Fingerprinting Paranesti Rubies through Oxygen Isotopes

In this study, the oxygen isotope (δ18O) composition of pink to red gem-quality rubies from Paranesti, Greece was investigated using in-situ secondary ionization mass spectrometry (SIMS) and laser-fluorination techniques. Paranesti rubies have a narrow range of δ18O values between ~0 and +1h and represent one of only a few cases worldwide where δ18O signatures can be used to distinguish them from other localities. SIMS analyses from this study and previous work by the authors suggests that the rubies formed under metamorphic/metasomatic conditions involving deeply penetrating meteoric waters along major crustal structures associated with the Nestos Shear Zone. SIMS analyses also revealed slight variations in δ18O composition for two outcrops located just ~500 m apart: PAR-1 with a mean value of 1.0h ± 0.42h and PAR-5 with a mean value of 0.14h ± 0.24h. This work adds to the growing use of in-situ methods to determine the origin of gem-quality corundum and re-confirms its usefulness in geographic “fingerprinting”.


Oxygen Isotopic Studies in Corundums
Oxygen is an abundant element in the Earth's crust, mantle and fluids.Oxygen consists of three naturally-occurring stable isotopes: 16 O (99.76%), 17 O (0.04%) and 18 O (0.2%).δ 18 O expressed as Vienna standard mean ocean water (VSMOW) in per mil is the standard for the oxygen isotopic composition which is a measure of the ratio of the stable isotopes oxygen-18 ( 18 O) and oxygen-16 ( 16 O).There are numerous applications of oxygen isotope geochemistry including paleoclimatology, urban forensics, geological genesis and many more [1][2][3].Oxygen isotope fractionation is a function of the initial Rayleigh evaporation-precipitation cycle, temperature of the system and degree of water-rock interaction and therefore great care must be taken when interpreting oxygen isotope values [4][5][6][7].
Although worldwide corundum oxygen isotope values have been found in a wide range from -27 (Khitostrov, Russia) to +23 (Mong Hsu, Myanmar), most are in the range of +3 to +21 [8][9][10].This criterion has often been used to determine the geological origin of coloured corundum and especially the gem corundums, rubies and sapphires.δ 18 O has been particularly useful in determining the likely primary geological origin of placer corundums where the primary origin is uncertain [11].As isotopic fractionation is a function of both temperature and geological processes, oxygen isotope data need to be treated with some degree of caution Thus, there are very few examples where oxygen isotopes have been used to "fingerprint" the geographic location [12].

Geological Setting and Sample Background
The Paranesti rubies are found within the Nestos Shear Zone (NSZ) of the Rhodope Mountain Complex (RMC) in north-eastern Greece (Figure 1).The tectonic and polymetamorphic record of this Northern Aegean region (including the RMC) reflects the Middle Jurassic to Neogene northeast dipping subduction and convergence of the African-Eurasian plates which resulted in the closure of the Tethys Ocean [13,14].The NSZ is thought to be a one of the syn-metamorphic thrusts in the RMC that are responsible for regional metamorphic inversion, placing higher amphibolite-facies intermediate terranes onto upper-greenschist to lower amphibolite-facies rocks of the lower terrane [15,16].
Minerals 2019, 9, x FOR PEER REVIEW 2 of 14 Although worldwide corundum oxygen isotope values have been found in a wide range from -27‰ (Khitostrov, Russia) to +23‰ (Mong Hsu, Myanmar), most are in the range of +3‰ to +21‰ [8][9][10].This criterion has often been used to determine the geological origin of coloured corundum and especially the gem corundums, rubies and sapphires.δ 18 O has been particularly useful in determining the likely primary geological origin of placer corundums where the primary origin is uncertain [11].As isotopic fractionation is a function of both temperature and geological processes, oxygen isotope data need to be treated with some degree of caution Thus, there are very few examples where oxygen isotopes have been used to "fingerprint" the geographic location [12].

Geological Setting and Sample Background
The Paranesti rubies are found within the Nestos Shear Zone (NSZ) of the Rhodope Mountain Complex (RMC) in north-eastern Greece (Figure 1).The tectonic and polymetamorphic record of this Northern Aegean region (including the RMC) reflects the Middle Jurassic to Neogene northeast dipping subduction and convergence of the African-Eurasian plates which resulted in the closure of the Tethys Ocean [13,14].The NSZ is thought to be a one of the syn-metamorphic thrusts in the RMC that are responsible for regional metamorphic inversion, placing higher amphibolite-facies intermediate terranes onto upper-greenschist to lower amphibolite-facies rocks of the lower terrane [15,16].Based on an earlier systematic study on Paranesti [18], the ruby-bearing occurrences were found to be hosted in pargasite schist with a mafic/ultramafic protolith.The surrounding non-corundumbearing chlorite schist was found to mainly be comprised of clinochlore.The ruby-bearing occurrence found on the hillside is referred to as PAR-1 (Figure 2a) and the road-side occurrence is termed PAR-5 (Figure 2b).Not all of the pargasite boudins nor the pegmatite intrusion found within the vicinity of the two sites contained corundum (Figure 2c,d).Based on an earlier systematic study on Paranesti [18], the ruby-bearing occurrences were found to be hosted in pargasite schist with a mafic/ultramafic protolith.The surrounding non-corundum-bearing chlorite schist was found to mainly be comprised of clinochlore.The ruby-bearing occurrence found on the hillside is referred to as PAR-1 (Figure 2a) and the road-side occurrence is termed PAR-5 (Figure 2b).Not all of the pargasite boudins nor the pegmatite intrusion found within the vicinity of the two sites contained corundum (Figure 2c,d).A summary of the main findings from this previous study is listed in Table 1.Detailed LA-ICP-MS trace element analyses showed that the rubies are of metamorphic origin (Figure 3a) with minor partial metasomatic influences (Figure 3b).The high R 2 value based on the Fe/Mg vs Ga/Mg elemental discrimination diagram shows both PAR-1 and PAR-5 rubies to contain highly consistent trace element compositions (Figure 3a).A summary of the main findings from this previous study is listed in Table 1.Detailed LA-ICP-MS trace element analyses showed that the rubies are of metamorphic origin (Figure 3a) with minor partial metasomatic influences (Figure 3b).The high R 2 value based on the Fe/Mg vs Ga/Mg elemental discrimination diagram shows both PAR-1 and PAR-5 rubies to contain highly consistent trace element compositions (Figure 3a).

Materials and Methods
Two different oxygen isotopic analytical methods have been used in this study in order to verify the oxygen isotope values of the Paranesti rubies.The rubies were mechanically extracted from the pargasite host matrix and carefully cleaned prior to being sent for analysis.In many samples, the ruby crystals occur in clusters of platy crystals where the grain sizes generally range between 0.5-1.5 cm (Figure 4a-c).Importantly, in the previous study [18] the ruby grains were generally found to be free of inclusions and thus amenable to in-situ analysis.

Materials and Methods
Two different oxygen isotopic analytical methods have been used in this study in order to verify the oxygen isotope values of the Paranesti rubies.The rubies were mechanically extracted from the pargasite host matrix and carefully cleaned prior to being sent for analysis.In many samples, the ruby crystals occur in clusters of platy crystals where the grain sizes generally range between 0.5-1.5 cm (Figure 4a-c).Importantly, in the previous study [18] the ruby grains were generally found to be free of inclusions and thus amenable to in-situ analysis.

Materials and Methods
Two different oxygen isotopic analytical methods have been used in this study in order to verify the oxygen isotope values of the Paranesti rubies.The rubies were mechanically extracted from the pargasite host matrix and carefully cleaned prior to being sent for analysis.In many samples, the ruby crystals occur in clusters of platy crystals where the grain sizes generally range between 0.5-1.5 cm (Figure 4a-c).Importantly, in the previous study [18] the ruby grains were generally found to be free of inclusions and thus amenable to in-situ analysis.

Laser-Fluorination Method (2009)
In 2009, a reconnaissance study was conducted, whereby five individual grains, one each from different corundum localities/geological environments in Greece, were studied for their δ 18 O composition.These included two colourless to blue sapphires in desilicified pegmatite from Naxos, one pink marble-hosted ruby from Kimi and one purple marble-hosted ruby from Xanthi.One medium red intensity ruby in pargasite schist from Paranesti (PAR-1) was also included.Oxygen isotope analyses were performed using a modification of the laser-fluorination technique described by Sharp [21] that was similar to that applied by Giuliani et al. in 2005 [11].
The method involves the complete reaction of ~1 mg of ground corundum.This powder is then heated by a CO2 laser, with ClF3 as the fluorine reagent.The released oxygen is passed through an inline Hg-diffusion pump before conversion to CO2 on platinized graphite.The yield is then measured by a capacitance manometer.The gas-handling vacuum line is connected to the inlet system of a dedicated VG PRISM 3 dual inlet isotope-ratio mass spectrometer.All oxygen isotope ratios are reported in δ 18 O (‰) relative to Vienna standard mean ocean water (VSMOW).The secondary standard used for the laser-fluorination method was an internal quartz standard, NBS28 quartz that gave an average δ 18 O value of 9.6‰.Oxygen yields differing significantly from the theoretical value of 14.07 µmol.per mg were taken as likely evidence of analytical artefact.Precision and accuracy on the internal quartz standard are ± 0.1‰ (1σ).Duplicate and triplicate analyses of sapphire and ruby suggested that this is appropriate for such materials.

Secondary Ion Mass Spectrometry (SIMS) Method (2017)
The 2017 analyses were performed exclusively on a range of coloured ruby grains from the two distinct Paranesti locations described in the previous study [13].Unlike the 2009 analyses, these analyses were made using secondary ionisation mass spectrometry (SIMS) to analyse different areas of selected ruby grains in-situ to measure oxygen isotope ratios with less than one per mil (‰) level precision.Oxygen isotope ratios ( 18 O/ 16 O) in ruby were determined using a Cameca IMS 1280 multicollector ion microprobe within the Centre for Microscopy, Characterisation and Analysis (CMCA), University of Western Australia (UWA).The materials examined in 2017 included 3 ruby grains from 3 different samples (57 analyses) from PAR-1 and 5 ruby grains from 5 different samples (44 analyses) from PAR-5.Each analysis point is shown in Figure 5a,b.

Laser-Fluorination Method (2009)
In 2009, a reconnaissance study was conducted, whereby five individual grains, one each from different corundum localities/geological environments in Greece, were studied for their δ 18 O composition.These included two colourless to blue sapphires in desilicified pegmatite from Naxos, one pink marble-hosted ruby from Kimi and one purple marble-hosted ruby from Xanthi.One medium red intensity ruby in pargasite schist from Paranesti (PAR-1) was also included.Oxygen isotope analyses were performed using a modification of the laser-fluorination technique described by Sharp [21] that was similar to that applied by Giuliani et al. in 2005 [11].
The method involves the complete reaction of ~1 mg of ground corundum.This powder is then heated by a CO 2 laser, with ClF 3 as the fluorine reagent.The released oxygen is passed through an in-line Hg-diffusion pump before conversion to CO 2 on platinized graphite.The yield is then measured by a capacitance manometer.The gas-handling vacuum line is connected to the inlet system of a dedicated VG PRISM 3 dual inlet isotope-ratio mass spectrometer.All oxygen isotope ratios are reported in δ 18 O ( ) relative to Vienna standard mean ocean water (VSMOW).The secondary standard used for the laser-fluorination method was an internal quartz standard, NBS28 quartz that gave an average δ 18 O value of 9.6 .Oxygen yields differing significantly from the theoretical value of 14.07 µmol.per mg were taken as likely evidence of analytical artefact.Precision and accuracy on the internal quartz standard are ± 0.1 (1σ).Duplicate and triplicate analyses of sapphire and ruby suggested that this is appropriate for such materials.

Secondary Ion Mass Spectrometry (SIMS) Method (2017)
The 2017 analyses were performed exclusively on a range of coloured ruby grains from the two distinct Paranesti locations described in the previous study [13].Unlike the 2009 analyses, these analyses were made using secondary ionisation mass spectrometry (SIMS) to analyse different areas of selected ruby grains in-situ to measure oxygen isotope ratios with less than one per mil ( ) level precision.Oxygen isotope ratios ( 18 O/ 16 O) in ruby were determined using a Cameca IMS 1280 multi-collector ion microprobe within the Centre for Microscopy, Characterisation and Analysis (CMCA), University of Western Australia (UWA).The materials examined in 2017 included 3 ruby grains from 3 different samples (57 analyses) from PAR-1 and 5 ruby grains from 5 different samples (44 analyses) from PAR-5.Each analysis point is shown in Figure 5a The sample mounts were carefully cleaned with detergent, distilled water and ethanol in an ultrasonic bath and then coated with gold (30 nm in thickness) prior to SIMS O isotope analyses.For oxygen isotopic analyses, secondary ions were sputtered from the sample by bombarding its surface with a Gaussian Cs + beam and a total impact energy of 20 keV.The surface of the sample was rastered with a 2.5 nA primary beam over a 15 × 15 µm area.An electron gun was used to ensure charge compensation during the analyses.Secondary ions were admitted in the double focusing mass spectrometer within a 100 µm entrance slit and focused in the centre of a 4000 µm field aperture (×100 magnification).They were energy filtered using a 30 eV band pass with a 5 eV gap toward the highenergy side. 16O and 18 O were collected simultaneously in multicollection mode in Faraday Cup detectors fitted with 10 10 Ω and 10 11 Ω, respectively.Each analysis includes a pre-sputtering over a 20 × 20 µm area during 30 s and the automatic centring of the secondary ions in the field aperture, contrast aperture and entrance slit and consisted of 20 four-second cycles which give an average internal precision of ~0.16‰ (2 SE).
External reproducibility during the analytical sessions was evaluated by repeating analyses in one single fragment of PAR-1.External reproducibility in this fragment was 0.3 and 0.4 per mil (2SD) during the two analytical sessions.In total, three large fragments of PAR-1 were analysed for their oxygen isotope composition, altogether yielding an average value of 0.9 ± 0.6 per mil (2SD, n = 57, Table 2).Raw oxygen isotope ratios were corrected for instrumental mass fractionation using the δ 18 O composition of PAR-1, which oxygen isotope composition was obtained by laser fluorination method (from the 2009 study).Uncertainty on each δ 18 O spot has been calculated by propagating the errors on instrumental mass fractionation determination, which include the standard deviation of the mean oxygen isotope ratio measured on the primary standard during the session and internal error on each sample data point.Corrected δ 18 O (quoted with respect to Vienna Standard Mean Ocean Water or VSMOW) are presented in Supplementary Material Table S1.

Laser-Fluorination Results
Using the laser-fluorination method, the oxygen isotope ratio for the pargasite schist hosted PAR-1 ruby was found to be δ 18 O 1.0‰.This analytical run also included a number of rubies and sapphires from different geological environments.Sapphires from desilicified pegmatites were found The sample mounts were carefully cleaned with detergent, distilled water and ethanol in an ultrasonic bath and then coated with gold (30 nm in thickness) prior to SIMS O isotope analyses.For oxygen isotopic analyses, secondary ions were sputtered from the sample by bombarding its surface with a Gaussian Cs + beam and a total impact energy of 20 keV.The surface of the sample was rastered with a 2.5 nA primary beam over a 15 × 15 µm area.An electron gun was used to ensure charge compensation during the analyses.Secondary ions were admitted in the double focusing mass spectrometer within a 100 µm entrance slit and focused in the centre of a 4000 µm field aperture (×100 magnification).They were energy filtered using a 30 eV band pass with a 5 eV gap toward the high-energy side. 16O and 18 O were collected simultaneously in multicollection mode in Faraday Cup detectors fitted with 10 10 Ω and 10 11 Ω, respectively.Each analysis includes a pre-sputtering over a 20 × 20 µm area during 30 s and the automatic centring of the secondary ions in the field aperture, contrast aperture and entrance slit and consisted of 20 four-second cycles which give an average internal precision of ~0.16 (2 SE).
External reproducibility during the analytical sessions was evaluated by repeating analyses in one single fragment of PAR-1.External reproducibility in this fragment was 0.3 and 0.4 per mil (2SD) during the two analytical sessions.In total, three large fragments of PAR-1 were analysed for their oxygen isotope composition, altogether yielding an average value of 0.9 ± 0.6 per mil (2SD, n = 57, Table 2).Raw oxygen isotope ratios were corrected for instrumental mass fractionation using the δ 18 O composition of PAR-1, which oxygen isotope composition was obtained by laser fluorination method (from the 2009 study).Uncertainty on each δ 18 O spot has been calculated by propagating the errors on instrumental mass fractionation determination, which include the standard deviation of the mean oxygen isotope ratio measured on the primary standard during the session and internal error on each sample data point.Corrected δ 18 O (quoted with respect to Vienna Standard Mean Ocean Water or VSMOW) are presented in Supplementary Material Table S1.

Laser-Fluorination Results
Using the laser-fluorination method, the oxygen isotope ratio for the pargasite schist hosted PAR-1 ruby was found to be δ 18 O 1.0 .This analytical run also included a number of rubies and sapphires from different geological environments.Sapphires from desilicified pegmatites were found to range from 4.8 to 5.0 and rubies from marble-hosted deposits were found to range from 20 to 22 (Table 3).

Secondary Ionisation Mass Spectrometry (SIMS) Results
The oxygen isotope ratios δ 18 O (VSMOW) are presented in Table 2. PAR-5 results show values of −0.31 to 0.85 (0.14 ± 0.24), on average slightly lower compared to PAR-1 results 0.44 to 1.68 (1.00 ± 0.42) even though the two occurrences are only 500 m apart.

Corundum Oxygen Isotopes as An Identifier for Geological Origin
A framework on the interpretation of the geological origin of gem corundums using the δ 18 O ratio proposed by Giuliani et al is now widely adopted [11].Based on this framework, rubies and pink sapphires can be classified into 5 categories based on their δ 18 O value range.
The reconnaissance 2009 laser fluorination results on the sapphires and rubies from different geological environments very closely fits the oxygen isotope value ranges from the model of Giuliani et al.where over 200 corundum samples were analysed under the same method [1].That is the marble-hosted value from 20 to 22 is within the range of 16.3 to 23 and the sapphires from the desilicified pegmatites with a value from 4.8 to 5 fits within the framework range from 4.2 to 7.5 .Therefore, the δ 18 O results obtained using the laser-fluorination method in 2009 are further validated as accurate measurements.

PAR-1 vs. PAR-5 Variations
The oxygen isotope values obtained using SIMS indicates that the Paranesti rubies have a narrow defined band of oxygen isotope signatures with a mean on +1 (ranging from −0.31 to 1.31 ).This is lower than any ratios based on the existing framework for rubies.There are further distinctive constrained values between PAR-1 (+0.65 to 1.31 ) and PAR-5 (−0.31 to 0.85 ).There is a slight overlap of the individual highest value in PAR-5 to the lowest value in PAR-1.The average for PAR-1 is +1 whilst the average for PAR-5 is +0.14 .
There may be some differences between core-rim oxygen isotope values observed in the PAR-5 SIMS results where the core average (−0.02) is lower than the rim average (0.29).However, this is within the range of uncertainty when the errors are taken into account.This narrow range within individual localities and between the two localities that are 500 m apart is in stark contrast to the findings of Bindeman et al (2010) [7] who found variation within single 10 cm samples of up to 3 and variation within single ruby grains of up to 1.5 .It also does not rule-out variances due to partitioning in individual crystals during growth.Therefore, a detailed cathodoluminescence analysis to determine the homogeneity or heterogeneity of the sample grains is suggested for future studies.As the traditional laser fluorination method consumes the entire grain, such subtle zoning would not be seen using this technique.Thus, the greater spatial resolution of the SIMS technique enables us to analyse discrete isotopic domains (i.e., rims, cores, sectors) within single corundum crystals.

Global Low to Ultra-Low Oxygen Isotope Corundum Comparison
δ 18 O (SMOW) values for gem corundums below 1 are very rare and not shown on the original systematic framework by Giuliani et al 2005 [11].Other than the Paranesti rubies shown above, the only negative value for corundums are from Karelia in north-western Russia and sapphire from a secondary deposit in Madagascar.Table 4 lists the global low to ultra-low oxygen isotope analyses for corundums.
The Madagascar sapphire deposit of Ilakaka with δ 18 O of −0.3 to 16.5 is a consolidated placer formed in a sandstone environment.The geological origin of the different ranges of isotopic values found for the sapphires corresponds to at least five different geological environments [26].The low δ 18 O delta values for some sapphires correspond up to an unknown geological sapphire type.
The PAR-1 result of +1.0 is lower than most known primary corundum oxygen isotope values other than the unique ultra-low values of corundums from Karelia [28,30] and one instance of ruby from the Soamiakatra area of Madagascar [26].The Karelia corundum formed under unique circumstances (see discussion below) and can be easily distinguished from the Paranesti rubies.The Madagascar rubies show much higher average δ 18 O values and the minimum value obtained corresponds to the maximum value from Paranesti.Therefore, oxygen isotope analysis is a valuable tool that can be used to fingerprint the Paranesti rubies Figure 6 from other worldwide occurrences.The Madagascar sapphire deposit of Ilakaka with δ 18 O of −0.3‰ to 16.5‰ is a consolidated placer formed in a sandstone environment.The geological origin of the different ranges of isotopic values found for the sapphires corresponds to at least five different geological environments [26].The low δ 18 O delta values for some sapphires correspond up to an unknown geological sapphire type.
The PAR-1 result of +1.0‰ is lower than most known primary corundum oxygen isotope values other than the unique ultra-low values of corundums from Karelia [28,30] and one instance of ruby from the Soamiakatra area of Madagascar [26].The Karelia corundum formed under unique circumstances (see discussion below) and can be easily distinguished from the Paranesti rubies.The Madagascar rubies show much higher average δ 18 O values and the minimum value obtained corresponds to the maximum value from Paranesti.Therefore, oxygen isotope analysis is a valuable tool that can be used to fingerprint the Paranesti rubies Figure 6 from other worldwide occurrences.

Possible Causes for Low Oxygen Isotope Corundum Formation
There are several current hypotheses on how corundums can form with low δ 18 O isotope ratios.These range from hydrothermal alteration of deeply penetrating surface meteoric waters to isotope separation by thermal diffusion during endogenous fluid flow [31,32].

Kinetic Isotope Fractionation
Kinetic isotope fractionation occurs when rapid thermal decomposition of hydrous phases results in isotope disproportionation into a high-δ 18 O residue and a low-δ 18 O fluid [33,34].However, the high-δ 18 O residue material should also be found within proximity of the studied samples for this hypothesis to apply.The water-rock interaction is kinetically restricted in supracrustal rocks and isotope fractionation factors are large at low temperatures, favouring higher-δ 18 O solids [35].In contrast, isotopic exchange is more rapid within a hydrothermal system.As the Paranesti rubies formed under amphibolite-facies conditions above 600 °C [18], significant kinetic isotope

Possible Causes for Low Oxygen Isotope Corundum Formation
There are several current hypotheses on how corundums can form with low δ 18 O isotope ratios.These range from hydrothermal alteration of deeply penetrating surface meteoric waters to isotope separation by thermal diffusion during endogenous fluid flow [31,32].

Kinetic Isotope Fractionation
Kinetic isotope fractionation occurs when rapid thermal decomposition of hydrous phases results in isotope disproportionation into a high-δ 18 O residue and a low-δ 18 O fluid [33,34].However, the high-δ 18 O residue material should also be found within proximity of the studied samples for this hypothesis to apply.The water-rock interaction is kinetically restricted in supracrustal rocks and isotope fractionation factors are large at low temperatures, favouring higher-δ 18 O solids [35].In contrast, isotopic exchange is more rapid within a hydrothermal system.As the Paranesti rubies formed under amphibolite-facies conditions above 600 • C [18], significant kinetic isotope fractionation is highly unlikely and therefore rules out this hypothesis.

Thermal Diffusion
For thermal diffusion, the oxygen in a temperature gradient is redistributed with low δ 18 O at the hotter end and high δ 18 O at the colder end of melt or hydrous solution [31].Akimova (2015) [32] has proposed a model of cascading thermo-diffusion within shear zones to explain the Karelian ultra-low δ 18 O corundums.This scenario would require several individual thermal cells to align in the correct position simultaneously with a similar convection rate and timing.Given that only two locations have shown ruby-bearing pargasite schist with other very similar pargasite boudins nearby being ruby absent, this model appears to be less likely than the hydrothermal scenario.

Other Ultra-Low δ 18 O Protoliths
Ultra-low δ 18 O protoliths could potentially provide the low δ 18 O during corundum syn-metamorphic formation [35].However, a source for the ultra-low δ 18 O protolith would be needed under this scenario such as a low δ 18 O mantle reservoir or previously surface-exposed and then rapidly buried metamorphic rocks.Neither were observed at Paranesti.As oxygen isotope analyses were not performed on the whole-rock and associated mineral phases for the Paranesti occurrences, this hypothesis cannot be ruled out.

Hydrothermal Alteration Model
This hypothesis involves the conservation of the initial isotopic ratios of the protolith in the corundum-bearing rocks and then isotopic exchange between these rocks and meteoric waters before metamorphism [10].Wang et al. [18] demonstrated that the rubies from Paranesti were syn-metamorphic and were largely free of inclusions.However, as shown by Bindeman et al (2010) [7], this does not preclude preservation of the initial ratios within the protolith for the Paranesti occurrence.Therefore, it is unlikely that the low isotopic ratios observed within the Paranesti rubies were due to preservation of the initial ratios within the protolith.
For granulite facies metamorphism, Wilson and Banksi (1983) [36] proposed three processes that could produce a low oxygen isotope value.These are (1) pre-granulite reaction between heated seawater and hot basic intrusives or an initial protolith such as a palaeosol for the sapphirine-spinel-(cordierite) assemblages; (2) syn-granulite depletion in 18 O related to dehydration during granulite metamorphism and removal of the resultant products of partial melting with a depletion in 18 O by up to 2 or 3 for the restite; and (3) post-granulite facies metamorphism with recrystallization under the effect of biotite and/or amphibole-metasomatism with depletion in δ 18 O up to 4 .Based on the previous study [18], the Paranesti rubies were found to have formed under amphibolite facies conditions and there is no evidence that they ever reached granulite facies within the specified zone.However, there are other locations within the Rhodope Mountain Complex (RMC) where regional metamorphism reached granulite facies conditions, though these are some distance away from Paranesti and no rubies are known from these locations.
The glacial meltwater influence during formation of corundums was proposed to explain the ultra-low δ 18 O isotopic ratios observed for corundums from Karelia in north-western Russia [10,12,27].However, there is no evidence suggesting the existence of glaciers in the Mediterranean region based on the reconstruction of the tectonic evolution of the East Mediterranean region since the late Cretaceous [37].Could the RMC be a higher mountain with glaciers that have melted during ruby genesis?There is no evidence in the literature to suggest such and this would only be a remote possibility.Therefore, it is unlikely that a glacial meltwater source played a role during ruby formation at Paranesti.However, it is likely that meteoric water (but not glacial melt) interaction caused by downward flow of surface waters along deep crustal fractures/structures during the formation of the corundum would contribute in producing low δ 18 O values for the Paranesti rubies.

Conclusions
The in-situ SIMS oxygen isotope analyses on the Paranesti rubies is the first time that a primary (and exclusively) ruby occurrence was found to have ~+1 for its δ 18 O-isotope composition.Based on the low δ 18 O value and the local geology, it is most likely that the Paranesti rubies formed under metamorphic/metasomatic conditions involving deeply penetrating meteoric waters along major crustal structures related to the Nestos Shear Zone.PAR-5 is potentially closer to the source of the hydrothermal influence during ruby formation compared to PAR-1 and thus has a lower δ 18 O.Importantly, this study shows that in-situ gem corundum oxygen isotope analysis using the SIMS method may be used to determine the likely geographic origin for corundums lacking any provenance details.Importantly, with the SIMS method being only minimally destructive, with the analysis spot (15 × 15 µm) amenable to repolishing, a wider adoption of this technique has important applications/implications for the international gem and jewellery industry.A future area of research would be to apply this methodology for more gem mineral varieties other than corundum and emeralds.The aim of such future work would be to determine if in-situ SIMS oxygen isotope analysis can be used to both better understand gem formation and to see if it can be used to clearly separate the same gem mineral from different geographic locations.

Figure 2 .
Figure 2. Locality diagram of the ruby occurrences.(a) PAR-1 location on top of the hill.(b) PAR-5 location on the roadside.(c) Pargasite schist boudin found approx.500 m north of PAR-5 without any corundum.(d) Pegmatite on top of the ruby-bearing pargasite schist at PAR-1.

Figure 2 .
Figure 2. Locality diagram of the ruby occurrences.(a) PAR-1 location on top of the hill.(b) PAR-5 location on the roadside.(c) Pargasite schist boudin found approx.500 m north of PAR-5 without any corundum.(d) Pegmatite on top of the ruby-bearing pargasite schist at PAR-1.

Figure 3 .
Figure 3. Trace element discrimination diagrams showing the fields for magmatic, metamorphic and metasomatic corundums, along with the plots for the Paranesti rubies.(a) FeMg vs. GaMg elemental diagram showing the metamorphic vs magmatic fields and SD lines with Paranesti ruby plots.Adapted with permission from Sutherland et al. 2014 [19].(b) FeO + TiO 2 + Ga 2 O 3 vs.FeO-Cr 2 O 3 -MgO-V 2 O 3 elemental diagram showing a metasomatic origin as well as a mafic-ultramafic influence on the Paranesti rubies.Adapted with permission from Giuliani et al. 2014 [20].

Figure 3 .
Figure 3. Trace element discrimination diagrams showing the fields for magmatic, metamorphic and metasomatic corundums, along with the plots for the Paranesti rubies.(a) FeMg vs. GaMg elemental diagram showing the metamorphic vs magmatic fields and SD lines with Paranesti ruby plots.Adapted with permission from Sutherland et al. 2014 [19].(b) FeO + TiO2 + Ga2O3 vs. FeO-Cr2O3-MgO-V2O3 elemental diagram showing a metasomatic origin as well as a mafic-ultramafic influence on the Paranesti rubies.Adapted with permission from Giuliani et al. 2014 [20].

Figure 4 .
Figure 4. Images of ruby samples from Paranesti.(a) Dark red ruby samples from PAR-1 in pargasite schist host rock 0.5-1.0cm; (b) Cluster of pale ruby samples from PAR-5 in pargasite schist host rock 0.5-1.5 cm; (c) Clean PAR-1 ruby sample free from inclusions used for the 2009 Oxygen Isotope analysis.

Figure 4 .
Figure 4. Images of ruby samples from Paranesti.(a) Dark red ruby samples from PAR-1 in pargasite schist host rock 0.5-1.0cm; (b) Cluster of pale ruby samples from PAR-5 in pargasite schist host rock 0.5-1.5 cm; (c) Clean PAR-1 ruby sample free from inclusions used for the 2009 Oxygen Isotope analysis.

Figure 5 .
Figure 5. (a) SIMS in-situ analysis spot location individually marked on the ruby grain-PAR-5 Grain A; (b) SIMS in-situ analysis spot location individually marked on the ruby grain-PAR-5 Grain C.

Figure 5 .
Figure 5. (a) SIMS in-situ analysis spot location individually marked on the ruby grain-PAR-5 Grain A; (b) SIMS in-situ analysis spot location individually marked on the ruby grain-PAR-5 Grain C.

Figure 6 .
Figure 6.Oxygen isotopic comparison of rubies from Paranesti occurrences compared with low δ 18 O corundums from Karelia in northwestern Russia and Soamiakatra in Madagascar.

Figure 6 .
Figure 6.Oxygen isotopic comparison of rubies from Paranesti occurrences compared with low δ 18 O corundums from Karelia in northwestern Russia and Soamiakatra in Madagascar.

Table 2 .
Oxygen isotope results from 2017 using the SIMS method.

Table 3 .
Oxygen isotope results from the 2009 reconnaissance using the laser-fluorination method, n = 1.

Table 4 .
Global comparison of corundums with low oxygen isotope values.