshes 50 and 80; in these conditions, the measured sand porosity was 44%. As mentioned previously, the experimental fluids were nitrogen and hexane. There are four sampling tubes located at different locations along the cell, where samples of fluid can be obtained. Fluids collected were analyzed by chromatographic techniques to determine the nitrogen concentration in the hexane. In addition, the system was placed in an oven to keep the temperature constant. Initially the cell is completely filled with hexane, injected through the lower valve (part 11, Fig. 2); the system is isolated until thermodynamic equilibrium is reached; then, nitrogen is injected in the upper valve (2, Fig. 2) until the nitrogen-hexane interface reaches the sand face. At this moment the system is isolated and the experiment begins. The nitrogen injection time (~1 minute) must be very small compared with characteristic diffusion time, to avoid interference between convection and diffusion.
1 2

Fig. 1- Diagram of the PVT cell and reference system used to determine the effective diffusion coefficient in porous media.

C N 2 t

= Dl ,eff

2 C N2 xl
2

(1)
x1 x2 x3 x4

Gas Chamber interface

where C N 2 is the nitrogen molar concentration, t is time, and x the vertical coordinate measured from the bottom of the PVT cell. At the beginning there is no nitrogen in the liquid zone (hexane); thus the initial condition to solve equation (1) is: C N 2 = 0 at t = 0. Since the bottom of the cell is impermeable, the nitrogen flow is zero, and the boundary condition at this point is:

Valve 1
7

Valve 2

Sand system

Valve 3

C N 2 =0 x l xl = 0

10

Valve 4

(2)
12 11

Local thermodynamic equilibrium is assumed at the liquid-gas interface; therefore, there is an infinitely small zone adjacent to the interface where hexane and nitrogen coexist in their equilibrium concentrations; this phenomenon occurs both in the liquid and gas phases. Under this assumption the boundary condition at the interface is: C N 2 (xl = hl ) = C N 2 , Eq . (3) Solving equation (1), with its initial and boundary conditions, by the method of separation of variables yields:
C N2 = C N2 ,eq C N2 , eq
m 2m + 1 2 tDl ,eff 4 ( 1) exp 2 m =0 2m + 1 h l 2

Fig. 2- Diagram of the experimental setup

2m + 1 xl cos hl 2

Once the experiment starts, the pressure in the nitrogen inlet tube is measured and recorded in a computer using a precision pressure gauge and a Sensotec system. The fluid samples obtained at different times at the different sampling tubes are stored and sealed in small containers of approximately 0.3 cm3; and afterwards they are analyzed in the chromatograph. Two experimental runs using different pressures and temperatures were carried out. The results obtained in the two experimental runs are described below. Run 1 Results The conditions in this run were: initial pressure 14.7 psia and temperature 30 oC . Table 1 shows a summary of the chromatographic results for three different fluid samples. Fig. 3 shows the behavior of pressure as a function of time.

(4) Experimental setup The experimental setup consists in an acrylic cylindrical PVT cell of internal diameter 6.57 cm and height 40 cm (see Fig.

www.petroman.ir

SPE 92196

Table 1- Variation of the nitrogen concentration and hexane for different times in the samples obtained through the valves placed at different distances from the interface (fig 2). Run No. 1.
Distance to the interface ( cm ) Concentration at 336 hours (% mass)
hexane nitrogen

Valve

Concentration at 386 hours (% mass)

Concentration at 696 hours (% mass)

Pressure ( psia )

45

35

Valve 1

4.1 9.1 14.1 19.1

99.77 99.81 99.83 99.86

0.23 0.19 0.17 0.14

99.70 99.75 99.80 99.83

0.30 0.25 0.20 0.17

99.60 99.68 99.70 99.76

0.40 0.32 0.30 0.24

Valve 2

Valve 3

25 0

20

40 Time (hours)

60

80

Valve 4

Fig. 4- Graph of the behavior of nitrogen pressure for the second run.

20

15

10

Analysis of results By using the Peng-Robinson equation of state, the equilibrium concentration of nitrogen in hexane was obtained where, for the conditions of pressure and temperature in both runs, was 0.49% of nitrogen in the liquid hexane; then, considering a zero distance, at the interface, assuming conditions of equilibrium concentration, and using the data of table 1 for experiment 1 and from table 2 for experiment 2, graphs were obtained to represent the variation of the nitrogen concentration with respect to distance and time, which are shown in Figs. 5 to 8.
0 200 400 600 800

Pressure ( psia )

0.5 0.4 0.3 0.2 0.1 0.0 0 5 10 15 20 25

Time (hours)
Mass Concentration of N2 (% mass) Fig. 3- Pressure vs. time in run 1

y = 0. 4718e

0. 0354x

R2 = 0. 9722

Run 2. In this case, the experimental conditions were: initial pressure 39.6 psia and temperature 35 oC. For this run some modifications were made in the distance between the interface and the sampling tubes (3 to 6 in Fig. 2). In this case three samples were taken from each valve at different times (Fig. 2), and by means of chromatographic analysis the concentrations of hexane and nitrogen shown in table 2 were obtained. The pressure in this run was controlled using a system that measures the pressure by means of a previously calibrated transducer. Fig. 4 shows the measured pressure values.
Table 2- Variation of the nitrogen concentration and hexane in run No. 2
Distance to the interface ( cm ) Concentration at 261 hours (% mass)
hexane nitrogen

Fig. 5- Behavior of nitrogen concentration in hexane for first run. 0.5 Mass Concentration o f N2 (% mass) 0.4
y=5E07x

+ 0. 0009x -0. 0014

9899 R 2 = 0.

Valve

Concentration at 501 hours (% mass)

Concentration at 693 hours (% mass)

0.3
Experi m ental

0.2 0.1 0.0 0 200 400 Time (hours)

Best fit A j uste

Valve 1

4.5 9.5 14.5 19.5

99.86 99.87 99.90 99.91

0.14 0.13 0.10 0.09

99.73 99.80 99.85 99.90

0.27 0.20 0.15 0.10

99.60 99.76 99.81 99.88

0.40 0.24 0.19 0.12

Valve 2

600

800

Valve 3

Valve 4

Fig. 6- Behavior of nitrogen concentration for the first run, at different times, and at a distance of 4.1 cm of the interface.

www.petroman.ir

SPE 92196

0.5
Mass concentration of N2 (% mass)

0.4 0.3 0.2 0.1 0.0 0 5 10 15

at 261 hours at 501 hours at 693 hours

With these equations, and using the second Ficks law, the effective coefficient of diffusion for run one can be calculated according to equation 1. By taking the second derivative of equation 5 with respect to distance x, gives

2C N 2 x
2

= 0.0005912e

0.0354 x

(7)

Similarly, by taking the first derivative of Eq. 6 with respect to the time, yields
20 25
Distance ( cm )

C N 2 t

Fig. 7- Behavior for second run of nitrogen concentration in the hexane

= 10 10 7 t + 0.0009 ;

(8)

substituting Eqs. 7 and 8 into Eq. 1, yields

1.0
Mass concentration of N2 (% mass)
Experi m ent al

10 10 7 t + 0.0009 = Dl , eff 0.0005912e 0.0354 x ; (9)

0.8 0.6
y = 1E07x
2

substituting in this equation the distance between valve 1 and the interface (4.1 cm), and a time of 696 hours: 10 107 696 + 0.0009 = Dl , eff 0.0005912 e 0.0354 4.1 . (10) Solving for the effective coefficient of diffusion of nitrogen in hexane gives

+ 0. 0005x + 0. 0013
2

0.4 0.2 0.0 0 200

R = 0. 9994

Dl , eff = 0.3990
400
Time (hours)

600

800

cm 2 cm 2 = 1.1083 10 4 hour seg

For run 2: Distance (Fig. 7):

Fig. 8- Behavior of nitrogen concentration for the second run, at different times, and a distance of 4.5 cm from the interface.

Using an exponential fit for figures 5 and 7, and polynomial for graphs of figures 6 and 8, the best fit equations for the behavior of data were obtained experimentally; that is, the variation of the nitrogen concentration in the hexane with respect to length x is obtained for the largest time, 696 hours in Fig. 5, and 501 hours in Fig. 7. In relation to the concentration variation of nitrogen in hexane with respect to time in both cases the best fit was obtained for the shortest distance, in which cases the correlation coefficient (R2) was nearly unity (Figs. 6 and 8). The equations obtained for run one are the following : For run 1: Distance (Fig. 5):

C N 2 = 0.4401e 0.0766 x ;
Time (Fig. 8):

(11)

C N 2 = 1 10 7 t 2 + 0.0005t + 0.0013

(12)

Using equations 11 and 12 and making a similar analysis to that of run one, but now considering a distance of 4.5 cm and a time of 501 hours, the corresponding effective diffusion coefficient is

Dl , eff = 0.3280 C N 2 = 0.4718e 0.0354 x ;

cm 2 cm 2 . = 0.9111 10 4 hora seg

On the other hand, the diffusion equation of liquid (hexane) in the gas zone of the model shown in Fig. 1 is

Time (Fig. 6):

C N 2 = 5 10 7 t 2 + 0.0009t 0.0014 .

(6)

CCn 2CCn = Dg 2 t xg

(13)

www.petroman.ir

SPE 92196

where CC n is the molar concentration of the hexane and xg is distance to the top of the cell (Fig 1). The initial condition necessary to solve equation 13 is: in t = 0; that is, initially there is no hexane in the gaseous zone, and boundary conditions can be considered similar to those of equation 1, namely

2 m + 1 2 2m + 1 2 d l t exp d gt (1 + 1 / a1 )exp 8a 2 2 p = peq + 21 (pini peq ) 2 ( ) m 2 1 + m=0

, where the parameters a1 , dl, y dg are defined as

(19)

CC n =0 x g xg = 0
and

(14)

a1 =

C N 2 , eq (z g zl )RThg

(p

ini

peq )H
and

(20)

C C n (x g = hg ) = C C n , Eq ,

dl =
(15)

Dl ,eff h
2 l

dg =

Dg , 2 hg

(21)

where C Cn , Eq it is the equilibrium concentration. The set of equations 13 to 15, describes the process of diffusion in the space free of rock; that is, in the gaseous phase (Fig. 1) and they can be solved using the method of separation of variables in a similar way as the one used for the liquid phase. Thus, for the gas phase the concentrations is given by
m 2m + 1 2 tDl ,eff C N2 4 ( 1) = 1 exp 2 m =0 2 m + 1 C N2 ,Eq hl 2

The solution for the problem of effective diffusion described by equation 19 is of the type4:

p = p(t : pini , peq , a1 , d l , d g )

(22)

2m + 1 xl . (16) cos 2 hl

This means that the pressure depends on its initial value and four more parameters. These parameters are unknowns for the problem; a method to determine the values of these parameters is the Levenberg8-Marquardt,9 that consists in finding the optimal parametric values through a process of nonlinear regression that diminishes the quadratic difference between the analytical function of pressure and an experimental data set, pexp :
Peq , a , d l , d g

During the diffusion process the liquid-gas interface tends to move as a function of time. This work consider that these variations are very small with respect to H; that is why for purposes of calculation they are considered as constant. Luna4 et al demonstrated that, when the chemical composition of the liquid and gaseous phases are modified, changes in the global pressure of the system occur; these changes in pressure can be studied through equations of state. The authors present the following equation for pressure variation
C dp (z g zl )RT CC n = Dg Dl , eff N 2 dt H x g x g = hg xl xl = hl ,

Min

{p

exp

p(pini : peq , a, d , d g )}

(23)

(17)

A program in Fortran4 has been developed that solves the problem defined by equation 23, and then by means of equation 21 the value of the effective coefficient of diffusion can be obtained. The analysis of the pressure data obtained in the two described runs allows the determination of a value of the effective coefficient of nitrogen diffusion in hexane. The results obtained for both runs are next presented: Run 1: Figure 9 shows experimental data of pressure against time, and the analytical curve that was calculated by the method of Levenberg8-Marquardt9 using a program4 in Fortran.

subject to the following conditions:

p(t = 0 ) = pini

and

p(t ) = peq ,

(18)
Experimental dates Levenberg-Marquardt match

where, p is the global pressure system, zg and zl are the compressibility of the gas and liquid phases, T is the temperature of the system, is the porosity of medium, R the universal constant of gases, pini is the initial pressure and peq is the equilibrium pressure. With the results of equations 4 and 16, the solution of the problem for the global pressure defined by Eqs. 17 and 18 is the following

R2=0.96 Pressure (psia)

Time (days)

Fig. 9- Match of data of pressure obtained experimentally in the 8 9 first run with the method of Levenberg -Marquardt .

www.petroman.ir

SPE 92196

The result obtained in the first run for the effective coefficient of diffusion of nitrogen in hexane, through the match appearing in figure 9, was

Table 3- Effective diffusion coefficients.

Dl , eff = 1.001 10 4

cm sec

Run

Effective diffusion by concentration measurements ( 10-4 cm2/sec)

Effective diffusion by pressure analysis ( 10-4 cm2/sec)

Difference (% )

Run 2: Figure 10 shows experimentally measured pressure data Vs. time in run 2, and the best fit curve obtained by means of the method of Levenberg8-Marquardt,9 using the program in Fortran4 already mentioned.

1.1083

1.001

10.70

0.9111

1.007

9.50

Experimental dates Levenberg-Marquardt match

R2 = 0.98 Pressure (psia)

Conclusions The main purpose of this paper has been to present the results of a study on the effective diffusion in homogeneous porous media. Based on the findings of this work, the following conclusions and/or observations can be made: 1. The experimental aspect of the procedure for estimating effective coefficient of diffusion presented in this work, is original. 2. The effective coefficient of diffusion estimated for a sand packing (44% of porosity), by means of best fit curves of behavior of concentration with respect to time and distance, and the coefficient determined by means of pressure behavior observed in the runs, through the method of Levenberg8-Marquardt9 has the same order of magnitude (the values differs in 10% approximately); therefore both methods are reliable. 3. The fact that the values of concentration as a function of distance and time can be matched, implies that the process of effective diffusion can be described by means of the second Ficks law, by replacing the binary coefficient of diffusion (D) by an effective coefficient of binary diffusion ( Deff ). 4. As a final comment it can be stated that, with the theoretical-experimental work described above, new subjects of applied investigation can be opened, as the possible application of the definitions of effective diffusion, average concentration for fractured media. Nomenclature

Time (days)

Fig. 10- Matching of pressure data obtained experimentally in the 8 9 second run by the method of Levenberg -Marquardt .

In this run a better control for taking pressure data was used, which resulted in a better match, as it is observed in the value of the correlation coefficient R2 that is shown in figure 10. The result obtained for the effective coefficient of diffusion of nitrogen in hexane through nonlinear regression is

Dl , eff

cm 2 = 1.007 10 sec
4

The values for the effective coefficient determined for diffusion by means of the experimental analysis present a good agreement among them (1.1083 10-4 and 0.91110-4 cm2/sec); then comparing the previous values with the ones obtained by means of the adjustment by nonlinear regression of the registered data of pressure Vs. time (1.001 10-4 and 1.00710-4 cm2/sec), it can be concluded that these results are acceptable, because these were determined through different methods from analysis of the results. Finally the coefficients of effective diffusion for nitrogen in hexane obtained experimentally in this work, are shown in the following table:

a1 CC n C N2 CCn ,Eq C N 2 , Eq
D

parameter concentration of the hexane in nitrogen hexane concentration in nitrogen equilibrium concentration of hexane equilibrium concentration of nitrogen binary coefficient of diffusion parameter

d1

www.petroman.ir

SPE 92196

dg
Deff Dg Dl Dl,eff r H hp p pini peq pexp R T t hg hl x xl xg y zg zl

parameter effective binary diffusion coefficient in porous media free binary diffusion coefficient in the gas free binary diffusion coefficient in the liquid binary diffusion effective coefficient in the liquid radius total height height of the porous matrix pressure initial pressure equilibrium pressure experimental pressure universal constant of gases temperature time height of the gas column height of the liquid column longitudinal coordinate longitudinal coordinate in the liquid phase longitudinal coordinate in the gas phase longitudinal coordinate compressibility of the gaseous phase compressibility of the liquid mixture porosity

References 1. Perkins, T. K. and Johnston, O. C.: A Review of Diffusion and Dispersion in Porous Media, SPEJ (1963), 77-91. 2. Somerton, W.H.: Thermal Properties and TemperatureRelated Behavior of Rock/Fluid Systems, Elsevier, Amsterdam (1992).

3. Islas Jurez, R.: Difusin Efectiva en Medios Porosos, M. Sc. Report, School of Engineering, National University of Mexico, D.F., (2003). 4. Luna, R. E. E., Serrano, S. M., Pineda, M. A., Hernndez, E. L. and Serrano, S. E. Medicin del Coeficiente de Difusin a Travs de Cambios de Presin en una Celda PVT, Technical report, project D.02101, from Mexican Petroleum Institute, (2003). 5. Thomas, L. K., Dixon T. N., and Pierson R. G. : Ekofisk Nitrogen Injection, paper SPE 19839 presented at the Annual Technical Conference and Exhibition, San Antonio, Texas, Oct. 8-11, 1989. 6. Buchwald, A.: Determination of the Ion Diffusion Coefficient in Moisture and Salt Loaded Masonry Materials by Impedance Spectroscopy, 3rd int. PhD symposium 11-13 Oct. 2000, Vienna (2000), vol. 2, 475482. 7. Lagalaye, Y., Nectoux, A. and James, N.: Characterization of Acid Gas Diffusion in a Carbonate Fractured Reservoir Through Experimental Studies, Numerical Simulation and Field Pilots, paper SPE 77339 presented at the Annual Technical Conference and Exhibition, San Antonio, Texas, 29 Sep. to 2 Oct., 2002. 8. Levenberg, K.: A method for Solution of Certain Problems in Least Squares, Quarterly of Applied Mathematics (1944) 2, 164-168. 9. Marquardt, D.: An Algorithm for Least-Squares Estimation of Nonlinear Parameters, SIAM Journal of Applied Mathematics (1963) 11, 431-441. 10 Compaq Computer Corporation: Compaq Visual Fortran Professional Edition 6.6.0, copyright 2000 11. OriginLab, Corporation: Origin 6.1, copyright 19912001.

www.petroman.ir