Tourmaline as a Recorder of Ore-Forming Processes in the Xuebaoding W-Sn-Be Deposit, Sichuan Province, China: Evidence from the Chemical Composition of Tourmaline

The Xuebaoding W-Sn-Be deposit located in the Songpan-Ganze Orogenic Belt (Sichuan Province, China) is a hydrothermal deposit with less developed pegmatite stage. The deposit is famous for the coarse-grained crystals of beryl, scheelite, cassiterite, apatite, fluorite, muscovite, and others. The orebody is spatially associated with the Pankou and Pukouling granites hosted in Triassic marbles and schists. The highly fractionated granites are peraluminous, Li-Rb-Cs-rich, and related to W-Sn-Be mineralization. The mineralization can chiefly be classified based on the wallrock and mineral assemblages as muscovite and beryl in granite (Zone I), then beryl, cassiterite and muscovite at the transition from granite to triassic strata (Zone II), and the main mineralized veins composed of an assemblage of beryl, cassiterite, scheelite, fluorite, and apatite hosted in metasedimentary rock units of marble and schist (Zone III). Due to the stability of tourmaline over a wide range of temperature and pressure conditions, its compositional variability can reflect the evolution of the ore-forming fluids. Tourmaline is an important gangue mineral in the Xuebaoding deposit and occurs in the late-magmatic to early-hydrothermal stage, and can thus be used as a proxy for the fluid evolution. Three types of tourmalines can be distinguished: tourmaline disseminations within the granite (type I), tourmaline clusters at the margin of the granite (type II), and tourmalines occurring in the mineralized veins (type III). Based on their chemical composition, both type I and II tourmalines belong to the alkali group and to the dravite-schorl solid solution. Type III tourmaline which is higher in X-site vacancy corresponds to foitite and schorl. It is proposed that the weakly zoned type I tourmalines result from an immiscible boron-rich aqueous fluid in the latest stage of granite crystallization, that the type II tourmalines showing skeletal texture directly formed from the undercooled melts, and that type III tourmalines occurring in the mineralized veins formed directly from the magmatic hydrothermal fluids. Both type I and type II tourmalines show similar compositional variations reflecting the highly fractionated Pankou and Pukouling granites. The higher Ca, Mg, and Fe contents of type III tourmaline are buffered by the composition of the metasedimentary host rocks. The decreasing Na content (<0.8 atoms per formula unit (apfu)) and increasing Fe3+/Fe2+ ratios of all tourmaline samples suggest that they precipitated from oxidized, low-salinity fluids. The decreasing trend of Al content from type I (5.60–6.36 apfu) and type II (6.01–6.43 apfu) to type III (5.58–5.87 apfu) tourmalines, and associated decrease in Na, may be caused by the crystallization of albite and muscovite. The combined petrographic, mineralogical, and chemical characteristics of the three types of tourmalines thus reflect the late-magmatic to early-hydrothermal evolution of the ore-forming fluids, and could be used as a geochemical fingerprint for prospecting W-Sn-Be mineralization in the Xuebaoding district. Minerals 2020, 10, 438; doi:10.3390/min10050438 www.mdpi.com/journal/minerals Minerals 2020, 10, 438 2 of 21

The Xuebaoding deposit is hosted in Triassic strata of marble and schist surrounding the two granite intrusions. The mineralization occurs primarily in hydrothermal veins cutting marbles, and breccia bodies and pods in schist. The Xuebaoding deposit is particular for two reasons: (a) in addition to being sparsely disseminated, the crystals of beryl, cassiterite, and scheelite occur in clusters, and in geodes in mineralized veins; (b) the coarse-grained crystals are often euhedral, transparent, large, and of gem-quality.
The Pankou and Pukouling granites are mainly composed of quartz (35%-40%), albite (30%-35%), muscovite (30%-35%), with minor K-feldspar (0%-5%), while mafic minerals are absent. Accessory minerals are zircon, apatite, pyrite, rutile and tourmaline. The Pankou and Pukouling granites are characterized by a central facies and a border facies with a progressively decreasing mineral grain size towards the border of the granite (Figure 2). The border facies is spatially associated with ore veins, with a high content of ore elements (such as W, Sn, Be) according to whole rock analysis, significantly higher than those of the Clarke [5][6][7][48][49][50]. This suggests that the Pankou and Pukouling granites are either the source of the mineralizing fluids or contributed indirectly to the formation of the deposit. In addition, the border facies of the Pankou and Pukouling granites has higher concentrations of Al 2 O 3 (14.63-21.3 wt.%), P 2 O 5 (0.07-0.11 wt.%), and B (65-114 ppm) compared to the central facies [6,7]. In term of trace elements, the Pankou and Pukouling granites have low contents of Cr, Ni, Sr, Ba, and Zr, and high contents of Li, Rb, and Cs, typical of highly fractionated granites as described by Wu et al. [51]. Further, with the increasing degree of differentiation, the content of albite in the granites is increasing [5]. 4 granites are characterized by a central facies and a border facies with a progressively decreasing mineral grain size towards the border of the granite (Figure 2). The border facies is spatially associated with ore veins, with a high content of ore elements (such as W, Sn, Be) according to whole rock analysis, significantly higher than those of the Clarke [5][6][7][48][49][50]. This suggests that the Pankou and Pukouling granites are either the source of the mineralizing fluids or contributed indirectly to the formation of the deposit. In addition, the border facies of the Pankou and Pukouling granites has higher concentrations of Al2O3 (14.63-21.3 wt.%), P2O5 (0.07-0.11 wt.%), and B (65-114 ppm) compared to the central facies [6,7]. In term of trace elements, the Pankou and Pukouling granites have low contents of Cr, Ni, Sr, Ba, and Zr, and high contents of Li, Rb, and Cs, typical of highly fractionated granites as described by Wu et al. [51]. Further, with the increasing degree of differentiation, the content of albite in the granites is increasing [5]. The mineralized veins occur in the extensive joints of the Triassic strata and granites. These consist of a quartz-dominated center, and a margin composed of coarse-grained beryl (0.5-15 cm), cassiterite (1-30 cm), scheelite (1-30 cm), K-feldspar (1-25 cm), albite (1-25 cm), muscovite (1-3 cm), fluorite (1-10 cm), and apatite (0.5-3 cm). This suggests that the monomineralic quartz veins formed after the coarse-grained crystals of the wall. A muscovite fringe separates the mineralized vein from the host rock. Beryl, cassiterite, scheelite, fluorite, apatite, K-feldspar, and albite occur as single crystals or aggregates overgrown on muscovite. Greisen and other wall rock alterations are not aggregates overgrown on muscovite. Greisen and other wall rock alterations are not developed in this deposit, and only a very weak muscovite alteration (up to only 0.5 cm thick) and skarn-type alteration were observed in the Xuebaoding deposit.
The crystallization sequence of feldspars in the veins was described by Liu et al. [5]. K-feldspars form at relatively high temperature (500-800 • C), and with decreasing temperature, K-feldspar is replaced by albite on a large scale. Ore minerals such as beryl, scheelite, and cassiterite mostly coexist with albite. In contrast, only beryl, and only in minor amount, is found to coexist with K-feldspar.
Mineralized veins in the Xuebaoding deposit can chiefly be categorized into three zones, based on their mineral assemblages and their host rocks ( Figure 3). Zone I is hosted in the granite and is dominated by muscovite, tourmaline, and beryl assemblage. Zone II is mainly composed of beryl, cassiterite, tourmaline, and muscovite and is located at the transition from the granite to metamorphic rock. Zone III represents the main host for the mineralization, with beryl, cassiterite, scheelite, fluorite, calcite, and tourmaline. No crosscutting relationship was observed between the three zones of veins. The calculated H and O isotopic compositions of beryl, scheelite and cassiterite, indicate that the ore-forming fluids are mainly composed of magmatic water with minor meteoric water and CO 2 derived from decarbonation of marble [7]. Thus, it is likely that the veins formed during a single hydrothermal stage.
5 developed in this deposit, and only a very weak muscovite alteration (up to only 0.5 cm thick) and skarn-type alteration were observed in the Xuebaoding deposit.
The crystallization sequence of feldspars in the veins was described by Liu et al. [5]. K-feldspars form at relatively high temperature (500-800 °C), and with decreasing temperature, K-feldspar is replaced by albite on a large scale. Ore minerals such as beryl, scheelite, and cassiterite mostly coexist with albite. In contrast, only beryl, and only in minor amount, is found to coexist with K-feldspar.
Mineralized veins in the Xuebaoding deposit can chiefly be categorized into three zones, based on their mineral assemblages and their host rocks ( Figure 3). Zone I is hosted in the granite and is dominated by muscovite, tourmaline, and beryl assemblage. Zone II is mainly composed of beryl, cassiterite, tourmaline, and muscovite and is located at the transition from the granite to metamorphic rock. Zone III represents the main host for the mineralization, with beryl, cassiterite, scheelite, fluorite, calcite, and tourmaline. No crosscutting relationship was observed between the three zones of veins. The calculated H and O isotopic compositions of beryl, scheelite and cassiterite, indicate that the ore-forming fluids are mainly composed of magmatic water with minor meteoric water and CO2 derived from decarbonation of marble [7]. Thus, it is likely that the veins formed during a single hydrothermal stage. . Schematic cross-section of a typical mineralized vein showing the distribution of beryl, cassiterite, scheelite, muscovite, tourmaline, albite and K-feldspar and host rock transition from granite to Triassic metamorphic strata and divided into three characteristic zones. After reference [6].

Sampling and Petrographic Description
Samples of tourmaline granite, tourmaline clusters and tourmaline inclusions are collected from different locations of the Xuebaoding deposit. Representative samples are shown in Figure 4. Because of the general simple texture of the thin section micrographs, we only provide a qualitative description, followed by quantitative observation of samples.
Based on their occurrences, tourmalines from the Xuebaoding deposit can be divided into three types. Type I tourmaline was collected in the border facies of the granites. It is fine-grained with needles of 200-400 μm in diameter and length of 1-2 mm (Figure 4a). Type I tourmaline crystals are disseminated in the granitic rock, always intergrown with albite and phengite, of euhedral shape, and surrounded by secondary hydrothermal minerals. . Schematic cross-section of a typical mineralized vein showing the distribution of beryl, cassiterite, scheelite, muscovite, tourmaline, albite and K-feldspar and host rock transition from granite to Triassic metamorphic strata and divided into three characteristic zones. After reference [6].

Sampling and Petrographic Description
Samples of tourmaline granite, tourmaline clusters and tourmaline inclusions are collected from different locations of the Xuebaoding deposit. Representative samples are shown in Figure 4. Because of the general simple texture of the thin section micrographs, we only provide a qualitative description, followed by quantitative observation of samples.
Based on their occurrences, tourmalines from the Xuebaoding deposit can be divided into three types. Type I tourmaline was collected in the border facies of the granites. It is fine-grained with needles of 200-400 µm in diameter and length of 1-2 mm (Figure 4a). Type I tourmaline crystals are disseminated in the granitic rock, always intergrown with albite and phengite, of euhedral shape, and surrounded by secondary hydrothermal minerals.
Type II tourmalines formed along the margin of the granite as radial clusters. Muscovite layers separate the mineralized vein from the tourmaline clusters ( Figure 4b). Locally, tourmaline clusters are overgrown by beryl, albite, scheelite, cassiterite, and quartz (Figure 4b,c). Type II tourmaline crystals are larger than type I with a length up to 0.5-1 cm. Type II tourmalines formed along the margin of the granite as radial clusters. Muscovite layers separate the mineralized vein from the tourmaline clusters ( Figure 4b). Locally, tourmaline clusters are overgrown by beryl, albite, scheelite, cassiterite, and quartz (Figure 4b,c). Type II tourmaline crystals are larger than type I with a length up to 0.5-1 cm.
Type III tourmalines are found as needles in mineralized veins ( Figure 4d). The very coarsegrained (2-5 cm) tourmaline crystals always formed as inclusions in other coarse-grained minerals, such as albite, quartz and beryl. Except the size, the difference between type III and type II tourmalines is that type III tourmalines always overgrow on the muscovite layers. Type III tourmaline crystals predominantly show parallel growth (Figure 4d), compared with the radial type II tourmaline.

Analytical Methods
Back scattered electron (BSE) images and chemical compositions of tourmalines were acquired at the Laboratory of Mineralization and Dynamics, Chang'an University (China), using a JXA-8100 electron microprobe analyzer (EMPA), with an acceleration voltage of 15 kV, a beam current of 10 nA, and a spot size less than 10 μm. EMPA standards include the following minerals: andradite for Si and Ca, rutile for Ti, corundum for Al, hematite for Fe, eskolaite for Cr, rhodonite for Mn, bunsenite for Ni, periclase for Mg, albite for Na, and K-feldspar for K.
The general formula of tourmaline is XY3Z6(T6O18)(BO3)3VW with X = Na + , K + , Ca 2+ , or vacancy; Y = Fe 2+ , Mg 2+ , Mn 2+ , Al 3+ , Fe 3+ , Cr 3+ ,V 3+ , Ti 4+ , and Li + ; Z = Al 3+ , Mg 2+ , Fe 3+ , Cr 3+ , V 3+ ; T = Si, Al; B = B; V = OH, O and W = OH, F, O [18]. Following the normalization procedure of [21], structural formulae of tourmaline were calculated on the basis of a total of 15 cations in the octahedral and tetrahedral (Y + Z + T) sites, all iron was assumed to be Fe 2+ . The proportion of X-site vacancies (❑ ) was calculated as [1 − (Na + Ca + K)]. F was not analyzed. Results of different crystals are given in Tables 1-3.  Type III tourmalines are found as needles in mineralized veins (Figure 4d). The very coarse-grained (2-5 cm) tourmaline crystals always formed as inclusions in other coarse-grained minerals, such as albite, quartz and beryl. Except the size, the difference between type III and type II tourmalines is that type III tourmalines always overgrow on the muscovite layers. Type III tourmaline crystals predominantly show parallel growth (Figure 4d), compared with the radial type II tourmaline.

Analytical Methods
Back scattered electron (BSE) images and chemical compositions of tourmalines were acquired at the Laboratory of Mineralization and Dynamics, Chang'an University (China), using a JXA-8100 electron microprobe analyzer (EMPA), with an acceleration voltage of 15 kV, a beam current of 10 nA, and a spot size less than 10 µm. EMPA standards include the following minerals: andradite for Si and Ca, rutile for Ti, corundum for Al, hematite for Fe, eskolaite for Cr, rhodonite for Mn, bunsenite for Ni, periclase for Mg, albite for Na, and K-feldspar for K.
Type III tourmalines are found as needles in mineralized veins ( Figure 4d). The very coarsegrained (2-5 cm) tourmaline crystals always formed as inclusions in other coarse-grained minerals, such as albite, quartz and beryl. Except the size, the difference between type III and type II tourmalines is that type III tourmalines always overgrow on the muscovite layers. Type III tourmaline crystals predominantly show parallel growth (Figure 4d), compared with the radial type II tourmaline.

Analytical Methods
Back scattered electron (BSE) images and chemical compositions of tourmalines were acquired at the Laboratory of Mineralization and Dynamics, Chang'an University (China), using a JXA-8100 electron microprobe analyzer (EMPA), with an acceleration voltage of 15 kV, a beam current of 10 nA, and a spot size less than 10 μm. EMPA standards include the following minerals: andradite for Si and Ca, rutile for Ti, corundum for Al, hematite for Fe, eskolaite for Cr, rhodonite for Mn, bunsenite for Ni, periclase for Mg, albite for Na, and K-feldspar for K.

Textural Features Of Tourmaline
Type I tourmalines occurs as euhedral grains varying from 200 to 300 µm in diameter on BSE images (Figure 5a,b). They are poorly zoned with dark cores surrounded by bright rims. The abundances of feldspar in the interstitial areas between type I tourmaline grains in the tourmaline-bearing granite show some differences between the Pankou and Pukouling granites. The K-feldspar content is lower, and albite is more abundant (Figure 5a-c), compared to the border facies of the granite (Figure 2). Muscovite is transformed into phengite on a large scale. Biotite does not occur in tourmaline-bearing granite. Further, scheelite forms grains of about 10 µm in size and generally occurring in fractures of phengite (Figure 5a).

Textural Features Of Tourmaline
Type I tourmalines occurs as euhedral grains varying from 200 to 300 μm in diameter on BSE images (Figure 5a,b). They are poorly zoned with dark cores surrounded by bright rims. The abundances of feldspar in the interstitial areas between type I tourmaline grains in the tourmalinebearing granite show some differences between the Pankou and Pukouling granites. The K-feldspar content is lower, and albite is more abundant (Figure 5a-c), compared to the border facies of the granite (Figure 2). Muscovite is transformed into phengite on a large scale. Biotite does not occur in tourmaline-bearing granite. Further, scheelite forms grains of about 10 μm in size and generally occurring in fractures of phengite (Figure 5a). Type II tourmalines occur as radial clusters along the margin of the granite. No distinct compositional domains or zoning are observed within single tourmaline crystals on the BSE images (Figure 5d). In particular, the larger tourmalines show a skeletal texture which are composed of numerous small (50-150 μm) disconnected fine-grained tourmaline crystals (Figure 5e,f). Type II Type II tourmalines occur as radial clusters along the margin of the granite. No distinct compositional domains or zoning are observed within single tourmaline crystals on the BSE images (Figure 5d). In particular, the larger tourmalines show a skeletal texture which are composed of numerous small (50-150 µm) disconnected fine-grained tourmaline crystals (Figure 5e,f). Type II tourmaline crystals show characteristics of hydrothermal alteration on the BSE images ( Figure 5). Tourmaline crystals are 1-2 cm in size and always coexist with albite, K-feldspar, and muscovite ( Figure 5e). K-feldspar is more abundant at the margin of the granite, with large scale albitization. (Figure 5d,e). Muscovite here has a relatively smaller size compared with the phengite coexisting with type I tourmaline (Figure 5e).
The very coarse-grain type III tourmalines (2-5 cm) occur in the mineralized veins. They are always found as inclusions within albite, quartz, and beryl in the mineralized veins. K-feldspar is not found to coexist with this type of tourmaline (Figure 4d). Sometimes, type III tourmaline crystals are found to grow above the muscovite layer, which can distinguish the two types (Figure 4b).  Table 1). The content of Mg, Fe, Na, Al and X-site vacancy seem to vary greatly. Color variations respond to increasing Ca, Ti, Fe, Mg, and X-site vacancies and decreasing Al towards brighter rims. In the Al-Fe-Mg ternary diagram after reference [20], tourmaline compositions plot in the field of Li-poor granitoid and their associated pegmatites, aplites, metapelites, and psammites ( Figure 6a). In the Fe-Mg-Ca ternary diagram, type I tourmaline compositions plot in the field of Ca-poor metapelites, psammites, and calc-silicate rocks, and the near Li-poor granitoid and their associated pegmatites and aplites (Figure 6b). All tourmalines belong to the alkali group ( Figure 6c Table 2). The Al values of type II tourmalines are relatively higher than that of type I, and the Na contents are relatively lower. In the Al-Fe-Mg and the Fe-Mg-Ca ternary diagrams (Figure 6a,b), type II and type I tourmaline compositions closely overlap with each other. Type II tourmaline belongs to the alkali group (Figure 6c), and plot at the limit of the schorl/Oxy-schorl and dravite/Oxy-dravite fields (Figure 6d). With the content of Al total (6.01-6.43 apfu) nearly to 6 apfu ( Figure 7a) and the (Fe + Mg) values (2.10-2.32 apfu) nearly to 3 apfu, type II could also be classified as dravite-schorl which have a higher content of X-site vacancy and Al than common dravite and schorl.  Table 3). The contents of Na and Al are obviously lower than type I and II tourmalines. The Fe, Mg, Ca and X-site vacancy values are higher than that in type I and II tourmalines. In the Fe-Mg-Ca ternary diagram (Figure 6a), type III tourmaline compositions plot in the field of Fe 3+ -rich quartz-tourmaline rocks. Tourmaline compositions are plotted in the Li-rich granitoid, pegmatites, aplites and Li-poor granitoids, pegmatites, aplites in the Fe-Mg-Ca ternary diagrams (Figure 6b). Due to the variation of X-site vacancy, Type III tourmaline belongs to the alkali group and vacancy group (Figure 6c ineralized veins (Figure 4d). The very coarseas inclusions in other coarse-grained minerals, ference between type III and type II tourmalines is muscovite layers. Type III tourmaline crystals mpared with the radial type II tourmaline.

Chemical Composition of Tourmaline
ical compositions of tourmalines were acquired Chang'an University (China), using a JXA-8100 leration voltage of 15 kV, a beam current of 10 ds include the following minerals: andradite for Fe, eskolaite for Cr, rhodonite for Mn, bunsenite r for K.
18)(BO3)3VW with X = Na + , K + , Ca 2+ , or vacancy; Z = Al 3+ , Mg 2+ , Fe 3+ , Cr 3+ , V 3+ ; T = Si, Al; B = B; V = ization procedure of [21], structural formulae of 15 cations in the octahedral and tetrahedral (Y + portion of X-site vacancies ❑ was calculated as ifferent crystals are given in Tables 1-3. -Fe-O root name fields (Figure 6d). With the high content of (Fe + Mg) values more than 3 apfu and the Al total (5.58-5.87 apfu) nearly to 6 apfu, type III could be classified as schorl (NaFe 2+ 3 Al 6 Si 6 O 18 (BO 3 ) 3 (OH) 3 OH) which have a higher content of X-site vacancy and foitite ( Type II tourmalines formed along the margin of the granite as radial clusters. Muscovite layers separate the mineralized vein from the tourmaline clusters (Figure 4b). Locally, tourmaline clusters are overgrown by beryl, albite, scheelite, cassiterite, and quartz (Figure 4b,c). Type II tourmaline crystals are larger than type I with a length up to 0.5-1 cm.
Type III tourmalines are found as needles in mineralized veins (Figure 4d). The very coarse grained (2-5 cm) tourmaline crystals always formed as inclusions in other coarse-grained minerals such as albite, quartz and beryl. Except the size, the difference between type III and type II tourmalines is that type III tourmalines always overgrow on the muscovite layers. Type III tourmaline crystals predominantly show parallel growth (Figure 4d), compared with the radial type II tourmaline.

Origin of Tourmaline
In previous studies, three hypotheses have been proposed for the origin of tourmaline in granitic rocks: (1) Crystallization can occur from a boron-rich granitic melt [26,33]; (2) Crystallization proceeds from an immiscible boron-rich hydrous melt or fluid that segregated in the late-magmatic stage [13,29,35,44]; lastly, (3) post-magmatic hydrothermal alteration of the granite by an externally supplied boron-rich fluid [29,34]. In Pankou and Pukouling granites, the initially formed K-feldspar is largely replaced by albite during the evolution of the highly fractionated magma, hence, only few K-feldspars are preserved [5]. Type I tourmalines occur as disseminated euhedral crystals in the granites and coexist with albite and phengite (Figure 5a-c), and no K-feldspar is found. In the highly fractionated granite, the abundance of K-feldspar decrease with the increasing degree of differentiation. Phengite is generally regarded as the alteration product of muscovite. In addition, fine-grained scheelite is found coexisting with type I tourmaline (Figure 5a).
Tourmaline formed from hydrothermal fluids shows fine-scale, oscillatory-type zoning [33,35,42]. Euhedral type I tourmaline crystals are poorly zoned with dark cores surrounded by bright rims. Evidence above implies that these types may not crystallize directly from a B-rich melt even in granites.
Chemical compositions of the Pankou and Pukouling granites as well as the intruded marble were investigated by previous studies [5][6][7].

Origin of Tourmaline
In previous studies, three hypotheses have been proposed for the origin of tourmaline in granitic rocks: (1) Crystallization can occur from a boron-rich granitic melt [26,33]; (2) Crystallization proceeds from an immiscible boron-rich hydrous melt or fluid that segregated in the late-magmatic stage [13,29,35,44]; lastly, (3) post-magmatic hydrothermal alteration of the granite by an externally supplied boron-rich fluid [29,34]. In Pankou and Pukouling granites, the initially formed K-feldspar is largely replaced by albite during the evolution of the highly fractionated magma, hence, only few K-feldspars are preserved [5]. Type I tourmalines occur as disseminated euhedral crystals in the granites and coexist with albite and phengite (Figure 5a-c), and no K-feldspar is found. In the highly fractionated granite, the abundance of K-feldspar decrease with the increasing degree of differentiation. Phengite is generally regarded as the alteration product of muscovite. In addition, fine-grained scheelite is found coexisting with type I tourmaline (Figure 5a).
Tourmaline formed from hydrothermal fluids shows fine-scale, oscillatory-type zoning [33,35,42]. Euhedral type I tourmaline crystals are poorly zoned with dark cores surrounded by bright rims. Evidence above implies that these types may not crystallize directly from a B-rich melt even in granites.
Chemical compositions of the Pankou and Pukouling granites as well as the intruded marble were investigated by previous studies [5][6][7]. Previous studies [7] show that the ore-forming fluids are mainly composed of magmatic water with minor meteoric water and CO 2 derived from decarbonation of marble. Thus, these tourmalines seem not to have crystallized from post-magmatic hydrothermal metasomatism by infiltrating boron-rich fluid.
Type I tourmalines are generally intergrowth with albite and phengite. In highly fractionated magma, albite alway replace K-feldspar with the increasing degree of differentiation. And phengite is generally regarded as the product of hydrothermal alteration. While the tourmaline crystals are texturally isolated within the granite, there are also no joints or fissures around these crystals. In addition, according to previous fluid inclusion studies [4], liquid immiscibility has been proposed in the Pankou and Pukouling melts during its evolution. Hence, because of the mixture signature of magmatic and hydrothermal, it is likely that type I tourmalines are products of an immiscible boron-rich aqueous fluid released during the late-magmatic to hydrothermal transition, as observed in many magmatic-hydrothermal deposits elsewhere [13,29,33,35,37,42,44].
Like most of the magmatic tourmalines, type II tourmalines exhibit no zoning [26,33,37]. Notably type II tourmalines show a skeletal texture (Figure 5d-f) similar to the Stone Mountain tourmalines described by Longfellow and Swanson [33], and the occurrence of skeletal tourmaline crystals is cited as evidence of undercooled crystallization of magma. These authors proposed a model for skeletal tourmaline formation: the crystallization of the melt starts along the margins, tourmalines nucleate and grow from a highly fractionated and undercooled melt, which is B-rich, resulting in skeletal crystals; crystallization of skeletal tourmaline along the margins deplete B and raises the solidus temperature, resulting in crystallization at that margin and the remaining fluids then crystallize at lower temperatures, and produce euhedral tourmaline and coarse-grained minerals.
Type II tourmaline, which is skeletal, always grows along the margin of the granite. Border facies of Pankou and Pukouling granites show compositional features of highly fractionated granites, and aplite is generally thought as the product of such a granitic magma [51]. Thus, the border facies of the Pankou and Pukouling granites may closely connect with aplite. Radial tourmaline clusters are often overgrown by coarse-grained muscovite, beryl, and albite (Figure 4b,c). The skeletal type II tourmalines coexisting with muscovite, albite and K-feldspar (Figure 5d-f). The K-feldspar are obviously a replacement remnant. Mica here is mainly muscovite and is relatively less than that coexisting with type I tourmalines. It is likely that the B-rich melt intrude into triassic strata and crystallized from the margin; B and Al of the melt firstly supply the growth of tourmaline which caused the relatively small size of muscovite. Due to the low degree of crystal differentiation in the margin, K-feldspar is not completely replaced by albite. In addition, type II tourmalines generally reflect the compositions of the host rocks with the high Al total (6.01-6.43 apfu) and low Mg (0.83-1.17 apfu), Fe total (1.07-1.33 apfu) and Ca (0.05-0.17 apfu) values. As proposed above, it is most likely that type II tourmaline clusters form directly from the melts in the late-magmatic stage. Furthermore, the remaining fluids crystallized at lower undercooling, which may produce euhedral type I tourmalines found in granites and tourmaline inclusions of other coarse-grained minerals found in the veins.
Type III tourmaline occurring in the veins is always separated by muscovite layers from the granite margin and type II tourmaline clusters. These tourmalines are always found as inclusions hosted in beryl, albite, K-feldspar and quartz (Figure 4d). Type III tourmaline crystals are larger than types I and II. Abundance of K-feldspar decrease with decreasing temperature, in the mineralized veins. Albite could be found intergrowing with all the ore minerals (beryl, scheelite and cassirtite), while K-feldspar only coexisting with minor beryl. Thus, type III tourmalines seem to have crystallized in the early hydrothermal process (crystallizing later than muscovite, and earlier than K-feldspar and other coarse-grain minerals). Chemical compositions further support this model with the content of Mg (0.74-0.90 apfu), Fe total (2.16-2.40 apfu), and Ca (0.06-0.32 apfu) higher and the Al total (5.58-5.87 apfu) values relatively lower than that of type I and II tourmalines. These high Mg and Fe values are inconsistent with the Pankou and Pukouling leucogranite, thus it may be caused by the interaction between hydrothermal fluid and biotite schist (Figure 2). With the growth sequence, mineralogical characteristics, and chemical composition variations of type I, II, and III tourmalines, this can effectiely target the evolution during late-magmatic to early-hydrothermal stage of fluids Based on the evidences above, we can deduce the genesis of two types of tourmalines. At first, highly fractionated Pankou and Pukouling magma rise and intrude into triassic strata causing some joints and fissures. During this process, skeletal tourmalines (type II) form from the undercooled and B-rich melt and gradually grow into radial clusters along the margin of highly fractionated granites. Then, type I tourmalines, as suggested products of an immiscible boron-rich aqueous fluid, form in the remaining fluids. Ore-forming fluids flow into fissures and joints of the metamorphic strata, lastly forming the type III tourmalines and other coarse-grained minerals.

Chemical Evolution of Tourmaline
Tourmaline is the most important borosilicate mineral because of its ubiquity and the diversity of petrologic information that it can yield. Once formed, it does not readily readjust its composition by volume diffusion, even at relatively high temperatures [18][19][20][21][22][23][24]. This chemically complex borosilicate is mechanically and chemically refractory, found in many rock types, and stable over a wide range of geological conditions [13,14,16,17,26,27]. During its formation, tourmaline is sensitive to its chemical environment and responds to chemical changes in coexisting minerals and fluids, activities of H 2 O and dissolved species, and pressure and temperature conditions [29][30][31][32][33][34][35]. As demonstrated in Section 5.1 above, variations of tourmaline compositions may record the late-magmatic to early-hydrothermal transition.
The chemical compositions of types I and II are similar and overlap in the variation diagrams (Figures 6 and 7). In the Al-Fe-Mg diagram (Figure 6a), chemical compositions of type I and II tourmalines plot in the field of Li-poor granitoid and their associated pegmatites and aplites and metapelites and psammites. In the Fe-Mg-Ca ternary diagram (Figure 6b), the compositions of type I and II tourmalines fall in the field of Ca-poor metapelites, psammites, and calc-silicate rocks, and the near Li-poor granitoid and their associated pegmatites and aplites. Compositions of type II tourmalines are similar to type I, implying a relatively similar crystallization environment with type II to type I tourmalines. The low Ca content of tourmaline is somewhat unexpected for the Xuebaoding deposit, because the ore veins are mainly hosted in the Ca-rich metamorphic strata. It is likely that the tourmaline composition was buffered by the composition of the Pankou and Pukouling granites which are low in Ca. For these leucogranites, tourmaline is the major mafic mineral, which leads to the schorl-dravite species. In addition, as pointed out by Wu at al. [51], tourmalines usually show compositional variations from the early Mg-and Fe-bearing to later Al-bearing elbaite, in highly fractionated magmas. Thus, the relatively high Al total (nearly 6 apfu) and low Mg + Fe (<3 apfu) values of type I and II tourmalines could further provide their late-magmatic origin.
Type III tourmaline compositions plot in the field of Fe 3+ -rich quartz-tourmaline rocks, in the Fe-Mg-Ca ternary diagram (Figure 6a), and the Li-rich granitoid, pegmatites, aplites, and Li-poor granitoids, pegmatites, aplites in the Fe-Mg-Ca ternary diagrams (Figure 6b). The concentrations of Al, Mg, Fe, Ca, and Na in type III tourmaline vary significantly, compared to type I and II tourmalines. It is likely that the increasing Mg + Fe (>3 apfu) and Ca (0.06-0.32 apfu) values are caused by the reaction between fluid and triassic strata (biotite schist and marble), and the decreasing of Al and Na values are caused by the crystallization of albite and muscovite.
Type I and II tourmalines from the Xuebaoding deposit share similar characteristics. In the Al total vs. X-site vacancy diagram (Figure 7a), compositions of type I and II tourmalines show positive correlation between Al total and X-site vacancies. Linear regression of the data gives two positive slopes, implying that the variations of Al, Mg, and Na are mainly due to the Type II tourmalines formed along the margin of the granite as radial clusters. Muscovite layers separate the mineralized vein from the tourmaline clusters ( Figure 4b). Locally, tourmaline clusters are overgrown by beryl, albite, scheelite, cassiterite, and quartz (Figure 4b,c). Type II tourmaline crystals are larger than type I with a length up to 0.5-1 cm.
Type III tourmalines are found as needles in mineralized veins (Figure 4d). The very coarsegrained (2-5 cm) tourmaline crystals always formed as inclusions in other coarse-grained minerals, such as albite, quartz and beryl. Except the size, the difference between type III and type II tourmalines is that type III tourmalines always overgrow on the muscovite layers. Type III tourmaline crystals predominantly show parallel growth (Figure 4d), compared with the radial type II tourmaline.

Analytical Methods
Back scattered electron (BSE) images and chemical compositions of tourmalines were acquired at the Laboratory of Mineralization and Dynamics, Chang'an University (China), using a JXA-8100 electron microprobe analyzer (EMPA), with an acceleration voltage of 15 kV, a beam current of 10 nA, and a spot size less than 10 μm. EMPA standards include the following minerals: andradite for Si and Ca, rutile for Ti, corundum for Al, hematite for Fe, eskolaite for Cr, rhodonite for Mn, bunsenite for Ni, periclase for Mg, albite for Na, and K-feldspar for K.
The general formula of tourmaline is XY3Z6(T6O18)(BO3)3VW with X = Na + , K + , Ca 2+ , or vacancy; Y = Fe 2+ , Mg 2+ , Mn 2+ , Al 3+ , Fe 3+ , Cr 3+ ,V 3+ , Ti 4+ , and Li + ; Z = Al 3+ , Mg 2+ , Fe 3+ , Cr 3+ , V 3+ ; T = Si, Al; B = B; V = OH, O and W = OH, F, O [18]. Following the normalization procedure of [21], structural formulae of tourmaline were calculated on the basis of a total of 15 cations in the octahedral and tetrahedral (Y + Z + T) sites, all iron was assumed to be Fe 2+ . The proportion of X-site vacancies ❑ was calculated as [1 − (Na + Ca + K)]. F was not analyzed. Results of different crystals are given in Tables 1-3. 1 Al(NaR) −1 substitution vector with influence by others. In the Al total vs. Fe total (Figure 7b e margin of the granite as radial clusters. Muscovite layers urmaline clusters (Figure 4b). Locally, tourmaline clusters , cassiterite, and quartz (Figure 4b,c). Type II tourmaline th up to 0.5-1 cm. eedles in mineralized veins (Figure 4d). The very coarseays formed as inclusions in other coarse-grained minerals, size, the difference between type III and type II tourmalines is ow on the muscovite layers. Type III tourmaline crystals ure 4d), compared with the radial type II tourmaline. and chemical compositions of tourmalines were acquired Dynamics, Chang'an University (China), using a JXA-8100 ith an acceleration voltage of 15  ; T = Si, Al; B = B; V = the normalization procedure of [21], structural formulae of f a total of 15 cations in the octahedral and tetrahedral (Y + e 2+ . The proportion of X-site vacancies ❑ was calculated as esults of different crystals are given in Tables 1-3. Al(NaMg) −1 vector. In the Al(total) − X-site vacancy vs. R − X-site vacancy diagram (Figure 7c), compositional data of tourmaline is discrete. In the Mg(total) vs. Fe(total) diagram In the Xuebaoding deposit, the high Fe 3+ /Fe 2+ ratios of type III tourmaline imply a oxidized condition during hydrothermal stage of fluids, which is beneficial to precipitation of SnO 2 . The Sn-rich fluid flow into Triassic strata through the joints and fissures, with the reaction between HCl and CaCO 3 from marble the pH of the fluid increases. As mentioned by Liu et al. [7], cassiterite-quartz veins frequently are lined with coarse muscovite selvages, it is likely that the alteration of feldspar to muscovite in granitic rocks also contributes to the increasing in pH. Furthermore, mixing of ore fluids with dilute meteoric waters was shown to play a role in the Xuebaoding deposit, which could cause the decreasing chloride molality in fluids [7]. Changes above would cause the destroying of Sn-bearing complexes, the decreasing of SnO 2 solubility, and finally the large scale precipitation of Sn-bearing minerals (mainly cassiterite).
In Wood and Samson [59], the occurrence of W in high temperature hydrothermal system in different environments was compared in detail. Wood and Samson [59] proposed: (1) tungstate forms mainly as H 2 WO 4 , HWO 4 − , WO 4 2− , NaWO 4 − , Na 2 WO 4 in fluid; tungsten complexes (-chloride, -fluoride, -carbonate complexes or more exotic species) are not necessary to form an tungsten deposit; (2) the tungsten concentration in equilibrium with scheelite increases strongly with increasing temperature, NaCl concentration and pH value; (3) simple cooling of a solution with a constant Ca/Fe ratio cannot result in the precipitation of scheelite, it requires an increasing in the Ca/Fe ratio concomitant with cooling. As discussed above, because of the low Ca value of fluids which could improve the solubility of tungsten, mineralization of scheelite does not happen in late-magmatic stage. Then, with undercooled fluids flowing into triassic strata, content of Ca and pH increase during the hydrothermal evolution stage. No mafic minerals are found in the veins except tourmaline. Thus, it is likely that the increasing Mg and Fe mainly supplies the crystallization of tourmaline. Together with the decreasing Na values, finally caused the mineralization of W-bearing minerals which is mainly scheelite in Xuebaoding deposit.
In addition, as proposed by Pirajno and Smithies [60], the FeO/(FeO + MgO) ratio of tourmaline could be a useful indicator of spatial variations in granite-related hydrothermal W-Sn deposits hosted in siliclastic metasedimentary rocks. Systematic variations of the FeO/(FeO + MgO) ratio are observed from endogranitic deposits to distal vein systems emplaced at some distance from the granitic source. In Figure 2, the Pankou and Pukouling granites are closely related to Triassic metasedimentary strata. The higher Mg and Fe compositions of type III tourmaline, compared with type I and type II tourmalines, indicate that the biotite schist does contribute to the Xuebaoding deposit. With the FeO/(FeO + MgO) ratios ranging from 0.8 and 0.6, all types of tourmalines plot into the proximal to intermediate field. However, although such a result can not show the specific distance between the ore body and granites, it could be used as a reference suggesting that tourmalines formed as a result of fluid flow a distance from the intrusion (Figure 8).
It is proposed that the location in the country rocks where type I, type II and type III tourmalines precipitated should indicate also the presence of W-Sn-Be mineralization. Furthermore, as discussed above, in a low salinity and gradually cooling fluid, Ca content and pH could be the main variables controlling the solubility of W. The Ca and pH values would increase in the fluid due to the continuous reaction with Ca-rich sedimentary rocks. Thus, W-bearing minerals are expected to possibly extend considerably into marble and other Ca-rich sedimentary rock.

18
The higher Mg and Fe compositions of type III tourmaline, compared with type I and type II tourmalines, indicate that the biotite schist does contribute to the Xuebaoding deposit. With the FeO/(FeO + MgO) ratios ranging from 0.8 and 0.6, all types of tourmalines plot into the proximal to intermediate field. However, although such a result can not show the specific distance between the ore body and granites, it could be used as a reference suggesting that tourmalines formed as a result of fluid flow a distance from the intrusion (Figure 8).

Conclusions
(1) Three types of tourmalines are identified in the Xuebaoding deposit. Type I tourmalines are interpreted to have formed from an immiscible boron-rich magmatic-hydrothermal fluid. Type II tourmalines with skeletal texture formed earlier than type I, in a transition from late-magmatic to early-hydrothermal conditions, and type III are hydrothermal tourmalines, occurring in the mineralized veins.
(2) The chemical compositions of tourmaline are buffered by the host rocks. Inferred increasing Fe 3+ /Fe 2+ ratios and the decreasing Na values of all tourmalines studied suggest that they precipitated from oxidized, low-salinity fluids. (3) Mineralogical characteristics and chemical composition variations of tourmalines as established in this work may help in W-mineralization exploration in the larger region around Pinguw-Xuebaoding, or more generally in related geological settings.