Description

The mineralogy of Mercury’s surface is not well constrained. The absence of absorption bands in the visible and near-infrared spectra acquired both by Earth-based telescopes and spacecrafts close to the planet prevents a clear identification of its constituents, although 1) models based on elemental and geochemical composition studies predicted the probable composition of Mercury’s surface (mainly plagioclase, iron-poor pyroxene, and olivine), and 2) recent works revealed the possible detection of weak spectral features in the visible range of local fresh material (e.g., hollows). To evidence this, Lucchetti et al. 2018 performed a k-mean clustering analysis on MDIS-WAC images of several crater floors presenting hollows. Our project aims to have a complementary approach and to serve as the development of a method which could be applied in the future to BepiColombo data. While k-mean clustering helps to classify and spot regions with similar spectral properties, our approach intends to identify features or terrains in MDIS images from spectra acquired in the laboratory using the Spectral Angle Mapper tool. Further, we plan to implement machine-learning tools to lead to a more systematic and trustworthy method.

_______________________

________________________________________________________________________________________________________________________________________________________________

Research highlights

​​​​As a first test, we decided to focus on the Dominici crater (1.38°N - 36.5°W), which hosts large hollows on its floor. Lucchetti et al. 2018 used a k-mean clustering analysis on MDIS-WAC data of Dominici and were able to detect the hollows thanks to their characteristic large, weak absorption around 0.7 µm. To explain this feature, they hypothesized the presence of pyroxenes presenting transitional elements like Cr, Ni and Ti in substitution of Fe and Mg atoms in the crystal lattice. We wanted to test this hypothesis by trying to identify, in the same data, pixels likely containing these specific minerals and see if these pixels indeed correspond to hollows. We used spectra of 17 sulfides measured in the laboratory (e.g., see Figure 1.d.) as inputs of another classification method called Spectral Angle Mapper (SAM). Our results (see Figure 1.c.) are then compared to those obtained with the k-mean clustering (see Figure 1.a-b.). The use of sulfides spectra is motivated by their believed involvement in the formation process of hollows.

Both classifications are in very good agreement. In particular, the SAM classification retrieves the morphological shape of Dominici hollows, as does the k-mean classification. It is very interesting to note that most of the pixels in the SAM classification are classified thanks to the 3 spectra of fresh sulfides, which is consistent with the young nature of hollows. In addition, these sulfides are those proposed by Barraud et al. 2020 for the possible composition of hollows.

Meetings of the projects

Date	Topic	Minutes
February 7th 2023	Update of project objectives (online)
February 10th 2021	Update on progress of method testing (online)
Octobre 22nd 2020	Update on progress of method testing
September 30th 2020	Update on project method, discussion of starting point (online)
September 14th 2020	Brainstorming, building project method (online)

​​

Publications/Abstracts related to the project

​​​​​​