By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Due to essential maintenance the subscribe/renew pages will be unavailable on Wednesday 26 October between 02:00- 08:00 BST/ 09:00 – 15:00 SGT/ 21:00- 03:00 EDT. Apologies for the inconvenience.
Abstract– Miller Range (MIL) 05029 is a slowly cooled melt rock with metal/sulfide depletion and an Ar-Ar age of 4517 ± 11 Ma. Oxygen isotopes and mineral composition indicate that it is an L chondrite impact melt, and a well-equilibrated igneous rock texture with a lack of clasts favors a melt pool over a melt dike as its probable depositional setting. A metallographic cooling rate of approximately 14 °C Ma−1 indicates that the impact occurred at least approximately 20 Ma before the Ar-Ar closure age of 4517 Ma, possibly even shortly after accretion of its parent body. A metal grain with a Widmanstätten-like pattern further substantiates slow cooling. The formation age of MIL 05029 is at least as old as the Ar-Ar age of unshocked L and H chondrites, indicating that endogenous metamorphism on the parent asteroid was still ongoing at the time of impact. Its metallographic cooling rate of approximately 14 °C Ma−1 is similar to that typical for L6 chondrites, suggesting a collisional event on the L chondrite asteroid that produced impact melt at a minimum depth of 5–12 km. The inferred minimum crater diameter of 25–60 km may have shattered the 100–200 km diameter L chondrite asteroid. Therefore, MIL 05029 could record the timing and petrogenetic setting for the observed lack of correlation of cooling rates with metamorphic grades in many L chondrites.
Impact craters are the most obvious surface features on planetary bodies lacking atmospheres and volcanic activity. Thus, it comes as no surprise that impact cratering is one of the fundamental surface processes in the solar system. Even very young surfaces such as those on Earth have numerous impact craters, attesting to the importance of impact cratering even today. Radiometric dating of lunar samples has shown that the impact flux varies with time. Surfaces older than 3.5 Ga (even those just a few 100 Ma older) have many more craters than those a few 100 Ma <3.5 Ga, indicating higher impact rates prior to 3.5 Ga (Strom et al. 2005). Understanding the source and rate of impacts throughout time will allow us to better understand the evolution of the solar system.
Ordinary chondrites (OCs) provide a unique opportunity to study impacts for three reasons:
1 The OC parent bodies were small, so they experienced mild heating and rapid cooling (Trieloff et al. 2003) allowing them to preserve the signature of impacts from the early solar system.
2 The chondrite parent bodies are located in the asteroid belt, so they record the impact history in a location different from the Earth–Moon system.
3 The high abundance of OCs compared to other meteoritic material provides samples from more diverse regions of the parent bodies.
The K-Ar radioisotopic system allows us to investigate the impact history of meteorites. The isotope 39K decays to 40Ar with a half-life of 1.28 Ga, ensuring that measurable quantities of both parent and daughter isotopes can be found in solar system materials of almost any age, provided there is K in the sample. The closure temperature of 200–300 °C for most K-Ar systems is low enough that mild shock-metamorphic overprints do not affect them, whereas strong shock-induced heating, which is more likely to occur on a wider scale in larger impact events, resets the K-Ar age. The details are unique to the specific time/temperature history of each sample. By irradiating a sample with neutrons in a nuclear reactor, a fraction of the 39K is converted to 39Ar, thus 39Ar is a proxy for K. Both isotopes of Ar can then be removed from the sample by stepwise heating, and analyzed with a mass spectrometer. The advantage of the 40Ar-39Ar technique (hereafter referred to as Ar-Ar) is twofold. First, partial resetting can be identified by the pattern of apparent ages determined from the fractions of Ar released at various temperatures, and second, these temperature extractions allow us to determine the diffusion rate(s) of the K-bearing phase(s). The diffusion information can then be used to identify the K-bearing phase(s), and potentially constrain the time–temperature history of the meteorite.
By examining heavily shocked or shock-melted chondrites (i.e., those that have experienced high shock pressures and related heating), we can rule out chondrites that only record solar system formation ages. Thus, we use the Ar-Ar age of highly shocked meteorites to study the collisional evolution of their parent asteroids. Most Ar-Ar ages of heavily shocked OCs are 50 Ma to 1.5 Ga, or 3.5–4 Ga, with very few other ages (Bogard 1995; Kring et al. 1996, 1999; Grier et al. 2004; Swindle et al. 2009).
Antarctic meteorite Miller Range (MIL) 05029 was initially selected for petrologic and chronological examination because it was reported to be of impact origin (Connolly et al. 2007). However, we have considered MIL 05029 at a higher level of detail than other meteorites we have studied because it is a complete melt with no shock features, and has an unusually old Ar-Ar age, making parentage more difficult to determine, as well as a distinction between an impact origin and endogenous heating difficult. As unusual samples often contain keys to understanding the evolution of their parent bodies, this makes unusual meteorites potentially important. Therefore, the modal abundance, mineral chemistry, and formation scenario for MIL 05029, including thermal history, are explored thoroughly.
Description and Methods
The Antarctic Search for Meteorites during the 2005–2006 field season found MIL 05029 in Miller Range, Antarctica (Fig. 1). Its dimensions are between 4.0 cm and 5.5 cm, its mass is 132.7 g, and it is paired with the 8.3 g MIL 05136. MIL 05029 is classified as an L impact melt (Connolly et al. 2007) consisting of large orthopyroxene grains with interstitial feldspar, both of which are poikilitically enclosing olivine grains (Fig. 2).
Microscopy and Microprobe Analysis
The NASA-Johnson Space Center’s Cameca SX-100 microprobe with five wavelength-dispersive spectrometers was used to analyze thin-section MIL 05029,4. Chemical analyses of silicate minerals were obtained using an accelerating voltage of 15 kV, a beam current of 15 nA and a beam diameter of 1 μm. Peak positions and intensities of individual elements were calibrated and monitored with well-characterized standards from the NASA Johnson Space Center probe standard collection, listed in the Appendix.
To avoid volatilization, feldspathic inclusions were analyzed with an accelerating voltage of 15 kV, a reduced beam current of 10 nA, and an enlarged beam diameter of 5 μm. Analyses of Fe-Ni metals and sulfides utilized an accelerating voltage of 15 kV, a beam current of 20 nA, and a beam diameter of 1 μm. Wavelength-dispersive element concentration maps of thin-section MIL 05029,4 were performed on the same instrument. To assure suitable X-ray count rates, the beam current was increased to 40 nA and the relative concentrations of a subset of elements (Si, K, Na, Ca, Fe, Al, S, Cr, Ni, and Mg) were analyzed on the whole thin-section area.
Three approaches were undertaken to determine the modal mineralogical composition of thin-section MIL 05029,4:
1 Point-counting under reflected light with a 50x objective and a step-size of 100 μm (2014 points) was used to distinguish the abundance of metal and troilite with respect to all other phases.
2 Point-counting on backscattered electron images with a magnification of 1500x and step-size of 100 μm (2146 points) resolved the relative abundances of silicate phases and apatite.
3 Wavelength-dispersive element concentration mapping was performed followed by processing with the Nikon “NIS-Elements Documentation” software package, an image analysis program that calculated the absolute proportions of these species over the area of the thin section. To capture the relative proportions of mineral components, appropriate thresholds for detection were set and the area proportions based on the reference area of the analyzed thin section was calculated (Maloy and Treiman 2007). This technique successfully determined the chromite proportion and confirmed the abundance of plagioclase.The main petrographic characteristics of missing lithic clasts, metal-sulfide depletion, main constituent phases, and textural relationships were confirmed with the optical microscopical examination of another thin section of MIL 05029 from the curatorial repository at the NASA Johnson Space Center. Modal compositions were recalculated from vol% to wt% by using the following specific gravities (in g cm−3): plagioclase 2.6, olivine 4.05, low-Ca pyroxene 3.8, high-Ca pyroxene 3.3, chromite 4.64, troilite 4.67, Fe-Ni metal 7.95, feldspathic inclusions 2.5.
Oxygen Isotope Analysis
Oxygen isotopes were analyzed in two whole-rock aliquots of MIL 05029,8. The samples were crushed under ethanol in a boron nitride mortar and pestle. Then, the material was ultrasonicated in dilute HCl, and the magnetic fraction removed with a hand magnet. The nonmagnetic fraction was loaded in a reaction chamber with duplicates of a laboratory working reference (Gore Mountain garnet, U.S. National Museum 107144). The loaded reaction chamber was repeatedly fluorinated with BrF5, including overnight fluorination, until there was no measurable oxygen blank. Finally, samples were heated in the presence of 30 torr of BrF5 with a 30 W Synrad CO2 laser. Oxygen was transferred on a 5A molecular sieve to the dual inlet of a Thermo Fisher Scientific MAT 252 mass spectrometer for analysis of oxygen isotopes at the Carnegie Institution of Washington.
Ar-Ar Experimental Procedures
Three splits of MIL 05029,5 (A—12.194 mg, B—10.571 mg, and C—9.186 mg) were irradiated without cadmium shielding at the U.S. Geological Survey Training Research and Isotope Production, General Atomics (USGS TRIGA) reactor in Denver, CO for 10.7 h, and analyzed using a VG5400 noble gas mass spectrometer at the University of Arizona. Very high 40Ar/39Ar ratios (Table S2) are the result of anticipating an approximately 500 Ma rock, and instead having an approximately 4500 Ma rock. High errors of 37Ar/39Ar ratios resulted because a 5 month wait was followed by chopping the gas into many steps, resulting in 37Ar values only about five times blank levels. Standards of PP-20, K2SO4, and CaF2 were spaced throughout the radiation package to monitor Ar isotope production rates, and allowed us to account for any irregularity of the neutron flux. Further procedural information can be found in Swindle et al. (2009). A J-factor of 2.156 × 10−3 ± 5 × 10−6 was determined by an unweighted average of the analyses of three samples of hornblende (2 of PP-20 [1074 Ma], 1 of Mmhb-1 [523.1 Ma]), with a standard deviation for error. Reactor interference corrections were (39Ar/37Ar)Ca of 7.187 × 10−4 ± 6.2 × 10−6, (36Ar/37Ar)Ca of 4.276 × 10−4 ± 9.24 × 10−5, and (40Ar/39Ar)K of 3.526 × 10−2 ± 3.526 × 10−2. The production rate of 39Ar per gram of K was 1.77 times greater than the production rate of 37Ar per gram of Ca, as determined by K2SO4 and CaF2. Decay constants were taken from Steiger and Jager (1977). After analysis, data were corrected for blanks, decay, reactor interference, and mass discrimination. A correction for trapped Ar could not be performed as the measured 38Ar/36Ar ratio of most steps was >1.5, indicating that Cl absorbed thermal neutrons and produced 38Ar during irradiation. To provide an upper limit for the amount of trapped Ar, spallation was “removed” by assuming the steps with the lowest 36Ar/37Ar contained only spallation at those isotopes. Spallation can then be “removed” by taking that ratio out of all temperature steps. This technique is described in further detail by Swindle et al. (2009).
Miller Range 05029 at first glance seems comparable to Patuxent Range (PAT) 91501, a clast-poor L chondrite impact melt rock (McBride et al. 2006), leading us to consider whether they might be part of the same impact. Our thin-section analysis of MIL 05029 showed that, on the contrary, PAT 91501 and MIL 05029 are quite different. Although both meteorites exhibit a medium-grained poikilitic texture, in contrast to PAT 91501, MIL 05029 has no vesicles, no relic clasts, no glassy melt, no shock features, metal grains with Widmanstätten-like patterns, metal depletion with plagioclase enrichment, and adjacent pyroxene crystals have 120° junctions. Hence, the formation of MIL 05029 differs from that of PAT 91501 (Mittlefehldt and Lindstrom 2001; Benedix et al. 2008).
The oxygen isotopic signatures of MIL 05029 (Table 1) are mostly consistent with those of L chondrites (Clayton et al. 1991). The large mismatch of δ18O between L chondrite values and sample MIL 05029-1 is likely due to the sensitivity of whole-rock samples to the modal abundance of olivine versus pyroxene versus plagioclase, the grain sizes of components, and the small sample sizes; however, Δ17O is not as sensitive to these variations and the MIL 05029 Δ17O values agree well with those of L chondrites.
Table 1. Oxygen isotope data of MIL 05029,8.
Note: Errors (in parentheses) are 1σ. Variation of chondrite groups [in brackets] are also 1σ.
The meteorite has a medium-grained, fairly well equilibrated igneous texture of mostly euhedral crystals, which tend to exhibit 120° junctions between grain contacts (e.g., between olivine crystals and pyroxene and plagioclase crystals). Macroscopically, the rock exhibits mm-size, dark green crystals set in a yellowish-green matrix. Several opaque nodules (up to ∼2.5 mm size) occur that are spaced >1 cm apart (Fig. 1). Microscopically, subhedral-to-anhedral low-Ca pyroxene and high-Ca pyroxene poikilitically enclose small, euhedral grains of olivine. Interstitial spaces are filled with plagioclase, chromite, nickel-iron metal, and troilite that are poikilitically intergrown with small euhedral olivines (Fig. 2). Minor chlorapatite occurs as anhedral, mm-size grains. Silicate phases and apatite occasionally exhibit <5–∼20 μm diameter round to rectangular inclusions of mostly birefringent phases. In olivine, these inclusions tend to be round, whereas in pyroxenes, these inclusions tend to be rectangular and aligned along crystallographic planes. Frequently, such inclusions contain <1 μm opaque phases as minor components. Mineral phases in MIL 05029 do not exhibit a distinct shape-preferred orientation or alignment. No brecciation or evidence for shock metamorphism is apparent; no lithic clasts were detected; and under the optical microscope, the major constituents show no compositional zonation. A network of cracks overprints the rock’s fabric. Orange-brown oxidation fronts (alteration) are visible along grain boundaries near the outer edge of the meteorite, confirming the designation of weathering grade A-BE (Connolly et al. 2007).
Miller Range 05029,4 appears strongly depleted in metal and troilite, depleted in high-Ca pyroxene, and enriched in plagioclase compared to the average normative mineralogy of L chondrites (Table 2). Apatite, chromite, low-Ca pyroxene, and olivine exhibit concentrations that agree within error with the average normative mineralogy of L chondrites.
Table 2. Modal abundance of MIL 05029,4.
Modal abundances of silicate fractionsa
MIL 05029b (vol%)
MIL 05029c (wt%)
MIL 05029 (wt%)
Note: n.d. = none detected; n.a. = not applicable.
aNormalized nonsilicate free.
bNormalized to 100 vol%, excluding 2.7 vol% cracks and voids.
cvol% were converted to wt% using the specific gravities given in the text.
Mason (1965) notes that modal pyroxene may be 1–2% higher than the normative amount, while the normative calculations by Mason (1965) and McSween et al. (1991) predict a lower proportion of olivine and a higher proportion of augite than is seen in MIL 05029, which may be attributable to natural variation. For example, in Table 1 of McSween et al. (1991), normative olivine contents in 34 L6 chondrites range from 38.8% to 53.1%. However, the maximum value for plagioclase proportions in the 34 L6 chondrites listed by McSween et al. (1991) is 10.55% and the lowest proportions for nickel-iron metal and troilite are 5.35% and 3.88%, respectively.
Low-Ca pyroxene grains (En73.1–77.1, Fs20.3–23.1, Wo1.9–6.6; n = 92 enstatite analyses and n = 1 pigeonite analysis; Table 3) are typically about 3 mm in size and aggregate grains reach sizes over 5 mm. Some lamellar-to-platelet-like exsolution of high-Ca pyroxene occurs in the central regions of larger grains (Fig. 2B). The exsolved regions are typically <1 μm wide and, thus, too small for microprobe analyses. Regular to irregular cracks cut across some grains. A few of the larger grains exhibit irregular, undulous extinction and sometimes patchy extinction that is unlike mosaicism, but may be due to chadocrysts. The chemical composition of MIL 05029 low-Ca pyroxene closely resembles type 4–6 L chondrites, e.g., the average ferrosilite component falls clearly within the typical range of Fs19–22 (Brearley and Jones 1998); an exception is that the MIL 05029 low-Ca pyroxenes appear to be richer in CaO (average Wo3.7 versus Wo<2 according to Brearley and Jones ). The low standard deviation and consistency of measurements (Table 3) suggest that this enrichment in CaO is not due to contamination with high-Ca pyroxene exsolutions.
Table 3. Major oxide compositions of MIL 05029,4.
Note: Mineral compositions of various phases in MIL 05029,4 determined by electron microprobe. All values in wt%. Errors (in parentheses) are 1σ of averaged analyses.
Low-Ca pyroxene (n = 92)
Olivine (n = 82)
High-Ca pyroxene (n = 48)
Plagioclase (n = 29)
Melt inclusions (n = 9)
Olivine crystals (Fa21.6–25.7, n = 82 analyses; Table 3) are typically euhedral, 50–100 μm in size, but can be up to 200 μm across. They are poikilitically overgrown by low-Ca pyroxene, high-Ca pyroxene, and plagioclase. A few olivine grains contain tiny sulfide and feldspathic inclusions, the latter of which sometimes contain spherical voids that may represent a vapor phase or plucked opaque crystals (Fig. 3). Nine measurements of olivine indicate CaO contents of 0.1–0.65 wt%, but in most analyses CaO is below 0.08 wt%. All of the elevated CaO contents were detected in traverse measurements over inclusion-rich, subhedral grains (although not all measurements on these grains exhibit elevated CaO). Some olivine crystals are intergrown with chromite and have unusually high Cr contents of 0.33–1.2 wt%. These crystals are also slightly more magnesian than most olivine in the meteorite (Fa23.2–21.6, n = 4 analyses compared to the average of Fa25.1, n = 78 analyses). However, the typical range of compositions matches the range of equilibrated L chondrites of Fa22–26 (Brearley and Jones 1998).
High-Ca pyroxene (En46.9–52.3, Fs8.9–12.1, Wo35.8–44.2, n = 48, all augite analyses; Table 3) shows straight extinction, some straight cleavage and abundant lamellae that are <5 μm thick and approximately 10 μm spaced. These appear to be normal polysynthetic twins, rather than shock-produced features. The subhedral-to-anhedral grains are up to 2.5 mm in size but frequently smaller. Some form irregular, indented boundaries toward low-Ca pyroxene crystals and poikilitic intergrowths with olivine occur as well.
Plagioclase exhibits typical polysynthetic twinning. Some larger grains show slightly sweeping extinction but straight extinction is more typical. Interstitial plagioclase domains can be 0.8 mm in size, but grain boundaries are hard to delimit because they are intergrown with abundant 50–200 μm size olivine grains. Compositions range from albite to oligoclase (Ab90.2–78.2An19.7–4.6Or5.7–1.7, n = 29 analyses), with oligoclase compositions being more typical.
Melt inclusions occur as rounded to angular domains with a typical size <10 μm in olivine (Fig. 3), low- and high-Ca pyroxene, and feldspar and apatite. High-resolution scanning electron microscopic imaging coupled with energy-dispersive spectroscopic analyses revealed that these melt inclusions are mostly composed of a feldspathic phase with a relative enrichment of Si and a depletion of Al and variable enrichments of K compared to the large plagioclase grains that crystallized in interstitial spaces around olivines, pyroxenes, and apatite. In some inclusions <<1 μm diameter troilite, Fe-Ni metal, high-Ca pyroxene, and apatite crystals are observable along with the feldspar phase. This suggests that these inclusions are most likely crystallized, trapped melt. Two inclusions were analyzed with a focused, 1 μm diameter beam, with the same analytical conditions used for silicate phase analysis. The first inclusion is in low-Ca pyroxene, 10 μm in size, and its shape appears to be crystallographically controlled because it is roughly rectangular in cross section. The second inclusion is triangular in cross section with subrounded corners, approximately 15 μm size, within high-Ca pyroxene. These analyses have totals of 99.63 and 100.18 wt% and feldspathic compositions with mole ratios of Na54, K44, Ca1 and Na46, K53, Ca1, respectively, but lack the stoichiometry of feldspar.
Seven other melt inclusions were analyzed with a 5 μm diameter defocused beam to limit the effects of volatilization. Three of these inclusions have rounded to subrounded shapes and occur in low-Ca pyroxene. They exhibit mole ratios of Na54–58, K41–45, and Ca0.3. Three melt inclusions with rectangular shapes and concave boundaries occur in interstitial positions in-between low-Ca pyroxene and olivine. They exhibit cation ratios of Na68–79, K16–30, and Ca1–6. One round inclusion in olivine has a cation ratio of Na73, K1, and Ca26. The melt inclusions exhibit similarities with the interstitial/poikilitic feldspar. However, potassium contents are typically much higher, whereas CaO contents are lower (anorthoclase- to sanidine-like compositions). Subtle differences are also present in that certain minor elements exhibit elevated average concentrations compared to the averages of the interstitial plagioclase: P2O5, 0.19 versus 0.05 wt%; FeO, 0.62 versus 0.26 wt%; and TiO2, 0.26 versus 0.07 wt%, respectively.
Spinel is dark brown to black and forms subhedral-to-anhedral grains in mostly interstitial positions. Single grains can be up to 0.5 mm in size but most are smaller (<0.1 mm). Some grains are riddled with inclusions of olivine, whereas others are associated with sulfide/metal. Energy-dispersive spectroscopy suggests dominantly chromitic compositions.
As indicated in the modal composition (Table 2), there is very little metal and sulfide in MIL 05029,4 compared to typical L chondrites. These opaque phases occur in different size grains and/or assemblages. The largest is 1.1 mm in diameter and exhibits a Widmanstätten pattern of lamellar taenite-martensite in a kamacite matrix (Fig. 4). Another large grain is composed mostly of fragmented metal and little sulfide. Smaller opaque grains of sizes <0.5 mm are anhedral, of interstitial types with inclusions, whereas larger grains tend to be subrounded. These smaller grains typically exhibit discrete domains of kamacite and taenite-martensite, similar to the “zoned taenite + kamacite particles” of Reisener and Goldstein (2003).
Metal and sulfide occur both as separate grains and paired assemblages. Intergrowths are not common. This type of texture is similar to that in higher petrologic type (∼6) OCs (Brearley and Jones 1998). No inclusions of spinel, silica, or phosphides were recognized and Si and Ca were typically below detection, suggesting redistribution of these components that is also typical for higher petrologic types of OCs (Zanda et al. 1994). No Neumann Bands were recognized, consistent with the lack of shock-metamorphic features in the silicate minerals.
Kamacite and taenite-martensite domains are usually distinct. Smaller grains exhibit well-developed taenite-martensite profiles, larger grains display several taenite shoulders, interdispersed martensite and few kamacite domains in profiles. Kamacite typically has a composition between 6 and 7 wt% Ni, taenite occurs on the rim of martensite grains or as separate domains in the metal grains with typical Ni concentrations between 20–24 wt% and up to approximately 32 wt%. Martensite has variable compositions of approximately 11–20 wt% Ni; P concentrations are generally very low (<0.02 wt%), Co correlates negatively with Ni contents, with average concentrations of 0.9 and 1.3 wt% in kamacite and approximately 0.3–0.5 wt% in taenite (Table S1). According to Rubin (1990), these values are closest to the characteristic concentration range in kamacite of 0.70–0.96 wt% Co for equilibrated petrologic type 4–6 L chondrites, but below what is characteristic for equilibrated LL chondrites (>1.4 wt%). The Co-contents of taenite-martensite (Ni contents of 11–32 wt%) are only close to the range that is typical for L chondrites (Brearley and Jones 1998).
The two-pyroxene thermometer of Lindsley and Andersen (1983) was used to determine an equilibration temperature for the low- and high-Ca pyroxenes in MIL 05029 (Fig. 5). We modeled a best fit for the electron microprobe data with the QUILF program (Andersen et al. 1993), using the data pairs of high-Ca pyroxene En49.6Fs12Wo38.4 and low-Ca pyroxene En76Fs20.7Wo3.3, which yields an equilibrium temperature of 1076 °C. Taking into account the relative uncertainties of the method and the compositional variety of low- and high-Ca pyroxenes in MIL 05029, an equilibration temperature of 1050–1100 °C appears appropriate. This probably translates to a blocking temperature for Ca diffusion in the pyroxenes of MIL 05029.
Metallographic Cooling Rate
Unfortunately, the well-equilibrated texture of MIL 05029 does not yield information about the thermal equilibration that occurs in large impact events within the first minutes after formation of the melt. However, we know that during this time the super-heated impact melt equilibrates with any cool lithic debris that was mixed in during the cratering process (Onorato et al. 1978).
Typically, immiscible sulfide and metal melt segregates from the silicate melt during this equilibration stage. From the characteristic textures that result in the sulfide-metal droplets, first-order cooling rates can be deduced. These rates are typically on the order of °C s−1 through the temperature range from 1400 to 950 °C, when these melts are still liquid (Scott 1982). Because no relics of clastic debris are retained in MIL 05029, it probably formed in a central part of a voluminous melt pod that equilibrated slowly enough to digest all clastic debris. A thermally equilibrated melt will cool at a slower rate to the surroundings. Characteristic subsolidus textures develop in the Fe-Ni metal phase during cooling through the temperature range of approximately 700–400 °C. Over this temperature range, taenite transforms to martensite and kamacite nucleates at the grain boundaries of polycrystalline taenite grains (Reisener and Goldstein 2003). The central part of the remnant martensite grain is zoned with the highest Ni concentrations (taenite) at the contact toward kamacite. However, contrary to the formational model of Reisener and Goldstein (2003) and Yang et al. (1997), we did not observe the presence of the high Ni alloy tetrataenite (>46 wt% Ni) associated with taenite rims.
Following the procedure of Smith and Goldstein (1977) and Taylor et al. (1979), cooling rates were calculated from traverses across metal particles (Fig. 6). For this method, the isothermal growth of kamacite at an interval of optimum growth temperatures is assumed. In this temperature interval, the observed amount of diffusion relates to a time of growth. Applying the method of Smith and Goldstein (1977), the equivalent times for growing the measured kamacite widths were determined from the corresponding equilibrium martensite Ni compositions. We used the parameterization and Ni diffusion coefficient of Hopfe and Goldstein (2001) to construct the P-free Fe-Ni phase diagram after Yang et al. (1996) and determined the equilibrium cooling time required to grow the observed widths of kamacite domains (Table 4) with the bulk Ni contents (Table S1) determined from martensite plateaus. These equilibrium cooling times were divided by the temperature interval between the start of growth of kamacite and the end of growth. The latter value was approximated from (1) the highest Ni concentration in metal measured in the respective traverse, and (2) the highest Ni concentration in all traverse measurements (35.99 wt%). The results scatter between 9 and 45 °C Ma−1 if one outlier is disregarded (Table 4). We regard the value of approximately 14 ± 6.5 °C Ma−1 (1σ) as a likely conservative cooling rate derived from averaging the consistent cooling rates for the five traverses that ought to be correct within the uncertainties of the method, which are commonly assumed to be one order of magnitude (Smith and Goldstein 1977; Taylor et al. 1979). The example of measurement traverses metal Q-1 and Q-2 in Table 4 shows that an inherent, considerable uncertainty arises from the geometry of the measurement traverses. In this example, a variation of a factor of approximately three occurred for the cooling rate estimate.
Table 4. MIL 05029 metallographic cooling rates.
Average Fe-Ni metal
Growth duration (s)
Taenite maximum Ni (wt%)
Min.a (°C Ma−1)
Avg.b (°C Ma−1
Max.c (°C Ma−1
Notes: n.d. = none detected. Particle metalQ was analyzed in two traverses that show different kamacite widths due to geometric variation, this cooling rate appears to be an outlier. We prefer the avg. values as the approximate metallographic cooling rate of MIL 05029, which suggests approximately 14 °C Ma−1.
aBased on the temperatures derived from the optimum growth interval.
bBased on the temperatures derived from the maximum Ni concentrations in taenite measured in the individual traverses.
cBased on the temperature that corresponds to 35.99 wt% Ni, the highest Ni concentration in taenite measured in all the traverses.
dBecause 22.61 wt% Ni in taenite corresponds to a temperature of 579 °C, this would lie within the optimum growth range, suggesting that the highest Ni concentration was not captured in the measurement traverse. The Ni concentration that corresponds to the lower limit of the optimum growth at 560 °C is approximately 24.3 wt% Ni.
eThis traverse indicates a transition toward a Widmanstätten-like texture; however, the calculated average cooling rate appears consistent.
fSecond highest Ni concentration in this traverse, the highest is 35.99 wt% Ni.
An attempt to apply the method of Wood (1964) to determine a metallographic cooling rate by comparing the central taenite concentrations to taenite half-widths on the Widmanstätten pattern-grain in MIL 05029 was unsuccessful because most of the lamellae are too thin. Central Ni concentrations and radii of the thickest taenite lamellae would yield cooling rates of several 100 °C Ma−1 (Willis and Goldstein 1981). The calculated cooling rate for the Widmanstätten pattern-grain is probably misleading because past applications of metallographic cooling were performed on massive metal grains that were at least in the cm-size range (e.g., in Portales Valley, or iron meteorites). Additionally, these larger metal grains had a much higher P content in the Fe-Ni metal than is seen in MIL 05029.
We used the Ar-Ar method to determine the age of three splits of MIL 05029. Complete Ar-Ar data are presented in Table S2. Plateau plots for all three splits are shown in Fig. 7. Although there is some scatter within each plateau, and between plateaus, our best interpretation gives an Ar-Ar age for MIL 05029 of 4517 ± 11 Ma. A variety of reduction scenarios are listed in Table 5, and will be explained below. First, we will discuss how we obtained each plateau age and which steps we included, and then we will consider complications potentially resulting from nonradiogenic 40Ar.
Table 5. Ar-Ar age reduction of MIL 05029,5.
Raw plateau age (Ma)
“Air” plateau (Ma)
Spallation plateau (Ma)
% of 39Ar in plateau
Total release age (Ma)
Notes: Various age reductions for the three analyzed splits. Column 2 is uncorrected for trapped gas, and is the preferred age. Column 3 is corrected assuming all 36Ar is trapped terrestrial atmosphere. Column 4 is an attempt to remove the contribution from spallation, and then assumes that the remaining 36Ar is trapped terrestrial gas. Column 5 is the percent of 39Ar used in determining the ages from the columns 2–4. Column 6 is determined by summing all temperature steps together except for the first few steps that obviously have been disturbed.
MIL 05029,5 split A
MIL 05029,5 split B
MIL 05029,5 split C
Weighted average (1σ)
We determined plateau ages using three different methods: (1) a mean age weighted by the inverse square of the errors, with final errors equal to the inverse sum of the squared errors; (2) unweighted mean with standard deviation for errors; and (3) summing the gas from all steps in the plateau together, which would be analogous to what we would measure if the steps that comprise the plateau had instead been released as a single step (these are the ages given in Table 5). Regardless of which method is used, the plateau age of a particular split only changes by 12 Ma or less, and there is no systematic trend for ages derived by different techniques, i.e., one particular method does not always give the oldest or youngest apparent age. Errors for each method are about 10 Ma for the weighted mean and summing of gas, whereas for the unweighted method the standard deviation is about 50 Ma for split A, and about 150 Ma for splits B and C. Although a significant scatter of individual steps is indicated by the large standard deviation, because the three different methods of reduction all give about the same plateau age, the real uncertainty is likely closer to 10 Ma. The reason for the scatter of individual ages is at this time unknown, because (as stated below) removal of suspected trapped Ar has no effect on the scatter. All plateau ages in Table 5 were determined by summing the gas together from the individual steps that comprise each plateau, giving a semi-total fusion age.
The next question is which steps should be included in the plateau. Because this meteorite has at least two sources of K (see the Diffusion section) it is possible that we would see grain boundary effects such as adsorbed Ar, partial resetting, or weathering twice in the same sample. Upon first glance at Fig. 7, it would seem that we are indeed seeing partial resetting in splits A and C and adsorbed Ar in split B between about 65% and 75% of the 39Ar release. However, split A did not have a regular heating schedule, so the step with the low age is actually 50 °C lower than the highest temperature previously achieved (Table S2), hence any grain boundary effects should have already been removed. For splits B and C, the deviation from the plateau age occurs when about 10% of the gas from the second phase should have already been removed. Hence, the deviation from the plateau age being a grain boundary effect does not seem plausible. Therefore, we include these steps when calculating the plateau age, although if we were to remove them it would produce a change comparable to the size of the error bars.
Terrestrial adsorbed Ar, or partial loss due to mild heating, typically alters the first few steps of any Ar-Ar release pattern. These three chips are no different, and the first approximately 5% of the Ar is not contained in the plateau. Additionally, above approximately 90% release, chips B and C have multiple steps that give apparent ages less than the plateau age. This is probably due to 39Ar recoil during irradiation. Recoil typically would also lead to artificially old ages at low temperatures, where near-surface portions of grains are preferentially degassed, but in these splits the partial resetting masks the effect of recoil. Because there is a slight downward trend in age for chips A and B above approximately 60% release, one may wonder if this is also due to recoil. We find this improbable because recoiled 39Ar is typically released above 1000 °C due to implantation in pyroxene and olivine, and in MIL 05029 only 5–10% is released above this temperature. It is interesting that the gas from approximately 60–85% in chip B gives an age of approximately 4420 Ma, whereas the gas from approximately 0–60% gives an age of approximately 4560 Ma. Because the latter release has a younger age opposite to what one would expect for partial resetting, we doubt the significance of this younger age. Because the age difference between the higher and lower temperature release is about a factor of 2 smaller in chip A and nonexistent in chip C, we prefer to combine all the gas together, giving the age shown in Table 5. For chip A, we only extend the plateau to approximately 60% because over this range the steps agree with each other to within 2σ. Although choosing various steps to include in the plateau can give different ages for a single chip, if we use the same plateau criteria for all three chips, the averaged age only changes within the stated averaged errors. Therefore, we consider the age in boldface in column 2 of Table 5 as most likely to be correct.
Although it is important to remove any trapped 40Ar in the data reduction, Cl-derived 38Ar prevented us from accurately distinguishing between all sources of 36Ar. However, by assuming all 36Ar is trapped terrestrial gas with a 40Ar/36Ar ratio of 295.5, we can determine a lower limit to the age, as shown in column 3 of Table 5. One potential way of removing the contribution from Ca spallation is to assume that the temperature step with the lowest 36Ar/37Ar gives the ratio of Ca-spallation-produced 36Ar to reactor-produced 37Ar (Garrison et al. 2000). Although this value may not be strictly true, and does not remove any spallation from K or Fe, it will give a more reasonable lower age limit when assuming the remaining 36Ar is all terrestrial gas (column 4 of Table 5). The assumption that any 36Ar is terrestrial is probably not correct since, after removal of Ca spallation–produced 36Ar, an isochron plot of 40Ar/36Ar versus 39Ar/36Ar is highly linear and gives an intercept of −79 ± 65, indicating that there is little or no 40Ar associated with the trapped 36Ar. Additionally, removal of trapped Ar, when actually present, typically has the effect of smoothing out the plateau plot. Removal of suspected trapped Ar in MIL 05029 has the effect of lowering the apparent age of each step without smoothing the plateau, although the lowering effect is more pronounced above about 90%39Ar released. Because the isochron plot intercept could be compromised by spallation from K and Fe, or by incorrect identification of spallation from Ca, lower limits in the form of “total release” ages without air corrections are provided in Table 5. A “total release” age is all steps except for the first few that had obvious disturbance, either from partial resetting or weathering. We doubt the validity of these total release ages because, as discussed above, there appears to be recoil and partial resetting/weathering. The total release age usually includes 100% of the gas, and removes the effect of recoil by combining the artificially old low-temperature steps with the artificially young high-temperature steps. In MIL 05029, recoil is overlain by disturbance at low temperature, so there are no old low-temperature steps to combine with young high-temperature steps, effectively making the total release age a lower limit.
We favor the raw plateau age of 4517 ± 11 Ma for the Ar-Ar age. However, we note that the age could be as young as approximately 4.48 Ga if all 36Ar was terrestrial air contamination, although we find this very unlikely because of the negative intercept on the isochron. The other reduction scenarios give a more realistic lower limit of approximately 4.50 Ga. When looking at any of the ages, it should also be kept in mind that the decay constant of K is currently under investigation (Min et al. 2000; Schwarz and Trieloff 2007), and it is now recommended to add approximately 30–60 Ma to approximately 4.5 Ga old rocks. If the revision is right, this correction would increase the apparent age of MIL 05029 to approximately 4547 ± 11 Ma or approximately 4577 ± 11 Ma.
For most materials, the diffusion coefficient can be described by the Arrhenius equation:
where D0 is the pre-exponential, E is the activation energy, T is the temperature in Kelvin, and R is the gas constant. The typical Ar-Ar experiment, such as ours, actually measures D/a2, where a is the grain radius. This does not affect the Arrhenius equation as both sides can be divided by a2 without changing the function. By plotting our data in an Arrhenius plot (log D/a2 versus 1/T), any material behaving according to the Arrhenius equation will result in a straight line, with the slope proportional to E. With known diffusion parameters, we can gain more insight into the K-bearing domains, and determine how many Ca-bearing domains also contain K. With this information we can identify the physical location of K in the meteorite, and determine how representative the splits we used for Ar-Ar dating were by comparison with the thin section. Additionally, diffusion information can be used to determine the closure temperature, which can be important for relating the Ar-Ar age to the impact age.
A high-resolution Arrhenius plot of chip B, determined by assuming spherical grains, is shown in Fig. 8. The Arrhenius plot for chip C gives a similar pattern; however, as it has a lower temperature resolution during the linear portions and has more scatter, we will be exclusively examining chip B. The temperature resolution of chip A was too low to make analysis worthwhile. Both 39Ar (K) and 37Ar (Ca) are plotted together. Because we are dealing with whole-rock samples, which can have multiple K domains (either different sizes or different mineralogical compositions), the data do not all fall on a single straight line. Portions of the plot, however, will fall on a straight line when only a single domain is outgassing. This separation of domains can also be seen by comparison with Fig. 9, where there appears to be two or three separate releases. Accurate diffusion parameters for these domains can be found by including as many points as possible that fall on a straight line, whereas excluding points during which two or more domains are contributing to the released gas. This can be done by using the goodness-of-fit parameter Q (Press et al. 1992), which uses the incomplete gamma function and the chi-square value. The incomplete gamma function is an integral often used in probability theory, and the chi-square value is a quantity determined by both the deviation of the measured values from an assumed model and by the error of each measurement. Essentially, Q is the probability that the calculated chi-square value would be that poor by chance (higher Q means higher probability of linear behavior). Rather than choosing data points that give the highest Q (which would be only three or four points), we instead choose data points that give the highest Q*N, where N is the number of data points included in the fit (Lovera et al. 1997). This ensures that we include the largest number of data points that still form a straight line. Errors in temperature were 5 °C, and errors in log D/a2 were determined by the method employed by Lovera et al. (1997) and line fitting routines were taken from Press et al. (1992).
A simple fitting of the linear portions of the Ar “Total” release curve of Fig. 8 will not produce accurate results. We can see an obvious example of this in the lower temperature portions of the 37Ar and 39Ar curves, which have similar slopes but different intercepts. This artifact results because the Arrhenius equation assumes a single domain, when we in fact have multiple domains. Qualitatively, the percentage of the total gas released in a single step is smaller than the percentage of gas released from a single domain, which tricks the Arrhenius equation into lowering the diffusion coefficient.
To remove this artifact, we try to partition the gas into the different domains. Figure 9 shows that the first release is exhausted by 660 °C, so we will take all of the gas below 660 °C and partition that into its own domain. As we will see later this is not strictly true, but is the best we can do. Figure 9 also indicates that the second domain is exhausted by 1000 °C, so we partition the gas released between 660 and 1000 °C into the second domain. The results can be seen in the curves labeled “Dom. Sep.” in Fig. 8, and the slope and intercept of both domains is listed in Table 6. We see that the curves of 37Ar and 39Ar for the first domain are now in agreement within 2σ. For the first domain, we trust the 39Ar data to be more accurate because it contains approximately 75% of the total 39Ar, but only approximately 7% of the total 37Ar. The curves of 37Ar and 39Ar do not agree with each other as well in the second domain, although the slope and intercept are still within 2σ. The diffusion parameters for the second domain are less reliable because of the inherent difficulties in partitioning the gas. For example, if we use the calculated diffusion parameters, the second domain would have lost 5–10% of its gas by 660 °C. The effect on the first domain would be much less, only increasing the amount of gas by 1–2%.
Table 6. Diffusion parameters of MIL 05029,5 split B.
Temp. range (°C)
E (kcal mol−1)
Log D0/a2 (s−1)
Note: Diffusion parameters of both 39Ar and 37Ar for both temperature releases after domain separation. Errors (in parentheses) are 1σ.
aInherent difficulties in partitioning the gas indicate formal errors are probably too small.
There might be a third domain at high T between 1025 and 1250 °C, which appears linear in 39Ar and contains about 6% of the total 39Ar; however, as most of those points are above the melting temperature of Na-rich feldspar (∼1100 °C) and do not match the activation energy in 37Ar, it is not known if they have any meaning. These steps also give young apparent ages (Fig. 7) that may indicate that this 39Ar could be recoiled into pyroxene or olivine.
Starting around 950 °C, the diffusion rate of 37Ar increases drastically, indicating a shift in the mineral source of Ca, with no accompanying shift in 39Ar. The high-temperature release of 37Ar is from pyroxene (mostly high-Ca Pyx) and apatite. At this high T, the E of 37Ar is >100 kcal mol−1. As the activation energy of 39Ar does not also increase this means the pyroxene and/or apatite do not contain measurable amounts of K, contrary to what has been suggested in many meteorites (Bogard et al. 1995; Kring et al. 1996; Kunz et al. 1997).
Is MIL 05029 an impact-melted L chondrite, or a differentiated rock from an as yet unsampled parent body with L chondrite affinities? Most chondritic impact melts have a small grain size due to rapid cooling and contain relict clasts with shock features. MIL 05029 does not have these distinguishing features, despite having L chondrite affinities and an oxygen isotopic composition well within the range observed for L chondrites. Moreover, its ancient Ar-Ar age and unusual petrologic properties suggest that it constitutes a unique record of the early evolution of its parent asteroid, regardless of which formation scenario is correct.
Oxygen isotope measurements indicate L chondritic affinity for MIL 05029. The compositions of the major mineral constituents (olivine and low-Ca pyroxene) are mostly similar to those in L chondrites. Higher CaO contents in low-Ca pyroxenes of MIL 05029 probably indicate high-temperature equilibration conditions (Lindsley and Andersen 1983). Fodor and Keil (1975, 1976) interpreted increased CaO contents in poikilitic low-Ca pyroxenes of the L chondrite melt rock Shaw and in similar melt rock fragments in the LL chondrite Ngawi and the H chondrite Plainview as due to equilibration of superheated impact melts. Furthermore, the relative abundances of olivine and low-Ca pyroxene phases also confirm the relationship with L chondrites.
Texturally, it is difficult to determine definitively the formation environment of MIL 05029. The medium-grained poikilitic texture of MIL 05029 implies complete crystallization of a melt in a relatively stable environment, which is also indicated by pyroxene crystals that frequently exhibit 120° junctions. Although this texture is consistent with cooling at a relatively slow rate in an insulated environment, it does not distinguish between plutonic magmas and an impact melt setting. If MIL 05029 was created in an impact, the lack of shock features is not surprising because it is a clast-free melt.
The meteorite has an extremely old age, indicating that the melt formed within tens of Ma of solar system formation. This means MIL 05029 could potentially be sampling an asteroid with L chondrite affinities that was heated enough to cause differentiation. The HED (howardites, eucrites, and diogenites) meteorites are well-studied examples of differentiated meteorites. The oldest cumulate and unbrecciated eucrites give a tight cluster of Ar-Ar ages around 4.48 Ga, which have been interpreted as recording closure of the Ar-Ar system by excavation from depth during an impact (Bogard and Garrison 2003). As MIL 05029 reached its closure temperature at least 30 Ma before the oldest eucrites, there are two scenarios that would enable it to be part of an unsampled parent body. The first is that it was large enough to differentiate, but small enough to cool below the closure temperature faster than the HED parent body, or MIL 05029 was also excavated from depth. Both of these scenarios are unsubstantiated, but possible.
Although an old age opens the possibility of magmatic formation, oxygen isotope data and mineral composition are consistent with an L chondrite. Recent thermochronologic modeling suggests that an L chondrite parent body was never sufficiently heated to experience differentiation and magmatism because accretion of the OC parent asteroids was delayed for approximately 2 Ma after formation of calcium-aluminum-rich inclusions (CAIs) (Hevey and Sanders 2006; Kleine et al. 2008). This deprived the 100–200 km diameter parent asteroids of sufficient radiogenic heating from short-lived isotopes to initiate endogenous magmatism. Because no evidence for multiple parent bodies with L chondrite affinities as potential sources for MIL 05029 is available, the petrological data suggest an impact on the L chondrite parent body as the most plausible formational scenario for MIL 05029.
The meteorite appears depleted in metal and sulfide, both in thin section and in hand sample, although it must be kept in mind that preliminary work on the Shaw meteorite also reported metal and sulfide depletion. While it is true that cm-sized regions were depleted in Shaw, other areas were enriched such that the overall abundances were within the normal limits (Taylor et al. 1979). If we assume that the metal and sulfide depletion in MIL 05029 occurs throughout the entire meteorite, the impact event probably either stripped metal and sulfide from the shock-metamorphosed target rocks (e.g., by shear), or created a melt volume that was sufficiently large and remained sufficiently hot long enough to allow for density-driven segregation of metal-sulfide from silicate melt. Below, we consider these two possibilities.
Shearing during an impact (Stöffler et al. 1991) has been seen in Portales Valley (Kring et al. 1999), and is a very reasonable explanation for the metal/sulfide depletion in MIL 05029. However, shearing does not explain the feldspar enrichment in the silicate fraction seen in Table 2. If the metal/sulfide depletion was instead caused by density separation, then if fractional crystallization also occurred it might allow the mafic silicates to separate by density, enriching the remaining melt in the feldspathic component. The silicate fraction in Table 2 does not show depletion of olivine or low-Ca pyroxene, but does appear to show depletion in high-Ca pyroxene. Nonetheless, we calculate separation speeds for all relevant mineral species.
A simple calculation using the Stokes velocity equation (Turcotte and Schubert 2002) can be applied to the metal/sulfide, olivine, and pyroxene grains. The velocity is then
where Δρ is the density difference between the mineral of interest and the silicate melt, r is the grain radius of the mineral of interest, g is the gravity on the asteroid, and η is the viscosity of the silicate melt. As we are interested in the maximum speed (giving a minimum separation time), we used the largest estimated parent body size of 190 km diameter with zero porosity (Bennett and McSween 1996) which gives a g of 0.09 m s−2. By using equations which determine viscosity as a function of composition and temperature, we calculate a viscosity of 0.05 Pa·s for an L chondrite silicate melt at 1600 °C (Bottinga and Weill 1972; Shaw 1972). An Fe-Ni metal grain of radius 0.55 mm would sink at a rate of approximately 3.3 m h−1, whereas an olivine grain of radius 100 μm would sink at 1.5 × 10−2 m h−1. Pure troilite of radius 0.55 mm would descend at approximately 1.1 m h−1. Although the pyroxene grains in MIL 05029 are much larger than the olivine grains, much of the growth could have been solid-state. Hence, we assume pyroxene grains of a similar size to the olivine grains, and because of the similar density of low-Ca pyroxene, high-Ca pyroxene, and olivine, the pyroxene grains would sink at a rate close to that of the olivine.
Even if the proposed mafic silicate depletion was not a meteorite-wide event, the thin section we examined shows that migration of at least a centimeter (the size of the thin section) could have occurred. If the mafic silicates migrated 1 cm via density separation at this speed, it would take at least 0.7 h to occur. If the depletion is throughout the entire meteorite (or if the migration took place at a temperature <1600 °C), it implies the meteorite remained liquid for an even longer period of time. If density separation did deplete the mafic silicates, it raises questions about the metal/sulfide abundance. In the amount of time it takes the mafic silicates to travel 1 cm, the metal would have traveled 2.2 m, which should have removed nearly all the metal. This issue becomes even more severe upon examining the whole-rock specimen, which has metal grains up to 2.5 mm in radius (Fig. 1) that would descend at 68 m h−1.
Both shear and density separation leave unanswered questions. Shear does not explain the feldspar enrichment, while density separation requires a fine balance between the grain size of metal/sulfide and pyroxene such that the pyroxene can be depleted without completely removing the metal/sulfide. At this time that balance, and how it would leave the olivine abundance unaltered, cannot be confirmed.
Using the thermal history to distinguish between a melt dike and a melt pool will be difficult because the old age means the parent body might have been up to 800 °C at the time of the impact (Bennett and McSween 1996), depending on when the impact occurred and the emplacement depth. In the melt pool scenario, the initial temperature of the parent body is somewhat mitigated because the breccia surrounding the melt pool would have been heated by the impact. The solidification time would then be controlled by conduction to the surrounding breccia, and the subsolidus cooling rate would be controlled by conduction to the surface and surrounding wall rock, and this energy would in turn be radiated to space. In the melt dike scenario, the solidification time would be less than the melt pool scenario as the thermal diffusion distance is less, but if the impact occurred into a region that was 800 °C, the cooling time scale may still be sufficiently slow to develop a medium-grained igneous texture. The subsolidus cooling would then be controlled by the parent body radiating leftover radiogenic energy to space, and would be identical to all material located at that depth.
The lack of relict clasts helps distinguish between these two scenarios. PAT 91501, while texturally similar to MIL 05029, contains rare relic clasts. For this reason it was interpreted to be a sample of a melt dike, albeit one from the center due to the small number of clasts (Benedix et al. 2008). MIL 05029, in contrast to many chondritic impact melt breccias, e.g., Shaw (Taylor et al. 1979), Cat Mountain (Kring et al. 1996), Abee (Rubin and Scott 1997), and PAT 91501 (Mittlefehldt and Lindstrom 2001), is not associated with clastic material. In this regard, MIL 05029 is much like the total impact melt Ilafegh 009, although this impact melt formed on an enstatite parent body (McCoy et al. 1995). Last but not least, slow cooling indicates deep burial, further indicating that MIL 05029 was involved in a large-scale impact. Comparison with large-scale terrestrial impact craters indicates that the amount of impact melt that occurs in dikes is dwarfed by the volume of melt that is retained as coherent pods, melt sheets, or as components of suevites (Wittmann et al. 2010). Thus, the lack of clasts in MIL 05029 implies a thick volume of melt, suggesting a melt pool rather than a melt dike, because MIL 05029 must have remained liquid long enough to assimilate clasts by melting.
Regardless of the formation scenario, because the metallographic cooling rate of approximately 14 °C Ma−1 is about what is inferred for the transitional conditions between typical L5 and L6 chondrites (Taylor et al. 1987), it appears to be a reasonable assumption that MIL 05029 cooled in the thermal regime of the L5–L6 boundary. Thermal modeling of the L chondrite parent asteroid (Bennett and McSween 1996) suggests that the minimum depth for this region is 5–12 km below the surface of a 100–200 km diameter L chondrite asteroid. Scaling laws for simple craters on asteroids and the Moon indicate depth-to-diameter ratios of 1:5 (Sullivan et al. 1996). Excavation to depths of 5–12 km in a low-gravity environment would therefore require minimum crater diameters of 25–60 km. On a parent body 100–200 km in diameter, this size impact would at least have caused rheological weakening (Housen 2009), making later shattering (where the pieces remain gravitationally bound) easier. However, it is also likely that such a magnitude impact event itself shattered the nascent L chondrite parent body if scaling relationships of Holsapple et al. (2002) are taken into account. It should be kept in mind that shattering does not necessarily indicate blocks that are put into orbit and later re-accreted. Shattering could also mean an “expansion” of the radius of the parent body, followed by collapse. If shattering occurred, whether by the impact that created MIL 05029 or a later event, it could have rotated blocks in place. This rotation would expose hot type 6 material to space that would cool rapidly, and possibly emplace larger pods of impact melt at depth (Grimm 1985). This shattering scenario would explain why L chondrites do not exhibit a correlation of metamorphic grade with cooling rates as is found in H chondrites (Trieloff et al. 2003), and instead have strongly metamorphosed L6 chondrites that have faster metallographic cooling rates than less metamorphosed L5 or L4 chondrites (Taylor et al. 1987). In summary, MIL 05029 could provide evidence for why the onion shell model does not work for the L chondrite parent body, because shattering would have disrupted the otherwise steady cooling. However, an impact into a thermally metamorphosing L chondrite body is not the only scenario that can produce a slowly cooled impact melt. Recent hydrocode models using conservative parameters show that impacts between cold porous bodies can produce melting followed by subsequent cooling of 10 °C Ma−1 (Ciesla et al. 2009; Davison et al. 2010). Additionally, porous objects will absorb more of the shock energy, requiring larger impacts to disrupt them.
Most L chondrite impact melts have ages close to 500 Ma (e.g., Chico). Previously, only two L impact melts were known with Ar-Ar ages >4.4Ga: Shaw (∼4.23–∼4.46 Ga; Bogard and Hirsch 1980; Turner et al. 1978) and PAT 91501 (4.461 ± 0.008 Ga; Benedix et al. 2008). MIL 05029 cooled much slower, has an even older Ar-Ar age of 4.517 ± 0.011 Ga, and is >5σ away from the oldest previous L impact melt. The age of MIL 05029 is about the same as the Ar-Ar ages of unshocked H5 chondrites (Turner et al. 1978; Trieloff et al. 2003), and as the H and L chondrite parent bodies initially had a similar thermal history, this indicates the impact event that created MIL 05029 occurred while the parent body was still cooling. This means the Ar-Ar age could be dating the end of metamorphism instead of the age of the impact. By incorporating the time required to form the cooling rate seen in the metal, we can somewhat constrain when the impact occurred. The metal tells us that over the temperature range of approximately 700–400 °C, MIL 05029 cooled at approximately 14 °C Ma−1 (Table 4). This means we need to add at least approximately 20 Ma to the Ar-Ar age, putting the impact at approximately 4.537 Ga. If we assume cooling down to the closure temperature of the Ar-Ar system, the impact must have occurred even earlier than this. Newtonian cooling, where the cooling rate is proportional to the temperature difference, will give us a decent estimate of the cooling rate below that measured by the metal. The equation for temperature as a function of time then becomes
where Ta is the ambient temperature, T0 is the initial temperature (700 °C in this case), k is a constant, and t is time. We can determine k from the metallographic cooling time, as the temperature after approximately 20 Ma was 400 °C. The closure temperature of feldspar (a common carrier of K) is typically 240 ± 120 °C (Turner et al. 1978), which for Ta between about −170 and 30 °C would indicate that the impact occurred approximately 40 Ma before Ar closure, setting the impact age at approximately 4.557 Ga. Using the diffusion parameters and cooling rate for the first release of MIL 05029, we get an even lower closure temperature of ∼40 ± ∼20 °C. As a side note, while an apparent closure temperature of approximately 40 °C is low, it is not unique for L chondrites. The L chondrite Mount Brown also has a very low closure temperature of 40–60 °C (Turner et al. 1978). This lower closure temperature, coupled with the Newtonian cooling, would indicate an impact at approximately 4.59 Ga for ambient temperatures of −170 °C. This would indicate an age older than the formation of the solar system, which clearly did not happen. Higher ambient temperatures (even temperatures that decrease with time) require an even older event, making the problem worse.
More likely, the cooling rate was not steady below 400 °C due to shattering, the Ar closure temperature was not 40 °C, or (because the metallographic cooling rates are only accurate to one order of magnitude) the true cooling rate was closer to 100 °C Ma−1. If the first scenario, the cooling rate will not help us calculate an impact age, but will instead only give a lower limit. If the last scenario, the cooling rate indicates that the impact occurred between 4.521 and 4.531 Ga, depending upon the ambient temperature and closure temperature. If it is shown the metallographic cooling rate of approximately 14 °C Ma−1 is correct, it almost requires excavation after the formation of MIL 05029 or a higher Ar closure temperature to explain the apparent Ar-Ar age. Without further work, we can only say the impact occurred sometime between the apparent Ar-Ar age of approximately 4.52 Ga and the accretion of the L chondrite parent asteroid, but our best estimate places it at approximately 4.54 Ga, due to the additional approximately 20 Ma indicated by the metallographic cooling time. To determine accurately the age of impact, isotopic systems with a higher closure temperature, such as Pb-Pb, Sm-Nd, or I-Xe, will be needed.
Miller Range 05029 is not the only meteorite that records a very early impact in solar system history. Happy Canyon, an impact melt breccia on either the EL or EH parent body, has an Ar-Ar age of 4.53 Ga, and an I-Xe age that is indistinguishable from solar system formation (McCoy et al. 1995). The accretion of most asteroids was rapid, typically <10 Ma after the formation of CAIs (Nichols 2006). But we also know that post-primary accretion could have continued for significantly longer, as evidenced by the final formation of the Earth–Moon System approximately 60 Ma after the formation of CAIs (Touboul et al. 2007). MIL 05029, sandwiched between these two important events, is one of only a few meteorites that record an impact during or shortly after post-primary accretion (McCoy et al. 1995).
K Carrier Phase
Mineral separates are typically not extracted from chondrites because of their limited quantity and fine-grained nature. Whole-rock analysis usually leads to difficulty in identification of which phase is being dated. What is known is that meteorites commonly show two releases of Ar, one below the melting temperature of sodium-rich plagioclase (∼1100 °C; Deer et al. 1962), the other above. The lower temperature release is almost certainly plagioclase feldspar, but there is no consensus on what produces the higher temperature release. It has been attributed to K contained in pyroxene, an effect of shock that converts the low-temperature phase into the high-temperature phase, a change of diffusion distance from grain size to melt vein width, and feldspar enclosed in enstatite grains (Bogard and Hirsch 1980; Bogard et al. 1995; McCoy et al. 1995; Kunz et al. 1997). MIL 05029 is special because, in addition to just two releases of Ar, we have also identified two K sources via the microprobe. The next logical step is to see if these two sets can be matched, so that is where we turn our attention next.
From the Arrhenius plot (Fig. 8) and release curve (Fig. 9), it is obvious that there are two major sources of K in MIL 05029, and both phases release their Ar below the melting temperature of feldspar. Examination of the thin section reveals two sources of K, oligoclase and feldspathic melt inclusions. The first release has an E of approximately 25 kcal mol−1, much less than the approximately 44 kcal mol−1 of feldspar (McDougall and Harrison 1999). The E for maskelynite in Allan Hills 84001 was found by Weiss et al. (2002) to be approximately 18 kcal mol−1, and we have found, in work to be published subsequently, shocked Na-rich feldspar to have an E of approximately 24 kcal mol−1 (Weirich et al. 2010). Additionally, if we assume the D0 for shocked feldspar from Weirich et al. (2010) to be the same as the D0 for the first release in Table 6, we get a grain diameter for the first release of 7–20 μm, in agreement with the grain size of the feldspathic melt inclusions. However, the melt inclusions in MIL 05029 are crystallized, not disordered like maskelynite or shocked feldspar. The feldspar in MIL 05029 is much larger and easy to analyze, exhibits twinning and birefringence, and certainly is not glass or maskelynite. We were not able to determine accurate diffusion parameters for the second release, although it is consistent with the activation energy of feldspar. Based upon the diffusion parameters, feldspar is most likely the second release and melt inclusions are the first release, although it is unclear why (or if) the melt inclusions have a disordered structure.
The relative amounts of K in each domain make comparison of the two sets even harder. Calculating the amount of K from the release of Ar is straightforward: 74% of the total 570 ppm K is “released” from 300 to 660 °C, another 20% from 675 to 1000 °C.
To calculate the amount of K in each domain based on petrographic data, we multiplied the average K wt% (Table 3) by the volume abundance (Table 2) and the ratio of domain density to whole-rock density (assumed to be 0.77). In this calculation, the melt inclusions contain 29% of the total 837 ppm K, with the remaining 71% coming from the feldspar.
By comparing the Ar-Ar data to the microprobe data, two things become obvious. First, the Ar-Ar data and the microprobe data (both with uncertainties on the order of ∼10%) disagree on the K abundance by approximately 40%. As long as we are comparing percentages, this will not affect the outcome. Second, the microprobe data clearly indicate that most of the K is located in the feldspar, whereas the Ar data indicate that most of the K is not located in the feldspar. Perhaps, there is an inhomogeneity in the sample, and the splits we used for Ar-Ar analysis had more melt inclusions. This is certainly possible because the samples used for Ar-Ar analysis were approximately 2 mm in diameter, smaller than the typical grain size of low-Ca pyroxene.
Unfortunately, we were not able to find a consistent correlation between the two domains using Ar-Ar and microprobe data. Despite not having an exact match between the two data sets, we can confidently say that two K domains were identified using two independent techniques. Based on the petrographic observations, those domains are oligoclase and melt inclusions.
Miller Range 05029 is a slowly cooled L impact melt, with an Ar-Ar age of 4.517 ± 0.011 Ga. From the oxygen isotopes and major mineral composition, we infer an L chondrite parentage for MIL 05029. However, although the abundance of olivine and pyroxene are similar to what is typical for L chondrites, MIL 05029 displays metal and sulfide depletion and plagioclase enrichment relative to the normative composition of L chondrites. Because of the clast-free nature, metallographic cooling rate of approximately 14 °C Ma−1, and presence of a Widmanstätten-like pattern in a mm-sized metal grain, we conclude that MIL 05029 most likely was part of an impact melt pool buried in the L5–L6 region or equivalent thermal regime. The size of an impact required to emplace shocked material at that depth most likely rheologically weakened the asteroid, making it easier to shatter or disrupt during later impacts. By combining the measured Ar-Ar age with the time required by the slow cooling, we estimate that the impact occurred approximately 4.54 Ga, although times between approximately 4.52 Ga and the formation of the L chondrite parent asteroid are possible. Systems with a higher closure temperature, such as Pb-Pb, Sm-Nd, or I-Xe will need to be utilized to determine accurately the impact age. Both the Ar-Ar release pattern and microprobe data indicate two major sources of K (oligoclase and melt inclusions), although the relative quantities of K in these two phases as measured by the two techniques do not agree, perhaps implying heterogeneity. Despite the apparent plagioclase enrichment, the whole-rock K abundance seems normal for an L chondrite.
Miller Range 05029 is an exceptional find because it records the so far oldest known impact on the L chondrite parent body, giving an Ar-Ar age close to that of the oldest unshocked H5 chondrites. This implies that the impact occurred while the parent body was still undergoing thermal metamorphism. Further, this impact occurred after the formation of the parent body, but prior to the formation of the Earth’s moon, making it one of the few known impacts during or shortly after the postprimary accretional phase of the solar system. Due to the inferred crater size of 25–60 km, MIL 05029 is evidence that at least one large impact occurred on the L chondrite parent body while it was still cooling, perhaps explaining why the onion shell model does not work for this asteroid.
Acknowledgments— We acknowledge ANSMET and their funding institutions NASA, NSF, and Smithsonian Institution, for recovering MIL 05029. We thank K. Righter, C. Satterwhite, R. Harrington, and L. Watts at the NASA Johnson Space Center Meteorite Curation Facility. We also acknowledge Electron Microprobe assistance by A. Peslier and J. Herrin, and SEM assistance by G. A. Robinson. D. R. is grateful to NASA’s Cosmochemistry program, grant NNX07AI48G, for supporting operation of the oxygen isotope at the Geophysical Laboratory. Funding by NASA Earth and Space Science Fellowship (NESSF) for J. R. W. and NASA grants to T. D. S. and D. A. K. (NNX07AG55G) are gratefully acknowledged. This is part of J. R. Weirich’s Ph.D. dissertation.
Editorial Handling— Dr. Alex Ruzicka
Elements used for silicate analysis were calibrated with the following standards: Mg, Si, Al, Ca, Fe, Ti with a natural kaersutite; Na with a natural oligoclase; K with a natural orthoclase from Grisows, Switzerland; P with a natural apatite from Minot, Maine; Cr with a natural chromite; Mn with a natural rhodonite; Ni with synthetic NiO; S with Canyon Diablo troilite.
Elements used for melt inclusions were calibrated with the following standards: Na, K, Al with obsidian; Si, Ca, Fe, Mg with a synthetic glass (K412); P with a natural apatite from Minot, Maine; Cl with natural tugtupite; Ti with rutile (pure oxide); Mn with a natural rhodonite; Ni with synthetic NiO; Cr with a natural chromite.
Elements used for Fe-Ni metals and sulfide were calibrated with the following standards: Fe, Ni, Co, Cr, Mn, Cu, Zn, Ti, Si, and V with pure metals; P with apatite; Ca and Mg with diopside; S with Canyon Diablo troilite.