1.1 Mars 2020 and the Opportunity for Core Orientation

The Perseverance rover is currently exploring Jezero crater to search for signs of past life and characterize the climate and geology of ancient Mars. Jezero is a 45-km diameter complex impact crater situated in the Nili Planum region that formed sometime during the Early Noachian to Late Noachian (e.g., Holm-Alwmark et al., 2021). The original crater floor was covered by igneous rocks interpreted to be volcanic and/or intrusive (Farley et al., 2022) followed by clastic sedimentary materials interpreted to be fluvio-deltaic in origin (Mangold et al., 2021). The Mars 2020 mission's central objective is to drill, document and cache approximately 30 rock and regolith samples that could be returned to Earth for future laboratory studies (Farley et al., 2020). Thus far, the rover has explored the crater floor, fan front, upper fan, and crater margin and collected 20 cores of igneous and sedimentary rocks.

Unlike Martian meteorites, the rocks sampled by the rover have geologic and stratigraphic contexts that constrain their formation and evolution. Furthermore, unlike meteorites and nearly all returned samples from the Moon and asteroids, 18 of the 20 rock cores collected thus far are samples of unambiguous in situ bedrock (i.e., solid rock that underlies the regolith). If the orientations of the cores can be estimated relative to Martian geographic coordinates, this would enable their records of geologic and geophysical processes to be oriented relative to Martian geographic coordinates at the time the rocks crystallized, emplaced and/or were altered.

Oriented samples provide the foundation for paleomagnetic studies, which seek to determine the direction and intensity of natural remanent magnetization. In particular, oriented bedrock samples would enable the first measurements of the paleodirection of another planetary body's past magnetic field. These could be used to test the hypothesis that the waning of an early Martian dynamo led to atmospheric loss on Mars (Gunell et al., 2018; Jakosky & Phillips, 2001; Mittelholz et al., 2018). Such studies rely on an understanding of when a rock acquired its paleomagnetic records. This can be powerfully constrained using field tests that measure the relative orientations of magnetization directions of samples of conglomerate clasts and folded and tilted rocks (Butler, 1992; Tauxe, 2018). Because the orientation of a dipolar magnetic field varies systematically with latitude, the field paleodirection could also be used to establish whether Mars experienced true polar wander (Perron et al., 2007; Thomas et al., 2018) and/or plate tectonics (Connerney et al., 2005; Golombek & Phillips, 2010) and constrain regional tectonic processes. Paleodirectional data could additionally be used to study the secular variation of the dynamo and search for geomagnetic reversals (Merrill et al., 1998), thereby constraining the evolution of Mars' interior and potentially developing a magnetostratigraphic timescale.

Magnetic analyses of oriented samples may also enable the identification and interpretation of secondary remagnetization processes in a way that would be difficult to do for unoriented samples. Examples include the interpretation of magnetizations associated with lightning strikes (Carporzen et al., 2012; Verrier & Rochette, 2002), the measurement of emplacement temperatures of volcanoclastic rocks by analyzing unidirectional thermally induced magnetization overprints after deposition (Lerner et al., 2022), the timing of formation of pigmentary hematite (Swanson-Hysell et al., 2019), and the timing of diagenetic alteration (Heij & Elmore, 2020). Experience with terrestrial rocks has shown that essentially all of the above paleomagnetic investigations can be achieved with samples whose orientations are estimated with uncertainties ∼3° (Tauxe, 2018).

Oriented samples could also be used for sedimentological, volcanological, and structural geological studies. The orientations of petrofabrics such as ripples and crossbedding in sedimentary rocks (Tucker, 2001) and crystal orientations in igneous rocks could constrain the direction of the flows that produced these rocks (e.g., Geoffroy et al., 2002). Crystal orientations can also provide information about the paleohorizontal direction during diagenesis (Tucker, 2001). The orientation of gradients in clastic sedimentary grain sizes can indicate changes in flow rates (Tucker, 2001), while the orientations of gradients in crystal sizes, vesicle sizes, and vesicularity reflect the relative locations of cooling surfaces (e.g., lava flowtops) (e.g., Cashman & Kauahikaua, 1997; Potter et al., 2019). The orientation of textures in sedimentary rocks and, in particular, the slope of their layers relative to paleohorizontal can even be indicators of biogenicity. For example, microbially-stabilized layers in stromatolites can be steeper than the angle of repose for loose sediment, while different biological and abiotic growth processes also contribute to the orientation and tilt of stromatolitic columns (Allwood et al., 2018; Bosak et al., 2013; Donaldson, 1976; Hoffman, 1976; Logan, 1961; Petroff et al., 2013). Finally, the orientations of microfractures, twinning, and deformation lamellae can constrain the stress and strain history of the rocks (e.g., from impacts and endogenic tectonic processes) (Ramsay & Huber, 1983). Again, orientation uncertainties of a few degrees are sufficient to achieve many of these investigations on Earth.

Despite the great value of oriented bedrock samples, neither obtaining information about the orientations of the samples nor the acquisition of samples of bedrock are requirements for the Mars 2020 mission. As a result, the development of techniques for orienting cores had not been planned by the time of launch. Following landing, we addressed this gap through discussions within Mars 2020 science working groups (e.g., the Returned Sample Science Working Group), instrument teams (Wide Angle Topographic Sensor for Operations and eNgineering [WATSON] and SuperCam), team-wide Science Discussions, and deliberations amongst the rover engineers, Long Term Planners, and the Mars 2020 Project Science office. Our collective efforts led to a recognition of the value of oriented samples, prioritization of bedrock samples over float rocks, and a directive to develop methods for orienting the cores. Here we describe a method that we developed for orienting Perseverance rock cores and its application to the first 20 rock cores drilled with the rover. Fundamentally, the method uses the rover's precise knowledge of the orientation of its Drill and WATSON camera during and after the drilling to constrain the orientation of the cores. We show how, with this information, we have oriented all rock cores drilled thus far to better than 2.7° (3σ) total uncertainty.

1.2 History of Acquiring Planetary Samples

Rock samples have been acquired by humans and robotic missions from the surfaces of the Earth, the Moon, three asteroids, and a comet. On the Moon, Apollo samples were acquired as single rocks, soil samples, and chips of boulders broken off with hammers (Heiken et al., 1991). Essentially all samples and parent boulders were float rather than bedrock with the exception of the regolith breccia 70019, which sampled the glass melt lining a <7 million-year-old crater (Schmitt et al., 2017), and possibly also mare basalt samples 15595 and 15596 that were chipped off a boulder outcropping from flows exposed on the edge of Hadley Rille (Lofgren et al., 1975). Some of the larger (≳10 cm) rock samples were oriented in lunar geographic coordinates using astronaut photographs taken with an optical camera (Kammerer & Zeiss, 1972) and a gnomon for measuring shadow angles to determine geographic north (Allton, 1989). Because the local horizontal was not marked on the samples, their orientations were reproduced after return to Earth by illuminating them under simulated lighting conditions like that during sampling (e.g., Nichols et al., 2021; Wolfe et al., 1981).

The Luna 24 (Florensky et al., 1977) and Chang'e-5 (Li et al., 2021) missions each returned unoriented regolith. The Stardust mission returned unoriented dust particles from comet 81/Wild2 (Brownlee et al., 2006), while Hayabusa 1 and 2 and the Origins-Spectral Interpretation-Resource Identification-Security-Regolith Explorer (OSIRIS-REx) missions returned unoriented regolith from asteroids (25143) Itokawa (Nakamura et al., 2011), (162173) Ryugu (Yada et al., 2022), and (101955) Bennu (Lauretta et al., 2021), respectively. In summary, other than Apollo sample 70019, no samples returned from bodies other than Earth are unambiguous bedrock samples. As a result, other than for 70019, the orientations of all samples at the time they formed relative to planetary geographic coordinates are unknown.

In contrast, countless oriented samples of bedrock have been acquired on Earth. Such samples have been acquired in three main ways. First, near-surface (≲20 cm deep) core samples are commonly acquired using a handheld diamond tipped rotary drill (for hard rock) or with non-rotary drive coring devices (for soft sediments) (Tauxe, 2018; Turner et al., 2015). These are oriented using a sun compass, magnetic compass, and/or a global navigation satellite system and can have uncertainties as low as ±3° (Cañón-Tapia, 2007; Fukuma & Muramatsu, 2022; Tauxe, 2018). Second, block samples can be chipped off bedrock and typically oriented with a sun compass and/or magnetic compass with uncertainties that can approach those typically achieved with hand held drills (Collinson, 1983). Third, deeper drill cores ranging up to several km in length are acquired with mechanized drill rigs and can be fully oriented using continuous scribing with an oriented core barrel (Davison & Haszeldine, 1984), spatial correlation of natural features in the core with either oriented borehole images and/or regionally extensive geologic structures (Davison & Haszeldine, 1984; Paulsen et al., 2002), and/or correlation of the direction of natural remanent magnetization in the core with present geographic north (Sugimoto et al., 2020). These methods can have uncertainties as low as ±10° (Davison & Haszeldine, 1984; Nelson et al., 1987).

1.3 Cores Drilled by Perseverance

The Mars 2020 mission is acquiring the first extensive sample suite from another planet (Figure 1). Perseverance has thus far acquired 23 samples from 15 different workspaces (Farley & Stack, 2022, 2023; Simon et al., 2023) (Table 1). Of these 23 attempts, 21 were cores drilled from solid rock targets and 2 of regolith. Of the 21 rock coring attempts, 20 were successful; the first coring attempt (Roubion) sampled only the atmosphere and possibly a few unoriented grains because the rock crumbled. Of the 20 successful rock cores, 18 were acquired from in situ bedrock: 6 samples taken from igneous rocks (likely volcanic and olivine-rich cumulates of broadly basaltic composition) and the remaining 12 from sedimentary rocks (siltstones and sandstones with either bulk basaltic or sulfate-rich compositions) (Table 1). The remaining 2 cores (Montdenier and Montagnac) were acquired from a ∼40-cm diameter tabular igneous boulder that was differentially tilted with respect to adjacent bedrock and moved during the course of drilling. Future laboratory measurements on this unique sample suite should collectively enable many of the aforementioned studies of the igneous, sedimentological, tectonic and magnetic evolution of Mars (Section 1.1).

2.1 Orienting the Core by Orienting the Coring Drill

Mars 2020 samples are acquired with the Coring Drill (referred to hereafter as the Drill), a rotary percussive drilling mechanism at the end of the rover arm (Moeller et al., 2021) (Figure 2). The Drill employs a Coring Bit (a 27-mm outer diameter, 185-mm long stainless steel cylinder coated with titanium nitride and tipped with 4 tungsten carbide teeth) containing an interior Sample Tube. Core samples are typically 13 mm in diameter and 61–81 mm long, such that the typical kerf is 7 mm. The Drill can acquire cores at an angle of up to 10° between the Drill axis (pointing vector) and surface normal and at an angle up to 110° from local gravity (i.e., 20° up from geographic horizontal).

Prior to drilling, the outcrop surface is imaged at high resolution (∼25–30 μm at a typical 6–7 cm standoff distance) using WATSON (Bhartia et al., 2021). The Drill is then placed just above the outcrop face with a lateral horizontal drill uncertainty of typically ±5.4 mm (3σ) but occasionally ±9.4 mm (3σ). Following placement, two Stabilizers on either side of the Coring Bit are pushed into the rock face (this is called preloading) (Figure 3). Then the Coring Bit is advanced progressively into the rock, chiseling out a cylindrical kerf of rock flour (Movies S1–S8). At the completion of drilling, the Stabilizers are unloaded and rotation of the Drill body about an axis offset from the center of the Sample Tube breaks the core with shear pressure. Finally, the Coring Bit, now with the rock sample inside the interior Sample Tube, is extracted from the outcrop. Following imaging of the core's bottom (i.e., inner end) using the Caching System Camera (CacheCam), the sample is cached inside the rover body (Moeller et al., 2021).

Because the cylindrical sidewalls of the Coring Bit are essentially parallel to the borehole, we estimate the orientation of the core by orienting the Drill itself. Because the Drill's orientation may change slightly during drilling, we have chosen the orientation of the pointing vector of the Drill just after completion of drilling but before Stabilizer unloading as an indication of the mean orientation of the pointing vector of the core. As we show in Section 3.2, we have found that the orientation angle of the Drill generally varies by ≲0.7° (3σ) during drilling. Therefore, the estimated orientation is not sensitive to the choice of when during the drilling process the Drill orientation is assigned as the core orientation.

2.2 Reference Frames

The process of estimating the core orientation involves six coordinate systems that orient the core relative to the Perseverance rover and Mars (Table 2). Here we define these orientation systems and show how they are interrelated. We focus on the coordinate systems' orientations rather than the locations of their origins since only the former are relevant for core orientation. All the reference frames described here are Cartesian and right-handed. We will use the notation $${\widehat{A}}_{k}$$ to generically denote a unit vector for a particular Cartesian axis, k = x, y, z, associated with a particular reference frame, A, such that $${\widehat{A}}_{x}$$ = [1, 0, 0], $${\widehat{A}}_{y}$$ = [0, 1, 0] and $${\widehat{A}}_{z}$$ = [0, 0, 1]. To express these vectors in another frame, B, we use $${\widehat{A}}_{k,\mathrm{B}}$$ with individual components A_k,l,B for the Cartesian axes in another frame B of l = x, y, z.

We make use of three reference frames describing the orientation of the rover and two of its subsystems. The Rover Mechanical frame (R_x, R_y, R_z) is fixed with respect to the rover body (Figure 4). Here, $${\widehat{R}}_{x}$$ points toward rover forward, $${\widehat{R}}_{y}$$ points toward starboard, and $${\widehat{R}}_{z}$$ points toward rover down. The Drill reference frame consists of a pointing vector, $${\widehat{D}}_{x}$$, pointing along the Drill axis away from the rover body, while $${\widehat{D}}_{y}$$ points perpendicular to the axis containing the two Stabilizers toward the Planetary Instrument for X-Ray Lithochemistry instrument (Allwood et al., 2020), and $${\widehat{D}}_{z}$$ points along the axis of the Stabilizers (Figures 2 and 4). Likewise, the rover's WATSON frame (W_x, W_y, W_z) consists of a vector, $${\widehat{W}}_{x}$$, pointing into the image plane toward the outcrop (Figure 4) with $${\widehat{W}}_{y}$$ pointing in the plane of the image toward the top of the image, and $${\widehat{W}}_{z}$$ pointing in the plane of the image toward the right. We refer to the image plane $${\widehat{W}}_{y}$$−$${\widehat{W}}_{z}$$ as the “WATSON plane.”

Two reference frames specify the rover's orientation with respect to Mars. First, the Martian Geographic frame (G_x, G_y, G_z) (Figure 5) is a fixed frame independent of the rover position and orientation. Martian north (N) lies along $${\widehat{G}}_{x}$$, east (E) along $${\widehat{G}}_{y}$$, and nadir (down [D]) lies along $${\widehat{G}}_{z}$$. North points toward Mars' rotation pole, east lies in Mars' mean equatorial plane, and nadir points toward Mars' center of mass (Archinal et al., 2018).

Second, the Site frames (S_x,i, S_y,i, S_z,i) are a collection of i frames fixed with respect to the local surface at various locations visited by the rover, each of which are estimates of the Martian Geographic frame (Deen, 2022) (Figure 5). Martian north is estimated to lie along $${\widehat{S}}_{x,i}$$, east is estimated to lie along $${\widehat{S}}_{y,i}$$, and nadir (down) lies along $${\widehat{S}}_{z,i}$$. While driving, the rover uses its gyroscopes to estimate the three angles quantifying its orientation: roll (right handed rotation angle about $${\hat{R}}_{x}$$), pitch (right handed rotation angle about $${\hat{R}}_{y}$$), and yaw (right handed angle about $${\hat{R}}_{z}$$). Because the rover accumulates uncertainty in its estimated orientation as it drives, the uncertainty estimation of the Geographic frame grows with time (Deen, 2022). This eventually necessitates a “Surface Attitude Positioning and Pointing Operations (SAPP) update” in which the rover obtains a new, more precise estimate of its orientation. The SAPP update process uses rover accelerometers to estimate the roll and pitch, gyroscopes to estimate the yaw $${\widehat{R}}_{z}$$ and, occasionally, additional observations of the position of the Sun (a “sunID update”) to obtain the most accurate estimate of the yaw (Ali et al., 2005; Trautman et al., 2022) (see Figure 4). Some of these SAPP updates are associated with an increase of the Site Index, i, which redefines the position and orientation of the Site frame actively being used by the rover. Unless otherwise stated, we will not distinguish between the Martian Geographic and various Site frames in the rest of this manuscript and we will drop the Site frame subscript, i. We will use the terms “Site horizontal” to denote the $${\widehat{S}}_{x}$$-$${\widehat{S}}_{y}$$ planes and “Site down” and “Site up” to denote the axial directions parallel and antiparallel, respectively, to $${\widehat{S}}_{z}$$ (Figure 6).

Finally, we define the core frame as having $${\widehat{C}}_{x}$$ pointing along the core into the outcrop, $${\widehat{C}}_{y}$$ pointing upward along the steepest direction of the plane normal to $${\widehat{C}}_{x}$$ (i.e., $${\widehat{C}}_{x}$$−$${\widehat{C}}_{y}$$ lies in the Site vertical plane), and $${\widehat{C}}_{z}$$ lying in the Martian geographic horizontal plane, completing the triad (Figure 6). We refer to the $${\widehat{C}}_{y}$$−$${\widehat{C}}_{z}$$ plane as the “Core plane,” which may be canted with respect to the outcrop surface and with respect to the WATSON plan. $${\widehat{C}}_{x}$$ is to high accuracy (<0.1°) aligned with $${\widehat{D}}_{x}$$ (see end of Section 2.1), but $${\widehat{C}}_{y}$$ and $${\widehat{C}}_{z}$$ can be at large angles with respect to, $${\widehat{D}}_{y}$$ and $${\widehat{D}}_{z}$$, respectively.

2.3 Geometric Quantities Defining Core Orientation

The orientation of a solid body in three-dimensional space can be parameterized with an axis-angle representation consisting of a unit vector indicating the direction of an axis of rotation and an angle prescribing the magnitude of a rotation around this axis (Amirouche, 2006). Because the unit vector is itself described by two angles, three angles are necessary in total to describe a body's orientation. For Perseverance core orientation, the relevant unit vector is the core's pointing vector, $${\widehat{C}}_{x}$$, whose direction is quantified by the hade, H, and azimuth, A (Butler, 1992), while the rotation around this axis is quantified by the roll, α (Figure 6). Note that this core roll is distinct from the rover body roll mentioned above.

2.4 Quaternions

2.5 Procedures for Estimating Core Orientation

2.5.2 Determining the Roll

Estimation of the roll requires determining the rotational spin of the core relative to a fiducial direction. This angle cannot be estimated using rover Drill metadata because the core rotates in an unknown way with respect to the sample tube and core body. Instead, we use images of the outcrop prior to drilling taken with WATSON (Figures 7-10). These images can be registered with respect to the core top using rotationally asymmetric features on the outcrop location where the core is drilled and knowledge of the orientation of the WATSON frame at the time of imaging, t_W. Again, we use the Drill orientation as a proxy for the core orientation.

Equation 3 shows that the roll is a function of two quantities, here expressed in the Site frame. The first is the fiducial direction, the upward direction in the WATSON image, $${\widehat{W}}_{y,\text{Site}}\left({t}_{W}\right)$$. The second is the projection of the direction of steepest ascent in the Drill plane, $${\widehat{D}}_{y,\text{Site}}$$, into the WATSON plane: $${\text{proj}}_{{\widehat{W}}_{y}-{\widehat{W}}_{z}}\left[{\widehat{D}}_{y,\text{Site}}\left({t}_{W}\right)\right]$$. Note that we have found that in practice, the Drill pointing vector $${\widehat{D}}_{x}\left({t}_{d}\right)$$ and the WATSON pointing vector $${\widehat{W}}_{x}\left({t}_{W}\right)$$ have been canted with respect to one another by angles ranging from 0.2 to 38° depending on the core.

The vertical image direction vector, $${\widehat{W}}_{y,\text{WATSON}}$$ = [0, 1, 0], can be transformed from the WATSON to the Site frame by first rotating it into the Rover Mechanical frame and subsequently rotating into the Site frame. The quaternions representing the transformation from the WATSON frame to the Rover Mechanical frame, Q_Wb(t_W), and the transformation from the Rover Mechanical frame to the Site frame, Q_bll(t_W), at the time of WATSON imaging are provided by rover metadata. Specifically, the orientations of WATSON and the rover are captured and reported when the WATSON image is taken. With the quaternions known, the upward WATSON image direction in the Site frame $${\widehat{W}}_{y,\text{Site}}$$ can be calculated as:

$$${\widehat{W}}_{y,\text{Site}}\left({t}_{W}\right)={\boldsymbol{Q}}_{\boldsymbol{b}\boldsymbol{l}\boldsymbol{l}}\left({t}_{W}\right)\times {\boldsymbol{Q}}_{\boldsymbol{W}\boldsymbol{b}}\left({t}_{W}\right)\times {\widehat{W}}_{y,\text{WATSON}}\times {\boldsymbol{Q}}_{\boldsymbol{W}\boldsymbol{b}}^{\ast }\left({t}_{W}\right)\times {\boldsymbol{Q}}_{\boldsymbol{b}\boldsymbol{l}\boldsymbol{l}}^{\ast }\left({t}_{W}\right)$$$ (8)

To obtain $${\text{proj}}_{{\widehat{W}}_{y}-{\widehat{W}}_{z}}\left[{\widehat{D}}_{y,\text{Site}}\left({t}_{W}\right)\right]$$, we first calculate $${\widehat{D}}_{y,\text{Site}}$$ using $${\widehat{S}}_{z,\text{Site}}=$$ [0, 0, 1]:

$$${\widehat{D}}_{y,\text{Site}}\left({t}_{W}\right)=\frac{\left[{\widehat{D}}_{x,\text{Site}}\left({t}_{W}\right)\times {\widehat{S}}_{z,\text{Site}}\left({t}_{W}\right)\right]\times {\widehat{D}}_{x,\text{Site}}\left({t}_{W}\right)}{\vert \left[{\widehat{D}}_{x,\text{Site}}\left({t}_{W}\right)\times {\widehat{S}}_{z,\text{Site}}\left({t}_{W}\right)\right]\times {\widehat{D}}_{x,\text{Site}}\left({t}_{W}\right)\vert }$$$

We then project this onto the WATSON plane:

$$${\text{proj}}_{{\widehat{W}}_{y}-{\widehat{W}}_{z}}\left[{\widehat{D}}_{y,\text{Site}}\left({t}_{W}\right)\right]={\text{proj}}_{{\widehat{W}}_{x}}\left[{\widehat{D}}_{y,\text{Site}}\left({t}_{W}\right)\right]-{\widehat{D}}_{y,\text{Site}}\left({t}_{W}\right)$$$ (9)

where $${\text{proj}}_{{\widehat{W}}_{x}}\left({\widehat{D}}_{y}\right)=\left[{\widehat{D}}_{y}\cdot {\widehat{W}}_{x}\right]{\widehat{W}}_{x}$$ is the projection of $${\widehat{D}}_{y}$$ on $${\widehat{W}}_{x}.$$ With $${\widehat{C}}_{y,\text{Site}}\left({t}_{W}\right)\approx {\widehat{D}}_{y,\text{Site}}\left({t}_{W}\right)$$, the roll can now be estimated by substituting Equations 8 and 9 into Equation 3.

Translation of the direction $${\widehat{C}}_{y}$$ in the WATSON image to the core after it is returned to Earth requires identification of azimuthally asymmetric features on the core top. These rotationally asymmetrical features on the natural core top including nodules or bumps, crevices, and other markings that will be easily recognizable.

For cores lacking such natural markings, rotationally asymmetric artificial markings can be made on the core top using the laser on the SuperCam instrument (Maurice et al., 2021; Wiens et al., 2021). In particular, we have developed a procedure in which three distinct pits are produced on the outcrop surface prior to drilling in the shape of an “L” (∼2.5 mm × ∼1 mm) using 125 laser pulses lasting 4 ns each. The number of pits and their arrangement were chosen to meet four competing priorities. The first priority, addressed by maximizing the number of pits, is that the core surface must be successfully marked despite the typical lateral uncertainty in placement of the Drill. The second priority, addressed by maximizing the depth of each pit, is that the outside surface of the core must retain the marks until it is returned to Earth given the possibility that it could be damaged during or after drilling. The third and fourth priorities, addressed by minimizing both the number of pits and their depths, were to minimize laser damage to the core surface area and core volume and to reduce rover operational complexity and power consumption.

2.5.3 Propagating Orientation Into Core Body

The above description of the core orientation procedure assumes the core is returned to Earth as a solid object that is not internally differentially rotated during or after drilling. Because the only constraint on the state of internal deformation following drilling are from CacheCam images of the core bottom, the overall mechanical states of the drilled cores will be essentially unknown until they are returned to Earth. In fact, coring of a diversity of rock lithologies using the Mars 2020 Qualification Model Dirty Testing (QMDT) testbed at the Jet Propulsion Laboratory (JPL) found that essentially all cores fractured during drilling (see Figure 48 of Moeller et al. (2021)). Well-indurated lithologies with weak foliation (e.g., basalts, basaltic sandstones, and rhyolitic tuffs) typically produced fragments with well-preserved fracture planes (Figure 11). However, softer and more foliated lithologies (e.g., mudstones, hydrated sulfates) commonly produced fragments in the form of wafers or occasionally even pulverized pebbles and granules (Figure 11).

For fragmented cores, the full core orientation (hade, azimuth, and roll) can be propagated into the core from the outer surface imaged by WATSON by rotating the fragments so that they fit together and lock into place like puzzle pieces. Computerized tomography scanning of the core in the sample receiving laboratory prior to extraction (e.g., Ziegler et al., 2021) would be very valuable for documenting the orientation relationships of fragments. The first depth at which pieces cannot be reassembled in this way to their original orientations sets the maximum depth to which the full orientation can be propagated using surface images. This maximum depth is highly dependent on core lithology; images of QMDT test cores show that it ranges from ∼20% to 100% of the core length for the indurated lithologies and ∼0%–40% for the soft and foliated lithologies (Figure 10). It is possible that other features (e.g., layering in the core i.e., not parallel to the core plane) could be used to orient core fragments beyond the maximum depth of propagation achieved with surface images.

2.6 Uncertainties on Core Orientations

We have identified two possible sources of uncertainty in the orientation of the Drill and WATSON: uncertainty in knowledge of the orientation of the Drill and WATSON in the Site frame and wobbling of the Drill during drilling. Although there may be additional uncertainty associated with reorienting the samples after return to Earth, it is currently not possible to quantify this because the sample receiving facility and its procedures are still under development. We discuss each below and then estimate the total uncertainty.

3.1 Summary of Acquired Samples

We determined the orientations for the 20 successful rock cores as well as for the unsuccessful Roubion core. Because the regolith sampling process does not maintain the relative orientations of the acquired grains during sample acquisition (Moeller et al., 2021), we did not attempt to orient their sampling tubes.

3.2 Estimation of Wobbling of Drill

We also estimated the variation of the orientation of the Drill during drilling (see Section 2.6.2). For the first 12 rock cores, we estimated these variations using a comparison of the Drill orientation just after loading the Stabilizers and just prior to commencing drilling versus just after the end of drilling and before unloading the Stabilizers. In all cases, we found that the changes in the hades and azimuths were <0.2° and 0.08°, respectively. For the remaining 8 rock cores, these variations were estimated from these two timesteps along with 20–30 intermediate timesteps, each of which is associated with the acquisition of a Front Hazcam (e.g., Figure 4) image. From the latter metadata, we found that the 3σ variations in the hades and azimuths are only 0.031° and 0.68°, respectfully, across the 8 cores (Figure 12; Movies S1–S8). Because the latter estimates are based on more measurements than the 2-point calculations from the first 12 cores, we report the latter estimates in Table 3.

3.3 Core Orientations

The orientations for the first 20 rock cores taken with the Perseverance rover are shown in Table 1. Annotated WATSON images for each of the rock cores denoting the estimated rolls and directions toward Martian geographic north relative to the coring surfaces are shown for the crater floor (Figure 13), fan front (Figure 14), upper fan (Figure 15), and margin (Figure 16) separately. We found that the natural outcrop surfaces of most coring targets had durable asymmetric markings that could be used for core orientation (Figures 7, 8, and 9c). However, we concluded that the outcrop targeted for 2 cores—Bearwallow and Melyn—lacked such natural markings and so we marked them using SuperCam prior to drilling (Figures 8, 9, and 9b). In addition, the mission baselined SuperCam core top marking for all cores from Pelican Point onward to minimize operational complexity.

As a consistency check on the accuracy of our orientations, we expect that paired samples that are taken from the same outcrop will have similar hades and azimuths. However, we have no similar expectation for the roll as that angle is heavily dependent on the orientation of the WATSON camera rather than on the intrinsic relief of the outcrop. The average difference in hade between paired samples is 4.14°, while the average difference in azimuth is 7.98° excluding the Malay and Robine cores. The Malay and Robine cores, while taken from the same outcrop, were sampled with the rover on different sides of the outcrop leading to a 147.93° difference in their azimuths.