Styles of Alteration of Ti Oxides of the Kimberlite Groundmass : Implications on the Petrogenesis and Classification of Kimberlites and Similar Rocks

The sequence of replacement in groundmass perovskite and spinel from SK-1 and SK-2 kimberlites (Eastern Dharwar craton, India) has been established. Two types of perovskite occur in the studied Indian kimberlites. Type 1 perovskite is found in the groundmass, crystallized directly from the kimberlite magma, it is light rare-earth elements (LREE)-rich and Fe-poor and its ∆NNO calculated value is from −3.82 to −0.73. The second generation of perovskite (type 2 perovskite) is found replacing groundmass atoll spinel, it was formed from hydrothermal fluids, it is LREE-free and Fe-rich and has very high ∆NNO value (from 1.03 to 10.52). Type 1 groundmass perovskite may be either replaced by anatase or kassite along with aeschynite-(Ce). These differences in the alteration are related to different f (CO2) and f (H2O) conditions. Furthermore, primary perovskite may be strongly altered to secondary minerals, resulting in redistribution of rare-earth elements (REE) and, potentially, U, Pb and Th. Therefore, accurate petrographic and chemical analyses are necessary in order to demonstrate that perovskite is magmatic before proceeding to sort geochronological data by using perovskite. Ti-rich hydrogarnets (12.9 wt %–26.3 wt % TiO2) were produced during spinel replacement by late hydrothermal processes. Therefore, attention must be paid to the position of Ca-Ti-garnets in the mineral sequence and their water content before using them to classify the rock based on their occurrence.


Introduction
Ti-rich minerals from kimberlite such as ilmenite, spinels, rutile and perovskite are important carriers of petrogenetic information.Xenocrystic Ti-rich oxides in kimberlitic rock, such as Cr-rich rutile, Ti-rich spinel and ilmenite, provide information about the metasomatic processes in the cratonic lithospheric mantle [1,2].In addition, Ti-rich oxides from kimberlite groundmass, such as Ti-rich spinel [3] and perovskite [4], could supply information about the evolution of kimberlitic magmas.Finally, perovskite [5,6] and rutile [7,8] could also be used to determine the kimberlite emplacement age.However, in many cases these minerals undergo complex alteration processes during the hydrothermal or supergene late stages of the kimberlite crystallization sequence that could disturb the petrogenetic interpretations based on geochemical data.
Perovskite is a principal host of LREE in SiO 2 -undersaturated ultramafic and alkaline rocks [9] and is a mineral that may be produced along different crystallization stages of these magmas.Therefore, the chemical and textural study of these generations can provide information about several stages of magmatic crystallization [10,11].The crystallization of magmatic perovskite is produced later than that of macrocrystal spinel; and simultaneously with "reaction" Fe-rich spinel and groundmass spinels of the magnesian ulvöspinel-magnetite series in kimberlite [12].
However, perovskite cannot be formed only by magmatic processes.A metasomatic origin has been inferred for perovskite from carbonatites [22,23] and perovskite from skarns is also widely described as a hydrothermal product [9].
Primary Ti-rich garnet is considered as a key mineral for the classification of ultramafic lamprophyres, such as aillikites [24] and orangeites [25].Ti-rich garnets classified as andradite, schorlomite, zirconian schorlomite and kimzeyite occur in the Torngat aillikite dykes [24]; kimzeyite and Ti-andradite from Aillik Bay are considered as primary magmatic minerals and are indicators of the aillikitic affinity of these rocks [26].A primary magmatic origin has also been suggested for the Ca-Ti-Fe rich garnets found in orangeite from Swartruggens [27].However, Ti-rich hydrogarnets can be produced by subsolidus reactions in a wide span of environments, as in metapyroxenites [28], serpentinites [29], basalts from the oceanic seafloor [30] and magmatic alkaline rocks and carbonatites [31].
Therefore, the Ti-rich minerals can supply information of petrological or economic interest.However, these minerals can be easily altered by subsolidus processes.In addition, the hydrothermal occurrences of these minerals in other geological environments suggest that these minerals can be produced by the reaction of the existing magmatic minerals with late hydrothermal or supergene fluids.In the present work, we describe the different alteration styles of groundmass Ti-rich oxide (perovskite and spinel) from SK-1 and SK-2 kimberlites (Eastern Dharwar craton, India), including the neoformation of pristine perovskite and Ti-rich hydrogarnets by subsolidus processes.The petrogenetic implications are discussed.

Geological Setting
Kimberlites occur in the Bastar craton (central India) and the Eastern Dharwar craton (EDC, southern India) in India.Kimberlite intrusions (around 100 bodies) found in the EDC of southern India are distributed in three fields: (1) the southern Wajrakarur kimberlite field; (2) the northern Narayanpet kimberlite field; and (3) the Raichur kimberlite field (RKF) [17,32].This work studied two sequences of alteration in samples from two nearby pipes (SK-1 and SK-2, about 1100 m of distance) from Siddanpalli, RKF, EDC, southern India (Figure 1).Emplacement of these two kimberlites took place during the Mesoproterozoic around 1.1 Ga [17].

Methods
Minerals have been identified in situ on thin polished section by SEM-BSE-EDS, EMPA and Raman spectroscopy at the Scientific and Technological Centers of the University of Barcelona (CCiTUB), Barcelona, Spain.
Petrographic and textural studies were carried out using optical and scanning electron microscopy (SEM), the latter employing an E-SEM-Quanta 200 FEI-XTE-325/D8395 (FEI, Hillsboro, OR, USA) with a BSE detector and coupled to a Genesis EDS microanalysis system.The operating conditions were 20-25 kV, 1 nA beam current and 10 mm distance to detector.

Methods
Minerals have been identified in situ on thin polished section by SEM-BSE-EDS, EMPA and Raman spectroscopy at the Scientific and Technological Centers of the University of Barcelona (CCiTUB), Barcelona, Spain.
Petrographic and textural studies were carried out using optical and scanning electron microscopy (SEM), the latter employing an E-SEM-Quanta 200 FEI-XTE-325/D8395 (FEI, Hillsboro, OR, USA) with a BSE detector and coupled to a Genesis EDS microanalysis system.The operating conditions were 20-25 kV, 1 nA beam current and 10 mm distance to detector.
Micro Raman analyses were obtained at the CCiTUB by using a HORIBA Jobin Yvon LabRam HR 800 dispersive spectrometer (HORIBA, Kyoto, Japan), equipped with an Olympus BXFM optical microscope.Non-polarized Raman spectra were obtained by applying a 532 nm laser, the pixel size was 1 µm.The exposure time was 5 s with 3 scans and laser power at sample was 2.5 mW for anatase.The exposure time was 5 s with 10 scans and laser power at sample was 5 mW for kassite.The exposure time was 20 s with 5 scans and laser power at sample was 1.25 mW for garnet.

Mineral Textures
Both kimberlites have similar xenocrysts, mainly consisting of Ti oxides such as rutile and ilmenite, scattered in a fine-grained groundmass.Primary groundmass minerals are also similar in both kimberlites and the dominant minerals include calcite, apatite, perovskite and altered atoll-shaped spinel-group minerals.The abundance of minerals of the serpentine group in the groundmass suggests that olivine was also common in these kimberlites.
Micro Raman analyses were obtained at the CCiTUB by using a HORIBA Jobin Yvon LabRam HR 800 dispersive spectrometer (HORIBA, Kyoto, Japan), equipped with an Olympus BXFM optical microscope.Non-polarized Raman spectra were obtained by applying a 532 nm laser, the pixel size was 1 µm.The exposure time was 5 s with 3 scans and laser power at sample was 2.5 mW for anatase.The exposure time was 5 s with 10 scans and laser power at sample was 5 mW for kassite.The exposure time was 20 s with 5 scans and laser power at sample was 1.25 mW for garnet.

Mineral Textures
Both kimberlites have similar xenocrysts, mainly consisting of Ti oxides such as rutile and ilmenite, scattered in a fine-grained groundmass.Primary groundmass minerals are also similar in both kimberlites and the dominant minerals include calcite, apatite, perovskite and altered atoll-shaped spinel-group minerals.The abundance of minerals of the serpentine group in the groundmass suggests that olivine was also common in these kimberlites.

MicroRaman Study
Identity of anatase was confirmed by microRaman spectroscopy and its spectrum was compared with that of submicroscopic anatase, mixed with calcite, altering perovskite in carbonatites [14,15] (Figure 3a).Calcite has a peak at 1088 cm −1 which corresponds to the vibration of the CO 3 group [15].However, calcite is absent from the products of alteration of the perovskite from Siddanpalli.
SK-2 may also be replaced by a sequence of typical groundmass minerals and sometimes by type 2 perovskite (Figure 2i,j); on its turn, this type 2 perovskite may be replaced by anatase.

MicroRaman Study
Identity of anatase was confirmed by microRaman spectroscopy and its spectrum was compared with that of submicroscopic anatase, mixed with calcite, altering perovskite in carbonatites [14,15] (Figure 3a).Calcite has a peak at 1088 cm −1 which corresponds to the vibration of the CO3 group [15].However, calcite is absent from the products of alteration of the perovskite from Siddanpalli.Kassite [CaTi2O4(OH)2] has a very similar chemical composition to cafetite (CaTi2O5•H2O).The identification of kassite from the SK-2 kimberlite was confirmed by comparing the kassite Raman spectrum with the available spectrum of kassite standards [15] (Figure 3b).
Raman spectra of Ti-rich garnet from SK-2 kimberlite is also recorded (Figure 3c) and shows a peak at 3576 cm −1 which corresponds to the OH vibration [30].The spectrum was compared with those of other hydrogarnets [33,34].
The microRaman study also confirmed the identification of the second generation of perovskite (type 2) by comparison with the spectrum of the magmatic perovskite (type 1) (Figure 3d).Minor The identification of kassite from the SK-2 kimberlite was confirmed by comparing the kassite Raman spectrum with the available spectrum of kassite standards [15] (Figure 3b).
Raman spectra of Ti-rich garnet from SK-2 kimberlite is also recorded (Figure 3c) and shows a peak at 3576 cm −1 which corresponds to the OH vibration [30].The spectrum was compared with those of other hydrogarnets [33,34].
The microRaman study also confirmed the identification of the second generation of perovskite (type 2) by comparison with the spectrum of the magmatic perovskite (type 1) (Figure 3d).Minor differences in the position and intensity of the bands can also be related to changes in the chemical composition or to different orientation.
Finally, the Raman analysis was helpful to establish the identity of aeschynite-(Ce).In this case, most of the bands have similarities with those from the published standards of members of the aeschynite group.Minor differences in the positions of the bands can be explained because the studied aeschynite is Ta-poor and La-and Nd-rich when compared with the standard aeschynites [35,36].

Mineral Chemistry
Spinel-group minerals from groundmass of both kimberlites have compositions in the ulvöspinel-titanomagnetite domain.However, they are strongly altered and the restitic cores may be seldom enriched in Zn (0.1 wt %-2.0 wt % ZnO in SK-1 and 0 wt %-7.5 wt % ZnO in SK-2).
Minerals 2018, 8, x FOR PEER REVIEW 7 of 15 differences in the position and intensity of the bands can also be related to changes in the chemical composition or to different orientation.Finally, the Raman analysis was helpful to establish the identity of aeschynite-(Ce).In this case, most of the bands have similarities with those from the published standards of members of the aeschynite group.Minor differences in the positions of the bands can be explained because the studied aeschynite is Ta-poor and La-and Nd-rich when compared with the standard aeschynites [35,36].

Mineral Chemistry
Spinel-group minerals from groundmass of both kimberlites have compositions in the ulvöspinel-titanomagnetite domain.However, they are strongly altered and the restitic cores may be seldom enriched in Zn (0.1 wt %-2.0 wt % ZnO in SK-1 and 0 wt %-7.5 wt % ZnO in SK-2).
Type 2 perovskite from the SK-2 kimberlite has different composition than type 1 perovskite.It is depleted in LREE (<1 wt % ∑LREE2O3) and Nb (<0.1 wt % Nb2O5) but it has higher Fe contents (2.0 wt %-4.0 wt % Fe2O3) than type 1 perovskite (Figure 4, Table 1).Frequency histograms of log fO 2 expressed relative to the NNO buffer (∆NNO) calculated using the perovskite oxybarometer developed by Bellis and Canil [4] are shown in Figure 5. Type 1 perovskite in SK-1 and SK-2 has similar ∆NNO value (from −0.73 to −3.40 in SK-1 and from −1.07 to −3.82 in SK-2) but type 2 perovskite has very high ∆NNO value (from 1.03 to 10.52) and was formed in a highly oxidizing environment.Frequency histograms of log fO2 expressed relative to the NNO buffer (ΔNNO) calculated using the perovskite oxybarometer developed by Bellis and Canil [4] are shown in Figure 5. Type 1 perovskite in SK-1 and SK-2 has similar ΔNNO value (from −0.73 to −3.40 in SK-1 and from −1.07 to −3.82 in SK-2) but type 2 perovskite has very high ΔNNO value (from 1.03 to 10.52) and was formed in a highly oxidizing environment.Kassite replacing type 1 perovskite in SK-2 has a stoichiometric composition (Table 1), whereas the associated aeschynite-(Ce) tends to concentrate LREE and, to a lesser extent, Nb.The aeschynite-(Ce) produces the next average structural formula: (Ca0.39Ce0.33La0.13Nd0.12Pr0.04)∑1.01(Ti1.82Nb0.07Fe3+ 0.02Zr0.02)∑1.93(O,OH)6(Table 1).Y, Er, Dy and Pb contents in aeschynite-(Ce) have been analysed but they are below detection limit.Therefore, the aeschynitegroup minerals from SK-2 kimberlite are poor in Nb, U and Th when compared to similar minerals typically occurring as metamictic phases in carbonatites [37] and metasomatised rocks [38].However, their compositions are similar to those of the late Ti-REE minerals described in the Iron Mountain kimberlite field [13].
Ti-rich hydrogarnets from the SK-2 kimberlite were analysed by EMPA, while those from the SK-1 pipe are too small to be analysed (Table 2).According to the IMA nomenclature for garnet group minerals [39], schorlomite end member has 2 apfu Ti in Y position, while andradite end member has 2 apfu Fe 3+ in Y position.The Ti-rich garnets (12.9 wt %-26.3 wt % TiO2) studied in the present work could correspond to a theoretical hydrous andradite (when it has < 1 apfu Ti) and hydrous schorlomite (when it has > 1 apfu).However, Ti-rich hydroandradite from the SK-2 kimberlite returns low total (88 wt %-96 wt %) and Si is also very low (1.6-1.9 apfu), thus suggesting the substitution of Si by OH in Z position and the existence of H2O molecules.They plot inside the field of Ti andradites from ultramafic lamprophyres [25] (Figure 6).Kassite replacing type 1 perovskite in SK-2 has a stoichiometric composition (Table 1), whereas the associated aeschynite-(Ce) tends to concentrate LREE and, to a lesser extent, Nb.The aeschynite-(Ce) produces the next average structural formula: (Ca 0.39 Ce 0.33 La 0.13 Nd 0.12 Pr 0.04 ) ∑1.01 (Ti 1.82 Nb 0.07 Fe 3+ 0.02 Zr 0.02 ) ∑1.93 (O,OH) 6 (Table 1).Y, Er, Dy and Pb contents in aeschynite-(Ce) have been analysed but they are below detection limit.Therefore, the aeschynite-group minerals from SK-2 kimberlite are poor in Nb, U and Th when compared to similar minerals typically occurring as metamictic phases in carbonatites [37] and metasomatised rocks [38].However, their compositions are similar to those of the late Ti-REE minerals described in the Iron Mountain kimberlite field [13].
Ti-rich hydrogarnets from the SK-2 kimberlite were analysed by EMPA, while those from the SK-1 pipe are too small to be analysed (Table 2).According to the IMA nomenclature for garnet group minerals [39], schorlomite end member has 2 apfu Ti in Y position, while andradite end member has 2 apfu Fe 3+ in Y position.The Ti-rich garnets (12.9 wt %-26.3 wt % TiO 2 ) studied in the present work could correspond to a theoretical hydrous andradite (when it has <1 apfu Ti) and hydrous schorlomite (when it has >1 apfu).However, Ti-rich hydroandradite from the SK-2 kimberlite returns low total (88 wt %-96 wt %) and Si is also very low (1.6-1.9 apfu), thus suggesting the substitution of Si by OH in Z position and the existence of H 2 O molecules.They plot inside the field of Ti andradites from ultramafic lamprophyres [25] (Figure 6)."bdl": below detection limit.

Discussion
Perovskite is a valuable mineral recorder of the crystallization conditions of the kimberlites.It commonly crystallizes directly from the kimberlite magma [40].Therefore, pristine primary perovskite grains in both kimberlites and carbonatites are often used for geochemical investigations and, in particular for U-Pb dating [5,6,10,12,13,16-21,41,42].However, it has been proved that different perovskite generations can occur in the same kimberlite.Simultaneous occurrence of two

Discussion
Perovskite is a valuable mineral recorder of the crystallization conditions of the kimberlites.It commonly crystallizes directly from the kimberlite magma [40].Therefore, pristine primary perovskite grains in both kimberlites and carbonatites are often used for geochemical investigations and, in particular for U-Pb dating [5,6,10,12,13,[16][17][18][19][20][21]41,42].However, it has been proved that different perovskite generations can occur in the same kimberlite.Simultaneous occurrence of two populations of primary perovskite has been explained by magma mingling [5].In addition, primary magmatic perovskite can be altered [12,13] during subsolidus processes to secondary minerals that may redistribute REE and potentially, U, Pb and Th [43].Our petrographic data shows that two types of texturally fresh (i.e., pristine) perovskite occur in the studied Indian kimberlites.Groundmass type 1 crystals may be interpreted as primary magmatic perovskite.However, type 2 perovskite occurs along with calcite and serpentine filling porosity produced by replacement of Ti-rich spinels.This assemblage suggests that type 2 perovskite could be produced by subsolidus hydrothermal phenomena and thus not necessarily related to the primary perovskite.Similar pristine secondary hydrothermal perovskites have been described in carbonatites and cannot be used to obtain the age of the intrusive [23].Therefore, our observations further restrict the use of groundmass perovskite for geochronological purposes, since they show for the first time that pristine perovskite can be also formed in kimberlites by hydrothermal processes.Therefore, an accurate petrographic study is necessary to exclude perovskite affected by subsolidus processes.Hence, we suggest taking additional cautions when using perovskite grains in concentrates.
The alteration of perovskite is strongly dependent upon pH, f (CO 2 ) and temperature.It is expected to occur in late-stage hydrothermal alteration processes and in the subaerial weathering environment [9].The replacement of perovskite occurs at the late stage of groundmass formation, resulting from a decrease in f (O 2 ) and temperature (<350 • C) at low pressure (P < 2 kbars) and over a wide range of a(Mg 2+ ) values [12].The replacement of olivine by serpentine as well as the replacements of perovskite and spinel suggest that P (CO 2 ) and P (H 2 O) remained relatively high [13].The two Indian kimberlites studied here have significant differences regarding their subsolidus history, mainly represented by the higher complexity of perovskite alteration in SK-2.The alteration process took place under different fluid/rock ratios in each kimberlite, in a relatively closed system.Under these conditions, Ti-rich minerals are unstable and, in particular, Ti-rich spinels are easily replaced in both kimberlites by mixtures of Ti-rich hydrogarnets, calcite and serpentine.A relatively low SiO 2 and high water activities were necessary to avoid the crystallization of titanite and to favour the crystallization of hydrogarnets.Slight replacement of perovskite by anatase in SK-1 could be indicative of a decrease of temperature under conditions of medium to high f (CO 2 ), following the thermodynamic calculations data [15].However, the same experimental data suggest that the strong replacement of perovskite by kassite in SK-2 needs a high f (H 2 O) and a low activity in alkalis.The formation of kassite or anatase during perovskite replacement may also involve different rates of Ca-leaching, as kassite formation is favoured by lower Ca-leaching [13].The LREE-rich perovskite is more unstable during these processes than pure end-member perovskite.Therefore, aeschynite-(Ce) inherits the composition of the replaced LREE-bearing perovskite cores and it is Nb-poor because the cores were also Nb-poor.
The occurrence of abundant Ti-rich garnets in the groundmass of the rock could suggest an aillikitic affinity during a preliminary examination, based on the International Union of Geological.Sciences (IUGS) rock classification [44].Ti-rich garnets in groundmass from Indian kimberlites have been used to classify the rocks as orangeites [25].Those Ti-rich garnets have similar composition to the Ti-rich garnets studied in the current work, which also plot inside the field of high Ti-andradite from kimberlite-UML rocks (Figure 6).However, hydrogarnets from SK-1 and SK-2 kimberlites replace Ti-rich oxides and are accompanied by hydrothermal minerals such as serpentine and type 2 perovskite, thus indicating that they were produced by late hydrothermal processes.Therefore, these garnets cannot be representative of the parental magma composition.In fact, Ti-rich hydrogarnets have also been found in ophiolite sequences as a result of hydrothermal alteration [33].Hence, attention must be paid to the position of Ca-Ti-rich garnet in the mineral sequence before using it to classify the rock based on its occurrence.

Conclusions
Two types of perovskite occur in the SK-1 and SK-2 Indian kimberlites.The first type crystallized directly from the kimberlite magma, whereas a hydrothermal origin was inferred from the second.Additionally, two different replacement trends of groundmass perovskite have been identified.Type 1 groundmass perovskite is replaced by anatase in SK-1 and by kassite along with aeschynite-(Ce) in SK-2.The different sequences are related to alteration under different f (CO 2 ) and f (H 2 O) conditions.In some cases, perovskite may be strongly altered to secondary minerals, resulting in a redistribution of REE and potentially, U, Pb and Th.Therefore, U-Pb dating studies involving perovskite require a detailed petrographic characterisation to confirm its primary (i.e., magmatic) origin.
Ti-rich hydrogarnets in SK-1 and SK-2 replace groundmass atoll spinel and could correspond to "hydroandradite" and "hydroschorlomite."They were produced by hydrothermal processes.Therefore, attention must be paid to the position of Ca-Ti-garnet in the mineral sequence and to its water content before using it to classify the rock based on its occurrence.