Sandstone dikes in dolerite sills: Evidence for high-pressure gradients and sediment mobilization during solidification of magmatic sheet intrusions in sedimentary basins

Subsurface sediment mobilization and fluidization have been recognized from many geological settings, ranging from overpressured clastic reservoirs (Jolly and Lonergan, 2002; Mazzini et al., 2003; Nichols et al., 1994) to contact metamorphism around magmatic sill intrusions (Jamtveit et al., 2004; Svensen et al., 2006). In sedimentary basins affected by magmatic sill intrusions (i.e., volcanic basins), like the Karoo Basin in South Africa, sediment dikes are reported from within doleritic sills (Van Biljon and Smitter, 1956). It is interesting that these dikes comprise metamorphic sandstone, demonstrating that the sand intruded the dolerite while the sills were still hot. The importance of these observations is that they form direct evidence for high pore fluid pressure during sill emplacement and subsequent contact metamorphism.

In a classic study by Walton and O'Sullivan (1950), it was suggested that pressure drop during sill cooling and fracturing (i.e., thermal contraction) led to boiling of aureole pore fluids that ultimately led to sediment fluidization. That study was based on field examples from a sill emplaced in sediments during formation of the Central Atlantic Magmatic Province. The role of pore fluid boiling in causing high aureole pressures and subsequent fluid movement was explored in more detail by Delaney (1982; and more recently, e.g., Jamtveit et al., 2004).

Understanding sediment mobilization from contact aureoles may put important constraints on pressure evolution of aureoles. The past decade has seen an increasing interest in degassing of volatiles from sedimentary basins with magmatic intrusions, where high pore fluid pressure plays a key role (Ganino and Arndt, 2009; McElwain et al., 2005; Retallack and Jahren, 2008; Svensen et al., 2004, 2007, 2009). Gas venting triggered by overpressure in contact aureoles within shale has been proposed to have caused global climate changes in the end-Permian, Early Jurassic (Toarcian), and at the Paleocene-Eocene boundary (Svensen et al., 2004, 2007, 2009).

The aim of this study is to understand the formation of sandstone intrusions in dolerite sills. We present several case studies of sediment dikes and sediment breccias within sills in the Karoo Basin. However, the results can be applied to other sedimentary basins where sediments have been injected into magmatic sheet intrusions, including the Vøring Basin offshore

Norway, the Tunguska Basin of east Siberia, and the Amazonas Basin in Brazil. The process of sediment injections is addressed by adopting a new theoretical model for sill pressure evolution during cooling and crystallization (Aarnes et al., 2008).

The Karoo Basin (Fig. 1) covers more than half of South Africa. The basin is bounded by the Cape Fold Belt along its southern margin and comprises as much as 6 km of clastic sedimentary strata capped by at least 1.4 km of basaltic lava (e.g., Johnson et al., 1997; Smith, 1990). The sediments were deposited from the late Carboniferous to the Middle Jurassic, in an environment ranging from dominantly marine (the Dwyka and Ecca Groups) to fluvial (the Beaufort Group and parts of the Stormberg Group) and eolian (upper part of the Stormberg Group) (Catuneanu et al., 1998; Veevers et al., 1994). The Beaufort Group is a thick sequence of dominantly sandstones. The overlying Stormberg Group includes the Molteno Formation (coarse sandstone, shale, and coal), the Elliot Formation (sandstone, shale; red beds), and the Clarens Formation (sandstone with occasional siltstone horizons).

Both southern Africa and Antarctica underwent extensive volcanic activity in early Jurassic times, starting ca. 182.5 Ma. Dolerites and lavas of the Karoo-Ferrar large igneous province were emplaced within a relatively short time span. The main phase of flood volcanism lasted <1 m.y. (Duncan et al., 1997; Jourdan et al., 2005), although volcanism in southern Africa continued for several million years (Jourdan et al., 2005). Sills and dikes are present throughout the sedimentary succession in the Karoo Basin (Fig. 1) (Chevallier and Woodford, 1999; Polteau et al., 2008b), where they locally compose as much as 70% of the stratigraphy (Rowsell and De Swardt, 1976).

Sediment dikes are common within thick (70–120 m) dolerite sills within the Beaufort Group sediments. The depth of magma emplacement is estimated as 600–1000 m below the paleosurface, based on present-day stratigraphic levels. We have done detailed studies of three localities with sediment dikes in dolerites: (1) the Waterdown Dam area, (2) the Elandsberg roadcut (Nico Malan Pass), and (3) the Golden Valley (Fig. 1). Many more localities with sediment dikes have been discovered during our field work in the Karoo Basin during the past decade (e.g., south of Cathcart), but the chosen localities are representative. One of the sediment dikes from the Waterdown Dam locality contains numerous fragments of sediments and dolerite. It was mapped in detail by covering it with transparent A4 plastic sheets and tracing individual clasts by hand. This method was preferred over photo analysis due to better accuracy and the benefit of doing on-site interpretations on clast type and clast outline. The resulting map represents a two-dimensional (2D) slice through the dike. We then used image analysis techniques and a MATLAB (http://www.mathworks.com/) code to quantify the clast content (i.e., area). Probability densities were calculated using a smoothing procedure, where data were binned in either 10 consecutive areas (for sediment clasts) or 5 consecutive areas (for dolerite clasts). The aspect ratio between the long and short axes of the fragments was also calculated. Since our mapping analyses are done in 2D, and we only have one slice through the dike, the results should be regarded as approximate.

Thin sections of collected samples were studied by optical and electron microscopes (scanning electron microscope, SEM) at the Department of Geology, University of Oslo. The SEM is a JEOL JSM 840, and was also utilized for cathodoluminescence (CL) imaging.

We used Perple_X (Connolly, 2005) to compute phase diagrams for rocks with a pelite composition to predict the temperature stability of the mineral assemblages identified in the sandstone dikes. The calculated phase diagram is projected from an average pelite composition (Caddick and Thompson, 2008), with SiO₂ = 59.8, Al₂O₃ = 16.6, FeO = 5.8, MgO = 2.6, CaO = 1.1, Na₂O = 1.7, K₂O = 3.53, TiO₂ = 0.75, H₂O = 5.0 (all in wt%). We calculated the reactions using quartz saturation, which means that the phase assemblages obtained are not dependent upon the bulk content of quartz. Hence the phase diagram is valid for sandstones as well as pelites, as long as the ratios of the other oxides do not change significantly.

We have developed a numerical model using the finite element method (FEM) in MATLAB. We couple standard heat conduction to pressure (or hydraulic) diffusion using the equation for thermal stress similar to that of Aarnes et al. (2008). We calculate the pressure anomalies arising from pore fluid expansion of pure water in the contact aureole, and the pressure changes related to phase transitions (melt to crystal) in the sill. The pressure anomalies diffuse over time according to Darcy's law. The equations are solved on a 2D square grid with a resolution of 25 × 200 elements. Initial conditions for the thermal solver is a host-rock temperature T_hr of 35 °C, and a sill temperature, T_m, of 1200 °C. For temperature boundary conditions we fixed both the upper and lower boundaries at initial host-rock temperature, as the geothermal gradient is negligible on the scale of a few hundred meters. We assume a hydrostatic pressure gradient with a fluid density of 1000 kg/m³ as initial conditions for pressure. The upper and lower boundaries are fixed according to initial hydrostatic pressure. The boundaries do not influence the calculations.

We have developed a numerical model to quantify the first-order effects associated with sill cooling and pressure evolution. The model is conceptual and does not attempt to describe the full system. We assume an instant emplacement model of the sill because sediment dikes are related to postemplacement processes occurring at subsolidus conditions. The thermal diffusivities are equal for the sill and the sedimentary host rock, as differences in thermal properties are negligible (see Table 1). However, the hydraulic diffusion coefficients of melts and pore fluids differ by approximately one order of magnitude in our model. We assume no heat advection by fluids in either the sill or the contact aureole. This is justified from studies showing that heat advection by fluids is a second-order effect (Connolly, 1997; Podladchikov and Wickham, 1994). Apart from the sandstone dikes, there is little evidence of high fluid circulation in the intruded sediments, which makes heat advection within the intrusion negligible (cf. Norton et al., 1984).

The major assumption concerning the equation of thermal stresses is that expansion of pore fluids and contraction of melt due to crystallization are prevented either by the sediment matrix or the crystal network. This assumption is valid until the expanding fluids break the sediment matrix and reduce the overpressure, either by fluidization or by pervasive flow along the overpressure gradient. We account for fracturing of the host rock by resetting pressures that exceed the tensile strength of the host rock to hydrostatic pressure. We assume the tensile strength of our model sandstone host rock to be on average 35 MPa (Ai and Ahrens, 2004). We expect a drop in overpressure gradients with time, depending on how freely the mobilized sediments can move and reequilibrate the overpressure anomalies. For the underpressure, we expect the assumption of prevented volume change to be valid for the intrusion until the thermal contraction produces fracturing of the sill. Tensile strength of gabbroic rocks is >125 MPa (Ai and Ahrens, 2004). Such under pressure is not achieved in our model, which suggests that we are using conservative values. The main equations used for the modeling are shown in the Appendix.

Several sediment dikes within dolerite sills are located in roadcuts along the Waterdown Dam north of the Elandsberg area in South Africa (Fig. 1). The main sites are numbered 1–3 in Figure 1B, where thick sediment dikes are exposed close to the lower contact of a transgressive dolerite sill. The intruded sediments are mainly sandstones, all from the Permian and Triassic Beaufort Group. An overview of the locality is given in Figure 2A. At all sites, the field evidence suggests upward movement of sediment, based on the presence of dolerite bridges. The maximum upward penetration is not known, but is estimated to 10%–15% of the sill thickness based on the exposed dike heights and sill thickness.

At site 1 (S32°18.2′, E26°52.6′), a sediment breccia dike can be traced for ∼150 m westward from the main road, cutting vertically through at least 15 vertical meters of dolerite. The strike is 80° east, and it pinches out in both directions. The maximum thickness is 0.5 m and it splits in two branches toward the west. Sediment and dolerite fragments as much as 40 cm long are common, and bridge-like portions of dolerite are present locally (Fig. 2E). The latter suggests an eastward direction of emplacement.

Several thin sandstone dikes crop out at site 2 (Fig. 2B). The maximum width is 0.5 m, the strike is 84° east, and their vertical extension can be traced for 10–15 m in the roadcut. A few pieces of fresh dolerite are located within the dikes. These represent fragments of wall-rock dolerite broken off during dike emplacement, as also seen at site 1.

At site 3, a vertical dike as thick as 2.2 m crops out along the road (Fig. 2C), striking 78° east. This is, to our knowledge, the thickest sediment dike ever found in a sill intrusion. The contact with the dolerite is sharp, although weathered, and it comprises a breccia with sedimentary fragments as long as 40 cm. Some of the fragments show sedimentary layering. The lateral extension of the dike is unknown due to poor exposures, but the dike is located ∼10 m above the dolerite-sediment contact cropping out to the west. In addition to the sedimentary clasts, the dike contains numerous fragments of dolerite (Fig. 2D). The dike has been mapped in detail, and the results are presented in Figure 3A. The area percentage occupied by clasts and their size distribution have been quantified (Fig. 3B). The results show that the sandstone matrix (including clasts <0.5 cm) composes 86.4% of the area, sediment fragments occupy 10.6% (320 clasts), and dolerite fragments compose 3.0% (70 clasts). There is a four order-of-magnitude variation in clast size for the sedimentary fragments, but a lesser variation for the dolerite fragments. Note that the probability versus size relationship is similar for both sediment and dolerite clasts. The aspect ratio of the clasts length and width is calculated and shown in Figure 3C. It is interesting that the aspect ratio is independent of the clast size. The sediment clasts are more elongated compared to the dolerite clasts (aspect ratios of 2.51 and 1.95, respectively), which is also evident from Figure 3A.

The locality is located in the great escarpment defined by thick sill intrusions in the Beaufort Group sediments. A sediment dike was found intruding into the lower contact of the upper sill encountered when driving north toward the Nico Malan Pass along the R67 (S32°30.2′, E26°50.2′). The dike has penetrated 2.3 m into the inclined dolerite sill, and has a slightly curved and irregular shape (Fig. 4A). The maximum width is ∼20 cm, and the dike pinches out upward. No dolerite fragments were found in the dike, and the sandstone texture was markedly different at the tip of the dike compared to the surrounding contact-aureole sandstone, becoming increasingly recrystallized. No flow structures were observed in the sediment beds below the dike or within the dike.

The Golden Valley sill complex (Galerne et al., 2008; Polteau et al., 2008a, 2008b) is characterized by a flat inner sill that is partly exposed along a small river in the southern end. Here (S31°58.4′, E26°16.4′) a well-exposed part of the sill-roof hosts several small (<30 cm wide) sandstone intrusions (Fig. 5A). Note that the sediment source is located above the sill contact, demonstrating downward sediment movement. Note that in general, downward sediment movement is not unique for this location (e.g., Harms, 1965; Peterson, 1968; Vitanage, 1954). The dikes are irregularly shaped, and are characterized by a network-like pattern. Brownish alteration haloes are common around the dikes. The intruded sediments are Beaufort Group sandstones and shales, where the sandstones contain abundant nodules with radial fracture patterns (Fig. 5B). These nodules were originally composed of carbonate, but were modified during metamorphism.

We studied thin sections of sediment dikes from the Waterdown Dam (site 3), the Nico Malan Pass, and Golden Valley. The main aims were to identify metamorphic minerals, characterize the texture, and characterize the metamorphic conditions. The diageneses of nonmetamorphic sandstones located far from sill intrusions in the Karoo Basin are characterized by authigenic minerals stable at relatively shallow burial (clay minerals, K-feldspar, calcite, albite, and quartz) (e.g., Rowsell and De Swardt, 1976; Svensen et al., 2008; Turner, 1972). Typically, detrital grains (like quartz and K-feldspar) get coated and overgrown by authigenic minerals during burial without affecting the composition or texture of the grain interiors. At 1 km of burial in the Karoo Basin, the original sandstone porosity could have been ∼10%–25%, presumably filled with low-salinity pore fluids. After contact metamorphism of sandstone within the sediment dikes, detrital components of the quartz grains are still easily recognized, whereas feldspar grains (plagioclase and K-feldspar) were recrystallized in mosaic patterns. Moreover, the rock porosity is negligible, and chlorite and biotite are commonly present. Further details of the effects of contact metamorphism of sandstone injections from the examined localities are given in the following.

The sediment dike at Waterdown Dam contains metamorphic sandstone. Former grains and grain boundaries, representing the original sedimentary components, are easily recognized (Fig. 6A). This is confirmed by CL imaging of quartz (Fig. 6B). In addition to quartz, the dominant minerals that recrystallized in the dike are K-feldspar, plagioclase, and chlorite. Detrital feldspar grains are recrystallized and contain a mosaic of K-feldspar and plagioclase. Based on the gray-scale variations on SEM backscatter images, the plagioclase is characterized by several different compositions, apparently in textural equilibrium (Fig. 6C), and thus recrystallized during high-temperature metamorphic conditions. Biotite was not identified in the studied sample, but abundant chlorite could possibly be a product of biotite retrogression. The altered dolerite fragments in the dike are dominated by chlorite.

The Elandsberg sandstone dike contains identifiable detrital sand grains with quartz overgrowths. A sample from 2.3 m into the dike was studied using SEM, where CL imaging revealed detrital quartz cores. The presence of metamorphic epidote and biotite is important. The biotite is partly altered to chlorite, although the dominating mode of chlorite occurrence is in fresh patches unrelated to alteration. Feldspar grains are recrystallized and comprise mixtures of K-feldspar, albite, and plagioclase. Generally, the textures within the dike sample are tight and typical hornfels-like.

The sandstone dikes from Golden Valley have the same mineral content as the one at Waterdown Dam. Quartz, plagioclase, and chlorite are the main minerals. One difference, however, is the plagioclase textures. In Golden Valley, the detrital plagioclase is apparently completely recrystallized and zoned, present as tabular crystals (Fig. 5C). Ilmenite and apatite are minor minerals. The chlorite is locally present as tabular crystals, possibly suggesting biotite replacement. We have compared this mineral assemblage with the assemblage within a former carbonate nodule with sandstone from the same locality. The sample (Fig. 5B) is located ∼2–3 m above the contact with the dolerite sill, and is characterized by radial fractures extending out from a zoned nodule. In thin section, the main minerals are quartz (with detrital cores and overgrowths), feldspar, chlorite, and zeolite (Fig. 5D). The boundaries between detrital cores and metamorphic quartz are marked by rims of metamorphic garnet. The plagioclase is partly dissolved, and the pores filled by zeolite. Chlorite is common within the zeolite.

The studied textures from the examined localities show that the sandstone dikes underwent medium-temperature metamorphism following injection. Original quartz grain boundaries and grain cores are still preserved and document that the sediment dikes did not undergo partial melting after emplacement. This is consistent with the absence of macroscopic melt patches in the dikes. Diagnostic peak metamorphic minerals are sparse in metasandstones due to the low iron and magnesium content. Recrystallization of quartz and feldspar grains is the dominant mode of mineralogical transformation. However, the occurrence of minerals like chlorite, biotite, plagioclase, and epidote is typical for greenschist facies conditions. Based on the general presence of these phases in the sediment dikes, we can use phase petrology to constrain the peak metamorphic conditions. We have made a phase diagram projected from a pelite composition, and compared the calculated phase assemblages with those identified in the rocks in order to determine the temperature during dike emplacement (Fig. 7).

We have developed a numerical model in order to calculate the pressure gradients developing between an igneous sill and the surrounding sedimentary rocks as a function of temperature. Here we present snapshots of the temperature and pressure state during sill cooling. The modeling is based on the parameters listed in Table 1.

At the time of instantaneous emplacement, the 100-m-thick sill is hot (1200 °C) with a sharp thermal boundary to the cold host rock (35 °C) (Fig. 8A). Note that the gradient will be similar even if the sill is emplaced by continuous infilling and inflation. After 15 yr the temperature increases rapidly in the host rock, causing thermal expansion of the pore fluids, which results in overpressure of ∼22 MPa relative to the hydrostatic pressure gradient (∼7–8 MPa) (Figs. 8A, 8B). A fracture pressure of 35 MPa is indicated by a red dashed line in Figure 8B. The sill is in a state of underpressure due to cooling and crystallization of interstitial melt in a solid crystal network. The major mechanism of underpressure within the sill (-22 MPa relative to the hydrostatic pressure gradient) is due to a density change when interstitial melt (2600 kg/m³) is crystallizing (2900 kg/m³) within a confining crystal network during cooling. Note that the tensile strength of a gabbroic rock is >125 MPa (Ai and Ahrens, 2004).

After 100 yr, the sill has solidified and the temperature gradients become less steep (Fig. 8C). Correspondingly, the pressure gradient anomalies are reduced through diffusive fluid flow. The internal gradient within the sill is dispersed, and the main gradient is now from the host rock into the sill (Fig. 8D). The difference between the maximum overpressure (∼4 MPa) in the host rock and underpressure in the sill (∼–7 MPa) is ∼10 MPa, relative to hydrostatic pressure.

The fluid pressure in the aureole is increasing after sill emplacement due to the density change associated with heating of the H₂O pore fluid. Assuming a pore fluid pressure of 25 MPa at ∼1 km depth, H₂O undergoes a density reduction from 1004 kg/m³ to 162 kg/m³ when heated from 35 °C to 400 °C (Wagner and Pruss, 2002). The thermal expansion of the sediment matrix for the same temperature interval is negligible relative to the phase transition in the fluid (i.e., boiling). The overpressure is released through diffusive flow, with rates depending on the permeability of the host rock. Similarly, the underpressure within the sill is created during cooling and crystallization. The thermal contraction associated with the melt to crystal transition is several magnitudes larger than the thermal contraction of the surrounding network for the same temperature interval. Hence, an underpressure will develop as a response to the density change of interstitial melt (>55% crystals; Marsh, 1996; Philpotts and Carroll, 1996). This underpressure can be relaxed through internal melt flow from the molten to the crystallizing regions of the sill. When the sill is 100% crystallized, the thermal stresses will continue to develop as long as the thermal contraction is larger than what can be accommodated by volume change. The stresses can be released through brittle fracturing of the rocks, which in turn can be filled in by, for example, fluids, interior melt, or fluidized sediments (e.g., Norton et al., 1984).

We have made a thermal model with a realistic sediment dike geometry to estimate the maximum temperature attained within the dike at a given sill temperature. We emplace a 20-m-tall and 2-m-thick sandstone dike with an initial temperature of 35 °C into a 100-m-thick sill with sill temperatures between 1100 and 1200 °C (Fig. 9). As expected, the dike rapidly reaches peak temperature (i.e., within 1 yr). Hence, the initial temperature of the sandstone dike is not important for the final maximum temperature recorded in the sill. If the dike is injected 15 yr after sill emplacement, the sediment dike reaches a temperature of ∼850 °C. Injection at the time of sill solidification (i.e., at ∼100 yr), the peak temperature in the dike is ∼650–675 °C. For the sandstone dike to be heated to a maximum of ∼450 °C (cf. Elands-berg), injection after 300 yr of sill cooling is indicated. After 600 yr the sill has cooled to such an extent that the temperature in the sediment dike never exceeds 350 °C (cf. Golden Valley and Waterdown Dam).

In contrast to the 30–70 m.y. time scale of fluid production and pressure buildup during regional metamorphism and orogenesis (e.g., Connolly and Thompson, 1989; Walther and Orville, 1982), contact metamorphism around igneous sill intrusions in sedimentary basins have dramatic and short-term effects on fluid flow. This is particularly important in basins with rapidly cooling sill intrusions compared to settings with >100 k.y. of contact metamorphism around plutons (e.g., Hanson, 1992, 1995). When sedimentary host rocks are heated around sills, pore fluid expansion and boiling occur on a time scale of years, dominating the fluid production compared to devolatilization reactions (e.g., Delaney, 1982; Hanson, 1995; Jamtveit et al., 2004). Overpressure related to boiling and pore fluid expansion may ultimately lead to hydrofracturing and the formation of hydrothermal vent complexes in the upper 1 km in the basin (e.g., Jamtveit et al., 2004). In the Karoo Basin, the hydrothermal vent complexes commonly crop out in the Stormberg Group sediments. In addition, numerous breccia pipes are rooted in contact aureoles of black shale, demonstrating that high pore fluid pressures developed during rapid cracking of organic matter to methane (Svensen et al., 2007). Thus contact metamorphism around sill intrusions is a process that causes rapid pressure buildup and drives fluid flow on a very short time scale. In this setting, sediment dikes represent direct evidence for the rapid release of aureole pressure and fluids.

It has been shown that sill cooling and crystallization result in an underpressure within the sill (Aarnes et al., 2008). Underpressure generation is caused by the following. At the earlier stages of the sill cooling, a solid crystal network (>55% crystals) with interstitial melt will form (Marsh, 1988, 1996; Philpotts and Carroll, 1996). With further cooling the interstitial melt undergoes a significant density change due to the melt-to-crystal transition. However, a strong crystal network prevents a volume change and causes a large underpressure to develop. Experiments have shown that a crystal network have considerable strength already at 35% crystals, and effectively behaves as a solid even with large amounts of interstitial melts (Philpotts and Carroll, 1996). Such an underpressure may induce melt flow, have consequences for the chemistry of the magmatic system, and induce sediment injections into the sill (Aarnes et al., 2008).

During the initial stages of sill cooling, the pore water in the aureole sediments will expand and flow either away from the sill or into the sill, depending on the pressure gradients. Melt may also flow within the sills along the pressure gradient toward the cooling margins (cf. Fig. 8). The fluid flow is a result of the developed pressure anomalies and will act to even out the pressure anomalies with time. After 100 yr the pressure gradient within the sill is reversed, going from the margins to the center (Fig. 8B). However, the melt is now unable to flow as solidification is complete. At this time, the pressure in the sedimentary host rock has effectively been diffused by fluid flow. Thus, the main pressure gradient is now from the host rock toward the sill, both above and below the intrusion. At this stage, heated pore fluids will flow into the sill if permeability allows the fluids to enter, i.e., if fractures develop.

Fluidization due to heating of water-rich sedimentary rocks is most likely to occur at depths where pressure is less than the pressure corresponding to the critical point of water (Jamtveit et al., 2004; Kokelaar, 1982). The paleodepth of the study areas with sediment dikes in dolerites is ∼600–900 m, thus shallower than the critical depth. In some geological settings overpressure can cause horizontal fracturing through fluids seeping away from the overpressurized source (e.g., Cobbold and Rodrigues, 2007; Mourgues and Cobbold, 2003), while in the case of boiling and very high overpressures, modeling has demonstrated that the gas release may localize vertically and eventually reach the atmosphere (e.g., Jamtveit et al., 2004; Rozhko et al., 2007). The key requirements for pressure-induced sediment mobilization in the aureole are low permeabilities, high porosities, and high thermal diffusivities (Delaney, 1982; Jamtveit et al., 2004). In the case of high permeabilities in shallow sandstones, the rate of heating must exceed the rate of pressure loss by fluid flow in order to build up significant overpressure. When the sedimentary host rock undergoes extensive pressure buildup, it may ultimately lose all cohesive strength and become fluidized and result in substantial sediment displacement (e.g., Harms, 1965; Kokelaar, 1982; Ross and White, 2005; Vitanage, 1954). The sandstones of the Beaufort Group in the Karoo Basin were still in the early to intermediate stages of diagenesis (i.e., reached quartz cementation) at the time of sill emplacement. Thus the conditions were right for fluidization to occur, at least where clay minerals limit relaxation of pressures through fluid flow (Jolly and Lonergan, 2002), or as mentioned, if heating was rapid compared to pressure drop by fluid flow (Jamtveit et al., 2004).

Heat-induced overpressure and subsequent fluidization of sediments in the contact aureole is here suggested to be the main formation mechanism of sandstone dikes in magmatic intrusions. We show that there is an additional strong gradient from the aureole into the intrusion, and that this gradient makes sediment mobilization more likely to happen compared to injections driven by pore fluid boiling and fracturing during thermal contraction. However, we argue that fracturing during thermal contraction is of lesser importance, as sediment dikes are not present in a hexagonal network even in areas with abundant fractures developed during thermal contraction (e.g., the Golden Valley locality). The overpressure scenario is schematically presented in Figure 10. Our results show pressure anomalies of as much as 10⁸ Pa after solidification of the sill, in agreement with the magnitude 10⁷ Pa overpressure commonly found for several rock types due to expansion of pore fluids from magmatic intrusion (Delaney, 1982). It is important that the high pressure is sufficient to break the tensile strength of sandstones above ∼1 km depth (e.g., Kokelaar, 1982), thus fluids can potentially flow from the aureole and into the sill.

We therefore argue that sandstone dikes form as a result of the difference in pressure between the sill and the aureole (∼10 MPa) that develops during sill cooling and sediment contact metamorphism. The pressure gradient is sufficient for fracturing the sill (pressures beyond the lithostatic) and to act as a suction force on the sediments from the moment the chilled margin of the sill fractures. Once initiated, the fracture will propagate as a result of the injected pore fluids and sediments; this also may lead to further tensile failure (cf. Rubin, 1993). The fracturing process may be violent, as indicated by the high proportion of both sedimentary and doleritic rock fragments in the dikes at Waterdown Dam. Sediment fragments compose 86% of the dike surface at site 3, and the size distribution between sedimentary and dolerite clasts suggests that the same process was responsible for brecciation of both dolerite sill and aureole sediments. The four orders of magnitude variation in clast size (Fig. 3B) demonstrate that the brecciation was rapid and that the bulk of the breccia was injected into the sill.

The sediment dikes described in this study are all affected by contact metamorphism. Thus they were heated while the sill intrusions were still hot, either in situ in the contact aureole prior to injection, or within the sediment dike. Metamorphism of the injected sediments is a common observation from all sediment dikes in magmatic sill intrusions (Van Biljon and Smitter, 1956; Walton and O'Sullivan, 1950). Based on the metamorphic minerals in the dikes and aureoles from the Karoo Basin (chlorite, biotite, plagioclase, epidote, and garnet), the metamorphic conditions were equivalent to those of the greenschist facies. Based on these minerals and the phase diagram (Fig. 7), a maximum temperature of ∼450 °C is suggested. There are no accurate thermometers that can be applied to the identified mineral assemblages, so the temperature is approximate. Comparing with active hydrothermal metamorphism of sandstone, biotite appears at ∼320 °C (e.g., Schiffman et al., 1985), so our estimate is reasonable. The absence of minerals like cordierite, clinopyroxene, and muscovite furthermore suggests temperatures <∼450 °C, although the potential for generating some of these minerals depends on the bulk rock composition. As the temperature of heated sedimentary rocks around a sill intrusion will never exceed about half the sill temperature, a doleritic sill (∼1200 °C) will commonly not be able to melt the host sediments, and maximum temperatures should be close to 600–700 °C, depending upon the host-rock temperature at the time of emplacement. However, this situation may be different in other geological systems (e.g., Hersum et al., 2007).

The temperature estimates from the mineralogy are of importance when assessing the timing of sediment dike emplacement. As we have shown, an early emplacement into a hot sill will result in high-temperature metamorphism in the dike. Based on our thermal modeling, injection after 250–600 yr of sill solidification will give 325–450 °C in the dike. Note that reaction kinetics or significant latent heat of vaporization may contribute to discrepancies between modeled heat from conduction and that of a natural system. Earlier timing of sediment injection is therefore possible.

To summarize, our data suggest that the emplacement of the sediment dikes occurred after the sill was 100% crystallized, which puts a lower boundary to the timing of injection of ∼100 yr.

This means that sediment injection into sills has only limited potential for contaminating the magma, since the sill is 100% crystallized at the time of sandstone injection. For contamination to happen, the sediments would have to be injected into a partly molten sill, for which we have no supporting observations.

Field evidence shows that sediment dikes can propagate tens of meters into dolerite sills from the lower contact. The vertical termination of dikes has, however, not been found in the field. However, since the metamorphic recrystallization led to very low permeabilities, the sediment dikes were prevented from becoming long-lasting fluid flow pathways.

The basin settings in which sediment dikes within igneous sills are not likely to form are (1) when the overpressure difference between sill and aureole is small, as when the sill intrusion is thin, or (2) the aureole has limited potential for generating overpressure during heating, e.g., when the porosity is very low or the content of organic matter is negligible. Thus the presence of sediment injections in igneous systems may provide important constraints on the pressure evolution and fluid flow history in sedimentary basins with sill intrusions.

Sediment dikes have been discovered within dolerite sill intrusions at several localities in the Karoo Basin in South Africa. The sediment dikes contain metamorphic sandstone and clasts of sediments and dolerite. Field, petrographic, and numerical evidence suggests the following.

Both upward and downward movement of sediments into sill intrusions is common.

The sediments intruded while the sills were hot, producing mineral assemblages typical for >300 °C metamorphism.

Thermal modeling, to account for the dike metamorphism, shows that the sediment dikes were injected more than 100 yr after sill emplacement, depending on sill thickness and the initial sill temperature.

The presence of sediment dikes in sills is a result of the coupled pressure evolution of dolerite sills and contact aureoles. Negative pressure anomalies in the sill form due to cooling, whereas high pressure develops in the aureole due to thermal expansion.

The pressure generated is of the correct order of magnitude required to explain fracturing of the solidified sill. The sediments were accordingly drawn into the sill.

This study was supported by a PetroMaks grant from the Norwegian Research Council to Svensen. We thank Goonie Marsh and Luc Chevallier for discussions during our field trips to South Africa, in particular Goonie for showing us the Waterdown Dam locality, and Dirk Liss for the company and assistance during sampling of the sediment dikes. Else-Ragnhild Neumann and the Golden Valley Study Group at PGP (Physics of Geological Processes, University of Oslo) contributed with valuable input to the project. We also thank Joe Cartwright and an anonymous referee for critical comments.