Abstract

Crude oil spills pose a severe threat to coastal ecosystems, degrading sand and soil through the persistence of toxic hydrocarbon compounds. Froth flotation offers an effective physicochemical approach for remediating hydrocarbon-contaminated soil by separating hydrophobic compounds from the solid matrix using injected air in a stirred reactor. This study conducted kinetic flotation tests in a 5 L batch cell to derive kinetic parameters. Eight kinetic flotation tests were performed using fine sand (particle size < 150 mu m) contaminated with a diesel-bunker oil mixture. The experimental variables included airflow rate, hydrocarbon concentration (8-17 g/L), and surfactant dosage, at a fixed pulp density of 3 wt% solids(wt%). The tests provided kinetic constants and maximum recovery values, enabling quantitative modeling of hydrocarbon removal over time. All eight tests achieved hydrocarbon extraction rates exceeding 75% within 12 min. Analysis of the kinetic curves revealed a systematic deviation from the first-order exponential behavior typical of mineral flotation. During the initial 0-5 min, removal followed a linear (zero-order) trend attributable to the natural buoyancy of organic droplets, indicating that hydrocarbon flotation is governed by two concurrent mechanisms: (i) conventional first-order bubble-particle attachment and (ii) zero-order buoyancy-driven ascent intrinsic to low-density organic compounds. Least-squares fitting of three kinetic models (first-order, Kelsall, and Klimpel) to the eight experimental datasets showed that the first-order and Kelsall models achieved equivalent fit quality (mean sum of squared residuals (SSR) 588 and 586, respectively), with the Kelsall slow-component rate constant converging to zero in seven of eight tests; the Klimpel model yielded a substantially higher mean SSR of 1,043. A covariance analysis identified initial hydrocarbon concentration as the dominant factor controlling the kinetic rate constant k, while the hydrocarbon-airflow interaction (D & times; A) was the strongest driver of maximum recovery R-infinity, reaching a standardized covariance coefficient of 97.5 for the first-order model.