A Time‐Resolved Paleomagnetic Record of Main Group Pallasites: Evidence for a Large‐Cored, Thin‐Mantled Parent Body

Several paleomagnetic studies have been conducted on five Main Group pallasites: Brenham, Marjalahti, Springwater, Imilac, and Esquel. These pallasites have distinct cooling histories, meaning that their paleomagnetic records may have been acquired at different times during the thermal evolution of their parent body. Here, we compile new and existing data to present the most complete time‐resolved paleomagnetic record for a planetesimal, which includes a period of quiescence prior to core solidification as well as dynamo activity generated by compositional convection during core solidification. We present new paleomagnetic data for the Springwater pallasite, which constrains the timing of core solidification. Our results suggest that in order to generate the observed strong paleointensities (∼65–95 μT), the pallasites must have been relatively close to the dynamo source. Our thermal and dynamo models predict that the Main Group pallasites originate from a planetesimal with a large core (>200 km) and a thin mantle (<70 km).

The pallasites have an enigmatic formation history. They have commonly been interpreted to represent the core-mantle boundary, but this has been questioned given their paleomagnetic record of the dynamo, which suggests that they must have formed at a much lower temperature than the molten core (Bryson et al., 2015;Mckibbin et al., 2019;Nichols et al., 2016;Tarduno et al., 2012). In addition, their varied cooling rates have been used to argue that they formed at a variety of depths in a disorganized asteroid reassembled after impact (Yang et al., 2010). Most recently, it has been suggested that the pallasites could originate from a mantle-stripped metallic asteroid, such as asteroid (16) Psyche (Elkins-Tanton et al., 2020). This would resolve the discrepancy between shallow and deep origins for the pallasites, if they originate in the middle of thin mantle that is a short distance from the core-mantle boundary. To resolve the pallasite controversy, Tarduno et al. (2012) suggested an impact whereby dykes from the impactor core penetrated into the parent body mantle. Alternatively, Johnson et al. (2019) recently suggested a mechanism for pallasite formation by which sulfur-rich metallic melts at the core-mantle boundary penetrate a shallow overlying mafic mantle. These authors also suggest that the paleomagnetic record of the pallasites may be consistent with a dynamo driven by inward core solidification Scheinberg et al., 2016).
Here, we present the most complete paleomagnetic record yet of the MG pallasite parent body. We report a new X-PEEM paleointensity for the Springwater pallasite cloudy zone and compile previous paleointensity measurements of the cloudy zone and olivine inclusions in the Brenham, Marjalahti, Imilac, and Esquel pallasites. We discuss the implications for the size, internal structure, and nature of core solidification on the parent body. We argue that the pallasites may originate from a parent body similar to the metallic asteroid (16) Psyche and could originate from within a thin mantle above, but closer to, the core-mantle boundary.
to prevent heating. The sample was then polished with diamond paste grade 9-0.25 μm. The sample was etched for ∼30 s with nital (2% nitric acid in ethanol) to reveal the FeNi microstructures and was examined using reflected light microscopy to ensure there was no evidence of terrestrial weathering or shock. The sample was then repolished and sputtered for 10 h at 1.2 keV, 12 h at 0.8 keV, and 1.5 h at 0.4 keV using a focused Ar-ion beam under ultra-high vacuum (pressure < 1.5 × 10 −5 mbar) to remove any oxidation and surface magnetization induced by polishing and to minimize surface topography due to differential etching. The sample was kept in vacuum between sputtering and measuring (measuring pressure < 1.0 × 10 −8 mbar).

Experimental Method
X-PEEM measurements were performed at the SPEEM UE49 beamline, BESSY II, Berlin (Kronast et al., 2010). A beam of monochromatic X-rays was focused at 16° to the sample surface, exciting secondary photoelectrons from the top ∼5 nm of the sample surface. The X-rays were tuned to the Fe L 3 edge (∼707 eV) and linearly polarized to acquire compositional images. Images of the projection of the magnetic moment onto the X-ray beam were acquired using X-ray magnetic circular dichroism (XMCD). Images were acquired with a 10 μm field of view (with a pixel size of 19 × 19 nm). In total, 160 images were collected (80 for each polarization) with an exposure time of 2 s, and then averaged. Averaged images were also corrected for intensity drift due to minor sample charging (see Section S1.1.1).
The X-PEEM technique has been further developed since previous experiments, which only measured one magnetization component (Bryson et al., 2014(Bryson et al., , 2015Nichols et al., 2016Nichols et al., , 2018. We imaged Springwater using the three rotation approach outlined by , which enables the full vector magnetization to be recovered. Once the length of the tetrataenite rim-cloudy zone interface had been imaged, the sample was rotated 109° and the same interface was remeasured; the sample was then rotated 120° and remeasured, resulting in three data sets at different angles relative to the X-ray beam. We collected 12, 8, and 7 averaged and corrected images for rotations 1, 2, and 3, respectively.

Paleointensity Estimates
The magnitude and direction of the paleomagnetic field recorded by the Springwater cloudy zone was determined following the method outlined by . The mean XMCD intensities in each of the three measured orientations (I 1 , I 2 , and I 3 ) were used to reconstruct the 3D vector representing the paleomagnetic field at the time of remanence acquisition:  M V  I  I  I B  I  I B  I  I B  I  I  I  I  I  I  k T (1)  I  I  I B  I  I B  I  I B  I  I  I  I  I  I  k T (2)  M V  I  I  I B  I  I B  I  I B  I  I  I  I  I  I  k T   (3) where M s is the saturation magnetization of tetrataenite (1.12 × 10 6 Am −1 ), V is the volume of tetrataenite islands at the time of remanence acquisition, which we take to be 78% of their present-day size (Maurel et al., 2019), k B is the Boltzmann constant, and T 0 is the tetrataenite formation temperature (593 K). The domains in the tetrataenite rim correspond to the XMCD intensities of the orthogonal easy axes I x , I −x , I y , I −y , I z , and I −z (Table S3). Although the tetrataenite rim intensities do not directly correspond to one another in different rotations (i.e., we did not measure the exact same field of view for each rotation), the uncertainty in how the intensities change with rotation has a negligible effect (<1 μT) on the calculated paleointensity.
Previously calculated paleointensities for the pallasites (Bryson et al., 2015;Nichols et al., 2016) are also reassessed here. Since these estimates were based on one measurement orientation, they represent lower limits on the paleointensity, since it is unknown how the paleofield vector relates to the measurement direction. If the paleomagnetic field direction is parallel to the measurement direction, 100% of the paleointensity is recovered, whereas if the field is perpendicular to the measurement direction, 0% of the paleointensity is recovered. The recovered paleointensity (I rec ) is proportional to the true paleointensity of the natural remanent magnetization (I NRM ) depending on the angle, θ, between the measurement direction and the paleofield vector: We calculated the 95% confidence for I NRM by bootstrapping over all possible values of θ. We sampled 100,000 values of θ distributed evenly over a sphere for the measured values of I rec for Brenham, Marjalahti, Imilac, and Esquel. We recovered a one-tailed distribution of values for I NRM for each sample. Details of the bootstrapping method are given in the supplementary material, section S1.1.2.

Sample Preparation
Samples of Springwater were borrowed from the American Museum of Natural History (SW) and the Bowers Q.D. Collection, Harvard Museum of Natural History. A sample of Imilac (IM) was purchased from the meteorite collector "Meteorite Madness." A summary of the samples used in this study and the associated preparation and experimental techniques is shown in Table 1. Samples were selected following the same protocol as Tarduno et al. (2012); regions of gem-quality olivines showing no visible inclusions were preferentially sampled. Gem-quality olivines are identified by conchoidal fracture, green coloration (rather than orange, which suggests alteration), and a high degree of translucency. Gem-like olivines 2-3 mm in diameter from the Esquel and Imilac pallasites, sampled >5 mm from the meteorite edge and several millimeters from the metal-olivine contact were found to have recoverable natural remanent magnetization (NRM) that was replicable within and between meteorite samples (2 from each meteorite). The success rate from the sampled olivines was 50% if the NRM was of order 10 −9 to 10 −10 A m 2 (only 15% of specimens analyzed). We therefore also preferentially selected samples with NRM moments of this magnitude.
The sample of Springwater from the Harvard Museum of Natural History (SWH) was a large slab (weight ∼ 250 g, thickness ∼ 1 cm) ( Figure S5a). A small (∼5 cm diameter) section was removed from one corner of the slab using an Buehler IsoMet® low speed saw ( Figure S5b). All further sampling was carried out using a Well precision diamond wire saw at the MIT Paleomagnetism Laboratory. Cutting took place in a magnetically shielded clean room (DC field <200 nT), and the saw was thoroughly cleaned between sampling. Mutually oriented samples with a diameter of ∼1-5 mm were prepared ( Figure S5c). Samples were photo-NICHOLS ET AL. Note. The first column lists the sample name used in this manuscript, the second column lists the pallasite studied, the third column lists the origin of the sample, the fourth column lists how the sample was prepared, and the fifth column lists the experimental methods used. graphed before and after cutting in order to maintain orientation. Samples were mounted on 2.5-cm-diameter quartz glass disks using a minimal quantity of cyanoacrylate cement ( Figure S5d). Quartz disks were cleaned using acetone to remove any contamination and alternating field (AF) demagnetized until they had a moment <1 × 10 −11 Am 2 .
A second piece of the Springwater sample from the Harvard Museum of Natural History (SWR) was removed using an ASC Scientific Dual Bladed Rock Saw. Samples were polished using 1 μm Buehler alumina powder to remove the thin layer of epoxy protecting the sample surface. Olivines were then extracted using a small copper hammer and chisel at the Rochester Paleomagnetism and Rock Magnetism Laboratory. The boundary between each olivine and the FeNi matrix was gently tapped until the olivine "popped" out (Figure S6a). Given the considerable time needed to prepare oriented specimens following methods of Tarduno et al. (2012), unoriented samples were prepared for analysis. Extracted olivines were subsequently soaked in 1 molar HCl acid for ∼12 h to remove any surface contamination ( Figure S6b). The NRM of olivine samples dropped by up to 2 orders of magnitude after acid treatment due to removal of alteration at the olivine-metal contact. Once the olivines had been soaked in acid, they were washed in deionized water and examined using a reflected light microscope. Any further obvious surface contamination was removed using a hand-held Dremmel drill. Samples were then mounted in a 1-mm 3 -fused quartz box, the magnetic moment of which was measured prior to mounting ( Figure S6c). Where possible, samples were wedged in the box with no adhesion, but otherwise either a small amount of demagnetized OMEGA® cement or sodium silicate was used to fix the sample in place.

Alternating Field Demagnetization
Experiments were carried out on subsamples of Springwater (SW and SWH) and Imilac (IM) at the MIT Paleomagnetic Laboratory using a 2G Enterprises superconducting rock magnetometer. The magnetometer is shielded within a room made of permalloy and the ambient field is <200 nT for DC fields and <40 nT for AC fields. NRMs were removed by three axis AF demagnetization in steps of 0.5 mT up to 25 mT, followed by steps of 1 mT up to 95 mT and then steps of 1.5 mT up to 145 mT. The magnetic moment was measured after each AF step and the three orthogonal measurements then averaged to correct for any gyroscopic remanent magnetization (Garrick-Bethell et al., 2009;Tikoo et al., 2012). Two samples (SW1Aa and SW3Bb) were further demagnetized up to 420 mT in steps of 7.5 mT to test for higher coercivity components.
Paleointensity estimates were made by comparing the demagnetization of an NRM to either that of an anhysteretic remanent magnetization (ARM) or an isothermal remanent magnetization (IRM). Samples were given ARMs of 50, 100, and 300 μT in an AC field of 260 mT and subsequently AF demagnetized up to 145 mT. Paleointensity estimates were calculated for high coercivity (>5 mT) components ( Figure S9a). The TRM equivalent paleointensity was calculated using where ΔNRM and ΔARM are the change in NRM and ARM between demagnetization steps. B lab is the DC bias field and f is the TRM/ARM ratio which is taken to be f = 1.34 because the paleomagnetic carriers in pallasite olivines are FeNi metal (Gattacceca & Rochette, 2004;Stephenson & Collinson, 1974;Tikoo et al., 2014).
TRM equivalent paleointensities were also determined by comparing AF demagnetization of the NRM to that of a 400 mT IRM for samples SW1Aa, SW1Ab, and SWH1frag. Paleointensities were determined using where ΔNRM and ΔIRM are the change in ARM and IRM between demagnetization steps, and a ∼ 3,000 μT is a calibration constant, inversely proportional to the TRM/IRM ratio (Gattacceca & Rochette, 2004;Stephenson & Collinson, 1974). It should be noted that this calibration was measured for magnetite, although a similar trend has been observed for FeNi carriers (Fuller & Cisowski, 1989). The uncertainty in paleointensities estimated using this method may be large depending on the exact composition of FeNi carriers and grain size, as well as more fundamental questions of whether demagnetization characteristics of TRM (unblocking temperatures) are adequately represented by ARM or IRM demagnetizations (which reflect magnetic coercivities). We therefore only use calculated paleointensities to assess whether samples acquired remanence in the presence or absence of a dynamo field.

Controlled Atmosphere IZZI Thellier-Thellier Experiment
Samples of Springwater (SWH) and Imilac (IM) were heated in a ASC Scientific TS-48SC thermal demagnetization oven (internal field <20 nT) in the MIT Paleomagnetic Laboratory. Samples were heated in a controlled H 2 -CO 2 atmosphere as described by Suavet et al. (2014). Samples were held at an oxygen fugacity of IW-2 where IW is the iron-wüstite buffer (Brett & Sato, 1984;Holmes & Arculus, 1982) using H 2 -CO 2 gas mixing. We followed the in-field, zero-field, zero-field, in-field (IZZI) protocol (Tauxe & Staudigel, 2004) and heated samples in steps of 25°C-50°C from 100°C to 500°C. Samples were held at each temperature for 20 min and took ∼10-20 min to reach temperature, and <30 min to cool back to room temperature. Partial thermal remanent magnetization (pTRM) checks (Coe et al., 1978) were carried out at 150°C, 250°C, 325°C, and 400°C. For in-field and pTRM check steps, an applied laboratory field of 100 μT was used.

Laser Heated IZZI Thellier-Thellier Experiment
Samples of Springwater (SWR) were heated using a CO 2 laser in the University of Rochester Paleomagnetism and Rock Magnetism Laboratory in a three-layered shielded room with an ambient field of <200 nT. The power of the laser was calibrated to temperature using a thermocouple before measuring the samples. Each heating step took <5 min. Approximately 90 s were needed to reach temperature, and the temperature was held for 90 s. The "cooling time" before measurement was 90-120 s, but because of the very small thermal mass of the sample, cooling from high temperature occurs much more rapidly than this duration (O'Brien et al., 2020). Samples were heated and cooled in a magnetically shielded tube. The tube contained a coil so that laboratory fields could also be applied during heating. Both Thellier-Thellier experiments following the IZZI protocol and thermal demagnetization experiments in the absence of a laboratory field were conducted (Tarduno et al., 2012;Tauxe & Staudigel, 2004). All experiments were conducted using the small-bore (6.3 mm) ultra-high-sensitive three-component DC WSGI SQUID magnetometer. An applied laboratory field of 60 μT was used for in-field steps.
Samples were initially heated in three large temperature steps to 100°C, 210°C, and 290°C to minimize heating times and the chance of alteration at low temperature steps. Between 300°C and 700°C samples were heated in temperature steps of 10°C-20°C.

Asteroid Thermal Modeling
We follow a similar approach to Bryson et al. (2015) and assume that the pallasite parent body is spherically symmetric and cools via conduction alone. For the thermal evolution of the mantle, we iteratively solve where r is the radial distance from the center of the body, δr is the incremental distance which we set as 1 km, t is time since accretion, δt is the incremental time step, t r T is the temperature of the body at a given radial distance and time, and κ is the mantle thermal diffusivity. The surface temperature is fixed at 250 K, and the entire asteroid initially has a temperature of 1,600 K, which is approximately the silicate solidus. A summary of the parameters used in our calculations is given in Table 2.

Outward Growth of a Solid Inner Core
For nucleation and growth of a solid inner core, we assume the core is isothermal throughout asteroid cooling. For each time step, the entire core cools by where k m is the mantle thermal conductivity, ρ c is the core density, C p is the core heat capacity, and r c is the core radius. We allow the core radius to vary from 40% to 60% of the parent body radius.
To model the concentric outward growth of the solid inner core, we follow the model described by Nimmo (2009). The inner core spontaneously nucleates with a radius of 1 km. Its subsequent growth is given by and is the scale height of convection, α is the thermal expansivity, and G is the gravitational constant, T c is the temperature at the core-mantle boundary, f is the fraction of core that has so- is the relative slope of the solidus to the adiabat (Δ > 1 for inner core solidification). The cooling rate is taken from our thermal model for the pallasite parent body at the core-mantle boundary. When the core cools to the FeS eutectic (1,200 K) it remains isothermal until it is entirely solid. We terminate the model when the core is fully solidified, dictated by the total heat extracted across the core-mantle boundary equaling the latent heat of the core: where Q cmb is the core-mantle boundary heat flux and L c is the latent heat of core solidification.   r c T r was calculated at each time step by taking the difference between the temperature directly above and below the core-mantle boundary. The extracted heat is summed over each subsequent time step until it exceeds the latent heat of the core.
NICHOLS ET AL.

Inward Growth of a Solid Outer Core
For concentric inward growth of the core beginning at the core-mantle boundary, we follow the approach of Scheinberg et al. (2016). We assume that the core begins to solidify at the core-mantle boundary when its liquidus temperature is reached. The core is held at the liquidus temperature until a shell (r shell ) of 1 km thickness has solidified at the top of the core. The subsequent thermal evolution of the core is governed by where Q ocb is the heat flux across and r oc is the radius of the solid-outer-core liquid-inner-core boundary. The solid outer core and the mantle continue to lose heat via conduction, driving further solidification of the core. When the liquid portion of the core reaches the eutectic temperature (1,200 K), this liquid is held at the eutectic temperature throughout the remaining solidification. Solidification is complete when the cumulative heat flux exceeds the latent heat of the core:

The Influence of Sulfur Content on Core Solidification
The core begins to solidify when it reaches the FeS liquidus temperature, which we vary from 1,200 to 1,590 K corresponding to 18-31 wt% S (Ehlers, 1972). We assume that the core is entirely liquid and the mantle is solid when the asteroid differentiates, and therefore the highest liquidus temperature of the core must be below the silicate solidus temperature (1,600 K), which corresponds to 18 wt% S. It is assumed that all the sulfur remains in the liquid part of the core and the sulfur content of the liquid core increases by where S liq X is the sulfur content of the liquid part of the core, S o X is the initial sulfur content of the core, and r liq is the radius of the liquid part of the core. This causes a decrease in the liquidus temperature as core solidification progresses until the FeS eutectic is reached.

The Depth of the Pallasites Within the Parent Body
The depths of the pallasites within the parent body were calculated based on their cooling rates (Table 4) which were determined at a temperature of 800 K (Yang et al., 2010). We use the 800 K isotherm to find the depth at which each of these cooling rates is met (Bryson et al., 2015). These depths were then interpolated along the 593 K isotherm to find the time at which each pallasite formed tetrataenite, which also corresponds to the time of remanence acquisition. For the pallasite olivines measured by Tarduno et al. (2012), the depths of the Imilac and Esquel pallasites were interpolated along the 633 K contour, which corresponds to the thermal blocking temperature for taenite with 50-55 wt% Ni.

A Dynamo Driven by Inner Core Growth
For outward solidification of a core, the mechanism of dynamo generation via compositional convection is well understood. For this mode of core solidification, we calculated the temporal evolution of a dynamo for each of our thermal models. The buoyancy flux, Q b , generated by core solidification was calculated following Christensen and Aubert (2006) and Aubert et al. (2009): which can be used to calculate the flux-based Rayleigh number: where Ω is the rotational frequency of the planetesimal and d is the thickness of the molten part of the core. We assume a fairly rapid rotational frequency consistent with that observed for the asteroid (16) Psyche (Hanuš et al. 2013 and Table 2). The power per unit volume, P, of the resulting dynamo is proportional to the Rayleigh number where P = γRa Q and where f i is the fraction of inner-boundary originated buoyancy which is taken to be ∼0.5.
The strength of the magnetic field generated at the core-mantle boundary (B cmb ) is given by where c 1 is a scaling prefactor (Aubert et al., 2009), ρ is the density of the fluid part of the core, μ is magnetic permeability, and f ohm is the fraction of ohmic dissipation. Assuming a planetocentric dipolar field, the strength of the magnetic field at the depth of the pallasites is given by where r pal is the distance from the center of the parent body to the pallasite depth (constrained by our thermal model).

Maximum Magnetic Field Generation Potential
Since dynamo generation in cores undergoing inward solidification is poorly understood and an area of active research, we do not attempt to model dynamo behavior for this scenario. In order to compare inward versus outward core solidification without invoking a dynamo generation mechanism, we calculated the maximum magnetic field intensity the pallasites could have experienced following the approach of Evans et al. (2018). We consider all energy sources available to the core: gravitational energy (E G ), thermal energy (E T ), latent energy (E L ), and radioactive energy (E R ). Given the small size of the pallasite parent body and the uranium content of the pallasites (85 ppb; Crozaz et al., 1982), we find that radioactive energy is negligible. We also find that thermal and gravitational energy are negligible compared to latent energy: The maximum average intensity of the magnetic field (B max ) at the depth of the pallasites is given by where A c is the surface area of the core and Δt c is the length of time over which the dynamo is active, f dip is the dipolar fraction of the magnetic field, c is a constant of proportionality, and F is an efficiency factor (Evans et al., 2018).

Cloudy Zone Paleointensities
Histograms of XMCD intensity were calculated for the region of cloudy zone adjacent to the tetrataenite rim ( Figure 1). The mean XMCD intensity was calculated from the histograms of pixel intensity for eight regions of cloudy zone for each rotation (Table S2). Paleointensities were determined using Equations 1-3 and we found the Springwater cloudy zone records a unidirectional paleointensity of 22 ± 8 μT ( Figure 2).
We have also reassessed previous studies on the Imilac, Esquel, Brenham, and Marjalahti pallasites and updated their paleointensities and corresponding uncertainties to reflect their analysis for only a single rotation (Bryson et al., 2015;Nichols et al., 2016). The original paleointensity estimates were improved upon by Maurel et al. (2019) after the size of the islands in the cloudy zone at the point of remanence acquisition was established to have been ∼78% of their present-day diameter. Here, we estimate the upper limit on these revised paleointensities to 95% confidence (Table 3).

Alternating Field Demagnetization
We carried out NRM AF demagnetization of 3 subsamples of Imilac and 11 subsamples of Springwater (Table S6). We found that the three subsamples of Springwater taken from the AMNH sample (SW) have been overprinted by a strong IRM, most likely from a collector's hand magnet ( Figure S8) and were therefore discounted from our assessment of the NRMs.  and high coercivity (7-35 mT) components were identified. High coercivity components were forced through the origin. The high coercivity component is poorly defined for many samples, highlighted by high maximum angular deviation (MAD) values (>20°). The directions of the low and high coercivity components were plotted on a stereonet for each set of oriented subsamples from Imilac and Springwater (Figure 3). No evidence was found for a stable direction for low or high coercivity components in either sample, suggesting that no primary NRM is recoverable.
We assessed the fidelity of Springwater olivines as remanence carriers by comparing NRM demagnetization to laboratory applied ARM and IRM demagnetization (Table S5). To quantify the fidelity, we followed the approach of Tikoo et al. (2012) and Bryson et al. (2017). We consider: D′, the difference between the applied laboratory field and the retrieved paleointensity; and E, the error associated with the retrieved paleointensity compared to the applied laboratory field, where and where L is the laboratory field, I is the retrieved paleointensity, and W is the 95% confidence interval on the recovered paleointensity. D′ and E were calculated for Springwater using both the ARM method (Equation 5 and Figure S9c) and the IRM method (Equation 6 and Figure S9d). Results are presented in Table S5. We consider generous acceptance criteria for reliable paleomagnetic recorders; −50% > D′ < 100% and E < 100% (Bryson et al., 2017). Figure 4 shows results of the fidelity tests, which do not reliably fall within the acceptance criteria suggesting these pallasite olivines are not robust paleomagnetic recorders of magnetic fields <225 μT, as estimated by ARM and IRM methods.

Controlled-Atmosphere Thermal Demagnetization
We conducted controlled atmosphere IZZI Thellier-Thellier experiments on five subsamples of Springwater (SWH) and six subsamples of Imilac ( Figures S17 and S16). In all cases, we found that pTRM checks failed at low temperatures (<250°C ; Table S8). We found that following heating, nanoscale FeNi particles appeared to have grown along cracks in the olivine crystals ( Figure S12). This is consistent with the observed change in the rock magnetic properties of the samples. After heating, the coercivity of the samples had increased ( Figure S13) and the hysteresis loops indicate that magnetic carriers with ideal, single-domain characteristics had formed ( Figure S15). Quantum diamond microscopy showed that the magnetic phases are concentrated along cracks; however, given that these were imaged after applying a saturating IRM of 300 mT, we cannot quantify their contribution to the measured NRM ( Figure S14).

CO 2 Laser Thermal Demagnetization
IZZI Thellier-Thellier experiments were also conducted on two subsamples of Springwater (SWR) in air using a CO 2 laser. An origin-trending component was identified in subsample SWRTopD3 between 350°C and 500°C; however, pTRM checks failed at 305°C-320°C and therefore demagnetization was not continued to higher temperatures ( Figure S18 and Table S9).

Thermal and Dynamo Model Results for the Pallasite Parent Body
Our experimental data are used to place two important constraints on the cooling model for the pallasite parent body. First, using previously measured X-PEEM images that demonstrated that Marjalahti and Brenham recorded a null field, we assume that these meteorites captured a magnetic field record prior to core solidification (Nichols et al., 2016). We assume that there was no driving force for a dynamo following early thermal convection, and prior to core solidification, which would initiate compositional convection Elkins-Tanton et al., 2011). Second, Springwater, Imilac, and Esquel acquired paleointensities while a dynamo was active, most likely during core solidification. The model-derived timing of remanence acquisition for each pallasite is given in Table 4.
For inner core nucleation, we found the thermal evolution of parent bodies with a radius between 180 and 360 km was consistent with our experimental results (Figure 6a and Table S10). The parent body cannot have a radius <180 km because there is insufficient time for Springwater, Imilac, and Esquel to acquire remanence before the core is entirely solid. The upper limit on parent body size (360 km) is constrained by the sulfur content of the core; the liquidus temperature must be sufficiently high for core solidification to begin before Springwater acquired its remanence. The FeS liquidus exceeds 1,600 K (the silicate solidus) when S < 18 wt% (Ehlers, 1972). Since we assume both the metal and silicate components of the parent body are entirely liquid upon accretion to ensure complete differentiation sulfur contents below 18 wt% are not considered.
For outer core nucleation, all the cooling model solutions for inner core nucleation remain valid; however, core solidification extends over significantly longer time scales (Figure 5b). We found that the minimum mantle thickness was also smaller for inward core solidification compared to outward core solidification for similar sized cores (Figure 7).
For each plausible cooling model solution, we calculated the evolution of a dynamo driven by inner core solidification. The resulting dynamo model was compared to our experimental data ( Figure 6). We found that for dynamo evolution to be consistent with our measured paleointensities, relatively long core solidification times (>180 Myr) and large cores (>170 km) are required.
Using our planetesimal models for both inward and outward core nucleation, we calculated the maximum intensity that the dynamo could have generated, assuming it was continuously active between the Springwater and Esquel cloudy zones cooling through the tetrataenite ordering temperature. The maximum intensities generated for inward and outward core solidification were 80 and 72 μT, respectively. Our results suggest that a planetesimal with a large core and thin mantle has sufficient energy to generate the observed high paleomagnetic field intensities ( Figure 7).
NICHOLS ET AL. Note. The first column is the name of the studied pallasite. The second column is the cooling rate of the pallasite (Yang et al., 2010). The third column is the inferred depth of the pallasite within the parent body. The fourth column is the inferred height of the pallasite above the coremantle boundary. The fifth column is the inferred time of remanence acquisition; the cloudy zone acquires remanence upon cooling through 320°C.  Figure 5. Thermal evolution models for planetesimals with outward and inward core solidification. The green line is the 593 K contour, which is the temperature at which the cloudy zone acquires remanence. The green squares represent the timing of remanence acquisition for the Marjalahti, Brenham, Springwater, Imilac, and Esquel cloudy zones. The yellow line is the 633 K contour, which is the temperature at which the pallasite olivines acquire remanence (Tarduno et al., 2012). The yellow squares are for the Imilac and Esquel olivines. The gray region shows the solid part of the core. (a) Thermal model for outward core solidification within a planetesimal of radius 220 km with a core radius of 100 km. The purple line is the 1,530 K contour, which is the liquidus of FeS with a composition of 23 wt% S. (b) Thermal model for inward core solidification within a planetesimal of radius 300 km with a core radius of 250 km. The blue line is the 1,200 K contour, which is the liquidus of FeS with a composition of 31 wt% S. 4. Discussion

Pallasites as Paleomagnetic Recorders
In recent years, significant progress has been made to understand how remanence is acquired by the cloudy zone and to improve methods for retrieving a reliable, quantified paleointensity. The first studies on the paleomagnetism of the cloudy zone in pallasites (Bryson et al., 2015;Nichols et al., 2016) assumed that the cloudy zone could record time-resolved information, with the oldest record in the coarsest cloudy zone immediately adjacent to the tetrataenite rim, and the youngest record in the fine cloudy zone, furthest from the tetrataenite rim. Einsle et al. (2018) subsequently showed that a significant portion of the cloudy zone acquires remanence simultaneously upon cooling through 320°C, when the islands order to tetrataenite. The diameter of the islands at the time of tetrataenite ordering was also found to be much larger (92-122 nm; Maurel et al., 2019) than previously assumed (35-45 nm; Bryson et al., 2015;Nichols et al., 2016), lowering the paleointensity estimates for Marjalahti, Brenham, Imilac, and Esquel (Table 3; (Table S10). Black squares and vertical error bars are the measured paleointensities of the pallasite cloudy zones using X-PEEM. The relative timings of remanence acquisition are calculated from the cooling model. The red line shows the evolution of a dynamo driven by core solidification. The black dashed line shows the minimum paleointensity that can be attributed to an active dynamo. Model 2 is too weak to explain the measured paleointensities. The shaded regions represent experimental constraints on when there was no dynamo and the core was entirely liquid (blue region) and when there was an active dynamo and the core was solidifying (red region). (d) Model 24 does not sustain a dynamo for long enough to explain the measured paleointensities. (e) Model 31 matches the measured paleointensities. (f) Model 56 generates a dynamo which is too short and intense to explain the measured paleointensities. X-PEEM, X-ray photoemission electron microscopy.

Figure 7.
The effect of mantle thickness on dynamo duration and mean dynamo intensity. Red points are for our thermal models for outward core solidification. Blue points are for our thermal models for inward core solidification. The size of the point represents the size of the planetesimal core. (a) Minimum dynamo duration (the time between Springwater and Esquel acquiring remanence) as a function of mantle thickness. For outward solidification, there is little dependence on mantle thickness or core size. For inward core solidification, the dynamo duration depends strongly on mantle thickness. (b) The yellow region represents the paleointensity results reported by Tarduno et al. (2012) for Imilac and Esquel. The green region represents the paleointensities reported here from X-PEEM studies of Springwater, Imilac, and Esquel. The yellow-green region shows where independent paleointensity estimates overlap.
To generate paleointensities consistent with both sets of observations, the pallasite parent body must have had a large core and thin mantle, regardless of the direction of core solidification. X-PEEM, X-ray photoemission electron microscopy.
The experimental method using X-PEEM to image the distribution of magnetization in the cloudy zone has also been improved. Previous studies imaged the distribution of magnetization in the cloudy zone in a single orientation (Bryson et al., 2015;Nichols et al., 2016). This does not allow the 3D vector of magnetization to be fully resolved. Therefore, only a minimum paleointensity estimate and minimal (constrained to within a hemisphere) directional information can be recovered. We imaged Springwater in three rotations, adopting the improved method outlined by . Therefore, the paleointensity and direction (relative to the sample) of magnetization in the Springwater pallasite have been fully resolved. We have also improved upon the paleointensity estimates for Marjalahti, Brenham, Imilac, and Esquel calculated by Maurel et al. (2019) by accounting for the fact these paleointensity estimates were only lower limits. We have calculated upper bounds on the true paleointensity and found that our results are consistent with the paleointensities reported for Imilac and Esquel olivines (Tarduno et al., 2012).
Given the developments in the understanding of cloudy zone paleomagnetism, we are now able to present the most detailed time-resolved record of paleointensities on the MG pallasite parent body. Further work is still required to understand the role of interactions between cloudy zone islands. However, since we have assumed no interaction, accounting for magnetostatic interactions between islands will only increase our paleointensity estimates (Dunlop & Ozdemir, 1997). Despite the significant uncertainty associated with some of the paleointensities, an increase in paleointensity over time-most likely triggered by the onset of dynamo activity-can be resolved and the implications of this record are considered in the following sections.
Single crystal thermal paleointensity methods applied to pallasite olivines are challenging because of the significant time needed in sample selection, preparation, and experiments, as well as the potential for thermal alteration. Using techniques previously developed at the University of Rochester (Tarduno et al., 2006), the time for conducting single crystal paleointensity studies typically spans multiple years, rather than the 2 weeks in which the laser-heated experiments discussed here were conducted at the University of Rochester. Additional experiments using AF and controlled-atmosphere heating techniques were conducted at MIT over 3 months. Our AF experiments demonstrated the poor recording fidelity of pallasite olivines, which cannot reliably record paleointensities <225 μT using ARM and IRM methods. Controlled-atmosphere experiments were also unable to prevent olivine alteration. In contrast, some olivines can be cycled through multiple TRMs and show properties meeting Thellier's criteria (Tarduno et al., 2012). Springwater olivines are 18% fayalitic (Righter et al., 1990) and Imilac olivines are 12.3% fayalitic (Wasson & Choi, 2003) suggesting that they have an equivalent oxygen fugacity at high temperatures (Righter et al., 1990). However, experimental evidence suggests that pallasite fugacity can vary significantly (Brett & Sato, 1984;Holmes & Arculus, 1982). Given the rapid alteration of olivines at low temperatures (150°C), it appears that our samples were not in equilibrium with the controlled atmosphere. The growth of nanoscale FeNi particles may be indicative that the atmosphere (IW-2) was too reducing.
Previous studies (Tarduno & Cottrell, 2013;Tarduno et al., 2012) reported successful thermal demagnetization of olivines from the Imilac, Esquel, and Springwater pallasites. The method used for these studies involved heating samples with a CO 2 laser in air with short heating and cooling times of a few minutes. The null success rate for our controlled-atmosphere thermal experiments is possibly attributed to the low number of samples measured (7 for Springwater [SWH] and 6 for Imilac). Similar unsuccessful results were reported by Tarduno et al. (2012) and further indicate that only some millimeter-sized pallasite olivines have inclusions from which paleointensities can be retrieved by heating. We suggest that at least twice as many specimens should be measured as studied here to obtain paleointensity values. This is supported by our identification of a stable remanence component in only one of our Springwater specimens, SWRTopD3 ( Figure S18), which suggests preservation of magnetic minerals recording a primary magnetization. The experimental challenges encountered when measuring pallasite olivines highlight the benefits of studying the cloudy zone. Our X-PEEM technique does not require sample heating and therefore offers an effective approach for recovering paleointensities without risk of sample alteration. However, the availability of results from pallasite olivines is important in our following discussion because results obeying Thellier laws are independent of those acquired using X-PEEM and the fact that both yield results indicating a past magnetization gives us confidence in our further consideration of the interior dynamics and structure of the pallasite parent body.

Time-Resolved Records of Magnetic Activity on the Main Group Pallasite Parent Body
We present a comprehensive time-resolved paleomagnetic record for the MG pallasite parent body. We have used these paleomagnetic observations to constrain the thermal evolution of the pallasite parent body following a similar approach to Tarduno et al. (2012) and Bryson et al. (2015). Assuming that the dynamo was driven by core solidification, Brenham and Marjalahti acquired a null remanence (or experienced magnetic field intensities too weak to attribute to an active dynamo) prior to the onset of solidification, while Springwater, Imilac, and Esquel acquired remanence while the core was still partially liquid consistent with predictions for asteroid dynamo behavior . Previous studies only considered eutectic solidification of FeS at 1,200 K, corresponding to a sulfur content of 31 wt% Bryson et al., 2015;Tarduno et al., 2012). For solidification at the eutectic, we found that for parent bodies that were sufficiently small for Springwater to acquire remanence after the onset of core solidification, Esquel could not have acquired remanence before core solidification was complete. It should also be noted that eutectic solidification of the core will in fact not drive a thermochemical dynamo, although it is a reasonable approximation for considering the thermal evolution of a core. We therefore consider lower sulfur compositions in our planetary cooling model that correspond to higher FeS liquidus temperatures. The raised liquidus temperature causes core solidification to begin at earlier times and allows a thicker mantle, resulting in slower cooling and longer solidification times. By varying the sulfur content and core size, we found a range of plausible parent body sizes from 180 to 360 km radius, a larger upper limit than that reported by Bryson et al. (2015). We found that permittable parent body sizes increase with decreasing sulfur content and increasing core size ( Figure 6).

Implications for Core Crystallization and Dynamo Generation
We have considered both outward and inward core solidification for the pallasite parent body. In both cases, we assume concentric growth and do not account for the possibility of dendritic growth or other more exotic styles of crystallization (Scheinberg et al., 2016). Our models can therefore be considered as lower limits on the time of core solidification. This does not affect the implications of our results, which primarily depend on the core beginning to solidify between the times at which Brenham and Springwater acquired remanences, and being sufficiently molten to drive an active dynamo when Esquel acquired its remanence. We found that both inward and outward core growth are consistent with our X-PEEM results, and therefore both solidification styles are considered in the context of dynamo generation.
We present a simple model of dynamo generation via inner core nucleation where the onset of core solidification is accompanied by a rapid increase in magnetic field intensity ( Figure 6). There is a gradual decrease in dynamo strength as the liquid, convecting outer shell reduces in volume throughout solidification. This mechanism of core solidification can sustain a long-lived dynamo, supported by current observations of Earth's magnetic field, and modeling of long-lived lunar dynamo activity (Laneuville et al., 2014). If the dynamo was only active for some fraction of the total core solidification time, then the maximum possible magnetic field intensities will be higher; however, our measured paleointensity for the Springwater pallasites (22 ± 8 μT), which corresponds to the maximum field strength shortly after the onset of a compositional dynamo, suggests that the mean intensity must be relatively weak (<20 μT). This is consistent with the long lifetime of the dynamo (>117 ± 4 Myr), which corresponds to the time between Springwater and Esquel acquiring remanence.
We found that a dynamo driven by inner core nucleation is difficult to reconcile with our upper paleointensity limits for Imilac and Esquel and the results reported by Tarduno et al. (2012). As the inner core crystallizes, the dynamo loses power, at odds with the apparent increase in paleointensity recorded by Springwater, Imilac, and Esquel, respectively. This discrepancy cannot be explained by the greater depth of each pallasite within the parent body (Table 4), which only accounts for a small increase in intensity. Additionally, the total energy in the core is insufficient to explain the paleointensities recorded by Imilac and Esquel at their modeled depths within the parent body, even for the shortest possible dynamo duration (Figure 7).
Given the small planetesimal sizes derived from our cooling models, the core is at low pressure, which suggests that core solidification is in fact more likely to be initiated at the core-mantle boundary (Hauck et al., 2006;Williams, 2009). We therefore also consider inward core solidification but do not attempt to model the corresponding dynamo evolution since this is a relatively unexplored phenomena, particularly for a mantled planetesimal. However, model results have demonstrated that inward core solidification can drive a dynamo Scheinberg et al., 2016) and that some asteroid cores solidified in such a manner (Yang et al., 2007). We find that both inward and outward core solidification of a planetesimal with a large core and thin mantle can explain both the timings of pallasite remanence acquisition relative to core solidification, and the high paleointensities and dynamo lifetime required by our recovered paleomagnetic record (Figure 7). We therefore conclude that a large core (>200 km radius) and small mantle thickness (<70 km) are more important for generating high paleointensities than inward versus outward core solidification.

Implications for the Origin of Pallasites
A recent study suggested that the pallasites may be the result of the propagation of metallic dykes from the core-mantle boundary into a shallow overlying mantle (Johnson et al., 2019). This study predicts that for high sulfur compositions (∼31 wt%), metallic dykes could penetrate tens of kilometers into the mantle. This is consistent with our model results for inward core solidification (Table 4) which suggest that the pallasite parent body had a large metallic core with a high sulfur content and the pallasites could have originated from short distances (∼30 km) above the core-mantle boundary. This also demonstrates that the pallasites could form close to the core-mantle boundary while still originating in the mid-mantle of their parent planetesimal as suggested by previous studies (Bryson et al., 2015;Tarduno et al., 2012;Yang et al., 2010).
Pallasites have also been suggested to have formed throughout the parent body mantle, with the metal representing residual melt left over from incomplete differentiation (Boesenberg et al., 2012;Mckibbin et al., 2019). Textural and geochemical evidence suggests that in addition to residual metallic melt in the mantle, differentiated metal from an impactor must also play a role in pallasite formation in order to explain the range of observed textures (Tarduno et al., 2012;Walte et al., 2020). To reconcile these parent body configurations, where slow-cooled pallasites form close to the core-mantle boundary but also acquire a paleomagnetic record of a convecting, liquid core requires inward core solidification (Mckibbin et al., 2019), consistent with our results. We also note however, that if the pallasite parent body only underwent partial differentiation, the metallic core will be relatively small, at odds with the energy required to drive a sustained, high-intensity dynamo (Figure 7). This may suggest that the parent body must be relatively large (radius >300 km) in order to partially differentiate and still form a core large enough to explain the observed paleointensities.
It has also been hypothesized that the pallasites may form within a planetesimal similar to the metallic asteroid (16) Psyche (Elkins-Tanton et al., 2020;Johnson et al., 2019;Walte et al., 2020). Current estimates suggest that Psyche has a density of 4,160 ± 640 kg m −3 (Drummond et al., 2018). This is comparable to the density for many of our planetesimal models, assuming the core has a density of 7,000 kg m −3 and the mantle has a density of 3,000 kg m −3 (Figure 8). For our cooling models where intensities are consistent with our X-PEEM observations, the model planetesimals have a similar density to Psyche. To explain both our X-PEEM observations and the paleointensities recovered from olivine inclusions (Tarduno et al., 2012) require a large-cored and thinly mantled planetesimal with a density slightly higher than current predictions for Psyche. Figure 8. The magnetic field intensity versus planetesimal density predicted for each of our cooling models assuming an average density of 7,000 kg m −3 for the core and 3,000 kg m −3 for the mantle. Red points are for our thermal models for outward core solidification. Blue points are for our thermal models for inward core solidification. The size of the point represents the size of the planetesimal core. The yellow region represents the paleointensity results reported by Tarduno et al. (2012) for Imilac and Esquel. The green region represents the paleointensities reported here from X-PEEM studies of Springwater, Imilac, and Esquel. The yellow-green region shows where independent paleointensity estimates overlap. The gray region represents the predicted density of asteroid (16) Psyche (Drummond et al., 2018). X-PEEM, X-ray photoemission electron microscopy.

Conclusions
We have presented new paleomagnetic data for the Springwater pallasite which "fills the gap" between Brenham and Marjalahti acquiring remanence prior to core solidification (Nichols et al., 2016), and Imilac and Esquel acquiring remanence after core solidification had begun and a compositional dynamo was active (Bryson et al., 2015;Tarduno et al., 2012). We have reassessed all X-PEEM data for the pallasites in light of advances in the technique and our understanding of remanence acquisition in the cloudy zone Einsle et al., 2018;Maurel et al., 2019). In addition, we conducted thermal and AF demagnetization experiments on olivines from the Springwater and Imilac pallasites. We found that AF techniques were not well suited for isolating directions or intensities in the samples analyzed. The extensive time (years) required to extract paleointensities from pallasite olivines, and issues with thermal alteration during controlled-atmosphere experiments, highlights the benefit of conducting paleomagnetic analyses on the cloudy zone rather than silicate inclusions in extraterrestrial samples with variable oxygen fugacities. However, previous paleointensity estimates from pallasite olivines (Tarduno et al., 2012) are critical for corroborating our cloudy zone paleointensities; the fact that both yield similar results indicating a past magnetization gives us confidence in our further consideration of the interior dynamics and structure of the pallasite parent body.
We present the most complete time-resolved paleomagnetic record for a planetesimal to date. Our results constrain the onset of core solidification and give a minimum time for core solidification, allowing the size and thermal evolution of the MG pallasite parent body to be well determined. We found that our X-PEEM results are consistent with both inward and outward core solidification, and the temporal trend in paleointensities is consistent with a dynamo driven by inner core solidification. However, to explain the upper limit on our paleointensity determinations for Imilac and Esquel and the strong paleointensities recovered by Tarduno et al. (2012), we show that a planetesimal with a large core and thin mantle is required. Strong paleointensities (∼65-95 μT) in the middle of a thin mantle can be generated by either inward or outward core crystallization. A dynamo driven by inward crystallization has been proposed for the asteroid (16) Psyche Scheinberg et al., 2016). We show that the pallasites could form relatively close to the core-mantle boundary, while still forming in the middle of a thin mantle as predicted by previous studies (Bryson et al., 2015;Tarduno et al., 2012;Yang et al., 2010). We conclude that the pallasites could form on a planetesimal like asteroid (16) Psyche and could form by the intrusion of sulfur-rich metallic dykes through the mantle from the underlying core (Johnson et al., 2019).
Future studies should focus on directly determining the thermochronology of the pallasites. Our thermal models for outward versus inward core solidification provide contrasting predictions for the timing of remanence acquisition and the size of the parent body (Figure 7). Combined thermochronologic and paleomagnetic studies on suites of meteorites from the same parent body can provide a wealth of information regarding the size, internal structure, and dynamics of planetesimals.