Cosmic ray exposure ages for ureilites—New data and a literature study

We report newly measured noble gas isotopic concentrations of He, Ne, and Ar for 21 samples from the 10 ureilites, DaG 084, DaG 319, DaG 340, Dho 132, HaH 126, JaH 422, JaH 424, Kenna, NWA 5928, and RaS 247, including the results of both single and stepwise heating extractions. Cosmic ray exposure (CRE) ages calculated using model calculations that fully account for all shielding depths and a wide range of preatmospheric radii, and are tailored to ureilite chemistry, range from 3.7 Ma for Dho 132 to 36.3 Ma for one of several measured Kenna samples. In a Ne‐three‐isotope plot, the data for DaG 340 and JaH 422 plot below the Necos/Neureilite mixing envelope, possibly indicating the presence of Ne produced from solar cosmic rays. In combination with literature data and correcting for pairing, we established a fully consistent database containing 100 samples from 40 different ureilites. The CRE age histogram shows a trend of decreasing meteorite number with increasing CRE age. We speculate that the parent body of the known ureilites is moving closer to a resonance and/or that there is a loss mechanism that acts on ureilites independent of their size. In addition, there is a slight indication for a peak in the range 30 Ma, which might indicate a larger impact on the ureilite daughter body. Finally, we confirm earlier results that the majority of the studied ureilites have relatively small preatmospheric radii less or equal ~20 cm.


INTRODUCTION
With more than 470 individually classified objects, the ureilite meteorites make up the second largest group of achondrites. The known ureilites are carbon-rich ultramafic achondrites largely consisting of Mg-rich olivine and pyroxene that might represent mantle material from a single parent asteroid, the ureilite parent body (UPB) (Warren and Kallemeyn 1989;Goodrich 1992;Goodrich et al. 2004). Ureilites are subdivided into the main group, unbrecciated ureilites (formerly called monomict), and the brecciated ureilites that are polymict or dimict (Goodrich et al. 2004(Goodrich et al. , 2015Downes et al. 2008). According to recent thermal modeling and age dating, the UPB assembled < 1 Ma after CAI formation followed by partial melt segregation and crystallization no later than 5 Ma after CAI formation (Wilson et al. 2008;Goodrich et al. 2010;Amelin et al. 2015). However, some aspects of ureilites are difficult to reconcile with an igneous petrogenesis, such as variable oxygen isotope (D 17 O) compositions and high abundances of trapped noble gases (e.g., Clayton and Mayeda 1988;Goodrich 1992;Mittlefehldt et al. 1998). Some of the noble gases, however, might be trapped in refractory graphite and/or diamond phases (e.g., G€ obel et al. 1978). All these findings make it difficult to distinguish whether ureilites represent primitive or igneous material and if they can be linked to a common source.
Considering the final steps of the UPB history, Goodrich et al. (2004) suggested that the UPB broke up early in the solar system, i.e., about 5 Ma after CAI formation, and reassembled as one or more daughter bodies (ureilite daughter body, UDB). The known ureilites are derived most likely from one of the heterogeneous brecciated UDBs (Downes et al. 2008;Herrin et al. 2010a;Warren 2012;Goodrich et al. 2015;Sanders et al. 2017). The heterogeneity is indicated by the wide variation in bulk composition, Fo content of olivine, and oxygen isotopes among the various ureilite meteorites (e.g., Clayton and Mayeda 1988;Downes et al. 2008;Warren 2012). The production and delivery of ureilites to Earth may be a result of either a single large impact or multiple small impacts. In the first case, the ureilites should show a single cosmic ray exposure (CRE) age; in the second scenario, the CRE ages will be variable. Remember that clustering in CRE ages for the howardite-eucrite-diogenite meteorites was one of the arguments to the idea that all three meteorite types come from a single parent body (e.g., Eugster and Michel 1995).
There are important reasons to establish a consistent CRE age histogram for ureilites. However, much of the existing noble gas data for ureilites predate the development of modern cosmogenic production rate models, such as that of Leya and Masarik (2009). As such, typical data reported in the literature either omit cosmogenic production rate data and exposure ages, use older models based solely on chemical composition but without considering shielding (e.g., Mazor et al. [1970] and references therein), or use empirical models such as Eugster (1988), Aylmer et al. (1990), or Eugster and Michel (1995), which take into account both shielding (via the cosmogenic 22 Ne/ 21 Ne ratio) and chemical composition. Three recent approaches also compiled noble gas data from literature but used a constant, i.e., shielding independent 21 Ne production rate (Herzog and Caffee 2014;Park et al. 2014;Beard and Swindle 2017).
Here we present new noble gas data for 10 ureilites, most of which have not previously been studied for noble gases and/or CRE histories. In combination with literature data, we established a consistent database of CRE ages for 100 samples from 41 ureilites (before pairing). The CRE ages were all calculated using production rates determined using the model by Leya and Masarik (2009) assuming average chemical composition for the ureilites and fully considering the dependence of the 21 Ne production rates on preatmospheric radii and shielding depths. The CRE age histogram indicates a decrease, which might be an exponential, of meteorite number with increasing CRE age, indicating a continuous delivery of ureilites to Earth. Superimposed on the decrease is a slight indication for a peak in the range 30 Ma, which might be statistical significant.

EXPERIMENTAL Sample Selection and Sample Preparation
We measured the He, Ne, and Ar isotopic concentrations of 10 ureilites. Eight ureilites are from the hot deserts of Libya (Dar al Gani "DaG" 084, DaG 319, DaG 340, and Hammadah al Hamra "HaH" 126) and Oman (Dhofar "Dho" 132, Ramlat as Sahmah "RaS" 247, Jiddat al Harasis "JaH" 422, and JaH 424). The samples from DaG 319, DaG 340, RaS 247, JaH 422, and JaH 424 were all provided by the Naturhistorisches Museum Bern/Switzerland (B. Hofmann). In addition, we studied one ureilite from Northwest Africa (NWA 5928, provided by T. Jakubowski) and we also included the ureilite Kenna in our study. Most samples are normal ureilites and ranged in weathering grade from W2 to W4. Two were petrographically distinct: DaG 319 is a polymict breccia and JaH 422 is an impact melt breccia, which makes interpreting the noble gas data likely more difficult. However, both samples were available to us and were therefore included in this study. Most of the ureilites were not studied for noble gases before. Preliminary data were reported by Cosarinsky et al. (2010).
The samples consisted of small chips almost free of fusion crust (excepting HaH 126, which contained some fusion crust). All samples were cleaned with isopropanol and dried in air. Samples for total melt extractions were in the range of 45-120 mg; samples for stepwise heating extractions ranged from 120 to 140 mg.

Noble Gas Measurements and Data Handling
Samples were prepared and measured using the same procedures and on the same spectrometers as described in detail by Huber et al. (2008). Briefly, samples were preheated in vacuum to release atmospheric gases trapped at the surface. All samples were degassed in vacuum by radio frequency induction in a Mo crucible at~1700°C for 35 min to extract noble gases. A second heating step at~1750°C confirmed complete degassing at 1700°C. For the four samples marked "SWH" in Tables 1 and 2, a stepwise heating experiment was performed with heating steps of 600°C, 800°C, 1000°C, 1200°C, 1400°C, 1600°C, and a final step at 1700°C, 1740°C, or 1800°C. The "total" value given in Tables 1 and 2 is the sum over all temperature steps. The gases released from the samples were purified on Ti getters operated in the temperature range between room temperature and 700°C. The He-Ne fraction was separated from the Ar fraction by trapping the latter on activated charcoals cooled to the temperature of boiling liquid nitrogen (LN 2 ,. After separation, gases were further purified by Ti getters and He-Ne was measured on a sector field mass spectrometer and Ar was measured on a tandem spectrometer. Both instruments were built in-house and both are equipped with a Faraday Cup and an electron multiplier working in analog mode (see Schwarzm€ uller [1971] for details). Helium-Ne measurements were carried out with the spectrometer connected to a cold trap cooled with LN 2 to minimize interferences from Blank measurements were performed using the same extraction and measurement procedure as for samples. For the measurements with one gas extraction at 1700°C, blanks were measured before and after each sample measurement. For the stepwise heating extractions, blanks were measured at 600°C, 1000°C, 1400°C, and 1700°C and interpolated linearly for the intermediate steps (800°C, 1200°C, 1600°C). The blank (whether measured or interpolated) for each step was then used to correct the sample gas amounts of this respective temperature step. After measuring samples that contained large gas amounts, the crucible was degassed at 1800°C. The blanks measured by melting the same commercial Ni foil used to wrap the samples did not differ from those measured without melting any foil. Average blank amounts are (in 10 À10 cm 3 STP) 4 He~0.1, 20 Ne~0.01, and 40 Ar~1. The isotopic composition of the blanks was mostly atmospheric or, in the case of Ar, a mixture of air and trapped ratios. Blanks are typically in the range of less than 10% for 4 He, less than 1% for Ne isotopes but can be substantial for Ar isotopes, especially for 40 Ar due to extremely low 40 Ar concentrations in the samples.
Instrumental mass discrimination and sensitivities of the He-Ne and Ar mass spectrometers were regularly determined by analyzing calibration gases with known amounts of He, Ne, and Ar; all having atmospheric isotopic compositions, except for He, which is enriched in 3 He ( 4 He/ 3 He~100). Since the sensitivity of the spectrometers depends on gas amounts, the data were also corrected for nonlinear behavior of the ion source and the detectors. The correction factors were empirically derived from dilution series. For most of the samples studied here the corrections are on the order of a few percent but can reach up to 12% for He and 25% for Ne for the lowest gas amounts measured in stepwise heating measurements. Note that the corrections are only for gas amounts but not for isotopic ratios.

New Data for 21 Samples from 10 Ureilites
Here we discuss the results for the 10 ureilites studied by us. We focus the discussion on the noble gas release characteristics of the stepwise heating experiments and on the determination of cosmogenic noble gas concentrations, i.e., on the correction for trapped components. The data are given in Tables 1 and 2.

Stepwise Heating Extractions
Helium: Cosmogenic 4 He/ 3 He ratios typically range between 3.5 for small meteorites and~6 or slightly higher for larger meteorites (e.g., Alexeev 1998;Welten et al. 2003;Leya and Masarik 2009). From the data in Fig. 1 (upper panel), we can see that 4 He/ 3 He ratios are cosmogenic or close to cosmogenic for all temperature steps below 1000°C for JaH 422, 800°C for DaG 340, 1200°C for Kenna, and 800°C for RaS 247. Higher temperature steps release He with higher 4 He/ 3 He ratios, reaching 455 for the 1600°C temperature step of DaG 340. Consequently, cosmogenic gases are released early during the stepwise pyrolysis, indicating that the radiogenic He and/or a planetary He component are both more tightly bound than the spallogenic component, which is in contrast to most stepwise heating data for other meteorite types. Most of the 4 He/ 3 He ratios are relatively low. Assuming that 3 He in ureilites is mostly cosmogenic and that most 4 He is radiogenic, the low 4 He/ 3 He ratios are consistent with low U and Th concentrations (see below) (e.g., Janssens et al. 1987;Mittlefehldt et al. 1998).
Neon: The 20 Ne/ 22 Ne ratios range between 0.819 for the 1740°C step of Kenna and 10.5 for the second 1700°C temperature step of JaH 422 (Fig. 1, middle panel). For Kenna and RaS 247, the 20 Ne/ 22 Ne ratios for all temperature steps are below 2 or even below 1 (for Ras 247), clearly indicating that the released Ne is dominantly cosmogenic. For JaH 422, the ratios range between 0.945 and 10.5, indicating the admixture of a trapped component. Most of the trapped gases are released at the 600°C temperature step and above 1200°C, while the temperature steps at 800°C and 1000°C release Ne with 20 Ne/ 22 Ne ratios close to cosmogenic. The release pattern is slightly different for DaG 340, the 600°C temperature step releases some trapped gases ( 20 Ne/ 22 Ne~2), the following step releases Ne mostly of cosmogenic origin ( 20 Ne/ 22 Ne1 ), followed by Ne with a significant trapped contribution, i.e., a 20 Ne/ 22 Ne ratio in the range 3.3-7.7. Above 1700°C, i.e., close to the final degassing temperature, the 20 Ne/ 22 Ne ratio is again close to cosmogenic. Since many of the Dar al Gani ureilites are polymict, DaG 340 might be a single clast from a polymict ureilite and might therefore contain additional and/or other trapped Ne components, which would explain why the release pattern for DaG 340 differs from that of the other studied samples.
Argon: Figure 1 (lower panel) depicts for the four ureilites studied via stepwise heating techniques the 38 Ar/ 36 Ar ratios for all temperature steps. The ratios for JaH 422 vary in the small range from 0.183 to 0.196, clearly indicating that the released Ar is dominated by a planetary component (also indicated by the low

Single Extractions and Totals of Stepwise Heating Extractions
Helium: The measured 4 He/ 3 He ratios vary betweeñ 6 for DaG 084 and~166 for Dho 132. The majority of the data are below 10, clearly indicating that contributions of radiogenic and/or planetary 4 He are low. Assuming that all 3 He is cosmogenic and that the cosmogenic 4 He/ 3 He ratio is 5.72, i.e., given by the lowest measured 4 He/ 3 He ratio, we can calculate the amount of radiogenic and/or planetary 4 He for each meteorite. However, we start the discussion with cosmogenic 3 He. The reproducibility of the 3 He concentrations for the three DaG 340 and JaH 422 samples is reasonable; the maximum differences are 18% and 6%, respectively. The situation for Kenna is different; the 3 He concentrations for the seven analyzed samples vary by more than 60%, which is much more than the experimental reproducibility (about 10%) that we routinely checked using a homogeneous sample powder from the meteorite Millbillililie. For RaS 247, the differences between the two samples, one measured by stepwise heating and one by one step pyrolysis, are 37%, i.e., also higher than the experimental reproducibility. We might speculate that part of the bad reproducibility is caused by an inhomogeneous chemical composition. However, varying the carbon concentration between zero and 10% changes the 3 He concentration only by about 10% (based on model calculations by Leya and Masarik 2009), i.e., not enough to fully account for the observed variation.
A comparison of our results to the few data found in literature reveals some serious discrepancies. The new 3 He concentrations for DaG 084 are more than a factor of~2 lower than the value given by Scherer et al. (2000). A similar observation holds for DaG 319: the value by Scherer et al. (1998) is more than a factor of 2.5 higher than ours. For DaG 340, all three of our data are lower by a factor of 1.8 than the value given by Scherer et al. (1998). For Kenna, the results are contradictory: the data by G€ obel et al. (1978) and Wilkening and Marti (1976) are in the range (62-74) 9 10 À8 cm 3 STP g À1 , whereas our measurements give values in the range (25-40) 9 10 À8 cm 3 STP g À1 , i.e., substantially lower.
The difficulty with the 3 He data can best be seen in Fig. 2 in which we plot ( 3 He/ 21 Ne) cos as a function of ( 22 Ne/ 21 Ne) cos for the studied ureilites together with literature data and results from the model calculations discussed further below. Also shown is the empirical correlation by Nishiizumi et al. (1980) that is based on data from 138 chondrites. Three important trends are visible. First, the modeled trend differs from the empirical correlation especially for low ( 22 Ne/ 21 Ne) cos ratios. This finding either indicates that the empirical correlation deduced for chondrites is not applicable to ureilites and/ or that the model fails in describing production rates for small ureilites. Second, all of our data as well as most of the literature fall below both the model predictions and the empirical correlation line (Nishiizumi et al. 1980). Third, the experimental data scatter widely, showing no visible trend of ( 3 He/ 21 N) cos as a function of ( 22 Ne/ 21 Ne) cos . Especially important is thereby the finding that also data from one ureilite scatter significantly, indicating large sample-to-sample heterogeneities. As an example we discuss the data from Kenna. The data from our study are shown by solid black dots and the literature data (Wilkening and Marti 1976;Okazaki et al. 2003;Rai et al. 2003) are shown by open dots. While the ( 22 Ne/ 21 Ne) cos ratios vary slightly more than expected considering the experimental uncertainties (1r standard deviation: 2-3%), the maximum deviation for the ( 3 He/ 21 Ne) cos ratios is~75% with a standard deviation of the mean of~20%. Consequently, while the reproducibility of the ( 22 Ne/ 21 Ne) cos ratio is reasonable, the variation of ( 3 He/ 21 Ne) cos is far too large to be caused by analytical artifacts. Note that we can for our samples exclude 3 He losses caused by sample preparation because all samples were treated similarly (same preheating temperature, same duration of preheating, similar vacuum conditions) and we regularly checked the noble gas extraction and cleaning procedures using calibration gases and standard meteorites. It can be seen that the literature data (open symbols, Okazaki et al. 2003;Rai et al. 2003;Wilkening and Marti 1976) show the same trend as our data, they plot into the same field far below the model predictions and the empirical correlation line, also indicating substantial and highly variable 3 He deficits. Consequently, while our data at first glance indicate to be highly variable and therefore unreliable they are fully consistent with literature data. Therefore, the significant scatter most likely rather indicates an inherent feature of the ureilites that is worth being studied in more detail (see below). The second important feature in Fig. 2 is that most of the experimental data (shown a solid black stars) scatter widely below the model predictions and/or the empirical correlation line, indicating that most ureilites show variable 3 He deficits relative to what is expected. This finding indicates that most of the 3 He cos concentrations published for ureilites might be too low due to 3 H and/or 3 He diffusive losses. Since in this study we are interested in reliable CRE ages for ureilites, we concentrate on 21 Ne cos , which we consider as more reliable than 3 He cos . Note that it would be worth studying in more detail if and why 3 He deficits seem to be more common in ureilites than in other chondrite types.
With the assumed cosmogenic ( 4 He/ 3 He) cos ratio of 5.72, we calculated excess 4 He and assumed all of it to be radiogenic, i.e., 4 He rad . Considering that 3 He is most likely too low due to 3 He losses, this procedure likely overestimates 4 He rad . Furthermore, assuming all excess 4 He to be radiogenic also overestimates 4 He rad because contributions from planetary 4 He are expected. The 4 He rad concentrations vary between zero for RaS 247 and DaG 084 to~1700 9 10 À8 cm 3 STP g À1 for one sample of DaG 340; they also vary significantly for different samples from the same meteorite. For example, the 4 He rad concentrations vary by a factor of~3 for the three DaG 340 samples, by a factor of~3 for the three JaH 422 samples, and by more than a factor of~4 for the seven Kenna samples. Assuming U and Th concentrations of 1 ng/g and 3 ng/g, respectively (cf. Mittlefehldt et al. 1998), we calculate 4 He gas retention ages ranging from zero to an unreasonably high apparent age of 8100 Ma. For 6 of the 21 samples, we calculate apparent ages older than the age of the solar system. With U and Th concentrations of 1 and 3 ng g À1 , respectively, we calculate a maximum possible 4 He rad concentration of 155 9 10 À8 cm 3 STP g À1 . Consequently, higher 4 He rad concentrations are only possible for higher U and Th concentrations. Turning the argument around, a 4 He rad concentration as high as 1700 9 10 À8 cm 3 STP g À1 measured for one DaG 340 sample requires U and Th concentrations in the range 10 and 30 ng/g, i.e., much higher than usually measured in ureilites (Mittlefehldt et al. 1989). Note that Janssens et al. (1987) measured for one ureilite (Dylapur) a high U concentration of almost 7 ng g À1 , i.e., much higher than the average. It is unlikely that most of the ureilites studied by us have such exceptionally high U and Th concentrations, especially if we consider that Janssens et al. (1987) measured for other ureilites U concentrations in accord with low values in the range 1 ng g À1 . We therefore conclude that some of the studied ureilites contain substantial amounts of trapped ureilite He with 3 He/ 4 He ratios likely similar to Q-He, i.e., 3 He/ 4 He~10 À4 . However, a deconvolution of the measured data into spallogenic, radiogenic, and planetary components is not possible.
Neon: Figure 3 depicts the Ne data in a Ne-threeisotope plot. In addition to the data for the single  (Nishiizumi et al. 1980) indicating that almost all ureilites experienced 3 He deficits. temperature step extractions, we also plot the data from the stepwise heating extractions, which helps to better determine the mixing line(s) and to better correct for trapped components. We also plot the Ne endmember composition for ureilite gases (index "ureilite"), which has a 20 Ne/ 22 Ne ratio of either 10.4 (Ott et al. 1985) or 10.7 (G€ obel et al. 1978;Busemann et al. 2000). The ratio ( 21 Ne/ 22 Ne) ureilite is not well known. Here we assume ( 21 Ne/ 22 Ne) ureilite = 0.031 (e.g., Busemann et al. 2000). All data plot on a mixing line between cosmogenic Ne and a trapped component, which is most likely Ne ureilite . The two dashed lines are drawn to cover all data using Ne ureilite as one endmember, i.e., both lines start at Ne ureilite . Doing so, the cosmogenic 21 Ne/ 22 Ne ratio (index "cos") at ( 20 Ne/ 22 Ne) cos = 0.87 varies between 0.73 and 0.92, which is lower than the average ratio for most chondrite types. For the determination of 21 Ne cos and ( 22 Ne/ 21 Ne) cos , we use a standard two-component deconvolution with ( 20 Ne/ 22 Ne) ureilite = 10.4 (Ott et al. 1985) and ( 20 Ne/ 22 Ne) cos = 0.84, which is the lowest ratio measured in this study for a bulk sample (RaS 247, both total of stepwise heating and single temperature extraction). The results for 21 Ne cos and ( 22 Ne/ 21 Ne) cos vary only very little with our choice for the Ne ureilite endmember, i.e., 10.7 (G€ obel et al. 1978;Busemann et al. 2000) or 10.4 (Ott et al. 1985). The 21 Ne cos concentrations determined in this study vary between 0.8 9 10 À8 cm 3 STP g À1 and 11.6 9 10 À8 cm 3 STP g À1 (Table 1). The reproducibility of the 21 Ne cos concentrations for samples that have been measured more than once is usually within or at least close to the uncertainties. For example, the standard deviations of the mean 21 Ne cos concentrations are 3.5%, 7.8%, and 7.0% for JaH 422, DaG 340, and Kenna, respectively. The only exception is RaS 247, for which 21 Ne cos of the two samples varies by almost a factor of two. The reproducibility of ( 22 Ne/ 21 Ne) cos , which for all studied samples range from 1.064 to 1.377, is slightly larger than the experimental uncertainties, the variation is less than 4% for the different JaH 422 samples and less than 3% for the three DaG 340 samples. For the seven Kenna and two RaS 247 samples, the maximum variation in ( 22 Ne/ 21 Ne) cos is~7%. Interestingly, despite variable ( 22 Ne/ 21 Ne) cos ratios for the different aliquots of the same meteorite, the deduced CRE ages usually agree within the uncertainties (see below).
Argon: The Ar isotopic concentrations are given in Table 2. The 40 Ar/ 36 Ar ratios range from 0.0153 for one of the DaG 340 samples to 252.5 for DaG 084. Excluding the exceptionally high value for DaG 084, the measured range for 40 Ar/ 36 Ar reduces to 0.0153-6.75; the average ratio is 2.34 AE 2.56. The lowest 40 Ar/ 36 Ar ratio measured so far in ureilites is (2.9 9 1.7) 9 10 À4 (G€ obel et al. 1978). Considering that ureilites have K concentrations in the range 15 ppm (e.g., Scherer et al. 1998), less than 120 9 10 À8 cm 3 STP g À1 radiogenic 40 K can have been produced over the last 4.5 Ga. Consequently, only for RaS 247 (stepwise heating) and DaG 340-a can the measured 40 Ar fully be explained by radiogenic contributions; the higher concentrations measured for all other samples must be due to atmospheric contamination. Note that most of the studied ureilites are from hot deserts. Correcting the Ar data for atmospheric contamination by assuming all 40 Ar to be atmospheric gives 38 Ar/ 36 Ar ratios in the range 0.177-0.203, clearly indicating a primordial rather than a cosmogenic origin. The data indicate that ( 38 Ar/ 36 Ar) ureilite must be lower than 0.177, which is the lowest ratio determined by us. However, since this Ar fraction must contain a cosmogenic contribution with a 38 Ar/ 36 Ar ratios close to 1.5, the ureilitic 38 Ar/ 36 Ar ratio must be even lower than 0.177. Note that it is usually assumed that ( 38 Ar/ 36 Ar) ureilite = 0.190 (e.g., G€ obel et al. 1978;Busemann et al. 2000).

CHEMICAL COMPOSITION AND PRODUCTION RATE SYSTEMATICS
The production rate of a cosmogenic nuclide, e.g., 21 Ne c , depends on the radius of the meteoroid in space, the depth of the sample within the preatmospheric object (radius and depth are usually merged together into the term shielding depth), the geometry of the meteoroid in space, the chemical composition of the meteoroid, and the chemical composition of the studied sample (e.g., Leya and Masarik 2009). Since the shielding conditions are very often not known most studies rely on correlations as the one discussed further below.
We first discuss the bulk chemical composition. We collected bulk chemistry data of 36 samples from 26 ureilites from various literature sources. The results are compiled in Table 3 where we also give the average chemical composition used for the model calculations (see below). The overall relative deviation from the mean is modest, with the relative 1r standard error of the mean for major elements (those with ≥10% of the total mass, i.e., O, Mg, Si, and Fe) ranging between 1.1% and 3.6%. For minor elements (those with ≤10% of the total mass, i.e., Na, Al, S, Ca, K, Ti, Mn, and Ni), the relative 1r standard error of the mean ranges between 1.8% and 16.3%. The result compiled by us is in excellent agreement with the average elemental composition given by Rai et al. (2003), which is based on nine ureilites.
Based on the determined average chemical composition, we calculated 21 Ne and 22 Ne production rates as a function of preatmospheric radius and shielding depth using the model from Leya and Masarik (2009). Briefly, the model is based on the spectra of primary and secondary particles calculated using Monte-Carlo techniques and the cross sections for all relevant nuclear reactions. The current version of the model only distinguishes particle spectra for ordinary chondrites, carbonaceous chondrites, and iron meteorites. This, however, is not a serious limitation. The particle spectra depend, beside radius and shielding depth, on the bulk chemical composition of the entire meteoroid but not, or only very little, on the chemical composition of an individual sample. As an example, for calculating the cosmogenic nuclide production rates for a metal inclusion in an ordinary chondrite one has to use the particle spectra for ordinary chondrites but considering only nuclear reactions relevant for this metal inclusion. Therefore, the particle spectra for ordinary chondrites not only cover bulk samples but also all inclusions and/or mineral separates in ordinary chondrites. In addition, the dependence of the particle spectra on the bulk chemical composition of a meteorite is relatively minor; i.e., the particle spectra for H-, L-, and LL-chondrites are essentially indistinguishable.
Here we use the model for carbonaceous chondrites because carbonaceous chondrites are a better match for ureilites than ordinary chondrites due to their more similar metal and carbon content. Since the production rates depend on preatmospheric radius of the meteoroid and shielding depth of the sample, two parameters usually not known, we cannot use the modeled results directly but rather rely on production rate systematics. We therefore calculated 21 Ne production rates, P21, as a function of ( 22 Ne/ 21 Ne) cos . The production rates have been calculated using the particle spectra for all shielding depths in carbonaceous chondrites with preatmospheric radii less or equal 100 cm and using the average chemical composition given in Table 3. The modeled results indicate that small meteorites have high ( 22 Ne/ 21 Ne) cos ratios and relatively low P21 values in the range 0.2 9 10 À8 cm 3 STP g À1 Ma À1 . The larger the meteorite, the smaller the ( 22 Ne/ 21 Ne) cos ratios and the larger P21. Plotting the data for all shielding depths gives a correlation close to linear. As it is very often in CRE age studies, we not try to calculate preatmospheric radii for all studied ureilites and the shielding depths for each studied sample because this is very often not possible with the limited data available. We instead use the modeled data and calculated a linear fit of P21 values as a function of ( 22 Ne/ 21 Ne) cos . In addition to the fitted linear correlation, we also show in Fig. 4 estimated uncertainties for the modeled P21, which we assume is 15%. From Fig. 4, three important points must be emphasized. First, the linear correlation cannot be used outside the range of ( 22 Ne/ 21 Ne) cos ratios given by the model, which is 1.016-1.235. Second, the model clearly demonstrates that P21 depends on shielding. For example, the ( 22 Ne/ 21 Ne) cos ratios for all samples studied here range from 1.017 to 1.231, which gives P21 values in the range 0.171 9 10 À8 cm 3 STP g À1 Ma À1 and 0.494 9 10 À8 cm 3 STP g À1 Ma À1 , i.e., they vary by almost a factor of three. The ( 22 Ne/ 21 Ne) cos ratios in our data set vary by more than 20%, which is more than the 10% variation in the data set by Beard and Swindle (2017). These authors argued that the variation in ( 22 Ne/ 21 Ne) cos ratios results in a 15% variation of the 21 Ne cos production rates. In contrast, the model used here predicts for 10% ( 22 Ne/ 21 Ne) cos variations 21 Ne cos production rate variations of almost 45%, i.e., slightly higher. In the other study of ureilite CRE ages, Herzog and Caffee (2014) used a constant 21 Ne cos production rate of 0.421 9 10 À8 cm 3 STP g À1 Ma À1 . While this value is well within the range predicted by the model calculations, it is slightly biased to larger preatmospheric radii. Averaging all 21 Ne cos production rates used by us gives a value of 0.316 9 10 À8 cm 3 STP g À1 Ma À1 with a standard deviation of about 25%, i.e., slightly lower that the value used by Herzog and Caffee (2014). Anyway, from the model predictions, we conclude that 21 Ne cos production rates are shielding dependent and that this shielding dependency must be taken into account. Third, the established correlation between P21 and ( 22 Ne/ 21 Ne) cos only slightly depends on the chemical composition. The  Table 3 corresponds to a Fo number of 78, which is at the lower edge of Fo numbers for ureilites (ranging from 75 to 92). However, reducing the Fe content to a Fo number of 92 increases the production rates only bỹ 20%. Since most of the studied ureilites have Fo numbers in the relatively narrow range from 75 to 80, we conclude that using an average chemical composition with Fo number 78 is well applicable for calculating the 21 Ne cos production rates for ureilites. Note that Downes et al. (2008) also argued that ureilitic olivine clast with magnesium numbers less than 85 are much more common than those with magnesium numbers larger than 85.

RESULTS-COSMIC RAY EXPOSURE AGES New Measured Data
The ( 22 Ne/ 21 Ne) cos ratios after the component deconvolution are summarized in Table 1. As discussed before, the model predicts for ureilites with preatmospheric radii lower than or equal to 100 cm and having average chemical composition (Table 3) ( 22 Ne/ 21 Ne) cos ratios between 1.016 and 1.235. Ratios higher or lower than this can be due to (1) a wrong assumption of the trapped component(s) used for the component deconvolution, (2) a wrong assumption of the chemical composition used for modeling, (3) errors in the assumed ( 20 Ne/ 22 Ne) cos ratio used for the component deconvolution, or (4) errors in the modeling calculations. From the 10 studied ureilites, eight have-after the component deconvolution using Ne ureilite as the endmember-( 22 Ne/ 21 Ne) cos ratios well within the range defined by the model; only the three samples from DaG 340 and JaH 422 have higher ratios. Changing the trapped component does not solve the problem with the model disagreement. If we assume for the component deconvolution that measured Ne is a mixture of atmospheric contamination and cosmogenic Ne instead of a mixture of Ne ureilite and cosmogenic Ne, the ( 22 Ne/ 21 Ne) cos ratios for JaH 422 are still higher than the upper limit given by the model. For DaG 340 the situation is different. The ( 22 Ne/ 21 Ne) cos ratios calculated by using Ne ureilite as a trapped endmember agree to within better than 3% for the three samples, as it is expected considering that all three samples are from the same piece. However, by using Ne air for the component deconvolution, we calculate ( 22 Ne/ 21 Ne) cos ratios of 1.024, 1.198, and 1.228 for the three DaG 340 samples. While the ratios are now within the range defined by the model, the variation of~20% strongly argues against this procedure. Interestingly, the stepwise heating data for JaH 422 and DaG 340 indicate that both meteorites contain solar cosmic ray (SCR) produced noble gases. The data in the Ne-three-isotope diagram are all close to the cosmogenic endmember but plot distinctly below and to the left of the Ne cos /Ne ureilite and Ne cos /Ne air envelopes (Fig. 3). The existence of solar type gases in DaG 340 has already been demonstrated by Scherer et al. (1998). Solar wind implanted noble gases are usually seen in polymict ureilites, distinguishing them as regolith breccias (e.g., Bischoff et al. 2006;Goodrich et al. 2015). In contrast, JaH 422 and DaG 340 are both classified as main group ureilites. However, DaG 340 might be a clast from a polymict ureilite, considering that many DaG ureilites are polymict.
The CRE ages are determined using the calculated P21 ( 22 Ne/ 21 Ne) cos , and 21 Ne cos values are compiled in Table 4. The ages range from 6.9 Ma for JaH 424 to 36.3 Ma for one of the Kenna samples. The reproducibility of the CRE ages for the two RaS 247 samples is~10%. Note that this good reproducibility is surprising. The Meteoritical Bulletin lists for RaS 247 a mass of 579 g, which corresponds to a sphere with a radius of~4 cm. However, in such a small object, we do not expect ( 22 Ne/ 21 Ne) cos or 21 Ne cos variations in the range 7% or 80%, respectively, as measured by us. With a ( 22 Ne/ 21 Ne) cos ratio as high as 1.202 the preatmospheric radius must have been in the range 10 cm, whereas ( 22 Ne/ 21 Ne) cos = 1.115 requires a preatmospheric radius larger than 20 cm. For the following discussion, we use for RaS 247, an (average) CRE age of 24.9 AE 1.2 Ma. Fig. 4. Modeled 21 Ne production rates as a function of ( 22 Ne/ 21 Ne) cos ratios for ureilites with preatmospheric radii less or equal 100 cm. For the model calculations, we use the particle spectra for carbonaceous chondrites but used the average chemical composition for ureilites. There is a linear dependence of 21 Ne production rates as a function of ( 22 Ne/ 21 Ne) cos ratios.
The reproducibility of the seven Kenna samples is not as good as one would like it to be. The ages range between 25.6 and 36.3 Ma, i.e., they vary by~40%. Since the variation in cosmogenic noble gas concentrations is significantly smaller than the variation in CRE ages, i.e.,~20%, we conclude that a reason for the large scatter is in the highly variable ( 22 Ne/ 21 Ne) cos ratios used to correct for shielding. All our samples come from the same piece of Kenna, i.e., they were less than a few mm apart. We therefore expect very similar shielding conditions and consequently very similar ( 22 Ne/ 21 Ne) cos ratios and 21 Ne cos concentrations. Indeed, the 21 Ne cos concentrations vary only by~20% but the ( 22 Ne/ 21 Ne) cos ratios vary by~6%, which is substantial and would indicate significant differences in shielding. To be more quantitative, the 6% variation in ( 22 Ne/ 21 Ne) cos results in a more than 40% variation in 21 Ne cos production rates, which is the reason for the highly variable CRE ages. However, while the reproducibility for the seven samples studied by us is at least reasonable, the CRE ages calculate from the data reported by Okazaki et al. (2003), Rai et al. (2003), and Wilkening and Marti (1976) range from 32.4 to 52.8 Ma, i.e., they vary by more than 60%. By combining all data, the CRE ages for Kenna range from 25.6 to 52.8 Ma, i.e., they vary by more than a factor of 2. Consequently, for reasons not yet understood, the reproducibly of the 11 studied Kenna samples is not satisfying. One might speculate that sample-to-sample heterogeneities might be responsible but it is hard to envision that variations in chemical composition can change the 21 Ne cos production rates by a factor of two. Note that the measured 21 Ne cos concentrations of all studied Kenna samples vary bỹ 70%, which is substantial but which is slightly less than the variations in CRE ages after correcting for shielding using ( 22 Ne/ 21 Ne) cos . Therefore, and as already discussed before, some scatter is introduced by the shielding correction because ( 22 Ne/ 21 Ne) cos ratios vary by 6% for the samples studied by us and by about 7% if we also consider the literature data. For the following discussion, we assume for Kenna-a, a CRE age of 35.6 AE 7.4 Ma, which is the average of all 11 individual data.
The CRE age obtained by us for DaG 084 of 16.6 Ma is in reasonable agreement with the CRE age of 18.5 Ma calculated using the data given by Scherer et al. (2000). For the following discussion, we use the average age of 17.5 AE 1.1 Ma. For DaG 319, we can compare our data to the results from Scherer et al. (1998). However, while they found abundant solar type gases, our Ne data are consistent with being a mixture of cosmogenic and Ne ureilite , though solar type gases cannot be excluded. The calculated ages are 17.6 Ma for our sample and 21.2 Ma for the samples from Scherer et al. (1998); the grand average value used for further discussion is 19.4 AE 1.8 Ma.

Cosmic Ray Exposure Ages for Ureilites-Easy to Interpret
For discussing literature data, we focus entirely on Ne isotopes because (1) 3 He cos data are unreliable due to common but highly variable 3 He cos deficits in most ureilites (see above) and (2) there are not enough data to apply other dating systems like, e.g., 36 Cl-36 Ar in metal or 81 Kr-Kr in bulk samples. By combining our results with literature data, we established a data set comprising 100 data for 41 ureilites, which is the current most complete database for cosmogenic Ne and CRE ages for ureilites. Even more important, our database is the only database that fully considers the shielding dependency of the 21 Ne production rates and therefore gives reliable 21 Ne CRE ages. The cosmogenic Ne composition for all samples was determined using a two-component deconvolution with the following endmembers: ( 20 Ne/ 22 Ne) cos = 0.8 (Eugster and Michel 1995), ( 20 Ne/ 22 Ne) air = 9.78 and ( 21 Ne/ 22 Ne) air = 0.029 (Meija et al. 2016), ( 20 Ne/ 22 Ne) ureilite = 0.4 and ( 21 Ne/ 22 Ne) ureilite = 0.027 (Ott et al. 1985), and ( 20 Ne/ 22 Ne) solar = 13.75 and ( 21 Ne/ 22 Ne) solar = 0.0333 (Grimberg et al. 2008). Of all analyzed 124 samples, 100 have Ne data that can easily be interpreted as being a mixture of Ne cos with either Ne air , Ne solar , or Ne ureilite . In addition, the calculated ( 22 Ne/ 21 Ne) cos ratios for the sample lie within the range allowed by the model predictions, i.e., in the range 1.016-1.235 (see above). The data for 100 samples from 41 ureilites are compiled in Table 4. There we also give the calculated 21 Ne production rate and the determined CRE age for each sample. For meteorites where there are more data, we also give the average CRE age together with the standard error of the mean.
Instead of a proper error propagation, which is difficult because sometimes no uncertainties are given for the literature data and error propagation for the component deconvolution is difficult, we assume that the total uncertainty for the CRE ages is entirely dominated by the uncertainties of the model calculations, which is 15% (see also Fig. 4). We therefore assign an uncertainty of 15% to the individual CRE ages. Note that for establishing a CRE age histogram, the individual uncertainties are not that relevant because the bin sizes are given by statistical constraints and they are usually larger than the uncertainties.

Literature Data-Difficult to Interpret
For some samples from some ureilites, the ( 22 Ne/ 21 Ne) cos ratios are outside the range allowed by the   Yamamoto et al. (1998) for ALHA77257, Asuka-881931, and Yamato-790981 are not considered because the maximum release temperature of 900°C used in their study is not sufficient to completely degas ureilite samples and therefore CRE ages determined using their data would be too low. Note that all three ureilites were included in the study by Beard and Swindle (2017). For ALHA78019, the samples measured by Wacker (1986) and Rai et al. (2003) plot outside the Ne cos /Ne ureilite and Ne cos /Ne air envelopes and can therefore not be interpreted as simple two-component mixtures. Sample (b) from Wacker (1986) plots in between the Ne cos /Ne ureilite and Ne cos /Ne air envelopes, suggesting a three-component mixture, while sample (a) plots significantly under both envelopes. The sample reported by Rai et al. (2003) also plots below the Ne cos / Ne air envelope and appears similar to sample (a) studied by Wacker (1986). ALHA78019 likely contains SCR produced Ne but additional data are needed. Wacker (1985) only performed stepwise heating on sample (b) but not on the anomalous sample (a), while the stepwise heating data reported by Rai et al. (2003) are only for temperatures 1000-1600°C and do not reveal any information about the nontrapped component. For DaG 340, two of the three measured ( 22 Ne/ 21 Ne) cos ratios are slightly higher than allowed by the model, therefore no age can be calculated from them. Only the data obtained from the stepwise heating experiment fall into the allowed range and can be used to determine the CRE age. As a possible explanation why the stepwise heating data seem to be more reliable than the other two data, we might speculate that the stepwise setup possibly better reduces the trapped component, which for DaG 340 is likely solar wind, and therefore provide more reliable cosmogenic data. Rai et al. (2004) analyzed bulk samples and HF/ HCl residues for FRO 90036. However, their ( 22 Ne/ 21 Ne) cos ratios after component deconvolution with either Ne solar or Ne ureilite as an endmember is too high, i.e., 2.446 or 1.452, respectively, and therefore no CRE age can be calculated. For JaH 422, all of our ( 22 Ne/ 21 Ne) cos ratios are higher than allowed by the model predictions, clearly indicating the small preatmospheric size of this ureilite and likely indicating that SCRs contribute to the measured Ne budget, which makes the calculation of a CRE age impossible. Note that the small preatmospheric size of JaH 422 is also indicated by its low mass of only 61.6 g (Meteoritical Bulletin). The difficulties for the Kenna data have been discussed before. For the following discussion, we assume for Kenna-a, a CRE age of 35.6 AE 7.4 Ma, which is the average of the seven data measured by us and the four literature data.
The sample of Lahrauli reported by Padia et al. (1983) appears to be almost exclusively cosmogenic, though its ( 22 Ne/ 21 Ne) cos ratio of 1.245 is slightly higher Abundances are in 10 À8 cm 3 STP g À1 and production rates are in 10 À8 cm 3 STP g À1 Ma À1 . The uncertainties for the 21 Ne cos production rates and therefore for the individual CRE ages are estimated to be~15%. The given uncertainties for the average values are the 1r standard deviation of the mean. DEM = deconvolution endmember.
than the upper limit allowed by the model. In addition, the data reported for this meteorite by Rai et al. (2003) plot below the Ne cos /Ne ureilite and the Ne cos /Ne air envelopes, suggesting the presence of SCR produced Ne. This is further confirmed by the high apparent ( 22 Ne/ 21 Ne) cos ratio of 1.311. Consequently, no age can be calculated for Lahrauli. Beard and Swindle (2017) included Lahrauli in their database despite the possible contributions of SCR Ne. For LEW 85328, only the data by Ott et al. (1993) can be used to determine a CRE age. The ( 22 Ne/ 21 Ne) cos ratio determined from the data by Rai et al. (2003) of 1.327 is higher than allowed by the model predictions.
The sample of Novo-Urei reported by M€ uller and Z€ ahringer (1969) has a ( 22 Ne/ 21 Ne) cos ratio outside the range allowed by the model predictions. This sample plots well under the Ne cos /Ne ureilite and Ne cos /Ne air envelopes in the Ne-three-isotope diagram. The other four samples of Novo-Urei (Stauffer 1961;Mazor et al. 1970;G€ obel et al. 1978;Aylmer et al. 1990) can well be explained as simple mixing between Ne cos and Ne ureilite and all have consistent CRE ages. Therefore, either some error was made while measuring the M€ uller and Z€ ahringer (1969) sample or that particular sample somehow differed significantly from the other three. For Yamato-74123, all measured samples (Hintenberger et al. 1978;Ott, personal communication;Aylmer et al. 1990) give ( 22 Ne/ 21 Ne) cos ratios higher than allowed by the model; consequently no CRE age can be calculated. The ( 22 Ne/ 21 Ne) cos ratio of 1.284 for EET 83225 determined from the data by Park et al. (2014) is also higher than the range predicted by the model calculations, therefore no CRE age can be calculated. Finally, the data for North Haig (e.g., Bogard et al. 1973) indicate significant 21 Ne cos losses and can therefore not be used to calculate a CRE age (Aylmer et al. 1990). Note that both EET 83225 and North Haig were included in the study by Beard and Swindle (2017).

COSMIC RAY EXPOSURE HISTOGRAM
The CRE ages calculated here are based on our current best knowledge on shielding corrected 21 Ne production rates. In this respect, our study differs from the two earlier approaches by Herzog and Caffee (2014) and Park et al. (2014), both of which used a constant and therefore shielding independent 21 Ne production rate of 0.412 9 10 À8 cm 3 STP g À1 Ma À1 . Beard and Swindle (2017) also used a constant 21 Ne production rate, which was calculated from the data for the lunar surface (Hohenberg et al. 1978) multiplied by a factor of two to account for a 4p exposure geometry. The 21 Ne production rates used by us vary between 0.171 9 10 À8 cm 3 STP g À1 Ma À1 for one Yamato-74123 sample and 0.493 9 10 À8 cm 3 STP g À1 Ma À1 for one sample of ALHA78019, i.e., the 21 Ne production rates vary by a factor of almost three. The average of all 21 Ne production rates used by us is 0.316 9 10 À8 cm 3 STP g À1 Ma À1 , i.e., about 30% lower than the (average) production rate used by Herzog and Caffee (2014) and Park et al. (2014).
Considering the data for Kenna, for which the shielding correction increased the scatter of the CRE ages (see above), it is reasonable to ask if the shielding correction is necessary and if so if it is reliable. Considering the first point (necessity), we argue that using a constant 21 Ne cos production rate might distort the CRE age histogram. In contrast, it is obvious that using shielding corrected 21 Ne cos production rates, if properly done, is beneficial and cannot have any negative effects. Therefore, if possible, a shielding correction should always be performed. Considering the second point (reliability), we checked the data by investigating if there is a correlation between CRE ages and 21 Ne production rates (Table 4). The argument is as follows, a false correction for shielding can produce too low or too high 21 Ne production rates for certain ( 22 Ne/ 21 Ne) cos ratios, i.e., for certain shielding conditions. In such a case, there could be a (false) correlation between 21 Ne production rates and ( 22 Ne/ 21 Ne) cos ratios. We find no correlation between 21 Ne production rates and CRE ages, clearly indicating the reliability of our data compilation.
The histogram of 21 Ne CRE ages for both updated literature data and our newly reported data are shown in the upper panel of Fig. 5. The CRE ages for ALHA82106 and ALHA82130 of 3.3 AE 0.5 and 2.7 AE 0.5 Ma confirm the pairing of the ALH82xxx ureilites (e.g., Mason 1985;Takeda et al. 1988). Since there are no indications for further paring, we assume that our database contains 40 individual ureilite falls. Based on the CRE age data available at that time, Goodrich (1992) and Scherer et al. (1998) stated that there was no apparent clustering of CRE ages. Later and based on a larger data set, Herzog and Caffee (2014) argued for a broad maximum between 9 and 25 Ma (~40% of all ureilites). According to them, there is also a possible peak near 10 Ma and they concluded that at least three events produced the known ureilites (Fig. 5, middle panel). By adding 13 new CRE ages to the database, Park et al. (2014) essentially confirmed the findings by Herzog and Caffee (2014). In contrast, Beard and Swindle (2017) (Fig. 5 upper panel) indicates that 11 out of the 40 studied ureilites have 21 Ne CRE ages of less than or equal to 5.8 Ma; five ureilites have unusually low CRE ages of even less than 2 Ma (Acfer 277, ALHA78019, Hajmah, HaH 064, RC 027). ALHA78019 has a CRE age of 560 kyr, which is in the range typical for CM chondrites and lunar meteorites (cf, Herzog and Caffee 2014). The longest CRE age is 45 Ma (EET 83309). Interpreting the CRE age histogram is not without problems. With a spread of almost 45 Ma for the CRE ages and a bin size for the histogram of 2.5 Ma, there are on average 2.2 meteorites per bin, which is not sufficient to clearly resolve detailed structures. From the limited data, we can conclude that there is no apparent clustering in the CRE age histogram, except for a slight indication of a peak in the range 30 Ma. There is a trend, however, of decreasing meteorite number with increasing CRE age, the trend is roughly exponential. There are more ureilites with short CRE ages than there are ureilites with long exposure ages. This is also obvious considering that the range of CRE ages is~44 Ma but the median of the data is~12.5 Ma. One could argue, however, that there is a peak in the range 30 Ma, which will be discussed further below.
Also the two histograms by Herzog and Caffee (2014) (Fig. 5 middle panel) and Beard and Swindle (2017) (Fig. 5 lower panel) show a decrease in the CRE age histogram. However, while the CRE age histogram by Herzog and Caffee (2014) is somewhat uncertain due to the relatively low number of meteorites, the CRE age distribution by Beard and Swindle (2017) seems steeper than the one established by us and the one published by Herzog and Caffee (2014). The differences are likely due to the choice of the used 21 Ne production rates. By way of example, the longest CRE age determined by us is for EET 83309. The 21 Ne production rate used by us for this meteorite is 0.347 9 10 À8 cm 3 STP g À1 Ma À1 , which is 20% lower than the 0.421 9 10 À8 cm 3 STP g À1 Ma À1 used by Herzog and Caffee (2014) and therefore results in a 20% higher CRE age. Another interesting example is for one of the Haver€ o data (Smith and Fireman 1973). From the ( 22 Ne/ 21 Ne) cos ratio of 1.223, we calculate a 21 Ne cos production rate of 0.183, i.e., more than a factor of two lower than the production rate used by Herzog and Caffee (2014). Consequently, while we calculate with our approach for Haver€ o a CRE age of 43.5 Ma (Table 4), we would calculate a CRE age of about 19 Ma using the 21 Ne cos production rate from Herzog and Caffee (2014). Since most of the production rates used by us are slightly lower than the average production rates used in the two other studies, we calculate on average slightly higher CRE ages, which naturally produces a less steep decrease in the CRE age distribution.
For the (roughly) exponential behavior of the CRE age histogram, we can find two possible explanations. First, it might be that meteorite delivery from the UDBs is actually increasing, i.e., that currently more ureilites are produced and delivered to Earth than in the past. This, in combination with a constant destruction rate of the ureilites, would naturally produce a decreasing behavior in the CRE age histogram. Higher meteorite delivery rates can be due to higher collision rates inside the asteroid belt or at least in the orbital  Herzog and Caffee (2014), and the lower panel shows the results from Beard and Swindle (2017). All three CRE age histograms show a decreasing number of ureilites with increasing CRE age. region of the UDB. We consider higher collision rates unlikely because this would also be seen in other meteorite groups. It could be possible, however, that the orbit of the UDB somehow differ from the orbits of the other meteorite parent bodies and that higher collision rates only occur in this special region of the asteroid belt. Another possibility is that the collisions that produce meteoroids are now taking place at locations closer to a resonance. A UDB that slowly moves closer to an effective resonance while suffering impacts at a constant rate would naturally produce meteorites with shorter and shorter CRE ages and therefore a CRE age histogram as seen in Fig. 5. It has been speculated that the UDB is located in the Nysa-Polana region, which is within about 0.1 AU of the 3:1 resonance (e.g., Gayon-Markt et al. 2012). If true, the UDB is indeed, at least currently, located close to a very effective resonance.
A second possibility is that meteorite delivery from the UDB is temporarily constant but that some type of loss mechanism reduces the lifetime of the ureilites. As loss mechanisms one can imagine erosion, collisions, or loss of ureilites either into the Sun or onto other solar system planets. A similar scenario has been proposed by Takeda et al. (1988), who argued that ureilites were created early in the solar system and that rapid erosion in space accounts for the absence of ureilites with long exposure ages. If erosion were the main loss mechanism, one would expect that all ureilites are affected to the same or at least similar degrees. By assuming constant erosion rates, meteorites with short CRE ages would have preatmospheric radii very close to their radii at ejection from the UDB because erosion has not yet reduced their dimensions. On the other hand, meteorites having a large preatmospheric radius must have short CRE ages because they are still large. However, the data indicate differently. The large ureilites ALHA81101, DaG 084, and Almahata Sitta with ( 22 Ne/ 21 Ne) cos ratios lower than or equal to 1.067 all have CRE ages in the range 10-20 Ma, i.e., not exceptionally short. Though, we cannot exclude that some of the large ureilites from our collections were produced even much larger and entered Earth atmosphere still large, i.e., erosion indeed reduced the size of the original meteoroid but even after CRE the meteoroid was still large, we would nevertheless expect some kind of relationship between CRE age and preatmospheric radius. We therefore conclude that the data contradict the hypothesis that erosion is the main loss mechanism. This result confirms earlier studies on erosion rates (e.g., Schaeffer et al. 1981). If meteorite breakup would be the most relevant loss mechanism, we can essentially follow the same arguments as before for the case of erosion; large objects must have short CRE ages because they have not yet been broken into smaller pieces. Since this is not what is seen in the data, we can exclude breakup as a relevant loss mechanism shaping the CRE age histogram of ureilites. On the other hand, a meteorite loss that is independent on meteorite size, i.e., small and large ureilites are lost at a very similar rate, into the Sun, onto other solar system planets or out of the solar system would naturally produce a CRE age histogram as seen in Fig 5. To summarize, a likely explanation for the overall structure of the CRE age histogram for ureilites, i.e., the decreasing number of ureilites with increasing CRE age, is to imagine either that the UDB actually moves closer to the meteorite delivering resonance and/or that there is an effective type of loss mechanism. Either effects, or a combination of the two, would naturally produce more ureilites with short CRE ages, exactly what is seen in Fig. 5.
Using our model predictions tailored for ureilites and assuming~85% atmospheric ablation losses, we find that ( 22 Ne/ 21 Ne) cos ratios higher or equal to 1.10 most likely indicate preatmospheric radii less or equal 20 cm, which can be considered as small. With the database from Table 4, we find that 28 out of the 40 studied ureilites, i.e., almost 70%, had preatmospheric radii less than 20 cm. This finding confirms earlier less quantitative results by Aylmer et al. (1990) and Park et al. (2014). A possible explanation as to why a large number of ureilites, at least from the ones studied by us, have small preatmospheric radii is that they are all derived from the UDB, which is composed of reaccreted material from the original ureilite parent body (e.g., Downes et al. 2008;Herrin et al. 2010aHerrin et al. , 2010bGoodrich et al. 2015). Impacts on such a brecciated asteroid would naturally produce small meteoroids, most likely in the size range of the clasts in the breccia. Note that the largest sizes of clasts from the Almahata Sitta fall were all in the range 200-300 g (e.g., Herrin et al. 2010b), which corresponds to a radius of about 3 cm, i.e., even smaller than the radii calculated by us.
The CRE age histogram showing the new data (Fig. 5, upper panel) depicts what might be a peak at 30 Ma. The possible peak consists of the three ureilites Goalpara (28.9 Ma), GRA 98032 (29.2 Ma), and MET 1085 (29.1 Ma). We can also add NWA 5928 and Haver€ o with CRE ages of 26.4 and 32.5 Ma, respectively, to the list. We can now ask ourselves whether the five ureilites have other features in common. The forsterite numbers of Goalpara, GRA 98032, and Haver€ o are all in a narrow range between 78.6 and 79.0, i.e., they indicate a similar petrologic origin of the samples (cf. Beard and Swindle 2017). Unfortunately, there are no data for the other two ureilites MET 1085 and NWA 5928. The conclusion that can be drawn, however, is not very significant because there is a large peak in the distribution of olivine core compositions around Fo 79-81 (Goodrich et al. 2015) and the finding that the five ureilites all have Fo number in that range is therefore not surprising. Although the result is compromised by the low statistic, we suggest that a big impact on the UDB~30 Ma ago produced the three (or five) homogeneous ureilites.

SUMMARY
We measured the noble gas isotopic concentrations of He, Ne, and Ar for 21 samples from the 10 ureilites, DaG 084, DaG 319, DaG 340, Dho 132, HaH 126, JaH 422, JaH 424, Kenna, NWA 5928, and RaS 247, using both single and stepwise heating extractions. Here we report the CRE ages. By also considering literature data, we compiled a consistent set of CRE ages for ureilites. All CRE ages were calculated using production rates for cosmogenic 21 Ne by fully considering their dependence on preatmospheric radius and shielding depth (Leya and Masarik 2009). In this respect, our study is more advanced than earlier studies, which are based on a constant, i.e., shielding independent, 21 Ne production rate (e.g., Herzog and Caffee 2014;Park et al. 2014;Beard and Swindle 2017).
The CRE age histogram is now based on 100 samples from 40 different ureilites. The result is surprising: The number of meteorites decreases with increasing CRE age, i.e., there are more ureilites with short CRE ages than there are ureilites with long CRE ages. As a possible explanation, we speculate that either the parent body of the known ureilites is currently moving closer to a resonance and/or that there is some loss mechanism involved, which affects all ureilites the same, i.e., the loss mechanism is independent on the size of the meteorite. In addition and superimposed to the finding that the number of ureilites is decreasing with increasing CRE age, there is a slight indication for a peak in the range 30 Ma, which contains up to 12% of the studied ureilites. Finally, we confirm earlier results (Aylmer et al. 1990;Park et al. 2014) that the majority of the studied ureilites have relatively small preatmospheric radii, i.e., they have preatmospheric radii less than or equal to~20 cm.