High-speed nanoindentation mapping of organic matter-rich rocks: A critical evaluation by correlative imaging and machine learning data analysis
Research output: Contribution to journal › Article › Research › peer-review
External Organisational units
- Universität Bielefeld
- China University of Geosciences, Beijing
- Erich Schmid Institute of Materials Science
- GEOS4 GmbH
Nanoindentation is a valuable tool, which enables insights into the material properties of natural, highly inhomogeneous composite materials such as shales and organic matter-rich rocks. However, the inherent complexity of these rocks and its constituents complicates the extraction of representative material parameters such as the reduced elastic modulus (E r) and hardness (H) for organic matter (OM) via nanoindentation. The present study aims to extract the representative H and E r values for OM within an over-mature sample set (1.33–2.23%Rr) from the Chinese Songliao Basin and evaluate influencing factors of the resulting parameters. This was realized by means of high-speed nanoindentation mapping in combination with comprehensive optical and high resolution-imaging methods. The average E r and H values for the different particles range from 3.86 ± 0.17 to 7.52 ± 3.80 GPa and from 0.36 ± 0.02 to 0.64 ± 0.09 GPa, respectively. The results were subsequently processed by the unsupervised machine learning algorithm k-means clustering in order to evaluate representative E r and H results. The post-processing suggests that inherent heterogeneity of OM is responsible for considerable data scattering. In fact, surrounding, underlying and inherent mineral matter lead to confinement effects and enhanced E r values, whereas cracks and pores are responsible for a lowered stiffness. Adjusted for these influencing factors, a declining trend with increasing maturity (up to 1.96%Rr) could be observed for E r, with average values calculated from representative clusters ranging from 5.88 ± 0.37 down to 4.07 ± 0.32 GPa. E r slightly increases again between 2.00 and 2.23%Rr (up to 4.85 ± 0.35 GPa). No clear relationship of H with thermal maturity was observed. The enhanced accuracy archived by a large data set facilitated machine learning approach not only improves further modelling attempts but also allows insights of impacting geological processes on the material parameter and general understanding of mechanical behavior of OM in rock formations. Thus, the presented multimethod approach promotes a fast and reliable assessment of representative material parameters from organic rock constituents.