Pertraction based separation of cobalt and aluminium from spent catalyst leach solutions

Abstract

Fischer-Tropsch (FT) spent catalysts contain three primary metals: platinum, aluminium, and cobalt. While platinum recovery is well-established, aluminium and cobalt remain largely unutilised. This study explores the feasibility of separating and recovering aluminium and cobalt from spent FT catalyst leach solutions for potential agricultural applications (specifically focusing on aluminium). Two separation technologies were evaluated and compared: (i) conventional mixer-settlers and (ii) pertraction. Initial shake-out mixer-settler experiments were conducted to determine optimal reaction conditions, followed by regression modelling to assess the influence of extraction variables. An industrial-scale mixer-settler system was also designed for the extraction step. Subsequently, benchtop pertraction experiments were performed under the optimal conditions identified from shake-out tests to evaluate the separation capability and obtain overall mass transfer coefficients. A dynamic model was developed to simulate benchtop membrane performance and provide insights into industrial scale pertraction implementation. Finally, a preliminary feasibility study compared both technologies from a Greenfield investment perspective. The shake-out tests identified optimal conditions, achieving 87% aluminium extraction with 17% cobalt co-extraction at 50 mol.% pre-neutralisation, pH 3.13, and 40°C. A scrubbing step with an aluminiumrich solution at pH 2.8 removed >99% of cobalt, followed by stripping with 1 M H2SO4, recovering over 94% of aluminium. The final product contained 54 ppm cobalt, meeting the <100 ppm product specification. An industrial-scale mixer-settler design based on McCabe-Thiele analysis required two mixer stages and a 16.7 m³ settling tank to ensure effective phase separation. Linear regression models failed to capture the complexity of the relationship between extraction efficiency and the targeted variables, necessitating the use of artificial neural networks. Among Levenberg-Marquardt, Scaled Conjugate Gradient, and Bayesian Regularisation algorithms, the latter demonstrated acceptable accuracy, with R² = 0.99 and a mean absolute error of 3.5%. The modelling results indicated that feed pH, composition, and pre-neutralisation equally influenced extraction efficiency, highlighting the need for control loops in industrial applications to maintain optimal performance. Benchtop pertraction experiments confirmed the technology's capability to separate aluminium and cobalt. The experimental mass transfer coefficients for aluminium and cobalt were determined as 5.6 × 10⁻⁵ m/s and 7.4 × 10⁻⁶ m/s, respectively. The validated dynamic model predicted these coefficients with ~4% error and was used to assess four industrial-scale membrane options. The LiquiCel 14x40 (X40) membrane emerged as the best choice, offering a low pressure drop (~8 bar), minimal production rate at a 16 m³/h minimum flow rate. The extraction phase of the PX technology required only ~0.6 m³ compared to the 16.7 m3 required for the mixer-settlers (due to the compact size of a membrane module while providing a very high surface contact area). From a Greenfield investment standpoint, the choice between pertraction and mixer-settlers depends on process needs and constraints. Pertraction is preferred when space is limited, solvent losses must be minimised, high selectivity is required, or modularity and sustainability (location dependent and customer driven) are prioritised, while mixer-settlers offer proven reliability, easier troubleshooting, and are more suitable for high-throughput processes where footprint and solvent inventory are less of a concern. membrane requirements (13 modules in series), high overall aluminium recovery (85%), an optimal