The assimilation of felsic xenoliths in kimberlites: insights into temperature and volatiles during kimberlite emplacement

Abstract

This study aims to constrain the nature of kimberlite–xenolith reactions and the fluid origin for Kimberley-type pyroclastic kimberlite (KPK). KPKs are characterized by an abundance of basement xenoliths (15–90%) and display distinct pipe morphology, textures, and mineralogy. To explain the KPK mineralogy deviating from the mineralogy of crystallized kimberlite melt, we study reactions between hypabyssal kimberlite transitional to KPK and felsic xenoliths. Here, we characterize the pectolite–diopside–phlogopite–serpentine–olivine common zonal patterns using petrography, bulk composition, thermodynamic modelling, and conserved element ratio analysis. To replicate the observed mineral assemblages, we extended the thermodynamic database to include pectolite, using calculated density functional theory methods. Our modelling reproduces the formation of the observed distinct mineralogy in reacted granitoid and gneiss. The assimilation of xenoliths is a process that starts from high temperatures (1200–600 °C) with the formation of clinopyroxene and wollastonite, continues at 600–200 °C with the growth of clinopyroxene, garnet, and phlogopite finishing at temperatures < 300 °C when pectolite and prehnite join in. Critically, the majority of the new mineral growth occurs in the sub-solidus, at temperatures below 600 °C. The metasomatic origin of the xenolith mineralogy is best explained by gradients in the chemical potentials of Si, Al, Ca, and Mg across the xenolith–kimberlite contacts. The low-temperature mineralogy of the fluid-limited thermodynamic calculations, where H₂O and CO₂ are controlled by kimberlite concentrations, reproduces the observed mineralogy better than a fluid-saturated model with a meteoric fluid composition. Our findings imply the deuteric origin of the fluids in KPK pipes controlling the kimberlite mineralogy and texture.

Appendices

Appendix 1: Determination of thermodynamic parameters for pectolite

Thermodynamic parameters for pectolite (NaCa₂Si₃O₈(OH)) are summarized in Table 3. The values for ∆_fH^298.15 and V^298.15 were calculated according to Benisek and Dachs (2018) using DFT methods. C_P was experimentally measured and was used to calculate the standard entropy S^298.15 as described below. Thermal expansion α and bulk modulus K were estimated as mean values from diopside and jadeite. The values were manually added to the thermodynamic database file hp62ver.dat (Holland and Powell 2011).

Calorimetric methods

Low-temperature heat capacities (C_P) were measured using a commercially designed relaxation calorimeter (Physical Properties Measurements System (PPMS); Quantum Design^®) at temperatures between 2 and 300 K using a measuring technique described by Dachs and Benisek (2011). The sample powder (~ 10 mg) was put into an Al cup (~ 8 mg) made from an Al-foil. It was pressed to a cylindrical pellet (0.5 mm thickness, 5 mm in diameter), the Al-foil surrounding the sample powder. The heat capacity at higher temperatures was measured with a differential scanning calorimeter from Perkin Elmer (Diamond DSC^®) according to the method described in Benisek et al. (2012). The measurements were performed at temperatures between 280 and 670 K on two pectolite samples from the Renard 65 pipe weighing 44 mg with compositions 52.65 wt% SiO₂, 0.06 wt% Al₂O₃, 0.26 wt% FeO, 0.17 wt% MnO, 0.31 wt% MgO, 32.47 wt% CaO, and 8.58 wt% Na₂O.

Evaluation of calorimetric data

The PPMS C_P data were used to calculate the standard entropy by solving the integral of C_P/T dT numerically from 0 to 298.15 K using a spline interpolation function of Mathematica^®. The resulting standard entropy usually agrees with published reference values within 0.21% for silicates (Dachs and Benisek 2011). The DSC C_P data were fitted to a polynomial of the following form:

$$C_{{\text{P}}} = a + b*T^{ - 2} + c*T^{ - 0.5} + d*T$$

Table 3 Thermodynamic parameters of pectolite (NaCa₂Si₃O₈(OH))

$$C_{{\text{P}}}=732.348+\frac{3146791.6}{{T}^{2}}-\frac{8306.619}{\sqrt{T}}-0.0777679 T$$

The a, b, c, and d coefficients in the C_P polynomial correspond to c1, c3, c5, and c2, respectively, in the hp62ver.dat file.

Appendix 2: Thermodynamic phase equilibria models

Phase equilibria models for system with components SiO₂, TiO₂, Al₂O₃, Cr₂O₃, FeO, MnO, MgO, CaO, Na₂O, K₂O, CO₂, H₂O) for four whole rock compositions were produced. The four bulk compositions represented fresh and moderately reacted xenoliths of both types, granitoid and gneiss (Table 4). The analyses are the subset of the larger bulk rock major element analyses database fully described in Appendix 4. Since P₂O₅ is not a modelled component by default in Perple_X, it is stoichiometrically deducted from the bulk chemistry in the form of apatite along with the corresponding CaO amount (3.333 CaO·P₂O₅, Table 4) using a procedure described in Dyck et al. (2020). The phase equilibria models were computed for analyzed contents of CO₂ and H₂O for fresh and moderately reacted xenoliths (Table 5) and additionally for fluid-saturated conditions for moderately reacted xenoliths (Table 6). The modelling output is represented as the modal composition of mineral phases at surface conditions in resulting Figs. 6 and 7. All phases below 5% or with very limited temperature stability intervals not greater than 10% are combined into “Other”.

Table 4 Whole rock composition of xenoliths used in the thermodynamic models

Analyses (2), (4), (6), and (8) used for modelling were recalculated compositions that excluded apatite, i.e., P₂O₅ were equated to zero and CaO was reduced stoichiometrically.

Table 5 Perple_X phase equilibria model parameters with H₂O and CO₂ as immobile analyzed components

Table 6 Perple_X phase equilibria model parameters in fluid-saturated conditions

Appendix 3: Mineralogical zonation models in chemical potential space

A generic chemical potential diagram in μ(SiO₂ + Al₂O₃) − μMgO space is a useful tool for explaining the formation of the disequilibrium mineralogical zonal patterns observed in the completely reacted xenoliths (Fig. 5a, c). To account for the mineral phases observed, we used the simplified MgO–SiO₂–Al₂O₃–CaO–Na₂O–K₂O–H₂O–CO₂ model system. The mineral phases involved in the reactions are quartz, plagioclase, pectolite, dioside, phlogopite, serpentine, and forsterite. Biotite and spinel are ignored. Graphical depiction of the mineral assemblage is done on a three-component compatibility diagram, illustrating positions of various stable minerals based on molar quantities of the components. We grouped oxides into MgO − (SiO₂ + Al₂O₃) − (CaO + Na₂O + K₂O) to make three components, and all the reactions occur in the presence of a fluid composed of a mixture of H₂O and CO₂. For simplification, plagioclase is excluded from the reacting assemblage. However, since pectolite is formed by reacting plagioclase, rather than quartz, with calcite, an alternative reaction producing pectolite would be a reaction between quartz and calcite in the presence of plagioclase. To retain the importance of plagioclase as a reacting phase, it is shown as a hollow dot on the compatibility diagram.

The compatibility diagram with all the involved phases mapped (Fig. 8a) is the basis for the μ(SiO₂ + Al₂O₃) − μMgO chemical potential diagram on which the topological relationships between the stable phases have to be preserved (Fig. 8b). To properly depict the slopes of the univariant reaction lines, stochiometric reactions between the phases, containing SiO₂, Al₂O₃, and MgO, and excess fluids and plagioclase must be balanced. The ratio of molar proportions of MgO to (SiO₂ + Al₂O₃) is the slope of a particular reaction.

Appendix 4: Bulk compositional profiles and conserved element ratio analysis

Two types of xenolith-kimberlite traverses were selected for analytical work, (i) granitoid xenoliths ranging from fresh/unaltered to those with a high degree of reaction and passing into the kimberlite reaction halo and further into hypabyssal kimberlite; and (ii) gneiss xenoliths ranging from fresh/unaltered to those with a high degree of reaction and passing into the kimberlite reaction halo and further into hypabyssal kimberlite. The samples included: for (i) CRGR1 and CRGR2—granitoid country rocks from drillcores, GRx2 and GRx3—slightly altered and moderately altered, respectively, granitoid xenoliths from drillcores, GRx1-c—moderately reacted core of a granitoid xenolith 5 mm from the contact, GRx1-r1 and GRx1-r2—highly reacted rims of a granitoid xenolith bordering kimberlite, HKh-65b—kimberlite reaction halo adjacent to granitoid xenolith, HK-65b1 to HK-65b3 and HK-65c—hypabyssal kimberlite; for (ii) CRGN1 and CRGN2—gneiss country rocks from drillcores, GNx2 and GNx3—a slightly altered and moderately altered, respectively, gneiss xenoliths from drillcores, GNx1-c—moderately reacted core of a gneiss xenolith 5 mm from the contact, GNx1-r—highly reacted rim of a gneiss xenolith bordering kimberlite, HKh-65b—kimberlite reaction halo adjacent to gneiss xenolith, HK-65b1 to HK-65b3, and HK-65c—hypabyssal kimberlite.

Bulk rock major elements were analysed in six samples (CRGR2, GRx2, GRx3, CRGN2, GNx2, GNx3) at Activation Laboratories, Ontario, Canada. Major oxides were determined by X-ray fluorescence (XRF) spectroscopy, FeO by titration, CO₂ content by CO₂-infrared, and structural and adsorbed water (H₂O⁺ and H₂O⁻) by infrared and gravimetry. CO₂ and H₂O are combined as a loss on ignition (LOI). FeO was determined through titration, using a cold acid digestion of ammonium metavanadate, and hydrofluoric acid in an open system. Ferrous ammonium sulphate is added after digestion and potassium dichromate is the titrating agent. H₂O-(moisture) was determined gravimetrically using a 2 g sample heated in an oven at 105 °C. For analysis of H₂O⁺, 0.3 g of the dried sample is thermally decomposed in a resistance furnace in a pure nitrogen environment at 1000 °C, using an ELTRA CW-800, directly releasing H₂O⁺. For infrared analysis of CO₂, 0.2 g sample is thermally decomposed in a resistance furnace in a pure nitrogen environment at 1000 °C, using an ELTRA instrument, directly releasing CO₂ after H₂O had been removed in a moisture trap. The concentration of CO₂ is detected as a reduction in the level of energy at the detector. The bulk rock data for samples CRGR1, CRGN1, HK-65b2, HK-65b3 and HK-65c were provided by Stornoway Diamond Corporation. Information on drill hole and depth, locations, the host phase of kimberlite for all analyzed samples can be found in Niyazova et al. (2021).

Major element compositions for five samples from the granitoid suite (GRx1-c, GRx1-r1, GRx1-r2, HKh-65b, HK-65b1) and four samples from the gneiss suite (GNx1-c, GNx1-r, HKh-65b, HK-65b1) were analysed in polished thin sections using specially designed electron microprobe techniques. The cores and rims of the two zonally altered granitoid and gneiss xenoliths (GRx1 and GNx1) and their adjacent host kimberlite (Kimb65b) were analysed using a sequence of 2 by 2 mm² square “frames”. Within each frame, sixty-four individual raster analyses were obtained on an 8 × 8 square grid and averaged to mimic a bulk rock analysis of a much-restricted area with the xenolith-kimberlite reaction products. Each frame was positioned over the mineralogy representative of a moderate and a high degree of reaction in the xenolith core (-c) and rims (-r), respectively, the adjacent hypabyssal kimberlite reaction halo surrounding the xenolith (HKh), and the main host (HK). For each of the nine frames major element oxide analyses were obtained on a fully automated CAMECA SX50 electron microprobe (EMP; Earth and Ocean Sciences Department, University of British Columbia), operating in a wavelength-dispersion mode with the following operating conditions: excitation voltage 15 kV, beam current 20 nA, peak count time 20 s, background count time 10 s, spot diameter 30 μm. Na and P were measured first. The resulting EPMA major oxide values are highly dependent upon the positioning of the points within the frame and the localisation of the frame upon an area capturing the representative mineralogy. To check the effect of the points positioning and the sensitivity of the analysis, we compared the averaged values of all the 64 points with two subsets of the points in the same frames. We averaged separately a subset of the odd-numbered and another subset of even-numbered points. The resulting relative difference between the 64-point averages and the subsets averages turned out to be not greater than 3% for SiO₂, 14% for MgO, and 16% for CaO. The isocon analyses were done on the geochemical profiles that are created using both the bulk rock data and the unconventional EPMA data of the averaged 64 raster points. To check whether the low precision of these EPMA analyses distorts the shapes of the geochemical profiles, we created the same profiles based only on the bulk rock compositions and compared them with the original “all-inclusive” profiles, i.e., the profiles that include EPMA samples. The shapes of the bulk-composition-only profiles did not change principally; therefore, we concluded that the low accuracy of the EPMA did not distort the geochemical interpretation. Our treatment and discussion of the geochemical data, including the isocon analyses, therefore, are based on the all-inclusive geochemical profiles. The analyses for all samples are provided in the Electronic Supplementary Material.

Conserved element ratio analysis was employed to counteract the closure effect in constraining chemical processes during xenoliths’ assimilation. To aid visualization of mass transfer, we used a translation procedure of Hilchie et al. (2018) applied to an isocon analysis (Grant 1986). In the procedure, a normalized reference composition is deducted from the normalized composition of the altered samples, thus demonstrating deviations from the original composition. Samples plotting above the zero-base line represent gains of material with respect to the reference sample, and those plotting below represent losses. The data and worked procedure of the translated isocon analysis are provided in the Electronic Supplementary Material.

The conserved element ratio analysis comprised several consecutive steps:

• (1) Selection of the fresh reference samples, representing the original composition prior to xenolith-kimberlite reactions. Fresh unreacted granitoid and gneiss country rock samples are the original reference for the respective xenoliths, and fresh hypabyssal kimberlite away from the xenoliths is the original reference composition for the host kimberlite.

• (2) Selection of the conserved elements. An ideal conserved element shows none or the least variation in the suite, i.e., its amount does not change along with reactions. In the granitoid-kimberlite zonation samples, phosphorus (P) varies the least in the granitoid, and Mn varies the least in the kimberlite. Therefore, P is chosen as the conserved element for the granitoid samples and Mn for the kimberlite samples. Mn content in country rock granites is below 0.01 wt%, too low for a numerator and too close for a minimum detection limit. In the gneiss-kimberlite suite, Mn shows the least variation at higher concentrations and is chosen as the conserved element for both lithologies.

• (3) Conversion to molar proportions of each element per 100 g of the material:

  $$\mathrm{mol\, per }\,100\mathrm{ g}\,=\,\frac{\mathrm{Oxide\, concentration }}{\mathrm{Molar\,mass\, of\, oxide}} \,\times\, \#\,\mathrm{ of \,cations\, per\, oxide\, formula}$$

• (4) Calculation of the molar Pearce Element Ratios (PER) by:

  $$\mathrm{PER}\,=\,\frac{\mathrm{mol\,per\,}100\mathrm{\,g\,of\, Element }}{\mathrm{mol\,per\,}100\mathrm{\,g\,of\, Conserved \,Element}}$$

• (5) Calculation of the Translated Conserved Element Ratios by subtracting the PER of the reference samples from each sample (Tables Isocons in ESM). The reference samples plot at zero while the altered samples fluctuate around the zero-base line. Note that the observed threefold maximal variations in P in granitoids (from 0.05 to 0.16 wt% P₂O₅) creates a large difference in the normalized SiO₂ (−1200/P) between CRGR1 and CRGR2. The change is only moderately lower than the magnitude of change in reacted granite xenoliths (samples GRx1-c, GRx1-r1, GRx1-r2) attributed to the outflow of Si.