**5. Arguments for model of Mg-ilmenite crystallization**

### **5.1 An asthenospheric and lithospheric source for kimberlites, and their megacryst suite**

The similarity of Rb-Sr, Sm-Nd, and Lu-Hf isotope systematics, the same age of formation [11, 15, 16, 18, 28, 34] for kimberlites and low-Cr megacryst association of minerals (to which Ilm belongs) testify to a single primary asthenospheric source for them. The similar or almost identical compositions of Ilm in different pipes of one cluster can be accounted for by the existence of a common magmatic supply channel. Various clusters of pipes were fed via different channels of ascending kimberlitic melt, which therefore disintegrated and assimilated different mantle rocks. In the Orapa A/K-1 pipe, the Cr2O3 content of Ilm has been shown to be independent of the variation of other oxide components [35]. Two groups of Ilm are recognized in this pipe, with average Cr2O3 contents of 1.91 and 3.62 wt%, whereas the content of MgO remains virtually constant. The Ilm nodules from the same pipe, although showing discrete zoning in MgO and Fe2O3, are found to have homogeneous Cr2O3 contents. Ilm from the Monastery pipe (South Africa) can be divided into three groups [27] based on Cr2O3 and Nb contents, while they demonstrate the same trend in terms of major components. Thus, this feature of the behavior of Cr2O3 in Ilm is common for different kimberlite

pipes. Moore et al. [27] suggested that there was a mixing of magmas or assimilation of host rocks in the magma chamber during Ilm crystallization. We suggest that the assimilation of lithospheric mantle rocks by the kimberlite melt might have occurred in the supply channel of kimberlite pipes. It appears that this peculiarity did not originate in the asthenosphere, but rather in the different channels and modes of ascent of the kimberlite melt, which led to the formation of the various clusters of pipes.

### **5.2 The presence and formation of Ilm/oxide melts**

The presence of large Ilm megacrysts (up to 4 cm), their abundance (up to 3% of the total rock volume), sometimes found in pipes, and, finally, the existence of veinlets, Ilm lenses in deformed lherzolites (**Figure 11**)–all this indicates the existence of a melt Ilm composition. A number of researchers refer to the presence of such melts [18, 36–39]. The appearance of Ilm melts, judging by the veinlets in deformed lherzolites, is recorded at depths corresponding to the boundary between the asthenosphere and lithosphere. The liquidation of a high-Ti melt corresponding in composition to Ilm in the initial asthenospheric melt was caused by deformation processes and a change in the PT parameters during its ascent. We assume that the latter initiated the formation of deformed lherzolites and the ascent of the asthenospheric melt. Ilm crystallization from the kimberlite melt continued to the later stages of ascent and possibly during and after kimberlite emplacement into the upper crust, as indicated by the presence of small groundmass Ilm [40, 41].

### **5.3 The model of Ilm crystallization**

It is commonly argued that fractional crystallization is the primary mechanism responsible for the formation of composition trends in minerals of the Cr-poor megacryst suite [3, 4, 27, 41, 42]. Geochemical data, as well as petrographic constraints (e.g., the abundance of Cpx inclusions in Ilm macrocrysts and Ilm-Cpx intergrowths), indicates that Ilm and Cpx were the final phases of Cr-poor megacryst suite to crystallize [1, 38]. However, the Ilm composition distributions considered above using the example of MgO-Cr2O3 plots showed that they cannot be readily explained by a process of fractional crystallization.

The features of the composition distribution of Ilm macrocrysts considered above, the heterogeneity of the composition of both individual macrocrysts (**Figures 2**–**6**) and polygranular megacrysts (**Figure 9**) were the basis for distinguishing three stages of Ilm crystallization, which occurred at the level of (1) asthenosphere (in the primary asthenospheric melt); (2) the lithosphere (in the melt, which changed its composition as a result of the capture and partial assimilation of rocks by the mantle lithosphere) and (3) the lithosphere and crust (as a result of changes in P–T–O crystallization parameters during ascent through the lithosphere and crust.

In the first stage, crystallization of minerals of the mega crystal low-Cr association of minerals took place, including Ilm. It is assumed that the leading crystallization mechanism was fractional crystallization. At the same time, Ilm and Cpx crystallized last, after Grt, Ol, and Opx.

The second stage of crystallization of Ilm occurred in a melt enriched in MgO and Cr2O3 (as a result of the assimilation of rocks of the lithospheric mantle), which was reflected in the corresponding graphs by the formation of the left branch of the Haggerty parabola (**Figure 3a**).

### *Mg-Ilmenite from Kimberlites, Its Origin DOI: http://dx.doi.org/10.5772/intechopen.102676*

During the third stage, recrystallization of macrocrysts occurred as a result of an increase in fO2 of the kimberlite melt as it ascended through the upper horizons of the lithosphere. This stage is reflected in the formation of heterogeneity in the composition of individual grains. Recrystallization of Ilm led to a decrease in the content of FeO and MnO with a corresponding increase in the content of MgO. Since the content of Cr2O3 remains unchanged, these changes in the composition are reflected in the plot of MgO and Cr2O3 by the formation of the right branch of the "Haggerty parabola". All three stages of Ilm crystallization occurred in different pipes (pipe clusters) in different ways, which is primarily due to a different section of the lithospheric mantle, with a different set of trapped and assimilated rocks of the lithospheric mantle. The formation of other Ilm compositional distribution patterns (e.g., "Steplike", and "Hockey stick") is attributed to different compositions of the entrained and partially assimilated lithospheric mantle material, and different ascent dynamics in each of the different kimberlite conduits (which were different for each different kimberlite cluster). Similar Ilm compositional distributions are also typical of other kimberlite provinces worldwide, and we infer that Ilm's three-stage crystallization model is responsible for these compositional distributions in all cases [4, 7–9, 27, 37]. These compositional features are attributed to the existence of a single magmatic conduit feeding all pipes of a given cluster, and different conduits feeding different clusters. Proto-kimberlite melt compositions evolved separately in each cluster (conduit) by the incorporation and partial assimilation of trapped fragments and minerals of the lithospheric mantle rocks.

Summing up, we come to the conclusion that the differences in Ilm compositions in individual pipes, pipe clusters are due to a different set of trapped and partially assimilated mantle xenoliths, or local heterogeneity of the lithospheric mantle. And thus, the similarity of Ilm compositions in the pipes of a particular kimberlite field can serve as a key to deciphering its structure (that is, identifying pipe clusters).
