Materials Chemistry and Physics 119 (2010) 519–523

Contents lists available at ScienceDirect

Materials Chemistry and Physics journal homepage: www.elsevier.com/locate/matchemphys

The effects of quenching treatment and AlF3 coating on LiNi0.5 Mn0.5 O2 cathode materials for lithium-ion battery Hecheng Lin a,b , Jianming Zheng a , Yong Yang a,∗ a State Key Laboratory for Physical Chemistry of Solid Surfaces, and Department of Chemistry, College of Chemistry and Chemical Engineering, Xiamen University, Xiamen 361005, PR China b Department of Biological Science and Chemical Engineering, Hunan University of Science and Engineering, Yongzhou 425006, PR China

a r t i c l e

i n f o

Article history: Received 24 August 2008 Received in revised form 20 March 2009 Accepted 10 October 2009 Keywords: Quenching methods LiNi0.5 Mn0.5 O2 High rate performance Surface coating Lithium-ion batteries

a b s t r a c t Submicron layered LiNi0.5 Mn0.5 O2 was synthesized via a co-precipitation and solid-state reaction method together with a quenching process. The crystal structure and morphology of the materials were investigated by X-ray diffraction (XRD), Brunauer–Emmett and Teller (BET) surface area and scanning electron microscopy (SEM) techniques. It is found that LiNi0.5 Mn0.5 O2 material prepared with quenching methods has smooth and regular structure in submicron scale with surface area of 0.43 m2 g−1 . The initial discharge capacities are 175.8 mAh g−1 at 0.1 C (28 mA g−1 ) and 120.3 mAh g−1 at 5.0 C (1400 mA g−1 ), respectively, for the quenched samples between 2.5 and 4.5 V. It is demonstrated that quenching method is a useful approach for the preparation of submicron layered LiNi0.5 Mn0.5 O2 cathode materials with excellent rate performance. In addition, the cycling performance of quenched-LiNi0.5 Mn0.5 O2 material was also greatly improved by AlF3 coating technique. © 2009 Elsevier B.V. All rights reserved.

1. Introduction The cathode materials are critically important to the performance and safety of lithium-ion batteries. Nowadays, LiCoO2 is still the most widely used commercial cathode material for lithiumion battery for its easy synthesis and excellent reversibility, but suffers from high cost and relative toxicity of cobalt, and structural instability at high potential limits. Recently, cobalt-less and cobalt-free materials have been investigated intensively as cathode of lithium-ion secondary batteries [1–12]. Among these materials, lithium nickel manganese oxides, LiNi0.5 Mn0.5 O2 attracts research interest due to its higher theoretical capacity (280 mAh g−1 ), lower cost, better thermal stability and more environmental friendliness. The synthesis of LiNi0.5 Mn0.5 O2 often proceeds through the formation of Li2 MnO3 and LiNiO2 by the solid-state reaction between them, thus it is easy to be present of Li2 MnO3 at reduced temperatures [13]. The LiNi0.5 Mn0.5 O2 compound is a promising material in the applications of lithium-ion batteries, but it is difficult to prepare due to either containing substantial Li/Ni disorder or existing structural impurity in LiNi0.5 Mn0.5 O2 . Recently, an ionexchange method has been reported to transform NaNi0.5 Mn0.5 O2 into LiNi0.5 Mn0.5 O2 , which possesses excellent electrochemical performance [14]. However, this is a multistep synthesis process and needs to consume a large amount of salts containing Li+ . Alter-

∗ Corresponding author. Tel.: +86 5922185753, fax: +86 5922185753. E-mail address: [email protected] (Y. Yang). 0254-0584/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.matchemphys.2009.10.007

natively, using mixed metal hydroxide, co-precipitated in solution, could acquire pure phase LiNi0.5 Mn0.5 O2 [2]. Another disadvantage of this material is lower electronic conductivity, which results in its poor rate capability and poor cycling stability. Many methods have been used to overcome the disadvantage, such as optimization the synthesis condition [10], partial substitution of transition [15], surface coating [12], and doping with Li2 TiO3 [16]. Recently, the AlF3 compound as a new coating material had been introduced to study the properties of electrode, and the effects of surface coating with AlF3 on LiCoO2 , LiNi1/3 Co1/3 Mn1/3 O2 and LiNi0.8 Mn0.1 Co0.1 O2 cathodes have demonstrated that the electrochemical performance of non-quenched electrode could be improved [17–19]. In this paper, we synthesized the layered LiNi0.5 Mn0.5 O2 by coprecipitation method and studied the quenching effects on the structure, surface morphology, rate capability and cycling performance of LiNi0.5 Mn0.5 O2 cathode material. Moreover, with respect to the unsatisfied cycling stability of the quenched samples, surface modification with AlF3 was performed to improve the cycling performance of the material. 2. Experimental LiNi0.5 Mn0.5 O2 was synthesized by the co-precipitation method described in detail as follows. A stoichiometric amount of nickel acetate and manganese acetate were dissolved in deionized water. This solution was slowly added into LiOH solution with drastically stirring. The mixture of co-precipitated double nickel manganese hydroxide, Ni0.5 Mn0.5 (OH)2 , was filtered, washed thoroughly, and dried at 120 ◦ C for 24 h. Then the double nickel manganese hydroxide was mixed with required amount

520

H. Lin et al. / Materials Chemistry and Physics 119 (2010) 519–523

of LiOH·H2 O, pressed into pellets, and heated to 1000 ◦ C for 12 h in air. Considering the evaporation loss of lithium at high temperature, 10% of excess amount of lithium was added to compensate lithium loss at high temperature in present study. The pellets were cooled naturally in furnace to room temperature (RT) to obtain nonquenched sample, which was labeled as sample A, and the sample B was gained by quenching process in air after being fired at 1000 ◦ C for 12 h. The AlF3 -coated material, coated sample, was prepared according to the following procedure [20–21]: firstly, quenched-LiNi0.5 Mn0.5 O2 powders was dispersed in the solution of aluminum nitrate (Al(NO3 )3 ·9H2 O). Then stoichiometric amount of ammonium fluoride (NH4 F) was added and stirred for 30 min. After that, the mixture was dried at 120 ◦ C for 3 h, and heat treated at 400 ◦ C for 5 h under nitrogen atmosphere to obtain sample C. The weight ratio of AlF3 /(sample B) was 3%. The AlF3 coating process may be expressed as following reactions: Al(NO3 )3 + NH4 F → AlF3 + NH4 NO3 (insolution, roomtemperature)

(1)

NH4 NO3 → NO2 + NH3 + H2 O(innitrogenatmosphere, 400 ◦ C)

(2)

The powder X-ray diffraction (XRD) characterization was carried out using Cu K␣ radiation on a Panalytical X’Pert (Philip, Netherlands) instrument to identify the crystal structure of the materials. The morphology and particle size of the samples were observed and measured by a LEO1530 Field Emission Scanning Electron Microscope (SEM, Oxford Instrument). Brunauer, Emmett and Teller (BET) surface area of powder was measured by a Tristar 3000 instrument (Micromeritic Co., USA). Electrochemical measurements of LiNi0.5 Mn0.5 O2 samples were evaluated using CR2025 coin-type cell at room temperature (RT). The working electrodes were prepared by pasting a slurry containing 80 wt% active material LiNi0.5 Mn0.5 O2 , 15 wt% acetylene black and 5 wt% binder (polyvinylidene difluoride, PVDF) in N-methyl pyrrolidone (NMP) solvent on circular Al foil, followed by being dried in an oven at 120 ◦ C for 2 h. Electrochemical cells were assembled using as-prepared working electrodes as cathode, lithium foil as anode, Cellgard 2300 as separator and 1 M LiPF6 dissolved in ethyl carbonate (EC) and dimethyl carbonate (DMC) (1:1 in volume) as electrolyte in Argon-filled glove box (Labmaster100, Mbraun, Germany). The cells were aged for 24 h before the tests. The cyclic voltammetry was directly measured by a Model 263A Potentiostat/Galvanostat (PARC, USA) using CR2025 coin-type cell. The charge/discharge experiments of the batteries were performed galvanostatically by using LAND CT 2001A battery testers (Wuhan, China).

3. Results and discussion The XRD patterns of non-quenched, quenched and coated sample powders are presented in Fig. 1. As can be seen, all of diffraction patterns of the samples can be indexed as hexagonal ␣-NaFeO2 structure with a space group R−3m. The well-resolved splitting of (0 0 6)/(1 0 2) and (1 0 8)/(1 1 0) diffraction pairs can be distinctly observed in XRD patterns of these samples, which indicate that all of the samples have a typical layered structure. The lattice parameters of LiNi0.5 Mn0.5 O2 are estimated to be a = 2.863 Å and c = 14.246 Å for the quenched sample and a = 2.888 Å and c = 14.284 Å for non-quenched quenched sample, respectively. For the non-quenched sample, the small peaks between 20◦ and 25◦ are obviously observed (Fig. 1A), which we attribute to the superlattice ordering of Li and Mn in the layers that contain transition metals. These superlattice peaks are observable in XRD patterns of Li2 MnO3 -based oxides [22]. On the contrary, no diffraction peak can be detected in this range, which indicate that Li2 MnO3 -based oxide does not exist in the quenched material (Fig. 1B). The ratio (R) of I(0 0 3)/I(1 0 4) is sensitive to the cation distribution in lattice and the degree of cation mixing of materials [23–25]. When the R value is higher, the degree of cation mixing is lower. In our experiment, the R values of non-quenched sample and quenched sample are 2.107 and 1.394, respectively. These changes are attributed to that cation disorder in the crystal lattice. When the sample was cooled naturally in furnace, the cations could reside in its normal position in the lattice, thus disorder degree of the cations is relatively lower in this case. While the sample was quenched, a small quantity of transition metal ion (mainly Ni2+ , because of the similarity between ion radius of Ni2+ (0.69 Å) and Li+ (0.76 Å)) remained in the lithium-ion layers, therefore the cation disorder degree of quenched sample should be higher, and the R value of quenched sample is lower than that of non-quenched sample. It was also reported that high temperature is a favorable condition for the Li/Ni disorder and the disorder increases as the heating temperature increases [26].

Fig. 1. XRD patterns of LiNi0.5 Mn0.5 O2 synthesized by different ways: (A) nonquenched sample; (B) quenched sample; (C) quenched sample with surface coating.

The XRD pattern of AlF3 -coated material is also shown in Fig. 1C. No extra reflection peaks corresponding to the Al-related impurity phases are seen for the AlF3 -coated sample, which is probably due to a small quantity and/or amorphous nature of AlF3 . The SEM images of the surface morphology of the nonquenched, quenched and coated samples are given in Fig. 2. The LiNi0.5 Mn0.5 O2 powders prepared by co-precipitation method are not very uniform in particle size, which distributes approximately in the size range of 0.2–0.5 ␮m. For the non-quenched sample, the agglomeration and sintering phenomenon are clearly observed from the images (Fig. 2A1 and A2). In comparison, the particles of quenched sample (Fig. 2B1 and B2) exist quite independently with very smooth surface. It manifests that the quenched material has more exposing surface than the nonquenched material, which is also supported by the BET surface area measurement (i.e. 0.1546 m2 g−1 for non-quenched material and 0.4332 m2 g−1 for quenched material, respectively). High surface area of samples could make that the electrodes of quenched material have more surface and easier contact with the electrolyte, which will lead to improve the charge/discharge capacity and rate capability. Compared with the particles of quenched sample in Fig. 2(B1 and B2), more agglomeration phenomenon and non-uniform distribution of particles’ size can be also observed in AlF3 -coated sample

H. Lin et al. / Materials Chemistry and Physics 119 (2010) 519–523

521

Fig. 2. SEM images of LiNi0.5 Mn0.5 O2 synthesized by different ways: (A) non-quenched sample; (B) quenched sample; (C) quenched sample with surface coating. Magnification of samples are: A1, B1 and C1: 2 K; A2: 10 K; B2 and C2: 50 K.

(Fig. 2C1). A large number of nano-scaled particles, which are attributed to aluminum fluoride (AlF3 ), are observed on the surface of LiNi0.5 Mn0.5 O2 particles (Fig. 2C2). In overall, the smaller particles size of LiNi0.5 Mn0.5 O2 can be obtained by quenching process, but the cation mixing degree increased. For the particle agglomeration of sample C (compared with sample B), its reason was that sample C was reheated at 400 ◦ C for 5 h. Fig. 3 presents the initial cyclic voltammograms (CVs) of nonquenched, quenched and AlF3 -coated samples. A couple of big sharp redox peaks are observed for the quenched electrode, however, the redox peaks of non-quenched sample are wider and smoother. The peak potential difference (Vp ) of redox peak of quenched sample is 0.16 V (Fig. 3B). But it is difficult to measure the Vp value from the curve due to the “overlapped” peaks

obtained for the non-quenched sample (Fig. 3A). A couple of sharp redox peaks can also be clearly seen in the cyclic voltammogram of AlF3 -coated sample, in spite of its weaker current peaks compared to that of quenched sample, the Vp value for coated sample is 0.225 V (Fig. 3C). Since the Vp value is a parameter to evaluate the electrochemical reversibility of material system, thus above Vp value imply that quenching method is able to improve the electrochemical reversibility of LiNi0.5 Mn0.5 O2 . Although the CV curve for non-quenched sample is quite broad in Fig. 3, but the integrated area under the peak is similar to that quenched sample. For the difference in the shape of oxidation/reduction peak in CV results, it is believed that the cation disorder is one of its negative effect factors, but the particle agglomeration is main factor in our work (for non-quenched sample and quenched sample), i.e. the difference in

522

H. Lin et al. / Materials Chemistry and Physics 119 (2010) 519–523

Fig. 3. The first cyclic voltammograms of LiNi0.5 Mn0.5 O2 synthesized by different ways: (A) non-quenched sample; (B) quenched sample; (C) quenched sample with surface coating. In the experiments, the potential scan rate is 0.1 mV s−1 and scanning potential range is 2.5–4.5 V.

the Li + insertion kinetics due to different diffusion length in these systems. The influences of cooling process on the electrochemical performance of LiNi0.5 Mn0.5 O2 cathode materials are obvious from charge/discharge results. Fig. 4 shows the initial discharge capacities of samples prepared by quenching and cooling slowly (no quenching) process at different current density. The theoretical discharge capacity of LiNi0.5 Mn0.5 O2 is 280 mAh g−1 . Fig. 4A shows that the initial discharge capacities of non-quenched sample are only 136 mAh g−1 at 0.1 C (28 mA g−1 , RT), and 83 mAh g−1 at 5.0 C (1400 mA g−1 , RT), respectively, in the voltage range of 2.5–4.5 V. Under the same condition, the quenched sample delivers the initial discharge capacities of 175.8 mAh g−1 at 0.1 C (28 mAh g−1 , RT), and 120.3 mAh g−1 at 5 C (1400 mAh g−1 , RT), respectively. Here, the ini-

Fig. 4. Voltage profiles–discharge capacity of the materials in the first cycle with different cooling method at different current rate (1 C = 280 mA g−1 ) in the voltage range of 2.5–4.5 V: (A) non-quenched sample; (B) quenched sample.

Fig. 5. Cycling stability curves of Li/LiNi0.5 Mn0.5 O2 cells at 28 mA g−1 between 2.5 and 4.5 V: (A) non-quenched sample; (B) quenched sample; (C) quenched sample with surface coating.

tial discharge capacities of quenched sample are much higher than that of non-quenched sample at the same current rate. Thus it can be seen that the quenching procedure has a big effect on the initial discharge capacities of material, especially at high current rates. For the improvement reason of initial capacity and rate performance of quenching sample, we think that the quenching process is advantage to the formation of fine particles (Fig. 2). The smaller particle and larger surface area of electrode material mostly contribute to the improvement of electrochemical performance, because the smaller particle means that the electrode materials have more active centers (for elevation of initial capacity) and the shorter Li+ diffusion path in inferior (for elevation of rate); and the larger surface area is advantage to the full contact between the electrode and the electrolyte. In addition, the quenching process also affects the cycling stabilities of the materials. The charge/discharge cycling stabilities of samples are shown in Fig. 5. Fig. 5(B) reveals that the discharge capacities of quenched sample decreased to 139 mAh g−1 after 30 cycles, with only 79.2% of capacity retention (vs. to initial discharge capacity). However, the capacities of non-quenched sample were from 136 to 125 mAh g−1 after 30 cycles, whose capacity retention is up to 91.4%, as shown in Fig. 5(A). Although the initial discharge capacity of quenched sample is higher than non-quenched one, its cycling stability is worse. For non-quenched material, apart from the influence of the lower cation disorder, its better cycling performance maybe resulted from a small quantity of Li2 MnO3 component (Fig. 1A) which is electrochemically inactive, but may serve to stabilize the structure of LiNi0.5 Mn0.5 O2 [27]. Apparently, the quenched material with excellent rate capability and initial capacity is promising if its poor cycling stability can be improved. Our previous studies [28,29] have showed clearly that the surface coating is useful technique to improve the cycling stability of cathode material by the formation of more stable electrode/electrolyte interfaces and suppression of oxidation of electrolytes. Therefore, we chose AlF3 as a surface coating material of quenched sample to investigate the cycling stability effect. Fig. 5 shows the discharge capacity versus cycle number of quenched sample with and without AlF3 coating. Fig. 6 shows a comparison of discharge profiles of both cells at difference cycles. Both cells had stable and smooth discharge curves (Fig. 6). No difference in the operation voltage also indicated that the AlF3 coating medium was not introduced into the LiNi0.5 Mn0.5 O2 structure. The AlF3 coating sample delivered the similar initial discharge

H. Lin et al. / Materials Chemistry and Physics 119 (2010) 519–523

523

4. Conclusions Submicron layered LiNi0.5 Mn0.5 O2 has been successfully synthesized via co-precipitation and solid-state reaction method with fast quenching process. The effects of quenching process on the structure, morphology and electrochemical performance have been investigated. It has been shown that the quenched material was able to deliver high discharge capacity and has excellent high rate charge/discharge performance, that is, it could deliver high initial discharge capacities of 175.8 mAh g−1 at 0.1 C (28 mA g−1 ), 120.3 mAh g−1 at 5.0 C(1400 mA g−1 ), respectively, in the range of 2.5–4.5 V. It has been proved that quenching process is a useful and efficient approach to improve the poor rate performance of LiNi0.5 Mn0.5 O2 material prepared by a conventional calcination method. The cycling performance of quenched-LiNi0.5 Mn0.5 O2 is further improved by AlF3 coating. The improvement mechanism was believed to be the formation of stable electrode/electrolyte interface by coating with AlF3 . Acknowledgments

Fig. 6. Discharge profiles of Li/LiNi0.5 Mn0.5 O2 cells at various cycles at 28 mAh g−1 (0.1 C) in the voltage range of 2.5–4.5 V: (A) quenched sample; (B) quenched sample with surface coating.

capacity around 175 mAh g−1 to the uncoated sample (see Fig. 5B and C and Fig. 6), namely the initial discharge capacities were 175.8 mAh g−1 for the uncoated sample and 174.2 mAh g−1 for the AlF3 -coated sample, respectively. The discharge capacity degradation of AlF3 -coated sample is also similar with that of bare quenched sample in the initial 10 cycles. The quenched electrode continued to lose its reversible capacity steadily during the subsequent cycles, which decreases dramatically to 127.2 mAh g−1 after 50th cycles (see Fig. 5(B) and Fig. 6(A)) with a capacity retention of 72.3%. However, the AlF3 -coated electrode shows very stable cycling performance during the following cycling, which can still deliver a capacity of 153.1 mAh g−1 at 50th cycles (see Fig. 5C and Fig. 6B) with 87.9% capacity retention. As a consequence, the AlF3 -coated LiNi0.5 Mn0.5 O2 becomes much more attractive for its higher discharge capacity in comparison with the non-quenched material and for its better cycling stability in comparison to the quenched material without coating. Recently, Sun et al. [19] reported that the excellent capacity retention of AlF3 -coated LiCoO2 at the cutoff voltage of 4.5 V originated from the lower Rct and reduced cobalt dissolution. Sun et al. [17] also reported that the enhanced electrochemical performance of AlF3 -coated LiNi1/3 Co1/3 Mn1/3 O2 originated from the lower and stable charge transfer resistance between cathode and electrolyte, and from the stabilized host structure of LiNi1/3 Co1/3 Mn1/3 O2 . In addition, it is suggested that the AlF3 coating layer has prevented cathode/electrolyte interfacial degradation, which is caused by a complicated process as a result of decomposition of the electrolyte at high potential and from the attack of acidic species in the electrolyte. Consequently, we also believe AlF3 coating layer can form a stable electrode/electrolyte interface for the coated material, which enhances the electrochemical performance of AlF3 -coated LiNi0.5 Mn0.5 O2 .

This work was financially supported by the National Basic Research Program of China (973 Program) (Grant No. 2007CB209702), and the National Natural Science Foundation of China Grants (No. 29925310 and Nos. 20473060, 20021002), and the construct program of the key discipline in Hunan province (no. 2006-180). References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29]

T. Ohzuku, Y. Makimura, Chem. Lett. 642 (2001) 744. Z. Lu, D.D. Macneil, J.R. Dahn, Electrochem. Solid State Lett. 4 (2001) A191. X. Yang, J. McBeen, W. Yoon, C.P. Grey, Electrochem. Commun. 4 (2002) 649. J.-S. Kim, C.S. Johnson, M.M. Thackeray, Electrochem. Commun. 4 (2002) 205. C.S. John, J.-S. Kim, A.J. Arthur, J. Kahaian, J.T. Vaughey, M.M. Thackeray, Electrochem. Commun. 4 (2002) 492. O.A. Shlyakhtin, S.-H. Choi, Y.S. Yoon, Y.-J. Oh, J. Power Sources 141 (2005) 122. K.M. Shaju, G.V. Subba, B.V.R. Chowdari, Electrochim. Acta 48 (2003) 1505. B.L. Cushing, J.B. Goodenough, Solid State Sci. 4 (2002) 1487. Y. Makimura, T. Ohzuku, J. Power Sources 121 (2003) 119–156. K.-S. Lee, S.-T. Myung, J.-S. Moon, Y.-K. Sun, Electrochim. Acta 53 (2008) 6033. Q. Wu, X.L. Yan, Z.Y. Jiang, Electrochem. Commun. 5 (2003) 878. B.-J. Hwang, S.-K. Hu, C.-H. Chen, C.-Y. Chen, H.-S. Sheu, J. Power Sources 174 (2007) 761. M. Yoshio, Y. Todorov, K. Yamato, H. Noguchi, J. Itoh, M. Okada, T. Mouri, J. Power Sources 74 (1998) 46. K. Kang, Y.S. Meng, J. Breger, C.P. Grey, G. Ceder, Science 311 (2006) 977–980. S.-H. Kang, K. Amine, J. Power Sources 119 (2003) 155. C.S. Johnson, J.S. Kim, A.J. Kropf, A.J. Kahaian, J.T. Vaughey, M.M. Thackeray, J. Power Sources 119–121 (2003) 139. Y.K. Sun, S.W. Cho, S.W. Lee, C.S. Yoon, K. Amine, J. Electrochem. Soc. 154 (3) (2007) A168. S.U. Woo, C.S. Yoon, K. Amine, I. Belharouak, Y.K. Sun, J. Electrochem. Soc. 154 (11) (2007) A1005. Y.K. Sun, J.M. Han, S.T. Myung, S.W. Lee, K. Amine, Electrochem. Commun. 8 (2006) 821. J.M. Zheng, Z.R. Zhang, X.B. Wu, Z.X. Dong, Z. Zhu, Y. Yang, J. Electrochem. Soc. 155 (10) (2008) A775. H.C. Lin, Y. Yang, Acta Chim. Sinica 67 (2) (2009) 104. Y.J. Park, Y.-S. Hong, X. Wu, M.G. Kim, K.S. Ryu, S.H. Chang, J. Electrochem. Soc. 151 (2004) A720. R.J. Gummov, M.M. Thackeray, W.I.F. David, S. Hull, Mater. Res. Bull. 27 (1992) 327. J.N. Reimer, E. Rossen, C.D. Jones, J.R. Dahn, Solid State Ionics 67 (1990) 123. B. Zhang, G. Chen, P. Xu, Z.S. Lv, Solid State Ionics 178 (2007) 1230. Y. Hinuma, Y.S. Meng, K. Kang, G. Ceder, Chem. Mater. 19 (2007) 1790. S.-H. Kang, M.M. Thackeray, J. Electrochem. Soc. 155 (4) (2008) A269. H.S. Liu, Z.R. Zhang, Z.L. Gong, Y. Yang, Solid State Ionics 166 (2004) 317. Z.R. Zhang, H.S. Liu, Z.L. Gong, Y. Yang, J. Power Sources 12 (2004) 101.