Thermal History and Origin of the IVB Iron Meteorites and Their Parent Body

Home , IVB meteorite, Plessite, Tetrataenite

Available online at www.sciencedirect.com

Geochimica et Cosmochimica Acta 74 (2010) 4493–4506 www.elsevier.com/locate/gca

Thermal history and origin of the IVB iron meteorites and their parent body

Jijin Yang a,*, Joseph I. Goldstein a, Joseph R. Michael b, Paul G. Kotula b, Edward R.D. Scott c

a Department of Mechanical and Industrial Engineering, University of Massachusetts, Amherst, MA 01003, USA b Materials Characterization Department, Sandia National Laboratories, P.O. Box 5800, MS 0886, Albuquerque, NM 87185, USA c Hawaii Institute of Geophysics and Planetology, University of Hawaii at Manoa, Honolulu, HI 96822, USA

Received 11 November 2009; accepted in revised form 6 April 2010; available online 21 April 2010

Abstract

We have determined metallographic cooling rates of 9 IVB irons by measuring Ni gradients 3 lm or less in length at kamacite–taenite boundaries with the analytical transmission electron microscope and by comparing these Ni gradients with those derived by modeling kamacite growth. Cooling rates at 600–400 °C vary from 475 K/Myr at the low-Ni end of group IVB to 5000 K/Myr at the high-Ni end. Sizes of high-Ni particles in the cloudy zone microstructure in taenite and the widths of the tetrataenite rims, which both increase with decreasing cooling rate, are inversely correlated with the bulk Ni concentrations of the IVB irons conﬁrming the correlation between cooling rate and bulk Ni. Since samples of a core that cooled inside a ther- mally insulating silicate mantle should have uniform cooling rates, the IVB core must have cooled through 500 °C without a silicate mantle. The correlation between cooling rate and bulk Ni suggests that the core crystallized concentrically outwards. Our thermal and fractional crystallization models suggest that in this case the radius of the core was 65 ± 15 km when it cooled without a mantle. The mantle was probably removed when the IVB body was torn apart in a glancing impact with a larger body. Clean separation of the mantle from the solid core during this impact could have been aided by a thin layer of residual metallic melt at the core-mantle boundary. Thus the IVB irons may have crystallized in a well-mantled core that was 70 ± 15 km in radius while it was inside a body of radius 140 ± 30 km. Ó 2010 Elsevier Ltd. All rights reserved.

1. INTRODUCTION Bulletin Database see http://tin.er.usgs.gov/meteor/metbull. php), the group is well studied, as the irons appear to be a The IVB iron meteorites are ataxites with microscopic fairly representative suite of samples from the core of an Widmansta¨tten patterns consisting of kamacite platelets interesting asteroid with a relatively simple history. <20 lm in width, which appear in section as platelets The IVB irons are chemically unique as they contain embedded in matrices of plessite, a fine mixture of kama- exceptionally low concentrations of moderately volatile ele- cite and taenite. The small size of the kamacite platelets ments and high Ni contents: element/Ni ratios are depleted reflects the high bulk Ni concentrations of the IVB irons, relative to CI chondrites by factors of 10–104 and decrease 15–18 wt.%. Low-Ni IVB irons have very few kamacite in order of increasing volatility: Au, P and As, Cu and Ga, platelets, and phosphides whereas high-Ni IVB irons have and Ge (Kelly and Larimer, 1977; Rasmussen et al., 1984; many more kamacite platelets and phosphides (Fig. 1). Wasson, 1985). The low volatile abundances reflect high Although only 13 IVB irons are known (Meteoritical nebular temperatures when their parent body accreted or volatile loss during one or more impacts (Goldstein et al., 2009a). The high Ni concentrations in group IVB favor for- * Corresponding author. mation in an oxidizing environment (Campbell and Huma- E-mail address: [email protected] (J. Yang). yun, 2005). Despite the unusual bulk composition of group

Fig. 1. Reﬂected light photomicrographs of low-Ni IVB iron meteorite, Cape of Good Hope (left) and high-Ni IVB iron, Weaver Mountains (right) showing kamacite platelets and phosphides in a plessite matrix. The abundance of kamacite platelets and phosphides in IVB irons increases with increasing bulk Ni concentration. Figure modiﬁed from Goldstein and Michael (2006).

IVB, variations in the concentrations of Ni, P and other ele- zone particles and tetrataenite widths in the IVB irons as ments within the group are well modeled by simple frac- these parameters provide additional constraints on cooling tional crystallization in a vigorously convecting core using rates at 300–400 °C(Goldstein et al., 2009a). The cloudy experimentally determined solid–liquid distribution coeffi- zone formed below 350 °C by spinodal decomposition cients (Rasmussen et al., 1984; Campbell and Humayun, and grew by diffusion controlled coarsening (Oswald Rip- 2005; Walker et al., 2008). ening). Thus the size of the high-Ni particles is inversely re- Although much is known about the formation of the IVB lated to cooling rate (Yang et al., 1997) and has been used irons, the direction of crystallization of the IVB core and its to help understand the origin and formation of the parent subsequent thermal and collisional history is poorly known. asteroid of the IVA irons (Yang et al., 2007; Goldstein Precise metallographic cooling rates are difficult to obtain for et al., 2009b). Similarly, the tetrataenite band between the the IVB irons with the 1 lm spatial resolution of the electron cloudy taenite and the kamacite platelet formed below probe microanalyzer (EPMA) because the Ni gradients in 400 °C and its width is inversely related to the cooling rate taenite at the boundary with the kamacite platelets are (Goldstein et al., 2009a). 3 lm or less in length. In addition, the kamacite platelets Our first goal was to find out whether the cooling rates are too far apart for impingement to occur so that M-shaped in the IVB irons were uniform or if they were correlated profiles are not generated in the intervening taenite. Given with bulk Ni, and to reconcile their thermal histories with these limitations, the conventional taenite central Ni vs half their simple fractional crystallization history. A second goal width Wood method (Wood, 1964), and the profile matching was to understand core formation and evolution of the IVB method for taenite lamellae (Goldstein and Ogilvie, 1965) parent body. cannot be applied. Cooling rates for the IVB irons have been obtained from the measurement of kamacite bandwidths and 2. SAMPLES AND ANALYSES have values as high as 10,000 K/Myr (Rasmussen, 1989). However, the measured band-widths and the corresponding We examined 9 out of the 13 known IVB irons (Table 1). cooling rates have large and unknown uncertainties because Samples were first prepared by standard metallographic the orientation of the kamacite platelets with respect to the procedures, mounting, grinding, polishing, and etching polished surface was not measured and their nucleation tem- with 2% nital, and observed using light optical microscopy. peratures were poorly constrained. Samples for electron probe microanalysis were prepared the The fast cooling rates for IVB irons inferred from their same way but without etching. kamacite bandwidths have been interpreted as evidence In order to determine the nucleation mechanism and the for core formation and cooling in a small parent body with nucleation temperature of the kamacite platelets, it is neces- a radius of 2–4 km (Rasmussen, 1989; Chabot and Haack, sary to have accurate bulk Ni and P contents for each mete- 2006). However, the inferred radius of 2–4 km is much orite. Although bulk Ni and P contents of the IVB irons have smaller than the minimum radius of 10–20 km for a body been measured (Moore et al., 1969; Buchwald, 1975; Camp- to be melted and differentiated by 26Al decay (Hevey and bell and Humayun, 2005; Walker et al., 2008), there are sig- Sanders, 2006). nificant differences among the measured values, especially The purpose of this study was to obtain accurate cooling the P content. Therefore, we re-measured the bulk Ni and P rates during kamacite formation for a diverse suite of IVB content of each meteorite using X-ray area scans obtained irons by measuring Ni gradients at the kamacite–taenite with wavelength dispersive spectrometers on a Cameca SX- boundary with the analytical transmission electron micro- 50 electron probe microanalyzer (EPMA) at the University scope in thin sections cut from oriented samples with a fo- of Massachusetts. An operating voltage of 15 kV and a beam cused ion beam instrument. These Ni profiles are then current of 40 nA were used. Two major elements (Fe and Ni) compared with calculated Ni profiles generated by model- and two minor elements (Co and P) were measured. Pure Fe, ing kamacite growth (Yang and Goldstein, 2006). In addi- Ni, Co and (Fe–Ni)3P in the Grant meteorite (15.5 wt.% P) tion, we wanted to measure the dimensions of the cloudy were used as standards. The bulk Fe, Ni, Co and P composi- Thermal history and origin of IVB iron meteorites 4495 tions were measured by averaging 8–27 side by side The size of the high-Ni particles in the cloudy zone next 100 80 lm2 area scans for each IVB iron. The total area to the cloudy zone/tetrataenite boundary (Yang et al., scanned for each meteorite was dependent on its structure 1997) was measured directly from the STEM image or the and ranged from 64,000 lm2 to 216,000 lm2 (Table 1). Since Ni X-ray scan obtained from the thin foil. The measure- few kamacite platelets and phosphides are present in the low- ment of the size of the high-Ni particles in the cloudy zone Ni IVBs, their Ni and P compositions are relatively homoge- was made next to the cloudy zone/tetrataenite boundary neous and 8 area scans were adequate for each meteorite. For since the local Ni content at that position is constant at the high-Ni irons, which have larger amounts of kamacite 40–42 wt.% Ni (Yang et al., 1997). High-Ni particles farther and phosphide, larger X-ray scan areas were needed to obtain away from the cloudy zone/tetrataenite boundary were not representative bulk compositions and up to 24 area scans measured as the high-Ni particle size and Ni content of tae- were made. nite decrease with increasing distance from the cloudy zone/ Ni profiles in taenite and adjacent kamacite across kama- tetrataenite boundary (Yang et al., 1997). The measurement cite–taenite interfaces were measured by using the analytical of the size of the high-Ni particles is dependent on the electron microscope. A dual beam FEI focused ion beam geometry of the particles as discussed by Goldstein et al. (FIB) instrument at Sandia National Labs was used to obtain (2009b) for cloudy zone in the IVA irons. The lack of a transmission electron microscope (TEM) thin sections gener- large sample of high-Ni particles limits the accuracy of ally 50–100 nm thick and approximately 10 lm long and the measurement of the mean size of the high-Ni particles. 5 lm in depth. These TEM sections were cut perpendicular The width of the outer taenite rim or tetrataenite region to kamacite–taenite interfaces of the kamacite platelets in between the cloudy zone and the kamacite–taenite interface the polished and etched meteorite sample surface (Fig. 2). was also measured at several points along the kamacite–tae- Using this methodology, the kamacite–taenite interface of nite interface in each FIB section. The widths are measured the platelet in the thin section is perpendicular to the surface. directly from STEM images or Ni X-ray scans. No correc- Selected kamacite–taenite interface regions in the thinnest tion for orientation effects was necessary since the tetratae- areas of each FIB section were analyzed using a FEI Tecnai nite width was measured in a direction normal to the plane F30ST field emission transmission – analytical electron of the kamacite–taenite interface (See Fig. 2). microscope (TEM–AEM) at Sandia National Labs or a In order to measure the cooling rates, information about VG 603 scanning transmission electron microscope (STEM) the nucleation process of the Widmansta¨tten pattern, the instrument at Lehigh University both operated at 300 keV. nucleation temperature, the effect of impingement by adjacent To obtain Ni composition gradients in kamacite and taenite kamacite plates, the Fe–Ni and Fe–Ni–P phase diagrams, and at the two phase boundary, X-ray scans were taken with an the interdiffusion coefficients that control the growth of the energy-dispersive spectrometer (EDS) in micrometer-sized Widmansta¨tten pattern are needed (Yang et al., 2008). regions of the thin foil. Quantitative Ni X-ray gradients According to Yang and Goldstein (2005), the Ni and P con- 0.5–2 lm in length in taenite were measured in a direction tents of the IVB irons indicate that the Widmanstatten ori- normal to the plane of the kamacite–taenite interface using ented platelets in the IVB irons nucleated by mechanism II the stored EDS X-ray scan data. The X-ray data in each pixel (c ? c +Ph? a + c + Ph) for high Ni–high P members of were converted to composition using the Cliff–Lorimer meth- the IVB group where phosphides form in taenite during cool- od (Cliff and Lorimer, 1975) with X-ray spatial resolution of ing before kamacite nucleated. For low Ni–low P members of 2–4 nm. A kNi–Fe factor in the Cliff–Lorimer method of 1.10 the IVB group kamacite nucleation is controlled by mecha- was measured at 300 keV using a 25 wt.% Ni–Fe standard. nism III (c ? (a + c) ? a + c + Ph), where kamacite nucleation is suppressed during cooling until the meteorite enters the 3 phase field a + c + Ph. Our bulk Ni and P content of each IVB iron meteorite along with the Fe–Ni–P phase diagram were used to determine the nucleation mechanism and the nucleation temperature for each meteorite. The Ni profile matching method (Goldstein and Ogilvie, 1965) was used to obtain a cooling rate for each individual measured taenite zoning profile. In this method, the measured Ni composition vs distance profile across taenite adjacent to a kamacite platelet is compared to Ni profiles calculated for several assumed cooling rates. A match between the observed and calculated profiles allows a unique cooling rate to be determined for each meteorite (Yang and Goldstein, 2006).

3. RESULTS

3.1. Bulk Ni and P contents Fig. 2. Scanning electron microscope image of a thin section cut by a focused ion beam (FIB) instrument across a kamacite–taenite interface in a polished section of the Warburton Range IVB iron The bulk Ni and P contents that we determined for the meteorite just before extraction. The FIB section is cut normal to nine IVB irons are listed in Table 1 and are plotted in the kamacite–taenite interface of the kamacite platelet. Fig. 3a. The bulk Ni varies from 15.5 to 17.8 wt.% and 4496 J. Yang et al. / Geochimica et Cosmochimica Acta 74 (2010) 4493–4506 the bulk P varies from 0.05 to 0.25 wt.%. Literature data Another Ni profile from Hoba is given in Fig. 4c along with are shown in Fig. 3b for comparison. Fig. 3a also shows two Ni profiles from the high-Ni IVB iron, Warburton the Ni–P composition regions where mechanisms II Range, Fig. 4d and e. About 500 measurements of the Ni con- (c ? c +Ph? a + c + Ph) and III (c ? (a + c) ? a + c centration were made for each 1–2 lm profile with a spatial + Ph) are applicable for the nucleation of kamacite platelets resolution of 4 nm/pixel depending on specimen thickness. in IVB irons (Yang and Goldstein, 2005). Irons with bulk The Ni content in tetrataenite in Hoba and Warburton Ni and P located above the dashed line nucleate kamacite Range exceeds 50 wt.% at the kamacite–taenite interface by mechanism II while irons with bulk Ni and P located be- indicating continuous slow cooling to temperatures of low the dashed line nucleate kamacite at lower tempera- 300 °C or below. One composition profile was measured tures by mechanism III. Nucleation temperatures for for 3 meteorites, two composition profiles for 5 meteorites, kamacite formation in each meteorite were obtained from and four composition profiles for 1 meteorite. The 17 mea- the Fe–Ni–P phase diagram (Romig and Goldstein, 1980) sured Ni composition profiles were then compared with pro- and are listed in Table 1. They increase with increasing files calculated for various assumed cooling rates (Yang and Ni and P content from 470 to 590 °C. We note that Ras- Goldstein, 2006). The two compositional profiles for Hoba in mussen (1989) proposed that taenite was saturated in P so Fig. 4b and c give average cooling rates of 400 and 900 K/ that schreibersite formed in all IVB irons before kamacite Myr for a mean value of 650 K/Myr. With the exception of formed as the equilibrium phase. This assumption is incor- analyses in the tetrataenite rim and a few data points near rect for the low P IVB irons as shown in Fig. 3a and leads to the edge of the cloudy taenite, the Ni concentration data inaccuracies in cooling rate measurements. for Hoba scatter from 300 to slightly above the 1000 K/ Myr curves. The two Warburton Range profiles in Fig. 4d 3.2. Measured metallographic cooling rates and e both give cooling rates of 3000 K/Myr with virtually all the data in the cloudy zone scattering between the 1000 A Ni X-ray scan taken in a 2000 500 nm region of the and 9000 K/Myr cooling rate curves. low-Ni Hoba meteorite is shown in Fig. 4a. The Ni X-ray The Ni profiles of two IVB irons, Skookum and Santa scan shows the kamacite (black – low Ni), the tetrataenite Clara, show evidence of re-heating at the kamacite/taenite (light gray – high Ni) and the cloudy zone composed of an interface (Fig. 5). For Skookum, the taenite Ni gradient intergrowth of high-Ni particles (light gray) and kamacite within 300 nm of the kamacite–taenite boundary shows evi- (dark gray). The measured Ni profile across the kamacite– dence of re-heating (Fig. 5a). The Ni profile in tetrataenite taenite boundary shown in Fig. 4a is plotted in Fig. 4b. decreases rather than increases approaching the kamacite– taenite boundary and the maximum Ni content in tetrataenite is 40 wt.%. The taenite Ni gradient beyond 300 nm, however, is not substantially influenced by re-heating so that the cooling rate can still be measured. For the Santa Clara IVB iron (Fig. 5b), the Ni gradient in taenite within 100 nm of the kamacite–taenite interface shows a diffusion profile suggesting a more modest shock re-heating than in Skookum. The maximum Ni content in tetrataenite is about 47 wt.%. The Ni gradient beyond 100 nm is not substantially influenced by re-heating so that the cooling rate can still be determined. In the case of Skookum, it is likely that heating occurred after atmospheric entry (Buchwald, 1975) but for Santa Clara we suspect earlier shock heating. Table 1 lists the measured cooling rates for each of the 17 Ni profiles and the average cooling rate for the 9 IVB irons. Since so few cooling rates are measured for each meteorite, we also list the range of the cooling rates which fit the Ni composition data for each meteorite e.g., 300–1200 K/Myr for Hoba and 1000–9000 K/Myr for Warburton Range (Fig. 4). Fig. 6 shows the measured cooling rates vs the bulk Ni content for the 9 IVB irons. The vertical bars are not error bars but show the cooling rate ranges, e.g., 300–1200 K/Myr for Hoba. The mean cooling rates differ by a factor of 10 ranging between 475 and 5000 K/Myr and increase systematically with increasing Ni content.

Fig. 3. Bulk analyses of Ni and P in IVB irons: (a) this study; (b) data from Walker et al. (2008), Campbell and Humayun (2005) Buchwald 3.3. Cloudy zone particle size and tetrataenite bandwidth (1975), and Moore et al. (1969). The dashed line in (a) shows the measurements compositional ﬁelds in which kamacite platelets form by mechanisms II (above the line) and III (below the line). The dotted lines in both (a) The cloudy zone regions and tetrataenite bands in all and (b) indicate the best straight-line ﬁt to the data in (a). but two IVB irons are well developed and preserved with Thermal history and origin of IVB iron meteorites 4497

70 70 (b) (d) 60 60

50 50 1000 K/Myr 40 40 3000 K/Myr

Ni (wt%) 30 30 300 K/Myr Ni (wt%)

20 750 K/Myr 20 9000 K/Myr 500 K/Myr 10 10 Hoba Warburton Range 0 0 -2 -1.8 -1.6 -1.4 -1.2 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 -0.2 0 0.2 0.4 0.6 0.8 1 1.2 Distance (µm) Distance (µm)

70 70 (c) (e) 60 60

50 50

40 300 K/My 40 750 K/My 1000 K/Myr 30 30 Ni (wt%) Ni (wt%) 1000 K/Myr 9000 K/Myr 20 20 3000 K/Myr 10 10 Hoba Warburton Range 0 0 -0.2 0 0.2 0.4 0.6 0.8 1 -0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 Distance (µm) Distance (µm)

Fig. 4. Ni analyses and cooling rate measurements in two IVB irons Hoba (low-Ni, low P) and Warburton Range (high Ni, high P). (a) Ni X-ray scanning image of the surface of a FIB thin section of Hoba with an X-ray resolution of 4 nm/pixel obtained using the analytical transmission electron microscope. The scanningimage shows from left to right: cloudy zone intergrowth, a 200 nm wide band of tetrataenite (white) and the edge of the kamacite platelet (black). The dashed rectangle shows the region over which the Ni profile shown in Fig. 4b was obtained. (b and c) Ni concentration profiles across the kamacite–taenite interfaces in Hoba. (d and e) Ni concentration profiles across the kamacite–taenite interfaces in Warburton Range. The vertical arrow marks the boundary between the cloudy taenite and the tetrataenite rim where some data points plot below the curves. Cooling rates were determined by comparing the observed Ni profiles in the cloudy taenite with those calculated for a range of cooling rates. no evidence of re-heating. For Skookum and Santa Clara measured and the average particle size and its 1r standard no cloudy zone regions were observed, consistent with deviation were obtained (Table 2). The relationship be- evidence of re-heating observed in the Ni taenite gradient tween the cloudy zone high-Ni particle size and the bulk near the kamacite–taenite boundary (Fig. 5). Figs. 4a and Ni of each IVB meteorite is plotted in Fig. 8a. The 7 show examples of Ni X-ray area scans from which the high-Ni particle size in the cloudy zone varies from 19 high-Ni particle size of the island phase of the cloudy zone to 33 nm and decreases with increasing Ni content. The microstructure and the tetrataenite bandwidths were mea- width of the tetrataenite bands measured in each of the sured. For each image, a limited number of high-Ni parti- images and the average tetrataenite width for each meteor- cles in the cloudy zone from several taenite bands were ite are also given in Table 2. The relationship between the 4498 J. Yang et al. / Geochimica et Cosmochimica Acta 74 (2010) 4493–4506

70 10000 (a) 60

50 1000 40

30 5000 K/Myr Ni (wt%) Cooling Rate (K/Myr) 20 100 15 16 17 18 19 10 Bulk Ni (wt%) Skookum 0 Fig. 6. Variation of the metallographic cooling rate with Ni -0.3 0 0.3 0.6 0.9 1.2 1.5 1.8 content for 9 IVB irons. The vertical bars show the range of cooling Distance (µm) rates determined from individual Ni analyses in taenite, e.g., 1000– 70 9000 K/Myr for Warburton Range (Fig. 4d and e). These ranges (b) are very conservative estimates of the 2r precision limits for the cooling rates derived for each iron (see text). The dashed line shows 60 the inferred relationship between cooling rate and bulk Ni.

50 1500 K/Myr range of 0.04–0.2 wt.% P found by Campbell and Humayun 40 (2005) and other authors. The diﬀerence in bulk Ni and P composition from these two studies can be attributed to 30 3000 K/Myr the total areas from which the data were obtained. Camp-

Ni (wt%) bell and Humayun (2005) used the laser ablation - induc- 20 tively coupled plasma – mass spectrometry technique to

10 Santa Clara 0 -0.1 0 0.1 0.2 0.3 0.4 0.5 Distance (µm)

Fig. 5. Ni concentration profiles across the kamacite–taenite interface in (a) Skookum and (b) Santa Clara. These profiles show kamacite–taenite boundaries that are less well defined than those in Fig. 4 with lower maximum Ni concentrations of 40–45% cf. 50% or higher in the other IVB irons. This difference reflects re-heating that caused redistribution of Ni within the vicinity of the interface. The regions between the dashed lines were used to obtain the cooling rates. tetrataenite bandwidth and the bulk Ni is shown in Fig. 8b. The tetrataenite bandwidth varies from 150 to 265 nm for the 7 meteorites and decreases with increasing Ni content. A positive correlation between cloudy zone particle size and tetrataenite bandwidth was also observed, like in IVA irons (Goldstein et al., 2009b).

4. ERRORS AND COMPARISON WITH PUBLISHED DATA

4.1. Bulk composition and metallographic cooling rate Fig. 7. Ni X-ray scanning images of the low-Ni IVB iron, Cape of Good Hope (a) and the high-Ni iron, Warburton Range (b) showing right to left kamacite (black), tetrataenite (white) and Our bulk Ni and P contents are very consistent with the cloudy taenite intergrowth. The wider tetrataenite rim and coarser data of Campbell and Humayun (2005) and most other cloudy taenite intergrowth in (a) show that Cape of Good Hope analysts (Fig. 3), but are significantly different from those cooled slower than Warburton Range around 300 °C. The dashed of Walker et al. (2008) who obtained bulk P contents from rectangle in (a) shows the region over which the Ni concentration 0.025 to 0.66 wt.%. This range is much broader than the profile was obtained. Thermal history and origin of IVB iron meteorites 4499

orite is smaller than these ranges, we take the plotted error bars to be very conservative estimates of the 2r precision limits for the cooling rates derived for each iron. For example, the mean cooling rate derived from the four Ni profiles for Weaver Mountain is 3500 ± 1000 K/Myr. The 2r uncertainty in the mean extends from 2500 to 4500 K/ Myr, which is considerably smaller than the total range of 1000–9000 K/Myr which is plotted in Fig. 6. The errors in the cooling rates of the IVB irons are much larger than the errors estimated for the IVA and IIIAB irons where it was possible to measure Ni gradients in taenite with the EPMA (Yang and Goldstein, 2006; Yang et al., 2008). Our results indicate that the cooling rates in group IVB irons increase systematically by a factor of 10 from 475 K/Myr at the low-Ni end of the group to 5000 K/ Myr at the high-Ni end. By contrast, Rasmussen et al. (1984) and Rasmussen (1989) inferred from their data that cooling rates did not vary with bulk Ni in group IVB. How- ever, the cooling rate data of Rasmussen (1989) are actually more consistent with our results as his Fig. 2 shows a significant positive correlation between bulk Ni and cooling rate for 10 IVB irons. As pointed out previously, the bandwidth method em- ployed by Rasmussen et al. (1984) and Rasmussen (1989) has significant errors due to the use of an incorrect kamacite nucleation mechanism for the low Ni–P irons. In addi- Fig. 8. Bulk Ni concentrations of seven IVB irons plotted against tion, these two studies lack accurate bandwidth sizes since (a) the size of the high-Ni particles in their cloudy zone, and (b) the the orientation of the kamacite platelets with respect to tetrataenite bandwidth. Both plots show negative correlations. the polished surface of the sample was not obtained. Our Since high-Ni particle size and tetrataenite width are both inversely study avoided these sources of error as we used oriented related to cooling rate, these data confirm that the high-Ni IVB samples and correct kamacite nucleation temperatures irons cooled faster than the low-Ni IVB irons, as was inferred from and mechanisms. kamacite growth modeling (Table 1). The bars for the cloudy zone The cooling rate of a metallic core surrounded by a well particle size for each meteorite represent 1r of the analyses. The insulating silicate mantle decreases with falling tempera- bars for the tetrataenite width for each meteorite show the range of measured values. ture. For the IVB irons, we should expect some increase of cooling rate with bulk Ni even if they had cooled in a well insulated core because of the increase in kamacite measure the bulk composition from a scanned area of nucleation temperature of 100 °C with increasing bulk 250,000 lm2 for each meteorite (50 lm diameter laser Ni. However, this effect should only cause a 15% variation beam, 1 mm scan length, 5 scans). Walker et al. (2008) used in cooling rate (Haack et al., 1990), much less than the 10- the same technique to measure the bulk composition from fold variation we observe across the group. Our analysis for scanned areas of 16,000 or 30,000 lm2 for each meteorite the IVA irons shows that a 10-fold variation in cooling rate (8 or 15 lm diameter laser beam, 0.5 mm scan length, 4 cannot be attributed to uncertainties in the parameters used scans). In our study, we measured the bulk composition to calculate cooling rates (Yang et al., 2007, 2008). in each iron meteorite from total scanned areas from 64,000 to 216,000 lm2 with larger areas for the high-Ni ir- 4.2. Cloudy zone particle size and tetrataenite bandwidth ons. We found that the scanned areas of Walker et al. (2008) were not large enough to cover a representative area Cloudy zone particle sizes and metallographic cooling for the high Ni and P iron meteorites. The Ni and P trend rates of IVB irons (Tables 1 and 2) are plotted in Fig. 9a, shown by the dotted line in Fig. 3a is consistent with a pro- together with other data from mesosiderites, IIIAB and cess of fractional crystallization in the core of the parent IVA irons and the main-group pallasites (Yang et al., 2007; asteroid (Campbell and Humayun, 2005; Walker et al., Goldstein et al., 2009a). Data for the IVB irons closely follow 2008). the general cooling rate vs particle size relationship. We cannot estimate the precision of the cooling rate The data for metallographic cooling rate and tetrataenite data for each IVB iron because the number of taenite Ni bandwidth in several groups, IVA (Goldstein et al., 2009b), gradients obtained for each meteorite was limited. As noted IVB (this study), IIIAB and MG pallasites (Yang et al., above, we show in Fig. 6, the cooling rate ranges given in 2009), are shown in Fig. 9b. The IVB data are clearly consis- Table 1 which give some estimate of the reproducibility of tent with the overall trend in which cloudy zone particle size the measured cooling rates. Since the spread in the cooling and tetrataenite bandwidth decrease with increasing cooling rates derived from individual profiles from the same mete- rate. The fact that cloudy zone particle size and tetrataenite 4500 J. Yang et al. / Geochimica et Cosmochimica Acta 74 (2010) 4493–4506

Table 1 EPMA analysis of bulk Ni and P concentration, inferred kamacite nucleation temperature, and cooling rate determined by taenite Ni proﬁle matching from 17 FIB sections of nine IVB iron meteorites. Meteorite! Ni(±1SEM) P(±1SEM) Total scan area* Kamacite Nuc. CR CR range (wt.%) (wt.%) (lm2) T (°C) (K/Myr) (K/Myr) Iquique (USNM 1230) 15.5(0.06) 0.05(0.003) 72000 470 600 250–700 350 475 Hoba (USNM 6506) 15.6(0.06) 0.07(0.006) 72000 475 400 300–1200 900 650 Tlacotepec (USNM 872) 15.9(0.07) 0.05(0.004) 64000 475 500 400–600 Cape of Good Hope (USNM 985) 16.0(0.07) 0.05(0.004) 216000 475 550 300–800 600 575 Santa Clara (USNM 7191) 16.8(0.09) 0.14(0.03) 184000 590 1500 750–3000 Weaver Mountain (USNM 3154) 17.5(0.06) 0.14(0.02) 192000 580 3000 1000–8000 5000 3000 3000 3500 Tawallah Valley (USNM 1458) 17.5(0.08) 0.19(0.02) 192000 585 8000 1000–9000 2000 5000 Skookum (USNM 536) 17.8(0.14) 0.25(0.07) 192000 580 5000 2500–10,000 Warburton Range (USNM 985) 17.8(0.11) 0.19(0.05) 192000 575 3000 1000–9000 3000 3000 !Meteorite name and source. *Individual scan area is 80 100 lm2. ±1 SEM – one standard error of the mean. bandwidth also decrease with increasing Ni content in the Our data from three independent techniques show that IVB irons (Fig. 8) strongly supports the positive correlation there is a signiﬁcant range of cooling rates among IVB irons between metallographic cooling rate and bulk Ni shown in and that these cooling rates correlate with Ni content. There- Fig. 6. As noted above, the relatively large cooling rate ranges fore the IVB irons could not have cooled in an insulated core. for each measured IVB cooling rate, which are plotted in In group IVB, the Ni-poor members, which crystallized early, Fig. 6, are very conservative estimates of the uncertainty in cooled slower than the Ni-rich members suggesting they were the cooling rates for the IVB irons. located closer to the center, contrary to what was observed in groups IIIAB and IVA. If the IVB core crystallized concentri- 5. DISCUSSION cally, these data suggest that it grew outwards from the center. Below we discuss various constraints on the mode of Fractionally crystallized iron meteorite groups like IVB crystallization of asteroidal cores. are widely thought to be derived from the cores of asteroids that crystallized and cooled while they were surrounded by 5.1. Core crystallization of an asteroid: outside-in vs inside- silicate mantles (Haack and McCoy, 2004). In such bodies, out iron meteorite samples from diverse positions in the core would have nearly identical cooling rates because of the high In planetary cores, the direction of crystallization of a thermal conductivity of metal compared with that of mantle planetary core depends on the relative magnitudes of the and crust materials. However, in groups IVA and IIIAB, the liquidus temperature gradient and the actual temperature irons cooled below 600 °C at diverse rates and could not gradient, which should approach the adiabatic gradient therefore have cooled inside a mantled core (Yang and Gold- assuming ideal convection. In the Earth’s core where pres- stein, 2006; Yang et al., 2007, 2008). The IVA irons appear to sures exceed 100 GPa, the adiabatic thermal gradient is have formed in a metallic body with virtually no silicate man- shallower than the liquidus gradient so that crystallization tle following a glancing impact (Yang et al., 2007). proceeds outwards. Under these conditions, rejection of Thermal history and origin of IVB iron meteorites 4501

Table 2 Cloudy zone (CZ) high-Ni particle size and tetrataenite (Tt) bandwidth measurements for 7 IVB irons. Meteorite Ni CZ particle size (±1r)(nm) Tt (nm) (nm) Iquique 15.5 – – 245 33.1 2.4 282 33.1 2.4 264 Hoba 15.6 26.1 2.5 202 26.9 2.6 183 26.5 2.6 193 Tlacotepec 15.9 29.3 4.5 232 Cape of Good Hope 16.0 31.5 4.1 237 27.1 2.4 268 20.7 2.9 169 – – 266 26.4 3.1 236 Weaver Mtn 17.5 16.0 4.2 130 – – 127 16.4 1.7 152 17.4 2.8 147 20.0 2.5 185 26.3 2.6 188 19.2 2.8 155 Tawallah Valley 17.5 – – 162 Fig. 9. Metallographic cooling rates of irons in groups IIIAB, – – 140 IVA, and IVB and the stony-iron meteorites, main-group pallasites – – 144 (MG Pall.) and mesosiderites (Meso.), show an inverse correlation 23.3 1.9 155 with (a) the sizes of high-Ni particle in the cloudy zone microstructure and (b) the tetrataenite bandwidths. Although there 23.2 1.9 151 is more scatter at the fast-cooled end of the line, the IVB data are Warburton Range 17.8 17.4 2.1 152 quite consistent with the trends established by the other groups 23.4 1.9 181 showing that tetrataenite bandwidth and cloudy zone particle size 21.1 2.9 180 can be used to infer relative cooling rates at 300 °C in diverse meteorites. 20.6 2.3 171 ‘–’ Indicates the image quality was not sufficient for an accurate measurement of CZ particle size. coefficient a. A value of a = 9.2 105 K1 for pure iron has been widely adopted (Anderson and Ahrens, 1994). However, Williams (2009) prefers the value for a of 13.2 105 K1 recommended by Assael et al. (2006).If light-elements like S and release of latent heat promote con- the smaller a value is used, dT ¼ 28:4 K/GPa at the melting vection and dynamo action. For much smaller bodies like dP point of pure iron, but the larger a value gives dT ¼ 40:8K/ asteroids which have central pressures <1 GPa, the situa- dP GPa at melting point. tion is much less clear. Haack and Scott (1992) and Wil- To derive the melting point gradient for pure iron, Wil- liams (2009) infer that the adiabatic gradient across the liams (2009) prefers the data of Strong et al. (1973) up to core would be steeper than the liquidus gradient and that 5 GPa, which give a gradient of 35 K/GPa. However, addi- the core should therefore crystallize inwards. However, tional data for pressures up to 5–10 GPa favor a slightly uncertainties in the properties of liquid Fe–Ni–S and kinetic higher melting point gradient of 39 K/GPa (Boehler, factors discussed below may negate this conclusion. 1993; Saxena et al., 1994; Saxena and Dubrovinsky, 2000). The adiabatic temperature gradient can be defined by To illustrate the very small temperature differences ex- Eq. (1) (Jacobs, 1987; Haack and Scott, 1992; Boehler, pected across convecting asteroidal cores, we show in 1993), Fig. 10, adiabatic and liquidus gradients calculated for a dT aT pure Fe core in an asteroid of radius 150 km. Pressures were ¼ ð1Þ dP qCp calculated using the hydrostatic equation (Jacobs, 1987) assuming the metallic core has half the radius of the body. where a is the pressure-dependent coefficient of volume The densities of core and mantle materials were taken as thermal expansion, T is the temperature, q is the density, 3000 kg/m3 for the silicate mantle (Haack and Scott, 3 and Cp is the specific heat at constant pressure. The largest 1992) and 7019 kg/m for liquid iron (Anderson and Ah- uncertainty in these parameters is for the thermal expansion rens, 1994). The curves in Fig. 10 were derived using the 4502 J. Yang et al. / Geochimica et Cosmochimica Acta 74 (2010) 4493–4506

1815 gest that S-build up promoted the formation of inwards growth of dendrites with lengths approaching the core ra- 1814 dius. Convection may then have been driven by S-rich li- Successive adiabatic temperature curve quid rising along the dendrites. 1813 Melting point curve Another factor that may have aﬀected the crystallization mode in asteroidal cores is the availability of nuclei for Fe– 1812 Ni crystals. Homogeneous nucleation of solid iron in liquid iron requires undercooling of 300 K (Turnbull and Cech, 1811 1950). Alternatively, solid iron may have nucleated heteroge- neously on solid metal, silicate or oxides. Silica inclusions are mantle 1810 present in the IVB iron, Santa Clara (Teshima and Larimer,

T (K) 1983), but their concentration around sulfide nodules suggests they may not have nucleated solid Fe–Ni. Solid Fe–Ni 1809 nuclei may have reached the core from dendrites that grew on the core-mantle boundary and became detached (Wasson, 1808 1993). Alternatively, peak core temperatures may never have exceeded the liquidus temperature so that residual metal crys- 1807 tals were available at the center for outwards growth. Olivine nuclei would also have been plentiful at the core- 1806 mantle boundary, but the activation energy barrier for het- CMB a erogeneous nucleation (DGhet ¼ SðhÞDGhom) on olivine is 1805 much higher than for solid metal because of the difference 0 0.1 0.2 0.3 0.4 0.5 0.6 in wetting angle, which determines the shape factor center r/R 2 SðhÞ¼ð2 þ cos hÞð1 cos hÞ =4. The wetting angle h be- Fig. 10. Melting point curve for a pure iron body of radius 150 km tween liquid iron and olivine is about 94° (Eckhardt, with adiabatic curves showing successive thermal gradients assum- 1995) so the shape factor SðhÞ¼0:55. The wetting angle be- ing ideal convection, a thermal expansion coefficient of 9.2 tween liquid iron and solid iron is unknown but it is ex- 105 K1 and a melting point gradient of 39 K/GPa (see text). pected to be close to 0° (Eustathopoulos et al., 1999). Under these conditions, equilibrium crystallization would begin at Assuming h = 5°, then the shape factor for solid metal the center. However, the small temperature difference across the nucleation SðhÞ¼1:2 105. Thus the activation energy core of only a few degrees K suggests that kinetic factors may have barrier for heterogeneous nucleation on solid Fe–Ni is 5 or- been more important in establishing the crystallization mode in ders of magnitude less than on olivine. asteroidal cores. CMB is the core-mantle boundary. In summary, adiabatic and thermal gradients in asteroi- 5 1 dal cores may have favored inwards growth except possibly lower a value of 9.2 10 K and a melting point gradient of 39 K/GPa. With these data, one would predict in low-S cores. However, the small temperature gradients outwards growth of the core assuming equilibrium crysta- across asteroidal cores suggest that kinetic factors and ini- llization. Use of the higher a value would indicate inwards tial conditions may have been more important than equilib- equilibrium crystallization. Williams (2009) has investigated rium relationships. Concentric inwards growth in a mantled the effects of Ni and S and infers that even with the lower a core would enhance the S concentration in the outer layer value, asteroidal cores with >5 wt.% S should crystallize precluding global convection and fractional crystallization. inwards under equilibrium conditions. However, evidence for fractional crystallization in iron Despite uncertainties in the adiabatic and liquidus tem- meteorites suggests that convection was vigorous. Inwards perature gradients, it is clear that temperature differences growth of multi-km-sized dendrites (Haack and Scott, across asteroidal cores assuming ideal convection should 1992) remains speculative as cooling rate data for iron not have exceeded a few degrees K. The mode of crystalli- meteorites in groups IIIAB, IVA and IVB suggest concen- zation may therefore have depended on other factors. For tric growth. Outwards growth in low-S cores may have re- example, Haack and Scott (1992) infer that lateral temper- sulted from incomplete melting of Fe–Ni and supercooling ature variations at the base of the mantle resulting from in the outer regions of the core due to the difficulty in nucle- mantle convection or topographic variations may have ex- ating on olivine at the core-mantle boundary. ceeded the temperature variation across the core. As a result, crystallization may have commenced at the coldest 5.2. Crystallization and origin of the IVB Irons sites at the base of the mantle. Once crystallization starts, S is rejected from the growing solid and may build up in Although we cannot exclude the possibility that the IVB the adjacent liquid without vigorous convection. Since 1 core grew inwards with multi-km long dendrites, the cool- wt.% S will lower the liquidus temperature by 20 K (Li ing rate data favor simple outwards concentric crystalliza- and Fei, 2003), even minor S gradients can quickly affect tion. Since the IVB meteorites have the lowest the course of crystallization. Concentric inwards growth concentrations of troilite of any group (Buchwald, 1975), in a mantled core would enhance the S concentration in the low S abundance in the core may have promoted the outer layer precluding global convection and fractional outwards growth. Outwards growth is also consistent with crystallization. Alternatively, Haack and Scott (1992) sug- the chemical trends in IVB, which require fractional Thermal history and origin of IVB iron meteorites 4503 crystallization (Walker et al., 2008), and vigorous convec- 28 tion. If outwards crystallization is typical of mantled aster- 2 wt% S 26 oidal cores, the inwards crystallization of the IVA and 1 wt% S IIIAB metallic bodies may simply result from mantle re- 24 moval before crystallization rather than afterwards (Yang et al., 2007, 2008). Steep thermal gradients in the outer 22 metallic portions may have promoted convection in the IIIAB and IVA bodies. 20 Our explanation for the diverse cooling rates of the IVB Ni (wt%) irons requires that the mantle was removed after the IVB 18 irons crystallized. Fractional crystallization modeling suggests that the IVB irons sample only 70–80% of the whole 16 core (Walker et al., 2008). Thus the IVB mantle could have 14 been removed before the outer 20–30% of the core had crys- 0 0.2 0.4 0.6 0.8 1 tallized. Mantle removal by a glancing hit-and-run impact r/R when the outer 20–30% was still liquid seems more plausible than post-solidification removal as the latter would presum- Fig. 12. Ni concentration in solid Fe–Ni metal as a function of ably have fragmented much of the outer parts of the core. radial distance r from the center of a liquid core of radius R that Our proposed history for the IVB irons is shown sche- crystallizes from the center of the core outwards. Initial bulk S concentrations of 1 and 2 wt.% are assumed. Dashed lines (r/R =0 matically in Fig. 11. Before core solidification, the parent to r/R = 0.9) show the range of bulk Ni concentrations, 15.5– body had a liquid or largely liquid core with radius R sur- 18.5 wt.% (Campbell and Humayun, 2005; Walker et al., 2008)in rounded by a mantle (Fig. 11a). During cooling, the solid group IVB irons. The IVB irons are core samples taken from the Fe–Ni grew from the center of the core which fractionally center to r/R = 0.90. The unsampled shaded section where r/ crystallized (Fig. 11b). When the radius of the solid core R = 0.9–1.0 accounts for 25% of the core volume. was R0 (Fig. 11c), a glancing impact with a larger body (Fig. 11d) exposed the solid core and dispersed the residual concentration (Chabot and Jones, 2003). Since the chemical molten metal (Fig. 11e). The solid core containing the IVB trends in group IVB are well modeled using bulk S contents iron meteorites cooled below 600 °C without a surrounding of 0–2 wt.% (Walker et al., 2008; Goldstein et al., 2009a), mantle (Fig. 11f). Below we estimate the radii of the solid we assumed S contents of the liquid core of 1 and 2 wt.% and liquid cores (R0 and R) and the locations of the IVB ir- S. The Ni content of the liquid core was fixed at ons using fractional crystallization and thermal models. 17.6 wt.% Ni so that the first solid to crystallize had a Ni Fractional crystallization of the IVB irons was modeled concentration equal or close to that of the lowest Ni IVB using elemental distribution coefficients that vary with S iron (15.5%). Following the procedure used by Yang

Fig. 11. Cartoon showing the evolution of the group IVB asteroid. (a) Parent body with a liquid core (red) of radius R surrounded by a mantle (blue). (b) Fractional crystallization of the core begins in the center of the metallic core (brown). (c) Fractional crystallization continues until about 80 % of the original liquid core has solidified and the solid core has a radius R0. (d) Glancing impact with a much larger asteroid breaks up the IVB parent body forming in (e) the exposed solid core (brown), dispersed liquid metal (red), and mantle fragments (blue). (f) The solid inner core of radius R0 cools to lower temperatures without a surrounding mantle and is subsequently broken up to form the IVB meteorites. We infer from thermal and fractional crystallization models that the radius (R) of the liquid core in (a) was 70 km while the radius (R0) of the solid core in (c–f) was 65 km. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) 4504 J. Yang et al. / Geochimica et Cosmochimica Acta 74 (2010) 4493–4506 et al. (2008) for the IVA irons, we generated a relationship r/R for this calculation is given by the Ni vs r/R plot between the Ni content of the solidified iron and the frac- (Fig. 12). Using Fig. 13, we determined that the IVB irons tion of the core that has solidified (fs) for the two S con- with the fastest cooling rate of 5000 K/Myr were located at tents. Assuming outwards growth, the Ni content in the a radial distance r/R0 = 0.95, or 62 km from the center of solidified core was then obtained as a function of the nor- the solid body. The fastest cooled IVB irons have Ni contents malized radius r/R (Fig. 12). In this model, the IVB iron of 17.5 to 17.8 wt.% Ni (Table 1). Using the calculated Ni dis- meteorites with the highest bulk Ni contents (18.5 wt.%; tribution in the solidified core as a function of the normalized Campbell and Humayun, 2005; Walker et al., 2008) would radius r/R (Fig. 12), it is observed that meteorites with the have crystallized at a radial position r/R = 0.90 when 75% fastest cooling rates are located at a position r/R of 0.88. of the original liquid had crystallized. We assume that resid- Since r is 62 km, the original liquid core radius is 70 km. ual liquid between R0 and R was dispersed in the impact By combining the results from the fractional crystalliza- that removed the mantle (Fig. 11e). tion model (Ni content vs r/R, Fig. 12), and the thermal model (cooling rate vs r/R, Fig. 13), we calculate the cool- 5.3. Thermal history of the IVB metallic asteroid ing rate vs bulk Ni relationship for samples from a solid metallic core of radius 65 km that crystallized outwards in The cooling rate of the IVB metallic asteroid was investi- a molten core with a radius 70 km and a bulk S concentra- gated using the thermal model developed for the IVA aster- tion of 1 and 2 wt.% (Fig. 14). Fig. 14 shows an excellent oid in which the initial temperature was taken as 1760 K match between the calculated cooling rate vs bulk Ni vari- (Yang et al., 2008). Assuming that the IVB iron with the slow- ations and that obtained from the metallographic cooling est cooling rate of 475 K/Myr (Table 1) was located close to rates of the IVB irons (Table 1). Although both S contents the center of the metallic body and that the metal core essen- match the data, the 1 wt.% curve is a better fit. tially lacked a silicate mantle, we calculate that the metallic The metallographic cooling rate range for the four low-Ni core in which the kamacite plates formed had a radius (R0) IVB irons is about 350–900 K/Myr (Column CR (K/Myr) in of 65 km (Fig 11f). Since the kamacite nucleation tempera- Table 1). By using this cooling rate range to determine the tures of the IVB irons vary with Ni content from 740 to uncertainty in the calculation of the inner core size, we calcu- 860 K (Table 1), we calculated the cooling rate at the kama- late that the radius of the solid metallic body size was between cite nucleation temperature vs r/R0 from the thermal model. 50 km and 80 km. Therefore, the radius of the inner core was Fig. 13 shows the calculated cooling rate as a function of 65 ± 15 km, and that of the original liquid core was the normalized radial distance r/R0 from the center of the core 70 ± 15 km. Assuming that the original liquid core was half (upper abscissa) as well as the radial distance r/R in the origi- the radius of the differentiated body (Fig. 11a), we infer that nal core (lower abscissa). The relationship of Ni content to the original IVB body was 140 ± 30 km in radius before breakup.

r/R' 5.4. Possible relation with angrites 0 0.25 0.5 0.75 1 10000 Campbell and Humayun (2005) suggested that the IVB irons might be derived from the same body as the angrites,

10000

1 wt% S 1000

2 wt% S 1000 Cooling rate (K/Myr)

100 Cooling Rate (K/Myr) 0 0.2 0.4 0.6 0.8 1 r/R 100 15 16 17 18 19 Fig. 13. Cooling rate as a function of normalized radial distance Bulk Ni (wt%) r/R0 from the center of the cooling body (radius R0 = 65 km) derived using the thermal model of Yang et al. (2008). The lower Fig. 14. Comparison between the metallographic cooling rates for horizontal axis shows the radial position normalized to the radius IVB irons plotted as a function of bulk Ni concentration (Table 1; R of the original liquid metallic core. The cooling rate is calculated Fig. 7) and those calculated using the fractional crystallization and at the kamacite nucleation temperature which varies from 740 to thermal models (Figs. 12 and 13) for a solid metallic core of radius 860 K. The horizontal dashed lines show the minimum and 65 km that crystallized outwards in a molten core with a radius maximum range of cooling rates for the IVB irons, Table 1. The 70 km and bulk S concentration of 1 and 2 wt.%. Both curves shaded outermost part of the original liquid core is assumed to provide an excellent fit to the data, though the 1 wt.% S curve is a have been lost during the impact event that removed the mantle. better match for the data for the irons with >16.5% Ni. Thermal history and origin of IVB iron meteorites 4505 which are more depleted in Na, K and other volatile elements simple outwards concentric crystallization. Then high-Ni ir- than any other basalts. The angrites record magnetic fields on ons that crystallized late would have been located nearer the order of 10 lT that lasted for at least 8 Myr (Weiss et al., the surface and cooled more rapidly than low-Ni irons. 2008) around the time when the IVB irons crystallized (Walk- If the core grew concentrically outwards as we infer, our er et al., 2008). Indeed, the IVB body may well have had a dy- thermal and crystallization models suggest that the radius namo-generated magnetic field during core crystallization, as of the core was 65 ± 15 km when it cooled without a man- fractional crystallization and dynamos both require efficient tle. The mantle was probably removed when the IVB body convection (Nimmo, 2009). However, oxygen isotope data was torn apart in a glancing impact with a larger body, as for IVB irons and more siderophile element data for angrites these “hit-and-run” impacts were probably the dominant are needed to test this possible link. mechanism for extracting core material from differentiated asteroids (Asphaug et al., 2006). Clean removal of the man- 6. SUMMARY tle from the solid core by a hit-and-run impact may have been aided by a thin layer of residual metallic melt at the We produced electron transparent thin sections of core-mantle boundary, which was lost during the impact. kamacite–taenite interfaces from the polished surfaces of The IVB irons may therefore have crystallized in a core that nine group IVB irons. This allowed us to measure the was slightly larger, 70 ± 15 km in radius, in a body that was 2–3 lm long Ni concentration profiles normal to the plane probably around 140 ± 30 km in radius. of the kamacite–taenite interface with a resolution of Outwards crystallization in asteroidal cores would have 2–4 nm using analytical transmission electron microscopy. promoted convection but it is counter to the predictions of By comparing these profiles with those generated from Williams (2009), who argued for top–down crystallization, computer modeling of kamacite growth using our measure- except possibly in cores with low S (like the IVB core). ments of bulk Ni and P concentrations, we have determined However, inwards growth would have enriched S in the that the nine IVB irons cooled at rates of between 475 and outer liquid zones inhibiting convection and fractional crys- 5000 K/Myr during kamacite growth at 600–400 °C. These tallization. Since the thermal gradient across the core would cooling rates are faster than those of most other iron and have been less than a few degrees K, it is possible that ki- stony-iron meteorites and are correlated with the bulk Ni netic and other factors controlled crystallization of asteroi- concentrations. dal cores rather than the relative gradients of the adiabatic Seven of the nine IVB irons have cloudy taenite inter- and liquidus curves. The outside-in crystallization that we growths and tetrataenite layers between the cloudy taenite inferred for the IVA and IIIAB irons may have resulted and the kamacite platelets. Sizes of the high-Ni particles from rapid cooling and crystallization following mantle re- in the cloudy taenite at the interface with the tetrataenite moval (Yang et al., 2008). rim and widths of the tetrataenite rims in the IVB irons are inversely correlated with bulk Ni concentrations. Since ACKNOWLEDGMENTS the widths of cloudy taenite particles and the tetrataenite rims in a wide variety of iron and stony-iron meteorites The financial support from NASA through Grant are inversely correlated with the cooling rates during kama- NNX08AG53G (J.I. Goldstein, P.I.) and NNX08AE08G (K. Keil, cite growth, these data strongly support our conclusion P.I.) and NNX08AI43G (E. Scott, P.I.) is acknowledged. Sandia is from kamacite growth modeling that the cooling rates of a multi-program laboratory operated by Sandia Corporation, a IVB irons are correlated with their bulk Ni concentrations. Lockheed Martin Company, for the United States Department of In addition they confirm that IVB irons cooled faster than Energy’s National Nuclear Security Administration under contract nearly all other iron and stony-iron meteorites. DE-AC0494AL85000. We thank Michael Rye (Sandia) for assis- The two IVB irons, Santa Clara and Skookum that lack tance with the FIB samples, Tim McCoy, Smithsonian Institution for the loan of the IVB meteorites, and Nancy Chabot, Henning cloudy taenite have lower peak Ni concentrations in taenite Haack, Misha Petaev, and associate editor Richard Walker for at the kamacite interface (40–45% cf. >50% in the other helpful reviews and suggestions. IVB irons) and less steep Ni profiles close to this peak indicating re-heating above 400 °C. Because their Ni profiles were undisturbed at distances 100 and 200 nm from the REFERENCES kamacite–taenite interface, cooling rates could still be determined from the rest of the taenite Ni profile. Skookum Anderson W. W. and Ahrens T. J. (1994) An equation of state for was probably heated after atmospheric entry (Buchwald, liquid iron and implications for the Earth’s core. J. Geophys. 1975), but for Santa Clara, which was much less affected, Res. 99, 4273–4284. earlier shock heating is more probable. Assael M. J., Kakosimos K., Banish R. M., Brillo J., Egry I., If the group IVB irons had cooled in a core that was en- Brooks R., Questad P. N., Mills K. C., Nagashima A., Sato Y. cased by a silicate mantle, their cooling rates would be uni- and Wakeham W. A. (2006) Reference data for the density and form. The 10-fold variation in cooling rate for the IVB viscosity of liquid aluminum and iron. J. Phys. Chem. Ref. Data 35, 285–300. irons (Table 1) suggests instead that the IVB core cooled Asphaug E., Agnor C. B. and Williams Q. (2006) Hit and run through 600 to 400 °C after the mantle had been removed. planetary collisions. Nature 439, 155–160. Although we cannot exclude the possibility that the IVB Boehler R. (1993) Temperatures in the Earth’s core from melting- core grew inwards with multi-km long dendrites, the posi- point measurements of iron at high static pressures. Nature 363, tive correlation between bulk Ni and cooling rate suggests 534–536. 4506 J. Yang et al. / Geochimica et Cosmochimica Acta 74 (2010) 4493–4506

Buchwald V. F. (1975) Handbook of Iron Meteorites. Their History, Rasmussen K. L., Malvin D. J., Buchwald V. F. and Wasson Distribution, Composition and Structure 1. University of Cali- J. T. (1984) Compositional trends and cooling rates of fornia Press, Berkeley, p. 262. group IVB iron meteorites. Geochim. Cosmochim. Acta 48, Campbell A. J. and Humayun M. (2005) Compositions of group 805–813. IVB iron meteorites and their parent melt. Geochim. Cosmo- Romig A. D. and Goldstein J. I. (1980) Determination of the Fe– chim. Acta 69, 4733–4744. Ni and Fe–Ni–P phase diagrams at low temperatures (700 to Chabot N. L. and Haack H. (2006) Evolution of asteroidal cores. 300 °C). Metall. Trans. 11A, 1151–1159. In Meteorites and the Early Solar System II (eds. D. S. Lauretta Saxena S. K. and Dubrovinsky L. S. (2000) Iron phases at high and H. Y. McSween). Univ. Ariz. Press, Tucson, pp. 747–771. pressures and temperatures: phase transition and melting. Am. Chabot N. L. and Jones J. H. (2003) The parameterization of solid Mineral. 85, 372–375. metal–liquid metal partitioning of siderophile elements. Mete- Saxena S. K., Shen G. and Lazor P. (1994) Temperatures in earth’s orit. Planet. Sci. 38, 1425–1436. core based on melting and phase-transformation experiments Cliff G. and Lorimer G. W. (1975) The quantitative analysis of thin on iron. Science 264, 405–407. specimens. J. Microsc. 103, 203–207. Strong H. M., Tuft R. E. and Hanneman R. E. (1973) Iron fusion Eckhardt M. M. (1995) Melt segregation in planetary interiors: curve and gamma-delta-l triple point. Metall. Trans. 4, 2657– evidence from the Brenham pallasite. Lunar and Planet. Sci. 2661. Conf. 26, 349–350. Teshima J. and Larimer J. W. (1983) The IVB iron meteorite Eustathopoulos N., Nicholas M. G. and Drevet B. (1999) Wettability parent body: Inferences based on inclusion mineralogy and at High Temperatures. Elsevier Science Ltd., Oxford, UK. composition (abstract). Meteoritics 18, 406–407. Goldstein J. I. and Michael J. R. (2006) The formation of plessite in Turnbull D. and Cech R. E. (1950) Microscopic observation of meteoritic metal. Meteorit. Planet. Sci. 41, 553–570. solidification of small metal droplets. J. Appl. Phys. 21, 804– Goldstein J. I. and Ogilvie R. E. (1965) The growth of the 810. Widmansta¨tten pattern in metallic meteorites. Geochim. Cos- Walker R. J., McDonough W. F., Honesto J., Chabot N. L., mochim. Acta 29, 893–920. McCoy T. J., Ash R. D. and Bellucci J. J. (2008) Modeling Goldstein J. I., Scott E. R. D. and Chabot N. L. (2009a) Iron fractional crystallization of group IVB iron meteorites. Geo- meteorites: crystallization, thermal history, parent bodies and chim. Cosmochim. Acta 72, 2198–2216. origin. Chemie der Erde 69, 293–325. Wasson J. T. (1985) Meteorites: Their Record of Early Solar- Goldstein J. I., Yang J., Kotula P., Michael J. R. and Scott E. R. System History. W.H. Freeman, p. 267. D. (2009b) Thermal histories of IVA iron meteorites from Wasson J. T. (1993) Role of body crystallization in asteroidal cores. transmission electron microscopy of the cloudy zone micro- Meteoritics 28, 455–456 (abstr.). structure. Meteorit. Planet. Sci. 44, 343–358. Weiss B. P., Berdahl J. S., Elkins-Tanton L., Stanley S., Lima E. A. Haack T. E. and McCoy T. J. (2004) Iron and stony-iron and Carporzen L. (2008) Magnetism on the angrite parent body meteorites. In Meteorites, Comets, and Planets. Treatise on and early differentiation of planetesimals. Science 322, 713–716. Geochemistry (ed. A. M. Davis) (eds. H. D. Holland and K. I. Williams Q. (2009) Bottom-up versus top–down solidification of Turekian). Elsevier-Pergamon, Oxford, pp. 325–345. the cores of small solar system bodies: constraints on paradox- Haack H. and Scott E. R. D. (1992) Asteroid core crystallization ical cores. Earth Planet. Sci. Lett. 284, 564–569. by inward dendritic growth. J. Geophys. Res. 97, 14727–14734. Wood J. A. (1964) The cooling rates and parent planets of several Haack H., Rasmussen K. L. and Warren P. H. (1990) Effects of iron meteorites. Icarus 3, 429–459. regolith/megaregolith insulation on the cooling histories of Yang C.-W., Williams D. B. and Goldstein J. I. (1997) A new differentiated asteroids. J. Geophys. Res. 95B, 5111–5124. empirical cooling rate indicator for meteorites based on the size Hevey P. J. and Sanders I. S. (2006) A model for planetesimal of the cloudy zone of the metallic phases. Meteorit. Planet. Sci. meltdown by 26Al, and its implications for meteorite parent 32, 423–429. bodies. Meteorit. Planet. Sci. 41, 95–106. Yang J. and Goldstein J. I. (2005) The formation of the Jacobs J. A. (1987) The Earth’s Core, second ed. Academic Press, Widmansta¨tten structure in meteorites. Meteorit. Planet. Sci. London. 40, 239–253. Kelly W. R. and Larimer J. W. (1977) Chemical fractionations Yang J. and Goldstein J. I. (2006) Metallographic cooling rates of in meteorites-VIII. Iron meteorites and the cosmochemical the IIIAB iron meteorites. Geochim. Cosmochim. Acta 70, 3197– history of the metal phase. Geochim. Cosmochim. Acta 41, 93– 3215. 111. Yang J., Goldstein J. I. and Scott E. R. D. (2007) Iron meteorite Li J. and Fei Y. (2003) Experimental Constraints on Core evidence for early catastrophic disruption of protoplanets. Composition. In Treatise on Geochemistry, vol. 2 (ed. R. W. Nature 446, 888–891. Carlson). Elsevier, pp. 521–546, p. 568. Yang J., Goldstein J. I. and Scott E. R. D. (2008) Metallographic Moore C. B., Lewis C. F. and Nava D. (1969) Superior analysis of cooling rates and origin of IVA iron meteorites. Geochim. iron meteorites. In Meteorite Research (ed. P. M. Millman). Cosmochim. Acta 72, 3043–3061. Springer-Verlag, New York, pp. 738–748. Yang J., Goldstein J. I., Scott E. R. D., Yu H. and Zhu M. (2009) Nimmo F. (2009) Energetics of asteroid dynamos and the role of Cooling rates and origin of main group pallasites. Meteorit. compositional convection. Geophys. Res. Lett. 36, L10201. Planet. Sci. 44, A224. Rasmussen K. L. (1989) Cooling rates and parent bodies of iron meteorites from Group IIICD, IAB, and IVB. Phys. Scr. 39, Associate editor: Richard J. Walker 410–416.