One bauxite and 10 Fe-lateritic duricrusts were sampled in the Brownsberg mountains. Samples have been collected from well-developed lateritic cover at Leo Falls, while isolated duricrust found next to the road or excavated product from mining activities has been collected (Figure 2).

A first set of samples was collected at Leo Falls: one isolated boulder from the nodular zone (BWG09B, ∼420 m a.s.l.), five duricrust samples (BWG10A, BWG10B, BWG11, BWG12, and BWG13 from 440 to 460 m a.s.l.), and one bauxite sample containing Fe-oxides (BWG14) collected at an elevation of 465 m a.s.l., at the top of the Brownsberg plateau. A second set of samples was collected in a gold mine (average elevation 170 m a.s.l.), where big boulders of duricrust crop out in an open pit ca. 8 m in the land surface. This site will be called the Mine. Three different blocks of duricrust with distinct facies were collected in this site (BWG21A, BWG21B, and BWG21C). These duricrusts are supposed to be eroded material from surrounding plateaus and brought in the depression of the Mine (Supplementary Figure S1B). A last sample was collected on a forest road at ∼250 m a.s.l. where the duricrust is outcropping. This site will be referred to as the Road. The relative position of each site is shown on Figure 2A.

The main characteristics of the samples are summarized in Table 1.

All samples were dried at room temperature and cut into 1- to 2-cm-thick slices. Polished sections were prepared for optical observation. Representative subsamples of nodules and duricrusts (∼1–1.5 cm) were included in Epoxy resin and polished for in situ characterization by scanning electron microscopy (SEM) with energy-dispersive X-ray spectroscopy (EDS), using the backscattered scanning electron mode (BSD) on a Phenom X Pro (Fondis) microscope at 15 keV and 30 s of acquisition time per point (PANOPLY platform, Université Paris Saclay, France).

A slice of each sample was finely ground for mineralogical and geochemical analysis. The bulk mineralogy of duricrust was determined by XRD on non-oriented bulk powder. The data were acquired on a PANalytical X’Pert Pro diffractometer with Ni-filter Cu Kα radiation and carried out at a voltage of 45 kV, a beam current of 40 mA, and a step size of 0.0167° 2θ, with a counting time of 55 s per step for a total counting time of 4 h, and conducted between 3° 2θ and 80°2θ (PANOPLY platform, Université Paris Saclay, France). The extent of substitution of Al for Fe in goethite was calculated using the method of Schulze (1984) based on the relative variability of 110 and 111 position peaks for Al-substitution, using bulk X-ray diffractograms. The estimated uncertainty is ±2.6 mol % Al when using the method of (Schulze, 1984).

The geochemical composition of bulk samples (major and trace elements, including rare Earth elements) was provided by the SARM analytical service of the CRPG (Nancy, France), using the procedure of Carignan et al. (2001). Major element concentrations were determined by ICP-OES (iCap6500 ThermoFisher) after a lithium tetraborate alkali fusion, and trace elements were quantified by quadrupole ICP-MS (iCapQ ThermoFischer). For further information, see https://sarm.cnrs.fr/pages/roches.html.

Based on macroscopic and microscopic (under a binocular microscope) observations of bulk samples, subsamples a priori appropriate for (U-Th)/He geochronology (macroscopically and microscopically homogenous in texture and color with a metallic aspect) were micro-drilled. The obtained material of Fe-oxide was further roughly crushed into several fragments of several hundred micrometers which were subsequently cleaned in an ultrasonic bath with milliQ-water to remove residual clays, without exceeding 30 min and often replacing the water to avoid heating and eventual He loss (Vasconcelos et al., 2013). Representative fragments of dated subsamples were analyzed by SEM-EDS, using the same method as that for bulk fragments.

A set of grains from two selected subsamples (BWG10B_B and BWG21B_B) were also investigated by both µXRD and µXAS at SOLEIL synchrotron (Saint Aubin, France) on DiffAbs and LUCIA beamlines, respectively. These analyses were performed on fine slices (100 µm) of selected aliquots embedded in epoxy resin and carefully polished. Two thin slices of 3 aliquots of the same region of interest were prepared to observed homogeneities or heterogeneities within an Fe-oxide grain. Elemental maps (200 × 200 µm²) were obtained at 7,250 eV using X-ray fluorescence (XRF) with the Flyscan mode developed at SOLEIL (Leclercq et al., 2015), with a step size of 5 µm, allowing us to obtain the distribution of some metallic elements prior to selecting the region of interest for XANES. These maps were extracted using PyMCa software (Solé et al., 2007). Spatially resolved Fe K-edge X-ray absorption Near Edge Structure (XANES) spectra were obtained on the LUCIA beamline (Vantelon et al., 2016). The data were collected at room temperature under vacuum (5 × 10⁻² mbar) using a Si (311) double-crystal monochromator (DCM) with a beam size of 3 × 3 µm². The monochromator was calibrated by setting the first inflexion point of an Fe metallic foil XANES spectrum to 7,112 eV. Five reference spectra were also analyzed: magnetite, chromite, illite, hematite, and goethite. Spectra were collected in both transmission (using a Si diode) and fluorescence (using a 60-mm² mono-element SDD) modes for references and only in fluorescence mode for aliquots. XANES spectra for aliquots and references were extracted using Athena software (Ravel and Newville, 2005). The spatially resolved X-ray diffraction (µXRD) analysis and µXRF were performed on the same material on the DIFFABS beamline, synchrotron SOLEIL. The data were collected at room temperature in a transmission mode using an XPAD 2D detector (µXRD) and in fluorescence mode using a SDD 4-elements detector (µXRF) in an angular domain of 7–40° 2θ, with a step of 0.02°2θ and an acquisition of 10 s per step. Both µXRD and µXRF mapping were performed at an energy value of 18 keV using a Si (111) DCM monochromator. Data acquisition was performed either on point mode, on the same point as LUCIA, or on transect mode (using variable step size, from 30 to 50 µm), crossing aliquots, with a beam size of 10 × 10 µm².

The (U-Th)/He method is based on the quantification of ⁴He produced by α decay of ²³⁸U, ²³⁵U, ²³²Th, and ¹⁴⁷Sm, trapped within the crystal structure of a mineral, that is, goethite or hematite. Indeed, U, Th, and Sm may substitute Fe atoms in the crystal structure of goethite, and the contents of He, U, Th, and Sm are required to determine the age, considering the decay constants of ²³⁵U, ²³⁸U, ²³²Th, and ¹⁴⁷Sm and their radioactive daughter products (Farley, 2002; Pidgeon et al., 2004; Gautheron and Zeitler, 2020). As the He production associated to ¹⁴⁷Sm is low in general compared to the other radioactive isotopes, the method is called (U-Th)/He.

A total of 179 Fe-oxide aliquots from 11 ferruginous blocks of nodule, duricrust, or bauxite samples from 3 sites have been dated. Between 2 and 20 (for the BWG14 crust) grains of ∼100 µg were selected from each macroscopically homogenous subsample under a binocular microscope. Furthermore, around twenty grains of ∼2 mg were used to test He extraction in bigger fragments (e.g., BWG12_E, Supplementary Figure S1). Aliquots were precisely sized, weighted, and loaded in a pure Nb tube. Helium content was determined at GEOPS (Gautheron et al., 2021; Université Paris Saclay, France). The method from the study by Allard et al. (2018) was applied, with only one heating cycle necessary to degas the smallest grains and up to three cycles for the largest ones (∼2 mg). Each capsule was heated using a diode ytterbium laser under vacuum for 30 min. The heating temperature was kept below 1,000°C in order to avoid actinide volatilization, as observed on hematite (Danišík et al., 2013; Vasconcelos et al., 2013; Hofmann et al., 2020). The heating phase was repeated until all ⁴He was degassed, especially for the biggest grains. The extracted gas was mixed in the purification line with a known amount of ³He spike and purified from most of the H₂O, CO₂, H₂, and Ar gases using one liquid nitrogen-cooled traps of activated charcoal, titanium sponge, and ST707 and ST701 SAES getters. These gases and helium isotopes (³He and ⁴He) were measured using a Pfeiffer Prisma Quadrupole mass spectrometer to ensure the purity of the analytical gas (Gautheron et al., 2021). Durango apatite standard materials were degassed to correct for mass spectrometer drift. An internal goethite age and U standard were also used to verify the method. After degassing, the Nb capsules containing aliquots were retrieved and dissolved using the method presented in the study by Allard et al. (2018), with a²³⁵U, ²³⁰Th, and ¹⁴⁷Sm spike volume of 50 µl. The U, Th, and Sm were measured using a high-resolution inductively coupled mass spectrometer Element XR at GEOPS (Université Paris Saclay, France). The analytical uncertainty is lower than 2% (1σ) for U, Th, Sm, and He concentration and 5% for the (U-Th)/He age where equilibrium in the U-Th series is assumed. For ages younger than 0.8 Ma, this statement is not valid anymore, but no correction can be applied as we use ²³⁵U and ²³⁰Th spikes (Farley, 2002) and thus add a 40% uncertainty on the calculated age. As only one aliquot is characterized with an age younger than 0.8 Ma, the age is reported with the <0.8 Ma value. In addition, a correction of the He loss associated with samples of polycrystallinity and Al content has been applied with 5% ± 5% for the Al-rich samples and 15% ± 15% for the Al-poor samples following the proposition of Bassal et al. (2022).

Optical images of thin sections of two representative nodular duricrusts highlight either the large-scale homogeneity of Fe-oxide nodules observed in the BWG10B duricrust sample (Figure 3A) or the large-scale heterogeneity of Fe-oxide nodules observed in the BWG11 duricrust sample (Figure 3B). XRD reveals the presence of goethite, gibbsite, and hematite in all samples except in BWG13 and BWG21B, which do not contain hematite (Figure 4). Some remnants of quartz are found in BWG10A, BWG17, and BWG21A, whereas kaolinite is found as booklets in the BWG10A and BWG13 in Leo Falls and in BWG21B in the Mine (Figures 4, 5E,I). Boehmite is present in the bauxite sample of Leo Falls (BWG14). Finally, Ti-oxides are found as anatase or rutile or both in all samples, except in BWG21B. Rietveld refinement performed on samples BWG10B and BWG11 reveals that goethite is the main component of the duricrust (70% and 51%, respectively), while hematite and gibbsite are less abundant with only 25 and 33% and 3 and 9%, respectively.

The calculated Al-substitution (Table 2) in goethite ranges between 0% (BWG09A, BWG21A, and BWG21C) and 32% (BWG10B) with Al-substitutions generally being >25% in Leo Falls duricrust, except for the massive duricrust (BWG13) and the bauxite (BWG14) in which the Al-substitution in goethite reaches 7 mol% and 0 mol%, respectively. At a microscopic scale, the extent of Al substitution for Fe in goethite is observed by reflected light microscopy with the gray phase being more substituted than the white ones (Figure 3F).

Even though at the macroscopic scale, the selected subsamples seem homogenous and composed of a single Fe-oxide generation (Figures 3C,E), optical microscopic and SEM observations reveal that hand-picked samples can vary in texture and composition and pure single generations were rare to detect (Figures 3, 5). Both XANES and µXRD investigations on selected aliquots of generation from two duricrusts (i.e., BWG10B_B from Leo Falls and BWG21B_B from the Mine) show that the selected grains mostly consist of goethite, even though rare hematite has been observed (Figure 5C). Colloform Fe-oxides, with varying amounts of Al, Ti, V, and Si, are found (Figures 5F,G), indicating different steps of crystallization. For example, the goethitic coatings in the bauxite (BWG14) are at least 3 different generations with distinct Al content (Figure 5F). In addition, the µXRD of the BWG10B_B subsample spectra display a 111 goethite peak shift associated with the substitution of Al for Fe in goethite (Supplementary Figures S5, S6). This finding is consistent with the substitution observed in the bulk sample (Table 2). In the same way, the substitution observed in µXRD of the BWG21B_B sample is similar to the results of the bulk sample, that is, no substitution is observed for goethite.

Kaolinite is preserved in the form of small booklets (ca. 3–4 µm) trapped within the goethite phase found in a few samples, especially in massive duricrust (e.g., BWG13) or massive hand-picked phases (e.g., BWG10A_B, Figure 5A). Kaolinite is embedded in goethite, preserving a clayey and porous texture (Figures 5A,E).

Scarce inclusions of zircon (Figure 5D) and quartz are also observed within Fe-oxides. Yttrium mineral as xenotime (<10 µm) has been reported in the generation BWG10A_A1 (Figure 5B). Still, these inclusions are minor compared to the size of the studied Fe-oxide, largely bigger than 100 µm.

Elemental composition is consistent with the mineralogy, with Fe₂O₃ ranging between 34% (BWG14) and 80% (BWG21B) and Al₂O₃ between 3% (BWG21B) and 37% (BWG14), respectively (Table 2). The REE content of the duricrust is lower than in the UCC (Taylor and McLennan, 1995), varying from 13 ppm (BWG21C) to 132 ppm (BWG11). The REE content is higher in the Leo Falls duricrust (43–132 ppm) than in the Mine (13–23 ppm) or Road (20 ppm) samples. Concentrations of Th and U are the lowest in the Mine samples with 0.33 ppm (BWG21B) and 0.37 ppm (BWG21A), respectively, and maximum in the BWG11 sample, with 14.8 and 2.8 ppm, respectively.

(U-Th)/He ages of Fe-oxides range from 34.4 ± 3.1 Ma to <0.8 Ma, and the age distribution reveals peaks at ∼4 Ma, 10–15 Ma, and ∼30 Ma (Figure 6; Supplementary Figure S3).

The ages of Fe-oxides extracted from Leo Falls duricrusts range from early Miocene to Pleistocene (19.5 ± 2.5 Ma and <0.8 Ma), both measured in the sample (BWG13). Mine samples display older ages, between 4.1 ± 0.6 Ma (BWG21A) and 36.0 ± 5.4 Ma (BWG21C). For the Road duricrust, He ages lie between 1.8 ± 0.3 Ma and 24.5 ± 3.7 Ma. Uranium, Th, and Sm contents are highly variable, ranging from 0.03 to 4.6 ppm, 0.01 to 30.3 ppm, and 0 to 2.8 ppm, respectively.

Most dated subsamples show relatively homogenous ages, though exceeding the analytical uncertainty, but intra-subsample variability is observed for ages and U, Th, or Sm content that cannot be explained by uncertainties of measurement (i.e., BWG10B_B, BWG13_A1, BWG13_A2, BWG14_A, BWG17_B, BWG17_C, BWG17_D, BWG21A_B, BWG21B_A, BWG21B_B, BWG21C_A, and BWG21C_B). The standard deviation on age between aliquots from the same subsample can exceed 6 Ma (i.e., BWG21C_A) but can also be lower than 1 Ma (i.e., BWG09B_A, BWG09B_B, BWG10A_A2, BWG10B_A1, and BWG13_B). Age variability is greater when the average age is older.

The three studied sites yield Fe-oxide He-ages from the late Eocene (36.0 ± 5.4 Ma) to the Pleistocene (<0.8 Ma) (Figure 6). A large variability was noticed between extracted subsamples even from the same duricrust slice but also, in some cases, between replicates of the same subsample. Indeed, dispersed ages are generally associated with distinct U, Th, or Sm content (Figure 6). In fact, many factors are likely to influence the He content in Fe-oxides, for example, helium loss (Hofmann et al., 2017), mineral inclusions richer in He, U, or Th than Fe-oxides (Cornu et al., 2009; Vasconcelos et al., 2013; Monteiro et al., 2014; Riffel et al., 2016), Fe-oxide grain size and crystallite size (Shuster et al., 2005), or mixture of Fe-oxides of different ages (Heller et al., 2022). To avoid U, Th, or Sm loss, the heating phase was carefully undertaken to degas all He content, and temperature was regularly monitored. In addition, internal goethite standards were analyzed to check that no U, Th, or Sm was lost during the heating process.

U–Th-rich mineral inclusions (e.g., apatite, zircon, titanite, xenotime, and monazite) might be responsible for heterogenous age within a single generation, because their He content is higher than that of Fe-oxides, leading to the measurement of older ages (Vermeesch et al., 2007). Although the dated grains seem of relatively high-purity Fe-oxide (i.e., uniform dark color and metallic luster), microscopic observations show mineral inclusions. SEM observations reveal that few xenotime and zircon minerals can be expected in some samples and subsamples (Figures 5B,D). Xenotime is often associated with old metamorphic rocks of low grade metamorphism, that is, greenschist (Spear and Pyle, 2002), which is compatible with the local geology (Eeckhout, 1999; Kroonenberg et al., 2016). Farley and Stockli (2002) showed that xenotime contains high Th and U concentrations but with higher temperature for complete He degassing than Fe-oxides (∼1,300°C for xenotime but ∼950°C for Fe-oxides). Thus, an excess of U and Th compared to He appears in subsamples containing xenotime inclusions, lowering the age of the dated grain. However, no such excess was observed (Supplementary Figure S3) and the yttrium concentrations did not show any enrichment in our subsamples, compared to the upper continental crust (Table2; Rudnick and Gao, 2003). Consequently, we hypothesize that remnant xenotime does not impact the present (U-Th)/He data. In all cases, data with abnormally high Th, U, or Sm compared to the He content but also compared to other grains of the same subsample were not considered (Supplementary Figure S3 , Supplementary Table S4) Inversely, grains with too high He content compared to the U–Th–Sm were also removed for further discussion. In total, these grains represent 14 data over 193.

Additionally, the variability in U, Th, and age found in aliquots from the same subsample can result from different generations of Fe-oxides, or simply different Fe-oxides, that developed tightly and are highlighted in SEM observations by a difference in contrast and texture (Figures 3F, 5A,C,F,G,I) (Monteiro et al., 2014; Heller et al., 2022).

Finally, Shuster et al. (2005) pointed out the potential size effect of the dated object (i.e., subsample) on the disparity of measured ages, considering that too large Fe-oxides are formed by multiple generations and not smaller ones. However, in our study, we have observed that the small-sized (∼500 µm) dated grains from the BWG12 duricrust (Supplementary Table S4) display similar age to larger-sized grains, that is, more than 1 mm (e.g., BWG12_A and BWG12_E10 or BWG12_D and BWG12_G07). In addition, synchrotron analysis also revealed that even at a smaller scale, the selected Fe-oxide grains display heterogeneities (in texture, composition, and presence of different generations). This means that the small-sized grains can also reflect the complex history of the formation of a whole subsample. Nonetheless, the biggest aliquots are prone to containing phase mixtures (e.g., kaolinite) and, consequently, display He deficiency compared to the U–Th–Sm content (Supplementary Figure S3) (Wells et al., 2019). All of these considerations have a potential effect on measured Fe-oxide age (Shuster et al., 2005; Monteiro et al., 2014; Riffel et al., 2016) and confirm that small aliquots are more appropriate for dating than material in the size range of a millimeter. However, we consider that most of the measured U, Th, and He content is robust and reflect Fe-oxide U and Th content with no addition or loss of He (Supplementary Figure S3).

Even though supergene goethite can retain He over geological time scales, 2–20% of natural He can be lost by diffusion for goethite (Shuster et al., 2005). The He diffusion leads to underestimated (U-Th)/He ages due to the polycrystalline nature of supergene Fe-oxides (Figure 5H) (Shuster et al., 2005). Bassal et al. (2022) proposed that He retention is linked to the damage dose and the Al-substitution of the Fe-oxide. In addition, porosity and crystallinity of the different generations can impact He retention and lead to younger ages (Farley, 2018). For instance, the porosity increases when goethite is replacing primary minerals such as kaolinite and are consequently tightly intergrown (e.g., Figures 5A,E,I). But the relatively high homogenous amount of He suggests that our samples were not subject to major He loss due to porosity. Synchrotron XRD analysis on Fe-oxide grains (Supplementary Figure S5, S6) and calculated %Al substitution in goethite from XRD bulk samples (Table 2) highlight significant Al substitutions in goethite from Leo Falls duricrusts (ca. 30% of Al-substitution in the nodular and vacuolar Fe-crusts), that should decrease goethite crystallinity and consequently may increase He retention, due to smaller crystallite size (Fitzpatrick and Schwertmann, 1982; Schulze, 1984; Schwertmann and Carlson, 1994; Vasconcelos et al., 2013). Therefore, we systematically applied the correction for He diffusion to our samples following the consideration of Bassal et al. (2022) (Supplementary Table S4).

The different grains of each subsample usually display homogenous U and Th concentrations within the same range of values and with similar Th/U ratios (Figure 7), attesting to similar weathering history. Even though some selected subsamples display variable Th and U concentrations or distinct ages, the U vs. age diagram shows that the generations are usually distributed between two endmembers, with an old U-poor generation and a young U-rich generation (Figures 6B, 7B,F). This trend suggests a progressive evolution of the system or the mixing of these two phases. Indeed, even though U is highly mobile under an oxidizing environment, U can be easily removed from fluids by Fe-oxides either by adsorption or Fe-substitution (Duff et al., 2002; Missana et al., 2003; Kerisit et al., 2011). Thus, the variability in U, Th, and age reflects a mixture of the young U-rich generation and the old U-poor generation and of different origins. The distinct Th vs. U (Figure 7C) can traduce a different weathering history of different subsamples with either Fe-precipitation from enriched fluids (BWG10A_B) or replacing silicate minerals, inheriting its Th content (BWG10A_A) (Riffel et al., 2016).

Thus, we can conclude that measured (U-Th)/He ages can inform on the long-term weathering history of the Brownsberg area, with several weathering episodes recorded and preserved through the multiple Fe-oxide generations.

The Guiana shield has been exposed to long-term lateritization and bauxitization processes, shaping the landscape with thick lateritic covers sometimes containing Fe- and Al-crusts that evolved through time (Beauvais and Colin, 1993).

The variability of facies and mineralogy at different scales of the studied area reveals the complex history of lateritic and bauxitic duricrusts that experienced several episodes of precipitation of Fe-Al oxides over the Cenozoic. The distribution of the measured ages indicates significant weathering events affecting previously formed secondary minerals, leading to a rejuvenation of the duricrust, due to the possible dissolution of the oldest Fe-oxides and crystallization of new Fe-oxides (Figure 8). The oldest age measured in the present study indicates that ferruginization processes were initiated, at least, ca. 35 Myr ago in this part of Suriname. The predominance of goethite in all the samples testifies to high water activity associated with a humid tropical context with only low seasonal contrast (Tardy and Nahon, 1985) that probably persists and intensifies in the early Miocene with bauxite formation (older than 15 Ma). Low-temperature thermochronological investigations (Derycke et al., 2021) and sedimentary data (Wong, 1986; Wong, 1994) reveal that the northern part of the Guiana Shield basement has been exposed to (near-)surface conditions since the late Cretaceous (c.a. 90 Ma), while our study indicates a maximum age of 36.0 ± 5.4 Ma for secondary Fe-oxide formation. The equatorial position of the Guiana Shield and the supposed globally warm and permanently humid climate prevailing during the 90–35 Ma interval would have favored strong weathering processes (Zachos et al., 2008; Hoorn et al., 2010; Monteiro et al., 2018; Scotese et al., 2021). It is anticipated that older duricrusts would have existed, but it was either not sampled in the present study or it was destroyed by erosion and/or subsequent weathering events. Indeed, at a global scale, compilation of lateritic and bauxitic weathering events shows a higher frequency of young ages than old ones, which probably reflects the successive processes of rejuvenation or erosion (Pidgeon et al., 2004; Shuster et al., 2005; Retallack, 2010; Monteiro et al., 2014; Riffel et al., 2015; Vasconcelos et al., 2015; Allard et al., 2018; Yans et al., 2021).

Only few dating investigations on duricrusts were made in the Guiana Shield and it is consequently particularly difficult to interpret our results in terms of local weathering events (Théveniaut and Freyssinet, 2002; Allard et al., 2018; Heller et al., 2022). Paleomagnetism dating suggests weathering episodes at 50–60 Ma (Bakhuis Mountains, Suriname; Kaw mountain, French Guiana) and ca. 10 Ma in the Guiana Shield (Moengo bauxite deposit, Suriname; Kaw mountain) (Théveniaut and Freyssinet, 2002). However, the distribution of our data suggests four major weathering episodes during the Cenozoic (Figure 8): 1) a lateritization event at ∼35 Ma, 2) a bauxitization episode older than 15 Ma but with no more precise constraints, 3) a lateritization event at 15–20 Ma, and 4) a bauxitization event (6–2 Ma).

1) An Amazonian weathering episode ∼35 Ma has been proposed in the literature (Shuster et al., 2012; Monteiro et al., 2014; Heller et al., 2022), traducing a regional weathering event in South America. However, in Suriname, the late Eocene–Oligocene is known to be the period of bauxite formation during the so called “bauxite hiatus” and must consist of particularly wet climate with a weak seasonal contrast, lower temperatures (still in the range of temperatures for tropical climates), and low sea level, increasing continental incision and drainage (Wong, 1986; Théveniaut and Freyssinet, 2002; Zachos et al., 2008). Probably, the formation of lateritic crust and bauxitic cap are concomitant and, aside from climate, depends on environmental factors such as topography or drainage efficiency.

The Leo Falls lateritic cover shows a bauxitic cap above duricrust. This Fe-crust displays distinct fabric at a macroscopic scale but also different geochemical, mineralogical, and (U-Th)/He ages that highlight the long weathering history in the area. The oldest age found at the top of the Leo Falls formation (∼460 m a.s.l.) in the massive duricrust (BWG13) is consistent with typical downward progression of the weathering front and chemical reorganization (Nahon and Millot, 1977; Lucas et al., 1989; Tardy, 1993; Monteiro et al., 2014). In addition, the Th/U ratio of dated generations is globally higher at the bottom of the duricrust unit (Figure 7D, H, Supplementary Table S4), as Th is relatively immobile compared to U which is mobilized under an oxidizing environment and can reveal successive dissolution/recrystallization processes of Fe-oxides (Riffel et al., 2016). Tardy et al. (1997) considered those massive duricrusts to be the first stage of duricrust formation with Fe-aggradation as goethite under a water-saturated environment, associated with a fluctuating water table. However, they considered that kaolinite cannot be preserved at this stage of weathering. In our study, remnant booklets of kaolinite have been found that suggest the epigenetic replacement of kaolinite by goethite (Nahon et al., 1989) (Figures 5E,I). This implies a rapid ferruginization of the saprolite unit with goethite cement at the groundwater–atmosphere interface (Heim et al., 2006) as suggested by U enrichment compared to Th, (Figure 6B), related to colloform goethite crystallization from U-rich fluids (Riffel et al., 2016). With the downward progression of the weathering front to form younger duricrust, REE are redistributed and enriched toward the lower part of the crust (Figure 8C), as observed for some other trace element in lateritic context, for example, Sc or Ni (Dublet et al., 2015; Chassé et al., 2019). The massive duricrust (BWG13) is thus progressively replaced at its base by younger duricrust in which kaolinite is replaced by the more stable gibbsite following weathering advancement, as shown by the decreasing SiO₂ concentration in other duricrusts (Table 2), while Fe precipitates both as hematite and goethite according to the (bio)geochemical conditions of the weathering system at the time of their formation (Figure 4; Trolard and Tardy, 1989).

The progression of the weathering front, from the top to the bottom, is also accompanied by an increase in Al-substitution in goethite of the younger duricrust (Figure 8), similar to the results of Heller et al. (2022) that give insight into physico-chemical conditions and weathering intensity. The progressive dissolution of kaolinite, which is no longer in equilibrium with the environment, starts to enrich weathering fluid in Al that can be incorporated in Al-goethite, Al-hematite, and gibbsite, more stable, under humid tropical climate (Fitzpatrick and Schwertmann, 1982; Tardy and Nahon, 1985; Fritsch et al., 2005). Aluminum enrichment in goethite can be associated with successive cycles of dissolution and reprecipitation in the same way as metal enrichment found in goethite in other studies (Dublet et al., 2015; Chassé et al., 2019). Interestingly, Al-substitution in goethite coating found in the bauxite accounts for 0 ± 2.6 mol% (Table 2), while the bulk Al content is high, revealing Fe and Al segregation.

This is consistent with the idea of a concomitant formation of the massive crust (BWG13) and of the goethitic coating in BWG14 during the same lateritization episode. However, the bauxite itself (i.e., gibbsitic pisolithes) was not formed at the same time as the Fe-crust. Elsewhere, pisolithic bauxites have been observed in Brazil and Ivory Coast and were interpreted as dismantled material and development in the zone of groundwater fluctuation (Boulangé, 1984; Bardossy and Aleva, 1990). This implies that a first bauxite has been formed earlier, possibly during the main bauxite level (Théveniaut and Freyssinet, 2002), and was eroded and chemically reorganized depending on oxidizing/reducing conditions. The segregation of Al and Fe, that is, the inner concentration of Al and outward migration of Fe from gibbsite, reflects fluctuating chemical conditions and increasing O₂ partial pressure in a well-drained environment (Fitzpatrick and Schwertmann, 1982; Delvigne, 1998). This reinforces the idea that the massive duricrust (BWG13, ∼460 m a.s.l.) and the Fe-cement formation in the bauxite (BWG14, ∼465 m a.s.l.) occurred in a relatively short time period but were not the first stage of lateritic crust formation in the Brownsberg. Similar to the work of (Fritsch et al., 2005; Heller et al., 2022), we conclude that Al-poor goethite is formed at an early stage of profile development. Thus, Al-goethite reflects weathering advancement, Al being provided by the progressive dissolution of unstable kaolinite from BWG13. This weathering episode is also accompanied by uranium and vanadium enrichment, probably associated with successive Fe-oxide dissolution and reprecipitation and progressive fluid enrichment. But the Leo Falls crusts are already rich in those elements and attest to a longer weathering history in the Brownsberg that can be related to the duricrust boulders found in the valley which generally show older (U-Th)/He ages (Figure 6, Supplementary Figure S1B).

The morphology of a sample, its mineralogical and chemical composition, and the U and Th content of Fe-oxyhydroxides can be a tool to investigate weathering and erosion history.