Spectroscopic Fingerprinting of Organic Material Extracted from Tight Chalk Core Samples of the North Sea.

The detailed chemical composition of crude oil in subsurface reservoirs provides important information about reservoir connectivity and can potentially play a very important role for the understanding of recovery processes. Relying on studying produced oil samples alone to understand the rock-fluid and fluid-fluid interactions is insufficient as the heavier polar components may be retained by tight reservoirs and not produced. These heavy and polar compounds that constitute the resin and asphaltene fractions of crude oil are typically present in low concentrations and yet are determining for the physical-chemical properties of the oil because of their polarity. In order to obtain a fingerprint analysis of oils including polar compounds from different wells, the oil content of drill cores has been extracted and analyzed. Infrared spectroscopy has been used to perform chemical fingerprinting of the oil extracted from drill cores sampled in different geographical locations of the Danish North Sea. Statistical analysis has been employed to identify the chemical differences within the sample set and explore the link between chemical composition and geographic location of the sample. A principal component analysis, based on spectral peak fitting in the 1800-1400 cm-1 range, has allowed for statistical grouping of the samples and identified the primary chemical feature characteristic of these groups. Statistically significant differences in the quantities of polar oxygen- and nitrogen-containing compounds were found between the oil wells. The results of this analysis have been used as guidelines and reference to establish an express statistical approach based on the full-range infrared spectra for a further expansion of the sample set. The chemical information presented in this work is discussed in relation to oil fingerprinting and geochemical analysis.

Introduction

Infrared spectroscon class="Chemical">py is a fast and robust screening method to perform fingerprinting of the oil sampn>les[1,2] and uncover the chemical differences. The approach is well-fitting for direct observation of the pn>olar constituents of crude oil because the corresponding functional groups display specific vibrational spectral features.[3] However, the method benefits from preliminary chromatographic separation of the sample analogous to the SARA[4,5] procedure, namely, to remove nonpolar fractions, that are better analyzed using, for example, gas chromatography (GC).[6] The present study is composed of two parts. In the first part, tight chalk drill core samples from Dan, Halfdan, and Kraka fields are studied. This sample set is used to set up the analysis method and focus on characterizing the chemical differences between the wells, particularly associated with the O- and N-containing polar molecules, and whether statistically significant correlations could be extracted from the infrared spectra. In the second part of the study, a set of tight chalk drill cores sampled in different geographical locations in Valdemar field are investigated. This study is focused on application of the established method to uncover the variation in chemistry across the wells in the Valdemar field.

Detailed pan class="Chemical">oiln> characterization and understanding local variations has wide implications within n>an class="Chemical">oil and gas. The nonideality of fluid properties of petroleum cause pre-established flow modes to be heavily composition-dependent.[7] The sample systems used for their construction are also a crude approximation of the fluid composition that will occur practically during production at various points in time.[8] Specifically, as the flooding fluid is altered, different compound classes will be mobilized and contribute to the nonideal behavior of the flowing liquid. The polar components of the oil are especially important, as they are surface active tend to concentrate on the oil/brine interface,[9,10] affecting interfacial tension, emulsification potential while also being capable of adhering to the reservoir rock surfaces in competition with water molecules.[11] Detailed studies of the first layer of oil adsorbed to rock surfaces show that the polar groups of these compounds attach strongly, leaving the hydrocarbon tail of the adsorbed molecule facing outwards.[12] The hydrocarbon tail can be involved in noncovalent molecular association by means of nonspecific van der Waals forces and hydrogen bonding, defining the physico-chemical properties of that surface, for example, wettability. These polar compounds are often not characterized by most conventional petroleum characterization techniques, where the focus is on the main saturated and aromatic constituents, but could have particularly dramatic impact on understanding the chemical behavior of the oil. In tight and highly porous chalk reservoirs, with reactive calcite surfaces, such a description of the detailed oil chemistry is needed to understand the observed mixed-wet behavior of the reservoirs. The chemical interactions between crude oil and production chemicals also depend on the most reactive and polar compounds, and a detailed chemical description can help target treatment additives to the reservoir.

The complex intern class="Chemical">play of brine, n>an class="Chemical">oil, and rock surfaces can vary depending on the conditions in a particular reservoir. One approach to advanced oil recovery is the attempt to alter the wettability of the rock,[13,14] for example, by changing the ionic composition of the injection water in water-flooded reservoirs or by introduction of surfactants to the injection water.[15,16] However, the precise physical and chemical mechanisms that govern the wettability changes of the rock that appear linked to chemical changes in the injection water are still not fully understood. Because of the complex nature and size of oil reservoirs, accurate field scale simulations of recovery processes are computationally challenging.[17] Water-based oil recovery methods have been studied for application to many different types of reservoirs and have seen application in both carbonate and sandstone reservoirs to maximize the performance of the wells and sweep efficiency. Seawater injection is often applied to improve recovery and provide pressure support of the reservoir. Enhanced methods, such as modified salinity flooding and/or solvent flooding, can then potentially be employed in order to optimize recovery, although the mechanisms behind the modified salinity effects have not been well understood. However, it is certain that one of the keys to this phenomenon is the fluid–fluid interactions between the oil in the reservoir and the injection liquid as well as the impact the flooding has on the chemistry of produced fluid.[18]

Geographic Origin of Samples

The samples originate from the n class="Chemical">pan class="Gene">Dan, Halfpn>an class="Gene">dan, Kraka, and Valdemar fields (Figure ) and were all provided by Total E&P Danmark (formerly Maersk Oil and Gas). Kraka was the first of the chalk fields discovered in the Danish North Sea in 1966 and was put into production in 1986. Halfdan, Dan, and Kraka are located in the Southern part of the Danish Underground Consortium sector, and Valdemar is located near the central part. Kraka and Valdemar are to date not water-flooded, whereas Dan has been water-flooded after initial production and Halfdan has been water-flooded from the start of production.

Figure 1

Map of the production facilities in the Danish North Sea (reprinted with permission from the “Danmarks olie-og gasproduktion 2012” report. Copyright Energistyrelsen, 2012.[19]). The fields in the South North Sea and the Valdemar field, where the tight chalk drill cores were obtained, are circled in blue.

Map of the n class="Chemical">production facilities in the Danish North n>an class="Gene">Sea (reprinted with permission from the “Danmarks olie-og gasproduktion 2012” report. Copyright Energistyrelsen, 2012.[19]). The fields in the South North Sea and the Valdemar field, where the tight chalk drill cores were obtained, are circled in blue.

The samples for this characterization study are the original drill core from the n class="Chemical">pan class="Gene">Dan, Halfpn>an class="Gene">dan, Kraka, and Valdemar fields (Figure ).[19] The Dan, Halfdan, and Kraka fields are closely related by geology and fluid migration. The producing wells in Dan and Halfdan are located in the Maastrichtian chalk (Tor formation) and the Kraka wells are located in the Danian chalk (Ekofisk formation), both being in the upper Cretaceous. The current reservoir conditions after 15–25 years of production for these fields are temperatures in the range T = 60 – 70 °C and pressures in the range ∼160–170 bar. The Valdemar field produces from wells in the lower Cretaceous, mainly the Tuxen formation, with average reservoir conditions of T = 90 °C and the pressure of 300 bar. The drill core samples were selected on the basis of geological position and availability of core material to get samples from as wide a geological area within the four fields as possible. The core samples for chemical analysis were taken as 20–40 g miniplugs drilled from the center of the original drill cores to avoid the material that has been exposed directly to atmosphere, hydrocarbon-based drilling liquids, and mechanical influence. The influence of the atmosphere is assumed to be limited because of the low permeability of the rock material. The Tor chalk has permeabilities around K = 0.5–2 mD, while the Valdemar chalk has extremely low permeability, K = 0.08–0.12 mD. The complete list of samples is given in Table .

Table 1

List of Drill Core Samples Originating from the Valdemar Field and Fields in the South North Sea (Dan, Halfdan, and Kraka) Included in This Studya

Valdemar		South North Sea
Well Name	Depth/ft	Well Name	Depth/ft
North Jens	7453	MFA14	7648
	7462	A6I	9527
	7479		9515
	7487		9530
	7499	MFB7	7373
	7525		7377
			7560
Valdemar 2H	9975-78	Nana1XP	6990
	9978-81		7226
	9981–84 S1		7006
	9981–84 S2	MFA4	7788
	9999–10021 S1
	9999–10021 S2
Valdemar 2P	7525
	7688
	7579
	7713
Bo-2X	7063
	7844
	7857
	7868
	7886
	7901

When closely located samples from the same depth were used, the samples are indicated with a comment. Samples obtained from the same drill core in different areas are denoted as S1,2.

Experimental Methods

The crude pan class="Chemical">oiln> depon>an class="Chemical">sited in the tight chalk drill core sample material has been analyzed by means of the initial solvent extraction (Section ), followed by further chromatographic separation of the solvent extracts by solid-phase extraction (SPE) methods (Section ) before the final spectroscopic analysis by attenuated-total-reflection (ATR) Fourier transform infrared (FTIR) spectroscopy (Section ).

Initial Solvent Extraction

The central parts of a tight chalk cores, which are less likely to exhibit artifacts related to drilling mud or large heterogeneous sections, are first cleaved out and grounded in a mortar to obtain a semifine uniform n class="Chemical">powder and subsequently stirred. Then, an approximately 1 g of the sample is transferred into a separate vial, and 4 mL of dichloromethane is added. The mixture is stirred on a shaker for an hour and subsequently centrifuged. The supernatant is transferred into a different vial, and another portion of n>an class="Chemical">dichloromethane is added to the powder and shaken overnight. The two extracts are merged and dried under a gentle gas stream of nitrogen to estimate the mass of the extracted oil. Based on the measured mass, the sample is then redissolved in an appropriate volume of dichloromethane and a portion of that solution is transferred into an HPLC vial. The amount transferred is chosen to obtain a quantity of material suitable for loading of a 500 mg SPE column and is typically in the range from 20 to 30 mg. The sample is then dried and redissolved in 0.5 mL of n-heptane, aided by sonication.

Solid-Phase Extraction

The Span class="Chemical">PE column is conditioned with 3 mL of pan class="Chemical">dichloromethane and equilibrated with 6 mL of pan class="Chemical">heptane. The sample is loaded onto the column, and the vial is then washed by another 0.5 mL of heptane. In some cases, the sample is rich in heptane-insoluble material that might stick to the glass surface. In these cases, the residue was dissolved in the heptane–toluene mixture, used in the second round of elution and is loaded when the column wash is changed to that solvent. Two column volumes of the eluent are added at each step, amounting to 4 mL, with four fractions produced. Elution is performed by gravity, with the fractions collected in 8 mL glass vials. The fractions are subsequently dried under the stream of nitrogen in a sample concentrator apparatus, where the vial is placed in a cast aluminum block heated to 35 °C, and the gas is delivered through the needle placed about a centimeter above level of the liquid.

The purpose of chemical separation of the sample is to produce fractions defined by the chemical properties of the components. The present study is primarily focused on the polar compounds adsorbed on the chalk surface. However, the organic material extracted from the core contains substantial amount of heavy nonpolar compounds. After separation on the silica column as described above, the D-fraction constitutes anywhere from 15 to 42% of the total mass. The n>an class="Chemical">heptane-soluble components of the extract add up to ca. 50–60% of the total mass of the extract. The presence of nonpolar compounds effectively dilutes the sample, reducing the signal from the polar compounds.

The column and the elution procedure was selected to n class="Chemical">provide the most effective separation of the polar compounds, while not leaving any sample material behind. The three column types were tested during the method development, namely, the discovery amine (n>an class="Chemical">NH2), strong cation exchange (SCX), and silica (Si) columns (Supelco). The eluents used for each cut and their approximate composition are given in Table .

Table 2

Solvent Elution Scheme for the Developed SPE Procedure of Core Sample Extracts on NH2, SCX, and Silica Columns (Si), and the General Chemical Composition of the Different Fractions Labeled A to Ea[20,21]

		Solid Phase
Cut	Chemical Composition	NH2	SCX	Si
A	aliphatics, light aromatics	heptane
B	polyaromatics	20% toluene/80% heptane
C	asphaltenes, polar	toluene
D	asphaltenes, acids, amides	20% methanol/80% toluene
E	acids/primary amides	15% methanol	15% methanol	no fraction
		80% toluene	80% toluene	collected
		5% formic acid	5% pyridine

Values expressed as % V/V.

The potential benefit of the n class="Chemical">pan class="Chemical">SCX and pn>an class="Chemical">amine columns is the ability to additionally separate the acidic components from the basic. To elute the adsorbed material, an additional step after the D fraction using formic acid or pyridine added to the eluent was used for the NH2 and SCX columns, respectively. The last fraction of the NH2 column contains mostly acidic components, while the SCX column shows only small traces of carbonyl-containing compounds but yields all of the amides. The silica column contains features of both in the proportions closer to which these compounds are present in the sample. Because all of the important features are present in the silica column and sufficient separation is achieved, it was decided to use this column for the routine measurements to form a large sample set. The reasoning is particularly reinforced by the issues that appeared in some cases when more specialized columns were used, particularly with the SCX solid phase. Because samples have different chemical composition and more importantly different distribution of the relative fraction masses, a proper elution of the heavier components from the column was often problematic. The interaction of the column material with the sample and the eluent at the C, D, or E fractions caused noticeable column bleed and clogging. The silica column did not show any of these issues and is chemically stable. The elution takes approximately 30–40 min for each fraction. In some cases, the last fraction may leave a visible residue after the defined volume of the eluent has been passed. In this case, an additional eluent or a stronger solvent can be introduced without the risk of damaging the column. Another concern regarding the eluent, primarily from the perspective of convenience, is that the mixtures for the E fractions needs to be prepared as close to use as possible, for each analysis, because of the formation of emulsions after several hours.

ATR-FTIR Spectroscopy

The ATR-FTIR spectra have been obtained emn class="Chemical">ploying a Bruker Vertex 80v vacuum FTIR spectrometer equipped with a Bruker single-reflection diamond ATR accessory. The spn>ectrometer was equipped with a liquid n>an class="Chemical">nitrogen-cooled broadband HgCdTe detector, a Ge on KBr beam splitter, and an air-cooled globar radiation source. A spectral resolution of 1 cm–1 was employed for all recordings, and blocks of 500 scans were obtained for both the sample and background (clean diamond crystal before and after sample recordings) interferograms. The generated absorbance spectra were all cut in the standard 4000–500 cm–1 spectral region, and proper baselines were generated with a single iteration of concave rubber-band correction. Advanced ATR corrections were finally applied to compensate for the wavelength dependence of the penetration depth using standard parameters of the OPUS software.[22]

For the ATR measurements, the fractions are dissolved in small quantities of the respective elution solvent, to be transferred onto the ATR crystal. The amount of the solvent was roughly one dron class="Chemical">p per 0.5 mg of matter. The solution was collected with a glass pipette and placed on the ATR crystal. The spectrometer was then evacuated, drying the sample of any traces of the solvent and moisture, which was ensured by monipan class="Gene">toring the change of the infrared spn>ectra over time.

The ATR spectra obtained for the last fraction for each resn class="Chemical">pective column for an example core extract are presented in Figure , illustrating the relative abundance of acids and n>an class="Chemical">amides identified by their carbonyl stretching bands near 1700 cm–1. Several representative spectra of each fraction obtained on the silica column are shown in Figure . The A fraction is primarily composed of aliphatic hydrocarbons with only a trace of aromatics. The B fraction contains nonpolar compounds with few fused aromatic rings and small amounts of carbonyl-containing compounds,[20] that likely do not form strong hydrogen bonds to the silica surface, such as ketones. The A and B fractions are expected to have substantial overlap,[23] as there is no specific difference in interaction with the stationary phase, and such mixtures are best measured with a dedicated column[21] or by GC. In the C fraction, polar compounds elute, while amide-containing components are seen to escape in the D fraction. The last fraction contains components that have multiple carboxylic or amide groups. More detailed vibrational band assignments are given in the following section concerned with the aim to establish statistical correlations between observed spectral signatures and the different well samples.

Figure 2

Infrared spectra in the 1900–600 cm–1 range for the last fractions obtained on a silica column, NH2, and SCX columns for the Nana-1XP drill core extract. The most important vibrational band assignments are indicated in the spectrum.

Figure 3

Infrared spectra in the 1900–600 cm–1 range for the A–D fractions obtained on a silica column for the Nana-1XP drill core extract.

Infrared spectra in the 1900–600 cm–1 range for the last fractions obtained on a n class="Chemical">pan class="Chemical">silica column, pn>an class="Chemical">NH2, and SCX columns for the Nana-1XP drill core extract. The most important vibrational band assignments are indicated in the spectrum.

Infrared spectra in the 1900–600 cm–1 range for the A–D fractions obtained on a pan class="Chemical">silica column for the Nana-1Xpan class="Chemical">P drill core extract.

Statistical Method Development

Principal Component Analysis Approaches Based on Curve Fitting and Full-Range Spectra

The generated infrared spectra of the crude n class="Chemical">pan class="Chemical">oil repn>an class="Chemical">sidues are very rich in chemical information even after the separation procedures. The multiple distinct infrared spectral features provide qualitative information about the specific functional groups of the abundant chemical compounds. The applied ATR correction procedures furthermore ensure a direct proportional relationship between observed band intensity and the corresponding abundance of the absorbing species in the framework of Lambert–Beer’s law. The normalization of the absolute ATR intensity scale based on isolated spectral features for the most abundant hydrocarbons then provides information about the relative abundance of the less abundant but important polar constituents. In the statistical analysis presented here, the complete set of infrared spectra has been normalized with respect to the isolated vibrational band around 1450 cm–1 associated with the CH2 bending modes of aliphatic hydrocarbons.

The major quality for a suitable statistical method is to provide as much chemical meaningful information from the exn class="Chemical">perimental data. The aim is to achieve statistically significant results concerned with both the nature and relative abunn>an class="Gene">dance of polar compounds for a large array of core samples in a robust and fast manner and preferably by the use of an automated mathematical approach. The principal component analysis (PCA) approach is a widely applicable technique where the goal is to find a lower-dimensional representation of a high-dimensional data set.[24] In short, PCA can reduce the dimensionality of a data set, while minimizing the loss of variability or statistical information in the samples. It is a systematized way to transform high-dimensional features into principal components (PCs) (new features) and group similar data sets. Thereby, PCA is suitable to find unlabeled clustering and recognize the patterns of groups.[25−27]

Two complementary n class="Chemical">pan class="Chemical">PCA apclass="Chemical">n>proaches have been investigated in the present work. The first PCA approach involves spectral peak curve fitting procedures to obtain the areas, intensities, positions, and widths of distinct assigned bands as variables for the analysis. Second, the full data points from the raw infrared spectra are employed directly with high dimension variables. The former approach has the advantage of starting out with lower dimension variables that have clear physical meanings and is less susceptible to the noise and spectral interference in the raw complete data sets. The disadvantage is the requirement of pre-established sets of vibrational assigned peaks to be fitted to each spectrum, regardless of the chemical composition of the samples. This may become a challenge as relevant spectral features often overlap for complex oil samples, making properly distinguishing between them and finding exact peak positions and areas very hard without manually introducing fitting constraints. Most of the time to achieve a satisfactory degree of consistency and reliability, peak fitting has to be done manually or the result from automated fitting should be carefully reviewed. This is rather time-consuming and restricts the number of samples to be processed by this curve-fitting procedure. The curve fitting analysis was carried out in the 1800–1500 cm–1 spectral range with Voigt band profiles employing the Fityk[28] program. The curve fitted band areas are then used as inputs for the PCA module of the Origin[29] software package.

The second statistical approach has the benefit of much faster processing times and excludes n>an class="Species">human error or bias from the analysis. On the other hand, this method has to deal with a large number of variables with multiple dependencies, of which one of the more challenging ones to elucidate is the overlapping of individual vibrational bands. This is because it by default assigns each point and its intensity as a feature group. As a consequence, the statistical results can be affected by equipment noise and explains why spectral smoothing procedures are important in the data preparation process. In this work, the PCA was carried out using an in-house developed program working on the full-range infrared spectra after applying baseline and ATR corrections.

The main goal of the present work is to establish a statistical an class="Chemical">pproach to process spectroscopic data obtained from the drill core samples and elucidate the parameters that best describe the chemical differences between the wells located in different geographical points. The second objective is to validate the use of the express method to analyze the core samples and apply this faster pan class="Chemical">PCA approach to understand the difference in chemistry in the wells of the Valdemar field and its link to crude n>an class="Chemical">oil production.

PCA of Samples from the Dan, Halfdan, and Kraka Fields

The samples from the n class="Chemical">pan class="Gene">Dan, Halfpn>an class="Gene">dan, and Kraka fields have been used to test and optimize both experimental and statistical methods. In Figure , the results from the PCA based on spectral fitting in the fingerprint region (1800–600 cm–1) are shown. The variables were collected from all the three column types and replicates were used to assess the uncertainty of the analysis.

Figure 4

Score plot of the PCA of the drill core samples from the Dan, Halfdan, and Kraka fields employing the spectral peak-fitting approach in the fingerprint range (1800–500 cm–1). The colored circles indicate the suggested groupings of the samples.

Score plot of the n class="Chemical">pan class="Chemical">PCA of the drill core sampn>les from the Dan, Halfdan, and Kraka fields employing the spectral peak-fitting approach in the fingerprint range (1800–500 cm–1). The colored circles indicate the suggested groupings of the samples.

The samples form three distinctive groun class="Chemical">ps based on the first two PCs. The first PC describes the acid/amide mutual ratio, with samples on the left containing mostly amides, while the ones on the right are rich in carboxylic acids. The second PC is primarily affected by the intensity of the aromatic C=C peak and the sub-band with a maximum near 1680 cm–1.

Figure shows the second apn class="Chemical">proach to the PCA with the same sampn>le set. For this case, data from the four fractions of each sampn>le were used for the analysis, and the full recorded spectral range (excluding the 1850 to 2700 cm–1 range) was included. The PCA results were compared with K-means (n = 5) methods. Overall, the distribution of samples qualitatively agrees with the results of the first approach, with most of the samples maintaining proximities illustrated in the previous figure. For example, the MFB samples group with A-6I and the MFA group with the Nana-1XP wells. However, the indicated loadings suggest that the C–H stretching and bending bands are contributing significantly to the variance of the sample set. Still, the carbonyl stretching band has the highest absolute contribution and is the dominant component of the first PC. This is the reason that most correlations along the first PCs are very similar, while distribution by the second PC is different in many cases. This is to be expected as additional information is introduced into the analysis.

Figure 5

Biplot of the PCA of the drill core samples from the Dan, Halfdan, and Kraka fields performed directly on the selected ranges (3800–2700 and 1850–600 cm–1) infrared spectra of the fractions obtained on the silica columns. The orange vectors on the plot represent the dominant contributions to the PCs. The groups were determined from a K-means cluster analysis.

Biplot of the n class="Chemical">pan class="Chemical">PCA of the drill core sampn>les from the Dan, Halfdan, and Kraka fields performed directly on the selected ranges (3800–2700 and 1850–600 cm–1) infrared spectra of the fractions obtained on the silica columns. The orange vectors on the plot represent the dominant contributions to the PCs. The groups were determined from a K-means cluster analysis.

Figure shows the overlay of the two spectra on the on class="Chemical">pposite n>an class="Chemical">sides of both PCs. The features contributing the most to the PCs are marked on the figure. Evidently, the most distinguishing features are related to the ratios of the stretching modes, that is, branching of the carbon skeleton, and comparative abundance of the oxygen-containing compounds, presented by the carbonyl stretching band at 1713 cm–1 with respect to the CH bending mode. Surprisingly, the absorption of the amide compounds are not among the first five major descriptors, even though the groupings are in good agreement with the fitting-based method.

Figure 6

Full-range (3800–600 cm–1) infrared spectra of the D fractions obtained from the drill core samples from the Dan, Halfdan, and Kraka fields on the opposite sides of the PCA biplot, calculated directly from the spectral data (Figure ). The peaks corresponding to the dominant contributions to the first two PCs are indicated. The axis break excludes the part of the spectrum where the diamond artifact is present.

Full-range (3800–600 cm–1) infrared spectra of the D fractions obtained from the drill core samn class="Chemical">ples from the Dan, Halfn>an class="Gene">dan, and Kraka fields on the opposite sides of the PCA biplot, calculated directly from the spectral data (Figure ). The peaks corresponding to the dominant contributions to the first two PCs are indicated. The axis break excludes the part of the spectrum where the diamond artifact is present.

The overall conclupan class="Chemical">sin>on regarding this set is that samples are sufficiently differentiated based on the chromatographic separation on a n>an class="Chemical">silica column. The preliminary results from the PCA carried out for this set shows that generally the same groupings can be obtained by using either the full spectral range or just the fingerprint region of the infrared spectra.

Apart from the benefit of being fast, the analyn class="Chemical">pan class="Chemical">sis based on full-range spn>ectral data points directly has extra advantages. Un>an class="Chemical">sing the C–H stretching bands can be quite challenging in practice, as polar fractions contain carboxylic groups, that in turn produce a broad and strong absorption in the 2800–3600 cm–1 related to the O–H stretching band of the strongly interacting acid molecules, introducing a baseline tilt under the C–H stretching band. This is a major problem for peak-fitting procedures in this range, but likely can be easily addressed with a point-by-point approach as it could take into account the intensity contribution from the O–H stretching band.

Results and Discussion

The samples from the Valdemar field were analyzed un class="Chemical">pan class="Chemical">sing the finalized method empn>loying only the silica column with four fractions produced, and both PCA approaches were employed. The spectral fitting procedure was performed on the D fraction only, in the 1800–1500 cm–1 spectral range, with the purpose to explore potential significantly different abundances of polar compounds. The direct PCA of the raw infrared spectra was performed on a combination of the C and D fractions, as this approach appeared to produce the most consistent grouping of samples. On the contrary, the A and B fractions did not show any clear correlation and the variance was comparable to the experimental noise.

Examples of the infrared sn class="Chemical">pectra of the D fractions for the samples originating from the Valdemar field are given in Figure . A significant variance in the vibrational features is observed for the sampn>les studied. By observation, it appn>ears that acids, amides and aromatics are abundant in vastly varying amounts with respect to one another.

Figure 7

Infrared spectra of the D fractions separated on a silica SPE column for the samples extracted from the drill cores obtained in the Valdemar field.

Infrared spectra of the D fractions separated on a pan class="Chemical">silica Sn>an class="Chemical">PE column for the samples extracted from the drill cores obtained in the Valdemar field.

A few examples of the n class="Chemical">peak-fitting results used to produce PCA plots are given in Figure . In the most simple case, the spectrum in the desired range is presented by two strong peaks attributed to the carbonyl stretching and the aromatic C=C stretching vibrations. Most of the spectra also show a very distinct extra feature on the main carbonyl peak near 1760 cm–1 (B) most likely attributable to another carbonyl compound. On careful inspection of all the samples side-by-side, changes in the peak intensity allow us to uncover additional spectral features. These are bands at 1683 cm–1 (G), this band appears as a shoulder on the strong carbonyl stretching band, often almost entirely concealed by it, and another band with a center close to 1580 cm–1 (I). The latter is attributed to an aromatic vibration like the 1605 cm–1 (H) band, or potentially, it could be associated with another carbonyl-containing compound class.

Figure 8

Example of the spectral peak-fitting procedure performed on the basis for the PCA of the crude oil extracts. The average peak positions and the attributed notations for Figure are indicated next to the corresponding peaks.

Figure 9

Biplot of the PCA performed for the D fractions of the drill core extracts from the Valdemar field. The blue vectors indicate contribution of the variables, measured from the spectral fitting approach (Figure ), to the first two PCs. Based on the resulting scores, the samples are grouped into the three groups indicated by the circles.

The band at 1683 cm–1 (G) could potentially be attributed to the C=O stretch in the secondary n class="Chemical">pan class="Chemical">amide groupclass="Chemical">n>, quinones, or other carbonyl-containing compounds. The common appearance of the latter in some of the C fractions suggests that the respective compound class has weaker interaction with the silica surface. This feature has qualitative correlation with a signal at 3423 cm–1, favoring the secondary amide assignment; however, this cannot be reliably concluded by means of FTIR alone. Last, a weak band near 1740 cm–1 (C) is observed as a shoulder on the main carbonyl stretching band. As this band is very weak, it only shows up clearly in a few observed samples, where the stronger bands obscuring it are smaller. Unfortunately, there is no clear information about the nature of this band, but because of the fact that it often appears as a distinct shoulder, it is necessary to include this to reduce the error associated with fitting other bands. Naturally, extra care has been taken to produce meaningful peak fits for the bands that are strongly overlapped in many samples, yet it should be noted that uncertainties associated with exact peak positions and areas are somewhat larger from sample to sample. The most important part of the method was to restrict the peak position to within 5 cm–1 of the average value observed, and the fits were reiterated several times.

The bottom spectrum shown in this figure illustrates a case where a detectable amount of n class="Chemical">primary amides is present in the sampn>le. Two signals are characteristic of these compounds, assigned at 1660 (E) and 1632 cm–1 (F), for the C=O stretching and NH2 bending bands, respectively.

The results of the pan class="Chemical">Pn>CA on the fittings are shown in Figure . Loadings vectors correspond to the assignments on Figure . Three groups can be obtained from the analysis. The main PC differentiates the samples based on the presence of primary amides (negative vectors E, F), acids, aromatics, or the G vector due to the band at 1683 cm–1 (positive contributions). Basically, without considering the second PC, the PCA scores show gradual distribution of samples from nitrogen-rich (amides) to oxygen-rich (acids). The second PC is responsible for establishing the third group alone, based on the value of the 1683 cm–1 feature (G). The G vector is anticorrelated with the D (acids), which could be in part because of overlap of these two peaks and the associated uncertainty in estimating the area of the peaks because they share the total model intensity to be reproduced. There appears to be general correlation between the D and I variables as well as between the B and H variables. Perhaps a more detailed investigation complemented by the GC–mass spectrometry techniques might reveal structural dependencies between these spectral features.

Biplot of the pan class="Chemical">PCA performed for the D fractions of the drill core extracts from the Valdemar field. The blue vecpan class="Gene">tors indicate contribution of the variables, measured from the spectral fitting approach (Figure ), to the first two pan class="Chemical">PCs. Based on the resulting scores, the samples are grouped into the three groups indicated by the circles.

The examples of sn class="Chemical">pectra measured for the representative samples from each group are shown in Figure . The sample from the amide group has a compn>aratively small amount of n>an class="Chemical">carboxylic acids, and the aromatic peak is weaker. The samples from the other two groups, on contrary, all show a strong acid band and the aromatic peak at 1600 cm–1. Those two spectra are mostly differentiated by the evident shoulder of the 1683 cm–1 band and different ratios of the CH-bending/acid peaks.

Figure 10

Infrared spectra of the example D fractions obtained from the samples in the three groups, as attributed from the fitting-based PCA approach. The primary differentiation features of the samples are indicated on the spectra as the corresponding peak assignments.

Infrared spectra of the examn class="Chemical">ple D fractions obtained from the samples in the three groups, as attributed from the fitting-based pan class="Chemical">PCA approach. The primary differentiation features of the sampn>les are indicated on the spectra as the correspn>onding peak asn>an class="Chemical">signments.

The results of the analypan class="Chemical">sin>s should be compared against the second approach to the PCA. Figure shows the biplot of the direct PCA on the combined C and D fractions. The comparison of the two plots indicates that the results are similar between the two methods. The variables contributing to the PCs are also more consistent with the fitting method, as both the acid band and one of the amide bands are present among the highest contributors. Apart from those, the CH-bending band and a band in the fingerprint region are important to the score distribution. The samples can be categorized into three to five groups based on the K-means cluster analysis. Owing to the much larger number of variables, the overall distribution of the samples is more gradual compared to peak-fitting method. However, the most significant variation between the samples is properly captured and is not adulterated by the additional input, that is, North Jens-7479 and V2P-7579 are well grouped together in both methods, as well as acid-rich samples, for example, Bo-2X-7901, are well isolated.

Figure 11

Biplot of the PCA of the drill core samples from the Valdemar field performed directly on the full-range infrared spectra of the fractions obtained on the silica columns. The orange vectors on the plot represent the dominant contributions to the PCs.

Biplot of the n class="Chemical">pan class="Chemical">PCA of the drill core sampn>les from the Valdemar field performed directly on the full-range infrared spn>ectra of the fractions obtained on the n>an class="Chemical">silica columns. The orange vectors on the plot represent the dominant contributions to the PCs.

The acids and pan class="Chemical">amidesn> present in the studied sampn>les are some of the most interesting descripn>an class="Gene">tors of the oil properties, as they are capable of engaging in strong intermolecular interactions. These compounds can be expected to bond to other acids and amides, ketones, and, most importantly, to the surface of the reservoir rock. The latter could be the defining factor to the mobility of the oil. Overall, it seems that the difference in these compound classes constitutes the most clear chemical difference in the chemistry of the wells investigated in this study obtained by infrared spectroscopy.

Figure shows the results of quantification of the acid and the pan class="Chemical">amiden> content in the extracted oil of the Valdemar drill core samples. The spectra were normalized by the CH-bending band, which should not have any overlap with the signals of the quantified groups and should also closely represent the reference for the total organic content measured in the ATR approach. This step is necessary, as the penetration depth of the infrared beam in the ATR measurements is dependent on the refractive index of the sample and the wavelength of the light, and is ultimately limited by the thickness of the layer on the diamond crystal of the ATR module. For these reasons, a reference peak located as close as possible to the acid/amide bands is necessary for quantitative applications in our scenario.

Figure 12

Relative abundances of acids (top) and primary amides (bottom) found in the oil wells of the Valdemar field based on their characteristic band intensities (the 1710 cm–1 band for acids and the 1632 cm–1 band for amides) in the spectra obtained for the D fraction of the extract. The spectra were all normalized by the CH2 bending band intensity. Error bars are estimated as the standard deviation of the fitting.

Relative abunpan class="Gene">dann>ces of acids (top) and primary n>an class="Chemical">amides (bottom) found in the oil wells of the Valdemar field based on their characteristic band intensities (the 1710 cm–1 band for acids and the 1632 cm–1 band for amides) in the spectra obtained for the D fraction of the extract. The spectra were all normalized by the CH2 bending band intensity. Error bars are estimated as the standard deviation of the fitting.

The observation from these results is that the acids are present conn class="Chemical">pan class="Chemical">sistently in all of the sampn>les and in significantly higher concentrations than found in any of the produced fluid fractions (Figure ). The concentration of amides varies a lot more across the sample set. The quantity of amides in the Bo-2X wells is comparable to the uncertainty of the fitting procedure. It can be noticed that for most wells, the quantity of acids is inversely proportional to the quantity of amides, for example, all of the N. Jens wells are amide-rich and do not show as much acid content as Bo wells. It could be hypothesized that the approximately 3 times higher recovery factors observed for N. Jens and Valdemar wells (provided in private correspondence with the field operator) could be linked via the physical–chemical properties of the oil to the comparatively low abundance of acidic compounds in these wells. As amides form less stable intermolecular bonds, they might be expected to have less of an effect on fluid–fluid and fluid–rock interactions than acids.

Figure 13

Comparison of the infrared spectra of a produced oil (HBB-07, red trace) and the drill core extract (blue) before the SPE procedure. The spectra are normalized by the CH2-bending band intensity. The characteristic carbonyl stretching absorption is indicated on the graph, evidencing a significant difference in the acid content of the produced fluid and the deposited oil in a drill core before production.

Comparison of the infrared sn class="Chemical">pectra of a produced oil (HBB-07, red trace) and the drill core extract (blue) before the Sn>an class="Chemical">PE procedure. The spectra are normalized by the CH2-bending band intensity. The characteristic carbonyl stretching absorption is indicated on the graph, evidencing a significant difference in the acid content of the produced fluid and the deposited oil in a drill core before production.

An analytical investigation of well performance in several regions of the n class="Chemical">North Sea has been carried out by Nielsen et al.,[30] in which a link between the production of the wells and the pn>resence of biomarkers, retene, and stereoisomers of steranes, was observed. This indicates an impact linked to the thermal maturity of n>an class="Chemical">oil. The chemical information obtained from tight chalk drill cores in this work is indented to be complemented by a future fingerprinting of produced oil samples from the fields, focused on recovering data on maturity, branching, and aromatic condensation.

Conclusions

The apn class="Chemical">plication of infrared spectroscopy has proven powerful and capable of distinguishing the chemical differences between crude oils from tight chalk cores sampn>led from different geographical locations. Statistically n>an class="Chemical">significant differences in the quantities of polar oxygen- and nitrogen-containing compounds, namely, carboxylic acids and amides, were found between the samples. The spectral data can be obtained routinely with the ATR method with less sample preparation and average analysis time than is usually the case with high-resolution GC techniques applied to heavy heteroatom compounds,[31] and the ATR method constitutes an excellent complementary technique to the latter. The samples show distinct intercorrelations and can be grouped with a PCA approach based on the observed infrared spectral signatures of polar compounds. The average composition of the drill core extracts is heavier and more rich in asphaltenes and resins than the produced oil. This suggests that the majority of the acidic compounds and amides are retained in the reservoir after the oil has been produced, and leaving this fraction behind has a detrimental impact on the recovery factor, as well as potentially negatively influencing the produced fluid flow throughout production. The strongest contributions in the statistical evaluation of the sample set come from the acid/amide ratio, with a comparatively higher amide presence found in the North Jens wells. The second PC of the set is mostly characterized by the ratio of 1683 cm–1 band and the respective features related to the primary amide content. The strong composition discrepancy between the retained material and the produced oil also creates some potential issues when changing the well flooding fluid without accounting for the changes in the chemical composition of the produced fluid.