Revisiting liquid-liquid extraction of pyrolysis bio-oil: Parameter sensitivities, residual water removal, and analytical implications

Biofuels produced from waste-derived pyrolysis liquids can contribute to near-term decarbonization, but their upgrading and commercialization depend on reproducible chemical characterization. Pyrolysis bio-oil typically contains substantial water, which depresses heating value, degrades stability, and compromises GC–MS repeatability; therefore, water must be controlled prior to compositional profiling. Liquid–liquid extraction (LLE) is widely used to isolate GC–MS-detectable fractions and remove water, yet protocols vary widely across the literature, limiting cross-study comparability needed for consistent process evaluation, performance benchmarking, and scale-up decisions in waste-to-energy pathways. Here, a Box–Behnken experimental design was applied to quantify the sensitivities of LLE performance to three commonly adjusted parameters: bio-oil-to-solvent ratio, Na 2 SO 4 drying-agent loading, and the number of extraction cycles. Residual water in the dried organic extracts was estimated using a rapid FT-NIR approach tailored for low-water organic matrices, and potential impacts on GC–MS chemical fingerprints were assessed using relative peak-area compound-class distributions. The results showed that residual water was significantly affected by solvent ratio, while the effects of Na 2 SO 4 loading and extraction number were weaker, with values remaining within a narrow range (<4%). In contrast, the relative distributions of GC–MS-detectable compound classes were not significantly altered by changes in extraction conditions, indicating that semi-quantitative chemical fingerprints remain comparatively stable. Overall, the findings demonstrate that LLE parameters exert limited influence on both excess water and chemical profiles. Given the small differences observed, lower solvent volumes, reduced Na 2 SO 4 usage, and fewer extraction steps may be sufficient, enabling more cost-effective and resource-efficient extraction protocols.