Magma Mingling in Kimberlites: Evidence from the Groundmass Cocrystallization of Two Spinel-Group Minerals

: We present the results of a detailed petrographic study of fresh coherent samples of the Menominee kimberlite sampled at site 73, located in Menominee County, MI, USA. Our objective is to account for its unusual and complex paragenetic sequence. Several generations of olivine, ilmenite, and spinel-group minerals are described. Early olivine and ilmenite are xenocrystic and were replaced or overgrown by primary minerals. Zoned microcrysts of olivine have a xenocrystic core mantled by a ﬁrst rim in which rutile, geikielite, and spinel s.s. (spinel sensu stricto) cocrystallized. The in situ U–Pb dating of a microcryst of primary rutile yielded 168.9 ± 4.4 Ma, which was interpreted as the age of emplacement. The groundmass consists of olivine, spinel s.s., a magnesian ulvöspinel–ulvöspinel–magnetite (MUM) spinel, calcite, and dolomite. An extremely low activity of Si is suggested by the crystallization of spinel s.s. instead of phlogopite in the groundmass. The presence of djerﬁsherite microcrysts indicates high activities of Cl and S during the late stages of melt crystallization. The occurrence of two distinct spinel-group minerals (spinel s.s. and qandilite-rich MUM) in the groundmass is interpreted as clear evidence of the mingling of a magnesiocarbonatitic melt with a dominant kimberlitic melt.


Introduction
Kimberlites are mantle-derived hybrid rocks consisting of a mixture of crystals, some derived from the disaggregation of xenoliths and some that grew directly in the carrier magma [1]. These rocks clearly present a petrographic challenge due to the hybrid nature and common strong overprint by late hydrothermal and supergene processes. Such alteration processes lead to the crystallization of abundant secondary minerals and to the modification of the original bulk composition, making the study of the early magmatic processes difficult [2,3]. The study of rare fresh kimberlite samples

Methods
Petrographic and textural studies were carried out using optical and scanning electron microscopy (SEM). We used an E-SEM-Quanta 200 FEI-XTE-325/D8395 with a Back-Scattered Electron (BSE) detector, coupled to a Genesis energy-dispersive spectrometer (EDS) microanalysis system, at the Scientific and Technical Centers of the University of Barcelona (CCiTUB). The operating conditions were 20 kV, 1 nA beam current, 1 µm beam diameter, and 10 mm distance to detector. We determined the chemical composition of the spinel-group minerals, ilmenite-group minerals, rutile, and olivine using an electron microprobe (EMP) JEOL JXA-8230 (JEOL, Tokyo, Japan) equipped with five wavelength-dispersive spectrometers (WDS) and one energy-dispersive spectrometer (EDS), also at the CCiTUB. The detection limits for olivine were as follows: 107 ppm for Si, 184 ppm for Mg, 127 ppm for Mn, 118 ppm for Fe, 118 ppm for Ni and 58 ppm for Ca. In situ U-Pb analyses of primary rutile were carried out by secondary ion mass spectrometry (SIMS) using a CAMECA IMS 1280HR ion microprobe at the Institute of Geology and Geophysics, Chinese Academy of Sciences (IGGCAS). Details of the SIMS analytical method are described in [17][18][19]. The O − primary ion beam was accelerated at −13 kV, with an intensity of ca. 30 nA. The spot was approximately 20 µm × 30 µm and oval in shape. Each measurement consisted of 10 cycles and took approximately 15 min. The Pb/U ratios were calibrated using a DXK rutile standard (1782.6 ± 2.8 Ma) [20].
The olivine is relatively fresh. The anhedral macrocrystic olivine ranges from 2 cm down to 1 mm across. Euhedral microphenocrysts of olivine (0.2-0.5 mm) are fresh and display complex zoning (labeled as Ol1 to Ol3 from the core to the outermost rim, Figure 2a). The core of the olivine crystals (Ol1) has inclusions of pentlandite and pyrrhotite; these sulfides have a roundish shape and are fine-grained. A first rim of euhedral olivine (Ol2) mantles the core and contains inclusions of subhedral rutile crystals, which are set mostly at the contact between the core and this rim. The outermost rim of euhedral olivine (Ol3) contains inclusions of spinel sensu stricto (Al-Mg spinel) and Mg-rich ilmenite (Figure 2b). The groundmass olivine ranges from 10 to 50 µm across; it is partially replaced by a serpentine.
Rutile shows three different textural types: (1) euhedral rutile, as very common inclusions in olivine, at the contact between the olivine core (Ol1) and the first rim (Ol2) ( Figure 2a); (2) rutile microphenocrysts (about 200 µm), which are replaced by geikielite along their margins and along fractures, and then by qandilite ( Figure 2f); (3) rutile replacing macro-to microcrysts of Mg-rich ilmenite ( Figure 2e) and which, in turn, is replaced by geikielite and MUM.
Subhedral laths of calcite (about 50 to 100 µm in length) occur in the kimberlite groundmass. Anhedral dolomite in the groundmass fills the interstitial spaces between the above-mentioned minerals (Figure 2c). MUM is found commonly in contact with dolomite. Barite forms small anhedral crystals (<1 to 5 µm) associated with dolomite and also in association with secondary minerals such as meixnerite and serpentine.  Two different spinel-group minerals have been identified in the groundmass; one is spinel sensu stricto, and the other is labeled MUM. These two spinels occur in similar proportions in the groundmass, and there is no overgrowth of one on the other. The Al-Mg-rich spinel (82-90 mol.% MgAl 2 O 4 ) is commonly euhedral to subhedral in the groundmass, but it is also found as small inclusions in fresh olivine, together with geikielite ( Figure 2c). The groundmass spinel s.s. is found locally replaced by fine-grained aggregates of meixnerite (Mg 6 Al 2 (OH) 16 (OH) 2 · 4H 2 O) at the rim. The second spinel-group mineral found in the groundmass is MUM, and it shows an euhedral outline ( Figure 2c); it is also found as the last phase replacing Mg-rich ilmenite and rutile macrocrysts (Figure 2d-f).
Four textural types of ilmenite-group minerals are present: (1) homogeneous macro-to microcrystic Mg-rich ilmenite ( Figure 2g); (2) macro-to microcrystic Mg-rich ilmenite replaced along veinlets and grain borders by a sequence of rutile, geikielite, and qandilite (Figure 2d Rutile shows three different textural types: (1) euhedral rutile, as very common inclusions in olivine, at the contact between the olivine core (Ol1) and the first rim (Ol2) ( Subhedral laths of calcite (about 50 to 100 µm in length) occur in the kimberlite groundmass. Anhedral dolomite in the groundmass fills the interstitial spaces between the above-mentioned minerals ( Figure 2c). MUM is found commonly in contact with dolomite. Barite forms small anhedral crystals (<1 to 5 µm) associated with dolomite and also in association with secondary minerals such as meixnerite and serpentine.   Two spinel-group minerals occur in the groundmass ( Figure 4, Table 1), with compositions typical of Al-Mg spinel (spinel s.s.) and MUM (magnesian ulvöspinel-ulvöspinel-magnetite). The first one contains 82-90 mol.% MgAl2O4, 3-10 mol.% FeCr2O4, 5-6 mol.% Fe3O4, and 1-3 mol.% Mg2TiO4. The spinel s.s. included in olivine phenocrysts has a similar composition to that found in the groundmass. The second spinel in the groundmass, the euhedral grains of MUM, has significant  Two spinel-group minerals occur in the groundmass ( Figure 4,  TiO 4 . Where this spinel occurs as a replacement product of geikielite xenocrysts, it is even more enriched in Mg and approaches the qandilite end-member (Mg up to 1.26 apfu, following the trend G indicated in Figure 4). In the case of the euhedral MUM in the groundmass, it increases in Mg content toward its rim, but the most external Mg-rich MUM rim is too thin to analyze.    (Figure 5d). All the compositions of the ilmenite-group minerals are plotted in the kimberlite field, which was established by [23]. They show a positive correlation between TiO 2 and MgO contents (Figure 5c). The negative correlation between Mg and Fe 2+ points to the substitution of Fe 2+ by Mg from types 1 and 2 to types 3 and 4a and, finally, to types 4b and 4c (Figure 5e). All the textural types of ilmenite-group minerals have low levels of Mn, Nb, and Zr (<0.3 wt.%); types 4b and 4c are slightly enriched in Mn (Figure 5f). Representative compositions of ilmenite-group minerals are listed in Table 2.  the compositions of the ilmenite-group minerals are plotted in the kimberlite field, which was established by [23]. They show a positive correlation between TiO2 and MgO contents (Figure 5c). The negative correlation between Mg and Fe 2+ points to the substitution of Fe 2+ by Mg from types 1 and 2 to types 3 and 4a and, finally, to types 4b and 4c (Figure 5e). All the textural types of ilmenitegroup minerals have low levels of Mn, Nb, and Zr (<0.3 wt.%); types 4b and 4c are slightly enriched in Mn (Figure 5f). Representative compositions of ilmenite-group minerals are listed in Table 2.  The representative results of the analyses of rutile (Table 2) are plotted in Figure 6. The three textural types of rutile have similar compositions: 0.9-2.0 wt.% Nb 2 O 5 , 0.8-4.5 wt.% Cr 2 O 3 , and 0.6-0.9 wt.% V 2 O 3 . Nevertheless, the Cr 2 O 3 content decreases from type 1 (up to 4.5 wt.%), to type 2 (up to 3.0 wt.%) and to type 3 (up to 1.2 wt.%). With regard to Nb, type 1 rutile has up to 2.0 wt.% Nb 2 O 5 , type 2 has up to 1.7 wt.%, and type 3 has up to 1.2 wt.%. In the TiO 2 -Cr 2 O 3 plot, most of the rutile crystals show kimberlite groundmass compositions, as in [24]. On the other hand, the range also matches the composition of rutile from cratonic, deep-seated magmatic rocks (Figure 6a). The Cr 2 O 3 -Nb 2 O 5 plot (Figure 6b) shows both in the fields of xenogenic material and pyrope inclusions and outside the field of rutile from metasomatic nodules.

Mineral Chemistry
Minerals 2020, 10, x FOR PEER REVIEW 10 of 19 Figure 6. Chemical composition of the three different rutile types described in this work for the studied Menominee samples. (a) TiO2-Cr2O3 plot (composition fields are according to [24]). Domain labeled "a" corresponds to rutile from cratonic deep-seated magmatic rocks, domain "b" to rutile from off-craton alkali basalts, and domain "c" to groundmass rutile from kimberlites. (b) Cr2O3-Nb2O5 plot. Domain labeled 1 corresponds to rutile from the xenogenic material of kimberlites of South Africa [25]; domain 2 to rutile in metasomatic nodules from the kimberlites of the Orapa pipe, Botswana [25]; domain 3 to rutile inclusions in pyrope from Garnet Ridge [26]; and domain 4 to rutile inclusions in pyrope from the Internatsionalnaya pipe [27].
The composition of our djerfisherite is shown in Table 3. Djerfisherite is Ni-rich (5.5-7.9 apfu Ni) and Cu-poor (0.4-1.    Figure 6. Chemical composition of the three different rutile types described in this work for the studied Menominee samples. (a) TiO2-Cr2O3 plot (composition fields are according to [24]). Domain labeled "a" corresponds to rutile from cratonic deep-seated magmatic rocks, domain "b" to rutile from off-craton alkali basalts, and domain "c" to groundmass rutile from kimberlites. (b) Cr2O3-Nb2O5 plot. Domain labeled 1 corresponds to rutile from the xenogenic material of kimberlites of South Africa [25]; domain 2 to rutile in metasomatic nodules from the kimberlites of the Orapa pipe, Botswana [25]; domain 3 to rutile inclusions in pyrope from Garnet Ridge [26]; and domain 4 to rutile inclusions in pyrope from the Internatsionalnaya pipe [27].

The U-Pb Age of Rutile
Four analyses were conducted in situ on a microcryst of rutile (textural type 2). The uranium contents varied from 5.2 to 5.5 ppm, and the Th/U values were consistent and equal to 0.05. The measured 207 Pb/ 206 Pb of four spots was in range of 0.08 to 0.14, corresponding to f 206 (the percentage of common 206 Pb in total 206 Pb), around 4% to 12%. The regression line with these rather clustered data gives an imprecise upper intercept of 2.4 ± 2.8 as the common Pb composition, which is obviously unreasonable. An anchored composition of common Pb expressed by the ratio 207 Pb/ 206 Pb as 0.84 ± 0.05 assigned as the SK (Stacey and Kramers) model [36] was used. The regression line derived from the data points on the Tera-Wasserburg (TW) plot (mean standard weighted deviation MSWD = 1.06) shows a lower intercept age of 168.9 ± 4.4 Ma (Figure 8 and Table 4). According to the evaluation of the variation in common Pb composition, the influence on the final age from common Pb should be less than 1%. The main uncertainty is from the Pb/U elemental instrumental fractionation and statistical counting error.

Paragenetic Sequence
The fresh samples from the site 73 Menominee kimberlite allow us to propose a complete paragenetic sequence ( Figure 9) and reconstruct the petrogenetic history of kimberlite magma. We interpret the Mg-rich ilmenite of textural types 1 and 2 and the olivine cores (Ol1) with inclusions of pentlandite and pyrrhotine to be xenocrysts of mantle formation. Olivine Ol1 has a composition typical of mantle peridotite, whereas the Mg-rich ilmenite is a typical mineral formed during a metasomatic process in the source [37]. These xenocrysts of olivine and ilmenite are replaced or mantled by minerals that crystallized from kimberlitic magma. Subhedral rutile crystallized at the early stage of growth of the first forsterite rim (Ol2), whereas geikielite and spinel s.s. started to crystallize during the late stage of the growth of that rim. Rutile in kimberlitic rocks is commonly interpreted as being xenocrystic, derived from the disaggregation of a wide variety of rocks either of crustal or mantle origin (eclogites, MARID (Mica-Amphibole-Rutile-Ilmenite-Diopside), pyroxenites, and metasomatized peridotites) [25,38]. On the other hand, rutile has been also described as inclusions within or intergrown with diamond [25,38]. Its composition is broadly used to constrain the source rock. Chromium-poor rutile is interpreted to be derived from both crustal and off-cratonic or cratonic mantle rocks, whereas Cr-rich rutile (>1.7 wt.% Cr2O3) is considered to be exclusively related to cratonic mantle [24]. Rutile derived from metasomatized mantle-derived xenoliths may contain up to 9.75 wt.% Cr2O3 [25,26]. These determinations and the potential for U-Pb dating have categorized mantle rutile as a diamond indicator mineral (e.g., [24]). Although the rutile crystals studied here are found mostly in the cratonic mantle compositional field, they occur as subhedral inclusions in growth bands in zoned olivine (Ol2). Therefore, these crystals can hardly be described as being xenocrystic. Moreover, the composition of the rutile microcrysts is the same as the composition of rutile inclusions in olivine. Hence, we suggest that all these rutile crystals formed during the early stages of magma crystallization, prior to the crystallization of the first rim on olivine (Ol2). This was followed by the crystallization of geikielite and spinel s.s. As a consequence, the

Paragenetic Sequence
The fresh samples from the site 73 Menominee kimberlite allow us to propose a complete paragenetic sequence ( Figure 9) and reconstruct the petrogenetic history of kimberlite magma. We interpret the Mg-rich ilmenite of textural types 1 and 2 and the olivine cores (Ol1) with inclusions of pentlandite and pyrrhotine to be xenocrysts of mantle formation. Olivine Ol1 has a composition typical of mantle peridotite, whereas the Mg-rich ilmenite is a typical mineral formed during a metasomatic process in the source [37]. These xenocrysts of olivine and ilmenite are replaced or mantled by minerals that crystallized from kimberlitic magma. Subhedral rutile crystallized at the early stage of growth of the first forsterite rim (Ol2), whereas geikielite and spinel s.s. started to crystallize during the late stage of the growth of that rim. Rutile in kimberlitic rocks is commonly interpreted as being xenocrystic, derived from the disaggregation of a wide variety of rocks either of crustal or mantle origin (eclogites, MARID (Mica-Amphibole-Rutile-Ilmenite-Diopside), pyroxenites, and metasomatized peridotites) [25,38]. On the other hand, rutile has been also described as inclusions within or intergrown with diamond [25,38]. Its composition is broadly used to constrain the source rock. Chromium-poor rutile is interpreted to be derived from both crustal and off-cratonic or cratonic mantle rocks, whereas Cr-rich rutile (>1.7 wt.% Cr 2 O 3 ) is considered to be exclusively related to cratonic mantle [24]. Rutile derived from metasomatized mantle-derived xenoliths may contain up to 9.75 wt.% Cr 2 O 3 [25,26]. These determinations and the potential for U-Pb dating have categorized mantle rutile as a diamond indicator mineral (e.g., [24]). Although the rutile crystals studied here are found mostly in the cratonic mantle compositional field, they occur as subhedral inclusions in growth bands in zoned olivine (Ol2). Therefore, these crystals can hardly be described as being xenocrystic. Moreover, the composition of the rutile microcrysts is the same as the composition of rutile inclusions in olivine. Hence, we suggest that all these rutile crystals formed during the early stages of magma crystallization, prior to the crystallization of the first rim on olivine (Ol2). This was followed by the crystallization of geikielite and spinel s.s. As a consequence, the occurrence of phenocrystic rutile in Menominee precludes the use of the aforementioned rutile-based diagrams unless previous accurate petrographic study ensures that the rutile crystals are xenocrysts. In any case, all the rutile types tend to be replaced by geikielite (ilmenite-group mineral 4), which also replaces the above categories of ilmenite. The composition of the replacing ilmenite in terms of minor elements inherits the composition of the replaced minerals, as noted in other kimberlites [37,39]. diagrams unless previous accurate petrographic study ensures that the rutile crystals are xenocrysts.
In any case, all the rutile types tend to be replaced by geikielite (ilmenite-group mineral 4), which also replaces the above categories of ilmenite. The composition of the replacing ilmenite in terms of minor elements inherits the composition of the replaced minerals, as noted in other kimberlites [37,39]. The site 73 kimberlite emplacement age has been estimated to be 155 Ma using zircon fissiontrack dating and 190 Ma using K-Ar of phlogopite [8]. However, these two estimates derive from a personal communication based on analyses from 1987, providing only two numbers of age without any analysis dates. Moreover, the closure temperature of the zircon fission-track is about 220 °C, which is easily affected by the later modification; thus, the age obtained by this method is usually lower. The K-Ar method may be uncertain because of the heterogeneous element content of the sample or the excess Ar; thus, the age obtained by this method is usually higher. Therefore, the U-Pb system of stable mineral rutile was selected in this work to determine the emplacement age. The rutile from Menominee was dated by SIMS U-Pb dating at 168.9 ± 4.4 Ma. This age is broadly consistent with the 155 and 190 Ma reported earlier in [8], and also with the nearby Lake Ellen kimberlite at 186 ± 6, 190 ± 5 (unpublished data in [10]), and 266 ± 9 Ma [11].

Magma Mingling Revealed by the Coexistence of Two Spinels in the Groundmass
The coexistence of two different spinel-group minerals (spinel s.s. and qandilite-rich MUM) in the groundmass of the Menominee kimberlite samples provides evidence for the mingling of two magmas, each spinel crystallizing from a separate immiscible magma. Magma mingling is not rare and has also been invoked to explain the simultaneous occurrence of two types of perovskite in groundmass [5,40]. The study of the mineral sequence points to rather coeval spinels formed at the stage of groundmass crystallization.
Qandilite-rich MUM is found in the groundmass of some Michigan kimberlites, such as Menominee and Lake Ellen [41]. Although the crystals have variable compositions (28-56 mol.% of the qandilite component), the highest magnesium content is found in the MUM replacing geikielite. However, the occurrence of a MUM spinel replacing geikielite is very common in kimberlites, without achieving such magnesium enrichment. Therefore, the occurrence of groundmass qandiliterich MUM indicates the existence of a high activity of Mg in the magma: The site 73 kimberlite emplacement age has been estimated to be 155 Ma using zircon fission-track dating and 190 Ma using K-Ar of phlogopite [8]. However, these two estimates derive from a personal communication based on analyses from 1987, providing only two numbers of age without any analysis dates. Moreover, the closure temperature of the zircon fission-track is about 220 • C, which is easily affected by the later modification; thus, the age obtained by this method is usually lower. The K-Ar method may be uncertain because of the heterogeneous element content of the sample or the excess Ar; thus, the age obtained by this method is usually higher. Therefore, the U-Pb system of stable mineral rutile was selected in this work to determine the emplacement age. The rutile from Menominee was dated by SIMS U-Pb dating at 168.9 ± 4.4 Ma. This age is broadly consistent with the 155 and 190 Ma reported earlier in [8], and also with the nearby Lake Ellen kimberlite at 186 ± 6, 190 ± 5 (unpublished data in [10]), and 266 ± 9 Ma [11].

Magma Mingling Revealed by the Coexistence of Two Spinels in the Groundmass
The coexistence of two different spinel-group minerals (spinel s.s. and qandilite-rich MUM) in the groundmass of the Menominee kimberlite samples provides evidence for the mingling of two magmas, each spinel crystallizing from a separate immiscible magma. Magma mingling is not rare and has also been invoked to explain the simultaneous occurrence of two types of perovskite in groundmass [5,40]. The study of the mineral sequence points to rather coeval spinels formed at the stage of groundmass crystallization.
Qandilite-rich MUM is found in the groundmass of some Michigan kimberlites, such as Menominee and Lake Ellen [41]. Although the crystals have variable compositions (28-56 mol.% of the qandilite component), the highest magnesium content is found in the MUM replacing geikielite. However, the occurrence of a MUM spinel replacing geikielite is very common in kimberlites, without achieving such magnesium enrichment. Therefore, the occurrence of groundmass qandilite-rich MUM indicates the existence of a high activity of Mg in the magma: The presence of spinel s.s. in the groundmass of a kimberlite is relatively rare, although not unknown (Igwisi [21,42] and Tli Kwi Cho [21]). The early crystallization of aluminous spinels in a kimberlite groundmass has been attributed to the absence of phlogopite crystallization [21,43].
Spinel s.s. is found in the groundmass of some lamprophyres, as in polzenite and melilitolite from the Ploučnice River region, Czechoslovakia [44]. However, in the Czech case, the spinel is zoned and has a Ti-rich rim of MUM type. Here, the higher Al content of the magma can account for the occurrence of spinel s.s. It is also found as an inclusion in leucite in the West Kimberly lamproite [45,46]; the authors attributed these crystals to exsolution from a non-stoichiometric leucite containing Mg, Fe, and Ti in solid solution. They argued that the exsolution could be favored by undercooling. However, it can be argued that the undercooling would favor exactly the contrary-the preservation of metastable high-T phases. The exsolution hypothesis is ruled out in the present case because of textural evidence and paragenetic position: both spinels occur as abundant euhedral crystals in the groundmass, and each term tends to be associated with different minerals. Moreover, the Al/Ti contents in spinels have been used to constrain the activity of these elements in the original magma; these proportions are different depending on the source of the magma [47]. Hence, the occurrence of two different spinels can be indicative of different magmas from different sources.
The crystallization of djerfisherite in the Menominee kimberlite could produce the depletion of K in the magma and therefore favor the crystallization of spinel. However, djerfisherite formed at a relatively late phase in this kimberlite, whereas spinel formed at an early stage. Djerfisherite occurrences in kimberlites worldwide are generally late to form, resulting from the replacement of early-formed sulfides [48]. The early crystallization of spinel s.s. could produce the depletion of Al in the magma, which favors the crystallization of K sulfides rather than phlogopite.
By assuming that these two spinels were produced by the crystallization of two different magmas, we next questioned the nature of the two magmas. The two spinels are unzoned, and they do not follow the typical trends described of kimberlites; the Igwisi [21,42] and Tli Kwi Cho kimberlites [21] are exceptional. The early crystallization of an aluminum spinel in a kimberlite groundmass has been explained as a result of the lack of the phlogopite crystallization [21,43]. Some authors [21] suggested that the trend from chromite to aluminous spinel is the result of rapid growth at a high degree of supersaturation of spinel. However, a chromite core has not been found in the crystals of spinel s.s. in our samples. Rather, the occurrence of qandilite-rich MUM spinel at the last stages of magma crystallization might indicate the existence of a highly evolved kimberlitic magma, such as that mentioned for the Jos and Benfontein kimberlites [1]. However, MUM is found in contact with dolomite at site 73. Qandilite-rich spinel associated with a dolomitic groundmass is also known in a kimberlite sill at Wemindji, Quebec (Canada) [44]. Similar compositions of qandilite-rich MUM were found in association with carbonates in the kimberlite groundmass at the Igwisi Hills and Benfontein [42,49]. This assemblage is consistent with the expected conditions of the crystallization of qandilite, favored by extremely high f (CO 2 ) and extremely low a (SiO 2 ) [50]. We believe that the mingling of an immiscible dolomite-rich melt (magma 2) might have led to the crystallization of dolomite and qandilite. Qandilite-rich MUM and dolomite appear as late minerals in the crystallization sequence. We propose an injection of a carbonatite magma rich in the dolomite component during the crystallization of the kimberlitic magma.

Volatiles in the Kimberlitic Melt
We next considered another important observation-the occurrence of a complex alkali-and chlorine-bearing sulfide such as djerfisherite. It is found in several geological environments-e.g., Si-poor igneous rocks, meteorites [51], carbonatites [52,53], kimberlites [29,54,55], metasomatized mantle rocks [56], melt inclusions in basaltic rocks [57], and ultramafic lamprophyres [58]. Djerfisherite is also found in chloride-carbonate "nodules" in kimberlite [4]. In mantle xenoliths, it rims Fe-Ni-Cu sulfides and sulfide globules [29][30][31][32][33][34]59], is a daughter phase in melt inclusions [28,[30][31][32]35,[59][60][61][62][63], and occurs as an inclusion in diamonds [35]. However, experimental data indicated that djerfisherite is not stable at pressures greater than 3 Gpa [63]. Thus, the djerfisherite included in diamond and mantle xenoliths could not have crystallized in the cratonic lithospheric mantle [64], but might have formed by interaction between xenoliths and a kimberlite melt [64,65]. Some authors proposed that djerfisherite formed as a replacement product of pre-existing Fe-Ni-Cu xenolithic sulfides by a K-Cl bearing melt or by the direct crystallization of the kimberlite melt [66]. A late magmatic origin of djerfisherite, at a shallow depth and at less than 800 • C, was also suggested to explain the occurrence of this mineral in the groundmass of the Udachnaya-East kimberlite [29]. The same has been suggested for the groundmass djerfisherite from the Lac de Gras field [67]. Finally, the presence of djerfisherite might indicate a Cl enrichment in the kimberlite melt, as djerfisherites described in Udachnaya-East and Kuoikskoe kimberlites [29,64]. The groundmass djerfisherite in the Udachnaya-East kimberlite is not contaminated by crustal components, as indicated by S isotopes, which are within the mantle range and are thus primary [68]. These authors proposed that djerfisherite crystallized at the late stage of the kimberlite evolution, during the eruption, when the evolved kimberlitic magma became sufficiently enriched in S, Na, K, and Cl [68]. A potassium-enriched upper mantle at 120-140 km is proposed by the study of xenoliths from the nearby Lake Ellen kimberlite [12]. The occurrence of djerfisherite at Menominee indicates a high activity of volatiles (S and Cl) and alkalis during melt crystallization, as well as a low Si activity, which could also explain the common occurrence of spinel s.s. instead of phlogopite. The higher volatile content probably led to more rapid eruption, thus favoring the preservation of diamond [40]. The occurrence of djerfisherite in kimberlite also provides evidence that the melt was enriched in alkalis and halogens [29,64,66].

Conclusions
A petrographic examination of rock samples from Menominee revealed a complex paragenetic sequence that strongly suggests the injection of a magnesiocarbonatitic liquid into a resident kimberlitic melt. We attribute the coexistence of two spinel-group minerals in the Menominee kimberlite groundmass to the mingling of two magmas. Spinel s.s. likely crystallized from the kimberlitic melt, whereas qandilite-rich MUM crystallized from the carbonate melt. The crystallization of spinel s.s. instead of phlogopite in the groundmass is consistent with the low activity of Si in this kimberlitic melt. The occurrence of djerfisherite microcrysts as an accessory phase indicates a high activity of volatile elements (Cl and S) during the late stages of the crystallization of the magma. The emplacement of the Menominee kimberlite occurred at 168.9 ± 4.4 Ma, according to the U-Pb data for a rutile microphenocryst.