Chalcedony vein horizons and clastic dikes in the White River Group as products of diagenetically driven deformation

Chalcedony vein and clastic dike arrays are well known from numerous localities in the Tertiary White River Group of South Dakota, Nebraska, and Colorado (e.g., Retallack, 1983). However, their origin is not well understood. Questions regarding formation, timing, localization, and geometric patterns are relatively unexplored, yet vein complexes carry significant information about strain patterns and history, associated stress fields, mechanical properties, and fluid-rock interactions. Chalcedony vein and clastic dike spatial distributions, orientations, and character are not easily explained by tectonic forces (Madison and Fischer, 2007; discussed herein). Recent advances in structural diagenesis (e.g., Cartwright and Lonergan, 1996; Cartwright and Dewhurst, 1998; Davies, 2005; Jonk et al., 2005; Laubach et al., 2010), the idea that shallow chemical processes associated with diagenesis can produce and be coupled with significant fracture and fault generation, provide a framework for understanding their formation. We use the term diagenetically driven deformation to focus on the shallow setting, and as a subset of the phenomena of chemically driven deformation. White River strata are known to exhibit a complex and variable diagenetic history (e.g., Lander and Hay, 1993; Loope et al., 2005), in large part due to the unusually volcanic ash–rich character of the sediment. This report describes the character of vein and dike arrays at six study sites within the White River Group (Fig. 1), and then explores how diagenetically driven deformation can explain many of their attributes.

The White River Group consists of a variety of Eocene to Oligocene, terrestrial sediments shed from Laramide uplifts to the west, and there is evidence of a very significant volcanic ash input that has been reworked to varying degrees. Figure 2 provides the stratigraphic context for the veins and dikes. The basal Chamberlin Pass Formation is complex and discontinuous, and it is composed of channel sands, overbank deposits, and paleosol complexes (Terry, 1998). In northwest Nebraska, basal conglomerates contain uranium deposits (e.g., Crow Butte deposit). The more regionally distributed, overlying Chadron Formation is dominated by mudstones, with infrequent, shoestring channel complexes, especially in the upper Big Cottonwood Creek Member (Terry, 1998). The mudstones consist of gray-green horizons intercalated with subordinate horizons having a pinkish cast, with common claystone breccias. Much of the coloration and banding throughout the White River Group is argued by Retallack (1983) to be due to paleosol development.

The upper Brule Formation is characterized by red and tan colors, better-defined bedding, more abundant channel sandstone complexes, loessal silts, and an overall fining-upward trend (e.g., Evanoff et al., 2010). Debate exists about the exact position of the Chadron-Brule contact (see reviews by Retallack, 1983; LaGarry, 1998). Disconformably overlying the Brule in the Badlands National Park area, the Sharps Formation is dominated by channel sandstones, with a basal marker horizon, the Rocky Ford Ash, locally present. In northwestern Nebraska, the Brule is overlain by Oligocene Arikaree Group strata siltstones and fine sandstones. Chalcedony veins, to our knowledge, are only found within the White River Group strata throughout the study area. More detailed stratigraphic descriptions can be found in Terry et al. (1988) and Evanoff et al. (2010).

Clastic dikes within the White River Group are described from the Badlands National Park area of South Dakota, and near Douglas, Wyoming (e.g., Lander and Hay, 1993), but they occur sporadically elsewhere, including a distinctive site in northwest Nebraska described here (Monroe Creek site, Fig. 1). Previous descriptions of the clastic dikes in the Badlands National Park area have been presented by Lawler (1923), Bump (1951), Smith (1952), Retallack (1983), Whelan et al. (1996), Madison and Fischer (2007), and Madison (2010).

White River Group diagenesis, as presently understood, is complex and stratigraphically and laterally variable. Paleosol development under evolving climatic conditions strongly influenced initial clay mineralogy (Retallack, 1993). Smectite and opal cristobolite-tridymite (CT) are common alteration products of the volcaniclastic sediments (Lander and Hay, 1993), and smectite cement is described from Oligocene sandstones of the Scotts Bluff area, as is local folding of playa deposits due to shallow diagenesis (Loope et al., 2005). Pervasive silica mobilization is evident as chalcedony nodules, cement, fossil replacement, and chalcedony veins (e.g., Lander and Hay, 1993). Chalcedony pseudomorphs of barite and gypsum are also present. Carbonate cements and veins are also locally common. Zeolites, specifically clinoptolite, are locally a significant diagenetic product (Lander and Hay, 1993). The complex and variable diagenetic signature for the White River Group can be attributed to shallow conditions, fluctuating groundwater table positions and groundwater fluxes, and the peculiarly volcanic ash–rich nature of the sediment (Lander and Hay, 1993).

Veins are subvertical, occur in distinct stratabound arrays of varying areal extent, and display internal zonation, tip curls, and an array of deformational features (Fig. 3). Their stratabound extent and deformational features are described in more detail in the next sections. Individual veins vary in width from millimeters up to 6 cm, and in horizontal length from a meter to tens of meters. Horizontal extents can be similar to their vertical extent, but can also be ten or more times greater. In the denser arrays, vein spacing is typically in the range of 1–2 m, and widths are 3–4 cm, and therefore veins can represent up to 2%–3% of the host horizon volume. Longer veins are often segmented, with overlap and tip curl linkages. The chalcedony veins show a mode-I dilational geometry with parallel walls, tapered tips, and septa. Internal zonation and textures are consistent within an array but can vary significantly from array to array. A common trait is a gradation from darker- to lighter-colored chalcedony from the exterior to the interior. Calcite is a medial phase in many localities (Fig. 3C), but it is also observed as bands along the vein margin. Medial vugs commonly occur, sometimes lined with crystals (quartz or calcite as coarse yellow spar) and other times with colloform chalce-dony. Coarse agate-like banding exists in some vein fields. The banding and zoning vary from vein-wall symmetric and laterally continuous to very asymmetric and laterally discontinuous.

Veins vary from being planar to exhibiting strong tip curls (Fig. 3D). The tip curls are consistent with low horizontal deviatoric stress levels (Olson and Pollard, 1989; Renshaw and Pollard, 1994), and they have vertical axes of curvature, consistent with horizontal propagation directions. Smaller, en echelon vein segments with strong tip curls can sometimes be seen to connect with a continuous planar vein when traced vertically down, consistent with a propagation arrest geometry.

The chalcedony vein arrays have been described as polygonal (e.g., Lawler, 1923; Evanoff et al., 2010). However, triple junction boundaries characteristic of a honeycomb pattern are very rare in the chalcedony vein arrays, and instead orthogonal vein junctions or approaches are the general rule (e.g., Figs. 3A and 3E). Lachenbruch (1962) distinguished between orthogonal and nonorthogonal fracture patterns associated with volumetric decrease, and Aydin and DeGraff (1988) termed the former tetragonal. Some chalcedony vein arrays may be thought of as having a mosaic pattern, with tetragonal subareas having different orientations (Fig. 4A; Hoff et al., 2007). At Imlay (Fig. 1), the subareas are generally tens of square meters in area. In contrast, at Toadstool, a given array is well organized as one coherent, orthogonal set (Fig. 4).

T-junctions (Fig. 3E) observed with a common fill and symmetric zoning indicate that different directions grew simultaneously. Oblique, smaller veins typically curve to approach a larger vein at right angles, tipping out within a few centimeters of the vein, suggesting that a preexisting larger vein was a barrier to propagation. Such geometries occur for differently oriented veins in a manner consistent with coeval development of the different vein directions. Crosscutting relationships have not been unequivocally observed despite ample opportunity, suggesting that an existing vein was a barrier to tip propagation. Wall geometries vary considerably from relatively smooth and ornamented with linear features (Fig. 3F), to very rough, with colloform chalcedony protuberances. It is fairly common for veins to have greenish alteration zones that extend 1–3 cm into the host rock, especially where the host rocks have a brownish or reddish cast.

A distinctive trait of the chalcedony vein arrays is their stratabound distribution centered on mud-rich horizons. Any explanation for chalcedony vein formation needs to explain this fundamental trait. Multiple beds are typically transgressed by a given array.

Mudstones and/or siltstones at the core of the array are typically greenish to dark brownish in color. The chalcedony vein horizons do not tend to be centered on fine-grained layers with a reddish coloration, although they can transgress reddish mudstones and even sandstones at the upper or lower bounds of their stratabound distribution. For different arrays, the host stratigraphic interval can vary considerably in vertical extent from slightly less than a meter up to greater than 30 m.

Near Rock Bass Reservoir (Fig. 1), to the east of Toadstool Geologic Park, a distinct array of chalcedony veins found within the Chadron Formation is only 80 cm in vertical extent (Fig. 3B) and centered on a distinctive green mudstone found in between two more distinctly reddish mudstones. The boundaries between the red and green mudstones layers are not sharp, but they are transitional and mottled. Vein tips extend 10–20 cm into the adjacent reddish mudstones. The veins extend horizontally for several tens of meters. This clear association of the chalcedony veins with one bed is unique to this study site. Chalcedony veins occur in a similar stratigraphic position within the Big Cottonwood Creek Member, nearby at Toadstool Geologic Park (Figs. 1 and 4), but with a vertical extent in excess of 3 m and traversing multiple beds. Chalcedony vein horizons are unknown higher in the stratigraphy at Toadstool Geologic Park, although chalcedony vein material does occur to a distinctly limited extent along fault planes and as isolated veins higher in the section (lower Whitney Member levels).

In distinct contrast to northwest Nebraska, stratiform chalcedony vein arrays occur minimally at five different stratigraphic levels (Fig. 2) in the general Sheep Mountain Table area of Badlands National Park (Fig. 1). The lowest occurs within greenish and light-colored mudstones of the Chadron Formation (Fig. 1, site 8). A very laterally extensive array occurs at the Chadron-Brule Formation contact, and it was studied at the Imlay, Chamberlin Pass, and Sheep Mountain entrance sites (Fig. 1, sites 3–5). At Imlay, the vein horizon is 8 m thick, while at Chamberlin Pass, it varies from several meters to 10 m thick. This vein horizon locally terminates where channel complexes of the basal Brule Formation locally cut down into underlying mudstones.

A third array occurs roughly at the stratigraphic level of the Saddle Pass marker (Evanoff et al., 2010), and it is ∼7 m thick, while another array, ∼3–4 m thick, occurs centered on the Heck Table marker (Evanoff et al., 2010). Both markers within the Scenic Member of the Brule Formation are based on the location of distinctive paleosols. The highest array occurs roughly 20 m (visual estimate) stratigraphically above the Heck Table marker, within the Poleslide Member (site 7, Figs. 1 and 2). Average vein width distinctly decreases with stratigraphic ascent from level to level. In the lower two levels, veins are typically 2–4 cm thick. The chalcedony veins within the highest most stratabound unit are only several millimeters in thickness, with a vertical extent of 1–2 m.

The areal extent of the vein arrays varies considerably. Vein density and thickness decrease at array margins. In Toadstool Geologic Park, arrays are typically several hundred to several thousand square meters in areal extent (Fig. 4). Here, the chalcedony veins formed contemporaneously with abundant normal faults, which influenced their areal distribution (Moak et al., 2004; Fig. 4A). Fault tips are one site where chalcedony vein arrays are concentrated. Other arrays truncate against a fault. In the Sheep Mountain area, the chalcedony vein horizon along the Chadron-Brule contact is minimally several tens of square kilometers in areal extent. Faulting is absent in the Sheep Mountain area.

Three different deformation features are commonly found in the chalcedony veins: (1) small faults with well-developed calcite slickensides that overall traverse the vein at a low angle; (2) buckle and fault-bend folds of the veins; and (3) striated walls (Fig. 3). These features are concentrated toward the middle of the chalcedony vein array where veins are thickest. In addition, stratigraphically lower chalcedony horizons exhibit better-developed and more frequent small-scale faults and folds than higher horizons. These features are found whether or not large-scale stratal faulting is present.

The faults uniformly have vertical slickensides and/or striae with consistent hanging-wall down motion. Given the veins’ vertical orientation, these are normal faults that produced vein-parallel vertical shortening. Offset is typically several centimeters, and up to 10 cm. Small fault-bend folds occur where the fault traverses the vein, and then bends parallel to the vein wall. The fault wall toward the host side of the vein has a cutoff relation to the fault, while toward the interior, there is a flat relationship. The crest of the resultant fold can be angular, with smaller chalcedony veins occupying the broken crest within the deformed larger vein. In other cases of small fault-propagation geometries, the associated fold is quite rounded. This range of forms suggests that the veins had a range of ductilities during deformation. There is no evidence that faults propagate out into the country rock, although there is a lack of well-defined markers in the adjacent massive siltstones and mudstones.

The vein folds consistently have subhorizontal axes. Layer thickness is typically maintained (Fig. 3G). Thin sections display distinct chalcedony microveins within folded veins in an extrados-type position (concentrated along the outer fold arc and with a vein intersection subparallel to the fold axis). Thus, distributed microveining has accommodated a significant portion of the folding, with multiple generations of chalcedony involved (Riggle, 2006). The folds are also consistent with vertical vein shortening.

Vein walls show a substantial variety of surface types and linear features that vary from smooth and striated, to corrugated with an amplitude of several millimeters, to rough with colloform growth protrusions with a relief of around a centimeter, to an unlineated planar geometry with relief less than a millimeter. Where linear features are present, they are uniformly subvertical in orientation. At the Indian Creek Road site (Fig. 1), well-developed chalcedony slickensides on opposing vein margins both clearly show host-rock down motion (Fig. 3F).

While these deformational features must have formed during or after vein formation, the involvement of chalcedony in the slickensides and microveins indicates deformation was concurrent with silica mobilization. The range of fold forms seen in the same vein array suggests that deformation occurred during a time of changing mechanical properties, which is most easily explained by silica phase changes. The timing and significance of these vertical shortening features are explored in the discussion.

Veins are consistently subvertical, and thus the focus here is on their strike distribution and map pattern. Figure 5 shows strike orientation plots from six locations to demonstrate the considerable variability in patterns seen. Ray plots of the number of strike measurements within 15° to either side of a ray orientation (encompassing a 30° sector) were plotted for 1° increments from 0° to 360°. Thirty degrees was chosen as a sector size that should capture a high percentage (>90%) of a typical unimodal (preferred orientation) fracture population. These diagrams can be qualitatively interpreted in a manner similar to standard rose diagrams, but they help to avoid the pitfalls of bin position selection associated with typical rose diagrams. They also permit more detailed modeling of multiple components (see following). In addition, chi square tests for vein strike uniformity (with 18 ten-degree bins, using Excel’s™ ChiTest function) were conducted for each site in Figure 5 (Table 1). The default hypothesis, that there is a uniform distribution, can be rejected with greater confidence as these values approach 0. The threshold value for rejecting the hypothesis that they are uniform was if ChiTest values were <0.05. An overall uniform distribution is consistent with an isotropic horizontal stress state, as would be expected during syneresis unaccompanied by additional horizontal stress components.

At one end of the spectrum of strike distributions, there is the Toadstool Geologic Park site. Separate arrays, all at roughly the same stratigraphic level but separated by intervening barren areas, are included in the plot. Each array displays a well-developed preferred orientation, although the orientation can differ between arrays (Fig. 5). Within an array, two orthogonal sets are typical, with one set dominating. The dominant preferred direction of chalcedony veins can be both parallel and oblique to the major preferred directions of the faults (Fig. 5; Moak et al., 2004). The very small ChiTest value attests to the well-developed preferred orientations and the obvious departure from a uniform distribution.

Nearby at the Rock Bass site (Fig. 1), one preferred direction exists (110°; Fig. 5), which is parallel to local faulting, and to one of the primary vein directions at Toadstool. However, the pattern is fundamentally different than at Toadstool, and a significant portion of the vein population cannot reasonably be assigned to any preferred direction. This is partly reflected in the chi square value of 0.07, which technically leads to the acceptance of the null hypothesis that the vein distribution is uniform despite a relatively large n. Another important difference is that the faulting at the Rock Bass site clearly postdates chalcedony vein formation, as indicated by truncation and offset geometries and by brecciation of the chalcedony vein material within fault zones.

The pattern seen at Imlay is at the other end of the spectrum and provides a stark contrast to that at Toadstool. Faulting is absent at Imlay, and the chalcedony vein strike plot from the area does not display a strong preferred direction. The ChiTest value leads to the acceptance of the hypothesis that the distribution is uniform/random despite the large number of observations (n = 1037). Nearer the middle of the spectrum, the pattern for the chalcedony veins near the entrance road to Sheep Mountain and Chamberlin Pass (Fig. 5), also positioned near the Brule-Chadron contact, shows distinct preferred directions, and the ChiTest values support the contention that the distribution is nonuniform. However, because bin positioning can favor acceptance of the null hypothesis, the ChiTest values must be used with caution.

To further inform interpretation, the distribution patterns evident in the ray plots were modeled as a combination of preferred orientation components (up to four) each with a normal distribution, plus a uniform distribution component (Fig. 5). Input parameters were the population mean and standard deviation for the preferred orientation components, along with the contribution percentage for each component, including a uniform component. Reiterative estimates of input parameters, while minimizing the average absolute difference between the model and observed ray values, were used to arrive at the best model. These model results are very sensitive to the choice of mean.

Constraining the standard deviation of a preferred orientation helps to better constrain the magnitude of any uniform component and avoid nonunique solutions. When looking at single joint sets from a variety of settings, Engelder and Delteil (2004) found standard deviations of 1.7°–8.6°. However, the joint distribution could reasonably be expected to have a different form for the chalcedony veins. The Toadstool Geologic Park site can be used to loosely calibrate the standard deviation, since its well-developed preferred directions could be modeled without a uniform contribution. The result is standard deviation model values between 8° and 12°. Thus, initial input values for the standard deviation are 10°. More than 15° is considered unreasonable.

Such modeling allows us to crudely explore the question of the extent to which a uniform component contributed to the observed strike distributions, providing a crude measure of the degree of organization of the composite strike distribution. Model input and results are depicted in Figure 5, along with model parameters. Typically, an average data versus model difference of less than 10% could be obtained.

The distribution modeling helps to inform and constrain the visual interpretation and further demonstrates that the vein arrays vary considerably by site—from dominantly uniform (Imlay), to a mixture of uniform and oriented components (e.g., Sheep Mountain Table), to being dominated by well-developed preferred directions with almost no uniform component (Toadstool Geologic Park). However, all but the Toadstool Geologic Park site require a significant uniform component. The strike distributions can thus be thought of as being organized to different degrees and having different mixtures of a uniform component and one to four coeval preferred directions.

Clastic dikes are common in and well known from the Badlands National Park area. Here, they are most numerous and thickest in the Sharps Formation and taper downward, with many terminating in the Poleslide Member of the Brule Formation, but with some continuing down to the Brule-Chadron Formation contact. Places in Badlands National Park where they are more common (e.g., Sheep Mountain Table, Sage Creek basin, Cedar Pass) are also places where younger strata are exposed, and since the dikes taper downward, the clastic dikes may have originally been much more widely distributed but have since been removed by erosion. Clastic dikes are not generally described from equivalent strata in northwest Nebraska. However, they are also well developed at the Monroe Creek site in Nebraska (Fig. 1), where they cut strata of the upper Chadron and lower Brule Formations.

Some clastic dikes exceed a kilometer in strike length, indicating that, like the chalcedony veins, their horizontal extent can be several times their vertical extent. Longer dikes can show distinct curvatures. Unlike the chalcedony veins, clastic dikes often crosscut across each other. Individual dikes contain vertical layers of both mud and sand, with an aggregate thickness from millimeters up to 40 cm thick (Fig. 6). Offsets of oblique crosscutting dikes are consistent with a dilational (mode I) fracturing. Splays are common, and the fill from one dike can become an additional layer within the fill of the other. Splaying tends to be concentrated in the sandier host units (such as the persistent white layer [Evanoff et al., 2010] in the Cedar Pass area; Fig. 6A) and is less common in the more mud-rich units below. Some dikes show significant alteration zones of green discoloration in adjacent host rocks, while adjacent dikes do not.

The dikes show a complex array of internal structures. These include dike-parallel lamination and lithologic layering, internal cross stratification, lenses of sand within mud, soft-sediment deformation, thin gypsum veins, crosscutting relationships between internal layers, grading, septa, and “xenoliths” (Fig. 6). A common and striking structure is the internal vertical banding, with dikes often consisting of multiple layers of sand, silt, and/or mud material. Internal layer continuity varies from visible lenses to a minimum strike continuity of tens of meters. Sand layers often show a well-developed internal lamination parallel to the dike walls, defined by the preferred orientation of inequant grains, particularly muscovite. Observed geometries where an oblique dike cuts some layers of a dike and then turns into parallelism to form an internal layer clearly indicate the occurrence of multiple and distinct fill events. Congruent fill of orthogonal dikes is observed. Angular clasts of mudstone millimeters to centimeters in size are common.

One particular distinctive feature found in some of the coarser and thicker sandstone dikes is upward-pointing, V-shaped, cross laminations that are ornamented by well-developed flute marks that taper upward (Figs. 6B and 6C). The flute marks often give a surface appearance similar to an oscillation ripple in cross section, with the pointed crest consistently pointing outward on either side of the dike. Traced downward, the flutes flatten out into a planar lamination. Commonly, these dikes are grain-size graded, and the cross laminae on the margins give way to a more massive and coarser interior, which can have large mudstone clasts in it. While this collection of features usually occupies the entirety of the dike, they also can occur as a distinct layer within a composite dike.

Within individual sand layers, cross laminae also occur. Laminae are mostly planar and do not curve toward parallelism with the layer boundary. Folding of the lamination within sandy layers also occurs, typically at locations where a dike splays or has an irregular margin geometry. Plunges vary, and the irregular fold style is consistent with soft-sediment deformation.

Dikes show several features indicating postemplacement modification, including evidence of subtle folding (Fig. 6A). Dike walls and internal structures are coherently folded, indicating that this deformation occurred after dike formation. The axes are typically subhorizontal. In addition, in some instances, dike walls do not match in detail, as would be expected with an unmodified brittle fracture. Instead, the walls display undulations with wavelengths of centimeters to tens of centimeters. Mismatches could be due to either wall-rock erosion during fill emplacement (consistent with observed clasts), or to postemplacement loading. In some cases, in cross section, these are cuspate and reminiscent of loading structures. Some smaller dike splays in the persistent white layer (Evanoff et al., 2010) in the Cedar Pass area are discontinuous, suggesting that, locally, the sediment fill has been squeezed out. This suggests the possibility that sediment fill can migrate along a fracture segment that closes in the lagging portion.

Distinct nodules are common at multiple levels within the Brule Formation, reflecting localized cementation and diagenesis. Dikes often bend around these nodules, or locally branch within the nodules, indicating nodule formation occurred before dike emplacement. Thus, dike formation is bracketed by earlier nodule formation, and concurrent or later compaction.

The dike strike distribution is notably variable from locality to locality, both in terms of the orientation of preferred directions, and the degree to which a preferred orientation is developed. We describe the pattern from three different areas to demonstrate the variability—Cedar Pass and Sage Creek sites in Badlands National Park, and the Monroe Creek site in northwest Nebraska.

In the Cedar Pass area (Fig. 1), the dike traces are complex. Locally, two orthogonal sets are well defined, with the dominant one trending 120°–130° (Fig. 7), subparallel to local normal faults and monoclinal structures found just to the southwest (Raymond and King, 1976; Evanoff et al., 2010). However, a significant portion of the distribution cannot be explained by these two sets. Members of the orthogonal set cut each other, or share common fill layers, indicating that they opened together. However, members of the dominant set (azimuth 120°–130°) to the south, where they are in closer proximity to the faults, more commonly display green alteration zones. This same subset of dikes typically cuts and therefore postdates other dikes, suggesting a late phase of more oriented dike formation that tapped into reducing fluids. One clastic dike was observed to cut a fault.

In the Sage Creek area, the pattern is different and more complex than at Cedar Pass. Madison (2010, p. 18), in a careful analysis, concluded that length-weighted data for dikes in this area “depict a very well-defined trend at 160 degrees and a relatively less well-defined sub-perpendicular trend.” However, for mapped dike data, where each individual dike is given even weight, he indicates “a uniform orientation distribution cannot be reliably rejected.” Importantly, a significant percentage of the dikes would not belong to either of the two orthogonal sets. The 160° azimuth preferred orientation at Sage Creek is also at a distinct angle to the Sage Creek Arch and associated normal faults, which trend at 130°–140° and lie ∼1 km to the north.

The clastic dikes at the Monroe Creek site (Fig. 7B) display three preferred orientations (Halligan et al., 2011). The 75° dike trend is roughly the same as that of the Toadstool fault, and of the Colorado Lineament (Warner, 1978), and the 121° trend is subparallel to a clastic dike direction seen at Cedar Pass and the fault trend seen at Rock Bass. However, the exposure-constrained study area is not considered a large enough sampling to identify if one direction is dominant, and its shape is likely biasing the distribution. The distribution model shown (Fig. 7) would likely change significantly if data from a larger area were gathered. Identical fill and overlapping crosscutting relationships indicate that the various directions formed contemporaneously. In some cases, faulting is localized along the dikes, and parallels the dikes, but field relationships here consistently indicate that faulting postdated dike emplacement. Some of the dikes show distinct curvature, up to 60° of arc, similar to that seen at the Sage Creek site. One dike pinches out upward, and overlying younger Arikaree Group strata just south of the study site are barren of dikes.

Clastic dikes and chalcedony veins can occur independently, but locally they interact in a variety of ways. At the Sheep Mountain site, chalcedony veins occur within the clastic dikes. Proximal veins tend to be parallel and/or perpendicular to the dike, indicating either formation in a similar stress field, or mechanical influence of the dike on chalcedony vein orientation. Perpendicular chalcedony veins do not cut dikes but instead tip out with approach to the dike, indicating the dike was a preexisting barrier to vein propagation, in a similar manner as seen with other chalcedony veins. Small irregular patches of mobilized sediments occur in some chalcedony veins, and discontinuous patches of chalcedony occur within clastic dikes (Fig. 3I).

In thin section, clastic fill and chalcedony vein material show microstructures consistent with deformation of a silica gel within a mobile, unlithified sediment (Fig. 8). Elongate blebs of chalcedony material are aligned with the preferred orientation of grains, and irregular and cuspate chalcedony-sediment boundaries are typical. Late-stage chalcedony planar microveins parallel to the overall dike cut through deformed chalcedony and sedimentary grains (Fig. 8E), indicating continued silica mobilization after or during lithification.

Most clastic dikes at the Monroe Creek site have a composite chalcedony and mud fill. The chalcedony is often quite irregular in form, makes up to half of the dike fill, and can appear brecciated within the mud fill matrix. Chalcedony distribution along the features is discontinuous and is preferentially associated with the intersection of the dike with greenish, mud-rich horizons, suggesting local derivation of the silica. The typical chalcedony vein array seen at other sites are absent here, but significant silica mobilization and chalcedony formation were a part of clastic dike emplacement at the Monroe Creek site.

Perhaps the most attention to the origin of the White River Group chalcedony veins was given by Lawler (1923, p. 172), who stressed the role of compaction (“settling”) in vein and dike formation, suggesting that, “Such settling …. gave rise to differential consolidation which in turn produced local torsion.” The settling was ascribed to a period of desiccation and water table lowering associated with arid Miocene conditions that produced “widespread settling and squeezing out of water.” We postulate that diagenetically driven deformation, and in specific, syneresis, played a role in their development.

Features such as tip curls, wall-perpendicular mineral orientation, dilational internal microveins, and vertical orientations are consistent with mode 1 fractures. Such vein arrays could reflect either overall horizontal layer extension or shrinkage. In the latter case the vein extension is matched by shrinkage and horizontal shortening of the host material. Considering the horizontal extension possibility, the implied strain, given up to 2%–3% host material, is significant and greater than that typically associated with simple jointing. A primary consideration is that this deformation is preferentially associated with finer-grained layers. The source mechanism of the silica in the veins is consistently described as devitrification of the volcanic glass in the adjacent sediments (Lander and Hay, 1993; Retallack, 1983). Given that chalcedony veins occur within clastic dikes that extend higher into the Sharps Formation, the burial depth was greater than 150 m. Lander and Hay (1993) cited the maximum burial depth of White River Group strata in northwest Nebraska as 800 m. The veins also are constrained to have formed during or after Sharps Formation deposition. The following discussion explores how the stratabound geometry, the vein strike distributions, the vertical shortening, and the known diagenetic history support a syneresis model for vein formation.

A key consideration is the stratabound character of the vein systems. Fractures can often be stratabound simply because of horizontal barriers to fracture propagation, such as bedding planes, a typical situation with joints in well-layered sedimentary rocks (Pollard and Aydin, 1990). This is clearly not the case here, where chalcedony veins tip out and transgress multiple beds, and bedding planes are often poorly defined or nonexistent. Veins and fractures can also be stratabound within a more competent layer confined within more incompetent material that accommodates the strain in a different manner. In such a case, layers above and below any given vein array would be expected to have features that accommodated the same or greater horizontal strain, but accommodated it by a different mechanism (e.g., ductile flowage). Evidence for such strain in the adjacent layers is absent. This is despite the proximity, in some cases, of sandstone layers that could be expected to behave more competently and preferentially display veins. In addition, other adjacent strata of the same basic lithologic and mechanical character do not display the veining. Fundamentally, the field relationships of the vein arrays are inconsistent with localization of regional strain in more competent layers.

At a larger scale, there is also a lack of correspondence between vein array development sites and sites with tectonism. Both Toadstool Geologic Park and the Sage Creek area in Badlands National Park have larger-scale faulting and folding of White River Group strata, but the former displays chalcedony veins and the latter does not, while the Sheep Mountain site, where faulting and folding are absent, has very well-developed vein arrays. This indicates an independence between vein array occurence and tectonism.

Syneresis concentrated in selective mud-rich layers based on their clay mineralogy would produce a stratabound character, with a lack of corresponding strain above and below. A corresponding distributed contraction (shrinkage) of the host mudstones would counterbalance the vein strain, with no net, finite, horizontal strain across the array.

A simple initial expectation is that syneresis alone would produce a polygonal pattern with an overall uniform and/or random strike distribution. Organized, tetragonal subareas may exist, but collectively they would have a uniform and/or random strike distribution. The Imlay site displays this character, but it is at one end of a spectrum of sites (Fig. 5) that varies from unorganized to highly organized strike distributions (e.g., Toadstool). The question arises: Why do some sites have aggregate preferred orientations if syneresis is the driving force? Next, we explore how the observed variability in strike distributions can be explained by variations in the ambient stress fields in which the syneresis occurs.

Three possibilities considered are that: (1) all the strain was due to syneresis, and any differential horizontal principal stress serves to organize the veins into multiple directions that in aggregate accommodate the shrinkage; or (2) concurrent tectonic strain components existed while syneresis occurred; or (3) that material horizontal anisotropy organized the syneresis-driven fractures. Olson et al. (2009) described the way in which small variations in the relative magnitudes of horizontal strains and/or anisotropies can influence fracture patterns. A reasonable mechanism that would produce a consistent horizontal material anisotropy in the fine-grained host rocks on the vertical (tens of meters and multiple beds involved) and horizontal (thousands of square meters) scales over which the preserved chalcedony vein orientations occur is unknown to us. Arguments explored next suggest that the first two possibilities are in play.

Figure 9 explores the traits of combinations of syneresis, an organizing stress field, and an additional regional strain. The occurrence of coeval multiple preferred directions without a dominant set is considered significant. A dominant preferred direction in the distribution would suggest that a unidirectional, tectonic strain component exists, since syneresis alone should be characterized by overall homogeneous, horizontal shrinkage. Multiple coeval preferred directions of equivalent expression, with or without a significant uniform component, would be consistent with syneresis. Uniform horizontal tectonic extension could theoretically also produce multiple, coeval preferred directions, but it would affect the entire rock body and would be a special case (nonuniform stretching being the more general case).The larger a tectonic/structural strain component, the greater is the expectation that corresponding strain and structures should be expressed in some manner in strata above and below the chalcedony array.

In the Toadstool area, where the veins are best organized, the concurrent normal faults indicate that there was an additional tectonic component, likely related to reactivation along the Colorado Lineament (Warner, 1978). Within an individual array, one set can dominate (Fig. 9), consistent with a local bulk rock strain. At other sites, faulting, or other evidence for significant strain in non–chalcedony-bearing intervals, is missing, and the strike distributions are less organized. The one site where only one preferred direction was modeled (Imlay) is also the location where the uniform model component strongly dominates, and forming the conclusion that the distribution is nonuniform is statistically questionable. Multiple, coeval preferred directions without a dominant set observed at other sites (Fig. 5) are consistent with syneresis-related contraction organized by differential horizontal stresses.

The contrast between the Toadstool and Rock Bass sites in specific supports the idea that local tectonic stresses served to organize the veins. At Toadstool, the distribution is described by well-developed, syn-faulting preferred orientations without a significant uniform component. At nearby Rock Bass, the associated faulting postdates chalcedony vein formation, one chalcedony vein preferred direction parallels that in the faults, and modeling indicates a much larger uniform component in the strike distribution (Fig. 5). We interpret the Rock Bass site as an example where veins formed without contemporary faulting, but where related subcritical stresses influenced the strike distribution to some degree. Toadstool is taken as an example where additional tectonic strains produced complete organization of the strike distribution into preferred directions, with one preferred direction dominating locally.

In summary, the large uniform components seen in the models of the strike distributions of many sites, and the coeval character of multiple preferred orientations are consistent with a syneresis origin. Site-varying tectonic stresses and/or strains can explain partial organization of the veins, with the sites showing a continuum from Imlay at one end, with its largely unorganized distribution, to Toadstool at the other end, with its largely organized distribution and multiple directions developed to different degrees

The veins show ubiquitous evidence for significant vertical shortening, as Lawler (1923) noted. Several arguments indicate that vertical shortening was roughly contemporaneous with vein development. The significant variation from ductile to brittle style can be best explained by deformation during a silica gel to chalcedony transformation, with the attendant increase in strength of the chalcedony veins and change in deformation mechanics. The folds show abundant microveins of chalcedony at the extrados position within the chalcedony vein, indicating deformation occurred during silica mobilization (Riggle, 2006). Slickensides associated with the faults are composed of calcite material, which is also the medial phase of many chalcedony veins.

The vertical linear features on vein walls can also be explained by settling of the adjacent sediment relative to the vein. Marginal slickensides seen at the Indian Creek locality (Fig. 1) are composed of chalcedony material identical to that in the vein interior, again consistent with formation during silica mobilization. The degree to which the vein was mechanically coupled or decoupled to the adjacent wall rock likely determined whether it folded or whether the vein surfaces became slip horizons and became adorned with linear features. Linear features occur preferentially on the smoother vein sides, suggesting surface roughness as a factor. The simplest explanation of these structures is that the surrounding layer compacted during diagenesis, dewatering, and related silica mobilization, causing recently formed chalcedony veins to deform and shorten. The vein system may have aided fluid escape, which in turn aided compaction.

A pattern of localized stratabound horizontal extension and vertical shortening is also associated with polygonal normal fault systems that are seen in fine-grained marine strata, and are well documented in numerous three-dimensional seismic surveys. These are attributed to syneresis of smectite-rich, colloidal marine mudstones and associated fluid-expulsion–related compaction (e.g., Cartwright and Lonergan, 1996; Dewhurst et al., 1999; Gay et al., 2004). The associated normal faults have throws from 10 to 100 m and spacing of 100–1000 m (Cartwright and Dewhurst, 1998). While on a much smaller scale, the chalcedony vein systems in the White River Group can be considered as an analogous example of the general phenomena of syneresis-related, diagenetically driven deformation. Both the veins and polygonal faults fall in the realm of structural diagenesis as described by Laubach et al. (2010). Tensile instead of shear fractures were generated in this terrestrial version, perhaps due to higher fluid pressures and/or a shallower level of development.

Mechanisms for syneresis described in the literature include smectite to illite transformation (e.g., Cartwright and Dewhurst, 1998) and opal A to opal CT conversion (e.g., Davies et al., 2006). In addition, porosity collapse and compaction due to opal cement dissolution were described by Spinelli et al. (2007). Smectite diagenesis is a prime candidate for producing syneresis in the case of the chalcedony veins. Smectite and/or smectite-illite mixed-layer clays (smectite dominated) are a significant portion of the clay size fraction of the White River Group (Retallack, 1983). X-ray diffraction (XRD) analysis has identified a complex clay signature in sample suites that traverse chalcedony vein horizons (Nicholson, 2010), which includes up to 90% (of clay fraction) smectite, up to 80% mixed-layer smectite-illite (>90% smectite layers), up to 40% illite, and up to 5% clinoptolite. Initial results suggest that increased illite or smectite-illite amounts are associated with the chalcedony veins. While the specific mechanism awaits further work, it is clear that these sediments have mineralogies with significant syneresis potential.

The smectite to illite transformation is traditionally considered to occur at deeper depths and temperatures (e.g., Pollastro, 1993) than are possible for the chalcedony veins. However, the smectite to illite transformation is notably complex (e.g., Morton, 1985), and smectite cation (e.g., Na⁺ and Ca⁺²) exchanges could theoretically also produce spacing and volume changes. Mixed smectite-illite is typically thought to be diagenetic in origin, reflecting a partial transformation (Morton, 1985). Given that reworked volcanic ash is a dominant component in the fine-grained White River strata, the observed mixed smectite-illite is not inherited but is due to Tertiary diagenesis of the ash. Clinoptolite also has significant cation exchange capacity and associated volume changes, although, to our knowledge, it has not been identified in the literature as driving natural syneresis. Triggers for syneresis could include changes in pore-water chemistry, in addition to changes in pressure, temperature, and water content with burial or exhumation. Not only the presence, but the amount of smectite may be a crucial determinant in whether syneresis triggers vein formation, as the stresses generated by shrinkage need to be large enough to overcome the tensile strength.

Given the shallow level of development, could the veins reflect drying and desiccation in the vadose zone as suggested by Lawler (1923)? Several arguments indicate that the chalcedony veins occurred under saturated phreatic conditions. First and foremost, there is the need for a medium for the substantial silica transport involved in vein formation. Second, the green alteration zones are seen associated with many veins that suggest a reducing environment, which would be more consistent with saturated, phreatic conditions. Finally, without pore pressures to help prop them open, mechanically it would be difficult to form the equivalent of stratiform mud cracks at depths of 150 m or more.

Multiple working hypotheses for clastic dike formation include: tectonic strain (e.g., Winslow, 1983), fluid pressure (injectites; e.g., Boehm and Moore 2002), seismically induced (e.g., Madison and Fischer, 2007) large desiccation cracks that reach the surface (e.g., Retallack, 1983), and structural diagenesis (e.g., Davies et al., 2006; this paper). These are not all mutually exclusive (e.g., Winslow, 1983), but two conflicting end members are desiccation-related cracks with a surface sediment source and an injectite origin with fluidized subsurface sources. We propose that the following attributes are consistent with a structural diagenesis origin for the Badlands clastic dikes: (1) complex strike distributions and simultaneous, multiple directions of opening, (2) significant dike development unassociated with tectonic features, (3) internal structures indicating fluidized sediment injection, and (4) evidence for recurrent opening and sediment fill. Each of these is discussed in turn.

The dike strike distributions at Cedar Pass, Sage Creek, and Monroe Creek all show multiple coeval preferred orientations (Fig. 7), but with a significant spread. For Cedar Pass, it is not possible to model the strike distribution without a significant uniform component. One test of the Sage Creek distribution is close to uniform (Madison, 2010). Detailed modeling could not explain the dikes as being seismically induced by earthquakes on faults that are ∼1 km to the north (Madison and Fischer, 2007). At the Cedar Pass and Monroe Creek sites, one of the clastic dike preferred directions does parallel nearby normal faults. Individual dikes that show substantial curvature in map view are also inconsistent with a significant regional tectonic stress field. In a similar manner to the chalcedony veins, the strike distribution is consistent with syneresis organized to varying degrees by a local existing stress field that at some sites also produced faulting. Dikes also occur in the Sheep Mountain and Cuny Table areas, but without associated faulting or folding, suggesting general independence of dike formation and faulting and/or folding.

The vertical lamination defined by the preferred orientation of grains (Fig. 8F), well-developed interior grading in coarser dikes, the entrainment of large clasts, internal soft-sediment deformation, and cross-laminae are all consistent with the injection of fluidized sediments (e.g., Peterson, 1968). The geometries reasonably expected from infilling by surface sediments washed in from above, such as dish-shaped internal layering, sieve deposits, and vertical zonation, are distinctly lacking. Diagenesis and associated compaction can produce overpressures that contribute to clastic dike formation (Davies et al., 2006). The observed postformation folding of the dikes is consistent with related compaction and dewatering, similar to that inferred for the chalcedony veins.

Uranium mineralization is associated with the dikes (Retallack, 1983). Green alteration zones are also associated with some dikes, consistent with associated reducing fluids. Both are inconsistent with an ostensibly oxidized source from surface sediments and waters. Dikes with green alteration are adjacent to and often cut dikes without alteration zones. This indicates that the alteration was associated with a specific dike formation event, reflecting the geochemistry of the fill at that time, and is not due to a later groundwater flow, which should influence adjacent dikes similarly.

The internal grading and oblique outward-dipping lamination of some coarser dikes (Fig. 6C) are consistent with upward injection. Consideration of the significance of dikes tapering downward in the Badlands National Park and tapering upward at the Monroe Creek site is complicated by the possibility that dikes could close behind themselves, especially vertically and due to the lithostatic gradient. They could be detached from their sources. Considering the much greater horizontal versus vertical extent of many dikes, lateral propagation and injection could be expected, and scenarios can be envisioned where downward injection may be favored (due to either pore pressures or syneresis). The sediment source has not yet been identified. Mud versus sand layers suggests different sources are tapped during the development of a dike.

The geometries associated with the internal banding found in the clastic dikes clearly indicate multiple episodes of opening for individual dikes. The brittle geometries observed (Fig. 6A) indicate that the preexisting fill must have lithified prior to injection. Such recurrent development is consistent with syneresis and diagenetic processes driven by burial and/or groundwater regime changes over a substantial period of time (processes partly constrained by fluid transport speeds).

When taken in aggregate, we believe that the lines of evidence presented here are consistent with an origin of clastic dikes in the White River Group by diagenetically driven deformation, in a way that other models are not. The model of Davies et al. (2006), where an opal A to opal CT transformation produces overpressures that produce sediments “susceptible to remobilization,” could be applicable to the White River Group, where opal and chalcedony cements abound (e.g., Lander and Hay, 1993). Instead of biogenic opal in marine sediments playing the crucial role, in the terrestrial White River case, distributed diagenetic opal from volcanic ash alteration and silica mobilization (Retallack, 1983; Lander and Hay, 1993), or from smectite-illite transformations, could be a primary driver.

For the chalcedony veins, the following factors are consistent with diagenetically driven deformation: their stratabound distribution, localization in finer-grained lithologies with mineral species that could drive syneresis, localities with a significant uniform component to their strike distributions and/or multiple directions of coeval opening, concurrent significant compaction (vein shortening) during silica mobilization, and strong tip curvature indicating development in a stress field with a low horizontal deviatoric stress. The preferred orientations seen at some localities can be interpreted as due to a local anisotropic horizontal stress field partially organizing syneresis-driven fractures. The Toadstool locality may be the exception, where additional tectonic strains existed.

For the clastic dikes, the following factors are also consistent with an origin by structural diagenesis: complex strike distributions and multiple coeval opening directions, a history of incremental formation with intervening fill lithification, the presence of broadly distributed silica and clay phases that could contribute to syneresis, and dike development independent of tectonism. Details of sediment source and injection directions remain to be elucidated.

Development of fracture systems during burial and due to diagenetic reactions has important implications for understanding fluid-flow histories. For the White River Group, the silica mobilization associated with chalcedony vein development, the sediment mobilization associated with clastic dike formation, and uranium mineralization in the White River Group (Zielinski, 1983) may all be related. The distinctive volcanic ash–rich character of the White River Group, and the associated complex diagenetic alterations that composition permitted, may ultimately be responsible for the chalcedony vein arrays and clastic dikes. If so, it is reasonable to look for similar features and history in other ash-rich units.

We would like to thank a long list of University of Nebraska at Omaha undergraduate students (>20) who enthusiastically engaged in related undergraduate research projects over a period of more than a decade. The Petroleum Research Fund of the American Chemical Society provided crucial support (PRF #49405-UR8), as did funding through a National Science Foundation STEP (Science, Technology, Engineering, and Mathematics Talent Expansion Program) grant. We would also like to thank and acknowledge Rachel Benton at Badlands National Park for her help with permitting, and her generous sharing of information, and Mary Ann Holmes for patiently helping our students and us with clay analysis.