40Ar-39Ar age of Northwest Africa 091: More evidence for a link between L chondrites and fossil meteorites


  • J. R. WEIRICH,

    Corresponding author
    1. Department of Planetary Sciences, The University of Arizona, 1629 E. University Blvd., Tucson, Arizona 85721, USA
      Corresponding author. E-mail: jweirich@asu.edu
    Search for more papers by this author
    • Present address: School of Earth and Space Exploration, Arizona State University, 781 E. Terrace Rd., Tempe, Arizona 85287, USA.

  • T. D. SWINDLE,

    1. Department of Planetary Sciences, The University of Arizona, 1629 E. University Blvd., Tucson, Arizona 85721, USA
    2. Department of Geosciences, The University of Arizona, 1040 E. 4th St., Tucson, Arizona 85721, USA
    Search for more papers by this author

    1. Department of Geosciences, The University of Arizona, 1040 E. 4th St., Tucson, Arizona 85721, USA
    Search for more papers by this author

Corresponding author. E-mail: jweirich@asu.edu


Abstract— Most 40Ar-39Ar ages of L chondrites record an event at approximately 500 Ma, indicating a large collisional impact at that time. However, there is a spread in ages from 400 to 600 Ma in these meteorites that is greater than the analytical uncertainty. Identification of, and correction for, trapped Ar in a few L chondrites has given an age of 470 ± 6 Ma. This age coincides with Ordivician fossil meteorites that fell to Earth at 467 ± 2 Ma. As these fossil meteorites were originally L chondrites, the apparent conclusion is that a large impact sent a flood of L chondrite material to Earth, while material that remained on the L chondrite parent body was strongly heated and reset. We have reduced 40Ar-39Ar data for Northwest Africa 091 using various techniques that appear in the literature, including identification and subtraction of trapped Ar. These techniques give a range of ages from 455 to 520 Ma, and show the importance of making accurate corrections. By using the most straightforward technique to identify and remove a trapped Ar component (which is neither terrestrial nor primordial), an 40Ar-39Ar age of 475 ± 6 Ma is found for Northwest Africa 091, showing a temporal link to fossil meteorites. In addition, high temperature releases of Northwest Africa 091 contain evidence for a second trapped component, and subtraction of this component indicates a possible second collisional impact at approximately 800 Ma. This earlier age coincides with 40Ar-39Ar ages of some H and L chondrites, and lunar samples.


Collisional impacts have played an important role in the formation of our solar system. In addition to impact craters being a dominant surface feature on objects such as the Moon, Mercury, and Mars, heating during collisions has been the dominant thermal event on asteroidal bodies for the past approximately 4.4 Ga. This heating will cause diffusive loss of argon (Ar), and hence the age of the collisional impact can be determined by 40Ar-39Ar dating (a variant of K-Ar dating). In addition to the heating, unmelted portions of the asteroid involved in a collision will record shock imprints such as mosaicism and planar deformation features (Stöffler et al. 1991). Therefore, shocked meteorites can be used to understand the collisional impact history of their parent asteroid. Ordinary chondrites (OCs) are the most common types of meteorites falling to Earth, and have been well studied with the 40Ar-39Ar method (Turner et al. 1978; Bogard 1995).

Most unshocked OC meteorites give 40Ar-39Ar ages of 4.4–4.5 Ga, attesting to cooling within the first approximately 100 Ma after formation (Turner et al. 1978; Trieloff et al. 2003). Heavily shocked OCs, on the other hand, often have ages much younger, usually <1.3 Ga (Bogard 1995). Because all young chondrites have also experienced heavy shock, thermal events on these parent bodies after approximately 4.4 Ga are due to collisional impacts. The impact that resets the 40Ar-39Ar chronometer is typically not the impact that sends the meteorite to Earth, since the cosmic ray exposure (CRE) ages for OCs are typically <50 Ma, while 40Ar-39Ar ages are much older (Bogard 1995). Meteorites can certainly arrive on Earth after an 40Ar-39Ar resetting event, but unless the impact occurred in the last few Ma, it would not survive weathering on the surface to be collected in the present time (Bland et al. 2006). Fossil meteorites are an obvious exception, and will be discussed below. While there are many collisional events in the asteroid belt, many of which are recorded on the L chondrite parent body, we will only talk in detail of those that are relevant to interpreting the 40Ar-39Ar age of the meteorite we are studying, Northwest Africa (NWA) 091.

For the shocked L chondrites (the most common type of OC), about half the 40Ar-39Ar ages are approximately 500 Ma, indicating a large collision at that time, possibly large enough to cause breakup of the parent body (Anders 1964; Heymann 1967; Bogard et al. 1976; Bogard and Hirsch 1980; McConville et al. 1988; Keil et al. 1994; Bogard 1995; Kunz et al. 1997; Korochantseva et al. 2007). However, rather than a tight clustering of ages, there is a spread from about 400 to 600 Ma, which is larger than the uncertainty for individual meteorites. While trapped (pre-existing) 40Ar is always a concern, and is a likely source of the older ages, identification of trapped Ar is often hampered by Cl-activation during neutron bombardment. Neutron bombardment is a necessary step of 40Ar-39Ar dating because it converts a fraction of the 39K to 39Ar. Unfortunately, low-energy thermal neutrons will also convert 37Cl to 38Cl, which then rapidly decays to 38Ar, one of the isotopes which would ideally be used to identify the trapped component, making identification of that component difficult. However, shielding samples with cadmium (Cd) during neutron irradiation can block the problematic thermal neutrons while allowing the neutron captures necessary for the 40Ar-39Ar technique, making more accurate identification of components possible, as will be explained below.

Fossil meteorites found in a Swedish quarry opened the possibility of getting independent chronological information about L chondrites that give an approximately 500 Ma age (Schmitz et al. 2001). Fossil meteorites are meteorites that fell to Earth and were then buried before they could deteriorate on the sediment surface. In this case, meteorites fell to the ocean floor and were then buried under sediment. Most of the minerals have been replaced, although original textures and relict chromite grains exist and the oxygen isotopes, chemical composition, and chondrule size are all consistent with L chondrites (Schmitz et al. 2001; Bridges et al. 2007; Heck et al. 2010). These fossil meteorites have very short CRE ages of 0.1–1 Ma, and are found in a limestone layer that is 467.3 ± 1.6 Ma (Gradstein et al. 2004; Heck et al. 2004, 2008). In addition, these CRE ages increase with higher stratigraphy. These results show that at approximately 470 Ma, meteorites were ejected from their host body, and meteorites that were in space longer did indeed fall to Earth at a later time. Because the 40Ar-39Ar age of many L chondrites are approximately 500 Ma, and because the number of fossil meteorites implies a substantially higher flux of L chondrite material than at present, it has been suggested that both the fossil meteorites and the recent L chondrite falls were involved in an event that disrupted the L chondrite parent body at approximately 470 Ma (Schmitz et al. 2003). An approximately 470 Ma impact age for the L chondrites is not surprising, since that age did frequently come up in the literature prior to discovery of the vast majority of fossil L chondrites. Peace River was found to have an 40Ar-39Ar age of 450 ± 30 Ma (McConville et al. 1988), while Chico had a Rb-Sr age of 467 ± 15 Ma (Fujiwara and Nakamura 1992; Bogard et al. 1995). The question then becomes, why do some meteorites give an age of approximately 470 Ma, while others do not? Korochantseva et al. (2007) were able to identify trapped components in a set of L chondrites, and removal of these components gave a weighted average of 470 ± 6 Ma, the same as the age of the limestone containing the fossil meteorites (Gradstein et al. 2004). The identification of the trapped components was only possible because of Cd-shielding during neutron irradiation, and led to strong circumstantial evidence for a link between an increased flux of L chondrite material to Earth at approximately 470 Ma, and L chondrite material currently falling to Earth that records disruption of a parent body at approximately 470 Ma.

While not nearly as dramatic as the approximately 500 Ma event on the L chondrite parent body, there is also evidence for an approximately 800 Ma impact event on the Moon (Zellner et al. 2009). Impact glasses from different Apollo landing sites, with different compositions (and hence different source regions), give a tight cluster of ages near 800 Ma (Zellner et al. 2009). This is also the accepted age of the Copernicus event (Bogard et al. 1994). In addition, Cat Mountain (L chondrite) and LAP 031308 (H chondrite), both have 40Ar-39Ar ages close to 800 Ma (Kring et al. 1996; Swindle et al. 2009). The lunar impact glasses combined with the chondrites suggest the possibility of an increase in the impact flux at that time (Zellner et al. 2009). While a temporal relationship between impacts on different bodies does not mean they are causally linked, it does open up the possibility that a disruption event will send material on a collision course with other objects. This study of NWA 091, an L chondrite, is intended to test the idea of a temporal relationship between impact events on the L chondrite parent body and other objects.

For reasons described below, Cd-shielding is often not used during neutron irradiation of meteorites, and therefore a variety of techniques have appeared in the literature that attempt to overcome this limitation. However, the most straightforward technique is to use Cd-shielding, which allows for identification of trapped Ar components (McDougall and Harrison 1999). For this reason, NWA 091 was irradiated with Cd-shielding to provide an opportunity to search for these trapped Ar components, without relying on techniques to circumvent Cl-activation. We reduce the Ar data using the variety of techniques that appear in the literature, and then test their accuracy by comparing these results to the more accurate results acquired using trapped Ar components.

Samples And Methods

The NWA 091 has been classed as an L6 S4 (Grossman and Zipfel 2001). Three whole-rock splits of approximately 20 mg of NWA 091 were used for 40Ar-39Ar analysis, and labeled NB2 through NB4. Ar diffusion information on two whole-rock splits of approximately 15 mg of NWA 091, which have been labeled NB1 and NB5, is presented in Weirich (2011). Ar data and correction factors related to the splits of NWA 091 used in this work can be found in Supporting Information Data S1. While the splits used for diffusion (NB1 and NB5) are different from those used for age analyses (NB2 through NB4), all samples were irradiated in the same quartz tube, so the same reactor parameters (including corrections and method of calculating J values) were used. Errors include uncertainty in J, but not in the decay constant.

To obtain accurate 40Ar-39Ar ages, we must identify and mathematically remove both the trapped and cosmogenic contribution to 38Ar and 36Ar. Cosmogenic Ar is produced by cosmic ray spallation of K, Ca, and Fe. The isotopic composition depends upon on the relative concentration of K, Ca, and Fe, and burial depth, but a typical meteoritic value is 38Ar/36Ar ≈ 1.54 (in calculations we assume an uncertainty of 0.02) (Wieler 2002). We use the term “trapped Ar” to refer to any component that has an identifiable 40Ar/36Ar and/or 38Ar/36Ar ratio that is not cosmogenic. Two common sources of trapped Ar are a primordial component called Q, and terrestrial atmospheric Ar. Atmospheric Ar has a 40Ar/36Ar of 295.5 ± 0.5 while 38Ar/36Ar is 0.1869 ± 0.0004 (Nier 1950; Steiger and Jager 1977), although a recent study favors slightly larger ratios (Lee et al. 2006). In all calculations we use the earlier ratios, although using the ratios from the recent study would not change our conclusions. Q has a 40Ar/36Ar << 1, while 38Ar/36Ar is 0.1873 ± 0.0007 (Busemann et al. 2000). Because the 38Ar/36Ar of Q is the same as atmospheric Ar within error, we use the atmospheric Ar ratio for both trapped components. If the only contributions to 38Ar and 36Ar are cosmogenic and trapped Ar, the two can be separated by a simple component balance between a cosmogenic 38Ar/36Ar ratio of 1.54 ± 0.02, and a trapped ratio of 0.1869 ± 0.0004. Errors in the trapped and cosmogenic components are propagated into the final 36Ar error. The Identification and Removal of Spallation section details the subtraction of cosmogenic Ar using the component balance mentioned above, as well as an independent subtraction method using 37Ar. Both methods yield very similar results, showing that the trapped 38Ar/36Ar ratio is indeed approximately 0.19 for most extractions.

The presence of Cl-derived 38Ar prevents such a separation between cosmogenic and trapped Ar from being performed, so special attention is given to identification of its presence. The techniques used for identification of trapped Ar are discussed in the relevant sections below. At this point, it might seem that elimination of Cl-derived 38Ar (for instance, by Cd-shielding) would have become standard long ago for 40Ar-39Ar dating. However, terrestrial samples, which are the subject of the vast majority of 40Ar-39Ar studies, are shielded from cosmic rays by the Earth’s atmosphere, so there is no need to separate out cosmogenic Ar, and hence no need to use 38Ar in most studies.

Identification of, and correction for, trapped Ar often makes use of a three-isotope plot, or isochron. Isochrons are fit with a linear least squares method. Because the formal errors are based solely upon the number and error of individual points, it does not determine if the results are significant, hence we need a test to indicate if the result has meaning. We use the goodness-of-fit parameter (i.e., the P-test, or just P) defined by Press et al. (1992) for this purpose, and only accept a fit if it has a P > 0.05. The final apparent age for each extraction includes the error of the trapped Ar component determined from the isochron. The ages of the three splits were combined in two ways. The first is a mean value weighted by the inverse square of the errors, and the second combination is an unweighted mean with standard deviation for errors. We follow a conservative approach, and report the values that have the largest error. All errors are listed as 1σ.


Data Reduction with Spallation Correction

The most straightforward (and therefore presumably the most accurate) application of the 40Ar-39Ar technique is discussed in this subsection. All three splits of NWA 091 have very similar release patterns for all isotopes of Ar. For this reason, we will only give a detailed description of NWA 091 split NB4, which had more extractions than the other two splits. Where relevant, differences between splits are discussed. A summary of all three splits is given in the Comparison of Splits section.

Identification and Removal of Spallation

Even without a spallation correction, some information can be gleaned from a sample by creating a plateau plot and isochron (Bogard et al. 2010). If a trapped component is present, but not removed, the apparent age of each extraction will be artificially old. Figure 1 shows the plateau plot with no trapped correction. We can see the effect of trapped gas below approximately 40%39Ar released in all three splits, manifested as decreasing ages with each successive extraction as a low temperature trapped component (perhaps terrestrial contamination) is exhausted. The gas above approximately 40%39Ar released appears to be partially reset, although the whole spectrum could also be interpreted as a saddle-shaped plateau, which would indicate trapped gas at both low and high temperatures.

Figure 1.

 Plateau plot of NWA 091, without a trapped correction. Boxes represent 1σ errors.

A reverse isochron is a plot of 36Ar/40Ar versus 39Ar/40Ar. In a well-behaved system without partial resetting, the y-intercept is the inverse of the trapped 40Ar/36Ar ratio, and the x-intercept can be used to calculate the age. However, if one does not first remove spallation, the resulting 36Ar/40Ar ratio for each extraction will be raised. If spallation has not been removed before plotting the data on a reverse isochron, the x-intercept can be pushed either higher or lower, depending upon whether spallation comes out in extractions that have a high or low 39Ar/40Ar ratio. Figure 2 shows the reverse isochron without a spallation correction for split NB4; symbols are for illustrative purposes only, although circles, triangles (upright and inverted combined), and squares represent consecutive extractions. The symbol shape for each extraction is the same in Figs. 2, 3, 4 and 7, although the fill state changes depending upon which extractions are relevant for each figure. To properly distinguish trapped 36Ar from spallation-produced 36Ar, we must use 38Ar (see the Samples and Methods section). To receive useful information from 38Ar, we must make sure 38Ar does not contain Cl-activated argon. Hence, we utilize Fig. 3, involving 37Ar, which is not found naturally in the rock, but is produced by the neutron irradiation of Ca. In a plot of 37Ar/36Ar versus 38Ar/36Ar, a two-component mixture will form a straight line. In this case, the first component is trapped gas (37Ar/36Ar = 0 and 38Ar/36Ar = 0.187) and the second is Ca-spallation (37Ar/36Ar is inversely proportional to the CRE age and 38Ar/36Ar = 1.54). Any deviation from a straight line is due either to large contributions from K and/or Fe spallation, or Cl-activation.

Figure 2.

 Reverse isochron without a spallation correction for NWA 091 split NB4. Error bars are typically smaller than the symbol. Filled circles and squares form linear patterns. A linear fit combining filled triangles and filled circles does not form a statistically rigorous line, but the two populations are still similar. Quarter filled circles and open inverted triangles are not included in any of the fits. The three groups indicated by circles (filled and quarter filled), triangles (upright and inverted), and squares contain consecutive extractions from split NB4; the symbol shape (but not necessarily the symbol fill) used for each extraction remains the same in Figs. 2 through 5. Even without a spallation correction two trapped components are visible, but are not as linear as after spallation correction.

Figure 3.

 Plot showing the relationship among 38Ar, 37Ar, and 36Ar. Symbol shape is the same as in Fig. 2. Filled symbols indicate extractions that form a straight line, while open symbols indicate extractions that deviate from a straight line. Symbols for individual points are the same as in Fig. 2. An unweighted fit to the filled symbols is also shown.

Figure 4.

 Reverse isochron after 36Ar spallation removal. Symbol shape is the same as in Fig. 2. One trapped component is represented by filled circles and triangles, the other by filled squares. A solid line shows the weighted fit of the filled circles and triangles, while a dashed line shows the weighted fit of the filled squares.

Figure 7.

 Isochron plot referenced to 37Ar instead of 36Ar. Symbol shape is the same as in Fig. 2. Filled symbols indicate points used for linear fit, open symbols indicate points that were not used. Symbols for individual points are the same as in Fig. 2. Line is the weighted linear fit to the filled symbols. Many other extractions plot off-chart, well above, and to the right. Isochron age is 476 ± 28 Ma.

First, we note that no extraction has a 38Ar/36Ar ratio above the spallation value of 1.54. If there were extractions with such a high 38Ar/36Ar ratio, it would be an obvious sign of Cl-activation, because that reaction would produce only 38Ar (again, see the Samples and Methods section). In fact, none have a 38Ar/36Ar above 0.7, indicating that all extractions have a significant amount of trapped gas, further confirmation that using the method of the maximum measured 37Ar/36Ar ratio to remove spallation (described in the Data Reduction section without using 38Ar) would lead to erroneous values.

The linearity of most of the extractions in Fig. 3 suggests that any deviation from a 38Ar/36Ar of 0.187 for those extractions is indeed due to Ca-spallation. For the six extractions that do not fall on the line, obvious culprits are spallation from K and/or Fe, and Cl-activation. The 39Ar/37Ar ratio (proportional to the K/Ca ratio) of these six extractions is lower than previous extractions, indicating that the spallation production from K is a smaller contribution, leaving Fe or Cl as the culprit. These six extractions were all released consecutively between 850 and 1050 °C, around the eutectic temperature of the troilite/Fe-Ni metal system, hinting at Fe spallation. Cl-activation is still a possibility, although below we show this is probably not the case.

We have two choices to remove the 36Ar due to spallation. The first is to extrapolate the linear portion of Fig. 3 to a 38Ar/36Ar of 1.54, and use that value of 37Ar/36Ar to subtract out the Ca-spallation. The other method is to use only 38Ar and 36Ar, and assume the measured 38Ar/36Ar ratio for each extraction is due to mixing various amounts of trapped gas (38Ar/36Ar of 0.187) and K/Ca/Fe spallation (38Ar/36Ar of 1.54). We use the latter method because it incorporates spallation from all three elements instead of just Ca, although the former method gives very similar results. The spallation-corrected reverse isochron is shown in Fig. 4. The biggest difference between the two methods of spallation reduction occurs, unsurprisingly, for the six extractions that fell off the line in Fig. 3. Three of the six extractions are the rightmost filled triangles of Fig. 4, which are now colinear with the rest of the points, which suggests that Fe spallation was indeed the culprit. The other three extractions are the open inverted triangles between a 39Ar/40Ar ratio of 0.015 and 0.02, which are now close to, but still slightly above, the line. Cl contamination does not explain these extractions, because that would have the effect of removing too much 36Ar, causing them to fall below the line. These extractions remain a mystery, although they are now at least within 2σ of the line. It should be noted that these six extractions are not bad measurements. While some of the isotopes are challenging to explain, others are quite normal. Hence, we only exclude them when they deviate from the established pattern, and only then with sufficient justification.

Identification and Removal of Trapped Gas

Having removed the contribution from spallation, we can now determine the 40Ar/36Ar ratio of the trapped component, and thus determine apparent ages at each temperature. To do this, we actually reduce the data twice. We first include all the gas that reasonably forms a line in the reverse isochron plot, excluding the problematic extractions discussed in the above paragraph, and perform the full reduction with that data (i.e., trapped component and plateau age). We then determine which extractions have an age that is zero to within 2σ, and perform the entire reduction again without these extractions. The reason for excluding the extractions that have an age of zero to within 2σ is because we do not wish to include portions of the release that have undergone recent partial resetting. In addition, these extractions have large errors, and will only inflate the final error without providing much additional information. For each split, the first and second reductions give plateau ages that are the same to within 1σ.

For each instance of line fitting on the reverse isochron plot, we perform a weighted fit of the previously selected values. In both cases the P (i.e., the goodness-of-fit, see the Samples and Methods section) does indeed indicate a statistically significant line. The y-intercept of this line then gives the 36Ar/40Ar ratio for the trapped component. We then use this trapped 36Ar/40Ar ratio and the measured amounts of 36Ar to calculate the amount of trapped 40Ar in each step, which we then subtract from the total amount of 40Ar, leaving only radiogenic 40Ar. For the plateau plot, the plateau age is calculated by summing together gas from all the extractions that were included in the linear fit. Various alternative methods of calculating a plateau age (weighting by errors of individual extractions, weighting by number of extractions, etc.) all give plateaus ages that are within the stated errors. For split NB4, the final selection of points is indicated by filled symbols in Fig. 4, and the final plateau plot is shown in Fig. 5. Also shown in Fig. 4 are the two linear fits showing the two different trapped components.

Figure 5.

 Plateau plot of NWA 091 split NB4. Boxes indicate 1σ errors. The jump at approximately 90%39Ar released is due to changing the value of the trapped component (see Table 1 for the two trapped component values). The broken portion of the long double arrow indicates the steps that were not included in the plateau. Data have been corrected for blank, decay, reactor interference, mass discrimination, spallation, and trapped argon.

The above scenario works very well for splits NB2 and NB4. Split NB3 requires special treatment. After the first round of data reduction, one of the extractions on the plateau plot falls below the plateau and has fairly large errors, but not so large that they overlap zero at 2σ. We do not wish to include extractions with partial resetting, yet at the same time this extraction has a very small amount of 39Ar and thus provides the most accurate measurement of the trapped component. Ultimately we have decided to reduce the data both with and without this extraction, and refer to these two reductions as “reduction 1” and “reduction 2,” respectively. Note that neither of these reductions include the points whose apparent age is within 2σ of zero.

Comparison of Splits

The trapped component and plateau ages for all splits are listed in Table 1, and all plateau plots are shown in Fig. 6. NB3 in Fig. 6 shows the plateau plot for “reduction 2.” Isochron ages are not reported here, but are the same as the plateau ages within error, as they should be, given the mathematical relationship between the two presentations. The first 3% of the gas of NB2 and NB4, and the first 5–7% of gas in NB3, had a very large trapped contribution to the total 40Ar, and hence after the removal of trapped gas those steps have very large errors on the plateau plot. Some of this gas also probably contained adsorbed terrestrial air. Hence, it is not used to calculate the plateau ages, although in split NB3 we did include a small amount of it for one of the reductions. The gas that is released between ∼32 and ∼42%39Ar released (depending on the split) gives an apparent age less than the plateau. On the reverse isochron plots, these extractions correspond to values that are slightly above the linear fit, and on the release plots they correspond to a transition from the low temperature to high temperature release, suggesting that they represent the earliest-degassing portion of a second phase or diffusion domain (see Weirich [2011] for the same effect in a different split). These extractions probably indicate weathering or partial resetting, and are also not included in either the plateau age or the linear fit on the reverse isochron. Including the first few extractions of each split and the gas between the high and low temperature release produces an approximately 3–12 Ma change in the plateau age, but increases the error bars from ∼10 Ma to ∼20–27 Ma.

Table 1.   Summary of trapped components and plateau ages of NWA 091.
NWA 091 splita 40Ar/36Ar trappedNo. of extractions%39Ar in plateauAge (Ma)c
  1. aNB2 through NB4 are internal identifiers used for the splits of NWA 091.

  2. bTwo reductions for split NB3 are included because there are reasons to both include and exclude one of the extractions.

  3. cAll errors are listed as 1σ.

  4. dThis is the isochron age for this gas. The plateau age is the same, although the error is about a factor of five lower, casting doubt on its accuracy.

bNB3 reduction 1278.7 ± 6.4781.6483.4 ± 13.2
bNB3 reduction 2289.5 ± 9.1679.3468.5 ± 14.5
NB2282.7 ± 4.81480.0475.8 ± 9.2
NB4285.6 ± 3.91981.6474.2 ± 9.1
Final plateau  ∼80475.4 ± 5.9
NB4 component age90.9 ± 5.5911.6799.0 ± 70.7d
Figure 6.

 Plateau plot of all three splits of NWA 091. Boxes indicate 1σ error. Data have been corrected for blank, decay, reactor interference, mass discrimination, spallation, and trapped argon. See text for further details, and Table 1 for trapped argon values.

The plateau plots of the three splits are all very consistent, although split NB4 deviates from the other two splits above approximately 90%39Ar released. This is because low resolution prevented us from identifying a second trapped component in the other two splits. Hence, we used a single trapped component for the entire release, when we most likely have two components for all three splits. Because we “removed” a trapped component that had an inappropriately high 40Ar/36Ar, these splits have an apparent age <∼500 Ma above 90%39Ar released. If we were to use the second trapped component identified in split NB4, and apply it to the last 10% of gas released in the other two splits, three of the four large extractions in split NB2 are 800 Ma to within 2σ, and the large extraction in split NB3 is approximately 1100 Ma. While NB3 gives a high temperature age that is not consistent with an 800 Ma event, NB2 does hint at an event at 800 Ma, indicating this event is probably real.

The gas below approximately 90%39Ar released (with the minor exceptions noted above) appears to form a plateau, with all steps within 2σ of the summed age, and all three splits agree on the age to within 1σ. To combine both reductions for NB3, we average the plateau ages together, and use the largest error of the two reductions. This gives 476.0 ± 14.5 Ma. Now, we can combine the three splits together to form a mean age weighted by the inverse of the errors, with a final error equal to the inverse sum of the squared errors. This gives an event at 475.4 ± 5.9 Ma. Interestingly, the gas from approximately 5–30%39Ar was identified in Weirich (2011) as corresponding to the low temperature release (shocked feldspar), while that of the gas from approximately 40–90%39Ar released was identified as corresponding to a high temperature release (the portion that could be modeled as pyroxene). Both of these Ar domains contribute to the approximately 470 Ma plateau age above. The gas above approximately 90%39Ar released corresponds to a second high temperature release mentioned in Weirich (2011), the release pattern that could not be easily modeled, and gives a component age of approximately 800 Ma.

Data Reduction without Using 38Ar

In this subsection we reduce the 40Ar-39Ar data using various techniques that have been used in the past, all of which attempt to circumvent the limitations imposed by Cl-activation. All of these approaches warrant caution, but are included here to show the techniques most frequently used on meteorites. Here, we use a reverse isochron without a spallation correction, attempt to remove spallation using the minimum 36Ar/37Ar ratio, and use a three-isotope plot normalized to 37Ar. Figure 2 shows the reverse isochron with no spallation correction for NWA 091 split NB4, which has the highest temperature resolution of the three splits. From Fig. 2, NWA 091 appears to have two trapped components, indicated by the two linear portions with different y-intercepts, represented by filled circles and filled squares. The filled circles form a linear pattern, but the filled triangles seem as though they could be part of this line as well. All of these points correspond to extractions at less than approximately 1200 °C, or approximately 90%39Ar released. A weighted fit to the filled circles gives an isochron age of ∼500 ± ∼30 Ma, and gives a statistically significant linear fit (see the Samples and Methods section). Fitting the filled circles along with the filled triangles gives an isochron age of ∼420 ± ∼10 Ma, but the P value indicates it is certainly not a statistically significant linear fit (P<<0.01).

These two isochrons nearly span the range of reported ages for most shocked L chondrites, although the best fit gives an age of approximately 500 Ma, the same (but now believed to be slightly old, see the Comparison of Splits section) age that was previously thought to be the catastrophic impact age. The other, steeper, linear portion (filled squares) gives an age of ∼800 ± ∼60 Ma. The other two splits show a similar story, and the best-fit ages are approximately 470 ± 30 Ma for split NB2 and ∼550 ± ∼35 Ma for split NB3, although no steeper linear portion is found in these splits. An unweighted average of the shallow portion of all three splits, with standard deviation for errors, gives ∼510 ± ∼40 Ma.

One technique to remove the contribution to 36Ar from spallation is to assume that the spallation is all due to Ca and that the 36Ar in the extraction with the minimum 36Ar/37Ar ratio is all due to spallation, then remove that fraction of gas from all extractions (Garrison et al. 2000; Benedix et al. 2008; Swindle et al. 2009). For many samples these assumptions appear to be valid, and if significant Cl-derived 38Ar is present it may be the only option. In fact, we have used this technique ourselves (Swindle et al. 2009). However, there are two potential problems with this technique. The first is that (without using 38Ar) there is no independent check to see if in fact the minimum 36Ar/37Ar ratio has no trapped gas. If this assumption is wrong, too much 36Ar will be removed. Even if that assumption is correct, it will not fully correct extractions that have a large spallation contribution from K and/or Fe. As with no removal of spallation, the resulting age can be pushed in either direction depending upon the specifics of the sample.

For split NB4, the minimum 36Ar/37Ar occurs for two extractions at approximately 0.011, both well within 1σ of each other. After applying this correction, many of the points that were close to the shallow linear portion in Fig. 2 would be well below any fitted line, a strong indication we have removed too much 36Ar. For the points that remain, we get an isochron age of ∼535 ±∼15 Ma, again older than that determined using trapped Ar components, although within the range reported for many shocked L chondrites. The steeper portion now gives an isochron age of ∼900 ± ∼65 Ma. The other two splits give ∼520 ± ∼30 Ma for split NB2 and ∼510 ± ∼30 Ma for split NB3, again with no steeper linear portion. A weighted mean of the three splits, with errors equal to the inverse sum of the squared errors, gives ∼519 ± ∼13 Ma. Again, we get an age very close to 500 Ma.

One final technique has been used on a few meteorites (Bogard et al. 2010), although it is unclear why any meteorite would meet the condition required to obtain a meaningful result. This technique is to make an isochron plot of 40Ar/37Ar versus 39Ar/37Ar, with the slope of any linear portion giving the age. In order for this technique to work, it requires the trapped Ar to be correlated with Ca for at least a portion of the release. Since Ca is primarily contained in pyroxene in L chondrites, this means that the trapped Ar would also have to be related to pyroxene. Although it is not obvious why this should be so, the linearity seen on many such plots suggests that the technique may have some validity.

As is seen in Fig. 7, there is indeed an obvious linear portion in split NB4 (indicated by filled symbols), and the isochron age is 476 ± 28 Ma. The linear portion of the gas is released between approximately 900 and 1200 °C, and corresponds to the temperature range over which pyroxene releases its gas (Weirich et al. 2012). Evidently the trapped Ar is contained in the pyroxene, and is correlated with Ca. Reasonable fits for split NB2 give ages from ∼400 to ∼475 Ma with individual errors of ∼20 to ∼35 Ma. While that does seem promising, the more reasonable fits for NB2 are at the lower end of the isochron ages, and split NB3 has no linear portion at all. Without good criteria for which isochron of NB2 to pick, we pick an age in the middle, and assign error bars large enough to encompass both extremes. This gives an age of ∼440 ± ∼40 Ma for NB2. An unweighted average of NB2 and NB4, with standard deviation for errors, is then ∼457 ± ∼27 Ma. Although this result is comparable to the ages found using trapped Ar components (although with fairly large errors), there remains the question of why trapped Ar would be correlated with Ca.

All of the techniques in this subsection leave something to be desired. After averaging all the splits together, the different techniques give a range from 455 to 520 Ma. For an individual technique on an individual split, we can often get a similar range of ages, depending upon which points we include in the linear fit. Previous authors have often only run a single split of a meteorite, which makes deciding on an age even more subjective. To give an accurate and precise age for a meteorite, Cd-shielding needs to be used during irradiation to allow for identification of spallation-produced Ar. The techniques that attempt to circumvent the limitations imposed by Cl-activation do give ages that are within error of those obtained using Cd-shielding during irradiation, but they have larger errors and scatter to both sides.


The NWA 091 upon first inspection appears to be a typical heavily shocked L6 meteorite that records an impact event at approximately 500 Ma. This age is obtained by using techniques found in the literature that circumvent Cl-activation, which is typical of meteorites that were not irradiated with Cd-shielding. The variety of techniques give a range of ages, and often the same technique applied to different splits of NWA 091 gives a similar range of ages. However, because NWA 091 was irradiated with Cd-shielding that prevented Cl-activation, we have been able to examine this meteorite in more detail, allowing us to identify any trapped components in a more straightforward way. The removal of this trapped Ar leads to an 40Ar-39Ar age of 475.4 ± 5.9 Ma (1σ), similar to what Korochantseva et al. (2007) found after removal of trapped Ar in other L chondrites. Korochantseva et al. (2007) noted that this age coincides with an increase of the influx of L chondrite material to Earth, which is now seen as fossil meteorites in sedimentary layers dated to 467.3 ±1.6 Ma (1σ) (Heck et al. 2004, 2008). In addition, Schmitz et al. (2001) noted a large number of terrestrial impact crater ages in the range of 450–480 Ma. This evidence is suggestive of parent body breakup at approximately 470 Ma, which leads to a cascade of material to Earth, some of which created impact craters and some of which was preserved as fossil meteorites. Even if it is shown that the synchronicity of well-dated L chondrites and fossil L chondrites is only coincidental, NWA 091 does provide further evidence for a large impact event on the parent body at 465 to 475 Ma instead of at 450 to 550 Ma.

The presence of a trapped component with a 40Ar/36Ar between 1 and 295 is surprising. Trapped components might be expected to be terrestrial air or primordial gas, with nothing in between, but these are all between 275 and 290. NWA 091 split NB3 “reduction 2” has a trapped 40Ar/36Ar ratio that has overlapping 1σ error bars with terrestrial air, but the other reduction and the other two splits do not overlap until >2σ (Table 1). This component appears to be distinct from “air,” but is close enough that they could be related.

Korochantseva et al. (2007) noted that the low temperature isochron they identified could be due to slightly fractioned terrestrial argon acquired through weathering. In the case of NWA 091, that is more difficult because the gas from approximately 40–90%39Ar released is the gas that is identified in Weirich (2011) as having an activation energy similar to pyroxene, and is certainly not the low temperature gas. The linearity of the isochron referenced to 37Ar shows that the trapped component is indeed contained in the pyroxene. The pyroxene in the thin section of NWA 091 we analyzed is unweathered, and NWA 091 has a weathering grade of W2 (Grossman and Zipfel 2001), meaning weathering has only affected metal and troilite (Bland et al. 2006). Hence, we find weathered pyroxene an unlikely possibility. The other explanation by Korochantseva et al. (2007) for the trapped Ar component is an equilibration between primordial argon (see the Samples and Methods section) and mobilized radiogenic argon. While this does explain our results, it begs another question. In order for a heating event to cause equilibration of argon, instead of simply causing argon to be lost to space, a nonzero argon pressure must be maintained at the grain surface. It is currently unclear how that would occur, and why the 40Ar/36Ar ratio would be so close to air.

In addition to a trapped component that appears to be contained in shocked feldspar and pyroxene, a second trapped component (which is definitely not terrestrial) is found in the gas that corresponds to the second high temperature release identified in Weirich (2011). This second trapped component has a 40Ar/36Ar ratio different from the first trapped component, but the 38Ar/36Ar ratio of both components is approximately 0.19. Therefore, the spallation correction is still accurate. The source of this second trapped component is unknown. Because this release occurs at high temperature, it is possible that this release is not an actual domain, but instead due to some other effect (e.g., recoil, melting, mixing of different Ar reservoirs). Recoil of 39Ar during irradiation from a fine-grained K-rich phase to a K-poor phase cannot explain the second high temperature release because 40Ar is associated with 39Ar (Figs. 5 and 6). Melting of a domain that has not fully exhausted seems to be ruled out by examination of Fig. 8, which shows the Arrhenius plot of log (D/a2) versus reciprocal temperature for NWA 091 split NB4 for both high temperature releases (the low temperature release plots off-chart). Melting typically raises (D/a2), but for this release (which is seen to the left of the vertical dotted line) (D/a2) is lowered. Volume diffusion from a single reservoir would form a straight line as long as gas released from multiple reservoirs did not mix during extraction. While we cannot determine if this second high temperature release is in fact an Ar domain, linearity in a reverse isochron (Fig. 5), and possible linearity of the extractions between 1213 and 1250 °C in an Arrhenius plot (Fig. 8) allow for this to be the case. If there indeed is a second high temperature Ar domain in NWA 091, there are interesting consequences, namely that a single meteorite recorded two different impact events.

Figure 8.

 Arrhenius plot of NWA 091 split NB4. Points to the left of dotted vertical line indicate steps that have a different trapped component. Change in diffusion behavior is accompanied by change in trapped component. Error bars in both directions are smaller than the symbol.

A single meteorite recording two different thermal events can be understood by multiple Ar reservoirs that have different retentiveness. We know NWA 091 has two Ar reservoirs (a low and high temperature release) with different diffusion rates; if there is a third (a second high temperature release), it has a diffusion rate different from the other two. The impact resetting event at approximately 470 Ma was energetic enough to reset two of the reservoirs, but the potential third reservoir sustained little Ar loss and could record an earlier event at approximately 800 Ma due to a low diffusion rate. This can be seen as an obvious drop in the diffusion rate between the two releases (Fig. 8). In addition, the activation energy of the second high temperature release is the same or larger than the first high temperature release, although it is hard to quantify by how much. Even a small difference in activation energy can allow one reservoir to fully outgas and leave the other mostly unaffected, depending upon the temperature regime over which Ar loss occurred. For example, heating to 800 °C for 109 seconds would completely reset a domain with log D0/a2 = 12 s−1 and E = 104 kcal mol−1, but cause only 15% loss in a domain with E = 116 kcal mol−1 and the same log D0/a2. Without a more accurate measurement of the diffusion rate, as well as an estimate of the time-temperature history, we cannot prove that one reservoir could be completely reset without affecting the second reservoir. However, given the information we have it is possible.

If the second high temperature release is in fact recording a second impact event at approximately 800 Ma, it would provide evidence for another, earlier, inner solar system bombardment that affected multiple objects. Impact glasses on the Moon, as well as the Copernicus impact crater, have ages near 800 Ma (Bogard et al. 1994; Zellner et al. 2009), as do Cat Mountain (L chondrite) and LAP 031308 (H chondrite) (Kring et al. 1996; Swindle et al. 2009). However, like the impact events at approximately 470 Ma, a temporal relationship does not mean a causal relationship. Additional studies will be needed to determine if any of these events on various objects are casually related.


Korochantseva et al. (2007) found a possible temporal link between L chondrites and Ordovician fossil chondrites by identifying and removing a nonterrestrial and nonprimordial trapped Ar component from the measured 40Ar. Here, we have presented further confirmation of that link. We found the L chondrite NWA 091 to have a nonterrestrial and nonprimordial trapped Ar component, and subtraction of this component gives an 40Ar-39Ar age of 475.4 ± 5.9 Ma. Although other plausible data reduction techniques give relatively precise ages that scatter around this age, using multiple splits and best-practice data reduction yields an age that agrees with the fossil meteorites, with smaller uncertainties. The trapped Ar components do not appear to be due to terrestrial weathering, and their source is currently unknown.

In addition, we have shown that NWA 091 records evidence of a possible event at approximately 800 Ma, which coincides with a large scale resetting event on the Moon, as well as the age of some H and L chondrites (Zellner et al. 2009). While a single meteorite recording two separate events is unusual, different Ar reservoirs with different diffusion rates allow for that possibility, depending upon the thermal history of the meteorite. Further investigation will be required to confirm an approximately 800 Ma event recorded by NWA 091.

Acknowledgments–– E. Olsen graciously provided NWA 091. Funding by NASA Earth and Space Science Fellowship (NESSF) for J. R. W. and NASA grant NNX09AG88G to T. D. S. are gratefully acknowledged. The editorial handling by Associate Editor J. Spray and reviews by J. Park, B. Schmitz, and an anonymous reviewer helped improve this manuscript. This is part of J. R. Weirich’s Ph.D. dissertation.

Editorial Handling–– Dr. John Spray