Crater ejecta as a clue to investigating Mercury’s subsurface composition

Introduction

Geochemical data from the MESSENGER X-Ray Spectrometer (XRS) [1] suggest heterogeneity in Mercury’s mantle composition and volcanic history [e.g., 2]. While vertical compositional variability within the crust may hint at temporal changes in magma sources, subsurface geochemical structure remains poorly constrained from surface mapping alone. To solve this problem, impact cratering offers a window into the subsurface by excavating and redistributing deep materials.

Given that surface roughness is a proxy of surface age, fresh craters with rough ejecta may provide clues to the subsurface composition as they are often associated with anomalies in Mg/Si and Al/Si ratios [3]. For example, the rough ejecta regions around Rustaveli and Rachmaninoff craters show elevated Mg/Si and depressed Al/Si relative to the surroundings (Figure 1). Although lateral regolith mixing can obscure excavated material, this effect would be minimal at these fresh craters, making them suitable for subsurface composition estimates when combined with crater formation modeling.

In this study, we compare XRS observations with surface composition models that incorporate impact simulations to predict the subsurface material ransport. Based on the comparison, we constrain the subsurface composition and its relation to Mercury's volcanic timeline.

Figure 1. Summary of surface roughness and composition maps around (a-c) Rustaveli crater and  (d-f) Rachmaninoff crater: (a, d) Surface roughness maps [3], (b, e) Mg/Si ratio maps [4], and (c, f) Al/Si ratio maps [4]. Red dashed lines denote the crater rims.

 

Method

Our surface composition model considers impact cratering using the iSALE-2D shock physics code [5–7]. As Mercury is thought to have experienced global resurfacing by basaltic volcanism [e.g., 8], we use basalt ANEOS for the equation of state of the crust. The projectile is assumed as a dunite sphere colliding at the mean impact speed on Mercury.

Based on the iSALE simulation results, we next model surface composition distribution after the excavation and deposition of subsurface materials (Figure 2). Assuming vertical Mg/Si and Al/Si profiles, the material distribution is converted into surface composition maps. For simplicity, we adopt a two-layer model for subsurface composition structure, varying upper-layer thickness and lower-layer composition. The upper-layer composition is fixed to the average composition derived from XRS data beyond three crater radii from the crater centers. The modeled maps of Mg/Si and Al/Si ratios are then spatially averaged within each XRS footprint polygon and compared with the corresponding XRS observation [4].

Figure 2. Example of iSALE simulation results. (a) Pre-impact tracer locations. (b) Post-impact tracer locations. (c) Fraction of initial depth of materials located shallower than 1-km depth. The colors correspond to the material depth in the pre-impact phase.

 

Results and Discussion

Figure 3 compares modeled and observed Mg/Si and Al/Si ratios for Rustaveli crater. Despite the large uncertainties of XRS measurements (black error bars in Figure 3-a and b), the general radial trends in the observations are reproduced with our iSALE-based models using specific combinations of the upper-layer thickness and lower-layer composition. For example, assuming the upper-layer thickness of 5 km, the Mg/Si and Al/Si ratios of the lower layer needs to be ~0.5 and ~0.2, respectively.

To find the best-fit parameter sets, the root-mean-square of differences between data and models are computed for all parameter sets Figure 3-c and d). As the mixing ratio of subsurface materials decreases with increasing origin depth, the best-fit composition of the lower layer depends on the assumed upper-layer thickness. We find that a lower-layer composition with Mg/Si > 0.5 and Al/Si < 0.2 best reproduces XRS observations.

The best-fit composition of the lower layer is consistent with that of the cratered terrain, rather than northern plains [e.g., 2], suggesting that similar crustal materials are hidden beneath the northern smooth plains. Given that Mercury’s crust has formed through multi-phase volcanism, each lava layer likely represents the volcanic conditions. This result implies that volcanic materials on Mercury have less heterogeneity than that seen on the surface and might have originated from similar magma source before the formation of the northern plains.

The model comparison for Rachmaninoff crater also shows similar ranges for lower-layer composition. As similar trends of Mg/Si and Al/Si ratios are observed at other fresh craters, such as Stieglitz and Tung Yüan, the combination of X-ray observations with impact simulations may reveal subsurface structures at other locations on Mercury. In this presentation, we will also report results from other craters for further insights into magmatism timeline on Mercury.

Figure 3. Comparison between iSALE-based model and XRS data for Rustaveli crater. (a, b) Mg/Si and Al/Si ratios of all footprints over distances from the crater center. The x- and y-error bars represent the footprint coverage and XRS measurement uncertainties [4], respectively. The colored points show the modeled Mg/Si and Al/Si ratios, varying lower-layer composition with an upper-layer thickness of 5 km. (c, d) Differences in Mg/Si and Al/Si ratios between XRS data and models. The differences are normalized by the standard deviation of XRS data beyond three crater radii. The black dashed line represents the assumed upper-layer composition. The cyan dotted line shows the best-fit lower-layer composition for each assumed boundary depth. The colored stars are parameter sets shown in (a) and (b).

 

Acknowledgement

We gratefully acknowledge the developers of iSALE‐2D (https://isale‐code.github.io), including Kai Wünnemann, Dirk Elbeshausen, Boris Ivanov, and Jay Melosh. We used pySALEplot to analyze the output file of iSALE and thank Tom Davison for the development of pySALEPlot.

 

References

[1] Schlemm II et al., 2007, SSR, 131, 393–415.

[2] Namur et al., 2016, EPSL, 439, 117–128.

[3] Nishiyama et al., 2026, PSJ, 7(3), 59.

[4] Nittler et al., 2020, Icarus, 345, 113716.

[5] Amsden et al., 1980, Los Alamos National Laboratories Report, LA‐8095, (p. 101).

[6] Ivanov et al., 1997, International Journal of Impact Engineering, 20(1–5), 411–430.

[7] Wünnemann et al., 2006, Icarus, 180(2), 514–527.

[8] Marchi et al., 2013, Nature, 499, 59.