Abstract DGP2026-68

Spectral Characterization of Mercury Analogs based on Machine learning for MERTIS

Nimisha Verma (1), Joern Helbert (2), Mario D'Amore (1), Alessandro Maturilli (1), Oceane Barraud (1), Siddhant Agarwal (1), Lida Fanara (1), Katharine Otto (1), Greta Lamers (3), Akin Domac (1), Aurelie Van den Neucker (1), Giulia Alemanno (1), Harald Heisinger (4), Solmaz Adeli (1)

(1) German Aerospace Center (DLR) – Institute of Space Research, Rutherfordstrasse 2, 12489 Berlin, Germany, (2) ESTEC, European Space Agency, Keplerlaan 1, 2201AZ, Noordwijk ZH, The Netherlands, (3) Department of Earth Sciences, Freie Universität, Malteserstr. 74-100 / Building A, 12249 Berlin, Germany, (4) Institut für Planetologie (IfP), Universität Münster, Wilhelm-KlemmStr. 10, 48149 Münster, Germany

We are developing a machine learning-based mineral identification approach using laboratory emissivity measurements and synthetic spectra generation to support MERTIS observations of Mercury's surface. MESSENGER data revealed that Mercury shows a unique elemental composition – depleted in iron (Fe) but enriched in magnesium (Mg) [1]. However, absence of diagnostic absorption bands in MESSENGER MASCS visible/near-infrared data [2] and limited laboratory emissivity measurements of Mercury analog material [3] have prevented mineral identification. Characterizing Mercury’s mineralogy is crucial to understand its formation and thermal evolution.

MERTIS (Mercury Radiometer and Thermal infrared Spectrometer), onboard the ESA-JAXA mission BepiColombo consists of an infrared spectrometer (TIS) with a spectral range of 7-14 μm and a radiometer (TIR) with a radiometric range of 7-40 μm [4]. The instrument addresses a critical gap in understanding of Mercury’s surface mineral composition.

Emissivity spectra of Mercury analogs are measured at the Planetary Spectroscopy Laboratories (PSL), DLR, Berlin [5]. The Mercury analogs are rocks and minerals with similar compositions to Mercury, as derived from previous studies based on MESSENGER data and models [1] [3] [6], and are measured at Mercury-like dayside temperatures (up to 500°C) [5]. We use a Variational Autoencoder (VAE) neural network to study the emissivity spectra. A Variational Autoencoder (VAE) uses a probabilistic approach to map diagnostic features like Christiansen Features (CF), Reststrahlen Bands (RB) and Transparency Features (TF) into a latent space where each feature is distributed according to its statistical properties. Emissivity spectra measured at different temperatures and grain sizes provide variability for the VAE to understand the shifts in these diagnostic features, preparing the algorithm for the variability expected in MERTIS observations. This enables mineral identification and generation of synthetic spectra to expand our dataset. To ensure the accuracy of these synthetic spectra, we compare the distribution of these key spectral features against the laboratory measurements.

While initial analysis with approximately 70 spectra (pure or mixed minerals) identified dataset limitations that hindered the algorithm’s ability to generalize spectral patterns and generate realistic synthetic spectra [7], we are now focused on expanding the training dataset by adding more emissivity spectra measured in the laboratory (up to 1000 spectra). This will enhance the model’s mineral identification accuracy and synthetic spectra generation capabilities. Once validated, the spectral identification framework with the real and synthetic spectra will serve as a reference point for identifying minerals from data obtained via MERTIS instrument.

References:

[1] L. R. Nittler et.al., "The Major-Element Composition of Mercury’s Surface from MESSENGER X-ray Spectrometry," Science, pp. 1847-1850, 2011.

[2] N. R. Izenberg et.al., "The low-iron, reduced surface of Mercury as seen in Spectral reflectance by MESSENGER," Icarus, vol. 228, pp. 364 - 374, 2013.

[3] A. Maturilli et.al., "Komatiites as Mercury surface analogues: Spectral measurements at PEL," Earth and Planetary Science Letters, vol. 398, pp. 58 - 65, 2014.

[4] H. Hiesinger and J. Helbert, "The Mercury Radiometer and Thermal Infrared Spectrometer (MERTIS) for the BepiColombo mission," Planetary and Space Science, vol. 58, no. 1-2, pp. 144-165, 2010.

[5] A. Maturilli et.al., "Emissivity spectra of analogue materials at Mercury T-P conditions," in LPSC, 2017.

[6] C. Carli et.al., "Laboratory Emissivity Spectra of Sulphide-Bearing Samples, New Constraints for the Surface of Mercury: Oldhamite in Mafic Aggregates," Minerals, vol. 14, no. 1, 2024.

[7] N. Verma et.al., "Spectral Fingerprints of pure and mixed minerals: Laboratory Characterization and ML Integration," in EPSC-DPS 2026, Helsinki, Finland, 2025.