Magnetic Resonance Imaging 34 (2016) 437441

Contents lists available at ScienceDirect

Magnetic Resonance Imaging

journal homepage: www.mrijournal.com

Original contribution

A rapid method for the measurement and estimation of CO2

diffusivity in liquid hydrocarbon-saturated porous media using MRI
Yuechao Zhao, Junlin Chen , Mingjun Yang, Yu Liu, Yongchen Song
Key Laboratory of Ocean Energy Utilization and Energy Conservation of Ministry of Education, Dalian University of Technology, Dalian, Liaoning Province, 116024, China

a r t i c l e i n f o a b s t r a c t

Article history: In this study, magnetic resonance imaging (MRI) was used to dynamically visualize the diffusion process of
Received 22 October 2015 CO2 in porous media saturated with liquid hydrocarbon. Based on the assumption of semi-innite media,
Accepted 14 December 2015 effective CO2 diffusivity was obtained directly by the nonlinear tting of one MR prole during the
diffusion process. These experimental ndings obtained based on MRI method showed a close agreement
Keywords:
with the conventional pressurevolumetemperature method. The novel MRI-based technique is a
Diffusivity
CO2
time-saving approach that can reduce the duration of CO2 diffusivity measurement more than 90%, and
Porous media realize rapid and accurate measurement and estimation of CO2 diffusivity.
Magnetic resonance imaging 2015 Elsevier Inc. All rights reserved.

1. Introduction non-invasive multidimensional visualization of ow and transport in

Gas injection into oil reservoirs to enhance oil recovery has
been widely practiced in the oil industry [1]. The injection of
certain gases, such as CH4, C2H6, and CO2, into oil reservoirs can
signicantly reduce oil viscosity [2]. Gas diffusivity has been a
subject of interest in oil recovery projects. Its study is central to
understanding the transport properties of two different media,
namely, gas and liquid [3]. Molecular diffusion helps gas dissolve in
crude oil, reduces the effect of viscous ngering, decelerates
gas breakthroughs, and increases the oil production rate [4].
Diffusion facilitates the analysis of the rate of transport caused by
concentration gradients.
Several methods have been developed to measure gas diffusivity
in liquid. These experimental methods can be roughly categorized
into two methods, namely, indirect and direct. Indirect diffusivity
measurements involve measuring the change in a parameter related
to diffusion, such as monitoring pressure decay [57], monitoring
gasoil interface position [8], and dynamic pendant drop volume
analysis [9], over time. Direct measurements involve measuring the
change in concentration over time, such as X-ray Computer Assisted
Tomography [10].
Magnetic resonance imaging (MRI), a powerful analytical tool,
has recently been shown to exhibit considerable potential for
Corresponding authors at: Dalian University of Technology. Tel.: +86 411
84706608; fax: +86 411 84708015.
E-mail addresses: jindanch@163.com (J. Chen), songyc@dlut.edu.cn (Y. Song).
porous media [11].
The indirect method for calculating gas diffusivity by measuring
pressure decay in a pressurevolumetemperature (PVT) cell is
extremely time consuming, as it will take multiple hours and even
days to reach the equilibration state. Thus, this study aimed to
demonstrate the potential of MRI applications in rapidly measuring
and estimating gas diffusivity in liquid hydrocarbon-saturated
porous media. This technique is a time-saving approach that
measures only one magnetic resonance (MR) prole during diffusion
and calculates gas diffusivity based on the assumption of
semi-innite media. The novel MRI-based technique reduces the
duration of gas diffusivity measurement from the span of many
hours into minutes.
2. Experimental investigation
2.1. Experimental apparatus
A schematic of the experimental setup is shown in Fig. 1. A Varian
400 MHz NMR system (Varian Inc., Palo Alto, CA, USA), which
provided a maximum gradient strength of 50 G/cm and a proton
millipede vertical imaging probe, was used for imaging. The gas
cylinder was a stainless steel cylinder (500 ml) wrapped with an
electric heater, which was used to supply pressurized CO2. The MRI
sample vessel was the diffusion cell and it was made of polyimide to
hold the diffusion samples. The outlet was blocked in the
experiments. The carrier of oil into the sample vessel was a glass
tube (10 cm height and diameter of 1 cm).

http://dx.doi.org/10.1016/j.mri.2015.12.024
0730-725X/ 2015 Elsevier Inc. All rights reserved.
438 Y. Zhao et al. / Magnetic Resonance Imaging 34 (2016) 437441
Fig. 1. Schematic of the experimental apparatus for diffusivity measurements.
2.2. Experimental materials
In this study, 99.99% pure CO2 (Dalian Guangming Special Gas
Co., Ltd., China) was used as the gas phase, and n-tetradecane with 98%
purity (Tianjin Kemiou Chemical Reagent Co., Ltd., China) was used as
the oil phase. The experimental temperature conditions of the gas and
oil phases were 20 C and 30 C, respectively. Pressure values ranged
from 2000 kPa to 5000 kPa. Glass beads of BZ02 (AS ONE, Japan), with
grain size distribution ranging from 0.177 mm to 0.250 mm, were
used to ll the glass tube as porous media.
2.3. Experimental procedure
Prior to starting the experiment, the sample vessel and all
connections were pressurized with nitrogen and tested for leakage at
a pressure of up to 6000 kPa for a minimum of 12 h. The bead-pack
porous media in the vessel were fully saturated with a predeter-
mined amount of n-tetradecane. Once temperature was conrmed to
be constant, gas was injected into the gas cylinder, and the desired
pressure was maintained using an ISCO syringe pump (260D,
Teledyne Isco Inc., America). The sample vessel and gas cylinder
were isolated. The liquid hydrocarbon sample was scanned to obtain
the initial MR image. The valve at the outlet of the gas cylinder was
opened, and gas was injected into the diffusion cell from the gas
cylinder. Diffusion of gas into the liquid began. The sample was then
scanned using MRI during the diffusion process. When the pressure
of the system uctuated within 10 kPa and the overall proton
density of liquid hydrocarbon phase generally remained changeless,
the system was considered to reach equilibrium, and the experiment
was nished. Finally, the experimental system was slowly depres-
surized to atmospheric pressure, and a new experiment was
restarted after detaching and cleaning.
2.4. NMR imaging parameter
Two-dimensional (2D) proles were obtained using standard
spin-echo multi-slice pulse sequence. Echo time (TE) was 4.9 ms;
repetition time (TR) was 3.0 s. The image data matrix was 96 96.
The eld of view was 40 mm 40 mm with 1 mm thickness, and
the acquisition time was 4 min and 48 s.
3. Mathematical analysis
The following assumptions are made for the analysis of the
diffusion process:
(1) The effective diffusivity of the gas is constant throughout the
experiment.
(2) No resistance to mass transfer occurs at the gasliquid
interface, meaning that the concentration at the interface is
the equilibrium concentration.
(3) Temperature remains constant during the experiment.
(4) Density-induced natural convection is negligible.
(5) The liquid phase is non-volatile.
(6) The porous media are homogeneous.
The model of semi-innite media [12] for calculating diffusivity is
shown as follows:
0 1
C y C 0 B y C
erfc@ qA 1
C 1 C 0 2 D t p
where Cy is the gas concentration at distance y from the inux
boundary, C0 is the gas concentration at the inux boundary, C1 is the
initial gas concentration in the porous media, erfc is the error
function, Dp is the overall diffusivity, and t is time.
MR images reect the information regarding the liquid phase
concentration. Considering the diffusion of gas into liquid, the signal
intensity of the liquid phase declined, but this change was small. To
minimize the inuence caused by uctuations of each position
and each time in the process of calculation, the following formula
was obtained using Eq. (1) at the distance y by subtracting the
 Y. Zhao et al. / Magnetic Resonance Imaging 34 (2016) 437441 439

earliest time liquid concentration for t1 from liquid concentration
corresponding to longer diffusion times for t2:
0 0 1 0 11
B B y C B y CC
C y t 1 C y t 2 K  @erfc@ qAerfc@ qAA 2
2 Dp  t 2 2 Dp  t 1
where K is the constant.
The concentration difference data of two different times in each
position y were nonlinearly tted using Eq. (2). The diffusivity Dp of
gas into liquid was used as an adjustable parameter to obtain the
best match of these data. Diffusivity was then obtained.
Given that Eq. (2) is deduced by the assumption of semi-innite
media, this method can only be used in situations in which gas has not
reached the bottom of the liquid phase. This situation generally exists in
the project of CO2 enhanced oil recovery (CO2-EOR). Using Eq. (2), the
distortion of images caused by interference from the proton-rich liquid
phase could be eliminated. Moreover, the constant effect caused by T1
yields was removed by subtraction in the differential prole approach.
4. Results and discussions
A series of images in Fig. 2 shows the diffusion process of CO2 in
n-tetradecane at an initial given pressure of 5000 kPa and temper-
ature of 30 C. With CO2 diffusing in the liquid phase, MR images
gradually darken, because CO2 is invisible on resonance frequency of
protons, thus, only the oil phase distribution is measured. The bright
region (red) indicates a high liquid content and corresponds to high
MRI signal intensity. By contrast, the dark region (purple) indicates
low liquid content. Through the change in brightness, evident
differences were observed between the part wherein CO2 diffused in
the liquid phase and the part where no CO2 could be observed. This
difference reects the concentration difference between the top and
bottom. Furthermore, the area and extent of gradual change in color
were enhanced with diffusion time and diffusion distance. This
phenomenon showed that CO2 could slowly dissolve into the liquid
phase. The image in Fig. 2 clearly shows that CO2 gradually diffused
to the bottom in porous media, and the concentration nally became
consistent. In addition, the interface gradually increased because the
volume of the liquid phase increased when CO2 dissolved into the
liquid phase.
To calculate diffusivity, t1 was set to 24 min. The time of CO2
diffusion to the bottom of porous media was approximately 192 min,
so t2 was set to 48, 96, 144, and 192 min. The tting results of CO2
into porous media saturated with n-tetradecane under 4000 kPa and
30 C are shown in Fig. 3. At each curve, the value of Cy(t1) Cy(t2)
gradually increased along the diffusion direction to a maximum
peak, nally decreasing to zero. This phenomenon is explained in
Fig. 4. By dividing the signal intensity of different times (t2 = 48, 96,
144, and 196 min) from the signal intensity at the initial time (t1 =
24 min), Fig. 4 was obtained. As shown in Fig. 4, the change in
interface concentration was indistinct, and the assumption of
equilibrium concentration at the interface was feasible. At a certain

4 min, 48 s 9 min, 36 s 14 min, 24 s 19 min, 12 s 24 min 28 min, 48 s 33 min, 36 s

38 min, 24 s 43 min, 12 s 134 min, 24 s 148 min, 48 s 192 min 249 min, 36 s 2064 min

Fig. 2. 2D MR images of CO2 diffusion process in porous media saturated with n-tetradecane.
440 Y. Zhao et al. / Magnetic Resonance Imaging 34 (2016) 437441

Fig. 3. Fitted curve of CO2 in porous media saturated with n-tetradecane in 4000 kPa 30 C and t2 of (a) 48, (b) 96, (c) 144, and (d) 192 min.

Fig. 4. Signal intensity at different time.
time, given the dissolution and diffusion of CO2, the signal intensity
decreased at the region where CO2 diffused. Each curve in Fig. 3
initially showed this increasing trend. With the restriction of
diffusion depth, a decreasing trend appeared later in the curve.
In theory, two liquid concentration proles of different time are
sufcient to retrieve all these parameters in Eq. (2). But in order to
reduce the inuence of experimental measurement error, a series of
images at various times was measured during the diffusion process;
the best matched diffusivity was obtained by tting experimental
data at four different time points under one experimental condition
using Eq. (2). The effective gas diffusivity was obtained by averaging
four diffusivity values. The correlation coefcients for the t of
experimental data to Eq. (2) ranged from 0.939 to 0.962. The
effective diffusion coefcients of different times (t2 = 48, 96, 144,
and 196 min) are 1.1204 10 9 m 2 /s, 1.2742 10 9 m 2 /s,
1.3565 10 9 m 2/s, and 1.4294 10 9 m 2/s, respectively. The
average is 1.2951 10 9 m 2/s. Using the PVT method of pressure
decay, the equilibrium pressure must be reached, and the equilib-
rium time was 2064 min (Fig.3). Using this mathematical procedure,
the shortest measuring time of CO2 effective diffusivity was 48 min.
Compared with 2064 min, the measuring time can be saved about
 Y. Zhao et al. / Magnetic Resonance Imaging 34 (2016) 437441
Table 1
Summary of measurements of CO2 diffusivities in n-tetradecane at different initial
pressures and temperatures.
Solvent Oil Liquid Temperature Pressure (kPa) Diffusion Coefcient
Viscosity (cp) (C) (109 m2/s)
Tetradecane 1.74 30 2000 0.9144
3000 1.1407
4000 1.2951
5000 1.7531
2.07 20 2000 0.4731
3000 0.4844
4000 0.7315
5000 1.0854
98% by this method. Even considering that the measuring time is
over until the diffusion front reaches the bottom of porous media,
the measuring time for determining diffusivity was 196 min.
Compared with 2064 min, the measuring time still can be save
more than 90% by this method. Therefore, this method, which based
on the assumption of semi-innite media and using the liquid
concentration data directly obtained by MRI, is more efcient than
the ordinary PVT method. Table 1 shows a summary of measurement
of CO2 diffusivities in n-tetradecane at different initial given
pressures and temperatures.
As shown in Table 1, given the range of test temperature and
pressure, the effective diffusivity increased as the initial given
pressure increased, because gas solubility increased with increasing
pressure. Regarding the effect of temperature, the effective diffusiv-
ity increased as temperature increased. Molecular thermodynamic
movement increased because of the increase in temperature,
thereby exacerbating diffusion. In addition, the viscosity of liquid
hydrocarbon decreased because of the increase in temperature. This
decreased viscosity facilitated gas diffusion in liquid hydrocarbon.
This study compared the diffusivity values with those of other
previous studies. As shown in Table 2, under similar experimental
conditions, the diffusivity of CO2 in the present study was small
compared with that of Grogan, but in the same order of magnitude
(10 9 m 2/s). This good agreement indicated that the experimental
apparatus and mathematical procedure employed in this study are
reliable for the investigation and calculation of the diffusivities of
CO2 in liquid hydrocarbon. Thus, this method can realize rapid and
accurate measurement and estimation of gas diffusivity.
The diffusivities for CO2 in liquid hydrocarbon are given by an
empirical function of pressure and temperature, respectively. The
correlating parameters vary over the following ranges: pressure (p),
304.77739.88 psia (20005000 kPa); temperature (T) 20 C30 C
(293.15303.15 K). An empirical expression is adopted and given by:
Dp ap T
A least-squares t of all the data as a function of p and T according
to Eq. (3) led to the following relationship with the largest
correlation coefcient:
0:8443 4:2146 22
Dp 1:8989p T  10
The correlation coefcient for the t of the experimental data to
Eq. (4) was 0.82. Note that p is expressed in psia, and T is expressed
in Kelvin. From the equation, we found a signicant phenomenon, in
which the effect factor of temperature was larger than the pressure.
In the CO2-EOR industry, raising the temperature rather than the
pressure is more effective under the same economic conditions.
441
Table 2
Comparison of diffusivity results with other studies.
Viscosity Temperature Pressure Diffusion Coefcient Reference
(cp) (C) (kPa) (109 m2/s)
8360 50 5000 0.5 Schmidt (1989) [13]
3 25 6000 2.0 Grogan (1988) [8]
767 2590 31004100 0.160.47 Upreti (2000) [7]
5. Conclusion
A non-invasive MRI experimental method along with a mathemat-
ical model was developed to measure the effective diffusivity of CO2 in
liquid hydrocarbon-saturated bead-pack porous media. Based on the
assumption of semi-innite media, the effective gas diffusivity was
obtained directly by nonlinear tting of one MR prole during the
diffusion process. The experimental ndings showed close agreement
with the results of the conventional PVT method.
The diffusivities for CO2 in liquid hydrocarbon were given by an
empirical function of pressure and temperature. Diffusivity in-
creased with increasing temperature and pressure. Notably, the
effect of temperature was larger than that of pressure. This
mathematical model can only be used in situations in which gas
has not reached the bottom of the liquid phase. However, this
situation generally exists in CO2-EOR. The proposed experimental
method can guide the use of NMR logging in CO2-EOR. The method
provided a rapid and accurate tool for further studies on gas diffusion
in liquid-saturated porous media.
Acknowledgments
The authors are grateful for the support provided by the National
Natural Science Foundation of China (Grant No. 51206018), the
Specialized Research Fund for the Doctoral Program of Higher
Education (Grant No. 20120041120057), the Ph. D. Startup Foundation
of Liaoning Province (Grant No. 20121022), and the Fundamental
Research Funds for Central Universities.
References
[1] Li Z, Dong M. Experimental study of carbon dioxide diffusion in oil-saturated
porous media under reservoir conditions [J]. Ind Eng Chem Res 2009;48(20):
930717.
[2] Yang C, Gu Y. New experimental method for measuring gas diffusivity in heavy
oil by the dynamic pendant drop volume analysis (DPDVA) [J]. Ind Eng Chem Res
2005;44(12):447483.
[3] Zainal S, Hon VY, Saaid IM, Jelani J. An evaluation of gas diffusivity measurement
in reservoir uid from low to high pressure systems for oil recovery applications;
proceedings of the International Petroleum Technology Conference, F [C].
International Petroleum Technology Conference; 2009.
[4] Unatrakarn D, Asghari K, Condor J. Experimental studies of CO 2 and CH 4 diffusion
coefcient in bulk oil and porous media [J]. Energy Procedia 2011;4:21707.
3 [5] Riaza M. R. A new method for experimental measurement of diffusion
coefcients in reservoir uids [J]. J Pet Sci Eng 1996;14(3):23550.
[6] Zhang Y, Hyndman C, Maini B. Measurement of gas diffusivity in heavy oils [J]. J
Pet Sci Eng 2000;25(1):3747.
[7] Upreti SR, Mehrotra AK. Experimental measurement of gas diffusivity in
bitumen: results for carbon dioxide [J]. Ind Eng Chem Res 2000;39(4):10807.
[8] Grogan A, Pinczewski V, Ruskauff GJ, Orr JF. Diffusion of CO2 at reservoir
conditions: models and measurements [J]. SPE Reserv Eng 1988;3(01):93102.
[9] Yang C, Gu Y. A new method for measuring solvent diffusivity in heavy oil by
4 dynamic pendant drop shape analysis (DPDSA) [J]. SPE J 2006;11(01):4857.
[10] Song L, Kantzas A, Bryan JL. Investigation of CO2 diffusivity in heavy oil using X-ray
computer-assisted tomography under reservoir conditions [C]//SPE International
Conference on CO2 Capture, Storage, and Utilization. Society of Petroleum
Engineers; 2010.
[11] Callaghan PT. Principles of nuclear magnetic resonance microscopy [M].
Clarendon Press Oxford; 1991.
[12] Crank J. The mathematics of diffusion [M]. Clarendon Press Oxford; 1975.
[13] Schmidt T, Leshchyshyn T, Puttagunta V. Diffusion of carbon dioxide into
athabasca bitumen; proceedings of the 33rd Annual Technical Meeting of the
Petroleum Society of CIM, Calgary, Canada, Paper, F [C]; 1982.