Li ion/vapor grown carbon fiber polymer actuators show higher performance than single-walled carbon nanotube polymer actuators

Abstract

The effects of Li tetrafluoroborate (Li[BF₄]) and lithium bis(trifluoromethanesulfonyl)imide (Li[TFSI]) salts on the electrochemical and electromechanical properties of an actuator using a poly(vinylidene fluoride-co-hexafluoropropylene)-supported vapor grown carbon fiber (VGCF)/ionic liquid (IL) gel electrode formed without ultrasonication were investigated. At a slow sweep rate of 1 mV s⁻¹, the measured double-layer capacitance values for the VGCF/Li[X]/IL actuator were in the range of 20.0–42.3 F g⁻¹, and were larger than that for an IL-only actuator. The capacitance was found to increase with the Li[BF₄] content, but was independent of the Li[TFSI] content. All of the VGCF/Li[X]/IL actuators exhibited a larger amount of strain than IL-only actuators, and for a VGCF/Li[TFSI]/EMI[TFSI] actuator with a Li[TFSI]/EMI[TFSI] molar ratio of 1.0, the maximum strain was greater than that for an IL actuator containing single-walled carbon nanotubes. Moreover, the frequency dependence of the displacement response for the VGCF/Li[X]/IL actuators was successfully simulated using an electrochemical kinetic model, similar to the case for SWCNT and VGCF based actuators containing metal oxide. The results yielded the strain in the low-frequency limit in addition to the time constant of the response.

---

Introduction

Recently, much attention has been focused on soft materials that can directly transform electrical energy into mechanical work for a wide range of applications including robotics, tactile and optical displays, prosthetic devices, medical devices, and microelectromechanical systems.¹ Low-voltage electroactive polymer (EAP) actuators that can respond quickly and are softly driven are particularly useful, because they can be used as artificial muscle-like actuators for various biomedical and human related applications.^2,3 We have previously reported^4–6 a dry actuator that can be simply fabricated by layer-by-layer casting, using ‘bucky-gel’,⁷ a gelatinous room-temperature ionic liquid (IL) containing single-walled carbon nanotubes (SWCNTs). The actuator has a bimorph configuration with a polymer-supported IL electrolyte layer sandwiched between polymer-supported bucky-gel electrode layers that allow quick and long-lived operation in air at low applied voltages. ILs have low volatility and exhibit high ionic conductivities and wide potential windows, which are advantageous for quick-response actuators and high electrochemical stability components.⁸

We previously reported the dependence of the electromechanical and electrochemical properties of actuators composed of polymer-supported bucky-gel electrodes and a gel electrolyte layer on the type of IL, nanocarbon and the polymer used.^6,9–12 Furthermore, we have recently reported that polymer actuators containing activated multi-walled carbon nanotubes (MWCNTs), or a combination of non-activated MWCNTs and a metal oxide, surpassed the performance of SWCNT-based actuators in terms of the strain and maximum stress generated.^13–15

Whereas SWCNTs require special preparation techniques and are very expensive, MWCNTs are very inexpensive and are commonly used in battery electrodes. Much attention has been focused on activated (acid treated) CNTs for electric double-layer capacitors (EDLCs), which have higher electrochemical capacitance than those based on non-activated CNTs.¹⁶ CNTs have been recognized as potential electrode materials for supercapacitors due to their unique properties (such as their mesoporous character, good chemical stability and conductivity and nanometer dimensions).¹⁷

A vapor grown carbon fiber (VGCF) is specifically designed to enhance the electrical and thermal properties of high performance materials. A disadvantage of SWCNT actuators is their poor dispersibility, while VGCF exhibits good dispersibility, high electrical conductivity and high mechanical strength. However, the main disadvantage of VGCF as an actuator material is its low capacitance.¹¹ Recently, we have reported that polymer actuators containing VGCF and metal oxide, formed without using ultrasonication, surpassed the performance of a SWCNT-containing actuator in terms of the amount of strain produced.¹⁸

Rechargeable Li-ion batteries are ubiquitous energy devices that are used worldwide in many types of portable electronic equipment. In state-of-the-art 4 V class Li-ion batteries, a mixture of organic aprotic solvents and the conducting salt lithium hexafluorophosphate (LiPF₆) is generally used as a non-aqueous electrolyte. There is also considerable interest in solid polymer gel electrolytes containing Li salts.¹⁹ We focus our attention on electrodes and electrolytes based on Li salts and ILs for use in high-energy-density devices such as EAP actuators and electrochemical capacitors. It is expected that the addition of a Li salt to a polymer-supported bucky-gel electrode would increase its double-layer capacitance. In an actuator based on such electrodes, this would lead to a larger amount of strain being generated, and the higher ionic conductivity would produce a faster response. We have already reported that the addition of a Li salt to a SWCNT-based polymer actuator containing an IL led to a substantial improvement in its performance.²⁰

In a previous study,⁶ we investigated the voltage–current and voltage–displacement characteristics of a bucky-gel actuator by applying a triangular voltage waveform at various frequencies. In order to quantitatively describe the frequency dependence of the strain generated in the actuator, we proposed an electrochemical equivalent circuit model that considered the lumped resistance and capacitance of the electrode layer and the lumped resistance of the electrolyte layer. We have recently reported that this model can also be applied to polymer actuators containing VGCF and metal oxide.²¹

In the present study, we investigated the effect of Li salts on the electrochemical and electromechanical properties of actuators using polymer-supported VGCF/IL gel electrodes.

Experimental

Materials

VGCFs (VGCF-X, Showa Denko Co. Ltd) were used as received, and the fibers were determined to have a diameter of 10–15 nm, an average length of 3 μm and an average surface area of 270 m² g⁻¹. The Li salts used were lithium tetrafluoroborate (Li[BF₄]) and lithium bis(trifluoromethanesulfonyl)imide (Li[TFSI]), whose chemical structures are shown in Fig. 1. The ILs used were 1-ethlyl-3-methylimidazolium tetrafluoroborate (EMI[BF₄]) and 1-ethlyl-3-methylimidazolium bis(trifluoromethanesulfonyl) imide (EMI[TFSI]), whose chemical structures are also shown in Fig. 1. Li[BF₄] and Li[TFSI] were obtained from Kishida Chemical, Co. Ltd., and were used as received. EMI[BF₄] and EMI[TFSI] were obtained from Fluka and Merck, Co. Ltd., respectively, and were used as received. Other reagents were used as received from Arkema Chemicals Inc. (poly(vinylidene fluoride-co-hexafluoropropylene (PVdF(HFP)), Kynar Flex 2801), Aldrich (methyl pentanone (MP) and propylene carbonate (PC)), Kishida Chemical Co. Ltd. (Dimethylacetamide (DMAc)).

Fig. 1 Configuration of the polymer-supported VGCF/IL/Li[X] gel actuator, and the molecular structures of the ILs and polymer used.

Preparation of the actuator film⁹

The configuration of the bucky-gel actuator is illustrated in Fig. 1. Typically, the polymer-supported bucky-gel electrode layer composed of 9.3 wt% VGCF, 32.1 wt% Li[TFSI], 43.7 wt% EMI[TFSI] and 14.9 wt% PVdF(HFP) was prepared as follows. A mixture of 50 mg of VGCF, 172 mg (0.6 mmol) of Li[TFSI], 235 mg (0.6 mmol) of EMI[TFSI] and 80 mg of PVdF(HFP) in 9 ml DMAc was dispersed using a stirrer for more than 5 h to produce a gelatinous mixture. In the case of X = BF₄, the casting solution was obtained by mixing 0.6 mmol of Li[BF₄] and EMI[BF₄] with the same amount of other components in 9 ml of DMAc. The electrode layer was fabricated by casting 1.6 ml of the solution in a Teflon mold (2.5 × 2.5 cm²), allowing the solvent to evaporate, and then removing any residual solvent by heating at 80 °C in vacuo. The thickness of the obtained electrode film was 60–80 μm. The gel electrolyte layer was fabricated by casting 0.5 ml of a solution containing Li[X] (1 mmol), EMI[X] (1 mmol) and PVdF(HFP) (200 mg) in a mixed solution composed of 6 ml of MP and 500 mg of PC in the Teflon mold and performing the same solvent evaporation treatment. The thickness of the obtained gel electrolyte film was 20–30 μm. The actuator structure was fabricated by hot pressing an electrolyte and two electrode layers with the same internal IL (70 °C, 120 N, 60 s). The final thickness of the actuator was 140–175 μm, which was less than the sum of the individual layers, since the thickness of each layer decreased during hot pressing.

Displacement measurement²²

The actuator experiments were conducted using an applied triangular voltage to a 10 × 1 mm² actuator strip clipped by two gold electrodes. The displacement, at a point 5 mm away (free length) from the fixed point, was continuously monitored from one side of the actuator strip by using a laser displacement meter (Keyence, LC2100/2220). A potentio/galvanostat (Hokuto Denko, HA-501G) and a waveform generator (Yokogawa Electric, FC 200) were used with the nanocarbon polymer actuator. The electrical parameters were simultaneously measured. The measured displacement δ was transformed into the strain difference between two nanocarbon/IL electrode layers (ε) by using the following equation, on the assumption that the cross-sections are planar at any position along the actuator, i.e., there is no distortion of the cross-section:

ε = 2dδ/(L² + δ²), (1)

where L is the free length and d is the thickness of the actuator strip.²³

Characterization of the electrode and electrolyte

The conductivity of the gel electrolyte layer was measured by impedance measurement, which was measured using a Solatron 1250 Impedance Analyzer. The double-layer capacitance of the polymer-supported bucky-gel electrode (φ 7 mm) was estimated by cyclic voltammetry (CV), which was measured using a two-electrode configuration with a potentiostat (Hokuto Denko, HSV-100). The electrical conductivities of the electrodes were evaluated using the four-probe DC current method, where a linear sweep wave of current was applied from outer probe electrodes, and the voltage was measured by inner probe electrodes. Current–voltage curves were obtained using a potentio/galvanostat (Hokuto Denko, HA-151) with a waveform generator (Yokogawa Electric, FC 200). Young's moduli for the electrodes were estimated from the stress–strain curve, which was measured using a thermal stress–strain instrument (Seiko, TMA/SS 6000). The morphology of the electrode film was observed using scanning electron microscopy (SEM) with a JEOL JSM-6510.

Results and discussion

Fig. S1† shows the ionic conductivity κ (= thickness/(R × area)) plotted against the Li[X]/EMI[X] molar ratio.¹⁴ For both BF₄ and TFSI, there is a clear dependence of κ on the molar ratio. Furthermore, for Li[BF₄]/EMI[BF₄] = 0.1 and 0.5, and for Li[TFSI]/EMI[TFSI] = 0.1, κ is higher than that for gel electrolyte layers containing only EMI[BF₄] and EMI[TFSI], respectively. One possible reason for this is that the van der Waals volume for Li cations is smaller than that for [EMI] cations. However, for Li[BF₄]/EMI[BF₄] = 1.0, and for Li[TFSI]/EMI[TFSI] = 0.5 and 1.0, κ is lower than that for gel electrolyte layers containing only EMI[BF₄] and EMI[TFSI], respectively. Thus, the ionic conductivity may depend on other factors, such as the diffusion coefficient of the gel electrolyte.

Fig. 2 shows the dependence of the double-layer capacitance C (the gravimetric capacitance of the VGCF, C_VGCF = C₁/(weight of the VGCF)) on the Li[X]/EMI[X] molar ratio. For non-zero molar ratios, the capacitance for the VGCF/Li[BF₄]/EMI[BF₄] and VGCF/Li[TFSI]/EMI[TFSI] electrodes is 20.0–42.3 and about 20 F g⁻¹, respectively, at a slow sweep rate of 1 mV s⁻¹. In general, the addition of the Li salt causes an increase in the capacitance. In addition, the capacitance increases with the Li[BF₄] content but is independent of the amount of Li[TFSI]. Thus, it can be considered that the intercalation of Li⁺ and [BF₄]⁻ increases the capacitance, and that the small amount of the intercalation of [TFSI]⁻ (because of the larger anion size) is independent of the amount of Li[TFSI].

Fig. 2 Comparison of gravimetric capacitance C (F g⁻¹) for polymer-supported electrode layers with different molar ratios of Li[X]/EMI[X] (applied triangular voltage: ±0.5 V and sweep rate = 1 mV s⁻¹).

Fig. S2† shows the dependence of the electrical conductivity on the Li[X]/EMI[X] molar ratio. For non-zero molar ratios, the conductivity for the VGCF/Li[BF₄]/EMI[BF₄] and VGCF/Li[TFSI]/EMI[TFSI] electrodes is approximately 6–9 and 2.5–7.5 S cm⁻¹, respectively. Thus, there is no strong dependence on the Li[BF₄] and the electrical conductivity decreases with the Li[TFSI] content.

Fig. 3 shows the strain generated in the actuators as a function of the frequency of the applied triangular voltage (±2 V). A clear dependence on the measurement frequency can be seen. For low frequencies of 0.005–0.01 Hz, the VGCF dispersed in the electrode layer is considered to be fully charged. In contrast, at higher frequencies, there is insufficient time for the VGCF to become fully charged.⁶ The strain values for the Li[BF₄]/EMI[BF₄] actuators (in the wide frequency range of 0.005–0.01 Hz) are larger than that for the EMI[BF₄]-only actuator. For such low frequencies, the double-layer capacitance for the Li[BF₄]/EMI[BF₄] electrode is considered to be larger than that for the EMI[BF₄]-only electrode. In the frequency range of 0.005–0.1 Hz, the highest strain values are obtained for the actuator with a Li[BF₄]/EMI[BF₄] molar ratio of 1.0. As shown in Table 1, similar results are obtained for the [TFSI] case. In fact, the maximum strain value, obtained for a Li[TFSI]/EMI[TFSI] molar ratio of 1.0, is larger than that for a SWCNT-based actuator (Table 1).

Fig. 3 Strain calculated from the peak-to-peak value of the displacement for polymer-supported gel actuators with different molar ratios of Li[X]/EMI[X] as a function of the frequency of the applied triangular voltage (±2 V).

Table 1 Comparison of maximum strain (%) for polymer-supported gel actuators with different molar ratios of Li[X]/EMI[X] as a function of the frequency of the applied triangular voltage (±2 V)

Li[X]/VGCF	X = [BF4]	X = [TFSI]
0	0.205	0.136
0.1	0.210	0.147
0.5	0.287	0.234
1	0.373	0.379
SWCNT	0.43	0.28

---

Thus, a VGCF/Li[TFSI]/EMI[TFSI] actuator produced without ultrasonication exhibited superior performance to a SWCNT-based actuator, and is therefore a promising candidate for practical applications without the need for specialized SWCNTs.

Fig. S3† shows the dependence of Young's modulus on the Li[X]/EMI[X] molar ratio. For non-zero molar ratios, the values are 90–105 MPa for Li[BF₄]/EMI[BF₄] and 93–102 MPa for Li[TFSI]/EMI[TFSI], which are lower than the case with no Li salt. This is consistent with the formation of an open mesoporous network by the addition of VGCF, as in the case with SWCNTs (Fig. 4).

Fig. 4 SEM micrographs (magnification 50 000×) of the polymer-supported VGCF electrode layer (Li[BF₄]/EMI[BF₄] = 1.0).

Optimizing the performance of the actuators requires controlling the amount of Li[X], EMI[X] and VGCF, the latter of which has a very high conductivity. To compete with SWCNT-based polymer actuators, the capacitance of the bucky-gel electrode layers should be more than about 20 F g⁻¹, with an electrical conductivity of greater than about 2 S cm⁻¹. This is clearly achieved for the actuator electrode with a composition of 9.3 wt% VGCNT, 32.1 wt% Li[TFSI], 43.7 wt% EMI[TFSI] and 14.9 wt% PVdF(HFP), without the need for expensive SWCNTs.

Fig. 5 shows equivalent circuit models for the VGCF/Li[X]/IL actuators. The model in Fig. 5(a) consists of a double-layer capacitance C₁ between the electrode and the electrolyte layer, and a resistance R associated with the electrolyte layer. Fig. 5(b) shows a more simplified model, in which the double-layer capacitance C₁ is replaced by a single layer capacitance C = C₁/2. When a triangular voltage with an amplitude of ±A and a frequency of f is applied to the equivalent circuit shown in Fig. 5(b), the maximum accumulated charge Q(f) is expressed by⁶

Q(f)/Q₀ = 1 − 4CRf (1 − exp(−1/4CRf)), (2)

where Q₀ is the accumulated charge in the low-frequency limit. If the strain ε in the electrode layer is proportional to the accumulated charge, then it is given by

ε = ε₀Q(f)/Q₀, (3)

where ε₀ is the strain in the low-frequency limit.

Fig. 5 Equivalent circuit models for the bucky-gel actuator. (a) Model composed of a double-layer capacitance C₁ and ionic resistance R. (b) Model in which the double-layer capacitance is represented by C = C₁/2. (c) Model composed of a double-layer capacitance C, ionic resistance R, and electrode resistance R_el.

The electrode conductivity measured using the four-probe DC method is around 2–10 S cm⁻¹, which is much larger than that for the electrolyte. However, when considering the conduction path for the electrode layer, the electrode resistance must be accounted for in the equivalent circuit. If the electrode resistance is considered explicitly, then the equivalent circuit should be treated as a distributed transmission line circuit.²⁴ Here, we use a simple treatment for the electrode resistance that employs the resistance element R_el, as shown in Fig. 5(c), where R in eqn (2) can be replaced with R + R_el. The R_el value can be estimated based on the electrical conductivity, thickness and area of the electrode.

To evaluate the double-layer charging kinetic model, the double-layer capacitance C of the VGCF/Li[X]/IL electrode and the ionic resistance R of two different IL gel electrolyte layers were measured. The frequency dependence of the strain was calculated using eqn (2) and (3). Fig. 6 shows the frequency dependence of the measured strain, together with the simulated curve. Line A was calculated from C and R. Table S1† summarizes the measured and estimated parameters. Line B was calculated by considering the effect of the electrode resistance, and the estimated parameters are summarized in Table S2.† It is clear from Fig. 6 that the frequency dependence of the strain is reproduced using the double-layer charging kinetic model. The values of ε₀ in eqn (3) were set to the measured values in the low-frequency limit, and these are summarized in Table S1.†

Fig. 6 Measured (blue symbols) and simulated results (curve A and B) for frequency dependence of strain for VGCF/Li[X]/EMI[X]. Curve A is calculated using the equivalent circuit shown in Fig. 5(b). Curve B is calculated using the equivalent circuit shown in Fig. 5(c). (a) Li[BF₄]/EMI[BF₄] ratio of 1.0 and (b) Li[TFSI]/EMI[TFSI] ratio of 1.0.

To optimize the performance of the actuator, the results summarized in Tables S1 and S2† should be considered with respect to both the kinetic and static components. From the kinetic viewpoint, the frequency dependence of the strain is determined by electrochemical charging, as shown in Fig. 6. To obtain good fits to the experimental data, it is necessary to take the electrode resistance into consideration. Therefore, the response of the VGCF/Li[X]/IL polymer actuator can be improved by the use of an electrode with higher conductivity.

In considering the equivalent circuit of numerous microcapacitor elements, the model in Fig. 7 consists of a double-layer capacitance C_n and a resistance R_n. When a triangular voltage with an amplitude of A is applied to the equivalent circuit, the charge Q is expressed by

Q = ΣQ_n

Q_n = (A/R_n)exp(−t/C_nR_n)

Fig. 7 Configuration of an equivalent circuit of a double-layer capacitor.

The high salt concentrations (≫10⁻³ M) in the material are expected to give rigid double layers and it is not dependent on high experimental concentrations. However, in this system, C(R + R_el) is found to increase with the Li[X] content (Li[BF₄]: C is found to increase with the Li[BF₄] content, and Li[TFSI]: R_el is found to increase with the Li[TFSI] content). Therefore, we considered that sufficient time is needed for the VGCF to become fully charged.

Conclusion

The effects of Li[BF₄] and Li[TFSI] salts on the electrochemical and electromechanical properties of an actuator using a PVdF(HFP)-supported VGCF/IL gel electrode were investigated.

At a slow sweep rate of 1 mV s⁻¹, the measured double-layer capacitance values for the VGCF/Li[X]/IL actuator were in the range of 20.0–42.3 F g⁻¹, and were larger than that for an IL-only actuator. The capacitance was found to increase with the Li[BF₄] content, but was independent of the Li[TFSI] content. This is thought to be related to the intercalation between the VGCF and Li salts (Li⁺, [BF₄]⁻ and [TFSI]⁻) in both cases. All of the VGCF/Li[X]/IL actuators exhibited a larger amount of strain compared to IL-only actuators. In addition, for an actuator with a Li[TFSI]/EMI[TFSI] molar ratio of 1.0, the maximum strain was greater than that for a SWCNT-based actuator. Such actuators are therefore promising candidates for real-world applications without the need for specialized SWCNTs.

The frequency dependence of the displacement response was measured for the VGCF/Li[X]/EMI[X] actuators, and could be successfully simulated using a double-layer charging kinetic model, similar to the case for SWCNT and VGCF based actuators containing metal oxide. The results yielded the strain in the low-frequency limit in addition to the time constant of the response. The time constant was determined based on an equivalent circuit containing a series combination of an ionic resistance R, a double-layer capacitance C, and the electrode resistance R_el. The strain in the low-frequency limit can be considered to be related to the electromechanical mechanism involved in the actuators.

Acknowledgements

This work was supported in part by a KAKENHI Grant-in Aid for Scientific Research C (no. 24550264) from JSPS.

References

1. Electroactive Polymer (EAP) Actuators as Artificial Muscles, Reality, Potential and Challenges, ed. Y. Bar-Cohen, SPIE Press, Washington, DC, 2^nd edn, 2004.
2. E. Smela, Adv. Mater., 2003, 15, 481–494.
3. M. Shahinpoor, Y. Bar-Cohen, H. Simpson and J. Smith, Smart Mater. Struct., 1998, 7, R15–R30.
4. T. Fukushima, K. Asaka, A. Kosaka and T. Aida, Angew. Chem., Int. Ed., 2005, 44, 2410–2413.
5. K. Mukai, K. Asaka, K. Kiyohara, T. Sugino, I. Takeuchi, T. Fukushima and T. Aida, Electrochim. Acta, 2008, 53, 5555–5562.
6. I. Takeuchi, K. Asaka, K. Kiyohara, T. Sugino, K. Mukai, T. Fukushima and T. Aida, Electrochim. Acta, 2009, 543, 1762–1768.
7. T. Fukushima, A. Kosaka, Y. Ishimura, T. Yamamoto, T. Takigawa, N. Ishii and T. Aida, Science, 2003, 300, 2072–2074.
8. W. Lu, A. G. Fadeev, B. Qi, E. Smela, B. R. Mattes, J. Ding, G. M. Spinks, J. Mazurkiewicz, D. Zhou, G. G. Wallace, D. R. MacFarlane, S. A. Forsyth and M. Forsyth, Science, 2002, 297, 983–987.
9. N. Terasawa, I. Takeuchi and H. Matsumoto, Sens. Actuators, B, 2009, 139, 624–630.
10. N. Terasawa, I. Takeuchi, H. Matsumoto, K. Mukai and K. Asaka, Sens. Actuators, B, 2011, 156, 539–545.
11. I. Takeuchi, K. Asaka, K. Kiyohara, T. Sugino, N. Terasawa, K. Mukai and S. Shiraishi, Carbon, 2009, 47, 1373–1380.
12. N. Terasawa, N. Ono, Y. Hayakawa, K. Mukai, T. Koga, H. Higashi and K. Asaka, Sens. Actuators, B, 2011, 160, 161–167.
13. N. Terasawa, N. Ono, K. Mukai, T. Koga, H. Higashi and K. Asaka, Carbon, 2012, 50, 311–320.
14. N. Terasawa, K. Mukai, K. Yamato and K. Asaka, Carbon, 2012, 50, 1888–1896.
15. N. Terasawa, K. Mukai, K. Yamato and K. Asaka, Sens. Actuators, B, 2012, 171–172, 595–601.
16. C. Li, D. Wang, T. Liang, X. Wang and L. Ji, Mater. Lett., 2004, 58, 3774–3777.
17. S. Iijima, Nature, 1991, 354, 56–58.
18. N. Terasawa, K. Mukai and K. Asaka, J. Mater. Chem., 2012, 22, 15104–15109.
19. C. A. Angell, C. Liu and E. Sanchez, Nature, 1993, 362, 137–139.
20. N. Terasawa, I. Takeuchi, K. Mukai and K. Asaka, Polymer, 2010, 51, 3372–3376.
21. N. Terasawa and I. Takeuchi, Sens. Actuators, B, 2013, 176, 1065–1073.
22. N. Terasawa and I. Takeuchi, Sens. Actuators, B, 2010, 145, 775–780.
23. Q. Pei and O. Inganas, J. Phys. Chem., 1992, 96, 10507–10514.
24. A. Bondi, J. Phys. Chem., 1964, 68, 441–451.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c3ta1338a

---