Organic-rich shale samples from a lacustrine sedimentary sequence of the Newark Basin (New Jersey, USA) are investigated by combining Broad Ion Beam polishing with Scanning Electron Microscopy (BIB-SEM). We model permeability from this 2D data and compare our results with measured petrophysical properties. Three samples with total organic carbon (TOC) contents ranging from 0.7% to 2.9% and permeabilities ranging from 4 to 160 nD are selected. Pore space is imaged at high resolution (at 20,000x magnification) and segmented from representative BIB-SEM maps. Modeled permeabilities, derived using the capillary tube model (CTM) on segmented pores, range from 2.3 nD to 310 nD and are relatively close to measured intrinsic permeabilities. SEM-visible porosities range from 0.1% to 1.8% increasing with TOC, in agreement with our measurements. The CTM predicts permeability correctly within one order of magnitude. The results of this work demonstrate the potential of 2D BIB-SEM for calculating transport properties of heterogeneous shales.
The pore network is the main control on transport processes in low-porous media, such as gas shales [
As the properties of a rock’s pore system directly govern permeability, image-derived pore-scale models (also known under the term
In contrast, this work aims to determine pore system properties and permeabilities by BIB-SEM imaging in 2D from three different shale samples. Permeability predictions are based on a simple model [
Three samples (Table
Drilling sites, geological origin, and current depth of the sample material.
Sample | Well # | Formation | Member | Depth [m] |
---|---|---|---|---|
NJ-001 | Titusville 2 | Passaic | Perkasie | 21.98 |
NJ-019 | Princeton 2 | Lockatong | Ewing creek | 54.86 |
NJ-023 | Nursery 1 | Lockatong | Wilburtha | 928.06 |
Bulk mineralogical compositions were derived from X-ray diffraction (XRD) patterns of randomly oriented powder preparations. The measurement was done on a
Total organic carbon (TOC) and total inorganic carbon (TIC) were measured on powdered samples with a
Vitrinite reflectance (
He-porosities were calculated by combining skeletal densities from helium expansion (pycnometry) and bulk densities from cylindrical sample plug dimensions (
Gas permeability coefficients were measured on the same cylindrical sample plugs drilled parallel to bedding with helium gas as permeate at 25°C (298 K) in dry condition. They were installed into triaxial flow cells and then loaded to a confining pressure level of 40 MPa. After installation, the system was flushed with helium and leak-tested. Permeability measurements were conducted at confining pressure levels of 40, 30, 20, and 10 MPa during unloading. At each confining pressure level, nonsteady state flow tests (pressure pulse decay) were then performed at various pore pressures from 0.5 to 5 MPa. From the pressure incline/decline versus time series, apparent permeability coefficients were calculated. Details on the experimental setup, pressure pulse decay tests, and gas permeability calculation procedure are described in [
Due to gas slippage, measured apparent gas permeability coefficients are higher than “intrinsic” permeability coefficients. Gas slippage effects were corrected by applying the Klinkenberg-correction on the apparent permeability data of a given confining pressure level [
The end of all plugs used for the permeability measurement were cut off and used for BIB-SEM investigation. Subsamples were cut dry into rectangular blocks of 3 × 5 × 5 mm using a low-speed microdiamond saw. Subsampling was based on macroscopic investigations, that is, checking for milling locations perpendicular to the bedding plane featuring visibly different kinds of layers, for example, darker (more clayey) and brighter (siltier) layers. These locations were then BIB-polished by a
The end trim of the plugs used for the permeability measurement of samples NJ-001 and NJ-019 was additionally polished perpendicular to the bedding using sand paper (SiC) in preparation of cm2 scale energy dispersive X-ray spectroscopy (EDX) analysis.
The Field Emission SEM used for image acquisition in this work is a
Pores can be identified on BIB-polished surfaces using a SE2 detector at low acceleration voltages of 3 to 5 kV [
Several image mosaics were scanned at different magnifications to gain microstructural information at multiple scales. For each sample, at least one BIB cross-section was prepared, and typical layers were selected based on their mineralogical composition and mapped at high resolution with the SE2 detector, and when relevant, also with the BSE and EDS detector: (
Assuming that the pore space is strongly linked to the mineral phases, the representative elementary area (REA) of mineralogy is an approximation for the representative area of porosity [
The segmentation of mineral phases included thresholding and automatic image treatment based on BSE and EDX maps using different toolsets within
Porosity was segmented from SE2 images utilizing a “seed and grow” algorithm [
Macroscale EDX analysis of the surface of the plug’s end trim with 125x magnification allowed distinguishing layers with typical mineralogy that can be used for upscaling to a respective plug’s mineral content. To compare permeability predictions of a single layer to the measured permeabilities of respective plugs (both parallel to the bedding), typical layers of sample NJ-001 and NJ-019 were assessed via EDX intensity measurements. The total plug area was then separated into typical layers to obtain the area proportions of each representative layer.
To identify the properties of distinct layers, each layer is characterized by a BSE/EDX mosaic map (Figure
We used the CTM similar to [
The tortuosity characterizes flow pathways in shales and describes the connectivity of the pore network. If tortuosity equals 1, flow may be fully described by the Hagen–Poiseuille equation [
Besides, modeling of the cumulative permeability coefficients can be optimized by correcting the visible porosities for isolated
The samples vary in TOC, with the highest organic content in sample NJ-001 and lowest organic content in sample NJ-023 (Table
Petrophysical measurements of the investigated samples featuring elemental analysis, helium pycnometry (He), non-steady-state gas permeability measurements, and vitrinite reflectance data.
Sample | TOC [wt.%] | TOC [vol.%] | Bulk density [g/cm3] | He grain density [g/cm3] | He-porosity [%] | Permeability |
|
---|---|---|---|---|---|---|---|
NJ-001 | 3.85 | 8.39 | 2.54 | 2.68 | 5.27 | 160.33 | 2.52 |
NJ-019 | 1.82 | 4.07 | 2.61 | 2.72 | 3.75 | 43.15 | 2.05 |
NJ-023 | 0.76 | 1.74 | 2.68 | 2.72 | 1.66 | 3.94 | 2.70 |
XRD data of all investigated samples listing mineral groups (bold) and their corresponding minerals (italic).
Sample | Quartz | Feldspar |
|
|
Carbonate |
|
|
Clay + mica |
|
Pyrite + others | Total |
---|---|---|---|---|---|---|---|---|---|---|---|
NJ-001 |
|
|
|
|
|
|
|
|
|
|
100.0 |
NJ-019 |
|
|
|
|
|
|
|
|
|
|
100.0 |
NJ-023 |
|
|
|
|
|
|
|
|
|
|
100.0 |
Bulk densities vary between 2.54 g/cm3 (NJ-001) and 2.68 g/cm3 (NJ-023) and grain densities between 2.68 and 2.72 g/cm3. As a result, calculated He-porosity values of sample NJ-001 are highest with approx. 5.1% followed by NJ-019 with 4.1% and NJ-023 with 1.4% (Table
Similarly, Klinkenberg-corrected permeabilities (extrapolated to zero stress) decrease from sample NJ-001 to NJ-023. Sample NJ-001 shows highest permeability with
All samples are characterized by a very heterogeneous fabric with a variety of different clasts embedded in a fine-grained matrix framework. Grains occur in a wide range of sizes and mostly touch each other. Layering lies within submillimeter to centimeter scale (NJ-001 slightly more distinct than NJ-019 or NJ-023). In between clasts, dispersed OM, intercalated by clay minerals, is evident and qualitatively analyzed to be gradually less from sample NJ-001 to NJ-023 according to the chosen TOC sequence. The given samples feature three typical submillimeter layers with an abundance of large albite (Figure
BIB-SEM investigation of sample NJ-001 including their local porosity maps and selected areas of interest depicted in Figure
Mineralogical, microstructural, and morphological features of clasts (red), pore space (yellow/purple), and rock matrix showing BSE (a, c, e, g, i, k) and SE2 (b, d, f, h, j, l) recordings of sample NJ-001. Grains occur in a wide range of sizes but were chosen according to their characteristic properties for this illustration (e.g., due to partial idiomorphism). Porosity is predominantly distributed between clasts and OM interfaces (
The feldspar-rich laminae are responsible for the high porosities in the investigated samples (Figure
Comparing the qualitative results to typical features of marine shales, lacustrine shales are often differentiated by (
Few albite grains embody large pores Some carbonates feature cluttered pores Quartz and pyrite show almost no associated porosity (cf. Figures Relatively small pores are located within the dispersed occurring OM (cf. Figures All other types of intergranular porosity
The visible porosity of sample NJ-019 and NJ-023 is significantly less but occurs similar to the cavities described in sample NJ-001 (Figure
High magnified images revealed porosity
Dispersed OM pervaded by porosity in sample NJ-001 (a) and almost no OM porosity in sample NJ-019 or NJ-023 (b).
BIB-SEM investigation of sample NJ-019 including their local porosity maps and selected areas of interest depicted in Figure
BIB-SEM investigation of sample NJ-023 including local porosity map and selected areas of interest depicted in Figure
Mineralogical, microstructural, and morphological features of clasts (red), pore space (yellow/purple), and rock matrix of sample NJ-019 and NJ-023. Grains occur in a wide range of sizes but were chosen according to their characteristic properties for this illustration. Porosity is predominantly distributed between clasts and OM interfaces (
This indicates the presence of at least two different OM types, interpreted as dispersed bitumen that migrated into cavities between grains; primary terrestrial vitrinite and inertinite particles.
Furthermore, the investigated lacustrine shale samples on average exhibit less visible porosity (~0.8%) than their marine organic-rich counterparts at similar maturity. In contrast to other studies on overmature organic-rich marine shales (e.g., [
The mineralogical proportions of the investigated samples vary when comparing the microstructure of sample NJ-001 to NJ-019 and NJ-023, respectively (Figure
Comparison of EDX and XRD (plus TOC) data, where XRD can be seen as the intrinsic mineralogical composition.
Photographs of all three samples (top) plus distinct layers based on EDX analysis (bottom) and their respective area fractions of the plug surface for sample NJ-001 and NJ-019.
EDX results show that the mineral content varies most with up to approx. 27%, 16%, and 8% for carbonate, feldspar (albite), and pyrite, respectively, in the maps of sample NJ-001. Variability of the mineralogical proportions between maps of sample NJ-019 is up to approx. 36%, 28%, and 21% for clays, quartz/silicates (correctly classified as K-feldspar by XRD analysis), and albite, respectively (Figure
Summary of BIB-SEM investigation results showing EDX mineral compositions (bars) and visible porosities as well as pore orientations of all pores (top), respectively, excluding microcracks and pores below the PPR (bottom) of each mosaic map.
All pores below 18 pixels (equal to 72 and 144 nm in diameter at 20,000x or 10,000x magnification, resp.) in size must be considered with caution since pores of these sizes are below the practical pore resolution (PPR). Microcracks (narrow elongated pores) that may originate from drilling, core recovery operations, drying, or sample preparation are identified and removed.
Total visible porosities between all mosaic maps range from approx. 2.0% (NJ-001) to almost no visible porosity (NJ-023), while considering only pores above the PPR lowers these values slightly (up to ca. 1.7%) (Table
Overview of parameters received through the pore shape analysis. All parameters excluding cracks (as well as the power law exponent) are also given above the practical pore resolution of 18 pixels (PPR).
Sample | Mosaic map | Magnification | Area [ |
Number of pores | Visible porosity | Avg. circularity | Avg. AR |
|
||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
total | >PPR | total [%] | >PPR [%] | total | >PPR | total | >PPR | |||||
NJ-001 | Map #1 | 10,000x |
|
32,187 | 12,228 | 2.01 | 1.72 | 0.60 | 0.31 | 0.55 | 0.40 | 2.36 |
Map #2 | 20,000x |
|
21,507 | 5,313 | 1.43 | 1.21 | 0.60 | 0.32 | 0.60 | 0.48 | 2.12 | |
Map #3 | 20,000x |
|
76,740 | 8,519 | 1.83 | 1.38 | 0.59 | 0.31 | 0.67 | 0.45 | 2.32 | |
|
||||||||||||
NJ-019 | Map #1 | 20,000x |
|
2,331 | 904 | 0.53 | 0.47 | 0.50 | 0.26 | 0.44 | 0.22 | 2.02 |
Map #2 | 20,000x |
|
4,562 | 1,559 | 0.85 | 0.79 | 0.52 | 0.23 | 0.47 | 0.29 | 1.87 | |
Map #3 | 20,000x |
|
4,368 | 1,632 | 0.45 | 0.40 | 0.52 | 0.30 | 0.46 | 0.29 | 2.03 | |
|
||||||||||||
NJ-023 | Map #1 | 20,000x |
|
853 | 241 | 0.07 | 0.07 | 0.60 | 0.55 | 0.65 | 0.50 | 1.84 |
Experimental permeabilities (intrinsic) as well as modeled permeabilities (cumulated and upscaled, resp.). Cumulated values show the modeled permeability of each mosaic map derived from the CTM. Upscaled values were normalized (weighted average) to characteristic layer proportions of a whole plug surface (Figure
Sample | Map | Major element | Intrinsic |
Cumulated |
Upscaled |
||
---|---|---|---|---|---|---|---|
|
|
|
|
||||
NJ-001 | Map #1 | Na | 160.3 | 800.8 |
|
665.8 | 310.1 |
Map #2 | Ca | 537.6 |
| ||||
Map #3 | S | 38.4 | 37.6 | ||||
|
|||||||
NJ-019 | Map #1 | Ca | 43.1 | 7.9 | 3.0 | 323.1 | 43.1 |
Map #2 | Na | 1,208.4 |
| ||||
Map #3 | K | 14.5 | 13.3 | ||||
|
|||||||
NJ-023 | Map #1 | — | 3.9 | 2.3 | 2.3 | 2.3 | 2.3 |
Initial classifications of pore types according to Desbois et al. 2009 [
Very high magnification images (40,000 to 80,000x) revealed omnipresent porosity within the OM in sample NJ-001, while samples NJ-019 and NJ-023 show no OM porosity (Figure
Pore frequency histograms with power law based bin sizes are given in Figure
Pore size histograms and cumulative visible porosity for all samples.
Log-log pore size distributions for all three samples.
Cumulated permeability graphs of each individual mosaic are presented in Figure isolated, relatively large pores below the PPR; visible microcracks (all pores with an AR threshold value of ≤0.2 and circularity of ≤0.3) which are assumed to be artefacts from sample handling.
Results of the capillary tube model (predicted permeabilities), indicating the importance of pore sizes when adding up the single permeabilities within each pore bin size. Solid lines mark the experimentally derived permeabilities.
Examples of pore space corrections illustrated for Map #1 and Map #2 of sample NJ-001 and Map #2 of sample NJ-019. Large, isolated
Through the investigation of characteristic plug layering, three representative layers of each formation were identified (Maps #1 to #3 of samples NJ-001 and NJ-019, resp.). For upscaling, the predicted permeability values of each mosaic are normalized (by the weighted average) to the total area proportions of each plug. Hence, the weighted average permeability of the three samples changes from 665.8, 323.1. and 2.3 nD to 310.1, 43.1, and 2.3 nD, respectively, after taking the mineralogical composition of the plugs NJ-001, NJ-019, and NJ-023 into account (Table
The sum of modeled permeability coefficients of microcracks ranges from
The mineralogical compositions of EDX and XRD are similar for samples NJ-001 and NJ-019 (Figure
Except for the fraction of OM, no other mineral phase shows a clear correlation between its fraction and the porosity trend (Figure
Comparison of EDX and XRD (plus TOC) data in regard to the correlation of mineral phases and porosity. Only OM reveals a clear trend with porosity, both, independently of applying direct (EDX) versus indirect (XRD plus TOC) types of investigation.
Moreover, solid trends between mineralogy, microstructure, and pore properties throughout all investigated scales imply representative sampling. The bulk measurements exhibit clear correlations between petrophysical properties such as the trend of decreasing He-porosities as well as permeability coefficients along with decreasing TOC content (Table
Unlike sandstones, shales do not feature clear poro-perm relationships [
We used the pore system characteristics of typical layers that were resolved by BIB-SEM for upscaling to plug scale (Figure
For the permeability calculation, several adjustments on the pore data were conducted as they were expected to influence the modeled permeability coefficient. Very narrow, elongated pore shapes, identified as microcracks, do not influence the predicted permeabilities significantly (the predicted
Albite and its distinct porosity seem to control permeability since the albite rich maps show the highest calculated permeability (Table
Limitations of stereological modeling apply to the CTM calculations as well. Obtaining adequate tortuosity values, especially from 2D data, is problematic. A 3D investigation could deliver different tortuosity factors. By choosing a tortuosity of
For validating the BIB-SEM based permeabilities, the results are compared to the experimentally derived permeabilities measured on the exact same sample plugs. Assessing the mineralogy and microstructure is key to allow upscaling to greater sample volumes for meaningful comparison. This was established by normalizing the results of the CTM to the plug scale in this study. The final results of the corrected and upscaled BIB-SEM permeabilities are 310 nD, 43 nD, and 2 nD for sample NJ-001, NJ-019, and NJ-023, respectively. These values are at maximum twice or half, respectively, when compared to the experimentally derived permeabilities (160.3, 43.1, and 3.9 nD). A good correlation of permeabilities modeled by the CTM compared to measured permeabilities by gas permeation was also found by Philipp et al. 2017 [
BIB-SEM is a powerful tool that can deliver meaningful permeability values parallel to bedding from 2D pore areas based on simple capillary tube models as shown in this study for the heterogeneous Newark Shale. Upscaling pore system characteristics and permeability predictions from BIB-SEM derived permeabilities of the heterogeneous Newark Shale on the plug scale (upscaled) are representative and close to experimentally measured permeabilities. Silty laminae and OM porosity control porosity and permeability of the lacustrine Newark Shale.
Pore elongation is based on a pore’s axial ratio (
Fluid flow through porous media is described by Darcy’s law for incompressible media:
Here,
The Hagen–Poiseuille equation describes laminar flow through a long cylindrical pipe of constant cross-section (a capillary tube):
Combining Darcy’s law (see (
With the assumption of a sample cube of the length
Solving (
Porosity of a sample cube with a single capillary can also be derived from the 2D BIB-SEM data by using areas instead of volumes and thereby neglecting the 3rd dimension:
The hydraulic radius can also be calculated from the BIB-SEM data by
The authors declare that there are no conflicts of interest regarding the publication of this paper.
The authors thank P. Bertier of the Clay and Interface Mineralogy (CIM) research group at the RWTH Aachen University for providing them with the thorough XRD analysis. P. E. Olsen is acknowledged for his allowance to sample the Newark Basin Coring Project cores and his help in the IODP core store at Rutgers University.