A series of natural zircon samples (with U concentrations of 140–2600 ppm and ranging from well crystalline to severely radiation damaged) were investigated by means of REE3+ photoluminescence spectroscopy. We found systematic changes in REE3+ emissions depending on the accumulated radiation damage expressed by the effective time-integrated α-dose of zircon samples. Structural reconstitution as caused by dry annealing resulted in intensity gains and decreases of half-widths of REE3+ emissions. The band half-widths of distinct luminescence Stark’s levels of the 4F9/2 → 6H13/2 transition of Dy3+ (~17 250 cm−1; ~580 nm wavelength) and the 4F3/2 → 4I9/2 transition of Nd3+ (~11 300 cm−1; ~885 nm wavelength) were found to correlate sensitively with the degree of radiation damage accumulated. These REE3+ emissions are proposed as potential measure of the irradiation-induced structural disorder of zircon. The two emissions are considered particularly suitable because (1) they are commonly detected in PL spectra of natural zircon, and (2) they are hardly biased by other emissions or Stark’s levels. Preliminary calibration curves that relate band-width increases to the α dose were established using a suite of well-characterized Sri Lankan zircon. Band broadening upon increasing corpuscular self-irradiation is assigned to increasing structural destruction, i.e., the increasing perturbation of REE3+ cationic lattice sites. Possible advantages of REE3+ luminescence spectroscopy, complementary to Raman spectroscopy, as method to quantify structural radiation damage are discussed.
Structural disorder in U- and Th-containing minerals is caused predominantly by α-decay events, in particular comprising atomic displacements created by recoils of heavy daughter nuclei. The α particles, in constrast, have minor contribution to the total bulk damage (ca. 10–15%; Nasdala et al. 2001), whereas β and γ radiation are considered insignificant for the creation of permanent structural damage. The transformation of initially crystalline minerals into an amorphous state, through accumulation of recoil clusters over geologic periods of time, is commonly referred to as “metamictization” (e.g., Brøgger 1893; Pabst 1952; Ewing 1993).
Zircon, ZrSiO4, is a widespread accessory mineral that occurs in many igneous, metamorphic, and sedimentary rocks. Zircon incorporates the actinides U4+ and Th4+ that substitute for Zr4+ in dodecahedral coordination (D2d; Finch et al. 2001). Concentrations of the two actinides lie predominantly in the range 10–4000 ppm. As much as several weight percent have however been reported in some cases (for instance reported by Törnroos 1985; Rubin et al. 1989; Geisler et al. 2005). The accumulation of structural damage as caused by the radioactive decay of U and Th and their instable daughter nuclei, results in dramatic changes of solid-state properties. This includes changes of physical parameters such as refraction and birefringence, hardness, density, and elastic moduli (Holland and Gottfried 1955; Chakoumakos et al. 1991; Palenik et al. 2003), and a general decrease of the chemical resistance (Balan et al. 2001; Mathieu et al. 2001; Geisler et al. 2003a). The generally increased susceptibility of radiation-damaged zircon to alteration in potential post-growth processes is of enormous importance as it for instance affects negatively the ability of this mineral to immobilize radioactive elements (storage of nuclear waste; e.g., Ewing 2001) or to retain radiogenic nuclei (bias of U–Pb geochronology results; Geisler et al. 2003b).
In addition to X-ray diffraction (e.g., Weber 1990; Ríos et al. 2000), several spectroscopic techniques have been applied to quantify radiation damage accumulated in zircon. For instance, Wasilewski et al. (1973), Deliens et al. (1977), Zhang et al. (2008), and Zhang and Salje (2001) used infrared (IR) absorption spectroscopy, and Farnan and Salje (2001) and Farnan et al. (2007) applied 29Si nuclear magnetic resonance (NMR) spectroscopy for the estimation of the amorphous fraction in radiation-damaged zircon. Raman spectroscopy was introduced by Nasdala et al. (1995) as a method to quantify the degree of radiation-induced structural disorder in zircon. This technique has opened up new opportunities for studying the structural state of heterogeneous zircon crystals and for investigating relationships between α-doses, age, and the annealing history of geothermal events on a micrometer-scale (cf. Nasdala et al. 2001). The present study has been performed with the objective to test the applicability of REE3+ luminescence spectroscopy for the very same purpose, i.e., to check if the proposed method may be used, complementary to Raman spectroscopy, to characterize and estimate non-destructively the radiation-induced disorder of zircon.
Zircon typically incorporates traces of—mainly heavy—trivalent rare-earth elements (REE) into the structure, which is explained commonly by the coupled, xenotime-type substitution (Hanchar et al. 2001)
The emissions of REE3+ of zircon have been studied with different luminescence techniques based on differing excitation mechanisms, such as photoluminescence (PL, e.g., Friis et al. 2010), cathodoluminescence (CL, e.g., Blanc et al. 2000), ionoluminescence (IL, e.g., Yang et al. 1994; Finch et al. 2004), thermally induced luminescence (TL, e.g., Vaz and Senftle 1971; Van Es et al. 2002), etc. Among them, CL imaging is the most widespread technique within the Earth sciences. Intensity-based imaging of CL emissions is commonly used to visualize internal textures of zircon, which provide valuable information on primary formation and post-growth history (e.g., Vavra 1990; Hanchar and Miller 1993; Hanchar and Rudnick 1995; Rakovan and Reeder 1996; Götze 2000, 2002; Corfu et al. 2003; Götze et al. 2013). Luminescence spectroscopy of REE emissions was (and still is) used successfully to detect traces of different REE species in zircon (e.g., Nicholas 1967; Yang et al. 1992; Götze et al. 1999; Kempe et al. 2000; Gaft et al. 2001). Synthetic zircon individually or multi-doped with REEs have also been studied routinely, aiming at a better understanding of the luminescence in their natural analogs (e.g., Cesbron et al. 1995; Blanc et al. 2000; Karali et al. 2000; Hanchar et al. 2001; Friis et al. 2010).
More recently, the luminescence of REE3+ was used as structural probe for the characterization of REE-substituted lattice sites in minerals and mineral-based ceramics. For instance, spectral parameters of Eu3+ emissions have been used for the characterization of REE-substituted cation-sites in glasses and crystalline materials in process of evaluation of potential nuclear waste forms (Ollier et al. 2003; Reisfeld et al. 2004; Ternane et al. 2005).
It has been found that REE3+ emission-band widths are influenced significantly by the real structure, i.e., deviations from the ideal chemical composition and structural state. In particular, it has been observed that individual REE emissions broaden appreciably, and decrease in intensity, upon (1) increasing incorporation of non-formula elements (e.g., Lenz et al. 2013), and (2) increasing radiation damage (e.g., Jazmati and Townsend 2000; Nasdala et al. 2002, 2013; Ruschel et al. 2010; Panczer et al. 2012).
In the present study we review briefly the REE3+ photoluminescence typical of natural zircon samples, and effects of radiation damage on the REE3+ luminescence pattern of zircon. The potential of hyperspectral luminescence mapping, e.g., to visualize radiation damage patterns on a micrometer scale, is demonstrated by comparing well-characterized, cut zircon crystal halves (with on half annealed through heat treatment and the other left in the natural radiation-damaged state, Nasdala et al. 2006). Finally, we show that the FWHM of selected REE3+ luminescence sublevels of zircon is related to the amount of radiation-damage accumulated over geologic periods of time (α-dose). The latter aspect provides an alternative possibility to estimate the quantity of radiation damage present in unknown zircon samples by measuring the FWHM of REE luminescence Stark’s levels.
Samples and methods
The effects of radiation damage on the REE3+ luminescence were first studied using four zoned samples of natural, accessory zircon single crystals (lengths ranging from 150 to 350 μm, U–Pb ages scattering between Archean and Neogene), which have been characterized in detail by Nasdala et al. (2006). In that study, single crystals were cut in two halves, along their c-axes. One half each was then subjected to heat treatment in air to anneal the radiation damage, whereas the other half remained in its natural, radiation-damaged state. Both, the annealed and naturally radiation-damaged half-crystal, of each sample were then embedded in epoxy, and prepared in close proximity to each other in one polished mount. These four zircon samples originated from the following lithologies and localities: (1) a potassium-rich leucogranite located near Dannemora, Adirondack Mountains, New York [sample A1; age ~1045–1050 Ma; (J.M. Hanchar and M.J. Whitehouse, personal communication); for rock description cf. McLelland et al. 2001], (2) the Bluffpoint quartz diorite, Atikwa Lawrence Batholith, Ontario [sample 81A; age 2732 ± 1 Ma; (D.W. Davis, personal communication); for rock description cf. Davis and Edwards 1985], (3) a gabbro from the Mulcahy Lake intrusion, Ontario (sample 31E; age 2733 ± 1 Ma; Morrison et al. 1985), and (4) a rhyolite tuff from the Gyulakeszi formation near Pécs, Hungary (sample M2; age 19.6 ± 1.4 Ma; Harangi 2001). An overview of the samples studied, and results of electron probe microanalysis (EPMA), are given in Table 1 (compare also Nasdala et al. 2006).
Second, we have investigated large, homogeneous, gem-quality zircon samples. These included 13 mildly to severely radiation-damaged stones from gem gravels in the Ratnapura area, Sri Lanka (U–Pb ages in the range 522–572 Ma; Nasdala et al. 2004a, 2008), and one well-crystalline stone from the Ban Lung area, Ratanakiri, Cambodia (age 1.2 ± 0.3 Ma, P.C. Piilonen, personal communication; for sample description see Wittwer et al. 2013, and references therein). Uranium and thorium concentrations, ages, α-doses and spectroscopic data of the gem samples are listed in Table 2.
To identify specific REE3+ emissions in natural zircon, PL spectra of synthetic REE-doped ZrSiO4 single crystals up to millimeters in size were obtained as internal references. Those materials have been synthesized in earlier studies using flux techniques; details are described elsewhere (Hanchar et al. 2001; Lenz et al. 2013).
Photoluminescence measurements in the visible to near infrared (NIR) range (single spectra and hyperspectral maps) were made using a Horiba LabRAM HR Evolution dispersive spectrometer. The spectrometer system was equipped with an Olympus BX41 optical microscope, two diffraction gratings with 600 and 1800 grooves per millimeter, and a Si-based, Peltier-cooled charge-coupled device detector. Photoluminescence was excited using a 473 nm diode-pumped solid-state laser (9 mW at the sample surface) and the 532 nm emission of a frequency-doubled Nd3+:YAG laser (10 mW at the sample surface). An Olympus 100× objective (numerical aperture 0.9) was used. The system was operated in the confocal mode (confocal aperture and entrance slit set at 100 μm); the resulting lateral resolution was ~1 μm, and the depth resolution (with the beam being focused at the sample surface) was ~2–3 μm. The spectral resolution for visible light was 0.8–1.1 cm−1 with the 1800/mm grating (used for low-T measurements). The hyperspectral PL maps, consisting typically of 60 000–80 000 single spectra, were obtained using a software-controlled x–y stage. Color-coded PL maps (here, distribution patterns of the FWHM of individual Stark’s levels) were then produced after appropriate data reduction, which included background correction and band-fitting assuming combined Lorentzian-Gaussian band shapes. Any mathematical correction of measured FWHMs for the system’s spectral resolution (in detail discussed by Nasdala et al. 2001) turned out to be unnecessary. This is because spectrometer-related artificial broadening of Stark’s level bands detected (FWHMs ≥12 cm−1 at room temperature) was negligibly small, owing to the high spectral resolution of the spectrometer. Low-temperature photoluminescence-measurements of homogeneous, gem-quality zircon samples were done using a long-distance, 50× objective (numerical aperture 0.55; free working distance 10.6 mm) and a Linkam FTIR-600 liquid-nitrogen cooling stage. The temperature accuracy was better than ±2 K.
Results and discussion
Figure 1 shows the laser-induced (λexc = 473 nm) REE3+ photoluminescence spectrum of sample M2. This spectrum can be considered typical of spectra of well-crystallized natural zircon. It is dominated by groups of narrow emission bands. In contrast, broad-band yellow and blue emissions (which are more often detected in CL spectra; e.g., Götze et al. 1999) were observed only occasionally in PL. Note that the sharpness of individual REE3+ emission lines (with FWHMs as small as 12 cm−1 at ambient temperature) may often causes confusion with Raman signals (for a detailed discussion see Nasdala et al. 2012).
Comparison of the PL spectrum of zircon M2 with spectra of synthetic REE-doped ZrSiO4 (Fig. 1) reveals that the former comprises the principal transitions of traces of Dy3+ and Sm3+ in the visible to NIR spectral range. The most intense emissions are assigned to electronic transitions of Dy3+, namely 4F9/2 → 6H15/2 near ~20 700 cm−1 and 4F9/2 → 6H13/2 near ~17 250 cm−1. They are accompanied by several transitions of Sm3+ (Fig. 1; compare, e.g., Cesbron et al. 1995; Blanc et al. 2000; Karali et al. 2000; Gaft et al. 2001; Friis et al. 2010). In addition, low-intensity PL of Tm3+ and Nd3+ (in the NIR near ~11 400 cm−1; assigned to the 4F3/2 → 4I9/2 transition) was observed. Assignments of electronic transitions and Russel-Saunders terms in Figure 1 have been extracted from “Dieke” energy-level diagrams (e.g., Dieke and Crosswhite 1963; Reisfeld and Jørgensen 1977; Wegh et al. 2000).
All electronic transitions detected show crystal-field splitting into sharp sublevel bands (which are commonly referred to as Stark’s levels). This is caused by a non-isotropic electronegative, crystallographic environment of REE3+ ions (crystal field splitting; e.g., reviewed in Burns 1993). As one consequence, REE3+ emissions show notable orientation dependence, i.e., individual Stark’s levels of a given transition may be detected with variable relative intensities, depending on crystal orientation (compare, for example, relative intensities of individual peaks in the main Dy3+ emissions of zircon M2 and Dy3+-doped ZrSiO4 in Fig. 1).
The predominance of Dy3+ bands in REE3+ luminescence spectra of natural zircon has been well documented already in CL (Mariano 1989; Rémond et al. 1992; Hanchar and Rudnick 1995; Götze et al. 1999) and IL studies (Yang et al. 1994; Finch et al. 2004). In contrast, emissions of Ho3+ (cf. spectrum of Ho3+-doped ZrSiO4 in Fig. 1) were not detected in any steady-state PL spectrum obtained from the natural zircon samples studied here, neither with 473 nm nor with 532 nm laser-excitation. This is remarkable insofar as some of the samples have Ho and Dy concentrations of comparable magnitude. For instance, zircon M275 contains 12.6 ± 0.8 ppm Dy and 4.4 ± 0.3 ppm Ho (Nasdala et al. 2008) and M127 contains 57 ± 6 ppm Dy and 21 ± 2 ppm Ho (D. Frei, personal communication); however both samples showed intense Dy emission but no notable Ho emission bands. This observation seems to support results of the IL study of Finch et al. (2004) who concluded that Ho3+ emissions may be suppressed in the presence of Dy3+.
Moreover, it is worthy of note that PL spectra strongly depend on the excitation wavelength. None of the PL spectra of natural zircon samples, including M257, obtained in the present study (with 473 and 532 nm laser excitation) showed emissions of Er3+ with significant intensity. In contrast, Er3+ emissions have been readily detected in PL spectra of zircon M275 obtained with 488 nm excitation (Nasdala et al. 2008). The contradiction is however only apparent, because selective excitation of REE emissions is well known in steady-state laser-induced PL (e.g., Blasse and Grabmaier 1994). Strong excitation bands of Er3+ centered at 490 nm (4I15/2 → 4F7/2) have been reported for synthetic zircon (Friis et al. 2010). These may allow PL excitation with a nearby 488 nm, but not with a 473 nm laser.
Low-intensity emissions of Er and Ho, as well as emissions of other REEs in zircon (including Eu, Tb, Pr, Tm, Ce), which are hardly or not detectable with steady-state luminescence techniques, can however be measured using time-resolved laser-induced spectroscopy (cf. Gaft et al. 2001).
Hyperspectral PL mapping
Reliable estimation of spectral parameters (i.e., band fitting to determine FWHMs of individual REE3+ sublevels) is possible only if the emission detected consists of well-separated, distinct peaks. This is not the case for Sm3+ (and, similarly, not for Er3+ and Ho3+ emissions) in synthetic ZrSiO4 whose emissions show numerous Stark’s levels that strongly superimpose each other (Fig. 2A). In contrast, Stark’s levels of the 4F9/2 → 6H13/2 transition of Dy3+ (especially those two in the range 17 240–17 260 cm−1) can be identified and fitted without bias, as there are intense and not obscured by other bands (Fig. 2B). Dy3+ Stark’s level I (Fig. 2B) is therefore used exemplarily in the present study, for the creation of PL maps based on the distribution of this band’s FWHM.
Figure 3 shows BSE images, CL images and hyperspectral PL maps of pairs of halves of heterogeneous zircon crystals, whereas one half is naturally radiation-damaged, the other one annealed. The PL maps show color-coded distribution patterns of the FWHM of sublevel I of the 4F9/2 → 6H13/2 transition of Dy3+. Corresponding PL spectra and the respective Stark’s levels used for mapping are presented in Figure 2B.
Figure 3 shows that accumulation of self-irradiation damage causes significant BSE increase and CL intensity decrease, as revealed by opposite trends upon structural reconstitution by dry annealing (Fig. 3). Nasdala et al. (2006) attempted to explain the BSE increase with increasing damage by electron channeling contrast (cf. Mitchell and Day 1998), and the CL-intensity loss upon damage accumulation was discussed by Nasdala et al. (2002). A similarly close, inverse correlation was observed also for the total PL emission, whose intensity decreases appreciably with increasing U concentration and radiation damage (see Fig. 2 again). The hypothesis that the presence of U is the main reason for the decreased PL emission (this effect was proposed for CL by Poller et al. 2001) is disproved by the fact that annealed halves have the same U concentrations as their unannealed counterparts but yield much more intense PL. Also, ion-irradiation of synthetic and natural zircon samples done by Finch et al. (2004) showed that REE3+ emissions observed in IL strongly decrease with increasing radiation damage. Finch et al. (2004) explained their observation by increasing defect concentrations upon increasing structural damage created, which give rise to non-radiative transitions and cause a reduced possibility of energy migration through the crystal.
In addition to the general intensity loss of PL bands increasing radiation damage results in clear increases in FWHMs of individual bands (see also, e.g., Nasdala et al. 2013). Hyperspectral PL maps (Fig. 3) show that in naturally radiation-damaged zircon, the FWHM of Dy3+ sublevel I is broadened particularly in zones with high uranium concentrations and, hence, with more extensive radiation damage. In contrast, zones with low uranium concentrations (with corresponding lower levels of accumulated radiation damage) are characterized by narrow Dy3+ Stark’s level bands (for instance see overgrowth rim of sample 31E, Fig. 3c). Depending on the degree of radiation damage of particular interior regions, the Dy3+ sublevel near 17 200 cm−1 yielded FWHMs of up to >40 cm−1, whereas predominantly values close to 12 cm−1 were obtained after annealing. Distribution patterns of the FWHM of certain REE-emission sublevels (here, Dy3+ sublevel I) are hence virtual distribution patterns of radiation damage or crystallinity, respectively.
It should be noted that, in contrast to FWHM or band position, the band intensity and intensity-related spectral parameters such as the area integral are not suitable to estimate directly the degree of disorder. This is because luminescence intensity strongly depends on further variables, such as the absolute concentration of the respective REE3+, effects of quenching/sensitizing by other elements (e.g., Marfunin 1979; Kempe and Götze 2002), and effects of crystal orientation (e.g., Lenz et al. 2013). The present study was, hence, not focused merely on intensity-based REE distributions (e.g., MacRae et al. 2012, 2013), but based on the interpretation of band widths of REE emissions.
The dependence of the broadening of the PL emission on the structural damage, rather than on the U concentration, is supported by the observation that annealed halves (containing the same amounts of U as their unannealed counterparts; Table 1) show narrow Dy3+ sublevel-bands. Furthermore, clustering of REE centers potentially may also result in PL band broadening. The similarity of our PL maps and Raman maps for the same pairs of crystal-halves (presented by Nasdala et al. 2006) rather supports the dependence of PL broadening on the defect accumulation upon increasing radiation damage. Both show the very same distribution patterns, which are in the case of Raman maps caused predominantly by the radiation damage accumulated.
In addition to effects of structural damage on the PL broadening, however, minor effects of the chemical composition cannot be neglected. An example is the PL map in Figure 3a. The annealed half-zircon of sample A1 still shows a slightly heterogeneous FWHM distribution pattern. Nasdala et al. (2006) discussed their similar Raman observation for the very same sample as possible indication of incomplete structural reconstitution during the heat treatment (four days at 1300 °C). As an alternative interpretation we consider the possibility that minor heterogeneity in their Raman map, and our PL map, may be due to minor heterogeneity of the chemical composition (i.e., FWHM increase due to the elevated presence of non-formula elements). These heterogeneous areas are characterized by slightly, but significantly, elevated concentrations of P2O5, REE2O3, and U/ThO2 (see EPMA measurement point N3 from the sample A1 in Table 1).
The impact of compositional heterogeneity on the FWHM is supported also by the PL map of the very mildly radiation-damaged zircon M2 (Fig. 4), whose internal FWHM variations (between 14 and 25 cm−1 in the natural and between 12 and 15 cm−1 in the annealed half-crystal) are assigned in part to chemical heterogeneity. First, FWHMs are not uniform in the annealed crystal half (as it should be expected if FWHM broadening was solely due to structural damage). Second, the bright-BSE growth zone is characterized by both elevated concentrations of UO2, ThO2, and REE2O3 (Table 1) and the strongest PL band broadening (Fig. 4). Note that, due to its Neogene age (Harangi 2001), sample M2 has accumulated much less radiation damage compared to the zircon samples shown in Figure 3. Effects of radiation damage on the PL are hence less extensive in sample M2, and, correspondingly, the (minor) effects of the chemical composition on band FWHMs are detected because they are less obscured.
Effects of compositional-induced structural-disorder have been well documented already for the Nd3+ luminescence of monazite–(Ce) (e.g., PL study of Lenz et al. 2013). Note, however, that natural monazite–(Ce) typically has much higher compositional variability, when compared to natural zircon, which hence affects the FWHMs of PL emissions of monazite–(Ce) to much higher magnitudes. The impact of chemical heterogeneity on the FWHM of REE3+ Stark’s levels in natural zircon is, therefore, comparably low.
Reference zircon spectra: Estimation of radiation damage
Effects of radiation damage on the Dy3+ and Nd3+ luminescence of zircon are visualized and determined more quantitatively by PL data obtained from a suite of gem-zircon samples from Ratnapura, Sri Lanka, and Ratanakiri, Cambodia (Fig. 5). Note that the Sri Lankan zircon samples represent various degrees of radiation damage, ranging from moderately (M144) to strongly damaged (G3; see Table 2; compare also Nasdala et al. 2004a). The Ratanakiri zircon, in contrast, is characterized by a remarkably low degree of self-irradiation damage (Wittwer et al. 2013), which is mainly due to the sample’s young age of 1.2 ± 0.3 Ma (P.C. Piilonen, personal communication). Note also that the Sri Lanka zircon has experienced partial structural reconstitution in its geological history, which is why α-doses calculated based on the time period since the Neoproterozoic to Cambrian closure of the U–Pb system overestimate the radiation damage present (for details Nasdala et al. 2004a). To account for the annealing, “effective α-doses” were corrected applying a correction factor of 0.55 to the calculated total α-doses, as elucidated by Nasdala et al. (2004a).
In accordance to our findings described above, band widths of Dy3+ and Nd3+ luminescence sublevels show a marked increase that is closely related to the self-irradiation dose (Fig. 5). Note that the particular extent of the dose-related FWHM increase for a given Stark’s level band is not necessarily uniform but may vary appreciably among emission centers and even among individual sublevels of a single transition. As an example, FWHMs of two sublevels belonging to the 4F9/2 → 6H13/2 transition of Dy3+ (labeled I and II in Fig. 5a), show divergent band-broadenings upon increasing α-doses (Fig. 5c). Dysprosium sublevel I [FWHM increase from 12 cm−1 (Rata) to 87 cm−1 (G3)] obviously is more sensitive to radiation-induced structural disorder, compared to Dy3+ sublevel II [FWHM increase from 20 cm−1 (Rata) to 68 cm−1 (G3)] and the ~11 350 cm−1 sublevel of Nd3+ (4F3/2 → 4I9/2) [FWHM increase from 14 cm−1 (Rata) to 51 cm−1 (G3)], respectively (Fig. 5d). Consequently, for each single Stark’s level that is intended to be used as measure of the radiation damage, a separate calibration of the FWHM-dose dependence needs to be available.
The FWHM increase of REE-related emissions in mildly to strongly radiation-damaged (but not yet fully amorphous) zircon is assigned to disturbance of the structure as caused by the accumulation of defects. These defects are considered to perturb the local crystallographic environment of the REE3+ centers. As a consequence, the crystal field around REE cations is distorted, and PL bands may shift slightly in spectral position. The spectral band detected may then be regarded as superposition of a multitude of single levels arising from variably distorted crystallographic sites within the spatially resolved area analyzed (e.g., Skinner and Moerner 1996; Lenz et al. 2013).
Completely metamict ZrSiO4, and amorphous volumes in moderately damaged zircon, are in contrast characterized by randomly perturbed cation sites, which results in the entire degeneration of REE3+ crystal-field states. This in turn leads to the total loss of fine structure of electronic transitions, i.e., no more distinct Stark’s levels but broad single bands are observed in PL spectra. Analogous effects are observed in PL spectra of glasses produced by melt quenching; see for instance the PL spectra of Dy-doped CaTiSiO5 glass (degenerate transitions) and crystal (Stark fine-split transitions) in Figure 15 of Nasdala et al. (2004b). Consequently, the broadened (but still split) Stark’s level bands obtained from zircon samples studied here are assigned to the crystalline fraction present, whereas amorphous clusters (for a general description of the damage-accumulation process see Murakami et al. 1991) are virtually “PL-invisible” at low self-irradiation doses. At elevated radiation damage, fitting of individual Stark’s level bands is feasible as long as a minimum of remnant (even though strongly stressed) crystalline volume units contribute to the luminescence emission, and provided the REE signal originating from the amorphous volume fraction does not hamper too much the background.
Note, however, that FWHMs of luminescence bands also depend on temperature. In general, low-temperature luminescence studies are conducted to improve the sharpness of individual bands (for the narrowing of REE3+ emissions at low temperatures see, e.g., Reisfeld and Jørgensen 1977; Marfunin 1979; Blasse and Grabmaier 1994; Lenz et al. 2013). Figure 6a shows an exemplary low-temperature series of PL spectra of Dy3+ in zircon samples representing various degrees of radiation damage. The plot of FWHMs against temperature (Fig. 6b) however indicates that FWHMs of Dy3+ sublevels differ appreciably among variably radiation-damaged samples. In contrast, the effect of temperature on the FWHMs (i.e., narrowing with lowered temperature) is comparably low. The band FWHMs decrease by approximately about −8 cm−1 from ambient temperature to 78 K (virtually independent from the degree of the sample’s radiation damage).
The effect of temperature on widths of REE3+ emissions (i.e., 4f elements) is hence comparatively small when compared with d-d or d-f transitions. This is because f-electrons and their inter-transitions are well shielded by the outermost filled 5s25p6 orbitals (Blasse and Grabmaier 1994). Low-temperature PL measurements are, therefore, of insignificant usefulness to increase the reliability of FWHM estimates for strongly radiation-damaged zircon (see data for samples OR1 and G3 in Fig. 6b). The effective temperature-induced narrowing of Stark’s levels of these two samples does not exceed significantly the statistical error of the FWHM estimates. In addition, low-temperature measurements performed by using cooling stages or other liquid-nitrogen setups entail considerable effort; especially for measurements with high lateral resolution such as for hyperspectral PL mapping. Therefore, all reference luminescence spectra presented in the sections above are recorded at room temperature (298 K) as this study promotes the application of REE3+ luminescence spectroscopy being used potentially for various other techniques as well (such as CL coupled to EPMA). Nevertheless, for unbiased quantitative estimates, it needs to be ensured that luminescence analyses are done at room temperature or, alternatively, at the same temperature as the calibration measurements. Consequently, uncontrolled sample heating, such as due to intense local absorption of the incident laser light, must be avoided.
The REE3+ luminescence of zircon was found once again to be very sensitive to structural disorder caused by corpuscular self-irradiation. The emission intensity generally decreases, and band FWHMs increase, possibly accompanied by minor band shifts, with increasing structural damage. Our results show that the band-widths of distinct sublevels of Dy3+ (4F9/2 → 6H13/2) and Nd3+ (4F3/2 → 4I9/2) emissions (1) can be determined reliably by conventional band-fitting procedures and (2) is strongly related to the amount of radiation-damage accumulated (the latter being expressed by the effective time-integrated α-dose; Fig. 5). We propose the use of the FWHM of the Stark’s level at ~17 200 cm−1 of the 4F9/2 → 6H13/2 transition of Dy3+ (here described as sublevel I; see again Fig. 5c) and the dominant Stark’s level of Nd3+ near 11 350 cm−1 (Figs. 5b and 5d) as most suitable means of estimating quantitatively the radiation-induced structural disorder of zircon on a micrometer-scale.
Estimation of the crystallinity based on REE-band FWHMs may open up new opportunities to characterize the structural state of radiation-damaged zircon, complementary to the use of Raman-band FWHMs for this purpose (Nasdala et al. 1995, 2001). This is supported by Figure 7, which shows that the radiation-damage induced FWHM increase of the Dy3+ sublevel I is related to that of the ν3(SiO4) Raman band of zircon. Potential advantages of using REE3+ luminescence spectroscopy include (1) the high-volume resolution (with confocal spectrometer systems, analyses can be done on a micrometer scale), (2) the option to do non-destructive analyses without sample preparation, and (3) the possibility to obtain the FWHM of Stark’s levels of multiple REE3+ centers simultaneously (e.g., Dy3+ and Nd3+). The latter (i.e., choosing another center or transition) may be advisable if a Stark’s level under analysis should be obscured by other phenomena (e.g., defect-related broad-band emissions created upon corpuscular irradiation; Götze et al. 1999; Finch et al. 2004; Nasdala et al. 2013). The proposed PL of Dy3+ sublevel I may be used widely in studying zircon, because according to our experience, trace-Dy is virtually omnipresent in natural zircon, and even minute quantities yield well-measurable luminescence signals. Finally, similar to FWHMs of SiO4-related Raman bands, FWHMs of the Dy3+ (and Nd3+) signals studied here were found to depend predominantly on the degree of radiation damage present, whereas chemical effects, which may be much more extensive in minerals that typically are compositionally variable, are almost negligible in the case of (unaltered igneous or metamorphic) zircon.
Thanks to D.W. Davis, D. Frei, J.M. Hanchar, W. Hofmeister, A.K. Kennedy, I. Oláh, P. Piilonen, R. Stern, D. Talla, B. Wanthanachaisaeng, and M.J. Whitehouse for providing samples and/or data. Sample preparation was done by D. Dettmar, M. Tichomirowa, and A. Wagner. Critical comments of Adrian Finch and two anonymous reviewers are gratefully acknowledged. Financial support was provided by the Materials Science Research Centre, Mainz, Germany, and the Austrian Science Fund (FWF), grant P24448-N19 to L.N.