Integrated molecular-level oil characterization and physics‑informed surrogate modeling for enhancing the tractability of supercritical water liquefaction of polyethylene

Supercritical water liquefaction (SCWL) offers a promising route for converting polyethylene‑rich plastic waste into liquid hydrocarbons, yet the intrinsic complexity of the process limits clear interpretation and predictive capability. In this study, high‑density polyethylene (HDPE) was liquefied under SCW conditions using explicitly measured real heating times. The resulting oils were subjected to comprehensive carbon-number-resolved molecular-level characterization using two-dimensional gas chromatography (GCXGC) coupled to flame-ionization detector (FID) and atomic-emission detector (AED) with selective adsorption pretreatments. The oils were dominated by paraffins (30–44 wt.%) and naphthenes (26–34 wt.%), with moderate olefins (9–14 wt.%), isoparaffins (5–10 wt.%), and aromatics (5–27 wt.%). Oxygenates remained below detection limits (<100 ppm), confirming that oxygen incorporation from water is negligible. Their comparatively low olefin content indicates enhanced suitability as steam‑cracking feedstocks relative to conventional pyrolysis oils. Mechanistic interpretation shows that SCW modulates radical chemistry by promoting cyclization and suppressing condensation pathways associated with char formation. To improve the tractability of SCWL a generalized severity index was introduced to capture the combined influence of different operating variables on oil composition. Building on this descriptor, a surrogate model based on a physics-informed Gaussian kernel regression was constructed enabling accurate predictions of hydrocarbon distributions (R² ∼ 0.9) under data‑sparse conditions. Overall, this integrated experimental-modeling framework enhances the tractability of SCWL and provides mechanistic and predictive foundation for advancing SCW‑based chemical recycling of PE-rich plastic waste.