Although zircon is fairly resistant to normal chemical attack altered varieties, so called metamict zircons, are not uncommon, the mineral eventually becoming converted to an isotropic form with a lower density. In addition to some degree of structural instability the presence of radioactive atoms involving autoxidation plays a considerable part in the zircon metamictization. Holland and Gottfried (1955) suggested that the observed effects are caused mainly by the displacement of atoms by recoil nuclei, and by the high temperatures produced in the path of the nuclear particles. As the irradiation proceeds, the X-ray powder patterns show movement of the peak positions and then distortion of the peak shapes followed by the substitution of a new set of peaks. Finally the second peaks completely disappear with the production of an isotropic glass: the density falls by 16 per cent. The breakdown of the structure is envisaged as a four-stage process in which the structure is first saturated with displacements; the saturated structure then breaks down into crystallites of ordered zircon which ultimately break down to a glass. There is a considerable range of properties observable in zircon, ranging from the fresh or normal zircon to the fully metamict type (Deer et al., 1962).

Zircons are extensively used in the U/Pb dating of complex metamorphic terrains, and ion microprobe measurements of cores of single crystals have given the oldest known terrestrial ages. The oldest dated zircons are from paragneisses and probably originate from a variety of sources. Since, in these samples, the radiogenic lead has remained immobile, it is probable that the original trace element chemistry has also been preserved. Zircon compositions should therefore give us some insight into the chemistry of a variety of ancient rocks (Hinton and Upton, 1991).

A number of studies have been made of zircon chemistry, some including coexisting minerals. The zircon structure has two possible sites for cation substitution, a tetragonal (Si) and a triangular dodecahedral site (Zr), (Speer, 1980). Zircon chemistry is relatively simple; usually ZrO₂, HfO₂, and SiO₂ compose more than 99% of the total oxides. The other substituting elements commonly present are Y, REEs, Th and U. Substitution predominantly involves replacement of Zr in the larger triangular dodecahedral sites. Coupled substitution of REEs and PO₄ into the zircon structure (Speer, 1980; Romans et al., 1975) can lead to substantial enrichment in the REEs.

Considering the effects that temperature, pressure, oxygen fugacity, and initial magma composition could have on mineral chemistry (Hanson, 1978), it seems reasonable to anticipate that zircon grains crystallizing from diverse magma compositions will inherit a diagnostic geochemical signature. Previous investigations of the minor and trace element content of zircon grains have documented the presence of as many as 50 elements in zircon (Speer, 1982) and, more importantly, have shown that there are profound differences in the abundance of some elements such as the REEs, Hf, Ta, and Sc in zircon concentrates isolated from carbonatites, alkaline basalts, nepheline syenites, and granitic pegmatites (Murali et al., 1983; Irving and Frey, 1984; Kapustin, 1985).

A large number of elements have been reported in zircon (Speer, 1980); however, as noted by some authors many of the trace elements detected are present within inclusions.

2. GENERAL GEOCHEMICAL CHARACTERISTICS OF ZIRCONS

Zircon always contains a certain amount of Hafnium: the HfO₂/ZrO₂ ratio varies but is normally about 0.01. The highest ratio is found in metamict varieties (0.06) while it increases from 0.015 for zircons in nepheline-syenites to 0.04 for those of granites (Rankama and Sahama, 1950). Phosphorus may be present in some varieties, probably replacing Si, the structure maintaining electrostatic neutrality by the entry of rare earths.

The REE substitution into zircon can clearly be shown to be related to the size of the REE ion. The Zr⁴⁺ which normally occupies the triangular dodecahedral site is smaller than the Lu³⁺ ion. Since the ionic radii of the REEs decrease from La³⁺ to Lu³⁺ (lanthanide contraction), the substitution into the zircon lattice becomes progressively easier for the higher atomic number REEs. This effect is illustrated in the analyses of zircon and other coexisting phases from igneous rocks (Fujimaki, l986; Sawka and Chappell, 1988; Heaman et al., 1990).

The chondrite-normalised REE concentrations for zircon increase rapidly from Sm to Lu even in rocks which have significant LREE enrichment. Calculated partition coefficients between zircon and melt (based on measured concentrations of mineral separates and interstitial glass from volcanic rocks or whole rocks) are high for the heavy REEs (HREEs), falling rapidly towards the light REEs (LREEs; approx. 400 for Lu and 4 for Ce; Hendersen, 1984). Despite the relatively high concentrations of the HREEs in zircon, it generally does not have a significant effect on the bulk rock REE chemistry due to its low abundance (Gromet and Silver, 1983; Exley, 1980). Experimental studies have also demonstrated zircon's preference for the HREEs (Watson, 1980). However, the measured variations in the partition coefficients between the HREEs and the LREEs are much lower than those observed in the natural samples, and the experimental studies suggest that significant partitioning of the LREEs into zircon is possible.

It has been shown that Ce can be anomalously high in zircon compared to other measured LREEs (Murali et al., 1983). The chondrite-normalised patterns reported by Murali et al. (1983) included some which contained both positive Ce and negative Eu anomalies. However, as not all the LREEs were analysed the exact relationship between Ce and the other LREEs (and hence sign and the magnitude of the Ce anomaly in these zircons) could not be fully investigated.

Variations in trace element chemistry due to growth zoning, inclusions, or alteration make even single-crystal trace element determinations average values. The electron microprobe has been used to measure some of the more abundant trace elements (the HREEs, Pb and U; Exley, 1980), but many important trace elements (e.g., LREEs) are below the detection limit of this method.

There is still some uncertainty as to the extent that this geochemical variation is an intrinsic feature of the zircon populations examined. In certain cases, 100% pure zircon separates were not used, some zircon fractions analysed consisted of more than one zircon population, and some of the existing data were determined by instrumental neutron activation analyses (INAA) where there is some difficulty in assigning appropriate corrections for U and U-fission products, especially in uranium-rich zircon fractions.
 
 

3. MINOR AND TRACE ELEMENT CONTENT OF ZIRCONS
 
 

Heaman et al. (1980), hand picked 100% pure zircon fractions by selecting zircon fractions from diverse rock types in order to provide a general geochemical database for igneous zircon suites. They describe a technique for analysing the trace element content of extremely small quantities of zircon (i.e., less than 1 mg) by INAA in which uranium is removed from the zircon fractions prior to analysis using anion exchange chromatography to avoid the problem of U-fission corrections. Finally, three zircon fractions with a range in magnetic susceptibility were selected from one sample to investigate the nature of geochemical variations within a single zircon population. One of the most striking features of these data is the enormous range in the minor and trace element content of these zircon populations, even when the two magnetic fractions from LH84-4 are excluded: Sm (<1-44 ppm), Yb (6-2056 ppm), Lu (2-479 ppm), Sc (<10-230 ppm), Hf (6000-11600 ppm), Th (2-3600 ppm), and U (< 10-1600 ppm). In addition to the four groupings listed above, a Group V is introduced there and consists of zircon fractions isolated from mafic volcanic rocks. The three zircon fractions that fall into the Group V category include a composite of three zircon megacrysts from the Elie Ness tuff Scotland (Irving and Frey, 1984), and groundmass zircon from two Kyushu andesite samples, Japan (Fujimaki, 1986). Based on whole-rock major element geochemistry alone there is some overlap between some of the Group III (mafic plutonic) and (mafic volcanic) samples. However, since there is such a large difference in the Hf content of zircon from the samples they studied, they have treated them as two groups throughout the discussion to facilitate a geochemical comparison of volcanic and plutonic zircon suites.

The results for zircon fractions analysed in this study are plotted on selected element-element and ratio-element diagrams together with the results for zircon fractions isolated from mafic volcanic rocks (Iving and Frey, 1984; Fujimaki, 1986). One of the more interesting and striking features of the data is that at least three of the five proposed zircon groups have diagnostic Hf signatures. As discussed above, zircon fractions from mafic volcanic rocks (Group V) have some of the lowest known Hf contents (4100-5200 ppm), carbonatite and nepheline syenite zircons (Group II) have intermediate Hf contents (6000-8000 ppm) and do not overlap with the Group V zircons while Group I, III, and IV zircon fractions have "normal" Hf contents compared to the range commonly reported for igneous zircon suites (10000-11600 ppm).

Although based on Hf content alone there is no clear geochemical distinction between zircon suites isolated from felsic, mafic, and kimberlite samples, the Group III zircon suite (mafic) has distinctly higher Sc contents (86-230 ppm) than all other zircon suites. In addition to the relatively high Sc abundances, some of these zircon populations have very high Lu/Sm ratios (i.e., >20) indicating that the mafic suite zircon populations can have more fractionated REE profiles compared to other zircon suites. The zircon fraction (GD233) with the highest Lu/Sm ratio (44.29) is in fact from the felsic suite: a plagiogranite sample from an ophiolite complex. A diagnostic feature of the kimberlite zircon suite is the unusually low HREE concentrations.

According to Heaman et al. (1990), another geochemical feature of zircon that has been used to distinguish metamorphic from igneous zircon is the Th/U ratio. This ratio is generally higher for metamorphic zircon but there is no comprehensive evaluation of the Th/U ratio in different igneous zircon suites. The Th/U ratio in zircon, as pointed out by Ahrens et al. (1967), is generally much lower (i.e., < 1 ) than the average Th/U ratio of 4 for crustal rocks implying a preferential incorporation of U versus Th in zircon during crystallization from a magma. The majority of zircon fractions analysed in this study have Th/U ratios between 0.28 and 1.17 with an average of 0.56 for sixteen samples. These results are in general agreement with the range and average Th/U ratio (0.15-1.20 and 0.47, respectively) reported by Ahrens et al. (1967) for 158 dominantly granitic zircon suites. This relatively restricted range in Th/U ratio is particularly striking considering the rather large range in U (7-1568 ppm) and Th (2-3641 ppm) abundances. Zircon fractions from two of the three mafic samples (MAN85-13 and NR) have anomalously high Th/U ratios (1.74 and 2.32, respectively) and may be the hallmark of late-stage crystallization of zircon from a mafic magma. One carbonatite sample (LH86-82) has an anomalously low Th/U ratio of 0.05.

The chondrite normalized REE profiles for the various zircon suites analysed by Heaman et al. (1990). For comparison, the REE profile for the zircon megacryst composite (solid circles, thick dashed line) from the Elie Ness basanite, Scotland (Irving and Frey, 1984), is also shown by these authors in each diagram. As reported in previous studies (e.g., Nagasawa, 1970), all the zircon suites analysed show a preference for concentrating the HREEs. The REE profiles for zircon suites from kimberlites (Group I, open circles), carbonatites (Group II, solid squares), and the nepheline syenite pegmatite (Group II, open squares and thin dashed lines) show that the REE profiles for all Group I and II zircon suites plot below the Elie Ness reference profile indicating that the overall REE abundances in these zircon suites are relatively low. In particular, the REE contents of the kimberlite zircons are extremely low. The profiles for both the carbonatite and nepheline syenite zircons closely resemble the Elie Ness pattern except the latter shows a noticeable negative Eu anomaly. This Eu anomaly is present in both undissolved zircon fragments from this sample and is considered "real" since the U-fission correction for Sm in this sample is negligible (Heaman et al., 1990)..

In contrast, the mafic (Group III) and felsic (Group IV) zircon suites have REE profiles with higher total REE abundances than the Elie Ness zircon. One sample from the Group III suite (NR) has a particularly interesting profile with a pronounced negative Eu anomaly and an elevated La content (16 ppm) compared to zircons from other mafic rocks (Irving and Frey, 1984; Fujimaki, 1986; Heaman et al., 1990). Three of the Group IV zircon fractions (C83-31, LLZ1, GFA86-675) also have negative Eu anomalies, a feature also reported in some other felsic zircon suites (e.g., Nagasawa, 1970; Sawka, 1988). Only one sample analysed by Heaman et al. (1990), so far (K71-48) has an apparent positive Eu anomaly and is the only igneous zircon fraction known to the authors that has such a pattern. However, aacording to them, some caution must be exercised in interpreting the Sm data from this zircon fraction since the total amount of Sm (0.42 nanograms) is quite low and the corresponding errors are exceedingly high. Nevertheless, it is interesting that this latter sample is part of an anorthosite suite that typically have huge positive Eu anomalies.

Heaman et al. (1990) investigated three samples from alkaline/ peralkaline complexes: cited as GFA86-675, 86AD 1, and CL2. Except for the Coldwell granite zircon population (CL2) which has an anomalously low Hf content (8179 ppm) compared to the other Group IV zircon fractions and plots close to the Group II carbonatite field, there appears to be no consistent geochemical feature of these zircon fractions that might represent a possible tracer to discriminate between alkaline and calc-alkaline felsic suites (Heaman et al., 1990).

4. CONCLUSIONS
 
 

An important consideration in evaluating the potential of an accessory mineral, such as zircon, for geochemical tracer studies is the magnitude of primary versus secondary geochemical variations present within a single population. Numerous factors such as magma composition, degree of melt polymerization, temperature, and pressure may together play an important role in determining mineral/melt partition coefficients and hence the indigenous trace element abundance in accessory minerals. However, this geochemical signature may be modified by geochemical zoning, presence of alteration, inclusions or other imperfections. Two processes that could modify the intrinsic geochemical characteristics of igneous zircon suites should be examined: geochemical zoning and alteration.

One source of trace element variation in accessory minerals is zoning. Geochemical zoning in zircon can occur either as a heterogeneous distribution of trace element enriched domains (e.g., Romans et al., 1975), a progressive increase in Hf, U, Th, and the REEs from the core to the rim (Exley, 1980; Speer, 1982), or, as in the case of grains with visible growth bands, the geochemical zoning is characterzied by alternating trace element enriched and depleted growth bands (e.g., Chakoumakos et al., 1987).

Microprobe studies of zircon crystals that do not show visible growth bands have demonstrated that both geochemically zoned (e.g., Fujimaki, 1986) and unzoned (e.g., Fujimaki, 1986; Sawka, 1988; Sawka and Chappell, 1988) zircon suites exist. The rims of geochemically zoned crystals can be enriched in certain trace elements by as much as ten times the levels found in the core (e.g., Exley, 1980), possibly as a consequence of the elevated trace element concentrations in the magma during the final stages of zircon crystallization. In these cases, the average zircon composition could be biased toward the composition of the rims and could mask the intrinsic geochemical signature of a zircon population. To avoid some of the complications associated with analysing geochemically zoned crystals which Heaman et al. (1990) have removed the outer surfaces of most grains by an abrasion technique of Krogh (1982), and their samples with zircon populations that do not show visible growth bands, and selected zircon crystals from fractions showing the least magnetic susceptibility where geochemical zoning is expected to be minimal.

Although chemical zonation is present in many zircon fractions, the relatively uniform Hf contents (9000-11000 ppm) for the majority of fractions analysed in Heaman et al. (1990)'s work implies that the scale and magnitude of this zonation is reduced if the central portions of transparent, non-magnetic crystals devoid of visible imperfections are selected.

Another source of geochemical heterogeneity in accessory minerals is alteration. Alteration in zircon crystals is quite common and in some populations it is ubiquitous. Zircon crystals containing altered domains that are enriched in Fe and depleted in Pb often have a high magnetic susceptibility which explains the now commonly observed correlation that exists between the degree of discordance and magnetic susceptibility in a zircon population first described by Silver and Deutsch (1963). In the case of zircon crystals with internal growth bands, variations in the intensity of alpha-decay in the alternating U-rich and U-poor bands can cause differential expansion and microfracture development (Chakoumakos et al., 1987) which would enhance the creation of pathways for fluid infiltration and alteration. However, the critical issue is whether the trace element enriched domains existed prior to alteration. Initially high-U and trace element substituted domains, which would tend to accumulate more radiation damage in a given period of time, could be more susceptible to subsequent fluid interaction and associated alteration, or alternatively there could be a net increase in trace element content in damaged domains during the process of alteration (Heaman et al., 1990).

One method to investigate the nature of trace element variations within a single, homogeneous zircon population is to analyse zircon fractions that represent crystals with slightly different magnetic susceptibilities. If the zircon composition is exclusively a function of crystal/chemical control during crystallization then one would predict, based on the Rayleigh fractionation law, that for a protracted period of zircon crystallization there would be a smooth and progressive enrichment in the trace element content of zircon throughout the crystallization history of the magma. One of the most sensitive indicators of crystal/chemical control would be the REE pattern. The absolute abundances of the REEs in zircon might increase during crystallization, but the REE profile does not vary significantly with changes in temperature or magma composition (Gromet and Silver, 1983), so the profile should remain constant. For this experiment Heaman et al. (1990) selected three fractions from a "simple" zircon population from a hornblende-biotite granite sample. This granite body does not form part of a differentiated plutonic suite and likely experienced a relatively rapid crystallization history. The three fractions selected include zircon grains with slight differences in magnetic susceptibility. All three fractions were devoid of visible inclusions or growth bands, but the 7 degree fraction contained some grains with visible fractures. Diffuse zones of turbidity could be seen associated with some of these fractures and could represent metamict or altered domains. Heaman et al. (1990) stated that all the elements analysed increase systematically from the least to most magnetic fraction. The most dramatic increase is observed for Sm (83% change) suggesting that the light rare earth elements (LREEs) are preferentially enriched compared to the HREEs in the more magnetic fractions. Assuming that the zircon/magma partition coefficients for the REEs remain constant during crystallization, the preferred interpretation for this change in the REE profile is an external influence (alteration?) on the trace element abundance of the more magnetic zircon crystals. The alternative interpretation is that the LREEs are more strongly zoned in this zircon population than the HREEs.

Although alteration might account for the high magnetic susceptibility of some zircon grains, there is evidence that this is not the case for all magnetic fractions. A large percentage of zircon crystals isolated from mafic rocks have high magnetic susceptibilities but do not show visible signs of inclusions, internal growth bands, or alteration. In fact, many of these fractions can be shown to represent closed systems with respect of U-Pb geochronology even though they have high trace element contents (i.e., > 1000 ppm U). The weak paramagnetism combined with the often skeletal habit of these grains likely reflects the increased trace element substitution, especially the LREEs, during rapid, late-stage crystallization from a mafic magma.

The distinction between primary and secondary geochemical variations in zircon, and probably other accessory minerals, is not always readily apparent. However, it is clear that in order to enhance the possibility of obtaining the primary geochemical signature of a zircon suite, it is paramount that stringent selection procedures be used to isolate the "best quality" grains that are devoid of alteration, inclusions, and other imperfections and preferably grains that do not show internal growth bands or have large magnitude within-grain geochemical variability (Heaman et al., 1990).

The minor and trace element content of the best quality zircon grains isolated from a variety of igneous rocks of Heaman et al. (1990)'s study indicate that there are some profound geochemical variations that are unique to certain rock types. The most salient geochemical features of each suite are outlined by Heaman et al. (1990) and are compared to the average composition of the felsic zircon suite: Kimberlitic zircons (Group I) can be distinguished by extremely low abundances of U, Th, and the HREE and relatively unfractionated REE patterns; carbonatite and nepheline syenite zircons (Group II) have anomalously low Hf contents (6000-8000 ppm) with correspondingly high Zr/Hf ratios (>60); zircon grains from mafic rocks have elevated levels of Sc (>80 ppm), highly fractionated chondrite normalized REE patterns, and high Th/U ratios (> 1); and zircon from mafic volcanic rocks (Group V) have some of the lowest known Hf contents (4100-5200 ppm). Therefore, from this cursory examination of zircon geochemistry, it appears that zircon is a feasible geochemical tracer.

Heaman et al. (1990) stated that, the zircon chemistry reflects the composition of the magma from which it crystallized. According to them, mafic rocks are typically enriched in Sc compared to more differentiated rocks, and the three zircon fractions from mafic rocks analysed have significantly higher Sc contents than all other zircon fractions. Similarly, the Lu/Sm ratios in mafic rocks are generally higher than most other rock types, and the zircon Lu/Sm ratios can be quite high for the mafic suite. However, in order to explain how zircon from vastly different magma types can have a similar geochemical anomaly, such as the low-Hf zircon fractions found in carbonatites, a nepheline syenite (86GD01 ), and granite (CL2) of Heaman et al. (1990), some of the geochemical attributes of zircon must reflect more than just bulk magma composition.

Some of the previously reported geochemical data for igneous zircon suites do not conform to the simplified subdivision of Heaman et al. (1990). Most notable among these are the results for zircon isolated from granitoid rocks of the Sierra Nevada batholith (Sawka, 1988; Sawka and Chappell, 1988) and the results for zircon from high-silica rhyolites (Mahood and Hildreth, 1983). For example, zircon separated from different phases of the difterentiated McMurry Meadows pluton have a wide range in composition. The Hf content of zircon populations from this pluton vary from 3277 ppm (granodiorite) to 8809 ppm (leucogranite) with a progressive increase in the zircon/ wholerock Hf and Sc contents with fractionation (Sawka, 1988). This is contrary to the results presented by Heaman et al. (1990) where the Sc contents in zircon tend to be highest in the more mafic rock types. Clearly if zircon is to become a widely used geochemical tracer, additional studies like those of Sawka.

It has been shown by some studies that certain trace elements are diagnostic of magmas generated in different tectonic settings (e.g., Pearce and Cann, 1973; Wood et al., 1979). For mafic magmas, some progress has been made in this direction by using elements that are hygromagmatophile (i.e., elements that have bulk partition coefficients less than 1 and hence are strongly partitioned into a magma during partial melting; Wood et al., 1979). Some of the elements that are particularly useful for delineating tectonic settings in mafic magmas such as Hf, Ta, and Th (Wood, 1980) are also present in zircon. Zircon from the two Kyushu andesites (Fujimaki, 1986) have unusually low Hf contents, since magmas generated at destructive plate margins are also characterized by low Hf contents (Wood, 1980).

Low-Hf zircon (i.e., <9000 ppm Hf) is also characteristic of igneous activity in continental rift settings. A number of samples including the carbonatite and nepheline syenite samples of Group II and the Coldwell granite sample (CL2), analysed by Heaman et al. (1990) all formed in continental rift settings. At the present time it is unclear whether the low-Hf zircons have acquired a trace element signature that reflects the geochemical nature of the subcontinental mantle source region where these magmas are generated or whether other factors influence the zircon/magma partition coefficients in these magmas.

Since zircon is a very robust mineral and is often resilient to mechanical destruction it is very useful mineral for geochemical and petrogenetical works as well as provenance studies.

REFERENCES
 
 

This study was prepared through the Term Project Paper of Advanced GeochemistryCourse (Instructor: Assis. Prof. Dr. A. Pırıl ÖNEN).