Geochemical signatures of rare earth elements and yttrium in the vicinity of an ion-adsorption type deposit: roles of source sediment control

1 School of Water Resources and Environmental Engineering, East China University of Technology, Nanchang 330013, P. R. China State Key Laboratory of Nuclear Resources and Environment, East China University of Technology, Nanchang 330013, P. R. China 3 School of Water Resources and Environment, China University of Geosciences (Beijing), Beijing 100083, P.R. China UniLaSalle, AGHYLE, Beauvais, France

Previous researches on the concentration and distribution of REE+Y in AMD of various mine sites show that the REE+Y concentrations in AMD are several orders of magnitude higher than the median values of near-neutral waters (Ayora et al., 2016), and the normalized REE+Y patterns (with respect to standards such as North American Shale) are characterized by middle REE (MREE: Sm to Dy) enrichment over light REE (LREE: La to Nd) and heavy REE (HREE: Ho to Lu). The MREE-enriched REE+Y patterns were shown to exhibit differently from patterns of rock/sediment that AMD flow through and have been documented in many studies (Olías et al., 2005;Silva et al., 2009;Delgado et al., 2012;Sahoo et al., 2012;Stewart et al., 2017). However, LREE- (Bozau et al., 2004) and HREE- (Medas et al., 2013) enriched patterns were also observed in AMD, and enrichments of both MREE and HREE in acidic systems have been reported as well (Gammons et al., 2003;Sharifi et al., 2013;Migaszewski et al., 2014).
In spite of the known normalized REE+Y patterns in AMD systems, the underlying mechanism regulating the formation of the patterns is poorly understood. Considering the extremely low concentrations of REE+Y in rainwater (Zhang and Liu, 2004;Zhu et al., 2016), REE+Y are ultimately sourced from rocks which fluids move through. Studies have proposed a preferential leaching of REE-bearing solid phases to be a cause for the enrichment patterns (Sun et al., 2012a;Wallrich et al., 2020 and references therein). In this framework, corresponding REE+Y-enriched mineral phases should present in the bulk solid samples. Sequential extraction techniques using carbonate and silicate with solution pH of 1.6, 3.6 and 5.5 were performed to stimulate formation process of acidic groundwater, showing that the sequential extracts have similar REE patterns to those of AMD (Worrall and Pearson, 2001).

Results of the author's experiments suggested that REE signatures in acidic water
were a mixture of leachates of various sedimentary components. This means that REE+Y enrichment patterns are of source-control and local strata lithologies play a significant role in contribution of REE+Y to AMD. Furthermore, the conservative behaviors of REE+Y in the acidic solutions have accounted for their long-distance transport (Verplancket al. 2004). However, other studies indicate that process-derived reactions such as fractionation of colloidal complexes (Åström and Corin, 2003), adsorption-desorption on surface coatings (Åström, 2001), differences in solution complexation (Dia et al., 2000;Zhao et al., 2007), and preferential precipitation of certain REE during formation of secondary minerals (Elderfield et al., 1990;Leybourne et al., 2000) are responsible for the REE+Y enrichment patterns. A recent study suggested that the roles of geology in controlling REE+Y distribution and patterns in AMD might be complicated by geochemical, biological and environmental factors (León et al., 2021). The unique properties of REE+Y constitute an opportunity to study the source, origin, fate and transport of AMD, which is important in addressing the mechanisms of AMD formation and contamination.
Therefore, the objectives of this study are to: (i) investigate the concentrations and distributions of REE+Y in AMD, (ii) characterize the geochemical controls on REE+Y sources and fractionation patterns in AMD, and (iii) assess the feasibility of REE+Y recovery from AMD.

Method and materials
2.1 Regional hydrogeological settings The mining site (longitude: 114°47′28″, latitude: 24°54′25″) is located in the Jiangxi province, southern China. It has a subtropical monsoon climate characterized with plentiful rainfall and abundant sunlight. The highest temperature recorded historically is 39 °C and the lowest is -3.1 °C. The annual average temperature is 19.7 °C. Precipitation ranges from 1058 to 2190 mm (annual average 1608 mm), dominating from April to July (accounts for 60%). Annual evaporation is 1497 mm (ranging between 1349 mm and 1619 mm), intensely occurring in July and August.
The drainage systems are well developed in the basin and have five main rivers, which are connected to the tributaries of Dongjiang River running from south to north with a flow rate of 13.93 m 3 /s. Strata of Senonian to Quaternary well outcrop in the mine area. The Cambrian, Devonian and Jurassic strata have the largest exposed area and are mostly distributed in eastern and southern part. The mine area is hosted in Jurassic to early-Cretaceous igneous rocks, which are mainly composed of acidic granites and cover a total area of 359 km 2 . The main rocks are micaceous granite, granite porphyry, monzonitic granite and diorite . The REE+Y are originally born in two types of rocks, namely muscovite granite and biotite granite, where HREE and LREE have distinct fractions in the completely weathering layers .
The HREE were shown to account for 75% to 89% (wt) with an average ratio of HREE to LREE (HREE/LREE) being 4.4 in the residues of muscovite granite. The average HREE proportion was 61%, and the HREE/LREE value was 1.6 for biotite granite counterpart. Among the HREE, the Y occurs as Y2O3 and increases with increasing grade of rare-earth rocks with the largest fraction being commonly >50%. These rocks have experienced different extends of weathering. The thickness of strong-weathered layer ranges from 5 to 25 m. Minerals including quartz, feldspar and mica mainly occur in the weathering residues. Those rock bodies having undergone intense weathering exhibit sound permeability and become water conduits.
Weathering fissure water is the major groundwater source in the mine area. It is mainly hosted in the fissure network of medium-grain micaceous granite and largely belongs to phreatic water. The weathering crust has a depth of 7.6 to 25.4 m. The coefficient of permeability ranges between 0.13 and 0.19 m/d. Hydrochemical facies include Ca-SO4, Ca-SO4-HCO3, Ca-Cl and Ca-HCO3-Cl with total dissolved solids (TDS) from 17 to 170 mg/L. Aquifers are mainly recharged by precipitation and discharged via springs and/or seepage. Runoffs are controlled by distributions and occurrences of fissures. Hydrologic units form in the piedmont area leading to interactions between surface water and groundwater. There is no centralized groundwater source for drinking water supplies 100 km downstream of the mining area. Hence, groundwater is utilized in a decentralized way by inhabitants living nearby. Surface water is used for industry, irrigation and landscape entertainment.
Open pit was used to be an active mining operation in exploring rare-earth metals in this mine site; whereas now, underground shafts are mainly being used as well.
During the former operation, the intense weathered crusts are excavated resulting in outcrops of the bedrock underlain, which thus modify the pristine hydrogeologic connections. The underground shaft technology is not easy to implement, because leaching solutions infiltrate into grounds and percolate with water flowing, generating environmental contaminations. Ammonium has been reported to exceed the threshold recorded in groundwater quality standard due to usage of ammonium sulfate as a leaching liquor (Liu et al., 2019). Additionally, slags and waste dumps produced over mining activities are potential risks to the surrounding rivers, especially during rainy seasons. In this respect, a sewage treatment plant (STP) was established in the piedmont area to receive fluxes of leaching solutions from the mine (Fig. 1). The treated water is discharged into rivers down-gradient.

Sample collection
Sixteen water samples were collected in July 2020. Among them, four were pristine mine water samples (MW1 to MW5), six were effluents of STP undergoing nitrification-denitrification treatment procedures (NDT1 to NDT6), four were effluents going through coagulating-precipitation treatment processes (CPT1 to CPT4), and one was a well water (WW) collected from a local well in a village near the mine. The mine water is discharged after being treated via nitrification-denitrification and coagulating-precipitation processes. All water samples were filtered with 0.22 μm pore size cellulose acetate membrane filters. Samples for cation and trace element analysis were stored in 50 mL high-density polyethylene (HDPE) bottles and were acidified to pH< 2 with HNO3. Samples for anion analysis were stored in the HDPE bottles without acidification. Samples for dissolved organic carbon (DOC) analysis were sampled in 30 mL amber glass bottles and immediately acidified to pH < 2.0 by dropping 1:9 (volume) H2SO4. Samples for REE and Y analysis were sampled in 500 mL HDPE bottles following an acidification with HNO3. All sampling locations are shown in the Fig.1.
Solid samples including two rock samples, three surface sediment samples and five sludge samples were collected. Rock samples (R1 and R2) were sampled from the mining area. Three surface sediment samples (MW1-SS,  were collected corresponding to the locations where mine water samples (MW1, MW2 and MW3) were taken. Sludge samples (MW4-SG, MW5-SG and CPT1-SG) were sampled by taking the precipitates from bottom of the settling pond and creekbed, and NDT2-SG and NDT3-SG were sludge samples collected by using a long sucker. All the rock, sediment and sludge samples were stored in the clean plastic bags and placed in the shaded environment.

Analytical methods
Parameters, including temperature, pH, Electrical Conductivity (EC), and redox potential, were measured with a HI 9828 portable multi-meter (HANNA, Woonsocket, RI, USA) during water sampling. To minimize the influence of atmospheric contact, the monitoring probe was immersed in an in-line flow cell where water flows through constantly. Physiochemical data were recorded when stabilized. Alkalinity was titrated immediately in the field by using a Model 16900 digital titrator (HACH, Loveland, USA) during sampling. Redox sensitive components such as total Fe, Fe(II), NH4-N, NO2-N and S 2were quantified using a portable UV/VIS spectrophotometer (HACH, DR2800).
Analysis of REE+Y concentrations were performed by employing two single-collector ICP-SFMS instruments (ELEMENT XR and ELEMENT 2, Thermo Fisher Scientific, Bremen, Germany) based on the protocol documented in Rodushkin et al. (2018). Cation exchange resin AG 50W-X8 (200 to 400 dry mesh size, Bio-Rad laboratory AB, Solna, Sweden) was used for preconcentration of REE+Y. Prior to use, the resins were immersed in 14 mol/L HNO3 for a night and were cleaned with deionized water. A 2 mL low-density polyethylene (LDPE) column was loaded with the prepared resins, and 4 mL 14 mol/L HNO3, 8 mL deionized water and 4 mL 9.5 mol/L HCl were loaded sequentially for conditioning. Finally, 200 mL sample was introduced into the pre-conditioned column and the retained REE+Y were eluted with 4 mL 0.5 mol/L HCl. By this method, a 50-fold preconcentration was achieved. Lu, as suggested by Wilkin et al. (2020). The detection limits were 0.1 ng/L for Lu, 0.2 ng/L for Eu, and 0.5 ng/L for other REE and Y. In addition, 1 ng/L In solution was used as internal standard to check the stability of the analytical system. Interferences of BaO + on the 151 Eu and 153 Eu were corrected automatically by setting up a correction equation during analysis. Briefly, we run ~100 μg/L Ba and determine the peak area of the Eu isotope being monitored. During a run of samples, Ba (without calibration) was monitored and the results were applied for a correction. The analytical precisions of REE+Y were generally better than 8%.
Bulk composition analysis of sediment and sludge samples were carried out after a digestion using LiBO2-LiB4O7 fusion/dilute nitric. The obtained solution was used for element determination by ICP-AES (iCAP6300, Thermo Fisher Scientific, Waltham, MA, USA) and ICP-MS (7500C, Agilent Technologies, Santa Clara, CA, USA). The oxides (SiO2, Al2O3, Fe2O3, K2O, MgO, MnO, Na2O, P2O5 and TiO2) were analyzed by the X-ray fluorescence (XRF) (ARL Advant X) technique with a glass flux sheet method. Mineral compositions in surface sediments and sludges were determined by X-ray diffraction (XRD) (D8 advance, Bruker) equipped with a XGEN-4000 generator.
Diffraction patterns for each sample were identified under operation condition at 40 kV and 40 mA, 2θ of 20° to 70°, and a step size of 0.02°. Minerals having content greater than 5 wt% of the bulk solid were identified.
The photomicrographic characteristics of the solid samples and residuals (by filtering water in-situ) of aqueous solution were examined by using scanning electron microscopy (SEM) and the elemental compositions were determined with energy-dispersive X-ray spectrometry (EDS, Jeol, Jsm-6510la).

Modelling approach
Speciation calculations were performed with hydrogeochemical code PHREEQC version 3.4 (Parkhurst and Appelo, 2013) using the Nagra/PSI database (Hummel et al., 2002). Stability constants of REE complexation to major anions (e.g. CO3 2-, SO4 2-, OH -, Cl -, F -, and NO3 -) were incorporated into the database. For examples, constants for REE(CO3)2and REECO3 + were taken from Luo and Byrne (2004), and those for REESO4 + were from Schijf and Byrne (2004). Surface complexations of REE onto iron oxyhydroxides were considered as previously done by , in which the relevant formation constants and detailed modelling procedures are provided.

Water chemistry
The major ions and physicochemical parameters are shown in Table 1. Results show that pristine mine waters (MW1 to 5) have pH values between 3.8 and 4.2, showing a characteristic of acidic water. A slight increase in pH value (between 4.1 and 4.9) has been observed for NDT samples (NDT1 to 6). The CPT samples have pH values higher than 7 (between 7.4 and 8.7), and the WW sample's pH is 6.5. Clconcentrations are below 6 meq%. Water types thus are Ca/Mg-HCO3/SO4 and Ca/Mg-SO4. Over the CPT pathway (CPT1 to 4), the major cation shifts towards Na + (meq from 36% to 50%), followed by Mg 2+ (meq ranging from 23% to 34%) and Ca 2+ (meq from 22% to 25%), while the dominant anion is HCO3 -(meq >62%), hydrochemical type for these samples being Na-Ca/Mg-HCO3. This type of water stretches to the downstream near the village (CPT4), approximately 1 km downstream from the STP. The NDT samples have coherent water type of Mg-Na-Ca-HCO3-SO4. The WW collected from the village shows a Na-HCO3 water type. All water compositions are shown with a piper plot for a comparison (Fig. 2).  The S 2concentrations mostly are below detection limit. Among MW samples, only MW1 and MW4 have S 2concentrations of 1 mg/L and 5.00 mg/L, respectively.
All NDT samples are below detection limit for S 2concentrations. The CPT samples have S 2concentrations below detection limit with an exception of CPT3 and CPT4 (3.00 mg/L). The WW sample has S 2concentrations of 1 mg/L (Table 1).   The calculated lanthanide species are presented as a function of their atomic numbers (Fig. 4)    and between 0.02% and 2.27% (average 1.03%) in MW and CPT samples, respectively (Fig. 6).
SEM images of sediment and sludge samples are shown in Fig. 7 and Fig. S3.
The sediment samples contain clastic textures on the surfaces with length < 1 μm.
EDS spectra indicate that the stripe-like shapes are probably Al-and/or Fe-bearing crystals. The percentage of Al and Fe is determined to be 7% to 10% and ~2%, respectively. REE+Y are all below 1% with exceptions of Dy and Er (Fig. 7). On the surfaces of sludge samples, needlelike and burr-like shapes occur with shorter length than those observed in sediment samples. EDS results show that Al and Fe account for from 7% to 10% (average 10%) and from 1% to 8% (average 5%), respectively, indicating that these are the amorphous Al-and Fe-containing minerals.    5), and similar HREE enrichment characteristics are found in MW and NDT samples as well (Fig. 4a and c). Therefore, the acid waters have similar UCC-normalized patterns to those of sediments rather than rocks, which means that the HREE enrichments in MW probably result from the sediment signatures in this area, but unlikely from the rock signatures. The higher degrees of HREE enrichment in sediments relative to rocks suggest that REE+Y have fractionated during weathering of parent rocks. Therefore, after REE+Y as well as other metal elements (i.e. Fe, Mn and Al) are initially mobilized from the rocks, the majority of them are sorbed onto the surfaces of sediments/soils which developed from rocks. During the process, the source-rock composition regulates HREE enrichment in sediments. When acid leaching occurs, the sorbed REE+Y are readily eluted from the surfaces of sediment.
This secondary mobilization is unexpected to significantly fractionate REE+Y in acid waters for two possible reasons: (i) sulfate complex and free ion are the main solution species (Wood, 1990), as indicated by speciation calculations (Figs. 4 b and d); REE+Y association with sulfate is not believed to highly modify REE+Y patterns in aqueous solutions owing to the relatively unchangeable of REE+Y-sulfate formation constants (Schijf and Byrne, 2004;Wilkin et al., 2020); (ii) REE+Y re-adsorption onto mineral phases is quite limited. This is due to the general low pH values (< 4) in the MW samples, where REE+Y behave conservatively even though Fe oxyhydroxides are present (Verplanck et al., 2004). Pourret and Davranche (2013) (Fig. 3a), indicating that Fe-and Al-bearing minerals are possible secondary phases immobilizing REE+Y from CPT samples. Results of XRD support that ferrihydrite and schwertmannite are the most plausible minerals which act as the sinks of aqueous REE+Y. These minerals have been determined in sludge samples (Fig. S3), and have been suggested to be the most common minerals precipitated from AMD during neutralization processes (Lozano et al., 2020). The dominant Fe species shifts from Fe(II) in MW and NDT samples to Fe(III) in CPT (Fig.   3a ), suggesting that the formation of ferrihydrite and schwertmannite may occur following Eqs. (1) and (2).
The impact of sorption and co-precipitation on ΣREE+Y concentrations is reflected by positive Y/Ho anomalies. Yttrium has the same charges and similar ionic radii to Ho. However, Y is thought to behave differently from Ho during sorption and co-precipitation with Ho having greater affinities than Y for surface complexes (Bau, 1999), and thus leads to fractionations between the two elements (Nozaki et al., 1997;Möller et al., 1998). The prevailing positive Y/Ho anomaly observed in water samples implies that sorption of REE+Y readily occurs during neutralization processes. Higher Y/Ho values are found in CPT samples (1.70 to 1.81) as compared to MW (1.14 to 1.53) and NDT (1.21 to 1.36) samples. This is attributed to intense sorption or/and co-precipitation of REE+Y in CPT samples, which is governed by solution pH values.
Indeed, ΣREE+Y concentrations are negatively correlated with Y/Ho anomalies (R 2 =0.66), showing that low-pH conditions are not the most favorable for REE+Y sorption. Results of field studies indicated that Y/Ho anomaly could be a tracer for understanding REE+Y sorption in groundwater (Tweed et al., 2006), although the obtained samples were weakly alkaline. The great affinities for REE+Y to amorphous Fe oxyhydroxides could be achieved by means of adsorption and/or co-precipitation (De Carlo et al., 1998;. Aluminum oxyhydroxides have been shown to effectively remove REE+Y from acid thermal water when it's neutralized to pH values of <6 (Ogawa et al., 2019). More recently, sorption behaviors of REE+Y onto schwertmannite were studied by Lozano et al. (2020a, b), showing that lanthanides sorption increased from pH 5to 6.5. It should be noted that changes of REE+Y concentrations may also be coupled to scavenging of Mn oxyhydroxides, which has a sorption edge as low as pH of 4 and a complete sorption at pH around 8 (Pourret and Davranche, 2013), although this mineral is not detected by XRD in sludge samples due to its low content in this study. Therefore, dissolved REE+Y concentrations are mainly determined by acidification-enhanced dissolution/leaching and scavenging of secondary minerals precipitated from aqueous solutions.

REE+Y patterns
All water samples exhibit enrichments in HREE relative to LREE when normalized to UCC (Fig.4). This fractionation patterns are inconsistent with the general enrichment in MREE over LREE and HREE observed in AMD worldwide (Pérez-López et al. 2010;Grawunder et al., 2015;Ayora et al., 2016;León et al., 2021). However, the REE+Y patterns of acid waters are still of diversity. HREE (Medas et al., 2013) and LREE (Bozau et al., 2004)  The possible mechanisms are addressed as follow.
The flat UCC-normalized REE+Y pattern of parent rock indicates that it is not a direct cause for HREE enrichment in acid waters, whereas it can be sources of REE+Y contributing to acid waters via weathering and leaching. This has been discussed above. Generally, natural waters have distinct normalized REE+Y patterns when interacted with different rock types (Duvert et al., 2015). Cases have been reported that REE+Y patterns in circumneutral waters directly inherit from rocks that they interact with (Banner et al., 1989), and the patterns change with water flowing due to influences of (bio)geochemcial processes. However, the present study probably shows that REE+Y patterns are at a late stage of outflow discharges from the mine site. Therefore, it is impossible that there is a direct relationship between the REE+Y patterns in the acid waters and those in the parent rock.
The effect of organic matter on the formation of HREE enrichment waters can be ruled out due to the low DOC concentrations, which are generally below 3.2 mg/L with exceptions of sample CPT3 and CPT4. Preferential complexation of MREE relative to LREE and HREE with humic acid could lead to MREE-enriched patterns in alkaline waters (Wood, 1993;Pourret et al., 2007), which is the case reported by Munemoto et al. (2020), but not for the present study. The SEM results of filter residuals show no substantial REE+Y in forms of particulate (>0.22 μm) (Fig. 6) (Ogawa et al., 2019). Moreover, coefficients of distribution of the REE+Y between solids consisting of Al and Fe oxyhydroxides and water show a very weak increase across lanthanide series at pH of 6.3 to 6.8 (Lozano et al., 2020a). It reflects that REE+Y fractionation owing to adsorption or/and co-precipitation by Fe and Al oxyhydroxide in the study is insignificant. showing no preferential complexation of individual lanthanide (Wood, 1990).
Constants of REE-NO3complexation are higher for LREE than HREE (Millero, 1992), suggesting a preferential complexation of LREE over HREE. This does not make for HREE enrichment in waters either. Complexation to CO3 2could contribute to formation of HREE enrichment in CPT and WW due to preferential stabilization of HREE relative to LREE. This mechanism has been demonstrated to be responsible for HREE enrichment in natural alkaline waters . The preferential sorption of positively charged REECO3 + by negatively charged surfaces under neutral to weakly alkaline conditions may also lead to HREE enrichment (Guo et al., 2010).
Other ligands including OH -, PO4 3--, Fand Cltend to preferentially complex HREE, but their low speciation predicted by the model or/and low concentrations exclude the possibility of HREE enrichment caused by solution complexation.
In consequence, the clear resemblance between REE+Y patterns in sediments and those in acid waters indicates that the aqueous REE+Y patterns are probably controlled by the characteristics of REE+Y source areas, and the patterns remain relatively unchangeable with water flowing irrespective of precipitation of secondary minerals due to a rise of pH values. A model can be developed as to the formation of HREE enrichment pattern in acid waters. (i) thick zones of clay rich sediment/soil would develop above the granites during weathering process (Van Gosen et al., 2014), as suggested by the common presence of kaolinite that determined by XRD (Fig. S3).
(ii) The mobilized REE+Y are initially adsorbed onto sediments, and a preferential accumulation of HREE by the metal oxides and/or clay minerals containing in sediments leads to HREE enrichment in sediments (Aström, 2001). This is reflected by the presence of ferrihydrite, kaolinite, schwertmannite in sediment samples (Fig.   S3) and higher contents of HREE such as Gd, Er and Dy being detected by SEM in sediments (Fig. 7). (iii) The acid mining process leaches previously adsorbed REE+Y into aqueous solutions, giving rise to enrichments of HREE relative to LREEs in acidic waters (Leybourne et al., 2000;Migaszewski et al., 2016). No substantial variability of REE+Y patterns over plant treatments are for a large part due to REE+Y source controls. The origins of REE+Y patterns in acid water remain debatable. Studies suggest that they originate from REE+Y-rich soil waters leached from acid soils (Åström, 2001), while others attributed them to sulfides (Pérez-López et al., 2010) or dissolution of parent rock minerals (Cánovas et al., 2020). Ferreira da Silva et al. (2009) results show that REE of mining wastes in Lousal mine area result from the mixing of sulfides and host rocks, which is also part of arguments presented by Pérez-López et al. (2010). The dispersion of REE+Y patterns observed in AMD collected from 30 mining districts of the IPB was suggested to be controlled by local lithology and mineralogy (Léon et al., 2021). Similarly, the HREE enrichment patterns of acid waters shown in this study is a consequence of leaching of HREE-rich sediments developing from host rock. Be and Mo are located in the negative field, indicating that these elements mainly come from leaching of sediments and are strongly associated with pH values, as the parameter is found to be in the same part of these elements. The only exception is Sc which remains in the positive field. Different behaviors of Sc with respect to the REE+Y at low pH may explain this phenomenon (Lozano et al., 2020 to co-precipitation of these secondary minerals may occur in CPT samples (Ohta and Kawabe, 2001;Takahashi et al., 2007). However, the process could not be a dominate control on negative Ce anomalies as well, because the Ce/Ce* values in CPT samples are comparable to those of MW and NDT samples (Fig. 9a), despite that Fe and Mn concentrations decrease substantially with acid waters flowing downwards (Fig. 3a). Furthermore, the SEM results show that no substantially higher contents of Ce than other REE+Y are observed in sludge samples (Fig. 7). Therefore, the negative Ce anomalies are suggested to be largely controlled by REE+Y sources in sediments. It's worthwhile to note that the role of source composition in development of Ce anomalies would be complexed by redox change, adsorption on particle surfaces, organic complexation, microbial-mediated activities, and preferential incorporation into secondary minerals (Dia et al., 2000;Protano and Riccobono, 2002;Pourret et al., 2007;Tanaka et al., 2010). Both negative and positive Ce anomalies in acid waters have been reported previously (Gammons et al., 2003;Olías et al., 2005) ( Fig. 9b). Smedley et al. (1991) argued that Ce anomalies in slightly acidic groundwaters (5-6.8) were related to rock aquifer-source, and their water samples exhibited similar REE+Y patterns to those of rocks composed of granites and metasediments. Therefore, the occurrences of Ce anomalies found in acid waters are most likely source-sediment inherited characteristics rather than arising from the process-related oxidative precipitation of Ce in waters.

Implication for REE+Y recovery
Although AMD has been flagged as an environmental concern globally due to its great hazard to ecosystems, it's a critically potential secondary source of REE+Y as between 1288 and 3764 g/day, contributing between 34% and 38%, and between 38% and 42% of the total ∑REE+Y, respectively (Fig. 11). The recoverable Y ranges between 793 and 2348 g/day (20% to 28%). The highest total recoverable value occurs at the location of MW5, which is nearest the inlet of WTP (Fig. 1).

Figure 11
The proportion of potentially recoverable REE+Y in MW and NDT samples It must be noted that the profitability of REE+Y recovery from AMD depends on their amounts born in AMD, the cost of separation and extraction, and the proportion of REE+Y in AMD (Léon et al., 2021). Hence, the high proportions of HREE found in this area makes recovery of REE+Y from the acid waters significantly attractive, considering HREE are more precious compared to LREE and are largely restricted to South China. The large differences in lanthanide price (SMM, 2020) derive studies focused on selective recovery of REE+Y from ADM and their treatment sludge. A preferential retention of REE+Y from AMD treatment process were reported in coal mines (Stewart et al. 2017;Hedin et al., 2019). The availability of recovering REE+Y as well as other critical metals like Cu, Zn and Ni from acid mine leachate and precipitates were documented by Zhang and Honaker (2018). Ayora et al. (2016) studied the recovery of REE+Y from passive remediation systems of AMD system with two substrate-(schwertmannite and basaluminite) based treatments, showing that AMD remediation process serves as a suitable REE+Y source. Biochar may server as a potential sorbent to immobilize these metals in aqueous solution where pH value is low to 3 (Pourret and Houben, 2018). In all, the practical examples of REE+Y recovery from AMD render an environmental problem worthwhile as a source of raw materials, considering finding alternative sources of REE+Y is in pressing need and an increasing demand for these critical metals, particular for the countries without primary deposits.

Conclusions
This study investigates the geochemistry of rare earth element (REE) and yttrium Carbonate complexation accounts for HREE enrichment as well in CPT and WW samples. The loads of LREE and HREE in AMD are calculated to range between 1116 g/day and 3373 g/day, and between 1288 g/day and 3764 g/day, respectively, which shows great potential for REE+Y recovery from AMD.