Extreme lithium isotopic fractionation in three zircon standards (Plešovice, Qinghu and Temora)

To understand the behavior of Li in zircon, we have analyzed the abundance and isotopic composition of Li in three zircon standards (Plešovice, Qinghu and Temora) widely used for microbeam analysis of U-Pb ages and O-Hf isotopes. We have mapped Li concentration ([Li]) on large grains, using a Cameca 1280HR Secondary Ion Mass Spectrometer (SIMS). All zircons have a rim 5–20 μm wide in which [Li] is 5 to 20 times higher than in the core. Up to ~20‰ isotopic fractionation is observed on a small scale in the rims of a single zircon grain. The measured δ7Li values range from –14.3 to 3.7‰ for Plešovice, –22.8 to 1.4‰ for Qinghu and –4.7 to 16.1‰ for Temora zircon. The [Li] and δ7Li are highly variable at the rims, but relatively homogenous in the cores of the grains. From zircon rim to core, [Li] decreases rapidly, while δ7Li increases, suggesting that the large isotopic variation of Li in zircons could be caused by diffusion. Our data demonstrate that homogeneous δ7Li in the cores of zircon can retain the original isotopic signatures of the magmas, while the bulk analysis of Li isotopes in mineral separates and in bulk-rock samples may produce misleading data.

In this study, we have analyzed Li isotopes (δ 7 Li) and abundance ( [Li]), in both whole-rock samples and in situ in zircon, and have imaged the distribution of seven trace elements (Li, P, Ti, Yb, Y, U, and Th), in three well-characterized zircon standards commonly used for U-Pb and Hf-O isotopic analysis. Our aim is to shed light on the Li-isotope characteristics of these zircons, to find out whether variations in the Li isotopes reflect differences in source values and/or is caused by later diffusion. This is fundamental to evaluating the feasibility of using Li-isotope analysis of whole-rock samples and zircon as a geochemical tracer to identify the source regions of magmas.

Sample Information
Three standard zircons (Plešovice 9 , Qinghu 10 and Temora 9,11,12 ) and have been chosen for this study, as they (i) are homogeneous and concordant in terms of radiogenic Pb/U ratios, (ii) are homogeneous in Hf and O isotopic composition, (iii) have crystalline (nonmetamict) structures, and (iv) occur as mm-sized grains. The M257 zircon [13][14][15][16] is used as zircon Li-isotope analysis standard. The isotopic characteristics of the zircons are summarized in Table 1.
Plešovice. Plešovice zircons were extracted from a high-temperature potassic granulite in the southern Bohemian Massif of the Czech Republic. The content of zircon in the granulite is up to 0.6 wt%, and it is variable on the scale of tens of centimeters. The Plešovice zircon is homogeneous in terms of U-Pb age and Hf isotopes, and thus is widely used as a standard for microbeam isotopic analyses 9 . Zircon grains are prismatic, pale pink to brown euhedral crystals and have been tested for Li-isotope homogeneity in a previous study, which showed highly heterogeneous Li-isotope compositions (~10‰ variation in δ 7 Li) 16 . Two grains (maximum lengths of 2094 μ m and 1770 μ m, respectively) were selected and mounted to acquire detailed diffusion profiles.
Qinghu. The Qinghu monzonite is located in the southwestern part of the Nanling Range in south China 17 . The rock types of the pluton are coarse-grained hornblende monzonite, medium-grained quartz-bearing hornblende monzonite and fine-grained quartz monzonite 10 . Zircons from the Qinghu monzonite are euhedral, range mostly from 100 to 300 μ m in length, and have length/ width ratios mostly between 1:2 and 1:4. Most zircons show a clear euhedral concentric zoning in cathodoluminscence (CL). Previous studies indicate that Qinghu zircon is quite homogeneous and can be used as a working reference standard for SIMS U-Pb and Hf-O isotopic analysis 10 . Temora. The Temora zircons were extracted from the Middledale Gabbroic Diorite in the Tasman Oragenic Belt, Australia. The rock is medium-to coarse-grained, relatively unaltered and composed dominantly of labradoritic plagioclase and brown pargasitic hornblende (partially replaced by chlorite) 18 . Zircons are mostly euhedral, and 100-550 microns in length. Temora zircons have been widely used as standards for U-Pb age and Hf-O isotope microanalyses 9,11,12,19 .

Results
Zircons with euhedral crystal shapes were analyzed for both Li isotopic composition and element distribution. The abundance and isotopic composition of Li were measured along 12 traverses across seven zircon grains (2 Plešovice, 4 Qinghu and 1 Temora). Nine ion-imaging analyses were used to partially map the zircon grains. Traverses of 5-15 analytical spots at 10-200 μ m intervals were performed across selected grains in order to investigate the degree of homogeneity. A smaller primary ion beam (∼ 5 μ m) was used to conduct a more thorough study of the rim phenomena. Detailed information for all the measurements is listed in the Appendix dataset. Generally, all of the zircon grains have a 5-20 μ m wide rim with extremely high [Li], about 5 to 20 times higher than the core. There also is extremely large isotopic fractionation (~20‰) at the zircon rims on a small scale (~50 μ m).
Negative correlations between [Li] and δ 7 Li were encountered at the rim of the grains. Variability in δ 7 Li, related to cracks with higher Li contents, was also observed and these spots were eliminated.
Plešovice. Four whole-rock samples of Plešovice have been analyzed ( R) have been studied in detail. ~20 μ m, ~10 μ m and ~5 μ m beam spots were used to traverse each grain. [Li] profiles across the two megacrysts are nearly flat at the center, but rise dramatically at the rim (Figs 1  and 2). The overall [Li] [Li] vs relative distance to the rim. Uncertainty of standard analysis (2SD) has been shown in the plots while 2SE of individual analyses is smaller than 2SD. Typical analytical uncertainty of δ 7 Li is ∼ 2‰ for the 20 μ m pits, ~2.5‰ for the 10 μ m pits and ∼ 5‰ for 5 μ m pits (2 SD). Distance to rim is calculated by the X-Y position of the analytical spot to the nearest rim. Fractionation of Li isotopes is up to 18‰ in a single grain. Both Plešovice megacrysts show a plateau of constant δ 7 Li in their cores, but striking heterogeneity at the rims. For the Plešovice-R zircon ( Fig. 1), [Li] ranges from 0.5 to 9.5 ppm and δ 7 Li ranges from − 14.1 to + 3.7‰. The δ 7 Li values in the rim of Plešovice-R show very complex spatial variation; therefore, ~10 m and ~5 μ m beam spots were used to map the changes in Li isotopes over small areas. The ion image is a visual display of changes in [Li] and [Y] at the rim. Both [Li] and δ 7 Li are uniform at distances of > 100 μ m from the rim (Fig. 1). The Plešovice-L zircon (Fig. 2) has [Li] ranging from 0.5 to 14.4 ppm and δ 7 Li from − 14.3 to + 3.7‰. The ion image displays [Li] and [Y] at a broken rim. Profile AA' along the zircon demonstrates that the Li isotopes change across the entire zircon. [Li] shows a rim-ward increase, while δ 7 Li is negatively correlated with [Li] and characterized by a typical diffusion profile. Both [Li] and δ 7 Li become uniform at distances of > 100 μ m from the rim (Fig. 2). Compared to Plešovice-R, Plešovice-L has a more euhedral shape and fewer cracks, and δ 7 Li variations at the rim of Plešovice-L are less complex than in Plešovice-R. Qinghu. The whole-rock [Li] of the Qinghu monzonite is ~23 ppm; δ 7 Li values range from 0.9 to + 2.5‰. Zircon grains have been traversed using both ~20 μ m and ~10 μ m beams. [Li] ranges from 0.3 to 11.8 ppm and δ 7 Li values range from − 22.8 to + 1.4‰ (Table 3). Compared to the Plešovice samples, Qinghu zircons have higher [Li] in the cores, mostly around 1-2 ppm, and extremely high [Li] in their rims, 6-15 times higher than in the core.
Across the Qinghu_1 zircon, [Li] ranges from 0.3 to 11.8 ppm and δ 7 Li from − 18.3 to + 1.4‰. The ion image indicates that [Li] and [Y] are to some extent correlated. It also shows higher [Li] near cracks while [Y] remains unaffected. The [Li] of Qinghu_1 increases rimward. The δ 7 Li values in the rim, obtained using the ~20 μ m and ~10 μ m beam spots, are negatively correlated with [Li] and define a typical diffusion profile (Fig. 3).
Another three Qinghu zircons were traversed using a ~10 μ m beam spot. Overall, their [Li] shows a rimward increase and becomes relatively uniform at distances from rim of > 60 μ m (Fig. 4). Generally, the δ 7 Li values are higher and more homogeneous in the core than at the rim.
Temora. The whole-rock [Li] of Temora is ~11.4 ppm, and δ 7 Li values range from 0.7 to 1.5‰. The Temora zircon grains have been traversed using 20 μ m and 10 μ m beam sizes. All the data are shown in the Appendix dataset. Ion images show the distributions of Li, P, Y, Yb (Fig. 5). Y and Yb are positively correlated with [Li] while P is quite homogenous from the rim to core.
Temora zircons have quite low Li contents, which leads to larger analytical errors. For all twenty-four analytical spots, [Li] ranges from 0.02-6.84 ppm and δ 7 Li from − 4.7 to + 16.1‰. There is a typical diffusion profile with gradually decreasing [Li] and increasing δ 7 Li from rim to core (Fig. 5c,d). Except for the spots with very low [Li], the δ 7 Li is relatively homogeneous (with a weighted average of 7.0 ± 1.8‰) in the centers of grains at least 150 μ m across.

Discussion
To date, only a few studies on the isotopic composition of Li in zircons have been reported 2,8,16,20 . The results of these studies are by no means conclusive with respect to either diffusion fractionation, or isotopic fractionation between zircon and melt or zircon and surrounding minerals. Significant Li-isotope heterogeneity occurs on both intra-and inter-granular levels. There is up to ~20‰ isotopic fractionation within ~50 microns of the rim of a single grain (Figs 1-4). Extreme Li-isotope fractionation (~35‰, ~48‰) also has been observed in olivine 21,22 and orthopyroxene (~40‰) 23 . Such extreme fractionation is thought to be produced by disequilibrium fractionation. The processes responsible may include metasomatism of mantle rocks by silicate and carbonatite melts and hydrous fluids 21 , preservation or modification of fluid/melt interactions in recycled oceanic crust 22 , or diffusion-driven fractionation and emplacement-related alteration coupled with isotopic exchange 23 .  Our data display the largest intra-granular fractionations (~20‰) found in published zircon Li-isotope data. To confirm the robustness of the data at the rims, we examined the Li isotopes in the epoxy and in the rim of the sloped surface of the M257 standard in order to check the following possibilities: 1) contamination by the epoxy; 2) analytical artifacts from sample geometry and topography [24][25][26][27] . All the analysis spots are within the zircon grains, so that the epoxy has no effect on the analysis of the zircon rims. Moreover, none of the [Li] and isotopic data of zircons, olivines and pyroxenes analyzed previously in our lab show any effect of epoxy on the rim analyses of mineral grains 16,21,22,[28][29][30] .
The charge effect was carefully avoided during the analysis. The sample mount has been taken out and re-coated if the analytical spots were too close to each other. Moreover, data have been screened by ion yield and analytical errors. The high- [Li] zones at the rim are not observed in the M257 standard zircon, which is a fragment, and are not observed where the zircon grain is bounded by a broken surface. It is only seen at growth surfaces (crystal faces). Since both types of grain boundary stand up from the epoxy mount, the high- [Li] zones cannot be ascribed to a topographic effect; therefore, they represent processes affecting the original boundary between the zircon crystal and its host rock. Li is also concentrated in fissures and interstices of the zircon grains. Considering the high Li abundance in the corresponding whole-rocks, the high [Li] and extreme δ 7 Li variations at the zircon rims are most likely caused by a gradient in [Li] via diffusion of Li from Li-rich mineral phases on cracks and surfaces, and/or a fluid phase with high δ 7 Li. The diffusive fractionation of Li isotopes can occur during cooling or in relation to overgrowth, which is controlled by variable mineral-melt and mineral-mineral partitioning of Li. The zircon standards have good crystalline structure and are homogeneous in terms of radiogenic Pb/U ratios and Hf-O isotopic composition. It is reasonable to conjecture that whole zircon grains were most likely homogeneous in Li isotopes during their crystallization, and in equilibrium with the host magma. The Li isotopes of zircon rims thus were affected by late diffusion-driven processes. This process might have been a "two-stage diffusion": 1) Diffusion between melt and zircon: zircon crystallized in equilibrium with magma, then Li diffused from the melt into zircon grains during magmatic cooling, driven by the [Li] gradient and the increasing partition coefficient of Li between zircon and melt; 2) Diffusion between solid mineral phases: after the melt fully crystallized, Li could diffuse from the adjacent minerals with high [Li] (feldspar/biotite/quartz, etc.) into zircon grains.
The preservation of the original Li isotopic data in zircon will depend on the cooling rate and the size of zircon grains 8 . Slow cooling rates might greatly modify the original "equilibrium" Li isotopes. In spite of this, if the zircon grain is large enough or the diffusion period is short enough, the Li isotopes in the core will retain the pristine δ 7 Li signature of the melt from which the zircon crystalized. For large zircon grains like Plešovice-R and -L, the δ 7 Li in the central parts of the grains shows a plateau of constant δ 7 Li which can represent the original δ 7 Li values of the zircon. Figure 6 presents the compiled data of all the analyzed spots on the Plešovice zircons. The analyzed spots more than 100 μ m from the rim show a good Gaussian distribution, with weighted average δ 7 Li = 1.15 ± 0.15‰ (1σ ). This should be the best estimate of the pristine Li-isotope composition of the Plešovice zircon. The Plešovice granulite was derived from the mature crustal sources by Sr-Nd 31 and Hf-O isotopes 15 and is consistent with zircon δ 7 Li.
It is noteworthy that Li isotopes from large zircon cores retain the initial Li signature, but they are more or less different from the whole-rock δ 7 Li. The cores of Plešovice zircon have higher δ 7 Li values ( 0.1‰ to 1.4‰) than the whole-rock (− 1.6‰ to − 0.6‰) while the δ 7 Li of most Qinghu zircon cores (− 6.4‰ to − 3.6‰) is lower than the whole-rock values (0.9‰ to 2.5‰). Studies of Li-isotope fractionation in peridotites and basalts show little fractionation of Li isotopes at high temperatures during partial melting and equilibrium fractionation 20,23,32 . Bouvior (2007) 20 assumed that Li-isotope fractionation between melt and zircon is less than 2‰. Therefore, the original δ 7 Li value of whole-rock is likely to be similar to the original value in zircons. However, as Li is a highly mobile element and has high diffusivity, the initial Li-isotope signature of the whole-rock is unlikely to survive through later geological processes such as fluid/melt interaction, high-grade metamorphism and hydrothermal alteration.  Therefore, we propose that the bulk-rock data are misleading because they are a weighted average of Li in all rock-forming and accessory minerals, and could reflect the integrated effects of multiple geological processes. The inconsistent δ 7 Li results between whole-rocks and zircon cores are most likely indicative of later processes, for example fluid addition after zircon crystallization, which could shift the Li isotopic composition of whole rock.
As the Plešovice, Qinghu and Temora zircons are homogeneous in terms of U-Pb ages and O-Hf isotopes, one can assume their parental melts remained homogeneous and no separate fluid/melt was added during the crystallization of the zircons. Therefore, the plateau of [Li] and δ 7 Li in the core of large Plešovice zircons could reflect the pristine Li-isotope composition of the magma from which zircon crystallized. Unlike the equant Plešovice zircons, the Qinghu zircons are elongated and have length to width ratios of ~3:1 to 4:1. The data were collected along long axis of the zircon while the plateau is obtained more than 60 μ m from the rim, which is larger than the half-width of the short axis. Therefore, the plateau δ 7 Li data of the Qinghu zircon is more likely to be compromised by Li diffusion parallel to the short axis and cannot represent the primary zircon δ 7 Li. Moreover, the Hf-O isotopes of the Qinghu zircon suggest it originated from a depleted-mantle source with little crustal contamination and it therefore should have a positive δ 7 Li, close to the normal mantle, which is inconsistent with the negative δ 7 Li found in this study. This implies that the δ 7 Li of Qinghu zircons have been lowered by later processes. Therefore, if zircon grains are small (e.g. Temora zircon), the δ 7 Li value in the center of the grain may have been shifted by diffusion and/or later processes and thus cannot represent the pristine Li-isotope value. Confirmation of homogeneous δ 7 Li and [Li] by multiple analyses is necessary to define the pristine igneous Li signature in zircon. Heterogeneity (δ 7 Li variations ~20‰ within individual grains) indicates that the bulk analysis of Li isotopes in mineral separates and in bulk-rock samples may produce misleading data.

Methods
The Li-isotope composition of whole-rock samples was measured using solution MC-ICP-MS methods at the Geochemical Analysis Unit, GEMOC, Macquarie University. The procedure used for purification of the dissolved solution is similar to those previously described by Seitz et al. (2004) 33 . The isotopic analysis of each unknown sample was bracketed by two 15 ppb L-SVEC standard solutions ( 7 Li/ 6 Li = 12.039 ± 0.03) 34 to correct for mass fractionation. The raw δ 7 Li values of unknowns are expressed as δ 7 Li = [(R measure /R L-SVEC )-1]*1000‰. The accuracy of the results was monitored using international rock standard RGM-1 with average δ 7 Li values of 2.3 ± 0.4‰(2SD) in agreement with the published values 35 . The average reproducibility based on samples and standards is ± 0.9‰ (2SD). The analytical data are listed in Table 2.
SIMS in situ analysis of Li isotopes were performed using the Cameca IMS 1280HR SIMS at the Institute of Geology and Geophysics, Chinese Academy of Sciences (IGG-CAS) in Beijing, following the procedures described by Li et al (2011) 16 . A ~20 μ m and ~10 μ m beam was used to traverse zircon grains, while a ~5 μ m beam was used for detailed analysis at the zircon rim. Drift of δ 7 Li values in the M257 standard over repeated analyses is typically < 2‰ in a single analytical session with an internal precision between 1.5 and 2.0‰ (2SD) for a single spot with ~20 μ m beam size, 2.5‰ for ~10 μ m beam size and 5.0‰ for ~5 μ m beam size. Measurements of Li isotopes and concentrations were conducted in several separate sessions. The full data set is presented in Appendix dataset.
Secondary ion images were obtained for construction of element concentration maps (Li, P, Ti, Yb, Y, U, and Th) over selected zircon grains within a single analysis using a peak-switching mode. The sample was pre-sputtered over a 200 μ m × 200 μ m area by rastering for 300 s in order to eliminate surface contamination. We used a focused beam of O − with a diameter of 5 μ m (∼ 1 nA) to scan a 150 μ m × 150 μ m area. The Dynamic Transfer Optical System was applied to obtain a wide field of view and 7 Li + was detected by the electron multiplier. Other mass spectrometric conditions were similar to those used for isotopic analyses of the samples.