Abstract
Alumina carbochlorination is a key reaction step in chloride-based aluminium production, where alumina is converted to volatile aluminium chloride using CO and Cl₂. Interpreting the kinetics of this heterogeneous reaction is challenging because measured conversion depends not only on the intrinsic chemical rate, but also on gas composition, evolving solid structure, heat and mass transport, and reactor hydrodynamics. This thesis develops an integrated experimental and modelling framework for γ-alumina carbochlorination, connecting controlled packed-bed experiments with transport-aware kinetic extraction and reactive gas–solid computational fluid dynamics.
γ-alumina particles were prepared from gibbsite through boehmite and classified into unimodal 50–100, 100–125, and 125–160 µm agglomerate-size fractions. Shallow packed-bed experiments, using an initial bed height of approximately 5 mm, were conducted with systematic variation of temperature, particle size, Cl₂ flow rate, and CO flow rate. Apparent conversion increased with temperature from 500 to 700 °C, with decreasing particle size, and with increasing reactive-gas fraction. At 700 °C and equimolar CO–Cl₂ feed, the final conversion after 110 min decreased from 35.6 % for the 50–100 µm fraction to 32.5 % for 100–125 µm and 28.1 % for 125–160 µm. Increasing either Cl₂ or CO flow from 30 to 50 mL min⁻¹ raised the final conversion from approximately 23–24 % to 32.5 %.
Post-reaction characterization showed that the conversion data reflected not only alumina consumption, but also structural evolution of the reacting solid. Surface recession, particle rounding, fragmentation, fines formation, and increased accessible surface area were observed, while γ-alumina remained the main crystalline phase. The reacted particles developed bimodal particle-size distributions with fines below 30 µm, and the specific surface area increased for all fractions, most strongly from 28 to 44 m² g⁻¹ for the smallest fraction. The shallow bed showed no substantial height reduction after reaction, whereas a higher-aspect-ratio packed-bed test showed more visible layer-wise consumption and possible channeling. This supports the shallow bed as the controlled basis for kinetic interpretation.
Kinetic interpretation was developed through a hierarchy of reduced and spatially resolved models, progressing from literature-based particle-scale analysis, to transport-corrected packed-bed fitting, and finally to reactor-resolved gas-composition correction. The particle-scale models showed that apparent kinetics depend on temperature regime, internal diffusion, heat release, and structural assumptions. The main global fitting model for the present γ-alumina experiments combined shrinking-core surface scaling, power-law CO and Cl₂ dependence, Arrhenius temperature dependence, and a run-specific effectiveness factor to correct bed-scale reactant depletion. The fitting reproduced the experimental matrix with R² = 0.965–0.993 and gave m = 0.67 for Cl₂, n = 0.71 for CO, and Eₐ = 34.0 ± 1.6 kJ mol⁻¹, with η_run = 0.72–0.89 and Φ_bed ≈ 0.37–0.68. A transient packed-bed model gave a consistent activation-energy level of 34.7–36.2 kJ mol⁻¹, while the resolved laboratory-reactor model showed that the nominal inlet gas composition differs from the gas composition experienced by the packed bed. Correcting the bed-inlet composition gave activation energies of 33.6–36.4 kJ mol⁻¹. Together, the models show that the most defensible kinetic description for the present shallow-bed experiments is mainly chemically controlled γ-alumina carbochlorination with measurable transport correction.
The reactor-scale part established the hydrodynamic basis for transferring kinetics toward fluidized-bed operation. Non-reactive Eulerian–Eulerian simulations of a literature fluidized-bed benchmark in ANSYS Fluent, MFiX, and OpenFOAM produced reasonable agreement with measured bed behaviour, but bed expansion, pressure drop, voidage, and solid-volume-fraction fields differed with solver implementation, drag closure, dimensionality, and particle–wall treatment. The voidage-profile error ranged from 5.06 to 13.57 %, with MFiX generally giving the lowest error.
A reactive Eulerian–Eulerian computational fluid dynamics framework was developed in OpenFOAM by coupling the corrected kinetics with gas species transport, heterogeneous source terms, heat release, pressure response, bed expansion, and Eulerian solid-inventory change. A superficial-velocity sweep from 0.006 to 0.100 m s⁻¹ showed that gas velocity controls both bed structure and reactive environment. Increasing velocity increased product formation and heat distribution, but also promoted CO–Cl₂ breakthrough. Inventory-based alumina conversion increased from 4.2 × 10⁻⁵ to 2.3 × 10⁻⁴, with outlet mass fractions of AlCl₃ = 0.28, Cl₂ = 0.18, CO₂ = 0.14, and CO = 0.07 at 12 s in the highest-velocity case. The volume-averaged gas-temperature increase reached about 4.1 K, compared with 0.3–0.5 K for the lowest velocity range.
Overall, the thesis provides a connected route from material response and apparent packed-bed conversion to transport-corrected kinetics and resolved gas–solid computational fluid dynamics. This framework supports kinetic–hydrodynamic interpretation of γ-alumina carbochlorination within the present experimental and modelling boundary, providing a basis for future validation in reactive fluidized-bed operation.