Impact of Composition on Pore Throat Size in Mature Shales

Impact of Composition on Pore Throat Size in Mature Shales The impact of composition on pore throat size and permeability in mature shales: an example in Middle and Upper Devonian Horn River Group shale, northeastern British Columbia, Canada Tian Donga, Nicholas B. Harrisa, Korhan Ayrancia, Cory E. Twemlowb, Brent R. Nassichukb a Department of Earth and Atmospheric Sciences, University of Alberta, Edmonton, AB T6G 2E3, Canada, b Trican Geological Solutions Ltd., Calgary, AB T2E 2M1, Canada, Abstract Shale reservoirs of the Middle and Upper Devonian Horn River Group provide an opportunity to study the influence of rock composition on permeability and pore throat size distribution in mature formation. Sedimentological, geochemical and petrophysical analyses reveal relationships between rock composition, pore throat size and matrix permeability. In our sample set, measured matrix permeability ranges between 1.69 and 42.81 nanodarcies and increases with increasing porosity. Total organic carbon (TOC) content positively correlates to permeability and exerts a stronger control on permeability than inorganic composition. A positive correlation between silica content and permeability, and the abundant presence of interparticle pores between quartz crystals, suggest that quartz content may be another factor enhancing the permeability. Pore throat size distributions are strongly related to TOC content. In organic rich samples, the dominant pore throat size is less than 10 nm, whereas in organic lean samples, pore throat size distribution is dominantly greater than 20 nm. SEM images suggest that in organic rich samples, organic matter pores are the dominant pore type, whereas in quartz rich samples, the dominant type is interparticle pores between quartz grains. In clay rich and carbonate rich samples, the dominant pore type is intr aparticle pores, which are fewer and smaller in size. High permeability shales are associated with specific depositional facies. Massive and pyritic mudstones, rich in TOC and quartz, have relatively high permeability. Laminated mudstone, bioturbated mudstone and carbonate facies, which are relatively enriched in clay or carbonate, have relatively low permeability. Key words: Pore throat size; Permeability; shale composition; Horn River Group shale; Western Canada Sedimentary Basin 1. Introduction Typical shales or mudstones are sedimentary rocks with a dominant grain size less than 63 ÃŽÂ ¼m, serving as source rocks if organic matter is rich and as seals preventing hydrocarbon migration because of fine-grained nature (Schieber, 1998). Permeability is a fundamental property in conventional reservoirs that strongly influences hydrocarbon production rate. Permeability is presumably also important in shale reservoirs for long term flow rates, although initial production rates are also influenced by natural and artificial fracture systems (Jarvie et al., 2007; Rickman et al., 2008). Permeabilities in mudstones are typically several orders of magnitude lower than in coarser grained lithologies, such as siltstones and sandstones (Dewhurst et al., 1999; Nelson, 2009; Yang and Aplin, 2010). Published absolute permeabilities, measured on a variety of shales and by different analytical methods, typically fall in the nano-darcy range (Kwon et al., 2004). Because of the extremely low p ermeability, accurate measurements of permeability in shale samples are challenging (Sakhaee-Pour and Bryant, 2011; Tinni et al., 2012; Moghadam and Chalaturnyk, 2015). Steady-state flow techniques are impractical because it is difficult to achieve flow through shale plugs in a period of time short enough to permit analysis of large numbers of samples (Mallon and Swarbrick, 2008; Sakhaee-Pour and Bryant, 2011). Consequently, transient pulse decay methods, which require much less time, are generally employed to measure shale permeability on both plugs and crushed particles (Cui et al., 2009). One potential problem in using core plugs for pulse-decay measurements is that induced fractures may influence the measurements (Ghanizadeh et al., 2015); therefore, a crushed rock technique (the GRI method) may be a favorable method to measure the matrix permeability (Cui et al., 2009). On the other hand, where microfractures exist naturally in a shale, the GRI method might not be appropriate. In mudstones, permeability primarily depends on the abundance and size of pores and pore throats (Yang and Aplin, 1998; Dewhurst et al., 1999); under reservoir conditions, pore throats and consequently permeabilities may be substantial lower than measured under ambient conditions due to compression of pore throats. Permeability under in-situ conditions is difficult to measure, but it can be estimated from more easily determined petrophysical properties such as pore size and pore throat size distribution as well as surface area (Yang and Aplin, 1998). Mercury injection capillary pressure (MICP) measurements provide a qualitative understanding of permeability by giving useful information about the pore throat size and connectivity. MICP data suggest that pore throat size distributions in mudstones are influenced by porosity, grain size and clay content (Dewhurst et al., 1999; Yang and Aplin, 2007). Previously published data indicate that pore throat sizes in shales ranges from 5 nm to more than 100 nm (Nelson, 2009). Reported permeabilities in mudstones vary by ten orders of magnitude, primarily controlled by the presence of clay minerals, which decreases permeability by clogging mineral associated pores (Neuzil, 1994; Yang and Aplin, 1998, 2007, 2010; Dewhurst et al., 1998; Dewhurst et al., 1999). Permeabilities are also impacted by diagenetic processes such as destruction of porosity by mechanical compaction and cementation, and enhancement of pore throats by mineral dissolution (Pommer and Milliken, 2015). Most samples in these studies are either organic lean mudstones or low maturity, and the dominant pores exist between particles. Recently, high resolution scanning electron microscopy combined with ion milling techniques applied to mudstone samples has documented another important set of pores, i.e. those developed within organic matter (Loucks et al., 2009; Loucks et al., 2012; Nelson, 2009; Slatt and OBrien, 2011; Chalmers et al., 2012a; Curtis et al., 2012a; Curtis et al., 2012b; Dong and Harris, 2013; Dong et al., 2015; Mastalerz et al., 2013; Klaver et al., 2015; Tian et al., 2015). However, little work has been done on the control exerted by organic matter and other compositional variables on pore throat size distribution and permeability . Some studies have described pore features and factors controlling the matrix permeability in the Horn River Group shale (Ross and Bustin, 2009; Chalmers et al., 2012b), but none have been sufficiently detailed to determine the compositional factors influencing pore throat size distribution and permeability. In this study, we present a large dataset of permeability measurements on crushed samples and pore throat structure determined by MICP data By integrating geochemical data and petrophysical data for the Horn River Group shale, we investigate the potential effects of shale composition and organic matter on pore geometry, pore throat size distribution and permeability. We then link permeability to lithofacies, which can be used to predict spatial variation in permeability. 2. Geological setting The Horn River Basin, an area of nearly 12,000 km2, is situated in the deep northwest portion of the Western Canada Sedimentary Basin in northeastern British Columbia, Canada (Fig. 1) (Oldale and Munday, 1994). It is bounded to the south and east by carbonate barrier reefs (Presquile barrier) and to the west by the Bovie Fault, a Cretaceous structure associated with Laramide tectonism (Ross and Bustin, 2008). During the Middle and Late Devonian, the southern part was proximal to the paleo-shoreline and received more siliclastic input than the more distal northern part of the Horn River Basin (Fig. 1) (OConnell, 1994; Dong et al., 2016). The Horn River Group shale includes the Evie and Otter Park Members of Horn River Formation and the Muskwa Formation (Fig. 2), all deposited within a roughly 8 m.y. interval spanning the Givetian to early Frasnian Stages (~ 392 to 384 Ma) (Oldale and Munday, 1994). In the Horn River Basin, most of the Horn River Group shale is within the dry gas windo w with a vitrinite reflectance (Ro) ranging between 1.6 and 2.5% (Ross and Bustin, 2008, 2009; Rivard et al., 2014). The Evie Member is a dark grey, organic rich, variably calcareous mudstone that overlies the shallow marine carbonates of the Lower Keg River Formation (McPhail et al., 2008; Hulsy, 2011). The Evie Member is up to 75 meters thick near the Presquile barrier, thinning to less than 40 meters to the west (McPhail et al., 2008). The average TOC content for the Evie Member is 3.7 wt.% (Dong et al., 2015). The Otter Park Member is typically a grey, pyritic, argillaceous to calcareous mudstone. It is much thicker than the underlying Evie Member and the overlying Muskwa Formation, as much as 270 meters in the southeast Horn River Basin (McPhail et al., 2008). The Otter Park shale generally has lower organic content than either the Evie or the Muskwa, averaging 2.4 wt.% TOC (Dong et al., 2015). Portions of the Otter Park Member are rich in organic carbon with up to 7.09 wt.% TOC (Dong et al., 2015). The Otter Park shale varies geographically in composition, becoming argillaceous in distal part s of the basin to the north and west. The Muskwa shale is a gray to black siliceous, pyritic, organic-rich shale that overlies the Otter Park Member. The Muskwa Formation varies in thickness from 50 to 90 meters (Oldale and Munday, 1994). Organic carbon enrichment in the Muskwa Formation is generally higher than in the Otter Park Member but slightly lower than in the Evie Member, averaging 3.41wt.% TOC (Dong et al., 2015). The Muskwa Formation is overlain by the Fort Simpson Formation which is poor in organic matter. 3. Methodology We obtained core samples from four wells drilled in the Horn River Basin distributed from the northern distal part of the basin to southern proximal part: EOG Maxhamish D-012-L/094-O-15, Nexen Gote A-27-I/094-O-8, ConocoPhillips McAdam C-87-K/094-O-7 and Imperial Komie D-069-K/094-O-02 (Fig. 1). All samples were slabs cut from a 10 cm diameter core and were, on average, approximately 10 cm long and 6 cm wide. Splits were cut vertically along the sides of the core samples for geochemical analysis, permeability measurements, MICP analysis and SEM image analysis, so that the different analyses were performed on the same interval of rock. Before sampling, these four cores were stratigraphically logged in order to identify the sedimentological and ichnological characteristics and define lithofacies (see Dong et al., 2015, 2016 for methods on sedimentological analysis). Weatherford Laboratories analyzed total organic carbon (TOC) content using LECO combustion. Acme Analytical Laboratories determined the major element concentrations, including SiO2, Al2O3, Fe2O3, MgO, CaO, Na2O, K2O, TiO2, P2O5, MnO and Cr2O3 by using Inductively Coupled Plasma Mass Spectrometry (ICP-MS). Detailed information on analytical procedures for TOC and major oxides was provided in Dong et al. (2015). We selected ten samples (Table 1) for bulk mineralogical analysis and Based on the lithofacies classification, we selected five samples (Table 2) representing different lithofacies for QEMSCAN analysis, carried out by Whiting Petroleum Corporation, Denver. QEMSCAN is an automated SEM-based mineralogical analysis tool, and can be used for the quantitative determination of mineral abundance and identification of micro-texture (Ahmad and Haghighi, 2012). Permeability and porosity were measured on one hundred samples (Table 3) by Trican Well Service Ltd., Calgary, Alberta. Samples were crushed, sieved with a 10 mesh screen and dried in an oven at 105ËÅ ¡C to remove any existing fluids. Matrix permeability was measured on the crushed and sieved samples using the GRI method (Luffel et al., 1993). Helium pycnometry was used to measure the grain densities of each crushed sample. Ultra-high purity helium was used to maximize penetration of pore space and minimize potential reactions with the samples (Cui et al., 2009). Permeability was calculated at ambient conditions based on a method refined from ResTech (1996) and Luffel et al. (1993), and was not calibrated to insitu conditions. Pore throat size distributions were measured by mercury porosimeter on shale chips. We selected thirty-six samples (Table 4) from the four wells representing a wide range of TOC contents and mineralogical compositions to do the mercury injection analysis (Klaver et al., 2015). Mercury injection capillary pressure (MICP) analyses force mercury into pore throats and pores under increasing applied pressure. Pore throat diameters, not pore diameters, are then interpreted from the MICP measurements. The samples were dried in a vacuum oven over 12 hours and then intruded with mercury from 2 to 60000 psi using Micromeritics AutoPore IV 9500 V1.09 apparatus at the Department of Physics, University of Alberta. The minimal pore throat diameter can be measured by this instrument is 3 nm. Scanning electron microscopy enabled visualization of pores on samples polished with ion milling, which produces extremely smooth surfaces (Loucks et al., 2009). Eleven shale samples (Table 5) from core plugs were first mechanically polished and then further polished using ion milling (Fischione Model 1060 SEM Mill at the Department of Earth and Atmospheric Sciences, University of Alberta). Composition of the 11 samples is provided in Table 5. Ion milled samples were mounted to SEM stubs using carbon paste and coated with carbon to provide conductive surfaces. The prepared samples were imaged with two different field-emission SEMs. One was a JEOL 6301 F field-emission scanning electron microscope at the Scanning Electron Microscope Facility at the University of Alberta. We performed the FE-SEM analysis using an accelerating voltage of 5.0 kV and working distance range from 10-15 mm. The other was a Zeiss Sigma field-emission scanning electron microscope coupled with an EDX EBSD at t he nanoFAB facility, University of Alberta. The FE-SEM was performed using an accelerating voltage of 10.0 kV and working distance around 8.5 mm. Secondary electron (SE) images document the pore systems and topographic variation. Backscatter Electron Detector (BSE) and Oxford Instruments 150mm X-Max Energy Dispersive X-Ray Detector (EDX) provided the compositional and mineralogical variation. 4. Results 4.1 Lithofacies classification We identified five lithofacies based on thin section analysis and core observation from the four cores within Horn River Basin: massive mudstone, massive mudstone with abundant pyrite lenses and laminae (pyritic mudstone), laminated to heterolithic bedded mudstone (laminated mudstone), bioturbated mudstone, and carbonates. More detailed descriptions and photographs of the lithofacies are presented in Dong et al. (2015). Massive mudstone, lacking physical sedimentary structures and primarily comprising quartz (Figs. 3A and 4A), dominates the Muskwa Formation and the Evie Member (Figs. 5 and 6). Pyritic mudstone is characterized by pyrite-rich laminae and pyrite nodules (Figs. 3B and 4B), and dominates the Muskwa Formation in all four cores, and also dominates the Otter Park Member in the EOG Maxhamish core (Figs. 5 and 6). This lithofacies has less quartz but more clay than massive mudstone. Laminated mudstone is common in the Otter Park Member (Figs. 5 and 6) and consists of millimeter scale clay-rich mudstone laminae with quartz- and calcite-rich silt laminae (Figs. 3C and 4C). Bioturbated mudstone is characterized by moderate to intensely bioturbation and weak lamination (Figs. 3D and 4D) and primarily occurs in the lower part of the Otter Park Member (Figs. 5 and 6). Compared to the massive and pyritic mudstones, the laminated and bioturbated mudstones are relatively rich in clay (Figs. 4C and D) . The carbonate lithofacies, rich in calcite (Figs. 3E and 4E), is restricted to the lower part of the Evie Member (Figs. 5 and 6). 4.2 TOC content, major oxides concentration and mineralogy TOC content for all samples in our data set ranges from 0.04 to 8.25 wt.%, with a mean value of 3.09% (Dong et al., 2015). Lithofacies vary systematically in TOC content (Fig. 7A). Massive mudstone samples are richest in TOC, ranging from 0.82 to 8.25%, averaging 4.23 wt.%. Pyritic mudstone samples have TOC values ranging from 0.3 to 6.81 %, averaging 3.44 wt.%. Laminated mudstone samples have relatively low TOC, between 0.24 and 7.09 % (mean TOC = 2.02 wt.%). Bioturbated mudstone and carbonate mudstone samples have the lowest TOC values, between 0.04 and 3.05 % (mean TOC = 1.11 wt.%). TOC content is highest in Evie Member, moderate in Muskwa Formation and lowest in Otter Park Member (Dong et al., 2015). The oxides SiO2, Al2O3 and CaO represent the major components of quartz, clay and carbonate minerals, indicated by the strong correlation coefficient between major oxides and quantitative mineralogy from XRD analysis (Fig. 8). Thus concentrations of these oxides can be used as proxies for quartz, clay and carbonates. Oxide compositions differ greatly among lithofacies (Figs. 7B-D). The massive mudstone and pyritic mudstone lithofacies are relatively rich in SiO2, ranging from 9.9-80.1% and 12.3-89.4% with average values of 56.3 and 66.5%, respectively. The laminated mudstone and bioturbated mudstone lithofacies are richer in Al2O3, with concentrations of Al2O3 ranging from 2.0-17.0% and 9.1-19.7% with average values of 9.2 and 17.1%, respectively. The carbonate lithofacies is richest in CaO, ranging from 43.8-52.6% with an average of 47.6%. SiO2 concentration is highest in Muskwa Formation, Al2O3 concentration is highest in Otter Park Member, whereas CaO concentration is highest in E vie Member (Dong et al., 2016). Mineral components identified by X-Ray Diffraction (XRD) are presented in Table 1 and include quartz, K-feldspar, plagioclase, calcite, dolomite, pyrite and clay minerals (Dong et al., 2016). The clay fraction is dominated by illite and mixed-layer illite/smectite, plus a trace of chlorite in some samples. 4.3 Permeability Matrix permeability profiles from the EOG Maxhamish, Imperial Komie, Nexen Gote and ConocoPhillips McAdam cores are shown in Figs. 5 and 6. The average permeability for all samples is 15.6 nD, ranging from 1.69 to 42.81 nD (Table 3 and Fig. 9). Permeability is highest in the Evie Member (average permeability = 17.15 nD), moderate in Muskwa Formation (average permeability = 15.18 nD), and lowest in the Otter Park Member (average permeability = 14.44 nD). 4.4 Pore systems Porosity measured on core samples ranges from 0.62% to 12.04%, averaging 5.1% (Dong et al., 2015). Pores are categorized as micropores (pore diameter 50 nm) by the International Union of Pure and Applied Chemistry (Sing, 1985). Loucks et al. (2012) recognized three general types of pores in shales: organic matter pores, interparticle pores developed between grains and crystals, and intraparticle pores contained with a particle boundary. All three pore types were observed in our shale samples (Figs. 10, 11 and 12). In our Horn River Group shale samples, mesopores and macropores were observed in the high resolution SEM images (Figs. 10, 11 and 12). Micropores are smaller, below the limit of the SEM images resolution (Dong and Harris, 2013). Pores are common in organic matter and are predominately round or elliptical in cross-section with a wide size range from a few nanometers (Figs. 10B, D and E) to greater than 1 micron (Fig. 10C). Pore abundance within organic matter is strongly heterogeneous, with both non-porous solid organic matter and porous organic matter commonly observed (Figs. 10A and F). Even within the same patch of organic matter, we observed dense area and porous area (Fig. 10B). The size of organic matter pores is also highly variable; for example, mesopores dominate the pore system in sample IK4 (Fig. 10E), whereas macropores dominate sample M2(Figs. 10A and C). Interparticle pores are observed between quartz crystals, calcite crystals and other detrital particles, such as feldspar (Fig. 11). These pores display triangular and elongated shapes (Fig. 11), substantially different in morphology and size from organic matter-hosted pores which are typically ovoid and elliptical in shape. The pore size and morphology of interparticle pores depends on the surrounding minerals, geometry and arrangement of adjacent particles. Most interparticle pores are much larger than organic matter pores, typically greater than 100nm. Interparticle pores are also present between fine-grained phyllosilicate particles that occupy primary pores between carbonate particles (Fig. 12F), displaying smaller size. Intraparticle pores are found within particles or mineral grains, such as clay minerals, carbonate grains, pyrite framboids and apatite. They include primary pores preserved during burial and diagenetic processes and secondary pores generated by dissolution of feldspar and carbonate. Pore spaces within clay flocculates are common in clay rich samples (Fig. 12A). Pyrite framboids, aggregates of submicron pyrite crystals, are relatively common in Horn River Group shale and contain mesopores developed between the submicron pyrite crystals (Fig. 12B). Apatite also provides sites for porosity development (Fig. 12E). Numerous intraparticle pores are present within carbonate grains due to carbonate dissolution (Figs. 12D and E). All fractures observed in the Horn River Group shale are completely open and lack cement filling (Figs. 12C and D). In clay rich samples, the fractures are probably artificial shrinkage cracks produced as the clays dehydrated (Fig. 12C). In the carbonate rich samples (Fig. 12D), fractures surrounding calcite grains are narrower and shorter than fractures in clay rich samples and are interpreted to be natural. 4.5 Pore throat size distributions Porosity and pore size distributions, calculated from nitrogen adsorption analyses, were presented in Dong et al. (2015). These date show that the Horn River Group shale samples contain mixtures of macropores, mesopores and micropores. Pore throat size distributions are more critical than pore size distributions to permeability (Nelson, 2009). Sample preparation and applied injection pressure of up to 60000 psi may either cause artificial fractures in our samples or results in collapse of large pores (Yang and Aplin, 2007; Chalmers et al., 2012a). In this study, pore throats related to artificial fractures were removed from the distributions (Fig. 13). Samples in Figs. 13 are grouped by increasing TOC content. Pore throat diameter distributions are increasingly skewed towards smaller values with increasing TOC content. Samples with low TOC content (Figs. 13A, B and C) are characterized by asymmetric distributions with dominant pore throat radii greater than 20 nm. Pore throat diameters less than 10nm dominate in the organic rich samples (Figs. 13D, E and F). Median pore throat diameter is thus negatively correlated to TOC content (Fig. 14A), but no association with major inorganic components is evident (Figs. 14B, C and D). Mercury intrusion porosimetry also can be used to calculate effective porosity. Porosity calculated from mercury injection ranges from 0.6% to 2.9%, averaging 1.5%, which is much lower than total porosity measured by helium pycnometer. There is a positive correlation between TOC content and effective porosity, yielding a correlation coefficient of 0.44 (Fig. 15). 5. Discussion 5.1 Relationship between porosity and permeability Previous studies have shown that the relationship between porosity and permeability in mudstones is primarily controlled by the clay content (Yang and Aplin, 2007; 2010). At a given porosity, Dewhurst et al. (1998, 1999) found that clay poor mudstones are much more permeable than clay rich mudstones. The samples in the Dewhurst et al. (1998, 1999) studies were shallowly buried London clay, with a TOC content between 0.2 and 0.9 wt.%. The samples in the study of Yang and Aplin (2007) are core samples from North Sea and Gulf of Mexico, with a range of TOC from 0.1 to 2.4 wt.%. Samples in those studies are organic lean mudstones and no organic matter pores were reported in their studies. The loss of porosity and permeability is largely driven by the preferential collapse of large primary pores. The wide range of permeability (3 orders of magnitude) likely can be explained by the variation in grain size, which is in turn affected by the clay content (Dewhurst et al., 1998, 1999; Yang and Aplin, 2007). In our Horn River Group shale dataset, however, the relationship between porosity and permeability do not vary systematically with the concentration of Al2O3 (Fig. 9B), which is an approximation for clay content. Unlike the studies cited above, samples with high clay content does not show lower permeability at a given porosity than samples with low clay content. The primary reasons for the contrast between our results and those of Dewhurst et al. (1998, 1999) and Yang and Aplin (2007) are probably the high organic content and the high maturity of the Horn River samples and the definition of clay content. In their studies, clay content is defined as particles less than 2 ÃŽÂ ¼m regardless of mineralogy, whereas we defined the clay content as the abundance of clay minerals including smectite, illite, mixed layer of smectite+illite and chlorite. The samples in this study have a TOC content range of 0.04-8.25 wt.%, with a mean value of 3.09%, approximately 3 to 10 times higher than in the Dewhurst et al. (1998, 1999) and Yang and Aplin (2007) data sets. Ross and Bustin (2008, 2009) showed that Horn River Group shale is highly mature, with vitrinite reflectance from approximately 1.6 to 2.5% in contrast to the low maturities in Dewhurst et al. (1998, 1999) and Yang and Aplin (2007). Dong et al. (2015) reported that hydrogen index (HI) and oxygen index (OI) are very low in Horn River Group shale, indicative of dry gas window. Compared to economically successful shale gas plays in North American such as Barnett Shale (Jarvie et al., 2007) and Eagle Ford Shale (Pommer and Milliken, 2015), Horn River Group shale is more mature, although it is less mature than the gas-productive Silurian black shales in Sichuan Basin, southwestern China, which have an equivalent vitrinite reflectance (%Ro) range of 2.84 3.54 (Tian et al., 2013). We propose that the extensive development of organic matter pores in mature shales impacts the relationship between clay content and porosit y-permeability behavior. Porosity-permeability relationships are shown in Fig. 9. Our permeability data show a positive correlation with porosity, yielding a correlation coefficient of 0.72 for all the samples (Fig. 9A). Porosity is the strongest individual predictor of matrix permeability, stronger than any correlation between any compositional parameter and permeability. 5.2 Relationship between shale composition and pore throat size distribution TOC and median pore throat size calculated from mercury injection capillary pressure data (Fig. 14A) are negatively correlated, suggesting that smaller median pore throat size occurs in organic rich samples than in organic lean samples. The smaller pore throat size in organic carbon rich samples (predominantly less than 10 nm) is also evident in histograms of pore throat size distribution (Figs. 13D, E and F). This relationship is consistent with observations from scanning electron microscopy (Fig. 10), where most of the organic matter pores are less than 100 nm. Similar phenomenon have been observed in Devonian shales, Appalachian Basin, where pore throat size is much smaller in organic rich samples (averaging 8 nm) than in organic poor samples (averaging 22 nm) (Nelson, 2009). Bernard et al. (2012) suggest that in the Barnett Shale, organic pores formed not in kerogen, but rather in bitumen which derived from thermally degraded kerogen in the oil window and in pyrobitumen, which resulted from secondary cracking of bitumen in the gas window. In this study, bitumen, solid bitumen and pyrobitumen are defined as secondary organic matter, following terminology in Pommer and Milliken (2015). Although it is operationally challenging to distinguish bitumen or pyrobitumen from kerogen on SEM images, organic matter in the Horn River Group shale probably consists of mixtures of kerogen, bitumen and pyrobitumen (Fig. 10), as all the stratigraphic units are currently in the dry gas window. A certain fraction of the buried detrital and marine kerogen apparently has been converted to hydrocarbon and secondary organic matter, generating the numerous bubble-like pores (Fig. 10). Pommer and Milliken (2015) identified similar processes in the Eagle Ford Shale, where, over a range of thermal maturities from oil window to gas window, original primary mineral-associated pores are largely infilled by secondary organic matter, in which much smaller organic matter pores (median size 13.2 nm) later develop. Primary intergranular pores between rigid grains such as quartz, calcite were clogged by kerogen, bitumen and pyrobitumen, where small organic matter pores were generated because of the thermal conversion from kerogen to hydrocarbon (Figs. 10B and E). Clay content does not appear to be significantly related to pore throat size in the Horn River Group shale, in contrast to some previous studies (Yang and Aplin, 2007; 2010) (Fig. 14C). At deposition, pore throat size and connectivity is a function of the shape, size and packing pattern of the constituent clasts. Clay-sized particles damage matrix permeability by clogging pores and throats (Yang and Aplin, 2007, 2010). Large primary pores may have been present in the Horn River Group shale at low maturities and relatively shallow burial depths, but at its present-day high thermal maturity (gas window), primary pores have been largely lost due to compaction, suggested by the twisted clay flakes (Fig. 12A). In clay rich samples, only a minor amount of secondary organic matter pores are present (Fig. 12B). Any correlation between clay content and pore throat size that may have existed at low maturity was effectively erased by diagenesis. 5.3 Shale composition and permeability Organic matter pores, which generally are interpreted to be generated during burial and maturation (Jarvie et al., 2007; Zargari et al., 2015), have been well documented in organic rich shales such as the Barnett Shale, Woodford Shale, Marcellus Shale and the Kimmeridge Clay Formation (Loucks et al., 2009; Passey et al., 2010; Curtis et al., 2012a; Fishman et al., 2012