Volcanic-Derived Placers as a Potential Resource of Rare Earth Elements: The Aksu Diamas Case Study, Turkey

: Rare earth elements (REE) are essential raw materials used in modern technology. Current production of REE is dominated by hard-rock mining, particularly in China, which typically requires high energy input. In order to expand the resource base of the REE, it is important to determine what alternative sources exist. REE placers have been known for many years, and require less energy than mining of hard rock, but the REE ore minerals are typically derived from eroded granitic rocks and are commonly radioactive. Other types of REE placers, such as those derived from volcanic activity, are rare. The Aksu Diamas heavy mineral placer in Turkey has been assessed for potential REE extraction as a by-product of magnetite production, but its genesis was not previously well understood. REE at Aksu Diamas are hosted in an array of mineral phases, including apatite, chevkinite group minerals (CGM), monazite, allanite and britholite, which are concentrated in lenses and channels in unconsolidated Quaternary sands. Fingerprinting of pyroxene, CGM, magnetite and zircon have identiﬁed the source of the placer as the nearby Gölcük alkaline volcanic complex, which has a history of eruption throughout the Plio-Quaternary. Heavy minerals were eroded from tephra and reworked into basinal sediments. This type of deposit may represent a potential resource of REE in other areas of alkaline volcanism. The chevkinite group minerals (CGMs) exhibit a “problematic spectral region” between 4.4–4.9 keV with overlapping X-ray energies of Ce, La, Ba and Ti (Ce L α 1 4.84 keV, La L α 1 4.65 keV, Ti K α 1 4.51 keV, Ti K β 4.931 keV and Ba L α 1 4.465 keV). WDS, with superior spectral resolution to EDS, was used to resolve these overlaps. The WDS analysis was conducted using the same FEI 600 Quanta SEM with a Wave 500 spectrometer, running with Oxford Instruments INCA (Version 4) software. The SEM was operated at 20 kV accelerating voltage, 10 nA beam current. Only one WDS spectrometer was available and to minimize the risk of beam drift, the analysis was limited to nine elements per analytical spot only. The counting times varied per element between 20/10 s (peak/background) and 40/20 s for major and trace elements, respectively, resulting in a total acquisition time of approximately 30 min per analysis. The following reference materials and WDS crystals were used for calibration of the selected nine element X-ray lines: (LaL α )–LaB & LIF; 2 & LIF; & LIF; (NdL & LIF; )–SmAlO 3 & LIF; (TiK α , SrL α 3 & PET; (BaL α )–BaSO 4 & PET; and (VK α )–V metal & LIF. The calibration was performed every 24 h. This protocol was used for two representative samples, one from the Gölcük pyroclastic rocks and one from the Aksu Diamas pit with approximately ten EDS-WDS analyses per sample. All other samples of chevkinite group minerals were analyzed by EDS only. A detailed evaluation of the EDS and WDS results taken from the same analytical spot gave us conﬁdence to use the EDS data for plotting of element ratios. One common disadvantage of the EDS analysis is the


Introduction
Rare earth elements (REE) are considered as critical metals, due to their widespread use in many "green" technologies. Primary global resources of the REE [1] are dominantly hosted in carbonatite [2] and alkaline igneous intrusions [3]. Worldwide, the main deposits currently producing REE are carbonatite-hosted, including Bayan Obo in China [4,5], Maoniuping in China [6,7] and Mount Weld in Australia [2]. From the 1980s, China became the world's largest producer of REE, and by 2010 about 95% of the global supply originated from China [8]. However, in recent years, the supply base for rare earth elements is noticeably broadening [9]. Between the years 2011 and 2015, a five-fold increase in delineated global REE resources was identified-with almost 40% of these in Europe (Greenland, Sweden, Germany and Turkey) [10]. Although the majority of these discoveries are in carbonatites and alkaline rocks, there is an ongoing emphasis on research into alternative sources and the genesis of other types of REE deposits [11,12]. Examples of alternative REE sources include ion adsorption clays [13], alumina production waste [14], off-shore heavy mineral sands [15], deep ocean

Geological Setting
The Eastern Mediterranean has a complicated and prolonged geological history. The geology of the region records the closure of the Tethys ocean, which extended from modern day Greece to the Middle East [34]. The Mesozoic to Early-Cenozoic Neotethys ocean was located in what is now southern Turkey [35]. The southern margin of the Neotethys was the northern margin of the Gondwanan supercontinent and comprised of an array of smaller crustal fragments or microcontinents. Northward subduction beneath the Eurasian plate resulted in continent-continent collision during the middle-late Miocene, forming the Tauride Mountains [35][36][37]. Subsequently, in the area of interest, the development of an extensional regime led to the formation of sedimentary basins that overlie the Tauride basement [38].
The widespread alkaline magmatism in the Circum-Mediterranean Anorogenic Cenozoic Igneous (CiMACI) province, which extends from the Canary Islands to the Middle East, is linked to extensional tectonics, subsequent to continental collision [39]. The Eastern Mediterranean region is considered to be a back-arc setting, with active northwards subduction in the South Aegean and post-collisional alkaline magmatism in Western Anatolia [40]. Modelling of teleseismic P-wave traveltime tomography data from the Mediterranean region has identified that there are extensive slab interruptions in the form of tears [41], allowing interaction of magmas derived from the asthenosphere and the enriched lithospheric mantle [40,[42][43][44][45].
In south west Turkey, the CiMACI province is expressed by potassic to ultra potassic volcanism in the Isparta-Gölcük region ( Figure 1A). The bedrock geology of this area comprises of authochtonous Mesozoic platform carbonates from multiple palaeogeographic domains, and allochtonous nappes that are highly deformed [46] ( Figure 1B). These nappes include platform and basinal sedimentary sequences, together with ophiolitic material, that were thrust onto the platform in Late Cretaceous-Palaeocene times. The Isparta-Gölcük region lies within the Lycian nappe [46].  [47], Poisson et al. [46] and Nielsen et al. [48] and references therein. Base map data SRTM 1 Arc-Second Global available from the U.S. Geological Survey Earth Explorer. (C) Geological map of the Gölcük igneous province from Platevoet et al. [49].

Gölcük Volcano
The Gölcük volcano lies to the north of the Western Taurus Mountains that are formed of Mesozoic limestone and ophiolitic lithologies of the Lycian Nappe [50]. The marine successions comprise shelf and slope deposits and are several hundred metres thick in some areas [48]. Mt. Akdağ is the highest point at 2240 m. The range in this area has a lateral extent >20 km and a width of approximately 4 km at the widest point. These mountains form a drainage barrier between the volcano located on the north side of the range and the Quaternary basin to the south that hosts the Aksu Diamas deposit (Figure 2).  [47], Poisson et al. [46] and Nielsen et al. [48] and references therein. Base map data SRTM 1 Arc-Second Global available from the U.S. Geological Survey Earth Explorer. (C) Geological map of the Gölcük igneous province from Platevoet et al. [49].

Gölcük Volcano
The Gölcük volcano lies to the north of the Western Taurus Mountains that are formed of Mesozoic limestone and ophiolitic lithologies of the Lycian Nappe [50]. The marine successions comprise shelf and slope deposits and are several hundred metres thick in some areas [48]. Mt. Akdag is the highest point at 2240 m. The range in this area has a lateral extent >20 km and a width of approximately 4 km at the widest point. These mountains form a drainage barrier between the volcano located on the north side of the range and the Quaternary basin to the south that hosts the Aksu Diamas deposit (Figure 2). The dormant Gölcük volcano [51] is located close to the city of Isparta ( Figure 1A), and is a part of the potassic volcanism of the CiMACI province. The volcano lies in the Isparta Angle, the intersection of the NE-trending Burdur Fault and the NW-treading Aksehir-Simav Fault [52]. The volcanic edifice is approximately 2.5 km across and is filled with a maar lake, surrounded by a topographic rim comprising mainly pyroclastic rocks with minor lavas and dykes ( Figure 1C) [53,54]. This volcanic edifice developed from multiple volcanic eruptions and its products cover an area of up to 100 km 2 [49,51,53,[55][56][57][58]. A recent geophysical survey over the volcano has identified the presence of a trachytic dome beneath the peak of Pilav Tepe ( Figure 1C) that is >850 m deep and 1.2 km in horizontal extent [59]. The Gölcük magmas are generally more potassic than other post-collisional volcanic provinces in the vicinity of Tethyan suture zones [51,60,61].
Multiple major episodes of volcanic activity, most recently phreatoplinian type eruptions, have been recognised at Gölcük and are well described in overviews by Kumral et al. [54], Platevoet et al. [58] and Kalyoncuoglu [62]. The oldest igneous activity at Gölcük comprises potassic lavas and domes was dated at 4.7 Ma and 2.77 ± 0.06 Ma [49,55,56]. The location of the igneous activity at Gölcük is controlled by the intersection of NE-and NW-trending faults that were active during the Quaternary [52,63].
The current Gölcük volcanic edifice is the final result of long-lived volcanic activity which began at around 2.77 ± 0.06 Ma [49]. Three eruptive cycles, termed eruptive Cycles I, II, and III, have been recognised by Platevoet et al. [49].
Cycle I comprises of a ca. 200 m-thick pile of pyroclastic flow deposits [49] formed between 2.18 Ma and 0.148 Ma in a series of major explosive events [58]. Cycle II consists of volumetrically minor tephriphonolitic lava flow-domes and dykes, which were emplaced not long after the end of Cycle I, between 0.115 Ma and 0.062 Ma [49,55,56]. Cycle III created a significant tuff ring, centered on the maar crater, which covered previous eruptive material and extends over the flanks of the volcano (Figure 3a-c). The tuff cone was developed between 0.0723 Ma, and 0.024 Ma (age of the last trachyte dome) [49]. Dates for trachytic intra-caldera lava domes (Pilav Tepe and Küçükçeşme Tepe), which are the most recent igneous activity at Gölcük, indicate that the minimum age for recent volcanic activity is ~0.013 Ma [51]. The dormant Gölcük volcano [51] is located close to the city of Isparta ( Figure 1A), and is a part of the potassic volcanism of the CiMACI province. The volcano lies in the Isparta Angle, the intersection of the NE-trending Burdur Fault and the NW-treading Aksehir-Simav Fault [52]. The volcanic edifice is approximately 2.5 km across and is filled with a maar lake, surrounded by a topographic rim comprising mainly pyroclastic rocks with minor lavas and dykes ( Figure 1C) [53,54]. This volcanic edifice developed from multiple volcanic eruptions and its products cover an area of up to 100 km 2 [49,51,53,[55][56][57][58]. A recent geophysical survey over the volcano has identified the presence of a trachytic dome beneath the peak of Pilav Tepe ( Figure 1C) that is >850 m deep and 1.2 km in horizontal extent [59]. The Gölcük magmas are generally more potassic than other post-collisional volcanic provinces in the vicinity of Tethyan suture zones [51,60,61].
Multiple major episodes of volcanic activity, most recently phreatoplinian type eruptions, have been recognised at Gölcük and are well described in overviews by Kumral et al. [54], Platevoet et al. [58] and Kalyoncuoglu [62]. The oldest igneous activity at Gölcük comprises potassic lavas and domes was dated at 4.7 Ma and 2.77 ± 0.06 Ma [49,55,56]. The location of the igneous activity at Gölcük is controlled by the intersection of NE-and NW-trending faults that were active during the Quaternary [52,63].
The current Gölcük volcanic edifice is the final result of long-lived volcanic activity which began at around 2.77 ± 0.06 Ma [49]. Three eruptive cycles, termed eruptive Cycles I, II, and III, have been recognised by Platevoet et al. [49].
Cycle I comprises of a ca. 200 m-thick pile of pyroclastic flow deposits [49] formed between 2.18 Ma and 0.148 Ma in a series of major explosive events [58]. Cycle II consists of volumetrically minor tephriphonolitic lava flow-domes and dykes, which were emplaced not long after the end of Cycle I, between 0.115 Ma and 0.062 Ma [49,55,56]. Cycle III created a significant tuff ring, centered on the maar crater, which covered previous eruptive material and extends over the flanks of the volcano (Figure 3a-c). The tuff cone was developed between 0.0723 Ma, and 0.024 Ma (age of the last trachyte dome) [49]. Dates for trachytic intra-caldera lava domes (Pilav Tepe and Küçükçeşme Tepe), which are the most recent igneous activity at Gölcük, indicate that the minimum age for recent volcanic activity is~0.013 Ma [51].  (after [58]). (b) Finer-grained unconsolidated tuff layers of Eruptive Cycle III (after [58]). (c) Graded layering in unconsolidated tuff with pumice-rich layers of Eruptive Cycle III (after [58] The petrology of the Gölcük pyroclastic rocks has been described in detail by Platevoet et al. [58]. The dominant material in both Cycles I and III is trachyandesitic pumice, with phenocrysts of green clinopyroxene, brown amphibole, magnetite, titanite, apatite, mica and a range of accessory minerals. Sanidine is abundant in Cycle III pumices. The pyroclastic rocks are strongly alkaline (ultrapotassic) in composition and are enriched in Ba, Sr, Th, U, Rb and light rare earth elements (LREE) [55,58].

Aksu Diamas
Approximately 25 km southwest of the Gölcük volcano are Quaternary basins of unconsolidated alluvium and possible lacustrine deposits ( Figures 1B and 2). These basins are surrounded by Cretaceous Bey Dağlari limestone. The Aksu Diamas deposit is located at the margins of such a basin, adjacent to moderately rugged limestone slopes. There has been very little previous research in this area and the following description is based on our fieldwork at the deposit, carried out in 2014.
The uppermost ca. 8 m of the weakly consolidated sedimentary succession is exposed in an exploration pit at Aksu Diamas, which is on the edge of the basin (Figure 3d). The upper part of the exposed sequence (4 m) comprises of a fine-grained, laminated, clay-rich muddy sand that contains organic material (Figure 3d). This upper part of the sequence is interpreted as having formed in an overbank setting. The lower part of the exposed succession comprises of a 5-m thick sequence of overlapping, lens-shaped gravel bodies that range in size from 90 cm to 4.5 m wide and 15-60 cm  [58]). (b) Finer-grained unconsolidated tuff layers of Eruptive Cycle III (after [58]). (c) Graded layering in unconsolidated tuff with pumice-rich layers of Eruptive Cycle III (after [58]). (d) Exposure of the Aksu Diamas exploration pit wall, with outlines of gravel lenses highlighted with black dashed lines.
The petrology of the Gölcük pyroclastic rocks has been described in detail by Platevoet et al. [58]. The dominant material in both Cycles I and III is trachyandesitic pumice, with phenocrysts of green clinopyroxene, brown amphibole, magnetite, titanite, apatite, mica and a range of accessory minerals. Sanidine is abundant in Cycle III pumices. The pyroclastic rocks are strongly alkaline (ultrapotassic) in composition and are enriched in Ba, Sr, Th, U, Rb and light rare earth elements (LREE) [55,58].

Aksu Diamas
Approximately 25 km southwest of the Gölcük volcano are Quaternary basins of unconsolidated alluvium and possible lacustrine deposits ( Figures 1B and 2). These basins are surrounded by Cretaceous Bey Daglari limestone. The Aksu Diamas deposit is located at the margins of such a basin, adjacent to moderately rugged limestone slopes. There has been very little previous research in this area and the following description is based on our fieldwork at the deposit, carried out in 2014.
The uppermost ca. 8 m of the weakly consolidated sedimentary succession is exposed in an exploration pit at Aksu Diamas, which is on the edge of the basin (Figure 3d). The upper part of the exposed sequence (4 m) comprises of a fine-grained, laminated, clay-rich muddy sand that contains organic material (Figure 3d). This upper part of the sequence is interpreted as having formed in an overbank setting. The lower part of the exposed succession comprises of a 5-m thick sequence of overlapping, lens-shaped gravel bodies that range in size from 90 cm to 4.5 m wide and 15-60 cm thick. The gravel is poorly sorted, varying from matrix-supported to clast-supported, and contains angular clasts that are predominantly limestone with subordinate volcanic material. The clasts range in size Minerals 2019, 9, 208 8 of 31 from <0.5-5 cm with occasional boulders up to 40 cm. The matrix is a light brown-coloured, fine to medium-grained, sub-angular, unconsolidated sand. The limestone clasts are likely locally sourced from the Cretaceous bedrock. The gravel bodies are interpreted as a series of high-energy channelised mass flow deposits, and the surrounding fine-grained sand is interpreted as a low-energy overbank setting. The sedimentary relationships at the base of the pit are obscured by washed and redistributed sediments, resulting from previous exploration work (Figure 3d). Heavy minerals are concentrated in the matrix of the lenses and are also disseminated in the loosely consolidated fine-grained sand that hosts the gravel lenses. These concentrations represent the placer deposit that was the subject of exploration by AMR Mineral and Metal, Inc. Samples were collected to investigate the heavy mineral composition of the Aksu Diamas placer and compare it with the most recent pyroclastic deposits of Gölcük.

Materials
Three samples of pyroclastic material (Table 2) were collected from sub-layers within the accessible exposures of Eruptive Cycle III (after Platevoet et al. [58]) (Figure 3a-c) at the Gölcük alkaline complex ( Figure 1). Samples were collected from cleaned faces over an area of approximately 50 × 30 cm. Each of the localities comprises of unconsolidated tuffs, with matrix-supported lithic clasts in a medium-to fine-grained matrix. The clasts are angular, predominantly pumice, with minor trachytic clasts. The deposit as a whole is bedded and generally poorly sorted, with some clast-rich layers. Clast size varies between the three sites, with the modal clast size being smaller (ca. 2 cm) in deposits on the inner wall of the crater, and larger (ca. 10 cm) on the outer margin.
Three bulk samples of unconsolidated, poorly-sorted gravel hosting heavy minerals (Aksu Diamas placers, Table 2) were collected from multiple channels in the lower levels of the test pit at Aksu Diamas ( Figure 3d). The faces at the sampling locations were scraped clean before sampling to avoid any recent weathering features. Representative samples from multiple channels were taken. A sample of heavy mineral concentrate (MPLV 054) was obtained from the pilot plant facility adjacent to the exploration pit.

Methods
Initially, the samples were dried at 40 • C for a minimum of 12 h and then split into two at the sample preparation facility at the British Geological Survey (BGS), Keyworth, UK, with half of each sample retained for archiving at the BGS. Samples were sieved to collect the 125-250 µm fraction, removing both the larger lithic clasts and the finer fractions. After manual removal of the magnetite using a handheld magnet, the heavy minerals were separated using a Frantz isodynamic magnetic separator at the Natural Environment Research Council (NERC) Isotope Geosciences Laboratory (NIGL), Keyworth, UK. Both the magnetic and non-magnetic fractions were sampled. The separated heavy minerals were mounted in resin blocks and prepared as polished thin sections at the BGS. Separate polished blocks were prepared for zircon analysis.
Qualitative and quantitative chemical analyses of pyroxene and chevkinite group minerals (CGMs) were performed using an FEI Quanta 600 scanning electron microscope (SEM) with Oxford Instruments energy (EDS) and wavelength dispersive X-ray spectrometer (WDS), the latter for selected samples only. Samples were analyzed in the form of carbon sputter coated (≈25 nm thick) polished thin sections ( Table 3). The SEM was operated under high vacuum (<1 × 10 −5 torr) at 20 kV accelerating voltage, with the optimal analytical working distance of 10 mm and an X-ray detector take-off angle of 45 • . Modal analysis of the heavy mineral separates was based on SEM X-ray mapping of three representative fields of view, but may not include all the trace phases present in the sample. For the quantitative EDS analyses, 1 nA beam current and 120 s acquisition time were used, giving ≈ 1 × 10 6 total X-ray counts. EDS process time of 4 resulted in deadtime <45%. The beam stability was monitored by the analysis of a cobalt (Co) reference standard at the beginning of each analytical session and after every ten analyses of the unknown. The instrument performance on the day was also assessed using a well-characterised in-house standard Frisch pyroxene (augite) prior to the analysis of unknown pyroxene. End members of pyroxene from this study, and those from Alıcı et al. [55] and Platevoet et al. [58], were calculated following the same method [64]. The chevkinite group minerals (CGMs) exhibit a "problematic spectral region" between 4.4-4.9 keV with overlapping X-ray energies of Ce, La, Ba and Ti (Ce Lα 1 4.84 keV, La Lα 1 4.65 keV, Ti Kα 1 4.51 keV, Ti Kβ 4.931 keV and Ba Lα 1 4.465 keV). WDS, with superior spectral resolution to EDS, was used to resolve these overlaps. The WDS analysis was conducted using the same FEI 600 Quanta SEM with a Wave 500 spectrometer, running with Oxford Instruments INCA (Version 4) software. The SEM was operated at 20 kV accelerating voltage, 10 nA beam current. Only one WDS spectrometer was available and to minimize the risk of beam drift, the analysis was limited to nine elements per analytical spot only. The counting times varied per element between 20/10 s (peak/background) and 40/20 s for major and trace elements, respectively, resulting in a total acquisition time of approximately 30 min per analysis. The following reference materials and WDS crystals were used for calibration of the selected nine element X-ray lines: (LaLα)-LaB 6  and (VKα)-V metal & LIF. The calibration was performed every 24 h. This protocol was used for two representative samples, one from the Gölcük pyroclastic rocks and one from the Aksu Diamas pit with approximately ten EDS-WDS analyses per sample. All other samples of chevkinite group minerals were analyzed by EDS only. A detailed evaluation of the EDS and WDS results taken from the same analytical spot gave us confidence to use the EDS data for plotting of element ratios. One common disadvantage of the EDS analysis is the possible misidentification of Ba when either or both Ti and La are also present in the material. In our case, this X-ray energy overlap resulted in an overestimation of Ba in all crystals analyzed of about 1 wt.%. Although Ba can be present in the CGMs, this was not the case in the samples investigated here, where only a trace amount was found by WDS in a couple of crystals. The presence or absence of Ba strongly affects the reported amount of Ti and La in the analysis, thus all analyses here were recalculated without Ba. The effects of X-ray energy overlaps on the results of chevkinite microanalysis using SEM EDS-WDS are further explained by Lacinska and Rushton (2018) [65].
Collectively, approx. 30 analyses were obtained with EDS only and approx. 20 with a combination of EDS-WDS, with the majority summing up to 100 ± 2 wt. % and largely corresponding to the expected mineral stoichiometry. The limited compositional range and good agreement with published data gave us confidence in the quality of the results presented here. It must be noted that trace levels of other REE might be present in the CGMs but are below the detection limit of the SEM-EDS system, i.e., 0.1 wt. %. Several check analyses by WDS confirmed the presence of Eu and Gd in trace amounts.
The CGMs are Ti-REE silicates with general unit cell formulae calculated based on the idealized A 4 BC 2 D 2 Si 4 O 22 , where A = REE, Ca, Sr, Th; B = Fe 2+ ; C = Fe 3+ , Fe 2+ , Mn, Mg, Zr, Nb; and D = Ti [66]; using 22 oxygens, and the ideal cation sum of 13. In this study, the cation sum ranges from 13-13.4, the higher values possibly reflecting elements that might be present at different valence states, e.g., Ce presented here, as Ce 3+ might also be present as Ce 4+ . The cations have been allocated to sites following recommendations in Parodi et al. [67] and Macdonald and Belkin [68]. Two chevkinite group minerals were identified in the samples studied: chevkinite and perrierite. These minerals differ in crystal structure, with the most robust discriminant being the β angle of ca. 100 • for chevkinite and ca. 113 • for perrierite [69]. The different crystal structures allow variable amounts of cations into the different structural sites. The two minerals can also be distinguished based on their chemistry [68], typically, chevkinite is richer in FeO and poorer in CaO than perrierite [68].
Trace element concentrations in zircons from both Gölcük (MPLV057) and Aksu Diamas (MPLV051) were measured in polished blocks at NIGL using a New Wave Research 193UC laser ablation system equipped with a TV2 cell coupled to a Nu Instruments Attom single collector inductively coupled plasma mass spectrometer (SC-ICP-MS). Ablation conditions were a 25 µm static beam, 5 Hz repetition rate, 3 J/cm 2 fluence and a 20 second dwell. The Attom was run in linkscan mode, which measures two sweeps of the total mass range (both up and down) every 2 ms, with dwell times automatically split across the measured isotopes. These were 29 Si, 49  Normalisation was achieved using sample standard bracketing using 29 Si as an internal standard and 91500 zircon [70] as the reference material. Zircon GJ-1 [71] was measured as an unknown as a check on reproducibility and accuracy, and is within 10% of in-house values for all elements. Data handling and reduction were performed using Iolite software (Version 2.5) [72].
Magnetite trace elements were measured at BGS using a New Wave Research UP193FX excimer laser ablation system coupled to an Agilent 7500 series quadrupole ICP-MS, calibrated using a trace element glass standard (NIST SRM-610 [73]). Ablation conditions were a 70 µm spot size, 10 Hz repetition rate, a 3 J/cm 2 fluence, and a dwell time of 40 s. The relative standard deviation for the repeat analysis of a secondary trace element standard (NIST SRM-612 [73]) is typically better than 5% for the following elements of interest: Ca, Al, Nb, V, Mg, Co, Cr, Mn, Zr, Ni, Y, Pb, Cu, and better than 10% for Sn and Zn, and better than 15% Ti. The relative standard deviation for P is slightly higher, being just outside 20%. Data handling and reduction were performed using Iolite software (Version 2.5) [72].

Mineralogy
The heavy mineral-rich separate (125-250 µm size fraction) from the Gölcük samples comprises, in decreasing amounts, clinopyroxene (diopside) and amphibole (dominantly Ti-rich), magnetite-ulvospinel, biotite, apatite, volcanic glass and traces of lithic fragments, K-feldspar, fluorite, titanite, garnet (andraditic composition), plagioclase, chevkinite group minerals (CGMs), zircon, chlorite, Fe sulphide, rutile, monazite, allanite, britholite, thorite, and Ce oxide (possibly cerianite). An example of a modal analysis of one sample from each locality is presented in Table 4. Table 4. Modal analysis (%) of representative samples from the Gölcük HMS and Aksu Diamas HMS as an example of mineral distribution. The analysis was based on SEM X-ray mapping of three representative fields of view in each of the samples listed and may not include all the trace phases present in the sample. For ease mineral classification, the pyroxene and amphibole were grouped into one column. n.d.-not detected.

Sample
Pyroxene The particle morphology varies from angular, commonly euhedral to subhedral (particularly in the case of the pyroxene grains), to sub-rounded. The volcanic glass is typically vesicular and occurs as either separate particles or surrounding the pyroclasts, e.g., pyroxene ( Figure 4a). The mineral composition of the samples analyzed is consistent with the composition of volcanic rocks from Gölcük, as described by Alıcı et al. [55].  The particle morphology varies from angular, commonly euhedral to subhedral (particularly in the case of the pyroxene grains), to sub-rounded. The volcanic glass is typically vesicular and occurs as either separate particles or surrounding the pyroclasts, e.g. pyroxene (Figure 4a). The mineral composition of the samples analyzed is consistent with the composition of volcanic rocks from Gölcük, as described by Alıcı et al. [55]. The heavy minerals separate from the Aksu Diamas deposit includes in decreasing amount, magnetite-ulvospinel, clinopyroxene and Ti-rich amphibole, titanite (some grains contain TiO2 inclusions), apatite (locally as inclusions in clinopyroxene), lithic fragments, zircon and traces of chevkinite, garnet (andraditic composition), allanite, magnesiochromite, biotite, K feldspar, chlorite, Fe sulphide (usually as <5 µm inclusions in magnetite), rutile and thorite. An example modal analysis of one sample is presented in Table 4. The REE-bearing minerals include REE-bearing apatite, CGMs, allanite, bastnäsite, and monazite. Cerianite and possible britholite were also detected. Volcanic glass The heavy minerals separate from the Aksu Diamas deposit includes in decreasing amount, magnetite-ulvospinel, clinopyroxene and Ti-rich amphibole, titanite (some grains contain TiO 2 inclusions), apatite (locally as inclusions in clinopyroxene), lithic fragments, zircon and traces of chevkinite, garnet (andraditic composition), allanite, magnesiochromite, biotite, K feldspar, chlorite, Fe sulphide (usually as <5 µm inclusions in magnetite), rutile and thorite. An example modal analysis of one sample is presented in Table 4. The REE-bearing minerals include REE-bearing apatite, CGMs, allanite, bastnäsite, and monazite. Cerianite and possible britholite were also detected. Volcanic glass was rarely observed in the Aksu Diamas samples. The morphology of the particles in the 125-250 µm size fraction varies but most particles fall into an angular to sub-angular group, some grains being subhedral to euhedral (Figure 4b).
In order to test the hypothesis that the REE-bearing minerals at Aksu Diamas are derived from the Gölcük volcano, we carried out mineralogical and geochemical fingerprinting of minerals that are present in abundance in both sample groups. The study includes the crystal chemistry of the two main groups, pyroxene and CGMs, as well as trace element data for magnetite and zircon.

Gölcük Heavy Mineral Separate (HMS)
The pyroxene from the Gölcük HMS belongs to the clinopyroxene group and the majority of crystals studied are diopside, with a few lower-Ca crystals plotting in the augite compositional field ( Figure 5). are present in abundance in both sample groups. The study includes the crystal chemistry of the two main groups, pyroxene and CGMs, as well as trace element data for magnetite and zircon.

Pyroxene
Gölcük Heavy Mineral Separate (HMS) The pyroxene from the Gölcük HMS belongs to the clinopyroxene group and the majority of crystals studied are diopside, with a few lower-Ca crystals plotting in the augite compositional field ( Figure 5).  (Table S1; Supplementary Materials), along with literature data [55,58]      Some of the pyroxene crystals are chemically zoned. The zonation is controlled chiefly by the amount of Mg, Fe, Al, and to lesser extent Ca, Ti and Na; and manifests itself by Backscattered Electron (BSE) contrast differences within a single crystal (Figure 8). Notably, a large number of crystals displayed an Fe-enriched core overgrown with multiple Mg-or Fe-rich bands (Figure 8). Both the MgO and the FeO vary by ca. 2 wt. % between the corresponding bands. The thickness of bands as well as their distribution varies greatly from crystal to crystal and ranges from fine 1-5 µm oscillatory parallel lamellae mimicking the orientation of crystal faces, to up to 100 µm wide banding with orientation that is difficult to discern. The latter feature is typically present in broken fragments of larger crystals. This type of zonation is typical of phenocrysts from volcanic rocks. Some of the pyroxene crystals are chemically zoned. The zonation is controlled chiefly by the amount of Mg, Fe, Al, and to lesser extent Ca, Ti and Na; and manifests itself by Backscattered Electron (BSE) contrast differences within a single crystal (Figure 8). Notably, a large number of crystals displayed an Fe-enriched core overgrown with multiple Mg-or Fe-rich bands (Figure 8). Both the MgO and the FeO vary by ca. 2 wt. % between the corresponding bands. The thickness of bands as well as their distribution varies greatly from crystal to crystal and ranges from fine 1-5 µm oscillatory parallel lamellae mimicking the orientation of crystal faces, to up to 100 µm wide banding with orientation that is difficult to discern. The latter feature is typically present in broken fragments of larger crystals. This type of zonation is typical of phenocrysts from volcanic rocks.

Aksu Diamas HMS
The pyroxenes from the Aksu Diamas HMS also belong to the clinopyroxene group with diopside being the most common end-member. The chemical variation of pyroxene in the En-Fs-Wo

Aksu Diamas HMS
The pyroxenes from the Aksu Diamas HMS also belong to the clinopyroxene group with diopside being the most common end-member. The chemical variation of pyroxene in the En-Fs-Wo ternary system has the following ranges, En 33.6-55.7, Fs 5.6-19.5 and Wo 38.8-48.9 ( Figure 5). In the enstatite-aegirine-ferrosilite ternary diagram, the Aksu Diamas pyroxene data overlap with those from Gölcük and show a similar trend of increasing Fe with a slight increase in Na ( Figure 6). The amount of Al 2 O 3 ranges from 0.4-6.4 wt. % with nearly 90% of crystals having contents below 4.5 wt. %, giving an Al Si tetrahedral site substitution of ≤10% similar to that observed in the Gölcük HMS. In the crystals analyzed, Cr 2  Some crystals exhibit well-defined chemical zones (Figure 9), largely controlled by variation in Fe, Mg, and Al content. Many of the crystals analyzed displayed an Fe-enriched core surrounded by more or less well-defined Mg or Fe-enriched bands, with the FeO and MgO contents varying by <2 wt. % between the corresponding zones. The zonation is commonly oscillatory with bands of varying thickness, from a few microns to a few tens of microns. The distribution of compositional zones in crystals varies from sharply defined, mimicking crystal growth zones to embayed and generally uneven. The pattern of zonation is similar to that observed in pyroxene from the Gölcük HMS.
Minerals 2019, 9, x FOR PEER REVIEW 15 of 33 ternary system has the following ranges, En 33.6-55.7, Fs 5.6-19.5 and Wo 38.8-48.9 ( Figure 5). In the enstatite-aegirine-ferrosilite ternary diagram, the Aksu Diamas pyroxene data overlap with those from Gölcük and show a similar trend of increasing Fe with a slight increase in Na ( Figure 6). The amount of Al2O3 ranges from 0.4-6.4 wt. % with nearly 90% of crystals having contents below 4.5 wt. %, giving an Al ⇋ Si tetrahedral site substitution of ≤10% similar to that observed in the Gölcük HMS.
In the crystals analyzed, Cr2O3 varies from 0.1-0.4 wt. %, MnO 0.1-0.8 wt. % and TiO2 0.1-0.2 wt. %. The pyroxenes show a clear positive correlation between Cr2O3 + TiO2 and Al2O3 (Figure 7) and overlap with the Gölcük data. Some crystals exhibit well-defined chemical zones (Figure 9), largely controlled by variation in Fe, Mg, and Al content. Many of the crystals analyzed displayed an Fe-enriched core surrounded by more or less well-defined Mg or Fe-enriched bands, with the FeO and MgO contents varying by <2 wt. % between the corresponding zones. The zonation is commonly oscillatory with bands of varying thickness, from a few microns to a few tens of microns. The distribution of compositional zones in crystals varies from sharply defined, mimicking crystal growth zones to embayed and generally uneven. The pattern of zonation is similar to that observed in pyroxene from the Gölcük HMS.

Chevkinite Group Minerals (CGMs)
Gölcük HMS Ten quantitative EDS-WDS analyses were obtained for CGMs from the Gölcük HMS (Table 5, sample MPLV 059). The major element composition of the CGMs from Gölcük varies slightly from

Chevkinite Group Minerals (CGMs)
Gölcük HMS Ten quantitative EDS-WDS analyses were obtained for CGMs from the Gölcük HMS (Table 5, Table 5). The REE content of the minerals is dominated by light rare earth elements (LREE) with La and Ce being major components. Traces of Eu, Gd and Dy were detected by the WDS system in some crystals. Most crystals display complex chemical zonation resulting from the % variation in several elements, but are mostly associated with the FeO and ThO 2 contents. All crystals analyzed contain vanadium, with values generally ≤0.44 wt. % with one high value of 0.94 wt. % V 2 O 5 . A plot of ∑REE against ∑(Ca, Ti, Zr) ( Figure 10) shows a negative correlation, reflecting the common element substitution [67] in the chevkinite mineral group: Ca A 2+ + Ti C 4+ + Zr C 4+ = REE A 3+ .
against ∑(Ca, Ti, Zr) ( Figure 10) shows a negative correlation, reflecting the common element substitution [67] in the chevkinite mineral group: CaA 2+ + TiC 4+ + ZrC 4+ = REEA 3+ . As no crystallographic data on the CGMs was obtained in this study, a plot of CaO wt.% vs. FeO wt. % was used to distinguish between chevkinite and perrierite [68]. It is noted that only two data points from the Gölcük area samples are plotted in the perrierite field; thus, their definite presence in this location should be verified with a larger number of samples ( Figure 11).  As no crystallographic data on the CGMs was obtained in this study, a plot of CaO wt.% vs. FeO wt. % was used to distinguish between chevkinite and perrierite [68]. It is noted that only two data points from the Gölcük area samples are plotted in the perrierite field; thus, their definite presence in this location should be verified with a larger number of samples ( Figure 11).
against ∑(Ca, Ti, Zr) ( Figure 10) shows a negative correlation, reflecting the common element substitution [67] in the chevkinite mineral group: CaA 2+ + TiC 4+ + ZrC 4+ = REEA 3+ . As no crystallographic data on the CGMs was obtained in this study, a plot of CaO wt.% vs. FeO wt. % was used to distinguish between chevkinite and perrierite [68]. It is noted that only two data points from the Gölcük area samples are plotted in the perrierite field; thus, their definite presence in this location should be verified with a larger number of samples ( Figure 11).  Aksu Diamas HMS Ten quantitative EDS-WDS analyses were obtained for the minerals belonging to the CGM from the Aksu Diamas HMS sample MPLV 049 ( Table 6). The major oxides decrease in abundance from  The minerals are dominated by LREE content, with traces of Gd, Eu and Dy detected in some crystals. Similar to the CGM in the Gölcük samples, a complex chemical zonation was observed in the CGM from the Aksu Diamas HMS. Although it is difficult to confirm the main elements contributing to the differences in backscatter electron contrast, it is likely that both FeO and ThO 2 are the major source of those differences ( Figure 12). All crystals analyzed contain V 2 O 5, ranging from 0.1-0.48 wt.%. On the ∑REE against ∑(Ca, Ti, Zr) plot (Figure 9), the Aksu Diamas samples show the same negative correlation as those from Gölcük. the major source of those differences ( Figure 12). All crystals analyzed contain V2O5, ranging from 0.1-0.48 wt.%. On the ∑REE against ∑(Ca, Ti, Zr) plot (Figure 9), the Aksu Diamas samples show the same negative correlation as those from Gölcük. As shown in Figure 11, >70% of crystals analyzed from the Aksu Diamas HMS sample plot within the chevkinite compositional range, with FeO and CaO content > 8 wt. % and < 5 wt. %, respectively. A few crystals have CaO > 5 wt. % and plot in the perrierite field.

Gölcük HMS
The average content of REE in zircon from Gölcük is approx. 7000 mg/kg ( Table 7). The range of chondrite-normalized [74] trace element abundances for zircons from both Gölcük and Aksu Diamas show patterns typical of magmatic zircon [75,76], with a steep rise in the middle (Sm) to heavy (Lu) REE ( Figure 13), a strong positive Ce anomaly, and a weak negative Eu anomaly. The Lan/Lun ratio for zircon from Gölcük ranges from 0.00004 to 0.00128 with an average of 0.00024.  [74] REE data for zircon in samples from Gölcük (blue) and Aksu Diamas (red). Chondrite normalized [74] REE (median values) in typical alkaline syenite and larvikite zircons plotted for comparison [75]. As shown in Figure 11, >70% of crystals analyzed from the Aksu Diamas HMS sample plot within the chevkinite compositional range, with FeO and CaO content > 8 wt. % and < 5 wt. %, respectively. A few crystals have CaO > 5 wt. % and plot in the perrierite field.

Gölcük HMS
The average content of REE in zircon from Gölcük is approx. 7000 mg/kg ( Table 7). The range of chondrite-normalized [74] trace element abundances for zircons from both Gölcük and Aksu Diamas show patterns typical of magmatic zircon [75,76], with a steep rise in the middle (Sm) to heavy (Lu) REE ( Figure 13), a strong positive Ce anomaly, and a weak negative Eu anomaly. The La n /Lu n ratio for zircon from Gölcük ranges from 0.00004 to 0.00128 with an average of 0.00024. Minerals 2019, 9, x FOR PEER REVIEW 19 of 33 the major source of those differences ( Figure 12). All crystals analyzed contain V2O5, ranging from 0.1-0.48 wt.%. On the ∑REE against ∑(Ca, Ti, Zr) plot (Figure 9), the Aksu Diamas samples show the same negative correlation as those from Gölcük. As shown in Figure 11, >70% of crystals analyzed from the Aksu Diamas HMS sample plot within the chevkinite compositional range, with FeO and CaO content > 8 wt. % and < 5 wt. %, respectively. A few crystals have CaO > 5 wt. % and plot in the perrierite field.

Gölcük HMS
The average content of REE in zircon from Gölcük is approx. 7000 mg/kg ( Table 7). The range of chondrite-normalized [74] trace element abundances for zircons from both Gölcük and Aksu Diamas show patterns typical of magmatic zircon [75,76], with a steep rise in the middle (Sm) to heavy (Lu) REE ( Figure 13), a strong positive Ce anomaly, and a weak negative Eu anomaly. The Lan/Lun ratio for zircon from Gölcük ranges from 0.00004 to 0.00128 with an average of 0.00024.  [74] REE data for zircon in samples from Gölcük (blue) and Aksu Diamas (red). Chondrite normalized [74] REE (median values) in typical alkaline syenite and larvikite zircons plotted for comparison [75].

Aksu Diamas HMS
The average content of REE in zircon from Aksu Diamas is ca. 6000 mg/kg ( Table 8). The La n /Lu n ratio for zircon from Aksu Diamas ranges from 0.00002 to 0.00214 with an average of 0.00022. The REE pattern for Aksu Diamas zircons is almost indistinguishable from that for Gölcük ( Figure 13). This shows a strong positive Ce anomaly, a weak negative Eu anomaly, and a steep rise in REE content from Sm to Lu; features all typical of magmatic zircon [75,76]. Table 7. Trace element analysis of zircon (mg/kg) from Gölcük heavy mineral separates. La n /Lu n = normalized to chondrite after Sun and McDonough (1989) [74].   Table 9). All other elements of interest (i.e., Ni, Cu, Y, Zr, Nb, Sn and Pb) are typically present at concentrations of less than 100 mg/kg. The majority of magnetite grains analyzed from Gölcük have Ni/Cr ratios of <1, which are very similar to published data for magnetite from andesitic rocks [77], although there are several grains from Gölcük with much higher ratios (i.e., between 1.39 and 4.04) (Figure 14).

Aksu Diamas HMS
Magnetite from the Aksu Diamas placer contains relatively higher concentrations of Ti (1.14-3.  (Table 9). Zn (696-1286 mg/kg) and P (76-1350 mg/kg) concentrations are overall higher in magnetite from Aksu Diamas. On average, the concentration of other elements of interest (i.e., Cr, Ni, Cu, Y, Zr, Nb and Sn) is below 100 mg/kg in magnetite from Aksu Diamas. The Ni/Cr ratio for Aksu Diamas magnetite grains ranges between 0.227 and 1.88, although the majority of grains have a Ni/Cr ratio of <1 ( Figure 14). This range of Ni/Cr ratio values is consistent with published data for magnetite from intermediate volcanic rocks [77].  [77].

Discussion
The clearly overlapping composition of the fingerprinting minerals, pyroxene, chevkinite group minerals, zircon and magnetite, both in major and trace element chemistry from the two localities is consistent with the hypothesis that the heavy minerals at Aksu Diamas are sourced from the Gölcük volcano.

Pyroxene
Pyroxene, due to its abundance and resistance to supergene processes, is one of the most important mineral phases for provenance studies and comparison between the Gölcük and Aksu Diamas sites. The Ens-Wol-Fs ternary diagram ( Figure 5) shows a clear match between the two sampling sites, with the majority of the analyses plotting within the diopside compositional field. These results are also compared to published data from the Gölcük igneous province (Platevoet et al. [58] and Alıcı et al. [55]) and again a strong agreement is observed, in particular with the Alıcı et al. data [55]. The Gölcük and Aksu Diamas pyroxenes clearly overlap in the enstatite-aegirine-ferrosilite ternary diagram ( Figure 6) and also overlap with the data for some of the youngest volcanic rocks from Gölcük [55]. Both data sets show a clear positive correlation between Cr2O3 + TiO2 and Al2O3 (Figure 7). The variability in pyroxene composition across all the samples suggests that each eruption sampled material from different levels in the magma chamber that recorded varying stages of magmatic evolution. Overall, pyroxene from Aksu Diamas records the same compositional variation as those from Gölcük. Furthermore, the chemical zonation of crystals from both localities follows the same pattern of chemical alternation between Fe-or Mg-enriched zones, normally with an Fe-enriched core (Figures 8 and 9).

Chevkinite
The analysis of both sets of samples has revealed the presence of a Ti-REE silicate, belonging to the chevkinite group minerals (CGM). These minerals are a widespread, but very rare, mineral in igneous and metamorphic parageneses [68,78] and exhibit complex chemistry with a wide range of possible element substitutions. The minerals are difficult to identify by optical petrography and it is rare that they occur in sufficient amounts to be listed and described as a significant component of the

Discussion
The clearly overlapping composition of the fingerprinting minerals, pyroxene, chevkinite group minerals, zircon and magnetite, both in major and trace element chemistry from the two localities is consistent with the hypothesis that the heavy minerals at Aksu Diamas are sourced from the Gölcük volcano.

Pyroxene
Pyroxene, due to its abundance and resistance to supergene processes, is one of the most important mineral phases for provenance studies and comparison between the Gölcük and Aksu Diamas sites. The Ens-Wol-Fs ternary diagram ( Figure 5) shows a clear match between the two sampling sites, with the majority of the analyses plotting within the diopside compositional field. These results are also compared to published data from the Gölcük igneous province (Platevoet et al. [58] and Alıcı et al. [55]) and again a strong agreement is observed, in particular with the Alıcı et al. data [55]. The Gölcük and Aksu Diamas pyroxenes clearly overlap in the enstatite-aegirine-ferrosilite ternary diagram ( Figure 6) and also overlap with the data for some of the youngest volcanic rocks from Gölcük [55]. Both data sets show a clear positive correlation between Cr 2 O 3 + TiO 2 and Al 2 O 3 (Figure 7). The variability in pyroxene composition across all the samples suggests that each eruption sampled material from different levels in the magma chamber that recorded varying stages of magmatic evolution. Overall, pyroxene from Aksu Diamas records the same compositional variation as those from Gölcük. Furthermore, the chemical zonation of crystals from both localities follows the same pattern of chemical alternation between Fe-or Mg-enriched zones, normally with an Fe-enriched core (Figures 8 and 9).

Chevkinite
The analysis of both sets of samples has revealed the presence of a Ti-REE silicate, belonging to the chevkinite group minerals (CGM). These minerals are a widespread, but very rare, mineral in igneous and metamorphic parageneses [68,78] and exhibit complex chemistry with a wide range of possible element substitutions. The minerals are difficult to identify by optical petrography and it is rare that they occur in sufficient amounts to be listed and described as a significant component of the rocks. CGMs are unlikely to be sourced from the bedrock that surrounds the Aksu Diamas deposit, as these are dominated by sedimentary and ophiolitic lithologies, particularly limestone ( Figure 1B). The CGM fractionate the LREE [78], the ΣREE 2 O 3 content of up to 49 wt. %, in some crystals analyzed here, demonstrates their high capacity to concentrate rare earth elements. Although CGM are the most abundant REE-bearing minerals present in the samples, they have not previously been described in detail from the Gölcük volcanic rocks. However, they appear to be resistant to weathering and therefore represent an excellent fingerprinting phase.
Importantly, the set of CGMs analyzed here from the two localities display a very similar compositional range (Figure 10), implying that they originate from the same source. To unambiguously distinguish between different CGMs, crystallographic data is required; however, chevkinite and perrierite can also be differentiated based on their chemistry [68]. Figure 11 presents data points from both localities, with a clear distinction between the two minerals. The stability of perrierite and its relationship to chevkinite are not well known, especially in natural systems [67]. It is however, generally uncommon to find chevkinite and perrierite in the same lithology, with the latter being more common in subalkaline suites, whilst chevkinite is found in alkaline and peralkaline rocks [79]. The origin of the perrierite found both at Gölcük and Aksu Diamas remains unclear; however, its presence in both sample sets is further evidence for a shared source. The CGMs from both Gölcük and Aksu Diamas display a complex and uneven chemical zoning ( Figure 12). The origin of the zoning could be related to magma mixing, hydrothermal alteration [79], or grain boundary diffusion of elements and recrystallisation facilitated by metamictisation related to the presence of Th and U [80].
Notably, all crystals analyzed contain vanadium in amounts varying from 0.1-0.94 wt. % V 2 O 5 (Tables 5 and 6). This element is considered trace, ≤1 wt. % [68] and generally not included in the literature data, being either absent, below the detection limits, or omitted in reporting. Its presence in comparable quantities in both data sets provides another line of evidence for a common source for the CGMs in this study.

Zircon and Magnetite
Geochemical fingerprinting was also performed using trace element concentrations in zircon and magnetite, and there is an excellent match between the data from Aksu Diamas and Gölcük. Notably, the REE pattern of zircon from the two localities ( Figure 13) are nearly identical. Other potential sources of zircon in the area are limited to detrital zircon in Miocene-Eocene sediments and zircon in igneous rocks present in the ophiolitic sequences of the Lycian and Antalya nappes. There is no evidence from the geological map for other magmatic sources of zircon in the Isparta region, other than those associated with the Gölcük volcanic complex [81].
Magnetite from Gölcük and Aksu Diamas have overlapping compositions that are dominated by Ti, Mg, Al, V and Mn. The relatively high concentration of Al (up to 2 wt. %) and Ti (up to 3 wt. %) ( Table 9) in magnetite from Gölcük and Aksu Diamas indicates a magmatic rather than hydrothermal origin, since hydrothermal magnetite typically contains <1 wt. % Al and <2 wt. % Ti [77]. The concentration of other elements of interest (i.e., Zn, Co, Cr, Ni, Cu, Nb and Sn) are remarkably consistent between the two localities, which indicates that Gölcük is the likely source of the magnetite found in the Aksu Diamas placer. The Ni/Cr ratio of magnetite from both localities is also very consistent, mostly being <1. The Ni/Cr ratios, when plotted against Ti, can also be used to discriminate between magnetite from mafic, intermediate, and felsic magmas. When plotted, magnetite from both localities cluster in the intermediate magma field, which would fit with their derivation from a trachyandesitic volcanic system [77] (Figure 14).

Formation of the Aksu Diamas Placer Deposit
Although AMR Mineral and Metal, Inc. interpreted the Aksu Diamas deposit as being formed between 7 Ma and 8 Ma using field mapping and structural interpretation, evidence for this interpretation was not detailed in the NI 43-101 report [32]. Volcanic activity at Gölcük began at 4.07 ± 0.20 to 4.7 ± 0.50 Ma [55,56], younger than the age proposed for the Aksu Diamas deposit by AMR. Recent work, however, indicates that the youngest activity at Gölcük occurred at ca. 0.013 Ma [51]. The unconsolidated nature of the sediments at Aksu Diamas is consistent with their being Quaternary in age (i.e., post-dating Pliocene deformation in the Isparta region [46,82]) and thus potentially sourced from the more recent eruptive cycles at Gölcük.
The mineralogical and geochemical evidence presented here strongly support the hypothesis that volcanic activity at Gölcük was the source of the heavy minerals at Aksu Diamas. However, no connected fluvial systems occur between the two localities, as they are separated by the western end of the Taurus Mountain range (Figure 2). We, therefore, consider that the heavy minerals must originally have been transported as airborne particles, before being deposited onto the palaeo-topography ( Figure 15). The unconsolidated nature of the sediments at Aksu Diamas is consistent with their being Quaternary in age (i.e. post-dating Pliocene deformation in the Isparta region [46,82]) and thus potentially sourced from the more recent eruptive cycles at Gölcük. The mineralogical and geochemical evidence presented here strongly support the hypothesis that volcanic activity at Gölcük was the source of the heavy minerals at Aksu Diamas. However, no connected fluvial systems occur between the two localities, as they are separated by the western end of the Taurus Mountain range (Figure 2). We, therefore, consider that the heavy minerals must originally have been transported as airborne particles, before being deposited onto the palaeo-topography ( Figure 15). The intense H2O variations in the magma were responsible for the explosive nature of the eruptions at Gölcük [58], and consequently the extensive development of ash clouds, from which REE-bearing minerals and magnetite were deposited. Volcanic material from Eruptive Cycle III, which formed late tuff-ring deposits at Gölcük [49], was found to form 30-100 cm thick pumice falls in the Isparta and Eğirdir grabens [49]. Evidence from Lake Eğirdir, 35 km northeast of Gölcük, shows that explosive eruptions resulted in the deposition of turbulent pyroclastic surges with large lithic clasts up to 3 cm in diameter [53]. This indicates that eruptions at Gölcük would have had sufficient energy to transport material up to 125 µm diameter over the top of the mountain watershed to basins only 23 km to the south east where Aksu Diamas is located.
Air-borne particles of ash that resulted from the explosive eruptions of Gölcük would have landed as multiple layers of tephra on the surrounding area, including the basin that now hosts the Aksu Diamas deposit and the slopes of the surrounding limestone mountains. The heavy minerals have been preserved in the matrix of lens-shaped channels with angular limestone clasts in a muddy sand matrix in the basin at the foot of the hills. Transport of the heavy minerals is evident from their redistribution into the matrix of the channels (Figure 3d), forming the placer deposits. The intense H 2 O variations in the magma were responsible for the explosive nature of the eruptions at Gölcük [58], and consequently the extensive development of ash clouds, from which REE-bearing minerals and magnetite were deposited. Volcanic material from Eruptive Cycle III, which formed late tuff-ring deposits at Gölcük [49], was found to form 30-100 cm thick pumice falls in the Isparta and Egirdir grabens [49]. Evidence from Lake Egirdir, 35 km northeast of Gölcük, shows that explosive eruptions resulted in the deposition of turbulent pyroclastic surges with large lithic clasts up to 3 cm in diameter [53]. This indicates that eruptions at Gölcük would have had sufficient energy to transport material up to 125 µm diameter over the top of the mountain watershed to basins only 23 km to the south east where Aksu Diamas is located.
Air-borne particles of ash that resulted from the explosive eruptions of Gölcük would have landed as multiple layers of tephra on the surrounding area, including the basin that now hosts the Aksu Diamas deposit and the slopes of the surrounding limestone mountains. The heavy minerals have been preserved in the matrix of lens-shaped channels with angular limestone clasts in a muddy sand matrix in the basin at the foot of the hills. Transport of the heavy minerals is evident from their redistribution into the matrix of the channels (Figure 3d), forming the placer deposits.
The sequence observed at Aksu Diamas is interpreted as a series of episodic mass-flow events. The limestone clasts are interpreted as being locally sourced from the Bey Daglari limestone ( Figure 1B). In contrast, the heavy minerals have been shown to come from the Gölcük volcano, over 23 km away. The poorly-sorted nature and angularity of both the limestone clasts and the heavy mineral grains suggests that the mass-flows deposits have travelled short distances. This fits with a two-phase process for the formation of the Aksu Diamas deposit. Firstly, the airborne transport and deposition of tephra from the Gölcük eruptions, and secondly, local reworking as part of series of mass-flow events. The sequence exposed in the pit is likely part of a larger alluvial fan system, draining locally to the north. Similar Quaternary-aged alluvial fan-systems are typically seen in the Mediterranean [83], the formation of which is attributed to climatic change driven by glacial-interglacial cyclicity [83][84][85].

Implications for Exploration
In the region around Gölcük and the Aksu Diamas deposit, AMR Minerals, Metals, Inc. identified exploration targets with a strong magnetic geophysical anomaly, similar to the anomaly at Aksu Diamas [32]. Some of these targets were located to the north and east of Gölcük, and others to the south of Aksu Diamas. This implies that tephra may have been widely distributed during past eruptive cycles, and that placer deposition was controlled by the presence of basins and sub-basins where tephra was reworked by alluvial processes.
The identification of this deposit type has implications for other areas in Europe, particularly in the circum-Mediterranean region, where certain factors make it particularly prospective for these types of low-grade REE deposits. Alkaline volcanism acts as the source of the REE-bearing phases, and this type of volcanism generally occurs due to extension following closure of the various strands of the Tethys [39]. Localised sedimentary basins could then act as traps for the heavy minerals. Therefore, tephra from other similar provinces, such as the Italian Roman Volcanic province and southern Italy [86], where there is the potential for these to have been reworked and redeposited, may be considered as prospective targets for low-grade REE mineralization.
The formation of these deposit types is highly dependent on the presence of an explosively erupting alkaline volcano and appropriate geomorphological and climatic conditions. Certain aspects of these deposit types underpin their future potential as REE resources. At Aksu Diamas, the deposit is very low-grade, but considered to be a significant size (454 Mt) and relatively easy to beneficiate. The weakly consolidated nature of the mineralization results in easier mining with water monitors, easy slurry formation, and great potential for water reuse. Gravity separation of the heavy minerals is a relatively low-cost, low-energy technology, which can easily separate out the individual phases. This was demonstrated at the pilot processing plant operated by AMR Minerals and Metals, Inc. at the Aksu Diamas site. In this particular case, the REE are a by-product of iron production, due to the abundance of magnetite in the deposit. This has a further benefit, as the strong magnetic geophysical signature of magnetite allows for relatively easy exploration for similar deposits.

Conclusions
The Aksu Diamas iron-REE deposit is an atypical placer deposit containing magnetite and a range of REE minerals, such as chevkinite, perrierite, apatite, allanite, and britholite. Our work shows that the most likely source for the Aksu Diamas deposit is the Gölcük alkaline volcano, some 23 km to the north. A comparative mineralogical and geochemical study of fingerprinting minerals provides compelling evidence for the volcanic provenance of the Aksu Diamas deposit. The deposit was formed when tephra material from Gölcük was reworked into a basin by alluvial processes, concentrating the heavy minerals.
Similar volcanic-derived placer deposits have rarely been described elsewhere, but may represent an important potential resource of the REE. This deposit type may be more common than previously described in the literature and therefore deserves further study. We suggest that such deposits may occur in association with many of the alkaline volcanic provinces in the circum-Mediterranean region.