Classification of CM chondrite breccias—Implications for the evaluation of samples from the OSIRIS‐REx and Hayabusa 2 missions

CM chondrites are complex impact (mostly regolith) breccias, in which lithic clasts show various degrees of aqueous alteration. Here, we investigated the degree of alteration of individual clasts within 19 different CM chondrites and CM‐like clasts in three achondrites by chemical analysis of the tochilinite‐cronstedtite‐intergrowths (TCIs; formerly named “poorly characterized phases”). To identify TCIs in various chondritic lithologies, we used backscattered electron (BSE) overview images of polished thin sections, after which appropriate samples underwent electron microprobe measurements. Thus, 75 lithic clasts were classified. In general, the excellent work and specific criteria of Rubin et al. (2007) were used and considered to classify CM breccias in a similar way as ordinary chondrite breccias (e.g., CM2.2‐2.7). In BSE images, TCIs in strongly altered fragments in CM chondrites (CM2.0‐CM2.2) appear dark grayish and show a low contrast to the surrounding material (typically clastic matrix), and can be distinguished from TCIs in moderately (CM2.4‐CM2.6) or less altered fragments (CM2.7‐CM2.9); the latter are bright and have high contrast to the surroundings. We found that an accurate subclassification can be obtained by considering only the “FeO”/SiO2 ratio of the TCI chemistry. One could also consider the TCIs’ S/SiO2 ratio and the metal abundance, but these were not used for classification due to several disadvantages. Most of the CM chondrites are finds that have suffered terrestrial weathering in hot and cold deserts. Thus, the observed abundance of metal is susceptible to weathering and may not be a reliable indicator of subtype classification. This study proposes an extended classification scheme based on Rubin’s scale from subtypes CM2.0‐CM2.9 that takes the brecciation into account and includes the minimum to maximum degree of alteration of individual clasts. The range of aqueous alteration in CM chondrites and small spatial scale of mixing of clasts with different alteration histories will be important for interpreting returned samples from the OSIRIS‐REx and Hayabusa 2 missions in the future.

One of the first classification schemes using the degree of alteration was developed by McSween (1979b), and it consisted of the three categories (1) partially altered, (2) altered, and (3) highly altered. Rubin et al. (2007) proposed an improved alteration scheme that defined CM2 subtypes by classification of the main lithology based on the abundance and size of the TCIs (former PCPs), the "FeO"/SiO 2 ratio of the TCIs, and the abundances of Ca-carbonates and metals. Since this classification scheme was established, it has been used frequently. Yet, while it does consider wellestablished, detailed aspects and is easy to understand, it has some major problems: 1. It only considers the degree of aqueous alteration of the major lithology of a thin section. Considering that the exposed specimens used for analysis are often small, 1 inch thin sections, they are not representative in many cases. 2. One of the key observations needed to use this classification scheme is the metal abundance. Unfortunately, because most of the CM chondrites are finds that have suffered weathering in hot and cold deserts (only 18 of 640 CM chondrites are fresh falls; Meteoritical Bulletin Database, December 2019), their original metal abundances cannot be determined. Thus, the observed abundance of metal is susceptible to terrestrial weathering and may not be a reliable indicator of subtype classification. 3. Besides the metal abundance, this classification scheme also incorporates both the calculated S/SiO 2 and "FeO"/SiO 2 ratios for determining the CMsubtype classification. However, the S/SiO 2 ratio may significantly vary in a sample because of the difference in the precursor materials of the TCI alteration product (sulfide versus metal). 4. Most CM chondrites are heavily brecciated (as will be shown below; Fig. 1), and the most abundant lithology can vary from thin section to thin section because the small sections are not representative (see point A above and the following paragraph).
Previous reports have indicated that CM breccias are highly heterogeneous (e.g., Metzler et al. 1992;Bischoff et al. 2006Bischoff et al. , 2017Lindgren et al. 2013;Zolensky et al. 2017;Lentfort et al. 2019). For instance, Bischoff et al. (2017) showed that in a single meteorite, the main lithology can vary from one sample to another. Nogoya is an excellent example indicating the heterogeneity of CM breccias in different thin sections; Rubin et al (2007) classified Nogoya as a CM 2.2, whereas the thin section investigated by Bischoff et al. (2017; which is also part of this study) is dominated by a CM2.5 fragment (~70% of the thin section; see Fig. 2). Therefore, Bischoff et al. (2017) proposed an extended classification scheme that gives a range of alteration subtypes (e.g., CM2.2-2.5) for a specific CM chondrite breccia, similar to the classification system used for ordinary chondrites breccias (e.g., H3-5; LL4-6).
During this study, 19 different CM chondrites and three achondrites containing CM-like clasts were examined. They were examined in order (1) to identify the possible brecciation of the sample and (2) to determine the degree of aqueous alteration of individual fragments. TCIs were searched for in 32 polished sections with a scanning electron microscope (SEM) and the help of overview images. Subsequently, 27 thin sections with TCI-bearing fragments/areas were analyzed with the microprobe in order to detect clasts with the highest and lowest alteration types of each sample based on the "FeO"/SiO 2 ratio. For samples with few (no) clearly visible lithologies (Murchison, Santa Cruz, Y-791198, and Maribo), the compositions of the TCIs of the main lithology (bulk rock) were obtained.
Our classification procedure for individual clasts mainly considers the "FeO"/SiO 2 ratios mentioned by Rubin et al. (2007) to make distinctions between several CM subtypes, as we have done in earlier studies (Bischoff et al. 2017;Lentfort et al. 2019). Thus, we classify CM chondrites based on the constituents of different clasts, as is done for ordinary chondrite breccias (e.g., Bischoff et al. 2006Bischoff et al. , 2018bMetzler et al. 2011). The classification scheme of Rubin et al. (2007) is limited to subtypes not higher than 2.6, though subsequent work suggested that the Paris CM chondrite has subtype of at least 2.7 (Rubin 2015). In this study, we use ratios of "FeO"/SiO 2 to define criteria for CM 2.7, 2.8, and 2.9 subtypes to help compare weakly altered CM chondrites. Finally, we compare our classification results with those of Rubin et al. (2007).

Samples
Thirty-two thin sections of 19 different CM chondrites (including the anomalous CM chondrites WIS 91600, Dhofar 225, and NWA 10907/10908) and CM fragments in three achondrites were investigated. Table 1 lists all investigated CM chondrites and achondrites having CM-like fragments, whether they are falls or finds, the country of discovery, their type, the source of the sample, and the number of investigated samples for SEM and/or EMP studies.
For most of the samples, distinct fragments were analyzed. A fragment is defined as an area that contains TCIs and can be clearly distinguished from the components within the surrounding material, which is the typical clastic matrix of a CM breccia (e.g., Fig. 1). For the few samples with optically unclear identification of distinct clasts of different degrees of aqueous alteration, the TCIs of the entire samples were randomly analyzed.
The anomalous and untypical CM samples (WIS 91600, Dhofar 225, and NWA 10907/10908) were removed from the main study and are described in Supplement S1 in supporting information. The same holds for some fragments in CM chondrite breccias. Two examples are shown in Fig. 3, and more fragments are shown in Supplement S1.

Scanning Electron Microscopy
A JEOL JSM-6610LV SEM at the Interdisciplinary Center for Electron Microscopy and Microanalysis at the Westf€ alische Wilhelms-Universit€ at M€ unster operating at 20 kV was used to image whole samples and areas of interest. Overview images were assembled using the Microsoft image composition editor. These images were used to search for and to register fragments with different degrees of alteration for further studies (e.g., microprobe analysis). Since CM chondrites occur as fragments in other polymict meteorites, CM-like fragments in three achondrites were also investigated.

Electron Microprobe Analysis
Quantitative TCI analyses of different fragments and measurements of phyllosilicates of the host unit were obtained using a JEOL JXA-8530F field emission electron microprobe at the Institut f€ ur Mineralogie (M€ unster), which was operating at 15 kV and with a probe current of 15 nA. For TCI and phyllosilicate analyses, different natural and synthetic standards were used: Na (jadeite), Mg (San Carlos olivine), Al (disthene), Si (hypersthene), P (apatite), S (celestine), K (sanidine), Ca (diopside), Ti (rutile), Fe (fayalite), Cr (Cr 2 O 3 ), Mn (rhodonite), Ni (NiO), and Cl (tugtupite). The measuring points were distributed on the TCIs with variable spot sizes between 3 and 15 µm depending on the size of the TCI (Fig. 4). Some fragments with coarse-grained or fishbone-like TCIs (e.g., in Maribo) were measured with larger spot sizes up to 30 µm. The analyses of TCIs of the host unit for samples with few/no fragments were done with variable spot sizes between 10 and 30 µm. Examples for the selection of analyzed TCIs in specific fragments are given in Fig. 4. Considering the results given in Table 2, the "FeO" value is given with quotation marks because it is not clear which form of Fe was measured; the measurements could include variable abundances of FeO in phyllosilicates, Fe 2+ in sulfides, and Fe 3+ in cronstedtite (compare with Rubin et al. 2007). A small number of measurements were excluded from the average calculation of the TCI compositions because of contamination with surrounding phases. This includes measurements with >3.0 wt% Ni (perhaps from pentlandite or metal) and CaO> 3.0 wt% (probably due to abundant calcite). Most analyses show SiO 2 values of > 10.0 wt%, similar to those of Rubin et al. (2007). Three samples show lower than 10 wt% SiO 2 , but these samples fit into the trend indicating a higher subtype classification.
In the past, the calculated S/SiO 2 and "FeO"/SiO 2 ratios and the metal abundance were mainly considered for the CM2 subtype classifications (compare tables 4 and 5 of Rubin et al. 2007). Although the "FeO"/SiO 2 ratios play a key role in the subtype classification by Rubin et al. (2007), that classification system also uses the other characteristics. In this study, we will only use the "FeO"/SiO 2 ratios. The results of the analyses of 1000 TCIs in distinct clasts and bulk meteorites are given in Table 2.

Investigated Meteorites
Nineteen different CM chondrites and various CM-like clasts in three achondrites were examined. Most samples are characterized by clear evidence of being polymict breccias (Figs. 1-5); however, the degree of brecciation can be highly variable on a thin section (cm) scale. Murchison, Santa Cruz, Y-791198, and Maribo only show very weak or unclear evidence of brecciation, and the degree of aqueous alteration was obtained by random analyses of TCIs in several areas of the bulk sample (Table 2).

Chemical Composition of TCI-Rich Fragments within CM Chondrites
From the different meteorites, 27 thin sections were chosen for microprobe analysis to subclassify the fragments within the CM chondrites (and brecciated achondrites) or for random TCI analysis in the case of the widely unbrecciated rocks (Table 2). Some fragments that optically appear to be TCI-rich fragments turned out not to be, but rather contain metal-or sulfide-rich intergrowths ( Fig. 3b; see Supplement S1). Those and other fragments were excluded from the subtype classification (see Supplement S1).
Most of the investigated fragments and the unbrecciated bulk rocks can be classified according to the classification scheme of Rubin et al. (2007), mainly based on the "FeO"/SiO 2 ratio. The "FeO"/SiO 2 ratio decreases as alteration increases (Table 2; Fig. 6). Some clasts show significantly higher "FeO"/SiO 2 ratios than the highest mentioned by Rubin et al. (2007) for the subtype classification of 2.6 ("FeO"/ SiO 2 ratio >3.3). Therefore, in this work, we defined new threshold values for subtypes CM2.7-2.9. The "FeO"/SiO 2 values for the borders are related to data from Rubin et al. (2007), who defined the borders of CM2.2/2.3 (at 1.5 and 1.7, mean of 1.6 used here), CM2.4/2.5 (at 2.0), and CM2.6/2.7 (at 3.3) in their table 5. All threshold values for the subgroups CM2.0-CM2.9 used in this work are given in Table 3. Based on these values, the classified subtypes for the different fragments and bulk rocks (or main lithologies, if some fragments may occur) of all investigated samples are listed in Table 2.
In general, a positive correlation is observed when the "FeO"/SiO 2 ratios of the analyzed TCIs in the individual clasts are plotted against the S/SiO 2 ratios (Fig. 6). However, there are deviations from the trend: Remarkable exceptions of the positive correlation are shown by the values obtained for several clasts (especially for fragment B of Jbilet Winselwan, fragment C of LON 94101 co. 34, both fragments [A and B] of NWA 12651, and some others; see the large triangle in Fig. 6). These fragments show low S/SiO 2 ratios compared to relatively high "FeO"/SiO 2 ratios and, therefore, do not fit into the correlation. These fragments will be considered in detail in the discussion.
The mean compositions of some main oxides (SiO 2 , "FeO," MgO) and ratios of all subtypes (CM2.0-CM2.9) are given in Table 3. The single subtypes generally show smooth transitions. In some cases, values slightly overlap, and a clear distinction is only possible by using the "FeO"/SiO 2 ratio. The mean "FeO"/SiO 2 and S/SiO 2 ratios as well as the "FeO" content increase from strongly altered to moderately or less altered subtypes, while the mean MgO and SiO 2 concentrations decrease from strongly altered to less altered samples (Table 3). Therefore, other correlations are obvious when considering the different aqueous alteration subtypes (Fig. 7). Figure 7 shows a negative correlation between the "FeO"/SiO 2 ratio and the MgO content (wt%). Strongly altered subtypes (2.0-2.2) have higher MgO contents of~15.0-27.0 wt% compared to moderately or weakly altered subtypes (>2.6), with MgO concentrations of~5.0 to 10.0 wt%. Furthermore, the abundance of clearly visible and weakly altered components (such as chondrules) increases with higher subtype classification. Table 3 also includes some additional aspects concerning Table 2. Mean S/SiO 2 and "FeO"/SiO 2 ratios of the studied fragments and of TCI measurements of the main lithology for samples with very few fragments (Murchison, Y-791198, Jbilet Winselwan). For samples with no clearly detectable fragments (Santa Cruz, Maribo), the TCIs of the bulk rock were randomly analyzed. Subtype classification of fragments/areas only based on the mean values of the "FeO"/SiO 2 ratio; fragments within individual polished sections are ordered based on the "FeO"/SiO 2 ratio. the optical appearance of the TCIs and their visible contrast to the surrounding materials.

Chemical Composition of TCIs in CM chondrites without Clearly Detectable Fragments
For some samples with few or no obvious fragments, the compositions of the TCIs of the host unit were measured. The S/SiO 2 ratio, "FeO"/SiO 2 ratio, and subtype of the TCI measurements from the host units are given in Table 2. In Maribo and Santa Cruz, no distinct fragment could be identified; thus, the composition of the TCIs of the bulk rock was determined. In Yamato 791198 Murchison, and Jbilet Winselwan, very few barely visible fragments were observed from which the compositions of the TCIs were obtained. Most samples, especially those from Murchison, show high "FeO"/SiO 2 ratios, indicating a low degree of alteration (Table 3). Besides these fragments, the compositions of the TCIs of the remaining main lithology were obtained by random analyses of TCIs. Figure 8 shows the "FeO"/SiO 2 versus S/SiO 2 ratio for TCIs in the bulk rock (in the case of Maribo and Santa Cruz) or the main lithology of these CM chondrites.
The positive correlation between the "FeO"/SiO 2 and S/SiO 2 ratio is visible in the diagram. It is Table 2. Continued. Mean S/SiO 2 and "FeO"/SiO 2 ratios of the studied fragments and of TCI measurements of the main lithology for samples with very few fragments (Murchison, Y-791198, Jbilet Winselwan). For samples with no clearly detectable fragments (Santa Cruz, Maribo), the TCIs of the bulk rock were randomly analyzed. Subtype classification of fragments/areas only based on the mean values of the "FeO"/SiO 2 ratio; fragments within individual polished sections are ordered based on the "FeO"/SiO 2 ratio. noticeable that the TCI measurements of the main lithologies (host unit) of Jbilet Winselwan and Santa Cruz also do not completely fit in the trend because of a low S/SiO 2 ratio in comparison to a relatively high "FeO"/SiO 2 ratio. Since only the "FeO"/SiO 2 ratio is used in this work, these rocks have to be classified as CM2.5 and 2.7, respectively.

Fragments Having TCIs with Low S/SiO 2 Ratios
Considering all analyzed fragments in the studied CM breccias, some clasts can be characterized by having low concentrations of sulfur, resulting in very low S/SiO 2 ratios compared to the "FeO"/SiO 2 ratios (Table 4). These special clasts can be identified in the triangle of Fig. 6.

CM-Like Fragments in Achondrites
CM-like clasts can be frequently observed in polymict eucrites, howardites, and rarely in H chondrites (e.g., Fodor and Keil 1976;Zolensky et al. 1996;Gounelle et al. 2003;Patzek et al. 2018a). Previously, we identified and characterized a large number of CM-like clasts (>60) in 15 different HEDs together with CI-like clasts in HEDs and other achondrite and chondrite groups ). Based on this survey, we selected seven clasts for which we here aimed to determine the petrologic subtypes ( Table 2). Three of these clasts observed in the polymict breccias NWA 7542 and EET 87513 are shown in Fig. 9. While clast NWA 7542A-10 turned out to be of petrologic subtype 2.2, the clasts NWA 7542B-13 and EET 87513-02 are of petrologic type 2.6 ( Fig. 9). This CM classification is in agreement with results on the abundance and appearance of individual chondrules and TCI lumps that are surrounded by fine-grained rims (Metzler et al. 1992;Metzler and Bischoff 1996). EET 87513-02 contains abundant chondrules as well as fragments of Fo-rich and Enrich olivine and pyroxene, which are embedded in a mixture of TCI lumps and fine-grained clastic matrix (e.g., Fig. 9c). Several sulfides including pyrrhotite (sometimes close to stoichiometric FeS), pentlandite, and a P-rich sulfide can be observed when viewed in detail. Additionally, a heavily fractured CAI can be observed in one of the clasts that consists of spinel and Ca-rich pyroxene and shows incipient aqueous alteration at the rims. Additionally, several of the unshocked or very weakly shocked clasts consist of different lithologies, themselves resembling the brecciated nature of common CM2 chondrites (Fig. 9b).

Optical and Chemical Characteristics of Individual Fragments
The investigation of different CM chondrites shows that TCI-rich fragments with different aqueous alteration stages can be distinguished using "FeO"/SiO 2 of TCIs and BSE contrast between TCIs and surrounding matrix.
Strongly altered and less altered TCI-rich fragments can be optically distinguished by differences in brightness of the TCIs in BSE images, which, in most cases, contrast with the surrounding clastic matrix. In detail, strongly altered TCI-rich fragments (CM2.0-2.3) appear dark-grayish and generally only show a low contrast to the surrounding material (Figs. 2 and 5a, 5b, 5c), whereas the moderately altered TCI-rich fragments (CM2.4-2.6) show a brighter appearance on BSE images compared to the surrounding matrix (Figs. 2 and 4a). Less altered TCI-rich fragments (CM2.7-2.9) show the brightest appearance and have the highest contrast to the surrounding material (Figs. 4b and 5d, 5e, 5f).
The occurrence of chondrules or other inclusions (CAIs, fragments) can also reflect the degree of aqueous alteration. Strongly altered fragments typically contain components with a high degree of alteration. In these cases, the phenocrysts within chondrules are often replaced by phyllosilicates, whereas the less altered fragments contain a higher abundance of unaltered and minimally altered coarse-grained components (i.e., chondrules, CAIs, and fragments).
As also noted by Rubin et al. (2007), the metal content of different subtypes can also reflect the degree of aqueous alteration. Typically, clasts with a low degree of alteration have a higher abundance of metals. However, the metal abundance is difficult to estimate, since the metal grains are often very small and are  Table 3. Appearance in BSE images and mean concentrations of selected elements (wt%) of TCIs in different subtypes of aqueously altered CM fragments. The "FeO"/SiO 2 values for the borders (threshold values) are related to data of Rubin et al. (2007), who defined the borders for CM2.2/2.3 (at 1.5 and 1.7, mean of 1.6 used here), for CM2.4/2.5 (at 2.0), and for CM2.6/2.7 (at 3.3) in their table 5. For the main lithology of Paris, Rubin (2015) found a high "FeO"/SiO 2 value and suggested that a CM2.7 should have an "FeO"/SiO 2 value between 4.0 and 7.0, which would be a CM2.8 or CM2.9 in this study.  affected by terrestrial weathering resulting in the oxidation of metal grains. The tendency that strongly altered fragments contain no/few metals and moderately and less altered fragments contain more metals can often be observed if fresh meteorite falls are considered. A good example is the Nogoya breccia, in which the strongly altered fragment B (CM2.2) contains no metal, and the moderately altered fragment A (CM2.5), which contains some metals. Nonetheless, using the metal abundance for subgroup classification of most CM chondrites is inappropriate because less than 3% of CM chondrites represent fresh meteorite falls. The optical characteristics observed in BSE images are certainly related to chemical parameters, especially the "FeO" concentrations of the TCIs within the fragments. The FeO, SiO 2 , and MgO concentrations as well as the "FeO"/SiO 2 and S/SiO 2 ratios indicate distinct chemical trends related to the degree of aqueous alteration (Figs. 7 and 8). In general-as was also found by Rubin et al (2007)-the MgO and SiO 2 contents decrease from strongly altered to moderately altered and less altered samples (Table 3). The mean SiO 2 content of strongly altered samples (subtype CM2.0) is~30 wt%, while moderately altered (subtype CM2.4) and less altered fragments (CM2.9) have concentrations of~21 and~10 wt%, respectively. Also, the mean MgO concentration decreases from~25 wt% (subtype CM2.0) to~5.5 wt% (CM2.9). The mean FeO content for subtype CM2.0 increases from~22 wt% to~53.5 wt% (CM2.9). Consequently, the "FeO"/SiO 2 ratio increases from 0.76 for CM2.0 up to 5.76 for subtype CM2.9 and, appropriately, the S/SiO 2 ratio from 0.10 to 0.67. For most of the investigated TCI-rich fragments and CM bulk rocks (or main lithologies; Table 2), the classification scheme of Rubin et al. (2007) works well. However, some TCI-rich fragments of higher subtypes (CM >2.5) have lower S/SiO 2 ratios, which does not fit this trend (Fig. 6).

Variable S Concentrations Among Individual Clasts
As mentioned above, some TCI-rich fragments of high subtypes (CM> 2.5) have too low S/SiO 2 ratios to fit this trend (Fig. 6). Fragments with low S/SiO 2 ratios are: LON 94101 co. 34 fragment C, Jbilet Winselwan fragment B, 12651 fragment A and B, LON 94101 MZ2 fragment E, ALH 85013 fragment D and F, Banten 2 fragment B and C, Santa Cruz 1 fragment A1 and B, and NWA 7542-B fragment 11. These fragments show low S/ SiO 2 ratios relative to their high "FeO"/SiO 2 ratios. The sulfur depletions could have arisen for at least two different reasons (1) the precursor materials from which the TCIs formed contained less S or (2) the S may have been redistributed (washed out) during aqueous alteration on the meteorite parent body.
Regarding the first reason, if we consider that typical TCI compositions include abundant Fe and S, this may indicate that the TCIs were formed from precursors containing sulfides such as troilite, pentlandite, or pyrrhotite. Thus, the precursor material for the S-depleted TCIs could have been a metal-rich component. As a result, the TCIs would show low S concentrations compared to Fe. This could explain the greater variety in S content for different TCI-rich fragments. However, this cannot explain why fragments of higher subtypes generally contain more S than fragments with lower subtypes (CM2.0-CM2.4).
The second possibility, that S was redistributed during the aqueous alteration (Fig. 6), can explain this observation much better. This process would generally lead to lower S abundances in fragments suffering from a higher degree of aqueous alteration. However, for the S-depleted fragments having a high petrologic subtype (CM> 2.5), these TCIs may have formed from precursor materials depleted in S-bearing phases such as metals.

TCI Compositions within Rocks with a Low Abundance of or without Distinguishable Clasts
As described above, no distinct fragments could be identified in Maribo and Santa Cruz; thus, the composition of the TCIs from the bulk rock was determined (Table 2; Fig. 10). In Yamato 791198, Murchison, and Jbilet Winselwan, only very few barely visible fragments were observed from which the compositions of the TCIs were obtained. The sample from Jbilet Winselwan shows a large range of the "FeO"/SiO 2 versus S/SiO 2 ratios for the two TCI-rich fragments, which can be classified as subtypes of CM2.4 to CM2.9, while the TCIs of the host lithology are more homogeneous and have a composition consistent with a subtype classification of CM2.5. Within the nearly unbrecciated CM chondrite Y-791198, only one barely visible area (probably a fragment) was discernible, and this area's TCIs are chemically similar to fragments of subtype CM2.8 ( Fig. 10; Table 2). A similar relationship is found for Murchison. The three visible fragments show a very low degree of aqueous alteration (CM2.9), while the main host rock lithology is of CM2.7. During the random measurements of TCIs in the main host rock lithologies of Murchison, Y-791198, and Jbilet Winselwan, we found that the TCIs in the analyzed areas are not completely homogeneous. However, the variations are not as large as those within samples having clearly definable fragments of different degrees of alteration. The variations probably indicate impact-related mixing of finegrained TCI-fragments within the (clastic) matrix of these three CM-rocks. The effect of impact-related mixing and, thus, "homogenization" is described well in Fig. 7 of Metzler et al. (1992).

CM Chondrites and Thermal Metamorphism of Related Samples
To be able to carry out a correct subclassification of a CM chondrite and their individual fragments, the effects of thermal metamorphism have to be considered. Thermal annealing is changing the textural, mineralogical, and chemical parameters of the chondrites and their individual clasts. In our study, WIS 91600 is recognized not to be a typical CM chondrite (Fig. S1 in supporting information). Optically, the sample does not contain TCI-bearing fragments or TCIs in the clastic matrix. The rock seems to have been influenced by a secondary process like thermal metamorphism. Already Rubin et al. (2007) described WIS 91600 as a thermally metamorphosed CM-like chondrite. Details on several other heated carbonaceous chondrites have been reported (e.g., Tomeoka 1989;Tomeoka et al. 1989;Zolensky et al. 1989aZolensky et al. , 1989bAkai 1990aAkai , 1990bAkai , 1992Bischoff and Metzler 1991;Ikeda 1992;Tonui et al. 2002Tonui et al. , 2003Tonui et al. , 2014Nakamura 2005;Harries and Langenhorst 2013;Ebert et al. 2019). Recently, a large light CM-clast in the Murchison breccia was described (Kerraouch et al. 2018;Bischoff et al. 2018a; Fig. 11). A detailed study shows that this clast having a granoblastic texture formed by metasomatic processes in the depth of a CM parent body (Kerraouch et al. 2019a). In the past, also CV chondrites and their dark inclusions were discussed as objects that suffered aqueous alteration and dehydration in their evolution (e.g., Bischoff et al. 1988;Krot et al. 2004).
For more information and detailed descriptions of the effects of thermal metamorphism on CM chondrites, the work of Tonui et al. (2014) is suggested to the reader.
Results of this Work Compared to those of Rubin et al. (2007) The classification of CM chondrites suggested by Rubin et al. (2007) and the classification in this study are different (Fig. 12). Rubin et al. (2007) classified CM2 subtypes based on the abundance and size of serpentine-tochilinite intergrowths/PCPs (TCIs), on the "FeO"/SiO 2 ratio of the TCIs, as well as on the abundances (and types) of Ca-carbonates and metals. Furthermore, they classify CM2 subtypes based on the dominating petrologic type of fragments, which results in the highest subtype CM2.6. In this study, CM2 chondrites are classified solely based on the "FeO"/SiO 2 ratio defined by Rubin et al. (2007), and we take into account the minimum and maximum degree of alteration of different fragments (Bischoff et al. 2017). As subtypes CM > 2.6 are not considered in the classification scheme of Rubin et al. (2007), this study proposes three further subtypes of CM2.7-CM2.9 for individual clasts in CM breccias. However, the Paris breccia has to be mentioned, for which Rubin (2015) determined a high "FeO"/SiO 2 ratio and suggested a subtype classification of at least CM2.7. As thin sections from Nogoya, Cold Bokkeveld, Murchison, and Y-791198 were part of the classification work of both studies (Rubin et al. [2007] and this study), we show a comparison of the results in Fig. 12. Figure 12 illustrates clearly the difference between the classifications of Rubin et al. (2007) and those in this study. This is especially the case considering the CM breccias Murchison and Y-791198. The very different results indicate that Rubin et al. (2007) must have studied completely different lithologies than those studied in this work. This is not surprising: As known from meteorite breccias, the main fragment can vary from one thin section to another, and individual samples can even show different subtypes, as described by Bischoff et al. (2017). One good example is LON 94101 from which several thin sections were investigated. In most sections, fragments occur that show a large range of subtypes from CM2.0 to CM2.8 (Fig. 1). One sample (LON 94101-Cari 6) appears to be quite homogeneous, with some fragments of subtype CM2.2 (Table 2). Another good example is Nogoya (Fig. 3), which is classified as a CM2.2 by Rubin et al. (2007) based on the degree of alteration of the dominating lithology in their studied thin section. The thin section from this work also contains a fragment with the subclassification of CM2.2 (fragment B; Fig. 3), but the main fragment clearly is of petrologic type CM2.5 (fragment A; Fig. 3).
Based on the complex nature of CM breccias, this study proposes an extended subtype classification  Table 3) and includes the characteristic of individual fragments into the classification in a similar way as done for ordinary chondrite breccias (Bischoff et al. 2017 in order to appropriately account for the complexity of the CM chondrite breccias. Clear differences in the classification (e.g., Y-791198; Fig. 12) may be due to large-scale differences in the petrologic properties of the rock (as an effect of sampling during thin section preparation).

Genetic Relationship of CM-Like Clasts in HED Breccias and CM Chondrites
Several of the CM-like clasts in brecciated achondrites consist of different lithologies themselves, where they sample the border of two lithologies of different petrologic type. This resembles the brecciated nature of common CM2 chondrites as discussed in this work. They also span a range in petrologic subtypes from 2.2 to 2.8. Additionally, the genetic relation between CM-like clasts in HEDs and common CM chondrites is evidenced not only by their typical mineralogy but also by their D/H ratios, oxygen isotopes, and e 54 Cr composition (e.g., Buchanan et al. 1993;Gounelle et al. 2005;Van Drongelen et al. 2016;Patzek et al. 2018aPatzek et al. , 2018b. Since many of the CM2-like clasts show no evidence for intensive heating or shattered structures, high-velocity impacts of larger impactors were not likely the major process of delivery to the HED parent bodies. Rather, the material was delivered via low-velocity impacts that mixed the clasts into the regolith of the HED parent body (e.g., Reddy et al. 2012). Features indicative for post-shock heating and shock are rare, which implies that low-impact velocities in combination with the general porous nature of carbonaceous chondrite material are the relevant process for delivery of CM material to the HED parent body. Furthermore, studies on the shock effects of carbonaceous chondrites have shown that most CM chondrites are unshocked or very weakly shocked (Scott et al. 1992), which is consistent with the observation of the CM-like clasts in HEDs. Nonetheless, heating features are observable in the finegrained, phyllosilicate-rich matrix using TEM techniques (Buchanan et al. 1993). Also, Rubin and Bottke (2009) studied CM-like clasts in breccias and showed that clast PV3 from Plainview appears to have lost serpentine during its impact on the H-chondrite parent asteroid indicating a high impact velocity. Overall though, it is likely that most of the CM-like clasts in HEDs were incorporated into the host rock by slow infall of (micro)meteorites rather than by highvelocity impacts, since some are apparently not heated or fractured (Gounelle et al. 2003;Patzek et al. 2018a).

Brecciated CM Chondrites as Analogs for the Asteroidal Samples Returned by the OSIRIS-REx and Hayabusa 2 Missions
CI and CM chondrites (as also shown in this study) are highly brecciated rocks (e.g., Metzler et al. 1992;Bischoff et al. 2006Bischoff et al. , 2017Morlok et al. 2006;Lindgren et al. 2013;Zolensky et al. 2015Zolensky et al. , 2017Alfing et al. 2019). The characterization of CM breccias gives insight into the formation processes of CM parent bodies as well as into the evolution of asteroid surfaces. Considering the current NASA and JAXA missions to the asteroids (101955) Bennu and (162173) Ryugu, respectively, the microscopic observations on CI and CM chondrites may help to decipher distinct surface features on Bennu and Ryugu. Both asteroids are regarded as consisting of materials affected by aqueous alteration (e.g., Hamilton et al. 2019;Kitazato et al. 2019). Ryugu is described as a rubble pile-like body with a very low density and an estimated high porosity of >50% in its interior . Early spectral data of OSIRIS-REx show evidence for abundant hydrated minerals on the surface of Bennu as demonstrated by the near-infrared absorption at 2.7 µm . These authors also state that the thermal infrared spectral features are most similar to those of aqueously altered CM-type carbonaceous chondrites. In addition, the low density of Bennu is consistent with a rubble-pile structure of high porosity assuming a particle density characteristic of CM chondrites . Fractured boulders as shown on surface images McCoy et al. 2019)-clearly demonstrating surface rock heterogeneities-have morphologies that suggest the influence of impact or thermal processes. Considering Bennu as a possible CM parent body, the latter would not be surprising, since some CM chondrites available for study completely lost their volatiles due to heating or contain fragments that were formed in water-free environments or were strongly metamorphosed as shown by the centimeter-sized light clast in Murchison (Fig. 11) that has been identified as a type 6 lithology of a CM parent body (Kerraouch et al. 2018(Kerraouch et al. , 2019a(Kerraouch et al. , 2019bBischoff et al. 2018a). In summary, the mineralogical, textural, and chemical work on CM and CI breccias will certainly help to evaluate and understand the hopefully returned samples from the OSIRIS-REx and Hayabusa 2 missions.

CONCLUSION
In this study, a large number of TCI-rich fragments/areas from 27 thin sections out of 19 different CM chondrites and three brecciated HEDs were examined in order to reveal information on their brecciation processes and on the degree of aqueous  Rubin et al. (2007); the black bars represent the petrologic subtype as defined by Rubin et al. (2007) and the colored areas show the range defined in this work by taking into account differently altered lithologies. Since the thin sections studied here are different from those analyzed by Rubin et al. (2007), the different results may be due to large-scale differences in the petrologic properties of the bulk rocks. This must especially be the case considering the CM breccias Murchison and Y-791198. * As shown in Fig. 3, LON 94101 may also contain C1-and C3related clasts; ** considering the CM6-clast Kerraouch et al. 2019aKerraouch et al. ,2019b. (Color figure can be viewed at wileyonlinelibrary.com.) alteration of their individual clasts. According to the scheme of Rubin et al. (2007), 80 fragments, main lithologies, and bulk rocks were classified by determination of the "FeO"/SiO 2 ratios of the TCIs. New subtypes (CM2.7-CM2.9) were defined in this study to allow for a precise classification of samples with a low degree of aqueous alteration.
Most studied CM chondrite breccias contain individual fragments which are affected by various degrees of aqueous alteration (Fig. 12). The abundances of different clasts with different degrees of aqueous alteration can vary from thin section to thin section, even if the thin sections are from the same CM chondrite rock. Therefore, this study proposes an extended classification scheme based on that of Rubin et al. (2007) but includes the fragments with the highest and lowest degree of alteration in the classification (e.g., CM2.3-2.7, CM2.0-2.8, etc.). This type of classification includes the brecciation characteristics and is similar to the classification of ordinary chondrites (e.g., H3-5; L4-6, etc.). Good examples are the CM chondrites LON 94101 (CM2.0-2.8; clearly even more complex considering the fragments shown in Fig. 3), Cold Bokkeveld (CM2.1-2.7), and Nogoya (CM2.2-2.5).
Furthermore, fragments exist in which the TCIs show a higher degree of aqueous alteration based on the "FeO"/SiO 2 ratios than as indicated by their S/SiO 2 ratios. This might be due to a redistribution of sulfur during the aqueous alteration. Also, the low S/SiO 2 ratio of the TCIs could be related to the alteration of different precursor materials that were S-free phases (like metals). Thus, the "FeO"/SiO 2 ratio of TCIs is more reliable than S/SiO 2 as a parameter for subtype classification of CM chondrites and their fragments.

SUPPORTING INFORMATION
Additional supporting information may be found in the online version of this article. Supplement S1. Anomalous Samples and Unique Fragments. Fig. S1. BSE image of WIS 91600 having small magnetite inclusions and magnetite nests in the clastic matrix Dhofar 225. This anomalous CM rock seems to be modified by secondary processes (Fig. S2). In the beginning of this study, the sample was measured but later excluded from the classification because the measurements indicate that the samples did not contain real TCIs, but areas looking like TCIs turned out to be composed of sulfidic or metallic components rather than phyllosilicates. Another reason for excluding this meteorite is its oxygen isotope composition, which is more enriched in 17 O and 18 O than those of other CM chondrites and plots next to the Belgica group (see Meteoritical Bulletin Database; https://www.lpi.usra.ed u/meteor/). Fig. S2. Anomalous CM chondrite Dhofar 225: Bright areas that look like TCIs probably represent areas with abundant sulfidic or metallic components rather than phyllosilicates. Fig. S3. BSE images of TCI-rich areas in (a) NWA 10907 and (b) NWA 10908: The light TCI-rich areas show very different optical characteristics compared with those shown in the CMs and their fragments in the main document (compare Figs. 4 and 5). Fig. S4. BSE images of some fragments from Jbilet Winselwan, NWA 7542, and Saric ßic ßek having bright areas looking like TCIs, but contain sulfidic and/or metallic components.