Texture and Trace Element Composition of Rutile in Orogenic Gold Deposits


 Rutile from a wide range of orogenic gold deposits and districts, including representative world-class deposits, was investigated for its texture and trace element composition using scanning electron microscopy, electron probe microanalysis, and laser ablation-inductively coupled plasma-mass spectrometry. Deposits are hosted in various country rocks including felsic to ultramafic igneous rocks and sedimentary rocks, which were metamorphosed from lower greenschist to middle amphibolite facies and with ages of mineralization that range from Archean to Phanerozoic. Rutile presents a wide range of size, texture, and chemical zoning. Rutile is the dominant TiO2 polymorph in orogenic gold mineralization. Elemental plots and partial least square-discriminant analysis suggest that the composition of the country rocks exerts a strong control on concentrations of V, Nb, Ta, and Cr in rutile, whereas the metamorphic facies of the country rocks controls concentrations of V, Zr, Sc, U, rare earth elements, Y, Ca, Th, and Ba in rutile. The trace element composition of rutile in orogenic gold deposits can be distinguished from rutile in other deposit types and geologic settings. Elemental ratios Nb/V, Nb/Sb, and Sn/V differentiate the rutile trace element composition of orogenic gold deposits compared with those from other geologic settings and environments. A binary plot of Nb/V vs. W enables distinction of rutile in metamorphic-hydrothermal and hydrothermal deposits from rutile in magmatic-hydrothermal deposits and magmatic environments. The binary plot Nb/Sb vs. Sn/V distinguishes rutile in orogenic gold deposits from other geologic settings and environments. Results are used to establish geochemical criteria to constrain the source of rutile for indicator mineral surveys and potentially guide mineral exploration.

Rutile is stable over a wide range of pressure and temperature conditions (up to 30 kbar and 900°C; Liou et al., 1998). Rutile is a TiO2 polymorph with anatase and brookite. Anatase and brookite are stable at low temperature and pressure, whereas rutile becomes dominant at medium to high pressure and temperature (Dachille et al., 1968;Jamieson and Olinger, 1968). Rutile is among the most stable minerals during weathering and surficial transport (Hubert, 1962;Force, 1980;Morton and Hallsworth, 1994). Rutile's specific gravity is 4.2, and its hardness on the Mohr scale is 6.5 (Bonewitz, 2008); therefore, it is commonly found with other heavy minerals in beaches and rivers and their lithified equivalents (Zack et al., 2002(Zack et al., , 2004b. Rutile specific gravity and hardness are features that favor long distance transport and concentration in physical traps. However, because rutile occurs in a broad range of mineral deposits and geologic environments, it is crucial to develop tools to determine its provenance in indicator mineral surveys. Previous studies have shown that in VMS, Sn, Archean porphyry Au-Mo, and orogenic gold deposits, rutile from the mineralization commonly has elevated concentrations of Sn, W, and Sb, whereas rutile from barren host rocks has lower concentrations of these elements (Clark and Williams-Jones, 2004;Pochon et al., 2017;Agangi et al., 2019;Porter et al., 2020). This shows that rutile minor and trace element compositions have a high potential to differentiate rutile in mineralized from barren rocks. In provenance studies, the Nb/Cr ratios of detrital rutile are known to distinguish pelitic from mafic source rocks (Zack et al., 2002(Zack et al., , 2004bStendal et al., 2006a;Triebold et al., 2007;Meinhold et al., 2008).
Rutile can be used to date mineralization by U/Pb geochronology . Rutile is also useful to calculate temperature in eclogite and granulite facies metamorphic rocks using the Zr-in-rutile (Zack et al., 2004a;Watson et al., 2006;Tomkins et al., 2007;Luvizotto and Zack, 2009). Detrital rutile can be used as a provenance indicator for sedimentary rocks based on the Nb and Cr contents (Zack et al., 2004b;Stendal et al., 2006b;Triebold et al., 2007;Meinhold et al., 2008;Morton and Chenery, 2009;Okay et al., 2011).
Here, we document the texture and the trace element composition of rutile associated with mineralization in orogenic gold deposits. Our study focuses on a range of representative orogenic gold deposits. We test the influence of the geologic settings of the orogenic gold deposits, including the composition and the metamorphic facies of the country rocks, as well as the deposit age, on the rutile trace element composition. Results are used to develop a robust geochemical tool to constrain rutile sources for indicator mineral surveys and, thus, orient mineral exploration targeting. Results for orogenic gold deposits are compared with compositional data from the literature to develop criteria based on the trace element composition in order to discriminate rutile in orogenic gold deposits from rutile in other deposit types and geologic environments.

Trace element composition of TiO2 polymorphs in gold deposits
Anatase and brookite commonly have low Fe, Cr, V, W, Sn, and Sb contents compared with rutile (Triebold et al., 2011;Plavsa et al., 2018). Brookite has high Al content, rutile has low Al content, and the Al content of anatase is intermediate between the two other polymorphs (Plavsa et al., 2018). Plavsa et al. (2018) highlighted that anatase and brookite trace element compositions coincide with rutile from barren rocks of Clark and Williams-Jones (2004), characterized by low W and Sn concentrations. Therefore, they pointed out the importance of classifying TiO2 polymorphs before using elemental ternary classification diagrams for gold deposits as described by Clark and Williams-Jones (2004) to avoid false-positive results. TiO2 polymorphs are reliably classified using the ternary diagram of Plavsa et al. (2018). TiO2 polymorphs can be identified using laser Raman spectroscopy or electron backscatter diffraction (Meinhold, 2010;Plavsa et al., 2018).

Geologic Settings of the Selected Orogenic Gold Deposits
Rutile from 41 orogenic gold deposits and districts was selected for this study, among which 22 are considered as world-class deposits or districts (>70 t Au), as defined by Goldfarb et al. (2005). Deposits were selected to test the vari-ability of the geologic settings including the composition and metamorphic facies of country rocks and the age and style of mineralization (Table 1). Country rocks include volcanic and plutonic rocks with ultramafic to felsic compositions and clastic sedimentary rocks. In the following, we differentiate between deposits hosted only in mafic rocks and deposits where ultramafic rocks occur with mafic rocks. The metamorphic facies of the country rocks range from lower greenschist (e.g., Kanowna Belle, Essakane) to middle amphibolite facies (e.g., Big Bell). Country rocks of selected deposits are most commonly Archean, but some deposits are hosted in Proterozoic (e.g., Rosebel, Obuasi) or Phanerozoic (e.g., Kumtor, Macraes) rocks. Most deposits formed during the Archean, although some formed during the Proterozoic (e.g., Meliadine, Essakane) and in the Phanerozoic (e.g., Kumtor, Juneau). In most cases, the mineralization age is close to the age of the country rock, with the exception of the Proterozoic Meliadine and Rosebel districts that are hosted in Archean rocks (Table 1).

Sample preparation
Samples were collected from the selected deposits, and others were donated by researchers or mining companies (Table 2).
In total, 147 thin sections containing rutile from 41 orogenic gold deposits were studied using optical microscopy. Details concerning provenance of the samples are presented in Table 1. The following results are based on a variable number  of thin sections per deposit (1-19) and include rutile grains from Big Bell (Mueller et al., 1996), which were mounted in epoxy polished section.

Scanning electron microscopy
Thirty-three thin sections were examined using scanning electron microscopy (SEM) to document mineral texture, zoning, and mineral associations using energy wavelength dispersive spectroscopy (EDS) and backscattered electron (BSE) imaging on a JEOL JSM-840A SEM at Université Laval (Québec, Canada).

Electron probe microanalysis
Ninety-two thin sections were selected for electron probe microanalysis (EPMA), and 304 points were analyzed. EPMA data are reported in Appendix 1,  Smith et al. (1984); Mason andMelnik (1986a, 1986b); Wood et al.  Bold font indicates deposit with more than 70 t Au considered to be world-class gold after Goldfarb et al. (2005); in brackets: data assumed according to regional geology 1 Controversial origin method is described in detail in Table A1. Chemical maps were acquired at 50 nA and 25 kV on rutile from Muruntau, Hollinger, and Big Bell (Fig. 1). Those specimens were chosen to document the chemical zoning associated with the optical zoning observed under SEM. As rutile occurs commonly with poikilitic to skeletal texture and can be very fine grained, matrix effects were carefully checked on the EPMA analyses. As described below, data below the detection limits (DLs) were discarded.

Laser ablation-inductively coupled plasma-mass spectrometry
A total of 295 points and lines were analyzed using laser ablation-inductively coupled plasma-mass spectrometry (LA-ICP-MS). Data are presented in Table A4. LA-ICP-MS trace element analyses were performed on rutile using an ASI RESOlution S-155 Excimer 193 nm laser coupled to an Agilent 7,700× ICP-MS at the LabMaTer (Université du Québec à Chicoutimi, Canada). Laser repetition rate was constant at 15 Hz with laser beam energy maintained at 3 J/ cm 2 . Lines with a laser beam size of 25 to 44 µm and spot diameters of 25 to 55 µm were used depending on grain size and distribution of inclusions. The laser speed was set up at 10 µm/s for the laser lines. 47 Ti and 49 Ti were measured with dwell times of 1 ms and other trace elements with dwell time of 4 ms. The background was counted for 30 s, and the acquisition time was 60 to 70 s. Mean concentration of Ti determined by thin section with EPMA was used for data reduction with the software Iolite (Igor Pro 6.31) using data reduction schemes from Woodhead et al. (2007). The standard glass GSD-1g was used as the external standard. The standards rutile R10 , and glasses including GSE-1g, Gprob6-A, BC-28 (Barnes et al., 2004), NIST-610, and NIST-612 were analyzed before and after a run and every hour. They were used as unknown samples to check data quality. Rutile grain R13 (Schmitt and Zack, 2012) Table A3.
Three maps were performed on rutile from Canadian Malartic ( Fig. 2), Tindals, and Big Bell with 11 and 19 µm beam sizes.
EPMA and LA-ICP-MS data are compared using the Pearson correlation coefficient for Ti, Si, Al, Mn, Mg, As, Fe, V, Cr, Nb, Zr, Sn, Sb, Ta, and W in Appendix 1, Figure A1. Aluminum and As are overestimated using EPMA (Fig. A1c, f), which may be the result of poor detection with EPMA for Al (median EPMA DL = 517 ppm; Table A1). The As content in rutile (median 75 ppm As) is commonly near the DL (33 ppm; Table A1), which explains the poor EPMA data quality. Silicon and Mg are overestimated using LA-ICP-MS (Fig. A1b, e), which may result from the high DL for Si using LA-ICP-MS analyses. The Mg LA-ICP-MS values plot along a vertical array at about 10 to 100 ppm. Thus, Mg LA-ICP-MS data are not used. EPMA vs. LA-ICP-MS data for Fe, V, Cr, Nb, Zr, Sn, Sb, Ta, and W have a Pearson correlation coefficient r above 0.59 (Fig. A1). Considering EPMA and LA-ICP-MS data quality, the Mg, Si, V, Cr, Fe, Zr, Nb, Sn, Sb, Ta, and W contents measured using EPMA, and the Al, Ca, Sc, Mn, Co, Ni, Zn, As, Sr, Y, Mo, Ba, REE, Hf, Pb, Th, and U contents measured using LA-ICP-MS are used and described in the following. If the Pearson correlation coefficient r is greater than 0.59 between EPMA and LA-ICP-MS data ( Fig. A1), both may be used in binary plots.

Statistical analysis
Rutile trace element composition was investigated using binary and ternary plots, as well as by partial least square-discriminant analysis (PLS-DA). PLS-DA is a supervised multivariate statistical method used to discriminate the mineral trace element composition (Makvandi et al., 2016b). Weights and loadings (qw*) are the components and highlight covariation between elements such that elements that plot in the same quadrant covary, whereas those that plot in the opposite quadrant show inverse covariations. The t scores enable visualization of the relationship between the classes defined by data labels and elements. Contributions show elements that have an impact on the classification. VIP-cumulative refers to the variable importance of projection that shows the importance of each element on the sample classification. VIP values greater than 0.8 are significant on the sample classification. Censored data were imputed using robCompositions from the R package, and elements with more than 40% data below the DL were excluded. Data were  (2002)  Bold font indicates deposits with more than 70 t Au, considered to be world-class gold after Goldfarb et al. (2005) Abbreviations: ab = albite, amp = amphibole, ap = apatite, apy = arsenopyrite, bt = biotite, cal = calcite, carb = carbonate, chl = chlorite, chr = chromite, cpx = clinopyroxene, cpy = chalcopyrite, ep = epidote, fsp = feldspar, gn = galena, grt = garnet, hem = hematite, ilm = ilmenite, mag = magnetite, mc = microcline, ms = muscovite, NA = not available; or = orthoclase, pl = plagioclase, po = pyrrhotite, py = pyrite, qz = quartz, rt = rutile, sch = scheelite, ser = sericite, sp = sphalerite, ttn = titanite, tur = tourmaline 1 Thin sections refers to the number of observed thin sections containing rutile 2 Controversial origin transformed using the log-centered log-ratio prior to PLS-DA (Makvandi et al., 2016a, b;Sciuba et al., 2020Sciuba et al., , 2021.

Chemical zoning
Zoning observed under the microscope and with SEM reflects chemical zoning (Figs. 1, 2). EPMA and LA-ICP-MS element maps show that zoned rutile is characterized by significant variations in Ta, W, Nb, and V contents and by subtle variations in Sn, Sb, Sc, and Fe contents such that bright zones under BSE imaging have high Ta, W, Nb, V, Sn, Sc, Sb, and Fe contents (Figs. 1, 2). Silicon, Mg, and Al weakly vary in some grains, whereas Sn and Mn contents remain constant. At Canadian Malartic, chemical zoning shows different variations in which the rutile core has high concentrations of W (>5,000 ppm), V (>600 ppm), Nb (>500 ppm), Sn (>100 ppm), Ta (>50 ppm), Sc (>30 ppm), and Sb (>20 ppm; Fig. 2). These element concentrations decrease toward the rim, although Nb and Ta concentrations are low in intermediate zones and high in rutile rims from Canadian Malartic (Fig. 2h, i).

TiO 2 polymorphs
Rutile from orogenic gold deposits has high concentrations of W relative to anatase and brookite (Fig. 5a). In the Ti-100×(Fe+Cr+V)-1000×W ternary diagram, rutile from orogenic gold deposits plots within or close to the field for ru-  (Fig. 5a). Some rutile grains from Rosebel, Sixteen-to-One, Lac Herbin, Giant, and Sunrise Dam plot in the anatase field. Others from Rosebel, Young Davidson, Harbour Light, Meliadine, and Beaufor plot in the brookite field. Our data plot in the field of rutile from orogenic gold mineralization from Clark and Williams-Jones (2004), with the exception of two grains from Rosebel and one grain from Sixteen-to-One (Figs. 5a, 6). In the ternary Al+Ti/ V-Fe+Cr+Sb+Mo+Sn-10×(W+Zr) diagram, rutile from orogenic gold deposits plots mostly in the rutile field with the exception of a few grains from Obuasi, Uti, and Sixteen-to-One deposits that plot in the anatase field at high Al+Ti/V values ( Fig. 5b; Plavsa et al., 2018). In the 100×Cr-Al-Fe ternary diagram as defined by Plavsa et al. (2018), our data plot mainly along the 100×Cr-Fe axis, in the rutile field (Fig. 5c). However, some data fall in the anatase field, mostly grains from Obuasi, Hollinger, Rosebel, and Hoyle Pond.
Rutile from deposits hosted in felsic rocks commonly has higher V concentrations (median 1,560 ppm V) than those from deposits hosted in intermediate rocks (median 1,009 ppm V; Fig. 8a; Table 3). Rutile from deposits hosted in sedimentary rocks has V composition similar to that of rutile from  Table 3). Rutile from deposits hosted in felsic rocks commonly has higher Nb and Ta concentrations (median 1,963 ppm Nb and 90 ppm Ta) than those from deposits hosted in mafic and mafic-ultramafic rocks that have low Nb and Ta concentrations (median 298 ppm Nb and 21 ppm Ta; 268 ppm Nb and 34 ppm Ta, respectively; Fig. 8b Table 3). Thus, the V vs. Nb diagram enables partial discrimination of rutile from deposits hosted in felsic, intermediate, and mafic and mafic-ultramafic rocks, whereas rutile from deposits hosted in sedimentary rocks overlaps the three other host rock compositions (Fig. 8a). Rutile from deposits hosted in felsic rocks tends to have low Cr concentrations (median 245 ppm) compared with those from deposits hosted in mafic and mafic-ultramafic rocks (median 808 and 5,405 ppm, respectively; Fig. 8c Fig. 8c; Table 3).
REE patterns for each deposit are presented in Fig. A2. Chondrite-normalized REEs most commonly display a flat pattern with no to small positive or negative Eu anomalies and an average rutile/chondrite ratio of about 10 (Fig. 10). A small negative Eu anomaly is characteristic of rutile from the intermediate rock-hosted Lac Herbin and Beaufor deposits (Figs. A2f, g, respectively) and is locally found in rutile from other intermediate rock-hosted deposits including Goldex and Tindals (Fig. A2e, p, respectively). Others, including deposits mostly hosted by mafic and sedimentary rocks such as Hollinger, Hoyle Pond, Obuasi, Essakane, Uti, and Raleigh, have a small positive Eu anomaly (Fig. A2b, c, k, l, n, and q).
Some trace elements also vary with the metamorphic facies of the country rocks. Vanadium concentrations in rutile from orogenic gold deposits increase with increasing metamorphic grade of the country rocks, regardless of the country rock composition. For instance, vanadium in rutile from deposits hosted by mafic rocks increases from lower to middle greenschist facies (median 320 ppm), to upper greenschist facies (median 1,814 ppm), and to lower to middle amphibolite facies (median 6,581 ppm; Fig. 8a;  8g, h) show that rutile from deposits hosted in upper greenschist to middle amphibolite facies country rocks has larger ratios compared with those from deposits hosted in lower to middle greenschist facies country rocks.

Multivariate statistical analysis of rutile trace element composition
PLS-DA is used with both EPMA and LA-ICP-MS data to highlight covariation between elements and enhance discrimination between geologic parameters including composition and metamorphic facies of the country rocks. Samples are labeled using the selected geologic parameter. Both EPMA and LA-ICP-MS data are investigated separately using PLS-DA as they contain different trace element sets and highlight different results. Data from the Kittilä (Finland) orogenic gold deposit (Auger, 2016) are combined with our data in PLS-DA on EPMA trace element composition of rutile as the reference contains the same set of trace elements as those in our study. Variation with the country rock compositions: PLS-DA of rutile divided by deposit country rock compositions shows that qw*1 is defined by large negative contributions of Nb, Ta, and Sn and large positive contributions of V, Sb, Si, and Cr. The qw*2 is defined by large negative contributions of Si, V, Sb, and Nb and large positive contributions of Cr, W, and Sn (Fig. 11a). Niobium, Ta, V, Si, and Ta have VIP-cumulative above 0.8 and, thus, are significant for the sample classification, whereas Sb, W, Fe, and Mg have VIP-cumulative below 0.8, which shows that they have minor importance on the classification (Fig. 11c). Rutile from deposits hosted in felsic rocks has dominantly negative t 1 and t2 (Fig. 11b) caused by positive contributions of Nb, Si, and Sb and negative contributions of Cr, V, and Mg (Fig. 11d). Rutile from deposits hosted in intermediate rocks has mostly negative t1 and variable t2 from positive contributions of Cr, Nb, and Ta and negative contributions of V and Si (Fig. 11e). Rutile from deposits hosted in intermediate rocks overlaps with those from deposits hosted in felsic rocks. Rutile from deposits hosted in mafic rocks mostly has positive t1 and variable t2 caused by positive contributions of Si, V, and Sb and negative contributions of Nb, Sn, and Ta (Fig. 11f). Rutile from deposits hosted in maficultramafic rocks has dominantly positive t1 scores and variable t2 caused by positive contributions of Cr, W, V, and Sb and negative contribution of Nb, Ta, Si, and Sn (Fig. 11g). Rutile from deposits hosted in mafic-ultramafic rocks overlaps with those from deposits hosted in mafic rocks. Rutile from deposits hosted in sedimentary rocks has dominantly negative to low positive t1 and variable t2 caused by positive contributions of Nb, Ta, and Sn and weak negative contributions of Sb, Si, V, and Cr (Fig. 11h).
Variation with the metamorphic facies of the country rock: PLS-DA of rutile divided by the metamorphic facies of the country rock shows that qw*2-qw*3 loadings most efficiently discriminate rutile (Fig. 12). The qw*1-qw*2 and qw*1-qw*3 loadings are listed in Fig. A3. The qw*2 is defined by positive contributions of Nb, Ta, W, and Sb and negative contributions of Fe, Cr, and V. The qw*3 is defined by positive contributions of Fe, Cr, V, and Nb and negative contributions of Mg, Si, Sb, and W (Fig. 12a). Fe, Cr, Sn, and Nb have VIP-cumulative above 0.8 and, thus, are significant for the sample classification, whereas Sb, Ta, Si, W, V, and Mg have VIP-cumulative below 0.8, which shows that they have minor importance on the classification (Fig. 12c). Rutile from deposits hosted in lower to middle greenschist facies country rocks has variable t2 and t3 about the origin, thus showing no classification (     Scott and Radford (2007), and Scott et al. (2011). Chondrite data are from McDonough and Sun (1995). 12b, d). Rutile from deposits hosted in upper greenschist facies country rocks has dominantly negative t2 and positive t3 caused by a high positive contribution of Cr and negative contributions of Nb, W, and Ta (Fig. 12e). Rutile from deposits hosted in lower to middle amphibolite facies country rocks has dominantly positive t2 and negative t3 caused by positive contribution of Sn, Ta, W, and Sb and negative contributions of Fe, Cr, and Si (Fig. 12f). Rutile from deposits hosted in lower to middle greenschist facies country rocks overlaps with both rutile hosted in upper greenschist and lower to middle amphibolite facies country rocks (Fig. 12b).
Variation with composition and metamorphic facies of the country rock: PLS-DA of LA-ICP-MS data for rutile from orogenic gold deposits hosted in country rocks with various compositions and metamorphosed to various facies is investigated using Nb, Ta, W, Zr, Hf, U, La, Ce, Yb, Y, Sc, Sn, Sb, Cr, Fe, V, Pb, Ba, Na, Al, and As (Fig. 13). The qw*1-qw*2 and qw*2-qw*3 are in Figure A3. The qw*1 is defined by negative contributions of Ba, La, Ce, Yb and Y and positive contributions of Sb, V, U, Zr, and Pb (Fig. 13a). The qw*3 is defined by large negative contributions of Cr, V, and Pb and large positive contributions of Sc, Fe, Hf, Zr, Sn, and La (Fig. 13a). Chromium, V, Sb, Sc, Nb, Ta, Hf, Zr, and Fe have VIP-cumulative higher than 1 and, thus, have high importance on the group classification. Lanthanum, Ce, and Pb have VIP-cumulative between 0.8 and 1 and, thus, have moderate importance on the group classification (Fig. 13c). Rutile from deposits hosted in intermediate rocks has mostly negative t1 and negative t3 caused by positive contributions of Ba, Y, and Ce and negative contributions of W, Zr, Hf, and V. Rutile from deposits hosted in mafic rocks has near 0 to positive t1 and t3 caused by positive contributions of W, Zr, Hf, U, Sn, Fe, Pb, and Au and negative contributions of Ce, Cr, La, Ce, Yb, Y, Ba, and As. Rutile from deposits hosted in mafic-ultramafic rocks has dominantly negative t1 and t3 caused by positive contributions of Na and As and negative contributions of Nb, U, and Pb. Rutile from deposits hosted in sedimentary rocks has variable t1 and t3 caused by low element contributions. Rutile from deposits hosted in lower to middle greenschist facies country rocks has dominantly negative t1 and variable t3 caused by positive contributions of La, Ce, Yb, Y, Ba, Na, and As and negative contributions of Zr, U, Sn, Sb, and Au. Rutile from deposits hosted in upper greenschist facies country rocks has positive to near 0 t1 caused by low element contributions of Cr, V, Pb, and Sc and positive to negative t3 caused by positive contributions of Nb, Ta, Sn, Na, and Au and negative contributions of Sb, Cr, V, Ce, La, and W. Rutile from deposits hosted in lower to middle amphibolite facies country rocks has dominantly positive t1 and negative t3 caused by positive contributions of Zr, Hf, U, Sc, Sn, Cr, and V and negative contributions La, Ce, Yb, Y, Ba, and Na (Fig. 13b).
Variation with the age of mineralization: PLS-DA of rutile classified by age of mineralization shows that qw*1 is defined by positive contributions of V, Si, and Mg and negative contributions of Ta, Nb, and Sn. The qw*2 is defined by positive contributions of V, Si, and Nb and negative contributions of Sb, Sn, Cr, and W (Fig. 14a). Rutile from deposits formed during the Archean has variable t1 and t2 caused by a negative contribution of V (Fig. 14b, c). Rutile from deposits formed during Proterozoic has dominantly positive t1 and variable t2 (Fig. 14b) caused by positive contributions of V, Si, and Mg and negative contributions of Nb, Ta, and Sn (Fig. 14d). Rutile from deposits formed during Phanerozoic has mostly negative t1 and variable t2 (Fig. 14b) caused by positive contributions of Ta, Nb, and Sn and negative contributions of Sb, Mg, Si, and (Fig. 14e). Rutile from Archean deposits overlaps with  Scott and Radford (2007), and Scott et al. (2011). Chondrite data are from McDonough and Sun (1995).   both rutile from deposits formed during the Proterozoic and Phanerozoic (Fig. 14b).

Discussion
Classifying TiO2 polymorphs in orogenic gold deposits Plavsa et al. (2018) highlighted the importance of characterizing TiO2 polymorphs in order to use them in indicator mineral surveys for mineral exploration. In the ternary diagrams Ti vs. 100×(Fe+Cr+V) vs. 1000×W, Al+Ti/V vs. Fe+Cr+Sb+Mo+Sn vs. 10×(W+Zr), and 100×Cr vs. Al vs. Fe (Fig. 5), our data from orogenic gold deposits plot most commonly in the rutile field, which suggests that rutile is the dominant TiO2 polymorph in orogenic gold mineralization. Additionally, our data plot mostly out of the field for unaltered country rocks in orogenic gold deposits, which also coincides with the anatase and brookite field in the Ti vs. 100×(Fe+Cr+V) vs. 1000×W ternary diagram (Figs. 5,6). This result strengthens the geochemical characterization of rutile in orogenic gold mineralization proposed by Clark and Williams-Jones (2004). However, Raman spectroscopy is necessary to identify TiO2 polymorphs.

Rutile grain size
Our observations show that rutile from orogenic gold deposits is commonly less than 250 µm in size and contains mineral inclusions (Table 2). In indicator mineral surveys, the 0.25 to 0.5 and 0.5 to 2.0 mm size fractions are the most useful because these size grains are large enough for identification under a binocular microscope and for chemical analysis (Thorleifson, 2010). Thus, rutile grains smaller than 0.25 mm will not be recovered in the size fractions commonly used. Scott (2005) observed that in Cu-Au porphyry systems, rutile proximal to mineralization is coarser and more likely zoned, whereas those distal from the mineralization are smaller in size. In contrast, Williams and Cesbron (1977) reported that rutile size slightly increases outward from porphyry Cu deposits. Banfield and Veblen (1991) noted that rutile grain size increases from low P-T chlorite and chloritoid zones to higher P-T garnet and staurolite zones. Our study only includes rutile from proximal mineralization zones and, thus, does not allow us to discuss rutile grain size relative to distance from mineralization. The investigated rutile does not show any systematic variation in grain size with the metamorphic facies of the country rocks. Additionally, rutile grain size less than 250 µm and textures associated with orogenic gold deposits may limit LA-ICP-MS analysis and trace element quantification.

Assimilation of country rock signature
Rutile trace element composition depends on multiple factors, such as pressure, temperature, oxygen fugacity, and fluid composition, as well as element partitioning with coprecipitating minerals (Brenan and Watson, 1991;Brenan et al., 1994Brenan et al., , 1995Adam et al., 1997;Ayers et al., 1997;Ayers, 1998;Stalder et al., 1998;Klemme et al., 2005;Liu et al., 2014;Mallmann et al., 2014). Fluids related to orogenic gold mineralization are aqueous-carbonic fluids that contain 5 to 20 mol % CO2 that typically have near-neutral pH, salinity of 3 to 7 wt % NaCl equivalent, and Na > K > Ca, Mg concentrations (Goldfarb and Groves, 2015). They have significant concentrations of CH4 and/or N2 and 0.01 to 0.36 mol % H2S (Goldfarb and Groves, 2015). Thus, the relatively uniform composition of the Au-bearing fluid is unlikely to explain the variation in trace element composition of rutile from orogenic gold deposits. Niobium, Ta, and Cr contents in rutile vary with the host rock composition such that those elements are used to discriminate source lithologies in sediment provenance studies (Zack et al., 2004b;Triebold et al., 2007;Meinhold et al., 2008). Our results also show that Nb, Ta, V, and Cr in rutile from orogenic gold deposits vary with the country rock composition (Figs. 8,9,11). Nb and Ta concentrations mimic directly the median Nb and Ta contents of the country rocks. Similar to rutile from orogenic gold deposits, felsic rocks have high concentrations in Nb and Ta (16 ppm Nb and 1.3 ppm Ta) compared with intermediate rocks that have intermediate concentrations in Nb and Ta (8 ppm Nb and 0.5 ppm Ta) and mafic and ultramafic rocks that have low concentrations in Nb and Ta (5 ppm Nb and 0.25 ppm Ta; Condie, 1993). Similar to Nb and Ta concentrations in rutile from orogenic gold deposits, sedimentary rocks have median Nb and Ta concentrations (10 ppm Nb and 0.8 ppm Ta) comparable to that of intermediate rocks (Condie, 1993). Chromium concentrations in rutile (Fig. 8c) also mimic the Cr concentrations in the country rocks such that the Cr concentration progressively increases from felsic to mafic to ultramafic rocks (felsic, 9 ppm;intermediate, 45 ppm;mafic and ultramafic, 170 ppm;Condie, 1993). However, binary diagrams, such as Cr vs. Nb (Fig. 8c), do not discriminate rutile from deposits hosted in various country rock compositions. In contrast, PLS-DA shows that Cr is important to classify rutile according to the country rock composition (Fig. 11). Several studies reviewed trace element partition coefficients between rutile/melt and rutile/fluid (Brenan and Watson, 1991;Brenan et al., 1994Brenan et al., , 1995Adam et al., 1997;Ayers, 1998;Ayers et al., 1997;Stalder et al., 1998;Foley et al., 2000;Klemme et al., 2005). Rutile/melt partition coefficients DNb and DTa are in the range of 100 to 500, DV is in the range of 50 to 120, whereas DZr and DHf are around 5 (Foley et al., 2000;Klemme et al., 2005). DCr ranges from 4.8 to >108 (Klemme et al., 2005). Other trace elements including U have rutile/melt partition coefficients less than 1, whereas others, such as Rb, Sr, REE, Y, Ba, Th, Pb, and Mn, have partition coefficients less than 0.1 (Foley et al., 2000;Klemme et al., 2005). Other studies showed that the rutile/fluid partition coefficients are similar to those for rutile/melt for Nb and Ta, whereas DZr and DHf are greater (Brenan and Watson, 1991;Brenan et al., 1994Brenan et al., , 1995Adam et al., 1997;Ayers, 1998;Ayers et al., 1997;Stalder et al., 1998). During orogenic gold mineralization, hydrothermal fluids interact with the country rock packages and evolve in trace element composition (Garofalo et al., 2014). Thus, our results are consistent with rutile from orogenic gold deposits that inherit the Nb, Ta, V, and Cr signature from the country rock from which they are most likely sourced. Sciuba et al. (2020) showed that in orogenic gold deposits, the Eu anomaly in scheelite reflects the country rock composition so that scheelite from felsic and intermediate hosted deposits commonly has a negative Eu anomaly, whereas  (Alderton et al., 1980). Europium concentration in rutile is low (median 0.06 ppm; Fig. 7), and Eu 2+ substitution is probably minor. However, the following equation may be considered: where Y is a pentavalent cation, such as Nb 5+ , Ta 5+ , or Sb 5+ . Our study shows that rutile trace element composition changes between middle and upper greenschist facies such that rutile in deposits formed at lower to middle greenschist facies has high REE, Y, Ca, Th, and Ba concentrations and low Zr, Sc, U, and V concentrations compared with deposits formed at upper greenschist to middle amphibolite facies that have low REE,Y,Ca,Th,and Ba concentrations and high Zr,Sc,U,and V concentrations (Figs. 8,9,13). Zirconium incorporation in rutile is known to be temperature dependent (Zack et al., 2004a;Watson et al., 2006;Tomkins et al., 2007).
There also appears to be temperature control on the Mo and U incorporation in rutile such that Mo and U concentrations increase with increasing temperature (Rabbia et al., 2009;Agangi et al., 2020). Thus, increasing Zr and U concentrations in rutile from orogenic gold deposits with increasing metamorphic facies may reflect the temperature control on Zr and U incorporation in rutile composition. At upper greenschist facies, Ca-amphiboles, such as tremolite and actinolite, appear in the mineral assemblage associated with gold mineralization, and Ca-garnets, such as andradite and grossularite, and clinopyroxene, appear at middle amphibolite facies (Eilu et al., 1999). Actinolite is common in the low amphibolite facies country rocks of the Roberto deposit (Fontaine et al., 2015; Table 2), and andradite is commonly associated with the gold mineralization at Big Bell, which is hosted in middle amphibolite facies country rocks (Mueller et al., 1996; Table  2). Calcium is a major element in tremolite, actinolite, andradite and grossular garnet, and clinopyroxene, which perhaps explains the low-Ca rutile in deposits formed from upper greenschist to middle amphibolite facies. The Red Lake deposit is hosted in upper greenschist facies country rocks, in Fig. 13. Partial least square-discriminant analysis of laser ablation-inductively coupled plasma-mass spectrometry data for rutile from orogenic gold deposits hosted in country rocks with various compositions and metamorphosed to various facies. (a) qw*1-qw*3 loadings, (b) t1-t3 scores, and (c) VIP-cumulative.   (Table 2). Ca-amphiboles and/or garnets are described in the mineralization assemblage of those three deposits (McCormick and Hanna, 1990;Mishra et al., 2005;Cadieux et al., 2006), but none were observed in our analyzed samples (Table 2), which perhaps explains the higher Ca concentrations compared with other deposits hosted in upper greenschist to middle amphibolite facies country rocks. Amphibole/fluid and garnet/fluid partition coefficients are higher for REE and Y than rutile/fluid partition coefficients (Roll-inson, 1993;Foley et al., 2000;Klemme et al., 2005). Thus, the low REE and Y contents in rutile from deposits hosted in country rocks metamorphosed from upper greenschist to middle amphibolite facies are possibly caused by the preferential incorporation of REE and Y in amphibole and garnet coprecipitating with rutile. Rutile compositions in orogenic gold deposits from other studies (Clark and Williams-Jones, 2004;Scott and Radford, 2007;Dostal et al., 2009;Scott et al., 2011;Decker, 2012;Martin, 2012;Auger, 2016;Pochon et al., 2017;Agangi et al., Fig. 14. Partial least square-discriminant analysis of electron probe microanalysis data for rutile in orogenic gold deposits of various ages. (a) qw*1-qw*2 loadings and (b) t1-t2 scores and score contributions for each group including rutile from deposits formed at (c) Archean, (d) Proterozoic, and (e) Phanerozoic. Data from this study and Auger (2016).

Comparison to rutile from other deposit types and geologic environments
The trace element composition of rutile from orogenic gold deposits is compared with that from other deposit types and geologic settings, including deposits from hydrothermal, Hemlo deposit (Urban et al. 1992, Clark andWilliams-Jones 2004) Other gold deposits (Urban et al. 1992, Rice et al. 1998, Clark and Williams-Jones 2004, Cabral et al. 2013 Pegmatite (Černý et al. 1999. Černý et al. 2007) Granite (

Hydrothermal
VMS (Clark and Williams-Jones 2004) U-unconformity related (Adlakha 2016) Paleoplacer (Clark and Williams-Jones 2004) Mn concentration (Smith and Perseil 1997) b d c a e f magmatic-hydrothermal, magmatic, and sedimentary environments, as well as metamorphic environments in Figures  15 and 16. Manganese and Fe (Fig. 15f) and Nb and Ta (Fig.  16c) covary in rutile from all deposit types and geologic environments. Rutile from other types of gold deposits (Hemlo) and other hydrothermal deposits including the unconformityrelated U McArthur River deposit (Adlakha, 2016) commonly has higher Zr and similar Mn, Fe, V, Al, Cr, Sn, Nb, Ta, and Sb contents. Rutile from McArthur River has slightly higher V and Nb contents (Adlakha, 2016), whereas rutile from VMS deposits has higher Sn concentration (Clark and Williams-Jones, 2004;Porter et al., 2020), and rutile from the Hemlo gold deposit has higher Sb and V contents than those from orogenic gold deposits. Rutile from magmatic-hydrothermal environments commonly has higher Sn, Nb, Ta, U, and Fe; lower Mn, V, and Sb; and similar Mn, V, Sn, W, Cr, and Al concentrations compared with rutile from orogenic gold deposits (Figs. 15, 16;Agangi et al., 2019). Rutile from magmatic environments, excluding pegmatites, has lower V, Sb, and Sn and similar Zr, Cr, U, Al, Nb, Ta, Mn, and Fe contents compared with rutile from orogenic gold deposits. Rutile from pegmatites has higher Sn, Nb, Ta, Fe, and Mn compared with rutile from orogenic gold deposits. Rutile from sedimentary deposit types, including paleoplacer and Mn mineralization, has higher Fe and Mn (Fig. 15f) and Sn, V, and Nb contents similar to rutile from orogenic gold deposits. Rutile from metamorphic environments commonly has lower Mn, Sb, and Sn and similar V, Nb, Ta, U, Al, Cr, Fe, and Zr concentrations compared with rutile from orogenic gold deposits (Figs. 15,16). Rutile from orogenic gold deposits can be partially discriminated from those from other deposit types and environments based on the Mn, V, Sn, Sb, and W concentrations (Fig. 15a, b, g, h). In the V vs. Sb diagram, rutile from metamorphic-hydrothermal and hydrothermal environments, including orogenic gold deposits, plots at high V and Sb, whereas rutile from magmatichydrothermal and magmatic environments plots mostly at low V and Sb (Fig. 15g). Vanadium data for rutile in pegmatites are lacking, but the range of Sb content in pegmatites (400 to 9,000 ppm; Černý et al., 1999, 2007) is within the Sb range for orogenic gold rutile (10 to 10,000 ppm; Fig. 7). In the Nb/V vs. W diagram (Fig. 15h), rutile from orogenic gold deposits plots together with those from metamorphic-hydrothermal, hydrothermal, and metamorphic environments at low Nb/V and high W, whereas those from magmatic-hydrothermal and magmatic environments commonly plot at high Nb/V and low to moderate W compared with rutile from orogenic gold deposits. In the Nb/Sb vs. Sn/V diagram (Fig. 15i), rutile from orogenic gold deposits has low Nb/Sb and Sn/V ratios, which discriminates it from rutile from other environments that commonly have high Nb/Sb and Sn/V ratios. Orogenic gold deposits are characterized by consistent geochemical anomalies in Ag, As, Au, B, Bi, Sb, Te, and W (Kerrich, 1983;Phillips and Groves, 1983;Groves et al., 1998;Goldfarb et al., 2005;Goldfarb and Groves, 2015), mostly reflecting the ore fluid composition that is dominated by sulfur, fluoro, and hydroxo complexes associated with low fluid salinity (Goldfarb and Groves, 2015). Similarly, rutile in orogenic gold deposits also has elevated Sb and W concentrations, as Agangi et al. (2019) and Porter et al. (2020) previously noted, which we suggest may indicate that Sb and W contents are controlled by the fluid composition. This agrees with Pochon et al. (2017), who suggested that Sb and W were transported by hydrothermal fluids and incorporated into rutile from the Sb-Au orogenic mineralization of the Armorican Massif (France). Thus, the chemical composition of rutile has characteristics that allow recognition from an orogenic gold deposit source, which can be applied to detect the occurrence of an eroded orogenic gold deposit in overburden sediments in indicator mineral surveys.

Conclusions
Orogenic gold deposit rutile has a wide range of grain size, texture, and chemical zoning. Rutile from orogenic gold deposits is often <250 µm in size, which may limit its recovery from the most commonly used size fractions in indicator mineral surveys and may require that smaller grain-size fractions be sampled for orogenic gold deposit targeting. Mineral inclusions in rutile are common; associated with small grain size, these inclusions may limit the trace element analyses by LA-ICP-MS. Our results show that rutile is the dominant TiO2 polymorph in orogenic gold mineralization; anatase and brookite are uncommonly found. In orogenic gold rutile, V, Nb, Ta, and Cr vary with the composition of the country rocks such that Nb and Ta tend to increase, and Cr tends to decrease from mafic to intermediate to felsic igneous rocks. Rutile from deposits hosted in sedimentary rocks commonly has V, Nb, and Ta compositions similar to those from deposits hosted in intermediate rocks. Rutile from deposits hosted in felsic rocks has high V concentrations compared with those from deposits hosted in intermediate rocks. Scandium, REE, Y, Ca, Ba, Th, and V vary with the metamorphic facies of the country rocks, whereas Zr and U incorporation may be temperature dependent. Vanadium content varies with both the composition and metamorphic facies of the country rock.
PLS-DA allows discrimination of rutile from different country rock compositions and metamorphic facies. The results indicate that some elements, such as Nb, Ta, and Cr, are probably derived from the regional country rock packages, whereas others, such as W and Sb, are transported by the Au-bearing hydrothermal fluids. Rutile from orogenic gold deposits has a distinctive Mn, V, Sn, Sb, and W composition compared with rutile from other deposit types and geologic settings. Binary diagrams, including V vs. Sb and Nb/V vs. Sn/V, discriminate rutile from orogenic gold deposits from those from magmatichydrothermal and magmatic deposit types. Other binary diagrams, such as Nb/V vs. W, discriminate orogenic gold deposit rutile from hydrothermal and metamorphic-hydrothermal environments. Our results show that the trace element composition of rutile can be used in indicator mineral surveys to orient exploration strategy.