A Fault Surface Exposed on Mercury

Mercury’s tectonic history has been dominated by global contraction as the planet shrinks and cools. Previously, fault dips have been indirectly estimated by measuring offsets to the rims of craters displaced by fault movement. Here we present the first observations of a fault surface exposed in three dimensions, in this case within a volcanic vent. We use shadow measurements to determine the depth of the vent and have measured the dip of the fault as 28 5. This is towards the shallower end of previous dip estimates used to estimate global contraction. If it is representative, it suggests global contraction at the higher end (5.7–7.1 km) of previously published estimates.

A lobate scarp is an asymmetrical ridge that is the surface expression of a thrust fault. They are known to be up to 2 km high on Mercury (Strom et al., 1975;Watters and Nimmo, 2010;. Wrinkle ridges have a similar surface expression to lobate scarps, but tend to be smaller and more symmetrical (Strom et al., 1975;Watters and Nimmo, 2010;Korteniemi et al., 2015). High relief ridges are broad, symmetrical topographic features, interpreted as high angle reverse faults (Dzurisin, 1978). Initially thought to be a distinct tectonic landforms, subsequent observations show high relief ridges can often transition into lobate scarps and thus are part of a continuum of landforms (Melosh and McKinnon, 1988;Watters et al., 2009;Byrne et al., 2014;. Key terminology related to faults in this manuscript includes: the fault plane, which is the surface separating the blocks and along which movement has occurred; strike, the azimuth at which the fault plane intersects the horizontal; and the dip, the angle at which the fault plane descends below the horizontal (Neuendorf et al., 2005).
The stresses in Mercury's crust that have generated this contractional tectonism have multiple possible origins, including global contraction (Watters and Nimmo, 2010), tidal despinning (e.g. Melosh and Dzurisin, 1978), true polar wander (Matsuyama and Nimmo, 2009), and mantle convection (King, 2008). Some of the faulting is localised at the junction between regional geochemical terrains (Galluzzi et al., 2019), or at the edges of smooth plains confined within impact basins (Fegan et al., 2017). Although a combination of factors probably influences many of the visible faults, the dominant cause of contractional tectonics on Mercury is widely accepted to be global contraction of the planet (Klimczak et al., 2015). Mercury's relatively high ratio of surface area to volume means that it has undergone significant cooling, which has led to shrinking (Watters and Nimmo, 2010). Thermal and compositional models of the planet provide a wide range of estimated shrinking; the planet could have shrunk by between 0.6 and 10 km radially (Solomon, 1977;Watters and Nimmo, 2010;Grott et al., 2011). Byrne et al. (2014) calculated Mercury's global contraction from a global inventory of contractional features combined with estimates of fault dips associated with each type of surface expression. The amount of horizontal movement across a thrust fault can be determined from the measured displacement of features on either side, most usefully offset crater rims (Galluzzi et al., 2015). The steeper the average fault dip, the less horizontal movement is accommodated by each fault of a given horizontal offset, and so the lower the amount of global contraction necessary to account for it. The derived reduction in global radius recorded by Mercury's crust is 7.1 km if the average fault dip is 25 , but as little as 4.7 km if the average fault dip is 35 (Byrne et al., 2014). Watters (2021) uses a more conservative number of faults on Mercury to claim a maximum global contraction of only 1-2 km, but this too is based on average fault dips of 25 -35 . Thus, an understanding of the average dip of the faults on Mercury is vital to constrain contraction estimates and therefore to understand Mercury's thermal history. Estimates of fault dips range between 6 and 57 with an average of 25 (Galluzzi et al., 2015(Galluzzi et al., , 2019 were made using the shortening and vertical offset of craters that straddle faults on lobate scarps. Crane (2020) used topographic data and modelling to estimate a dip for 29 faults; all results being less than 20 , with average dips ranging from 3 to 10 . Until now no fault dip on Mercury has been measured by direct means.

A fault's dip revealed
Here we describe the first reported example on Mercury where the local topography exposes a fault scarp sufficiently well that its dip of the fault plane can be measured. Our example comes from the inner wall of a pit interpreted as an explosive volcanic vent. Explosive volcanism is manifest on Mercury through endogenic pits (vents) that are up to tens of kilometres across and several kilometres deep. The vents are often surrounded by faculae: high albedo, spectrally red features that are probably explosive ejecta from the eruptions (Head et al., 2008;Goudge et al., 2014). Unlike impact craters, these vents tend to be non-circular and lack a raised rim, and hence we refer to the break of slope at the edge of a pit as the 'brink' to avoid confusion with the rim of an impact crater. High-resolution studies of these vents suggest that they did not form through caldera collapse (Rothery et al., 2014;Thomas et al., 2014b), in which the emptying of a magma chamber leads to subsidence creating a large volcanic crater. Instead, the vents are interpreted to be excavated by explosive eruption, perhaps analogous to kimberlite pipes on Earth (White and Ross, 2011).
Explosive volcanism peaked during Mercury's Calorian (3.9-~3.5-3.0 Ga) and Mansurian (~3.5-3.0 to~1.0-0.28 Ga) epochs but continued into the Kuiperian (Thomas et al., 2014a;Jozwiak et al., 2018), which may have begun as recently as 280 Ma (Banks et al., 2017). As Mercury's lithosphere has cooled, it has thickened (Peplowski et al., 2011), and contraction hinders volcanism by impeding magma ascent (Solomon, 1978). The formation of volcanic vents on Mercury therefore probably requires crustal weakness to aid continued rising of the magmas (Klimczak et al., 2018), which are generally presumed to be volatile-rich to account for the explosivity of these eruptions (e.g. Kerber et al., 2009;Weider et al., 2016). Possible crustal weaknesses include fractures associated with impact craters, and faulting. Vents are found both within impact craters and on lobate scarps (Thomas et al., 2014b;Goudge et al., 2014). There is a significant correlation between impact craters and vents, but the high density of tectonic features on Mercury means it is not possible to distinguish from random whether there are also tectonic controls on the locations of vents (Klimczak et al., 2018).
During the process of reviewing volcanic vents on Mercury (Pegg et al., 2020) we identified several vents that occur directly on a tectonic fault (Fig. 1). The fault scarp is not usually visible within such vents. There are many factors that could contribute to this, such as: the vent post-dating the fault movement, space weathering, volcanic activity or rockfalls covering the scarp, poor illumination and/or low image resolution, lack of shadows or stereo pairs. However we have identified one vent within Vazov crater (147.6 o E, À65.6 N) that shows what we interpret as a small fault scarp visible beyond the north side of the vent and expressed also on the internal vent wall and on the vent floor (Fig. 2). This situation enables us to reconstruct the geometry of the fault plane from the fault's intersection with the topographic surface.
Outside of the vent on the crater floor there is a contractional ridge 3 km wide, 30 km long, which strikes NW-SW. The topographic feature continues as a ridge NW of the crater. Because it is expressed by a symmetrical ridge, the facing direction of the fault would be uncertain if that were the only evidence, but fortunately in this case the scarp of the NE dipping fault is apparent inside the vent. The section revealed in the vent wall shows that the main fault dips north. The fault trace is visible on the west wall of the vent but is hidden by the vent-wall shadow in the bottom and east of the vent (Fig. 2a, c).

Geological setting
To study the fault, we used images from the Narrow-Angle Camera (NAC) onboard MErcury Surface, Space ENvironment, GEochemistry, and Ranging (MESSENGER) spacecraft (Solomon et al., 2001). The vent's location in the southern hemisphere means no laser altimeter data are available. We processed the image data using the USGS's Integrated Software for Imagers and Spectrometers (v 3.5.1.7995) and produced a geologic map of the vent site based on image EN0231657081M (resolution of 163 m/pixel) using ESRI's ArcGIS software. We chose a sinusoidal projection centred on the vent. We used the mapping methods and units outlined by (Galluzzi et al., 2016) following USGS (Tanaka et al., 2011) and Planmap (Rothery et al., 2018) guidelines and classified crater degradation state using (Kinczyk et al., 2020).
The geological map (Fig. 2) and image data show that the faulted vent is one of several vents within and around a degraded impact crater. Beyond the vent, on the crater floor, the fault is clearly marked by the SW side of a 4 km wide ridge. A slope on the north side of the ridge, may represent an antithetic fault on the hanging wall (Fig. 3), as is common in reverse faulting, and indeed is possibly the case with high relief ridges on Mercury (Strom et al., 1975). Within the vent itself the main fault is clearly visible where it cuts obliquely down the western internal wall of the vent and across its floor, but the antithetic fault is hard to make out. In the east of the vent, the location of the fault rupture is in shadow, and cannot be discerned, even by increasing the image 'stretch', but it does appear to re-emerge onto the impact crater floor east of the vent. The fault structure appears to continue beyond the vent-hosting impact crater to the northwest but is less apparent there than within the crater.
The presence of the vent within the impact crater shows that the vent postdates the formation of the impact crater. The crater degradation state corresponds to the 'partly-degraded' C2 class of the Kinczyk et al. (2020), suggesting that it formed during the Tolstojan period (~3.9-4.0 Ga); however, we could not perform a reliable crater size-frequency analysis due to the small area involved. The fault cuts across the floor of the crater, showing that it was active after the crater was formed (though could have existed before).
The timing of activity on the fault relative to the vent is also unclear. The continuity along strike of the fault on both sides of the vent might suggest the fault formed before the vent, in which case the vent-forming eruption could then have exploited the fault structure as a weakness enabling it to reach the surface, though the presence of other vents in the same crater shows this is not essential. The visibility of the fault trace on the vent wall allows us to measure the fault plane, as a result of its present-day topographic expression inside the vent wall. This could have formed during the excavation of the vent, with the excavation occurring preferentially within the fractured rock in the fault surface, leaving some of the hanging wall behind and forming a slight step. Other possible interpretations are that there has been post-eruption movement on a preexisting fault to produce the step, or that the fault could have been initiated entirely after the vent formation.

Measuring the fault dip
We produced a digital terrain model (DTM) for the vent site using the AMES stereo pipeline (Broxton and Edwards, 2008) using images: EN0231657081M and EN0231356027M. We could not provide vertical control for this DTM due to the lack of laser altimeter data, so we could not eliminate the possibility that systematic tilts related to uncertainties in the spacecraft pointing remained in these data. Thus, we did not use these data to perform dip measurements. Instead, we used shadow measurements on images EN0246250635M (176 m/pixel) and EN0261431434M (197 m/pixel). These are the highest resolution close to nadir images of this feature that MESSENGER obtained. Although we examined off-nadir images as well, we elected not to attempt to use these via the method of Barnouin et al. (2012) as we would have had to correct both the shadow and horizontal measurements, and this would likely compound any errors from any uncertainty in spacecraft pointing.
As shown in Fig. 4, we constructed a line propagating the fault strike from the NW over the vent. Then we produced a second line (x) orthogonal to the strike to where the shadow from the vent intersects the fault trace. Finally, we constructed a line to record the length of the shadow from the intersection of the fault trace with the vent shadow, to the brink of the vent. We drew this line orthogonal to the shadow to get the shadow length at the point where the shadow intersects the fault trace. The incidence angle could then be used to calculate the depth of the shadow (and thus the depth of the fault trace) below the vent brink at this point (ΔZ). This could then be combined with x to get the dip angle.
For each image we repeated the measurements ten times to reduce measurement uncertainty from digitisation. We calculated the depth of vent at the intersection of the fault and the shadow using basic shadow measurement techniques e.g. (Pike, 1974;Chappelow and Sharpton, 2002) and then calculated the dip. Unfortunately, there are no markers that enable us to estimate the amount of displacement on the fault.
Our estimates for the fault plane dip from the two images are shown in a box and whisker plot (Fig. 5). The means for the two images suggests a mean dip of 28 AE5 . Uncertainties associated with this method arise from (i) possible uncertainties in spacecraft pointing, (ii) measurement error, and (iii) image resolution limits. As a test of the pointing issues, we used two nadir images to make measurements, the comparison between two images acquired at different times allowing us to test whether there was a pointing error offset. The dip measured in EN0246250635M has a mean of 27.8 , and median of 27.4 . The dip measured in EN0261431434M has a mean of 27.7 and a median of 27.5 . The similarity of the two measurements provides confidence in the technique. By making multiple repeat measurements we tested the repeatability of human operators to make the measurement and interpret the location of the strike and shadow intersection. This provides a measurement of the measurement error. In addition, we consider that no line can be drawn more accurately than 1-pixel width. A 1-pixel error would result in the equivalent of þ5.0/-4.7 error in dip for image EN0246250635M and þ5.4/-5.0 error in dip for EN0261431434M, (Fig. 5).

Discussion
With a dip of 28 AE 5 , this fault is slightly steeper than the mean value used by Galluzzi et al. (2019Galluzzi et al. ( , 2015, but within the range of measured dips in their database. The upper end of the error for the dip measurement is consistent with the 30 predicted to be typical by the Anderson faulting model for a standard thrust fault (Anderson, 1905).
The scale of this fault is much smaller than that of Mercury's most characteristic lobate scarps and its surface expression is similar to high relief ridges end members of the lobate scarphigh relief ridge continuum, and its dip suggests that it is a similar compressive feature to those larger examples. This dip measurement is greater than the 3-10 estimated by Crane (2020), however those are based on features within Smooth Plains units, which are likely layered (Denevi et al., 2009) and so prone to shallow d ecollements (Crane and Klimczak, 2019) that would tend to encourage gentler dipping structures. The dips of faults are known on parts of Mercury to be affected by local geochemical conditions (Galluzzi et al., 2019), however the location of this faulted vent in the southern hemisphere means the geochemical data from MESSENGER is to low resolution to make inferences (Nittler et al., 2020).
While measurement of a single fault is not conclusive, our data does lend credence to the view that, on Mercury, moderately dipping faults are more typical than steeper or very shallow ones. Additionally, this work demonstrates that measurements such as these are possible and can be made even in the absence of DEM data. The image resolution does not allow measurement of fault displacement, which would enable quantification of how much strain the fault has accommodated, but higherresolution imaging data from BepiColombo  may allow this. BepiColombo should provide sufficient data to make more measurements like this, for example in occurrences such as the example in Fig. 1a, generating a database of the fault geometries in locations across Mercury and allowing a more detailed assessment of global tectonic and thermal models. If measurements of other faults prove possible and are in line with this example, generalizing such a result to the model of Byrne et al. (2014) would suggest that global contraction is likely to be about 5.7-7.1 km, which at the higher end of the currently estimated range.

Conclusions
Direct measurements of the dip of a fault on Mercury find a value of 28 AE 5. This is within the range of fault dips calculated indirectly through measurements of crater displacement and for classical thrust fault models. Although only a single example, this dip estimate agrees with the growing evidence to suggests that the higher-end estimates of radial shrinking by global contraction are more representative than lower estimates. BepiColombo will provide an opportunity to gain better data on this site and identify other locations where the volcanic vents expose fault surfaces in 3 dimensions, offering fuller insight into planetary tectonics.

Author Statement
The research was undertaken primarily by David Pegg who first identified the feature and made the measurement and wrote the paper. Professor David Rothery, Dr Susan Conway, and Professor Matthew Balme supervised the project and provided input into developing the methodology and review & editing the paper.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.