Desilication Kinetics in Bayer Liquor Replenishment

Within the closed-loop Bayer circuit, the continuous removal of dissolved silica via the precipitation of desilication product (DSP)—a sodalite-type sodium aluminosilicate—is a non-negotiable unit operation preceding alumina trihydrate recovery. Pregnant liquor arising from bauxite digestion at temperatures between 240°C and 270°C carries metastable silica concentrations frequently exceeding 4.0 g/L SiO₂ in high-temperature processing of diasporic or boehmitic bauxites, whereas the permissible ceiling for precipitation feed lies below 0.6 g/L to avoid co-precipitation of silica with gibbsite and subsequent contamination of the smelter-grade product. Replenishment of the liquor circuit with fresh caustic soda (typically 50 wt% NaOH) and makeup water, necessary to compensate for entrainment losses, evaporation, and chemical consumption, introduces abrupt shifts in the aluminate-to-caustic ratio (A/C) and total alkali concentration (C_Na₂O), each of which modulates the thermodynamic driving force and the kinetic regime of desilication. In continuous desilication vessels maintained at 95–105°C and atmospheric pressure, the replenishment stream is blended with spent liquor returning from the precipitation area; this mixed stream then passes through a series of mechanically agitated or air-lifted reactors wherein desilication proceeds predominantly via a seeded, surface-reaction-controlled mechanism. Residence time distributions obtained from radiotracer studies (using ^⁸²Br as a liquid-phase tracer) on full-scale tanks at Western Australian refineries reveal substantial dead volumes—up to 18% of the nominal capacity in poorly baffled cylindrical vessels—leading to short-circuiting that lowers the effective mean residence time below the design value of 6–8 hours and elevates the probability of silica breakthrough. This operational reality demands that kinetic models employed for liquor replenishment control be calibrated against tracer-validated reactor hydraulics rather than ideal continuous stirred-tank reactor (CSTR) assumptions.

Seeded desilication and the surface-area-limited regime

Addition of DSP seed recovered from the desilication underflow thickener, typically dosed at 50–200 g/L dry solids, provides the crystalline substrate onto which dissolved silicate anions and aluminate complexes co-precipitate. The overall rate expression derived from batch reactor data conforms to a shrinking-core or product-layer diffusion model, with the instantaneous desilication rate −d[SiO₂]/dt being proportional to the total seed surface area A_seed and the supersaturation driving force ([SiO₂] – [SiO₂]_eq) raised to an apparent order n ≈ 1.3–1.8 for liquors with C_Na₂O between 180 g/L and 250 g/L and A/C ratios in the range 0.65–0.72. A_seed, in turn, is governed not only by the mass loading but also by the particle size distribution (PSD) of the recycled DSP; laser diffraction measurements (ISO 13320:2020, with Mie theory applied to a refractive index of 1.53 for sodalite) on seed samples taken from refinery operations show that the volume-median diameter D_v50 typically lies between 12 µm and 35 µm, with a specific surface area determined by BET nitrogen adsorption (ISO 9277:2022) ranging from 8 m²/g to 22 m²/g. In refineries where replenishment with fresh caustic depresses the free hydroxide concentration transiently due to localized dilution effects—sometimes occurring in a dedicated caustic mixing tee upstream of the desilication feed tank—the equilibrium silica solubility [SiO₂]_eq can increase by as much as 0.15 g/L for a 10 g/L drop in C_Na₂O, eroding the supersaturation driving force and extending the required holding time by 15–30 minutes to reach the target residual silica of 0.45 g/L. Operators mitigate this by staging replenishment caustic injection across two or three points, thereby smoothing the alkali profile and preventing a local collapse of the desilication rate. Particle attrition within recirculation pumps (typically centrifugal pumps with an impeller-tip speed exceeding 25 m/s) generates sub-5 µm fines that act as secondary nucleation sites but also increase the viscosity of the desilication slurry; slurry rheology at 95°C exhibits pseudo-plastic behavior with a yield stress that can exceed 12 Pa at seed concentrations above 180 g/L, necessitating the installation of draft tubes with upward pumping impellers (Lightnin A315 or equivalent hydrofoil designs) to maintain solids suspension without exceeding the torque limits of gearboxes specified for continuous operation at a service factor of 1.5.

Direct measurement of intrinsic kinetics is confounded by the simultaneous precipitation of multiple sodalite solid-solution endmembers—chloride-sodalite (Na₈Al₆Si₆O₂₄Cl₂), sulfate-sodalite (nosean), and carbonate-sodalite (cancrinite-type)—each exhibiting a distinct molar volume and lattice parameter, which alters the diffusive resistance of the product layer. X-ray diffraction quantification using the Rietveld method (employing an internal corundum standard per ASTM D4926-20) on DSP solids collected from a refinery processing Weipa bauxite showed that the carbonate-bearing phase can comprise 55–70 wt% of the total DSP mass when the liquor carbonate concentration (expressed as Na₂CO₃) exceeds 25 g/L, and that this phase exhibits a slightly higher specific surface area after drying than the chloride-rich variant. Consequently, the apparent rate constant k, derived from an Arrhenius plot across the temperature interval 85–110°C, shifts from 0.018 min⁻¹ to 0.027 min⁻¹ for a seed mass loading of 100 g/L when the carbonate fraction rises from 30% to 60%, an effect not captured by standard single-species kinetic models. Refinery laboratories monitor the ratio of carbonate to total anion occupancy in DSP through a combination of wavelength-dispersive X-ray fluorescence (ASTM E1621-21) and thermogravimetric analysis with evolved gas analysis (TGA-EGA), reporting values weekly to the process engineering group to allow adjustment of the seed recycle ratio and the desilication temperature setpoint.

What limits the effectiveness of low-temperature desilication when make-up caustic is introduced during a bypass event?

Replenishment activities are frequently interrupted by short-duration bypass events—typically lasting 20–40 minutes—during which the stream from the bauxite residue washing circuit is diverted directly to the desilication feed tank without passing through the mud thickener overflow clarifier. This introduces fine red mud solids (predominantly hematite, goethite, and residual quartz) that adsorb silicate species and provide a competing surface for heterogeneous nucleation of DSP. An internal study conducted on a side-stream pilot rig at a South American refinery documented that the presence of 0.8–1.2 g/L suspended red mud fines (D_v90 < 20 µm) during a simulated caustic replenishment step depressed the desilication rate by 22–28% relative to a solids-free baseline, measured by the time required to reduce dissolved SiO₂ from 3.8 g/L to 0.5 g/L at 102°C with a seed charge of 120 g/L. The inhibitory mechanism is attributed to partial dissolution of iron oxyhydroxide surface coatings that release ferric ions into solution; ferric ions then complex with aluminate species, reducing the availability of Al(OH)₄⁻ for sodalite framework construction. When bypass events coincide with the injection of 50 wt% caustic soda, the resultant short-lived pH excursion to 13.8–14.0 (measured in-line with retractable high-temperature glass electrodes calibrated against NIST-traceable buffer solutions at 100°C) enhances the dissolution of these iron phases, magnifying the inhibitory effect. Refineries have responded by installing automatic actuated pinch valves on the bypass line, interlocked with the caustic makeup pump such that replenishment flow is reduced to a maintenance rate of 5 m³/h when the bypass is active, and by dosing a dilute sodium carbonate solution (15 wt%) to the reactor ahead of the bypass entry point to preferentially precipitate dissolved iron as siderite prior to its interaction with aluminate.

Desilication kinetics in replenished liquors are further complicated by the accumulation of organic carbon species—humates, fulvates, and low-molecular-weight organic acids—that survive the digestion process and concentrate in the liquor to levels of 15–35 g/L total organic carbon (TOC). These organics adsorb onto the active sites of DSP seed, especially the {001} basal planes of the sodalite cage structure, as demonstrated by diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) on seed samples subjected to Soxhlet extraction with tetrahydrofuran. At a TOC loading of 25 g/L, the apparent activation energy E_a for desilication in a synthetic Bayer liquor (C_Na₂O 200 g/L, A/C 0.70) was reported to increase from 58 ± 3 kJ/mol to 73 ± 4 kJ/mol, derived from isothermal batch experiments in a 1 L Parr autoclave equipped with a magnetically driven impeller and in-situ sampling via a dip tube with a 0.2 µm sintered metal filter. This elevation of the temperature sensitivity implies that, in high-TOC circuits, the thermal gradient within the desilication vessel—especially in the lower portion where liquor enters at 92°C and is heated by steam sparging to 105°C—creates a zone of sluggish kinetics that can persist for 2–3 hours unless mechanical agitation is intensified or the sparger is relocated to create a more uniform thermal field. Published data for this specific configuration of organic carbon interference in a continuous desilication cascade with lithium-based tracer validation is limited; however, plant-scale observations indicate that raising the operating temperature by a mere 3°C, from 100°C to 103°C, compensates for the kinetic penalty at TOC 20 g/L without requiring an increase in seed inventory, thereby averting the additional capital cost of a larger thickener.

When a desilication circuit is subject to seasonal variations in bauxite mineralogy—for instance, a shift from a largely gibbsitic feed to one containing increased proportions of kaolinite (Al₂Si₂O₅(OH)₄)—the reactive silica input can double within a 48-hour window, compelling rapid adjustment of the replenishment regime. Kaolinite dissolves completely during digestion, releasing two moles of silica per mole of alumina, and the resulting surge in dissolved SiO₂ to above 6.0 g/L in the pregnant liquor cannot be accommodated by the design residence time of standard desilication tanks sized for a steady-state feed of 3.5 g/L SiO₂. In such transients, refinery operators are forced to implement an emergency seed-generating protocol: a side-stream of the high-silica liquor is heated in a dedicated, low-headroom atmospheric reactor to 110–112°C with an unseeded induction period of 45–60 minutes, after which massive spontaneous nucleation of sodalite occurs, creating fresh seed with a D_v50 of 4–7 µm that is then transferred into the main desilication train. The replenishment caustic flow is simultaneously reduced to maintain the total alkali concentration at the upper end of the operating envelope (255–260 g/L Na₂O), because higher hydroxide concentration reduces the equilibrium silica solubility and provides additional thermodynamic driving force. The emergency protocol’s success hinges on the rapid attainment of a seed surface area exceeding 1500 m² per m³ of liquor; performance is verified online by an on-stream laser diffraction analyzer (e.g., Malvern Insitec or Sympatec HELOS) providing a PSD every 2 minutes, coupled with a slurry electrical resistance tomographic (ERT) sensor array mounted on the tank wall to detect stratification that could cause localized silica blinding.

Deposit adhesion strengths on shell-and-tube heat exchanger surfaces during unsteady replenishment operation

Scaling of DSP on the process side of shell-and-tube heat exchangers used for desilication liquor reheating—typically BEM-type single-pass exchangers with 19.05 mm OD tubes in a 45° rotated square pitch—presents a critical operational limit. When caustic replenishment is carried out through a direct sparging of 50% NaOH into the shell-side inlet nozzle, thermal shock and localized concentration gradients induce a transient where the solubility product of sodalite is exceeded adjacent to the tube wall. The incipient deposit, characterized by X-ray microtomography (resolution 2.5 µm voxel size), exhibits adhesion strengths measured via fluid dynamic gauging (using a calibrated stainless steel nozzle at a standoff distance of 0.25 mm) in the range 120–200 kPa for layers 50–100 µm thick. Over multiple deposition-cleaning cycles, the scale layer transitions from a porous, calcite-like morphology (specific surface area 15 m²/g) to a densified, vitreous-type scale with a surface area below 3 m²/g, at which point chemical cleaning with sulfamic acid (10 wt% at 60°C) requires circulation for 8–12 hours instead of the design 4 hours. To break this densification feedback, some refineries have switched to welded plate heat exchangers (Alfa Laval M15 or equivalent) with chevron-pattern plates pressed from 0.6 mm 254 SMO stainless steel, which produce higher wall shear stresses (above 15 Pa) at the same liquor velocity, thereby suppressing the formation of a stagnant boundary layer. The plate pack is configured with a port-hole modification that introduces the replenishment caustic through a separate, smaller-diameter port (DN25) located 150 mm upstream of the main liquor inlet, allowing the two streams to mix within the entrance region of the plate channel before encountering a heat transfer surface; this arrangement has been validated by computational fluid dynamics simulations using a species transport model with temperature-dependent viscosity and a non-Newtonian power-law index of 0.72.

The interaction between replenishment-induced alkalinity waves and the desilication reactor’s internal hydrodynamics has been studied by tracer-stimulated response modeling at refineries that installed on-line conductivity cells in the outflow of each of three CSTRs in series. When a step increase in make-up caustic flow from 18 m³/h to 24 m³/h was introduced at the suction of the first reactor’s circulation pump, the conductivity trace exhibited a double-peak response after 12 minutes, indicative of a rapidly traveling component that bypasses the central agitation zone and a slower, dispersed component that undergoes full mixing. The ratio of the fast-to-slow peak areas, computed by deconvolution with a two-compartment model (active volume fraction 0.82, dead volume fraction 0.18), correlated strongly with the measured outlet silica concentration an hour later (Pearson’s r = 0.91). This real-time signature allows closed-loop adjustment of the seed addition rate from the underflow thickener without waiting for the laboratory ICP-OES result, which typically has a 25-minute turnaround time; a PID controller receiving the conductivity peak ratio as a process variable reduces the seed pump speed by 10–15% when the fast-peak contribution exceeds 35% of the total area, anticipating that the temporary short-circuiting is reducing the effective residence time and thus less seed is required to compensate for the lower conversion per pass. Published data verifying the long-term stability of this control strategy across bauxite feed changes is not yet available, but twelve-month operational logs from a single refinery indicate that the incidence of silica excursions above 0.55 g/L in the precipitation feed decreased from 7.2 events per month to 1.5 events per month after implementation, with no measurable increase in DSP seed inventory.