Kimberlites : understanding their petrogenesis and uncovering the identity of their source composition

Kimberlites are one of the rarest and most volumetrically insignificant igneous rocks, which are derived from deep-seated (>150 km) magmas that originate in the subcontinental lithospheric mantle (SCLM) or asthenosphere. This deep origin, in conjunction with their intermittent relationship with deep crustal and mantle xenoliths and xenocrysts, including diamonds, renders kimberlites an invaluable tool for studying the composition and nature of intracontinental magmatism and the mantle. Despite more than half a century of research, the identity of parental/primary kimberlite melts in the mantle remains hotly debated. This is largely attributed to numerous processes which inescapably modify kimberlite magmas during magmatic ascent and upon emplacement in the crust. These processes include the entrainment and assimilation of mantle and crustal rocks, volatile (H\(_2\)O, CO\(_2\)) degassing, magma differentiation and syn- and/or post-magmatic alteration (i.e. serpentinisation). Constraining and quantifying these effects is therefore an essential task for constructing a complete understanding of kimberlite magmatism and petrogenesis. This thesis presents petrographic, geochemical and melt/fluid inclusion data of kimberlitic rocks and minerals from localities worldwide (Russia, South Africa, Finland, Canada) in order to: i) provide new constraints on the composition and evolution of kimberlite melts during and/or prior to magmatic ascent, solidification and post-magmatic alteration, ii) reconstruct the sequence of crystallisation of kimberlite minerals by examining textural relationships, zoning patterns and hosted inclusions, and iii) understand the processes that modify kimberlite magmas and rocks after their emplacement. An important technique used for examining the composition and evolution of kimberlite magmas prior to post-magmatic processes is the study of different generations (i.e. primary, pseudosecondary and secondary) which can provide snapshots of the melt at a particular stage of magma evolution. Melt/fluid inclusions were examined in xenocrystic olivine and megacrysts, as well as various magmatic groundmass minerals such as olivine, spinel, perovskite, monticellite, apatite and carbonates. Although these inclusions in kimberlitic minerals are extremely heterogeneous in composition and contain a diversity of daughter phase assemblages, they are all consistently shown contain abundant Ca-Mg- and Na-K-Ba- Sr- carbonates, Na-K-chlorides, F-bearing halides, phosphates, sulphates/sulphides and oxides. In contrast, (hydrous-)silicate minerals are either rare or absent. The daughter mineral assemblages in primary melt inclusions hosted by magmatic minerals show a systematic trend, where early crystallising phases (e.g., olivine, chromite) exhibit more silicate carbonate compositions, whereas late-stage minerals (e.g., apatite, carbonate) contain more evolved, carbonate-rich compositions that are enriched in phosphates, alkalis/alkali-earths and halogens. These compositional differences likely occurred in response to fractional crystallisation of constituent minerals (e.g., silicates, oxides). Examination of the Benfontein Sill Complex (South Africa) demonstrated that under quiescent intrusive settings, kimberlite magmas may undergo even more extreme degrees of fractionation towards essentially carbonatitic compositions, which are enriched in light rare earth elements (LREEs), high field strength elements (HFSEs), alkalis/alkali-earths and halogens. The abundance of alkalis/alkali-earths and halogens in melt inclusions hosted by kimberlitic minerals is at odds with their very low concentrations in kimberlite whole-rock. Melt inclusion evidence suggests that parental kimberlite melts were potentially much more enriched in alkalis/alkaliearths and halogens. However, these components were likely exsolved from the magma system during magmatic ascent/emplacement, or leached from the groundmass during synand/ or post-magmatic alteration (i.e. serpentinisation). A revealing feature of melt inclusions, in particular in groundmass monticellite and monticellite rims replacing kimberlitic olivine, is that they may provide evidence of meltcrystal reactions that occurred during groundmass crystallisation. Combined textural and petrographic data of monticellite partially replacing olivine combined with the presence of abundant periclase inclusions demonstrates that that monticellite in some kimberlites formed as result a decarbonation reaction between the carbonate component of the kimberlite melt and olivine in order to produce monticellite, periclase and CO\(_2\). It is inferred that CO\(_2\) was ultimately lost due to degassing and periclase also existed in the groundmass, but was subsequently altered to brucite during post-magmatic alteration. Furthermore, CO\(_2\) removal is a likely driver of this decarbonation reaction, where additional degassing of CO\(_2\) causes this reaction to proceed further in order to maintain equilibrium. This process may in turn be a commonly overlooked process in the exsolution of CO\(_2\) in kimberlite eruptions. One of the most intriguing cases that challenge previous kimberlite melt models is the petrologically unique Udachnaya-East (Russia) kimberlite. This locality is characterised by unserpentinised units, which contain fresh olivine and groundmass that is composed of the same alkali- and halogen-rich minerals (e.g., carbonates, chlorides, sulphides/sulphates) that are ubiquitously found in melt inclusions hosted in kimberlitic minerals in this kimberlite, as well as from localities worldwide. Combined petrographic, geochemical and melt inclusion evidence presents new evidence countering the ‘crustal contamination model’, which previously asserted that Udachnaya-East was either intruded evaporites or were permeated by platform brines. Comparisons between the unserpentinised and serpentinised varieties of the Udachnaya-East kimberlite show that they share numerous mineralogical and geochemical similarities, but are distinguished by: i) the presence of fresh olivine and abundant groundmass Na-K-Cl-S-rich minerals and absence of H\(_2\)O-rich phases (i.e. serpentine, iowaite (Mg\(_4\)Fe\(^{3+}\)(OH)\(_8\)OCl•3(H\(_2\)O))) in unserpentinised units, and ii) the absence of alkali- Cl-rich groundmass phases and incipient-to-ubiquitous olivine alteration in serpentinised varieties. Examination of melt inclusions in olivine and groundmass minerals in both serpentinised and unserpentinised kimberlite varieties show that they virtually identical and both enriched in alkalis-Cl-S, therefore indicating that these components were an intrinsic part of the kimberlite melt, originating in the mantle. Although the unserpentinised units of the Udachnaya-East kimberlite may represent an example of ‘pristinely preserved’ kimberlite, it is unique, and thus producing a universal primary kimberlite melt model using this locality is tenuous. Additional evidence of alkali and halogen enrichment in the parental/primary kimberlite melts in the mantle was obtained from the study of metasomatised mantle minerals. The presence of djerfisherite (K\(_6\)(Fe,Ni,Cu)\(_{25}\)S\(_{26}\)Cl) rims surrounding Fe-Ni-Cu sulphides in mantle xenoliths and as inclusions in rock-forming minerals are interpreted to be the result of modal metasomatism by infiltrating K-Cl-bearing kimberlitic melts/fluids reacting with precursory sulphides, which likely occurred close to the timing of, or during kimberlite magma ascent. In addition, large polymineralic inclusions and micro melt inclusions hosted in kimberlite-hosted megacrysts (e.g., clinopyroxene, olivine) are texturally and geochemically interpreted to have formed due to kimberlite melt infiltrating along crystal defects. Significant disequilibria between the megacrysts and the permeating kimberlite melt resulted in the formation of hybrid daughter mineral assemblages within these inclusions. Micro melt inclusions became completely isolated during fracture healing and are shown to contain entrapped remnants of variably differentiated kimberlite melt that was more enriched in alkalis-Cl-S-CO\(_2\) than polymineralic inclusions and the kimberlite whole-rock, which were modified during post-magmatic alteration. The combined study of melt/fluid inclusions in xenocrystic and magmatic kimberlitic minerals from localities worldwide have all demonstrated a consistent trend, which shows that they interacted and/or entrapped a variably differentiated aluminosilicate- and H\(_2\)O-poor, Ca-Mg-, halogen- (F, Cl) and alkali- (Na, K) bearing melt, that contained varying amounts of alkali-earths (Ba, Sr), phosphorus and sulphur. This composition may be a more likely candidate for primary/parental kimberlite melt compositions, as opposed to the classical ‘ultramafic, H\(_2\)O- and silicate-bearing’ model. In summary, this thesis demonstrates the strong potential of melt/fluid inclusion studies in circumventing processes that intermittently contaminate kimberlite rocks and gain insight into the composition and evolution of kimberlite melts during ascent and emplacement