Molecular Discrimination of Geological Formations in Reservoir Samples Using ESI(-) FT-ICR MS and Supervised Modeling

Understanding the compositional heterogeneity of reservoir fluids is crucial for the exploration and development of petroleum fields. Variations in crude oil molecular composition reflect the geological history of source rocks as well as migration, accumulation, and post-accumulation alteration processes. Traditional geochemical analyses have largely focused on the saturated and aromatic fractions of crude oil, often overlooking the polar fractions, which can represent up to 20% of the total composition. In this context, petroleomics, through ultrahigh-resolution mass spectrometry (FT-ICR MS), has emerged as a transformative approach, enabling the detection of thousands of molecular formulas, especially those containing nitrogen and oxygen.

This study investigates the molecular differentiation of crude oils from two distinct geological formations (referred to as A and B) using FT-ICR MS with negative electrospray ionization (ESI−), combined with supervised statistical modeling. A total of 83 crude oil samples from a Brazilian field—covering four geological formations—were analyzed. Sample preparation involved dissolving 1 mg of oil in toluene, followed by dilution in methanol and the addition of ammonium hydroxide (NH₄OH). Analyses were conducted in triplicate using a SolariX 7T FT-ICR mass spectrometer (Bruker), ensuring data reproducibility.

Mass spectra were recalibrated and processed using Composer software, and replicate alignment was done using in-house tools. Statistical modeling relied on partial least squares discriminant analysis (PLS-DA) and the ordered predictors selection for discriminant analysis (OPSDA) method to identify key variables. Model performance was assessed based on accuracy and sensitivity.

FT-ICR MS data in ESI(−) mode led to the identification of 4,495 unique molecular formulas. Among the detected chemical classes, those containing nitrogen (N, N₂) and oxygen (O, O₂, NO) were predominant in both abundance and frequency. Detailed analysis of carbon number and double bond equivalents (DBE) within these classes revealed formation-specific molecular patterns. Oils from Formation A showed a higher relative abundance of high-carbon-number compounds (≥ C23), suggesting greater molecular complexity or more advanced thermal maturity compared to Formation B.

Supervised modeling demonstrated excellent performance in differentiating the oils by geological origin. PLS-DA models based on ESI(−) data achieved high classification accuracy and sensitivity. Additionally, OPSDA enabled the selection of the most significant molecular variables for discriminating between formations. The 13 most relevant abundance ratios, illustrated in Figure 1, consistently differentiated between Formations A and B, underscoring the importance of specific molecular pairs. Figure 2 further supported these findings with class-based separation plots showing the contribution of nitrogen- and oxygen-containing species to the classification performance.

Relative distribution analyses of chemical classes across the samples also revealed informative trends. Nitrogenous classes, especially N and N₂, were dominant across all samples. Ternary diagrams showed subtle yet meaningful compositional variations that aligned with the observed geological separations. These results, along with the robust classification performance of optimized molecular ratios, demonstrate the strength and flexibility of the applied methodology—even in geologically complex reservoirs.

In conclusion, combining FT-ICR MS with ESI(−) ionization and advanced supervised modeling techniques like PLS-DA and OPSDA enabled a clear and reliable differentiation of crude oil samples from distinct geological formations. The molecular-level insights provided a refined understanding of oil composition, uncovering specific markers unique to each formation—even when traditional metrics such as DBE or carbon number offered only minor distinctions.

The identification of these molecular signatures contributes to a better understanding of oil generation, migration, and alteration processes, offering valuable input for integrated reservoir management and enhanced recovery strategies. The demonstrated ability to differentiate geological formations based on oil composition opens avenues for more complex applications, such as fluid tracking in multi-zone production systems and monitoring changes induced by CO₂-WAG injection processes. Overall, this study highlights the potential of molecular petroleomics as a powerful tool for advanced reservoir characterization, bridging fundamental science with practical applications in the oil and gas industry.