Introduction

Transmission Electron Microscopy is an ideal way to characterise the microstructural details of a pyroxene sample. Since electrons have a considerably shorter wavelength than X-rays (0.037Â for lOOkeV electrons compared with 0.7093Â for MoK(x X-rays), many more diffracted beams may be observed simultaneously resulting in a large section of the reciprocal lattice being analysed. Since the resolution of the image is defined by the wavelength of the incident radiation, the small wavelength of the electrons allows imaging of Ângstrôm-scale detail of the sample. As electrons also interact strongly with crystals, diffraction patterns recording both space group and microstructural information may be photographed in merely a few seconds. The most useful crystallographic orientation for characterisation of pyroxenes is the projection down the 6-axis onto the chc plane, which allows the imaging of both stacking disorder and twinning. In most transmission electron microscopes (TEM's), the sample crystal may be tilted by between 10° and 60° on two different tilt axes in order to obtain the required crystallographic orientation.

Diffraction contrast, obtained by forming the image using different individual reflections in the diffraction pattern, is very useful for examining the microstructure of the crystal on a local scale. There are two different diffracting conditions possible for obtaining this type of image: the first, known as bright-field {BF) imaging, forms the image with only the direct electron beam passing through the objective aperture, while dark-field {DF) imaging requires that the direct or straight-through electron beam is excluded, and only a diffracted beam is used for image formation; details of the two methods may be found in eg., Putnis, 1992 and Buseck et al., 1980. Dark- field imaging is particularly useful for imaging parts of a crystal that have slightly different structures or orientations relative to the rest of the crystal, such as twins or antiphase domains (see Chapter 5). By tilting the crystal in the correct manner, it is thus possible to identify specific diffracted beams with individual microstructural features.

In order to fully understand the microstructures observed in the samples, it is convenient to describe the structures of the pyroxene polymorphs in a slightly different way to that outlined in Chapter 1. The orthopyroxene structure may be considered to be made up of clinopyroxene unit cells twinned on a unit-cell scale (eg, I to, 1935, 1950; Section 1.2). These monoclinic cells may be defined by blocks - 4.5Â wide in the a* direction, with the boundaries parallel to (100) passing through the tetrahedral chains common to all the pyroxene polymorphs (Figure 2.7a and b). These blocks may be in two different orientations denoted T and - , with the T and - directions being related by a 6-glide on (100). In low-clinopyroxene all the blocks have the same orientation, denoted + + + ... or —..., depending on their orientation relative to the unit cell axes (Buseck et al., 1980). The low-cl inopyroxene unit cell consists of two of the monoclinic blocks, since they are not equivalent by translation, resulting in cell dimension u - 9 k (Figure 2.7a). Orthopyroxenes consist of blocks in the stacking

sequence with a ~ 18 A (Figure 2.7b).

T w i n P l a n #

c h a in le n g t h s

Figure 2.7: Schematic diagram of the structures of a) low-clinoenstatite and b) orthoenstatite projected along [010], after Buseck et al., 1980. The + and - orientations of the blocks reflect the relative displacements of adjacent chains; the conventional unit cell in each case is shaded.

Lx)w-clinopyroxene and orthopyroxene are therefore both constructed from the same monoclinic building blocks; mistakes in the overall stacking sequences of the monoclinic units may cause twinning or stacking faults, which may be observed as streaking along the a* direction in the diffraction pattern of the pyroxene (Buseck et al., 1980). Although large scale twin intergrowths would not cause streaking in the diffraction patterns, the streaking in the a* direction (see Figures 2.9b - 2.15b) is due to interlayering of smaller units of ortho- and clinopyroxene.

Although it is not possible to directly image the high-pressure C2/c clinopyroxene phase (since it quenches to the lower symmetry P 2 /c phase upon pressure release; see Chapter 5), there is much information about this phase preserved in the microstructure of the low-cl inopyroxene. The presence of antiphase domains within the P 2/c clinopyroxene suggest that a transformation from the higher-symmetry C2/c phase to the lower symmetry P 2 /c phase has occurred (as first suggested for the C2/c to P2j/c transition which occurs on cooling in pigeonitic pyroxenes; Morimoto and Tokonami, 1969a); however such antiphase domains can also arise from growth mistakes of the pyroxene, depending on the morphology of the particular crystal. During the C2/c - ^ P 2 /c transition, ordering of the previously symmetry-equivalent silicate chains into "A" and ”B" chains occurs simultaneously in many parts of the crystal, forming numerous primitive domains. These domains will be either in registry, or out of registry by \l2{a-¥h), (Morimoto and Tokonami, 1969a). This vector of 1/2(a4-6) corresponds to the C-centring translational symmetry that is lost as a result of the transition.

Antiphase domain structures have also been observed for omphacites (Champness, 1973; Phakey and Ghose, 1973) in which cation ordering may cause reversion from the space group C2/c to P2/n, either on a local scale causing relatively small domains, or a more widespread one. However, the formation of such antiphase domains in omphacite by cation ordering is due to the differentiation of the Ml and M2 sites into Mia, M lb, M2a and M2b sites respectively, rather than a change in conformation of the silicate chains. Such a symmetry-breaking transformation is therefore not really analogous to the displacive C2/c - ^ P 2 / c transition which occurs in these (Mg,Fe)Si03 clinopyroxene phases during pressure release (see Chapter 5).

Due to the increased symmetry of the high-pressure clinopyroxene phase caused by its C-face centring, the associated h+k=(xM reflections are not observed in its diffraction pattern. However, since there are no systematic absences caused by the primitive cell of the low-clinopyroxene (although there are absences due to both the screw axis and the c-glide), information about the C2/c transition (and also mistakes occurring during the formation of antiphase domains) will be contained in these h+ k= odd reflections, which are absent from the diffraction pattern of the C2/c pyroxene. Antiphase domains are thus imaged in the TEM using the h+k=odd

reflections (eg, Putnis, 1992). Dark-field images obtained using the h-\~k=even

reflections of the low-cl inopyroxene diffraction pattern ought to show no contrast due to the antiphase structure, since these reflections are present in both the P 2 /c and C2/c diffraction patterns. Antiphase domains produced as a result of pressure decrease at ambient temperature are generally very small (eg, Carpenter, 1979); overlap of the antiphase boundaries generally tend to form an image in which the domains themselves are difficult to identify. If the crystal grain is also thicker than the domain size, this overlapping effect will become more pronounced.