Fig 2 - Phase identification was done using Bruker EVA software.


Fig 5 - Le Bail extractions and Rietveld refinements of the structure were performed using JANA 2006 and the Cheary Coelho fundamental approach for XRD profile parameters. 

Fig 10 - Corrections were made for ATR and background using Perkin-Elmer Spectrum software. 

Fig 11 - To understand how the Na:Al molar ratio affected the reaction of kaolinite to form a hydrosodalite, the proportion of kaolinite consumed was estimated using PXRD, TGA and 29SI MAS-NMR. For PXRD, Rietveld refinement was used as already described. For TGA, in the dTG spectrum (plotted in %mass loss / minute to normalise between samples), the peak attributed to the dehydroxylation of kaolinite was integrated. This peak area was then expressed as an area fraction of the equivalent peak in the dTG spectrum for the starting kaolinite precursor. The area fraction was then assumed as equivalent to proportion of kaolinite remaining in the sample. For 29Si MAS-NMR, peaks corresponding to kaolinite and hydrosodalite were integrated (deconvoluted as required when overlapping). The area fraction of the kaolinite peak from the total peak area in a given sample’s spectrum was assumed as equivalent to phase proportion, since 29Si is a spin-half nucleus and does not suffer quadrupolar effects. A Lorentzian profile was used for deconvolution as it gave a better fit to the measured curves than a Gaussian profile.