Introduction

China’s first Mars exploration mission Tianwen-1 has landed and released the Zhurong (ZR) rover in southern Utopia Planitia on the northern hemisphere of Mars on 15 May 20211,2,3,4, within the Late Hesperian-aged geologic unit named the Vastitas Borealis Formation (VBF)1,4,5,6. The VBF is one of the most widespread sedimentary units on Mars that covers most of the northern plains (Fig. 1a)7,8,9. The boundary of the VBF resembles closely the shoreline of a potential ancient ocean due to the cliffs separating the uplands and plains10,11,12. Several fretted terrains within VBF including curvilinear ridges and stepped massifs with terraced slopes are also proposed to be related to the potential ocean10,11,12. Moreover, VBF has kilometer-scale surface roughness relevant to a non-volcanic origin10. Based on those observations, several previous studies proposed that the VBF material is a sedimentary deposit from a standing body of water that might be an Upper Hesperian ocean6,10,11,12,13. A circumpolar ocean was also proposed based on a coupled model with a dynamic ocean and atmosphere14. In addition, the pebbles, cobbles, and boulders in VBF may be deposited by a catastrophic flood15, suggesting that the VBF has flood deposits from outflow channels5,16. However, the polygonal cracking, potential mud volcanoes, and viscous flow features within the VBF also suggest a hypothesis that VBF materials may be the compaction and drainage of a fluid-laden sedimentary unit17,18. In addition, several geological ice-related features are distributed across the VBF including pancake-like ejecta, rampart craters, pitted-wall craters, giant polygonal troughs, and concentric crater fill, which are interpreted to represent an energetic release of volatiles and sublimation of ice-layer from subsurface1,4,19. Moreover, the craters with low depth/diameter ratios and potential mud volcanos in the VBF also suggest a substantial amount of ground ice4,20, consistent with the inferred subsurface water/ice in Utopia Planitia from shallow radar detections21. In summary, direct evidence of the origin of the VBF is still unclear based on available mission data.

Fig. 1: Landing site, traverse map, and MI images of six VBF-ZR targets.
figure 1

a Distribution map of the VBF unit on the Northern Plains of Mars. The map base is a Mars Orbiter Laser Altimeter (MOLA) shaded-relief image. b Traverse map of the ZR rover up to sol 300 overlain on the base image of High-Resolution Imaging Camera (HiRIC) data. The HiRIC raw data is provided by China National Space Administration. The yellow points represent the locations of MarSCoDe targets, and the red points indicate the locations of six VBF-ZR targets discussed in this study. ce The images of VBF-ZR-01, VBF-ZR-21, and VBF-ZR-28 targets collected by the Multispectral Camera (R 650 nm, G 525 nm, B 480 nm). f The MI images of VBF-ZR-01 target (the first target detected at the landing site, 64 × 64 pixels), g, h coarse-grained soils (VBF-ZR-09 at TAR1, VBF-ZR-17 at TAR2, 1024 × 1024 pixels), i fine-grained soils (VBF-ZR-26 at TAR3, 1024 × 1024 pixels), j a layered rock (VBF-ZR-21, 1024 × 1024 pixels), and k a coherent rock (VBF-ZR-28 at TAR4, 1024 × 1024 pixels). In these MI images, the red cross indicates the laser-ablated crater.

VBF materials are geological indicators of the VBF origin. It was hypothesized that the chemical compositions at Viking 1 and Pathfinder landing sites are representative of the average compositions of the VBF22. The soils at these landing sites are rich in S and Cl elements23,24, where the sulfate in soil crusts may be concentrated through intergranular leaching, upward transport, and intergranular evaporative precipitation23. Measurements from Viking 1 and Pathfinder landing sites were unable to provide enough information to constrain the origin of the VBF. The ZR rover landed at a new site in the VBF and has traversed over 1800 meters (Fig. 1b) in the first 300 sols, providing a unique opportunity to investigate the VBF materials. The ZR rover carries a Mars Surface Composition Detector (MarSCoDe) instrument suite, consisting of a Laser-Induced Breakdown Spectrometer (LIBS), a Short-Wave Infrared Spectrometer (SWIR), and a Micro-Imaging Camera (MI), with the intent to investigate the elemental compositions (Si, Al, Fe, Ca, Mg, K, Na, O, H, etc.), mineralogy, and morphology of encountered rocks and soils25,26,27.

This study combined the analyses of LIBS and SWIR spectra from MarSCoDe to investigate the composition and mineralogy of VBF-ZR targets, with the intent to provide ground truth and further constraints for the origin of VBF, and evaluate whether the VBF could have been deposited by an ancient ocean.

Results and discussion

Six VBF-ZR targets and their geomorphology

The terrain of ZR rover traverse is flat, covered by reddish soils and dust. Dark-toned materials are exposed in the surface with dust removed by the turbulent airflow and wheel track (Supplementary Fig. 1a, b), suggesting potential volcanic materials under the covered soils and dust. Four barchan-like Transverse Aeolian Ridges (TARs) are observed by the ZR rover (Fig. 1b). Those TARs have brighter albedo than the surrounding materials and intermittent distribution of dark and bright sand, suggesting that they are covered by dust and hence inactive and ancient28,29. During the first 300 sols, 38 targets have been measured by the LIBS and SWIR spectrometers (Supplementary Table 1), and several targets are near the TARs. This study selected six VBF-ZR targets along the traverse, including coarse and fine-grained soils and two types of rocks, to show their morphological, chemical and mineralogical characteristics (the data of all other VBF-ZR targets are presented in Supplementary Table 1, Supplementary Table 5, Supplementary Table 8, and included in the discussion). The first target (VBF-ZR-01, Fig. 1c, f) is an indurated soil with the surface dust removed by the exhaust plumes during the landing procedure and indurated by wheel tracks. Two types of soils (coarse- and fine-grained) are observed near TARs, including two coarse-grained soils with a grain size of ~6 mm (VBF-ZR-09, Fig. 1g; VBF-ZR-17, Fig. 1h) and a fine-grained soil (VBF-ZR-26, Fig. 1i) with grains smaller than the resolution of MI (~150 μm per pixel, see Supplementary Notes 1). A layered rock (VBF-ZR-21, Fig. 1d, j) between TARs may be an outcrop with irregular layers that are probably related to wind erosion and subsequent deposition of fine-grained materials (e.g., dust). A coherent rock near TAR4 (VBF-ZR-28, Fig. 1e, k) displays channels and angular shapes that are probably caused by the long-termed erosion of wind. In addition, the undetected mineral grains in the layered rock and coherent rock may be less than the resolution of MI (~150 μm) or covered by a weathered (or dust-covered) surface.

Mineral identification by SWIR data analysis

The visible and near-infrared spectra (Fig. 2) acquired by SWIR provide the mineralogy of VBF-ZR materials. After wavelength calibration (Supplementary Fig. 2 and Supplementary Table 2), the SWIR spectra are calibrated using two on-board SWIR calibration targets (white and gray pieces of cal-targets, Supplementary Fig. 3), and similar absorption features are observed (Supplementary Fig. 4). All spectra from the selected six VBF-ZR targets (Fig. 2) exhibit a deep iron-related broad absorption band near ~980 nm30 and a negative slope from ~1600 nm to 2000 nm, which can be assigned to the absorption feature of pyroxene, similar to the Fe-rich orthopyroxene (OPX, Wo2.5En39Fs58.5, No. C1DL14) and clinopyroxene spectra (CPX, Wo42.6En7Fs50.4, No. C1DD148) from the Brown University Reflectance Experiment Laboratory (RELAB) database. The absorption center of VBF-ZR targets (~980 nm) is closer to the center of Ca-rich CPX (Wo42.6, 1015 nm) compared to that of Ca-poor OPX (Wo2.5, 939 nm), suggesting high-Ca pyroxene in these VBF-ZR targets. Several H2O/OH absorption features appear in these SWIR spectra, including a broad band at ~2200 nm (OH combination stretching plus bending band31), ~1400 nm (H2O stretching overtone band31), and ~1900 nm (H2O combination stretching plus bending band31). These three bands are similarly found in the spectra of synthetic allophane (containing 5 wt.% Fe, No. C1JBA75), imogolite (No. C1JBA34), and opal (No. C1JBA33) in the RELAB spectral database. Especially the broad band at ~2227 nm is consistent with the poorly-crystalline phases (i.e., allophane and imogolite). While the band center of this feature is shorter than imogolite and opal (~2260 nm) and longer than allophane (~2202 nm), suggesting a mixture of allophane and imogolite/opal.

Fig. 2: Absolute SWIR spectra and mineral identification.
figure 2

The SWIR spectra before and after continuum removal (CR) of selected VBF-ZR rocks (a) and soils (b) compared to the spectra of pyroxene, imogolite, allophane, and opal from RELAB database. Dotted lines indicate the absorption centers of these phases. The gray regions indicate absorption features of the Martian atmosphere.

Note that SWIR spectra of VBF-ZR targets have triple features at ~1946 nm, ~2000 nm, and ~2040 nm, mainly contributed by the Martian atmosphere32. Therefore, the H2O bands at ~1900 nm are weaker than laboratory spectra of hydrated aluminosilicates due to the potential masked features of the Martian atmosphere. In addition, the similarity of SWIR spectra along the ZR traverse should be caused by the mixture of local materials and widespread dust. While the SWIR spectrum of fine-grained soil (VBF-ZR-26) has a deeper feature at ~939 nm and similar strength at ~1400 nm, ~1900 nm, and ~2227 nm compared to other VBF-ZR targets (Fig. 2b), suggesting more pyroxene proportion in fine-grained soils. The SWIR spectrum of indurated soil (VBF-ZR-01) has a slightly weaker pyroxene feature and slightly stronger OH/H2O bands, indicating a higher proportion of aluminosilicate.

We also compared the spectra of VBF-ZR targets to other phyllosilicates (nontronite, hectorite, chlorite, clinochlore, and montmorillonite) and gypsum from the U.S. Geological Survey (USGS) Spectral Library (Supplementary Fig. 5)33, and found that the spectra of these minerals are different to the SWIR spectra. In particular, the spectra of nontronite (No. SWa-1a and NG-1a) have a deep band at ~2295 nm, which is absent in SWIR spectra. The spectra of hectorite (No. SHCa-1), chlorite (No. SMR-13a), and clinochlore (No. GDS159) exhibit a feature at ~2300 nm, which is not observed in SWIR spectra. Montmorillonite spectrum (No. SWy-1) exhibits an obvious spectral feature at ~2204 nm consistent with the allophane spectrum (~2202 nm) but has a narrow band at ~1400 nm that is different from the broad H2O bands in SWIR spectra. Gypsum spectrum (No. SU2202) has similar features to SWIR spectra including ~1400 nm and ~1900 nm bands, but exhibits an obvious feature at ~1747 nm and a convexity feature at ~2300 nm that are not observed in SWIR spectra. Overall, the distinct spectral features between VBF-ZR targets and these minerals suggest the absence or very few of these minerals in VBF rocks and soils.

In summary, the SWIR spectra of all VBF-ZR targets suggest mixtures of Fe, Ca-rich pyroxene with allophane and imogolite/opal (at variable mixing ratios) in the soils covering the surface unit of the ZR landing site. Especially the layered rock and coherent rock have similar mineral types at the scale of SWIR (~69.4 mm and ~113.1 mm in diameter), which are composed of pyroxene and poorly-crystalline phases with absence or very few other secondary minerals, suggesting these two rocks may keep igneous nature after weak water-rock interaction.

Chemical compositions derived from LIBS spectral analysis

Each LIBS spectrum investigated here was averaged from 60 single laser shots on one target. The continuum background in these spectra was removed (Supplementary Fig. 6). Cluster analysis was firstly conducted by plotting the spectral line intensity ratio of (Fe + Mg+Ca+Na+Al)/Ti vs. intensity ratio of Si/Ti from the normalized LIBS spectrum of each VBF-ZR target (details in Methods), together with the corresponding spectral line intensity ratios of seven silicate calibration targets (see Supplementary Notes 1)25 carried by MarSCoDe (Fig. 3c), which can directly provide the qualitative classification of VBF-ZR targets.

Fig. 3: LIBS spectra and chemical compositions of selected VBF-ZR targets.
figure 3

a Selected and normalized emission lines of Si, Fe, Mg, H, C, K, Na, Ca, and Al of six targets. b Major element concentrations in VBF-ZR targets, where the error bars are RMSEs of the multivariate model used in this study. c Cluster analysis using LIBS spectral line intensity ratios of the selected VBF-ZR targets and seven silicate cal-targets carried by MarSCoDe. The black line is a linear fit result using scatters of basalt, hypersthene, and norite cal-targets (R2 = 0.89). d Relative concentrations of eight major elements at four landing sites normalized to the average Martian soil compositions (data in Supplementary Table 7). The error bars are the ratios between error values and average Martian soil compositions.

Chemical compositions of VBF-ZR targets were derived from these LIBS spectra using a multivariate model, which was built using the relative intensities of 14 emission lines of nine major elements (Supplementary Table 3) in the normalized ChemCam laboratory spectra of 295 standards34. The performance of this multivariate model is evaluated by comparing the Root Mean Square Error (RMSE) values for eight major oxides (Supplementary Table 4) using four sets of LIBS spectra taken by different LIBS systems (details in Methods, Supplementary Fig. 7). The obtained similar RMSE values from different database support the suitableness of the developed multivariate model in this study. The proportion of dust in VBF-ZR soils (Supplementary Table 6) is assessed using compositions of indurated soil (VBF-ZR-01) and global average dust (using ChemCam results at Gale crater, Supplementary Table 7) as endmembers based on mass-balance analysis. Subsequently, we calculated the modal mineralogy using CIPW Norm Calculation (Supplementary Fig. 8 and Supplementary Table 8)35. It should be noted that CIPW does not take into account sedimentary processes.

The relative intensities of LIBS emission lines (Fig. 3a) and derived compositions (Fig. 3b) show the variation of major elements among the selected six VBF-ZR targets. Between the two rock targets, the coherent rock (VBF-ZR-28) has a mafic composition (Fig. 4a) with a relatively high concentration of SiO2 (51 wt.%) and more pyroxene (~40%, Supplementary Fig. 8) than the layered rock (SiO2 = 42 wt.%, pyroxene = ~25%) and occurs adjacent to the cluster of the basalt cal-targets (Fig. 3c). The layered rock (VBF-ZR-21) contains a higher concentration of mafic components (23 wt.% of FeOT and 5 wt.% of MgO) than the coherent rock (18 wt.% of FeOT and 3 wt.% of MgO). It occurs near the cluster of Fe-Mg-rich norite calibration targets in Fig. 3c, representing a mafic composition. Note that the H emission line (with good S/N) was observed in the LIBS spectrum of layered rock (Fig. 3a). All eight rocks exhibit Fe, Ca, Ti-rich (FeOT = 22 ± 3 wt.%, CaO = 15 ± 1 wt.%, TiO2 = 1.6 ± 0.3 wt.%) and Mg-deficit features (MgO = 5 ± 1 wt.%) compared to the Mars global dust. The Ca-rich compositions in these rocks are consistent with the high-Ca pyroxene determined by their SWIR spectra. Most rocks (including the outcrop of VBF-ZR-21) have similar chemical compositions within the error of the multivariate model and similar mineral types although they are detected at different locations, suggesting they may reflect the local materials. Thus, the compositions of these rocks can reflect the special volcanism at the Tianwen-1 landing region.

Fig. 4: Geochemical results of selected VBF-ZR targets.
figure 4

a Total alkalis (Na2O and K2O) vs. silica diagram of VBF-ZR targets that does not take into account sedimentary processes. Data of soils, dust, and rocks acquired by other missions are from Bridges et al.75, Berger et al.43, Lasue et al.40, and Mangold et al.55. Compositions of fine-grained soils at Gale crater (red circle) are from PDS Geosciences Node at Washington University in St. Louis. b, c Ternary diagrams of Al2O3-(CaO+Na2O)-K2O and Al2O3-(CaO+Na2O+K2O)-(FeOT+MgO), with CIA and MIA values on the vertical axis. ChemCam results from Murray formation and Yellowknife Bay in Gale crater are respectively from Mangold et al.55 and McLennan et al.53. The average compositions of Mars crust are from Taylor et al.51. The big size of VBF-ZR data represents the currently used percent level (wt%) accuracy for them, which will be improved along with the Tianwen-1 mission development.

For VBF-ZR soils, average compositions of soils can be characterized by multiple LIBS shots at one single point because the new portion of the soil will be ablated by the laser due to the disturbs of the shock wave caused by previous shots. The chemical compositions of soils are similar and the slight differences between them are within the RMSEs of the multivariate model in our work (6 wt.% of SiO2, 3 wt.% of FeOT, and 2 wt.% of MgO). The coarse soils at TAR1 (e.g., VBF-ZR-09) and fine-grained soils (e.g., VBF-ZR-26) at TAR3 have similar compositions (e.g., 4–7 wt.% MgO and 18–24 wt.% FeOT), which are adjacent to the cluster of basaltic cal-targets on MarSCoDe (Fig. 3c), suggesting that the coarse soils could be aggregates or precursor of fine particles36,37. The fine-grained soil (VBF-ZR-26) has a stronger H emission line (Fig. 3a) that can be assigned to adsorbed water38 at grain surfaces and interstitial gaps. However, another coarse-grained soil (VBF-ZR-17) exhibits distinct compositions compared to fine-grained soils analyzed by the ZR rover, which has a higher SiO2 (55 wt.% vs 44–52 wt.%), lower FeOT (15 wt.% vs 18–22 wt.%), and MgO (3 wt.% vs 4–6 wt.%; Fig. 3b and Supplementary Table 5). In cluster analysis, it lies in a zone that is distanced from the cluster of basalt, Fe-Mg-rich norite, and Si-rich andesite cal-targets (Fig. 3c). The slight difference in compositions of these three soils near different TARs may reflect their different proportions of silicic and mafic minerals at the submillimeter scale, suggesting their different mixing proportions of local materials and Mars global dust (22–75% in Supplementary Table 6) probably caused by sorting processes of aeolian sand along a transport pathway. This phenomenon is similar to the mixing effects of fine- and coarse-grained soils at Gale crater36,37.

Interestingly, the indurated soil (VBF-ZR-01) has the strongest emission lines of Si and Al while relatively weak of Fe, K, and Na, indicating its distinct compositions from all other VBF-ZR targets. In cluster analysis (Fig. 3c), data from this target occur quite far from other VBF-ZR targets. It also has more SiO2 (59 wt.%) and Al2O3 (15 wt.%) compared to other targets (SiO2 < 55 wt.% and Al2O3 < 13 wt.%), which is consistent with more aluminosilicate in this target determined by SWIR spectra. In addition, the H emission line (S/N ~ 3) in its LIBS spectrum (Fig. 3a) also suggests the existence of H2O/OH-bearing species in this dust-removed indurated soil. Compared to the basaltic indurated soil of the Rocknest Sand Shadow in Gale crater (CIA = ~33.4%)39, the overall compositions of this target may be affected by chemical weathering. Considering the surface dust was removed from this target (by the turbulent airflow during landing), its distinct compositions may reflect the real compositional characteristics of local soils (beneath surface dust) at the Tianwen-1 landing site.

In addition, sulfur emission lines (e.g., S(II) at 543.0 nm, 545.5 nm, 548.0 nm, 553.8 nm, and 566.6 nm) are absent in the LIBS spectra of VBF-ZR targets (Supplementary Fig. 9). This may be due to the low response efficiency at the end wavelength region of MarSCoDe CCD (540–850 nm). It may be also caused by the average of 60 LIBS spectra at each point, which weakens the potential S lines. Thus, the sulfur in the detected targets may be lower than the limit of detection of MarSCoDe. This is consistent with the absence or very few sulfates in VBF-ZR targets determined using SWIR spectra.

Comparisons with GRS data and fine-grained soils at other landing sites

Two comparisons were made to evaluate the reliability of the above compositional measurements. Firstly, the chemistry of the fine-grained soil target (VBF-ZR-26) was compared with the finest fraction of soils and dust (1.4–2.5 μm) at other landing sites (Gale crater40, Gusev crater41, and Meridiani Planum42) on Mars that have similar compositions due to the global circulation of surface dust by regional and global dust storms43,44,45,46. It was found that VBF-ZR-26 (fine-grained soil) has a very similar composition to the average Martian surface soils, and the discrepancies are not very different between the VBF-ZR-26 target and the surface dust at the other three landing sites (Supplementary Table 7 and Fig. 3d). While the VBF-ZR-26 target has slightly higher CaO and lower MgO than dust, which may be contributed by the 23% Ca-rich and Mg-poor local materials (e.g., the indurated soil has 11 wt.% CaO and 2 wt.% MgO in Supplementary Fig. 10) mixed with 73 % widespread dust (e.g., dust in Gale crater has 7.2 wt.% CaO and 8.5 wt.% MgO, shown in Supplementary Table 7). The mixture of local materials can also be reflected by the larger grain size of VBF-ZR fine-grained soils (~150 μm) than dust (1.4–2.5 μm). Secondly, an average of derived compositions of all VBF-ZR targets was compared with the published compositions (SiO2, FeOT, Na2O, and K2O) acquired by Mars Odyssey’s Gamma-Ray Spectrometer (GRS)47 at the Tianwen-1 landing site (Supplementary Table 9). In this limited comparison, the direct ratios of FeOT/SiO2 from both data sets are almost the same (0.43 for MarSCoDe and 0.41 for GRS), although the range of ZR rover observations is <2 km while the GRS data were extracted from a region of 5° × 5° (~300 km at the equator)47.

The two comparisons above show that the quantitative compositional information of VBF-ZR targets can be extracted using the multivariate model developed in this study, and can be compared with the chemical compositions of widespread dust and fine-grained soils from other rover missions at different landing sites. The slightly distinct compositions of VBF-ZR targets compared to dust can reflect the Ca-rich and Mg-poor characteristics in local materials. Although at this early stage of the Tianwen-1 mission and different detection distances of VBF-ZR targets, we limit ourselves to the percent (wt%) level for compositional accuracy (Fig. 4a and Supplementary Table 5) for eight major oxides, that accuracy will be improved along with the mission development as more targets are analyzed.

Degree of chemical alteration

The degree of chemical alteration of VBF-ZR targets is characterized by the Chemical Index of Alteration (CIA) and the Mafic Index of Alteration (MIA). CIA is defined as the molar ratio of Al2O3 and the total of Al2O3, Na2O, K2O, and CaO48, reflecting the release of Ca, Na, and K relative to Al. MIA is used to characterize the loss of Mg and Fe from mafic minerals in a reducing environment49. CIA and MIA can be integrated into Al2O3-(CaO+Na2O)-K2O and Al2O3-(CaO+Na2O+K2O)-(FeOT+MgO) diagrams derived from LIBS spectral analyses (Fig. 4c, d).

In general, igneous rocks with a low degree of alteration have low CIA values. For instance, igneous rocks at Gale crater and Ares Vallis (in VBF) respectively have CIA values of 41.7 ± 6.3% (using ChemCam data50) and 37.4 ± 3.5 % (using APXS data24). In addition, the estimated Martian crust compositions also have a low CIA value of ~36.9% (using APXS data51,52). For comparison, the CIA values of Martian dust are measured around 36–37% at Gale crater40, Gusev crater41, and Meridiani Planum42. While the sediments from chemical weathering of mafic materials in open systems have higher CIA values than the parent rock (~40–45%, even ~50–55%)53,54, and this trend from chemical weathering can be recognized in ternary diagrams (e.g., black arrows in Fig. 4b, c). For example, some altered materials at Yellowknife Bay at Gale crater exhibit extremely high CIA values (<70% based on ChemCam data, a green-colored patch in Fig. 4b)53. Most Murray formation rocks in Gale crater with substantial alteration also exhibit high CIA values of >45%55 (a pink-colored patch in Fig. 4b). However, sedimentary materials with a high abundance of alkaline components (e.g., K-feldspar) would exhibit low CIA values similar to the CIA values of igneous rocks, which could not affect the CIA values of VBF-ZR targets due to their low abundance of alkaline components (2–3 wt% of Na2O + K2O). VBF-ZR rocks have low CIA values of 28% ± 8% (the layered rock) and 30% ± 8% (coherent rock), suggesting a small amount of amorphous aluminosilicates mixed with high-Ca pyroxene and thus these rocks may be igneous rocks with a low degree of alteration and long-termed erosion of wind. Alternatively, low CIA values can be contributed by the local Ca-rich materials mixed with Mars global dust (CIA = ~36%).

Specifically, the VBF-ZR-01 target has the highest CIA value of 39% ± 9% as well as high SiO2 and Al2O3 concentrations, which is consistent with its SWIR spectrum, and shows a higher portion of allophane and imogolite mixed with mafic materials due to the dust removal by the turbulent airflow during landing. Moreover, the VBF-ZR-01 target occurs at the upper end of a distribution line of VBF-ZR targets along the weathering trend (Fig. 4b, c) defined by Hurowitz et al.56 and Ehlmann et al.57, which is consistent with the results that VBF-ZR-01 contains more incipient weathering products based on its LIBS and SWIR spectra. With the surface dust being removed, this target can generally represent the local soils, a mixture of aluminosilicates and igneous parent materials or dust.

Discussion

The VBF was previously proposed to be formed from the sedimentary deposits in a standing body of water either from an Upper Hesperian ocean6,10,13,14, outflow channels5,16, or compaction and drainage activities17,18, which experienced subsurface ice-related processes1,4,19,20 and was covered by globally homogenized Mars surface dust. Based on SWIR analyses, the VBF-ZR targets are composed of amorphous aluminosilicates (allophane and imogolite/opal) and High-Ca pyroxene that are widely distributed in soils and rocks along the traverse path of the ZR rover. Most materials at the Tianwen-1 landing site have very low CIA values based on their LIBS spectra, also suggesting a mixture of local basaltic materials (e.g., VBF-ZR-09 and VBF-ZR-26) and Martian surface dust41,43,51. In addition, the indurated soil (VBF-ZR-01) target with the dust removal by the turbulent airflow during landing may represent the intrinsic VBF soils broadly distributed in the Tianwen-1 landing site, which has the highest CIA (39% ± 9%) with high Al2O3 abundance and the most amorphous aluminosilicates among all VBF-ZR targets.

Terrestrial allophane and imogolite are mainly associated with highly leached and weathered volcanic materials, which are characteristic of soils formed from the rapid weathering of volcanic glass in cold and low W/R systems31,58,59,60 or in cold and icy terrains61. The release of Ca and Mg cations from basaltic glass can provide neutral to slightly acidic conditions (pH = 5–7) for allophane and imogolite formation60,62,63. Some allophane can also be precipitated from soil water (e.g., stream-deposits from a river in Japan64) with silica and hydroxy-Al ion65. On Mars, the allophane and imogolite have been identified at Mawrth Vallis, and were proposed to be generated from limited aqueous alteration of basaltic glass from rainy environment or melting snow/ice63. Based on those, we suggest that the amorphous aluminosilicates along the traverse path of the ZR rover are probably generated from the dissolution of basaltic glasses in an open system with a low W/R condition at low temperature. Alternatively, these aluminosilicates are precipitated from the soil solution with high silica and hydroxy-Al ion. Moreover, the absence of large amounts of other phyllosilicates (e.g., montmorillonite and nontronite) and sulfates also suggests a limited water-rock interaction which may be generated from limited water or short-term soil solution.

Overall, in-situ observations of ZR rover suggest that VBF-ZR materials still keep the igneous nature with limited alteration features, and the products are then suffered a long-time wind aeolian erosion leading to mixing and sorting along a transport pathway. The geochemistry and mineralogy of local rocks and soils support that the Tianwen-1 landing site (to the depth that ZR rover can measure) within VBF may have the deposits generated by chemical weathering from extensive volcaniclastic soils by limited water (e.g., rainy or melting from frost/snow/ice) or from a short-lived standing body of water (e.g., melting of glaciers or sourced from southern highlands). In terms of the four hypotheses of the VBF origin, our study suggests that the limited water interaction under cold conditions at the Tianwen-1 landing site mostly supports the hypotheses relating to the ephemeral frozen ocean and sublimation.

Methods

SWIR spectra processing

Mineralogy information of VBF-ZR materials can be investigated using SWIR spectra. The original radiance data acquired by SWIR are converted to reflectance after wavelength calibration, and continuum spectra are subsequently deducted to highlight absorption features. The processing steps of SWIR spectra are as follows.

1) Wavelength calibration. The wavelength shift of SWIR spectra is determined using the absorption features of the Martian atmosphere. The SWIR spectra of VBF-ZR are firstly converted into reflectance by dividing the solar spectrum provided by Gueymard66, and absorption centers are identified and compared to the features provided by CRISM (Compact Reconnaissance Imaging Spectrometer for Mars) Analysis Toolkit (CAT)32 (Supplementary Table 2). We find SWIR spectra shift ~10 nm to a longer wavelength than CRISM spectra (Supplementary Fig. 2a), which is subsequently corrected (Supplementary Fig. 2b).

2) Transition into absolute reflectance. The SWIR radiance spectra are converted to absolute reflectance using Exp.1. Relative reflectance spectra of VBF-ZR targets are acquired from their radiance spectra divided by the radiance of two calibration targets (white and gray pieces), with the hypothesis that calibration targets are total reflectors. Subsequently, the relative reflectance data are converted to absolute reflectance by multiplying the absolute reflectance of SWIR calibration targets (Supplementary Fig. 3). The pre-processing of SWIR spectra can reduce the effects of the absorption features of Martian atmosphere.

$${R}_{SWIR}(Absolute\,reflectance)=\frac{{I}_{SWIR}}{{I}_{SWIRC}}{R}_{SWIRC}(Absolute\,reflectance)$$
(1)

where ISWIR and ISWIRC are radiance data of VBF-ZR targets and two calibration targets (white and gray pieces) collected by SWIR spectrometer. RSWIRC represents the absolute reflectance of calibration targets given in Supplementary Fig. 3.

3) Continuum removal. The continuum removal of SWIR spectra is conducted to highlight the spectral absorption features (Supplementary Fig. 11). The continuum is defined as a polyline using four tie points at ~1187 nm, ~1251 nm, ~1653 nm, and ~2236 nm. The linear continuum at a lower wavelength than the first tie points (~1187 nm) is defined as an extended line of the straight line between the first(at ~1187 nm) and the second tie points (at ~1251 nm).

LIBS spectra processing

The chemical composition of a target can be derived from the emission lines in its LIBS spectrum, which was averaged from 60 spectra of 60 single laser shots. In this study, a multivariate model is built using the intensities of selected emission lines in the ChemCam LIBS database collected in the laboratory. The model is applied to interpret the LIBS spectra collected using a LIBS system at Shandong University, China (SDU-LIBS) to verify the applicability of the model among different LIBS systems and the uncertainties. Finally, the chemical compositions of VBF-ZR targets are derived using this model. The RMSEs of these different LIBS databases were compared. The processing of VBF-ZR-LIBS spectra has the following steps.

1) Continuum removal. The emission lines in LIBS spectra are affected by background continuum emission caused by bremsstrahlung and ion-electron recombination processes34,67. The continuum background spectra are obtained by the interpolation of local minima using the spline function and are subtracted from the LIBS spectra. The instrument white noise spectra are not considered due to the reduced effects of instrumental noise in terms of the averaging of 60 LIBS spectra from a single target. The LIBS spectrum before and after continuum removal is provided in Supplementary Fig. 6.

2) Identification of the emission lines. The 14 emission lines of major elements (Si, Ti, Al, Fe, Mg, Ca, Na, K, H, C, and O) are identified and selected according to the NIST database68 (Supplementary Table 3). Those emission lines have high-intensity values and do not overlap with other lines (as shown in Supplementary Fig. 12). For elements with multiple lines (Fe, Mg, Ca, Al, Na), two independent and strong emission lines are selected. Only one strong emission line is selected for Ti due to its low abundance in most rocks and soils. For Si, only the neutral emission peak at ~288.3 nm has a high intensity and is not affected by the emission lines of Fe. Only one emission line is selected for K and O at ~766.7 nm and ~777.6 nm, respectively, because no other strong emission lines can be seen as candidates. A summary list of selected emission lines is shown in Supplementary Table 3. Moreover, obvious self-absorption is not observed in all selected emission lines, which may be due to the low atmospheric pressure on Mars69,70.

3) Cluster analysis of VBF-ZR targets and cal-targets. Cluster analysis was conducted by plotting the spectral line intensity ratio of (Fe + Mg+Ca+Na+Al)/Ti vs. the line intensity ratio of Si/Ti from the normalized LIBS spectra of VBF-ZR targets and cal-targets (andesite, hypersthene, basalt, norite, montmorillonite, and nontronite). These intensity ratios correspond to the chemical space of (olivine + plagioclase + clinopyroxene)/Ti vs Si/Ti, which defines a 1:1 basalt line and a direction of chemical weathering according to igneous fractionation and crystal sorting in magmatic systems71.

Multivariate model using ChemCam LIBS database

The ChemCam LIBS database of 295 standards is selected to build a multivariate model and interpret the LIBS spectra acquired by MarSCoDe. Those ChemCam LIBS spectra are acquired using the ChemCam engineering model at a 1.5 m standoff distance. LIBS spectra are acquired at five different spots on each sample with 50 laser shots at each location, and 250 spectra are obtained and averaged for each sample34. The steps to build our multivariate model are as follows.

1) Selection of ChemCam database. The ChemCam standards are composed of igneous, sedimentary, and metasedimentary samples as well as some doped mixtures. Some types of these samples (e.g., sulfates and chloride) are not observed at the Tianwen-1 landing site, and hence not selected. In addition, we did not select the samples with the total oxide abundance of eight major elements (SiO2, TiO2, FeOT, MgO, Al2O3, CaO, Na2O, K2O) larger than 110 wt.% and lower than 90 wt.%. Subsequently, the compositions of the 295 selected samples are normalized to the total oxide abundance of 8 major elements. Finally, the LIBS spectra of these ChemCam standards are adopted in our model.

2) Earth-to-Mars correction. LIBS spectra are different in terrestrial laboratories and on Mars due to different plasma temperatures achieved by the flight model (FM) laser on Mars and in the laboratory. The Earth-to-Mars correction is performed to compensate for these spectral differences. The selected LIBS spectra are multiplied by the pixel-by-pixel correction factors developed by the LIBS spectra of ChemCam Calibration Targets (CCCT) obtained on Mars and in the laboratory34.

3) Normalization of the emission lines’ intensity. Normalization was conducted using the total intensity of selected emission lines of each sample. Subsequently, the intensities of elements with multiple emission lines (e.g., Fe405 and Fe438) are summed. Finally, nine intensities for nine major elements (Si, Ti, Fe, Mg, Al, Ca, Na, K, and O) are acquired for each sample.

4) Build a multivariate model. The normalized emission intensities for nine major elements are used to build a multivariate model using a polynomial fitting algorithm to predict the oxide abundance of eight major oxides (SiO2, TiO2, FeOT, MgO, Al2O3, CaO, Na2O, and K2O). Polynomial fitting is selected with the intent to reduce the disturbance effects in LIBS intensities between elements (i.e., chemical matrix effects). The performance of this model is determined by RMSE between actual and derived compositions as follows.

$$RMSE=\sqrt{\frac{1}{n}{\mathop{\sum }\limits_{i=1}^{n}({y}_{ip}-{y}_{i})}^{2}}$$
(2)

where n is the number of samples, yi and yip are actual and derived compositions.

Deriving compositions of VBF-ZR targets

(1) Response transformation from MarSCoDe to ChemCam

In order to properly extract chemical compositions of the VBF-ZR targets based on the ZR-LIBS spectra using the model that we built using the ChemCam database, the response difference between these two LIBS systems should be removed. First, the radiance calibration methods and the units of LIBS intensity are different for ZR-LIBS and ChemCam. The system response values of ChemCam (Digital Number, DN) are converted into radiance in photons·s−1 cm−2 sr−1 μm−167,72. While the unit of ZR-LIBS radiance is W·m−2 sr−1 nm−125, which is converted into photons·s−1 cm−2 sr−1 μm−1 by dividing the photon energy as follows.

$${I}_{ChemCam}(photons)=\frac{{I}_{MarSCoDe}(W)}{E}$$
(3)

where IChemCam and IMarSCoDe respectively represent the LIBS intensities of ChemCam and MarSCoDe. E indicates the photon energy, which can be expressed as73

$$E=\frac{hc}{\lambda }$$
(4)

where h is the Planck constant, c is the speed of light, and λ is the wavelength.

Therefore, the response transformation can be expressed as

$${I}_{MarSCoDe}(photons)=\frac{{I}_{MarSCoDe}(W)\,\lambda }{hc}$$
(5)

(2) Intensity transformation using LIBS spectra of a norite cal-target

The norite sample is the only on-board calibration target carried by both ChemCam and MarSCoDe (Supplementary Table 10)25, and its compositions are provided by Clegg et al.34. Thus, the LIBS spectral intensities of this sample can be used for cross-calibration between these two flight LIBS systems. The same set of 14 emission lines in norite spectra (both ZR-LIBS and ChemCam) is obtained and normalized using the same method for the 295 spectra of the ChemCam LIBS database. The different intensity values of emission lines between ZR-LIBS and ChemCam (Supplementary Fig. 13a) may be due to the different radiance calibration methods of these two LIBS systems25,72. Specifically, solid angle and imaged source area of the MarSCoDe telescope are not considered in ZR-LIBS radiance calibration, while these parameters are constant for all bands in each ZR-LIBS spectrum and can be removed in normalization.

The ratios between ChemCam and MarSCoDe intensities (Supplementary Fig. 13b) are used to further transform from MarSCoDe to ChemCam response. This type of transfer method has been applied in the Earth-to-Mars correction of ChemCam and SuperCam34,74. The intensity ratio of Ti(I) line at ~335.5 nm is larger than the values of other elements, which is ~10 times higher than the ratio of Na(I) line at ~819.1 nm. This variation may be due to different laser irradiance on targets (~6.4 GW cm−2 of MarSCoDe25, ~3.1 GW cm−2 of ChemCam72), which can be reduced by intensity transformation from MarSCoDe to ChemCam as follows.

$${I}_{ChemCam}(photons)={I}_{MarSCoDe}(photons)\frac{{I}_{ChemCam\_norite}}{{I}_{MarSCoDe\_norite}}$$
(6)

where IChemCam_norite and IMarSCoDe_norite are the intensity values of selected emission lines from the ChemCam and MarSCoDe-LIBS spectra of norite cal-target, respectively.

3) Deriving compositions of VBF-ZR targets

Similar to ChemCam data processing, nine normalized spectral intensities for nine elements are extracted from each ZR-LIBS spectrum and are used to derive the abundance for oxides of eight major elements (SiO2, TiO2, FeOT, MgO, Al2O3, CaO, Na2O, and K2O) of VBF-ZR targets using the multivariate model that we built.

Validation of the multivariate model

It is critical to evaluate whether the multivariate model that we built using the ChemCam spectral database is applicable for ZR-LIBS compositional extraction. The results of this evaluation are shown in Supplementary Fig. 7.

Our multivariate model is firstly applied to the laboratory LIBS spectra of 295 ChemCam standards, and the ChemCam spectra of seven ChemCam Calibration Targets (CCCT). In Supplementary Fig. 7a, b, the actual compositions of these samples and the derived compositions (in wt.%) using our multivariate model are shown along a 1:1 line. We noticed that ~92% of 295 ChemCam standards and all data from CCCT fall into a buffer zone of ±10 wt.%. Supplementary Table 4 lists the RMSEs using our multivariate model (e.g., 6.3 wt.% and 6.0 wt.% for SiO2 in 295 ChemCam standards and CCCT) are comparable to the published values by ChemCam and SuperCam teams (i.e., 5.3 wt.% and 6.1 wt.% for SiO2 using ChemCam model34 and SuperCam model74).

Secondly, our multivariate model is applied to the laboratory LIBS spectra of 168 SDU-LIBS standards collected using SDU-LIBS system (the other four samples are the same standards in 295 ChemCam standards used for LIBS system response transformation, Supplementary Notes 2), and also applied to the LIBS spectrum of the norite cal-target collected by ZR-LIBS system. Using the same methodology (as Supplementary Fig. 7a, b) to plot the actual and derived compositions in Supplementary Fig. 7c, d along a 1:1 line, we found that >72% data (but lower for FeOT of ~43%) from 168 samples and all data from norite sample fall into a buffer zone of ±10 wt%. Similarly, comparable RMSEs (Supplementary Table 4) are obtained from those LIBS spectra of four different LIBS systems (e.g., 6.3 wt.%, 6.0 wt.%, 9.1 wt.%, and 7.4 wt.% for SiO2 in 295 ChemCam standards, CCCT, 168 SDU-LIBS standards, and norite cal-target). Slightly higher RMSEs are obtained for the spectra in SDU-LIBS standards, due to the low Signal to Noise Ratio of the averaged spectra (30 spectra per sample vs. 250 spectra per sample in the ChemCam spectral database), especially for Fe405, and Na819 emission lines.

Overall, the comparable RMSEs (Supplementary Table 4) indicate that our multivariate model (with proper pre-processing) works well on the spectra from different LIBS systems, and accurate elemental compositions of VBF-ZR targets can be derived from ZR-LIBS spectra.