This thesis explored, as its major aim, the crystallographic and compositional characteristics of a particular type of peridotitic garnet associated with kimberlite. This garnet has a highly sinusoidal rare-earth element (REE) pattern as its distinguishing feature. Before the main research question could be addressed, a technique had to be developed that enabled the rapid and straight-forward acquisition of a full profile digital X-ray diffractogram from a single sub-milligram crystal fragment. After extensive experimentation and testing, successful development and realization of a method that is capable of producing such data was achieved. The next step of this research project was to empirically investigate and assess the crystal lattice strain model (CLSM) of Blundy and Wood (1994). Careful analyses of the REE present in a suite of clinopyroxenes were compared to the crystal structure data given from the XRD technique outlined above. Clinopyroxene is particularly useful for such an assessment because the radius of the M2 site in this mineral is between the largest and smallest REE, making the distribution of REE particularly sensitive to variation of the M2 site, which in turn is a direct consequence of the overall pyroxene structure. Subsequent to illustrating that XRD data could be collected on such small material and the XRD data and structural data given from the CLSM correlate strongly, peridotitic garnets with sinusoidal REE patters were investigated. The conclusions drawn in the first two contributions – namely that it was possible to collect accurate and precise XRD data from sub-milligram specimens and that the crystal structure and REE distribution were directly related – were imperative for the deduction of conclusions in the final, major research question. The XRD analysis of many garnets with and without sinusoidal REE patterns showed the presence of a small amount of an additional phase in some of these garnets. While this phase (2 out of 3 peaks indexed as possibly orthorhombic perovskite) is not present in sufficient quantities to give rise to such a strong sinusoidal segment in the garnet REE pattern, it prompted the critical step forward in formulating a working hypothesis for the otherwise inexplicable REE patterns. Specifically, I posit that many of the sinusoidal garnets may originally have precipitated as a very high pressure phase (in the mantle transition zone or deeper) that subsequently underwent a subsolidus isochemical transformation to garnet. Possible original precursor mineralogy is a combination of two perovskites or a perovskite + iv garnet assemblage. Theoretical calculation using experimental partition coefficients demonstrated that a mixture of Ca-perovskite (CaPv) and Mg-perovskite (MgPv) REE patterns in the approximate proportions of 10% CaPv and 90% MgPv produce a REE diagram that is strikingly similar to those observed in sinusoidal single phase garnet. It has been shown experimentally that with increasing depth in the mantle, garnet plus a progressively increasing CaPv component is the stable mineral assemblage. Initial precipitation of two perovskites or CaPv + garnet as cumulates from a deep magma ocean would preserve the REE distribution of these minerals. Subsequent exhumation of such an assemblage would result in the retrogressive subsolidus phase transformation to a mineral stable at conditions of T and P of the shallow mantle environment, i.e. garnet, while retaining the REE pattern of the initial precipitate mineral assemblage. While this working hypothesis will require many more tests, its proposal may have significant implications for the mantle structure.

This thesis explored, as its major aim, the crystallographic and compositional characteristics of a particular type of peridotitic garnet associated with kimberlite. This garnet has a highly sinusoidal rare-earth element (REE) pattern as its distinguishing feature. Before the main research question could be addressed, a technique had to be developed that enabled the rapid and straight-forward acquisition of a full profile digital X-ray diffractogram from a single sub-milligram crystal fragment. After extensive experimentation and testing, successful development and realization of a method that is capable of producing such data was achieved. The next step of this research project was to empirically investigate and assess the crystal lattice strain model (CLSM) of Blundy and Wood (1994). Careful analyses of the REE present in a suite of clinopyroxenes were compared to the crystal structure data given from the XRD technique outlined above. Clinopyroxene is particularly useful for such an assessment because the radius of the M2 site in this mineral is between the largest and smallest REE, making the distribution of REE particularly sensitive to variation of the M2 site, which in turn is a direct consequence of the overall pyroxene structure. Subsequent to illustrating that XRD data could be collected on such small material and the XRD data and structural data given from the CLSM correlate strongly, peridotitic garnets with sinusoidal REE patters were investigated. The conclusions drawn in the first two contributions – namely that it was possible to collect accurate and precise XRD data from sub-milligram specimens and that the crystal structure and REE distribution were directly related – were imperative for the deduction of conclusions in the final, major research question. The XRD analysis of many garnets with and without sinusoidal REE patterns showed the presence of a small amount of an additional phase in some of these garnets. While this phase (2 out of 3 peaks indexed as possibly orthorhombic perovskite) is not present in sufficient quantities to give rise to such a strong sinusoidal segment in the garnet REE pattern, it prompted the critical step forward in formulating a working hypothesis for the otherwise inexplicable REE patterns. Specifically, I posit that many of the sinusoidal garnets may originally have precipitated as a very high pressure phase (in the mantle transition zone or deeper) that subsequently underwent a subsolidus isochemical transformation to garnet. Possible original precursor mineralogy is a combination of two perovskites or a perovskite + iv garnet assemblage. Theoretical calculation using experimental partition coefficients demonstrated that a mixture of Ca-perovskite (CaPv) and Mg-perovskite (MgPv) REE patterns in the approximate proportions of 10% CaPv and 90% MgPv produce a REE diagram that is strikingly similar to those observed in sinusoidal single phase garnet. It has been shown experimentally that with increasing depth in the mantle, garnet plus a progressively increasing CaPv component is the stable mineral assemblage. Initial precipitation of two perovskites or CaPv + garnet as cumulates from a deep magma ocean would preserve the REE distribution of these minerals. Subsequent exhumation of such an assemblage would result in the retrogressive subsolidus phase transformation to a mineral stable at conditions of T and P of the shallow mantle environment, i.e. garnet, while retaining the REE pattern of the initial precipitate mineral assemblage. While this working hypothesis will require many more tests, its proposal may have significant implications for the mantle structure.