Explosive volcanism in complex impact craters on Mercury and the Moon : influence of tectonic regime on depth of magmatic intrusion

9 Vents and deposits attributed to explosive volcanism occur within numerous impact craters on both 10 the Moon and Mercury. Given the similarities between the two bodies it is probable that similar 11 processes control this spatial association on both. However, the precise morphology and localization 12 of the activity differs on the two bodies, indicating that the nature of structures beneath impact craters 13 and/or volcanic activity may also be different. To explore this, we analyze sites of explosive 14 volcanism within complex impact craters on the Moon and Mercury, comparing the scale and 15 localization of volcanic activity and evidence for post-formation modification of the host crater. We 16 show that the scale of vents and deposits is consistently greater on Mercury than on the Moon, 17 indicating greater eruption energy, powered by a higher concentration of volatiles. Additionally, while 18 the floors of lunar craters hosting explosive volcanism are commonly fractured, those on Mercury are 19 not. The most probable explanation for these differences is that the state of regional compression 20 acting on Mercury’s crust through most of the planet’s history results in deeper magma storage 21 beneath craters on Mercury than on the Moon. The probable role of the regional stress regime in 22 dictating the depth of intrusion on Mercury suggests that it may also play a role in the depth of sub23 crater intrusion on the Moon and on other planetary bodies. Examples on the Moon (and also on 24 Mars) commonly occur at locations where flexural extension may facilitate shallower intrusion than 25 would be driven by the buoyancy of the magma alone. 26


33
It has long been recognized that vents and deposits attributed to explosive volcanism frequently occur 34 within complex impact craters on the Moon [e.g., Schultz, 1976;Head and Wilson, 1979 In the lunar case, it has been proposed that localization of explosive volcanism within impact craters 43 results from density-trapping of magma in the brecciated zone below the crater [Head and Wilson, 44 1979]. In this model, a vertically-propagating dike encounters the low density, weak material of the 45 breccia lens beneath the crater floor and is diverted to form a sill because the density and rigidity 46 contrast favors lateral propagation rather than continued vertical ascent [Schultz, 1976;Wichman and 47 Lunar pyroclastic deposits are commonly identified by their low albedo relative to highlands material 105 and a spectral character suggesting varying mixtures of highlands, basaltic and glass components 106 [Gaddis et al., 2003]. We identified the extent of putative deposits on the basis of a low-albedo,

118
For both bodies, we digitized the areal extent of the spectral anomaly, taking a conservative approach 119 by excluding the tenuous outer fringe. This was further refined in lunar examples where the extent of 120 the low albedo material is apparent as fine-grained material mantling the underlying terrain in high-121 resolution narrow-angle camera (NAC) images from the Lunar Reconnaissance Orbiter Camera 122 (LROC). As a means of calculating the maximum specific energy with which particles were ejected from vents, we additionally measured the maximum distance between a candidate vent (Section 2.3) 124 and the outer margin of its surrounding continuous deposit at each site. Because the available data 125 types and the spectral character of deposits differ on the two bodies, the same level of error cannot be 126 assumed in determination of the position of the outer boundary of the deposit. We estimated it as 2 127 pixels, but it may be higher, particularly on Mercury where there are no high-resolution images with 128 which the position of this outer boundary can be refined. This introduces a bias in favor of larger 129 detected deposits on the Moon. Comparisons of deposit areal extent on the two bodies are therefore 130 made with caution. 131

Volcanic vents 132
On Mercury, irregular, rimless depressions lacking the characteristic ejecta blanket of impact craters 133 (known as 'pits') are considered candidate volcanic vents [Kerber et al., 2011]. These are readily 134 identifiable in monochrome orbital imagery taken by the NAC and WAC in MESSENGER's Mercury 135 Dual Imaging System (MDIS) (Figure 2a-b). We obtained topographic data with which to determine 136 the volume of these vents by using stereo images (NAC or WAC frames using the 750 nm filter) to 137 create high-resolution DEMs by photogrammetry using the Ames Stereo Pipeline [Moratto et al., 138 2010]. Point data were averaged on a 3x3 block of pixels, giving the DEM a horizontal resolution 3 139 times larger than that of the stereo images used to create it. On the basis of error determinations made 140 by Thomas et al. [2014b], the vertical error is up to 80 m. 141 We identified candidate lunar vents by reference to the LROC WAC Global mosaic at 100 m/pixel,

153
Volcanic vents commonly form by erosion of wall-rock during eruption and/or by collapse into an 154 evacuated magma chamber. Therefore the volume of the vent can indicate the energy or volume of 155 eruption. In order to calculate the volume of material that was lost to form the identified vents, we 156 calculated their volume below a rim elevation determined with reference to orbital imagery and 157 topographic products. On both bodies, though to a greater degree in floor-fractured craters on the 158 Moon, the original surface prior to vent-formation was uneven. To account for this when calculating 159 the volume lost to form the vent, we used a Natural Neighbor technique within ArcGIS software to 160 interpolate a surface at the vent rim level on the basis of the surrounding topography, and subtracted 161 elevations on the vent floor from the elevation of that surface. Because this interpolation technique 162 estimates elevation values on a local basis, any relief owing to a pre-existing graben crossing the vent 163 is greatest at the margins of the interpolated area and reduces towards the interior. This means that the original graben volume is only partially accounted for, and the calculated volume of vents within 165 grabens should be viewed as a maximum value. 166

Host crater dimensions 167
The intrusion of magma beneath impact craters on the Moon is proposed to result in a reduction in 168 crater depth [Schultz, 1976]. To explore this, we calculated the host crater depth for all sites in the two 169 samples, defined as the vertical distance between the average rim crest elevation and the average floor 170 elevation. In finding the average rim elevation, we excluded parts of the rim crest where major post-171 formation modification was evident. The average crater floor elevation was defined as the 100 m bin 172 within which the highest number of DEM pixels in the interior of the crater fell. We compared the 173 depth thus calculated to the depth calculated using depth-diameter relationships observed in large 174 populations of mature complex craters [Pike, 1980[Pike, , 1988. infilling.

Regional setting 191
To assess possible regional controls on the occurrence of explosive volcanism, we studied the 192 geological setting of each site in detail. This included noting the proximity to and spatial relationship 193 with extensive lava plains, association with specific substrates and types of tectonic structure, and

212
The maximum ballistic range measured for particles forming the observed deposit is generally higher 213 on Mercury (median value of 18.6 ± 1.2 km, maximum of 50.3 ± 1.2 km) than on the Moon (median 214 10.7 ± 0.04 km, maximum 46.6 ± 0.04 km) despite the observational bias in favor of detection of 215 pyroclastic material to greater distances on the Moon and despite higher gravity on Mercury, which 216 means that particles ejected at equal velocity will have a smaller range than on the Moon. Because 217 lunar vents commonly occur as a relatively subtle widening of a graben, it is probable that in some 218 cases particle sources have been missed and the ballistic range overestimated. We therefore also 219 compare the average geodetic area of deposits within our sample sets. This, too is larger for Mercury 220 (median 1210 ± 53.2 km 2 , maximum 6990 ± 138 km 2 ) than for the Moon (median 231 ± 5km 2 , 221 maximum 3949 ± 22 km 2 ), supporting the inference that particles were, on average, ejected to greater 222 distances on Mercury. The maximum ballistic range (X) can be used to calculate the maximum speed 223 (v) at which pyroclasts were ejected from a vent in a vacuum using the relationship: 224 where g is gravitational acceleration and θ is the angle at which dispersal is greatest (45°). This gives 226 a value of 284 m s -1 for the median and 468 m s -1 for the greatest ballistic range observed in the 227 Mercury sample set, and 143 m s -1 for the median and 297 m s -1 for the greatest ballistic range 228 observed in the lunar sample set. As the specific energy of particle ejection is approximately 229 proportional to the volatile mass fraction in the released magma [Wilson, 1980]

Association with regional geological units and tectonic structures 288
Craters hosting pyroclastic deposits in the lunar sample set commonly superpose, are adjacent to, or 289 are in areas annular to extensive basin-filling mare deposits. The distance to the edge of a major mare 290 deposit ranges up to 340 km, with a mean distance of 90 km. Conversely, sites on Mercury are not 291 commonly adjacent to morphologically young large-scale lava plains, which range from 90 to 1540 292 km distant, 800 km on average (Figure 6). depth to which such craters excavate can be estimated as > 15 km [Croft, 1985], indicating that this 300 substrate is present to considerable depth. At three of the sampled sites the crater also hosts hollows,

Implications for sub-crater magma storage on Mercury 334
The high incidence of floor-fracturing in complex craters hosting pyroclastic deposits on the Moon 335 and its absence at such sites on Mercury requires explanation. Floor-fracturing on the Moon is 336 proposed to occur due to sub-crater magmatic intrusion. An alternative hypothesis, that it occurs due 337 to viscous relaxation [Hall et al., 1981], has been found to be inconsistent with the geometry and

356
One possible contributing factor to a lack of surface deformation in response to a subsurface intrusion 357 on Mercury is that the overburden is stronger than on the Moon. This could result from more 358 voluminous impact melt [Grieve and Cintala, 1997] or less porosity [Collins, 2014] due to higher 359 impact velocity and gravity, or from infilling by massive lavas prior to the proposed explosive 360 volcanic activity. Numerical and physical modeling is necessary to determine the degree to which 361 these factors could affect the bulk strength of sub-crater-floor materials, though the differences would 362 need to be large if they were to account for the total lack of surface deformation seen in host craters 363 on Mercury. 364 The major factor governing surface deformation above a magma body is the depth of intrusion. sampled sites on Mercury occur [Weider et al., 2015], supporting the inference from spectral data that 381 these heavily-cratered surfaces may simply be ancient volcanic plains [Murchie et al., 2015]. This 382 suggests that, contrary to deeper magma storage being favored, hot, volatile-bearing Hermean 383 magmas are expected to be so buoyant that effusive eruption will occur without a period of sub-384 surface storage, except where the crust has anomalously low density. Thus in addition to the evidence 385 presented here for deeper magma storage beneath impact craters on Mercury than on the Moon, the 386 additional problem arises that the observed frequent occurrence of volcanic activity within impact 387 craters [Thomas et al., 2014b], where ascent should be least favored (due to underlying low-density 388 breccia), is the opposite of what is expected on the grounds of magma buoyancy. 389 However, the above applies only if an LNB is reached, whereas there is abundant evidence [e.g., 390 Takada, 1989] that it is rarely reached in nature. The level of magma rise is commonly controlled by the presence of rheological or rigidity contrasts in the overburden [Menand, 2011]; indeed the rigidity

471
A comparison of the scale of vents and surrounding deposits attributable to pyroclastic volcanism 472 within complex impact craters on the Moon and Mercury indicates that eruptions had a significantly 473 higher average energy on Mercury. On the Moon, this activity commonly occurs in craters with 474 uplifted, fractured floors, but no such deformation is detected in host craters on Mercury. This 475 evidence is most consistent with deeper magma storage prior to eruption on Mercury, in a magma 476 chamber inhibited from upwards rupture by regional compression. Once stalled in such a reservoir, 477 the eventual upward propagation of magma that results in a high-energy eruption is likely to have 478 been promoted by concentration of volatiles by fractional crystallization and/or by incorporation of 479 volatiles from wall rock. 480 The comparison with Mercury indicates that the absence of regional compressive stress was important 481 in allowing shallow intrusions to form on the Moon. Further, because lunar FFCs are most common in 482 circum-mare basin regions, which have been in flexural extension for much of their history due to the 483 mare load, it is possible that it is not only the absence of compression but the action of extensional 484 stresses that favored shallow intrusion in these craters. The concentration of FFCs on Mars in zones 485 that have undergone long-term regional extension is supportive of this hypothesis, and suggests that 486 crustal extension may play a controlling role in the formation of floor-fractured craters on terrestrial 487 bodies in general. 488