The electron backscatter diffraction (EBSD) analysis procedure for crystal preferred orientations to reduce loss of EBSD map data: An example of antigorite

Abstract

Indexing Kikuchi or electron backscatter diffraction (EBSD) patterns from rock samples provides information on mineral phases and their crystal orientation and maps of this information are widely used for mineral microstructural analysis. However, uncertainties in the results due to possible problems of mis-indexing have not been closely studied. One approach to identify unreliable data is to use a filter based on maximum mean angular deviation (MAD) values. Mechanically weak minerals, in particular hydrous sheet silicates, are particularly challenging to analyze due to surface damage formed during sample preparation and electron beam irradiation. The application of stringent MAD filters to extract high-quality accurate data for minerals such as antigorite drastically reduces the proportion of data that can be used for analysis in particular in surfaces cut perpendicular to rock foliation. Combining the MAD data with the number of the EBSD Kikuchi diffraction bands provides a new improved protocol to filter EBSD data that identifies a higher proportion of high-quality data (up to >∼ 50%) compared to the conventional method using only MAD (∼ 5-29%), even from sample surfaces normal to the foliation, while maintaining the same accuracy as the conventional method. Use of the improved protocol will enable more reliable EBSD data sets to be provided which will improve any associated microstructural analyses. Previously reported crystallographic pole figure (CPO) patterns for hydrous sheet minerals such as antigorite may have been calculated from data sets that include a significant proportion of incorrect crystal orientations. Use of the approach proposed in this study may improve our understanding of rock structure, in particular where hydrous sheet minerals are important, including rheological studies of subduction interfaces that involve mineral deformation mechanisms and bulk rock anisotropy.

INTRODUCTION

Determining the mineral phase and the crystal orientation on a sample surface is based on the goodness of fit between the detected and simulated electron backscatter diffraction (EBSD) patterns, known as the Kikuchi pattern (e.g., Kikuchi, 1928; Nishikawa and Kikuchi, 1928a, 1928b). The simulated Kikuchi pattern is calculated using the crystal structure database of appropriate candidate minerals. The degree of fit at each EBSD measurement point is generally expressed as the average value of the angular differences with units of degrees between the detected and theoretically simulated Kikuchi diffraction bands. This value is also known as the mean angular deviation (MAD) value. Therefore, the EBSD indexing points with lower MAD values are interpreted as data points with more accurate mineral phase and orientation determinations. The indexed information (such as the mineral phase and its crystal orientation) of some measured points with high MAD values by EBSD is generally filtered out and not used during the subsequent analysis flow for microstructural analysis as it is thought to contain inaccurate data.

The acceptable upper limit of MAD values (MAD_max) used in the accurate crystal orientation analyses is typically 1.3° (e.g., Franke et al., 2007). However, studies of the impact of filtering conditions on EBSD data using MAD_max on the results of subsequent analyses are limited. There are few examples of studies of MAD_max setting conditions for hydrous sheet silicates even though the conditions for setting the appropriate MAD_max of hydrous sheet silicates, which are difficult to measure by the EBSD system, may differ from those for other common minerals (Nagaya et al., 2017, 2020). In analysis of crystal preferred orientation (CPO) of antigorite based on EBSD measurements, only a few previous studies of antigorite CPOs provide information on the value of MAD_max used, but where given it is generally MAD_max < 1.3° (see references in Nagaya et al., 2017). However, Nagaya et al. (2017) proposed when using the sample plane normal to the foliation a more stringent filtering condition of MAD_max of <0.7° is needed to obtain accurate antigorite CPO patterns in the EBSD measurement compared with using the sample plane parallel to the foliation.

While a lower MAD_max condition can extract only EBSD data with more accurate information of mineral phase and orientation, a too low MAD_max condition not only reduces the number of analysis points available, but also increases the number of regions which are under-represented in the EBSD map. The existence of such gaps introduces uncertainty into interpretations of microstructure that require information such as grain size, shape, and crystal orientation derived from EBSD maps. Examples include estimating mineral deformation mechanisms and the bulk anisotropy. Therefore, setting the filtering conditions for EBSD data is essential not only for the calculation of mineral CPO patterns but also for general microstructural analysis using EBSD.

In this study, we investigated whether the number of Kikuchi diffraction bands used in EBSD indexing could be used as an index of the accuracy of measurement results in addition to MAD_max. If EBSD indexing is successful when more Kikuchi bands from the detected bands are used without a significant increase in the MAD value, the indexing results may be more accurate than when fewer bands are used. However, as with MAD_max, the more stringent conditions used during EBSD indexing (i.e., in this case, the more Kikuchi diffraction bands detected and used for indexing), the fewer data points will be available after filtering. Here, the minimum number of Kikuchi diffraction bands, defined as Band_min, is also used as a filtering condition.

In this study, we extract data with high Band_min from the indexing data removed from the original full data set by previously filtering using MAD_max in the CPO analysis process. These extracted indexing data with high Band_min may be reused for CPO analysis as relatively accurate analysis points, even among data with high MAD_max. If this approach is successful, fewer analysis points will be removed than with previous analysis methods using MAD_max, reducing missing areas on the EBSD map and enabling more complex and accurate microstructural analysis.

This study takes antigorite as an example, where the accuracy of indexing results by EBSD analysis has been questioned (e.g., Van de Moortele et al., 2010; Nishii et al., 2011; Padrón-Navarta et al., 2012; Soda and Wenk, 2014; Nagaya et al., 2017). We investigate the combinations of MAD_max and Band_min that show results accurate enough to be used for CPO analysis. This may allow us to find an approach for processing EBSD mapping data that has a higher proportion of available analysis points than conventional methods, while maintaining accuracy of CPO patterns comparable to previous analysis methods.

DESCRIPTIONS OF MATERIALS, AREAS, AND METHODS OF STUDY

EBSD system and antigorite indexing data

EBSD mapping of antigorite crystal orientations based on auto-indexing was constructed using a scanning electron microscope (SEM) (JSM-6510LV, JEOL) equipped with an EBSD system [a NordlysNano detector controlled by AZtecHKL software (ver. 2.3 in Nagaya et al., 2017), Oxford Instruments] and AZtec and HKL Channel5 analysis software (Oxford Instruments) installed at originally at Nagoya University (now at the University of Tokyo) and following the procedure described by Nagaya et al. (2017, 2020, 2022), which also described methods to minimize the formation of damaged layers during sample preparation, such as damage caused by polishing and electron beam irradiation. The collection of Kikuchi patterns and automatic indexing were performed using resolution images with 4 × 4 pixel binning. The ranges of detected bands and maximum MAD values were 5-10, and 0° < MAD_max < 2.0°. The antigorite crystallographic parameters reported by Uehara (1998) were used for EBSD indexing the detected Kikuchi patterns.

Increasing the number of candidate minerals for phase identification at the time of EBSD measurement causes partly incorrect phase identification to different minerals as well as mis-indexing of crystal orientation. In general, the probability of noise generation on the EBSD map due to incorrect mineral identifications is low, and such noise is corrected by noise-reducing processing in the EBSD map analyses after the measurement. However, especially in cases where a clear Kikuchi pattern cannot be obtained and identification by EBSD is difficult, such as some hydrous minerals, insufficient surface polishing conditions and too fine grain sizes, the noise that is incorrectly identified to different mineral phases within such an EBSD measurement area, including antigorite, often occur. In this case, as with the mis-indexing of crystal orientations of the same mineral in this study, it may be difficult to remove the noise in the map sufficiently after the measurement using the usual noise reduction processes. Noise from such incorrect mineral phase identifications may be an important concern because it introduces uncertainty into the post-measurement microstructural analysis using EBSD maps. In this study, we focus on examining incorrect antigorite crystal orientations by mis-indexing of antigorite and included only antigorite in the EBSD data.

In this study we reanalyzed the antigorite EBSD mapping datasets reported in Nagaya et al. (2017). These data were collected from three antigorite-schist samples from the Besshi (BD1), Happo (HP5) and Saganoseki (serpentinite B) areas, and one antigorite vein sample with a fiber lineation fabric (NSY8-12) from the Shiraga area. All study areas are located in Japan (see Nagaya et al., 2017 for details). In our analyses we only used MAD_max and Band_min to filter the data and did not apply any noise reduction and cleanup functions such as the Pseudosymmetry and Wild Spikes correction functions incorporated in software such as Channel5 and AZtecCrystal for EBSD map data analyses after EBSD measurements. The CPO figures were prepared using the Fortran software developed and provided by D. Mainprice (Mainprice, 1990) on macOS (Mojave 10.14.6).

Threshold values of MAD_max (from <0.5 to <2.0°) were used to filter the antigorite EBSD data of antigorite and calculate the post filtering antigorite CPO patterns. The same antigorite crystallographic parameters (Uehara, 1998) used for EBSD indexing of the detected Kikuchi patterns were also used for CPO calculations of antigorite. Contours of the CPOs are shown as of multiples of uniform distribution (m.u.d.).

Antigorite EBSD maps from the Besshi (BD1) (Nagaya et al., 2017) and Happo (HP5) (Nagaya et al., 2014, 2017) areas were collected using the XZ-plane (normal to the foliation and parallel to the lineation) and the XY-plane (parallel to the foliation). Antigorite EBSD maps from Saganoseki (serpentinite B) (Soda and Takagi, 2010; Soda and Wenk, 2014; Nagaya et al., 2017) and Shiraga (NSY8-12) (Nagaya et al., 2017) areas were collected using the XZ-plane. The foliation is defined by the orientation of platy antigorite grains and the mineral lineation is defined by the boudinage direction of chrome spinel and magnetite grains. In Nagaya et al. (2017), verification of the exact orientations of antigorite indexing data was carried out by comparison with TEM measurements, and antigorite CPO pattern consistent with TEM results was B-type (defined by the concentration of the b-axes parallel to the lineation and the c-axes normal to the foliation) for these three antigorite-schists and one antigorite vein sample. The good characterization of these data mean they are well suited for examining the effects of filtering using a combined Band_min and MAD_max conditions.

Sample descriptions and details of the associated geological backgrounds are given in Nagaya et al. (2017) for the Besshi and Shiraga samples; in Nagaya et al. (2014, 2017) for the Happo sample; and in Soda and Takagi (2010), Soda and Wenk (2014) and Nagaya et al. (2017) for the Saganoseki sample, respectively.

RESULTS AND DATA

Effect of filtering conditions in Band_min and MAD_max on resulting antigorite CPO patterns

The EBSD measurements of Nagaya et al. (2017), which were used to examine the optimal MAD_max filtering condition of antigorite indexing data (relaxed filtering conditions have higher MAD_max and strict filtering conditions have lower MAD_max) were re-examined to determine the optimal Band_min filtering condition (where relaxed filtering conditions have lower Band_min and strict filtering conditions have higher Band_min) combined with MAD_max. A total of 3-8 Kikuchi bands are generally used in EBSD measurements (e.g., Randle and Engler, 2000). The data by Nagaya et al. (2017) used in this study are based on a greater number (5-10) of Kikuchi bands. The average number of Kikuchi bands used for the antigorite indexing in this study shows a range of ∼ 7-8. Regardless of MAD_max filtering conditions, the average number of Kikuchi bands of the antigorite indexing data remaining after MAD_max filtering also showed a similar range of ∼ 7-8 in this study.

The type classification results of antigorite CPO patterns calculated from EBSD indexing data filtered using different combinations of Band_min and MAD_max and the ratios of available indexing data of antigorite after filtering are summarized in Figures 1-4, S1, and S2 (Supplementary Figs. S1-S9 are available online from https://doi.org/10.2465/jmps.241210). The original data of antigorite CPO patterns in Figures 1-4, S1, and S2 used for the type classification are shown in Figures S3-S8, respectively. Antigorite CPO patterns using the XZ-plane are summarized in Figures 1 and S3 for the Besshi area, Figures 2 and S4 for the Happo area, Figures S1 and S7 for the Saganoseki area, and Figures S2 and S8 for the Shiraga area. Antigorite CPO patterns using the XY-plane are summarized in Figures 3 and S5 for the Besshi area, and Figures 4 and S6 for the Happo area. In this study, cases where more than 1% of all indexed data remained after filtering by the combination of Band_min and MAD_max are included in the following examination of the change in antigorite CPO patterns.

Figure 1. Changes in type of CPO patterns associated with filtering mis-indexing using Figure 3 of Nagaya et al. (2017) (the XZ-plane from the Besshi area, BD1). Each box shows the most concentrated orientation (black square) in the CPO of the [100] axis (upper left), [010] axis (upper right), and the pole to the (001) plane (lower left). The lower right shows the number and the percentage of data remaining and used for analysis after filtering. The background color is changed when the most concentrated orientation of the a-axis rotates nearly 90°: pattern changes from the B-type pattern CPO (blue) to CPOs related to the rotation due to antigorite mis-indexing around the c-axes are shown in red. When the number of antigorite indexing data remaining after filtering is below 1%, the background colour is black. Even when MAD_max of 1.3° is used, the change in the background color in this figure due to changes in Band_min is the same as when MAD_max of 2.0° is used.

Figure 2. Changes in type of CPO patterns associated with filtering mis-indexing using Figure 4 of Nagaya et al. (2017) (the XZ-plane from the Happo area, HP5). The background color is changed when the most concentrated orientation of the a-axis rotates nearly 90°: pattern changes from the B-type pattern CPO (blue) to CPOs related to the rotation due to antigorite mis-indexing around the b-axes are shown in green. All other features are shown in the same way as in Figure 1.

Figure 3. Changes in type of CPO patterns associated with filtering mis-indexing using Figure 6 of Nagaya et al. (2017) (the XY-plane from the Besshi area, BD1). All other features are shown in the same way as in Figure 1.

Figure 4. Changes in type of CPO patterns associated with filtering mis-indexing using Figure 7 of Nagaya et al. (2017) (the XY-plane from the Happo area, HP5). All other features are shown in the same way as in Figure 1.

1) XZ-plane of antigorite schists. When using EBSD mapping data obtained from the XZ-plane (Figs. 1, 2, S1, S3, and S4), the effect of mis-indexing associated with a rotation about the antigorite c-axis on the resulting antigorite CPO is generally more pronounced for relaxed filtering conditions (higher MAD_max and lower Band_min) (Figs. 1, 2, S3, and S4). Antigorite CPO calculated using relaxed filtering conditions show a change from those with characteristics of B-type antigorite CPO to those more similar A-type antigorite CPO (defined by the concentration of the a-axes parallel to the lineation and the c-axes normal to the foliation) (e.g., Katayama et al., 2009) and G-type antigorite CPO (defined by the girdle distribution of the a- and b-axes parallel to the foliation and the concentration of the c-axes normal to the foliation) (e.g., Morales et al., 2013). These changes are associated with rotation of the concentrations of the antigorite a- and b-axes (Figs. 1, 2, S3, and S4).

On the other hand, when using the data from the XZ-plane (Figs. 1, 2, S1, S3, S4, and S7), the effect of the mis-indexing associated with rotation around the antigorite b-axis on the resulting antigorite CPO can be more pronounced under stringent filtering conditions (lower MAD_max and higher Band_min) (Figs. 2 and S4). Antigorite CPO calculated using progressively more stringent filtering conditions showed a change from those with characteristics of B-type antigorite CPO to those more similar to L-type antigorite CPO (defined by the concentration of the b-axes parallel to the lineation and the girdle distribution of the a- and c-axes rotated around the b-axes) (Liu et al., 2020) reflecting rotation of the concentration of the antigorite a-axes (Figs. 2 and S4).

In addition, when MAD_max and Band_min were combined, accurate antigorite CPOs with characteristics of B-type antigorite CPO in samples of Nagaya et al. (2017) are obtained even in the cases of relaxed filtering conditions of MAD_max more than 0.7°. The combination of MAD_max (∼ 1.0° or less) and Band_min (∼ 7 or more) also can allow more indexing points (up to 50% or more) with accurate orientations to remain after filtering than when only MAD_max is used to filter the data (Figs. 1, 2, S1, S3, S4, and S7). This means that the combination of MAD_max and Band_min can obtain a larger number of reliable crystal orientations, which can be used for CPO analysis from the same EBSD mapping area than filtering methods using only MAD_max. This combined approach to filtering can reduce the risk of determining CPO patterns from biased sample data, which is caused by the fact that the stringent MAD filtering restrictions on the data from a limited mapping area set reduce the number of data available for CPO analysis. This technique is particularly useful for EBSD mapping of small areas where the sample area of interest is limited.

2) XY-plane of antigorite schists. Antigorite CPOs calculated from EBSD mapping of XY-planes show no significant changes from B-type antigorite CPO, regardless of the filtering conditions of MAD_max and Band_min (Figs. 3, 4, S5, and S6). This is the same result as in a previous study that used only MAD_max (Nagaya et al., 2017).

3) XZ-plane of the antigorite vein. When a single antigorite crystal on the XZ-plane was used to collect EBSD mapping data, the probability of obtaining the accurate orientation is not high even with Band_min combined unless MAD_max is set to 0.7° (Figs. S2 and S8). There are not always more accurate indexing points available for CPO analysis using a combination of MAD_max and Band_min that for MAD_max alone (Figs. S2 and S8).

INTERPRETATION AND DISCUSSION

Effect of filtering conditions in Band_min and MAD_max on resulting antigorite CPO patterns

1) Filtering conditions of EBSD mapping data to obtain the accurate indexing data in the previous studies. The use of EBSD mapping as a method for measuring the crystal orientations of minerals has become widespread. However, a measurement protocol for obtaining reliable results that minimize the effect of mis-indexing has not been established. This issue is particularly important for hydrous sheet minerals, such as antigorite, talc, brucite, and mica, which are known to be difficult to index accurately using EBSD. Several studies discuss the SEM measurement environment, such as the absence or presence (and its thickness) of a coating on the sample surface, the degree of vacuum, accelerating voltage, and magnification (e.g., van de Moortèle et al., 2010; Padrón-Navarta et al., 2012, 2015; Nagaya et al., 2017; Horn et al., 2020; Nagaya et al., 2020, 2022), as well as the EBSD measurement conditions, such as indexing when using EBSD in combination with energy dispersive X-ray spectroscopy (EDS), the step size during mapping, and the resolution of the Kikuchi pattern images acquired by the EBSD detector’s charge-coupled device (CCD) camera (e.g., van de Moortèle et al., 2010; Brownlee et al., 2013; Nagaya et al., 2017). In addition, it has been demonstrated that the polishing procedure of sample surfaces for sample preparation changes the probability of identifying minerals during measurement and the optimal conditions to obtain accurate crystal orientation during the analysis of the measured data, such as for antigorite (e.g., Nishii et al., 2011; Padrón-Navarta et al., 2012; Nagaya et al., 2017), talc (Nagaya et al., 2020), and brucite (Nagaya et al., 2022).

Previous studies have shown that highly accurate antigorite crystal orientations can be obtained from EBSD indexing data even with relatively less stringent filtering conditions (e.g., MAD_max of 1.3°) by using a polished sample surface parallel to the foliation (Nagaya et al., 2017). This approach has also been utilized in the successful examination of other minerals with a strongly layered crystal structure such as graphite (Kouketsu et al., 2019) and talc (Nagaya et al., 2020). However, when using sample surfaces perpendicular to the foliation (such as the XZ-plane), which is difficult to polish, the indexing data often contain inaccuracies. Therefore, the more stringent filtering conditions of the measured EBSD data are needed to calculate the CPO. Nagaya et al. (2017) showed that when EBSD mapping is performed using the XZ-plane, it has been observed that antigorite mis-indexing rotated around the c-axis occurs most frequently and mis-indexing rotated around the b-axis is the second most common. Therefore, changes in the CPO pattern can occur due to these mis-indexing issues depending on the analytical conditions, such as when data filtering settings are too relaxed or too stringent (Nagaya et al., 2017).

MAD_max of 1.3° is commonly used for filtering indexing data measured by EBSD (e.g., Franke et al., 2007). However, even under more stringent MAD_max conditions of 1.1-1.2° for antigorite indexing data, only ∼ 10-20% of the data is removed by filtering when using a sample surface perpendicular to the foliation. In these MAD_max and surface conditions, inaccurate data cannot be sufficiently removed (Nagaya et al., 2017). Therefore, stringent filtering conditions of MAD_max are needed, such as MAD_max of 0.7° (Nagaya et al., 2017). On the other hand, more stringent MAD_max filtering conditions significantly reduce the number of data that can be used to determine antigorite crystal orientations and used for CPO analyses. Therefore, the disadvantages of this approach are that the crystal orientations by EBSD may not be representative of the microstructure of the entire measurement area and that a wide measurement area with many measurements and long measurement time are required to obtain accurate crystal orientations of the entire microstructure.

2) Approach to examining filtering conditions of EBSD mapping data in this study. In this study, it was investigated whether adding Band_min to the filtering conditions used in the CPO analysis of antigorite indexing data obtained by EBSD, in addition to MAD_max, would allow a large number of accurate measurements to be identified. In EBSD measurements the mineral phase and its orientation are determined by fitting the detected Kikuchi pattern with the theoretically predicted Kikuchi pattern. In this process, the greater the number of bands of the best-fitting Kikuchi pattern used for indexing, the more accurate the results may be. However, this is partly due to the sharpness of the Kikuchi patterns detected. As shown in the result of antigorite single crystal in Figures S2 and S8, the quality of the detected Kikuchi patterns and the optimal combination of MAD_max and Band_min to obtain the accurate CPO pattern can vary on the XZ-plane depending on the types of minerals and their polishing method in addition to the measurement environment. Differences between the crystal structure of the measured mineral and the mineral database used in fitting the theoretical Kikuchi pattern to the detected Kikuchi pattern may also affect the most appropriate values for MAD_max and Band_min.

The indexing rates of antigorite in this study are summarized in Supplementary Tables S1-S6 (Supplementary Tables S1-S6 are available online from https://doi.org/10.2465/jmps.241210). The percentages listed in Tables S1-S6 are limited to the indexing rate for antigorite because the antigorite-rich domain was the target area of the EBSD mapping to investigate the mis-indexing of antigorite in this study. This EBSD mapping of a single phase, antigorite, reduce noise in the EBSD map (see the section, EBSD system and antigorite indexing data). Most of the zero solution data points count the areas of antigorite that could not be indexed as antigorite. Zero solutions also include grain boundaries. In addition, most areas of minor constituent minerals other than antigorite, which are present in small amounts, are also counted as zero solutions. This is because clear Kikuchi patterns can be generally detected from areas of other constituent minerals (e.g., olivine, Cr-spinel, and magnetite), and other constituent minerals are less likely to be misidentified as antigorite by EBSD. Therefore, the antigorite indexing rate for the entire EBSD map area shown in Tables S1-S6 of this study is lower when compared to the EBSD mapping of antigorite-only areas.

In this study, areas of constituent minerals other than antigorite may be incorrectly counted as antigorite indexing points rather than zero solutions. However, such incorrect identification as antigorite in areas of other constituent minerals is a small part of the antigorite indexing data. Theoretically, in such cases there will only be a poor match for antigorite, and these can be removed by the data filtering used in this study and by common noise removal processes. In addition, in this study, the influence of incorrect antigorite indexing data in non-antigorite constituent minerals on the calculated results of the antigorite CPO patterns could not be confirmed. However, it may be necessary to continue to examine the pattern and probability of mis-indexing of each mineral when multiple minerals are used as candidate minerals for EBSD measurements and analysis of the map data.

3) Optimal combinations of Band_min and MAD_max to calculate an antigorite CPO from EBSD mapping data. CPO patterns derived from EBSD map data will more strongly reflect large grains that have more mapping points. This is an appropriate approach when studying the relationship between CPO and bulk rock anisotropy. Another way to calculate CPO from EBSD map data is to use the orientation dataset collected from a single orientation for each grain, which is more appropriate for relating CPO to finite strain. However, this approach may suffer disproportionately from an insufficient number of data by mis-indexing problems during EBSD indexing due to the reduced number of available measurement points from the same mapping area. To assess these issues, we also checked possible differences between the antigorite CPOs calculated from the two distinct types of datasets. For this comparison, we used the mapping dataset from the Besshi area, because this map contained a particularly large number of antigorite grains with well-defined orientations and sufficient data for analysis even after filtering using MAD_max < 0.7°. Antigorite CPOs calculated from a single measurement for each grain did not show any significant differences in antigorite CPO patterns from those calculated from area mapping, which is consistent with results reported from previous studies (Soda and Wenk, 2014; Nagaya et al., 2017).

In this study, the results of the commonly used MAD_max of 1.3° are similar to those for MAD_max of 2.0° for all cases of the XZ- and XY-planes, even when Band_min is combined. Therefore, as a conclusion of the optimal filtering condition of antigorite indexing data by EBSD mapping to calculate the accurate antigorite CPO in this study, the combinations of Band_min of ∼ 7 or more and MAD_max of ∼ 1.0° or less can allow the use of many accurate antigorite crystal orientations after the filtering. When the different filtering combinations from this study are used for antigorite CPO analyses, we recommend using multiple combinations of at least MAD_max less than ∼ 1.0° and Band_min greater than ∼ 7 to examine whether the CPO pattern changes shown in this study occur before setting the filtering combination of MAD_max and Band_min. The analytical flow for calculating antigorite CPO proposed in this study is summarized in Figure S9. When applying the approach used in this study to minerals other than antigorite, the optimal conditions for EBSD analyses can be found by completing the analytical protocol of Figure S9 or by examining how CPO patterns change for a range of conditions, as shown in Figures 1-4 and S1-S8 in this study.

A literature review shows there are numerous examples where different types of antigorite CPO patterns have been reported from the same region. A summary is given below including aspects of the measurement procedure relevant to obtaining a reliable data set with minimum mis-indexing issues.

1. 1)    A-, B-, and L-type antigorite CPOs were reported from Val Malenco, Northern Italy by Jung (2011) (using the XZ-plane and without any description of MAD_max or Band_min conditions), Liu et al. (2018) [using the XZ-plane and MAD_max of 1.3° (average MAD values of 0.73-0.88°) and without any description of the Band_min condition], Liu et al. (2020) [using the XZ- and YZ-planes and MAD_max of 1.3° (average MAD values of 0.73-0.88°) and without any description of the Band_min condition] and Morales et al. (2018) (using the XZ-, YZ- and XY-planes and without any description of MAD_max or Band_min conditions).
2. 2)    A- and B-type antigorite CPOs were reported from Cerro del Almirez massif, SE Spain by Padrón-Navarta et al. (2012) [using the XZ- and YZ-planes and MAD_max of 1.3° (average MAD values of 0.73-0.88°) and without any description of the Band_min condition] and Padrón-Navarta et al. (2015) (using the XZ-plane and MAD_max of 1.3° and without any description of the Band_min condition).
3. 3)    A- and G-type antigorite CPOs were reported from the Escambray massif, Central Cuba by Bezacier et al. (2010) (without any description of the used surface plane, MAD_max or Band_min conditions) and van de Moortèle et al. (2010) (using the plane normal to the foliation and 5-7 Kikuchi bands and without any description of the MAD_max condition).
4. 4)    B- and G-type antigorite CPOs were reported from Punta Bettolina, Western Italy by Jung (2011) (using the XZ-plane and without any description of MAD_max or Band_min conditions), from the Motagua fault system, Central Guatemala by Brownlee et al. (2013) (using Kikuchi bands of 7 or less and without any description of the MAD_max condition) (some of the used planes were parallel to the foliation), from Liaoning Province, NE China by Shao et al. (2014) (using MAD_max of 1° and without any description of used surface planes and the Band_min condition) and from the Sapat Complex, Northern Pakistan by Horn et al. (2020) (using the planes parallel to the XZ-plane and normal to the foliation, and Kikuchi bands of 7 or less and without any description of the MAD_max condition).

There are very few reported examples of antigorite CPO patterns obtained by EBSD in which the details of used sample surfaces, MAD_max and Band_min are all described. A rare exception is Morales et al. (2013) that reported G-type antigorite CPO from the Colorado Plateau, US (using the XZ-plane, MAD_max of 1.3° and 5-7 Kikuchi bands). None of the above previous studies clearly meets the optimal combinations of MAD_max and Band_min to calculate an antigorite CPO proposed in this study and there is a possibility that the results are affected to some degree by mis-indexing issues.

CONCLUDING REMARKS

In this study, in addition to the maximum MAD value (MAD_max), the minimum number of the Kikuchi bands (Band_min) used in the EBSD patterns is also considered as a way to filter EBSD data. This study proposes that the optimal filtering conditions are combinations of Band_min of ∼ 7 or more and MAD_max of ∼ 1.0° or less. Use of these optimal combinations of Band_min and MAD_max does not result in significant changes to the antigorite CPO patterns obtained when only MAD_max < 0.7° was used as a filtering condition. However, use of the combination proposed in this study allows a larger proportion of accurate antigorite to be identified (up to >∼ 50% of the indexed points) compared with using only MAD_max < 0.7° (∼ 5-29%).

The data filtering by combining Band_min and MAD_max can result in the reduction of the missing regions caused by the elimination of incorrect indexing data in the EBSD map. This approach leads to the reduction of the loss of EBSD measurement points and the uncertainty in the EBSD map analysis, which are needed for accurate antigorite CPO patterns and microstructural analyses. The verification method used in this study to determine the optimal filtering conditions of MAD_max and Band_min for antigorite can be similarly applied to find the optimal EBSD analysis conditions for other minerals. Changes in the filtering conditions can lead to B-type antigorite CPO patterns being incorrectly identified as A-type, G-type, or L-type antigorite CPO patterns, this result suggests previously reported antigorite CPO which do not consider issues of potential mis-indexing should be treated with caution.

ACKNOWLEDGMENTS

We thank D. Mainprice for providing the software to analyze the antigorite CPO data. We thank T. Kato for editorial handling and two anonymous reviewers for their valuable comments, which have helped to improve the manuscript. This work was supported by JSPS grants-in-aid Nos. 20K14575, 20KK0079, 23K20895, 24K07187, and 21H05202.

SUPPLEMENTARY/DEPOSITORY MATERIALS

Supplementary Figures S1-S9 and Tables S1-S6 are available online from https://doi.org/10.2465/jmps.241210.

The Fortran software developed by D. Mainprice (Mainprice, 1990) used to analyze the antigorite CPO patterns in this study is available from ftp://www.gm.univ-montp2.fr/mainprice//CareWare_Unicef_Programs/. The data and the analytical conditions used in this study are presented in Nagaya et al. (2017) and in the figures and methods of the main text and Supplementary figures and tables.

REFERENCES

•  Bezacier,  L.,  Reynard,  B.,  Bass,  J.,  Sanchez-Valle,  C. and  Moortèle,  B. (2010) Elasticity of antigorite, seismic detection of serpentinites, and anisotropy in subduction zones. Earth and Planetary Science Letters, 289, 198-208.
•  Brownlee,  S.J.,  Hacker,  B.R.,  Harlow,  G.E. and  Seward,  G. (2013) Seismic signatures of a hydrated mantle wedge from antigorite crystal-preferred orientation (CPO). Earth and Planetary Science Letters, 375, 395-407.
•  Franke,  C.,  Pennock,  G.M.,  Drury,  M.R.,  Engelmann,  R., et al (2007) Identification of magnetic Fe-Ti oxides in marine sediments by electron backscatter diffraction in scanning electron microscopy. Geophysical Journal International, 170, 545-555.
•  Horn,  C.,  Bouilhol,  P. and  Skemer,  P. (2020) Serpentinization, deformation, and seismic anisotropy in the subduction mantle wedge. Geochemistry, Geophysics, Geosystems, 21, e2020GC008950.
•  Jung,  H. (2011) Seismic anisotropy produced by serpentine in mantle wedge. Earth and Planetary Science Letters, 307, 535-543.
•  Katayama,  I.,  Hirauchi,  K.-I.,  Michibayashi,  K. and  Ando,  J.-I. (2009) Trench-parallel anisotorpy produced by serpentine deformation in the hydrated mantle wedge. Nature, 461, 1114-1117.
•  Kikuchi,  S. (1928) Diffraction of cathode rays by mica. Japanese Journal of Physics, 5, 83-96.
•  Kouketsu,  Y.,  Miyake,  A.,  Igami,  Y.,  Taguchi,  T., et al (2019) Drastic effect of shearing on graphite microtexture: attention and application to Earth science. Progress in Earth and Planetary Science, 6, 23.
•  Liu,  W.,  Zhang,  J. and  Barou,  F. (2018) B-type olivine fabric induced by low temperature dissolution creep during serpentinization and deformation in mantle wedge. Tectonophysics, 722, 1-10.
•  Liu,  W.,  Zhang,  J.,  Cao,  Y. and  Jin,  Z. (2020) Geneses of two contrasting antigorite crystal preferred orientations and their implications for seismic anisotropy in the forearc mantle. Journal of Geophysical Research: Solid Earth, 125, e2020JB019354.
•  Mainprice,  D. (1990) Fortran program to calculate seismic anisotropy from the lattice preferred orientation of minerals. Computers & Geosciences, 16, 385-393.
•  Morales,  L.F.G.,  Mainprice,  D. and  Boudier,  F. (2013) The influence of hydrous phases on the microstructure and seismic properties of a hydrated mantle rock. Tectonophysics, 594, 103-117.
•  Morales,  L.F.G.,  Mainprice,  D. and  Kern,  H. (2018) Olivine-antigorite orientation relationships: Microstructures, phase boundary misorientations and the effect of cracks in the seismic properties of serpentinites. Tectonophysics, 724-725, 93-115.
•  Nagaya,  T.,  Wallis,  S.R.,  Kobayashi,  H.,  Michibayashi,  K., et al (2014) Dehydration breakdown of antigorite and the formation of B-type olivine CPO. Earth and Planetary Science Letters, 387, 67-76.
•  Nagaya,  T.,  Wallis,  S.R.,  Seto,  Y.,  Miyake,  A., et al (2017) Minimizing and quantifying mis-indexing in electron backscatter diffraction (EBSD) determinations of antigorite crystal directions. Journal of Structural Geology, 95, 127-141.
•  Nagaya,  T.,  Okamoto,  A.,  Oyanagi,  R.,  Seto,  Y., et al (2020) Crystallographic preferred orientation of talc determined by an improved EBSD procedure for sheet silicates: Implications for anisotropy at the slab-mantle interface due to Si-metasomatism. American Mineralogist, 105, 873-893.
•  Nagaya,  T.,  Okamoto,  A.,  Kido,  M.,  Muto,  J. and  Wallis,  S.R. (2022) Dehydration of brucite + antigorite under mantle wedge conditions: insights from the direct comparison of microstructures before and after experiments. Contributions to Mineralogy and Petrology, 177, 87.
•  Nishii,  A.,  Wallis,  S.R.,  Mizukami,  T. and  Michibayashi,  K. (2011) Subduction related antigorite CPO patterns from forearc mantle in the Sanbagawa belt, southwest Japan. Journal of Structural Geology, 33, 1436-1445.
•  Nishikawa,  S. and  Kikuchi,  S. (1928a) Diffraction of cathode rays by mica. Nature, 121, 1019-1020.
•  Nishikawa,  S. and  Kikuchi,  S. (1928b) Diffraction of cathode rays by calcite. Nature, 122, 726.
•  Padrón-Navarta,  J.A.,  Tommasi,  A.,  Garrido,  C.J. and  Sanchez-Vizcaíno,  V.L. (2012) Plastic deformation and development of antigorite crystal preferred orientation in high-pressure serpentinites. Earth and Planetary Science Letters, 349-350, 75-86.
•  Padrón-Navarta,  J.A.,  Tommasi,  A.,  Garrido,  C.J. and  Mainprice,  D. (2015) On topotaxy and compaction during antigorite and chlorite dehydration: an experimental and natural study. Contributions to Mineralogy and Petrology, 169, 35.
• Randle, V. and Engler, O. (2000) Introduction to texture analysis: macrotexture, microtexture and orientation mapping. pp. 388, CRC Press, London.
•  Shao,  T.,  Ji,  S.,  Kondo,  Y.,  Michibayashi,  K., et al (2014) Antigorite-induced seismic anisotropy and implications for deformation in subduction zones and the Tibetan Plateau. Journal of Geophysical Research: Solid Earth, 119, 2068-2099.
•  Soda,  Y. and  Takagi,  H. (2010) Sequential deformation from serpentinite mylonite to metasomatic rocks along the Sashu Fault, SW Japan. Journal of Structural Geology, 32, 792-802.
•  Soda,  Y. and  Wenk,  H.R. (2014) Antigorite crystallographic preferred orientations in serpentinites from Japan. Tectonophysics, 615-616, 199-212.
•  Uehara,  S. (1998) TEM and XRD study of antigorite superstructures. The Canadian Mineralogist, 36, 1595-1605.
•  van de Moortèle,  B.,  Bezacier,  L.,  Trullenque,  G. and  Reynard,  B. (2010) Electron back-scattering diffraction (EBSD) measurements of antigorite lattice-preferred orientations (LPO). Journal of Microscopy, 239, 245-248.