Optimized multi-microbeam analytical techniques for rapid and accurate identification of lunar minerals: insights from Chang'e-5 basaltic clasts

Abstract

Lunar soils are rare but have great importance for unraveling the evolutionary history of the Moon. The identification of various minerals in lunar soil is vital to explore petrology, geochronology and isotope chemistry of the Moon. However, the rapid and accurate identification of lunar minerals remains technically challenging. Although an automated mineral identification (AMI) method was initially developed to perform mineral identification, the accuracy and analytical conditions of the AMI method are not well-defined. Other microbeam methods, such as electron probe microanalysis (EPMA), Raman spectroscopy (RS) and transmission electron microscopy (TEM), have also been used to identify minerals with various sizes. The applicability of these analytical methods is unclear due to complex mineral phases and diverse grain sizes. In this paper, three Chang'e-5 lunar basaltic clasts were systematically investigated using AMI, EPMA, RS and TEM techniques. Monte Carlo simulation on various minerals was conducted for the first time to optimize the analytical conditions (e.g. accelerating voltage and step size) of the AMI method, which could promote the quick identification of specific lunar minerals. By comparing the minerals identified by AMI, EPMA and RS, the reliability of the AMI method was well validated. Finally, based on the characteristics and applicability of the four methods, an AMI-RS-TEM technical route for identification of lunar minerals was established. This study provides an optimal method for the rapid and accurate identification of lunar minerals, and could also offer valuable insights into mineral identification in other extraterrestrial precious samples (e.g. asteroidal materials and meteorite samples).

---

1. Introduction

Mineralogical and petrological studies of lunar soils are important to unravel the geological evolution on the Moon.^1,2 Concretely, the compositions, structures and isotope ratios of minerals can reflect the physical and chemical conditions and times at which lunar rocks formed.^3–6 For example, Pb isotopic analyses of baddeleyite and zirconolite in Chang'e-5 (CE-5) lunar basalts yielded the youngest crystallization age of ∼2 Ga, indicating prolonged magmatic activity on the Moon.⁷ Some minor lunar minerals (such as pentlandite) could even record unique geological information and provide constraints on lunar evolution processes.^3,8 In addition, lunar minerals are also valuable materials that could drive the mining and utilization of resources on the Moon.³

In general, lunar soils are fine-grained clasts with sizes varying from 4.8 to 432 μm.^9,10 The rapid and accurate identification of mineral species in lunar soils is crucial for further studies on chemical compositions, isotopic characteristics (e.g., Rb–Sr, Sm–Nd and U–Pb), volatile components and so on.^11–16 Recently, scanning electron microscopy (SEM) combined with energy dispersive X-ray spectroscopy (EDS) microanalysis has emerged as a powerful AMI technology to characterize the morphology and microstructure of minerals.^17–19 Commercialized systems, such as the Mineral Liberation Analysis (MLA) system,^17,19,20 Quantitative Evaluation of Minerals by Scanning electron microscopy (QEMSCAN),^21,22 the Advanced Mineral Identification and Characterization System (AMICS),^19,23 Tescan Integrated Mineral Analyzer (TIMA)²⁴ and Maps Mineralogy,^25,26 have been applied in geoscience and materials science.¹⁹ These AMI systems enable rapid visualization of mineral distributions and sizes at the micrometer scale. However, the reliability of the AMI method based on the automatic classification of EDS spectra remains unverified, potentially compromising the identification of precious lunar minerals. Furthermore, the volumes of X-ray generation within different types of minerals may be different at a single accelerating voltage (AV) of SEM, which makes it hard to discern minerals of various sizes.^27,28 Therefore, the analytical conditions and accuracy of the AMI method need to be further evaluated so that lunar minerals can be well recognized.

Other microbeam methods for mineralogical identification are electron probe microanalysis, Raman spectroscopy (RS) and transmission electron microscopy.^29,30 The EPMA method determines the major and minor element concentrations in minerals by wavelength-dispersive spectroscopy (WDS), making it easy to identify the mineral species. Different from the former, RS relies on the inelastic scattering of photons and reveals the crystal structure and lattice vibrations of minerals.^31–33 TEM is usually used to determine the crystallographic structure of materials and thus could be used to identify the mineral phases.^34,35 These analytical methods have unique advantages and can be applied to various minerals. However, the rapid and accurate identification of lunar minerals remains challenging due to diverse grain sizes (varying from micrometers to nanometers) and complex mineral phases. The identification results are significantly constrained by instrument resolution and analytical conditions. Thus, it is essential to systematically evaluate these analytical methods to achieve rapid, accurate and optimal mineral identification.

In this work, three CE-5 lunar basalt clasts were systematically investigated by using a combination of SEM, EPMA, RS and TEM. The SEM-EDS-based automatic mineral identification method was validated and assessed by comparison with other microbeam methods for mineralogical identification. We further evaluated the advantages and applicability of SEM, EPMA, RS and TEM methods for lunar mineral identification. Finally, a technical route was proposed to perform rapid and accurate identification of various minerals in lunar soils and other extraterrestrial samples, as well as terrestrial rocks.

2. Materials and experiments

2.1 Materials

The samples studied in this paper were collected from the lunar regolith and returned by the Chang'e-5 mission. The samples were assigned by the China National Space Administration Agency and designated CE5C0400YJFM00405, CE5C0400YJFM00406 and CE5C1000YJFM00404 (hereinafter abbreviated as “405#”, “406#” and “C1000#” samples, respectively). Lunar soil fragments were randomly hand-picked under an optical microscope in an ultraclean room environment at the Institute of Geology and Geophysics, Chinese Academy of Sciences (IGGCAS). These fragments were further mounted in epoxy resin and polished for mineralogical, chemical and structural analyses (see the Appendix). All sample preparation and analyses were performed at IGGCAS.

2.2 SEM analyses and automated mineral identification

The polished mounts were coated with carbon to a thickness of 8 nm. Backscattered electron (BSE) images were acquired on a Thermo Fisher Apreo S SEM at an accelerating voltage of 15 kV and a working distance of 8 mm. This SEM was equipped with a Bruker XFlash 6160 energy dispersive X-ray spectrometer for chemical analyses and automated mineral identification. The EDS detector had an elevation angle of 35°, and the working distance was 10 mm. EDS line-scanning and mapping analyses were conducted at an accelerating voltage of 20 kV and a beam current of 6.4 nA. Automated mineral identification of the lunar samples was conducted using Maps Mineralogy software and analyzed using Nanomin software at IGGCAS.²⁵ In addition, the SEM beam current was chosen to be 13 nA and the EDS acquisition time was 7–8 ms, which ensured that X-ray counts at each point for EDS analysis could reach 2500 during AMI analysis.

The automated mineral identification system discriminates minerals based on high-resolution BSE imaging and EDS semi-quantitative data of elements. The capability to recognize mineral size depends on the interaction volume (the depth and lateral width) of X-rays. Usually, the interaction volumes of incident electrons and excited X-rays within materials depends on accelerating voltage and material composition, i.e., the higher the accelerating voltage or the lighter the atoms in minerals, the greater the depth and width of electron scattering.^36,37 Using Monte Carlo simulation, it is possible to estimate the interaction volumes for different electron energies in various materials (see Appendix for details).^38–40 As shown in Fig. 1a and b, Monte Carlo simulations portray the interaction volumes of electron scattering at 5, 10, 15 and 20 kV for ilmenite and pyroxene. The depths and lateral widths of the emitted characteristic X-ray areas are shown in Table 1. With an increase in AV, the size of the interaction area increases while the spatial resolution is reduced. In this study, since lunar minerals contain heavier elements, such as Fe (E_c: 7.11 keV, Kα: 6.4 keV), Co (E_c: 7.71 keV; Kα: 6.93 keV) and Ni (E_c: 8.33 keV, Kα: 7.47 keV), the accelerating voltage was chosen to be 20 kV based on the empirical formula that AV should preferably be 2–3 times greater than the critical excitation energy (E_c) of the heaviest elements.^41,42Fig. 1c shows the electron interaction volumes and emitted X-ray volumes of Fe Kα (for kamacite and troilite) and Ca Kα (for plagioclase) at 20 kV. The widths of the emitted X-ray areas, on behalf of EDS lateral resolution, are ∼2.25, ∼3.3, ∼3.33, ∼3.9, ∼4.41, ∼4.7, and ∼5.43 μm for kamasite, ilmenite, troilite, olivine, pyroxene, apatite and plagioclase, respectively (Table 1 and SI Fig. S1). It is obvious from these simulations that minerals with heavier elemental compositions produce lower X-ray interaction volumes, while minerals with lighter elements produce larger volumes. When AMI analysis for the samples is conducted at the same accelerating voltage and EDS step size (such as 2R₁), a minute inclusion C could be identified in phase A (e.g., plagioclase) but may be missed in phase B (e.g., ilmenite) (Fig. 2a). Therefore, the EDS step size should be carefully considered according the target mineral during AMI analysis.

Fig. 1 Monte Carlo simulation of electron scattering at accelerating voltages of 5, 10, 15, 20 kV for ilmenite (Fe Kα) (a) and pyroxene (Fe Kα) (b). (c) Electrons and X-ray generation volumes calculated using Monte Carlo simulation at 20 kV for kamasite (Fe Kα), troilite (Fe Kα) and plagioclase (Ca Kα). 1 × 10⁵ electron trajectories were calculated. Blue trajectories indicate the total extent of electron penetration, and red trajectories represent the volume of emitted characteristic X-rays. Simulations were performed using the software package Electron Flight Simulator (version 3.1-E).^38,43,44 Ilm = ilmenite; Px = pyroxene; Ka = kamasite; Tro = troilite; Pl = plagioclase.

Table 1 The depths and widths of emitted characteristic X-rays at different voltages for typical lunar minerals^a

AV/sizes (μm)	Ilm-D	Ilm-W	Px-D	Px-W	Ka-D	Ka-W	Tro-D	Tro-W	Pl-D	Pl-W
a The depths and widths at 5 kV indicate the volume of primary electrons, while those at 10, 15 and 20 kV indicate the volume of emitted characteristic X-rays (Fig. 1 and S1). D = depth; W = width.
5 kV	0.27	0.31	0.34	0.40	0.17	0.23	0.25	0.31	0.44	0.50
10 kV	0.32	0.40	0.41	0.57	0.22	0.26	0.35	0.41	0.92	1.21
15 kV	1.30	1.62	1.60	2.03	0.86	1.05	1.36	1.69	2.78	3.30
20 kV	2.70	3.30	3.47	4.41	1.77	2.25	2.69	3.33	4.70	5.43

---

Fig. 2 (a) The schematic diagram of X-ray generation volumes excited by electron beam in phases A and B. The emitted X-ray radius (R₁) in phase A is larger than the excited radius (R₂) in phase B under the same working conditions. (b) BSE image of pyroxene and ilmenite. The yellow arrow indicates the location and direction of EDS line scan analysis. (c) The quantitative mass fractions (wt%) of Mg, Si, Ca, Ti and Fe along the yellow arrow in (b). (d) BSE image of plagioclase and ilmenite. (e) EDS mass fraction (wt%) of Al, Si, Ca, Ti and Fe on the yellow arrow in (d). (f) EDS elemental mass fraction (wt%) on the red arrow in (d). (g–i) BSE image of troilite and pyroxene, and mass fractions (wt%) of Mg, Si, S, Ca and Fe. The yellow and red arrows indicate the EDS line scan locations in (h) and (i), respectively. Px = pyroxene, Ilm = ilmenite, Pl = plagioclase, Tro = troilite.

To further determine the reliability of the emitted characteristic X-ray volumes simulated by the Monte Carlo method, SEM-EDS line scan analyses (20 kV accelerating voltage; the position of the EDS detector is in front of the direction of line movement, at an azimuth angle of 0°) were conducted on pyroxene, ilmenite, plagioclase and troilite, as shown in Fig. 2b–i. The length of the scanning line was 5.9 μm and the EDS step size was 0.5 μm. Fig. 2b shows the BSE image of pyroxene and ilmenite, with the line scan position starting at pyroxene (0 μm) and ending at the interface (5.9 μm) of pyroxene and ilmenite. Fig. 2c shows the mass fractions (wt%) of Mg, Si, Ca, Ti and Fe along the yellow arrow in Fig. 2b. It can be seen that the contents of Mg, Si and Ca decreased constantly with increasing scanning distance, while the contents of Ti and Fe increased significantly. At ∼3.75 μm, the Ti content (1.34 wt%) exceeds the normal concentration in pyroxene (∼0.69–1.1 wt% determined by EPMA), indicating that the X-ray generation volume has reached the adjacent ilmenite. This means that the interaction radius of X-rays in pyroxene is 2.15 (5.9 − 3.75 = 2.15) μm, and the corresponding lateral width is 4.3 μm. The EDS line analysis (yellow arrow in Fig. 2d) from plagioclase to the interface between plagioclase and ilmenite indicates that the content of Ti rises significantly from ∼3.22 μm (distance (μm)/content (wt%): 3.22/0.27; 3.75/0.58; 4.29/1.35; 4.83/3.37; 5.36/7.14; 5.9/14.25, as shown in Fig. 2e), that is, the lateral width of X-rays in plagioclase is 5.36 μm. As shown in Fig. 2f, the contents of Al, Si and Ca in ilmenite increased from ∼4.29 μm, indicating that the lateral width of the X-ray generation area in ilmenite is 3.22 μm. Furthermore, EDS line analyses on troilite and pyroxene revealed that the lateral widths of their X-ray generation areas are 3.18 μm and 4.26 μm, respectively (Fig. 2g–i). These experimental results demonstrate that the X-ray generation volumes in real minerals are basically consistent with those predicted by Monte Carlo simulations. Therefore, Monte Carlo simulations can be used to guide the working conditions (e.g. accelerating voltage and EDS step size) for AMI analysis.

2.3 Electron probe microanalysis

Most lunar silicate minerals in the fragments were measured using a JEOL JXA 8100 electron probe microanalyzer with an acceleration voltage of 15 kV at IGGCAS. The beam current was 20 nA and the beam size was 1 μm. The analytical crystal settings were as follows: two TAP crystals for Na, Mg, Al and Si; one LIFH crystal for Cr, Mn, Fe and Ni; one PETJ crystal for K, Ca, Ba and Ti.⁴⁵ The calibration standards were albite (Na), diopside (Mg, Ca, and Si), bustamite (Mn), K-feldspar (K), Fe₂O₃ (Fe), Al₂O₃ (Al), NiO (Ni), rutile (Ti), and Cr₂O₃ (Cr). Peak counting times were 10–20 s for major and trace elements. A program based on ZAF procedure was used for data correction (CITZAF).⁴⁶ Sulfides were determined using a CAMECA SXFive FE EPMA operated at an accelerating voltage of 20 kV, a beam current of 40 nA, and a beam size of ∼1 μm at IGGCAS. The calibration standards were FeS₂ (S and Fe), apatite (P), rutile (Ti), pure Ni, Co and Cu metals (Ni, Co and Cu), Ca–Al silicate glass (Si), and PbCrO₄ (Cr). The Kα characteristic X-rays were selected for all elements during analyses. Analytical crystals and peak count times for each element are as follows: Si (TAP, 20 s), P (LPET, 20 s), S (PET, 10 s), Ti (LPET, 20 s), Fe (LIF, 10 s), Cr (LLIF, 40 s), Ni (LLIF, 10 s) and Co (LLIF, 40 s). Peak overlap between Fe Kβ and Co Kα lines was corrected using Peak Sight software. All data were corrected on-line using the Phi-Rho-Z matrix correction procedure.⁴⁷

2.4 Micro-Raman spectroscopy

Micro-Raman analyses were performed on a confocal Raman microscope (WITec Alpha 300 R) at IGGCAS. Before testing, the Raman spectrometer was calibrated at a peak of 520.7 cm⁻¹ using a silicon reference. Mineral identification in 405#, 406# and C1000# clasts were performed with 532 nm radiation. The grating had 600 grooves per mm with a spectral resolution of 3 cm⁻¹. The laser beam was focused on the sample surface using a 100× Zeiss microscope with a spot size of ca. 360 nm. Raman measurements at power ranging from 1.25 mW to 25 mW on troilite in the C1000# clast were conducted using a laser with a wavelength of 488 nm. Raman shifts were recorded in the range of 65–1500 cm⁻¹.

2.5 Transmission electron microscopy analyses

The regions of interest in the lunar clasts were coated with Pt and cut using a Zeiss Auriga Compact focused ion beam microscope (FIB) equipped with an Omniprobe AutoProbe 200 micromanipulator at IGGCAS. The ultrathin foil pieces were thinned to ∼10 μm × 5 μm × 0.1 μm at 5–30 kV with beam currents of 50 pA to 2 nA. Before TEM observation, these foil pieces were cleaned using a multifunctional plasma cleaning apparatus that we developed. TEM bright-field (BF) imaging, selected area electron diffraction (SAED), high-resolution transmission electron microscopy (HRTEM) and EDS analyses were carried out using a JEOL JEM-2100 TEM instrument equipped with an Oxford X-MAX energy dispersive X-ray spectrometer at IGGCAS. This TEM was operated at 200 kV with an electron beam generated from a LaB₆ gun.

3. Results

3.1 BSE images and mineral distribution maps of lunar basalt

Backscattered electron images show the morphology of three lunar basalt clasts with sizes of ∼310, ∼690 and ∼1200 μm (Fig. 3a, d and g, respectively). The distinct contrasts in the BSE images indicate different types of minerals with clear grain boundaries. Fig. 3b, e and h are the corresponding mineral distribution maps of (a), (d) and (g) acquired using a SEM-based AMI system. A mineral map from the panel was fully scanned in just five minutes. Each colour displayed on the map represents an individual mineral phase that was identified automatically based on the lunar mineralogy database embedded in Nanomin software. The nomenclature of the studied lunar minerals follows the International Mineralogical Association (IMA) classification.^48,49 The mineralogical composition indicated that the analyzed basalts contained feldspar, pyroxene, olivine, ilmenite, apatite, spinel, Fe sulfide (Fe_1−xS (0 ≤ x < 0.125)), SiO₂ phase and Zr-bearing minerals. Since SiO₂ (such as quartz, cristobalite and so on) and Fe_1−xS compound (troilite/pyrrhotite) include multiple polymorphs that cannot be differentiated by SEM-EDS, these mineral phases identified by the AMI method are temporarily named by their chemical formulas. By comparing the mineral map with the corresponding BSE image, it can be seen that SiO₂ has the darkest contrast among all minerals, and those with gradually brightening contrast are mainly feldspar, pyroxene, olivine, ilmenite and Fe_1−xS compounds, in that order. All minerals show intergrowth textures with each other, suggesting that they may have formed during the same crystallization stage of the magma. Quantitative mineral abundances (area%) in lunar clasts were further acquired and are shown in Fig. 3c, f and i. It is evident that feldspar, pyroxene, olivine and ilmenite are the main components of the CE-5 lunar soil basalt. The mass abundances of the minerals are also easily available from Nanomin software and are shown in SI Table S1.

Fig. 3 Backscattered electron images (a, d and g), mineralogical distribution maps (b, e and h) and quantitative mineral abundances (area%) (c, f and i) of lunar basalt clasts. The lunar clasts in (a–c), (d–f) and (g–i) were picked from samples 406#, 405# and C1000#, respectively. Each colour in (b, e and h) represents an individual mineral phase. Minerals in the red box areas in (a) and (g) will be further analyzed by TEM. Fsp = feldspar, Px = pyroxene, Ol = olivine, Ilm = ilmenite, Spl = spinel, Ap = apatite.

3.2 EPMA analyses

EPMA analyses were conducted on minerals with different contrasts in BSE images of the 406#, 405# and C1000# CE-5 clasts. The chemical compositions of representative EPMA points for different minerals are shown in Tables 2 and 3. The measured pyroxenes have ∼7.78–12.13 wt% of MgO, ∼46.49–48.87 wt% of SiO₂, ∼12.67–18.81 wt% of CaO, and ∼15.35–28.16 wt% of FeO (SI Table S2). The values of wollastonite (Wo)_27.8–41.7 and enstatite (En)_24.1–36.4 indicate that the analyzed pyroxene crystals belong to augites. Feldspars contain ∼31.09–33.61 wt% of Al₂O₃, ∼45.55–49.29 wt% of SiO₂, and ∼16.15–18.1 wt% of CaO (SI Table S2). The feldspars identified by the AMI technique were determined to be plagioclase, subclassified as anorthite with An_81.4–87.1Ab_10.2–17.4Or_0.4–1.1. Similarly, olivines were identified with compositional variations of Fa_44.9–68.2Fo_31.8–54.7. The compositions of pyroxene, plagioclase and olivine are consistent with those reported for most CE-5 basaltic clasts.^11,14 Ilmenite and the SiO₂ phase were also verified based on the major and minor element concentrations obtained by EPMA. In summary, lunar minerals identified by the EPMA method are identical to those recognized by the SEM-EDS-based AMI method. In addition, Table 3 shows the normalized composition of the Fe_1−xS (0 ≤ x < 0.125) compound in atomic percent (at%) (wt% data are shown in Table S3 in the SI). The Fe/S atomic ratios of Fe_1−xS compounds in 406#, 405# and C1000# range from 0.96 to 0.993, 0.98 to 0.996 and 0.975 to 0.99, respectively. Obviously, Fe sulfides with similar composition in lunar soils, such as troilite (FeS) and pyrrhotite, are hard to distinguish by the EPMA method.

Table 2 EPMA chemical compositions of pyroxene, plagioclase, olivine, ilmenite and SiO₂ in 406# and 405# lunar clasts^a

Comment	Na2O	MgO	Al2O3	SiO2	K2O	CaO	TiO2	Cr2O3	MnO	FeO	NiO	Total	Fs^b	En	Wo
a “bdl” means below detection limit. b Fs = 100 × Fe/(Fe + Mg + Ca); En = 100 × Mg/(Mg + Fe + Ca); Wo = 100 × Ca/(Ca + Fe + Mg). c An = 100 × Ca/(Ca + Na + K); Ab = 100 × Na/(Na + Ca + K); Or = 100 × K/(K + Ca + Na). d Fa = 100 × Fe/(Fe + Mg); Fo = 100 × Mg/(Mg + Fe).
406#-Px-1	0.04	7.78	1.41	47.40	bdl	12.67	1.15	0.17	0.44	28.16	bdl	99.21	48.1	24.1	27.8
406#-Px-2	bdl	8.84	1.51	47.67	bdl	12.69	1.23	0.21	0.45	26.55	bdl	99.13	45.1	27.2	27.7
405#-Px-1	0.10	10.53	2.80	47.74	bdl	16.80	2.32	0.38	0.29	18.19	bdl	99.14	30.9	32.4	36.7
405#-Px-2	0.05	11.31	2.44	48.21	bdl	14.26	1.91	0.37	0.36	20.29	bdl	99.18	34.3	34.6	31.1
Pl^c												Total	An	Ab	Or
406#-Pl-1	1.14	0.18	33.61	45.55	0.06	18.10	0.08	bdl	bdl	1.23	bdl	99.98	89.4	10.2	0.4
406#-Pl-2	1.20	0.21	33.61	45.67	0.07	18.04	0.07	bdl	bdl	1.18	bdl	100.05	89.0	10.7	0.4
405#-Pl-1	1.36	0.25	32.63	47.05	0.11	17.34	0.10	bdl	bdl	0.74	bdl	99.61	87.0	12.3	0.7
405#-Pl-2	1.41	0.23	32.52	47.07	0.11	17.23	0.09	bdl	bdl	0.83	bdl	99.51	86.6	12.8	0.6
Ol^d												Total	Fa	Fo
406#-Ol-1	bdl	23.06	bdl	34.76	bdl	0.44	0.10	0.10	0.41	41.52	bdl	100.39	49.9	50.1
406#-Ol-2	bdl	25.50	bdl	35.59	bdl	0.43	0.09	0.12	0.39	38.47	0.03	100.65	45.5	54.5
405#-Ol-1	bdl	13.67	bdl	32.77	bdl	0.44	0.12	0.05	0.53	53.01	bdl	100.59	68.2	31.8
405#-Ol-2	0.03	14.96	1.96	32.53	bdl	0.42	0.08	0.05	0.52	50.81	bdl	101.35	65.2	34.8
Ilm												Total
406#-Ilm-1	0.04	1.93	0.11	bdl	bdl	0.13	52.67	0.46	0.39	44.30	bdl	100.02
406#-Ilm-2	bdl	1.89	0.11	bdl	bdl	0.16	52.63	0.48	0.38	44.22	bdl	99.83
405#-Ilm-1	0.03	0.23	0.11	bdl	bdl	0.17	52.76	0.21	0.36	46.10	bdl	99.93
405#-Ilm-2	bdl	0.25	0.12	bdl	bdl	0.16	52.94	0.22	0.38	46.11	bdl	100.16
SiO2												Total
405#-SiO2-1	0.14	bdl	0.54	98.89	bdl	0.15	0.23	bdl	bdl	0.31	bdl	100.24
405#-SiO2-2	0.05	bdl	0.73	97.58	0.08	0.17	0.23	bdl	bdl	0.31	bdl	99.16

---

Table 3 Electron microprobe analyses of Fe_1−xS (normalized at%) in CE-5-406#, 405# and C1000# samples^a

Comment	Si	P	S	Ti	Fe	Cr	Ni	Co	Total	Fe/S
a “bdl” means below detection limit.
406#-Fe1−xS-1	0.18	bdl	50.73	0.14	48.92	bdl	0.03	bdl	100	0.964
406#-Fe1−xS-2	0.12	bdl	50.40	0.15	49.3	0.01	0.02	bdl	100	0.978
406#-Fe1−xS-3	0.14	bdl	50.83	0.18	48.82	0.01	0.02	bdl	100	0.960
406#-Fe1−xS-4	0.15	bdl	50.09	0.02	49.72	0.01	bdl	bdl	99.99	0.993
405#-Fe1−xS-1	0.09	bdl	49.91	0.76	49.23	bdl	bdl	0.01	100	0.986
405#-Fe1−xS-2	0.13	bdl	49.60	0.86	49.40	bdl	bdl	0.01	100	0.996
405#-Fe1−xS-3	0.10	bdl	49.85	0.72	49.31	0.01	bdl	0.01	100	0.989
405#-Fe1−xS-4	0.13	bdl	49.80	0.87	49.19	bdl	bdl	0.01	100	0.988
405#-Fe1−xS-5	0.09	bdl	50.07	0.75	49.05	0.01	0.01	0.01	99.99	0.980
C1000#-Fe1−xS-1	0.05	bdl	50.30	0.51	49.12	bdl	bdl	0.01	99.99	0.977
C1000#-Fe1−xS-2	0.15	bdl	50.19	0.38	49.22	bdl	bdl	0.01	99.95	0.981
C1000#-Fe1−xS-3	0.28	bdl	50.42	0.09	49.18	bdl	bdl	bdl	99.97	0.975
C1000#-Fe1−xS-4	0.19	bdl	50.34	0.09	49.35	bdl	bdl	bdl	99.97	0.980
C1000#-Fe1−xS-5	0.31	bdl	49.99	0.22	49.47	bdl	bdl	bdl	99.99	0.990

---

3.3 Micro-Raman spectroscopy analyses of lunar minerals

Lunar minerals in the three clasts were further checked using a Raman spectroscope. Fig. 4 shows the Raman spectra of various minerals. Olivine, pyroxene (augite), ilmenite and feldspar (plagioclase/anorthite), constituting the main lunar minerals, were well identified and show typical Raman vibrational bands at ∼843–849, ∼1007–1010, ∼676–681 and ∼505–507 cm⁻¹, respectively (refer to the CE-5 basalt material of IGGCAS and the RRUFF Project database).^33,50,51 The SiO₂ phase in the 405# clast was clearly discriminated and shows Raman bands at 231 and 416 cm⁻¹, indicating a cristobalite structure (Fig. 4b). While in the C1000# clast, it was identified as quartz with Raman bands at 204 and 466 cm⁻¹ (Fig. 4c). The locations of the analyzed SiO₂ phase in 405# and C1000# clasts are shown in Fig. S2 in the SI. The Fe_1−xS compounds in 406# and 405# clasts show Raman peaks localized at 200 cm⁻¹, indicating that all of them are troilite (reference material: troilite in Dar el Kahal H5-6 ordinary meteorite⁵²). In C1000# clast, besides the identified troilite, pyrrhotite was also recognized with a Raman peak at 334 cm⁻¹ (Fig. 4c).^53,54 The RS analysis positions of the Fe_1−xS compound in the 406#, 405# and C1000# clast are the same as those of the EPMA analysis, as shown in SI Fig. S3. In addition, a sub-microsized mineral with a cyan Raman peak at 374 cm⁻¹ in the 406# clast wasn't identified (Fig. 4a) and requires further investigation with TEM.

Fig. 4 (a–c) Raman spectra of various lunar minerals in 406#, 405# and C1000# samples, respectively. (d) Raman spectra of troilite in the C1000# clast exposed to different laser powers in air.

Fig. 4d shows the Raman spectra of troilite (within C1000# clast, as shown in panel (h) in SI Fig. S3) at different laser powers in air. From 1.25 to 5 mW, troilite maintains a stable structure and typical Raman shifts are located at around 150 cm⁻¹ and 200 cm⁻¹. At a laser power of 8 mW, a new 225 cm⁻¹ peak appeared, indicating Fe–S vibrations (stage II in Fig. 4d). With increasing laser power, the main Raman peaks are localized at 296, 533 and ∼660 cm⁻¹ from 12 to 20 mW, and at 280, 390, 584 and 1270 cm⁻¹ at 25 mW, corresponding to magnetite and hematite, respectively. SEM analyses further demonstrated that troilite was oxidized in air by the laser heating (Fig. 5). As shown in Fig. 5a and b, BSE images clearly show the irradiated spot of troilite in the C1000# clast. EDS spectra and element maps indicate the irradiated product on troilite belongs to Fe-oxides (Fig. 5c–f).

Fig. 5 (a) BSE image of the irradiated troilite in the C1000# clast under 25 mW laser power in air. (b) The enlarged BSE image of the irradiated spot in panel (a). (c) EDS spectra of troilite and Fe-oxide in (b). (d–f) EDS element maps of minerals in panel (b) that show the distribution of Fe, O and S.

3.4 TEM analyses of nanosized lunar minerals

The nanosized mineral inclusion in troilite in the 406# clast was not identified by the AMI method, EPMA and micro-Raman analyses due to its smaller grain size and ambiguity in interpreting the Raman spectrum (Fig. 3a and 4a). TEM analyses were performed on the cross section of FIB foil-1 (Fig. 6b) cut from the nanosized minerals in Fig. 6a. Fig. 6c shows EDS spectra of two minerals, one containing Fe and S, and the other containing Fe, S, and Ni. Fig. 6d shows the typical selected area electron diffraction (SAED) pattern of troilite (hexagonal structure with space group P6̄2c, data from the American Mineralogist Crystal Structure Database, AMCSD) viewed along the [1̄100] zone axis (reference material: CE5 basaltic troilite from IGGCAS).⁸ SAED patterns in Fig. 6e and f (viewed along the [1̄11] and [013̄] directions, respectively) demonstrate that the unknown mineral is pentlandite with a face centered cubic structure (space group: Fm3̄m) (reference material: CE5 lunar pentlandite from IGGCAS⁸).

Fig. 6 (a) The enlarged BSE image of the specimen (406# clast) indicated in the red box in Fig. 3a. (b) Secondary electron (SE) image of FIB foil-1 cut along the dotted white line in (a). (c) TEM-EDS spectra of troilite and pentlandite in the specimen. (d) SAED pattern of troilite, viewed along the [1̄100] zone axis. (e) and (f) SAED patterns of pentlandite, viewed along the [1̄11] and [013̄] directions, respectively. (g) BSE image of the specimen (C1000# clast) indicated by the red box in Fig. 3g. (h and i) TEM BF image and EDS spectra of FIB foil-2 cut along the dotted white line in (g). (j) SAED pattern of clinopyroxene, viewed along the [310] direction. (k) and (l) SAED patterns of ilmenite, viewed along the [1̄11] and [4̄41] directions, respectively. (m–o) HRTEM images of host clinopyroxene and ilmenite inclusion.

Fig. 6g shows the BSE image of the specimen (C1000# clast) indicated by the red box in Fig. 3g. The nanosized inclusions (∼100 nm) and the host mineral can be seen clearly, but cannot be identified by the SEM-EDS AMI method. Fig. 6h and i show the TEM BF image and EDS spectra of minerals in FIB foil-2. Combined with elemental EDS spectra and the TEM-SAED pattern in Fig. 6j, the host mineral was identified as clinopyroxene, viewed along the [310] zone axis. The inclusions were analyzed along different crystallographic directions ([1̄11] and [4̄41]) and identified as ilmenite with space group R3̄ (Fig. 6k and l). HRTEM images in Fig. 6m–o further show clear lattice-fringes and the interfacial orientation relationship between clinopyroxene and ilmenite. Obviously, these nano-minerals can be easily and robustly identified using TEM techniques.

4. Discussion

4.1 Reliability evaluation of the AMI method in identifying lunar minerals

Three lunar clasts were analyzed using the SEM-EDS-based AMI method. The identified minerals are mainly feldspar, pyroxene, olivine, ilmenite, Fe_1−xS compound, SiO₂ phase, apatite and spinel, and shown in mineral distribution maps (Fig. 3). We further used EPMA and RS to analyze those minerals with different contrasts in clasts (Tables 2, 3 and Fig. 4). The identified mineral species are consistent with those identified by the AMI method, indicating that the AMI method is substantially accurate and reliable. It provides a convenient way to search for specific mineral phases of interest, such as apatite and baddeleyite in CE-5 lunar soil.^7,11,55 Additionally, mixed spectra caused by the electron beam incident at grain boundaries may lead to uncertainties in identifying the fine-grained minerals (e.g., <5 μm).^56,57 Although it is uncommon, we recommend either adding mixed spectra to the mineral reference list to aid in identifying fine-grained minerals, or rechecking such minerals with SEM-EDS or other microbeam techniques.

To accurately identify various types of minerals during AMI analysis, the analytical conditions of both SEM and EDS are also vital. Based on the critical excitation potential of heavy elements (E_c: 8.33 keV for Ni; 7.11 keV for Fe) in lunar minerals, the accelerating voltage of the SEM is preferably set at 20 kV. Monte Carlo simulations for lunar minerals have demonstrated that the interaction volumes of excited X-rays were different within different minerals, which determines the lateral resolution of EDS (Fig. 1, S1 and Table 1). Therefore, before AMI analysis, it is preferable to set the EDS step size with reference to the lateral width of X-ray interaction area to achieve rapid identification of minerals of various sizes. For instance, the EDS step size can be set to ∼5.4 μm (2R₁ ≈ 5.43 μm) for feldspar, ∼4.4 μm for pyroxene, ∼3 μm for ilmenite and troilite, ∼2 μm for FeNi metal, and ∼1 μm for Zr-bearing minerals (∼600 nm width of X-ray volumes of Zr Lα, SI Fig. S1i and j). In some cases, the EDS step size is even recommended to be half or two-thirds the lateral width. As shown in Fig. 3b, the EDS step size of 1 μm was used to scan the 405# clast and a Zr-bearing mineral with a size of ∼1.6 μm was accurately recognized (the location is shown in panel (d) in SI Fig. S3). However, when the EDS scanning step size was set to 3 μm, the mineral distribution map demonstrated no detectable Zr-bearing minerals (SI Fig. S4). If there is a specific need for mineral species in the next experiment, such as searching for Zr-bearing mineral with size above 5 μm for in situ SIMS dating (∼3 μm of beam size), the EDS step size can be set to at least 5 μm to reliably find zircon and baddeleyite. Certainly, if minerals with smaller size are not included in the screening target, the EDS step size can be increased to improve identification efficiency.

4.2 Evaluation of various microbeam-based mineral identification methods

Four common microbeam methods, such as SEM-based AMI, EPMA, RS and TEM, have been used for lunar mineral identification. Each method however, has its own limitations and there is no universal method to efficiently identify all lunar minerals. The AMI method can rapidly identify mineral species without standards, and present visualized mineral distribution maps and mineral abundances (Fig. 3). In addition, petrographic textures and locations of specific minerals in clasts are acquired by combining with BSE images, which is highly convenient for other in situ microbeam experiments. On the downside, some minerals with polymorphic structures and similar compositions, such as SiO₂ and Fe_1−xS, cannot be distinguished. EPMA can precisely quantify the elemental composition of minerals at a micrometer scale, and obtain mineral formulae by calculation of abundance. However, the standards are essential and the WDS parameters need to be adjusted for different types of minerals during EPMA testing. It is time-consuming. Additionally, it has been demonstrated that the EPMA method also cannot recognize polymorphism phases and minerals with similar compositions (Tables 2 and 3).

For Raman spectroscopy, the significant advantages are the absence of standards and fast measurement. Unlike the AMI and EPMA methods, RS can also clearly distinguish the structures of quartz, cristobalite, troilite and pyrrhotite in lunar rock (Fig. 4b and c). However, there are two shortcomings in Raman analysis. First, Raman spectroscopy has multiple possible interpretations to some extent, making some tiny minerals difficult to identify. Second, high laser power can induce phase transitions of sulfide minerals (Fig. 4d), as also reported in previous studies of troilite.^52,58,59 Therefore, Raman analysis of sulfide minerals should be performed at a low laser power of <5 mW. For nanoscale minerals (<1000 nm) which are beyond the resolution of AMI, EPMA and RS, the TEM technique can easily identify the crystal structure of minerals. As shown in Fig. 6e and f, the pentlandite mineral that cannot be identified by AMI, EPMA and RS was well indexed in TEM SAED patterns. TEM can accurately identify mineral polymorphs and interfacial orientation relationship between two minerals.^60,61 The only limitation is that TEM experiments require ultrathin and cleaned specimens (less than 100 nm in thickness) prepared using a focused ion beam (FIB) microscope, which is a complicated process. For AMI, EPMA and RS microbeam methods, no complicated sample preparation is needed.

The characteristics and applicability of AMI, EPMA, RS and TEM methods in identifying lunar minerals are summarized in Table 4. SEM-based AMI enables rapid visualization of mineral distributions across large-scale areas (micrometers to centimeters), whereas EPMA and RS are confined to smaller regions and exhibit slower acquisition speeds in mineral mapping. TEM imaging characterizes mineral distribution at sub-micrometer scales. Empirically, AMI, EPMA and RS analyses (lower laser power for sulfides) are considered to be “non-destructive” and TEM analysis is micro-destructive. The efficiency of mineral identification is high for AMI and RS methods, but slow for EPMA and TEM methods due to complex analytical conditions or sample preparation. The evaluation of characteristics and applicability of the four techniques helps to provide the optimum identification method for lunar minerals with different sizes.

Table 4 Characteristics and applicability of AMI, EPMA, RS and TEM methods

Method	Sample preparation	Standard	Mineral size	Visualization	Destructiveness	Polymorph	Efficiency
AMI	Polishing	No	Micrometer	Yes	Non-destructive	No	High
EPMA	Polishing	Yes	Micrometer	Yes	Non-destructive	No	Low
RS	Polishing	No	Micrometer	Yes	Non-destructive except sulfides	Yes	High
TEM	FIB cutting	No	Nanometer	Yes	Micro-destructive	Yes	Low

---

4.3 Technical route for rapid and accurate identification of lunar minerals

Lunar minerals range in size from a few hundred micrometers to nanometers. To identify these lunar minerals rapidly and accurately, comprehensive analyses using multiple microbeam methods are essential. Here, we preferentially recommend the SEM-EDS-based AMI method for mineral identification because it can acquire visualized mineral distribution maps, grain size and mineral abundances, which is conducive to providing accurate mineral localization for other in situ microanalytical experiments (Fig. 7). For polymorphic phases (such as SiO₂) and minerals with similar compositions (Fe_1−xS) that are not recognized by the AMI method, Raman spectroscopy is used to identify SiO₂ as quartz or cristobalite, and Fe_1−xS as troilite or pyrrhotite (Fig. 4 and 7). Finally, for nanoparticles in minerals, the FIB-TEM method is used to accurately identify their crystal structures, such as pentlandite inclusions, seifertite, stishovite, nanosized ilmenite and nanophase Fe particles.^62–64 The technical route (AMI-RS-TEM) provides a reference method for mineral identification, which can effectively identify different sizes of lunar minerals, and provide important mineralogical instructions for further studies of lunar evolution.

Fig. 7 AMI-RS-TEM technical roadmap for the identification of lunar minerals.

5. Conclusion

Lunar minerals with sizes ranging from a few hundred micrometers to the nanometer scale were full identified using AMI, EPMA, RS, and TEM methods. The advantages, disadvantages and applicability of the four methods were systematically evaluated and summarized. The novel aspects of this work are as follows: (i) Monte Carlo simulation was performed to optimize the analytical conditions of the AMI method, which helps to identify various lunar minerals rapidly; (ii) the reliability of the AMI method for mineral identification was well demonstrated through evaluation with EPMA and RS methods; (iii) a comprehensive technical route for rapid and accurate identification of lunar minerals was established. It will help implement the optimal mineral identification method for rare and valuable lunar samples. This work also provides valuable insights and methodological foundations for the rapid and precise identification of fine-grained significant minerals in both terrestrial and other extraterrestrial samples (e.g., asteroidal materials, meteorite samples, etc.), contributing to advances in planetary science, resource exploration, and geological research.

Author contributions

All authors listed contributed to this study. Conceptualization: Xu Tang. Methodology: Xu Tang, Lixin Gu, Di Zhang, Xiaoguang Li and Lihui Jia. Validation: Jinhua Li and Qiuli Li. Investigation: Xu Tang. Resources, Wei Yang, Jinhua Li, Hengci Tian, Shuhui Cai, and Li Wang. Data curation: Lixin Gu, Di Zhang, Xiaoguang Li, Lihui Jia, and Li Wang. Writing – original draft: Xu Tang. Writing – review & editing: Xu Tang. Funding acquisition: Xu Tang. All authors have read and approved the final manuscript.

Conflicts of interest

There are no conflicts of interest to declare.

Data availability

Data necessary to generate the results used for this study are available online (https://doi.org/10.5281/zenodo.14993524, https://doi.org/10.5281/zenodo.7912813).

Supplementary information: Monte Carlo simulations, mineralogical distribution maps, mass abundance of lunar minerals, EPMA chemical compositions, sample preparation, SEM and EDS calibration, and the calibrate procedures for BSE detector. See DOI: https://doi.org/10.1039/d5ja00092k.

Acknowledgements

We thank Prof. Xianhua Li from the Institute of Geology and Geophysics, Chinese Academy of Sciences, for providing some lunar soil mounts. We also thank Dr Cong Liu from the Bruker (Beijing) Scientific Technology Co., Ltd, for his kind suggestions regarding EDS line analysis. This work was supported by the Experimental Technology Innovation Fund of the Institute of Geology and Geophysics, Chinese Academy of Sciences (TEC202304), the National Natural Science Foundation of China (42203025 and 42241101) and the Key Research Program of the Institute of Geology and Geophysics, Chinese Academy of Sciences (IGGCAS-202101).

References

1. J. A. Wood, U. B. Marvin, J. B. Reid, G. J. Taylor, J. F. Bower, B. N. Powell, J. S Dickey, Smithsonian Institution Astrophysical Observatory Special Report 333, 1971, pp. 1–272.
2. P. J. Isaacson, A. B. Sarbadhikari, C. M. Pieters, R. L. Klima, T. Hiroi, Y. Liu and L. A. Taylor, Meteorit. Planet. Sci., 2011, 46, 228–251.
3. J. J. Papike, G. Ryder and C. K. Shearer, Rev. Mineral. Geochem., 1998, 36, 1–234.
4. H. Hui, C. R. Neal, C.-Y. Shih and L. E. Nyquist, Earth Planet. Sci. Lett., 2013, 373, 150–159.
5. H. C. Tian, C. Zhang, W. Yang, J. Du, Y. Chen, Z. Xiao, R. N. Mitchell, H. Hui, H. G. Changela, T. X. Zhang, X. Tang, D. Zhang, Y. Lin, X. Li and F. Wu, Nat. Commun., 2023, 14, 3734.
6. A.-C. Zhang, H.-Y. He, S. Hu, X.-H. Li, Y.-T. Lin, L.-P. Qin, G.-Q. Wang and Z.-Y. Xiao, Space: Sci. Technol., 2024, 4, 0170.
7. Q. L. Li, Q. Zhou, Y. Liu, Z. Xiao, Y. Lin, J. H. Li, H. X. Ma, G. Q. Tang, S. Guo, X. Tang, J. Y. Yuan, J. Li, F. Y. Wu, Z. Ouyang, C. Li and X. H. Li, Nature, 2021, 600, 54–58.
8. X. Tang, H. Tian, S. Sun, L. Gu, Q. Li, X. Li and J. Li, Lithos, 2023, 458–459.
9. J. C. Graf, Lunar soils grain size catalog, National Aeronautics and Space Administration, 1993, pp. 1–463.
10. S. J. Liu, Q. Zhou, Q. L. Li, S. Hu and W. Yang, Sci. China: Earth Sci., 2021, 64, 2083–2089.
11. H. C. Tian, H. Wang, Y. Chen, W. Yang, Q. Zhou, C. Zhang, H. L. Lin, C. Huang, S. T. Wu, L. H. Jia, L. Xu, D. Zhang, X. G. Li, R. Chang, Y. H. Yang, L. W. Xie, D. P. Zhang, G. L. Zhang, S. H. Yang and F. Y. Wu, Nature, 2021, 600, 59–63.
12. J. L. Ji, H. C. He, S. Hu, Y. T. Lin, H. J. Hui, J. L. Hao, R. Y. Li, W. Yang, Y. H. Yan, H. C. Tian, C. Zhang, M. Anand, R. Tartèse, L. X. Gu, J. H. Li, D. Zhang, Q. Mao, L. H. Jia, Y. Chen, S. T. Wu, H. Wang, H. Y. He, X. H. Li and F. Y. Wu, Earth Planet. Sci. Lett., 2022, 591, 117636.
13. H. C. Tian, W. Yang, D. Zhang, H. J. Zhang, L. H. Jia, S. T. Wu, Y. T. Lin, X. H. Li and F. Y. Wu, Am. Mineral., 2023, 108(9), 1669–1677.
14. Q. He, Y. H. Li, I. Baziotis, Y. Q. Qian, L. Xiao, Z. C. Wang, W. Zhang, B. J. Luo, C. R. Neal, J. M. D. Day, F. Pan, Z. B. She, X. Wu, Z. C. Hu, K. Q. Zong and L. Wang, Icarus, 2022, 383, 115082.
15. B. J. Luo, Z. C. Wang, J. L. Song, Y. Q. Qian, Q. He, Y. H. Li, J. W. Head, F. Moynier, L. Xiao, H. Becker, B. X. Huang, B. Ruan, Y. X. Hu, F. B. Pan, C. Xu, W. L. Liu, K. Q. Zong, J. W. Zhao, W. Zhang, Z. C. Hu, Z. B. She, X. Wu and H. F. Zhang, Nat. Geosci., 2023, 16, 301–308.
16. J. H. Li, Q. L. Li, L. Zhao, J. H. Zhang, X. Tang, L. X. Gu, Q. Guo, H. X. Ma, Q. Zhou, Y. Liu, P. Y. Liu, H. Qiu, G. Li, L. Gu, S. Guo, C. L. Li, X. H. Li, F. Y. Wu and Y. X. Pan, Geosci. Front., 2022, 13, 101367.
17. Y. Gu, J. Miner. Mater. Char. Eng., 2003, 2, 33–41.
18. C. Cardell, I. Guerra, J. Romero-Pastor, G. Cultrone and A. Rodriguez-Navarro, Anal. Chem., 2009, 81, 604–611.
19. B. Schulz, D. Sandmann and S. Gilbricht, Minerals, 2020, 10, 1004.
20. B. Schulz, G. Merker and J. Gutzmer, Minerals, 2019, 9, 527.
21. P. Gottlieb, G. Wilkie, D. Sutherland, E. Ho-Tun, S. Suthers, K. Perera, B. Jenkins, S. Spencer, A. Butcher and J. Rayner, JOM, 2000, 52, 24–25.
22. B. Ayling, P. Rose, S. Petty, E. Zemach and P. Drakos, Thirty-Seventh Workshop on Geothermal Reservoir Engineering, SGP-TR-194, 2012.
23. D. L. Min, G. X. Kun, F. Guang, L. Ting, Y. A. Peng and W. Tao, World Nucl. Geosci., 2023, 40, 98–105.
24. T. Hrstka, P. Gottlieb, R. Skála, K. Breiter and D. Motl, J. Geosci., 2018, 63, 47–63.
25. L. X. Gu, S. Hu, M. Anand, X. Tang, J. L. Ji, B. Zhang, N. Wang and Y. T. Lin, Am. Mineral., 2022, 107, 1018–1029.
26. S. Guo, J. Hermann, P. Tang, X. Chu, Y. Chen and B. Su, Earth Planet. Sci. Lett., 2022, 597, 117816.
27. D. E. Newbury and N. W. M. Ritchie, J. Anal. At. Spectrom., 2013, 28, 973–988.
28. D. Zhang, Y. Chen, W. Yang, J. H. Fournelle, J. L. Ji, B. Su, Q. Mao, L. H. Jia, J. Y. Yuan and X. G. Li, At. Spectrosc., 2022, 43, 28–41.
29. R. F. Berry, L. V. Danyushevsky, K. Goemann, A. Parbhakar-Fox and T. Rodemann, in Environmental Indicators in Metal Mining, ed. B. Lottermoser, Springer International Publishing, Cham, 2017, pp. 55–72,  DOI:10.1007/978-3-319-42731-7_4.
30. X. Llovet, in Encyclopedia of Analytical Science, 3rd edn, 2019, vol. 7, pp. 30–38.
31. J. Y. Yuan, W. Su, X. M. Hu, X. G. Li and C. H. Fei, Microsc. Res. Tech., 2022, 85, 2729–2739.
32. A. Wang, B. L. Jolliff and L. A. Haskin, J. Geophys. Res.: Planets, 1995, 100, 21189–21199.
33. Z. C. Ling, A. Wang and B. L. Jolliff, Icarus, 2011, 211, 101–113.
34. R. Brydson, A. Brown, L. G. Benning and K. Livi, Rev. Mineral. Geochem., 2014, 78, 219–269.
35. X. Tang and J. H. Li, Earth Sci., 2021, 46, 1374–1415.
36. I. Barkshire, P. Karduck, W. P. Rehbach and S. Richter, Mikrochim. Acta, 2000, 132, 113–128.
37. X. Llovet, A. Moy, P. T. Pinard and J. H. Fournelle, Prog. Mater. Sci., 2021, 116, 100673.
38. D. C. Joy, Monte Carlo Modeling for Electron Microscopy and Microanalysis, Oxford University Press, 1995.
39. S. Burgess, J. Sagar, J. Holland, X. B. Li and F. Bauer, Microsc. Today, 2017, 25, 20–29.
40. A. Popowich, T. Lam and E. P. Vicenzi, Anal. Chem., 2021, 93, 12906–12913.
41. J. I. Goldstein, D. E. Newbury, P. Echlin, D. C. Joy, C. E. Lyman, E. Lifshin, L. Sawyer and J. R. Michael, Scanning Electron Microscopy and X-ray Microanalysis, Springer New York, NY, 3rd edn, 2003, pp. 1–690,  DOI:10.1007/978-1-4615-0215-9.
42. A. Ul-Hamid, A Beginners' Guide to Scanning Electron Microscopy, Springer Cham, 2018, pp. 1–422,  DOI:10.1007/978-3-319-98482-7.
43. L. Khouchaf, Gaseous Scanning Electron Microscope (GSEM): Applications and Improvement, in Scanning Electron Microscopy, ed. V. Kazmiruk, IntechOpen, Rijeka, 2012,  DOI:10.5772/34930.
44. B. Van Eerdenbrugh, L. Froyen, J. Van Humbeeck, J. A. Martens, P. Augustijns and G. Van Den Mooter, Eur. J. Pharm. Sci., 2008, 35, 344–353.
45. L. Jia, X. Ma, S. Wu, Y. Li, J. Yuan, Q. Mao, X. Li and Y. Chen, Geostand. Geoanal. Res., 2024, 48, 207–225.
46. J. T. Armstrong, Microbeam Anal., 1995, 4, 177–200.
47. C. Merlet, Mikrochim. Acta, 1994, 114, 363–376.
48. W. A. Deer, R. A. Howie and J. Zussman, An Introduction to the Rock-Forming Minerals, Mineralogical Society of Great Britain and Ireland, 3rd edn, 2013.
49. https://rruff.info/ima/ .
50. H. Cao, C. Wang, J. Chen, X. Che, X. Fu, Y. Shi, D. Liu, Z. Ling, L. Qiao, X. Lu, X. Qi, C. Yin, P. Liu, C. Liu, Y. Xin and J. Liu, Geophys. Res. Lett., 2022, 49, e2022GL099282.
51. K. Cao, M. Dong, Z. She, Q. Xiao, X. Wang, Y. Qian, Y. Li, Z. Wang, Q. He, X. Wu, K. Zong, Z. Hu and L. Xiao, Sci. China: Earth Sci., 2022, 65, 1704–1714.
52. X. G. Li, Y. Chen, X. Tang, L. X. Gu, J. Y. Yuan, W. Su, H. C. Tian, H. Q. Luo, S. H. Cai and S. Komarneni, Icarus, 2023, 390, 115299.
53. E. L. Walton, T. G. Sharp, J. Hu and J. Filiberto, Geochim. Cosmochim. Acta, 2014, 140, 334–348.
54. T. Pasinszki, M. Krebsz, D. Chand, L. Kótai, Z. Homonnay, I. E. Sajó and T. Váczi, J. Mater. Sci., 2020, 55, 1425–1435.
55. S. Hu, H. C. He, J. L. Ji, Y. T. Lin, H. J. Hui, M. Anand, R. Tartese, Y. H. Yan, J. L. Hao, R. Y. Li, L. X. Gu, Q. Guo, H. Y. He and Z. Y. Ouyang, Nature, 2021, 600, 49–53.
56. M. Kern, R. Möckel, J. Krause, J. Teichmann and J. Gutzmer, Miner. Eng., 2018, 116, 213–225.
57. R. Blannin, M. Frenzel, L. Tuşa, S. Birtel, P. Ivăşcanu, T. Baker and J. Gutzmer, Miner. Eng., 2021, 167, 106836.
58. G. Genchev and A. Erbe, J. Electrochem. Soc., 2016, 163, C333–C338.
59. I. Weber, U. Böttger, S. G. Pavlov, H. W. Hübers, H. Hiesinger and E. K. Jessberger, J. Raman Spectrosc., 2017, 48, 1509–1517.
60. R. L. Pang, J. Yang, W. Du, A. C. Zhang, S. R. Liu and R. Li, Geophys. Res. Lett., 2022, 49, e2022GL098722.
61. X. Liu, B. X. Su, Y. Bai, P. T. Robinson, X. Tang, Y. Xiao, D. S. Xue and M. M. Cui, Lithos, 2020, 378–379, 105796.
62. L. X. Gu, Y. J. Chen, Y. C. Xu, X. Tang, Y. T. Lin, T. Noguchi and J. H. Li, Geophys. Res. Lett., 2022, 49, e2022GL097875.
63. R. Pang, J. Yang, W. Du, A. Zhang, S. Liu and R. Li, Geophys. Res. Lett., 2022, 49, e2022GL098722.
64. C. Li, Z. Guo, Y. Li, K. Tai, K. Wei, X. Li, J. Liu and W. Ma, Nat. Astron., 2022, 6, 1156–1162.

---