A Molecular-Scale Approach to Rare-Earth Beneficiation: Thinking Small to Avoid Large Losses.

Separating rare-earth-element-rich minerals from unwanted gangue in mined ores relies on selective binding of collector molecules at the interface to facilitate froth flotation. Salicylhydroxamic acid (SHA) exhibits enhanced selectivity for bastnäsite over calcite in microflotation experiments. Through a multifaceted approach, leveraging density functional theory calculations, and advanced spectroscopic methods, we provide molecular-level mechanistic insight to this selectivity. The hydroxamic acid moiety introduces strong interactions at metal-atom surface sites and hinders subsurface-cation stabilization at vacancy-defect sites, in calcite especially. Resulting from hydrogen-bond-induced interactions, SHA lies flat on the bastnäsite surface and shows a tendency for multilayer formation at high coverages. In this conformation, SHA complexation with bastnäsite metal ions is stabilized, leading to advanced flotation performance. In contrast, SHA lies perpendicular to the calcite surface due to a difference in cationic spacing. We anticipate that these insights will motivate rational design and selection of future collector molecules for enhanced ore beneficiation.

Introduction

Rare-earth elements (REEs) have widespread applications, ranging from catalysts in automobiles and petroleum refining to polishing media, magnets in computer equipment and wind turbines, batteries, phosphors in lighting, and medical diagnostic equipment (Eggert et al., 2016; Ganguli and Cook, 2018; Jordens et al., 2013; Tunsu et al., 2015). These “rare” elements are so named for the paucity of mineral deposits containing them in mineable quantities, despite their fair abundance in the Earth's crust (Eggert et al., 2016). The existence of so few economically viable reserves, coupled with a year-over-year increase in demand (Eggert et al., 2016), highlights the need for more efficient strategies for extracting REEs from mined ores.

REEs are subclassified into two groups: the light rare-earth elements (LREEs), La to Eu, and the heavy rare-earth elements (HREEs), Gd to Lu and Y (Krishnamurthy and Gupta, 2016). The largest mineral sources of LREEs are bastnäsite deposits in Bayan Obo, China, and Mountain Pass, California, USA (Castor, 2008; Drew et al., 1990; Eggert et al., 2016; Kynicky et al., 2012; Mariano and Mariano, 2012; Yang et al., 2011). Bastnäsite is a fluoro-carbonate mineral with the chemical formula MFCO3 (M = Ce (~50%), La (~30%), and other LREEs) (Eggert et al., 2016; Long et al., 2010; Weng et al., 2015). Obtaining REEs from these mineral sources first requires beneficiation, or the separation of bastnäsite from unwanted gangue materials in the ore. Bastnäsite beneficiation is currently achieved by means of froth flotation (Cui and Anderson, 2017a; Liu et al., 2019). In this process, a slurry of ore is prepared, containing a collector molecule that selectively adsorbs to the bastnäsite particles, increasing their hydrophobicity. Air is then bubbled through the slurry, and the hydrophobic bastnäsite adheres to the bubbles, which float to the top of the flotation cell and are skimmed off for further processing. The grade of beneficiated ore thus relies on selective adsorption of the collector onto the surfaces of bastnäsite. Separation of bastnäsite from calcite, the most abundant gangue mineral (Pradip and Fuerstenau, 2013), is particularly difficult, because both of these salt-type minerals (Gallios and Matis, 1992) include carbonate anions as integral components of their functional units.

Beginning with initial reports in the 1960s of their flotability (Fuerstenau; Peterson, 1969), hydroxamic acid compounds have shown selectivity for REE minerals (Assis et al., 1996; Cui and Anderson, 2017b; Liu et al., 2019; Nagaraj, 1992; Pradip and Fuerstenau, 2013; Wanhala et al., 2019; Yoon and Hilderbrand, 1986; Zhang et al., 2014). Nonetheless, beneficiation with hydroxamic acid results in the recovery of only 60%–72% REE from the primary ore, with the rest lost in the gangue (Nagaraj and Farinato, 2016; Talens Peiró; Villalba Méndez, 2013). This loss highlights the need for advanced collector ligands with enhanced selectivity for bastnäsite over gangue materials.

Since the inception of collection with hydroxamic acid ligands, extensive progress has been made in describing the interactions between the hydroxamic acid group (Scheme 1, with pKa values from Humbert et al., 1998, Khalil et al., 2007) and metal ions in terms of chemisorption on REE mineral surfaces and weaker adsorption on gangue (Anderson, 2015; Cao et al., 2019; Marion et al., 2020; Nagaraj, 2018; Pradip and Fuerstenau, 1983, 1985; Xiong et al., 2020). Typically, chemisorption results in chelation, forming a five-membered ring through interactions between both hydroxamic oxygen atoms and a surface cation (Nagaraj, 2018; Ren et al., 1997). Electronic effects of the hydroxamic acid moiety in selective adsorption have also been discussed thoroughly (Nagaraj, 2018).

Scheme 1

Skeletal Structures and pKa Values for the Three Collector Ligands Investigated in This Work

SHA and BHA both exhibit the C(=O)-NR-OH hydroxamic acid moiety.

New hydroxamic acid collector ligands are traditionally selected through a trial-and-error approach, guided by empirical trends and observations of flotation in other materials. Studies have begun to extend a molecular-level approach beyond the hydroxamic acid moiety, toward a comprehensive understanding of adsorption in terms of specific structural interactions among functional groups in the collector, water molecules, and atoms in the mineral surface. This detail-oriented perspective, which has been employed for other metal- and mineral-ligand systems (Frey et al., 2000; Rio-Echevarria et al., 2008; Waterson et al., 2016), can lead to the rational design of future collector ligands.

Such an approach entails the molecular recognition of surfaces, wherein structural features of a ligand molecule are aligned with those of the mineral surface, leading favorably to adsorption. Motivated by biomineralization phenomena resulting in the natural production of materials such as oyster shells, ivory, and pearls in living creatures as a result of structural similarities in organic polymers and inorganic minerals, Pradip and co-workers have pioneered a description of collector selectivity and flotation efficacy in terms of differences in ligand recognition across mineral surfaces (Elam, 2014; Pradip, 1992; Pradip and Rai, 2003; Pradip et al., 2002). Previous work has enlisted coverage-dependent isotherms to describe the orientation of ligands bound to a surface (Cases and Villieras, 1992). Investigations of the molecular recognition of surfaces extend these approaches by enlisting computational and spectroscopic methods, allowing us to “see” the adsorption state of molecules on the surface.

Joint spectroscopic-computational studies have recently begun to reveal adsorption structures and energies on bastnäsite surfaces (Espiritu et al., 2018). Wanhala et al., for instance, explored the mechanisms of adsorption of octyl hydroxamic acid to bastnäsite as a function of hydroxamic acid coverage (Wanhala et al., 2019). Implementing both attenuated total reflection-Fourier transform infrared (ATR-FTIR) and vibrational sum frequency generation (vSFG) spectroscopies, along with density functional theory (DFT), this work revealed coverage-dependent ligand binding and highlighted the role of the long alkyl chain in adsorption.

Aromatic hydroxamic ligands have shown recent promise for bastnäsite flotation (Everly, 2017). Studies employing these collectors have optimized the conditions for flotation and describe the adsorption of some of these ligands onto the surface (Cao et al., 2018; Jordens et al., 2014, 2016; Li et al., 2018; Marion et al., 2020; Xia et al., 2014; Yang et al., 2017; Yu et al., 2020). Here, we provide mechanistic insight into the adsorption of one aromatic ligand, salicylhydroxamic acid (N,2-dihydroxybenzamide, SHA), through a series of investigations designed to relate structure to adsorption selectivity. SHA has shown good performance for LREE recovery (Xia et al., 2014), and previous work employing flotation and contact-angle studies reveals increased bastnäsite hydrophobicity following treatment with SHA, contributing to its flotability (Cao et al., 2018; Xiong et al., 2020). Recent work has compared the optimal conditions for SHA flotation of bastnäsite with those of gangue minerals, including calcite, using spectroscopic, microflotation, and zeta-potential measurements (Xiong et al., 2020). Importantly, this study revealed stronger chemisorptive binding of SHA to bastnäsite and weaker physisorption to gangue minerals.

In the present work, a combined approach that enlists DFT, along with Raman, ATR-FTIR, and vSFG spectroscopies, affords a comprehensive molecular-level description of adsorption. Results of parallel studies on the adsorption of SHA, benzohydroxamic acid (N-hydroxybenzamide, BHA), and salicylic acid (2-hydroxybenzoic acid, SA) ligands onto bastnäsite demonstrate the roles of functional groups important to beneficiation by flotation with SHA. In this manner, we are able to conceptually deconstruct the ligand into a sum of functionalities in order to relate the structure of the ligand to surface recognition. The structure of SHA, shown in Scheme 1 along with the other two ligands, consists of a benzene ring with ortho-hydroxamic acid and phenolic moieties. BHA has a similar structure, lacking only the phenol group. By comparing DFT and experimental results for SHA and BHA, new insight into the role of the ortho-phenol group in the adsorption of SHA is elucidated. Similarly, SA differs from SHA through the substitution of the hydroxamic acid group for a carboxylic acid group. A comparison of results using these substituted ligands provides insight into the role of the hydroxamic acid moiety in the recognition of bastnäsite and calcite surfaces, as well as on the beneficiative properties of SHA.

With limited mineable rare-earth resources, strategies must be developed that will enhance the availability of these critical materials. We anticipate that detailed insights such as those gained from our systematic, molecular-scale approach will influence the design of compounds that can more effectively separate rare-earth-containing mineral compounds from gangue.

Results

Microflotation Experiments

The results of microflotation experiments for the three ligands on bastnäsite are given in Table 1. The beneficiation potential of the three ligands decreases in the order SHA > BHA > SA, as reflected by both sample grade and total rare-earth oxide (REO) recovery. From an ore sample containing 7.7% rare-earth oxide by weight, flotation with SHA results in the greatest total recovery of REO (66.6%), with an increase in grade to 33.4% REO. This corresponds to 5.5 times as much REO recovered as calcium oxides (CaO), with a concentration of REO 2.5 times that of CaO in the recovered froth, indicating selectivity of SHA for bastnäsite over calcite. This result is in good agreement with recent microflotation studies of bastnäsite ores with SHA (Xiong et al., 2020). BHA also shows some capability as a collector, increasing the grade to 22.5% REO by weight, and recovering 57.0% of the REO from the original ore. BHA selectivity for bastnäsite is also noted, recovering 4.7 times as much REO as CaO from the ore. SA, which shows almost no selectivity for bastnäsite, yields microflotation froth with a composition resembling that of the starting ore.

Table 1

Results of Microflotation Experiments, Reported in Terms of Rare-Earth Oxide (REO) and Calcium Oxide (CaO) Content

	Grade		Recovery
	REO (%)	REO/CaO	REO (%)	REO/CaO
Ore	7.7	0.5	–	–
Salicylhydroxamic acid (SHA)	33.4	2.5	66.6	5.5
Benzohydroxamic acid (BHA)	22.5	1.9	57.0	4.7
Salicylic acid (SA)	7.3	0.8	50.5	1.8

Density Functional Theory Calculations

Periodic-DFT-calculated adsorption energies for SHA, BHA, and SA on bastnäsite-[100], the most prevalent bastnäsite surface (Srinivasan et al., 2017; Srinivasan et al., 2016), with acidic protons proximal to the adsorbate, are provided in Table 2, and optimized structures are shown in Figure 1 and detailed in Figure S1 with interactions contributing to adsorption specified.

Table 2

Collector Adsorption Energies (kJ/mol) on Bastnäsite-[100] and Calcite-[104] Surfaces with Dipole and Solvent Corrections Included

Mineral	Salicylhydroxamic Acid (SHA)	Benzohydroxamic Acid (BHA)	Salicylic Acid (SA)
Bastnäsite-[101¯0]	−48.9	−50.1	−32.5
Calcite-[101¯4]	−59.2	−19.2	−43.2

Figure 1

DFT Structures of the Most Stable Adsorption Configurations of Each Ligand on Bastnäsite-[100]

Charge neutrality is maintained by having acidic proton(s) either on a ligand or surface-CO32-/F−. For clarity, side views show only the top three layers of bastnäsite slabs. Color scheme: Ce, light green; F, purple; N, orange; C, brown/cyan; O, red; H, white. See also Figure S1 for a detailed depiction of interactions contributing to adsorption.

The optimized structures for SHA and BHA, which exhibit different binding modes on the bastnäsite surface, result in similar adsorption energies. In its most stable adsorption orientation, BHA binds in a bridging fashion to two surface Ce3+ cations via both hydroxamic acid oxygen atoms, as well as through a hydrogen bond donated by a protonated carbonate moiety to the carbonyl oxygen, resulting in an adsorption energy of −50.1 kJ/mol. Conversely, SHA adsorbs with an energy of −48.9 kJ/mol through a single monodentate interaction with surface Ce3+, along with two hydrogen bonds with surface anionic groups. Previous work highlights the role of hydrogen bonding in SHA adsorption to mineral surfaces and describes a bidentate chelation binding mechanism between the hydroxamic acid moiety and the surface (Cao et al., 2019). In contrast, the lowest-energy structure for SHA on bastnäsite that we present in Figure 1 is 6.5 kJ/mol more stable than the complex in which SHA exhibits bidentate chelation, revealing in a surface-perpendicular orientation. (i.e., structure D in Figure S6).

Both BHA and SHA adsorb onto the surface with their aromatic ring moieties oriented essentially parallel to the surface plane. In SHA, the hydrogen bond afforded by the additional phenolic moiety anchors the ring to the surface, resulting in a more surface-parallel ring in SHA than in BHA (Figure S6).

SA, like BHA, shows adsorption to bastnäsite to two surface Ce3+ cations and a single hydrogen bonding interaction. Without the constraint of a bound hydroxamic group, the aromatic ring is now oriented more vertically with respect to the surface, implicating the binding configuration of the hydroxamic acid group in the orientation of the ring in SHA along the surface. Although the binding modes are similar, BHA adsorbs more strongly to the surface (Eads = −50.1 kJ/mol) than does SA (Eads = −32.5 kJ/mol), supporting the role of the hydroxamic acid moiety in strong SHA adsorption.

For insight into the selectivity of the three ligands for bastnäsite over calcite, we also explored the adsorption of SHA, BHA, and SA on the most stable calcite surface: [104]. Adsorption energies of the lowest-energy configurations for each ligand-calcite system, with acidic protons near the site of adsorption, are provided in Table 2, and corresponding structures are shown in Figure 2 and detailed in Figure S2 with interactions contributing to adsorption specified.

Figure 2

DFT Structures of the Most Stable Adsorption Configurations of Each Ligand on Calcite-[104]

Charge neutrality is maintained by having acidic proton(s) either on a ligand or surface-CO32-. For clarity, side views show only the top three layers of calcite slabs. Color scheme: Ca, green; C, brown/cyan; O, red; N, orange; H, white. See also Figure S2 for a detailed depiction of interactions contributing to adsorption.

BHA shows much weaker adsorption onto calcite (Eads = −19.2 kJ/mol) than onto bastnäsite (Eads = −50.1 kJ/mol), despite a similar adsorption mechanism including bridging interactions and a hydrogen bond. Closer cationic spacing on the calcite surface (d(Ca-Ca) = 4.03 and 4.99 Å) compared with the Ce-bastnäsite surface (d(Ce-Ce) = 4.78 and 7.07 Å) may allow for closer adsorption on the bastnäsite surface and potentially cause differing degrees of strain on the structure of the ligand upon adsorption. Previous work on bastnäsite-collector design highlights variations in cationic spacing in these surfaces (Srinivasan et al., 2017).

SHA shows stronger adsorption (Eads = −59.2 kJ/mol) on calcite than on bastnäsite (Eads = −48.9 kJ/mol), resulting from a different adsorption mechanism. SHA binding to bastnäsite occurs through a single interaction with a surface cation and two hydrogen bonds, whereas binding to calcite occurs through cationic interactions with all three oxygen atoms in SHA to different Ca2+ sites, and two hydrogen bonds with surface carbonate groups. In this case, closer cationic spacing in calcite leads to stronger adsorption through increased ligand-surface interaction. In order to facilitate an interaction between the phenolic oxygen and a nearby calcium ion, the aromatic ring moiety is now oriented more perpendicular to the surface. Moreover, the adsorption structures of the hydroxamic acid moieties in BHA and SHA on calcite are essentially identical to one another, directly revealing the role of the phenolic moiety in SHA adsorption on calcite. In the absence of the phenol, the ring in BHA lies more flat on the surface while ligand adsorption becomes less energetically favorable by 40.0 kJ/mol.

SA on calcite also shows similar adsorption features to SHA, with a bridging structure involving two oxygens and surface Ca2+ cations. However, a third interaction with a surface cation is not possible with SA, as the carboxylic acid group, in the place of the larger hydroxamic acid group in SHA, is not able to reach to an additional surface cation (Figure S2). Thus, SA adsorbs more weakly onto calcite (Eads = −43.2 kJ/mol) than SHA (Eads = −59.2 kJ/mol). Nonetheless, SA adsorption on calcite is stronger than on bastnäsite (Eads = −32.5 kJ/mol).

For DFT results discussed up to this point, we have considered only structures in which the acidic ligand protons are found either associated with or placed nearby on the surfaces of bastnäsite or calcite, in order to maintain charge neutrality in the system. We now turn to an alternative strategy, wherein SHA2− and SA2− replace, in their doubly deprotonated forms, a surface CO32− anion (see Transparent Methods in Supplemental Information). This approach affords the advantage of considering adsorption at a single-site vacancy defect, providing initial insight into a nonideal-surface system such as those expected from mined ores.

Replacement of CO32− by a ligand on the bastnäsite surface results in an energetic expense much less than that on the calcite surface (structures in Figures S3–S5, energies in Table S1). Despite showing more extensive SHA-surface interaction than on bastnäsite, the greater replacement energy on calcite is likely due to the inherent differences in structure between the [100] surface of bastnäsite and the [104] surface of calcite (Srinivasan et al., 2017). In the former, Ce3+ cations in the subsurface layer lie 1.9 Å below those in the surface layer, whereas in the latter, subsurface Ca2+ cations lie 3.0 Å below the surface. The removed carbonate anion, which served to neutralize subsurface cations as it extended below either surface, is replaced with a ligand that does not penetrate below the surface into the vacancy (See Figures S12 and S14 and Tables S8 and S10). As a result, the subsurface cation at the vacancy site remains under-coordinated to some degree. Calcite, with subsurface cations farther from the surface-bound adsorbate, is even more destabilized by replacement of CO32− with SHA2− than is bastnäsite.

SA, with a smaller carboxylate functional group (d(O-O) = 2.24 Å) than the hydroxamate group in SHA (d(O-O) = 2.74 Å), is more apt to fit into a carbonate vacancy site, allowing the carboxylate group to reach into the defect to interact with subsurface cations in both bastnäsite and calcite (see Figures S13 and S15 and Tables S9 and S11). Nevertheless, SA and SHA show the same number of interactions with cations at carbonate vacancy sites on the same surface (see Figures S4 and S5), and both molecules show similar replacement energies on bastnäsite prior to correcting for experimental pH (i.e., applying Equation S2 without including Ecorr values yields 39.3 kJ/mol for SHA and 36.6 kJ/mol for SA). In contrast, the replacement energy of SA on calcite is 23.0 kJ/mol less than that of SHA, before considering pH corrections (i.e., 96.5 kJ/mol for SHA and 73.5 kJ/mol for SA). With a deeper subsurface layer, defective calcite is more effectively stabilized by the penetration of SA into the vacancy site than is defective bastnäsite. As a result, the energy difference between replacement of SA on bastnäsite and on calcite (ΔErep, SA, bastnäsite-calcite = −36.9 kJ/mol) is smaller than for SHA (ΔErep, SHA, bastnäsite-calcite = −57.2 kJ/mol), which does not penetrate into either surface. This translates directly to greater selectivity for bastnäsite over calcite by SHA than by SA, in agreement with the microflotation results above. Furthermore, after accounting for pH corrections, according to Equation S3, we find more facile replacement of CO32− by SHA than by SA on both surfaces (see Table S1).

Vibrational Sum Frequency Generation Spectroscopy Experiments

The vSFG spectra probing the C-H stretching region for the three ligands studied at the bastnäsite-[100]-water interface are shown in Figure 3A at surface saturation. In all cases, the signals from the ligand-adsorption species were much weaker compared with measurements on ligands with alkyl tails (Sutton et al., 2020; Wanhala et al., 2019). The weak signal qualitatively suggests that the ligands are arranged such that the transition dipoles are oriented more parallel to the interfacial plane. This is supported by DFT calculations that show that SHA and BHA prefer to orient with the aromatic rings more parallel to the interface, where vSFG signals would vanish due to in-plane rotational isotropy. Bands near 2,850–2,950 cm−1 are associated with vibrations in the respective ligands and can be attributed to a redshifted O-H stretch of the carboxylic acid or phenolic groups or aromatic C-H stretches (Doughty et al., 2017; Gotch et al., 1991; Kandziolka et al., 2014). DFT results presented above support the possibility that these peaks are O-H stretches, whereas Raman measurements (vide infra) suggest there might be some weakening of the aromatic ring that would lower the C-H stretching frequency. Where present, features near 3,050 cm−1 are often attributed to the C-H stretches of aromatic rings; however, the absence of this mode in SA suggests that it could be due to a redshifted N-H stretch arising from hydrogen-bonding interactions, as further supported in the literature (Voss et al., 2018). Future measurements and calculations would aid in a definitive assignment.

Figure 3

vSFG Results for SA (Blue), SHA (Red), and BHA (Black) Ligands at Bastnäsite-[100]-Aqueous Interfaces

(A) Spectra at surface saturation in the 2,800 to 3,100-cm−1 region.

(B) Adsorption isotherms from 0 to 1 mM ligand concentration. Error bars show a standard deviation across three trials.

See also Figure S16.

Spectrally integrating the vSFG spectra and plotting the resulting intensity versus bulk ligand concentration generates the isotherms shown in Figure 3B. The binding equilibria captured by these isotherms can be described bysuch that is the adsorption equilibrium constant. The adsorption isotherms were fit up to a concentration of 0.3 mM using a standard Langmuir isotherm of the form (fits shown in Figure S16):where α is an amplitude related to the maximum coverage and C is the ligand concentration (Doughty et al., 2016; Wanhala et al., 2019). The extracted equilibrium adsorption constants at 296 K were for BHA and for SHA. The corresponding adsorption free energies are then for BHA and for SHA. In agreement with DFT results (Table 2), the adsorption strengths of the two ligands are very similar to one another, indicating that they should perform similarly in flotation, assuming they occupy the same surface area and operate in a monolayer regime. Binding of SA to bastnäsite was observed as a small change in signal and could not be fit to an isotherm model. Notably, at a concentration near 0.5 mM—above the isotherm saturation (i.e., ~0.3 mM)—we see an increase in signal for BHA, whereas a gradual decrease in signal is observed for SHA (Figure 3B). The increase in signal of BHA can be rationalized as a coverage-induced reorientation of the ligands on the surface. Specifically, at lower coverages, the molecules arrange such that the aromatic rings are lying more parallel to the interface, but as the surface becomes crowded they begin standing more upright to accommodate an overall larger number of ligands. This increase in the vSFG signal would then reflect an increase in the maximum number density of ligands and their orientational rearrangement out of the surface plane. Such a case of aromatic rings packing perpendicular to the surface at higher coverages may preclude BHA multilayer formation.

The lack of signal increase for SHA at higher concentrations suggests that SHA assumes a mostly surface-parallel orientation at low and high coverages, consistent with the DFT result showing anchoring by the phenolic group to interfacial binding sites, thereby limiting its ability to reorient at higher coverages. The slight decrease in signal suggests that, in contrast to BHA, SHA multilayers are being formed, decreasing vSFG signals due to increasing local symmetry. SA shows a weak signal throughout, with a subtle increase at higher concentrations, suggesting that, consistent with DFT results, adsorption is much weaker than for the other two ligands, even at elevated SA concentrations.

Attenuated Total Reflection-Fourier Transform Infrared Spectroscopy Experiments

Spectra of SA adsorbed to bastnäsite in situ at pH 8.5 are shown in Figure 4B. Peaks corresponding to adsorbed SA are very weak, suggesting poor adsorption of SA onto bastnäsite, in agreement with vSFG intensities and DFT energies. At 1 mM concentration (green traces), the symmetric and asymmetric COO− stretching bands are visible at 1,392 and 1,563 cm−1, respectively. Although the peak at 1,392 cm−1 overlaps the intense CO3 stretching vibration of the underlying bastnäsite solid (shaded region), it can still be clearly differentiated from the subtracted background. Decreased separation of these frequencies relative to those of the free ligand (Figure 4A, black trace) indicates binding of the carboxylate group via either a bidentate chelate or bridging mechanism (Korpany et al., 2017). Phenolic C-O stretching shows a broadened peak at 1,253 cm−1, similar to that in the spectrum of the aqueous SA-Ce3+ complex (red trace).

Figure 4

ATR-FTIR Spectra of Pure Ligands (Black), Ligands Complexed with Ce3+ (Red), and Spectra of Ligands Adsorbed onto Bastnäsite after 20, 40, and 60 min at Two Different Ligand Concentrations (Green and Blue)

(A) Pure SA and SA-Ce3+ complex. (B) SA on bastnäsite.

(C) Pure BHA and BHA-Ce3+ complex. (D) BHA on bastnäsite.

(E Pure SHA and SHA-Ce3+ complex. (F) SHA on bastnäsite. Curves are vertically offset for clarity. Peaks in shaded regions cannot be resolved owing to imperfect subtraction of the bastnäsite background.

At 2 mM concentration (Figure 4B, blue traces), more SA adsorbs, slightly improving the signal to noise ratio of adsorbed-ligand peaks. However, the peak positions are changed relative to the 1 mM spectra. The COO− asymmetric stretching peak at 1,563 cm−1 is smaller than the neighboring peak at 1,598 cm−1. The symmetric stretching frequency is decreased to 1,388 cm−1, with unresolved intensity on the low-wavenumber side of this peak. These bands are more similar to those of the uncomplexed ligand than to the low-concentration SA-bastnäsite spectra, indicating a weaker physisorption at higher coverages, perhaps in addition to the chemisorbed layer. The phenolic C-O stretch now shows a predominant peak that is redshifted relative to the aqueous Ce3+ complex, to 1,248 cm−1, primarily suggesting interaction of the phenolic group with the surface and/or among surface-bound layers of SA, likely through hydrogen bonding (Thomas and Kelley, 2008).

Adsorption of hydroxamate ligands onto mineral surfaces has previously been measured using ATR-FTIR (Borer et al., 2009; Sreenivas and Padmanabhan, 2002; Wanhala et al., 2019). In previous investigations, the adsorption mechanisms of alkyl hydroxamates were clearly reflected in the strong C=O stretching band (Cui et al., 2012). Figure 4D shows time-resolved spectra of BHA adsorbed to bastnäsite. Strong peaks at 1,607 and 1,571 cm−1 appear, consistent with the spectrum of the Ce-complexed BHA solution (Figure 4C, red trace). The sharp peak at 1,159 cm−1, also in agreement with the spectrum of the solution-phase complex, is reported in the literature for BHA as a combination band with contributions from C-N, N-O, and aromatic C-C stretching and C-H bending (Al-Saadi, 2012). Finally, a peak appears at 913 cm−1 in the spectra of bastnäsite-adsorbed species, assigned to N-O stretching and CNO deformation of the hydroxamate functional group. Although this peak exhibits poor signal to noise ratio, it appears much more clearly, and at higher frequency, than in the spectrum of aqueous BHA (black trace). Together, these results suggest that BHA adsorbs via bridging or a bidentate chelate, which is expected from Ce-hydroxamic acid complexes in solution (Sarvaramini et al., 2016).

Spectra of SHA adsorbed to bastnäsite are shown in Figure 4F. At 0.5 mM (green traces), clear peaks corresponding to the adsorbed ligand are apparent in the spectra. The sharp peak at 1,605 cm−1 resembles that in the spectrum of the aqueous-phase Ce-complexed ligand (Figure 4E, red trace). This well-resolved band involves stretching motions of both the hydroxamate C=O bond and the aromatic ring, suggesting chelated SHA, in agreement with DFT results shown in Figure S3 and FTIR spectra of SHA-bastnäsite systems reported previously (Xiong et al., 2020). A partially resolved band at 1,575 cm−1, assigned to C-N stretching/N-H bending (amide II) of the hydroxamate moiety (Brennan et al., 2016; Zhao et al., 2019), as well as aromatic-ring stretching, is consistent with the spectrum of the aqueous complex; however, this band shows increased separation from the C=O stretching feature when compared with the spectrum of the aqueous-phase Ce-complexed ligand. A redshift from higher frequencies may be due to hydrogen-bond donation by the N-H moiety (Arunan et al., 2011; Pauling, 1931), in line with DFT results (Figure 2) and vSFG spectral features.

At 2 mM concentration, the spectra of adsorbed SHA (Figure 4F, blue traces) are slightly different. For example, the peak at 1,507 cm−1, which appears upon complexation, contributes less to the overall intensity than at low concentration, suggesting that some of the SHA at the surface is uncomplexed, consistent with multilayer formation at elevated concentrations, in agreement with vSFG results. The hydroxamate mode around 1,580 cm−1 is less resolved from the carbonyl stretching, with similar separation between both frequencies as in the aqueous SHA-Ce complex. This perhaps suggests that the hydrogen bond donated by the hydroxamate N-H moiety is weaker or absent in physisorbed SHA.

Raman Spectroscopy Experiments

The Raman spectrum for SA at the bastnäsite-[100] surface is shown as the red trace in Figure 5A. The spectrum of bastnäsite prior to SA exposure (Figure S17, green trace) was subtracted from the SA-bastnäsite spectrum and scaled to match the intensity of the pure bastnäsite peak at 1,430 cm−1 (Frost et al., 2013).

Figure 5

Raman Spectra of Aqueous Ligand Solutions (Blue), Complex of Ligand + Ce3+ Prepared in Aqueous Solution, pH 8.5 (Black), and Bastnäsite-[100] with Ligand (Red)

Red bastnäsite + ligand spectra result from the subtraction of the spectrum of the bastnäsite crystal (See Figure S17) prior to ligand exposure.

(A) SA, 17 mM.

(B) BHA, 18 mM.

The spectrum of SA on bastnäsite shows strong features at 1,343 and 1,495 cm−1. The band at 1,495 cm−1 is assigned to changes in C-C stretching in the aromatic ring in SA (Green, 1977; Humbert et al., 1998). This band is shifted to higher frequency in comparison with that of the aqueous Ce-complex (Figure 5A, black trace), which suggests a greater distortion of the ring vibration of SA in the presence of bastnäsite. The distortion of the aromatic ring vibrations as a result of the SA interaction with metallic surfaces has been reported in the literature (Alvarez-Ros et al., 2000). The intensity of the ring vibration may be influenced by how molecules are packed on the surface. For instance, a confinement of the ring to the surface may suppress ring vibrations, whereas a vertical SA orientation, with the ring pointing away from the bastnäsite surface, would allow for the stretching of C-C bonds as observed in our case. A vertical orientation of SA on the surface is consistent with DFT structures of SA on bastnäsite (Figure 1 and S3). In the case of SA-bastnäsite, the strong band at 1,343 cm−1, which contains contributions from both the C–O (carboxylic) and C–OH (phenolic) groups, is shifted to a higher frequency relative to the band near 1,313 cm−1 in the aqueous SA-Ce3+ complex, suggesting the participation of these groups in the interaction with the surface, in agreement with DFT and ATR-FTIR results. In addition, we acknowledge that ring vibrations might also contribute to this spectral feature. Literature reports (Alvarez-Ros et al., 2000; Green, 1977; Humbert et al., 1998) suggest that SA can bind to the surface electrostatically, through the involvement of coordination bonds, and/or through the formation of hydrogen bonds. The Raman spectrum of SA on the surface is notably distinct from that of SA bound to Ce3+ in solution, suggesting that either the adsorption mechanism of SA to Ce3+ in bastnäsite differs greatly from that in solution or that SA is not specific for Ce3+ and can bind to surface or other ions in bastnäsite. The role of other ions in collector adsorption onto bastnäsite has been discussed previously (Cao et al., 2018).

The crystal-subtracted spectrum of BHA on the surface of bastnäsite is presented as the red trace in Figure 5B. Vibrational bands of pure BHA (blue trace) at 1,330 and 1,602 cm−1 are shown in the bastnäsite-adsorbed spectrum. Furthermore, a comparison of the subtracted-bastnäsite spectrum with that of the aqueous BHA-Ce3+ complex (black trace) reveals features that are common to both spectra at around 1,494, 1,525, and 1,576 cm−1. Together, this suggests similar complexes are formed both on the surface and in aqueous solution. However, the positions of some bands differ from those in the solution-phase spectrum, indicative of different strengths of interaction. Differences in interaction strength between BHA and metal ions in solution and at surfaces are also reported in the literature (Blajiev et al., 2006). The band at 1,494 cm−1 (Blajiev et al., 2006) suggests an involvement of the hydroxamate moiety in the interaction with the surface, consistent with ATR-FTIR and DFT results.

The bastnäsite-subtracted SHA-crystal spectrum (Figure S18) does not show well-defined peaks that correlate with those found in the aqueous-phase spectrum with Ce3+. There are, however, some weak peaks that can be associated with the chemical fingerprint of pure SHA in water. The low intensity of the peaks cannot be explained by a fluorescence background, as signal to noise ratios are on average twice as high for SHA as for SA or BHA. Although such a spectrum is suggestive of weak adsorption of SHA on the bastnäsite surface, an alternative explanation could be that SHA lies flat on the surface, as suggested by DFT and vSFG results, thus reducing the Raman signal from the interface.

Discussion

A multifaceted approach enlisting microflotation experiments, DFT calculations, and vSFG, ATR-FTIR, and Raman spectroscopies explored the adsorption of SHA onto the surface of bastnäsite. By comparing the results from complementary theoretical and surface-sensitive spectroscopic methods with analogous ones for BHA and SA, we obtain a mutually consistent, molecular-level understanding of the relationship between the structure of SHA and its adsorption onto bastnäsite and its selectivity for bastnäsite over calcite in froth flotation.

The DFT, vSFG, and ATR-FTIR results show that the hydroxamic acid moiety in SHA/BHA leads to stronger adsorption than the carboxylic acid moiety in SA onto bastnäsite surfaces, although DFT structures and ATR-FTIR spectra show that both moieties present similar binding modes. Furthermore, DFT, vSFG, and Raman results, combined, reveal that the aromatic ring in SHA tends to lie parallel to the bastnäsite-[100] surface at both monolayer and multilayer concentrations, whereas BHA, which differs in structure from SHA only through the omission of the ortho phenol group, shows a ring oriented more perpendicular to the surface at high ligand concentrations.

The contrast of orientations of BHA and SHA on bastnäsite and calcite highlights the role of the phenolic moiety in adsorption. Despite the disparate adsorption orientations, however, the phenolic moiety does not affect the adsorption strength of SHA onto an ideal bastnäsite-[100] surface. SHA, as revealed through DFT and vSFG results, has a similar binding energy as BHA, perhaps resulting from the recognition of the surface by the common hydroxamic acid moiety. The similarity in energies, in contrast to dissimilar microflotation results, suggests that the collector's performance in bastnäsite flotation experiments may not be captured by considering solely pristine mineral surfaces. DFT results highlight the need to also consider defective surfaces, as are likely present in mined ores. Studies at single-carbonate-vacancy-defect sites reveal that the large hydroxamic acid moiety in SHA is unable to penetrate beyond the surface into the defect, leaving an unstable metal cation at the subsurface. This instability, due to differences in the cationic spacing on each surface, is greater on the most stable calcite surface than on the most stable bastnäsite surface, leading to adsorption differences on these surfaces and thereby contributing to the efficacy of hydroxamic acid collectors in selective froth flotation.

Finally, our unique surface-sensitive approach to SHA recognition of the bastnäsite surface paves the way for a detailed account of nonbonding interactions in beneficiation systems. Although mechanistic studies of flotation have generally focused on the chelation of the hydroxamate moiety with metal ions, our piecewise approach to understanding SHA adsorption reveals the integral role of hydrogen-bonding interactions in flotation, which have only begun to be described. Future work in this field shall continue to employ parallel studies of ligand-bastnäsite and ligand-gangue interactions on both pristine and defective surfaces in order to relate a comparative molecular-level description to flotation efficacy in terms of the recognition of surfaces. Probes of hydrogen bonding must also be extended toward a complete description of interactions among the ligand, the surface, and explicit water molecules on the surface and in immediate solvation shells. We anticipate that our fundamental results relating the structure of SHA to its adsorption onto bastnäsite surfaces and its beneficiation efficacy will provide insight into the selection and design of new ligands as candidates for bastnäsite collectors.

Limitations of the Study

Microflotation results presented herein reflect a single trial for each system under the conditions presented in the Experimental Procedures section of this work. However, these results are taken from an extensive design of experiments system that included 18 collectors each, tested at a combination of three different collector concentrations and three pH values. The values reported herein reflect the most successful conditions tested in terms of recovery and grade.

DFT calculations are limited by a realistic description of surface charge and the protonation state of the ligand under experimental conditions. We expect, at a pH of 8.5, SHA and SA to be monoanionic. Owing to prohibitive computational expense, the model used in this work does not provide explicit water molecules, which would serve to stabilize this anion and its conjugate proton. As such, we instead provide two charge-balanced systems, one in which all protons are found near the adsorbed ligand and another in which two protons have been donated. In the latter, the dianionic ligand replaces a CO32− dianion on the surface.

As adsorbate frequencies likely shift as a result of adsorption, the peaks presented in vSFG experiments cannot be definitively assigned without isotopic substitution or costly vibrational spectral calculations, which are beyond the scope of this paper.

In Raman experiments, all signals coming from the bare crystal cannot be completely removed owing to intrinsic inhomogeneities of the mineral. As such, the intensity of the Raman signals are dependent on the location probed on the crystal surface. Additionally, imperfect background subtraction of bastnäsite background in ATR-FTIR spectra leaves bands that overlap with adsorbate vibrational frequencies.

Resource Availability

Lead Contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Vyacheslav S. Bryantsev (bryantsevv@ornl.gov).

Materials Availability

This study did not generate new unique reagents.

Data and Code Availability

The published article includes all datasets generated or analyzed during this study.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.