The Sirius Passet Lagerst€atte: silica death masking opens the window on the earliest matground community of the Cambrian explosion

The full-text may be used and/or reproduced, and given to third parties in any format or medium, without prior permission or charge, for personal research or study, educational, or not-for-pro t purposes provided that: • a full bibliographic reference is made to the original source • a link is made to the metadata record in DRO • the full-text is not changed in any way The full-text must not be sold in any format or medium without the formal permission of the copyright holders. Please consult the full DRO policy for further details.

The Cambrian Explosion records the diversification of complex, mainly bilaterian, animal body plans and the expansion of animal-based marine ecosystems (Butterfield 2003). Much of what we know about this event is recorded in a succession of black shale conservation Lagerst€ atten. The Sirius Passet (SP) is the oldest of these and is located in J.P. Koch Fjord, N. Greenland, at 82°47.6 0 N, 42°13.7 0 W ( Fig. 1A) and during the Cambrian lay at approximately 10°S (Fig. 1B;Peel & Ineson 2011). The black slates have been mapped as the 'Transitional' Buen, previously interpreted as fine-grained turbidites (Peel & Ineson 2011). Buenellus higginsi is the most common macrofossil, indicative of the Nevadella Biozone in Laurentia, equivalent to the middle part of Stage 3 (lowest stage of Series 2; 520-535 Ma; see Babcock 2005).
The SP fauna is similar to that of the Burgess Shale, comprising ca. 50 species including trilobites, sponges, worms, halkieriids, lobopods and nontrilobite bivalved arthropods. The SP Lagerst€ atte predates the Burgess Shale (510 Ma) and Chengjiang biotas (520 Ma) and is therefore possibly the earliest example of high-fidelity, soft-tissue preservation in the Cambrian. As recognized from studies of other Lagerst€ atten, an understanding of taphonomy is crucial in aiding anatomical interpretations of the SP fauna and provides for a better understanding of the palaeoenvironment, community reconstruction and early diagenesis (e.g. Zhu et al. 2006).
Similarities have been drawn between the depositional setting of the Burgess Shale and SP, for example proximity to a submarine cliff line and supposed basinal, turbiditic sedimentation (Peel & Ineson 2011). These similarities have led to the untested inference that SP fossils show Burgess Shale-type preservation (Gaines et al. 2008;Peel & Ineson 2011). Burgess Shale-type preservation is defined as 'exceptionally preserved fossils whose primary taphonomic mode is one of non-mineralizing organisms preserved as carbonaceous compressions in fully marine sediments' (Butterfield 1995). In the Burgess Shale, the soft-bodied fossils are preserved as thin, multi-layered silvery films (Whittington 1980). Significant discussion has focused on the composition of the films. Are they predomi-nantly organic and result from coalification during metamorphism (Butterfield 1990(Butterfield , 1995Page et al. 2008) or predominantly composed of aligned clay minerals (Towe 1996;Orr et al. 1998)? Disagreement regarding the composition of the fossils has led to a lack of consensus regarding the mode of fossilization. Butterfield (1990Butterfield ( , 1995 proposed that the clay which surrounded the fossils during burial acted as a catalyst that inhibited microbial decomposition of the labile tissues, leaving the organic material to be 'tanned' (Butterfield 1990) and producing a kerogen film during diagenesis. In comparison, Orr et al. (1998) concluded that fossilization occurred during decomposition as a consequence of chemical interactions between the tissues and the surrounding clays. In this model, variations in composition and reactivity of the different tissues during decomposition resulted in varying accumulations of clays on the carcass either by accumulation or direct precipitation from the pore water. The enhanced preservation of nonmineralized tissues may have also resulted from a combination of environmental factors. Near bottom anoxia would have prevented sediment irrigation by bioturbators. Reduced permeability of the seafloor and exclusion of oxygen may have also resulted from an absence of coarse grains such as silt, faecal pellets or bioclasts and the presence of reduced bottom waters that may have acted to deflocculate clay mineral aggregations and facilitated the precipitation of early diagenetic pore occluding carbonate cements .
Some of the controversy surrounding the mechanisms of Burgess Shale-type preservation is in part due to the loss of primary features during postdepositional diagenesis and metamorphism. As a consequence of low greenschist facies metamorphism and cleavage formation, the phyllosilicates of the Burgess Shale currently associated with the fossils are not the same as the minerals that initially buried the fossils (Powell 2003). A detailed understanding of the metamorphic history and reconstruction of the primary bulk mineralogy of the Burgess Shale precludes the presence of highly reactive clay species necessary for the Butterfield model involving organic preservation due to clay-related suppression of decomposition-related reactions (Powell 2003). Instead, it indicates there was nothing unusual about the initial mud sediment and lends support to the Orr hypothesis of clay templating during decomposition (Powell 2003). This re-evaluation challenges the nature of Burgess Shale-type preservation and shows the importance of determining the mineralogical changes in Cambrian black shale Lagerst€ atten during metamorphism. This is critical for understanding the taphonomy, in which previous discussions of the SP, metamorphism has been largely ignored.
In this contribution, we use a combination of observations, petrography and SEM analyses to demonstrate the sequence of events leading to an early 'death mask' preservation of the trilobites within the SP and discount the influence of metamorphic processes. A key element of our proposed  Cocks and Torsvik (2011). trilobite taphonomy is the sealing of the specimens and early silicification, within the microbial mats on which they lived. This taphonomy though common in the Ediacaran is unique to the Cambrian and may only have existed for a short time due the absence of the mat grazing guild that appeared later in the Cambrian Explosion.

Methods
Samples were collected during expeditions to SP led by Harper in 2009 and 2011. Thin (25 lm) doublepolished thin sections, cut perpendicular to bedding, provided petrographic and textural data. Samples were studied for texture and composition using optical microscopy and scanning electron microscopy. Prior to SEM imaging, all samples were coated with a ca. 20 nm layer of carbon.
SEM imaging and analysis were carried out using the Hitachi SU-70 FEG SEM in Durham University using secondary electron and backscattered electron detectors at 15 kV. Both primary and secondary backscatter techniques were used to produce general images prior to elemental mapping (SEM-EDAX) and SEM-CL analysis. EDAX was carried out using the backscatter detector and the same voltage settings used for imaging. For point analysis, a cobalt standard was run before analysis to confirm quantitative results, and Tables 1 and S2 (See supplementary information) provide detailed mineralogical data. This was performed using the QUANT software and running the standard several times to ensure maximum accuracy (100 AE 5%).
SEM-CL (cathodoluminescence) was carried out using a mirror-type detector (Gatan Mono-CL) based in Durham University (Department of Physics). The machine was set to low magnification with a 10 kV voltage to allow the site of interest to be determined. The working distance was set to 20 mm and then adjusted as necessary to allow focusing of the sample. Working distance is then set to 16.2-16.7 mm and CL luminosities collected. Results were obtained using both the panchromatic mode (mono-CL) with both mirrors adjusted to panchromatic settings and then each colour filter was inserted one by one (red >600 nm, green >480-580 nm, and blue <480 nm). For individual grain studies, mono-CL was used to produce greyscale pictures rather than true colour RGB CL. The former is faster at producing images with textural details, especially at high magnification (Schieber et al. 2000). Various studies have been carried out which show that quartz grains display a variety of luminescence intensities dependent on their provenance. The standards used were adapted from Seyedolali et al. (1997) and Schieber et al. (2000). CL intensity is dependent on the density of intrinsic and extrinsic defects within the band gap of the mineral. These defects are usually structural imperfections in the quartz crystal due to vacancies within the crystal lattice. These include point defects, translations, radiation damage, shock damage, melt inclusions and fluid inclusions. These types of defects can provide information on the conditions during mineralization and subsequent post-mineralization events such as deformation and metamorphism (Frelinger et al. 2014). The results were compared to other quartz CL provenance data in the literature to identify the luminosities.
Bulk mineralogy was checked using XRD of the <2l clay fraction following the standard procedures outlined by Moore & Reynolds (1997). Rock samples were gently disaggregated in an agate mortar and pestle to avoid shearing the clays. Organic matter was removed by leaving the sample in a solution of 5% hydrogen peroxide until reaction ceased. The sample was then centrifuged at 2000 rpm, before decanting the supernatant liquid. Distilled water was then added and the sample was stirred. This was repeated until the supernatant liquid was neutral and the clay no longer fluctuated. Separation of clay fraction and saturation for XRD analysis was carried out using the following procedure. About 5 ml of dispersion agent was added and the sample was stirred well before centrifuging for 84 seconds at 800 rpm. The supernatant liquid was decanted into another tube. These steps were repeated until the sample was clean. Separated clays were then airdried and samples placed on coverslips. Samples were analysed in a Bruker D8 Advance Diffractometer. CuKa radiation counting ranged from 2 to 60°T able 1. 2h with a 0.02°2h steps at 0.85 seconds per step. Lower angles were run than in bulk analysis in order to see low-angle clay peaks. Standard d spacing was calculated following Moore & Reynolds (1997).

Sedimentary petrography
The cleavage parallels sedimentary bedding and primary sedimentary textures are preserved. Two endmember lithofacies are recognized, a 'spotted' and a silt-rich facies (Fig. 3). The 'spotted' facies consists of structureless to discontinuously laminated phyllosilicates containing Al-rich chlorite-mica aggregates that are 10-20 lm in diameter (Fig. 3A, B). Significantly, outsized clasts of silicified microbial mat occur infrequently in both the 'spotted' facies and the silt-rich facies (Fig. 3C). The silt-rich facies consists of clay to very fine silt particles composed predominantly of phyllosilicates with minor detrital quartz grains in the silt fraction ( Fig. 3D, E). Outsized, lenticular mud clasts are present in a few beds ( Fig. 3F). Bedsets are planar and cross-laminated ( Fig. 3G), with sharp to gradational boundaries between coarser and finer layers (Fig. 3H).

Bulk mineralogy
Bulk-rock composition is similar to that of average shale (Post-Archean Australian Shale (PAAS), but differs from the Burgess Shale or PAAS in higher detrital quartz, lower K/Al (0.165, see Supporting information) and an absence of calcite (Table 1). The mineral assemblage of chlorite-mica-quartz and minor/trace of carbon, albite and illite and the absence of sulphates is found to be stable from upper sub-greenschist to middle greenschist facies (Powell 2003). The absence of carbonate is a consequence of the sediment having passed through the smectite-illite transition (S-I; below). Abundant chloritoid porphyroblasts are indicative of low greenschist facies metamorphism (Higgins et al. 2001; equivalent tõ 10 km burial) and their random orientation relative to bedding indicates relatively low-pressure conditions or possible formation during retrogressive regional metamorphism. Chloritoid is represented by the general formula (Fe,Mg,Mn)Al 2 SiO 5 (OH) 2 and occurs in high-Al, low-temperature metapelites (Halferdahl 1961). It is normally formed through the reaction of pyrophyllite, a phyllosilicate clay and chlorite-forming chloritoid, quartz and water. During this reaction, any excess silica can be precipitated as inclusions within the chloritoid. Quartz inclusions are common in the chloritoid needles and appear as irregular shaped grains within the needle.
Total organic carbon ranges from 0.5 to 1%. However, Raiswell & Berner (1986) documented an exponential loss of organic carbon relative to vitrinite reflectance in shales. Lower greenschist facies metamorphism would correspond to a minimum vitrinite reflectance of 5% (Kisch 1987). Projecting the Raiswell & Berner curve to a greenschist facies equivalent yields an estimated organic carbon preservation of 15%. The initial organic carbon content of the SP would therefore have been >3%, a value typical for many oil source rocks.

SEM-EDAX
Quantitative elemental mapping across a convex epirelief cast of Buenellus (Fig. 4A) shows typical values for Si (84%), Al (11%) and Fe (5%) compared with the matrix (Supporting Information Tables S1-S3). There is evidence of small framboidal pyrite associated with the mat material, which is absent from the matrix (Fig. 4B).

SEM-CL
Panchromatic cathodoluminescence analysis shows the presence of two quartz phases. Detrital quartz grains in the matrix exhibit true colour RGB-CL brown shade of luminescence at around 540-550 nm (Fig. 4C). Silica in the mineralized mat and concave external moulds exhibits almost no luminescence (Fig. 4D). Mono-CL of metamorphic quartz grains found as inclusions within the chloritoid needle shows medium grey luminosity and a mottled texture (Fig. 4E). These were compared with individual grains of silica associated with the fossils, which exhibit no luminosity under mono-CL and appear black and textureless (Fig. 4D).

Discussion
Marine carbonaceous shales in the Cambrian were limited to shallow-water regions with turbulent circulation (Raiswell & Berner 1986). The presence in the SP of normal and reverse grading, scouring and cross-lamination indicates deposition from low-density sediment gravity flows at or just below storm wave base. Cross-lamination suggests sediment transport by ripple migration. Compacted mud ripples have been observed as extremely lowangle cross-lamination in ancient mudstones (Kom arek & Anagnostidis 1986) and are consistent with sediment transport by mud-laden sediment gravity flows (Kremer & Kazmierczak 2005). The presence of oversized clasts in the normally graded sediment suggests flows had sufficient competence to transport larger clasts, such as a denser, mudrich 'slurry-flow', transitional between a turbidity current and a debris flow (Kom arek & Anagnostidis 1986).
The chlorite-mica aggregates in the spotted facies could have originated as primary detrital grains modified during weathering and transport or, formed as the result of diagenetic or metamorphic minerals mimicking existing sedimentary textures (see Craig et al. 2009). Aggregates are concentrated into discontinuous beds. These beds are sharpbased, normally graded with scour fills and faint cross-lamination. The aggregates were probably transported and size-sorted at the seafloor. Normal grading suggests deposition was from sediment gravity flows producing successive beds of rapidly accumulated sediment (Best 2005). The fabric is primary and the texture produced by metamorphism is inherited from a pre-existing sedimentary texture. Today, aggregates are produced largely by the activities of organisms living in the water column (e.g. through filter feeding, faecal pellet production, test building) or random collisions between grains in the water column (McCave 1985). The dense accumulations of pellets suggest these were originally hemipelagites deposited during quiescent periods.
If an average uncompacted thickness of 50 m (10 m compacted) for the SP and an age span of ca. 3 myr for the interval are taken, then a minimum average, long-term sedimentation rate of about 15 mm/kyr is realistic. We infer that this high sedimentation rate improved the chance of the fossils being buried, isolating them from the oxidizing environment at the seafloor. Given the lack of penetrative burrowing, it is likely that pore waters were already anoxic a few millimetres below the seafloor.
In modern pericontinental shelves, distance from shore is a good proxy for water depth. The SP pericontinental shelf was much wider than modern analogues (Fig. 5). Due to their distance from shore, distal, yet shallow-water deposits on wide shelves may have biological and geochemical characteristics indicative of deeper-water facies. On very wide shelves, with sufficiently low seafloor gradients, incoming waves dissipate their energy well before reaching the shore (Keulegan & Krumbein 1949). This may explain why, despite the large fetch expected from the palaeogeographical setting of SP, the shallow seafloor would be little subjected to swell, thereby allowing mud deposition. The seafloor was also consolidated by mat-forming cyanobacteria shortly after deposition.
Similar storm-influenced environments are proposed for the dark grey to black laminated mudstones of the Emu Bay Shale (Jago et al. 2012), the Alum (Thickpenny & Leggett 1987) and Burgess Shales . In common with the SP, these deposits also lack penetrative bioturbation, suggesting anoxic conditions immediately below the seafloor. The presence of a diverse invertebrate benthos in the SP and an absence of pyrite framboids in the matrix indicate the lower water column was oxic with a sharp redox boundary at the sediment-water interface. Bioturbation is surficial and typically only occurs in association with generally large arthropod carcasses (Mangano et al. 2012). Benthic cyanobacteria were the dominant photoautotrophic and oxygenic microorganisms in the SP and could have produced or enhanced the oxygenated conditions at the seafloor. However, in extant cyanobacterial mats, sulphate reduction in the biomass can produce periodic hydrogen sulphide (H 2 S) emissions above the mat surface, particularly at night (Jørgensen 1979). This could have been a major factor impeding the settlement of benthic organisms on the seafloor. The mat-dwelling fauna was ephemeral and either able to escape H 2 S emissions or had adapted a resistance to short-lived H 2 S toxicity. Occasional massive H 2 S expulsion from the decaying mats might have led to mass mortalities of the mat-dwelling community (see also Weeks et al. 2002).
Microbial mat textures observed on the concave external moulds, epirelief convex casts and evidence of microbial mat growing over trilobites (see Mangano et al. 2012; fig. 1) indicate that the fauna was an autochthonous matground community previously only known from trace fossil evidence (Buatois et al. 2014).
The dense cellular aggregates comprising the microbial textures are compared with colonies of modern coccoid cyanobacteria, particularly with the exclusively benthic Entophysalis or Chlorogloea (e.g. Kremer & Kazmierczak 2005 Rajaibi et al. (2015), and provide evidence of microbially mediated syndepositional precipitation of silica. These genera are known from freshwater and marine environments, and some pleurocapsaleans are adapted to low light levels within the photic zone (Stal & Walsby 2000). Similar material has been described from the Proterozoic (Gehling 1999) and Silurian (Kremer & Kazmierczak 2005). As cyanobacteria are a morphologically conservative group, SP specimens are probably part of the same groups.
We propose the arthropod-lobopodian fauna that characterizes the Sirius Passet inhabited a warm, muddy, matground habitat close to or just below storm wave base, but within the photic zone. We have found no evidence for aeolian deposition of clay or silt in the field or in our thin sections, in contrast to the study by Boudec et al. (2014). Primary productivity was principally by benthic cyanobacteria. Contemporary shallow subtidal to intertidal car- bonate environments had a distinct shelly fauna including archaeocyathans, halkieriids, tommotiids, hyoliths, molluscs, rare trilobites, and a variety of other small shelly fossils of unknown affinity (examples cited in Mount & Signer 1985). The rapid appearance in the geological record and ecological stability of the 'Burgess Shale-type assemblages,' over some 10 million years, suggests that rather than being long-lived holdovers, successively displaced from the shallow water (Mount & Signer 1985;Morris 2009), these animals were members of new Bambachian megaguilds (groups of organisms with mutually similar adaptive strategies), highly specialized and adapted to the dynamic and unstable nature of the tropical, muddy lower shoreface and shelf (Fig. 5). The three-dimensional preservation of cyanobacterial structure, the hyporelief external moulds and particularly the epirelief casts indicate the trilobite carapace remained intact long enough for the internal labile tissues to decompose and the body cavity to be injected with mat material. Dewatering and compaction occurs during early burial. Experimental and core studies indicate porosity in muds is reduced to 33-36% by 175 m depth. Effective porosity (~70%) is lost by 1 km depth of burial (Hedberg 1936). At the inferred sedimentation rates, specimens that had not been permineralized would have been flattened to onethird the original volume after 100 kyr-1 myr, thus providing sufficient time for early silicification and critically before the S-I transition between 1.8 and 3.7 km depth of burial. Major mineralogical changes with depth take place over the S-I transition (Hower et al. 1976). The most abundant mineral, illite/smectite, undergoes a conversion from less than 20% to about 80% illite layers over this interval, after which the proportion of illite layers remains constant. As the transition proceeds, illite packets grow within a shrinking matrix of smectite, dislocations decrease and pathways for ion transport are restricted, this causes a loss of local permeability and a rise in the fluid pressure gradient. Consequent loss of hydraulic continuity with the surface and a more efficient geopressure seals the clay to fluid flow (Freed & Peacor 1989). In the S-I transition, calcite decreases from about 20 percent of the rock to almost zero, disappearing from progressively larger size fractions with increasing depth; potassium feldspar (but not albite) decreases to zero; and chlorite appears to increase in amount. At these depths, smectite reacts with Al 3+ and K + from the decomposition of potassium feldspar to form illite with the release of Si 4+ . Magnesium and iron lost from the smectite form chlorite and ultimately chloritoid. The stable clay mineral assemblage following the S-I transition is thus illite, quartz and chlorite. All the major mineralogical and chemical changes, as the response to burial and metamorphism, occurred in a closed system for all components except H 2 O, CaO, Na 2 O and CO 2 (Hower et al. 1976). Reactions during the S-I transition therefore explain the absence of calcite in the SP. The remaining reactions into lower greenschist facies metamorphism include conversion of remaining smectite to illite (up to~175°C) and the recrystallization of illite to micas (200-300°C: Totten & Blatt 1993).
The timing of silicification is critical to understanding the taphonomic pathway. Silica supersaturation of pore fluids occurred due to the synsedimentary remobilization of biogenic silica; from silica released during the S-I transition or during greenschist facies metamorphism. Our key evidence for silicification being early, shortly after death includes the following: 1 The 3-D preservation of moulds and casts and cellular textures in the microbial mat material indicate silicification occurred before compaction and dewatering of the muds. The observation that 3-D cells are filled with monocrystalline quartz preserved within the cyanobacterial mat material indicates silica deposition prior to dewatering and compaction. Because the mat material is porous and the decay of labile cell contents occurred soon after death, it is unlikely that internal cell fluid pressure would have supported the primary structures on burial. Alternately, decay gasses from decomposition may have formed cavities (Reineck & Singh 1980) that prevented the collapse of the cells prior to silicification. A similar mechanism of mineral deposition in gas has been proposed for the formation of pyrite framboids in muds (e.g. Rickard 1970) and silica infilling of Tasmanite cysts (Schieber 1996). And the fact that spherical gas bubbles are found close to the sediment surface has been documented by Forstner et al. (1968). Silica deposition is therefore most likely to have occurred very early in diagenesis and within centimetres to decimetres of the sediment-water interface.
At the proposed sedimentation rates for the SP, this would have been equivalent to a few thousand years of deposition; 2 The presence of synsedimentary rip-up clasts; 3 Microbial mat and convex external moulds are formed of non-luminescent, textureless silica.
Non-luminescing authigenic quartz lacks crystal lattice defects generated during metamorphism or volcanism (Seyedolali et al. 1997;Augustsson & Bahlburg 2003;Frelinger et al. 2014); 4 Metamorphic quartz inclusions in the chloritoid needles exhibit strong luminosity with a smooth to mottled texture. This is distinctly different from the silica associated with the fossils; and 5 The fact that chloritoid needles cross-cut all other textures indicates metamorphism occurred after silicification.
The presence of very early diagenetic silica deposition excludes a silica source during the S-I transition and during diagenetic to low-grade metamorphic recrystallization of illite to muscovite mica. Further, the early and largely biogenic porosity within the SP muds was filled by silica and not calcium carbonate (compare data from the Wheeler Formation: . That chalcedony appears to have been the initial silica phase deposited in the voids is not unusual, first documented by Folk & Weaver (1952) and subsequently in numerous other studies (e.g. Heath & Moberly 1971;Meyers 1977;Frondel 1978;Milliken 1979;Noble & Van Stempvoort 1989). The presence of pyrite framboids in the mat material indicates sulphate reduction and that pyrite remained a stable phase through to greenschist facies metamorphism.
Decay experiments show that silica precipitates from low-pH pore waters in extracellular and intracellular environments within a mat (Krauskopf 1959), particularly when the pore fluids have reduced dissolved sulphate (Birnbaum & Wireman 1985;Schultze-Lam et al. 1995). Soluble silica, in the form of monosilicic acid, H 4 SiO 4 , dissociates to H 3 SiO 4 À at pH values above ca. 9.7 (Birnbaum & Wireman 1985;Arp et al. 2003). H 3 SiO 4 À is a highly soluble form of silicic acid, and it reacts with hydrogen ions to form SiO 2 (Rickert et al. 2002; fig. 3). Silica nucleation and precipitation is sensitive to iron content of bottom/pore water and, in early Cambrian seawater, Fe 2+ appears to have been high relative to FeOOH (Muscente et al. 2014). We hypothesize that sediment-water was saturated in silica, high in Fe 2+ and low in oxygen and sulphate. Localized pH reduction was associated with the decay-initiated silica nucleation and precipitation within the mat. Silicasaturated pore waters could have been derived from the siliciclastic mud, by remobilization of biogenic silica (e.g. sponge spicules), or from the overlying seawater. Early silicification as a mechanism of exceptional preservation is rare outside volcanic hot springs and a number of hypotheses for silicification have been proposed: (1) direct precipitation into void spaces; (2) microbial mediated silica precipitation (Schultze-Lam et al. 1995;Yee et al. 2003); and (3) chemically mediated silica precipitation (Patwardhan et al. 2011). It has been shown that microbial surfaces do not directly nucleate silica mineral formation; however, they play an important role in the aggregation of polymeric silica and the deposition of silica colloids on microbial surfaces (Patwardhan et al. 2011). For example, in modern hot spring environments, silica sinters actively form in close spatial relation to microorganisms (Yee et al. 2003 and references therein). Decay compounds, proteins and amino acids are all known to affect the kinetics, surface area of material, pore structure and particle size and are therefore implicated in silicification (Perry & Keeling-Tucker 2003). At present, we cannot distinguish between the different mechanisms proposed for silicification and all may have been important.  Figure 6 shows our preferred taphonomic pathway. We propose that Buenellus was the dominant member of the matground community preserved in the SP. After death, the carapace was sealed by the growth of the microbial mat. During early burial, internal labile tissues decayed, and mat material was injected into the carcass to produce epirelief casts. The subsequent decay of the trilobite cuticle produced external moulds. All voids spaces were progressively infilled by silica as a consequence of localized pH reduction, associated with decay, occurring in the pore waters. Specimens were clearly permineralized, preserved in 3-D, before significant dewatering of the muds had taken place. Subsequent chemical reactions associated with the S-I transition and low greenschist facies metamorphism did not materially affect the preservation of the specimens.
Similar 'death mask' preservation has been described for the external moulds of soft-bodied Ediacaran biotas from the uppermost Proterozoic Rawnsley Quartzite, South Australia. However, in this example the sole veneer resulted from bacterial precipitation of iron minerals (Gehling 1999). A number of Cambrian Lagerst€ atten are characterized by pyritization of soft-bodied fossils, for example Chengjiang and Guanshan (Gabbott et al. 2004;Forchielli et al. 2014), secondary coating by iron (e.g. Emu Bay: Brett et al. 2009), or phosphatization (Zhu et al. 2014). Silica death mask preservation thus appears to be currently unique to the SP.

Conclusions
The arthropod-lobopodian fauna that characterizes the Sirius Passet inhabited a warm, muddy, matground habitat close to or just below storm wave base, but within the photic zone. Primary productivity was principally by benthic cyanobacteria. Trilobites from the SP are preserved as complete, concave hyporelief external moulds and convex epirelief casts. External moulds are shown to consist of a thin veneer of authigenic silica. The casts are silicified cyanobacterial mat material. Early silicification is supported by the presence of synsedimentary mat rip-up clasts, 3-D preservation, which indicates silicification prior to sediment compaction and textural and mono-CL evidence. The growth of metamorphic chloritoid needles, which cross-cut the silica in the matrix, indicates that metamorphism occurred much later. It is hypothesized that silicification was initiated by falling pH in the decaying mat. Pore waters are interpreted to have been initially alkali, silica-saturated, high in ferric iron but low in oxygen and sulphate. Excess silica was likely derived from remobilized biogenic silica, probably sponge spicules in the muddy sediment. Decay resulted in a lowering of pore water pH and the conditions necessary for silicification. It is not clear whether silicification was microbially or chemically mediated.
The remarkable preservation during very early silica mineralization provides a new window on the environment and location of the Cambrian Explosion. The iconic organisms of the Cambrian Explosion formed the first mat ground communities. At least in the case of the SP the presence of cyanobacterial mats, sealing both the sediments and fossils, was fundamental to the preservation of this community. With the rise of mat grazing organisms during the later Cambrian, this taphonomic window disappeared.

Supporting Information
Additional Supporting Information may be found in the online version of this article: Table S1. Results from EDAX (sample 4.56) silt-rich facies converted to oxides using INCA processing software. Table S2. EDAX spectra from sample 1.12 m from the spotted facies converted to oxides using INCA processing software. Table S3. EDAX spectra data from the mat material (taken from a cross section through an epirelief cast of Buenellu) from sample 340.103a.  . XRD plot of 2h against intensity with labels showing normalized peaks for each main clay mineral, interpreted using Moore and Reynolds (1997). These data are from sample 0.06 which is the 'lozenge rich' facies. Fig. S3. XRD plot of 2h against intensity with labels showing normalized peaks for each main clay mineral, interpreted using Moore and Reynolds (1997). These data are for sample 4.56, the 'mud-rich' facies.