Electrochimica Acta 118 (2014) 143– 149

Contents lists available at ScienceDirect

Electrochimica Acta

jo u r n al hom ep age: www.elsev ier .com/ locate /e lec tac ta

Comparative Study of Li and Na Electrochemical Reactions with IronOxide Nanowires

Bo Huang, Kaiping Tai, Mingou Zhang, Yiran Xiao, Shen J. Dillon ∗

Department of Materials Science and Engineering, University of Illinois Urbana-Champaign, Urbana, IL, United States

a r t i c l e i n f o

Article history:Received 10 September 2013Received in revised form 2 December 2013Accepted 3 December 2013Available online 17 December 2013

Keywords:lithium batterysodium batteryiron oxide nanowireconversion reactionTEM investigation.

a b s t r a c t

This study emphasizes optimization of Fe2O3 and Fe3O4 nanowire conversion electrodes by directly-growing them on current collectors, preparing them as single crystals, and coating their surfaces withconductive carbon coatings. The systems with the least polarization during Li-ion cycling are then testedas electrodes for Na-ion chemistry. Precipitation of nanograined material during the first cycle reducesthe polarization associated with Li insertion upon subsequent cycles. After the first cycle, delithiationprimarily contributes to polarization associated with the conversion reaction with lithiation occurringclose to the equilibrium potential. The initial reduction reaction does not proceed to completion for Nachemistries. Electron microscopy reveals significant Na insertion that occurs along with the formationof defect networks. However, the results indicate that an insufficient amount is present to form criticalnuclei necessary to induce the conversion reaction.

© 2013 Elsevier Ltd. All rights reserved.

1. Introduction

Conversion reactions possess the potential to greatly enhancethe energy density of electrochemical storage technologies. Manyconversion electrodes exhibit capacities in the range of 500-1000mAhg−1 and can achieve good cycle life.[1] Metal oxide based con-version reactions with Li typically proceed as follows [2]:

MxO + 2Li+ + 2e− ↔ Li2O + xM(M = Fe, Ni, Cu, Co, etc.) (1)

High capacity high voltage conversion cathodes based on metalfluorides show particular promise in improving Li-ion energydensity.[3] Conversion reactions typically suffer from hysteresisthat limits their round trip efficiency.[1] Since conversion reactionsdo not require intercalation, ideally they should be well suited toperform well as electrodes for Na-ion reactions. Developing highcapacity Na-ion batteries is desirable because of Na’s relative abun-dance and low cost. However, limited success has been achievedin applying conversion reaction electrodes to Na-ion systems.[4]This work investigates the rate limiting processes associated withLi-ion and Na-ion cycling of model iron oxides (Fe2O3 and Fe3O4)conversion reaction electrode materials.

Iron oxides are interesting conversion reaction anodes due tothe high capacity (1007 mAh/g for Fe2O3 and 900 mAh/g for Fe3O4[5,6]), abundance, low toxicity, and low cost. Similarly, iron flu-orides could serve as ideal next-generation Li-ion cathodes [3].

∗ Corresponding author.E-mail addresses: sdillon@illinois.edu, huangboscu@163.com (S.J. Dillon).

Unfortunately, poor electronic conductivity and short characteris-tic diffusion lengths at room temperature handicaps iron oxides andlimit their commercial applicability. Therefore, various approachesto nanostructuring have been married with carbon coating strate-gies to achieve reasonable performance [7–15]. Nanostructuresgrown directly on current collector are favored to reduce contactresistance and achieve optimal charge transport[16].

FeS2, Ni3S2 and NiCo2O4 have been shown to function as con-version reaction electrodes for Na-ion batteries.[4,17–20] Thesereactions result in nanocrystalline metal surrounded by Na2O orNa2S. It has been suggested by some that iron oxides do not func-tion as Na-ion conversion electrodes.[4] Na insertion was observedin Fe3O4 and Fe2O3 nanoparticles ranging from 10-400 nm withincreased capacity resulting from particle size reduction [21].However, no conversion reaction was observed. Recent workdemonstrates Na conversion reaction with Fe3O4 in the particlesize range from 4 to 10 nm [22]. However, much of the capacityachieved was in the range where Na intercalates into carbon, whichwas present in large amounts. The primary evidence for the reac-tion is the observation of Na2O in electron diffraction and a singleFe ring that overlaps with Fe3O4. The Na2O could possibly formupon exposure of the sample to air while loading it into the trans-mission electron microscope. However, the overall capacity doesexceed what would be expected from the carbon alone. Hollow !-Fe2O3 nanoparticles with abundant cation vacancies have recentlybeen reported to be promising electrode materials for Na-ion bat-teries in terms of the both capacity and cycle life [23]. The variousresults motivate further investigation into Na conversion reactionswith Fe-based oxides.

0013-4686/$ – see front matter © 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.electacta.2013.12.007

144 B. Huang et al. / Electrochimica Acta 118 (2014) 143– 149

Fig. 1. SEM micrographs, TEM micrographs, and selected area electron diffraction (SAED) patterns of as-prepared nanowires: "-Fe2O3 (a-c), "-Fe2O3-C (d-f), Fe3O4-C (g-i)and Fe3O4 (j-l).

In this work, we grow "-Fe2O3 and Fe3O4 single crystalnanowires with and without carbon coating directly on currentcollectors. The systems are optimized for Li-ion cycling and sub-sequently characterized as Na-ion hosts.

2. Experimental

2.1. Preparation of ˛-Fe2O3 single crystal nanowires withoutcarbon coating

220 #m diameter (99.99%, Goodfellow) 10 cm long iron wirewas cleaned in dilute hydrochloric acid (2% in volume) and rinsedwith acetone, alcohol, and deionized water. AC power (∼4 W at60 Hz) was then applied for ∼10 mins to promote surface oxida-tion during Joule heating [24]. Nanowire growth proceeds by rapidcation diffusion that emerges from grain boundaries of a thickerunderlying oxide.

2.2. Preparation of ˛-Fe2O3 single crystal nanowires with carboncoating

The resulting "-Fe2O3 single crystal nanowires were heatedto 500 ◦C for 5 h in a tube furnace containing Ar that had flowedthrough toluene. Decomposition of the toluene produces a carboncoating on the nanowires.

2.3. Preparation of Fe3O4 single crystal nanowires with carboncoating [25]

220 #m diameter iron wire (99.99%, Goodfellow) was firstcleaned as described above. It was the pre-oxidized at 250 ◦C for0.5 h on the hotplate in ambient conditions. The samples were sub-sequently annealed at 550 ◦C in a crucible also containing pure Cu.Annealing was performed in Ar that flowed through toluene priorto entering the furnace. During annealing Cu deposits on the Fesubstrate and catalyzes the growth of Fe3O4 nanowires. This pro-cess does not result in a thick underlying oxide typical of the Fe2O3growth process.

2.4. Preparation of Fe3O4 single crystal nanowires withoutcarbon coating

Reactive ion etching in 5% oxygen and 95% argon for 40 sec wasused to remove the carbon coating from the Fe3O4 nanowires pre-pared as described above.

2.5. Electrochemical testing

The nanowires were tested in a vial cell within a dry Ar-filledglovebox (Mbraun Labstar). The iron wire substrate served as theelectrode current collector. This was cycled against either a metal-lic lithium or sodium counter electrode in ethylene carbonate (EC)

B. Huang et al. / Electrochimica Acta 118 (2014) 143– 149 145

Fig. 2. Cyclic voltammagrams of nanowires cycled against Li metal at scan rates of 500 #V/s: "-Fe2O3 (a), "-Fe2O3-C (b), Fe3O4 (c) and Fe3O4-C (d).

dimethyl carbonate (DMC) (1:1 by volume) 1 M LiClO4 or NaClO4electrolyte, respectively. The electrochemical tests were carried outusing a computer-controlled potentiostat/galvanostat (SP200, Bio-logic Co.). Cyclic voltammetry (CV) was performed at a scan rateof 500 #V/s. In order to promote a more complete reaction dur-ing Na-ion cycling, the samples were also maintained at a constantpotential of 0.01 V (vs. Na/Na+) for 10 h, after the potential sweeps.The rate performance of Fe3O4 nanowires was tested at various cur-rent densities (0.2 mAh/cm2-1.6 mAh/cm2). After testing, all of thesamples were washed by propylene carbonate (PC) and acetone,and then dried in the glove box.

2.6. Charaterization

The pristine and reacted nanowires were characterized byscanning electron microscopy, SEM (JEOL-6060LV), transmissionelectron microscopy, TEM (JEOL-2010Lab6 and JEOL-2010Cryo),and Energy-dispersive X-ray spectroscopy (EDS) in the scanningtransmission electron microscopy (JEOL-2010F EF-FEG).

3. Results and discussion

3.1. Cycling Iron Oxide Nanowires Against Li

Fig. 1 shows SEM and TEM micrographs of the pristine "-Fe2O3, "-Fe2O3-C, Fe3O4-C, and Fe3O4 single crystal nanowires. Ahigh density of relatively long nanowires (2∼10 #m for Fe2O3 and5∼20 #m for Fe3O4) resulted from the different growth processes.The Fe3O4-C nanowires were present in the highest density andwere the longest nanowires. The "-Fe2O3, "-Fe2O3-C nanowiresare needle shaped, while the Fe3O4 nanowires are relatively uni-form in diameter. Fig. 1e indicates a ∼3 nm carbon layer uniformlysurrounds the coated nanowires. The copper nanoparticles ter-minating the carbon-coated Fe3O4 and Fe3O4 nanowires (Fig. 1h

and k) function as catalysts during the synthesis at low oxygenpartial pressure [25]. According to each of the selected area elec-tron diffraction (SAED) patterns (Fig. 1c, f, i and l), the preparednanowires are single crystalline.

Fig. 2 shows cyclic voltammetry performed on each sampleat a scan rate of 500 #V/s between 0.25 and 2.5 V vs. Li/Li+. Thefirst cathodic peak in each sample occurs at 0.75 V. The initialcathodic peaks shift to 0.85-0.9 V in the second cycle. This value isassociated with the reversible electrochemical potential for Fe2O3and Fe3O4 [26,27]. During subsequent cycles some increase incathodic polarization is observed. The carbon coating provideslimited improvement in the polarization associated with Fe2O3, buthas no observable impact on the cathodic polarization in Fe3O4.

After the first cycle, the single crystal nanowires are convertedto polycrystalline nanowires [28]. The excess cathodic polarizationobserved in the first cycle likely results from the conversion of asingle crystal nanowire to a polycrystalline nanowire.

Fig. 3. Rate performance, columbic efficiency, and cycle life of Fe3O4-C nanowirearray cycle against Li at various current densities.

146 B. Huang et al. / Electrochimica Acta 118 (2014) 143– 149

Fig. 4. Cyclic voltammagrams of "-Fe2O3-C (a) and Fe3O4-C (b) nanowires cycled against Na metal at scan rates of 500 #V/s.

All of the anodic peaks are much broader in width than thecathodic peaks, indicating sluggish oxidation during delithiation.The peak separation (!Ep = Ecath - Eanod) is smaller than values 1 and1.6 V reported elsewhere [15,29]. This may result from two factors;high surface area nanostructures can limit diffusional path lengthsand the reduced contact resistance, associated with directly grow-ing the electrodes on the current collectors [16], optimizing chargetransport.

It is well known that carbonous coatings often enhance elec-tron transport, suppress polarization, and extend cycle life. Relativeto the pristine "-Fe2O3, carbon-coated "-Fe2O3 nanowires exhibitlower hysteresis and improved stability during multiple cycles.Uncoated and carbon-coated Fe3O4 perform similarly to oneanother and the "-Fe2O3-C. The larger current density in the Fe3O4results from the larger nanowire density. The intrinsic electronicconductivity of Fe3O4 ($ ∼ 10−2 S/m) significantly exceeds that of

Fig. 5. Bright field (BF) images, dark field (DF) images, and SAED patterns of nanowires after the first half of cyclic voltammetry: Li-reacted "-Fe2O3-C (a-c) and Fe3O4-C (d-f),and Na-reacted "-Fe2O3-C (g-i) and Fe3O4-C (j-l).

B. Huang et al. / Electrochimica Acta 118 (2014) 143– 149 147

Fig. 6. BF image of pristine Fe3O4-C nanowire (a), BF image (b), SAED pattern (c) and HRTEM image (d) of the Na-reacted Fe3O4-C nanowire discharged at 0.01 V (vs. Na/Na+)for 10 h.

"-Fe2O3 ($ ∼ 10−8 S/m) at room temperature [30,31]. This differ-ence accounts for the difference in polarization between the twomaterials. It also accounts for the lack of improvement of Fe3O4 bycarbon coating.

Overall the CV results indicate that Fe3O4-C nanowires exhibitthe best performance in terms of hysteresis and cycling stability.The capacity retention, rate capability, and columbic efficiency of

these electrodes under galvanostatic cycling against Li was char-acterized as shown in Fig. 3. The columbic efficiency increasesfrom 60% to 96% in the first 5 cycles. The possible sources ofirreversible capacity include: solid electrolyte interface (SEI) for-mation on the large surface area of nanowire array, which mayalso be observed in Fig. 5d, lithium insertion into irreversible sites,or side reactions with absorbed and impurity species.[16] Taberna

Fig. 7. BF image (a) and EDS maps (b-d) of Na-reacted Fe3O4-C nanowire discharged at 0.01 V (vs. Na/Na+) for 10 h.

148 B. Huang et al. / Electrochimica Acta 118 (2014) 143– 149

et al. electrochemically deposited Fe3O4 nano-particles on coppernanowire arrays, achieving a large areal capacity of 0.35 mAh/cm2

[5]. Our direct nanowire growth improves the areal energy den-sity to 1.1 mAh/cm2, with similar electrochemical performance. Wenote that at the lowest rate (∼C/5) that these nanowires approxi-mately achieve the theoretical capacity.

3.2. Cycling Iron Oxide Nanowires Against Na

The conversion electrode systems optimized for lithium ionchemistry described above is extended to sodium ion systems. Fig. 4shows CV for "-Fe2O3-C and Fe3O4-C nanowires electrodes cycledagainst sodium in the range of 0.2-2.2 V (vs. Na/Na+), which wasutilized in order to keep absolute potentials consistent with the Liexperiments. The cathodic current density for Na insertion duringthe first cycle is one order of magnitude smaller than for Li insertion.No appreciable anodic current is measured, and the cathodic cur-rent degrades upon subsequent cycling. Limited Na is inserted intothe nanowires and almost none is extracted anodically. The decayin the cathodic peak suggests that the electrode progressively sat-urates with Na during the CV cycles. Testing the Fe substrate aloneresulted in a current response an order of magnitude lower thanwhen the nanowires are present. This is consistent with the factthat there is negligible solubility and no intermediate phases in theNa-Fe system. We also tested the samples in propylene carbon-ate 1 M NaClO4 to compare with ref. [23], and observed a similarresponse to that in Fig. 4.

All the nanowires were investigated by transmission electronmicroscopy (TEM) after the first half cycle of potential sweep asshown in Fig. 5. The morphology, phase, and structure are char-acterized by bright field (BF) imaging, dark field (DF) imaging andselected area electron diffraction (SAED), respectively. Prior inves-tigation of the mechanism for the conversion reaction in "-Fe2O3nanowires [28] indicates that reduction of Fe2O3 is completed andconvert to BCC Fe occurs by 0.6 V. The results in Fig. 5 indicate that"-Fe2O3 and Fe3O4 completely reduce to BCC Fe. The size of the Fenanocrystals ranges from 5 to 25 nm. No appreciable SEI is observedon the nanowires cycled against Na. These nanowires remain sin-gle crystalline, but display a significant amount of BF diffractioncontrast relative to the initial nanowires.

In attempt to complete Na insertion into the Fe3O4-C nanowires,they were cycled to 0.01 V vs. Na/Na+ and held for 10 h. The contrastof BF image demonstrates the occurrence of a large concentration ofdislocations within the nanowire (Fig. 6b), which are absent in thepristine Fe3O4-C (Fig. 6a). The SAED pattern (Fig. 6c) demonstratesthat the Fe3O4 remains single crystalline. No reduced Fe appears inthe SAED or high-resolution images (Fig. 6d). Additional evidenceis that the 0.30 nm d-space of [220] measured from HRTEM images,matches the standard value of d-space for Fe3O4 (XRD-PDF 01-078-6086).

Fig. 7 shows EDS maps of Na, Fe, and O in the Fe3O4 nanowires,discharged to 0.01 V vs. Na/Na+ for 10 h. The results suggest thata small amount of Na intercalates into the nanowire. The uniformintensity should not result from SEI, which would cause the Na tobe concentrated at the surface. Additionally, SEI is not observedat the surface by high-resolution imaging (Fig. 6d). Overall, theresults indicate that Na diffuses into the nanowires and inducessignificant dislocation production, but does not initiate the conver-sion reaction. This Na is subsequently trapped in the nanowire andcannot diffuse out during the cathodic sweep. This differs signifi-cantly from reactions in the Li system, where the anodic reactionoccurs near the theoretical potential and the cathodic reaction ischemically reversible with some degree of hysteresis. The signif-icant discrepancies likely arise from the differences in ionic radiibetween Li (76 pm) and Na (102 pm) [1]. A larger hysteresis wasobserved for Li insertion during the first cycle, which is attributed

to the conversion of the single crystalline nanowires to polycrys-talline nanowires that have grain boundary pathways for enhancedtransport and heterogeneous nucleation. It should be noted thatwe also first cycled nanowire samples against Li to induce theconversion reaction that results in polycrystalline nanowires, andthen attempted to cycle these nanowires versus Na. However, theapproach did not appreciably affect the Na insertion reaction.

Damage accumulation in Fe2O3 and Fe3O4 nanowires due tostrain induced by Na insertion is consistent with results for Li inser-tion in SnO2 where the large strain induces dislocation productionand amorphization [32]. The effect is attributed to the large in-planemisfit associated with Li insertion. Na insertion into Fe2O3 or Fe3O4likely introduces large shear stresses that both induce damage andultimately suppress the conversion reaction by not allowing theNa to reach a critical concentration necessary to nucleate Na2O andreduced Fe.

Hariharan et al. recently investigated Na insertion intonanocrystalline Fe3O4, in the size range of 4-10 nm. Anodic currentswere attributed to Na insertion following the conversion reaction[22]:

Fe3O4 + 8e + 8Na+ ↔ 3Fe0 + 4Na2O (2)

Their results suggest that all Fe2+/3+ can be reduced to Fe0 dur-ing the discharge. SAED in this study indicates the formation ofNa2O with only a weak ring associated with Fe, which overlapswith Fe3O4. The capacity achieved is half of that associated with Li,indicating that the Fe2+/3+ should only be partially reduced. Addi-tionally, some of the capacity is likely associated with intercalationinto carbon. However, carbon cannot account for all of the observedcapacity indicating that the conversion reaction must proceed tosome degree. This fine (4-10 nm) powder must be small enoughthat the volumetric strain associated with the initial Na intercala-tion does not hinder the accumulation of sufficient Na to overcomethe barrier associated with nucleating Fe and Na2O. The nanowiresin this study are not sufficiently small to allow the reaction toproceed to this step. Extraction of Li from Fe2O3 and Fe3O4 hasa larger associated polarization than insertion. The same problemmay also plague the Na conversion reaction, even in the case of finenanoparticles. Such irreversibility may account for the relativelylow columbic efficiency and rapid capacity fade associated withthe Na conversion reaction in Fe3O4 [22].

4. Conclusions

The electrochemical response of "-Fe2O3, "-Fe2O3-C, Fe3O4,and Fe3O4-C single crystal nanowire conversion anodes was char-acterized during cycling against Li and Na. "-Fe2O3-C and Fe3O4-Cnanowires exhibited the best performance in terms of capacity,reversibility, and cyclability versus Li. While Na could be partiallyinserted into the nanowires, the conversion reaction could not beinduced and the reaction was not chemically reversible. The accu-mulation of damage during insertion suggests that the reaction isstrain limited. The work supports earlier results that indicate Nainsertion is possible and that at smaller particle sizes it may bepossible to induce the conversion reaction.

Acknowledgements

The authors are grateful for funding provided by the U.S.Department of Energy, Basic Energy Sciences (Contract No. DE-SC0006509). The research was carried out in the Frederick SeitzMaterials Research Laboratory Central Facilities, University of Illi-nois.

B. Huang et al. / Electrochimica Acta 118 (2014) 143– 149 149

References

[1] S.P. Ong, V.L. Chevrier, G. Hautier, A. Jain, C. Moore, S. Kim, X. Ma, G. Ceder,Energy and Environmental Science 4 (2011) 3680–3688.

[2] P. Poizot, S. Laruelle, S. Grugeon, L. Dupont, J.M. Tarascon, Nature 407 (2000)496–499.

[3] L. Li, F. Meng, S. Jin, Nano Letters 12 (2012) 6030–6037.[4] S.W. Kim, D.H. Seo, X. Ma, G. Ceder, K. Kang, Advanced Energy Materials 2 (2012)

710–721.[5] P.L. Taberna, S. Mitra, P. Poizot, P. Simon, J.M. Tarascon, Nature Materials 5

(2006) 567–573.[6] D. Larcher, C. Masquelier, D. Bonnin, Y. Chabre, V. Masson, J.B. Leriche, J.M.

Tarascon, Journal of the Electrochemical Society 150 (2003) A133–A139.[7] H. Liu, G. Wang, J. Park, J. Wang, C. Zhang, Electrochimica Acta 54 (2009)

1733–1736.[8] X.L. Wu, Y.G. Guo, L.J. Wan, C.W. Hu, Journal of Physical Chemistry C 112 (2008)

16824–16829.[9] Y. NuLi, R. Zeng, P. Zhang, Z. Guo, H. Liu, Journal of Power Sources 184 (2008)

456–461.[10] J. Li, H.M. Dahn, L.J. Krause, D.B. Le, J.R. Dahn, Journal of the Electrochemical

Society 155 (2008) A812–A816.[11] S. Zeng, K. Tang, T. Li, Z. Liang, D. Wang, Y. Wang, W. Zhou, Journal of Physical

Chemistry C 111 (2007) 10217–10225.[12] P.C. Wang, H.P. Ding, T. Bark, C.H. Chen, Electrochimica Acta 52 (2007)

6650–6655.[13] F. Jiao, J. Bao, P.G. Bruce, Electrochemical and Solid-State Letters 10 (2007)

A264–A266.[14] J. Morales, L. Sánchez, F. Martín, F. Berry, X. Ren, Journal of the Electrochemical

Society 152 (2005) A1748–A1754.[15] J. Chen, L. Xu, W. Li, X. Gou, Advanced Materials 17 (2005) 582–586.

[16] H. Han, T. Song, E.K. Lee, A. Devadoss, Y. Jeon, J. Ha, Y.C. Chung, Y.M. Choi, Y.G.Jung, U. Paik, ACS Nano 6 (2012) 8308–8315.

[17] H. Kim, D.H. Seo, S.W. Kim, J. Kim, K. Kang, Carbon 49 (2011) 326–332.[18] J.S. Kim, G.B. Cho, K.W. Kim, J.H. Ahn, G. Wang, H.J. Ahn, Current Applied Physics

11 (2011) S215–S218.[19] T.B. Kim, J.W. Choi, H.S. Ryu, G.B. Cho, K.W. Kim, J.H. Ahn, K.K. Cho, H.J. Ahn,

Journal of Power Sources 174 (2007) 1275–1278.[20] R. Alcántara, M. Jaraba, P. Lavela, J.L. Tirado, Chemistry of Materials 14 (2002)

2847–2848.[21] S. Komaba, T. Mikumo, N. Yabuuchi, A. Ogata, H. Yoshida, Y. Yamada, Journal of

the Electrochemical Society 157 (2010) A60–A65.[22] S. Hariharan, K. Saravanan, V. Ramar, P. Balaya, Physical Chemistry Chemical

Physics 15 (2013) 2945–2953.[23] B. Koo, S. Chattopadhyay, T. Shibata, V.B. Prakapenka, C.S. Johnson, T. Rajh, E.V.

Shevchenko, Chemistry of Materials 25 (2013) 245–252.[24] A.G. Nasibulin, S. Rackauskas, H. Jiang, Y. Tian, P.R. Mudimela, S.D. Shandakov,

L.I. Nasibulina, S. Jani, E.I. Kauppinen, Nano Research 2 (2009) 373–379.[25] K. Tai, B., Huang, S.J. Dillon.[26] B. Mauvernay, M.L. Doublet, L. Monconduit, Journal of Physics and Chemistry

of Solids 67 (2006) 1252–1257.[27] H. Kitaura, K. Takahashi, F. Mizuno, A. Hayashi, K. Tadanaga, M. Tatsumisago,

Journal of the Electrochemical Society 154 (2007) A725–A729.[28] B. Huang, K. Tai, S.J. Dillon, Journal of Power Sources 245 (2014) 308–314.[29] M.V. Reddy, T. Yu, C.H. Sow, Z.X. Shen, C.T. Lim, G.V.S. Rao, B.V.R. Chowdari,

Advanced Functional Materials 17 (2007) 2792–2799.[30] E.J.W. Verwey, P.W. Haayman, Physica 8 (1941) 979–987.[31] S. Ito, Y. Yui, J. Mizuguchi, Materials Transactions 51 (2010) 1163–1167.[32] J.Y. Huang, L. Zhong, C.M. Wang, J.P. Sullivan, W. Xu, L.Q. Zhang, S.X. Mao, N.S.

Hudak, X.H. Liu, A. Subramanian, H. Fan, L. Qi, A. Kushima, J. Li, Science 330(2010) 1515–1520.