Analysis of layered boulders on asteroid (101955) Bennu and their implications for fluid flow on the parent body

The observation of carbonate veins on asteroid Bennu supports the idea that large‐scale water flow may have occurred in carbonaceous asteroids in the early solar system. We identified and analyzed 11 boulders with layered structures on asteroid Bennu's surface using high‐resolution (centimeter‐scale) image and altimetry data obtained by the OSIRIS‐REx mission. The boulders' linear layer boundaries and parallel bedding follow the principle of original horizontality and suggest that they formed from sediment deposition by fluid flow on Bennu's parent body. We developed a simple model of the parent body (100‐km diameter with the density of CM chondrite material) and found that the water flow velocity had to be at least 21.1 cm s−1 to transport the largest clast observed embedded in a layered rock, which is 85 cm in average length. The flow velocity could have been as high as 26.5 cm s−1 if a larger clast observed on top of a layered rock was once embedded therein. Our results strongly support open‐system aqueous alteration on carbonaceous chondrite parent bodies.


INTRODUCTION
It is important to understand how water behaved in carbonaceous asteroids in the early solar system because they could have been the source of water on Earth.However, whether the water was flowing or stagnant in hydrated asteroids has been under debate.For example, the static water scenario is supported by variability in meteoritic carbonate compositions (Dufresne & Anders, 1962;Johnson & Prinz, 1993).Unfractionated bulk chemical compositions and oxygen isotopic compositions of carbonaceous chondrites have been interpreted to reflect alteration in a closed system at the scale of meteorite samples (Clayton & Mayeda, 1999;Rubin et al., 2007).A numerical model that considered low permeability of chondritic materials predicted very limited water flow (hundreds of micrometers at most) (Bland et al., 2009).Meanwhile, other numerical studies predict open-system, large-scale fluid flow (tens of kilometers; Bland & Travis, 2017;Grimm & McSween, 1989;Mcsween et al., 2002;Palguta et al., 2010;Travis & Schubert, 2005;Young, 2001;Young et al., 1999Young et al., , 2003)).Petrography, mineralogy, and chemistry of carbonates in CI chondrites suggest that liquid water flowed over meter scales on the parent asteroid through fractures, forming carbonate veins (Endreß & Bischoff, 1996).
Carbonate veins have been discovered on the near-Earth carbonaceous asteroid (101955) Bennu by NASA's Origins, Spectral Interpretation, Resource Identification, and Security-Regolith Explorer (OSIRIS-REx) mission (Kaplan et al., 2020).Given Bennu's small diameter of $500 m (Lauretta et al., 2019), if it had formed in the main asteroid belt 4.56 Gyr ago, it would have been destroyed by collisions within 0.1-1 Gyr (Bottke et al., 2005).Therefore, Bennu is most likely a fragment of a larger parent asteroid, estimated to be about 100 km in diameter (Walsh et al., 2013), that experienced a catastrophic disruption.The presence of veins up to 150 cm long in its boulders suggests fluid flow and hydrothermal deposition over kilometer scales for thousands to millions of years on the parent asteroid (Kaplan et al., 2020).This significant piece of evidence observed in situ on an asteroid supports an open-system alteration scenario early in solar system history.
Here, we describe layered boulders on Bennu that also support the large-scale fluid flow scenario.In this work, we define a boulder as a rock that is larger than 21 cm in diameter, per the OSIRIS-REx mission (Burke et al., 2021).We use "unit" to describe material with a uniform texture and a characteristic albedo."Layer" is used to describe a unit which has measurable thickness and a distinct linear boundary.We define a layered boulder as a boulder that has at least two distinct units, defined by a linear boundary.Some layered boulders include thin beddings in their units, that is, beddings with thickness of 5-60 cm (Mckee & Weir, 1953).Clasts are rocky fragments and components of units.
Figure 1 shows Gargoyle Saxum, an example of a layered boulder.The top unit and bottom unit are layers, defined by a linear boundary marked by the arrow.Thin beddings can be seen in the circle.A comparatively bright rock nicknamed Jewel is apparent on the surface on the top unit; it may be a unit of Gargoyle (though not a layer), or a clast originating from the top unit, or an unrelated rock that landed on Gargoyle after being launched by impact.
We analyzed the units of layered boulders using image and altimetry data at centimeter scales and evaluated multiple formation hypotheses.We propose that the layered structures seen in Bennu's boulders formed via sediment deposition in hydrothermal convective flow on the parent asteroid.Based on the observed clast size, we provide a simplified model of particle settling and estimation of the flow velocity in the convection.
Bennu was explored by the OSIRIS-REx spacecraft from December 2018 to May 2021 (Lauretta et al., 2021).Spectral data suggest that Bennu's composition corresponds to aqueously altered carbonaceous chondrites (Hamilton et al., 2019(Hamilton et al., , 2021).Bennu's material most likely went through aqueous alteration on the parent asteroid (Hamilton et al., 2019;Kaplan et al., 2020).Bennu spent most of its lifetime in the main belt and has been in near-Earth space for 1.75 AE 0.75 Myr (Ballouz et al., 2020;Walsh et al., 2019).

Image Data
Boulders with layered structure were identified by manually examining a global mosaic of Bennu (Bennett et al., 2021) that was projected using the Small Body Mapping Tool (Ernst et al., 2018) onto a lidar-based global shape model (version "OLA v16") (Daly et al., 2017(Daly et al., , 2019(Daly et al., , 2020)).The search was mainly focused on the area between AE70°latitude to ensure preferable illumination conditions for normal albedo calculations.Once a layered boulder was identified on the global mosaic, corresponding images acquired by the OSIRIS-REx Camera Suite (OCAMS) PolyCam imager (Golish et al., 2020;Rizk et al., 2018Rizk et al., , 2019) ) were located that highlighted the textures.The PolyCam images used in this work are listed in Table 1.Boulder names approved by the International Astronomical Union are used when applicable; otherwise, we use an internal alphanumeric designation to identify boulders.

Altimetry Data
Boulders were examined further using digital terrain models (DTMs) based on lidar data.Global coverage of the surface was obtained by the OSIRIS-REx Laser Altimeter (OLA; Daly et al., 2017), a scanning lidar that measures distance to the asteroid's surface to produce 3-D topography data.OLA data have average spot spacings of less than 5 cm and vertical precision of AE1.25 cm (Daly et al., 2020).All the OLA scans available for each boulder (5 scans or more) were registered to ensure that the DTM covered the entire surface of the boulder.The point cloud was meshed into a surface using the Poisson reconstruction meshing technique (Kazhdan & Hoppe, 2013) to create a DTM (Figure 2a).The DTMs were used to estimate the longest dimensions of the boulders using the measuring tool on the Meshlab software package (Cignoni et al., 2008) by selecting two farthest points within each boulder.

Registration, Photometric Correction, and Albedo Measurement
The PolyCam images were registered to the corresponding DTM using reconstructed SPICE kernels (Acton, 1996;Acton et al., 2018) on the USGS's Integrated Software for Imagers and Spectrometers version 3 (ISIS3).Backplanes of images contain the incidence angle, emission angle, phase angle, oblique pixel scale, and body-fixed geographic coordinates.These values were calculated per pixel using ray-tracing methods (DellaGiustina et al., 2018).After the registration, the images were photometrically corrected using the RObotic Lunar Observatory (ROLO) photometric function (Golish et al., 2021), setting incidence angle, emission angle, and phase angle to 0°( Figure 2b).The resulting pixel values represent the normal albedo.
Normal albedo is a measure of reflectance of a surface when it is illuminated and observed from nadir.
Here, we use it to distinguish the lithology in each layer.Differences in normal albedo can arise from mineralogy, particle size, and surface roughness.Using the region tool  Fluid flow on Bennu's parent body on SAO DS9 (Joye & Mandel, 2003), representative areas within each unit were selected to calculate the average normal albedo and standard deviation.The median value of the representative area is 312 pixels; the smallest number of pixel analyzed was 20, from LR7's bottom unit, and the largest number of pixels analyzed was 2261, from Simurgh Saxum's middle unit.

Clast Size and Layer Thickness Measurements
Large clasts on the surface contribute to a rough appearance, and a lack of visible clasts leads to a smooth surface.Clasts in the units were manually identified and labeled using the ArcMap geographic information system (GIS) tool (ESRI, 2011).The analysis focused on the largest 10 clasts per unit in three of the boulders-Gargoyle Saxum, LR1, and Simurgh Saxum-because they constrain the fastest speed of water flow required to transport the clasts.Jewel was included in the analysis of the clasts in Gargoyle's top unit because it may have originated therefrom (discussed in Results, Gargoyle Saxum Sections).Smaller clasts analyzed in the LR1 west unit are close to the pixel scale of the image used (5.9 cm per pixel; Table 1).Therefore, only the largest five clasts, which are sufficiently larger than the resolution limit, were analyzed in this unit.The largest clasts were identified by their longest lengths in pixels.
For each clast, lengths at eight angles were calculated, starting with the longest axis; the other seven lengths were determined by rotating the line of the longest axis by 22.5°(Figure 3a).Lengths were calculated using the Cartesian distance formula with the endpoints determined from the backplanes of (x, y, z) coordinates included in a multilayer image cube file.The average and SD were calculated from the eight lengths for each clast.
For layer thickness measurements, we used images that show the thickest part of the layers.Layer thickness was determined by the distance between two straight lines over the boundaries that define the layer (Figure 3b).When the two lines were parallel to each other, the thickness was the perpendicular distance between them.When the two boundary lines were not parallel to each other, the longer length of the two perpendicular lines to each boundary line was chosen at the thickest end of the layer.When the layer is defined by one boundary and the opposite side is the outline of the boulder, a parallel line to the boundary was drawn to the farthest edge of the boulder.The thickness of the layer is the distance between the parallel lines.Thickness of each thin bedding was determined by the same method as layer thickness, and the average thickness was calculated for each set of thin beddings.The lengths of the lines that represent the thickness were calculated in ArcMap using the same method as the clast size measurements described above.

Settling Velocity (Minimum Flow Velocity) Estimation
To constrain the flow velocity, the settling velocity was calculated for varying particle sizes in microgravity.Stokes' equation is only applicable to laminar flow, which is not the case for this study because the observed clasts are 10-100 cm in size and would cause turbulent flow.Therefore, in this model, Rubey's equation was used (Rubey, 1933).This is a general equation for settling velocities of all grain sizes.In addition to Stokes' law, it also considers the impact of the rising water that the sinking particle experiences.This provides settling velocities for larger particles that do not follow Stokes's law.
where v (cm s À1 ) is settling velocity, g (cm s À2 ) is gravitational acceleration, ρ F (g cm À3 ) is density of the fluid, ρ P (g cm À3 ) is density of particle, r (cm) is radius of the particle, and η (g cm s À1 ) is dynamic viscosity of the fluid.The density of water is ρ F ¼ 1 (g cm À3 ).We assumed the parent body to be 100 km in diameter (Walsh et al., 2013) and composed of CM chondrite material (Praet et al., 2021).The average density of CM chondrite is ρ P = 2.92 (g cm À3 ; Macke et al., 2011).
Using the assumed diameter and the average density, the gravitational acceleration on the surface of the parent body was calculated to be g = 4.08 cm s À2 .The gravitational acceleration within the asteroid linearly depends on the radial distance from the center.It is weakest near the center and strongest on the surface.This leads to radial dependence of settling velocity.Settling velocity of a particle increases as the radius increases (Figure 4 and Discussion section).The settling velocities of the largest clasts in each unit reported in this paper are maximum possible values because the gravitational acceleration at the surface was used to calculate them.If the units formed deeper in the interior, the settling velocities would have been slower (Figure 4).
For the dynamic viscosity of water, η = 4.0 × 10 À3 (g cm s À1 ) was used, which is the value at 70°C (Huber et al., 2009;Kestin et al., 1978).This is the highest temperature of aqueous alteration in CM chondrites as estimated using the carbonate clumped isotope thermometer (Guo & Eiler, 2007;Suttle et al., 2021).Parameters in this equation that depend on temperature are density of water, density of particle, and dynamic viscosity of water.The temperature dependence of the densities is negligible.The dynamic viscosity parameter comes from Stokes' viscous resistance, which does not affect the settling velocity of larger particles (>1 cm) that are focus of this work.Therefore, temperature of water does not meaningfully affect our flow velocity estimation or our discussion.
If the flow velocity is slower than settling velocity of a particle in water flow, the particle will sink and deposit.If the flow velocity is faster than the settling velocity, the particle will remain in suspension.Therefore, settling velocity is the minimum required flow velocity to transport particles with corresponding size.We note that this equation considers a spherical particle and does not consider collision with other particles.Irregularly shaped particles have lower settling velocity than spherical particles (Dietrich, 1982).Collisions among particles would slow down their settling velocity.Our calculated velocity here is therefore the fastest estimation.

RESULTS
We identified 11 boulders with layered structures of apparently distinct lithologic units on the global surface of Bennu (Figures 1, 5-7).They are a subset of the total number of layered boulders on Bennu, and they constitute a sufficient number for statistics.Five boulders are composed of more than two layers.Thin beddings are observed in Gargoyle Saxum, Roc Saxum, and LR10 (explained in detail below).The boulders range from about 4 m to about 95 m in longest dimension.Layer thickness is typically a few meters, and some are thicker than 10 m.The thickest layer observed is Simurgh Saxum's east unit, which is 19.8 m (Table 2).
The normal albedo of each unit is presented in Table 2. Difference in normal albedo and surface texture -or the size, shape, appearance, arrangements, and sorting of the grains in rocks (Serra, 1986)-makes units appear distinct within a boulder.Although some units cannot be distinguished by normal albedo values when the SDs are considered, most adjacent units have distinct values from each other.The texture of the units varies.Generally, brighter units have smoother surfaces than darker units in the observed layered boulders.There are multiple units that are dark and rough (Gargoyle top, LR7 top, LR8 top, and LR9 west and east units).Surface roughness is defined as the SD of the heights of data points from a model surface (Pollyea & Fairley, 2011).The dark and rough units have higher values of standard deviation in their normal albedo because of the rough surface.
Three boulders-Gargoyle Saxum, LR1, and Simurgh Saxum-were selected for detailed analyses because their unit boundaries are clearly seen in images   3. Here, we provide detailed description of them, as well as Roc Saxum, the largest boulder on Bennu.

Gargoyle Saxum
Gargoyle Saxum is about 15 m tall and 17 m wide, located close to the equator (Table 1 and Figure 5).It consists of at least two units-the top and bottom-and potentially Jewel, the bright rock on the surface of the top unit (Figures 1 and 6).There might be an additional unit under the bottom unit (Figure 6a), but we did not focus on it as it is difficult to clearly see.The top and bottom units are labeled based on their current orientation.They are separated by a linear boundary, which make them layers.
There is a depression on the top unit, right above Jewel (Figure 6b).The depression's shape and size are very similar to Jewel's.Currently, Jewel seems to be sitting on top of the overhang of the top unit, but in the past it could have been embedded where the depression exists.Therefore, we tentatively treat Jewel as one of the clasts of the top unit.
The top unit is about 11.6 m thick, and its normal albedo is 3.7 AE 0.8% (Table 2).This mean value is darker than the global average albedo of 4.4 AE 0.2% (DellaGiustina et al., 2019) and represents the darkest material on Bennu (DellaGiustina et al., 2020), whereas Jewel is comparatively bright with a normal albedo of 8.7 AE 0.8%.The top unit has numerous visible clasts.The average lengths of the largest clasts in the top unit of Gargoyle Saxum range from 32 to 85 cm excluding Jewel, or up to 134 cm including Jewel.The settling velocity of the 85-cm clast is 21.1 cm s À1 , and the settling velocity of Jewel is 26.5 cm s À1 (Figure 8 and Table 3).At the boundary with the bottom unit, there are thin beddings that are 48.9 cm thick on average (Figure 1).
The bottom unit is about 4.4 m thick and has a brighter and smoother surface.Its normal albedo is 5.6 AE 0.4%, which is distinct from that of the top unit and Jewel.The thin beddings at the boundary are also present in this unit (Figure 1).The bottom unit has only a few resolvable clasts in the highest resolution images.The 10 largest range from 10 to 19.8 cm.The settling velocity of the largest clast is 10.2 cm s À1 .
In PolyCam images and OLA data, Gargoyle Saxum appears to be tipping over toward the south, with its bottom above the ground on the north end (Figure 6d,f).There are smaller (a few meters diameter) boulders around Gargoyle Saxum that have a rough texture similar to Gargoyle's top unit.These observations led us to explore whether Gargoyle Saxum could have been moving toward the south, producing fragments in the process.
The outline of Gargoyle Saxum that meets the ground is very rough, and the bottom unit there abruptly ends (where the thin beddings exist, Figure 1).It seems as though a large chunk of rock was removed and caved the material of both the top and bottom units.This caved feature is about 9 m wide.One of the smaller nearby boulders that have similar textures, LR7, is almost 9 m wide and is located only 4 m south of Gargoyle Saxum.The normal albedo values of the LR7 top unit and bottom unit are indistinguishable from those of the Gargoyle Saxum top and bottom units, respectively, when the standard deviations are considered (Table 2).Therefore, it is possible that LR7 was a part of Gargoyle Saxum.The fact that Gargoyle Saxum is tipping over southward suggests that it might have migrated downslope from higher latitudes to the equator.This direction of migration is consistent with the global mass movement trend observed on Bennu (Jawin et al., 2020).It is unclear whether Gargoyle Saxum was rolling or sliding down.During the movement, LR7 could have broken apart from Gargoyle Saxum and continued moving south, where it eventually stopped.Disruption of the boulder by its movement, such as LR7 cleaving off or other fragment production, could have caused Jewel to pop out from the depression that we observe in the top unit.

LR1
LR1 is located in the midlatitudes of the northern hemisphere.It has two units: west and east (Figure 7a).The west unit is about 4.2 m thick and appears smoother and brighter, with a normal albedo of 6.9 AE 1.1%.Clasts in this unit are overall very small and close to the resolution limit.Therefore, we measured the five largest clasts in this unit, instead of 10, which range from 20 to 33 cm in average length.The settling velocity of the largest clast is 13.1 cm s À1 .There is a linear fracture in the west unit, which could have resulted from thermal fatigue (Molaro et al., 2020).The east unit contains resolved clasts and appears rougher.There are some smaller, brighter boulders on the northeast surface of the east unit, and they are not considered as a part of the unit.The east unit is about 9.1 m thick, and its normal albedo is 5.2 AE 0.7%.The 10 largest clasts in the east unit have average lengths that range from 25 to 43 cm.The largest clast has a settling velocity of 15.0 cm s À1 .

Simurgh Saxum
Simurgh Saxum is located in the midlatitudes of the southern hemisphere, and its western edge defines Bennu's prime meridian.Three units are identified for this boulder: a west unit, a middle unit, and an east unit (Figure 7b).The west unit is about 11 m thick, and the normal albedo is 6.5 AE 0.6%.The west unit's largest 10 clasts range from 20 to 85 cm, with the settling velocity of the largest clast being 21.1 cm s À1 .This unit appears to have a smoother surface, and there are some linear fractures that run NNW-SSE, mainly observed in the west and east units.
The boundaries that define the middle unit are linear, but not parallel to each other.The middle unit is thinner toward the southwest.It is about 11.7 m thick at the widest part.It is the darkest of the three units, with a normal albedo of 4.2 AE 0.8%.Clasts are visible in this unit.The 10 largest clasts in the middle unit range from 32 to 82 cm.The settling velocity of the largest clast is 20.6 cm s À1 .
The east unit is shaped like a triangle.The distance from its eastern edge to the boundary with the middle unit is about 19.8 m, which is the thickest of any observed layer.There are linear fractures on the southern edge, similar to the ones on the west unit.The largest 10 clasts in east range from 20 to 76 cm.The largest clast has a settling velocity of 19.9 cm s À1 .The surface of the east unit appears smoother and brighter than the middle unit, with a normal albedo of 5.7 AE 0.6%.
In terms of normal albedo, the middle unit is distinct from the west and east units, but the west and east units cannot be distinguished from each other within our measurement uncertainty.

Roc Saxum
Roc Saxum is about 95 m across, and the largest boulder observed on Bennu.It is located in the southern hemisphere, east of Simurgh Saxum.We define two units in this boulder; a bedding unit and host rock (Figure 7c).The bedding unit is a layer observed on the eastern edge, with visible linear fractures (beddings).The rest of the boulder is the host rock and is not a layer.The bedding unit has a thickness of 2.2 m and a length extending to almost 20 m.This unit appears to be sticking out from the host rock.The average thickness of the thin beddings within the unit is about 14 cm.The normal albedo is 5.4 AE 0.5%, whereas that of the host rock is 4.2 AE 0.5%.

Layered Boulder Formation
Bennu is a small, rubble-pile asteroid with microgravity (Barnouin et al., 2019).It is unlikely that geologic processes large-scale enough to produce meterscale layers occurred on Bennu.Therefore, we focus on formation hypotheses set on Bennu's Main Belt parent asteroid.Here, we evaluate four possible formation scenarios for the observed layered rocks: (1) protoplanetary disk origin, (2) brecciation, (3) sediment deposition, and (4) intrusion.

Protoplanetary Disk Origin
The protoplanetary disk was a rotating disk of gas, dust, and ice around the Sun that existed before the planets and other solar system bodies formed from these materials.In the disk, dust particles and aggregates FIGURE 7. PolyCam images of layered rocks (Table 1).The images are oriented so that Bennu north is up, unless otherwise indicated in the image.Arrows point to the boundary, and names of units are annotated.K. Ishimaru and D. S. Lauretta collide and coagulate, and they grow to form refractory solids and chondrules which are then accreted into asteroids (Dominik & Tielens, 1997;Trigo-Rodriguez et al., 2006).Chondrules are a major component of meteorites from undifferentiated asteroids, which indicates that they were abundant in the disk.They are spherical, with diameters of micrometers to a centimeter, and mainly composed of silicate minerals such as olivine and pyroxene.It is possible that chondrules were size-sorted in the protoplanetary disk or on a parent asteroid.Turbulent concentration is one explanation for chondrule size sorting in the disk, where isotropic, homogeneous, threedimensional turbulence in disk gas causes particles with certain aerodynamic properties to concentrate in zones (Cuzzi et al., 2001;Teitler et al., 2010).Alternatively, chondrule accretion on to an existing planetesimal leads to aerodynamical sorting of chondrules, and larger chondrules and smaller chondrules are thought to accrete on the surface in layers (Johansen et al., 2015).Metzler et al. (2019) reported that such size sorting of chondrules in the disk or during accretion is recorded in the brecciated ordinary chondrite Northwest Africa (NWA) 5205.They found three clasts (centimeter-to decimetersized) with dramatically different mean chondrule sizes in two-dimensional measurements.Although these particular clasts are not layered relative to one another, these observations provide evidence that particles with a certain size can accumulate and lithify with other clasts due to nebular accretion.If the layers in Bennu's boulders formed in the protoplanetary disk as size-sorted chondrules, their mineralogy should include olivine or pyroxene (Lauretta et al., 2006).However, on Bennu, there are no such minerals identified, with the exception of rare pyroxenerich clasts linked to asteroid (4) Vesta (DellaGiustina et al., 2021).This observation does not rule out the possibility that the layers formed as a result of chondrule size sorting and the chondrules later turned into pseudomorphs composed of phyllosilicates (Trigo-Rodrıǵuez et al., 2019).However, the presence of large carbonate veins, formed by precipitation from an aqueous solution, leads us to favor a fluid flow origin.

Accretionary Collision
Comets 67P/Churyumov-Gerasimenko (hereafter, 67P) and 9P/Tempe 1 show morphology that suggests that the nuclei are composed of stacked layered structures on 10 to >100 m scales (Belton et al., 2007;Thomas et al., 2015).These are thought to have formed by accretionary collision in low velocity (a few meters per second) (Jutzi & Asphaug, 2015).In this process, some material of one of the colliding bodies accretes in layers in the combined body.
Bennu's parent asteroid spent most of its time in the main asteroid belt (Ballouz et al., 2020), where the most probable impact velocity is about 4.4 km s À1 (Bottke et al., 1994).This velocity is too fast for the collisional accretion to occur.Also, this mechanism would not produce the thin, linear beddings observed in Gargoyle Saxum and Roc Saxum.Therefore, layering due to low-velocity collisions is unlikely as a formation mechanism of the layered boulders on Bennu.

Propagation of Phase-Change Fronts
Comet 67P exbibits a layered structure that extends for $290 m in the Hathor cliff.The average vertical thickness of the strata is $14 m (Belton et al., 2018), which is similar to our measurements from Bennu.The strata on 67P are delineated by topographic depressions.Belton et al. (2018) proposed that the strata on 67P formed by cracking that was induced by a thermodynamic phase change of ice.When the comet approached the inner solar system, starting at a range  near $8 AU, the surface temperature increased to 115-120 K.This thermal pulse initiated crystallization of the amorphous water ice.The phase change front propagated through the interior and sculpted the layered structure (Belton et al., 2018).
In the layered boulders on Bennu, the layers are defined by boundaries between optically distinct lithologies, whereas in the case of comet 67P's Hathor cliff, the layers are defined by cracks.Therefore, this formation mechanism is unlikely to be responsible for Bennu's layered rocks.

Brecciation
Breccia are rocks that consist of clasts cemented by a fine-grained matrix.In meteorites, clast size ranges from micrometer scale to centimeter scales.On Bennu and in terrestrial settings, meter-scale clasts have been observed in breccias (DellaGiustina et al., 2020;Druitt & Sparks, 1982).Meteorite breccias are thought to form by slow velocity impacts during accretion of their parent asteroids or by fast impacts with other asteroids (Bischoff et al., 2006).
We cannot completely rule out the possibility that the layered boulders on Bennu are breccias, where one unit is a clast and the adjacent unit consists of matrix.This would explain the coexistence of different lithologies in one boulder.The linear boundaries between the observed layers can be explained if the boulder is a breccia with angular clasts that are several meters long.However, some boulders (Gargoyle Saxum, Roc Saxum, and LR10) exhibit thin beddings that are parallel to the boundary.These linear features are not seen in breccias.The middle units of LRs 9, 10, and 12 and the five units of LR11 are elongated with nearly parallel sides.Such multiple parallel sides are unlikely to be seen in breccias because usually they consist of randomly oriented, irregularly shaped clasts (Bischoff et al., 2006).Also, none of the observed layers show a complete outline as a clast in a matrix, whereas the breccias on Bennu that are already reported show distinct clasts (DellaGiustina et al., 2020;Walsh et al., 2019).Therefore, it is unlikely that the layered structures observed in boulders on Bennu are formed through brecciation.

Metamorphism by High Pressure
Rock metamorphism is a process where rock is altered mineralogically and chemically as well as structurally, typically by high temperature, high heat, or reaction with fluids.This process can produce a layered structure called foliation when high pressure (> $1 GPa) applied to rocks for millions of years aligns the minerals within the rocks.Aligned minerals form cleavages which can produce parallel, linear cracks on the surface of the rock.On Earth, large-scale geologic processes induced by the plate tectonics provide high pressure and temperature to alter rocks on geologic timescales.
Carbonaceous asteroids on a scale of 100 km in diameter are not large or dense enough to produce such a pressure even at the center.If we assume its density to be 2.92 (g cm À3 ), the same as CM chondrite material, and the diameter to be 100 km, the pressure at the center would be only a few MPa.Instead, impacts can provide high pressure (55 GPa or higher; St öffler et al., 1991) to alter the mineralogy of asteroid materials (Tomioka & Miyahara, 2017).However, high pressure by impacts does not last long enough to align the minerals to form cleavages.Therefore, it is unlikely that the thin beddings observed in Bennu's boulders are of metamorphic origin.

Impact Ejecta
Impact ejecta forms when an impactor hits a surface of a planetary body and the target materials are ejected beyond the crater rim.It is proposed that the target material on the surface is ejected ballistically and deposits with relatively less shock, and that the target material beneath it melts and deposits on top (Osinski et al., 2011).Some impact ejecta with fluidized appearance are observed on several bodies including Mars and Ceres.There are multiple formation mechanisms proposed for ejecta fluidization, such as volatiles in the target materials, interactions with atmosphere, and granular flow (Hughson et al., 2019 and references therein).
Impact ejecta with different lithology layering on top of one another can create the layered structures of distinct units observed in Bennu's layered boulders.Whether the Bennu material is impact ejecta can be tested with the sample of Bennu returned by the OSIRIS-REx mission (Lauretta et al., 2021) by investigating its shock level (Beitz et al., 2016).However, the thin beddings within the same lithology might not be produced by this mechanism.

Sediment Deposition
The possibility of water flow at the centimeter scale in carbonaceous meteorite parent asteroids is supported by some meteorite studies, such as the observation of a lens feature in CM chondrites (Rubin et al., 2007;Trigo-Rodrıǵuez et al., 2019;Trigo-Rodrıǵuez & Rubin, 2006).The discovery of carbonate veins that are more than 1 m long on Bennu suggests that there was water flow at kilometer scales on the parent body (Kaplan et al., 2020).This large-scale water flow should have affected most of the material we see on Bennu.It has been proposed that melted ice and rocky materials in primordial asteroids could have caused global convection (Bland & Travis, 2017).In this model, the heat from radioactive decay of short-lived radionuclides melted the accreted ice, and the resultant liquid water mobilized accreted fines and chondrules.As the water and rocky particles reacted, the exothermic reaction of serpentinization also became a heat source.The convective flows suspended and lofted particles in upwelling flows, which then moved sideways and sank.Bland and Travis's (2017) model shows that, in many cases, larger chondrules (1 and 0.5 mm) can settle in the inner core, surrounded by smaller chondrules (0.1 mm, 50 μm, and 10 μm).Fine particles and water (i.e., mud) make the mantle, and the crust is made of ice at the end of the convection, resulting in a kilometer-scale size-sorted interior structure.Coarse particles become immobile when they settle to the core and reach their maximum packing density.This core formation takes from 25,000 years to more than 100,000 years.Water with suspended fine particles can still flow through the pores in the core.If this flow is mostly water, this could be how the observed veins formed.Also, the fluid flow scenario is consistent with numerical studies that predict that aqueous alteration occurred at large ($10 km) scales (Bland & Travis, 2017;Grimm & McSween, 1989;Mcsween et al., 2002;Palguta et al., 2010;Travis & Schubert, 2005;Young, 2001;Young et al., 1999Young et al., , 2003)).
The observed layered boulders could be the result of sediment deposition in the fluid flow.In the convective system, some particles can be suspended in flowing liquid water.The size of the particles that can be transported by the flow depends on its velocity; faster flow can transport larger clasts.The largest clast in sediment thus constrains the lowest required flow velocity.Particles are deposited when their settling velocity is larger than the upwelling flow velocity (Figure 8).As shown in Figure 3, the settling velocity of a particle in water increases from the center to the surface due to the increase in gravitational acceleration.If a flow with constant velocity moves radially toward the surface, the settling velocity of a suspended particle might become equal to the flow velocity at a certain radius.At that point, the particle will stay there until the velocity balance is disturbed.
Flow velocity can change as production rate of water changes, such as when the convection is starting or ending, and when serpentinization starts.As sediment is deposited, changing permeability could affect the flow velocity.Another possible mechanism is episodic depositions with distinct flow speeds.Settling velocity can change when a nonspherical particle changes orientation and its hydrodynamics change, for example, when a flaky particle turns sideways.When the flow speed decreases, and/or the particle becomes more hydrodynamic and settling velocity increases, the particle sinks and deposits.As many particles settle, this would create a linear flat surface.Larger particles settle closer to the center than smaller particles due to their larger settling velocities.Layered structure forms when multiple sets of particles with distinct sizes settle on top of one another.
In the layered boulders on Bennu, the largest clasts in each unit vary in size, which could reflect velocity changes (Figure 9).In the case of Gargoyle Saxum, the minimum required flow velocity to transport the largest clast in the top unit would have had to be at least 21.1, or 26.5 cm s À1 if we include Jewel as a clast.In comparison, the largest clast in the bottom unit corresponds to a flow velocity of 10.2 cm s À1 .This suggests that there was a change in flow velocity between the two deposition events.
The observed boundaries between units are linear for a few meters to more than 10 m.Gargoyle Saxum exhibits thin ($50 cm) beddings that are also linear and parallel to the boundary.In terrestrial environments, beddings are found in sedimentary rocks and form by sediment deposition.Therefore, based on the principle of original horizontality, sediment deposition from the fluid is the most likely formation mechanism.This model provides a simple mechanism of sediment deposition without chronological context.Dating carbonates in the returned sample will constrain the timing of the aqueous alteration in the parent body.

Intrusion
Intrusion is generally thought of as an igneous rock intruding into a pre-existing rock on Earth.In this paper, we refer to intrusion as a mobile mixture of rocky particles and water intruding into a pre-existing body on the parent asteroid.The pre-existing body of rock can be unconsolidated.Intrusion and sediment deposition in hydrothermal convection share the same process, where a mixture of water and rocks flows.The difference is that the water-to-rock ratio is lower in the case of intrusion.At a high particle density, there is no longer particle settling, and the mud flows as one unit.Bland and Travis (2017) show that radial upwelling flows can move through settled particles during global convection.This process can be preserved as intrusive rocks.Among Bennu's layered boulders, examples of possible intrusion can be seen in Simurgh Saxum and LRs 9, 10, and 12.These boulders' east and west units have overlapping normal albedo values, and their textures appear similar.However, their middle units have distinct albedo values (Figure 7, Table 2).In Simurgh Saxum, there are linear fractures that extend in the direction of NNW-SSE on the west unit, and they seem to be continuous with the linear fractures seen in the east unit.Therefore, the two units could have been one rock, and the middle unit intruded into it.The two boundaries are each linear, but they are not parallel to each other.The middle unit is thickest, about 9 m on the northeast, and it becomes thinner toward the southwest.This could indicate that the southwest side was near the tip of the intrusive flow.
In LR10, a part of the middle unit sticks out on the northern edge of the boulder (Figure 7g,h).This suggests that the material that composes the middle unit is more resistant than and distinct from that of the west and east units.Therefore, it is possible that the west and east units were a part of one larger unit and the middle unit flowed through it.

CONCLUSION
A layered structure consisting of distinct lithological units with linear boundaries was identified in 11 boulders on asteroid Bennu.Those layered boulders range from about 4 to 95 m in longest dimension, and the layer thickness ranges from 46 cm to 19.8 m.Most units have a distinct normal albedo value from their adjacent units.Given Bennu's microgravity, small size, and rubble pile structure, it is unlikely that dynamic geological processes that could produce the observed meters-thick layers occurred on this asteroid.Based on the observations of the linear boundaries between layers and presence of thin beddings within units, we propose that the layered boulders formed via sediment deposition and intrusion as a result of global-scale hydrothermal convection on Bennu's parent body.Hydrothermal convection on the parent body is supported by numerical studies and observation of carbonate veins on Bennu.The difference in clast size among units in the same boulder suggests that there were episodic deposition events in the fluid flow.The largest clast embedded in the layered boulders is 85 cm in average length, and it would require a flow velocity of 21.1 cm s À1 to be transported.A higher velocity of 26.5 cm s À1 would be required if a 134.0-cm rock on the surface of a layered boulder was originally an embedded clast.This study strongly supports the open-system model for aqueous alteration on carbonaceous chondrite parent asteroids.

FIGURE 1 .
FIGURE 1. Image of Gargoyle Saxum, which is one of the layered boulders.The top unit, bottom unit, and Jewel-a brighter rock on the surface of the top unit-are annotated.The arrow points to the linear boundary that demarcates the top and bottom units, and north is indicated.The circle highlights the thin beddings at the boundary.

FIGURE 3 .
FIGURE 3. (a) Example of our clast size analysis.Eight measured lengths are shown overlaid on Jewel, the brighter rock on Gargoyle Saxum's top unit.(b) Example of our layer thickness measurements.The yellow lines indicate layer boundaries.The magenta lines represent the layer thickness of Simurgh Saxum's middle unit and east unit.

FIGURE 4 .
FIGURE 4. Settling velocity as a function of radius within an asteroid with a 100-km diameter.The plot shows that as the distance from the center increases, the settling velocity increases.A particle radius of 40 cm is used as an example here.(Color figure can be viewed at wileyonlinelibrary.com)

FIGURE 5 .
FIGURE 5. Distribution of the layered boulders on Bennu, shown on a global mosaic in equirectangular projection (Bennett et al., 2021).(Color figure can be viewed at wileyonlinelibrary.com)

FIGURE 6 .
FIGURE 6.(a) PolyCam image of Gargoyle Saxum (Table 1).The top and bottom units and Jewel are annotated.The arrow points to the unit boundary, and north is indicated.The circle highlights a possible additional unit beneath the bottom unit.(b) Jewel and a depression are shown in a close-up image of Figure 1.The dotted line outlines the depression that is similar to Jewel in size and shape.(c-g) DTMs of Gargoyle Saxum's south face, east face, north face, west face, and top, respectively.In (d) and (e), the surface beneath the boulder going into the ground appears smooth.It is caused by the decreased number of data points there because the area is hidden by the boulder.In (d) and (f), the arrows show that the bottom of the boulder on the north end is above the ground.(Color figure can be viewed at wileyonlinelibrary.com) FIGURE 7. PolyCam images of layered rocks (Table1).The images are oriented so that Bennu north is up, unless otherwise indicated in the image.Arrows point to the boundary, and names of units are annotated.(a) LR1.(b) Simurgh Saxum.(c) Roc Saxum (northeast face).(d) LR7.(e) LR8.(f) LR9.(g) LR10.(h) LR10 lidar-based DTM in the same orientation as the image in (g).It shows how the boulder is elevated and the ridge is the middle unit, which is not clearly noticeable in the image.(i) LR11.(j) LR12.(k, l) LR13.(k) was used to estimate normal albedo and layer thickness because the boulder is oriented so that a face is shown with less tilt.(l) The linear boundary is seen well in this image.(Color figure can be viewed at wileyonlinelibrary.com)

FIGURE 8 .
FIGURE 8. Proposed model of sediment deposition based on the mud convection model from Bland and Travis (2017).(Color figure can be viewed at wileyonlinelibrary.com)

FIGURE 9 .
FIGURE 9. (a) Settling velocity curve at the asteroid surface as a function of particle size.The colored lines indicate the average lengths of the largest clasts in each unit and their settling velocity, which is also the minimum flow velocity to transport them.The figure includes Gargoyle Saxum top unit excluding Jewel as well as including Jewel.(b) A close-up of the squared area in (a).

TABLE 1 .
List of PolyCam images used in this analysis and their properties.
Note: An asterisk indicates images used to estimate the normal albedo in Table2.

TABLE 2 .
The results of the layered boulder analysis.Boulder longest dimensions, normal albedo of all units within the studied boulders, and layer thickness are presented.The longest dimensions of the boulders have an uncertainty of 20 cm.Images used to estimate the normal albedo values are indicated in Table1.The large SDs of the units in LR9 are due to its very rough surface, which includes many overexposed spots and shadowed spots in the photometrically corrected image.Layer thickness measurements of Jewel and Roc Saxum's host rock are not included because they are not layers.The layer thickness values have uncertainty of 4 pixels. Note:

TABLE 3 .
Largest clasts observed and their settling velocities.