Supercritical CO₂ extraction of organic carbonate-based electrolytes of lithium-ion batteries

Abstract

Supercritical fluid extraction (SFE) was applied to reclaim organic carbonate-based electrolytes of spent lithium-ion batteries. To optimize the SFE operational conditions, the response surface methodology was adopted. The parameters studied were as follow: pressure, ranging from 15 to 35 MPa; temperature, between 40 °C and 50 °C and static extraction time, within 45 to 75 min. The optimal conditions for extraction yield were 23 MPa, 40 °C and was dynamically extracted for 45 min. Extracts were collected at a constant flow rate of 4.0 L min⁻¹. Under these conditions, the extraction yield was 85.07 ± 0.36%, which matched with the predicted value. Furthermore, the components of the extracts were systematically characterized and analyzed by using FT-IR, GC-MS and ICP-OES, and the effect of SFE on the electrolyte reclamation was evaluated. The results suggest that the SFE is an effective method for recovery of organic carbonate-based electrolytes from spent lithium-ion batteries, to prevent environmental pollution and resource waste.

---

Introduction

Lithium-ion batteries (LIBs) are widely used as electrochemical power sources in consumer electronics, electric vehicles and other modern-life appliances. LIBs will probably be sent to recycling facilities at the end of life, which specializes in the specific battery type. As is known, the spent LIBs contain considerable valuable chemical substances; besides cathode active material, they also comprise copper and aluminium foil (anode and cathode current collect) and electrolyte. Besides the expensive cathode materials, the electrolyte is also the most valuable component in LIBs.¹ Many recycling methods for the spent LIBs have been reported.^2–5 Mostly their concern was valuable metals, while the remainder of the battery including the electrolyte was deemed worthless and disposed in any way possible to get rid them.

In addition to the profit motive, there is the need for preventing the pollution caused by the hydrolysis of conductive salt and also the toxic electrolyte mixture that virtually corrupts the earth and water for any use along with the danger to animal and insect life as well as human life. The electrolytes in present the LIBs are mixtures, which contain aprotic solvents in addition to a conductive salt. The most frequently used solvents are propylene carbonate, ethylene carbonate, diethyl carbonate and dimethyl carbonate.^6–8 Although an entire series of conductive salts is being discussed, LiPF₆ is till date the mostly used one.⁹ When exposed to water or moist air, LiPF₆ can readily hydrolyze and produce toxic hydrogen fluoride gasses. When connected with moisture, including skin tissue, hydrogen fluoride gasses immediately convert to hydrofluoric acid, which is highly corrosive to battery reclaiming facilities and toxic to operators. Evidently, it is necessary to separate or remove of spent LIBs electrolytes in a safe manner before the dismantling of the LIBs by reclaiming facilities, which can prevent above mentioned pollution and hazards.

Several electrolyte separation and extraction techniques have been employed in the recycling process of the spent LIBs. Lian immersed mechanical-shredded LIBs into a suitable solvent for several hours, the electrolytes were extracted. The solvents were recovered by evaporation using the process of pressure reducing, and the pure electrolytes were left eventually.¹⁰ The limitations are that the solvent boiling point at reduced pressure must be below the lithium salt decomposition temperature (≤80 °C), and the materials are available in an anhydrous state. Similarly, Schmidt et al. developed suitable solvents, such as 1,2-dimethoxyethane, dimethyl carbonate, ethyl acetate and acetone, to extract the organic electrolyte solvents such as polyvinylidene fluoride and other binders, and the dissolved lithium hexafluorophosphate. The solvents used for the extraction can be recovered by reduced pressure distillation.¹¹ For Sun et al., electrolytes of spent LIBs were separated from LIBs using a vacuum pyrolysis process in a pyrolysis system at the following conditions: temperature of 600 °C, vacuum evaporation time of 30 min, and residual gas pressure of 1.0 kPa. The components of the pyrolysis products were analyzed by FT-IR, which indicated that the main components are fluorocarbon organic compounds.¹² Most of the fluorinated compounds can be enriched and recovered to prevent environmental pollution and resource waste. The organic solvent extraction process always introduced a solvent impurity, which not only complicated the separation process but also brought new pollutants. In the vacuum pyrolysis process, electrolytes are thoroughly decomposed in a vacuum pyrolysis process, but the components of the decomposition products are too complicated to reuse.

SFE is a separation technology using the relationship of density to dissolving capacity of supercritical fluid. SFE offers extraction yield comparable with those obtained by conventional extraction methods of organic solvents. In the supercritical fluid systems, CO₂ has a moderate critical pressure (73.8 atm) and a low critical temperature (31.1 °C) as compared to the others.¹³ Under the supercritical state, CO₂ has a large dissolving capacity of low and non-polar substances.¹⁴ The CO₂ is easily separated from the extract by its high volatility and can be completely recycled and reused without hazardous solvent wastes emission. A number of experiments have been carried out to eliminate toxic materials from waste by supercritical CO₂ extraction.^15,16 Therefore, this method has become a more viable option in the separation industry and the environmental protection industry. Leaching of the spent LIBs by supercritical CO₂ and several other supercritical fluids for removal of the electrolytes were illustrated in a patent by Sloop.¹⁷ The batteries were placed in an extraction vessel, which was full of fluids. Temperature and pressure of the fluids in the extraction vessel were adjusted to achieve the supercritical state. The electrolytes were exposed to and extracted by the fluids at the supercritical state. All supercritical fluids were then transferred to a collection vessel, in which the temperature and the pressure of the fluids reverted to the original state, and the electrolytes eventually left. There is slight information available in the patent about optimization of the extraction process and the component analysis of the extraction product.

In this paper, the organic carbonate-based electrolytes were separated using supercritical CO₂ from the LIBs separator, simulating electrolyte extraction from a spent lithium-ion battery. The extraction parameters were optimized with the response surface methodology (RSM) in order to obtain a considerable extraction yield in an economical operation range. Furthermore, the effect of SFE on the electrolyte reclamation was evaluated from the aspect of consistency and integrity of the electrolyte components.

Experimental

Reagent and materials

The TC-E201# electrolyte mainly composed of 1 M LiPF₆ in EC/DMC/EMC (1 : 1 : 1 vol%), which was provided by Tinci Materials Technology Co., Ltd. (Guangzhou, China). Commercial grade CO₂, supplied by Liming Gas Co., Ltd. (Harbin, China), was used with the purity of more than 99.95%. The JH ordinary type polypropylene separator was purchased from Jinhui Hi-tech Optoelectronic Material Co., Ltd. (Foshan, China), which used as adsorbent of electrolytes.

Extraction procedures

For each extraction experiment, electrolytes were adsorbed in a lithium-ion battery separator, and enclosed into the extraction vessel in an argon-filled glove box with moisture and oxygen level less than 1 ppm. The extraction vessel was transferred to a Spe-ed SCF Prime supercritical CO₂ extraction system (Applied Separations, Inc., Allentown, PA, USA) for electrolyte extraction. A schematic diagram of the apparatus is shown in Fig. 1. To study the influence of pressure, temperature and time on the extraction efficiency, a series of experiments was designed to conduct under a pressure of 15 to 35 MPa, a temperature of 40 to 50 °C and a static extraction time of 45 to 75 min. In all experiments, the designed temperature was lower or equal to 70 °C, which is the temperature employed in LiPF₆, because thermal degradation could take place at higher temperatures.¹⁸ The extracts were then collected into a sample vial at a constant flow rate of 4.0 L min⁻¹. The collected sample was tightly sealed and stored in the glove box before analysis. The extraction yield was calculated according to the following equation:

Y% = (m_ads − m_res)/(m_ads − m_sep) × 100 (1)

where Y is the extraction yield, m_sep is the mass of the separator, m_ads is the mass of the separator after adsorbing electrolytes and m_res is the mass of the residues after extraction.

Fig. 1 Schematic diagram of supercritical CO₂ extraction apparatus: (1) CO₂ cylinder, (2) cooling bath, (3) air driven fluid pump (gas booster pump), (4) air compressor, (5) air regulator, (6) CO₂ pressure, (7) inlet valve, (8) extraction vessel, (9) heating jacket, (10) vessel heat, (11) vent valve, (12) outlet valve, (13) flow valve, (14) valve heat, (15) heating jacket, (16) collecting vial, (17) alumina filter, (18) gas flow meter.

Experimental design

In this research, response surface methodology with Box–Behnken^19,20 design was applied to optimize the extraction conditions, which obtained the maximum extraction yield from lithium-ion battery separator in SC-CO₂ medium. The variables studied were pressure (MPa, X₁), temperature (°C, X₂) and extraction time (min, X₃), and each variable set at three levels. A total of 15 experiments were designed in Table 1, including the triplicate runs for the center points (Runs 3, 6 and 14). The center points provided an internal estimate of pure error used to test for lack of fit and also contributed toward estimation of the squared terms. All the experiments were done in triplicates and the average extraction yield (%) was taken as the response Y. The experimental data obtained were fitted to a second order polynomial equation. The equation, coefficient of determination, analysis of variance (ANOVA), surface plot and conditions for maximum extraction yield were obtained using Design-Expert 8 (Stat-Ease, Inc., Minneapolis, MN, USA).

Table 1 Box–Behnken design and observed responses^a

Run	Independent variable			Response (Y%)
	X1 (press, MPa)	X2 (temperature, °C)	X3 (time, min)
a Average of triplicate experiments.
1	15(−1)	50(+1)	60(0)	83.98
2	25(0)	40(−1)	75(+1)	86.71
3	25(0)	45(0)	60(0)	87.96
4	15(−1)	45(0)	75(+1)	84.17
5	25(0)	40(−1)	45(−1)	85.69
6	25(0)	45(0)	60(0)	87.53
7	25(0)	50(+1)	75(+1)	88.98
8	15(−1)	40(−1)	60(0)	82.24
9	35(+1)	40(−1)	60(0)	86.84
10	35(+1)	50(+1)	60(0)	88.26
11	35(+1)	45(0)	45(−1)	86.13
12	35(+1)	45(0)	75(+1)	87.72
13	15(−1)	45(0)	45(−1)	81.66
14	25(0)	45(0)	60(0)	87.55
15	25(0)	50(+1)	45(−1)	86.74

---

Analytical methodology

IR spectra were collected on a PerkinElmer Spectrum One FT-IR spectrometer (Waltham, MA, USA) equipped with a KBr crystal in the absorbance mode range from 400 to 4000 cm⁻¹ with a resolution of 4 cm⁻¹.

The extracts were analyzed using an Agilent (Agilent Technologies, Palo Alto, CA, USA) GC-MS system (GC 6890N, MS 5973N) with a DB-5 ms capillary column (30 m × 0.25 mm i.d., 1.0 μm film thickness; J&W Scientific, Folsom, CA, USA). The GC operating conditions were as follows: column temperature 60 °C maintained for 3 min, then increased to 300 °C at a rate of 10 °C min⁻¹ and finally sustained at 300 °C for 2 min; carrier gas helium at a flow rate 1.0 mL min⁻¹; injector temperature 280 °C; injected volume 2 μL; splitless. The temperature of the transfer line was 260 °C. The MS operating conditions were: ionization voltage 70 eV; ion source temperature 230 °C; mass range 15–750 amu.

The concentration of LiPF₆ in the electrolyte solution and the extract was obtained by measuring the Li⁺ concentration in the back-extraction solution. The LiPF₆ in organic liquid was extracted back to the water with the pH < 1, and the concentration of lithium ions in water was accurately analyzed by inductively coupled plasma optical emission spectrometry (ICP-OES, PerkinElmer Optima 5300DV, PerkinElmer, Waltham, Ma, USA).

¹⁹F NMR (376.4 MHz, acetone-d₆, 25 °C) and ³¹P NMR (161.9 MHz, acetone-d₆, 25 °C) spectra were recorded on an Avance III 400 MHz digital NMR spectrometer using either 5 mm glass tubes (Wilmad Glass Co., Buena, NJ, USA).

Results and discussion

Optimization of the experimental conditions

The supercritical CO₂ extraction of organic carbonate-based electrolytes from the lithium-ion battery separators were optimized by varying the operating parameters according to the Box–Behnken design. The number of experiments required to investigate the above mentioned three parameters at the three levels would be 27 (3³ factorial). Thus, this was reduced to 15 using a Box–Behnken experimental design. The results of this limited number of experiments provided a statistical model that was used to identify trends of high yield for the extraction process. Table 1 presents the experiment design and corresponding response yield data for SFE. The analysis of variance (ANOVA) for the experimental results of the Box–Behnken design is shown in Table 2. The fit of the model can be checked by the determination coefficient (R²), which was 0.9944, indicating that the model adequately represented the real relationship between the chosen parameters. The lack-of-fit measures the failure of the model to represent data in the experimental domain at points that are not included in the regression. The non-significant value of lack-of-fit (p > 0.05) revealed that the model equation was adequate for predicting the yield under any combination of values of the variables. eqn (2) illustrates the relationship of the three variables and Y.

Y = +87.61 + 2.11X₁ + 0.81X₂ + 0.92X₃ − 0.23X₁X₃ + 0.31X₂X₃ − 2.22X₁² − 0.52X₃² (2)

where Y is the extraction yield, X₁ is the pressure, X₂ is the temperature and X₃ is the time of extraction.

Table 2 Analysis of variance (ANOVA) for the experimental results

Source	Sum of squares	df	Mean squares	F value	p-value prob. > F
Model	67.11	7	9.59	176.42	<0.0001
X1	35.70	1	35.70	656.93	<0.0001
X2	5.25	1	5.25	96.58	<0.0001
X3	6.77	1	6.77	124.60	<0.0001
X1X3	0.21	1	0.21	3.89	0.0891
X2X3	0.37	1	0.37	6.85	0.0346
X1^2	18.32	1	18.32	337.07	<0.0001
X3^2	1.01	1	1.01	18.54	0.0035
Residual	0.38	7	0.054
Lack of fit	0.26	5	0.053	0.89	0.6040
Pure error	0.12	2	0.059
Cor total	67.49	14
Adjust-R^2	0.9887
R^2	0.9944

---

Eqn (2) shows that the extraction yield depends more on pressure variations than extraction time variations. Dependence of yield on temperature is least. In order to get a better understanding of the influences of the independent variables and their interactions on the dependent variables, three-dimensional (3D) response surface plots for the measured responses were constructed according to eqn (2). Since the regression model has three independent variables, one variable was held constant at the central level for each plot. Fig. 2 shows the 3D response surfaces as the function of two variables at the centre level of other variables, respectively.

Fig. 2 Response surfaces and contour plots for: (a) extraction time vs. pressure; (b) extraction time vs. temperature.

It is generally believed that the solubility of a solute in the SCF tends to increase with the density of the fluid (at constant temperature). In the present study, the influence of pressure on the composition of electrolytes displayed that, in a definite extraction time, the extraction yield was increased drastically with the pressure increasing. This result was predictable because raising the extraction pressure resulted in a higher fluid density, which can improve the solubility of the electrolyte composition. The variation of temperature during the SFE affects the fluid density and the volatility of the electrolytes from the lithium-ion battery separator. By increasing the temperature, the volatilities of the electrolyte composition keeps an upward tendency but the SCF density decreases. In the temperature range of this study, raising temperature steadily increases the extraction yield of electrolytes due to the enhanced volatilities of the electrolyte composition.

In practice, getting the required output with minimum input is the most economical production mode. Based on the polynomial regression model, the mildest experimental conditions of considerable extraction yield were found to be at 23.4 MPa, 40 °C and 45 min. Under these conditions, the predicted extraction yield was 85.22%. On the basis of these results, a set of verification experiments (3 replicates) were carried out at 23 MPa, 40 °C and 45 min, when the average extraction yield was 85.07 ± 0.36%. This result indicated that the experimental values were in good agreement with the predicted values, and also suggested that the model was satisfactory and accurate.

Composition analysis

Organic solvent. The appearance of the extracts collected by the sample vial is a colourless liquid with a mild odour. The FT-IR spectrum (Fig. 3) of the extracts shows peaks around 3421 cm⁻¹, which were attributed to ν_O–H of the intermolecular hydrogen bonds, broad ν_C–H peaks in the range 3163–2963 cm⁻¹, typical peaks around 1805–1776 cm⁻¹ were attributed to ν_C=O of organic carbonate, typical peaks around 1597 and 1553 cm⁻¹ were attributed to skeletal C=C vibrations, peaks around 1482–1163 cm⁻¹ were attributed to δ_C–H of CH₂ and CH₃ groups, peaks around 1075 and 1010 cm⁻¹ were attributed to ν_C–O and peaks around 972–737 cm⁻¹ were attributed to γ_C–H of organic species.²¹ In addition to the above characteristic peaks of organic species, there is also a pronounced peak at 846 cm⁻¹, which should be attributed to ν_P–F bands,²² in the FT-IR spectrum of the electrolytes. The FT-IR analysis indicated that the main components of extracts are organic carbonate. The FT-IR spectrum of the extracts is not entirely consistent with the electrolytes', which means a change in the structures of some components. Therefore, further research is needed to identify each composition of the extracts.

Fig. 3 Comparison between the FT-IR spectrum of electrolyte (top) and extract (bottom).

The volatile components produced in the extraction process of this electrolyte was analyzed by GC-MS and structurally assigned through matching to the National Institutes of Standards (NIST) library. The gas chromatograms of electrolyte components before and after SFE were compared and analyzed to confirm the consistency and integrity of the electrolyte components. Fig. 4 shows that electrolytes and extracts have a high degree of consistency in the gas chromatographic retention times. There is no significant difference between electrolyte and extracts in the components. The extracts in order of volatility (early to late retention times) are characterized as dimethyl carbonate, ethyl methyl carbonate, vinylene carbonate, ethylene carbonate, 1,11-biphenyl, which are listed in Table 3.

Fig. 4 Comparison between the gas chromatogram of electrolyte (top) and extract (bottom).

Table 3 Components analysis of electrolyte extracts by mass spectrometry in order of volatility marked in Fig. 4

No.	Components	Retention times/min	Molecular weight	Molecular ion peak
1	Dimethyl carbonate	3.087	90	90
2	Ethyl methyl carbonate	4.123	103	103
3	Vinylene carbonate	5.282	86	86
4	Ethylene carbonate	10.125	88	88
5	1,11-biphenyl	16.736	154	154

---

Conductive salt. In order to determine the content of LiPF₆ in the extracts, the remaining Li⁺ concentrations were analyzed using ICP-OES. The results show that the Li⁺ concentration in the electrolyte solution is 0.9038 mol L⁻¹, but only 0.0636 mol L⁻¹ in the extract. Comparing the test results of the two Li⁺ concentrations, it turns out that LiPF₆ was decomposed during the supercritical CO₂ extraction process.

To further prove that the LiPF₆ was hydrolyzed, the electrolyte, soak solution of the separator with DMC after SFE and extracts were analyzed by nuclear magnetic resonance spectroscopy. The ¹⁹F and ³¹P NMR spectra and peak assignments for the above mentioned samples are depicted in Fig. 5 and 6, respectively. In the ¹⁹F NMR spectrum of electrolytes (Fig. 5a), the doublet with chemical shift −72.8 ppm are assigned to PF₆⁻, another doublet observed at −84.3 ppm are ascribed to PO₂F₂⁻, which is the product of LiPF₆ hydrolysis. In the case of soak solutions of the separator with DMC after SFE (Fig. 5b), two doublets at −75.7 ppm and −85.1 ppm are attributed to PO₃F²⁻ and PO₂F₂⁻, respectively. A singlet of F⁻ with chemical shift −188.2 ppm was also observed. Furthermore, in the spectrum of extract (Fig. 5c), the same singlet can be found at −187.8 ppm. In the ³¹P NMR spectrum of electrolytes (Fig. 6a), the septet at −144.3 ppm is ascribed to the PF₆⁻. Moreover, for the soak solutions of separator with DMC after SFE (Fig. 6b), a singlet and a doublet are observed at 1.3 and −7.4 ppm, which can be assigned to the H₃PO₄ and PO₃F²⁻, respectively. However, the intensities of the signals are too low to be clearly seen in the spectrum of the extract (Fig. 6c). Therefore, only PO₂F₂⁻, PO₃F²⁻ and HF are detected clearly, other products of hydrolysis cannot be clearly observed. These spectra are similar to the NMR measurements of hydrolysis in propylene carbonate–dimethyl carbonate–H₂O reported by other authors.²³

Fig. 5 ¹⁹F NMR spectrum of electrolyte, separator and extract in acetone-d₆ (376.4 MHz, 25 °C).

Fig. 6 ³¹P NMR spectrum of electrolyte, separator and extract in acetone-d₆ (376.4 MHz, 25 °C).

In general, LiPF₆ is electrolytic dissociative in organic solvents in the equation:

LiPF₆ ⇌ Li⁺ + PF₆⁻ (3)

However, part of non-electrolytic dissociative LiPF₆ is instable, and gets decomposed into LiF and PF₅:²⁴

LiPF₆ ⇌ LiF + PF₅ (4)

PF₅ is a strong Lewis acid,²⁵ and hydrolyzed by trace water in the impurity CO₂ according to following equation:^26,27

PF₅ + H₂O ⇌ POF₃ + 2HF↑ (5)

POF₃ + 3H₂O ⇌ H₃PO₄ + 3HF↑ (6)

As intermediates, PO₂F₂⁻ and PO₃F²⁻ are synthesized in the transformation course of POF₃:²³

POF₃ + H₂O ⇌ PO₂F₂⁻ + HF + H⁺ (7)

PO₂F₂⁻ + H₂O ⇌ PO₃F²⁻ + HF + H⁺ (8)

The above analysis and results of the NMR spectra suggest that LiPF₆ was hydrolyzed in the extraction process. Some products of the hydrolysis of LiPF₆ were absorbed on the separator, especially most of the phosphorus-containing products of the hydrolysis stayed in the separator. HF, another product of the hydrolysis of LiPF₆, will cause damage to both human health and equipment. Therefore, exhaust gas should be treated with appropriate method before emitting into the atmosphere or entering into the CO₂ circulation system. In this study, the HF was absorbed by filtering tube filled with alumina.

Conclusions

The supercritical CO₂ extraction is an efficient and environment-friendly electrolyte separation method for recycling LIBs. The extraction yield of electrolytes from the LIBs separator can achieve 85.07 ± 0.36% using the mildest operating conditions of 23 MPa, 40 °C and 45 min. This result matched with the predicted values and confirmed that the response model is adequate to reflect the expected optimization. The results of the experiment showed that the extraction pressure is the major contributing factor of the electrolyte extraction. Moreover, the results of the component analysis reveal that the contents of organic solvents in the electrolyte basically remained unchanged in the supercritical CO₂ extraction process. The electrolyte is a mixture, and the best choice for the purification and reuse is that the mixture can be selectively extracted into individual component by using the supercritical CO₂. This is the next goal of electrolyte recycling in further research.

Acknowledgements

This research was supported by the National Natural Science Foundation of China (no. 51274075), the National Environmental Technology Special Project (no. 201009028), and Guangdong Province-department University-industry Collaboration Project (grant no. 2012B091100315). Yuanlong Liu thanks Dr Zhaohui Wen, at Department of Neuro Intern, First Affiliated Hospital of Harbin Medical University, Harbin, China, for her constant help and suggestions.

Notes and references

1. M. J. Lain, J. Power Sources, 2001, 97–98, 736–738.
2. X. Zeng, J. Li and N. Singh, Crit. Rev. Environ. Sci. Technol., 2013, 44, 1129–1165.
3. M. Vanitha and N. Balasubramanian, Environ. Technol. Rev., 2013, 2, 101–115.
4. O. E. Bankole, C. Gong and L. Lei, J. Environ. Ecol., 2013, 4, 14–28.
5. S. Al-Thyabat, T. Nakamura, E. Shibata and A. Iizuka, Miner. Eng., 2013, 45, 4–17.
6. D. Aurbach, A. Zaban, Y. Gofer, Y. E. Ely, I. Weissman, O. Chusid and O. Abramson, J. Power Sources, 1995, 54, 76–84.
7. G. E. Blomgren, J. Power Sources, 1999, 81–82, 112–118.
8. G. E. Blomgren, J. Power Sources, 2003, 119–121, 326–329.
9. K. Hayashi, Y. Nemoto, S. Tobishima and J. Yamaki, Electrochim. Acta, 1999, 44, 2337–2344.
10. M. J. Lian, US Pat., US6447669 B1, 2002.
11. M. Schmidt, R. P. Hemmer, M. Wohlfahrt-Mehrens, G. Arnold and C. Vogler, US Pat., US6511639 B1, 2003.
12. L. Sun and K. Qiu, J. Hazard. Mater., 2011, 194, 378–384.
13. E. Reverchon and I. De Marco, J. Supercrit. Fluids, 2006, 38, 146–166.
14. S. Wang, Y. Ling and Y. Giang, Forensic Science Journal, 2003, 2, 5–18.
15. H. Fu and M. A. Matthews, J. Hazard. Mater., 1999, 67, 197–213.
16. V. Jain, Environ. Sci. Technol., 1993, 27, 806–808.
17. S. E. Sloop, US Pat., US7858216 B2, 2010.
18. E. Zinigrad, L. Larush-Asraf, J. S. Gnanaraj, M. Sprecher and D. Aurbach, Thermochim. Acta, 2005, 438, 184–191.
19. G. E. P. Box and D. W. Behnken, Technometrics, 1960, 2, 455–475.
20. G. E. P. Box and K. B. Wilson, On the Experimental Attainment of Optimum Conditions, ed. S. Kotz and N. Johnson, Springer, New York, 1992, pp. 270–310.
21. B. Zhang, Y. Zhou, X. Li, X. Ren, H. Nian, Y. Shen and Q. Yun, Spectrochim. Acta, Part A, 2014, 122, 59–64.
22. D. Aurbach, B. Markovsky, A. Shechter, Y. Ein Eli and H. Cohen, J. Electrochem. Soc., 1996, 143, 3809–3820.
23. A. V. Plakhotnyk, L. Ernst and R. Schmutzler, J. Fluorine Chem., 2005, 126, 27–31.
24. L. Terborg, S. Weber, F. Blaske, S. Passerini, M. Winter, U. Karst and S. Nowak, J. Power Sources, 2013, 242, 832–837.
25. K. O. Christe, D. A. Dixon, D. McLemore, W. W. Wilson, J. A. Sheehy and J. A. Boatz, J. Fluorine Chem., 2000, 101, 151–153.
26. D. Aurbach, A. Zaban, Y. Ein-Eli, I. Weissman, O. Chusid, B. Markovsky, M. Levi, E. Levi, A. Schechter and E. Granot, J. Power Sources, 1997, 68, 91–98.
27. T. Kawamura, S. Okada and J. Yamaki, J. Power Sources, 2006, 156, 547–554.

---