Analytical Chemistry Solutions to Hazard Evaluation of Petroleum Refining Products

Abstract

Products of petroleum refining are substances that are both complex and variable. These substances are produced and distributed in high volumes; therefore, they are heavily scrutinized in terms of their potential hazards and risks. Because of inherent compositional complexity and variability, unique challenges exist in terms of their registration and evaluation. Continued dialogue between the industry and the decision-makers has revolved around the most appropriate approach to fill data gaps and ensure safe use of these substances. One of the challenging topics has been the extent of chemical compositional characterization of products of petroleum refining that may be necessary for substance identification and hazard evaluation. There are several novel analytical methods that can be used for comprehensive characterization of petroleum substances and identification of most abundant constituents. However, translation of the advances in analytical chemistry to regulatory decision-making has not been as evident. Therefore, this review’s goal is to bridge the divide between the science of chemical characterization of petroleum and the needs and expectations of the decision-makers. Collectively, mutual appreciation of the regulatory guidance and the realities of what information these new methods can deliver should facilitate the path forward in ensuring safety of the products of petroleum refining.

Introduction

Crude oils are naturally occurring and highly complex substances which vary considerably in molecular composition according to their origins; they comprise a myriad of constituents, primarily hydrocarbons, but also other organic and inorganic molecules (Smith et al., 1959). Close to one hundred billion barrels of crude oils are annually extracted, distributed, and processed into a wide variety of refined petroleum products (Kaiser, 2017; Salvito et al., 2020). The chemical composition of petroleum refining products therefore depends on both the type (i.e., origins) of crude oil from which it was derived, and the refining process (i.e., fractional distillation and/or cracking followed by additional processing through solvent extraction, hydro-desulfurization, or hydrogenation) used to meet performance characteristics of the end-products (McKee et al., 2015). Products of petroleum refining are high production volume substances and thus are heavily scrutinized in terms of their potential human and environmental health hazards and risks. Because of inherent compositional complexity and variability, petroleum substances are prototypical representatives of a diverse class known as substances of unknown, variable composition, complex reaction products, or biological materials (UVCBs); these substances present unique challenges to regulatory agencies, especially in terms of characterization of their chemical composition (Clark et al., 2013; ECHA, 2017c; Lai et al., 2022). It is worth noting that the different fields within the academic, industrial, and regulatory science communities can use differing terminology. Analytical researchers often refer to petroleum refining products as “hydrocarbon mixtures” which are, in turn, part of the broader “complex mixtures” family of samples. The term “mixture” is avoided by the industry and decision-makers as they reason that most substances in commerce that are made from oil are products of refining, rather than mixing, and thus to differentiate from the mixtures found in the environment, such substances are called petroleum UVCBs.

Studies of molecular composition of crude oils and petroleum refining products have a long history spanning over 80 years (Figure 1). The analytical characterization of petroleum substances historically tracked the physico-chemical properties that pertained to the functionality of the product, such as flash points and vapor pressure. With the advent of spectroscopy and mass spectrometry techniques, there came the possibility to gradually gain more detailed understanding of composition; however, the granularity of information on the constituents in registered petroleum products is still lacking. Recent improvements have been made in the resolution of mass spectrometers, ionization methods in order to access a wider range of components, separation methods to offer structural insights, and developments in data visualization through standardized diagrams (Palacio Lozano et al., 2020). As a result of these advances, new opportunities emerged to provide comprehensive characterization of these complex substances and satisfy regulatory needs on the composition, quantity of potentially hazardous constituents, and the extent of variability among manufacturing batches of these products. The range and types of mass spectrometry techniques that can be used for the analysis of petroleum-related samples is quite extensive; collectively, the methods for study of petroleum are now being referred to as “petroleomics,” a sub-field of analytical chemistry aiming to identify the totality of constituents of crude oil and petroleum refining products using high resolution mass spectrometry methods (Hsu et al., 2011; Marshall and Rodgers, 2004; Palacio Lozano et al., 2020).

Figure 1.

A timeline of major developments in the fields of analytical chemistry and data analysis of petroleum UVCB, and the concomitant evolution of the regulatory frameworks for registration and hazard classification of these substances See abbreviations in the text.

Despite major advances in the ever-improving analytical resolution of individual molecules and their classes in oils and complex petroleum UVCBs (Wise et al., 2022), there has been relatively little use of the data from these new methods and instruments in regulatory submissions, with the exception of GC×GC-FID-derived data (Redman et al., 2014; Ventura et al., 2011), or even their mention in the reviews or original research publications (Figure 2A). The naming conventions and approaches to identification of complex petroleum UVCBs remain rather imprecise (Rasmussen et al., 1999); only general compositional characteristics are used to define broad manufacturing categories (Salvito et al., 2020). While such information is generally sufficient for naming and identification of petroleum UVCBs (ECHA, 2017a), it is often not sufficient for evaluation of human health and environmental hazards, a prerequisite to registration and authorization for their use (Salvito et al., 2020).

Figure 2.

Literature review of the major analytical methods and their application for the analysis of petroleum substances. (A) A dendrogram of the major searches. The numbers indicate the quantity of publications for each search. See Supplemental Table 1 for information on the exact search terms and hyperlinks to the publications. (B) Cumulative histograms indicating the number of publications across time through December 2021. Colors correspond to the methods indicated in the inset.

The regulatory science and analytical chemistry fields run in parallel and both are highly specialized, requiring significant expertise. Consequently, intricacies of the legislative mandates governing regulatory decision-making are often unfamiliar to the researchers who develop and refine advanced methods for petroleomics. Similarly, decision-makers may not be aware of the latest opportunities that analytical chemistry has to offer. Collectively, there is a considerable gap in the translation of knowledge from petroleomics-focused analytical laboratories to applied decision-making. This review aims to first summarize the regulatory guidance for characterization of the chemical composition of petroleum UVCBs and then demonstrate how existing petroleomics techniques could be applied to address these needs. Recent additions to the European Union (EU) Registration, Evaluation, Authorisation and Restriction of Chemicals (REACH) regulatory guidance for demonstrating the composition of complex UVCB (ECHA, 2022), together with recent advancements in petroleomics applications that are poised to address these needs (Palacio Lozano et al., 2020), create a unique opportunity to bridge the divide. Here, we highlight the opportunities that are already within reach for using modern analytical and data analysis/visualization techniques for impactful decisions on petroleum UVCBs. We reason that it is imperative for decision-makers to be aware of the possibilities and limitations of current analytical approaches so that regulations can be sufficiently strict, yet realistic in terms of their attainment using best available science.

What Information do the Decision-Makers Seek on Petroleum UVCBs?

Most impactful guidance to the industry on petroleum UVCBs has been produced in the United States and the European Union. These include the United States Environmental Protection Agency (US EPA) High Production Volume (HPV) Challenge Program, a voluntary industry-government information sharing effort that was launched in 1998 (Petroleum HPV Testing Group, 2017), and the REACH legislation and associated guidance documents in the European Union (ECHA, 2017c) which is now being adopted in other countries around the globe. Because REACH is the most recent and stringent legislative regime, it has become a de facto global driver for the regulatory scrutiny of both new and existing chemicals, including petroleum UVCBs. Therefore, most decision-making contexts discussed herein pertain to EU REACH regulation and its implementation by the European Chemical Agency (ECHA) through guidance documents.

Simply put, REACH-based guidance for UVCBs states that data provided by the registrants shall enable (i) identification of the substances that are submitted for registration, and (ii) evaluation of the potential hazards to human health and the environment (ECHA, 2017c). In both instances, information on chemical composition of a substance is required; however, in a slightly different context (Table 1). For the former, data requirements are typically less stringent because complex substances may be registered based on the manufacturing process, intended use, and/or physical-chemical properties. For the latter, the molecular identification (both elemental composition and structure) data requirements are far greater because the individual constituents that may be present in a “representative” sample, their amounts, and presence of known or suspected hazardous substances must be reasonable ascertained to enable grouping and read-across among petroleum UVCB substances. Based on these two initial components of REACH, an EU regulation that came into force in June of 2007 (European Council, 2007), decisions on authorization of use or restriction(s) are made. An additional challenge for petroleum-derived and other UVCBs is that their composition is inherently variable from batch to batch of the nominally the “same” product; this is also true for product-to-product variability within broader categories (FuelsEurope, 2015). This variability may be due to the source crude oil used in manufacturing; petroleomics-related research has revealed significant differences in compositions of petroleum according to its origins, slight variations in manufacturing processes among refineries, and naturally occurring degradation and weathering processes. Because safety testing is typically conducted with a “representative” sample of a UVCB product or category, the confidence with which these data can be extrapolated to other samples of the same product or to other products that are “similar,” especially when sample composition data may be commercially sensitive, depends on how much and what type of data is available to ascertain such similarity.

Table 1.

Summary of the chemical characterization needs for UVCB that are registered and evaluated under REACH in the European Union.

Information Needed	Substance Identification for Registration (REACH Annex VI, Section 2)	Hazard Characterization for Evaluation (REACH Annex XI, Section 1.5)
		Read-Across Assessment Framework: Considerations on multi-constituent substances and UVCBs (ECHA, 2017c)		Advice on Using Read-Across for UVCB substances (ECHA, 2022)
		Category Approach	Analogue Approach
Overall Substance Identification	Substance name [from a trade association and/or nomenclature system] Substance identifiers [source, manufacturing process, carbon and boiling range, phys-chem properties, etc.]	Substance(s) to be grouped in a category Compositions to be included Manufacturing process description	Source substance(s) and target substance Compositions to be included Manufacturing process description	Similarity may be established based on: (i) presence of identical constituents OR (ii) variation in concentration and variability in constituents Constituents present at >1% must be identified; lower thresholds if constituents of concern are present (>0.1% for carcinogenic/mutagenic, >0.3% for repro/developmental) >80% of constituents in the substance must be identified
Composition Characterization	Constituents present >10% Constituents <10% that may be impacting hazard classification	Category “domain” needs to be defined (constituent-specific concentration determination not specified)	Source substance(s) Target substance (constituent-specific concentration determination not specified)	Must be characterized up to 100% When full characterization is impractical/impossible need to provide (i) justification AND (ii) demonstration of similarity “by other means”
Variability	Not Required	Structural similarity for category based on worst-case scenario Determine if quantitative differences or patterns in predicted properties may be reflected in structural similarity	Structural Similarity between source & target based on worst-case scenario Determine if quantitative differences or patterns in predicted properties may be reflected in structural similarity	Structural similarity explained based on quantitative and qualitative comparison of composition When full characterization is impractical/impossible need to provide (i) comparison of constituents AND (ii) demonstration of similarity “by other means” (e.g. analytical information for >95% constituents, constituents of high concern, high resolution for confidence fingerprinting) Analysis of at least 5 independent (i.e., production batches) samples analyzed from ALL registrants of a substance

For the purpose of substance identification, the regulatory frameworks in the United States and the European Union have historically (Figure 1) named, grouped and categorized petroleum UVCBs based on the manufacturing processes used in oil refining, as well as physical-chemical properties and other broad chemical fingerprinting data (Dimitrov et al., 2015; ECHA, 2008; Rasmussen et al., 1999). Manufacturing process-centric naming conventions for these complex substances were originally developed by the American Petroleum Institute (API) and the US EPA for the purpose of creating an inventory of petroleum products under the Toxic Substances Control Act inventory (API, 1983; EPA, 1995; U.S. EPA, 1978). Both Chemical Abstract Service (CAS) and European Inventory of Existing Commercial Chemical Substances (EINECS) identifications have been assigned to a large number of petroleum UVCBs, even though the descriptions of each of these substances under either one of these “unique identifiers” are rather imprecise and far from being unique (Rasmussen et al., 1999). The broad substance categories were somewhat refined under the US EPA’s HPV Challenge Program through addition of more detailed information on physical-chemical properties, as well as some human and environmental hazards data (Petroleum HPV Testing Group, 2017).

Far greater information requirements, with respect to both chemical composition and potential hazards, were imposed by REACH (European Council, 2007). Petroleum UVCBs are high production volume substances that were subject to the earliest deadline for registration and the most stringent requirements for hazard evaluation. From 2007 to 2010, about 8,000 registrations were submitted in the EU for petroleum substances that were produced or imported at >100 tons/year (CONCAWE, 2022). Subsequently, the number of registrations of petroleum products was reduced to 191 substances through consolidation of redundant submissions and further grouping of substances deemed to be “similar” based on a variety of considerations. To put the scale of the challenge in context, tens of thousands or even hundreds of thousands of unique molecular formulae can be observed in a single fraction of a petroleum sample using ultrahigh resolution mass spectrometry, indicating that potentially millions of different structures are present in individual crude oils when allowing for isomers (Palacio Lozano et al., 2019b; Palacio Lozano et al., 2020). While the registration submissions for petroleum UVCBs were completed more than 10 years ago and are regularly updated (CONCAWE, 2021), discussions between trade associations and regulatory agencies are ongoing to determine the most sensible ways to improve the dossier quality and ensure the information is compliant with the REACH regulation, primarily by generating more testing information and reinforcing read-across and category approaches. Still, industry’s attempts to waive animal testing requirements through read-across have been rejected by ECHA because of considerable data gaps in hazard assessment and compositional characterization (ECHA, 2020a; ECHA, 2020b; ECHA, 2021).

For registration under REACH (Annex VI, Section 2), the data should be sufficient to enable substance identification (Table 1). For petroleum substances, ECHA guidance is that the following data should be provided: (1) accepted nomenclature; (2) appropriate identifiers such as source/feedstock, refining history, boiling and carbon number ranges, physio-chemical characteristics, chromatographic or spectral information, flash point, and viscosity; and (3) compositional information including identification and concentration of the individual constituents present at >10% and that are known to be hazardous, persistent and/or bioaccumulative, identification of any additives, and generic description of unknown constituents (CONCAWE, 2012; ECHA, 2017a). Additionally, Articles 7(2) and 33 of REACH have defined a concentration threshold of 0.1% w/w for constituents classified as “substance of very high concern” (ECHA, 2017b). Based on these guidance documents, the registrants (companies or trade associations) have traditionally relied on a wide range of analytical techniques (Figure 2A) to furnish the information on petroleum substance identification (Clark et al., 2013; CONCAWE, 2012; CONCAWE, 2014; CONCAWE, 2020).

For the evaluation step, REACH specifies (Annex XI, Section 1.5) that substances may be grouped based on “structural similarity between substances which results in a likelihood that the substances have similar physicochemical, toxicological and ecotoxicological properties so that the substances may be considered as a group or category.” Next, prediction of possible hazards of data-poor substances are made through the application of “read-across” from an analogous substance that has been tested in a requisite assay and is deemed “similar.” For this step, REACH regulation states that “it is required that the relevant properties of a substance within the group may be predicted from data for reference substance(s) within the group (read-across approach).” The registrant shall establish a read-across hypothesis which explains what structural similarities or differences exist between the source and target substance(s) to which read-across is applied and why a prediction for a toxicological or ecotoxicological property can be made with confidence.

Recently, REACH regulation (Annex XI, Section 1.5) has been amended to state that for the application in grouping, “structural similarity for UVCB substances shall be established on the basis of similarities in the structures of the constituents, together with the concentration of these constituents and variability in the concentration of these constituents” (European Commission, 2021). Once such information becomes available, and in cases where structural differences are present between the source and target substances, the read-across hypothesis should explain why the differences in the chemical structures within a group will not influence the toxicological or ecotoxicological properties or may do so in a regular pattern.

Even though the regulatory language above may seem rather straightforward, in practice the bar on establishing “structural similarity” is very elusive in the case of petroleum UVCBs. It is widely acknowledged that the chemical complexity of petroleum substances far exceeds the capabilities of any one method, even the highest resolution mass spectrometers; therefore, a combination of techniques and approaches is often employed. However, despite the considerable scientific advances achieved in the past decade (Figure 1) in both molecular separation (encompassing chromatography, spectrometry and ionization) and detection (various modalities of mass spectrometry, flame ionization detection, and spectroscopy), many challenges with precision and confidence in comprehensive molecular characterization of petroleum samples persist.

For example, a recent study using Fourier transform ion cyclotron resonance mass spectrometry (FT-ICR MS) revealed nearly a quarter of a million molecular formulae in a fraction from one petroleum sample (Palacio Lozano et al., 2019b). This study is an example of the reality that millions of structures (as opposed to formulae, as structure also determines toxicity) are present in petroleum UVCBs, an intractable challenge both for the chemical analysts to resolve and identify, as well as for the decision-makers to evaluate. Indeed, even the REACH regulation itself (Annex XI, Section 1.5) acknowledges that “if it can be demonstrated that the identification of all individual constituents is not technically possible or impractical, the structural similarity may be demonstrated by other means, to enable a quantitative and qualitative comparison of the actual composition between substances” (European Commission, 2021). Clarification regarding the definition of “other means” was recently released May 2022 (ECHA, 2022), where specific justification for lack of knowledge about constituents, lack of published methods to identify constituents, and explanation about technical hindrance to resolution and identification of constituents comprising >20% of the substance is necessary to apply such means for characterization. Where “other” means are justified, quantitative comparison of constituents in common between source and target substances should be included, as well as a qualitative comparison of structures that vary between the substances. Fingerprinting, for example, can be used if the following are addressed: information on >95% of all constituents, information on the constituents of high concern, and high analytical resolution to enable accurate alignment, quantitation, variability, and structural data (beyond molecular formulas) of constituents between substances (ECHA, 2022). Therefore, the subsequent sections of this review are framed around three overarching critical needs/questions that REACH regulation challenges the registrants of petroleum UVCBs to address to gain regulatory acceptance of the grouping and read-across hypotheses that are addressable by means of standardized analytical chemistry methods:

• Critical Need 1: Providing detailed information on the structure of the constituents;

• Critical Need 2: Providing information on the concentration of the individual constituents; and

• Critical Need 3: Demonstrating compositional similarity of complex petroleum UVCBs through other means when the identification of all individual constituents is not technically possible or impractical.

Conventional Methods for Characterization of Petroleum Substance Identity and Composition

There are many methods for characterization of physical-chemical properties and chemical composition of oil, that have also been applied to petroleum UVCBs (Figure 2A). Multiple techniques are used due to the broad range of substances with widely different composition, volatility, and polarity (CONCAWE, 2014; Stout and Wang, 2007; Wang et al., 2011). Most publications concerning characterization of petroleum UVCBs have explored approaches to define elemental composition, physical properties, and gross structural information using nuclear magnetic resonance (NMR) or infrared spectra of these substances. The elemental analyses for characterization of petroleum UVCBs evaluate the concentrations of major elements ranging from carbon and heteroatoms to metals (CONCAWE, 2019). There are specific methods for assessing physical properties, which typically relate to the quality of the product and are thus more regularly measured during manufacturing and more available. For example, measurements of specific gravity using American Society for Testing and Methods (ASTM) methods such as ASTM D287 Standard Method for API Gravity of Crude Petroleum and Petroleum Products are in wide use (Giles, 2016). Other physical-chemical information, such as boiling and carbon number ranges, is typically deduced through physical (e.g., ASTM methods D86, D1160 and D2892), or simulated (CONCAWE, 2019; CONCAWE, 2020) distillation methods.

Spectroscopic techniques have been widely employed to obtain broad compositional information for regulatory characterization and identification of UVCB substances but their utility for the analysis of petroleum UVCBs has been questioned (CONCAWE, 2020). NMR methods (IP392, ASTM D5292) measure the percent of carbon or hydrogen atoms in an aromatic ring (CONCAWE, 2020). Infra-red spectroscopy measures the presence of functional groups to define the degree of saturation in the constituents (CONCAWE, 2012). Ultra-violet spectroscopic analysis quantifies compounds by detecting unsaturated bonds such as those in olefins and aromatics, as well as ketonic and heteroatom groups, but is limited in resolution for other constituents.

More detailed compositional information, which gives greater insight into the chemical classes and carbon ranges of the substance, is obtained using chromatographic techniques that enable separation of constituent groups in complex petroleum UVCBs, these include gas- and liquid-based approaches. Gas chromatography (GC)-based analyses predominate; gas-based analysis of hydrocarbons were first published in the 1960s leading to the development of a standardized method (ASTM 2887–84) for determination of Boiling Range Distribution of Petroleum Fractions by Gas Chromatography (Giles, 2016). GC is a powerful tool used for the separation and semi-quantitative assessment of non-polar constituents such as hydrocarbons and polycyclic aromatic hydrocarbons (PAH) (CONCAWE, 2012). It offers better separation than liquid chromatography (LC) but is limited by the boiling point of compounds (affecting the accessible mass range), which is a greater hindrance for the characterization of the heavy petroleum that is increasingly relied upon, and sometimes compounds must be derivatized to ensure GC compatibility. LC is dependent on the polarities of the constituents present, predominantly used to characterize less volatile polar compounds. High-performance liquid chromatography (HPLC; e.g., ASTMD6379 and IP391 methods) is used to quantify mono-, di- and tri-aromatic hydrocarbons (CONCAWE, 2012; CONCAWE, 2019). Meanwhile, thin layer or liquid column chromatography (ASTM D2007) separation generates information on basic chemical properties (CONCAWE, 2012; CONCAWE, 2019). Gas chromatography coupled to mass spectrometry detection (GC-MS) is widely used in forensic fingerprinting (US EPA 8270 and 8051B) and to characterize the composition of petroleum UVCBs (US EPA, 1996; US EPA, 2014). Flame ionization detection (FID) has been coupled with both LC and GC for the detection and quantification of hydrocarbons (CONCAWE, 2012).

However, conventional standardized methods detailed above are insufficient to establish truly comprehensive compositional characterization of UVCBs as needed by REACH to accept grouping and read-across hypotheses from registrants. Recent decisions by ECHA on testing proposals provided several reasons as to why that is the case (ECHA, 2020a; ECHA, 2020b; ECHA, 2021). In these decisions, ECHA noted that (i) physical-chemical characterization of whole complex substances does not demonstrate similarity of chemical constituents of these substances; (ii) elemental and other traditional analysis methods do not provide information on the identity and concentration of individual chemical constituents, but rather provide physical-chemical characterization of the substance as a whole; and (iii) standard methods used in the submissions provided insufficient information to estimate the variability of constituents both within and among substances and groups. ECHA deems these criteria necessary to establish the applicability domain of a given category to confirm membership of the source substance(s) and enable subsequent read-across to the target substance(s) (ECHA, 2020a; ECHA, 2020b; ECHA, 2021).

Collectively, despite the use of a battery of analytical assays, expending large sample quantities on some of these analyses, using specialized sample preparation techniques, and incurring considerable costs to acquire these data, the registrants did not establish substance characterization that would be acceptable by ECHA. Indeed, the industry itself acknowledges that spectroscopic techniques are limited to bulk characterization and “most substances [i.e., petroleum UVCBs] cannot be effectively differentiated from each other by UV, IR, ¹H-NMR or ¹³C-NMR spectroscopies” (CONCAWE, 2020). Further, even though chromatography-based methods provide considerable amount of information for characterization of nonpolar and relatively volatile compounds (e.g., ASTM D2134, D6729, D6730, among others (CONCAWE, 2012), as well as aliphatic and aromatic fractions (Reddy and Quinn, 1999; Wang and Fingas, 2003), their limited resolution leaves much of the complex substance uncharacterized (Wang et al., 2011; Weng et al., 2015). When assessing the needs for analysis of complex substances, it is important to consider that different analytical approaches are complementary and bring their own advantages and disadvantages. With the limitations of many of the more routine methods, there has been increasing use of ultrahigh resolution mass spectrometry and other advanced methods.

More recently, two-dimensional GC (GC×GC) technique has been applied to petroleum substances because it allows for an even greater separation of the multitude of constituents. Specifically, the coupling of GC×GC with FID allows qualitative constituent information at the level of carbon number and chemical class (ASTM International, 2011); this has been useful for simplifying the composition of petroleum substances by binning molecules by “hydrocarbon block” (Redman et al., 2012). Coupling GC×GC with mass spectrometry provides more structural information on the specific constituents (Jennerwein et al., 2014; Mao et al., 2009).

High and Ultrahigh Resolution Mass Spectrometry Techniques

Advancements in mass spectrometry over the past 80 years (Figure 1) have spawned the application of high-resolution approaches for the study of petroleum substances at a molecular level (Palacio Lozano et al., 2019a; Wise et al., 2022). Resolving power is one of the key performance metrics of any mass spectrometry and is typically defined as $\frac{m}{\Delta m}$, where m is the m/z of the ion of interest and Δm is the width of the peak at half its height, using the full width at half maximum (FWHM) definition. In essence, the higher the resolving power, the more peaks can be observed for complex samples, as this reduces overlap of peaks (Phillips et al., 2022). “High resolution” has typically been accepted to mean a resolving power of >10,000 (Xian et al., 2012), but it has become well-established over the past two decades that “ultrahigh resolution,” which often refers to a resolving power of >100,000, is essential for characterization of the most complex samples such as petroleum.

The second performance metric that should always be considered is that of mass accuracy. An instrument that offers high mass accuracy indicates that it typically provides data with low mass errors. When an elemental composition (molecular formula) has been assigned to an observed peak, the observed m/z and the theoretical m/z for the assignment can be used to calculate the mass error, measured in parts per million (ppm), using the following equation:

$$\frac{m_{\mathit{\text{observed}}}-m_{\mathit{\text{theoretical}}}}{m_{\mathit{\text{theoretical}}}}\times 1,000,000$$

Note that a negative mass error indicates the peak appears at a lower m/z than the theoretical value and a positive mass error indicates the peak appears at a higher m/z than the theoretical value; for each peak, the closer the mass error is to zero, the greater the confidence in the given assignment of the molecular formula. It is also important to note that, while researchers desire resolving power to be as high as possible, they strive to keep mass errors as low as possible. Thus, the highest resolving powers afford researchers the ability to observe more components within complex samples, while the highest mass accuracies (lowest mass errors) afford greater confidence in the assignments of elemental composition and structures, together establishing detailed compositional “profiles,” “fingerprints,” or “signatures” for complex samples. Within the field, there has been discussion about the most appropriate terminology to use, drawing parallels with how fingerprints do not normally change but that signatures do, and how this understanding may be related to compositions of complex substances changing when subject to anthropogenic or environmental processes.

The comprehensive characterization of molecular composition of petroleum through high resolution mass spectrometry is an active area of investigation and includes several approaches to precise detection, naming, and structural characterization of the individual constituents (Hsu et al., 2011; Marshall and Rodgers, 2004; Niyonsaba et al., 2019; Palacio Lozano et al., 2019a; Palacio Lozano et al., 2019b; Palacio Lozano et al., 2020; Roman-Hubers et al., 2022; Xian et al., 2012). Modern time-of-flight (TOF) mass spectrometers, which are widespread and considered high resolution, offer resolving powers typically in the range of 10,000–60,000. By contrast, FT-ICR MS is the highest performance variety of mass spectrometer and is considered to offer researchers ultrahigh resolution, at one or two orders of magnitude higher performance. FT-ICR MS is based upon ions orbiting inside a cell, which is in turn housed within the bore of a superconducting magnet, and the technique offers ultrahigh resolving power (~10⁶ FWHM) and mass accuracy (sub-ppm). Orbitrap mass spectrometers are a newer variety of mass spectrometer based upon a Kingdon trap design, rather than using magnetic fields; these instruments typically offer a resolving power of ~10⁵ FWHM. FT-ICR MS and Orbitrap MS offer differing degrees of ultrahigh resolution, with FT-ICR MS offering the highest performance; however, the advantage of TOF MS is in rapid acquisition time which allows for coupling with additional separation techniques such as two-dimensional gas chromatography and ion mobility spectrometry (Palacio Lozano et al., 2019a). These three analytical techniques (Figure 2B) are actively used in petroleomics analyses because they offer somewhat different approaches to determining molecular formulae present in complex substances (Palacio Lozano et al., 2019b; Rodgers and McKenna, 2011), but have not yet been used for detailed petroleum substance characterization for registration or evaluation purposes.

Ultrahigh Resolution MS Data Processing and Visualizations

Petroleum substances contain highly homologous series of hydrocarbon molecules; thus, complex substance analysis can be facilitated by exploiting the patterns of various chemical groups. The method of Kendrick mass defect (KMD) analysis facilitates sorting molecules into homologous series (Kendrick, 1963). The composition of complex petroleum substances can be visualized using the KMD approach because most molecules belong to homologous series comprised of (CH₂) alkyl groups and other functional groups (Figure 3A), this method has been widely used in petroleomics (Hughey et al., 2001; Marshall and Rodgers, 2004; Marshall and Rodgers, 2008; Palacio Lozano et al., 2020). Due to the high resolution and mass accuracy, the molecular composition assigned to the ions that fall in or out of the homologous series can be used to predict their elemental content (HC_#, O_#, N_#, O_#, S_#), carbon number, and double bond equivalents (DBE = C_# − H_#/2 + N_#/2 + 1). A number of other visualizations have been proposed to express the molecular composition of various substances based on the rings and double bonds in the carbon framework (i.e., DBE) of the constituents that can be plotted against their carbon number (Figure 3B). Van Krevelen diagrams (Figure 3C) are used to display the degree of aromaticity and oxidation of constituents by plotting the H/C versus O/C ratio of the organic compounds in the complex substance (Kim et al., 2003; Van Krevelen, 1950; Van Krevelen, 1984). Relative abundance of various classes of compounds in the samples is typically plotted as a chemical class bar chart (Figure 3D), or as two-dimensional “hydrocarbon blocks” (Figure 3E).

Figure 3.

Example visualizations commonly used to represent hydrocarbon composition of petroleum substances analyzed using high resolution MS techniques. (A) A Kendrick mass defect (KMD) plot demonstrating repetitive patterns of CH₂-containing molecules in a petroleum sample. (B) A DBE vs carbon number plot indicating the relative proportions of molecules varying by their degree of aromaticity and carbon number in a sample. (C) van Krevelen plot display the degree of oxidation by plotting the H/C versus O/C ratio in a sample. (D) A stacked bar plot showing relative proportions of constituents from various chemical groups. (E) A plot of relative amounts of various hydrocarbon blocks in a sample.

Molecular-level analyses using ultrahigh resolution MS typically generate voluminous datasets even after processing by the software provided by instrument manufacturers. A number of third-party software packages have been developed recently to facilitate data analysis and visualizations. Peak-by-Peak fee-based software (Spectroswiss, 2019) is tailored for processing FTMS raw data on transients and mass spectra and generates output files for follow-up statistical analyses. PetroOrg (Riches et al., 2015) and Composer (Sierra Analytics, 2022), launched in 2014 and 2008, respectively, are two stand-alone fee-based packages that allow processing and visualization of the molecular composition of petroleum substances data acquired through high resolution MS. UltraMassExplorer is an open source web-based package that uses R Studio (Leefmann et al., 2019). It applies static formula libraries for molecular formula assignment of the molecular formulas based on neutral masses coupled to PubChem searches for putative structural assignment. Data visualization is enabled by van Krevelen, KMD and DBE plots. DropMS is another web-based tool that facilitates high resolution MS data processing and molecular assignments with its corresponding DBE, error, signal intensity, as well as a number of visualizations (Rosa et al., 2020). KairosMS utilizes an R Shiny interface to process complex data sets produced through hyphenated MS experiments that, when combined with software for formula assignments, can be used for visualization, comparison, and statistical analyses for both direct infusion and hyphenated data set, using a wide variety of approaches (Gavard et al., 2020). Interactive van Krevelen (Kew et al., 2017) and Open van Krevelen (Brockman et al., 2018) are two packages that offer interactive diagrams for molecular-level exploration of the data from high resolution MS.

The Regulatory Needs for Characterizing Chemical Composition of Petroleum UVCBs

General considerations and examples of the application of high resolution MS for the analysis of petroleum substances have been reviewed elsewhere (Hsu and Shi, 2013; Niyonsaba et al., 2019; Palacio Lozano et al., 2019a; Palacio Lozano et al., 2019b; Palacio Lozano et al., 2020; Rodgers and McKenna, 2011; Xian et al., 2012). However, these previous reviews did not specifically place the application of these techniques in the context of the regulatory needs for registration and evaluation of petroleum UVCBs. Comprehensive characterization, or at least more detailed information on some specific types of constituents, of complex substances based on high resolving power and mass accuracy of FT-ICR MS, Orbitrap MS and TOF MS affords the opportunity to attain valuable information on both molecular composition and relative abundance of the constituents. Such data can also be used to quantitatively evaluate both variability within a substance or group of substances, and similarity between substances. Then, it should be possible to conduct read-across and complete hazard characterization. To provide specific examples of how each technique can be used to address specific regulatory needs, we performed a systematic literature search (with end date of December 2021) focused on the application of each technique to petroleum substances (Figure 4). We found that even though there are many dozens to hundreds of publications on each technique in general, and on their use to analyze petroleum samples in particular, few studies presented data or drew conclusions in a manner relevant to address each regulatory need. The Web of Science search terms and results of the literature search are included in Supplemental Table 1.

Figure 4.

The use of various novel MS methods to address specific regulatory needs identified in this review. (A) Scholarly publications that were identified as relevant to each regulatory need/question (identified by colors). A total number of publications identified by a literature search is listed in the first circle (see Supplemental Table 2 for details). Upon examination of each study’s content, a number of publications were deemed not relevant (red circles); the remaining studies are shown in green circles. (B) The number of relevant publications as a function of the high-/ultra-high MS technique. Top left, a stacked bar graph indicating the number of publications as they pertain to each regulatory need/question in (A). Remaining stacked bar plots show the number of studies that used various separation (HPLC, GC, GC×GC or IMMS) or direct injection with each MS technique.

Critical Need 1: Providing detailed information on the structure of the constituents.

High mass measurement accuracy and resolving power of modern mass spectrometry analyzers are key steps to address this regulatory need for petroleum UVCBs. Regulatory agencies have repeatedly stressed that “broad” compositional information is insufficient to justify groupings and support read-across hypotheses (ECHA, 2014; ECHA, 2020a). While many publications tout ever increasing resolution as an advance in the science of analytical chemistry, we highlight recent examples of studies that focused on comprehensive structural information of the individual constituents in petroleum samples.

With the advances in the performance of mass spectrometers and development of new methodologies, significant strides have been made with respect to characterization of mixtures and petroleum samples. For the analysis of whole crude oils and refined products, FT-ICR MS could resolve hundreds of molecular formulae in the late 1990s (Rodgers et al., 1998) and several thousand in the early 2000s (Hughey et al., 2002; Qian et al., 2001). By 2019, newer FT-ICR MS instrumentation and experiment design led to the identification of ~245,000 molecular formulae in a non-distillable fraction of the maltenes from a heavy petroleum sample, using a resolving power of more than 3,000,000 across the entire m/z range (Palacio Lozano et al., 2019b). Thus, in less than two decades, there was a 60-fold increase in resolving power and 80-fold increase in the number of molecular formulae being assigned within a single sample.

While ultrahigh resolution mass spectrometry has offered unprecedented levels of insight into highly complex samples, such instrumentation typically affords molecular formulae but does not lead to definitive identification of structures. Experimental parameters, such as polarity and ionization behavior, can yield information about probable functionalities of the components, but true structural identification would require additional data. This could be obtained by fragmentation patterns associated with individual peaks (and hence molecular formulae), acquired during “tandem mass spectrometry” (also known as “MS/MS”) experiments. As complex samples may comprise tens of thousands or even hundreds of thousands of molecular formulae, performing hundreds of thousands of tandem mass spectrometry experiments to target all of the peaks is not viable, due to the time required and workload due to data analysis.

Combination with an orthogonal separation method, such as a form of chromatography or ion mobility, is an alternative approach for accessing information about functional groups and isomeric contributions. Optimization of these experiments (such as choice of columns, temperature program, etc.) is similarly non-trivial, however, and resolution of components in complex substances can be challenging; the concept of “unresolved complex mixtures”, also referred to as an “UCM hump” on the chromatograms, is well-known (Gough and Rowland, 1990). Furthermore, while chromatography is a well-known method for distinguishing isomers, this is typically combined with a need for authentic standards (e.g., to determine expected retention times on the chromatography column) and databases, although certainly not all available compounds are found within databases. Where there may be dozens of isomers per molecular formula, this would in practice mean a requirement for millions of authentic standards to be run in order to address the most complex petroleum samples that have been characterized by ultrahigh mass spectrometry. This is, again, impractical due to the amount of time required. It is also worth drawing attention to a subtle distinction between molecular formulae (e.g., C₁₈H₃₆O₂) and the concept of “peaks;” a peak may represent a single molecular formula in a direct infusion experiment by mass spectrometry (and data is represented by m/z and intensity only), but in combination with mass spectrometry with chromatography, there will be an additional dimension to the data (m/z and intensity, but now also time) and multiple peaks per molecular formula may occur due to the presence of isomers. Combining mass spectrometry with chromatography can yield more peaks by counting of isomers, but not necessarily more molecular formulae.

The isobaric information afforded by ultrahigh resolution direct injection FT-ICR MS has played a significant role in comprehensive understanding of petroleum substance composition; however, the presence of a large number of isomeric hydrocarbons in petroleum substances challenges the utility of ultrahigh resolution methods that do not provide other means for separation of isomeric species. Some studies addressed a growing need to define the isomeric composition of petroleum substances by coupling FT-ICR MS with separation techniques. Gas chromatography prior to FT-ICR MS was used for separation of isomers in petroleum by resolving individual constituents (Barrow et al., 2014; Palacio Lozano et al., 2022). Information of the structural composition and the isomeric diversity can be also attained by coupling with trapped wave ion mobility spectrometry (TWIMS), a variety of ion mobility spectrometry, to FT-ICR MS (Maillard et al., 2021). Multidimensional separation with ultrahigh resolution allows for characterization of individual constituents with accurate mass measurements and structural features. Additionally, high mass accuracy and ultrahigh resolution measurements have elucidated structural information of the individual constituents through post-instrumental data analysis (Cho et al., 2011; Hu et al., 2018).

A number of studies have coupled chromatography to Orbitrap MS for analysis of complex environmental samples (MacLennan et al., 2018; Pereira et al., 2013; Sorensen et al., 2019; Yang et al., 2019). Some studies applied this technique for the analysis of oils or fractions thereof. A combination of GC separation with different ionization methods for Orbitrap MS was used to study volatile components of a petroleum refining product, gas condensate. The authors showed that separation of different isomeric compounds could be achieved using this hyphenated method thus aiding in deeper characterization of a complex substance (Kondyli and Schrader, 2019). More recently, a reverse-phase liquid chromatography method was applied to the analysis of petroleum refining-derived UVCBs and compared to direct injection Orbitrap and FT-ICR MS (Xia et al., 2021). The authors showed that not only could they obtain elemental formulas for a large number of hydrocarbon and heteroatom species, but chromatography-informed retention patterns could also be used to distinguish among isomeric species, hence increasing confidence in structural identification of the individual constituents in petroleum substances.

TOF MS detection, despite its lower mass resolution as compared to FT-ICR and Orbitrap MS, affords an advantage of rapid data acquisition; this technique has been explored extensively by coupling with chromatography or ion mobility spectrometry separations (Palacio Lozano et al., 2019a). The most common type of front-end chromatography in petroleum substance analysis is GC×GC. The additional separation afforded by this technique offers substantial improvement for resolving isomeric constituents and compounds that would otherwise coelute (Ball and Aluwihare, 2014; Luna et al., 2014a; Ngo et al., 2012; Rowland et al., 2011). For example, this technique was used to identify a large number of isomeric molecules, both aliphatic and aromatic, that could not be resolved by a typical GC-MS technique in a complex milieu of hydrocarbons containing a wide (C₁₂-C₃₆) range of carbon numbers (Alam et al., 2016). Similarly, the ability GC×GC-TOF MS to resolve isomers of classical and sulfur-containing naphthenic acids in oil-contaminated environmental samples enabled contamination-source fingerprinting (Bowman et al., 2019) and identification of potentially hazardous substances (Bowman et al., 2020). TOF MS has also been coupled with ultrahigh pressure liquid chromatography to isolate isomeric species in petroleum derivatives (Lv et al., 2013; Mahmoud and Dabek-Zlotorzynska, 2018).

The compositional characterization of petroleum samples has typically relied upon either the use of chromatography or, more recently, the use of ultrahigh resolution mass spectrometry. The separation on the basis of retention time using gas chromatography can be sufficient for monitoring targeted components (e.g., steranes for differentiation of sample origins) and revealing the most significant contributions. As samples become more complex, however, particularly with the increasing use of heavy petroleum sources, there is the increasing probability of coeluting components, as mentioned above, where compounds have the same retention time and therefore cannot be distinguished by one-dimensional chromatography alone. The use of a second column of a different type to result in a two-dimensional approach, GC×GC, can significantly improve the separation and therefore the number of compounds observed.

Due to the temperature ranges of the ovens, gas chromatography-based methods have limited retention times that are accessible, as these are, in turn, linked to the boiling points of the individual compounds. Direct infusion methods coupled with ultrahigh resolution mass spectrometry (i.e., where samples are injected directly into an ion source, with no preceding chromatography) do not have the limitations associated with boiling point, and ultrahigh resolution approaches with direct injection delivery have become increasingly significant, especially for heavy petroleum samples.

There has also been limited coupling of chromatography with ultrahigh resolution mass spectrometry, which affords researchers the ability to separate compounds on the basis of two different dimensions: retention time and m/z. In this way, co-eluting compounds (i.e., the same retention time associated with the GC column) can be separated due to the additional use of the m/z dimension (Barrow et al., 2014). The data sets acquired can be large (e.g., ~25–50 GB) and therefore present data processing and data analysis challenges. The combination of orthogonal approaches, yielding so-called “hyphenated” techniques, represents means by which to access the advantages of the individual methods. Ultrahigh resolution mass spectrometry affords the ability to resolve a greater number of peaks and assign many thousands of unique molecular formulae with confidence, while chromatography or ion mobility affords the ability to separate isomers (same molecular formulae and therefore same m/z, but differing structures due to different arrangements of the atoms). While GC and low resolution GC-MS methods have seen widespread usage for many years, it can be expected that multidimensional GC and the coupling of orthogonal methods with ultrahigh resolution mass spectrometry will both become increasingly used.

Obtaining structural isomeric information is also possible using ion mobility mass spectrometry (IMMS), a post-ionization separation technique, often coupled to TOF MS and used for the analysis of petroleum samples (Santos et al., 2015). IMMS provides structural information that is complementary to the observed m/z of a compound by characterizing the spatial conformation of individual constituents via their drift time through an inert gas and the subsequent derivation of collision cross section (CCS) values (Dodds and Baker, 2019). CCS values can also be compared with those determined through computational means to determine structures for observed compounds. Currently, this is laborious for complex samples and so not viable for samples with, for example, tens of thousands of molecular formulae and their associated isomers. The combination of ion mobility, mass spectrometry, and computational chemistry does, however, hold potential for providing greater structural insights during characterization of samples. IMMS allows for separation of isomeric compounds by their structural composition (Hoskins et al., 2011; Lalli et al., 2015; Lalli et al., 2017; Mahmoud and Dabek-Zlotorzynska, 2018). The high mass accuracy and resolution of TOF MS, coupled with structural characterization using ion mobility, has allowed comprehensive elucidation of the composition of complex petroleum substances and byproducts (Lalli et al., 2015; Lalli et al., 2017). Confident molecular formula assignment to the IMMS-derived features in a gasoline standard and a crude oil sample was demonstrated through the use of KMD analyses based on CH₂ and H functional units (Roman-Hubers et al., 2021).

While instrumentation has increased in performance, the data analysis methods have often struggled to keep pace. It is evident that highly complex samples are being characterized better than ever before. The field is acutely aware that handling the increasingly complex data is already challenging, but that also there is a need to obtain greater structural insights, going beyond molecular formulae alone and/or ensuring all of the many thousands of peaks are associated with a definitive structure, not only those that are mostly easily targeted. While the progress with complex substance analysis has been remarkable in recent years, there remain mountains to climb.

Critical Need 2: Providing information on the concentration of the individual constituents.

Quantification of individual constituents in complex and multi-constituent substances is a required step for hazard evaluation to ensure no underestimation of the potential human health and environmental hazards (CONCAWE, 2012; ECHA, 2017a). Thus, the ability of high resolution mass spectrometers to determine the thousands of individual constituents is not sufficient without determination of their abundance. Traditional quantitative approaches relying on mass spectrometry detection require the use of various extraction and detection standards, preferably isotopically-labelled ones (Urban, 2016). However, because of the complexity of petroleum substances, the use of standards for absolute quantitation of numerous, rather than a small number of targeted constituents, would potentially mean the need for millions of authentic standards, which is both impractical and impossible. Instead of quantifying absolute concentrations of multiple individual constituents, traditional analysis techniques such as GC-MS and GC×GC-FID derive relative (e.g., fraction of total) amounts for groups of compounds based on a limited number of standards for key classes of constituents. For example, quantitation of individual n-alkanes, selected isoprenoids, polycyclic aromatic and alkyl polycyclic aromatic hydrocarbons, and biomarker compounds is possible by GC-MS (Wang et al., 1994); however, the “UCM hump” of high molecular weight hydrocarbons limits the utility of this technique for hazard evaluation. GC×GC-FID technique is also commonly used for petroleum analyses to derive “hydrocarbon blocks” (ASTM International, 2011). This technique is more amenable for hazard evaluation as it can separate polycyclic compounds with known or suspected hazardous properties (Bierkens and Geerts, 2014); however, this technique is not considered sufficiently informative by some decision-makers, especially for the substances that contain C30 or greater hydrocarbons (ECHA, 2020a). Therefore, there is a great need to determine the ability of high resolution MS techniques to provide quantitation of the individual constituents in petroleum substances.

There are significant challenges with respect to quantifying the abundance of constituents in petroleum UVCBs. The signal observed for a given compound will be influenced by a number of experimental parameters. One of these is, of course, concentration. Other factors include solubility in the chosen solvents, pH of the sample solution, the suitability of ionization method chosen, the polarity of the ion source (e.g., when using electrospray ionization, acidic species will be observed in negative-ion mode while basic species will be observed in positive-ion mode), and instrument tuning, amongst other variables. As one example, alkanes can represent as much as half or more of the composition of some petroleum samples and yet if performing analysis using electrospray ionization, a widespread ionization method which is suitable for observation of polar and ionic species, then the alkanes would essentially go unobserved.

The matter of quantification is also closely linked to structure, as previously discussed. Different isomers of a molecular formula may have differing functional groups which, in turn, may influence the ionization response of the compounds. For instance, an organic molecule containing two oxygen atoms may incorporate two hydroxyl groups or a carboxylic acid, where the carboxylic acid would ionize much more readily than the former, and therefore give a much stronger signal. For this reason, quantification typically involves the coupling of chromatography to mass spectrometry, in order that isomers are separated the signals associated with each isomer can be measured and considered separately. Such experiments are labor intensive, however. Authentic standards must be used for each isomer of each molecular formula, as mentioned previously, but the standards must also be prepared as a series of sample solutions spanning a range of concentrations to measure the signal intensity as a function of concentration, leading to calibration curves. The needs for authentic standards and for calculation of the calibration curves leads to increases in researcher workload akin to orders of magnitude. “Untargeted” analyses could necessitate the need for potentially millions of experiments for quantification purposes, depending on the objectives. It is much more common that “targeted” analyses are instead used, where a short list of compounds of concern are searched for, such as benzene, toluene, ethylbenzene, and xylene compounds. Many advanced mass spectrometry approaches which are commonly used are accepted to be semi-quantitative, balancing sample complexity, experimental design, data processing, and time.

Most studies that use FT-ICR MS for the analysis of petroleum samples report tens to hundreds of thousands of detectable constituents; however, this technique is semi-quantitative where abundance of each molecule would depend on multiple factors (e.g., solubility, concentration, ionization response, tuning, etc.). FT-ICR MS studies typically report relative abundances of various hydrocarbon or heteroatom classes rather than that of individual constituents (Bae et al., 2010; Chen et al., 2012; Jennerwein et al., 2014; Kim et al., 2015; Oldenburg et al., 2014; Walters et al., 2015). Some publications focused on detection and quantitation of specific constituents, such as organic sulfur compounds (Lu et al., 2013), or metalloporphyrin complexes (Cho et al., 2014); however, these constituents have uncertain relevance for the purpose of hazard evaluation. The ultrahigh resolution Orbitrap MS technique has not been used for the characterization of the individual constituents; instead, the abundances of the detected ions with inferred elemental composition are used for semi-quantitative evaluation of various fractions and broad chemical classes (Castiblanco et al., 2020; Liu et al., 2020; Rodrigues Covas et al., 2020; Vanini et al., 2020).

Similarly, quantification of the individual constituents using TOF MS is challenging and most studies focused on quantifying abundances of broad classes of compounds (Scarlett et al., 2008). A combination of GC×GC-TOF MS and GC×GC-FID was used to conduct qualitative and quantitative analysis of polycyclic aromatic hydrocarbons for several petroleum UVCBs; however, quantitation was performed only for the constituents for which chemical standards were used (Ristic et al., 2018). Even though absolute quantitation of the individual constituents in complex petroleum samples may be unattainable, the combination of confident molecular formula assignments aided by structural information provided by IMMS and data on relative abundance does enable quantitative evaluation of the most abundant constituents (Roman-Hubers et al., 2021). For example, a study of variability in chemical composition of petroleum UVCBs used the average relative abundance of each constituent in a product across production cycles to determine what molecules may be present at relatively high (e.g., REACH threshold of concern at 0.1% (ECHA, 2017b)) amounts and whether those molecules vary significantly between production cycles (Roman-Hubers et al., 2022). This study is an example of how high resolution MS can be used to not only characterize individual constituents, but also to determine their abundance for consideration as potential substances of concern in hazard evaluation. Indeed, the ability to quantify, even in relative terms, the abundance of the identifiable constituents that comprise most of the petroleum sample is of utmost interest under the REACH framework (Table 1). The most recent advice from ECHA states that “all constituents present in a concentration at or above 1% must be identified” when grouping or read-across is proposed (ECHA, 2022). The bar is even greater for specific constituents that may possess hazardous properties, i.e., “0.1% for constituents that are classified as carcinogenic or mutagenic and 0.3% for substances that are toxic to reproduction or development.” Collectively, “identified constituents above the thresholds given above must account for a minimum of 80% of the mass of a UVCB substance” (ECHA, 2022).

Critical Need 3: Demonstrating compositional similarity of complex petroleum UVCBs through other means when the identification of all individual constituents is not technically possible or impractical.

Even though challenges remain in the confident identification of structures for individual constituents in complex substances, the multi-dimensional data from high resolution MS analyses (FT-ICR, Orbitrap and TOF MS) is highly valuable for evaluating broad similarities among complex petroleum UVCB substances and for identifying the degree of variability within a class of substances or between production batches of the same substance (Figure 4). Indeed, a number of studies have used these data to perform statistical analyses and visualize the relationships between samples in large datasets to demonstrate that even broad molecular compositional data can be effective as a means for screening, evaluating similarity and determining what constituents may be variable to prioritize selected samples and constituents that may warrant further targeted quantitative analyses.

The variation in composition of hydrocarbons and heteroatoms resolved using FT-ICR MS has been explored for the analysis of crude oils (Hosseini et al., 2021; Rocha et al., 2018; Silva et al., 2020) and refining products (Benassi et al., 2013; Mennito and Qian, 2013; Oldenburg et al., 2014; Oldenburg et al., 2017; Orrego- Ruíz, 2018; Silva et al., 2020; Walters et al., 2015; Wang et al., 2020), including studies that used such data for categorization of new products (Abib et al., 2020; Hourani et al., 2013; Liu et al., 2014). Through detailed chemical profiles, compositional characterization of different environmental samples allowed detection of the asphaltenes (Neumann et al., 2021; Ruger et al., 2015), fulvic acids (Stenson et al., 2003), oil sands process-affected water (Barrow et al., 2016), bitumen (Lacroix-Andrivet et al., 2021), biochar-derived organic matter (Li et al., 2022a), as well as soil and sedimentary organic matter (Zhong et al., 2011). Studies of oil weathering (Wozniak et al., 2019) and transformation in the environment (Jaggi et al., 2019; Li et al., 2022b; Wozniak et al., 2019) enabled to not only determine the trends in compositional changes, but also to group samples, allow for forensic identification of the related samples, predict the potential environmental impact of oil spills, and designing mitigation strategies.

While there will be limitations to the lowest m/z that can be detected by FT-ICR MS, based upon the magnetic field strength and highest frequency that can be detected (e.g., ~m/z 37 for a modern 12 T instrument), with appropriate instrument tuning, both Orbitrap and FT-ICR mass spectrometers can offer comprehensive insights into the lower molecular weight compounds of the chemical profile of complex substances, not afforded by other high resolution MS instruments (Chen et al., 2018; Cheng and Hous, 2021; Cho et al., 2017; Headley et al., 2011). Through the application of Orbitrap MS the variations of the mass spectrum profiles can be traced at a molecular level with low mass error to determine the overall chemical composition (Castiblanco et al., 2020; Dong et al., 2019; Porto et al., 2019; Silva et al., 2019; Vanini et al., 2020). Liquid chromatography separation techniques coupled to Orbitrap MS have provided qualitative and quantitative monitoring of the organic species through the chemical profile complex substances. (Folkerts et al., 2019; Miles et al., 2020; Sorensen et al., 2019; Xia et al., 2021)

When designing future strategies for screening petroleum-related compounds, it is important to consider what is being measured by different approaches. For example, with electrospray ionization, acidic species would typically deprotonate and be observed using negative-ion mode, while basic species would protonate and be observed using positive-ion mode; acidic and basic species would not be observed in the same experiment, therefore. While this might initially be considered a disadvantage because of the need for two experiments instead of one, on the other hand this can be used to the researcher’s advantage when needing to differentiate by functionality. An organic molecule with a heteroatom content of only one nitrogen atom (e.g., C_cH_hN₁) could be pyrrolic (weakly acidic) if observed in negative-ion mode, but would more likely be pyridinic (basic) if observed in positive-ion mode. Where electrospray ionization is suitable for polar and ionic compounds, it does not readily ionize non-polar species and so, for these, an alternative ionization method, such as atmospheric pressure photoionization, would be used. To cover acidic, basic, and non-polar species in an untargeted manner, three experiments may be used, but the number of methods employed could be reduced if instead adopting a targeted approach. Through an awareness of the advantages and disadvantages of the different ionization methods, combined with clear objectives of the screening process (which could become of a targeted nature), it is possible to tailor a system which balances required information and workload.

Multidimensional GC separation coupled with high resolution detection in TOF MS has been widely employed to map out the composition of complex substances. The high-throughput acquisition has been readily employed to screen for saturate and aromatic hydrocarbons to characterize the broad compositional profile (Haitao et al., 2013; Hao et al., 2017; Kulkarni and Thies, 2012; Luo et al., 2016; Qian et al., 2004; Rui et al., 2012). Nevertheless, to achieve a better compositional information and distribution patterns of complex volatile constituents, TOF MS has been coupled with gas chromatography for qualitative and quantitative assessment (Haitao et al., 2013; Hao et al., 2017). When coupled with an HPLC, non-volatile components can be readily separated based on isomeric and isobaric information for comprehensive high resolution characterization (Cao et al., 2020). The drawbacks observed from the above addressed separation techniques prompted the coupling of capillary electrophoresis with TOF MS for qualitative assessment of high molecular weight constituents in a heavy gas oil (Nolte et al., 2013). Compared to other TOF MS hyphenations, GC×GC offers multidimensional high resolution characterization to map out the composition of complex substances (Alam et al., 2018; Frenzel et al., 2010; Gabetti et al., 2021; Luna et al., 2014b; Muller et al., 2020; O’Reilly et al., 2019; Qian and Wang, 2019; Ristic et al., 2018; Zhu et al., 2020). The plots generated from GC×GC-TOF MS analysis plotting constituents in a two-dimensional space (1^st retention time versus 2^nd retention time) help elucidate the composition of complex substances and facilitates the molecular classification of compounds (Damasceno et al., 2014). Notably, physical characteristics (i.e., density) can be directly correlated based on the detailed chemical composition defined through GC×GC-TOF MS (Vozka et al., 2019).

Petroleum UVCB composition: A regulatory challenge, but what are the solutions?

As summarized above, novel high resolution mass spectrometers offer comprehensive characterization of petroleum substances to qualitatively assess and measure the broad chemical composition of a substance and its individual constituents. The information available through high resolution characterization of the chemical composition of UVCBs can address the shortcomings with regards to the prediction of toxicological properties when practicing read-across assessment. Comprehensive characterization can provide sufficient information on the molecular composition and their relative abundance within a substance to define the commonality between substances and their similarity at a molecular level. With this information, it is then possible to determine constituents which may be used to infer human health hazard properties of the whole substance.

Still, complex UVCBs, especially those produced by refining of oil, remain to be an evolving competency in regulatory science. In the recent past, both decision-makers and industry relied on rather imprecise substance categorization and read-across hypotheses to predict toxicological properties (Clark et al., 2013; McKee et al., 2015; Salvito et al., 2020). Recognizing the lack of clarity in the original regulations and guidelines (ECHA, 2017c), and persistent issues identified in regulatory submissions of petroleum UVCBs (ECHA, 2020a; ECHA, 2020b; ECHA, 2021), additional advice was recently provided with respect to the information on chemical composition for UVCB substances (ECHA, 2022). Indeed, there appears to be a major gap in what information is perceived as sufficient in terms of chemical characterization of petroleum UVCBs by either decision-makers or industry. At the same time, the science of analytical chemistry delivered a number of major advances in terms of novel techniques and visualizations to aid in deep characterization of complex petroleum UVCBs. Scores of research studies have been published, reviews written, and lectures delivered. A “meeting in the middle” is required between those working in regulatory and analytical disciplines, so that the translation of petroleomics science into the practice of decision making on petroleum UVCBs can begin to be realized.

One barrier to such a translation could be the difference of opinions on what “science” is “evidence” and whether the scientific data are “sufficient” to satisfy the regulatory requirements for making decisions (National Research Council, 2009). What is sufficient for a peer-reviewed scholarly publication in a specialized scientific journal may not be sufficient for a regulatory decision. For example, in the EU, it is generally noted that “available scientific and technical data” shall be taken into account “in preparing its policy on the environment” (Allio et al., 2006). It was noted that this legislative mandate does not call for the available data to be of “best available” quality, even though some of the agencies are required to “provide the EU institutions and the Member States with the best possible scientific opinions” (Allio et al., 2006). In the US, the concept of “best available science” has been more defined as a statutory requirement for risk evaluation of chemical substances (US EPA, 2017). Specifically, the statute defines best available science as “science that is reliable and unbiased. Use of best available science involves the use of supporting studies conducted in accordance with sound and objective science practices, including, when available, peer reviewed science and supporting studies and data collected by accepted methods or best available methods (if the reliability of the method and the nature of the decision justifies use of the data). Additionally, EPA will consider as applicable: (1) The extent to which the scientific information, technical procedures, measures, methods, protocols, methodologies, or models employed to generate the information are reasonable for and consistent with the intended use of the information; (2) The extent to which the information is relevant for the Administrator’s use in making a decision about a chemical substance or mixture; (3) The degree of clarity and completeness with which the data, assumptions, methods, quality assurance, and analyses employed to generate the information are documented; (4) The extent to which the variability and uncertainty in the information, or in the procedures, measures, methods, protocols, methodologies, or models, are evaluated and characterized; and (5) The extent of independent verification or peer review of the information or of the procedures, measures, methods, protocols, methodologies or models.” In principle, all of these can apply to characterization of the chemical composition of petroleum UVCBs when data are submitted to the authorities for a decision.

Related to the concept of best available science is the concept of “reasonably available information” which means information that the agency “possesses or can reasonably generate, obtain, and synthesize for use in risk evaluations, considering the deadlines […] for completing such evaluation” (US EPA, 2017). To relate the latter concept to the data on chemical composition of petroleum UVCBs, one may argue that a range of technically valid analytical methods applicable to the challenge of characterizing their composition and variability are already available and the data could be generated in a reasonable period of time. There is a balance between cost, availability, time, and the level of information required; the most advanced techniques which ultimately yield the fullest possible understanding may not be suitable for routine screening, whilst more common techniques may afford incomplete information or increased risk of misinterpretation. In this regard, novel analytical advances, such as those described herein, have greatly increased our ability to access valuable information on the composition of UVCBs at a molecular level. At the same time, the advances have revealed in much greater detail the compositional complexity of petroleum and areas where development is still required in order to fully access the range of components and, most significantly, establish structures. Still, strong arguments have been made about the impractical and potentially unnecessary regulatory requirements to deconvolute every possible constituent in a UVCB, even if at some arbitrary abundance cut-off level(s) (Table 1), and that hazard assessment may be sufficiently informed by using bulk compositional data and hydrocarbon blocks (Redman et al., 2012; Salvito et al., 2020). Clearly, the fields of analytical chemistry and regulatory science have not always worked in tandem.

To bridge this chasm, all sides will need to meet in the middle to establish baseline requirements with an understanding of what is analytically possible (Figure 4), including considerations of instrumentation availability, cost and time required for the analysis and data processing. It has been well-documented that different analytical methods, when applied to petroleum substances, have different windows of applicability with respect to the types of molecules they may ionize and/or detect, and may suffer from being semi-quantitative (Aeppli, 2022; Fernandez-Lima et al., 2009; McKenna et al., 2013; Prince and Walters, 2022; Rodgers and McKenna, 2011). Indeed, Figure 4 illustrates the point that novel high resolution analytical methods (FT-ICR, Orbitrap-MS and TOF-MS) have been already applied by a number of authors to the research questions that are directly relevant to the specific regulatory needs in petroleum UVCB space. However, the number of such publications remains very small as compared to studies of the traditional analytical approaches (Figure 2), and seldom do the authors acknowledge the potential for their work to be applied in the regulatory context(s). It is clear that the analytical chemists could better appreciate the regulatory issues and strive to provide solutions that not only advance the science but also address specific challenges. For example, the focus on the critical needs identified herein will serve well both the researchers and the ultimate end-users if targeted collaborations and regulatory-informed case studies to build confidence in their performance and utility. In addition, cross-validation collaborative trials are needed to achieve standardization of the new techniques and analysis methods to build confidence among stakeholders, a relevant example comes from the evolution of the analytical methods used for oil spill response (Faksness et al., 2002).

Concomitantly, the government bodies tasked with enforcing REACH and other relevant regulations, while providing additional useful guidance about the details of chemical characterization of petroleum UVCBs (ECHA, 2022), should have realistic expectations as to what is achievable using even the most advanced analytical methods. The most comprehensive analytical information may or may not be actionable in terms of hazard evaluation in absence of anchoring to the toxicological data. In this regard, existing data on a handful of known hazardous components in petroleum UVCBs, be it “priority PAHs” (ATSDR, 2005) or other constituents with existing human hazard evaluations, are widely regarded as insufficient for assessment of the whole substance(s). Thus, additional data from in vitro and other “bioactivity” data streams may be needed to determine the toxicologically relevant compositional features and variability within and among substances and categories (Grimm et al., 2016; House et al., 2022; House et al., 2021).

Concomitantly, the size and complexity of data sets pertaining to sample composition, which are increasingly accompanied with orthogonal high-dimensional information such as (eco)toxicological data, present challenges with respect to data handling and subsequent interpretation. These comprehensive high-/multi-dimensional datasets will typically be subject to dimensionality reduction and used to support grouping and/or classification of the individual petroleum UVCBs. A recent study demonstrated that the choices of the data analysis and visualization methods can not only potentially aid in the communication of “sufficient similarity” among complex substances, but also yield different outcomes in terms of grouping and classification (Onel et al., 2019). The proverbial “black box” of data processing and analysis may create “a variety of methodological and scientific concerns which mean that it is impossible to independently assess the methods and results” (ECHA, 2020a) in terms of reliably predicting the properties of the substances that are being evaluated. Hence, it is important to involve not only the analytical chemists when examining the data, but those in other scientific disciplines, such as statistics, mathematics, and computer science, and ensure that the data processing and analysis methods are transparent, and that the decision-makers are sufficiently familiar with the bioinformatics aspects of the information presented to them. Such involvement can expedite assessments of which components are particularly relevant for regulatory needs and may warrant greater focus and analytical precision.

Overall, while statutory and advisory language in the government agencies-produced documents may seem clear, the analytical science may not be available to fully identify and quantify all, or even most, components within complex petroleum UVCBs. It is being acknowledged that “it is not required to provide detailed structural information on all constituents of a UVCB substance, but there must be sufficient characterization of constituents so as to demonstrate structural similarity, and consecutively provide a basis for predicting the properties of the substance in read-across” (ECHA, 2020a). Even though additional clarifications have been recently made with respect to what “sufficient” may mean in the context of petroleum UVCBs (ECHA, 2022), ultimately, the flexibility is needed to consider new science and determine if it is “best available” and also fit for a specific decision-context purpose so that the statutes and regulations are ultimately enforceable. While adherence to the most common and established methods is understandable, the industry needs to be more open to the application of the modern analytical chemistry methods to the analysis of the samples and inclusion of such data into regulatory submissions. The traditional approaches that are used for broad characterization of petroleum UVCBs still have an important role to play with respect to substance identification, greater consideration of nontargeted approaches (and how these may also shed new light on emerging hazards), followed by targeted approaches in some cases, is needed. It is also important, however, to recognize that the measurement of the composition of a complex substance is very much influenced by the methods and techniques used; comparisons of data from different laboratories must demonstrate appreciation of such factors if the comparisons are to be meaningful. Therefore, detailed characterization of the technical performance of novel analytical methods is required, in addition to the generation of data for a wide range of samples and applications.

Highlights.

• Petroleum UVCBs are compositionally complex and variable

• Analytical characterization of these substance is a challenge for registration and evaluation

• Recently, the regulatory authorities provided more specificity about UVCB characterization

• Science advances and novel analytical methods in petroleomics offer solutions for data gaps

• This review aims to bridge the divide between the science of petroleomics and the regulators

Abbreviations:

API

American Petroleum Institute

ASTM

American Society for Testing and Methods

CAS

Chemical Abstract Service

CCS

collision cross section

DBE

double bond equivalents

ECHA

European Chemical Agency

EINECS

European Inventory of Existing Commercial Chemical Substances

US EPA

United States Environmental Protection Agency

FID

Flame ionization detection

FT-ICR MS

Fourier transform ion cyclotron resonance mass spectrometry

FWHM

full width at half maximum

GC

gas chromatography

GC×GC

two-dimensional gas chromatography

GC-MS

gas chromatography coupled to mass spectrometry

HPLC

high-performance liquid chromatography

IMMS

ion mobility mass spectrometry

IR

infrared

KMD

Kendrick mass defect

LC

liquid chromatography

MS

mass spectrometry

NMR

nuclear magnetic resonance

REACH

Registration, Evaluation, Authorisation and Restriction of Chemicals

TIMS

trapped ion mobility spectrometry

TOF MS

time-of-flight mass spectrometry

UV

ultraviolet

UVCBs

substances of unknown, variable composition, complex reaction products, or biological materials