Chemical Geology 257 (2008) Contents lists available at ScienceDirect. Chemical Geology. journal homepage:

Transcription

1 Chemical Geology 257 (2008) Contents lists available at ScienceDirect Chemical Geology journal homepage: Determining high precision, in situ, oxygen isotope ratios with a SHRIMP II: Analyses of MPI-DING silicate-glass reference materials and zircon from contrasting granites R.B. Ickert, J. Hiess, I.S. Williams, P. Holden, T.R. Ireland, P. Lanc, N. Schram, J.J. Foster, S.W. Clement Research School of Earth Sciences, The Australian National University, Canberra, ACT, 0200, Australia article info abstract Article history: Received 26 June 2008 Received in revised form 21 August 2008 Accepted 24 August 2008 Editor: R.L. Rudnick Keywords: SIMS SHRIMP Zircon Oxygen isotopes Granite The development of new techniques and instrumentation on the ANU SHRIMP II ion microprobe has made it possible to measure the oxygen isotope ratios of insulating and conducting phases (e.g. silicates, carbonates, phosphates and oxides) on a 25 µm scale with better than 0.4 precision and accuracy at 95% confidence. Instrumentation changes include the installation of a multiple collector, charge neutralization using an oblique-incidence low-energy electron gun, and the addition of Helmholtz coils to counter mass dispersion by the Earth's magnetic field. A redesign of sample mounts and mount holders has effectively eliminated differences in variable isotope fractionation across the mount surface during analysis. Techniques have been developed to minimize the effect of electron-induced secondary ionization of oxygen. During a 6-minute analysis involving s of data collection, δ 18 O values can be measured on one 25 µm spot with an internal precision of better than 0.2 (2 standard errors). Analyses of MPI-DING silicate-glass reference material demonstrate that the external reproducibility of single spots can be better than 0.4 at 95% confidence, and that for matrix-matched samples and reference material, accuracy is commensurate with precision. MPI-DING glasses are acceptable ion microprobe reference materials for oxygen isotope measurements of glasses, although KL2-G is possibly heterogeneous. Zircon reference materials TEMORA 2 and FC1 appear to be acceptable as preliminary oxygen isotope reference materials. SHRIMP II analyses of FC1 indicate that it has a δ 18 O value of 5.4 (VSMOW). Analyses of zircon oxygen isotopic compositions from a gabbro, a tonalite and a granodiorite from southeastern Australia are presented. Zircon from the gabbro has a δ 18 O value of 5.6, the tonalite has an I-type affinity and slightly heterogeneous δ 18 O values around 6.6, and the granodiorite has an S-type affinity and a range of igneous, melt precipitated zircon δ 18 O values between 8.2 and These results suggest that the gabbro is mantle-derived and slightly contaminated with crustal material, and that the I-type granodiorite has evolved in a similar manner from a mantle-derived source. The δ 18 O values of the zircon from the S-type granodiorite are not only higher than from the I-type, but also more heterogeneous, consistent with partial melting of a poorly-mixed, metasedimentary source Elsevier B.V. All rights reserved. 1. Introduction SHRIMP (Sensitive High Resolution Ion MicroProbe) instruments have been used in a wide variety of applications in the Earth sciences. Principal among these is U Pb geochronology because of the presence of coherent age domains in the mineral zircon. SHRIMP has also been effective in isotope cosmochemistry and trace element abundance measurements. One of the first types of measurement carried out on SHRIMP was the analysis of stable S isotopes. Initially this was attempted with an Ar + primary beam and negative secondary ions (S ), Corresponding author. RSES, Building 61, Mills Road, Acton, ACT, 0200, Australia. Tel.: ; fax: address: Ryan.Ickert@ualberta.net (R.B. Ickert). although it was noted by Coles et al. (1981) that Cs + primary ions would produce much higher secondary ion yields. The success of the instrument for the analysis of positive ions, however, and its ability to make per mill measurements of S isotopes in that mode as well, stalled further development of negative secondary ion isotope ratio measurements. We have now revived negative ion work because of the clear benefits of making multiple types of analyses, including oxygen isotope ratio measurements, on single zircon domains (e.g., Mojzsis et al., 2001; Peck et al., 2001). A wide range of geological processes can modify the relative abundances of oxygen isotopes, making oxygen isotope ratios a valuable tracer in the geological and environmental sciences (Hoefs, 2004). Such processes include equilibrium isotopic exchange, and incomplete or unidirectional effects such as diffusion, evaporation/condensation, and metabolism. Geological applications of oxygen isotope analysis include studies of magma genesis, hydrothermal systems, ore genesis and /$ see front matter 2008 Elsevier B.V. All rights reserved. doi: /j.chemgeo

2 R.B. Ickert et al. / Chemical Geology 257 (2008) paleoclimatology (Sharp, 2007). Oxygen in extraterrestrial materials has provided key information about the origin of the solar system and nucleosynthesis (MacPherson et al., 2008). Oxygen isotope ratios of zircon have proved particularly valuable in studies of magma genesis because zircon can preserve the original oxygen isotopic composition of the melt from which it precipitated, even if the subsequent igneous rock has been heavily altered (Valley et al., 1994; Valley, 2003). Secondary Ion Mass Spectrometry (SIMS) is a technique that can be used to measure oxygen isotope ratios of solid samples in situ on the micrometre scale (Ireland, 2004 and references therein). SIMS works by focusing a beam of high-energy ions onto the solid under investigation, resulting in the ejection or sputtering of material of interest. The sputtered products include neutral atoms, molecular fragments, and a small fraction of ionized particles that can be introduced into a mass spectrometer, selected for energy and mass and measured by an ion counter or Faraday cup. Oxygen is strongly electronegative and therefore best measured as a negative ion, with strongly electropositive Cs + as the most common choice for primary ion species. When measuring insulating samples, the configuration of positive primary ions and negative secondary ions causes a build-up of a positive charge on the sample surface, despite a conductive coat applied to that surface. For in situ work, the charging problem is generally avoided by delivering electrons to the sample surface (Migeon et al., 1990; Hervig et al., 1992), although in special cases, small samples can be pressed into soft metals such as In or Au (McKeegan, 1987). Early oxygen isotope ratio measurement by SIMS instruments were limited by count times on the order of min, and a precision of about 1 (1σ; McKeegan, 1987; Valley and Graham, 1991; Hervig et al., 1992; Graham et al.,1996). These limitations were primarily a function of the inherently low transmission of the small turning radius SIMS (such as the Cameca IMS f series), the need for extreme energy filtering, and the measurement of oxygen isotope ratios dynamically on a single electron multiplier. The development of the large turning radius Cameca IMS1270 for oxygen isotope analysis (de Chambost et al., 1994; McKeegan et al., 1998) overcame these obstacles, providing both higher transmission and dispersion while generating stronger secondary beam currents (higher count rates) and permitting the use of static multiple collection (de Chambost et al., 1998). These instruments have achieved improved precision and reproducibility for substantially faster analyses. Many of the developments necessary for these measurements were specific to the Cameca ion optical configuration, for example, use of an electron gun with normal incidence in order to overcome the high field gradient in the secondary ion extraction region. Here we report major improvements to both the analytical techniques and the instrumentation on the ANU SHRIMP II that permit the high precision measurement of δ 18 O. Oxygen isotope ratio measurements of MPI-DING silicate-glass reference material and zircons used as U Pb reference material (TEMORA 2 and FC1) demonstrate that better than 0.4 (at 95% confidence) external reproducibility can be attained. As an example, we have applied these techniques to measure the δ 18 O values in zircon from two granites and a gabbro from the Lachlan Fold Belt, southeastern Australia. 2. Configuration of the ANU SHRIMP II for oxygen isotopic analysis The ANU SHRIMP II is a high mass resolution, double-focusing mass spectrometer which is based on an ion optic design by Matsuda (1974). The turning radii of the electrostatic and magnetic sectors are 1272 and 1000 mm respectively (Fig. 1). Although a secondary ion energy window is available on SHRIMP II, it is generally not used (i.e. is wide open) in isotope measurements because of the high quality of the chromatic refocus. SHRIMP II incorporates an astigmatic secondary ion beam-matching system to improve transmission through the source slit and mass analyzer. The geometry and ion extraction system are optimized for both high mass resolution and sensitivity. The SHRIMP II was commissioned in 1992 as a single collector instrument with a duoplasmatron primary ion source (Clement and Compston, 1990; Ireland et al., 2008). Both a Cs + source and multiple collection were envisaged for this instrument, with construction of a dual ion source and multiple collector soon thereafter. In a similar fashion to the development of a stable isotope analysis capability on the Cameca instruments, it has been necessary to install additional components on SHRIMP II in order to measure oxygen isotope ratios with high precision. These include a demountable Cs + primary ion source, electron gun, multiple collector, and Helmholtz coils around the source chamber. The configuration of this hardware is specific to the SHRIMP II geometry and is discussed below Cs + primary ion source The ANU SHRIMP II is equipped with a Kimball Physics IGS-4 alkali metal ion gun fitted with a Cs + firing unit. The gun can be removed and stored in air when not in use. Ions are produced by thermal ionization by heating a zeolite solid. The primary beam current is adjusted by changing the temperature of the ion source. The gun is mounted on an adaptor flange that makes it directly interchangeable with the Fig. 1. Schematic diagram of the main elements of the ANU SHRIMP II. Key components discussed in the text are labelled.

3 116 R.B. Ickert et al. / Chemical Geology 257 (2008) Fig. 2. Schematic diagram of the main elements of the secondary ion extraction system, electron gun, and Cs + source. The electron beam is deflected by the difference in potential between the sample and the intermediate extraction electrode, and is steered to be coincident with the impact point of the Cs + primary beam. Secondary ions are initially accelerated to 0.7 kv by the intermediate extraction lens held 2 mm from the sample surface. Full acceleration to 10 kv occurs after the intermediate extraction lens. After Ireland (2004). duoplasmatron ion source used for generating the oxygen (O,O 2 ) primary ions used in geochronology and trace element analysis. The adaptor places the point of Cs + emission in the same effective ion optic position as the point of O ion emission from the duoplasmatron. The source assembly is operated at close to ground potential, 10 kv above sample potential. Caesium ions are extracted from the gun at an additional 5 kv, giving the Cs + ions a total energy of 15 kev at the target (Fig. 2). Lenses and apertures in the primary column demagnify and collimate the Cs + beam before projecting a sharp-edged spot of uniform density onto the sample surface at an angle of 45. First, a pair of einzel lenses (one inside the Cs + source assembly and the other part of the normal primary column) works as a zoom-pair and project a magnified image of the object formed by the Cs + source onto the differential pumping aperture at the entrance to the source chamber. An ion image of that aperture is then projected forward by two einzel lenses that have an aperture (the Köhler aperture) inserted between them at the focal point of the second lens. In this form of Köhler illumination, the primary ion beam is focused to an evenly illuminated spot that is a demagnified image of the Köhler aperture (Ireland, 1995; Williams, 1998). During a typical analysis, a 3.5 na beam of Cs + is focused into a spot 25 µm diameter on the target surface, generating approximately 250 pa of O secondary ions from most silicate, phosphate and carbonate targets. The secondary ions are initially accelerated away by a low extraction voltage ( 700 V). The full acceleration to true ground potential (i.e kv relative to the sample) is produced by an extraction cone behind the extraction plate (Fig. 2) Electron gun The ANU SHRIMP II is equipped with a Kimball Physics ELG-5 electron gun. It is mounted on the side of the secondary extraction lens assembly in the source chamber, with the nose of the gun 20 mm Fig. 3. Schematic diagram showing the main components of the ANU SHRIMP II multiple collector.

4 R.B. Ickert et al. / Chemical Geology 257 (2008) from the sample surface at an incidence angle of 45 (i.e., 90 from the primary beam trajectory). Both the gun and its power supply are floated at column ground, nominally 10 kv relative to true ground (Fig. 2). The electron gun uses a refractory metal cathode to generate an electron beam with low-energy spread. Adjusting the cathode temperature controls the beam intensity. The electron emission energy is adjustable up to 3 kev. Under typical analytical conditions, the gun provides 1 µa of electrons to the target surface. The gun assembly includes two focusing elements that together operate as a zoom lens, plus two pairs of deflection plates to control the beam trajectory. The latter are particularly important as the electron beam is strongly deflected as it enters the secondary ion extraction field ( 350 V/mm). When focused, the electron beam, as imaged on a phosphorescent plate, is concentrated in an ellipse approximately µm. In practice, the electron energy is normally set in the range kev which, because of the 700 V secondary ion extraction potential, illuminates the analyzed area with electron energies in the range ev. At such energies the electrons have no observable effect on the conducting coating on the sample, or on the epoxy mounting medium (cf. Hervig et al., 1992). Electron energy is kept constant during an analytical session. The electron beam is adjusted to provide the highest intensity secondary ion current. Focusing and steering are very stable and the gun can be left on while moving the sample. Due to the stability of the charge neutralization, it is often the case that the electron gun does not need to be adjusted during an analytical session, and after a sample change only a small change in steering is required to reoptimize charge neutralization Multiple collector The SHRIMP II multiple collector (Ireland et al., 2008) has five collectors, three in a central array adjustable for single-mass-unit spacing over the mass range , and two moving heads on either side of the central array, with a maximum mass separation of 1 in 8 (Fig. 3). The two moving heads and the axial array are each equipped with a single Faraday cup, linked by a low-capacitance high-vacuum feed-through to an external Keithley 642 low-noise electrometer. The moving heads are used for 18 O/ 16 O analysis. The electrometers measuring 16 O and 18 O have Ω and Ω input resistors, respectively. When necessary, the Faraday cups can be interchanged with Sjuts sintered ceramic channel electron multipliers without breaking vacuum. The mass resolution required for measuring 18 O/ 16 O is relatively low. Potential isobaric interferences on 18 O from 17 OH and 16 OD can be resolved with resolutions of 2300 and 1830 respectively. This is achieved with 300 µm collector slits and a 150 µm source slit, which truncates the secondary ion beam by less than 5%. Representative peak shapes are shown in Fig. 4. An array of four beam-defining slits of different widths is provided for each collector (100, 200, 300, 400 µm on the moveable heads, and 50, 100, 200, and 300 µm on the central array), allowing mass resolution to be adjusted in steps from 1500 to N Helmholtz coils The ambient magnetic field of the Earth is strong enough to generate significant mass dispersion of oxygen isotopes between the sample and the source slit, and can induce a component of instrumental mass fractionation (IMF) that is sensitive to secondary ion steering. After extraction and acceleration from the sample surface, the secondary ions are focused by a triplet of quadrupole lenses before passing through the source slit and entering the secondary mass analyzer. Over the 350 mm from the sample surface to the source slit, the ions are subject to the ambient magnetic field of Fig. 4. Simultaneous mass scans of 16 O and 18 O showing flat-topped peaks and lack of interferences. The vertical axes are linear in (a) and (b), and logarithmic in (c). the Earth, the vertical component of which is 0.53 G in Canberra. This field generates mass dispersion, with 16 O being deflected about 3.5 µm more than 18 O. As the edges of the secondary ion beam are truncated by the source slit, this introduces a component of IMF that is very sensitive to the horizontal steering of the beam. A similar effect has been reported on other large radius ion microprobes (e.g., Schuhmacher et al., 2004). The installation of coupled horizontal Helmholtz coils above and below the source chamber has enabled the cancellation of the terrestrial magnetic field and effected a substantial reduction of IMF at the source slit of SHRIMP II. The IMF is relatively insensitive to the vertical steering; no compensation is required in that direction. The effectiveness of the coils in reducing steering-dependent isotope fractionation is illustrated in Fig. 5. Without coils, the fractionation of measured 18 O/ 16 O caused by changing the secondary beam horizontal steering enough to truncate the beam by about 10% is 58 for a 50 µm source slit, falling to 32 for a 150 µm slit. At a coil current of 0.38 A (corresponding to a magnetic field strength of 0.5 G), the fractionation for the same deflection for all slit widths falls below detection

5 118 R.B. Ickert et al. / Chemical Geology 257 (2008) the very centre of the mount. Only within 4 mm of the mount edge, where the mount did not fully cover the 8 mm diameter extraction aperture, did significant fractionation become apparent Electron-induced secondary ion emission (EISIE) Fig. 5. Drift of secondary ions from the sample to the source slit through the Earth's magnetic field results in mass dispersion, which is manifested by a change in the measured isotope ratio as a function of ion beam steering against the source slit. The magnitude of the difference in measured isotope ratio between a secondary beam that is directed at the centre of the source slit and a secondary beam that is slightly offset is a function of the magnitude of the mass dispersion. This difference (for three different sized slits) is plotted vs. the current in the Helmholtz coils, which produce a different strength magnetic field according to coil current. The difference in measured isotope ratio (here depicted as δ 18 O values) between a secondary beam going directly through the slit, and one that is slightly offset, is zero at a coil current of 0.4 A. This is the current at which the Helmholtz coils produce a magnetic field that completely compensates for the Earth's magnetic field. When using the SHRIMP II electron gun during the measurement of insulating materials, there is often a small but measurable O signal that can be detected at the multicollector when the Cs + beam is deflected away from the target, the Cs + source power supply is off, or the Cs + source is physically isolated (by a gate valve) from the target. This signal is not present when measuring oxygen isotope ratios in conducting materials and the electron gun is not used, nor is it present when the electron gun is running on a completely fresh, insulating target before the Cs + source has been turned on. The intensity of the O signal is positively correlated with the proximity of the electron (b1 ). As the coil current is increased further, the fractionation increases, but in the opposite sense. By combining the use of Helmholtz coils with a 150 µm source slit, variable IMF as a function of horizontal beam steering has been effectively eliminated. This has desensitized the IMF to small adjustments in secondary beam steering that are still required to compensate for microtopography on the target surface and small differences in positioning the target plane Redesign of sample mounts The standard SHRIMP II sample mount consists of a 25 mm diameter epoxy disc supported by a stainless steel mount holder. The face of the disc sits flush behind a thin (250 µm) flange in the mount holder, so that the surface presented to the face of the extraction lens is a 22 mm diameter Au-coated polished epoxy surface, surrounded by a 6 mm wide stainless steel annulus, with a 250 µm step in between. The IMF of oxygen secondary ions across even the central 10 mm square of this sample mount was found to be large, with fractionations in 18 O/ 16 O relative to the mount centre ranging from 4.5 at the NE corner to +2.9 at the NW corner (Fig. 6). Within 2 mm of the inner edge of the stainless steel annulus the fractionations were extreme, 48 in the E and +22 in the W (not shown in Fig. 6). To test whether the fractionations were related to the 250 µm step near the mount edge, the presence of the stainless steel, or both, a custom mount was prepared with a recessed edge such that the epoxy and stainless steel flange defined a flat plane, eliminating the 250 µm step. The result was to remove the extreme edge effects close to the flange and the E W gradient in relative mass fractionation, but the fractionation gradient from centre to edge remained (about +4, Fig. 7). This showed that the step in the mount surface produced the pronounced mount edge effect, but that the contrast in material between Au-coated epoxy and stainless steel also was a major contributing factor and would still be a problem for SHRIMP II analysis. To eliminate both the step and the material contrast between the sample mount and its holder, the epoxy mount was redesigned to be larger (35 mm diameter) and to attach to the face of the mount holder. This megamount presented a uniform surface to the extraction lens with no internal boundaries. The result was to effectively eliminate the isotopic gradient over the central 19 mm (Fig. 8; Table 1). Isotopic variation at any point in this area is the same as what is expected at Fig. 6. Conventional mount holders for epoxy grain mounts present a material and topographic contrast to the extraction field. This introduces a gradient in instrumental mass fractionation across the mount and extreme, asymmetric edge effects near the steel. Black diamonds are zircon grains with a size exaggerated for clarity. In the cross section, the distance by which the steel sits proud of the mount face is exaggerated for clarity.

6 R.B. Ickert et al. / Chemical Geology 257 (2008) Electron-induced secondary ion emission of oxygen is observed on SHRIMP II for a wide range of insulating target materials and operating conditions. It effectively results in the simultaneous measurement of two sources of oxygen: one directly sputtered from the spot by primary Cs +, and a second produced by desorption over a wider area. The EISIE is temporally variable, particularly during the sputtering Fig. 7. An experimental mount holder was designed to present a flat face to the extraction field where the epoxy is flushed with the steel. This design eliminated the extreme edge effects of the conventional mount holder, but the isotopic gradient across the face of the mount remained. bombardment area to any sputtered crater (Fig. 9), and with electron energy. There is no sample damage (e.g., removal of conductive coat) associated with this effect. We suspect that this anomalous secondary signal is the result of electron-stimulated desorption ESD (Madey, 1986). We refer to the specific effect seen on SHRIMP II as electron-induced secondary ion emission (EISIE). ESD is the production of low-energy ions and neutrals through the electronic excitation of terminal bulk atoms or adsorbed monolayers by electron bombardment (Madey, 1986). It is commonly observed when an electron gun is used to compensate for charging during SIMS analysis of electrical insulators (Williams, 1981; Williams et al., 1983; Williams and Gillen, 1987). ESD can be a serious problem, for example inhibiting the analysis of fluorine (e.g. McPhail et al., 1986) and carbon (e.g. de la Mata and Dowsett, 2007), or a useful tool, for example providing a means of visualizing the electron beam when adjusting an electron gun (e.g. Reger et al., 1997). Fig. 8. The newly designed megamount attached to the front of a modified sample holder eliminates topographic and material contrast across the analytical surface. Mass fractionation associated with position on the mount is only significant near the edge, where the extraction aperture overlaps with the edge of the mount.

7 120 R.B. Ickert et al. / Chemical Geology 257 (2008) Table 1 Analyses of zircon in transects across the face of a megamount Spot a δ 18 O b X1 7.9 X2 8.0 X3 7.7 X4 7.6 X5 7.8 X6 7.8 X7 8.2 X8 7.7 X9 8.2 X X X X X X Y1 8.3 Y2 9.2 Y3 8.9 Y4 8.6 Y5 8.7 Y6 8.4 Y7 8.2 Y8 8.6 Y9 8.7 Y Y Y a Numbers represent horizontal (X) positions from left to right on the mount, and vertical (Y) positions from top to bottom. b Data are normalized to a mean δ 18 O value of 8.2. δ 18 O( )=(( 18 O/ 16 O Sample )/( 18 O/ 16 O VSMOW ) 1) O/ 16 O VSMOW = (Baertschi, 1976). process when it increases approximately and exponentially with time (Fig. 10). The desorbed oxygen is isotopically fractionated by approximately 200 relative to the Cs + sputtered oxygen and has a much lower mean energy and energy range. Despite the low intensity of the EISIE usually it contributes about 0.1% of the total measured 16 O ions its fractionated isotopic composition and temporal and spatial variability mean that it has the potential to affect analyses of target oxygen isotope ratios. We have found that this effect is substantially reduced by using an Al Fig. 9. Electron-induced secondary ion emission (EISIE) occurs near a sputtered region, but dies away to a background EISIE within a few hundred micrometres. The data for this figure was collected by sputtering the zircon grain for several minutes, then turning the Cs + beam off, moving the stage to different positions around the sputtered area and measuring the count rate of 16 O due only to electron-induced ionization. The contours were drawn by hand around the original spot analyses, and are labelled in per cent relative to the maximum EISIE count rate determined near the analytical spot. Fig.10. Electron-induced secondary ion emission (EISIE) occurs when the electron beam ionizes adsorbed oxygen on the sample surface. The measured abundance of these ions increases approximately and exponentially with time, although it is low during the length of a typical analysis. The data here were collected by sputtering a zircon under normal analytical conditions, and periodically turning off the Cs + beam and measuring only the 16 O ions desorbed by the electron beam. Time zero is when sputtering began. conductive coat, rather than the Au coat that is used for most SHRIMP II work, and by keeping the electron energy as low as possible while still maintaining charge neutralization. In addition, analytical strategies have been adopted such that spots in both unknowns and standards, where possible, are positioned so that the relative spacing is similar or identical so that there is no bias between sets of analyses. Finally, the EISIE intensity is usually measured 3 times during an analysis, (by turning off the Cs + and measuring the remaining 16 O intensity) providing a check that the effect has been properly minimized and data from which peak-stripping can be carried out, although this is commonly not required (see Section 2.9 and 3.2 for examples) Analyses Under the analytical conditions described above, the Cs + beam removes the 12 nm Al coating at the analytical site within a few seconds and the secondary ion current rises steadily for about 3 min, after which the rise slows or the beam steadies and then slowly declines, depending on the matrix. During the rapid rise phase, the 18 O / 16 O also rises as steady state conditions between the isotopic composition of the analyzed area, secondary ion beam, and implanted Cs become established (Fig. 11). Before the secondary beam stabilizes, the primary beam is rastered across an area slightly larger than the analytical pit for 30 s in order to remove any surface contamination. The spot is then sputtered for up to an additional 90 s without rastering. After the secondary beam stabilizes, the secondary ion steering is optimized, the secondary ion beams are centred in the collector slits and the data collection commences. Isotope ratios are normally measured for s as one set of ten, or two sets of five to seven, 10-second measurements, with re-optimization of the beam centring on the source and collector slits between the two sets. The EISIE is monitored throughout, with measurements taking place: 1) a few seconds after the primary beam is turned on, 2) before data collection starts, and 3) after the isotope ratio data have been collected. Typical count rates on 18 O are about cps, and on 16 O about cps. Background count rates for 18 O and 16 O are about cps (10 11 Ω resistor) and cps (10 10 Ω resistor) respectively, and are measured at the start and end of each analytical session IMF and gain correction Due to the large IMF in SIMS, accurate isotope ratio measurement requires standardization to a matrix-matched reference material of

8 R.B. Ickert et al. / Chemical Geology 257 (2008) Fig. 11. Plot of the time-dependent rise in 18 O/ 16 O following first-impact of a 3 nacs + ion beam on an Al-coated grain of TEMORA-2 zircon. Data collection commences once the signal has stabilized, normally after about 3 min. Each data point represents a count time of 10 s, and error bars are at 1SD and are based on counting statistics. More scans are shown here than taken for a normal analysis, to demonstrate the negligible change in the ratio after about 200 s. Fourteen spots were analyzed on each of the six glasses, for a total of 84 individual oxygen isotope analyses (Table 1). Oxygen isotope ratios are reported in terms of the IMF, as the per mill deviation of the gain and background corrected 18 O/ 16 O from the true 18 O/ 16 O (e.g., Eiler et al., 1997). Because the difference between the true isotopic compositions of the materials and VSMOW is small (a maximum of 9.4 ), the size of a 1 variation in IMF is nearly identical to a 1 variation in a δ 18 O value relative to VSMOW. Each analysis took less than 6 min, and the results are illustrated in Fig. 12. Ten scans, of 10 s each, were measured for each analysis, and count rates on 16 O ranged from 2.5 to cps. Electron-induced secondary ions accounted for approximately 0.03% of the total secondary beam, and induced very small changes in the measured isotopic ratios, on the order of 0.1. The change is systematic for all spots. For each spot, the additional uncertainty due to fluctuations in the EISIE is very small and on the order of 0.01 ; as such, no correction was applied. The precision of an individual analysis (i.e. an individual spot) can be described by three different values in an hierarchical manner (e.g., Stern and Amelin, 2003); in decreasing order of precision they are, internal precision, spot-to-spot precision, and external precision. For the following discussion, we define the standard deviation, or dispersion of the data, as SD, and the standard error (or standard deviation of the mean, SD/ N, where N is the number of samples), or known and nominally homogeneous isotopic composition. During an analytical session, analyses of the reference material are interspersed with those of unknown samples. The analyses of reference material serve to determine the magnitude of the IMF, which can change on a day-to-day basis, and to monitor any changes during an analytical session. Normal practice is to measure the reference material (standard) several times at the beginning of an analytical session, and thereafter following every three to five sample analyses. Both IMF and relative gain between electrometers are corrected by normalizing all background corrected data to the measured isotopic composition of the reference material. For single collector data, following Eiler et al. (1997), the IMF can be depicted as an α ratio α u R True Std =RMeasured Std where R= 18 O/ 16 O. In this case, the measured isotope ratio of a sample is corrected by simply multiplying it by the α-factor. For multiple collection, the relative gains of the two collectors must be taken into account. The gain correction takes an identical mathematical form, and therefore by referencing it to a known isotope ratio, both gain and IMF can be corrected at once. Usually, IMF and gain are corrected in a sample by multiplying the measured, background corrected 18 O/ 16 O by the ratio of the true 18 O/ 16 O of a standard to the measured 18 O/ 16 O of the standard. Blocks of data where the IMF changes in step-wise shifts are treated separately, and corrections are applied when the gain or IMF drifts with time. Such shifts have proved to be infrequent on SHRIMP II, and are generally related to definable external factors such as high voltage fluctuations or temperature variation Precision and accuracy: measurements of MPI-DING reference material To assess the instrument performance, including precision, accuracy, IMF magnitude and sensitivity to the matrix, a series of analyses was run on six of the MPI-DING natural-silicate-glass reference materials (Jochum et al., 2006). The MPI-DING glasses included were the basaltic KL2-G and ML3B-G, komatiitic GOR132-G, andesitic StHs6/80-G, rhyolitic ATHO-G, and dioritic T1-G. Jochum et al. (2006) have reported bulk oxygen isotope ratios measured by the laser fluorination method for each sample to a precision ranging from 0.07 to 0.22 (2σ). Fig. 12. Plot of MPI-DING glass analyses, in chronological order as they were analyzed. Error bars represent internal precision at 95% confidence. Data is reported in terms of the IMF as defined in the text.

9 122 R.B. Ickert et al. / Chemical Geology 257 (2008) the dispersion of the mean, as SE. Confidence limits are determined by multiplying the SD or SE by a Students-t factor for a particular number of samples. All values of SD and SE here are reported at 95% confidence limits. The internal (or within-spot ) precision is defined as the SE of a set of isotope ratios measured during the course of a single spot analysis. For the MPI-DING analyses, the internal precisions ranged from 0.08 to 0.28, with a median value of 0.16 (Fig. 13). These values do not correlate with count rates, secondary ion beam steering or glass composition. For the count rates and measurement times during this session, the limit of precision dictated by counting statistics is approximately 0.04, meaning that the internal precisions measured here approach the theoretically permissible limit (e.g. Fitzsimons et al., 2000). It is apparent from Figs. 12 and 13 that for all glasses, the internal precisions are not able to account for the dispersion in the data. This contrasts with U Pb analyses of zircon on SHRIMP II, where internal precisions commonly account for all the scatter in several measurements of nominally homogeneous reference zircon. The spot-to-spot precision is the dispersion of individual analyses of a nominally homogeneous material. It is determined by calculating the SD of the measured reference material oxygen isotope ratios for a particular session. The designation of material as nominally homogeneous is important, as there is no independent method of determining oxygen isotope homogeneity on a scale similar to, or smaller than, the sample size of a SIMS analytical spot. The spot-tospot precisions of the MPI-DING analyses are depicted in Fig. 14. Four of the glasses have a spot-to-spot precision (i.e., standard deviations at 95% confidence limits) near 0.4, one slightly higher at around 0.6, and one much higher at 1.1. These values are two to three times larger than the typical internal precisions (Fig. 13). There is a large discrepancy between the internal precisions of the individual analyses and the actual dispersion of data for each glass. This requires that either 1) the glasses are heterogeneous in oxygen isotope composition at the scale of 's of µm 3, 2) the internal precisions do not adequately reflect the true uncertainty on the measurements of the isotopic composition of individual spots, or 3) a combination of both. If there is an additional component of analytical uncertainty associated with the measurement of different spots (case 2, above), then it should be the case that the dispersion of the data for each glass should be identical (provided the additional uncertainty is not related to differing compositions). Bartlett's test for equal variances is a parametric statistical test that can determine the probability that all of the samples have equal variances (Snedecor and Cochran, 1989). When all six samples are taken together, they fail a Bartlett's test with a probability less than 0.01%, but pass with a probability greater than 5% when analyses of glass KL2-G are removed. To confirm this result using another technique that does not require normality, 95% Fig. 13. Histogram of internal precisions for the MPI-DING glass data. The best precision possible at these count rates and measurement times is depicted by the light grey band at 0.08, and the reproducibility is depicted by a light grey band at Fig. 14. Spot-to-spot precision of MPI-DING glasses. Values plotted are the standard deviations (at 95% confidence), and the error bars are the 95% confidence limits on these standard deviations. The error bars have been generated by a bootstrap, and are asymmetric. confidence limits on the range in the standard deviations were constructed using a bootstrap (Wehrens et al., 2000) with 1000 random re-samples. The results of the bootstrap are depicted in Fig. 14 and confirm the results of the Bartlett's test, sample KL2-G clearly has a significantly different degree of scatter from the remaining six samples, and the spot-to-spot precisions estimated using those six samples are statistically equivalent. It is likely that glass KL2-G has heterogeneous oxygen isotope ratios on a scale and of an amplitude that can be measured by SIMS. As has been found previously for SIMS analysis, the internal precisions significantly underestimate the true analytical uncertainty. The excess uncertainty is most likely due to slight differences in charge compensation and secondary ion beam trajectory when the sample is moved. One method to accommodate the extra uncertainty would be to expand the internal precisions by an excess scatter term, for example the square root of the reduced chi-squared (χ 2 /ν; or MSWD in the geochronological literature: Wendt and Carl, 1991; Bevington and Robinson, 2003). This is a useful recourse when the spot-to-spot precision approaches the internal precision, but for larger values of the former the excess scatter term dominates. Additionally, the variation in internal precision (e.g., Fig. 13) probably reflects simple statistical fluctuation generated by the small number of scans, and not variation in the confidence with which the mean of the scans is known. A simpler solution is to use the spot-to-spot precision (the standard deviation of a series of analyses on a nominally homogeneous standard) as the true uncertainty in the measured isotope ratio of any given analytical spot. This solution is the one preferred here, if only for simplicity, and is similar to that in use by other laboratories (e.g., Page et al., 2007; Trail et al., 2007). We expect that with improvements to reproducibility our treatment of uncertainties will evolve. Finally, the external precision takes into account the uncertainty in the IMF correction, both in terms of precision of the measured IMF (as measured on the reference material), and the uncertainty in the true isotopic composition of the reference material. There are two ways that this external precision can be presented. If the data of interest are analytical results from single spots, the external precision on any one spot is the spot-to-spot precision, the SE of the reference material analyses, and the uncertainty in the true isotopic composition added in quadrature. If the data of interest are a pooled set of spots (for example, during the determination of the oxygen isotopic composition of a homogenous material measured by multiple spots) then the SE of the sample analyses, the SE of the reference material analyses, and the uncertainty in the true isotopic composition of the reference material are added in quadrature. For single analyses, where the true isotopic composition of the reference material is known to better than 0.1, this results in an inflation of the single spot analytical uncertainty by only

10 R.B. Ickert et al. / Chemical Geology 257 (2008) Accuracy and IMF Accurate SIMS isotope ratio measurements depend on how well the IMF can be measured and how much it differs between the reference material and material of unknown isotopic composition. Including the scatter of the individual analyses of each MPI-DING glass sample, and adding the uncertainty in the true isotopic composition (Jochum et al., 2006) in quadrature, the precision of the individual determinations of IMF range from 0.13 to 0.30, at 95% confidence limits. The IMF of the MPI-DING glasses varies strongly with composition. This illustrates the well-known matrix problem in SIMS, whereby the IMF varies strongly with the composition of the material that is being measured (Shimizu and Hart, 1982; Eiler et al., 1997). This relationship is illustrated in Fig. 15 by plotting the IMF vs. SiO 2 concentration. The IMF varies linearly from rhyolitic to komatiitic compositions (76 46 wt.% SiO 2 ) over a range of IMF of approximately 7. This relationship emphasizes the need for matrix-matched samples and points the way to a metric for how close in composition samples and unknowns need to be. The two basaltic composition glasses, KL2-G (despite probably being slightly heterogeneous) and ML3B-G provide an example of measuring two matrix-matched samples. The measured difference in isotope ratios agrees well within error, suggesting that when using reference material of identical compositions to unknown material, accuracy can be commensurate with precision to better than SE Australian granites The Paleozoic granites of southeastern Australia, where I- and S- type granites were first defined, have been intensely studied for over three decades (Chappell and White, 1974; Kemp et al., 2007) and have become textbook examples of granite magmatism (e.g., Winter, 2001). Despite these studies, however, the nature and origin of these granites continue to be controversial. Extant problems include the relative roles of open- and closed-system processes in producing the granites and their compositional variation, and the genetic relationship between the I- and S-type granites. Measurement of oxygen isotope ratios in zircon by ion microprobe can be a powerful tool in the study of granitic rocks. Although rocks can easily have their bulk oxygen isotopic compositions modified by exchange with water during alteration, the chemical stability of zircon and its very low oxygen diffusion rate (Page et al., 2007) make it highly resistant to 18 O exchange with hydrothermal fluids (King et al., 1997). In addition, when the slow diffusion of oxygen is coupled with the long residence time of zircon in magma chambers (Matzel et al., 2006), it becomes possible to track the isotopic evolution of a magmatic system through inter- or intra-grain variation in zircon Fig. 15. Variation of instrumental mass fractionation (IMF) with weight percent SiO 2. The IMF varies linearly with respect to chemical composition. Error bars are at 95% confidence on the mean, and include the uncertainty in the true isotopic composition of the sample. Where error bars are not present, they are smaller than the symbols. Table 2 Measured, apparent oxygen isotope ratios in MPI-DING reference glasses Spot Name Spot Number 18 O / 16 O a 2σ b IMF ( ) 2σ b T ± ±0.08 T ± ±0.16 T ± ±0.19 T ± ±0.16 T ± ±0.27 T ± ±0.11 T ± ±0.22 T ± ±0.14 T ± ±0.15 T ± ±0.22 T ± ± 0.23 T ± ±0.16 T ± ±0.17 T ± ± 0.13 Mean ± ±0.39 GOR ± ±0.11 GOR ± ± 0.20 GOR ± ±0.17 GOR ± ±0.11 GOR ± ±0.10 GOR ± ± 0.16 GOR ± ±0.16 GOR ± ± 0.14 GOR ± ±0.08 GOR ± ±0.12 GOR ± ±0.17 GOR ± ±0.14 GOR ± ±0.19 GOR ± ± 0.16 Mean ± ±0.42 StH ± ±0.16 StH ± ±0.12 StH ± ±0.16 StH ± ± 0.12 StH ± ±0.17 StH ± ±0.16 StH ± ±0.16 StH ± ±0.10 StH ± ±0.11 StH ± ±0.19 StH ± ±0.20 StH ± ±0.18 StH ± ±0.25 StH ± ±0.20 Mean ± ±0.37 ML3B ± ±0.28 ML3B ± ±0.18 ML3B ± ±0.13 ML3B ± ±0.14 ML3B ± ± 0.12 ML3B ± ±0.12 ML3B ± ±0.16 ML3B ± ± 0.15 ML3B ± ±0.17 ML3B ± ±0.17 ML3B ± ±0.14 ML3B ± ±0.18 ML3B ± ±0.19 ML3B ± ±0.19 Mean ± ± 0.57 ATHO ± ±0.15 ATHO ± ±0.14 ATHO ± ±0.10 ATHO ± ±0.15 ATHO ± ±0.14 ATHO ± ±0.13 ATHO ± ±0.14 ATHO ± ±0.21 ATHO ± ±0.22 ATHO ± ± 0.14 ATHO ± ±0.09 ATHO ± ±0.14 ATHO ± ± 0.12 ATHO ± ±0.08 (continued on next page)

11 124 R.B. Ickert et al. / Chemical Geology 257 (2008) Table 2 (continued) Spot Name Spot Number oxygen isotope ratios. Oxygen isotope ratios are particularly useful for determining the relative contributions of juvenile and sedimentary components to granitic magma systems, because the bulk δ 18 O of the mantle has a low variance at approximately 5.5 (Mattey et al., 1994), and bulk sedimentary rocks commonly range from 12 to 15. We have measured oxygen isotope ratios in zircon from three granitoids from the Kosciuszko Batholith in southeastern Australia: the mafic Blind Gabbro, the intermediate composition I-type Jindabyne Tonalite, and the S-type Jillamatong Granodiorite. All are of Siluro Devonian age (ca Ma) and have been placed in separate granite suites. The geology and petrography of these samples are described by Hine et al. (1978) and White and Chappell (1989). The bulk oxygen isotopic compositions of I- and S-type granites in the adjacent region have been described by O'Neil and Chappell (1977). I- type granites range from 7.9 to 9.4, whereas the S-types have a narrow range of 9.9 to These compositions have been interpreted to reflect the dominantly igneous and sedimentary precursors of I- and S-type granites, respectively Analytical procedure 18 O / 16 O a 2σ b IMF ( ) 2σ b Mean ± ±0.39 KL ± ±0.19 KL ± ±0.20 KL ± ±0.22 KL ± ±0.15 KL ± ±0.16 KL ± ±0.14 KL ± ±0.13 KL ± ±0.21 KL ± ±0.13 KL ± ±0.12 KL ± ±0.10 KL ± ±0.11 KL ± ±0.24 KL ± ±0.24 Mean ± ±1.07 IMF ( )=(( 18 O/ 16 O measured )/( 18 O/ 16 O true ) 1) Compositions of MPI-DING glasses reported by Jochum et al. (2006), inδ 18 O values are: T1-G=7.53±0.07; GOR132-G=8.52±0.08; StHs6/80-G=6.12±0.07; ML3B-G=8.35±0.22; ATHO-G=3.20±0.07; KL2-G=8.63±0.09. δ 18 O ( )=(( 18 O/ 16 O Sample )/( 18 O/ 16 O VSMOW ) 1) O/ 16 O VSMOW = (Baertschi, 1976). a This is the measured 18 O/ 16 O, corrected for gain and background but not instrumental mass fractionation. b Internal error. Approximately 2 kg of rock were sampled by sledgehammer or drill from fresh boulders or roadcuts. Weathered surfaces were removed, and approximately 0.5 kg of each sample were reduced by hand to 1 cm sized pieces, followed by fine crushing in a tungsten carbide swing mill and sieving to b250 µm. Very fine material was decanted in water and the samples dried under a heat lamp. Zircon was separated from the remaining powder, using standard heavy liquid and magnetic techniques. Zircon was hand picked under a binocular microscope and mounted on double-sided tape prior to casting in an epoxy megamount as described above. Zircon grains from the granite samples were mounted along with zircon reference materials TEMORA 2 and FC1, and all grains were contained within an 8 mm by 8 mm square at the centre of the mount. The mount was coarsely polished to expose the grains just above their centres and brought to a fine polish using 1 µm diamond paste. The zircon grains were then all photographed in reflected light, transmitted light, and by scanning electron microscope cathodoluminescence (SEM-CL). U Pb ages were measured on SHRIMP II following the procedures of Williams (1998). The results will be presented in a separate publication as part of a regional study. The analyzed grains were then individually imaged by SEM-CL, and then polished again to Fig. 16. Results of an oxygen analytical session on the ANU SHRIMP II measuring the isotopic composition of FC1 relative to TEMORA 2. The IMF is very well controlled and does not drift with time. The data are normalized to a δ 18 O value of 8.2 for TEMORA 2. The measured difference between FC1 and TEMORA 2 is 2.8 ±0.3 at the 95% confidence level. remove the approximately 2 µm deep sputtering pits and implanted oxygen. A separate analytical session was run to establish the oxygen isotopic composition of the FC1 zircon for use as a secondary standard. TEMORA 2 zircon (δ 18 O=8.2 ; Black et al., 2004) was analyzed along with FC1 zircon (Paces and Miller, 1993). The results are presented in Table 2 and plotted on Fig. 16. No corrections for IMF/gain drift or EISIE were necessary. Oxygen isotope analyses of FC1 on SHRIMP II, normalized to TEMORA 2, yield a mean δ 18 O value of 5.4±0.3. Laser fluorination oxygen isotope ratio analyses of zircon from similar rocks by Booth et al. (2005) and Trail et al. (2007) also yield a mean δ 18 O value of 5.4 (neglecting one low outlier) Results Oxygen isotope ratios were measured in zircon crystals as close as possible to the spots analyzed for U Pb, using the techniques described in Section 2. Each measurement consisted of ten, 10 s integrations. Baseline count rates on the electrometers were measured at the beginning of the session and have been subtracted from all the data presented here. Average count rates for 16 O and 18 O were cps and cps, respectively. Corrections for IMF and detector gains were carried out as described above and all results were normalized to a δ 18 O value of 8.2 for TEMORA 2. No correction for IMF drift or gain drift was necessary. Fig. 17. Plot of uncorrected data vs. data that have had the EISIE (Electron-Induced Secondary Ion Emission) contribution removed. The regression is represented by the solid grey line and is nearly indistinguishable from a line of unity, indicating an insignificant effect of the EISIE correction on these analyses.