Influence of open and sealed fractures on fluid flow and water saturation in sandstone cores using Magnetic Resonance Imaging

Summary

We use Magnetic Resonance Imaging (MRI) to image the imbibition of water by capillary action in a right-cylindrical sample of a porous sedimentary rock with low iron content. In the method some 55 repeat images are taken over a period of approximately two hours, covering five vertical sections. The evolution of the water flood front and the degree of water saturation can be observed by examining snapshots of proton density. The results clearly show (a) the development of a rising wetting front in the rock matrix (b) preferential flow along open fractures observed on the core surface, and (c) reduced flow associated with sealed fractures. The inferred location, orientation and connectivity of conducting and sealing fractures are confirmed by impregnating the sample after the test with an appropriate low-viscosity setting resin and taking serial thin sections in destructive mode. The results validate the utility of MRI as a non-destructive analytical tool for visualizing the distribution of water inside fractured porous media with low iron content. The technique identifies paths of high and low permeability in the sample, and quantifies the fracture location, orientation, and connectivity in sedimentary rocks. Preferential fluid flow in open fractures during capillary imbibition implies that the fractures are more water-wet than the clasts within the matrix. This may be due to due to differences in the age, morphology and mineral structure on the surface of the pores and the fractures.

fracture geometry, MRI, water saturation.

1 Introduction

Magnetic resonance imaging (MRI) is now established as an important high-resolution diagnostic imaging technique in biophysics and medicine. MRI does not use ionising radiation, so it is free of the potential health hazards of alternatives such as X-ray computer tomography (CT) scanning (Stuart & Young 1988). The nuclear magnetic resonance (NMR) signal has a higher information density because it has five separate components per imaged pixel. These are: the density of the nuclear species (dominated by hydrogen, whose nucleus is composed of a single proton, ¹H), two relaxation times (T₁ and T₂), and a phenomenon called the chemical shift that allows images of aqueous and non-aqueous fluids, and their flow velocities, to be obtained independently.

The MRI technique is sensitive to proton density in natural porous materials of low natural magnetization (iron content). It can therefore image mobile aqueous and non-aqueous fluids containing hydrogen nuclei and quantify their interaction with the surface of the pores in three dimensions (Chardaire & Roussel 1990). As a consequence NMR has long been used in the oil industry for the measurement of petrophysical rock properties. These include the bulk porosity (Timur 1969; Cowgill et al. 1981), permeability (Headley 1973; Kenyon et al. 1986), wettability (Brown & Fatt 1956; Kumar et al. 1969; Williams & Fung 1982), the distribution of pore size (Seevers 1966; Kozlov & Ivanchuk 1982; Brown et al. 1981; Gallegos et al. 1987; Gallegos & Smith 1988; Kenyon et al. 1989); and the degree of water saturation (Saraf & Fatt 1966, 1967).

One of the first applications of MRI to the imaging of porous media was the use of NMR to follow the capillary absorption of water as a function of time (Gummerson et al. 1979). Later Horsfield et al. (1989) and Dechter et al. (1989) used MRI to selectively identify the degree of water and oil saturation and to visualize two-phase flow using chemical displacement data. Chardaire & Roussel (1990) studied fluid distributions in porous media using a high magnetic field of 0.9T. Their work was focused on the determination of the local porosity through the intensity of the NMR signal and on the visualization of the distribution of oil and water using chemical-shift selective excitations. Dijk et al. (1999) measured water flow velocities and investigated the effects of wall morphology on flow patterns inside a typical rock fracture. Their study demonstrates that MRI effectively and accurately measures even highly heterogeneous flow fields in such systems.

In this paper we use MRI to determine the geometry of a flood front during a water imbibition experiment in a cylindrical specimen of a water-wet porous fractured rock (sandstone), and to quantify the distribution of water saturation as the flooding progresses. We image time-dependent variations in water saturation and identify conducting and sealing fractures. The results are used to infer the location, orientation and connectivity of the fractures in three dimensions. These inferences are successfully validated in destructive mode after the test using serial thin sectioning after resin injection (Baraka-Lokmane 1999). The fractures are natural rock joints, with no appreciable shear dislocation. The degree of water saturation is calculated both in the rock matrix and around the fractures, in order to study the influence of the fracture network on the spatio-temporal distribution of water saturation of the sandstone cores and to determine if fractures are open or sealed by a filling material.

We chose an imbibition experiment at ambient pressure because it is difficult to construct a rig that will be transparent to the technique and still be capable of providing a fluid pressure gradient within the sample and maintaining a confining pressure and differential stress. Our results are therefore more applicable to capillary infiltration at the very near surface above the water table, rather than subsurface fluid flow under pressure with a hydraulic gradient, and are hence more directly applicable to surface environmental rather than problems in hydrocarbon production or the safe deep storage of high-level waste.

2 Principles of nmr and mri techniques

Nuclear Magnetic Resonance is a spectrometric technique. When nuclei with magnetic moments are put in a static homogeneous magnetic field, the population of energy states is shared between discrete levels. A suitable radio-frequency field is applied to induce transitions between the levels, thus changing the magnetization. At the end of radio frequency excitation, nuclei return to equilibrium and the magnetization decays as a function of time. The frequency of this decay depends on the strength of the magnetic field and is characteristic of both the atomic nuclei in the material and its chemical environment. This method is a very powerful tool for quantitative, qualitative and structural analysis of materials, including rocks.

To obtain an image of the distribution of a particular atomic nucleus such as hydrogen, a spatial discrimination is required instead of a chemical one. This discrimination is produced by superposition of three orthogonal linear magnetic field gradients on the main magnetic field. The three gradients are applied in the x, y, and z directions. The ultimate resolution then depends on the strength of the three gradients, on the density of the nucleus, and on the transverse relaxation rate. There are two kinds of relaxation: longitudinal relaxation (T₁), which governs the recovery of the system towards the equilibrium and the frequency of the repetition rate, and transverse relaxation (T₂), which characterises the extinction of the signal and the width of the frequency signal, and hence the resolution of the method. The time needed to apply and interrupt gradients, especially in echo methods, can be long, and it is necessary to adjust this time with the relaxation rate to obtain a signal. In a porous rock with low natural magnetization subject to water imbibition, the proton density is directly proportional to the degree of water saturation. In contrast porosity, permeability, wettability and paramagnetic impurities all govern the relaxation time, and hence only qualitative inferences can be made on these properties. In this paper we concentrate on a quantitative analysis of the spatial heterogeneity of the degree of saturation due to channelled flow in a fracture, and use this to make qualitative inferences on fracture geometry and permeability.

3 Experimental method

The experiments were performed on a whole body scanner (Siemens Magnetom Expert) operating at a magnetic field strength of 1.0T. During the tests no absorbent compound was used, and water flow within the fractured porous media was not disturbed. A standard ‘spin-echo’ sequence was applied to obtain images of local proton density. For this type of pulse sequence, the sample was excited by a 90° radio frequency (rf) pulse and the magnetization was refocused by a 180° rf pulse. Both pulses had sinc-shaped envelopes, resulting in a band-selective frequency response which, in conjunction with a z-gradient, selected the section of interest. Discrimination in the x–y plane was obtained by applying field gradients in y-direction for phase encoding, and in x-direction for frequency encoding (the latter applied during data recording). A 2-D Fourier transform of the frequency spectrum was used to construct the final spatial image of the individual section. The 2-D images were acquired with a repetition time TR (the delay between two successive excitations) of 1.0s, an echo time TE (the delay between 90° pulse and the middle of generated spin-echo) of 12.0ms and a measuring time (TA) of 4min and 16s. The resulting images have a matrix of 252×256, a field of view (FoV) of 300×300mm². The resolution was limited by the software for medical use, which adjusted the gradient coils to scan a volume of around 0.1×8×0.1mm³ (in the x, y, z directions as in Fig.1). The resolution of the technique allowed images to be obtained by averaging over 8mm in the y direction (Fig.1). This maximizes the 3-D coverage of the sampling while maintaining optimal resolution in the orientation of the section. Using five sections, this corresponds to a total of 40 per cent of the sample being imaged. This resolution is larger than the typical fracture aperture (on the order of 50µm), so individual fractures are imaged more coarsely than they exist in nature.

Figure 1

Diagram showing the five positions where images are taken. (a) Three dimensional geometry of experimental set-up. (b) Horizontal section showing location of vertical NMR scans positions.

4 Core selection

The sandstone that was chosen for the present study was obtained from the middle Stubensandstein unit of the Middle Keuper succession in the south-west German Trias. There are two major sandstone formations, which are important regional aquifers: the Buntsandstein and the Stubensandstein. In contrast to the more quartzitic Buntsandstein, the Stubensandstein is characterised by a mixed mineralogy and high matrix porosity, therefore it is well suited for this study. A separate mineralogical study showed that the samples are arkose, with well- to poorly sorted grains. Quartz is the dominant framework of the sandstone samples (70 per cent to 80 per cent), with some feldspar (10 per cent to 25 per cent), and micas are rare. The remaining fraction is cement, the most common being clay minerals including kaolinite, smectite and illite, carbonate (calcite or dolomite) and quartz in the form of overgrowths (Baraka-Lokmane 1999). The samples were obtained by coring at surface exposure subject to long-term interaction with the atmosphere, and are hence water-wet.

The samples contain fractures in the form of opening-mode joints, with no appreciable shear offset and no cataclastic deformation bands. The type, nature and orientation of these fractures all have an effect on the permeability of the sample. The samples were selected so that they contained visible fractures that could be cored roughly parallel to the axis of the coring instrument. These criteria were designed to allow flow in the fractures to be more easily distinguishable from flow in the matrix given the geometry of the experiment shown in Fig.1.

We present results from two samples in this paper, numbered 2 and 20. These represent end members of open and sealed fractures. Their dimensions are 10cm in diameter and, respectively, 6.5cm and 5.7cm in length (a longer core is not necessary for a water imbibition experiment under ambient pressure). The hydraulic and physical properties of these two samples are presented in Table1 (after Baraka-Lokmane 1999). The rock samples were air-dried for 15days in an oven, and then the lower 1.5cm was placed in a receptacle of water. During the MRI measurements, the sample was then raised a few millimetres from the bottom of the receptacle of water, allowing water to infiltrate the sample by capillary action from its base. The sample with its receptacle of water was then inserted in the MRI apparatus system. Over a period of 2h 10m, a total of 55 images were made for the sample, covering the five sections illustrated in Fig.1(b). These images record the evolution of the water front and the level of saturation during imbibition of water by capillary action in a water-wet rock sample.

Table 1

The hydraulic and physical parameters of the two selected samples

5 Results

The MRI method can determine the distribution of water within the sample and, by inference, the geometry and the hydraulic conductivity of the fractures. The development and evolution of wetting front surrounding open fractures is shown in Fig.2. As a consequence of the initial deep immersion, the sample is initially saturated from the sides of the lower 1.5cm of the sample, but this effect decays quickly. Fig.2 shows that as long as the sample remains ‘air-dry’, the structure within the sample is not visible. Initially the fractures act as locations for preferential flow, with the water flood front (transition from grey to near-white tones) being channelled preferentially along the location of fractures (black tones indicative of high water saturation S_w). At later times the flood front is flatter, showing the absence of fracture-dominated flow in the upper half of the sample in the orientation of the sections shown here. The images of Fig.2, taken at the position −16 of Fig.1(b), show that the sample in fact has two conducting fractures in the lower part of the sample.

Figure 2

Distribution of water inside sample 2 at the position −16 of Fig.1(b) at different times after start of the experiment. Dark values correspond to high levels of water saturation S_w, and light areas to lower values. The base of the sample is initially immersed in water, as indicated by the flat black area at the sample base. The residual black areas to the sides of the sample are due to the sample being dipped and raised above the water level at the start time.

The 55 images for each sample for the 5 vertical sections were then interpolated to form 3-D images of the location of the wetting front the inferred fracture geometry. The results are shown in Fig.3, whose first image also shows the location of the particular section used in Fig.2 for cross-reference. Initially the flow front geometry is highly channelled by the fractures, subsequently becoming flat at later times. This confirms the presence of these two near-vertical fractures in three dimensions in the lower half of the sample.

Figure 3

3-D visualization of the development of the water front in sample 2 with time after start.

6 Comparison with the resin casting method for the determination of the fracture geometry

So far we have merely inferred the presence of fractures by (a) increased water saturation and (b) increased fluid flow by capillary action, using the non-destructive MRI method. In this section we describe the results of serial sectionings of the sample after resin injection, in order to validate this interpretation independently. Serial sectioning gives a direct, but destructive, test of the inferences of fracture geometry in porous sandstone made from the MRI described above.

To date, most research on fracture geometry has concentrated on otherwise low-permeability rock such as granite, where the hydraulic properties are dominated by fractures. For example, Pyrak-Nolte et al. (1987) injected Woods metal into fractures in crystalline rock held under a finite confining pressure. This alloy is a low-viscosity liquid at 100°C and a solid at ambient conditions, so it can access the fractures easily when hot, and then retain the geometry when it freezes at ambient temperatures. In their experiment the fracture is subjected to a known stress level and then filled with the hot liquid metal. This method has the disadvantage that fitting the two surfaces back together is difficult and, when the two fracture surfaces are taken apart, the metal can break and parts of it may stick to each surface, rather than providing a perfect cast.

An alternative method is to use a silicone polymer resin. For example Gentier et al. (1989) used a brand named RTV 141 injected into a granite sample under normal stress. In this technique two sides of the fracture are fitted together to act as a mould in which the fracture void space is filled with the coloured resin. Pressure is maintained at approximately 0.05 bar normal to the fracture while the resin is cured at 100°C for 4h. Cox et al. (1990) applied the same casting technique to create a replica of the void space enclosed between the walls of actual fracture specimen. The cast consists of translucent epoxy resin, the thickness of which can be measured by means of the degree of light attenuation under transmission. The main disadvantage of this technique is that the fracture aperture is no longer natural. Furthermore, the technique is not applicable for very small fracture apertures (on the order of micrometers).

Some experiments have also used an artificial fault gouge. For example Dollinger et al. (1994) injected gouge an into a granodiorite core with resin (Sikadur 52 of the Firm Sika) tuned to variable viscosity by adding different amounts of hardener. The resulting resins have a viscosity varying between 1700 cPs (at 20°C) and 83 cPs (at 13°C). They found the best results were obtained for the lowest viscosity of 83 cPs. The artificial fault gouge was injected under a pressure of 5 bar, with a flow rate varying between 1mLmin⁻¹ and 0.2mLmin⁻¹ for a total injection time of 1h and 30min under a normal load.

We also used a resin injection technique, but our experiment differs from the above examples in some important aspects. First, the rock matrix has a permeability that is comparable to (but smaller than) the fracture permeability (see Fig.2). The fractures are also stress-free, to simulate near-surface conditions. In order not to distort the fractures after the imbibition experiment, it important to inject the resin under as low a pressure as possible in order not to distort the fractures by opening them up in response to the associated poro-elastic stresses. We therefore constructed a new and more simple injection apparatus to accommodate these differences. After injection, the core can be sectioned without significantly damaging the resin, which perfectly fills the void space. In order to achieve this a new technique of resin impregnation was developed, as follows.

After the water-flood test, the sample was oven-dried for 15days and placed under vacuum before injecting the resin. After testing several resins, the most suitable was found to be SPURR, because of its low viscosity (60cp at 20°C) and its long pot life (2days). The sample was wrapped in a latex membrane and then fixed into a core holder. Resin was then injected through the sample from the top to the bottom using air pressure. At the same time, the apparatus is connected at its base to the vacuum pump. The injection period was one hour, and the injection pressure was increased in steps between 0.1 bar and 1 bar, with each step lasting 10min This procedure provided an optimum between the duration of the experiment and any disturbance to the fracture during sectioning. The results using the injection of coloured resin into the sample are very satisfactory.

After impregnation the core was cut into 5mm horizontal slices. The resulting cross-sections from the sample were then photographed and the fracture traces digitised using the graphic software program DESIGNER, Version 4.1a (Micrografx, Inc.). For each section, the fracture aperture and width were calculated using the image analysis program OPTIMAS (OPTIcal Measurement and Analysis System). The values of the effective fracture apertures vary between 10µm and 50µm (Baraka-Lokmane 1999). This confirms that the fracture aperture is much smaller than the resolution of the MRI method. As a consequence the saturations obtained by MRI will represent averages within a pixel resolution that includes some of the surrounding matrix.

Two examples of resin-impregnated horizontal sections of sample 2 are shown in Fig.4 and 5. The results clearly show both sedimentary layering in the intact rock, and the location of the fractures, both at a high resolution than could be imaged by MRI technique. A three-dimensional picture (Fig.6) of the fracture geometry was then produced using the software program IDL (Research systems, Inc.). This clearly shows the two fractures in the lower half of the sample imaged in vertical section on Fig.2. On the latter figure the two fractures appear disconnected in two dimensions, but this 3-D image shows that they are in fact connected. A third fracture is present in the upper half of the sample. It is approximately orthogonal to the first two, and so is not visible in the section of Fig.2 because the section is taken parallel to this fracture. We conclude that in general it will be important to image using the MRI technique in two orthogonal directions when fracture sets may be orthogonal.

Figure 4

Pixel map of the upper section of sample 2 impregnated with SPURR resin coloured in black. Note the filled fracture, and the sedimentary layers.

Figure 5

Pixel map of the lower section of sample 2 impregnated with SPURR resin coloured in black. Note the filled fractures, and the sedimentary layers.

Figure 6

Three dimensional visualization of the fractures of sample 2. (The irregularities and vertical segmentation are artefacts of the interpolation process).

7 Quantitative water saturation in open and sealed fractures

Water saturation is measured quantitatively using proton density maps (as a proxy for H₂O). Saturations are shown by the grey tones in Fig.2, but can be quantified within the analysis software by focusing the NMR sample on the volumes outlined as circles on Figs7 and 8. Fig.7 corresponds to the case of two open fractures as discussed above (sample 2) and Fig.8 for a sealed fracture (sample 20).

Figure 7

Local water saturation values measured on sample 2. Note the two conducting fractures. The circles indicate the areas where water saturation is measured; the corresponding mean saturations are also shown. (TR: repetition time, TE: echo time, TA: measuring time, SP: slice position, SL: slice thickness, FoV: field of view, 256×256: matrix (number of pixels) of the image).

Figure 8

Local water saturation values measured on sample 20. Note the sealed fracture. The circles indicate the areas where water saturation is measured; the corresponding mean saturations are also shown. (TR: repetition time, TE: echo time, TA: measuring time, SP: slice position, SL: slice thickness, FoV: field of view, 256×256: matrix (number of pixels) of the image).

In Fig.7 the water saturation in the matrix (volume 2 outlined) is 69 per cent in the pore space, compared to an ambient air humidity of 6.6 per cent (volume 3). From resin impregnation the two fractures shown in Fig.7 have an aperture of 43.6µm.

An example of water imbibition into a sample with a single sealed fracture is shown in Fig.8. The fracture is on the lower left-hand side of the image, this time filled by clay minerals. The fracture shows up as a lighter tone, implying lower water saturation. Some residual water remains in the fracture, but at a reduced intensity. (In fact some of this finite ‘saturation’ may actually be due to iron in the clay fill). The degree of saturation measured on either side of the fracture are different (23.2 per cent at position 2 on the lower left hand side of the fracture and 48.6 per cent at position 3 on the right hand side). The gradient across the fracture implies that the clay-filled fracture works like a barrier, inhibiting flow between these two areas. The ambient humidity was 6.6 per cent, showing a good cross-calibration of the sampling method with sample 2.

In this case the resin casting showed that the fracture aperture was 54.9µm. The sealing nature of the fracture was verified by resin casting, showing the clay fill directly after the test as well as the results of the flow measurements (BarakaLokmane 1999).

8 Discussion

It may seem surprising at first sight that the open fractures act as preferential conduits to fluid flow by capillary action, because their aperture is larger than the typical pore radius. According to standard capillary theory (based on cylindrical smooth-sided capillaries such as the inside of a pipette), the water column height is inversely proportional to the capillary radius. Hence the highest capillary rise should be in the tightest part of the pore space, in this case the rock matrix. We would therefore expect the wetting front to move faster through the matrix than the fractures, in direct contrast to the results of Figs2 and 3. This requires a systematic difference in wettability of the natural pore space and the fractures to overcome the aperture effect. This may be due to their different ages: the fractures are younger and have had less time to equilibrate with underground pore fluids. It may also be due to differences in specific surface energy due to the surface morphology, or differences in surface electrochemistry caused by mineral arrangements within the lattice. For example the quartz in the overgrowths around the mineral clasts is more likely to be amorphous, whereas it is more likely to be crystalline in the fractures that cut through individual grains. We conclude that simple capillary theory with a uniform wettability may not be applicable in predicting relative flow by capillary action in fracture surfaces and the rock matrix.

9 Conclusions

We have positively tested the hypothesis that the MRI technique can be used as a non-destructive technique for imaging the presence of fractures, and for visualizing their effect on fluid flow, within porous media with low iron content. Multiple sections can be used to build up a quasi-three-dimensional image of the water saturation, and hence indicate the presence of open or sealing fractures and the location of a wetting front during water imbibition by capillary action. The acquisition time of the images is relatively short (on the order of one hour), allowing a dynamic representation of the water flow to be obtained in defined sections of the sample. No absorbent compound is used, and water flow within the fractured porous media is not disturbed. The location, orientation and connectivity of the fractures inferred from the MRI images compares well with direct imaging of sections taken after resin-casting using a low-viscosity resin. The open fractures act as fluid conduits even under capillary flow, implying that they are more water-wet than the pore space. The sealed fractures, in contrast, show reduced water saturation, associated with blocking of the pore space by clay minerals. We conclude that the Magnetic Resonance Imaging technique can be used for the study of real-time flow in fractured porous rocks, and in particular provides a relatively rapid technique for determining the presence, geometry, saturation and wettability of open or sealed fractures. Some of these are quantitative to a high degree of resolution, and some more qualitative, but the information density is much higher than alternative techniques such as X-ray CT scanning.

Acknowledgments

This work was carried out as part of the project ‘Festgesteins-Aquiferanalogue: Experimente und Modellierung’, funded by Deutsche Forschungsgemeinschaft (DFG), Germany. The authors wish to thank PD Dr Dr Fritz Schick from the clinic of radiology of the University of Tuebingen, Germany for the different MRI measurements and for his kindness. S. Baraka-Lokmane was supported by Engineering and Physical Sciences Research Council, grant GR/M62150 during the writing of the paper. We thank Dr Sandy Steacy and two anonymous reviewers for their constructive reviews of an earlier draft.