Separation of GaCl3 from AlCl3 by Solid-Liquid Extraction and Stripping Using Anhydrous n-Dodecane and NaCl.

Separation of GaCl3 from other associating chloride compounds (e.g. AlCl3, SbCl3, and InCl3) is generally achieved by hydrometallurgical processes. In this study, we explore the separation of GaCl3 from these compounds on the basis of the exceptionally high solubility of GaCl3 in hydrocarbon solvents. We found that GaCl3 can be efficiently extracted by anhydrous n-dodecane from a solid mixture of GaCl3 and AlCl3; on the contrary, SbCl3 and InCl3 significantly reduce the extraction of GaCl3. On the basis of Lewis acidity theory and study of the Raman spectra, it is shown that formation of the ionic compound [SbCl2][GaCl4] is responsible for the reduced GaCl3 extraction. Formation of [InCl2][GaCl4] is also likely, but further study is needed to support the existence of this compound. Further making use of the strong Lewis acidity of GaCl3, GaCl3 can be efficiently stripped from the loaded n-dodecane phase by solid NaCl through formation of NaGaCl4. The extraction of GaCl3 by n-dodecane, in combination with its stripping by NaCl, is a solvometallurgical process that is essentially different from the hydrometallurgical processes for the separation of GaCl3 and AlCl3.

Introduction

Known as “the backbone of the electronics industry”, gallium is used widely in optoelectronics and particularly for semiconductors.[1−4] It also finds applications in alloys[5,6] and biomedical applications.[7] Gallium does not form its own ores, but it is produced primarily as a byproduct of the treatment of bauxite for alumina production in the Bayer process.[1,8] Besides, secondary resources such as waste semiconductors (e.g. InGaN, GaSb) are potential sources for gallium.[9−11] Therefore, purification of gallium via a chloride route involves the separation of GaCl3 and these associating chloride compounds, including AlCl3, InCl3, or SbCl3. Hydrometallurgical processes, such as solvent extraction and ion exchange, are the main processes for the separation of gallium from these associating elements.[8,10,12−14] While established, these processes usually require multiple separation stages due to the similar affinity of these compounds to extractants and ion exchangers.

In contrast to the generally negligible solubility of inorganic salts in hydrocarbons, GaCl3 has the unique property of being highly soluble in aliphatic hydrocarbons: up to 1230 g of GaCl3 can be dissolved in 1 L of hexane at 60 °C.[15] GaCl3 is likely dissolved in hexane as a mixture of monomer, dimer, and even trimer, as indicated by a 71Ga NMR study,[16] whereas GaCl3 forms a π-complex in aromatic solvents.[16,17] This high solubility of GaCl3 in hydrocarbons could be exploited for purification of gallium. Iwantscheff and Dötzer proposed that volatile hydrocarbons, such as n-pentane, n-hexane, and cyclohexane can be used to extract GaCl3 from AlCl3 and InCl3, followed by removal of the hydrocarbon solvents by distillation.[15] However, these solvents raise significant environmental concerns: they are volatile organic compounds (VOC) and highly flammable, and some of them, such as n-hexane, are neurotoxic.[18−20] Moreover, distillation is an energy-intensive operation. More importantly, the actual performance of GaCl3 separation from other chloride compounds using hydrocarbons is not known, because no studies can be found except for the brief descriptions in the patent of Iwantscheff and Dötzer. Therefore, the feasibility of GaCl3 separation from other chloride compounds using hydrocarbons warrants more investigations.

In this study, we investigate the use of n-dodecane, which is a much less volatile and an environmentally more benign solvent, to extract GaCl3, and we explore a suitable stripping process based on precipitation rather than distillation. In addition, we discuss the extraction and stripping mechanisms. This work fits into our research program on the development of solvometallurgy.[21]

Experiment and Methods

Extraction

Separations of GaCl3 from metal chloride compounds were tested by mixing 500 mg of GaCl3 with different amounts of other chloride compounds with various molar ratios of MCl3/GaCl3 (M = Al, In, and Sb). Then, 10 mL of anhydrous n-dodecane (or n-hexane) was added to these mixtures and they were shaken for 30 min, although extraction equilibrium could be reached in 1 min. To analyze the extraction efficiency, the loaded n-dodecane phases were brought in contact with an equal volume of Milli-Q water and shaken for 60 min. The metals in the n-dodecane phase were found to be completely stripped, and the resultant aqueous solution was measured by total reflection X-ray fluorescence spectrometry (TXRF, Bruker S2 Picofox) or inductively coupled plasma optical emission spectrometry (ICP-OES, PerkinElmer Optima 8300). Raman spectra were recorded (by Bruker Vertex 70 spectrometer) for the pure compounds of AlCl3, SbCl3, and InCl3 and the solid residues after extraction of GaCl3 by anhydrous n-dodecane (or n-hexane) from the solid binary mixtures of GaCl3/AlCl3, GaCl3/SbCl3, and GaCl3/InCl3.

Stripping

LiCl, NaCl, and KCl powders were ground by a mortar grinder (Fritsch Pulverisette 2) and sieved by a sieve shaker (Fritsch Analysette 3) with a sieve mesh of 250 μm. Then the ground and sieved salts were put into a round-bottom flask and connected to a Schlenk line to dry them for 24 h at 100 °C. Subsequently, the salts were used for stripping experiments. Each of the samples containing 10 mL of loaded n-dodecane with 50 g·L–1 GaCl3 was contacted with the ground MCl (M = Li, Na, K) in varying MCl/GaCl3 molar ratios and shaken for 120 min, followed by settling for 5 min. The resultant n-dodecane solution was then stripped by Milli-Q water by shaking for 60 min, and the aqueous solution was measured by TXRF for Ga concentration. Raman spectra were recorded for the pure salts of LiCl, NaCl, and KCl and the solid mixtures after stripping of GaCl3 by these salts.

Computational Details

The geometries for the compounds of Ga2Cl6, InGaCl6, [GaCl4]−, and [InCl2]+ were studied in the gas phase by density functional theory (DFT).[22] The structures were optimized at the B3LYP/def2tzvp theory of level.[23−25] The frequency calculations were also performed to confirm that the obtained structures correspond to minima on their potential energy surfaces (PES). Based on the optimized structures, the Raman spectra of the compounds were calculated to compare with the experimental results. All calculations were performed by Gaussian 09.[26] The Multiwfn code was used for data postprocessing.[27]

Results and Discussion

Extraction of GaCl3

Results of GaCl3 extraction by anhydrous n-dodecane from the solid binary mixtures of GaCl3/AlCl3, GaCl3/SbCl3, and GaCl3/InCl3 are presented in Figure . GaCl3 was efficiently extracted from GaCl3/AlCl3 mixtures, even at very high AlCl3-to-GaCl3 ratios. However, the presence of InCl3 and SbCl3 significantly reduced the extraction efficiency of GaCl3. The difference in the GaCl3 extraction efficiency from the tested chloride compounds might be attributed to the differences in Lewis acidity of these compounds. The sequence of Lewis acidity is AlCl3 > GaCl3 > InCl3 > SbCl3.[28−30] The GaCl3 dissolved in the anhydrous n-dodecane may react with InCl3 or SbCl3, which are weaker Lewis acids than GaCl3, forming the ionic compounds of [InCl2][GaCl4] and [SbCl2][GaCl4] by chloride ion transfer.[31] These ionic compounds are insoluble in nonpolar solvents such as n-dodecane due to their high lattice energy and the poorly solvating properties of these solvents, thus inhibiting the extraction of GaCl3 by n-dodecane. In contrast, as AlCl3 is a stronger Lewis acid than GaCl3, one expects that the chloride ion transfers from GaCl3 to AlCl3, leading to the Lewis adduct of [GaCl2][AlCl4]. However, the efficient extraction of GaCl3 from the GaCl3/AlCl3 mixture indicates that the influence of AlCl3 on GaCl3 extraction is almost negligible.

Figure 1

Solid–liquid extraction of GaCl3 from binary mixtures of GaCl3 and metal chloride compounds by anhydrous n-dodecane.

To understand how the chloride compounds affect GaCl3 extraction, Raman spectra were measured for the solid residues of the binary mixtures after extraction of GaCl3 by anhydrous n-dodecane (Figure ). For the spectrum of the GaCl3/SbCl3 mixture, four extra bands were observed in comparison with the spectrum of pure SbCl3. The Raman bands at 127 and 354 cm–1 can be assigned to the [GaCl4]− species, and the Raman bands at 206 and 376 cm–1 can be assigned to [SbCl2]+.[32,33] Similarly, for the Raman spectrum of the GaCl3/InCl3 mixture, the bands at 133 cm–1 can be assigned to [GaCl4]−, which is reported to be at 128 cm–1.[33] The band at 362 cm–1 may be due to the vibration of [InCl2]+, as reported to be at 367 cm–1.[34] The Raman spectrum assignments indicate the formation of the ionic compounds of [SbCl2][GaCl4] and [InCl2][GaCl4]. Binary GaCl3/InCl3 mixtures have been reported to form an eutectic system,[35,36] but this does not exclude the possibility of forming an ionic compound under a different condition. For example, the mixed dimer InGaCl6 can be formed in the vapor phase.[37] GaCl3 dissolved in n-dodecane should have a higher propensity to react with SbCl3 or InCl3 than in the solid or molten state where the eutectic is formed. In contrast to the mixtures of GaCl3/SbCl3 and GaCl3/InCl3 that showed extra Raman bands, the Raman spectrum of a GaCl3/AlCl3 mixture was almost identical to that of pure AlCl3, indicating that the formation of either [AlCl2][GaCl4] or [GaCl2][AlCl4] is unlikely. The absence of formation of an ionic compound for the GaCl3/AlCl3 mixture can be explained by the very similar Lewis acidity of GaCl3 and AlCl3,[28−30] and this result is consistent with the efficient solid–liquid extraction of GaCl3 from the binary mixture of GaCl3/AlCl3 (Figure ). GaCl3 and AlCl3 also form eutectic mixtures,[36,38] but AlCl3 does not affect the extraction of GaCl3, indicating that the eutectic does not suppress GaCl3 extraction. Similarly, formation of a simple eutectic between GaCl3 and InCl3 cannot explain the reduced GaCl3 extraction, and formation of an ionic compound might be a more reasonable explanation. To validate the Lewis adduct formation, the same extraction experiments were conducted using anhydrous n-hexane, and similar Raman spectra were observed for the extraction residues (Figure S1).

Figure 2

Raman spectra of the pure chloride compounds and of the different separation residues after solid–liquid extraction of GaCl3 by anhydrous n-dodecane.

Because the ionic compound [InCl2][GaCl4] has not been reported yet, more studies are required to confirm the formation of this species. Besides the formation of the ionic compound [InCl2][GaCl4], the mixed dimer InGaCl6 may be formed in the mixture of InCl3 and GaCl3, which has been reported in the vapor phase of the InCl3–GaCl3 system by Buraya et al.[37] We optimized structures of the relevant compounds in this study by DFT calculations (Figure ) and generated their Raman spectra, which are used to discuss the two possible compounds. Pure GaCl3 is a dimer (Ga2Cl6),[39] and its Raman spectrum has been reported by various researchers.[40,41] There is a good match between the calculated Raman spectrum of Ga2Cl6 and the corresponding experimental spectrum (Figure S2), although the shifts for the bands at 244 and 329 cm–1 are relatively large. The good match indicates that the DFT method used in this study is suitable for calculating this compound. The computed spectrum has an overall red shift, which is consistent with the observation of Timoshkin et al.,[42] who found that at the B3LYP level of theory, the low frequency vibrations (<500 cm–1) are underestimated and the high frequency vibrations are overestimated by DFT computation. Despite the fact that there are available coefficients for scaling the frequency of this level of theory (e.g., coefficients given by Bao et al.[43]), these coefficients are suitable for high frequency vibrations but not suitable for low frequency vibrations, which is the case in this study. Therefore, we followed the treatment in the study of Timoshkin et al.,[42] to regress scaling coefficients using the following equation, based on the experimental spectrum of Ga2Cl6.where wExp and wDFT are the experimental and computed wavenumber, respectively. The obtained coefficients are A = 0.9878 and B = 14.66 cm–1, and the correlation coefficient is 0.996. After scaling (Figure a), the computed spectrum matches better with the experimental spectrum. The computed spectra of InGaCl6, [GaCl4]−, and [InCl2]+ were also scaled using the same coefficients (Figure ).

Figure 3

DFT optimized structures of the compounds Ga2Cl6, InGaCl6, [GaCl4]−, and [InCl2]+.

Figure 4

Raman spectra: (a) experimental and DFT calculated spectra of Ga2Cl6 and comparison of the two spectra; (b) DFT calculated spectrum of InGaCl6; (c) DFT calculated spectrum of [GaCl4]− and its comparison with literature data of Shamir and Rafaeloff,[32,33] and Gerding and Koningstein;[45] (d) DFT calculated spectrum of [InCl2]+.

InGaCl6 exhibits a similar structure as Ga2Cl6 (Figure ), and its Raman spectrum resembles that of Ga2Cl6 except for an extra band at 361 cm–1 (Figure b), which is mainly due to the symmetric stretching of In−μ1−Cl. Since the Raman intensity at 403 cm–1 for InGaCl6, which corresponds to the band at 404 cm–1 for the experimental Ga2Cl6 spectrum, is high, a band at around 404 cm–1 should be observable if InGaCl6 were formed. The absence of the band in Figure b indicates that InGaCl6 was not formed. In addition, the structures of the mixed dimer of InGaCl6 and the dimer of Ga2Cl6 are so similar that they are expected to have similar solubility properties; that is, InGaCl6 should be soluble in anhydrous n-dodecane since Ga2Cl6 is highly soluble in anhydrous n-dodecane. However, InCl3 suppressed the extraction of GaCl3, meaning that a compound that is insoluble in anhydrous n-dodecane was formed.

The calculated [GaCl4]− exhibits a tetrahedral structure, and the Raman spectrum shows four bands at 115 cm–1, 158 cm–1, 336 cm–1, and 373 cm–1, corresponding to experimental observations at around 128 cm–1, 153 cm–1, 346 cm–1, and 390 cm–1, respectively. The calculated [InCl2]+ exhibits a linear structure, and the Raman spectrum shows only one band at 374 cm–1 due to the symmetric stretching of In–Cl, which matches well with the reported band at 367 cm–1.[34] Kloo and Taylor reported spectroscopic data of solid [InCl2(15-crown-5)][InCl4] but did not report the exact structure.[34] In a similar study, Kloo and Taylor reported that the I–In–I angle is 170.1° in the solid compound of [InI2(18-crown-6)][InI4].[44] In this compound, the [InI2]+ structure is almost linear. Besides, the coordination number of indium in this compound was estimated to be less than four based on the relatively short bond length of In–I. Considering the similarity of [InCl2]+ and [InI2]+, the computed linear [InCl2]+ is reasonable. Therefore, the bands at 362 and 366 cm–1 in Figure b and Figure S1-b, respectively, can be assigned to [InCl2]+, indicating the formation of [InCl2][GaCl4]. Formation of the ionic compound [InCl2][GaCl4] also explains why the extraction of GaCl3 was suppressed by InCl3, because ionic compounds are insoluble in hydrocarbons. In summary, formation of [InCl2][GaCl4] in the InCl3/GaCl3 mixture after extraction is very likely. However, formation of [InCl2][GaCl4] contradicts with the formation of an eutectic between GaCl3 and InCl3. Further studies are needed to confirm the speciation of GaCl3 and InCl3 mixtures.

Stripping of GaCl3

With the knowledge that the chloride is transferred and the resulting formation of ionic complexes can inhibit the dissolution of GaCl3, advantage of it was taken to strip GaCl3 from the loaded n-dodecane solutions using alkali metal chlorides. LiCl, NaCl, and KCl that were ground and sieved with a 250 μm sieve mesh were tested for the stripping of GaCl3 for various MCl/GaCl3 (M = Li, Na, and K) molar ratios (Figure ). The efficiency of stripping by LiCl, NaCl, and KCl is comparable, although NaCl is slightly more efficient when the MCl/GaCl3 ratio is less than 10. About 96% of GaCl3 can be stripped from the loaded n-dodecane phase when the MCl/GaCl3 is 20, and the stripping is almost complete when the ratio is 30. Raman spectra of pure MCl (M = Li, Na, and K) and the solid residues after stripping of GaCl3 by MCl were recorded, and the species of [GaCl4]− was clearly identified (Figure S3). It is therefore evident that the alkali chlorides interact with the dissolved GaCl3 by forming the ionic compounds of LiGaCl4, NaGaCl4, and KGaCl4, which are insoluble in the aliphatic hydrocarbon, effectively inhibiting GaCl3 to be dissolved. Moreover, the formation of KGa2Cl7 and KGa3Cl10 is also possible, because these species have been observed as well.[46,47] NaCl is the most suitable salt for stripping GaCl3 because it is easily available, environmentally benign, and cheap. If NaCl is used for stripping GaCl3, the resultant NaGaCl4 salt can be further processed to recover gallium metal. For instance, it is known that gallium can be electrodeposited from alkaline solutions of GaCl3, so that NaGaCl4 could be dissolved in an aqueous NaOH solution to prepare an electrolyte for electrodeposition of gallium metal.[48] The n-dodecane after stripping can be recycled and reused for extraction of GaCl3.

Figure 5

Stripping of GaCl3 from the loaded n-dodecane solution by alkali metal chlorides.

Conclusions

A preliminary solvometallurgical process using only anhydrous n-dodecane and NaCl has been developed for the separation of GaCl3 and AlCl3, based on the exceptionally high solubility of GaCl3 in hydrocarbon solvents and on its Lewis acidity. GaCl3 can be efficiently extracted by anhydrous n-dodecane from binary solid mixtures of GaCl3/AlCl3 without formation of ionic compounds by chloride transfer thanks to the similarity of the Lewis acidity of these compounds. The loaded GaCl3 in n-dodecane can be stripped by NaCl by forming NaGaCl4. This solvometallurgical process is essentially different from, and can be supplementary to, the hydrometallurgical processes for GaCl3 and AlCl3 separations.