Development of a technique to identify µm-sized organic matter in asteroidal material - An approach using machine learning

Aug 5, 2026

Supplementary study material describing a machine-learning workflow for identifying micron-sized organic matter (OM) in asteroidal material, using carbonaceous chondrites and a Ryugu sample. The document covers sample preparation with ultra-microtoming, acquisition of BSE images and carbon element maps under defined SEM conditions, and the need to separate OM signal from abundant noise. It details image mosaicking, signal enhancement via maxpooling-style binning, and noise filtering to improve OM map generation.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages24

Reading Time1 h 0 min

AccessFree to download

Contents

Sample Preparation

Analytical Condition

Image Processing

Stitching images to have a mosaic

Signal enhancement

Removal of isolated pixels

Removal of signal from non-OM phases

Optimization of the image processing

Principal of algorithms on machine learning and classification

Properties of instances on Orgueil, Murray, and Ryugu

References

Supplementary tables

Supplementary figures

Supplementary materials Development of a technique to identify µm-sized organic matter in asteroidal material: An approach using machine learning Rahul Kumara*, Katsura Kobayashi a, Christian Potiszil a, and Tak Kunihiro a a The Pheasant Memorial Laboratory, Institute for Planetary Materials, Okayama University, Yamada 827, Misasa, Tottori 682-0193, Japan. *corresponding author [email:](email: kumar_rahul@s.okayama-u.ac.jp)[ kumar_rahul@s.okayama-u.ac.jp](email: kumar_rahul@s.okayama-u.ac.jp) Contents 1 Sample Preparation .................................................................................................................. 2 2 Analytical Condition................................................................................................................ 2 3 Image Processing ..................................................................................................................... 2 3.1 Stitching images to have a mosaic.................................................................................. 2 3.2 Signal enhancement ........................................................................................................ 2 3.3 Removal of isolated pixels .............................................................................................. 3 3.4 Removal of signal from non-OM phases ........................................................................ 4 4 Optimization of the image processing ..................................................................................... 5 5 Principal of algorithms on machine learning and classification.............................................. 6 6 Properties of instances on Orgueil, Murray, and Ryugu ......................................................... 6 7 References ............................................................................................................................... 8 8 Supplementary tables............................................................................................................... 9 9 Supplementary figures ........................................................................................................... 16 1 Sample Preparation In this study two carbonaceous chondrites (CC) and one Ryugu sample were surveyed. The CC samples included, the Orgueil (CI) and the Murray (CM) chondrites. In this work one Ryugu sample from‘chamber A’ was chosen. All of the samples were fixed within a steel cup via embedding them in indium pool, as described in the methods section of Nakamura et al.,(2022) . Each sample was then ultra-microtomed without using any fluid or further polishing methods, in order to avoid removing any SOM (Sephton et al., 2003) or imparting any physical or chemical change to the OM (Kerridge, 1983) . A diamond knife was used to perform the microtoming and prepare a flat surface for each sample. Finally, the 4 mm-diameter steel cups, containing the samples, were placed within a 1-inch diameter aluminum disk, as shown in Figure S1a. Figure S1b shows an optical image of the Orgueil that is polished by microtomy. 2 Analytical Condition Back-scattered electron (BSE) images and element maps were collected using a JEOLJSM-7001F scanning-electron microscope equipped with an Oxford Instruments energy dispersive spectroscope. A 15 kV accelerating voltage and a 3 nA beam current with a pixel dwell time of 100 µs were applied to scan area of a given sample surface, at magnification of x2000 . No coating was applied for surface although typically surface should be coated by C to prevent charging. Abundant C in matrix on CC and Ryugu makes sample surface conductive on operational condition described above. The dimension of each image in a single acquisition is 2048x1536 pixels, in which 1 pixel is equivalent to 0.03 µm. The BSE images and element maps were exported as 16-bit grayscale TIFF images. C element map acquired from CC and Ryugu include b

FAQs

What samples and preparation steps are used for the OM identification study?

Two carbonaceous chondrites (Orgueil and Murray) and one Ryugu sample from chamber A are embedded in indium and ultra-microtomed to avoid removing SOM or altering OM. Samples are mounted in steel cups for subsequent SEM analysis.

Which microscopy conditions are applied to collect the imaging and element data?

Back-scattered electron images and carbon element maps are collected with a JEOL JSM-7001F SEM equipped with an energy-dispersive spectroscope. A 15 kV accelerating voltage and 3 nA beam current are used, with a defined dwell time and a pixel scale of 0.03 µm per pixel.

Why is image processing necessary before identifying micron-sized OM?

Carbon element maps include both OM-related signal and abundant noise, making small (sub-10 µm) OM difficult to distinguish from false positives or misses. The workflow enhances signal-to-noise ratio and filters noise to generate reliable OM maps.

Aug 5, 2026

Upload by Theodora