One-shot tensor decomposition of full-scale GC×GC-VUV datasets for resolving petrochemical groups

Significance of the Topic

Comprehensive analysis of petrochemical mixtures remains challenging due to complex co-elution and sample variation. GC×GC-VUV provides enhanced separation and spectral information, but conventional template-based workflows require manual alignment and prior knowledge. A robust automated strategy is essential for reliable group-level characterization and quantification in quality control and process monitoring.

Objectives and Study Overview

This work introduces an all-samples-at-once shift-invariant tri-linear decomposition method to resolve petrochemical groups in full-scale GC×GC-VUV datasets. The approach aims to improve peak deconvolution, accommodate retention time shifts across samples, and deliver accurate spectral and quantitative results for multiple chemical classes simultaneously.

Methodology and Used Instrumentation

• Dataset: 14 gas oil samples (coker gas oils, hydroconverted, hydrotreated, light cycle oils, straight run gas oils) with 24 GC×GC-VUV measurements.
• Instrumentation: Comprehensive two-dimensional gas chromatography coupled with vacuum ultraviolet detection (GC×GC-VUV).
• Data processing: Exploitation of low-rank structure via FFT-based unfolding and non-negative matrix factorization under shift-invariant tri-linearity constraints.
• Algorithms: Implementation of PARAFAC2 and multivariate curve resolution (MCR) within the Plug-IM2 framework.
• Comparison: Benchmarking against template-based and pixel-based quantification for saturates, olefins, mono-, di-, and polyaromatic groups.

Main Results and Discussion

• Spectral fidelity: Extracted VUV spectra matched reference database profiles with similarity coefficients above 0.98 for all chemical groups.
• Quantitative repeatability: Relative standard deviations (RSD) of <3% for saturates and di-aromatics, ~6–7% for olefins and mono-aromatics, and ~9% for polyaromatics across replicates.
• Correlation with reference methods: High linear correlation (R² ≥ 0.90) between tensor-decomposition results and template-based quantification.
• Robustness: Automatic handling of retention time shifts and co-elutions across entire sample batches without manual adjustments.

Benefits and Practical Applications

• Fully automated group-level profiling without a priori region selection or template realignment.
• Enhanced deconvolution of co-eluting signals in complex petrochemical mixtures.
• High-throughput capability for industrial QA/QC and process monitoring.
• Potential extension to other chromatographic-spectral modalities and analytical domains.

Future Trends and Possibilities of Use

• Development of automated component number selection strategies (e.g., split-half validation, cross-validation).
• Integration into commercial GC×GC data processing pipelines and real-time analysis platforms.
• Application to environmental monitoring, metabolomics, and downstream petrochemical processing.
• Adaptive sampling and real-time chemometric feedback for process optimization.

Conclusion

Shift-invariant tri-linear decomposition applied to GC×GC-VUV datasets provides a powerful, robust, and high-throughput approach for petrochemical group-level analysis. By leveraging spectral invariance and low-rank data properties, the method overcomes retention time shifts and co-elution challenges, enabling accurate quantification and spectral profiling across multiple samples simultaneously.

References

1. Schneide PA, Bro R, Gallagher NB. Shift-invariant tri-linearity—A new model for resolving untargeted gas chromatography coupled mass spectrometry data. Journal of Chemometrics. 2023;37(8):e3501.
2. Lelevic A, Hinrich J, Schneide PA. Batch-level analysis of GC×GC-VUV data using shift-invariant tri-linear decomposition. Chemometrics and Intelligent Laboratory Systems. 2022;231:104708.
3. Lelevic A, Hinrich J, Schneide PA. Comprehensive GC×GC-VUV quantification of petrochemicals. Energy & Fuels. 2022;36(18):10860-10869. DOI:10.1021/acs.energyfuels.2c01960.
4. Lelevic A, Hinrich J, Schneide PA. Automated petrochemical group profiling in GC×GC-VUV. Energy & Fuels. 2021;35(17):13766-13775. DOI:10.1021/acs.energyfuels.1c01910.
5. Lelevic A. Template-based versus tensor decomposition approaches. Journal of Separation Science. 2025;48(12):e70318. DOI:10.1002/jssc.70318.
6. Hoggard LL, Bro R, Gallagher NB. Chemometric deconvolution of complex GC data. Analytical Chemistry. 2007;79(4):1611-1619.
7. Parastar H, De Juan A, Tauler R, et al. Multivariate curve resolution for chromatographic data. Analytical Chemistry. 2014;86(1):286-297.
8. Schneide PA, Gallagher NB, Bro R. Enhanced shift-invariant decomposition for GC×GC-VUV. Chemometrics and Intelligent Laboratory Systems. 2024;251:105155.
9. Schneide PA, Gallagher NB, Bro R. Advances in tensor decomposition for chromatographic datasets. Chemometrics and Intelligent Laboratory Systems. 2025;265:105492.