Evaluating Molecular Evolution of Kerogen by Raman Spectroscopy: Correlation with Optical Microscopy and Rock-Eval Pyrolysis

Vitrinite maturity and programmed pyrolysis are conventional methods to evaluate organic matter (OM) regarding its thermal maturity. Moreover, vitrinite reflectance analysis can be difficult if prepared samples have no primary vitrinite or dispersed widely. Raman spectroscopy is a nondestructive method that has been used in the last decade for maturity evaluation of organic matter by detecting structural transformations, however, it might suffer from fluorescence background in low mature samples. In this study, four samples of different maturities from both shale formations of Bakken (the upper and lower members) Formation were collected and analyzed with Rock-Eval (RE) and Raman spectroscopy. In the next step, portions of the same samples were then used for the isolation of kerogen and analyzed by Raman spectroscopy. Results showed that Raman spectroscopy, by detecting structural information of OM, could reflect thermal maturity parameters that were derived from programmed pyrolysis. Moreover, isolating kerogen will reduce the background noise (fluorescence) in the samples dramatically and yield a better spectrum. The study showed that thermal properties of OM could be precisely reflected in Raman signals.


Introduction
Organic matter (OM), usually known as kerogen, is derived from the animals, plants, and micro-organisms burial and preservation within the rock matrix [1]. Organic-rich intervals are usually found in shale reservoirs, including solid bitumen, kerogen, and moveable hydrocarbons. Understanding kerogen properties regarding maturity has a key role in the development of unconventional resources. It is well understood that the quantities of hydrocarbon generation are a function of kerogen content and type in the formation, its maturity [2], and the diffusion and permeability of the hydrocarbons [3]. Thermal maturity of kerogen is usually understood by the reflectance measurement of vitrinite. This method can be difficult if the samples have no primary vitrinite or it is dispersed widely in the sample [4,5]. The second method to evaluate the maturity of organic matter is programmed pyrolysis [6,7]. The most widely used pyrolysis technique is Rock-Eval, in which a sample is subjected to a programmed heating from 300 • C to 650 • C in an inert gas atmosphere [8][9][10][11]. The results are reported in various parameters characterizing the kerogen type, quality, and its potential to produce hydrocarbons. All of these properties reflect the molecular structure of kerogen and the chemical compounds that exist in the organic matter.
In this study, four samples from Bakken Formation in Williston Basin in North Dakota were collected. Then, Raman signals were related to the %VRo, and maturity parameters from Rock-Eval pyrolysis (Tmax, S1, S2, and production index). The conclusions are then confirmed with the Fourier Transform Infrared Spectroscopy (FTIR) spectrum of two samples with the highest and lowest maturities to provide us with a better insight of molecular transformation of kerogen. In order to avoid the errors that may exist in the Raman spectrum when samples are mechanically polished or when carbonaceous minerals are present [5,23,28] and fluorescent background, kerogen was isolated from the mineral matrix. The Raman spectra of both the in-situ and pure kerogen were compared and the discrepancies in the data were presented.
It should be noted in this study that, for the first time, we are proposing the process in which Raman signals can be correlated to Rock-Eval results. However, due to the lack of Raman and pyrolysis data that are necessary for a strong conclusion, to better support our hypothesis, additional data from published literature has been also utilized.

Geological Setting
The late Devonian-Early Mississippian Bakken Formation is one of the main source rocks of the Williston Basin [29][30][31]. The Williston Basin is covers the southern portion of Canada (Manitoba and Saskatchewan) and the northern part of the United States (Montana and North Dakota), Figure 1. The Bakken Formation is underlain by the Three Forks Formation and is overlain by the Lodgepole Formation. The Bakken as an organic rich formation consists upper and lower members serving as source rocks and middle member as the reservoir with low porosity and permeability. Thickness of the Bakken Formation is about 150 ft and has no surface outcrop. The Bakken Formation has high TOC (Total Organic Carbon) content, which ranges from 5 to 20 wt % with an average of 11.33 wt % across the basin [32][33][34]. Different literature showed that most abundant kerogen type in the Bakken Formation are types I and II [35][36][37].

Sample Preparation and Experiments
The samples are taken from four different Wells (A, B, C, and D) and they vary in depth and maturity, core images are shown in Figure 2. Small pieces of samples were homogenized and divided into three parts to be used for Raman spectroscopy, vitrinite reflectance (%VRo), and Rock-Eval analysis. The same portions of the samples that were tested for Raman measurement were then gone through the kerogen isolation process by acid to separate the organic matter from mineral matrix.
To evaluate the maturity of samples, they were crushed (850 µm) and hardened by mixing with the epoxy resin [4]. In the next step, to ensure the surfaces are scratch, samples were polished by polishing equipment. The random reflectance (%VRo) and qualitative fluorescence were then analyzed with a Carl Zeiss Axio imager A2m microscope. As primary vitrinite is absent in sedimentary formations that are older than the Devonian, the same as in four samples from Bakken Formation (Figure 3), bitumen reflectance (%BRo) was used to obtain an equivalent vitrinite maturity (%VRo-eq) using the Jacob equation (Equation (1)) [38], which is one of numerous such equations available. In addition, VRo-eq was calculated by converting Tmax from Rock-Eval pyrolysis through Equation (2) [39]. Equation (2) was used because it is derived from the Mississippian-age Barnett Shale in Texas, which contains type II marine kerogen, similar to that found in the Bakken Shale.
%VRo-eq = %BRo × 0.618 + 0.4 (1) %VRo-eq = 0.0180 × Tmax − 7.16 (2) In the next step, programmed Pyrolysis was performed on samples by Rock-Eval 6. Table 1 summarizes the properties that were derived from the Rock-Eval analysis of four samples. As in Figure 4 Type II is the main type of Kerogen of our samples, TOC values vary from 12.69 wt % to 16.36 wt %, and %VRo-eq maturity fluctuates from 0.53 to 0.99. S1 is the free hydrocarbons that are present in the rock, while S2 is the amount of hydrocarbon generated through pyrolysis and is used to derive hydrogen index (HI) [10]. S3 is the CO 2 that will be released during pyrolysis and is in relationship to the oxygen in the kerogen structure that is used to derive oxygen index (OI) [10]. Production index (PI) can also be a maturity indicator [10]. More description of Rock-Eval pyrolysis can be found in [11].

Isolation of Kerogen
Polishing rocks will create dislocation of constituent components within the samples that will affect their Raman response. Moreover, fluorescent background noise due to presence of bitumen and sulfur in lower maturity samples masks the signal [15,23,40]. In order to avoid these two issues, kerogen was isolated from the samples, according to [41,42].
The process can be summarized as below: • Pre-acid preparation of the samples: sample quality needs to be evaluated to ensure they are properly washed and are brought to the appropriate mesh size (crushed if necessary).
Typical starting sample size is 20-30 g of material, depending on organic richness.

•
The sample is then placed into concentrated 37% HCl (small increments should be added to prevent excessive foaming, while timing is dependent on sample reaction).

•
Samples are placed into concentrated 48% HF (HF is for digestion of remaining silicates).

•
The samples are placed into concentrated HCl for the second time. • Centrifugation: this is done essentially by taking a portion of the remnant of the sample and placing it into a centrifuge tube to create the kerogen slide. This process will separate the solid kerogen from remaining liquids, Figure 5.

Raman Spectroscopy
Inelastic scattering of monochromatic light in Raman spectroscopy, such as a laser beam, from the surface of the sample, interacts with molecular vibrations [43]. The interaction causes an exchange of energy in which the energy of laser photons will either shift up or down [44][45][46]. Based on the frequency of vibration of the molecule(s), change in energy varies, which leads to unique spectral patterns for each molecule.
Raman spectra were acquired using 532 nm excitation energy by focusing the laser beam on the surface of four samples (both in situ and extracted), Figure 6. It should be noted that Raman signals might interfere with fluorescence noise in generally low thermal maturity samples.

Raman Signals on Pure Kerogen
As mentioned before, bitumen and sulfur content in mud rocks tend to express high fluorescent background noise in Raman spectrum. As shown in Figure 6, the isolation of kerogen removes this issue by eliminating the effects of bitumen. Comparing spectra before and after kerogen isolation in Figure 6 indicates high background fluorescent levels in in-situ samples which have been reduced, specifically for Well A with lower maturity. Table 2 contains the details of the Raman spectroscopy data for both isolated and shale kerogen. It can be concluded from the table that standard deviations of D and G bands for isolated kerogen are less than the in-situ (non-isolated) samples. The higher deviation of D bands when compared to G bands can be attributed to the fact that depending on the orientation of the crystallites to the exciting laser beam, spectral features of disordered sp2 carbons (D band) may vary notably [51].
Moreover, it has shown by [52] that polishing the samples to 0.05 µm roughness influences the Raman spectra. This originates from carbonaceous matters that are very sensitive to the polishing process, thus isolation will circumvent this problem. Furthermore, bitumen and sulfur content in mud rocks tend to express high fluorescent background noise in Raman spectrum. As shown in Figure 6, isolation also removes this issue by eliminating the effects of bitumen. Comparing spectra before and after kerogen isolation in Figure 6 indicates high background fluorescent levels in in-situ samples that have been reduced, specifically for Well A with lower maturity.

Raman Spectroscopy for Evaluating Production Potential of OM
Previous studies have documented that thermal maturity reflects a systematic change in the positions, separations, and other important parameters of the bands in the Raman spectra for carbonaceous materials. As kerogen matures, it starts to make clusters of aromatic bands and also loses its heteroatoms [12][13][14]. Such evolution is perceived by Raman spectroscopy at different levels of maturation, reflecting molecular structural transformations. Figure 8 displays a non-linear correlation between band separation (G-D) versus maturity (%VRo) for samples that are taken from 12 different fields from [5,15,53], and including samples from this study. Tmax, which is also a major parameter representing the thermal maturity of the samples and is obtained from Rock-Eval analysis, shows correlation with band separation, as in earlier studies [15,17,26,27,40,[54][55][56].
It should be noted that band separation increases with a higher rate at lower maturities. This has been explained by the fact that organic matter cracks with different rate before and after the peak oil window which is now reflected in Raman signals as well [15,27,56,57].
In addition to understanding the thermal maturity of organic matter, its production potential regarding Rock-Eval parameters, such as S1, S2, and production index (PI) is also important in evaluating productive intervals. As mentioned before, S1 is the amount of free hydrocarbons (oil) already in the sample (in milligrams of hydrocarbon per gram of rock), which can be detected at 350 • C (initial heating stage) [10,58]. S2 is the quantity of hydrocarbons that will be generated through pyrolysis (550 • C). Production index is the ratio of generated hydrocarbons to potential hydrocarbons that can be produced, Equation (3) [59,60], which can give an insight about the kerogen conversion level when is plotted vs. Tmax. Figure 9 shows that our samples are in a low level of conversion stage. PI = S1 mg HC g rock / S1 mg HC g rock + S2 mg HC g rock (3) In the kerogen structure, aromatics that are linked by aliphatic and heteroatoms [15], in which solid bitumen can be trapped [61,62]. The increase in carbon aromaticity that takes place during maturation has been also detected during natural maturation that is caused by igneous intrusions [63][64][65][66][67][68], or by the increase in depth of burial [69][70][71].
Fortunately, the complex structure of kerogen correlates strongly with its maturity. As hydrocarbon is being generated, aliphatic carbon linkages are lost (reduction in HI is a good indication of the reduction in aliphatic carbon). The main contributor in the process of hydrocarbon generation is aliphatic [57,[72][73][74][75][76], which can be attributed to S2, since they are both representing kerogen potential hydrocarbon production. Such structural evolution can be monitored by different spectroscopy methods [17,51].
For instance, Infrared Spectroscopy (IR) showed the decrease in that aliphatic concentrations with maturity [72]. In another study, quantitative 13C DP MAS NMR spectrum was used on samples that were exposed to the hydrous pyrolysis, the results showed aliphatic groups are the important compound that are contributing to the hydrocarbon generation [56,57].
FTIR is an analytical technique to identify chemical compounds in organic materials as a complimentary tool to Raman spectroscopy by measuring the absorption of infrared (IR) light due to the molecular vibrations of the sample. The spectrum of IR absorption is used as a fingerprint of a chemical substance and reports the existing functional groups. The peaks correspond to the frequencies of vibrations and the amplitude is a direct indication of the amount that is present in the material. In our study, Fourier Transform Infrared Spectroscopy (FTIR) was performed on pure kerogen (to avoid the influence of the inorganic mineral matter) from Well B and Well D, which illustrated a qualitative and quantitative change in aromatic and aliphatic content with respect to the maturity. It is important to remember that these two samples have %VRo of 0.59 and %VRo of 0.92, respectively, reflecting lower and higher maturity levels. In Figure 10, it can be found that sample with %VRo of 0.59 has a higher peak for aliphatic groups, while a sample with %VRo of 0.92 has higher peaks for aromatic carbons.
Using high-resolution transmission electronic microscope (TEM) also showed such change in kerogen molecular structure specifically clustering of aromatics [17,77]. Kerogen that was isolated from the sample of Well D, which has the highest maturity, was imaged in a novel approach under Hitachi HD-2300A Dual EDS Cryo STEM (scanning transmission electron microscope), Figure 11. The image revealed the imperfect stacked aromatic layers in the sample with higher maturity. We were not able to capture the same image for lower maturity samples.
Raman spectroscopy can detect such structural change as well, which might have the potential to replace with conventional and bulk methods approaches that are usually exhaustive and time consuming for sample preparation and performing the measurement.
The origin of the G is due to the in-plane E 2g vibrational modes of the carbon atoms in aromatic ring structures [5,[26][27][28]77]. If the structure of organic matter has high aromatic content, it is corresponding to less potential of hydrocarbon production, thus, S2 can be correlated to the G band as aromatic indicator, Figure 13. As mentioned earlier, it has been shown G band shifts slightly toward higher shifts by increasing maturity concurrent with the increase in aromaticity [15]. The negative correlation between S2 and the G band is a result of this matter.
Furthermore, the abundance of aromatic structures in conjunction with high maturity corresponds to lower hydrogen content [84], Figure 12. Therefore, HI (Hydrogen Index) could be correlated with G band position, Figure 12.  The image on the right shows imperfect stacked aromatic layer in the sample. By increasing maturity, aromatics stack and make larger clusters.
By increasing maturation and consequently hydrocarbon production, the heteroatom-rich OM lose their oxygenated/hydrogenated groups [87][88][89], as a consequence attachment that separate from aromatic carbons increase [27,90]. The D band signals (~1350 cm −1 ) refers to a disorder in the atoms representing discontinuities and defects of the sp 2 carbon network, for instance, as in heteroatoms [5,27]. Consequently, D band can be utilized as an indicator of attachments separated from OM, which is happening throughout hydrocarbon generation. Thus, already generated hydrocarbon (S1) can be correlated with D band, Figure 14. According to previous studies [40,54], when maturity increases, the D band shifts towards lower wavenumber (1370 to 1330 cm −1 ) and displays lower intensities. This is related to the increase of separating attachments from aromatic groups and transforming the organic matter from disordered state to more ordered molecule [91]. The disappearance of the D band in the uppermost stage of maturity also verifies this phenomenon. (A) The same plot using data from other sources (red circles) [5,85,86], including samples in this study (blue circles) with R 2 75%; (Right) HI for data from other sources was calculated based on the definition, as they did not report HI directly. Note the negative correlation that corresponds to less hydrogen content of organic matter by increasing aromaticity. (Right) The same plot using data from other sources (red circles) including samples in this study (blue circles) and [5,85] with 87% R 2 . Note the negative correlation that corresponds to less potential of production by increasing aromaticity. Figure 14. (Left) Correlation of S1 with D band of four samples in this study; (Right) The same plot using data from other sources (red circles) [5,85,86] including samples in this study (blue circles) with 71% R 2 .
Correlating S1 and S2 with G and D band positions, provides us with a path to correlate the PI from Rock-Eval with PI derived from Raman signals based on the definition of PI (Equation (3)), Figure 15. Figure 15. Correlation between PI from Rock-Eval and PI based on definition using Raman signals. Data from other sources (red circles) [5,85,86], including samples in this study (blue circles) with 78% R 2 . PI data from other sources were calculated based on the definition, as they did not report PI directly.
Results from this study indicated that the maturation process of organic matter can be described as a nonlinear growth of aromatic clusters thus a disappearance of disordered molecules. Moreover, Raman spectroscopy is directly dependent to the presence of various organic compounds with different structures in the organic matter. This makes it suitable and a good instrument that has the potential to yield significant information about the maturity levels of the organic matter. Having a good insight into the thermal maturity can lead to a better realization of the production potential of the organic matter. Additionally, integration of different analytical methods can assist researchers to study the molecular changes that kerogen will suffer during the maturation process. The results from these studies will improve the efforts to build kerogen models when considering fractions of various carbons at different maturities. Finally, it should be mentioned that Rock-Eval pyrolysis is usually suggested to be done on the samples that were approximately taken every 30 ft. (~10 m) in each well [9]. Certainly, when it comes to the assessment of unconventional targets, closer sample spacing of 3 ft. (~1 m) is suggested to properly identify sweet spots. When considering the huge amount of analysis that is required to study this number of samples, Raman spectroscopy, can be replaced by pyrolysis. Raman spectroscopy is a fast and nondestructive equipment to create the profile of maturity, S1, S2, and PI, and accelerate the analysis with good precision. Moreover, it has been shown that organic matter has a non-negligible effect on hydraulic fracturing of organic rich unconventional reservoirs [91][92][93][94][95][96][97][98], therefore such fast methods to characterize organic matter can be useful in the geomechanical simulation of hydraulic fracturing operations.
This study was a preliminary effort and it was done on a limited number of samples. Specifically, the STEM imaging was done based on a trial and error operation since there are no similar studies or known procedures to follow. We will continue this research by adding more number of samples for pyrolysis, Raman and FTIR. Creating a good database of Gas Chromatograph data is highly recommended and underway in the next project.

Conclusions
Samples from four wells in the Bakken Formation were studied using a series of analytical techniques, including Raman spectroscopy, RE pyrolysis, %VRo, and FTIR. Then, we captured aromatic molecules under STEM to confirm our conclusions from these analytical tools. This was done in order to complete a set of information that can provide a precise and fast method to relate molecular structure of kerogen to its maturity and production potential.
Since this study was mainly focused on the applications of Raman spectroscopy to give a better insight into molecular structure of organic matter, thus to avoid possible erroneous Raman signals and to reduce the background noise, the study was performed on isolated kerogen material. This resulted in a better visible signal and more accurate results. Based on this study, we proposed that Raman spectroscopy has a good potential to predict the kerogen maturity, S1, S2, and Productivity index that is normally extracted from Rock-Eval pyrolysis, and give us a better insight about the molecular structure of organic matter. This method suggests a quick but precise approach to understand kerogen production potential, which is necessary to be evaluated for unconventional reservoirs. This study is underway on higher maturity samples with different kerogen types to support our methodology of integrating STEM imaging techniques and Raman spectroscopy.