Testing Trace-Element Distribution and the Zr-Based Thermometry of Accessory Rutile from Chromitite

: Trace element distribution and Zr-in-rutile temperature have been investigated in accessory rutile from stratiform (UG2, Merensky Reef, Jacurici), podiform (Loma Peguera), and metamorphic chromitites in cratonic shields (Cedrolina, Nuasahi). Rutile from chromitite has typical ﬁnger-print of Cr-V-Nb-W-Zr, whose relative abundance distinguishes magmatic from metamorphic chromitite. In magmatic deposits, rutile precipitates as an intercumulus phase, or forms by exsolution from chromite, between 870 ◦ C and 540 ◦ C. The Cr-V in rutile reﬂects the composition of chromite, both Nb and Zr are moderately enriched, and W is depleted, except for in Jacurici, where moderate W excess was a result of crustal contamination of the maﬁc magma. In metamorphic deposits, rutile forms by removal of Ti-Cr-V from chromite during metamorphism between 650 ◦ C and 400 ◦ C, consistent with greenschist-amphibolite facies, and displays variable Cr-Nb, low V-Zr, and anomalous enrichment in W caused by reaction with felsic ﬂuids emanating from granitoid intrusions. All deposits, except Cedrolina, contain Rutile+PGM composite grains (<10 µ m) locked in chromite, possibly representing relics of orthomagmatic assemblages. The high Cr-V content and the distinctive W-Nb-Zr signature that typiﬁes accessory rutile in chromitite provide a new pathﬁnder to trace the provenance of detrital rutile in placer deposits.


Introduction
Titanium was discovered by William Gregor in 1791, and since the second half of the twentieth century, it has become a basic commodity essential to several modern technologies in the metal-alloys, steel, and pigment industries [1]. Rutile, the most common tetragonal form of titanium dioxide (TiO 2 ), and ilmenite (FeTiO 3 ) are the principal natural sources for titanium recovery. Rutile is a relatively abundant mineral in the Earth's crust [2], occurring as an accessory phase in various igneous and high-grade metamorphic rocks, although half of the present day annual production comes from detrital sedimentary sources in shoreline and fluvial placer deposits.
Besides its indisputable economic importance, rutile plays an important role in geosciences as a powerful petrogenetic indicator [3]. Although characterized by a simple chemical composition, TiO 2 , rutile may carry a variety of trace elements replacing Ti in the crystal lattice. The extent of the substitution of some specific elements is an effective guide to petrology, geochronology, and geothermometry of the rutile-bearing host rock, and can be used in some cases to identify the source rock of detrital rutile in sedimentary deposits [2][3][4][5][6][7][8][9][10][11].
It has been experimentally demonstrated [12] that the solubility of ZrO 2 in rutile is temperature-dependent, particularly when rutile is properly buffered with zircon (ZrSiO 4 ) and silica (SiO 2 ). Furthermore, recent investigations have shown that the Zr-based thermometer can be successfully applied to rutile from silica-free, ultramafic rocks such as chromitites [13,14], since the obtained temperatures were in agreement with the geological setting and the temperatures calculated with other geothermometers. Before and after these contributions, accessory rutile had been reported from a number of chromitite occurrences [15][16][17][18][19][20][21][22]. However, detailed trace-element systematic has been provided only for single chromite deposits of Brazil and India [13,23]. This contribution presents the results of the first systematic study of the distribution of Zr and other trace elements in accessory rutile from chromitite of different ages and geological settings. The analyzed samples represent chromitite associated with high-grade metamorphic terranes of India and Brazil mentioned above, stratiform chromitite from continental layered intrusions of South Africa and Brazil, and one podiform chromitite from the Caribbean ophiolite belt of the Dominican Republic.
The main objectives were to (1) characterize the trace-element finger-print of rutile derived from chromitites with different magmatic and post-magmatic histories, (2) test the validity of the Zr-in-rutile thermometer applied to silica under-saturated rocks composed mainly of chromite, (3) evaluate the role of metamorphism in the re-distribution of Zr and other trace elements in chromitite-hosted rutile, and (4) discuss rutile-chromite relationships (texture, composition) and provide further argument to the debate on the origin of accessory rutile in chromitite: a primary magmatic phase, a product of subsolidus exsolution from the host chromite during magmatic cooling, or a result of regional metamorphism?

Methodology
Rutile was previously located on thin sections and polished blocks by reflected-light microscope and confirmed by Raman spectroscopy. Afterwards, electron microprobe analyses of chromite and rutile were carried out at the Eugen F. Stumpfl laboratory (Leoben University, Leoben, Austria), using a Jeol JXA 8200 Superprobe (Jeol, Tokyo, Japan). Wavelength dispersive system (WDS) was used for quantitative analyses. Rutile grains less than 10 microns in size were qualitatively analyzed by energy dispersive system (EDS). The same instrument was used to obtain electron images.
Chromite was analyzed with 15 kV accelerating voltage and 10 nA beam current. All the elements were analyzed using the Kα line, and were calibrated on natural chromite, rhodonite, ilmenite, kaersutite, and olivine. The following diffracting crystals were used: TAP for Mg and Al; PETJ for Si; and LIFH for Ti, V, Cr, Mn, and Fe. Counting times of 20 and 10 s were used on peak and backgrounds, respectively. The amount of Fe 3+ in chromite was calculated assuming the ideal spinel stoichiometry.
Preliminary electron microprobe analysis of accessory rutile indicated the systematic occurrence of trace amounts of Mg, Al, V, Cr, Mn, Fe, Zr, Nb, and W. Qualitative analyses failed in finding other elements commonly reported from terrestrial rutiles (e.g., Hf, Ta, Th). Since most of the recorded metals are among the major constituents of the host chromite, and the analyses were performed in situ, special attention was paid to avoid spurious fluorescence from the matrix and unwanted interferences between analytical lines. Rutile was analyzed with 20 kV accelerating voltage and 30 nA beam current. The counting times were increased up to 60 and 30 s on peak and backgrounds, respectively. The Kα line was used for Mg, Al, Ti, V, Cr, Mn, and Fe, and the Lα line for Nb, Zr and W. The following standards were selected: natural olivine, chromite, rutile, rhodonite, magnetite and zircon, synthetic metallic Nb, V, and W. The TAP crystal was used for Mg and Al; the PETJ for Nb, Mn, and Zr; and the LIFH for Cr, Ti, W, Fe, and V. The detection limits of elements occurring as trace concentrations were automatically calculated by the Jeol software as follows: Mg, Al, V, Cr, Zr, and Fe = 50 ppm; Nb, Ti, and Mn = 100 ppm; and W = 150 ppm. The same instrument was used to obtain the Back-Scattered Electron (BSE) images. Monitoring the accuracy and precision of several tens of analyses carried out under the abovementioned analytical conditions, it was concluded that only rutile grains larger than 10 µm yielded reliable results, free of fluorescence effects and chemical contamination from the chromite host.

Geological Setting and Petrographic Notes
In order to provide a statistically significant collection of data, only chromitites containing high proportion of accessory rutile were considered in this work (Table 1). Podiform chromitites in ophiolitic mantle tectonite from supra-subduction-zone (SSZ) were found to be unsuitable due to the absence or scarcity of accessory rutile. Three of the investigated deposits represent unmetamorphosed, magmatic chromitite deposited in distinctive geological settings and ages. They are Paleoproterozoic stratiform deposits associated with the layered intrusions of Bushveld (Transvaal Province, South Africa), the ultramafic-mafic sill of Jacurici (Bahia State, Brazil), and a Mesozoic podiform chromitite of Loma Peguera in the Loma Caribe ophiolite massif (Dominican Republic). Both Bushveld and Jacurici complexes intrude high-grade metamorphic terranes; however, texture, mineralogy, and chemical variations proper of cumulate rocks are generally well preserved, suggesting that regional metamorphism was not effective in these cases. The podiform chromitite of Loma Peguera is unaffected by high-grade regional metamorphism, but nonetheless the host peridotite shows widespread low-temperature serpentinization typical of sub-oceanic upper mantle, and lateritic weathering under supergene conditions. Two of the examined chromite deposits are examples of chromitites involved in highgrade regional metamorphism, following primary magmatic deposition. The Nuasahi chromite deposit consists of a strongly deformed and metamorphosed layered sill associated with the Archaean cratonic shield of Singhbhum (Orissa province, Eastern India). The Cedrolina chromitite is located in the Archean-Paleoproterozoic greenstone belt of Pilar de Goiás (Central Brazil). The amphibolite facies metamorphic overprint of the area, metasomatic reactions with exotic fluids, and poly-cyclical tectonic deformations have prevented the identification of the original chromite magmatic affinity (podiform or stratiform).

Bushveld Layered Intrusion (UG2 Chromitite Layer, Merensky Reef)
The Bushveld Complex, located in the Transvaal Province of South Africa, is the largest mafic-ultramafic layered intrusion in the world. It consists of a 9 km thick sequence of layered mafic and ultramafic cumulates that hosts, in specific mineralized reefs, world-class deposits of chromium, platinum group elements (PGE), nickel, copper, and vanadium. Despite ages as old as 2.06 Ga, magmatic features are well preserved, showing minimal deformation and metamorphic effects [24,25].
The occurrence of accessory rutile in the chromitites was documented by various authors (e.g., [14][15][16][19][20][21]). Analytical data have been provided for rutile in pyroxenite from various localities of the UG2 and Merensky Reef units [14]. In this work, we analyzed rutile from massive chromitite collected along the Rustenburg Section of the Rustenburg Platinum Mine during the Third International Platinum Symposium, Pretoria, July 1981. They represent the UG2 chromitite layer and the chromite seams associated with the Ni-Cu-PGE deposit of Merensky Reef (UG2 and MR labels).
The UG2 is one of the uppermost chromitite layers in the Critical Zone of the Bushveld Complex and contains the largest single PGE resource in the world, while Cr is a byproduct [25][26][27]. In the Rustemburg Section, the UG2 chromitite is about 0.75 m thick and is located below the Merensky Reef, at an estimated stratigraphic distance of about 120 m [26]. The examined UG2 samples consist of a massive 60 to 90 modal percent chromite showing a granular texture derived from accumulation of euhedral to subhedral chromite variable in size from few microns up to 1 mm. Internal layering is marked by alternating zones of massive chromite with zones of high-grade disseminated chromite with increasing modal percent of interstitial silicate mainly composed of pyroxene and plagioclase ( Figure 1A).
In the Rustemburg Section, the Merensky Reef consists of a pegmatoidal feldspatic pyroxenite of about 0.3 m average thickness, densely spotted with Ni-Cu magmatic sulfides and grains of platinum-group minerals (PGM). The Reef is bounded by two chromite layers, about 0.2-2 cm thick, marking the top and the bottom contacts with feldspatic pyroxenite hangingwall and spotted anorthosite footwall, respectively. Our chromitite samples consist of euhedral and subhedral grains, up to 1 mm in size, arranged in a typical cumulus texture ( Figure 1B). Chromite grains are completely fresh and do not show any sign of alteration; however, the associated silicate matrix, mainly coarse pyroxene and plagioclase, contains minor amounts of hydrous minerals (actinolite, micas, talc, chlorite, and a serpentine subgroup phases) that have been recognized as a result of deuteric reaction with late-stage hydrous melt during solidification of the Merensky Reef [28][29][30].

Jacurici Mafic-Ultramafic Sill (São Francisco Craton)
The Jacurici Complex consists of a suite of ultramafic bodies intruding Archaean-Paleoproterozoic terranes of the São Francisco Craton in the NE of Bahia State, Brazil [31][32][33]. According to zircon U-Pb data, the crystallization age of the Jacurici Complex can be placed at about 2085 Ma [34]. Structural reconstruction indicates that the ultramafic bodies of Jacurici represent dismembered segments of a single mafic-ultramafic layered sill, having an estimated thickness of about 300 m, and extending laterally over about 100 km [35][36][37].
Ideal stratigraphy comprises a cumulus sequence of dunite, harzburgite, pyroxenite, and gabbro norite, with granulitic gneisses at the footwall, and calc-silicate marble at the hanging-wall. Despite its age and tectonic deformation, the Jacurici mafic-ultramafic sill does not appear to have suffered the high-grade metamorphism that affects the host rocks of the São Francisco Craton. However, the presence of partially melted xenoliths of carbonate country rocks and geochemical isotopic data support significant contamination of the mafic magma with crust-compatible lithophile elements [37].
A huge chromitite layer up to 2-8 m thick is found in the upper part of the ultramafic section (dunite, harzburgite), representing the largest chromite deposit in Brazil. The chromitite is generally massive and typical cumulus textures are common, showing decreasing grain size-grading, passing from massive to disseminated ore ( Figure 1C), or locally characterized by accumulation of equigranular euhedral crystals ( Figure 1D). Chromite is always well preserved, devoid of deformation and alteration, whilst most of the interstitial silicates (olivine, orthopyroxene) were transformed into chlorite by low temperature alteration. The altered silicate matrix contains scattered crystals of zircon up to 100 µm in size. Accessory rutile in the chromitite of the Jacurici Complex has been reported by [22] and interpreted as an intercumulus mineral. Three samples of massive chromitite (JC label) from boreholes at the localities of Ipuera/Socò and Varzea do Macaco/Teiù were found to contain accessory rutile.

Loma Peguera (Loma Caribe Ophiolite)
The 100 km long peridotite belt of Loma Caribe in the Central Cordillera of the Dominican Republic is one of the ophiolite-related, ultramafic massifs exposed along the northern margin of the Caribbean Plate [38]. Unlike most ophiolite complexes worldwide, the Loma Caribe ophiolite is exclusively composed of harzburgitic mantle tectonite, with subordinate dunite and lherzolite, but lacks the other elements typical of ophiolites, i.e., the cumulus transition zone, sheeted dyke complex, pillow lavas, and pelagic sediments. According to the recently proposed geodynamic classification of ophiolites [39], the Loma Caribe peridotite belt may represent a rare case of mantle-plume emplaced in the oceanic plateau of the Mesozoic Caribbean basin [40][41][42].
Prolonged exposition to supergene alteration and weathering generated a thick layer of laterite on the top of the ultramafic rocks, leading to economic deposits of silicate-Ni [43,44]. In the mid1990s, mining exploration for Ni-laterite revealed the presence of small pods and lenses of chromitite (<2 m) randomly distributed within small masses of serpentinized dunite, in the area of Loma Peguera.
Two samples labelled LP were collected from the chromitites exposed at the outcrops visited during the field trip in the framework of the 18va Conferencia Geologica del Caribe in 2008. The samples consist of massive ore with less than 10% interstitial silicate, locally affected by fracturing and stretching ( Figure 1E). Despite widespread brittle deformation, the chromite is generally fresh, and cumulus-like texture is observed in micro-domains consisting of tiny polygonal crystals in a silicate matrix mostly composed of chlorite and minor serpentine ( Figure 1F). Several minute rutile crystals up to 20 µm in size were observed in between or partially included in chromite grains.
The Nuasahi Complex forms an N-S elongated body extending over about 60 km NE of the city of Sukinda. A W-E cross section of the Nuasahi mining area [48] shows ã 500 m thick ultramafic zone composed of enstatitite, olivine orthopyroxenite, dunite with chromitite layers, and orthopyroxenite, limited to the E by a several decametres-thick zone of tectonic breccia that marks the transition to the Suite-2 gabbroic formation. Structural data indicate that the breccia resulted from the gabbro intrusion [47] and consists of a gabbroic matrix engulfing angular to sub-rounded fragments of massive, and banded chromitite. The Breccia Zone is intruded by pegmatitic gabbro and quartz-diorite veins and displays widespread evidence of greenschist-amphibolite grade metamorphism [48]. The chromitite sample analyzed in this contribution (label NS9) comes from a strongly deformed area of the Suite-2 gabbro, in the hanging-wall of the Breccia Zone, exposed to the south of the Nuasahi mining area [23].
Under the microscope, the chromitite is massive to disseminated and composed of about 70% chromite by volume. Chromite grains appear as irregular fragments with jagged boundaries and extremely variable size, from a few hundred microns to more than one millimeter ( Figure 2A-C), and interpreted as a result of recrystallization [23]. The interstitial silicate matrix essentially consists of enstatite and minor Mg-hornblende ( Figure 2B), with minimal formation of low-temperature hydrous silicates, i.e., chlorite, suggesting that the metamorphic grade was higher than hydrothermal. The presence of rutile in chromitites from Nuasahi and Singhbhum craton was previously reported by [23,49].

Cedrolina Ultramafic Body (Pilar de Goiás Greenstone Belt)
The Pilar de Goiás greenstone belt is exposed in the Tocantins Province, State of Goiás, central Brazil. The belt consists of a volcano-sedimentary sequence metamorphosed under greenschist to amphibolite facies conditions. It is intruded by small bodies of albite-granite and affected by a multistage deformation during Paleoproterozoic and Neoproterozoic times [50].
The Cedrolina chromitite body is located within the lower stratigraphic units of the greenstone belt, where metakomatiites and metabasalts (talc-chlorite-schist, amphibolite and amphibole-schists) predominate over inter-bedded metasediments (metacherts, BIF). The chromitite has a tabular geometry of about 230 × 100 m and 1-2.4 m thickness, being in contact with micaschists to the SE, and amphibole rich BIF to the NW [13,51]. At the scale of hand samples, the chromitite consists of rounded nodules, up to 1 cm in diameter, embedded in a matrix of chlorite and talc, with a mean modal composition of 60% chromite, 25% talc, and 15% Cr-rich chlorite.
Under the microscope, the chromite nodules display a complex internal texture possibly resulting from coalescence of polygonal chromite crystals that still preserve sharp angular boundaries and triple junctions ( Figure 2D). The nodules may be strongly fractured, with fissures mostly filled with chlorite. The borders may be extremely rugged, suggesting late corrosion or tectonic deformation, or smooth, with the nodules displaying less intense fissuring ( Figure 2E). Different patterns of corrosion/deformation are seen causing dense riddling of the spinel body with microscopic holes ( Figure 2F, left side) or developing minute fissures according to the cubic symmetry of spinel faces ( Figure 2F, center and top).
Single crystals of chromite, variable in size from few microns up to 1 mm and, eventually, with hexagonal and octahedral shapes, have been also described [13,51], but are not shown in the pictures. According to these authors, the Cedrolina chromitite suffered intense metasomatism related to hydrothermal fluids emanated by albite-granite intrusions, which generated a complex suite of exotic accessory minerals such as monazite-Ce [52], uraninite, thorianite, and zircon. The samples studied here (CD label) are taken from the same collection of [13].

Chromite Mineral Chemistry
Representative compositions of chrome spinels from the rutile-bearing chromitite are listed in Table 2 and presented as relevant binary diagrams in

Unmetamorphosed (Magmatic) Chromitite
Based on the concept that the chromite mineral chemistry is strictly related to the composition of the parental magmas ( [53], and references therein) compositional variations observed in the chromitites devoid of high-grade metamorphism can be interpreted as a reflection of magmatic processes leading to precipitation of massive chromitite. Paleoproterozoic-Archaean layered intrusions of the world formed by mixing and fractional crystallization of variable proportions of high-Mg, basaltic andesite (U-type magma) with tholeiite-type mafic magmas [27], that may drive the rhythmical stabilization of chromitite during fractionation [54].
The UG2 and MR chromitites crystallized from relatively evolved parent melts, in the highest stratigraphic levels of the Critical Zone. They have high concentrations of TiO 2 , Fe 2 O 3 , and V 2 O 3 , (Figure 3B-D). The high oxidation state of the MR chromite (Fe 2 O 3 > 15 wt%) and the abundance of hydrous mafic silicates in the pyroxenite host suggest crystallization under high fluid activity and oxygen fugacity [28][29][30]55].
The oversized chromitite layers of the Jacurici ultramafic sill require deposition from an enormous volume of magma that is not consistent with the actual size of the intrusive body [37]. According to these authors, the sill acted as a conduit through which large amounts of a primitive, high-Mg magma flowed upwards to the crust, for a long time, and formed a thick zone of chromitite by close events of chromite precipitation. The low concentration of TiO 2 , V 2 O 3 , and Fe 2 O 3 in the chromite ( Figure 3B-D) reflects poor differentiation degree similar to magmas parental to low-Ti chromitites from other Proterozoic layered intrusions (e.g., Campo Formoso [56]). Cr saturation in the mafic magma was enhanced, if not triggered, by changes in magma composition associated with the crustal contamination event in the parent melt of the intrusion [37].
Precipitation of massive chromitite in the Mesozoic mantle-plume of Loma Caribe was a result of reaction between harzburgitic mantle and a percolating mafic magma, in the same way as podiform chromitite in upper mantle tectonite of ophiolites ( [57], and references therein). The unusual geodynamic setting of this mantle fragment [39]) is responsible for the peculiar composition of the chromite ore. The high-Cr nature of the chromite (Cr# = 0.76-0.82) would be consistent with crystallization from "boninite", commonly invoked as the parent melt of high-Cr podiform chromitite from supra-subduction-zone (SSZ). However, the high Fe 2 # = 0.63 and high titanium (TiO 2 = 0.68 wt%) and vanadium (V 2 O 3 = 0.18 wt%) ( Figure 3A,B,D) indicate a more evolved parent melt compared with boninite, which cannot be readily reconciled with a SSZ geodynamic setting [39,41,42].

Metamorphic Chromitite
The Nuasahi chromite deposit originally formed by magmatic accumulation of chromite, as usual for stratiform chromitite in layered intrusions, however, the high-grade metamorphism and recrystallization that have obliterated the igneous cumulus texture induced changes in the primary magmatic composition of chromite [23]. The low Cr# = 0.48-0.58 may be a magmatic relic consistent with the stratiform affinity of the chromite. Nevertheless, the high Fe 2 # = 0.65-0.72 relative to low Cr# ( Figure 3A) is a result of secondary Fe 2 enrichment and/or loss of Mg during metamorphism [23], as expected for chrome spinels from high-grade metamorphic rocks [53,58]. Furthermore, a distinctive depletion in titanium, from 0.71 to 0.04 wt% TiO 2 ( Figure 3B), has shifted the NS compositions out of the ideal TiO 2 versus Cr 2 O 3 trend formed by the unmetamorphosed stratiform chromitites of JC, UG2, and MR. We suggest that this depletion was caused by significant remobilization of Ti from the NS chromite during metamorphic recrystallization at high temperature, which likely led to subsequent rutile crystallization.
The samples from Cedrolina show the most scattered patterns in chromite composition despite of the extremely small size of the deposit [13]. The chromite composition has the highest chromium and iron numbers (Cr# = 0.77-0.91, Fe 2 # = 0.56-0.85) due to depletion in MgO and Al 2 O 3 , while the oxidation state increases with decreasing Cr 2 O 3 in the range of Fe 2 O 3 = 0.68-12.45 wt% ( Figure 3A,B). These wide changes in chromite composition cannot be reconciled with magmatic differentiation processes in a small batch of magma. According to [13], they are consistent with processes leading to the formation of "ferrian-chromite" under polycyclic regional metamorphism, hydrothermal metasomatism, and final supergene alteration, as already observed in other chromite deposits of Brazil [17,37,56,59]. In particular, an increasing oxidation state (i.e., Fe 2 O 3 ) coupled with Cr 2 O 3 decrease is consistent with chromium oxidation to Cr 6+ and mobilization during soil formation.

Rutile Modal Abundance and Textural Relations
Despite small grain sizes (sometimes < 10 µm), rutile was easily recognized under the reflected-light microscope because of its yellow-orange plane-polarized internal reflection ( Figure 4A,B). On this basis, simple point-counting analysis yielded modal abundances between 0.5% and 2% by volume, increasing from the magmatic chromitites (in the order LP, JC, UG2, MR) to the metamorphic ones (CD, NS). Similarly, the maximum grain size of rutile roughly follows the same trend, increasing from below 20 µm in LP chromitite up to more than 200 µm in those of NS. Consistently, morphology and textural relations with the host chromite vary remarkably from magmatic chromitites (UG2, MR, JC, LP) to the metamorphic ones (CD, NS), and show a variegated typology of cases much more complex compared with rutile described from mafic and ultramafic rocks of the Bushveld complex ( [14], and references therein).
In cumulus chromitites of UG2, MR, and JC, rutile has a very similar mode of occurrence. Most commonly, it occurs as irregular patches typically molded onto chromite grains accumulated by gravity settling (Figure 5A), or as euhedral to sub-euhedral grains interstitial to chromite ( Figure 5B,C), resembling an intercumulus phase in both cases. At Merensky Reef, rutile forms anhedral grains overgrowing chromite at the contact with clinopyroxene of the silicate matrix ( Figures 5E and 1B boxes 1,2,3) or occurs in silicate pockets (orthopyroxene, amphibole replaced by chlorite) occluded among chromite grains, although maintaining contact with chromite ( Figure 5D). Another interesting association in the Merensky Reef consists of rutile flame-like lamellae located inside ilmenite interstitial to chromite and large sulfide aggregates ( Figure 5F). Ilmenite appears to have replaced rutile and includes a droplet of Pt-Fe and Ni-sulfide, while amphibole and chlorite mark the contact ilmenite-sulfide. Rutile in the LP chromitite may occur as irregular grains included in partially altered massive chromite, sometimes in contact with altered silicates ( Figure 5G). Complex assemblages of chromite-rutile, both replaced by ilmenite, occur in shear zones dominated by interlocking fine crystals of chromite and altered silicates forming a "mortar-like" texture ( Figure 5H).
In the metamorphic chromitite of Cedrolina, [13] described relatively large rutile grains (usually > 10 µm) in variable textural relations with altered chromite. Rutile may occur inside fractured chromite nodules or adjacent to grain boundaries, typically in contact with the chlorite-talc matrix ( Figure 6A-D). Notably, some rutile grains include fragments of chromite, possibly indicating crystallization during or soon after chromite deformation ( Figure 6B,D). Rutile crystals of various sizes also occur in the chlorite-talc matrix around chromite nodules or filling crosscutting fractures. These rutiles are frequently found in association with zircon ( Figure 6E,F), or appear to have been replaced by ilmenite with proceeding hydrothermal metasomatism ( Figure 6G,H). The rutile in the metamorphic chromitite fragment of the Nuasahi Breccia Zone essentially consists of large (up to >200 µm) anhedral grains included in altered chromite ( Figure 7A), filling chromite fractures ( Figure 7B), or located at the chromite-silicate contact ( Figure 7C,D). The typically irregular morphology suggests coarse-grain recrystallization and further alteration of the chromite and silicate matrix.    All of the studied types of chromitite were found to host rutile as minute inclusions that rarely exceed 10 µm in size. In the magmatic chromitites, rutile commonly occurs as isolated polygonal crystals engulfed in unaltered solid chromite, away from cracks and fissures ( Figure 8A). The largest rutile inclusions (up to~50 µm) were observed in the Nuasahi sample, characterized by rounded or irregular morphology, and frequently associated with fissures in the altered domains of chromite ( Figure 8B-D). At Cedrolina, minute rutile grains (<10 µm) along with chlorite fill the holes in riddled chromite ( Figure 8F). Furthermore, idiomorphic rutile with a tiny droplet of uraninite occurs in the core of a large chromite grain with increasing alteration at the rim ( Figure 8G,H). Conspicuously, tiny rutile grains with irregular shape are visible in the chlorite-talc matrix ( Figure 8G, left and upper corner), indicating two different stages of secondary rutile crystallization. A peculiar type of rutile inclusion was observed in metamorphic chromitites (NS, CD), characterized by swarms of acicular rutile needles developed along cubic cleavage planes of highly altered chromite ( Figure 8E).  The present examples reported from Loma Peguera ( Figure 9A) and Nuasahi ( Figure 9B,C) confirm previous studies [23,60,61], which documented primary magmatic inclusions of rutile plus laurite, Os-Ir alloy, Ru-rich pentlandite (Bushveld), rutile plus laurite-irarsite-pyrrhotite or laurite-cuproiridsite (Loma Peguera), and rutile plus lauriteirarsite-pentlandite, or laurite-silicate (Nuasahi).

Trace Element Composition of Rutile from Chromitite
A total of 328 electron microprobe analyses of trace elements were performed on rutile grains larger than 10 µm, in which fluorescence effects were absent. Compositions of small rutile inclusions, e.g., the minute needles in Figure 6C, had to be discarded because of the systematic contamination of Cr. Representative compositions of large grains (Table 3) show that rutile contains two suites of trace elements, one consisting of chromitite-compatible metals (Cr, V, Al, Fe, Mg, Mn), the another represented by high field-strength elements (HFSE), e.g., Nb, W, and Zr, usually found in rutile from other petrologic assemblages.  except for LP, where rutile contains up to 2347 ppm Mn. Iron concentrations above detection limit were encountered in 95% of the rutile analyses. The average iron content is always high in rutile from LP, but, excluding two anomalous samples from NS, concentrations decrease down to 1398-196 ppm in rutile from NS, JC, MR, CD, and UG2. Aluminum is found in all rutile analyses, with average concentrations of a few hundred ppm in MR (465), JC (519), LP (669), NS (515), and CD (224), and as high as 3665 ppm in UG2 (Table 4). 601 ppm in MR, and 2020 ppm in JC. Rutile has the lowest Nb content at LP (<168 ppm), and NS (<280 ppm) (Tables 3 and 4). Relatively high Nb up to 1780 ppm is found in a few samples of the CD deposit, although 92% of the analyses have less than 615 ppm Nb ( Figure 10B). Wolfram was detected in 76% of rutile analyses, varying from close to detection limit (UG2, MR, LP) to 1808 ppm (JC), and maximum concentrations of 8334 ppm and 6836 ppm in NS and CD, respectively (Table 4).
Zirconium concentrations above detection limit were encountered in 92% of rutile analyses. The highest concentrations are found in rutile from UG2, MR, JC, and LP, decreasing from 3368 ppm to 1081 ppm (Table 4). Rutile from the chromitites of NS and CD is the most depleted in Zr, with average contents of 277 ppm and 165 ppm, respectively, and maximum values not exceeding 511 ppm and 540 ppm, respectively.
The variation patterns of V, Nb, W, and Zr versus Cr in rutile ( Figure 10) are reflections of the different chromitite type (stratiform versus podiform, and unmetamorphic versus metamorphic). Rutile from the podiform chromitite of LP characterizes for the highest Cr content, medium V and Zr concentrations, and maximum depletion in W, compared with the group of stratiform deposits. The rutile from stratiform chromitites (UG2, MR, JC) has the lowest Cr content, and displays the widest ranges of V, Nb, and Zr, whereas W is depleted at the minimum levels similar to LP, except for the JC rutile which shows moderate W enrichment. Rutile from the metamorphic chromitites (NS, CD) distinguishes for the widest variation of Cr content and the lowest concentrations of V, Nb, and Zr, except for five samples from CD which show Nb enrichment similar to JC. A striking feature of rutile from NS and CD is the trend of anomalous W enrichment that largely exceeds maximum concentrations encountered in the unmetamorphic chromitites.

Zr-Based Rutile Geothermometry
The temperature of rutile crystallization was calculated with the Zr-based geothermometer of [12] for the 328 analyzed grains. Although the authors warn about the use of the Zr-in-rutile thermometer in SiO 2 undersatured systems (e.g., chromitite), [13,14] have shown that application to chromitite provides reasonable results in good agreement with geological and petrogenetic conditions. According to these authors, the influence of pressure on calculated temperature is limited to values below 35 • C; therefore, a pressure of 0.5 GPa was selected to treat all the rutile analyses. The results obtained in this contribution are presented in Table 3

Significance of Rutile-Chromite Textural Relations
Rutile is not a typical accessory mineral in igneous ultramafic contests such as chromitite (Cr ore composed of more than 50% chromite). However, as this study demonstrates, chromitite crystallized from mantle-derived magmas in particular geological settings may become mineralized with rutile. The accessory rutile in chromitite has distinctive textural setting that allows us to speculate on the origin of rutile in the different types of host chromitite, and to establish the petrogenetic pathways from primary magmatic deposition up to post-magmatic metamorphism.
In chromitites devoid of significant metamorphic overprint (UG2, MR, JC, LP), primary textural relations ( Figure 5) indicate that rutile has crystallized either in the intercumulus spaces of chromite or at the rims of chromite grains, precipitating directly from the melt that forms the interstitial silicate matrix of chromitite. This type of rutile can be considered as co-magmatic with chromite, although later in the order of crystallization (magmatic rutile). The pegmatoidal nature of the MR cumulate unit suggests that the reef developed by magma-mixing under high partial pressure of volatiles. Residual fluids released during early solidification of the interstitial melt were forced to percolate upwards and laterally, diffusing along crystal boundaries and intercumulus spaces, possibly driven by thermal and pressure gradient. In their migration, they mixed with trapped interstitial liquid, becoming enriched in Zr and Nb, which were incorporated in rutile precipitating in the last stage of crystallization ( Figure 5E,F). This mechanism enhanced by high fluid activity might have delayed interstitial precipitation of rutile well below the solidus temperature of chromite. Another type of rutile occurring as small needles or euhedral to subhedral crystals included in the core of chromite grains (Figure 8) is considered to have unmixed from the chromite in the subsolidus (exsolved rutile). Rutile subsolidus exsolution is common in all types of chromitite, occurring either on post-magmatic cooling, or during syn-metamorphic recrystallization. The composite inclusions of rutile with highly refractory PGM±sulfide±silicate (Figure 9) may suggest entrapment into chromite at high temperature (early magmatic rutile). The main issue, in this case, concerns the physical state of the various components at the time of entrapment (solid, solid+liquid, or liquid droplet). The polygonal shape of most composite inclusions may indicate engulfment into chromite as a solid or solid+liquid particle.
Magmatic rutile is absent in all samples of metamorphic chromitite (NS, CD). All primary textures have been obliterated and rutile characterized by extremely variable morphology and grain size occur indifferently associated with altered and deformed chromite or engulfed in the altered silicate matrix of the chromitite (Figures 6 and 7). The association with exotic metasomatic minerals, and the frequent appearance of altered chromite included in rutile, indicate an origin by remobilization of primary rutile during metamorphism (secondary-metamorphic rutile).

Chromium and Vanadium in Rutile from Magmatic and Metamorphic Chromitites
Although percentage levels of trace elements such as Cr, V, and Nb have been reported [62] for rutiles, the concentrations of Cr and V in accessory rutile from the studied chromitites are distinctly high (2-10×) when compared with rutile occurring in lithotypes of regular composition, such as metapelites, metabasites, gneisses, and granulites. In these rocks, Cr and V do not exceed 5000 ppm and 4000 ppm, respectively [2,11]. The systematic Cr-V-rich nature of rutile from chromitites is a strong indication of the high activity of these metals in the chromite-forming system, then reflects the composition of the host chromite ( Figure 12A,B), varying in relation to the magmatic and metamorphic history of the deposit. Magmatic rutile of LP has crystallized in equilibrium with poorly differentiated high-Cr magma similar to parent melts of podiform chromitite in the upper mantle. The moderate V content, however, would indicate a more evolved composition compared with SSZ boninite [63]. Magmatic differentiation is responsible for the high-V and the decrease in Cr content in rutile of UG2 and MR chromitite layers. In contrast, the low Cr-V content of rutile in the JC ultramafic sill is a reflection of the relatively low Cr# and V-content of chromite that precipitated from poorly differentiated high-Mg magma that prevented V fractionation and "diluted" its concentration in an overwhelming volume of chromite.
Complete overprint of the Cedrolina chromitite by multiple regional metamorphic events and metasomatic reactions with hydrous fluids from albite-granite intrusions has obliterated any remnant of magmatic chromite and possible magmatic rutile. Rutile is characterized by a wide range of Cr content and a constantly low concentration of V. These variation patterns of Cr and V in rutile parallel those of the host chromite, supporting genetic heredity. According to [13], rutile formed at the expense of chromite by remobilization of Ti, Cr, and V during metamorphism, thus shifting chromite composition towards a "ferrian-chromite" phase characterized by extremely high Fe 2 O 3 , but strongly depleted in Ti and V.
The large rutile crystals observed in the chromitite fragments of the Nuasahi breccia are characterized by a low-Cr, low-V composition that mimic Cr-V relations of the host chromite. Like CD rutile, they are interpreted as secondary crystals formed by re-distribution of Ti, Cr, and V during metamorphism. The scarcity of low-temperature hydrous silicates in the chromitite is suggestive of minimal fluid activity, indicating that mobilization of Ti, Cr, and V from chromite to form rutile started at temperatures higher than hydrothermal during gabbro intrusion, and continued under the greenschist-amphibolite grade metamorphism [23]. Intrusion of quartz-diorite veins into the chromitite breccia triggered further metasomatic reactions and contamination of the rutile forming system by incompatible elements with felsic signature (see below).

Significance of HFSE (Niobium-Zirconium-Wolfram) in Chromitite-Hosted Rutile
The data in Table 4 show that the total amount of HFSE is very low (av. 963 ppm) in rutile from the podiform chromitite of LP, and is mainly accounted for by Zr. The HFSE content is slightly higher in rutile from the stratiform chromitites of UG2 (av. 1255 ppm) and MR (av. 1216 ppm), determined by relatively high concentrations of Zr+Nb. Rutile from the JC ultramafic has the highest concentration of HFSE in the group of magmatic chromitites, determined by equal amounts of Zr, Nb, and W summing up to av. 2848 ppm. The highest HFSE concentrations are achieved in rutile from metamorphic chromitite of NS (up to 9125 ppm) and CD (up to 9158 ppm), being essentially due to the anomalous concentration of W, while both Zr and Nb are depleted. Broad positive correlation does exist between Nb and Zr in magmatic rutile from the stratiform chromitites of UG2, MR, and JC ( Figure 13A), while correlation between W and Zr seems to be present only in the samples from JC, although very weak ( Figure 13B). No obvious correlation is visible between Nb-Zr, W-Zr, and W-Nb in rutile from metamorphic deposits of NS and CD, which characterize for a trend of exceptional increase in W content (Figure 13 B,C).
Reciprocal substitutions among HFSE in rutile from magmatic chromitite are presented as compositional fields in the Zr-Nb-W ternary diagram, and compared with data point distribution of the metamorphic chromitites and estimated average composition of depleted mantle, tholeiitic basalt, and continental crust taken from a wide literature ( Figure 14).
The Zr-specialized composition observed in magmatic rutile from podiform chromitite (LP) is interpreted as a reflection of the general paucity of lithophile metals (Nb, W) in magmas parent to chromite deposits in the sub-oceanic upper mantle. On the other hand, it is worthy of remark that the occurrence of rutile in a high-Cr chromitite enriched in Ti-V-Zr, and the presence of Pt minerals at LP [41,42] is irreconcilable with composition and mineral assemblage of high-Cr deposits commonly associated with ophiolites, which are characterized by low Ti-V-Zr, and carry Ru-Os-Ir PGM [63]. Thus, the peculiar mineral chemistry of chromite and rutile at LP requires that the parent melt had relatively high concentration of incompatible elements, implying an origin by partial melting of undepleted mantle source, not consistent with the strongly residual mantle related with SSZ [39]. This raises once again the question of the real geodynamic setting of the Loma Caribe mantle tectonite.
Magmatic rutile from the stratiform chromitites of Bushveld (UG2, MR) defines a trend of Nb-for-Zr substitution at relatively low W contents (Figure 14) that is a result of magma mixing and differentiation. As mentioned above, these processes were responsible for the collection and redistribution of Zr and Nb in residual melts yielding rutile precipitation in the intercumulus spaces of chromite or at the rims of chromite grains.
The wide range of Zr-Nb exchange in rutile of MR was promoted by the high fluid activity in the late residual melts originated by fractional crystallization of the pegmatoid reef. The rutile in the stratiform chromitite of JC defines trends of increasing Nb and Zr contents similar to the Bushveld stratiform deposits, ( Figure 13A); however, it distinguishes for a sensible and consistent enrichment in W that causes shifting of rutile compositions towards the Zr-poor field typical of rutile from metamorphic chromitites ( Figure 14). Most rutile from the chromitites of NS and CD display systematic low contents of Zr and Nb, possibly suggesting removal of HFSE from rutile crystallizing under metamorphic conditions. Data point distribution indicates that Nb was almost totally removed from the rutile of NS, compared with CD, where Nb appears to decrease gradually. In both cases, the trend of W substitution for Zr and Nb is striking and marks the role of W as an indicator of the different origin of rutile in magmatic and metamorphic chromitites.  Geochemical affinity and estimated W abundance in depleted mantle, tholeiitic basalt, and continental crust ( Figure 14) predict that W is generally depleted in mafic magmas derived from mantle partial melting, but has strong tendency to concentrate in melts and fluids of felsic and granitic composition. Therefore, the presence of substantial concentrations of W in ultramafic igneous rocks such as the chromitites must be related with "crustal contamination" processes. The high-W ranges of up to more than 8000 ppm in rutile of CD and NS ( Figure 12B,C) argue for the addition of W to the chromitite in the post-magmatic stage, during metamorphism. At Cedrolina, W was probably part of the incompatible elements assemblage (REE, U, Th, Zr) that was introduced into the chromitite via infiltrating albite-granite hydrothermal fluids [13]. At Nuasahi, the source of W is uncertain. Considering that the analyzed sample is a chromitite fragment of the breccia [23], W could have been brought into the chromitite by metasomatic reaction with the gabbroic matrix, during intrusion. The reaction was possibly enhanced during the greenschist-amphibolite metamorphic event, and late infiltration of quartz-diorite melt that is likely to have caused the main W contamination in the chromitite fragments.
Crustal contamination provides a reasonable explanation for the moderate W enrichment (up to 1808 ppm) detected in rutile associated with the Jacurici chromitite layer. However, according to currently accepted reconstruction [37], contamination with crustcompatible lithophile elements was not a post-magmatic event at Jacurici, but occurred in the orthomagmatic stage by assimilation of partially melted metasediments bordering the magma chamber. Compositional changes induced in the mafic magma contributed to sustain Cr saturation long enough for the formation of a thick chromitite layer. There is no evidence of late infiltration of felsic melts in the samples of UG2, MR, and LP chromitite, neither assimilation of country rocks appears to have been effective. The small amounts of W encountered in rutile of these magmatic chromitites may be due to the enrichment of incompatible elements into residual melts during magmatic differentiation, or alternatively might have derived from limited contamination in the Bushveld parental magmas during its ascent into the crust [27].

Zr-in-Rutile Thermometry of Chromitite
Results of Zr-in-rutile thermometry indicate that rutile from metamorphic and magmatic chromitites crystallized in diverse thermal ranges ( Figure 11). Oscillatory zoning in the studied rutile was not observed, suggesting that the degree of isomorphic substitution of Zr and other trace elements in rutile associated with chromitite is not related with the growth of the crystals and diffusion rate as reported by [62], but is controlled by the temperature of crystallization.
Correlation of calculated temperatures with the W content of rutile highlights the existence of a thermal gap of about 140 • C between the two types of chromitite ( Figure 15A). Rutile from both metamorphic chromitites of CD and NS presented average crystallization temperatures of 546 • C and 602 • C, respectively. These values are in good agreement with the thermal peak of 550-600 • C proposed for the amphibolite-facies regional metamorphism of the Pilar de Goiás Greenstone Belt [13], and the greenschist-amphibolite metamorphic overprint of the Nuasahi complex [48].
The distribution of data points in the correlation diagram between W and T • C poses a maximum thermal limit of 600-650 • C for the metasomatic enrichment of W in rutile from metamorphic chromitites. Rutile associated with magmatic chromitites (UG2, MR, LP) is generally W-depleted compared with the metamorphic ones, which apparently exclude contamination from felsic melts of an external source. The Zr-in-rutile thermometry yielded temperatures of magmatic crystallization in the range of 700-869 • C, with some data points extending down to 600 • C and less. The authors of [14] have shown that rutile from ultramafic cumulates of the Bushveld Critical Zone split into two distinct fields, based on a diagram involving variation of the Nb content as a function of Zr-in-rutile temperature. These authors interpreted the Nb-rich rutile (Nb > 4000 ppm) as a product of direct crystallization from the mafic magma at temperatures of 800-1000 • C, whereas the Nb-poor rutiles (Nb < 1000 ppm) exsolved from chromite in a subsolidus thermal range of 800-500 • C. According to this model, the great bulk of rutile from chromitite deposits examined in this study would plot in the field of rutile exsolving from chromite at subsolidus temperatures below 800 • C ( Figure 15B).
This conclusion is probably applicable to exsolved rutile scattered in the chromite of magmatic and metamorphic deposits, but disagrees with the origin of magmatic rutile proposed in this work. In magmatic chromitite, rutile exsolution takes place upon cooling in a thermal interval from about 700 • C down to less than 600 • C. This type of rutile, significantly depleted in Nb, is encountered in both magmatic and metamorphic chromitites (e.g., UG2, MR, LP, CD, NS) ( Figure 15B). Rutile needles (<5 µm) crystallographically oriented according to the cubic symmetry of the host spinel, are found included in the chromite of metamorphic chromitite (CD, NS). The texture strongly supports exsolution during chromite metamorphism at temperatures well below 600 • C, down to 400 • C.
On the other hand, the occurrence of large rutile grains in the altered matrix of metamorphic chromites is consistent with direct precipitation of rutile from hydrous solutions, thus shifting the temperature of rutile crystallization below the amphibolite facies peak (600-650 • C), but consistent with greenschist facies retrograde metamorphism, as suggested by [13]. The exsolution model is seemingly in contrast with the intercumulus textures of rutile described in the magmatic chromitites which indicates rutile precipitation from the interstitial melt. On the other hand, the Nb-poor (Nb < 2000 ppb) composition and low crystallization temperature (800-700 • C) would be consistent with rutile exsolution from chromite. It is unlikely that this discrepancy resides in the erroneous calculation of the rutile temperature. Conversely, we believe that these types of rutile are late-magmatic in origin, implying that the thermal interval for magmatic precipitation of rutile that should be extended to temperatures lower than previously supposed. It is worthy of remark that magmatic rutile interstitial to chromite in MR appears to have precipitated at temperature quite lower than those proposed by [14] for pure magmatic rutile. We suggest that low crystallization temperatures are to be expected when rutile precipitates from post-chromite interstitial melts enriched in hydrous fluids at relatively high oxygen fugacity, as those reported from MR chromitite in this study.

Origin of Rutile-PGM Inclusions in Chromite
The chromitites of UG2, MR, NS, and LP contain a particular type of polygonal inclusions composed of rutile, PGM, and magmatic sulfides. The rutile-PGM association has been previously reported from a number of chromite deposits associated with layered intrusions [17,23,56,60,64,65]. The examples reported in this work regard the chromitites of UG-2, MR, NS, and LP, the latter representing the first documented quotation of rutile PGM inclusions in podiform chromitites of ophiolitic upper mantle.
The presence of PGM inclusions in chromitite is not surprising in consideration of the common association of PGE with chromite. However, the systematic association of PGM with rutile observed in our samples establishes a genetic linkage, and raises again the important question about the origin of rutile: exsolution-related or magmatic precipitation. Previous works have considered the composite inclusions as material crystallized at high temperature in the melt and trapped as solid particles into precipitating chromite. Unfortunately, we were not able to precisely determine the composition of rutile for temperature calculation due to the extremely small size of the grains. Yet, the common euhedral morphology of both rutile and PGM would be consistent with crystallization in a fluid milieu at high temperature. Of particular interest is the occurrence of rutile PGM inclusions observed in the metamorphic chromitite of NS [23] and this work. The association, previously observed in highly metamorphosed chromitites of Brazil [17,56], was ascribed to low temperature remobilization of Ti and PGE during metamorphism and alteration of primary chromite and PGM. In contrast, based on textural relation of the rutile-PGM inclusions in the NS chromitite, [23] concluded that they are relic magmatic minerals endured the metamorphic event. In summary, the rutile PGM inclusions in the magmatic chromitites of UG2, MR, and LP provide further evidence for early deposition of rutile in the pre-chromite stage, an origin that can be extended even to single rutile inclusions allegedly interpreted as chromite exsolution products [14,15].

Comparison of Accessory Rutile in Chromitite with Rutile from Other Petrologic Associations
Accessory rutile associated with chromitite represents a further category of igneous rutile characterized by distinctive trace-element fingerprints dominated by chromitecompatible metals (Cr, V) and HFSE (Nb, W, Zr). The Cr-Nb correlation has been used in the literature to trace the provenance of rutile in sediments, in an attempt to discriminate among rutile originally associated with mafic rocks, metapelites, gneisses, or felsic-and mafic-granulites [2,7,11]. Although this approach may lead to contrasting results when dealing with high-grade metamorphic assemblages, we found it useful to illustrate the unusual Cr/Nb composition of accessory rutile from chromitites. Using the fields drawn by [7], we have compiled a modified version of the diagram Cr/Nb, extending the Cr scale upwards in order to include composition of rutiles from chromitites ( Figure 16). The bulk of our analyses cluster in the Cr-rich, Nb-poor sector of the diagram, with only limited overlap with Cr-poor rutile from mafic and metapelitic rocks characterized by high Nb contents [7,11]. The Cr and V contents increase according to the composition of the host chromite, establishing a genetic linkage with either the parental magma of the deposit, or compositional reworking of chromite during metamorphism.  visit the Jacurici mining sites (August 1988, September 1989). The authors G.G. and F.Z. acknowledge the 18va Conferencia Geologica del Caribe, Republica Dominicana, for having organized the field trip to the chromitite occurrences of Loma Peguera (March 2008). We are grateful to the academic editor and two anonymous referees for their fruitful comments. The editorial staff of Minerals is thanked for the help in editing the manuscript.

Conflicts of Interest:
The authors declare no conflict of interest.