American Mineralogist, Volume 93, pages 187–197, 2008

0003-004X/08/0001–187$05.00/DOI: 10.2138/am.2008.2658 187

* E-mail: S.Reddy@curtin.edu.au

Electron backscatter diffraction analysis of zircon: A systematic assessment of match unit characteristics and pattern indexing optimization

Steven M. reddy,1,* nIcholaS e. tIMMS,1 and Bruce M. eglIngton2

1Department of Applied Geology, Curtin University of Technology, GPO Box U1987, Perth, WA 6845, Australia2Saskatchewan Isotope Laboratory, Geological Sciences, University of Saskatchewan, 114 Science Place, Saskatoon, Saskatchewan S7N 5E2,

Canada

aBStract

Quantitative microstructural analysis of zircon using electron backscatter diffraction (EBSD) re-quires a comparison of empirically collected electron backscatter patterns with theoretical patterns or “match units” derived from known crystallographic parameters. There are 23 possible crystallographic data sets for zircon, and associated match units, derived from natural and synthetic zircon and from theoretical calculations over a range of pressures and different rare earth element (REE) compositions. A systematic assessment of these match units has been undertaken by EBSD analysis of each of four zircons from a range of geological environments combined with principal components analysis and self-organizing map networks. Comparison of the different match units shows a systematic relation-ship across all samples that are related to changes in unit-cell dimensions associated with pressure and compositional variations. Systematic variations in the data generated from 96 EBSD maps, each comprising 10 000 electron backscatter patterns, indicate that match units associated with increasing pressure or REE dopants yield poorer quality EBSD data. The match units from low-pressure, undoped, natural zircon consistently yield the best EBSD results and are recommended for natural zircon EBSD studies irrespective of the zircon source or U content. The results provide a clear strategy for optimizing the acquisition and analysis of EBSD data from zircon from both crustal and mantle sources. In addi-tion, the developed approach to match unit analysis may be applied to all other crystalline materials, potentially optimizing EBSD analyses from a range of materials.

Keywords: EBSD, microstructure, zircon, reflector file, match unit, REE, pressure, rare earth element

IntroductIon

The mineral zircon (ZrSiO4) is widely used by the earth science community to provide geochemical and temporal constraints on a range of important geological processes. The justifications for its widespread use are its ability to incorporate a wide range of elements into its lattice (e.g., the radioactive element, U) and its ability to retain compositional information to temperatures approaching 1000 °C. Despite the successful application of zircon geochemical and geochronological data to various disciplines, recent studies have documented that zircon may deform by crystal plasticity within the Earth’s crust (Reddy et al. 2007). Studies linking quantitative microstructural and geochemical analysis of an Indian Ocean zircon illustrate that crystal plasticity of zircon may facilitate the geochemical modification of zircon rare earth elements (REE) at tempera-tures significantly below those suggested for volume diffusion (Reddy et al. 2006). In a similar study, U-Th variations have also been linked to deformation-related crystal plasticity in zircon (Timms et al. 2006). These findings have the potential to lead

to the refinement of geochemical models and the development of new applications of zircon geochemical data to earth-science research. However, such applications require the quantitative and non-destructive microstructural analysis of zircon by electron backscatter diffraction (EBSD) to allow integration of orientation data with high-spatial resolution geochemical information.

Electron backscatter diffraction is a quantitative microstruc-tural technique (e.g., Prior et al. 1999) that relies on a comparison of an empirically obtained electron diffraction pattern (EBSP) with a theoretically calculated reference diffraction pattern, often referred to as a “match unit,” to constrain the orientation of a crystalline material at the point at which the diffraction pattern is generated. Critical to the successful ESBD analysis of a material is the ability to generate theoretical match unit files that reliably represent the material being analyzed. However, the parameters used to generate match unit files, including unit-cell dimensions and atomic positions, may change with both intrinsic (e.g., com-position) and extrinsic (e.g., pressure) factors.

Several different studies have constrained the crystallographic parameters of zircon over a range of chemical (REE) and pressure variations, such that there are currently 23 possible structural data sets, each of which allow the calculation of a different zircon match unit file (Table 1; Fig. 1). These 23 different sets

REDDy ET AL.: ELECTRON BACkSCATTER DIFFRACTION ANALySIS OF ZIRCON188

of structural parameters have been obtained using a range of different approaches. Robinson et al. (1971) used natural, non-metamict, zircon from the Proterozoic kragero syenite of Norway to refine the structure of zircon. Hazen and Finger (1979) used fragments of a non-metamict natural zircon from the Finsch kimberlite, South Africa to undertake structural measurements at a range of different pressure conditions up to ~5 GPa. Still higher-pressure zircon structures were subsequently derived us-ing density function theory (Farnan et al. 2003). A Sri Lankan natural zircon (age of 570 ± 20 Ma and a U content of 1100 ppm) was used to study the effects of radiation damage (1.8 × 1018 α-decay events g–1) on zircon structural properties (Ríos et al. 2000) while doped synthetic zircon was used to investigate the variation in single-crystal X-ray diffraction associated with rare earth element (REE) substitution (Finch et al. 2001). Synthetic zircon was also used by kolesov et al. (2001) to provide some

additional constraints on the structure of non-metamict zircon. In addition, a zircon “Best in Family” match unit, referred to as HkL BIF, provides an untested “best guess” default in the EBSD application phase database supplied by Oxford Instruments (for-merly HkL Technologies Ltd.), and this provides an additional 24th match unit. There are no data available as to how this “Best in Family” match unit was derived. However, it is included in this study for completeness.

Among these structural data sets there are variations in crystallographic parameters (Table 1), for example there is >5% variation in unit-cell length in the c direction. However, to date there has been no systematic assessment of the applicability of these different properties to natural zircons via EBSD analysis. As a result there are currently no data available to assess if there are significant differences in data quality associated with the range of possible match units or which of the 24 different possible

Table 1. Variations in crystallographic data for the 24 different match units used in this study Match Unit Name 5259a 5260b 5261b 5262b 5263b 5264b 5265b 5266b 5267b 5271c 5272d 6131e

Pressure (GPa) – 1 atm 0.98 1.74 2.32 2.89 3.71 4.2 4.81 – – 0.53Dopant (ionic radii in pm) na na na na na na na na na na na na

a (Å) 6.6070 6.6042 6.5927 6.5849 6.5808 6.5737 6.5650 6.5592 6.5531 6.6180 6.6039 6.5670c (Å) 5.9820 5.9796 5.9742 5.9692 5.9670 5.9638 5.9682 5.9553 5.9519 6.0190 5.9782 5.9260O y/b 0.0661 0.0660 0.0661 0.0666 0.0657 0.0654 0.0656 0.0656 0.0659 0.0658 0.0659 0.0660O z/c 0.1953 0.1951 0.1962 0.1963 0.1951 0.1954 0.1961 0.1960 0.1956 0.1955 0.1953 0.1950Unit-cell volume (Å3) 261.13 260.80 259.66 258.83 258.41 257.72 257.23 256.22 255.59 263.62 260.72 255.56Mol. volume (cm3/mol) 39.32 39.27 39.10 38.98 38.91 38.81 38.73 38.58 38.49 39.70 39.26 38.48 Match Unit Name 6132e 6133e no. 38f Dy(12)f Dy(15)f Sm+P(28)f Gd+P(30)f Dy+P(33)f Er+P(36)f Yb+P(40)f Y+P(43)f HKLPressure (GPa) 20.85 40.44 na na na na na na na na na “Best in Dopant (ionic radii in pm) na na na Dy (91) Dy (91) SM (96)+ P Gd (94)+P Dy (91)+P Er (88)+P Yb (86)+P Y (89)+P Family”

a (Å) 6.3920 6.2690 6.1020 6.6175 6.6139 6.6119 6.6213 6.6260 6.6355 6.6265 6.6329 6.6000c (Å) 5.8160 5.7200 5.9860 5.9890 5.9850 5.9830 5.9879 5.9860 5.9890 5.9790 5.9860 5.9800O y/b 0.0640 0.0620 0.0657 0.0658 0.0658 0.0658 0.0660 0.0664 0.0663 0.0664 0.0630 O z/c 0.1960 0.1970 0.1961 0.1955 0.1954 0.1953 0.1959 0.1963 0.1964 0.1965 0.1964 Unit-cell volume (Å3) 237.63 224.80 222.89 262.27 261.81 261.56 262.52 262.81 263.69 262.54 263.36 260.80Mol. volume (cm3/mol) 35.78 33.85 33.56 39.49 39.42 39.39 39.53 39.57 39.71 39.53 39.66

Notes: a and c are the unit-cell lengths in angstroms. O y/b and O z/c are the only two varying oxygen atom positions between the different match unit files. Crystal-lographic data that is not shown is constant among all 23 match units. Refer to the CRY files in supplementary data for constant values. Zircon Space Group: *I41/amd; Variant: no. 2. Match unit superscripts (a–f ) refer to source of data used to create the match unit: a = Robinson et al. (1971); b = Hazen and Finger (1979); c = Ríos et al. (2000); d = Kolesov et al. (2001); e = Farnan et al. (2003); f = Finch et al. (2001).

FIgure 1. Representation of the different match units with respect to pressure and trace element dopant (see Table 1 for details). Symbols used for the different match units are the same as those used in Figure 6. Match units 6132, 6133, 38, and Dy(12) are labeled for ease of comparison with subsequent figures.

REDDy ET AL.: ELECTRON BACkSCATTER DIFFRACTION ANALySIS OF ZIRCON 189

match unit files is most appropriate to use under particular geo-logical circumstances. With the increasing application of EBSD to zircon studies such data are critical. In this paper we address this issue by systematically analyzing four different samples for each of the 24 different match unit files and assessing the results. Various properties of the 24 maps are compared to interrogate the quality of the data for each of the different zircon samples and provide a systematic estimate of the best match unit files for use in different geological environments. We also provide the individual and correctly formatted data files (.CRy files) used to generate 23 of the 24 match unit files (all except HkL BIF), as well as a phase database for zircon that include these 23 match units (see supplementary data1). These will allow the relatively simple incorporation of zircon match units into EBSD acquisition software (Channel 5) and use of the theoretical reflector files by the scientific community. The approach used represents the first systematic analysis of match units in the EBSD literature and provides a method for assessing match units characteristics in all other crystalline materials.

SaMple characterIStIcS

The four zircon samples used for this study (Fig. 2) were selected to test the compatibility of natural zircon samples with match units derived from natural, synthetic, and calculated structure parameters, which represent variations in pressure and REE chemistry (Fig. 1). Samples were selected which documented crystallization over a range of geological environments. Samples were also selected so that the potential effect of radiation damage on EBSD results was kept to a minimum, that is old grains had low U-Th concentrations while high U-Th grains recorded young crystallization ages. The samples comprise a high-pressure zircon collected from a mantle xenolith, two crustal Archean zircons with low-U and Th, and one Late Miocene, high-U zircon of igneous origin.

High-pressure mantle zircon (sample UX)UX is a large (>900 µm long) anhedral zircon grain (Fig. 2a)

from a mantle xenolith from the Udachnaya kimberlite, Siberia (latitude of 66°25′N and a longitude of 112°51′E). Cathodolu-minescence (CL) imaging shows a complex structure with dark anastamosing bands locally traversing the sample (Fig. 2a). The zircon lies within a zircon-rich zone of a partially metasomatised garnet websterite. Udachnaya peridotite xenoliths preserve as-semblages from a range of mantle temperature and pressure conditions (750–1380 °C and ~75–210 km depth, Boyd et al. 1997), but these have not been constrained in the studied sample. Perovskite in the Udachnaya kimberlite has been dated by U-Pb SHRIMP and gives an age of 360 Ma (kinny et al. 1997). However, zircons from this sample yield Palaeoproterozoic ages

FIgure 2. Cathodoluminescence images of the four studied zircons. In all cases the white square shows the area analyzed by EBSD using automated orientation mapping. (a) Udachnaya sample (UX); (b) Unknown sample (VIET1); (c) Lewisian sample (GST15); (d) Javanese sample (JAVA). The small circles on the surface of c are the sites of ionprobe analyses reported by Timms et al. (2006).

1 Deposit item AM-08-007, Supplementary Data. (Data files and a phase database for zircon and a figure.) Deposit items are available two ways: For a paper copy contact the Business Office of the Mineralogical Society of America (see inside front cover of recent issue) for price information. For an electronic copy visit the MSA web site at http://www.minsocam.org, go to the American Mineralogist Contents, find the table of contents for the specific volume/issue wanted, and then click on the deposit link there.

REDDy ET AL.: ELECTRON BACkSCATTER DIFFRACTION ANALySIS OF ZIRCON190

and have maximum U and Th concentrations of 12 and 40 ppm, respectively (P. kinny, pers. comm.).

Low-U crustal zircon (sample VIET1)VIET1 is a sub-rounded zircon, ~300 µm across, and is of

unknown origin. The zircon is characterized in CL by a bright core, surrounded by a narrow, discontinuous, CL-dark rim (Fig. 2b). SHRIMP data from the region of the grain analyzed by EBSD yields a late Archean age (Timms, unpublished data). Uranium and Th concentrations in the core are 15 and 18 ppm, respectively.

Low-U crustal zircon (sample GST15)Sample GST15 is a subhedral zircon grain from the margin

a syntectonic pyroxenite intrusion in the Lewisian gneiss terrain of northwest Scotland (kinny and Friend 1997), located at Loch an Daimh Mór (58° 19′54′′N, 5°07′57′′W). The grain is largely uniform in CL, and is disrupted by CL-dark bands that relate to microstructures associated with intragrain crystal-plastic defor-mation soon after initial crystallization (Fig. 2c) (Timms et al. 2006). Fifty SHRIMP analyses across the grain show that U and Th concentrations are <100 ppm, Th/U ratios vary between 1 and 2, and it preserves a weighted mean combined 207Pb/206Pb age of 2451 ± 14 Ma (kinny and Friend 1997; Timms et al. 2006). The area analyzed in this study comes from a region of the grain that shows no discernable deformation and low U and Th.

High-U crustal grain (sample JAVA)JAVA is the ~200 µm long broken tip of a larger euhedral

igneous zircon (Fig. 2d) from a diorite in the Ponorogo district of East Java (sample Jhs2Pon4 of Smyth et al. 2007) located approximately at latitude 07°50′S and longitude 111°20′E. This grain forms one of a suite of zircons that show euhedral shapes and well-defined concentric oscillatory growth zoning in CL (Fig. 2d). U-Pb SHRIMP dating of these grains yields a crystallization age of the host diorite of ~9 Ma (P.J. Hamilton, pers. comm.). The analyzed grain has a U content of 2112 ppm and a single analysis gives a SHRIMP U-Pb age of 9.2 Ma (P.J. Hamilton, pers. comm.). The sample from which the zircon is derived also contains a population of xenocrystic grains that record Cambrian to Archean U-Pb SHRIMP ages (Smyth et al. 2007). No xenocrysts younger than a Cambrian age have been reported from this sample.

analytIcal procedureThe analyzed zircons were selected from a series of SHRIMP epoxy disks

(GST15, VIET1, JAVA) and a polished thin section (UX) that had been previously made by standard preparation techniques. These samples were repolished with pro-gressively finer grades of diamond paste (down to 1 µm) followed by a further 4 h of polishing with 0.06 µm colloidal silica in a NaOH solution (pH 10) on a Vibromet II polisher. The samples were then analyzed by EBSD at the Microstructural Analysis Facility, Curtin University of Technology, Australia. Acquisition, processing and post-processing analysis of EBSD data were undertaken using Oxford Instruments Channel 5.9 software. In each case, a small area of each zircon was analyzed under constant conditions (shown in Table 2) in a tungsten-sourced Phillips XL30 scanning electron microscope. For three of the zircons, the areas comprised a 100 × 100 step matrix with a step size of 1.5 µm (to yield a 150 × 150 µm2 orientation map). For the JAVA sample, a 1.0 µm step size was used because of the smaller size of this grain. A single electron backscatter pattern (EBSP) was measured at each step and the 10 000 EBSPs from each sample were saved and subsequently indexed off line. Indexing parameters used for EBSP acquisition (Table 2) were kept constant for

all samples and were selected based on empirical studies over several years that show they yield good results for zircon. By keeping acquisition settings constant, the only variable for each sample is the theoretical match unit file used to index each of the collected diffraction patterns. For each sample, the collection of 10 000 EBSPs were systematically analyzed by each of 24 different match unit files that vary according to both pressure and trace element geochemistry (Fig.1 and Table 1). These different match unit files represent theoretical reflector intensity data and were derived from Channel 5 Twist software via structure factor calculations that utilize a kinematical electron diffraction model. The crystallographic, space group, atom coordinate, and occupancy data needed to create these match unit reflector files are summarized in Table 1 (also refer to supplementary .CRy files1). For 17 of the match units, the names used to label them are taken from the card number used to identify the zircon structure in the Mincryst database (Chichagov et al. 2001). The exceptions are those data derived by Finch et al. (2001), which use the original sample names.

In all cases, empirically obtained EBSPs from the zircon were indexed to theoretical reflector intensities using eight detected kikuchi bands. This is done automatically via a Hough transform (Hough 1962), a technique that finds the dominant edges (i.e., bands) within the EBSP and compares the results to the theoretical reflector files. The mean angular deviation (MAD) discriminator be-tween the theoretical reflector file and the empirically obtained EBSP was set to a value of 1.3°, above which analyses were assigned a zero solution. This was done to preclude the indexing of poor quality EBSPs. Analysis of the EBSPs for each zircon resulted in 24 sets of EBSD data for each sample. Several repeat analyses of individual projects were undertaken to test the reproducibility of indexing statistics for different match unit files. In all cases, reanalysis yielded exactly the same results, indicating that variations in indexing properties are the results of the different match units and not random variations in indexing.

Several different variables were measured for each of the 96 orientation maps. These variables are those commonly recorded in quanitative microstructural analy-ses to provide constraints on the quality of EBSD data. Recorded data include the mean band contrast and mean band slope; the percent indexing, MAD, and the cumulative misorientation boundary length. The mean band contrast and mean band slope are a measure of the quality of the original EBSP used in the index-ing procedure. Band contrast values stem from difference in the mean height of detected peaks in the Hough transform, each peak corresponding to a particular kikuchi band in the EBSP, and are a measure of the strength of bands in the EBSP. Band slope value is derived from the mean gradient of detected peaks in the Hough transform, and is a measure of the sharpness of bands. The % indexing represents the percentage of the 10 000 analyses from each map that were indexed as zircon. The mean, maximum and minimum values of the MAD are a measure of the angular fit between the empirically derived EBSP and the theoretical match

Table 2. Scanning electron microscope and EBSD settings used for this study

SEM settingsSEM system Philips XL30 SEMEBSD system HKL Channel 5.9C coat YesAcc. voltage (kV) 20Working distance (mm) 20Spot size (nm) Spot size 5Tilt (°) 70

EBSD settings Samples UX, VIET1, GST15 JAVAEBSP collection time per frame (ms) 60 60Background (frames) 64 64EBSP noise reduction (frames) 4 4 (binning) 4 × 4 4 × 4 (gain) low lowHough resolution 65 65Band detection min/max 6/8 6/8X steps 100 100Y steps 100 100Step distance (µm) 1.5 1.0†MAD discriminator (°) 1.3 1.3Noise reduction*—“wildspike” Yes Yes–n neighbor zero solution extrapolation 4 4

* Only applied for noise-reduced cumulative misorientation boundary data.† Smaller step used because of smaller grain size.

REDDy ET AL.: ELECTRON BACkSCATTER DIFFRACTION ANALySIS OF ZIRCON 191

unit. The cumulative misorientation boundary length is a measure of the number of orientation boundaries (>0.25°) between adjacent pixels and at these low angles is a semi-quantitative measure of angular noise in each map. Two values are given for cumulative misorientation boundary length, one derived from the raw data and a second after noise reduction. The noise reduction process for this single property included a “wildspike” and four-nearest neighbor zero solution correction (refer to Reddy et al. 2007 for details). This noise-reduction was undertaken to determine if the noise generated by different match units was effectively removed by a standard noise reduction process.

Analysis of the empirical EBSD data for the four different samples by each of 24 different match units was initially undertaken by graphical comparison of characteristics derived from the EBSD data. A multivariate statistical analysis was also undertaken to better understand the variables that control the resulting EBSD data. This approach involved simultaneous R- and Q-mode principal component analysis (e.g., Davis 2002) using the “Coran” statistical software developed by Bruce Eglington (http://sil.usask.ca/eglington/bme_software.htm). This multi-variate, dimension-reduction approach projects the original set of variables onto a set of new, uncorrelated, principal components so as to maximize the variance on the first component axis, then maximize the remaining variance on a second component axis, and so on until all of the original variance is captured. For the analyses performed here, the first two principal components generally account for a large percentage (~95%) of the total variance. Calculation of component scores and component loadings from the eigenvalues and eigenvectors using simultane-ous R- and Q-mode principal component analysis allows graphical representation of samples and variables in a common component space. Such an approach has the significant advantage of allowing the variance in a large number of variables to be visualized with respect to the samples in only a few graphs of principal component space.

As a final analysis of the data, self-organizing map (SOM or kohonen) neural networks have been used to allow “unsupervised” analysis of the variables and pro-vide insight into the recognition of how data cluster. In SOM, variables are projected onto a 2-D net and an associated unified matrix illustrates where major changes in the SOM occur. By so doing, patterns in the distribution of different variables can be compared and the qualitative relationships among them understood (kohonen 2006). The SOM approach permits a non-linear view of the data space whereas principal component analysis is a linear technique, and thus provides an alternative check on data integrity. All SOM analyses were performed using MATLAB and the SOM Toolbox (http://www.cis.hut.fi/projects/somtoolbox). All of the zircons analyzed during this study are adequately described by vectors within a linear space but, since the methodology utilized here has potential for EBSD studies of other minerals that may require non-linear approaches, we have also provided the SOM results as supplementary diagrams so as to facilitate future comparisons.

reSultS and data analySIS

A representative EBSP from each sample shows a qualita-tively similar, and good quality, EBSP across all samples [Figs. 3a–3d(i)]. Band contrast and MAD data, derived from each EBSP, and misorientation boundary data are shown as example maps for each sample for just one of the match units (HkL BIF) and show the range of different values recorded by each sample [Figs. 3a–3d(ii–iv)]. For analysis purposes a representative (mean or total) value has been derived for each EBSD variable for each match unit. The results of this analysis of each of the 24 match units for each of the four samples are shown in Table 3. The key aspects of these data are shown graphically for each sample and the different match units in Figure 4. In all samples the mean band contrast is generally high, indicating the collec-tion of good quality EBSPs from all samples. The % indexing is also high across all samples (mostly >92%), with some samples consistently above 96%. The variation in % indexing within samples is generally <2%, except for GST15 that records an 11% variation. The minor variation among samples reflects physical differences in the analyzed areas, for example the white corners in some of the texture component maps (Fig. 3) repre-sent analysis of a non-zircon phase, while non-indexed areas in

Java represent sample damage associated with microstructural modification (Fig. 3d). Consequently % indexing can only be used to consider within sample variations between match units and not cross-sample differences. For similar reasons the varia-tions in cumulative misorientation length cannot be compared among different samples and only within sample variations are useful comparisons of sample noise. Variations in mean MAD are also seen among the samples but these are all below the cut off MAD of 1.3°.

Despite the difficulty in comparing the data among the differ-ent samples, the variation in the four measured variables (Fig. 4) show several notable patterns. There is a significantly reduced mean band contrast associated with match units 6132 and 6133 for all samples (Fig. 4a). These two match units correspond to structure properties calculated for high-pressure conditions (Farnan et al. 2003). Mean band contrast values for match units 38 and Dy(12) are similar to other match units for samples UX and JAVA but record reduced values for VIET1 and increased values for GST15 (Fig. 4a). Similarly decreasing values are seen in the % indexing data for VIET1 for the same four match units, and reductions in % indexing are also recorded by 6132, 6133, and 38 in the GST 15 and JAVA data (Fig. 4b). Only the UX sample shows little change in % indexing associated with these match units. The MAD data show a little more complexity, but again the data for 6132, 6133 are anomalous in three of the samples with JAVA showing a decrease while UX and GST15 show an increase (Fig. 4c). Even more complexity is recorded in the cumulative misorientation boundary length data (Fig. 4d). However, once again 6132 and 6133 match units tend to be anomalous, with 38 and Dy(12) also showing anomalies in GST15 and VIET1, respectively. Comparison of raw and noise reduced cumulative misorientation boundary length data indi-cates that these values are correlated such that high raw values correspond to high noise reduced values.

The complexities highlighted between the 4 measured vari-ables, the four different samples and the 24 different match units are difficult to understand using the graphical representations of Figure 4 alone. In addition, linking the differences in measured EBSD variables to specific characteristics of the different match units associated with changes in lattice parameters are not pos-sible using the graphical representations shown in Figure 4.

Principal component analysis, using input variables derived from both the lattice parameters used to construct the match units and the measured EBSD data, indicate that the four analyzed zircons can be discriminated based on a combination of the first three principal components (see Fig. 5 for relationship between components 1 and 2). These three components account for >95% of the variance among the zircon data and show clear differences between the eigenvectors associated with the original crystal-lographic data and the measured EBSD parameters. Within each zircon a strong linear trend parallel to eigenvectors defined by the crystallographic parameters is apparent (Fig. 5). However, match units 6132 and 6133 are anomalous in all samples and plot away from the main trend, as are match units 38 and Dy12 for VIET1 and GST15. In the case of UX, GST15 and VIET1 the spread of data shows a strong inverse correlation to crystallographic eigenvectors (a, c, O y/b, and O z/c), JAVA is correlated to these same variables (Fig. 5). These variations systematically mimic

REDDy ET AL.: ELECTRON BACkSCATTER DIFFRACTION ANALySIS OF ZIRCON192

the pressure conditions or ionic radius of the dopants used in the generation of the structural data (Table 1). These relationships are also seen in the less significant components (3 and 4, data not shown). The differences between the four samples primar-ily relate to variations in EBSD properties (component 2) with different zircons having varying band contrast and band slope values (Fig. 5). However, the way in which the different match units control EBSD data are better illustrated in the principal component data of the EBSD variables associated with individual samples, which allow % indexing and misorientation boundary length to be included (Fig. 6).

Principal component analysis of the Udachnaya zircon (UX) EBSD data shows that most of the variability in the data (98.9%) can be assigned to only two principal components (Fig. 6a). These

components indicate that match units 6132 and 6133 are inversely related to band contrast, a feature that is easily recognizable in Figure 4. However, the component analysis also shows that the match unit data are generally dominated by two variables; the mean MAD and the cumulative misorientation boundary length (Fig. 6a). The former controls the variability in pressure-related crystallographic data of Hazen and Finger (1979) and shows a systematic trend with low-pressure match units closer to the component space origin than higher pressure analyses. The data of Ríos et al. (2000) also show a correlation with high MAD (Fig. 6a). The cumulative misorientation boundary length also shows a systematic variation, with match units associated with lower ionic radii REE plotting further away from the origin (Fig. 6a), indicating increasing noise in the EBSD data linked to increas-

FIgure 3. Electron backscatter diffraction data for the four zircon samples: (a) UX; (b) VIET1; (c) GST15; and (d) JAVA collected using the HkL “Best in Family” match unit. (i) Sample EBSPs collected from each sample and used in this study. (ii, iii, and iv) Maps for each sample showing variations in band contrast, MAD, and low-angle boundaries among the samples. In (ii) and (iii) non-indexed points are colored white. In (iv), different lines represent boundaries of different misorientation angle (0.25–0.5° = gray lines; 0.5–1° = thin black lines; >1° = thick black lines).

REDDy ET AL.: ELECTRON BACkSCATTER DIFFRACTION ANALySIS OF ZIRCON 193

ingly heavy REE. In EBSD data, 95.2% of the variation in zircon VIET1 can be

described by two components (Fig. 6b). The extent to which indi-vidual variables control these components is not clearly resolved. However, cumulative misorientation boundary length (raw and noise-reduced) dominate component 1, the other variables acting in the opposite direction. Mean MAD dominates component 2. The data are systematically dispersed within the space defined by components 1 and 2. Match units associated with increasing pressure (Hazen and Finger 1979) generally correlate with higher mean MAD (although we note that 5265 plots higher than both

5266 and 5267) representing poorer quality fits between the mea-sured EBSPs and the theoretical reflector file, poorer indexing and increasing noise. Match units associated with increasingly heavy REE substitutions show systematic decreases in mean MAD, increases in % indexing and less cumulative misorienta-tion boundary length (angular noise). To some extent, the data from GST15 (Fig. 6c) shows similar patterns to those in VIET1 (Fig. 6b). However, this is less clear because all of the variables except CMO are strongly linked to component 1.

Principal component analysis of the Java zircon (JAVA) EBSD data (Fig. 6d) shows a less linear distribution than present in the other three samples (Figs. 6a–6c). 80.4% the variation be-tween the different match units is described by component 1, but largely reflects the anomalous behavior of the two high-pressure match units (6132 and 6133). The data show a systematic disper-sion associated with component 2 that is strongly correlated to mean MAD (Fig. 6d). In the case of the JAVA zircon, increasing mean MAD, that is poorer quality fitting of empirically observed EBSPs and theoretically derived match units, correlates with substitution of increasingly heavy REE into the zircon lattice (Fig. 6d). The pressure data define a trend that seems to reflect both an inverse correlation to mean MAD and a correlation with increasing angular noise. However, in detail there is no system-atic relationship of these variables with pressure.

The relationship between the REE-doped zircon match units and EBSD data in the VIET1, GST15, and JAVA samples are difficult to resolve because in two of the samples they lie parallel to the eigenvectors for band contrast, band slope, % indexing, and noise-reduced cumulative misorientation boundary length. Inspection of the data (Table 3) shows that in the case of VIET1 and GST15, the trend of REE doped match units is inversely related to the last of these variables and not increasing pattern quality or % indexing. For GST15, the apparent trend toward increasing % indexing (Fig. 6c) also reflects an inverse relation-

FIgure 4. Graphs showing the relationships between the four main variables collected by EBSD analysis from the four different samples for each of 24 different match units. (a) Mean band contrast; (b) % indexing; (c) mean MAD; (d) cumulative misorientation boundary length. The abscissa shows the different match units (see Table 1 for source).

FIgure 5. Principal component analysis (only components 1 and 2 are shown) of crystal and EBSD variables from all match units and the four samples. In all cases, crystal property data are derived from Table 1, while EBSD variables are the mean values derived for each match unit/sample pair (Table 3). MAD = mean MAD; BC = mean band contrast; BS = mean band slope.

REDDy ET AL.: ELECTRON BACkSCATTER DIFFRACTION ANALySIS OF ZIRCON194

ship with MAD.The SOM analysis also delineates distinct difference between

the various zircons (Supplementary Fig. 11) that are consistent with the principal component analysis shown in Figure 5. SOM analyses of the individual zircon EBSD data also exhibit distinc-tive, associations of the various variables that are unique to each zircon (Supplementary Fig. 21). These data therefore provide further evidence of the complex relationships among match units and EBSD quality.

dIScuSSIon

Band contrast and band slope are measured prior to any processing and is generally used as an indicator of EBSP quality with higher values representing better quality (e.g., Wright and Nowell 2005). Variations in mean band contrast and band slope among the samples may reflect variations in SEM and EBSD settings. However, since these were kept constant throughout the analyses (Table 1), it is likely that mean band contrast and band slope variations among samples [e.g., Fig. 3(ii)] reflect dif-ferences in the orientation of the lattice in the different zircons (Claves and Deal 2005), local dislocation concentrations (Reddy et al. 2007) and variations in surface polish or damage (Nowell

et al. 2005). Band contrast maps [Fig. 3(ii)] are a useful way of characterizing within-sample variations in EBSP quality, and again can be interpreted in terms of crystallographic orientation, structural integrity and surface character. However, the mean band contrast associated with the different match units from each sample are derived from the same 10 000 EBSPs and therefore cannot be related to these characteristics. The general decrease in mean band contrast associated with the match units derived from the density function calculations (6132 and 6133) (Farnan et al. 2003) and the synthetic zircons 38 and Dy(12) (Finch et al. 2001) must therefore reflect variations in the parameters involved in the band contrast calculation.

Band contrast and band slope values are susceptible to changing the number of bands that are detected. For example, the omission of a lower intensity band will increase the mean height of the detected peaks and increase the band contrast and slope. However, inspection of the EBSD data record file (not shown) indicates that 8 bands were used for indexing all EBSPs in all samples, so this explanation is not valid. An alternative explanation for generating different band contrast and slope values from the same set of EBSPs is that the different match units generate reflector files that have sufficiently different band

Table 3. Main measured EBSD variables for the four different samples for each of the 24 different match units 5259a 5260b 5261b 5262b 5263b 5264b 5265b 5266b 5267b 5271c 5272d 6131e 6132e 6133e no. 38f Dy(12)f Dy(15)f Sm+P(28)f Gd+P(30)f Dy+P(33)f Er+P(36)f Yb+P(40)f Y+P(43)f HKLBIF*

UX Indexing (%) 97.45 97.45 97.45 98.07 97.43 98.06 98.03 97.73 97.35 97.29 97.44 97.57 97.68 97.54 97.45 97.44 97.47 97.44 97.45 97.52 97.56 97.58 97.57 98.07 UX zircon mean BC 92.01 92.01 92.01 92.03 92.01 92.03 92.03 92.03 92.01 92.01 92.01 92.01 85.17 85.17 92.01 92.01 92.17 92.02 92.01 92.01 92.01 92.01 92.01 92.03 UX zircon mean BS 131.2 131.2 131.2 131.1 131.2 131.1 131.1 131.1 131.2 131.2 131.2 131.1 130.6 130.6 131.2 131.2 129.9 131.2 131.2 131.1 131.1 131.1 131.1 131.1 UX Mean MAD 0.4809 0.481 0.4814 0.487 0.4867 0.4906 0.5017 0.4945 0.4952 0.5029 0.4802 0.4714 0.5246 0.5428 0.4815 0.4793 0.4567 0.4788 0.4768 0.474 0.4718 0.4713 0.4716 0.4823 UX S.D. 0.1083 0.1083 0.1084 0.1093 0.1093 0.1096 0.1098 0.1098 0.11 0.1099 0.1081 0.1057 0.1261 0.1262 0.1084 0.1079 0.1097 0.1077 0.1073 0.1967 0.1058 0.1056 0.1057 0.1084 UX Min MAD 0.1693 0.1696 0.1703 0.1565 0.1647 0.1632 0.1656 0.1651 0.1647 0.1679 0.1683 0.165 0.1891 0.1911 0.171 0.1667 0.1761 0.1676 0.1654 0.1645 0.166 0.1707 0.1652 0.1487 UX Max MAD 0.9364 0.9365 0.937 0.939 0.9408 0.9459 0.9133 0.9529 0.9567 0.923 0.9353 0.9324 0.9666 0.9966 0.9373 0.9366 0.9792 0.935 0.9399 0.932 0.9318 0.9307 0.9315 1.2726 Cumulative MO length (μm) 5460 5552 5480 5490 5139 5409 5420 5322 4971 5027 5580 6908 6669 7058 5414 5685 5595 5693 6030 6426 6723 6969 6854 5928 Cumulative MO length (μm) NR 636 641 633 648 620 621 629 621 602 588 636 786 957 1011 648 635 681 647 666 726 786 788 807 689 VIET1 Indexing (%) 97.73 97.73 97.73 97.7 97.69 97.69 97.63 97.67 97.66 97.62 97.73 97.82 96.19 96.57 97.06 96.59 97.73 97.73 97.74 97.79 97.82 97.82 97.82 97.73 VIET1 zircon mean BC 104.1 104.4 104.4 104.4 104.4 104.4 104.4 104.4 104.4 104.4 104.4 104.4 84.22 82.14 85.87 85.21 104.4 104.4 104.4 104.4 104.4 104.4 104.4 104.4 VIET1 zircon mean BS 107.2 107.2 107.2 107.2 107.2 107.3 107.3 107.3 107.3 107.3 107.2 107.2 104.5 105.7 102.7 103.6 107.2 107.2 107.2 107.2 107.2 107.2 107.2 107.2 VIET1 Mean MAD 0.6149 0.615 0.6159 0.6231 0.6248 0.6287 0.6433 0.6342 0.6368 0.6465 0.6138 0.5937 0.6548 0.6304 0.6224 0.6294 0.6112 0.611 0.6071 0.6008 0.595 0.593 0.5943 0.615 VIET1 S.D. 0.1749 0.1749 0.1749 0.1753 0.1752 0.1755 0.1757 0.1755 0.1755 0.1758 0.1748 0.174 0.1642 0.1633 0.1755 0.1887 0.1746 0.1747 0.1746 0.1744 0.1741 0.1739 0.174 0.1749 VIET1 Min MAD 0.2109 0.211 0.2123 0.2188 0.221 0.2299 0.2387 0.2298 0.2321 0.2416 0.2098 0.1904 0.2333 0.2427 0.1918 0.2073 0.2083 0.2081 0.2032 0.1975 0.1916 0.1906 0.1912 0.211 VIET1 Max MAD 1.2803 1.2804 1.281 1.2861 1.2854 1.2885 1.2953 1.2906 1.2917 1.2967 1.2799 1.269 1.2117 1.2642 1.2404 1.2234 1.2784 1.2783 1.2758 1.2726 1.2687 1.2678 1.2685 1.2804 Cumulative MO length (μm) 16091 16055 16167 16383 16394 16494 16854 16614 16751 17010 16191 15468 18957 16278 18686 19439 15962 15995 15944 15686 15404 15251 15392 16055 Cumulative MO length (μm) NR 2955 2945 2940 3122 3126 3246 3489 3282 3401 3570 2859 2555 5042 3534 4589 5166 2778 2871 2778 2694 2601 2540 2600 2943 GST15 Indexing (%) 93.66 93.65 93.63 93.62 93.58 93.49 93.05 93.35 93.28 92.92 93.7 93.73 91.96 89.53 84.29 94.71 93.73 93.73 93.71 93.75 93.74 93.73 93.73 93.63 GST15 zircon mean BC 91.26 91.26 91.26 91.25 91.25 91.25 91.27 91.26 91.26 91.27 91.25 91.25 88.28 88.34 99.67 94.31 91.26 91.26 91.25 91.25 91.25 91.25 91.25 91.26 GST15 zircon mean BS 101.9 101.9 101.9 101.8 101.9 101.9 101.9 101.9 101.9 101.9 101.9 101.8 103.1 103.4 102.1 101.3 101.8 101.8 101.8 101.8 101.8 101.8 101.8 101.9 GST15 Mean MAD 0.6735 0.6738 0.6752 0.6867 0.6898 0.6958 0.7189 0.7046 0.7089 0.724 0.6718 0.6402 0.7687 0.7993 0.6755 0.6958 0.6679 0.6675 0.6614 0.6512 0.6419 0.6391 0.641 0.6737 GST15 S.D. 0.1916 0.1916 0.192 0.1945 0.1949 0.1956 0.1972 0.1967 0.1972 0.1974 0.1912 0.1823 0.1869 0.1825 0.1921 0.1919 0.1904 0.1903 0.189 0.1861 0.1828 0.1819 0.1825 0.1916 GST15 Min MAD 0.1755 0.1758 0.1763 0.1857 0.1872 0.1926 0.2105 0.2021 0.2054 0.2085 0.1751 0.1769 0.2597 0.2676 0.1761 0.2131 0.173 0.1733 0.1717 0.1737 0.1785 0.1779 0.1768 0.175 GST15 Max MAD 1.3124 1.3128 1.3161 1.3424 1.3489 1.3641 1.414 1.3818 1.3947 1.4249 1.3083 1.2681 1.3664 1.3872 1.3169 1.3008 1.3191 1.3188 1.3074 1.2889 1.2715 1.2659 1.2695 1.3128 Cumulative MO length (μm) 19019 18992 19022 19229 19401 19379 19566 19499 19512 19592 18996 18608 19613 19217 15287 19947 18951 18921 18869 18782 18627 18518 18599 18999 Cumulative MO length (μm) NR 6039 6084 6101 6387 6411 6519 6983 6672 6729 7016 5966 5465 7515 8036 5489 6606 5928 5927 5852 5675 5571 5531 5511 6077 JAVA Indexing (%) 95.37 95.37 95.37 95.39 95.37 95.38 95.37 95.38 95.36 95.34 95.37 95.37 94.57 94.55 95.36 95.37 95.39 95.39 95.37 95.36 95.36 95.38 95.37 95.38 JAVA zircon mean BC 96.34 96.34 96.34 96.35 96.35 96.34 96.34 96.34 96.34 96.35 96.34 96.34 93.32 93.32 96.35 96.35 96.35 96.35 96.35 96.34 96.34 96.34 96.34 96.1 JAVA zircon mean BS 109.6 109.6 109.6 109.6 109.6 109.6 109.6 109.6 109.6 109.6 109.6 109.6 111.6 111.6 109.6 109.6 109.6 109.6 109.6 109.6 109.6 109.6 109.6 109.6 JAVA Mean MAD 0.4078 0.4078 0.4077 0.4074 0.4072 0.4071 0.4074 0.4071 0.407 0.4074 0.4079 0.4109 0.3839 0.3851 0.4077 0.408 0.4082 0.4082 0.4086 0.4096 0.4106 0.4111 0.4108 0.4166 JAVA S.D. 0.1001 0.1001 0.1001 0.1006 0.1001 0.1005 0.1007 0.1005 0.1002 0.1003 0.1001 0.0999 0.1053 0.1055 0.1001 0.1 0.1001 0.1001 0.1 0.1 0.1001 0.1001 0.1001 0.106 JAVA Min MAD 0.1227 0.1226 0.1225 0.1262 0.1261 0.1297 0.1378 0.1321 0.1342 0.1412 0.1233 0.1406 0.1144 0.1205 0.1225 0.1236 0.1243 0.1247 0.1284 0.1331 0.1396 0.1405 0.1402 0.1283 JAVA Max MAD 1.1664 1.1664 1.1664 1.4342 1.1194 1.4329 1.4296 1.4319 1.1248 1.1285 1.1666 1.1645 1.256 1.2508 1.1666 1.669 1.1665 1.1661 1.1657 1.1682 1.1654 1.1644 1.1658 1.3601 Cumulative MO length (μm) 14129 14090 14004 14042 14049 14172 13986 14073 13941 13941 14064 14129 12020 11876 14127 14055 14237 14168 14153 14100 14201 14189 14199 14006 Cumulative MO length (μm) NR 4535 4524 4494 4517 4508 4496 4478 4461 4436 4377 4475 4431 4097 4074 4481 4482 4541 4497 4458 4517 4466 4482 4485 4518

Notes: BC = band contrast, BS = band slope, S.D. = standard deviation, MAD = mean angular deviation, MO = misorientation, NR = noise reduction. Match unit superscripts (a–f ) refer to source of data used to create the match unit: a = Robinson et al. (1971); b = Hazen and Finger (1979); c = Rios et al. (2000); d = Kolesov et al. (2001); e = Farnan et al. (2003); f = Finch et al. (2001).* HKL Technology’s “Best in Family.”

REDDy ET AL.: ELECTRON BACkSCATTER DIFFRACTION ANALySIS OF ZIRCON 195

intensities so that for any given orientation, different match units may use different bands for indexing. From the data obtained in this study, this seems the most likely scenario. Significantly, the data from all of the samples indicate that 6132, 6133, 38, and Dy(12) match units tend to select bands that give an overall poorer quality EBSD data.

The lower values of band contrast associated with 6132, 6133, 38, and Dy(12) are reflected in the % indexing, particularly for the sample GST15 and three of the four samples show higher MAD values indicating a poorer angular fit between the observed EBSP and these theoretical match units. These four match units are also anomalous in the results of the multivariate principal component analysis of the crystallographic parameters used to derive the 23 match units and the fundamental EBSD properties of band contrast, band slope and mean angular deviation (Fig. 5). Furthermore, 6132 and 6133 record increased misorientation boundary length (Fig. 4), indicating higher levels of noise in the data derived from these match units. These observations further indicate that for zircons from a range of geological settings, the match units 6132, 6133, 38, and Dy(12) generally give worse EBSD results than the other match units.

The relationships between the different match units and the

EBSD variables indicate systematic trends with poorer qual-ity EBSD resulting from the use of match units derived from structural parameters associated with increasing pressure or de-creasing dopant ionic radius. The analysis of four natural zircons from both crustal and mantle conditions by 24 different match unit, derived from a range of pressures and zircon compositions, therefore yields the result that the best EBSD data are invariably obtained by using the structural parameters derived from low-pressure natural zircon. Structural parameters from high-pressure zircon (Hazen and Finger 1979), doped or undoped synthetic zircon (Finch et al. 2001; kolesov et al. 2001), radiation damaged zircon (Ríos et al. 2000), and numerical simulation (Farnan et al. 2003) yield poorer results irrespective of the geological environ-ment under which the natural zircon formed. This feature is even common to high U content and high-pressure mantle zircons. In addition, the “Best in Family” match unit (HkL BIF), the default match unit for zircon in the Channel 5 phase database, yields relatively poor results for the JAVA sample. The data suggest that the consistently best results come from the use of match unit 5260, the lowest pressure natural zircon characterized by Hazen and Finger (1979). A conclusion of this study is that this match unit be adopted in future by researchers undertaking EBSD

Table 3. Main measured EBSD variables for the four different samples for each of the 24 different match units 5259a 5260b 5261b 5262b 5263b 5264b 5265b 5266b 5267b 5271c 5272d 6131e 6132e 6133e no. 38f Dy(12)f Dy(15)f Sm+P(28)f Gd+P(30)f Dy+P(33)f Er+P(36)f Yb+P(40)f Y+P(43)f HKLBIF*

UX Indexing (%) 97.45 97.45 97.45 98.07 97.43 98.06 98.03 97.73 97.35 97.29 97.44 97.57 97.68 97.54 97.45 97.44 97.47 97.44 97.45 97.52 97.56 97.58 97.57 98.07 UX zircon mean BC 92.01 92.01 92.01 92.03 92.01 92.03 92.03 92.03 92.01 92.01 92.01 92.01 85.17 85.17 92.01 92.01 92.17 92.02 92.01 92.01 92.01 92.01 92.01 92.03 UX zircon mean BS 131.2 131.2 131.2 131.1 131.2 131.1 131.1 131.1 131.2 131.2 131.2 131.1 130.6 130.6 131.2 131.2 129.9 131.2 131.2 131.1 131.1 131.1 131.1 131.1 UX Mean MAD 0.4809 0.481 0.4814 0.487 0.4867 0.4906 0.5017 0.4945 0.4952 0.5029 0.4802 0.4714 0.5246 0.5428 0.4815 0.4793 0.4567 0.4788 0.4768 0.474 0.4718 0.4713 0.4716 0.4823 UX S.D. 0.1083 0.1083 0.1084 0.1093 0.1093 0.1096 0.1098 0.1098 0.11 0.1099 0.1081 0.1057 0.1261 0.1262 0.1084 0.1079 0.1097 0.1077 0.1073 0.1967 0.1058 0.1056 0.1057 0.1084 UX Min MAD 0.1693 0.1696 0.1703 0.1565 0.1647 0.1632 0.1656 0.1651 0.1647 0.1679 0.1683 0.165 0.1891 0.1911 0.171 0.1667 0.1761 0.1676 0.1654 0.1645 0.166 0.1707 0.1652 0.1487 UX Max MAD 0.9364 0.9365 0.937 0.939 0.9408 0.9459 0.9133 0.9529 0.9567 0.923 0.9353 0.9324 0.9666 0.9966 0.9373 0.9366 0.9792 0.935 0.9399 0.932 0.9318 0.9307 0.9315 1.2726 Cumulative MO length (μm) 5460 5552 5480 5490 5139 5409 5420 5322 4971 5027 5580 6908 6669 7058 5414 5685 5595 5693 6030 6426 6723 6969 6854 5928 Cumulative MO length (μm) NR 636 641 633 648 620 621 629 621 602 588 636 786 957 1011 648 635 681 647 666 726 786 788 807 689 VIET1 Indexing (%) 97.73 97.73 97.73 97.7 97.69 97.69 97.63 97.67 97.66 97.62 97.73 97.82 96.19 96.57 97.06 96.59 97.73 97.73 97.74 97.79 97.82 97.82 97.82 97.73 VIET1 zircon mean BC 104.1 104.4 104.4 104.4 104.4 104.4 104.4 104.4 104.4 104.4 104.4 104.4 84.22 82.14 85.87 85.21 104.4 104.4 104.4 104.4 104.4 104.4 104.4 104.4 VIET1 zircon mean BS 107.2 107.2 107.2 107.2 107.2 107.3 107.3 107.3 107.3 107.3 107.2 107.2 104.5 105.7 102.7 103.6 107.2 107.2 107.2 107.2 107.2 107.2 107.2 107.2 VIET1 Mean MAD 0.6149 0.615 0.6159 0.6231 0.6248 0.6287 0.6433 0.6342 0.6368 0.6465 0.6138 0.5937 0.6548 0.6304 0.6224 0.6294 0.6112 0.611 0.6071 0.6008 0.595 0.593 0.5943 0.615 VIET1 S.D. 0.1749 0.1749 0.1749 0.1753 0.1752 0.1755 0.1757 0.1755 0.1755 0.1758 0.1748 0.174 0.1642 0.1633 0.1755 0.1887 0.1746 0.1747 0.1746 0.1744 0.1741 0.1739 0.174 0.1749 VIET1 Min MAD 0.2109 0.211 0.2123 0.2188 0.221 0.2299 0.2387 0.2298 0.2321 0.2416 0.2098 0.1904 0.2333 0.2427 0.1918 0.2073 0.2083 0.2081 0.2032 0.1975 0.1916 0.1906 0.1912 0.211 VIET1 Max MAD 1.2803 1.2804 1.281 1.2861 1.2854 1.2885 1.2953 1.2906 1.2917 1.2967 1.2799 1.269 1.2117 1.2642 1.2404 1.2234 1.2784 1.2783 1.2758 1.2726 1.2687 1.2678 1.2685 1.2804 Cumulative MO length (μm) 16091 16055 16167 16383 16394 16494 16854 16614 16751 17010 16191 15468 18957 16278 18686 19439 15962 15995 15944 15686 15404 15251 15392 16055 Cumulative MO length (μm) NR 2955 2945 2940 3122 3126 3246 3489 3282 3401 3570 2859 2555 5042 3534 4589 5166 2778 2871 2778 2694 2601 2540 2600 2943 GST15 Indexing (%) 93.66 93.65 93.63 93.62 93.58 93.49 93.05 93.35 93.28 92.92 93.7 93.73 91.96 89.53 84.29 94.71 93.73 93.73 93.71 93.75 93.74 93.73 93.73 93.63 GST15 zircon mean BC 91.26 91.26 91.26 91.25 91.25 91.25 91.27 91.26 91.26 91.27 91.25 91.25 88.28 88.34 99.67 94.31 91.26 91.26 91.25 91.25 91.25 91.25 91.25 91.26 GST15 zircon mean BS 101.9 101.9 101.9 101.8 101.9 101.9 101.9 101.9 101.9 101.9 101.9 101.8 103.1 103.4 102.1 101.3 101.8 101.8 101.8 101.8 101.8 101.8 101.8 101.9 GST15 Mean MAD 0.6735 0.6738 0.6752 0.6867 0.6898 0.6958 0.7189 0.7046 0.7089 0.724 0.6718 0.6402 0.7687 0.7993 0.6755 0.6958 0.6679 0.6675 0.6614 0.6512 0.6419 0.6391 0.641 0.6737 GST15 S.D. 0.1916 0.1916 0.192 0.1945 0.1949 0.1956 0.1972 0.1967 0.1972 0.1974 0.1912 0.1823 0.1869 0.1825 0.1921 0.1919 0.1904 0.1903 0.189 0.1861 0.1828 0.1819 0.1825 0.1916 GST15 Min MAD 0.1755 0.1758 0.1763 0.1857 0.1872 0.1926 0.2105 0.2021 0.2054 0.2085 0.1751 0.1769 0.2597 0.2676 0.1761 0.2131 0.173 0.1733 0.1717 0.1737 0.1785 0.1779 0.1768 0.175 GST15 Max MAD 1.3124 1.3128 1.3161 1.3424 1.3489 1.3641 1.414 1.3818 1.3947 1.4249 1.3083 1.2681 1.3664 1.3872 1.3169 1.3008 1.3191 1.3188 1.3074 1.2889 1.2715 1.2659 1.2695 1.3128 Cumulative MO length (μm) 19019 18992 19022 19229 19401 19379 19566 19499 19512 19592 18996 18608 19613 19217 15287 19947 18951 18921 18869 18782 18627 18518 18599 18999 Cumulative MO length (μm) NR 6039 6084 6101 6387 6411 6519 6983 6672 6729 7016 5966 5465 7515 8036 5489 6606 5928 5927 5852 5675 5571 5531 5511 6077 JAVA Indexing (%) 95.37 95.37 95.37 95.39 95.37 95.38 95.37 95.38 95.36 95.34 95.37 95.37 94.57 94.55 95.36 95.37 95.39 95.39 95.37 95.36 95.36 95.38 95.37 95.38 JAVA zircon mean BC 96.34 96.34 96.34 96.35 96.35 96.34 96.34 96.34 96.34 96.35 96.34 96.34 93.32 93.32 96.35 96.35 96.35 96.35 96.35 96.34 96.34 96.34 96.34 96.1 JAVA zircon mean BS 109.6 109.6 109.6 109.6 109.6 109.6 109.6 109.6 109.6 109.6 109.6 109.6 111.6 111.6 109.6 109.6 109.6 109.6 109.6 109.6 109.6 109.6 109.6 109.6 JAVA Mean MAD 0.4078 0.4078 0.4077 0.4074 0.4072 0.4071 0.4074 0.4071 0.407 0.4074 0.4079 0.4109 0.3839 0.3851 0.4077 0.408 0.4082 0.4082 0.4086 0.4096 0.4106 0.4111 0.4108 0.4166 JAVA S.D. 0.1001 0.1001 0.1001 0.1006 0.1001 0.1005 0.1007 0.1005 0.1002 0.1003 0.1001 0.0999 0.1053 0.1055 0.1001 0.1 0.1001 0.1001 0.1 0.1 0.1001 0.1001 0.1001 0.106 JAVA Min MAD 0.1227 0.1226 0.1225 0.1262 0.1261 0.1297 0.1378 0.1321 0.1342 0.1412 0.1233 0.1406 0.1144 0.1205 0.1225 0.1236 0.1243 0.1247 0.1284 0.1331 0.1396 0.1405 0.1402 0.1283 JAVA Max MAD 1.1664 1.1664 1.1664 1.4342 1.1194 1.4329 1.4296 1.4319 1.1248 1.1285 1.1666 1.1645 1.256 1.2508 1.1666 1.669 1.1665 1.1661 1.1657 1.1682 1.1654 1.1644 1.1658 1.3601 Cumulative MO length (μm) 14129 14090 14004 14042 14049 14172 13986 14073 13941 13941 14064 14129 12020 11876 14127 14055 14237 14168 14153 14100 14201 14189 14199 14006 Cumulative MO length (μm) NR 4535 4524 4494 4517 4508 4496 4478 4461 4436 4377 4475 4431 4097 4074 4481 4482 4541 4497 4458 4517 4466 4482 4485 4518

Notes: BC = band contrast, BS = band slope, S.D. = standard deviation, MAD = mean angular deviation, MO = misorientation, NR = noise reduction. Match unit superscripts (a–f ) refer to source of data used to create the match unit: a = Robinson et al. (1971); b = Hazen and Finger (1979); c = Rios et al. (2000); d = Kolesov et al. (2001); e = Farnan et al. (2003); f = Finch et al. (2001).* HKL Technology’s “Best in Family.”

Table 3.—ExtEndEd

REDDy ET AL.: ELECTRON BACkSCATTER DIFFRACTION ANALySIS OF ZIRCON196

FIgure 6. Principal component analysis (only components 1 and 2 are shown) of EBSD variables associated with each of 24 match units. (a) Udachnaya sample (UX); (b) VIET1; (c) Lewisian sample (GST15); (d) Javanese sample (JAVA). Symbols for match units (see legend) follow those used in Figure 1. In all cases, EBSD variables are the mean values derived for each match unit/sample pair. MAD = mean MAD; BC = mean band contrast; BS = mean band slope; CMO = cumulative misorientation boundary length; CMO-NR = noise-reduced cumulative misorientation boundary length. Percentage values for the components (C1; C2) indicate the variance within the data that is described by the component.

REDDy ET AL.: ELECTRON BACkSCATTER DIFFRACTION ANALySIS OF ZIRCON 197

analysis of natural zircon. The integration of EBSD data with principal component

analysis and self-organizing map networks presented in this paper represents the first systematic analysis of the theoretical match units required for indexing of empirically obtained EB-SPs. In this case, the mineral zircon was specifically targeted. However, the method provides a means of assessing match units characteristics in all other crystalline materials. The application of this approach to other crystalline phases, therefore, has the potential to yield better EBSD results across the material science and geoscience disciplines.

acknowledgMentSThe Australian Research Council (DP0664078) and a Curtin University

Targeted Research Fellowship provided funding to undertake this research. Pete kinny and Joseph Hamilton are thanked for supplying the various zircon samples used in this study. Rachel Beane and Nicola Cayzer are thanked for detailed and constructive reviews. This paper is “The Institute of Geoscience Research” (TIGeR) publication no. 62.

reFerenceS cItedBoyd, F.R., Pokhilenko, N.P., Pearson, D.G., Mertzman, S.A., Sobolev, N.V., and

Finger, L.W. (1997) Composition of the Siberian cratonic mantle: Evidence from Udachnaya peridotite xenoliths. Contributions to Mineralogy and Petrol-ogy, 128, 228–246.

Chichagov, A.V., Varlamov, D.A., Dilanyan, R.A., Dokina, T.N., Drozhzhina, N.A., Samokhvalova, O.L., and Ushakovskaya, T.V. (2001) MINCRyST: A crystallographic database for minerals, local and network (WWW) Versions. Crystallography Reports, 46, 876–879.

Claves, S.R. and Deal, A. (2005) Orientation dependence of EBSD pattern quality. Microscopy and Microanalysis, 11(S02), 514–515.

Davis, J.C. (2002) Statistics and Data Analysis in Geology, 638 p. Wiley, New york.

Farnan, I., Balan, E., Pickard, C.J., and Mauri, F. (2003) The effect of radiation damage on local structure in the crystalline fraction of ZrSiO4: Investigating the 29Si NMR response to pressure in zircon and reidite. American Mineralo-gist, 88, 1663–1667.

Finch, R.J., Hanchar, J.M., Hoskin, P.W.O., and Burns, P.C. (2001) Rare-earth elements in synthetic zircon: Part 2. A single-crystal X-ray study of xenotime substitution. American Mineralogist, 86, 681–689.

Hazen, R.M. and Finger, L.W. (1979) Crystal structure and compressibility of

zircon at high pressure. American Mineralogist, 64, 196–201.Hough, P.V.C. (1962). Method and Means for Recognizing Complex Patterns.

U.S. Patent 3,069,654. kinny, P.D. and Friend, C. (1997) U-Pb isotopic evidence for the accretion of dif-

ferent crustal blocks to form the Lewisian Complex of northwest Scotland. Contributions to Mineralogy and Petrology, 129, 326–340.

Kinny, P.D., Griffin, B.J., Heaman, L.M., Brakhfogel, F.F., and Spetsius, Z.V. (1997) SHRIMP U-Pb ages of perovskite from yakutian kimberlites. Russian Geology and Geophysics, 38, 97–105.

kohonen, T. (2006) Self-Organizing Maps, 362 p. Springer, Berlin.kolesov, B.A., Geiger, C.A., and Armbruster, T. (2001) The dynamic properties

of zircon studied by single-crystal X-ray diffraction and Raman spectroscopy. European Journal of Mineralogy, 13, 939–948.

Nowell, M.M., Witt, R.A., and True, B.W. (2005) EBSD sample preparation: Techniques, tips, and tricks. Microscopy Today, 13, 44–48.

Prior, D.J., Boyle, A.P., Brenker, F., Cheadle, M.C., Day, A., Lopez, G., Peruzzo, L., Potts, G.J., Reddy, S., Spiess, R., Timms, N.E., Trimby, P., Wheeler, J., and Zetterström, L. (1999) The application of electron backscatter diffraction and orientation contrast imaging in the SEM to textural problems in rocks. American Mineralogist, 84, 1741–1759.

Reddy, S.M., Timms, N.E., Trimby, P., kinny, P.D., Buchan, C., and Blake, k. (2006) Crystal-plastic deformation of zircon: A defect in the assumption of chemical robustness. Geology, 34, 257–260.

Reddy, S.M., Timms, N.E., kinny, P.D., Buchan, C., and Trimby, P. (2007) Natural plastic deformation of zircon and its geological significance. Contributions to Mineralogy and Petrology, 153, 625–645.

Ríos, S., Malcherek, T., Salje, E.k.H., and Domeneghetti, C. (2000) Localized defects in radiation-damaged zircon. Acta Crystallographica, B56, 947–952.

Robinson, k., Gibbs, G.V., and Ribbe, P.H. (1971) The structure of zircon: A comparison with garnet. American Mineralogist, 56, 782–790.

Smyth, H.R., Hamilton, P.J., Hall, R., and kinny, P.D. (2007) The deep crust beneath island arcs: Inherited zircons reveal a Gondwana continental fragment beneath East Java, Indonesia. Earth and Planetary Science Letters, 258, 269–282.

Timms, N.E., kinny, P.D., and Reddy, S.M. (2006) Enhanced diffusion of Uranium and Thorium linked to crystal plasticity in zircon. Geochemical Transactions, 7, 10.

Wright, S.I., and Nowell, M.M. (2005) EBSD Image Quality Mapping. Microscopy and Microanalysis, 12, 1–13.

Manuscript received april 1, 2007Manuscript accepted septeMber 26, 2007Manuscript handled by paul hoskin