Distinction between amorphous and healed planar deformation features in shocked quartz using composite color scanning electron microscope cathodoluminescence (SEM-CL) imaging

– Planar deformation features (PDFs) in quartz are one of the most reliable and most widely used forms of evidence for hypervelocity impact. PDFs can be identiﬁed in scanning electron microscope cathodoluminescence (SEM-CL) images, but not all PDFs show the same CL behavior: there are nonluminescent and red luminescent PDFs. This study aims to explain the origin of the different CL emissions in PDFs. Focused ion beam (FIB) thin foils were prepared of speciﬁc sample locations selected in composite color SEM-CL images and were analyzed in a transmission electron microscope (TEM). The FIB preparation technique allowed a direct, often one-to-one correlation between the CL images and the defect structure observed in TEM. This correlation shows that composite color SEM-CL imaging allows distinction between amorphous PDFs on one hand and healed PDFs and basal Brazil twins on the other: nonluminescent PDFs are amorphous, while healed PDFs and basal Brazil twins are red luminescent, with a dominant emission peak at 650 nm. We suggest that the red luminescence is the result of preferential beam damage along dislocations, ﬂuid inclusions, and twin boundaries. Furthermore, a high-pressure phase (possibly stishovite) in PDFs can be detected in color SEM-CL images by its blue luminescence.


INTRODUCTION
The presence of planar deformation features (PDFs) in quartz is one of the most reliable forms of evidence for the impact origin of suspected geological structures or stratigraphic layers. PDFs are planar microstructures that form in quartz and other minerals (e.g., feldspar) as a result of a shock wave passing through the rocks (e.g., papers in French and Short 1968;von Engelhardt and Bertsch 1969;St€ offler and Langenhorst 1994;Grieve et al. 1996) and can be observed in standard thin sections in a standard petrographic microscope. Because PDFs are so widely used as impact evidence, their correct identification is crucial. However, using light microscopy PDFs can look very similar to nonimpactrelated (sub)planar microstructures, such as healed fractures or tectonic deformation lamellae (e.g., Christie and Raleigh 1959;Christie and Ardell 1974;Drury 1993;Vernooij and Langenhorst 2005). Therefore, transmission electron microscopy (TEM) observations on the submicron scale structure of the lamellae are often necessary to distinguish between shock and nonshock-related structures and to confirm the shock origin of planar microstructures in quartz (e.g., St€ offler and Langenhorst 1994;Grieve et al. 1996;French and Koeberl 2010). However, TEM facilities are expensive and not readily available to many researchers. In addition, TEM studies involve difficult and timeconsuming sample preparation and often require extensive training. cathodoluminescence (CL) imaging in the scanning electron microscope (SEM) is a reliable method to distinguish between shock-related and tectonically formed planar microstructures in quartz. Although Boggs et al. (2001) concluded that PDFs show no CL emission at all, Hamers and Drury (2011) demonstrated that composite color SEM-CL imaging shows two main types of CL behavior in PDFs: red luminescence and nonluminescence. Also Gucsik et al. (2015) observed nonluminescent PDFs in shocked quartz grains from the Ries crater. The reason for the different CL emissions, however, remained unclear.
In this paper, we aim to clarify the differences in CL emissions from PDFs in quartz as observed in Hamers and Drury (2011). In order to identify the relationship between CL emission and the microstructure of PDFs in quartz, we performed TEM analysis on both red and nonluminescent PDFs. To control the exact location of the TEM samples and to enable a direct, often one-to-one comparison of PDFs in SEM-CL and TEM images, the TEM samples were prepared using a focused ion beam (FIB) (Wirth 2004(Wirth , 2009). In addition, CL spectroscopic measurements of the emission of red luminescent PDFs were done to provide an indication of the defects that are responsible for the red CL emission. Furthermore, we analyze the effect on CL emission of prolonged exposure of the quartz grains to the electron beam and we show results suggesting that a high-pressure phase, possibly stishovite, can be detected in SEM-CL images.
Directly comparing SEM-CL images with results from TEM techniques that are universally accepted as effective tools for PDF identification and analysis (see reviews by St€ offler and Langenhorst 1994;Grieve et al. 1996;French and Koeberl 2010), with the addition of CL spectroscopy, will help to determine the source of the CL emission and thereby validate the SEM-CL method to identify and characterize PDFs in quartz.

Planar Deformation Features in Quartz
PDFs consist of thin (<1 lm), parallel, closely spaced (typically <2 lm apart) lamellae. Usually multiple (normally~2-4) sets of PDFs develop in the same grain (von Engelhardt and Bertsch 1969;St€ offler and Langenhorst 1994;Grieve et al. 1996;French and Koeberl 2010). Although PDFs are generally described as strictly straight and planar, some bending or slight curvature has been reported (von Engelhardt and Bertsch 1969;St€ offler and Langenhorst 1994;. PDFs can be penetrative through a whole grain or occur in parts of a grain, but do not cross grain boundaries, cracks, faults, or fractures that were present before PDF formation. PDFs have a crystallographically determined orientation parallel to low-index rational planes (most frequently observed are the basal c(0001) and x{10 12} and p{10 13} type planes) (St€ offler and Langenhorst 1994;Ferri ere et al. 2009). Unaltered or "fresh" PDFs are thin lamellae of amorphous silica, formed under high pressure during shock compression. Basal PDFs, parallel to c(0001), are not amorphous, but consist of mechanical Brazil twins. In shock experiments on quartz, Langenhorst (1994) observed that at relatively low peak shock pressures (<25 GPa), rhombohedral PDFs are extremely narrow (~30 nm), but around 25 GPa, thickness increases to >200 nm and PDFs can become slightly curved and show thickness variations along their length.
During postshock high temperature or metamorphic events, amorphous PDFs can easily recrystallize. Healed PDFs are recognized in TEM as planes with a very high dislocation density (e.g., Goltrant et al. 1991Goltrant et al. , 1992Trepmann and Spray 2006). In light microscopy, they are observed as traces of fluid inclusions along the original PDF plane (known as "decorated PDFs"), of which the crystallographic orientation can still be measured in the universal stage (e.g., Ferri ere et al. 2009). Because amorphous PDFs only occur in relatively fresh, unaltered impact material, it is useful to be able to easily detect the presence of amorphous PDFs in shocked quartz. Furthermore, the presence of amorphous material can help identify microlamellae in quartz as PDFs, as microstructures that are often misinterpreted as PDFs (e.g., tectonic deformation lamellae) do not contain amorphous material. At present, the most widely applied method to detect amorphous PDFs is TEM analysis (e.g., Goltrant et al. 1991Goltrant et al. , 1992Langenhorst 1994Langenhorst , 1996Trepmann and Spray 2006).

Cathodoluminescence
In a material that is scanned by the electron beam in the SEM, electrons in the sample are excited to a higher band level by the incoming energy of the beam. When the excited electrons fall back to their original energy state, photons are emitted. This phenomenon is known as cathodoluminescence. The wavelength spectrum and intensity of photon emission depend on the material and crystallographic defects and on the presence of trace elements or fluids in the sample. In quartz, CL emission depends on many factors such as the degree of ordering in the crystal lattice; vacant lattice sites; impurities such as Fe 3+ , Al 3+ , and H + ; water content; or damage resulting from radiation, strain, or shock (Marshall 1988;Ramseyer and Mullis 1990;Perny et al. 1992). Common emission bands in crystalline a-quartz are around 340 nm (Al and Li impurities), 420 nm (intrinsic emission), and 620-650 nm (nonbridging oxygen hole center, NBOHC, O 3 Si-OÁ, i.e., a dangling oxygen bond, that is bonded to a silicon atom attached to three other oxygen atoms) (Stevens Kalceff and Phillips 1995;Stevens Kalceff et al. 2000;G€ otze et al. 2001;Stevens Kalceff 2009), but many more emission bands are described in the literature. The relative intensities of the dominant emission bands in the visible light range (380-700 nm, blue to red) determine the CL color that is perceived by the eye (G€ otze et al. 2001).
Shocked quartz is generally blue luminescent, with an additional emission band at 630-650 nm Ramseyer and Mullis 2000;; G€ otte 2009). Boggs et al. (2001) first demonstrated the potential of SEM-CL to image PDFs in quartz and concluded that PDFs are nonluminescent in grayscale SEM-CL images. Hamers and Drury (2011), however, using a CL detector with a broader detection range and color filters, showed that in composite color SEM-CL images, PDFs can have different CL characteristics. Some are indeed nonluminescent, but many are red luminescent and also "intermediate" cases were observed, with both red and nonluminescent (parts of) PDFs in the same grain.

MATERIALS AND METHODS
The quartz grains described in this study are mostly naturally shocked grains from different impact structures (El'gygytgyn, Ries, Popigai, Rochechouart, and Vredefort). The same shocked quartz samples were studied here as were described by Hamers and Drury (2011) and in addition SEM-CL images were recorded of a granitic sample from the Vredefort structure in South Africa, quartz grains in a reworked suevite sample from the El'gygytgyn crater in Russia (ICDP drill core, Raschke et al. 2013), an experimentally shocked (23.5 GPa) Hospital Hill quartzite, and an experimentally shocked (24 GPa) single crystal. The experimentally shocked samples were previously studied and described by H€ orz (1986a, 1986b) (Hospital Hill quartzite) and Fritz et al. (2011) (single crystal). In total, 12 TEM foils were studied. Table 1 lists the new samples and the TEM foils, with their locations, main characteristics, and the observed microstructures.
Composite color SEM-CL images were recorded at the Electron Microscopy Utrecht lab (EMU), the Netherlands, using a FEI Nova Nanolab 600 dual beam SEM with focused ion beam (FIB) and Gatan PanaCL detector with a detection range of 185-850 nm. Acceleration voltages of 5-10 kV were used, with beam currents of 1.6-6.3 nA, at room temperature. The procedures are described in detail in Hamers and Drury (2011). The CL signal is emitted from the whole interaction volume. Monte Carlo simulations using Casino v2.42 software (Drouin et al. 2007) indicate that an electron beam with an energy of 5-10 kV penetrates into the sample to a maximum depth of~0.1-3 lm. This gives an indication of the imaging resolution under the conditions used in this study.
TEM foils with a thickness of~100-200 nm were prepared in the same FIB-SEM instrument, using a standard in situ lift-out procedure (Wirth 2004(Wirth , 2009, with an acceleration voltage of 30 kV and beam currents ranging from 20 nA down to 0.3-0.5 nA for the final polishing steps. The FIB-SEM preparation of the TEM foils allows a direct, often one-to-one correspondence of PDFs in TEM and SEM-CL images. TEM analysis was carried out on an FEI Tecnai 20 FEG instrument, also at EMU, operated at 200 kV, with double tilt holder. Bright-field (BF) TEM imaging was used to show the microstructure in the samples. Diffuse dark-field (DDF) TEM images showed the presence of amorphous material. In DDF images, amorphous areas are bright, while crystalline material is dark. In addition, tilting the sample in BF mode showed whether there was any diffraction contrast associated with crystallinity.
CL spectroscopic measurements were performed on a polished, carbon-coated thin section of sample RI-37 (Ries, see Table 1) with red luminescent PDFs, on a Zeiss EVO 50 instrument with Gatan MonoCL 3 system, at the Institute of Geological Sciences in Bern, Switzerland. Spectra of the red luminescent PDFs and the host quartz were recorded with a PIXIS 100 CCD camera (Princeton Instruments). The CCD has nearly constant detector efficiency for the visible light wavelength range. Grating of the spectroscopic system was set to 300 lines/mm with a slit size of 2 mm. Spectra were recorded in a 600 nm range and with a central wavelength of 500 nm, resulting in simultaneous measurement of the 200-800 nm spectra. Acceleration voltages of 5-8 kV were used, with beam currents of 1.4-3 nA at room temperature. PDFs are too narrow to measure directly by spot measurements, because the interaction volume of the electron beam with the quartz is too large to only sample the PDF. Inevitably some of the surrounding quartz is sampled by the beam. To overcome this problem, spectra were measured from areas with a high density of PDFs and areas without PDFs in the same grain. The ratio of the intensity of the main peaks between PDFs and host grain was then determined. Measurements were done in areas of the sample that had not been scanned by the electron beam before, with a measurement time of 30 seconds. Because only peak ratios were considered and not absolute values, no correction for detector response was carried out. GPa (see also H€ orz 1986a, 1986b) Figure 1 shows CL (a and d) and TEM (b, c, e, and f) images of two grains with nonluminescent PDFs, in a suevite from the Ries crater in Germany (a, b, c) and in a quartzite experimentally shocked to 23.5 GPa (d, e, f). In bright-field TEM images (Figs. 1b and 1e), the PDFs are always a uniform gray, irrespective of tilt angle. In corresponding diffuse dark-field images of the same area (Figs. 1c and 1f), PDFs are brighter than the surrounding material. Both bright-field and diffuse dark-field characteristics of nonluminescent PDFs therefore indicate that they are amorphous. The corresponding diffraction patterns (insets in Figs. 1b and 1e) show diffuse scattering intensity (similar to, e.g., diffraction patterns shown by Cordier and Gratz [1995] and Langenhorst [1996]) and are consistent with the presence of amorphous material in the sample. The diffraction patterns from nonluminescent PDFs show considerably more diffuse intensity than those from red luminescent PDFs (e.g., Fig. 2b). In addition to diffuse intensity, a broad, diffuse halo is present in the diffraction pattern as shown in Fig. 1e, just inside the 10 10 diffraction maxima (4.26 A), which is consistent with d-spacings for fused silica reported in the literature (4.12

Nonluminescent PDFs
A, e.g., Lana and Seddon 1998). Diffuse halos can be hard to distinguish from inelastic scattering in a diffraction pattern, which can occur in thick foils. However, the foils used in this study have a thickness of 100-200 nm, which is not particularly thick for quartz when using 200 kV electrons. Furthermore, there is no evidence of Kikuchi bands in the diffraction patterns, which are normally observed in thick sections.
The quartz in between the PDFs is diffracting and therefore crystalline. Slight streaking of the (outer) diffraction spots (insets in Figs. 1b and 1e) points to a small misorientation between the crystalline domains. Dislocations and other defects were not observed in the crystalline quartz.

Red Luminescent PDFs
Figures 2-4 show that red luminescent PDFs can be either healed structures (Figs. 2 and 3) or basal Brazil twins (Fig. 4).  show that the red luminescent PDFs in this sample are planes of high dislocation density or arrays of dislocations, similar to microstructures observed in healed PDFs in earlier studies (e.g., Goltrant et al. 1991;Trepmann and Spray 2006). In grains like the one shown in Figs. 2a-c, with few sets and a relatively low density of PDFs, the crystalline host quartz is a single crystal and not microcrystalline; there are no misorientations in the grain (no streaking of diffraction spots occurs, see inset in Fig. 2b). Aside from in the PDF planes, only a few dislocations are present in the grain. In Figs. 2b and 2c, a Dauphin e twin boundary can be traced from the upper right corner along a PDF, then crossing to another PDF and following this PDF's orientation. In rare cases, such as the grain from the Ries crater in Fig. 2d, SEM-CL images show that red luminescent PDFs consist of small sublamellae (black arrow in Fig. 2d) in a configuration similar to the dislocation arrays observed in TEM.
In a grain from the Vredefort Dome, red luminescent PDFs (Fig. 3a) consist of basal Brazil twins (shown edge-on in bright-field TEM in Fig. 3b and tilted in Fig. 3c). Some dislocations are present in this grain and parts of the Brazil twin boundaries are decorated with dislocations. However, there are also segments of the twin boundaries that are free of dislocations. It therefore appears that the red CL emission comes from the twin boundaries themselves and not (only) from dislocations, as opposed to the red luminescent PDFs shown in Fig. 2.

Red and Nonluminescent PDFs
Besides the red and nonluminescent endmembers of CL behavior that PDFs exhibit in SEM-CL, an  intermediate type occurs: grains with both red and nonluminescent PDFs, and PDFs that have both red and nonluminescent parts (see also Hamers and Drury 2011). SEM-CL and TEM images of this type of grain are shown in Fig. 4. This grain contains both amorphous PDFs (light in the diffuse dark-field image in Fig. 4b) and PDFs that are planes of high defect density (Fig. 4c). Again, the presence of bright diffuse scattering intensity in the diffraction pattern is consistent with the presence of amorphous material in the sample.

CL Spectra of Red Luminescent PDFs
CL spectra were measured in the SEM in five grains with red luminescent PDFs in a granite fragment from the Ries crater (Fig. 5). The main peaks in the spectra in both PDF and host quartz are at 410 nm and 650 nm, which are commonly occurring peaks in quartz (see the Cathodoluminescence section). After a measurement time of 30 seconds, the intensity ratio of the 650 nm and 410 nm peaks is higher in the PDFs than in the host quartz (Fig. 5d). This indicates that the 1 µm CL spectra of red luminescent PDFs have a stronger 650 nm peak than the surrounding quartz. This is observed in all studied grains. On average, with a measurement time of 30 s, the 650/410 nm peak ratio is 1.7 for PDFs versus 1.3 for the host quartz (Fig. 6); that is, the 650/410 nm peak ratio is 24% higher in PDFs.
The intensity of the red 650 nm peak in quartz increases with increasing irradiation time in the SEM, while the blue 410 nm peak slightly decreases. The intensity increase of the 650 nm emission is quicker in the PDFs than in the host quartz and therefore PDFs are red in CL images. This was also noted by Hamers and Drury (2011), who observed that the red luminescent PDFs become brighter in composite color CL images after scanning the same area several times. This effect can furthermore be observed in Figs. 5a (first scan) and 5b (after several scans): the (panchromatic) CL emission from PDFs is lower than that of the host quartz in the first scan, but brighter in PDFs than in the rest of the grain after several slow scans. The effect can be shown with point measurement times around 30 s, but when longer measurements are taken, the difference between PDFs and host quartz decreases and eventually disappears.
Unfortunately, no peak ratios could be determined for samples with basal Brazil twins (Vredefort), because no areas could be found with a high enough twin density for reliable measurements.
Changes in CL Emission after Prolonged Exposure to the Electron Beam Figure 7 illustrates the effect of irradiation with the electron beam on the CL emission of quartz. In Figs. 7a and 7b, red squares are areas that have been scanned previously at higher magnification. Prolonged exposure to the electron beam strongly increases the red CL emission of the whole previously scanned area. In Fig. 7b, the central red square is made up of red spots that show where spot measurements were taken for an electron backscattered diffraction (EBSD) map (not discussed in this paper) recorded before scanning the CL image. Figure 7c is a higher magnification scan of the red square on the right in Fig. 7b. In the center of measurements. The CL emission of the PDFs, which are mostly red luminescent in the rest of the image, has disappeared in the area that has suffered long exposure to the beam during point mapping: the PDFs have become nonluminescent. This effect is also observed in Fig. 7a in a small rectangular area in the top right of the image, indicated by the black arrow. Quenching of the CL emission was not observed as a result of scanning images (even after repeated slow scanning). The CL emission quenched only in sample areas on which the beam was stationary for a relatively long time (~0.2 s on each point in this case, whereas in scanned images, the dwell time was four orders of magnitude lower, typically 30-60 ls per pixel) and with higher voltages and beam currents (20 kV and 2.4 nA for these two examples) than those used for CL imaging. Figure 8 shows backscattered electron (BSE), secondary electron (SE), and composite color SEM-CL images of a quartz grain that was shocked experimentally to 24 GPa. Most of the PDFs in Fig. 8a are darker (less dense) than the surrounding quartz, as is commonly observed in BSE images of shocked quartz grains (e.g., St€ ahle et al. 2008;Ebert et al. 2013), because PDFs are filled with amorphous SiO 2 that is less dense than crystalline quartz. However, some PDFs in Fig. 8a are filled with a bright material. With EDX measurements of different parts of the grain, both PDFs and host quartz, only the presence of SiO 2 was detected (one spectrum is shown in Fig. 8b), indicating that any chemical differences must be very small and are unlikely to be big enough to account for the strong contrast observed in the BSE image. The BSE contrast in Fig. 8a is therefore likely to be from density differences and unlikely to be from atomic number contrast. Thus, darker shades of gray in the BSE image mean lowdensity material, while lighter shades of gray indicate higher densities. Figures 8c and 8d, respectively, show an SE and composite color SEM-CL image of the same grain as is shown in Fig. 8a. Due to an extra polish that was applied to the sample between the SE/CL imaging and the BSE imaging, exact correlation of the PDFs between BSE and SE/CL is not possible, because a thin layer (a few microns) of the sample surface was removed. However, the PDFs in the SE image show the same characteristics as in the BSE image: some (parts of) PDFs are brighter. Although SE images are commonly used to show topography, also density or phase contrast can be imaged when acceleration voltages above~5 kV are used, mainly because backscattered electrons excite some secondary electrons on their trajectories out of the surface layer (e.g., Reimer and Hawkes 1998). The density contrast might be enhanced by our use of an Everhart-Thornley detector (Everhart and Thornley 1960), which inevitably detects some backscattered electrons in addition to the secondary electrons. Furthermore, the higher density material could be more resistant to polishing and therefore induce slight topography on the sample surface, enhancing the visibility of the brighter material in the SE image even more.

PDFs Filled with a Blue Luminescent Material
When comparing the SE image to the composite color SEM-CL image in Fig. 8d, the brighter SE regions clearly show blue luminescence, in contrast to the majority of PDFs, which are nonluminescent.

Cathodoluminescence of PDFs
Our results show that the SEM-CL behavior of PDFs in quartz can be directly related to the microstructure of the PDFs as observed in the TEM.

Red Luminescent PDFs
The one-to-one correspondence of SEM-CL and TEM images (Fig. 2) confirms that red luminescence occurs in healed PDFs and is related to the high dislocation density in the PDF plane. The dominant emission peak in red luminescent PDFs is at 650 nm (Figs. 5 and 6), and is therefore related to the nonbridging oxygen hole center, or NBOHC (see the Cathodoluminescence section) (Stevens Kalceff and Phillips 1995;Stevens Kalceff et al. 2000;G€ otze et al. 2001).
The radiolytic beam damage processes is the main source for the red CL emission. This production of NBOHC defects accompanied by red CL emission is a wellknown phenomenon and also takes place in undamaged crystalline quartz (e.g., Stevens Kalceff and Phillips 1995;G€ otze et al. 2001;Stevens Kalceff 2009)-as can also be observed in Fig. 7, where previously scanned areas even without PDFs show a strong red CL emission. However, PDFs are quicker to show effects of irradiation damage than the surrounding quartz, and to show an increasingly bright red luminescence. In TEM, localized electron beam damage, mainly by radiolytic processes, is a commonly observed phenomenon at dislocations, fluid inclusions, and twin boundaries, where continued beam damage finally results in complete (local) amorphization (McLaren et al. 1970;Comer 1972;Cherns et al. 1980;Carter and Kohlstedt 1981). The combination of the red CL emission at 650 nm with strongly increasing intensity in PDFs, the high dislocation density along healed PDFs in TEM, and the known phenomenon of localized beam damage in TEM at defects strongly suggest that preferential electron beam damage at dislocations produces NBOHC defects by breaking Si-O bonds, forming the source of the red luminescence in healed PDFs. In rare cases, even individual dislocations within the healed PDFs are apparently detected in SEM-CL images, as is shown by the PDFs in the upper right corner of Fig. 2d. These PDFs consist of arrays of tiny red luminescent "sublamellae" that look similar to the arrays of dislocations observed in healed PDFs in TEM (e.g., Figs. 2b and 2c, and figures in Goltrant et al. [1991] and Trepmann and Spray [2006]). The presence of OH bonds may also play a role in the production of NBOHC defects and the related red CL emission. Healed PDFs are thought to form by water-assisted crystallization of the initially amorphous PDFs under high postshock temperatures or metamorphic events (Grieve et al. 1996), so obviously water must be present in healed PDFs. This is supported by the fluid inclusions that usually decorate healed PDFs (Goltrant et al. 1992;Grieve et al. 1996). As OH groups may preferentially bond to nonbridging atoms in dislocation cores (e.g., Trepied and Doukhan  Fig. 7. Composite color images of grains in a granite fragment from the Ries crater. a) Low magnification image showing red squares where higher magnification images have been scanned before at a slow scanning rate. A small rectangle has become completely nonluminescent after intense spot irradiation due to point measurements for an EBSD map (top left, black arrow). b) Low magnification image showing red squares where higher magnification images have been scanned before at a slow scanning rate (right) and where an EBSD map was recorded using point measurements (center). c) Higher magnification image of the red square on the right in (c). The square in the center shows changes in CL emission of the part of the sample where point measurements have been taken for EBSD mapping. The PDFs in the square, which are mostly red in the rest of the image, have become nonluminescent.
1978; Heggie 1992), it is likely that a relatively high concentration of OH groups are present in healed PDFs and that these (at least initially) contribute to the red luminescence. To check the presence of OH groups in healed PDFs, micro-Raman spectroscopy could be applied in a future study.
Not only healed PDFs but also Brazil twins are red luminescent in SEM-CL images (Fig. 3). If dislocations are present along the twin plane, they could be a source of the red luminescence, but more likely the red CL emission comes also from the twin plane itself, because the dislocation density in the Brazil twin plane is usually much lower than in healed PDFs and large parts of the twin plane are free of dislocations (Fig. 3). Furthermore, McLaren and Phakey (1965) observed very quick preferential damage along Brazil twin boundaries in amethyst in TEM, similar to the damage occurring at dislocations. The necessary precursors for the NBOHC may be strained bonds in the locally distorted tetrahedral structure in the twin boundary, or alternatively OH groups, peroxy linkages, or impurities along the twin boundary.  Figure 1 shows that nonluminescent PDFs contain amorphous material. Although the 650 nm CL emission is a commonly observed peak in fused (low pressure, high temperature produced) silica glass, i.e., amorphous silica (Stevens Kalceff et al. 2000;Fitting et al. 2001), in the amorphous material in nonluminescent PDFs, such as the grain from the Ries crater and the experimentally shocked grain in Fig. 1, the structure is apparently so damaged and disordered that no CL emission is possible. Due to the extreme pressures and temperatures involved in impact processes, the crystal structure in PDFs formed at high pressure could be completely destroyed, preventing any photon emission under electron bombardment. The conclusion that the material in nonluminescent PDFs is amorphous is supported by our observations of the effect of beam damage on the CL emission of quartz. Figure 7 shows that prolonged intense electron irradiation, with the beam in spot mode, first induces an increase in red CL emission, but finally results in nonluminescence. This suggests that ongoing beam damage eliminates all luminescence centers. As damage processes in TEM and SEM are largely the same and beam damage in TEM finally results in local amorphization, it is likely that the same happens in SEM and that localized amorphization is observed as localized quenching of CL emission.

Nonluminescent PDFs
An alternative explanation for the nonluminescence of amorphous PDFs might be variations in trace element concentrations, for example iron content. Although interstitial Fe 3+ in (crystalline) quartz is mentioned as the source of a red luminescence peak around 700-750 nm (G€ otze 2009;Stevens Kalceff 2009), it has also been suggested that the presence of iron can quench other CL emissions (Haddad et al. 2006). Chemical analysis of experimentally shocked quartz grains suggests that grains with a high density of PDFs contain an increased amount of metals from the iron meteorite projectile that was involved (Ebert et al. 2013). If meteoritic iron, or iron from another source, is in some way distributed preferentially into the PDFs, it could quench the luminescence. In this context, it would be interesting in a future study to map the trace element distribution in shocked quartz grains with nonluminescent amorphous PDFs.
The breaking of Si-O bonds due to electron irradiation is in fact very local amorphization in the direct vicinity of the defect. Because defects in SiO 2 disturb the short-range order, the defect structure and most luminescence centers in crystalline and amorphous SiO 2 are similar (Sigel and Marrone 1981;Stevens Kalceff and Phillips 1995;Stevens Kalceff 2013). It has even been suggested that the NBOHC and most other defects in quartz are associated with locally amorphous parts of the crystal structure (Stevens Kalceff 2009, 2013. Furthermore, as mentioned earlier, the 650 nm peak is a commonly observed CL emission in fused SiO 2 glass (Stevens Kalceff et al. 2000;Fitting et al. 2001;Stevens Kalceff 2013). It is therefore theoretically possible that some amorphous PDFs are red luminescent. However, red luminescent amorphous PDFs were not observed in our samples. Apparently, the structure or trace element composition of amorphous PDFs is not the same as that of most types of fused silica glass.
It is worth noting, as was already reported by Hamers and Drury (2011), that only highly shocked grains were observed with exclusively nonluminescent PDFs. In these grains, the PDFs are always thick and slightly wavy and irregularly shaped: the characteristics of PDFs formed at high shock pressures > 25 GPa (Langenhorst 1994). No grains were observed with only nonluminescent PDFs showing characteristics of a lower shock stage (i.e., fewer sets, more widely spaced and thin and straight PDFs). Even in relatively recent structures, such as the 3.6 Ma El'gygytgyn crater (images not shown), grains with characteristics of lower shock pressure and with only nonluminescent PDFs were not observed. The most probable explanation for the apparent lack of moderately shocked quartz grains with exclusively nonluminescent PDFs is that, even in relatively recent impact structures, a significant proportion of the amorphous material has crystallized and many (parts of) PDFs are healed. Under similar temperature conditions, thinner PDFs, formed at lower shock pressures, will likely heal more easily than wider PDFs, which formed at much higher shock pressures.
Several other explanations for the lack of moderately shocked grains with nonluminescent PDFs might be considered: A) amorphous PDFs in grains shocked to lower pressures are too narrow to be detected as nonluminescent in CL images, but instead defects or damage zones along the PDFs give rise to a red luminescent signal; B) unaltered PDFs at lower shock stages are not (completely) amorphous; C) lower shock amorphous PDFs contain amorphous material with a different structure, that is red instead of nonluminescent -in this case, these PDFs would be indistinguishable from healed ones in the CL images in this study. Scenario A is unlikely, as thin and straight, but nonluminescent and amorphous PDFs were observed in the grain in Fig. 4. No direct evidence for scenarios B or C can be found in our data, but from the literature and our results, it is clear that amorphous SiO 2 can exhibit different types of CL behavior. TEM and SEM-CL observations on very recent or, preferably, experimentally shocked quartz should be able to solve this issue.
Besides TEM analysis, also (micro-)Raman spectroscopy could help in determining the presence of amorphous material in PDFs and the structural configuration of the crystal lattice in shocked quartz grains. Gucsik et al. (2011) and Pittarello et al. (2015) showed promising results using this technique on shocked quartz from the Ries crater and the El'Gygytgyn impact structure, respectively. A future combination of CL and Raman analyses of shocked quartz (e.g., similar to the methods used by Okumura et al. [2009] on ballen silica) might provide more information on the characteristics of the amorphous material in nonluminescent PDFs.

Blue Luminescent High-Pressure Phase
In the experimentally shocked quartz grain in Fig. 8, PDFs were observed filled with a material that is bright in BSE images and blue luminescent in CL images. Because the material is brighter than the host quartz in BSE images, it must have a higher density, but with EDX spectra only SiO 2 was detected in both the bright material and the host grain. Therefore, the bright material can be interpreted as a high-pressure SiO 2 phase. Two high-pressure polymorphs of quartz are frequently observed in impact environments: coesite and stishovite (e.g., St€ offler and Langenhorst 1994;St€ ahle et al. 2008). Stishovite is estimated to occur in quartz shocked to pressures between~12 and 45 GPa, while coesite only appears at pressures above~30 GPa (St€ offler and Langenhorst 1994). The grain in Fig. 8 was shocked to 24 GPa, which is in the pressure range in which stishovite can form. Furthermore, the highdensity phase has the same geometry and appearance as the stishovite-filled PDFs shown by St€ ahle et al. (2008) in their fig. 2. The blue CL is also consistent with stishovite: Fitting et al. (2001) showed that the CL emission of stishovite from Barringer Crater (Arizona, USA) at room temperature has a broad band that is a combination of peaks around 420, 460, and 500 nm. Trukhin et al. (2003Trukhin et al. ( , 2014 also observed a blue luminescence band in stishovite. Because of the shock pressure, geometry, and CL emission color, the highpressure phase observed in the grain in Fig. 8 can be interpreted as stishovite. Alternatively, as it is unknown whether this high-density phase is crystalline or not, it could be a high-pressure amorphous form of SiO 2 . The observation of the blue luminescent highdensity phase in Fig. 8 illustrates that, like BSE images, composite color SEM-CL imaging can be used as an indication for the presence of high-pressure phases in shocked quartz. However, care should be taken with a definitive interpretation, because remains of polishing material or dust particles on the sample surface can also be blue luminescent (see the bright blue spots present in most color CL images in this paper and fig. 4 in Hamers and Drury 2011), as might possible secondary minerals in altered rocks. Chemical analysis of the composition of the phase and TEM or EBSD observations, providing information on crystal structure, would be required to confirm the nature of a potential high-pressure form of SiO 2 .
Use of SEM-CL Imaging as Evidence for Impact Structures Although above results show that SEM-CL imaging is a very useful technique to identify and characterize PDFs in quartz, it does not provide a quick and easy "short-cut" to prove the shock origin of planar microstructures in quartz. Also deformation lamellae can be imaged with this technique and even though in high quality images PDFs can be distinguished from nonimpact-related planar or subplanar microstructures (Hamers and Drury 2011), this is not a quantitative method. To obtain the best possible SEM-CL images, good sample preparation and a carbon coat as thin as possible are essential, in order to obtain the strongest signal (see Frelinger et al. [2015] for a detailed discussion on recording high-quality SEM-CL images). Higher beam currents increase signal strength, but reduce spatial resolution. Slower scanning times strongly increase image quality, but also beam damage effects.
If a researcher aims to identify shocked quartz to provide evidence for a new impact structure, it remains important to start with proper light microscopy studies to identify potentially shocked grains, study as many grains as possible, and to use multiple methods to cross-check any evidence. Additional methods that could be used are universal stage measurements of PDF orientation distributions (made much easier by recent work of Ferri ere et al. 2009;Huber et al. 2011;and Losiak et al. 2016) and TEM analysis (e.g., Goltrant et al. 1991Goltrant et al. , 1992Langenhorst 1994Langenhorst , 1996Trepmann and Spray 2006). Also Raman spectroscopy (e.g., Gucsik et al. 2011), EBSD, and other SEM techniques, and chemical analyses like EDX and WDS can provide useful information.

CONCLUSIONS
The combination of color SEM-CL imaging and TEM analysis of shocked quartz grains confirms the great potential of color SEM-CL imaging as a method to identify PDFs in quartz, and also illustrates the complexities that are involved in interpreting CL emissions, especially in images. A very useful result of this study for crater research is that color SEM-CL imaging can distinguish between fresh (amorphous) and healed PDFs or basal Brazil twins: nonluminescent PDFs are amorphous, while healed PDFs and basal Brazil twins are red luminescent. TEM and CL spectrometry confirm that the dominant peak in the emission from red luminescent PDFs is around 650 nm, related to the NBOHC, and is most likely the result of preferential beam damage at dislocations and twin boundaries. A high-pressure phase, possibly stishovite, can be detected in color SEM-CL images as blue luminescent parts of PDFs. Although not all details of the CL emission of PDFs can be explained at this time, SEM-CL imaging is a great method to visualize subtle differences in shock microstructures in quartz. For more quantitative information, additional analysis can be done on the same samples using other techniques, such as TEM, EBSD, or light microscopy with universal stage for crystallographic orientations of grains or PDFs, Raman spectroscopy for extra information on amorphous material and crystal order, and EDX/WDS mapping or CL spectroscopy for defect and trace element distributions.