Investigating Exchange Efficiencies of Sodium and Magnesium to Access Lithium from β-Spodumene and Li-Stuffed β-Quartz (γ-Spodumene)

Abstract

After the high-temperature pretreatment of α-spodumene to induce a phase transition to β-spodumene, a derivative of the silica polymorph keatite, often coexisting with metastable Li-stuffed β-quartz (γ-spodumene)_, the conventional approach to access lithium is through ion exchange with hydrogen using concentrated sulfuric acid, which presents drawbacks associated with the production of low-value leaching residues. As sodium and magnesium can produce more interesting aluminosilicate byproducts, this study investigates Na⁺ ↔ Li⁺ and Mg²⁺ ↔ 2 Li⁺ substitution efficiencies in β-spodumene and β-quartz. Thermal annealing at 850 °C of the LiAlSi₂O₆ silica derivatives mixed with an equimolar proportion of Na endmember glass of equivalent stoichiometry (NaAlSi₂O₆) indicates that sodium incorporation in β-quartz is limited, whereas the main constraint for not attaining complete growth to a Na_0.5Li_0.5AlSi₂O₆ β-spodumene solid solution is co-crystallization of minor nepheline. For similar experiments in the equimolar LiAlSi₂O₆-Mg_0.5AlSi₂O₆ system, the efficient substitution of Mg for Li is observed in both β-spodumene and β-quartz, consistent with the alkaline earth having an ionic radius closer to lithium than sodium. Ion exchange at lower temperatures was also evaluated by exposing coexisting β-spodumene and β-quartz to molten salts. In NaNO₃ at 320 °C, sodium for lithium exchange reaches ≈90% in β-spodumene but less than ≈2% in β-quartz, suggesting that to be an efficient lithium recovery route, the formation of β-quartz during the conversion of α-spodumene needs to be minimized. At 525 °C in a molten MgCl₂/KCl medium, although full LiAlSi₂O₆-Mg_0.5AlSi₂O₆ solid solution is observed in β-quartz, structural constraints restrict the incorporation of magnesium in β-spodumene to a Li_0.2Mg_0.4AlSi₂O₆ stoichiometry, limiting lithium recovery to 80%.

1. Introduction

Lithium-cesium-tantalum (LCT) pegmatites represent the largest mineral-based source of Li, hosting α-spodumene (LiAlSi₂O₆) as the main carrier phase. Due to its refractory nature, conventional processing methods involve an energy-intensive pretreatment step in which the concentrate is brought to temperatures exceeding 1000 °C (“decrepitation” [1,2,3]) to induce a phase transition to a β-spodumene solid solution (β-spodumene_ss), a stuffed derivative of the tetragonal silica polymorph keatite [4,5] that has a more open crystalline structure with higher ionic porosity characterized by zeolite-like channels that enhance lithium mobility.

Although the decrepitation step is typically performed between 1000 and 1100 °C, extrapolation of high-pressure phase equilibria studies [6,7,8] suggest that the low-temperature boundary of the β-spodumene_ss stability field can be as low as 450 to 500 °C. This indicates that the high temperatures employed during conventional lithium recovery processes to achieve the α-β phase transition in spodumene are not dictated by thermodynamics but rather by sluggish kinetics [9]. Recent investigations into the α-β phase transition in natural spodumene, aimed at optimizing the energy consumption of lithium recovery processes, focused on lower decrepitation temperatures down to 800 °C [10,11,12,13,14,15,16]. These studies document that, at the onset of the α-spodumene conversion at such lower temperatures, a metastable Li-stuffed derivative of hexagonal β-quartz_ss, often referred to as “γ-spodumene”, is commonly observed as the initial transition phase or growing simultaneously with β-spodumene_ss. Therefore, by minimizing the temperature of the phase transition, ion-exchange efficiency to access lithium needs to be considered not only in β-spodumene_ss, but also in β-quartz_ss.

While exchange with hydrogen (i.e., H⁺ ↔ Li⁺) using concentrated sulfuric acid baking has been widely applied for β-spodumene_ss, this approach suffers drawbacks such as the production of large amounts of low-value sulfate salt (e.g., Na₂SO₄) and hydrogen aluminosilicate (HAlSi₂O₆) byproducts [17], estimated by Xing et al. [18] at 0.8 T leaching residue produced per 1 T of α-spodumene processed. In this context, exchange with sodium is gaining consideration in order to produce a cleaner crystalline alkali aluminosilicate residue, which can be converted into marketable products, for example, zeolites [18,19,20] or analcime [17,21]. The proposed processes are typically performed in a pressure vessel (autoclave) with sodium exchange media ranging from NaCl [17], NaCl/NaOH [17], or Na₂CO₃ [21] aqueous solutions using, as a feed, a spodumene concentrate initially pretreated at >1000 °C to induce the α-β phase transition, leading to a grain size of ≈100 μm or less. To attain a Li recovery of >90%, optimal temperatures are typically in the range of 200 °C, with aggressive agitation (350 to 600 rpm, in some cases, with the addition of steel balls [21]) and residence times ranging from 1 to 3 h.

Although currently not considered for lithium recovery applications, exchange with magnesium, a divalent cation that has an ionic radius much closer to lithium (Li⁺: 0.059 nm, Na⁺: 0.099 nm, Mg²⁺: 0.057 nm, all in fourfold coordination; [22]), allowing recovery of two moles of lithium per mole of magnesium, could represent an interesting alternative. The incorporation of magnesium in stuffed silica derivatives has been investigated in the context of improving the properties of low thermal expansion glass-ceramics [5,23,24,25,26,27,28,29]. Through liquid crystallization and glass annealing in the Li₂O-MgO-Al₂O₃-SiO₂ system, sometimes in the presence of nucleating agents such as TiO₂ and ZrO₂, crystallization up to the magnesium endmember in β-quartz [23,24,25,26] and keatite solid solutions attaining magnesium for lithium substitution around 40 to 50% [5,27] could be achieved.

The goal of this study was to evaluate the relative efficiencies of sodium and magnesium exchange routes to access lithium, not only from β-spodumene_ss, but also from β-quartz_ss that can coexist in significant concentrations after decrepitation of the α-spodumene concentrate, especially if performed at lower temperatures. The approach taken was to conduct controlled experiments combined with detailed characterization of the reactive product in order to obtain a rigorous comparison of the potential for the alkali and alkaline earth to substitute for lithium in both Li-stuffed silica polymorphs, with the hope of providing objective information for future process optimization. First, annealing experiments were performed to investigate the extent of solid-state diffusion and crystal growth within synthetic mixtures consisting of equimolar proportions of LiAlSi₂O₆ silica derivative (either β-spodumene_ss or β-quartz_ss) and endmember MAlSi₂O₆ glass (M = Na⁺ or 0.5 Mg²⁺). Secondly, fragments, where both β-spodumene and β-quartz coexist, were exposed to molten salts to determine their relative efficiency in fully exchanging lithium for sodium or magnesium.

In the context of lithium recovery, the name “β-spodumene” is widely accepted to describe the stuffed keatite derivative obtained from the conversion of α-spodumene (LiAlSi₂O₆) after the high-temperature heat treatment. Moreover, β-spodumene_ss is a useful term to express the tetragonal solid solution stable within a wide window along the LiAlO₂-SiO₂ join [5]. On the other hand, for solid solutions resulting from the substitution of Li by Na and Mg, a nomenclature that refers specifically to the keatite silica polymorph (keatite_ss) is usually adopted for glass-ceramics applications [30]. However, with the hope of preventing confusion, considering that the purpose of this work is more focused on lithium recovery, we opted to keep the term β-spodumene_ss. In the case of the hexagonal β-quartz polymorph, the situation is simpler, and β-quartz_ss is used throughout this paper to refer to the solid solution.

2. Methods

2.1. Characterization

2.1.1. X-Ray Diffraction

The crystalline structure characteristics of the experimental samples were assessed by powder X-ray diffraction (XRD) on finely ground material (<45 μm) using a Rigaku D/MAX 2500 rotating anode system with monochromatic Cu Kα radiation (λ: 0.154059 nm) operated at 40 kV and 200 mA. Powder patterns were typically collected using a step scan of 0.02° in 2θ with a dwell time of 1 s per step. Phase identification was achieved using JADE version 9 interfaced with the current International Centre for Diffraction Data (ICDD) database. Unit cell dimensions of β-spodumene_ss (tetragonal P4₃2₁2) and β-quartz_ss (hexagonal P6₂22) were refined using initial parameters from Li and Peacor [4] and Li [31], respectively, through CrystalDiffract version 7. In some cases, a more complete Rietveld analysis was also performed, the details of which can be found in Supplementary Information Section S2.

2.1.2. Raman Spectroscopy

Raman spectra were collected on an Edinburgh Instruments RM5 system fitted with a 785 nm laser source using a 100× objective, a grating of 1200 grooves mm⁻¹, and a 50 μm slit, with periodic calibration of the spectrometer to the 520.5 cm⁻¹ band of a Si single crystal. For every discrete analysis, 15 acquisitions of 20 s were collected. Artifacts arising from cosmic rays were removed using the Edinburgh Instruments Ramacle software. Considering that the grains characterized can be a fairly small, minor contribution from the surrounding resin within the polished section of the products, when identified, it was labeled as “epoxy”.

2.1.3. Electron Probe X-Ray Microanalysis

The textural and compositional nature of the synthesized material and the products obtained from crystallization and ion-exchange experiments were characterized by backscattered electron (BSE) imaging and quantitative X-ray microanalyses by wavelength-dispersive spectrometry (WDS) using a JEOL JXA 8230 Electron Probe Microanalyzer (EPMA) operated with an accelerating voltage of 20 kV and a probe current of 8 to 50 nA. Counting times for discrete analyses ranged from 8 to 30 s, typically with a beam defocused at 20 to 30 μm to prevent alkali mobility. Quantitative X-ray maps were collected with 1 to 3 μm steps and dwell times of 35 milliseconds. The characteristic X-ray lines and standards used for the analyses were Na Kα (cleavelandite), Al Kα (kunzite, cleavelandite, orthoclase), Si Kα (kunzite, cleavelandite, orthoclase), and Mg Kα (forsterite). For proper matrix correction, considering that no lithium characteristic X-ray lines can be detected by WDS, the abundance of Li₂O was estimated by the difference in the calculated total relative to 100 wt%.

2.1.4. Automated Phase Analysis

Modal abundances of the phases identified within the products of synthesis and exchange experiments were obtained using a TESCAN Integrated Mineral Analyzer (TIMA) equipped with four silicon drift X-ray energy-dispersive spectrometers (EDSs). Phase maps were collected from polished cross-sections using an accelerating voltage of 25 kV and a beam current of 5.5 nA with a step size of 1 µm. The BSE signal and EDS spectra were acquired simultaneously and were both used to discriminate phases based on their compositional differences.

2.1.5. Compositional Space

The compositional space and the main exchange vectors investigated in this study are visualized by expressing the phase chemistry in terms of atomic ratios of silicon, sodium, and half-magnesium (Mg_0.5), all relative to aluminum (Figure 1). The silicon-to-aluminum ratio ([Si:Al]_atomic) represents the extent of silica (SiO₂) substitution for the alkali or alkaline earth aluminate (NaAlO₂, Mg_0.5AlO₂), characteristic of stuffed silica derivatives. On the other hand, although the compositions are characterized by WDS-EPMA, where Li cannot be detected (Section 2.1.3), considering that the incorporation of a cation (e.g., Li¹⁺, Na¹⁺, Mg²⁺) within the silica derivatives is charge-balanced by Al³⁺ substituting for Si⁴⁺ in the tetrahedral framework [32], the efficiency of sodium or magnesium to exchange for lithium is equivalent to either the sodium over aluminum ([Na:Al]_atomic) or, in the case of the magnesium system, the ratio of half the Mg relative to Al ([Mg_0.5:Al]_atomic), due to its divalent nature.

2.2. Materials Synthesis

2.2.1. Glass

The preparation of glasses with NaAlSi₂O₆, Li_0.5Na_0.5AlSi₂O₆, Mg_0.5AlSi₂O₆, and Li_0.5Mg_0.25AlSi₂O₆ compositions was conducted using high-purity silica glass (SiO₂ > 99.9%), fired alumina (Al₂O₃ > 99.9%), as well as carbonates with an estimated purity of ≥99% Li₂CO₃, ≥99.5% Na₂CO₃, and ≥99% MgCO₃, mixed in the appropriate stoichiometric proportions, loaded in a Pt crucible and heated above their liquidus (1300–1500 °C) in a bottom loading furnace for dwell times up to 3 h. The melt was then quenched to a glass by rapidly dropping and partially immersing the Pt crucible in a eutectic NaCl-H₂O brine (23.3 wt% NaCl) at ≈−21 °C [33]. To improve homogeneity, the glasses were finely ground and re-melted at least twice. A lithium endmember glass was also produced following the same protocol but using, as starting material, a natural single crystal of kunzite (α-spodumene) having a composition, as determined by WDS-EPMA, close to the endmember LiAlSi₂O₆ formula (≈99.8 mol% LiAlSi₂O₆) with Na (0.15 wt% as Na₂O) as the only minor element with a concentration of more than 0.1 wt%.

The XRD patterns shown in Figure 2a,b confirm that all synthesized glasses are fully amorphous with no identified crystalline phases. Their chemical compositions, as determined by WDS-EPMA (Figure 2c,d), are very close to the expected ideal stoichiometries displayed in Figure 1a,b. The level of homogeneity is within 2% relative, the only exception being the sodium endmember glass, where the [Si:Al]_atomic ratio displays a wider variation of 8% relative (Figure 2c), likely related to a higher melt viscosity at peak temperature during the synthesis. Finally, considering that the starting material for synthesizing the lithium endmember glass was kunzite, which contains 0.15 wt% as Na₂O, the [Na:Al]_atomic ratio is actually slightly above 0 (0.01).

2.2.2. Li-Stuffed Silica Derivatives Endmembers

β-quartz_ss was produced by thermal annealing (850 °C, 4 h) of the LiAlSi₂O₆ glass that was initially ground to less than 75 μm and pelletized [31,34]. The only identified crystalline phase by XRD after heat treatment is indeed hexagonal β-quartz (Figure 3a). This is further supported by a frequency at 483 cm⁻¹ for the main Raman band (Figure 3b), which can be assigned to the A1 symmetric stretching mode (u_s) of an oxygen bridging two tetrahedral (T) cations (u_s [T-O-T], where T = Si⁴⁺_, Al³⁺; [35,36]). Compositionally, with an average [Si:Al]_atomic ratio of 1.98 and sodium below the detection limit (Na₂O < 0.03 wt%) yielding a [Na:Al]_atomic ratio of less than 0.002, it has almost the ideal LiAlSi₂O₆ stoichiometry. The excess sodium contained in the kunzite glass precursor (Section 2.2.1) is accommodated in small residual Na-rich glass cores (Figure 4b) having a [Na:Al]_atomic ratio ≈ 0.32, which, based on automated phase analysis (Section 2.1.4), represents only a small modal proportion of 2% calculated on a molar basis (Figure 4a).

β-spodumene_ss, being the stable LiAlSi₂O₆ silica derivative at atmospheric conditions, was synthesized by melting one gram of the kunzite precursor at 1430 °C, followed by slow cooling at a rate of 5 °C/h to promote crystal growth. However, as discussed by Thibault and Gamage McEvoy [9], to prevent the formation of any metastable β-quartz_ss, the kunzite starting material was initially doped with minor Na₂CO₃ in order to slightly increase the [Na:Al]_atomic ratio from 0.01 to 0.05 resulting, as confirmed by XRD and the characteristic u_s [T-O-T] Raman band at 495 cm⁻¹, with minor bands at ≈186 and ≈288 cm⁻¹ [35], in the formation of single-phase polycrystalline β-spodumene_ss (Figure 3c,d and Figure 4c) with a minor substitution of ≈5% Na for Li ([Na:Al]_atomic ratio of 0.05; Figure 4a).

A starting material with coexisting crystals of β-spodumene_ss and β-quartz_ss was also produced by slow cooling (10 °C/h) of a LiAlSi₂O₆ melt. However, in this case, to promote the co-crystallization of metastable β-quartz_ss, the precursor was prepared using high-purity SiO₂, Al₂O₃, and Li₂CO₃, ensuring no contribution from sodium. The XRD pattern shown in Figure 5a confirms the presence of both silica polymorphs in the product. BSE imaging of an mm-sized fragment (Figure 5b) indicates the coexistence of two phases, the brightest and darkest being, based on Raman characterization, β-spodumene_ss and β-quartz_ss, respectively (Figure 5c). This BSE contrast between both phases reflects a slight difference in mean atomic number, where the average [Si:Al]_atomic ratio of β-spodumene_ss is slightly higher (Figure 5d).

2.3. Solid-State Crystallization

Equimolar proportions of the Na or Mg endmember glass (Figure 2) and one of either β-spodumene_ss or β-quartz_ss (Figure 3 and Figure 4) were ground to less than 75 μm, pelletized, and loaded in a Pt crucible. The pellet was then annealed in a muffle furnace at 850 °C for 14 h, after which it was broken into fragments, a portion of which was kept for powder-XRD analysis, with the remaining being embedded in resin and polished for textural, structural, and chemical characterization at the crystal-scale.

2.4. Ion Exchange in Molten Media

Ion-exchange experiments in molten salt media were performed to investigate the extent of Na or Mg substitution for Li in coexisting β-spodumene_ss and β-quartz_ss. Approximately 0.2 g of material containing both Li-stuffed silica derivatives (Figure 5) and 3 g of NaNO₃ or a [40%_molar KCl/60%_molar MgCl₂] salt mixture were loaded in a covered alumina crucible and brought to a peak temperature of 320 °C (NaNO₃) to 525 °C (KCl/MgCl₂), within a horizontal tube furnace kept under inert argon atmosphere. After cooling, the solidified nitrate or chloride salt was dissolved in DI water, and the silicate fragments were recovered for characterization.

3. Results and Discussion

3.1. Solid-State Crystallization

In contrast to solid-state experiments in which a single-phase glass is crystallized by heat treatment [23,24,25,26,27], the approach taken was to thermally anneal β-spodumene_ss or β-quartz_ss (LiAlSi₂O₆) mixed with an equimolar proportion of Na or Mg endmember glass of equivalent stoichiometry in terms of alumina and silica (MAlSi₂O₆, M = Na, Mg_0.5). Consequently, in addition to the amount of alkali or alkaline earth that can be structurally accommodated in the silica derivative, the potential for crystal growth toward the bulk composition (Li_0.5Na_0.5AlSi₂O₆ or Li_0.5Mg_0.25AlSi₂O₆) is also controlled by the rate of Na⁺-Li⁺ or Mg²⁺-Li⁺ interdiffusion within and at the interface of both phases. All experiments were performed following the protocol described in Section 2.3.

3.1.1. LiAlSi₂O₆—NaAlSi₂O₆

Characterization of the product after heat treatment of an equimolar mixture of Li-stuffed β-quartz_ss and NaAlSi₂O₆ glass indicates that both phases are still present (Figure 6a and Figure 7a), displaying distinct intensities in the BSE image shown in Figure 6b. The measured molar proportion of the dark grains, which can be identified as β-quartz_ss by Raman spectroscopy (Figure 6c), is estimated at 52%, representing negligible growth from the initial modal content of 50%. This is consistent with the limited increase in its [Na:Al]_atomic ratio (Figure 6d), reflecting the substitution of Li by Na, ranging from only ≈3% to ≈14%, confirming the inability of β-quartz to structurally incorporate the larger alkali [9,28] rather than a slow Na-Li interdiffusion rate.

In contrast, in a similar solid-state experiment, extensive growth of β-spodumene_ss is observed from an initial molar proportion of 50% to 87% at the expense of the glass phase reduced to only 7% and coexisting with 6% of newly formed crystalline nepheline (Figure 7b and Figure 8). This is consistent with the substitution of Na for Li in β-spodumene_ss reaching an average of 42% (Figure 8d). Compared with the pristine β-spodumene_ss (Figure 3c), the XRD pattern shown in Figure 8a reveals a significant shift to lower 2θ of the (102) lattice plane from 22.62° to 22.27° as well as the (112) going from unresolved to a shoulder on the low 2θ flank of the (201) diffraction peak, which, based on Rietveld analysis (Section S2 of Supplementary Information), reflects a unit cell expansion along the c-axis direction to accommodate the larger sodium cation (

Table 1. Unit cell dimensions of β-spodumene_ss and β-quartz_ss before and after solid-state heat treatment with equimolar proportion of sodium or magnesium endmember glasses.

Unit Cell Parameter	β-quartzss			β-spodumeness
	a (Å)	c (Å)	V (Å3)	a (Å)	c (Å)	V (Å3)
Li-stuffed silica derivative endmembers
β-quartzss	5.220	5.451	128.62
β-spodumeness				7.549	9.203	524.39
Solid-state heat treatment
LiAlSi2O6 (β-quartzss)—NaAlSi2O6	5.236	5.451	129.41
LiAlSi2O6 (β-spodumeness)—NaAlSi2O6				7.516	9.399	530.95
LiAlSi2O6 (β-quartzss)—Mg0.5AlSi2O6	5.207	5.412	127.07
LiAlSi2O6 (β-spodumeness)—Mg0.5AlSi2O6	5.197	5.417	126.72	7.538	9.176	521.39

). Nepheline, a stuffed derivative of the silica polymorph tridymite, displays an average [Si:Al]_atomic ratio of 1.95 (Figure 8d), much closer to a NaAlSi₂O₆ stoichiometry than the ideal NaAlSiO₄ endmember ([Si:Al]_atomic = 1), emphasizing the high degree of solid solution along the SiO₂-NaAlO₂ join. Detailed BSE imaging of the nepheline habit (inset of Figure 8b) reveals that it surrounds and therefore isolates the residual pools of glass, indicating that its growth likely represents the main limiting factor for not attaining complete β-spodumene_ss crystallization at the system bulk composition (Li_0.5Na_0.5AlSi₂O₆).

3.1.2. LiAlSi₂O₆—Mg_0.5AlSi₂O₆

The extent of crystallization and substitution of divalent Mg²⁺ for Li⁺ in the tetragonal and hexagonal Li-stuffed silica derivatives when in contact with the equimolar proportion of the Mg glass endmember (Figure 2d) at 850 °C for 14 h was also investigated following the same experimental protocol (Section 2.3).

Characterization of the product from the experiment performed with the Li-stuffed β-quartz_ss reveals a significant increase in the measured modal proportion of the silica derivative from 50% up to 78% through crystal growth initiated at the rim of the receding glass (Figure 9). The exchange of Mg for Li is much more pronounced than what was observed in the case of Na (Section 3.1.1), with [Mg_0.5:Al]_atomic ratios from 0.21 up to 0.45 in the existing grains and reaching 0.49 within the new growth (Figure 9d), while, as confirmed by Raman spectroscopy, maintaining the β-quartz_ss structure (Figure 9c).

The degree of crystallization for the experiment performed with β-spodumene_ss is comparable to what was observed with β-quartz_ss, leading to a modal proportion of silica derivatives estimated at 79% (Figure 10). However, the XRD pattern collected indicates that β-spodumene_ss now coexists with β-quartz_ss (Figure 10a), suggesting that the new growth observed within the glass fragments (Figure 10b) is dominated by the hexagonal phase. Both phases show the significant substitution of Mg for Li, which ranges from 35 to 45% in the original β-spodumene_ss grains and up to 52% in β-quartz growing within the glass fragments (Figure 10d).

In conclusion, in the equimolar LiAlSi₂O₆-Mg_0.5AlSi₂O₆ system, rapid magnesium and lithium diffusion is observed, and significant substitution of Mg for Li is achieved, not only in the new growth but also throughout the original β-quartz_ss and β-spodumene_ss, suggesting that further annealing would induce complete crystallization to the bulk Li_0.5Mg_0.25AlSi₂O₆ stoichiometry. This is consistent with magnesium having an ionic radius very close to lithium, requiring a relatively small unit cell contraction (Table 1).

3.2. Melt Crystallization

To investigate whether β-spodumene_ss with substitution for lithium exceeding 40%, observed in both the sodium (LiAlSi₂O₆-NaAlSi₂O₆) and magnesium (LiAlSi₂O₆-Mg_0.5AlSi₂O₆) equimolar solid-state systems, represents a stable phase, synthetic Na_0.5Li_0.5AlSi₂O₆ and Mg_0.25Li_0.5AlSi₂O₆ glasses (Section 2.2.1) were brought over their liquidus and slowly cooled at a rate of 10 °C/h to induce crystallization (Section 2.2.2).

In the case of the Li_0.5Na_0.5AlSi₂O₆ composition, large β-spodumene_ss crystals coexisting with glass were produced (Figure 11). Compositionally, Na strongly partitioned in the melt, with no more than 6% substitution for Li in the β-spodumene_ss phase (Figure 11d). In contrast, for the Li_0.5Mg_0.25AlSi₂O₆ stoichiometry, complete crystallization dominated by β-spodumene_ss with minor mullite (3 Al₂O₃ · 2 SiO₂) is observed (Figure 12). Mg fully partitioned in β-spodumene_ss ([Mg_0.5:Al]_atomic ratio: 0.5) with a slight alumina depletion relative to silica ([Si:Al]_atomic ratio: 2.05–2.2) that can be explained by Al-rich mullite crystallizing first, as evidenced by its euhedral habit (Figure 12b,d). Based on these observations, β-spodumene_ss with an appreciable amount of substitution of Na for Li represents a metastable phase, whereas solid solutions of the keatite polymorph can be stable with the incorporation of magnesium up to at least Li_0.5Mg_0.25AlSi₂O₆ stoichiometry [5].

3.3. Ion Exchange in Molten Salts

In the context of optimal lithium recovery, after the conversion of α-spodumene to Li-stuffed silica derivatives, the potential of an ion-exchange route with sodium or magnesium needs to be evaluated at lower temperatures, hence requiring a liquid media in order to promote diffusion and interfacial contact. Sodium nitrate salt (NaNO₃), with a low melting point at 308 °C, has been used extensively as an infinite reservoir for ion exchange in glass and glass-ceramics [28]. Similarly, Baumgartner and Müller [37] have synthesized a sodium keatite endmember (NaAlSi₂O₆) from an isostructural HAlSi₂O₆ precursor immersed in a molten NaNO₃ bath at 320 °C for 24 h. For magnesium, choices of low-temperature molten phases seem more limited; nevertheless, although MgCl₂ melts at 714 °C, the presence of eutectics below 500 °C along the MgCl₂-KCl join [38,39] provides options to minimize the ion-exchange temperature. Since a significant amount of β-quartz_ss (γ-spodumene) is often observed in the product of the conversion from α- to β-spodumene, the molten salt ion-exchange experiments were performed on synthesized fragments where both lithium-stuffed silica derivatives coexist (Section 2.2.2).

After a 24 h treatment in molten NaNO₃ at 320 °C, XRD characterization indicates that, structurally, both polymorphs are preserved (Figure 13a). However, although the unit cell dimensions of β-quartz_ss are essentially unchanged, for β-spodumene_ss, there is a significant expansion along the c-axis, accompanied by a slight contraction of the a-dimension (

Table 2. Unit cell dimensions in coexisting β-spodumene_ss and β-quartz_ss before and after ion exchange in molten media.

Unit Cell Parameter	β-quartzss			β-spodumeness
	a (Å)	c (Å)	V (Å3)	a (Å)	c (Å)	V (Å3)
Synthetic coexistingβ-spodumeness/β-quartzss (ion exchange duration: 24 h)
Pristine	5.227	5.459	129.16	7.549	9.179	523.04
Exchange with NaNO3 (320 °C)	5.231	5.464	129.49	7.494	9.648	541.90
Exchange with MgCl2/KCl (525 °C)	5.168	5.383	124.49	7.501	9.118	513.00
Natural spodumene/quartz intergrowth heat-treated at 1100 °C (ion exchange duration: 2 h)
Pristine	5.223	5.462	129.05	7.546	9.174	522.38
Exchange with NaNO3 (320 °C)	5.219	5.470	129.00	7.502	9.595	539.98
Exchange with MgCl2/KCl (525 °C)	5.171	5.392	124.84	7.503	9.130	513.93

, Figure S16), resulting in values very close to those reported by Baumgartner and Müller [37] for a pure NaAlSi₂O₆ keatite_ss endmember produced by ion exchange with HAlSi₂O₆ (a = 7.483 Å; c = 9.629 Å). BSE imaging and quantitative elemental mapping of an exchanged fragment reveal a sharp compositional contrast between the coexisting phases (Figure 13c,d), whose identity can be confirmed by Raman spectroscopy (Figure 13b). Whereas the [Na:Al]_atomic ratio reaches values ranging from 0.89 to 0.92 across the entire β-spodumene_ss region, the substitution of Na for Li is once again negligible, at less than 2%, throughout the β-quartz_ss structure.

The magnesium exchange experiment was performed at 525 °C for 24 h within a molten media consisting of 60 mol% MgCl₂ and 40 mol% KCl, a composition close to the 470 °C eutectic [38,39,40]. In this case, the XRD patterns collected on the pristine sample and the exchanged product indicate a positive shift in 2θ for the main diffraction peaks in both structures (Figure 14a) due to a contraction of their unit cells (Table 2). This is consistent with the important yet distinct extent of substitution of magnesium for lithium, which, based on the quantitative distribution of the [Mg:Al]_atomic ratio shown in Figure 14d, reaches 81% in β-spodumene_ss and 97% in β-quartz_ss, the identity of both phases being confirmed by Raman spectroscopy (Figure 14b). The level of compositional homogeneity observed within each coexisting phase (Figure 14d) suggests that this difference in Mg-Li exchange reflects a structural constraint of the keatite polymorph to fully accommodate the alkaline earth, at least at that temperature, as opposed to a kinetic control. The presence of potassium in the molten salt is not believed to have played a role, as it is not detectable in either phase.

Compared with the synthetic mm-sized fragments of coexisting Li-stuffed silica derivatives, the grain size of a natural spodumene concentrate is typically in the range of 200 to 400 μm and eventually reduces to ≈100 μm or less by decrepitation during the pretreatment stage to induce the α-β phase transition. In this context, a sample from the Tanco LCT pegmatite (Bernic Lake, MB, Canada), consisting of intergrowths of α-spodumene and quartz, ground and sieved to a size fraction of 355 to 600 μm, was heat-treated at 1100 °C for 4 h, resulting in the conversion to β-spodumene_ss with minor β-quartz_ss (Figure 15). With this material, the sodium and magnesium molten salt exchange experiments were repeated at similar temperatures, 320 °C and 525 °C, respectively, but for a time period of only 2 h. The variation in unit cell dimensions (Table 2) obtained by XRD characterization (Figure 15) indicates that the extent of sodium and magnesium substitution for lithium is comparable to what was observed for the mm-sized fragments at the longer duration of 24 h.

4. Conclusions

Both sodium and magnesium present advantages as alternatives to recover lithium through a clean direct exchange without collapsing the β-spodumene_ss (keatite_ss) and β-quartz_ss structures. However, in each case, there are challenges that need to be considered.

A sodium exchange route in a molten or liquid media can be efficient to access more than 90% of lithium at temperatures down to at least 320 °C, but the formation of β-quartz_ss during decrepitation of the α-spodumene concentrate needs to be minimal as, otherwise, it will retain lithium due to its inability to accommodate the larger alkali. This may be a concern in the context of inducing the α-β phase transition in spodumene below 1000 °C, where the metastable hexagonal phase tends to become more prominent [10,11,12,13,14,15,16]. However, as observed by Thibault and Gamage McEvoy [9], minor sodium addition during the conversion of the α-spodumene concentrate, even at lower temperatures, could strongly depress the growth of β-quartz_ss.

Exchange with magnesium represents a flexible alternative, considering that, as a divalent cation, it allows access to two moles of lithium for each mole of magnesium. Additionally, due to its ionic radius very close to lithium, extensive Mg-Li substitution with minor unit cell modifications is achieved within both tetragonal and hexagonal silica derivatives. Yet, there are not many magnesium compounds with low melting points or high solubility, and at 525 °C, although full LiAlSi₂O₆-Mg_0.5AlSi₂O₆ solid solution is observed in β-quartz_ss, the incorporation of magnesium in β-spodumene_ss seems restricted to a Li_0.2Mg_0.4AlSi₂O₆ stoichiometry, limiting access to lithium at 80%.