INTRODUCTION

It has long been known that faults in sedimentary basins can act as either conduits or barriers to fluid flow (Hubbert, 1953; Smith, 1966, 1980; Weber and Mandl, 1978; Harding and Tuminas, 1988, 1989; Hooper, 1991). However, the detailed mechanisms by which faults conduct or seal hydrocarbons or other fluids are not well understood. Does significant fluid migration only occur along "active" faults? What are the detailed mechanisms responsible for fault sealing? The answers to these questions are critical in the prediction of trap charge and seal integrity.

This paper presents outcrop and microstructural data on the relationship between faulting and fluid flow in porous sandstone adjacent to the North Scapa fault, a major intrabasinal extensional fault in the Orcadian basin, northeast Scotland. Detailed microstructural observations from optical, transmission electron (Barber, 1985), scanning electron (Lloyd, 1985), and cathodoluminescence (Marshall, 1988) microscopy, and quantitative mineral analysis (Calvert et al., 1989) are described with the aim of (1) inferring the deformation mechanisms and fluid-flow processes that contribute to fault-zone sealing, and (2) assessing the influence of faulting on hydrocarbon migration.

REGIONAL SETTING

The Orcadian basin contains Old Red Sandstone (ORS) rocks of lacustrine and fluvial/eolian facies that now lie exposed in Caithness, northeast Scotland, and the Orkney Islands (Figure 1), as well as in

End_Page 625------------------------------

the Shetland Islands. The sediments are subdivided into the lower, middle, and upper Old Red Sandstone groups (LORS, MORS, and UORS, respectively) (Figure 2). Basinwide correlation depends on the MORS Achnarras/Sandwick Fishbed horizon, a lacustrine deposit that contains a distinctly diverse fish fauna (Rayner, 1963; Trewin, 1976, 1986; Westoll, 1977; Astin, 1990). The earliest age placed on the LORS sediments is early Devonian (?Emsian), whereas the UORS may extend into or be coeval with the Lower Carboniferous (T. R. Astin, 1990, personal communication) (Figure 2).

Exposures on the Scottish mainland and Orkney islands and offshore seismic data indicate that initial extension in the Orcadian basin began in the Early Devonian (Enfield and Coward, 1987; Astin, 1990). Deposition of the ORS took place over a series of east-southeasterly-dipping faults that defined tilted fault blocks (Astin, 1985; Enfield and Coward, 1987; Rogers, 1987). Uplift of the basin began in the middle Carboniferous and continued through the middle Cretaceous (Figure 3). During the Permian-Triassic, the Middle Devonian sediments of the Orcadian basin were remagnetized, suggesting that the present topography was relatively close to the land surface at that time (Robinson, 1986). Many faults that were reactivated during the later inversion are crosscut by Permian-Triassic dikes in the Orkney area (Brown, 1975; Baxter and Mitchell, 1984) (Figure 3).

Earlier in the Carboniferous, the lower MORS lacustrine sequences in the Orcadian basin (the upper and lower Stromness Groups) (Figure 2), which include carbonate laminites of significant source rock potential, were buried to their greatest depths of 2-3 km (Parnell, 1985). At present, these source rocks are mature (Marshall et al., 1985), that is, they entered the oil-generation window at some time in the past. From sedimentological data, Astin (1990) concluded that the source rocks in the Orcadian basin reached peak maturation during the latest Devonian to middle Carboniferous (Figure 3). Thus, hydrocarbons in the basin would have been expelled and become available for migration just prior to the onset of basin uplift.

The North Scapa fault, a major southwest to northeast-trending fault that is exposed on the southern portion of Mainland island (Figure 4), was active during evolution and uplift of the Orcadian basin. Deformation of the North Scapa Sandstone in the hanging

Fig. 1. Distribution of Old Red Sandstone (ORS) rocks in northeast Scotland and the Orkney Islands, and location of the study area (Orphir Bay) on Mainland island.

Fig. 2. Stratigraphy of the Old Red Sandstone within Orkney, after Mykura (1983) and Rogers (1987). A major hiatus is present at the Middle to Upper Devonian boundary, and the extent of the Upper Devonian succession is unknown. The dashed line within the Eday Group indicates the approximate location of the Eday Sandstone, which has been recently correlated with the North Scapa Sandstone.

End_Page 626------------------------------

wall of the fault was examined to address the questions posed in the introduction of this paper. The North Scapa Sandstone has been recently correlated with sandstones of the upper MORS Eday Group (Figure 2), the Eday Sandstones (T. R. Astin, personal communication, 1990), which were buried to maximum depths of 2.5 km (Astin, 1990) (Figure 3).

NORTH SCAPA FAULT

Field Observations

Along the length of its exposure on Mainland island, the North Scapa fault juxtaposes upper Stromness (middle MORS) lacustrine facies shale in the footwall against North Scapa (upper MORS) fluvial facies sandstone in the hanging wall (Figure 4). This study was carried out in Orphir Bay, where a well-exposed section of the fault is observed (Figure 5).

At this locality, the fault dips 64° to the southeast, and slickensides on the fault surface plunge toward the northeast. Fault-related deformation of the footwall block extends up to 600 m away from the fault, and is characterized by gentle folding of the shales. Ten meters from the fault, the shales are tightly folded, and bedding is sheared into parallelism with the fault plane (Figure 5). Bedding-parallel slip surfaces and conjugate fractures in the shales are observed up to 5 m from the fault plane. Two hundred meters from the fault-plane exposure, three Permian-Triassic dikes cut through the folded shales and are undeformed.

Fig. 3. Burial history diagram of the Middle Devonian Eday Sandstone Group and Sandwick Fish Bed in Orkney (after Astin, 1990, his Figure 9). The curve is constructed using the revised thicknesses for the lacustrine flagstones between the Sandwick Fishbed and the Eday Sandstone, and allowing for the presently compacted state of the sediments (Astin, 1990). The dashed curves assume 1000 m of Carboniferous sediments in the region. The ages of the dikes come from Brown (1975), Mykura (1976), and Baxter and Mitchell (1984). The dike magnetic ages and diagenetic magnetic ages preserved in the Eday Group are from Robinson (1985, 1986).

Fig. 4. Map of the Scapa Flow fault array on Mainland Island, Orkney. Exposures are limited to coastal outcrops, and the crosscutting relationship between the North Scapa fault and the East Scapa fault is unknown. The East Scapa fault has a complex movement history shown by varied kinematic indicators suggesting both reverse and normal movements (Hippler, 1989). A well-exposed section of the North Scapa fault is located in Orphir Bay (Figure 5), and deformation of the North Scapa Sandstone in the hanging wall of the fault is the subject of this study.

End_Page 627------------------------------

In the hanging-wall block, the North Scapa Sandstone is relatively undeformed more than 15 m from the fault plane, and bedding dips 20-25° to the northwest. This study focused on the deformed sandstone adjacent to the fault plane, where detailed mapping indicates that the deformed sandstone can be divided into two domains on the basis of the relative intensity of fractures, cementation, and evidence for the migration of hydrocarbons through the sandstone (Figure 5). The first domain of the fault zone extends up to 1 m from the fault plane, and consists of well-cemented, fine-grained sandstone. The second domain is located from 1 to 15 m from the fault plane, and contains fractured and brecciated sandstone. Both extension and shear fractures are observed in the second domain, with the shear fractures dominating the total fracture population (approximately 75%). Fracture lengths range from 1 m to greater than the height of the outcrop (5 m). Where offset markers are available, they indicate normal slip along the shear fractures, and offsets that range from several centimeters up to 1 m. Very rarely, a narrow (<5 mm) zone of gouge is contained in the shear fractures.

A decrease in fracture spacing corresponds with the development of breccia zones at 1 and 3 m from the fault plane. The breccia zones trend subparallel to the fault plane, and are 10-20-cm wide. The clasts in the breccias range in shape from rounded to subangular, and are <1 to 5 cm in diameter (Figure 6). The second domain also contains cataclastic zones that are coincident with hydrocarbon staining; these occur as narrow (<10 cm), anastamosing black bands oriented subparallel to the fault plane (Figure 7).

Microstructural and Petrographic Observations

Outcrop samples were taken at a spacing of approximately 1 m throughout domains one and two, and each was examined in thin section. A thin section of an undeformed sample just outside the fault zone and five thin sections from the fault zone were point counted (Figure 5; Table 1). Quantitative mineral analysis was performed on the undeformed sample and on a sample from the first domain, as this provides highly accurate estimates (within 3% absolute or ±10% of the amount of a given phase present) of mineral types and proportions in the bulk rock

Click to view image in GIF format. Fig. 5. [Color] Photomontage of the North Scapa fault as exposed in Orphir Bay on Mainland island, Orkney, looking northeast. The umbrella is 1 m long. The fault plane (the solid line indicated by FP) dips 65° to the southeast. The solid line nearest to the fault plane defines the boundary between the first and second domain of the fault zone, whereas the dashed line indicates the extent of the hydrocarbon staining observed in the second domain. The second domain extends up to the far right of the montage. The location of the samples referred to in Table 1 is indicated by the solid dots, with sample number increasing from left to right. See text for discussion.

End_Page 628------------------------------

(Calvert et al., 1989) (Table 1). In order to compare the total organic carbon present in the hydrocarbon stained zones in the second domain to the deformed sandstone in the first domain, two samples were submitted for microelemental analysis by gas chromatography (Table 1).

The grain size of the undeformed sandstone ranges from 180 to 200 µm. Although compaction has resulted in slight suturing of grain contacts, original grain boundaries appear to have been fairly well rounded (Figure 8). Quartz overgrowth rims are common. Present-day porosity in the undeformed sandstone ranges from 5 to 10% (Ridgeway, 1974). Quantitative mineral analysis indicates that the undeformed sandstone contains 88% quartz, 5% potassium feldspar, and 2% kaolinite (Table 1). Detrital mica is observed in thin section, but is rare.

In contrast to the undeformed sandstone, the first domain is characterized by a wide range of grain sizes (Figure 9). Large, 180-200-µm diameter quartz grains are typically surrounded by an extremely fine-grained matrix. Comparison of conventional scanning electron microscopy (SEM) micrographs with cathodoluminescence SEM micrographs reveals that the larger grains contain many transgranular fractures that are filled with quartz cement (Figure 10). Transmission electron microscopy (TEM) reveals that the matrix consists of quartz grains 2-4 mm in diameter surrounded by pore-bridging illite (Figure 11). Thus, grain-size reduction up to 2 orders of magnitude has occurred within the first domain. Pore space is virtually nonexistent in this domain, as TEM observations detect pervasive icrofibrous illite cement filling even the smallest pores (Figure 12). Quantitative mineral analysis indicates that the deformed sandstone from the first domain contains 89% quartz, 5% potassium feldspar, 6% illite, and a trace of kaolinite (Table 1).

The clast types observed in the breccia zones in the second domain are either fragments of relatively undeformed parent rock or the cemented cataclastic sandstone found only in the first domain. The matrix zones in the breccia consist of either an incohesive, fine-grained gouge or an organic-rich material. Optical microscopy of a breccia zone shows the clasts of cemented cataclastic sandstone, 0.5-2 mm in diameter, in a matrix of organic matter (Figure 13).

In thin section, the narrow cataclastic zones in the second domain are characterized by hydrocarbon staining and by sharp, planar boundaries that mark a significant decrease in grain size (Figure 14). Microelemental analysis of a sample containing these hydrocarbon-stained zones and a sample of the deformed sandstone in the first domain shows negligible organic carbon in the first domain, whereas the stained zones in the second domain contain a small but significant amount of organic carbon (Table 1).

DISCUSSION

Fault Timing and Kinematics

Outcrop and microstructural evidence suggests that two distinct periods of fault movement occurred along the North Scapa fault. First, the highly strained, cemented sandstone in the first domain is found as clasts in the fault-parallel breccia zone between the first and second domains. This indicates that the first domain was deformed and

Click to view image in GIF format. Fig. 6. [Color] Field photograph of the fault-parallel breccia zone separating the first and second domains of the fault zone. Note the angular to rounded clasts of cemented sandstone from the first domain within a dark matrix of organic matter. The pencil is 15 cm long.

Click to view image in GIF format. Fig. 7. [Color] Field photograph of the fault-parallel hydrocarbon-migration pathways in the sandstone from the second domain of the fault zone. The pencil is 15 cm long.

End_Page 629------------------------------

cemented as a result of earlier faulting, prior to a later event that subsequently incorporated clasts of the cemented sandstone into the breccia zone. Also note that the hydrocarbon staining is observed only in the second domain.

Astin (1990) showed that at least 55 m of thickening of the North Scapa Sandstone occurred in the hanging wall of the North Scapa fault during deposition of the first half of the Eday Sandstone on Orkney. This thickening suggests that the first movements on the North Scapa fault were syndepositional and occurred during basin extension in the upper MORS (Middle to Upper Devonian). Extensional faulting probably continued as the basin subsided and sediments became lithified while being buried to depths of up to 2.5 km (Astin, 1990). Thus, the first fault rock formed (in the first domain) corresponds to this syndepositional to postdepositional extensional faulting.

The presence of hydrocarbon staining in the cataclastic

Table 1. Distance From the Fault Plane of Outcrop Samples, First and Second Domains, Hanging Wall, North Scapa Fault, Orphir Bay, and Point Count, Quantitative Mineral, and Microelemental Analyses

Click to view image in GIF format. Fig. 8. [Color] Optical micrograph of the undeformed North Scapa Sandstone outside of the fault zone (Table 1, sample 7). Note the quartz grains with 150-200 µm diameters, quartz overgrowths, and pore space (blue). The scale bar is 0.25 mm.

Click to view image in GIF format. Fig. 9. [Color] Optical micrograph of the highly strained sandstone from the first domain of the fault zone (Table 1, sample 1). Note the wide range of grain sizes, with large 150-200 µm grains completely surrounded by fine-grained quartz/clay matrix. The scale bar is 0.25 mm.

End_Page 630------------------------------

zones within the second domain suggests that the faulting responsible for them took place at some time during, or just after, hydrocarbon expulsion. The MORS source rocks in the Orcadian basin reached peak maturation just prior to basin uplift (see Figure 3). Many of the faults in the Orcadian basin were reactivated during the uplift, including the North Scapa fault (Hippler, 1989). Several Permian dikes (Brown, 1975; Baxter and Mitchell, 1984) exposed at this locality crosscut deformation features associated with the North Scapa fault and are undeformed (Hippler, 1989). Thus, the fault reactivation

Click to view image in GIF format. Fig. 10. [Grey Scale] (a) Back-scattered SEM micrograph of sample 1 from the first domain of the fault zone. The scale bar is 10 µm. (b) Cathodoluminescence image of the back-scattered SEM micrograph shown in (a). Note that numerous transgranular extension fractures are now visible. Microprobe analyses indicates that the fractures are filled with quartz. The scale bar is 10 µm.

Click to view image in GIF format. Fig. 11. [Grey Scale] TEM micrograph of the fine-grained matrix zone in sample 1 from the first domain of the fault zone. Note the 2-3 µm quartz grains completely surrounded by pore-bridging illite. The illite was identified by qualitative microanalyses. The scale bar is 2 µm.

Click to view image in GIF format. Fig. 12. [Grey Scale] TEM micrograph of the pervasive illite cement in sample 1 from the first domain of the fault zone. The illite was identified by qualitative microanalyses. The scale bar is 0.5 µm.

End_Page 631------------------------------

probably occurred during the Carboniferous, just after hydrocarbons were expelled from the underlying source rocks and prior to dike intrusion.

The slickensides on the fault plane are presumed to be preserved from this last faulting event, and indicate sinistral, oblique-slip motion on the fault. A lack of marker horizons does not allow an estimate of the magnitude of displacement associated with the last faulting event.

Deformation Mechanisms

This section outlines the deformation mechanisms contributing to the fault rock evolution by reviewing and interpreting the microstructures that are observed in each domain and are associated with faulting events described above.

The extent of the grain-size reduction observed in the first domain of the fault zone indicates that strain was accommodated in the sandstone by fracture mechanisms. Microstructural observations suggest that as these fault rocks evolved, several fracture mechanisms contributed to the grain-size reduction process.

Large quartz grains remain intact within the fault rock and are similar in size to those observed in the undeformed host rock (Figure 9). The original grain boundaries appear to be preserved in the deformed rock (i.e., they are rounded or slightly irregular). These observations suggest that during initial straining, the quartz grains were disaggregated and incorporated into the matrix zones by grain-boundary fracture mechanisms (Atkinson, 1987). Further analysis by SEM cathodoluminescence reveals that many of the larger quartz grains were affected by transgranular fractures (Atkinson, 1987). The presence of smaller quartz clasts with angular or planar boundaries in the finer grained matrix zones of the fault rocks also suggests that transgranular fracture mechanisms operated (Figures , 10b). Transgranular fractures may be initiated as a result of stress concentrations at point contacts between grains or clasts in the manner suggested by the experiments of Gallagher et al. (1974) and described in natural fault rocks by Hippler and Knipe (1990) and Lloyd and Knipe (1992).

The fracture mechanisms contributed to the evolution of cataclastic fault rock that is associated with the first fault movement, a syndepositional to postdepositional deformation. Syndepositional faulting implies that the sediments were poorly consolidated and unlithified to weakly lithified at the time of faulting. Cataclasis has been documented in the deformation of poorly consolidated, weakly lithified sediments in accretionary margin settings (Byrne, 1984; Lucas and Moore, 1986; Moore and Byrne, 1987). However, the evolution of cataclastic bands and breccias is more commonly reported in association with deformation of lithified rocks at shallow crustal levels where brittle deformation mechanisms are expected to operate (Engelder, 1974; Sibson, 1977, 1986; Knipe, 1989). For example Aydin (1978), Pittman (1981), Jamison and Stearns (1982), Aydin and Johnson (1983), Bevan (1985), and Underhill and Woodcock

Click to view image in GIF format. Fig. 13. [Color] Optical micrograph of a fault-parallel breccia zone from the boundary between the first and second domain of the fault zone (Table 1, sample 3). Note the clasts of cemented sandstone within a matrix of organic matter. The scale bar is 0.5 mm.

Click to view image in GIF format. Fig. 14. [Color] Optical micrograph of the narrow cataclastic bands from the second domain of the fault zone (Table 1, sample 4), which trend subparallel to the fault plane. The bands contain organic matter (see text for discussion). The scale bar is 0.5 mm.

End_Page 632------------------------------

(1987) described cataclasis of high-porosity sandstones in a variety of tectonic settings.

It is difficult to determine whether the cataclasis observed in the first domain of the fault zone occurred during the syndepositional faulting. However, recognition that specific fracture mechanisms operated during evolution of the fault rock aids in elucidating the lithification state of the sediments at the time of faulting and the possible processes involved in producing the final microstructure. In particular, Lucas and Moore (1986) described the processes leading to cataclastic deformation of the weakly lithified sediments in accretionary margin settings. The fracture mechanisms observed in this study suggest that these processes may have also operated during initial movement along the North Scapa fault.

Lucas and Moore (1986) described a deformation sequence beginning with weakly lithified sands that are under low effective-confining stress due to excess pore-fluid pressure. During deformation in this situation, the cohesion and friction between grains can be easily overcome, and grain boundary fracture mechanisms operate to disaggregate the sediments (Knipe, 1989). Strain in the disaggregated sediments is then accommodated by nondestructive particulate flow mechanisms (Borradaille, 1981). As a result of the particulate flow, grain reorientation and repacking would allow the deformed sediments to achieve a greater degree of consolidation than the undeformed sediments. Continued strain of the consolidated material then leads to the development of stress concentrations at point contact between grains. As outlined earlier, transgranular fractures may be initiated as a result of these stress concentrations (Gallagher et al., 1974), and at this point, cataclasis takes over as the dominant deformation process in the sediment.

This deformation sequence would explain the preservation of the quartz-grain boundaries in the fault rocks in the first domain of the North Scapa fault, as well as the presence of numerous transgranular fractures. Although the contribution of the disaggregation and consolidation processes has been recognized in association with cataclasis of weakly lithified sediments, this example outlines the specific fracture mechanisms that may have been involved in the grain-size reduction process.

During later faulting, strain was accommodated in the second domain by the development of narrow cataclastic zones and broader breccia zones. The sandstones experienced maximum burial (2.5 km) just prior to this faulting and, under these conditions, brittle deformation mechanisms (i.e., fracturing and cataclasis) dominate faulting processes in rocks (Sibson, 1977). Microstructural observations confirm that fracture mechanisms operated during the evolution of the deformation features, as illustrated by angular clasts in the breccia zones and grain-size reduction in the cataclastic zones.

The deformation features associated with this later faulting are easily identified in outcrop and thin section due to the presence of hydrocarbon staining within them. The absence of such features in the first domain suggests that the later faulting did not affect the previously strained area. This observation is similar to the development of deformation bands in the sandstones of southeastern Utah described by Aydin and Johnson (1978). Aydin and Johnson (1978) recognized a progression from deformation bands to zones of deformation bands to discrete planes of slip, a progression which they argued reflects strain hardening of the sandstone. Strain hardening is thought to be caused by an increase in grain-to-grain contacts brought on by a reduction in grain size and porosity within the eformation zones. Due to the consolidation and cataclasis within the bands, the material becomes stronger than the surrounding, undeformed sandstone. As a result, other zones of deformation bands develop. Jamison and Stearns (1982) also found that a decrease in porosity and permeability in microfaults resulted in strain hardening in the Wingate Sandstone. Thus, at the time of the second deformation, the mechanical properties of the sandstone in the first domain of the North Scapa fault were such that slip was accommodated in the surrounding undeformed rock (presently domain 2).

Fault-Zone Fluid Flow

Microstructural analysis of fault-zone material has shown how specific deformation mechanisms alter the permeability structure of undeformed rock (Pittman, 1981; Mitra, 1988; Roberts, 1991). Integration of microstructural analysis with studies of cements and vein fillings in fault zones have provided information regarding past fluid flows and the mechanisms that influence fluid flow during a deformation (Sibson et al., 1975; Tillman and Barnes, 1983; Etheridge et al., 1984; Parry and Bruhn, 1986; Reynolds and Lister, 1987; Grant, 1989). For example, Grant (1989) used fluid-inclusion data from fault rocks in the Pyrenees to suggest that the dilatancy pumping mechanism (Etheridge et al., 1984; Reynolds and Lister, 1987) operated to transport packets of fluids into dilatant regions of th ust fault zones during a seismic deformation. Dilatancy pumping is driven by stress buildup that results in the opening of extension fractures or tension gashes into which fluids are drawn due to the fluid-pressure gradient that is created by the dilatancy. The fluids are then rapidly expelled during collapse of the dilatant pore space after failure, or they are moved along the fault zone due to a change in location of the dilatant zone during the next cycle of stress buildup and failure.

The faulting mechanisms which may have influenced fluid flow during deformation in sedimentary

End_Page 633------------------------------

basins are not well understood because similar analyses of fault zone material are limited. Instead, fluid inclusion and/or isotopic studies of cements in reservoir rocks have been used to provide indirect evidence regarding the mechanisms that control fault-zone fluid flow in sedimentary basins. Burley et al. (1989) studied fluid inclusions in Tartan field in the North Sea, and suggested that seismic pumping (Sibson et al., 1975) operated to drive large-scale fluid flow along faults. In the seismic pumping model, fluid flows into a dilating region surrounding a fault prior to an earthquake; subsequent seismic-strain release and collapse of the dilatant fracture volume then forces fluid out of the dilating region. Wood and Boles (1991) suggested that seismic pumping also operated in N rth Coles Levee field in the Bakersfield arch, California; however, they used carbon isotope data from reservoir cements to support their conclusions. Several paragenetic studies and age dating of reservoir cements have provided further evidence for large-scale fluid flow along faults in sedimentary basins, but the authors have not argued for a specific mechanism of fault-zone fluid flow (Flourney and Ferrell, 1980; Porter and Weimer, 1982; Jourdan et al., 1987; Lee et al., 1989).

Integration of the microstructural and diagenetic data obtained from the North Scapa fault provides direct evidence of the mechanisms that can influence fluid flow along faults in a sedimentary basin. Extensive illite cementation is localized in the first domain of the fault zone and suggests a strong structural control on fluid flow within or through the fault zone. Cementation of fault zones has been attributed either to microscale diffusive mass transfer of material between dissolution and precipitation sites (Rutter, 1976, 1983), or larger scale transport of material in solution via the mechanisms of seismic pumping (Sibson et al., 1975) or dilatancy pumping (Etheridge et al., 1984; Reynolds and Lister, 1987). In both cases, cementation occurs as a result of fluid-rock interaction that are initiated by a chemical disequilibrium between the migrating fluids and the host rock. If the amount of cement observed in a fault zone cannot be accounted for by local dissolution and reprecipitation of material, then another mechanism that can transport material in solution to the precipitation site is likely to have operated, for example, seismic or dilatancy pumping.

Because quantitative mineral analysis provides highly accurate estimates of the mineral types and proportions of the bulk rock, data from these analyses can help to constrain the magnitude of fluid flow responsible for the mineralogical differences between the undeformed and deformed sandstone adjacent to the North Scapa fault. The analyses indicate that the undeformed and deformed rock from the first domain of the fault zone contain approximately the same amount of quartz and potassium feldspar, but the undeformed sandstone contains 2% kaolinite, whereas the deformed rock contains only a trace of kaolinite and a significant amount of illite (6%). At temperatures of 120-130°C, kaolinite becomes unstable (Hower et al., 1976; Hoffman and Hower, 1979), and if potassium is present, k olinite then transforms to illite:

[EQUATION]

The presence of hydrocarbon staining at this locality indicates that liquid hydrocarbons were generated in the area, most probably near maximum burial of the sediments (Figure 3). Liquid hydrocarbons are generated at temperatures in the range 50-150°C (Hood et al., 1975). Thus it is possible that the sediments adjacent to the North Scapa fault were in temperatures that would allow the kaolinite-illite transformation to take place. However, the amount of potassium feldspar in the undeformed rock is the same as that in the deformed rock, and the amount of kaolinite available for dissolution cannot easily explain the amount of illite precipitated in the first domain of the fault zone. Additionally, there is no evidence for leaching of potassium feldspar grains in or around the fault zone. These mass balance arguments, although crude, suggest that the cement precipitation in the first domain must have involved significant fluid flux into the fault zone in order to transport the solids needed to precipitate the illite.

To focus the fluid flow within the low-permeability first domain, the cataclastic zones must have had a localized pore-volume increase compared to what is observed at present. Volume changes, specifically, pore-volume increase, or dilatancy, is known to accompany failure of intact rock in the laboratory (Brace, 1978). Experimental studies have shown that the magnitude of the porosity increase due to dilatancy during failure of porous rocks can range from 20 to 50% (Brace and Orange, 1968; Edmond and Patterson, 1972; Shipman et al., 1974; Schock et al., 1973a, b). Additional laboratory studies by Teufel (1981) and Raleigh and Marone (1987) have determined that dilatancy also precedes slip along faults and/or occurs during frictional sliding. These laboratory observations appear to be a alogous to the dilatancy which occurs in natural fault zones during seismic or dilatancy pumping.

Thus, the indication that both an influx of fluids into the fault zone and dilatancy are necessary in order to account for the illite precipitation suggests that either the seismic or dilatancy pumping mechanisms operated during deformation along the North Scapa fault. Because the origin of most cataclastic fault rocks in terms of their significance with respect to the seismic cycle remains an unresolved problem (Sibson, 1989), it is unclear whether seismicity occurred during slip along the North Scapa fault. Hence, it is difficult to assess whether seismic pumping is a viable mechanism for driving the fluid flow.

End_Page 634------------------------------

Alternatively, the dilatancy pumping mechanism could also explain the channelized fluid flow through the fault zone. Although dilatancy pumping may only be effective for transporting fluids on the scale of tens of meters, Reynolds and Lister (1987) noted that dilatancy pumping could result in overall updip flow or transfer of fluid packets along faults with high rates of displacement. This is because there is greater potential for an increased connectivity of fluid pathways, a higher frequency of dilatancy creation-collapse cycles, and a thermal contrast between rocks on either side of a large-displacement fault.

The source of the hydrocarbons in the second domain of the North Scapa fault is clearly not local, that is, they were not generated from the sandstone, but rather from the underlying organic-rich zones of the upper and lower Stromness Groups. Additionally, the fluids were focused through the now-impermeable cataclastic zones. These observations again suggest that the seismic or dilatancy pumping mechanism was involved in transporting fluids (in this case, hydrocarbons) through the fault zone.

To estimate rates and volumes of fault-related fluid flow, it is critical to identify how specific faulting mechanisms influence fluid flow. These estimates are important, because the rate at which fluid-rock interactions (and thus cementation) can take place is controlled by the rate at which reactants in solution can be delivered to the reaction site (Phillips, 1990). The magnitude and timing of fault-zone cementation subsequently determines the rheological properties (Angevine et al., 1982) and sealing potential of a fault.

Estimates of fluid flux associated with seismic pumping are the most prevalent. For example, Sibson et al. (1975) estimated that 5 × 10⁹ L of fluid could be released from a seismic pulse associated with a magnitude 6 earthquake. Stark and Stark (1991) presented the first quantitative assessment of seismic fluid flux using percolation theory to model the size distribution of slip planes. They modeled a fault segment with a length of 20 km (horizontally), a width of 10 km (vertically), a coseismic fracture zone width of 5 m, and a fracture porosity of 1%. They calculated that approximately 10⁹ L of fluid are released on such a fault after a major earthquake, which is in very good agreement with that of Sibson et al. (1975).

Sibson et al. (1975) used their estimate of seismic fluid flux to then calculate the amount of quartz that can be dissolved and precipitated in each seismic pulse. They calculated the solubility of quartz in pure water (grams SiO₂/kilogram solution) along a geothermobaric gradient of 35°C and 300 bars/km, and found that at 10 km, 10¹⁰ grams of quartz can be dissolved; if the expelled fluid cools to 100°C during ascent along a fault, more than 95% of the quartz will be precipitated. It is important to note that in an extensional basin where hydrocarbons are produced, the pressure gradients and temperatures are generally much less than those used in the calculations by Sibson et al. (1975). These parameters affect quartz solubility and thus the volume of uartz that can be dissolved and precipitated along faults.

In areas where seismicity is absent and/or more complex fluid-rock interactions are involved, estimates of fluid flux are limited, and more work is needed to be able to predict fault-zone cementation. This is beyond the scope of this study, which focuses on demonstrating how faulting mechanisms may affect sealing processes and hydrocarbon migration in sedimentary basins.

Fault-Sealing Processes

The observations from this study have important implications for fault-seal models that suggest that the presence of an impermeable rock directly adjacent to the fault is in many cases necessary to seal a fault. The impermeable rock is commonly a shaly top seal that is either "smeared" into the fault zone or simply juxtaposed against the reservoir rock (e.g., Smith, 1980; Harding and Tuminas, 1988, 1989). This study suggests that (1) cataclasis and extensive cementation of porous sands can provide a permeability barrier to migrating hydrocarbons, and (2) fault reactivation may not breach an existing seal where strain hardening has occurred.

Faults as Hydrocarbon Migration Pathways

From the above discussion (given a fault with access to a hydrocarbon "kitchen"), it is evident that fault activity that is contemporaneous with hydrocarbon migration can greatly influence migration directions. Yet it is unknown how broadly or narrowly coincident in time these processes must be in order to influence migration. In addition, the mechanics by which slip is accommodated along a fault (i.e., Tullis, 1986; Scholz, 1989) and the mechanisms which operate during the deformation (i.e., Knipe, 1989) will clearly have an influence on the migration of fluids through a fault zone. This study, in which cataclastic faulting mechanisms in sandstone were dominant, suggests that determining the timing of faulting relative to hydrocarbon maturation and migration could be extremely import nt in predicting which faults might act as migration pathways. In fault zones where only "ductile" processes have operated (i.e., particulate flow with the absence of grain fracture), the detailed mechanisms of fluid flow through the fault zone may be very different.

CONCLUSIONS

The North Scapa fault is characterized by fault-parallel domains that differ in terms of the distribution

End_Page 635------------------------------

of cataclasis, organic carbon content, and diagenetic cements. Outcrop and microstructural analysis of these features identified two periods of fault movement and associated fluid flow through the fault zone. Dilatancy associated with cataclasis during fault activity controlled the location of fluid pathways, which in turn governed the location of diagenetic cements and hydrocarbon pathways during the basin history. This study has emphasized the importance of establishing the timing of fault activity relative to hydrocarbon maturation and migration for the prediction of fault seal and in the assessment of secondary migration pathways.