Geochemical Characteristics of Mineral Assemblages from the Yamansu Iron Deposit, NW China, and Their Metallogenic Implications

: The Yamansu deposit, which is hosted in the volcanic-sedimentary sequence of the Carboniferous Yamansu Formation in Eastern Tianshan, NW China, contains many skarns, and the orebodies occur in the ore district in stratoidal, banded or lenticular forms. Four alteration stages, namely, albite–tourmaline–apatite–Grt1 (Stage I), K-feldspar–Grt2 (Stage II), magnetite–chlorite–epidote (Stage III), and quartz–calcite–axinite–Grt3 (Stage IV), are distinguished in the Yamansu deposit. The mineral geochemistry associated with each di ﬀ erent stage is presented to provide a better understanding of the corresponding metallogenic processes. The ore-forming ﬂuid in Stage I was derived from a magmatic–hydrothermal source and formed at high temperatures with many volatiles. This ore-forming ﬂuid, which contained considerable metallogenic materials during the early stage, likely experienced di ﬀ usive metasomatism in a closed system with low water / rock (W / R) ratios. Mineral geochemical analyses show that the Fe content gradually increases from Stage I to Stage II, indicating that accumulated ore-forming materials were available during changes in the physicochemical conditions from a reducing environment with neutral pH to oxidizing conditions with mildly acidic pH. During the main metallogenic stage (Stage III), mineral assemblages reﬂect moderate-to high-temperature conditions, and the ore-forming ﬂuid was created and destroyed periodically; the magnetite ores were deposited in a ﬂuctuating ﬂuid system. The multilayered orebodies, multigenerational garnets, and minerals with oscillatory zoning indicate that the ore-forming ﬂuid may have developed periodic ﬂuctuations, and this special multistage ﬂuctuation of the hydrothermal ﬂuid in the Yamansu deposit was the key factor controlling the multiple extraction, enrichment and precipitation of metallogenic materials.


Introduction
The Yamansu deposit is one of the submarine volcanic-hosted iron deposits in the Aqishan-Yamansu belt of the Eastern Tianshan, Xinjiang [1][2][3].Skarn minerals are strongly developed within this deposit, and the occurrences of mineral assemblages in the Yamansu deposit have been mentioned [4,5].However, the detailed relationships between associated skarn minerals and orebodies from different stages have not been delineated, which restricts any discussion of the corresponding metallogenic processes.Although earlier studies have largely focused on the deposit geology, magnetite geochemistry, and diagenetic and metallogenic ages of the deposit, few studies have examined the skarn mineral assemblages and their important implications for the evolution of ore-forming fluids and physicochemical conditions therein.Skarn minerals, especially those that preserve complex chemical

Regional Geology
The eastern Tianshan Orogenic Belt is located at the intersection of the Siberia, Eastern Europe, Tarim and North China Cratons, and it is an important part of the Central Asian Orogenic Belt (Figure 1a) [13][14][15].From North to South, the Eastern Tianshan orogenic belt successively occupies the Bogda-Harlik tectonic belt, the Turpan-Hami basin, the Jueluotage tectonic belt, and the Central Tianshan Massif (Figure 1b).As indicated by the distribution of ore deposit types and faults, the Jueluotage tectonic belt comprises the southern margin of the Turpan-Hami basin (copper belt), the Kangguer belt (gold belt), and the Aqishan-Yamansu belt (iron-copper-silver belt) (Figure 1b) [16,17].The Aqishan-Yamansu is one of the important subunit belts in eastern Tianshan and is well known for its iron-copper deposits that are hosted in its carboniferous submarine volcanics [1,18,19].Typical and well-known iron deposits include the Yamansu, Hongyuntan, Kumutag, and Aqishan deposits, which have similar mineralization and present orebodies that are structurally developed into banded or lens-like forms and mostly conformable with their carboniferous volcanic-sedimentary country rocks [1,3].

Ore Deposit
The Yamansu iron deposit is located in the east-central part of the Aqishan-Yamansu belt, Xinjiang, and it was discovered by the Sixth Geological Brigade of the Xinjiang Geological and Mineral Resources Bureau in 1957.The estimated ore reserve is 3549 × 10 4 t with an average grade of Figure 1.Tectonic sketch geological map of the Central Asian Orogenic Belt (a) and geological map with deposit distributions in the Eastern Tianshan Orogenic Belt (b), modified after [16].

Ore Deposit
The Yamansu iron deposit is located in the east-central part of the Aqishan-Yamansu belt, Xinjiang, and it was discovered by the Sixth Geological Brigade of the Xinjiang Geological and Mineral Resources Bureau in 1957.The estimated ore reserve is 3549 × 10 4 t with an average grade of up to 77% total Fe by the end of 2013 [20].The main lithostratigraphic units in the Yamansu ore district are the Yamansu Formation (C 1 y), and the Dikaner Formation (C 2 d) strata (Figure 2).There are five carboniferous lithologies or formation units from bottom upward: (1) marble, andesite, basalt, basaltic breccia and ignimbrite intercalated with rhyolitic ignimbrite; (2) andesitic volcanic ignimbrite, andesitic pyroclastic ignimbrite and rhyolitic ignimbrite; (3) basaltic andesite and andesitic ignimbrite; (4) limestone and limestone; and (5) andesitic ignimbrite and volcanic breccia.The exposed rocks comprising the Yamansu Formation, which hosted the Yamansu orebodies, are bimodal volcanic rocks characterized by two main geochemical serie trends (calcalkaline and alkaline) with high-Na affinity [21].The orebody-hosted volcanic rocks are mainly basalt, ignimbrite, and rhyolitic ignimbrite, which are originally derived from the parental iron-oxide melt [3,22].The main intrusion rocks and dykes are pyroxene diorite and diabase.The dominant structure is the Yamansu anticline, which has a width of 1500-2000 m, with a NE 80° trend and a dip angle of 25° [24].The volcanic breccia and crystal ignimbrite of the Yamansu Formation constitute the core of the Yamansu anticlinorium.The skarns are extensively developed in the deposit, and they are near EW direction with a strike length of about 1500 m and a width of 150-200 m (Figure 2a).The similar distribution of skarns and volcanic rocks and the basalts are partially replaced by andradite support the idea of a replacement of basaltic protolith into garnet skarn [25].The orebodies of Yamansu deposit nearly occur at the axis of anticline, which nearly follow EW-direction (trend of 80°, dip angle of 60°-75°) (Figure 2b).About 16 orebodies can be in the deposit, and they are near EW direction with a strike length of about 1500 m and a width of 150-200 m (Figure 2a).The similar distribution of skarns and volcanic rocks and the basalts are partially replaced by andradite support the idea of a replacement of basaltic protolith into garnet skarn [25].The orebodies of Yamansu deposit nearly occur at the axis of anticline, which nearly follow EW-direction (trend of 80 • , dip angle of 60 • -75 • ) (Figure 2b).About 16 orebodies can be recognized, with Fe1 and Fe2 + 3 being the major orebodies of the deposit.The orebodies are developed into structures with stratoidal, banded, or lenticular shapes and mostly show sharp contact with the country rocks.The orebodies are nearly parallel and display En Echelony distributions, and their maximum depth in the profile can reach 500 m (Figure 2b).

Paragenesis of Alteration
The dominant skarn minerals are garnet, chlorite, epidote, and K-feldspar with subordinate albite-plagioclase, amphibole, apatite, tourmaline, and axinite.The skarns usually show sharp contact with the orebodies.Based on the paragenesis and crosscutting relationships of the skarn minerals, we can identify four main mineralization stages (Figure 3).
The first metallogenic stage (Stage I) is represented by albite-tourmaline-apatite-garnet paragenesis.The alteration in this stage is characterized by minerals that contain volatile components (such as tourmaline) together with other accessory minerals.Isolated tourmaline and apatite from Stage I exhibit granular forms.The tourmaline crystals are characterized by large sizes, whereas apatite grains are smaller than tourmaline grains, and both are replaced by epidote and K-feldspar (Figure 4a).The mineralization of this stage consists dominantly of disseminated albite and magnetite (Figure 4b).The first-generation of garnets (Grt1) in the Stage I are brown or chocolate-brown with sizes ranging from 0.5-5 mm and are generally cut by magnetite veins (Figure 4c).
The first metallogenic stage (Stage I) is represented by albite-tourmaline-apatite-garnet paragenesis.The alteration in this stage is characterized by minerals that contain volatile components (such as tourmaline) together with other accessory minerals.Isolated tourmaline and apatite from Stage I exhibit granular forms.The tourmaline crystals are characterized by large sizes, whereas apatite grains are smaller than tourmaline grains, and both are replaced by epidote and K-feldspar (Figure 4a).The mineralization of this stage consists dominantly of disseminated albite and magnetite (Figure 4b).The first-generation of garnets (Grt1) in the Stage I are brown or chocolate-brown with sizes ranging from 0.5-5 mm and are generally cut by magnetite veins (Figure 4c).Within the Yamansu deposit, the second metallogenic stage (Stage II) is represented by abundant K-feldspar and garnet with subordinate diopside, which occur mainly surrounding the Fe1 orebody and in the hanging wall of the Fe2 + 3 orebodies.The second-generation of garnets (Grt2) in Stage II are generally found as massive garnet aggregates or intergrown with breccia and banded magnetite (Figure 4d,e).Figure 4f shows that Grt2 is cut by epidote vein followed by quartz + calcite vein.
The third metallogenic stage (Stage III) is the major stage of iron mineralization in the Yamansu deposit.During this stage, abundant hydrothermal minerals (such as chlorite, epidote, tremolite, and actinolite) and considerable metallic minerals (mainly magnetite) formed with subordinate quartz and sulfides.The mineral parageneses are dominated by magnetite, chlorite + epidote, and magnetite + chlorite, which are mainly cut by later quartz veins or calcite veins (Figure 4g,i).

Mineral Samples
Samples for analyses with in situ EMPA (Shimadzu, Kyoto, Japan) and LA-ICP-MS (Agilent, Santa Clara, CA, USA) were collected within the Yamansu open pit.Detailed mineral samples, descriptions, and analytical items are listed in Table 1.

Analytical Methods
The EMPA, which includes primarily the in situ analysis of major elements and back-scattered electron (BSE) observation, was performed at the EMPA laboratory at the Institute of Geology and Geophysics of the Chinese Academy of Sciences.Before the analysis, the polished thin sections were sprayed with carbon films.The experiments were conducted using a JXA-8100 instrument, with an acceleration voltage of 15 kV, an acceleration current of 20 nA, a peak counting time of 20 s, and a beam spot diameter of 3 µm were selected as the analysis conditions.Fe 2 O 3 (Fe), MnTiO 3 (Mn, Ti), diopside (Mg, Si), Cr 2 O 3 (Cr), orthoclase (Al, K), Ni (Ni), anorthite (Ca) and albite (Na) were used as standards.During the experiment, the test object was held under the secondary electron image and then selected for testing.The precision of all analyzed elements was better than 1.5%, and the lower detection limits were calculated to be on the order of 0.01 wt% for all elements.During the experiment, the test object was held under the secondary electron image and then selected for testing.Chemical formulae and end-member proportions for minerals were calculated from mineral analyses following the method of [26].
After EMPA was completed, polished thin sections of representative garnets were selected for LA-ICP-MS trace element analyses.In situ LA-ICP-MS analyses were performed at the Guangzhou Institute of Geochemistry, Chinese Academy of Sciences, using an ArF excimer laser ablation system (GeoLasPro) coupled with an Agilent 7700a ICP-MS.The details of the analytical methods are presented in [27].Element concentrations were calibrated against multiple reference materials.United States Geological Survey (USGS) standard reference materials (SRMs) NIST-610 were used as the external standards, and 29 Si was used as the internal standard.ICPMSDataCal software was used to perform the offline selection and integration of the background and analysed signals, the time-drift correction and quantitative calibration for trace element analysis [27].

Mineralogy and Petrographic Features
The tourmaline crystals from Stage I are characterized by granular textures with diameters of 0.2-5 mm and they show growth zonation (Figure 5a).In some cases, albite crystals are enclosed by tourmaline along crystal rims (Figure 5b).Apatite is also recorded to exhibit a granular texture in Stage I and is developed mainly in altered andesitic ignimbrite and tourmaline + apatite and tourmaline + garnet (Grt1) gangue minerals.Apatite crystals, which intergrown with tourmaline, show fine-grained crystals with diameters of 0.1-0.5 mm and are replaced by epidote crystals (Figure 5a).Moreover, apatite crystals intergrown with Grt1 show granular and columnar shapes (Figure 5c).The alteration and mineralization of Stage I are recorded mainly by disseminated ores.Disseminated magnetite crystals (10-200 µm in diameter) are intergrown with albite and they show euhedral to subhedral textures (Figure 5d).Replacement of magnetite by hematite is common in disseminated ores where magnetite is mostly rimmed by hematite (Figure 5d).
The Grt1 crystals in Stage I show well-preserved growth zonation.The cores of Grt1 show nearly euhedral hexagonal habitus, whereas the rims of Grt1 have large overgrowth areas under BSE observation (Figure 5e).However, the Grt2 crystals in Stage II have light yellow cores and pale or colorless rims, and Grt2 are always replaced by magnetite and quartz or cut by chlorite veins (Figure 5f). Figure 5g shows that later magnetite and quartz fill the fractures in Grt2.Occasionally, the microscopic signatures of Grt2 show irregular cores (Figure 5h), which are interpreted to indicate crystallization under nonequilibrium conditions.
As the major mineralization stage, Stage III is characterized by massive magnetite ores.The massive magnetite can be divided into two subtypes.The earlier magnetite crystals are characterized by fine euhedral habitus with growth zonation (Figure 5i), whereas the later magnetite crystals are intergrown with pyrite and are characterized by xenomorphic textures (Figure 5j).The main massive ore in Stage III consists of magnetite with a minor amount of quartz and pyrite.Microscopic features of magnetite crystals from the massive ore show granular or irregular platy textures, which are commonly replaced by quartz along their rims or fractures (Figure 5i).The axinite in the quartz + axinite vein of Stage IV is observed as euhedral wedges or rhomboidal crystals (Figure 5k).In addition, Grt3 crystals intergrown with quartz in Stage IV record thin oscillatory zoning along the rims (Figure 5l).Unlike Grt1 and Grt2 in the early stages, the Grt3 crystals in Stage IV display nearly monotonic textures and larger crystal grains >1 mm in diameter (Figure 5l).

Apatite
The electron microprobe data of apatite are listed in Table S1.Apatites in the Yamansu deposit generally show little compositional variation in P 2 O 5 (38.84%-43.43%)and CaO (55.09%-57.52%),and have low Al 2 O 3 , SiO 2 , Na 2 O, and MnO contents.The calculated F contents of apatite vary between 3.06% and 3.95% and the calculated Cl contents are very low (0.05%-0.28%).Apatite crystals have high F/Cl ratios (14.29-57.59)and F/(F+Cl) (X F ) values (0.93-0.99), which classify as fluor-apatites and formed at high temperature condition [28].All the apatite crystals from the Yamansu deposit plot on the F-rich field, which is similar to that of Kiruna deposits, in the F-Cl-OH ternary diagram (Figure 6).irregular platy textures, which are commonly replaced by quartz along their rims or fractures (Figure 5i).The axinite in the quartz + axinite vein of Stage IV is observed as euhedral wedges or rhomboidal crystals (Figure 5k).In addition, Grt3 crystals intergrown with quartz in Stage IV record thin oscillatory zoning along the rims (Figure 5l).Unlike Grt1 and Grt2 in the early stages, the Grt3 crystals in Stage IV display nearly monotonic textures and larger crystal grains >1 mm in diameter (Figure 5l).

Apatite
The electron microprobe data of apatite are listed in Table S1.Apatites in the Yamansu deposit generally show little compositional variation in P2O5 (38.84%-43.43%)and CaO (55.09%-57.52%),and have low Al2O3, SiO2, Na2O, and MnO contents.The calculated F contents of apatite vary between 3.06% and 3.95% and the calculated Cl contents are very low (0.05%-0.28%).Apatite crystals have high F/Cl ratios (14.29-57.59)and F/(F+Cl) (XF) values (0.93-0.99), which classify as fluor-apatites and formed at high temperature condition [28].All the apatite crystals from the Yamansu deposit plot on the F-rich field, which is similar to that of Kiruna deposits, in the F-Cl-OH ternary

Tourmaline
The tourmaline results of major elements analyses are shown in Table S2.Tourmaline crystals in the Yamansu deposit generally show narrow SiO 2 (33.09%-36.94%)and MgO (8.38%-10.51%)contents.According to the relationship between Fe and Mg, tourmaline crystals have low contents of FeO (5.43%-14.53%),medium-low contents of FeO/(FeO + Mg) ratios, and medium-high MgO/FeO ratios (0.67-1.54), which classify Yamansu tourmaline as dravite or Mg-rich tourmaline.The calculated B 2 O 3 contents of tourmaline vary between 9.39% and 10.39% (Table S2).The Al-Fe-Mg compositional ternary of tourmaline is generally used to reflect the host rock types or origin of tourmaline [30], and most of the Yamansu tourmaline crystals plot in the metapelites, metapsammites, quartz-tourmaline, and calc-silicate rocks fields (Figure 7).Moreover, as the main cations of X and Y sites in tourmaline crystals, the changes of Na and Fe sensitively reflect the relationship between tourmaline and their host rocks.The Yamansu tourmaline crystals have high contents of Na 2 O (0.98%-1.80%), which is consistent with the high Na 2 O contents in volcanic rocks in Yamansu ore district [31].Normally, zoned tourmalines are present.As shown in Figure 8, FeO, SiO 2 , TiO 2 , and CaO contents increase rimward, but Al 2 O 3 and Na 2 O contents decrease from the core to the rim, indicating a compositional change of sodium depletion and iron enrichment during the growth of tourmaline.
Mg compositional ternary of tourmaline is generally used to reflect the host rock types or origin of tourmaline [30], and most of the Yamansu tourmaline crystals plot in the metapelites, metapsammites, quartz-tourmaline, and calc-silicate rocks fields (Figure 7).Moreover, as the main cations of X and Y sites in tourmaline crystals, the changes of Na and Fe sensitively reflect the relationship between tourmaline and their host rocks.The Yamansu tourmaline crystals have high contents of Na2O (0.98%-1.80%), which is consistent with the high Na2O contents in volcanic rocks in Yamansu ore district [31].Normally, zoned tourmalines are present.As shown in Figure 8, FeO, SiO2, TiO2, and CaO contents increase rimward, but Al2O3 and Na2O contents decrease from the core to the rim, indicating a compositional change of sodium depletion and iron enrichment during the growth of tourmaline.

Garnet
The contents of major elements and calculated end-member compositions of garnets are listed in Table S3.Electron microprobe analyses show that garnet crystals from the Yamansu deposit belong to the andradite-grossular solid solutions with less pyralspite (sum of pyrope, spessartine, almandine, and uvarovite) (Figure 9).

Garnet
The contents of major elements and calculated end-member compositions of garnets are listed in Table S3.Electron microprobe analyses show that garnet crystals from the Yamansu deposit belong to the andradite-grossular solid solutions with less pyralspite (sum of pyrope, spessartine, almandine, and uvarovite) (Figure 9).

Garnet
The contents of major elements and calculated end-member compositions of garnets are listed in Table S3.Electron microprobe analyses show that garnet crystals from the Yamansu deposit belong to the andradite-grossular solid solutions with less pyralspite (sum of pyrope, spessartine, almandine, and uvarovite) (Figure 9).Grt1 have narrow compositional ranges in SiO2 and CaO but wide compositional ranges in FeOT and Al2O3.The contents of MnO and MgO are remarkably low (less than 1%).The end-member compositions of Grt1 show dramatic changes consisting of And25.33-99.88Gro0.00-72.42 with minor Pys0.12-2.40(Figure 9a).Profile along the crystal zonation of Grt1 (sample Y7729-1) records compositional Grt1 have narrow compositional ranges in SiO 2 and CaO but wide compositional ranges in FeO T and Al 2 O 3 .The contents of MnO and MgO are remarkably low (less than 1%).The end-member compositions of Grt1 show dramatic changes consisting of And 25.33-99.88Gro 0.00-72.42with minor Pys 0.12-2.40(Figure 9a).Profile along the crystal zonation of Grt1 (sample Y7729-1) records compositional changes where Fe content decreases first and then increases rimward, but Al content increases first and then decreases with negligible changes in Ca and Si content in the core (Figure 10a).
Compared with Grt1, Grt2 show narrow compositional ranges in SiO 2 and CaO but have relatively higher FeO T and lower Al 2 O 3 contents.Similarly, MnO and MgO contents of Grt2 are remarkably low (less than 1.00%).The calculated end-member compositions of Grt2 also show dramatic changes consisting of And 38.71-99.94Gro 0.00-58.80with minor Pys 0.06-2.49(Figure 9b).Well-preserved chemical growth zonations and their analysis results generally show a general decrease in FeO T or andradite contents and increase in Al 2 O 3 or grossular contents from core to rim (Figure 10b,c).For the sample Y5817-5, SiO 2 and CaO contents gradually increase, and FeO T content decreases first and then increases but Al 2 O 3 content increases first and then decreases from core to rim (Figure 10b).The variations of SiO 2 and CaO contents in the sample Y5818-3 are smooth from core to rim.However, FeO T (or andradite) and Al 2 O 3 (or grossular) contents show "M" and "W" patterns in variations, respectively, from core to rim (Figure 10c).These features indicate clear indications of multi-stage growth.In general, there is a negative correlation between FeO T and Al 2 O 3 contents in Yamansu garnets.

Axinite
The representative major element chemical compositions of axinites are shown in Table S5.The SiO2 and CaO contents range from 42.66% to 43.84% and from 19.16 to 19.88%, respectively.The average contents of FeO and Al2O3 in Yamansu axinite are 7.16% and 15.79%, respectively.The calculated B2O3 contents vary from 6.63 to 7.02%.According to the FeO/(MnO + FeO) ratios (0.52-0.72) and Mg-Al-Fe triangular diagram, Yamansu axinites belong to the feraxinite (Figure 12).

Axinite
The representative major element chemical compositions of axinites are shown in Table S5.The SiO 2 and CaO contents range from 42.66% to 43.84% and from 19.16 to 19.88%, respectively.The average contents of FeO and Al 2 O 3 in Yamansu axinite are 7.16% and 15.79%, respectively.The calculated B 2 O 3 contents vary from 6.63 to 7.02%.According to the FeO/(MnO + FeO) ratios (0.52-0.72) and Mg-Al-Fe triangular diagram, Yamansu axinites belong to the feraxinite (Figure 12).

Trace Elements
The trace element analyses by LA-ICP-MS in the Yamansu deposit are based on different generations of garnets.The cores from each generation garnet were analyzed and their chemical compositions are presented in Table S6.

Trace Elements
The trace element analyses by LA-ICP-MS in the Yamansu deposit are based on different generations of garnets.The cores from each generation garnet were analyzed and their chemical compositions are presented in Table S6.

Trace Elements
The trace element analyses by LA-ICP-MS in the Yamansu deposit are based on different generations of garnets.The cores from each generation garnet were analyzed and their chemical compositions are presented in Table S6.

Fluid Properties
The ore-forming fluid in a magmatic-hydrothermal system is characterized by large amounts of volatiles and dissolved metal elements [36,37].The presence of apatite and tourmaline is well developed in Stage I, indicating that the Yamansu ore-forming fluid formed under high temperatures with considerable volatiles.The Grt1 in Stage I show variations in composition and have grossular-rich cores without oscillatory zoning, which reflects retrograde high-temperature minerals [38].The characteristics of the REEs from Grt1 (Figure 14a) are similar to those resulting from REE partitioning in high-temperature fluids and silicate melts (relative depletions in REEs and variable Eu anomalies), implying that the ore-forming fluid in Stage I was derived from a magmatic-hydrothermal source [39,40].This interpretation is confirmed by melt inclusions in the garnets of the early stage [41].In addition, abrupt compositional variations between the cores and rims (Figure 5e) indicate that Grt1 grew under rapidly changing physicochemical conditions, which are interpreted as changes in the composition-dependent fluid influx [42,43].

Fluid Properties
The ore-forming fluid in a magmatic-hydrothermal system is characterized by large amounts of volatiles and dissolved metal elements [36,37].The presence of apatite and tourmaline is well developed in Stage I, indicating that the Yamansu ore-forming fluid formed under high temperatures with considerable volatiles.The Grt1 in Stage I show variations in composition and have grossular-rich cores without oscillatory zoning, which reflects retrograde high-temperature minerals [38].The characteristics of the REEs from Grt1 (Figure 14a) are similar to those resulting from REE partitioning in high-temperature fluids and silicate melts (relative depletions in REEs and variable Eu anomalies), implying that the ore-forming fluid in Stage I was derived from a magmatic-hydrothermal source [39,40].This interpretation is confirmed by melt inclusions in the garnets of the early stage [41].In addition, abrupt compositional variations between the cores and rims (Figure 5e) indicate that Grt1 grew under rapidly changing physicochemical conditions, which are interpreted as changes in the composition-dependent fluid influx [42,43].In Stage II, the ore-forming fluid probably inherited the properties of the early fluid with a high temperature but a different fluid composition.K-feldspar is represented in this stage, and hydrothermal fluids with alkali-rich contents can promote the enrichment and migration of iron contents [45].During the main metallogenic stage (Stage III), the occurrences of magnetite and retrograde skarn minerals, such as chlorite and epidote, reflect moderate-to high-temperature conditions for the ore-forming fluid [46].Yttrium and REEs have similar geochemical properties and they show almost linear correlations.The Yamansu Grt2 samples, however, show insignificant correlations between REEs and Y (Figure 14b), which might reflect either local protolith chemical heterogeneities or changes in the fluid composition during Stage II [10,47].Moreover, the fractionation of Y and Ho twin pairs from the chondritic ratio (Figure 14c) is an indication that Grt2 growth occurred in a complex hydrothermal fluid system [44,48].This result is supported by carbon and oxygen isotope data [5].
The representative quartz + sulfides, quartz + calcite, and quartz + axinite veins that cut early-stage mineral assemblages (Figure 4h,i) signify that the ore-forming fluid in Stage IV was generated under low-temperature conditions after iron precipitation.Although Grt3 crystals are present in Stage IV, they have low ΣREE, Y, and Y/Ho contents, and high contents of HFSEs (Figure 14c,d).This disparity results from differences in the fluid composition and fluid flow rate [12].
Minerals 2020, 10, 39 16 of 23 6.1.2.Oxygen Fugacity Studies of end-member compositions in garnets suggest that increasing oxygen fugacity (f O 2 ) may result in an increase in andradite content during garnet growth [42,49].The Yamansu Grt1 cores are dominated by grossular end-member compositions, and andradite compositions increase from the Grt1 rims to the Grt2 cores (Figure 5e-h), suggesting an increase in f O 2 from Stage I to Stage II [50].Theoretical and experimental data for variable valence elements (such as Eu, U, and Cr) in hydrothermal fluids also imply that increasing high-valence activity with increasing f O 2 during the fluid process may result in increases in the corresponding element contents [51][52][53].The results showing increasing δEu and U contents from Grt1 to Grt2 (Figure 15a

pH
In addition to the fluid properties in a hydrothermal system, the temperature, cation type, and fluid density also play important roles in the pH of the fluid.Therefore, it is inappropriate to separately describe the pH based on fluid composition [56].Fortunately, REEs and some immobile trace elements in hydrothermal fluids are scarcely affected in magmatic-hydrothermal fluids, and pH is largely correlated with their total contents in the fluid [44,57].It was demonstrated by [52] that LREE-depleted and HREE-enriched patterns with negative or no Eu anomalies form under nearly neutral pH, whereas LREE-enriched and HREE-depleted patterns develop under mildly acidic pH, and the Eu anomalies are positive or significantly controlled by the complexing agents (mainly F, Cl and B).The Yamansu Grt1 crystals are characterized by variable LREE/HREE and δEu values (both show values >1 and <1) (Figures 14a and 15a), which may be due to the influences of the complexing agents (such as F, Cl, and P) in the early stage [7,58].This interpretation agrees well with the presence of apatite and tourmaline in Stage I.In comparison, the Grt2 crystals show LREE-enriched and HREE-depleted patterns with positive Eu anomalies (Figure 13), suggesting that a mildly acidic pH is recorded for Stage II.This particular pH condition is favorable for the enrichment and mineralization of Fe contents in ore-forming fluids [59].
An investigation by [57] of thermal fluids found that pH strongly controls the changes in REEs and Ce anomalies and shows obvious negative correlations with ΣREE and δCe.Moreover, theoretical and thermodynamic experiments suggested that pH significantly affects the fractionation of ΣREE and Nd and that pH is negatively correlated with ΣREE and Nd in hydrothermal fluids [52,60].An early increasing trend followed by a decreasing trend for the ΣREE, δCe, and Nd contents from Stage I to Stage IV (Figure 15c,d) indicate a decrease and then an increase in pH during the evolution of the Yamansu hydrothermal fluid.

Ore-Forming Fluid Evolution
The mineral assemblages of the Yamansu deposit and their geochemical characteristics may reflect the evolution of the hydrothermal fluid.The results for the Grt1 crystals with low ΣREE and LILEs contents but high HFSEs, Y, and Ho contents (Figures 14 and 15) indicate that the ore-forming fluid in Stage I formed under near equilibrium condition within a closed system with low W/R ratios [40,61].During the early stage, the ore-forming fluid likely experienced diffusive metasomatism, and the fluid flow and mineral growth rates were slow [62].From Stage II to Stage III, the fluid commonly shows major changes, namely a fluctuating composition associated with minerals ranging from garnet (oscillatory zoning, Figure 5g,h) to magnetite (oscillatory zoning, Figure 5i) and chlorite.Such dynamic and repetitive enrichments and depletions in composition require the fluid flow to have been episodic, which may have been caused by periods of metasomatism between hydrothermal fluid and wall rocks [63].In particular, irregular Grt2 cores, which are characterized by high ΣREE, LREE/HREE, δEu, and δCe contents, and deviations of Y/Ho ratios from the chondritic linear trend (Figures 14 and 15), have been proposed to be responses to the nonequilibrium system [40,62].This situation was mostly followed by high W/R ratios and relatively high f O 2 under infiltrative metasomatism with the introduction of a batch of externally derived fluid [64].Note that the decreasing andradite (Figures 5h and 10b) indicates that the Fe content in the fluid decreased gradually and could be explained by introducing ore-forming material (Fe) into the late stage (Stage III).This evolutionary interpretation is supported by the replacement of garnets in Stage II by subsequent magnetite in Stage III (Figure 4d,e and Figure 5f,g).

Implications for Mineralization
Although skarn minerals such as garnet, chlorite, and epidote are present, skarn zones and associated intrusive rocks have not been identified in the Yamansu ore district [2], implying the absence of an obvious relationship between the formation of skarns and intrusive rocks.In addition, the presence of albite, apatite, and tourmaline coupled with the near absence of diopside in the early stage and the multigeneration garnets, together with the chlorite and epidote skarns cut by later garnet veins (Figure 4h,i), indicate that Yamansu differs from typical contact metasomatic skarn-type deposits.Previous studies on the field geology and geochronology of the diagenesis and mineralization of this deposit attributed the Yamansu mineralization to volcanic rocks [1][2][3]5].As discussed above, the geochemistry of tourmaline has a homologous relationship with volcanic rocks; additionally, the ore-forming fluid formed under high temperatures with considerable volatiles in Stage I. Therefore, the Yamansu ore-forming fluid was likely derived from a volcanic magmatic-hydrothermal fluid.
Studies have suggested that metallogenic materials in fluids are transported mainly as complex compounds in most hydrothermal deposits [65,66].Volatile elements (such as B and Cl) and alkaline elements (such as Na and K) easily form complex compounds, and these complex compounds may have transported considerable metallogenic materials in the early stage of the Yamansu deposit.This hypothesis is supported by the Cl-rich fluid in the early stage of Yamansu mineralization [67].The Grt1 crystals have the highest contents of siderophile elements (Sc, V, Cr, Co, and Zn) (Figure 15f), indicating that these elements were enriched in the fluid from the early stage.The preserved concentration of Fe in the early stage fluid was low, as recorded by the Grt1 cores and tourmaline cores (Figure 16a); therefore, reaching saturation of the Fe concentration for precipitation was difficult.Furthermore, the temperature did not significantly decrease in Stage I, which hindered the precipitation of ore minerals [67].From Stage I to Stage II, mineral geochemistry shows that the Fe content increased gradually (Figure 16b), implying that ore-forming materials accumulated within the ore-forming fluid.The Yamansu ore-forming fluid experienced changing physicochemical conditions, namely, from reducing conditions with neutral pH to oxidizing conditions with mildly acidic pH from Stage I to Stage II (Figure 16a,b).These changes were favorable for the transport and accumulation of ore-forming materials.During the migration of an ore-forming fluid, mineral precipitation occurs when fluid properties or physicochemical conditions change with time.Thus, the permeability may eventually be reduced, obstructing fluid channels [68].Then, a strong pressure increase could have caused hydraulic fracturing when the hydrostatic pressure exceeded the lithostatic pressure.It follows, then, that breccia magnetite ores were present (Figure 4d).However, the permeability increased as a result of the opening of fractures and a decrease in pressure, which in turn promoted mineral formation.Ultimately, the fluid system was created and destroyed periodically (Figure 16c); and consequently, the magnetite composition may have been controlled by dissolution-reprecipitation processes under fluctuating physicochemical conditions [69,70].Therefore, massive ores containing magnetite exhibiting oscillatory zoning were deposited under these fluctuating conditions during Stage III (Figure 5i).Nevertheless, Grt3 formed in Stage IV due to the depletion of ore-forming materials (Fe) in the fluid and the presence of unfavorable physicochemical conditions (such as low fO 2 and low W/R ratios) (Figure 16d) for the precipitation of iron; but quartz, sulfides, axinite, and calcite precipitated in the late stage.
destroyed periodically (Figure 16c); and consequently, the magnetite composition may have been controlled by dissolution-reprecipitation processes under fluctuating physicochemical conditions [69,70].Therefore, massive ores containing magnetite exhibiting oscillatory zoning were deposited under these fluctuating conditions during Stage III (Figure 5i).Nevertheless, Grt3 formed in Stage IV due to the depletion of ore-forming materials (Fe) in the fluid and the presence of unfavorable physicochemical conditions (such as low fO2 and low W/R ratios) (Figure 16d) for the precipitation of iron; but quartz, sulfides, axinite, and calcite precipitated in the late stage.

Conclusions
Four alteration and mineralization stages are distinguished in the Yamansu deposit based on the field geology, petrographic observations, and mineral geochemistry studies.The ore-forming fluid was derived from a magmatic-hydrothermal source and contained considerable volatiles and metallogenic materials in the early stage but experienced different physicochemical conditions in the subsequent stages.The magnetite ores of the major mineralization stage were deposited in a fluctuating fluid system.Multistage fluctuations of the hydrothermal fluid may have been the main mechanism for the enrichment and precipitation of metallogenic materials.

Supplementary Materials:
The following are available online at http://www.mdpi.com/2075-163X/10/1/39/s1,Table S1: EMPA data of the apatite in the Yamansu deposit, Table S2: EMPA data of the tourmaline in the Yamansu deposit, Table S3: EMPA data of the different generations of garnet in the Yamansu deposit, Table S4: EMPA data of the plagioclase and K-feldspar in the Yamansu deposit, Table S5: EMPA data of the axinite in the Yamansu deposit, Table S6: LA-ICP-MS trace element (ppm) compositions of different generations of garnets in the Yamansu deposit.
Author Contributions: J.W., Y.W., and L.L. offered advice to this work; Z.S., Y.W., and L.L. finished the field work; Z.S. performed the experiments and organized this paper.All authors have read and agreed to the published version of the manuscript.

Minerals 2020 , 24 Figure 1 .
Figure 1.Tectonic sketch geological map of the Central Asian Orogenic Belt (a) and geological map with deposit distributions in the Eastern Tianshan Orogenic Belt (b), modified after [16].

Figure 2 .
Figure 2. Geological map (a) and geological section (b) of the Yamansu iron deposit (modified from [23]).The dominant structure is the Yamansu anticline, which has a width of 1500-2000 m, with a NE 80 • trend and a dip angle of 25 • [24].The volcanic breccia and crystal ignimbrite of the Yamansu Formation constitute the core of the Yamansu anticlinorium.The skarns are extensively developed

Figure 3 .
Figure 3. Alteration and mineralization paragenesis of the Yamansu iron deposit.Figure 3. Alteration and mineralization paragenesis of the Yamansu iron deposit.

Figure 3 .
Figure 3. Alteration and mineralization paragenesis of the Yamansu iron deposit.Figure 3. Alteration and mineralization paragenesis of the Yamansu iron deposit.

Figure 3 .
Figure 3. Alteration and mineralization paragenesis of the Yamansu iron deposit.

Figure 4 .
Figure 4. Photographs of representative samples and their relationships in the Yamansu deposit.(a) Granular tourmaline and apatite (fine-grained) replaced by epidote and K-feldspar; (b)

Figure 5 .
Figure 5. Photomicrographs of representative samples from different stages and their relationships in the Yamansu deposit.(a) Tourmaline (growth zonation) and apatite (fine-grained) crystals replaced by epidote (plane polarized light); (b) Albite crystals replaced or enclosed by tourmaline along their rims (cross polarized light); (c) Granular and rod-shaped apatite enclosed by Grt1 (back-scattered electron (BSE) image); (d) Disseminated euhedral to anhedral magnetite crystals with diversely 5-100 μm grains that are intergrown with albite and replaced by hematite (reflected light); (e) Grt1 characterized by core-rim growth zonation (BSE image); (f) Grt2 (with yellow cores and pale rims) show replacement by magnetite and quartz, or cut by chlorite vein (plane polarized light); (g) A Grt2 shows well-preserved growth zonation and replaced by magnetite and quartz (BSE image); (h) Euhedral Grt2 with growth zonation of irregular core (BSE image); (i) Magnetite crystals in massive ores have continuous alternate zonation with a brighter central zone and replaced by quartz (BSE image); (j) Magnetite crystals coexisting with sulfide (mainly fine to coarse grained pyrite) have xenomorphic shapes (BSE image); (k) Axinite (stumpy wedge or rhomboidal) and quartz superimposed on the Grt3 (plane polarized light); (l) Zoned Grt3 (with oscillatory zoning in the rims) accompanied with quartz (plane polarized light in 1 and cross polarized light in 2).Abbreviations: Tur-Tourmaline; Ap-Apatite; Ep-epidote; Ab-Albite; Grt1-Garnets of the first generation; Grt2-Garnets of the second generation; Grt3-Garnets of the third generation; Mt-magnetite; Hem-hematite; Qtz-quartz; Chl-chlorite; Py-pyrite; Ax-Axinite.

Figure 5 .
Figure 5. Photomicrographs of representative samples from different stages and their relationships in the Yamansu deposit.(a) Tourmaline (growth zonation) and apatite (fine-grained) crystals replaced by epidote (plane polarized light); (b) Albite crystals replaced or enclosed by tourmaline along their rims (cross polarized light); (c) Granular and rod-shaped apatite enclosed by Grt1 (back-scattered electron (BSE) image); (d) Disseminated euhedral to anhedral magnetite crystals with diversely 5-100 µm grains that are intergrown with albite and replaced by hematite (reflected light); (e) Grt1 characterized by core-rim growth zonation (BSE image); (f) Grt2 (with yellow cores and pale rims) show replacement by magnetite and quartz, or cut by chlorite vein (plane polarized light); (g) A Grt2 shows well-preserved growth zonation and replaced by magnetite and quartz (BSE image); (h) Euhedral Grt2 with growth zonation of irregular core (BSE image); (i) Magnetite crystals in massive ores have continuous alternate zonation with a brighter central zone and replaced by quartz (BSE image); (j) Magnetite crystals coexisting with sulfide (mainly fine to coarse grained pyrite) have xenomorphic shapes (BSE image); (k) Axinite (stumpy wedge or rhomboidal) and quartz superimposed on the Grt3 (plane polarized light); (l) Zoned Grt3 (with oscillatory zoning in the rims) accompanied with quartz (plane polarized light in 1 and cross polarized light in 2).Abbreviations: Tur-Tourmaline; Ap-Apatite; Ep-epidote; Ab-Albite; Grt1-Garnets of the first generation; Grt2-Garnets of the second generation; Grt3-Garnets of the third generation; Mt-magnetite; Hem-hematite; Qtz-quartz; Chl-chlorite; Py-pyrite; Ax-Axinite.

Figure 8 .
Figure 8. Major chemical element compositional profiles for the Yamansu zoned tourmaline.(a) Growth zonation and test points of tourmaline; (b,c) Variations of elements in compositional profile of tourmaline.

Minerals 2020 , 24 Figure 10 . 4 . 24 Figure 10 .
Figure 10.Back-scattered electron and cross polarized light images for different generations of garnets and their compositional profiles in the Yamansu deposit.

Figure 13 .
Figure 13.Chondrite-normalized rare earth element (REE) distribution patterns for three generations of garnets from the Yamansu deposit.The chondrite-normalized values were calculated by [35].

24 Figure 15 .
Figure 15.Binary plots of (a) proportions of andradite in garnet versus Eu anomalies (δEu), (b) proportions of andradite in garnet versus U, (c) ΣREE versus δCe, (d) ΣREE versus Nd, (e) proportions of andradite in garnet versus large ion lithophile elements (LILE), and (f) proportions of andradite in garnet versus siderophile elements (sum of Sc, V, Cr, Co, and Zn) for garnets from the Yamansu deposit.6.1.3.pH In addition to the fluid properties in a hydrothermal system, the temperature, cation type, and fluid density also play important roles in the pH of the fluid.Therefore, it is inappropriate to separately describe the pH based on fluid composition [56].Fortunately, REEs and some immobile trace elements in hydrothermal fluids are scarcely affected in magmatic-hydrothermal fluids, and

Figure 16 .Figure 16 .
Figure 16.Schematic timeline of fluid evolution model for the Yamansu deposit.(a) Ore-forming fluid in Stage I was characterized by high temperatures with considerable volatiles, low low fO2 and low W/R ratios.Ore-forming fluid showed increasing Fe contents in the early stage; (b) The Fe content in the ore-forming fluid increased from Stage I to Stage II.The physicochemical conditions were characterized by high fO2, mildly acidic pH, and high W/R ratios in Stage II.The decreased Fe contents were favorable for accumulation of ore-forming materials in Stage II; (c) The decreasing temperature, periodical changes of fluid system, and fluctuating physicochemical conditions caused Figure 16.Schematic timeline of fluid evolution model for the Yamansu deposit.(a) Ore-forming fluid in Stage I was characterized by high temperatures with considerable volatiles, low low f O 2 and low W/R ratios.Ore-forming fluid showed increasing Fe contents in the early stage; (b) The Fe content in the ore-forming fluid increased from Stage I to Stage II.The physicochemical conditions were characterized by high f O 2 , mildly acidic pH, and high W/R ratios in Stage II.The decreased Fe contents were favorable for accumulation of ore-forming materials in Stage II; (c) The decreasing temperature, periodical changes of fluid system, and fluctuating physicochemical conditions caused the deposition of oscillatory zoning magnetite in Stage III; (d) Ore-forming fluid in Stage IV was characterized by depletion of ore-forming materials (Fe contents), low f O 2 , and low W/R ratios.

Table 1 .
The mineral samples and their characteristics from the Yamansu deposit.