Our samples of the SCS crusts and nodules were collected by seafloor trawling on the Guangzhou Marine Geological Survey research vessel “Haiyangsihao” during the 2011-2012 SCS survey. Detailed information of the samples is given in Guan et al. (2017b). Polymetallic crust samples (ST1, ZSQD251A, ZSQD253A, ZSQD42A, and HYD66-2) have platy or crustal Fe-Mn encrustations and hard substrate of altered basalt and carbonates (e.g., reef limestone). Most samples have friable thin (0.1-1.9 cm) Fe-Mn encrustations and are (dark) brown or black. Polymetallic nodule samples (STD275, ZJ86, and HYD104) are spherical, ellipsoidal, elongate, conodont, or irregular. They have Fe-Mn encrustations around the cores (iron oxides, clayey, muddy, or altered basalt). The nodules grow on top or surrounded by coarse sandy to clayey sediments ( Table 1 ).

The samples are cut along the growth profile to make polished thick sections (thickness ~100 μm), and then observed with a ZEISS Imager and an A2m optical microscope (reflected light). Photographs were processed with the Axio Vision 4.8 software.

The analysis was conducted at the School of Marine Science, Sun Yat-sen University (SYSU), using a ThermoFisher DXR2xi laser Raman micro-spectrometer equipped with a 532 nm argon ion laser and edge filters. Both spot analysis and mapping were performed. The map was formed by multiple line analyses. The analysis conditions include 3 mW laser energy, 6 μm beam size, 0.2 s exposure time, 10 scans, 600 × 600 μm sweep range, and 8 μm mapping resolution. Data processing utilized the ThermoFisher’s OMSNICxi and OMSNIC software.

The analysis were conducted at the School of Marine Science, SYSU, using a Rigaku D/max Rapid II micro-area X-ray diffractometer (Japan). In the XRD analysis of bulk samples, the crusts and nodules were washed with ultrapure water and dried in a blower for 24 h (drying temperature set at 30°C), then ground to 200-mesh using an agate mortar. In the in-situ micro XRD analysis, particle samples with a diameter of <3 mm were observed with an optical microscope equipped on the X-ray diffractometer, and then adjust the selected microarea to the center of the field of view.

The instrument parameters include Mo target and 0.1 mm collimator, 50 kV voltage, 30 mA current, and XRD data acquisition on a 2-D image plate detector. All samples were analyzed twice under the same conditions, and the signals were combined to improve the signal-to-noise ratio. The 2-D data were then converted to 1-D spectra using Rigaku’s 2DP software. Identification of the major minerals was done with the PDXL2 software.

The high-resolution TEM (HRTEM) imaging and selected-area electron diffraction (SAED) analyses were completed at the Analysis Center of SYSU on a JEOL JEM-2010 transmission electron microscope (Japan), and at the Guangzhou Institute of Geochemistry, Chinese Academy of Sciences, on a FEI Talos F200S transmission electron microscope (USA).

(1) JEM-2010: The powder sample (200-mesh) was dispersed in alcohol and dropped onto a Cu grid (with a carbon film microgrid), using a disposable dropper and then allowed to dry naturally. The TEM operating voltage is 200 kV, and the instrument is equipped with an INCA energy spectrum. The analysis site was selected first under microscope at low magnification, and observed with energy-dispersive X-ray spectroscopy (EDS) at higher magnification, followed by high resolution phase analysis at ultra-high magnification.

(2) Talos F200S: The samples were made by secondary embedding-ultra-thin sectioning. Each sample was dispersed in ultrapure water by ultrasound, which was then dropped onto the pre-consolidated resin, dried with a blower, and re-embedded in resin afterward. The resin was heated with a blower, solidified, and cut with a Leica ultrathin microtome (USA), and then placed on a Cu mesh (with carbon-coated film) and dried. The instrument operates at 200 kV with maximum 1.5M× magnification and ≤ 0.25 nm resolution. The instrument is equipped with an EDAX energy spectrum with ≤ 136 eV energy resolution. Fast Fourier transform (FFT) and other TEM image processing were performed with the TIA TEM Imaging and Analysis software (version 4.15).

The encrustations of the SCS polymetallic crusts and nodules contain light and dark Fe-Mn mineral layers, with medium-coarse crystalline silicates and bioclasts being cemented or encapsulated by Fe-Mn minerals ( Figures 2 , 3 ). Microstructural features of the samples are very similar, with primary growth structures being lamellar, columnar, or mottled, and secondary ones being concentric or dendritic.

By comparing the Raman spectra (Mn-O peak signal) maps and microscopic photographs, we find that the Mn-O peak intensity correlates positively to the density of growth structure ( Figure 4 ). The Mn-O peak intensity is similar to the light and dark changes of the growth layer, which indicates that the changes in the growth structure and the growth layer of the crust (core) are the alternating changes in the mineral composition.

Previous studies have shown that Fe-Mn crusts and nodules consists mainly of Mn (Fe-vernadite and todorokite) and Fe (goethite, lepidocrocite, amorphous ferrihydrite) phase minerals, together with minor clay minerals, calcite, apatite and detrital quartz and feldspars (Rajani et al., 2005; Pattan and Parthiban, 2007; Xu, 2013; Zhou, 2016; Zhong et al., 2017). Characteristic XRD peaks of different Mn minerals in our samples include ( Figure 5 ): 0.239 and 0.140 nm (vernadite), 0.72, 0.35 and 0.24 nm (birnessite), and 0.97, 0.48, 0.24 and 0.14 nm (todorokite and buserite). The characteristic diffraction peaks of todorokite are at 9.56 Å (100) and 4.78 Å (covering (200) and (002)). Meanwhile, the 4.26 Å (100)_quartz, 3.34 Å (011)_quartz, 2.45 Å (110)_quartz, and 1.37 Å (203)_quartz diffraction peaks represent the better-crystallized quartz, whilst the 3.84 Å (104)_calcite, 3.02 Å (113)_calcite, 2.08 Å (202)_calcite, 1.90 Å (018)_calcite, and 1.86 Å (116)_calcite diffraction peaks represent Mg-calcite, which is mainly derived from biological detritus (mainly foraminifera). Most samples have an undulating background peak with peaks centered at ca. 2.4 and 1.4 Å. This background represents the presence of amorphous ferrihydrite (Lee et al., 2016). Our results show that the SCS polymetallic crusts and nodules comprise mainly Fe-vernadite (δ-MnO₂), quartz, feldspars, and amorphous ferrihydrite. This is consistent with the findings by Guan et al. (2017b), although some minerals (e.g., apatite) reported by that study were not found here.

Under the microscope, the SCS polymetallic crusts and nodules display black Fe-Mn encrustations with grape-like or massive structure. In situ micro-XRD analysis results indicate that the Fe-Mn crusts are mainly composed of Fe-vernadite and todorokite, with many light-colored minerals (mainly quartz and feldspars) interspersed among the Fe-Mn minerals ( Figure 6 ).

In the TEM dark field, there are numerous nano-particles in the samples ( Figure 7 ). Based on the XRD results, we further analyze the SAED patterns and the TEM-EDS data, and the major minerals in the samples are recognized. In particular, the form and composition of birnessite and todorokite are very similar, so it will be confirmed by SAED pattern after each EDS analysis ( Figures 7 , 8 ). In samples ST1 and ZSQD253A ( Figures 7A–C ), there are numerous nano-particles of quartz, thin clay mineral layers, network of Fe-vernadite (δ-MnO₂), and amorphous flocs of FeOOH. Sample ZJ86 contains birnessite and todorokite as well ( Figures 7F–H ). The EDS results show that the birnessite is mainly associated with Na, K, Ca, and Ni, while todorokite is mainly associated with Ca and Ni.

We found that independent Ti minerals are common in the SCS crusts and nodules ( Figure 8 ). In the TEM dark field, Ti minerals exist in a flocculent form, which is very similar to the colloidal morphology of amorphous FeOOH and Si-Al hydrate. Meanwhile, Ti minerals occur in the form of nanofragments in the high-resolution bright field. The EDS results show that the flocculent Ti minerals contain certain Fe and Si. Its high-resolution photograph and the FFT results show [_37_4] the crystallographic stripes in (_403), (111), and (51_2) under the crystallographic band axis, and the measured crystallographic spacing under the Z-axis is 3.58 Å. Therefore, the Ti minerals were determined as hydrated Ti oxide (e.g. H₂Ti₃O₇, Kataoka et al., 2013).

In this study, the growth profiles of the SCS polymetallic crusts and nodules were found to contain different microstructure assemblages ( Figure 9 ). Xu (2013) summarized that the growth structures of the crusts and nodules can be divided into primary and secondary ones, while different structure assemblages representing different growth rates and bottom currents. The alternating growth of different Fe-Mn layers implies fluctuations in the growing conditions, which may reflect periodic changes of seawater environment in the South China Sea.

Based on the growth structure characteristics, we suggest that the SCS polymetallic crusts and nodules may have formed in four growth stages ( Figure 9 ). Columnar, palisade, and dendritic structures are predominant in the early growth stage. In this stage, the dendritic structure indicates the inconsistent growth rates at various locations on the same growth surface, while the formation of columnar structure indicates a low hydrodynamic growth environment and a higher formation growth rate (Xu, 2013). The second growth stage developed mainly lamellar (and some mottled) structures, which indicates that the bottom current and oxidation conditions are weak (Xu, 2013). The third growth stage mainly developed mottled structures, which reflects strong bottom current and turbulent seawater environment (Xu, 2013; Guan et al., 2017b). Laminar and mottled structures are mainly developed in the fourth growth stage, indicating that the bottom current and oxidation environment in this stage are more reduced than the previous stage (Guan et al., 2017b).

We also found that the SCS polymetallic crusts and nodules from near the continental margin contain mainly mottled structures. This indicates that they were grown in a more turbulent depositional environment, and were strongly influenced by coastal currents and terrestrial debris (Zhong et al., 2017). In contrast, polymetallic crusts and nodules near the central basin (e.g., HYD66-2 and HYD104) mostly develop laminae and columnar structures. This indicates that they grew in a more stable bottom flow, and were less influenced by terrestrial input (Zhong et al., 2017).

Mineralogical characteristics are important in determining the genetic type of Fe-Mn crusts and nodules (Bau et al., 2014) and the ore-metal enrichment mechanism. The XRD spectral peaks of Mn minerals are all broad, indicating poor crystallization, and the presence of todorokite suggests that the SCS polymetallic crusts and nodules may have grown in a short period of sub-oxic seawater environment (Conrad et al., 2017).

According to the intensity of XRD ( Figure 5 ) and chemical composition (Guan et al., 2017b), samples from the northern SCS samples (e.g., ST1, STD275) have large amount of detrital quartz and feldspar. This suggests that these SCS crusts and nodules obtain many terrestrial materials. Samples ZSQD42A, and ZSQD253 from around the Zhongsha Island areas contain large amounts of calcite and biological debris ( Table 1 ), suggesting input from marine organisms on the polymetallic crust/nodule mineralization. In contrast, samples near the central basin (e.g., HYD66-2 and HYD104) contain significant amounts of Mn minerals and less detrital minerals (quartz and feldspar, Figure 5 ), consistent with their columnar and dendritic growth structures that suggest a stable growth environment and insignificant terrestrial input. Todorokite in crust samples HYD66-2 ( Figure 5 ) and ST1 ( Figure 6 ) indicates the prevalence of todorokite in the SCS polymetallic crusts and nodules, yet its content is likely very low (undetected by XRD). Todorokite in crusts and nodules usually forms in suboxic condition (Conrad et al., 2017), the low content of todorokite in the SCS polymetallic crusts and nodules suggested that they have experienced suboxic environment. As mentioned above, the SCS crusts and nodules are of hydrogenetic origin. Therefore, we propose that the SCS crusts and nodules have been exposed a suboxic environment (Guan et al., 2019), but the process may have been of short duration or not suboxic enough to affect their overall hydrogenetic signatures. According to Figure 6 , the SCS crusts contain a low content of todorokite expect for HYD66-2. The reason for more todorokite in HYD66-2 is that the location of this sample grows near the central basin and more possible to be exposed to a suboxic environment.

Silicon can precipitate directly into the polymetallic crusts and nodules as quartz and clay minerals, and occur in seawater in colloidal precipitates or as anion complexes (Akagi, 2013; Tréguer and de la Rocha, 2013). EDS spectra of Si is present in all the samples ( Figure 7 ), which indicates good supply of Si during the crust/nodule growth and accompanied by the precipitation of Fe-Mn oxides into the crusts and nodules, and the layered clay minerals often coexist with the Mn minerals. This may be due to the similar layered structure of vernadite and birnessite (both clay minerals). In addition, the mineral interlayers of birnessite are negatively charged (Halbach et al., 2017) and can accommodate cations (e.g., K⁺, Ca⁺) and even water molecules (Manceau et al., 2014; Lee and Xu, 2016b). This shows strong adsorption properties similar to those of clay minerals. In addition to the layered minerals, many flocculent minerals can be seen in the TEM dark-field phase, which are mostly colloidal minerals of Si, Al and FeOOH.

The hydrated oxide of Ti (H₂Ti₃O₇) was first reported by Izawa et al. (1985), and its crystal structure determined by Kataoka et al. (2013). H₂Ti₃O₇ is a hydrate of TiO₂, and occurs mainly in colloidal form in aqueous solution. In this study, we first observed Ti minerals in the SCS polymetallic crusts and nodules, and the EDS result shows that these Ti minerals mainly contained Ti and O. However, the SAED pattern could not match any Ti-O minerals data. Nevertheless, the evidences in this study were not enough to support the proposal of a newfound Ti mineral. Because the nodules and crusts grow in the seabed environment, and previous studies have already suggested that Ti entered the nodules in a colloidal chemical way (Koschinsky and Hein, 2003). We speculate that these independent Ti minerals might be a kind of Ti hydroxide, which was temporarily indicated by TiO₂·xH₂O in this study. These Ti minerals perhaps indicated that Ti could form colloidal minerals in seawater and enter directly into the crusts and nodules.

The surface of hydrated FeOOH and Ti oxides is positively charged in seawater, which facilitates selective adsorption of some seawater anions or anion complexes (Koschinsky and Hein, 2003; Jiang et al., 2011). In addition, colloids can aggregate (through covalent bonds) and eventually co-precipitate (Bruland et al., 2014). In this study, we inferred that these colloidal hydrated Ti and Fe oxides were concentrated into the polymetallic crusts and nodules (together with colloidal aggregates, incl. SiO₂-xH₂O or Al₂O₃-xH₂O in seawater), as evidenced by the widespread presence of colloidal flocs of Fe, Ti, Si, and Al in the samples (Koschinsky and Halbach, 1995). Fluvial and aeolian input can carry substantial terrestrial substance (incl. Si and Al) to the South China Sea, which dissolves rapidly and forms colloids in the seawater, and then re-precipitates in Fe-Mn oxides/hydroxides. Therefore, terrestrial influence on mineral composition of the SCS polymetallic crusts and nodules includes not only the direct detrital input, but also the later re-precipitation of dissolved materials or colloids.

Previous studies have shown that vernadite, birnessite, and todorokite are the three major Mn-phase minerals in marine Fe-Mn crusts and nodules (Usui et al., 1989; Bodeï et al., 2007). In general, vernadite (δ-MnO₂) is a good indicator mineral for hydrogenetic origin, while todorokite and 10 Å phyllomanganate often occur in diagenetic and hydrothermal origins respectively (Takahashi et al., 2007; Manceau et al., 2014; Pelleter et al., 2017). For example, some Fe-Mn growth layers at the bottom of oceanic polymetallic nodules (buried in sediments) were subjected to early diagenesis to form todorokite, as reported in the Pacific and Atlantic Fe-Mn nodules and crusts (Marino et al., 2018). The Mn minerals in the SCS crusts and nodules are mainly disordered vernadite (δ-MnO₂) with minor birnessite and buserite. This result suggests that the SCS polymetallic crusts and nodules are mainly of hydrogenetic origin, with some Fe-Mn growth layers possibly affected by diagenesis (Guan et al., 2019).

Under the TEM bright field, we found direct evidence for the conversion of vernadite (δ-MnO₂) to birnessite and todorokite ( Figure 10 ). Vernadite (δ-MnO₂) is an amorphous phyllomanganate that is randomly stacked in the c-axis. This series of δ-MnO₂ minerals is mainly divided into three species (Halbach et al., 2017): (1) 7 Å vernadite with [001] and [002] crystal plane spacing of ca. 7 and 3.5 Å, respectively; (2) 10 Å vernadite with [001] and [002] crystal plane spacing of ca. 10 and 5 Å, respectively; (3) Fe-vernadite (δ-MnO₂) without crystalline surface spacing, which is most common in the SCS polymetallic crusts and nodules (referred to as Fe-vernadite hereafter).

Birnessite are ordered, layered, well-crystallized δ-MnO₂ (Giovanoli, 1980). Since there are two species of 7 Å and 10 Å vernadite, the ordered birnessite is divided into 7 Å birnessite and 10 Å buserite. Birnessite is structurally more stable and occurs mostly in diagenetic (or mixed hydrogenetic-diagenetic) Fe-Mn crusts and nodules, which are often used as precursors in experiments on synthetic todorokite (Liu et al., 2005; Atkins et al., 2016). 10 Å buserite is more common in diagenetic Fe-Mn nodules, and can be used as a precursor (Bai et al., 2002) or intermediate transition phase (Bodeï et al., 2007) in experimental synthesis of the structurally less-stable todorokite.

The ideal todorokite has a 3-D tunnel structure, consisting of three M-O octahedra (arranged as tunnel) and three M-O octahedra (arranged as tunnel bottom) at common angles, with the tunnel and tunnel bottom oriented along the a-axis and c-axis, respectively (Yu, 1979; Cui et al., 2009; Qiao et al., 2016). Its [001] and [002] crystal plane spacing is ~9.7 Å and ~4.8 Å, respectively, and its diffraction pattern has a strong peak at ~2.4 Å and weak peaks at ~2.2 Å and ~1.7 Å.

As shown in Figure 10 , Region 1 of the sample is a random stack of Mn oxides, whose diffraction patterns show two broad rings at ca. 2.4 and 1.4 Å, which indicates that δ-MnO₂ is weakly crystallized. In contrast, the minerals in Region 2 have a bamboo-leaf shape, and their diffraction patterns show four broad rings at ~4.9 Å, ~2.4 Å, ~2.1 Å, and ~1.4 Å, indicating that the nano-minerals in this region are relatively well crystallized. The HRTEM photos show clearer crystal plane stripes ( Figures 10C, D ), with crystal plane spacing of ca. 6.2 and 7.4 Å (birnessite) and 4.9 and 9.8 Å (todorokite).

To elucidate the transformation processes between these different Mn minerals, we combined our trace element data with published ones (Golden et al., 1987; Cui et al., 2005; Cui et al., 2006; Cui et al., 2010; Atkins et al., 2014; Lee and Xu, 2016b), and proposed a phase transformation model for Mn minerals in the SCS polymetallic crusts and nodules ( Figure 11 ): the interlayer space of 7 Å birnessite contains single H₂O molecules and/or cations (e.g., Na⁺, K⁺, Mg²⁺, Ca²⁺, Ni²⁺) (Manceau et al., 2014; Lee and Xu, 2016b; Halbach et al., 2017). Some cations (e.g., Ni²⁺) can increase the layer spacing of birnessite after entering the layers of birnessite, which eventually leads to the conversion of 7 Å birnessite to todorokite (Golden et al., 1987; Atkins et al., 2016). This transformation process was also reported from Fe-Mn nodules in freshwater lakes (Lee and Xu, 2016b). Previous studies have shown that synthetic phyllomanganate mineral phases can be converted to todorokite by atmospheric reflux, hydrothermal treatment, or aging treatments (Cui et al., 2005; Cui et al., 2006; Cui et al., 2010; Atkins et al., 2016), and can be promoted by neutral-alkaline and reducing conditions and the presence of clay minerals. Since todorokite crystallization in the SCS polymetallic crusts and nodules is poor, we suggested that the SCS crusts and nodules were subjected to short period of suboxic conditions and diagenetic process.

The TEM-EDS results show that both birnessite and todorokite contain high Ni content, while vernadite (δ-MnO₂) contains only low trace element contents, which can be explained by their crystal structures. In the birnessite structure, some divalent cations (e.g., Co²⁺, Ni²⁺, Zn²⁺, Fe²⁺, Cu²⁺) can replace Mn⁴⁺ to enter the Mn mineral interlayer (Halbach et al., 2017). In the todorokite, Mg²⁺, Ca²⁺, Co²⁺, or Ba²⁺ are the main cations in the crystal lattice (Post, 1999; Manceau et al., 2014), while other cations are bound to the lamellar sites of Mn (incl. co-hexagonal, co-bilateral, co-trilateral, co-dual, and co-trigonal with Mn-O octahedra) (Cui et al., 2009; Qiao et al., 2016). In addition, vernadite has negative charge in seawater (Koschinsky and Hein, 2003), which facilitate its adsorption of free cations and cation complexes (e.g., VO²⁺) from the seawater (Wehrli and Stumm, 1989).