The Balengou deposit is situated in the highlands of West Cameroon, where the humid tropical climate yields thick red soils. Successive magmatic events occurred there since the Tertiary. The Balengou clay deposit (Fig. 1) outcrops in a couple of hills. The eastern hill is abandoned since a few years after a landslide that buried three persons. Its survey through drills with manual auger revealed reworked sandy clays, and blocks of felsic volcanic rocks. The western hill is limited westwards by a cliff likely related to a fault striking N30° E. The eastern mined slope is covered with barren sandy clays and by rhyolite blocks (Fig. 2a and b), whereas the northern slope is partially covered with a basalt flow. On this hill, craftsmen dug galleries following veins of pink and light grey blocky sand-poor clays, which crosscut sand-rich clays (Fig. 2c and d). At the bottom of the hill, a thin horizon of red lateritic soil covers the silty light brown clay (Fig. 2e). Rhyolite blocks are scattered on this red lateritic cover (Fig. 2f). This study is focussed on the western hill. In order to obtain a complete and undisturbed profile, the pit N1 was located on the western flank of the hill (N05°05′49′′; E10°25′55.7′′, altitude 1284 m). It presents four major types of clay described in Fig. 3. Unfortunately, right to a 20-m depth, the light grey clay mined by craftsmen was still not found. There was only red to dark brown sandy clay.

In order to collect fresh samples of mined sand-poor clays and, if possible, its contact with the felsic parent rock, the second pit N2 (N05°05′50.3′′; Longitude E10°25′56.2′′, 1289 ± 5 m altitude) was dug at the heart of the mining zone, where blocks of whitish grey clays are extracted and sold for geophagism. Three types of clays were identified and described in Fig. 3. At this pit too, the parent rock was not reached.

4.2.1 Particle-size distribution and Atterberg's limits

Two groups of particles are distinct in pit N1 on cumulative curves (not shown). Small particles (< 66 μm) are more abundant at shallow depths (< 9 m) than at deep ones (90% vs. 60%). Samples from pit N2 are characterized by huge amounts (90–98%) of small particles (< 66 μm). In Fig. 4a, samples from shallow depth (pit N1) and those of pit N2 are similar, and they overlap with a mode around 10 μm, whereas samples from deep depth display two modes, around 10 and 1000 μm. In Fig. 4b, shallow samples from pit N1 and all the samples from pit N2 fall in the fields of fine clayey silts and silts, whereas the deep samples of pit N1 correspond to coarse silts.

4.2.2 X-ray diffraction

The X-ray patterns of representative bulk samples (Fig. 5) show that the main minerals are halloysite, quartz, cristobalite, and tridymite in some samples, anatase, hematite, associated with gibbsite in samples from shallow depth and illite in samples from the great depth (10–20 m). Some low peaks occurring at 14.98 Å (N14), 15.018 Å (N112), 15.61 Å (samples from pit N1), and at 15.88 Å (N23) and 15.52 Å (N29) (samples from pit N2) are assigned to smectite. The variations would be related to (i) the fluctuations in the number of the water molecules adsorbed on the interlayer cations, (ii) the different substitution rates of interlayer cations, and (iii) likely the nature of the cations at the tetrahedral and octahedral sites. Theoretically, d₀₀₁ spacing of smectite may vary continuously between 10 Å and about 21 Å (Deer et al., 1992, p. 370).

4.2.3 Thermal analysis

Curves of thermal analyses show typical peaks of well-crystallised halloysite. The weak endothermic peak around 45–61 °C (Fig. 6a and b) corresponds to hydroscopic water adsorbed on halloysite. It differs from that of kaolinite, which is at 85–91 °C (Njoya et al., 2006; Pialy et al., 2008). Samples N14 and N23 display an endothermic peak at 276 °C and 271 °C respectively, showing the dehydration of gibbsite (Mackenzie, 1957; Njopwouo et al., 1998). The most important endothermic effect is observed between 490–502 °C, testifying to the volatilization of OH in halloysite. This peak is symmetric in all samples, suggesting that the studied halloysite was well crystallised. For hydrated halloysite (poorly crystallised) the corresponding peak is at 515 °C, accompanied with an exothermic peak at 865 °C, while those of kaolinite are at 560 °C and 920 °C respectively (Quantin et al., 1988). According to Soro (2003), the low exothermic peak at 865 °C might be triggered by Fe diffusion in halloysite network. At higher temperature such as 945 °C in sample N118, there is an exothermic effect interpreted in terms of a spinel phase formation (Njopwouo et al., 1998; Pialy et al., 2008).

4.2.4 Mineral crystal-chemistry characterization by infrared spectroscopy

All samples present remarkable vibration bands around 3691, 3620 cm⁻¹ and 907–910 cm⁻¹ (Fig. 7a), belonging to well-crystallised halloysite (Quantin et al., 1988). The 3652 cm⁻¹ band in N110 and N118 characterizes kaolinite. Weak bands at 3525 and 3441 cm⁻¹ (N14 and N23) indicate the presence of gibbsite. This is comforted by three bands at 749, 668 and 525 cm⁻¹. The H–O–H deformation band at 1630–1635 cm⁻¹ is common (Qtaitat and Al-Trawneh, 2005). The 400 cm⁻¹–1150 cm⁻¹ intervals of Si–O symmetrical elongation vibrations are similar in all samples. Other bands of halloysite and kaolinite are at 749, 688 cm⁻¹. There is a general shift towards low numbers at 525, 457 and 424 cm⁻¹, characteristic of illite and quartz instead of 540, 470 and 433 cm⁻¹, respectively, for halloysite and kaolinite. The bands at 798, 794, 1119 and 1172 cm⁻¹ correspond to quartz (Fig. 7b).

Table 2 shows the chemical compositions of Balengou clays. All samples are made up of SiO₂ (52.2–66.3 wt. %), Al₂O₃ (16.0–27.3 wt. %), Fe₂O₃ (9.5–2.0 wt. %), and minor TiO₂ (1.0–0.2 wt. %) and K₂O (3.0–0.0 wt. %). The SiO₂/Al₂O₃ molar ratio values (3.31–6.67) are higher than those of kaolinite and illite (1.18), and point to free silica, be it quartz crystals or amorphous silica. Fe₂O₃ and TiO₂ contents increase with depth (< 3 wt.% Fe₂O₃ at <10 m depth, 4.9–9.5 wt.% Fe₂O₃ at great depth >10 m), and are positively correlated, suggesting that they are controlled by Fe–Ti oxides such as ilmenite (Table 3). K₂O contents increase with depth from the surface down to 16 m, and the cycle starts again. Conversely, the loss on ignition (L.O.I.) decreases with depth down to 16 m, and the cycle starts again. The values of the chemical index of alteration (CIA = 100*Al₂O₃/(Al₂O₃ + CaO + Na₂O + K₂O)) are high (85–100%). Many trace elements (Ba, Ce, Cr, Cu, Ga, La, Nb, Zn, Zr, Nd, Table 2) show high and varied contents. Ni, V and Pb values are positively correlated with Fe₂O₃ and TiO₂, whereas the Nb contents decrease with a concomitant increase in the Fe₂O₃ and TiO₂ contents (figure not shown). When the sample N120 is discarded, Rb contents are positively correlated with Ba (R² = 0.88) and Sr (R² = 0.94) contents. Other positive correlations are observed between Nb and Ta (R² = 0.98) and between Ta and Zr (R² = 0.87). Except the 10-m depth sample, which displays an exceptional high REE content (e.g., La_N = 3312), there is a decrease in the REE contents downwards in pit N1. Their patterns (Fig. 8a and b) are marked by an important fractionation of LREE/HREE (La_N/Yb_N = 19.1–9.4), and by deep Ce (Ce/Ce* = 0.13–0.73) and Eu (Eu/Eu* = 0.17–0.70) anomalies, LREE being more fractionated (La_N/Sm_N = 4.1–3.5) than HREE (Gd_N/Yb_N = 3.0–1.5). However, the pattern of the 10-m-depth sample shows a constant dip (La_N/Sm_N = Gd_N/Yb_N = 5.3). Samples from pit N2 too are rich in REE and their patterns are similar to those of Pit N1 (Fig. 8 and Table 3). The sample taken at 10 m in depth is rich in REE and we speculate, anticipating the discussion section that it was likely induced by convection of hydrothermal fluids.

By combining the chemical analyses with the data of qualitative mineralogical compositions established through XRD, DTA and TG studies, modal compositions were determined from chemical analyses according to constrained multi-linear calculations (Yvon et al., 1990) using the following basic formula:

M_i = wt.% mineral “i” in the sample, and

P_i(a) = proportion of element a in the mineral i.

Since chemical analyses of minerals were not available, theoretical compositions of halloysite, montmorillonite, illite, anatase, hematite, and gibbsite were used.

The mineralogical reconstitution scheme is as follows:

• • the TiO₂ content, when available, is used for anatase or rutile TiO₂;
• • Cr is used for chromite FeCr₂O₄;
• • Zr is allocated to zircon (ZrSiO₄);
• • P₂O₅ is allocated to apatite (Ca₅(PO₄)₃(OH)) taking into account possible substitutions of Ca by Mn, Ba, Sr, and REE (Deer et al., 1992);
• • gibbsite is calculated using corresponding mass loss measured by GTA;
• • illite (K_1.5-1.0Al₄[Si_6.5-7.0Al_1.5-1.0O₂₀](OH)₄) is calculated using K₂O;
• • MgO is allocated to montmorillonite;
• • the remaining of Al₂O₃ after the reconstitution of illite and montmorillonite is allocated to halloysite, with a hydration coefficient of 0.72 calculated following the method of Njopwouo et al. (1998);

  x=2ΔP1/ΔP3

  (1)

  where ΔP1 = H₂O⁻ and ΔP3 = H₂O⁺ − H₂O*, H₂O* corresponds to the mass loss between 200 and 350 °C of sample N14 (Fig. 5a). H₂O⁺ and H₂O⁻ are from Deer et al. (1992, analysis No. 3, p. 354);
• • Fe₂O₃ after the reconstitution of chromite is used for ilmenite;
• • the remaining SiO₂ is allocated to quartz.

The results are listed in Table 3. The difference between the chemical composition and the sum of the reconstituted minerals is in the range of 0.5 to 4.3 and is usually attributed to the simplification of the reconstitution scheme (Njoya et al., 2006) and even to the presence of colloidal matrix (Pialy et al., 2008). In this study, the use of an arbitrary constant halloysite hydration coefficient for all samples constitutes an additional factor enhancing the difference. Therefore, we suggest that mineral amounts > 10% are affected by an error of ± 5% and results < 10%, by an error of ± 10%. As a whole, the modal compositions show that clay materials from Balengou are very heterogeneous: halloysite amounts range from 10 to 61% in pits N1 and from 48 to 73% in pit N2; illite amounts range from 0.2 to 39%, montmorillonite from 5 to 21% in pit N1. Montmorillonite occurs in trace amounts in pit N2. Samples of this pit, except N23, display modal compositions very close to those analysed by Njopwouo et al. (1998).

Previous works (Njopwouo et al., 1998) and our field observations allow us to claim that the Balengou clay deposit developed at the expense of volcanic acid rocks. These rocks underwent substantial alteration that resulted in chemical modification favoured by the mobility of some major (Si, Mg, Ca, Fe, Al, Ca, Na, K) and trace (e.g., Rb, Sr, Ba, Ni, Co, Sc, V, Cu) elements. As the pits did not reach the bedrock, the determination of the nature of the volcanic felsic rocks and their granitic basement is based on immobile high-field strength elements (e.g., Ti, Zr, Y, Th, Ta, Nb, Hf). Many couples of elements show good positive correlations (Nb–Ta, Ta–Zr, Zr–TiO₂) usually considered as a proof of immobility during alteration, although some of these correlations concerning clays, derived from the alteration of Pan-African granites and of Tertiary volcanic felsic rocks, are a simple coincidence. In such cases, the robust argument is that of Workman and Hart (2005), who established that canonical ratios of immobile elements (e.g., Zr/Th, Ta/Th, Nb/Y) remain constant during magmatic and subsequent metamorphism and/or alteration processes. These findings comforted the choice of the diagram Nb/Y versus Zr/TiO₂ of Winchester and Floyd (1977) for the determination of the protolith (Fig. 9). Samples from shallow depth of pit N1 and all samples of pit N2 plot in the field of commendite/pantelerite (alkaline granite), whereas deep samples (> 6 m) of pit N1 plot in the fields of rhyolite (granite) and dacite (quartz diorite). Strictly speaking, the deposit is made up of many late volcanic dikes intrusive in a granitic dome. Their alteration (hypogene and/or supergene?) yielded silica-free and silica-rich clays with various modal compositions (Table 3). Therefore, similarly to Mayouom (Njoya et al., 2006; Nkoumbou et al., 2009) and Lembo (Pialy et al., 2008) kaolin deposits, sand-poor, cristobalite and tridymite-bearing clays are assigned to a volcanic origin, whereas sand-rich clays are the products of the alteration of the granitic basement.

The problem of meteoric or hydrothermal weathering is inferred here from geological setting, mineral assemblages, and geochemical features. The Balengou clay deposit is under tropical humid climate. Meteoric weathering of rocks here, whatever their nature, yields a thick red ferrallitic soil. This clay deposit constitutes an exception with grey and white clays observed from the surface to a depth of more than 20 m. This exception casts doubt on the meteoric weathering as the main process that has controlled the formation of the clay deposit. If the deposit were resulting from the meteoric alteration of the granitic rock, it would have had a ubiquitous occurrence on the granitic massif (see Tchouankoué et al., 2016). In addition, the meteoric alteration of granitic or gneissic domes is always accompanied by an intense creeping leading to a thin alteration cover at the top. The contradiction between the dip slopes at Balengou and the thick clayey cover (> 20 m) is an additional argument against meteoric weathering as the origin of the clay deposit.

Mineral assemblage in the deposit comprises halloysite + kaolinite + quartz + illite + montmorillonite + hematite + anatase. This assemblage is as common as meteoric weathering or low-temperature hydrothermal products (Pialy et al., 2008; Singer et al., 2004). The occurrence of gibbsite at shallow depth and that of illite in deep layers can be interpreted in terms of meteoritic alteration near the surface, thus as a superposition of meteoric alteration over hydrothermal alteration. This conclusion is strengthened by the absence of the iron hydroxide minerals that prevail in meteoric alteration products under tropical humid climate (Bilong et al., 1992; Bitom et al., 2003; Braun et al., 1990).

Studies based on the chemical compositions of clays have proposed some trace elements (e.g., Fe, Ti, Cr, Nb, P, Ce, Y, La, Ba, Sr, Zr) that can be used to distinguish between hydrothermal and weathering clays (Dill et al., 1997, 2000). Recent studies have proven that trace elements chemical behaviours vary with the deposit; thus, the criteria of Dill et al. (1997) cannot be transferred to classify another deposit (Cravero et al., 2010). Balengou clays contain some amounts of alkalis and alkali earths despite their location on steep slope hills under tropical humid climate. These are good drainage conditions that must have favoured the leaching of these elements during meteoric weathering. All samples present similar REE patterns inferring to a contemporaneous formation under control of the same processes. Generally, weathering and lateritic soil formation processes occur in an oxidizing environment where Ce³⁺ is oxidized to Ce⁴⁺ and therefore behave differently from other REEs, cerium oxides being less soluble than those of other REEs. It can precipitate as cerianite (CeO₂), which is adsorbed on particles, and triggers a positive Ce anomaly (Braun et al., 1990). Conversely, hydrothermal processes usually occur in more reducing conditions (Dill et al., 1997, 2000) where Eu³⁺ is reduced to Eu²⁺ and fractionates differently from other REEs. Clay samples from Balengou display peculiar deep negative Ce and Eu anomalies (Fig. 8). These anomalies reflect a selective leaching of these two elements in their less oxidized forms, since the corresponding salts are more soluble than those of the more oxidized forms. A highly selective behaviour of trace and rare earth elements is common in hydrothermally altered felsic rocks (Parsapoor et al., 2009). It is very unlikely that these anomalies are attributable to the source rocks. They result from a reducing process at a stage of geological evolution of the system. The high values of Ce and Eu anomalies have been assigned to an advanced argillic alteration, which generally operates under low-temperature conditions (Dill et al., 2000; Shao et al., 2015). In Cameroon, clays formed by advanced argillic hydrothermal alteration are described at Mayouom and they too display Ce and Eu anomalies (Njoya et al., 2006; Tassongwa et al., 2014). Alteration conditions were quite similar to those of the Mayouom kaolin deposit: (i) the fault of the western flank and other tectonic events yielded many fractures in the granitic dome, which have acted as pathways for infiltrating meteoric water in the aquifer; (ii) then the magmatic veins emplacement was accompanied by heat and magmatic fluids which were heated and mixed with the ground water; (iii) magmatic fluids rich in sulphides and sulphates induced an acidic environment where Eu and Ce were reduced and fractionated differently from other REEs; (iv) these hot fluids underwent a convective movement during which the felsic volcanic materials and the neighbouring granitic basement were altered into halloysite, which subsequently unrolled into kaolinite. However, the very plausible hypothesis of low-temperature hydrothermal alteration is yet to be constrained by D/H and ¹⁸O/¹⁶O studies.

Most hydrothermally formed clay deposits host other ores of high economic values such as gold, copper, silver, tourmaline, molybdenum, tin, tungsten, uranium (Robb, 2005) depending on the host rocks, the basement, and the geologic events recorded by the deposits (Dill et al., 2015). At Balengou, the protoliths are alkaline–peralkaline trachytes and rhyolites. In such rocks, the expected metals are REE, Zr, Nb, Ta, Hf, Sn, Mo, W, and F (Nkoumbou et al., 1995). The Pan-African basement is a potassic calc-alkaline granite with expected Au, Ag, Cu, Mo, Pb, Zn, Sn, W (Robb, 2005). Along the Cameroon volcanic line, Sn mineralization is described in the Mayo-Darlé granitic massif (Nguene, 1982). Any attempt of ore prospection at Balengou should apply the new approach recommended by Dill et al. (2015), taking into account the neighbouring tertiary Bana complex (Kuepou et al., 2006) and its Pan-African basement.