It is shown that polytypes or stacking sequences of cronstedtite, an Fe-bearing trioctahedral 1:1 phyllosilicate, can be determined using near-atomic high-resolution transmission electron microscopy (HRTEM). By viewing along the [010], [310] and directions (orthohexagonal indexing), the four groups of the standard polytypes can be distinguished. Imaging along the [100], [110] and directions allows determination of the polytypes in each group. The polytypic sequences of groups A and C are intergrown at the monolayer level in cronstedtite from Lostwithiel, England, which is a new insight if compared with previous suggestions that layer stackings characteristic of different groups do not occur together. The HRTEM images also revealed the relationship between the layer polarity and the morphology of the cronstedtite crystals, where the tetrahedral sheet side points towards the top of the truncated pyramidal shape of the crystal.

Cronstedtite is a member of the kaolin-serpentine group. It yields a wealth of more or less disordered polytypes. The crystals of polytype 1T (space group P31m, a = 5.512, c = 7.106 Å) contain, within coherently scattering blocks, variable concentrations of stacking faults so that domains of the basic 3D periodic structure can be shifted by 1/3 (a2 − a1) or 1/3(a1 − a2). These so-called OD parallel intergrowths have been confirmed by high-resolution transmission electron microscopy. The effect manifests itself in the diffraction pattern so that reflections with h-k = 3n — the family reflections — are always sharp, whereas remaining reflections — the characteristic polytype reflections — may be smeared out parallel to c*. The intensities of the latter are thus underestimated during diffractometer measurements. An analysis of such multiple OD intergrowths reveals that the moduli of structure factors for all characteristic (i.e. non-family) polytype reflections are reduced relative to those calculated for the non-intergrown basic structure, by a common factor. This fact usually leads to the appearance of ghost peaks in Fourier maps and to their erroneous interpretation. The structure of the basic model can, however, be refined much better if two scale factors are assigned to the family and non-family reflections, respectively.