The analytical landscape for fuels, bio-based materials, and environmental monitoring is undergoing a massive shift. As the energy sector transitions from conventional crude oils to heavy distillates, oil sands, and alternative renewable feedstocks—such as pyrolysis and hydrothermal liquefaction (HTL) oils—traditional analytical techniques are being pushed to their absolute limits.
To understand how researchers are overcoming these analytical challenges associated with complex molecular mixtures, we look to the insights of Dr. Rafal Gieleciak, a senior research scientist at CanmetENERGY Devon, Natural Resources Canada. Dr. Gieleciak's pioneering work focuses on the deep chemical characterization of complex fuels and clean energy technologies, offering a roadmap for labs transitioning into the next era of separation science.
The Breakdown of 1D-GC: Navigating the "Unresolved Complex Mixture"
For decades, one-dimensional gas chromatography (1D-GC) has been the undisputed workhorse of the petroleum laboratory. However, it struggles severely when confronting modern, complex feedstocks.
"Traditional one-dimensional GC is very useful for relatively simple mixtures, but it starts to struggle when the sample contains thousands of compounds over a wide boiling range," Dr. Gieleciak explains. "The problem gets even worse with renewable feedstock-derived streams such as pyrolysis oils, hydrothermal liquefaction (HTL) oils, and partially upgraded biocrude intermediates."
Working in analytical science?
Register for a FREE Separation Science account to subscribe to the Separation Science Newsletter.
Instead of clean, distinct peaks, the chromatogram of these highly complex samples features a massive, unresolved hump known as the unresolved complex mixture (UCM). Because many compounds elute from a 1D-GC column at nearly the same time, individual compounds become impossible to isolate.
This severe co-elution hides critical analytical information, particularly regarding heteroatom-containing compounds (specifically sulfur, nitrogen, and oxygen). Though present in low concentrations relative to the bulk hydrocarbon matrix, they exert a disproportionate influence on refining and environmental outcomes.
"They can have a major impact on upgrading, refining, catalyst performance, equipment corrosion, process energy consumption, emissions, and product quality," Gieleciak warns. "In 1D-GC, these compounds can easily be hidden under co-eluting hydrocarbons."
The Multidimensional Solution
Comprehensive two-dimensional gas chromatography (GC×GC) solves this fundamental limitation by adding a second, independent dimension of separation.
In a typical setup, the first-dimension column separates compounds primarily by volatility (boiling point). The second-dimension column provides an orthogonal separation mechanism, typically based on polarity or chemical structure. By spreading the sample across a two-dimensional space, GC×GC creates highly structured, recognizable patterns, separating compounds into a grid-like peak pattern rather than a baseline hump.
When coupled with Time-of-Flight Mass Spectrometry (TOF-MS)—which is fast enough to acquire multiple spectral scans across incredibly narrow chromatographic peaks—researchers gain a powerful three-dimensional data space.
"Coupling GC×GC with mass spectrometric detection adds another important layer of information," says Gieleciak. "Mass spectrometry provides molecular-level data that supports compound identification, whereas the improved chromatographic separation reduces spectral overlap. For complex petroleum and biomass-derived samples, this combination is often essential to meaningful chemical characterization."
Environmental Water Tracking: The Case of OSPW
The chemical complexity of heavy hydrocarbon mixtures isn't confined to the refinery gate; it also presents severe environmental monitoring challenges, such as analyzing oil sands process-affected water (OSPW) to track persistent, toxic naphthenic acids.
"Environmental water samples are difficult to analyze because they are both complex and highly variable," Gieleciak explains. "Instead of a single compound or a few, there are large groups of molecules that vary in carbon number, structure, and heteroatom content."
According to Gieleciak, laboratories must address three distinct technical realities to navigate these challenges:
- The matrix effect: The sample matrix can suppress signals and distort the perceived chemical composition.
- Beyond the molecular formula: For environmentally persistent compounds, a molecular formula alone is insufficient. Analysts require detailed information on subclasses, isomers, and transformation products.
- The workflow dictates the result: Because these acid extracts comprise a diverse family of molecules, upstream preparation heavily biases what the instrument ultimately detects. Steps such as pH adjustment, filtration, and derivatization significantly affect the final results.
Ultimately, tracking these highly persistent contaminants requires an integrated strategy in which meticulous, repeatable sample preparation and high-resolution chromatography align to yield a trustworthy chemical fingerprint.
Drowning in Data: A Structured Chemometrics Roadmap
The sheer volume of raw data generated by a single GC×GC-TOF-MS run can easily overwhelm a bench scientist. Rather than jumping straight into complex "black-box" machine learning, Gieleciak advocates a structured, chemically grounded roadmap to chemometrics.
"Scientists do not need to start with advanced machine learning right away," Gieleciak advises. "Often, the most helpful chemometric tools are those that are simple to visualize and explain, like Principal Component Analysis (PCA)."
Before attempting advanced predictive modeling, scientists must establish clean, reliable data through meticulous preprocessing. "Alignment, normalization, background correction, and feature filtering are essential," Gieleciak says. "The biggest improvement often comes not from using a more complicated algorithm, but from making sure the dataset is clean, reliable, and chemically meaningful."
Recommended Data Analysis Workflow
| Phase | Methodology | Primary Purpose / Gieleciak's Guidance |
| 1. Clean | Data Preprocessing | An essential starting point to eliminate noise, baseline drift, and retention time shifts. |
| 2. Explore | Principal Component Analysis (PCA) | Used to check whether the analytical method is capturing real chemical differences (for example, feedstock, treatment, aging). |
| 3. Model | Partial Least Squares (PLS) Regression | Links chemical patterns to measured properties, performance indicators, or process variables. |
| 4. Extend | Advanced AI & Neural Networks | Useful for modeling complex, nonlinear relationships, provided the models are chemically grounded and rigorously validated. |
"I suggest starting with PCA or clustering to get a sense of the data structure and then moving on to supervised tools like PLS regression only when you have a clear prediction question," Gieleciak recommends. "The main goal should always be to understand the chemistry, not just to find patterns."
Bridging the Software Gap: Why Researchers Build Custom Tools
While commercial chromatography software packages are optimized for routine testing and regulatory compliance, they often fall short in advanced research settings.
"We built research-focused MATLAB tools to solve specific problems that came up when working with complex petroleum and biomass-derived samples," Gieleciak asserts. "The objective was to add more flexibility for exploring data, visualizing results, classifying samples, and interpreting molecular patterns."
Gieleciak highlights transparency as a major driver behind this custom approach. "Some commercial platforms are very efficient at providing answers, but some parts of the process may function like a black box. This makes it hard to develop, explain, and defend a new analytical method."
Custom Modeling via QSRR
Custom software tools also enable research groups to integrate Quantitative Structure-Retention Relationship (QSRR) modeling, which links a compound's molecular structure to its chromatographic behavior.
"In highly unresolved mixtures, compound identification cannot rely solely on library matching," Gieleciak explains. "Many compounds are not represented in spectral libraries, and well-resolved spectra attributable to individual compounds are often unavailable"
QSRR introduces a deterministic layer of chemical verification by predicting where a molecule should logically elute based on its physical properties. "If a proposed structure fits the mass spectrum but appears in the wrong region of the chromatogram, that is a warning sign," Gieleciak clarifies. "I see QSRR as a supporting tool, not a replacement for experimental analysis."
The Path Forward: AI and Feedstock Flexibility
The future of the analytical laboratory lies at the intersection of digital transformation and chemical transition, fundamentally redefining the chemist's role over the next decade.
Will AI Replace the Bench Chemist?
Artificial intelligence is making inroads into everyday laboratory software, but Gieleciak notes that its transition to a routine commercial tool will be evolutionary rather than revolutionary.
"The technology exists on a subtle level, thanks to better peak identification capabilities, spectral deconvolution, and pattern recognition," Gieleciak says. However, for AI to be widely trusted in regulated environments, developers must address data variability. "Analytical datasets are usually noisy and vary significantly between different instruments. The software must be able to provide the user with a clear picture when it comes to the confidence level of the results."
Navigating Feedstock Flexibility
As the global petrochemical sector increasingly co-processes fossil fuels alongside biomass-derived feeds, pyrolysis oils, or waste intermediates, analytical methods must adapt.
"Most established petroleum workflows were designed for feedstocks with known behaviors," Gieleciak shares. "One cannot just assume that a method developed for conventional crude can capture all the important chemistry in biomass feeds. These samples usually contain higher concentrations of oxygenated compounds, polar components, or chemically unstable molecules."
This shift requires thoughtful adjustments in sample preparation, validation, and detection rather than a total workflow overhaul. The labs that succeed will be those that adapt their methods rather than using a one-size-fits-all approach.
The Next Five Years for Commercial Labs
Gieleciak notes that increasingly complex unconventional feedstocks, combined with growing pressure for faster turnaround, are creating a significant bottleneck in commercial analytical laboratories.
"The biggest challenge now for these labs is turning complex analytical data into reliable, decision-making-ready information fast enough for clients’ needs," Gieleciak concludes. "Generating data is not the main issue anymore; the real difficulty is interpreting it accurately and consistently."
To succeed, Gieleciak recommends that testing labs invest in selective, information-rich analytical tools, improve their chemometrics systems to reduce manual processing times, and focus heavily on training staff in advanced separation science and mass spectrometry.
