Hand-portable HPLC with broadband spectral detection enables analysis of complex polycyclic aromatic hydrocarbon mixtures

Polycyclic aromatic hydrocarbons (PAHs) are a significant concern due to their carcinogenic, mutagenic, and teratogenic properties. Their presence is widespread, from the interstellar medium to various environmental sources like fossil fuels and combustion processes. Historically, occupational exposure to PAHs led to a high incidence of cancer, highlighting their harmful effects. A broad range of diseases, including various cancers, metabolic disorders, and cardiovascular disease, have been linked to PAH exposure. Furthermore, in utero and childhood exposure has been associated with developmental problems. General population exposure occurs through contaminated air, water, and food. These health effects and bioaccumulation in food chains have led to stringent regulations globally. PAHs are non-polar organic molecules composed of fused aromatic rings. They exist in complex mixtures, making their analysis challenging. Environmental agencies list numerous PAHs as pollutants (e.g., the US EPA's 16 priority pollutants), including isomers with identical molecular formulas. Higher molecular weight PAHs are less mobile but persistent, while lower molecular weight PAHs are more water-soluble and bioavailable. Water analysis for PAHs is thus crucial. Current lab-based methods utilize high-performance liquid chromatography (HPLC) and gas chromatography-mass spectrometry (GC-MS). Samples are collected in the field and processed in centralized labs. While effective, these methods lack the ability for on-site analysis. Developing field-deployable systems is hampered by size, weight, and reliance on lab infrastructure (power, gas, solvent supply, waste disposal). While portable GC-MS systems exist, they are limited to volatile and thermally stable PAHs. Portable LC systems offer promise, requiring less infrastructure, but currently lack the sophisticated detection needed for complex mixtures. Existing portable LC systems often use single-wavelength UV-Vis detectors, relying solely on retention time for identification, making them vulnerable to matrix effects.

Numerous studies highlight the health risks associated with PAH exposure. Research on PAH metabolism and its link to lung cancer is extensive. Studies have examined the association between PAH exposure and various cancers (e.g., esophageal, skin, and other cancers). The cardiovascular effects of PAHs have also been explored. The impact of prenatal and childhood PAH exposure on neurodevelopment and respiratory health is well documented. Analytical methods for PAH detection have advanced, with HPLC and GC-MS being widely used. However, the limitations of these methods, particularly in terms of field applicability, have driven the development of portable analytical systems. Past research into portable GC, LC, and MS systems has demonstrated progress but hasn't fully addressed the need for on-site analysis of complex PAH mixtures with high analytical performance.

This research focuses on developing and validating a hand-portable HPLC system with a broadband spectral detector. The system integrates all components (pump, injector, detector, computer) into a single unit, weighing 4.2 kg. The pump uses pre-compressed gas, enabling zero-electrical power operation for extended periods (over 24h). The pulsed Xenon light source provides spectral detection from 180-890nm. The system's stability was evaluated, showing minimal drift in signal intensity over time regardless of power source (mains or battery). Experiments confirmed that photodegradation of PAHs was negligible during typical operation times. The automated constant volume injector was rigorously tested for reproducibility using a 4-component mixture, with RSDs for retention time and peak area comparable to a commercial HPLC (Agilent 1260/90). The chromatographic performance of the portable system was compared to the commercial HPLC system using a 24-component PAH mixture and two microbore columns: Zorbax Eclipse PAH and Poroshell 120 EC-C18. While complete resolution of all 24 PAHs wasn't achieved, the separation capability of the portable system was comparable to the commercial system. To address the issue of overlapping peaks in complex mixtures, spectral absorption detection was employed, providing full-spectrum chromatograms (3D data). Gaussian Mixture Models (GMMs) were initially used for spectral deconvolution but showed limitations due to overfitting and sensitivity to varying relative contributions of species in overlapping peaks. Therefore, Principal Component Analysis (PCA) followed by K-means clustering was used to determine the number of distinct species in unresolved peaks. Multivariate Curve Resolution with Alternating Regression (MCR-AR) was further implemented to achieve spectral deconvolution of hidden peaks. Linear Discriminant Analysis (LDA) using a database of reference spectra, supplemented with newly acquired spectra, was used for spectral fingerprinting and analyte classification. Field testing was conducted on water samples from various locations in Wales, England (London), and Cyprus. Samples were spiked with the 24-component PAH mixture, filtered, and analyzed. Recovery rates and variation in recovery rates were determined and compared with the results obtained from HPLC-grade water. The impact of matrix effects was evaluated. Spectral classification was consistently 100% accurate for all detected PAHs, even in the presence of matrix effects.

The development of a hand-portable HPLC system with a broadband UV-Vis spectral detector (180-890nm) represents a significant advancement in field-based PAH analysis. The system achieves chromatographic performance comparable to a commercial HPLC (Agilent 1260/90) in terms of reproducibility and retention time variation. Spectral fingerprinting, using LDA classification on a database of reference spectra, resulted in 100% classification of the 24 PAHs tested, including all those on the US EPA priority pollutant list. The spectral data overcame the limitations of retention time-based identification, mitigating matrix effects. Unsupervised methods, including PCA and K-means clustering with MCR-AR, effectively deconvolved overlapping and hidden peaks, enabling complete identification even in challenging samples. Field testing demonstrated the robustness of the method. The recovery rates of spiked PAHs in diverse water samples ranged from 82.16% to 170.36%, with most measurements showing variation less than 5%. The spectral classification remained at 100% accuracy despite matrix effects from varied water sources. The limits of detection varied for each PAH, with the lowest achievable detection being in the range of 2.5 ng/mL to 55.8 ng/mL. Possible improvements in the system to lower limits of detection include extending the absorption length of the flow cell, improving the spectrometer array's sensitivity and using pre-treatment methods to concentrate PAHs.

This research successfully addresses the need for rapid, on-site analysis of complex PAH mixtures. The portable HPLC system, combined with spectral fingerprinting and unsupervised peak deconvolution methods, provides a powerful tool for environmental monitoring and risk assessment. The ability to perform accurate PAH analysis in the field, without reliance on centralized laboratories, significantly improves the efficiency of regulatory enforcement and environmental remediation efforts. The 100% classification rate, even in the presence of matrix effects, demonstrates the robustness and reliability of the method. The system's portability and ease of use make it a valuable asset for various applications, including environmental monitoring, industrial hygiene, and emergency response.

This study demonstrates the successful development and validation of a hand-portable HPLC system with broadband spectral detection for PAH analysis. This system offers superior performance over existing portable LC systems by leveraging full-spectrum data to overcome the limitations of retention time-based identification and matrix effects. The unsupervised methods for peak deconvolution and spectral classification provide a powerful approach for analyzing complex mixtures in real-time. Future research could focus on further improving the system's sensitivity, expanding the library of reference spectra, and exploring applications beyond PAH analysis. Data sharing initiatives are also vital to fully realize the potential of dynamically populated spectral databases for field-based analysis.

The limits of detection (LODs) for some PAHs are higher than those reported for some lab-based HPLC methods. While the system’s performance was comparable to a commercial HPLC, complete resolution of all 24 PAHs wasn't always achieved in every run. Although the method demonstrated robustness against matrix effects in field testing, more extensive testing across a wider range of sample matrices is needed for complete validation. Finally, the availability of reference spectra in the literature might limit the expandability to other compounds in the absence of large training datasets.