Magnetic

PI1 SO730-725X( 96) 00218-2

ELSEVIER

Resonance Imaging, Vol. 14, Nos. 718, pp. 869-873, 1996

Copyright 0 1996 Elsevier Science Inc.
Printed in the USA. All rights reserved
0730-725X/96 $15.00 + .OO

Contributed

Paper

MULTINUCLEAR

NMR MICROSCOPY
OF TWO-PHASE
SYSTEMS IN POROUS ROCK

TOMUTSA
National Institute for Petroleum and Energy Research (NIPER), Bartlesville, OK, 74005, USA
DARYL

BDM-Oklahoma,

A. DOUGHTY

FLUID

AND LIVIU

The high-field magnetic resonance (MR) characteristics of fluids in porous reservoir rock exhibit short T2
relaxation times and broad natural line widths. Thesecharacteristics severely restrict which MR imaging
(MRI) methodology can be used to obtain high-resolution porescaleimagesof fluids in porous rock. An
MR microscopy protocol based on 3D backprojection using strong imaging gradients was developed to
overcome many of theseconstraints. To improve the imagequality of two-phasesystems,multinuclear MRI
using proton MR to imagethe brine phaseand q MR of a fluorinated hydrocarbon to image the oil phase
was used. Resolution as high as 25 microns per pixel has been obtained for fluid systemsin Bentheim and
Fontainebleau sandstones.Separate proton and % imagesof brine and oil phasesshow good agreement
with total saturation images.Software has been developed to perform 3D erosion/dilations and to extract
the pore size distribution from binarized 3D imagesof fluid filled porosity. Resultsfrom pore size measurements show significant differences in the nature of the pore network in Fontainebleau and Bentheim
sandstones. Copyright 0 1996 Elsevier Science Inc.
Keywords:

Two-phasefluid systems;Porous rock; NMR microscopy.

EXPERIMENTAL

INTRODUCTION

In petroleum reservoir assessment and characterization, MRI is a valuable nondestructive imaging technology used to image fluids within cores, providing
information about porosity distribution and how it is
affected by rock heterogeneity. MR microscopy provides information about the pore network connectivity,
which is directly related to fluid flow characteristics
within the rock matrix, pore sizes, and fluid distributions in pores.2 The high resolution achievable allows
visualization of the effect of rock/fluid interactions on
oil and water distributions within the pore spaces of
reservoir rocks. Such a capability aids in understanding
oil displacement processes taking place at the pore
level, which are essential in understanding the mechanisms of various oil recovery processes. High-resolution MR microscopy was recently expanded to include
multinuclear imaging for direct imaging of separate
phases using fluids th.at are tagged by an MR-active
marker such as 19F.

The most efficient 3D imaging process for fluids in

porous rock is the 3D projection reconstruction sequence.3 The imaging gradient is of fixed amplitude
and its direction in space is controlled by three X-, Y-,
and Z-components. Because the gradient is switched
on prior to the RF pulses in the spin echo and remains
on until the signal is acquired, rapid switching of the
gradient is not required and the time to echo (TE) can
be shorter, maximizing signal strength for fluid/rock
systems, which have a short T2 relaxation time. However, achieving high resolutions using this protocol
requires strong gradients and high RF bandwidths. The
algorithm used to process the projection reconstruction
image data is a two-stage implementation of the convolution method for parallel beam reconstruction tomography from the Donner Laboratory algorithms.4 In
adapting this algorithm for MR imaging, the MR projection is treated like a parallel beam emission projection of source intensity across the dimension of the
object defined by the gradient direction. The result is

Address correspondence to D.A. Doughty, BDM-OkIahoma, National Institute for Petroleum and Energy Research

(NIPER),
869

P.O. Box 2565, Bartlesville, OK 74005.

Magnetic Resonance Imaging 0 Volume 14. Numbers 7/S, 1996

870

a 3D Cartesian image of the fluid-filled porosity of the

sample, where pixel intensity is expressed as a O-255
level, &bit value.
The small core plugs of reservoir rock used for MR
microscopy are typically about 5 mm in diameter and
7- 10 mm long. They are mounted inside a core holder
made from two Teflon end caps and two layers of
Teflon shrink tubing. The inner layer is an FEP-type
Teflon, which shrinks at 206C, and the outer layer
is PTFE Teflon, which shrinks at 320C. This higher
temperature softens the inner layer, bonding the end
caps and core plug into a leak-proof assembly. Small
diameter Teflon tubing is clamped into the end caps
and permits the exchange of fluid components in the
core plug while mounted in the spectrometer. Typical
imaging parameters are: 200 KHz projection bandwidth, 60 gauss/cm gradient, 256 X 128 gradient orientations, 256 complex points per projection, 4-8
summed accumulations per projection, 1.28 ms TE,
and 0.4 s PD for proton imaging of brine. A longer
PD is used for imaging the oil phase. The brine used
is 0.5% NaCl by weight with 0.023% MnC12 added as
a T1 relaxation agent. The oil phase is usually Soltrol
or a fluorinated compound such as hexafluorobenzene
(HFB) or 3,5-bis (trifluoromethyl)
bromobenzene
(TFMBB ) .

Using a longer TE to suppress the more quickly

relaxing brine phase is a common approach to separately imaging two-phase fluid systems. However. attenuation of the projection areas at certain gradient
orientations was noticed in these experiments compared to the projections for short TE experiments. Figure 2 shows the results of an experiment on a small
spherical, water-filled phantom at a TE of 10.24 ms
for varying gradients strengths. Off-axis gradients involving combinations of the Z-gradient with either Xor Y-gradients were discovered to cause a suppression
of projection signal strength, increasing with TE and
gradient strength. In Fig. 2, the area under the MR
projection signal, normalized to the area for a TE of

1.28 ms, was plotted vs. the gradient angle index in

the XZ plane (0- 180). The figure shows that the
average signal area for all gradient orientations is
strongly affected by gradient strength with the stronger
gradient causing a greater suppression of the signal at
a given TE. Also, the influence of gradient orientation
on the signal is much more pronounced for the stronger
gradients. The average suppression of the MR signal

BRINE

OIL at
RESIDUAL
BRINE

RESIDUAL

BRINE

RESULTS
Figure 1 shows the two-phase brine/Soltrol
fluid
filled porosity for three successive regions down the
axis of a core plug of Fontainebleau sandstone for
brine and Soltrol/residual brine conditions. The residual brine condition was established by flowing Soltrol
through the brine saturated core plug until no additional brine was produced at the outlet. The image is
displayed as a surface-rendered view of porosity at a
voxel intensity matching the brine saturated porosity.
Image resolution was about 30 microns per voxel edge.
The first column (A, D, and G) shows the initial brinesaturated total porosity. The direction of fluid flow in
the figure was from (A) to (D) to (G) back to front.
The second column (B, E, and H) shows the corresponding oil occupied porosity obtained using a longer
TE to suppress the brine signal for the same respective
regions. The last column (C, F, and I) shows the corresponding residual brine occupied porosity in the three
regions obtained by difference. Even though at least
10 pore volumes of oil were flowed through the core
plug before the second experiment was run, Fig. 1
shows that the relative volume of the total porosity
occupied by Soltrol is decreasing and the volume occupied by brine is increasing as the distance from the inlet
port increases, which is a typical end effect noticed in
fluid flow in core plugs.

I
800 microns

Fig. 1. Surface-rendered3D views of fluid-saturatedFontainebleausandstone.The first column(A, D, and G) shows

the brine-saturatedrock. The secondcolumn (B, E, and H)
showsthe oil phaseat residualbrine obtainedusinga longer
TE to suppressthe brine signal.The last column(C, F, and
I) showsthe residualbrine by difference.

Multinuclear NMR microscopy of two-phase fluid systems0

D.A.

DOUGHTY

AND

L. TOMUTSA

871

80
,

Fig. 3. Part of a cross-sectional

slicefrom the 3D MR image
of a brine filled, wedge-shapedphantom.The brine is the
lighter shadesdisplayedusinga O-255 grey scale.The white
triangle showsthe shapeof the wedge.The arrow pointsto
where the wedgeis one voxel thick (38 microns).

20

20

40

60
Angie

80

100

120

index

voxel is less than that in wider regions of the wedge.

Another artifact of the reconstruction process is that
the edges of objects are blurred over a width of about
two voxels, as shown in Fig. 3. At the spot where the
wedge is one voxel wide, the image spreadsover three
or four voxels causing a reduced intensity. The intensity falls off even more as the thickness of the wedge
falls below one voxel. This variation in intensity could
be used to make a quantitative estimate of fractional

Fig. 2. The areaunder the MRI projection at a TE of 10.24

ms is plotted as a function of XZ-gradient angleindex (O180degrees)for severalgradientstrengthsin gauss/cm.A
markedgradientorientationeffect is evident for the stronger
gradients.
in the presenceof a gradient is causedby the diffusion
of fluid molecules from one location in the sample to
a region of different magnetic field during the TE.
This causesthe contribution from that molecule to be
missing from the echo signal. An explanation for why
this effect should be influenced by gradient orientation
is not available at this; time. Using multinuclear MRI
with separatefrequencies to image the two phaseswith
similar short TEs was chosen to eliminate the impact
of this effect.
Another image effect in a high-resolution MR image
of a fluid-filled porous medium is the variation in voxel
brightness across the different pores or from pore to
pore in the images. One such interpretation is that the
dimmer voxels represent places within the rock/fluid
system where only a part of the voxel is occupied by
fluid-filled porosity contributing to the signal. Figure
3 shows an enlarged view of part of one cross-sectional
slice of a brine saturated phantoms image showing a
brine-filled wedge formed by putting a small Teflon
spacer, 0.250 mm thidk, between small squaresof glass
plate and filling the void with brine. The white triangle
shows the actual shape of the wedge, and the arrow
points to the region where the thickness of the wedge
became equal to the voxel dimension (38 microns in
this image). Even in this region the brightness of the

800 microns

Fig. 4. Slice views of fluid-saturatedBentheim sandstone.

The slices are 64 x 64 voxels extracted from near the center
of 3D multinuclear
MR images. (A) Brine (H); (B)
TFMBB
at residual brine ( 19F); (C) Brine at residual
TFMBB
(H); (D) Residual TFMBB
( 19F). Fluid-filled
pores are the lighter shades displayed using a O-255 grey

872

Magnetic Resonance Imaging 0 Volume 14, Numbers 7/8, 1996

800microns

Fig. 5. Two brine-saturatedsandstones

showingdifferences
in pore structure.The lighter areasare the brine-filled pore
spacesdisplayedusing a O-255 grey scale.(A) Bentheim
sandstone
showinga well-connectedpore system.(B) Binarized imageof A. (C) Fontainebleausandstoneshowingisolated largerporeswith numeroussmallpores.(D) Binarized
imageof C.
voxel filling in different regions of MR images and
effectively increase resolution.
Figure 4 shows the results of a multinuclear MR
microscopy experiment on a core plug of Bentheim
sandstoneusing brine and TFMBB. The sameimaging
gradient strengths were used for both the proton and
F images. Square 2D slices 64 x 64 voxels on each
side and extracted from near the center of the 3D imagesare shown. Figure 4A shows the proton image of
the brine saturated core plug. Figure 5B shows the oil
phaseusing F MRI. Figure 4C showsthe brine image
at residual oil, and Fig. 4D shows the residual oil,
again imaged using F MR. In the images the voxel
intensity is displayed using a O-255 level gray scale.
The F images had to be scaled by the ratio of the
H/F MR frequencies to obtain the same size and
resolution. If the images in 4C and 4D were added,
which would represent the total fluid-filled porosity
in the core plug, good agreement with 4A would be
obtained. This lends confidence to using multinuclear
MR imaging to separately image the two phases.
Software was developed at NIPER to track pores
through 3D space and develop information on pore

size and connectivity using an erosion/dilation process

similar to that developed by Ehrlich5 for petrographic
thin-section analysis but extended to three dimensions.
The erosion/dilation process works on binarized images where voxels at or above a selected intensity are
set to an intensity value of 255 and those below the
reference intensity are set to zero. A 3D kernel consisting of a 3 x 3 x 3 cross with seven voxels is
scanned progressively through the image volume. As
the kernel encounters filled voxels in the image plane,
in the erosion process the value of the central voxel is
replaced by the minimum value of the kernel voxels.
For a binarized image this is equivalent to being either
on (255 ) or off (0). Erosion is used to break narrow
connections between larger pore spaces.In the dilation
process the value of the central voxel is replaced by
the maximum value of the kernel voxels. Dilation is
used to replace the previously eroded outer layer of
voxels for the larger pore spaces. The software will
permit up to six successivestagesof erosion/dilation.
In the pore size determination used on the binarized
image data, as the layers of the image are scanned, a
pore index is established for each separate activated
voxel encountered during the scan. If the newly encountered voxel is determined to be a nondiagonal
neighbor to a previously counted voxel considering all
three dimensions, the voxel count is incremented for
that pore index. If a bridge is later discovered connecting two previously isolated pores, the voxel count for
the highest pore index is added to the lower pore index
and the higher pore index is zeroed and freed for later
use. The end result of the pore size determination is a

._

Erosions/

Dilations

Fig. 6. Pore measurements

on binarized pore images.The
largest pore as a fraction of total porosity remainingafter
erosion/dilationis plotted vs. the numberof erosion/dilation
cycles. The Bentheimand Fontainebleausandstones
show
markeddifferencesin pore structure.

873

Multinuclear NMR microscopy of two-phase fluid systemsl D.A. DOUGHTY AKD L. TOMUTSA

histogram of pore size: for the image volume and a

table showing the ranges in X, Y, and Z for each pore.
Figure 5 shows 2D MR slices taken from near the
center of 3D images of two brine-saturated pore systems. Figure 5A shows the image of a core plug of
Bentheim sandstone of 24% porosity. The same gray
scale of voxel intensity was used for this figure. The
binarized version of the same image is shown in 5B.
Figure 5C shows a similar image of a core plug of
Fontainebleau sandstone of 12% porosity. Figure 5D
shows the binarized version of this image. The images
show a considerable variation in appearance. The Fontainebleau image shows much greater variation in
voxel intensity and pore size with isolated large pores
separated by low porosity areas of much smaller pores.
The Bentheim image shows a higher image uniformity
with more connectivity between neighboring pores. No
erosion/dilation
was done in obtaining the images in
Fig. 5.
Figure 6 shows results of pore size/connectivity
analysis on the two systems, where the largest pore in
the histogram, expressled as a fraction of the total porosity remaining at each stage, is plotted vs. the number
of successive erosion/dilation
stages for the two sandstone binarized images. For the Bentheim sandstone,
even after one stage of erosion/dilation,
most of the
porosity is contained in one large pore throughout the
image pore space. The largest pore size then falls off
sharply after an additional erosion/dilation
and then
starts to increase. In contrast to the Bentheim system,
the results for the Fontainebleau sandstone show an
opposite trend with the largest pore size starting off
small and then increasing modestly after several erosion/dilation stages. After two stages of erosion/dilation the Fontainebleau system behaves similarly to the
Bentheim result. These statistical results confirm the
visual differences noted above and indicate the high
connectivity in the Bentheim core plug compared to
the Fontainebleau sample.

CONCLUSIONS
A technique of MR microscopy that has successfully imaged fluids at pore scale in sandstones has been
developed at NIPER using the projection reconstruction pulse sequence on small core plugs with strong
imaging gradients. Voxel resolutions as high as 25
microns have been achieved.
Multinuclear MR imaging of two-phase fluid systems using 19F substituted hydrocarbons permits obtaining unambiguous images of the separate water and
oil phases leading to better understanding of fluid distributions in porous rock.
Software has been developed at NIPER which can
perform a true 3D erosion/dilation
process on the binarized, pore scale, MR images, and then obtain pore
size measurements and track pore connectivity.
Results from pore size measurements show significant differences in the nature of the pore network in
Fontainebleau and Bentheim sandstones.
REFERENCES
1. Edelstein,W.A.; Vinegar, H.J.; Tutunjian, P.N.; Roemer,

P.B.; Mueller, O.M. NMR imagingfor coreanalysis,63rd

Annual Technical Conference,Society of PetroleumEngineers,Houston,TX; October 2-5, 1988.
2. Doughty, D.A.; Tomutsa,L.; Madden, M.P. Pore scale
fluid imagingin reservoirrock by NMR microscopy;ACS
National Spring Meeting Symposiumon Applications of
Magnetic Resonance
Imagingin EnhancedOil Recovery,
Denver, CO; March 28-April 2, 1993.
3. Doughty, D.A.; Tomutsa,L. NMR microscopyfor fluid
imagingat pore scalein reservoir rock. Joint Society of
CoreAnalysts/Society of ProfessionalWell Log Analysts
Symposium,

Oklahoma

City, OK; June 14-17,

1992.

4. Huesman,R.H.; Gullberg. G.T.; Greenberg,W.L.; Budinger, T.F. RECLBL Library UsersManual: Donner Algorithmsfor ReconstructionTomography,Berkeley, CA.
LawrenceBerkeley Laboratory, University of California;
1977.
5. Ehrlich, R.; Crabtree,S.J.; Kennedy, S.K.; Cannon,R.L.
PetrographicimageanalysisI. Analysis of reservoirpore
complexes.J. Sediment.Petrol. 54:1365- 1376; 1984.