Silica diagenesis-driven fracturing in limestone: an example from the Ordovician of Central Pennsylvania

Abstract. Fracture patterns, interactions, and crosscutting relationships are tools for interpretation of fractures as paleostress indicators for past tectonic events and as past or present-day fluid-flow networks. In the Appalachian Basin in Central Pennsylvania along Mount Nittany Expressway Route 322 lies a significantly stratified fracture set hosted in Ordovician age limestone. Tectonic strain is a problematic mechanism for these fractures because they are hosted in individual beds lacking apparent mechanical significance relative to other limestone beds in the outcrop. Many of the fractures are layer-parallel, a characteristic commonly observed in shales, due to shales' mechanical anisotropy and tendency to develop fluid overpressures; however, these fracture-hosting limestones lack obvious mechanical anisotropy. Fracture orientations vary, but desiccation, bentonite swelling, and dolomitization are eliminated by an interpreted transgressional paleoenvironment and a deficiency of the hypothesized minerals. X-ray diffraction determined the composition of samples collected, point-count quantification determined fracture intensity, and optical petrography recorded scaled petrographic photographs. Comparison between fracture intensity and host-rock minerals reveal that silica content is consistently depleted in fractured layers relative to unfractured layers. The diagenetic transition of biogenic silica to quartz is suggested to be the driving mechanism based on silica being present as biogenic grains, as well as cement and detrital grains, and fractures being filled with calcite cement. Silica migration explains the volume lost from fractured layers in a proposed horizontal fracturing mechanism whereby the host rock shrinks but is excluded from vertical contraction.


large faults and having significant shear displacement. As well, these fractures may merge or bifurcate to form closely spaced networks having a cataclastic texture (Fig. 4). 90 A subset of the observed fractures is found to have a highly stratified pattern and, unlike the other fracture measurements, a layer-parallel orientation. Six "fractured layers," F1 through F6, moving downsection, contain abundant layer-parallel fractures and subsidiary layer-perpendicular fractures. F1 through F6 lie within the transition between the Salona and Coburn Formations, whose contact is uncertain and probably gradational within this roadcut. These fractured layers are observed in the southeastern limb of the Mount Nittany Syncline but cannot be directly correlated due to safety concerns within the 95 quarry. Intervening layers, including limestone layers, show only sparse fractures, of any orientation.
In fractured layers 5 and 6, there is a stratiform color change within the limestone beds. The fractured beds of fractured layer 5 ( Fig. 5) are observed as a medium grey limestone that is underlain by a sharp shift to a tan limestone with black and brown clay clasts. The top 2 3 ⁄ of fractured layer 5 are gray and fractured, whilst the bottom 1 3 ⁄ is pale gray with banding.
However, in both fractured layers 5 and 6, the tan limestone then grades downward to a clay layer. 100

Fracture Orientation
The bedding of the outcrop and the poles to the fractures are shown in a stereonet (Fig. 6). Bedding dips gently (15-20°) to the east. Fractures are dominantly layer-parallel, but layer-perpendicular fracturs are common. Within F1 though F6, layerparallel fractures may be present in the absence of layer-perpendicular fractures, but the latter are not encountered without the former. There is no apparent dominant strike orientation for bedding-perpendicular fractures. There is a potential 105 correspondence of these fractures with the J1 and J2 set of Pennsylvania's regional structural geology (Engelder, 2004), however, additional measurements are needed to rigorously determine an association. Layer-parallel and -perpendicular fractures show mutual crosscutting within F1 through F6.

Fracosity results
Results of point counts, taken to quantify fracture volume (fracosity), are shown in table 2. The total area represents the total 110 points within the field picture; the N/A area being the points of neither a fractured or unfractured point of the photo such as dirt, moss, or a shade; the unfractured area represents the unfractured, or host rock, points in the photo. Fracosity ranges from 1.29 to 6.15% within fractured layers with an average of 3.36%. The minimum fracture volume hosted within a fractured layer, 1.29%, is used as a conservative upper limit to for the fracture volume that might be hosted in the unfractured layers, but undetected. 115

XRD results
XRD results indicate that the silica content is consistently depleted in fractured layers F1 through F6 in comparison to the nearby unfractured layers. However, the unfractured tops and bases of the fractured layers have the highest quartz https://doi.org/10.5194/se-2020-50 Preprint. Discussion started: 28 April 2020 c Author(s) 2020. CC BY 4.0 License. composition. Albite does not systematically vary throughout each group, but albite content varies with a range of 0-3.9 wt. % (table 3). The albite composition from each layer is not observed to correlate with quartz ( Fig. 7) such that the quartz to 120 albite ratio is the lowest in the fractured layers (additional mineral compositions in Appendix 1).
The mineral compositions (in wt.%) of quartz and calcite, from the individual fractured and unfractured layers, as measured using XRD, are plotted against fracture volumes in Fig. 8. There is a negative correlation between fracosity and quartz content, and a positive correlation between fracosity and calcite content.
The individual fractured layer thickness displays no obvious correlation with the fracosity percentage for the corresponding 125 layer.
These observations are consistent with field photographs of fractured layers. A field photo for fractured layer 2 is shown in Fig. 9 and for fractured layer 6 in Fig. 10. These photos display the layer-parallel and perpendicular attitude of the fractures, and their mutual crosscutting. The representative fracture volume can be observed in the photos in which layer 2 has significantly more fracosity, mostly manifest in greater apertures, in comparison to layer 6. These observations are quantified 130 in table 2 which displays the fracosity for each fractured layer.

Petrography
Petrography revealed quartz present in at least three forms: detrital, pore filling, and skeletal grains. The thin section observed in Fig. 11 originates from a sample collected from fractured layer 1. Quartz in Fig. 11 appears as a clear mineral lacking cleavage and having gray or white interference color under crossed polars. Here, quartz has a euhedral termination 135 with abundant solid and fluid inclusions, and is surrounded with a carbonate mineral, based on the limestone lithology and the mineral's higher-order birefringence.
In the silicified base layer underneath fractured layer 2 there is a presently silica test (Fig. 12) and in fractured layer 1 there is a dissolved test (Fig. 13); in the silicified base layers, silica is also present as disseminated pore-filling cement. These observations are consistent with the fractured layers being sites previously occupied by silica that has since migrated from 140 these layers and left behind voids which have been replaced by calcite. The result of this migration can be observed in Fig. 9 in which fractured layer 2, hosting 5% quartz, is underlain by a silicified base layer, hosting 15% quartz.
The bulk mineral composition of the base of layer 2 is richer in quartz compared to fractured layer 2, which lies immediately above, and to fractured layer 1. The boundary in thin section for the base of layer 2 and fractured layer 2 can be observed in lacking abundant silica. This change in mineralogy within the fractured layer 2 coincides with a change in color apparent in the outcrop (Fig. 9). The fractured central part of the layer is light gray in color whereas the unfractured base of the layer has a tan, brown color. The same pattern is reflected in fractured layer 6 ( Fig. 10), in which the bulk mineral composition of the base of layer 6 is richer in quartz compared to fractured layer 6 (table 3).
A petrographic photo of fractured layer 1 with a crosscutting fracture is observed in Fig. 15. The fill material is sub-angular 150 and blocky, indication void-filling cement. The crosscutting relationship between the largest two fractures is ambiguous, https://doi.org/10.5194/se-2020-50 Preprint. Discussion started: 28 April 2020 c Author(s) 2020. CC BY 4.0 License.
with a series of cement blocks within the large, horizontal fracture apparently connecting the separated vertical fracture. It may be that the two fractures were open at the same time; the horizontal fracture appears to crosscut smaller vertical fractures. The presence of microfractures (Fig. 14, Fig. 15) suggests that our field-photograph based fracosity Fig.s should be regarded as minimum estimates. 155

Discussion
The layer-perpendicular fractures in this study-apart from those that from as a subsidiary set alongside the layer-parallel fractures-preferentially form within brittle, stiffer layers. Commonly, as layer-parallel extension increases sequential infilling occurs as a result which induces new fractures to form between existing fractures. Fracture spacing is inversely related to the applied layer-parallel strain and expected to diminish as the overburden pressure increases and the tensile layer 160 strength decreases (Schöpfer et al., 2011). Such an explanation is consistent with most layer-perpendicular fractures in this study. However, the layer-extension model is unsatisfactory for the arrays of stratified fractures present in fractured layers one through six, because (i) those fractures are dominantly layer-parallel, and (ii) there is no obvious stiffness contrast between fractured and unfractured limestones, judged by the abundance of other fracture sets in the outcrop, which readily form in any limestones and are only absent from the intervening shales. 165 The first proposed fracture mechanism is the effect of stress and strain in the rock layers experienced during tectonic events, such as folding in the Alleghenian orogeny (Faill, 1997). However, a tectonic interpretation suffers from numerous inconsistencies for the fractured layers because numerous fractures in the outcrop are associated with tectonic faults and are not stratified (Fig. 4) or are present throughout the limestone layers and absent within the shale layers (Fig. 3). This contrasts with catagenesis (Flores, 2014) which could induce volume swelling and produce fracture; though, such fractures would be 170 expected to preferentially form in organic matter preserving shale layers (Ma et al., 2017). The theory of systematic joints in sedimentary rocks predicts that joint abundance is proportional to the inverse of the bed thickness (Hobbs, 1967). However, the fractures are not strongly aligned in any systematic direction with respect to the fold (Fig. 6). The final mechanism considered is diagenesis, because the problems associated with physical explanations point us toward chemical principles.
Based on these fractures' bed-parallel attitude, we look for a source of internal pressuring, which would explain why they 175 apparently open against gravity (Cosgrove, 1995). An increase in fluid pressure could be generated by disequilibrium compaction, hydrocarbon generation, or a combination of the two (Cobbold et al., 2013). Disequilibrium compaction arises where porosities are higher than expected at a given depth and deviate from the standard porosity trend (Zhang, 2013). The process of hydrocarbon generation increases fluid pressure in low permeability shales as kerogen is buried and heated (Ma et al., 2017). In either case, high pore fluid pressure produces overpressure, a state where the fluid pressure is greater than the 180 hydrostatic pressure at a given depth. If the fluid pressure exceeds the overburden stress then horizontal fractures can form (Cobbold et al., 2013). The fractured beds observed in this study are hosted in limestone that lack significant organic https://doi.org/10.5194/se-2020-50 Preprint. Discussion started: 28 April 2020 c Author(s) 2020. CC BY 4.0 License. material or signs of disequilibrium compaction, such as soft-sediment deformation. In the absence of likely candidates for fluid overpressure, it may be that a geochemical reaction forced the fractures to open.
The diagenetic growth of dolomite can also be a primary expansional mechanism (Bellamy, 1977). However, dolomitic beds 185 are a minor component of the studied formation. Tension fractures are known to form as a result of shrinkage within the rock due to contraction of mud or silt sediments through desiccation (Singhal & Gupta, 2010). However, the fractures in this study do not host mudcrack geometries, nor does the depositional environment display evidence of experiencing extreme dryness. The XRD results detected no traces of evaporites and the paleoenvironment is interpreted to be a transgressional phase. Fractures can form as bentonite absorbs water into the interlayer region of the crystal lattice. However, the style of 190 fracturing present in the fractured layers is not observed near documented bentonite beds (Gold et al., 2017).
Throughout the stratigraphic column, there is varying silica content, and a correlation is observed in which the fractured layers contain low silica content and a high calcite content. The XRD results demonstrate a significant quantity of quartz hosted in each layer, excluding the acid washed clay layer. These results also yield that the silica is consistently depleted in all the fractured layers throughout the outcrop when compared to nearby unfractured layers. This produces a negative 195 correlation in which fracosity increases as silica content decreases (Fig. 8).
The quartz could be detrital; however, most quartz observed through petrography is in the form of cements or biosiliceous tests that have become mobilized silica during the transformation to quartz at depth (Fig. 12, Fig. 13). This observation is consistent with the primary source of silica being biogenic and not detrital. A significant quantity of detrital quartz is present, based on petrographic analysis (Fig. 11). We assume that essentially all albite grains are detrital and so reflect siliclastic 200 deposition, which would have been accompanied by detrital quartz. This conjecture is consistent with albite composition plotted versus quartz composition (Fig. 7), in which there is no overall correlation but a positive correlation between quartz content and albite content within the sub-populations of both the unfractured and fractured layers.
An explanation for the stratified fracture pattern should account for the systematic paucity of silica within those layers (i.e., the "de-silicified" fractured layers) and the abundant silica within adjacent beds. Many possibilities exist, but we suggest that 205 silica diagenesis drove fracturing by either (i) modifying the mechanical properties of the host layers, such that they became susceptible to fracturing by imposed tractions, (ii) modifying the hydraulic properties of the host layers, such that they became susceptible to fluid overpressures, or (iii) producing volume changes within the host layers, directly causing stresses and driving fractures open as a result.
A significant present-day mechanical property contrast is unlikely, for the reason discussed above: fractures are abundant 210 elsewhere in the section which are confined to brittle limestone layers, and absent from intervening shales. However, we can only guess at the mechanical properties the beds would have had during the transition from biogenic silica to quartz, which occurs during burial to relatively shallow depths, at temperatures less than 80°C (Davies & Cartwright, 2002). Precipitation of quartz would likely have coincided with mechanical consolidation if the sediment were not consolidated by that time.
Lithification of a pair of relatively silicified limestone beds, and their bonding to an unsilicified limestone bed in between, 215 under confining stress, might have produced residual stresses that led to fracturing once the stress state changed (Bourne, https://doi.org/10.5194/se-2020-50 Preprint. Discussion started: 28 April 2020 c Author(s) 2020. CC BY 4.0 License.

2003)
. Likewise, silicification may have reduced permeability to create local overpressures, assisting the development of disequilibrium compaction. However, both of these processes would seem equally capable of forming fractures in unsilicified units underneath the lowest silicified bed, and yet such fractures are not observed.
Silica diagenesis has been proposed to create opening-mode fractures (Hooker et al., 2017). The correlation between the 220 fractures and silica content suggests that diagenesis could have generated fracturing the dissolution and reprecipitation of biogenic silica. Mobilized silica would produce quartz in the form of lightly silicified layers and result in a fracturing at stratigraphic levels that are dominated by dissolution and thus, volume loss. Hooker et al (2017) presented evidence that silica diagenesis can explain fracture volumes of several percent; we find similar results here (see calculations in Appendix 2). However, fractures in that study were almost exclusively layer perpendicular. The explanation given was that volume loss 225 within a laterally pinned rock layer resulted in subvertical failure planes. A similar process could have produced the present fractures, but their layer-parallel orientation implies that the host rock would have been vertically pinned-that is, prevented from contracting vertically.
Energetically, the meaning of "vertically pinned" is that the energy required to collapse the de-silicified layer, SC, is greater than the energy required to generate fractures (2G, where G is the surface energy of the rock), plus any strain energy required 230 in association with fracturing, SF. SC and SF may be manifest in the form of elastic deformation, grain-boundary sliding, shear microfracturing, mineral solution, or crystal plasticity. We assume that SF arises from internal deformation of the layer, which does not shorten vertically. As such, collapse of the de-silicified layer would entail a relative energetic benefit, H, by lowering the overlying rock mass by the distance to which the de-silicified layer is shortened. We assume that the fractures formed in lieu of this lowering, and so can treat this lowering distance as equivalent to h, the cumulative aperture of 235 horizontal fractures: where ρ and z are the density and height of the overlying rock, respectively, and g is the acceleration of gravity. 240 For horizontal fracture development to be energetically favorable over the collapse of the de-silicified bed, we therefore should meet the criterion: The units of each term are energy per unit area; the two strain terms (S) are volumetric strain energy times layer thickness.
Re-arranging (2), we can state that the energetic cost of shape change in the form of collapse, in excess of that for shape change in response to fracturing, is, at least, the sum of the fracture surface energy and the potential-energy benefit of lowering the overburden by collapse: The right side of (3) can be estimated using G ~10 J/m 2 (Friedman et al., 1972), ρ < 2700 kg/m 3 , h < 10 -2 m in most beds ( Fig. 9, Fig. 10), and z < 3000 m for the depth of the silica transition (Davies and Cartwright, 2002). A high-end estimate of 255 the right side of (3) is therefore on the order of 10 6 J/m 2 , with G being negligible compared to H. We have little constraint as to the left side of (3), but we can estimate the range of elastic strain energy that would be required to vertically contract the host material, and thus give an estimate of SC assuming elastic behavior. Young's moduli for clays and soils range from approximately 1 × 10 6 to 1 × 10 7 Pa, and limestone ranges from approximately 2 × 10 10 to 6 × 10 10 Pa (Gudmundsson, 2011). To achieve a 1% shortening strain (Fig. 9, Fig. 10) would require between 1 × 10 4 and 6 × 10 8 Pa of pressure. Layer 260 thicknesses on the order of 0.1 m imply an energy requirement between 1 × 10 3 and 6 × 10 7 J/m 2 . Therefore, the high-end estimate of the right side of (3) is considerably lower than the high-end estimate of the left side. So, if there is a large difference between SC and SF, the fracture mechanism of volume loss amid vertical pinning is plausible.

Conclusion
Outcrop, petrographic, and mineralogical observations in Ordovician limestones in Pennsylvania show that stratified 265 fractures are located in beds that lost silica, likely during the diagenetic transition of biogenic silica to quartz. Silica is observed as dominantly biogenic grains and cement, with subsidiary detrital grains, and there is an anti-correlation between fracture volume and silica content. Fractures are observed in many orientations and are poorly explained by formation in response to tectonic strains or fluid overpressures. A transgressional environment rules out a desiccation mechanism. The fractures are not related to bentonite deposition, and dolomitization is ruled out via the absence of dolomite in XRD. 270 Migration of silica explains the volume lost from the fractured layers; calcite cement currently fills the fractures. The details of how volume loss produced layer-parallel fractures are unclear, but a mechanical analysis indicates such a mechanism is possible if the host rock is prohibited from vertical contraction during volume loss.

Data availability
Data from this study are listed in Appendix 1. 275

2-base
Unfractured dark gray layer below fracture layer 2, thin section of contact F2 Fractured layer 2

6-base
Unfractured dark gray layer below fractured layer 6, thin section for contact 6base & F6 F6 Fractured layer 6, medium gray limestone then sharp shift

Volume Loss Calculation
Volume loss calculations were used to analyze whether the volume that migrated from fractured layers could explain the fracosity amount observed. The volume represented by fractures within any fractured layer was hypothesized to be an estimate of the amount of silica that migrated from the fractured layers to the silicified base layers. We began by assuming that the distribution of silica within each bed was initially homogeneous. For fractured layer 6, fractured layer 2, 6 base 410 layer, and 2 base layer, we then assumed that the final distribution of silica within the fractured layers (F2 and F6) and their underlying base layers (F2-base and F6-base) resulted from silica migration during the transition from opal to quartz. The mineral weight percent generated by XRD was then converted to volume percent using the following relation: