Unified Kinetics of Serpentine Dehydroxylation

Calcination of serpentinite as a means of activating the mineral for leaching of magnesium has been studied for more than 80 years. At temperatures in excess of 800 K, dehydroxylation gives rise to the the formation of the largely amorphous phase anhydrous serpentine or, more commonly , metaserpentine in which the MgO units are relatively loosely bound. However, the process requires significant energy, both to heat the rock and effect the dehydroxylation. This energy demand is very significant in the evaluation and process design of ex situ mineral carbonation processes that depend on activation, in particularly the aqueous NETL process. However, it must be remembered that “dehydroxylation” and “thermal activation” are not interchangeable terms – the amorphous, activated product of dehydroxylation is metastable and converts rapidly to inactive crystalline silicates (forsterite and enstatite) at higher temperatures and, as has long been known, the activation window for serpentinite is relatively narrow.  It is therefore essential that the kinetics of dehydroxylation be understood in order to provide a basis for optimisation of the activation temperature and residence time.

We have evaluated data from earlier studies and found that the wide diversity of kinetic parameters (first-order rate constant, A, and activation energy, E) reported is simply correlated in terms of a kinetic compensation effect (i.e. a linear relationship between lnA and E). We suggest that this effect arises from treatment of the experimental data rather than from any intrinsic variability in the nature of the dehydroxylation in different samples. In particular, adoption of different kinetic forms for the progress of the reaction in the reacting volume (e.g. Ginstling & Brounshtein  vs the JMAK approach of Johnson & Mehl, Avrami, and Kolmogorov) leads to the extraction of different parameters from the same data set. Differences in ambient water vapour concentration and in particle size are also associated with variations in values of E and A reported from different experiments.

We show that literature data for isothermal dehydroxylation can be reduced to a common kinetic form corresponding to finite-rate kinetics at a retreating spherical interface. With the relative resistance of the diffusion and chemical reaction rates expressed as φ, this form is essentially a blend of the diffusion-limited Ginstling & Brounshtein model (D4, φ→∞) and the spherical, retreating interface model (R3,  φ→0). Since the diffusional resistance varies with particle size, as R/𝔇, it is essential to know the particle size in order to pursue this analysis.  

We have studied temperature-programmed dehydroxylation (TPD_OH) of size-segregated samples from South-West Oregon Lizardite (SWOL) and a serpentinite from the Great Serpentinite Belt in NSW, Australia. For each material, we examined ten size classes; mean particle diameters ranged from 4 to 250 mm. The reactions were carried in vacuo in order to minimise effects of ambient water vapour concentration. A relatively slow, linear heating rate of 20 K min^-1 was employed to ensure that the particles in the sample holder experienced uniform temperature-time history, as validated by a comprehensive in situ Curie-point study using 5 reference metals. Dehydration was monitored quantitatively by observing the sample weight loss as a function of time; we also examined the evolved gases by mass spectrometry and confirmed that water vapour was the only product detectable.

The TPD_OH spectra for SWOL showed two peaks, the first accounting for 87(±7)% of the mass loss and the second 13(±2)%. Similar behaviour was observed for the GSB sample (81 and 4% for the first and second peaks respectively) except that a third peak (15%) was also observed. While the minor (second and third) peaks were always located at the same temperatures (980±7 K and 1029 K, respectively), independent of particle size, the main peak shifted markedly when the particle size was changed. For SWOL, this shift was from 873 K with 4 mm particles to 953 K with 250 mm particles; the GSB samples showed very similar behaviour.

We interpret dependence of TPD_OH peak location on particle size in terms of the canonical spherical, retreating interface model (R3), implying slow kinetics at the interface relative to the rate of diffusion of water through the dehydroxylated, amorphous outer layer. The reaction activation energy and velocity factor are determined to be E = 355 ± 39 kJ.mol^-1 and A = 10^{13.96 ± 0.18} m.s^-1. The SWOL and GSB specimens exhibit statistically equivalent results, suggesting that these parameters are intrinsic to the mineral.

The absence of diffusional limitation in the observed kinetics is consistent with the observation that the overall rate is suppressed in the presence of water vapour at partial pressures as low as 2.1 kPa H₂O, most obviously at temperatures below 900 K. The ability of such low concentrations of ambient moisture to inhibit the reaction lends further support to the conclusion that the reaction is chemically controlled. Moreover, the degree of conversion under isothermal conditions, when expressed as a function of the normalised reaction time (t/t_0.5, for example), remains unchanged in the presence of water. We conclude that the underlying mechanism remains chemically controlled, with water vapour simply acting to suppress the rate. Similar rate suppression in the dehydroxylation of kaolinite derives from the reversibility of the rate-controlling chemical step.