TEXTURE ANALYSIS OF QUARTZ IN A GRANITE MYLONITE BY EBSP-ORIENTATION IMAGING MICROSCOPY

The crystallographic preferred orientation (texture) of the quartz phase in a mylonitic leucogranite from the Santa Rosa mylonite zone was investigated using automated analysis of electron backscattering patterns (EBSP) in the scanning electron microscope (SEM). The separation of the quartz diffraction patterns from patterns of other constituents (feldspar, mica, etc.) in this polymineralic rock was achieved using an image quality parameter. The quartz phase displays a texture typical for high temperature mylonites (c-axis maximum in the intermediate strain direction). The misorientation distribution between next neighbors is dominated by Dauphin6 twins.


INTRODUCTION
The measurement of crystallographic preferred orientation (texture) with bulk diffraction methods is considerably limited for low symmetry and polyphase materials by the resolution and overlap of diffraction peaks. Particularly the texture analysis of polymineralic granitic rocks, which are of great interest to structural geologists, due to their abundance in crystal deformation zones, is complicated by the low symmetry of its main constituents like mica and feldspar. The problem of peak overlap has partially been overcome by the use of neutron diffraction which generally has a higher resolution in 2 (Drechsler et al., 1988;Wenk and Pannetier, 1990) and allows for better peak deconvolution, or new approaches with profile analysis (Lutterotti et al., 1997). However, the use of neutron diffraction is limited by the small number of available facilities. A different approach is the measurement of individual crystal orientations either with optical methods (U-stage) or by electron diffraction. Automated analysis of EBSPs or OIM (orientation imaging microscopy) was originally developed to study metals Wright, 1992;Schmidt and Olesen, 1989). Since then it has also been applied to minerals with trigonal symmetry 1994).
Unlike traditional techniques for measuring textures (X-ray and neutron diffraction), this method links lattice orientation of single crystals with the microstructure of a sample.
The present study, conducted at Los Alamos National Laboratory's Center for Material Science, is the first application of automated EBSP analysis to a complex polyphase rock. An image quality (IQ) parameter is used in order to separate orientation measurements of quartz from measurements of other minerals in a sample of mylonitic leucogranite.
EBSPs are generated in a scanning electron microscope (SEM) when a stationary electron beam (diameter of 0.2 tm) penetrates the surface of a crystal. Backscattered electrons which strike crystallographic planes are preferentially diffracted according to Bragg's law and form cones of radiation resulting in a pattern known as EBSP. Each band in an EBSP represents a specific lattice plane (hkl). Band directions are a function of crystal orientation. Band widths are a function of interplanar spacing. Therefore, the geometrical arrangement of bands can be used to index a pattern (assign hkl's to bands) and to determine the lattice orientation of a local region. Schmidt and Olesen (1989) proposed a method for digitally analyzing EBSPs. Using this method, Wright and Adams (1992) introduced a completely automated system which requires no intervention from the operator once a data collection run begins. A computer controls translations of a x-y stage, a camera records the EBSP which is processed and digitized. The recognition of the diffraction bands in the pattern is achieved by a Hough transform, which converts lines TEXTURE ANALYSIS BY EBSP 187 (bands) into single maxima (Illingworth and Kittler, 1988). The computer indexes the pattern, calculates orientation angles and determines an image quality parameters that depends on the contrast of the EBSP (Wright, 1993).
Regardless of pattern quality, the computer will try to index the EBSP. In order to assess the accuracy of computer indexing, two parameters are used: an image quality (IQ) parameter and a match quality (MQ) parameter. The IQ parameter evaluates the clarity of the digitized pattern. If a Hough transform contains only a few broad maxima, a low IQ is assigned. Conversely a sharp pattern with many Hough maxima has a high IQ.
The MQ parameter evaluates the similarity between the SEM generated EBSP and the computer indexed EBSP. Once determined, the MQ parameter can be used to remove measurements with a poor match. Unfortunately, the TexSEM Laboratories software system used in this study did not provide MQ parameters.

Sample Description
The sample used in this study, PC374, is a mylonitic leucogranite from the Palm Canyon Formation in the Santa Rosa mylonite zone (SRMZ) in southern California. Mylonitization has been associated with shear movements and the emplacement of the Peninsular Ranges batholith in the late Cretaceous (Erskine and Wenk, 1985;Simpson, 1984). The leucogranite, primarily composed of quartz and altered feldspar, has a strong mylonitic foliation and lineation. Small amounts of biotite are found along the perimeter of high-angle grain boundaries. The microstructure of this sample ( Fig. 1) is represented by large coherent feldspar grains (1-2mm "clasts") and elongated quartz grains which are flattened parallel to the foliation characteristic of the so-called 'granulite petrographic texture'.

Sample Preparation
Three perpendicular circular sections, 2.5 cm in diameter and 0.5 cm thick ( Fig. 2), were cut normal to the foliation and parallel to the 188 A. PEARSON et al.

I mm FIGURE
Transmitted light photomicrograph of PC374 in thin section, illustrating the microstructure. Crossed polars. Diagonal platelets are mainly quartz, large rounded grains are feldspar. lineation (A), parallel to the foliation (B), and perpendicular to both the foliation and the lineation (C) (Fig. 3). The samples were mechanically ground using diamond paste (smallest grain size 0.25 lam) and enclosed in an epoxy mount. Surface damage induced by grinding was removed by polishing the samples for 24h with a Buehler vibratory polisher in a solution of 0.05 lam colloidal silica. The surface of the epoxy was painted with carbon in order to reduce charging. After SEM measurements, sections were etched with hydrofluoric acid (HF) in order to increase the contrast for observations in reflected light.

Experimental Settings
Charging effects in the SEM were reduced by using a low accelerating voltage of 12 kV. Each of the three samples was run at least twice. Measured areas ranged from mm 2 to cm 2 and were scanned in a hexagonal grid witil an average step-size of 25 lam, resulting in at least 6000 measured points per run. Table I

Microstructures
Microstructural maps were constructed from orientation measurements. Figure 4(a) is a grain boundary map of a single quartz grain.
Each point is compared to its six neighbors on the hexagonal grid. If the orientation difference (misorientation) between two measurements is greater than 10 , a black line is drawn between the two points. In the same figure, points with an IQ greater than 35 are highlighted in black.
Notice that the majority of black lines, separat.ing cells, are concentrated outside of the gray area, i.e. not in quartz. This is because in areas with low IQ values (presumably feldspar and mica grains) the solutions found by the algorithm are more or less random, producing large misorientations between data points. The boundaries outlining hexagonal cells can therefore not be interpreted as high angle grain boundaries. The test is important because it shows that the indexing algorithm does not interpret a feldspar EBSP as a quartz orientation. The sample area can also be imaged as a function of IQ value (Fig. 4(b)), assigning each cell different shades of gray for different IQ values. In this figure, cells with an IQ greater than 35 are white. Darker gray shades indicate lower IQ values. Comparison between Fig. 4(a) and (b) shows that the orientation imaging technique maps out reliably the contiguous quartz grain with IQ > 35.
The same technique is applied to a large sample area in section C (Fig. 5(b)). The image contrast (IQ) map corresponds closely with the microstructure, photographed with a reflected light microscope (Fig. 5(a)). Shades of gray darken as IQ value decreases. There is a good correlation between high IQ and quartz.

TEXTURE ANALYSIS BY EBSP 193
Textures Measurements with a low IQ value were rejected from orientation calculations. All other individual orientation measurements from the three sections were rotated into the same orientation (C section, normal to foliation and lineation). Orientations were then entered into boxes in the ODF using discrete methods (Matthies and Vinel, 1993) and assigning to each orientation a Gaussian of 15 full width at halfmaximum (FWHM). From the ODF, pole figures for various lattice planes were calculated with programs in the BEARTEX package (Wenk et al., 1996). Eight pole figures were calculated for each section and for a combination of the three sections (Figs. 6 and 7). The combination ODF was also smoothed by applying a Gauss bell of 10 FWHM. Figure 6 displays pole figures for poles (hkil) with hexagonal symmetry (i.e. pyramids (hh2k 1) and prisms) and Fig. 7 displays those with trigonal symmetry, separating positive rhombs (h 0h 1) and negative rhombs (0 h h 1). The lineation X is in the center of the pole figure, Y is perpendicular to the lineation in the foliation plane, and Z indicates the normal to the foliation plane.
Observe the good agreement between the pole figures from the three sections. (0001) pole figures consistently display slightly asymmetric caxis maxima which lie preferentially in the foliation plane and perpendicular to the lineation. Pole figures from first-order prisms (1010) and second-order prisms (1020) exhibit girdles which trend N-S.
Maxima are located at approximately 60 intervals along the girdle, corresponding to the angle between a-axes. Notice that maxima on this girdle in (1010)

Misorientations
The orientation difference (misorientation) between adjacent crystallites is expressed with the statistical misorientation distribution in axisangle space (Fig. 8)

DISCUSSION
The results of texture analysis in this mylonitic leucogranite agree with results from earlier neutron diffraction experiments on granitic rocks from Palm Canyon (Wenk and Pannetier, 1990). The texture pattern is also similar to previously documented quartz textures in quartzite mylonites with an asymmetric c-axis double maximum in the plane of foliation and at a high angle to the lineation (e.g. Helming et al., 1994;Schmid and Casey, 1986). Rather than an a-axis maximum (1120) parallel to the lineation we observe a concentration of (1010). It has been suggested that this texture is generated by recrystallization (Helming et al., 1994). According to models, quartz textures with c-axis maxima at high angles to the direction of maximum compression may form during dynamic recrystallization (Jessell and Lister, 1990;Wenk et al., 1989;Wenk, 1994). Unfortunately, experimental studies have so far been unable to produce this texture in the laboratory (Gleason et al., 1993).
The results of the misorientation distribution display a sharp maximum at = 55-60 (Fig. 7). This angle represents the 2-fold rotation around the c-axis which produces Dauphin6 twins. An earlier study by Kunze et al. (1993) used automated analysis of EBSPs to reveal Dauphin6 twins and image the location of twin boundaries in a naturally recrystallized quartzite. Dauphin6 twins cannot be observed with conventional light microscopy. They have been documented as mechanical twins ia experimentally deformed quartzites (Tullis and Tullis, 1972). They also form during a cooling phase transition from hexagonal -quartz to trigonal c-quartz (573C at atmospheric pressure) in quartzite (Barber and Wenk, 1991). Considering the geologic history of this metamorphic sample, Dauphin6 twins are attributed to mechanical or thermal stresses.
The results of this study also demonstrate that OIM-EBSP can be used to reliably determine the texture in polymineralic rocks. The IQ factor can be applied to discriminate between different phases. In our case the sample polishing technique preferentially prepared quartz for generating EBSPs. The agreement between the contrast map ( Fig. 5(b)) and the photomicrograph (Fig. 4(a)) is remarkable. Quartz grains (mainly colored white in both figures), whether elongated stringers, included clasts or large grains, can be easily distinguished from the surrounding material in the sample. Additionally, subtleties such as grain boundaries are not lost on the contrast map. For example, the large quartz grain at the bottom of the image is cut by a grain boundary which appears in both the photograph and the map.
However, the IQ mapping is not foolproof for phase identification. The IQ limit picked for each measurement is somewhat arbitrary and there are a couple of feldspar grains that do show high IQ values. One is located along the right border of the photograph and the other is directly to its left. However, the wrong solutions assigned to the feldspar or mica patterns tend to be random and so inclusion of some of them into the texture calculation does not change the type of texture found. If an MQ factor had been available in this study, it would have been possible to exclude such grains from the measurements. Alternatively by simultaneous recording of EBSP and an energy dispersive X-ray fluorescence signal, phases could be assigned. Without such additional information it is conceivable that EBSP patterns from other minerals were recorded as quartz and falsified the total texture to a small degree.

CONCLUSIONS
This study documents the successof using an IQ factor generated by automated analysis of EBSPs to isolate orientation measurements of quartz from measurements of other phases in a polyphase rock. The accuracy of texture analysis performed on isolated measurements is confirmed by previously documented textures of mylonitic quartz. This technique will be useful in studying how quartz grains interact with or are affected by other minerals during deformation.