During sand emplacement, the system must have transitioned from a flowing to geometrically locked grain geometry (the jamming transition; sensu Corwin et al. ). In the case of the Yellow Bank Creek Complex, the transition occurred when a sand-water mobile slurry (Figure 8b) decreased in fluid velocity until the grains made contact and locked into a network of framework sand grains with fluid flowing through the pores (Figure 8c). The Yellow Bank Creek Complex contains multiple fabrics, most of which we interpret to have formed either during or after the jamming transition between granular flow and pore fluid diffusion during dewatering.
 We observed high apparent porosity in the laminated sandstone. Thompson et al. also observed high apparent porosity, and explained this observation by post-emplacement grain dissolution. The pore spaces in our samples are significantly smaller than grains (Figure 6) so we consider this explanation unlikely although we can not rule out the possibility that some fine fraction was dissolved subsequent to emplacement. Scott et al.  measured porosity of the laminated sandstones (mean: 29%), and found lower porosity than our estimates (48% and 51%). However, Scott et al.'s measurements were performed on cemented samples, which under-represent the primary porosity of the sand. They also suggest that porosity at the time of consolidation must have been at least 46%, where 54% is the approximate maximum granular concentration for fluidization [Leva, 1959]. Based on this data and the preservation of the delicate laminae and other grain structures, our data are reasonable estimates for the original porosity of the sand just after the jamming transition. Blower et al.  showed that for spherical bubbles (we may consider bubbles to be geometrically analogous to grains) following a power law size distribution, 2D grainsizes will underestimate 3D grainsizes. Therefore, our 2D porosity values are likely slight overestimates of 3D porosity.
6.2.1. Textures Relating to Injectite Flow
 Wall rock clasts were ripped from the walls during high velocity slurry flow [Scott et al., 2009; Thompson et al., 2007] and can be treated as large, elongate grains in a viscous flow. Previous authors have used terminal settling velocity calculations to bound the minimum upward sand velocity, which requires estimating the density and viscosity of the sand slurry during flow as these quantities are not preserved [e.g., Duranti and Hurst, 2004]. Scott et al.  used this type of analysis to conclude that the Reynolds numbers of the Yellow Bank Creek Complex during injection required turbulent flow. However, they used fluid viscosity values near that of water [Kestin et al., 1978] and too low for a granular slurry viscosity, as demonstrated by experiments on sediment slurries. We therefore recalculate the Reynolds numbers using a range of viscosities of 1–30 Pa-s [Major and Pierson, 1992]. If inertia is potentially important, the appropriate terminal velocity is governed by the balance of inertial drag and buoyancy,
where g is the acceleration due to gravity, d is the diameter of the largest observed clast (0.7 m [Scott et al., 2009]), ρs is the grain density (2300 kg/m3 [Scott et al., 2009]), ρ is the fluid density (1250–1890 kg/m3 [Scott et al., 2009]) and Cd is the drag coefficient (0.5 for rough spheres). The resultant terminal velocities are 2–4 m/s where the range is provided by the range of densities from dilute to dense suspensions. The Reynolds number is
where L is the estimated width of the feeder to the injection (10 m) and ηis the range of slurry viscosities from dilute to dense suspensions (1–30 Pa-s). The corresponding range of Reynolds numbers is 1.3 × 103 to 4.9 × 104. Therefore, the turbulent flow inferred by Scott et al.  based on scouring of the wall rock was possible during initial transport, but a transition to laminar flow likely occurred as the slurry decelerated and density increased toward the jamming threshold. Whereas sand dikes observed by Boehm and Moore exhibit grain size sorting with coarser grains concentrated toward the center, consistent with laminar flow between the dike walls, the lack of sorting in the large sill at Yellow Bank Beach suggests that well-organized velocity gradients did not persist for long enough to effectively sort the grains. This observation allows that the sand slurry was emplaced as an initially turbulent flow which transitioned to laminar flow behavior during emplacement.
 Regardless of initial conditions, the flowing slurry behaved as a laminar flow as the velocity decreased. Mudstone clasts are gently dipping (17° ± 17°, N = 82) on the outcrop surface, consistent with rotating toward flow lines as the injectite sill flowed horizontally along bedding. Studies of phenocrysts transported in viscous flows have been shown to develop a characteristic alignment parallel with flow direction [Yamato et al., 2012] and sorting of grains by density and shape according to velocity profiles [Geoffroy et al., 2002]. Therefore, we hypothesize that local mudstone clast alignment is the best preserved record of last direction of sand slurry flow during injectite emplacement. We use this alignment as a tentative marker of flow direction, to which we can then compare the other granular features. There can have been almost no relative motion between wall rock clasts and the sand encasing them following emplacement. Any relative rotation would have destroyed laminae microstructures, and this is not observed, so we can assume the orientation we observe is primary (Figure 2c).
 The velocity of the slurry eventually declined until the sand stopped moving. This transition, the jamming transition, occurs where particle contacts have sufficient number and strength to stop grains moving relative to one another. The slowing velocity of the slurry reduced the capacity of the flow to suspend the grains, allowing them to settle into contact at the critical porosity. The simulations by Majmudar et al.  and experiments by Wilhelm and Wilmański  exploring the same transition in the opposite sense (mobilization by pore fluid flow) both indicate that the transition from jammed to unjammed state (stable to flowing or vice verse) is a sudden, threshold transition.
6.2.2. Textures Formed After Jamming
 Our observations of crosscutting relationships confirm that the iron oxide-cemented and uncemented laminations are a primary fabric defined by small grain-scale differences [cf.Scott et al., 2009]. The field observations of crosscutting relationships show that the laminae are the earliest-formed fabric in the injectite. They formed prior to large-scale dewatering, as they are deformed by compaction banding and convolute structures (discussed below). The iron oxide cementation exploited this primary fabric. In contrast toScott et al. we did not detect any packing difference between the laminae types. The only grain-scale difference we detected is a distinction in the apparent aspect ratio of grains. The apparent aspect ratio seen in the 2D thin section is a function of the true aspect ratio and the angle of the grain long axis to the plane of the thin section (rotation about the z-axis,Figure 6). Since the grain size distributions are the same, this suggests the grains were not sorted to form the laminae. We infer that the difference between iron oxide-cemented and uncemented laminae reflects a difference in the degree of alignment, or lineation, of grains within the plane of the laminations, although this distinction can not be directly measured in our 2D sections.
 The iron oxide, precipitated from fresh pore water at a later stage after uplift of the marine section, is preferentially concentrated in alternating laminae. Iron oxide cement occurs in the laminae with slightly higher apparent aspect ratio (consistent with stronger shape lineation). We suggest this reflects a difference in the permeability of the laminae, caused by and preserved from the initial orientation of the sand grains, that affected late stage groundwater flow.
 We propose that a permeability difference between otherwise similar laminae was established by a percolating fluid rearranging grains immediately following jamming of the granular slurry (Figures 8b and 8c). The apparent 2D porosity (∼50%) is similar to the critical packing densities reported for ellipsoidal grains [Garboczi et al., 1995; Saar and Manga, 2002]. Near the critical porosity small variations in porosity or pore geometry give rise to large permeability variations, even with identical grain populations [Ross et al., 2011; Wilhelm and Wilmański, 2002]. Permeability is a function of the porosity, as well as geometric details controlling the pore geometry and connectivity, such as grain shape, aspect ratio and orientation, and fluid pressure gradients [Iverson, 1997]. In granular material, when the porosity approaches a critical value, parameters that are dependent on porosity, such as permeability, exhibit power law changes in response to small porosity variations. This is the percolation threshold.
 We therefore suggest that near the percolation threshold, the pore fluid rearranged grains in such a way to cause the slight alignment contrast between higher and lower permeability laminae. This formed a feedback whereby the flow through relatively high permeability layers was promoted, further rearranging grains and enhancing the permeability difference. Pore fluid drained preferentially into and along the higher permeability layers (Figure 8c). These were the same layers that were later cemented with iron oxide, possibly because of preferred groundwater flow along them.
 We propose that the uniform thickness of the laminae reflects a characteristic length scale for the process controlling the permeability instability (laminae thickness, 5.7 ± 1 mm). The length scale was established by the relative contribution of pore fluid advection along the laminae and diffusion across the laminae. We use the length scale to estimate the pore fluid flow velocity from the Péclet number, P. The Péclet number is described as
where U is the pore fluid velocity (m/s), L is the critical length scale (m, 1/2 laminae width for a relative permeability channel with two walls), and D is the diffusivity (m2/s, along dewatering pathways as indicated in Figure 8c). The system forms dewatering layers with a spacing perhaps determined by the balance of the advection rate and diffusion rate, i.e. P = 1. The hydraulic diffusivity of the sand is dependent on multiple unconstrained factors including in situ water content, slurry velocity, and pressure gradients. Therefore we refer to experimental debris flows to determine a reasonable estimate for a moving sand slurry. Based on Iverson [1997, Table 8], and assuming that their sand-gravel mixtures were sand dominated (contained high sand concentrations), we estimate the hydraulic diffusivity is on order 10−4 m2/s but allow for an order of magnitude variation. Using the observed average half-width scale of laminae (∼3 mm), we estimate a pore fluid velocity of ∼3 cm/s (permissible range is 3 mm/s–30 cm/s). This is generally consistent with the transport velocity estimates above: in the jammed state, the fluid flow velocity through the porous medium must be significantly less than the emplacement velocity of the fully mobilized flow. Pore fluids traveling at these velocities would exert a flow stress on the grains and potentially rotate those grains that are most loosely held in the granular skeleton, or have the least surrounding grain contacts, imposing a subtle lineation to the preferred conduits. Similar flow velocities (cm/s) are observed to have macroscopic effects on permeability in other natural systems [Brodsky et al., 2003; Manga et al., 2012].
 The laminae are accentuated by iron oxide cement. Two previous studies offered conflicting interpretations for the origins of the iron oxide-cemented laminae:Thompson et al. interpreted them as Liesegang banding (a self-organized cyclic variation in concentrations along a gradient) whileScott et al.  argued that the cement actually followed a previously existing grain structure. Our results support the interpretation of Scott et al. , although our microstructural data indicate that the primary structure highlighted by the iron oxide is not a porosity variation as they suggest, but a more subtle contrast in permeability related to grain orientation and pore shape. We can rule out the Liesegang banding interpretation based on our observations of crosscutting relations. Liesegang rings are precipitated normal to concentration gradients when a precipitation front moves across a homogeneous medium [Dee, 1986]. This is most readily observed where they form parallel to joints or concentrically in joint-bounded blocks, nodules, and amygdules, indicating diffusion gradients normal to surfaces of permeability discontinuities in the rock [McBride, 2003; Eichhubl et al., 2009]. Our observations show that the iron oxide-cemented laminae are not concentric to fractures or other permeability discontinuities in the rock. They end sharply at the boundaries of dolomite-cemented sands and do not cross regions of homogenous sand where primary grain structures forming the laminae have been destroyed (e.g., convolute structures or small faults).
6.2.3. Textures Resulting From Compaction and Dewatering
 After the jamming of the sand grains and formation of the laminae, compaction commenced and the pore water escaped from the injectite. Two classes of features were formed simultaneously by dewatering: 6–13 cm banding, which crosscuts both the dolomite-cemented and laminated sands, and convolute structures, which are observed only in the laminated sand.
 The bands represent a spaced compaction fabric, which is primarily horizontal but undulates, possibly in response to topography on the floor of the injectite. Olsson et al.  showed that compaction bands form perpendicular to the maximum strain direction, locally reducing permeability. Multiple localization fronts can create trapped fluid in higher porosity zones resulting in an increased pore pressure. Olsson et al.  argue that this increased pore pressure will locally decrease the effective mean stress driving compaction and can slow or stop the compaction process. In the Yellow Bank Creek Complex, the increased pore pressure between compaction bands was released by localized fluid venting in pipes which deformed or destroyed laminae, forming the convolute structures. The geometry and spacing of the convolute structures strongly resemble fumarolic pipes formed in compacting ash flows, where material is also well mixed and microstructure destroyed along the upward flow path [Nermoen et al., 2010], as well as pipes, consolidation laminae and burst through structures associated with water escape from saturated sediments.
 In areas of the injectite where compaction bands are weakly expressed or absent, convolute structures grew much taller and are more widely spaced (Figure 5b). These may have facilitated the draining of larger volumes of water. In the southern part of the outcrop, where dolomite-cemented sand forms vertical columns (Figure 3), tall, isolated convolutions are more abundant. This may record vertical motion of the sand during upward emplacement. Toward the northern part of the outcrop, where the subhorizontal compaction bands are strongly developed, the convolutions are short and closely spaced, and lean toward the north, as do the “fingers” of oil sand injected into the wet sand (Figure 3a). This may record a component of subhorizontal flow during injection of the sill, and the continued lateral momentum during initial dewatering and compaction. Convolutions and areas of wiped-out laminae are also found immediately below mudstone clasts. This may indicate that fluid pressure was locally elevated due to the mudstone clast blocking upward fluid escape.