Rapid oxygen diffusion during high temperature alteration of zircon

The mineral zircon through its isotopic and elemental signatures comprises the greatest archive recording the evolution of Earth’s continental crust. Recognising primary from secondary zircon compositional signatures is thus important for the accurate interpretation of this archive. We report two examples of metasedimentary rocks from high-grade shear zones within the Southern Granulite Belt of India, where anomalously high and homogeneous oxygen isotope signatures indicate disturbance of this isotopic system. Utilising the combined U-Pb-Hf-O and trace element signatures from these zircon grains, we postulate that fluid-assisted alteration has led to complete resetting of the oxygen isotope signatures. This case study presents a rarely observed natural example of potentially fast diffusion of oxygen under hydrous conditions. Given the pervasive nature of fluid interaction within high-grade and highly deformed rocks, we expect that such isotopic disturbance might be more common to nature than is currently reported. A lack of correlation between isotopic disturbance with cathodoluminescence or Th/U values, suggests that these altered zircon grains would not clearly be classified as metamorphic, in which case they would be expected to yield primary compositions. Caution is therefore advised when using detrital δ18O zircon compilations without a high level of scrutiny for primary versus secondary compositions.

and lacking any imprints of the later Neoarchaean, Palaeoproterozoic and Neoproterozoic tectonothermal events that are widely represented in the surrounding crustal blocks in the SGT 21 . The margins of the Coorg Block and the bounding suture zones have been the focus of recent studies in preserving some of the oldest crustal remnants in Peninsular India 22 , and as the zone along which Mesoarchaean high grade metamorphism occurred, possibly associated with collisional orogeny 26 . The adjacent crustal blocks, Nilgiri, Biligiri Rangan, Salem and Madras (Fig. 1a), preserve common Neoarchaean crustal formation/recycling and metamorphism e.g. [27][28][29][30] . However, the Coorg Block is believed to be an exotic crustal domain within the SGT, in the absence of any records of the 2.5 Ga pervasive regional metamorphism 21,22 .
The Coorg Block is bound by the Moyar Shear Zone 21,31 to the south, and the Mercara Shear Zone 32 to the north (Fig. 1b). The Mercara Shear Zone formed from the accretion of the Coorg Block to the Western Dharwar Craton, which is inferred to have occurred at 3.0 Ga based on zircon metamorphic ages 26 . Steep gravity gradients have been imaged beneath the Mercara Suture Zone, and are considered to mark underplated high-density mafic lower crust that has welded the Coorg Block with the Dharwar Craton 33 . Peak metamorphic conditions recorded in Mercara Shear Zone granulites are estimated at 10-12 kbar at 700-900 °C 26 .
The Moyar Shear Zone marks the boundary between the Nilgiri and Coorg Blocks in the west, and the Nilgiri Block with the Dharwar Craton in its eastern extent. In the Wynad region at the confluence of the Mercara and Moyar Shear Zones, peak pressure temperature conditions have been estimated in metagabbros at 10.6-11.8 kbar at 900 °C, with the timing of this metamorphism estimated at 2. 36-2.22 Ga, and the origin of this metamorphism being ascribed to suturing of the Coorg and Nilgiri Blocks 34 . Ultramafic rocks in this region record peak metamorphism of 6.5-8.7 kbar at 720-840 °C 35 . Further along strike to the southeast, peak conditions are recorded in garnet-bearing quartzofeldspathic gneiss at 9-11 kbar at 900 °C 36 . These authors present a model whereby the Moyar Shear Zone is part of a doubly vergent orogen marking the suturing in this region of the Nilgiri Block and the Dharwar Craton.
Metasedimentary belts comprising pelite-rich and quartz-rich lithologies are found within both the Mercara and Moyar Shear Zones. TNL-1-1 is taken from a set of nearly E-W trending laminated, pale red to white quartz mica schist bands, ranging in width from several tens up to 500 m. The protolith of these rocks likely correspond to ferruginous sandstone. Several thin bands of medium to fine-grained, finely laminated, pale green to yellowish white quartz mica schist intercalated with calcareous schist are also exposed around the Tirunelli temple area. Weathered quartz mica schist interbanded with fuchsite quartzite is exposed on the western side of Mambaram, and continues in the Kuppadi reserve forest NW of Vengur. A small hillock with fine laminated garnet-bearing brownish quartz mica schist occurs at Neduvaloor along the Irikur-Thaliparamba road. Sample NDV-2 is taken from here, where quartz mica schist occurs in association with fuchsite and micaceous quartzite, talc-tremolite schist, meta-BIF, fuchsite quartzite and amphibolite. Quartz mica schist in association with greenish fuchsite quartzite and medium grained white quartzites are also exposed in other locations in the area including Tirunelly, Manandawady and Sulthan Bathery. Metamorphic garnet, rutile and paragonite within quartz-mica-schists 22 attest to them experiencing at least amphibolite-facies metamorphism. The peak temperature of one sample studied here (NDV-2) was previously estimated at 710-760 °C 22 , and sillimanite in some quartz-muscovite schist units suggests pressures up to 8 kbar 22 .

Results
U-Pb. TNL-1-1 exhibits an array of generally discordant data (Fig. 4), with 207 Pb/ 206 Pb ages of sub-concordant (>90% concordance) analyses ranging from 2386 to 3060 Ma. There is no precise lower intercept obtainable, since the data are scattered, reflecting variable discordance from a range of ~2.4 to 3.0 Ga primary ages. A chord drawn between the lowest concordant and most discordant ages regresses to a lower intercept of 802 ± 86 Ma with an upper intercept of 2497 ± 42 Ma. There is no particular correlation between age and Th/U contents; sub-concordant ages have a range of both 'magmatic' (>0.1) and 'metamorphic' (<0.1) values (Fig. 5a). NDV-2 exhibits an array of concordant to sub-concordant ages ranging from 2735 to 3212 Ma (Fig. 4), and clustering into two groups at ca. 2750-2900 Ga and ca. 3000-3150 Ga. There is no pattern linking Th/U content with age ( Fig. 5a). δ 18 O. Oxygen isotope data are plotted against apparent age ( 207 Pb/ 206 Pb) in Fig. 5b. Analyses in TNL-1-1 were made in zircon grains with both concordant and heavily discordant ages ranging from 2992 to 1439 Ma. The data form a horizontal array with a very limited range of δ 18 O of 8.98 to 9.42‰. NDV-2 exhibits consistently higher values, with a range from 10.51 to 11.65‰. The oldest analysis from NDV-2 forms an outlier with a lower δ 18 O of 9.25. Data from both samples are much higher than mantle values (5.3 ± 0.3‰) 37 ; for magmatic zircon these data at face value imply a contribution to the crystallising melt of low-temperature altered (i.e. supracrustal) material 7,38 .

Lu-Hf.
Hf isotope data are plotted as apparent zircon age ( 207 Pb/ 206 Pb) against the age-corrected 176 Hf/ 177 Hf ratio in Fig. 5c. TNL-1-1 shows a horizontal array extending from ca. 3060 to 1440 Ma at ca. 0.2807 that can be interpreted as a lead-loss trend (Lu/Hf = 0). A second cluster has apparent ages around ca. 2600 to 2350 Ga, also with discordant analyses falling on a lead-loss trend from this population. Two analyses from TNL-1-1 have higher initial 176 Hf/ 177 Hf ratios (0.28122 and 0.28175), with one ca. 1650 Ma analysis being a distinct outlier. NDV-2 has a single horizontal array ranging from ca. 3200 to 2700 Ma, with a cluster around 3.1 Ga that has an average composition just lower than that of CHUR. The array is compatible with a lead-loss trend. A single outlier falls above CHUR at 2.7 Ga, and at face value would represent a more juvenile composition outside of that expected from reworking of the 3100 Ma population alone.
Trace elements. Titanium contents in zircon can be used as an indicator of the temperature of the melt during zircon crystallization 11,39,40 . This is most accurate in magmatic and metamorphic rocks where the activity of Ti can be assessed through the presence of Ti-phases such as rutile. In detrital rocks, as in this study here, the temperatures can only arbitrarily be used, as the Si and Ti activities during crystallisation that are required for the temperature calculation are not known. Using typical Si and Ti activities of 1 and 0.7, respectively, the Ti-in-zircon temperatures range from 767 to 853 °C for TNL-1-1 and from 708 to 959 °C for NDV-2. These temperatures are higher than the average compositions of both mafic and felsic rocks 12 . However, it should be noted that the measurement uncertainty on the Ti contents leads to an average temperature uncertainty of 30 °C, and that drastically changing the Si and Ti contents can lead to an up to 70 °C lowering in the temperature.
Chondrite-normalised Dy/Yb (i.e. MREE/HREE) and REE patterns are shown in Figs 6 and 7, respectively. The data for TNL-1-1 are tightly spread, and show distinct positive Ce and negative Eu anomalies that are typical for zircon formed in both magmatic settings and amphibolite-to granulite-facies metamorphic conditions. The MREE to HREE are relatively flat for all analyses, which is typical of granulite-facies zircon grown in the presence of HREE-sequestering minerals, primarily garnet 41,42 . Only one sample is an outlier, and lacks an Eu anomaly and also displays less depleted HREE. Sample NDV-2 has greater scatter in the data, but overall with similar patterns. A few analyses lack the HREE depletion, and have steeper MREE-HREE patterns typical of magmatic zircon. The MREE to HREE ratio, i.e. a monitor of HREE depletion typical of high-grade metamorphic zircon 42 , and plotted as chondrite normalized Dy/Yb, is shown against apparent 207 Pb/ 206 Pb age in Fig. 6. There is no correlation between HREE depletion and age, irrespective of the discordance of the age.
Light rare earth element (LREE) contents can be used to discriminate hydrothermal zircon from magmatic zircon, as LREE are typically enriched during hydrothermal alteration 43 . A plot of chondrite-normalised La versus Sm/La (Fig. 8) shows that the bulk of the data do not cross into the hydrothermal field.

Discussion
For several reasons outlined below, we interpret the combined isotopic and elemental zircon data presented to reflect alteration and disturbance of zircon, particularly of the oxygen isotope system, rather than primary signatures formed during crystallization. First, we list the various lines of evidence, and this is followed by an evaluation of the potential processes responsible and the constraints on the timing of these processes.
The following lines of evidence indicate alteration of the zircon after their deposition as sedimentary units: 1) Significant discordance in the U-Pb system for TNL-1-1 is indicative of lead-loss, the timing of this being loosely constrained to early Neoproterozoic.   2) A lead-loss trend in U-Pb-Hf space (Fig. 5b), combined with concordant U-Pb data, implies an old (Neoarchaean-Palaeoproterozoic) lead-loss event affecting NDV-2. 3) Oxygen isotope data are remarkably consistent across each sample, and much higher than that predicted for zircon grains of this age (see Fig. 5b). Detrital zircon grains are predicted to exhibit variable δ 18 O signatures reflecting the variability among their source rocks, but this is not the case for both samples. 4) REE profiles are remarkably consistent for TNL-1-1, especially considering the large age range of ages, and show flat MREE-HREE implying the signature was imparted during conditions with growth of other HREE-bearing phases, typical of granulite-facies. NDV-2 is more scattered, with a few magmatic signatures, but mostly showing the same flat MREE-HREE pattern. Detrital zircon grains are predicted to have variable patterns reflecting their source regions, and would likely comprise both magmatic and metamorphic signatures.
After establishing that the zircon grains exhibit some alteration to their isotopic signatures, it is pertinent to ask the question, would these zircon grains be classified as metamorphic based on their appearance or Th/U ratios? In CL, sample TNL-1-1 has consistent weak zoning with only faint oscillatory patterns. The analysed domains are not from clear overgrowths but typically from central domains. The optical images (see Supplementary Information), indicate that no oxygen isotope analyses were located near cracks which can act as potential alteration pathways, and only two samples are clearly metamict. Th/U ratios are mixed, with the minority having classical metamorphic values of <0.1. Sample NDV-2 has a wider variety of CL textures, with some analyses being located on strongly oscillatory zoned domains that would be classified as magmatic. Optical imagery shows that two analyses were located on cracked domains, and roughly half the zircon grains analysed are potentially metamict. Th/U ratios are again mixed, with less than half being from metamorphic values. Thus, we can conclude that although there is some evidence in CL for TNL-1-1, and optically for NDV-2, overall the zircon grains would not generally be classified as metamorphic and not suffering from strong metamictisation. The importance of this will be returned to in the concluding remarks.
Oxygen isotopes, due to their slow diffusion in dry conditions 2,3 are thought to be robust within zircon, even after high-grade metamorphism, and this has been evidenced by several studies 8,19,44 . The exception to this is zircon that has been altered by hydrothermal fluids, in this case, newly grown zircon can deviate from the primary signature, and trend towards that of the metamorphic fluid 45 . For this reason, studies that wish to observe only primary signatures will tend to exclude data that result from cracked or pitted domains, from zircon grains that have discordant U-Pb age data or from zircon with metamorphic textures. There is debate over the use of wet or dry diffusion data for zircon 42,46 , because most natural data have pointed to slow diffusion more compatible with dry diffusion rates 5,8,10 , but experimental data have shown that diffusion under high water activity would be much faster 2,3 . However, a recent study 46 from the Ukrainian Shield interpreted a set of high δ 18 O analyses as due to modification under high temperatures (~700 °C) in hydrothermal conditions, advocating the fast rates of diffusion under wet conditions. The non-metamict nature of zircon grains of that study, the high δ 18 O, and the Archaean ages that distinguish them from global datasets, is similar to the zircon data we present here. The authors 46 recognised zircon with modified oxygen isotope compositions which were altered by a process independent of crystallinity, state of metamictisation and CL texture, and that had caused limited Hf isotope disturbance. Given the similar characteristics of the zircon systematics in the present study, we consider the δ 18 O data of this present study at face value to indicate a similar case of fluid-assisted resetting of oxygen isotope signatures.
The U-Pb, Hf-isotope and elemental signatures can be used to further constrain this postulated fluid-assisted alteration. Ti-in-zircon temperatures of both samples are typical of amphibolite-to granulite-facies conditions. Given the uncertainty over the accuracy of these data without evidence for equilibrium during Ti-incorporation, equilibrium with a Ti-phase, and the unknown pressure during metamorphism 47,48 , along with the unknown nature of the primary zircon signatures, we cannot be certain whether these reflect alteration of the Ti contents.
REE patterns for both samples are mostly typical of granulite-facies zircon, i.e. zircon that has low HREE due to its sequestration in HREE-bearing phases such as garnet. Previous estimates on the metamorphic history of these metasedimentary belts suggest conditions up to granulite-conditions 22,26,34,35 , and the peak temperature of NDV-2 was estimated at 720-760 °C 22 . Diffusion of REE is slower than that of oxygen, and thus we may expect primary signatures to be preserved through metamorphism if the zircon is not recrystallised or reset. However, because of the general homogeneity of the REE signatures exhibited here, we consider them to be potentially obtained during subsequent metamorphism. In this case, the HREE depletion in comparison to magmatic zircon could thus be related to two mechanisms: (1) fluid-assisted resetting of the zircon during metamorphism under granulite-facies conditions, with HREE sequestering into other phases such as garnet; or (2) fluid-assisted alteration disturbed the REE unequally so that HREE were lost preferentially to LREE and MREE from the zircon. The mineral assemblage of both samples (muscovite + quartz ± chlorite ± garnet ± sillimanite ± rutile), and the potential granulite-facies temperatures recorded by both samples, indicate that the first option is viable. However, because HREE diffuse faster than LREE 13 , the second mechanism cannot be excluded.
As with the Ukrainian case study 46 , we conclude that disturbance of the oxygen isotope signatures and potentially the REE contents is likely to have occurred via rapid oxygen diffusion in hydrous conditions during high-temperature metamorphism. The timing of this metamorphic event can partly be tied down by the U-Pb-Hf data. The oxygen isotope disturbance affects zircon of all ages; for TNL-1-1, this disturbance therefore must be younger than the maximum depositional age of ca. 2.5 Ga, and is likely to have occurred during the strong lead-loss inducing event in the Neoproterozoic. For NDV-2, the lead-loss array in time-Hf space suggests that the 2.9 Ga ages are not primary, but record lead-loss from 3.0-3.1 Ga ages. Such lead-loss parallel to concordia is typical of Achaean high-grade rocks 49 , and makes it hard to date events precisely. The alteration can only be constrained in this case to equal or younger than 2.9 Ga. Since the two samples come from different shear-zones, there is potential that the fluid-assisted alteration occurred at different times within each sample. Therefore the 2.36 to 2.22 Ga metamorphism recorded in the nearby region 34 is potentially responsible for altering sample NDV-2. Both samples reveal some retrogression to hydrous greenschist facies mineral assemblages. This presents a later post-peak metamorphic event that could potentially be related to zircon alteration; however, given the expected behaviour of oxygen isotope diffusion 2,3,5 , we suggest that this is unlikely and that earlier high temperature alteration is more appropriate.
In summary, we postulate that zircon from two different localities has experienced fluid-assisted oxygen isotope diffusion during high temperature conditions in the shear-zones surrounding the Coorg Block. Studies that have shown rapid fluid-assisted oxygen isotope diffusion are scarce 46 , and therefore this work adds weight to the argument that wet diffusion rates may be more applicable to zircon in some settings. Given that high-grade metamorphism and deformation are both typically associated with fluid circulation, it is perhaps surprising that such observations from nature have not already been made. Deciphering the exact processes of alteration, i.e. by wet diffusion alone, by micro-fractures or micro-porosity, or by elemental substitution, will require much higher spatial resolution data obtainable by methods such as Atom Probe Tomography 50,51 .
It is important to recognise the potential for disturbed zircon isotopic systematics in detrital zircon, because many studies use archives of mineral isotopic and/or elemental data to study Earth evolution [16][17][18][19][20]52 . Such studies may involve, for example, filtering of oxygen data to obtain primary zircon signatures by various means: (1) removing discordant U-Pb analyses; (2) removing analyses with metamorphic (<0.1) Th/U ratios; (3) removing metamorphic-zoned domains by CL imaging; (4) removing physically cracked zircon domains; and (5) removing hydrothermally altered (high LREE) domains. However, in the case study presented here, such filtering will not have removed the entire set of high-δ 18 O zircon grains. Thus, this study presents a cautionary tale against the use of detrital δ 18 O zircon compilations, where a high level of scrutiny of individual datasets is typically not achieved, and where deciphering primary from secondary characteristics is hampered by the lack of host rock constraints.

Methods
Zircon oxygen isotope studies were conducted using a Cameca IMS1280 ion microprobe housed at the Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing. Detailed working conditions and analytical procedures are the same as those described by 53 . The Cs + primary ion beam was accelerated at 10 kV, with an intensity of ca. 2 nA and rastered over a 10 μm area. The spot is about 20 μm in diameter. The instrumental mass fractionation factor (IMF) was corrected using Penglai zircon standard with (δ 18 O) VSMOW = 5.31 ± 0.10‰ 54 . Measured 18 O/ 16 O was normalized by using Vienna Standard Mean Ocean Water (VSMOW) compositions, and then corrected for the instrumental mass fractionation factor. The internal precision of a single analysis was generally better than 0.2‰ for 18 O/ 16 O ratio. Values of δ 18 O were standardized to VSMOW, and the δ 18 O values are reported in standard per mil notation U-Pb geochronology, Lu-Hf isotope analyses and trace element analyses were conducted at the NERC Isotope Geoscience Laboratory (NIGL) in Nottingham, UK. U-Pb geochronology was measured using a Nu Attom single collector inductively coupled plasma mass spectrometry coupled to a New Wave Research (ESI) 193UC excimer laser ablation system, following procedures documented in 55 . Lu-Hf isotope analyses were conducted using the same laser coupled to a Thermo Scientific Neptune Plus multi-collector ICP-MS, following procedures outlined in 56 . Trace elements, including Ti contents, were conducted using the same set-up as for U-Pb. Procedures are modified from those previously described 57,58 . Data were collected in the following order: δ 18 O (ion microprobe), U-Pb (LA-SC-ICP-MS), Lu-Hf (LA-MC-ICP-MS), and trace elements (LA-SC-ICP-MS). All analyses were placed on top of each other, except for trace element pits, which were placed in the same zircon domain, as imaged by CL, if space allowed.
All data generated or analysed during this study are included in this published article (and its Supplementary Information files).