Magnetic Properties of Tektites and Other Related Impact Glasses ∗ P

Home , Darwin (lunar crater), Wabar craters

Earth and Planetary Science Letters 432 (2015) 381–390

Contents lists available at ScienceDirect

Earth and Planetary Science Letters

www.elsevier.com/locate/epsl

Magnetic properties of tektites and other related impact glasses ∗ P. Rochette a, , J. Gattacceca a, B. Devouard a, F. Moustard a, N.S. Bezaeva b,c,d, C. Cournède a, B. Scaillet e a Aix-Marseille Université, CNRS, IRD, CEREGE UM34, 13545 Aix en Provence, France b Faculty of Physics, M.V. Lomonosov Moscow State University, Leninskie Gory, 119991 Moscow, Russia c Ural Federal University, 19 Mira Str., 620002 Ekaterinburg, Russia d Kazan Federal University, 18 Kremlyovskaya Str., 420008 Kazan, Russia e ISTO CNRS, Université d’Orléans, BRGM, UMR7327, 45071 Orleans, France a r t i c l e i n f o a b s t r a c t

Article history: We present a comprehensive overview of the magnetic properties of the four known tektite fields Received 7 July 2015 and related fully melted impact glasses (Aouelloul, Belize, Darwin, Libyan desert and Wabar glasses, Received in revised form 8 October 2015 irghizites, and atacamaites), namely magnetic susceptibility and hysteresis properties as well as properties Accepted 19 October 2015 + dependent on magnetic grain-size. Tektites appear to be characterized by pure Fe2 paramagnetism, Available online 3 November 2015 with ferromagnetic traces below 1 ppm. The different tektite fields yield mostly non-overlapping narrow Editor: B. Marty susceptibility ranges. Belize and Darwin glasses share similar characteristics. On the other hand the 2+ Keywords: other studied glasses have wider susceptibility ranges, with median close to paramagnetism (Fe and 3+ magnetic properties Fe ) but with a high-susceptibility population bearing variable amounts of magnetite. This signs a tektite fundamental difference between tektites (plus Belize and Darwin glasses) and other studied glasses in impact glass terms of oxygen fugacity and heterogeneity during formation, thus bringing new light to the formation crater processes of these materials. It also appears that selecting the most magnetic glass samples allows to find impactor-rich material, opening new perspectives to identify the type of impactor responsible for the glass generation. © 2015 Elsevier B.V. All rights reserved.

1. Introduction formation conditions. Tektites have been identified as a specific type of natural glass, obsidian-like but unrelated to volcanism Magnetic properties provide a rapid and versatile technique to (Suess, 1900; Koeberl, 1986; Glass, 1990; McCall, 2001). Tektites characterize non-destructively the composition of rare materials in are homogeneous materials made only of glass, not or poorly terms of content of magnetic elements (mostly Fe, but also Mn, Cr, vesiculated, that can be found geographically spread over a large Ni, and other trace magnetic elements), as well as oxidation state strewnfield (500–5000 km size range). The natural shape of tek- and distribution of these elements among various phases. Such an tites usually demonstrates flight in liquid state in the atmosphere approach has been exemplified in meteorites (e.g., Rochette et al., (splash-forms). Initial examples were the central European mol- 2008, 2012) where magnetic properties, mostly susceptibility, have davites and the Australasian tektites (Suess, 1900), but soon two been used for classification purposes. Magnetic measurements al- other strewnfields were identified, in Ivory Coast and North Amer- low to screen large collections, in museums in particular, and to ica (Glass, 1990; Koeberl et al., 1997). An impact origin has been single out anomalous samples worth of further investigations. established based on various arguments, including connection with High velocity impacts on Earth are able to generate high tem- large (>10 km diameter, see Table 1) impact craters (except in perature melted material that can be subsequently ejected away the Australasian case). Besides these four canonical examples, a from the crater and quenched as natural glasses (Dressler and fifth central American strewnfield has been proposed (e.g., Poven- Reimold, 2001). These impact glasses have specific composition mire and Cornec, 2015), originally named Tikal glass (Sentfle et al., and properties with respect to volcanic glasses, due to the na- 2000), although its characteristics are not as well defined as for ture of their source materials and high temperature and pressure the other four cases. Here is it will be named Belize glass as hun- dreds of specimens have been found in soils from this country. No * Corresponding author. consensus has been reached yet to rank the Belize glass among E-mail address: [email protected] (P. Rochette). tektites. However, it yields redox state and water content typical http://dx.doi.org/10.1016/j.epsl.2015.10.030 0012-821X/© 2015 Elsevier B.V. All rights reserved. 382 P. Rochette et al. / Earth and Planetary Science Letters 432 (2015) 381–390

Table 1 Magnetic susceptibility (χ ) measurements on various tektites and related impact glasses, ordered by average FeO wt%. Corresponding crater diameter (in km), when known, − is indicated within brackets. Average χ value in 10 9 m3/kg with s.d. normalized to the mean in %, range, N number of samples measured; poorly deﬁned s.d. within bracket (N < 8). FeO Average χ s.d. χ range N − (wt%) (10 9 m3/kg) (%) Tektites Moldavitea (24) 1.8 35 34 23 to 78 15 Moldavite (24) 1.8 31 19 25 to 60 39 Georgiaite (40) 2.6 49 (8) 44 to 52 3 Bediasite (40) 4 65 24 43 to 129 62 Australasitea 4.9 82 10 57 to 103 152 Ivoirite (10) 6.2 103 12 62 to 138 109

Other glasses Libyan desert normal 0.1 −2.3 39 −3.3 to −0.6 10 Libyan desert dark 0.1 4.4 71 −0.1 to 10.8 8 Darwin (1.2) 2.6 53 23 34 to 79 45 Aouelloul (0.4) 2.9 82 89 38 to 463 65 Irghizite (6–14) 5.6 164 59 103 to 791 91 Wabar (0.1) 6.1 468 58 125 to 1025 14 Belize glass 7.0 127 5 115 to 137 11 Atacamaite 8.5 302 286 84 to 20 500 3291 Atacamaite >0.5 g 8.5 191 71 84 to 1270 1401 Safford 0.6 380 181 5 to 2382 12 a Indicates measurements from the literature while the other lines correspond to this work, with one literature data for Belize glass. Note that Safford is a site of obsidian like glass presented in the discussion. FeO data is derived from the following references: Koeberl (1986); Koeberl et al. (1997, 1998), Giuli et al. (2002), Howard (2008), Hamman et al. (2013), Senflte et al. (2000). Note that total iron is reported as FeO. of tektites (Giuli et al., 2014). Other “tektite-related” impact glasses tions using portable instruments (MSB2 and SM30 instruments) (i.e. with fully melted splash-forms) have been described and will and following procedures and calibration described in Sagnotti et be discussed here. al. (2003), Gattacceca et al. (2004), and Rochette et al. (2009). The size of tektites is typically in the centimetre range, al- Raw signal after sample holder correction is proportional to sam- though much smaller samples have been found in the sedimentary ple mass, thus limiting the lower size measurable with sufficient − record and called microtektites (Glass, 1990). Microtektites have accuracy. MFK1, SM30 and MS2B noise level reaches 10 9 m3/kg been described associated to three tektite fields (Australasian, Ivory for a 0.1, 5 and 10 g sample, respectively. MFK1 and MS2B are Coast and North American), and are scattered almost globally as restricted to sample masses below about 40 g (depending on sam- illustrated by the Australasian strewnfield traced from China to ple shape), while SM30 can be used for larger samples. We mea- Australia, Madagascar and Antarctica (Folco et al., 2008). sured ivoirites, georgiaites and bediasites directly in the London Magnetic properties of tektites have been investigated by and Paris Natural History museums, as well as in the private col- Werner and Borradaile (1998) concerning the australasite and lection of A. Carion in Paris. When using SM30 or MS2B, we did moldavite strewnfields, while the Belize strewnfield has been in- not consider samples that yield a precision worse than 10%. In vestigated by Sentfle et al. (2000) and Hoffmann et al. (2013). the case of MS2B this represents a lower mass limit of 1g con- − One purpose of the present contribution is to extend the database sidering the typical χ of 100 × 10 9 m3/kg for tektites. Data to the Ivory Coast (ivoirite) and North American (bediasite and from australasites, moldavites and Belize glass are derived from georgiaite) tektites, as well as to impact glasses with charac- the literature (Werner and Borradaile, 1998; Senftle et al., 2000; teristics close to tektites, i.e. fully melted material with splash- Hoffmann et al., 2013). We also measured a collection of mol- forms and moderate vesiculation. The glasses studied in this paper davites and Belize glass using a Kappabridge. Irghizites from D.D. are the previously described irghizites, Aouelloul, Darwin, Libyan Badyukov collection (Vernadsky Institute, Moscow, Russia) were Desert and Wabar glasses (e.g., Koeberl, 1986; Koeberl et al., 1998; measured in the Schmidt Institute of Physics of the Earth RAS Howard and Haines, 2007; Giuli et al., 2003; Hamman et al., 2013), (Moscow, Russia) using the MFK1, while our own field collec- as well as the newly discovered atacamaites (Devouard et al., tion for the newly discovered impact glass strewnfield in Atacama 2014). We also report on “tektite-like” volcanic glass found in Saf- (Devouard et al., 2014)was measured with the MFK1 at CEREGE. ford (Arizona), to evaluate the ability of magnetic properties to For other impact glasses, i.e. Aouelloul, Darwin, Libyan desert, and discriminate volcanic from impact glasses. Wabar we report essentially MFK1 measurements made on sam- One central question in impact glasses is to trace a possi- ples acquired or loaned from private collectors. The seminal work ble meteoritic contamination in the overall terrestrial composition of Senftle and Thorpe (1959) on moldavites, australasites and a (Koeberl, 1998). This quest is not often fruitful as the average con- few other glasses will not be directly integrated in the present tamination may be only at trace level. One potential application low field susceptibility database. Indeed their measurements were of magnetic measurements on a large number of tektites and re- performed on mg size samples using a high variable field Curie lated glass samples is the detection of samples anomalously rich in balance (minimum field of 0.39 T), providing by extrapolation the magnetic elements (Fe, Ni, ...) pointing toward a larger meteoritic high field susceptibility and saturation magnetization. contamination, as done by Tagle et al. (2014) using Ni content The measured low field susceptibility is the sum of the para-, screening of australasites. dia- and ferro-magnetic susceptibilities. Paramagnetic susceptibility (χp) arising from magnetic elements diluted in glass can be 2. Sample and methods predicted based on chemical composition and oxidation state. While FeO content in tektites is usually above 2 wt%, Mn is around Glass samples were weighted and their low field magnetic sus- 600 ppm, and Cr, Ni, Co are mostly below 100 ppm (Koeberl, ceptibility (χ ) measured either in the laboratory at CEREGE using 1986, 1990). Therefore we can obtain good estimates of χp us- a high sensitivity MFK1 kappabridge, or directly in the collec- ing FeO analyses alone. At room temperature and assuming neg- P. Rochette et al. / Earth and Planetary Science Letters 432 (2015) 381–390 383 ligible magnetic interaction, a reasonable assumption for Fe con- 2+ tents below 10 wt%, χp can be predicted to equal (25.2 tFe + − + 33.4 tFe3 ) 10 9 m3/kg, where tFe is the elemental amount in wt% (Rochette, 1987). The paramagnetic contribution of 600 − ppm Mn is about 2 × 10 9 m3/kg (Rochette, 1987). The diamag- − netic susceptibility can be estimated at −5.5 × 10 9 m3/kg (by analogy with quartz, e.g., Rochette, 1987). The ferromagnetic susceptibility can be estimated as the difference of the measured total susceptibility and the predicted para- and diamagnetic susceptibilities. All together, these numbers can provide estimates of + + + Fe3 /(Fe2 + Fe3 ) oxidation ratio as well as ferromagnetic contribution linked to the presence of a Fe rich phase like metal, magnetite or hematite. The ferromagnetic contribution can also be estimated by mea- suring hysteresis loops, using a vibrating sample magnetometer (VSM Micromag, at CEREGE). This instrument allows measurement of a sample up to 2g (and at least 0.2 g), with a magnetization − sensitivity of the order of 10 8 Am2, in fields up to 1T. The high field susceptibility χHF is estimated, after subtraction of sample holder signal, by linear fit on the loop branch above 0.5 to 0.7 T. The remaining part is representative of the ferromagnetic contribution, and allows to define coercivity (BC ), saturation magnetization (M S ) and saturation remanence (MRS). Back-field remanence curves allow to define remanent coercivity (BCR). χHF is the sum of the para- and diamagnetic susceptibilities discussed above, plus an eventual antiferromagnetic term. However, this term appears absent in the studied samples. MRS of weakly magnetic samples has also been measured using a SQUID magnetometer after magnetization in a pulse field of 2T. Viscous decay of MRS has been evaluated by remeasuring MRS 15 days after initial magnetization, or by continuous measurements with the VSM over a few minutes. Another test for the presence of magnetic grains near the superparamagnetic threshold is the frequency dependence (fd%) that has been evaluated using MFK1 χ measurement at 1 kHz (the frequency used for regular measurements) and 15 kHz, noted χ1 and χ15. fd% is 100 ∗ (1 − χ15/χ1). Polished sections of selected atacamaite glasses were investigated using an analytical SEM JSM-5910LV, operated under 15 kV accelerating voltage in backscatterred electrons imaging mode, at Laboratoire Magmas et Volcans (Clermont–Ferrand, France).

3. Results Fig. 1. a) Average magnetic susceptibility for tektites (solid squares), Libyan desert, 3.1. Tektites Darwin and Belize glasses (gray diamond, ordered by increasing FeO) as a function of FeO content. Error bar is the calculated s.d. for χ and an arbitrary 10% for FeO; 2+ 3+ Table 1 and Fig. 1 summarize our results and literature average theoretical lines for Fe and Fe (including diamagnetic and 600 ppm Mn contri- butions) in blue and red, respectively. Note that total iron is reported as FeO. b) The χ data. For tektites the relative s.d. is about 10% for australasites same plot in log units for Aouelloul, Irghizite, Wabar and Atacamaite glasses (or- georgiaites and ivoirites, 25% for bediasites and 19–34% for mol- dered by increasing FeO), showing median susceptibility values (red ticks) as well davites. These two tektite groups are known to show a significantly as 5, 25, 75, 95 percentiles (black ticks). (For interpretation of the references to larger heterogeneity in iron amount (Koeberl, 1986). Low signal as color in this figure legend, the reader is referred to the web version of this article.) an additional source of dispersion may be evoked for moldavites, the less magnetic tektites. Our measurements of a larger set of Hysteresis loops (Fig. 2 and Table 2) confirm the above in- moldavite than the one reported in Werner and Borradaile (1998) ference that tektites are essentially paramagnetic. After careful yield indistinct average, with a lower s.d. in our case. We will use subtraction of the sample holder signal (re-measured after each our average data in the following. sample measurement), the ferromagnetic part of the loop appears No outliers at the three s.d. level are found, confirming that negligible within experimental errors. This can be seen by the magnetic susceptibility of each tektite group is very well defined. agreement between the high and low field susceptibility (χ HF and Only three georgiaites samples were measured, giving lower χ χ determined by the VSM and MFK1, respectively). The ratio of than the one of bediasites, although within one s.d. Correlation be- the two is on average 0.9, signing an instrumental bias linked to tween average χ and FeO amount is as predicted for paramagnetic large sample volume rather than a real ferromagnetic signal. In- + Fe2 within error bars (Fig. 1a), except for the poorly defined geor- deed calibration is performed on a point sample. The apparent giaite value. This provides an indirect proof that the susceptibility saturation magnetization MS determined with the VSM (by lin- + of tektites is essentially paramagnetic and that Fe3 is minor to ab- ear interpolation over 0.5 T), ranging from 0.17 to 0.32 mA m2/kg, sent in tektites. Indeed, Mössbauer, ESR and XANES spectroscopies is less than 1% of the total magnetization at 1 T. We interpret indicate that the oxidation ratio in tektite is near zero (Fudali et this slight deviation from linearity as a likely instrumental bias al., 1987; Dunlap and Sibley, 2004; Giuli et al., 2002, 2010). rather than a real ferromagnetic content of the tektite. Taken at 384 P. Rochette et al. / Earth and Planetary Science Letters 432 (2015) 381–390

Table 2 Room temperature hysteresis parameters as well as frequency dependence. Note that MRS is measured with a cryogenic magnetometer when too weak for the VSM.

−9 −3 −3 a −9 Name χHF × 10 MS × 10 MRS × 10 χHF/χ N χHF ×10 Fd (m3/kg) (A m2/kg) (A m2/kg) (%) (m3/kg) (%) Moldavite 19.1–29.3 0.17–0.27 ≤0.001 78–107 2 33 ± 6– Bediasite 47.5–68.1 0.21–0.32 ≤0.001 84–95 2 66 ± 9– Australasite 76.7–88.7 0.17-0.25 ≤0.004 84–92 2 85 ± 10 – Ivoirite 108–113 0.4–0.6 0.002 90 2 – –

Darwin 49–107 0.15–0.6 0.002 77–109 2 – – Aouelloul 34–63 0.3–19.6 0.04–2.9 14–58 2 48 14 Irghizite 84–148 1–73 0.006–16 18–95 5 – 3–20 Wabar 82 12 1.9 13 1 – 20 Belize glass 111 0.33 ≤0.005 87 1 117 ± 8– Atacamaite 64–877 0.4–2350 0.02–280 2–109 32 – 7–16 Safford 14–22 18–68 2–12 7–10 2– – a High ﬁeld susceptibility from Senftle and Thorpe (1959) and Senftle et al. (2000) is also indicated.

(their noise level). In Table 2 we also report the χHF values of Senftle and Thorpe (1959), which ﬁt reasonably our data. Mag- netic grains are likely of very small size, on the order of 20 nm, the superparamagnetic threshold for magnetite, as the 15 days decay of MRS is on average 48% for the measured tektites. In the above interpretations, the ferromagnetic grains present in minute amounts are assumed to be magnetite, but metallic iron is not ex- + cluded, in accordance with the lack of Fe3 in tektites (Giuli et al., 2002) and the report of metallic iron spherules by Chao et al. (1964; see also discussion in Ganapathy and Larimer, 1983; Koeberl, 1990). However, we note that Chao et al. (1964) report shows that metal bearing samples are rarities (less than 1% of the collection studied).

3.2. Other impact glasses

Fig. 2. Hysteresis loops for tektites, according to increasing slope: moldavite, bedia- 3.2.1. Strongly magnetic impact glasses site, australasite, ivoirite. Compared to the four accepted tektite groups, the atacamaite, irghizite, Aouelloul and Wabar glass χ data show strong dispersion (s.d.% from 59 to 89, Table 1 and Fig. 1b) that implies a significant face value the MS values would be equivalent to 2–3 ppm of pure magnetite (less than 1 ppm of metallic iron), which would trans- contribution of ferromagnetic grains to susceptibility at least for −9 3 samples with high χ values. Distribution appears nearly log nor- late into a ferromagnetic susceptibility around 10 m /kg. The + mal (Fig. 3a). A large proportion of Fe3 cannot explain the higher saturation remanence (MRS) measured with the SQUID magne- values (Fig. 1b). The median χ is significantly lower than the av- tometer, from 0.5 to 4.5 μA m2/kg, is lower by one to two orders erage: 136 versus 164 for irghizites, 55 versus 88 for Aouelloul of magnitude than the apparent MRS determined with the VSM, and 169 versus 302 for atacamaites. The plot of χ versus sample again indicating instrumental noise rather than a real sample sig- masses (Fig. 3b and c) shows a higher probability of high χ values nal. M values measured with the SQUID magnetometer indicate RS for smaller samples for atacamaite and irghizite glasses, signing that the true ferromagnetic content is in the 0.03–0.3 ppm range heterogeneities at the mm scale, typical of nugget effect. However, assuming a M /M ratio of 0.2. This is at odds with Werner and RS S the opposite trend is observed in Aouelloul glasses, possibly re- Borradaile (1998) who advocated for significant ferromagnetism, lated to the fact that the studied Aouelloul samples are fragments, with M in the range 0.5–21 mA m2/kg and M in the range S RS while atacamaite and irghizite samples are nearly full flight shaped 0.01–1 mAm2/kg. We suspect that the ferromagnetic signal re- specimens. The structure of the data is compatible with the major- ported by these authors may be due to contamination of the ity of samples being dominated by paramagnetism, and a minority sample or sample holder and imperfect subtraction of the sam- showing significant and variable amount of ferromagnetic inclu- ple holder signal. Indeed, they used the AGFM Micromag system sions. The ferromagnetic contamination remains low for irghizites, − that measures samples of a few mg whose expected signal is not Aouelloul and Wabar glasses (χ < 1025 × 10 9 m3/kg) while χ much larger than the holder signal. In our case the large size varies over two orders of magnitude for atacamaites. This situation of the samples (typically one gram) decreases strongly the pos- may be due to the one or two orders of magnitude difference in sible effect of contamination. De Gasparis et al. (1975) report MRS number of samples measured. 2 values for australasites in the 7 to 70 μA m /kg range, in agree- Hysteresis loops (Fig. 4) confirm this interpretation, samples ment with our suspicion of overestimation of the ferromagnetic with susceptibility close to the median showing essentially para- signal by Werner and Borradaile (1998). With such low rema- magnetism (Fig. 4b), and samples chosen within the high suscep- nence, the paleomagnetic study of tektites is very delicate and tibility range yielding significant ferromagnetism. The shape of the can only be successful in tektite with anomalously high magnetite hysteresis loops is strongly suggestive of magnetite as the ferro- content. Moreover we measured a strong viscosity of remanence magnetic mineral, with saturation reached below 0.3 T, except in (with an average MRS decay of the order of 20% in 24 h), in- an irghizite sample where a minor high coercivity component is dicative of poor paleomagnetic stability. We note also that all MS visible (Fig. 4c). This high coercivity component is interpreted as reported by Senftle and Thorpe (1959) are below 0.1 mA m2/kg the signature of hematite as this mineral, together with magnetite, P. Rochette et al. / Earth and Planetary Science Letters 432 (2015) 381–390 385

Fig. 4. a) Hysteresis loops for impact glasses according to increasing magnetization at 1 T: Darwin, Aouelloul, Wabar, Belize; b) hysteresis loops of atacamaites and one irghizite (in gray); c) normalized back-field remanence curves (after saturation in 1T). In each case the most magnetic sample was measured. Crossing of the x-axis defines BCR value. Fig. 3. a) Histograms of susceptibility for Atacamaites, Irghizites and Aouelloul glass; b) distribution of susceptibility versus mass for Atacamaites (including density con- some samples show clear departure toward mixture with super- tours); the line exemplifies nugget effect by diluting a 50 mg mass with χ of 15 000 − paramagnetic grain (SP; Dunlop, 2002). B values vary between into a material with χ of 169 × 10 9 m3/kg; c) the same for Irghizites and Aouel- CR loul glass (gray and empty circles, respectively). 10 and 50 mT (except one atacamaite sample at 78 mT), again typical of fine-grained magnetite. Frequency dependence up to 20% has been identified microscopically in irghizites by Badjukov et al. has been measured in irghizites (16% in atacamaites), with MRS de- (1996).No evidence of metal is observed, although metal inclu- cay of 29% over 15 days. Fig. 6 shows the viscous decay over 5min sions have been described in Aouelloul and Wabar glasses (Chao in the four types of glass, again indicative of a large proportion of et al., 1966; Hamman et al., 2013). However metal appears very ferromagnetic grains near the SP–SD threshold. The inferred PSD rare in Aouelloul: a single sample among a hundred revealed metal grain size is confirmed by SEM investigations (Fig. 7 and Devouard (Chao et al., 1966). Grain size, based on Day–Fuller diagram (Fig. 5), et al., in preparation). χHF, a good proxy for total iron amount, is in the pseudo-single domain range (PSD, 0.1 to 10 μm) although varies by a factor over 10 for atacamaites (2 for irghizites). The 386 P. Rochette et al. / Earth and Planetary Science Letters 432 (2015) 381–390

Fig. 7. Backscattered electrons SEM image of a strongly magnetic atacamaite, show- Fig. 5. Day plot of hysteresis parameters for Atacamaites (open circles), Irghizites, ing crystallites of magnetite and/or iron-rich silicates (appearing white on the im- Aouelloul and Wabar glasses (gray circles). Indicated are the limits for PSD, as well age) in the most Fe-rich regions of the glass; width of field 4.4 mm. Inset: higher as theoretical trend for SD–MD mixture according to Dunlop (2002). magnification of magnetite crystallites in another atacamaite; width of field 100 μm.

This is in agreement with the large variability of FeO content reported in Darwin glass: from 0.8 to 6 wt% (Howard, 2008). Satura- tion remanence is as low as in tektites. The low field susceptibility of Belize glass (including data − from Hoffman et al., 2013), at 127 ± 5 × 10 9 m3/kg appears + typical of tektites: small dispersion, agreement with the Fe2 line, and purely paramagnetic behaviour (Figs. 1 and 4a). Tikal glass yields an average value of high field susceptibility of 117 ± − 8 × 10 9 m3/kg (Senftle et al., 2000). This value fits with the assumption that they represent the same material, taking into account the unknown cross-calibration between the two methods and ferromagnetic component. Our MRS measurements are also typical of tektites (<0.001 mA m2/kg except one sample at 0.005). Accordingly, Senftle et al. (2000) observed a MS below 2 0.1 mA m /kg in 9 out of 12 measured fragments. Their higher MS reached 1.8 mA m2/kg but was obtained on a possibly contami- nated crushed powder. The MRS reported by Hoffman et al. (2013) on one Belize tektite is anomalously strong: 20–26 mA m2/kg. As − their susceptibility value (137 × 10 9 m3/kg) fits with our measurements, we suspect an error in their M report. Fig. 6. Viscous decay of saturation remanence MRS measured with the VSM as a RS function of time after 1T application (from four samples away for the theoreti- Finally the very iron-poor Libyan Desert glass (0.1 wt% FeO cal trend in Fig. 5). By order of increasing viscosity: atacamaite, Wabar, Aouelloul, according to Giuli et al., 2003) shows a significant difference be- irghizite. tween the standard clear facies and the green facies with more or less abundant dark streaks. The clear type is diamagnetic (as amount of magnetite estimated with MS varies over at least two already observed by Senftle and Thorpe, 1959), while the dark fa- orders of magnitude (four orders if estimated with MRS), and is cies has positive susceptibility, likely of ferromagnetic origin, as weakly correlated to total iron amount (R2 of 0.39), indicating that 2 demonstrated by the significant MRS (0.01 mA m /kg), and by a crystallization of magnetite is favoured by the presence of iron in non-isotropic susceptibility (6% of anisotropy). Unfortunately the the liquid but is also controlled in large part by other parameters magnetization is too weak to produce significant hysteresis mea- such as oxygen fugacity that may vary from sample to sample. The surements. A significant frequency dependence (11%), as well as a maximum equivalent amount of pure magnetite is 2.5 wt%, in ata- 24% decrease of MRS over 15 days, indicate a grain size near the camaites. superparamagnetic threshold (i.e. ≈20 nm).

3.2.2. Weakly magnetic impact glasses 4. Discussion Darwin and Belize glasses show magnetic characteristics more in line with tektites. The susceptibilities of the 45 studied Darwin 4.1. Origin of variable redox state glass samples measured are no more scattered than for moldavites or bediasites, with a clear correspondence between color and χ Magnetic properties of tektites and impact glasses are mostly value (low χ for clear color). Still, Fig. 1a suggests that the Dar- a function of Fe and Ni amounts, but constitute also a very sen- + win glass bears significant Fe3 or ferromagnetic impurity, as al- sitive probe of the redox state. Contrary to spectroscopic meth- ready suggested by spectroscopic investigations (Giuli et al., 2002; ods (XANES or Mössbauer spectroscopy) who essentially probe the + + Dunlap and McGraw, 2007). Hysteresis loops of Darwin glass show Fe2 and Fe3 atoms dispersed in the silicate glass, they are sensi- essentially paramagnetism, even for samples with the higher χ , tive to minor amount (down to the ppm level) of ferromagnetic Fe indicating that variability is caused by variable total iron amount. rich crystalline phases that encompass the whole oxidation range P. Rochette et al. / Earth and Planetary Science Letters 432 (2015) 381–390 387 from metal to magnetite and hematite. Unequilibrated phases may coexist at the microscopic scale, as exemplified by the case of Wabar glass in which metal and magnetite are observed, within + a mostly Fe2 rich glass (Hamman et al., 2013). The present results, together with published spectroscopic re- + sults, show that while tektites are essentially free of Fe3 and of ferromagnetic inclusions (despite the very rare occurrence of metallic droplets, see Ganapathi and Larimer, 1983), the non- + + tektite splash-form impact glasses usually contain a Fe2 /Fe3 mixture, allowing the crystallization of magnetite in variable amount, from <1 ppm to a few wt%. However, the vast majority of samples are nearly magnetite free. The single case with presence of hematite, in irghizites, has been interpreted based on microscopic observation as residual crystals from the target, while the equilibrium phase crystallizing in the melt was magnetite (Badjukov et al., 1996). Our results confirm, on a much larger dataset, the spectroscopic results of Giuli et al. (2002), Dunlap and Sibley (2004) and Dunlap and McGraw (2007). They indicate a system- atic dichotomy in oxidation level between tektites (plus Belize and Darwin glasses) and other splash-form impact glasses. What does it tell us about the difference between the formation conditions of these two types of impact glasses that have both been 2+ projected in the atmosphere as liquid droplets? The parameters Fig. 8. Temperature evolution of the proportion of Fe in a silicic melt (78 wt% SiO2 −5 involved are target lithology, temperature, pressure, oxygen fugac- and 1.4 wt% FeOtot) for different f O2 (0.2 or 10 bar) and total pressure (0.1 MPa and 10 GPa). See text for explanations. ity, cooling rate, equilibration and heterogeneity of the melt with the possible presence of reduced inclusions, i.e. metal enriched, + other impact glasses. An additional contributing factor to the pro- and oxidized inclusions, i.e. Fe3 enriched phases. The source ma- duction of reduced glasses/melts is the incorporation (dissolution terial of tektites, being at the Earth surface (Ma et al., 2004), + into the silicate melt) of reduced volatiles during pressure melting, should initially contain a significant Fe3 amount. A simple expla- and their subsequent degassing during pressure relaxation into the nation is that tektites derive from melts that reached significantly + + open atmosphere. In particular, the degassing of sulphur from the higher temperatures, as the Fe2 –Fe3 equilibrium (correspond- tektite liquid may serve as an oxygen pump through the reaction ing to the Fe3O4 → 3FeO + ½O2 equation) is displaced toward S + O2 → SO2 (Gaillard and Scaillet, 2014). S degassing is sug- higher fO2 values for higher temperature (e.g., Rhamdani et al., gested by the very low S content in tektites (a few ppm: Koeberl, 2008 for the solid state). This is in line with the higher kinetic 1986) despite the finding that tektites are derived from surface energy involved in tektite formation, characterized by a higher ho- soils and sediments (Ma et al., 2004)thattypically contain 102 mogeneity and much larger ejection distance. To illustrate this, + + 4 we have calculated the Fe2 /Fe3 ratio of a highly silicic glass to 10 ppm of sulphur (Sumner, 2000). Similarly, incorporation of reduced organic matter, either added by the impactor or already (SiO2 = 78 wt%, FeOtot = 1.4 wt%) using the method of Kress and present in the target, will upon degassing drive the melt toward Carmichael (1991), that relates f O2, temperature and melt com- + + position to its Fe2 /Fe3 (we note that the output is not sensitive more reduced conditions (see Iacono-Marziano et al., 2012). It has to be noted that strong melt reduction, down to equilibrium to melt compositional variation). Since the P–T– f O2 conditions during impact melting are not well constrained, but also likely with metallic Fe, requires the incorporation of comparatively small to vary between different impacts, three trends are shown: the amounts of organic matter, less than 1 wt% of C. The occasional occurrence of FeNi spherules in some tektites (Chao et al., 1964; first trend is obtained with an f O2 fixed at 0.2 bar, simulating equilibration with surrounding air (i.e. at 1 bar total pressure). It Ganapathy and Larimer, 1983)may witness such a process. Cal- + shows that to produce a glass with more than 90% Fe2 under ox- culations of prevailing f O2 in those tektites by combining the ◦ idized conditions, temperatures over 2000 C are required (Fig. 8). composition of FeNi droplets (Ni/Fe about 0.02, Ganapathy and ◦ Temperatures in excess of 2000 C have been indeed inferred for Larimer, 1983), with the method of Weitz et al. (1997), and the 2+ 3+ some impact melts (e.g., Dressler and Reimold, 2001). The sec- Fe /Fe liquid equilibrium (Kress and Carmichael, 1991), yield −9.6 ◦ −14.7 ◦ ond trend explores the role of pressure which is also known to f O2 ranging from 10 at 2000 C to 10 at 1300 C. + + 3+ favour Fe2 in lieu of Fe3 due to the larger molar volume of the Still, the near total lack of Fe in tektite seems to indicate latter relative to the former melt species (Kress and Carmichael, that the liquid cannot be in equilibrium with an atmosphere bear- + 1991). On Fig. 8 the evolution of Fe2 proportion with tempera- ing 20% of oxygen, except if quenched at very high temperature and pressure (Fig. 8). One has thus to consider a mechanism to ture at an f O2 = 0.2bar but at 10 GPa shows that the production + ◦ of a glass with 90% Fe2 requires a temperature of about 1400 C, prevent re-oxidation of iron during the hot flight in air. This sec- ◦ which is 600 C less than at 1bar. Pressure melting during impact ondary oxidation process may be the reason for the correlation processes may largely exceed 10 GPa, possibly up to 50 GPa for seen in atacamaites and irghizites (both aerodynamically shaped dense quartz-rich lithologies (e.g., Keil et al., 1997), hence com- splash-forms) between size and magnetite content: smaller projec- + plete Fe3 reduction at high pressure is conceivable, though melt tiles have a higher surface/volume ratio. However, we feel that a relaxation to lower pressure will occur to some extent. Athird nugget effect is more likely the source of this correlation. trend shows the calculation in which f O2 is arbitrarily fixed at −5 10 bar, i.e. midway between air and the f O2 of common ter- 4.2. Extraterrestrial contamination − − restrial magmas (10 8 to 10 12 bar). It shows that a moderately + reduced target produces glasses with nearly 100% Fe2 at tem- Extraterrestrial contamination can be a source of heterogene- ◦ peratures around 1600 C and 10 GPa. Hence, higher equilibrium ity in both iron amount and oxidation level. As an example, the pressure and temperature are likely to differentiate tektites from dark facies of Libyan Desert glass has been shown to yield both 388 P. Rochette et al. / Earth and Planetary Science Letters 432 (2015) 381–390

ples than in Fig. 9a). Anyhow, it clearly appears in atacamaites that selecting the more magnetic samples increases the probability to get samples with higher extraterrestrial contamination. Moreover, SEM imaging of the most magnetic samples shows magnetite-rich bands where Ni content locally peaks up to 3 wt%, indicating a large contamination. We foresee that the same strategy could be used in other impact glasses. The percentage of highly magnetic outliers demonstrates the interest of scanning a large number of samples by χ measurements. Setting the outlying threshold at two (four) times the median, this percentage is 17(7), 14(6) and 9(1)% for atacamaites, Aouelloul glass and irghizites, respectively. This two-fold percentage is below 1–2.5% for tektites (only datasets with N ≥ 40 were considered).

4.3. Magnetic discrimination

One side application of χ measurement of tektites is to discriminate non-destructively the provenance of samples if any ambiguity exists. Indeed, the different fields have discrete susceptibility ranges. As an example of application, a private collection of ivoirites recently exhumed by A. Carion yields an aver- − age χ of 102 ± 10 × 10 9 m3/kg (N = 94) while historic samples from London and Paris collection yields an average χ of − 106 ± 17 × 10 9 m3/kg (N = 15). These identical values, clearly distinct from australasite range, confirm the ivoirite provenance of the tektites in this private collection. We measured a tektite re- puted to be collected in Tibet (A. Carion collection) and found its − χ value (86 × 10 9 m3/kg) indistinct from the average of australasite (Table 1), thus supporting that this material comes from the Australasian strewn field. We also measured a faked mol- − davite. Its susceptibility (20 × 10 9 m3/kg) is clearly out of the range of moldavites. Finally, as a test for the possibility to use magnetic properties to discriminate impact glasses from obsidian like volcanic glasses, we measured a set of small glasses pebbles Fig. 9. Correlation between a) χ and bulk Ni content and b) MS and χHF in ataca- collected at Safford, Arizona. The 12 samples measured yield an maites (not from the same sample collection). Note that the minimum MS was set at 1μAm2/kg, i.e. about the noise level. extreme spread (Table 1), typical of susceptibility being carried by ferromagnetic impurities. Compared to known magnetic properties lower oxidation level (Giuli et al., 2003) and significant enrich- of obsidian (Zanella et al., 2012; Frahm and Feinberg, 2013)the ment in Fe, Ni, and iridium. On the other hand, the presence spread of susceptibility is similar (e.g. s.d.% of 181, compared to of rare metallic FeNi droplets in australasites, first interpreted as 176 in Acigol obsidian from Frahm and Feinberg, 2013). Hysteresis the sign of extraterrestrial contamination (Chao et al., 1964) has points toward variable near SD magnetite amount and variable low been later reinterpreted as terrestrial in origin (Ganapathy and paramagnetic content (equivalent to about 0.6% FeO). So properties Larimer, 1983). Extraterrestrial contamination is often scrutinized of Safford glass are compatible with those of obsidian. However, by correlations among Cr, Ni, Co, and possibly Fe. The most robust obsidian and magnetite-bearing impact glasses (atacamaite, etc.) correlations are between Ni and Co and between Ni and Fe, as ob- mainly share the same magnetic characteristics. The main differ- served in irghizites (Mizera et al., 2012), Darwin glass (Howard, ences are a lower FeO amount in obsidian (typically 0.1–2% range) 2008), Wabar glass (Hamman et al., 2013), as well as in ataca- and the lack or rarity of purely paramagnetic samples that consti- maites (Devouard et al., 2014 and in preparation), although the tute the median case in the impact glasses (except Wabar). This ambiguity with an ultramafic lithology in the target has to be dis- dichotomy may be related to a difference in cooling rate, which cussed (Howard, 2008). One way to look for samples with a higher may often be too high to allow the diffusion of Fe to crystallize extraterrestrial contamination would be to look for the most mag- magnetite in impact glasses. netic samples. In fact magnetic separation from crushed glass has been used to locate inclusions of iron oxides (Badjukov et al., 5. Conclusions 1996). In the case of atacamaites, faced with the large susceptibility range, we chose for bulk major and trace elements to analyze Our study of the magnetic susceptibility of the five type of samples spanning the whole range of susceptibility. This allowed tektites and seven tektite related impact glasses allows to obtain to discover a significant correlation between χ and e.g. Ni (Fig. 9a, robust conclusions on their magnetic characteristics based on the R2 = 0.47), still much less significant than the Fe/Ni or Co/Ni cor- large number of samples studied in each field (>40 except for relations (Devouard et al., in preparation, R2 = 0.97). This is due georgiaites, Belize, Wabar and Libyan desert glasses). Georgiaites, to the fact that susceptibility is the sum of χ HF, the paramagnetic with 3 samples, clearly require further study. + part strictly proportional to total Fe in the glass plus the ferro- Tektites appear to be characterized by pure Fe2 paramag- magnetic susceptibility χf proportional to the amount of magnetite netism, with ferromagnetic traces below 1 ppm. The different inclusions. The amount of magnetite is only loosely correlated to fields yield narrow susceptibility ranges, allowing eventually to dis- total iron, as demonstrated by the weak correlation between M S criminate samples of uncertain origin. Belize and Darwin glasses 2 and χHF (R = 0.18; Fig. 9b, note that it is a different set of sam- share similar characteristics. P. Rochette et al. / Earth and Planetary Science Letters 432 (2015) 381–390 389

On the other hand the other studied impact glasses have wider Ganapathy, R., Larimer, J.W., 1983. Nickel–iron spherules in tektites: non-meteoritic susceptibility ranges, with majority of samples close to paramag- origin. Earth Planet. Sci. Lett. 65, 225–228. + + netism (Fe2 and Fe3 ) but with a significant high-susceptibility Gattacceca, J., Eisenlohr, P., Rochette, P., 2004. Calibration of in situ magnetic susceptibility measurements. Geophys. J. Int. 158, 42–49. population bearing variable amounts of magnetite. The grain size Giuli, G., Pratesi, G., Cipriani, C., Paris, E., 2002. Iron local structure in tektites of magnetite is encompassing the SP–SD threshold, i.e. circa 20 nm. and impact glasses by extended X-ray absorption fine structure and high- This signs a fundamental difference between tektites (plus Be- resolution X-ray absorption near-edge structure spectroscopy. Geochim. Cos- lize and Darwin glasses) and other studied glasses in terms of mochim. Acta 66, 4347–4353. oxygen fugacity, pressure, temperature and heterogeneity during Giuli, G., Paris, E., Pratesi, G., Koeberl, C., Cipriani, C., 2003. Iron oxidation state in the Fe-rich layer and silica matrix of Libyan Desert Glass: a high-resolution formation, thus bringing new light to the formation processes of XANES study. Meteorit. Planet. Sci. 38, 1181–1186. these materials. To acknowledge the dichotomy between these two Giuli, G., Eeckhout, S.G., Cicconi, M.R., Koeberl, C., Pratesi, G., Paris, E., 2010. Iron ox- types of splash-form impact glasses, we propose a generic name idation state and local structure in North American tektites. In: Reimold, W.U., for the non-tektite type: tektoid, to make it clear it is tektite-like, Gibson, R. (Eds.), Large Meteorite Impacts and Planetary Evolution IV. In: Geo- but fails to pass all criteria for tektite, including the reduced state logical Society of America Special Paper, vol. 465, pp. 645–652. advocated here. Giuli, G., Cicconi, M.R., Stabile, P., Trapananti, A., Pratesi, G., Cestelli-Guidi, M., Koeberl, C., 2014. New data on the Fe oxidation state and water content It also appears that selecting the most magnetic glass samples of Belize Tektites. In: 45th Lunar and Planetary Science Conference. Held allows to find rare impactor-rich material, opening new perspec- 17–21 March, 2014 at The Woodlands, Texas. In: LPI Contribution, vol. 1777, tives to identify the type of impactor responsible for the glass p. 2322. generation. Glass, B.P., 1990. Tektites and microtektites: key facts and inferences. Tectono- physics 171, 393–404. Hoffmann, V.H., Funaki, M., Cornec, J.H., Kaliwoda, M., Hochleitner, R., 2013. Mag- Acknowledgements netic properties and micro Raman spectroscopy of a central American tektites from Belize. In: 44th Lunar and Planetary Science Conference. #2528. We deeply acknowledge the curators of London and Paris Nat- Hamman, C., Hecht, L., Ebert, M., Wirth, R., 2013. Chemical projectile–target inter- ural History Museums, C. Smith and B. Zanda, for allowing us action and liquid immiscibility in impact glass from the Wabar craters, Saudi to measure on site their tektite and glass collections, as well as Arabia. Geochim. Cosmochim. Acta 121, 291–310. Howard, K.T., 2008. Geochemistry of Darwin glass and target rocks from Darwin D.D. Badyukov (Vernadsky Institute, Moscow, Russia) and the pri- crater, Tasmania, Australia. Meteorit. Planet. Sci. 43, 479–496. vate collectors, A. Carion, J. Cornec, L. Labenne, N. Lehrman and Howard, K.T., Haines, P.W., 2007. Geology of Darwin crater, western Tasmania, Aus- M. Warner, who helped in completing our database through gifts tralia. Earth Planet. Sci. Lett. 260, 328–339. and loans. N. Lehrman deserves a special acknowledgment for Iacono-Marziano, G., Gaillard, F., Scaillet, B., Polozov, A.G., Marecal, V., Pirre, M., coining the term “tektoid”. Editor and reviewers, B. Marty, G. Giuli Arndt, N.T., 2012. Extremely reducing conditions reached during basaltic in- and T. Kohout, are thanked for timely and constructive assess- trusion in organic matter-bearing sediments. Earth Planet. Sci. Lett. 357–358, 319–326. ment of the submitted manuscript. This work was partially per- Keil, K., Stoeffler, D., Love, S.G., Scott, E.R.D., 1997. Constraints on the role of impact formed according to the Russian Government Program of Competi- heating and melting in asteroids. Meteorit. Planet. Sci. 32, 349–363. tive Growth of Kazan Federal University. We acknowledge the sup- Koeberl, C., 1986. Geochemistry of tektites and impact glasses. Annu. Rev. Earth port of INSU/CNES Planetology Program on the atacamaite study. Planet. Sci. 14, 323–350. Koeberl, C., 1990. The geochemistry of tektites: an overview. Tectonophysics 171, 405–422. References Koeberl, C., 1998. Identification of meteoritical components in impactites. In: Grady, M.M., Hutchison, R., McCall, G.H.J., Rothery, D.A. (Eds.), Meteorites: Flux Badjukov, D.D., Brandstaetter, F., Petrova, T.L., Kurat, G., 1996. Iron Oxides in with Time and Impact Effects. In: Geol. Soc. London, Spec. Publ., vol. 140, Irghizites. In: Lunar Planetary Science Conference, vol. 27. pp. 133–152. Chao, E.C.T., Dwornik, E.J., Littler, J., 1964. New data on the nickel–iron spherules Koeberl, C., Bottomley, R., Glass, B.P., Storzer, D., 1997. Geochemistry and age of Ivory from south–east Asian tektites and their implications. Geochim. Cosmochim. Coast tektites and microtektites. Geochim. Cosmochim. Acta 61, 1745–1772. Acta 28, 971–980. Koeberl, C., Reimold, W.U., Shirey, S.B., 1998. The Aouelloul crater, Mauritania: on Chao, E.C.T., Dwornik, E.J., Merrill, C.W., 1966. Nickel–iron spherules from Aouelloul the problem of confirming the impact origin of a small crater. Meteorit. Planet. glass. Science 154, 759–760. Sci. 33, 513–517. De Gasparis, A.A., Fuller, M., Cassidy, W., 1975. Natural remanent magnetism of Kress, V.C., Carmichael, I.S.E., 1991. The compressibility of silicate liquid containing tektites of the Muong–Nong type and its bearing on models of their origin. Ge- Fe2O3 and the effect of composition, temperature, oxygen fugacity and pressure ology 3, 605–607. on their redox states. Contrib. Mineral. Petrol. 108, 82–92. Devouard, B., Rochette, P., Gattacceca, J., Barrat, J.-A., Moustard, F., Valenzuela, E.M., Ma, P., Aggrey, K., Tonzola, C., Schnabel, C., de Nicola, P., Herzog, G.T., Wasson, J.T., Alard, O., Balestrieri, M.L., Bigazzi, G., Dos Santos, E., Gounelle, M., Jambon, A., Glass, B.P., Brown, L., Tera, F., Middleton, R., Klein, J., 2004. Beryllium-10 in Laridhi-Ouazza, N., Shuster, D.L., Warner, M., Warner, M., 2014. A new Tektite Australasian tektites: constraints on the location of the source crater. Geochim. Strewnfield in Atacama, Chile. In: 77th Annual Meteoritical Society Meeting. Cosmochim. Acta 68, 3883–3896. Casablanca. #5394. McCall, G.J.M., 2001. Tektites in the Geological Record: Showers of Glass from the Dressler, B.O., Reimold, W.U., 2001. Terrestrial impact melt rocks and glasses. Earth- Sky. The Geological Society, London. Sci. Rev. 56, 205–284. Mizera, J., Randa, Z., Tomandl, I., 2012. Geochemical characterization of impact Dunlap, R.A., Sibley, A.D.E., 2004. A Mössbauer effect study of Fe-site occupancy in glasses from the Zhamanshin crater by various modes of activation analysis. Re- Australasian tektites. J. Non-Cryst. Solids 337, 36–41. marks on genesis of irghizites. J. Radioanal. Nucl. Chem. 293, 359–376. Dunlap, R.A., McGraw, J.D., 2007. A Mössbauer effect study of Fe environments in Povenmire, H., Cornec, J.H., 2015. The 2014 report on the Belize tektite strewnfield. impact glasses. J. Non-Cryst. Solids 353, 2201–2205. In: 46th Lunar and Planetary Science Conference. Abstract #1132. Dunlop, D.J., 2002. Theory and application of the Day plot (Mrs/Ms versus Hcr/Hc). 1: Theoretical curves and tests using titanomagnetite data. J. Geophys. Res. 107, Rhamdhani, M.A., Hayes, P.C., Jak, E., 2008. Subsolidus phase equilibria of the Fe– 1–22. Ni–O system. Metall. Mater. Trans., B 39B, 690–701. Folco, L., Rochette, P., Perchiazzi, N., d’Orazio, M., Laurenzi, M.A., Tiepolo, M., Rochette, P., 1987. Magnetic susceptibility of the rock matrix related to magnetic 2008. Microtektites from Victoria Land Transantarctic Mountains. Geology 36, fabric studies. J. Struct. Geol. 9, 1015–1020. 291–294. Rochette, P., Gattacceca, J., Bonal, L., Bourot-Denise, M., Chevrier, V., Clerc, J.-P., Frahm, E., Feinberg, J.M., 2013. From flow to quarry: magnetic properties of ob- Consolmagno, G., Folco, L., Gounelle, M., Kohout, T., Pesonen, L., Quirico, E., sidian and changing the scales of archaeological sourcing. J. Archaeol. Sci. 40, Sagnotti, L., Skripnik, A., 2008. Magnetic classification of stony meteorites: 2. 3706–3721. Non-ordinary chondrites. Meteorit. Planet. Sci. 43, 959–980. Fudali, R.F., Dyar, M.D., Griscom, D.L., Schreiber, D., 1987. The oxidation state of iron Rochette, P., Gattacceca, J., Bourot-Denise, M., Consolmagno, G., Folco, L., Kohout, in tektite glass. Geochim. Cosmochim. Acta 51, 2749–2756. T., Pesonen, L., Sagnotti, L., 2009. Magnetic classification of stony meteorites: 3. Gaillard, F., Scaillet, B., 2014. A theoretical framework for volcanic degassing chem- Achondrites. Meteorit. Planet. Sci. 44, 405–427. istry in a comparative planetology perspective and implications for planetary Rochette, P., Gattacceca, J., Lewandowski, M., 2012. Magnetic classification of mete- atmospheres. Earth Planet. Sci. Lett. 403, 307–316. orites and application to Soltmany fall. Meteorites 2, 67–71. 390 P. Rochette et al. / Earth and Planetary Science Letters 432 (2015) 381–390

Sagnotti, L., Rochette, P., Jackson, M., Vadeboin, F., Dinares-Turrel, J., Winkler, Tagle, R., Goderis, S., Fritz, J., Bartoschewitz, R., Artemieva, N., Vanhaecke, F., Claeys, A., MAGNET Science Team, 2003. Inter-laboratory calibration of low ﬁeld P., 2014. An extraterrestrial component in Australasian tektites. In: 45th Lunar and anhysteretic susceptibility measurements. Phys. Earth Planet. Inter. 138, and Planetary Science Conference. #2222. 25–38. Weitz, C.M., Rutherford, M.J., Head, J.W., 1997. Oxidation states and ascent history of Senftle, F.E., Thorpe, A., 1959. Magnetic susceptibility of tektites and some other the Apollo 17 volcanic beads as inferred from metal–glass equilibria. Geochim. glasses. Geochim. Cosmochim. Acta 17, 234–247. Cosmochim. Acta 61, 2765–2775. Senftle, F.E., Thorpe, A.N., Grant, J.R., Hildebrand, A., Moholy-Nagy, H., Evans, B.J., Werner, T., Borradaile, G.J., 1998. Homogeneous magnetic susceptibilities of tektites: May, L., 2000. Magnetic measurements of glass from Tikal, Guatemala: possible implications for extreme homogenization of source material. Phys. Earth Planet. tektites. J. Geophys. Res. 105, 18921–18925. Inter. 108, 235–243. Suess, F.E., 1900. Die herkunft der moldavite. Jahrb. Geol. Reichanst. (Bundesanst.), Zanella, E., Ferrara, E., Bagnasco, L., Ollà, A., Lanza, R., Beatrice, C., 2012. Magnetite Wien 50, 193–382. grain-size analysis and sourcing of Mediterranean obsidians. J. Archaeol. Sci. 39, Sumner, M.E., 2000. Handbook of Soil Sciences. CRC Press, Boca Raton. 2148 pp. 1493–1498.